From 70a16ff8a162ad0b6a39d17a1699a2949e2a2674 Mon Sep 17 00:00:00 2001 From: Patrick Steinhardt Date: Fri, 7 Feb 2025 12:03:26 +0100 Subject: [PATCH 001/312] path: refactor `repo_common_path()` family of functions The functions provided by the "path" subsystem to derive repository paths for the commondir, gitdir, worktrees and submodules are quite inconsistent. Some functions have a `strbuf_` prefix, others have different return values, some don't provide a variant working on top of `strbuf`s. We're thus about to refactor all of these family of functions so that they follow a common pattern: - `repo_*_path()` returns an allocated string. - `repo_*_path_append()` appends the path to the caller-provided buffer while returning a constant pointer to the buffer. This clarifies whether the buffer is being appended to or rewritten, which otherwise wasn't immediately obvious. - `repo_*_path_replace()` replaces contents of the buffer with the computed path, again returning a pointer to the buffer contents. The returned constant pointer isn't being used anywhere yet, but it will be used in subsequent commits. Its intent is to allow calling patterns like the following somewhat contrived example: if (!stat(&st, repo_common_path_replace(repo, &buf, ...)) && !unlink(repo_common_path_replace(repo, &buf, ...))) ... Refactor the commondir family of functions accordingly and adapt all callers. Note that `repo_common_pathv()` is converted into an internal implementation detail. It is only used to implement `the_repository` compatibility shims and will eventually be removed from the public interface. Signed-off-by: Patrick Steinhardt Signed-off-by: Junio C Hamano --- loose.c | 6 +++--- path.c | 32 ++++++++++++++++++++++++++++---- path.h | 30 +++++++++++++++++------------- refs.c | 4 ++-- setup.c | 4 ++-- worktree.c | 5 ++--- 6 files changed, 54 insertions(+), 27 deletions(-) diff --git a/loose.c b/loose.c index 897ba389daa18c..51ef490f935adc 100644 --- a/loose.c +++ b/loose.c @@ -75,7 +75,7 @@ static int load_one_loose_object_map(struct repository *repo, struct object_dire insert_loose_map(dir, repo->hash_algo->empty_blob, repo->compat_hash_algo->empty_blob); insert_loose_map(dir, repo->hash_algo->null_oid, repo->compat_hash_algo->null_oid); - strbuf_git_common_path(&path, repo, "objects/loose-object-idx"); + repo_common_path_replace(repo, &path, "objects/loose-object-idx"); fp = fopen(path.buf, "rb"); if (!fp) { strbuf_release(&path); @@ -133,7 +133,7 @@ int repo_write_loose_object_map(struct repository *repo) if (!should_use_loose_object_map(repo)) return 0; - strbuf_git_common_path(&path, repo, "objects/loose-object-idx"); + repo_common_path_replace(repo, &path, "objects/loose-object-idx"); fd = hold_lock_file_for_update_timeout(&lock, path.buf, LOCK_DIE_ON_ERROR, -1); iter = kh_begin(map); if (write_in_full(fd, loose_object_header, strlen(loose_object_header)) < 0) @@ -174,7 +174,7 @@ static int write_one_object(struct repository *repo, const struct object_id *oid struct stat st; struct strbuf buf = STRBUF_INIT, path = STRBUF_INIT; - strbuf_git_common_path(&path, repo, "objects/loose-object-idx"); + repo_common_path_replace(repo, &path, "objects/loose-object-idx"); hold_lock_file_for_update_timeout(&lock, path.buf, LOCK_DIE_ON_ERROR, -1); fd = open(path.buf, O_WRONLY | O_CREAT | O_APPEND, 0666); diff --git a/path.c b/path.c index 07964f5d32c977..273b649e00c2d7 100644 --- a/path.c +++ b/path.c @@ -414,7 +414,7 @@ static void strbuf_worktree_gitdir(struct strbuf *buf, else if (!wt->id) strbuf_addstr(buf, repo->commondir); else - strbuf_git_common_path(buf, repo, "worktrees/%s", wt->id); + repo_common_path_append(repo, buf, "worktrees/%s", wt->id); } void repo_git_pathv(const struct repository *repo, @@ -596,14 +596,38 @@ void repo_common_pathv(const struct repository *repo, strbuf_cleanup_path(sb); } -void strbuf_git_common_path(struct strbuf *sb, - const struct repository *repo, - const char *fmt, ...) +char *repo_common_path(const struct repository *repo, + const char *fmt, ...) +{ + struct strbuf sb = STRBUF_INIT; + va_list args; + va_start(args, fmt); + repo_common_pathv(repo, &sb, fmt, args); + va_end(args); + return strbuf_detach(&sb, NULL); +} + +const char *repo_common_path_append(const struct repository *repo, + struct strbuf *sb, + const char *fmt, ...) +{ + va_list args; + va_start(args, fmt); + repo_common_pathv(repo, sb, fmt, args); + va_end(args); + return sb->buf; +} + +const char *repo_common_path_replace(const struct repository *repo, + struct strbuf *sb, + const char *fmt, ...) { va_list args; + strbuf_reset(sb); va_start(args, fmt); repo_common_pathv(repo, sb, fmt, args); va_end(args); + return sb->buf; } static struct passwd *getpw_str(const char *username, size_t len) diff --git a/path.h b/path.h index 5f6c85e5f8d795..3c75495e1a55e7 100644 --- a/path.h +++ b/path.h @@ -25,22 +25,20 @@ char *mkpathdup(const char *fmt, ...) __attribute__((format (printf, 1, 2))); /* - * The `strbuf_git_common_path` family of functions will construct a path into a + * The `repo_common_path` family of functions will construct a path into a * repository's common git directory, which is shared by all worktrees. */ - -/* - * Constructs a path into the common git directory of repository `repo` and - * append it in the provided buffer `sb`. - */ -void strbuf_git_common_path(struct strbuf *sb, - const struct repository *repo, - const char *fmt, ...) +char *repo_common_path(const struct repository *repo, + const char *fmt, ...) + __attribute__((format (printf, 2, 3))); +const char *repo_common_path_append(const struct repository *repo, + struct strbuf *sb, + const char *fmt, ...) + __attribute__((format (printf, 3, 4))); +const char *repo_common_path_replace(const struct repository *repo, + struct strbuf *sb, + const char *fmt, ...) __attribute__((format (printf, 3, 4))); -void repo_common_pathv(const struct repository *repo, - struct strbuf *buf, - const char *fmt, - va_list args); /* * The `repo_git_path` family of functions will construct a path into a repository's @@ -243,6 +241,12 @@ struct strbuf *get_pathname(void); # include "strbuf.h" # include "repository.h" +/* Internal implementation detail that should not be used. */ +void repo_common_pathv(const struct repository *repo, + struct strbuf *buf, + const char *fmt, + va_list args); + /* * Return a statically allocated path into the main repository's * (the_repository) common git directory. diff --git a/refs.c b/refs.c index f4094a326a9f88..daf6a842050a16 100644 --- a/refs.c +++ b/refs.c @@ -2184,8 +2184,8 @@ struct ref_store *get_worktree_ref_store(const struct worktree *wt) if (wt->id) { struct strbuf common_path = STRBUF_INIT; - strbuf_git_common_path(&common_path, wt->repo, - "worktrees/%s", wt->id); + repo_common_path_append(wt->repo, &common_path, + "worktrees/%s", wt->id); refs = ref_store_init(wt->repo, wt->repo->ref_storage_format, common_path.buf, REF_STORE_ALL_CAPS); strbuf_release(&common_path); diff --git a/setup.c b/setup.c index 8a488f3e7c74b1..74b5ba532555e1 100644 --- a/setup.c +++ b/setup.c @@ -792,7 +792,7 @@ int upgrade_repository_format(int target_version) struct repository_format repo_fmt = REPOSITORY_FORMAT_INIT; int ret; - strbuf_git_common_path(&sb, the_repository, "config"); + repo_common_path_append(the_repository, &sb, "config"); read_repository_format(&repo_fmt, sb.buf); strbuf_release(&sb); @@ -2242,7 +2242,7 @@ void initialize_repository_version(int hash_algo, struct strbuf config = STRBUF_INIT; struct repository_format repo_fmt = REPOSITORY_FORMAT_INIT; - strbuf_git_common_path(&config, the_repository, "config"); + repo_common_path_append(the_repository, &config, "config"); read_repository_format(&repo_fmt, config.buf); if (repo_fmt.v1_only_extensions.nr) diff --git a/worktree.c b/worktree.c index 248bbb39d43ae8..f8d6e7127f4ab8 100644 --- a/worktree.c +++ b/worktree.c @@ -104,7 +104,7 @@ struct worktree *get_linked_worktree(const char *id, if (!id) die("Missing linked worktree name"); - strbuf_git_common_path(&path, the_repository, "worktrees/%s/gitdir", id); + repo_common_path_append(the_repository, &path, "worktrees/%s/gitdir", id); if (strbuf_read_file(&worktree_path, path.buf, 0) <= 0) /* invalid gitdir file */ goto done; @@ -731,8 +731,7 @@ static ssize_t infer_backlink(const char *gitfile, struct strbuf *inferred) id++; /* advance past '/' to point at */ if (!*id) goto error; - strbuf_reset(inferred); - strbuf_git_common_path(inferred, the_repository, "worktrees/%s", id); + repo_common_path_replace(the_repository, inferred, "worktrees/%s", id); if (!is_directory(inferred->buf)) goto error; From bdfc07bfdf3f4f4ef94580c0cb46eef5977bb810 Mon Sep 17 00:00:00 2001 From: Patrick Steinhardt Date: Fri, 7 Feb 2025 12:03:27 +0100 Subject: [PATCH 002/312] path: refactor `repo_git_path()` family of functions As explained in an earlier commit, we're refactoring path-related functions to provide a consistent interface for computing paths into the commondir, gitdir and worktree. Refactor the "gitdir" family of functions accordingly. Note that the `repo_git_pathv()` function is converted into an internal implementation detail. It is only used to implement `the_repository` compatibility shims and will eventually be removed from the public interface. Signed-off-by: Patrick Steinhardt Signed-off-by: Junio C Hamano --- editor.c | 6 ++---- hook.c | 3 +-- path.c | 19 ++++++++++++++++--- path.h | 32 +++++++++++--------------------- submodule.c | 4 ++-- 5 files changed, 32 insertions(+), 32 deletions(-) diff --git a/editor.c b/editor.c index 6b9ce81d5fc0fd..b79d97b0e721c2 100644 --- a/editor.c +++ b/editor.c @@ -142,10 +142,8 @@ int strbuf_edit_interactively(struct repository *r, struct strbuf sb = STRBUF_INIT; int fd, res = 0; - if (!is_absolute_path(path)) { - strbuf_repo_git_path(&sb, r, "%s", path); - path = sb.buf; - } + if (!is_absolute_path(path)) + path = repo_git_path_append(r, &sb, "%s", path); fd = open(path, O_WRONLY | O_CREAT | O_TRUNC, 0666); if (fd < 0) diff --git a/hook.c b/hook.c index 9ddbdee06d5774..b3de1048bf44b9 100644 --- a/hook.c +++ b/hook.c @@ -16,8 +16,7 @@ const char *find_hook(struct repository *r, const char *name) int found_hook; - strbuf_reset(&path); - strbuf_repo_git_path(&path, r, "hooks/%s", name); + repo_git_path_replace(r, &path, "hooks/%s", name); found_hook = access(path.buf, X_OK) >= 0; #ifdef STRIP_EXTENSION if (!found_hook) { diff --git a/path.c b/path.c index 273b649e00c2d7..779aa94b567a9d 100644 --- a/path.c +++ b/path.c @@ -443,14 +443,27 @@ char *repo_git_path(const struct repository *repo, return strbuf_detach(&path, NULL); } -void strbuf_repo_git_path(struct strbuf *sb, - const struct repository *repo, - const char *fmt, ...) +const char *repo_git_path_append(const struct repository *repo, + struct strbuf *sb, + const char *fmt, ...) { va_list args; va_start(args, fmt); repo_git_pathv(repo, NULL, sb, fmt, args); va_end(args); + return sb->buf; +} + +const char *repo_git_path_replace(const struct repository *repo, + struct strbuf *sb, + const char *fmt, ...) +{ + va_list args; + strbuf_reset(sb); + va_start(args, fmt); + repo_git_pathv(repo, NULL, sb, fmt, args); + va_end(args); + return sb->buf; } char *mkpathdup(const char *fmt, ...) diff --git a/path.h b/path.h index 3c75495e1a55e7..c45311b0a839a2 100644 --- a/path.h +++ b/path.h @@ -52,29 +52,16 @@ const char *repo_common_path_replace(const struct repository *repo, * For an exhaustive list of the adjustments made look at `common_list` and * `adjust_git_path` in path.c. */ - -/* - * Return a path into the git directory of repository `repo`. - */ char *repo_git_path(const struct repository *repo, const char *fmt, ...) __attribute__((format (printf, 2, 3))); - -/* - * Print a path into the git directory of repository `repo` into the provided - * buffer. - */ -void repo_git_pathv(const struct repository *repo, - const struct worktree *wt, struct strbuf *buf, - const char *fmt, va_list args); - -/* - * Construct a path into the git directory of repository `repo` and append it - * to the provided buffer `sb`. - */ -void strbuf_repo_git_path(struct strbuf *sb, - const struct repository *repo, - const char *fmt, ...) +const char *repo_git_path_append(const struct repository *repo, + struct strbuf *sb, + const char *fmt, ...) + __attribute__((format (printf, 3, 4))); +const char *repo_git_path_replace(const struct repository *repo, + struct strbuf *sb, + const char *fmt, ...) __attribute__((format (printf, 3, 4))); /* @@ -241,11 +228,14 @@ struct strbuf *get_pathname(void); # include "strbuf.h" # include "repository.h" -/* Internal implementation detail that should not be used. */ +/* Internal implementation details that should not be used. */ void repo_common_pathv(const struct repository *repo, struct strbuf *buf, const char *fmt, va_list args); +void repo_git_pathv(const struct repository *repo, + const struct worktree *wt, struct strbuf *buf, + const char *fmt, va_list args); /* * Return a statically allocated path into the main repository's diff --git a/submodule.c b/submodule.c index b361076c5b97a0..211ead54a0f8ae 100644 --- a/submodule.c +++ b/submodule.c @@ -1315,7 +1315,7 @@ static int repo_has_absorbed_submodules(struct repository *r) int ret; struct strbuf buf = STRBUF_INIT; - strbuf_repo_git_path(&buf, r, "modules/"); + repo_git_path_append(r, &buf, "modules/"); ret = file_exists(buf.buf) && !is_empty_dir(buf.buf); strbuf_release(&buf); return ret; @@ -2629,6 +2629,6 @@ void submodule_name_to_gitdir(struct strbuf *buf, struct repository *r, * administrators can explicitly set. Nothing has been decided, * so for now, just append the name at the end of the path. */ - strbuf_repo_git_path(buf, r, "modules/"); + repo_git_path_append(r, buf, "modules/"); strbuf_addstr(buf, submodule_name); } From 93a8cfaf3c24f8c1f999b2ca5532ff8f46e0808d Mon Sep 17 00:00:00 2001 From: Patrick Steinhardt Date: Fri, 7 Feb 2025 12:03:28 +0100 Subject: [PATCH 003/312] path: refactor `repo_worktree_path()` family of functions As explained in an earlier commit, we're refactoring path-related functions to provide a consistent interface for computing paths into the commondir, gitdir and worktree. Refactor the "worktree" family of functions accordingly. Signed-off-by: Patrick Steinhardt Signed-off-by: Junio C Hamano --- path.c | 28 ++++++++++++++++++++++------ path.h | 20 +++++++++----------- repository.c | 4 ++-- 3 files changed, 33 insertions(+), 19 deletions(-) diff --git a/path.c b/path.c index 779aa94b567a9d..499116dd1e0760 100644 --- a/path.c +++ b/path.c @@ -519,28 +519,44 @@ char *repo_worktree_path(const struct repository *repo, const char *fmt, ...) struct strbuf path = STRBUF_INIT; va_list args; + va_start(args, fmt); + do_worktree_path(repo, &path, fmt, args); + va_end(args); + + return strbuf_detach(&path, NULL); +} + +const char *repo_worktree_path_append(const struct repository *repo, + struct strbuf *sb, + const char *fmt, ...) +{ + va_list args; + if (!repo->worktree) return NULL; va_start(args, fmt); - do_worktree_path(repo, &path, fmt, args); + do_worktree_path(repo, sb, fmt, args); va_end(args); - return strbuf_detach(&path, NULL); + return sb->buf; } -void strbuf_repo_worktree_path(struct strbuf *sb, - const struct repository *repo, - const char *fmt, ...) +const char *repo_worktree_path_replace(const struct repository *repo, + struct strbuf *sb, + const char *fmt, ...) { va_list args; + strbuf_reset(sb); if (!repo->worktree) - return; + return NULL; va_start(args, fmt); do_worktree_path(repo, sb, fmt, args); va_end(args); + + return sb->buf; } /* Returns 0 on success, negative on failure. */ diff --git a/path.h b/path.h index c45311b0a839a2..d3f85f06762c67 100644 --- a/path.h +++ b/path.h @@ -75,24 +75,22 @@ const char *worktree_git_path(struct repository *r, __attribute__((format (printf, 3, 4))); /* - * Return a path into the worktree of repository `repo`. + * The `repo_worktree_path` family of functions will construct a path into a + * repository's worktree. * - * If the repository doesn't have a worktree NULL is returned. + * Returns a `NULL` pointer in case the repository has no worktree. */ char *repo_worktree_path(const struct repository *repo, const char *fmt, ...) __attribute__((format (printf, 2, 3))); - -/* - * Construct a path into the worktree of repository `repo` and append it - * to the provided buffer `sb`. - * - * If the repository doesn't have a worktree nothing will be appended to `sb`. - */ -void strbuf_repo_worktree_path(struct strbuf *sb, - const struct repository *repo, +const char *repo_worktree_path_append(const struct repository *repo, + struct strbuf *sb, const char *fmt, ...) __attribute__((format (printf, 3, 4))); +const char *repo_worktree_path_replace(const struct repository *repo, + struct strbuf *sb, + const char *fmt, ...) + __attribute__((format (printf, 3, 4))); /* * Return a path into a submodule's git directory located at `path`. `path` diff --git a/repository.c b/repository.c index 1a6a62bbd03a5d..648cd88474eda3 100644 --- a/repository.c +++ b/repository.c @@ -312,8 +312,8 @@ int repo_submodule_init(struct repository *subrepo, struct strbuf worktree = STRBUF_INIT; int ret = 0; - strbuf_repo_worktree_path(&gitdir, superproject, "%s/.git", path); - strbuf_repo_worktree_path(&worktree, superproject, "%s", path); + repo_worktree_path_append(superproject, &gitdir, "%s/.git", path); + repo_worktree_path_append(superproject, &worktree, "%s", path); if (repo_init(subrepo, gitdir.buf, worktree.buf)) { /* From f9467895d884908d5588fc920997b2e53dfb3302 Mon Sep 17 00:00:00 2001 From: Patrick Steinhardt Date: Fri, 7 Feb 2025 12:03:29 +0100 Subject: [PATCH 004/312] submodule: refactor `submodule_to_gitdir()` to accept a repo The `submodule_to_gitdir()` function implicitly uses `the_repository` to resolve submodule paths. Refactor the function to instead accept a repo as parameter to remove the dependency on global state. Signed-off-by: Patrick Steinhardt Signed-off-by: Junio C Hamano --- builtin/submodule--helper.c | 2 +- path.c | 2 +- refs.c | 2 +- submodule.c | 11 ++++++----- submodule.h | 3 ++- 5 files changed, 11 insertions(+), 9 deletions(-) diff --git a/builtin/submodule--helper.c b/builtin/submodule--helper.c index f9b970f8a64a54..3a64f7e605e64e 100644 --- a/builtin/submodule--helper.c +++ b/builtin/submodule--helper.c @@ -1301,7 +1301,7 @@ static void sync_submodule(const char *path, const char *prefix, remote_key = xstrfmt("remote.%s.url", default_remote); free(default_remote); - submodule_to_gitdir(&sb, path); + submodule_to_gitdir(the_repository, &sb, path); strbuf_addstr(&sb, "/config"); if (git_config_set_in_file_gently(sb.buf, remote_key, NULL, sub_origin_url)) diff --git a/path.c b/path.c index 499116dd1e0760..a7fa42162e61e1 100644 --- a/path.c +++ b/path.c @@ -567,7 +567,7 @@ static int do_submodule_path(struct strbuf *buf, const char *path, struct strbuf git_submodule_dir = STRBUF_INIT; int ret; - ret = submodule_to_gitdir(&git_submodule_dir, path); + ret = submodule_to_gitdir(the_repository, &git_submodule_dir, path); if (ret) goto cleanup; diff --git a/refs.c b/refs.c index daf6a842050a16..e1293e53aaa930 100644 --- a/refs.c +++ b/refs.c @@ -2146,7 +2146,7 @@ struct ref_store *repo_get_submodule_ref_store(struct repository *repo, if (!is_nonbare_repository_dir(&submodule_sb)) goto done; - if (submodule_to_gitdir(&submodule_sb, submodule)) + if (submodule_to_gitdir(repo, &submodule_sb, submodule)) goto done; subrepo = xmalloc(sizeof(*subrepo)); diff --git a/submodule.c b/submodule.c index 211ead54a0f8ae..0530e8cf24e045 100644 --- a/submodule.c +++ b/submodule.c @@ -536,7 +536,8 @@ static struct repository *open_submodule(const char *path) struct strbuf sb = STRBUF_INIT; struct repository *out = xmalloc(sizeof(*out)); - if (submodule_to_gitdir(&sb, path) || repo_init(out, sb.buf, NULL)) { + if (submodule_to_gitdir(the_repository, &sb, path) || + repo_init(out, sb.buf, NULL)) { strbuf_release(&sb); free(out); return NULL; @@ -2572,7 +2573,8 @@ int get_superproject_working_tree(struct strbuf *buf) * Put the gitdir for a submodule (given relative to the main * repository worktree) into `buf`, or return -1 on error. */ -int submodule_to_gitdir(struct strbuf *buf, const char *submodule) +int submodule_to_gitdir(struct repository *repo, + struct strbuf *buf, const char *submodule) { const struct submodule *sub; const char *git_dir; @@ -2592,14 +2594,13 @@ int submodule_to_gitdir(struct strbuf *buf, const char *submodule) strbuf_addstr(buf, git_dir); } if (!is_git_directory(buf->buf)) { - sub = submodule_from_path(the_repository, null_oid(), - submodule); + sub = submodule_from_path(repo, null_oid(), submodule); if (!sub) { ret = -1; goto cleanup; } strbuf_reset(buf); - submodule_name_to_gitdir(buf, the_repository, sub->name); + submodule_name_to_gitdir(buf, repo, sub->name); } cleanup: diff --git a/submodule.h b/submodule.h index 4deb1b5f84e71e..db980c1d083bc0 100644 --- a/submodule.h +++ b/submodule.h @@ -136,7 +136,8 @@ int push_unpushed_submodules(struct repository *r, * path of that submodule in 'buf'. Return -1 on error or when the * submodule is not initialized. */ -int submodule_to_gitdir(struct strbuf *buf, const char *submodule); +int submodule_to_gitdir(struct repository *repo, + struct strbuf *buf, const char *submodule); /* * Given a submodule name, create a path to where the submodule's gitdir lives From f5c714e2a7d6239548b94c37ae906484e94b5bc7 Mon Sep 17 00:00:00 2001 From: Patrick Steinhardt Date: Fri, 7 Feb 2025 12:03:30 +0100 Subject: [PATCH 005/312] path: refactor `repo_submodule_path()` family of functions As explained in an earlier commit, we're refactoring path-related functions to provide a consistent interface for computing paths into the commondir, gitdir and worktree. Refactor the "submodule" family of functions accordingly. Note that in contrast to the other `repo_*_path()` families, we have to pass in the repository as a non-constant pointer. This is because we end up calling `repo_read_gitmodules()` deep down in the callstack, which may end up modifying the repository. Signed-off-by: Patrick Steinhardt Signed-off-by: Junio C Hamano --- builtin/submodule--helper.c | 2 +- path.c | 37 +++++++++++++++++++++++++++++-------- path.h | 30 ++++++++++++++++++------------ t/helper/test-ref-store.c | 7 +++---- worktree.c | 3 ++- 5 files changed, 53 insertions(+), 26 deletions(-) diff --git a/builtin/submodule--helper.c b/builtin/submodule--helper.c index 3a64f7e605e64e..c1a8029714bfe9 100644 --- a/builtin/submodule--helper.c +++ b/builtin/submodule--helper.c @@ -1826,7 +1826,7 @@ static int clone_submodule(const struct module_clone_data *clone_data, connect_work_tree_and_git_dir(clone_data_path, sm_gitdir, 0); - p = git_pathdup_submodule(clone_data_path, "config"); + p = repo_submodule_path(the_repository, clone_data_path, "config"); if (!p) die(_("could not get submodule directory for '%s'"), clone_data_path); diff --git a/path.c b/path.c index a7fa42162e61e1..0d81e9fc328df4 100644 --- a/path.c +++ b/path.c @@ -560,14 +560,15 @@ const char *repo_worktree_path_replace(const struct repository *repo, } /* Returns 0 on success, negative on failure. */ -static int do_submodule_path(struct strbuf *buf, const char *path, +static int do_submodule_path(struct repository *repo, + struct strbuf *buf, const char *path, const char *fmt, va_list args) { struct strbuf git_submodule_common_dir = STRBUF_INIT; struct strbuf git_submodule_dir = STRBUF_INIT; int ret; - ret = submodule_to_gitdir(the_repository, &git_submodule_dir, path); + ret = submodule_to_gitdir(repo, &git_submodule_dir, path); if (ret) goto cleanup; @@ -586,13 +587,14 @@ static int do_submodule_path(struct strbuf *buf, const char *path, return ret; } -char *git_pathdup_submodule(const char *path, const char *fmt, ...) +char *repo_submodule_path(struct repository *repo, + const char *path, const char *fmt, ...) { int err; va_list args; struct strbuf buf = STRBUF_INIT; va_start(args, fmt); - err = do_submodule_path(&buf, path, fmt, args); + err = do_submodule_path(repo, &buf, path, fmt, args); va_end(args); if (err) { strbuf_release(&buf); @@ -601,16 +603,35 @@ char *git_pathdup_submodule(const char *path, const char *fmt, ...) return strbuf_detach(&buf, NULL); } -int strbuf_git_path_submodule(struct strbuf *buf, const char *path, - const char *fmt, ...) +const char *repo_submodule_path_append(struct repository *repo, + struct strbuf *buf, + const char *path, + const char *fmt, ...) { int err; va_list args; va_start(args, fmt); - err = do_submodule_path(buf, path, fmt, args); + err = do_submodule_path(repo, buf, path, fmt, args); va_end(args); + if (err) + return NULL; + return buf->buf; +} - return err; +const char *repo_submodule_path_replace(struct repository *repo, + struct strbuf *buf, + const char *path, + const char *fmt, ...) +{ + int err; + va_list args; + strbuf_reset(buf); + va_start(args, fmt); + err = do_submodule_path(repo, buf, path, fmt, args); + va_end(args); + if (err) + return NULL; + return buf->buf; } void repo_common_pathv(const struct repository *repo, diff --git a/path.h b/path.h index d3f85f06762c67..4fe523626c5059 100644 --- a/path.h +++ b/path.h @@ -93,20 +93,26 @@ const char *repo_worktree_path_replace(const struct repository *repo, __attribute__((format (printf, 3, 4))); /* - * Return a path into a submodule's git directory located at `path`. `path` - * must only reference a submodule of the main repository (the_repository). - */ -char *git_pathdup_submodule(const char *path, const char *fmt, ...) - __attribute__((format (printf, 2, 3))); - -/* - * Construct a path into a submodule's git directory located at `path` and - * append it to the provided buffer `sb`. `path` must only reference a - * submodule of the main repository (the_repository). + * The `repo_submodule_path` family of functions will construct a path into a + * submodule's git directory located at `path`. `path` must be a submodule path + * as found in the index and must be part of the given repository. + * + * Returns a `NULL` pointer in case the submodule cannot be found. */ -int strbuf_git_path_submodule(struct strbuf *sb, const char *path, - const char *fmt, ...) +char *repo_submodule_path(struct repository *repo, + const char *path, + const char *fmt, ...) __attribute__((format (printf, 3, 4))); +const char *repo_submodule_path_append(struct repository *repo, + struct strbuf *sb, + const char *path, + const char *fmt, ...) + __attribute__((format (printf, 4, 5))); +const char *repo_submodule_path_replace(struct repository *repo, + struct strbuf *sb, + const char *path, + const char *fmt, ...) + __attribute__((format (printf, 4, 5))); void report_linked_checkout_garbage(struct repository *r); diff --git a/t/helper/test-ref-store.c b/t/helper/test-ref-store.c index 1cc05f043ac204..e00fce592b10ed 100644 --- a/t/helper/test-ref-store.c +++ b/t/helper/test-ref-store.c @@ -75,11 +75,10 @@ static const char **get_store(const char **argv, struct ref_store **refs) *refs = get_main_ref_store(the_repository); } else if (skip_prefix(argv[0], "submodule:", &gitdir)) { struct strbuf sb = STRBUF_INIT; - int ret; - ret = strbuf_git_path_submodule(&sb, gitdir, "objects/"); - if (ret) - die("strbuf_git_path_submodule failed: %d", ret); + if (!repo_submodule_path_append(the_repository, + &sb, gitdir, "objects/")) + die("computing submodule path failed"); add_to_alternates_memory(sb.buf); strbuf_release(&sb); diff --git a/worktree.c b/worktree.c index f8d6e7127f4ab8..8f4fc10c44b513 100644 --- a/worktree.c +++ b/worktree.c @@ -487,7 +487,8 @@ int submodule_uses_worktrees(const char *path) int ret = 0; struct repository_format format = REPOSITORY_FORMAT_INIT; - submodule_gitdir = git_pathdup_submodule(path, "%s", ""); + submodule_gitdir = repo_submodule_path(the_repository, + path, "%s", ""); if (!submodule_gitdir) return 0; From 7f17900b5ba2569d9d07352be51ce90a249a0b46 Mon Sep 17 00:00:00 2001 From: Patrick Steinhardt Date: Fri, 7 Feb 2025 12:03:31 +0100 Subject: [PATCH 006/312] path: drop unused `strbuf_git_path()` function The `strbuf_git_path()` function isn't used anywhere, and neither should it grow any callers because it depends on `the_repository`. Remove it. Signed-off-by: Patrick Steinhardt Signed-off-by: Junio C Hamano --- path.h | 13 ------------- 1 file changed, 13 deletions(-) diff --git a/path.h b/path.h index 4fe523626c5059..8798db74694b23 100644 --- a/path.h +++ b/path.h @@ -272,19 +272,6 @@ static inline char *git_path_buf(struct strbuf *buf, const char *fmt, ...) return buf->buf; } -/* - * Construct a path into the main repository's (the_repository) git directory - * and append it to the provided buffer `sb`. - */ -__attribute__((format (printf, 2, 3))) -static inline void strbuf_git_path(struct strbuf *sb, const char *fmt, ...) -{ - va_list args; - va_start(args, fmt); - repo_git_pathv(the_repository, NULL, sb, fmt, args); - va_end(args); -} - /* * Return a statically allocated path into the main repository's * (the_repository) git directory. From bba59f58a4eeda6fafaa3d41e14f3d00a179923f Mon Sep 17 00:00:00 2001 From: Patrick Steinhardt Date: Fri, 7 Feb 2025 12:03:32 +0100 Subject: [PATCH 007/312] path: drop `git_pathdup()` in favor of `repo_git_path()` Remove `git_pathdup()` in favor of `repo_git_path()`. The latter does essentially the same, with the only exception that it does not rely on `the_repository` but takes the repo as separate parameter. Signed-off-by: Patrick Steinhardt Signed-off-by: Junio C Hamano --- bisect.c | 2 +- builtin/am.c | 2 +- builtin/clone.c | 2 +- builtin/config.c | 8 ++++---- builtin/fast-import.c | 4 ++-- builtin/fsck.c | 2 +- builtin/gc.c | 4 ++-- builtin/notes.c | 2 +- builtin/replace.c | 2 +- builtin/tag.c | 2 +- builtin/worktree.c | 4 ++-- dir.c | 2 +- http-backend.c | 2 +- notes-merge.c | 2 +- object-file.c | 2 +- path.h | 16 +--------------- 16 files changed, 22 insertions(+), 36 deletions(-) diff --git a/bisect.c b/bisect.c index 7a3c77c6d84da0..269a98bf978de6 100644 --- a/bisect.c +++ b/bisect.c @@ -930,7 +930,7 @@ static enum bisect_error check_good_are_ancestors_of_bad(struct repository *r, if (!current_bad_oid) return error(_("a %s revision is needed"), term_bad); - filename = git_pathdup("BISECT_ANCESTORS_OK"); + filename = repo_git_path(the_repository, "BISECT_ANCESTORS_OK"); /* Check if file BISECT_ANCESTORS_OK exists. */ if (!stat(filename, &st) && S_ISREG(st.st_mode)) diff --git a/builtin/am.c b/builtin/am.c index 390b463144202d..2921bb89ef16e6 100644 --- a/builtin/am.c +++ b/builtin/am.c @@ -158,7 +158,7 @@ static void am_state_init(struct am_state *state) memset(state, 0, sizeof(*state)); - state->dir = git_pathdup("rebase-apply"); + state->dir = repo_git_path(the_repository, "rebase-apply"); state->prec = 4; diff --git a/builtin/clone.c b/builtin/clone.c index fd001d800c635e..5ae6ee9db9b4fb 100644 --- a/builtin/clone.c +++ b/builtin/clone.c @@ -938,7 +938,7 @@ static void write_refspec_config(const char *src_ref_prefix, static void dissociate_from_references(void) { - char *alternates = git_pathdup("objects/info/alternates"); + char *alternates = repo_git_path(the_repository, "objects/info/alternates"); if (!access(alternates, F_OK)) { struct child_process cmd = CHILD_PROCESS_INIT; diff --git a/builtin/config.c b/builtin/config.c index 16e6e3055598f1..53a90094e31454 100644 --- a/builtin/config.c +++ b/builtin/config.c @@ -775,13 +775,13 @@ static void location_options_init(struct config_location_options *opts, opts->source.file = opts->file_to_free = git_system_config(); opts->source.scope = CONFIG_SCOPE_SYSTEM; } else if (opts->use_local_config) { - opts->source.file = opts->file_to_free = git_pathdup("config"); + opts->source.file = opts->file_to_free = repo_git_path(the_repository, "config"); opts->source.scope = CONFIG_SCOPE_LOCAL; } else if (opts->use_worktree_config) { struct worktree **worktrees = get_worktrees(); if (the_repository->repository_format_worktree_config) opts->source.file = opts->file_to_free = - git_pathdup("config.worktree"); + repo_git_path(the_repository, "config.worktree"); else if (worktrees[0] && worktrees[1]) die(_("--worktree cannot be used with multiple " "working trees unless the config\n" @@ -790,7 +790,7 @@ static void location_options_init(struct config_location_options *opts, "section in \"git help worktree\" for details")); else opts->source.file = opts->file_to_free = - git_pathdup("config"); + repo_git_path(the_repository, "config"); opts->source.scope = CONFIG_SCOPE_LOCAL; free_worktrees(worktrees); } else if (opts->source.file) { @@ -1087,7 +1087,7 @@ static int show_editor(struct config_location_options *opts) git_config(git_default_config, NULL); config_file = opts->source.file ? xstrdup(opts->source.file) : - git_pathdup("config"); + repo_git_path(the_repository, "config"); if (opts->use_global_config) { int fd = open(config_file, O_CREAT | O_EXCL | O_WRONLY, 0666); if (fd >= 0) { diff --git a/builtin/fast-import.c b/builtin/fast-import.c index a6a84058cde7b5..c6f5147e8b0736 100644 --- a/builtin/fast-import.c +++ b/builtin/fast-import.c @@ -328,7 +328,7 @@ static void write_branch_report(FILE *rpt, struct branch *b) static void write_crash_report(const char *err) { - char *loc = git_pathdup("fast_import_crash_%"PRIuMAX, (uintmax_t) getpid()); + char *loc = repo_git_path(the_repository, "fast_import_crash_%"PRIuMAX, (uintmax_t) getpid()); FILE *rpt = fopen(loc, "w"); struct branch *b; unsigned long lu; @@ -3280,7 +3280,7 @@ static char* make_fast_import_path(const char *path) { if (!relative_marks_paths || is_absolute_path(path)) return prefix_filename(global_prefix, path); - return git_pathdup("info/fast-import/%s", path); + return repo_git_path(the_repository, "info/fast-import/%s", path); } static void option_import_marks(const char *marks, diff --git a/builtin/fsck.c b/builtin/fsck.c index 7a4dcb0716052f..c12203e0121058 100644 --- a/builtin/fsck.c +++ b/builtin/fsck.c @@ -326,7 +326,7 @@ static void check_unreachable_object(struct object *obj) printable_type(&obj->oid, obj->type), describe_object(&obj->oid)); if (write_lost_and_found) { - char *filename = git_pathdup("lost-found/%s/%s", + char *filename = repo_git_path(the_repository, "lost-found/%s/%s", obj->type == OBJ_COMMIT ? "commit" : "other", describe_object(&obj->oid)); FILE *f; diff --git a/builtin/gc.c b/builtin/gc.c index 0bf35334942ca0..57f6aee174fdb2 100644 --- a/builtin/gc.c +++ b/builtin/gc.c @@ -546,7 +546,7 @@ static const char *lock_repo_for_gc(int force, pid_t* ret_pid) if (xgethostname(my_host, sizeof(my_host))) xsnprintf(my_host, sizeof(my_host), "unknown"); - pidfile_path = git_pathdup("gc.pid"); + pidfile_path = repo_git_path(the_repository, "gc.pid"); fd = hold_lock_file_for_update(&lock, pidfile_path, LOCK_DIE_ON_ERROR); if (!force) { @@ -607,7 +607,7 @@ static int report_last_gc_error(void) int ret = 0; ssize_t len; struct stat st; - char *gc_log_path = git_pathdup("gc.log"); + char *gc_log_path = repo_git_path(the_repository, "gc.log"); if (stat(gc_log_path, &st)) { if (errno == ENOENT) diff --git a/builtin/notes.c b/builtin/notes.c index d051abf6dff8f2..18bcbb2f913430 100644 --- a/builtin/notes.c +++ b/builtin/notes.c @@ -197,7 +197,7 @@ static void prepare_note_data(const struct object_id *object, struct note_data * struct strbuf buf = STRBUF_INIT; /* write the template message before editing: */ - d->edit_path = git_pathdup("NOTES_EDITMSG"); + d->edit_path = repo_git_path(the_repository, "NOTES_EDITMSG"); fd = xopen(d->edit_path, O_CREAT | O_TRUNC | O_WRONLY, 0600); if (d->msg_nr) diff --git a/builtin/replace.c b/builtin/replace.c index a4eaadff91f1be..15ec0922ce14d8 100644 --- a/builtin/replace.c +++ b/builtin/replace.c @@ -345,7 +345,7 @@ static int edit_and_replace(const char *object_ref, int force, int raw) } strbuf_release(&ref); - tmpfile = git_pathdup("REPLACE_EDITOBJ"); + tmpfile = repo_git_path(the_repository, "REPLACE_EDITOBJ"); if (export_object(&old_oid, type, raw, tmpfile)) { free(tmpfile); return -1; diff --git a/builtin/tag.c b/builtin/tag.c index e8a344b9264b9c..d3e0943b734759 100644 --- a/builtin/tag.c +++ b/builtin/tag.c @@ -667,7 +667,7 @@ int cmd_tag(int argc, if (create_tag_object) { if (force_sign_annotate && !annotate) opt.sign = 1; - path = git_pathdup("TAG_EDITMSG"); + path = repo_git_path(the_repository, "TAG_EDITMSG"); create_tag(&object, object_ref, tag, &buf, &opt, &prev, &object, &trailer_args, path); } diff --git a/builtin/worktree.c b/builtin/worktree.c index c043d4d523f578..c84e6aa2cb5b90 100644 --- a/builtin/worktree.c +++ b/builtin/worktree.c @@ -337,7 +337,7 @@ static void check_candidate_path(const char *path, static void copy_sparse_checkout(const char *worktree_git_dir) { - char *from_file = git_pathdup("info/sparse-checkout"); + char *from_file = repo_git_path(the_repository, "info/sparse-checkout"); char *to_file = xstrfmt("%s/info/sparse-checkout", worktree_git_dir); if (file_exists(from_file)) { @@ -353,7 +353,7 @@ static void copy_sparse_checkout(const char *worktree_git_dir) static void copy_filtered_worktree_config(const char *worktree_git_dir) { - char *from_file = git_pathdup("config.worktree"); + char *from_file = repo_git_path(the_repository, "config.worktree"); char *to_file = xstrfmt("%s/config.worktree", worktree_git_dir); if (file_exists(from_file)) { diff --git a/dir.c b/dir.c index 5b2181e5899ce9..4122f6513d82b2 100644 --- a/dir.c +++ b/dir.c @@ -3455,7 +3455,7 @@ void setup_standard_excludes(struct dir_struct *dir) char *get_sparse_checkout_filename(void) { - return git_pathdup("info/sparse-checkout"); + return repo_git_path(the_repository, "info/sparse-checkout"); } int get_sparse_checkout_patterns(struct pattern_list *pl) diff --git a/http-backend.c b/http-backend.c index 33cf3782826f6c..50b2858fad634a 100644 --- a/http-backend.c +++ b/http-backend.c @@ -183,7 +183,7 @@ static void send_strbuf(struct strbuf *hdr, static void send_local_file(struct strbuf *hdr, const char *the_type, const char *name) { - char *p = git_pathdup("%s", name); + char *p = repo_git_path(the_repository, "%s", name); size_t buf_alloc = 8192; char *buf = xmalloc(buf_alloc); int fd; diff --git a/notes-merge.c b/notes-merge.c index 8d701ed428aa62..c997c0c1e3925f 100644 --- a/notes-merge.c +++ b/notes-merge.c @@ -309,7 +309,7 @@ static void write_buf_to_worktree(const struct object_id *obj, const char *buf, unsigned long size) { int fd; - char *path = git_pathdup(NOTES_MERGE_WORKTREE "/%s", oid_to_hex(obj)); + char *path = repo_git_path(the_repository, NOTES_MERGE_WORKTREE "/%s", oid_to_hex(obj)); if (safe_create_leading_directories_const(path)) die_errno("unable to create directory for '%s'", path); diff --git a/object-file.c b/object-file.c index 6ce1caacae7f0b..335cc2a5da24f1 100644 --- a/object-file.c +++ b/object-file.c @@ -717,7 +717,7 @@ static void read_info_alternates(struct repository *r, void add_to_alternates_file(const char *reference) { struct lock_file lock = LOCK_INIT; - char *alts = git_pathdup("objects/info/alternates"); + char *alts = repo_git_path(the_repository, "objects/info/alternates"); FILE *in, *out; int found = 0; diff --git a/path.h b/path.h index 8798db74694b23..65a8f21c4cee60 100644 --- a/path.h +++ b/path.h @@ -292,24 +292,10 @@ static inline const char *git_path(const char *fmt, ...) { \ static char *ret; \ if (!ret) \ - ret = git_pathdup(filename); \ + ret = repo_git_path(the_repository, filename); \ return ret; \ } -/* - * Return a path into the main repository's (the_repository) git directory. - */ -__attribute__((format (printf, 1, 2))) -static inline char *git_pathdup(const char *fmt, ...) -{ - struct strbuf path = STRBUF_INIT; - va_list args; - va_start(args, fmt); - repo_git_pathv(the_repository, NULL, &path, fmt, args); - va_end(args); - return strbuf_detach(&path, NULL); -} - # endif /* USE_THE_REPOSITORY_VARIABLE */ #endif /* PATH_H */ From 3859e3965993493defd39cd54a2ab2097957e270 Mon Sep 17 00:00:00 2001 From: Patrick Steinhardt Date: Fri, 7 Feb 2025 12:03:33 +0100 Subject: [PATCH 008/312] path: drop `git_path_buf()` in favor of `repo_git_path_replace()` Remove `git_path_buf()` in favor of `repo_git_path_replace()`. The latter does essentially the same, with the only exception that it does not rely on `the_repository` but takes the repo as separate parameter. Signed-off-by: Patrick Steinhardt Signed-off-by: Junio C Hamano --- builtin/worktree.c | 2 +- compat/precompose_utf8.c | 6 +++--- notes-merge.c | 4 ++-- object-file.c | 4 ++-- path.h | 16 ---------------- setup.c | 33 ++++++++++++++++----------------- 6 files changed, 24 insertions(+), 41 deletions(-) diff --git a/builtin/worktree.c b/builtin/worktree.c index c84e6aa2cb5b90..7959b10d267561 100644 --- a/builtin/worktree.c +++ b/builtin/worktree.c @@ -457,7 +457,7 @@ static int add_worktree(const char *path, const char *refname, BUG("How come '%s' becomes empty after sanitization?", sb.buf); strbuf_reset(&sb); name = sb_name.buf; - git_path_buf(&sb_repo, "worktrees/%s", name); + repo_git_path_replace(the_repository, &sb_repo, "worktrees/%s", name); len = sb_repo.len; if (safe_create_leading_directories_const(sb_repo.buf)) die_errno(_("could not create leading directories of '%s'"), diff --git a/compat/precompose_utf8.c b/compat/precompose_utf8.c index f7cc7b3be59751..12e38e0ea3c570 100644 --- a/compat/precompose_utf8.c +++ b/compat/precompose_utf8.c @@ -50,15 +50,15 @@ void probe_utf8_pathname_composition(void) int output_fd; if (precomposed_unicode != -1) return; /* We found it defined in the global config, respect it */ - git_path_buf(&path, "%s", auml_nfc); + repo_git_path_replace(the_repository, &path, "%s", auml_nfc); output_fd = open(path.buf, O_CREAT|O_EXCL|O_RDWR, 0600); if (output_fd >= 0) { close(output_fd); - git_path_buf(&path, "%s", auml_nfd); + repo_git_path_replace(the_repository, &path, "%s", auml_nfd); precomposed_unicode = access(path.buf, R_OK) ? 0 : 1; git_config_set("core.precomposeunicode", precomposed_unicode ? "true" : "false"); - git_path_buf(&path, "%s", auml_nfc); + repo_git_path_replace(the_repository, &path, "%s", auml_nfc); if (unlink(path.buf)) die_errno(_("failed to unlink '%s'"), path.buf); } diff --git a/notes-merge.c b/notes-merge.c index c997c0c1e3925f..8c22a171c147ba 100644 --- a/notes-merge.c +++ b/notes-merge.c @@ -695,7 +695,7 @@ int notes_merge_commit(struct notes_merge_options *o, const char *msg = strstr(buffer, "\n\n"); int baselen; - git_path_buf(&path, NOTES_MERGE_WORKTREE); + repo_git_path_replace(the_repository, &path, NOTES_MERGE_WORKTREE); if (o->verbosity >= 3) printf("Committing notes in notes merge worktree at %s\n", path.buf); @@ -757,7 +757,7 @@ int notes_merge_abort(struct notes_merge_options *o) struct strbuf buf = STRBUF_INIT; int ret; - git_path_buf(&buf, NOTES_MERGE_WORKTREE); + repo_git_path_replace(the_repository, &buf, NOTES_MERGE_WORKTREE); if (o->verbosity >= 3) printf("Removing notes merge worktree at %s/*\n", buf.buf); ret = remove_dir_recursively(&buf, REMOVE_DIR_KEEP_TOPLEVEL); diff --git a/object-file.c b/object-file.c index 335cc2a5da24f1..dc9fcaf3e97380 100644 --- a/object-file.c +++ b/object-file.c @@ -476,14 +476,14 @@ int odb_mkstemp(struct strbuf *temp_filename, const char *pattern) * restrictive except to remove write permission. */ int mode = 0444; - git_path_buf(temp_filename, "objects/%s", pattern); + repo_git_path_replace(the_repository, temp_filename, "objects/%s", pattern); fd = git_mkstemp_mode(temp_filename->buf, mode); if (0 <= fd) return fd; /* slow path */ /* some mkstemp implementations erase temp_filename on failure */ - git_path_buf(temp_filename, "objects/%s", pattern); + repo_git_path_replace(the_repository, temp_filename, "objects/%s", pattern); safe_create_leading_directories(temp_filename->buf); return xmkstemp_mode(temp_filename->buf, mode); } diff --git a/path.h b/path.h index 65a8f21c4cee60..cdc26acb740e63 100644 --- a/path.h +++ b/path.h @@ -256,22 +256,6 @@ static inline const char *git_common_path(const char *fmt, ...) return pathname->buf; } -/* - * Construct a path into the main repository's (the_repository) git directory - * and place it in the provided buffer `buf`, the contents of the buffer will - * be overridden. - */ -__attribute__((format (printf, 2, 3))) -static inline char *git_path_buf(struct strbuf *buf, const char *fmt, ...) -{ - va_list args; - strbuf_reset(buf); - va_start(args, fmt); - repo_git_pathv(the_repository, NULL, buf, fmt, args); - va_end(args); - return buf->buf; -} - /* * Return a statically allocated path into the main repository's * (the_repository) git directory. diff --git a/setup.c b/setup.c index 74b5ba532555e1..30889386f761ff 100644 --- a/setup.c +++ b/setup.c @@ -2264,7 +2264,7 @@ static int is_reinit(void) char junk[2]; int ret; - git_path_buf(&buf, "HEAD"); + repo_git_path_replace(the_repository, &buf, "HEAD"); ret = !access(buf.buf, R_OK) || readlink(buf.buf, junk, sizeof(junk) - 1) != -1; strbuf_release(&buf); return ret; @@ -2316,8 +2316,7 @@ static int create_default_files(const char *template_path, int init_shared_repository) { struct stat st1; - struct strbuf buf = STRBUF_INIT; - char *path; + struct strbuf path = STRBUF_INIT; int reinit; int filemode; const char *work_tree = repo_get_work_tree(the_repository); @@ -2358,14 +2357,14 @@ static int create_default_files(const char *template_path, initialize_repository_version(fmt->hash_algo, fmt->ref_storage_format, reinit); /* Check filemode trustability */ - path = git_path_buf(&buf, "config"); + repo_git_path_replace(the_repository, &path, "config"); filemode = TEST_FILEMODE; - if (TEST_FILEMODE && !lstat(path, &st1)) { + if (TEST_FILEMODE && !lstat(path.buf, &st1)) { struct stat st2; - filemode = (!chmod(path, st1.st_mode ^ S_IXUSR) && - !lstat(path, &st2) && + filemode = (!chmod(path.buf, st1.st_mode ^ S_IXUSR) && + !lstat(path.buf, &st2) && st1.st_mode != st2.st_mode && - !chmod(path, st1.st_mode)); + !chmod(path.buf, st1.st_mode)); if (filemode && !reinit && (st1.st_mode & S_IXUSR)) filemode = 0; } @@ -2384,24 +2383,24 @@ static int create_default_files(const char *template_path, if (!reinit) { /* Check if symlink is supported in the work tree */ - path = git_path_buf(&buf, "tXXXXXX"); - if (!close(xmkstemp(path)) && - !unlink(path) && - !symlink("testing", path) && - !lstat(path, &st1) && + repo_git_path_replace(the_repository, &path, "tXXXXXX"); + if (!close(xmkstemp(path.buf)) && + !unlink(path.buf) && + !symlink("testing", path.buf) && + !lstat(path.buf, &st1) && S_ISLNK(st1.st_mode)) - unlink(path); /* good */ + unlink(path.buf); /* good */ else git_config_set("core.symlinks", "false"); /* Check if the filesystem is case-insensitive */ - path = git_path_buf(&buf, "CoNfIg"); - if (!access(path, F_OK)) + repo_git_path_replace(the_repository, &path, "CoNfIg"); + if (!access(path.buf, F_OK)) git_config_set("core.ignorecase", "true"); probe_utf8_pathname_composition(); } - strbuf_release(&buf); + strbuf_release(&path); return reinit; } From 8e4710f011dce286d24838fdafd5ce52cfac5285 Mon Sep 17 00:00:00 2001 From: Patrick Steinhardt Date: Fri, 7 Feb 2025 12:03:34 +0100 Subject: [PATCH 009/312] worktree: return allocated string from `get_worktree_git_dir()` The `get_worktree_git_dir()` function returns a string constant that does not need to be free'd by the caller. This string is computed for three different cases: - If we don't have a worktree we return a path into the Git directory. The returned string is owned by `the_repository`, so there is no need for the caller to free it. - If we have a worktree, but no worktree ID then the caller requests the main worktree. In this case we return a path into the common directory, which again is owned by `the_repository` and thus does not need to be free'd. - In the third case, where we have an actual worktree, we compute the path relative to "$GIT_COMMON_DIR/worktrees/". This string does not need to be released either, even though `git_common_path()` ends up allocating memory. But this doesn't result in a memory leak either because we write into a buffer returned by `get_pathname()`, which returns one out of four static buffers. We're about to drop `git_common_path()` in favor of `repo_common_path()`, which doesn't use the same mechanism but instead returns an allocated string owned by the caller. While we could adapt `get_worktree_git_dir()` to also use `get_pathname()` and print the derived common path into that buffer, the whole schema feels a lot like premature optimization in this context. There are some callsites where we call `get_worktree_git_dir()` in a loop that iterates through all worktrees. But none of these loops seem to be even remotely in the hot path, so saving a single allocation there does not feel worth it. Refactor the function to instead consistently return an allocated path so that we can start using `repo_common_path()` in a subsequent commit. Signed-off-by: Patrick Steinhardt Signed-off-by: Junio C Hamano --- branch.c | 7 +++++-- builtin/fsck.c | 8 ++++++-- builtin/receive-pack.c | 4 +++- builtin/worktree.c | 10 ++++++++-- reachable.c | 6 +++++- revision.c | 7 ++++++- worktree.c | 11 ++++++----- worktree.h | 2 +- 8 files changed, 40 insertions(+), 15 deletions(-) diff --git a/branch.c b/branch.c index 77716966feb918..91297d55ac9f60 100644 --- a/branch.c +++ b/branch.c @@ -397,7 +397,7 @@ static void prepare_checked_out_branches(void) worktrees = get_worktrees(); while (worktrees[i]) { - char *old; + char *old, *wt_gitdir; struct wt_status_state state = { 0 }; struct worktree *wt = worktrees[i++]; struct string_list update_refs = STRING_LIST_INIT_DUP; @@ -437,7 +437,8 @@ static void prepare_checked_out_branches(void) } wt_status_state_free_buffers(&state); - if (!sequencer_get_update_refs_state(get_worktree_git_dir(wt), + wt_gitdir = get_worktree_git_dir(wt); + if (!sequencer_get_update_refs_state(wt_gitdir, &update_refs)) { struct string_list_item *item; for_each_string_list_item(item, &update_refs) { @@ -448,6 +449,8 @@ static void prepare_checked_out_branches(void) } string_list_clear(&update_refs, 1); } + + free(wt_gitdir); } free_worktrees(worktrees); diff --git a/builtin/fsck.c b/builtin/fsck.c index c12203e0121058..eea1d43647f224 100644 --- a/builtin/fsck.c +++ b/builtin/fsck.c @@ -1057,7 +1057,7 @@ int cmd_fsck(int argc, struct worktree *wt = *p; struct index_state istate = INDEX_STATE_INIT(the_repository); - char *path; + char *path, *wt_gitdir; /* * Make a copy since the buffer is reusable @@ -1065,9 +1065,13 @@ int cmd_fsck(int argc, * while we're examining the index. */ path = xstrdup(worktree_git_path(the_repository, wt, "index")); - read_index_from(&istate, path, get_worktree_git_dir(wt)); + wt_gitdir = get_worktree_git_dir(wt); + + read_index_from(&istate, path, wt_gitdir); fsck_index(&istate, path, wt->is_current); + discard_index(&istate); + free(wt_gitdir); free(path); } free_worktrees(worktrees); diff --git a/builtin/receive-pack.c b/builtin/receive-pack.c index b7ea7746094305..d65a441f32d1d4 100644 --- a/builtin/receive-pack.c +++ b/builtin/receive-pack.c @@ -1435,7 +1435,8 @@ static const char *push_to_checkout(unsigned char *hash, static const char *update_worktree(unsigned char *sha1, const struct worktree *worktree) { - const char *retval, *git_dir; + const char *retval; + char *git_dir; struct strvec env = STRVEC_INIT; int invoked_hook; @@ -1453,6 +1454,7 @@ static const char *update_worktree(unsigned char *sha1, const struct worktree *w retval = push_to_deploy(sha1, &env, worktree->path); strvec_clear(&env); + free(git_dir); return retval; } diff --git a/builtin/worktree.c b/builtin/worktree.c index 7959b10d267561..2cea9441a6f215 100644 --- a/builtin/worktree.c +++ b/builtin/worktree.c @@ -657,8 +657,9 @@ static int can_use_local_refs(const struct add_opts *opts) if (!opts->quiet) { struct strbuf path = STRBUF_INIT; struct strbuf contents = STRBUF_INIT; + char *wt_gitdir = get_worktree_git_dir(NULL); - strbuf_add_real_path(&path, get_worktree_git_dir(NULL)); + strbuf_add_real_path(&path, wt_gitdir); strbuf_addstr(&path, "/HEAD"); strbuf_read_file(&contents, path.buf, 64); strbuf_stripspace(&contents, NULL); @@ -670,6 +671,7 @@ static int can_use_local_refs(const struct add_opts *opts) path.buf, contents.buf); strbuf_release(&path); strbuf_release(&contents); + free(wt_gitdir); } return 1; } @@ -1157,6 +1159,9 @@ static void validate_no_submodules(const struct worktree *wt) struct index_state istate = INDEX_STATE_INIT(the_repository); struct strbuf path = STRBUF_INIT; int i, found_submodules = 0; + char *wt_gitdir; + + wt_gitdir = get_worktree_git_dir(wt); if (is_directory(worktree_git_path(the_repository, wt, "modules"))) { /* @@ -1166,7 +1171,7 @@ static void validate_no_submodules(const struct worktree *wt) */ found_submodules = 1; } else if (read_index_from(&istate, worktree_git_path(the_repository, wt, "index"), - get_worktree_git_dir(wt)) > 0) { + wt_gitdir) > 0) { for (i = 0; i < istate.cache_nr; i++) { struct cache_entry *ce = istate.cache[i]; int err; @@ -1185,6 +1190,7 @@ static void validate_no_submodules(const struct worktree *wt) } discard_index(&istate); strbuf_release(&path); + free(wt_gitdir); if (found_submodules) die(_("working trees containing submodules cannot be moved or removed")); diff --git a/reachable.c b/reachable.c index ecf7ccf5041a13..9ee04c89ec6151 100644 --- a/reachable.c +++ b/reachable.c @@ -65,8 +65,10 @@ static void add_rebase_files(struct rev_info *revs) struct worktree **worktrees = get_worktrees(); for (struct worktree **wt = worktrees; *wt; wt++) { + char *wt_gitdir = get_worktree_git_dir(*wt); + strbuf_reset(&buf); - strbuf_addstr(&buf, get_worktree_git_dir(*wt)); + strbuf_addstr(&buf, wt_gitdir); strbuf_complete(&buf, '/'); len = buf.len; for (size_t i = 0; i < ARRAY_SIZE(path); i++) { @@ -74,6 +76,8 @@ static void add_rebase_files(struct rev_info *revs) strbuf_addstr(&buf, path[i]); add_one_file(buf.buf, revs); } + + free(wt_gitdir); } strbuf_release(&buf); free_worktrees(worktrees); diff --git a/revision.c b/revision.c index 474fa1e767c8d8..c4390f0938cbde 100644 --- a/revision.c +++ b/revision.c @@ -1874,15 +1874,20 @@ void add_index_objects_to_pending(struct rev_info *revs, unsigned int flags) for (p = worktrees; *p; p++) { struct worktree *wt = *p; struct index_state istate = INDEX_STATE_INIT(revs->repo); + char *wt_gitdir; if (wt->is_current) continue; /* current index already taken care of */ + wt_gitdir = get_worktree_git_dir(wt); + if (read_index_from(&istate, worktree_git_path(the_repository, wt, "index"), - get_worktree_git_dir(wt)) > 0) + wt_gitdir) > 0) do_add_index_objects_to_pending(revs, &istate, flags); + discard_index(&istate); + free(wt_gitdir); } free_worktrees(worktrees); } diff --git a/worktree.c b/worktree.c index 8f4fc10c44b513..3b94535963313a 100644 --- a/worktree.c +++ b/worktree.c @@ -59,8 +59,9 @@ static void add_head_info(struct worktree *wt) static int is_current_worktree(struct worktree *wt) { char *git_dir = absolute_pathdup(repo_get_git_dir(the_repository)); - const char *wt_git_dir = get_worktree_git_dir(wt); + char *wt_git_dir = get_worktree_git_dir(wt); int is_current = !fspathcmp(git_dir, absolute_path(wt_git_dir)); + free(wt_git_dir); free(git_dir); return is_current; } @@ -175,14 +176,14 @@ struct worktree **get_worktrees(void) return get_worktrees_internal(0); } -const char *get_worktree_git_dir(const struct worktree *wt) +char *get_worktree_git_dir(const struct worktree *wt) { if (!wt) - return repo_get_git_dir(the_repository); + return xstrdup(repo_get_git_dir(the_repository)); else if (!wt->id) - return repo_get_common_dir(the_repository); + return xstrdup(repo_get_common_dir(the_repository)); else - return git_common_path("worktrees/%s", wt->id); + return xstrdup(git_common_path("worktrees/%s", wt->id)); } static struct worktree *find_worktree_by_suffix(struct worktree **list, diff --git a/worktree.h b/worktree.h index 38145df80f4107..16368588a0d46b 100644 --- a/worktree.h +++ b/worktree.h @@ -39,7 +39,7 @@ int submodule_uses_worktrees(const char *path); * Return git dir of the worktree. Note that the path may be relative. * If wt is NULL, git dir of current worktree is returned. */ -const char *get_worktree_git_dir(const struct worktree *wt); +char *get_worktree_git_dir(const struct worktree *wt); /* * Search for the worktree identified unambiguously by `arg` -- typically From 07242c2a5afb2a633feb110b1aa74e2adcc37575 Mon Sep 17 00:00:00 2001 From: Patrick Steinhardt Date: Fri, 7 Feb 2025 12:03:35 +0100 Subject: [PATCH 010/312] path: drop `git_common_path()` in favor of `repo_common_path()` Remove `git_common_path()` in favor of the `repo_common_path()` family of functions, which makes the implicit dependency on `the_repository` go away. Note that `git_common_path()` used to return a string allocated via `get_pathname()`, which uses a rotating set of statically allocated buffers. Consequently, callers didn't have to free the returned string. The same isn't true for `repo_common_path()`, so we also have to add logic to free the returned strings. This refactoring also allows us to remove `repo_common_pathv()` from the public interface. Signed-off-by: Patrick Steinhardt Signed-off-by: Junio C Hamano --- builtin/worktree.c | 16 ++++++++++++---- path.c | 8 ++++---- path.h | 19 ------------------- worktree.c | 32 ++++++++++++++++++++++++-------- 4 files changed, 40 insertions(+), 35 deletions(-) diff --git a/builtin/worktree.c b/builtin/worktree.c index 2cea9441a6f215..761e302a369e39 100644 --- a/builtin/worktree.c +++ b/builtin/worktree.c @@ -151,7 +151,7 @@ static int delete_git_dir(const char *id) struct strbuf sb = STRBUF_INIT; int ret; - strbuf_addstr(&sb, git_common_path("worktrees/%s", id)); + repo_common_path_append(the_repository, &sb, "worktrees/%s", id); ret = remove_dir_recursively(&sb, 0); if (ret < 0 && errno == ENOTDIR) ret = unlink(sb.buf); @@ -1102,6 +1102,7 @@ static int lock_worktree(int ac, const char **av, const char *prefix, OPT_END() }; struct worktree **worktrees, *wt; + char *path; ac = parse_options(ac, av, prefix, options, git_worktree_lock_usage, 0); if (ac != 1) @@ -1122,9 +1123,11 @@ static int lock_worktree(int ac, const char **av, const char *prefix, die(_("'%s' is already locked"), av[0]); } - write_file(git_common_path("worktrees/%s/locked", wt->id), - "%s", reason); + path = repo_common_path(the_repository, "worktrees/%s/locked", wt->id); + write_file(path, "%s", reason); + free_worktrees(worktrees); + free(path); return 0; } @@ -1135,6 +1138,7 @@ static int unlock_worktree(int ac, const char **av, const char *prefix, OPT_END() }; struct worktree **worktrees, *wt; + char *path; int ret; ac = parse_options(ac, av, prefix, options, git_worktree_unlock_usage, 0); @@ -1149,8 +1153,12 @@ static int unlock_worktree(int ac, const char **av, const char *prefix, die(_("The main working tree cannot be locked or unlocked")); if (!worktree_lock_reason(wt)) die(_("'%s' is not locked"), av[0]); - ret = unlink_or_warn(git_common_path("worktrees/%s/locked", wt->id)); + + path = repo_common_path(the_repository, "worktrees/%s/locked", wt->id); + ret = unlink_or_warn(path); + free_worktrees(worktrees); + free(path); return ret; } diff --git a/path.c b/path.c index 0d81e9fc328df4..2d07ba723dfd25 100644 --- a/path.c +++ b/path.c @@ -634,10 +634,10 @@ const char *repo_submodule_path_replace(struct repository *repo, return buf->buf; } -void repo_common_pathv(const struct repository *repo, - struct strbuf *sb, - const char *fmt, - va_list args) +static void repo_common_pathv(const struct repository *repo, + struct strbuf *sb, + const char *fmt, + va_list args) { strbuf_addstr(sb, repo->commondir); if (sb->len && !is_dir_sep(sb->buf[sb->len - 1])) diff --git a/path.h b/path.h index cdc26acb740e63..bed0a4c6f9eca3 100644 --- a/path.h +++ b/path.h @@ -233,29 +233,10 @@ struct strbuf *get_pathname(void); # include "repository.h" /* Internal implementation details that should not be used. */ -void repo_common_pathv(const struct repository *repo, - struct strbuf *buf, - const char *fmt, - va_list args); void repo_git_pathv(const struct repository *repo, const struct worktree *wt, struct strbuf *buf, const char *fmt, va_list args); -/* - * Return a statically allocated path into the main repository's - * (the_repository) common git directory. - */ -__attribute__((format (printf, 1, 2))) -static inline const char *git_common_path(const char *fmt, ...) -{ - struct strbuf *pathname = get_pathname(); - va_list args; - va_start(args, fmt); - repo_common_pathv(the_repository, pathname, fmt, args); - va_end(args); - return pathname->buf; -} - /* * Return a statically allocated path into the main repository's * (the_repository) git directory. diff --git a/worktree.c b/worktree.c index 3b94535963313a..d5d07d7a840be5 100644 --- a/worktree.c +++ b/worktree.c @@ -183,7 +183,7 @@ char *get_worktree_git_dir(const struct worktree *wt) else if (!wt->id) return xstrdup(repo_get_common_dir(the_repository)); else - return xstrdup(git_common_path("worktrees/%s", wt->id)); + return repo_common_path(the_repository, "worktrees/%s", wt->id); } static struct worktree *find_worktree_by_suffix(struct worktree **list, @@ -314,6 +314,7 @@ int validate_worktree(const struct worktree *wt, struct strbuf *errmsg, { struct strbuf wt_path = STRBUF_INIT; struct strbuf realpath = STRBUF_INIT; + struct strbuf buf = STRBUF_INIT; char *path = NULL; int err, ret = -1; @@ -343,7 +344,7 @@ int validate_worktree(const struct worktree *wt, struct strbuf *errmsg, if (!is_absolute_path(wt->path)) { strbuf_addf_gently(errmsg, _("'%s' file does not contain absolute path to the working tree location"), - git_common_path("worktrees/%s/gitdir", wt->id)); + repo_common_path_replace(the_repository, &buf, "worktrees/%s/gitdir", wt->id)); goto done; } @@ -365,14 +366,16 @@ int validate_worktree(const struct worktree *wt, struct strbuf *errmsg, goto done; } - strbuf_realpath(&realpath, git_common_path("worktrees/%s", wt->id), 1); + strbuf_realpath(&realpath, repo_common_path_replace(the_repository, &buf, "worktrees/%s", wt->id), 1); ret = fspathcmp(path, realpath.buf); if (ret) strbuf_addf_gently(errmsg, _("'%s' does not point back to '%s'"), - wt->path, git_common_path("worktrees/%s", wt->id)); + wt->path, repo_common_path_replace(the_repository, &buf, + "worktrees/%s", wt->id)); done: free(path); + strbuf_release(&buf); strbuf_release(&wt_path); strbuf_release(&realpath); return ret; @@ -384,11 +387,13 @@ void update_worktree_location(struct worktree *wt, const char *path_, struct strbuf path = STRBUF_INIT; struct strbuf dotgit = STRBUF_INIT; struct strbuf gitdir = STRBUF_INIT; + char *wt_gitdir; if (is_main_worktree(wt)) BUG("can't relocate main worktree"); - strbuf_realpath(&gitdir, git_common_path("worktrees/%s/gitdir", wt->id), 1); + wt_gitdir = repo_common_path(the_repository, "worktrees/%s/gitdir", wt->id); + strbuf_realpath(&gitdir, wt_gitdir, 1); strbuf_realpath(&path, path_, 1); strbuf_addf(&dotgit, "%s/.git", path.buf); if (fspathcmp(wt->path, path.buf)) { @@ -400,6 +405,7 @@ void update_worktree_location(struct worktree *wt, const char *path_, strbuf_release(&path); strbuf_release(&dotgit); strbuf_release(&gitdir); + free(wt_gitdir); } int is_worktree_being_rebased(const struct worktree *wt, @@ -585,6 +591,7 @@ static void repair_gitfile(struct worktree *wt, struct strbuf backlink = STRBUF_INIT; char *dotgit_contents = NULL; const char *repair = NULL; + char *path = NULL; int err; /* missing worktree can't be repaired */ @@ -596,7 +603,8 @@ static void repair_gitfile(struct worktree *wt, goto done; } - strbuf_realpath(&repo, git_common_path("worktrees/%s", wt->id), 1); + path = repo_common_path(the_repository, "worktrees/%s", wt->id); + strbuf_realpath(&repo, path, 1); strbuf_addf(&dotgit, "%s/.git", wt->path); strbuf_addf(&gitdir, "%s/gitdir", repo.buf); dotgit_contents = xstrdup_or_null(read_gitfile_gently(dotgit.buf, &err)); @@ -626,6 +634,7 @@ static void repair_gitfile(struct worktree *wt, done: free(dotgit_contents); + free(path); strbuf_release(&repo); strbuf_release(&dotgit); strbuf_release(&gitdir); @@ -657,11 +666,13 @@ void repair_worktree_after_gitdir_move(struct worktree *wt, const char *old_path struct strbuf gitdir = STRBUF_INIT; struct strbuf dotgit = STRBUF_INIT; int is_relative_path; + char *path = NULL; if (is_main_worktree(wt)) goto done; - strbuf_realpath(&gitdir, git_common_path("worktrees/%s/gitdir", wt->id), 1); + path = repo_common_path(the_repository, "worktrees/%s/gitdir", wt->id); + strbuf_realpath(&gitdir, path, 1); if (strbuf_read_file(&dotgit, gitdir.buf, 0) < 0) goto done; @@ -680,6 +691,7 @@ void repair_worktree_after_gitdir_move(struct worktree *wt, const char *old_path done: strbuf_release(&gitdir); strbuf_release(&dotgit); + free(path); } void repair_worktrees_after_gitdir_move(const char *old_path) @@ -871,7 +883,11 @@ int should_prune_worktree(const char *id, struct strbuf *reason, char **wtpath, ssize_t read_result; *wtpath = NULL; - strbuf_realpath(&repo, git_common_path("worktrees/%s", id), 1); + + path = repo_common_path(the_repository, "worktrees/%s", id); + strbuf_realpath(&repo, path, 1); + FREE_AND_NULL(path); + strbuf_addf(&gitdir, "%s/gitdir", repo.buf); if (!is_directory(repo.buf)) { strbuf_addstr(reason, _("not a valid directory")); From d460267613da14eba959eb225e2cbf6a1e132eb1 Mon Sep 17 00:00:00 2001 From: Christian Couder Date: Tue, 18 Feb 2025 12:32:02 +0100 Subject: [PATCH 011/312] Add 'promisor-remote' capability to protocol v2 When a server S knows that some objects from a repository are available from a promisor remote X, S might want to suggest to a client C cloning or fetching the repo from S that C may use X directly instead of S for these objects. Note that this could happen both in the case S itself doesn't have the objects and borrows them from X, and in the case S has the objects but knows that X is better connected to the world (e.g., it is in a $LARGEINTERNETCOMPANY datacenter with petabit/s backbone connections) than S. Implementation of the latter case, which would require S to omit in its response the objects available on X, is left for future improvement though. Then C might or might not, want to get the objects from X. If S and C can agree on C using X directly, S can then omit objects that can be obtained from X when answering C's request. To allow S and C to agree and let each other know about C using X or not, let's introduce a new "promisor-remote" capability in the protocol v2, as well as a few new configuration variables: - "promisor.advertise" on the server side, and: - "promisor.acceptFromServer" on the client side. By default, or if "promisor.advertise" is set to 'false', a server S will not advertise the "promisor-remote" capability. If S doesn't advertise the "promisor-remote" capability, then a client C replying to S shouldn't advertise the "promisor-remote" capability either. If "promisor.advertise" is set to 'true', S will advertise its promisor remotes with a string like: promisor-remote=[;]... where each element contains information about a single promisor remote in the form: name=[,url=] where is the urlencoded name of a promisor remote and is the urlencoded URL of the promisor remote named . For now, the URL is passed in addition to the name. In the future, it might be possible to pass other information like a filter-spec that the client may use when cloning from S, or a token that the client may use when retrieving objects from X. It is C's responsibility to arrange how it can reach X though, so pieces of information that are usually outside Git's concern, like proxy configuration, must not be distributed over this protocol. It might also be possible in the future for "promisor.advertise" to have other values. For example a value like "onlyName" could prevent S from advertising URLs, which could help in case C should use a different URL for X than the URL S is using. (The URL S is using might be an internal one on the server side for example.) By default or if "promisor.acceptFromServer" is set to "None", C will not accept to use the promisor remotes that might have been advertised by S. In this case, C will not advertise any "promisor-remote" capability in its reply to S. If "promisor.acceptFromServer" is set to "All" and S advertised some promisor remotes, then on the contrary, C will accept to use all the promisor remotes that S advertised and C will reply with a string like: promisor-remote=[;]... where the elements are the urlencoded names of all the promisor remotes S advertised. In a following commit, other values for "promisor.acceptFromServer" will be implemented, so that C will be able to decide the promisor remotes it accepts depending on the name and URL it received from S. So even if that name and URL information is not used much right now, it will be needed soon. Helped-by: Taylor Blau Helped-by: Patrick Steinhardt Signed-off-by: Christian Couder Signed-off-by: Junio C Hamano --- Documentation/config/promisor.adoc | 17 ++ Documentation/gitprotocol-v2.adoc | 54 ++++++ connect.c | 9 + promisor-remote.c | 194 ++++++++++++++++++++ promisor-remote.h | 37 +++- serve.c | 26 +++ t/meson.build | 1 + t/t5710-promisor-remote-capability.sh | 244 ++++++++++++++++++++++++++ upload-pack.c | 3 + 9 files changed, 584 insertions(+), 1 deletion(-) create mode 100755 t/t5710-promisor-remote-capability.sh diff --git a/Documentation/config/promisor.adoc b/Documentation/config/promisor.adoc index 98c5cb2ec20d34..9cbfe3e59ea834 100644 --- a/Documentation/config/promisor.adoc +++ b/Documentation/config/promisor.adoc @@ -1,3 +1,20 @@ promisor.quiet:: If set to "true" assume `--quiet` when fetching additional objects for a partial clone. + +promisor.advertise:: + If set to "true", a server will use the "promisor-remote" + capability, see linkgit:gitprotocol-v2[5], to advertise the + promisor remotes it is using, if it uses some. Default is + "false", which means the "promisor-remote" capability is not + advertised. + +promisor.acceptFromServer:: + If set to "all", a client will accept all the promisor remotes + a server might advertise using the "promisor-remote" + capability. Default is "none", which means no promisor remote + advertised by a server will be accepted. By accepting a + promisor remote, the client agrees that the server might omit + objects that are lazily fetchable from this promisor remote + from its responses to "fetch" and "clone" requests from the + client. See linkgit:gitprotocol-v2[5]. diff --git a/Documentation/gitprotocol-v2.adoc b/Documentation/gitprotocol-v2.adoc index 1652fef3aeb166..c20b74aac05dfa 100644 --- a/Documentation/gitprotocol-v2.adoc +++ b/Documentation/gitprotocol-v2.adoc @@ -781,6 +781,60 @@ retrieving the header from a bundle at the indicated URI, and thus save themselves and the server(s) the request(s) needed to inspect the headers of that bundle or bundles. +promisor-remote= +~~~~~~~~~~~~~~~~~~~~~~~~~~ + +The server may advertise some promisor remotes it is using or knows +about to a client which may want to use them as its promisor remotes, +instead of this repository. In this case should be of the +form: + + pr-infos = pr-info | pr-infos ";" pr-info + + pr-info = "name=" pr-name | "name=" pr-name "," "url=" pr-url + +where `pr-name` is the urlencoded name of a promisor remote, and +`pr-url` the urlencoded URL of that promisor remote. + +In this case, if the client decides to use one or more promisor +remotes the server advertised, it can reply with +"promisor-remote=" where should be of the form: + + pr-names = pr-name | pr-names ";" pr-name + +where `pr-name` is the urlencoded name of a promisor remote the server +advertised and the client accepts. + +Note that, everywhere in this document, `pr-name` MUST be a valid +remote name, and the ';' and ',' characters MUST be encoded if they +appear in `pr-name` or `pr-url`. + +If the server doesn't know any promisor remote that could be good for +a client to use, or prefers a client not to use any promisor remote it +uses or knows about, it shouldn't advertise the "promisor-remote" +capability at all. + +In this case, or if the client doesn't want to use any promisor remote +the server advertised, the client shouldn't advertise the +"promisor-remote" capability at all in its reply. + +The "promisor.advertise" and "promisor.acceptFromServer" configuration +options can be used on the server and client side to control what they +advertise or accept respectively. See the documentation of these +configuration options for more information. + +Note that in the future it would be nice if the "promisor-remote" +protocol capability could be used by the server, when responding to +`git fetch` or `git clone`, to advertise better-connected remotes that +the client can use as promisor remotes, instead of this repository, so +that the client can lazily fetch objects from these other +better-connected remotes. This would require the server to omit in its +response the objects available on the better-connected remotes that +the client has accepted. This hasn't been implemented yet though. So +for now this "promisor-remote" capability is useful only when the +server advertises some promisor remotes it already uses to borrow +objects from. + GIT --- Part of the linkgit:git[1] suite diff --git a/connect.c b/connect.c index 91f399001449c8..125150ac25044b 100644 --- a/connect.c +++ b/connect.c @@ -22,6 +22,7 @@ #include "protocol.h" #include "alias.h" #include "bundle-uri.h" +#include "promisor-remote.h" static char *server_capabilities_v1; static struct strvec server_capabilities_v2 = STRVEC_INIT; @@ -487,6 +488,7 @@ void check_stateless_delimiter(int stateless_rpc, static void send_capabilities(int fd_out, struct packet_reader *reader) { const char *hash_name; + const char *promisor_remote_info; if (server_supports_v2("agent")) packet_write_fmt(fd_out, "agent=%s", git_user_agent_sanitized()); @@ -500,6 +502,13 @@ static void send_capabilities(int fd_out, struct packet_reader *reader) } else { reader->hash_algo = &hash_algos[GIT_HASH_SHA1]; } + if (server_feature_v2("promisor-remote", &promisor_remote_info)) { + char *reply = promisor_remote_reply(promisor_remote_info); + if (reply) { + packet_write_fmt(fd_out, "promisor-remote=%s", reply); + free(reply); + } + } } int get_remote_bundle_uri(int fd_out, struct packet_reader *reader, diff --git a/promisor-remote.c b/promisor-remote.c index c714f4f00728e4..918be6528fe1fc 100644 --- a/promisor-remote.c +++ b/promisor-remote.c @@ -11,6 +11,8 @@ #include "strvec.h" #include "packfile.h" #include "environment.h" +#include "url.h" +#include "version.h" struct promisor_remote_config { struct promisor_remote *promisors; @@ -221,6 +223,18 @@ int repo_has_promisor_remote(struct repository *r) return !!repo_promisor_remote_find(r, NULL); } +int repo_has_accepted_promisor_remote(struct repository *r) +{ + struct promisor_remote *p; + + promisor_remote_init(r); + + for (p = r->promisor_remote_config->promisors; p; p = p->next) + if (p->accepted) + return 1; + return 0; +} + static int remove_fetched_oids(struct repository *repo, struct object_id **oids, int oid_nr, int to_free) @@ -292,3 +306,183 @@ void promisor_remote_get_direct(struct repository *repo, if (to_free) free(remaining_oids); } + +static int allow_unsanitized(char ch) +{ + if (ch == ',' || ch == ';' || ch == '%') + return 0; + return ch > 32 && ch < 127; +} + +static void promisor_info_vecs(struct repository *repo, + struct strvec *names, + struct strvec *urls) +{ + struct promisor_remote *r; + + promisor_remote_init(repo); + + for (r = repo->promisor_remote_config->promisors; r; r = r->next) { + char *url; + char *url_key = xstrfmt("remote.%s.url", r->name); + + strvec_push(names, r->name); + strvec_push(urls, git_config_get_string(url_key, &url) ? NULL : url); + + free(url); + free(url_key); + } +} + +char *promisor_remote_info(struct repository *repo) +{ + struct strbuf sb = STRBUF_INIT; + int advertise_promisors = 0; + struct strvec names = STRVEC_INIT; + struct strvec urls = STRVEC_INIT; + + git_config_get_bool("promisor.advertise", &advertise_promisors); + + if (!advertise_promisors) + return NULL; + + promisor_info_vecs(repo, &names, &urls); + + if (!names.nr) + return NULL; + + for (size_t i = 0; i < names.nr; i++) { + if (i) + strbuf_addch(&sb, ';'); + strbuf_addstr(&sb, "name="); + strbuf_addstr_urlencode(&sb, names.v[i], allow_unsanitized); + if (urls.v[i]) { + strbuf_addstr(&sb, ",url="); + strbuf_addstr_urlencode(&sb, urls.v[i], allow_unsanitized); + } + } + + strvec_clear(&names); + strvec_clear(&urls); + + return strbuf_detach(&sb, NULL); +} + +enum accept_promisor { + ACCEPT_NONE = 0, + ACCEPT_ALL +}; + +static int should_accept_remote(enum accept_promisor accept, + const char *remote_name UNUSED, + const char *remote_url UNUSED) +{ + if (accept == ACCEPT_ALL) + return 1; + + BUG("Unhandled 'enum accept_promisor' value '%d'", accept); +} + +static void filter_promisor_remote(struct strvec *accepted, const char *info) +{ + struct strbuf **remotes; + const char *accept_str; + enum accept_promisor accept = ACCEPT_NONE; + + if (!git_config_get_string_tmp("promisor.acceptfromserver", &accept_str)) { + if (!*accept_str || !strcasecmp("None", accept_str)) + accept = ACCEPT_NONE; + else if (!strcasecmp("All", accept_str)) + accept = ACCEPT_ALL; + else + warning(_("unknown '%s' value for '%s' config option"), + accept_str, "promisor.acceptfromserver"); + } + + if (accept == ACCEPT_NONE) + return; + + /* Parse remote info received */ + + remotes = strbuf_split_str(info, ';', 0); + + for (size_t i = 0; remotes[i]; i++) { + struct strbuf **elems; + const char *remote_name = NULL; + const char *remote_url = NULL; + char *decoded_name = NULL; + char *decoded_url = NULL; + + strbuf_strip_suffix(remotes[i], ";"); + elems = strbuf_split(remotes[i], ','); + + for (size_t j = 0; elems[j]; j++) { + int res; + strbuf_strip_suffix(elems[j], ","); + res = skip_prefix(elems[j]->buf, "name=", &remote_name) || + skip_prefix(elems[j]->buf, "url=", &remote_url); + if (!res) + warning(_("unknown element '%s' from remote info"), + elems[j]->buf); + } + + if (remote_name) + decoded_name = url_percent_decode(remote_name); + if (remote_url) + decoded_url = url_percent_decode(remote_url); + + if (decoded_name && should_accept_remote(accept, decoded_name, decoded_url)) + strvec_push(accepted, decoded_name); + + strbuf_list_free(elems); + free(decoded_name); + free(decoded_url); + } + + strbuf_list_free(remotes); +} + +char *promisor_remote_reply(const char *info) +{ + struct strvec accepted = STRVEC_INIT; + struct strbuf reply = STRBUF_INIT; + + filter_promisor_remote(&accepted, info); + + if (!accepted.nr) + return NULL; + + for (size_t i = 0; i < accepted.nr; i++) { + if (i) + strbuf_addch(&reply, ';'); + strbuf_addstr_urlencode(&reply, accepted.v[i], allow_unsanitized); + } + + strvec_clear(&accepted); + + return strbuf_detach(&reply, NULL); +} + +void mark_promisor_remotes_as_accepted(struct repository *r, const char *remotes) +{ + struct strbuf **accepted_remotes = strbuf_split_str(remotes, ';', 0); + + for (size_t i = 0; accepted_remotes[i]; i++) { + struct promisor_remote *p; + char *decoded_remote; + + strbuf_strip_suffix(accepted_remotes[i], ";"); + decoded_remote = url_percent_decode(accepted_remotes[i]->buf); + + p = repo_promisor_remote_find(r, decoded_remote); + if (p) + p->accepted = 1; + else + warning(_("accepted promisor remote '%s' not found"), + decoded_remote); + + free(decoded_remote); + } + + strbuf_list_free(accepted_remotes); +} diff --git a/promisor-remote.h b/promisor-remote.h index 88cb599c391aea..263d331a551b6c 100644 --- a/promisor-remote.h +++ b/promisor-remote.h @@ -9,11 +9,13 @@ struct object_id; * Promisor remote linked list * * Information in its fields come from remote.XXX config entries or - * from extensions.partialclone. + * from extensions.partialclone, except for 'accepted' which comes + * from protocol v2 capabilities exchange. */ struct promisor_remote { struct promisor_remote *next; char *partial_clone_filter; + unsigned int accepted : 1; const char name[FLEX_ARRAY]; }; @@ -32,4 +34,37 @@ void promisor_remote_get_direct(struct repository *repo, const struct object_id *oids, int oid_nr); +/* + * Prepare a "promisor-remote" advertisement by a server. + * Check the value of "promisor.advertise" and maybe the configured + * promisor remotes, if any, to prepare information to send in an + * advertisement. + * Return value is NULL if no promisor remote advertisement should be + * made. Otherwise it contains the names and urls of the advertised + * promisor remotes separated by ';'. See gitprotocol-v2(5). + */ +char *promisor_remote_info(struct repository *repo); + +/* + * Prepare a reply to a "promisor-remote" advertisement from a server. + * Check the value of "promisor.acceptfromserver" and maybe the + * configured promisor remotes, if any, to prepare the reply. + * Return value is NULL if no promisor remote from the server + * is accepted. Otherwise it contains the names of the accepted promisor + * remotes separated by ';'. See gitprotocol-v2(5). + */ +char *promisor_remote_reply(const char *info); + +/* + * Set the 'accepted' flag for some promisor remotes. Useful on the + * server side when some promisor remotes have been accepted by the + * client. + */ +void mark_promisor_remotes_as_accepted(struct repository *repo, const char *remotes); + +/* + * Has any promisor remote been accepted by the client? + */ +int repo_has_accepted_promisor_remote(struct repository *r); + #endif /* PROMISOR_REMOTE_H */ diff --git a/serve.c b/serve.c index f6dfe34a2bee6b..e3ccf1505ca1a0 100644 --- a/serve.c +++ b/serve.c @@ -10,6 +10,7 @@ #include "upload-pack.h" #include "bundle-uri.h" #include "trace2.h" +#include "promisor-remote.h" static int advertise_sid = -1; static int advertise_object_info = -1; @@ -29,6 +30,26 @@ static int agent_advertise(struct repository *r UNUSED, return 1; } +static int promisor_remote_advertise(struct repository *r, + struct strbuf *value) +{ + if (value) { + char *info = promisor_remote_info(r); + if (!info) + return 0; + strbuf_addstr(value, info); + free(info); + } + return 1; +} + +static void promisor_remote_receive(struct repository *r, + const char *remotes) +{ + mark_promisor_remotes_as_accepted(r, remotes); +} + + static int object_format_advertise(struct repository *r, struct strbuf *value) { @@ -155,6 +176,11 @@ static struct protocol_capability capabilities[] = { .advertise = bundle_uri_advertise, .command = bundle_uri_command, }, + { + .name = "promisor-remote", + .advertise = promisor_remote_advertise, + .receive = promisor_remote_receive, + }, }; void protocol_v2_advertise_capabilities(struct repository *r) diff --git a/t/meson.build b/t/meson.build index a03ebc81fd6f28..75ad6726c4b10e 100644 --- a/t/meson.build +++ b/t/meson.build @@ -728,6 +728,7 @@ integration_tests = [ 't5703-upload-pack-ref-in-want.sh', 't5704-protocol-violations.sh', 't5705-session-id-in-capabilities.sh', + 't5710-promisor-remote-capability.sh', 't5730-protocol-v2-bundle-uri-file.sh', 't5731-protocol-v2-bundle-uri-git.sh', 't5732-protocol-v2-bundle-uri-http.sh', diff --git a/t/t5710-promisor-remote-capability.sh b/t/t5710-promisor-remote-capability.sh new file mode 100755 index 00000000000000..51cf2269e1d370 --- /dev/null +++ b/t/t5710-promisor-remote-capability.sh @@ -0,0 +1,244 @@ +#!/bin/sh + +test_description='handling of promisor remote advertisement' + +. ./test-lib.sh + +GIT_TEST_MULTI_PACK_INDEX=0 +GIT_TEST_MULTI_PACK_INDEX_WRITE_INCREMENTAL=0 + +# Setup the repository with three commits, this way HEAD is always +# available and we can hide commit 1 or 2. +test_expect_success 'setup: create "template" repository' ' + git init template && + test_commit -C template 1 && + test_commit -C template 2 && + test_commit -C template 3 && + test-tool genrandom foo 10240 >template/foo && + git -C template add foo && + git -C template commit -m foo +' + +# A bare repo will act as a server repo with unpacked objects. +test_expect_success 'setup: create bare "server" repository' ' + git clone --bare --no-local template server && + mv server/objects/pack/pack-* . && + packfile=$(ls pack-*.pack) && + git -C server unpack-objects --strict <"$packfile" +' + +check_missing_objects () { + git -C "$1" rev-list --objects --all --missing=print > all.txt && + perl -ne 'print if s/^[?]//' all.txt >missing.txt && + test_line_count = "$2" missing.txt && + if test "$2" -lt 2 + then + test "$3" = "$(cat missing.txt)" + else + test -f "$3" && + sort <"$3" >expected_sorted && + sort actual_sorted && + test_cmp expected_sorted actual_sorted + fi +} + +initialize_server () { + count="$1" + missing_oids="$2" + + # Repack everything first + git -C server -c repack.writebitmaps=false repack -a -d && + + # Remove promisor file in case they exist, useful when reinitializing + rm -rf server/objects/pack/*.promisor && + + # Repack without the largest object and create a promisor pack on server + git -C server -c repack.writebitmaps=false repack -a -d \ + --filter=blob:limit=5k --filter-to="$(pwd)/pack" && + promisor_file=$(ls server/objects/pack/*.pack | sed "s/\.pack/.promisor/") && + >"$promisor_file" && + + # Check objects missing on the server + check_missing_objects server "$count" "$missing_oids" +} + +copy_to_lop () { + oid_path="$(test_oid_to_path $1)" && + path="server/objects/$oid_path" && + path2="lop/objects/$oid_path" && + mkdir -p $(dirname "$path2") && + cp "$path" "$path2" +} + +test_expect_success "setup for testing promisor remote advertisement" ' + # Create another bare repo called "lop" (for Large Object Promisor) + git init --bare lop && + + # Copy the largest object from server to lop + obj="HEAD:foo" && + oid="$(git -C server rev-parse $obj)" && + copy_to_lop "$oid" && + + initialize_server 1 "$oid" && + + # Configure lop as promisor remote for server + git -C server remote add lop "file://$(pwd)/lop" && + git -C server config remote.lop.promisor true && + + git -C lop config uploadpack.allowFilter true && + git -C lop config uploadpack.allowAnySHA1InWant true && + git -C server config uploadpack.allowFilter true && + git -C server config uploadpack.allowAnySHA1InWant true +' + +test_expect_success "clone with promisor.advertise set to 'true'" ' + git -C server config promisor.advertise true && + + # Clone from server to create a client + GIT_NO_LAZY_FETCH=0 git clone -c remote.lop.promisor=true \ + -c remote.lop.fetch="+refs/heads/*:refs/remotes/lop/*" \ + -c remote.lop.url="file://$(pwd)/lop" \ + -c promisor.acceptfromserver=All \ + --no-local --filter="blob:limit=5k" server client && + test_when_finished "rm -rf client" && + + # Check that the largest object is still missing on the server + check_missing_objects server 1 "$oid" +' + +test_expect_success "clone with promisor.advertise set to 'false'" ' + git -C server config promisor.advertise false && + + # Clone from server to create a client + GIT_NO_LAZY_FETCH=0 git clone -c remote.lop.promisor=true \ + -c remote.lop.fetch="+refs/heads/*:refs/remotes/lop/*" \ + -c remote.lop.url="file://$(pwd)/lop" \ + -c promisor.acceptfromserver=All \ + --no-local --filter="blob:limit=5k" server client && + test_when_finished "rm -rf client" && + + # Check that the largest object is not missing on the server + check_missing_objects server 0 "" && + + # Reinitialize server so that the largest object is missing again + initialize_server 1 "$oid" +' + +test_expect_success "clone with promisor.acceptfromserver set to 'None'" ' + git -C server config promisor.advertise true && + + # Clone from server to create a client + GIT_NO_LAZY_FETCH=0 git clone -c remote.lop.promisor=true \ + -c remote.lop.fetch="+refs/heads/*:refs/remotes/lop/*" \ + -c remote.lop.url="file://$(pwd)/lop" \ + -c promisor.acceptfromserver=None \ + --no-local --filter="blob:limit=5k" server client && + test_when_finished "rm -rf client" && + + # Check that the largest object is not missing on the server + check_missing_objects server 0 "" && + + # Reinitialize server so that the largest object is missing again + initialize_server 1 "$oid" +' + +test_expect_success "init + fetch with promisor.advertise set to 'true'" ' + git -C server config promisor.advertise true && + + test_when_finished "rm -rf client" && + mkdir client && + git -C client init && + git -C client config remote.lop.promisor true && + git -C client config remote.lop.fetch "+refs/heads/*:refs/remotes/lop/*" && + git -C client config remote.lop.url "file://$(pwd)/lop" && + git -C client config remote.server.url "file://$(pwd)/server" && + git -C client config remote.server.fetch "+refs/heads/*:refs/remotes/server/*" && + git -C client config promisor.acceptfromserver All && + GIT_NO_LAZY_FETCH=0 git -C client fetch --filter="blob:limit=5k" server && + + # Check that the largest object is still missing on the server + check_missing_objects server 1 "$oid" +' + +test_expect_success "clone with promisor.advertise set to 'true' but don't delete the client" ' + git -C server config promisor.advertise true && + + # Clone from server to create a client + GIT_NO_LAZY_FETCH=0 git clone -c remote.lop.promisor=true \ + -c remote.lop.fetch="+refs/heads/*:refs/remotes/lop/*" \ + -c remote.lop.url="file://$(pwd)/lop" \ + -c promisor.acceptfromserver=All \ + --no-local --filter="blob:limit=5k" server client && + + # Check that the largest object is still missing on the server + check_missing_objects server 1 "$oid" +' + +test_expect_success "setup for subsequent fetches" ' + # Generate new commit with large blob + test-tool genrandom bar 10240 >template/bar && + git -C template add bar && + git -C template commit -m bar && + + # Fetch new commit with large blob + git -C server fetch origin && + git -C server update-ref HEAD FETCH_HEAD && + git -C server rev-parse HEAD >expected_head && + + # Repack everything twice and remove .promisor files before + # each repack. This makes sure everything gets repacked + # into a single packfile. The second repack is necessary + # because the first one fetches from lop and creates a new + # packfile and its associated .promisor file. + + rm -f server/objects/pack/*.promisor && + git -C server -c repack.writebitmaps=false repack -a -d && + rm -f server/objects/pack/*.promisor && + git -C server -c repack.writebitmaps=false repack -a -d && + + # Unpack everything + rm pack-* && + mv server/objects/pack/pack-* . && + packfile=$(ls pack-*.pack) && + git -C server unpack-objects --strict <"$packfile" && + + # Copy new large object to lop + obj_bar="HEAD:bar" && + oid_bar="$(git -C server rev-parse $obj_bar)" && + copy_to_lop "$oid_bar" && + + # Reinitialize server so that the 2 largest objects are missing + printf "%s\n" "$oid" "$oid_bar" >expected_missing.txt && + initialize_server 2 expected_missing.txt && + + # Create one more client + cp -r client client2 +' + +test_expect_success "subsequent fetch from a client when promisor.advertise is true" ' + git -C server config promisor.advertise true && + + GIT_NO_LAZY_FETCH=0 git -C client pull origin && + + git -C client rev-parse HEAD >actual && + test_cmp expected_head actual && + + cat client/bar >/dev/null && + + check_missing_objects server 2 expected_missing.txt +' + +test_expect_success "subsequent fetch from a client when promisor.advertise is false" ' + git -C server config promisor.advertise false && + + GIT_NO_LAZY_FETCH=0 git -C client2 pull origin && + + git -C client2 rev-parse HEAD >actual && + test_cmp expected_head actual && + + cat client2/bar >/dev/null && + + check_missing_objects server 1 "$oid" +' + +test_done diff --git a/upload-pack.c b/upload-pack.c index 728b2477fcc71f..7498b45e2e1e21 100644 --- a/upload-pack.c +++ b/upload-pack.c @@ -32,6 +32,7 @@ #include "write-or-die.h" #include "json-writer.h" #include "strmap.h" +#include "promisor-remote.h" /* Remember to update object flag allocation in object.h */ #define THEY_HAVE (1u << 11) @@ -319,6 +320,8 @@ static void create_pack_file(struct upload_pack_data *pack_data, strvec_push(&pack_objects.args, "--delta-base-offset"); if (pack_data->use_include_tag) strvec_push(&pack_objects.args, "--include-tag"); + if (repo_has_accepted_promisor_remote(the_repository)) + strvec_push(&pack_objects.args, "--missing=allow-promisor"); if (pack_data->filter_options.choice) { const char *spec = expand_list_objects_filter_spec(&pack_data->filter_options); From 36463e32dfef16dc3a903763f14f5bbf6e0248e7 Mon Sep 17 00:00:00 2001 From: Christian Couder Date: Tue, 18 Feb 2025 12:32:03 +0100 Subject: [PATCH 012/312] promisor-remote: check advertised name or URL A previous commit introduced a "promisor.acceptFromServer" configuration variable with only "None" or "All" as valid values. Let's introduce "KnownName" and "KnownUrl" as valid values for this configuration option to give more choice to a client about which promisor remotes it might accept among those that the server advertised. In case of "KnownName", the client will accept promisor remotes which are already configured on the client and have the same name as those advertised by the client. This could be useful in a corporate setup where servers and clients are trusted to not switch names and URLs, but where some kind of control is still useful. In case of "KnownUrl", the client will accept promisor remotes which have both the same name and the same URL configured on the client as the name and URL advertised by the server. This is the most secure option, so it should be used if possible. Signed-off-by: Christian Couder Signed-off-by: Junio C Hamano --- Documentation/config/promisor.adoc | 22 ++++++--- promisor-remote.c | 60 ++++++++++++++++++++--- t/t5710-promisor-remote-capability.sh | 68 +++++++++++++++++++++++++++ 3 files changed, 138 insertions(+), 12 deletions(-) diff --git a/Documentation/config/promisor.adoc b/Documentation/config/promisor.adoc index 9cbfe3e59ea834..9192acfd243097 100644 --- a/Documentation/config/promisor.adoc +++ b/Documentation/config/promisor.adoc @@ -12,9 +12,19 @@ promisor.advertise:: promisor.acceptFromServer:: If set to "all", a client will accept all the promisor remotes a server might advertise using the "promisor-remote" - capability. Default is "none", which means no promisor remote - advertised by a server will be accepted. By accepting a - promisor remote, the client agrees that the server might omit - objects that are lazily fetchable from this promisor remote - from its responses to "fetch" and "clone" requests from the - client. See linkgit:gitprotocol-v2[5]. + capability. If set to "knownName" the client will accept + promisor remotes which are already configured on the client + and have the same name as those advertised by the client. This + is not very secure, but could be used in a corporate setup + where servers and clients are trusted to not switch name and + URLs. If set to "knownUrl", the client will accept promisor + remotes which have both the same name and the same URL + configured on the client as the name and URL advertised by the + server. This is more secure than "all" or "knownName", so it + should be used if possible instead of those options. Default + is "none", which means no promisor remote advertised by a + server will be accepted. By accepting a promisor remote, the + client agrees that the server might omit objects that are + lazily fetchable from this promisor remote from its responses + to "fetch" and "clone" requests from the client. See + linkgit:gitprotocol-v2[5]. diff --git a/promisor-remote.c b/promisor-remote.c index 918be6528fe1fc..6a0a61382f40c0 100644 --- a/promisor-remote.c +++ b/promisor-remote.c @@ -368,30 +368,73 @@ char *promisor_remote_info(struct repository *repo) return strbuf_detach(&sb, NULL); } +/* + * Find first index of 'nicks' where there is 'nick'. 'nick' is + * compared case insensitively to the strings in 'nicks'. If not found + * 'nicks->nr' is returned. + */ +static size_t remote_nick_find(struct strvec *nicks, const char *nick) +{ + for (size_t i = 0; i < nicks->nr; i++) + if (!strcasecmp(nicks->v[i], nick)) + return i; + return nicks->nr; +} + enum accept_promisor { ACCEPT_NONE = 0, + ACCEPT_KNOWN_URL, + ACCEPT_KNOWN_NAME, ACCEPT_ALL }; static int should_accept_remote(enum accept_promisor accept, - const char *remote_name UNUSED, - const char *remote_url UNUSED) + const char *remote_name, const char *remote_url, + struct strvec *names, struct strvec *urls) { + size_t i; + if (accept == ACCEPT_ALL) return 1; - BUG("Unhandled 'enum accept_promisor' value '%d'", accept); + i = remote_nick_find(names, remote_name); + + if (i >= names->nr) + /* We don't know about that remote */ + return 0; + + if (accept == ACCEPT_KNOWN_NAME) + return 1; + + if (accept != ACCEPT_KNOWN_URL) + BUG("Unhandled 'enum accept_promisor' value '%d'", accept); + + if (!strcmp(urls->v[i], remote_url)) + return 1; + + warning(_("known remote named '%s' but with url '%s' instead of '%s'"), + remote_name, urls->v[i], remote_url); + + return 0; } -static void filter_promisor_remote(struct strvec *accepted, const char *info) +static void filter_promisor_remote(struct repository *repo, + struct strvec *accepted, + const char *info) { struct strbuf **remotes; const char *accept_str; enum accept_promisor accept = ACCEPT_NONE; + struct strvec names = STRVEC_INIT; + struct strvec urls = STRVEC_INIT; if (!git_config_get_string_tmp("promisor.acceptfromserver", &accept_str)) { if (!*accept_str || !strcasecmp("None", accept_str)) accept = ACCEPT_NONE; + else if (!strcasecmp("KnownUrl", accept_str)) + accept = ACCEPT_KNOWN_URL; + else if (!strcasecmp("KnownName", accept_str)) + accept = ACCEPT_KNOWN_NAME; else if (!strcasecmp("All", accept_str)) accept = ACCEPT_ALL; else @@ -402,6 +445,9 @@ static void filter_promisor_remote(struct strvec *accepted, const char *info) if (accept == ACCEPT_NONE) return; + if (accept != ACCEPT_ALL) + promisor_info_vecs(repo, &names, &urls); + /* Parse remote info received */ remotes = strbuf_split_str(info, ';', 0); @@ -431,7 +477,7 @@ static void filter_promisor_remote(struct strvec *accepted, const char *info) if (remote_url) decoded_url = url_percent_decode(remote_url); - if (decoded_name && should_accept_remote(accept, decoded_name, decoded_url)) + if (decoded_name && should_accept_remote(accept, decoded_name, decoded_url, &names, &urls)) strvec_push(accepted, decoded_name); strbuf_list_free(elems); @@ -439,6 +485,8 @@ static void filter_promisor_remote(struct strvec *accepted, const char *info) free(decoded_url); } + strvec_clear(&names); + strvec_clear(&urls); strbuf_list_free(remotes); } @@ -447,7 +495,7 @@ char *promisor_remote_reply(const char *info) struct strvec accepted = STRVEC_INIT; struct strbuf reply = STRBUF_INIT; - filter_promisor_remote(&accepted, info); + filter_promisor_remote(the_repository, &accepted, info); if (!accepted.nr) return NULL; diff --git a/t/t5710-promisor-remote-capability.sh b/t/t5710-promisor-remote-capability.sh index 51cf2269e1d370..d2cc69a17e4668 100755 --- a/t/t5710-promisor-remote-capability.sh +++ b/t/t5710-promisor-remote-capability.sh @@ -160,6 +160,74 @@ test_expect_success "init + fetch with promisor.advertise set to 'true'" ' check_missing_objects server 1 "$oid" ' +test_expect_success "clone with promisor.acceptfromserver set to 'KnownName'" ' + git -C server config promisor.advertise true && + + # Clone from server to create a client + GIT_NO_LAZY_FETCH=0 git clone -c remote.lop.promisor=true \ + -c remote.lop.fetch="+refs/heads/*:refs/remotes/lop/*" \ + -c remote.lop.url="file://$(pwd)/lop" \ + -c promisor.acceptfromserver=KnownName \ + --no-local --filter="blob:limit=5k" server client && + test_when_finished "rm -rf client" && + + # Check that the largest object is still missing on the server + check_missing_objects server 1 "$oid" +' + +test_expect_success "clone with 'KnownName' and different remote names" ' + git -C server config promisor.advertise true && + + # Clone from server to create a client + GIT_NO_LAZY_FETCH=0 git clone -c remote.serverTwo.promisor=true \ + -c remote.serverTwo.fetch="+refs/heads/*:refs/remotes/lop/*" \ + -c remote.serverTwo.url="file://$(pwd)/lop" \ + -c promisor.acceptfromserver=KnownName \ + --no-local --filter="blob:limit=5k" server client && + test_when_finished "rm -rf client" && + + # Check that the largest object is not missing on the server + check_missing_objects server 0 "" && + + # Reinitialize server so that the largest object is missing again + initialize_server 1 "$oid" +' + +test_expect_success "clone with promisor.acceptfromserver set to 'KnownUrl'" ' + git -C server config promisor.advertise true && + + # Clone from server to create a client + GIT_NO_LAZY_FETCH=0 git clone -c remote.lop.promisor=true \ + -c remote.lop.fetch="+refs/heads/*:refs/remotes/lop/*" \ + -c remote.lop.url="file://$(pwd)/lop" \ + -c promisor.acceptfromserver=KnownUrl \ + --no-local --filter="blob:limit=5k" server client && + test_when_finished "rm -rf client" && + + # Check that the largest object is still missing on the server + check_missing_objects server 1 "$oid" +' + +test_expect_success "clone with 'KnownUrl' and different remote urls" ' + ln -s lop serverTwo && + + git -C server config promisor.advertise true && + + # Clone from server to create a client + GIT_NO_LAZY_FETCH=0 git clone -c remote.lop.promisor=true \ + -c remote.lop.fetch="+refs/heads/*:refs/remotes/lop/*" \ + -c remote.lop.url="file://$(pwd)/serverTwo" \ + -c promisor.acceptfromserver=KnownUrl \ + --no-local --filter="blob:limit=5k" server client && + test_when_finished "rm -rf client" && + + # Check that the largest object is not missing on the server + check_missing_objects server 0 "" && + + # Reinitialize server so that the largest object is missing again + initialize_server 1 "$oid" +' + test_expect_success "clone with promisor.advertise set to 'true' but don't delete the client" ' git -C server config promisor.advertise true && From a16a2ee312b088ae6ac2ab3302ed191d19a71718 Mon Sep 17 00:00:00 2001 From: Seyi Kuforiji Date: Tue, 25 Feb 2025 11:10:41 +0100 Subject: [PATCH 013/312] t/unit-tests: implement clar specific oid helper functions `get_oid_arbitrary_hex()` and `init_hash_algo()` are both required for oid-related tests to run without errors. In the current implementation, both functions are defined and declared in the `t/unit-tests/lib-oid.{c,h}` which is utilized by oid-related tests in the homegrown unit tests structure. Adapt functions in lib-oid.{c,h} to use clar. Both these functions become available for oid-related test files implemented using the clar testing framework, which requires them. This will be used by subsequent commits. Mentored-by: Patrick Steinhardt Signed-off-by: Seyi Kuforiji Signed-off-by: Junio C Hamano --- Makefile | 2 +- t/meson.build | 2 +- t/unit-tests/lib-oid.c | 32 +++++++++++--------------------- t/unit-tests/lib-oid.h | 9 ++++++--- t/unit-tests/unit-test.c | 2 ++ 5 files changed, 21 insertions(+), 26 deletions(-) diff --git a/Makefile b/Makefile index bcf5ed3f85fafc..81799488f0a338 100644 --- a/Makefile +++ b/Makefile @@ -1365,6 +1365,7 @@ CLAR_TEST_PROG = $(UNIT_TEST_BIN)/unit-tests$(X) CLAR_TEST_OBJS = $(patsubst %,$(UNIT_TEST_DIR)/%.o,$(CLAR_TEST_SUITES)) CLAR_TEST_OBJS += $(UNIT_TEST_DIR)/clar/clar.o CLAR_TEST_OBJS += $(UNIT_TEST_DIR)/unit-test.o +CLAR_TEST_OBJS += $(UNIT_TEST_DIR)/lib-oid.o UNIT_TEST_PROGRAMS += t-oid-array UNIT_TEST_PROGRAMS += t-oidmap @@ -1381,7 +1382,6 @@ UNIT_TEST_PROGRAMS += t-trailer UNIT_TEST_PROGRAMS += t-urlmatch-normalization UNIT_TEST_PROGS = $(patsubst %,$(UNIT_TEST_BIN)/%$X,$(UNIT_TEST_PROGRAMS)) UNIT_TEST_OBJS += $(UNIT_TEST_DIR)/test-lib.o -UNIT_TEST_OBJS += $(UNIT_TEST_DIR)/lib-oid.o UNIT_TEST_OBJS += $(UNIT_TEST_DIR)/lib-reftable.o # xdiff and reftable libs may in turn depend on what is in libgit.a diff --git a/t/meson.build b/t/meson.build index 780939d49fda56..862cf1cfd4060d 100644 --- a/t/meson.build +++ b/t/meson.build @@ -14,6 +14,7 @@ clar_test_suites = [ clar_sources = [ 'unit-tests/clar/clar.c', 'unit-tests/unit-test.c', + 'unit-tests/lib-oid.c' ] clar_decls_h = custom_target( @@ -68,7 +69,6 @@ foreach unit_test_program : unit_test_programs unit_test = executable(unit_test_name, sources: [ 'unit-tests/test-lib.c', - 'unit-tests/lib-oid.c', 'unit-tests/lib-reftable.c', unit_test_program, ], diff --git a/t/unit-tests/lib-oid.c b/t/unit-tests/lib-oid.c index 8f0ccac5328247..e0b3180f23c018 100644 --- a/t/unit-tests/lib-oid.c +++ b/t/unit-tests/lib-oid.c @@ -1,9 +1,9 @@ -#include "test-lib.h" +#include "unit-test.h" #include "lib-oid.h" #include "strbuf.h" #include "hex.h" -int init_hash_algo(void) +int cl_setup_hash_algo(void) { static int algo = -1; @@ -11,42 +11,32 @@ int init_hash_algo(void) const char *algo_name = getenv("GIT_TEST_DEFAULT_HASH"); algo = algo_name ? hash_algo_by_name(algo_name) : GIT_HASH_SHA1; - if (!check(algo != GIT_HASH_UNKNOWN)) - test_msg("BUG: invalid GIT_TEST_DEFAULT_HASH value ('%s')", - algo_name); + cl_assert(algo != GIT_HASH_UNKNOWN); } return algo; } -static int get_oid_arbitrary_hex_algop(const char *hex, struct object_id *oid, +static void cl_parse_oid(const char *hex, struct object_id *oid, const struct git_hash_algo *algop) { - int ret; size_t sz = strlen(hex); struct strbuf buf = STRBUF_INIT; - if (!check(sz <= algop->hexsz)) { - test_msg("BUG: hex string (%s) bigger than maximum allowed (%lu)", - hex, (unsigned long)algop->hexsz); - return -1; - } + cl_assert(sz <= algop->hexsz); strbuf_add(&buf, hex, sz); strbuf_addchars(&buf, '0', algop->hexsz - sz); - ret = get_oid_hex_algop(buf.buf, oid, algop); - if (!check_int(ret, ==, 0)) - test_msg("BUG: invalid hex input (%s) provided", hex); + cl_assert_equal_i(get_oid_hex_algop(buf.buf, oid, algop), 0); strbuf_release(&buf); - return ret; } -int get_oid_arbitrary_hex(const char *hex, struct object_id *oid) + +void cl_parse_any_oid(const char *hex, struct object_id *oid) { - int hash_algo = init_hash_algo(); + int hash_algo = cl_setup_hash_algo(); - if (!check_int(hash_algo, !=, GIT_HASH_UNKNOWN)) - return -1; - return get_oid_arbitrary_hex_algop(hex, oid, &hash_algos[hash_algo]); + cl_assert(hash_algo != GIT_HASH_UNKNOWN); + cl_parse_oid(hex, oid, &hash_algos[hash_algo]); } diff --git a/t/unit-tests/lib-oid.h b/t/unit-tests/lib-oid.h index 4e77c04bd224e6..4031775104906d 100644 --- a/t/unit-tests/lib-oid.h +++ b/t/unit-tests/lib-oid.h @@ -5,6 +5,7 @@ /* * Convert arbitrary hex string to object_id. + * * For example, passing "abc12" will generate * "abc1200000000000000000000000000000000000" hex of length 40 for SHA-1 and * create object_id with that. @@ -12,14 +13,16 @@ * algo is not allowed. The hash algo is decided based on GIT_TEST_DEFAULT_HASH * environment variable. */ -int get_oid_arbitrary_hex(const char *s, struct object_id *oid); + +void cl_parse_any_oid (const char *s, struct object_id *oid); /* * Returns one of GIT_HASH_{SHA1, SHA256, UNKNOWN} based on the value of * GIT_TEST_DEFAULT_HASH environment variable. The fallback value in the * absence of GIT_TEST_DEFAULT_HASH is GIT_HASH_SHA1. It also uses - * check(algo != GIT_HASH_UNKNOWN) before returning to verify if the + * cl_assert(algo != GIT_HASH_UNKNOWN) before returning to verify if the * GIT_TEST_DEFAULT_HASH's value is valid or not. */ -int init_hash_algo(void); + +int cl_setup_hash_algo(void); #endif /* LIB_OID_H */ diff --git a/t/unit-tests/unit-test.c b/t/unit-tests/unit-test.c index fa8818842a4247..5af645048adf4e 100644 --- a/t/unit-tests/unit-test.c +++ b/t/unit-tests/unit-test.c @@ -1,5 +1,7 @@ #include "unit-test.h" +#include "hex.h" #include "parse-options.h" +#include "strbuf.h" #include "string-list.h" #include "strvec.h" From 869a1edf4470fab379fdead671036d6906112da2 Mon Sep 17 00:00:00 2001 From: Seyi Kuforiji Date: Tue, 25 Feb 2025 11:10:42 +0100 Subject: [PATCH 014/312] t/unit-tests: convert oid-array test to use clar test framework Adapt oid-array test script to clar framework by using clar assertions where necessary. Remove descriptions from macros to reduce redundancy, and move test input arrays to global scope for reuse across multiple test functions. Introduce `test_oid_array__initialize()` to explicitly initialize the hash algorithm. These changes streamline the test suite, making individual tests self-contained and reducing redundant code. Mentored-by: Patrick Steinhardt Signed-off-by: Seyi Kuforiji Signed-off-by: Junio C Hamano --- Makefile | 2 +- t/meson.build | 2 +- t/unit-tests/t-oid-array.c | 126 ------------------------------------ t/unit-tests/u-oid-array.c | 129 +++++++++++++++++++++++++++++++++++++ 4 files changed, 131 insertions(+), 128 deletions(-) delete mode 100644 t/unit-tests/t-oid-array.c create mode 100644 t/unit-tests/u-oid-array.c diff --git a/Makefile b/Makefile index 81799488f0a338..c9b0320fcb203d 100644 --- a/Makefile +++ b/Makefile @@ -1356,6 +1356,7 @@ CLAR_TEST_SUITES += u-example-decorate CLAR_TEST_SUITES += u-hash CLAR_TEST_SUITES += u-hashmap CLAR_TEST_SUITES += u-mem-pool +CLAR_TEST_SUITES += u-oid-array CLAR_TEST_SUITES += u-prio-queue CLAR_TEST_SUITES += u-reftable-tree CLAR_TEST_SUITES += u-strbuf @@ -1367,7 +1368,6 @@ CLAR_TEST_OBJS += $(UNIT_TEST_DIR)/clar/clar.o CLAR_TEST_OBJS += $(UNIT_TEST_DIR)/unit-test.o CLAR_TEST_OBJS += $(UNIT_TEST_DIR)/lib-oid.o -UNIT_TEST_PROGRAMS += t-oid-array UNIT_TEST_PROGRAMS += t-oidmap UNIT_TEST_PROGRAMS += t-oidtree UNIT_TEST_PROGRAMS += t-reftable-basics diff --git a/t/meson.build b/t/meson.build index 862cf1cfd4060d..1b17a53acccd60 100644 --- a/t/meson.build +++ b/t/meson.build @@ -4,6 +4,7 @@ clar_test_suites = [ 'unit-tests/u-hash.c', 'unit-tests/u-hashmap.c', 'unit-tests/u-mem-pool.c', + 'unit-tests/u-oid-array.c', 'unit-tests/u-prio-queue.c', 'unit-tests/u-reftable-tree.c', 'unit-tests/u-strbuf.c', @@ -49,7 +50,6 @@ clar_unit_tests = executable('unit-tests', test('unit-tests', clar_unit_tests) unit_test_programs = [ - 'unit-tests/t-oid-array.c', 'unit-tests/t-oidmap.c', 'unit-tests/t-oidtree.c', 'unit-tests/t-reftable-basics.c', diff --git a/t/unit-tests/t-oid-array.c b/t/unit-tests/t-oid-array.c deleted file mode 100644 index 45b59a2a51bc52..00000000000000 --- a/t/unit-tests/t-oid-array.c +++ /dev/null @@ -1,126 +0,0 @@ -#define USE_THE_REPOSITORY_VARIABLE - -#include "test-lib.h" -#include "lib-oid.h" -#include "oid-array.h" -#include "hex.h" - -static int fill_array(struct oid_array *array, const char *hexes[], size_t n) -{ - for (size_t i = 0; i < n; i++) { - struct object_id oid; - - if (!check_int(get_oid_arbitrary_hex(hexes[i], &oid), ==, 0)) - return -1; - oid_array_append(array, &oid); - } - if (!check_uint(array->nr, ==, n)) - return -1; - return 0; -} - -static int add_to_oid_array(const struct object_id *oid, void *data) -{ - struct oid_array *array = data; - - oid_array_append(array, oid); - return 0; -} - -static void t_enumeration(const char **input_args, size_t input_sz, - const char **expect_args, size_t expect_sz) -{ - struct oid_array input = OID_ARRAY_INIT, expect = OID_ARRAY_INIT, - actual = OID_ARRAY_INIT; - size_t i; - - if (fill_array(&input, input_args, input_sz)) - return; - if (fill_array(&expect, expect_args, expect_sz)) - return; - - oid_array_for_each_unique(&input, add_to_oid_array, &actual); - if (!check_uint(actual.nr, ==, expect.nr)) - return; - - for (i = 0; i < actual.nr; i++) { - if (!check(oideq(&actual.oid[i], &expect.oid[i]))) - test_msg("expected: %s\n got: %s\n index: %" PRIuMAX, - oid_to_hex(&expect.oid[i]), oid_to_hex(&actual.oid[i]), - (uintmax_t)i); - } - - oid_array_clear(&actual); - oid_array_clear(&input); - oid_array_clear(&expect); -} - -#define TEST_ENUMERATION(input, expect, desc) \ - TEST(t_enumeration(input, ARRAY_SIZE(input), expect, ARRAY_SIZE(expect)), \ - desc " works") - -static void t_lookup(const char **input_hexes, size_t n, const char *query_hex, - int lower_bound, int upper_bound) -{ - struct oid_array array = OID_ARRAY_INIT; - struct object_id oid_query; - int ret; - - if (!check_int(get_oid_arbitrary_hex(query_hex, &oid_query), ==, 0)) - return; - if (fill_array(&array, input_hexes, n)) - return; - ret = oid_array_lookup(&array, &oid_query); - - if (!check_int(ret, <=, upper_bound) || - !check_int(ret, >=, lower_bound)) - test_msg("oid query for lookup: %s", oid_to_hex(&oid_query)); - - oid_array_clear(&array); -} - -#define TEST_LOOKUP(input_hexes, query, lower_bound, upper_bound, desc) \ - TEST(t_lookup(input_hexes, ARRAY_SIZE(input_hexes), query, \ - lower_bound, upper_bound), \ - desc " works") - -static void setup(void) -{ - /* The hash algo is used by oid_array_lookup() internally */ - int algo = init_hash_algo(); - if (check_int(algo, !=, GIT_HASH_UNKNOWN)) - repo_set_hash_algo(the_repository, algo); -} - -int cmd_main(int argc UNUSED, const char **argv UNUSED) -{ - const char *arr_input[] = { "88", "44", "aa", "55" }; - const char *arr_input_dup[] = { "88", "44", "aa", "55", - "88", "44", "aa", "55", - "88", "44", "aa", "55" }; - const char *res_sorted[] = { "44", "55", "88", "aa" }; - const char *nearly_55; - - if (!TEST(setup(), "setup")) - test_skip_all("hash algo initialization failed"); - - TEST_ENUMERATION(arr_input, res_sorted, "ordered enumeration"); - TEST_ENUMERATION(arr_input_dup, res_sorted, - "ordered enumeration with duplicate suppression"); - - TEST_LOOKUP(arr_input, "55", 1, 1, "lookup"); - TEST_LOOKUP(arr_input, "33", INT_MIN, -1, "lookup non-existent entry"); - TEST_LOOKUP(arr_input_dup, "55", 3, 5, "lookup with duplicates"); - TEST_LOOKUP(arr_input_dup, "66", INT_MIN, -1, - "lookup non-existent entry with duplicates"); - - nearly_55 = init_hash_algo() == GIT_HASH_SHA1 ? - "5500000000000000000000000000000000000001" : - "5500000000000000000000000000000000000000000000000000000000000001"; - TEST_LOOKUP(((const char *[]){ "55", nearly_55 }), "55", 0, 0, - "lookup with almost duplicate values"); - TEST_LOOKUP(((const char *[]){ "55", "55" }), "55", 0, 1, - "lookup with single duplicate value"); - - return test_done(); -} diff --git a/t/unit-tests/u-oid-array.c b/t/unit-tests/u-oid-array.c new file mode 100644 index 00000000000000..e48a433f2153b1 --- /dev/null +++ b/t/unit-tests/u-oid-array.c @@ -0,0 +1,129 @@ +#define USE_THE_REPOSITORY_VARIABLE + +#include "unit-test.h" +#include "lib-oid.h" +#include "oid-array.h" +#include "hex.h" + +static void fill_array(struct oid_array *array, const char *hexes[], size_t n) +{ + for (size_t i = 0; i < n; i++) { + struct object_id oid; + + cl_parse_any_oid(hexes[i], &oid); + oid_array_append(array, &oid); + } + cl_assert_equal_i(array->nr, n); +} + +static int add_to_oid_array(const struct object_id *oid, void *data) +{ + struct oid_array *array = data; + + oid_array_append(array, oid); + return 0; +} + +static void t_enumeration(const char **input_args, size_t input_sz, + const char **expect_args, size_t expect_sz) +{ + struct oid_array input = OID_ARRAY_INIT, expect = OID_ARRAY_INIT, + actual = OID_ARRAY_INIT; + size_t i; + + fill_array(&input, input_args, input_sz); + fill_array(&expect, expect_args, expect_sz); + + oid_array_for_each_unique(&input, add_to_oid_array, &actual); + cl_assert_equal_i(actual.nr, expect.nr); + + for (i = 0; i < actual.nr; i++) + cl_assert(oideq(&actual.oid[i], &expect.oid[i])); + + oid_array_clear(&actual); + oid_array_clear(&input); + oid_array_clear(&expect); +} + +#define TEST_ENUMERATION(input, expect) \ + t_enumeration(input, ARRAY_SIZE(input), expect, ARRAY_SIZE(expect)); + +static void t_lookup(const char **input_hexes, size_t n, const char *query_hex, + int lower_bound, int upper_bound) +{ + struct oid_array array = OID_ARRAY_INIT; + struct object_id oid_query; + int ret; + + cl_parse_any_oid(query_hex, &oid_query); + fill_array(&array, input_hexes, n); + ret = oid_array_lookup(&array, &oid_query); + + cl_assert(ret <= upper_bound); + cl_assert(ret >= lower_bound); + + oid_array_clear(&array); +} + +#define TEST_LOOKUP(input_hexes, query, lower_bound, upper_bound) \ + t_lookup(input_hexes, ARRAY_SIZE(input_hexes), query, \ + lower_bound, upper_bound); + +void test_oid_array__initialize(void) +{ + /* The hash algo is used by oid_array_lookup() internally */ + int algo = cl_setup_hash_algo(); + repo_set_hash_algo(the_repository, algo); +} + +static const char *arr_input[] = { "88", "44", "aa", "55" }; +static const char *arr_input_dup[] = { "88", "44", "aa", "55", + "88", "44", "aa", "55", + "88", "44", "aa", "55" }; +static const char *res_sorted[] = { "44", "55", "88", "aa" }; + +void test_oid_array__enumerate_unique(void) +{ + TEST_ENUMERATION(arr_input, res_sorted); +} + +void test_oid_array__enumerate_duplicate(void) +{ + TEST_ENUMERATION(arr_input_dup, res_sorted); +} + +void test_oid_array__lookup(void) +{ + TEST_LOOKUP(arr_input, "55", 1, 1); +} + +void test_oid_array__lookup_non_existent(void) +{ + TEST_LOOKUP(arr_input, "33", INT_MIN, -1); +} + +void test_oid_array__lookup_duplicates(void) +{ + TEST_LOOKUP(arr_input_dup, "55", 3, 5); +} + +void test_oid_array__lookup_non_existent_dup(void) +{ + TEST_LOOKUP(arr_input_dup, "66", INT_MIN, -1); +} + +void test_oid_array__lookup_almost_dup(void) +{ + const char *nearly_55; + + nearly_55 = cl_setup_hash_algo() == GIT_HASH_SHA1 ? + "5500000000000000000000000000000000000001" : + "5500000000000000000000000000000000000000000000000000000000000001"; + + TEST_LOOKUP(((const char *[]){ "55", nearly_55 }), "55", 0, 0); +} + +void test_oid_array__lookup_single_dup(void) +{ + TEST_LOOKUP(((const char *[]){ "55", "55" }), "55", 0, 1); +} From 69bc044def16e8d8f85cd321b00ae53abe96a567 Mon Sep 17 00:00:00 2001 From: Seyi Kuforiji Date: Tue, 25 Feb 2025 11:10:43 +0100 Subject: [PATCH 015/312] t/unit-tests: convert oidmap test to use clar test framework Adapt oidmap test script to clar framework by using clar assertions where necessary. `cl_parse_any_oid()` ensures the hash algorithm is set before parsing. This prevents issues from an uninitialized or invalid hash algorithm. Introduce 'test_oidmap__initialize` handles the to set up of the global oidmap map with predefined key-value pairs, and `test_oidmap__cleanup` frees the oidmap and its entries when all tests are completed. The test loops through all entries to detect multiple errors. With this change, it stops at the first error encountered, making it easier to address it. Mentored-by: Patrick Steinhardt Signed-off-by: Seyi Kuforiji Signed-off-by: Junio C Hamano --- Makefile | 2 +- t/meson.build | 2 +- t/unit-tests/t-oidmap.c | 181 ---------------------------------------- t/unit-tests/u-oidmap.c | 136 ++++++++++++++++++++++++++++++ 4 files changed, 138 insertions(+), 183 deletions(-) delete mode 100644 t/unit-tests/t-oidmap.c create mode 100644 t/unit-tests/u-oidmap.c diff --git a/Makefile b/Makefile index c9b0320fcb203d..e4e85e6007d2ac 100644 --- a/Makefile +++ b/Makefile @@ -1357,6 +1357,7 @@ CLAR_TEST_SUITES += u-hash CLAR_TEST_SUITES += u-hashmap CLAR_TEST_SUITES += u-mem-pool CLAR_TEST_SUITES += u-oid-array +CLAR_TEST_SUITES += u-oidmap CLAR_TEST_SUITES += u-prio-queue CLAR_TEST_SUITES += u-reftable-tree CLAR_TEST_SUITES += u-strbuf @@ -1368,7 +1369,6 @@ CLAR_TEST_OBJS += $(UNIT_TEST_DIR)/clar/clar.o CLAR_TEST_OBJS += $(UNIT_TEST_DIR)/unit-test.o CLAR_TEST_OBJS += $(UNIT_TEST_DIR)/lib-oid.o -UNIT_TEST_PROGRAMS += t-oidmap UNIT_TEST_PROGRAMS += t-oidtree UNIT_TEST_PROGRAMS += t-reftable-basics UNIT_TEST_PROGRAMS += t-reftable-block diff --git a/t/meson.build b/t/meson.build index 1b17a53acccd60..d5b83cdb72bc07 100644 --- a/t/meson.build +++ b/t/meson.build @@ -5,6 +5,7 @@ clar_test_suites = [ 'unit-tests/u-hashmap.c', 'unit-tests/u-mem-pool.c', 'unit-tests/u-oid-array.c', + 'unit-tests/u-oidmap.c', 'unit-tests/u-prio-queue.c', 'unit-tests/u-reftable-tree.c', 'unit-tests/u-strbuf.c', @@ -50,7 +51,6 @@ clar_unit_tests = executable('unit-tests', test('unit-tests', clar_unit_tests) unit_test_programs = [ - 'unit-tests/t-oidmap.c', 'unit-tests/t-oidtree.c', 'unit-tests/t-reftable-basics.c', 'unit-tests/t-reftable-block.c', diff --git a/t/unit-tests/t-oidmap.c b/t/unit-tests/t-oidmap.c deleted file mode 100644 index b22e52d08bdf77..00000000000000 --- a/t/unit-tests/t-oidmap.c +++ /dev/null @@ -1,181 +0,0 @@ -#include "test-lib.h" -#include "lib-oid.h" -#include "oidmap.h" -#include "hash.h" -#include "hex.h" - -/* - * Elements we will put in oidmap structs are made of a key: the entry.oid - * field, which is of type struct object_id, and a value: the name field (could - * be a refname for example). - */ -struct test_entry { - struct oidmap_entry entry; - char name[FLEX_ARRAY]; -}; - -static const char *const key_val[][2] = { { "11", "one" }, - { "22", "two" }, - { "33", "three" } }; - -static void setup(void (*f)(struct oidmap *map)) -{ - struct oidmap map = OIDMAP_INIT; - int ret = 0; - - for (size_t i = 0; i < ARRAY_SIZE(key_val); i++){ - struct test_entry *entry; - - FLEX_ALLOC_STR(entry, name, key_val[i][1]); - if ((ret = get_oid_arbitrary_hex(key_val[i][0], &entry->entry.oid))) { - free(entry); - break; - } - entry = oidmap_put(&map, entry); - if (!check(entry == NULL)) - free(entry); - } - - if (!ret) - f(&map); - oidmap_free(&map, 1); -} - -static void t_replace(struct oidmap *map) -{ - struct test_entry *entry, *prev; - - FLEX_ALLOC_STR(entry, name, "un"); - if (get_oid_arbitrary_hex("11", &entry->entry.oid)) - return; - prev = oidmap_put(map, entry); - if (!check(prev != NULL)) - return; - check_str(prev->name, "one"); - free(prev); - - FLEX_ALLOC_STR(entry, name, "deux"); - if (get_oid_arbitrary_hex("22", &entry->entry.oid)) - return; - prev = oidmap_put(map, entry); - if (!check(prev != NULL)) - return; - check_str(prev->name, "two"); - free(prev); -} - -static void t_get(struct oidmap *map) -{ - struct test_entry *entry; - struct object_id oid; - - if (get_oid_arbitrary_hex("22", &oid)) - return; - entry = oidmap_get(map, &oid); - if (!check(entry != NULL)) - return; - check_str(entry->name, "two"); - - if (get_oid_arbitrary_hex("44", &oid)) - return; - check(oidmap_get(map, &oid) == NULL); - - if (get_oid_arbitrary_hex("11", &oid)) - return; - entry = oidmap_get(map, &oid); - if (!check(entry != NULL)) - return; - check_str(entry->name, "one"); -} - -static void t_remove(struct oidmap *map) -{ - struct test_entry *entry; - struct object_id oid; - - if (get_oid_arbitrary_hex("11", &oid)) - return; - entry = oidmap_remove(map, &oid); - if (!check(entry != NULL)) - return; - check_str(entry->name, "one"); - check(oidmap_get(map, &oid) == NULL); - free(entry); - - if (get_oid_arbitrary_hex("22", &oid)) - return; - entry = oidmap_remove(map, &oid); - if (!check(entry != NULL)) - return; - check_str(entry->name, "two"); - check(oidmap_get(map, &oid) == NULL); - free(entry); - - if (get_oid_arbitrary_hex("44", &oid)) - return; - check(oidmap_remove(map, &oid) == NULL); -} - -static int key_val_contains(struct test_entry *entry, char seen[]) -{ - for (size_t i = 0; i < ARRAY_SIZE(key_val); i++) { - struct object_id oid; - - if (get_oid_arbitrary_hex(key_val[i][0], &oid)) - return -1; - - if (oideq(&entry->entry.oid, &oid)) { - if (seen[i]) - return 2; - seen[i] = 1; - return 0; - } - } - return 1; -} - -static void t_iterate(struct oidmap *map) -{ - struct oidmap_iter iter; - struct test_entry *entry; - char seen[ARRAY_SIZE(key_val)] = { 0 }; - int count = 0; - - oidmap_iter_init(map, &iter); - while ((entry = oidmap_iter_next(&iter))) { - int ret; - if (!check_int((ret = key_val_contains(entry, seen)), ==, 0)) { - switch (ret) { - case -1: - break; /* error message handled by get_oid_arbitrary_hex() */ - case 1: - test_msg("obtained entry was not given in the input\n" - " name: %s\n oid: %s\n", - entry->name, oid_to_hex(&entry->entry.oid)); - break; - case 2: - test_msg("duplicate entry detected\n" - " name: %s\n oid: %s\n", - entry->name, oid_to_hex(&entry->entry.oid)); - break; - default: - test_msg("BUG: invalid return value (%d) from key_val_contains()", - ret); - break; - } - } else { - count++; - } - } - check_int(count, ==, ARRAY_SIZE(key_val)); - check_int(hashmap_get_size(&map->map), ==, ARRAY_SIZE(key_val)); -} - -int cmd_main(int argc UNUSED, const char **argv UNUSED) -{ - TEST(setup(t_replace), "replace works"); - TEST(setup(t_get), "get works"); - TEST(setup(t_remove), "remove works"); - TEST(setup(t_iterate), "iterate works"); - return test_done(); -} diff --git a/t/unit-tests/u-oidmap.c b/t/unit-tests/u-oidmap.c new file mode 100644 index 00000000000000..dc805b7e3cb424 --- /dev/null +++ b/t/unit-tests/u-oidmap.c @@ -0,0 +1,136 @@ +#include "unit-test.h" +#include "lib-oid.h" +#include "oidmap.h" +#include "hash.h" +#include "hex.h" + +/* + * Elements we will put in oidmap structs are made of a key: the entry.oid + * field, which is of type struct object_id, and a value: the name field (could + * be a refname for example). + */ +struct test_entry { + struct oidmap_entry entry; + char name[FLEX_ARRAY]; +}; + +static const char *const key_val[][2] = { { "11", "one" }, + { "22", "two" }, + { "33", "three" } }; + +static struct oidmap map; + +void test_oidmap__initialize(void) +{ + oidmap_init(&map, 0); + + for (size_t i = 0; i < ARRAY_SIZE(key_val); i++){ + struct test_entry *entry; + + FLEX_ALLOC_STR(entry, name, key_val[i][1]); + cl_parse_any_oid(key_val[i][0], &entry->entry.oid); + cl_assert(oidmap_put(&map, entry) == NULL); + } +} + +void test_oidmap__cleanup(void) +{ + oidmap_free(&map, 1); +} + +void test_oidmap__replace(void) +{ + struct test_entry *entry, *prev; + + FLEX_ALLOC_STR(entry, name, "un"); + cl_parse_any_oid("11", &entry->entry.oid); + prev = oidmap_put(&map, entry); + cl_assert(prev != NULL); + cl_assert_equal_s(prev->name, "one"); + free(prev); + + FLEX_ALLOC_STR(entry, name, "deux"); + cl_parse_any_oid("22", &entry->entry.oid); + prev = oidmap_put(&map, entry); + cl_assert(prev != NULL); + cl_assert_equal_s(prev->name, "two"); + free(prev); +} + +void test_oidmap__get(void) +{ + struct test_entry *entry; + struct object_id oid; + + cl_parse_any_oid("22", &oid); + entry = oidmap_get(&map, &oid); + cl_assert(entry != NULL); + cl_assert_equal_s(entry->name, "two"); + + cl_parse_any_oid("44", &oid); + cl_assert(oidmap_get(&map, &oid) == NULL); + + cl_parse_any_oid("11", &oid); + entry = oidmap_get(&map, &oid); + cl_assert(entry != NULL); + cl_assert_equal_s(entry->name, "one"); +} + +void test_oidmap__remove(void) +{ + struct test_entry *entry; + struct object_id oid; + + cl_parse_any_oid("11", &oid); + entry = oidmap_remove(&map, &oid); + cl_assert(entry != NULL); + cl_assert_equal_s(entry->name, "one"); + cl_assert(oidmap_get(&map, &oid) == NULL); + free(entry); + + cl_parse_any_oid("22", &oid); + entry = oidmap_remove(&map, &oid); + cl_assert(entry != NULL); + cl_assert_equal_s(entry->name, "two"); + cl_assert(oidmap_get(&map, &oid) == NULL); + free(entry); + + cl_parse_any_oid("44", &oid); + cl_assert(oidmap_remove(&map, &oid) == NULL); +} + +static int key_val_contains(struct test_entry *entry, char seen[]) +{ + for (size_t i = 0; i < ARRAY_SIZE(key_val); i++) { + struct object_id oid; + + cl_parse_any_oid(key_val[i][0], &oid); + + if (oideq(&entry->entry.oid, &oid)) { + if (seen[i]) + return 2; + seen[i] = 1; + return 0; + } + } + return 1; +} + +void test_oidmap__iterate(void) +{ + struct oidmap_iter iter; + struct test_entry *entry; + char seen[ARRAY_SIZE(key_val)] = { 0 }; + int count = 0; + + oidmap_iter_init(&map, &iter); + while ((entry = oidmap_iter_next(&iter))) { + if (key_val_contains(entry, seen) != 0) { + cl_failf("Unexpected entry: name = %s, oid = %s", + entry->name, oid_to_hex(&entry->entry.oid)); + } + count++; + } + cl_assert_equal_i(count, ARRAY_SIZE(key_val)); + cl_assert_equal_i(hashmap_get_size(&map.map), ARRAY_SIZE(key_val)); +} From 149585079facfbdb5713bb8852caea2068cd4f28 Mon Sep 17 00:00:00 2001 From: Seyi Kuforiji Date: Tue, 25 Feb 2025 11:10:44 +0100 Subject: [PATCH 016/312] t/unit-tests: convert oidtree test to use clar test framework Adapt oidtree test script to clar framework by using clar assertions where necessary. `cl_parse_any_oid()` ensures the hash algorithm is set before parsing. This prevents issues from an uninitialized or invalid hash algorithm. Introduce 'test_oidtree__initialize` handles the to set up of the global oidtree variable and `test_oidtree__cleanup` frees the oidtree when all tests are completed. With this change, `check_each` stops at the first error encountered, making it easier to address it. Mentored-by: Patrick Steinhardt Signed-off-by: Seyi Kuforiji Signed-off-by: Junio C Hamano --- Makefile | 2 +- t/meson.build | 2 +- t/unit-tests/t-oidtree.c | 122 --------------------------------------- t/unit-tests/u-oidtree.c | 107 ++++++++++++++++++++++++++++++++++ 4 files changed, 109 insertions(+), 124 deletions(-) delete mode 100644 t/unit-tests/t-oidtree.c create mode 100644 t/unit-tests/u-oidtree.c diff --git a/Makefile b/Makefile index e4e85e6007d2ac..2b134efc70535b 100644 --- a/Makefile +++ b/Makefile @@ -1358,6 +1358,7 @@ CLAR_TEST_SUITES += u-hashmap CLAR_TEST_SUITES += u-mem-pool CLAR_TEST_SUITES += u-oid-array CLAR_TEST_SUITES += u-oidmap +CLAR_TEST_SUITES += u-oidtree CLAR_TEST_SUITES += u-prio-queue CLAR_TEST_SUITES += u-reftable-tree CLAR_TEST_SUITES += u-strbuf @@ -1369,7 +1370,6 @@ CLAR_TEST_OBJS += $(UNIT_TEST_DIR)/clar/clar.o CLAR_TEST_OBJS += $(UNIT_TEST_DIR)/unit-test.o CLAR_TEST_OBJS += $(UNIT_TEST_DIR)/lib-oid.o -UNIT_TEST_PROGRAMS += t-oidtree UNIT_TEST_PROGRAMS += t-reftable-basics UNIT_TEST_PROGRAMS += t-reftable-block UNIT_TEST_PROGRAMS += t-reftable-merged diff --git a/t/meson.build b/t/meson.build index d5b83cdb72bc07..91699917ff5f78 100644 --- a/t/meson.build +++ b/t/meson.build @@ -6,6 +6,7 @@ clar_test_suites = [ 'unit-tests/u-mem-pool.c', 'unit-tests/u-oid-array.c', 'unit-tests/u-oidmap.c', + 'unit-tests/u-oidtree.c', 'unit-tests/u-prio-queue.c', 'unit-tests/u-reftable-tree.c', 'unit-tests/u-strbuf.c', @@ -51,7 +52,6 @@ clar_unit_tests = executable('unit-tests', test('unit-tests', clar_unit_tests) unit_test_programs = [ - 'unit-tests/t-oidtree.c', 'unit-tests/t-reftable-basics.c', 'unit-tests/t-reftable-block.c', 'unit-tests/t-reftable-merged.c', diff --git a/t/unit-tests/t-oidtree.c b/t/unit-tests/t-oidtree.c deleted file mode 100644 index a38754b066b3ea..00000000000000 --- a/t/unit-tests/t-oidtree.c +++ /dev/null @@ -1,122 +0,0 @@ -#include "test-lib.h" -#include "lib-oid.h" -#include "oidtree.h" -#include "hash.h" -#include "hex.h" -#include "strvec.h" - -#define FILL_TREE(tree, ...) \ - do { \ - const char *hexes[] = { __VA_ARGS__ }; \ - if (fill_tree_loc(tree, hexes, ARRAY_SIZE(hexes))) \ - return; \ - } while (0) - -static int fill_tree_loc(struct oidtree *ot, const char *hexes[], size_t n) -{ - for (size_t i = 0; i < n; i++) { - struct object_id oid; - if (!check_int(get_oid_arbitrary_hex(hexes[i], &oid), ==, 0)) - return -1; - oidtree_insert(ot, &oid); - } - return 0; -} - -static void check_contains(struct oidtree *ot, const char *hex, int expected) -{ - struct object_id oid; - - if (!check_int(get_oid_arbitrary_hex(hex, &oid), ==, 0)) - return; - if (!check_int(oidtree_contains(ot, &oid), ==, expected)) - test_msg("oid: %s", oid_to_hex(&oid)); -} - -struct expected_hex_iter { - size_t i; - struct strvec expected_hexes; - const char *query; -}; - -static enum cb_next check_each_cb(const struct object_id *oid, void *data) -{ - struct expected_hex_iter *hex_iter = data; - struct object_id expected; - - if (!check_int(hex_iter->i, <, hex_iter->expected_hexes.nr)) { - test_msg("error: extraneous callback for query: ('%s'), object_id: ('%s')", - hex_iter->query, oid_to_hex(oid)); - return CB_BREAK; - } - - if (!check_int(get_oid_arbitrary_hex(hex_iter->expected_hexes.v[hex_iter->i], - &expected), ==, 0)) - ; /* the data is bogus and cannot be used */ - else if (!check(oideq(oid, &expected))) - test_msg("expected: %s\n got: %s\n query: %s", - oid_to_hex(&expected), oid_to_hex(oid), hex_iter->query); - - hex_iter->i += 1; - return CB_CONTINUE; -} - -LAST_ARG_MUST_BE_NULL -static void check_each(struct oidtree *ot, const char *query, ...) -{ - struct object_id oid; - struct expected_hex_iter hex_iter = { .expected_hexes = STRVEC_INIT, - .query = query }; - const char *arg; - va_list hex_args; - - va_start(hex_args, query); - while ((arg = va_arg(hex_args, const char *))) - strvec_push(&hex_iter.expected_hexes, arg); - va_end(hex_args); - - if (!check_int(get_oid_arbitrary_hex(query, &oid), ==, 0)) - return; - oidtree_each(ot, &oid, strlen(query), check_each_cb, &hex_iter); - - if (!check_int(hex_iter.i, ==, hex_iter.expected_hexes.nr)) - test_msg("error: could not find some 'object_id's for query ('%s')", query); - strvec_clear(&hex_iter.expected_hexes); -} - -static void setup(void (*f)(struct oidtree *ot)) -{ - struct oidtree ot; - - oidtree_init(&ot); - f(&ot); - oidtree_clear(&ot); -} - -static void t_contains(struct oidtree *ot) -{ - FILL_TREE(ot, "444", "1", "2", "3", "4", "5", "a", "b", "c", "d", "e"); - check_contains(ot, "44", 0); - check_contains(ot, "441", 0); - check_contains(ot, "440", 0); - check_contains(ot, "444", 1); - check_contains(ot, "4440", 1); - check_contains(ot, "4444", 0); -} - -static void t_each(struct oidtree *ot) -{ - FILL_TREE(ot, "f", "9", "8", "123", "321", "320", "a", "b", "c", "d", "e"); - check_each(ot, "12300", "123", NULL); - check_each(ot, "3211", NULL); /* should not reach callback */ - check_each(ot, "3210", "321", NULL); - check_each(ot, "32100", "321", NULL); - check_each(ot, "32", "320", "321", NULL); -} - -int cmd_main(int argc UNUSED, const char **argv UNUSED) -{ - TEST(setup(t_contains), "oidtree insert and contains works"); - TEST(setup(t_each), "oidtree each works"); - return test_done(); -} diff --git a/t/unit-tests/u-oidtree.c b/t/unit-tests/u-oidtree.c new file mode 100644 index 00000000000000..e6eede27404842 --- /dev/null +++ b/t/unit-tests/u-oidtree.c @@ -0,0 +1,107 @@ +#include "unit-test.h" +#include "lib-oid.h" +#include "oidtree.h" +#include "hash.h" +#include "hex.h" +#include "strvec.h" + +static struct oidtree ot; + +#define FILL_TREE(tree, ...) \ + do { \ + const char *hexes[] = { __VA_ARGS__ }; \ + if (fill_tree_loc(tree, hexes, ARRAY_SIZE(hexes))) \ + return; \ + } while (0) + +static int fill_tree_loc(struct oidtree *ot, const char *hexes[], size_t n) +{ + for (size_t i = 0; i < n; i++) { + struct object_id oid; + cl_parse_any_oid(hexes[i], &oid); + oidtree_insert(ot, &oid); + } + return 0; +} + +static void check_contains(struct oidtree *ot, const char *hex, int expected) +{ + struct object_id oid; + + cl_parse_any_oid(hex, &oid); + cl_assert_equal_i(oidtree_contains(ot, &oid), expected); +} + +struct expected_hex_iter { + size_t i; + struct strvec expected_hexes; + const char *query; +}; + +static enum cb_next check_each_cb(const struct object_id *oid, void *data) +{ + struct expected_hex_iter *hex_iter = data; + struct object_id expected; + + cl_assert(hex_iter->i < hex_iter->expected_hexes.nr); + + cl_parse_any_oid(hex_iter->expected_hexes.v[hex_iter->i], + &expected); + cl_assert_equal_s(oid_to_hex(oid), oid_to_hex(&expected)); + hex_iter->i += 1; + return CB_CONTINUE; +} + +LAST_ARG_MUST_BE_NULL +static void check_each(struct oidtree *ot, const char *query, ...) +{ + struct object_id oid; + struct expected_hex_iter hex_iter = { .expected_hexes = STRVEC_INIT, + .query = query }; + const char *arg; + va_list hex_args; + + va_start(hex_args, query); + while ((arg = va_arg(hex_args, const char *))) + strvec_push(&hex_iter.expected_hexes, arg); + va_end(hex_args); + + cl_parse_any_oid(query, &oid); + oidtree_each(ot, &oid, strlen(query), check_each_cb, &hex_iter); + + if (hex_iter.i != hex_iter.expected_hexes.nr) + cl_failf("error: could not find some 'object_id's for query ('%s')", query); + + strvec_clear(&hex_iter.expected_hexes); +} + +void test_oidtree__initialize(void) +{ + oidtree_init(&ot); +} + +void test_oidtree__cleanup(void) +{ + oidtree_clear(&ot); +} + +void test_oidtree__contains(void) +{ + FILL_TREE(&ot, "444", "1", "2", "3", "4", "5", "a", "b", "c", "d", "e"); + check_contains(&ot, "44", 0); + check_contains(&ot, "441", 0); + check_contains(&ot, "440", 0); + check_contains(&ot, "444", 1); + check_contains(&ot, "4440", 1); + check_contains(&ot, "4444", 0); +} + +void test_oidtree__each(void) +{ + FILL_TREE(&ot, "f", "9", "8", "123", "321", "320", "a", "b", "c", "d", "e"); + check_each(&ot, "12300", "123", NULL); + check_each(&ot, "3211", NULL); /* should not reach callback */ + check_each(&ot, "3210", "321", NULL); + check_each(&ot, "32100", "321", NULL); + check_each(&ot, "32", "320", "321", NULL); +} From 8ee018d863e521f32a9cb92db66c25e848b5e0d0 Mon Sep 17 00:00:00 2001 From: Patrick Steinhardt Date: Fri, 7 Feb 2025 12:03:36 +0100 Subject: [PATCH 017/312] rerere: let `rerere_path()` write paths into a caller-provided buffer Same as with `get_worktree_git_dir()` a couple of commits ago, the `rerere_path()` function returns paths that need not be free'd by the caller because `git_path()` internally uses `get_pathname()`. Refactor the function to instead accept a caller-provided buffer that the path will be written into, passing on ownership to the caller. This refactoring prepares us for the removal of `git_path()`. Signed-off-by: Patrick Steinhardt Signed-off-by: Junio C Hamano --- builtin/rerere.c | 11 +++--- rerere.c | 87 +++++++++++++++++++++++++++++++++--------------- rerere.h | 3 +- 3 files changed, 69 insertions(+), 32 deletions(-) diff --git a/builtin/rerere.c b/builtin/rerere.c index 41127e24e53098..1312e79d89f38e 100644 --- a/builtin/rerere.c +++ b/builtin/rerere.c @@ -4,9 +4,9 @@ #include "config.h" #include "gettext.h" #include "parse-options.h" - -#include "string-list.h" #include "rerere.h" +#include "strbuf.h" +#include "string-list.h" #include "xdiff/xdiff.h" #include "xdiff-interface.h" #include "pathspec.h" @@ -112,15 +112,18 @@ int cmd_rerere(int argc, merge_rr.items[i].util = NULL; } } else if (!strcmp(argv[0], "diff")) { + struct strbuf buf = STRBUF_INIT; if (setup_rerere(the_repository, &merge_rr, flags | RERERE_READONLY) < 0) return 0; for (size_t i = 0; i < merge_rr.nr; i++) { const char *path = merge_rr.items[i].string; const struct rerere_id *id = merge_rr.items[i].util; - if (diff_two(rerere_path(id, "preimage"), path, path, path)) - die(_("unable to generate diff for '%s'"), rerere_path(id, NULL)); + if (diff_two(rerere_path(&buf, id, "preimage"), path, path, path)) + die(_("unable to generate diff for '%s'"), rerere_path(&buf, id, NULL)); } + + strbuf_release(&buf); } else usage_with_options(rerere_usage, options); diff --git a/rerere.c b/rerere.c index e7fa6426b399e0..6dd6b959b52d82 100644 --- a/rerere.c +++ b/rerere.c @@ -91,16 +91,18 @@ static void assign_variant(struct rerere_id *id) id->variant = variant; } -const char *rerere_path(const struct rerere_id *id, const char *file) +const char *rerere_path(struct strbuf *buf, const struct rerere_id *id, const char *file) { if (!file) - return git_path("rr-cache/%s", rerere_id_hex(id)); + return repo_git_path_replace(the_repository, buf, "rr-cache/%s", + rerere_id_hex(id)); if (id->variant <= 0) - return git_path("rr-cache/%s/%s", rerere_id_hex(id), file); + return repo_git_path_replace(the_repository, buf, "rr-cache/%s/%s", + rerere_id_hex(id), file); - return git_path("rr-cache/%s/%s.%d", - rerere_id_hex(id), file, id->variant); + return repo_git_path_replace(the_repository, buf, "rr-cache/%s/%s.%d", + rerere_id_hex(id), file, id->variant); } static int is_rr_file(const char *name, const char *filename, int *variant) @@ -624,9 +626,10 @@ static int try_merge(struct index_state *istate, { enum ll_merge_result ret; mmfile_t base = {NULL, 0}, other = {NULL, 0}; + struct strbuf buf = STRBUF_INIT; - if (read_mmfile(&base, rerere_path(id, "preimage")) || - read_mmfile(&other, rerere_path(id, "postimage"))) { + if (read_mmfile(&base, rerere_path(&buf, id, "preimage")) || + read_mmfile(&other, rerere_path(&buf, id, "postimage"))) { ret = LL_MERGE_CONFLICT; } else { /* @@ -637,6 +640,7 @@ static int try_merge(struct index_state *istate, istate, NULL); } + strbuf_release(&buf); free(base.ptr); free(other.ptr); @@ -657,6 +661,7 @@ static int merge(struct index_state *istate, const struct rerere_id *id, const c { FILE *f; int ret; + struct strbuf buf = STRBUF_INIT; mmfile_t cur = {NULL, 0}; mmbuffer_t result = {NULL, 0}; @@ -664,8 +669,8 @@ static int merge(struct index_state *istate, const struct rerere_id *id, const c * Normalize the conflicts in path and write it out to * "thisimage" temporary file. */ - if ((handle_file(istate, path, NULL, rerere_path(id, "thisimage")) < 0) || - read_mmfile(&cur, rerere_path(id, "thisimage"))) { + if ((handle_file(istate, path, NULL, rerere_path(&buf, id, "thisimage")) < 0) || + read_mmfile(&cur, rerere_path(&buf, id, "thisimage"))) { ret = 1; goto out; } @@ -678,9 +683,9 @@ static int merge(struct index_state *istate, const struct rerere_id *id, const c * A successful replay of recorded resolution. * Mark that "postimage" was used to help gc. */ - if (utime(rerere_path(id, "postimage"), NULL) < 0) + if (utime(rerere_path(&buf, id, "postimage"), NULL) < 0) warning_errno(_("failed utime() on '%s'"), - rerere_path(id, "postimage")); + rerere_path(&buf, id, "postimage")); /* Update "path" with the resolution */ f = fopen(path, "w"); @@ -694,6 +699,7 @@ static int merge(struct index_state *istate, const struct rerere_id *id, const c out: free(cur.ptr); free(result.ptr); + strbuf_release(&buf); return ret; } @@ -720,9 +726,11 @@ static void update_paths(struct repository *r, struct string_list *update) static void remove_variant(struct rerere_id *id) { - unlink_or_warn(rerere_path(id, "postimage")); - unlink_or_warn(rerere_path(id, "preimage")); + struct strbuf buf = STRBUF_INIT; + unlink_or_warn(rerere_path(&buf, id, "postimage")); + unlink_or_warn(rerere_path(&buf, id, "preimage")); id->collection->status[id->variant] = 0; + strbuf_release(&buf); } /* @@ -739,6 +747,7 @@ static void do_rerere_one_path(struct index_state *istate, const char *path = rr_item->string; struct rerere_id *id = rr_item->util; struct rerere_dir *rr_dir = id->collection; + struct strbuf buf = STRBUF_INIT; int variant; variant = id->variant; @@ -746,12 +755,12 @@ static void do_rerere_one_path(struct index_state *istate, /* Has the user resolved it already? */ if (variant >= 0) { if (!handle_file(istate, path, NULL, NULL)) { - copy_file(rerere_path(id, "postimage"), path, 0666); + copy_file(rerere_path(&buf, id, "postimage"), path, 0666); id->collection->status[variant] |= RR_HAS_POSTIMAGE; fprintf_ln(stderr, _("Recorded resolution for '%s'."), path); free_rerere_id(rr_item); rr_item->util = NULL; - return; + goto out; } /* * There may be other variants that can cleanly @@ -787,22 +796,25 @@ static void do_rerere_one_path(struct index_state *istate, path); free_rerere_id(rr_item); rr_item->util = NULL; - return; + goto out; } /* None of the existing one applies; we need a new variant */ assign_variant(id); variant = id->variant; - handle_file(istate, path, NULL, rerere_path(id, "preimage")); + handle_file(istate, path, NULL, rerere_path(&buf, id, "preimage")); if (id->collection->status[variant] & RR_HAS_POSTIMAGE) { - const char *path = rerere_path(id, "postimage"); + const char *path = rerere_path(&buf, id, "postimage"); if (unlink(path)) die_errno(_("cannot unlink stray '%s'"), path); id->collection->status[variant] &= ~RR_HAS_POSTIMAGE; } id->collection->status[variant] |= RR_HAS_PREIMAGE; fprintf_ln(stderr, _("Recorded preimage for '%s'"), path); + +out: + strbuf_release(&buf); } static int do_plain_rerere(struct repository *r, @@ -810,6 +822,7 @@ static int do_plain_rerere(struct repository *r, { struct string_list conflict = STRING_LIST_INIT_DUP; struct string_list update = STRING_LIST_INIT_DUP; + struct strbuf buf = STRBUF_INIT; int i; find_conflict(r, &conflict); @@ -843,7 +856,7 @@ static int do_plain_rerere(struct repository *r, string_list_insert(rr, path)->util = id; /* Ensure that the directory exists. */ - mkdir_in_gitdir(rerere_path(id, NULL)); + mkdir_in_gitdir(rerere_path(&buf, id, NULL)); } for (i = 0; i < rr->nr; i++) @@ -854,6 +867,7 @@ static int do_plain_rerere(struct repository *r, string_list_clear(&conflict, 0); string_list_clear(&update, 0); + strbuf_release(&buf); return write_rr(rr, fd); } @@ -1033,6 +1047,7 @@ static int rerere_forget_one_path(struct index_state *istate, struct rerere_id *id; unsigned char hash[GIT_MAX_RAWSZ]; int ret; + struct strbuf buf = STRBUF_INIT; struct string_list_item *item; /* @@ -1056,8 +1071,8 @@ static int rerere_forget_one_path(struct index_state *istate, if (!has_rerere_resolution(id)) continue; - handle_cache(istate, path, hash, rerere_path(id, "thisimage")); - if (read_mmfile(&cur, rerere_path(id, "thisimage"))) { + handle_cache(istate, path, hash, rerere_path(&buf, id, "thisimage")); + if (read_mmfile(&cur, rerere_path(&buf, id, "thisimage"))) { free(cur.ptr); error(_("failed to update conflicted state in '%s'"), path); goto fail_exit; @@ -1074,7 +1089,7 @@ static int rerere_forget_one_path(struct index_state *istate, goto fail_exit; } - filename = rerere_path(id, "postimage"); + filename = rerere_path(&buf, id, "postimage"); if (unlink(filename)) { if (errno == ENOENT) error(_("no remembered resolution for '%s'"), path); @@ -1088,7 +1103,7 @@ static int rerere_forget_one_path(struct index_state *istate, * conflict in the working tree, run us again to record * the postimage. */ - handle_cache(istate, path, hash, rerere_path(id, "preimage")); + handle_cache(istate, path, hash, rerere_path(&buf, id, "preimage")); fprintf_ln(stderr, _("Updated preimage for '%s'"), path); /* @@ -1099,9 +1114,11 @@ static int rerere_forget_one_path(struct index_state *istate, free_rerere_id(item); item->util = id; fprintf(stderr, _("Forgot resolution for '%s'\n"), path); + strbuf_release(&buf); return 0; fail_exit: + strbuf_release(&buf); free(id); return -1; } @@ -1147,16 +1164,26 @@ int rerere_forget(struct repository *r, struct pathspec *pathspec) static timestamp_t rerere_created_at(struct rerere_id *id) { + struct strbuf buf = STRBUF_INIT; struct stat st; + timestamp_t ret; + + ret = stat(rerere_path(&buf, id, "preimage"), &st) ? (time_t) 0 : st.st_mtime; - return stat(rerere_path(id, "preimage"), &st) ? (time_t) 0 : st.st_mtime; + strbuf_release(&buf); + return ret; } static timestamp_t rerere_last_used_at(struct rerere_id *id) { + struct strbuf buf = STRBUF_INIT; struct stat st; + timestamp_t ret; + + ret = stat(rerere_path(&buf, id, "postimage"), &st) ? (time_t) 0 : st.st_mtime; - return stat(rerere_path(id, "postimage"), &st) ? (time_t) 0 : st.st_mtime; + strbuf_release(&buf); + return ret; } /* @@ -1164,9 +1191,11 @@ static timestamp_t rerere_last_used_at(struct rerere_id *id) */ static void unlink_rr_item(struct rerere_id *id) { - unlink_or_warn(rerere_path(id, "thisimage")); + struct strbuf buf = STRBUF_INIT; + unlink_or_warn(rerere_path(&buf, id, "thisimage")); remove_variant(id); id->collection->status[id->variant] = 0; + strbuf_release(&buf); } static void prune_one(struct rerere_id *id, @@ -1264,10 +1293,14 @@ void rerere_clear(struct repository *r, struct string_list *merge_rr) for (i = 0; i < merge_rr->nr; i++) { struct rerere_id *id = merge_rr->items[i].util; + struct strbuf buf = STRBUF_INIT; + if (!has_rerere_resolution(id)) { unlink_rr_item(id); - rmdir(rerere_path(id, NULL)); + rmdir(rerere_path(&buf, id, NULL)); } + + strbuf_release(&buf); } unlink_or_warn(git_path_merge_rr(r)); rollback_lock_file(&write_lock); diff --git a/rerere.h b/rerere.h index 5d6cb638793a49..d4b5f7c932006a 100644 --- a/rerere.h +++ b/rerere.h @@ -32,7 +32,8 @@ int repo_rerere(struct repository *, int); * path to that filesystem entity. With "file" specified with NULL, * return the path to the directory that houses these files. */ -const char *rerere_path(const struct rerere_id *, const char *file); +const char *rerere_path(struct strbuf *buf, const struct rerere_id *, + const char *file); int rerere_forget(struct repository *, struct pathspec *); int rerere_remaining(struct repository *, struct string_list *); void rerere_clear(struct repository *, struct string_list *); From 88dd321cfedc6ee190dfafe4670a83ea33cdf4a3 Mon Sep 17 00:00:00 2001 From: Patrick Steinhardt Date: Fri, 7 Feb 2025 12:03:37 +0100 Subject: [PATCH 018/312] path: drop `git_path()` in favor of `repo_git_path()` Remove `git_path()` in favor of the `repo_git_path()` family of functions, which makes the implicit dependency on `the_repository` go away. Note that `git_path()` returned a string allocated via `get_pathname()`, which uses a rotating set of statically allocated buffers. Consequently, callers didn't have to free the returned string. The same isn't true for `repo_common_path()`, so we also have to add logic to free the returned strings. This refactoring also allows us to remove `repo_common_pathv()` as well as `get_pathname()` from the public interface. Signed-off-by: Patrick Steinhardt Signed-off-by: Junio C Hamano --- builtin/commit.c | 8 +++++--- builtin/gc.c | 21 +++++++++++++++------ builtin/notes.c | 7 ++++++- builtin/rebase.c | 2 +- builtin/remote.c | 6 ++++-- builtin/rev-parse.c | 6 +++--- builtin/worktree.c | 11 +++++++++-- notes-merge.c | 20 ++++++++++++-------- path.c | 8 ++++---- path.h | 27 --------------------------- read-cache.c | 24 +++++++++++++++++------- remote.c | 21 ++++++++++++--------- rerere.c | 14 +++++++++++--- shallow.c | 4 +++- wt-status.c | 42 ++++++++++++++++++++++++++---------------- 15 files changed, 128 insertions(+), 93 deletions(-) diff --git a/builtin/commit.c b/builtin/commit.c index 9fb405dd4a628d..2f459682221d6a 100644 --- a/builtin/commit.c +++ b/builtin/commit.c @@ -352,6 +352,7 @@ static const char *prepare_index(const char **argv, const char *prefix, struct pathspec pathspec; int refresh_flags = REFRESH_QUIET; const char *ret; + char *path = NULL; if (is_status) refresh_flags |= REFRESH_UNMERGED; @@ -524,9 +525,9 @@ static const char *prepare_index(const char **argv, const char *prefix, if (write_locked_index(the_repository->index, &index_lock, 0)) die(_("unable to write new index file")); - hold_lock_file_for_update(&false_lock, - git_path("next-index-%"PRIuMAX, - (uintmax_t) getpid()), + path = repo_git_path(the_repository, "next-index-%"PRIuMAX, + (uintmax_t) getpid()); + hold_lock_file_for_update(&false_lock, path, LOCK_DIE_ON_ERROR); create_base_index(current_head); @@ -542,6 +543,7 @@ static const char *prepare_index(const char **argv, const char *prefix, out: string_list_clear(&partial, 0); clear_pathspec(&pathspec); + free(path); return ret; } diff --git a/builtin/gc.c b/builtin/gc.c index 57f6aee174fdb2..5923d9a05ba97d 100644 --- a/builtin/gc.c +++ b/builtin/gc.c @@ -99,9 +99,11 @@ static void process_log_file(void) /* There was some error recorded in the lock file */ commit_lock_file(&log_lock); } else { + char *path = repo_git_path(the_repository, "gc.log"); /* No error, clean up any old gc.log */ - unlink(git_path("gc.log")); + unlink(path); rollback_lock_file(&log_lock); + free(path); } } @@ -299,8 +301,11 @@ static int too_many_loose_objects(struct gc_config *cfg) int num_loose = 0; int needed = 0; const unsigned hexsz_loose = the_hash_algo->hexsz - 2; + char *path; - dir = opendir(git_path("objects/17")); + path = repo_git_path(the_repository, "objects/17"); + dir = opendir(path); + free(path); if (!dir) return 0; @@ -821,11 +826,12 @@ struct repository *repo UNUSED) } if (daemonized) { - hold_lock_file_for_update(&log_lock, - git_path("gc.log"), + char *path = repo_git_path(the_repository, "gc.log"); + hold_lock_file_for_update(&log_lock, path, LOCK_DIE_ON_ERROR); dup2(get_lock_file_fd(&log_lock), 2); atexit(process_log_file_at_exit); + free(path); } gc_before_repack(&opts, &cfg); @@ -887,8 +893,11 @@ struct repository *repo UNUSED) warning(_("There are too many unreachable loose objects; " "run 'git prune' to remove them.")); - if (!daemonized) - unlink(git_path("gc.log")); + if (!daemonized) { + char *path = repo_git_path(the_repository, "gc.log"); + unlink(path); + free(path); + } out: gc_config_release(&cfg); diff --git a/builtin/notes.c b/builtin/notes.c index 18bcbb2f913430..ff61ec5f2da953 100644 --- a/builtin/notes.c +++ b/builtin/notes.c @@ -979,6 +979,8 @@ static int merge(int argc, const char **argv, const char *prefix, else { /* Merge has unresolved conflicts */ struct worktree **worktrees; const struct worktree *wt; + char *path; + /* Update .git/NOTES_MERGE_PARTIAL with partial merge result */ refs_update_ref(get_main_ref_store(the_repository), msg.buf, "NOTES_MERGE_PARTIAL", &result_oid, NULL, @@ -994,10 +996,13 @@ static int merge(int argc, const char **argv, const char *prefix, if (refs_update_symref(get_main_ref_store(the_repository), "NOTES_MERGE_REF", notes_ref, NULL)) die(_("failed to store link to current notes ref (%s)"), notes_ref); + + path = repo_git_path(the_repository, NOTES_MERGE_WORKTREE); fprintf(stderr, _("Automatic notes merge failed. Fix conflicts in %s " "and commit the result with 'git notes merge --commit', " "or abort the merge with 'git notes merge --abort'.\n"), - git_path(NOTES_MERGE_WORKTREE)); + path); + free(path); } free_notes(t); diff --git a/builtin/rebase.c b/builtin/rebase.c index 6c9eaf3788c900..d4715ed35d77ed 100644 --- a/builtin/rebase.c +++ b/builtin/rebase.c @@ -644,7 +644,7 @@ static int run_am(struct rebase_options *opts) return run_command(&am); } - rebased_patches = xstrdup(git_path("rebased-patches")); + rebased_patches = repo_git_path(the_repository, "rebased-patches"); format_patch.out = open(rebased_patches, O_WRONLY | O_CREAT | O_TRUNC, 0666); if (format_patch.out < 0) { diff --git a/builtin/remote.c b/builtin/remote.c index 71d84fb3cfbef2..0489fcc8f3331e 100644 --- a/builtin/remote.c +++ b/builtin/remote.c @@ -644,9 +644,11 @@ static int migrate_file(struct remote *remote) git_config_set_multivar(buf.buf, remote->fetch.items[i].raw, "^$", 0); #ifndef WITH_BREAKING_CHANGES if (remote->origin == REMOTE_REMOTES) - unlink_or_warn(git_path("remotes/%s", remote->name)); + unlink_or_warn(repo_git_path_replace(the_repository, &buf, + "remotes/%s", remote->name)); else if (remote->origin == REMOTE_BRANCHES) - unlink_or_warn(git_path("branches/%s", remote->name)); + unlink_or_warn(repo_git_path_replace(the_repository, &buf, + "branches/%s", remote->name)); #endif /* WITH_BREAKING_CHANGES */ strbuf_release(&buf); diff --git a/builtin/rev-parse.c b/builtin/rev-parse.c index 428c866c05f4a4..490da33beccf72 100644 --- a/builtin/rev-parse.c +++ b/builtin/rev-parse.c @@ -789,8 +789,8 @@ int cmd_rev_parse(int argc, if (!strcmp(arg, "--git-path")) { if (!argv[i + 1]) die(_("--git-path requires an argument")); - strbuf_reset(&buf); - print_path(git_path("%s", argv[i + 1]), prefix, + print_path(repo_git_path_replace(the_repository, &buf, + "%s", argv[i + 1]), prefix, format, DEFAULT_RELATIVE_IF_SHARED); i++; @@ -1083,7 +1083,7 @@ int cmd_rev_parse(int argc, die(_("Could not read the index")); if (the_repository->index->split_index) { const struct object_id *oid = &the_repository->index->split_index->base_oid; - const char *path = git_path("sharedindex.%s", oid_to_hex(oid)); + const char *path = repo_git_path_replace(the_repository, &buf, "sharedindex.%s", oid_to_hex(oid)); print_path(path, prefix, format, DEFAULT_RELATIVE); } continue; diff --git a/builtin/worktree.c b/builtin/worktree.c index 761e302a369e39..48448a83550944 100644 --- a/builtin/worktree.c +++ b/builtin/worktree.c @@ -163,7 +163,9 @@ static int delete_git_dir(const char *id) static void delete_worktrees_dir_if_empty(void) { - rmdir(git_path("worktrees")); /* ignore failed removal */ + char *path = repo_git_path(the_repository, "worktrees"); + rmdir(path); /* ignore failed removal */ + free(path); } static void prune_worktree(const char *id, const char *reason) @@ -212,8 +214,13 @@ static void prune_worktrees(void) struct strbuf reason = STRBUF_INIT; struct strbuf main_path = STRBUF_INIT; struct string_list kept = STRING_LIST_INIT_DUP; - DIR *dir = opendir(git_path("worktrees")); + char *path; + DIR *dir; struct dirent *d; + + path = repo_git_path(the_repository, "worktrees"); + dir = opendir(path); + free(path); if (!dir) return; while ((d = readdir_skip_dot_and_dotdot(dir)) != NULL) { diff --git a/notes-merge.c b/notes-merge.c index 8c22a171c147ba..67a472020dedd9 100644 --- a/notes-merge.c +++ b/notes-merge.c @@ -275,34 +275,38 @@ static void diff_tree_local(struct notes_merge_options *o, static void check_notes_merge_worktree(struct notes_merge_options *o) { + struct strbuf buf = STRBUF_INIT; + if (!o->has_worktree) { /* * Must establish NOTES_MERGE_WORKTREE. * Abort if NOTES_MERGE_WORKTREE already exists */ - if (file_exists(git_path(NOTES_MERGE_WORKTREE)) && - !is_empty_dir(git_path(NOTES_MERGE_WORKTREE))) { + if (file_exists(repo_git_path_replace(the_repository, &buf, NOTES_MERGE_WORKTREE)) && + !is_empty_dir(repo_git_path_replace(the_repository, &buf, NOTES_MERGE_WORKTREE))) { if (advice_enabled(ADVICE_RESOLVE_CONFLICT)) die(_("You have not concluded your previous " "notes merge (%s exists).\nPlease, use " "'git notes merge --commit' or 'git notes " "merge --abort' to commit/abort the " "previous merge before you start a new " - "notes merge."), git_path("NOTES_MERGE_*")); + "notes merge."), repo_git_path_replace(the_repository, &buf, "NOTES_MERGE_*")); else die(_("You have not concluded your notes merge " - "(%s exists)."), git_path("NOTES_MERGE_*")); + "(%s exists)."), repo_git_path_replace(the_repository, &buf, "NOTES_MERGE_*")); } - if (safe_create_leading_directories_const(git_path( + if (safe_create_leading_directories_const(repo_git_path_replace(the_repository, &buf, NOTES_MERGE_WORKTREE "/.test"))) die_errno("unable to create directory %s", - git_path(NOTES_MERGE_WORKTREE)); + repo_git_path_replace(the_repository, &buf, NOTES_MERGE_WORKTREE)); o->has_worktree = 1; - } else if (!file_exists(git_path(NOTES_MERGE_WORKTREE))) + } else if (!file_exists(repo_git_path_replace(the_repository, &buf, NOTES_MERGE_WORKTREE))) /* NOTES_MERGE_WORKTREE should already be established */ die("missing '%s'. This should not happen", - git_path(NOTES_MERGE_WORKTREE)); + repo_git_path_replace(the_repository, &buf, NOTES_MERGE_WORKTREE)); + + strbuf_release(&buf); } static void write_buf_to_worktree(const struct object_id *obj, diff --git a/path.c b/path.c index 2d07ba723dfd25..a42f72800db673 100644 --- a/path.c +++ b/path.c @@ -30,7 +30,7 @@ static int get_st_mode_bits(const char *path, int *mode) return 0; } -struct strbuf *get_pathname(void) +static struct strbuf *get_pathname(void) { static struct strbuf pathname_array[4] = { STRBUF_INIT, STRBUF_INIT, STRBUF_INIT, STRBUF_INIT @@ -417,9 +417,9 @@ static void strbuf_worktree_gitdir(struct strbuf *buf, repo_common_path_append(repo, buf, "worktrees/%s", wt->id); } -void repo_git_pathv(const struct repository *repo, - const struct worktree *wt, struct strbuf *buf, - const char *fmt, va_list args) +static void repo_git_pathv(const struct repository *repo, + const struct worktree *wt, struct strbuf *buf, + const char *fmt, va_list args) { int gitdir_len; strbuf_worktree_gitdir(buf, repo, wt); diff --git a/path.h b/path.h index bed0a4c6f9eca3..f28d5a7ca9525c 100644 --- a/path.h +++ b/path.h @@ -221,37 +221,10 @@ char *xdg_cache_home(const char *filename); */ void safe_create_dir(const char *dir, int share); -/* - * Do not use this function. It is only exported to other subsystems until we - * can get rid of the below block of functions that implicitly rely on - * `the_repository`. - */ -struct strbuf *get_pathname(void); - # ifdef USE_THE_REPOSITORY_VARIABLE # include "strbuf.h" # include "repository.h" -/* Internal implementation details that should not be used. */ -void repo_git_pathv(const struct repository *repo, - const struct worktree *wt, struct strbuf *buf, - const char *fmt, va_list args); - -/* - * Return a statically allocated path into the main repository's - * (the_repository) git directory. - */ -__attribute__((format (printf, 1, 2))) -static inline const char *git_path(const char *fmt, ...) -{ - struct strbuf *pathname = get_pathname(); - va_list args; - va_start(args, fmt); - repo_git_pathv(the_repository, NULL, pathname, fmt, args); - va_end(args); - return pathname->buf; -} - #define GIT_PATH_FUNC(func, filename) \ const char *func(void) \ { \ diff --git a/read-cache.c b/read-cache.c index d54be2c1726872..66ad0015a7f75a 100644 --- a/read-cache.c +++ b/read-cache.c @@ -3251,15 +3251,18 @@ static int clean_shared_index_files(const char *current_hex) while ((de = readdir(dir)) != NULL) { const char *sha1_hex; - const char *shared_index_path; + char *shared_index_path; if (!skip_prefix(de->d_name, "sharedindex.", &sha1_hex)) continue; if (!strcmp(sha1_hex, current_hex)) continue; - shared_index_path = git_path("%s", de->d_name); + + shared_index_path = repo_git_path(the_repository, "%s", de->d_name); if (should_delete_shared_index(shared_index_path) > 0 && unlink(shared_index_path)) warning_errno(_("unable to unlink: %s"), shared_index_path); + + free(shared_index_path); } closedir(dir); @@ -3271,6 +3274,7 @@ static int write_shared_index(struct index_state *istate, { struct split_index *si = istate->split_index; int ret, was_full = !istate->sparse_index; + char *path; move_cache_to_base_index(istate); convert_to_sparse(istate, 0); @@ -3291,13 +3295,15 @@ static int write_shared_index(struct index_state *istate, error(_("cannot fix permission bits on '%s'"), get_tempfile_path(*temp)); return ret; } - ret = rename_tempfile(temp, - git_path("sharedindex.%s", oid_to_hex(&si->base->oid))); + + path = repo_git_path(the_repository, "sharedindex.%s", oid_to_hex(&si->base->oid)); + ret = rename_tempfile(temp, path); if (!ret) { oidcpy(&si->base_oid, &si->base->oid); clean_shared_index_files(oid_to_hex(&si->base->oid)); } + free(path); return ret; } @@ -3378,9 +3384,12 @@ int write_locked_index(struct index_state *istate, struct lock_file *lock, if (new_shared_index) { struct tempfile *temp; int saved_errno; + char *path; /* Same initial permissions as the main .git/index file */ - temp = mks_tempfile_sm(git_path("sharedindex_XXXXXX"), 0, 0666); + path = repo_git_path(the_repository, "sharedindex_XXXXXX"); + temp = mks_tempfile_sm(path, 0, 0666); + free(path); if (!temp) { ret = do_write_locked_index(istate, lock, flags, ~WRITE_SPLIT_INDEX_EXTENSION); @@ -3401,9 +3410,10 @@ int write_locked_index(struct index_state *istate, struct lock_file *lock, /* Freshen the shared index only if the split-index was written */ if (!ret && !new_shared_index && !is_null_oid(&si->base_oid)) { - const char *shared_index = git_path("sharedindex.%s", - oid_to_hex(&si->base_oid)); + char *shared_index = repo_git_path(the_repository, "sharedindex.%s", + oid_to_hex(&si->base_oid)); freshen_shared_index(shared_index, 1); + free(shared_index); } out: diff --git a/remote.c b/remote.c index 1779f0e7bbb8b8..81d151a507b1b7 100644 --- a/remote.c +++ b/remote.c @@ -321,10 +321,11 @@ static void read_remotes_file(struct remote_state *remote_state, struct remote *remote) { struct strbuf buf = STRBUF_INIT; - FILE *f = fopen_or_warn(git_path("remotes/%s", remote->name), "r"); + FILE *f = fopen_or_warn(repo_git_path_append(the_repository, &buf, + "remotes/%s", remote->name), "r"); if (!f) - return; + goto out; warn_about_deprecated_remote_type("remotes", remote); @@ -343,8 +344,10 @@ static void read_remotes_file(struct remote_state *remote_state, else if (skip_prefix(buf.buf, "Pull:", &v)) refspec_append(&remote->fetch, skip_spaces(v)); } - strbuf_release(&buf); fclose(f); + +out: + strbuf_release(&buf); } static void read_branches_file(struct remote_state *remote_state, @@ -352,20 +355,19 @@ static void read_branches_file(struct remote_state *remote_state, { char *frag, *to_free = NULL; struct strbuf buf = STRBUF_INIT; - FILE *f = fopen_or_warn(git_path("branches/%s", remote->name), "r"); + FILE *f = fopen_or_warn(repo_git_path_append(the_repository, &buf, + "branches/%s", remote->name), "r"); if (!f) - return; + goto out; warn_about_deprecated_remote_type("branches", remote); strbuf_getline_lf(&buf, f); fclose(f); strbuf_trim(&buf); - if (!buf.len) { - strbuf_release(&buf); - return; - } + if (!buf.len) + goto out; remote->configured_in_repo = 1; remote->origin = REMOTE_BRANCHES; @@ -393,6 +395,7 @@ static void read_branches_file(struct remote_state *remote_state, refspec_appendf(&remote->push, "HEAD:refs/heads/%s", frag); remote->fetch_tags = 1; /* always auto-follow */ +out: strbuf_release(&buf); free(to_free); } diff --git a/rerere.c b/rerere.c index 6dd6b959b52d82..6bd395bebee5e7 100644 --- a/rerere.c +++ b/rerere.c @@ -127,8 +127,12 @@ static int is_rr_file(const char *name, const char *filename, int *variant) static void scan_rerere_dir(struct rerere_dir *rr_dir) { struct dirent *de; - DIR *dir = opendir(git_path("rr-cache/%s", rr_dir->name)); + char *path; + DIR *dir; + path = repo_git_path(the_repository, "rr-cache/%s", rr_dir->name); + dir = opendir(path); + free(path); if (!dir) return; while ((de = readdir(dir)) != NULL) { @@ -1234,6 +1238,7 @@ void rerere_gc(struct repository *r, struct string_list *rr) timestamp_t now = time(NULL); timestamp_t cutoff_noresolve = now - 15 * 86400; timestamp_t cutoff_resolve = now - 60 * 86400; + struct strbuf buf = STRBUF_INIT; if (setup_rerere(r, rr, 0) < 0) return; @@ -1243,7 +1248,7 @@ void rerere_gc(struct repository *r, struct string_list *rr) repo_config_get_expiry_in_days(the_repository, "gc.rerereunresolved", &cutoff_noresolve, now); git_config(git_default_config, NULL); - dir = opendir(git_path("rr-cache")); + dir = opendir(repo_git_path_replace(the_repository, &buf, "rr-cache")); if (!dir) die_errno(_("unable to open rr-cache directory")); /* Collect stale conflict IDs ... */ @@ -1272,9 +1277,12 @@ void rerere_gc(struct repository *r, struct string_list *rr) /* ... and then remove the empty directories */ for (i = 0; i < to_remove.nr; i++) - rmdir(git_path("rr-cache/%s", to_remove.items[i].string)); + rmdir(repo_git_path_replace(the_repository, &buf, + "rr-cache/%s", to_remove.items[i].string)); + string_list_clear(&to_remove, 0); rollback_lock_file(&write_lock); + strbuf_release(&buf); } /* diff --git a/shallow.c b/shallow.c index b54244ffa90d66..4bd9342c9a745a 100644 --- a/shallow.c +++ b/shallow.c @@ -364,7 +364,9 @@ const char *setup_temporary_shallow(const struct oid_array *extra) struct strbuf sb = STRBUF_INIT; if (write_shallow_commits(&sb, 0, extra)) { - temp = xmks_tempfile(git_path("shallow_XXXXXX")); + char *path = repo_git_path(the_repository, "shallow_XXXXXX"); + temp = xmks_tempfile(path); + free(path); if (write_in_full(temp->fd, sb.buf, sb.len) < 0 || close_tempfile_gently(temp) < 0) diff --git a/wt-status.c b/wt-status.c index 3ee918176422c7..1da5732f57b115 100644 --- a/wt-status.c +++ b/wt-status.c @@ -1289,7 +1289,8 @@ static void show_am_in_progress(struct wt_status *s, static char *read_line_from_git_path(const char *filename) { struct strbuf buf = STRBUF_INIT; - FILE *fp = fopen_or_warn(git_path("%s", filename), "r"); + FILE *fp = fopen_or_warn(repo_git_path_append(the_repository, &buf, + "%s", filename), "r"); if (!fp) { strbuf_release(&buf); @@ -1383,27 +1384,33 @@ static void abbrev_oid_in_line(struct strbuf *line) static int read_rebase_todolist(const char *fname, struct string_list *lines) { - struct strbuf line = STRBUF_INIT; - FILE *f = fopen(git_path("%s", fname), "r"); + struct strbuf buf = STRBUF_INIT; + FILE *f = fopen(repo_git_path_append(the_repository, &buf, "%s", fname), "r"); + int ret; if (!f) { - if (errno == ENOENT) - return -1; + if (errno == ENOENT) { + ret = -1; + goto out; + } die_errno("Could not open file %s for reading", - git_path("%s", fname)); + repo_git_path_replace(the_repository, &buf, "%s", fname)); } - while (!strbuf_getline_lf(&line, f)) { - if (starts_with(line.buf, comment_line_str)) + while (!strbuf_getline_lf(&buf, f)) { + if (starts_with(buf.buf, comment_line_str)) continue; - strbuf_trim(&line); - if (!line.len) + strbuf_trim(&buf); + if (!buf.len) continue; - abbrev_oid_in_line(&line); - string_list_append(lines, line.buf); + abbrev_oid_in_line(&buf); + string_list_append(lines, buf.buf); } fclose(f); - strbuf_release(&line); - return 0; + + ret = 0; +out: + strbuf_release(&buf); + return ret; } static void show_rebase_information(struct wt_status *s, @@ -1434,9 +1441,12 @@ static void show_rebase_information(struct wt_status *s, i < have_done.nr; i++) status_printf_ln(s, color, " %s", have_done.items[i].string); - if (have_done.nr > nr_lines_to_show && s->hints) + if (have_done.nr > nr_lines_to_show && s->hints) { + char *path = repo_git_path(the_repository, "rebase-merge/done"); status_printf_ln(s, color, - _(" (see more in file %s)"), git_path("rebase-merge/done")); + _(" (see more in file %s)"), path); + free(path); + } } if (yet_to_do.nr == 0) From b411ed60c7438eda3fd85a308050e88159f275fd Mon Sep 17 00:00:00 2001 From: Patrick Steinhardt Date: Fri, 7 Feb 2025 12:03:38 +0100 Subject: [PATCH 019/312] repo-settings: introduce function to clear struct We don't provide a way to clear a `struct repo_settings`, and instead open-code this in `repo_clear()`. This is mixing up concerns and means that developers have to touch multiple files whenever they add a new field to the structure in case the associated resources need to be released. Provide a new `repo_settings_clear()` function to improve this. Signed-off-by: Patrick Steinhardt Signed-off-by: Junio C Hamano --- repo-settings.c | 10 ++++++++-- repo-settings.h | 1 + repository.c | 2 +- 3 files changed, 10 insertions(+), 3 deletions(-) diff --git a/repo-settings.c b/repo-settings.c index 9d16d5399e3751..719cd7c85c7be6 100644 --- a/repo-settings.c +++ b/repo-settings.c @@ -21,7 +21,6 @@ static void repo_cfg_int(struct repository *r, const char *key, int *dest, void prepare_repo_settings(struct repository *r) { - const struct repo_settings defaults = REPO_SETTINGS_INIT; int experimental; int value; const char *strval; @@ -35,7 +34,7 @@ void prepare_repo_settings(struct repository *r) if (r->settings.initialized) return; - memcpy(&r->settings, &defaults, sizeof(defaults)); + repo_settings_clear(r); r->settings.initialized++; /* Booleans config or default, cascades to other settings */ @@ -143,6 +142,13 @@ void prepare_repo_settings(struct repository *r) r->settings.packed_git_limit = ulongval; } +void repo_settings_clear(struct repository *r) +{ + struct repo_settings empty = REPO_SETTINGS_INIT; + FREE_AND_NULL(r->settings.fsmonitor); + r->settings = empty; +} + enum log_refs_config repo_settings_get_log_all_ref_updates(struct repository *repo) { const char *value; diff --git a/repo-settings.h b/repo-settings.h index 93ea0c32741125..c4f7e3bd8a8fe8 100644 --- a/repo-settings.h +++ b/repo-settings.h @@ -73,6 +73,7 @@ struct repo_settings { } void prepare_repo_settings(struct repository *r); +void repo_settings_clear(struct repository *r); /* Read the value for "core.logAllRefUpdates". */ enum log_refs_config repo_settings_get_log_all_ref_updates(struct repository *repo); diff --git a/repository.c b/repository.c index 648cd88474eda3..6cbaf2e3daa93a 100644 --- a/repository.c +++ b/repository.c @@ -380,7 +380,7 @@ void repo_clear(struct repository *repo) parsed_object_pool_clear(repo->parsed_objects); FREE_AND_NULL(repo->parsed_objects); - FREE_AND_NULL(repo->settings.fsmonitor); + repo_settings_clear(repo); if (repo->config) { git_configset_clear(repo->config); From 6f3fbed8eda577703426d77dacc71ce0ba46634e Mon Sep 17 00:00:00 2001 From: Patrick Steinhardt Date: Fri, 7 Feb 2025 12:03:39 +0100 Subject: [PATCH 020/312] environment: move access to "core.hooksPath" into repo settings The "core.hooksPath" setting is stored in a global variable and populated via the `git_default_core_config`. This may cause issues in the case where one is handling multiple different repositories in a single process with different values for that config key, as we may or may not see the correct value in that case. Furthermore, global state blocks our path towards libification. Refactor the code so that we instead store the value in `struct repo_settings`. The value is computed as-needed and cached. The result should be functionally the same as there aren't ever any code paths where we'd execute hooks outside the context of a repository. Note that this requires us to change the passed-in repository in the `repo_git_path()` family of functions to be non-constant, as we call `adjust_git_path()` there. Signed-off-by: Patrick Steinhardt Signed-off-by: Junio C Hamano --- config.c | 5 ----- environment.c | 1 - environment.h | 1 - path.c | 15 ++++++++------- path.h | 6 +++--- repo-settings.c | 8 ++++++++ repo-settings.h | 4 ++++ 7 files changed, 23 insertions(+), 17 deletions(-) diff --git a/config.c b/config.c index 50f2d17b399446..d932d4b1347d13 100644 --- a/config.c +++ b/config.c @@ -1436,11 +1436,6 @@ static int git_default_core_config(const char *var, const char *value, return git_config_pathname(&git_attributes_file, var, value); } - if (!strcmp(var, "core.hookspath")) { - FREE_AND_NULL(git_hooks_path); - return git_config_pathname(&git_hooks_path, var, value); - } - if (!strcmp(var, "core.bare")) { is_bare_repository_cfg = git_config_bool(var, value); return 0; diff --git a/environment.c b/environment.c index 8389a272700eac..39755873ee727d 100644 --- a/environment.c +++ b/environment.c @@ -42,7 +42,6 @@ char *git_log_output_encoding; char *apply_default_whitespace; char *apply_default_ignorewhitespace; char *git_attributes_file; -char *git_hooks_path; int zlib_compression_level = Z_BEST_SPEED; int pack_compression_level = Z_DEFAULT_COMPRESSION; int fsync_object_files = -1; diff --git a/environment.h b/environment.h index 2f43340f0b553a..66989afbacd1f4 100644 --- a/environment.h +++ b/environment.h @@ -160,7 +160,6 @@ extern int warn_on_object_refname_ambiguity; extern char *apply_default_whitespace; extern char *apply_default_ignorewhitespace; extern char *git_attributes_file; -extern char *git_hooks_path; extern int zlib_compression_level; extern int pack_compression_level; extern size_t packed_git_window_size; diff --git a/path.c b/path.c index a42f72800db673..e81ebd3b5cfdca 100644 --- a/path.c +++ b/path.c @@ -387,10 +387,11 @@ void report_linked_checkout_garbage(struct repository *r) strbuf_release(&sb); } -static void adjust_git_path(const struct repository *repo, +static void adjust_git_path(struct repository *repo, struct strbuf *buf, int git_dir_len) { const char *base = buf->buf + git_dir_len; + if (is_dir_file(base, "info", "grafts")) strbuf_splice(buf, 0, buf->len, repo->graft_file, strlen(repo->graft_file)); @@ -399,8 +400,8 @@ static void adjust_git_path(const struct repository *repo, repo->index_file, strlen(repo->index_file)); else if (dir_prefix(base, "objects")) replace_dir(buf, git_dir_len + 7, repo->objects->odb->path); - else if (git_hooks_path && dir_prefix(base, "hooks")) - replace_dir(buf, git_dir_len + 5, git_hooks_path); + else if (repo_settings_get_hooks_path(repo) && dir_prefix(base, "hooks")) + replace_dir(buf, git_dir_len + 5, repo_settings_get_hooks_path(repo)); else if (repo->different_commondir) update_common_dir(buf, git_dir_len, repo->commondir); } @@ -417,7 +418,7 @@ static void strbuf_worktree_gitdir(struct strbuf *buf, repo_common_path_append(repo, buf, "worktrees/%s", wt->id); } -static void repo_git_pathv(const struct repository *repo, +static void repo_git_pathv(struct repository *repo, const struct worktree *wt, struct strbuf *buf, const char *fmt, va_list args) { @@ -432,7 +433,7 @@ static void repo_git_pathv(const struct repository *repo, strbuf_cleanup_path(buf); } -char *repo_git_path(const struct repository *repo, +char *repo_git_path(struct repository *repo, const char *fmt, ...) { struct strbuf path = STRBUF_INIT; @@ -443,7 +444,7 @@ char *repo_git_path(const struct repository *repo, return strbuf_detach(&path, NULL); } -const char *repo_git_path_append(const struct repository *repo, +const char *repo_git_path_append(struct repository *repo, struct strbuf *sb, const char *fmt, ...) { @@ -454,7 +455,7 @@ const char *repo_git_path_append(const struct repository *repo, return sb->buf; } -const char *repo_git_path_replace(const struct repository *repo, +const char *repo_git_path_replace(struct repository *repo, struct strbuf *sb, const char *fmt, ...) { diff --git a/path.h b/path.h index f28d5a7ca9525c..373404dd9d2a5f 100644 --- a/path.h +++ b/path.h @@ -52,14 +52,14 @@ const char *repo_common_path_replace(const struct repository *repo, * For an exhaustive list of the adjustments made look at `common_list` and * `adjust_git_path` in path.c. */ -char *repo_git_path(const struct repository *repo, +char *repo_git_path(struct repository *repo, const char *fmt, ...) __attribute__((format (printf, 2, 3))); -const char *repo_git_path_append(const struct repository *repo, +const char *repo_git_path_append(struct repository *repo, struct strbuf *sb, const char *fmt, ...) __attribute__((format (printf, 3, 4))); -const char *repo_git_path_replace(const struct repository *repo, +const char *repo_git_path_replace(struct repository *repo, struct strbuf *sb, const char *fmt, ...) __attribute__((format (printf, 3, 4))); diff --git a/repo-settings.c b/repo-settings.c index 719cd7c85c7be6..876d5275816b84 100644 --- a/repo-settings.c +++ b/repo-settings.c @@ -146,6 +146,7 @@ void repo_settings_clear(struct repository *r) { struct repo_settings empty = REPO_SETTINGS_INIT; FREE_AND_NULL(r->settings.fsmonitor); + FREE_AND_NULL(r->settings.hooks_path); r->settings = empty; } @@ -173,3 +174,10 @@ int repo_settings_get_warn_ambiguous_refs(struct repository *repo) &repo->settings.warn_ambiguous_refs, 1); return repo->settings.warn_ambiguous_refs; } + +const char *repo_settings_get_hooks_path(struct repository *repo) +{ + if (!repo->settings.hooks_path) + repo_config_get_pathname(repo, "core.hookspath", &repo->settings.hooks_path); + return repo->settings.hooks_path; +} diff --git a/repo-settings.h b/repo-settings.h index c4f7e3bd8a8fe8..0cef970443be80 100644 --- a/repo-settings.h +++ b/repo-settings.h @@ -61,6 +61,8 @@ struct repo_settings { size_t delta_base_cache_limit; size_t packed_git_window_size; size_t packed_git_limit; + + char *hooks_path; }; #define REPO_SETTINGS_INIT { \ .index_version = -1, \ @@ -79,5 +81,7 @@ void repo_settings_clear(struct repository *r); enum log_refs_config repo_settings_get_log_all_ref_updates(struct repository *repo); /* Read the value for "core.warnAmbiguousRefs". */ int repo_settings_get_warn_ambiguous_refs(struct repository *repo); +/* Read the value for "core.hooksPath". */ +const char *repo_settings_get_hooks_path(struct repository *repo); #endif /* REPO_SETTINGS_H */ From f1ce861c34bffbc02998173016b0bca0f6d9f6c4 Mon Sep 17 00:00:00 2001 From: Patrick Steinhardt Date: Fri, 7 Feb 2025 12:03:40 +0100 Subject: [PATCH 021/312] environment: move access to "core.sharedRepository" into repo settings Similar as with the preceding commit, we track "core.sharedRepository" via a pair of global variables. Move them into `struct repo_settings` so that we can instead track them per-repository. Signed-off-by: Patrick Steinhardt Signed-off-by: Junio C Hamano --- builtin/init-db.c | 8 ++++---- builtin/log.c | 6 +++--- environment.c | 26 -------------------------- environment.h | 10 ---------- path.c | 10 +++++----- repo-settings.c | 26 ++++++++++++++++++++++++++ repo-settings.h | 9 +++++++++ setup.c | 21 +++++++++++---------- 8 files changed, 58 insertions(+), 58 deletions(-) diff --git a/builtin/init-db.c b/builtin/init-db.c index 096f96b9c4a203..196dccdd77acb8 100644 --- a/builtin/init-db.c +++ b/builtin/init-db.c @@ -132,8 +132,8 @@ int cmd_init_db(int argc, * and we know shared_repository should always be 0; * but just in case we play safe. */ - saved = get_shared_repository(); - set_shared_repository(0); + saved = repo_settings_get_shared_repository(the_repository); + repo_settings_set_shared_repository(the_repository, 0); switch (safe_create_leading_directories_const(argv[0])) { case SCLD_OK: case SCLD_PERMS: @@ -145,7 +145,7 @@ int cmd_init_db(int argc, die_errno(_("cannot mkdir %s"), argv[0]); break; } - set_shared_repository(saved); + repo_settings_set_shared_repository(the_repository, saved); if (mkdir(argv[0], 0777) < 0) die_errno(_("cannot mkdir %s"), argv[0]); mkdir_tried = 1; @@ -175,7 +175,7 @@ int cmd_init_db(int argc, } if (init_shared_repository != -1) - set_shared_repository(init_shared_repository); + repo_settings_set_shared_repository(the_repository, init_shared_repository); /* * GIT_WORK_TREE makes sense only in conjunction with GIT_DIR diff --git a/builtin/log.c b/builtin/log.c index e41f88945eb4ed..04a6ef97bc1442 100644 --- a/builtin/log.c +++ b/builtin/log.c @@ -2309,8 +2309,8 @@ int cmd_format_patch(int argc, * We consider as 'outside of gitdir', therefore avoid * applying adjust_shared_perm in s-c-l-d. */ - saved = get_shared_repository(); - set_shared_repository(0); + saved = repo_settings_get_shared_repository(the_repository); + repo_settings_set_shared_repository(the_repository, 0); switch (safe_create_leading_directories_const(output_directory)) { case SCLD_OK: case SCLD_EXISTS: @@ -2319,7 +2319,7 @@ int cmd_format_patch(int argc, die(_("could not create leading directories " "of '%s'"), output_directory); } - set_shared_repository(saved); + repo_settings_set_shared_repository(the_repository, saved); if (mkdir(output_directory, 0777) < 0 && errno != EEXIST) die_errno(_("could not create directory '%s'"), output_directory); diff --git a/environment.c b/environment.c index 39755873ee727d..c79acc69e74f35 100644 --- a/environment.c +++ b/environment.c @@ -206,32 +206,6 @@ const char *get_commit_output_encoding(void) return git_commit_encoding ? git_commit_encoding : "UTF-8"; } -static int the_shared_repository = PERM_UMASK; -static int need_shared_repository_from_config = 1; - -void set_shared_repository(int value) -{ - the_shared_repository = value; - need_shared_repository_from_config = 0; -} - -int get_shared_repository(void) -{ - if (need_shared_repository_from_config) { - const char *var = "core.sharedrepository"; - const char *value; - if (!git_config_get_value(var, &value)) - the_shared_repository = git_config_perm(var, value); - need_shared_repository_from_config = 0; - } - return the_shared_repository; -} - -void reset_shared_repository(void) -{ - need_shared_repository_from_config = 1; -} - int use_optional_locks(void) { return git_env_bool(GIT_OPTIONAL_LOCKS_ENVIRONMENT, 1); diff --git a/environment.h b/environment.h index 66989afbacd1f4..45e690f203fd1d 100644 --- a/environment.h +++ b/environment.h @@ -134,16 +134,6 @@ void setup_git_env(const char *git_dir); */ int have_git_dir(void); -/* - * Accessors for the core.sharedrepository config which lazy-load the value - * from the config (if not already set). The "reset" function can be - * used to unset "set" or cached value, meaning that the value will be loaded - * fresh from the config file on the next call to get_shared_repository(). - */ -void set_shared_repository(int value); -int get_shared_repository(void); -void reset_shared_repository(void); - extern int is_bare_repository_cfg; int is_bare_repository(void); extern char *git_work_tree_cfg; diff --git a/path.c b/path.c index e81ebd3b5cfdca..a2f402baecf21a 100644 --- a/path.c +++ b/path.c @@ -844,17 +844,17 @@ int calc_shared_perm(int mode) { int tweak; - if (get_shared_repository() < 0) - tweak = -get_shared_repository(); + if (repo_settings_get_shared_repository(the_repository) < 0) + tweak = -repo_settings_get_shared_repository(the_repository); else - tweak = get_shared_repository(); + tweak = repo_settings_get_shared_repository(the_repository); if (!(mode & S_IWUSR)) tweak &= ~0222; if (mode & S_IXUSR) /* Copy read bits to execute bits */ tweak |= (tweak & 0444) >> 2; - if (get_shared_repository() < 0) + if (repo_settings_get_shared_repository(the_repository) < 0) mode = (mode & ~0777) | tweak; else mode |= tweak; @@ -867,7 +867,7 @@ int adjust_shared_perm(const char *path) { int old_mode, new_mode; - if (!get_shared_repository()) + if (!repo_settings_get_shared_repository(the_repository)) return 0; if (get_st_mode_bits(path, &old_mode) < 0) return -1; diff --git a/repo-settings.c b/repo-settings.c index 876d5275816b84..67e9cfd2e63d9c 100644 --- a/repo-settings.c +++ b/repo-settings.c @@ -4,6 +4,7 @@ #include "repository.h" #include "midx.h" #include "pack-objects.h" +#include "setup.h" static void repo_cfg_bool(struct repository *r, const char *key, int *dest, int def) @@ -181,3 +182,28 @@ const char *repo_settings_get_hooks_path(struct repository *repo) repo_config_get_pathname(repo, "core.hookspath", &repo->settings.hooks_path); return repo->settings.hooks_path; } + +int repo_settings_get_shared_repository(struct repository *repo) +{ + if (!repo->settings.shared_repository_initialized) { + const char *var = "core.sharedrepository"; + const char *value; + if (!repo_config_get_value(repo, var, &value)) + repo->settings.shared_repository = git_config_perm(var, value); + else + repo->settings.shared_repository = PERM_UMASK; + repo->settings.shared_repository_initialized = 1; + } + return repo->settings.shared_repository; +} + +void repo_settings_set_shared_repository(struct repository *repo, int value) +{ + repo->settings.shared_repository = value; + repo->settings.shared_repository_initialized = 1; +} + +void repo_settings_reset_shared_repository(struct repository *repo) +{ + repo->settings.shared_repository_initialized = 0; +} diff --git a/repo-settings.h b/repo-settings.h index 0cef970443be80..ddc11967e015df 100644 --- a/repo-settings.h +++ b/repo-settings.h @@ -37,6 +37,9 @@ struct repo_settings { int pack_use_bitmap_boundary_traversal; int pack_use_multi_pack_reuse; + int shared_repository; + int shared_repository_initialized; + /* * Does this repository have core.useReplaceRefs=true (on by * default)? This provides a repository-scoped version of this @@ -65,6 +68,7 @@ struct repo_settings { char *hooks_path; }; #define REPO_SETTINGS_INIT { \ + .shared_repository = -1, \ .index_version = -1, \ .core_untracked_cache = UNTRACKED_CACHE_KEEP, \ .fetch_negotiation_algorithm = FETCH_NEGOTIATION_CONSECUTIVE, \ @@ -84,4 +88,9 @@ int repo_settings_get_warn_ambiguous_refs(struct repository *repo); /* Read the value for "core.hooksPath". */ const char *repo_settings_get_hooks_path(struct repository *repo); +/* Read, set or reset the value for "core.sharedRepository". */ +int repo_settings_get_shared_repository(struct repository *repo); +void repo_settings_set_shared_repository(struct repository *repo, int value); +void repo_settings_reset_shared_repository(struct repository *repo); + #endif /* REPO_SETTINGS_H */ diff --git a/setup.c b/setup.c index 30889386f761ff..aa65b93f530aa1 100644 --- a/setup.c +++ b/setup.c @@ -2332,7 +2332,7 @@ static int create_default_files(const char *template_path, */ copy_templates(template_path); git_config_clear(); - reset_shared_repository(); + repo_settings_reset_shared_repository(the_repository); git_config(git_default_config, NULL); reinit = is_reinit(); @@ -2342,7 +2342,8 @@ static int create_default_files(const char *template_path, * values we might have just re-read from the config. */ if (init_shared_repository != -1) - set_shared_repository(init_shared_repository); + repo_settings_set_shared_repository(the_repository, + init_shared_repository); is_bare_repository_cfg = !work_tree; @@ -2350,7 +2351,7 @@ static int create_default_files(const char *template_path, * We would have created the above under user's umask -- under * shared-repository settings, we would need to fix them up. */ - if (get_shared_repository()) { + if (repo_settings_get_shared_repository(the_repository)) { adjust_shared_perm(repo_get_git_dir(the_repository)); } @@ -2597,7 +2598,7 @@ int init_db(const char *git_dir, const char *real_git_dir, initial_branch, flags & INIT_DB_QUIET); create_object_directory(); - if (get_shared_repository()) { + if (repo_settings_get_shared_repository(the_repository)) { char buf[10]; /* We do not spell "group" and such, so that * the configuration can be read by older version @@ -2605,12 +2606,12 @@ int init_db(const char *git_dir, const char *real_git_dir, * and compatibility values for PERM_GROUP and * PERM_EVERYBODY. */ - if (get_shared_repository() < 0) + if (repo_settings_get_shared_repository(the_repository) < 0) /* force to the mode value */ - xsnprintf(buf, sizeof(buf), "0%o", -get_shared_repository()); - else if (get_shared_repository() == PERM_GROUP) + xsnprintf(buf, sizeof(buf), "0%o", -repo_settings_get_shared_repository(the_repository)); + else if (repo_settings_get_shared_repository(the_repository) == PERM_GROUP) xsnprintf(buf, sizeof(buf), "%d", OLD_PERM_GROUP); - else if (get_shared_repository() == PERM_EVERYBODY) + else if (repo_settings_get_shared_repository(the_repository) == PERM_EVERYBODY) xsnprintf(buf, sizeof(buf), "%d", OLD_PERM_EVERYBODY); else BUG("invalid value for shared_repository"); @@ -2622,12 +2623,12 @@ int init_db(const char *git_dir, const char *real_git_dir, int len = strlen(git_dir); if (reinit) - printf(get_shared_repository() + printf(repo_settings_get_shared_repository(the_repository) ? _("Reinitialized existing shared Git repository in %s%s\n") : _("Reinitialized existing Git repository in %s%s\n"), git_dir, len && git_dir[len-1] != '/' ? "/" : ""); else - printf(get_shared_repository() + printf(repo_settings_get_shared_repository(the_repository) ? _("Initialized empty shared Git repository in %s%s\n") : _("Initialized empty Git repository in %s%s\n"), git_dir, len && git_dir[len-1] != '/' ? "/" : ""); From 028f618658e34230e1d65678f14b6876e0f9856d Mon Sep 17 00:00:00 2001 From: Patrick Steinhardt Date: Fri, 7 Feb 2025 12:03:41 +0100 Subject: [PATCH 022/312] path: adjust last remaining users of `the_repository` With the preceding refactorings we now only have a couple of implicit users of `the_repository` left in the "path" subsystem, all of which depend on global state via `calc_shared_perm()`. Make the dependency on `the_repository` explicit by passing the repo as a parameter instead and adjust callers accordingly. Note that this change bubbles up into a couple of subsystems that were previously declared as free from `the_repository`. Instead of marking all of them as `the_repository`-dependent again, we instead use the repository that is available in the calling context. There are three exceptions though with "copy.c", "pack-write.c" and "tempfile.c". Adjusting these would require us to adapt callsites all over the place, so this is left for a future iteration. Mark "path.c" as free from `the_repository`. Signed-off-by: Patrick Steinhardt Signed-off-by: Junio C Hamano --- builtin/clone.c | 2 +- commit-graph.c | 2 +- copy.c | 4 +++- loose.c | 2 +- midx-write.c | 2 +- object-file.c | 8 ++++---- pack-bitmap-write.c | 2 +- pack-write.c | 10 ++++++---- path.c | 25 ++++++++++++------------- path.h | 6 +++--- read-cache.c | 2 +- refs/files-backend.c | 10 +++++----- refs/reftable-backend.c | 10 +++++----- server-info.c | 2 +- setup.c | 12 ++++++------ tempfile.c | 4 +++- tmp-objdir.c | 17 ++++++++++------- 17 files changed, 64 insertions(+), 56 deletions(-) diff --git a/builtin/clone.c b/builtin/clone.c index 5ae6ee9db9b4fb..23eeb782aa4fd2 100644 --- a/builtin/clone.c +++ b/builtin/clone.c @@ -1220,7 +1220,7 @@ int cmd_clone(int argc, strbuf_reset(&buf); strbuf_addf(&buf, "%s/refs", git_dir); - safe_create_dir(buf.buf, 1); + safe_create_dir(the_repository, buf.buf, 1); /* * additional config can be injected with -c, make sure it's included diff --git a/commit-graph.c b/commit-graph.c index 2a2999a6b88690..1021ccb983d4ee 100644 --- a/commit-graph.c +++ b/commit-graph.c @@ -2084,7 +2084,7 @@ static int write_commit_graph_file(struct write_commit_graph_context *ctx) return -1; } - if (adjust_shared_perm(get_tempfile_path(graph_layer))) { + if (adjust_shared_perm(the_repository, get_tempfile_path(graph_layer))) { error(_("unable to adjust shared permissions for '%s'"), get_tempfile_path(graph_layer)); return -1; diff --git a/copy.c b/copy.c index d9d20920126905..b668209b6c24fd 100644 --- a/copy.c +++ b/copy.c @@ -1,3 +1,5 @@ +#define USE_THE_REPOSITORY_VARIABLE + #include "git-compat-util.h" #include "copy.h" #include "path.h" @@ -57,7 +59,7 @@ int copy_file(const char *dst, const char *src, int mode) if (close(fdo) != 0) return error_errno("%s: close error", dst); - if (!status && adjust_shared_perm(dst)) + if (!status && adjust_shared_perm(the_repository, dst)) return -1; return status; diff --git a/loose.c b/loose.c index 51ef490f935adc..bb602aaa366bf7 100644 --- a/loose.c +++ b/loose.c @@ -190,7 +190,7 @@ static int write_one_object(struct repository *repo, const struct object_id *oid goto errout; if (close(fd)) goto errout; - adjust_shared_perm(path.buf); + adjust_shared_perm(repo, path.buf); rollback_lock_file(&lock); strbuf_release(&buf); strbuf_release(&path); diff --git a/midx-write.c b/midx-write.c index 61b59d557d3ba4..48d6558253ec93 100644 --- a/midx-write.c +++ b/midx-write.c @@ -1336,7 +1336,7 @@ static int write_midx_internal(struct repository *r, const char *object_dir, return -1; } - if (adjust_shared_perm(get_tempfile_path(incr))) { + if (adjust_shared_perm(r, get_tempfile_path(incr))) { error(_("unable to adjust shared permissions for '%s'"), get_tempfile_path(incr)); return -1; diff --git a/object-file.c b/object-file.c index dc9fcaf3e97380..5d782417e390a5 100644 --- a/object-file.c +++ b/object-file.c @@ -388,7 +388,7 @@ int mkdir_in_gitdir(const char *path) } strbuf_release(&sb); } - return adjust_shared_perm(path); + return adjust_shared_perm(the_repository, path); } static enum scld_error safe_create_leading_directories_1(char *path, int share) @@ -437,7 +437,7 @@ static enum scld_error safe_create_leading_directories_1(char *path, int share) ret = SCLD_VANISHED; else ret = SCLD_FAILED; - } else if (share && adjust_shared_perm(path)) { + } else if (share && adjust_shared_perm(the_repository, path)) { ret = SCLD_PERMS; } *slash = slash_character; @@ -2105,7 +2105,7 @@ int finalize_object_file_flags(const char *tmpfile, const char *filename, } out: - if (adjust_shared_perm(filename)) + if (adjust_shared_perm(the_repository, filename)) return error(_("unable to set permission to '%s'"), filename); return 0; } @@ -2181,7 +2181,7 @@ static int create_tmpfile(struct strbuf *tmp, const char *filename) strbuf_add(tmp, filename, dirlen - 1); if (mkdir(tmp->buf, 0777) && errno != EEXIST) return -1; - if (adjust_shared_perm(tmp->buf)) + if (adjust_shared_perm(the_repository, tmp->buf)) return -1; /* Try again */ diff --git a/pack-bitmap-write.c b/pack-bitmap-write.c index a06a1f35c619b3..34e86d49947d23 100644 --- a/pack-bitmap-write.c +++ b/pack-bitmap-write.c @@ -1072,7 +1072,7 @@ void bitmap_writer_finish(struct bitmap_writer *writer, finalize_hashfile(f, NULL, FSYNC_COMPONENT_PACK_METADATA, CSUM_HASH_IN_STREAM | CSUM_FSYNC | CSUM_CLOSE); - if (adjust_shared_perm(tmp_file.buf)) + if (adjust_shared_perm(the_repository, tmp_file.buf)) die_errno("unable to make temporary bitmap file readable"); if (rename(tmp_file.buf, filename)) diff --git a/pack-write.c b/pack-write.c index a2faeb1895e41f..c3f4e66f02e28b 100644 --- a/pack-write.c +++ b/pack-write.c @@ -1,3 +1,5 @@ +#define USE_THE_REPOSITORY_VARIABLE + #include "git-compat-util.h" #include "environment.h" #include "gettext.h" @@ -287,7 +289,7 @@ char *write_rev_file_order(const struct git_hash_algo *hash_algo, write_rev_index_positions(f, pack_order, nr_objects); write_rev_trailer(hash_algo, f, hash); - if (adjust_shared_perm(path) < 0) + if (adjust_shared_perm(the_repository, path) < 0) die(_("failed to make %s readable"), path); finalize_hashfile(f, NULL, FSYNC_COMPONENT_PACK_METADATA, @@ -350,7 +352,7 @@ static char *write_mtimes_file(const struct git_hash_algo *hash_algo, write_mtimes_objects(f, to_pack, objects, nr_objects); write_mtimes_trailer(hash_algo, f, hash); - if (adjust_shared_perm(mtimes_name) < 0) + if (adjust_shared_perm(the_repository, mtimes_name) < 0) die(_("failed to make %s readable"), mtimes_name); finalize_hashfile(f, NULL, FSYNC_COMPONENT_PACK_METADATA, @@ -565,12 +567,12 @@ void stage_tmp_packfiles(const struct git_hash_algo *hash_algo, char *rev_tmp_name = NULL; char *mtimes_tmp_name = NULL; - if (adjust_shared_perm(pack_tmp_name)) + if (adjust_shared_perm(the_repository, pack_tmp_name)) die_errno("unable to make temporary pack file readable"); *idx_tmp_name = (char *)write_idx_file(hash_algo, NULL, written_list, nr_written, pack_idx_opts, hash); - if (adjust_shared_perm(*idx_tmp_name)) + if (adjust_shared_perm(the_repository, *idx_tmp_name)) die_errno("unable to make temporary index file readable"); rev_tmp_name = write_rev_file(hash_algo, NULL, written_list, nr_written, diff --git a/path.c b/path.c index a2f402baecf21a..910756c8b3249c 100644 --- a/path.c +++ b/path.c @@ -2,8 +2,6 @@ * Utilities for paths and pathnames */ -#define USE_THE_REPOSITORY_VARIABLE - #include "git-compat-util.h" #include "abspath.h" #include "environment.h" @@ -840,21 +838,22 @@ const char *enter_repo(const char *path, unsigned flags) return NULL; } -int calc_shared_perm(int mode) +int calc_shared_perm(struct repository *repo, + int mode) { int tweak; - if (repo_settings_get_shared_repository(the_repository) < 0) - tweak = -repo_settings_get_shared_repository(the_repository); + if (repo_settings_get_shared_repository(repo) < 0) + tweak = -repo_settings_get_shared_repository(repo); else - tweak = repo_settings_get_shared_repository(the_repository); + tweak = repo_settings_get_shared_repository(repo); if (!(mode & S_IWUSR)) tweak &= ~0222; if (mode & S_IXUSR) /* Copy read bits to execute bits */ tweak |= (tweak & 0444) >> 2; - if (repo_settings_get_shared_repository(the_repository) < 0) + if (repo_settings_get_shared_repository(repo) < 0) mode = (mode & ~0777) | tweak; else mode |= tweak; @@ -862,17 +861,17 @@ int calc_shared_perm(int mode) return mode; } - -int adjust_shared_perm(const char *path) +int adjust_shared_perm(struct repository *repo, + const char *path) { int old_mode, new_mode; - if (!repo_settings_get_shared_repository(the_repository)) + if (!repo_settings_get_shared_repository(repo)) return 0; if (get_st_mode_bits(path, &old_mode) < 0) return -1; - new_mode = calc_shared_perm(old_mode); + new_mode = calc_shared_perm(repo, old_mode); if (S_ISDIR(old_mode)) { /* Copy read bits to execute bits */ new_mode |= (new_mode & 0444) >> 2; @@ -891,7 +890,7 @@ int adjust_shared_perm(const char *path) return 0; } -void safe_create_dir(const char *dir, int share) +void safe_create_dir(struct repository *repo, const char *dir, int share) { if (mkdir(dir, 0777) < 0) { if (errno != EEXIST) { @@ -899,7 +898,7 @@ void safe_create_dir(const char *dir, int share) exit(1); } } - else if (share && adjust_shared_perm(dir)) + else if (share && adjust_shared_perm(repo, dir)) die(_("Could not make %s writable by group"), dir); } diff --git a/path.h b/path.h index 373404dd9d2a5f..65fe968a13a191 100644 --- a/path.h +++ b/path.h @@ -141,8 +141,8 @@ const char *git_path_shallow(struct repository *r); int ends_with_path_components(const char *path, const char *components); -int calc_shared_perm(int mode); -int adjust_shared_perm(const char *path); +int calc_shared_perm(struct repository *repo, int mode); +int adjust_shared_perm(struct repository *repo, const char *path); char *interpolate_path(const char *path, int real_home); @@ -219,7 +219,7 @@ char *xdg_cache_home(const char *filename); * directories under $GIT_DIR. Don't use it for working tree * directories. */ -void safe_create_dir(const char *dir, int share); +void safe_create_dir(struct repository *repo, const char *dir, int share); # ifdef USE_THE_REPOSITORY_VARIABLE # include "strbuf.h" diff --git a/read-cache.c b/read-cache.c index 66ad0015a7f75a..900738e7a8d6aa 100644 --- a/read-cache.c +++ b/read-cache.c @@ -3290,7 +3290,7 @@ static int write_shared_index(struct index_state *istate, if (ret) return ret; - ret = adjust_shared_perm(get_tempfile_path(*temp)); + ret = adjust_shared_perm(the_repository, get_tempfile_path(*temp)); if (ret) { error(_("cannot fix permission bits on '%s'"), get_tempfile_path(*temp)); return ret; diff --git a/refs/files-backend.c b/refs/files-backend.c index 29f08dced40418..6c6e67dc1c48ed 100644 --- a/refs/files-backend.c +++ b/refs/files-backend.c @@ -1831,7 +1831,7 @@ static int log_ref_setup(struct files_ref_store *refs, } if (*logfd >= 0) - adjust_shared_perm(logfile); + adjust_shared_perm(the_repository, logfile); free(logfile); return 0; @@ -3488,8 +3488,8 @@ static int files_ref_store_create_on_disk(struct ref_store *ref_store, * they do not understand the reference format extension. */ strbuf_addf(&sb, "%s/refs", ref_store->gitdir); - safe_create_dir(sb.buf, 1); - adjust_shared_perm(sb.buf); + safe_create_dir(the_repository, sb.buf, 1); + adjust_shared_perm(the_repository, sb.buf); /* * There is no need to create directories for common refs when creating @@ -3501,11 +3501,11 @@ static int files_ref_store_create_on_disk(struct ref_store *ref_store, */ strbuf_reset(&sb); files_ref_path(refs, &sb, "refs/heads"); - safe_create_dir(sb.buf, 1); + safe_create_dir(the_repository, sb.buf, 1); strbuf_reset(&sb); files_ref_path(refs, &sb, "refs/tags"); - safe_create_dir(sb.buf, 1); + safe_create_dir(the_repository, sb.buf, 1); } strbuf_release(&sb); diff --git a/refs/reftable-backend.c b/refs/reftable-backend.c index d39a14c5a469d7..9c54e2c1737164 100644 --- a/refs/reftable-backend.c +++ b/refs/reftable-backend.c @@ -380,7 +380,7 @@ static struct ref_store *reftable_be_init(struct repository *repo, default: BUG("unknown hash algorithm %d", repo->hash_algo->format_id); } - refs->write_options.default_permissions = calc_shared_perm(0666 & ~mask); + refs->write_options.default_permissions = calc_shared_perm(the_repository, 0666 & ~mask); refs->write_options.disable_auto_compact = !git_env_bool("GIT_TEST_REFTABLE_AUTOCOMPACTION", 1); refs->write_options.lock_timeout_ms = 100; @@ -470,21 +470,21 @@ static int reftable_be_create_on_disk(struct ref_store *ref_store, struct strbuf sb = STRBUF_INIT; strbuf_addf(&sb, "%s/reftable", refs->base.gitdir); - safe_create_dir(sb.buf, 1); + safe_create_dir(the_repository, sb.buf, 1); strbuf_reset(&sb); strbuf_addf(&sb, "%s/HEAD", refs->base.gitdir); write_file(sb.buf, "ref: refs/heads/.invalid"); - adjust_shared_perm(sb.buf); + adjust_shared_perm(the_repository, sb.buf); strbuf_reset(&sb); strbuf_addf(&sb, "%s/refs", refs->base.gitdir); - safe_create_dir(sb.buf, 1); + safe_create_dir(the_repository, sb.buf, 1); strbuf_reset(&sb); strbuf_addf(&sb, "%s/refs/heads", refs->base.gitdir); write_file(sb.buf, "this repository uses the reftable format"); - adjust_shared_perm(sb.buf); + adjust_shared_perm(the_repository, sb.buf); strbuf_release(&sb); return 0; diff --git a/server-info.c b/server-info.c index 31c3fdc118447d..1ca0e00d51e6c0 100644 --- a/server-info.c +++ b/server-info.c @@ -125,7 +125,7 @@ static int update_info_file(struct repository *r, char *path, uic.cur_fp = NULL; if (uic_is_stale(&uic)) { - if (adjust_shared_perm(get_tempfile_path(f)) < 0) + if (adjust_shared_perm(r, get_tempfile_path(f)) < 0) goto out; if (rename_tempfile(&f, path) < 0) goto out; diff --git a/setup.c b/setup.c index aa65b93f530aa1..71a3d66f05da30 100644 --- a/setup.c +++ b/setup.c @@ -2088,7 +2088,7 @@ static void copy_templates_1(struct strbuf *path, struct strbuf *template_path, * with the way the namespace under .git/ is organized, should * be really carefully chosen. */ - safe_create_dir(path->buf, 1); + safe_create_dir(the_repository, path->buf, 1); while ((de = readdir(dir)) != NULL) { struct stat st_git, st_template; int exists = 0; @@ -2352,7 +2352,7 @@ static int create_default_files(const char *template_path, * shared-repository settings, we would need to fix them up. */ if (repo_settings_get_shared_repository(the_repository)) { - adjust_shared_perm(repo_get_git_dir(the_repository)); + adjust_shared_perm(the_repository, repo_get_git_dir(the_repository)); } initialize_repository_version(fmt->hash_algo, fmt->ref_storage_format, reinit); @@ -2413,15 +2413,15 @@ static void create_object_directory(void) strbuf_addstr(&path, repo_get_object_directory(the_repository)); baselen = path.len; - safe_create_dir(path.buf, 1); + safe_create_dir(the_repository, path.buf, 1); strbuf_setlen(&path, baselen); strbuf_addstr(&path, "/pack"); - safe_create_dir(path.buf, 1); + safe_create_dir(the_repository, path.buf, 1); strbuf_setlen(&path, baselen); strbuf_addstr(&path, "/info"); - safe_create_dir(path.buf, 1); + safe_create_dir(the_repository, path.buf, 1); strbuf_release(&path); } @@ -2588,7 +2588,7 @@ int init_db(const char *git_dir, const char *real_git_dir, */ git_config(platform_core_config, NULL); - safe_create_dir(git_dir, 0); + safe_create_dir(the_repository, git_dir, 0); reinit = create_default_files(template_dir, original_git_dir, &repo_fmt, init_shared_repository); diff --git a/tempfile.c b/tempfile.c index ed88cf84314753..82dfa3d82f2ac9 100644 --- a/tempfile.c +++ b/tempfile.c @@ -42,6 +42,8 @@ * file created by its parent. */ +#define USE_THE_REPOSITORY_VARIABLE + #include "git-compat-util.h" #include "abspath.h" #include "path.h" @@ -148,7 +150,7 @@ struct tempfile *create_tempfile_mode(const char *path, int mode) return NULL; } activate_tempfile(tempfile); - if (adjust_shared_perm(tempfile->filename.buf)) { + if (adjust_shared_perm(the_repository, tempfile->filename.buf)) { int save_errno = errno; error("cannot fix permission bits on %s", tempfile->filename.buf); delete_tempfile(&tempfile); diff --git a/tmp-objdir.c b/tmp-objdir.c index 0ea078a5c5f4eb..31d16a4c2c576e 100644 --- a/tmp-objdir.c +++ b/tmp-objdir.c @@ -207,10 +207,12 @@ static int read_dir_paths(struct string_list *out, const char *path) return 0; } -static int migrate_paths(struct strbuf *src, struct strbuf *dst, +static int migrate_paths(struct tmp_objdir *t, + struct strbuf *src, struct strbuf *dst, enum finalize_object_file_flags flags); -static int migrate_one(struct strbuf *src, struct strbuf *dst, +static int migrate_one(struct tmp_objdir *t, + struct strbuf *src, struct strbuf *dst, enum finalize_object_file_flags flags) { struct stat st; @@ -219,11 +221,11 @@ static int migrate_one(struct strbuf *src, struct strbuf *dst, return -1; if (S_ISDIR(st.st_mode)) { if (!mkdir(dst->buf, 0777)) { - if (adjust_shared_perm(dst->buf)) + if (adjust_shared_perm(t->repo, dst->buf)) return -1; } else if (errno != EEXIST) return -1; - return migrate_paths(src, dst, flags); + return migrate_paths(t, src, dst, flags); } return finalize_object_file_flags(src->buf, dst->buf, flags); } @@ -233,7 +235,8 @@ static int is_loose_object_shard(const char *name) return strlen(name) == 2 && isxdigit(name[0]) && isxdigit(name[1]); } -static int migrate_paths(struct strbuf *src, struct strbuf *dst, +static int migrate_paths(struct tmp_objdir *t, + struct strbuf *src, struct strbuf *dst, enum finalize_object_file_flags flags) { size_t src_len = src->len, dst_len = dst->len; @@ -255,7 +258,7 @@ static int migrate_paths(struct strbuf *src, struct strbuf *dst, if (is_loose_object_shard(name)) flags_copy |= FOF_SKIP_COLLISION_CHECK; - ret |= migrate_one(src, dst, flags_copy); + ret |= migrate_one(t, src, dst, flags_copy); strbuf_setlen(src, src_len); strbuf_setlen(dst, dst_len); @@ -283,7 +286,7 @@ int tmp_objdir_migrate(struct tmp_objdir *t) strbuf_addbuf(&src, &t->path); strbuf_addstr(&dst, repo_get_object_directory(t->repo)); - ret = migrate_paths(&src, &dst, 0); + ret = migrate_paths(t, &src, &dst, 0); strbuf_release(&src); strbuf_release(&dst); From 41c793eae900e632dbc73ec7f900ea1e7d38aa92 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Sat, 1 Mar 2025 10:36:02 -0500 Subject: [PATCH 023/312] doc: update howto-index.sh for .adoc extensions The .txt extensions were changed to .adoc in 1f010d6bdf (doc: use .adoc extension for AsciiDoc files, 2025-01-20). This left broken links in the generated howto-index.html. Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- Documentation/howto/howto-index.sh | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) diff --git a/Documentation/howto/howto-index.sh b/Documentation/howto/howto-index.sh index eecd123a936079..ace49830a8234f 100755 --- a/Documentation/howto/howto-index.sh +++ b/Documentation/howto/howto-index.sh @@ -9,9 +9,9 @@ people describing how they use Git in their workflow. EOF -for txt +for adoc do - title=$(expr "$txt" : '.*/\(.*\)\.txt$') + title=$(expr "$adoc" : '.*/\(.*\)\.adoc$') from=$(sed -ne ' /^$/q /^From:[ ]/{ @@ -21,7 +21,7 @@ do s/^/by / p } - ' "$txt") + ' "$adoc") abstract=$(sed -ne ' /^Abstract:[ ]/{ @@ -39,13 +39,13 @@ do x p q - }' "$txt") + }' "$adoc") - if grep 'Content-type: text/asciidoc' >/dev/null $txt + if grep 'Content-type: text/asciidoc' >/dev/null $adoc then - file=$(expr "$txt" : '\(.*\)\.txt$').html + file=$(expr "$adoc" : '\(.*\)\.adoc$').html else - file="$txt" + file="$adoc" fi echo "* link:howto/$(basename "$file")[$title] $from From fa779fa88d18f550421ae777f7e048dd08ad0006 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Sat, 1 Mar 2025 10:36:03 -0500 Subject: [PATCH 024/312] contrib/contacts: rename .txt to .adoc The .txt extensions were changed to .adoc in 1f010d6bdf (doc: use .adoc extension for AsciiDoc files, 2025-01-20). Do the same for contrib/contacts. Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- contrib/contacts/Makefile | 2 +- contrib/contacts/{git-contacts.txt => git-contacts.adoc} | 0 contrib/contacts/meson.build | 4 ++-- 3 files changed, 3 insertions(+), 3 deletions(-) rename contrib/contacts/{git-contacts.txt => git-contacts.adoc} (100%) diff --git a/contrib/contacts/Makefile b/contrib/contacts/Makefile index a2990f0dcb53b1..9c4ca4f3bcc7c2 100644 --- a/contrib/contacts/Makefile +++ b/contrib/contacts/Makefile @@ -34,7 +34,7 @@ GIT_CONTACTS := git-contacts GIT_CONTACTS_DOC := git-contacts.1 GIT_CONTACTS_XML := git-contacts.xml -GIT_CONTACTS_TXT := git-contacts.txt +GIT_CONTACTS_TXT := git-contacts.adoc GIT_CONTACTS_HTML := git-contacts.html doc: $(GIT_CONTACTS_DOC) $(GIT_CONTACTS_HTML) diff --git a/contrib/contacts/git-contacts.txt b/contrib/contacts/git-contacts.adoc similarity index 100% rename from contrib/contacts/git-contacts.txt rename to contrib/contacts/git-contacts.adoc diff --git a/contrib/contacts/meson.build b/contrib/contacts/meson.build index 6ec92f47c43994..73d82dfe52b85f 100644 --- a/contrib/contacts/meson.build +++ b/contrib/contacts/meson.build @@ -16,7 +16,7 @@ if get_option('docs').contains('man') '@INPUT@', ], depends: documentation_deps, - input: 'git-contacts.txt', + input: 'git-contacts.adoc', output: 'git-contacts.xml', ) @@ -47,7 +47,7 @@ if get_option('docs').contains('html') '@INPUT@', ], depends: documentation_deps, - input: 'git-contacts.txt', + input: 'git-contacts.adoc', output: 'git-contacts.html', install: true, install_dir: get_option('datadir') / 'doc/git-doc', From 561de07b57fcc2057fb6b96043f008a5b83f140c Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Sat, 1 Mar 2025 10:36:04 -0500 Subject: [PATCH 025/312] contrib/subtree: rename .txt to .adoc The .txt extensions were changed to .adoc in 1f010d6bdf (doc: use .adoc extension for AsciiDoc files, 2025-01-20). Do the same for contrib/subtree. Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- contrib/subtree/Makefile | 2 +- contrib/subtree/{git-subtree.txt => git-subtree.adoc} | 0 contrib/subtree/meson.build | 4 ++-- 3 files changed, 3 insertions(+), 3 deletions(-) rename contrib/subtree/{git-subtree.txt => git-subtree.adoc} (100%) diff --git a/contrib/subtree/Makefile b/contrib/subtree/Makefile index 8fe0bfd4013fe8..c0c9f21cb78022 100644 --- a/contrib/subtree/Makefile +++ b/contrib/subtree/Makefile @@ -50,7 +50,7 @@ GIT_SUBTREE := git-subtree GIT_SUBTREE_DOC := git-subtree.1 GIT_SUBTREE_XML := git-subtree.xml -GIT_SUBTREE_TXT := git-subtree.txt +GIT_SUBTREE_TXT := git-subtree.adoc GIT_SUBTREE_HTML := git-subtree.html GIT_SUBTREE_TEST := ../../git-subtree diff --git a/contrib/subtree/git-subtree.txt b/contrib/subtree/git-subtree.adoc similarity index 100% rename from contrib/subtree/git-subtree.txt rename to contrib/subtree/git-subtree.adoc diff --git a/contrib/subtree/meson.build b/contrib/subtree/meson.build index a752a188df1894..9c72b2362595a6 100644 --- a/contrib/subtree/meson.build +++ b/contrib/subtree/meson.build @@ -32,7 +32,7 @@ if get_option('docs').contains('man') '@INPUT@', ], depends: documentation_deps, - input: 'git-subtree.txt', + input: 'git-subtree.adoc', output: 'git-subtree.xml', ) @@ -63,7 +63,7 @@ if get_option('docs').contains('html') '@INPUT@', ], depends: documentation_deps, - input: 'git-subtree.txt', + input: 'git-subtree.adoc', output: 'git-subtree.html', install: true, install_dir: get_option('datadir') / 'doc/git-doc', From 06d9252bccc39c335fa5ab36adf724df4570248c Mon Sep 17 00:00:00 2001 From: Junio C Hamano Date: Sat, 1 Mar 2025 10:25:10 -0800 Subject: [PATCH 026/312] doc: fix build-docdep.perl We renamed from .txt to .adoc all the asciidoc source files and necessary includes. We also need to adjust the build-docdep tool to work on files whose suffix is .adoc when computing the documentation dependencies. Signed-off-by: Junio C Hamano --- Documentation/build-docdep.perl | 24 ++++++++++++------------ 1 file changed, 12 insertions(+), 12 deletions(-) diff --git a/Documentation/build-docdep.perl b/Documentation/build-docdep.perl index 315efaa2fa2f2e..781da12b2ee8c9 100755 --- a/Documentation/build-docdep.perl +++ b/Documentation/build-docdep.perl @@ -4,15 +4,15 @@ my %include = (); my %included = (); -for my $text (<*.txt>) { - open I, '<', $text || die "cannot read: $text"; +for my $adoc (<*.adoc>) { + open I, '<', $adoc || die "cannot read: $adoc"; while () { if (/^include::/) { chomp; s/^include::\s*//; s/\[\]//; s/{build_dir}/${build_dir}/; - $include{$text}{$_} = 1; + $include{$adoc}{$_} = 1; $included{$_} = 1; } } @@ -23,14 +23,14 @@ my $changed = 1; while ($changed) { $changed = 0; - while (my ($text, $included) = each %include) { + while (my ($adoc, $included) = each %include) { for my $i (keys %$included) { - # $text has include::$i; if $i includes $j - # $text indirectly includes $j. + # $adoc has include::$i; if $i includes $j + # $adoc indirectly includes $j. if (exists $include{$i}) { for my $j (keys %{$include{$i}}) { - if (!exists $include{$text}{$j}) { - $include{$text}{$j} = 1; + if (!exists $include{$adoc}{$j}) { + $include{$adoc}{$j} = 1; $included{$j} = 1; $changed = 1; } @@ -40,10 +40,10 @@ } } -foreach my $text (sort keys %include) { - my $included = $include{$text}; - if (! exists $included{$text} && - (my $base = $text) =~ s/\.txt$//) { +foreach my $adoc (sort keys %include) { + my $included = $include{$adoc}; + if (! exists $included{$adoc} && + (my $base = $adoc) =~ s/\.adoc$//) { print "$base.html $base.xml : ", join(" ", sort keys %$included), "\n"; } } From 87eccc3a81d133e636150268c24d2220cad70fe5 Mon Sep 17 00:00:00 2001 From: Phillip Wood Date: Sun, 2 Mar 2025 16:02:30 +0000 Subject: [PATCH 027/312] meson: fix building technical and howto docs When our asciidoc files were renamed from "*.txt" to "*.adoc" in 1f010d6bdf7 (doc: use .adoc extension for AsciiDoc files, 2025-01-20) the "meson.build" file in "Documentation" was updated but the "meson.build" files in the "technical" and "howto" subdirectories were not. This causes the meson build to fail when configured with -Ddocs=html. Fix this by updating the relevant "meson.build" files. Signed-off-by: Phillip Wood Acked-by: Patrick Steinhardt Signed-off-by: Junio C Hamano --- Documentation/howto/meson.build | 34 ++++++++-------- Documentation/technical/meson.build | 62 ++++++++++++++--------------- 2 files changed, 48 insertions(+), 48 deletions(-) diff --git a/Documentation/howto/meson.build b/Documentation/howto/meson.build index c023c104161e61..961e0533629fa0 100644 --- a/Documentation/howto/meson.build +++ b/Documentation/howto/meson.build @@ -1,20 +1,20 @@ howto_sources = [ - 'coordinate-embargoed-releases.txt', - 'keep-canonical-history-correct.txt', - 'maintain-git.txt', - 'new-command.txt', - 'rebase-from-internal-branch.txt', - 'rebuild-from-update-hook.txt', - 'recover-corrupted-blob-object.txt', - 'recover-corrupted-object-harder.txt', - 'revert-a-faulty-merge.txt', - 'revert-branch-rebase.txt', - 'separating-topic-branches.txt', - 'setup-git-server-over-http.txt', - 'update-hook-example.txt', - 'use-git-daemon.txt', - 'using-merge-subtree.txt', - 'using-signed-tag-in-pull-request.txt', + 'coordinate-embargoed-releases.adoc', + 'keep-canonical-history-correct.adoc', + 'maintain-git.adoc', + 'new-command.adoc', + 'rebase-from-internal-branch.adoc', + 'rebuild-from-update-hook.adoc', + 'recover-corrupted-blob-object.adoc', + 'recover-corrupted-object-harder.adoc', + 'revert-a-faulty-merge.adoc', + 'revert-branch-rebase.adoc', + 'separating-topic-branches.adoc', + 'setup-git-server-over-http.adoc', + 'update-hook-example.adoc', + 'use-git-daemon.adoc', + 'using-merge-subtree.adoc', + 'using-signed-tag-in-pull-request.adoc', ] howto_index = custom_target( @@ -26,7 +26,7 @@ howto_index = custom_target( env: script_environment, capture: true, input: howto_sources, - output: 'howto-index.txt', + output: 'howto-index.adoc', ) custom_target( diff --git a/Documentation/technical/meson.build b/Documentation/technical/meson.build index 3a65ee59b3cd2c..a13aafcfbb8c75 100644 --- a/Documentation/technical/meson.build +++ b/Documentation/technical/meson.build @@ -1,37 +1,37 @@ api_docs = [ - 'api-error-handling.txt', - 'api-merge.txt', - 'api-parse-options.txt', - 'api-simple-ipc.txt', - 'api-trace2.txt', + 'api-error-handling.adoc', + 'api-merge.adoc', + 'api-parse-options.adoc', + 'api-simple-ipc.adoc', + 'api-trace2.adoc', ] articles = [ - 'bitmap-format.txt', - 'build-systems.txt', - 'bundle-uri.txt', - 'commit-graph.txt', - 'directory-rename-detection.txt', - 'hash-function-transition.txt', - 'long-running-process-protocol.txt', - 'multi-pack-index.txt', - 'packfile-uri.txt', - 'pack-heuristics.txt', - 'parallel-checkout.txt', - 'partial-clone.txt', - 'platform-support.txt', - 'racy-git.txt', - 'reftable.txt', - 'remembering-renames.txt', - 'repository-version.txt', - 'rerere.txt', - 'scalar.txt', - 'send-pack-pipeline.txt', - 'shallow.txt', - 'sparse-checkout.txt', - 'sparse-index.txt', - 'trivial-merge.txt', - 'unit-tests.txt', + 'bitmap-format.adoc', + 'build-systems.adoc', + 'bundle-uri.adoc', + 'commit-graph.adoc', + 'directory-rename-detection.adoc', + 'hash-function-transition.adoc', + 'long-running-process-protocol.adoc', + 'multi-pack-index.adoc', + 'packfile-uri.adoc', + 'pack-heuristics.adoc', + 'parallel-checkout.adoc', + 'partial-clone.adoc', + 'platform-support.adoc', + 'racy-git.adoc', + 'reftable.adoc', + 'remembering-renames.adoc', + 'repository-version.adoc', + 'rerere.adoc', + 'scalar.adoc', + 'send-pack-pipeline.adoc', + 'shallow.adoc', + 'sparse-checkout.adoc', + 'sparse-index.adoc', + 'trivial-merge.adoc', + 'unit-tests.adoc', ] api_index = custom_target( @@ -43,7 +43,7 @@ api_index = custom_target( ], env: script_environment, input: api_docs, - output: 'api-index.txt', + output: 'api-index.adoc', ) custom_target( From c84209a8fda86dd1c7647f52bf0fd6eb8aaaaad2 Mon Sep 17 00:00:00 2001 From: David Mandelberg Date: Sun, 2 Mar 2025 15:54:44 -0500 Subject: [PATCH 028/312] editorconfig: add .bash extension Both files in the command below appear to be indented with tabs, and I'd expect .bash files to have roughly the same style as .sh files. $ find . -name \*.bash ./contrib/completion/git-completion.bash ./ci/check-directional-formatting.bash Signed-off-by: David Mandelberg Signed-off-by: Junio C Hamano --- .editorconfig | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/.editorconfig b/.editorconfig index a3c578a43cee57..2d3929b5916a14 100644 --- a/.editorconfig +++ b/.editorconfig @@ -4,7 +4,7 @@ insert_final_newline = true # The settings for C (*.c and *.h) files are mirrored in .clang-format. Keep # them in sync. -[{*.{c,h,sh,perl,pl,pm,txt,adoc},config.mak.*,Makefile}] +[{*.{c,h,sh,bash,perl,pl,pm,txt,adoc},config.mak.*,Makefile}] indent_style = tab tab_width = 8 From 5040f9f164ea7d633237e52bed311d1e9773c78e Mon Sep 17 00:00:00 2001 From: Christian Couder Date: Tue, 18 Feb 2025 12:32:04 +0100 Subject: [PATCH 029/312] doc: add technical design doc for large object promisors Let's add a design doc about how we could improve handling liarge blobs using "Large Object Promisors" (LOPs). It's a set of features with the goal of using special dedicated promisor remotes to store large blobs, and having them accessed directly by main remotes and clients. Signed-off-by: Christian Couder Signed-off-by: Junio C Hamano --- .../technical/large-object-promisors.txt | 656 ++++++++++++++++++ 1 file changed, 656 insertions(+) create mode 100644 Documentation/technical/large-object-promisors.txt diff --git a/Documentation/technical/large-object-promisors.txt b/Documentation/technical/large-object-promisors.txt new file mode 100644 index 00000000000000..dea8dafa669908 --- /dev/null +++ b/Documentation/technical/large-object-promisors.txt @@ -0,0 +1,656 @@ +Large Object Promisors +====================== + +Since Git has been created, users have been complaining about issues +with storing large files in Git. Some solutions have been created to +help, but they haven't helped much with some issues. + +Git currently supports multiple promisor remotes, which could help +with some of these remaining issues, but it's very hard to use them to +help, because a number of important features are missing. + +The goal of the effort described in this document is to add these +important features. + +We will call a "Large Object Promisor", or "LOP" in short, a promisor +remote which is used to store only large blobs and which is separate +from the main remote that should store the other Git objects and the +rest of the repos. + +By extension, we will also call "Large Object Promisor", or LOP, the +effort described in this document to add a set of features to make it +easier to handle large blobs/files in Git by using LOPs. + +This effort aims to especially improve things on the server side, and +especially for large blobs that are already compressed in a binary +format. + +This effort aims to provide an alternative to Git LFS +(https://git-lfs.com/) and similar tools like git-annex +(https://git-annex.branchable.com/) for handling large files, even +though a complete alternative would very likely require other efforts +especially on the client side, where it would likely help to implement +a new object representation for large blobs as discussed in: + +https://lore.kernel.org/git/xmqqbkdometi.fsf@gitster.g/ + +0) Non goals +------------ + +- We will not discuss those client side improvements here, as they + would require changes in different parts of Git than this effort. ++ +So we don't pretend to fully replace Git LFS with only this effort, +but we nevertheless believe that it can significantly improve the +current situation on the server side, and that other separate +efforts could also improve the situation on the client side. + +- In the same way, we are not going to discuss all the possible ways + to implement a LOP or their underlying object storage, or to + optimize how LOP works. ++ +Our opinion is that the simplest solution for now is for LOPs to use +object storage through a remote helper (see section II.2 below for +more details) to store their objects. So we consider that this is the +default implementation. If there are improvements on top of this, +that's great, but our opinion is that such improvements are not +necessary for LOPs to already be useful. Such improvements are likely +a different technical topic, and can be taken care of separately +anyway. ++ +So in particular we are not going to discuss pluggable ODBs or other +object database backends that could chunk large blobs, dedup the +chunks and store them efficiently. Sure, that would be a nice +improvement to store large blobs on the server side, but we believe +it can just be a separate effort as it's also not technically very +related to this effort. ++ +We are also not going to discuss data transfer improvements between +LOPs and clients or servers. Sure, there might be some easy and very +effective optimizations there (as we know that objects on LOPs are +very likely incompressible and not deltifying well), but this can be +dealt with separately in a separate effort. + +In other words, the goal of this document is not to talk about all the +possible ways to optimize how Git could handle large blobs, but to +describe how a LOP based solution can already work well and alleviate +a number of current issues in the context of Git clients and servers +sharing Git objects. + +Even if LOPs are used not very efficiently, they can still be useful +and worth using in some cases, as we will see in more details +later in this document: + + - they can make it simpler for clients to use promisor remotes and + therefore avoid fetching a lot of large blobs they might not need + locally, + + - they can make it significantly cheaper or easier for servers to + host a significant part of the current repository content, and + even more to host content with larger blobs or more large blobs + than currently. + +I) Issues with the current situation +------------------------------------ + +- Some statistics made on GitLab repos have shown that more than 75% + of the disk space is used by blobs that are larger than 1MB and + often in a binary format. + +- So even if users could use Git LFS or similar tools to store a lot + of large blobs out of their repos, it's a fact that in practice they + don't do it as much as they probably should. + +- On the server side ideally, the server should be able to decide for + itself how it stores things. It should not depend on users deciding + to use tools like Git LFS on some blobs or not. + +- It's much more expensive to store large blobs that don't delta + compress well on regular fast seeking drives (like SSDs) than on + object storage (like Amazon S3 or GCP Buckets). Using fast drives + for regular Git repos makes sense though, as serving regular Git + content (blobs containing text or code) needs drives where seeking + is fast, but the content is relatively small. On the other hand, + object storage for Git LFS blobs makes sense as seeking speed is not + as important when dealing with large files, while costs are more + important. So the fact that users don't use Git LFS or similar tools + for a significant number of large blobs has likely some bad + consequences on the cost of repo storage for most Git hosting + platforms. + +- Having large blobs handled in the same way as other blobs and Git + objects in Git repos instead of on object storage also has a cost in + increased memory and CPU usage, and therefore decreased performance, + when creating packfiles. (This is because Git tries to use delta + compression or zlib compression which is unlikely to work well on + already compressed binary content.) So it's not just a storage cost + increase. + +- When a large blob has been committed into a repo, it might not be + possible to remove this blob from the repo without rewriting + history, even if the user then decides to use Git LFS or a similar + tool to handle it. + +- In fact Git LFS and similar tools are not very flexible in letting + users change their minds about the blobs they should handle or not. + +- Even when users are using Git LFS or similar tools, they are often + complaining that these tools require significant effort to set up, + learn and use correctly. + +II) Main features of the "Large Object Promisors" solution +---------------------------------------------------------- + +The main features below should give a rough overview of how the +solution may work. Details about needed elements can be found in +following sections. + +Even if each feature below is very useful for the full solution, it is +very likely to be also useful on its own in some cases where the full +solution is not required. However, we'll focus primarily on the big +picture here. + +Also each feature doesn't need to be implemented entirely in Git +itself. Some could be scripts, hooks or helpers that are not part of +the Git repo. It would be helpful if those could be shared and +improved on collaboratively though. So we want to encourage sharing +them. + +1) Large blobs are stored on LOPs +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Large blobs should be stored on special promisor remotes that we will +call "Large Object Promisors" or LOPs. These LOPs should be additional +remotes dedicated to contain large blobs especially those in binary +format. They should be used along with main remotes that contain the +other objects. + +Note 1 +++++++ + +To clarify, a LOP is a normal promisor remote, except that: + +- it should store only large blobs, + +- it should be separate from the main remote, so that the main remote + can focus on serving other objects and the rest of the repos (see + feature 4) below) and can use the LOP as a promisor remote for + itself. + +Note 2 +++++++ + +Git already makes it possible for a main remote to also be a promisor +remote storing both regular objects and large blobs for a client that +clones from it with a filter on blob size. But here we explicitly want +to avoid that. + +Rationale ++++++++++ + +LOPs aim to be good at handling large blobs while main remotes are +already good at handling other objects. + +Implementation +++++++++++++++ + +Git already has support for multiple promisor remotes, see +link:partial-clone.html#using-many-promisor-remotes[the partial clone documentation]. + +Also, Git already has support for partial clone using a filter on the +size of the blobs (with `git clone --filter=blob:limit=`). Most +of the other main features below are based on these existing features +and are about making them easy and efficient to use for the purpose of +better handling large blobs. + +2) LOPs can use object storage +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +LOPs can be implemented using object storage, like an Amazon S3 or GCP +Bucket or MinIO (which is open source under the GNU AGPLv3 license) to +actually store the large blobs, and can be accessed through a Git +remote helper (see linkgit:gitremote-helpers[7]) which makes the +underlying object storage appear like a remote to Git. + +Note +++++ + +A LOP can be a promisor remote accessed using a remote helper by +both some clients and the main remote. + +Rationale ++++++++++ + +This looks like the simplest way to create LOPs that can cheaply +handle many large blobs. + +Implementation +++++++++++++++ + +Remote helpers are quite easy to write as shell scripts, but it might +be more efficient and maintainable to write them using other languages +like Go. + +Some already exist under open source licenses, for example: + + - https://github.com/awslabs/git-remote-s3 + - https://gitlab.com/eric.p.ju/git-remote-gs + +Other ways to implement LOPs are certainly possible, but the goal of +this document is not to discuss how to best implement a LOP or its +underlying object storage (see the "0) Non goals" section above). + +3) LOP object storage can be Git LFS storage +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +The underlying object storage that a LOP uses could also serve as +storage for large files handled by Git LFS. + +Rationale ++++++++++ + +This would simplify the server side if it wants to both use a LOP and +act as a Git LFS server. + +4) A main remote can offload to a LOP with a configurable threshold +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +On the server side, a main remote should have a way to offload to a +LOP all its blobs with a size over a configurable threshold. + +Rationale ++++++++++ + +This makes it easy to set things up and to clean things up. For +example, an admin could use this to manually convert a repo not using +LOPs to a repo using a LOP. On a repo already using a LOP but where +some users would sometimes push large blobs, a cron job could use this +to regularly make sure the large blobs are moved to the LOP. + +Implementation +++++++++++++++ + +Using something based on `git repack --filter=...` to separate the +blobs we want to offload from the other Git objects could be a good +idea. The missing part is to connect to the LOP, check if the blobs we +want to offload are already there and if not send them. + +5) A main remote should try to remain clean from large blobs +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +A main remote should try to avoid containing a lot of oversize +blobs. For that purpose, it should offload as needed to a LOP and it +should have ways to prevent oversize blobs to be fetched, and also +perhaps pushed, into it. + +Rationale ++++++++++ + +A main remote containing many oversize blobs would defeat the purpose +of LOPs. + +Implementation +++++++++++++++ + +The way to offload to a LOP discussed in 4) above can be used to +regularly offload oversize blobs. About preventing oversize blobs from +being fetched into the repo see 6) below. About preventing oversize +blob pushes, a pre-receive hook could be used. + +Also there are different scenarios in which large blobs could get +fetched into the main remote, for example: + +- A client that doesn't implement the "promisor-remote" protocol + (described in 6) below) clones from the main remote. + +- The main remote gets a request for information about a large blob + and is not able to get that information without fetching the blob + from the LOP. + +It might not be possible to completely prevent all these scenarios +from happening. So the goal here should be to implement features that +make the fetching of large blobs less likely. For example adding a +`remote-object-info` command in the `git cat-file --batch` protocol +and its variants might make it possible for a main repo to respond to +some requests about large blobs without fetching them. + +6) A protocol negotiation should happen when a client clones +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +When a client clones from a main repo, there should be a protocol +negotiation so that the server can advertise one or more LOPs and so +that the client and the server can discuss if the client could +directly use a LOP the server is advertising. If the client and the +server can agree on that, then the client would be able to get the +large blobs directly from the LOP and the server would not need to +fetch those blobs from the LOP to be able to serve the client. + +Note +++++ + +For fetches instead of clones, a protocol negotiation might not always +happen, see the "What about fetches?" FAQ entry below for details. + +Rationale ++++++++++ + +Security, configurability and efficiency of setting things up. + +Implementation +++++++++++++++ + +A "promisor-remote" protocol v2 capability looks like a good way to +implement this. The way the client and server use this capability +could be controlled by configuration variables. + +Information that the server could send to the client through that +protocol could be things like: LOP name, LOP URL, filter-spec (for +example `blob:limit=`) or just size limit that should be used as +a filter when cloning, token to be used with the LOP, etc. + +7) A client can offload to a LOP +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +When a client is using a LOP that is also a LOP of its main remote, +the client should be able to offload some large blobs it has fetched, +but might not need anymore, to the LOP. + +Note +++++ + +It might depend on the context if it should be OK or not for clients +to offload large blobs they have created, instead of fetched, directly +to the LOP without the main remote checking them in some ways +(possibly using hooks or other tools). + +This should be discussed and refined when we get closer to +implementing this feature. + +Rationale ++++++++++ + +On the client, the easiest way to deal with unneeded large blobs is to +offload them. + +Implementation +++++++++++++++ + +This is very similar to what 4) above is about, except on the client +side instead of the server side. So a good solution to 4) could likely +be adapted to work on the client side too. + +There might be some security issues here, as there is no negotiation, +but they might be mitigated if the client can reuse a token it got +when cloning (see 6) above). Also if the large blobs were fetched from +a LOP, it is likely, and can easily be confirmed, that the LOP still +has them, so that they can just be removed from the client. + +III) Benefits of using LOPs +--------------------------- + +Many benefits are related to the issues discussed in "I) Issues with +the current situation" above: + +- No need to rewrite history when deciding which blobs are worth + handling separately than other objects, or when moving or removing + the threshold. + +- If the protocol between client and server is developed and secured + enough, then many details might be setup on the server side only and + all the clients could then easily get all the configuration + information and use it to set themselves up mostly automatically. + +- Storage costs benefits on the server side. + +- Reduced memory and CPU needs on main remotes on the server side. + +- Reduced storage needs on the client side. + +IV) FAQ +------- + +What about using multiple LOPs on the server and client side? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +That could perhaps be useful in some cases, but for now it's more +likely that in most cases a single LOP will be advertised by the +server and should be used by the client. + +A case where it could be useful for a server to advertise multiple +LOPs is if a LOP is better for some users while a different LOP is +better for other users. For example some clients might have a better +connection to a LOP than others. + +In those cases it's the responsibility of the server to have some +documentation to help clients. It could say for example something like +"Users in this part of the world might want to pick only LOP A as it +is likely to be better connected to them, while users in other parts +of the world should pick only LOP B for the same reason." + +When should we trust or not trust the LOPs advertised by the server? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +In some contexts, like in corporate setup where the server and all the +clients are parts of an internal network in a company where admins +have all the rights on every system, it's OK, and perhaps even a good +thing, if the clients fully trust the server, as it can help ensure +that all the clients are on the same page. + +There are also contexts in which clients trust a code hosting platform +serving them some repos, but might not fully trust other users +managing or contributing to some of these repos. For example, the code +hosting platform could have hooks in place to check that any object it +receives doesn't contain malware or otherwise bad content. In this +case it might be OK for the client to use a main remote and its LOP if +they are both hosted by the code hosting platform, but not if the LOP +is hosted elsewhere (where the content is not checked). + +In other contexts, a client should just not trust a server. + +So there should be different ways to configure how the client should +behave when a server advertises a LOP to it at clone time. + +As the basic elements that a server can advertise about a LOP are a +LOP name and a LOP URL, the client should base its decision about +accepting a LOP on these elements. + +One simple way to be very strict in the LOP it accepts is for example +for the client to check that the LOP is already configured on the +client with the same name and URL as what the server advertises. + +In general default and "safe" settings should require that the LOP are +configured on the client separately from the "promisor-remote" +protocol and that the client accepts a LOP only when information about +it from the protocol matches what has been already configured +separately. + +What about LOP names? +~~~~~~~~~~~~~~~~~~~~~ + +In some contexts, for example if the clients sometimes fetch from each +other, it can be a good idea for all the clients to use the same names +for all the remotes they use, including LOPs. + +In other contexts, each client might want to be able to give the name +it wants to each remote, including each LOP, it interacts with. + +So there should be different ways to configure how the client accepts +or not the LOP name the server advertises. + +If a default or "safe" setting is used, then as such a setting should +require that the LOP be configured separately, then the name would be +configured separately and there is no risk that the server could +dictate a name to a client. + +Could the main remote be bogged down by old or paranoid clients? +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Yes, it could happen if there are too many clients that are either +unwilling to trust the main remote or that just don't implement the +"promisor-remote" protocol because they are too old or not fully +compatible with the 'git' client. + +When serving such a client, the main remote has no other choice than +to first fetch from its LOP, to then be able to provide to the client +everything it requested. So the main remote, even if it has cleanup +mechanisms (see section II.4 above), would be burdened at least +temporarily with the large blobs it had to fetch from its LOP. + +Not behaving like this would be breaking backward compatibility, and +could be seen as segregating clients. For example, it might be +possible to implement a special mode that allows the server to just +reject clients that don't implement the "promisor-remote" protocol or +aren't willing to trust the main remote. This mode might be useful in +a special context like a corporate environment. There is no plan to +implement such a mode though, and this should be discussed separately +later anyway. + +A better way to proceed is probably for the main remote to show a +message telling clients that don't implement the protocol or are +unwilling to accept the advertised LOP(s) that they would get faster +clone and fetches by upgrading client software or properly setting +them up to accept LOP(s). + +Waiting for clients to upgrade, monitoring these upgrades and limiting +the use of LOPs to repos that are not very frequently accessed might +be other good ways to make sure that some benefits are still reaped +from LOPs. Over time, as more and more clients upgrade and benefit +from LOPs, using them in more and more frequently accessed repos will +become worth it. + +Corporate environments, where it might be easier to make sure that all +the clients are up-to-date and properly configured, could hopefully +benefit more and earlier from using LOPs. + +What about fetches? +~~~~~~~~~~~~~~~~~~~ + +There are different kinds of fetches. A regular fetch happens when +some refs have been updated on the server and the client wants the ref +updates and possibly the new objects added with them. A "backfill" or +"lazy" fetch, on the contrary, happens when the client needs to use +some objects it already knows about but doesn't have because they are +on a promisor remote. + +Regular fetch ++++++++++++++ + +In a regular fetch, the client will contact the main remote and a +protocol negotiation will happen between them. It's a good thing that +a protocol negotiation happens every time, as the configuration on the +client or the main remote could have changed since the previous +protocol negotiation. In this case, the new protocol negotiation +should ensure that the new fetch will happen in a way that satisfies +the new configuration of both the client and the server. + +In most cases though, the configurations on the client and the main +remote will not have changed between 2 fetches or between the initial +clone and a subsequent fetch. This means that the result of a new +protocol negotiation will be the same as the previous result, so the +new fetch will happen in the same way as the previous clone or fetch, +using, or not using, the same LOP(s) as last time. + +"Backfill" or "lazy" fetch +++++++++++++++++++++++++++ + +When there is a backfill fetch, the client doesn't necessarily contact +the main remote first. It will try to fetch from its promisor remotes +in the order they appear in the config file, except that a remote +configured using the `extensions.partialClone` config variable will be +tried last. See +link:partial-clone.html#using-many-promisor-remotes[the partial clone documentation]. + +This is not new with this effort. In fact this is how multiple remotes +have already been working for around 5 years. + +When using LOPs, having the main remote configured using +`extensions.partialClone`, so it's tried last, makes sense, as missing +objects should only be large blobs that are on LOPs. + +This means that a protocol negotiation will likely not happen as the +missing objects will be fetched from the LOPs, and then there will be +nothing left to fetch from the main remote. + +To secure that, it could be a good idea for LOPs to require a token +from the client when it fetches from them. The client could get the +token when performing a protocol negotiation with the main remote (see +section II.6 above). + +V) Future improvements +---------------------- + +It is expected that at the beginning using LOPs will be mostly worth +it either in a corporate context where the Git version that clients +use can easily be controlled, or on repos that are infrequently +accessed. (See the "Could the main remote be bogged down by old or +paranoid clients?" section in the FAQ above.) + +Over time, as more and more clients upgrade to a version that +implements the "promisor-remote" protocol v2 capability described +above in section II.6), it will be worth it to use LOPs more widely. + +A lot of improvements may also help using LOPs more widely. Some of +these improvements are part of the scope of this document like the +following: + + - Implementing a "remote-object-info" command in the + `git cat-file --batch` protocol and its variants to allow main + remotes to respond to requests about large blobs without fetching + them. (Eric Ju has started working on this based on previous work + by Calvin Wan.) + + - Creating better cleanup and offload mechanisms for main remotes + and clients to prevent accumulation of large blobs. + + - Developing more sophisticated protocol negotiation capabilities + between clients and servers for handling LOPs, for example adding + a filter-spec (e.g., blob:limit=) or size limit for + filtering when cloning, or adding a token for LOP authentication. + + - Improving security measures for LOP access, particularly around + token handling and authentication. + + - Developing standardized ways to configure and manage multiple LOPs + across different environments. Especially in the case where + different LOPs serve the same content to clients in different + geographical locations, there is a need for replication or + synchronization between LOPs. + +Some improvements, including some that have been mentioned in the "0) +Non Goals" section of this document, are out of the scope of this +document: + + - Implementing a new object representation for large blobs on the + client side. + + - Developing pluggable ODBs or other object database backends that + could chunk large blobs, dedup the chunks and store them + efficiently. + + - Optimizing data transfer between LOPs and clients/servers, + particularly for incompressible and non-deltifying content. + + - Creating improved client side tools for managing large objects + more effectively, for example tools for migrating from Git LFS or + git-annex, or tools to find which objects could be offloaded and + how much disk space could be reclaimed by offloading them. + +Some improvements could be seen as part of the scope of this document, +but might already have their own separate projects from the Git +project, like: + + - Improving existing remote helpers to access object storage or + developing new ones. + + - Improving existing object storage solutions or developing new + ones. + +Even though all the above improvements may help, this document and the +LOP effort should try to focus, at least first, on a relatively small +number of improvements mostly those that are in its current scope. + +For example introducing pluggable ODBs and a new object database +backend is likely a multi-year effort on its own that can happen +separately in parallel. It has different technical requirements, +touches other part of the Git code base and should have its own design +document(s). From c268e3285d4eafeec048d98672faa43f21857990 Mon Sep 17 00:00:00 2001 From: Junio C Hamano Date: Fri, 28 Feb 2025 09:28:21 -0800 Subject: [PATCH 030/312] BreakingChanges: clarify the procedure The point behind a compile-time switch is to ensure that we have a mechanism to hide myriad of backward incompatible changes that may be prepared and accumulated over time, yet make them available for testing any time during the development toward the big version boundary. Add a few words to stress that point. Since the document was first written, we have added the CI job that the document anticipated us to have. Rephrase to state the current status. The discussion in [*1*] made us abandon the "feature.git3" based runtime switching of behaviour and instead adopt the compile-time switching mechanism, but a stray sentence about runtime switching still remained in the final text by mistake. Remove it. [Reference] *1* https://lore.kernel.org/git/xmqqldzel6ug.fsf@gitster.g/ Signed-off-by: Junio C Hamano Acked-by: Patrick Steinhardt Signed-off-by: Junio C Hamano --- Documentation/BreakingChanges.txt | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-) diff --git a/Documentation/BreakingChanges.txt b/Documentation/BreakingChanges.txt index eeb26c915550b2..3cf9dfb35f6df5 100644 --- a/Documentation/BreakingChanges.txt +++ b/Documentation/BreakingChanges.txt @@ -66,22 +66,21 @@ changes are made at a certain version boundary, and recording these decisions in this document, are necessary but not sufficient. Because such changes are expected to be numerous, and the design and implementation of them are expected to span over time, they have to -be deployable trivially at such a version boundary. +be deployable trivially at such a version boundary, prepared over long +time. The breaking changes MUST be guarded with the a compile-time switch, WITH_BREAKING_CHANGES, to help this process. When built with it, the resulting Git binary together with its documentation would behave as if these breaking changes slated for the next big version -boundary are already in effect. We may also want to have a CI job -or two to exercise the work-in-progress version of Git with these -breaking changes. +boundary are already in effect. We also have a CI job to exercise +the work-in-progress version of Git with these breaking changes. == Git 3.0 The following subsections document upcoming breaking changes for Git 3.0. There -is no planned release date for this breaking version yet. The early -adopter configuration used for changes for this release is `feature.git3`. +is no planned release date for this breaking version yet. Proposed changes and removals only include items which are "ready" to be done. In other words, this is not supposed to be a wishlist of features that should From 82deaae3b94d59caf13ea21c980ab386df6dac9e Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:43:59 -0500 Subject: [PATCH 031/312] t0450: *.txt -> *.adoc fixes After 1f010d6bdf (doc: use .adoc extension for AsciiDoc files, 2025-01-20), we no longer matched any files in this test. The result is that we did not test for mismatches in the documentation and --help output. Adjust the test to look at the renamed *.adoc files. Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- t/t0450-txt-doc-vs-help.sh | 50 +++++++++---------- ...t-help-mismatches => adoc-help-mismatches} | 0 2 files changed, 25 insertions(+), 25 deletions(-) rename t/t0450/{txt-help-mismatches => adoc-help-mismatches} (100%) diff --git a/t/t0450-txt-doc-vs-help.sh b/t/t0450-txt-doc-vs-help.sh index 853101b86ef53d..2f7504ae7e9090 100755 --- a/t/t0450-txt-doc-vs-help.sh +++ b/t/t0450-txt-doc-vs-help.sh @@ -1,6 +1,6 @@ #!/bin/sh -test_description='assert (unbuilt) Documentation/*.txt and -h output +test_description='assert (unbuilt) Documentation/*.adoc and -h output Run this with --debug to see a summary of where we still fail to make the two versions consistent with one another.' @@ -11,11 +11,11 @@ test_expect_success 'setup: list of builtins' ' git --list-cmds=builtins >builtins ' -test_expect_success 'list of txt and help mismatches is sorted' ' - sort -u "$TEST_DIRECTORY"/t0450/txt-help-mismatches >expect && - if ! test_cmp expect "$TEST_DIRECTORY"/t0450/txt-help-mismatches +test_expect_success 'list of adoc and help mismatches is sorted' ' + sort -u "$TEST_DIRECTORY"/t0450/adoc-help-mismatches >expect && + if ! test_cmp expect "$TEST_DIRECTORY"/t0450/adoc-help-mismatches then - BUG "please keep the list of txt and help mismatches sorted" + BUG "please keep the list of adoc and help mismatches sorted" fi ' @@ -40,20 +40,20 @@ help_to_synopsis () { echo "$out" } -builtin_to_txt () { - echo "$GIT_BUILD_DIR/Documentation/git-$1.txt" +builtin_to_adoc () { + echo "$GIT_BUILD_DIR/Documentation/git-$1.adoc" } -txt_to_synopsis () { +adoc_to_synopsis () { builtin="$1" && out_dir="out/$builtin" && - out="$out_dir/txt.synopsis" && + out="$out_dir/adoc.synopsis" && if test -f "$out" then echo "$out" && return 0 fi && - b2t="$(builtin_to_txt "$builtin")" && + b2t="$(builtin_to_adoc "$builtin")" && sed -n \ -E '/^\[(verse|synopsis)\]$/,/^$/ { /^$/d; @@ -109,29 +109,29 @@ do fi ' - txt="$(builtin_to_txt "$builtin")" && - preq="$(echo BUILTIN_TXT_$builtin | tr '[:lower:]-' '[:upper:]_')" && + adoc="$(builtin_to_adoc "$builtin")" && + preq="$(echo BUILTIN_ADOC_$builtin | tr '[:lower:]-' '[:upper:]_')" && - if test -f "$txt" + if test -f "$adoc" then test_set_prereq "$preq" fi && - # *.txt output assertions - test_expect_success "$preq" "$builtin *.txt SYNOPSIS has dashed labels" ' - check_dashed_labels "$(txt_to_synopsis "$builtin")" + # *.adoc output assertions + test_expect_success "$preq" "$builtin *.adoc SYNOPSIS has dashed labels" ' + check_dashed_labels "$(adoc_to_synopsis "$builtin")" ' - # *.txt output consistency assertions + # *.adoc output consistency assertions result= - if grep -q "^$builtin$" "$TEST_DIRECTORY"/t0450/txt-help-mismatches + if grep -q "^$builtin$" "$TEST_DIRECTORY"/t0450/adoc-help-mismatches then result=failure else result=success fi && test_expect_$result "$preq" "$builtin -h output and SYNOPSIS agree" ' - t2s="$(txt_to_synopsis "$builtin")" && + t2s="$(adoc_to_synopsis "$builtin")" && if test "$builtin" = "merge-tree" then test_when_finished "rm -f t2s.new" && @@ -140,17 +140,17 @@ do fi && h2s="$(help_to_synopsis "$builtin")" && - # The *.txt and -h use different spacing for the + # The *.adoc and -h use different spacing for the # alignment of continued usage output, normalize it. - align_after_nl "$builtin" <"$t2s" >txt && + align_after_nl "$builtin" <"$t2s" >adoc && align_after_nl "$builtin" <"$h2s" >help && - test_cmp txt help + test_cmp adoc help ' - if test_have_prereq "$preq" && test -e txt && test -e help + if test_have_prereq "$preq" && test -e adoc && test -e help then test_debug ' - if test_cmp txt help >cmp 2>/dev/null + if test_cmp adoc help >cmp 2>/dev/null then echo "=== DONE: $builtin ===" else @@ -161,7 +161,7 @@ do # Not in test_expect_success in case --run is being # used with --debug - rm -f txt help tmp 2>/dev/null + rm -f adoc help tmp 2>/dev/null fi done Date: Mon, 3 Mar 2025 15:44:00 -0500 Subject: [PATCH 032/312] .gitattributes: more *.txt -> *.adoc updates All Documentation files now end in .adoc. Update the entries for git-merge.adoc, gitk.adoc, and user-manual.adoc to properly set the conflict-marker-size attribute. Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- .gitattributes | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/.gitattributes b/.gitattributes index 43fa883a84f173..c6a0b35116fd90 100644 --- a/.gitattributes +++ b/.gitattributes @@ -12,7 +12,7 @@ CODE_OF_CONDUCT.md -whitespace /GIT-VERSION-GEN text eol=lf /mergetools/* text eol=lf /t/oid-info/* text eol=lf -/Documentation/git-merge.txt conflict-marker-size=32 -/Documentation/gitk.txt conflict-marker-size=32 -/Documentation/user-manual.txt conflict-marker-size=32 +/Documentation/git-merge.adoc conflict-marker-size=32 +/Documentation/gitk.adoc conflict-marker-size=32 +/Documentation/user-manual.adoc conflict-marker-size=32 /t/t????-*.sh conflict-marker-size=32 From 7d90a272ac4f9a2755d7b4f5807d670e0212e70f Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:01 -0500 Subject: [PATCH 033/312] doc: remove unneeded .gitattributes The top-level .gitattributes file contains entries for the Documentation tree. Documentation/.gitattributes has not been touched since it was added in 14f9e128d3 (Define the project whitespace policy, 2008-02-10). Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- Documentation/.gitattributes | 1 - 1 file changed, 1 deletion(-) delete mode 100644 Documentation/.gitattributes diff --git a/Documentation/.gitattributes b/Documentation/.gitattributes deleted file mode 100644 index ddb030137d54ef..00000000000000 --- a/Documentation/.gitattributes +++ /dev/null @@ -1 +0,0 @@ -*.txt whitespace From d40da0bd4b3c9c591fe36d4772d44209fbd70504 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:02 -0500 Subject: [PATCH 034/312] Makefile: update reference to technical/racy-git.adoc Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/Makefile b/Makefile index a9b2de06922798..a8012569da2876 100644 --- a/Makefile +++ b/Makefile @@ -194,7 +194,7 @@ include shared.mak # Linux, kernel 2.6.11 or newer is required for reliable sub-second file times # on file systems with exactly 1 ns or 1 s resolution. If you intend to use Git # on other file systems (e.g. CEPH, CIFS, NTFS, UDF), don't enable USE_NSEC. See -# Documentation/technical/racy-git.txt for details. +# Documentation/technical/racy-git.adoc for details. # # Define USE_ST_TIMESPEC if your "struct stat" uses "st_ctimespec" instead of # "st_ctim" From c50fbb2dd225e7e82abba4380423ae105089f4d7 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:03 -0500 Subject: [PATCH 035/312] README: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- README.md | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) diff --git a/README.md b/README.md index 665ce5f5a83647..d87bca1b8c3ebf 100644 --- a/README.md +++ b/README.md @@ -17,15 +17,15 @@ Please read the file [INSTALL][] for installation instructions. Many Git online resources are accessible from including full documentation and Git related tools. -See [Documentation/gittutorial.txt][] to get started, then see -[Documentation/giteveryday.txt][] for a useful minimum set of commands, and -`Documentation/git-.txt` for documentation of each command. +See [Documentation/gittutorial.adoc][] to get started, then see +[Documentation/giteveryday.adoc][] for a useful minimum set of commands, and +`Documentation/git-.adoc` for documentation of each command. If git has been correctly installed, then the tutorial can also be read with `man gittutorial` or `git help tutorial`, and the documentation of each command with `man git-` or `git help `. -CVS users may also want to read [Documentation/gitcvs-migration.txt][] +CVS users may also want to read [Documentation/gitcvs-migration.adoc][] (`man gitcvs-migration` or `git help cvs-migration` if git is installed). @@ -66,9 +66,9 @@ and the name as (depending on your mood): - "goddamn idiotic truckload of sh*t": when it breaks [INSTALL]: INSTALL -[Documentation/gittutorial.txt]: Documentation/gittutorial.txt -[Documentation/giteveryday.txt]: Documentation/giteveryday.txt -[Documentation/gitcvs-migration.txt]: Documentation/gitcvs-migration.txt +[Documentation/gittutorial.adoc]: Documentation/gittutorial.adoc +[Documentation/giteveryday.adoc]: Documentation/giteveryday.adoc +[Documentation/gitcvs-migration.adoc]: Documentation/gitcvs-migration.adoc [Documentation/SubmittingPatches]: Documentation/SubmittingPatches [Documentation/CodingGuidelines]: Documentation/CodingGuidelines [po/README.md]: po/README.md From 7c78c599bb9b51e5cbdae3e7dc1d723eefcf7c61 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:04 -0500 Subject: [PATCH 036/312] CodingGuidelines: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- Documentation/CodingGuidelines | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/Documentation/CodingGuidelines b/Documentation/CodingGuidelines index ba047ed224afbd..a0e7041c54b497 100644 --- a/Documentation/CodingGuidelines +++ b/Documentation/CodingGuidelines @@ -44,7 +44,7 @@ code are expected to match the style the surrounding code already uses (even if it doesn't match the overall style of existing code). But if you must have a list of rules, here are some language -specific ones. Note that Documentation/ToolsForGit.txt document +specific ones. Note that Documentation/ToolsForGit.adoc document has a collection of tips to help you use some external tools to conform to these guidelines. @@ -755,7 +755,7 @@ Externally Visible Names Writing Documentation: Most (if not all) of the documentation pages are written in the - AsciiDoc format in *.txt files (e.g. Documentation/git.txt), and + AsciiDoc format in *.adoc files (e.g. Documentation/git.adoc), and processed into HTML and manpages (e.g. git.html and git.1 in the same directory). From 8b4b41aefb87c9b3dbdf40bb65686cb37038cb71 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:05 -0500 Subject: [PATCH 037/312] MyFirstContribution: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- Documentation/MyFirstContribution.adoc | 24 ++++++++++++------------ 1 file changed, 12 insertions(+), 12 deletions(-) diff --git a/Documentation/MyFirstContribution.adoc b/Documentation/MyFirstContribution.adoc index e41654c00a68e0..afcf4b46c11ab2 100644 --- a/Documentation/MyFirstContribution.adoc +++ b/Documentation/MyFirstContribution.adoc @@ -21,7 +21,7 @@ This tutorial aims to summarize the following documents, but the reader may find useful additional context: - `Documentation/SubmittingPatches` -- `Documentation/howto/new-command.txt` +- `Documentation/howto/new-command.adoc` [[getting-help]] === Getting Help @@ -331,7 +331,7 @@ function body: apply standard precedence rules. `git_config_get_string_tmp()` will look up a specific key ("user.name") and give you the value. There are a number of single-key lookup functions like this one; you can see them all (and more info -about how to use `git_config()`) in `Documentation/technical/api-config.txt`. +about how to use `git_config()`) in `Documentation/technical/api-config.adoc`. You should see that the name printed matches the one you see when you run: @@ -461,10 +461,10 @@ $ ./bin-wrappers/git help psuh Your new command is undocumented! Let's fix that. -Take a look at `Documentation/git-*.txt`. These are the manpages for the +Take a look at `Documentation/git-*.adoc`. These are the manpages for the subcommands that Git knows about. You can open these up and take a look to get acquainted with the format, but then go ahead and make a new file -`Documentation/git-psuh.txt`. Like with most of the documentation in the Git +`Documentation/git-psuh.adoc`. Like with most of the documentation in the Git project, help pages are written with AsciiDoc (see CodingGuidelines, "Writing Documentation" section). Use the following template to fill out your own manpage: @@ -543,7 +543,7 @@ Try and run `./bin-wrappers/git psuh -h`. Your command should crash at the end. That's because `-h` is a special case which your command should handle by printing usage. -Take a look at `Documentation/technical/api-parse-options.txt`. This is a handy +Take a look at `Documentation/technical/api-parse-options.adoc`. This is a handy tool for pulling out options you need to be able to handle, and it takes a usage string. @@ -1088,14 +1088,14 @@ This gives reviewers a summary of what they're in for when reviewing your topic. The one generated for `psuh` from the sample implementation looks like this: ---- - Documentation/git-psuh.txt | 40 +++++++++++++++++++++ - Makefile | 1 + - builtin.h | 1 + - builtin/psuh.c | 73 ++++++++++++++++++++++++++++++++++++++ - git.c | 1 + - t/t9999-psuh-tutorial.sh | 12 +++++++ + Documentation/git-psuh.adoc | 40 +++++++++++++++++++++ + Makefile | 1 + + builtin.h | 1 + + builtin/psuh.c | 73 ++++++++++++++++++++++++++++++++++++++ + git.c | 1 + + t/t9999-psuh-tutorial.sh | 12 +++++++ 6 files changed, 128 insertions(+) - create mode 100644 Documentation/git-psuh.txt + create mode 100644 Documentation/git-psuh.adoc create mode 100644 builtin/psuh.c create mode 100755 t/t9999-psuh-tutorial.sh ---- From 5ac2c61b550eb801d3acab1aa68f52379f9ad33b Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:06 -0500 Subject: [PATCH 038/312] MyFirstObjectWalk: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- Documentation/MyFirstObjectWalk.adoc | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/Documentation/MyFirstObjectWalk.adoc b/Documentation/MyFirstObjectWalk.adoc index dec8afe5b10533..d6e9dfdbbe232e 100644 --- a/Documentation/MyFirstObjectWalk.adoc +++ b/Documentation/MyFirstObjectWalk.adoc @@ -15,7 +15,7 @@ revision walk is used for operations like `git log`. === Related Reading -- `Documentation/user-manual.txt` under "Hacking Git" contains some coverage of +- `Documentation/user-manual.adoc` under "Hacking Git" contains some coverage of the revision walker in its various incarnations. - `revision.h` - https://eagain.net/articles/git-for-computer-scientists/[Git for Computer Scientists] @@ -112,7 +112,7 @@ $ GIT_TRACE=1 ./bin-wrappers/git walken ---- NOTE: For a more exhaustive overview of the new command process, take a look at -`Documentation/MyFirstContribution.txt`. +`Documentation/MyFirstContribution.adoc`. NOTE: A reference implementation can be found at https://github.com/nasamuffin/git/tree/revwalk. @@ -132,7 +132,7 @@ used to track the allocated size of the list. Per entry, we find: `item` is the object provided upon which to base the object walk. Items in Git -can be blobs, trees, commits, or tags. (See `Documentation/gittutorial-2.txt`.) +can be blobs, trees, commits, or tags. (See `Documentation/gittutorial-2.adoc`.) `name` is the object ID (OID) of the object - a hex string you may be familiar with from using Git to organize your source in the past. Check the tutorial @@ -141,7 +141,7 @@ from. `whence` indicates some information about what to do with the parents of the specified object. We'll explore this flag more later on; take a look at -`Documentation/revisions.txt` to get an idea of what could set the `whence` +`Documentation/revisions.adoc` to get an idea of what could set the `whence` value. `flags` are used to hint the beginning of the revision walk and are the first @@ -153,7 +153,7 @@ can be used during the walk, as well. This one is quite a bit longer, and many fields are only used during the walk by `revision.c` - not configuration options. Most of the configurable flags in -`struct rev_info` have a mirror in `Documentation/rev-list-options.txt`. It's a +`struct rev_info` have a mirror in `Documentation/rev-list-options.adoc`. It's a good idea to take some time and read through that document. == Basic Commit Walk @@ -710,7 +710,7 @@ objects grows along with the Git project. === Adding a Filter There are a handful of filters that we can apply to the object walk laid out in -`Documentation/rev-list-options.txt`. These filters are typically useful for +`Documentation/rev-list-options.adoc`. These filters are typically useful for operations such as creating packfiles or performing a partial clone. They are defined in `list-objects-filter-options.h`. For the purposes of this tutorial we will use the "tree:1" filter, which causes the walk to omit all trees and blobs From 9100c91cd4c83827b6c3c296e3b2ce89f9e98f52 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:07 -0500 Subject: [PATCH 039/312] howto/new-command: update reference to builtin docs Commit ec14d4ecb5 (builtin.h: take over documentation from api-builtin.txt, 2017-08-02) deleted api-builtin.txt and moved the contents into builtin.h. Most of the references were fixed in d85e9448dd (new-command.txt: update reference to builtin docs, 2023-02-04), but one remained. Fix it. Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- Documentation/howto/new-command.adoc | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/Documentation/howto/new-command.adoc b/Documentation/howto/new-command.adoc index 880c51112ba478..ac73c98be72ded 100644 --- a/Documentation/howto/new-command.adoc +++ b/Documentation/howto/new-command.adoc @@ -48,7 +48,7 @@ binary); this organization makes it easy for people reading the code to find things. See the CodingGuidelines document for other guidance on what we consider -good practice in C and shell, and api-builtin.txt for the support +good practice in C and shell, and builtin.h for the support functions available to built-in commands written in C. What every extension command needs From 59d92809088f28b06b0cd102dd166ba8b3144fc7 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:08 -0500 Subject: [PATCH 040/312] technical/partial-clone: update reference to rev-list-options.adoc Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- Documentation/technical/partial-clone.adoc | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/Documentation/technical/partial-clone.adoc b/Documentation/technical/partial-clone.adoc index bf5ec5c82d9e0f..e513e391ea82ca 100644 --- a/Documentation/technical/partial-clone.adoc +++ b/Documentation/technical/partial-clone.adoc @@ -85,7 +85,7 @@ See "filter" in linkgit:gitprotocol-pack[5]. server to request filtering during packfile construction. + There are various filters available to accommodate different situations. -See "--filter=" in Documentation/rev-list-options.txt. +See "--filter=" in Documentation/rev-list-options.adoc. - On the server pack-objects applies the requested filter-spec as it creates "filtered" packfiles for the client. From 97350e18e2fc2e753e34df4ee8c1107457aa0a91 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:09 -0500 Subject: [PATCH 041/312] doc: *.txt -> *.adoc fixes Update a few more instances of Documentation/*.txt files which have been renamed to *.adoc. Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- Documentation/gitattributes.adoc | 6 +++--- Documentation/technical/api-simple-ipc.adoc | 2 +- Documentation/technical/hash-function-transition.adoc | 2 +- 3 files changed, 5 insertions(+), 5 deletions(-) diff --git a/Documentation/gitattributes.adoc b/Documentation/gitattributes.adoc index 7eaca899725664..a22d1ef1e15438 100644 --- a/Documentation/gitattributes.adoc +++ b/Documentation/gitattributes.adoc @@ -513,7 +513,7 @@ If the filter command (a string value) is defined via `filter..process` then Git can process all blobs with a single filter invocation for the entire life of a single Git command. This is achieved by using the long-running process protocol -(described in technical/long-running-process-protocol.txt). +(described in Documentation/technical/long-running-process-protocol.adoc). When Git encounters the first file that needs to be cleaned or smudged, it starts the filter and performs the handshake. In the handshake, the @@ -1177,11 +1177,11 @@ integer has a meaningful effect. For example, this line in `.gitattributes` can be used to tell the merge machinery to leave much longer (instead of the usual 7-character-long) -conflict markers when merging the file `Documentation/git-merge.txt` +conflict markers when merging the file `Documentation/git-merge.adoc` results in a conflict. ------------------------ -Documentation/git-merge.txt conflict-marker-size=32 +Documentation/git-merge.adoc conflict-marker-size=32 ------------------------ diff --git a/Documentation/technical/api-simple-ipc.adoc b/Documentation/technical/api-simple-ipc.adoc index c4fb152b23291c..972178b042bc99 100644 --- a/Documentation/technical/api-simple-ipc.adoc +++ b/Documentation/technical/api-simple-ipc.adoc @@ -36,7 +36,7 @@ Comparison with sub-process model --------------------------------- The Simple-IPC mechanism differs from the existing `sub-process.c` -model (Documentation/technical/long-running-process-protocol.txt) and +model (Documentation/technical/long-running-process-protocol.adoc) and used by applications like Git-LFS. In the LFS-style sub-process model, the helper is started by the foreground process, communication happens via a pair of file descriptors bound to the stdin/stdout of the diff --git a/Documentation/technical/hash-function-transition.adoc b/Documentation/technical/hash-function-transition.adoc index 7102c7c8f5d66d..f047fd80cadd1e 100644 --- a/Documentation/technical/hash-function-transition.adoc +++ b/Documentation/technical/hash-function-transition.adoc @@ -394,7 +394,7 @@ inflated again in step 3, for a total of two inflations. Step 4 is probably necessary for good read-time performance. "git pack-objects" on the server optimizes the pack file for good data -locality (see Documentation/technical/pack-heuristics.txt). +locality (see Documentation/technical/pack-heuristics.adoc). Details of this process are likely to change. It will take some experimenting to get this to perform well. From d795c65b3a29dca33991c441a9951fb9dd14eead Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:10 -0500 Subject: [PATCH 042/312] advice.h: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- advice.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/advice.h b/advice.h index cf2284ec43864b..d233cfc693c726 100644 --- a/advice.h +++ b/advice.h @@ -7,7 +7,7 @@ struct string_list; * To add a new advice, you need to: * Define a new advice_type. * Add a new entry to advice_setting array. - * Add the new config variable to Documentation/config/advice.txt. + * Add the new config variable to Documentation/config/advice.adoc. * Call advise_if_enabled to print your advice. */ enum advice_type { From 08ce333d361890714829897e9f3b9fc9ec06eb4f Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:11 -0500 Subject: [PATCH 043/312] apply.c: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- apply.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/apply.c b/apply.c index b124678b93bd93..f274a3794877dc 100644 --- a/apply.c +++ b/apply.c @@ -82,7 +82,7 @@ static int parse_whitespace_option(struct apply_state *state, const char *option } /* * Please update $__git_whitespacelist in git-completion.bash, - * Documentation/git-apply.txt, and Documentation/git-am.txt + * Documentation/git-apply.adoc, and Documentation/git-am.adoc * when you add new options. */ return error(_("unrecognized whitespace option '%s'"), option); From e8015223c74938548dfd8d6099b1ea5eb6ab9231 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:12 -0500 Subject: [PATCH 044/312] builtin.h: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- builtin.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/builtin.h b/builtin.h index 89928ccf92f532..993a583872c304 100644 --- a/builtin.h +++ b/builtin.h @@ -63,7 +63,7 @@ * * . Add tests to `t/` directory. * - * . Write documentation in `Documentation/git-foo.txt`. + * . Write documentation in `Documentation/git-foo.adoc`. * * . Add an entry for `git-foo` to `command-list.txt`. * From bbd6174b2501a3523d778dbd479e85119c86840f Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:13 -0500 Subject: [PATCH 045/312] config.c: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- config.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/config.c b/config.c index 36f76fafe53ee7..dcb5cb52657348 100644 --- a/config.c +++ b/config.c @@ -1652,7 +1652,7 @@ static int git_default_core_config(const char *var, const char *value, return 0; } - /* Add other config variables here and to Documentation/config.txt. */ + /* Add other config variables here and to Documentation/config.adoc. */ return platform_core_config(var, value, ctx, cb); } @@ -1663,7 +1663,7 @@ static int git_default_sparse_config(const char *var, const char *value) return 0; } - /* Add other config variables here and to Documentation/config/sparse.txt. */ + /* Add other config variables here and to Documentation/config/sparse.adoc. */ return 0; } @@ -1679,7 +1679,7 @@ static int git_default_i18n_config(const char *var, const char *value) return git_config_string(&git_log_output_encoding, var, value); } - /* Add other config variables here and to Documentation/config.txt. */ + /* Add other config variables here and to Documentation/config.adoc. */ return 0; } @@ -1715,7 +1715,7 @@ static int git_default_branch_config(const char *var, const char *value) return 0; } - /* Add other config variables here and to Documentation/config.txt. */ + /* Add other config variables here and to Documentation/config.adoc. */ return 0; } @@ -1744,7 +1744,7 @@ static int git_default_push_config(const char *var, const char *value) return 0; } - /* Add other config variables here and to Documentation/config.txt. */ + /* Add other config variables here and to Documentation/config.adoc. */ return 0; } @@ -1760,7 +1760,7 @@ static int git_default_mailmap_config(const char *var, const char *value) return git_config_string(&git_mailmap_blob, var, value); } - /* Add other config variables here and to Documentation/config.txt. */ + /* Add other config variables here and to Documentation/config.adoc. */ return 0; } @@ -1773,7 +1773,7 @@ static int git_default_attr_config(const char *var, const char *value) /* * Add other attribute related config variables here and to - * Documentation/config/attr.txt. + * Documentation/config/attr.adoc. */ return 0; } @@ -1831,7 +1831,7 @@ int git_default_config(const char *var, const char *value, if (starts_with(var, "sparse.")) return git_default_sparse_config(var, value); - /* Add other config variables here and to Documentation/config.txt. */ + /* Add other config variables here and to Documentation/config.adoc. */ return 0; } From 87e0910fb821d125fc06d25a8309d3c668045f57 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:14 -0500 Subject: [PATCH 046/312] contrib/long-running-filter: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- contrib/long-running-filter/example.pl | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/contrib/long-running-filter/example.pl b/contrib/long-running-filter/example.pl index a677569ddd95f4..4b83e4c5e89427 100755 --- a/contrib/long-running-filter/example.pl +++ b/contrib/long-running-filter/example.pl @@ -1,7 +1,7 @@ #!/usr/bin/perl # # Example implementation for the Git filter protocol version 2 -# See Documentation/gitattributes.txt, section "Filter Protocol" +# See Documentation/gitattributes.adoc, section "Filter Protocol" # # Please note, this pass-thru filter is a minimal skeleton. No proper # error handling was implemented. From 5c0375266594f19f3d73969718a5806662cd6d63 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:15 -0500 Subject: [PATCH 047/312] diff.h: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- diff.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/diff.h b/diff.h index 0a566f5531ac42..ff0348e4a9a022 100644 --- a/diff.h +++ b/diff.h @@ -333,7 +333,7 @@ struct diff_options { int xdl_opts; int ignore_driver_algorithm; - /* see Documentation/diff-options.txt */ + /* see Documentation/diff-options.adoc */ char **anchors; size_t anchors_nr, anchors_alloc; From 3936e95a7f056877425c6faa6994d9f3712a6fb0 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:16 -0500 Subject: [PATCH 048/312] diffcore.h: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- diffcore.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/diffcore.h b/diffcore.h index 2feb3250313125..9c0a0e7aaff85a 100644 --- a/diffcore.h +++ b/diffcore.h @@ -107,7 +107,7 @@ struct diff_filepair { struct diff_filespec *one; struct diff_filespec *two; unsigned short int score; - char status; /* M C R A D U etc. (see Documentation/diff-format.txt or DIFF_STATUS_* in diff.h) */ + char status; /* M C R A D U etc. (see Documentation/diff-format.adoc or DIFF_STATUS_* in diff.h) */ unsigned broken_pair : 1; unsigned renamed_pair : 1; unsigned is_unmerged : 1; From c09c29b43046373a0ace268c6de6266b5817dbc3 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:17 -0500 Subject: [PATCH 049/312] fsck.h: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- fsck.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fsck.h b/fsck.h index a44c231a5f1391..a95ae7eb847c49 100644 --- a/fsck.h +++ b/fsck.h @@ -15,7 +15,7 @@ enum fsck_msg_type { }; /* - * Documentation/fsck-msgids.txt documents these; when + * Documentation/fsck-msgids.adoc documents these; when * modifying this list in any way, make sure to keep the * two in sync. */ From 02ed88f6a23d142f6e5c7e0b351704654796a6fe Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:18 -0500 Subject: [PATCH 050/312] list-objects-filter-options.h: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- list-objects-filter-options.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/list-objects-filter-options.h b/list-objects-filter-options.h index 55fab8563d20a5..7b2108b98662c9 100644 --- a/list-objects-filter-options.h +++ b/list-objects-filter-options.h @@ -82,7 +82,7 @@ void list_objects_filter_init(struct list_objects_filter_options *filter_options * "filter" SP * * The filter keyword will be used by many commands. - * See Documentation/rev-list-options.txt for allowed values for . + * See Documentation/rev-list-options.adoc for allowed values for . * * Capture the given arg as the "filter_spec". This can be forwarded to * subordinate commands when necessary (although it's better to pass it through From 550fac1d131d764c7c5ec5d69f53290c30855be6 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:19 -0500 Subject: [PATCH 051/312] object-name.c: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- object-name.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/object-name.c b/object-name.c index 945d5bdef25217..76749fbfe652bf 100644 --- a/object-name.c +++ b/object-name.c @@ -1273,7 +1273,7 @@ static int peel_onion(struct repository *r, const char *name, int len, } /* - * Documentation/revisions.txt says: + * Documentation/revisions.adoc says: * '', e.g. 'v1.7.4.2-679-g3bee7fb':: * Output from `git describe`; i.e. a closest tag, optionally * followed by a dash and a number of commits, followed by a dash, a From 458f8b0eab9af7b770a25e9277f70bb4df3c8d90 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:20 -0500 Subject: [PATCH 052/312] parse-options.h: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- parse-options.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/parse-options.h b/parse-options.h index fca944d9a93d64..997ffbee8050d7 100644 --- a/parse-options.h +++ b/parse-options.h @@ -6,7 +6,7 @@ struct repository; /** - * Refer to Documentation/technical/api-parse-options.txt for the API doc. + * Refer to Documentation/technical/api-parse-options.adoc for the API doc. */ enum parse_opt_type { From dc657d56254d335016c02fb14bb257e37b8cde89 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:21 -0500 Subject: [PATCH 053/312] pseudo-merge.h: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- pseudo-merge.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/pseudo-merge.h b/pseudo-merge.h index 29df8a32ecf3be..c9fbe9d3129583 100644 --- a/pseudo-merge.h +++ b/pseudo-merge.h @@ -101,7 +101,7 @@ void select_pseudo_merges(struct bitmap_writer *writer); /* * Represents a serialized view of a file containing pseudo-merge(s) - * (see Documentation/technical/bitmap-format.txt for a specification + * (see Documentation/technical/bitmap-format.adoc for a specification * of the format). */ struct pseudo_merge_map { From 72d385824a75e1d39b2a4395852dc954907fcc7d Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:22 -0500 Subject: [PATCH 054/312] refs.h: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- refs.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/refs.h b/refs.h index 09be47afbee51e..31c67a71b24543 100644 --- a/refs.h +++ b/refs.h @@ -577,7 +577,7 @@ int refs_for_each_reflog(struct ref_store *refs, each_reflog_fn fn, void *cb_dat /* * Return 0 iff refname has the correct format for a refname according - * to the rules described in Documentation/git-check-ref-format.txt. + * to the rules described in Documentation/git-check-ref-format.adoc. * If REFNAME_ALLOW_ONELEVEL is set in flags, then accept one-level * reference names. If REFNAME_REFSPEC_PATTERN is set in flags, then * allow a single "*" wildcard character in the refspec. No leading or From 0543300b594eb1337ce33415997b47594bf29056 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:23 -0500 Subject: [PATCH 055/312] setup.c: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- setup.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/setup.c b/setup.c index 7da7aa8984b6df..801bd8c1ee8e24 100644 --- a/setup.c +++ b/setup.c @@ -1822,7 +1822,7 @@ const char *setup_git_directory_gently(int *nongit_ok) * * Regardless of the state of nongit_ok, startup_info->prefix and * the GIT_PREFIX environment variable must always match. For details - * see Documentation/config/alias.txt. + * see Documentation/config/alias.adoc. */ if (nongit_ok && *nongit_ok) startup_info->have_repository = 0; From 9f04cd7c61d35a9afc81d0625dc914c5221d22ab Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:24 -0500 Subject: [PATCH 056/312] simple-ipc.h: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- simple-ipc.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/simple-ipc.h b/simple-ipc.h index 3916eaf70d9863..701e005cb8e5f3 100644 --- a/simple-ipc.h +++ b/simple-ipc.h @@ -2,7 +2,7 @@ #define GIT_SIMPLE_IPC_H /* - * See Documentation/technical/api-simple-ipc.txt + * See Documentation/technical/api-simple-ipc.adoc */ enum ipc_active_state { From e680c625424e153580e8f2d78ae2ffaa42335221 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:25 -0500 Subject: [PATCH 057/312] t/helper/test-rot13-filter.c: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- t/helper/test-rot13-filter.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/t/helper/test-rot13-filter.c b/t/helper/test-rot13-filter.c index ff407b575ce318..722b1cbe7788bc 100644 --- a/t/helper/test-rot13-filter.c +++ b/t/helper/test-rot13-filter.c @@ -1,6 +1,6 @@ /* * Example implementation for the Git filter protocol version 2 - * See Documentation/gitattributes.txt, section "Filter Protocol" + * See Documentation/gitattributes.adoc, section "Filter Protocol" * * Usage: test-tool rot13-filter [--always-delay] --log= * From 8ea7d41f171fd52d3c920d40fdb390fd969a3270 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:26 -0500 Subject: [PATCH 058/312] t6012: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- t/t6012-rev-list-simplify.sh | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/t/t6012-rev-list-simplify.sh b/t/t6012-rev-list-simplify.sh index de1e87f1621d5d..4cecb6224cb464 100755 --- a/t/t6012-rev-list-simplify.sh +++ b/t/t6012-rev-list-simplify.sh @@ -177,7 +177,7 @@ test_expect_success '--full-diff is not affected by --parents' ' # \ / /\ / # `---X--' `---Y--' # -# This example is explained in Documentation/rev-list-options.txt +# This example is explained in Documentation/rev-list-options.adoc test_expect_success 'setup rebuild repo' ' rm -rf .git * && From 366074dc1843764c199c5b7a6354c58b8ac0b5c2 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:27 -0500 Subject: [PATCH 059/312] t6434: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- t/t6434-merge-recursive-rename-options.sh | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/t/t6434-merge-recursive-rename-options.sh b/t/t6434-merge-recursive-rename-options.sh index a11707835b4abe..6e913c30a136f5 100755 --- a/t/t6434-merge-recursive-rename-options.sh +++ b/t/t6434-merge-recursive-rename-options.sh @@ -22,7 +22,7 @@ R075 2-old 2-new R100 3-old 3-new Actual similarity indices are parsed from diff output. We rely on the fact that -they are rounded down (see, e.g., Documentation/diff-generate-patch.txt, which +they are rounded down (see, e.g., Documentation/diff-generate-patch.adoc, which mentions this in a different context). ' From 508cf7f5d8907a281602a54551abcc58a16d0493 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:28 -0500 Subject: [PATCH 060/312] trace2.h: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- trace2.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/trace2.h b/trace2.h index 901f39253a6593..e4f23784e4620f 100644 --- a/trace2.h +++ b/trace2.h @@ -31,7 +31,7 @@ * * For more info about: trace2 targets, conventions for public functions and * macros, trace2 target formats and examples on trace2 API usage refer to - * Documentation/technical/api-trace2.txt + * Documentation/technical/api-trace2.adoc * */ From 15db9a895d36a29eac2061ec31eba5e9628b4a5b Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:29 -0500 Subject: [PATCH 061/312] trace2/tr2_sysenv.c: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- trace2/tr2_sysenv.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/trace2/tr2_sysenv.c b/trace2/tr2_sysenv.c index 01379c5cad8438..4abc218514fdbc 100644 --- a/trace2/tr2_sysenv.c +++ b/trace2/tr2_sysenv.c @@ -7,7 +7,7 @@ /* * Each entry represents a trace2 setting. - * See Documentation/technical/api-trace2.txt + * See Documentation/technical/api-trace2.adoc */ struct tr2_sysenv_entry { const char *env_var_name; From ee00ef41f269bda33370acbacb41f5ca9a200a9b Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:30 -0500 Subject: [PATCH 062/312] transport.h: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- transport.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/transport.h b/transport.h index 44100fa9b7fdd6..892f19454a75d6 100644 --- a/transport.h +++ b/transport.h @@ -168,7 +168,7 @@ struct transport *transport_get(struct remote *, const char *); * Check whether a transport is allowed by the environment. * * Type should generally be the URL scheme, as described in - * Documentation/git.txt + * Documentation/git.adoc * * from_user specifies if the transport was given by the user. If unknown pass * a -1 to read from the environment to determine if the transport was given by From d6b67cefb5c27d1479d41a8fae1a03b8aece82f6 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:31 -0500 Subject: [PATCH 063/312] unpack-trees.c: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- unpack-trees.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/unpack-trees.c b/unpack-trees.c index 334cb84f6531b5..cf5b73c84be2ff 100644 --- a/unpack-trees.c +++ b/unpack-trees.c @@ -2904,7 +2904,7 @@ int threeway_merge(const struct cache_entry * const *stages, * The rule is to "carry forward" what is in the index without losing * information across a "fast-forward", favoring a successful merge * over a merge failure when it makes sense. For details of the - * "carry forward" rule, please see . + * "carry forward" rule, please see . * */ int twoway_merge(const struct cache_entry * const *src, From 61cd812130bda9c8996c90283501ee00c029c7a4 Mon Sep 17 00:00:00 2001 From: Todd Zullinger Date: Mon, 3 Mar 2025 15:44:32 -0500 Subject: [PATCH 064/312] xdiff: *.txt -> *.adoc fixes Signed-off-by: Todd Zullinger Signed-off-by: Junio C Hamano --- xdiff/xdiff.h | 2 +- xdiff/xpatience.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/xdiff/xdiff.h b/xdiff/xdiff.h index bb56b23f34c96c..2cecde5afe5da1 100644 --- a/xdiff/xdiff.h +++ b/xdiff/xdiff.h @@ -85,7 +85,7 @@ typedef struct s_xpparam { regex_t **ignore_regex; size_t ignore_regex_nr; - /* See Documentation/diff-options.txt. */ + /* See Documentation/diff-options.adoc. */ char **anchors; size_t anchors_nr; } xpparam_t; diff --git a/xdiff/xpatience.c b/xdiff/xpatience.c index 82f663004ec2ef..77dc411d1937ab 100644 --- a/xdiff/xpatience.c +++ b/xdiff/xpatience.c @@ -64,7 +64,7 @@ struct hashmap { /* * If 1, this entry can serve as an anchor. See - * Documentation/diff-options.txt for more information. + * Documentation/diff-options.adoc for more information. */ unsigned anchor : 1; } *entries, *first, *last; From 6dff5de1da7157de20310b10e5a0c09953dbb4bb Mon Sep 17 00:00:00 2001 From: Junio C Hamano Date: Mon, 3 Mar 2025 14:51:29 -0800 Subject: [PATCH 065/312] refs: show --no-reflog in the help text We forgot that we must keep the documentation and help text in sync. Signed-off-by: Junio C Hamano --- builtin/refs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/builtin/refs.c b/builtin/refs.c index c459507d514b37..44d592a94cf381 100644 --- a/builtin/refs.c +++ b/builtin/refs.c @@ -8,7 +8,7 @@ #include "worktree.h" #define REFS_MIGRATE_USAGE \ - N_("git refs migrate --ref-format= [--dry-run]") + N_("git refs migrate --ref-format= [--no-reflog] [--dry-run]") #define REFS_VERIFY_USAGE \ N_("git refs verify [--strict] [--verbose]") From bad79103998cd329584e844a315e85c3e1ac3839 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ren=C3=A9=20Scharfe?= Date: Tue, 4 Mar 2025 17:11:54 +0100 Subject: [PATCH 066/312] reftable: release name on reftable_reader_new() error MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit If block_source_read_block() or parse_footer() fail, we leak the "name" member of struct reftable_reader in reftable_reader_new(). Release it. Reported by: H Z Helped-by: Jeff King Signed-off-by: René Scharfe Signed-off-by: Junio C Hamano --- reftable/reader.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/reftable/reader.c b/reftable/reader.c index 3f2e4b280055ff..24bae50ac2fab3 100644 --- a/reftable/reader.c +++ b/reftable/reader.c @@ -666,6 +666,8 @@ int reftable_reader_new(struct reftable_reader **out, reftable_block_done(&footer); reftable_block_done(&header); if (err) { + if (r) + reftable_free(r->name); reftable_free(r); block_source_close(source); } From 92f8da8de36decbfcc4e30dc069b2c9ea28e3f6a Mon Sep 17 00:00:00 2001 From: Phillip Wood Date: Wed, 5 Mar 2025 10:42:37 +0000 Subject: [PATCH 067/312] docs: fix repository-layout when building with breaking changes Since commit 8ccc75c2452 (remote: announce removal of "branches/" and "remotes/", 2025-01-22) enabling WITH_BREAKING_CHANGES when building git removes support for reading branches from ".git/branches" and remotes from ".git/remotes". However those locations are still documented in gitrepository-layout.adoc even though the build does not support them. Rectify this by adding a new document attribute "with-breaking-changes" and use it to make the inclusion of those sections of the documentation conditional. Note that the name of the attribute does not match the test prerequisite WITHOUT_BREAKING_CHANGES added in c5bc9a7f94a (Makefile: wire up build option for deprecated features, 2025-01-22). This is to avoid the awkward double negative ifndef::without_breaking_changes for documentation that should be included when WITH_BREAKING_CHANGES is enabled. The test prerequisite will be renamed to match the documentation attribute in a future patch series. Signed-off-by: Phillip Wood Signed-off-by: Junio C Hamano --- Documentation/Makefile | 4 ++++ Documentation/gitrepository-layout.adoc | 4 ++++ Documentation/meson.build | 4 ++++ 3 files changed, 12 insertions(+) diff --git a/Documentation/Makefile b/Documentation/Makefile index c9a7cf662f0c43..671267a8ac7af6 100644 --- a/Documentation/Makefile +++ b/Documentation/Makefile @@ -224,6 +224,10 @@ asciidoc.conf: asciidoc.conf.in FORCE $(QUIET_GEN)$(call version_gen,"$(shell pwd)/..",$<,$@) endif +ifdef WITH_BREAKING_CHANGES +ASCIIDOC_EXTRA += -awith-breaking-changes +endif + ASCIIDOC_DEPS += docinfo.html SHELL_PATH ?= $(SHELL) diff --git a/Documentation/gitrepository-layout.adoc b/Documentation/gitrepository-layout.adoc index 6348ef1dcdf842..7421ef956d35ac 100644 --- a/Documentation/gitrepository-layout.adoc +++ b/Documentation/gitrepository-layout.adoc @@ -152,6 +152,7 @@ config.worktree:: working directory in multiple working directory setup (see linkgit:git-worktree[1]). +ifndef::with-breaking-changes[] branches:: A deprecated way to store shorthands to be used to specify a URL to 'git fetch', 'git pull' and 'git push'. @@ -164,6 +165,7 @@ branches:: "$GIT_COMMON_DIR/branches" will be used instead. + Git will stop reading remotes from this directory in Git 3.0. +endif::with-breaking-changes[] hooks:: Hooks are customization scripts used by various Git @@ -231,6 +233,7 @@ info/sparse-checkout:: This file stores sparse checkout patterns. See also: linkgit:git-read-tree[1]. +ifndef::with-breaking-changes[] remotes:: Stores shorthands for URL and default refnames for use when interacting with remote repositories via 'git fetch', @@ -241,6 +244,7 @@ remotes:: "$GIT_COMMON_DIR/remotes" will be used instead. + Git will stop reading remotes from this directory in Git 3.0. +endif::with-breaking-changes[] logs:: Records of changes made to refs are stored in this directory. diff --git a/Documentation/meson.build b/Documentation/meson.build index 0a0f2bfa14ab88..594546d68b12c1 100644 --- a/Documentation/meson.build +++ b/Documentation/meson.build @@ -284,6 +284,10 @@ elif docs_backend == 'asciidoctor' ] endif +if get_option('breaking_changes') + asciidoc_common_options += ['--attribute', 'with-breaking-changes'] +endif + xmlto = find_program('xmlto', dirs: program_path) cmd_lists = [ From e969bc875963a10890d61ba84eab3a460bd9e535 Mon Sep 17 00:00:00 2001 From: Junio C Hamano Date: Wed, 5 Mar 2025 10:37:53 -0800 Subject: [PATCH 068/312] A few more after -rc1 Signed-off-by: Junio C Hamano --- Documentation/RelNotes/2.49.0.adoc | 13 +++++++++++++ 1 file changed, 13 insertions(+) diff --git a/Documentation/RelNotes/2.49.0.adoc b/Documentation/RelNotes/2.49.0.adoc index 3d253455d373a9..a173eb193075c2 100644 --- a/Documentation/RelNotes/2.49.0.adoc +++ b/Documentation/RelNotes/2.49.0.adoc @@ -93,6 +93,18 @@ Performance, Internal Implementation, Development Support etc. * Rename processing in the recursive merge backend has seen a micro optimization. + * The path.[ch] API takes an explicit repository parameter passed + throughout the callchain, instead of relying on the_repository + singleton instance. + + * Large-object promisor protocol extension has been introduced. + + * The editorconfig file is updated to tell us that bash scripts are + similar to general Bourne shell scripts. + + * Meson-based build procedure forgot to build some docs, which has + been corrected. + Fixes since v2.48 ----------------- @@ -272,3 +284,4 @@ Fixes since v2.48 (merge 45761988ac en/doc-renormalize later to maint). (merge 832f56f06a jc/doc-boolean-synonyms later to maint). (merge 3eeed876a9 ac/doc-http-ssl-type-config later to maint). + (merge c268e3285d jc/breaking-changes-early-adopter-option later to maint). From 4478ad37a7d233b8db4d46dd563ece0bc8b00af4 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Thu, 6 Mar 2025 10:26:18 +0000 Subject: [PATCH 069/312] ident: stop assuming that `gw_gecos` is writable In 590e081dea7c (ident: add NO_GECOS_IN_PWENT for systems without pw_gecos in struct passwd, 2011-05-19), code was introduced to iterate over the `gw_gecos` field; The loop variable is of type `char *`, which assumes that `gw_gecos` is writable. However, it is not necessarily writable (and it is a bad idea to have it writable in the first place), so let's switch the loop variable type to `const char *`. This is not a new problem, but what is new is the Meson build. While it does not trigger in CI builds, imitating the commands of `ci/run-build-and-tests.sh` in a regular Git for Windows SDK (`meson setup build . --fatal-meson-warnings --warnlevel 2 --werror --wrap-mode nofallback -Dfuzzers=true` followed by `meson compile -C build --` results in this beautiful error: "cc" [...] -o libgit.a.p/ident.c.obj "-c" ../ident.c ../ident.c: In function 'copy_gecos': ../ident.c:68:18: error: assignment discards 'const' qualifier from pointer target type [-Werror=discarded-qualifiers] 68 | for (src = get_gecos(w); *src && *src != ','; src++) { | ^ cc1.exe: all warnings being treated as errors Now, why does this not trigger in CI? The answer is as simple as it is puzzling: The `win+Meson` job completely side-steps Git for Windows' development environment, opting instead to use the GCC that is on the `PATH` in GitHub-hosted `windows-latest` runners. That GCC is pinned to v12.2.0 and targets the UCRT (unlikely to change any time soon, see https://github.com/actions/runner-images/blob/win25/20250303.1/images/windows/toolsets/toolset-2022.json#L132-L141). That is in stark contrast to Git for Windows, which uses GCC v14.2.0 and targets MSVCRT. Git for Windows' `Makefile`-based build also obviously uses different compiler flags, otherwise this compile error would have had plenty of opportunity in almost 14 years to surface. In other words, contrary to my expectations, the `win+Meson` job is ill-equipped to replace the `win build` job because it exercises a completely different tool version/compiler flags vector than what Git for Windows needs. Nevertheless, there is currently this huge push, including breaking changes after -rc1 and all, for switching to Meson. Therefore, we need to make it work, somehow, even in Git for Windows' SDK, hence this patch, at this point in time. Signed-off-by: Johannes Schindelin Signed-off-by: Junio C Hamano --- ident.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/ident.c b/ident.c index caf41fb2a98f56..967895d885033d 100644 --- a/ident.c +++ b/ident.c @@ -59,7 +59,7 @@ static struct passwd *xgetpwuid_self(int *is_bogus) static void copy_gecos(const struct passwd *w, struct strbuf *name) { - char *src; + const char *src; /* Traditionally GECOS field had office phone numbers etc, separated * with commas. Also & stands for capitalized form of the login name. From 31761f391192909f6486ecb532e72dfaee756be6 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Thu, 6 Mar 2025 10:26:19 +0000 Subject: [PATCH 070/312] meson: fix sorting In 904339edbd80 (Introduce support for the Meson build system, 2024-12-06) the `meson.build` file was introduced, adding also a Windows-specific list of source files. This list was obviously meant to be sorted alphabetically, but there is one mistake. Let's fix that. Signed-off-by: Johannes Schindelin Signed-off-by: Junio C Hamano --- meson.build | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/meson.build b/meson.build index e86085b0a474a1..efe2871c9dba13 100644 --- a/meson.build +++ b/meson.build @@ -1109,11 +1109,11 @@ elif host_machine.system() == 'windows' libgit_sources += [ 'compat/mingw.c', 'compat/winansi.c', + 'compat/win32/dirent.c', 'compat/win32/flush.c', 'compat/win32/path-utils.c', 'compat/win32/pthread.c', 'compat/win32/syslog.c', - 'compat/win32/dirent.c', 'compat/win32mmap.c', 'compat/nedmalloc/nedmalloc.c', ] From 970916368764347c38d36c727b2e6e0086e784e3 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Thu, 6 Mar 2025 10:26:20 +0000 Subject: [PATCH 071/312] cmake: generalize the handling of the `CLAR_TEST_OBJS` list A late-comer to the v2.49.0 party, `sk/unit-test-oid`, added yet another array item to `CLAR_TEST_OBJS`, causing the `win+VS build` job to fail with symptoms like this one: unit-tests-lib.lib(u-oid-array.obj) : error LNK2019: unresolved external symbol cl_parse_any_oid referenced in function fill_array This is a similar scenario to the one that forced me to write 8afda42fce60 (cmake: generalize the handling of the `UNIT_TEST_OBJS` list, 2024-09-18): The hard-coded echo of `CLAR_TEST_OBJS` in `CMakeLists.txt` that recapitulates faithfully what was already hard-coded in `Makefile` would either have to be updated whack-a-mole style, or generalized. Just like I chose the latter option for `UNIT_TEST_OBJS`, I now do the same for `CLAR_TEST_OBJS`. Signed-off-by: Johannes Schindelin Signed-off-by: Junio C Hamano --- contrib/buildsystems/CMakeLists.txt | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-) diff --git a/contrib/buildsystems/CMakeLists.txt b/contrib/buildsystems/CMakeLists.txt index c6fbd57e158bc1..25b495fa737d07 100644 --- a/contrib/buildsystems/CMakeLists.txt +++ b/contrib/buildsystems/CMakeLists.txt @@ -1001,10 +1001,14 @@ parse_makefile_for_sources(unit-test_SOURCES ${CMAKE_SOURCE_DIR}/Makefile "UNIT_ list(TRANSFORM unit-test_SOURCES REPLACE "\\$\\(UNIT_TEST_DIR\\)/" "${CMAKE_SOURCE_DIR}/t/unit-tests/") add_library(unit-test-lib STATIC ${unit-test_SOURCES}) +parse_makefile_for_sources(clar-test_SOURCES ${CMAKE_SOURCE_DIR}/Makefile "CLAR_TEST_OBJS") +list(TRANSFORM clar-test_SOURCES REPLACE "\\$\\(UNIT_TEST_DIR\\)/" "${CMAKE_SOURCE_DIR}/t/unit-tests/") +add_library(clar-test-lib STATIC ${clar-test_SOURCES}) + parse_makefile_for_scripts(unit_test_PROGRAMS "UNIT_TEST_PROGRAMS" "") foreach(unit_test ${unit_test_PROGRAMS}) add_executable("${unit_test}" "${CMAKE_SOURCE_DIR}/t/unit-tests/${unit_test}.c") - target_link_libraries("${unit_test}" unit-test-lib common-main) + target_link_libraries("${unit_test}" unit-test-lib clar-test-lib common-main) set_target_properties("${unit_test}" PROPERTIES RUNTIME_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/t/unit-tests/bin) if(MSVC) @@ -1046,13 +1050,13 @@ add_custom_command(OUTPUT "${CMAKE_BINARY_DIR}/t/unit-tests/clar.suite" VERBATIM) add_library(unit-tests-lib ${clar_test_SUITES} - "${CMAKE_SOURCE_DIR}/t/unit-tests/clar/clar.c" "${CMAKE_BINARY_DIR}/t/unit-tests/clar-decls.h" "${CMAKE_BINARY_DIR}/t/unit-tests/clar.suite" ) +target_include_directories(clar-test-lib PUBLIC "${CMAKE_BINARY_DIR}/t/unit-tests") target_include_directories(unit-tests-lib PUBLIC "${CMAKE_BINARY_DIR}/t/unit-tests") -add_executable(unit-tests "${CMAKE_SOURCE_DIR}/t/unit-tests/unit-test.c") -target_link_libraries(unit-tests unit-tests-lib common-main) +add_executable(unit-tests) +target_link_libraries(unit-tests unit-tests-lib clar-test-lib common-main) set_target_properties(unit-tests PROPERTIES RUNTIME_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/t/unit-tests/bin) if(MSVC) From bd52d9a0583a2d12f584fdf6b47d2c2f51c0d791 Mon Sep 17 00:00:00 2001 From: Taylor Blau Date: Fri, 7 Mar 2025 18:27:03 -0500 Subject: [PATCH 072/312] fetch: fix following tags when fetching specific OID In 3f763ddf28 (fetch: set remote/HEAD if it does not exist, 2024-11-22), unconditionally adds "HEAD" to the list of ref prefixes we send to the server. This breaks a core assumption that the list of prefixes we send to the server is complete. We must either send all prefixes we care about, or none at all (in the latter case the server then advertises everything). The tag following code is careful to only add "refs/tags/" to the list of prefixes if there are already entries in the prefix list. But because the new code from 3f763ddf28 runs after the tag code, and because it unconditionally adds to the prefix list, we may end up with a prefix list that _should_ have "refs/tags/" in it, but doesn't. When that is the case, the server does not advertise any tags, and our auto-following breaks because we never learned about any tags in the first place. Fix this by only adding "HEAD" to the ref prefixes when we know that we are already limiting the advertisement. In either case we'll learn about HEAD (either through the limited advertisement, or implicitly through a full advertisement). Reported-by: Igor Todorovski Co-authored-by: Jeff King Signed-off-by: Jeff King Signed-off-by: Taylor Blau Signed-off-by: Junio C Hamano --- builtin/fetch.c | 4 +++- t/t5503-tagfollow.sh | 14 ++++++++++++++ 2 files changed, 17 insertions(+), 1 deletion(-) diff --git a/builtin/fetch.c b/builtin/fetch.c index fe2b26c74aecab..a06d13050165fc 100644 --- a/builtin/fetch.c +++ b/builtin/fetch.c @@ -1777,7 +1777,9 @@ static int do_fetch(struct transport *transport, if (uses_remote_tracking(transport, rs)) { must_list_refs = 1; - strvec_push(&transport_ls_refs_options.ref_prefixes, "HEAD"); + if (transport_ls_refs_options.ref_prefixes.nr) + strvec_push(&transport_ls_refs_options.ref_prefixes, + "HEAD"); } if (must_list_refs) { diff --git a/t/t5503-tagfollow.sh b/t/t5503-tagfollow.sh index 195fc64dd44ae7..845ca43ea0a7d7 100755 --- a/t/t5503-tagfollow.sh +++ b/t/t5503-tagfollow.sh @@ -160,4 +160,18 @@ test_expect_success 'new clone fetch main and tags' ' test_cmp expect actual ' +test_expect_success 'fetch specific OID with tag following' ' + git init --bare clone3.git && + ( + cd clone3.git && + git remote add origin .. && + git fetch origin $B:refs/heads/main && + + git -C .. for-each-ref >expect && + git for-each-ref >actual && + + test_cmp expect actual + ) +' + test_done From 87a0bdbf0f72b7561f3cd50636eee33dcb7dbcc3 Mon Sep 17 00:00:00 2001 From: Junio C Hamano Date: Mon, 10 Mar 2025 08:47:08 -0700 Subject: [PATCH 073/312] Git 2.49-rc2 Signed-off-by: Junio C Hamano --- GIT-VERSION-GEN | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/GIT-VERSION-GEN b/GIT-VERSION-GEN index 7fb7fff6dda300..8979fb2803285a 100755 --- a/GIT-VERSION-GEN +++ b/GIT-VERSION-GEN @@ -1,6 +1,6 @@ #!/bin/sh -DEF_VER=v2.49.0-rc1 +DEF_VER=v2.49.0-rc2 LF=' ' From 1240cfe162e2e090e54ea0627cd2e9a1ab1c8083 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Wed, 6 Nov 2024 20:34:50 +0100 Subject: [PATCH 074/312] sideband: mask control characters The output of `git clone` is a vital component for understanding what has happened when things go wrong. However, these logs are partially under the control of the remote server (via the "sideband", which typically contains what the remote `git pack-objects` process sends to `stderr`), and is currently not sanitized by Git. This makes Git susceptible to ANSI escape sequence injection (see CWE-150, https://cwe.mitre.org/data/definitions/150.html), which allows attackers to corrupt terminal state, to hide information, and even to insert characters into the input buffer (i.e. as if the user had typed those characters). To plug this vulnerability, disallow any control character in the sideband, replacing them instead with the common `^` (e.g. `^[` for `\x1b`, `^A` for `\x01`). There is likely a need for more fine-grained controls instead of using a "heavy hammer" like this, which will be introduced subsequently. Signed-off-by: Johannes Schindelin --- sideband.c | 17 +++++++++++++++-- t/t5409-colorize-remote-messages.sh | 12 ++++++++++++ 2 files changed, 27 insertions(+), 2 deletions(-) diff --git a/sideband.c b/sideband.c index 251e9615ed0239..81b1ff0805b3f6 100644 --- a/sideband.c +++ b/sideband.c @@ -66,6 +66,19 @@ void list_config_color_sideband_slots(struct string_list *list, const char *pref list_config_item(list, prefix, keywords[i].keyword); } +static void strbuf_add_sanitized(struct strbuf *dest, const char *src, int n) +{ + strbuf_grow(dest, n); + for (; n && *src; src++, n--) { + if (!iscntrl(*src) || *src == '\t' || *src == '\n') + strbuf_addch(dest, *src); + else { + strbuf_addch(dest, '^'); + strbuf_addch(dest, 0x40 + *src); + } + } +} + /* * Optionally highlight one keyword in remote output if it appears at the start * of the line. This should be called for a single line only, which is @@ -81,7 +94,7 @@ static void maybe_colorize_sideband(struct strbuf *dest, const char *src, int n) int i; if (!want_color_stderr(use_sideband_colors())) { - strbuf_add(dest, src, n); + strbuf_add_sanitized(dest, src, n); return; } @@ -114,7 +127,7 @@ static void maybe_colorize_sideband(struct strbuf *dest, const char *src, int n) } } - strbuf_add(dest, src, n); + strbuf_add_sanitized(dest, src, n); } diff --git a/t/t5409-colorize-remote-messages.sh b/t/t5409-colorize-remote-messages.sh index fa5de4500a4f50..d0745c391b2625 100755 --- a/t/t5409-colorize-remote-messages.sh +++ b/t/t5409-colorize-remote-messages.sh @@ -98,4 +98,16 @@ test_expect_success 'fallback to color.ui' ' grep "error: error" decoded ' +test_expect_success 'disallow (color) control sequences in sideband' ' + write_script .git/color-me-surprised <<-\EOF && + printf "error: Have you \\033[31mread\\033[m this?\\n" >&2 + exec "$@" + EOF + test_config_global uploadPack.packObjectshook ./color-me-surprised && + test_commit need-at-least-one-commit && + git clone --no-local . throw-away 2>stderr && + test_decode_color decoded && + test_grep ! RED decoded +' + test_done From 48428d73941e7602f4c19c47d9dbd188aa99b785 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Wed, 6 Nov 2024 21:07:51 +0100 Subject: [PATCH 075/312] sideband: introduce an "escape hatch" to allow control characters The preceding commit fixed the vulnerability whereas sideband messages (that are under the control of the remote server) could contain ANSI escape sequences that would be sent to the terminal verbatim. However, this fix may not be desirable under all circumstances, e.g. when remote servers deliberately add coloring to their messages to increase their urgency. To help with those use cases, give users a way to opt-out of the protections: `sideband.allowControlCharacters`. Signed-off-by: Johannes Schindelin --- Documentation/config.adoc | 2 ++ Documentation/config/sideband.adoc | 5 +++++ sideband.c | 10 ++++++++++ t/t5409-colorize-remote-messages.sh | 8 +++++++- 4 files changed, 24 insertions(+), 1 deletion(-) create mode 100644 Documentation/config/sideband.adoc diff --git a/Documentation/config.adoc b/Documentation/config.adoc index cc769251be339b..a8b04c4e5183d2 100644 --- a/Documentation/config.adoc +++ b/Documentation/config.adoc @@ -522,6 +522,8 @@ include::config/sequencer.adoc[] include::config/showbranch.adoc[] +include::config/sideband.adoc[] + include::config/sparse.adoc[] include::config/splitindex.adoc[] diff --git a/Documentation/config/sideband.adoc b/Documentation/config/sideband.adoc new file mode 100644 index 00000000000000..3fb5045cd79581 --- /dev/null +++ b/Documentation/config/sideband.adoc @@ -0,0 +1,5 @@ +sideband.allowControlCharacters:: + By default, control characters that are delivered via the sideband + are masked, to prevent potentially unwanted ANSI escape sequences + from being sent to the terminal. Use this config setting to override + this behavior. diff --git a/sideband.c b/sideband.c index 81b1ff0805b3f6..d1c326fa191068 100644 --- a/sideband.c +++ b/sideband.c @@ -26,6 +26,8 @@ static struct keyword_entry keywords[] = { { "error", GIT_COLOR_BOLD_RED }, }; +static int allow_control_characters; + /* Returns a color setting (GIT_COLOR_NEVER, etc). */ static int use_sideband_colors(void) { @@ -39,6 +41,9 @@ static int use_sideband_colors(void) if (use_sideband_colors_cached >= 0) return use_sideband_colors_cached; + git_config_get_bool("sideband.allowcontrolcharacters", + &allow_control_characters); + if (!git_config_get_string_tmp(key, &value)) use_sideband_colors_cached = git_config_colorbool(key, value); else if (!git_config_get_string_tmp("color.ui", &value)) @@ -68,6 +73,11 @@ void list_config_color_sideband_slots(struct string_list *list, const char *pref static void strbuf_add_sanitized(struct strbuf *dest, const char *src, int n) { + if (allow_control_characters) { + strbuf_add(dest, src, n); + return; + } + strbuf_grow(dest, n); for (; n && *src; src++, n--) { if (!iscntrl(*src) || *src == '\t' || *src == '\n') diff --git a/t/t5409-colorize-remote-messages.sh b/t/t5409-colorize-remote-messages.sh index d0745c391b2625..fb31e8525418a1 100755 --- a/t/t5409-colorize-remote-messages.sh +++ b/t/t5409-colorize-remote-messages.sh @@ -105,9 +105,15 @@ test_expect_success 'disallow (color) control sequences in sideband' ' EOF test_config_global uploadPack.packObjectshook ./color-me-surprised && test_commit need-at-least-one-commit && + git clone --no-local . throw-away 2>stderr && test_decode_color decoded && - test_grep ! RED decoded + test_grep ! RED decoded && + + rm -rf throw-away && + git -c sideband.allowControlCharacters clone --no-local . throw-away 2>stderr && + test_decode_color decoded && + test_grep RED decoded ' test_done From 5b8e1b4299dcec5a78ed30cb00e342976ab16cb1 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Mon, 18 Nov 2024 21:42:57 +0100 Subject: [PATCH 076/312] sideband: do allow ANSI color sequences by default The preceding two commits introduced special handling of the sideband channel to neutralize ANSI escape sequences before sending the payload to the terminal, and `sideband.allowControlCharacters` to override that behavior. However, some `pre-receive` hooks that are actively used in practice want to color their messages and therefore rely on the fact that Git passes them through to the terminal. In contrast to other ANSI escape sequences, it is highly unlikely that coloring sequences can be essential tools in attack vectors that mislead Git users e.g. by hiding crucial information. Therefore we can have both: Continue to allow ANSI coloring sequences to be passed to the terminal, and neutralize all other ANSI escape sequences. Signed-off-by: Johannes Schindelin --- Documentation/config/sideband.adoc | 17 ++++++-- sideband.c | 61 ++++++++++++++++++++++++++--- t/t5409-colorize-remote-messages.sh | 16 +++++++- 3 files changed, 84 insertions(+), 10 deletions(-) diff --git a/Documentation/config/sideband.adoc b/Documentation/config/sideband.adoc index 3fb5045cd79581..f347fd6b33004a 100644 --- a/Documentation/config/sideband.adoc +++ b/Documentation/config/sideband.adoc @@ -1,5 +1,16 @@ sideband.allowControlCharacters:: By default, control characters that are delivered via the sideband - are masked, to prevent potentially unwanted ANSI escape sequences - from being sent to the terminal. Use this config setting to override - this behavior. + are masked, except ANSI color sequences. This prevents potentially + unwanted ANSI escape sequences from being sent to the terminal. Use + this config setting to override this behavior: ++ +-- + color:: + Allow ANSI color sequences, line feeds and horizontal tabs, + but mask all other control characters. This is the default. + false:: + Mask all control characters other than line feeds and + horizontal tabs. + true:: + Allow all control characters to be sent to the terminal. +-- diff --git a/sideband.c b/sideband.c index d1c326fa191068..9084ca234dcecc 100644 --- a/sideband.c +++ b/sideband.c @@ -26,7 +26,11 @@ static struct keyword_entry keywords[] = { { "error", GIT_COLOR_BOLD_RED }, }; -static int allow_control_characters; +static enum { + ALLOW_NO_CONTROL_CHARACTERS = 0, + ALLOW_ALL_CONTROL_CHARACTERS = 1, + ALLOW_ANSI_COLOR_SEQUENCES = 2 +} allow_control_characters = ALLOW_ANSI_COLOR_SEQUENCES; /* Returns a color setting (GIT_COLOR_NEVER, etc). */ static int use_sideband_colors(void) @@ -41,8 +45,24 @@ static int use_sideband_colors(void) if (use_sideband_colors_cached >= 0) return use_sideband_colors_cached; - git_config_get_bool("sideband.allowcontrolcharacters", - &allow_control_characters); + switch (git_config_get_maybe_bool("sideband.allowcontrolcharacters", &i)) { + case 0: /* Boolean value */ + allow_control_characters = i ? ALLOW_ALL_CONTROL_CHARACTERS : + ALLOW_NO_CONTROL_CHARACTERS; + break; + case -1: /* non-Boolean value */ + if (git_config_get_string_tmp("sideband.allowcontrolcharacters", + &value)) + ; /* huh? `get_maybe_bool()` returned -1 */ + else if (!strcmp(value, "color")) + allow_control_characters = ALLOW_ANSI_COLOR_SEQUENCES; + else + warning(_("unrecognized value for `sideband." + "allowControlCharacters`: '%s'"), value); + break; + default: + break; /* not configured */ + } if (!git_config_get_string_tmp(key, &value)) use_sideband_colors_cached = git_config_colorbool(key, value); @@ -71,9 +91,37 @@ void list_config_color_sideband_slots(struct string_list *list, const char *pref list_config_item(list, prefix, keywords[i].keyword); } +static int handle_ansi_color_sequence(struct strbuf *dest, const char *src, int n) +{ + int i; + + /* + * Valid ANSI color sequences are of the form + * + * ESC [ [ [; ]*] m + */ + + if (allow_control_characters != ALLOW_ANSI_COLOR_SEQUENCES || + n < 3 || src[0] != '\x1b' || src[1] != '[') + return 0; + + for (i = 2; i < n; i++) { + if (src[i] == 'm') { + strbuf_add(dest, src, i + 1); + return i; + } + if (!isdigit(src[i]) && src[i] != ';') + break; + } + + return 0; +} + static void strbuf_add_sanitized(struct strbuf *dest, const char *src, int n) { - if (allow_control_characters) { + int i; + + if (allow_control_characters == ALLOW_ALL_CONTROL_CHARACTERS) { strbuf_add(dest, src, n); return; } @@ -82,7 +130,10 @@ static void strbuf_add_sanitized(struct strbuf *dest, const char *src, int n) for (; n && *src; src++, n--) { if (!iscntrl(*src) || *src == '\t' || *src == '\n') strbuf_addch(dest, *src); - else { + else if ((i = handle_ansi_color_sequence(dest, src, n))) { + src += i; + n -= i; + } else { strbuf_addch(dest, '^'); strbuf_addch(dest, 0x40 + *src); } diff --git a/t/t5409-colorize-remote-messages.sh b/t/t5409-colorize-remote-messages.sh index fb31e8525418a1..a755c49a74e634 100755 --- a/t/t5409-colorize-remote-messages.sh +++ b/t/t5409-colorize-remote-messages.sh @@ -100,7 +100,7 @@ test_expect_success 'fallback to color.ui' ' test_expect_success 'disallow (color) control sequences in sideband' ' write_script .git/color-me-surprised <<-\EOF && - printf "error: Have you \\033[31mread\\033[m this?\\n" >&2 + printf "error: Have you \\033[31mread\\033[m this?\\a\\n" >&2 exec "$@" EOF test_config_global uploadPack.packObjectshook ./color-me-surprised && @@ -108,12 +108,24 @@ test_expect_success 'disallow (color) control sequences in sideband' ' git clone --no-local . throw-away 2>stderr && test_decode_color decoded && + test_grep RED decoded && + test_grep "\\^G" stderr && + tr -dc "\\007" actual && + test_must_be_empty actual && + + rm -rf throw-away && + git -c sideband.allowControlCharacters=false \ + clone --no-local . throw-away 2>stderr && + test_decode_color decoded && test_grep ! RED decoded && + test_grep "\\^G" stderr && rm -rf throw-away && git -c sideband.allowControlCharacters clone --no-local . throw-away 2>stderr && test_decode_color decoded && - test_grep RED decoded + test_grep RED decoded && + tr -dc "\\007" actual && + test_file_not_empty actual ' test_done From d1f8c8fe3d27b34aac19d6c0d83ddff63264ad48 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Wed, 30 Oct 2024 19:48:46 +0100 Subject: [PATCH 077/312] unix-socket: avoid leak when initialization fails When a Unix socket is initialized, the current directory's path is stored so that the cleanup code can `chdir()` back to where it was before exit. If the path that needs to be stored exceeds the default size of the `sun_path` attribute of `struct sockaddr_un` (which is defined as a 108-sized byte array on Linux), a larger buffer needs to be allocated so that it can hold the path, and it is the responsibility of the `unix_sockaddr_cleanup()` function to release that allocated memory. In Git's CI, this stack allocation is not necessary because the code is checked out to `/home/runner/work/git/git`. Concatenate the path `t/trash directory.t0301-credential-cache/.cache/git/credential/socket` and a terminating NUL, and you end up with 96 bytes, 12 shy of the default `sun_path` size. However, I use worktrees with slightly longer paths: `/home/me/projects/git/yes/i/nest/worktrees/to/organize/them/` is more in line with what I have. When I recently tried to locally reproduce a failure of the `linux-leaks` CI job, this t0301 test failed (where it had not failed in CI). The reason: When `credential-cache` tries to reach its daemon initially by calling `unix_sockaddr_init()`, it is expected that the daemon cannot be reached (the idea is to spin up the daemon in that case and try again). However, when this first call to `unix_sockaddr_init()` fails, the code returns early from the `unix_stream_connect()` function _without_ giving the cleanup code a chance to run, skipping the deallocation of above-mentioned path. The fix is easy: do not return early but instead go directly to the cleanup code. Signed-off-by: Johannes Schindelin --- unix-socket.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/unix-socket.c b/unix-socket.c index 8860203c3f46dc..1fa0cf6c15c721 100644 --- a/unix-socket.c +++ b/unix-socket.c @@ -84,7 +84,7 @@ int unix_stream_connect(const char *path, int disallow_chdir) struct unix_sockaddr_context ctx; if (unix_sockaddr_init(&sa, path, &ctx, disallow_chdir) < 0) - return -1; + goto fail; fd = socket(AF_UNIX, SOCK_STREAM, 0); if (fd < 0) goto fail; From 0c8ecff4ba35fd4127667223708e698b1152880f Mon Sep 17 00:00:00 2001 From: Jeff King Date: Mon, 13 Jan 2025 01:26:01 -0500 Subject: [PATCH 078/312] grep: prevent `^$` false match at end of file In some implementations, `regexec_buf()` assumes that it is fed lines; Without `REG_NOTEOL` it thinks the end of the buffer is the end of a line. Which makes sense, but trips up this case because we are not feeding lines, but rather a whole buffer. So the final newline is not the start of an empty line, but the true end of the buffer. This causes an interesting bug: $ echo content >file.txt $ git grep --no-index -n '^$' file.txt file.txt:2: This bug is fixed by making the end of the buffer consistently the end of the final line. The patch was applied from https://lore.kernel.org/git/20250113062601.GD767856@coredump.intra.peff.net/ Reported-by: Olly Betts Signed-off-by: Johannes Schindelin --- grep.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/grep.c b/grep.c index 4e155ee9e66367..c4bb9f10814024 100644 --- a/grep.c +++ b/grep.c @@ -1646,6 +1646,8 @@ static int grep_source_1(struct grep_opt *opt, struct grep_source *gs, int colle bol = gs->buf; left = gs->size; + if (left && gs->buf[left-1] == '\n') + left--; while (left) { const char *eol; int hit; From dc465c4c0b7923a41a8ee45cce2ba38898bbd3ef Mon Sep 17 00:00:00 2001 From: Sverre Rabbelier Date: Sun, 24 Jul 2011 15:54:04 +0200 Subject: [PATCH 079/312] t9350: point out that refs are not updated correctly This happens only when the corresponding commits are not exported in the current fast-export run. This can happen either when the relevant commit is already marked, or when the commit is explicitly marked as UNINTERESTING with a negative ref by another argument. This breaks fast-export basec remote helpers. Signed-off-by: Sverre Rabbelier --- t/t9350-fast-export.sh | 11 +++++++++++ 1 file changed, 11 insertions(+) diff --git a/t/t9350-fast-export.sh b/t/t9350-fast-export.sh index 40427883ec6dbf..abffd267e259c4 100755 --- a/t/t9350-fast-export.sh +++ b/t/t9350-fast-export.sh @@ -801,4 +801,15 @@ test_expect_success 'fast-export handles --end-of-options' ' test_cmp expect actual ' +cat > expected << EOF +reset refs/heads/master +from $(git rev-parse master) + +EOF + +test_expect_failure 'refs are updated even if no commits need to be exported' ' + git fast-export master..master > actual && + test_cmp expected actual +' + test_done From b3241645d008c8467a19ab6273493b87c17fd25a Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Sun, 10 Jul 2022 01:15:08 +0200 Subject: [PATCH 080/312] ci: stop linking the `prove` cache It is not useful because we do not have any persisted directory anymore, not since dropping our Travis CI support. Signed-off-by: Johannes Schindelin --- ci/run-build-and-tests.sh | 5 ----- ci/run-test-slice.sh | 5 ----- 2 files changed, 10 deletions(-) diff --git a/ci/run-build-and-tests.sh b/ci/run-build-and-tests.sh index 1c698467235943..066d4c4c8547e7 100755 --- a/ci/run-build-and-tests.sh +++ b/ci/run-build-and-tests.sh @@ -5,11 +5,6 @@ . ${0%/*}/lib.sh -case "$CI_OS_NAME" in -windows*) cmd //c mklink //j t\\.prove "$(cygpath -aw "$cache_dir/.prove")";; -*) ln -s "$cache_dir/.prove" t/.prove;; -esac - run_tests=t case "$jobname" in diff --git a/ci/run-test-slice.sh b/ci/run-test-slice.sh index e167e646f79e3d..0444c79c023c82 100755 --- a/ci/run-test-slice.sh +++ b/ci/run-test-slice.sh @@ -5,11 +5,6 @@ . ${0%/*}/lib.sh -case "$CI_OS_NAME" in -windows*) cmd //c mklink //j t\\.prove "$(cygpath -aw "$cache_dir/.prove")";; -*) ln -s "$cache_dir/.prove" t/.prove;; -esac - group "Run tests" make --quiet -C t T="$(cd t && ./helper/test-tool path-utils slice-tests "$1" "$2" t[0-9]*.sh | tr '\n' ' ')" || From 25a3a0d03e12bb0fa90f12c3da11dd8a09fbc566 Mon Sep 17 00:00:00 2001 From: Sverre Rabbelier Date: Sat, 28 Aug 2010 20:49:01 -0500 Subject: [PATCH 081/312] transport-helper: add trailing -- [PT: ensure we add an additional element to the argv array] Signed-off-by: Sverre Rabbelier Signed-off-by: Johannes Schindelin --- transport-helper.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/transport-helper.c b/transport-helper.c index d457b425501a74..fed5d51a2b3d34 100644 --- a/transport-helper.c +++ b/transport-helper.c @@ -499,6 +499,8 @@ static int get_exporter(struct transport *transport, for (size_t i = 0; i < revlist_args->nr; i++) strvec_push(&fastexport->args, revlist_args->items[i].string); + strvec_push(&fastexport->args, "--"); + fastexport->git_cmd = 1; return start_command(fastexport); } From 2eeeea5b29ab91d9d2986e75582ab5f61344c4c6 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Thu, 30 Jan 2025 12:15:41 +0100 Subject: [PATCH 082/312] contrib/buildsystems: drop support for building .vcproj/.vcxproj files Before we had CMake support, the only way to build Git in Visual Studio was via this hacky `generate` script. For a while I tried to fix whenever things got broken, in particular to allow building confidence in embargoed releases by running the CI builds in Azure Pipelines in a private Azure DevOps project. I even carried the patches in Git for Windows with the intention of upstreaming them, eventually. However, it is a lot of work with too little benefit. CMake is much better supported by Visual Studio. So let's drop this hacky script (plus support code). Signed-off-by: Johannes Schindelin --- contrib/buildsystems/Generators.pm | 42 -- contrib/buildsystems/Generators/QMake.pm | 189 ------- contrib/buildsystems/Generators/Vcproj.pm | 579 --------------------- contrib/buildsystems/Generators/Vcxproj.pm | 402 -------------- contrib/buildsystems/engine.pl | 395 -------------- contrib/buildsystems/generate | 29 -- contrib/buildsystems/parse.pl | 228 -------- 7 files changed, 1864 deletions(-) delete mode 100644 contrib/buildsystems/Generators.pm delete mode 100644 contrib/buildsystems/Generators/QMake.pm delete mode 100644 contrib/buildsystems/Generators/Vcproj.pm delete mode 100644 contrib/buildsystems/Generators/Vcxproj.pm delete mode 100755 contrib/buildsystems/engine.pl delete mode 100755 contrib/buildsystems/generate delete mode 100755 contrib/buildsystems/parse.pl diff --git a/contrib/buildsystems/Generators.pm b/contrib/buildsystems/Generators.pm deleted file mode 100644 index aa4cbaa2adacb1..00000000000000 --- a/contrib/buildsystems/Generators.pm +++ /dev/null @@ -1,42 +0,0 @@ -package Generators; -require Exporter; - -use strict; -use File::Basename; -no strict 'refs'; -use vars qw($VERSION @AVAILABLE); - -our $VERSION = '1.00'; -our(@ISA, @EXPORT, @EXPORT_OK, @AVAILABLE); -@ISA = qw(Exporter); - -BEGIN { - local(*D); - my $me = $INC{"Generators.pm"}; - die "Couldn't find myself in \@INC, which is required to load the generators!" if ("$me" eq ""); - $me = dirname($me); - if (opendir(D,"$me/Generators")) { - foreach my $gen (readdir(D)) { - next unless ($gen =~ /\.pm$/); - require "${me}/Generators/$gen"; - $gen =~ s,\.pm,,; - push(@AVAILABLE, $gen); - } - closedir(D); - my $gens = join(', ', @AVAILABLE); - } - - push @EXPORT_OK, qw(available); -} - -sub available { - return @AVAILABLE; -} - -sub generate { - my ($gen, $git_dir, $out_dir, $rel_dir, %build_structure) = @_; - return eval("Generators::${gen}::generate(\$git_dir, \$out_dir, \$rel_dir, \%build_structure)") if grep(/^$gen$/, @AVAILABLE); - die "Generator \"${gen}\" is not available!\nAvailable generators are: @AVAILABLE\n"; -} - -1; diff --git a/contrib/buildsystems/Generators/QMake.pm b/contrib/buildsystems/Generators/QMake.pm deleted file mode 100644 index ff3b657e610575..00000000000000 --- a/contrib/buildsystems/Generators/QMake.pm +++ /dev/null @@ -1,189 +0,0 @@ -package Generators::QMake; -require Exporter; - -use strict; -use vars qw($VERSION); - -our $VERSION = '1.00'; -our(@ISA, @EXPORT, @EXPORT_OK, @AVAILABLE); -@ISA = qw(Exporter); - -BEGIN { - push @EXPORT_OK, qw(generate); -} - -sub generate { - my ($git_dir, $out_dir, $rel_dir, %build_structure) = @_; - - my @libs = @{$build_structure{"LIBS"}}; - foreach (@libs) { - createLibProject($_, $git_dir, $out_dir, $rel_dir, %build_structure); - } - - my @apps = @{$build_structure{"APPS"}}; - foreach (@apps) { - createAppProject($_, $git_dir, $out_dir, $rel_dir, %build_structure); - } - - createGlueProject($git_dir, $out_dir, $rel_dir, %build_structure); - return 0; -} - -sub createLibProject { - my ($libname, $git_dir, $out_dir, $rel_dir, %build_structure) = @_; - print "Generate $libname lib project\n"; - $rel_dir = "../$rel_dir"; - - my $sources = join(" \\\n\t", sort(map("$rel_dir/$_", @{$build_structure{"LIBS_${libname}_SOURCES"}}))); - my $defines = join(" \\\n\t", sort(@{$build_structure{"LIBS_${libname}_DEFINES"}})); - my $includes= join(" \\\n\t", sort(map("$rel_dir/$_", @{$build_structure{"LIBS_${libname}_INCLUDES"}}))); - my $cflags = join(" ", sort(@{$build_structure{"LIBS_${libname}_CFLAGS"}})); - - my $cflags_debug = $cflags; - $cflags_debug =~ s/-MT/-MTd/; - $cflags_debug =~ s/-O.//; - - my $cflags_release = $cflags; - $cflags_release =~ s/-MTd/-MT/; - - my @tmp = @{$build_structure{"LIBS_${libname}_LFLAGS"}}; - my @tmp2 = (); - foreach (@tmp) { - if (/^-LTCG/) { - } elsif (/^-L/) { - $_ =~ s/^-L/-LIBPATH:$rel_dir\//; - } - push(@tmp2, $_); - } - my $lflags = join(" ", sort(@tmp)); - - my $target = $libname; - $target =~ s/\//_/g; - $defines =~ s/-D//g; - $defines =~ s/"/\\\\"/g; - $includes =~ s/-I//g; - mkdir "$target" || die "Could not create the directory $target for lib project!\n"; - open F, ">$target/$target.pro" || die "Could not open $target/$target.pro for writing!\n"; - print F << "EOM"; -TEMPLATE = lib -TARGET = $target -DESTDIR = $rel_dir - -CONFIG -= qt -CONFIG += static - -QMAKE_CFLAGS = -QMAKE_CFLAGS_RELEASE = $cflags_release -QMAKE_CFLAGS_DEBUG = $cflags_debug -QMAKE_LIBFLAGS = $lflags - -DEFINES += \\ - $defines - -INCLUDEPATH += \\ - $includes - -SOURCES += \\ - $sources -EOM - close F; -} - -sub createAppProject { - my ($appname, $git_dir, $out_dir, $rel_dir, %build_structure) = @_; - print "Generate $appname app project\n"; - $rel_dir = "../$rel_dir"; - - my $sources = join(" \\\n\t", sort(map("$rel_dir/$_", @{$build_structure{"APPS_${appname}_SOURCES"}}))); - my $defines = join(" \\\n\t", sort(@{$build_structure{"APPS_${appname}_DEFINES"}})); - my $includes= join(" \\\n\t", sort(map("$rel_dir/$_", @{$build_structure{"APPS_${appname}_INCLUDES"}}))); - my $cflags = join(" ", sort(@{$build_structure{"APPS_${appname}_CFLAGS"}})); - - my $cflags_debug = $cflags; - $cflags_debug =~ s/-MT/-MTd/; - $cflags_debug =~ s/-O.//; - - my $cflags_release = $cflags; - $cflags_release =~ s/-MTd/-MT/; - - my $libs; - foreach (sort(@{$build_structure{"APPS_${appname}_LIBS"}})) { - $_ =~ s/\//_/g; - $libs .= " $_"; - } - my @tmp = @{$build_structure{"APPS_${appname}_LFLAGS"}}; - my @tmp2 = (); - foreach (@tmp) { - # next if ($_ eq "-NODEFAULTLIB:MSVCRT.lib"); - if (/^-LTCG/) { - } elsif (/^-L/) { - $_ =~ s/^-L/-LIBPATH:$rel_dir\//; - } - push(@tmp2, $_); - } - my $lflags = join(" ", sort(@tmp)); - - my $target = $appname; - $target =~ s/\.exe//; - $target =~ s/\//_/g; - $defines =~ s/-D//g; - $defines =~ s/"/\\\\"/g; - $includes =~ s/-I//g; - mkdir "$target" || die "Could not create the directory $target for app project!\n"; - open F, ">$target/$target.pro" || die "Could not open $target/$target.pro for writing!\n"; - print F << "EOM"; -TEMPLATE = app -TARGET = $target -DESTDIR = $rel_dir - -CONFIG -= qt embed_manifest_exe -CONFIG += console - -QMAKE_CFLAGS = -QMAKE_CFLAGS_RELEASE = $cflags_release -QMAKE_CFLAGS_DEBUG = $cflags_debug -QMAKE_LFLAGS = $lflags -LIBS = $libs - -DEFINES += \\ - $defines - -INCLUDEPATH += \\ - $includes - -win32:QMAKE_LFLAGS += -LIBPATH:$rel_dir -else: QMAKE_LFLAGS += -L$rel_dir - -SOURCES += \\ - $sources -EOM - close F; -} - -sub createGlueProject { - my ($git_dir, $out_dir, $rel_dir, %build_structure) = @_; - my $libs = join(" \\ \n", map("\t$_|$_.pro", @{$build_structure{"LIBS"}})); - my $apps = join(" \\ \n", map("\t$_|$_.pro", @{$build_structure{"APPS"}})); - $libs =~ s/\.a//g; - $libs =~ s/\//_/g; - $libs =~ s/\|/\//g; - $apps =~ s/\.exe//g; - $apps =~ s/\//_/g; - $apps =~ s/\|/\//g; - - my $filename = $out_dir; - $filename =~ s/.*\/([^\/]+)$/$1/; - $filename =~ s/\/$//; - print "Generate glue project $filename.pro\n"; - open F, ">$filename.pro" || die "Could not open $filename.pro for writing!\n"; - print F << "EOM"; -TEMPLATE = subdirs -CONFIG += ordered -SUBDIRS += \\ -$libs \\ -$apps -EOM - close F; -} - -1; diff --git a/contrib/buildsystems/Generators/Vcproj.pm b/contrib/buildsystems/Generators/Vcproj.pm deleted file mode 100644 index 737647e76afd42..00000000000000 --- a/contrib/buildsystems/Generators/Vcproj.pm +++ /dev/null @@ -1,579 +0,0 @@ -package Generators::Vcproj; -require Exporter; - -use strict; -use vars qw($VERSION); -use Digest::SHA qw(sha256_hex); - -our $VERSION = '1.00'; -our(@ISA, @EXPORT, @EXPORT_OK, @AVAILABLE); -@ISA = qw(Exporter); - -BEGIN { - push @EXPORT_OK, qw(generate); -} - -sub generate_guid ($) { - my $hex = sha256_hex($_[0]); - $hex =~ s/^(.{8})(.{4})(.{4})(.{4})(.{12}).*/{$1-$2-$3-$4-$5}/; - $hex =~ tr/a-z/A-Z/; - return $hex; -} - -sub generate { - my ($git_dir, $out_dir, $rel_dir, %build_structure) = @_; - my @libs = @{$build_structure{"LIBS"}}; - foreach (@libs) { - createLibProject($_, $git_dir, $out_dir, $rel_dir, \%build_structure); - } - - my @apps = @{$build_structure{"APPS"}}; - foreach (@apps) { - createAppProject($_, $git_dir, $out_dir, $rel_dir, \%build_structure); - } - - createGlueProject($git_dir, $out_dir, $rel_dir, %build_structure); - return 0; -} - -sub createLibProject { - my ($libname, $git_dir, $out_dir, $rel_dir, $build_structure) = @_; - print "Generate $libname vcproj lib project\n"; - $rel_dir = "..\\$rel_dir"; - $rel_dir =~ s/\//\\/g; - - my $target = $libname; - $target =~ s/\//_/g; - $target =~ s/\.a//; - - my $uuid = generate_guid($libname); - $$build_structure{"LIBS_${target}_GUID"} = $uuid; - - my @srcs = sort(map("$rel_dir\\$_", @{$$build_structure{"LIBS_${libname}_SOURCES"}})); - my @sources; - foreach (@srcs) { - $_ =~ s/\//\\/g; - push(@sources, $_); - } - my $defines = join(",", sort(@{$$build_structure{"LIBS_${libname}_DEFINES"}})); - my $includes= join(";", sort(map(""$rel_dir\\$_"", @{$$build_structure{"LIBS_${libname}_INCLUDES"}}))); - my $cflags = join(" ", sort(@{$$build_structure{"LIBS_${libname}_CFLAGS"}})); - $cflags =~ s/\"/"/g; - $cflags =~ s//>/g; - - my $cflags_debug = $cflags; - $cflags_debug =~ s/-MT/-MTd/; - $cflags_debug =~ s/-O.//; - - my $cflags_release = $cflags; - $cflags_release =~ s/-MTd/-MT/; - - my @tmp = @{$$build_structure{"LIBS_${libname}_LFLAGS"}}; - my @tmp2 = (); - foreach (@tmp) { - if (/^-LTCG/) { - } elsif (/^-L/) { - $_ =~ s/^-L/-LIBPATH:$rel_dir\//; - } - push(@tmp2, $_); - } - my $lflags = join(" ", sort(@tmp)); - - $defines =~ s/-D//g; - $defines =~ s/\"/\\"/g; - $defines =~ s//>/g; - $defines =~ s/\'//g; - $includes =~ s/-I//g; - mkdir "$target" || die "Could not create the directory $target for lib project!\n"; - open F, ">$target/$target.vcproj" || die "Could not open $target/$target.pro for writing!\n"; - binmode F, ":crlf"; - print F << "EOM"; - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -EOM - foreach(@sources) { - print F << "EOM"; - -EOM - } - print F << "EOM"; - - - - - -EOM - close F; -} - -sub createAppProject { - my ($appname, $git_dir, $out_dir, $rel_dir, $build_structure) = @_; - print "Generate $appname vcproj app project\n"; - $rel_dir = "..\\$rel_dir"; - $rel_dir =~ s/\//\\/g; - - my $target = $appname; - $target =~ s/\//_/g; - $target =~ s/\.exe//; - - my $uuid = generate_guid($appname); - $$build_structure{"APPS_${target}_GUID"} = $uuid; - - my @srcs = sort(map("$rel_dir\\$_", @{$$build_structure{"APPS_${appname}_SOURCES"}})); - my @sources; - foreach (@srcs) { - $_ =~ s/\//\\/g; - push(@sources, $_); - } - my $defines = join(",", sort(@{$$build_structure{"APPS_${appname}_DEFINES"}})); - my $includes= join(";", sort(map(""$rel_dir\\$_"", @{$$build_structure{"APPS_${appname}_INCLUDES"}}))); - my $cflags = join(" ", sort(@{$$build_structure{"APPS_${appname}_CFLAGS"}})); - $cflags =~ s/\"/"/g; - $cflags =~ s//>/g; - - my $cflags_debug = $cflags; - $cflags_debug =~ s/-MT/-MTd/; - $cflags_debug =~ s/-O.//; - - my $cflags_release = $cflags; - $cflags_release =~ s/-MTd/-MT/; - - my $libs; - foreach (sort(@{$$build_structure{"APPS_${appname}_LIBS"}})) { - $_ =~ s/\//_/g; - $libs .= " $_"; - } - my @tmp = @{$$build_structure{"APPS_${appname}_LFLAGS"}}; - my @tmp2 = (); - foreach (@tmp) { - if (/^-LTCG/) { - } elsif (/^-L/) { - $_ =~ s/^-L/-LIBPATH:$rel_dir\//; - } - push(@tmp2, $_); - } - my $lflags = join(" ", sort(@tmp)) . " -LIBPATH:$rel_dir"; - - $defines =~ s/-D//g; - $defines =~ s/\"/\\"/g; - $defines =~ s//>/g; - $defines =~ s/\'//g; - $defines =~ s/\\\\/\\/g; - $includes =~ s/-I//g; - mkdir "$target" || die "Could not create the directory $target for lib project!\n"; - open F, ">$target/$target.vcproj" || die "Could not open $target/$target.pro for writing!\n"; - binmode F, ":crlf"; - print F << "EOM"; - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -EOM - foreach(@sources) { - print F << "EOM"; - -EOM - } - print F << "EOM"; - - - - - -EOM - close F; -} - -sub createGlueProject { - my ($git_dir, $out_dir, $rel_dir, %build_structure) = @_; - print "Generate solutions file\n"; - $rel_dir = "..\\$rel_dir"; - $rel_dir =~ s/\//\\/g; - my $SLN_HEAD = "Microsoft Visual Studio Solution File, Format Version 10.00\n# Visual Studio 2008\n"; - my $SLN_PRE = "Project(\"{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}\") = "; - my $SLN_POST = "\nEndProject\n"; - - my @libs = @{$build_structure{"LIBS"}}; - my @tmp; - foreach (@libs) { - $_ =~ s/\//_/g; - $_ =~ s/\.a//; - push(@tmp, $_); - } - @libs = @tmp; - - my @apps = @{$build_structure{"APPS"}}; - @tmp = (); - foreach (@apps) { - $_ =~ s/\//_/g; - $_ =~ s/\.exe//; - if ($_ eq "git" ) { - unshift(@tmp, $_); - } else { - push(@tmp, $_); - } - } - @apps = @tmp; - - open F, ">git.sln" || die "Could not open git.sln for writing!\n"; - binmode F, ":crlf"; - print F "$SLN_HEAD"; - - my $uuid_libgit = $build_structure{"LIBS_libgit_GUID"}; - my $uuid_xdiff_lib = $build_structure{"LIBS_xdiff_lib_GUID"}; - foreach (@apps) { - my $appname = $_; - my $uuid = $build_structure{"APPS_${appname}_GUID"}; - print F "$SLN_PRE"; - print F "\"${appname}\", \"${appname}\\${appname}.vcproj\", \"${uuid}\"\n"; - print F " ProjectSection(ProjectDependencies) = postProject\n"; - print F " ${uuid_libgit} = ${uuid_libgit}\n"; - print F " ${uuid_xdiff_lib} = ${uuid_xdiff_lib}\n"; - print F " EndProjectSection"; - print F "$SLN_POST"; - } - foreach (@libs) { - my $libname = $_; - my $uuid = $build_structure{"LIBS_${libname}_GUID"}; - print F "$SLN_PRE"; - print F "\"${libname}\", \"${libname}\\${libname}.vcproj\", \"${uuid}\""; - print F "$SLN_POST"; - } - - print F << "EOM"; -Global - GlobalSection(SolutionConfigurationPlatforms) = preSolution - Debug|Win32 = Debug|Win32 - Release|Win32 = Release|Win32 - EndGlobalSection -EOM - print F << "EOM"; - GlobalSection(ProjectConfigurationPlatforms) = postSolution -EOM - foreach (@apps) { - my $appname = $_; - my $uuid = $build_structure{"APPS_${appname}_GUID"}; - print F "\t\t${uuid}.Debug|Win32.ActiveCfg = Debug|Win32\n"; - print F "\t\t${uuid}.Debug|Win32.Build.0 = Debug|Win32\n"; - print F "\t\t${uuid}.Release|Win32.ActiveCfg = Release|Win32\n"; - print F "\t\t${uuid}.Release|Win32.Build.0 = Release|Win32\n"; - } - foreach (@libs) { - my $libname = $_; - my $uuid = $build_structure{"LIBS_${libname}_GUID"}; - print F "\t\t${uuid}.Debug|Win32.ActiveCfg = Debug|Win32\n"; - print F "\t\t${uuid}.Debug|Win32.Build.0 = Debug|Win32\n"; - print F "\t\t${uuid}.Release|Win32.ActiveCfg = Release|Win32\n"; - print F "\t\t${uuid}.Release|Win32.Build.0 = Release|Win32\n"; - } - - print F << "EOM"; - EndGlobalSection -EndGlobal -EOM - close F; -} - -1; diff --git a/contrib/buildsystems/Generators/Vcxproj.pm b/contrib/buildsystems/Generators/Vcxproj.pm deleted file mode 100644 index b2e68a16715e39..00000000000000 --- a/contrib/buildsystems/Generators/Vcxproj.pm +++ /dev/null @@ -1,402 +0,0 @@ -package Generators::Vcxproj; -require Exporter; - -use strict; -use vars qw($VERSION); -use Digest::SHA qw(sha256_hex); - -our $VERSION = '1.00'; -our(@ISA, @EXPORT, @EXPORT_OK, @AVAILABLE); -@ISA = qw(Exporter); - -BEGIN { - push @EXPORT_OK, qw(generate); -} - -sub generate_guid ($) { - my $hex = sha256_hex($_[0]); - $hex =~ s/^(.{8})(.{4})(.{4})(.{4})(.{12}).*/{$1-$2-$3-$4-$5}/; - $hex =~ tr/a-z/A-Z/; - return $hex; -} - -sub generate { - my ($git_dir, $out_dir, $rel_dir, %build_structure) = @_; - my @libs = @{$build_structure{"LIBS"}}; - foreach (@libs) { - createProject($_, $git_dir, $out_dir, $rel_dir, \%build_structure, 1); - } - - my @apps = @{$build_structure{"APPS"}}; - foreach (@apps) { - createProject($_, $git_dir, $out_dir, $rel_dir, \%build_structure, 0); - } - - createGlueProject($git_dir, $out_dir, $rel_dir, %build_structure); - return 0; -} - -sub createProject { - my ($name, $git_dir, $out_dir, $rel_dir, $build_structure, $static_library) = @_; - my $label = $static_library ? "lib" : "app"; - my $prefix = $static_library ? "LIBS_" : "APPS_"; - my $config_type = $static_library ? "StaticLibrary" : "Application"; - print "Generate $name vcxproj $label project\n"; - my $cdup = $name; - $cdup =~ s/[^\/]+/../g; - $cdup =~ s/\//\\/g; - $rel_dir = $rel_dir eq "." ? $cdup : "$cdup\\$rel_dir"; - $rel_dir =~ s/\//\\/g; - - my $target = $name; - if ($static_library) { - $target =~ s/\.a//; - } else { - $target =~ s/\.exe//; - } - - my $uuid = generate_guid($name); - $$build_structure{"$prefix${target}_GUID"} = $uuid; - my $vcxproj = $target; - $vcxproj =~ s/(.*\/)?(.*)/$&\/$2.vcxproj/; - $vcxproj =~ s/([^\/]*)(\/lib)\/(lib.vcxproj)/$1$2\/$1_$3/; - $$build_structure{"$prefix${target}_VCXPROJ"} = $vcxproj; - - my @srcs = sort(map("$rel_dir\\$_", @{$$build_structure{"$prefix${name}_SOURCES"}})); - my @sources; - foreach (@srcs) { - $_ =~ s/\//\\/g; - push(@sources, $_); - } - my $defines = join(";", sort(@{$$build_structure{"$prefix${name}_DEFINES"}})); - my $includes= join(";", sort(map { s/^-I//; s/\//\\/g; File::Spec->file_name_is_absolute($_) ? $_ : "$rel_dir\\$_" } @{$$build_structure{"$prefix${name}_INCLUDES"}})); - my $cflags = join(" ", sort(map { s/^-[GLMOWZ].*//; s/.* .*/"$&"/; $_; } @{$$build_structure{"$prefix${name}_CFLAGS"}})); - $cflags =~ s//>/g; - - my $libs_release = "\n "; - my $libs_debug = "\n "; - if (!$static_library && $name ne 'headless-git') { - $libs_release = join(";", sort(grep /^(?!libgit\.lib|xdiff\/lib\.lib|vcs-svn\/lib\.lib|reftable\/libreftable\.lib)/, @{$$build_structure{"$prefix${name}_LIBS"}})); - $libs_debug = $libs_release; - $libs_debug =~ s/zlib\.lib/zlibd\.lib/g; - $libs_debug =~ s/libexpat\.lib/libexpatd\.lib/g; - $libs_debug =~ s/libcurl\.lib/libcurl-d\.lib/g; - } - - $defines =~ s/-D//g; - $defines =~ s//>/g; - $defines =~ s/\'//g; - - die "Could not create the directory $target for $label project!\n" unless (-d "$target" || mkdir "$target"); - - open F, ">$vcxproj" or die "Could not open $vcxproj for writing!\n"; - binmode F, ":crlf :utf8"; - print F chr(0xFEFF); - print F << "EOM"; - - - - - Debug - Win32 - - - Release - Win32 - - - Debug - x64 - - - Release - x64 - - - - $uuid - Win32Proj - x86-windows - x64-windows - $cdup\\compat\\vcbuild\\vcpkg\\installed\\\$(VCPKGArch) - \$(VCPKGArchDirectory)\\debug\\bin - \$(VCPKGArchDirectory)\\debug\\lib - \$(VCPKGArchDirectory)\\bin - \$(VCPKGArchDirectory)\\lib - \$(VCPKGArchDirectory)\\include - $libs_debug - $libs_release - - - - true - true - - - false - true - - - $config_type - v140 - - ..\\ - - - - - - - - - - - - - false - true - - - - $cflags %(AdditionalOptions) - $cdup;$cdup\\compat;$cdup\\compat\\regex;$cdup\\compat\\win32;$cdup\\compat\\poll;$cdup\\compat\\vcbuild\\include;\$(VCPKGIncludeDirectory);%(AdditionalIncludeDirectories) - - OnlyExplicitInline - - ProgramDatabase - - - true - - - \$(VCPKGLibDirectory);%(AdditionalLibraryDirectories) - \$(VCPKGLibs);\$(AdditionalDependencies) - invalidcontinue.obj %(AdditionalOptions) - wmainCRTStartup - $cdup\\compat\\win32\\git.manifest - Console - -EOM - if ($target eq 'libgit') { - print F << "EOM"; - - Initialize VCPKG - del "$cdup\\compat\\vcbuild\\vcpkg" - call "$cdup\\compat\\vcbuild\\vcpkg_install.bat" - -EOM - } - print F << "EOM"; - - - - MachineX86 - - - - - Disabled - WIN32;_DEBUG;$defines;%(PreprocessorDefinitions) - MultiThreadedDebugDLL - - - true - - - - - MaxSpeed - true - WIN32;NDEBUG;$defines;%(PreprocessorDefinitions) - MultiThreadedDLL - true - Speed - - - true - true - true - - - -EOM - foreach(@sources) { - print F << "EOM"; - -EOM - } - print F << "EOM"; - -EOM - if ((!$static_library || $target =~ 'vcs-svn' || $target =~ 'xdiff') && !($name =~ /headless-git/)) { - my $uuid_libgit = $$build_structure{"LIBS_libgit_GUID"}; - my $uuid_libreftable = $$build_structure{"LIBS_reftable/libreftable_GUID"}; - my $uuid_xdiff_lib = $$build_structure{"LIBS_xdiff/lib_GUID"}; - - print F << "EOM"; - - - $uuid_libgit - false - -EOM - if (!($name =~ /xdiff|libreftable/)) { - print F << "EOM"; - - $uuid_libreftable - false - -EOM - } - if (!($name =~ 'xdiff')) { - print F << "EOM"; - - $uuid_xdiff_lib - false - -EOM - } - if ($name =~ /(test-(line-buffer|svn-fe)|^git-remote-testsvn)\.exe$/) { - my $uuid_vcs_svn_lib = $$build_structure{"LIBS_vcs-svn/lib_GUID"}; - print F << "EOM"; - - $uuid_vcs_svn_lib - false - -EOM - } - print F << "EOM"; - -EOM - } - print F << "EOM"; - -EOM - if (!$static_library) { - print F << "EOM"; - - - - - - - -EOM - } - if ($target eq 'git') { - print F " \n"; - } - if ($target eq 'git-remote-http') { - print F " \n"; - } - print F << "EOM"; - -EOM - close F; -} - -sub createGlueProject { - my ($git_dir, $out_dir, $rel_dir, %build_structure) = @_; - print "Generate solutions file\n"; - $rel_dir = "..\\$rel_dir"; - $rel_dir =~ s/\//\\/g; - my $SLN_HEAD = "Microsoft Visual Studio Solution File, Format Version 11.00\n# Visual Studio 2010\n"; - my $SLN_PRE = "Project(\"{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}\") = "; - my $SLN_POST = "\nEndProject\n"; - - my @libs = @{$build_structure{"LIBS"}}; - my @tmp; - foreach (@libs) { - $_ =~ s/\.a//; - push(@tmp, $_); - } - @libs = @tmp; - - my @apps = @{$build_structure{"APPS"}}; - @tmp = (); - foreach (@apps) { - $_ =~ s/\.exe//; - if ($_ eq "git" ) { - unshift(@tmp, $_); - } else { - push(@tmp, $_); - } - } - @apps = @tmp; - - open F, ">git.sln" || die "Could not open git.sln for writing!\n"; - binmode F, ":crlf :utf8"; - print F chr(0xFEFF); - print F "$SLN_HEAD"; - - foreach (@apps) { - my $appname = $_; - my $uuid = $build_structure{"APPS_${appname}_GUID"}; - print F "$SLN_PRE"; - my $vcxproj = $build_structure{"APPS_${appname}_VCXPROJ"}; - $vcxproj =~ s/\//\\/g; - $appname =~ s/.*\///; - print F "\"${appname}\", \"${vcxproj}\", \"${uuid}\""; - print F "$SLN_POST"; - } - foreach (@libs) { - my $libname = $_; - my $uuid = $build_structure{"LIBS_${libname}_GUID"}; - print F "$SLN_PRE"; - my $vcxproj = $build_structure{"LIBS_${libname}_VCXPROJ"}; - $vcxproj =~ s/\//\\/g; - $libname =~ s/\//_/g; - print F "\"${libname}\", \"${vcxproj}\", \"${uuid}\""; - print F "$SLN_POST"; - } - - print F << "EOM"; -Global - GlobalSection(SolutionConfigurationPlatforms) = preSolution - Debug|x64 = Debug|x64 - Debug|x86 = Debug|x86 - Release|x64 = Release|x64 - Release|x86 = Release|x86 - EndGlobalSection -EOM - print F << "EOM"; - GlobalSection(ProjectConfigurationPlatforms) = postSolution -EOM - foreach (@apps) { - my $appname = $_; - my $uuid = $build_structure{"APPS_${appname}_GUID"}; - print F "\t\t${uuid}.Debug|x64.ActiveCfg = Debug|x64\n"; - print F "\t\t${uuid}.Debug|x64.Build.0 = Debug|x64\n"; - print F "\t\t${uuid}.Debug|x86.ActiveCfg = Debug|Win32\n"; - print F "\t\t${uuid}.Debug|x86.Build.0 = Debug|Win32\n"; - print F "\t\t${uuid}.Release|x64.ActiveCfg = Release|x64\n"; - print F "\t\t${uuid}.Release|x64.Build.0 = Release|x64\n"; - print F "\t\t${uuid}.Release|x86.ActiveCfg = Release|Win32\n"; - print F "\t\t${uuid}.Release|x86.Build.0 = Release|Win32\n"; - } - foreach (@libs) { - my $libname = $_; - my $uuid = $build_structure{"LIBS_${libname}_GUID"}; - print F "\t\t${uuid}.Debug|x64.ActiveCfg = Debug|x64\n"; - print F "\t\t${uuid}.Debug|x64.Build.0 = Debug|x64\n"; - print F "\t\t${uuid}.Debug|x86.ActiveCfg = Debug|Win32\n"; - print F "\t\t${uuid}.Debug|x86.Build.0 = Debug|Win32\n"; - print F "\t\t${uuid}.Release|x64.ActiveCfg = Release|x64\n"; - print F "\t\t${uuid}.Release|x64.Build.0 = Release|x64\n"; - print F "\t\t${uuid}.Release|x86.ActiveCfg = Release|Win32\n"; - print F "\t\t${uuid}.Release|x86.Build.0 = Release|Win32\n"; - } - - print F << "EOM"; - EndGlobalSection - GlobalSection(SolutionProperties) = preSolution - HideSolutionNode = FALSE - EndGlobalSection -EndGlobal -EOM - close F; -} - -1; diff --git a/contrib/buildsystems/engine.pl b/contrib/buildsystems/engine.pl deleted file mode 100755 index 069be7e4befcd7..00000000000000 --- a/contrib/buildsystems/engine.pl +++ /dev/null @@ -1,395 +0,0 @@ -#!/usr/bin/perl -w -###################################################################### -# Do not call this script directly! -# -# The generate script ensures that @INC is correct before the engine -# is executed. -# -# Copyright (C) 2009 Marius Storm-Olsen -###################################################################### -use strict; -use File::Basename; -use File::Spec; -use Cwd; -use Generators; -use Text::ParseWords; - -my (%build_structure, %compile_options, @makedry); -my $out_dir = getcwd(); -my $git_dir = $out_dir; -$git_dir =~ s=\\=/=g; -$git_dir = dirname($git_dir) while (!-e "$git_dir/git.c" && "$git_dir" ne ""); -die "Couldn't find Git repo" if ("$git_dir" eq ""); - -my @gens = Generators::available(); -my $gen = "Vcproj"; - -sub showUsage -{ - my $genlist = join(', ', @gens); - print << "EOM"; -generate usage: - -g --gen Specify the buildsystem generator (default: $gen) - Available: $genlist - -o --out Specify output directory generation (default: .) - --make-out Write the output of GNU Make into a file - -i --in Specify input file, instead of running GNU Make - -h,-? --help This help -EOM - exit 0; -} - -# Parse command-line options -my $make_out; -while (@ARGV) { - my $arg = shift @ARGV; - if ("$arg" eq "-h" || "$arg" eq "--help" || "$arg" eq "-?") { - showUsage(); - exit(0); - } elsif("$arg" eq "--out" || "$arg" eq "-o") { - $out_dir = shift @ARGV; - } elsif("$arg" eq "--make-out") { - $make_out = shift @ARGV; - } elsif("$arg" eq "--gen" || "$arg" eq "-g") { - $gen = shift @ARGV; - } elsif("$arg" eq "--in" || "$arg" eq "-i") { - my $infile = shift @ARGV; - open(F, "<$infile") || die "Couldn't open file $infile"; - @makedry = ; - close(F); - } else { - die "Unknown option: " . $arg; - } -} - -# NOT using File::Spec->rel2abs($path, $base) here, as -# it fails badly for me in the msysgit environment -$git_dir = File::Spec->rel2abs($git_dir); -$out_dir = File::Spec->rel2abs($out_dir); -my $rel_dir = makeOutRel2Git($git_dir, $out_dir); - -# Print some information so the user feels informed -print << "EOM"; ------ -Generator: $gen -Git dir: $git_dir -Out dir: $out_dir ------ -Running GNU Make to figure out build structure... -EOM - -# Pipe a make --dry-run into a variable, if not already loaded from file -# Capture the make dry stderr to file for review (will be empty for a release build). - -my $ErrsFile = "msvc-build-makedryerrors.txt"; -@makedry = `make -C $git_dir -n MSVC=1 SKIP_VCPKG=1 V=1 2>$ErrsFile` -if !@makedry; -# test for an empty Errors file and remove it -unlink $ErrsFile if -f -z $ErrsFile; - -if (defined $make_out) { - open OUT, ">" . $make_out; - print OUT @makedry; - close OUT; -} - -# Parse the make output into usable info -parseMakeOutput(); - -# Finally, ask the generator to start generating.. -Generators::generate($gen, $git_dir, $out_dir, $rel_dir, %build_structure); - -# main flow ends here -# ------------------------------------------------------------------------------------------------- - - -# 1) path: /foo/bar/baz 2) path: /foo/bar/baz 3) path: /foo/bar/baz -# base: /foo/bar/baz/temp base: /foo/bar base: /tmp -# rel: .. rel: baz rel: ../foo/bar/baz -sub makeOutRel2Git -{ - my ($path, $base) = @_; - my $rel; - if ("$path" eq "$base") { - return "."; - } elsif ($base =~ /^$path/) { - # case 1 - my $tmp = $base; - $tmp =~ s/^$path//; - foreach (split('/', $tmp)) { - $rel .= "../" if ("$_" ne ""); - } - } elsif ($path =~ /^$base/) { - # case 2 - $rel = $path; - $rel =~ s/^$base//; - $rel = "./$rel"; - } else { - my $tmp = $base; - foreach (split('/', $tmp)) { - $rel .= "../" if ("$_" ne ""); - } - $rel .= $path; - } - $rel =~ s/\/\//\//g; # simplify - $rel =~ s/\/$//; # don't end with / - return $rel; -} - -sub parseMakeOutput -{ - print "Parsing GNU Make output to figure out build structure...\n"; - my $line = 0; - while (my $text = shift @makedry) { - my $ate_next; - do { - $ate_next = 0; - $line++; - chomp $text; - chop $text if ($text =~ /\r$/); - if ($text =~ /\\$/) { - $text =~ s/\\$//; - $text .= shift @makedry; - $ate_next = 1; - } - } while($ate_next); - - if ($text =~ /^test /) { - # options to test (eg -o) may be mistaken for linker options - next; - } - - if ($text =~ /^(mkdir|msgfmt) /) { - # options to the Portable Object translations - # the line "mkdir ... && msgfmt ..." contains no linker options - next; - } - - if($text =~ / -c /) { - # compilation - handleCompileLine($text, $line); - - } elsif ($text =~ / -o /) { - # linking executable - handleLinkLine($text, $line); - - } elsif ($text =~ /\.o / && $text =~ /\.a /) { - # libifying - handleLibLine($text, $line); -# -# } elsif ($text =~ /^cp /) { -# # copy file around -# -# } elsif ($text =~ /^rm -f /) { -# # shell command -# -# } elsif ($text =~ /^make[ \[]/) { -# # make output -# -# } elsif ($text =~ /^echo /) { -# # echo to file -# -# } elsif ($text =~ /^if /) { -# # shell conditional -# -# } elsif ($text =~ /^tclsh /) { -# # translation stuff -# -# } elsif ($text =~ /^umask /) { -# # handling boilerplates -# -# } elsif ($text =~ /\$\(\:\)/) { -# # ignore -# -# } elsif ($text =~ /^FLAGS=/) { -# # flags check for dependencies -# -# } elsif ($text =~ /^'\/usr\/bin\/perl' -MError -e/) { -# # perl commands for copying files -# -# } elsif ($text =~ /generate-cmdlist\.sh/) { -# # command for generating list of commands -# -# } elsif ($text =~ /new locations or Tcl/) { -# # command for detecting Tcl/Tk changes -# -# } elsif ($text =~ /mkdir -p/) { -# # command creating path -# -# } elsif ($text =~ /: no custom templates yet/) { -# # whatever -# -# } else { -# print "Unhandled (line: $line): $text\n"; - } - } - -# use Data::Dumper; -# print "Parsed build structure:\n"; -# print Dumper(%build_structure); -} - -# variables for the compilation part of each step -my (@defines, @incpaths, @cflags, @sources); - -sub clearCompileStep -{ - @defines = (); - @incpaths = (); - @cflags = (); - @sources = (); -} - -sub removeDuplicates -{ - my (%dupHash, $entry); - %dupHash = map { $_, 1 } @defines; - @defines = keys %dupHash; - - %dupHash = map { $_, 1 } @incpaths; - @incpaths = keys %dupHash; - - %dupHash = map { $_, 1 } @cflags; - @cflags = keys %dupHash; -} - -sub handleCompileLine -{ - my ($line, $lineno) = @_; - my @parts = shellwords($line); - my $sourcefile; - shift(@parts); # ignore cmd - while (my $part = shift @parts) { - if ("$part" eq "-o") { - # ignore object file - shift @parts; - } elsif ("$part" eq "-c") { - # ignore compile flag - } elsif ("$part" eq "-c") { - } elsif ($part =~ /^.?-I/) { - push(@incpaths, $part); - } elsif ($part =~ /^.?-D/) { - push(@defines, $part); - } elsif ($part =~ /^-/) { - push(@cflags, $part); - } elsif ($part =~ /\.(c|cc|cpp)$/) { - $sourcefile = $part; - } else { - die "Unhandled compiler option @ line $lineno: $part"; - } - } - @{$compile_options{"${sourcefile}_CFLAGS"}} = @cflags; - @{$compile_options{"${sourcefile}_DEFINES"}} = @defines; - @{$compile_options{"${sourcefile}_INCPATHS"}} = @incpaths; - clearCompileStep(); -} - -sub handleLibLine -{ - my ($line, $lineno) = @_; - my (@objfiles, @lflags, $libout, $part); - # kill cmd and rm 'prefix' - $line =~ s/^rm -f .* && .* rcs //; - my @parts = shellwords($line); - while ($part = shift @parts) { - if ($part =~ /^-/) { - push(@lflags, $part); - } elsif ($part =~ /\.(o|obj)$/) { - push(@objfiles, $part); - } elsif ($part =~ /\.(a|lib)$/) { - $libout = $part; - $libout =~ s/\.a$//; - } else { - die "Unhandled lib option @ line $lineno: $part"; - } - } -# print "LibOut: '$libout'\nLFlags: @lflags\nOfiles: @objfiles\n"; -# exit(1); - foreach (@objfiles) { - my $sourcefile = $_; - $sourcefile =~ s/\.o$/.c/; - push(@sources, $sourcefile); - push(@cflags, @{$compile_options{"${sourcefile}_CFLAGS"}}); - push(@defines, @{$compile_options{"${sourcefile}_DEFINES"}}); - push(@incpaths, @{$compile_options{"${sourcefile}_INCPATHS"}}); - } - removeDuplicates(); - - push(@{$build_structure{"LIBS"}}, $libout); - @{$build_structure{"LIBS_${libout}"}} = ("_DEFINES", "_INCLUDES", "_CFLAGS", "_SOURCES", - "_OBJECTS"); - @{$build_structure{"LIBS_${libout}_DEFINES"}} = @defines; - @{$build_structure{"LIBS_${libout}_INCLUDES"}} = @incpaths; - @{$build_structure{"LIBS_${libout}_CFLAGS"}} = @cflags; - @{$build_structure{"LIBS_${libout}_LFLAGS"}} = @lflags; - @{$build_structure{"LIBS_${libout}_SOURCES"}} = @sources; - @{$build_structure{"LIBS_${libout}_OBJECTS"}} = @objfiles; - clearCompileStep(); -} - -sub handleLinkLine -{ - my ($line, $lineno) = @_; - my (@objfiles, @lflags, @libs, $appout, $part); - my @parts = shellwords($line); - shift(@parts); # ignore cmd - while ($part = shift @parts) { - if ($part =~ /^-IGNORE/) { - push(@lflags, $part); - } elsif ($part =~ /^-[GRIMDO]/) { - # eat compiler flags - } elsif ("$part" eq "-o") { - $appout = shift @parts; - } elsif ("$part" eq "-lz") { - push(@libs, "zlib.lib"); - } elsif ("$part" eq "-lcrypto") { - push(@libs, "libcrypto.lib"); - } elsif ("$part" eq "-lssl") { - push(@libs, "libssl.lib"); - } elsif ("$part" eq "-lcurl") { - push(@libs, "libcurl.lib"); - } elsif ("$part" eq "-lexpat") { - push(@libs, "libexpat.lib"); - } elsif ("$part" eq "-liconv") { - push(@libs, "iconv.lib"); - } elsif ($part =~ /^[-\/]/) { - push(@lflags, $part); - } elsif ($part =~ /\.(a|lib)$/) { - $part =~ s/\.a$/.lib/; - push(@libs, $part); - } elsif ($part eq 'invalidcontinue.obj') { - # ignore - known to MSVC - } elsif ($part =~ /\.o$/) { - push(@objfiles, $part); - } elsif ($part =~ /\.obj$/) { - # do nothing, 'make' should not be producing .obj, only .o files - } else { - die "Unhandled link option @ line $lineno: $part"; - } - } -# print "AppOut: '$appout'\nLFlags: @lflags\nLibs : @libs\nOfiles: @objfiles\n"; -# exit(1); - foreach (@objfiles) { - my $sourcefile = $_; - $sourcefile =~ s/^headless-git\.o$/compat\/win32\/headless.c/; - $sourcefile =~ s/\.o$/.c/; - push(@sources, $sourcefile); - push(@cflags, @{$compile_options{"${sourcefile}_CFLAGS"}}); - push(@defines, @{$compile_options{"${sourcefile}_DEFINES"}}); - push(@incpaths, @{$compile_options{"${sourcefile}_INCPATHS"}}); - } - removeDuplicates(); - - removeDuplicates(); - push(@{$build_structure{"APPS"}}, $appout); - @{$build_structure{"APPS_${appout}"}} = ("_DEFINES", "_INCLUDES", "_CFLAGS", "_LFLAGS", - "_SOURCES", "_OBJECTS", "_LIBS"); - @{$build_structure{"APPS_${appout}_DEFINES"}} = @defines; - @{$build_structure{"APPS_${appout}_INCLUDES"}} = @incpaths; - @{$build_structure{"APPS_${appout}_CFLAGS"}} = @cflags; - @{$build_structure{"APPS_${appout}_LFLAGS"}} = @lflags; - @{$build_structure{"APPS_${appout}_SOURCES"}} = @sources; - @{$build_structure{"APPS_${appout}_OBJECTS"}} = @objfiles; - @{$build_structure{"APPS_${appout}_LIBS"}} = @libs; - clearCompileStep(); -} diff --git a/contrib/buildsystems/generate b/contrib/buildsystems/generate deleted file mode 100755 index bc10f25ff24a74..00000000000000 --- a/contrib/buildsystems/generate +++ /dev/null @@ -1,29 +0,0 @@ -#!/usr/bin/perl -w -###################################################################### -# Generate buildsystem files -# -# This script generate buildsystem files based on the output of a -# GNU Make --dry-run, enabling Windows users to develop Git with their -# trusted IDE with native projects. -# -# Note: -# It is not meant as *the* way of building Git with MSVC, but merely a -# convenience. The correct way of building Git with MSVC is to use the -# GNU Make tool to build with the maintained Makefile in the root of -# the project. If you have the msysgit environment installed and -# available in your current console, together with the Visual Studio -# environment you wish to build for, all you have to do is run the -# command: -# make MSVC=1 -# -# Copyright (C) 2009 Marius Storm-Olsen -###################################################################### -use strict; -use File::Basename; -use Cwd; - -my $git_dir = getcwd(); -$git_dir =~ s=\\=/=g; -$git_dir = dirname($git_dir) while (!-e "$git_dir/git.c" && "$git_dir" ne ""); -die "Couldn't find Git repo" if ("$git_dir" eq ""); -exec join(" ", ("PERL5LIB=${git_dir}/contrib/buildsystems ${git_dir}/contrib/buildsystems/engine.pl", @ARGV)); diff --git a/contrib/buildsystems/parse.pl b/contrib/buildsystems/parse.pl deleted file mode 100755 index c9656ece9931c4..00000000000000 --- a/contrib/buildsystems/parse.pl +++ /dev/null @@ -1,228 +0,0 @@ -#!/usr/bin/perl -w -###################################################################### -# Do not call this script directly! -# -# The generate script ensures that @INC is correct before the engine -# is executed. -# -# Copyright (C) 2009 Marius Storm-Olsen -###################################################################### -use strict; -use File::Basename; -use Cwd; - -my $file = $ARGV[0]; -die "No file provided!" if !defined $file; - -my ($cflags, $target, $type, $line); - -open(F, "<$file") || die "Couldn't open file $file"; -my @data = ; -close(F); - -while (my $text = shift @data) { - my $ate_next; - do { - $ate_next = 0; - $line++; - chomp $text; - chop $text if ($text =~ /\r$/); - if ($text =~ /\\$/) { - $text =~ s/\\$//; - $text .= shift @data; - $ate_next = 1; - } - } while($ate_next); - - if($text =~ / -c /) { - # compilation - handleCompileLine($text, $line); - - } elsif ($text =~ / -o /) { - # linking executable - handleLinkLine($text, $line); - - } elsif ($text =~ /\.o / && $text =~ /\.a /) { - # libifying - handleLibLine($text, $line); - -# } elsif ($text =~ /^cp /) { -# # copy file around -# -# } elsif ($text =~ /^rm -f /) { -# # shell command -# -# } elsif ($text =~ /^make[ \[]/) { -# # make output -# -# } elsif ($text =~ /^echo /) { -# # echo to file -# -# } elsif ($text =~ /^if /) { -# # shell conditional -# -# } elsif ($text =~ /^tclsh /) { -# # translation stuff -# -# } elsif ($text =~ /^umask /) { -# # handling boilerplates -# -# } elsif ($text =~ /\$\(\:\)/) { -# # ignore -# -# } elsif ($text =~ /^FLAGS=/) { -# # flags check for dependencies -# -# } elsif ($text =~ /^'\/usr\/bin\/perl' -MError -e/) { -# # perl commands for copying files -# -# } elsif ($text =~ /generate-cmdlist\.sh/) { -# # command for generating list of commands -# -# } elsif ($text =~ /^test / && $text =~ /|| rm -f /) { -# # commands removing executables, if they exist -# -# } elsif ($text =~ /new locations or Tcl/) { -# # command for detecting Tcl/Tk changes -# -# } elsif ($text =~ /mkdir -p/) { -# # command creating path -# -# } elsif ($text =~ /: no custom templates yet/) { -# # whatever - - } else { -# print "Unhandled (line: $line): $text\n"; - } -} -close(F); - -# use Data::Dumper; -# print "Parsed build structure:\n"; -# print Dumper(%build_structure); - -# ------------------------------------------------------------------- -# Functions under here -# ------------------------------------------------------------------- -my (%build_structure, @defines, @incpaths, @cflags, @sources); - -sub clearCompileStep -{ - @defines = (); - @incpaths = (); - @cflags = (); - @sources = (); -} - -sub removeDuplicates -{ - my (%dupHash, $entry); - %dupHash = map { $_, 1 } @defines; - @defines = keys %dupHash; - - %dupHash = map { $_, 1 } @incpaths; - @incpaths = keys %dupHash; - - %dupHash = map { $_, 1 } @cflags; - @cflags = keys %dupHash; - - %dupHash = map { $_, 1 } @sources; - @sources = keys %dupHash; -} - -sub handleCompileLine -{ - my ($line, $lineno) = @_; - my @parts = split(' ', $line); - shift(@parts); # ignore cmd - while (my $part = shift @parts) { - if ("$part" eq "-o") { - # ignore object file - shift @parts; - } elsif ("$part" eq "-c") { - # ignore compile flag - } elsif ("$part" eq "-c") { - } elsif ($part =~ /^.?-I/) { - push(@incpaths, $part); - } elsif ($part =~ /^.?-D/) { - push(@defines, $part); - } elsif ($part =~ /^-/) { - push(@cflags, $part); - } elsif ($part =~ /\.(c|cc|cpp)$/) { - push(@sources, $part); - } else { - die "Unhandled compiler option @ line $lineno: $part"; - } - } - #print "Sources: @sources\nCFlags: @cflags\nDefine: @defines\nIncpat: @incpaths\n"; - #exit(1); -} - -sub handleLibLine -{ - my ($line, $lineno) = @_; - my (@objfiles, @lflags, $libout, $part); - # kill cmd and rm 'prefix' - $line =~ s/^rm -f .* && .* rcs //; - my @parts = split(' ', $line); - while ($part = shift @parts) { - if ($part =~ /^-/) { - push(@lflags, $part); - } elsif ($part =~ /\.(o|obj)$/) { - push(@objfiles, $part); - } elsif ($part =~ /\.(a|lib)$/) { - $libout = $part; - } else { - die "Unhandled lib option @ line $lineno: $part"; - } - } - #print "LibOut: '$libout'\nLFlags: @lflags\nOfiles: @objfiles\n"; - #exit(1); - removeDuplicates(); - push(@{$build_structure{"LIBS"}}, $libout); - @{$build_structure{"LIBS_${libout}"}} = ("_DEFINES", "_INCLUDES", "_CFLAGS", "_SOURCES", - "_OBJECTS"); - @{$build_structure{"LIBS_${libout}_DEFINES"}} = @defines; - @{$build_structure{"LIBS_${libout}_INCLUDES"}} = @incpaths; - @{$build_structure{"LIBS_${libout}_CFLAGS"}} = @cflags; - @{$build_structure{"LIBS_${libout}_SOURCES"}} = @sources; - @{$build_structure{"LIBS_${libout}_OBJECTS"}} = @objfiles; - clearCompileStep(); -} - -sub handleLinkLine -{ - my ($line, $lineno) = @_; - my (@objfiles, @lflags, @libs, $appout, $part); - my @parts = split(' ', $line); - shift(@parts); # ignore cmd - while ($part = shift @parts) { - if ($part =~ /^-[GRIDO]/) { - # eat compiler flags - } elsif ("$part" eq "-o") { - $appout = shift @parts; - } elsif ($part =~ /^-/) { - push(@lflags, $part); - } elsif ($part =~ /\.(a|lib)$/) { - push(@libs, $part); - } elsif ($part =~ /\.(o|obj)$/) { - push(@objfiles, $part); - } else { - die "Unhandled lib option @ line $lineno: $part"; - } - } - #print "AppOut: '$appout'\nLFlags: @lflags\nLibs : @libs\nOfiles: @objfiles\n"; - #exit(1); - removeDuplicates(); - push(@{$build_structure{"APPS"}}, $appout); - @{$build_structure{"APPS_${appout}"}} = ("_DEFINES", "_INCLUDES", "_CFLAGS", "_LFLAGS", - "_SOURCES", "_OBJECTS", "_LIBS"); - @{$build_structure{"APPS_${appout}_DEFINES"}} = @defines; - @{$build_structure{"APPS_${appout}_INCLUDES"}} = @incpaths; - @{$build_structure{"APPS_${appout}_CFLAGS"}} = @cflags; - @{$build_structure{"APPS_${appout}_LFLAGS"}} = @lflags; - @{$build_structure{"APPS_${appout}_SOURCES"}} = @sources; - @{$build_structure{"APPS_${appout}_OBJECTS"}} = @objfiles; - @{$build_structure{"APPS_${appout}_LIBS"}} = @libs; - clearCompileStep(); -} From d30aad47169acf332361a6bd332ec419ac2c14dd Mon Sep 17 00:00:00 2001 From: Sverre Rabbelier Date: Sun, 24 Jul 2011 00:06:00 +0200 Subject: [PATCH 083/312] remote-helper: check helper status after import/export Signed-off-by: Johannes Schindelin Signed-off-by: Sverre Rabbelier --- t/t5801-remote-helpers.sh | 2 +- transport-helper.c | 15 +++++++++++++++ 2 files changed, 16 insertions(+), 1 deletion(-) diff --git a/t/t5801-remote-helpers.sh b/t/t5801-remote-helpers.sh index d21877150ed82e..3917da47276825 100755 --- a/t/t5801-remote-helpers.sh +++ b/t/t5801-remote-helpers.sh @@ -262,7 +262,7 @@ test_expect_success 'push update refs failure' ' echo "update fail" >>file && git commit -a -m "update fail" && git rev-parse --verify testgit/origin/heads/update >expect && - test_expect_code 1 env GIT_REMOTE_TESTGIT_FAILURE="non-fast forward" \ + test_must_fail env GIT_REMOTE_TESTGIT_FAILURE="non-fast forward" \ git push origin update && git rev-parse --verify testgit/origin/heads/update >actual && test_cmp expect actual diff --git a/transport-helper.c b/transport-helper.c index fed5d51a2b3d34..72588c4d14e3ff 100644 --- a/transport-helper.c +++ b/transport-helper.c @@ -505,6 +505,19 @@ static int get_exporter(struct transport *transport, return start_command(fastexport); } +static void check_helper_status(struct helper_data *data) +{ + int pid, status; + + pid = waitpid(data->helper->pid, &status, WNOHANG); + if (pid < 0) + die("Could not retrieve status of remote helper '%s'", + data->name); + if (pid > 0 && WIFEXITED(status)) + die("Remote helper '%s' died with %d", + data->name, WEXITSTATUS(status)); +} + static int fetch_with_import(struct transport *transport, int nr_heads, struct ref **to_fetch) { @@ -541,6 +554,7 @@ static int fetch_with_import(struct transport *transport, if (finish_command(&fastimport)) die(_("error while running fast-import")); + check_helper_status(data); /* * The fast-import stream of a remote helper that advertises @@ -1160,6 +1174,7 @@ static int push_refs_with_export(struct transport *transport, if (finish_command(&exporter)) die(_("error while running fast-export")); + check_helper_status(data); if (push_update_refs_status(data, remote_refs, flags)) return 1; From 14ed3bc671656679b47da4a4c7837e644b36c3ba Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Tue, 18 Apr 2017 12:09:08 +0200 Subject: [PATCH 084/312] mingw: demonstrate a problem with certain absolute paths On Windows, there are several categories of absolute paths. One such category starts with a backslash and is implicitly relative to the drive associated with the current working directory. Example: c: git clone https://github.com/git-for-windows/git \G4W should clone into C:\G4W. There is currently a problem with that, in that mingw_mktemp() does not expect the _wmktemp() function to prefix the absolute path with the drive prefix, and as a consequence, the resulting path does not fit into the originally-passed string buffer. The symptom is a "Result too large" error. Reported by Juan Carlos Arevalo Baeza. Signed-off-by: Johannes Schindelin --- t/t5580-unc-paths.sh | 19 ++++++++++++++----- 1 file changed, 14 insertions(+), 5 deletions(-) diff --git a/t/t5580-unc-paths.sh b/t/t5580-unc-paths.sh index 65ef1a3628ee94..2c30f3b51d3183 100755 --- a/t/t5580-unc-paths.sh +++ b/t/t5580-unc-paths.sh @@ -20,14 +20,11 @@ fi UNCPATH="$(winpwd)" case "$UNCPATH" in [A-Z]:*) + WITHOUTDRIVE="${UNCPATH#?:}" # Use administrative share e.g. \\localhost\C$\git-sdk-64\usr\src\git # (we use forward slashes here because MSYS2 and Git accept them, and # they are easier on the eyes) - UNCPATH="//localhost/${UNCPATH%%:*}\$/${UNCPATH#?:}" - test -d "$UNCPATH" || { - skip_all='could not access administrative share; skipping' - test_done - } + UNCPATH="//localhost/${UNCPATH%%:*}\$$WITHOUTDRIVE" ;; *) skip_all='skipping UNC path tests, cannot determine current path as UNC' @@ -35,6 +32,18 @@ case "$UNCPATH" in ;; esac +test_expect_failure 'clone into absolute path lacking a drive prefix' ' + USINGBACKSLASHES="$(echo "$WITHOUTDRIVE"/without-drive-prefix | + tr / \\\\)" && + git clone . "$USINGBACKSLASHES" && + test -f without-drive-prefix/.git/HEAD +' + +test -d "$UNCPATH" || { + skip_all='could not access administrative share; skipping' + test_done +} + test_expect_success setup ' test_commit initial ' From f0cb0061b62e13fe5bcde6bb3c56c90c8b116668 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Sat, 9 May 2020 14:08:36 +0200 Subject: [PATCH 085/312] config.mak.uname: drop the `vcxproj` target Now that we dropped `contrib/buildsystems/generate` to generate Visual Studio Solution files, it is time to also drop the `vcxproj` Makefile target that depended on that script. Signed-off-by: Johannes Schindelin --- config.mak.uname | 76 ------------------------------------------------ 1 file changed, 76 deletions(-) diff --git a/config.mak.uname b/config.mak.uname index b12d4e168ae119..c7a9fa1b775981 100644 --- a/config.mak.uname +++ b/config.mak.uname @@ -759,79 +759,3 @@ ifeq ($(uname_S),QNX) NO_STRCASESTR = YesPlease NO_STRLCPY = YesPlease endif - -vcxproj: - # Require clean work tree - git update-index -q --refresh && \ - git diff-files --quiet && \ - git diff-index --cached --quiet HEAD -- - - # Make .vcxproj files and add them - perl contrib/buildsystems/generate -g Vcxproj - git add -f git.sln {*,*/lib,t/helper/*}/*.vcxproj - - # Generate the LinkOrCopyBuiltins.targets and LinkOrCopyRemoteHttp.targets file - (echo '' && \ - echo ' ' && \ - for name in $(BUILT_INS);\ - do \ - echo ' '; \ - done && \ - echo ' ' && \ - echo '') >git/LinkOrCopyBuiltins.targets - (echo '' && \ - echo ' ' && \ - for name in $(REMOTE_CURL_ALIASES); \ - do \ - echo ' '; \ - done && \ - echo ' ' && \ - echo '') >git-remote-http/LinkOrCopyRemoteHttp.targets - git add -f git/LinkOrCopyBuiltins.targets git-remote-http/LinkOrCopyRemoteHttp.targets - - # Add generated headers - $(MAKE) MSVC=1 SKIP_VCPKG=1 prefix=/mingw64 $(GENERATED_H) - git add -f $(GENERATED_H) - - # Add scripts - rm -f perl/perl.mak - $(MAKE) MSVC=1 SKIP_VCPKG=1 prefix=/mingw64 $(SCRIPT_LIB) $(SCRIPTS) - # Strip out the sane tool path, needed only for building - sed -i '/^git_broken_path_fix ".*/d' git-sh-setup - git add -f $(SCRIPT_LIB) $(SCRIPTS) - - # Add Perl module - $(MAKE) $(LIB_PERL_GEN) - git add -f perl/build - - # Add bin-wrappers, for testing - rm -rf bin-wrappers/ - $(MAKE) MSVC=1 SKIP_VCPKG=1 prefix=/mingw64 $(test_bindir_programs) - # Ensure that the GIT_EXEC_PATH is a Unix-y one, and that the absolute - # path of the repository is not hard-coded (GIT_EXEC_PATH will be set - # by test-lib.sh according to the current setup) - sed -i -e 's/^\(GIT_EXEC_PATH\)=.*/test -n "$${\1##*:*}" ||\ - \1="$$(cygpath -u "$$\1")"/' \ - -e "s|'$$(pwd)|\"\$$GIT_EXEC_PATH\"'|g" bin-wrappers/* - # Ensure that test-* helpers find the .dll files copied to top-level - sed -i 's|^PATH=.*|&:"$$GIT_EXEC_PATH"|' bin-wrappers/test-* - # We do not want to force hard-linking builtins - sed -i 's|\(git\)-\([-a-z]*\)\.exe"|\1.exe" \2|g' \ - bin-wrappers/git-{receive-pack,upload-archive} - git add -f $(test_bindir_programs) - - # Add templates - $(MAKE) -C templates - git add -f templates/boilerplates.made templates/blt/ - - # Add the translated messages - make MSVC=1 SKIP_VCPKG=1 prefix=/mingw64 $(MOFILES) - git add -f $(MOFILES) - - # Add build options - $(MAKE) MSVC=1 SKIP_VCPKG=1 prefix=/mingw64 GIT-BUILD-OPTIONS - git add -f GIT-BUILD-OPTIONS - - # Commit the whole shebang - git commit -m "Generate Visual Studio solution" \ - -m "Auto-generated by \`$(MAKE)$(MAKEFLAGS) $@\`" From 6faecb6f2d4b65df94c175e486d49d4b6cdcffc8 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Mon, 9 Apr 2012 13:04:35 -0500 Subject: [PATCH 086/312] Always auto-gc after calling a fast-import transport After importing anything with fast-import, we should always let the garbage collector do its job, since the objects are written to disk inefficiently. This brings down an initial import of http://selenic.com/hg from about 230 megabytes to about 14. In the future, we may want to make this configurable on a per-remote basis, or maybe teach fast-import about it in the first place. Signed-off-by: Johannes Schindelin --- transport-helper.c | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/transport-helper.c b/transport-helper.c index 72588c4d14e3ff..7513fd7eea05e4 100644 --- a/transport-helper.c +++ b/transport-helper.c @@ -22,6 +22,8 @@ #include "packfile.h" static int debug; +/* TODO: put somewhere sensible, e.g. git_transport_options? */ +static int auto_gc = 1; struct helper_data { char *name; @@ -588,6 +590,13 @@ static int fetch_with_import(struct transport *transport, } } strbuf_release(&buf); + if (auto_gc) { + struct child_process cmd = CHILD_PROCESS_INIT; + + cmd.git_cmd = 1; + strvec_pushl(&cmd.args, "gc", "--auto", "--quiet", NULL); + run_command(&cmd); + } return 0; } From 5d1c542153494994c381c49f37743e5534c79f41 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Tue, 18 Apr 2017 12:38:30 +0200 Subject: [PATCH 087/312] mingw: allow absolute paths without drive prefix When specifying an absolute path without a drive prefix, we convert that path internally. Let's make sure that we handle that case properly, too ;-) This fixes the command git clone https://github.com/git-for-windows/git \G4W Signed-off-by: Johannes Schindelin --- compat/mingw.c | 10 +++++++++- t/t5580-unc-paths.sh | 2 +- 2 files changed, 10 insertions(+), 2 deletions(-) diff --git a/compat/mingw.c b/compat/mingw.c index f524c54d06d965..12adc941381aa6 100644 --- a/compat/mingw.c +++ b/compat/mingw.c @@ -1141,11 +1141,19 @@ unsigned int sleep (unsigned int seconds) char *mingw_mktemp(char *template) { wchar_t wtemplate[MAX_PATH]; + int offset = 0; + if (xutftowcs_path(wtemplate, template) < 0) return NULL; + + if (is_dir_sep(template[0]) && !is_dir_sep(template[1]) && + iswalpha(wtemplate[0]) && wtemplate[1] == L':') { + /* We have an absolute path missing the drive prefix */ + offset = 2; + } if (!_wmktemp(wtemplate)) return NULL; - if (xwcstoutf(template, wtemplate, strlen(template) + 1) < 0) + if (xwcstoutf(template, wtemplate + offset, strlen(template) + 1) < 0) return NULL; return template; } diff --git a/t/t5580-unc-paths.sh b/t/t5580-unc-paths.sh index 2c30f3b51d3183..e9df367d5777fd 100755 --- a/t/t5580-unc-paths.sh +++ b/t/t5580-unc-paths.sh @@ -32,7 +32,7 @@ case "$UNCPATH" in ;; esac -test_expect_failure 'clone into absolute path lacking a drive prefix' ' +test_expect_success 'clone into absolute path lacking a drive prefix' ' USINGBACKSLASHES="$(echo "$WITHOUTDRIVE"/without-drive-prefix | tr / \\\\)" && git clone . "$USINGBACKSLASHES" && From 99b7daa6c84952322ac23c6173a4e529cfdf987a Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Sun, 20 Oct 2019 22:08:58 +0200 Subject: [PATCH 088/312] win32/pthread: avoid name clashes with winpthread The mingw-w64 GCC seems to link implicitly to libwinpthread, which does implement a pthread emulation (that is more complete than Git's). Let's keep preferring Git's. To avoid linker errors where it thinks that the `pthread_self` and the `pthread_create` symbols are defined twice, let's give our version a `win32_` prefix, just like we already do for `pthread_join()`. Signed-off-by: Johannes Schindelin --- compat/win32/pthread.c | 6 +++--- compat/win32/pthread.h | 8 +++++--- 2 files changed, 8 insertions(+), 6 deletions(-) diff --git a/compat/win32/pthread.c b/compat/win32/pthread.c index 58980a529c3eb9..4cc6356f757d80 100644 --- a/compat/win32/pthread.c +++ b/compat/win32/pthread.c @@ -21,8 +21,8 @@ static unsigned __stdcall win32_start_routine(void *arg) return 0; } -int pthread_create(pthread_t *thread, const void *attr UNUSED, - void *(*start_routine)(void *), void *arg) +int win32_pthread_create(pthread_t *thread, const void *attr UNUSED, + void *(*start_routine)(void *), void *arg) { thread->arg = arg; thread->start_routine = start_routine; @@ -53,7 +53,7 @@ int win32_pthread_join(pthread_t *thread, void **value_ptr) } } -pthread_t pthread_self(void) +pthread_t win32_pthread_self(void) { pthread_t t = { NULL }; t.tid = GetCurrentThreadId(); diff --git a/compat/win32/pthread.h b/compat/win32/pthread.h index e2b5c4f64c9b91..476e78b911b857 100644 --- a/compat/win32/pthread.h +++ b/compat/win32/pthread.h @@ -50,8 +50,9 @@ typedef struct { DWORD tid; } pthread_t; -int pthread_create(pthread_t *thread, const void *unused, - void *(*start_routine)(void*), void *arg); +int win32_pthread_create(pthread_t *thread, const void *unused, + void *(*start_routine)(void*), void *arg); +#define pthread_create win32_pthread_create /* * To avoid the need of copying a struct, we use small macro wrapper to pass @@ -62,7 +63,8 @@ int pthread_create(pthread_t *thread, const void *unused, int win32_pthread_join(pthread_t *thread, void **value_ptr); #define pthread_equal(t1, t2) ((t1).tid == (t2).tid) -pthread_t pthread_self(void); +pthread_t win32_pthread_self(void); +#define pthread_self win32_pthread_self static inline void NORETURN pthread_exit(void *ret) { From 281b6fb8af5a6f1b8a8e77475a3be818216071dc Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Fri, 12 Aug 2022 12:44:15 +0200 Subject: [PATCH 089/312] git-compat-util: avoid redeclaring _DEFAULT_SOURCE We are about to vendor in `mimalloc`'s source code which we will want to include `git-compat-util.h` after defining that constant. Signed-off-by: Johannes Schindelin --- git-compat-util.h | 2 ++ 1 file changed, 2 insertions(+) diff --git a/git-compat-util.h b/git-compat-util.h index e123288e8f1393..6aee7e1b27a26b 100644 --- a/git-compat-util.h +++ b/git-compat-util.h @@ -201,7 +201,9 @@ DISABLE_WARNING(-Wsign-compare) #define _ALL_SOURCE 1 #define _GNU_SOURCE 1 #define _BSD_SOURCE 1 +#ifndef _DEFAULT_SOURCE #define _DEFAULT_SOURCE 1 +#endif #define _NETBSD_SOURCE 1 #define _SGI_SOURCE 1 From 5971c3d7a788807d756abc2652494d2967c1f751 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Mon, 24 Jun 2019 21:31:30 +0200 Subject: [PATCH 090/312] Import the source code of mimalloc v2.1.2 This commit imports mimalloc's source code as per v2.1.2, fetched from the tag at https://github.com/microsoft/mimalloc. The .c files are from the src/ subdirectory, and the .h files from the include/ and include/mimalloc/ subdirectories. We will subsequently modify the source code to accommodate building within Git's context. Since we plan on using the `mi_*()` family of functions, we skip the C++-specific source code, some POSIX compliant functions to interact with mimalloc, and the code that wants to support auto-magic overriding of the `malloc()` function (mimalloc-new-delete.h, alloc-posix.c, mimalloc-override.h, alloc-override.c, alloc-override-osx.c, alloc-override-win.c and static.c). To appease the `check-whitespace` job of Git's Continuous Integration, this commit was washed one time via `git rebase --whitespace=fix`. Signed-off-by: Johannes Schindelin --- Makefile | 1 + compat/mimalloc/LICENSE | 21 + compat/mimalloc/alloc-aligned.c | 298 +++++ compat/mimalloc/alloc.c | 1060 ++++++++++++++++++ compat/mimalloc/arena.c | 935 ++++++++++++++++ compat/mimalloc/bitmap.c | 432 +++++++ compat/mimalloc/bitmap.h | 115 ++ compat/mimalloc/heap.c | 626 +++++++++++ compat/mimalloc/init.c | 709 ++++++++++++ compat/mimalloc/mimalloc.h | 565 ++++++++++ compat/mimalloc/mimalloc/atomic.h | 385 +++++++ compat/mimalloc/mimalloc/internal.h | 979 ++++++++++++++++ compat/mimalloc/mimalloc/prim.h | 323 ++++++ compat/mimalloc/mimalloc/track.h | 147 +++ compat/mimalloc/mimalloc/types.h | 670 +++++++++++ compat/mimalloc/options.c | 571 ++++++++++ compat/mimalloc/os.c | 689 ++++++++++++ compat/mimalloc/page-queue.c | 332 ++++++ compat/mimalloc/page.c | 939 ++++++++++++++++ compat/mimalloc/prim/windows/prim.c | 622 +++++++++++ compat/mimalloc/random.c | 254 +++++ compat/mimalloc/segment-cache.c | 0 compat/mimalloc/segment-map.c | 153 +++ compat/mimalloc/segment.c | 1617 +++++++++++++++++++++++++++ compat/mimalloc/stats.c | 467 ++++++++ 25 files changed, 12910 insertions(+) create mode 100644 compat/mimalloc/LICENSE create mode 100644 compat/mimalloc/alloc-aligned.c create mode 100644 compat/mimalloc/alloc.c create mode 100644 compat/mimalloc/arena.c create mode 100644 compat/mimalloc/bitmap.c create mode 100644 compat/mimalloc/bitmap.h create mode 100644 compat/mimalloc/heap.c create mode 100644 compat/mimalloc/init.c create mode 100644 compat/mimalloc/mimalloc.h create mode 100644 compat/mimalloc/mimalloc/atomic.h create mode 100644 compat/mimalloc/mimalloc/internal.h create mode 100644 compat/mimalloc/mimalloc/prim.h create mode 100644 compat/mimalloc/mimalloc/track.h create mode 100644 compat/mimalloc/mimalloc/types.h create mode 100644 compat/mimalloc/options.c create mode 100644 compat/mimalloc/os.c create mode 100644 compat/mimalloc/page-queue.c create mode 100644 compat/mimalloc/page.c create mode 100644 compat/mimalloc/prim/windows/prim.c create mode 100644 compat/mimalloc/random.c create mode 100644 compat/mimalloc/segment-cache.c create mode 100644 compat/mimalloc/segment-map.c create mode 100644 compat/mimalloc/segment.c create mode 100644 compat/mimalloc/stats.c diff --git a/Makefile b/Makefile index 7315507381e146..fbb9227e7d7909 100644 --- a/Makefile +++ b/Makefile @@ -1342,6 +1342,7 @@ BUILTIN_OBJS += builtin/write-tree.o # upstream unnecessarily (making merging in future changes easier). THIRD_PARTY_SOURCES += compat/inet_ntop.c THIRD_PARTY_SOURCES += compat/inet_pton.c +THIRD_PARTY_SOURCES += compat/mimalloc/% THIRD_PARTY_SOURCES += compat/nedmalloc/% THIRD_PARTY_SOURCES += compat/obstack.% THIRD_PARTY_SOURCES += compat/poll/% diff --git a/compat/mimalloc/LICENSE b/compat/mimalloc/LICENSE new file mode 100644 index 00000000000000..670b668a0c928e --- /dev/null +++ b/compat/mimalloc/LICENSE @@ -0,0 +1,21 @@ +MIT License + +Copyright (c) 2018-2021 Microsoft Corporation, Daan Leijen + +Permission is hereby granted, free of charge, to any person obtaining a copy +of this software and associated documentation files (the "Software"), to deal +in the Software without restriction, including without limitation the rights +to use, copy, modify, merge, publish, distribute, sublicense, and/or sell +copies of the Software, and to permit persons to whom the Software is +furnished to do so, subject to the following conditions: + +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. diff --git a/compat/mimalloc/alloc-aligned.c b/compat/mimalloc/alloc-aligned.c new file mode 100644 index 00000000000000..e975af5f7c2ad4 --- /dev/null +++ b/compat/mimalloc/alloc-aligned.c @@ -0,0 +1,298 @@ +/* ---------------------------------------------------------------------------- +Copyright (c) 2018-2021, Microsoft Research, Daan Leijen +This is free software; you can redistribute it and/or modify it under the +terms of the MIT license. A copy of the license can be found in the file +"LICENSE" at the root of this distribution. +-----------------------------------------------------------------------------*/ + +#include "mimalloc.h" +#include "mimalloc/internal.h" +#include "mimalloc/prim.h" // mi_prim_get_default_heap + +#include // memset + +// ------------------------------------------------------ +// Aligned Allocation +// ------------------------------------------------------ + +// Fallback primitive aligned allocation -- split out for better codegen +static mi_decl_noinline void* mi_heap_malloc_zero_aligned_at_fallback(mi_heap_t* const heap, const size_t size, const size_t alignment, const size_t offset, const bool zero) mi_attr_noexcept +{ + mi_assert_internal(size <= PTRDIFF_MAX); + mi_assert_internal(alignment != 0 && _mi_is_power_of_two(alignment)); + + const uintptr_t align_mask = alignment - 1; // for any x, `(x & align_mask) == (x % alignment)` + const size_t padsize = size + MI_PADDING_SIZE; + + // use regular allocation if it is guaranteed to fit the alignment constraints + if (offset==0 && alignment<=padsize && padsize<=MI_MAX_ALIGN_GUARANTEE && (padsize&align_mask)==0) { + void* p = _mi_heap_malloc_zero(heap, size, zero); + mi_assert_internal(p == NULL || ((uintptr_t)p % alignment) == 0); + return p; + } + + void* p; + size_t oversize; + if mi_unlikely(alignment > MI_ALIGNMENT_MAX) { + // use OS allocation for very large alignment and allocate inside a huge page (dedicated segment with 1 page) + // This can support alignments >= MI_SEGMENT_SIZE by ensuring the object can be aligned at a point in the + // first (and single) page such that the segment info is `MI_SEGMENT_SIZE` bytes before it (so it can be found by aligning the pointer down) + if mi_unlikely(offset != 0) { + // todo: cannot support offset alignment for very large alignments yet + #if MI_DEBUG > 0 + _mi_error_message(EOVERFLOW, "aligned allocation with a very large alignment cannot be used with an alignment offset (size %zu, alignment %zu, offset %zu)\n", size, alignment, offset); + #endif + return NULL; + } + oversize = (size <= MI_SMALL_SIZE_MAX ? MI_SMALL_SIZE_MAX + 1 /* ensure we use generic malloc path */ : size); + p = _mi_heap_malloc_zero_ex(heap, oversize, false, alignment); // the page block size should be large enough to align in the single huge page block + // zero afterwards as only the area from the aligned_p may be committed! + if (p == NULL) return NULL; + } + else { + // otherwise over-allocate + oversize = size + alignment - 1; + p = _mi_heap_malloc_zero(heap, oversize, zero); + if (p == NULL) return NULL; + } + + // .. and align within the allocation + const uintptr_t poffset = ((uintptr_t)p + offset) & align_mask; + const uintptr_t adjust = (poffset == 0 ? 0 : alignment - poffset); + mi_assert_internal(adjust < alignment); + void* aligned_p = (void*)((uintptr_t)p + adjust); + if (aligned_p != p) { + mi_page_t* page = _mi_ptr_page(p); + mi_page_set_has_aligned(page, true); + _mi_padding_shrink(page, (mi_block_t*)p, adjust + size); + } + // todo: expand padding if overallocated ? + + mi_assert_internal(mi_page_usable_block_size(_mi_ptr_page(p)) >= adjust + size); + mi_assert_internal(p == _mi_page_ptr_unalign(_mi_ptr_segment(aligned_p), _mi_ptr_page(aligned_p), aligned_p)); + mi_assert_internal(((uintptr_t)aligned_p + offset) % alignment == 0); + mi_assert_internal(mi_usable_size(aligned_p)>=size); + mi_assert_internal(mi_usable_size(p) == mi_usable_size(aligned_p)+adjust); + + // now zero the block if needed + if (alignment > MI_ALIGNMENT_MAX) { + // for the tracker, on huge aligned allocations only from the start of the large block is defined + mi_track_mem_undefined(aligned_p, size); + if (zero) { + _mi_memzero_aligned(aligned_p, mi_usable_size(aligned_p)); + } + } + + if (p != aligned_p) { + mi_track_align(p,aligned_p,adjust,mi_usable_size(aligned_p)); + } + return aligned_p; +} + +// Primitive aligned allocation +static void* mi_heap_malloc_zero_aligned_at(mi_heap_t* const heap, const size_t size, const size_t alignment, const size_t offset, const bool zero) mi_attr_noexcept +{ + // note: we don't require `size > offset`, we just guarantee that the address at offset is aligned regardless of the allocated size. + if mi_unlikely(alignment == 0 || !_mi_is_power_of_two(alignment)) { // require power-of-two (see ) + #if MI_DEBUG > 0 + _mi_error_message(EOVERFLOW, "aligned allocation requires the alignment to be a power-of-two (size %zu, alignment %zu)\n", size, alignment); + #endif + return NULL; + } + + if mi_unlikely(size > PTRDIFF_MAX) { // we don't allocate more than PTRDIFF_MAX (see ) + #if MI_DEBUG > 0 + _mi_error_message(EOVERFLOW, "aligned allocation request is too large (size %zu, alignment %zu)\n", size, alignment); + #endif + return NULL; + } + const uintptr_t align_mask = alignment-1; // for any x, `(x & align_mask) == (x % alignment)` + const size_t padsize = size + MI_PADDING_SIZE; // note: cannot overflow due to earlier size > PTRDIFF_MAX check + + // try first if there happens to be a small block available with just the right alignment + if mi_likely(padsize <= MI_SMALL_SIZE_MAX && alignment <= padsize) { + mi_page_t* page = _mi_heap_get_free_small_page(heap, padsize); + const bool is_aligned = (((uintptr_t)page->free+offset) & align_mask)==0; + if mi_likely(page->free != NULL && is_aligned) + { + #if MI_STAT>1 + mi_heap_stat_increase(heap, malloc, size); + #endif + void* p = _mi_page_malloc(heap, page, padsize, zero); // TODO: inline _mi_page_malloc + mi_assert_internal(p != NULL); + mi_assert_internal(((uintptr_t)p + offset) % alignment == 0); + mi_track_malloc(p,size,zero); + return p; + } + } + // fallback + return mi_heap_malloc_zero_aligned_at_fallback(heap, size, alignment, offset, zero); +} + + +// ------------------------------------------------------ +// Optimized mi_heap_malloc_aligned / mi_malloc_aligned +// ------------------------------------------------------ + +mi_decl_nodiscard mi_decl_restrict void* mi_heap_malloc_aligned_at(mi_heap_t* heap, size_t size, size_t alignment, size_t offset) mi_attr_noexcept { + return mi_heap_malloc_zero_aligned_at(heap, size, alignment, offset, false); +} + +mi_decl_nodiscard mi_decl_restrict void* mi_heap_malloc_aligned(mi_heap_t* heap, size_t size, size_t alignment) mi_attr_noexcept { + if mi_unlikely(alignment == 0 || !_mi_is_power_of_two(alignment)) return NULL; + #if !MI_PADDING + // without padding, any small sized allocation is naturally aligned (see also `_mi_segment_page_start`) + if mi_likely(_mi_is_power_of_two(size) && size >= alignment && size <= MI_SMALL_SIZE_MAX) + #else + // with padding, we can only guarantee this for fixed alignments + if mi_likely((alignment == sizeof(void*) || (alignment == MI_MAX_ALIGN_SIZE && size > (MI_MAX_ALIGN_SIZE/2))) + && size <= MI_SMALL_SIZE_MAX) + #endif + { + // fast path for common alignment and size + return mi_heap_malloc_small(heap, size); + } + else { + return mi_heap_malloc_aligned_at(heap, size, alignment, 0); + } +} + +// ensure a definition is emitted +#if defined(__cplusplus) +static void* _mi_heap_malloc_aligned = (void*)&mi_heap_malloc_aligned; +#endif + +// ------------------------------------------------------ +// Aligned Allocation +// ------------------------------------------------------ + +mi_decl_nodiscard mi_decl_restrict void* mi_heap_zalloc_aligned_at(mi_heap_t* heap, size_t size, size_t alignment, size_t offset) mi_attr_noexcept { + return mi_heap_malloc_zero_aligned_at(heap, size, alignment, offset, true); +} + +mi_decl_nodiscard mi_decl_restrict void* mi_heap_zalloc_aligned(mi_heap_t* heap, size_t size, size_t alignment) mi_attr_noexcept { + return mi_heap_zalloc_aligned_at(heap, size, alignment, 0); +} + +mi_decl_nodiscard mi_decl_restrict void* mi_heap_calloc_aligned_at(mi_heap_t* heap, size_t count, size_t size, size_t alignment, size_t offset) mi_attr_noexcept { + size_t total; + if (mi_count_size_overflow(count, size, &total)) return NULL; + return mi_heap_zalloc_aligned_at(heap, total, alignment, offset); +} + +mi_decl_nodiscard mi_decl_restrict void* mi_heap_calloc_aligned(mi_heap_t* heap, size_t count, size_t size, size_t alignment) mi_attr_noexcept { + return mi_heap_calloc_aligned_at(heap,count,size,alignment,0); +} + +mi_decl_nodiscard mi_decl_restrict void* mi_malloc_aligned_at(size_t size, size_t alignment, size_t offset) mi_attr_noexcept { + return mi_heap_malloc_aligned_at(mi_prim_get_default_heap(), size, alignment, offset); +} + +mi_decl_nodiscard mi_decl_restrict void* mi_malloc_aligned(size_t size, size_t alignment) mi_attr_noexcept { + return mi_heap_malloc_aligned(mi_prim_get_default_heap(), size, alignment); +} + +mi_decl_nodiscard mi_decl_restrict void* mi_zalloc_aligned_at(size_t size, size_t alignment, size_t offset) mi_attr_noexcept { + return mi_heap_zalloc_aligned_at(mi_prim_get_default_heap(), size, alignment, offset); +} + +mi_decl_nodiscard mi_decl_restrict void* mi_zalloc_aligned(size_t size, size_t alignment) mi_attr_noexcept { + return mi_heap_zalloc_aligned(mi_prim_get_default_heap(), size, alignment); +} + +mi_decl_nodiscard mi_decl_restrict void* mi_calloc_aligned_at(size_t count, size_t size, size_t alignment, size_t offset) mi_attr_noexcept { + return mi_heap_calloc_aligned_at(mi_prim_get_default_heap(), count, size, alignment, offset); +} + +mi_decl_nodiscard mi_decl_restrict void* mi_calloc_aligned(size_t count, size_t size, size_t alignment) mi_attr_noexcept { + return mi_heap_calloc_aligned(mi_prim_get_default_heap(), count, size, alignment); +} + + +// ------------------------------------------------------ +// Aligned re-allocation +// ------------------------------------------------------ + +static void* mi_heap_realloc_zero_aligned_at(mi_heap_t* heap, void* p, size_t newsize, size_t alignment, size_t offset, bool zero) mi_attr_noexcept { + mi_assert(alignment > 0); + if (alignment <= sizeof(uintptr_t)) return _mi_heap_realloc_zero(heap,p,newsize,zero); + if (p == NULL) return mi_heap_malloc_zero_aligned_at(heap,newsize,alignment,offset,zero); + size_t size = mi_usable_size(p); + if (newsize <= size && newsize >= (size - (size / 2)) + && (((uintptr_t)p + offset) % alignment) == 0) { + return p; // reallocation still fits, is aligned and not more than 50% waste + } + else { + // note: we don't zero allocate upfront so we only zero initialize the expanded part + void* newp = mi_heap_malloc_aligned_at(heap,newsize,alignment,offset); + if (newp != NULL) { + if (zero && newsize > size) { + // also set last word in the previous allocation to zero to ensure any padding is zero-initialized + size_t start = (size >= sizeof(intptr_t) ? size - sizeof(intptr_t) : 0); + _mi_memzero((uint8_t*)newp + start, newsize - start); + } + _mi_memcpy_aligned(newp, p, (newsize > size ? size : newsize)); + mi_free(p); // only free if successful + } + return newp; + } +} + +static void* mi_heap_realloc_zero_aligned(mi_heap_t* heap, void* p, size_t newsize, size_t alignment, bool zero) mi_attr_noexcept { + mi_assert(alignment > 0); + if (alignment <= sizeof(uintptr_t)) return _mi_heap_realloc_zero(heap,p,newsize,zero); + size_t offset = ((uintptr_t)p % alignment); // use offset of previous allocation (p can be NULL) + return mi_heap_realloc_zero_aligned_at(heap,p,newsize,alignment,offset,zero); +} + +mi_decl_nodiscard void* mi_heap_realloc_aligned_at(mi_heap_t* heap, void* p, size_t newsize, size_t alignment, size_t offset) mi_attr_noexcept { + return mi_heap_realloc_zero_aligned_at(heap,p,newsize,alignment,offset,false); +} + +mi_decl_nodiscard void* mi_heap_realloc_aligned(mi_heap_t* heap, void* p, size_t newsize, size_t alignment) mi_attr_noexcept { + return mi_heap_realloc_zero_aligned(heap,p,newsize,alignment,false); +} + +mi_decl_nodiscard void* mi_heap_rezalloc_aligned_at(mi_heap_t* heap, void* p, size_t newsize, size_t alignment, size_t offset) mi_attr_noexcept { + return mi_heap_realloc_zero_aligned_at(heap, p, newsize, alignment, offset, true); +} + +mi_decl_nodiscard void* mi_heap_rezalloc_aligned(mi_heap_t* heap, void* p, size_t newsize, size_t alignment) mi_attr_noexcept { + return mi_heap_realloc_zero_aligned(heap, p, newsize, alignment, true); +} + +mi_decl_nodiscard void* mi_heap_recalloc_aligned_at(mi_heap_t* heap, void* p, size_t newcount, size_t size, size_t alignment, size_t offset) mi_attr_noexcept { + size_t total; + if (mi_count_size_overflow(newcount, size, &total)) return NULL; + return mi_heap_rezalloc_aligned_at(heap, p, total, alignment, offset); +} + +mi_decl_nodiscard void* mi_heap_recalloc_aligned(mi_heap_t* heap, void* p, size_t newcount, size_t size, size_t alignment) mi_attr_noexcept { + size_t total; + if (mi_count_size_overflow(newcount, size, &total)) return NULL; + return mi_heap_rezalloc_aligned(heap, p, total, alignment); +} + +mi_decl_nodiscard void* mi_realloc_aligned_at(void* p, size_t newsize, size_t alignment, size_t offset) mi_attr_noexcept { + return mi_heap_realloc_aligned_at(mi_prim_get_default_heap(), p, newsize, alignment, offset); +} + +mi_decl_nodiscard void* mi_realloc_aligned(void* p, size_t newsize, size_t alignment) mi_attr_noexcept { + return mi_heap_realloc_aligned(mi_prim_get_default_heap(), p, newsize, alignment); +} + +mi_decl_nodiscard void* mi_rezalloc_aligned_at(void* p, size_t newsize, size_t alignment, size_t offset) mi_attr_noexcept { + return mi_heap_rezalloc_aligned_at(mi_prim_get_default_heap(), p, newsize, alignment, offset); +} + +mi_decl_nodiscard void* mi_rezalloc_aligned(void* p, size_t newsize, size_t alignment) mi_attr_noexcept { + return mi_heap_rezalloc_aligned(mi_prim_get_default_heap(), p, newsize, alignment); +} + +mi_decl_nodiscard void* mi_recalloc_aligned_at(void* p, size_t newcount, size_t size, size_t alignment, size_t offset) mi_attr_noexcept { + return mi_heap_recalloc_aligned_at(mi_prim_get_default_heap(), p, newcount, size, alignment, offset); +} + +mi_decl_nodiscard void* mi_recalloc_aligned(void* p, size_t newcount, size_t size, size_t alignment) mi_attr_noexcept { + return mi_heap_recalloc_aligned(mi_prim_get_default_heap(), p, newcount, size, alignment); +} diff --git a/compat/mimalloc/alloc.c b/compat/mimalloc/alloc.c new file mode 100644 index 00000000000000..961f6d53d0f2c7 --- /dev/null +++ b/compat/mimalloc/alloc.c @@ -0,0 +1,1060 @@ +/* ---------------------------------------------------------------------------- +Copyright (c) 2018-2022, Microsoft Research, Daan Leijen +This is free software; you can redistribute it and/or modify it under the +terms of the MIT license. A copy of the license can be found in the file +"LICENSE" at the root of this distribution. +-----------------------------------------------------------------------------*/ +#ifndef _DEFAULT_SOURCE +#define _DEFAULT_SOURCE // for realpath() on Linux +#endif + +#include "mimalloc.h" +#include "mimalloc/internal.h" +#include "mimalloc/atomic.h" +#include "mimalloc/prim.h" // _mi_prim_thread_id() + +#include // memset, strlen (for mi_strdup) +#include // malloc, abort + +#define MI_IN_ALLOC_C +#include "alloc-override.c" +#undef MI_IN_ALLOC_C + +// ------------------------------------------------------ +// Allocation +// ------------------------------------------------------ + +// Fast allocation in a page: just pop from the free list. +// Fall back to generic allocation only if the list is empty. +extern inline void* _mi_page_malloc(mi_heap_t* heap, mi_page_t* page, size_t size, bool zero) mi_attr_noexcept { + mi_assert_internal(page->xblock_size==0||mi_page_block_size(page) >= size); + mi_block_t* const block = page->free; + if mi_unlikely(block == NULL) { + return _mi_malloc_generic(heap, size, zero, 0); + } + mi_assert_internal(block != NULL && _mi_ptr_page(block) == page); + // pop from the free list + page->used++; + page->free = mi_block_next(page, block); + mi_assert_internal(page->free == NULL || _mi_ptr_page(page->free) == page); + #if MI_DEBUG>3 + if (page->free_is_zero) { + mi_assert_expensive(mi_mem_is_zero(block+1,size - sizeof(*block))); + } + #endif + + // allow use of the block internally + // note: when tracking we need to avoid ever touching the MI_PADDING since + // that is tracked by valgrind etc. as non-accessible (through the red-zone, see `mimalloc/track.h`) + mi_track_mem_undefined(block, mi_page_usable_block_size(page)); + + // zero the block? note: we need to zero the full block size (issue #63) + if mi_unlikely(zero) { + mi_assert_internal(page->xblock_size != 0); // do not call with zero'ing for huge blocks (see _mi_malloc_generic) + mi_assert_internal(page->xblock_size >= MI_PADDING_SIZE); + if (page->free_is_zero) { + block->next = 0; + mi_track_mem_defined(block, page->xblock_size - MI_PADDING_SIZE); + } + else { + _mi_memzero_aligned(block, page->xblock_size - MI_PADDING_SIZE); + } + } + +#if (MI_DEBUG>0) && !MI_TRACK_ENABLED && !MI_TSAN + if (!zero && !mi_page_is_huge(page)) { + memset(block, MI_DEBUG_UNINIT, mi_page_usable_block_size(page)); + } +#elif (MI_SECURE!=0) + if (!zero) { block->next = 0; } // don't leak internal data +#endif + +#if (MI_STAT>0) + const size_t bsize = mi_page_usable_block_size(page); + if (bsize <= MI_MEDIUM_OBJ_SIZE_MAX) { + mi_heap_stat_increase(heap, normal, bsize); + mi_heap_stat_counter_increase(heap, normal_count, 1); +#if (MI_STAT>1) + const size_t bin = _mi_bin(bsize); + mi_heap_stat_increase(heap, normal_bins[bin], 1); +#endif + } +#endif + +#if MI_PADDING // && !MI_TRACK_ENABLED + mi_padding_t* const padding = (mi_padding_t*)((uint8_t*)block + mi_page_usable_block_size(page)); + ptrdiff_t delta = ((uint8_t*)padding - (uint8_t*)block - (size - MI_PADDING_SIZE)); + #if (MI_DEBUG>=2) + mi_assert_internal(delta >= 0 && mi_page_usable_block_size(page) >= (size - MI_PADDING_SIZE + delta)); + #endif + mi_track_mem_defined(padding,sizeof(mi_padding_t)); // note: re-enable since mi_page_usable_block_size may set noaccess + padding->canary = (uint32_t)(mi_ptr_encode(page,block,page->keys)); + padding->delta = (uint32_t)(delta); + #if MI_PADDING_CHECK + if (!mi_page_is_huge(page)) { + uint8_t* fill = (uint8_t*)padding - delta; + const size_t maxpad = (delta > MI_MAX_ALIGN_SIZE ? MI_MAX_ALIGN_SIZE : delta); // set at most N initial padding bytes + for (size_t i = 0; i < maxpad; i++) { fill[i] = MI_DEBUG_PADDING; } + } + #endif +#endif + + return block; +} + +static inline mi_decl_restrict void* mi_heap_malloc_small_zero(mi_heap_t* heap, size_t size, bool zero) mi_attr_noexcept { + mi_assert(heap != NULL); + #if MI_DEBUG + const uintptr_t tid = _mi_thread_id(); + mi_assert(heap->thread_id == 0 || heap->thread_id == tid); // heaps are thread local + #endif + mi_assert(size <= MI_SMALL_SIZE_MAX); + #if (MI_PADDING) + if (size == 0) { size = sizeof(void*); } + #endif + mi_page_t* page = _mi_heap_get_free_small_page(heap, size + MI_PADDING_SIZE); + void* const p = _mi_page_malloc(heap, page, size + MI_PADDING_SIZE, zero); + mi_track_malloc(p,size,zero); + #if MI_STAT>1 + if (p != NULL) { + if (!mi_heap_is_initialized(heap)) { heap = mi_prim_get_default_heap(); } + mi_heap_stat_increase(heap, malloc, mi_usable_size(p)); + } + #endif + #if MI_DEBUG>3 + if (p != NULL && zero) { + mi_assert_expensive(mi_mem_is_zero(p, size)); + } + #endif + return p; +} + +// allocate a small block +mi_decl_nodiscard extern inline mi_decl_restrict void* mi_heap_malloc_small(mi_heap_t* heap, size_t size) mi_attr_noexcept { + return mi_heap_malloc_small_zero(heap, size, false); +} + +mi_decl_nodiscard extern inline mi_decl_restrict void* mi_malloc_small(size_t size) mi_attr_noexcept { + return mi_heap_malloc_small(mi_prim_get_default_heap(), size); +} + +// The main allocation function +extern inline void* _mi_heap_malloc_zero_ex(mi_heap_t* heap, size_t size, bool zero, size_t huge_alignment) mi_attr_noexcept { + if mi_likely(size <= MI_SMALL_SIZE_MAX) { + mi_assert_internal(huge_alignment == 0); + return mi_heap_malloc_small_zero(heap, size, zero); + } + else { + mi_assert(heap!=NULL); + mi_assert(heap->thread_id == 0 || heap->thread_id == _mi_thread_id()); // heaps are thread local + void* const p = _mi_malloc_generic(heap, size + MI_PADDING_SIZE, zero, huge_alignment); // note: size can overflow but it is detected in malloc_generic + mi_track_malloc(p,size,zero); + #if MI_STAT>1 + if (p != NULL) { + if (!mi_heap_is_initialized(heap)) { heap = mi_prim_get_default_heap(); } + mi_heap_stat_increase(heap, malloc, mi_usable_size(p)); + } + #endif + #if MI_DEBUG>3 + if (p != NULL && zero) { + mi_assert_expensive(mi_mem_is_zero(p, size)); + } + #endif + return p; + } +} + +extern inline void* _mi_heap_malloc_zero(mi_heap_t* heap, size_t size, bool zero) mi_attr_noexcept { + return _mi_heap_malloc_zero_ex(heap, size, zero, 0); +} + +mi_decl_nodiscard extern inline mi_decl_restrict void* mi_heap_malloc(mi_heap_t* heap, size_t size) mi_attr_noexcept { + return _mi_heap_malloc_zero(heap, size, false); +} + +mi_decl_nodiscard extern inline mi_decl_restrict void* mi_malloc(size_t size) mi_attr_noexcept { + return mi_heap_malloc(mi_prim_get_default_heap(), size); +} + +// zero initialized small block +mi_decl_nodiscard mi_decl_restrict void* mi_zalloc_small(size_t size) mi_attr_noexcept { + return mi_heap_malloc_small_zero(mi_prim_get_default_heap(), size, true); +} + +mi_decl_nodiscard extern inline mi_decl_restrict void* mi_heap_zalloc(mi_heap_t* heap, size_t size) mi_attr_noexcept { + return _mi_heap_malloc_zero(heap, size, true); +} + +mi_decl_nodiscard mi_decl_restrict void* mi_zalloc(size_t size) mi_attr_noexcept { + return mi_heap_zalloc(mi_prim_get_default_heap(),size); +} + + +// ------------------------------------------------------ +// Check for double free in secure and debug mode +// This is somewhat expensive so only enabled for secure mode 4 +// ------------------------------------------------------ + +#if (MI_ENCODE_FREELIST && (MI_SECURE>=4 || MI_DEBUG!=0)) +// linear check if the free list contains a specific element +static bool mi_list_contains(const mi_page_t* page, const mi_block_t* list, const mi_block_t* elem) { + while (list != NULL) { + if (elem==list) return true; + list = mi_block_next(page, list); + } + return false; +} + +static mi_decl_noinline bool mi_check_is_double_freex(const mi_page_t* page, const mi_block_t* block) { + // The decoded value is in the same page (or NULL). + // Walk the free lists to verify positively if it is already freed + if (mi_list_contains(page, page->free, block) || + mi_list_contains(page, page->local_free, block) || + mi_list_contains(page, mi_page_thread_free(page), block)) + { + _mi_error_message(EAGAIN, "double free detected of block %p with size %zu\n", block, mi_page_block_size(page)); + return true; + } + return false; +} + +#define mi_track_page(page,access) { size_t psize; void* pstart = _mi_page_start(_mi_page_segment(page),page,&psize); mi_track_mem_##access( pstart, psize); } + +static inline bool mi_check_is_double_free(const mi_page_t* page, const mi_block_t* block) { + bool is_double_free = false; + mi_block_t* n = mi_block_nextx(page, block, page->keys); // pretend it is freed, and get the decoded first field + if (((uintptr_t)n & (MI_INTPTR_SIZE-1))==0 && // quick check: aligned pointer? + (n==NULL || mi_is_in_same_page(block, n))) // quick check: in same page or NULL? + { + // Suspicous: decoded value a in block is in the same page (or NULL) -- maybe a double free? + // (continue in separate function to improve code generation) + is_double_free = mi_check_is_double_freex(page, block); + } + return is_double_free; +} +#else +static inline bool mi_check_is_double_free(const mi_page_t* page, const mi_block_t* block) { + MI_UNUSED(page); + MI_UNUSED(block); + return false; +} +#endif + +// --------------------------------------------------------------------------- +// Check for heap block overflow by setting up padding at the end of the block +// --------------------------------------------------------------------------- + +#if MI_PADDING // && !MI_TRACK_ENABLED +static bool mi_page_decode_padding(const mi_page_t* page, const mi_block_t* block, size_t* delta, size_t* bsize) { + *bsize = mi_page_usable_block_size(page); + const mi_padding_t* const padding = (mi_padding_t*)((uint8_t*)block + *bsize); + mi_track_mem_defined(padding,sizeof(mi_padding_t)); + *delta = padding->delta; + uint32_t canary = padding->canary; + uintptr_t keys[2]; + keys[0] = page->keys[0]; + keys[1] = page->keys[1]; + bool ok = ((uint32_t)mi_ptr_encode(page,block,keys) == canary && *delta <= *bsize); + mi_track_mem_noaccess(padding,sizeof(mi_padding_t)); + return ok; +} + +// Return the exact usable size of a block. +static size_t mi_page_usable_size_of(const mi_page_t* page, const mi_block_t* block) { + size_t bsize; + size_t delta; + bool ok = mi_page_decode_padding(page, block, &delta, &bsize); + mi_assert_internal(ok); mi_assert_internal(delta <= bsize); + return (ok ? bsize - delta : 0); +} + +// When a non-thread-local block is freed, it becomes part of the thread delayed free +// list that is freed later by the owning heap. If the exact usable size is too small to +// contain the pointer for the delayed list, then shrink the padding (by decreasing delta) +// so it will later not trigger an overflow error in `mi_free_block`. +void _mi_padding_shrink(const mi_page_t* page, const mi_block_t* block, const size_t min_size) { + size_t bsize; + size_t delta; + bool ok = mi_page_decode_padding(page, block, &delta, &bsize); + mi_assert_internal(ok); + if (!ok || (bsize - delta) >= min_size) return; // usually already enough space + mi_assert_internal(bsize >= min_size); + if (bsize < min_size) return; // should never happen + size_t new_delta = (bsize - min_size); + mi_assert_internal(new_delta < bsize); + mi_padding_t* padding = (mi_padding_t*)((uint8_t*)block + bsize); + mi_track_mem_defined(padding,sizeof(mi_padding_t)); + padding->delta = (uint32_t)new_delta; + mi_track_mem_noaccess(padding,sizeof(mi_padding_t)); +} +#else +static size_t mi_page_usable_size_of(const mi_page_t* page, const mi_block_t* block) { + MI_UNUSED(block); + return mi_page_usable_block_size(page); +} + +void _mi_padding_shrink(const mi_page_t* page, const mi_block_t* block, const size_t min_size) { + MI_UNUSED(page); + MI_UNUSED(block); + MI_UNUSED(min_size); +} +#endif + +#if MI_PADDING && MI_PADDING_CHECK + +static bool mi_verify_padding(const mi_page_t* page, const mi_block_t* block, size_t* size, size_t* wrong) { + size_t bsize; + size_t delta; + bool ok = mi_page_decode_padding(page, block, &delta, &bsize); + *size = *wrong = bsize; + if (!ok) return false; + mi_assert_internal(bsize >= delta); + *size = bsize - delta; + if (!mi_page_is_huge(page)) { + uint8_t* fill = (uint8_t*)block + bsize - delta; + const size_t maxpad = (delta > MI_MAX_ALIGN_SIZE ? MI_MAX_ALIGN_SIZE : delta); // check at most the first N padding bytes + mi_track_mem_defined(fill, maxpad); + for (size_t i = 0; i < maxpad; i++) { + if (fill[i] != MI_DEBUG_PADDING) { + *wrong = bsize - delta + i; + ok = false; + break; + } + } + mi_track_mem_noaccess(fill, maxpad); + } + return ok; +} + +static void mi_check_padding(const mi_page_t* page, const mi_block_t* block) { + size_t size; + size_t wrong; + if (!mi_verify_padding(page,block,&size,&wrong)) { + _mi_error_message(EFAULT, "buffer overflow in heap block %p of size %zu: write after %zu bytes\n", block, size, wrong ); + } +} + +#else + +static void mi_check_padding(const mi_page_t* page, const mi_block_t* block) { + MI_UNUSED(page); + MI_UNUSED(block); +} + +#endif + +// only maintain stats for smaller objects if requested +#if (MI_STAT>0) +static void mi_stat_free(const mi_page_t* page, const mi_block_t* block) { + #if (MI_STAT < 2) + MI_UNUSED(block); + #endif + mi_heap_t* const heap = mi_heap_get_default(); + const size_t bsize = mi_page_usable_block_size(page); + #if (MI_STAT>1) + const size_t usize = mi_page_usable_size_of(page, block); + mi_heap_stat_decrease(heap, malloc, usize); + #endif + if (bsize <= MI_MEDIUM_OBJ_SIZE_MAX) { + mi_heap_stat_decrease(heap, normal, bsize); + #if (MI_STAT > 1) + mi_heap_stat_decrease(heap, normal_bins[_mi_bin(bsize)], 1); + #endif + } + else if (bsize <= MI_LARGE_OBJ_SIZE_MAX) { + mi_heap_stat_decrease(heap, large, bsize); + } + else { + mi_heap_stat_decrease(heap, huge, bsize); + } +} +#else +static void mi_stat_free(const mi_page_t* page, const mi_block_t* block) { + MI_UNUSED(page); MI_UNUSED(block); +} +#endif + +#if MI_HUGE_PAGE_ABANDON +#if (MI_STAT>0) +// maintain stats for huge objects +static void mi_stat_huge_free(const mi_page_t* page) { + mi_heap_t* const heap = mi_heap_get_default(); + const size_t bsize = mi_page_block_size(page); // to match stats in `page.c:mi_page_huge_alloc` + if (bsize <= MI_LARGE_OBJ_SIZE_MAX) { + mi_heap_stat_decrease(heap, large, bsize); + } + else { + mi_heap_stat_decrease(heap, huge, bsize); + } +} +#else +static void mi_stat_huge_free(const mi_page_t* page) { + MI_UNUSED(page); +} +#endif +#endif + +// ------------------------------------------------------ +// Free +// ------------------------------------------------------ + +// multi-threaded free (or free in huge block if compiled with MI_HUGE_PAGE_ABANDON) +static mi_decl_noinline void _mi_free_block_mt(mi_page_t* page, mi_block_t* block) +{ + // The padding check may access the non-thread-owned page for the key values. + // that is safe as these are constant and the page won't be freed (as the block is not freed yet). + mi_check_padding(page, block); + _mi_padding_shrink(page, block, sizeof(mi_block_t)); // for small size, ensure we can fit the delayed thread pointers without triggering overflow detection + + // huge page segments are always abandoned and can be freed immediately + mi_segment_t* segment = _mi_page_segment(page); + if (segment->kind == MI_SEGMENT_HUGE) { + #if MI_HUGE_PAGE_ABANDON + // huge page segments are always abandoned and can be freed immediately + mi_stat_huge_free(page); + _mi_segment_huge_page_free(segment, page, block); + return; + #else + // huge pages are special as they occupy the entire segment + // as these are large we reset the memory occupied by the page so it is available to other threads + // (as the owning thread needs to actually free the memory later). + _mi_segment_huge_page_reset(segment, page, block); + #endif + } + + #if (MI_DEBUG>0) && !MI_TRACK_ENABLED && !MI_TSAN // note: when tracking, cannot use mi_usable_size with multi-threading + if (segment->kind != MI_SEGMENT_HUGE) { // not for huge segments as we just reset the content + memset(block, MI_DEBUG_FREED, mi_usable_size(block)); + } + #endif + + // Try to put the block on either the page-local thread free list, or the heap delayed free list. + mi_thread_free_t tfreex; + bool use_delayed; + mi_thread_free_t tfree = mi_atomic_load_relaxed(&page->xthread_free); + do { + use_delayed = (mi_tf_delayed(tfree) == MI_USE_DELAYED_FREE); + if mi_unlikely(use_delayed) { + // unlikely: this only happens on the first concurrent free in a page that is in the full list + tfreex = mi_tf_set_delayed(tfree,MI_DELAYED_FREEING); + } + else { + // usual: directly add to page thread_free list + mi_block_set_next(page, block, mi_tf_block(tfree)); + tfreex = mi_tf_set_block(tfree,block); + } + } while (!mi_atomic_cas_weak_release(&page->xthread_free, &tfree, tfreex)); + + if mi_unlikely(use_delayed) { + // racy read on `heap`, but ok because MI_DELAYED_FREEING is set (see `mi_heap_delete` and `mi_heap_collect_abandon`) + mi_heap_t* const heap = (mi_heap_t*)(mi_atomic_load_acquire(&page->xheap)); //mi_page_heap(page); + mi_assert_internal(heap != NULL); + if (heap != NULL) { + // add to the delayed free list of this heap. (do this atomically as the lock only protects heap memory validity) + mi_block_t* dfree = mi_atomic_load_ptr_relaxed(mi_block_t, &heap->thread_delayed_free); + do { + mi_block_set_nextx(heap,block,dfree, heap->keys); + } while (!mi_atomic_cas_ptr_weak_release(mi_block_t,&heap->thread_delayed_free, &dfree, block)); + } + + // and reset the MI_DELAYED_FREEING flag + tfree = mi_atomic_load_relaxed(&page->xthread_free); + do { + tfreex = tfree; + mi_assert_internal(mi_tf_delayed(tfree) == MI_DELAYED_FREEING); + tfreex = mi_tf_set_delayed(tfree,MI_NO_DELAYED_FREE); + } while (!mi_atomic_cas_weak_release(&page->xthread_free, &tfree, tfreex)); + } +} + +// regular free +static inline void _mi_free_block(mi_page_t* page, bool local, mi_block_t* block) +{ + // and push it on the free list + //const size_t bsize = mi_page_block_size(page); + if mi_likely(local) { + // owning thread can free a block directly + if mi_unlikely(mi_check_is_double_free(page, block)) return; + mi_check_padding(page, block); + #if (MI_DEBUG>0) && !MI_TRACK_ENABLED && !MI_TSAN + if (!mi_page_is_huge(page)) { // huge page content may be already decommitted + memset(block, MI_DEBUG_FREED, mi_page_block_size(page)); + } + #endif + mi_block_set_next(page, block, page->local_free); + page->local_free = block; + page->used--; + if mi_unlikely(mi_page_all_free(page)) { + _mi_page_retire(page); + } + else if mi_unlikely(mi_page_is_in_full(page)) { + _mi_page_unfull(page); + } + } + else { + _mi_free_block_mt(page,block); + } +} + + +// Adjust a block that was allocated aligned, to the actual start of the block in the page. +mi_block_t* _mi_page_ptr_unalign(const mi_segment_t* segment, const mi_page_t* page, const void* p) { + mi_assert_internal(page!=NULL && p!=NULL); + const size_t diff = (uint8_t*)p - _mi_page_start(segment, page, NULL); + const size_t adjust = (diff % mi_page_block_size(page)); + return (mi_block_t*)((uintptr_t)p - adjust); +} + + +void mi_decl_noinline _mi_free_generic(const mi_segment_t* segment, mi_page_t* page, bool is_local, void* p) mi_attr_noexcept { + mi_block_t* const block = (mi_page_has_aligned(page) ? _mi_page_ptr_unalign(segment, page, p) : (mi_block_t*)p); + mi_stat_free(page, block); // stat_free may access the padding + mi_track_free_size(block, mi_page_usable_size_of(page,block)); + _mi_free_block(page, is_local, block); +} + +// Get the segment data belonging to a pointer +// This is just a single `and` in assembly but does further checks in debug mode +// (and secure mode) if this was a valid pointer. +static inline mi_segment_t* mi_checked_ptr_segment(const void* p, const char* msg) +{ + MI_UNUSED(msg); + mi_assert(p != NULL); + +#if (MI_DEBUG>0) + if mi_unlikely(((uintptr_t)p & (MI_INTPTR_SIZE - 1)) != 0) { + _mi_error_message(EINVAL, "%s: invalid (unaligned) pointer: %p\n", msg, p); + return NULL; + } +#endif + + mi_segment_t* const segment = _mi_ptr_segment(p); + mi_assert_internal(segment != NULL); + +#if (MI_DEBUG>0) + if mi_unlikely(!mi_is_in_heap_region(p)) { + #if (MI_INTPTR_SIZE == 8 && defined(__linux__)) + if (((uintptr_t)p >> 40) != 0x7F) { // linux tends to align large blocks above 0x7F000000000 (issue #640) + #else + { + #endif + _mi_warning_message("%s: pointer might not point to a valid heap region: %p\n" + "(this may still be a valid very large allocation (over 64MiB))\n", msg, p); + if mi_likely(_mi_ptr_cookie(segment) == segment->cookie) { + _mi_warning_message("(yes, the previous pointer %p was valid after all)\n", p); + } + } + } +#endif +#if (MI_DEBUG>0 || MI_SECURE>=4) + if mi_unlikely(_mi_ptr_cookie(segment) != segment->cookie) { + _mi_error_message(EINVAL, "%s: pointer does not point to a valid heap space: %p\n", msg, p); + return NULL; + } +#endif + + return segment; +} + +// Free a block +// fast path written carefully to prevent spilling on the stack +void mi_free(void* p) mi_attr_noexcept +{ + if mi_unlikely(p == NULL) return; + mi_segment_t* const segment = mi_checked_ptr_segment(p,"mi_free"); + const bool is_local= (_mi_prim_thread_id() == mi_atomic_load_relaxed(&segment->thread_id)); + mi_page_t* const page = _mi_segment_page_of(segment, p); + + if mi_likely(is_local) { // thread-local free? + if mi_likely(page->flags.full_aligned == 0) // and it is not a full page (full pages need to move from the full bin), nor has aligned blocks (aligned blocks need to be unaligned) + { + mi_block_t* const block = (mi_block_t*)p; + if mi_unlikely(mi_check_is_double_free(page, block)) return; + mi_check_padding(page, block); + mi_stat_free(page, block); + #if (MI_DEBUG>0) && !MI_TRACK_ENABLED && !MI_TSAN + memset(block, MI_DEBUG_FREED, mi_page_block_size(page)); + #endif + mi_track_free_size(p, mi_page_usable_size_of(page,block)); // faster then mi_usable_size as we already know the page and that p is unaligned + mi_block_set_next(page, block, page->local_free); + page->local_free = block; + if mi_unlikely(--page->used == 0) { // using this expression generates better code than: page->used--; if (mi_page_all_free(page)) + _mi_page_retire(page); + } + } + else { + // page is full or contains (inner) aligned blocks; use generic path + _mi_free_generic(segment, page, true, p); + } + } + else { + // not thread-local; use generic path + _mi_free_generic(segment, page, false, p); + } +} + +// return true if successful +bool _mi_free_delayed_block(mi_block_t* block) { + // get segment and page + const mi_segment_t* const segment = _mi_ptr_segment(block); + mi_assert_internal(_mi_ptr_cookie(segment) == segment->cookie); + mi_assert_internal(_mi_thread_id() == segment->thread_id); + mi_page_t* const page = _mi_segment_page_of(segment, block); + + // Clear the no-delayed flag so delayed freeing is used again for this page. + // This must be done before collecting the free lists on this page -- otherwise + // some blocks may end up in the page `thread_free` list with no blocks in the + // heap `thread_delayed_free` list which may cause the page to be never freed! + // (it would only be freed if we happen to scan it in `mi_page_queue_find_free_ex`) + if (!_mi_page_try_use_delayed_free(page, MI_USE_DELAYED_FREE, false /* dont overwrite never delayed */)) { + return false; + } + + // collect all other non-local frees to ensure up-to-date `used` count + _mi_page_free_collect(page, false); + + // and free the block (possibly freeing the page as well since used is updated) + _mi_free_block(page, true, block); + return true; +} + +// Bytes available in a block +mi_decl_noinline static size_t mi_page_usable_aligned_size_of(const mi_segment_t* segment, const mi_page_t* page, const void* p) mi_attr_noexcept { + const mi_block_t* block = _mi_page_ptr_unalign(segment, page, p); + const size_t size = mi_page_usable_size_of(page, block); + const ptrdiff_t adjust = (uint8_t*)p - (uint8_t*)block; + mi_assert_internal(adjust >= 0 && (size_t)adjust <= size); + return (size - adjust); +} + +static inline size_t _mi_usable_size(const void* p, const char* msg) mi_attr_noexcept { + if (p == NULL) return 0; + const mi_segment_t* const segment = mi_checked_ptr_segment(p, msg); + const mi_page_t* const page = _mi_segment_page_of(segment, p); + if mi_likely(!mi_page_has_aligned(page)) { + const mi_block_t* block = (const mi_block_t*)p; + return mi_page_usable_size_of(page, block); + } + else { + // split out to separate routine for improved code generation + return mi_page_usable_aligned_size_of(segment, page, p); + } +} + +mi_decl_nodiscard size_t mi_usable_size(const void* p) mi_attr_noexcept { + return _mi_usable_size(p, "mi_usable_size"); +} + + +// ------------------------------------------------------ +// Allocation extensions +// ------------------------------------------------------ + +void mi_free_size(void* p, size_t size) mi_attr_noexcept { + MI_UNUSED_RELEASE(size); + mi_assert(p == NULL || size <= _mi_usable_size(p,"mi_free_size")); + mi_free(p); +} + +void mi_free_size_aligned(void* p, size_t size, size_t alignment) mi_attr_noexcept { + MI_UNUSED_RELEASE(alignment); + mi_assert(((uintptr_t)p % alignment) == 0); + mi_free_size(p,size); +} + +void mi_free_aligned(void* p, size_t alignment) mi_attr_noexcept { + MI_UNUSED_RELEASE(alignment); + mi_assert(((uintptr_t)p % alignment) == 0); + mi_free(p); +} + +mi_decl_nodiscard extern inline mi_decl_restrict void* mi_heap_calloc(mi_heap_t* heap, size_t count, size_t size) mi_attr_noexcept { + size_t total; + if (mi_count_size_overflow(count,size,&total)) return NULL; + return mi_heap_zalloc(heap,total); +} + +mi_decl_nodiscard mi_decl_restrict void* mi_calloc(size_t count, size_t size) mi_attr_noexcept { + return mi_heap_calloc(mi_prim_get_default_heap(),count,size); +} + +// Uninitialized `calloc` +mi_decl_nodiscard extern mi_decl_restrict void* mi_heap_mallocn(mi_heap_t* heap, size_t count, size_t size) mi_attr_noexcept { + size_t total; + if (mi_count_size_overflow(count, size, &total)) return NULL; + return mi_heap_malloc(heap, total); +} + +mi_decl_nodiscard mi_decl_restrict void* mi_mallocn(size_t count, size_t size) mi_attr_noexcept { + return mi_heap_mallocn(mi_prim_get_default_heap(),count,size); +} + +// Expand (or shrink) in place (or fail) +void* mi_expand(void* p, size_t newsize) mi_attr_noexcept { + #if MI_PADDING + // we do not shrink/expand with padding enabled + MI_UNUSED(p); MI_UNUSED(newsize); + return NULL; + #else + if (p == NULL) return NULL; + const size_t size = _mi_usable_size(p,"mi_expand"); + if (newsize > size) return NULL; + return p; // it fits + #endif +} + +void* _mi_heap_realloc_zero(mi_heap_t* heap, void* p, size_t newsize, bool zero) mi_attr_noexcept { + // if p == NULL then behave as malloc. + // else if size == 0 then reallocate to a zero-sized block (and don't return NULL, just as mi_malloc(0)). + // (this means that returning NULL always indicates an error, and `p` will not have been freed in that case.) + const size_t size = _mi_usable_size(p,"mi_realloc"); // also works if p == NULL (with size 0) + if mi_unlikely(newsize <= size && newsize >= (size / 2) && newsize > 0) { // note: newsize must be > 0 or otherwise we return NULL for realloc(NULL,0) + mi_assert_internal(p!=NULL); + // todo: do not track as the usable size is still the same in the free; adjust potential padding? + // mi_track_resize(p,size,newsize) + // if (newsize < size) { mi_track_mem_noaccess((uint8_t*)p + newsize, size - newsize); } + return p; // reallocation still fits and not more than 50% waste + } + void* newp = mi_heap_malloc(heap,newsize); + if mi_likely(newp != NULL) { + if (zero && newsize > size) { + // also set last word in the previous allocation to zero to ensure any padding is zero-initialized + const size_t start = (size >= sizeof(intptr_t) ? size - sizeof(intptr_t) : 0); + _mi_memzero((uint8_t*)newp + start, newsize - start); + } + else if (newsize == 0) { + ((uint8_t*)newp)[0] = 0; // work around for applications that expect zero-reallocation to be zero initialized (issue #725) + } + if mi_likely(p != NULL) { + const size_t copysize = (newsize > size ? size : newsize); + mi_track_mem_defined(p,copysize); // _mi_useable_size may be too large for byte precise memory tracking.. + _mi_memcpy(newp, p, copysize); + mi_free(p); // only free the original pointer if successful + } + } + return newp; +} + +mi_decl_nodiscard void* mi_heap_realloc(mi_heap_t* heap, void* p, size_t newsize) mi_attr_noexcept { + return _mi_heap_realloc_zero(heap, p, newsize, false); +} + +mi_decl_nodiscard void* mi_heap_reallocn(mi_heap_t* heap, void* p, size_t count, size_t size) mi_attr_noexcept { + size_t total; + if (mi_count_size_overflow(count, size, &total)) return NULL; + return mi_heap_realloc(heap, p, total); +} + + +// Reallocate but free `p` on errors +mi_decl_nodiscard void* mi_heap_reallocf(mi_heap_t* heap, void* p, size_t newsize) mi_attr_noexcept { + void* newp = mi_heap_realloc(heap, p, newsize); + if (newp==NULL && p!=NULL) mi_free(p); + return newp; +} + +mi_decl_nodiscard void* mi_heap_rezalloc(mi_heap_t* heap, void* p, size_t newsize) mi_attr_noexcept { + return _mi_heap_realloc_zero(heap, p, newsize, true); +} + +mi_decl_nodiscard void* mi_heap_recalloc(mi_heap_t* heap, void* p, size_t count, size_t size) mi_attr_noexcept { + size_t total; + if (mi_count_size_overflow(count, size, &total)) return NULL; + return mi_heap_rezalloc(heap, p, total); +} + + +mi_decl_nodiscard void* mi_realloc(void* p, size_t newsize) mi_attr_noexcept { + return mi_heap_realloc(mi_prim_get_default_heap(),p,newsize); +} + +mi_decl_nodiscard void* mi_reallocn(void* p, size_t count, size_t size) mi_attr_noexcept { + return mi_heap_reallocn(mi_prim_get_default_heap(),p,count,size); +} + +// Reallocate but free `p` on errors +mi_decl_nodiscard void* mi_reallocf(void* p, size_t newsize) mi_attr_noexcept { + return mi_heap_reallocf(mi_prim_get_default_heap(),p,newsize); +} + +mi_decl_nodiscard void* mi_rezalloc(void* p, size_t newsize) mi_attr_noexcept { + return mi_heap_rezalloc(mi_prim_get_default_heap(), p, newsize); +} + +mi_decl_nodiscard void* mi_recalloc(void* p, size_t count, size_t size) mi_attr_noexcept { + return mi_heap_recalloc(mi_prim_get_default_heap(), p, count, size); +} + + + +// ------------------------------------------------------ +// strdup, strndup, and realpath +// ------------------------------------------------------ + +// `strdup` using mi_malloc +mi_decl_nodiscard mi_decl_restrict char* mi_heap_strdup(mi_heap_t* heap, const char* s) mi_attr_noexcept { + if (s == NULL) return NULL; + size_t n = strlen(s); + char* t = (char*)mi_heap_malloc(heap,n+1); + if (t == NULL) return NULL; + _mi_memcpy(t, s, n); + t[n] = 0; + return t; +} + +mi_decl_nodiscard mi_decl_restrict char* mi_strdup(const char* s) mi_attr_noexcept { + return mi_heap_strdup(mi_prim_get_default_heap(), s); +} + +// `strndup` using mi_malloc +mi_decl_nodiscard mi_decl_restrict char* mi_heap_strndup(mi_heap_t* heap, const char* s, size_t n) mi_attr_noexcept { + if (s == NULL) return NULL; + const char* end = (const char*)memchr(s, 0, n); // find end of string in the first `n` characters (returns NULL if not found) + const size_t m = (end != NULL ? (size_t)(end - s) : n); // `m` is the minimum of `n` or the end-of-string + mi_assert_internal(m <= n); + char* t = (char*)mi_heap_malloc(heap, m+1); + if (t == NULL) return NULL; + _mi_memcpy(t, s, m); + t[m] = 0; + return t; +} + +mi_decl_nodiscard mi_decl_restrict char* mi_strndup(const char* s, size_t n) mi_attr_noexcept { + return mi_heap_strndup(mi_prim_get_default_heap(),s,n); +} + +#ifndef __wasi__ +// `realpath` using mi_malloc +#ifdef _WIN32 +#ifndef PATH_MAX +#define PATH_MAX MAX_PATH +#endif +#include +mi_decl_nodiscard mi_decl_restrict char* mi_heap_realpath(mi_heap_t* heap, const char* fname, char* resolved_name) mi_attr_noexcept { + // todo: use GetFullPathNameW to allow longer file names + char buf[PATH_MAX]; + DWORD res = GetFullPathNameA(fname, PATH_MAX, (resolved_name == NULL ? buf : resolved_name), NULL); + if (res == 0) { + errno = GetLastError(); return NULL; + } + else if (res > PATH_MAX) { + errno = EINVAL; return NULL; + } + else if (resolved_name != NULL) { + return resolved_name; + } + else { + return mi_heap_strndup(heap, buf, PATH_MAX); + } +} +#else +/* +#include // pathconf +static size_t mi_path_max(void) { + static size_t path_max = 0; + if (path_max <= 0) { + long m = pathconf("/",_PC_PATH_MAX); + if (m <= 0) path_max = 4096; // guess + else if (m < 256) path_max = 256; // at least 256 + else path_max = m; + } + return path_max; +} +*/ +char* mi_heap_realpath(mi_heap_t* heap, const char* fname, char* resolved_name) mi_attr_noexcept { + if (resolved_name != NULL) { + return realpath(fname,resolved_name); + } + else { + char* rname = realpath(fname, NULL); + if (rname == NULL) return NULL; + char* result = mi_heap_strdup(heap, rname); + free(rname); // use regular free! (which may be redirected to our free but that's ok) + return result; + } + /* + const size_t n = mi_path_max(); + char* buf = (char*)mi_malloc(n+1); + if (buf == NULL) { + errno = ENOMEM; + return NULL; + } + char* rname = realpath(fname,buf); + char* result = mi_heap_strndup(heap,rname,n); // ok if `rname==NULL` + mi_free(buf); + return result; + } + */ +} +#endif + +mi_decl_nodiscard mi_decl_restrict char* mi_realpath(const char* fname, char* resolved_name) mi_attr_noexcept { + return mi_heap_realpath(mi_prim_get_default_heap(),fname,resolved_name); +} +#endif + +/*------------------------------------------------------- +C++ new and new_aligned +The standard requires calling into `get_new_handler` and +throwing the bad_alloc exception on failure. If we compile +with a C++ compiler we can implement this precisely. If we +use a C compiler we cannot throw a `bad_alloc` exception +but we call `exit` instead (i.e. not returning). +-------------------------------------------------------*/ + +#ifdef __cplusplus +#include +static bool mi_try_new_handler(bool nothrow) { + #if defined(_MSC_VER) || (__cplusplus >= 201103L) + std::new_handler h = std::get_new_handler(); + #else + std::new_handler h = std::set_new_handler(); + std::set_new_handler(h); + #endif + if (h==NULL) { + _mi_error_message(ENOMEM, "out of memory in 'new'"); + if (!nothrow) { + throw std::bad_alloc(); + } + return false; + } + else { + h(); + return true; + } +} +#else +typedef void (*std_new_handler_t)(void); + +#if (defined(__GNUC__) || (defined(__clang__) && !defined(_MSC_VER))) // exclude clang-cl, see issue #631 +std_new_handler_t __attribute__((weak)) _ZSt15get_new_handlerv(void) { + return NULL; +} +static std_new_handler_t mi_get_new_handler(void) { + return _ZSt15get_new_handlerv(); +} +#else +// note: on windows we could dynamically link to `?get_new_handler@std@@YAP6AXXZXZ`. +static std_new_handler_t mi_get_new_handler() { + return NULL; +} +#endif + +static bool mi_try_new_handler(bool nothrow) { + std_new_handler_t h = mi_get_new_handler(); + if (h==NULL) { + _mi_error_message(ENOMEM, "out of memory in 'new'"); + if (!nothrow) { + abort(); // cannot throw in plain C, use abort + } + return false; + } + else { + h(); + return true; + } +} +#endif + +mi_decl_export mi_decl_noinline void* mi_heap_try_new(mi_heap_t* heap, size_t size, bool nothrow ) { + void* p = NULL; + while(p == NULL && mi_try_new_handler(nothrow)) { + p = mi_heap_malloc(heap,size); + } + return p; +} + +static mi_decl_noinline void* mi_try_new(size_t size, bool nothrow) { + return mi_heap_try_new(mi_prim_get_default_heap(), size, nothrow); +} + + +mi_decl_nodiscard mi_decl_restrict void* mi_heap_alloc_new(mi_heap_t* heap, size_t size) { + void* p = mi_heap_malloc(heap,size); + if mi_unlikely(p == NULL) return mi_heap_try_new(heap, size, false); + return p; +} + +mi_decl_nodiscard mi_decl_restrict void* mi_new(size_t size) { + return mi_heap_alloc_new(mi_prim_get_default_heap(), size); +} + + +mi_decl_nodiscard mi_decl_restrict void* mi_heap_alloc_new_n(mi_heap_t* heap, size_t count, size_t size) { + size_t total; + if mi_unlikely(mi_count_size_overflow(count, size, &total)) { + mi_try_new_handler(false); // on overflow we invoke the try_new_handler once to potentially throw std::bad_alloc + return NULL; + } + else { + return mi_heap_alloc_new(heap,total); + } +} + +mi_decl_nodiscard mi_decl_restrict void* mi_new_n(size_t count, size_t size) { + return mi_heap_alloc_new_n(mi_prim_get_default_heap(), size, count); +} + + +mi_decl_nodiscard mi_decl_restrict void* mi_new_nothrow(size_t size) mi_attr_noexcept { + void* p = mi_malloc(size); + if mi_unlikely(p == NULL) return mi_try_new(size, true); + return p; +} + +mi_decl_nodiscard mi_decl_restrict void* mi_new_aligned(size_t size, size_t alignment) { + void* p; + do { + p = mi_malloc_aligned(size, alignment); + } + while(p == NULL && mi_try_new_handler(false)); + return p; +} + +mi_decl_nodiscard mi_decl_restrict void* mi_new_aligned_nothrow(size_t size, size_t alignment) mi_attr_noexcept { + void* p; + do { + p = mi_malloc_aligned(size, alignment); + } + while(p == NULL && mi_try_new_handler(true)); + return p; +} + +mi_decl_nodiscard void* mi_new_realloc(void* p, size_t newsize) { + void* q; + do { + q = mi_realloc(p, newsize); + } while (q == NULL && mi_try_new_handler(false)); + return q; +} + +mi_decl_nodiscard void* mi_new_reallocn(void* p, size_t newcount, size_t size) { + size_t total; + if mi_unlikely(mi_count_size_overflow(newcount, size, &total)) { + mi_try_new_handler(false); // on overflow we invoke the try_new_handler once to potentially throw std::bad_alloc + return NULL; + } + else { + return mi_new_realloc(p, total); + } +} + +// ------------------------------------------------------ +// ensure explicit external inline definitions are emitted! +// ------------------------------------------------------ + +#ifdef __cplusplus +void* _mi_externs[] = { + (void*)&_mi_page_malloc, + (void*)&_mi_heap_malloc_zero, + (void*)&_mi_heap_malloc_zero_ex, + (void*)&mi_malloc, + (void*)&mi_malloc_small, + (void*)&mi_zalloc_small, + (void*)&mi_heap_malloc, + (void*)&mi_heap_zalloc, + (void*)&mi_heap_malloc_small, + // (void*)&mi_heap_alloc_new, + // (void*)&mi_heap_alloc_new_n +}; +#endif diff --git a/compat/mimalloc/arena.c b/compat/mimalloc/arena.c new file mode 100644 index 00000000000000..879ee9e7e773d4 --- /dev/null +++ b/compat/mimalloc/arena.c @@ -0,0 +1,935 @@ +/* ---------------------------------------------------------------------------- +Copyright (c) 2019-2023, Microsoft Research, Daan Leijen +This is free software; you can redistribute it and/or modify it under the +terms of the MIT license. A copy of the license can be found in the file +"LICENSE" at the root of this distribution. +-----------------------------------------------------------------------------*/ + +/* ---------------------------------------------------------------------------- +"Arenas" are fixed area's of OS memory from which we can allocate +large blocks (>= MI_ARENA_MIN_BLOCK_SIZE, 4MiB). +In contrast to the rest of mimalloc, the arenas are shared between +threads and need to be accessed using atomic operations. + +Arenas are used to for huge OS page (1GiB) reservations or for reserving +OS memory upfront which can be improve performance or is sometimes needed +on embedded devices. We can also employ this with WASI or `sbrk` systems +to reserve large arenas upfront and be able to reuse the memory more effectively. + +The arena allocation needs to be thread safe and we use an atomic bitmap to allocate. +-----------------------------------------------------------------------------*/ +#include "mimalloc.h" +#include "mimalloc/internal.h" +#include "mimalloc/atomic.h" + +#include // memset +#include // ENOMEM + +#include "bitmap.h" // atomic bitmap + +/* ----------------------------------------------------------- + Arena allocation +----------------------------------------------------------- */ + +// Block info: bit 0 contains the `in_use` bit, the upper bits the +// size in count of arena blocks. +typedef uintptr_t mi_block_info_t; +#define MI_ARENA_BLOCK_SIZE (MI_SEGMENT_SIZE) // 64MiB (must be at least MI_SEGMENT_ALIGN) +#define MI_ARENA_MIN_OBJ_SIZE (MI_ARENA_BLOCK_SIZE/2) // 32MiB +#define MI_MAX_ARENAS (112) // not more than 126 (since we use 7 bits in the memid and an arena index + 1) + +// A memory arena descriptor +typedef struct mi_arena_s { + mi_arena_id_t id; // arena id; 0 for non-specific + mi_memid_t memid; // memid of the memory area + _Atomic(uint8_t*) start; // the start of the memory area + size_t block_count; // size of the area in arena blocks (of `MI_ARENA_BLOCK_SIZE`) + size_t field_count; // number of bitmap fields (where `field_count * MI_BITMAP_FIELD_BITS >= block_count`) + size_t meta_size; // size of the arena structure itself (including its bitmaps) + mi_memid_t meta_memid; // memid of the arena structure itself (OS or static allocation) + int numa_node; // associated NUMA node + bool exclusive; // only allow allocations if specifically for this arena + bool is_large; // memory area consists of large- or huge OS pages (always committed) + _Atomic(size_t) search_idx; // optimization to start the search for free blocks + _Atomic(mi_msecs_t) purge_expire; // expiration time when blocks should be decommitted from `blocks_decommit`. + mi_bitmap_field_t* blocks_dirty; // are the blocks potentially non-zero? + mi_bitmap_field_t* blocks_committed; // are the blocks committed? (can be NULL for memory that cannot be decommitted) + mi_bitmap_field_t* blocks_purge; // blocks that can be (reset) decommitted. (can be NULL for memory that cannot be (reset) decommitted) + mi_bitmap_field_t blocks_inuse[1]; // in-place bitmap of in-use blocks (of size `field_count`) +} mi_arena_t; + + +// The available arenas +static mi_decl_cache_align _Atomic(mi_arena_t*) mi_arenas[MI_MAX_ARENAS]; +static mi_decl_cache_align _Atomic(size_t) mi_arena_count; // = 0 + + +//static bool mi_manage_os_memory_ex2(void* start, size_t size, bool is_large, int numa_node, bool exclusive, mi_memid_t memid, mi_arena_id_t* arena_id) mi_attr_noexcept; + +/* ----------------------------------------------------------- + Arena id's + id = arena_index + 1 +----------------------------------------------------------- */ + +static size_t mi_arena_id_index(mi_arena_id_t id) { + return (size_t)(id <= 0 ? MI_MAX_ARENAS : id - 1); +} + +static mi_arena_id_t mi_arena_id_create(size_t arena_index) { + mi_assert_internal(arena_index < MI_MAX_ARENAS); + return (int)arena_index + 1; +} + +mi_arena_id_t _mi_arena_id_none(void) { + return 0; +} + +static bool mi_arena_id_is_suitable(mi_arena_id_t arena_id, bool arena_is_exclusive, mi_arena_id_t req_arena_id) { + return ((!arena_is_exclusive && req_arena_id == _mi_arena_id_none()) || + (arena_id == req_arena_id)); +} + +bool _mi_arena_memid_is_suitable(mi_memid_t memid, mi_arena_id_t request_arena_id) { + if (memid.memkind == MI_MEM_ARENA) { + return mi_arena_id_is_suitable(memid.mem.arena.id, memid.mem.arena.is_exclusive, request_arena_id); + } + else { + return mi_arena_id_is_suitable(0, false, request_arena_id); + } +} + +bool _mi_arena_memid_is_os_allocated(mi_memid_t memid) { + return (memid.memkind == MI_MEM_OS); +} + +/* ----------------------------------------------------------- + Arena allocations get a (currently) 16-bit memory id where the + lower 8 bits are the arena id, and the upper bits the block index. +----------------------------------------------------------- */ + +static size_t mi_block_count_of_size(size_t size) { + return _mi_divide_up(size, MI_ARENA_BLOCK_SIZE); +} + +static size_t mi_arena_block_size(size_t bcount) { + return (bcount * MI_ARENA_BLOCK_SIZE); +} + +static size_t mi_arena_size(mi_arena_t* arena) { + return mi_arena_block_size(arena->block_count); +} + +static mi_memid_t mi_memid_create_arena(mi_arena_id_t id, bool is_exclusive, mi_bitmap_index_t bitmap_index) { + mi_memid_t memid = _mi_memid_create(MI_MEM_ARENA); + memid.mem.arena.id = id; + memid.mem.arena.block_index = bitmap_index; + memid.mem.arena.is_exclusive = is_exclusive; + return memid; +} + +static bool mi_arena_memid_indices(mi_memid_t memid, size_t* arena_index, mi_bitmap_index_t* bitmap_index) { + mi_assert_internal(memid.memkind == MI_MEM_ARENA); + *arena_index = mi_arena_id_index(memid.mem.arena.id); + *bitmap_index = memid.mem.arena.block_index; + return memid.mem.arena.is_exclusive; +} + + + +/* ----------------------------------------------------------- + Special static area for mimalloc internal structures + to avoid OS calls (for example, for the arena metadata) +----------------------------------------------------------- */ + +#define MI_ARENA_STATIC_MAX (MI_INTPTR_SIZE*MI_KiB) // 8 KiB on 64-bit + +static uint8_t mi_arena_static[MI_ARENA_STATIC_MAX]; +static _Atomic(size_t) mi_arena_static_top; + +static void* mi_arena_static_zalloc(size_t size, size_t alignment, mi_memid_t* memid) { + *memid = _mi_memid_none(); + if (size == 0 || size > MI_ARENA_STATIC_MAX) return NULL; + if ((mi_atomic_load_relaxed(&mi_arena_static_top) + size) > MI_ARENA_STATIC_MAX) return NULL; + + // try to claim space + if (alignment == 0) { alignment = 1; } + const size_t oversize = size + alignment - 1; + if (oversize > MI_ARENA_STATIC_MAX) return NULL; + const size_t oldtop = mi_atomic_add_acq_rel(&mi_arena_static_top, oversize); + size_t top = oldtop + oversize; + if (top > MI_ARENA_STATIC_MAX) { + // try to roll back, ok if this fails + mi_atomic_cas_strong_acq_rel(&mi_arena_static_top, &top, oldtop); + return NULL; + } + + // success + *memid = _mi_memid_create(MI_MEM_STATIC); + const size_t start = _mi_align_up(oldtop, alignment); + uint8_t* const p = &mi_arena_static[start]; + _mi_memzero(p, size); + return p; +} + +static void* mi_arena_meta_zalloc(size_t size, mi_memid_t* memid, mi_stats_t* stats) { + *memid = _mi_memid_none(); + + // try static + void* p = mi_arena_static_zalloc(size, MI_ALIGNMENT_MAX, memid); + if (p != NULL) return p; + + // or fall back to the OS + return _mi_os_alloc(size, memid, stats); +} + +static void mi_arena_meta_free(void* p, mi_memid_t memid, size_t size, mi_stats_t* stats) { + if (mi_memkind_is_os(memid.memkind)) { + _mi_os_free(p, size, memid, stats); + } + else { + mi_assert(memid.memkind == MI_MEM_STATIC); + } +} + +static void* mi_arena_block_start(mi_arena_t* arena, mi_bitmap_index_t bindex) { + return (arena->start + mi_arena_block_size(mi_bitmap_index_bit(bindex))); +} + + +/* ----------------------------------------------------------- + Thread safe allocation in an arena +----------------------------------------------------------- */ + +// claim the `blocks_inuse` bits +static bool mi_arena_try_claim(mi_arena_t* arena, size_t blocks, mi_bitmap_index_t* bitmap_idx) +{ + size_t idx = 0; // mi_atomic_load_relaxed(&arena->search_idx); // start from last search; ok to be relaxed as the exact start does not matter + if (_mi_bitmap_try_find_from_claim_across(arena->blocks_inuse, arena->field_count, idx, blocks, bitmap_idx)) { + mi_atomic_store_relaxed(&arena->search_idx, mi_bitmap_index_field(*bitmap_idx)); // start search from found location next time around + return true; + }; + return false; +} + + +/* ----------------------------------------------------------- + Arena Allocation +----------------------------------------------------------- */ + +static mi_decl_noinline void* mi_arena_try_alloc_at(mi_arena_t* arena, size_t arena_index, size_t needed_bcount, + bool commit, mi_memid_t* memid, mi_os_tld_t* tld) +{ + MI_UNUSED(arena_index); + mi_assert_internal(mi_arena_id_index(arena->id) == arena_index); + + mi_bitmap_index_t bitmap_index; + if (!mi_arena_try_claim(arena, needed_bcount, &bitmap_index)) return NULL; + + // claimed it! + void* p = mi_arena_block_start(arena, bitmap_index); + *memid = mi_memid_create_arena(arena->id, arena->exclusive, bitmap_index); + memid->is_pinned = arena->memid.is_pinned; + + // none of the claimed blocks should be scheduled for a decommit + if (arena->blocks_purge != NULL) { + // this is thread safe as a potential purge only decommits parts that are not yet claimed as used (in `blocks_inuse`). + _mi_bitmap_unclaim_across(arena->blocks_purge, arena->field_count, needed_bcount, bitmap_index); + } + + // set the dirty bits (todo: no need for an atomic op here?) + if (arena->memid.initially_zero && arena->blocks_dirty != NULL) { + memid->initially_zero = _mi_bitmap_claim_across(arena->blocks_dirty, arena->field_count, needed_bcount, bitmap_index, NULL); + } + + // set commit state + if (arena->blocks_committed == NULL) { + // always committed + memid->initially_committed = true; + } + else if (commit) { + // commit requested, but the range may not be committed as a whole: ensure it is committed now + memid->initially_committed = true; + bool any_uncommitted; + _mi_bitmap_claim_across(arena->blocks_committed, arena->field_count, needed_bcount, bitmap_index, &any_uncommitted); + if (any_uncommitted) { + bool commit_zero = false; + if (!_mi_os_commit(p, mi_arena_block_size(needed_bcount), &commit_zero, tld->stats)) { + memid->initially_committed = false; + } + else { + if (commit_zero) { memid->initially_zero = true; } + } + } + } + else { + // no need to commit, but check if already fully committed + memid->initially_committed = _mi_bitmap_is_claimed_across(arena->blocks_committed, arena->field_count, needed_bcount, bitmap_index); + } + + return p; +} + +// allocate in a speficic arena +static void* mi_arena_try_alloc_at_id(mi_arena_id_t arena_id, bool match_numa_node, int numa_node, size_t size, size_t alignment, + bool commit, bool allow_large, mi_arena_id_t req_arena_id, mi_memid_t* memid, mi_os_tld_t* tld ) +{ + MI_UNUSED_RELEASE(alignment); + mi_assert_internal(alignment <= MI_SEGMENT_ALIGN); + const size_t bcount = mi_block_count_of_size(size); + const size_t arena_index = mi_arena_id_index(arena_id); + mi_assert_internal(arena_index < mi_atomic_load_relaxed(&mi_arena_count)); + mi_assert_internal(size <= mi_arena_block_size(bcount)); + + // Check arena suitability + mi_arena_t* arena = mi_atomic_load_ptr_acquire(mi_arena_t, &mi_arenas[arena_index]); + if (arena == NULL) return NULL; + if (!allow_large && arena->is_large) return NULL; + if (!mi_arena_id_is_suitable(arena->id, arena->exclusive, req_arena_id)) return NULL; + if (req_arena_id == _mi_arena_id_none()) { // in not specific, check numa affinity + const bool numa_suitable = (numa_node < 0 || arena->numa_node < 0 || arena->numa_node == numa_node); + if (match_numa_node) { if (!numa_suitable) return NULL; } + else { if (numa_suitable) return NULL; } + } + + // try to allocate + void* p = mi_arena_try_alloc_at(arena, arena_index, bcount, commit, memid, tld); + mi_assert_internal(p == NULL || _mi_is_aligned(p, alignment)); + return p; +} + + +// allocate from an arena with fallback to the OS +static mi_decl_noinline void* mi_arena_try_alloc(int numa_node, size_t size, size_t alignment, + bool commit, bool allow_large, + mi_arena_id_t req_arena_id, mi_memid_t* memid, mi_os_tld_t* tld ) +{ + MI_UNUSED(alignment); + mi_assert_internal(alignment <= MI_SEGMENT_ALIGN); + const size_t max_arena = mi_atomic_load_relaxed(&mi_arena_count); + if mi_likely(max_arena == 0) return NULL; + + if (req_arena_id != _mi_arena_id_none()) { + // try a specific arena if requested + if (mi_arena_id_index(req_arena_id) < max_arena) { + void* p = mi_arena_try_alloc_at_id(req_arena_id, true, numa_node, size, alignment, commit, allow_large, req_arena_id, memid, tld); + if (p != NULL) return p; + } + } + else { + // try numa affine allocation + for (size_t i = 0; i < max_arena; i++) { + void* p = mi_arena_try_alloc_at_id(mi_arena_id_create(i), true, numa_node, size, alignment, commit, allow_large, req_arena_id, memid, tld); + if (p != NULL) return p; + } + + // try from another numa node instead.. + if (numa_node >= 0) { // if numa_node was < 0 (no specific affinity requested), all arena's have been tried already + for (size_t i = 0; i < max_arena; i++) { + void* p = mi_arena_try_alloc_at_id(mi_arena_id_create(i), false /* only proceed if not numa local */, numa_node, size, alignment, commit, allow_large, req_arena_id, memid, tld); + if (p != NULL) return p; + } + } + } + return NULL; +} + +// try to reserve a fresh arena space +static bool mi_arena_reserve(size_t req_size, bool allow_large, mi_arena_id_t req_arena_id, mi_arena_id_t *arena_id) +{ + if (_mi_preloading()) return false; // use OS only while pre loading + if (req_arena_id != _mi_arena_id_none()) return false; + + const size_t arena_count = mi_atomic_load_acquire(&mi_arena_count); + if (arena_count > (MI_MAX_ARENAS - 4)) return false; + + size_t arena_reserve = mi_option_get_size(mi_option_arena_reserve); + if (arena_reserve == 0) return false; + + if (!_mi_os_has_virtual_reserve()) { + arena_reserve = arena_reserve/4; // be conservative if virtual reserve is not supported (for some embedded systems for example) + } + arena_reserve = _mi_align_up(arena_reserve, MI_ARENA_BLOCK_SIZE); + if (arena_count >= 8 && arena_count <= 128) { + arena_reserve = ((size_t)1<<(arena_count/8)) * arena_reserve; // scale up the arena sizes exponentially + } + if (arena_reserve < req_size) return false; // should be able to at least handle the current allocation size + + // commit eagerly? + bool arena_commit = false; + if (mi_option_get(mi_option_arena_eager_commit) == 2) { arena_commit = _mi_os_has_overcommit(); } + else if (mi_option_get(mi_option_arena_eager_commit) == 1) { arena_commit = true; } + + return (mi_reserve_os_memory_ex(arena_reserve, arena_commit, allow_large, false /* exclusive */, arena_id) == 0); +} + + +void* _mi_arena_alloc_aligned(size_t size, size_t alignment, size_t align_offset, bool commit, bool allow_large, + mi_arena_id_t req_arena_id, mi_memid_t* memid, mi_os_tld_t* tld) +{ + mi_assert_internal(memid != NULL && tld != NULL); + mi_assert_internal(size > 0); + *memid = _mi_memid_none(); + + const int numa_node = _mi_os_numa_node(tld); // current numa node + + // try to allocate in an arena if the alignment is small enough and the object is not too small (as for heap meta data) + if (size >= MI_ARENA_MIN_OBJ_SIZE && alignment <= MI_SEGMENT_ALIGN && align_offset == 0) { + void* p = mi_arena_try_alloc(numa_node, size, alignment, commit, allow_large, req_arena_id, memid, tld); + if (p != NULL) return p; + + // otherwise, try to first eagerly reserve a new arena + if (req_arena_id == _mi_arena_id_none()) { + mi_arena_id_t arena_id = 0; + if (mi_arena_reserve(size, allow_large, req_arena_id, &arena_id)) { + // and try allocate in there + mi_assert_internal(req_arena_id == _mi_arena_id_none()); + p = mi_arena_try_alloc_at_id(arena_id, true, numa_node, size, alignment, commit, allow_large, req_arena_id, memid, tld); + if (p != NULL) return p; + } + } + } + + // if we cannot use OS allocation, return NULL + if (mi_option_is_enabled(mi_option_limit_os_alloc) || req_arena_id != _mi_arena_id_none()) { + errno = ENOMEM; + return NULL; + } + + // finally, fall back to the OS + if (align_offset > 0) { + return _mi_os_alloc_aligned_at_offset(size, alignment, align_offset, commit, allow_large, memid, tld->stats); + } + else { + return _mi_os_alloc_aligned(size, alignment, commit, allow_large, memid, tld->stats); + } +} + +void* _mi_arena_alloc(size_t size, bool commit, bool allow_large, mi_arena_id_t req_arena_id, mi_memid_t* memid, mi_os_tld_t* tld) +{ + return _mi_arena_alloc_aligned(size, MI_ARENA_BLOCK_SIZE, 0, commit, allow_large, req_arena_id, memid, tld); +} + + +void* mi_arena_area(mi_arena_id_t arena_id, size_t* size) { + if (size != NULL) *size = 0; + size_t arena_index = mi_arena_id_index(arena_id); + if (arena_index >= MI_MAX_ARENAS) return NULL; + mi_arena_t* arena = mi_atomic_load_ptr_acquire(mi_arena_t, &mi_arenas[arena_index]); + if (arena == NULL) return NULL; + if (size != NULL) { *size = mi_arena_block_size(arena->block_count); } + return arena->start; +} + + +/* ----------------------------------------------------------- + Arena purge +----------------------------------------------------------- */ + +static long mi_arena_purge_delay(void) { + // <0 = no purging allowed, 0=immediate purging, >0=milli-second delay + return (mi_option_get(mi_option_purge_delay) * mi_option_get(mi_option_arena_purge_mult)); +} + +// reset or decommit in an arena and update the committed/decommit bitmaps +// assumes we own the area (i.e. blocks_in_use is claimed by us) +static void mi_arena_purge(mi_arena_t* arena, size_t bitmap_idx, size_t blocks, mi_stats_t* stats) { + mi_assert_internal(arena->blocks_committed != NULL); + mi_assert_internal(arena->blocks_purge != NULL); + mi_assert_internal(!arena->memid.is_pinned); + const size_t size = mi_arena_block_size(blocks); + void* const p = mi_arena_block_start(arena, bitmap_idx); + bool needs_recommit; + if (_mi_bitmap_is_claimed_across(arena->blocks_committed, arena->field_count, blocks, bitmap_idx)) { + // all blocks are committed, we can purge freely + needs_recommit = _mi_os_purge(p, size, stats); + } + else { + // some blocks are not committed -- this can happen when a partially committed block is freed + // in `_mi_arena_free` and it is conservatively marked as uncommitted but still scheduled for a purge + // we need to ensure we do not try to reset (as that may be invalid for uncommitted memory), + // and also undo the decommit stats (as it was already adjusted) + mi_assert_internal(mi_option_is_enabled(mi_option_purge_decommits)); + needs_recommit = _mi_os_purge_ex(p, size, false /* allow reset? */, stats); + _mi_stat_increase(&stats->committed, size); + } + + // clear the purged blocks + _mi_bitmap_unclaim_across(arena->blocks_purge, arena->field_count, blocks, bitmap_idx); + // update committed bitmap + if (needs_recommit) { + _mi_bitmap_unclaim_across(arena->blocks_committed, arena->field_count, blocks, bitmap_idx); + } +} + +// Schedule a purge. This is usually delayed to avoid repeated decommit/commit calls. +// Note: assumes we (still) own the area as we may purge immediately +static void mi_arena_schedule_purge(mi_arena_t* arena, size_t bitmap_idx, size_t blocks, mi_stats_t* stats) { + mi_assert_internal(arena->blocks_purge != NULL); + const long delay = mi_arena_purge_delay(); + if (delay < 0) return; // is purging allowed at all? + + if (_mi_preloading() || delay == 0) { + // decommit directly + mi_arena_purge(arena, bitmap_idx, blocks, stats); + } + else { + // schedule decommit + mi_msecs_t expire = mi_atomic_loadi64_relaxed(&arena->purge_expire); + if (expire != 0) { + mi_atomic_addi64_acq_rel(&arena->purge_expire, delay/10); // add smallish extra delay + } + else { + mi_atomic_storei64_release(&arena->purge_expire, _mi_clock_now() + delay); + } + _mi_bitmap_claim_across(arena->blocks_purge, arena->field_count, blocks, bitmap_idx, NULL); + } +} + +// purge a range of blocks +// return true if the full range was purged. +// assumes we own the area (i.e. blocks_in_use is claimed by us) +static bool mi_arena_purge_range(mi_arena_t* arena, size_t idx, size_t startidx, size_t bitlen, size_t purge, mi_stats_t* stats) { + const size_t endidx = startidx + bitlen; + size_t bitidx = startidx; + bool all_purged = false; + while (bitidx < endidx) { + // count consequetive ones in the purge mask + size_t count = 0; + while (bitidx + count < endidx && (purge & ((size_t)1 << (bitidx + count))) != 0) { + count++; + } + if (count > 0) { + // found range to be purged + const mi_bitmap_index_t range_idx = mi_bitmap_index_create(idx, bitidx); + mi_arena_purge(arena, range_idx, count, stats); + if (count == bitlen) { + all_purged = true; + } + } + bitidx += (count+1); // +1 to skip the zero bit (or end) + } + return all_purged; +} + +// returns true if anything was purged +static bool mi_arena_try_purge(mi_arena_t* arena, mi_msecs_t now, bool force, mi_stats_t* stats) +{ + if (arena->memid.is_pinned || arena->blocks_purge == NULL) return false; + mi_msecs_t expire = mi_atomic_loadi64_relaxed(&arena->purge_expire); + if (expire == 0) return false; + if (!force && expire > now) return false; + + // reset expire (if not already set concurrently) + mi_atomic_casi64_strong_acq_rel(&arena->purge_expire, &expire, 0); + + // potential purges scheduled, walk through the bitmap + bool any_purged = false; + bool full_purge = true; + for (size_t i = 0; i < arena->field_count; i++) { + size_t purge = mi_atomic_load_relaxed(&arena->blocks_purge[i]); + if (purge != 0) { + size_t bitidx = 0; + while (bitidx < MI_BITMAP_FIELD_BITS) { + // find consequetive range of ones in the purge mask + size_t bitlen = 0; + while (bitidx + bitlen < MI_BITMAP_FIELD_BITS && (purge & ((size_t)1 << (bitidx + bitlen))) != 0) { + bitlen++; + } + // try to claim the longest range of corresponding in_use bits + const mi_bitmap_index_t bitmap_index = mi_bitmap_index_create(i, bitidx); + while( bitlen > 0 ) { + if (_mi_bitmap_try_claim(arena->blocks_inuse, arena->field_count, bitlen, bitmap_index)) { + break; + } + bitlen--; + } + // actual claimed bits at `in_use` + if (bitlen > 0) { + // read purge again now that we have the in_use bits + purge = mi_atomic_load_acquire(&arena->blocks_purge[i]); + if (!mi_arena_purge_range(arena, i, bitidx, bitlen, purge, stats)) { + full_purge = false; + } + any_purged = true; + // release the claimed `in_use` bits again + _mi_bitmap_unclaim(arena->blocks_inuse, arena->field_count, bitlen, bitmap_index); + } + bitidx += (bitlen+1); // +1 to skip the zero (or end) + } // while bitidx + } // purge != 0 + } + // if not fully purged, make sure to purge again in the future + if (!full_purge) { + const long delay = mi_arena_purge_delay(); + mi_msecs_t expected = 0; + mi_atomic_casi64_strong_acq_rel(&arena->purge_expire,&expected,_mi_clock_now() + delay); + } + return any_purged; +} + +static void mi_arenas_try_purge( bool force, bool visit_all, mi_stats_t* stats ) { + if (_mi_preloading() || mi_arena_purge_delay() <= 0) return; // nothing will be scheduled + + const size_t max_arena = mi_atomic_load_acquire(&mi_arena_count); + if (max_arena == 0) return; + + // allow only one thread to purge at a time + static mi_atomic_guard_t purge_guard; + mi_atomic_guard(&purge_guard) + { + mi_msecs_t now = _mi_clock_now(); + size_t max_purge_count = (visit_all ? max_arena : 1); + for (size_t i = 0; i < max_arena; i++) { + mi_arena_t* arena = mi_atomic_load_ptr_acquire(mi_arena_t, &mi_arenas[i]); + if (arena != NULL) { + if (mi_arena_try_purge(arena, now, force, stats)) { + if (max_purge_count <= 1) break; + max_purge_count--; + } + } + } + } +} + + +/* ----------------------------------------------------------- + Arena free +----------------------------------------------------------- */ + +void _mi_arena_free(void* p, size_t size, size_t committed_size, mi_memid_t memid, mi_stats_t* stats) { + mi_assert_internal(size > 0 && stats != NULL); + mi_assert_internal(committed_size <= size); + if (p==NULL) return; + if (size==0) return; + const bool all_committed = (committed_size == size); + + if (mi_memkind_is_os(memid.memkind)) { + // was a direct OS allocation, pass through + if (!all_committed && committed_size > 0) { + // if partially committed, adjust the committed stats (as `_mi_os_free` will increase decommit by the full size) + _mi_stat_decrease(&stats->committed, committed_size); + } + _mi_os_free(p, size, memid, stats); + } + else if (memid.memkind == MI_MEM_ARENA) { + // allocated in an arena + size_t arena_idx; + size_t bitmap_idx; + mi_arena_memid_indices(memid, &arena_idx, &bitmap_idx); + mi_assert_internal(arena_idx < MI_MAX_ARENAS); + mi_arena_t* arena = mi_atomic_load_ptr_acquire(mi_arena_t,&mi_arenas[arena_idx]); + mi_assert_internal(arena != NULL); + const size_t blocks = mi_block_count_of_size(size); + + // checks + if (arena == NULL) { + _mi_error_message(EINVAL, "trying to free from non-existent arena: %p, size %zu, memid: 0x%zx\n", p, size, memid); + return; + } + mi_assert_internal(arena->field_count > mi_bitmap_index_field(bitmap_idx)); + if (arena->field_count <= mi_bitmap_index_field(bitmap_idx)) { + _mi_error_message(EINVAL, "trying to free from non-existent arena block: %p, size %zu, memid: 0x%zx\n", p, size, memid); + return; + } + + // need to set all memory to undefined as some parts may still be marked as no_access (like padding etc.) + mi_track_mem_undefined(p,size); + + // potentially decommit + if (arena->memid.is_pinned || arena->blocks_committed == NULL) { + mi_assert_internal(all_committed); + } + else { + mi_assert_internal(arena->blocks_committed != NULL); + mi_assert_internal(arena->blocks_purge != NULL); + + if (!all_committed) { + // mark the entire range as no longer committed (so we recommit the full range when re-using) + _mi_bitmap_unclaim_across(arena->blocks_committed, arena->field_count, blocks, bitmap_idx); + mi_track_mem_noaccess(p,size); + if (committed_size > 0) { + // if partially committed, adjust the committed stats (is it will be recommitted when re-using) + // in the delayed purge, we now need to not count a decommit if the range is not marked as committed. + _mi_stat_decrease(&stats->committed, committed_size); + } + // note: if not all committed, it may be that the purge will reset/decommit the entire range + // that contains already decommitted parts. Since purge consistently uses reset or decommit that + // works (as we should never reset decommitted parts). + } + // (delay) purge the entire range + mi_arena_schedule_purge(arena, bitmap_idx, blocks, stats); + } + + // and make it available to others again + bool all_inuse = _mi_bitmap_unclaim_across(arena->blocks_inuse, arena->field_count, blocks, bitmap_idx); + if (!all_inuse) { + _mi_error_message(EAGAIN, "trying to free an already freed arena block: %p, size %zu\n", p, size); + return; + }; + } + else { + // arena was none, external, or static; nothing to do + mi_assert_internal(memid.memkind < MI_MEM_OS); + } + + // purge expired decommits + mi_arenas_try_purge(false, false, stats); +} + +// destroy owned arenas; this is unsafe and should only be done using `mi_option_destroy_on_exit` +// for dynamic libraries that are unloaded and need to release all their allocated memory. +static void mi_arenas_unsafe_destroy(void) { + const size_t max_arena = mi_atomic_load_relaxed(&mi_arena_count); + size_t new_max_arena = 0; + for (size_t i = 0; i < max_arena; i++) { + mi_arena_t* arena = mi_atomic_load_ptr_acquire(mi_arena_t, &mi_arenas[i]); + if (arena != NULL) { + if (arena->start != NULL && mi_memkind_is_os(arena->memid.memkind)) { + mi_atomic_store_ptr_release(mi_arena_t, &mi_arenas[i], NULL); + _mi_os_free(arena->start, mi_arena_size(arena), arena->memid, &_mi_stats_main); + } + else { + new_max_arena = i; + } + mi_arena_meta_free(arena, arena->meta_memid, arena->meta_size, &_mi_stats_main); + } + } + + // try to lower the max arena. + size_t expected = max_arena; + mi_atomic_cas_strong_acq_rel(&mi_arena_count, &expected, new_max_arena); +} + +// Purge the arenas; if `force_purge` is true, amenable parts are purged even if not yet expired +void _mi_arena_collect(bool force_purge, mi_stats_t* stats) { + mi_arenas_try_purge(force_purge, true /* visit all */, stats); +} + +// destroy owned arenas; this is unsafe and should only be done using `mi_option_destroy_on_exit` +// for dynamic libraries that are unloaded and need to release all their allocated memory. +void _mi_arena_unsafe_destroy_all(mi_stats_t* stats) { + mi_arenas_unsafe_destroy(); + _mi_arena_collect(true /* force purge */, stats); // purge non-owned arenas +} + +// Is a pointer inside any of our arenas? +bool _mi_arena_contains(const void* p) { + const size_t max_arena = mi_atomic_load_relaxed(&mi_arena_count); + for (size_t i = 0; i < max_arena; i++) { + mi_arena_t* arena = mi_atomic_load_ptr_acquire(mi_arena_t, &mi_arenas[i]); + if (arena != NULL && arena->start <= (const uint8_t*)p && arena->start + mi_arena_block_size(arena->block_count) > (const uint8_t*)p) { + return true; + } + } + return false; +} + + +/* ----------------------------------------------------------- + Add an arena. +----------------------------------------------------------- */ + +static bool mi_arena_add(mi_arena_t* arena, mi_arena_id_t* arena_id) { + mi_assert_internal(arena != NULL); + mi_assert_internal((uintptr_t)mi_atomic_load_ptr_relaxed(uint8_t,&arena->start) % MI_SEGMENT_ALIGN == 0); + mi_assert_internal(arena->block_count > 0); + if (arena_id != NULL) { *arena_id = -1; } + + size_t i = mi_atomic_increment_acq_rel(&mi_arena_count); + if (i >= MI_MAX_ARENAS) { + mi_atomic_decrement_acq_rel(&mi_arena_count); + return false; + } + arena->id = mi_arena_id_create(i); + mi_atomic_store_ptr_release(mi_arena_t,&mi_arenas[i], arena); + if (arena_id != NULL) { *arena_id = arena->id; } + return true; +} + +static bool mi_manage_os_memory_ex2(void* start, size_t size, bool is_large, int numa_node, bool exclusive, mi_memid_t memid, mi_arena_id_t* arena_id) mi_attr_noexcept +{ + if (arena_id != NULL) *arena_id = _mi_arena_id_none(); + if (size < MI_ARENA_BLOCK_SIZE) return false; + + if (is_large) { + mi_assert_internal(memid.initially_committed && memid.is_pinned); + } + + const size_t bcount = size / MI_ARENA_BLOCK_SIZE; + const size_t fields = _mi_divide_up(bcount, MI_BITMAP_FIELD_BITS); + const size_t bitmaps = (memid.is_pinned ? 2 : 4); + const size_t asize = sizeof(mi_arena_t) + (bitmaps*fields*sizeof(mi_bitmap_field_t)); + mi_memid_t meta_memid; + mi_arena_t* arena = (mi_arena_t*)mi_arena_meta_zalloc(asize, &meta_memid, &_mi_stats_main); // TODO: can we avoid allocating from the OS? + if (arena == NULL) return false; + + // already zero'd due to os_alloc + // _mi_memzero(arena, asize); + arena->id = _mi_arena_id_none(); + arena->memid = memid; + arena->exclusive = exclusive; + arena->meta_size = asize; + arena->meta_memid = meta_memid; + arena->block_count = bcount; + arena->field_count = fields; + arena->start = (uint8_t*)start; + arena->numa_node = numa_node; // TODO: or get the current numa node if -1? (now it allows anyone to allocate on -1) + arena->is_large = is_large; + arena->purge_expire = 0; + arena->search_idx = 0; + arena->blocks_dirty = &arena->blocks_inuse[fields]; // just after inuse bitmap + arena->blocks_committed = (arena->memid.is_pinned ? NULL : &arena->blocks_inuse[2*fields]); // just after dirty bitmap + arena->blocks_purge = (arena->memid.is_pinned ? NULL : &arena->blocks_inuse[3*fields]); // just after committed bitmap + // initialize committed bitmap? + if (arena->blocks_committed != NULL && arena->memid.initially_committed) { + memset((void*)arena->blocks_committed, 0xFF, fields*sizeof(mi_bitmap_field_t)); // cast to void* to avoid atomic warning + } + + // and claim leftover blocks if needed (so we never allocate there) + ptrdiff_t post = (fields * MI_BITMAP_FIELD_BITS) - bcount; + mi_assert_internal(post >= 0); + if (post > 0) { + // don't use leftover bits at the end + mi_bitmap_index_t postidx = mi_bitmap_index_create(fields - 1, MI_BITMAP_FIELD_BITS - post); + _mi_bitmap_claim(arena->blocks_inuse, fields, post, postidx, NULL); + } + return mi_arena_add(arena, arena_id); + +} + +bool mi_manage_os_memory_ex(void* start, size_t size, bool is_committed, bool is_large, bool is_zero, int numa_node, bool exclusive, mi_arena_id_t* arena_id) mi_attr_noexcept { + mi_memid_t memid = _mi_memid_create(MI_MEM_EXTERNAL); + memid.initially_committed = is_committed; + memid.initially_zero = is_zero; + memid.is_pinned = is_large; + return mi_manage_os_memory_ex2(start,size,is_large,numa_node,exclusive,memid, arena_id); +} + +// Reserve a range of regular OS memory +int mi_reserve_os_memory_ex(size_t size, bool commit, bool allow_large, bool exclusive, mi_arena_id_t* arena_id) mi_attr_noexcept { + if (arena_id != NULL) *arena_id = _mi_arena_id_none(); + size = _mi_align_up(size, MI_ARENA_BLOCK_SIZE); // at least one block + mi_memid_t memid; + void* start = _mi_os_alloc_aligned(size, MI_SEGMENT_ALIGN, commit, allow_large, &memid, &_mi_stats_main); + if (start == NULL) return ENOMEM; + const bool is_large = memid.is_pinned; // todo: use separate is_large field? + if (!mi_manage_os_memory_ex2(start, size, is_large, -1 /* numa node */, exclusive, memid, arena_id)) { + _mi_os_free_ex(start, size, commit, memid, &_mi_stats_main); + _mi_verbose_message("failed to reserve %zu k memory\n", _mi_divide_up(size, 1024)); + return ENOMEM; + } + _mi_verbose_message("reserved %zu KiB memory%s\n", _mi_divide_up(size, 1024), is_large ? " (in large os pages)" : ""); + return 0; +} + + +// Manage a range of regular OS memory +bool mi_manage_os_memory(void* start, size_t size, bool is_committed, bool is_large, bool is_zero, int numa_node) mi_attr_noexcept { + return mi_manage_os_memory_ex(start, size, is_committed, is_large, is_zero, numa_node, false /* exclusive? */, NULL); +} + +// Reserve a range of regular OS memory +int mi_reserve_os_memory(size_t size, bool commit, bool allow_large) mi_attr_noexcept { + return mi_reserve_os_memory_ex(size, commit, allow_large, false, NULL); +} + + +/* ----------------------------------------------------------- + Debugging +----------------------------------------------------------- */ + +static size_t mi_debug_show_bitmap(const char* prefix, mi_bitmap_field_t* fields, size_t field_count ) { + size_t inuse_count = 0; + for (size_t i = 0; i < field_count; i++) { + char buf[MI_BITMAP_FIELD_BITS + 1]; + uintptr_t field = mi_atomic_load_relaxed(&fields[i]); + for (size_t bit = 0; bit < MI_BITMAP_FIELD_BITS; bit++) { + bool inuse = ((((uintptr_t)1 << bit) & field) != 0); + if (inuse) inuse_count++; + buf[MI_BITMAP_FIELD_BITS - 1 - bit] = (inuse ? 'x' : '.'); + } + buf[MI_BITMAP_FIELD_BITS] = 0; + _mi_verbose_message("%s%s\n", prefix, buf); + } + return inuse_count; +} + +void mi_debug_show_arenas(void) mi_attr_noexcept { + size_t max_arenas = mi_atomic_load_relaxed(&mi_arena_count); + for (size_t i = 0; i < max_arenas; i++) { + mi_arena_t* arena = mi_atomic_load_ptr_relaxed(mi_arena_t, &mi_arenas[i]); + if (arena == NULL) break; + size_t inuse_count = 0; + _mi_verbose_message("arena %zu: %zu blocks with %zu fields\n", i, arena->block_count, arena->field_count); + inuse_count += mi_debug_show_bitmap(" ", arena->blocks_inuse, arena->field_count); + _mi_verbose_message(" blocks in use ('x'): %zu\n", inuse_count); + } +} + + +/* ----------------------------------------------------------- + Reserve a huge page arena. +----------------------------------------------------------- */ +// reserve at a specific numa node +int mi_reserve_huge_os_pages_at_ex(size_t pages, int numa_node, size_t timeout_msecs, bool exclusive, mi_arena_id_t* arena_id) mi_attr_noexcept { + if (arena_id != NULL) *arena_id = -1; + if (pages==0) return 0; + if (numa_node < -1) numa_node = -1; + if (numa_node >= 0) numa_node = numa_node % _mi_os_numa_node_count(); + size_t hsize = 0; + size_t pages_reserved = 0; + mi_memid_t memid; + void* p = _mi_os_alloc_huge_os_pages(pages, numa_node, timeout_msecs, &pages_reserved, &hsize, &memid); + if (p==NULL || pages_reserved==0) { + _mi_warning_message("failed to reserve %zu GiB huge pages\n", pages); + return ENOMEM; + } + _mi_verbose_message("numa node %i: reserved %zu GiB huge pages (of the %zu GiB requested)\n", numa_node, pages_reserved, pages); + + if (!mi_manage_os_memory_ex2(p, hsize, true, numa_node, exclusive, memid, arena_id)) { + _mi_os_free(p, hsize, memid, &_mi_stats_main); + return ENOMEM; + } + return 0; +} + +int mi_reserve_huge_os_pages_at(size_t pages, int numa_node, size_t timeout_msecs) mi_attr_noexcept { + return mi_reserve_huge_os_pages_at_ex(pages, numa_node, timeout_msecs, false, NULL); +} + +// reserve huge pages evenly among the given number of numa nodes (or use the available ones as detected) +int mi_reserve_huge_os_pages_interleave(size_t pages, size_t numa_nodes, size_t timeout_msecs) mi_attr_noexcept { + if (pages == 0) return 0; + + // pages per numa node + size_t numa_count = (numa_nodes > 0 ? numa_nodes : _mi_os_numa_node_count()); + if (numa_count <= 0) numa_count = 1; + const size_t pages_per = pages / numa_count; + const size_t pages_mod = pages % numa_count; + const size_t timeout_per = (timeout_msecs==0 ? 0 : (timeout_msecs / numa_count) + 50); + + // reserve evenly among numa nodes + for (size_t numa_node = 0; numa_node < numa_count && pages > 0; numa_node++) { + size_t node_pages = pages_per; // can be 0 + if (numa_node < pages_mod) node_pages++; + int err = mi_reserve_huge_os_pages_at(node_pages, (int)numa_node, timeout_per); + if (err) return err; + if (pages < node_pages) { + pages = 0; + } + else { + pages -= node_pages; + } + } + + return 0; +} + +int mi_reserve_huge_os_pages(size_t pages, double max_secs, size_t* pages_reserved) mi_attr_noexcept { + MI_UNUSED(max_secs); + _mi_warning_message("mi_reserve_huge_os_pages is deprecated: use mi_reserve_huge_os_pages_interleave/at instead\n"); + if (pages_reserved != NULL) *pages_reserved = 0; + int err = mi_reserve_huge_os_pages_interleave(pages, 0, (size_t)(max_secs * 1000.0)); + if (err==0 && pages_reserved!=NULL) *pages_reserved = pages; + return err; +} diff --git a/compat/mimalloc/bitmap.c b/compat/mimalloc/bitmap.c new file mode 100644 index 00000000000000..878f0ab3250a47 --- /dev/null +++ b/compat/mimalloc/bitmap.c @@ -0,0 +1,432 @@ +/* ---------------------------------------------------------------------------- +Copyright (c) 2019-2023 Microsoft Research, Daan Leijen +This is free software; you can redistribute it and/or modify it under the +terms of the MIT license. A copy of the license can be found in the file +"LICENSE" at the root of this distribution. +-----------------------------------------------------------------------------*/ + +/* ---------------------------------------------------------------------------- +Concurrent bitmap that can set/reset sequences of bits atomically, +represeted as an array of fields where each field is a machine word (`size_t`) + +There are two api's; the standard one cannot have sequences that cross +between the bitmap fields (and a sequence must be <= MI_BITMAP_FIELD_BITS). + +The `_across` postfixed functions do allow sequences that can cross over +between the fields. (This is used in arena allocation) +---------------------------------------------------------------------------- */ + +#include "mimalloc.h" +#include "mimalloc/internal.h" +#include "bitmap.h" + +/* ----------------------------------------------------------- + Bitmap definition +----------------------------------------------------------- */ + +// The bit mask for a given number of blocks at a specified bit index. +static inline size_t mi_bitmap_mask_(size_t count, size_t bitidx) { + mi_assert_internal(count + bitidx <= MI_BITMAP_FIELD_BITS); + mi_assert_internal(count > 0); + if (count >= MI_BITMAP_FIELD_BITS) return MI_BITMAP_FIELD_FULL; + if (count == 0) return 0; + return ((((size_t)1 << count) - 1) << bitidx); +} + + +/* ----------------------------------------------------------- + Claim a bit sequence atomically +----------------------------------------------------------- */ + +// Try to atomically claim a sequence of `count` bits in a single +// field at `idx` in `bitmap`. Returns `true` on success. +inline bool _mi_bitmap_try_find_claim_field(mi_bitmap_t bitmap, size_t idx, const size_t count, mi_bitmap_index_t* bitmap_idx) +{ + mi_assert_internal(bitmap_idx != NULL); + mi_assert_internal(count <= MI_BITMAP_FIELD_BITS); + mi_assert_internal(count > 0); + mi_bitmap_field_t* field = &bitmap[idx]; + size_t map = mi_atomic_load_relaxed(field); + if (map==MI_BITMAP_FIELD_FULL) return false; // short cut + + // search for 0-bit sequence of length count + const size_t mask = mi_bitmap_mask_(count, 0); + const size_t bitidx_max = MI_BITMAP_FIELD_BITS - count; + +#ifdef MI_HAVE_FAST_BITSCAN + size_t bitidx = mi_ctz(~map); // quickly find the first zero bit if possible +#else + size_t bitidx = 0; // otherwise start at 0 +#endif + size_t m = (mask << bitidx); // invariant: m == mask shifted by bitidx + + // scan linearly for a free range of zero bits + while (bitidx <= bitidx_max) { + const size_t mapm = (map & m); + if (mapm == 0) { // are the mask bits free at bitidx? + mi_assert_internal((m >> bitidx) == mask); // no overflow? + const size_t newmap = (map | m); + mi_assert_internal((newmap^map) >> bitidx == mask); + if (!mi_atomic_cas_strong_acq_rel(field, &map, newmap)) { // TODO: use weak cas here? + // no success, another thread claimed concurrently.. keep going (with updated `map`) + continue; + } + else { + // success, we claimed the bits! + *bitmap_idx = mi_bitmap_index_create(idx, bitidx); + return true; + } + } + else { + // on to the next bit range +#ifdef MI_HAVE_FAST_BITSCAN + mi_assert_internal(mapm != 0); + const size_t shift = (count == 1 ? 1 : (MI_INTPTR_BITS - mi_clz(mapm) - bitidx)); + mi_assert_internal(shift > 0 && shift <= count); +#else + const size_t shift = 1; +#endif + bitidx += shift; + m <<= shift; + } + } + // no bits found + return false; +} + +// Find `count` bits of 0 and set them to 1 atomically; returns `true` on success. +// Starts at idx, and wraps around to search in all `bitmap_fields` fields. +// `count` can be at most MI_BITMAP_FIELD_BITS and will never cross fields. +bool _mi_bitmap_try_find_from_claim(mi_bitmap_t bitmap, const size_t bitmap_fields, const size_t start_field_idx, const size_t count, mi_bitmap_index_t* bitmap_idx) { + size_t idx = start_field_idx; + for (size_t visited = 0; visited < bitmap_fields; visited++, idx++) { + if (idx >= bitmap_fields) { idx = 0; } // wrap + if (_mi_bitmap_try_find_claim_field(bitmap, idx, count, bitmap_idx)) { + return true; + } + } + return false; +} + +// Like _mi_bitmap_try_find_from_claim but with an extra predicate that must be fullfilled +bool _mi_bitmap_try_find_from_claim_pred(mi_bitmap_t bitmap, const size_t bitmap_fields, + const size_t start_field_idx, const size_t count, + mi_bitmap_pred_fun_t pred_fun, void* pred_arg, + mi_bitmap_index_t* bitmap_idx) { + size_t idx = start_field_idx; + for (size_t visited = 0; visited < bitmap_fields; visited++, idx++) { + if (idx >= bitmap_fields) idx = 0; // wrap + if (_mi_bitmap_try_find_claim_field(bitmap, idx, count, bitmap_idx)) { + if (pred_fun == NULL || pred_fun(*bitmap_idx, pred_arg)) { + return true; + } + // predicate returned false, unclaim and look further + _mi_bitmap_unclaim(bitmap, bitmap_fields, count, *bitmap_idx); + } + } + return false; +} + +// Set `count` bits at `bitmap_idx` to 0 atomically +// Returns `true` if all `count` bits were 1 previously. +bool _mi_bitmap_unclaim(mi_bitmap_t bitmap, size_t bitmap_fields, size_t count, mi_bitmap_index_t bitmap_idx) { + const size_t idx = mi_bitmap_index_field(bitmap_idx); + const size_t bitidx = mi_bitmap_index_bit_in_field(bitmap_idx); + const size_t mask = mi_bitmap_mask_(count, bitidx); + mi_assert_internal(bitmap_fields > idx); MI_UNUSED(bitmap_fields); + // mi_assert_internal((bitmap[idx] & mask) == mask); + const size_t prev = mi_atomic_and_acq_rel(&bitmap[idx], ~mask); + return ((prev & mask) == mask); +} + + +// Set `count` bits at `bitmap_idx` to 1 atomically +// Returns `true` if all `count` bits were 0 previously. `any_zero` is `true` if there was at least one zero bit. +bool _mi_bitmap_claim(mi_bitmap_t bitmap, size_t bitmap_fields, size_t count, mi_bitmap_index_t bitmap_idx, bool* any_zero) { + const size_t idx = mi_bitmap_index_field(bitmap_idx); + const size_t bitidx = mi_bitmap_index_bit_in_field(bitmap_idx); + const size_t mask = mi_bitmap_mask_(count, bitidx); + mi_assert_internal(bitmap_fields > idx); MI_UNUSED(bitmap_fields); + //mi_assert_internal(any_zero != NULL || (bitmap[idx] & mask) == 0); + size_t prev = mi_atomic_or_acq_rel(&bitmap[idx], mask); + if (any_zero != NULL) { *any_zero = ((prev & mask) != mask); } + return ((prev & mask) == 0); +} + +// Returns `true` if all `count` bits were 1. `any_ones` is `true` if there was at least one bit set to one. +static bool mi_bitmap_is_claimedx(mi_bitmap_t bitmap, size_t bitmap_fields, size_t count, mi_bitmap_index_t bitmap_idx, bool* any_ones) { + const size_t idx = mi_bitmap_index_field(bitmap_idx); + const size_t bitidx = mi_bitmap_index_bit_in_field(bitmap_idx); + const size_t mask = mi_bitmap_mask_(count, bitidx); + mi_assert_internal(bitmap_fields > idx); MI_UNUSED(bitmap_fields); + const size_t field = mi_atomic_load_relaxed(&bitmap[idx]); + if (any_ones != NULL) { *any_ones = ((field & mask) != 0); } + return ((field & mask) == mask); +} + +// Try to set `count` bits at `bitmap_idx` from 0 to 1 atomically. +// Returns `true` if successful when all previous `count` bits were 0. +bool _mi_bitmap_try_claim(mi_bitmap_t bitmap, size_t bitmap_fields, size_t count, mi_bitmap_index_t bitmap_idx) { + const size_t idx = mi_bitmap_index_field(bitmap_idx); + const size_t bitidx = mi_bitmap_index_bit_in_field(bitmap_idx); + const size_t mask = mi_bitmap_mask_(count, bitidx); + mi_assert_internal(bitmap_fields > idx); MI_UNUSED(bitmap_fields); + size_t expected = mi_atomic_load_relaxed(&bitmap[idx]); + do { + if ((expected & mask) != 0) return false; + } + while (!mi_atomic_cas_strong_acq_rel(&bitmap[idx], &expected, expected | mask)); + mi_assert_internal((expected & mask) == 0); + return true; +} + + +bool _mi_bitmap_is_claimed(mi_bitmap_t bitmap, size_t bitmap_fields, size_t count, mi_bitmap_index_t bitmap_idx) { + return mi_bitmap_is_claimedx(bitmap, bitmap_fields, count, bitmap_idx, NULL); +} + +bool _mi_bitmap_is_any_claimed(mi_bitmap_t bitmap, size_t bitmap_fields, size_t count, mi_bitmap_index_t bitmap_idx) { + bool any_ones; + mi_bitmap_is_claimedx(bitmap, bitmap_fields, count, bitmap_idx, &any_ones); + return any_ones; +} + + +//-------------------------------------------------------------------------- +// the `_across` functions work on bitmaps where sequences can cross over +// between the fields. This is used in arena allocation +//-------------------------------------------------------------------------- + +// Try to atomically claim a sequence of `count` bits starting from the field +// at `idx` in `bitmap` and crossing into subsequent fields. Returns `true` on success. +// Only needs to consider crossing into the next fields (see `mi_bitmap_try_find_from_claim_across`) +static bool mi_bitmap_try_find_claim_field_across(mi_bitmap_t bitmap, size_t bitmap_fields, size_t idx, const size_t count, const size_t retries, mi_bitmap_index_t* bitmap_idx) +{ + mi_assert_internal(bitmap_idx != NULL); + + // check initial trailing zeros + mi_bitmap_field_t* field = &bitmap[idx]; + size_t map = mi_atomic_load_relaxed(field); + const size_t initial = mi_clz(map); // count of initial zeros starting at idx + mi_assert_internal(initial <= MI_BITMAP_FIELD_BITS); + if (initial == 0) return false; + if (initial >= count) return _mi_bitmap_try_find_claim_field(bitmap, idx, count, bitmap_idx); // no need to cross fields (this case won't happen for us) + if (_mi_divide_up(count - initial, MI_BITMAP_FIELD_BITS) >= (bitmap_fields - idx)) return false; // not enough entries + + // scan ahead + size_t found = initial; + size_t mask = 0; // mask bits for the final field + while(found < count) { + field++; + map = mi_atomic_load_relaxed(field); + const size_t mask_bits = (found + MI_BITMAP_FIELD_BITS <= count ? MI_BITMAP_FIELD_BITS : (count - found)); + mi_assert_internal(mask_bits > 0 && mask_bits <= MI_BITMAP_FIELD_BITS); + mask = mi_bitmap_mask_(mask_bits, 0); + if ((map & mask) != 0) return false; // some part is already claimed + found += mask_bits; + } + mi_assert_internal(field < &bitmap[bitmap_fields]); + + // we found a range of contiguous zeros up to the final field; mask contains mask in the final field + // now try to claim the range atomically + mi_bitmap_field_t* const final_field = field; + const size_t final_mask = mask; + mi_bitmap_field_t* const initial_field = &bitmap[idx]; + const size_t initial_idx = MI_BITMAP_FIELD_BITS - initial; + const size_t initial_mask = mi_bitmap_mask_(initial, initial_idx); + + // initial field + size_t newmap; + field = initial_field; + map = mi_atomic_load_relaxed(field); + do { + newmap = (map | initial_mask); + if ((map & initial_mask) != 0) { goto rollback; }; + } while (!mi_atomic_cas_strong_acq_rel(field, &map, newmap)); + + // intermediate fields + while (++field < final_field) { + newmap = MI_BITMAP_FIELD_FULL; + map = 0; + if (!mi_atomic_cas_strong_acq_rel(field, &map, newmap)) { goto rollback; } + } + + // final field + mi_assert_internal(field == final_field); + map = mi_atomic_load_relaxed(field); + do { + newmap = (map | final_mask); + if ((map & final_mask) != 0) { goto rollback; } + } while (!mi_atomic_cas_strong_acq_rel(field, &map, newmap)); + + // claimed! + *bitmap_idx = mi_bitmap_index_create(idx, initial_idx); + return true; + +rollback: + // roll back intermediate fields + // (we just failed to claim `field` so decrement first) + while (--field > initial_field) { + newmap = 0; + map = MI_BITMAP_FIELD_FULL; + mi_assert_internal(mi_atomic_load_relaxed(field) == map); + mi_atomic_store_release(field, newmap); + } + if (field == initial_field) { // (if we failed on the initial field, `field + 1 == initial_field`) + map = mi_atomic_load_relaxed(field); + do { + mi_assert_internal((map & initial_mask) == initial_mask); + newmap = (map & ~initial_mask); + } while (!mi_atomic_cas_strong_acq_rel(field, &map, newmap)); + } + // retry? (we make a recursive call instead of goto to be able to use const declarations) + if (retries <= 2) { + return mi_bitmap_try_find_claim_field_across(bitmap, bitmap_fields, idx, count, retries+1, bitmap_idx); + } + else { + return false; + } +} + + +// Find `count` bits of zeros and set them to 1 atomically; returns `true` on success. +// Starts at idx, and wraps around to search in all `bitmap_fields` fields. +bool _mi_bitmap_try_find_from_claim_across(mi_bitmap_t bitmap, const size_t bitmap_fields, const size_t start_field_idx, const size_t count, mi_bitmap_index_t* bitmap_idx) { + mi_assert_internal(count > 0); + if (count <= 2) { + // we don't bother with crossover fields for small counts + return _mi_bitmap_try_find_from_claim(bitmap, bitmap_fields, start_field_idx, count, bitmap_idx); + } + + // visit the fields + size_t idx = start_field_idx; + for (size_t visited = 0; visited < bitmap_fields; visited++, idx++) { + if (idx >= bitmap_fields) { idx = 0; } // wrap + // first try to claim inside a field + if (count <= MI_BITMAP_FIELD_BITS) { + if (_mi_bitmap_try_find_claim_field(bitmap, idx, count, bitmap_idx)) { + return true; + } + } + // if that fails, then try to claim across fields + if (mi_bitmap_try_find_claim_field_across(bitmap, bitmap_fields, idx, count, 0, bitmap_idx)) { + return true; + } + } + return false; +} + +// Helper for masks across fields; returns the mid count, post_mask may be 0 +static size_t mi_bitmap_mask_across(mi_bitmap_index_t bitmap_idx, size_t bitmap_fields, size_t count, size_t* pre_mask, size_t* mid_mask, size_t* post_mask) { + MI_UNUSED(bitmap_fields); + const size_t bitidx = mi_bitmap_index_bit_in_field(bitmap_idx); + if mi_likely(bitidx + count <= MI_BITMAP_FIELD_BITS) { + *pre_mask = mi_bitmap_mask_(count, bitidx); + *mid_mask = 0; + *post_mask = 0; + mi_assert_internal(mi_bitmap_index_field(bitmap_idx) < bitmap_fields); + return 0; + } + else { + const size_t pre_bits = MI_BITMAP_FIELD_BITS - bitidx; + mi_assert_internal(pre_bits < count); + *pre_mask = mi_bitmap_mask_(pre_bits, bitidx); + count -= pre_bits; + const size_t mid_count = (count / MI_BITMAP_FIELD_BITS); + *mid_mask = MI_BITMAP_FIELD_FULL; + count %= MI_BITMAP_FIELD_BITS; + *post_mask = (count==0 ? 0 : mi_bitmap_mask_(count, 0)); + mi_assert_internal(mi_bitmap_index_field(bitmap_idx) + mid_count + (count==0 ? 0 : 1) < bitmap_fields); + return mid_count; + } +} + +// Set `count` bits at `bitmap_idx` to 0 atomically +// Returns `true` if all `count` bits were 1 previously. +bool _mi_bitmap_unclaim_across(mi_bitmap_t bitmap, size_t bitmap_fields, size_t count, mi_bitmap_index_t bitmap_idx) { + size_t idx = mi_bitmap_index_field(bitmap_idx); + size_t pre_mask; + size_t mid_mask; + size_t post_mask; + size_t mid_count = mi_bitmap_mask_across(bitmap_idx, bitmap_fields, count, &pre_mask, &mid_mask, &post_mask); + bool all_one = true; + mi_bitmap_field_t* field = &bitmap[idx]; + size_t prev = mi_atomic_and_acq_rel(field++, ~pre_mask); // clear first part + if ((prev & pre_mask) != pre_mask) all_one = false; + while(mid_count-- > 0) { + prev = mi_atomic_and_acq_rel(field++, ~mid_mask); // clear mid part + if ((prev & mid_mask) != mid_mask) all_one = false; + } + if (post_mask!=0) { + prev = mi_atomic_and_acq_rel(field, ~post_mask); // clear end part + if ((prev & post_mask) != post_mask) all_one = false; + } + return all_one; +} + +// Set `count` bits at `bitmap_idx` to 1 atomically +// Returns `true` if all `count` bits were 0 previously. `any_zero` is `true` if there was at least one zero bit. +bool _mi_bitmap_claim_across(mi_bitmap_t bitmap, size_t bitmap_fields, size_t count, mi_bitmap_index_t bitmap_idx, bool* pany_zero) { + size_t idx = mi_bitmap_index_field(bitmap_idx); + size_t pre_mask; + size_t mid_mask; + size_t post_mask; + size_t mid_count = mi_bitmap_mask_across(bitmap_idx, bitmap_fields, count, &pre_mask, &mid_mask, &post_mask); + bool all_zero = true; + bool any_zero = false; + _Atomic(size_t)*field = &bitmap[idx]; + size_t prev = mi_atomic_or_acq_rel(field++, pre_mask); + if ((prev & pre_mask) != 0) all_zero = false; + if ((prev & pre_mask) != pre_mask) any_zero = true; + while (mid_count-- > 0) { + prev = mi_atomic_or_acq_rel(field++, mid_mask); + if ((prev & mid_mask) != 0) all_zero = false; + if ((prev & mid_mask) != mid_mask) any_zero = true; + } + if (post_mask!=0) { + prev = mi_atomic_or_acq_rel(field, post_mask); + if ((prev & post_mask) != 0) all_zero = false; + if ((prev & post_mask) != post_mask) any_zero = true; + } + if (pany_zero != NULL) { *pany_zero = any_zero; } + return all_zero; +} + + +// Returns `true` if all `count` bits were 1. +// `any_ones` is `true` if there was at least one bit set to one. +static bool mi_bitmap_is_claimedx_across(mi_bitmap_t bitmap, size_t bitmap_fields, size_t count, mi_bitmap_index_t bitmap_idx, bool* pany_ones) { + size_t idx = mi_bitmap_index_field(bitmap_idx); + size_t pre_mask; + size_t mid_mask; + size_t post_mask; + size_t mid_count = mi_bitmap_mask_across(bitmap_idx, bitmap_fields, count, &pre_mask, &mid_mask, &post_mask); + bool all_ones = true; + bool any_ones = false; + mi_bitmap_field_t* field = &bitmap[idx]; + size_t prev = mi_atomic_load_relaxed(field++); + if ((prev & pre_mask) != pre_mask) all_ones = false; + if ((prev & pre_mask) != 0) any_ones = true; + while (mid_count-- > 0) { + prev = mi_atomic_load_relaxed(field++); + if ((prev & mid_mask) != mid_mask) all_ones = false; + if ((prev & mid_mask) != 0) any_ones = true; + } + if (post_mask!=0) { + prev = mi_atomic_load_relaxed(field); + if ((prev & post_mask) != post_mask) all_ones = false; + if ((prev & post_mask) != 0) any_ones = true; + } + if (pany_ones != NULL) { *pany_ones = any_ones; } + return all_ones; +} + +bool _mi_bitmap_is_claimed_across(mi_bitmap_t bitmap, size_t bitmap_fields, size_t count, mi_bitmap_index_t bitmap_idx) { + return mi_bitmap_is_claimedx_across(bitmap, bitmap_fields, count, bitmap_idx, NULL); +} + +bool _mi_bitmap_is_any_claimed_across(mi_bitmap_t bitmap, size_t bitmap_fields, size_t count, mi_bitmap_index_t bitmap_idx) { + bool any_ones; + mi_bitmap_is_claimedx_across(bitmap, bitmap_fields, count, bitmap_idx, &any_ones); + return any_ones; +} diff --git a/compat/mimalloc/bitmap.h b/compat/mimalloc/bitmap.h new file mode 100644 index 00000000000000..9ba15d5d6f09ea --- /dev/null +++ b/compat/mimalloc/bitmap.h @@ -0,0 +1,115 @@ +/* ---------------------------------------------------------------------------- +Copyright (c) 2019-2023 Microsoft Research, Daan Leijen +This is free software; you can redistribute it and/or modify it under the +terms of the MIT license. A copy of the license can be found in the file +"LICENSE" at the root of this distribution. +-----------------------------------------------------------------------------*/ + +/* ---------------------------------------------------------------------------- +Concurrent bitmap that can set/reset sequences of bits atomically, +represeted as an array of fields where each field is a machine word (`size_t`) + +There are two api's; the standard one cannot have sequences that cross +between the bitmap fields (and a sequence must be <= MI_BITMAP_FIELD_BITS). +(this is used in region allocation) + +The `_across` postfixed functions do allow sequences that can cross over +between the fields. (This is used in arena allocation) +---------------------------------------------------------------------------- */ +#pragma once +#ifndef MI_BITMAP_H +#define MI_BITMAP_H + +/* ----------------------------------------------------------- + Bitmap definition +----------------------------------------------------------- */ + +#define MI_BITMAP_FIELD_BITS (8*MI_SIZE_SIZE) +#define MI_BITMAP_FIELD_FULL (~((size_t)0)) // all bits set + +// An atomic bitmap of `size_t` fields +typedef _Atomic(size_t) mi_bitmap_field_t; +typedef mi_bitmap_field_t* mi_bitmap_t; + +// A bitmap index is the index of the bit in a bitmap. +typedef size_t mi_bitmap_index_t; + +// Create a bit index. +static inline mi_bitmap_index_t mi_bitmap_index_create(size_t idx, size_t bitidx) { + mi_assert_internal(bitidx < MI_BITMAP_FIELD_BITS); + return (idx*MI_BITMAP_FIELD_BITS) + bitidx; +} + +// Create a bit index. +static inline mi_bitmap_index_t mi_bitmap_index_create_from_bit(size_t full_bitidx) { + return mi_bitmap_index_create(full_bitidx / MI_BITMAP_FIELD_BITS, full_bitidx % MI_BITMAP_FIELD_BITS); +} + +// Get the field index from a bit index. +static inline size_t mi_bitmap_index_field(mi_bitmap_index_t bitmap_idx) { + return (bitmap_idx / MI_BITMAP_FIELD_BITS); +} + +// Get the bit index in a bitmap field +static inline size_t mi_bitmap_index_bit_in_field(mi_bitmap_index_t bitmap_idx) { + return (bitmap_idx % MI_BITMAP_FIELD_BITS); +} + +// Get the full bit index +static inline size_t mi_bitmap_index_bit(mi_bitmap_index_t bitmap_idx) { + return bitmap_idx; +} + +/* ----------------------------------------------------------- + Claim a bit sequence atomically +----------------------------------------------------------- */ + +// Try to atomically claim a sequence of `count` bits in a single +// field at `idx` in `bitmap`. Returns `true` on success. +bool _mi_bitmap_try_find_claim_field(mi_bitmap_t bitmap, size_t idx, const size_t count, mi_bitmap_index_t* bitmap_idx); + +// Starts at idx, and wraps around to search in all `bitmap_fields` fields. +// For now, `count` can be at most MI_BITMAP_FIELD_BITS and will never cross fields. +bool _mi_bitmap_try_find_from_claim(mi_bitmap_t bitmap, const size_t bitmap_fields, const size_t start_field_idx, const size_t count, mi_bitmap_index_t* bitmap_idx); + +// Like _mi_bitmap_try_find_from_claim but with an extra predicate that must be fullfilled +typedef bool (mi_cdecl *mi_bitmap_pred_fun_t)(mi_bitmap_index_t bitmap_idx, void* pred_arg); +bool _mi_bitmap_try_find_from_claim_pred(mi_bitmap_t bitmap, const size_t bitmap_fields, const size_t start_field_idx, const size_t count, mi_bitmap_pred_fun_t pred_fun, void* pred_arg, mi_bitmap_index_t* bitmap_idx); + +// Set `count` bits at `bitmap_idx` to 0 atomically +// Returns `true` if all `count` bits were 1 previously. +bool _mi_bitmap_unclaim(mi_bitmap_t bitmap, size_t bitmap_fields, size_t count, mi_bitmap_index_t bitmap_idx); + +// Try to set `count` bits at `bitmap_idx` from 0 to 1 atomically. +// Returns `true` if successful when all previous `count` bits were 0. +bool _mi_bitmap_try_claim(mi_bitmap_t bitmap, size_t bitmap_fields, size_t count, mi_bitmap_index_t bitmap_idx); + +// Set `count` bits at `bitmap_idx` to 1 atomically +// Returns `true` if all `count` bits were 0 previously. `any_zero` is `true` if there was at least one zero bit. +bool _mi_bitmap_claim(mi_bitmap_t bitmap, size_t bitmap_fields, size_t count, mi_bitmap_index_t bitmap_idx, bool* any_zero); + +bool _mi_bitmap_is_claimed(mi_bitmap_t bitmap, size_t bitmap_fields, size_t count, mi_bitmap_index_t bitmap_idx); +bool _mi_bitmap_is_any_claimed(mi_bitmap_t bitmap, size_t bitmap_fields, size_t count, mi_bitmap_index_t bitmap_idx); + + +//-------------------------------------------------------------------------- +// the `_across` functions work on bitmaps where sequences can cross over +// between the fields. This is used in arena allocation +//-------------------------------------------------------------------------- + +// Find `count` bits of zeros and set them to 1 atomically; returns `true` on success. +// Starts at idx, and wraps around to search in all `bitmap_fields` fields. +bool _mi_bitmap_try_find_from_claim_across(mi_bitmap_t bitmap, const size_t bitmap_fields, const size_t start_field_idx, const size_t count, mi_bitmap_index_t* bitmap_idx); + +// Set `count` bits at `bitmap_idx` to 0 atomically +// Returns `true` if all `count` bits were 1 previously. +bool _mi_bitmap_unclaim_across(mi_bitmap_t bitmap, size_t bitmap_fields, size_t count, mi_bitmap_index_t bitmap_idx); + +// Set `count` bits at `bitmap_idx` to 1 atomically +// Returns `true` if all `count` bits were 0 previously. `any_zero` is `true` if there was at least one zero bit. +bool _mi_bitmap_claim_across(mi_bitmap_t bitmap, size_t bitmap_fields, size_t count, mi_bitmap_index_t bitmap_idx, bool* pany_zero); + +bool _mi_bitmap_is_claimed_across(mi_bitmap_t bitmap, size_t bitmap_fields, size_t count, mi_bitmap_index_t bitmap_idx); +bool _mi_bitmap_is_any_claimed_across(mi_bitmap_t bitmap, size_t bitmap_fields, size_t count, mi_bitmap_index_t bitmap_idx); + +#endif diff --git a/compat/mimalloc/heap.c b/compat/mimalloc/heap.c new file mode 100644 index 00000000000000..dab8c4bf8ae388 --- /dev/null +++ b/compat/mimalloc/heap.c @@ -0,0 +1,626 @@ +/*---------------------------------------------------------------------------- +Copyright (c) 2018-2021, Microsoft Research, Daan Leijen +This is free software; you can redistribute it and/or modify it under the +terms of the MIT license. A copy of the license can be found in the file +"LICENSE" at the root of this distribution. +-----------------------------------------------------------------------------*/ + +#include "mimalloc.h" +#include "mimalloc/internal.h" +#include "mimalloc/atomic.h" +#include "mimalloc/prim.h" // mi_prim_get_default_heap + +#include // memset, memcpy + +#if defined(_MSC_VER) && (_MSC_VER < 1920) +#pragma warning(disable:4204) // non-constant aggregate initializer +#endif + +/* ----------------------------------------------------------- + Helpers +----------------------------------------------------------- */ + +// return `true` if ok, `false` to break +typedef bool (heap_page_visitor_fun)(mi_heap_t* heap, mi_page_queue_t* pq, mi_page_t* page, void* arg1, void* arg2); + +// Visit all pages in a heap; returns `false` if break was called. +static bool mi_heap_visit_pages(mi_heap_t* heap, heap_page_visitor_fun* fn, void* arg1, void* arg2) +{ + if (heap==NULL || heap->page_count==0) return 0; + + // visit all pages + #if MI_DEBUG>1 + size_t total = heap->page_count; + size_t count = 0; + #endif + + for (size_t i = 0; i <= MI_BIN_FULL; i++) { + mi_page_queue_t* pq = &heap->pages[i]; + mi_page_t* page = pq->first; + while(page != NULL) { + mi_page_t* next = page->next; // save next in case the page gets removed from the queue + mi_assert_internal(mi_page_heap(page) == heap); + #if MI_DEBUG>1 + count++; + #endif + if (!fn(heap, pq, page, arg1, arg2)) return false; + page = next; // and continue + } + } + mi_assert_internal(count == total); + return true; +} + + +#if MI_DEBUG>=2 +static bool mi_heap_page_is_valid(mi_heap_t* heap, mi_page_queue_t* pq, mi_page_t* page, void* arg1, void* arg2) { + MI_UNUSED(arg1); + MI_UNUSED(arg2); + MI_UNUSED(pq); + mi_assert_internal(mi_page_heap(page) == heap); + mi_segment_t* segment = _mi_page_segment(page); + mi_assert_internal(segment->thread_id == heap->thread_id); + mi_assert_expensive(_mi_page_is_valid(page)); + return true; +} +#endif +#if MI_DEBUG>=3 +static bool mi_heap_is_valid(mi_heap_t* heap) { + mi_assert_internal(heap!=NULL); + mi_heap_visit_pages(heap, &mi_heap_page_is_valid, NULL, NULL); + return true; +} +#endif + + + + +/* ----------------------------------------------------------- + "Collect" pages by migrating `local_free` and `thread_free` + lists and freeing empty pages. This is done when a thread + stops (and in that case abandons pages if there are still + blocks alive) +----------------------------------------------------------- */ + +typedef enum mi_collect_e { + MI_NORMAL, + MI_FORCE, + MI_ABANDON +} mi_collect_t; + + +static bool mi_heap_page_collect(mi_heap_t* heap, mi_page_queue_t* pq, mi_page_t* page, void* arg_collect, void* arg2 ) { + MI_UNUSED(arg2); + MI_UNUSED(heap); + mi_assert_internal(mi_heap_page_is_valid(heap, pq, page, NULL, NULL)); + mi_collect_t collect = *((mi_collect_t*)arg_collect); + _mi_page_free_collect(page, collect >= MI_FORCE); + if (mi_page_all_free(page)) { + // no more used blocks, free the page. + // note: this will free retired pages as well. + _mi_page_free(page, pq, collect >= MI_FORCE); + } + else if (collect == MI_ABANDON) { + // still used blocks but the thread is done; abandon the page + _mi_page_abandon(page, pq); + } + return true; // don't break +} + +static bool mi_heap_page_never_delayed_free(mi_heap_t* heap, mi_page_queue_t* pq, mi_page_t* page, void* arg1, void* arg2) { + MI_UNUSED(arg1); + MI_UNUSED(arg2); + MI_UNUSED(heap); + MI_UNUSED(pq); + _mi_page_use_delayed_free(page, MI_NEVER_DELAYED_FREE, false); + return true; // don't break +} + +static void mi_heap_collect_ex(mi_heap_t* heap, mi_collect_t collect) +{ + if (heap==NULL || !mi_heap_is_initialized(heap)) return; + + const bool force = collect >= MI_FORCE; + _mi_deferred_free(heap, force); + + // note: never reclaim on collect but leave it to threads that need storage to reclaim + const bool force_main = + #ifdef NDEBUG + collect == MI_FORCE + #else + collect >= MI_FORCE + #endif + && _mi_is_main_thread() && mi_heap_is_backing(heap) && !heap->no_reclaim; + + if (force_main) { + // the main thread is abandoned (end-of-program), try to reclaim all abandoned segments. + // if all memory is freed by now, all segments should be freed. + _mi_abandoned_reclaim_all(heap, &heap->tld->segments); + } + + // if abandoning, mark all pages to no longer add to delayed_free + if (collect == MI_ABANDON) { + mi_heap_visit_pages(heap, &mi_heap_page_never_delayed_free, NULL, NULL); + } + + // free all current thread delayed blocks. + // (if abandoning, after this there are no more thread-delayed references into the pages.) + _mi_heap_delayed_free_all(heap); + + // collect retired pages + _mi_heap_collect_retired(heap, force); + + // collect all pages owned by this thread + mi_heap_visit_pages(heap, &mi_heap_page_collect, &collect, NULL); + mi_assert_internal( collect != MI_ABANDON || mi_atomic_load_ptr_acquire(mi_block_t,&heap->thread_delayed_free) == NULL ); + + // collect abandoned segments (in particular, purge expired parts of segments in the abandoned segment list) + // note: forced purge can be quite expensive if many threads are created/destroyed so we do not force on abandonment + _mi_abandoned_collect(heap, collect == MI_FORCE /* force? */, &heap->tld->segments); + + // collect segment local caches + if (force) { + _mi_segment_thread_collect(&heap->tld->segments); + } + + // collect regions on program-exit (or shared library unload) + if (force && _mi_is_main_thread() && mi_heap_is_backing(heap)) { + _mi_thread_data_collect(); // collect thread data cache + _mi_arena_collect(true /* force purge */, &heap->tld->stats); + } +} + +void _mi_heap_collect_abandon(mi_heap_t* heap) { + mi_heap_collect_ex(heap, MI_ABANDON); +} + +void mi_heap_collect(mi_heap_t* heap, bool force) mi_attr_noexcept { + mi_heap_collect_ex(heap, (force ? MI_FORCE : MI_NORMAL)); +} + +void mi_collect(bool force) mi_attr_noexcept { + mi_heap_collect(mi_prim_get_default_heap(), force); +} + + +/* ----------------------------------------------------------- + Heap new +----------------------------------------------------------- */ + +mi_heap_t* mi_heap_get_default(void) { + mi_thread_init(); + return mi_prim_get_default_heap(); +} + +static bool mi_heap_is_default(const mi_heap_t* heap) { + return (heap == mi_prim_get_default_heap()); +} + + +mi_heap_t* mi_heap_get_backing(void) { + mi_heap_t* heap = mi_heap_get_default(); + mi_assert_internal(heap!=NULL); + mi_heap_t* bheap = heap->tld->heap_backing; + mi_assert_internal(bheap!=NULL); + mi_assert_internal(bheap->thread_id == _mi_thread_id()); + return bheap; +} + +mi_decl_nodiscard mi_heap_t* mi_heap_new_in_arena(mi_arena_id_t arena_id) { + mi_heap_t* bheap = mi_heap_get_backing(); + mi_heap_t* heap = mi_heap_malloc_tp(bheap, mi_heap_t); // todo: OS allocate in secure mode? + if (heap == NULL) return NULL; + _mi_memcpy_aligned(heap, &_mi_heap_empty, sizeof(mi_heap_t)); + heap->tld = bheap->tld; + heap->thread_id = _mi_thread_id(); + heap->arena_id = arena_id; + _mi_random_split(&bheap->random, &heap->random); + heap->cookie = _mi_heap_random_next(heap) | 1; + heap->keys[0] = _mi_heap_random_next(heap); + heap->keys[1] = _mi_heap_random_next(heap); + heap->no_reclaim = true; // don't reclaim abandoned pages or otherwise destroy is unsafe + // push on the thread local heaps list + heap->next = heap->tld->heaps; + heap->tld->heaps = heap; + return heap; +} + +mi_decl_nodiscard mi_heap_t* mi_heap_new(void) { + return mi_heap_new_in_arena(_mi_arena_id_none()); +} + +bool _mi_heap_memid_is_suitable(mi_heap_t* heap, mi_memid_t memid) { + return _mi_arena_memid_is_suitable(memid, heap->arena_id); +} + +uintptr_t _mi_heap_random_next(mi_heap_t* heap) { + return _mi_random_next(&heap->random); +} + +// zero out the page queues +static void mi_heap_reset_pages(mi_heap_t* heap) { + mi_assert_internal(heap != NULL); + mi_assert_internal(mi_heap_is_initialized(heap)); + // TODO: copy full empty heap instead? + memset(&heap->pages_free_direct, 0, sizeof(heap->pages_free_direct)); + _mi_memcpy_aligned(&heap->pages, &_mi_heap_empty.pages, sizeof(heap->pages)); + heap->thread_delayed_free = NULL; + heap->page_count = 0; +} + +// called from `mi_heap_destroy` and `mi_heap_delete` to free the internal heap resources. +static void mi_heap_free(mi_heap_t* heap) { + mi_assert(heap != NULL); + mi_assert_internal(mi_heap_is_initialized(heap)); + if (heap==NULL || !mi_heap_is_initialized(heap)) return; + if (mi_heap_is_backing(heap)) return; // dont free the backing heap + + // reset default + if (mi_heap_is_default(heap)) { + _mi_heap_set_default_direct(heap->tld->heap_backing); + } + + // remove ourselves from the thread local heaps list + // linear search but we expect the number of heaps to be relatively small + mi_heap_t* prev = NULL; + mi_heap_t* curr = heap->tld->heaps; + while (curr != heap && curr != NULL) { + prev = curr; + curr = curr->next; + } + mi_assert_internal(curr == heap); + if (curr == heap) { + if (prev != NULL) { prev->next = heap->next; } + else { heap->tld->heaps = heap->next; } + } + mi_assert_internal(heap->tld->heaps != NULL); + + // and free the used memory + mi_free(heap); +} + + +/* ----------------------------------------------------------- + Heap destroy +----------------------------------------------------------- */ + +static bool _mi_heap_page_destroy(mi_heap_t* heap, mi_page_queue_t* pq, mi_page_t* page, void* arg1, void* arg2) { + MI_UNUSED(arg1); + MI_UNUSED(arg2); + MI_UNUSED(heap); + MI_UNUSED(pq); + + // ensure no more thread_delayed_free will be added + _mi_page_use_delayed_free(page, MI_NEVER_DELAYED_FREE, false); + + // stats + const size_t bsize = mi_page_block_size(page); + if (bsize > MI_MEDIUM_OBJ_SIZE_MAX) { + if (bsize <= MI_LARGE_OBJ_SIZE_MAX) { + mi_heap_stat_decrease(heap, large, bsize); + } + else { + mi_heap_stat_decrease(heap, huge, bsize); + } + } +#if (MI_STAT) + _mi_page_free_collect(page, false); // update used count + const size_t inuse = page->used; + if (bsize <= MI_LARGE_OBJ_SIZE_MAX) { + mi_heap_stat_decrease(heap, normal, bsize * inuse); +#if (MI_STAT>1) + mi_heap_stat_decrease(heap, normal_bins[_mi_bin(bsize)], inuse); +#endif + } + mi_heap_stat_decrease(heap, malloc, bsize * inuse); // todo: off for aligned blocks... +#endif + + /// pretend it is all free now + mi_assert_internal(mi_page_thread_free(page) == NULL); + page->used = 0; + + // and free the page + // mi_page_free(page,false); + page->next = NULL; + page->prev = NULL; + _mi_segment_page_free(page,false /* no force? */, &heap->tld->segments); + + return true; // keep going +} + +void _mi_heap_destroy_pages(mi_heap_t* heap) { + mi_heap_visit_pages(heap, &_mi_heap_page_destroy, NULL, NULL); + mi_heap_reset_pages(heap); +} + +#if MI_TRACK_HEAP_DESTROY +static bool mi_cdecl mi_heap_track_block_free(const mi_heap_t* heap, const mi_heap_area_t* area, void* block, size_t block_size, void* arg) { + MI_UNUSED(heap); MI_UNUSED(area); MI_UNUSED(arg); MI_UNUSED(block_size); + mi_track_free_size(block,mi_usable_size(block)); + return true; +} +#endif + +void mi_heap_destroy(mi_heap_t* heap) { + mi_assert(heap != NULL); + mi_assert(mi_heap_is_initialized(heap)); + mi_assert(heap->no_reclaim); + mi_assert_expensive(mi_heap_is_valid(heap)); + if (heap==NULL || !mi_heap_is_initialized(heap)) return; + if (!heap->no_reclaim) { + // don't free in case it may contain reclaimed pages + mi_heap_delete(heap); + } + else { + // track all blocks as freed + #if MI_TRACK_HEAP_DESTROY + mi_heap_visit_blocks(heap, true, mi_heap_track_block_free, NULL); + #endif + // free all pages + _mi_heap_destroy_pages(heap); + mi_heap_free(heap); + } +} + +// forcefully destroy all heaps in the current thread +void _mi_heap_unsafe_destroy_all(void) { + mi_heap_t* bheap = mi_heap_get_backing(); + mi_heap_t* curr = bheap->tld->heaps; + while (curr != NULL) { + mi_heap_t* next = curr->next; + if (curr->no_reclaim) { + mi_heap_destroy(curr); + } + else { + _mi_heap_destroy_pages(curr); + } + curr = next; + } +} + +/* ----------------------------------------------------------- + Safe Heap delete +----------------------------------------------------------- */ + +// Transfer the pages from one heap to the other +static void mi_heap_absorb(mi_heap_t* heap, mi_heap_t* from) { + mi_assert_internal(heap!=NULL); + if (from==NULL || from->page_count == 0) return; + + // reduce the size of the delayed frees + _mi_heap_delayed_free_partial(from); + + // transfer all pages by appending the queues; this will set a new heap field + // so threads may do delayed frees in either heap for a while. + // note: appending waits for each page to not be in the `MI_DELAYED_FREEING` state + // so after this only the new heap will get delayed frees + for (size_t i = 0; i <= MI_BIN_FULL; i++) { + mi_page_queue_t* pq = &heap->pages[i]; + mi_page_queue_t* append = &from->pages[i]; + size_t pcount = _mi_page_queue_append(heap, pq, append); + heap->page_count += pcount; + from->page_count -= pcount; + } + mi_assert_internal(from->page_count == 0); + + // and do outstanding delayed frees in the `from` heap + // note: be careful here as the `heap` field in all those pages no longer point to `from`, + // turns out to be ok as `_mi_heap_delayed_free` only visits the list and calls a + // the regular `_mi_free_delayed_block` which is safe. + _mi_heap_delayed_free_all(from); + #if !defined(_MSC_VER) || (_MSC_VER > 1900) // somehow the following line gives an error in VS2015, issue #353 + mi_assert_internal(mi_atomic_load_ptr_relaxed(mi_block_t,&from->thread_delayed_free) == NULL); + #endif + + // and reset the `from` heap + mi_heap_reset_pages(from); +} + +// Safe delete a heap without freeing any still allocated blocks in that heap. +void mi_heap_delete(mi_heap_t* heap) +{ + mi_assert(heap != NULL); + mi_assert(mi_heap_is_initialized(heap)); + mi_assert_expensive(mi_heap_is_valid(heap)); + if (heap==NULL || !mi_heap_is_initialized(heap)) return; + + if (!mi_heap_is_backing(heap)) { + // tranfer still used pages to the backing heap + mi_heap_absorb(heap->tld->heap_backing, heap); + } + else { + // the backing heap abandons its pages + _mi_heap_collect_abandon(heap); + } + mi_assert_internal(heap->page_count==0); + mi_heap_free(heap); +} + +mi_heap_t* mi_heap_set_default(mi_heap_t* heap) { + mi_assert(heap != NULL); + mi_assert(mi_heap_is_initialized(heap)); + if (heap==NULL || !mi_heap_is_initialized(heap)) return NULL; + mi_assert_expensive(mi_heap_is_valid(heap)); + mi_heap_t* old = mi_prim_get_default_heap(); + _mi_heap_set_default_direct(heap); + return old; +} + + + + +/* ----------------------------------------------------------- + Analysis +----------------------------------------------------------- */ + +// static since it is not thread safe to access heaps from other threads. +static mi_heap_t* mi_heap_of_block(const void* p) { + if (p == NULL) return NULL; + mi_segment_t* segment = _mi_ptr_segment(p); + bool valid = (_mi_ptr_cookie(segment) == segment->cookie); + mi_assert_internal(valid); + if mi_unlikely(!valid) return NULL; + return mi_page_heap(_mi_segment_page_of(segment,p)); +} + +bool mi_heap_contains_block(mi_heap_t* heap, const void* p) { + mi_assert(heap != NULL); + if (heap==NULL || !mi_heap_is_initialized(heap)) return false; + return (heap == mi_heap_of_block(p)); +} + + +static bool mi_heap_page_check_owned(mi_heap_t* heap, mi_page_queue_t* pq, mi_page_t* page, void* p, void* vfound) { + MI_UNUSED(heap); + MI_UNUSED(pq); + bool* found = (bool*)vfound; + mi_segment_t* segment = _mi_page_segment(page); + void* start = _mi_page_start(segment, page, NULL); + void* end = (uint8_t*)start + (page->capacity * mi_page_block_size(page)); + *found = (p >= start && p < end); + return (!*found); // continue if not found +} + +bool mi_heap_check_owned(mi_heap_t* heap, const void* p) { + mi_assert(heap != NULL); + if (heap==NULL || !mi_heap_is_initialized(heap)) return false; + if (((uintptr_t)p & (MI_INTPTR_SIZE - 1)) != 0) return false; // only aligned pointers + bool found = false; + mi_heap_visit_pages(heap, &mi_heap_page_check_owned, (void*)p, &found); + return found; +} + +bool mi_check_owned(const void* p) { + return mi_heap_check_owned(mi_prim_get_default_heap(), p); +} + +/* ----------------------------------------------------------- + Visit all heap blocks and areas + Todo: enable visiting abandoned pages, and + enable visiting all blocks of all heaps across threads +----------------------------------------------------------- */ + +// Separate struct to keep `mi_page_t` out of the public interface +typedef struct mi_heap_area_ex_s { + mi_heap_area_t area; + mi_page_t* page; +} mi_heap_area_ex_t; + +static bool mi_heap_area_visit_blocks(const mi_heap_area_ex_t* xarea, mi_block_visit_fun* visitor, void* arg) { + mi_assert(xarea != NULL); + if (xarea==NULL) return true; + const mi_heap_area_t* area = &xarea->area; + mi_page_t* page = xarea->page; + mi_assert(page != NULL); + if (page == NULL) return true; + + _mi_page_free_collect(page,true); + mi_assert_internal(page->local_free == NULL); + if (page->used == 0) return true; + + const size_t bsize = mi_page_block_size(page); + const size_t ubsize = mi_page_usable_block_size(page); // without padding + size_t psize; + uint8_t* pstart = _mi_page_start(_mi_page_segment(page), page, &psize); + + if (page->capacity == 1) { + // optimize page with one block + mi_assert_internal(page->used == 1 && page->free == NULL); + return visitor(mi_page_heap(page), area, pstart, ubsize, arg); + } + + // create a bitmap of free blocks. + #define MI_MAX_BLOCKS (MI_SMALL_PAGE_SIZE / sizeof(void*)) + uintptr_t free_map[MI_MAX_BLOCKS / sizeof(uintptr_t)]; + memset(free_map, 0, sizeof(free_map)); + + #if MI_DEBUG>1 + size_t free_count = 0; + #endif + for (mi_block_t* block = page->free; block != NULL; block = mi_block_next(page,block)) { + #if MI_DEBUG>1 + free_count++; + #endif + mi_assert_internal((uint8_t*)block >= pstart && (uint8_t*)block < (pstart + psize)); + size_t offset = (uint8_t*)block - pstart; + mi_assert_internal(offset % bsize == 0); + size_t blockidx = offset / bsize; // Todo: avoid division? + mi_assert_internal( blockidx < MI_MAX_BLOCKS); + size_t bitidx = (blockidx / sizeof(uintptr_t)); + size_t bit = blockidx - (bitidx * sizeof(uintptr_t)); + free_map[bitidx] |= ((uintptr_t)1 << bit); + } + mi_assert_internal(page->capacity == (free_count + page->used)); + + // walk through all blocks skipping the free ones + #if MI_DEBUG>1 + size_t used_count = 0; + #endif + for (size_t i = 0; i < page->capacity; i++) { + size_t bitidx = (i / sizeof(uintptr_t)); + size_t bit = i - (bitidx * sizeof(uintptr_t)); + uintptr_t m = free_map[bitidx]; + if (bit == 0 && m == UINTPTR_MAX) { + i += (sizeof(uintptr_t) - 1); // skip a run of free blocks + } + else if ((m & ((uintptr_t)1 << bit)) == 0) { + #if MI_DEBUG>1 + used_count++; + #endif + uint8_t* block = pstart + (i * bsize); + if (!visitor(mi_page_heap(page), area, block, ubsize, arg)) return false; + } + } + mi_assert_internal(page->used == used_count); + return true; +} + +typedef bool (mi_heap_area_visit_fun)(const mi_heap_t* heap, const mi_heap_area_ex_t* area, void* arg); + + +static bool mi_heap_visit_areas_page(mi_heap_t* heap, mi_page_queue_t* pq, mi_page_t* page, void* vfun, void* arg) { + MI_UNUSED(heap); + MI_UNUSED(pq); + mi_heap_area_visit_fun* fun = (mi_heap_area_visit_fun*)vfun; + mi_heap_area_ex_t xarea; + const size_t bsize = mi_page_block_size(page); + const size_t ubsize = mi_page_usable_block_size(page); + xarea.page = page; + xarea.area.reserved = page->reserved * bsize; + xarea.area.committed = page->capacity * bsize; + xarea.area.blocks = _mi_page_start(_mi_page_segment(page), page, NULL); + xarea.area.used = page->used; // number of blocks in use (#553) + xarea.area.block_size = ubsize; + xarea.area.full_block_size = bsize; + return fun(heap, &xarea, arg); +} + +// Visit all heap pages as areas +static bool mi_heap_visit_areas(const mi_heap_t* heap, mi_heap_area_visit_fun* visitor, void* arg) { + if (visitor == NULL) return false; + return mi_heap_visit_pages((mi_heap_t*)heap, &mi_heap_visit_areas_page, (void*)(visitor), arg); // note: function pointer to void* :-{ +} + +// Just to pass arguments +typedef struct mi_visit_blocks_args_s { + bool visit_blocks; + mi_block_visit_fun* visitor; + void* arg; +} mi_visit_blocks_args_t; + +static bool mi_heap_area_visitor(const mi_heap_t* heap, const mi_heap_area_ex_t* xarea, void* arg) { + mi_visit_blocks_args_t* args = (mi_visit_blocks_args_t*)arg; + if (!args->visitor(heap, &xarea->area, NULL, xarea->area.block_size, args->arg)) return false; + if (args->visit_blocks) { + return mi_heap_area_visit_blocks(xarea, args->visitor, args->arg); + } + else { + return true; + } +} + +// Visit all blocks in a heap +bool mi_heap_visit_blocks(const mi_heap_t* heap, bool visit_blocks, mi_block_visit_fun* visitor, void* arg) { + mi_visit_blocks_args_t args = { visit_blocks, visitor, arg }; + return mi_heap_visit_areas(heap, &mi_heap_area_visitor, &args); +} diff --git a/compat/mimalloc/init.c b/compat/mimalloc/init.c new file mode 100644 index 00000000000000..4670d5510db187 --- /dev/null +++ b/compat/mimalloc/init.c @@ -0,0 +1,709 @@ +/* ---------------------------------------------------------------------------- +Copyright (c) 2018-2022, Microsoft Research, Daan Leijen +This is free software; you can redistribute it and/or modify it under the +terms of the MIT license. A copy of the license can be found in the file +"LICENSE" at the root of this distribution. +-----------------------------------------------------------------------------*/ +#include "mimalloc.h" +#include "mimalloc/internal.h" +#include "mimalloc/prim.h" + +#include // memcpy, memset +#include // atexit + + +// Empty page used to initialize the small free pages array +const mi_page_t _mi_page_empty = { + 0, false, false, false, + 0, // capacity + 0, // reserved capacity + { 0 }, // flags + false, // is_zero + 0, // retire_expire + NULL, // free + 0, // used + 0, // xblock_size + NULL, // local_free + #if (MI_PADDING || MI_ENCODE_FREELIST) + { 0, 0 }, + #endif + MI_ATOMIC_VAR_INIT(0), // xthread_free + MI_ATOMIC_VAR_INIT(0), // xheap + NULL, NULL + #if MI_INTPTR_SIZE==8 + , { 0 } // padding + #endif +}; + +#define MI_PAGE_EMPTY() ((mi_page_t*)&_mi_page_empty) + +#if (MI_SMALL_WSIZE_MAX==128) +#if (MI_PADDING>0) && (MI_INTPTR_SIZE >= 8) +#define MI_SMALL_PAGES_EMPTY { MI_INIT128(MI_PAGE_EMPTY), MI_PAGE_EMPTY(), MI_PAGE_EMPTY() } +#elif (MI_PADDING>0) +#define MI_SMALL_PAGES_EMPTY { MI_INIT128(MI_PAGE_EMPTY), MI_PAGE_EMPTY(), MI_PAGE_EMPTY(), MI_PAGE_EMPTY() } +#else +#define MI_SMALL_PAGES_EMPTY { MI_INIT128(MI_PAGE_EMPTY), MI_PAGE_EMPTY() } +#endif +#else +#error "define right initialization sizes corresponding to MI_SMALL_WSIZE_MAX" +#endif + +// Empty page queues for every bin +#define QNULL(sz) { NULL, NULL, (sz)*sizeof(uintptr_t) } +#define MI_PAGE_QUEUES_EMPTY \ + { QNULL(1), \ + QNULL( 1), QNULL( 2), QNULL( 3), QNULL( 4), QNULL( 5), QNULL( 6), QNULL( 7), QNULL( 8), /* 8 */ \ + QNULL( 10), QNULL( 12), QNULL( 14), QNULL( 16), QNULL( 20), QNULL( 24), QNULL( 28), QNULL( 32), /* 16 */ \ + QNULL( 40), QNULL( 48), QNULL( 56), QNULL( 64), QNULL( 80), QNULL( 96), QNULL( 112), QNULL( 128), /* 24 */ \ + QNULL( 160), QNULL( 192), QNULL( 224), QNULL( 256), QNULL( 320), QNULL( 384), QNULL( 448), QNULL( 512), /* 32 */ \ + QNULL( 640), QNULL( 768), QNULL( 896), QNULL( 1024), QNULL( 1280), QNULL( 1536), QNULL( 1792), QNULL( 2048), /* 40 */ \ + QNULL( 2560), QNULL( 3072), QNULL( 3584), QNULL( 4096), QNULL( 5120), QNULL( 6144), QNULL( 7168), QNULL( 8192), /* 48 */ \ + QNULL( 10240), QNULL( 12288), QNULL( 14336), QNULL( 16384), QNULL( 20480), QNULL( 24576), QNULL( 28672), QNULL( 32768), /* 56 */ \ + QNULL( 40960), QNULL( 49152), QNULL( 57344), QNULL( 65536), QNULL( 81920), QNULL( 98304), QNULL(114688), QNULL(131072), /* 64 */ \ + QNULL(163840), QNULL(196608), QNULL(229376), QNULL(262144), QNULL(327680), QNULL(393216), QNULL(458752), QNULL(524288), /* 72 */ \ + QNULL(MI_MEDIUM_OBJ_WSIZE_MAX + 1 /* 655360, Huge queue */), \ + QNULL(MI_MEDIUM_OBJ_WSIZE_MAX + 2) /* Full queue */ } + +#define MI_STAT_COUNT_NULL() {0,0,0,0} + +// Empty statistics +#if MI_STAT>1 +#define MI_STAT_COUNT_END_NULL() , { MI_STAT_COUNT_NULL(), MI_INIT32(MI_STAT_COUNT_NULL) } +#else +#define MI_STAT_COUNT_END_NULL() +#endif + +#define MI_STATS_NULL \ + MI_STAT_COUNT_NULL(), MI_STAT_COUNT_NULL(), \ + MI_STAT_COUNT_NULL(), MI_STAT_COUNT_NULL(), \ + MI_STAT_COUNT_NULL(), MI_STAT_COUNT_NULL(), \ + MI_STAT_COUNT_NULL(), MI_STAT_COUNT_NULL(), \ + MI_STAT_COUNT_NULL(), MI_STAT_COUNT_NULL(), \ + MI_STAT_COUNT_NULL(), MI_STAT_COUNT_NULL(), \ + MI_STAT_COUNT_NULL(), MI_STAT_COUNT_NULL(), \ + MI_STAT_COUNT_NULL(), \ + { 0, 0 }, { 0, 0 }, { 0, 0 }, { 0, 0 }, \ + { 0, 0 }, { 0, 0 }, { 0, 0 }, { 0, 0 }, { 0, 0 }, { 0, 0 } \ + MI_STAT_COUNT_END_NULL() + + +// Empty slice span queues for every bin +#define SQNULL(sz) { NULL, NULL, sz } +#define MI_SEGMENT_SPAN_QUEUES_EMPTY \ + { SQNULL(1), \ + SQNULL( 1), SQNULL( 2), SQNULL( 3), SQNULL( 4), SQNULL( 5), SQNULL( 6), SQNULL( 7), SQNULL( 10), /* 8 */ \ + SQNULL( 12), SQNULL( 14), SQNULL( 16), SQNULL( 20), SQNULL( 24), SQNULL( 28), SQNULL( 32), SQNULL( 40), /* 16 */ \ + SQNULL( 48), SQNULL( 56), SQNULL( 64), SQNULL( 80), SQNULL( 96), SQNULL( 112), SQNULL( 128), SQNULL( 160), /* 24 */ \ + SQNULL( 192), SQNULL( 224), SQNULL( 256), SQNULL( 320), SQNULL( 384), SQNULL( 448), SQNULL( 512), SQNULL( 640), /* 32 */ \ + SQNULL( 768), SQNULL( 896), SQNULL( 1024) /* 35 */ } + + +// -------------------------------------------------------- +// Statically allocate an empty heap as the initial +// thread local value for the default heap, +// and statically allocate the backing heap for the main +// thread so it can function without doing any allocation +// itself (as accessing a thread local for the first time +// may lead to allocation itself on some platforms) +// -------------------------------------------------------- + +mi_decl_cache_align const mi_heap_t _mi_heap_empty = { + NULL, + MI_SMALL_PAGES_EMPTY, + MI_PAGE_QUEUES_EMPTY, + MI_ATOMIC_VAR_INIT(NULL), + 0, // tid + 0, // cookie + 0, // arena id + { 0, 0 }, // keys + { {0}, {0}, 0, true }, // random + 0, // page count + MI_BIN_FULL, 0, // page retired min/max + NULL, // next + false +}; + +#define tld_empty_stats ((mi_stats_t*)((uint8_t*)&tld_empty + offsetof(mi_tld_t,stats))) +#define tld_empty_os ((mi_os_tld_t*)((uint8_t*)&tld_empty + offsetof(mi_tld_t,os))) + +mi_decl_cache_align static const mi_tld_t tld_empty = { + 0, + false, + NULL, NULL, + { MI_SEGMENT_SPAN_QUEUES_EMPTY, 0, 0, 0, 0, tld_empty_stats, tld_empty_os }, // segments + { 0, tld_empty_stats }, // os + { MI_STATS_NULL } // stats +}; + +mi_threadid_t _mi_thread_id(void) mi_attr_noexcept { + return _mi_prim_thread_id(); +} + +// the thread-local default heap for allocation +mi_decl_thread mi_heap_t* _mi_heap_default = (mi_heap_t*)&_mi_heap_empty; + +extern mi_heap_t _mi_heap_main; + +static mi_tld_t tld_main = { + 0, false, + &_mi_heap_main, & _mi_heap_main, + { MI_SEGMENT_SPAN_QUEUES_EMPTY, 0, 0, 0, 0, &tld_main.stats, &tld_main.os }, // segments + { 0, &tld_main.stats }, // os + { MI_STATS_NULL } // stats +}; + +mi_heap_t _mi_heap_main = { + &tld_main, + MI_SMALL_PAGES_EMPTY, + MI_PAGE_QUEUES_EMPTY, + MI_ATOMIC_VAR_INIT(NULL), + 0, // thread id + 0, // initial cookie + 0, // arena id + { 0, 0 }, // the key of the main heap can be fixed (unlike page keys that need to be secure!) + { {0x846ca68b}, {0}, 0, true }, // random + 0, // page count + MI_BIN_FULL, 0, // page retired min/max + NULL, // next heap + false // can reclaim +}; + +bool _mi_process_is_initialized = false; // set to `true` in `mi_process_init`. + +mi_stats_t _mi_stats_main = { MI_STATS_NULL }; + + +static void mi_heap_main_init(void) { + if (_mi_heap_main.cookie == 0) { + _mi_heap_main.thread_id = _mi_thread_id(); + _mi_heap_main.cookie = 1; + #if defined(_WIN32) && !defined(MI_SHARED_LIB) + _mi_random_init_weak(&_mi_heap_main.random); // prevent allocation failure during bcrypt dll initialization with static linking + #else + _mi_random_init(&_mi_heap_main.random); + #endif + _mi_heap_main.cookie = _mi_heap_random_next(&_mi_heap_main); + _mi_heap_main.keys[0] = _mi_heap_random_next(&_mi_heap_main); + _mi_heap_main.keys[1] = _mi_heap_random_next(&_mi_heap_main); + } +} + +mi_heap_t* _mi_heap_main_get(void) { + mi_heap_main_init(); + return &_mi_heap_main; +} + + +/* ----------------------------------------------------------- + Initialization and freeing of the thread local heaps +----------------------------------------------------------- */ + +// note: in x64 in release build `sizeof(mi_thread_data_t)` is under 4KiB (= OS page size). +typedef struct mi_thread_data_s { + mi_heap_t heap; // must come first due to cast in `_mi_heap_done` + mi_tld_t tld; + mi_memid_t memid; +} mi_thread_data_t; + + +// Thread meta-data is allocated directly from the OS. For +// some programs that do not use thread pools and allocate and +// destroy many OS threads, this may causes too much overhead +// per thread so we maintain a small cache of recently freed metadata. + +#define TD_CACHE_SIZE (16) +static _Atomic(mi_thread_data_t*) td_cache[TD_CACHE_SIZE]; + +static mi_thread_data_t* mi_thread_data_zalloc(void) { + // try to find thread metadata in the cache + bool is_zero = false; + mi_thread_data_t* td = NULL; + for (int i = 0; i < TD_CACHE_SIZE; i++) { + td = mi_atomic_load_ptr_relaxed(mi_thread_data_t, &td_cache[i]); + if (td != NULL) { + // found cached allocation, try use it + td = mi_atomic_exchange_ptr_acq_rel(mi_thread_data_t, &td_cache[i], NULL); + if (td != NULL) { + break; + } + } + } + + // if that fails, allocate as meta data + if (td == NULL) { + mi_memid_t memid; + td = (mi_thread_data_t*)_mi_os_alloc(sizeof(mi_thread_data_t), &memid, &_mi_stats_main); + if (td == NULL) { + // if this fails, try once more. (issue #257) + td = (mi_thread_data_t*)_mi_os_alloc(sizeof(mi_thread_data_t), &memid, &_mi_stats_main); + if (td == NULL) { + // really out of memory + _mi_error_message(ENOMEM, "unable to allocate thread local heap metadata (%zu bytes)\n", sizeof(mi_thread_data_t)); + } + } + if (td != NULL) { + td->memid = memid; + is_zero = memid.initially_zero; + } + } + + if (td != NULL && !is_zero) { + _mi_memzero_aligned(td, sizeof(*td)); + } + return td; +} + +static void mi_thread_data_free( mi_thread_data_t* tdfree ) { + // try to add the thread metadata to the cache + for (int i = 0; i < TD_CACHE_SIZE; i++) { + mi_thread_data_t* td = mi_atomic_load_ptr_relaxed(mi_thread_data_t, &td_cache[i]); + if (td == NULL) { + mi_thread_data_t* expected = NULL; + if (mi_atomic_cas_ptr_weak_acq_rel(mi_thread_data_t, &td_cache[i], &expected, tdfree)) { + return; + } + } + } + // if that fails, just free it directly + _mi_os_free(tdfree, sizeof(mi_thread_data_t), tdfree->memid, &_mi_stats_main); +} + +void _mi_thread_data_collect(void) { + // free all thread metadata from the cache + for (int i = 0; i < TD_CACHE_SIZE; i++) { + mi_thread_data_t* td = mi_atomic_load_ptr_relaxed(mi_thread_data_t, &td_cache[i]); + if (td != NULL) { + td = mi_atomic_exchange_ptr_acq_rel(mi_thread_data_t, &td_cache[i], NULL); + if (td != NULL) { + _mi_os_free(td, sizeof(mi_thread_data_t), td->memid, &_mi_stats_main); + } + } + } +} + +// Initialize the thread local default heap, called from `mi_thread_init` +static bool _mi_heap_init(void) { + if (mi_heap_is_initialized(mi_prim_get_default_heap())) return true; + if (_mi_is_main_thread()) { + // mi_assert_internal(_mi_heap_main.thread_id != 0); // can happen on freeBSD where alloc is called before any initialization + // the main heap is statically allocated + mi_heap_main_init(); + _mi_heap_set_default_direct(&_mi_heap_main); + //mi_assert_internal(_mi_heap_default->tld->heap_backing == mi_prim_get_default_heap()); + } + else { + // use `_mi_os_alloc` to allocate directly from the OS + mi_thread_data_t* td = mi_thread_data_zalloc(); + if (td == NULL) return false; + + mi_tld_t* tld = &td->tld; + mi_heap_t* heap = &td->heap; + _mi_memcpy_aligned(tld, &tld_empty, sizeof(*tld)); + _mi_memcpy_aligned(heap, &_mi_heap_empty, sizeof(*heap)); + heap->thread_id = _mi_thread_id(); + _mi_random_init(&heap->random); + heap->cookie = _mi_heap_random_next(heap) | 1; + heap->keys[0] = _mi_heap_random_next(heap); + heap->keys[1] = _mi_heap_random_next(heap); + heap->tld = tld; + tld->heap_backing = heap; + tld->heaps = heap; + tld->segments.stats = &tld->stats; + tld->segments.os = &tld->os; + tld->os.stats = &tld->stats; + _mi_heap_set_default_direct(heap); + } + return false; +} + +// Free the thread local default heap (called from `mi_thread_done`) +static bool _mi_heap_done(mi_heap_t* heap) { + if (!mi_heap_is_initialized(heap)) return true; + + // reset default heap + _mi_heap_set_default_direct(_mi_is_main_thread() ? &_mi_heap_main : (mi_heap_t*)&_mi_heap_empty); + + // switch to backing heap + heap = heap->tld->heap_backing; + if (!mi_heap_is_initialized(heap)) return false; + + // delete all non-backing heaps in this thread + mi_heap_t* curr = heap->tld->heaps; + while (curr != NULL) { + mi_heap_t* next = curr->next; // save `next` as `curr` will be freed + if (curr != heap) { + mi_assert_internal(!mi_heap_is_backing(curr)); + mi_heap_delete(curr); + } + curr = next; + } + mi_assert_internal(heap->tld->heaps == heap && heap->next == NULL); + mi_assert_internal(mi_heap_is_backing(heap)); + + // collect if not the main thread + if (heap != &_mi_heap_main) { + _mi_heap_collect_abandon(heap); + } + + // merge stats + _mi_stats_done(&heap->tld->stats); + + // free if not the main thread + if (heap != &_mi_heap_main) { + // the following assertion does not always hold for huge segments as those are always treated + // as abondened: one may allocate it in one thread, but deallocate in another in which case + // the count can be too large or negative. todo: perhaps not count huge segments? see issue #363 + // mi_assert_internal(heap->tld->segments.count == 0 || heap->thread_id != _mi_thread_id()); + mi_thread_data_free((mi_thread_data_t*)heap); + } + else { + #if 0 + // never free the main thread even in debug mode; if a dll is linked statically with mimalloc, + // there may still be delete/free calls after the mi_fls_done is called. Issue #207 + _mi_heap_destroy_pages(heap); + mi_assert_internal(heap->tld->heap_backing == &_mi_heap_main); + #endif + } + return false; +} + + + +// -------------------------------------------------------- +// Try to run `mi_thread_done()` automatically so any memory +// owned by the thread but not yet released can be abandoned +// and re-owned by another thread. +// +// 1. windows dynamic library: +// call from DllMain on DLL_THREAD_DETACH +// 2. windows static library: +// use `FlsAlloc` to call a destructor when the thread is done +// 3. unix, pthreads: +// use a pthread key to call a destructor when a pthread is done +// +// In the last two cases we also need to call `mi_process_init` +// to set up the thread local keys. +// -------------------------------------------------------- + +// Set up handlers so `mi_thread_done` is called automatically +static void mi_process_setup_auto_thread_done(void) { + static bool tls_initialized = false; // fine if it races + if (tls_initialized) return; + tls_initialized = true; + _mi_prim_thread_init_auto_done(); + _mi_heap_set_default_direct(&_mi_heap_main); +} + + +bool _mi_is_main_thread(void) { + return (_mi_heap_main.thread_id==0 || _mi_heap_main.thread_id == _mi_thread_id()); +} + +static _Atomic(size_t) thread_count = MI_ATOMIC_VAR_INIT(1); + +size_t _mi_current_thread_count(void) { + return mi_atomic_load_relaxed(&thread_count); +} + +// This is called from the `mi_malloc_generic` +void mi_thread_init(void) mi_attr_noexcept +{ + // ensure our process has started already + mi_process_init(); + + // initialize the thread local default heap + // (this will call `_mi_heap_set_default_direct` and thus set the + // fiber/pthread key to a non-zero value, ensuring `_mi_thread_done` is called) + if (_mi_heap_init()) return; // returns true if already initialized + + _mi_stat_increase(&_mi_stats_main.threads, 1); + mi_atomic_increment_relaxed(&thread_count); + //_mi_verbose_message("thread init: 0x%zx\n", _mi_thread_id()); +} + +void mi_thread_done(void) mi_attr_noexcept { + _mi_thread_done(NULL); +} + +void _mi_thread_done(mi_heap_t* heap) +{ + // calling with NULL implies using the default heap + if (heap == NULL) { + heap = mi_prim_get_default_heap(); + if (heap == NULL) return; + } + + // prevent re-entrancy through heap_done/heap_set_default_direct (issue #699) + if (!mi_heap_is_initialized(heap)) { + return; + } + + // adjust stats + mi_atomic_decrement_relaxed(&thread_count); + _mi_stat_decrease(&_mi_stats_main.threads, 1); + + // check thread-id as on Windows shutdown with FLS the main (exit) thread may call this on thread-local heaps... + if (heap->thread_id != _mi_thread_id()) return; + + // abandon the thread local heap + if (_mi_heap_done(heap)) return; // returns true if already ran +} + +void _mi_heap_set_default_direct(mi_heap_t* heap) { + mi_assert_internal(heap != NULL); + #if defined(MI_TLS_SLOT) + mi_prim_tls_slot_set(MI_TLS_SLOT,heap); + #elif defined(MI_TLS_PTHREAD_SLOT_OFS) + *mi_tls_pthread_heap_slot() = heap; + #elif defined(MI_TLS_PTHREAD) + // we use _mi_heap_default_key + #else + _mi_heap_default = heap; + #endif + + // ensure the default heap is passed to `_mi_thread_done` + // setting to a non-NULL value also ensures `mi_thread_done` is called. + _mi_prim_thread_associate_default_heap(heap); +} + + +// -------------------------------------------------------- +// Run functions on process init/done, and thread init/done +// -------------------------------------------------------- +static void mi_cdecl mi_process_done(void); + +static bool os_preloading = true; // true until this module is initialized +static bool mi_redirected = false; // true if malloc redirects to mi_malloc + +// Returns true if this module has not been initialized; Don't use C runtime routines until it returns false. +bool mi_decl_noinline _mi_preloading(void) { + return os_preloading; +} + +mi_decl_nodiscard bool mi_is_redirected(void) mi_attr_noexcept { + return mi_redirected; +} + +// Communicate with the redirection module on Windows +#if defined(_WIN32) && defined(MI_SHARED_LIB) && !defined(MI_WIN_NOREDIRECT) +#ifdef __cplusplus +extern "C" { +#endif +mi_decl_export void _mi_redirect_entry(DWORD reason) { + // called on redirection; careful as this may be called before DllMain + if (reason == DLL_PROCESS_ATTACH) { + mi_redirected = true; + } + else if (reason == DLL_PROCESS_DETACH) { + mi_redirected = false; + } + else if (reason == DLL_THREAD_DETACH) { + mi_thread_done(); + } +} +__declspec(dllimport) bool mi_cdecl mi_allocator_init(const char** message); +__declspec(dllimport) void mi_cdecl mi_allocator_done(void); +#ifdef __cplusplus +} +#endif +#else +static bool mi_allocator_init(const char** message) { + if (message != NULL) *message = NULL; + return true; +} +static void mi_allocator_done(void) { + // nothing to do +} +#endif + +// Called once by the process loader +static void mi_process_load(void) { + mi_heap_main_init(); + #if defined(__APPLE__) || defined(MI_TLS_RECURSE_GUARD) + volatile mi_heap_t* dummy = _mi_heap_default; // access TLS to allocate it before setting tls_initialized to true; + if (dummy == NULL) return; // use dummy or otherwise the access may get optimized away (issue #697) + #endif + os_preloading = false; + mi_assert_internal(_mi_is_main_thread()); + #if !(defined(_WIN32) && defined(MI_SHARED_LIB)) // use Dll process detach (see below) instead of atexit (issue #521) + atexit(&mi_process_done); + #endif + _mi_options_init(); + mi_process_setup_auto_thread_done(); + mi_process_init(); + if (mi_redirected) _mi_verbose_message("malloc is redirected.\n"); + + // show message from the redirector (if present) + const char* msg = NULL; + mi_allocator_init(&msg); + if (msg != NULL && (mi_option_is_enabled(mi_option_verbose) || mi_option_is_enabled(mi_option_show_errors))) { + _mi_fputs(NULL,NULL,NULL,msg); + } + + // reseed random + _mi_random_reinit_if_weak(&_mi_heap_main.random); +} + +#if defined(_WIN32) && (defined(_M_IX86) || defined(_M_X64)) +#include +mi_decl_cache_align bool _mi_cpu_has_fsrm = false; + +static void mi_detect_cpu_features(void) { + // FSRM for fast rep movsb support (AMD Zen3+ (~2020) or Intel Ice Lake+ (~2017)) + int32_t cpu_info[4]; + __cpuid(cpu_info, 7); + _mi_cpu_has_fsrm = ((cpu_info[3] & (1 << 4)) != 0); // bit 4 of EDX : see +} +#else +static void mi_detect_cpu_features(void) { + // nothing +} +#endif + +// Initialize the process; called by thread_init or the process loader +void mi_process_init(void) mi_attr_noexcept { + // ensure we are called once + static mi_atomic_once_t process_init; + #if _MSC_VER < 1920 + mi_heap_main_init(); // vs2017 can dynamically re-initialize _mi_heap_main + #endif + if (!mi_atomic_once(&process_init)) return; + _mi_process_is_initialized = true; + _mi_verbose_message("process init: 0x%zx\n", _mi_thread_id()); + mi_process_setup_auto_thread_done(); + + mi_detect_cpu_features(); + _mi_os_init(); + mi_heap_main_init(); + #if MI_DEBUG + _mi_verbose_message("debug level : %d\n", MI_DEBUG); + #endif + _mi_verbose_message("secure level: %d\n", MI_SECURE); + _mi_verbose_message("mem tracking: %s\n", MI_TRACK_TOOL); + #if MI_TSAN + _mi_verbose_message("thread santizer enabled\n"); + #endif + mi_thread_init(); + + #if defined(_WIN32) + // On windows, when building as a static lib the FLS cleanup happens to early for the main thread. + // To avoid this, set the FLS value for the main thread to NULL so the fls cleanup + // will not call _mi_thread_done on the (still executing) main thread. See issue #508. + _mi_prim_thread_associate_default_heap(NULL); + #endif + + mi_stats_reset(); // only call stat reset *after* thread init (or the heap tld == NULL) + mi_track_init(); + + if (mi_option_is_enabled(mi_option_reserve_huge_os_pages)) { + size_t pages = mi_option_get_clamp(mi_option_reserve_huge_os_pages, 0, 128*1024); + long reserve_at = mi_option_get(mi_option_reserve_huge_os_pages_at); + if (reserve_at != -1) { + mi_reserve_huge_os_pages_at(pages, reserve_at, pages*500); + } else { + mi_reserve_huge_os_pages_interleave(pages, 0, pages*500); + } + } + if (mi_option_is_enabled(mi_option_reserve_os_memory)) { + long ksize = mi_option_get(mi_option_reserve_os_memory); + if (ksize > 0) { + mi_reserve_os_memory((size_t)ksize*MI_KiB, true /* commit? */, true /* allow large pages? */); + } + } +} + +// Called when the process is done (through `at_exit`) +static void mi_cdecl mi_process_done(void) { + // only shutdown if we were initialized + if (!_mi_process_is_initialized) return; + // ensure we are called once + static bool process_done = false; + if (process_done) return; + process_done = true; + + // release any thread specific resources and ensure _mi_thread_done is called on all but the main thread + _mi_prim_thread_done_auto_done(); + + #ifndef MI_SKIP_COLLECT_ON_EXIT + #if (MI_DEBUG || !defined(MI_SHARED_LIB)) + // free all memory if possible on process exit. This is not needed for a stand-alone process + // but should be done if mimalloc is statically linked into another shared library which + // is repeatedly loaded/unloaded, see issue #281. + mi_collect(true /* force */ ); + #endif + #endif + + // Forcefully release all retained memory; this can be dangerous in general if overriding regular malloc/free + // since after process_done there might still be other code running that calls `free` (like at_exit routines, + // or C-runtime termination code. + if (mi_option_is_enabled(mi_option_destroy_on_exit)) { + mi_collect(true /* force */); + _mi_heap_unsafe_destroy_all(); // forcefully release all memory held by all heaps (of this thread only!) + _mi_arena_unsafe_destroy_all(& _mi_heap_main_get()->tld->stats); + } + + if (mi_option_is_enabled(mi_option_show_stats) || mi_option_is_enabled(mi_option_verbose)) { + mi_stats_print(NULL); + } + mi_allocator_done(); + _mi_verbose_message("process done: 0x%zx\n", _mi_heap_main.thread_id); + os_preloading = true; // don't call the C runtime anymore +} + + + +#if defined(_WIN32) && defined(MI_SHARED_LIB) + // Windows DLL: easy to hook into process_init and thread_done + __declspec(dllexport) BOOL WINAPI DllMain(HINSTANCE inst, DWORD reason, LPVOID reserved) { + MI_UNUSED(reserved); + MI_UNUSED(inst); + if (reason==DLL_PROCESS_ATTACH) { + mi_process_load(); + } + else if (reason==DLL_PROCESS_DETACH) { + mi_process_done(); + } + else if (reason==DLL_THREAD_DETACH) { + if (!mi_is_redirected()) { + mi_thread_done(); + } + } + return TRUE; + } + +#elif defined(_MSC_VER) + // MSVC: use data section magic for static libraries + // See + static int _mi_process_init(void) { + mi_process_load(); + return 0; + } + typedef int(*_mi_crt_callback_t)(void); + #if defined(_M_X64) || defined(_M_ARM64) + __pragma(comment(linker, "/include:" "_mi_msvc_initu")) + #pragma section(".CRT$XIU", long, read) + #else + __pragma(comment(linker, "/include:" "__mi_msvc_initu")) + #endif + #pragma data_seg(".CRT$XIU") + mi_decl_externc _mi_crt_callback_t _mi_msvc_initu[] = { &_mi_process_init }; + #pragma data_seg() + +#elif defined(__cplusplus) + // C++: use static initialization to detect process start + static bool _mi_process_init(void) { + mi_process_load(); + return (_mi_heap_main.thread_id != 0); + } + static bool mi_initialized = _mi_process_init(); + +#elif defined(__GNUC__) || defined(__clang__) + // GCC,Clang: use the constructor attribute + static void __attribute__((constructor)) _mi_process_init(void) { + mi_process_load(); + } + +#else +#pragma message("define a way to call mi_process_load on your platform") +#endif diff --git a/compat/mimalloc/mimalloc.h b/compat/mimalloc/mimalloc.h new file mode 100644 index 00000000000000..c0f5e96e51e975 --- /dev/null +++ b/compat/mimalloc/mimalloc.h @@ -0,0 +1,565 @@ +/* ---------------------------------------------------------------------------- +Copyright (c) 2018-2023, Microsoft Research, Daan Leijen +This is free software; you can redistribute it and/or modify it under the +terms of the MIT license. A copy of the license can be found in the file +"LICENSE" at the root of this distribution. +-----------------------------------------------------------------------------*/ +#pragma once +#ifndef MIMALLOC_H +#define MIMALLOC_H + +#define MI_MALLOC_VERSION 212 // major + 2 digits minor + +// ------------------------------------------------------ +// Compiler specific attributes +// ------------------------------------------------------ + +#ifdef __cplusplus + #if (__cplusplus >= 201103L) || (_MSC_VER > 1900) // C++11 + #define mi_attr_noexcept noexcept + #else + #define mi_attr_noexcept throw() + #endif +#else + #define mi_attr_noexcept +#endif + +#if defined(__cplusplus) && (__cplusplus >= 201703) + #define mi_decl_nodiscard [[nodiscard]] +#elif (defined(__GNUC__) && (__GNUC__ >= 4)) || defined(__clang__) // includes clang, icc, and clang-cl + #define mi_decl_nodiscard __attribute__((warn_unused_result)) +#elif defined(_HAS_NODISCARD) + #define mi_decl_nodiscard _NODISCARD +#elif (_MSC_VER >= 1700) + #define mi_decl_nodiscard _Check_return_ +#else + #define mi_decl_nodiscard +#endif + +#if defined(_MSC_VER) || defined(__MINGW32__) + #if !defined(MI_SHARED_LIB) + #define mi_decl_export + #elif defined(MI_SHARED_LIB_EXPORT) + #define mi_decl_export __declspec(dllexport) + #else + #define mi_decl_export __declspec(dllimport) + #endif + #if defined(__MINGW32__) + #define mi_decl_restrict + #define mi_attr_malloc __attribute__((malloc)) + #else + #if (_MSC_VER >= 1900) && !defined(__EDG__) + #define mi_decl_restrict __declspec(allocator) __declspec(restrict) + #else + #define mi_decl_restrict __declspec(restrict) + #endif + #define mi_attr_malloc + #endif + #define mi_cdecl __cdecl + #define mi_attr_alloc_size(s) + #define mi_attr_alloc_size2(s1,s2) + #define mi_attr_alloc_align(p) +#elif defined(__GNUC__) // includes clang and icc + #if defined(MI_SHARED_LIB) && defined(MI_SHARED_LIB_EXPORT) + #define mi_decl_export __attribute__((visibility("default"))) + #else + #define mi_decl_export + #endif + #define mi_cdecl // leads to warnings... __attribute__((cdecl)) + #define mi_decl_restrict + #define mi_attr_malloc __attribute__((malloc)) + #if (defined(__clang_major__) && (__clang_major__ < 4)) || (__GNUC__ < 5) + #define mi_attr_alloc_size(s) + #define mi_attr_alloc_size2(s1,s2) + #define mi_attr_alloc_align(p) + #elif defined(__INTEL_COMPILER) + #define mi_attr_alloc_size(s) __attribute__((alloc_size(s))) + #define mi_attr_alloc_size2(s1,s2) __attribute__((alloc_size(s1,s2))) + #define mi_attr_alloc_align(p) + #else + #define mi_attr_alloc_size(s) __attribute__((alloc_size(s))) + #define mi_attr_alloc_size2(s1,s2) __attribute__((alloc_size(s1,s2))) + #define mi_attr_alloc_align(p) __attribute__((alloc_align(p))) + #endif +#else + #define mi_cdecl + #define mi_decl_export + #define mi_decl_restrict + #define mi_attr_malloc + #define mi_attr_alloc_size(s) + #define mi_attr_alloc_size2(s1,s2) + #define mi_attr_alloc_align(p) +#endif + +// ------------------------------------------------------ +// Includes +// ------------------------------------------------------ + +#include // size_t +#include // bool +#include // INTPTR_MAX + +#ifdef __cplusplus +extern "C" { +#endif + +// ------------------------------------------------------ +// Standard malloc interface +// ------------------------------------------------------ + +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_malloc(size_t size) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size(1); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_calloc(size_t count, size_t size) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size2(1,2); +mi_decl_nodiscard mi_decl_export void* mi_realloc(void* p, size_t newsize) mi_attr_noexcept mi_attr_alloc_size(2); +mi_decl_export void* mi_expand(void* p, size_t newsize) mi_attr_noexcept mi_attr_alloc_size(2); + +mi_decl_export void mi_free(void* p) mi_attr_noexcept; +mi_decl_nodiscard mi_decl_export mi_decl_restrict char* mi_strdup(const char* s) mi_attr_noexcept mi_attr_malloc; +mi_decl_nodiscard mi_decl_export mi_decl_restrict char* mi_strndup(const char* s, size_t n) mi_attr_noexcept mi_attr_malloc; +mi_decl_nodiscard mi_decl_export mi_decl_restrict char* mi_realpath(const char* fname, char* resolved_name) mi_attr_noexcept mi_attr_malloc; + +// ------------------------------------------------------ +// Extended functionality +// ------------------------------------------------------ +#define MI_SMALL_WSIZE_MAX (128) +#define MI_SMALL_SIZE_MAX (MI_SMALL_WSIZE_MAX*sizeof(void*)) + +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_malloc_small(size_t size) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size(1); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_zalloc_small(size_t size) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size(1); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_zalloc(size_t size) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size(1); + +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_mallocn(size_t count, size_t size) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size2(1,2); +mi_decl_nodiscard mi_decl_export void* mi_reallocn(void* p, size_t count, size_t size) mi_attr_noexcept mi_attr_alloc_size2(2,3); +mi_decl_nodiscard mi_decl_export void* mi_reallocf(void* p, size_t newsize) mi_attr_noexcept mi_attr_alloc_size(2); + +mi_decl_nodiscard mi_decl_export size_t mi_usable_size(const void* p) mi_attr_noexcept; +mi_decl_nodiscard mi_decl_export size_t mi_good_size(size_t size) mi_attr_noexcept; + + +// ------------------------------------------------------ +// Internals +// ------------------------------------------------------ + +typedef void (mi_cdecl mi_deferred_free_fun)(bool force, unsigned long long heartbeat, void* arg); +mi_decl_export void mi_register_deferred_free(mi_deferred_free_fun* deferred_free, void* arg) mi_attr_noexcept; + +typedef void (mi_cdecl mi_output_fun)(const char* msg, void* arg); +mi_decl_export void mi_register_output(mi_output_fun* out, void* arg) mi_attr_noexcept; + +typedef void (mi_cdecl mi_error_fun)(int err, void* arg); +mi_decl_export void mi_register_error(mi_error_fun* fun, void* arg); + +mi_decl_export void mi_collect(bool force) mi_attr_noexcept; +mi_decl_export int mi_version(void) mi_attr_noexcept; +mi_decl_export void mi_stats_reset(void) mi_attr_noexcept; +mi_decl_export void mi_stats_merge(void) mi_attr_noexcept; +mi_decl_export void mi_stats_print(void* out) mi_attr_noexcept; // backward compatibility: `out` is ignored and should be NULL +mi_decl_export void mi_stats_print_out(mi_output_fun* out, void* arg) mi_attr_noexcept; + +mi_decl_export void mi_process_init(void) mi_attr_noexcept; +mi_decl_export void mi_thread_init(void) mi_attr_noexcept; +mi_decl_export void mi_thread_done(void) mi_attr_noexcept; +mi_decl_export void mi_thread_stats_print_out(mi_output_fun* out, void* arg) mi_attr_noexcept; + +mi_decl_export void mi_process_info(size_t* elapsed_msecs, size_t* user_msecs, size_t* system_msecs, + size_t* current_rss, size_t* peak_rss, + size_t* current_commit, size_t* peak_commit, size_t* page_faults) mi_attr_noexcept; + +// ------------------------------------------------------------------------------------- +// Aligned allocation +// Note that `alignment` always follows `size` for consistency with unaligned +// allocation, but unfortunately this differs from `posix_memalign` and `aligned_alloc`. +// ------------------------------------------------------------------------------------- + +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_malloc_aligned(size_t size, size_t alignment) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size(1) mi_attr_alloc_align(2); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_malloc_aligned_at(size_t size, size_t alignment, size_t offset) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size(1); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_zalloc_aligned(size_t size, size_t alignment) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size(1) mi_attr_alloc_align(2); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_zalloc_aligned_at(size_t size, size_t alignment, size_t offset) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size(1); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_calloc_aligned(size_t count, size_t size, size_t alignment) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size2(1,2) mi_attr_alloc_align(3); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_calloc_aligned_at(size_t count, size_t size, size_t alignment, size_t offset) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size2(1,2); +mi_decl_nodiscard mi_decl_export void* mi_realloc_aligned(void* p, size_t newsize, size_t alignment) mi_attr_noexcept mi_attr_alloc_size(2) mi_attr_alloc_align(3); +mi_decl_nodiscard mi_decl_export void* mi_realloc_aligned_at(void* p, size_t newsize, size_t alignment, size_t offset) mi_attr_noexcept mi_attr_alloc_size(2); + + +// ------------------------------------------------------------------------------------- +// Heaps: first-class, but can only allocate from the same thread that created it. +// ------------------------------------------------------------------------------------- + +struct mi_heap_s; +typedef struct mi_heap_s mi_heap_t; + +mi_decl_nodiscard mi_decl_export mi_heap_t* mi_heap_new(void); +mi_decl_export void mi_heap_delete(mi_heap_t* heap); +mi_decl_export void mi_heap_destroy(mi_heap_t* heap); +mi_decl_export mi_heap_t* mi_heap_set_default(mi_heap_t* heap); +mi_decl_export mi_heap_t* mi_heap_get_default(void); +mi_decl_export mi_heap_t* mi_heap_get_backing(void); +mi_decl_export void mi_heap_collect(mi_heap_t* heap, bool force) mi_attr_noexcept; + +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_heap_malloc(mi_heap_t* heap, size_t size) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size(2); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_heap_zalloc(mi_heap_t* heap, size_t size) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size(2); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_heap_calloc(mi_heap_t* heap, size_t count, size_t size) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size2(2, 3); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_heap_mallocn(mi_heap_t* heap, size_t count, size_t size) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size2(2, 3); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_heap_malloc_small(mi_heap_t* heap, size_t size) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size(2); + +mi_decl_nodiscard mi_decl_export void* mi_heap_realloc(mi_heap_t* heap, void* p, size_t newsize) mi_attr_noexcept mi_attr_alloc_size(3); +mi_decl_nodiscard mi_decl_export void* mi_heap_reallocn(mi_heap_t* heap, void* p, size_t count, size_t size) mi_attr_noexcept mi_attr_alloc_size2(3,4); +mi_decl_nodiscard mi_decl_export void* mi_heap_reallocf(mi_heap_t* heap, void* p, size_t newsize) mi_attr_noexcept mi_attr_alloc_size(3); + +mi_decl_nodiscard mi_decl_export mi_decl_restrict char* mi_heap_strdup(mi_heap_t* heap, const char* s) mi_attr_noexcept mi_attr_malloc; +mi_decl_nodiscard mi_decl_export mi_decl_restrict char* mi_heap_strndup(mi_heap_t* heap, const char* s, size_t n) mi_attr_noexcept mi_attr_malloc; +mi_decl_nodiscard mi_decl_export mi_decl_restrict char* mi_heap_realpath(mi_heap_t* heap, const char* fname, char* resolved_name) mi_attr_noexcept mi_attr_malloc; + +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_heap_malloc_aligned(mi_heap_t* heap, size_t size, size_t alignment) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size(2) mi_attr_alloc_align(3); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_heap_malloc_aligned_at(mi_heap_t* heap, size_t size, size_t alignment, size_t offset) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size(2); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_heap_zalloc_aligned(mi_heap_t* heap, size_t size, size_t alignment) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size(2) mi_attr_alloc_align(3); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_heap_zalloc_aligned_at(mi_heap_t* heap, size_t size, size_t alignment, size_t offset) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size(2); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_heap_calloc_aligned(mi_heap_t* heap, size_t count, size_t size, size_t alignment) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size2(2, 3) mi_attr_alloc_align(4); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_heap_calloc_aligned_at(mi_heap_t* heap, size_t count, size_t size, size_t alignment, size_t offset) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size2(2, 3); +mi_decl_nodiscard mi_decl_export void* mi_heap_realloc_aligned(mi_heap_t* heap, void* p, size_t newsize, size_t alignment) mi_attr_noexcept mi_attr_alloc_size(3) mi_attr_alloc_align(4); +mi_decl_nodiscard mi_decl_export void* mi_heap_realloc_aligned_at(mi_heap_t* heap, void* p, size_t newsize, size_t alignment, size_t offset) mi_attr_noexcept mi_attr_alloc_size(3); + + +// -------------------------------------------------------------------------------- +// Zero initialized re-allocation. +// Only valid on memory that was originally allocated with zero initialization too. +// e.g. `mi_calloc`, `mi_zalloc`, `mi_zalloc_aligned` etc. +// see +// -------------------------------------------------------------------------------- + +mi_decl_nodiscard mi_decl_export void* mi_rezalloc(void* p, size_t newsize) mi_attr_noexcept mi_attr_alloc_size(2); +mi_decl_nodiscard mi_decl_export void* mi_recalloc(void* p, size_t newcount, size_t size) mi_attr_noexcept mi_attr_alloc_size2(2,3); + +mi_decl_nodiscard mi_decl_export void* mi_rezalloc_aligned(void* p, size_t newsize, size_t alignment) mi_attr_noexcept mi_attr_alloc_size(2) mi_attr_alloc_align(3); +mi_decl_nodiscard mi_decl_export void* mi_rezalloc_aligned_at(void* p, size_t newsize, size_t alignment, size_t offset) mi_attr_noexcept mi_attr_alloc_size(2); +mi_decl_nodiscard mi_decl_export void* mi_recalloc_aligned(void* p, size_t newcount, size_t size, size_t alignment) mi_attr_noexcept mi_attr_alloc_size2(2,3) mi_attr_alloc_align(4); +mi_decl_nodiscard mi_decl_export void* mi_recalloc_aligned_at(void* p, size_t newcount, size_t size, size_t alignment, size_t offset) mi_attr_noexcept mi_attr_alloc_size2(2,3); + +mi_decl_nodiscard mi_decl_export void* mi_heap_rezalloc(mi_heap_t* heap, void* p, size_t newsize) mi_attr_noexcept mi_attr_alloc_size(3); +mi_decl_nodiscard mi_decl_export void* mi_heap_recalloc(mi_heap_t* heap, void* p, size_t newcount, size_t size) mi_attr_noexcept mi_attr_alloc_size2(3,4); + +mi_decl_nodiscard mi_decl_export void* mi_heap_rezalloc_aligned(mi_heap_t* heap, void* p, size_t newsize, size_t alignment) mi_attr_noexcept mi_attr_alloc_size(3) mi_attr_alloc_align(4); +mi_decl_nodiscard mi_decl_export void* mi_heap_rezalloc_aligned_at(mi_heap_t* heap, void* p, size_t newsize, size_t alignment, size_t offset) mi_attr_noexcept mi_attr_alloc_size(3); +mi_decl_nodiscard mi_decl_export void* mi_heap_recalloc_aligned(mi_heap_t* heap, void* p, size_t newcount, size_t size, size_t alignment) mi_attr_noexcept mi_attr_alloc_size2(3,4) mi_attr_alloc_align(5); +mi_decl_nodiscard mi_decl_export void* mi_heap_recalloc_aligned_at(mi_heap_t* heap, void* p, size_t newcount, size_t size, size_t alignment, size_t offset) mi_attr_noexcept mi_attr_alloc_size2(3,4); + + +// ------------------------------------------------------ +// Analysis +// ------------------------------------------------------ + +mi_decl_export bool mi_heap_contains_block(mi_heap_t* heap, const void* p); +mi_decl_export bool mi_heap_check_owned(mi_heap_t* heap, const void* p); +mi_decl_export bool mi_check_owned(const void* p); + +// An area of heap space contains blocks of a single size. +typedef struct mi_heap_area_s { + void* blocks; // start of the area containing heap blocks + size_t reserved; // bytes reserved for this area (virtual) + size_t committed; // current available bytes for this area + size_t used; // number of allocated blocks + size_t block_size; // size in bytes of each block + size_t full_block_size; // size in bytes of a full block including padding and metadata. +} mi_heap_area_t; + +typedef bool (mi_cdecl mi_block_visit_fun)(const mi_heap_t* heap, const mi_heap_area_t* area, void* block, size_t block_size, void* arg); + +mi_decl_export bool mi_heap_visit_blocks(const mi_heap_t* heap, bool visit_all_blocks, mi_block_visit_fun* visitor, void* arg); + +// Experimental +mi_decl_nodiscard mi_decl_export bool mi_is_in_heap_region(const void* p) mi_attr_noexcept; +mi_decl_nodiscard mi_decl_export bool mi_is_redirected(void) mi_attr_noexcept; + +mi_decl_export int mi_reserve_huge_os_pages_interleave(size_t pages, size_t numa_nodes, size_t timeout_msecs) mi_attr_noexcept; +mi_decl_export int mi_reserve_huge_os_pages_at(size_t pages, int numa_node, size_t timeout_msecs) mi_attr_noexcept; + +mi_decl_export int mi_reserve_os_memory(size_t size, bool commit, bool allow_large) mi_attr_noexcept; +mi_decl_export bool mi_manage_os_memory(void* start, size_t size, bool is_committed, bool is_large, bool is_zero, int numa_node) mi_attr_noexcept; + +mi_decl_export void mi_debug_show_arenas(void) mi_attr_noexcept; + +// Experimental: heaps associated with specific memory arena's +typedef int mi_arena_id_t; +mi_decl_export void* mi_arena_area(mi_arena_id_t arena_id, size_t* size); +mi_decl_export int mi_reserve_huge_os_pages_at_ex(size_t pages, int numa_node, size_t timeout_msecs, bool exclusive, mi_arena_id_t* arena_id) mi_attr_noexcept; +mi_decl_export int mi_reserve_os_memory_ex(size_t size, bool commit, bool allow_large, bool exclusive, mi_arena_id_t* arena_id) mi_attr_noexcept; +mi_decl_export bool mi_manage_os_memory_ex(void* start, size_t size, bool is_committed, bool is_large, bool is_zero, int numa_node, bool exclusive, mi_arena_id_t* arena_id) mi_attr_noexcept; + +#if MI_MALLOC_VERSION >= 182 +// Create a heap that only allocates in the specified arena +mi_decl_nodiscard mi_decl_export mi_heap_t* mi_heap_new_in_arena(mi_arena_id_t arena_id); +#endif + +// deprecated +mi_decl_export int mi_reserve_huge_os_pages(size_t pages, double max_secs, size_t* pages_reserved) mi_attr_noexcept; + + +// ------------------------------------------------------ +// Convenience +// ------------------------------------------------------ + +#define mi_malloc_tp(tp) ((tp*)mi_malloc(sizeof(tp))) +#define mi_zalloc_tp(tp) ((tp*)mi_zalloc(sizeof(tp))) +#define mi_calloc_tp(tp,n) ((tp*)mi_calloc(n,sizeof(tp))) +#define mi_mallocn_tp(tp,n) ((tp*)mi_mallocn(n,sizeof(tp))) +#define mi_reallocn_tp(p,tp,n) ((tp*)mi_reallocn(p,n,sizeof(tp))) +#define mi_recalloc_tp(p,tp,n) ((tp*)mi_recalloc(p,n,sizeof(tp))) + +#define mi_heap_malloc_tp(hp,tp) ((tp*)mi_heap_malloc(hp,sizeof(tp))) +#define mi_heap_zalloc_tp(hp,tp) ((tp*)mi_heap_zalloc(hp,sizeof(tp))) +#define mi_heap_calloc_tp(hp,tp,n) ((tp*)mi_heap_calloc(hp,n,sizeof(tp))) +#define mi_heap_mallocn_tp(hp,tp,n) ((tp*)mi_heap_mallocn(hp,n,sizeof(tp))) +#define mi_heap_reallocn_tp(hp,p,tp,n) ((tp*)mi_heap_reallocn(hp,p,n,sizeof(tp))) +#define mi_heap_recalloc_tp(hp,p,tp,n) ((tp*)mi_heap_recalloc(hp,p,n,sizeof(tp))) + + +// ------------------------------------------------------ +// Options +// ------------------------------------------------------ + +typedef enum mi_option_e { + // stable options + mi_option_show_errors, // print error messages + mi_option_show_stats, // print statistics on termination + mi_option_verbose, // print verbose messages + // the following options are experimental (see src/options.h) + mi_option_eager_commit, // eager commit segments? (after `eager_commit_delay` segments) (=1) + mi_option_arena_eager_commit, // eager commit arenas? Use 2 to enable just on overcommit systems (=2) + mi_option_purge_decommits, // should a memory purge decommit (or only reset) (=1) + mi_option_allow_large_os_pages, // allow large (2MiB) OS pages, implies eager commit + mi_option_reserve_huge_os_pages, // reserve N huge OS pages (1GiB/page) at startup + mi_option_reserve_huge_os_pages_at, // reserve huge OS pages at a specific NUMA node + mi_option_reserve_os_memory, // reserve specified amount of OS memory in an arena at startup + mi_option_deprecated_segment_cache, + mi_option_deprecated_page_reset, + mi_option_abandoned_page_purge, // immediately purge delayed purges on thread termination + mi_option_deprecated_segment_reset, + mi_option_eager_commit_delay, + mi_option_purge_delay, // memory purging is delayed by N milli seconds; use 0 for immediate purging or -1 for no purging at all. + mi_option_use_numa_nodes, // 0 = use all available numa nodes, otherwise use at most N nodes. + mi_option_limit_os_alloc, // 1 = do not use OS memory for allocation (but only programmatically reserved arenas) + mi_option_os_tag, // tag used for OS logging (macOS only for now) + mi_option_max_errors, // issue at most N error messages + mi_option_max_warnings, // issue at most N warning messages + mi_option_max_segment_reclaim, + mi_option_destroy_on_exit, // if set, release all memory on exit; sometimes used for dynamic unloading but can be unsafe. + mi_option_arena_reserve, // initial memory size in KiB for arena reservation (1GiB on 64-bit) + mi_option_arena_purge_mult, + mi_option_purge_extend_delay, + _mi_option_last, + // legacy option names + mi_option_large_os_pages = mi_option_allow_large_os_pages, + mi_option_eager_region_commit = mi_option_arena_eager_commit, + mi_option_reset_decommits = mi_option_purge_decommits, + mi_option_reset_delay = mi_option_purge_delay, + mi_option_abandoned_page_reset = mi_option_abandoned_page_purge +} mi_option_t; + + +mi_decl_nodiscard mi_decl_export bool mi_option_is_enabled(mi_option_t option); +mi_decl_export void mi_option_enable(mi_option_t option); +mi_decl_export void mi_option_disable(mi_option_t option); +mi_decl_export void mi_option_set_enabled(mi_option_t option, bool enable); +mi_decl_export void mi_option_set_enabled_default(mi_option_t option, bool enable); + +mi_decl_nodiscard mi_decl_export long mi_option_get(mi_option_t option); +mi_decl_nodiscard mi_decl_export long mi_option_get_clamp(mi_option_t option, long min, long max); +mi_decl_nodiscard mi_decl_export size_t mi_option_get_size(mi_option_t option); +mi_decl_export void mi_option_set(mi_option_t option, long value); +mi_decl_export void mi_option_set_default(mi_option_t option, long value); + + +// ------------------------------------------------------------------------------------------------------- +// "mi" prefixed implementations of various posix, Unix, Windows, and C++ allocation functions. +// (This can be convenient when providing overrides of these functions as done in `mimalloc-override.h`.) +// note: we use `mi_cfree` as "checked free" and it checks if the pointer is in our heap before free-ing. +// ------------------------------------------------------------------------------------------------------- + +mi_decl_export void mi_cfree(void* p) mi_attr_noexcept; +mi_decl_export void* mi__expand(void* p, size_t newsize) mi_attr_noexcept; +mi_decl_nodiscard mi_decl_export size_t mi_malloc_size(const void* p) mi_attr_noexcept; +mi_decl_nodiscard mi_decl_export size_t mi_malloc_good_size(size_t size) mi_attr_noexcept; +mi_decl_nodiscard mi_decl_export size_t mi_malloc_usable_size(const void *p) mi_attr_noexcept; + +mi_decl_export int mi_posix_memalign(void** p, size_t alignment, size_t size) mi_attr_noexcept; +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_memalign(size_t alignment, size_t size) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size(2) mi_attr_alloc_align(1); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_valloc(size_t size) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size(1); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_pvalloc(size_t size) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size(1); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_aligned_alloc(size_t alignment, size_t size) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size(2) mi_attr_alloc_align(1); + +mi_decl_nodiscard mi_decl_export void* mi_reallocarray(void* p, size_t count, size_t size) mi_attr_noexcept mi_attr_alloc_size2(2,3); +mi_decl_nodiscard mi_decl_export int mi_reallocarr(void* p, size_t count, size_t size) mi_attr_noexcept; +mi_decl_nodiscard mi_decl_export void* mi_aligned_recalloc(void* p, size_t newcount, size_t size, size_t alignment) mi_attr_noexcept; +mi_decl_nodiscard mi_decl_export void* mi_aligned_offset_recalloc(void* p, size_t newcount, size_t size, size_t alignment, size_t offset) mi_attr_noexcept; + +mi_decl_nodiscard mi_decl_export mi_decl_restrict unsigned short* mi_wcsdup(const unsigned short* s) mi_attr_noexcept mi_attr_malloc; +mi_decl_nodiscard mi_decl_export mi_decl_restrict unsigned char* mi_mbsdup(const unsigned char* s) mi_attr_noexcept mi_attr_malloc; +mi_decl_export int mi_dupenv_s(char** buf, size_t* size, const char* name) mi_attr_noexcept; +mi_decl_export int mi_wdupenv_s(unsigned short** buf, size_t* size, const unsigned short* name) mi_attr_noexcept; + +mi_decl_export void mi_free_size(void* p, size_t size) mi_attr_noexcept; +mi_decl_export void mi_free_size_aligned(void* p, size_t size, size_t alignment) mi_attr_noexcept; +mi_decl_export void mi_free_aligned(void* p, size_t alignment) mi_attr_noexcept; + +// The `mi_new` wrappers implement C++ semantics on out-of-memory instead of directly returning `NULL`. +// (and call `std::get_new_handler` and potentially raise a `std::bad_alloc` exception). +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_new(size_t size) mi_attr_malloc mi_attr_alloc_size(1); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_new_aligned(size_t size, size_t alignment) mi_attr_malloc mi_attr_alloc_size(1) mi_attr_alloc_align(2); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_new_nothrow(size_t size) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size(1); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_new_aligned_nothrow(size_t size, size_t alignment) mi_attr_noexcept mi_attr_malloc mi_attr_alloc_size(1) mi_attr_alloc_align(2); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_new_n(size_t count, size_t size) mi_attr_malloc mi_attr_alloc_size2(1, 2); +mi_decl_nodiscard mi_decl_export void* mi_new_realloc(void* p, size_t newsize) mi_attr_alloc_size(2); +mi_decl_nodiscard mi_decl_export void* mi_new_reallocn(void* p, size_t newcount, size_t size) mi_attr_alloc_size2(2, 3); + +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_heap_alloc_new(mi_heap_t* heap, size_t size) mi_attr_malloc mi_attr_alloc_size(2); +mi_decl_nodiscard mi_decl_export mi_decl_restrict void* mi_heap_alloc_new_n(mi_heap_t* heap, size_t count, size_t size) mi_attr_malloc mi_attr_alloc_size2(2, 3); + +#ifdef __cplusplus +} +#endif + +// --------------------------------------------------------------------------------------------- +// Implement the C++ std::allocator interface for use in STL containers. +// (note: see `mimalloc-new-delete.h` for overriding the new/delete operators globally) +// --------------------------------------------------------------------------------------------- +#ifdef __cplusplus + +#include // std::size_t +#include // PTRDIFF_MAX +#if (__cplusplus >= 201103L) || (_MSC_VER > 1900) // C++11 +#include // std::true_type +#include // std::forward +#endif + +template struct _mi_stl_allocator_common { + typedef T value_type; + typedef std::size_t size_type; + typedef std::ptrdiff_t difference_type; + typedef value_type& reference; + typedef value_type const& const_reference; + typedef value_type* pointer; + typedef value_type const* const_pointer; + + #if ((__cplusplus >= 201103L) || (_MSC_VER > 1900)) // C++11 + using propagate_on_container_copy_assignment = std::true_type; + using propagate_on_container_move_assignment = std::true_type; + using propagate_on_container_swap = std::true_type; + template void construct(U* p, Args&& ...args) { ::new(p) U(std::forward(args)...); } + template void destroy(U* p) mi_attr_noexcept { p->~U(); } + #else + void construct(pointer p, value_type const& val) { ::new(p) value_type(val); } + void destroy(pointer p) { p->~value_type(); } + #endif + + size_type max_size() const mi_attr_noexcept { return (PTRDIFF_MAX/sizeof(value_type)); } + pointer address(reference x) const { return &x; } + const_pointer address(const_reference x) const { return &x; } +}; + +template struct mi_stl_allocator : public _mi_stl_allocator_common { + using typename _mi_stl_allocator_common::size_type; + using typename _mi_stl_allocator_common::value_type; + using typename _mi_stl_allocator_common::pointer; + template struct rebind { typedef mi_stl_allocator other; }; + + mi_stl_allocator() mi_attr_noexcept = default; + mi_stl_allocator(const mi_stl_allocator&) mi_attr_noexcept = default; + template mi_stl_allocator(const mi_stl_allocator&) mi_attr_noexcept { } + mi_stl_allocator select_on_container_copy_construction() const { return *this; } + void deallocate(T* p, size_type) { mi_free(p); } + + #if (__cplusplus >= 201703L) // C++17 + mi_decl_nodiscard T* allocate(size_type count) { return static_cast(mi_new_n(count, sizeof(T))); } + mi_decl_nodiscard T* allocate(size_type count, const void*) { return allocate(count); } + #else + mi_decl_nodiscard pointer allocate(size_type count, const void* = 0) { return static_cast(mi_new_n(count, sizeof(value_type))); } + #endif + + #if ((__cplusplus >= 201103L) || (_MSC_VER > 1900)) // C++11 + using is_always_equal = std::true_type; + #endif +}; + +template bool operator==(const mi_stl_allocator& , const mi_stl_allocator& ) mi_attr_noexcept { return true; } +template bool operator!=(const mi_stl_allocator& , const mi_stl_allocator& ) mi_attr_noexcept { return false; } + + +#if (__cplusplus >= 201103L) || (_MSC_VER >= 1900) // C++11 +#define MI_HAS_HEAP_STL_ALLOCATOR 1 + +#include // std::shared_ptr + +// Common base class for STL allocators in a specific heap +template struct _mi_heap_stl_allocator_common : public _mi_stl_allocator_common { + using typename _mi_stl_allocator_common::size_type; + using typename _mi_stl_allocator_common::value_type; + using typename _mi_stl_allocator_common::pointer; + + _mi_heap_stl_allocator_common(mi_heap_t* hp) : heap(hp) { } /* will not delete nor destroy the passed in heap */ + + #if (__cplusplus >= 201703L) // C++17 + mi_decl_nodiscard T* allocate(size_type count) { return static_cast(mi_heap_alloc_new_n(this->heap.get(), count, sizeof(T))); } + mi_decl_nodiscard T* allocate(size_type count, const void*) { return allocate(count); } + #else + mi_decl_nodiscard pointer allocate(size_type count, const void* = 0) { return static_cast(mi_heap_alloc_new_n(this->heap.get(), count, sizeof(value_type))); } + #endif + + #if ((__cplusplus >= 201103L) || (_MSC_VER > 1900)) // C++11 + using is_always_equal = std::false_type; + #endif + + void collect(bool force) { mi_heap_collect(this->heap.get(), force); } + template bool is_equal(const _mi_heap_stl_allocator_common& x) const { return (this->heap == x.heap); } + +protected: + std::shared_ptr heap; + template friend struct _mi_heap_stl_allocator_common; + + _mi_heap_stl_allocator_common() { + mi_heap_t* hp = mi_heap_new(); + this->heap.reset(hp, (_mi_destroy ? &heap_destroy : &heap_delete)); /* calls heap_delete/destroy when the refcount drops to zero */ + } + _mi_heap_stl_allocator_common(const _mi_heap_stl_allocator_common& x) mi_attr_noexcept : heap(x.heap) { } + template _mi_heap_stl_allocator_common(const _mi_heap_stl_allocator_common& x) mi_attr_noexcept : heap(x.heap) { } + +private: + static void heap_delete(mi_heap_t* hp) { if (hp != NULL) { mi_heap_delete(hp); } } + static void heap_destroy(mi_heap_t* hp) { if (hp != NULL) { mi_heap_destroy(hp); } } +}; + +// STL allocator allocation in a specific heap +template struct mi_heap_stl_allocator : public _mi_heap_stl_allocator_common { + using typename _mi_heap_stl_allocator_common::size_type; + mi_heap_stl_allocator() : _mi_heap_stl_allocator_common() { } // creates fresh heap that is deleted when the destructor is called + mi_heap_stl_allocator(mi_heap_t* hp) : _mi_heap_stl_allocator_common(hp) { } // no delete nor destroy on the passed in heap + template mi_heap_stl_allocator(const mi_heap_stl_allocator& x) mi_attr_noexcept : _mi_heap_stl_allocator_common(x) { } + + mi_heap_stl_allocator select_on_container_copy_construction() const { return *this; } + void deallocate(T* p, size_type) { mi_free(p); } + template struct rebind { typedef mi_heap_stl_allocator other; }; +}; + +template bool operator==(const mi_heap_stl_allocator& x, const mi_heap_stl_allocator& y) mi_attr_noexcept { return (x.is_equal(y)); } +template bool operator!=(const mi_heap_stl_allocator& x, const mi_heap_stl_allocator& y) mi_attr_noexcept { return (!x.is_equal(y)); } + + +// STL allocator allocation in a specific heap, where `free` does nothing and +// the heap is destroyed in one go on destruction -- use with care! +template struct mi_heap_destroy_stl_allocator : public _mi_heap_stl_allocator_common { + using typename _mi_heap_stl_allocator_common::size_type; + mi_heap_destroy_stl_allocator() : _mi_heap_stl_allocator_common() { } // creates fresh heap that is destroyed when the destructor is called + mi_heap_destroy_stl_allocator(mi_heap_t* hp) : _mi_heap_stl_allocator_common(hp) { } // no delete nor destroy on the passed in heap + template mi_heap_destroy_stl_allocator(const mi_heap_destroy_stl_allocator& x) mi_attr_noexcept : _mi_heap_stl_allocator_common(x) { } + + mi_heap_destroy_stl_allocator select_on_container_copy_construction() const { return *this; } + void deallocate(T*, size_type) { /* do nothing as we destroy the heap on destruct. */ } + template struct rebind { typedef mi_heap_destroy_stl_allocator other; }; +}; + +template bool operator==(const mi_heap_destroy_stl_allocator& x, const mi_heap_destroy_stl_allocator& y) mi_attr_noexcept { return (x.is_equal(y)); } +template bool operator!=(const mi_heap_destroy_stl_allocator& x, const mi_heap_destroy_stl_allocator& y) mi_attr_noexcept { return (!x.is_equal(y)); } + +#endif // C++11 + +#endif // __cplusplus + +#endif diff --git a/compat/mimalloc/mimalloc/atomic.h b/compat/mimalloc/mimalloc/atomic.h new file mode 100644 index 00000000000000..c6b8146ffdb049 --- /dev/null +++ b/compat/mimalloc/mimalloc/atomic.h @@ -0,0 +1,385 @@ +/* ---------------------------------------------------------------------------- +Copyright (c) 2018-2023 Microsoft Research, Daan Leijen +This is free software; you can redistribute it and/or modify it under the +terms of the MIT license. A copy of the license can be found in the file +"LICENSE" at the root of this distribution. +-----------------------------------------------------------------------------*/ +#pragma once +#ifndef MIMALLOC_ATOMIC_H +#define MIMALLOC_ATOMIC_H + +// -------------------------------------------------------------------------------------------- +// Atomics +// We need to be portable between C, C++, and MSVC. +// We base the primitives on the C/C++ atomics and create a mimimal wrapper for MSVC in C compilation mode. +// This is why we try to use only `uintptr_t` and `*` as atomic types. +// To gain better insight in the range of used atomics, we use explicitly named memory order operations +// instead of passing the memory order as a parameter. +// ----------------------------------------------------------------------------------------------- + +#if defined(__cplusplus) +// Use C++ atomics +#include +#define _Atomic(tp) std::atomic +#define mi_atomic(name) std::atomic_##name +#define mi_memory_order(name) std::memory_order_##name +#if !defined(ATOMIC_VAR_INIT) || (__cplusplus >= 202002L) // c++20, see issue #571 + #define MI_ATOMIC_VAR_INIT(x) x +#else + #define MI_ATOMIC_VAR_INIT(x) ATOMIC_VAR_INIT(x) +#endif +#elif defined(_MSC_VER) +// Use MSVC C wrapper for C11 atomics +#define _Atomic(tp) tp +#define MI_ATOMIC_VAR_INIT(x) x +#define mi_atomic(name) mi_atomic_##name +#define mi_memory_order(name) mi_memory_order_##name +#else +// Use C11 atomics +#include +#define mi_atomic(name) atomic_##name +#define mi_memory_order(name) memory_order_##name +#if !defined(ATOMIC_VAR_INIT) || (__STDC_VERSION__ >= 201710L) // c17, see issue #735 + #define MI_ATOMIC_VAR_INIT(x) x +#else + #define MI_ATOMIC_VAR_INIT(x) ATOMIC_VAR_INIT(x) +#endif +#endif + +// Various defines for all used memory orders in mimalloc +#define mi_atomic_cas_weak(p,expected,desired,mem_success,mem_fail) \ + mi_atomic(compare_exchange_weak_explicit)(p,expected,desired,mem_success,mem_fail) + +#define mi_atomic_cas_strong(p,expected,desired,mem_success,mem_fail) \ + mi_atomic(compare_exchange_strong_explicit)(p,expected,desired,mem_success,mem_fail) + +#define mi_atomic_load_acquire(p) mi_atomic(load_explicit)(p,mi_memory_order(acquire)) +#define mi_atomic_load_relaxed(p) mi_atomic(load_explicit)(p,mi_memory_order(relaxed)) +#define mi_atomic_store_release(p,x) mi_atomic(store_explicit)(p,x,mi_memory_order(release)) +#define mi_atomic_store_relaxed(p,x) mi_atomic(store_explicit)(p,x,mi_memory_order(relaxed)) +#define mi_atomic_exchange_release(p,x) mi_atomic(exchange_explicit)(p,x,mi_memory_order(release)) +#define mi_atomic_exchange_acq_rel(p,x) mi_atomic(exchange_explicit)(p,x,mi_memory_order(acq_rel)) +#define mi_atomic_cas_weak_release(p,exp,des) mi_atomic_cas_weak(p,exp,des,mi_memory_order(release),mi_memory_order(relaxed)) +#define mi_atomic_cas_weak_acq_rel(p,exp,des) mi_atomic_cas_weak(p,exp,des,mi_memory_order(acq_rel),mi_memory_order(acquire)) +#define mi_atomic_cas_strong_release(p,exp,des) mi_atomic_cas_strong(p,exp,des,mi_memory_order(release),mi_memory_order(relaxed)) +#define mi_atomic_cas_strong_acq_rel(p,exp,des) mi_atomic_cas_strong(p,exp,des,mi_memory_order(acq_rel),mi_memory_order(acquire)) + +#define mi_atomic_add_relaxed(p,x) mi_atomic(fetch_add_explicit)(p,x,mi_memory_order(relaxed)) +#define mi_atomic_sub_relaxed(p,x) mi_atomic(fetch_sub_explicit)(p,x,mi_memory_order(relaxed)) +#define mi_atomic_add_acq_rel(p,x) mi_atomic(fetch_add_explicit)(p,x,mi_memory_order(acq_rel)) +#define mi_atomic_sub_acq_rel(p,x) mi_atomic(fetch_sub_explicit)(p,x,mi_memory_order(acq_rel)) +#define mi_atomic_and_acq_rel(p,x) mi_atomic(fetch_and_explicit)(p,x,mi_memory_order(acq_rel)) +#define mi_atomic_or_acq_rel(p,x) mi_atomic(fetch_or_explicit)(p,x,mi_memory_order(acq_rel)) + +#define mi_atomic_increment_relaxed(p) mi_atomic_add_relaxed(p,(uintptr_t)1) +#define mi_atomic_decrement_relaxed(p) mi_atomic_sub_relaxed(p,(uintptr_t)1) +#define mi_atomic_increment_acq_rel(p) mi_atomic_add_acq_rel(p,(uintptr_t)1) +#define mi_atomic_decrement_acq_rel(p) mi_atomic_sub_acq_rel(p,(uintptr_t)1) + +static inline void mi_atomic_yield(void); +static inline intptr_t mi_atomic_addi(_Atomic(intptr_t)*p, intptr_t add); +static inline intptr_t mi_atomic_subi(_Atomic(intptr_t)*p, intptr_t sub); + + +#if defined(__cplusplus) || !defined(_MSC_VER) + +// In C++/C11 atomics we have polymorphic atomics so can use the typed `ptr` variants (where `tp` is the type of atomic value) +// We use these macros so we can provide a typed wrapper in MSVC in C compilation mode as well +#define mi_atomic_load_ptr_acquire(tp,p) mi_atomic_load_acquire(p) +#define mi_atomic_load_ptr_relaxed(tp,p) mi_atomic_load_relaxed(p) + +// In C++ we need to add casts to help resolve templates if NULL is passed +#if defined(__cplusplus) +#define mi_atomic_store_ptr_release(tp,p,x) mi_atomic_store_release(p,(tp*)x) +#define mi_atomic_store_ptr_relaxed(tp,p,x) mi_atomic_store_relaxed(p,(tp*)x) +#define mi_atomic_cas_ptr_weak_release(tp,p,exp,des) mi_atomic_cas_weak_release(p,exp,(tp*)des) +#define mi_atomic_cas_ptr_weak_acq_rel(tp,p,exp,des) mi_atomic_cas_weak_acq_rel(p,exp,(tp*)des) +#define mi_atomic_cas_ptr_strong_release(tp,p,exp,des) mi_atomic_cas_strong_release(p,exp,(tp*)des) +#define mi_atomic_exchange_ptr_release(tp,p,x) mi_atomic_exchange_release(p,(tp*)x) +#define mi_atomic_exchange_ptr_acq_rel(tp,p,x) mi_atomic_exchange_acq_rel(p,(tp*)x) +#else +#define mi_atomic_store_ptr_release(tp,p,x) mi_atomic_store_release(p,x) +#define mi_atomic_store_ptr_relaxed(tp,p,x) mi_atomic_store_relaxed(p,x) +#define mi_atomic_cas_ptr_weak_release(tp,p,exp,des) mi_atomic_cas_weak_release(p,exp,des) +#define mi_atomic_cas_ptr_weak_acq_rel(tp,p,exp,des) mi_atomic_cas_weak_acq_rel(p,exp,des) +#define mi_atomic_cas_ptr_strong_release(tp,p,exp,des) mi_atomic_cas_strong_release(p,exp,des) +#define mi_atomic_exchange_ptr_release(tp,p,x) mi_atomic_exchange_release(p,x) +#define mi_atomic_exchange_ptr_acq_rel(tp,p,x) mi_atomic_exchange_acq_rel(p,x) +#endif + +// These are used by the statistics +static inline int64_t mi_atomic_addi64_relaxed(volatile int64_t* p, int64_t add) { + return mi_atomic(fetch_add_explicit)((_Atomic(int64_t)*)p, add, mi_memory_order(relaxed)); +} +static inline void mi_atomic_maxi64_relaxed(volatile int64_t* p, int64_t x) { + int64_t current = mi_atomic_load_relaxed((_Atomic(int64_t)*)p); + while (current < x && !mi_atomic_cas_weak_release((_Atomic(int64_t)*)p, ¤t, x)) { /* nothing */ }; +} + +// Used by timers +#define mi_atomic_loadi64_acquire(p) mi_atomic(load_explicit)(p,mi_memory_order(acquire)) +#define mi_atomic_loadi64_relaxed(p) mi_atomic(load_explicit)(p,mi_memory_order(relaxed)) +#define mi_atomic_storei64_release(p,x) mi_atomic(store_explicit)(p,x,mi_memory_order(release)) +#define mi_atomic_storei64_relaxed(p,x) mi_atomic(store_explicit)(p,x,mi_memory_order(relaxed)) + +#define mi_atomic_casi64_strong_acq_rel(p,e,d) mi_atomic_cas_strong_acq_rel(p,e,d) +#define mi_atomic_addi64_acq_rel(p,i) mi_atomic_add_acq_rel(p,i) + + +#elif defined(_MSC_VER) + +// MSVC C compilation wrapper that uses Interlocked operations to model C11 atomics. +#define WIN32_LEAN_AND_MEAN +#include +#include +#ifdef _WIN64 +typedef LONG64 msc_intptr_t; +#define MI_64(f) f##64 +#else +typedef LONG msc_intptr_t; +#define MI_64(f) f +#endif + +typedef enum mi_memory_order_e { + mi_memory_order_relaxed, + mi_memory_order_consume, + mi_memory_order_acquire, + mi_memory_order_release, + mi_memory_order_acq_rel, + mi_memory_order_seq_cst +} mi_memory_order; + +static inline uintptr_t mi_atomic_fetch_add_explicit(_Atomic(uintptr_t)*p, uintptr_t add, mi_memory_order mo) { + (void)(mo); + return (uintptr_t)MI_64(_InterlockedExchangeAdd)((volatile msc_intptr_t*)p, (msc_intptr_t)add); +} +static inline uintptr_t mi_atomic_fetch_sub_explicit(_Atomic(uintptr_t)*p, uintptr_t sub, mi_memory_order mo) { + (void)(mo); + return (uintptr_t)MI_64(_InterlockedExchangeAdd)((volatile msc_intptr_t*)p, -((msc_intptr_t)sub)); +} +static inline uintptr_t mi_atomic_fetch_and_explicit(_Atomic(uintptr_t)*p, uintptr_t x, mi_memory_order mo) { + (void)(mo); + return (uintptr_t)MI_64(_InterlockedAnd)((volatile msc_intptr_t*)p, (msc_intptr_t)x); +} +static inline uintptr_t mi_atomic_fetch_or_explicit(_Atomic(uintptr_t)*p, uintptr_t x, mi_memory_order mo) { + (void)(mo); + return (uintptr_t)MI_64(_InterlockedOr)((volatile msc_intptr_t*)p, (msc_intptr_t)x); +} +static inline bool mi_atomic_compare_exchange_strong_explicit(_Atomic(uintptr_t)*p, uintptr_t* expected, uintptr_t desired, mi_memory_order mo1, mi_memory_order mo2) { + (void)(mo1); (void)(mo2); + uintptr_t read = (uintptr_t)MI_64(_InterlockedCompareExchange)((volatile msc_intptr_t*)p, (msc_intptr_t)desired, (msc_intptr_t)(*expected)); + if (read == *expected) { + return true; + } + else { + *expected = read; + return false; + } +} +static inline bool mi_atomic_compare_exchange_weak_explicit(_Atomic(uintptr_t)*p, uintptr_t* expected, uintptr_t desired, mi_memory_order mo1, mi_memory_order mo2) { + return mi_atomic_compare_exchange_strong_explicit(p, expected, desired, mo1, mo2); +} +static inline uintptr_t mi_atomic_exchange_explicit(_Atomic(uintptr_t)*p, uintptr_t exchange, mi_memory_order mo) { + (void)(mo); + return (uintptr_t)MI_64(_InterlockedExchange)((volatile msc_intptr_t*)p, (msc_intptr_t)exchange); +} +static inline void mi_atomic_thread_fence(mi_memory_order mo) { + (void)(mo); + _Atomic(uintptr_t) x = 0; + mi_atomic_exchange_explicit(&x, 1, mo); +} +static inline uintptr_t mi_atomic_load_explicit(_Atomic(uintptr_t) const* p, mi_memory_order mo) { + (void)(mo); +#if defined(_M_IX86) || defined(_M_X64) + return *p; +#else + uintptr_t x = *p; + if (mo > mi_memory_order_relaxed) { + while (!mi_atomic_compare_exchange_weak_explicit(p, &x, x, mo, mi_memory_order_relaxed)) { /* nothing */ }; + } + return x; +#endif +} +static inline void mi_atomic_store_explicit(_Atomic(uintptr_t)*p, uintptr_t x, mi_memory_order mo) { + (void)(mo); +#if defined(_M_IX86) || defined(_M_X64) + *p = x; +#else + mi_atomic_exchange_explicit(p, x, mo); +#endif +} +static inline int64_t mi_atomic_loadi64_explicit(_Atomic(int64_t)*p, mi_memory_order mo) { + (void)(mo); +#if defined(_M_X64) + return *p; +#else + int64_t old = *p; + int64_t x = old; + while ((old = InterlockedCompareExchange64(p, x, old)) != x) { + x = old; + } + return x; +#endif +} +static inline void mi_atomic_storei64_explicit(_Atomic(int64_t)*p, int64_t x, mi_memory_order mo) { + (void)(mo); +#if defined(x_M_IX86) || defined(_M_X64) + *p = x; +#else + InterlockedExchange64(p, x); +#endif +} + +// These are used by the statistics +static inline int64_t mi_atomic_addi64_relaxed(volatile _Atomic(int64_t)*p, int64_t add) { +#ifdef _WIN64 + return (int64_t)mi_atomic_addi((int64_t*)p, add); +#else + int64_t current; + int64_t sum; + do { + current = *p; + sum = current + add; + } while (_InterlockedCompareExchange64(p, sum, current) != current); + return current; +#endif +} +static inline void mi_atomic_maxi64_relaxed(volatile _Atomic(int64_t)*p, int64_t x) { + int64_t current; + do { + current = *p; + } while (current < x && _InterlockedCompareExchange64(p, x, current) != current); +} + +static inline void mi_atomic_addi64_acq_rel(volatile _Atomic(int64_t*)p, int64_t i) { + mi_atomic_addi64_relaxed(p, i); +} + +static inline bool mi_atomic_casi64_strong_acq_rel(volatile _Atomic(int64_t*)p, int64_t* exp, int64_t des) { + int64_t read = _InterlockedCompareExchange64(p, des, *exp); + if (read == *exp) { + return true; + } + else { + *exp = read; + return false; + } +} + +// The pointer macros cast to `uintptr_t`. +#define mi_atomic_load_ptr_acquire(tp,p) (tp*)mi_atomic_load_acquire((_Atomic(uintptr_t)*)(p)) +#define mi_atomic_load_ptr_relaxed(tp,p) (tp*)mi_atomic_load_relaxed((_Atomic(uintptr_t)*)(p)) +#define mi_atomic_store_ptr_release(tp,p,x) mi_atomic_store_release((_Atomic(uintptr_t)*)(p),(uintptr_t)(x)) +#define mi_atomic_store_ptr_relaxed(tp,p,x) mi_atomic_store_relaxed((_Atomic(uintptr_t)*)(p),(uintptr_t)(x)) +#define mi_atomic_cas_ptr_weak_release(tp,p,exp,des) mi_atomic_cas_weak_release((_Atomic(uintptr_t)*)(p),(uintptr_t*)exp,(uintptr_t)des) +#define mi_atomic_cas_ptr_weak_acq_rel(tp,p,exp,des) mi_atomic_cas_weak_acq_rel((_Atomic(uintptr_t)*)(p),(uintptr_t*)exp,(uintptr_t)des) +#define mi_atomic_cas_ptr_strong_release(tp,p,exp,des) mi_atomic_cas_strong_release((_Atomic(uintptr_t)*)(p),(uintptr_t*)exp,(uintptr_t)des) +#define mi_atomic_exchange_ptr_release(tp,p,x) (tp*)mi_atomic_exchange_release((_Atomic(uintptr_t)*)(p),(uintptr_t)x) +#define mi_atomic_exchange_ptr_acq_rel(tp,p,x) (tp*)mi_atomic_exchange_acq_rel((_Atomic(uintptr_t)*)(p),(uintptr_t)x) + +#define mi_atomic_loadi64_acquire(p) mi_atomic(loadi64_explicit)(p,mi_memory_order(acquire)) +#define mi_atomic_loadi64_relaxed(p) mi_atomic(loadi64_explicit)(p,mi_memory_order(relaxed)) +#define mi_atomic_storei64_release(p,x) mi_atomic(storei64_explicit)(p,x,mi_memory_order(release)) +#define mi_atomic_storei64_relaxed(p,x) mi_atomic(storei64_explicit)(p,x,mi_memory_order(relaxed)) + + +#endif + + +// Atomically add a signed value; returns the previous value. +static inline intptr_t mi_atomic_addi(_Atomic(intptr_t)*p, intptr_t add) { + return (intptr_t)mi_atomic_add_acq_rel((_Atomic(uintptr_t)*)p, (uintptr_t)add); +} + +// Atomically subtract a signed value; returns the previous value. +static inline intptr_t mi_atomic_subi(_Atomic(intptr_t)*p, intptr_t sub) { + return (intptr_t)mi_atomic_addi(p, -sub); +} + +typedef _Atomic(uintptr_t) mi_atomic_once_t; + +// Returns true only on the first invocation +static inline bool mi_atomic_once( mi_atomic_once_t* once ) { + if (mi_atomic_load_relaxed(once) != 0) return false; // quick test + uintptr_t expected = 0; + return mi_atomic_cas_strong_acq_rel(once, &expected, (uintptr_t)1); // try to set to 1 +} + +typedef _Atomic(uintptr_t) mi_atomic_guard_t; + +// Allows only one thread to execute at a time +#define mi_atomic_guard(guard) \ + uintptr_t _mi_guard_expected = 0; \ + for(bool _mi_guard_once = true; \ + _mi_guard_once && mi_atomic_cas_strong_acq_rel(guard,&_mi_guard_expected,(uintptr_t)1); \ + (mi_atomic_store_release(guard,(uintptr_t)0), _mi_guard_once = false) ) + + + +// Yield +#if defined(__cplusplus) +#include +static inline void mi_atomic_yield(void) { + std::this_thread::yield(); +} +#elif defined(_WIN32) +#define WIN32_LEAN_AND_MEAN +#include +static inline void mi_atomic_yield(void) { + YieldProcessor(); +} +#elif defined(__SSE2__) +#include +static inline void mi_atomic_yield(void) { + _mm_pause(); +} +#elif (defined(__GNUC__) || defined(__clang__)) && \ + (defined(__x86_64__) || defined(__i386__) || defined(__arm__) || defined(__armel__) || defined(__ARMEL__) || \ + defined(__aarch64__) || defined(__powerpc__) || defined(__ppc__) || defined(__PPC__)) || defined(__POWERPC__) +#if defined(__x86_64__) || defined(__i386__) +static inline void mi_atomic_yield(void) { + __asm__ volatile ("pause" ::: "memory"); +} +#elif defined(__aarch64__) +static inline void mi_atomic_yield(void) { + __asm__ volatile("wfe"); +} +#elif (defined(__arm__) && __ARM_ARCH__ >= 7) +static inline void mi_atomic_yield(void) { + __asm__ volatile("yield" ::: "memory"); +} +#elif defined(__powerpc__) || defined(__ppc__) || defined(__PPC__) || defined(__POWERPC__) +#ifdef __APPLE__ +static inline void mi_atomic_yield(void) { + __asm__ volatile ("or r27,r27,r27" ::: "memory"); +} +#else +static inline void mi_atomic_yield(void) { + __asm__ __volatile__ ("or 27,27,27" ::: "memory"); +} +#endif +#elif defined(__armel__) || defined(__ARMEL__) +static inline void mi_atomic_yield(void) { + __asm__ volatile ("nop" ::: "memory"); +} +#endif +#elif defined(__sun) +// Fallback for other archs +#include +static inline void mi_atomic_yield(void) { + smt_pause(); +} +#elif defined(__wasi__) +#include +static inline void mi_atomic_yield(void) { + sched_yield(); +} +#else +#include +static inline void mi_atomic_yield(void) { + sleep(0); +} +#endif + + +#endif // __MIMALLOC_ATOMIC_H diff --git a/compat/mimalloc/mimalloc/internal.h b/compat/mimalloc/mimalloc/internal.h new file mode 100644 index 00000000000000..f076bc6a40f977 --- /dev/null +++ b/compat/mimalloc/mimalloc/internal.h @@ -0,0 +1,979 @@ +/* ---------------------------------------------------------------------------- +Copyright (c) 2018-2023, Microsoft Research, Daan Leijen +This is free software; you can redistribute it and/or modify it under the +terms of the MIT license. A copy of the license can be found in the file +"LICENSE" at the root of this distribution. +-----------------------------------------------------------------------------*/ +#pragma once +#ifndef MIMALLOC_INTERNAL_H +#define MIMALLOC_INTERNAL_H + + +// -------------------------------------------------------------------------- +// This file contains the interal API's of mimalloc and various utility +// functions and macros. +// -------------------------------------------------------------------------- + +#include "mimalloc/types.h" +#include "mimalloc/track.h" + +#if (MI_DEBUG>0) +#define mi_trace_message(...) _mi_trace_message(__VA_ARGS__) +#else +#define mi_trace_message(...) +#endif + +#define MI_CACHE_LINE 64 +#if defined(_MSC_VER) +#pragma warning(disable:4127) // suppress constant conditional warning (due to MI_SECURE paths) +#pragma warning(disable:26812) // unscoped enum warning +#define mi_decl_noinline __declspec(noinline) +#define mi_decl_thread __declspec(thread) +#define mi_decl_cache_align __declspec(align(MI_CACHE_LINE)) +#elif (defined(__GNUC__) && (__GNUC__ >= 3)) || defined(__clang__) // includes clang and icc +#define mi_decl_noinline __attribute__((noinline)) +#define mi_decl_thread __thread +#define mi_decl_cache_align __attribute__((aligned(MI_CACHE_LINE))) +#else +#define mi_decl_noinline +#define mi_decl_thread __thread // hope for the best :-) +#define mi_decl_cache_align +#endif + +#if defined(__EMSCRIPTEN__) && !defined(__wasi__) +#define __wasi__ +#endif + +#if defined(__cplusplus) +#define mi_decl_externc extern "C" +#else +#define mi_decl_externc +#endif + +// pthreads +#if !defined(_WIN32) && !defined(__wasi__) +#define MI_USE_PTHREADS +#include +#endif + +// "options.c" +void _mi_fputs(mi_output_fun* out, void* arg, const char* prefix, const char* message); +void _mi_fprintf(mi_output_fun* out, void* arg, const char* fmt, ...); +void _mi_warning_message(const char* fmt, ...); +void _mi_verbose_message(const char* fmt, ...); +void _mi_trace_message(const char* fmt, ...); +void _mi_options_init(void); +void _mi_error_message(int err, const char* fmt, ...); + +// random.c +void _mi_random_init(mi_random_ctx_t* ctx); +void _mi_random_init_weak(mi_random_ctx_t* ctx); +void _mi_random_reinit_if_weak(mi_random_ctx_t * ctx); +void _mi_random_split(mi_random_ctx_t* ctx, mi_random_ctx_t* new_ctx); +uintptr_t _mi_random_next(mi_random_ctx_t* ctx); +uintptr_t _mi_heap_random_next(mi_heap_t* heap); +uintptr_t _mi_os_random_weak(uintptr_t extra_seed); +static inline uintptr_t _mi_random_shuffle(uintptr_t x); + +// init.c +extern mi_decl_cache_align mi_stats_t _mi_stats_main; +extern mi_decl_cache_align const mi_page_t _mi_page_empty; +bool _mi_is_main_thread(void); +size_t _mi_current_thread_count(void); +bool _mi_preloading(void); // true while the C runtime is not initialized yet +mi_threadid_t _mi_thread_id(void) mi_attr_noexcept; +mi_heap_t* _mi_heap_main_get(void); // statically allocated main backing heap +void _mi_thread_done(mi_heap_t* heap); +void _mi_thread_data_collect(void); + +// os.c +void _mi_os_init(void); // called from process init +void* _mi_os_alloc(size_t size, mi_memid_t* memid, mi_stats_t* stats); +void _mi_os_free(void* p, size_t size, mi_memid_t memid, mi_stats_t* stats); +void _mi_os_free_ex(void* p, size_t size, bool still_committed, mi_memid_t memid, mi_stats_t* stats); + +size_t _mi_os_page_size(void); +size_t _mi_os_good_alloc_size(size_t size); +bool _mi_os_has_overcommit(void); +bool _mi_os_has_virtual_reserve(void); + +bool _mi_os_purge(void* p, size_t size, mi_stats_t* stats); +bool _mi_os_reset(void* addr, size_t size, mi_stats_t* tld_stats); +bool _mi_os_commit(void* p, size_t size, bool* is_zero, mi_stats_t* stats); +bool _mi_os_decommit(void* addr, size_t size, mi_stats_t* stats); +bool _mi_os_protect(void* addr, size_t size); +bool _mi_os_unprotect(void* addr, size_t size); +bool _mi_os_purge(void* p, size_t size, mi_stats_t* stats); +bool _mi_os_purge_ex(void* p, size_t size, bool allow_reset, mi_stats_t* stats); + +void* _mi_os_alloc_aligned(size_t size, size_t alignment, bool commit, bool allow_large, mi_memid_t* memid, mi_stats_t* stats); +void* _mi_os_alloc_aligned_at_offset(size_t size, size_t alignment, size_t align_offset, bool commit, bool allow_large, mi_memid_t* memid, mi_stats_t* tld_stats); + +void* _mi_os_get_aligned_hint(size_t try_alignment, size_t size); +bool _mi_os_use_large_page(size_t size, size_t alignment); +size_t _mi_os_large_page_size(void); + +void* _mi_os_alloc_huge_os_pages(size_t pages, int numa_node, mi_msecs_t max_secs, size_t* pages_reserved, size_t* psize, mi_memid_t* memid); + +// arena.c +mi_arena_id_t _mi_arena_id_none(void); +void _mi_arena_free(void* p, size_t size, size_t still_committed_size, mi_memid_t memid, mi_stats_t* stats); +void* _mi_arena_alloc(size_t size, bool commit, bool allow_large, mi_arena_id_t req_arena_id, mi_memid_t* memid, mi_os_tld_t* tld); +void* _mi_arena_alloc_aligned(size_t size, size_t alignment, size_t align_offset, bool commit, bool allow_large, mi_arena_id_t req_arena_id, mi_memid_t* memid, mi_os_tld_t* tld); +bool _mi_arena_memid_is_suitable(mi_memid_t memid, mi_arena_id_t request_arena_id); +bool _mi_arena_contains(const void* p); +void _mi_arena_collect(bool force_purge, mi_stats_t* stats); +void _mi_arena_unsafe_destroy_all(mi_stats_t* stats); + +// "segment-map.c" +void _mi_segment_map_allocated_at(const mi_segment_t* segment); +void _mi_segment_map_freed_at(const mi_segment_t* segment); + +// "segment.c" +mi_page_t* _mi_segment_page_alloc(mi_heap_t* heap, size_t block_size, size_t page_alignment, mi_segments_tld_t* tld, mi_os_tld_t* os_tld); +void _mi_segment_page_free(mi_page_t* page, bool force, mi_segments_tld_t* tld); +void _mi_segment_page_abandon(mi_page_t* page, mi_segments_tld_t* tld); +bool _mi_segment_try_reclaim_abandoned( mi_heap_t* heap, bool try_all, mi_segments_tld_t* tld); +void _mi_segment_thread_collect(mi_segments_tld_t* tld); + +#if MI_HUGE_PAGE_ABANDON +void _mi_segment_huge_page_free(mi_segment_t* segment, mi_page_t* page, mi_block_t* block); +#else +void _mi_segment_huge_page_reset(mi_segment_t* segment, mi_page_t* page, mi_block_t* block); +#endif + +uint8_t* _mi_segment_page_start(const mi_segment_t* segment, const mi_page_t* page, size_t* page_size); // page start for any page +void _mi_abandoned_reclaim_all(mi_heap_t* heap, mi_segments_tld_t* tld); +void _mi_abandoned_await_readers(void); +void _mi_abandoned_collect(mi_heap_t* heap, bool force, mi_segments_tld_t* tld); + +// "page.c" +void* _mi_malloc_generic(mi_heap_t* heap, size_t size, bool zero, size_t huge_alignment) mi_attr_noexcept mi_attr_malloc; + +void _mi_page_retire(mi_page_t* page) mi_attr_noexcept; // free the page if there are no other pages with many free blocks +void _mi_page_unfull(mi_page_t* page); +void _mi_page_free(mi_page_t* page, mi_page_queue_t* pq, bool force); // free the page +void _mi_page_abandon(mi_page_t* page, mi_page_queue_t* pq); // abandon the page, to be picked up by another thread... +void _mi_heap_delayed_free_all(mi_heap_t* heap); +bool _mi_heap_delayed_free_partial(mi_heap_t* heap); +void _mi_heap_collect_retired(mi_heap_t* heap, bool force); + +void _mi_page_use_delayed_free(mi_page_t* page, mi_delayed_t delay, bool override_never); +bool _mi_page_try_use_delayed_free(mi_page_t* page, mi_delayed_t delay, bool override_never); +size_t _mi_page_queue_append(mi_heap_t* heap, mi_page_queue_t* pq, mi_page_queue_t* append); +void _mi_deferred_free(mi_heap_t* heap, bool force); + +void _mi_page_free_collect(mi_page_t* page,bool force); +void _mi_page_reclaim(mi_heap_t* heap, mi_page_t* page); // callback from segments + +size_t _mi_bin_size(uint8_t bin); // for stats +uint8_t _mi_bin(size_t size); // for stats + +// "heap.c" +void _mi_heap_destroy_pages(mi_heap_t* heap); +void _mi_heap_collect_abandon(mi_heap_t* heap); +void _mi_heap_set_default_direct(mi_heap_t* heap); +bool _mi_heap_memid_is_suitable(mi_heap_t* heap, mi_memid_t memid); +void _mi_heap_unsafe_destroy_all(void); + +// "stats.c" +void _mi_stats_done(mi_stats_t* stats); +mi_msecs_t _mi_clock_now(void); +mi_msecs_t _mi_clock_end(mi_msecs_t start); +mi_msecs_t _mi_clock_start(void); + +// "alloc.c" +void* _mi_page_malloc(mi_heap_t* heap, mi_page_t* page, size_t size, bool zero) mi_attr_noexcept; // called from `_mi_malloc_generic` +void* _mi_heap_malloc_zero(mi_heap_t* heap, size_t size, bool zero) mi_attr_noexcept; +void* _mi_heap_malloc_zero_ex(mi_heap_t* heap, size_t size, bool zero, size_t huge_alignment) mi_attr_noexcept; // called from `_mi_heap_malloc_aligned` +void* _mi_heap_realloc_zero(mi_heap_t* heap, void* p, size_t newsize, bool zero) mi_attr_noexcept; +mi_block_t* _mi_page_ptr_unalign(const mi_segment_t* segment, const mi_page_t* page, const void* p); +bool _mi_free_delayed_block(mi_block_t* block); +void _mi_free_generic(const mi_segment_t* segment, mi_page_t* page, bool is_local, void* p) mi_attr_noexcept; // for runtime integration +void _mi_padding_shrink(const mi_page_t* page, const mi_block_t* block, const size_t min_size); + +// option.c, c primitives +char _mi_toupper(char c); +int _mi_strnicmp(const char* s, const char* t, size_t n); +void _mi_strlcpy(char* dest, const char* src, size_t dest_size); +void _mi_strlcat(char* dest, const char* src, size_t dest_size); +size_t _mi_strlen(const char* s); +size_t _mi_strnlen(const char* s, size_t max_len); + + +#if MI_DEBUG>1 +bool _mi_page_is_valid(mi_page_t* page); +#endif + + +// ------------------------------------------------------ +// Branches +// ------------------------------------------------------ + +#if defined(__GNUC__) || defined(__clang__) +#define mi_unlikely(x) (__builtin_expect(!!(x),false)) +#define mi_likely(x) (__builtin_expect(!!(x),true)) +#elif (defined(__cplusplus) && (__cplusplus >= 202002L)) || (defined(_MSVC_LANG) && _MSVC_LANG >= 202002L) +#define mi_unlikely(x) (x) [[unlikely]] +#define mi_likely(x) (x) [[likely]] +#else +#define mi_unlikely(x) (x) +#define mi_likely(x) (x) +#endif + +#ifndef __has_builtin +#define __has_builtin(x) 0 +#endif + + +/* ----------------------------------------------------------- + Error codes passed to `_mi_fatal_error` + All are recoverable but EFAULT is a serious error and aborts by default in secure mode. + For portability define undefined error codes using common Unix codes: + +----------------------------------------------------------- */ +#include +#ifndef EAGAIN // double free +#define EAGAIN (11) +#endif +#ifndef ENOMEM // out of memory +#define ENOMEM (12) +#endif +#ifndef EFAULT // corrupted free-list or meta-data +#define EFAULT (14) +#endif +#ifndef EINVAL // trying to free an invalid pointer +#define EINVAL (22) +#endif +#ifndef EOVERFLOW // count*size overflow +#define EOVERFLOW (75) +#endif + + +/* ----------------------------------------------------------- + Inlined definitions +----------------------------------------------------------- */ +#define MI_UNUSED(x) (void)(x) +#if (MI_DEBUG>0) +#define MI_UNUSED_RELEASE(x) +#else +#define MI_UNUSED_RELEASE(x) MI_UNUSED(x) +#endif + +#define MI_INIT4(x) x(),x(),x(),x() +#define MI_INIT8(x) MI_INIT4(x),MI_INIT4(x) +#define MI_INIT16(x) MI_INIT8(x),MI_INIT8(x) +#define MI_INIT32(x) MI_INIT16(x),MI_INIT16(x) +#define MI_INIT64(x) MI_INIT32(x),MI_INIT32(x) +#define MI_INIT128(x) MI_INIT64(x),MI_INIT64(x) +#define MI_INIT256(x) MI_INIT128(x),MI_INIT128(x) + + +#include +// initialize a local variable to zero; use memset as compilers optimize constant sized memset's +#define _mi_memzero_var(x) memset(&x,0,sizeof(x)) + +// Is `x` a power of two? (0 is considered a power of two) +static inline bool _mi_is_power_of_two(uintptr_t x) { + return ((x & (x - 1)) == 0); +} + +// Is a pointer aligned? +static inline bool _mi_is_aligned(void* p, size_t alignment) { + mi_assert_internal(alignment != 0); + return (((uintptr_t)p % alignment) == 0); +} + +// Align upwards +static inline uintptr_t _mi_align_up(uintptr_t sz, size_t alignment) { + mi_assert_internal(alignment != 0); + uintptr_t mask = alignment - 1; + if ((alignment & mask) == 0) { // power of two? + return ((sz + mask) & ~mask); + } + else { + return (((sz + mask)/alignment)*alignment); + } +} + +// Align downwards +static inline uintptr_t _mi_align_down(uintptr_t sz, size_t alignment) { + mi_assert_internal(alignment != 0); + uintptr_t mask = alignment - 1; + if ((alignment & mask) == 0) { // power of two? + return (sz & ~mask); + } + else { + return ((sz / alignment) * alignment); + } +} + +// Divide upwards: `s <= _mi_divide_up(s,d)*d < s+d`. +static inline uintptr_t _mi_divide_up(uintptr_t size, size_t divider) { + mi_assert_internal(divider != 0); + return (divider == 0 ? size : ((size + divider - 1) / divider)); +} + +// Is memory zero initialized? +static inline bool mi_mem_is_zero(const void* p, size_t size) { + for (size_t i = 0; i < size; i++) { + if (((uint8_t*)p)[i] != 0) return false; + } + return true; +} + + +// Align a byte size to a size in _machine words_, +// i.e. byte size == `wsize*sizeof(void*)`. +static inline size_t _mi_wsize_from_size(size_t size) { + mi_assert_internal(size <= SIZE_MAX - sizeof(uintptr_t)); + return (size + sizeof(uintptr_t) - 1) / sizeof(uintptr_t); +} + +// Overflow detecting multiply +#if __has_builtin(__builtin_umul_overflow) || (defined(__GNUC__) && (__GNUC__ >= 5)) +#include // UINT_MAX, ULONG_MAX +#if defined(_CLOCK_T) // for Illumos +#undef _CLOCK_T +#endif +static inline bool mi_mul_overflow(size_t count, size_t size, size_t* total) { + #if (SIZE_MAX == ULONG_MAX) + return __builtin_umull_overflow(count, size, (unsigned long *)total); + #elif (SIZE_MAX == UINT_MAX) + return __builtin_umul_overflow(count, size, (unsigned int *)total); + #else + return __builtin_umulll_overflow(count, size, (unsigned long long *)total); + #endif +} +#else /* __builtin_umul_overflow is unavailable */ +static inline bool mi_mul_overflow(size_t count, size_t size, size_t* total) { + #define MI_MUL_NO_OVERFLOW ((size_t)1 << (4*sizeof(size_t))) // sqrt(SIZE_MAX) + *total = count * size; + // note: gcc/clang optimize this to directly check the overflow flag + return ((size >= MI_MUL_NO_OVERFLOW || count >= MI_MUL_NO_OVERFLOW) && size > 0 && (SIZE_MAX / size) < count); +} +#endif + +// Safe multiply `count*size` into `total`; return `true` on overflow. +static inline bool mi_count_size_overflow(size_t count, size_t size, size_t* total) { + if (count==1) { // quick check for the case where count is one (common for C++ allocators) + *total = size; + return false; + } + else if mi_unlikely(mi_mul_overflow(count, size, total)) { + #if MI_DEBUG > 0 + _mi_error_message(EOVERFLOW, "allocation request is too large (%zu * %zu bytes)\n", count, size); + #endif + *total = SIZE_MAX; + return true; + } + else return false; +} + + +/*---------------------------------------------------------------------------------------- + Heap functions +------------------------------------------------------------------------------------------- */ + +extern const mi_heap_t _mi_heap_empty; // read-only empty heap, initial value of the thread local default heap + +static inline bool mi_heap_is_backing(const mi_heap_t* heap) { + return (heap->tld->heap_backing == heap); +} + +static inline bool mi_heap_is_initialized(mi_heap_t* heap) { + mi_assert_internal(heap != NULL); + return (heap != &_mi_heap_empty); +} + +static inline uintptr_t _mi_ptr_cookie(const void* p) { + extern mi_heap_t _mi_heap_main; + mi_assert_internal(_mi_heap_main.cookie != 0); + return ((uintptr_t)p ^ _mi_heap_main.cookie); +} + +/* ----------------------------------------------------------- + Pages +----------------------------------------------------------- */ + +static inline mi_page_t* _mi_heap_get_free_small_page(mi_heap_t* heap, size_t size) { + mi_assert_internal(size <= (MI_SMALL_SIZE_MAX + MI_PADDING_SIZE)); + const size_t idx = _mi_wsize_from_size(size); + mi_assert_internal(idx < MI_PAGES_DIRECT); + return heap->pages_free_direct[idx]; +} + +// Segment that contains the pointer +// Large aligned blocks may be aligned at N*MI_SEGMENT_SIZE (inside a huge segment > MI_SEGMENT_SIZE), +// and we need align "down" to the segment info which is `MI_SEGMENT_SIZE` bytes before it; +// therefore we align one byte before `p`. +static inline mi_segment_t* _mi_ptr_segment(const void* p) { + mi_assert_internal(p != NULL); + return (mi_segment_t*)(((uintptr_t)p - 1) & ~MI_SEGMENT_MASK); +} + +static inline mi_page_t* mi_slice_to_page(mi_slice_t* s) { + mi_assert_internal(s->slice_offset== 0 && s->slice_count > 0); + return (mi_page_t*)(s); +} + +static inline mi_slice_t* mi_page_to_slice(mi_page_t* p) { + mi_assert_internal(p->slice_offset== 0 && p->slice_count > 0); + return (mi_slice_t*)(p); +} + +// Segment belonging to a page +static inline mi_segment_t* _mi_page_segment(const mi_page_t* page) { + mi_segment_t* segment = _mi_ptr_segment(page); + mi_assert_internal(segment == NULL || ((mi_slice_t*)page >= segment->slices && (mi_slice_t*)page < segment->slices + segment->slice_entries)); + return segment; +} + +static inline mi_slice_t* mi_slice_first(const mi_slice_t* slice) { + mi_slice_t* start = (mi_slice_t*)((uint8_t*)slice - slice->slice_offset); + mi_assert_internal(start >= _mi_ptr_segment(slice)->slices); + mi_assert_internal(start->slice_offset == 0); + mi_assert_internal(start + start->slice_count > slice); + return start; +} + +// Get the page containing the pointer (performance critical as it is called in mi_free) +static inline mi_page_t* _mi_segment_page_of(const mi_segment_t* segment, const void* p) { + mi_assert_internal(p > (void*)segment); + ptrdiff_t diff = (uint8_t*)p - (uint8_t*)segment; + mi_assert_internal(diff > 0 && diff <= (ptrdiff_t)MI_SEGMENT_SIZE); + size_t idx = (size_t)diff >> MI_SEGMENT_SLICE_SHIFT; + mi_assert_internal(idx <= segment->slice_entries); + mi_slice_t* slice0 = (mi_slice_t*)&segment->slices[idx]; + mi_slice_t* slice = mi_slice_first(slice0); // adjust to the block that holds the page data + mi_assert_internal(slice->slice_offset == 0); + mi_assert_internal(slice >= segment->slices && slice < segment->slices + segment->slice_entries); + return mi_slice_to_page(slice); +} + +// Quick page start for initialized pages +static inline uint8_t* _mi_page_start(const mi_segment_t* segment, const mi_page_t* page, size_t* page_size) { + return _mi_segment_page_start(segment, page, page_size); +} + +// Get the page containing the pointer +static inline mi_page_t* _mi_ptr_page(void* p) { + return _mi_segment_page_of(_mi_ptr_segment(p), p); +} + +// Get the block size of a page (special case for huge objects) +static inline size_t mi_page_block_size(const mi_page_t* page) { + const size_t bsize = page->xblock_size; + mi_assert_internal(bsize > 0); + if mi_likely(bsize < MI_HUGE_BLOCK_SIZE) { + return bsize; + } + else { + size_t psize; + _mi_segment_page_start(_mi_page_segment(page), page, &psize); + return psize; + } +} + +static inline bool mi_page_is_huge(const mi_page_t* page) { + return (_mi_page_segment(page)->kind == MI_SEGMENT_HUGE); +} + +// Get the usable block size of a page without fixed padding. +// This may still include internal padding due to alignment and rounding up size classes. +static inline size_t mi_page_usable_block_size(const mi_page_t* page) { + return mi_page_block_size(page) - MI_PADDING_SIZE; +} + +// size of a segment +static inline size_t mi_segment_size(mi_segment_t* segment) { + return segment->segment_slices * MI_SEGMENT_SLICE_SIZE; +} + +static inline uint8_t* mi_segment_end(mi_segment_t* segment) { + return (uint8_t*)segment + mi_segment_size(segment); +} + +// Thread free access +static inline mi_block_t* mi_page_thread_free(const mi_page_t* page) { + return (mi_block_t*)(mi_atomic_load_relaxed(&((mi_page_t*)page)->xthread_free) & ~3); +} + +static inline mi_delayed_t mi_page_thread_free_flag(const mi_page_t* page) { + return (mi_delayed_t)(mi_atomic_load_relaxed(&((mi_page_t*)page)->xthread_free) & 3); +} + +// Heap access +static inline mi_heap_t* mi_page_heap(const mi_page_t* page) { + return (mi_heap_t*)(mi_atomic_load_relaxed(&((mi_page_t*)page)->xheap)); +} + +static inline void mi_page_set_heap(mi_page_t* page, mi_heap_t* heap) { + mi_assert_internal(mi_page_thread_free_flag(page) != MI_DELAYED_FREEING); + mi_atomic_store_release(&page->xheap,(uintptr_t)heap); +} + +// Thread free flag helpers +static inline mi_block_t* mi_tf_block(mi_thread_free_t tf) { + return (mi_block_t*)(tf & ~0x03); +} +static inline mi_delayed_t mi_tf_delayed(mi_thread_free_t tf) { + return (mi_delayed_t)(tf & 0x03); +} +static inline mi_thread_free_t mi_tf_make(mi_block_t* block, mi_delayed_t delayed) { + return (mi_thread_free_t)((uintptr_t)block | (uintptr_t)delayed); +} +static inline mi_thread_free_t mi_tf_set_delayed(mi_thread_free_t tf, mi_delayed_t delayed) { + return mi_tf_make(mi_tf_block(tf),delayed); +} +static inline mi_thread_free_t mi_tf_set_block(mi_thread_free_t tf, mi_block_t* block) { + return mi_tf_make(block, mi_tf_delayed(tf)); +} + +// are all blocks in a page freed? +// note: needs up-to-date used count, (as the `xthread_free` list may not be empty). see `_mi_page_collect_free`. +static inline bool mi_page_all_free(const mi_page_t* page) { + mi_assert_internal(page != NULL); + return (page->used == 0); +} + +// are there any available blocks? +static inline bool mi_page_has_any_available(const mi_page_t* page) { + mi_assert_internal(page != NULL && page->reserved > 0); + return (page->used < page->reserved || (mi_page_thread_free(page) != NULL)); +} + +// are there immediately available blocks, i.e. blocks available on the free list. +static inline bool mi_page_immediate_available(const mi_page_t* page) { + mi_assert_internal(page != NULL); + return (page->free != NULL); +} + +// is more than 7/8th of a page in use? +static inline bool mi_page_mostly_used(const mi_page_t* page) { + if (page==NULL) return true; + uint16_t frac = page->reserved / 8U; + return (page->reserved - page->used <= frac); +} + +static inline mi_page_queue_t* mi_page_queue(const mi_heap_t* heap, size_t size) { + return &((mi_heap_t*)heap)->pages[_mi_bin(size)]; +} + + + +//----------------------------------------------------------- +// Page flags +//----------------------------------------------------------- +static inline bool mi_page_is_in_full(const mi_page_t* page) { + return page->flags.x.in_full; +} + +static inline void mi_page_set_in_full(mi_page_t* page, bool in_full) { + page->flags.x.in_full = in_full; +} + +static inline bool mi_page_has_aligned(const mi_page_t* page) { + return page->flags.x.has_aligned; +} + +static inline void mi_page_set_has_aligned(mi_page_t* page, bool has_aligned) { + page->flags.x.has_aligned = has_aligned; +} + + +/* ------------------------------------------------------------------- +Encoding/Decoding the free list next pointers + +This is to protect against buffer overflow exploits where the +free list is mutated. Many hardened allocators xor the next pointer `p` +with a secret key `k1`, as `p^k1`. This prevents overwriting with known +values but might be still too weak: if the attacker can guess +the pointer `p` this can reveal `k1` (since `p^k1^p == k1`). +Moreover, if multiple blocks can be read as well, the attacker can +xor both as `(p1^k1) ^ (p2^k1) == p1^p2` which may reveal a lot +about the pointers (and subsequently `k1`). + +Instead mimalloc uses an extra key `k2` and encodes as `((p^k2)<<> (MI_INTPTR_BITS - shift)))); +} +static inline uintptr_t mi_rotr(uintptr_t x, uintptr_t shift) { + shift %= MI_INTPTR_BITS; + return (shift==0 ? x : ((x >> shift) | (x << (MI_INTPTR_BITS - shift)))); +} + +static inline void* mi_ptr_decode(const void* null, const mi_encoded_t x, const uintptr_t* keys) { + void* p = (void*)(mi_rotr(x - keys[0], keys[0]) ^ keys[1]); + return (p==null ? NULL : p); +} + +static inline mi_encoded_t mi_ptr_encode(const void* null, const void* p, const uintptr_t* keys) { + uintptr_t x = (uintptr_t)(p==NULL ? null : p); + return mi_rotl(x ^ keys[1], keys[0]) + keys[0]; +} + +static inline mi_block_t* mi_block_nextx( const void* null, const mi_block_t* block, const uintptr_t* keys ) { + mi_track_mem_defined(block,sizeof(mi_block_t)); + mi_block_t* next; + #ifdef MI_ENCODE_FREELIST + next = (mi_block_t*)mi_ptr_decode(null, block->next, keys); + #else + MI_UNUSED(keys); MI_UNUSED(null); + next = (mi_block_t*)block->next; + #endif + mi_track_mem_noaccess(block,sizeof(mi_block_t)); + return next; +} + +static inline void mi_block_set_nextx(const void* null, mi_block_t* block, const mi_block_t* next, const uintptr_t* keys) { + mi_track_mem_undefined(block,sizeof(mi_block_t)); + #ifdef MI_ENCODE_FREELIST + block->next = mi_ptr_encode(null, next, keys); + #else + MI_UNUSED(keys); MI_UNUSED(null); + block->next = (mi_encoded_t)next; + #endif + mi_track_mem_noaccess(block,sizeof(mi_block_t)); +} + +static inline mi_block_t* mi_block_next(const mi_page_t* page, const mi_block_t* block) { + #ifdef MI_ENCODE_FREELIST + mi_block_t* next = mi_block_nextx(page,block,page->keys); + // check for free list corruption: is `next` at least in the same page? + // TODO: check if `next` is `page->block_size` aligned? + if mi_unlikely(next!=NULL && !mi_is_in_same_page(block, next)) { + _mi_error_message(EFAULT, "corrupted free list entry of size %zub at %p: value 0x%zx\n", mi_page_block_size(page), block, (uintptr_t)next); + next = NULL; + } + return next; + #else + MI_UNUSED(page); + return mi_block_nextx(page,block,NULL); + #endif +} + +static inline void mi_block_set_next(const mi_page_t* page, mi_block_t* block, const mi_block_t* next) { + #ifdef MI_ENCODE_FREELIST + mi_block_set_nextx(page,block,next, page->keys); + #else + MI_UNUSED(page); + mi_block_set_nextx(page,block,next,NULL); + #endif +} + + +// ------------------------------------------------------------------- +// commit mask +// ------------------------------------------------------------------- + +static inline void mi_commit_mask_create_empty(mi_commit_mask_t* cm) { + for (size_t i = 0; i < MI_COMMIT_MASK_FIELD_COUNT; i++) { + cm->mask[i] = 0; + } +} + +static inline void mi_commit_mask_create_full(mi_commit_mask_t* cm) { + for (size_t i = 0; i < MI_COMMIT_MASK_FIELD_COUNT; i++) { + cm->mask[i] = ~((size_t)0); + } +} + +static inline bool mi_commit_mask_is_empty(const mi_commit_mask_t* cm) { + for (size_t i = 0; i < MI_COMMIT_MASK_FIELD_COUNT; i++) { + if (cm->mask[i] != 0) return false; + } + return true; +} + +static inline bool mi_commit_mask_is_full(const mi_commit_mask_t* cm) { + for (size_t i = 0; i < MI_COMMIT_MASK_FIELD_COUNT; i++) { + if (cm->mask[i] != ~((size_t)0)) return false; + } + return true; +} + +// defined in `segment.c`: +size_t _mi_commit_mask_committed_size(const mi_commit_mask_t* cm, size_t total); +size_t _mi_commit_mask_next_run(const mi_commit_mask_t* cm, size_t* idx); + +#define mi_commit_mask_foreach(cm,idx,count) \ + idx = 0; \ + while ((count = _mi_commit_mask_next_run(cm,&idx)) > 0) { + +#define mi_commit_mask_foreach_end() \ + idx += count; \ + } + + + +/* ----------------------------------------------------------- + memory id's +----------------------------------------------------------- */ + +static inline mi_memid_t _mi_memid_create(mi_memkind_t memkind) { + mi_memid_t memid; + _mi_memzero_var(memid); + memid.memkind = memkind; + return memid; +} + +static inline mi_memid_t _mi_memid_none(void) { + return _mi_memid_create(MI_MEM_NONE); +} + +static inline mi_memid_t _mi_memid_create_os(bool committed, bool is_zero, bool is_large) { + mi_memid_t memid = _mi_memid_create(MI_MEM_OS); + memid.initially_committed = committed; + memid.initially_zero = is_zero; + memid.is_pinned = is_large; + return memid; +} + + +// ------------------------------------------------------------------- +// Fast "random" shuffle +// ------------------------------------------------------------------- + +static inline uintptr_t _mi_random_shuffle(uintptr_t x) { + if (x==0) { x = 17; } // ensure we don't get stuck in generating zeros +#if (MI_INTPTR_SIZE==8) + // by Sebastiano Vigna, see: + x ^= x >> 30; + x *= 0xbf58476d1ce4e5b9UL; + x ^= x >> 27; + x *= 0x94d049bb133111ebUL; + x ^= x >> 31; +#elif (MI_INTPTR_SIZE==4) + // by Chris Wellons, see: + x ^= x >> 16; + x *= 0x7feb352dUL; + x ^= x >> 15; + x *= 0x846ca68bUL; + x ^= x >> 16; +#endif + return x; +} + +// ------------------------------------------------------------------- +// Optimize numa node access for the common case (= one node) +// ------------------------------------------------------------------- + +int _mi_os_numa_node_get(mi_os_tld_t* tld); +size_t _mi_os_numa_node_count_get(void); + +extern _Atomic(size_t) _mi_numa_node_count; +static inline int _mi_os_numa_node(mi_os_tld_t* tld) { + if mi_likely(mi_atomic_load_relaxed(&_mi_numa_node_count) == 1) { return 0; } + else return _mi_os_numa_node_get(tld); +} +static inline size_t _mi_os_numa_node_count(void) { + const size_t count = mi_atomic_load_relaxed(&_mi_numa_node_count); + if mi_likely(count > 0) { return count; } + else return _mi_os_numa_node_count_get(); +} + + + +// ----------------------------------------------------------------------- +// Count bits: trailing or leading zeros (with MI_INTPTR_BITS on all zero) +// ----------------------------------------------------------------------- + +#if defined(__GNUC__) + +#include // LONG_MAX +#define MI_HAVE_FAST_BITSCAN +static inline size_t mi_clz(uintptr_t x) { + if (x==0) return MI_INTPTR_BITS; +#if (INTPTR_MAX == LONG_MAX) + return __builtin_clzl(x); +#else + return __builtin_clzll(x); +#endif +} +static inline size_t mi_ctz(uintptr_t x) { + if (x==0) return MI_INTPTR_BITS; +#if (INTPTR_MAX == LONG_MAX) + return __builtin_ctzl(x); +#else + return __builtin_ctzll(x); +#endif +} + +#elif defined(_MSC_VER) + +#include // LONG_MAX +#include // BitScanReverse64 +#define MI_HAVE_FAST_BITSCAN +static inline size_t mi_clz(uintptr_t x) { + if (x==0) return MI_INTPTR_BITS; + unsigned long idx; +#if (INTPTR_MAX == LONG_MAX) + _BitScanReverse(&idx, x); +#else + _BitScanReverse64(&idx, x); +#endif + return ((MI_INTPTR_BITS - 1) - idx); +} +static inline size_t mi_ctz(uintptr_t x) { + if (x==0) return MI_INTPTR_BITS; + unsigned long idx; +#if (INTPTR_MAX == LONG_MAX) + _BitScanForward(&idx, x); +#else + _BitScanForward64(&idx, x); +#endif + return idx; +} + +#else +static inline size_t mi_ctz32(uint32_t x) { + // de Bruijn multiplication, see + static const unsigned char debruijn[32] = { + 0, 1, 28, 2, 29, 14, 24, 3, 30, 22, 20, 15, 25, 17, 4, 8, + 31, 27, 13, 23, 21, 19, 16, 7, 26, 12, 18, 6, 11, 5, 10, 9 + }; + if (x==0) return 32; + return debruijn[((x & -(int32_t)x) * 0x077CB531UL) >> 27]; +} +static inline size_t mi_clz32(uint32_t x) { + // de Bruijn multiplication, see + static const uint8_t debruijn[32] = { + 31, 22, 30, 21, 18, 10, 29, 2, 20, 17, 15, 13, 9, 6, 28, 1, + 23, 19, 11, 3, 16, 14, 7, 24, 12, 4, 8, 25, 5, 26, 27, 0 + }; + if (x==0) return 32; + x |= x >> 1; + x |= x >> 2; + x |= x >> 4; + x |= x >> 8; + x |= x >> 16; + return debruijn[(uint32_t)(x * 0x07C4ACDDUL) >> 27]; +} + +static inline size_t mi_clz(uintptr_t x) { + if (x==0) return MI_INTPTR_BITS; +#if (MI_INTPTR_BITS <= 32) + return mi_clz32((uint32_t)x); +#else + size_t count = mi_clz32((uint32_t)(x >> 32)); + if (count < 32) return count; + return (32 + mi_clz32((uint32_t)x)); +#endif +} +static inline size_t mi_ctz(uintptr_t x) { + if (x==0) return MI_INTPTR_BITS; +#if (MI_INTPTR_BITS <= 32) + return mi_ctz32((uint32_t)x); +#else + size_t count = mi_ctz32((uint32_t)x); + if (count < 32) return count; + return (32 + mi_ctz32((uint32_t)(x>>32))); +#endif +} + +#endif + +// "bit scan reverse": Return index of the highest bit (or MI_INTPTR_BITS if `x` is zero) +static inline size_t mi_bsr(uintptr_t x) { + return (x==0 ? MI_INTPTR_BITS : MI_INTPTR_BITS - 1 - mi_clz(x)); +} + + +// --------------------------------------------------------------------------------- +// Provide our own `_mi_memcpy` for potential performance optimizations. +// +// For now, only on Windows with msvc/clang-cl we optimize to `rep movsb` if +// we happen to run on x86/x64 cpu's that have "fast short rep movsb" (FSRM) support +// (AMD Zen3+ (~2020) or Intel Ice Lake+ (~2017). See also issue #201 and pr #253. +// --------------------------------------------------------------------------------- + +#if !MI_TRACK_ENABLED && defined(_WIN32) && (defined(_M_IX86) || defined(_M_X64)) +#include +extern bool _mi_cpu_has_fsrm; +static inline void _mi_memcpy(void* dst, const void* src, size_t n) { + if (_mi_cpu_has_fsrm) { + __movsb((unsigned char*)dst, (const unsigned char*)src, n); + } + else { + memcpy(dst, src, n); + } +} +static inline void _mi_memzero(void* dst, size_t n) { + if (_mi_cpu_has_fsrm) { + __stosb((unsigned char*)dst, 0, n); + } + else { + memset(dst, 0, n); + } +} +#else +static inline void _mi_memcpy(void* dst, const void* src, size_t n) { + memcpy(dst, src, n); +} +static inline void _mi_memzero(void* dst, size_t n) { + memset(dst, 0, n); +} +#endif + +// ------------------------------------------------------------------------------- +// The `_mi_memcpy_aligned` can be used if the pointers are machine-word aligned +// This is used for example in `mi_realloc`. +// ------------------------------------------------------------------------------- + +#if (defined(__GNUC__) && (__GNUC__ >= 4)) || defined(__clang__) +// On GCC/CLang we provide a hint that the pointers are word aligned. +static inline void _mi_memcpy_aligned(void* dst, const void* src, size_t n) { + mi_assert_internal(((uintptr_t)dst % MI_INTPTR_SIZE == 0) && ((uintptr_t)src % MI_INTPTR_SIZE == 0)); + void* adst = __builtin_assume_aligned(dst, MI_INTPTR_SIZE); + const void* asrc = __builtin_assume_aligned(src, MI_INTPTR_SIZE); + _mi_memcpy(adst, asrc, n); +} + +static inline void _mi_memzero_aligned(void* dst, size_t n) { + mi_assert_internal((uintptr_t)dst % MI_INTPTR_SIZE == 0); + void* adst = __builtin_assume_aligned(dst, MI_INTPTR_SIZE); + _mi_memzero(adst, n); +} +#else +// Default fallback on `_mi_memcpy` +static inline void _mi_memcpy_aligned(void* dst, const void* src, size_t n) { + mi_assert_internal(((uintptr_t)dst % MI_INTPTR_SIZE == 0) && ((uintptr_t)src % MI_INTPTR_SIZE == 0)); + _mi_memcpy(dst, src, n); +} + +static inline void _mi_memzero_aligned(void* dst, size_t n) { + mi_assert_internal((uintptr_t)dst % MI_INTPTR_SIZE == 0); + _mi_memzero(dst, n); +} +#endif + + +#endif diff --git a/compat/mimalloc/mimalloc/prim.h b/compat/mimalloc/mimalloc/prim.h new file mode 100644 index 00000000000000..1e55cb5f8802d7 --- /dev/null +++ b/compat/mimalloc/mimalloc/prim.h @@ -0,0 +1,323 @@ +/* ---------------------------------------------------------------------------- +Copyright (c) 2018-2023, Microsoft Research, Daan Leijen +This is free software; you can redistribute it and/or modify it under the +terms of the MIT license. A copy of the license can be found in the file +"LICENSE" at the root of this distribution. +-----------------------------------------------------------------------------*/ +#pragma once +#ifndef MIMALLOC_PRIM_H +#define MIMALLOC_PRIM_H + + +// -------------------------------------------------------------------------- +// This file specifies the primitive portability API. +// Each OS/host needs to implement these primitives, see `src/prim` +// for implementations on Window, macOS, WASI, and Linux/Unix. +// +// note: on all primitive functions, we always have result parameters != NUL, and: +// addr != NULL and page aligned +// size > 0 and page aligned +// return value is an error code an int where 0 is success. +// -------------------------------------------------------------------------- + +// OS memory configuration +typedef struct mi_os_mem_config_s { + size_t page_size; // 4KiB + size_t large_page_size; // 2MiB + size_t alloc_granularity; // smallest allocation size (on Windows 64KiB) + bool has_overcommit; // can we reserve more memory than can be actually committed? + bool must_free_whole; // must allocated blocks be freed as a whole (false for mmap, true for VirtualAlloc) + bool has_virtual_reserve; // supports virtual address space reservation? (if true we can reserve virtual address space without using commit or physical memory) +} mi_os_mem_config_t; + +// Initialize +void _mi_prim_mem_init( mi_os_mem_config_t* config ); + +// Free OS memory +int _mi_prim_free(void* addr, size_t size ); + +// Allocate OS memory. Return NULL on error. +// The `try_alignment` is just a hint and the returned pointer does not have to be aligned. +// If `commit` is false, the virtual memory range only needs to be reserved (with no access) +// which will later be committed explicitly using `_mi_prim_commit`. +// `is_zero` is set to true if the memory was zero initialized (as on most OS's) +// pre: !commit => !allow_large +// try_alignment >= _mi_os_page_size() and a power of 2 +int _mi_prim_alloc(size_t size, size_t try_alignment, bool commit, bool allow_large, bool* is_large, bool* is_zero, void** addr); + +// Commit memory. Returns error code or 0 on success. +// For example, on Linux this would make the memory PROT_READ|PROT_WRITE. +// `is_zero` is set to true if the memory was zero initialized (e.g. on Windows) +int _mi_prim_commit(void* addr, size_t size, bool* is_zero); + +// Decommit memory. Returns error code or 0 on success. The `needs_recommit` result is true +// if the memory would need to be re-committed. For example, on Windows this is always true, +// but on Linux we could use MADV_DONTNEED to decommit which does not need a recommit. +// pre: needs_recommit != NULL +int _mi_prim_decommit(void* addr, size_t size, bool* needs_recommit); + +// Reset memory. The range keeps being accessible but the content might be reset. +// Returns error code or 0 on success. +int _mi_prim_reset(void* addr, size_t size); + +// Protect memory. Returns error code or 0 on success. +int _mi_prim_protect(void* addr, size_t size, bool protect); + +// Allocate huge (1GiB) pages possibly associated with a NUMA node. +// `is_zero` is set to true if the memory was zero initialized (as on most OS's) +// pre: size > 0 and a multiple of 1GiB. +// numa_node is either negative (don't care), or a numa node number. +int _mi_prim_alloc_huge_os_pages(void* hint_addr, size_t size, int numa_node, bool* is_zero, void** addr); + +// Return the current NUMA node +size_t _mi_prim_numa_node(void); + +// Return the number of logical NUMA nodes +size_t _mi_prim_numa_node_count(void); + +// Clock ticks +mi_msecs_t _mi_prim_clock_now(void); + +// Return process information (only for statistics) +typedef struct mi_process_info_s { + mi_msecs_t elapsed; + mi_msecs_t utime; + mi_msecs_t stime; + size_t current_rss; + size_t peak_rss; + size_t current_commit; + size_t peak_commit; + size_t page_faults; +} mi_process_info_t; + +void _mi_prim_process_info(mi_process_info_t* pinfo); + +// Default stderr output. (only for warnings etc. with verbose enabled) +// msg != NULL && _mi_strlen(msg) > 0 +void _mi_prim_out_stderr( const char* msg ); + +// Get an environment variable. (only for options) +// name != NULL, result != NULL, result_size >= 64 +bool _mi_prim_getenv(const char* name, char* result, size_t result_size); + + +// Fill a buffer with strong randomness; return `false` on error or if +// there is no strong randomization available. +bool _mi_prim_random_buf(void* buf, size_t buf_len); + +// Called on the first thread start, and should ensure `_mi_thread_done` is called on thread termination. +void _mi_prim_thread_init_auto_done(void); + +// Called on process exit and may take action to clean up resources associated with the thread auto done. +void _mi_prim_thread_done_auto_done(void); + +// Called when the default heap for a thread changes +void _mi_prim_thread_associate_default_heap(mi_heap_t* heap); + + +//------------------------------------------------------------------- +// Thread id: `_mi_prim_thread_id()` +// +// Getting the thread id should be performant as it is called in the +// fast path of `_mi_free` and we specialize for various platforms as +// inlined definitions. Regular code should call `init.c:_mi_thread_id()`. +// We only require _mi_prim_thread_id() to return a unique id +// for each thread (unequal to zero). +//------------------------------------------------------------------- + +// defined in `init.c`; do not use these directly +extern mi_decl_thread mi_heap_t* _mi_heap_default; // default heap to allocate from +extern bool _mi_process_is_initialized; // has mi_process_init been called? + +static inline mi_threadid_t _mi_prim_thread_id(void) mi_attr_noexcept; + +#if defined(_WIN32) + +#define WIN32_LEAN_AND_MEAN +#include +static inline mi_threadid_t _mi_prim_thread_id(void) mi_attr_noexcept { + // Windows: works on Intel and ARM in both 32- and 64-bit + return (uintptr_t)NtCurrentTeb(); +} + +// We use assembly for a fast thread id on the main platforms. The TLS layout depends on +// both the OS and libc implementation so we use specific tests for each main platform. +// If you test on another platform and it works please send a PR :-) +// see also https://akkadia.org/drepper/tls.pdf for more info on the TLS register. +#elif defined(__GNUC__) && ( \ + (defined(__GLIBC__) && (defined(__x86_64__) || defined(__i386__) || defined(__arm__) || defined(__aarch64__))) \ + || (defined(__APPLE__) && (defined(__x86_64__) || defined(__aarch64__))) \ + || (defined(__BIONIC__) && (defined(__x86_64__) || defined(__i386__) || defined(__arm__) || defined(__aarch64__))) \ + || (defined(__FreeBSD__) && (defined(__x86_64__) || defined(__i386__) || defined(__aarch64__))) \ + || (defined(__OpenBSD__) && (defined(__x86_64__) || defined(__i386__) || defined(__aarch64__))) \ + ) + +static inline void* mi_prim_tls_slot(size_t slot) mi_attr_noexcept { + void* res; + const size_t ofs = (slot*sizeof(void*)); + #if defined(__i386__) + __asm__("movl %%gs:%1, %0" : "=r" (res) : "m" (*((void**)ofs)) : ); // x86 32-bit always uses GS + #elif defined(__APPLE__) && defined(__x86_64__) + __asm__("movq %%gs:%1, %0" : "=r" (res) : "m" (*((void**)ofs)) : ); // x86_64 macOSX uses GS + #elif defined(__x86_64__) && (MI_INTPTR_SIZE==4) + __asm__("movl %%fs:%1, %0" : "=r" (res) : "m" (*((void**)ofs)) : ); // x32 ABI + #elif defined(__x86_64__) + __asm__("movq %%fs:%1, %0" : "=r" (res) : "m" (*((void**)ofs)) : ); // x86_64 Linux, BSD uses FS + #elif defined(__arm__) + void** tcb; MI_UNUSED(ofs); + __asm__ volatile ("mrc p15, 0, %0, c13, c0, 3\nbic %0, %0, #3" : "=r" (tcb)); + res = tcb[slot]; + #elif defined(__aarch64__) + void** tcb; MI_UNUSED(ofs); + #if defined(__APPLE__) // M1, issue #343 + __asm__ volatile ("mrs %0, tpidrro_el0\nbic %0, %0, #7" : "=r" (tcb)); + #else + __asm__ volatile ("mrs %0, tpidr_el0" : "=r" (tcb)); + #endif + res = tcb[slot]; + #endif + return res; +} + +// setting a tls slot is only used on macOS for now +static inline void mi_prim_tls_slot_set(size_t slot, void* value) mi_attr_noexcept { + const size_t ofs = (slot*sizeof(void*)); + #if defined(__i386__) + __asm__("movl %1,%%gs:%0" : "=m" (*((void**)ofs)) : "rn" (value) : ); // 32-bit always uses GS + #elif defined(__APPLE__) && defined(__x86_64__) + __asm__("movq %1,%%gs:%0" : "=m" (*((void**)ofs)) : "rn" (value) : ); // x86_64 macOS uses GS + #elif defined(__x86_64__) && (MI_INTPTR_SIZE==4) + __asm__("movl %1,%%fs:%0" : "=m" (*((void**)ofs)) : "rn" (value) : ); // x32 ABI + #elif defined(__x86_64__) + __asm__("movq %1,%%fs:%0" : "=m" (*((void**)ofs)) : "rn" (value) : ); // x86_64 Linux, BSD uses FS + #elif defined(__arm__) + void** tcb; MI_UNUSED(ofs); + __asm__ volatile ("mrc p15, 0, %0, c13, c0, 3\nbic %0, %0, #3" : "=r" (tcb)); + tcb[slot] = value; + #elif defined(__aarch64__) + void** tcb; MI_UNUSED(ofs); + #if defined(__APPLE__) // M1, issue #343 + __asm__ volatile ("mrs %0, tpidrro_el0\nbic %0, %0, #7" : "=r" (tcb)); + #else + __asm__ volatile ("mrs %0, tpidr_el0" : "=r" (tcb)); + #endif + tcb[slot] = value; + #endif +} + +static inline mi_threadid_t _mi_prim_thread_id(void) mi_attr_noexcept { + #if defined(__BIONIC__) + // issue #384, #495: on the Bionic libc (Android), slot 1 is the thread id + // see: https://github.com/aosp-mirror/platform_bionic/blob/c44b1d0676ded732df4b3b21c5f798eacae93228/libc/platform/bionic/tls_defines.h#L86 + return (uintptr_t)mi_prim_tls_slot(1); + #else + // in all our other targets, slot 0 is the thread id + // glibc: https://sourceware.org/git/?p=glibc.git;a=blob_plain;f=sysdeps/x86_64/nptl/tls.h + // apple: https://github.com/apple/darwin-xnu/blob/main/libsyscall/os/tsd.h#L36 + return (uintptr_t)mi_prim_tls_slot(0); + #endif +} + +#else + +// otherwise use portable C, taking the address of a thread local variable (this is still very fast on most platforms). +static inline mi_threadid_t _mi_prim_thread_id(void) mi_attr_noexcept { + return (uintptr_t)&_mi_heap_default; +} + +#endif + + + +/* ---------------------------------------------------------------------------------------- +The thread local default heap: `_mi_prim_get_default_heap()` +This is inlined here as it is on the fast path for allocation functions. + +On most platforms (Windows, Linux, FreeBSD, NetBSD, etc), this just returns a +__thread local variable (`_mi_heap_default`). With the initial-exec TLS model this ensures +that the storage will always be available (allocated on the thread stacks). + +On some platforms though we cannot use that when overriding `malloc` since the underlying +TLS implementation (or the loader) will call itself `malloc` on a first access and recurse. +We try to circumvent this in an efficient way: +- macOSX : we use an unused TLS slot from the OS allocated slots (MI_TLS_SLOT). On OSX, the + loader itself calls `malloc` even before the modules are initialized. +- OpenBSD: we use an unused slot from the pthread block (MI_TLS_PTHREAD_SLOT_OFS). +- DragonFly: defaults are working but seem slow compared to freeBSD (see PR #323) +------------------------------------------------------------------------------------------- */ + +static inline mi_heap_t* mi_prim_get_default_heap(void); + +#if defined(MI_MALLOC_OVERRIDE) +#if defined(__APPLE__) // macOS + #define MI_TLS_SLOT 89 // seems unused? + // #define MI_TLS_RECURSE_GUARD 1 + // other possible unused ones are 9, 29, __PTK_FRAMEWORK_JAVASCRIPTCORE_KEY4 (94), __PTK_FRAMEWORK_GC_KEY9 (112) and __PTK_FRAMEWORK_OLDGC_KEY9 (89) + // see +#elif defined(__OpenBSD__) + // use end bytes of a name; goes wrong if anyone uses names > 23 characters (ptrhread specifies 16) + // see + #define MI_TLS_PTHREAD_SLOT_OFS (6*sizeof(int) + 4*sizeof(void*) + 24) + // #elif defined(__DragonFly__) + // #warning "mimalloc is not working correctly on DragonFly yet." + // #define MI_TLS_PTHREAD_SLOT_OFS (4 + 1*sizeof(void*)) // offset `uniqueid` (also used by gdb?) +#elif defined(__ANDROID__) + // See issue #381 + #define MI_TLS_PTHREAD +#endif +#endif + + +#if defined(MI_TLS_SLOT) + +static inline mi_heap_t* mi_prim_get_default_heap(void) { + mi_heap_t* heap = (mi_heap_t*)mi_prim_tls_slot(MI_TLS_SLOT); + if mi_unlikely(heap == NULL) { + #ifdef __GNUC__ + __asm(""); // prevent conditional load of the address of _mi_heap_empty + #endif + heap = (mi_heap_t*)&_mi_heap_empty; + } + return heap; +} + +#elif defined(MI_TLS_PTHREAD_SLOT_OFS) + +static inline mi_heap_t** mi_prim_tls_pthread_heap_slot(void) { + pthread_t self = pthread_self(); + #if defined(__DragonFly__) + if (self==NULL) return NULL; + #endif + return (mi_heap_t**)((uint8_t*)self + MI_TLS_PTHREAD_SLOT_OFS); +} + +static inline mi_heap_t* mi_prim_get_default_heap(void) { + mi_heap_t** pheap = mi_prim_tls_pthread_heap_slot(); + if mi_unlikely(pheap == NULL) return _mi_heap_main_get(); + mi_heap_t* heap = *pheap; + if mi_unlikely(heap == NULL) return (mi_heap_t*)&_mi_heap_empty; + return heap; +} + +#elif defined(MI_TLS_PTHREAD) + +extern pthread_key_t _mi_heap_default_key; +static inline mi_heap_t* mi_prim_get_default_heap(void) { + mi_heap_t* heap = (mi_unlikely(_mi_heap_default_key == (pthread_key_t)(-1)) ? _mi_heap_main_get() : (mi_heap_t*)pthread_getspecific(_mi_heap_default_key)); + return (mi_unlikely(heap == NULL) ? (mi_heap_t*)&_mi_heap_empty : heap); +} + +#else // default using a thread local variable; used on most platforms. + +static inline mi_heap_t* mi_prim_get_default_heap(void) { + #if defined(MI_TLS_RECURSE_GUARD) + if (mi_unlikely(!_mi_process_is_initialized)) return _mi_heap_main_get(); + #endif + return _mi_heap_default; +} + +#endif // mi_prim_get_default_heap() + + + +#endif // MIMALLOC_PRIM_H diff --git a/compat/mimalloc/mimalloc/track.h b/compat/mimalloc/mimalloc/track.h new file mode 100644 index 00000000000000..fa1a048d846a9c --- /dev/null +++ b/compat/mimalloc/mimalloc/track.h @@ -0,0 +1,147 @@ +/* ---------------------------------------------------------------------------- +Copyright (c) 2018-2023, Microsoft Research, Daan Leijen +This is free software; you can redistribute it and/or modify it under the +terms of the MIT license. A copy of the license can be found in the file +"LICENSE" at the root of this distribution. +-----------------------------------------------------------------------------*/ +#pragma once +#ifndef MIMALLOC_TRACK_H +#define MIMALLOC_TRACK_H + +/* ------------------------------------------------------------------------------------------------------ +Track memory ranges with macros for tools like Valgrind address sanitizer, or other memory checkers. +These can be defined for tracking allocation: + + #define mi_track_malloc_size(p,reqsize,size,zero) + #define mi_track_free_size(p,_size) + +The macros are set up such that the size passed to `mi_track_free_size` +always matches the size of `mi_track_malloc_size`. (currently, `size == mi_usable_size(p)`). +The `reqsize` is what the user requested, and `size >= reqsize`. +The `size` is either byte precise (and `size==reqsize`) if `MI_PADDING` is enabled, +or otherwise it is the usable block size which may be larger than the original request. +Use `_mi_block_size_of(void* p)` to get the full block size that was allocated (including padding etc). +The `zero` parameter is `true` if the allocated block is zero initialized. + +Optional: + + #define mi_track_align(p,alignedp,offset,size) + #define mi_track_resize(p,oldsize,newsize) + #define mi_track_init() + +The `mi_track_align` is called right after a `mi_track_malloc` for aligned pointers in a block. +The corresponding `mi_track_free` still uses the block start pointer and original size (corresponding to the `mi_track_malloc`). +The `mi_track_resize` is currently unused but could be called on reallocations within a block. +`mi_track_init` is called at program start. + +The following macros are for tools like asan and valgrind to track whether memory is +defined, undefined, or not accessible at all: + + #define mi_track_mem_defined(p,size) + #define mi_track_mem_undefined(p,size) + #define mi_track_mem_noaccess(p,size) + +-------------------------------------------------------------------------------------------------------*/ + +#if MI_TRACK_VALGRIND +// valgrind tool + +#define MI_TRACK_ENABLED 1 +#define MI_TRACK_HEAP_DESTROY 1 // track free of individual blocks on heap_destroy +#define MI_TRACK_TOOL "valgrind" + +#include +#include + +#define mi_track_malloc_size(p,reqsize,size,zero) VALGRIND_MALLOCLIKE_BLOCK(p,size,MI_PADDING_SIZE /*red zone*/,zero) +#define mi_track_free_size(p,_size) VALGRIND_FREELIKE_BLOCK(p,MI_PADDING_SIZE /*red zone*/) +#define mi_track_resize(p,oldsize,newsize) VALGRIND_RESIZEINPLACE_BLOCK(p,oldsize,newsize,MI_PADDING_SIZE /*red zone*/) +#define mi_track_mem_defined(p,size) VALGRIND_MAKE_MEM_DEFINED(p,size) +#define mi_track_mem_undefined(p,size) VALGRIND_MAKE_MEM_UNDEFINED(p,size) +#define mi_track_mem_noaccess(p,size) VALGRIND_MAKE_MEM_NOACCESS(p,size) + +#elif MI_TRACK_ASAN +// address sanitizer + +#define MI_TRACK_ENABLED 1 +#define MI_TRACK_HEAP_DESTROY 0 +#define MI_TRACK_TOOL "asan" + +#include + +#define mi_track_malloc_size(p,reqsize,size,zero) ASAN_UNPOISON_MEMORY_REGION(p,size) +#define mi_track_free_size(p,size) ASAN_POISON_MEMORY_REGION(p,size) +#define mi_track_mem_defined(p,size) ASAN_UNPOISON_MEMORY_REGION(p,size) +#define mi_track_mem_undefined(p,size) ASAN_UNPOISON_MEMORY_REGION(p,size) +#define mi_track_mem_noaccess(p,size) ASAN_POISON_MEMORY_REGION(p,size) + +#elif MI_TRACK_ETW +// windows event tracing + +#define MI_TRACK_ENABLED 1 +#define MI_TRACK_HEAP_DESTROY 1 +#define MI_TRACK_TOOL "ETW" + +#define WIN32_LEAN_AND_MEAN +#include +#include "../src/prim/windows/etw.h" + +#define mi_track_init() EventRegistermicrosoft_windows_mimalloc(); +#define mi_track_malloc_size(p,reqsize,size,zero) EventWriteETW_MI_ALLOC((UINT64)(p), size) +#define mi_track_free_size(p,size) EventWriteETW_MI_FREE((UINT64)(p), size) + +#else +// no tracking + +#define MI_TRACK_ENABLED 0 +#define MI_TRACK_HEAP_DESTROY 0 +#define MI_TRACK_TOOL "none" + +#define mi_track_malloc_size(p,reqsize,size,zero) +#define mi_track_free_size(p,_size) + +#endif + +// ------------------- +// Utility definitions + +#ifndef mi_track_resize +#define mi_track_resize(p,oldsize,newsize) mi_track_free_size(p,oldsize); mi_track_malloc(p,newsize,false) +#endif + +#ifndef mi_track_align +#define mi_track_align(p,alignedp,offset,size) mi_track_mem_noaccess(p,offset) +#endif + +#ifndef mi_track_init +#define mi_track_init() +#endif + +#ifndef mi_track_mem_defined +#define mi_track_mem_defined(p,size) +#endif + +#ifndef mi_track_mem_undefined +#define mi_track_mem_undefined(p,size) +#endif + +#ifndef mi_track_mem_noaccess +#define mi_track_mem_noaccess(p,size) +#endif + + +#if MI_PADDING +#define mi_track_malloc(p,reqsize,zero) \ + if ((p)!=NULL) { \ + mi_assert_internal(mi_usable_size(p)==(reqsize)); \ + mi_track_malloc_size(p,reqsize,reqsize,zero); \ + } +#else +#define mi_track_malloc(p,reqsize,zero) \ + if ((p)!=NULL) { \ + mi_assert_internal(mi_usable_size(p)>=(reqsize)); \ + mi_track_malloc_size(p,reqsize,mi_usable_size(p),zero); \ + } +#endif + +#endif diff --git a/compat/mimalloc/mimalloc/types.h b/compat/mimalloc/mimalloc/types.h new file mode 100644 index 00000000000000..7616f37e4b978f --- /dev/null +++ b/compat/mimalloc/mimalloc/types.h @@ -0,0 +1,670 @@ +/* ---------------------------------------------------------------------------- +Copyright (c) 2018-2023, Microsoft Research, Daan Leijen +This is free software; you can redistribute it and/or modify it under the +terms of the MIT license. A copy of the license can be found in the file +"LICENSE" at the root of this distribution. +-----------------------------------------------------------------------------*/ +#pragma once +#ifndef MIMALLOC_TYPES_H +#define MIMALLOC_TYPES_H + +// -------------------------------------------------------------------------- +// This file contains the main type definitions for mimalloc: +// mi_heap_t : all data for a thread-local heap, contains +// lists of all managed heap pages. +// mi_segment_t : a larger chunk of memory (32GiB) from where pages +// are allocated. +// mi_page_t : a mimalloc page (usually 64KiB or 512KiB) from +// where objects are allocated. +// -------------------------------------------------------------------------- + + +#include // ptrdiff_t +#include // uintptr_t, uint16_t, etc +#include "mimalloc/atomic.h" // _Atomic + +#ifdef _MSC_VER +#pragma warning(disable:4214) // bitfield is not int +#endif + +// Minimal alignment necessary. On most platforms 16 bytes are needed +// due to SSE registers for example. This must be at least `sizeof(void*)` +#ifndef MI_MAX_ALIGN_SIZE +#define MI_MAX_ALIGN_SIZE 16 // sizeof(max_align_t) +#endif + +// ------------------------------------------------------ +// Variants +// ------------------------------------------------------ + +// Define NDEBUG in the release version to disable assertions. +// #define NDEBUG + +// Define MI_TRACK_ to enable tracking support +// #define MI_TRACK_VALGRIND 1 +// #define MI_TRACK_ASAN 1 +// #define MI_TRACK_ETW 1 + +// Define MI_STAT as 1 to maintain statistics; set it to 2 to have detailed statistics (but costs some performance). +// #define MI_STAT 1 + +// Define MI_SECURE to enable security mitigations +// #define MI_SECURE 1 // guard page around metadata +// #define MI_SECURE 2 // guard page around each mimalloc page +// #define MI_SECURE 3 // encode free lists (detect corrupted free list (buffer overflow), and invalid pointer free) +// #define MI_SECURE 4 // checks for double free. (may be more expensive) + +#if !defined(MI_SECURE) +#define MI_SECURE 0 +#endif + +// Define MI_DEBUG for debug mode +// #define MI_DEBUG 1 // basic assertion checks and statistics, check double free, corrupted free list, and invalid pointer free. +// #define MI_DEBUG 2 // + internal assertion checks +// #define MI_DEBUG 3 // + extensive internal invariant checking (cmake -DMI_DEBUG_FULL=ON) +#if !defined(MI_DEBUG) +#if !defined(NDEBUG) || defined(_DEBUG) +#define MI_DEBUG 2 +#else +#define MI_DEBUG 0 +#endif +#endif + +// Reserve extra padding at the end of each block to be more resilient against heap block overflows. +// The padding can detect buffer overflow on free. +#if !defined(MI_PADDING) && (MI_SECURE>=3 || MI_DEBUG>=1 || (MI_TRACK_VALGRIND || MI_TRACK_ASAN || MI_TRACK_ETW)) +#define MI_PADDING 1 +#endif + +// Check padding bytes; allows byte-precise buffer overflow detection +#if !defined(MI_PADDING_CHECK) && MI_PADDING && (MI_SECURE>=3 || MI_DEBUG>=1) +#define MI_PADDING_CHECK 1 +#endif + + +// Encoded free lists allow detection of corrupted free lists +// and can detect buffer overflows, modify after free, and double `free`s. +#if (MI_SECURE>=3 || MI_DEBUG>=1) +#define MI_ENCODE_FREELIST 1 +#endif + + +// We used to abandon huge pages but to eagerly deallocate if freed from another thread, +// but that makes it not possible to visit them during a heap walk or include them in a +// `mi_heap_destroy`. We therefore instead reset/decommit the huge blocks if freed from +// another thread so most memory is available until it gets properly freed by the owning thread. +// #define MI_HUGE_PAGE_ABANDON 1 + + +// ------------------------------------------------------ +// Platform specific values +// ------------------------------------------------------ + +// ------------------------------------------------------ +// Size of a pointer. +// We assume that `sizeof(void*)==sizeof(intptr_t)` +// and it holds for all platforms we know of. +// +// However, the C standard only requires that: +// p == (void*)((intptr_t)p)) +// but we also need: +// i == (intptr_t)((void*)i) +// or otherwise one might define an intptr_t type that is larger than a pointer... +// ------------------------------------------------------ + +#if INTPTR_MAX > INT64_MAX +# define MI_INTPTR_SHIFT (4) // assume 128-bit (as on arm CHERI for example) +#elif INTPTR_MAX == INT64_MAX +# define MI_INTPTR_SHIFT (3) +#elif INTPTR_MAX == INT32_MAX +# define MI_INTPTR_SHIFT (2) +#else +#error platform pointers must be 32, 64, or 128 bits +#endif + +#if SIZE_MAX == UINT64_MAX +# define MI_SIZE_SHIFT (3) +typedef int64_t mi_ssize_t; +#elif SIZE_MAX == UINT32_MAX +# define MI_SIZE_SHIFT (2) +typedef int32_t mi_ssize_t; +#else +#error platform objects must be 32 or 64 bits +#endif + +#if (SIZE_MAX/2) > LONG_MAX +# define MI_ZU(x) x##ULL +# define MI_ZI(x) x##LL +#else +# define MI_ZU(x) x##UL +# define MI_ZI(x) x##L +#endif + +#define MI_INTPTR_SIZE (1< 4 +#define MI_SEGMENT_SHIFT ( 9 + MI_SEGMENT_SLICE_SHIFT) // 32MiB +#else +#define MI_SEGMENT_SHIFT ( 7 + MI_SEGMENT_SLICE_SHIFT) // 4MiB on 32-bit +#endif + +#define MI_SMALL_PAGE_SHIFT (MI_SEGMENT_SLICE_SHIFT) // 64KiB +#define MI_MEDIUM_PAGE_SHIFT ( 3 + MI_SMALL_PAGE_SHIFT) // 512KiB + + +// Derived constants +#define MI_SEGMENT_SIZE (MI_ZU(1)<= 655360) +#error "mimalloc internal: define more bins" +#endif + +// Maximum slice offset (15) +#define MI_MAX_SLICE_OFFSET ((MI_ALIGNMENT_MAX / MI_SEGMENT_SLICE_SIZE) - 1) + +// Used as a special value to encode block sizes in 32 bits. +#define MI_HUGE_BLOCK_SIZE ((uint32_t)(2*MI_GiB)) + +// blocks up to this size are always allocated aligned +#define MI_MAX_ALIGN_GUARANTEE (8*MI_MAX_ALIGN_SIZE) + +// Alignments over MI_ALIGNMENT_MAX are allocated in dedicated huge page segments +#define MI_ALIGNMENT_MAX (MI_SEGMENT_SIZE >> 1) + + +// ------------------------------------------------------ +// Mimalloc pages contain allocated blocks +// ------------------------------------------------------ + +// The free lists use encoded next fields +// (Only actually encodes when MI_ENCODED_FREELIST is defined.) +typedef uintptr_t mi_encoded_t; + +// thread id's +typedef size_t mi_threadid_t; + +// free lists contain blocks +typedef struct mi_block_s { + mi_encoded_t next; +} mi_block_t; + + +// The delayed flags are used for efficient multi-threaded free-ing +typedef enum mi_delayed_e { + MI_USE_DELAYED_FREE = 0, // push on the owning heap thread delayed list + MI_DELAYED_FREEING = 1, // temporary: another thread is accessing the owning heap + MI_NO_DELAYED_FREE = 2, // optimize: push on page local thread free queue if another block is already in the heap thread delayed free list + MI_NEVER_DELAYED_FREE = 3 // sticky, only resets on page reclaim +} mi_delayed_t; + + +// The `in_full` and `has_aligned` page flags are put in a union to efficiently +// test if both are false (`full_aligned == 0`) in the `mi_free` routine. +#if !MI_TSAN +typedef union mi_page_flags_s { + uint8_t full_aligned; + struct { + uint8_t in_full : 1; + uint8_t has_aligned : 1; + } x; +} mi_page_flags_t; +#else +// under thread sanitizer, use a byte for each flag to suppress warning, issue #130 +typedef union mi_page_flags_s { + uint16_t full_aligned; + struct { + uint8_t in_full; + uint8_t has_aligned; + } x; +} mi_page_flags_t; +#endif + +// Thread free list. +// We use the bottom 2 bits of the pointer for mi_delayed_t flags +typedef uintptr_t mi_thread_free_t; + +// A page contains blocks of one specific size (`block_size`). +// Each page has three list of free blocks: +// `free` for blocks that can be allocated, +// `local_free` for freed blocks that are not yet available to `mi_malloc` +// `thread_free` for freed blocks by other threads +// The `local_free` and `thread_free` lists are migrated to the `free` list +// when it is exhausted. The separate `local_free` list is necessary to +// implement a monotonic heartbeat. The `thread_free` list is needed for +// avoiding atomic operations in the common case. +// +// +// `used - |thread_free|` == actual blocks that are in use (alive) +// `used - |thread_free| + |free| + |local_free| == capacity` +// +// We don't count `freed` (as |free|) but use `used` to reduce +// the number of memory accesses in the `mi_page_all_free` function(s). +// +// Notes: +// - Access is optimized for `mi_free` and `mi_page_alloc` (in `alloc.c`) +// - Using `uint16_t` does not seem to slow things down +// - The size is 8 words on 64-bit which helps the page index calculations +// (and 10 words on 32-bit, and encoded free lists add 2 words. Sizes 10 +// and 12 are still good for address calculation) +// - To limit the structure size, the `xblock_size` is 32-bits only; for +// blocks > MI_HUGE_BLOCK_SIZE the size is determined from the segment page size +// - `thread_free` uses the bottom bits as a delayed-free flags to optimize +// concurrent frees where only the first concurrent free adds to the owning +// heap `thread_delayed_free` list (see `alloc.c:mi_free_block_mt`). +// The invariant is that no-delayed-free is only set if there is +// at least one block that will be added, or as already been added, to +// the owning heap `thread_delayed_free` list. This guarantees that pages +// will be freed correctly even if only other threads free blocks. +typedef struct mi_page_s { + // "owned" by the segment + uint32_t slice_count; // slices in this page (0 if not a page) + uint32_t slice_offset; // distance from the actual page data slice (0 if a page) + uint8_t is_committed : 1; // `true` if the page virtual memory is committed + uint8_t is_zero_init : 1; // `true` if the page was initially zero initialized + + // layout like this to optimize access in `mi_malloc` and `mi_free` + uint16_t capacity; // number of blocks committed, must be the first field, see `segment.c:page_clear` + uint16_t reserved; // number of blocks reserved in memory + mi_page_flags_t flags; // `in_full` and `has_aligned` flags (8 bits) + uint8_t free_is_zero : 1; // `true` if the blocks in the free list are zero initialized + uint8_t retire_expire : 7; // expiration count for retired blocks + + mi_block_t* free; // list of available free blocks (`malloc` allocates from this list) + uint32_t used; // number of blocks in use (including blocks in `local_free` and `thread_free`) + uint32_t xblock_size; // size available in each block (always `>0`) + mi_block_t* local_free; // list of deferred free blocks by this thread (migrates to `free`) + + #if (MI_ENCODE_FREELIST || MI_PADDING) + uintptr_t keys[2]; // two random keys to encode the free lists (see `_mi_block_next`) or padding canary + #endif + + _Atomic(mi_thread_free_t) xthread_free; // list of deferred free blocks freed by other threads + _Atomic(uintptr_t) xheap; + + struct mi_page_s* next; // next page owned by this thread with the same `block_size` + struct mi_page_s* prev; // previous page owned by this thread with the same `block_size` + + // 64-bit 9 words, 32-bit 12 words, (+2 for secure) + #if MI_INTPTR_SIZE==8 + uintptr_t padding[1]; + #endif +} mi_page_t; + + + +// ------------------------------------------------------ +// Mimalloc segments contain mimalloc pages +// ------------------------------------------------------ + +typedef enum mi_page_kind_e { + MI_PAGE_SMALL, // small blocks go into 64KiB pages inside a segment + MI_PAGE_MEDIUM, // medium blocks go into medium pages inside a segment + MI_PAGE_LARGE, // larger blocks go into a page of just one block + MI_PAGE_HUGE, // huge blocks (> 16 MiB) are put into a single page in a single segment. +} mi_page_kind_t; + +typedef enum mi_segment_kind_e { + MI_SEGMENT_NORMAL, // MI_SEGMENT_SIZE size with pages inside. + MI_SEGMENT_HUGE, // > MI_LARGE_SIZE_MAX segment with just one huge page inside. +} mi_segment_kind_t; + +// ------------------------------------------------------ +// A segment holds a commit mask where a bit is set if +// the corresponding MI_COMMIT_SIZE area is committed. +// The MI_COMMIT_SIZE must be a multiple of the slice +// size. If it is equal we have the most fine grained +// decommit (but setting it higher can be more efficient). +// The MI_MINIMAL_COMMIT_SIZE is the minimal amount that will +// be committed in one go which can be set higher than +// MI_COMMIT_SIZE for efficiency (while the decommit mask +// is still tracked in fine-grained MI_COMMIT_SIZE chunks) +// ------------------------------------------------------ + +#define MI_MINIMAL_COMMIT_SIZE (1*MI_SEGMENT_SLICE_SIZE) +#define MI_COMMIT_SIZE (MI_SEGMENT_SLICE_SIZE) // 64KiB +#define MI_COMMIT_MASK_BITS (MI_SEGMENT_SIZE / MI_COMMIT_SIZE) +#define MI_COMMIT_MASK_FIELD_BITS MI_SIZE_BITS +#define MI_COMMIT_MASK_FIELD_COUNT (MI_COMMIT_MASK_BITS / MI_COMMIT_MASK_FIELD_BITS) + +#if (MI_COMMIT_MASK_BITS != (MI_COMMIT_MASK_FIELD_COUNT * MI_COMMIT_MASK_FIELD_BITS)) +#error "the segment size must be exactly divisible by the (commit size * size_t bits)" +#endif + +typedef struct mi_commit_mask_s { + size_t mask[MI_COMMIT_MASK_FIELD_COUNT]; +} mi_commit_mask_t; + +typedef mi_page_t mi_slice_t; +typedef int64_t mi_msecs_t; + + +// Memory can reside in arena's, direct OS allocated, or statically allocated. The memid keeps track of this. +typedef enum mi_memkind_e { + MI_MEM_NONE, // not allocated + MI_MEM_EXTERNAL, // not owned by mimalloc but provided externally (via `mi_manage_os_memory` for example) + MI_MEM_STATIC, // allocated in a static area and should not be freed (for arena meta data for example) + MI_MEM_OS, // allocated from the OS + MI_MEM_OS_HUGE, // allocated as huge os pages + MI_MEM_OS_REMAP, // allocated in a remapable area (i.e. using `mremap`) + MI_MEM_ARENA // allocated from an arena (the usual case) +} mi_memkind_t; + +static inline bool mi_memkind_is_os(mi_memkind_t memkind) { + return (memkind >= MI_MEM_OS && memkind <= MI_MEM_OS_REMAP); +} + +typedef struct mi_memid_os_info { + void* base; // actual base address of the block (used for offset aligned allocations) + size_t alignment; // alignment at allocation +} mi_memid_os_info_t; + +typedef struct mi_memid_arena_info { + size_t block_index; // index in the arena + mi_arena_id_t id; // arena id (>= 1) + bool is_exclusive; // the arena can only be used for specific arena allocations +} mi_memid_arena_info_t; + +typedef struct mi_memid_s { + union { + mi_memid_os_info_t os; // only used for MI_MEM_OS + mi_memid_arena_info_t arena; // only used for MI_MEM_ARENA + } mem; + bool is_pinned; // `true` if we cannot decommit/reset/protect in this memory (e.g. when allocated using large OS pages) + bool initially_committed;// `true` if the memory was originally allocated as committed + bool initially_zero; // `true` if the memory was originally zero initialized + mi_memkind_t memkind; +} mi_memid_t; + + +// Segments are large allocated memory blocks (8mb on 64 bit) from +// the OS. Inside segments we allocated fixed size _pages_ that +// contain blocks. +typedef struct mi_segment_s { + // constant fields + mi_memid_t memid; // memory id for arena allocation + bool allow_decommit; + bool allow_purge; + size_t segment_size; + + // segment fields + mi_msecs_t purge_expire; + mi_commit_mask_t purge_mask; + mi_commit_mask_t commit_mask; + + _Atomic(struct mi_segment_s*) abandoned_next; + + // from here is zero initialized + struct mi_segment_s* next; // the list of freed segments in the cache (must be first field, see `segment.c:mi_segment_init`) + + size_t abandoned; // abandoned pages (i.e. the original owning thread stopped) (`abandoned <= used`) + size_t abandoned_visits; // count how often this segment is visited in the abandoned list (to force reclaim it it is too long) + size_t used; // count of pages in use + uintptr_t cookie; // verify addresses in debug mode: `mi_ptr_cookie(segment) == segment->cookie` + + size_t segment_slices; // for huge segments this may be different from `MI_SLICES_PER_SEGMENT` + size_t segment_info_slices; // initial slices we are using segment info and possible guard pages. + + // layout like this to optimize access in `mi_free` + mi_segment_kind_t kind; + size_t slice_entries; // entries in the `slices` array, at most `MI_SLICES_PER_SEGMENT` + _Atomic(mi_threadid_t) thread_id; // unique id of the thread owning this segment + + mi_slice_t slices[MI_SLICES_PER_SEGMENT+1]; // one more for huge blocks with large alignment +} mi_segment_t; + + +// ------------------------------------------------------ +// Heaps +// Provide first-class heaps to allocate from. +// A heap just owns a set of pages for allocation and +// can only be allocate/reallocate from the thread that created it. +// Freeing blocks can be done from any thread though. +// Per thread, the segments are shared among its heaps. +// Per thread, there is always a default heap that is +// used for allocation; it is initialized to statically +// point to an empty heap to avoid initialization checks +// in the fast path. +// ------------------------------------------------------ + +// Thread local data +typedef struct mi_tld_s mi_tld_t; + +// Pages of a certain block size are held in a queue. +typedef struct mi_page_queue_s { + mi_page_t* first; + mi_page_t* last; + size_t block_size; +} mi_page_queue_t; + +#define MI_BIN_FULL (MI_BIN_HUGE+1) + +// Random context +typedef struct mi_random_cxt_s { + uint32_t input[16]; + uint32_t output[16]; + int output_available; + bool weak; +} mi_random_ctx_t; + + +// In debug mode there is a padding structure at the end of the blocks to check for buffer overflows +#if (MI_PADDING) +typedef struct mi_padding_s { + uint32_t canary; // encoded block value to check validity of the padding (in case of overflow) + uint32_t delta; // padding bytes before the block. (mi_usable_size(p) - delta == exact allocated bytes) +} mi_padding_t; +#define MI_PADDING_SIZE (sizeof(mi_padding_t)) +#define MI_PADDING_WSIZE ((MI_PADDING_SIZE + MI_INTPTR_SIZE - 1) / MI_INTPTR_SIZE) +#else +#define MI_PADDING_SIZE 0 +#define MI_PADDING_WSIZE 0 +#endif + +#define MI_PAGES_DIRECT (MI_SMALL_WSIZE_MAX + MI_PADDING_WSIZE + 1) + + +// A heap owns a set of pages. +struct mi_heap_s { + mi_tld_t* tld; + mi_page_t* pages_free_direct[MI_PAGES_DIRECT]; // optimize: array where every entry points a page with possibly free blocks in the corresponding queue for that size. + mi_page_queue_t pages[MI_BIN_FULL + 1]; // queue of pages for each size class (or "bin") + _Atomic(mi_block_t*) thread_delayed_free; + mi_threadid_t thread_id; // thread this heap belongs too + mi_arena_id_t arena_id; // arena id if the heap belongs to a specific arena (or 0) + uintptr_t cookie; // random cookie to verify pointers (see `_mi_ptr_cookie`) + uintptr_t keys[2]; // two random keys used to encode the `thread_delayed_free` list + mi_random_ctx_t random; // random number context used for secure allocation + size_t page_count; // total number of pages in the `pages` queues. + size_t page_retired_min; // smallest retired index (retired pages are fully free, but still in the page queues) + size_t page_retired_max; // largest retired index into the `pages` array. + mi_heap_t* next; // list of heaps per thread + bool no_reclaim; // `true` if this heap should not reclaim abandoned pages +}; + + + +// ------------------------------------------------------ +// Debug +// ------------------------------------------------------ + +#if !defined(MI_DEBUG_UNINIT) +#define MI_DEBUG_UNINIT (0xD0) +#endif +#if !defined(MI_DEBUG_FREED) +#define MI_DEBUG_FREED (0xDF) +#endif +#if !defined(MI_DEBUG_PADDING) +#define MI_DEBUG_PADDING (0xDE) +#endif + +#if (MI_DEBUG) +// use our own assertion to print without memory allocation +void _mi_assert_fail(const char* assertion, const char* fname, unsigned int line, const char* func ); +#define mi_assert(expr) ((expr) ? (void)0 : _mi_assert_fail(#expr,__FILE__,__LINE__,__func__)) +#else +#define mi_assert(x) +#endif + +#if (MI_DEBUG>1) +#define mi_assert_internal mi_assert +#else +#define mi_assert_internal(x) +#endif + +#if (MI_DEBUG>2) +#define mi_assert_expensive mi_assert +#else +#define mi_assert_expensive(x) +#endif + +// ------------------------------------------------------ +// Statistics +// ------------------------------------------------------ + +#ifndef MI_STAT +#if (MI_DEBUG>0) +#define MI_STAT 2 +#else +#define MI_STAT 0 +#endif +#endif + +typedef struct mi_stat_count_s { + int64_t allocated; + int64_t freed; + int64_t peak; + int64_t current; +} mi_stat_count_t; + +typedef struct mi_stat_counter_s { + int64_t total; + int64_t count; +} mi_stat_counter_t; + +typedef struct mi_stats_s { + mi_stat_count_t segments; + mi_stat_count_t pages; + mi_stat_count_t reserved; + mi_stat_count_t committed; + mi_stat_count_t reset; + mi_stat_count_t purged; + mi_stat_count_t page_committed; + mi_stat_count_t segments_abandoned; + mi_stat_count_t pages_abandoned; + mi_stat_count_t threads; + mi_stat_count_t normal; + mi_stat_count_t huge; + mi_stat_count_t large; + mi_stat_count_t malloc; + mi_stat_count_t segments_cache; + mi_stat_counter_t pages_extended; + mi_stat_counter_t mmap_calls; + mi_stat_counter_t commit_calls; + mi_stat_counter_t reset_calls; + mi_stat_counter_t purge_calls; + mi_stat_counter_t page_no_retire; + mi_stat_counter_t searches; + mi_stat_counter_t normal_count; + mi_stat_counter_t huge_count; + mi_stat_counter_t large_count; +#if MI_STAT>1 + mi_stat_count_t normal_bins[MI_BIN_HUGE+1]; +#endif +} mi_stats_t; + + +void _mi_stat_increase(mi_stat_count_t* stat, size_t amount); +void _mi_stat_decrease(mi_stat_count_t* stat, size_t amount); +void _mi_stat_counter_increase(mi_stat_counter_t* stat, size_t amount); + +#if (MI_STAT) +#define mi_stat_increase(stat,amount) _mi_stat_increase( &(stat), amount) +#define mi_stat_decrease(stat,amount) _mi_stat_decrease( &(stat), amount) +#define mi_stat_counter_increase(stat,amount) _mi_stat_counter_increase( &(stat), amount) +#else +#define mi_stat_increase(stat,amount) (void)0 +#define mi_stat_decrease(stat,amount) (void)0 +#define mi_stat_counter_increase(stat,amount) (void)0 +#endif + +#define mi_heap_stat_counter_increase(heap,stat,amount) mi_stat_counter_increase( (heap)->tld->stats.stat, amount) +#define mi_heap_stat_increase(heap,stat,amount) mi_stat_increase( (heap)->tld->stats.stat, amount) +#define mi_heap_stat_decrease(heap,stat,amount) mi_stat_decrease( (heap)->tld->stats.stat, amount) + +// ------------------------------------------------------ +// Thread Local data +// ------------------------------------------------------ + +// A "span" is is an available range of slices. The span queues keep +// track of slice spans of at most the given `slice_count` (but more than the previous size class). +typedef struct mi_span_queue_s { + mi_slice_t* first; + mi_slice_t* last; + size_t slice_count; +} mi_span_queue_t; + +#define MI_SEGMENT_BIN_MAX (35) // 35 == mi_segment_bin(MI_SLICES_PER_SEGMENT) + +// OS thread local data +typedef struct mi_os_tld_s { + size_t region_idx; // start point for next allocation + mi_stats_t* stats; // points to tld stats +} mi_os_tld_t; + + +// Segments thread local data +typedef struct mi_segments_tld_s { + mi_span_queue_t spans[MI_SEGMENT_BIN_MAX+1]; // free slice spans inside segments + size_t count; // current number of segments; + size_t peak_count; // peak number of segments + size_t current_size; // current size of all segments + size_t peak_size; // peak size of all segments + mi_stats_t* stats; // points to tld stats + mi_os_tld_t* os; // points to os stats +} mi_segments_tld_t; + +// Thread local data +struct mi_tld_s { + unsigned long long heartbeat; // monotonic heartbeat count + bool recurse; // true if deferred was called; used to prevent infinite recursion. + mi_heap_t* heap_backing; // backing heap of this thread (cannot be deleted) + mi_heap_t* heaps; // list of heaps in this thread (so we can abandon all when the thread terminates) + mi_segments_tld_t segments; // segment tld + mi_os_tld_t os; // os tld + mi_stats_t stats; // statistics +}; + +#endif diff --git a/compat/mimalloc/options.c b/compat/mimalloc/options.c new file mode 100644 index 00000000000000..3a3090d9acfc94 --- /dev/null +++ b/compat/mimalloc/options.c @@ -0,0 +1,571 @@ +/* ---------------------------------------------------------------------------- +Copyright (c) 2018-2021, Microsoft Research, Daan Leijen +This is free software; you can redistribute it and/or modify it under the +terms of the MIT license. A copy of the license can be found in the file +"LICENSE" at the root of this distribution. +-----------------------------------------------------------------------------*/ +#include "mimalloc.h" +#include "mimalloc/internal.h" +#include "mimalloc/atomic.h" +#include "mimalloc/prim.h" // mi_prim_out_stderr + +#include // FILE +#include // abort +#include + + +static long mi_max_error_count = 16; // stop outputting errors after this (use < 0 for no limit) +static long mi_max_warning_count = 16; // stop outputting warnings after this (use < 0 for no limit) + +static void mi_add_stderr_output(void); + +int mi_version(void) mi_attr_noexcept { + return MI_MALLOC_VERSION; +} + + +// -------------------------------------------------------- +// Options +// These can be accessed by multiple threads and may be +// concurrently initialized, but an initializing data race +// is ok since they resolve to the same value. +// -------------------------------------------------------- +typedef enum mi_init_e { + UNINIT, // not yet initialized + DEFAULTED, // not found in the environment, use default value + INITIALIZED // found in environment or set explicitly +} mi_init_t; + +typedef struct mi_option_desc_s { + long value; // the value + mi_init_t init; // is it initialized yet? (from the environment) + mi_option_t option; // for debugging: the option index should match the option + const char* name; // option name without `mimalloc_` prefix + const char* legacy_name; // potential legacy option name +} mi_option_desc_t; + +#define MI_OPTION(opt) mi_option_##opt, #opt, NULL +#define MI_OPTION_LEGACY(opt,legacy) mi_option_##opt, #opt, #legacy + +static mi_option_desc_t options[_mi_option_last] = +{ + // stable options + #if MI_DEBUG || defined(MI_SHOW_ERRORS) + { 1, UNINIT, MI_OPTION(show_errors) }, + #else + { 0, UNINIT, MI_OPTION(show_errors) }, + #endif + { 0, UNINIT, MI_OPTION(show_stats) }, + { 0, UNINIT, MI_OPTION(verbose) }, + + // the following options are experimental and not all combinations make sense. + { 1, UNINIT, MI_OPTION(eager_commit) }, // commit per segment directly (4MiB) (but see also `eager_commit_delay`) + { 2, UNINIT, MI_OPTION_LEGACY(arena_eager_commit,eager_region_commit) }, // eager commit arena's? 2 is used to enable this only on an OS that has overcommit (i.e. linux) + { 1, UNINIT, MI_OPTION_LEGACY(purge_decommits,reset_decommits) }, // purge decommits memory (instead of reset) (note: on linux this uses MADV_DONTNEED for decommit) + { 0, UNINIT, MI_OPTION_LEGACY(allow_large_os_pages,large_os_pages) }, // use large OS pages, use only with eager commit to prevent fragmentation of VMA's + { 0, UNINIT, MI_OPTION(reserve_huge_os_pages) }, // per 1GiB huge pages + {-1, UNINIT, MI_OPTION(reserve_huge_os_pages_at) }, // reserve huge pages at node N + { 0, UNINIT, MI_OPTION(reserve_os_memory) }, + { 0, UNINIT, MI_OPTION(deprecated_segment_cache) }, // cache N segments per thread + { 0, UNINIT, MI_OPTION(deprecated_page_reset) }, // reset page memory on free + { 0, UNINIT, MI_OPTION_LEGACY(abandoned_page_purge,abandoned_page_reset) }, // reset free page memory when a thread terminates + { 0, UNINIT, MI_OPTION(deprecated_segment_reset) }, // reset segment memory on free (needs eager commit) +#if defined(__NetBSD__) + { 0, UNINIT, MI_OPTION(eager_commit_delay) }, // the first N segments per thread are not eagerly committed +#else + { 1, UNINIT, MI_OPTION(eager_commit_delay) }, // the first N segments per thread are not eagerly committed (but per page in the segment on demand) +#endif + { 10, UNINIT, MI_OPTION_LEGACY(purge_delay,reset_delay) }, // purge delay in milli-seconds + { 0, UNINIT, MI_OPTION(use_numa_nodes) }, // 0 = use available numa nodes, otherwise use at most N nodes. + { 0, UNINIT, MI_OPTION(limit_os_alloc) }, // 1 = do not use OS memory for allocation (but only reserved arenas) + { 100, UNINIT, MI_OPTION(os_tag) }, // only apple specific for now but might serve more or less related purpose + { 16, UNINIT, MI_OPTION(max_errors) }, // maximum errors that are output + { 16, UNINIT, MI_OPTION(max_warnings) }, // maximum warnings that are output + { 8, UNINIT, MI_OPTION(max_segment_reclaim)}, // max. number of segment reclaims from the abandoned segments per try. + { 0, UNINIT, MI_OPTION(destroy_on_exit)}, // release all OS memory on process exit; careful with dangling pointer or after-exit frees! + #if (MI_INTPTR_SIZE>4) + { 1024L * 1024L, UNINIT, MI_OPTION(arena_reserve) }, // reserve memory N KiB at a time + #else + { 128L * 1024L, UNINIT, MI_OPTION(arena_reserve) }, + #endif + { 10, UNINIT, MI_OPTION(arena_purge_mult) }, // purge delay multiplier for arena's + { 1, UNINIT, MI_OPTION_LEGACY(purge_extend_delay, decommit_extend_delay) }, +}; + +static void mi_option_init(mi_option_desc_t* desc); + +void _mi_options_init(void) { + // called on process load; should not be called before the CRT is initialized! + // (e.g. do not call this from process_init as that may run before CRT initialization) + mi_add_stderr_output(); // now it safe to use stderr for output + for(int i = 0; i < _mi_option_last; i++ ) { + mi_option_t option = (mi_option_t)i; + long l = mi_option_get(option); MI_UNUSED(l); // initialize + // if (option != mi_option_verbose) + { + mi_option_desc_t* desc = &options[option]; + _mi_verbose_message("option '%s': %ld\n", desc->name, desc->value); + } + } + mi_max_error_count = mi_option_get(mi_option_max_errors); + mi_max_warning_count = mi_option_get(mi_option_max_warnings); +} + +mi_decl_nodiscard long mi_option_get(mi_option_t option) { + mi_assert(option >= 0 && option < _mi_option_last); + if (option < 0 || option >= _mi_option_last) return 0; + mi_option_desc_t* desc = &options[option]; + mi_assert(desc->option == option); // index should match the option + if mi_unlikely(desc->init == UNINIT) { + mi_option_init(desc); + } + return desc->value; +} + +mi_decl_nodiscard long mi_option_get_clamp(mi_option_t option, long min, long max) { + long x = mi_option_get(option); + return (x < min ? min : (x > max ? max : x)); +} + +mi_decl_nodiscard size_t mi_option_get_size(mi_option_t option) { + mi_assert_internal(option == mi_option_reserve_os_memory || option == mi_option_arena_reserve); + long x = mi_option_get(option); + return (x < 0 ? 0 : (size_t)x * MI_KiB); +} + +void mi_option_set(mi_option_t option, long value) { + mi_assert(option >= 0 && option < _mi_option_last); + if (option < 0 || option >= _mi_option_last) return; + mi_option_desc_t* desc = &options[option]; + mi_assert(desc->option == option); // index should match the option + desc->value = value; + desc->init = INITIALIZED; +} + +void mi_option_set_default(mi_option_t option, long value) { + mi_assert(option >= 0 && option < _mi_option_last); + if (option < 0 || option >= _mi_option_last) return; + mi_option_desc_t* desc = &options[option]; + if (desc->init != INITIALIZED) { + desc->value = value; + } +} + +mi_decl_nodiscard bool mi_option_is_enabled(mi_option_t option) { + return (mi_option_get(option) != 0); +} + +void mi_option_set_enabled(mi_option_t option, bool enable) { + mi_option_set(option, (enable ? 1 : 0)); +} + +void mi_option_set_enabled_default(mi_option_t option, bool enable) { + mi_option_set_default(option, (enable ? 1 : 0)); +} + +void mi_option_enable(mi_option_t option) { + mi_option_set_enabled(option,true); +} + +void mi_option_disable(mi_option_t option) { + mi_option_set_enabled(option,false); +} + +static void mi_cdecl mi_out_stderr(const char* msg, void* arg) { + MI_UNUSED(arg); + if (msg != NULL && msg[0] != 0) { + _mi_prim_out_stderr(msg); + } +} + +// Since an output function can be registered earliest in the `main` +// function we also buffer output that happens earlier. When +// an output function is registered it is called immediately with +// the output up to that point. +#ifndef MI_MAX_DELAY_OUTPUT +#define MI_MAX_DELAY_OUTPUT ((size_t)(32*1024)) +#endif +static char out_buf[MI_MAX_DELAY_OUTPUT+1]; +static _Atomic(size_t) out_len; + +static void mi_cdecl mi_out_buf(const char* msg, void* arg) { + MI_UNUSED(arg); + if (msg==NULL) return; + if (mi_atomic_load_relaxed(&out_len)>=MI_MAX_DELAY_OUTPUT) return; + size_t n = _mi_strlen(msg); + if (n==0) return; + // claim space + size_t start = mi_atomic_add_acq_rel(&out_len, n); + if (start >= MI_MAX_DELAY_OUTPUT) return; + // check bound + if (start+n >= MI_MAX_DELAY_OUTPUT) { + n = MI_MAX_DELAY_OUTPUT-start-1; + } + _mi_memcpy(&out_buf[start], msg, n); +} + +static void mi_out_buf_flush(mi_output_fun* out, bool no_more_buf, void* arg) { + if (out==NULL) return; + // claim (if `no_more_buf == true`, no more output will be added after this point) + size_t count = mi_atomic_add_acq_rel(&out_len, (no_more_buf ? MI_MAX_DELAY_OUTPUT : 1)); + // and output the current contents + if (count>MI_MAX_DELAY_OUTPUT) count = MI_MAX_DELAY_OUTPUT; + out_buf[count] = 0; + out(out_buf,arg); + if (!no_more_buf) { + out_buf[count] = '\n'; // if continue with the buffer, insert a newline + } +} + + +// Once this module is loaded, switch to this routine +// which outputs to stderr and the delayed output buffer. +static void mi_cdecl mi_out_buf_stderr(const char* msg, void* arg) { + mi_out_stderr(msg,arg); + mi_out_buf(msg,arg); +} + + + +// -------------------------------------------------------- +// Default output handler +// -------------------------------------------------------- + +// Should be atomic but gives errors on many platforms as generally we cannot cast a function pointer to a uintptr_t. +// For now, don't register output from multiple threads. +static mi_output_fun* volatile mi_out_default; // = NULL +static _Atomic(void*) mi_out_arg; // = NULL + +static mi_output_fun* mi_out_get_default(void** parg) { + if (parg != NULL) { *parg = mi_atomic_load_ptr_acquire(void,&mi_out_arg); } + mi_output_fun* out = mi_out_default; + return (out == NULL ? &mi_out_buf : out); +} + +void mi_register_output(mi_output_fun* out, void* arg) mi_attr_noexcept { + mi_out_default = (out == NULL ? &mi_out_stderr : out); // stop using the delayed output buffer + mi_atomic_store_ptr_release(void,&mi_out_arg, arg); + if (out!=NULL) mi_out_buf_flush(out,true,arg); // output all the delayed output now +} + +// add stderr to the delayed output after the module is loaded +static void mi_add_stderr_output(void) { + mi_assert_internal(mi_out_default == NULL); + mi_out_buf_flush(&mi_out_stderr, false, NULL); // flush current contents to stderr + mi_out_default = &mi_out_buf_stderr; // and add stderr to the delayed output +} + +// -------------------------------------------------------- +// Messages, all end up calling `_mi_fputs`. +// -------------------------------------------------------- +static _Atomic(size_t) error_count; // = 0; // when >= max_error_count stop emitting errors +static _Atomic(size_t) warning_count; // = 0; // when >= max_warning_count stop emitting warnings + +// When overriding malloc, we may recurse into mi_vfprintf if an allocation +// inside the C runtime causes another message. +// In some cases (like on macOS) the loader already allocates which +// calls into mimalloc; if we then access thread locals (like `recurse`) +// this may crash as the access may call _tlv_bootstrap that tries to +// (recursively) invoke malloc again to allocate space for the thread local +// variables on demand. This is why we use a _mi_preloading test on such +// platforms. However, C code generator may move the initial thread local address +// load before the `if` and we therefore split it out in a separate funcion. +static mi_decl_thread bool recurse = false; + +static mi_decl_noinline bool mi_recurse_enter_prim(void) { + if (recurse) return false; + recurse = true; + return true; +} + +static mi_decl_noinline void mi_recurse_exit_prim(void) { + recurse = false; +} + +static bool mi_recurse_enter(void) { + #if defined(__APPLE__) || defined(MI_TLS_RECURSE_GUARD) + if (_mi_preloading()) return false; + #endif + return mi_recurse_enter_prim(); +} + +static void mi_recurse_exit(void) { + #if defined(__APPLE__) || defined(MI_TLS_RECURSE_GUARD) + if (_mi_preloading()) return; + #endif + mi_recurse_exit_prim(); +} + +void _mi_fputs(mi_output_fun* out, void* arg, const char* prefix, const char* message) { + if (out==NULL || (void*)out==(void*)stdout || (void*)out==(void*)stderr) { // TODO: use mi_out_stderr for stderr? + if (!mi_recurse_enter()) return; + out = mi_out_get_default(&arg); + if (prefix != NULL) out(prefix, arg); + out(message, arg); + mi_recurse_exit(); + } + else { + if (prefix != NULL) out(prefix, arg); + out(message, arg); + } +} + +// Define our own limited `fprintf` that avoids memory allocation. +// We do this using `snprintf` with a limited buffer. +static void mi_vfprintf( mi_output_fun* out, void* arg, const char* prefix, const char* fmt, va_list args ) { + char buf[512]; + if (fmt==NULL) return; + if (!mi_recurse_enter()) return; + vsnprintf(buf,sizeof(buf)-1,fmt,args); + mi_recurse_exit(); + _mi_fputs(out,arg,prefix,buf); +} + +void _mi_fprintf( mi_output_fun* out, void* arg, const char* fmt, ... ) { + va_list args; + va_start(args,fmt); + mi_vfprintf(out,arg,NULL,fmt,args); + va_end(args); +} + +static void mi_vfprintf_thread(mi_output_fun* out, void* arg, const char* prefix, const char* fmt, va_list args) { + if (prefix != NULL && _mi_strnlen(prefix,33) <= 32 && !_mi_is_main_thread()) { + char tprefix[64]; + snprintf(tprefix, sizeof(tprefix), "%sthread 0x%llx: ", prefix, (unsigned long long)_mi_thread_id()); + mi_vfprintf(out, arg, tprefix, fmt, args); + } + else { + mi_vfprintf(out, arg, prefix, fmt, args); + } +} + +void _mi_trace_message(const char* fmt, ...) { + if (mi_option_get(mi_option_verbose) <= 1) return; // only with verbose level 2 or higher + va_list args; + va_start(args, fmt); + mi_vfprintf_thread(NULL, NULL, "mimalloc: ", fmt, args); + va_end(args); +} + +void _mi_verbose_message(const char* fmt, ...) { + if (!mi_option_is_enabled(mi_option_verbose)) return; + va_list args; + va_start(args,fmt); + mi_vfprintf(NULL, NULL, "mimalloc: ", fmt, args); + va_end(args); +} + +static void mi_show_error_message(const char* fmt, va_list args) { + if (!mi_option_is_enabled(mi_option_verbose)) { + if (!mi_option_is_enabled(mi_option_show_errors)) return; + if (mi_max_error_count >= 0 && (long)mi_atomic_increment_acq_rel(&error_count) > mi_max_error_count) return; + } + mi_vfprintf_thread(NULL, NULL, "mimalloc: error: ", fmt, args); +} + +void _mi_warning_message(const char* fmt, ...) { + if (!mi_option_is_enabled(mi_option_verbose)) { + if (!mi_option_is_enabled(mi_option_show_errors)) return; + if (mi_max_warning_count >= 0 && (long)mi_atomic_increment_acq_rel(&warning_count) > mi_max_warning_count) return; + } + va_list args; + va_start(args,fmt); + mi_vfprintf_thread(NULL, NULL, "mimalloc: warning: ", fmt, args); + va_end(args); +} + + +#if MI_DEBUG +void _mi_assert_fail(const char* assertion, const char* fname, unsigned line, const char* func ) { + _mi_fprintf(NULL, NULL, "mimalloc: assertion failed: at \"%s\":%u, %s\n assertion: \"%s\"\n", fname, line, (func==NULL?"":func), assertion); + abort(); +} +#endif + +// -------------------------------------------------------- +// Errors +// -------------------------------------------------------- + +static mi_error_fun* volatile mi_error_handler; // = NULL +static _Atomic(void*) mi_error_arg; // = NULL + +static void mi_error_default(int err) { + MI_UNUSED(err); +#if (MI_DEBUG>0) + if (err==EFAULT) { + #ifdef _MSC_VER + __debugbreak(); + #endif + abort(); + } +#endif +#if (MI_SECURE>0) + if (err==EFAULT) { // abort on serious errors in secure mode (corrupted meta-data) + abort(); + } +#endif +#if defined(MI_XMALLOC) + if (err==ENOMEM || err==EOVERFLOW) { // abort on memory allocation fails in xmalloc mode + abort(); + } +#endif +} + +void mi_register_error(mi_error_fun* fun, void* arg) { + mi_error_handler = fun; // can be NULL + mi_atomic_store_ptr_release(void,&mi_error_arg, arg); +} + +void _mi_error_message(int err, const char* fmt, ...) { + // show detailed error message + va_list args; + va_start(args, fmt); + mi_show_error_message(fmt, args); + va_end(args); + // and call the error handler which may abort (or return normally) + if (mi_error_handler != NULL) { + mi_error_handler(err, mi_atomic_load_ptr_acquire(void,&mi_error_arg)); + } + else { + mi_error_default(err); + } +} + +// -------------------------------------------------------- +// Initialize options by checking the environment +// -------------------------------------------------------- +char _mi_toupper(char c) { + if (c >= 'a' && c <= 'z') return (c - 'a' + 'A'); + else return c; +} + +int _mi_strnicmp(const char* s, const char* t, size_t n) { + if (n == 0) return 0; + for (; *s != 0 && *t != 0 && n > 0; s++, t++, n--) { + if (_mi_toupper(*s) != _mi_toupper(*t)) break; + } + return (n == 0 ? 0 : *s - *t); +} + +void _mi_strlcpy(char* dest, const char* src, size_t dest_size) { + if (dest==NULL || src==NULL || dest_size == 0) return; + // copy until end of src, or when dest is (almost) full + while (*src != 0 && dest_size > 1) { + *dest++ = *src++; + dest_size--; + } + // always zero terminate + *dest = 0; +} + +void _mi_strlcat(char* dest, const char* src, size_t dest_size) { + if (dest==NULL || src==NULL || dest_size == 0) return; + // find end of string in the dest buffer + while (*dest != 0 && dest_size > 1) { + dest++; + dest_size--; + } + // and catenate + _mi_strlcpy(dest, src, dest_size); +} + +size_t _mi_strlen(const char* s) { + if (s==NULL) return 0; + size_t len = 0; + while(s[len] != 0) { len++; } + return len; +} + +size_t _mi_strnlen(const char* s, size_t max_len) { + if (s==NULL) return 0; + size_t len = 0; + while(s[len] != 0 && len < max_len) { len++; } + return len; +} + +#ifdef MI_NO_GETENV +static bool mi_getenv(const char* name, char* result, size_t result_size) { + MI_UNUSED(name); + MI_UNUSED(result); + MI_UNUSED(result_size); + return false; +} +#else +static bool mi_getenv(const char* name, char* result, size_t result_size) { + if (name==NULL || result == NULL || result_size < 64) return false; + return _mi_prim_getenv(name,result,result_size); +} +#endif + +// TODO: implement ourselves to reduce dependencies on the C runtime +#include // strtol +#include // strstr + + +static void mi_option_init(mi_option_desc_t* desc) { + // Read option value from the environment + char s[64 + 1]; + char buf[64+1]; + _mi_strlcpy(buf, "mimalloc_", sizeof(buf)); + _mi_strlcat(buf, desc->name, sizeof(buf)); + bool found = mi_getenv(buf, s, sizeof(s)); + if (!found && desc->legacy_name != NULL) { + _mi_strlcpy(buf, "mimalloc_", sizeof(buf)); + _mi_strlcat(buf, desc->legacy_name, sizeof(buf)); + found = mi_getenv(buf, s, sizeof(s)); + if (found) { + _mi_warning_message("environment option \"mimalloc_%s\" is deprecated -- use \"mimalloc_%s\" instead.\n", desc->legacy_name, desc->name); + } + } + + if (found) { + size_t len = _mi_strnlen(s, sizeof(buf) - 1); + for (size_t i = 0; i < len; i++) { + buf[i] = _mi_toupper(s[i]); + } + buf[len] = 0; + if (buf[0] == 0 || strstr("1;TRUE;YES;ON", buf) != NULL) { + desc->value = 1; + desc->init = INITIALIZED; + } + else if (strstr("0;FALSE;NO;OFF", buf) != NULL) { + desc->value = 0; + desc->init = INITIALIZED; + } + else { + char* end = buf; + long value = strtol(buf, &end, 10); + if (desc->option == mi_option_reserve_os_memory || desc->option == mi_option_arena_reserve) { + // this option is interpreted in KiB to prevent overflow of `long` + if (*end == 'K') { end++; } + else if (*end == 'M') { value *= MI_KiB; end++; } + else if (*end == 'G') { value *= MI_MiB; end++; } + else { value = (value + MI_KiB - 1) / MI_KiB; } + if (end[0] == 'I' && end[1] == 'B') { end += 2; } + else if (*end == 'B') { end++; } + } + if (*end == 0) { + desc->value = value; + desc->init = INITIALIZED; + } + else { + // set `init` first to avoid recursion through _mi_warning_message on mimalloc_verbose. + desc->init = DEFAULTED; + if (desc->option == mi_option_verbose && desc->value == 0) { + // if the 'mimalloc_verbose' env var has a bogus value we'd never know + // (since the value defaults to 'off') so in that case briefly enable verbose + desc->value = 1; + _mi_warning_message("environment option mimalloc_%s has an invalid value.\n", desc->name); + desc->value = 0; + } + else { + _mi_warning_message("environment option mimalloc_%s has an invalid value.\n", desc->name); + } + } + } + mi_assert_internal(desc->init != UNINIT); + } + else if (!_mi_preloading()) { + desc->init = DEFAULTED; + } +} diff --git a/compat/mimalloc/os.c b/compat/mimalloc/os.c new file mode 100644 index 00000000000000..bf9de1be0fdb49 --- /dev/null +++ b/compat/mimalloc/os.c @@ -0,0 +1,689 @@ +/* ---------------------------------------------------------------------------- +Copyright (c) 2018-2023, Microsoft Research, Daan Leijen +This is free software; you can redistribute it and/or modify it under the +terms of the MIT license. A copy of the license can be found in the file +"LICENSE" at the root of this distribution. +-----------------------------------------------------------------------------*/ +#include "mimalloc.h" +#include "mimalloc/internal.h" +#include "mimalloc/atomic.h" +#include "mimalloc/prim.h" + + +/* ----------------------------------------------------------- + Initialization. + On windows initializes support for aligned allocation and + large OS pages (if MIMALLOC_LARGE_OS_PAGES is true). +----------------------------------------------------------- */ + +static mi_os_mem_config_t mi_os_mem_config = { + 4096, // page size + 0, // large page size (usually 2MiB) + 4096, // allocation granularity + true, // has overcommit? (if true we use MAP_NORESERVE on mmap systems) + false, // must free whole? (on mmap systems we can free anywhere in a mapped range, but on Windows we must free the entire span) + true // has virtual reserve? (if true we can reserve virtual address space without using commit or physical memory) +}; + +bool _mi_os_has_overcommit(void) { + return mi_os_mem_config.has_overcommit; +} + +bool _mi_os_has_virtual_reserve(void) { + return mi_os_mem_config.has_virtual_reserve; +} + + +// OS (small) page size +size_t _mi_os_page_size(void) { + return mi_os_mem_config.page_size; +} + +// if large OS pages are supported (2 or 4MiB), then return the size, otherwise return the small page size (4KiB) +size_t _mi_os_large_page_size(void) { + return (mi_os_mem_config.large_page_size != 0 ? mi_os_mem_config.large_page_size : _mi_os_page_size()); +} + +bool _mi_os_use_large_page(size_t size, size_t alignment) { + // if we have access, check the size and alignment requirements + if (mi_os_mem_config.large_page_size == 0 || !mi_option_is_enabled(mi_option_allow_large_os_pages)) return false; + return ((size % mi_os_mem_config.large_page_size) == 0 && (alignment % mi_os_mem_config.large_page_size) == 0); +} + +// round to a good OS allocation size (bounded by max 12.5% waste) +size_t _mi_os_good_alloc_size(size_t size) { + size_t align_size; + if (size < 512*MI_KiB) align_size = _mi_os_page_size(); + else if (size < 2*MI_MiB) align_size = 64*MI_KiB; + else if (size < 8*MI_MiB) align_size = 256*MI_KiB; + else if (size < 32*MI_MiB) align_size = 1*MI_MiB; + else align_size = 4*MI_MiB; + if mi_unlikely(size >= (SIZE_MAX - align_size)) return size; // possible overflow? + return _mi_align_up(size, align_size); +} + +void _mi_os_init(void) { + _mi_prim_mem_init(&mi_os_mem_config); +} + + +/* ----------------------------------------------------------- + Util +-------------------------------------------------------------- */ +bool _mi_os_decommit(void* addr, size_t size, mi_stats_t* stats); +bool _mi_os_commit(void* addr, size_t size, bool* is_zero, mi_stats_t* tld_stats); + +static void* mi_align_up_ptr(void* p, size_t alignment) { + return (void*)_mi_align_up((uintptr_t)p, alignment); +} + +static void* mi_align_down_ptr(void* p, size_t alignment) { + return (void*)_mi_align_down((uintptr_t)p, alignment); +} + + +/* ----------------------------------------------------------- + aligned hinting +-------------------------------------------------------------- */ + +// On 64-bit systems, we can do efficient aligned allocation by using +// the 2TiB to 30TiB area to allocate those. +#if (MI_INTPTR_SIZE >= 8) +static mi_decl_cache_align _Atomic(uintptr_t)aligned_base; + +// Return a MI_SEGMENT_SIZE aligned address that is probably available. +// If this returns NULL, the OS will determine the address but on some OS's that may not be +// properly aligned which can be more costly as it needs to be adjusted afterwards. +// For a size > 1GiB this always returns NULL in order to guarantee good ASLR randomization; +// (otherwise an initial large allocation of say 2TiB has a 50% chance to include (known) addresses +// in the middle of the 2TiB - 6TiB address range (see issue #372)) + +#define MI_HINT_BASE ((uintptr_t)2 << 40) // 2TiB start +#define MI_HINT_AREA ((uintptr_t)4 << 40) // upto 6TiB (since before win8 there is "only" 8TiB available to processes) +#define MI_HINT_MAX ((uintptr_t)30 << 40) // wrap after 30TiB (area after 32TiB is used for huge OS pages) + +void* _mi_os_get_aligned_hint(size_t try_alignment, size_t size) +{ + if (try_alignment <= 1 || try_alignment > MI_SEGMENT_SIZE) return NULL; + size = _mi_align_up(size, MI_SEGMENT_SIZE); + if (size > 1*MI_GiB) return NULL; // guarantee the chance of fixed valid address is at most 1/(MI_HINT_AREA / 1<<30) = 1/4096. + #if (MI_SECURE>0) + size += MI_SEGMENT_SIZE; // put in `MI_SEGMENT_SIZE` virtual gaps between hinted blocks; this splits VLA's but increases guarded areas. + #endif + + uintptr_t hint = mi_atomic_add_acq_rel(&aligned_base, size); + if (hint == 0 || hint > MI_HINT_MAX) { // wrap or initialize + uintptr_t init = MI_HINT_BASE; + #if (MI_SECURE>0 || MI_DEBUG==0) // security: randomize start of aligned allocations unless in debug mode + uintptr_t r = _mi_heap_random_next(mi_prim_get_default_heap()); + init = init + ((MI_SEGMENT_SIZE * ((r>>17) & 0xFFFFF)) % MI_HINT_AREA); // (randomly 20 bits)*4MiB == 0 to 4TiB + #endif + uintptr_t expected = hint + size; + mi_atomic_cas_strong_acq_rel(&aligned_base, &expected, init); + hint = mi_atomic_add_acq_rel(&aligned_base, size); // this may still give 0 or > MI_HINT_MAX but that is ok, it is a hint after all + } + if (hint%try_alignment != 0) return NULL; + return (void*)hint; +} +#else +void* _mi_os_get_aligned_hint(size_t try_alignment, size_t size) { + MI_UNUSED(try_alignment); MI_UNUSED(size); + return NULL; +} +#endif + + +/* ----------------------------------------------------------- + Free memory +-------------------------------------------------------------- */ + +static void mi_os_free_huge_os_pages(void* p, size_t size, mi_stats_t* stats); + +static void mi_os_prim_free(void* addr, size_t size, bool still_committed, mi_stats_t* tld_stats) { + MI_UNUSED(tld_stats); + mi_assert_internal((size % _mi_os_page_size()) == 0); + if (addr == NULL || size == 0) return; // || _mi_os_is_huge_reserved(addr) + int err = _mi_prim_free(addr, size); + if (err != 0) { + _mi_warning_message("unable to free OS memory (error: %d (0x%x), size: 0x%zx bytes, address: %p)\n", err, err, size, addr); + } + mi_stats_t* stats = &_mi_stats_main; + if (still_committed) { _mi_stat_decrease(&stats->committed, size); } + _mi_stat_decrease(&stats->reserved, size); +} + +void _mi_os_free_ex(void* addr, size_t size, bool still_committed, mi_memid_t memid, mi_stats_t* tld_stats) { + if (mi_memkind_is_os(memid.memkind)) { + size_t csize = _mi_os_good_alloc_size(size); + void* base = addr; + // different base? (due to alignment) + if (memid.mem.os.base != NULL) { + mi_assert(memid.mem.os.base <= addr); + mi_assert((uint8_t*)memid.mem.os.base + memid.mem.os.alignment >= (uint8_t*)addr); + base = memid.mem.os.base; + csize += ((uint8_t*)addr - (uint8_t*)memid.mem.os.base); + } + // free it + if (memid.memkind == MI_MEM_OS_HUGE) { + mi_assert(memid.is_pinned); + mi_os_free_huge_os_pages(base, csize, tld_stats); + } + else { + mi_os_prim_free(base, csize, still_committed, tld_stats); + } + } + else { + // nothing to do + mi_assert(memid.memkind < MI_MEM_OS); + } +} + +void _mi_os_free(void* p, size_t size, mi_memid_t memid, mi_stats_t* tld_stats) { + _mi_os_free_ex(p, size, true, memid, tld_stats); +} + + +/* ----------------------------------------------------------- + Primitive allocation from the OS. +-------------------------------------------------------------- */ + +// Note: the `try_alignment` is just a hint and the returned pointer is not guaranteed to be aligned. +static void* mi_os_prim_alloc(size_t size, size_t try_alignment, bool commit, bool allow_large, bool* is_large, bool* is_zero, mi_stats_t* stats) { + mi_assert_internal(size > 0 && (size % _mi_os_page_size()) == 0); + mi_assert_internal(is_zero != NULL); + mi_assert_internal(is_large != NULL); + if (size == 0) return NULL; + if (!commit) { allow_large = false; } + if (try_alignment == 0) { try_alignment = 1; } // avoid 0 to ensure there will be no divide by zero when aligning + + *is_zero = false; + void* p = NULL; + int err = _mi_prim_alloc(size, try_alignment, commit, allow_large, is_large, is_zero, &p); + if (err != 0) { + _mi_warning_message("unable to allocate OS memory (error: %d (0x%x), size: 0x%zx bytes, align: 0x%zx, commit: %d, allow large: %d)\n", err, err, size, try_alignment, commit, allow_large); + } + mi_stat_counter_increase(stats->mmap_calls, 1); + if (p != NULL) { + _mi_stat_increase(&stats->reserved, size); + if (commit) { + _mi_stat_increase(&stats->committed, size); + // seems needed for asan (or `mimalloc-test-api` fails) + #ifdef MI_TRACK_ASAN + if (*is_zero) { mi_track_mem_defined(p,size); } + else { mi_track_mem_undefined(p,size); } + #endif + } + } + return p; +} + + +// Primitive aligned allocation from the OS. +// This function guarantees the allocated memory is aligned. +static void* mi_os_prim_alloc_aligned(size_t size, size_t alignment, bool commit, bool allow_large, bool* is_large, bool* is_zero, void** base, mi_stats_t* stats) { + mi_assert_internal(alignment >= _mi_os_page_size() && ((alignment & (alignment - 1)) == 0)); + mi_assert_internal(size > 0 && (size % _mi_os_page_size()) == 0); + mi_assert_internal(is_large != NULL); + mi_assert_internal(is_zero != NULL); + mi_assert_internal(base != NULL); + if (!commit) allow_large = false; + if (!(alignment >= _mi_os_page_size() && ((alignment & (alignment - 1)) == 0))) return NULL; + size = _mi_align_up(size, _mi_os_page_size()); + + // try first with a hint (this will be aligned directly on Win 10+ or BSD) + void* p = mi_os_prim_alloc(size, alignment, commit, allow_large, is_large, is_zero, stats); + if (p == NULL) return NULL; + + // aligned already? + if (((uintptr_t)p % alignment) == 0) { + *base = p; + } + else { + // if not aligned, free it, overallocate, and unmap around it + _mi_warning_message("unable to allocate aligned OS memory directly, fall back to over-allocation (size: 0x%zx bytes, address: %p, alignment: 0x%zx, commit: %d)\n", size, p, alignment, commit); + mi_os_prim_free(p, size, commit, stats); + if (size >= (SIZE_MAX - alignment)) return NULL; // overflow + const size_t over_size = size + alignment; + + if (mi_os_mem_config.must_free_whole) { // win32 virtualAlloc cannot free parts of an allocate block + // over-allocate uncommitted (virtual) memory + p = mi_os_prim_alloc(over_size, 1 /*alignment*/, false /* commit? */, false /* allow_large */, is_large, is_zero, stats); + if (p == NULL) return NULL; + + // set p to the aligned part in the full region + // note: this is dangerous on Windows as VirtualFree needs the actual base pointer + // this is handled though by having the `base` field in the memid's + *base = p; // remember the base + p = mi_align_up_ptr(p, alignment); + + // explicitly commit only the aligned part + if (commit) { + _mi_os_commit(p, size, NULL, stats); + } + } + else { // mmap can free inside an allocation + // overallocate... + p = mi_os_prim_alloc(over_size, 1, commit, false, is_large, is_zero, stats); + if (p == NULL) return NULL; + + // and selectively unmap parts around the over-allocated area. (noop on sbrk) + void* aligned_p = mi_align_up_ptr(p, alignment); + size_t pre_size = (uint8_t*)aligned_p - (uint8_t*)p; + size_t mid_size = _mi_align_up(size, _mi_os_page_size()); + size_t post_size = over_size - pre_size - mid_size; + mi_assert_internal(pre_size < over_size&& post_size < over_size&& mid_size >= size); + if (pre_size > 0) { mi_os_prim_free(p, pre_size, commit, stats); } + if (post_size > 0) { mi_os_prim_free((uint8_t*)aligned_p + mid_size, post_size, commit, stats); } + // we can return the aligned pointer on `mmap` (and sbrk) systems + p = aligned_p; + *base = aligned_p; // since we freed the pre part, `*base == p`. + } + } + + mi_assert_internal(p == NULL || (p != NULL && *base != NULL && ((uintptr_t)p % alignment) == 0)); + return p; +} + + +/* ----------------------------------------------------------- + OS API: alloc and alloc_aligned +----------------------------------------------------------- */ + +void* _mi_os_alloc(size_t size, mi_memid_t* memid, mi_stats_t* tld_stats) { + MI_UNUSED(tld_stats); + *memid = _mi_memid_none(); + mi_stats_t* stats = &_mi_stats_main; + if (size == 0) return NULL; + size = _mi_os_good_alloc_size(size); + bool os_is_large = false; + bool os_is_zero = false; + void* p = mi_os_prim_alloc(size, 0, true, false, &os_is_large, &os_is_zero, stats); + if (p != NULL) { + *memid = _mi_memid_create_os(true, os_is_zero, os_is_large); + } + return p; +} + +void* _mi_os_alloc_aligned(size_t size, size_t alignment, bool commit, bool allow_large, mi_memid_t* memid, mi_stats_t* tld_stats) +{ + MI_UNUSED(&_mi_os_get_aligned_hint); // suppress unused warnings + MI_UNUSED(tld_stats); + *memid = _mi_memid_none(); + if (size == 0) return NULL; + size = _mi_os_good_alloc_size(size); + alignment = _mi_align_up(alignment, _mi_os_page_size()); + + bool os_is_large = false; + bool os_is_zero = false; + void* os_base = NULL; + void* p = mi_os_prim_alloc_aligned(size, alignment, commit, allow_large, &os_is_large, &os_is_zero, &os_base, &_mi_stats_main /*tld->stats*/ ); + if (p != NULL) { + *memid = _mi_memid_create_os(commit, os_is_zero, os_is_large); + memid->mem.os.base = os_base; + memid->mem.os.alignment = alignment; + } + return p; +} + +/* ----------------------------------------------------------- + OS aligned allocation with an offset. This is used + for large alignments > MI_ALIGNMENT_MAX. We use a large mimalloc + page where the object can be aligned at an offset from the start of the segment. + As we may need to overallocate, we need to free such pointers using `mi_free_aligned` + to use the actual start of the memory region. +----------------------------------------------------------- */ + +void* _mi_os_alloc_aligned_at_offset(size_t size, size_t alignment, size_t offset, bool commit, bool allow_large, mi_memid_t* memid, mi_stats_t* tld_stats) { + mi_assert(offset <= MI_SEGMENT_SIZE); + mi_assert(offset <= size); + mi_assert((alignment % _mi_os_page_size()) == 0); + *memid = _mi_memid_none(); + if (offset > MI_SEGMENT_SIZE) return NULL; + if (offset == 0) { + // regular aligned allocation + return _mi_os_alloc_aligned(size, alignment, commit, allow_large, memid, tld_stats); + } + else { + // overallocate to align at an offset + const size_t extra = _mi_align_up(offset, alignment) - offset; + const size_t oversize = size + extra; + void* const start = _mi_os_alloc_aligned(oversize, alignment, commit, allow_large, memid, tld_stats); + if (start == NULL) return NULL; + + void* const p = (uint8_t*)start + extra; + mi_assert(_mi_is_aligned((uint8_t*)p + offset, alignment)); + // decommit the overallocation at the start + if (commit && extra > _mi_os_page_size()) { + _mi_os_decommit(start, extra, tld_stats); + } + return p; + } +} + +/* ----------------------------------------------------------- + OS memory API: reset, commit, decommit, protect, unprotect. +----------------------------------------------------------- */ + +// OS page align within a given area, either conservative (pages inside the area only), +// or not (straddling pages outside the area is possible) +static void* mi_os_page_align_areax(bool conservative, void* addr, size_t size, size_t* newsize) { + mi_assert(addr != NULL && size > 0); + if (newsize != NULL) *newsize = 0; + if (size == 0 || addr == NULL) return NULL; + + // page align conservatively within the range + void* start = (conservative ? mi_align_up_ptr(addr, _mi_os_page_size()) + : mi_align_down_ptr(addr, _mi_os_page_size())); + void* end = (conservative ? mi_align_down_ptr((uint8_t*)addr + size, _mi_os_page_size()) + : mi_align_up_ptr((uint8_t*)addr + size, _mi_os_page_size())); + ptrdiff_t diff = (uint8_t*)end - (uint8_t*)start; + if (diff <= 0) return NULL; + + mi_assert_internal((conservative && (size_t)diff <= size) || (!conservative && (size_t)diff >= size)); + if (newsize != NULL) *newsize = (size_t)diff; + return start; +} + +static void* mi_os_page_align_area_conservative(void* addr, size_t size, size_t* newsize) { + return mi_os_page_align_areax(true, addr, size, newsize); +} + +bool _mi_os_commit(void* addr, size_t size, bool* is_zero, mi_stats_t* tld_stats) { + MI_UNUSED(tld_stats); + mi_stats_t* stats = &_mi_stats_main; + if (is_zero != NULL) { *is_zero = false; } + _mi_stat_increase(&stats->committed, size); // use size for precise commit vs. decommit + _mi_stat_counter_increase(&stats->commit_calls, 1); + + // page align range + size_t csize; + void* start = mi_os_page_align_areax(false /* conservative? */, addr, size, &csize); + if (csize == 0) return true; + + // commit + bool os_is_zero = false; + int err = _mi_prim_commit(start, csize, &os_is_zero); + if (err != 0) { + _mi_warning_message("cannot commit OS memory (error: %d (0x%x), address: %p, size: 0x%zx bytes)\n", err, err, start, csize); + return false; + } + if (os_is_zero && is_zero != NULL) { + *is_zero = true; + mi_assert_expensive(mi_mem_is_zero(start, csize)); + } + // note: the following seems required for asan (otherwise `mimalloc-test-stress` fails) + #ifdef MI_TRACK_ASAN + if (os_is_zero) { mi_track_mem_defined(start,csize); } + else { mi_track_mem_undefined(start,csize); } + #endif + return true; +} + +static bool mi_os_decommit_ex(void* addr, size_t size, bool* needs_recommit, mi_stats_t* tld_stats) { + MI_UNUSED(tld_stats); + mi_stats_t* stats = &_mi_stats_main; + mi_assert_internal(needs_recommit!=NULL); + _mi_stat_decrease(&stats->committed, size); + + // page align + size_t csize; + void* start = mi_os_page_align_area_conservative(addr, size, &csize); + if (csize == 0) return true; + + // decommit + *needs_recommit = true; + int err = _mi_prim_decommit(start,csize,needs_recommit); + if (err != 0) { + _mi_warning_message("cannot decommit OS memory (error: %d (0x%x), address: %p, size: 0x%zx bytes)\n", err, err, start, csize); + } + mi_assert_internal(err == 0); + return (err == 0); +} + +bool _mi_os_decommit(void* addr, size_t size, mi_stats_t* tld_stats) { + bool needs_recommit; + return mi_os_decommit_ex(addr, size, &needs_recommit, tld_stats); +} + + +// Signal to the OS that the address range is no longer in use +// but may be used later again. This will release physical memory +// pages and reduce swapping while keeping the memory committed. +// We page align to a conservative area inside the range to reset. +bool _mi_os_reset(void* addr, size_t size, mi_stats_t* stats) { + // page align conservatively within the range + size_t csize; + void* start = mi_os_page_align_area_conservative(addr, size, &csize); + if (csize == 0) return true; // || _mi_os_is_huge_reserved(addr) + _mi_stat_increase(&stats->reset, csize); + _mi_stat_counter_increase(&stats->reset_calls, 1); + + #if (MI_DEBUG>1) && !MI_SECURE && !MI_TRACK_ENABLED // && !MI_TSAN + memset(start, 0, csize); // pretend it is eagerly reset + #endif + + int err = _mi_prim_reset(start, csize); + if (err != 0) { + _mi_warning_message("cannot reset OS memory (error: %d (0x%x), address: %p, size: 0x%zx bytes)\n", err, err, start, csize); + } + return (err == 0); +} + + +// either resets or decommits memory, returns true if the memory needs +// to be recommitted if it is to be re-used later on. +bool _mi_os_purge_ex(void* p, size_t size, bool allow_reset, mi_stats_t* stats) +{ + if (mi_option_get(mi_option_purge_delay) < 0) return false; // is purging allowed? + _mi_stat_counter_increase(&stats->purge_calls, 1); + _mi_stat_increase(&stats->purged, size); + + if (mi_option_is_enabled(mi_option_purge_decommits) && // should decommit? + !_mi_preloading()) // don't decommit during preloading (unsafe) + { + bool needs_recommit = true; + mi_os_decommit_ex(p, size, &needs_recommit, stats); + return needs_recommit; + } + else { + if (allow_reset) { // this can sometimes be not allowed if the range is not fully committed + _mi_os_reset(p, size, stats); + } + return false; // needs no recommit + } +} + +// either resets or decommits memory, returns true if the memory needs +// to be recommitted if it is to be re-used later on. +bool _mi_os_purge(void* p, size_t size, mi_stats_t * stats) { + return _mi_os_purge_ex(p, size, true, stats); +} + +// Protect a region in memory to be not accessible. +static bool mi_os_protectx(void* addr, size_t size, bool protect) { + // page align conservatively within the range + size_t csize = 0; + void* start = mi_os_page_align_area_conservative(addr, size, &csize); + if (csize == 0) return false; + /* + if (_mi_os_is_huge_reserved(addr)) { + _mi_warning_message("cannot mprotect memory allocated in huge OS pages\n"); + } + */ + int err = _mi_prim_protect(start,csize,protect); + if (err != 0) { + _mi_warning_message("cannot %s OS memory (error: %d (0x%x), address: %p, size: 0x%zx bytes)\n", (protect ? "protect" : "unprotect"), err, err, start, csize); + } + return (err == 0); +} + +bool _mi_os_protect(void* addr, size_t size) { + return mi_os_protectx(addr, size, true); +} + +bool _mi_os_unprotect(void* addr, size_t size) { + return mi_os_protectx(addr, size, false); +} + + + +/* ---------------------------------------------------------------------------- +Support for allocating huge OS pages (1Gib) that are reserved up-front +and possibly associated with a specific NUMA node. (use `numa_node>=0`) +-----------------------------------------------------------------------------*/ +#define MI_HUGE_OS_PAGE_SIZE (MI_GiB) + + +#if (MI_INTPTR_SIZE >= 8) +// To ensure proper alignment, use our own area for huge OS pages +static mi_decl_cache_align _Atomic(uintptr_t) mi_huge_start; // = 0 + +// Claim an aligned address range for huge pages +static uint8_t* mi_os_claim_huge_pages(size_t pages, size_t* total_size) { + if (total_size != NULL) *total_size = 0; + const size_t size = pages * MI_HUGE_OS_PAGE_SIZE; + + uintptr_t start = 0; + uintptr_t end = 0; + uintptr_t huge_start = mi_atomic_load_relaxed(&mi_huge_start); + do { + start = huge_start; + if (start == 0) { + // Initialize the start address after the 32TiB area + start = ((uintptr_t)32 << 40); // 32TiB virtual start address + #if (MI_SECURE>0 || MI_DEBUG==0) // security: randomize start of huge pages unless in debug mode + uintptr_t r = _mi_heap_random_next(mi_prim_get_default_heap()); + start = start + ((uintptr_t)MI_HUGE_OS_PAGE_SIZE * ((r>>17) & 0x0FFF)); // (randomly 12bits)*1GiB == between 0 to 4TiB + #endif + } + end = start + size; + mi_assert_internal(end % MI_SEGMENT_SIZE == 0); + } while (!mi_atomic_cas_strong_acq_rel(&mi_huge_start, &huge_start, end)); + + if (total_size != NULL) *total_size = size; + return (uint8_t*)start; +} +#else +static uint8_t* mi_os_claim_huge_pages(size_t pages, size_t* total_size) { + MI_UNUSED(pages); + if (total_size != NULL) *total_size = 0; + return NULL; +} +#endif + +// Allocate MI_SEGMENT_SIZE aligned huge pages +void* _mi_os_alloc_huge_os_pages(size_t pages, int numa_node, mi_msecs_t max_msecs, size_t* pages_reserved, size_t* psize, mi_memid_t* memid) { + *memid = _mi_memid_none(); + if (psize != NULL) *psize = 0; + if (pages_reserved != NULL) *pages_reserved = 0; + size_t size = 0; + uint8_t* start = mi_os_claim_huge_pages(pages, &size); + if (start == NULL) return NULL; // or 32-bit systems + + // Allocate one page at the time but try to place them contiguously + // We allocate one page at the time to be able to abort if it takes too long + // or to at least allocate as many as available on the system. + mi_msecs_t start_t = _mi_clock_start(); + size_t page = 0; + bool all_zero = true; + while (page < pages) { + // allocate a page + bool is_zero = false; + void* addr = start + (page * MI_HUGE_OS_PAGE_SIZE); + void* p = NULL; + int err = _mi_prim_alloc_huge_os_pages(addr, MI_HUGE_OS_PAGE_SIZE, numa_node, &is_zero, &p); + if (!is_zero) { all_zero = false; } + if (err != 0) { + _mi_warning_message("unable to allocate huge OS page (error: %d (0x%x), address: %p, size: %zx bytes)\n", err, err, addr, MI_HUGE_OS_PAGE_SIZE); + break; + } + + // Did we succeed at a contiguous address? + if (p != addr) { + // no success, issue a warning and break + if (p != NULL) { + _mi_warning_message("could not allocate contiguous huge OS page %zu at %p\n", page, addr); + mi_os_prim_free(p, MI_HUGE_OS_PAGE_SIZE, true, &_mi_stats_main); + } + break; + } + + // success, record it + page++; // increase before timeout check (see issue #711) + _mi_stat_increase(&_mi_stats_main.committed, MI_HUGE_OS_PAGE_SIZE); + _mi_stat_increase(&_mi_stats_main.reserved, MI_HUGE_OS_PAGE_SIZE); + + // check for timeout + if (max_msecs > 0) { + mi_msecs_t elapsed = _mi_clock_end(start_t); + if (page >= 1) { + mi_msecs_t estimate = ((elapsed / (page+1)) * pages); + if (estimate > 2*max_msecs) { // seems like we are going to timeout, break + elapsed = max_msecs + 1; + } + } + if (elapsed > max_msecs) { + _mi_warning_message("huge OS page allocation timed out (after allocating %zu page(s))\n", page); + break; + } + } + } + mi_assert_internal(page*MI_HUGE_OS_PAGE_SIZE <= size); + if (pages_reserved != NULL) { *pages_reserved = page; } + if (psize != NULL) { *psize = page * MI_HUGE_OS_PAGE_SIZE; } + if (page != 0) { + mi_assert(start != NULL); + *memid = _mi_memid_create_os(true /* is committed */, all_zero, true /* is_large */); + memid->memkind = MI_MEM_OS_HUGE; + mi_assert(memid->is_pinned); + #ifdef MI_TRACK_ASAN + if (all_zero) { mi_track_mem_defined(start,size); } + #endif + } + return (page == 0 ? NULL : start); +} + +// free every huge page in a range individually (as we allocated per page) +// note: needed with VirtualAlloc but could potentially be done in one go on mmap'd systems. +static void mi_os_free_huge_os_pages(void* p, size_t size, mi_stats_t* stats) { + if (p==NULL || size==0) return; + uint8_t* base = (uint8_t*)p; + while (size >= MI_HUGE_OS_PAGE_SIZE) { + mi_os_prim_free(base, MI_HUGE_OS_PAGE_SIZE, true, stats); + size -= MI_HUGE_OS_PAGE_SIZE; + base += MI_HUGE_OS_PAGE_SIZE; + } +} + +/* ---------------------------------------------------------------------------- +Support NUMA aware allocation +-----------------------------------------------------------------------------*/ + +_Atomic(size_t) _mi_numa_node_count; // = 0 // cache the node count + +size_t _mi_os_numa_node_count_get(void) { + size_t count = mi_atomic_load_acquire(&_mi_numa_node_count); + if (count <= 0) { + long ncount = mi_option_get(mi_option_use_numa_nodes); // given explicitly? + if (ncount > 0) { + count = (size_t)ncount; + } + else { + count = _mi_prim_numa_node_count(); // or detect dynamically + if (count == 0) count = 1; + } + mi_atomic_store_release(&_mi_numa_node_count, count); // save it + _mi_verbose_message("using %zd numa regions\n", count); + } + return count; +} + +int _mi_os_numa_node_get(mi_os_tld_t* tld) { + MI_UNUSED(tld); + size_t numa_count = _mi_os_numa_node_count(); + if (numa_count<=1) return 0; // optimize on single numa node systems: always node 0 + // never more than the node count and >= 0 + size_t numa_node = _mi_prim_numa_node(); + if (numa_node >= numa_count) { numa_node = numa_node % numa_count; } + return (int)numa_node; +} diff --git a/compat/mimalloc/page-queue.c b/compat/mimalloc/page-queue.c new file mode 100644 index 00000000000000..5619a81f9917fe --- /dev/null +++ b/compat/mimalloc/page-queue.c @@ -0,0 +1,332 @@ +/*---------------------------------------------------------------------------- +Copyright (c) 2018-2020, Microsoft Research, Daan Leijen +This is free software; you can redistribute it and/or modify it under the +terms of the MIT license. A copy of the license can be found in the file +"LICENSE" at the root of this distribution. +-----------------------------------------------------------------------------*/ + +/* ----------------------------------------------------------- + Definition of page queues for each block size +----------------------------------------------------------- */ + +#ifndef MI_IN_PAGE_C +#error "this file should be included from 'page.c'" +#endif + +/* ----------------------------------------------------------- + Minimal alignment in machine words (i.e. `sizeof(void*)`) +----------------------------------------------------------- */ + +#if (MI_MAX_ALIGN_SIZE > 4*MI_INTPTR_SIZE) + #error "define alignment for more than 4x word size for this platform" +#elif (MI_MAX_ALIGN_SIZE > 2*MI_INTPTR_SIZE) + #define MI_ALIGN4W // 4 machine words minimal alignment +#elif (MI_MAX_ALIGN_SIZE > MI_INTPTR_SIZE) + #define MI_ALIGN2W // 2 machine words minimal alignment +#else + // ok, default alignment is 1 word +#endif + + +/* ----------------------------------------------------------- + Queue query +----------------------------------------------------------- */ + + +static inline bool mi_page_queue_is_huge(const mi_page_queue_t* pq) { + return (pq->block_size == (MI_MEDIUM_OBJ_SIZE_MAX+sizeof(uintptr_t))); +} + +static inline bool mi_page_queue_is_full(const mi_page_queue_t* pq) { + return (pq->block_size == (MI_MEDIUM_OBJ_SIZE_MAX+(2*sizeof(uintptr_t)))); +} + +static inline bool mi_page_queue_is_special(const mi_page_queue_t* pq) { + return (pq->block_size > MI_MEDIUM_OBJ_SIZE_MAX); +} + +/* ----------------------------------------------------------- + Bins +----------------------------------------------------------- */ + +// Return the bin for a given field size. +// Returns MI_BIN_HUGE if the size is too large. +// We use `wsize` for the size in "machine word sizes", +// i.e. byte size == `wsize*sizeof(void*)`. +static inline uint8_t mi_bin(size_t size) { + size_t wsize = _mi_wsize_from_size(size); + uint8_t bin; + if (wsize <= 1) { + bin = 1; + } + #if defined(MI_ALIGN4W) + else if (wsize <= 4) { + bin = (uint8_t)((wsize+1)&~1); // round to double word sizes + } + #elif defined(MI_ALIGN2W) + else if (wsize <= 8) { + bin = (uint8_t)((wsize+1)&~1); // round to double word sizes + } + #else + else if (wsize <= 8) { + bin = (uint8_t)wsize; + } + #endif + else if (wsize > MI_MEDIUM_OBJ_WSIZE_MAX) { + bin = MI_BIN_HUGE; + } + else { + #if defined(MI_ALIGN4W) + if (wsize <= 16) { wsize = (wsize+3)&~3; } // round to 4x word sizes + #endif + wsize--; + // find the highest bit + uint8_t b = (uint8_t)mi_bsr(wsize); // note: wsize != 0 + // and use the top 3 bits to determine the bin (~12.5% worst internal fragmentation). + // - adjust with 3 because we use do not round the first 8 sizes + // which each get an exact bin + bin = ((b << 2) + (uint8_t)((wsize >> (b - 2)) & 0x03)) - 3; + mi_assert_internal(bin < MI_BIN_HUGE); + } + mi_assert_internal(bin > 0 && bin <= MI_BIN_HUGE); + return bin; +} + + + +/* ----------------------------------------------------------- + Queue of pages with free blocks +----------------------------------------------------------- */ + +uint8_t _mi_bin(size_t size) { + return mi_bin(size); +} + +size_t _mi_bin_size(uint8_t bin) { + return _mi_heap_empty.pages[bin].block_size; +} + +// Good size for allocation +size_t mi_good_size(size_t size) mi_attr_noexcept { + if (size <= MI_MEDIUM_OBJ_SIZE_MAX) { + return _mi_bin_size(mi_bin(size)); + } + else { + return _mi_align_up(size,_mi_os_page_size()); + } +} + +#if (MI_DEBUG>1) +static bool mi_page_queue_contains(mi_page_queue_t* queue, const mi_page_t* page) { + mi_assert_internal(page != NULL); + mi_page_t* list = queue->first; + while (list != NULL) { + mi_assert_internal(list->next == NULL || list->next->prev == list); + mi_assert_internal(list->prev == NULL || list->prev->next == list); + if (list == page) break; + list = list->next; + } + return (list == page); +} + +#endif + +#if (MI_DEBUG>1) +static bool mi_heap_contains_queue(const mi_heap_t* heap, const mi_page_queue_t* pq) { + return (pq >= &heap->pages[0] && pq <= &heap->pages[MI_BIN_FULL]); +} +#endif + +static mi_page_queue_t* mi_page_queue_of(const mi_page_t* page) { + uint8_t bin = (mi_page_is_in_full(page) ? MI_BIN_FULL : mi_bin(page->xblock_size)); + mi_heap_t* heap = mi_page_heap(page); + mi_assert_internal(heap != NULL && bin <= MI_BIN_FULL); + mi_page_queue_t* pq = &heap->pages[bin]; + mi_assert_internal(bin >= MI_BIN_HUGE || page->xblock_size == pq->block_size); + mi_assert_expensive(mi_page_queue_contains(pq, page)); + return pq; +} + +static mi_page_queue_t* mi_heap_page_queue_of(mi_heap_t* heap, const mi_page_t* page) { + uint8_t bin = (mi_page_is_in_full(page) ? MI_BIN_FULL : mi_bin(page->xblock_size)); + mi_assert_internal(bin <= MI_BIN_FULL); + mi_page_queue_t* pq = &heap->pages[bin]; + mi_assert_internal(mi_page_is_in_full(page) || page->xblock_size == pq->block_size); + return pq; +} + +// The current small page array is for efficiency and for each +// small size (up to 256) it points directly to the page for that +// size without having to compute the bin. This means when the +// current free page queue is updated for a small bin, we need to update a +// range of entries in `_mi_page_small_free`. +static inline void mi_heap_queue_first_update(mi_heap_t* heap, const mi_page_queue_t* pq) { + mi_assert_internal(mi_heap_contains_queue(heap,pq)); + size_t size = pq->block_size; + if (size > MI_SMALL_SIZE_MAX) return; + + mi_page_t* page = pq->first; + if (pq->first == NULL) page = (mi_page_t*)&_mi_page_empty; + + // find index in the right direct page array + size_t start; + size_t idx = _mi_wsize_from_size(size); + mi_page_t** pages_free = heap->pages_free_direct; + + if (pages_free[idx] == page) return; // already set + + // find start slot + if (idx<=1) { + start = 0; + } + else { + // find previous size; due to minimal alignment upto 3 previous bins may need to be skipped + uint8_t bin = mi_bin(size); + const mi_page_queue_t* prev = pq - 1; + while( bin == mi_bin(prev->block_size) && prev > &heap->pages[0]) { + prev--; + } + start = 1 + _mi_wsize_from_size(prev->block_size); + if (start > idx) start = idx; + } + + // set size range to the right page + mi_assert(start <= idx); + for (size_t sz = start; sz <= idx; sz++) { + pages_free[sz] = page; + } +} + +/* +static bool mi_page_queue_is_empty(mi_page_queue_t* queue) { + return (queue->first == NULL); +} +*/ + +static void mi_page_queue_remove(mi_page_queue_t* queue, mi_page_t* page) { + mi_assert_internal(page != NULL); + mi_assert_expensive(mi_page_queue_contains(queue, page)); + mi_assert_internal(page->xblock_size == queue->block_size || (page->xblock_size > MI_MEDIUM_OBJ_SIZE_MAX && mi_page_queue_is_huge(queue)) || (mi_page_is_in_full(page) && mi_page_queue_is_full(queue))); + mi_heap_t* heap = mi_page_heap(page); + + if (page->prev != NULL) page->prev->next = page->next; + if (page->next != NULL) page->next->prev = page->prev; + if (page == queue->last) queue->last = page->prev; + if (page == queue->first) { + queue->first = page->next; + // update first + mi_assert_internal(mi_heap_contains_queue(heap, queue)); + mi_heap_queue_first_update(heap,queue); + } + heap->page_count--; + page->next = NULL; + page->prev = NULL; + // mi_atomic_store_ptr_release(mi_atomic_cast(void*, &page->heap), NULL); + mi_page_set_in_full(page,false); +} + + +static void mi_page_queue_push(mi_heap_t* heap, mi_page_queue_t* queue, mi_page_t* page) { + mi_assert_internal(mi_page_heap(page) == heap); + mi_assert_internal(!mi_page_queue_contains(queue, page)); + #if MI_HUGE_PAGE_ABANDON + mi_assert_internal(_mi_page_segment(page)->kind != MI_SEGMENT_HUGE); + #endif + mi_assert_internal(page->xblock_size == queue->block_size || + (page->xblock_size > MI_MEDIUM_OBJ_SIZE_MAX) || + (mi_page_is_in_full(page) && mi_page_queue_is_full(queue))); + + mi_page_set_in_full(page, mi_page_queue_is_full(queue)); + // mi_atomic_store_ptr_release(mi_atomic_cast(void*, &page->heap), heap); + page->next = queue->first; + page->prev = NULL; + if (queue->first != NULL) { + mi_assert_internal(queue->first->prev == NULL); + queue->first->prev = page; + queue->first = page; + } + else { + queue->first = queue->last = page; + } + + // update direct + mi_heap_queue_first_update(heap, queue); + heap->page_count++; +} + + +static void mi_page_queue_enqueue_from(mi_page_queue_t* to, mi_page_queue_t* from, mi_page_t* page) { + mi_assert_internal(page != NULL); + mi_assert_expensive(mi_page_queue_contains(from, page)); + mi_assert_expensive(!mi_page_queue_contains(to, page)); + + mi_assert_internal((page->xblock_size == to->block_size && page->xblock_size == from->block_size) || + (page->xblock_size == to->block_size && mi_page_queue_is_full(from)) || + (page->xblock_size == from->block_size && mi_page_queue_is_full(to)) || + (page->xblock_size > MI_LARGE_OBJ_SIZE_MAX && mi_page_queue_is_huge(to)) || + (page->xblock_size > MI_LARGE_OBJ_SIZE_MAX && mi_page_queue_is_full(to))); + + mi_heap_t* heap = mi_page_heap(page); + if (page->prev != NULL) page->prev->next = page->next; + if (page->next != NULL) page->next->prev = page->prev; + if (page == from->last) from->last = page->prev; + if (page == from->first) { + from->first = page->next; + // update first + mi_assert_internal(mi_heap_contains_queue(heap, from)); + mi_heap_queue_first_update(heap, from); + } + + page->prev = to->last; + page->next = NULL; + if (to->last != NULL) { + mi_assert_internal(heap == mi_page_heap(to->last)); + to->last->next = page; + to->last = page; + } + else { + to->first = page; + to->last = page; + mi_heap_queue_first_update(heap, to); + } + + mi_page_set_in_full(page, mi_page_queue_is_full(to)); +} + +// Only called from `mi_heap_absorb`. +size_t _mi_page_queue_append(mi_heap_t* heap, mi_page_queue_t* pq, mi_page_queue_t* append) { + mi_assert_internal(mi_heap_contains_queue(heap,pq)); + mi_assert_internal(pq->block_size == append->block_size); + + if (append->first==NULL) return 0; + + // set append pages to new heap and count + size_t count = 0; + for (mi_page_t* page = append->first; page != NULL; page = page->next) { + // inline `mi_page_set_heap` to avoid wrong assertion during absorption; + // in this case it is ok to be delayed freeing since both "to" and "from" heap are still alive. + mi_atomic_store_release(&page->xheap, (uintptr_t)heap); + // set the flag to delayed free (not overriding NEVER_DELAYED_FREE) which has as a + // side effect that it spins until any DELAYED_FREEING is finished. This ensures + // that after appending only the new heap will be used for delayed free operations. + _mi_page_use_delayed_free(page, MI_USE_DELAYED_FREE, false); + count++; + } + + if (pq->last==NULL) { + // take over afresh + mi_assert_internal(pq->first==NULL); + pq->first = append->first; + pq->last = append->last; + mi_heap_queue_first_update(heap, pq); + } + else { + // append to end + mi_assert_internal(pq->last!=NULL); + mi_assert_internal(append->first!=NULL); + pq->last->next = append->first; + append->first->prev = pq->last; + pq->last = append->last; + } + return count; +} diff --git a/compat/mimalloc/page.c b/compat/mimalloc/page.c new file mode 100644 index 00000000000000..211204aa79e59d --- /dev/null +++ b/compat/mimalloc/page.c @@ -0,0 +1,939 @@ +/*---------------------------------------------------------------------------- +Copyright (c) 2018-2020, Microsoft Research, Daan Leijen +This is free software; you can redistribute it and/or modify it under the +terms of the MIT license. A copy of the license can be found in the file +"LICENSE" at the root of this distribution. +-----------------------------------------------------------------------------*/ + +/* ----------------------------------------------------------- + The core of the allocator. Every segment contains + pages of a certain block size. The main function + exported is `mi_malloc_generic`. +----------------------------------------------------------- */ + +#include "mimalloc.h" +#include "mimalloc/internal.h" +#include "mimalloc/atomic.h" + +/* ----------------------------------------------------------- + Definition of page queues for each block size +----------------------------------------------------------- */ + +#define MI_IN_PAGE_C +#include "page-queue.c" +#undef MI_IN_PAGE_C + + +/* ----------------------------------------------------------- + Page helpers +----------------------------------------------------------- */ + +// Index a block in a page +static inline mi_block_t* mi_page_block_at(const mi_page_t* page, void* page_start, size_t block_size, size_t i) { + MI_UNUSED(page); + mi_assert_internal(page != NULL); + mi_assert_internal(i <= page->reserved); + return (mi_block_t*)((uint8_t*)page_start + (i * block_size)); +} + +static void mi_page_init(mi_heap_t* heap, mi_page_t* page, size_t size, mi_tld_t* tld); +static void mi_page_extend_free(mi_heap_t* heap, mi_page_t* page, mi_tld_t* tld); + +#if (MI_DEBUG>=3) +static size_t mi_page_list_count(mi_page_t* page, mi_block_t* head) { + size_t count = 0; + while (head != NULL) { + mi_assert_internal(page == _mi_ptr_page(head)); + count++; + head = mi_block_next(page, head); + } + return count; +} + +/* +// Start of the page available memory +static inline uint8_t* mi_page_area(const mi_page_t* page) { + return _mi_page_start(_mi_page_segment(page), page, NULL); +} +*/ + +static bool mi_page_list_is_valid(mi_page_t* page, mi_block_t* p) { + size_t psize; + uint8_t* page_area = _mi_page_start(_mi_page_segment(page), page, &psize); + mi_block_t* start = (mi_block_t*)page_area; + mi_block_t* end = (mi_block_t*)(page_area + psize); + while(p != NULL) { + if (p < start || p >= end) return false; + p = mi_block_next(page, p); + } +#if MI_DEBUG>3 // generally too expensive to check this + if (page->free_is_zero) { + const size_t ubsize = mi_page_usable_block_size(page); + for (mi_block_t* block = page->free; block != NULL; block = mi_block_next(page, block)) { + mi_assert_expensive(mi_mem_is_zero(block + 1, ubsize - sizeof(mi_block_t))); + } + } +#endif + return true; +} + +static bool mi_page_is_valid_init(mi_page_t* page) { + mi_assert_internal(page->xblock_size > 0); + mi_assert_internal(page->used <= page->capacity); + mi_assert_internal(page->capacity <= page->reserved); + + mi_segment_t* segment = _mi_page_segment(page); + uint8_t* start = _mi_page_start(segment,page,NULL); + mi_assert_internal(start == _mi_segment_page_start(segment,page,NULL)); + //const size_t bsize = mi_page_block_size(page); + //mi_assert_internal(start + page->capacity*page->block_size == page->top); + + mi_assert_internal(mi_page_list_is_valid(page,page->free)); + mi_assert_internal(mi_page_list_is_valid(page,page->local_free)); + + #if MI_DEBUG>3 // generally too expensive to check this + if (page->free_is_zero) { + const size_t ubsize = mi_page_usable_block_size(page); + for(mi_block_t* block = page->free; block != NULL; block = mi_block_next(page,block)) { + mi_assert_expensive(mi_mem_is_zero(block + 1, ubsize - sizeof(mi_block_t))); + } + } + #endif + + #if !MI_TRACK_ENABLED && !MI_TSAN + mi_block_t* tfree = mi_page_thread_free(page); + mi_assert_internal(mi_page_list_is_valid(page, tfree)); + //size_t tfree_count = mi_page_list_count(page, tfree); + //mi_assert_internal(tfree_count <= page->thread_freed + 1); + #endif + + size_t free_count = mi_page_list_count(page, page->free) + mi_page_list_count(page, page->local_free); + mi_assert_internal(page->used + free_count == page->capacity); + + return true; +} + +extern bool _mi_process_is_initialized; // has mi_process_init been called? + +bool _mi_page_is_valid(mi_page_t* page) { + mi_assert_internal(mi_page_is_valid_init(page)); + #if MI_SECURE + mi_assert_internal(page->keys[0] != 0); + #endif + if (mi_page_heap(page)!=NULL) { + mi_segment_t* segment = _mi_page_segment(page); + + mi_assert_internal(!_mi_process_is_initialized || segment->thread_id==0 || segment->thread_id == mi_page_heap(page)->thread_id); + #if MI_HUGE_PAGE_ABANDON + if (segment->kind != MI_SEGMENT_HUGE) + #endif + { + mi_page_queue_t* pq = mi_page_queue_of(page); + mi_assert_internal(mi_page_queue_contains(pq, page)); + mi_assert_internal(pq->block_size==mi_page_block_size(page) || mi_page_block_size(page) > MI_MEDIUM_OBJ_SIZE_MAX || mi_page_is_in_full(page)); + mi_assert_internal(mi_heap_contains_queue(mi_page_heap(page),pq)); + } + } + return true; +} +#endif + +void _mi_page_use_delayed_free(mi_page_t* page, mi_delayed_t delay, bool override_never) { + while (!_mi_page_try_use_delayed_free(page, delay, override_never)) { + mi_atomic_yield(); + } +} + +bool _mi_page_try_use_delayed_free(mi_page_t* page, mi_delayed_t delay, bool override_never) { + mi_thread_free_t tfreex; + mi_delayed_t old_delay; + mi_thread_free_t tfree; + size_t yield_count = 0; + do { + tfree = mi_atomic_load_acquire(&page->xthread_free); // note: must acquire as we can break/repeat this loop and not do a CAS; + tfreex = mi_tf_set_delayed(tfree, delay); + old_delay = mi_tf_delayed(tfree); + if mi_unlikely(old_delay == MI_DELAYED_FREEING) { + if (yield_count >= 4) return false; // give up after 4 tries + yield_count++; + mi_atomic_yield(); // delay until outstanding MI_DELAYED_FREEING are done. + // tfree = mi_tf_set_delayed(tfree, MI_NO_DELAYED_FREE); // will cause CAS to busy fail + } + else if (delay == old_delay) { + break; // avoid atomic operation if already equal + } + else if (!override_never && old_delay == MI_NEVER_DELAYED_FREE) { + break; // leave never-delayed flag set + } + } while ((old_delay == MI_DELAYED_FREEING) || + !mi_atomic_cas_weak_release(&page->xthread_free, &tfree, tfreex)); + + return true; // success +} + +/* ----------------------------------------------------------- + Page collect the `local_free` and `thread_free` lists +----------------------------------------------------------- */ + +// Collect the local `thread_free` list using an atomic exchange. +// Note: The exchange must be done atomically as this is used right after +// moving to the full list in `mi_page_collect_ex` and we need to +// ensure that there was no race where the page became unfull just before the move. +static void _mi_page_thread_free_collect(mi_page_t* page) +{ + mi_block_t* head; + mi_thread_free_t tfreex; + mi_thread_free_t tfree = mi_atomic_load_relaxed(&page->xthread_free); + do { + head = mi_tf_block(tfree); + tfreex = mi_tf_set_block(tfree,NULL); + } while (!mi_atomic_cas_weak_acq_rel(&page->xthread_free, &tfree, tfreex)); + + // return if the list is empty + if (head == NULL) return; + + // find the tail -- also to get a proper count (without data races) + uint32_t max_count = page->capacity; // cannot collect more than capacity + uint32_t count = 1; + mi_block_t* tail = head; + mi_block_t* next; + while ((next = mi_block_next(page,tail)) != NULL && count <= max_count) { + count++; + tail = next; + } + // if `count > max_count` there was a memory corruption (possibly infinite list due to double multi-threaded free) + if (count > max_count) { + _mi_error_message(EFAULT, "corrupted thread-free list\n"); + return; // the thread-free items cannot be freed + } + + // and append the current local free list + mi_block_set_next(page,tail, page->local_free); + page->local_free = head; + + // update counts now + page->used -= count; +} + +void _mi_page_free_collect(mi_page_t* page, bool force) { + mi_assert_internal(page!=NULL); + + // collect the thread free list + if (force || mi_page_thread_free(page) != NULL) { // quick test to avoid an atomic operation + _mi_page_thread_free_collect(page); + } + + // and the local free list + if (page->local_free != NULL) { + if mi_likely(page->free == NULL) { + // usual case + page->free = page->local_free; + page->local_free = NULL; + page->free_is_zero = false; + } + else if (force) { + // append -- only on shutdown (force) as this is a linear operation + mi_block_t* tail = page->local_free; + mi_block_t* next; + while ((next = mi_block_next(page, tail)) != NULL) { + tail = next; + } + mi_block_set_next(page, tail, page->free); + page->free = page->local_free; + page->local_free = NULL; + page->free_is_zero = false; + } + } + + mi_assert_internal(!force || page->local_free == NULL); +} + + + +/* ----------------------------------------------------------- + Page fresh and retire +----------------------------------------------------------- */ + +// called from segments when reclaiming abandoned pages +void _mi_page_reclaim(mi_heap_t* heap, mi_page_t* page) { + mi_assert_expensive(mi_page_is_valid_init(page)); + + mi_assert_internal(mi_page_heap(page) == heap); + mi_assert_internal(mi_page_thread_free_flag(page) != MI_NEVER_DELAYED_FREE); + #if MI_HUGE_PAGE_ABANDON + mi_assert_internal(_mi_page_segment(page)->kind != MI_SEGMENT_HUGE); + #endif + + // TODO: push on full queue immediately if it is full? + mi_page_queue_t* pq = mi_page_queue(heap, mi_page_block_size(page)); + mi_page_queue_push(heap, pq, page); + mi_assert_expensive(_mi_page_is_valid(page)); +} + +// allocate a fresh page from a segment +static mi_page_t* mi_page_fresh_alloc(mi_heap_t* heap, mi_page_queue_t* pq, size_t block_size, size_t page_alignment) { + #if !MI_HUGE_PAGE_ABANDON + mi_assert_internal(pq != NULL); + mi_assert_internal(mi_heap_contains_queue(heap, pq)); + mi_assert_internal(page_alignment > 0 || block_size > MI_MEDIUM_OBJ_SIZE_MAX || block_size == pq->block_size); + #endif + mi_page_t* page = _mi_segment_page_alloc(heap, block_size, page_alignment, &heap->tld->segments, &heap->tld->os); + if (page == NULL) { + // this may be out-of-memory, or an abandoned page was reclaimed (and in our queue) + return NULL; + } + mi_assert_internal(page_alignment >0 || block_size > MI_MEDIUM_OBJ_SIZE_MAX || _mi_page_segment(page)->kind != MI_SEGMENT_HUGE); + mi_assert_internal(pq!=NULL || page->xblock_size != 0); + mi_assert_internal(pq!=NULL || mi_page_block_size(page) >= block_size); + // a fresh page was found, initialize it + const size_t full_block_size = ((pq == NULL || mi_page_queue_is_huge(pq)) ? mi_page_block_size(page) : block_size); // see also: mi_segment_huge_page_alloc + mi_assert_internal(full_block_size >= block_size); + mi_page_init(heap, page, full_block_size, heap->tld); + mi_heap_stat_increase(heap, pages, 1); + if (pq != NULL) { mi_page_queue_push(heap, pq, page); } + mi_assert_expensive(_mi_page_is_valid(page)); + return page; +} + +// Get a fresh page to use +static mi_page_t* mi_page_fresh(mi_heap_t* heap, mi_page_queue_t* pq) { + mi_assert_internal(mi_heap_contains_queue(heap, pq)); + mi_page_t* page = mi_page_fresh_alloc(heap, pq, pq->block_size, 0); + if (page==NULL) return NULL; + mi_assert_internal(pq->block_size==mi_page_block_size(page)); + mi_assert_internal(pq==mi_page_queue(heap, mi_page_block_size(page))); + return page; +} + +/* ----------------------------------------------------------- + Do any delayed frees + (put there by other threads if they deallocated in a full page) +----------------------------------------------------------- */ +void _mi_heap_delayed_free_all(mi_heap_t* heap) { + while (!_mi_heap_delayed_free_partial(heap)) { + mi_atomic_yield(); + } +} + +// returns true if all delayed frees were processed +bool _mi_heap_delayed_free_partial(mi_heap_t* heap) { + // take over the list (note: no atomic exchange since it is often NULL) + mi_block_t* block = mi_atomic_load_ptr_relaxed(mi_block_t, &heap->thread_delayed_free); + while (block != NULL && !mi_atomic_cas_ptr_weak_acq_rel(mi_block_t, &heap->thread_delayed_free, &block, NULL)) { /* nothing */ }; + bool all_freed = true; + + // and free them all + while(block != NULL) { + mi_block_t* next = mi_block_nextx(heap,block, heap->keys); + // use internal free instead of regular one to keep stats etc correct + if (!_mi_free_delayed_block(block)) { + // we might already start delayed freeing while another thread has not yet + // reset the delayed_freeing flag; in that case delay it further by reinserting the current block + // into the delayed free list + all_freed = false; + mi_block_t* dfree = mi_atomic_load_ptr_relaxed(mi_block_t, &heap->thread_delayed_free); + do { + mi_block_set_nextx(heap, block, dfree, heap->keys); + } while (!mi_atomic_cas_ptr_weak_release(mi_block_t,&heap->thread_delayed_free, &dfree, block)); + } + block = next; + } + return all_freed; +} + +/* ----------------------------------------------------------- + Unfull, abandon, free and retire +----------------------------------------------------------- */ + +// Move a page from the full list back to a regular list +void _mi_page_unfull(mi_page_t* page) { + mi_assert_internal(page != NULL); + mi_assert_expensive(_mi_page_is_valid(page)); + mi_assert_internal(mi_page_is_in_full(page)); + if (!mi_page_is_in_full(page)) return; + + mi_heap_t* heap = mi_page_heap(page); + mi_page_queue_t* pqfull = &heap->pages[MI_BIN_FULL]; + mi_page_set_in_full(page, false); // to get the right queue + mi_page_queue_t* pq = mi_heap_page_queue_of(heap, page); + mi_page_set_in_full(page, true); + mi_page_queue_enqueue_from(pq, pqfull, page); +} + +static void mi_page_to_full(mi_page_t* page, mi_page_queue_t* pq) { + mi_assert_internal(pq == mi_page_queue_of(page)); + mi_assert_internal(!mi_page_immediate_available(page)); + mi_assert_internal(!mi_page_is_in_full(page)); + + if (mi_page_is_in_full(page)) return; + mi_page_queue_enqueue_from(&mi_page_heap(page)->pages[MI_BIN_FULL], pq, page); + _mi_page_free_collect(page,false); // try to collect right away in case another thread freed just before MI_USE_DELAYED_FREE was set +} + + +// Abandon a page with used blocks at the end of a thread. +// Note: only call if it is ensured that no references exist from +// the `page->heap->thread_delayed_free` into this page. +// Currently only called through `mi_heap_collect_ex` which ensures this. +void _mi_page_abandon(mi_page_t* page, mi_page_queue_t* pq) { + mi_assert_internal(page != NULL); + mi_assert_expensive(_mi_page_is_valid(page)); + mi_assert_internal(pq == mi_page_queue_of(page)); + mi_assert_internal(mi_page_heap(page) != NULL); + + mi_heap_t* pheap = mi_page_heap(page); + + // remove from our page list + mi_segments_tld_t* segments_tld = &pheap->tld->segments; + mi_page_queue_remove(pq, page); + + // page is no longer associated with our heap + mi_assert_internal(mi_page_thread_free_flag(page)==MI_NEVER_DELAYED_FREE); + mi_page_set_heap(page, NULL); + +#if (MI_DEBUG>1) && !MI_TRACK_ENABLED + // check there are no references left.. + for (mi_block_t* block = (mi_block_t*)pheap->thread_delayed_free; block != NULL; block = mi_block_nextx(pheap, block, pheap->keys)) { + mi_assert_internal(_mi_ptr_page(block) != page); + } +#endif + + // and abandon it + mi_assert_internal(mi_page_heap(page) == NULL); + _mi_segment_page_abandon(page,segments_tld); +} + + +// Free a page with no more free blocks +void _mi_page_free(mi_page_t* page, mi_page_queue_t* pq, bool force) { + mi_assert_internal(page != NULL); + mi_assert_expensive(_mi_page_is_valid(page)); + mi_assert_internal(pq == mi_page_queue_of(page)); + mi_assert_internal(mi_page_all_free(page)); + mi_assert_internal(mi_page_thread_free_flag(page)!=MI_DELAYED_FREEING); + + // no more aligned blocks in here + mi_page_set_has_aligned(page, false); + + mi_heap_t* heap = mi_page_heap(page); + + // remove from the page list + // (no need to do _mi_heap_delayed_free first as all blocks are already free) + mi_segments_tld_t* segments_tld = &heap->tld->segments; + mi_page_queue_remove(pq, page); + + // and free it + mi_page_set_heap(page,NULL); + _mi_segment_page_free(page, force, segments_tld); +} + +// Retire parameters +#define MI_MAX_RETIRE_SIZE (MI_MEDIUM_OBJ_SIZE_MAX) +#define MI_RETIRE_CYCLES (16) + +// Retire a page with no more used blocks +// Important to not retire too quickly though as new +// allocations might coming. +// Note: called from `mi_free` and benchmarks often +// trigger this due to freeing everything and then +// allocating again so careful when changing this. +void _mi_page_retire(mi_page_t* page) mi_attr_noexcept { + mi_assert_internal(page != NULL); + mi_assert_expensive(_mi_page_is_valid(page)); + mi_assert_internal(mi_page_all_free(page)); + + mi_page_set_has_aligned(page, false); + + // don't retire too often.. + // (or we end up retiring and re-allocating most of the time) + // NOTE: refine this more: we should not retire if this + // is the only page left with free blocks. It is not clear + // how to check this efficiently though... + // for now, we don't retire if it is the only page left of this size class. + mi_page_queue_t* pq = mi_page_queue_of(page); + if mi_likely(page->xblock_size <= MI_MAX_RETIRE_SIZE && !mi_page_queue_is_special(pq)) { // not too large && not full or huge queue? + if (pq->last==page && pq->first==page) { // the only page in the queue? + mi_stat_counter_increase(_mi_stats_main.page_no_retire,1); + page->retire_expire = 1 + (page->xblock_size <= MI_SMALL_OBJ_SIZE_MAX ? MI_RETIRE_CYCLES : MI_RETIRE_CYCLES/4); + mi_heap_t* heap = mi_page_heap(page); + mi_assert_internal(pq >= heap->pages); + const size_t index = pq - heap->pages; + mi_assert_internal(index < MI_BIN_FULL && index < MI_BIN_HUGE); + if (index < heap->page_retired_min) heap->page_retired_min = index; + if (index > heap->page_retired_max) heap->page_retired_max = index; + mi_assert_internal(mi_page_all_free(page)); + return; // dont't free after all + } + } + _mi_page_free(page, pq, false); +} + +// free retired pages: we don't need to look at the entire queues +// since we only retire pages that are at the head position in a queue. +void _mi_heap_collect_retired(mi_heap_t* heap, bool force) { + size_t min = MI_BIN_FULL; + size_t max = 0; + for(size_t bin = heap->page_retired_min; bin <= heap->page_retired_max; bin++) { + mi_page_queue_t* pq = &heap->pages[bin]; + mi_page_t* page = pq->first; + if (page != NULL && page->retire_expire != 0) { + if (mi_page_all_free(page)) { + page->retire_expire--; + if (force || page->retire_expire == 0) { + _mi_page_free(pq->first, pq, force); + } + else { + // keep retired, update min/max + if (bin < min) min = bin; + if (bin > max) max = bin; + } + } + else { + page->retire_expire = 0; + } + } + } + heap->page_retired_min = min; + heap->page_retired_max = max; +} + + +/* ----------------------------------------------------------- + Initialize the initial free list in a page. + In secure mode we initialize a randomized list by + alternating between slices. +----------------------------------------------------------- */ + +#define MI_MAX_SLICE_SHIFT (6) // at most 64 slices +#define MI_MAX_SLICES (1UL << MI_MAX_SLICE_SHIFT) +#define MI_MIN_SLICES (2) + +static void mi_page_free_list_extend_secure(mi_heap_t* const heap, mi_page_t* const page, const size_t bsize, const size_t extend, mi_stats_t* const stats) { + MI_UNUSED(stats); + #if (MI_SECURE<=2) + mi_assert_internal(page->free == NULL); + mi_assert_internal(page->local_free == NULL); + #endif + mi_assert_internal(page->capacity + extend <= page->reserved); + mi_assert_internal(bsize == mi_page_block_size(page)); + void* const page_area = _mi_page_start(_mi_page_segment(page), page, NULL); + + // initialize a randomized free list + // set up `slice_count` slices to alternate between + size_t shift = MI_MAX_SLICE_SHIFT; + while ((extend >> shift) == 0) { + shift--; + } + const size_t slice_count = (size_t)1U << shift; + const size_t slice_extend = extend / slice_count; + mi_assert_internal(slice_extend >= 1); + mi_block_t* blocks[MI_MAX_SLICES]; // current start of the slice + size_t counts[MI_MAX_SLICES]; // available objects in the slice + for (size_t i = 0; i < slice_count; i++) { + blocks[i] = mi_page_block_at(page, page_area, bsize, page->capacity + i*slice_extend); + counts[i] = slice_extend; + } + counts[slice_count-1] += (extend % slice_count); // final slice holds the modulus too (todo: distribute evenly?) + + // and initialize the free list by randomly threading through them + // set up first element + const uintptr_t r = _mi_heap_random_next(heap); + size_t current = r % slice_count; + counts[current]--; + mi_block_t* const free_start = blocks[current]; + // and iterate through the rest; use `random_shuffle` for performance + uintptr_t rnd = _mi_random_shuffle(r|1); // ensure not 0 + for (size_t i = 1; i < extend; i++) { + // call random_shuffle only every INTPTR_SIZE rounds + const size_t round = i%MI_INTPTR_SIZE; + if (round == 0) rnd = _mi_random_shuffle(rnd); + // select a random next slice index + size_t next = ((rnd >> 8*round) & (slice_count-1)); + while (counts[next]==0) { // ensure it still has space + next++; + if (next==slice_count) next = 0; + } + // and link the current block to it + counts[next]--; + mi_block_t* const block = blocks[current]; + blocks[current] = (mi_block_t*)((uint8_t*)block + bsize); // bump to the following block + mi_block_set_next(page, block, blocks[next]); // and set next; note: we may have `current == next` + current = next; + } + // prepend to the free list (usually NULL) + mi_block_set_next(page, blocks[current], page->free); // end of the list + page->free = free_start; +} + +static mi_decl_noinline void mi_page_free_list_extend( mi_page_t* const page, const size_t bsize, const size_t extend, mi_stats_t* const stats) +{ + MI_UNUSED(stats); + #if (MI_SECURE <= 2) + mi_assert_internal(page->free == NULL); + mi_assert_internal(page->local_free == NULL); + #endif + mi_assert_internal(page->capacity + extend <= page->reserved); + mi_assert_internal(bsize == mi_page_block_size(page)); + void* const page_area = _mi_page_start(_mi_page_segment(page), page, NULL ); + + mi_block_t* const start = mi_page_block_at(page, page_area, bsize, page->capacity); + + // initialize a sequential free list + mi_block_t* const last = mi_page_block_at(page, page_area, bsize, page->capacity + extend - 1); + mi_block_t* block = start; + while(block <= last) { + mi_block_t* next = (mi_block_t*)((uint8_t*)block + bsize); + mi_block_set_next(page,block,next); + block = next; + } + // prepend to free list (usually `NULL`) + mi_block_set_next(page, last, page->free); + page->free = start; +} + +/* ----------------------------------------------------------- + Page initialize and extend the capacity +----------------------------------------------------------- */ + +#define MI_MAX_EXTEND_SIZE (4*1024) // heuristic, one OS page seems to work well. +#if (MI_SECURE>0) +#define MI_MIN_EXTEND (8*MI_SECURE) // extend at least by this many +#else +#define MI_MIN_EXTEND (4) +#endif + +// Extend the capacity (up to reserved) by initializing a free list +// We do at most `MI_MAX_EXTEND` to avoid touching too much memory +// Note: we also experimented with "bump" allocation on the first +// allocations but this did not speed up any benchmark (due to an +// extra test in malloc? or cache effects?) +static void mi_page_extend_free(mi_heap_t* heap, mi_page_t* page, mi_tld_t* tld) { + MI_UNUSED(tld); + mi_assert_expensive(mi_page_is_valid_init(page)); + #if (MI_SECURE<=2) + mi_assert(page->free == NULL); + mi_assert(page->local_free == NULL); + if (page->free != NULL) return; + #endif + if (page->capacity >= page->reserved) return; + + size_t page_size; + _mi_page_start(_mi_page_segment(page), page, &page_size); + mi_stat_counter_increase(tld->stats.pages_extended, 1); + + // calculate the extend count + const size_t bsize = (page->xblock_size < MI_HUGE_BLOCK_SIZE ? page->xblock_size : page_size); + size_t extend = page->reserved - page->capacity; + mi_assert_internal(extend > 0); + + size_t max_extend = (bsize >= MI_MAX_EXTEND_SIZE ? MI_MIN_EXTEND : MI_MAX_EXTEND_SIZE/(uint32_t)bsize); + if (max_extend < MI_MIN_EXTEND) { max_extend = MI_MIN_EXTEND; } + mi_assert_internal(max_extend > 0); + + if (extend > max_extend) { + // ensure we don't touch memory beyond the page to reduce page commit. + // the `lean` benchmark tests this. Going from 1 to 8 increases rss by 50%. + extend = max_extend; + } + + mi_assert_internal(extend > 0 && extend + page->capacity <= page->reserved); + mi_assert_internal(extend < (1UL<<16)); + + // and append the extend the free list + if (extend < MI_MIN_SLICES || MI_SECURE==0) { //!mi_option_is_enabled(mi_option_secure)) { + mi_page_free_list_extend(page, bsize, extend, &tld->stats ); + } + else { + mi_page_free_list_extend_secure(heap, page, bsize, extend, &tld->stats); + } + // enable the new free list + page->capacity += (uint16_t)extend; + mi_stat_increase(tld->stats.page_committed, extend * bsize); + mi_assert_expensive(mi_page_is_valid_init(page)); +} + +// Initialize a fresh page +static void mi_page_init(mi_heap_t* heap, mi_page_t* page, size_t block_size, mi_tld_t* tld) { + mi_assert(page != NULL); + mi_segment_t* segment = _mi_page_segment(page); + mi_assert(segment != NULL); + mi_assert_internal(block_size > 0); + // set fields + mi_page_set_heap(page, heap); + page->xblock_size = (block_size < MI_HUGE_BLOCK_SIZE ? (uint32_t)block_size : MI_HUGE_BLOCK_SIZE); // initialize before _mi_segment_page_start + size_t page_size; + const void* page_start = _mi_segment_page_start(segment, page, &page_size); + MI_UNUSED(page_start); + mi_track_mem_noaccess(page_start,page_size); + mi_assert_internal(mi_page_block_size(page) <= page_size); + mi_assert_internal(page_size <= page->slice_count*MI_SEGMENT_SLICE_SIZE); + mi_assert_internal(page_size / block_size < (1L<<16)); + page->reserved = (uint16_t)(page_size / block_size); + mi_assert_internal(page->reserved > 0); + #if (MI_PADDING || MI_ENCODE_FREELIST) + page->keys[0] = _mi_heap_random_next(heap); + page->keys[1] = _mi_heap_random_next(heap); + #endif + page->free_is_zero = page->is_zero_init; + #if MI_DEBUG>2 + if (page->is_zero_init) { + mi_track_mem_defined(page_start, page_size); + mi_assert_expensive(mi_mem_is_zero(page_start, page_size)); + } + #endif + + mi_assert_internal(page->is_committed); + mi_assert_internal(page->capacity == 0); + mi_assert_internal(page->free == NULL); + mi_assert_internal(page->used == 0); + mi_assert_internal(page->xthread_free == 0); + mi_assert_internal(page->next == NULL); + mi_assert_internal(page->prev == NULL); + mi_assert_internal(page->retire_expire == 0); + mi_assert_internal(!mi_page_has_aligned(page)); + #if (MI_PADDING || MI_ENCODE_FREELIST) + mi_assert_internal(page->keys[0] != 0); + mi_assert_internal(page->keys[1] != 0); + #endif + mi_assert_expensive(mi_page_is_valid_init(page)); + + // initialize an initial free list + mi_page_extend_free(heap,page,tld); + mi_assert(mi_page_immediate_available(page)); +} + + +/* ----------------------------------------------------------- + Find pages with free blocks +-------------------------------------------------------------*/ + +// Find a page with free blocks of `page->block_size`. +static mi_page_t* mi_page_queue_find_free_ex(mi_heap_t* heap, mi_page_queue_t* pq, bool first_try) +{ + // search through the pages in "next fit" order + #if MI_STAT + size_t count = 0; + #endif + mi_page_t* page = pq->first; + while (page != NULL) + { + mi_page_t* next = page->next; // remember next + #if MI_STAT + count++; + #endif + + // 0. collect freed blocks by us and other threads + _mi_page_free_collect(page, false); + + // 1. if the page contains free blocks, we are done + if (mi_page_immediate_available(page)) { + break; // pick this one + } + + // 2. Try to extend + if (page->capacity < page->reserved) { + mi_page_extend_free(heap, page, heap->tld); + mi_assert_internal(mi_page_immediate_available(page)); + break; + } + + // 3. If the page is completely full, move it to the `mi_pages_full` + // queue so we don't visit long-lived pages too often. + mi_assert_internal(!mi_page_is_in_full(page) && !mi_page_immediate_available(page)); + mi_page_to_full(page, pq); + + page = next; + } // for each page + + mi_heap_stat_counter_increase(heap, searches, count); + + if (page == NULL) { + _mi_heap_collect_retired(heap, false); // perhaps make a page available? + page = mi_page_fresh(heap, pq); + if (page == NULL && first_try) { + // out-of-memory _or_ an abandoned page with free blocks was reclaimed, try once again + page = mi_page_queue_find_free_ex(heap, pq, false); + } + } + else { + mi_assert(pq->first == page); + page->retire_expire = 0; + } + mi_assert_internal(page == NULL || mi_page_immediate_available(page)); + return page; +} + + + +// Find a page with free blocks of `size`. +static inline mi_page_t* mi_find_free_page(mi_heap_t* heap, size_t size) { + mi_page_queue_t* pq = mi_page_queue(heap,size); + mi_page_t* page = pq->first; + if (page != NULL) { + #if (MI_SECURE>=3) // in secure mode, we extend half the time to increase randomness + if (page->capacity < page->reserved && ((_mi_heap_random_next(heap) & 1) == 1)) { + mi_page_extend_free(heap, page, heap->tld); + mi_assert_internal(mi_page_immediate_available(page)); + } + else + #endif + { + _mi_page_free_collect(page,false); + } + + if (mi_page_immediate_available(page)) { + page->retire_expire = 0; + return page; // fast path + } + } + return mi_page_queue_find_free_ex(heap, pq, true); +} + + +/* ----------------------------------------------------------- + Users can register a deferred free function called + when the `free` list is empty. Since the `local_free` + is separate this is deterministically called after + a certain number of allocations. +----------------------------------------------------------- */ + +static mi_deferred_free_fun* volatile deferred_free = NULL; +static _Atomic(void*) deferred_arg; // = NULL + +void _mi_deferred_free(mi_heap_t* heap, bool force) { + heap->tld->heartbeat++; + if (deferred_free != NULL && !heap->tld->recurse) { + heap->tld->recurse = true; + deferred_free(force, heap->tld->heartbeat, mi_atomic_load_ptr_relaxed(void,&deferred_arg)); + heap->tld->recurse = false; + } +} + +void mi_register_deferred_free(mi_deferred_free_fun* fn, void* arg) mi_attr_noexcept { + deferred_free = fn; + mi_atomic_store_ptr_release(void,&deferred_arg, arg); +} + + +/* ----------------------------------------------------------- + General allocation +----------------------------------------------------------- */ + +// Large and huge page allocation. +// Huge pages are allocated directly without being in a queue. +// Because huge pages contain just one block, and the segment contains +// just that page, we always treat them as abandoned and any thread +// that frees the block can free the whole page and segment directly. +// Huge pages are also use if the requested alignment is very large (> MI_ALIGNMENT_MAX). +static mi_page_t* mi_large_huge_page_alloc(mi_heap_t* heap, size_t size, size_t page_alignment) { + size_t block_size = _mi_os_good_alloc_size(size); + mi_assert_internal(mi_bin(block_size) == MI_BIN_HUGE || page_alignment > 0); + bool is_huge = (block_size > MI_LARGE_OBJ_SIZE_MAX || page_alignment > 0); + #if MI_HUGE_PAGE_ABANDON + mi_page_queue_t* pq = (is_huge ? NULL : mi_page_queue(heap, block_size)); + #else + mi_page_queue_t* pq = mi_page_queue(heap, is_huge ? MI_HUGE_BLOCK_SIZE : block_size); // not block_size as that can be low if the page_alignment > 0 + mi_assert_internal(!is_huge || mi_page_queue_is_huge(pq)); + #endif + mi_page_t* page = mi_page_fresh_alloc(heap, pq, block_size, page_alignment); + if (page != NULL) { + mi_assert_internal(mi_page_immediate_available(page)); + + if (is_huge) { + mi_assert_internal(_mi_page_segment(page)->kind == MI_SEGMENT_HUGE); + mi_assert_internal(_mi_page_segment(page)->used==1); + #if MI_HUGE_PAGE_ABANDON + mi_assert_internal(_mi_page_segment(page)->thread_id==0); // abandoned, not in the huge queue + mi_page_set_heap(page, NULL); + #endif + } + else { + mi_assert_internal(_mi_page_segment(page)->kind != MI_SEGMENT_HUGE); + } + + const size_t bsize = mi_page_usable_block_size(page); // note: not `mi_page_block_size` to account for padding + if (bsize <= MI_LARGE_OBJ_SIZE_MAX) { + mi_heap_stat_increase(heap, large, bsize); + mi_heap_stat_counter_increase(heap, large_count, 1); + } + else { + mi_heap_stat_increase(heap, huge, bsize); + mi_heap_stat_counter_increase(heap, huge_count, 1); + } + } + return page; +} + + +// Allocate a page +// Note: in debug mode the size includes MI_PADDING_SIZE and might have overflowed. +static mi_page_t* mi_find_page(mi_heap_t* heap, size_t size, size_t huge_alignment) mi_attr_noexcept { + // huge allocation? + const size_t req_size = size - MI_PADDING_SIZE; // correct for padding_size in case of an overflow on `size` + if mi_unlikely(req_size > (MI_MEDIUM_OBJ_SIZE_MAX - MI_PADDING_SIZE) || huge_alignment > 0) { + if mi_unlikely(req_size > PTRDIFF_MAX) { // we don't allocate more than PTRDIFF_MAX (see ) + _mi_error_message(EOVERFLOW, "allocation request is too large (%zu bytes)\n", req_size); + return NULL; + } + else { + return mi_large_huge_page_alloc(heap,size,huge_alignment); + } + } + else { + // otherwise find a page with free blocks in our size segregated queues + #if MI_PADDING + mi_assert_internal(size >= MI_PADDING_SIZE); + #endif + return mi_find_free_page(heap, size); + } +} + +// Generic allocation routine if the fast path (`alloc.c:mi_page_malloc`) does not succeed. +// Note: in debug mode the size includes MI_PADDING_SIZE and might have overflowed. +// The `huge_alignment` is normally 0 but is set to a multiple of MI_SEGMENT_SIZE for +// very large requested alignments in which case we use a huge segment. +void* _mi_malloc_generic(mi_heap_t* heap, size_t size, bool zero, size_t huge_alignment) mi_attr_noexcept +{ + mi_assert_internal(heap != NULL); + + // initialize if necessary + if mi_unlikely(!mi_heap_is_initialized(heap)) { + heap = mi_heap_get_default(); // calls mi_thread_init + if mi_unlikely(!mi_heap_is_initialized(heap)) { return NULL; } + } + mi_assert_internal(mi_heap_is_initialized(heap)); + + // call potential deferred free routines + _mi_deferred_free(heap, false); + + // free delayed frees from other threads (but skip contended ones) + _mi_heap_delayed_free_partial(heap); + + // find (or allocate) a page of the right size + mi_page_t* page = mi_find_page(heap, size, huge_alignment); + if mi_unlikely(page == NULL) { // first time out of memory, try to collect and retry the allocation once more + mi_heap_collect(heap, true /* force */); + page = mi_find_page(heap, size, huge_alignment); + } + + if mi_unlikely(page == NULL) { // out of memory + const size_t req_size = size - MI_PADDING_SIZE; // correct for padding_size in case of an overflow on `size` + _mi_error_message(ENOMEM, "unable to allocate memory (%zu bytes)\n", req_size); + return NULL; + } + + mi_assert_internal(mi_page_immediate_available(page)); + mi_assert_internal(mi_page_block_size(page) >= size); + + // and try again, this time succeeding! (i.e. this should never recurse through _mi_page_malloc) + if mi_unlikely(zero && page->xblock_size == 0) { + // note: we cannot call _mi_page_malloc with zeroing for huge blocks; we zero it afterwards in that case. + void* p = _mi_page_malloc(heap, page, size, false); + mi_assert_internal(p != NULL); + _mi_memzero_aligned(p, mi_page_usable_block_size(page)); + return p; + } + else { + return _mi_page_malloc(heap, page, size, zero); + } +} diff --git a/compat/mimalloc/prim/windows/prim.c b/compat/mimalloc/prim/windows/prim.c new file mode 100644 index 00000000000000..d060833c5b644d --- /dev/null +++ b/compat/mimalloc/prim/windows/prim.c @@ -0,0 +1,622 @@ +/* ---------------------------------------------------------------------------- +Copyright (c) 2018-2023, Microsoft Research, Daan Leijen +This is free software; you can redistribute it and/or modify it under the +terms of the MIT license. A copy of the license can be found in the file +"LICENSE" at the root of this distribution. +-----------------------------------------------------------------------------*/ + +// This file is included in `src/prim/prim.c` + +#include "mimalloc.h" +#include "mimalloc/internal.h" +#include "mimalloc/atomic.h" +#include "mimalloc/prim.h" +#include // fputs, stderr + + +//--------------------------------------------- +// Dynamically bind Windows API points for portability +//--------------------------------------------- + +// We use VirtualAlloc2 for aligned allocation, but it is only supported on Windows 10 and Windows Server 2016. +// So, we need to look it up dynamically to run on older systems. (use __stdcall for 32-bit compatibility) +// NtAllocateVirtualAllocEx is used for huge OS page allocation (1GiB) +// We define a minimal MEM_EXTENDED_PARAMETER ourselves in order to be able to compile with older SDK's. +typedef enum MI_MEM_EXTENDED_PARAMETER_TYPE_E { + MiMemExtendedParameterInvalidType = 0, + MiMemExtendedParameterAddressRequirements, + MiMemExtendedParameterNumaNode, + MiMemExtendedParameterPartitionHandle, + MiMemExtendedParameterUserPhysicalHandle, + MiMemExtendedParameterAttributeFlags, + MiMemExtendedParameterMax +} MI_MEM_EXTENDED_PARAMETER_TYPE; + +typedef struct DECLSPEC_ALIGN(8) MI_MEM_EXTENDED_PARAMETER_S { + struct { DWORD64 Type : 8; DWORD64 Reserved : 56; } Type; + union { DWORD64 ULong64; PVOID Pointer; SIZE_T Size; HANDLE Handle; DWORD ULong; } Arg; +} MI_MEM_EXTENDED_PARAMETER; + +typedef struct MI_MEM_ADDRESS_REQUIREMENTS_S { + PVOID LowestStartingAddress; + PVOID HighestEndingAddress; + SIZE_T Alignment; +} MI_MEM_ADDRESS_REQUIREMENTS; + +#define MI_MEM_EXTENDED_PARAMETER_NONPAGED_HUGE 0x00000010 + +#include +typedef PVOID (__stdcall *PVirtualAlloc2)(HANDLE, PVOID, SIZE_T, ULONG, ULONG, MI_MEM_EXTENDED_PARAMETER*, ULONG); +typedef NTSTATUS (__stdcall *PNtAllocateVirtualMemoryEx)(HANDLE, PVOID*, SIZE_T*, ULONG, ULONG, MI_MEM_EXTENDED_PARAMETER*, ULONG); +static PVirtualAlloc2 pVirtualAlloc2 = NULL; +static PNtAllocateVirtualMemoryEx pNtAllocateVirtualMemoryEx = NULL; + +// Similarly, GetNumaProcesorNodeEx is only supported since Windows 7 +typedef struct MI_PROCESSOR_NUMBER_S { WORD Group; BYTE Number; BYTE Reserved; } MI_PROCESSOR_NUMBER; + +typedef VOID (__stdcall *PGetCurrentProcessorNumberEx)(MI_PROCESSOR_NUMBER* ProcNumber); +typedef BOOL (__stdcall *PGetNumaProcessorNodeEx)(MI_PROCESSOR_NUMBER* Processor, PUSHORT NodeNumber); +typedef BOOL (__stdcall* PGetNumaNodeProcessorMaskEx)(USHORT Node, PGROUP_AFFINITY ProcessorMask); +typedef BOOL (__stdcall *PGetNumaProcessorNode)(UCHAR Processor, PUCHAR NodeNumber); +static PGetCurrentProcessorNumberEx pGetCurrentProcessorNumberEx = NULL; +static PGetNumaProcessorNodeEx pGetNumaProcessorNodeEx = NULL; +static PGetNumaNodeProcessorMaskEx pGetNumaNodeProcessorMaskEx = NULL; +static PGetNumaProcessorNode pGetNumaProcessorNode = NULL; + +//--------------------------------------------- +// Enable large page support dynamically (if possible) +//--------------------------------------------- + +static bool win_enable_large_os_pages(size_t* large_page_size) +{ + static bool large_initialized = false; + if (large_initialized) return (_mi_os_large_page_size() > 0); + large_initialized = true; + + // Try to see if large OS pages are supported + // To use large pages on Windows, we first need access permission + // Set "Lock pages in memory" permission in the group policy editor + // + unsigned long err = 0; + HANDLE token = NULL; + BOOL ok = OpenProcessToken(GetCurrentProcess(), TOKEN_ADJUST_PRIVILEGES | TOKEN_QUERY, &token); + if (ok) { + TOKEN_PRIVILEGES tp; + ok = LookupPrivilegeValue(NULL, TEXT("SeLockMemoryPrivilege"), &tp.Privileges[0].Luid); + if (ok) { + tp.PrivilegeCount = 1; + tp.Privileges[0].Attributes = SE_PRIVILEGE_ENABLED; + ok = AdjustTokenPrivileges(token, FALSE, &tp, 0, (PTOKEN_PRIVILEGES)NULL, 0); + if (ok) { + err = GetLastError(); + ok = (err == ERROR_SUCCESS); + if (ok && large_page_size != NULL) { + *large_page_size = GetLargePageMinimum(); + } + } + } + CloseHandle(token); + } + if (!ok) { + if (err == 0) err = GetLastError(); + _mi_warning_message("cannot enable large OS page support, error %lu\n", err); + } + return (ok!=0); +} + + +//--------------------------------------------- +// Initialize +//--------------------------------------------- + +void _mi_prim_mem_init( mi_os_mem_config_t* config ) +{ + config->has_overcommit = false; + config->must_free_whole = true; + config->has_virtual_reserve = true; + // get the page size + SYSTEM_INFO si; + GetSystemInfo(&si); + if (si.dwPageSize > 0) { config->page_size = si.dwPageSize; } + if (si.dwAllocationGranularity > 0) { config->alloc_granularity = si.dwAllocationGranularity; } + // get the VirtualAlloc2 function + HINSTANCE hDll; + hDll = LoadLibrary(TEXT("kernelbase.dll")); + if (hDll != NULL) { + // use VirtualAlloc2FromApp if possible as it is available to Windows store apps + pVirtualAlloc2 = (PVirtualAlloc2)(void (*)(void))GetProcAddress(hDll, "VirtualAlloc2FromApp"); + if (pVirtualAlloc2==NULL) pVirtualAlloc2 = (PVirtualAlloc2)(void (*)(void))GetProcAddress(hDll, "VirtualAlloc2"); + FreeLibrary(hDll); + } + // NtAllocateVirtualMemoryEx is used for huge page allocation + hDll = LoadLibrary(TEXT("ntdll.dll")); + if (hDll != NULL) { + pNtAllocateVirtualMemoryEx = (PNtAllocateVirtualMemoryEx)(void (*)(void))GetProcAddress(hDll, "NtAllocateVirtualMemoryEx"); + FreeLibrary(hDll); + } + // Try to use Win7+ numa API + hDll = LoadLibrary(TEXT("kernel32.dll")); + if (hDll != NULL) { + pGetCurrentProcessorNumberEx = (PGetCurrentProcessorNumberEx)(void (*)(void))GetProcAddress(hDll, "GetCurrentProcessorNumberEx"); + pGetNumaProcessorNodeEx = (PGetNumaProcessorNodeEx)(void (*)(void))GetProcAddress(hDll, "GetNumaProcessorNodeEx"); + pGetNumaNodeProcessorMaskEx = (PGetNumaNodeProcessorMaskEx)(void (*)(void))GetProcAddress(hDll, "GetNumaNodeProcessorMaskEx"); + pGetNumaProcessorNode = (PGetNumaProcessorNode)(void (*)(void))GetProcAddress(hDll, "GetNumaProcessorNode"); + FreeLibrary(hDll); + } + if (mi_option_is_enabled(mi_option_allow_large_os_pages) || mi_option_is_enabled(mi_option_reserve_huge_os_pages)) { + win_enable_large_os_pages(&config->large_page_size); + } +} + + +//--------------------------------------------- +// Free +//--------------------------------------------- + +int _mi_prim_free(void* addr, size_t size ) { + MI_UNUSED(size); + DWORD errcode = 0; + bool err = (VirtualFree(addr, 0, MEM_RELEASE) == 0); + if (err) { errcode = GetLastError(); } + if (errcode == ERROR_INVALID_ADDRESS) { + // In mi_os_mem_alloc_aligned the fallback path may have returned a pointer inside + // the memory region returned by VirtualAlloc; in that case we need to free using + // the start of the region. + MEMORY_BASIC_INFORMATION info = { 0 }; + VirtualQuery(addr, &info, sizeof(info)); + if (info.AllocationBase < addr && ((uint8_t*)addr - (uint8_t*)info.AllocationBase) < (ptrdiff_t)MI_SEGMENT_SIZE) { + errcode = 0; + err = (VirtualFree(info.AllocationBase, 0, MEM_RELEASE) == 0); + if (err) { errcode = GetLastError(); } + } + } + return (int)errcode; +} + + +//--------------------------------------------- +// VirtualAlloc +//--------------------------------------------- + +static void* win_virtual_alloc_prim(void* addr, size_t size, size_t try_alignment, DWORD flags) { + #if (MI_INTPTR_SIZE >= 8) + // on 64-bit systems, try to use the virtual address area after 2TiB for 4MiB aligned allocations + if (addr == NULL) { + void* hint = _mi_os_get_aligned_hint(try_alignment,size); + if (hint != NULL) { + void* p = VirtualAlloc(hint, size, flags, PAGE_READWRITE); + if (p != NULL) return p; + _mi_verbose_message("warning: unable to allocate hinted aligned OS memory (%zu bytes, error code: 0x%x, address: %p, alignment: %zu, flags: 0x%x)\n", size, GetLastError(), hint, try_alignment, flags); + // fall through on error + } + } + #endif + // on modern Windows try use VirtualAlloc2 for aligned allocation + if (try_alignment > 1 && (try_alignment % _mi_os_page_size()) == 0 && pVirtualAlloc2 != NULL) { + MI_MEM_ADDRESS_REQUIREMENTS reqs = { 0, 0, 0 }; + reqs.Alignment = try_alignment; + MI_MEM_EXTENDED_PARAMETER param = { {0, 0}, {0} }; + param.Type.Type = MiMemExtendedParameterAddressRequirements; + param.Arg.Pointer = &reqs; + void* p = (*pVirtualAlloc2)(GetCurrentProcess(), addr, size, flags, PAGE_READWRITE, ¶m, 1); + if (p != NULL) return p; + _mi_warning_message("unable to allocate aligned OS memory (%zu bytes, error code: 0x%x, address: %p, alignment: %zu, flags: 0x%x)\n", size, GetLastError(), addr, try_alignment, flags); + // fall through on error + } + // last resort + return VirtualAlloc(addr, size, flags, PAGE_READWRITE); +} + +static void* win_virtual_alloc(void* addr, size_t size, size_t try_alignment, DWORD flags, bool large_only, bool allow_large, bool* is_large) { + mi_assert_internal(!(large_only && !allow_large)); + static _Atomic(size_t) large_page_try_ok; // = 0; + void* p = NULL; + // Try to allocate large OS pages (2MiB) if allowed or required. + if ((large_only || _mi_os_use_large_page(size, try_alignment)) + && allow_large && (flags&MEM_COMMIT)!=0 && (flags&MEM_RESERVE)!=0) { + size_t try_ok = mi_atomic_load_acquire(&large_page_try_ok); + if (!large_only && try_ok > 0) { + // if a large page allocation fails, it seems the calls to VirtualAlloc get very expensive. + // therefore, once a large page allocation failed, we don't try again for `large_page_try_ok` times. + mi_atomic_cas_strong_acq_rel(&large_page_try_ok, &try_ok, try_ok - 1); + } + else { + // large OS pages must always reserve and commit. + *is_large = true; + p = win_virtual_alloc_prim(addr, size, try_alignment, flags | MEM_LARGE_PAGES); + if (large_only) return p; + // fall back to non-large page allocation on error (`p == NULL`). + if (p == NULL) { + mi_atomic_store_release(&large_page_try_ok,10UL); // on error, don't try again for the next N allocations + } + } + } + // Fall back to regular page allocation + if (p == NULL) { + *is_large = ((flags&MEM_LARGE_PAGES) != 0); + p = win_virtual_alloc_prim(addr, size, try_alignment, flags); + } + //if (p == NULL) { _mi_warning_message("unable to allocate OS memory (%zu bytes, error code: 0x%x, address: %p, alignment: %zu, flags: 0x%x, large only: %d, allow large: %d)\n", size, GetLastError(), addr, try_alignment, flags, large_only, allow_large); } + return p; +} + +int _mi_prim_alloc(size_t size, size_t try_alignment, bool commit, bool allow_large, bool* is_large, bool* is_zero, void** addr) { + mi_assert_internal(size > 0 && (size % _mi_os_page_size()) == 0); + mi_assert_internal(commit || !allow_large); + mi_assert_internal(try_alignment > 0); + *is_zero = true; + int flags = MEM_RESERVE; + if (commit) { flags |= MEM_COMMIT; } + *addr = win_virtual_alloc(NULL, size, try_alignment, flags, false, allow_large, is_large); + return (*addr != NULL ? 0 : (int)GetLastError()); +} + + +//--------------------------------------------- +// Commit/Reset/Protect +//--------------------------------------------- +#ifdef _MSC_VER +#pragma warning(disable:6250) // suppress warning calling VirtualFree without MEM_RELEASE (for decommit) +#endif + +int _mi_prim_commit(void* addr, size_t size, bool* is_zero) { + *is_zero = false; + /* + // zero'ing only happens on an initial commit... but checking upfront seems expensive.. + _MEMORY_BASIC_INFORMATION meminfo; _mi_memzero_var(meminfo); + if (VirtualQuery(addr, &meminfo, size) > 0) { + if ((meminfo.State & MEM_COMMIT) == 0) { + *is_zero = true; + } + } + */ + // commit + void* p = VirtualAlloc(addr, size, MEM_COMMIT, PAGE_READWRITE); + if (p == NULL) return (int)GetLastError(); + return 0; +} + +int _mi_prim_decommit(void* addr, size_t size, bool* needs_recommit) { + BOOL ok = VirtualFree(addr, size, MEM_DECOMMIT); + *needs_recommit = true; // for safety, assume always decommitted even in the case of an error. + return (ok ? 0 : (int)GetLastError()); +} + +int _mi_prim_reset(void* addr, size_t size) { + void* p = VirtualAlloc(addr, size, MEM_RESET, PAGE_READWRITE); + mi_assert_internal(p == addr); + #if 0 + if (p != NULL) { + VirtualUnlock(addr,size); // VirtualUnlock after MEM_RESET removes the memory directly from the working set + } + #endif + return (p != NULL ? 0 : (int)GetLastError()); +} + +int _mi_prim_protect(void* addr, size_t size, bool protect) { + DWORD oldprotect = 0; + BOOL ok = VirtualProtect(addr, size, protect ? PAGE_NOACCESS : PAGE_READWRITE, &oldprotect); + return (ok ? 0 : (int)GetLastError()); +} + + +//--------------------------------------------- +// Huge page allocation +//--------------------------------------------- + +static void* _mi_prim_alloc_huge_os_pagesx(void* hint_addr, size_t size, int numa_node) +{ + const DWORD flags = MEM_LARGE_PAGES | MEM_COMMIT | MEM_RESERVE; + + win_enable_large_os_pages(NULL); + + MI_MEM_EXTENDED_PARAMETER params[3] = { {{0,0},{0}},{{0,0},{0}},{{0,0},{0}} }; + // on modern Windows try use NtAllocateVirtualMemoryEx for 1GiB huge pages + static bool mi_huge_pages_available = true; + if (pNtAllocateVirtualMemoryEx != NULL && mi_huge_pages_available) { + params[0].Type.Type = MiMemExtendedParameterAttributeFlags; + params[0].Arg.ULong64 = MI_MEM_EXTENDED_PARAMETER_NONPAGED_HUGE; + ULONG param_count = 1; + if (numa_node >= 0) { + param_count++; + params[1].Type.Type = MiMemExtendedParameterNumaNode; + params[1].Arg.ULong = (unsigned)numa_node; + } + SIZE_T psize = size; + void* base = hint_addr; + NTSTATUS err = (*pNtAllocateVirtualMemoryEx)(GetCurrentProcess(), &base, &psize, flags, PAGE_READWRITE, params, param_count); + if (err == 0 && base != NULL) { + return base; + } + else { + // fall back to regular large pages + mi_huge_pages_available = false; // don't try further huge pages + _mi_warning_message("unable to allocate using huge (1GiB) pages, trying large (2MiB) pages instead (status 0x%lx)\n", err); + } + } + // on modern Windows try use VirtualAlloc2 for numa aware large OS page allocation + if (pVirtualAlloc2 != NULL && numa_node >= 0) { + params[0].Type.Type = MiMemExtendedParameterNumaNode; + params[0].Arg.ULong = (unsigned)numa_node; + return (*pVirtualAlloc2)(GetCurrentProcess(), hint_addr, size, flags, PAGE_READWRITE, params, 1); + } + + // otherwise use regular virtual alloc on older windows + return VirtualAlloc(hint_addr, size, flags, PAGE_READWRITE); +} + +int _mi_prim_alloc_huge_os_pages(void* hint_addr, size_t size, int numa_node, bool* is_zero, void** addr) { + *is_zero = true; + *addr = _mi_prim_alloc_huge_os_pagesx(hint_addr,size,numa_node); + return (*addr != NULL ? 0 : (int)GetLastError()); +} + + +//--------------------------------------------- +// Numa nodes +//--------------------------------------------- + +size_t _mi_prim_numa_node(void) { + USHORT numa_node = 0; + if (pGetCurrentProcessorNumberEx != NULL && pGetNumaProcessorNodeEx != NULL) { + // Extended API is supported + MI_PROCESSOR_NUMBER pnum; + (*pGetCurrentProcessorNumberEx)(&pnum); + USHORT nnode = 0; + BOOL ok = (*pGetNumaProcessorNodeEx)(&pnum, &nnode); + if (ok) { numa_node = nnode; } + } + else if (pGetNumaProcessorNode != NULL) { + // Vista or earlier, use older API that is limited to 64 processors. Issue #277 + DWORD pnum = GetCurrentProcessorNumber(); + UCHAR nnode = 0; + BOOL ok = pGetNumaProcessorNode((UCHAR)pnum, &nnode); + if (ok) { numa_node = nnode; } + } + return numa_node; +} + +size_t _mi_prim_numa_node_count(void) { + ULONG numa_max = 0; + GetNumaHighestNodeNumber(&numa_max); + // find the highest node number that has actual processors assigned to it. Issue #282 + while(numa_max > 0) { + if (pGetNumaNodeProcessorMaskEx != NULL) { + // Extended API is supported + GROUP_AFFINITY affinity; + if ((*pGetNumaNodeProcessorMaskEx)((USHORT)numa_max, &affinity)) { + if (affinity.Mask != 0) break; // found the maximum non-empty node + } + } + else { + // Vista or earlier, use older API that is limited to 64 processors. + ULONGLONG mask; + if (GetNumaNodeProcessorMask((UCHAR)numa_max, &mask)) { + if (mask != 0) break; // found the maximum non-empty node + }; + } + // max node was invalid or had no processor assigned, try again + numa_max--; + } + return ((size_t)numa_max + 1); +} + + +//---------------------------------------------------------------- +// Clock +//---------------------------------------------------------------- + +static mi_msecs_t mi_to_msecs(LARGE_INTEGER t) { + static LARGE_INTEGER mfreq; // = 0 + if (mfreq.QuadPart == 0LL) { + LARGE_INTEGER f; + QueryPerformanceFrequency(&f); + mfreq.QuadPart = f.QuadPart/1000LL; + if (mfreq.QuadPart == 0) mfreq.QuadPart = 1; + } + return (mi_msecs_t)(t.QuadPart / mfreq.QuadPart); +} + +mi_msecs_t _mi_prim_clock_now(void) { + LARGE_INTEGER t; + QueryPerformanceCounter(&t); + return mi_to_msecs(t); +} + + +//---------------------------------------------------------------- +// Process Info +//---------------------------------------------------------------- + +#include +#include + +static mi_msecs_t filetime_msecs(const FILETIME* ftime) { + ULARGE_INTEGER i; + i.LowPart = ftime->dwLowDateTime; + i.HighPart = ftime->dwHighDateTime; + mi_msecs_t msecs = (i.QuadPart / 10000); // FILETIME is in 100 nano seconds + return msecs; +} + +typedef BOOL (WINAPI *PGetProcessMemoryInfo)(HANDLE, PPROCESS_MEMORY_COUNTERS, DWORD); +static PGetProcessMemoryInfo pGetProcessMemoryInfo = NULL; + +void _mi_prim_process_info(mi_process_info_t* pinfo) +{ + FILETIME ct; + FILETIME ut; + FILETIME st; + FILETIME et; + GetProcessTimes(GetCurrentProcess(), &ct, &et, &st, &ut); + pinfo->utime = filetime_msecs(&ut); + pinfo->stime = filetime_msecs(&st); + + // load psapi on demand + if (pGetProcessMemoryInfo == NULL) { + HINSTANCE hDll = LoadLibrary(TEXT("psapi.dll")); + if (hDll != NULL) { + pGetProcessMemoryInfo = (PGetProcessMemoryInfo)(void (*)(void))GetProcAddress(hDll, "GetProcessMemoryInfo"); + } + } + + // get process info + PROCESS_MEMORY_COUNTERS info; + memset(&info, 0, sizeof(info)); + if (pGetProcessMemoryInfo != NULL) { + pGetProcessMemoryInfo(GetCurrentProcess(), &info, sizeof(info)); + } + pinfo->current_rss = (size_t)info.WorkingSetSize; + pinfo->peak_rss = (size_t)info.PeakWorkingSetSize; + pinfo->current_commit = (size_t)info.PagefileUsage; + pinfo->peak_commit = (size_t)info.PeakPagefileUsage; + pinfo->page_faults = (size_t)info.PageFaultCount; +} + +//---------------------------------------------------------------- +// Output +//---------------------------------------------------------------- + +void _mi_prim_out_stderr( const char* msg ) +{ + // on windows with redirection, the C runtime cannot handle locale dependent output + // after the main thread closes so we use direct console output. + if (!_mi_preloading()) { + // _cputs(msg); // _cputs cannot be used at is aborts if it fails to lock the console + static HANDLE hcon = INVALID_HANDLE_VALUE; + static bool hconIsConsole; + if (hcon == INVALID_HANDLE_VALUE) { + CONSOLE_SCREEN_BUFFER_INFO sbi; + hcon = GetStdHandle(STD_ERROR_HANDLE); + hconIsConsole = ((hcon != INVALID_HANDLE_VALUE) && GetConsoleScreenBufferInfo(hcon, &sbi)); + } + const size_t len = _mi_strlen(msg); + if (len > 0 && len < UINT32_MAX) { + DWORD written = 0; + if (hconIsConsole) { + WriteConsoleA(hcon, msg, (DWORD)len, &written, NULL); + } + else if (hcon != INVALID_HANDLE_VALUE) { + // use direct write if stderr was redirected + WriteFile(hcon, msg, (DWORD)len, &written, NULL); + } + else { + // finally fall back to fputs after all + fputs(msg, stderr); + } + } + } +} + + +//---------------------------------------------------------------- +// Environment +//---------------------------------------------------------------- + +// On Windows use GetEnvironmentVariable instead of getenv to work +// reliably even when this is invoked before the C runtime is initialized. +// i.e. when `_mi_preloading() == true`. +// Note: on windows, environment names are not case sensitive. +bool _mi_prim_getenv(const char* name, char* result, size_t result_size) { + result[0] = 0; + size_t len = GetEnvironmentVariableA(name, result, (DWORD)result_size); + return (len > 0 && len < result_size); +} + + + +//---------------------------------------------------------------- +// Random +//---------------------------------------------------------------- + +#if defined(MI_USE_RTLGENRANDOM) // || defined(__cplusplus) +// We prefer to use BCryptGenRandom instead of (the unofficial) RtlGenRandom but when using +// dynamic overriding, we observed it can raise an exception when compiled with C++, and +// sometimes deadlocks when also running under the VS debugger. +// In contrast, issue #623 implies that on Windows Server 2019 we need to use BCryptGenRandom. +// To be continued.. +#pragma comment (lib,"advapi32.lib") +#define RtlGenRandom SystemFunction036 +mi_decl_externc BOOLEAN NTAPI RtlGenRandom(PVOID RandomBuffer, ULONG RandomBufferLength); + +bool _mi_prim_random_buf(void* buf, size_t buf_len) { + return (RtlGenRandom(buf, (ULONG)buf_len) != 0); +} + +#else + +#ifndef BCRYPT_USE_SYSTEM_PREFERRED_RNG +#define BCRYPT_USE_SYSTEM_PREFERRED_RNG 0x00000002 +#endif + +typedef LONG (NTAPI *PBCryptGenRandom)(HANDLE, PUCHAR, ULONG, ULONG); +static PBCryptGenRandom pBCryptGenRandom = NULL; + +bool _mi_prim_random_buf(void* buf, size_t buf_len) { + if (pBCryptGenRandom == NULL) { + HINSTANCE hDll = LoadLibrary(TEXT("bcrypt.dll")); + if (hDll != NULL) { + pBCryptGenRandom = (PBCryptGenRandom)(void (*)(void))GetProcAddress(hDll, "BCryptGenRandom"); + } + if (pBCryptGenRandom == NULL) return false; + } + return (pBCryptGenRandom(NULL, (PUCHAR)buf, (ULONG)buf_len, BCRYPT_USE_SYSTEM_PREFERRED_RNG) >= 0); +} + +#endif // MI_USE_RTLGENRANDOM + +//---------------------------------------------------------------- +// Thread init/done +//---------------------------------------------------------------- + +#if !defined(MI_SHARED_LIB) + +// use thread local storage keys to detect thread ending +#include +#if (_WIN32_WINNT < 0x600) // before Windows Vista +WINBASEAPI DWORD WINAPI FlsAlloc( _In_opt_ PFLS_CALLBACK_FUNCTION lpCallback ); +WINBASEAPI PVOID WINAPI FlsGetValue( _In_ DWORD dwFlsIndex ); +WINBASEAPI BOOL WINAPI FlsSetValue( _In_ DWORD dwFlsIndex, _In_opt_ PVOID lpFlsData ); +WINBASEAPI BOOL WINAPI FlsFree(_In_ DWORD dwFlsIndex); +#endif + +static DWORD mi_fls_key = (DWORD)(-1); + +static void NTAPI mi_fls_done(PVOID value) { + mi_heap_t* heap = (mi_heap_t*)value; + if (heap != NULL) { + _mi_thread_done(heap); + FlsSetValue(mi_fls_key, NULL); // prevent recursion as _mi_thread_done may set it back to the main heap, issue #672 + } +} + +void _mi_prim_thread_init_auto_done(void) { + mi_fls_key = FlsAlloc(&mi_fls_done); +} + +void _mi_prim_thread_done_auto_done(void) { + // call thread-done on all threads (except the main thread) to prevent + // dangling callback pointer if statically linked with a DLL; Issue #208 + FlsFree(mi_fls_key); +} + +void _mi_prim_thread_associate_default_heap(mi_heap_t* heap) { + mi_assert_internal(mi_fls_key != (DWORD)(-1)); + FlsSetValue(mi_fls_key, heap); +} + +#else + +// Dll; nothing to do as in that case thread_done is handled through the DLL_THREAD_DETACH event. + +void _mi_prim_thread_init_auto_done(void) { +} + +void _mi_prim_thread_done_auto_done(void) { +} + +void _mi_prim_thread_associate_default_heap(mi_heap_t* heap) { + MI_UNUSED(heap); +} + +#endif diff --git a/compat/mimalloc/random.c b/compat/mimalloc/random.c new file mode 100644 index 00000000000000..2a18b5aa992dad --- /dev/null +++ b/compat/mimalloc/random.c @@ -0,0 +1,254 @@ +/* ---------------------------------------------------------------------------- +Copyright (c) 2019-2021, Microsoft Research, Daan Leijen +This is free software; you can redistribute it and/or modify it under the +terms of the MIT license. A copy of the license can be found in the file +"LICENSE" at the root of this distribution. +-----------------------------------------------------------------------------*/ +#include "mimalloc.h" +#include "mimalloc/internal.h" +#include "mimalloc/prim.h" // _mi_prim_random_buf +#include // memset + +/* ---------------------------------------------------------------------------- +We use our own PRNG to keep predictable performance of random number generation +and to avoid implementations that use a lock. We only use the OS provided +random source to initialize the initial seeds. Since we do not need ultimate +performance but we do rely on the security (for secret cookies in secure mode) +we use a cryptographically secure generator (chacha20). +-----------------------------------------------------------------------------*/ + +#define MI_CHACHA_ROUNDS (20) // perhaps use 12 for better performance? + + +/* ---------------------------------------------------------------------------- +Chacha20 implementation as the original algorithm with a 64-bit nonce +and counter: https://en.wikipedia.org/wiki/Salsa20 +The input matrix has sixteen 32-bit values: +Position 0 to 3: constant key +Position 4 to 11: the key +Position 12 to 13: the counter. +Position 14 to 15: the nonce. + +The implementation uses regular C code which compiles very well on modern compilers. +(gcc x64 has no register spills, and clang 6+ uses SSE instructions) +-----------------------------------------------------------------------------*/ + +static inline uint32_t rotl(uint32_t x, uint32_t shift) { + return (x << shift) | (x >> (32 - shift)); +} + +static inline void qround(uint32_t x[16], size_t a, size_t b, size_t c, size_t d) { + x[a] += x[b]; x[d] = rotl(x[d] ^ x[a], 16); + x[c] += x[d]; x[b] = rotl(x[b] ^ x[c], 12); + x[a] += x[b]; x[d] = rotl(x[d] ^ x[a], 8); + x[c] += x[d]; x[b] = rotl(x[b] ^ x[c], 7); +} + +static void chacha_block(mi_random_ctx_t* ctx) +{ + // scramble into `x` + uint32_t x[16]; + for (size_t i = 0; i < 16; i++) { + x[i] = ctx->input[i]; + } + for (size_t i = 0; i < MI_CHACHA_ROUNDS; i += 2) { + qround(x, 0, 4, 8, 12); + qround(x, 1, 5, 9, 13); + qround(x, 2, 6, 10, 14); + qround(x, 3, 7, 11, 15); + qround(x, 0, 5, 10, 15); + qround(x, 1, 6, 11, 12); + qround(x, 2, 7, 8, 13); + qround(x, 3, 4, 9, 14); + } + + // add scrambled data to the initial state + for (size_t i = 0; i < 16; i++) { + ctx->output[i] = x[i] + ctx->input[i]; + } + ctx->output_available = 16; + + // increment the counter for the next round + ctx->input[12] += 1; + if (ctx->input[12] == 0) { + ctx->input[13] += 1; + if (ctx->input[13] == 0) { // and keep increasing into the nonce + ctx->input[14] += 1; + } + } +} + +static uint32_t chacha_next32(mi_random_ctx_t* ctx) { + if (ctx->output_available <= 0) { + chacha_block(ctx); + ctx->output_available = 16; // (assign again to suppress static analysis warning) + } + const uint32_t x = ctx->output[16 - ctx->output_available]; + ctx->output[16 - ctx->output_available] = 0; // reset once the data is handed out + ctx->output_available--; + return x; +} + +static inline uint32_t read32(const uint8_t* p, size_t idx32) { + const size_t i = 4*idx32; + return ((uint32_t)p[i+0] | (uint32_t)p[i+1] << 8 | (uint32_t)p[i+2] << 16 | (uint32_t)p[i+3] << 24); +} + +static void chacha_init(mi_random_ctx_t* ctx, const uint8_t key[32], uint64_t nonce) +{ + // since we only use chacha for randomness (and not encryption) we + // do not _need_ to read 32-bit values as little endian but we do anyways + // just for being compatible :-) + memset(ctx, 0, sizeof(*ctx)); + for (size_t i = 0; i < 4; i++) { + const uint8_t* sigma = (uint8_t*)"expand 32-byte k"; + ctx->input[i] = read32(sigma,i); + } + for (size_t i = 0; i < 8; i++) { + ctx->input[i + 4] = read32(key,i); + } + ctx->input[12] = 0; + ctx->input[13] = 0; + ctx->input[14] = (uint32_t)nonce; + ctx->input[15] = (uint32_t)(nonce >> 32); +} + +static void chacha_split(mi_random_ctx_t* ctx, uint64_t nonce, mi_random_ctx_t* ctx_new) { + memset(ctx_new, 0, sizeof(*ctx_new)); + _mi_memcpy(ctx_new->input, ctx->input, sizeof(ctx_new->input)); + ctx_new->input[12] = 0; + ctx_new->input[13] = 0; + ctx_new->input[14] = (uint32_t)nonce; + ctx_new->input[15] = (uint32_t)(nonce >> 32); + mi_assert_internal(ctx->input[14] != ctx_new->input[14] || ctx->input[15] != ctx_new->input[15]); // do not reuse nonces! + chacha_block(ctx_new); +} + + +/* ---------------------------------------------------------------------------- +Random interface +-----------------------------------------------------------------------------*/ + +#if MI_DEBUG>1 +static bool mi_random_is_initialized(mi_random_ctx_t* ctx) { + return (ctx != NULL && ctx->input[0] != 0); +} +#endif + +void _mi_random_split(mi_random_ctx_t* ctx, mi_random_ctx_t* ctx_new) { + mi_assert_internal(mi_random_is_initialized(ctx)); + mi_assert_internal(ctx != ctx_new); + chacha_split(ctx, (uintptr_t)ctx_new /*nonce*/, ctx_new); +} + +uintptr_t _mi_random_next(mi_random_ctx_t* ctx) { + mi_assert_internal(mi_random_is_initialized(ctx)); + #if MI_INTPTR_SIZE <= 4 + return chacha_next32(ctx); + #elif MI_INTPTR_SIZE == 8 + return (((uintptr_t)chacha_next32(ctx) << 32) | chacha_next32(ctx)); + #else + # error "define mi_random_next for this platform" + #endif +} + + +/* ---------------------------------------------------------------------------- +To initialize a fresh random context. +If we cannot get good randomness, we fall back to weak randomness based on a timer and ASLR. +-----------------------------------------------------------------------------*/ + +uintptr_t _mi_os_random_weak(uintptr_t extra_seed) { + uintptr_t x = (uintptr_t)&_mi_os_random_weak ^ extra_seed; // ASLR makes the address random + x ^= _mi_prim_clock_now(); + // and do a few randomization steps + uintptr_t max = ((x ^ (x >> 17)) & 0x0F) + 1; + for (uintptr_t i = 0; i < max; i++) { + x = _mi_random_shuffle(x); + } + mi_assert_internal(x != 0); + return x; +} + +static void mi_random_init_ex(mi_random_ctx_t* ctx, bool use_weak) { + uint8_t key[32]; + if (use_weak || !_mi_prim_random_buf(key, sizeof(key))) { + // if we fail to get random data from the OS, we fall back to a + // weak random source based on the current time + #if !defined(__wasi__) + if (!use_weak) { _mi_warning_message("unable to use secure randomness\n"); } + #endif + uintptr_t x = _mi_os_random_weak(0); + for (size_t i = 0; i < 8; i++) { // key is eight 32-bit words. + x = _mi_random_shuffle(x); + ((uint32_t*)key)[i] = (uint32_t)x; + } + ctx->weak = true; + } + else { + ctx->weak = false; + } + chacha_init(ctx, key, (uintptr_t)ctx /*nonce*/ ); +} + +void _mi_random_init(mi_random_ctx_t* ctx) { + mi_random_init_ex(ctx, false); +} + +void _mi_random_init_weak(mi_random_ctx_t * ctx) { + mi_random_init_ex(ctx, true); +} + +void _mi_random_reinit_if_weak(mi_random_ctx_t * ctx) { + if (ctx->weak) { + _mi_random_init(ctx); + } +} + +/* -------------------------------------------------------- +test vectors from +----------------------------------------------------------- */ +/* +static bool array_equals(uint32_t* x, uint32_t* y, size_t n) { + for (size_t i = 0; i < n; i++) { + if (x[i] != y[i]) return false; + } + return true; +} +static void chacha_test(void) +{ + uint32_t x[4] = { 0x11111111, 0x01020304, 0x9b8d6f43, 0x01234567 }; + uint32_t x_out[4] = { 0xea2a92f4, 0xcb1cf8ce, 0x4581472e, 0x5881c4bb }; + qround(x, 0, 1, 2, 3); + mi_assert_internal(array_equals(x, x_out, 4)); + + uint32_t y[16] = { + 0x879531e0, 0xc5ecf37d, 0x516461b1, 0xc9a62f8a, + 0x44c20ef3, 0x3390af7f, 0xd9fc690b, 0x2a5f714c, + 0x53372767, 0xb00a5631, 0x974c541a, 0x359e9963, + 0x5c971061, 0x3d631689, 0x2098d9d6, 0x91dbd320 }; + uint32_t y_out[16] = { + 0x879531e0, 0xc5ecf37d, 0xbdb886dc, 0xc9a62f8a, + 0x44c20ef3, 0x3390af7f, 0xd9fc690b, 0xcfacafd2, + 0xe46bea80, 0xb00a5631, 0x974c541a, 0x359e9963, + 0x5c971061, 0xccc07c79, 0x2098d9d6, 0x91dbd320 }; + qround(y, 2, 7, 8, 13); + mi_assert_internal(array_equals(y, y_out, 16)); + + mi_random_ctx_t r = { + { 0x61707865, 0x3320646e, 0x79622d32, 0x6b206574, + 0x03020100, 0x07060504, 0x0b0a0908, 0x0f0e0d0c, + 0x13121110, 0x17161514, 0x1b1a1918, 0x1f1e1d1c, + 0x00000001, 0x09000000, 0x4a000000, 0x00000000 }, + {0}, + 0 + }; + uint32_t r_out[16] = { + 0xe4e7f110, 0x15593bd1, 0x1fdd0f50, 0xc47120a3, + 0xc7f4d1c7, 0x0368c033, 0x9aaa2204, 0x4e6cd4c3, + 0x466482d2, 0x09aa9f07, 0x05d7c214, 0xa2028bd9, + 0xd19c12b5, 0xb94e16de, 0xe883d0cb, 0x4e3c50a2 }; + chacha_block(&r); + mi_assert_internal(array_equals(r.output, r_out, 16)); +} +*/ diff --git a/compat/mimalloc/segment-cache.c b/compat/mimalloc/segment-cache.c new file mode 100644 index 00000000000000..e69de29bb2d1d6 diff --git a/compat/mimalloc/segment-map.c b/compat/mimalloc/segment-map.c new file mode 100644 index 00000000000000..3cd2127e56c1a7 --- /dev/null +++ b/compat/mimalloc/segment-map.c @@ -0,0 +1,153 @@ +/* ---------------------------------------------------------------------------- +Copyright (c) 2019-2023, Microsoft Research, Daan Leijen +This is free software; you can redistribute it and/or modify it under the +terms of the MIT license. A copy of the license can be found in the file +"LICENSE" at the root of this distribution. +-----------------------------------------------------------------------------*/ + +/* ----------------------------------------------------------- + The following functions are to reliably find the segment or + block that encompasses any pointer p (or NULL if it is not + in any of our segments). + We maintain a bitmap of all memory with 1 bit per MI_SEGMENT_SIZE (64MiB) + set to 1 if it contains the segment meta data. +----------------------------------------------------------- */ +#include "mimalloc.h" +#include "mimalloc/internal.h" +#include "mimalloc/atomic.h" + +#if (MI_INTPTR_SIZE==8) +#define MI_MAX_ADDRESS ((size_t)40 << 40) // 40TB (to include huge page areas) +#else +#define MI_MAX_ADDRESS ((size_t)2 << 30) // 2Gb +#endif + +#define MI_SEGMENT_MAP_BITS (MI_MAX_ADDRESS / MI_SEGMENT_SIZE) +#define MI_SEGMENT_MAP_SIZE (MI_SEGMENT_MAP_BITS / 8) +#define MI_SEGMENT_MAP_WSIZE (MI_SEGMENT_MAP_SIZE / MI_INTPTR_SIZE) + +static _Atomic(uintptr_t) mi_segment_map[MI_SEGMENT_MAP_WSIZE + 1]; // 2KiB per TB with 64MiB segments + +static size_t mi_segment_map_index_of(const mi_segment_t* segment, size_t* bitidx) { + mi_assert_internal(_mi_ptr_segment(segment + 1) == segment); // is it aligned on MI_SEGMENT_SIZE? + if ((uintptr_t)segment >= MI_MAX_ADDRESS) { + *bitidx = 0; + return MI_SEGMENT_MAP_WSIZE; + } + else { + const uintptr_t segindex = ((uintptr_t)segment) / MI_SEGMENT_SIZE; + *bitidx = segindex % MI_INTPTR_BITS; + const size_t mapindex = segindex / MI_INTPTR_BITS; + mi_assert_internal(mapindex < MI_SEGMENT_MAP_WSIZE); + return mapindex; + } +} + +void _mi_segment_map_allocated_at(const mi_segment_t* segment) { + size_t bitidx; + size_t index = mi_segment_map_index_of(segment, &bitidx); + mi_assert_internal(index <= MI_SEGMENT_MAP_WSIZE); + if (index==MI_SEGMENT_MAP_WSIZE) return; + uintptr_t mask = mi_atomic_load_relaxed(&mi_segment_map[index]); + uintptr_t newmask; + do { + newmask = (mask | ((uintptr_t)1 << bitidx)); + } while (!mi_atomic_cas_weak_release(&mi_segment_map[index], &mask, newmask)); +} + +void _mi_segment_map_freed_at(const mi_segment_t* segment) { + size_t bitidx; + size_t index = mi_segment_map_index_of(segment, &bitidx); + mi_assert_internal(index <= MI_SEGMENT_MAP_WSIZE); + if (index == MI_SEGMENT_MAP_WSIZE) return; + uintptr_t mask = mi_atomic_load_relaxed(&mi_segment_map[index]); + uintptr_t newmask; + do { + newmask = (mask & ~((uintptr_t)1 << bitidx)); + } while (!mi_atomic_cas_weak_release(&mi_segment_map[index], &mask, newmask)); +} + +// Determine the segment belonging to a pointer or NULL if it is not in a valid segment. +static mi_segment_t* _mi_segment_of(const void* p) { + if (p == NULL) return NULL; + mi_segment_t* segment = _mi_ptr_segment(p); + mi_assert_internal(segment != NULL); + size_t bitidx; + size_t index = mi_segment_map_index_of(segment, &bitidx); + // fast path: for any pointer to valid small/medium/large object or first MI_SEGMENT_SIZE in huge + const uintptr_t mask = mi_atomic_load_relaxed(&mi_segment_map[index]); + if mi_likely((mask & ((uintptr_t)1 << bitidx)) != 0) { + return segment; // yes, allocated by us + } + if (index==MI_SEGMENT_MAP_WSIZE) return NULL; + + // TODO: maintain max/min allocated range for efficiency for more efficient rejection of invalid pointers? + + // search downwards for the first segment in case it is an interior pointer + // could be slow but searches in MI_INTPTR_SIZE * MI_SEGMENT_SIZE (512MiB) steps trough + // valid huge objects + // note: we could maintain a lowest index to speed up the path for invalid pointers? + size_t lobitidx; + size_t loindex; + uintptr_t lobits = mask & (((uintptr_t)1 << bitidx) - 1); + if (lobits != 0) { + loindex = index; + lobitidx = mi_bsr(lobits); // lobits != 0 + } + else if (index == 0) { + return NULL; + } + else { + mi_assert_internal(index > 0); + uintptr_t lomask = mask; + loindex = index; + do { + loindex--; + lomask = mi_atomic_load_relaxed(&mi_segment_map[loindex]); + } while (lomask != 0 && loindex > 0); + if (lomask == 0) return NULL; + lobitidx = mi_bsr(lomask); // lomask != 0 + } + mi_assert_internal(loindex < MI_SEGMENT_MAP_WSIZE); + // take difference as the addresses could be larger than the MAX_ADDRESS space. + size_t diff = (((index - loindex) * (8*MI_INTPTR_SIZE)) + bitidx - lobitidx) * MI_SEGMENT_SIZE; + segment = (mi_segment_t*)((uint8_t*)segment - diff); + + if (segment == NULL) return NULL; + mi_assert_internal((void*)segment < p); + bool cookie_ok = (_mi_ptr_cookie(segment) == segment->cookie); + mi_assert_internal(cookie_ok); + if mi_unlikely(!cookie_ok) return NULL; + if (((uint8_t*)segment + mi_segment_size(segment)) <= (uint8_t*)p) return NULL; // outside the range + mi_assert_internal(p >= (void*)segment && (uint8_t*)p < (uint8_t*)segment + mi_segment_size(segment)); + return segment; +} + +// Is this a valid pointer in our heap? +static bool mi_is_valid_pointer(const void* p) { + return ((_mi_segment_of(p) != NULL) || (_mi_arena_contains(p))); +} + +mi_decl_nodiscard mi_decl_export bool mi_is_in_heap_region(const void* p) mi_attr_noexcept { + return mi_is_valid_pointer(p); +} + +/* +// Return the full segment range belonging to a pointer +static void* mi_segment_range_of(const void* p, size_t* size) { + mi_segment_t* segment = _mi_segment_of(p); + if (segment == NULL) { + if (size != NULL) *size = 0; + return NULL; + } + else { + if (size != NULL) *size = segment->segment_size; + return segment; + } + mi_assert_expensive(page == NULL || mi_segment_is_valid(_mi_page_segment(page),tld)); + mi_assert_internal(page == NULL || (mi_segment_page_size(_mi_page_segment(page)) - (MI_SECURE == 0 ? 0 : _mi_os_page_size())) >= block_size); + mi_reset_delayed(tld); + mi_assert_internal(page == NULL || mi_page_not_in_queue(page, tld)); + return page; +} +*/ diff --git a/compat/mimalloc/segment.c b/compat/mimalloc/segment.c new file mode 100644 index 00000000000000..6b901f6cc80f13 --- /dev/null +++ b/compat/mimalloc/segment.c @@ -0,0 +1,1617 @@ +/* ---------------------------------------------------------------------------- +Copyright (c) 2018-2020, Microsoft Research, Daan Leijen +This is free software; you can redistribute it and/or modify it under the +terms of the MIT license. A copy of the license can be found in the file +"LICENSE" at the root of this distribution. +-----------------------------------------------------------------------------*/ +#include "mimalloc.h" +#include "mimalloc/internal.h" +#include "mimalloc/atomic.h" + +#include // memset +#include + +#define MI_PAGE_HUGE_ALIGN (256*1024) + +static void mi_segment_try_purge(mi_segment_t* segment, bool force, mi_stats_t* stats); + + +// ------------------------------------------------------------------- +// commit mask +// ------------------------------------------------------------------- + +static bool mi_commit_mask_all_set(const mi_commit_mask_t* commit, const mi_commit_mask_t* cm) { + for (size_t i = 0; i < MI_COMMIT_MASK_FIELD_COUNT; i++) { + if ((commit->mask[i] & cm->mask[i]) != cm->mask[i]) return false; + } + return true; +} + +static bool mi_commit_mask_any_set(const mi_commit_mask_t* commit, const mi_commit_mask_t* cm) { + for (size_t i = 0; i < MI_COMMIT_MASK_FIELD_COUNT; i++) { + if ((commit->mask[i] & cm->mask[i]) != 0) return true; + } + return false; +} + +static void mi_commit_mask_create_intersect(const mi_commit_mask_t* commit, const mi_commit_mask_t* cm, mi_commit_mask_t* res) { + for (size_t i = 0; i < MI_COMMIT_MASK_FIELD_COUNT; i++) { + res->mask[i] = (commit->mask[i] & cm->mask[i]); + } +} + +static void mi_commit_mask_clear(mi_commit_mask_t* res, const mi_commit_mask_t* cm) { + for (size_t i = 0; i < MI_COMMIT_MASK_FIELD_COUNT; i++) { + res->mask[i] &= ~(cm->mask[i]); + } +} + +static void mi_commit_mask_set(mi_commit_mask_t* res, const mi_commit_mask_t* cm) { + for (size_t i = 0; i < MI_COMMIT_MASK_FIELD_COUNT; i++) { + res->mask[i] |= cm->mask[i]; + } +} + +static void mi_commit_mask_create(size_t bitidx, size_t bitcount, mi_commit_mask_t* cm) { + mi_assert_internal(bitidx < MI_COMMIT_MASK_BITS); + mi_assert_internal((bitidx + bitcount) <= MI_COMMIT_MASK_BITS); + if (bitcount == MI_COMMIT_MASK_BITS) { + mi_assert_internal(bitidx==0); + mi_commit_mask_create_full(cm); + } + else if (bitcount == 0) { + mi_commit_mask_create_empty(cm); + } + else { + mi_commit_mask_create_empty(cm); + size_t i = bitidx / MI_COMMIT_MASK_FIELD_BITS; + size_t ofs = bitidx % MI_COMMIT_MASK_FIELD_BITS; + while (bitcount > 0) { + mi_assert_internal(i < MI_COMMIT_MASK_FIELD_COUNT); + size_t avail = MI_COMMIT_MASK_FIELD_BITS - ofs; + size_t count = (bitcount > avail ? avail : bitcount); + size_t mask = (count >= MI_COMMIT_MASK_FIELD_BITS ? ~((size_t)0) : (((size_t)1 << count) - 1) << ofs); + cm->mask[i] = mask; + bitcount -= count; + ofs = 0; + i++; + } + } +} + +size_t _mi_commit_mask_committed_size(const mi_commit_mask_t* cm, size_t total) { + mi_assert_internal((total%MI_COMMIT_MASK_BITS)==0); + size_t count = 0; + for (size_t i = 0; i < MI_COMMIT_MASK_FIELD_COUNT; i++) { + size_t mask = cm->mask[i]; + if (~mask == 0) { + count += MI_COMMIT_MASK_FIELD_BITS; + } + else { + for (; mask != 0; mask >>= 1) { // todo: use popcount + if ((mask&1)!=0) count++; + } + } + } + // we use total since for huge segments each commit bit may represent a larger size + return ((total / MI_COMMIT_MASK_BITS) * count); +} + + +size_t _mi_commit_mask_next_run(const mi_commit_mask_t* cm, size_t* idx) { + size_t i = (*idx) / MI_COMMIT_MASK_FIELD_BITS; + size_t ofs = (*idx) % MI_COMMIT_MASK_FIELD_BITS; + size_t mask = 0; + // find first ones + while (i < MI_COMMIT_MASK_FIELD_COUNT) { + mask = cm->mask[i]; + mask >>= ofs; + if (mask != 0) { + while ((mask&1) == 0) { + mask >>= 1; + ofs++; + } + break; + } + i++; + ofs = 0; + } + if (i >= MI_COMMIT_MASK_FIELD_COUNT) { + // not found + *idx = MI_COMMIT_MASK_BITS; + return 0; + } + else { + // found, count ones + size_t count = 0; + *idx = (i*MI_COMMIT_MASK_FIELD_BITS) + ofs; + do { + mi_assert_internal(ofs < MI_COMMIT_MASK_FIELD_BITS && (mask&1) == 1); + do { + count++; + mask >>= 1; + } while ((mask&1) == 1); + if ((((*idx + count) % MI_COMMIT_MASK_FIELD_BITS) == 0)) { + i++; + if (i >= MI_COMMIT_MASK_FIELD_COUNT) break; + mask = cm->mask[i]; + ofs = 0; + } + } while ((mask&1) == 1); + mi_assert_internal(count > 0); + return count; + } +} + + +/* -------------------------------------------------------------------------------- + Segment allocation + + If a thread ends, it "abandons" pages with used blocks + and there is an abandoned segment list whose segments can + be reclaimed by still running threads, much like work-stealing. +-------------------------------------------------------------------------------- */ + + +/* ----------------------------------------------------------- + Slices +----------------------------------------------------------- */ + + +static const mi_slice_t* mi_segment_slices_end(const mi_segment_t* segment) { + return &segment->slices[segment->slice_entries]; +} + +static uint8_t* mi_slice_start(const mi_slice_t* slice) { + mi_segment_t* segment = _mi_ptr_segment(slice); + mi_assert_internal(slice >= segment->slices && slice < mi_segment_slices_end(segment)); + return ((uint8_t*)segment + ((slice - segment->slices)*MI_SEGMENT_SLICE_SIZE)); +} + + +/* ----------------------------------------------------------- + Bins +----------------------------------------------------------- */ +// Use bit scan forward to quickly find the first zero bit if it is available + +static inline size_t mi_slice_bin8(size_t slice_count) { + if (slice_count<=1) return slice_count; + mi_assert_internal(slice_count <= MI_SLICES_PER_SEGMENT); + slice_count--; + size_t s = mi_bsr(slice_count); // slice_count > 1 + if (s <= 2) return slice_count + 1; + size_t bin = ((s << 2) | ((slice_count >> (s - 2))&0x03)) - 4; + return bin; +} + +static inline size_t mi_slice_bin(size_t slice_count) { + mi_assert_internal(slice_count*MI_SEGMENT_SLICE_SIZE <= MI_SEGMENT_SIZE); + mi_assert_internal(mi_slice_bin8(MI_SLICES_PER_SEGMENT) <= MI_SEGMENT_BIN_MAX); + size_t bin = mi_slice_bin8(slice_count); + mi_assert_internal(bin <= MI_SEGMENT_BIN_MAX); + return bin; +} + +static inline size_t mi_slice_index(const mi_slice_t* slice) { + mi_segment_t* segment = _mi_ptr_segment(slice); + ptrdiff_t index = slice - segment->slices; + mi_assert_internal(index >= 0 && index < (ptrdiff_t)segment->slice_entries); + return index; +} + + +/* ----------------------------------------------------------- + Slice span queues +----------------------------------------------------------- */ + +static void mi_span_queue_push(mi_span_queue_t* sq, mi_slice_t* slice) { + // todo: or push to the end? + mi_assert_internal(slice->prev == NULL && slice->next==NULL); + slice->prev = NULL; // paranoia + slice->next = sq->first; + sq->first = slice; + if (slice->next != NULL) slice->next->prev = slice; + else sq->last = slice; + slice->xblock_size = 0; // free +} + +static mi_span_queue_t* mi_span_queue_for(size_t slice_count, mi_segments_tld_t* tld) { + size_t bin = mi_slice_bin(slice_count); + mi_span_queue_t* sq = &tld->spans[bin]; + mi_assert_internal(sq->slice_count >= slice_count); + return sq; +} + +static void mi_span_queue_delete(mi_span_queue_t* sq, mi_slice_t* slice) { + mi_assert_internal(slice->xblock_size==0 && slice->slice_count>0 && slice->slice_offset==0); + // should work too if the queue does not contain slice (which can happen during reclaim) + if (slice->prev != NULL) slice->prev->next = slice->next; + if (slice == sq->first) sq->first = slice->next; + if (slice->next != NULL) slice->next->prev = slice->prev; + if (slice == sq->last) sq->last = slice->prev; + slice->prev = NULL; + slice->next = NULL; + slice->xblock_size = 1; // no more free +} + + +/* ----------------------------------------------------------- + Invariant checking +----------------------------------------------------------- */ + +static bool mi_slice_is_used(const mi_slice_t* slice) { + return (slice->xblock_size > 0); +} + + +#if (MI_DEBUG>=3) +static bool mi_span_queue_contains(mi_span_queue_t* sq, mi_slice_t* slice) { + for (mi_slice_t* s = sq->first; s != NULL; s = s->next) { + if (s==slice) return true; + } + return false; +} + +static bool mi_segment_is_valid(mi_segment_t* segment, mi_segments_tld_t* tld) { + mi_assert_internal(segment != NULL); + mi_assert_internal(_mi_ptr_cookie(segment) == segment->cookie); + mi_assert_internal(segment->abandoned <= segment->used); + mi_assert_internal(segment->thread_id == 0 || segment->thread_id == _mi_thread_id()); + mi_assert_internal(mi_commit_mask_all_set(&segment->commit_mask, &segment->purge_mask)); // can only decommit committed blocks + //mi_assert_internal(segment->segment_info_size % MI_SEGMENT_SLICE_SIZE == 0); + mi_slice_t* slice = &segment->slices[0]; + const mi_slice_t* end = mi_segment_slices_end(segment); + size_t used_count = 0; + mi_span_queue_t* sq; + while(slice < end) { + mi_assert_internal(slice->slice_count > 0); + mi_assert_internal(slice->slice_offset == 0); + size_t index = mi_slice_index(slice); + size_t maxindex = (index + slice->slice_count >= segment->slice_entries ? segment->slice_entries : index + slice->slice_count) - 1; + if (mi_slice_is_used(slice)) { // a page in use, we need at least MAX_SLICE_OFFSET valid back offsets + used_count++; + for (size_t i = 0; i <= MI_MAX_SLICE_OFFSET && index + i <= maxindex; i++) { + mi_assert_internal(segment->slices[index + i].slice_offset == i*sizeof(mi_slice_t)); + mi_assert_internal(i==0 || segment->slices[index + i].slice_count == 0); + mi_assert_internal(i==0 || segment->slices[index + i].xblock_size == 1); + } + // and the last entry as well (for coalescing) + const mi_slice_t* last = slice + slice->slice_count - 1; + if (last > slice && last < mi_segment_slices_end(segment)) { + mi_assert_internal(last->slice_offset == (slice->slice_count-1)*sizeof(mi_slice_t)); + mi_assert_internal(last->slice_count == 0); + mi_assert_internal(last->xblock_size == 1); + } + } + else { // free range of slices; only last slice needs a valid back offset + mi_slice_t* last = &segment->slices[maxindex]; + if (segment->kind != MI_SEGMENT_HUGE || slice->slice_count <= (segment->slice_entries - segment->segment_info_slices)) { + mi_assert_internal((uint8_t*)slice == (uint8_t*)last - last->slice_offset); + } + mi_assert_internal(slice == last || last->slice_count == 0 ); + mi_assert_internal(last->xblock_size == 0 || (segment->kind==MI_SEGMENT_HUGE && last->xblock_size==1)); + if (segment->kind != MI_SEGMENT_HUGE && segment->thread_id != 0) { // segment is not huge or abandoned + sq = mi_span_queue_for(slice->slice_count,tld); + mi_assert_internal(mi_span_queue_contains(sq,slice)); + } + } + slice = &segment->slices[maxindex+1]; + } + mi_assert_internal(slice == end); + mi_assert_internal(used_count == segment->used + 1); + return true; +} +#endif + +/* ----------------------------------------------------------- + Segment size calculations +----------------------------------------------------------- */ + +static size_t mi_segment_info_size(mi_segment_t* segment) { + return segment->segment_info_slices * MI_SEGMENT_SLICE_SIZE; +} + +static uint8_t* _mi_segment_page_start_from_slice(const mi_segment_t* segment, const mi_slice_t* slice, size_t xblock_size, size_t* page_size) +{ + ptrdiff_t idx = slice - segment->slices; + size_t psize = (size_t)slice->slice_count * MI_SEGMENT_SLICE_SIZE; + // make the start not OS page aligned for smaller blocks to avoid page/cache effects + // note: the offset must always be an xblock_size multiple since we assume small allocations + // are aligned (see `mi_heap_malloc_aligned`). + size_t start_offset = 0; + if (xblock_size >= MI_INTPTR_SIZE) { + if (xblock_size <= 64) { start_offset = 3*xblock_size; } + else if (xblock_size <= 512) { start_offset = xblock_size; } + } + if (page_size != NULL) { *page_size = psize - start_offset; } + return (uint8_t*)segment + ((idx*MI_SEGMENT_SLICE_SIZE) + start_offset); +} + +// Start of the page available memory; can be used on uninitialized pages +uint8_t* _mi_segment_page_start(const mi_segment_t* segment, const mi_page_t* page, size_t* page_size) +{ + const mi_slice_t* slice = mi_page_to_slice((mi_page_t*)page); + uint8_t* p = _mi_segment_page_start_from_slice(segment, slice, page->xblock_size, page_size); + mi_assert_internal(page->xblock_size > 0 || _mi_ptr_page(p) == page); + mi_assert_internal(_mi_ptr_segment(p) == segment); + return p; +} + + +static size_t mi_segment_calculate_slices(size_t required, size_t* pre_size, size_t* info_slices) { + size_t page_size = _mi_os_page_size(); + size_t isize = _mi_align_up(sizeof(mi_segment_t), page_size); + size_t guardsize = 0; + + if (MI_SECURE>0) { + // in secure mode, we set up a protected page in between the segment info + // and the page data (and one at the end of the segment) + guardsize = page_size; + if (required > 0) { + required = _mi_align_up(required, MI_SEGMENT_SLICE_SIZE) + page_size; + } + } + + if (pre_size != NULL) *pre_size = isize; + isize = _mi_align_up(isize + guardsize, MI_SEGMENT_SLICE_SIZE); + if (info_slices != NULL) *info_slices = isize / MI_SEGMENT_SLICE_SIZE; + size_t segment_size = (required==0 ? MI_SEGMENT_SIZE : _mi_align_up( required + isize + guardsize, MI_SEGMENT_SLICE_SIZE) ); + mi_assert_internal(segment_size % MI_SEGMENT_SLICE_SIZE == 0); + return (segment_size / MI_SEGMENT_SLICE_SIZE); +} + + +/* ---------------------------------------------------------------------------- +Segment caches +We keep a small segment cache per thread to increase local +reuse and avoid setting/clearing guard pages in secure mode. +------------------------------------------------------------------------------- */ + +static void mi_segments_track_size(long segment_size, mi_segments_tld_t* tld) { + if (segment_size>=0) _mi_stat_increase(&tld->stats->segments,1); + else _mi_stat_decrease(&tld->stats->segments,1); + tld->count += (segment_size >= 0 ? 1 : -1); + if (tld->count > tld->peak_count) tld->peak_count = tld->count; + tld->current_size += segment_size; + if (tld->current_size > tld->peak_size) tld->peak_size = tld->current_size; +} + +static void mi_segment_os_free(mi_segment_t* segment, mi_segments_tld_t* tld) { + segment->thread_id = 0; + _mi_segment_map_freed_at(segment); + mi_segments_track_size(-((long)mi_segment_size(segment)),tld); + if (MI_SECURE>0) { + // _mi_os_unprotect(segment, mi_segment_size(segment)); // ensure no more guard pages are set + // unprotect the guard pages; we cannot just unprotect the whole segment size as part may be decommitted + size_t os_pagesize = _mi_os_page_size(); + _mi_os_unprotect((uint8_t*)segment + mi_segment_info_size(segment) - os_pagesize, os_pagesize); + uint8_t* end = (uint8_t*)segment + mi_segment_size(segment) - os_pagesize; + _mi_os_unprotect(end, os_pagesize); + } + + // purge delayed decommits now? (no, leave it to the arena) + // mi_segment_try_purge(segment,true,tld->stats); + + const size_t size = mi_segment_size(segment); + const size_t csize = _mi_commit_mask_committed_size(&segment->commit_mask, size); + + _mi_abandoned_await_readers(); // wait until safe to free + _mi_arena_free(segment, mi_segment_size(segment), csize, segment->memid, tld->stats); +} + +// called by threads that are terminating +void _mi_segment_thread_collect(mi_segments_tld_t* tld) { + MI_UNUSED(tld); + // nothing to do +} + + +/* ----------------------------------------------------------- + Commit/Decommit ranges +----------------------------------------------------------- */ + +static void mi_segment_commit_mask(mi_segment_t* segment, bool conservative, uint8_t* p, size_t size, uint8_t** start_p, size_t* full_size, mi_commit_mask_t* cm) { + mi_assert_internal(_mi_ptr_segment(p + 1) == segment); + mi_assert_internal(segment->kind != MI_SEGMENT_HUGE); + mi_commit_mask_create_empty(cm); + if (size == 0 || size > MI_SEGMENT_SIZE || segment->kind == MI_SEGMENT_HUGE) return; + const size_t segstart = mi_segment_info_size(segment); + const size_t segsize = mi_segment_size(segment); + if (p >= (uint8_t*)segment + segsize) return; + + size_t pstart = (p - (uint8_t*)segment); + mi_assert_internal(pstart + size <= segsize); + + size_t start; + size_t end; + if (conservative) { + // decommit conservative + start = _mi_align_up(pstart, MI_COMMIT_SIZE); + end = _mi_align_down(pstart + size, MI_COMMIT_SIZE); + mi_assert_internal(start >= segstart); + mi_assert_internal(end <= segsize); + } + else { + // commit liberal + start = _mi_align_down(pstart, MI_MINIMAL_COMMIT_SIZE); + end = _mi_align_up(pstart + size, MI_MINIMAL_COMMIT_SIZE); + } + if (pstart >= segstart && start < segstart) { // note: the mask is also calculated for an initial commit of the info area + start = segstart; + } + if (end > segsize) { + end = segsize; + } + + mi_assert_internal(start <= pstart && (pstart + size) <= end); + mi_assert_internal(start % MI_COMMIT_SIZE==0 && end % MI_COMMIT_SIZE == 0); + *start_p = (uint8_t*)segment + start; + *full_size = (end > start ? end - start : 0); + if (*full_size == 0) return; + + size_t bitidx = start / MI_COMMIT_SIZE; + mi_assert_internal(bitidx < MI_COMMIT_MASK_BITS); + + size_t bitcount = *full_size / MI_COMMIT_SIZE; // can be 0 + if (bitidx + bitcount > MI_COMMIT_MASK_BITS) { + _mi_warning_message("commit mask overflow: idx=%zu count=%zu start=%zx end=%zx p=0x%p size=%zu fullsize=%zu\n", bitidx, bitcount, start, end, p, size, *full_size); + } + mi_assert_internal((bitidx + bitcount) <= MI_COMMIT_MASK_BITS); + mi_commit_mask_create(bitidx, bitcount, cm); +} + +static bool mi_segment_commit(mi_segment_t* segment, uint8_t* p, size_t size, mi_stats_t* stats) { + mi_assert_internal(mi_commit_mask_all_set(&segment->commit_mask, &segment->purge_mask)); + + // commit liberal + uint8_t* start = NULL; + size_t full_size = 0; + mi_commit_mask_t mask; + mi_segment_commit_mask(segment, false /* conservative? */, p, size, &start, &full_size, &mask); + if (mi_commit_mask_is_empty(&mask) || full_size == 0) return true; + + if (!mi_commit_mask_all_set(&segment->commit_mask, &mask)) { + // committing + bool is_zero = false; + mi_commit_mask_t cmask; + mi_commit_mask_create_intersect(&segment->commit_mask, &mask, &cmask); + _mi_stat_decrease(&_mi_stats_main.committed, _mi_commit_mask_committed_size(&cmask, MI_SEGMENT_SIZE)); // adjust for overlap + if (!_mi_os_commit(start, full_size, &is_zero, stats)) return false; + mi_commit_mask_set(&segment->commit_mask, &mask); + } + + // increase purge expiration when using part of delayed purges -- we assume more allocations are coming soon. + if (mi_commit_mask_any_set(&segment->purge_mask, &mask)) { + segment->purge_expire = _mi_clock_now() + mi_option_get(mi_option_purge_delay); + } + + // always clear any delayed purges in our range (as they are either committed now) + mi_commit_mask_clear(&segment->purge_mask, &mask); + return true; +} + +static bool mi_segment_ensure_committed(mi_segment_t* segment, uint8_t* p, size_t size, mi_stats_t* stats) { + mi_assert_internal(mi_commit_mask_all_set(&segment->commit_mask, &segment->purge_mask)); + // note: assumes commit_mask is always full for huge segments as otherwise the commit mask bits can overflow + if (mi_commit_mask_is_full(&segment->commit_mask) && mi_commit_mask_is_empty(&segment->purge_mask)) return true; // fully committed + mi_assert_internal(segment->kind != MI_SEGMENT_HUGE); + return mi_segment_commit(segment, p, size, stats); +} + +static bool mi_segment_purge(mi_segment_t* segment, uint8_t* p, size_t size, mi_stats_t* stats) { + mi_assert_internal(mi_commit_mask_all_set(&segment->commit_mask, &segment->purge_mask)); + if (!segment->allow_purge) return true; + + // purge conservative + uint8_t* start = NULL; + size_t full_size = 0; + mi_commit_mask_t mask; + mi_segment_commit_mask(segment, true /* conservative? */, p, size, &start, &full_size, &mask); + if (mi_commit_mask_is_empty(&mask) || full_size==0) return true; + + if (mi_commit_mask_any_set(&segment->commit_mask, &mask)) { + // purging + mi_assert_internal((void*)start != (void*)segment); + mi_assert_internal(segment->allow_decommit); + const bool decommitted = _mi_os_purge(start, full_size, stats); // reset or decommit + if (decommitted) { + mi_commit_mask_t cmask; + mi_commit_mask_create_intersect(&segment->commit_mask, &mask, &cmask); + _mi_stat_increase(&_mi_stats_main.committed, full_size - _mi_commit_mask_committed_size(&cmask, MI_SEGMENT_SIZE)); // adjust for double counting + mi_commit_mask_clear(&segment->commit_mask, &mask); + } + } + + // always clear any scheduled purges in our range + mi_commit_mask_clear(&segment->purge_mask, &mask); + return true; +} + +static void mi_segment_schedule_purge(mi_segment_t* segment, uint8_t* p, size_t size, mi_stats_t* stats) { + if (!segment->allow_purge) return; + + if (mi_option_get(mi_option_purge_delay) == 0) { + mi_segment_purge(segment, p, size, stats); + } + else { + // register for future purge in the purge mask + uint8_t* start = NULL; + size_t full_size = 0; + mi_commit_mask_t mask; + mi_segment_commit_mask(segment, true /*conservative*/, p, size, &start, &full_size, &mask); + if (mi_commit_mask_is_empty(&mask) || full_size==0) return; + + // update delayed commit + mi_assert_internal(segment->purge_expire > 0 || mi_commit_mask_is_empty(&segment->purge_mask)); + mi_commit_mask_t cmask; + mi_commit_mask_create_intersect(&segment->commit_mask, &mask, &cmask); // only purge what is committed; span_free may try to decommit more + mi_commit_mask_set(&segment->purge_mask, &cmask); + mi_msecs_t now = _mi_clock_now(); + if (segment->purge_expire == 0) { + // no previous purgess, initialize now + segment->purge_expire = now + mi_option_get(mi_option_purge_delay); + } + else if (segment->purge_expire <= now) { + // previous purge mask already expired + if (segment->purge_expire + mi_option_get(mi_option_purge_extend_delay) <= now) { + mi_segment_try_purge(segment, true, stats); + } + else { + segment->purge_expire = now + mi_option_get(mi_option_purge_extend_delay); // (mi_option_get(mi_option_purge_delay) / 8); // wait a tiny bit longer in case there is a series of free's + } + } + else { + // previous purge mask is not yet expired, increase the expiration by a bit. + segment->purge_expire += mi_option_get(mi_option_purge_extend_delay); + } + } +} + +static void mi_segment_try_purge(mi_segment_t* segment, bool force, mi_stats_t* stats) { + if (!segment->allow_purge || mi_commit_mask_is_empty(&segment->purge_mask)) return; + mi_msecs_t now = _mi_clock_now(); + if (!force && now < segment->purge_expire) return; + + mi_commit_mask_t mask = segment->purge_mask; + segment->purge_expire = 0; + mi_commit_mask_create_empty(&segment->purge_mask); + + size_t idx; + size_t count; + mi_commit_mask_foreach(&mask, idx, count) { + // if found, decommit that sequence + if (count > 0) { + uint8_t* p = (uint8_t*)segment + (idx*MI_COMMIT_SIZE); + size_t size = count * MI_COMMIT_SIZE; + mi_segment_purge(segment, p, size, stats); + } + } + mi_commit_mask_foreach_end() + mi_assert_internal(mi_commit_mask_is_empty(&segment->purge_mask)); +} + + +/* ----------------------------------------------------------- + Span free +----------------------------------------------------------- */ + +static bool mi_segment_is_abandoned(mi_segment_t* segment) { + return (segment->thread_id == 0); +} + +// note: can be called on abandoned segments +static void mi_segment_span_free(mi_segment_t* segment, size_t slice_index, size_t slice_count, bool allow_purge, mi_segments_tld_t* tld) { + mi_assert_internal(slice_index < segment->slice_entries); + mi_span_queue_t* sq = (segment->kind == MI_SEGMENT_HUGE || mi_segment_is_abandoned(segment) + ? NULL : mi_span_queue_for(slice_count,tld)); + if (slice_count==0) slice_count = 1; + mi_assert_internal(slice_index + slice_count - 1 < segment->slice_entries); + + // set first and last slice (the intermediates can be undetermined) + mi_slice_t* slice = &segment->slices[slice_index]; + slice->slice_count = (uint32_t)slice_count; + mi_assert_internal(slice->slice_count == slice_count); // no overflow? + slice->slice_offset = 0; + if (slice_count > 1) { + mi_slice_t* last = &segment->slices[slice_index + slice_count - 1]; + last->slice_count = 0; + last->slice_offset = (uint32_t)(sizeof(mi_page_t)*(slice_count - 1)); + last->xblock_size = 0; + } + + // perhaps decommit + if (allow_purge) { + mi_segment_schedule_purge(segment, mi_slice_start(slice), slice_count * MI_SEGMENT_SLICE_SIZE, tld->stats); + } + + // and push it on the free page queue (if it was not a huge page) + if (sq != NULL) mi_span_queue_push( sq, slice ); + else slice->xblock_size = 0; // mark huge page as free anyways +} + +/* +// called from reclaim to add existing free spans +static void mi_segment_span_add_free(mi_slice_t* slice, mi_segments_tld_t* tld) { + mi_segment_t* segment = _mi_ptr_segment(slice); + mi_assert_internal(slice->xblock_size==0 && slice->slice_count>0 && slice->slice_offset==0); + size_t slice_index = mi_slice_index(slice); + mi_segment_span_free(segment,slice_index,slice->slice_count,tld); +} +*/ + +static void mi_segment_span_remove_from_queue(mi_slice_t* slice, mi_segments_tld_t* tld) { + mi_assert_internal(slice->slice_count > 0 && slice->slice_offset==0 && slice->xblock_size==0); + mi_assert_internal(_mi_ptr_segment(slice)->kind != MI_SEGMENT_HUGE); + mi_span_queue_t* sq = mi_span_queue_for(slice->slice_count, tld); + mi_span_queue_delete(sq, slice); +} + +// note: can be called on abandoned segments +static mi_slice_t* mi_segment_span_free_coalesce(mi_slice_t* slice, mi_segments_tld_t* tld) { + mi_assert_internal(slice != NULL && slice->slice_count > 0 && slice->slice_offset == 0); + mi_segment_t* segment = _mi_ptr_segment(slice); + bool is_abandoned = mi_segment_is_abandoned(segment); + + // for huge pages, just mark as free but don't add to the queues + if (segment->kind == MI_SEGMENT_HUGE) { + // issue #691: segment->used can be 0 if the huge page block was freed while abandoned (reclaim will get here in that case) + mi_assert_internal((segment->used==0 && slice->xblock_size==0) || segment->used == 1); // decreased right after this call in `mi_segment_page_clear` + slice->xblock_size = 0; // mark as free anyways + // we should mark the last slice `xblock_size=0` now to maintain invariants but we skip it to + // avoid a possible cache miss (and the segment is about to be freed) + return slice; + } + + // otherwise coalesce the span and add to the free span queues + size_t slice_count = slice->slice_count; + mi_slice_t* next = slice + slice->slice_count; + mi_assert_internal(next <= mi_segment_slices_end(segment)); + if (next < mi_segment_slices_end(segment) && next->xblock_size==0) { + // free next block -- remove it from free and merge + mi_assert_internal(next->slice_count > 0 && next->slice_offset==0); + slice_count += next->slice_count; // extend + if (!is_abandoned) { mi_segment_span_remove_from_queue(next, tld); } + } + if (slice > segment->slices) { + mi_slice_t* prev = mi_slice_first(slice - 1); + mi_assert_internal(prev >= segment->slices); + if (prev->xblock_size==0) { + // free previous slice -- remove it from free and merge + mi_assert_internal(prev->slice_count > 0 && prev->slice_offset==0); + slice_count += prev->slice_count; + if (!is_abandoned) { mi_segment_span_remove_from_queue(prev, tld); } + slice = prev; + } + } + + // and add the new free page + mi_segment_span_free(segment, mi_slice_index(slice), slice_count, true, tld); + return slice; +} + + + +/* ----------------------------------------------------------- + Page allocation +----------------------------------------------------------- */ + +// Note: may still return NULL if committing the memory failed +static mi_page_t* mi_segment_span_allocate(mi_segment_t* segment, size_t slice_index, size_t slice_count, mi_segments_tld_t* tld) { + mi_assert_internal(slice_index < segment->slice_entries); + mi_slice_t* const slice = &segment->slices[slice_index]; + mi_assert_internal(slice->xblock_size==0 || slice->xblock_size==1); + + // commit before changing the slice data + if (!mi_segment_ensure_committed(segment, _mi_segment_page_start_from_slice(segment, slice, 0, NULL), slice_count * MI_SEGMENT_SLICE_SIZE, tld->stats)) { + return NULL; // commit failed! + } + + // convert the slices to a page + slice->slice_offset = 0; + slice->slice_count = (uint32_t)slice_count; + mi_assert_internal(slice->slice_count == slice_count); + const size_t bsize = slice_count * MI_SEGMENT_SLICE_SIZE; + slice->xblock_size = (uint32_t)(bsize >= MI_HUGE_BLOCK_SIZE ? MI_HUGE_BLOCK_SIZE : bsize); + mi_page_t* page = mi_slice_to_page(slice); + mi_assert_internal(mi_page_block_size(page) == bsize); + + // set slice back pointers for the first MI_MAX_SLICE_OFFSET entries + size_t extra = slice_count-1; + if (extra > MI_MAX_SLICE_OFFSET) extra = MI_MAX_SLICE_OFFSET; + if (slice_index + extra >= segment->slice_entries) extra = segment->slice_entries - slice_index - 1; // huge objects may have more slices than avaiable entries in the segment->slices + + mi_slice_t* slice_next = slice + 1; + for (size_t i = 1; i <= extra; i++, slice_next++) { + slice_next->slice_offset = (uint32_t)(sizeof(mi_slice_t)*i); + slice_next->slice_count = 0; + slice_next->xblock_size = 1; + } + + // and also for the last one (if not set already) (the last one is needed for coalescing and for large alignments) + // note: the cast is needed for ubsan since the index can be larger than MI_SLICES_PER_SEGMENT for huge allocations (see #543) + mi_slice_t* last = slice + slice_count - 1; + mi_slice_t* end = (mi_slice_t*)mi_segment_slices_end(segment); + if (last > end) last = end; + if (last > slice) { + last->slice_offset = (uint32_t)(sizeof(mi_slice_t) * (last - slice)); + last->slice_count = 0; + last->xblock_size = 1; + } + + // and initialize the page + page->is_committed = true; + segment->used++; + return page; +} + +static void mi_segment_slice_split(mi_segment_t* segment, mi_slice_t* slice, size_t slice_count, mi_segments_tld_t* tld) { + mi_assert_internal(_mi_ptr_segment(slice) == segment); + mi_assert_internal(slice->slice_count >= slice_count); + mi_assert_internal(slice->xblock_size > 0); // no more in free queue + if (slice->slice_count <= slice_count) return; + mi_assert_internal(segment->kind != MI_SEGMENT_HUGE); + size_t next_index = mi_slice_index(slice) + slice_count; + size_t next_count = slice->slice_count - slice_count; + mi_segment_span_free(segment, next_index, next_count, false /* don't purge left-over part */, tld); + slice->slice_count = (uint32_t)slice_count; +} + +static mi_page_t* mi_segments_page_find_and_allocate(size_t slice_count, mi_arena_id_t req_arena_id, mi_segments_tld_t* tld) { + mi_assert_internal(slice_count*MI_SEGMENT_SLICE_SIZE <= MI_LARGE_OBJ_SIZE_MAX); + // search from best fit up + mi_span_queue_t* sq = mi_span_queue_for(slice_count, tld); + if (slice_count == 0) slice_count = 1; + while (sq <= &tld->spans[MI_SEGMENT_BIN_MAX]) { + for (mi_slice_t* slice = sq->first; slice != NULL; slice = slice->next) { + if (slice->slice_count >= slice_count) { + // found one + mi_segment_t* segment = _mi_ptr_segment(slice); + if (_mi_arena_memid_is_suitable(segment->memid, req_arena_id)) { + // found a suitable page span + mi_span_queue_delete(sq, slice); + + if (slice->slice_count > slice_count) { + mi_segment_slice_split(segment, slice, slice_count, tld); + } + mi_assert_internal(slice != NULL && slice->slice_count == slice_count && slice->xblock_size > 0); + mi_page_t* page = mi_segment_span_allocate(segment, mi_slice_index(slice), slice->slice_count, tld); + if (page == NULL) { + // commit failed; return NULL but first restore the slice + mi_segment_span_free_coalesce(slice, tld); + return NULL; + } + return page; + } + } + } + sq++; + } + // could not find a page.. + return NULL; +} + + +/* ----------------------------------------------------------- + Segment allocation +----------------------------------------------------------- */ + +static mi_segment_t* mi_segment_os_alloc( size_t required, size_t page_alignment, bool eager_delayed, mi_arena_id_t req_arena_id, + size_t* psegment_slices, size_t* ppre_size, size_t* pinfo_slices, + bool commit, mi_segments_tld_t* tld, mi_os_tld_t* os_tld) + +{ + mi_memid_t memid; + bool allow_large = (!eager_delayed && (MI_SECURE == 0)); // only allow large OS pages once we are no longer lazy + size_t align_offset = 0; + size_t alignment = MI_SEGMENT_ALIGN; + + if (page_alignment > 0) { + // mi_assert_internal(huge_page != NULL); + mi_assert_internal(page_alignment >= MI_SEGMENT_ALIGN); + alignment = page_alignment; + const size_t info_size = (*pinfo_slices) * MI_SEGMENT_SLICE_SIZE; + align_offset = _mi_align_up( info_size, MI_SEGMENT_ALIGN ); + const size_t extra = align_offset - info_size; + // recalculate due to potential guard pages + *psegment_slices = mi_segment_calculate_slices(required + extra, ppre_size, pinfo_slices); + } + + const size_t segment_size = (*psegment_slices) * MI_SEGMENT_SLICE_SIZE; + mi_segment_t* segment = (mi_segment_t*)_mi_arena_alloc_aligned(segment_size, alignment, align_offset, commit, allow_large, req_arena_id, &memid, os_tld); + if (segment == NULL) { + return NULL; // failed to allocate + } + + // ensure metadata part of the segment is committed + mi_commit_mask_t commit_mask; + if (memid.initially_committed) { + mi_commit_mask_create_full(&commit_mask); + } + else { + // at least commit the info slices + const size_t commit_needed = _mi_divide_up((*pinfo_slices)*MI_SEGMENT_SLICE_SIZE, MI_COMMIT_SIZE); + mi_assert_internal(commit_needed>0); + mi_commit_mask_create(0, commit_needed, &commit_mask); + mi_assert_internal(commit_needed*MI_COMMIT_SIZE >= (*pinfo_slices)*MI_SEGMENT_SLICE_SIZE); + if (!_mi_os_commit(segment, commit_needed*MI_COMMIT_SIZE, NULL, tld->stats)) { + _mi_arena_free(segment,segment_size,0,memid,tld->stats); + return NULL; + } + } + mi_assert_internal(segment != NULL && (uintptr_t)segment % MI_SEGMENT_SIZE == 0); + + segment->memid = memid; + segment->allow_decommit = !memid.is_pinned; + segment->allow_purge = segment->allow_decommit && (mi_option_get(mi_option_purge_delay) >= 0); + segment->segment_size = segment_size; + segment->commit_mask = commit_mask; + segment->purge_expire = 0; + mi_commit_mask_create_empty(&segment->purge_mask); + mi_atomic_store_ptr_release(mi_segment_t, &segment->abandoned_next, NULL); // tsan + + mi_segments_track_size((long)(segment_size), tld); + _mi_segment_map_allocated_at(segment); + return segment; +} + + +// Allocate a segment from the OS aligned to `MI_SEGMENT_SIZE` . +static mi_segment_t* mi_segment_alloc(size_t required, size_t page_alignment, mi_arena_id_t req_arena_id, mi_segments_tld_t* tld, mi_os_tld_t* os_tld, mi_page_t** huge_page) +{ + mi_assert_internal((required==0 && huge_page==NULL) || (required>0 && huge_page != NULL)); + + // calculate needed sizes first + size_t info_slices; + size_t pre_size; + size_t segment_slices = mi_segment_calculate_slices(required, &pre_size, &info_slices); + + // Commit eagerly only if not the first N lazy segments (to reduce impact of many threads that allocate just a little) + const bool eager_delay = (// !_mi_os_has_overcommit() && // never delay on overcommit systems + _mi_current_thread_count() > 1 && // do not delay for the first N threads + tld->count < (size_t)mi_option_get(mi_option_eager_commit_delay)); + const bool eager = !eager_delay && mi_option_is_enabled(mi_option_eager_commit); + bool commit = eager || (required > 0); + + // Allocate the segment from the OS + mi_segment_t* segment = mi_segment_os_alloc(required, page_alignment, eager_delay, req_arena_id, + &segment_slices, &pre_size, &info_slices, commit, tld, os_tld); + if (segment == NULL) return NULL; + + // zero the segment info? -- not always needed as it may be zero initialized from the OS + if (!segment->memid.initially_zero) { + ptrdiff_t ofs = offsetof(mi_segment_t, next); + size_t prefix = offsetof(mi_segment_t, slices) - ofs; + size_t zsize = prefix + (sizeof(mi_slice_t) * (segment_slices + 1)); // one more + _mi_memzero((uint8_t*)segment + ofs, zsize); + } + + // initialize the rest of the segment info + const size_t slice_entries = (segment_slices > MI_SLICES_PER_SEGMENT ? MI_SLICES_PER_SEGMENT : segment_slices); + segment->segment_slices = segment_slices; + segment->segment_info_slices = info_slices; + segment->thread_id = _mi_thread_id(); + segment->cookie = _mi_ptr_cookie(segment); + segment->slice_entries = slice_entries; + segment->kind = (required == 0 ? MI_SEGMENT_NORMAL : MI_SEGMENT_HUGE); + + // _mi_memzero(segment->slices, sizeof(mi_slice_t)*(info_slices+1)); + _mi_stat_increase(&tld->stats->page_committed, mi_segment_info_size(segment)); + + // set up guard pages + size_t guard_slices = 0; + if (MI_SECURE>0) { + // in secure mode, we set up a protected page in between the segment info + // and the page data, and at the end of the segment. + size_t os_pagesize = _mi_os_page_size(); + mi_assert_internal(mi_segment_info_size(segment) - os_pagesize >= pre_size); + _mi_os_protect((uint8_t*)segment + mi_segment_info_size(segment) - os_pagesize, os_pagesize); + uint8_t* end = (uint8_t*)segment + mi_segment_size(segment) - os_pagesize; + mi_segment_ensure_committed(segment, end, os_pagesize, tld->stats); + _mi_os_protect(end, os_pagesize); + if (slice_entries == segment_slices) segment->slice_entries--; // don't use the last slice :-( + guard_slices = 1; + } + + // reserve first slices for segment info + mi_page_t* page0 = mi_segment_span_allocate(segment, 0, info_slices, tld); + mi_assert_internal(page0!=NULL); if (page0==NULL) return NULL; // cannot fail as we always commit in advance + mi_assert_internal(segment->used == 1); + segment->used = 0; // don't count our internal slices towards usage + + // initialize initial free pages + if (segment->kind == MI_SEGMENT_NORMAL) { // not a huge page + mi_assert_internal(huge_page==NULL); + mi_segment_span_free(segment, info_slices, segment->slice_entries - info_slices, false /* don't purge */, tld); + } + else { + mi_assert_internal(huge_page!=NULL); + mi_assert_internal(mi_commit_mask_is_empty(&segment->purge_mask)); + mi_assert_internal(mi_commit_mask_is_full(&segment->commit_mask)); + *huge_page = mi_segment_span_allocate(segment, info_slices, segment_slices - info_slices - guard_slices, tld); + mi_assert_internal(*huge_page != NULL); // cannot fail as we commit in advance + } + + mi_assert_expensive(mi_segment_is_valid(segment,tld)); + return segment; +} + + +static void mi_segment_free(mi_segment_t* segment, bool force, mi_segments_tld_t* tld) { + MI_UNUSED(force); + mi_assert_internal(segment != NULL); + mi_assert_internal(segment->next == NULL); + mi_assert_internal(segment->used == 0); + + // Remove the free pages + mi_slice_t* slice = &segment->slices[0]; + const mi_slice_t* end = mi_segment_slices_end(segment); + #if MI_DEBUG>1 + size_t page_count = 0; + #endif + while (slice < end) { + mi_assert_internal(slice->slice_count > 0); + mi_assert_internal(slice->slice_offset == 0); + mi_assert_internal(mi_slice_index(slice)==0 || slice->xblock_size == 0); // no more used pages .. + if (slice->xblock_size == 0 && segment->kind != MI_SEGMENT_HUGE) { + mi_segment_span_remove_from_queue(slice, tld); + } + #if MI_DEBUG>1 + page_count++; + #endif + slice = slice + slice->slice_count; + } + mi_assert_internal(page_count == 2); // first page is allocated by the segment itself + + // stats + _mi_stat_decrease(&tld->stats->page_committed, mi_segment_info_size(segment)); + + // return it to the OS + mi_segment_os_free(segment, tld); +} + + +/* ----------------------------------------------------------- + Page Free +----------------------------------------------------------- */ + +static void mi_segment_abandon(mi_segment_t* segment, mi_segments_tld_t* tld); + +// note: can be called on abandoned pages +static mi_slice_t* mi_segment_page_clear(mi_page_t* page, mi_segments_tld_t* tld) { + mi_assert_internal(page->xblock_size > 0); + mi_assert_internal(mi_page_all_free(page)); + mi_segment_t* segment = _mi_ptr_segment(page); + mi_assert_internal(segment->used > 0); + + size_t inuse = page->capacity * mi_page_block_size(page); + _mi_stat_decrease(&tld->stats->page_committed, inuse); + _mi_stat_decrease(&tld->stats->pages, 1); + + // reset the page memory to reduce memory pressure? + if (segment->allow_decommit && mi_option_is_enabled(mi_option_deprecated_page_reset)) { + size_t psize; + uint8_t* start = _mi_page_start(segment, page, &psize); + _mi_os_reset(start, psize, tld->stats); + } + + // zero the page data, but not the segment fields + page->is_zero_init = false; + ptrdiff_t ofs = offsetof(mi_page_t, capacity); + _mi_memzero((uint8_t*)page + ofs, sizeof(*page) - ofs); + page->xblock_size = 1; + + // and free it + mi_slice_t* slice = mi_segment_span_free_coalesce(mi_page_to_slice(page), tld); + segment->used--; + // cannot assert segment valid as it is called during reclaim + // mi_assert_expensive(mi_segment_is_valid(segment, tld)); + return slice; +} + +void _mi_segment_page_free(mi_page_t* page, bool force, mi_segments_tld_t* tld) +{ + mi_assert(page != NULL); + + mi_segment_t* segment = _mi_page_segment(page); + mi_assert_expensive(mi_segment_is_valid(segment,tld)); + + // mark it as free now + mi_segment_page_clear(page, tld); + mi_assert_expensive(mi_segment_is_valid(segment, tld)); + + if (segment->used == 0) { + // no more used pages; remove from the free list and free the segment + mi_segment_free(segment, force, tld); + } + else if (segment->used == segment->abandoned) { + // only abandoned pages; remove from free list and abandon + mi_segment_abandon(segment,tld); + } +} + + +/* ----------------------------------------------------------- +Abandonment + +When threads terminate, they can leave segments with +live blocks (reachable through other threads). Such segments +are "abandoned" and will be reclaimed by other threads to +reuse their pages and/or free them eventually + +We maintain a global list of abandoned segments that are +reclaimed on demand. Since this is shared among threads +the implementation needs to avoid the A-B-A problem on +popping abandoned segments: +We use tagged pointers to avoid accidentally identifying +reused segments, much like stamped references in Java. +Secondly, we maintain a reader counter to avoid resetting +or decommitting segments that have a pending read operation. + +Note: the current implementation is one possible design; +another way might be to keep track of abandoned segments +in the arenas/segment_cache's. This would have the advantage of keeping +all concurrent code in one place and not needing to deal +with ABA issues. The drawback is that it is unclear how to +scan abandoned segments efficiently in that case as they +would be spread among all other segments in the arenas. +----------------------------------------------------------- */ + +// Use the bottom 20-bits (on 64-bit) of the aligned segment pointers +// to put in a tag that increments on update to avoid the A-B-A problem. +#define MI_TAGGED_MASK MI_SEGMENT_MASK +typedef uintptr_t mi_tagged_segment_t; + +static mi_segment_t* mi_tagged_segment_ptr(mi_tagged_segment_t ts) { + return (mi_segment_t*)(ts & ~MI_TAGGED_MASK); +} + +static mi_tagged_segment_t mi_tagged_segment(mi_segment_t* segment, mi_tagged_segment_t ts) { + mi_assert_internal(((uintptr_t)segment & MI_TAGGED_MASK) == 0); + uintptr_t tag = ((ts & MI_TAGGED_MASK) + 1) & MI_TAGGED_MASK; + return ((uintptr_t)segment | tag); +} + +// This is a list of visited abandoned pages that were full at the time. +// this list migrates to `abandoned` when that becomes NULL. The use of +// this list reduces contention and the rate at which segments are visited. +static mi_decl_cache_align _Atomic(mi_segment_t*) abandoned_visited; // = NULL + +// The abandoned page list (tagged as it supports pop) +static mi_decl_cache_align _Atomic(mi_tagged_segment_t) abandoned; // = NULL + +// Maintain these for debug purposes (these counts may be a bit off) +static mi_decl_cache_align _Atomic(size_t) abandoned_count; +static mi_decl_cache_align _Atomic(size_t) abandoned_visited_count; + +// We also maintain a count of current readers of the abandoned list +// in order to prevent resetting/decommitting segment memory if it might +// still be read. +static mi_decl_cache_align _Atomic(size_t) abandoned_readers; // = 0 + +// Push on the visited list +static void mi_abandoned_visited_push(mi_segment_t* segment) { + mi_assert_internal(segment->thread_id == 0); + mi_assert_internal(mi_atomic_load_ptr_relaxed(mi_segment_t,&segment->abandoned_next) == NULL); + mi_assert_internal(segment->next == NULL); + mi_assert_internal(segment->used > 0); + mi_segment_t* anext = mi_atomic_load_ptr_relaxed(mi_segment_t, &abandoned_visited); + do { + mi_atomic_store_ptr_release(mi_segment_t, &segment->abandoned_next, anext); + } while (!mi_atomic_cas_ptr_weak_release(mi_segment_t, &abandoned_visited, &anext, segment)); + mi_atomic_increment_relaxed(&abandoned_visited_count); +} + +// Move the visited list to the abandoned list. +static bool mi_abandoned_visited_revisit(void) +{ + // quick check if the visited list is empty + if (mi_atomic_load_ptr_relaxed(mi_segment_t, &abandoned_visited) == NULL) return false; + + // grab the whole visited list + mi_segment_t* first = mi_atomic_exchange_ptr_acq_rel(mi_segment_t, &abandoned_visited, NULL); + if (first == NULL) return false; + + // first try to swap directly if the abandoned list happens to be NULL + mi_tagged_segment_t afirst; + mi_tagged_segment_t ts = mi_atomic_load_relaxed(&abandoned); + if (mi_tagged_segment_ptr(ts)==NULL) { + size_t count = mi_atomic_load_relaxed(&abandoned_visited_count); + afirst = mi_tagged_segment(first, ts); + if (mi_atomic_cas_strong_acq_rel(&abandoned, &ts, afirst)) { + mi_atomic_add_relaxed(&abandoned_count, count); + mi_atomic_sub_relaxed(&abandoned_visited_count, count); + return true; + } + } + + // find the last element of the visited list: O(n) + mi_segment_t* last = first; + mi_segment_t* next; + while ((next = mi_atomic_load_ptr_relaxed(mi_segment_t, &last->abandoned_next)) != NULL) { + last = next; + } + + // and atomically prepend to the abandoned list + // (no need to increase the readers as we don't access the abandoned segments) + mi_tagged_segment_t anext = mi_atomic_load_relaxed(&abandoned); + size_t count; + do { + count = mi_atomic_load_relaxed(&abandoned_visited_count); + mi_atomic_store_ptr_release(mi_segment_t, &last->abandoned_next, mi_tagged_segment_ptr(anext)); + afirst = mi_tagged_segment(first, anext); + } while (!mi_atomic_cas_weak_release(&abandoned, &anext, afirst)); + mi_atomic_add_relaxed(&abandoned_count, count); + mi_atomic_sub_relaxed(&abandoned_visited_count, count); + return true; +} + +// Push on the abandoned list. +static void mi_abandoned_push(mi_segment_t* segment) { + mi_assert_internal(segment->thread_id == 0); + mi_assert_internal(mi_atomic_load_ptr_relaxed(mi_segment_t, &segment->abandoned_next) == NULL); + mi_assert_internal(segment->next == NULL); + mi_assert_internal(segment->used > 0); + mi_tagged_segment_t next; + mi_tagged_segment_t ts = mi_atomic_load_relaxed(&abandoned); + do { + mi_atomic_store_ptr_release(mi_segment_t, &segment->abandoned_next, mi_tagged_segment_ptr(ts)); + next = mi_tagged_segment(segment, ts); + } while (!mi_atomic_cas_weak_release(&abandoned, &ts, next)); + mi_atomic_increment_relaxed(&abandoned_count); +} + +// Wait until there are no more pending reads on segments that used to be in the abandoned list +// called for example from `arena.c` before decommitting +void _mi_abandoned_await_readers(void) { + size_t n; + do { + n = mi_atomic_load_acquire(&abandoned_readers); + if (n != 0) mi_atomic_yield(); + } while (n != 0); +} + +// Pop from the abandoned list +static mi_segment_t* mi_abandoned_pop(void) { + mi_segment_t* segment; + // Check efficiently if it is empty (or if the visited list needs to be moved) + mi_tagged_segment_t ts = mi_atomic_load_relaxed(&abandoned); + segment = mi_tagged_segment_ptr(ts); + if mi_likely(segment == NULL) { + if mi_likely(!mi_abandoned_visited_revisit()) { // try to swap in the visited list on NULL + return NULL; + } + } + + // Do a pop. We use a reader count to prevent + // a segment to be decommitted while a read is still pending, + // and a tagged pointer to prevent A-B-A link corruption. + // (this is called from `region.c:_mi_mem_free` for example) + mi_atomic_increment_relaxed(&abandoned_readers); // ensure no segment gets decommitted + mi_tagged_segment_t next = 0; + ts = mi_atomic_load_acquire(&abandoned); + do { + segment = mi_tagged_segment_ptr(ts); + if (segment != NULL) { + mi_segment_t* anext = mi_atomic_load_ptr_relaxed(mi_segment_t, &segment->abandoned_next); + next = mi_tagged_segment(anext, ts); // note: reads the segment's `abandoned_next` field so should not be decommitted + } + } while (segment != NULL && !mi_atomic_cas_weak_acq_rel(&abandoned, &ts, next)); + mi_atomic_decrement_relaxed(&abandoned_readers); // release reader lock + if (segment != NULL) { + mi_atomic_store_ptr_release(mi_segment_t, &segment->abandoned_next, NULL); + mi_atomic_decrement_relaxed(&abandoned_count); + } + return segment; +} + +/* ----------------------------------------------------------- + Abandon segment/page +----------------------------------------------------------- */ + +static void mi_segment_abandon(mi_segment_t* segment, mi_segments_tld_t* tld) { + mi_assert_internal(segment->used == segment->abandoned); + mi_assert_internal(segment->used > 0); + mi_assert_internal(mi_atomic_load_ptr_relaxed(mi_segment_t, &segment->abandoned_next) == NULL); + mi_assert_internal(segment->abandoned_visits == 0); + mi_assert_expensive(mi_segment_is_valid(segment,tld)); + + // remove the free pages from the free page queues + mi_slice_t* slice = &segment->slices[0]; + const mi_slice_t* end = mi_segment_slices_end(segment); + while (slice < end) { + mi_assert_internal(slice->slice_count > 0); + mi_assert_internal(slice->slice_offset == 0); + if (slice->xblock_size == 0) { // a free page + mi_segment_span_remove_from_queue(slice,tld); + slice->xblock_size = 0; // but keep it free + } + slice = slice + slice->slice_count; + } + + // perform delayed decommits (forcing is much slower on mstress) + mi_segment_try_purge(segment, mi_option_is_enabled(mi_option_abandoned_page_purge) /* force? */, tld->stats); + + // all pages in the segment are abandoned; add it to the abandoned list + _mi_stat_increase(&tld->stats->segments_abandoned, 1); + mi_segments_track_size(-((long)mi_segment_size(segment)), tld); + segment->thread_id = 0; + mi_atomic_store_ptr_release(mi_segment_t, &segment->abandoned_next, NULL); + segment->abandoned_visits = 1; // from 0 to 1 to signify it is abandoned + mi_abandoned_push(segment); +} + +void _mi_segment_page_abandon(mi_page_t* page, mi_segments_tld_t* tld) { + mi_assert(page != NULL); + mi_assert_internal(mi_page_thread_free_flag(page)==MI_NEVER_DELAYED_FREE); + mi_assert_internal(mi_page_heap(page) == NULL); + mi_segment_t* segment = _mi_page_segment(page); + + mi_assert_expensive(mi_segment_is_valid(segment,tld)); + segment->abandoned++; + + _mi_stat_increase(&tld->stats->pages_abandoned, 1); + mi_assert_internal(segment->abandoned <= segment->used); + if (segment->used == segment->abandoned) { + // all pages are abandoned, abandon the entire segment + mi_segment_abandon(segment, tld); + } +} + +/* ----------------------------------------------------------- + Reclaim abandoned pages +----------------------------------------------------------- */ + +static mi_slice_t* mi_slices_start_iterate(mi_segment_t* segment, const mi_slice_t** end) { + mi_slice_t* slice = &segment->slices[0]; + *end = mi_segment_slices_end(segment); + mi_assert_internal(slice->slice_count>0 && slice->xblock_size>0); // segment allocated page + slice = slice + slice->slice_count; // skip the first segment allocated page + return slice; +} + +// Possibly free pages and check if free space is available +static bool mi_segment_check_free(mi_segment_t* segment, size_t slices_needed, size_t block_size, mi_segments_tld_t* tld) +{ + mi_assert_internal(block_size < MI_HUGE_BLOCK_SIZE); + mi_assert_internal(mi_segment_is_abandoned(segment)); + bool has_page = false; + + // for all slices + const mi_slice_t* end; + mi_slice_t* slice = mi_slices_start_iterate(segment, &end); + while (slice < end) { + mi_assert_internal(slice->slice_count > 0); + mi_assert_internal(slice->slice_offset == 0); + if (mi_slice_is_used(slice)) { // used page + // ensure used count is up to date and collect potential concurrent frees + mi_page_t* const page = mi_slice_to_page(slice); + _mi_page_free_collect(page, false); + if (mi_page_all_free(page)) { + // if this page is all free now, free it without adding to any queues (yet) + mi_assert_internal(page->next == NULL && page->prev==NULL); + _mi_stat_decrease(&tld->stats->pages_abandoned, 1); + segment->abandoned--; + slice = mi_segment_page_clear(page, tld); // re-assign slice due to coalesce! + mi_assert_internal(!mi_slice_is_used(slice)); + if (slice->slice_count >= slices_needed) { + has_page = true; + } + } + else { + if (page->xblock_size == block_size && mi_page_has_any_available(page)) { + // a page has available free blocks of the right size + has_page = true; + } + } + } + else { + // empty span + if (slice->slice_count >= slices_needed) { + has_page = true; + } + } + slice = slice + slice->slice_count; + } + return has_page; +} + +// Reclaim an abandoned segment; returns NULL if the segment was freed +// set `right_page_reclaimed` to `true` if it reclaimed a page of the right `block_size` that was not full. +static mi_segment_t* mi_segment_reclaim(mi_segment_t* segment, mi_heap_t* heap, size_t requested_block_size, bool* right_page_reclaimed, mi_segments_tld_t* tld) { + mi_assert_internal(mi_atomic_load_ptr_relaxed(mi_segment_t, &segment->abandoned_next) == NULL); + mi_assert_expensive(mi_segment_is_valid(segment, tld)); + if (right_page_reclaimed != NULL) { *right_page_reclaimed = false; } + + segment->thread_id = _mi_thread_id(); + segment->abandoned_visits = 0; + mi_segments_track_size((long)mi_segment_size(segment), tld); + mi_assert_internal(segment->next == NULL); + _mi_stat_decrease(&tld->stats->segments_abandoned, 1); + + // for all slices + const mi_slice_t* end; + mi_slice_t* slice = mi_slices_start_iterate(segment, &end); + while (slice < end) { + mi_assert_internal(slice->slice_count > 0); + mi_assert_internal(slice->slice_offset == 0); + if (mi_slice_is_used(slice)) { + // in use: reclaim the page in our heap + mi_page_t* page = mi_slice_to_page(slice); + mi_assert_internal(page->is_committed); + mi_assert_internal(mi_page_thread_free_flag(page)==MI_NEVER_DELAYED_FREE); + mi_assert_internal(mi_page_heap(page) == NULL); + mi_assert_internal(page->next == NULL && page->prev==NULL); + _mi_stat_decrease(&tld->stats->pages_abandoned, 1); + segment->abandoned--; + // set the heap again and allow delayed free again + mi_page_set_heap(page, heap); + _mi_page_use_delayed_free(page, MI_USE_DELAYED_FREE, true); // override never (after heap is set) + _mi_page_free_collect(page, false); // ensure used count is up to date + if (mi_page_all_free(page)) { + // if everything free by now, free the page + slice = mi_segment_page_clear(page, tld); // set slice again due to coalesceing + } + else { + // otherwise reclaim it into the heap + _mi_page_reclaim(heap, page); + if (requested_block_size == page->xblock_size && mi_page_has_any_available(page)) { + if (right_page_reclaimed != NULL) { *right_page_reclaimed = true; } + } + } + } + else { + // the span is free, add it to our page queues + slice = mi_segment_span_free_coalesce(slice, tld); // set slice again due to coalesceing + } + mi_assert_internal(slice->slice_count>0 && slice->slice_offset==0); + slice = slice + slice->slice_count; + } + + mi_assert(segment->abandoned == 0); + if (segment->used == 0) { // due to page_clear + mi_assert_internal(right_page_reclaimed == NULL || !(*right_page_reclaimed)); + mi_segment_free(segment, false, tld); + return NULL; + } + else { + return segment; + } +} + + +void _mi_abandoned_reclaim_all(mi_heap_t* heap, mi_segments_tld_t* tld) { + mi_segment_t* segment; + while ((segment = mi_abandoned_pop()) != NULL) { + mi_segment_reclaim(segment, heap, 0, NULL, tld); + } +} + +static mi_segment_t* mi_segment_try_reclaim(mi_heap_t* heap, size_t needed_slices, size_t block_size, bool* reclaimed, mi_segments_tld_t* tld) +{ + *reclaimed = false; + mi_segment_t* segment; + long max_tries = mi_option_get_clamp(mi_option_max_segment_reclaim, 8, 1024); // limit the work to bound allocation times + while ((max_tries-- > 0) && ((segment = mi_abandoned_pop()) != NULL)) { + segment->abandoned_visits++; + // todo: an arena exclusive heap will potentially visit many abandoned unsuitable segments + // and push them into the visited list and use many tries. Perhaps we can skip non-suitable ones in a better way? + bool is_suitable = _mi_heap_memid_is_suitable(heap, segment->memid); + bool has_page = mi_segment_check_free(segment,needed_slices,block_size,tld); // try to free up pages (due to concurrent frees) + if (segment->used == 0) { + // free the segment (by forced reclaim) to make it available to other threads. + // note1: we prefer to free a segment as that might lead to reclaiming another + // segment that is still partially used. + // note2: we could in principle optimize this by skipping reclaim and directly + // freeing but that would violate some invariants temporarily) + mi_segment_reclaim(segment, heap, 0, NULL, tld); + } + else if (has_page && is_suitable) { + // found a large enough free span, or a page of the right block_size with free space + // we return the result of reclaim (which is usually `segment`) as it might free + // the segment due to concurrent frees (in which case `NULL` is returned). + return mi_segment_reclaim(segment, heap, block_size, reclaimed, tld); + } + else if (segment->abandoned_visits > 3 && is_suitable) { + // always reclaim on 3rd visit to limit the abandoned queue length. + mi_segment_reclaim(segment, heap, 0, NULL, tld); + } + else { + // otherwise, push on the visited list so it gets not looked at too quickly again + mi_segment_try_purge(segment, true /* force? */, tld->stats); // force purge if needed as we may not visit soon again + mi_abandoned_visited_push(segment); + } + } + return NULL; +} + + +void _mi_abandoned_collect(mi_heap_t* heap, bool force, mi_segments_tld_t* tld) +{ + mi_segment_t* segment; + int max_tries = (force ? 16*1024 : 1024); // limit latency + if (force) { + mi_abandoned_visited_revisit(); + } + while ((max_tries-- > 0) && ((segment = mi_abandoned_pop()) != NULL)) { + mi_segment_check_free(segment,0,0,tld); // try to free up pages (due to concurrent frees) + if (segment->used == 0) { + // free the segment (by forced reclaim) to make it available to other threads. + // note: we could in principle optimize this by skipping reclaim and directly + // freeing but that would violate some invariants temporarily) + mi_segment_reclaim(segment, heap, 0, NULL, tld); + } + else { + // otherwise, purge if needed and push on the visited list + // note: forced purge can be expensive if many threads are destroyed/created as in mstress. + mi_segment_try_purge(segment, force, tld->stats); + mi_abandoned_visited_push(segment); + } + } +} + +/* ----------------------------------------------------------- + Reclaim or allocate +----------------------------------------------------------- */ + +static mi_segment_t* mi_segment_reclaim_or_alloc(mi_heap_t* heap, size_t needed_slices, size_t block_size, mi_segments_tld_t* tld, mi_os_tld_t* os_tld) +{ + mi_assert_internal(block_size < MI_HUGE_BLOCK_SIZE); + mi_assert_internal(block_size <= MI_LARGE_OBJ_SIZE_MAX); + + // 1. try to reclaim an abandoned segment + bool reclaimed; + mi_segment_t* segment = mi_segment_try_reclaim(heap, needed_slices, block_size, &reclaimed, tld); + if (reclaimed) { + // reclaimed the right page right into the heap + mi_assert_internal(segment != NULL); + return NULL; // pretend out-of-memory as the page will be in the page queue of the heap with available blocks + } + else if (segment != NULL) { + // reclaimed a segment with a large enough empty span in it + return segment; + } + // 2. otherwise allocate a fresh segment + return mi_segment_alloc(0, 0, heap->arena_id, tld, os_tld, NULL); +} + + +/* ----------------------------------------------------------- + Page allocation +----------------------------------------------------------- */ + +static mi_page_t* mi_segments_page_alloc(mi_heap_t* heap, mi_page_kind_t page_kind, size_t required, size_t block_size, mi_segments_tld_t* tld, mi_os_tld_t* os_tld) +{ + mi_assert_internal(required <= MI_LARGE_OBJ_SIZE_MAX && page_kind <= MI_PAGE_LARGE); + + // find a free page + size_t page_size = _mi_align_up(required, (required > MI_MEDIUM_PAGE_SIZE ? MI_MEDIUM_PAGE_SIZE : MI_SEGMENT_SLICE_SIZE)); + size_t slices_needed = page_size / MI_SEGMENT_SLICE_SIZE; + mi_assert_internal(slices_needed * MI_SEGMENT_SLICE_SIZE == page_size); + mi_page_t* page = mi_segments_page_find_and_allocate(slices_needed, heap->arena_id, tld); //(required <= MI_SMALL_SIZE_MAX ? 0 : slices_needed), tld); + if (page==NULL) { + // no free page, allocate a new segment and try again + if (mi_segment_reclaim_or_alloc(heap, slices_needed, block_size, tld, os_tld) == NULL) { + // OOM or reclaimed a good page in the heap + return NULL; + } + else { + // otherwise try again + return mi_segments_page_alloc(heap, page_kind, required, block_size, tld, os_tld); + } + } + mi_assert_internal(page != NULL && page->slice_count*MI_SEGMENT_SLICE_SIZE == page_size); + mi_assert_internal(_mi_ptr_segment(page)->thread_id == _mi_thread_id()); + mi_segment_try_purge(_mi_ptr_segment(page), false, tld->stats); + return page; +} + + + +/* ----------------------------------------------------------- + Huge page allocation +----------------------------------------------------------- */ + +static mi_page_t* mi_segment_huge_page_alloc(size_t size, size_t page_alignment, mi_arena_id_t req_arena_id, mi_segments_tld_t* tld, mi_os_tld_t* os_tld) +{ + mi_page_t* page = NULL; + mi_segment_t* segment = mi_segment_alloc(size,page_alignment,req_arena_id,tld,os_tld,&page); + if (segment == NULL || page==NULL) return NULL; + mi_assert_internal(segment->used==1); + mi_assert_internal(mi_page_block_size(page) >= size); + #if MI_HUGE_PAGE_ABANDON + segment->thread_id = 0; // huge segments are immediately abandoned + #endif + + // for huge pages we initialize the xblock_size as we may + // overallocate to accommodate large alignments. + size_t psize; + uint8_t* start = _mi_segment_page_start(segment, page, &psize); + page->xblock_size = (psize > MI_HUGE_BLOCK_SIZE ? MI_HUGE_BLOCK_SIZE : (uint32_t)psize); + + // decommit the part of the prefix of a page that will not be used; this can be quite large (close to MI_SEGMENT_SIZE) + if (page_alignment > 0 && segment->allow_decommit) { + uint8_t* aligned_p = (uint8_t*)_mi_align_up((uintptr_t)start, page_alignment); + mi_assert_internal(_mi_is_aligned(aligned_p, page_alignment)); + mi_assert_internal(psize - (aligned_p - start) >= size); + uint8_t* decommit_start = start + sizeof(mi_block_t); // for the free list + ptrdiff_t decommit_size = aligned_p - decommit_start; + _mi_os_reset(decommit_start, decommit_size, &_mi_stats_main); // note: cannot use segment_decommit on huge segments + } + + return page; +} + +#if MI_HUGE_PAGE_ABANDON +// free huge block from another thread +void _mi_segment_huge_page_free(mi_segment_t* segment, mi_page_t* page, mi_block_t* block) { + // huge page segments are always abandoned and can be freed immediately by any thread + mi_assert_internal(segment->kind==MI_SEGMENT_HUGE); + mi_assert_internal(segment == _mi_page_segment(page)); + mi_assert_internal(mi_atomic_load_relaxed(&segment->thread_id)==0); + + // claim it and free + mi_heap_t* heap = mi_heap_get_default(); // issue #221; don't use the internal get_default_heap as we need to ensure the thread is initialized. + // paranoia: if this it the last reference, the cas should always succeed + size_t expected_tid = 0; + if (mi_atomic_cas_strong_acq_rel(&segment->thread_id, &expected_tid, heap->thread_id)) { + mi_block_set_next(page, block, page->free); + page->free = block; + page->used--; + page->is_zero = false; + mi_assert(page->used == 0); + mi_tld_t* tld = heap->tld; + _mi_segment_page_free(page, true, &tld->segments); + } +#if (MI_DEBUG!=0) + else { + mi_assert_internal(false); + } +#endif +} + +#else +// reset memory of a huge block from another thread +void _mi_segment_huge_page_reset(mi_segment_t* segment, mi_page_t* page, mi_block_t* block) { + MI_UNUSED(page); + mi_assert_internal(segment->kind == MI_SEGMENT_HUGE); + mi_assert_internal(segment == _mi_page_segment(page)); + mi_assert_internal(page->used == 1); // this is called just before the free + mi_assert_internal(page->free == NULL); + if (segment->allow_decommit) { + size_t csize = mi_usable_size(block); + if (csize > sizeof(mi_block_t)) { + csize = csize - sizeof(mi_block_t); + uint8_t* p = (uint8_t*)block + sizeof(mi_block_t); + _mi_os_reset(p, csize, &_mi_stats_main); // note: cannot use segment_decommit on huge segments + } + } +} +#endif + +/* ----------------------------------------------------------- + Page allocation and free +----------------------------------------------------------- */ +mi_page_t* _mi_segment_page_alloc(mi_heap_t* heap, size_t block_size, size_t page_alignment, mi_segments_tld_t* tld, mi_os_tld_t* os_tld) { + mi_page_t* page; + if mi_unlikely(page_alignment > MI_ALIGNMENT_MAX) { + mi_assert_internal(_mi_is_power_of_two(page_alignment)); + mi_assert_internal(page_alignment >= MI_SEGMENT_SIZE); + if (page_alignment < MI_SEGMENT_SIZE) { page_alignment = MI_SEGMENT_SIZE; } + page = mi_segment_huge_page_alloc(block_size,page_alignment,heap->arena_id,tld,os_tld); + } + else if (block_size <= MI_SMALL_OBJ_SIZE_MAX) { + page = mi_segments_page_alloc(heap,MI_PAGE_SMALL,block_size,block_size,tld,os_tld); + } + else if (block_size <= MI_MEDIUM_OBJ_SIZE_MAX) { + page = mi_segments_page_alloc(heap,MI_PAGE_MEDIUM,MI_MEDIUM_PAGE_SIZE,block_size,tld, os_tld); + } + else if (block_size <= MI_LARGE_OBJ_SIZE_MAX) { + page = mi_segments_page_alloc(heap,MI_PAGE_LARGE,block_size,block_size,tld, os_tld); + } + else { + page = mi_segment_huge_page_alloc(block_size,page_alignment,heap->arena_id,tld,os_tld); + } + mi_assert_internal(page == NULL || _mi_heap_memid_is_suitable(heap, _mi_page_segment(page)->memid)); + mi_assert_expensive(page == NULL || mi_segment_is_valid(_mi_page_segment(page),tld)); + return page; +} diff --git a/compat/mimalloc/stats.c b/compat/mimalloc/stats.c new file mode 100644 index 00000000000000..6817e07aa1ee9f --- /dev/null +++ b/compat/mimalloc/stats.c @@ -0,0 +1,467 @@ +/* ---------------------------------------------------------------------------- +Copyright (c) 2018-2021, Microsoft Research, Daan Leijen +This is free software; you can redistribute it and/or modify it under the +terms of the MIT license. A copy of the license can be found in the file +"LICENSE" at the root of this distribution. +-----------------------------------------------------------------------------*/ +#include "mimalloc.h" +#include "mimalloc/internal.h" +#include "mimalloc/atomic.h" +#include "mimalloc/prim.h" + +#include // snprintf +#include // memset + +#if defined(_MSC_VER) && (_MSC_VER < 1920) +#pragma warning(disable:4204) // non-constant aggregate initializer +#endif + +/* ----------------------------------------------------------- + Statistics operations +----------------------------------------------------------- */ + +static bool mi_is_in_main(void* stat) { + return ((uint8_t*)stat >= (uint8_t*)&_mi_stats_main + && (uint8_t*)stat < ((uint8_t*)&_mi_stats_main + sizeof(mi_stats_t))); +} + +static void mi_stat_update(mi_stat_count_t* stat, int64_t amount) { + if (amount == 0) return; + if (mi_is_in_main(stat)) + { + // add atomically (for abandoned pages) + int64_t current = mi_atomic_addi64_relaxed(&stat->current, amount); + mi_atomic_maxi64_relaxed(&stat->peak, current + amount); + if (amount > 0) { + mi_atomic_addi64_relaxed(&stat->allocated,amount); + } + else { + mi_atomic_addi64_relaxed(&stat->freed, -amount); + } + } + else { + // add thread local + stat->current += amount; + if (stat->current > stat->peak) stat->peak = stat->current; + if (amount > 0) { + stat->allocated += amount; + } + else { + stat->freed += -amount; + } + } +} + +void _mi_stat_counter_increase(mi_stat_counter_t* stat, size_t amount) { + if (mi_is_in_main(stat)) { + mi_atomic_addi64_relaxed( &stat->count, 1 ); + mi_atomic_addi64_relaxed( &stat->total, (int64_t)amount ); + } + else { + stat->count++; + stat->total += amount; + } +} + +void _mi_stat_increase(mi_stat_count_t* stat, size_t amount) { + mi_stat_update(stat, (int64_t)amount); +} + +void _mi_stat_decrease(mi_stat_count_t* stat, size_t amount) { + mi_stat_update(stat, -((int64_t)amount)); +} + +// must be thread safe as it is called from stats_merge +static void mi_stat_add(mi_stat_count_t* stat, const mi_stat_count_t* src, int64_t unit) { + if (stat==src) return; + if (src->allocated==0 && src->freed==0) return; + mi_atomic_addi64_relaxed( &stat->allocated, src->allocated * unit); + mi_atomic_addi64_relaxed( &stat->current, src->current * unit); + mi_atomic_addi64_relaxed( &stat->freed, src->freed * unit); + // peak scores do not work across threads.. + mi_atomic_addi64_relaxed( &stat->peak, src->peak * unit); +} + +static void mi_stat_counter_add(mi_stat_counter_t* stat, const mi_stat_counter_t* src, int64_t unit) { + if (stat==src) return; + mi_atomic_addi64_relaxed( &stat->total, src->total * unit); + mi_atomic_addi64_relaxed( &stat->count, src->count * unit); +} + +// must be thread safe as it is called from stats_merge +static void mi_stats_add(mi_stats_t* stats, const mi_stats_t* src) { + if (stats==src) return; + mi_stat_add(&stats->segments, &src->segments,1); + mi_stat_add(&stats->pages, &src->pages,1); + mi_stat_add(&stats->reserved, &src->reserved, 1); + mi_stat_add(&stats->committed, &src->committed, 1); + mi_stat_add(&stats->reset, &src->reset, 1); + mi_stat_add(&stats->purged, &src->purged, 1); + mi_stat_add(&stats->page_committed, &src->page_committed, 1); + + mi_stat_add(&stats->pages_abandoned, &src->pages_abandoned, 1); + mi_stat_add(&stats->segments_abandoned, &src->segments_abandoned, 1); + mi_stat_add(&stats->threads, &src->threads, 1); + + mi_stat_add(&stats->malloc, &src->malloc, 1); + mi_stat_add(&stats->segments_cache, &src->segments_cache, 1); + mi_stat_add(&stats->normal, &src->normal, 1); + mi_stat_add(&stats->huge, &src->huge, 1); + mi_stat_add(&stats->large, &src->large, 1); + + mi_stat_counter_add(&stats->pages_extended, &src->pages_extended, 1); + mi_stat_counter_add(&stats->mmap_calls, &src->mmap_calls, 1); + mi_stat_counter_add(&stats->commit_calls, &src->commit_calls, 1); + mi_stat_counter_add(&stats->reset_calls, &src->reset_calls, 1); + mi_stat_counter_add(&stats->purge_calls, &src->purge_calls, 1); + + mi_stat_counter_add(&stats->page_no_retire, &src->page_no_retire, 1); + mi_stat_counter_add(&stats->searches, &src->searches, 1); + mi_stat_counter_add(&stats->normal_count, &src->normal_count, 1); + mi_stat_counter_add(&stats->huge_count, &src->huge_count, 1); + mi_stat_counter_add(&stats->large_count, &src->large_count, 1); +#if MI_STAT>1 + for (size_t i = 0; i <= MI_BIN_HUGE; i++) { + if (src->normal_bins[i].allocated > 0 || src->normal_bins[i].freed > 0) { + mi_stat_add(&stats->normal_bins[i], &src->normal_bins[i], 1); + } + } +#endif +} + +/* ----------------------------------------------------------- + Display statistics +----------------------------------------------------------- */ + +// unit > 0 : size in binary bytes +// unit == 0: count as decimal +// unit < 0 : count in binary +static void mi_printf_amount(int64_t n, int64_t unit, mi_output_fun* out, void* arg, const char* fmt) { + char buf[32]; buf[0] = 0; + int len = 32; + const char* suffix = (unit <= 0 ? " " : "B"); + const int64_t base = (unit == 0 ? 1000 : 1024); + if (unit>0) n *= unit; + + const int64_t pos = (n < 0 ? -n : n); + if (pos < base) { + if (n!=1 || suffix[0] != 'B') { // skip printing 1 B for the unit column + snprintf(buf, len, "%d %-3s", (int)n, (n==0 ? "" : suffix)); + } + } + else { + int64_t divider = base; + const char* magnitude = "K"; + if (pos >= divider*base) { divider *= base; magnitude = "M"; } + if (pos >= divider*base) { divider *= base; magnitude = "G"; } + const int64_t tens = (n / (divider/10)); + const long whole = (long)(tens/10); + const long frac1 = (long)(tens%10); + char unitdesc[8]; + snprintf(unitdesc, 8, "%s%s%s", magnitude, (base==1024 ? "i" : ""), suffix); + snprintf(buf, len, "%ld.%ld %-3s", whole, (frac1 < 0 ? -frac1 : frac1), unitdesc); + } + _mi_fprintf(out, arg, (fmt==NULL ? "%12s" : fmt), buf); +} + + +static void mi_print_amount(int64_t n, int64_t unit, mi_output_fun* out, void* arg) { + mi_printf_amount(n,unit,out,arg,NULL); +} + +static void mi_print_count(int64_t n, int64_t unit, mi_output_fun* out, void* arg) { + if (unit==1) _mi_fprintf(out, arg, "%12s"," "); + else mi_print_amount(n,0,out,arg); +} + +static void mi_stat_print_ex(const mi_stat_count_t* stat, const char* msg, int64_t unit, mi_output_fun* out, void* arg, const char* notok ) { + _mi_fprintf(out, arg,"%10s:", msg); + if (unit > 0) { + mi_print_amount(stat->peak, unit, out, arg); + mi_print_amount(stat->allocated, unit, out, arg); + mi_print_amount(stat->freed, unit, out, arg); + mi_print_amount(stat->current, unit, out, arg); + mi_print_amount(unit, 1, out, arg); + mi_print_count(stat->allocated, unit, out, arg); + if (stat->allocated > stat->freed) { + _mi_fprintf(out, arg, " "); + _mi_fprintf(out, arg, (notok == NULL ? "not all freed" : notok)); + _mi_fprintf(out, arg, "\n"); + } + else { + _mi_fprintf(out, arg, " ok\n"); + } + } + else if (unit<0) { + mi_print_amount(stat->peak, -1, out, arg); + mi_print_amount(stat->allocated, -1, out, arg); + mi_print_amount(stat->freed, -1, out, arg); + mi_print_amount(stat->current, -1, out, arg); + if (unit==-1) { + _mi_fprintf(out, arg, "%24s", ""); + } + else { + mi_print_amount(-unit, 1, out, arg); + mi_print_count((stat->allocated / -unit), 0, out, arg); + } + if (stat->allocated > stat->freed) + _mi_fprintf(out, arg, " not all freed!\n"); + else + _mi_fprintf(out, arg, " ok\n"); + } + else { + mi_print_amount(stat->peak, 1, out, arg); + mi_print_amount(stat->allocated, 1, out, arg); + _mi_fprintf(out, arg, "%11s", " "); // no freed + mi_print_amount(stat->current, 1, out, arg); + _mi_fprintf(out, arg, "\n"); + } +} + +static void mi_stat_print(const mi_stat_count_t* stat, const char* msg, int64_t unit, mi_output_fun* out, void* arg) { + mi_stat_print_ex(stat, msg, unit, out, arg, NULL); +} + +static void mi_stat_peak_print(const mi_stat_count_t* stat, const char* msg, int64_t unit, mi_output_fun* out, void* arg) { + _mi_fprintf(out, arg, "%10s:", msg); + mi_print_amount(stat->peak, unit, out, arg); + _mi_fprintf(out, arg, "\n"); +} + +static void mi_stat_counter_print(const mi_stat_counter_t* stat, const char* msg, mi_output_fun* out, void* arg ) { + _mi_fprintf(out, arg, "%10s:", msg); + mi_print_amount(stat->total, -1, out, arg); + _mi_fprintf(out, arg, "\n"); +} + + +static void mi_stat_counter_print_avg(const mi_stat_counter_t* stat, const char* msg, mi_output_fun* out, void* arg) { + const int64_t avg_tens = (stat->count == 0 ? 0 : (stat->total*10 / stat->count)); + const long avg_whole = (long)(avg_tens/10); + const long avg_frac1 = (long)(avg_tens%10); + _mi_fprintf(out, arg, "%10s: %5ld.%ld avg\n", msg, avg_whole, avg_frac1); +} + + +static void mi_print_header(mi_output_fun* out, void* arg ) { + _mi_fprintf(out, arg, "%10s: %11s %11s %11s %11s %11s %11s\n", "heap stats", "peak ", "total ", "freed ", "current ", "unit ", "count "); +} + +#if MI_STAT>1 +static void mi_stats_print_bins(const mi_stat_count_t* bins, size_t max, const char* fmt, mi_output_fun* out, void* arg) { + bool found = false; + char buf[64]; + for (size_t i = 0; i <= max; i++) { + if (bins[i].allocated > 0) { + found = true; + int64_t unit = _mi_bin_size((uint8_t)i); + snprintf(buf, 64, "%s %3lu", fmt, (long)i); + mi_stat_print(&bins[i], buf, unit, out, arg); + } + } + if (found) { + _mi_fprintf(out, arg, "\n"); + mi_print_header(out, arg); + } +} +#endif + + + +//------------------------------------------------------------ +// Use an output wrapper for line-buffered output +// (which is nice when using loggers etc.) +//------------------------------------------------------------ +typedef struct buffered_s { + mi_output_fun* out; // original output function + void* arg; // and state + char* buf; // local buffer of at least size `count+1` + size_t used; // currently used chars `used <= count` + size_t count; // total chars available for output +} buffered_t; + +static void mi_buffered_flush(buffered_t* buf) { + buf->buf[buf->used] = 0; + _mi_fputs(buf->out, buf->arg, NULL, buf->buf); + buf->used = 0; +} + +static void mi_cdecl mi_buffered_out(const char* msg, void* arg) { + buffered_t* buf = (buffered_t*)arg; + if (msg==NULL || buf==NULL) return; + for (const char* src = msg; *src != 0; src++) { + char c = *src; + if (buf->used >= buf->count) mi_buffered_flush(buf); + mi_assert_internal(buf->used < buf->count); + buf->buf[buf->used++] = c; + if (c == '\n') mi_buffered_flush(buf); + } +} + +//------------------------------------------------------------ +// Print statistics +//------------------------------------------------------------ + +static void _mi_stats_print(mi_stats_t* stats, mi_output_fun* out0, void* arg0) mi_attr_noexcept { + // wrap the output function to be line buffered + char buf[256]; + buffered_t buffer = { out0, arg0, NULL, 0, 255 }; + buffer.buf = buf; + mi_output_fun* out = &mi_buffered_out; + void* arg = &buffer; + + // and print using that + mi_print_header(out,arg); + #if MI_STAT>1 + mi_stats_print_bins(stats->normal_bins, MI_BIN_HUGE, "normal",out,arg); + #endif + #if MI_STAT + mi_stat_print(&stats->normal, "normal", (stats->normal_count.count == 0 ? 1 : -(stats->normal.allocated / stats->normal_count.count)), out, arg); + mi_stat_print(&stats->large, "large", (stats->large_count.count == 0 ? 1 : -(stats->large.allocated / stats->large_count.count)), out, arg); + mi_stat_print(&stats->huge, "huge", (stats->huge_count.count == 0 ? 1 : -(stats->huge.allocated / stats->huge_count.count)), out, arg); + mi_stat_count_t total = { 0,0,0,0 }; + mi_stat_add(&total, &stats->normal, 1); + mi_stat_add(&total, &stats->large, 1); + mi_stat_add(&total, &stats->huge, 1); + mi_stat_print(&total, "total", 1, out, arg); + #endif + #if MI_STAT>1 + mi_stat_print(&stats->malloc, "malloc req", 1, out, arg); + _mi_fprintf(out, arg, "\n"); + #endif + mi_stat_print_ex(&stats->reserved, "reserved", 1, out, arg, ""); + mi_stat_print_ex(&stats->committed, "committed", 1, out, arg, ""); + mi_stat_peak_print(&stats->reset, "reset", 1, out, arg ); + mi_stat_peak_print(&stats->purged, "purged", 1, out, arg ); + mi_stat_print(&stats->page_committed, "touched", 1, out, arg); + mi_stat_print(&stats->segments, "segments", -1, out, arg); + mi_stat_print(&stats->segments_abandoned, "-abandoned", -1, out, arg); + mi_stat_print(&stats->segments_cache, "-cached", -1, out, arg); + mi_stat_print(&stats->pages, "pages", -1, out, arg); + mi_stat_print(&stats->pages_abandoned, "-abandoned", -1, out, arg); + mi_stat_counter_print(&stats->pages_extended, "-extended", out, arg); + mi_stat_counter_print(&stats->page_no_retire, "-noretire", out, arg); + mi_stat_counter_print(&stats->mmap_calls, "mmaps", out, arg); + mi_stat_counter_print(&stats->commit_calls, "commits", out, arg); + mi_stat_counter_print(&stats->reset_calls, "resets", out, arg); + mi_stat_counter_print(&stats->purge_calls, "purges", out, arg); + mi_stat_print(&stats->threads, "threads", -1, out, arg); + mi_stat_counter_print_avg(&stats->searches, "searches", out, arg); + _mi_fprintf(out, arg, "%10s: %5zu\n", "numa nodes", _mi_os_numa_node_count()); + + size_t elapsed; + size_t user_time; + size_t sys_time; + size_t current_rss; + size_t peak_rss; + size_t current_commit; + size_t peak_commit; + size_t page_faults; + mi_process_info(&elapsed, &user_time, &sys_time, ¤t_rss, &peak_rss, ¤t_commit, &peak_commit, &page_faults); + _mi_fprintf(out, arg, "%10s: %5ld.%03ld s\n", "elapsed", elapsed/1000, elapsed%1000); + _mi_fprintf(out, arg, "%10s: user: %ld.%03ld s, system: %ld.%03ld s, faults: %lu, rss: ", "process", + user_time/1000, user_time%1000, sys_time/1000, sys_time%1000, (unsigned long)page_faults ); + mi_printf_amount((int64_t)peak_rss, 1, out, arg, "%s"); + if (peak_commit > 0) { + _mi_fprintf(out, arg, ", commit: "); + mi_printf_amount((int64_t)peak_commit, 1, out, arg, "%s"); + } + _mi_fprintf(out, arg, "\n"); +} + +static mi_msecs_t mi_process_start; // = 0 + +static mi_stats_t* mi_stats_get_default(void) { + mi_heap_t* heap = mi_heap_get_default(); + return &heap->tld->stats; +} + +static void mi_stats_merge_from(mi_stats_t* stats) { + if (stats != &_mi_stats_main) { + mi_stats_add(&_mi_stats_main, stats); + memset(stats, 0, sizeof(mi_stats_t)); + } +} + +void mi_stats_reset(void) mi_attr_noexcept { + mi_stats_t* stats = mi_stats_get_default(); + if (stats != &_mi_stats_main) { memset(stats, 0, sizeof(mi_stats_t)); } + memset(&_mi_stats_main, 0, sizeof(mi_stats_t)); + if (mi_process_start == 0) { mi_process_start = _mi_clock_start(); }; +} + +void mi_stats_merge(void) mi_attr_noexcept { + mi_stats_merge_from( mi_stats_get_default() ); +} + +void _mi_stats_done(mi_stats_t* stats) { // called from `mi_thread_done` + mi_stats_merge_from(stats); +} + +void mi_stats_print_out(mi_output_fun* out, void* arg) mi_attr_noexcept { + mi_stats_merge_from(mi_stats_get_default()); + _mi_stats_print(&_mi_stats_main, out, arg); +} + +void mi_stats_print(void* out) mi_attr_noexcept { + // for compatibility there is an `out` parameter (which can be `stdout` or `stderr`) + mi_stats_print_out((mi_output_fun*)out, NULL); +} + +void mi_thread_stats_print_out(mi_output_fun* out, void* arg) mi_attr_noexcept { + _mi_stats_print(mi_stats_get_default(), out, arg); +} + + +// ---------------------------------------------------------------- +// Basic timer for convenience; use milli-seconds to avoid doubles +// ---------------------------------------------------------------- + +static mi_msecs_t mi_clock_diff; + +mi_msecs_t _mi_clock_now(void) { + return _mi_prim_clock_now(); +} + +mi_msecs_t _mi_clock_start(void) { + if (mi_clock_diff == 0.0) { + mi_msecs_t t0 = _mi_clock_now(); + mi_clock_diff = _mi_clock_now() - t0; + } + return _mi_clock_now(); +} + +mi_msecs_t _mi_clock_end(mi_msecs_t start) { + mi_msecs_t end = _mi_clock_now(); + return (end - start - mi_clock_diff); +} + + +// -------------------------------------------------------- +// Basic process statistics +// -------------------------------------------------------- + +mi_decl_export void mi_process_info(size_t* elapsed_msecs, size_t* user_msecs, size_t* system_msecs, size_t* current_rss, size_t* peak_rss, size_t* current_commit, size_t* peak_commit, size_t* page_faults) mi_attr_noexcept +{ + mi_process_info_t pinfo; + _mi_memzero_var(pinfo); + pinfo.elapsed = _mi_clock_end(mi_process_start); + pinfo.current_commit = (size_t)(mi_atomic_loadi64_relaxed((_Atomic(int64_t)*)&_mi_stats_main.committed.current)); + pinfo.peak_commit = (size_t)(mi_atomic_loadi64_relaxed((_Atomic(int64_t)*)&_mi_stats_main.committed.peak)); + pinfo.current_rss = pinfo.current_commit; + pinfo.peak_rss = pinfo.peak_commit; + pinfo.utime = 0; + pinfo.stime = 0; + pinfo.page_faults = 0; + + _mi_prim_process_info(&pinfo); + + if (elapsed_msecs!=NULL) *elapsed_msecs = (pinfo.elapsed < 0 ? 0 : (pinfo.elapsed < (mi_msecs_t)PTRDIFF_MAX ? (size_t)pinfo.elapsed : PTRDIFF_MAX)); + if (user_msecs!=NULL) *user_msecs = (pinfo.utime < 0 ? 0 : (pinfo.utime < (mi_msecs_t)PTRDIFF_MAX ? (size_t)pinfo.utime : PTRDIFF_MAX)); + if (system_msecs!=NULL) *system_msecs = (pinfo.stime < 0 ? 0 : (pinfo.stime < (mi_msecs_t)PTRDIFF_MAX ? (size_t)pinfo.stime : PTRDIFF_MAX)); + if (current_rss!=NULL) *current_rss = pinfo.current_rss; + if (peak_rss!=NULL) *peak_rss = pinfo.peak_rss; + if (current_commit!=NULL) *current_commit = pinfo.current_commit; + if (peak_commit!=NULL) *peak_commit = pinfo.peak_commit; + if (page_faults!=NULL) *page_faults = pinfo.page_faults; +} From 45a2fd84abb152bb21f85a92e3cdee6af23f5541 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Mon, 24 Jun 2019 23:41:27 +0200 Subject: [PATCH 091/312] mimalloc: adjust for building inside Git We want to compile mimalloc's source code as part of Git, rather than requiring the code to be built as an external library: mimalloc uses a CMake-based build, which is not necessarily easy to integrate into the flavors of Git for Windows (which will be the main benefitting port). Signed-off-by: Johannes Schindelin --- compat/mimalloc/alloc.c | 4 ---- compat/mimalloc/mimalloc.h | 3 ++- 2 files changed, 2 insertions(+), 5 deletions(-) diff --git a/compat/mimalloc/alloc.c b/compat/mimalloc/alloc.c index 961f6d53d0f2c7..ae272c1fb54504 100644 --- a/compat/mimalloc/alloc.c +++ b/compat/mimalloc/alloc.c @@ -16,10 +16,6 @@ terms of the MIT license. A copy of the license can be found in the file #include // memset, strlen (for mi_strdup) #include // malloc, abort -#define MI_IN_ALLOC_C -#include "alloc-override.c" -#undef MI_IN_ALLOC_C - // ------------------------------------------------------ // Allocation // ------------------------------------------------------ diff --git a/compat/mimalloc/mimalloc.h b/compat/mimalloc/mimalloc.h index c0f5e96e51e975..7e3b5dd66e91a0 100644 --- a/compat/mimalloc/mimalloc.h +++ b/compat/mimalloc/mimalloc.h @@ -95,7 +95,8 @@ terms of the MIT license. A copy of the license can be found in the file // Includes // ------------------------------------------------------ -#include // size_t +#include "git-compat-util.h" + #include // bool #include // INTPTR_MAX From b935f58ec4b8b490a0acb3221efd8494841c4c3d Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Mon, 24 Jun 2019 23:43:06 +0200 Subject: [PATCH 092/312] mimalloc: offer a build-time option to enable it By defining `USE_MIMALLOC`, Git can now be compiled with that nicely-fast and small allocator. Note that we have to disable a couple `DEVELOPER` options to build mimalloc's source code, as it makes heavy use of declarations after statements, among other things that disagree with Git's conventions. We even have to silence some GCC warnings in non-DEVELOPER mode. For example, the `-Wno-array-bounds` flag is needed because in `-O2` builds, trying to call `NtCurrentTeb()` (which `_mi_thread_id()` does on Windows) causes the bogus warning about a system header, likely related to https://sourceforge.net/p/mingw-w64/mailman/message/37674519/ and to https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99578: C:/git-sdk-64-minimal/mingw64/include/psdk_inc/intrin-impl.h:838:1: error: array subscript 0 is outside array bounds of 'long long unsigned int[0]' [-Werror=array-bounds] 838 | __buildreadseg(__readgsqword, unsigned __int64, "gs", "q") | ^~~~~~~~~~~~~~ Also: The `mimalloc` library uses C11-style atomics, therefore we must require that standard when compiling with GCC if we want to use `mimalloc` (instead of requiring "only" C99). This is what we do in the CMake definition already, therefore this commit does not need to touch `contrib/buildsystems/`. Signed-off-by: Johannes Schindelin --- Makefile | 37 +++++++++++++++++++++++++++++++++++++ config.mak.dev | 2 ++ config.mak.uname | 2 +- git-compat-util.h | 10 ++++++++++ 4 files changed, 50 insertions(+), 1 deletion(-) diff --git a/Makefile b/Makefile index fbb9227e7d7909..ec5951d5c8a057 100644 --- a/Makefile +++ b/Makefile @@ -2147,6 +2147,43 @@ ifdef USE_NED_ALLOCATOR OVERRIDE_STRDUP = YesPlease endif +ifdef USE_MIMALLOC + MIMALLOC_OBJS = \ + compat/mimalloc/alloc-aligned.o \ + compat/mimalloc/alloc.o \ + compat/mimalloc/arena.o \ + compat/mimalloc/bitmap.o \ + compat/mimalloc/heap.o \ + compat/mimalloc/init.o \ + compat/mimalloc/options.o \ + compat/mimalloc/os.o \ + compat/mimalloc/page.o \ + compat/mimalloc/random.o \ + compat/mimalloc/prim/windows/prim.o \ + compat/mimalloc/segment.o \ + compat/mimalloc/segment-cache.o \ + compat/mimalloc/segment-map.o \ + compat/mimalloc/stats.o + + COMPAT_CFLAGS += -Icompat/mimalloc -DMI_DEBUG=0 -DUSE_MIMALLOC --std=gnu11 + COMPAT_OBJS += $(MIMALLOC_OBJS) + +$(MIMALLOC_OBJS): COMPAT_CFLAGS += -DBANNED_H + +$(MIMALLOC_OBJS): COMPAT_CFLAGS += \ + -Wno-attributes \ + -Wno-unknown-pragmas \ + -Wno-array-bounds + +ifdef DEVELOPER +$(MIMALLOC_OBJS): COMPAT_CFLAGS += \ + -Wno-pedantic \ + -Wno-declaration-after-statement \ + -Wno-old-style-definition \ + -Wno-missing-prototypes +endif +endif + ifdef OVERRIDE_STRDUP COMPAT_CFLAGS += -DOVERRIDE_STRDUP COMPAT_OBJS += compat/strdup.o diff --git a/config.mak.dev b/config.mak.dev index 0fd8cc4d355ebb..a5d6dd67dfd77e 100644 --- a/config.mak.dev +++ b/config.mak.dev @@ -22,8 +22,10 @@ endif ifneq ($(uname_S),FreeBSD) ifneq ($(or $(filter gcc6,$(COMPILER_FEATURES)),$(filter clang7,$(COMPILER_FEATURES))),) +ifndef USE_MIMALLOC DEVELOPER_CFLAGS += -std=gnu99 endif +endif else # FreeBSD cannot limit to C99 because its system headers unconditionally # rely on C11 features. diff --git a/config.mak.uname b/config.mak.uname index 7fba3e595c5cd5..cdfb606f010ba4 100644 --- a/config.mak.uname +++ b/config.mak.uname @@ -494,7 +494,7 @@ endif CC = compat/vcbuild/scripts/clink.pl AR = compat/vcbuild/scripts/lib.pl CFLAGS = - BASIC_CFLAGS = -nologo -I. -Icompat/vcbuild/include -DWIN32 -D_CONSOLE -DHAVE_STRING_H -D_CRT_SECURE_NO_WARNINGS -D_CRT_NONSTDC_NO_DEPRECATE + BASIC_CFLAGS = -nologo -I. -Icompat/vcbuild/include -DWIN32 -D_CONSOLE -DHAVE_STRING_H -D_CRT_SECURE_NO_WARNINGS -D_CRT_NONSTDC_NO_DEPRECATE -MP -std:c11 COMPAT_OBJS = compat/msvc.o compat/winansi.o \ compat/win32/flush.o \ compat/win32/path-utils.o \ diff --git a/git-compat-util.h b/git-compat-util.h index 6aee7e1b27a26b..0505966dd8b02a 100644 --- a/git-compat-util.h +++ b/git-compat-util.h @@ -443,6 +443,16 @@ char *gitdirname(char *); # include #endif +#ifdef USE_MIMALLOC +#include "mimalloc.h" +#define malloc mi_malloc +#define calloc mi_calloc +#define realloc mi_realloc +#define free mi_free +#define strdup mi_strdup +#define strndup mi_strndup +#endif + /* On most systems would have given us this, but * not on some systems (e.g. z/OS). */ From c5761ccfcfd1de8bd1f28a5724b6881fa4f7f339 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Fri, 7 Dec 2018 13:39:30 +0100 Subject: [PATCH 093/312] clean: do not traverse mount points It seems to be not exactly rare on Windows to install NTFS junction points (the equivalent of "bind mounts" on Linux/Unix) in worktrees, e.g. to map some development tools into a subdirectory. In such a scenario, it is pretty horrible if `git clean -dfx` traverses into the mapped directory and starts to "clean up". Let's just not do that. Let's make sure before we traverse into a directory that it is not a mount point (or junction). This addresses https://github.com/git-for-windows/git/issues/607 Signed-off-by: Johannes Schindelin --- builtin/clean.c | 14 ++++++++++++++ compat/mingw.c | 22 ++++++++++++++++++++++ compat/mingw.h | 3 +++ git-compat-util.h | 4 ++++ path.c | 39 +++++++++++++++++++++++++++++++++++++++ path.h | 1 + t/t7300-clean.sh | 9 +++++++++ 7 files changed, 92 insertions(+) diff --git a/builtin/clean.c b/builtin/clean.c index 053c94fc6bd12a..aeb7eb2d1ed2c3 100644 --- a/builtin/clean.c +++ b/builtin/clean.c @@ -40,6 +40,8 @@ static const char *msg_remove = N_("Removing %s\n"); static const char *msg_would_remove = N_("Would remove %s\n"); static const char *msg_skip_git_dir = N_("Skipping repository %s\n"); static const char *msg_would_skip_git_dir = N_("Would skip repository %s\n"); +static const char *msg_skip_mount_point = N_("Skipping mount point %s\n"); +static const char *msg_would_skip_mount_point = N_("Would skip mount point %s\n"); static const char *msg_warn_remove_failed = N_("failed to remove %s"); static const char *msg_warn_lstat_failed = N_("could not lstat %s\n"); static const char *msg_skip_cwd = N_("Refusing to remove current working directory\n"); @@ -184,6 +186,18 @@ static int remove_dirs(struct strbuf *path, const char *prefix, int force_flag, goto out; } + if (is_mount_point(path)) { + if (!quiet) { + quote_path(path->buf, prefix, "ed, 0); + printf(dry_run ? + _(msg_would_skip_mount_point) : + _(msg_skip_mount_point), quoted.buf); + } + *dir_gone = 0; + + goto out; + } + dir = opendir(path->buf); if (!dir) { /* an empty dir could be removed even if it is unreadble */ diff --git a/compat/mingw.c b/compat/mingw.c index f524c54d06d965..5824a9162a6d19 100644 --- a/compat/mingw.c +++ b/compat/mingw.c @@ -2688,6 +2688,28 @@ pid_t waitpid(pid_t pid, int *status, int options) return -1; } +int mingw_is_mount_point(struct strbuf *path) +{ + WIN32_FIND_DATAW findbuf = { 0 }; + HANDLE handle; + wchar_t wfilename[MAX_PATH]; + int wlen = xutftowcs_path(wfilename, path->buf); + if (wlen < 0) + die(_("could not get long path for '%s'"), path->buf); + + /* remove trailing slash, if any */ + if (wlen > 0 && wfilename[wlen - 1] == L'/') + wfilename[--wlen] = L'\0'; + + handle = FindFirstFileW(wfilename, &findbuf); + if (handle == INVALID_HANDLE_VALUE) + return 0; + FindClose(handle); + + return (findbuf.dwFileAttributes & FILE_ATTRIBUTE_REPARSE_POINT) && + (findbuf.dwReserved0 == IO_REPARSE_TAG_MOUNT_POINT); +} + int xutftowcsn(wchar_t *wcs, const char *utfs, size_t wcslen, int utflen) { int upos = 0, wpos = 0; diff --git a/compat/mingw.h b/compat/mingw.h index ebfb8ba423b754..92830de32a40e3 100644 --- a/compat/mingw.h +++ b/compat/mingw.h @@ -454,6 +454,9 @@ static inline void convert_slashes(char *path) if (*path == '\\') *path = '/'; } +struct strbuf; +int mingw_is_mount_point(struct strbuf *path); +#define is_mount_point mingw_is_mount_point #define PATH_SEP ';' char *mingw_query_user_email(void); #define query_user_email mingw_query_user_email diff --git a/git-compat-util.h b/git-compat-util.h index e123288e8f1393..df7f4437536dcf 100644 --- a/git-compat-util.h +++ b/git-compat-util.h @@ -626,6 +626,10 @@ static inline int git_has_dir_sep(const char *path) #define has_dir_sep(path) git_has_dir_sep(path) #endif +#ifndef is_mount_point +#define is_mount_point is_mount_point_via_stat +#endif + #ifndef query_user_email #define query_user_email() NULL #endif diff --git a/path.c b/path.c index 910756c8b3249c..f60c07b76b6c29 100644 --- a/path.c +++ b/path.c @@ -1300,6 +1300,45 @@ char *strip_path_suffix(const char *path, const char *suffix) return offset == -1 ? NULL : xstrndup(path, offset); } +int is_mount_point_via_stat(struct strbuf *path) +{ + size_t len = path->len; + dev_t current_dev; + struct stat st; + + if (!strcmp("/", path->buf)) + return 1; + + strbuf_addstr(path, "/."); + if (lstat(path->buf, &st)) { + /* + * If we cannot access the current directory, we cannot say + * that it is a bind mount. + */ + strbuf_setlen(path, len); + return 0; + } + current_dev = st.st_dev; + + /* Now look at the parent directory */ + strbuf_addch(path, '.'); + if (lstat(path->buf, &st)) { + /* + * If we cannot access the parent directory, we cannot say + * that it is a bind mount. + */ + strbuf_setlen(path, len); + return 0; + } + strbuf_setlen(path, len); + + /* + * If the device ID differs between current and parent directory, + * then it is a bind mount. + */ + return current_dev != st.st_dev; +} + int daemon_avoid_alias(const char *p) { int sl, ndot; diff --git a/path.h b/path.h index 65fe968a13a191..535b8456c646f2 100644 --- a/path.h +++ b/path.h @@ -172,6 +172,7 @@ int normalize_path_copy(char *dst, const char *src); int strbuf_normalize_path(struct strbuf *src); int longest_ancestor_length(const char *path, struct string_list *prefixes); char *strip_path_suffix(const char *path, const char *suffix); +int is_mount_point_via_stat(struct strbuf *path); int daemon_avoid_alias(const char *path); /* diff --git a/t/t7300-clean.sh b/t/t7300-clean.sh index 00d4070156243b..7c3a1ca91df534 100755 --- a/t/t7300-clean.sh +++ b/t/t7300-clean.sh @@ -800,4 +800,13 @@ test_expect_success 'traverse into directories that may have ignored entries' ' ) ' +test_expect_success MINGW 'clean does not traverse mount points' ' + mkdir target && + >target/dont-clean-me && + git init with-mountpoint && + cmd //c "mklink /j with-mountpoint\\mountpoint target" && + git -C with-mountpoint clean -dfx && + test_path_is_file target/dont-clean-me +' + test_done From 81c4c44a6a863ceec58e4afd2c5e769a098407da Mon Sep 17 00:00:00 2001 From: Jeff Hostetler Date: Fri, 12 May 2023 15:54:11 -0400 Subject: [PATCH 094/312] mimalloc: use "weak" random seed when statically linked Always use the internal "use_weak" random seed when initializing the "mimalloc" heap when statically linked on Windows. The imported "mimalloc" routines support several random sources to seed the heap data structures, including BCrypt.dll and RtlGenRandom. Crashes have been reported when using BCrypt.dll if it initialized during an `atexit()` handler function. Granted, such DLL initialization should not happen in an atexit handler, but yet the crashes remain. It should be noted that on Windows when statically linked, the mimalloc startup code (called by the GCC CRT to initialize static data prior to calling `main()`) always uses the internal "weak" random seed. "mimalloc" does not try to load an alternate random source until after the OS initialization has completed. Heap data is stored in `__declspec(thread)` TLS data and in theory each Git thread will have its own heap data. However, testing shows that the "mimalloc" library doesn't actually call `os_random_buf()` (to load a new random source) when creating these new per-thread heap structures. However, if an atexit handler is forced to run on a non-main thread, the "mimalloc" library *WILL* try to create a new heap and seed it with `os_random_buf()`. (The reason for this is still a mystery to this author.) The `os_random_buf()` call can cause the (previously uninitialized BCrypt.dll library) to be dynamically loaded and a call made into it. Crashes have been reported in v2.40.1.vfs.0.0 while in this call. As a workaround, the fix here forces the use of the internal "use_weak" random code for the subsequent `os_random_buf()` calls. Since we have been using that random generator for the majority of the program, it seems safe to use it for the final few mallocs in the atexit handler (of which there really shouldn't be that many. Signed-off-by: Jeff Hostetler Signed-off-by: Johannes Schindelin --- compat/mimalloc/init.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/compat/mimalloc/init.c b/compat/mimalloc/init.c index 4670d5510db187..4ec5812e3ce1d0 100644 --- a/compat/mimalloc/init.c +++ b/compat/mimalloc/init.c @@ -302,7 +302,11 @@ static bool _mi_heap_init(void) { _mi_memcpy_aligned(tld, &tld_empty, sizeof(*tld)); _mi_memcpy_aligned(heap, &_mi_heap_empty, sizeof(*heap)); heap->thread_id = _mi_thread_id(); + #if defined(_WIN32) && !defined(MI_SHARED_LIB) + _mi_random_init_weak(&heap->random); // match mi_heap_main_init() + #else _mi_random_init(&heap->random); + #endif heap->cookie = _mi_heap_random_next(heap) | 1; heap->keys[0] = _mi_heap_random_next(heap); heap->keys[1] = _mi_heap_random_next(heap); From d6ae4d0fb3e1da84c75084c618f2125db4ad604f Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Tue, 11 Dec 2018 12:55:26 +0100 Subject: [PATCH 095/312] clean: remove mount points when possible Windows' equivalent to "bind mounts", NTFS junction points, can be unlinked without affecting the mount target. This is clearly what users expect to happen when they call `git clean -dfx` in a worktree that contains NTFS junction points: the junction should be removed, and the target directory of said junction should be left alone (unless it is inside the worktree). Signed-off-by: Johannes Schindelin --- builtin/clean.c | 13 +++++++++++++ compat/mingw.h | 1 + t/t7300-clean.sh | 1 + 3 files changed, 15 insertions(+) diff --git a/builtin/clean.c b/builtin/clean.c index aeb7eb2d1ed2c3..df39efcc50fbc4 100644 --- a/builtin/clean.c +++ b/builtin/clean.c @@ -40,8 +40,10 @@ static const char *msg_remove = N_("Removing %s\n"); static const char *msg_would_remove = N_("Would remove %s\n"); static const char *msg_skip_git_dir = N_("Skipping repository %s\n"); static const char *msg_would_skip_git_dir = N_("Would skip repository %s\n"); +#ifndef CAN_UNLINK_MOUNT_POINTS static const char *msg_skip_mount_point = N_("Skipping mount point %s\n"); static const char *msg_would_skip_mount_point = N_("Would skip mount point %s\n"); +#endif static const char *msg_warn_remove_failed = N_("failed to remove %s"); static const char *msg_warn_lstat_failed = N_("could not lstat %s\n"); static const char *msg_skip_cwd = N_("Refusing to remove current working directory\n"); @@ -187,6 +189,7 @@ static int remove_dirs(struct strbuf *path, const char *prefix, int force_flag, } if (is_mount_point(path)) { +#ifndef CAN_UNLINK_MOUNT_POINTS if (!quiet) { quote_path(path->buf, prefix, "ed, 0); printf(dry_run ? @@ -194,6 +197,16 @@ static int remove_dirs(struct strbuf *path, const char *prefix, int force_flag, _(msg_skip_mount_point), quoted.buf); } *dir_gone = 0; +#else + if (!dry_run && unlink(path->buf)) { + int saved_errno = errno; + quote_path(path->buf, prefix, "ed, 0); + errno = saved_errno; + warning_errno(_(msg_warn_remove_failed), quoted.buf); + *dir_gone = 0; + ret = -1; + } +#endif goto out; } diff --git a/compat/mingw.h b/compat/mingw.h index 92830de32a40e3..92aa29bb8bb790 100644 --- a/compat/mingw.h +++ b/compat/mingw.h @@ -457,6 +457,7 @@ static inline void convert_slashes(char *path) struct strbuf; int mingw_is_mount_point(struct strbuf *path); #define is_mount_point mingw_is_mount_point +#define CAN_UNLINK_MOUNT_POINTS 1 #define PATH_SEP ';' char *mingw_query_user_email(void); #define query_user_email mingw_query_user_email diff --git a/t/t7300-clean.sh b/t/t7300-clean.sh index 7c3a1ca91df534..6f16f3893191e7 100755 --- a/t/t7300-clean.sh +++ b/t/t7300-clean.sh @@ -806,6 +806,7 @@ test_expect_success MINGW 'clean does not traverse mount points' ' git init with-mountpoint && cmd //c "mklink /j with-mountpoint\\mountpoint target" && git -C with-mountpoint clean -dfx && + test_path_is_missing with-mountpoint/mountpoint && test_path_is_file target/dont-clean-me ' From c0ffa199143cf6a73a128ceb38ca6a74ba891022 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Mon, 16 Feb 2015 14:06:59 +0100 Subject: [PATCH 096/312] mingw: include the Python parts in the build While Git for Windows does not _ship_ Python (in order to save on bandwidth), MSYS2 provides very fine Python interpreters that users can easily take advantage of, by using Git for Windows within its SDK. Signed-off-by: Johannes Schindelin --- config.mak.uname | 1 + 1 file changed, 1 insertion(+) diff --git a/config.mak.uname b/config.mak.uname index b12d4e168ae119..7fba3e595c5cd5 100644 --- a/config.mak.uname +++ b/config.mak.uname @@ -739,6 +739,7 @@ ifeq ($(uname_S),MINGW) USE_GETTEXT_SCHEME = fallthrough USE_LIBPCRE = YesPlease USE_NED_ALLOCATOR = YesPlease + NO_PYTHON = ifeq (/mingw64,$(subst 32,64,$(prefix))) # Move system config into top-level /etc/ ETC_GITCONFIG = ../etc/gitconfig From 3e5fd904c602867490f18aef75c99c3c2755da37 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Mon, 24 Jun 2019 23:45:21 +0200 Subject: [PATCH 097/312] mingw: use mimalloc Thorough benchmarking with repacking a subset of linux.git (the commit history reachable from 93a6fefe2f ([PATCH] fix the SYSCTL=n compilation, 2007-02-28), to be precise) suggest that this allocator is on par, in multi-threaded situations maybe even better than nedmalloc: `git repack -adfq` with mimalloc, 8 threads: 31.166991900 27.576763800 28.712311000 27.373859000 27.163141900 `git repack -adfq` with nedmalloc, 8 threads: 31.915032900 27.149883100 28.244933700 27.240188800 28.580849500 In a different test using GitHub Actions build agents (probably single-threaded, a core-strength of nedmalloc)): `git repack -q -d -l -A --unpack-unreachable=2.weeks.ago` with mimalloc: 943.426 978.500 939.709 959.811 954.605 `git repack -q -d -l -A --unpack-unreachable=2.weeks.ago` with nedmalloc: 995.383 952.179 943.253 963.043 980.468 While these measurements were not executed with complete scientific rigor, as no hardware was set aside specifically for these benchmarks, it shows that mimalloc and nedmalloc perform almost the same, nedmalloc with a bit higher variance and also slightly higher average (further testing suggests that nedmalloc performs worse in multi-threaded situations than in single-threaded ones). In short: mimalloc seems to be slightly better suited for our purposes than nedmalloc. Seeing that mimalloc is developed actively, while nedmalloc ceased to see any updates in eight years, let's use mimalloc on Windows instead. Signed-off-by: Johannes Schindelin --- config.mak.uname | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/config.mak.uname b/config.mak.uname index cdfb606f010ba4..2ebb82feb9f1b2 100644 --- a/config.mak.uname +++ b/config.mak.uname @@ -738,7 +738,7 @@ ifeq ($(uname_S),MINGW) HAVE_LIBCHARSET_H = YesPlease USE_GETTEXT_SCHEME = fallthrough USE_LIBPCRE = YesPlease - USE_NED_ALLOCATOR = YesPlease + USE_MIMALLOC = YesPlease NO_PYTHON = ifeq (/mingw64,$(subst 32,64,$(prefix))) # Move system config into top-level /etc/ From 228fc1099f9e650a083edc230966df1bfe4b2fb4 Mon Sep 17 00:00:00 2001 From: Thomas Braun Date: Thu, 8 May 2014 21:43:24 +0200 Subject: [PATCH 098/312] transport: optionally disable side-band-64k Since commit 0c499ea60fda (send-pack: demultiplex a sideband stream with status data, 2010-02-05) the send-pack builtin uses the side-band-64k capability if advertised by the server. Unfortunately this breaks pushing over the dump git protocol if used over a network connection. The detailed reasons for this breakage are (by courtesy of Jeff Preshing, quoted from https://groups.google.com/d/msg/msysgit/at8D7J-h7mw/eaLujILGUWoJ): MinGW wraps Windows sockets in CRT file descriptors in order to mimic the functionality of POSIX sockets. This causes msvcrt.dll to treat sockets as Installable File System (IFS) handles, calling ReadFile, WriteFile, DuplicateHandle and CloseHandle on them. This approach works well in simple cases on recent versions of Windows, but does not support all usage patterns. In particular, using this approach, any attempt to read & write concurrently on the same socket (from one or more processes) will deadlock in a scenario where the read waits for a response from the server which is only invoked after the write. This is what send_pack currently attempts to do in the use_sideband codepath. The new config option `sendpack.sideband` allows to override the side-band-64k capability of the server, and thus makes the dumb git protocol work. Other transportation methods like ssh and http/https still benefit from the sideband channel, therefore the default value of `sendpack.sideband` is still true. Signed-off-by: Thomas Braun Signed-off-by: Oliver Schneider Signed-off-by: Johannes Schindelin --- Documentation/config.adoc | 2 ++ Documentation/config/sendpack.adoc | 5 +++++ send-pack.c | 6 +++--- 3 files changed, 10 insertions(+), 3 deletions(-) create mode 100644 Documentation/config/sendpack.adoc diff --git a/Documentation/config.adoc b/Documentation/config.adoc index a8b04c4e5183d2..644216f02a322a 100644 --- a/Documentation/config.adoc +++ b/Documentation/config.adoc @@ -518,6 +518,8 @@ include::config/safe.adoc[] include::config/sendemail.adoc[] +include::config/sendpack.adoc[] + include::config/sequencer.adoc[] include::config/showbranch.adoc[] diff --git a/Documentation/config/sendpack.adoc b/Documentation/config/sendpack.adoc new file mode 100644 index 00000000000000..e306f657fba7dd --- /dev/null +++ b/Documentation/config/sendpack.adoc @@ -0,0 +1,5 @@ +sendpack.sideband:: + Allows to disable the side-band-64k capability for send-pack even + when it is advertised by the server. Makes it possible to work + around a limitation in the git for windows implementation together + with the dump git protocol. Defaults to true. diff --git a/send-pack.c b/send-pack.c index 856a65d5f5abce..5ae6d9f8e7408a 100644 --- a/send-pack.c +++ b/send-pack.c @@ -497,7 +497,7 @@ int send_pack(struct repository *r, int need_pack_data = 0; int allow_deleting_refs = 0; int status_report = 0; - int use_sideband = 0; + int use_sideband = 1; int quiet_supported = 0; int agent_supported = 0; int advertise_sid = 0; @@ -521,6 +521,7 @@ int send_pack(struct repository *r, goto out; } + repo_config_get_bool(r, "sendpack.sideband", &use_sideband); repo_config_get_bool(r, "push.negotiate", &push_negotiate); if (push_negotiate) { trace2_region_enter("send_pack", "push_negotiate", r); @@ -542,8 +543,7 @@ int send_pack(struct repository *r, allow_deleting_refs = 1; if (server_supports("ofs-delta")) args->use_ofs_delta = 1; - if (server_supports("side-band-64k")) - use_sideband = 1; + use_sideband = use_sideband && server_supports("side-band-64k"); if (server_supports("quiet")) quiet_supported = 1; if (server_supports("agent")) From b1a6666962ca32dae4190f54e802c374c122e479 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Wed, 1 Jan 2020 21:07:22 +0100 Subject: [PATCH 099/312] mingw: do resolve symlinks in `getcwd()` As pointed out in https://github.com/git-for-windows/git/issues/1676, the `git rev-parse --is-inside-work-tree` command currently fails when the current directory's path contains symbolic links. The underlying reason for this bug is that `getcwd()` is supposed to resolve symbolic links, but our `mingw_getcwd()` implementation did not. We do have all the building blocks for that, though: the `GetFinalPathByHandleW()` function will resolve symbolic links. However, we only called that function if `GetLongPathNameW()` failed, for historical reasons: the latter function was supported for a long time, but the former API function was introduced only with Windows Vista, and we used to support also Windows XP. With that support having been dropped, we are free to call the symbolic link-resolving function right away. Signed-off-by: Johannes Schindelin --- compat/mingw.c | 18 +++++++----------- 1 file changed, 7 insertions(+), 11 deletions(-) diff --git a/compat/mingw.c b/compat/mingw.c index f524c54d06d965..579a6a815f777a 100644 --- a/compat/mingw.c +++ b/compat/mingw.c @@ -1211,18 +1211,16 @@ char *mingw_getcwd(char *pointer, int len) { wchar_t cwd[MAX_PATH], wpointer[MAX_PATH]; DWORD ret = GetCurrentDirectoryW(ARRAY_SIZE(cwd), cwd); + HANDLE hnd; if (!ret || ret >= ARRAY_SIZE(cwd)) { errno = ret ? ENAMETOOLONG : err_win_to_posix(GetLastError()); return NULL; } - ret = GetLongPathNameW(cwd, wpointer, ARRAY_SIZE(wpointer)); - if (!ret && GetLastError() == ERROR_ACCESS_DENIED) { - HANDLE hnd = CreateFileW(cwd, 0, - FILE_SHARE_READ | FILE_SHARE_WRITE | FILE_SHARE_DELETE, NULL, - OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL); - if (hnd == INVALID_HANDLE_VALUE) - return NULL; + hnd = CreateFileW(cwd, 0, + FILE_SHARE_READ | FILE_SHARE_WRITE | FILE_SHARE_DELETE, NULL, + OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL); + if (hnd != INVALID_HANDLE_VALUE) { ret = GetFinalPathNameByHandleW(hnd, wpointer, ARRAY_SIZE(wpointer), 0); CloseHandle(hnd); if (!ret || ret >= ARRAY_SIZE(wpointer)) @@ -1231,13 +1229,11 @@ char *mingw_getcwd(char *pointer, int len) return NULL; return pointer; } - if (!ret || ret >= ARRAY_SIZE(wpointer)) - return NULL; - if (GetFileAttributesW(wpointer) == INVALID_FILE_ATTRIBUTES) { + if (GetFileAttributesW(cwd) == INVALID_FILE_ATTRIBUTES) { errno = ENOENT; return NULL; } - if (xwcstoutf(pointer, wpointer, len) < 0) + if (xwcstoutf(pointer, cwd, len) < 0) return NULL; convert_slashes(pointer); return pointer; From 9c84f9571157c7ffe30d3aa0409b000223c2c584 Mon Sep 17 00:00:00 2001 From: Bjoern Mueller Date: Wed, 22 Jan 2020 13:49:13 +0100 Subject: [PATCH 100/312] mingw: fix fatal error working on mapped network drives on Windows In 1e64d18 (mingw: do resolve symlinks in `getcwd()`) a problem was introduced that causes git for Windows to stop working with certain mapped network drives (in particular, drives that are mapped to locations with long path names). Error message was "fatal: Unable to read current working directory: No such file or directory". Present change fixes this issue as discussed in https://github.com/git-for-windows/git/issues/2480 Signed-off-by: Bjoern Mueller --- compat/mingw.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) diff --git a/compat/mingw.c b/compat/mingw.c index f524c54d06d965..41943760f8c8fa 100644 --- a/compat/mingw.c +++ b/compat/mingw.c @@ -1225,8 +1225,13 @@ char *mingw_getcwd(char *pointer, int len) return NULL; ret = GetFinalPathNameByHandleW(hnd, wpointer, ARRAY_SIZE(wpointer), 0); CloseHandle(hnd); - if (!ret || ret >= ARRAY_SIZE(wpointer)) - return NULL; + if (!ret || ret >= ARRAY_SIZE(wpointer)) { + ret = GetLongPathNameW(cwd, wpointer, ARRAY_SIZE(wpointer)); + if (!ret || ret >= ARRAY_SIZE(wpointer)) { + errno = ret ? ENAMETOOLONG : err_win_to_posix(GetLastError()); + return NULL; + } + } if (xwcstoutf(pointer, normalize_ntpath(wpointer), len) < 0) return NULL; return pointer; From 4c3f5c3876cf1b4e124c30d6d1ded85c315447b6 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Fri, 31 Jan 2020 12:02:47 +0100 Subject: [PATCH 101/312] mingw: demonstrate a `git add` issue with NTFS junctions NTFS junctions are somewhat similar in spirit to Unix bind mounts: they point to a different directory and are resolved by the filesystem driver. As such, they appear to `lstat()` as if they are directories, not as if they are symbolic links. _Any_ user can create junctions, while symbolic links can only be created by non-administrators in Developer Mode on Windows 10. Hence NTFS junctions are much more common "in the wild" than NTFS symbolic links. It was reported in https://github.com/git-for-windows/git/issues/2481 that adding files via an absolute path that traverses an NTFS junction: since 1e64d18 (mingw: do resolve symlinks in `getcwd()`), we resolve not only symbolic links but also NTFS junctions when determining the absolute path of the current directory. The same is not true for `git add `, where symbolic links are resolved in ``, but not NTFS junctions. Signed-off-by: Johannes Schindelin --- t/t3700-add.sh | 11 +++++++++++ 1 file changed, 11 insertions(+) diff --git a/t/t3700-add.sh b/t/t3700-add.sh index df580a5806b4f1..d934f3417faa45 100755 --- a/t/t3700-add.sh +++ b/t/t3700-add.sh @@ -548,4 +548,15 @@ test_expect_success CASE_INSENSITIVE_FS 'path is case-insensitive' ' git add "$downcased" ' +test_expect_failure MINGW 'can add files via NTFS junctions' ' + test_when_finished "cmd //c rmdir junction && rm -rf target" && + test_create_repo target && + cmd //c "mklink /j junction target" && + >target/via-junction && + git -C junction add "$(pwd)/junction/via-junction" && + echo via-junction >expect && + git -C target diff --cached --name-only >actual && + test_cmp expect actual +' + test_done From 8e69e5fae1a66370794937ab42375946141dea49 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Tue, 21 Feb 2017 13:28:58 +0100 Subject: [PATCH 102/312] mingw: ensure valid CTYPE A change between versions 2.4.1 and 2.6.0 of the MSYS2 runtime modified how Cygwin's runtime (and hence Git for Windows' MSYS2 runtime derivative) handles locales: d16a56306d (Consolidate wctomb/mbtowc calls for POSIX-1.2008, 2016-07-20). An unintended side-effect is that "cold-calling" into the POSIX emulation will start with a locale based on the current code page, something that Git for Windows is very ill-prepared for, as it expects to be able to pass a command-line containing non-ASCII characters to the shell without having those characters munged. One symptom of this behavior: when `git clone` or `git fetch` shell out to call `git-upload-pack` with a path that contains non-ASCII characters, the shell tried to interpret the entire command-line (including command-line parameters) as executable path, which obviously must fail. This fixes https://github.com/git-for-windows/git/issues/1036 Signed-off-by: Johannes Schindelin --- compat/mingw.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/compat/mingw.c b/compat/mingw.c index f524c54d06d965..ad00d56ed0cc58 100644 --- a/compat/mingw.c +++ b/compat/mingw.c @@ -2824,6 +2824,9 @@ static void setup_windows_environment(void) if (!tmp && (tmp = getenv("USERPROFILE"))) setenv("HOME", tmp, 1); } + + if (!getenv("LC_ALL") && !getenv("LC_CTYPE") && !getenv("LANG")) + setenv("LC_CTYPE", "C.UTF-8", 1); } static PSID get_current_user_sid(void) From d8983c950a0db03ec62d76e0fde24cd1d7f31156 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Fri, 31 Jan 2020 11:44:31 +0100 Subject: [PATCH 103/312] strbuf_realpath(): use platform-dependent API if available Some platforms (e.g. Windows) provide API functions to resolve paths much quicker. Let's offer a way to short-cut `strbuf_realpath()` on those platforms. Signed-off-by: Johannes Schindelin --- abspath.c | 3 +++ git-compat-util.h | 4 ++++ 2 files changed, 7 insertions(+) diff --git a/abspath.c b/abspath.c index 1202cde23dbc9b..0c17e98654e4b0 100644 --- a/abspath.c +++ b/abspath.c @@ -93,6 +93,9 @@ static char *strbuf_realpath_1(struct strbuf *resolved, const char *path, goto error_out; } + if (platform_strbuf_realpath(resolved, path)) + return resolved->buf; + strbuf_addstr(&remaining, path); get_root_part(resolved, &remaining); diff --git a/git-compat-util.h b/git-compat-util.h index e123288e8f1393..079b837e5a1f52 100644 --- a/git-compat-util.h +++ b/git-compat-util.h @@ -630,6 +630,10 @@ static inline int git_has_dir_sep(const char *path) #define query_user_email() NULL #endif +#ifndef platform_strbuf_realpath +#define platform_strbuf_realpath(resolved, path) NULL +#endif + #ifdef __TANDEM #include #include From 00932a228cb1bdd4585d778efc15cf3ad1441867 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Sat, 1 Feb 2020 00:31:16 +0100 Subject: [PATCH 104/312] mingw: allow `git.exe` to be used instead of the "Git wrapper" Git for Windows wants to add `git.exe` to the users' `PATH`, without cluttering the latter with unnecessary executables such as `wish.exe`. To that end, it invented the concept of its "Git wrapper", i.e. a tiny executable located in `C:\Program Files\Git\cmd\git.exe` (originally a CMD script) whose sole purpose is to set up a couple of environment variables and then spawn the _actual_ `git.exe` (which nowadays lives in `C:\Program Files\Git\mingw64\bin\git.exe` for 64-bit, and the obvious equivalent for 32-bit installations). Currently, the following environment variables are set unless already initialized: - `MSYSTEM`, to make sure that the MSYS2 Bash and the MSYS2 Perl interpreter behave as expected, and - `PLINK_PROTOCOL`, to force PuTTY's `plink.exe` to use the SSH protocol instead of Telnet, - `PATH`, to make sure that the `bin` folder in the user's home directory, as well as the `/mingw64/bin` and the `/usr/bin` directories are included. The trick here is that the `/mingw64/bin/` and `/usr/bin/` directories are relative to the top-level installation directory of Git for Windows (which the included Bash interprets as `/`, i.e. as the MSYS pseudo root directory). Using the absence of `MSYSTEM` as a tell-tale, we can detect in `git.exe` whether these environment variables have been initialized properly. Therefore we can call `C:\Program Files\Git\mingw64\bin\git` in-place after this change, without having to call Git through the Git wrapper. Obviously, above-mentioned directories must be _prepended_ to the `PATH` variable, otherwise we risk picking up executables from unrelated Git installations. We do that by constructing the new `PATH` value from scratch, appending `$HOME/bin` (if `HOME` is set), then the MSYS2 system directories, and then appending the original `PATH`. Side note: this modification of the `PATH` variable is independent of the modification necessary to reach the executables and scripts in `/mingw64/libexec/git-core/`, i.e. the `GIT_EXEC_PATH`. That modification is still performed by Git, elsewhere, long after making the changes described above. While we _still_ cannot simply hard-link `mingw64\bin\git.exe` to `cmd` (because the former depends on a couple of `.dll` files that are only in `mingw64\bin`, i.e. calling `...\cmd\git.exe` would fail to load due to missing dependencies), at least we can now avoid that extra process of running the Git wrapper (which then has to wait for the spawned `git.exe` to finish) by calling `...\mingw64\bin\git.exe` directly, via its absolute path. Testing this is in Git's test suite tricky: we set up a "new" MSYS pseudo-root and copy the `git.exe` file into the appropriate location, then verify that `MSYSTEM` is set properly, and also that the `PATH` is modified so that scripts can be found in `$HOME/bin`, `/mingw64/bin/` and `/usr/bin/`. This addresses https://github.com/git-for-windows/git/issues/2283 Signed-off-by: Johannes Schindelin --- compat/mingw.c | 72 +++++++++++++++++++++++++++++++++++++++++++ config.mak.uname | 4 +-- t/t0060-path-utils.sh | 31 ++++++++++++++++++- 3 files changed, 104 insertions(+), 3 deletions(-) diff --git a/compat/mingw.c b/compat/mingw.c index ad00d56ed0cc58..a3e5baf32e8e87 100644 --- a/compat/mingw.c +++ b/compat/mingw.c @@ -2773,6 +2773,50 @@ int xwcstoutf(char *utf, const wchar_t *wcs, size_t utflen) return -1; } +#ifdef ENSURE_MSYSTEM_IS_SET +#if !defined(RUNTIME_PREFIX) || !defined(HAVE_WPGMPTR) +static size_t append_system_bin_dirs(char *path UNUSED, size_t size UNUSED) +{ + return 0; +} +#else +static size_t append_system_bin_dirs(char *path, size_t size) +{ + char prefix[32768]; + const char *slash; + size_t len = xwcstoutf(prefix, _wpgmptr, sizeof(prefix)), off = 0; + + if (len == 0 || len >= sizeof(prefix) || + !(slash = find_last_dir_sep(prefix))) + return 0; + /* strip trailing `git.exe` */ + len = slash - prefix; + + /* strip trailing `cmd` or `mingw64\bin` or `mingw32\bin` or `bin` or `libexec\git-core` */ + if (strip_suffix_mem(prefix, &len, "\\mingw64\\libexec\\git-core") || + strip_suffix_mem(prefix, &len, "\\mingw64\\bin")) + off += xsnprintf(path + off, size - off, + "%.*s\\mingw64\\bin;", (int)len, prefix); + else if (strip_suffix_mem(prefix, &len, "\\mingw32\\libexec\\git-core") || + strip_suffix_mem(prefix, &len, "\\mingw32\\bin")) + off += xsnprintf(path + off, size - off, + "%.*s\\mingw32\\bin;", (int)len, prefix); + else if (strip_suffix_mem(prefix, &len, "\\cmd") || + strip_suffix_mem(prefix, &len, "\\bin") || + strip_suffix_mem(prefix, &len, "\\libexec\\git-core")) + off += xsnprintf(path + off, size - off, + "%.*s\\mingw%d\\bin;", (int)len, prefix, + (int)(sizeof(void *) * 8)); + else + return 0; + + off += xsnprintf(path + off, size - off, + "%.*s\\usr\\bin;", (int)len, prefix); + return off; +} +#endif +#endif + static void setup_windows_environment(void) { char *tmp = getenv("TMPDIR"); @@ -2825,6 +2869,34 @@ static void setup_windows_environment(void) setenv("HOME", tmp, 1); } + if (!getenv("PLINK_PROTOCOL")) + setenv("PLINK_PROTOCOL", "ssh", 0); + +#ifdef ENSURE_MSYSTEM_IS_SET + if (!(tmp = getenv("MSYSTEM")) || !tmp[0]) { + const char *home = getenv("HOME"), *path = getenv("PATH"); + char buf[32768]; + size_t off = 0; + + xsnprintf(buf, sizeof(buf), + "MINGW%d", (int)(sizeof(void *) * 8)); + setenv("MSYSTEM", buf, 1); + + if (home) + off += xsnprintf(buf + off, sizeof(buf) - off, + "%s\\bin;", home); + off += append_system_bin_dirs(buf + off, sizeof(buf) - off); + if (path) + off += xsnprintf(buf + off, sizeof(buf) - off, + "%s", path); + else if (off > 0) + buf[off - 1] = '\0'; + else + buf[0] = '\0'; + setenv("PATH", buf, 1); + } +#endif + if (!getenv("LC_ALL") && !getenv("LC_CTYPE") && !getenv("LANG")) setenv("LC_CTYPE", "C.UTF-8", 1); } diff --git a/config.mak.uname b/config.mak.uname index b12d4e168ae119..23748b65a8e480 100644 --- a/config.mak.uname +++ b/config.mak.uname @@ -501,7 +501,7 @@ endif compat/win32/pthread.o compat/win32/syslog.o \ compat/win32/trace2_win32_process_info.o \ compat/win32/dirent.o - COMPAT_CFLAGS = -D__USE_MINGW_ACCESS -DDETECT_MSYS_TTY -DNOGDI -DHAVE_STRING_H -Icompat -Icompat/regex -Icompat/win32 -DSTRIP_EXTENSION=\".exe\" + COMPAT_CFLAGS = -D__USE_MINGW_ACCESS -DDETECT_MSYS_TTY -DENSURE_MSYSTEM_IS_SET -DNOGDI -DHAVE_STRING_H -Icompat -Icompat/regex -Icompat/win32 -DSTRIP_EXTENSION=\".exe\" BASIC_LDFLAGS = -IGNORE:4217 -IGNORE:4049 -NOLOGO -ENTRY:wmainCRTStartup -SUBSYSTEM:CONSOLE # invalidcontinue.obj allows Git's source code to close the same file # handle twice, or to access the osfhandle of an already-closed stdout @@ -730,7 +730,7 @@ ifeq ($(uname_S),MINGW) endif CC = gcc COMPAT_CFLAGS += -D__USE_MINGW_ANSI_STDIO=0 -DDETECT_MSYS_TTY \ - -fstack-protector-strong + -DENSURE_MSYSTEM_IS_SET -fstack-protector-strong EXTLIBS += -lntdll EXTRA_PROGRAMS += headless-git$X INSTALL = /bin/install diff --git a/t/t0060-path-utils.sh b/t/t0060-path-utils.sh index 8545cdfab559b4..b25b6da58ec4be 100755 --- a/t/t0060-path-utils.sh +++ b/t/t0060-path-utils.sh @@ -602,7 +602,8 @@ test_expect_success !VALGRIND,RUNTIME_PREFIX,CAN_EXEC_IN_PWD 'RUNTIME_PREFIX wor echo "echo HERE" | write_script pretend/libexec/git-core/git-here && GIT_EXEC_PATH= ./pretend/bin/git here >actual && echo HERE >expect && - test_cmp expect actual' + test_cmp expect actual +' test_expect_success !VALGRIND,RUNTIME_PREFIX,CAN_EXEC_IN_PWD '%(prefix)/ works' ' git config yes.path "%(prefix)/yes" && @@ -611,4 +612,32 @@ test_expect_success !VALGRIND,RUNTIME_PREFIX,CAN_EXEC_IN_PWD '%(prefix)/ works' test_cmp expect actual ' +test_expect_success MINGW,RUNTIME_PREFIX 'MSYSTEM/PATH is adjusted if necessary' ' + mkdir -p "$HOME"/bin pretend/mingw64/bin \ + pretend/mingw64/libexec/git-core pretend/usr/bin && + cp "$GIT_EXEC_PATH"/git.exe pretend/mingw64/bin/ && + cp "$GIT_EXEC_PATH"/git.exe pretend/mingw64/libexec/git-core/ && + # copy the .dll files, if any (happens when building via CMake) + case "$GIT_EXEC_PATH"/*.dll in + */"*.dll") ;; # no `.dll` files to be copied + *) + cp "$GIT_EXEC_PATH"/*.dll pretend/mingw64/bin/ && + cp "$GIT_EXEC_PATH"/*.dll pretend/mingw64/libexec/git-core/ + ;; + esac && + echo "env | grep MSYSTEM=" | write_script "$HOME"/bin/git-test-home && + echo "echo mingw64" | write_script pretend/mingw64/bin/git-test-bin && + echo "echo usr" | write_script pretend/usr/bin/git-test-bin2 && + + ( + MSYSTEM= && + GIT_EXEC_PATH= && + pretend/mingw64/libexec/git-core/git.exe test-home >actual && + pretend/mingw64/libexec/git-core/git.exe test-bin >>actual && + pretend/mingw64/bin/git.exe test-bin2 >>actual + ) && + test_write_lines MSYSTEM=$MSYSTEM mingw64 usr >expect && + test_cmp expect actual +' + test_done From 02fd27ac49262ec70796930e3bf5629c249777f5 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Sat, 9 May 2020 16:19:06 +0200 Subject: [PATCH 105/312] t5505/t5516: allow running without `.git/branches/` in the templates When we commit the template directory as part of `make vcxproj`, the `branches/` directory is not actually commited, as it is empty. Two tests were not prepared for that situation. This developer tried to get rid of the support for `.git/branches/` a long time ago, but that effort did not bear fruit, so the best we can do is work around in these here tests. Signed-off-by: Johannes Schindelin --- t/t5505-remote.sh | 4 ++-- t/t5516-fetch-push.sh | 8 ++++---- 2 files changed, 6 insertions(+), 6 deletions(-) diff --git a/t/t5505-remote.sh b/t/t5505-remote.sh index bb7e0c6879ed8a..75e2177e8ce254 100755 --- a/t/t5505-remote.sh +++ b/t/t5505-remote.sh @@ -1155,7 +1155,7 @@ test_expect_success WITHOUT_BREAKING_CHANGES 'migrate a remote from named file i ( cd six && git remote rm origin && - mkdir .git/branches && + mkdir -p .git/branches && echo "$origin_url#main" >.git/branches/origin && git remote rename origin origin && test_path_is_missing .git/branches/origin && @@ -1170,7 +1170,7 @@ test_expect_success WITHOUT_BREAKING_CHANGES 'migrate a remote from named file i ( cd seven && git remote rm origin && - mkdir .git/branches && + mkdir -p .git/branches && echo "quux#foom" > .git/branches/origin && git remote rename origin origin && test_path_is_missing .git/branches/origin && diff --git a/t/t5516-fetch-push.sh b/t/t5516-fetch-push.sh index 85ed049627d2e6..3aa7cc14c032a7 100755 --- a/t/t5516-fetch-push.sh +++ b/t/t5516-fetch-push.sh @@ -979,7 +979,7 @@ test_expect_success WITHOUT_BREAKING_CHANGES 'fetch with branches' ' mk_empty testrepo && git branch second $the_first_commit && git checkout second && - mkdir testrepo/.git/branches && + mkdir -p testrepo/.git/branches && echo ".." > testrepo/.git/branches/branch1 && ( cd testrepo && @@ -993,7 +993,7 @@ test_expect_success WITHOUT_BREAKING_CHANGES 'fetch with branches' ' test_expect_success WITHOUT_BREAKING_CHANGES 'fetch with branches containing #' ' mk_empty testrepo && - mkdir testrepo/.git/branches && + mkdir -p testrepo/.git/branches && echo "..#second" > testrepo/.git/branches/branch2 && ( cd testrepo && @@ -1010,7 +1010,7 @@ test_expect_success WITHOUT_BREAKING_CHANGES 'push with branches' ' git checkout second && test_when_finished "rm -rf .git/branches" && - mkdir .git/branches && + mkdir -p .git/branches && echo "testrepo" > .git/branches/branch1 && git push branch1 && @@ -1026,7 +1026,7 @@ test_expect_success WITHOUT_BREAKING_CHANGES 'push with branches containing #' ' mk_empty testrepo && test_when_finished "rm -rf .git/branches" && - mkdir .git/branches && + mkdir -p .git/branches && echo "testrepo#branch3" > .git/branches/branch2 && git push branch2 && From 0d79dc7902724d85c31f51e384f5a7e013a44bd6 Mon Sep 17 00:00:00 2001 From: Jeff Hostetler Date: Thu, 30 Jan 2020 14:22:27 -0500 Subject: [PATCH 106/312] clink.pl: fix MSVC compile script to handle libcurl-d.lib Update clink.pl to link with either libcurl.lib or libcurl-d.lib depending on whether DEBUG=1 is set. Signed-off-by: Jeff Hostetler Signed-off-by: Johannes Schindelin --- compat/vcbuild/scripts/clink.pl | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/compat/vcbuild/scripts/clink.pl b/compat/vcbuild/scripts/clink.pl index 3bd824154be381..c4c99d1a11f18c 100755 --- a/compat/vcbuild/scripts/clink.pl +++ b/compat/vcbuild/scripts/clink.pl @@ -56,7 +56,8 @@ # need to use that instead? foreach my $flag (@lflags) { if ($flag =~ /^-LIBPATH:(.*)/) { - foreach my $l ("libcurl_imp.lib", "libcurl.lib") { + my $libcurl = $is_debug ? "libcurl-d.lib" : "libcurl.lib"; + foreach my $l ("libcurl_imp.lib", $libcurl) { if (-f "$1/$l") { $lib = $l; last; From 1a828e57e1fa6cc8bba1413f4e3f6e6279c7cd1e Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Fri, 31 Jan 2020 11:49:04 +0100 Subject: [PATCH 107/312] mingw: implement a platform-specific `strbuf_realpath()` There is a Win32 API function to resolve symbolic links, and we can use that instead of resolving them manually. Even better, this function also resolves NTFS junction points (which are somewhat similar to bind mounts). This fixes https://github.com/git-for-windows/git/issues/2481. Signed-off-by: Johannes Schindelin --- compat/mingw.c | 76 +++++++++++++++++++++++++++++++++++++++++++ compat/mingw.h | 3 ++ t/t0060-path-utils.sh | 8 +++++ t/t3700-add.sh | 2 +- t/t5601-clone.sh | 7 ++++ 5 files changed, 95 insertions(+), 1 deletion(-) diff --git a/compat/mingw.c b/compat/mingw.c index f524c54d06d965..7a4dc87fa17e10 100644 --- a/compat/mingw.c +++ b/compat/mingw.c @@ -1207,6 +1207,82 @@ struct tm *localtime_r(const time_t *timep, struct tm *result) } #endif +char *mingw_strbuf_realpath(struct strbuf *resolved, const char *path) +{ + wchar_t wpath[MAX_PATH]; + HANDLE h; + DWORD ret; + int len; + const char *last_component = NULL; + char *append = NULL; + + if (xutftowcs_path(wpath, path) < 0) + return NULL; + + h = CreateFileW(wpath, 0, + FILE_SHARE_READ | FILE_SHARE_WRITE | FILE_SHARE_DELETE, NULL, + OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL); + + /* + * strbuf_realpath() allows the last path component to not exist. If + * that is the case, now it's time to try without last component. + */ + if (h == INVALID_HANDLE_VALUE && + GetLastError() == ERROR_FILE_NOT_FOUND) { + /* cut last component off of `wpath` */ + wchar_t *p = wpath + wcslen(wpath); + + while (p != wpath) + if (*(--p) == L'/' || *p == L'\\') + break; /* found start of last component */ + + if (p != wpath && (last_component = find_last_dir_sep(path))) { + append = xstrdup(last_component + 1); /* skip directory separator */ + /* + * Do not strip the trailing slash at the drive root, otherwise + * the path would be e.g. `C:` (which resolves to the + * _current_ directory on that drive). + */ + if (p[-1] == L':') + p[1] = L'\0'; + else + *p = L'\0'; + h = CreateFileW(wpath, 0, FILE_SHARE_READ | + FILE_SHARE_WRITE | FILE_SHARE_DELETE, + NULL, OPEN_EXISTING, + FILE_FLAG_BACKUP_SEMANTICS, NULL); + } + } + + if (h == INVALID_HANDLE_VALUE) { +realpath_failed: + FREE_AND_NULL(append); + return NULL; + } + + ret = GetFinalPathNameByHandleW(h, wpath, ARRAY_SIZE(wpath), 0); + CloseHandle(h); + if (!ret || ret >= ARRAY_SIZE(wpath)) + goto realpath_failed; + + len = wcslen(wpath) * 3; + strbuf_grow(resolved, len); + len = xwcstoutf(resolved->buf, normalize_ntpath(wpath), len); + if (len < 0) + goto realpath_failed; + resolved->len = len; + + if (append) { + /* Use forward-slash, like `normalize_ntpath()` */ + strbuf_complete(resolved, '/'); + strbuf_addstr(resolved, append); + FREE_AND_NULL(append); + } + + return resolved->buf; + +} + char *mingw_getcwd(char *pointer, int len) { wchar_t cwd[MAX_PATH], wpointer[MAX_PATH]; diff --git a/compat/mingw.h b/compat/mingw.h index ebfb8ba423b754..35c427376cbc6d 100644 --- a/compat/mingw.h +++ b/compat/mingw.h @@ -457,6 +457,9 @@ static inline void convert_slashes(char *path) #define PATH_SEP ';' char *mingw_query_user_email(void); #define query_user_email mingw_query_user_email +struct strbuf; +char *mingw_strbuf_realpath(struct strbuf *resolved, const char *path); +#define platform_strbuf_realpath mingw_strbuf_realpath #if !defined(__MINGW64_VERSION_MAJOR) && (!defined(_MSC_VER) || _MSC_VER < 1800) #define PRIuMAX "I64u" #define PRId64 "I64d" diff --git a/t/t0060-path-utils.sh b/t/t0060-path-utils.sh index 8545cdfab559b4..eb2ab9d437ea8e 100755 --- a/t/t0060-path-utils.sh +++ b/t/t0060-path-utils.sh @@ -281,6 +281,14 @@ test_expect_success SYMLINKS 'real path works on symlinks' ' test_cmp expect actual ' +test_expect_success MINGW 'real path works near drive root' ' + # we need a non-existing path at the drive root; simply skip if C:/xyz exists + if test ! -e C:/xyz + then + test C:/xyz = $(test-tool path-utils real_path C:/xyz) + fi +' + test_expect_success SYMLINKS 'prefix_path works with absolute paths to work tree symlinks' ' ln -s target symlink && echo "symlink" >expect && diff --git a/t/t3700-add.sh b/t/t3700-add.sh index d934f3417faa45..10185f31f019a9 100755 --- a/t/t3700-add.sh +++ b/t/t3700-add.sh @@ -548,7 +548,7 @@ test_expect_success CASE_INSENSITIVE_FS 'path is case-insensitive' ' git add "$downcased" ' -test_expect_failure MINGW 'can add files via NTFS junctions' ' +test_expect_success MINGW 'can add files via NTFS junctions' ' test_when_finished "cmd //c rmdir junction && rm -rf target" && test_create_repo target && cmd //c "mklink /j junction target" && diff --git a/t/t5601-clone.sh b/t/t5601-clone.sh index d0c18660e33113..b3ecd6b8b06c33 100755 --- a/t/t5601-clone.sh +++ b/t/t5601-clone.sh @@ -78,6 +78,13 @@ test_expect_success 'clone respects GIT_WORK_TREE' ' ' +test_expect_success CASE_INSENSITIVE_FS 'core.worktree is not added due to path case' ' + + mkdir UPPERCASE && + git clone src "$(pwd)/uppercase" && + test "unset" = "$(git -C UPPERCASE config --default unset core.worktree)" +' + test_expect_success 'clone from hooks' ' test_create_repo r0 && From 5cf32c76b3d590d26538c533123f4b501547506c Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Tue, 25 Aug 2020 12:13:26 +0200 Subject: [PATCH 108/312] mingw: ignore HOMEDRIVE/HOMEPATH if it points to Windows' system directory Internally, Git expects the environment variable `HOME` to be set, and to point to the current user's home directory. This environment variable is not set by default on Windows, and therefore Git tries its best to construct one if it finds `HOME` unset. There are actually two different approaches Git tries: first, it looks at `HOMEDRIVE`/`HOMEPATH` because this is widely used in corporate environments with roaming profiles, and a user generally wants their global Git settings to be in a roaming profile. Only when `HOMEDRIVE`/`HOMEPATH` is either unset or does not point to a valid location, Git will fall back to using `USERPROFILE` instead. However, starting with Windows Vista, for secondary logons and services, the environment variables `HOMEDRIVE`/`HOMEPATH` point to Windows' system directory (usually `C:\Windows\system32`). That is undesirable, and that location is usually write-protected anyway. So let's verify that the `HOMEDRIVE`/`HOMEPATH` combo does not point to Windows' system directory before using it, falling back to `USERPROFILE` if it does. This fixes git-for-windows#2709 Initial-Path-by: Ivan Pozdeev Signed-off-by: Johannes Schindelin --- compat/mingw.c | 15 ++++++++++++++- 1 file changed, 14 insertions(+), 1 deletion(-) diff --git a/compat/mingw.c b/compat/mingw.c index a3e5baf32e8e87..2b18f81c38e206 100644 --- a/compat/mingw.c +++ b/compat/mingw.c @@ -2817,6 +2817,18 @@ static size_t append_system_bin_dirs(char *path, size_t size) #endif #endif +static int is_system32_path(const char *path) +{ + WCHAR system32[MAX_PATH], wpath[MAX_PATH]; + + if (xutftowcs_path(wpath, path) < 0 || + !GetSystemDirectoryW(system32, ARRAY_SIZE(system32)) || + _wcsicmp(system32, wpath)) + return 0; + + return 1; +} + static void setup_windows_environment(void) { char *tmp = getenv("TMPDIR"); @@ -2857,7 +2869,8 @@ static void setup_windows_environment(void) strbuf_addstr(&buf, tmp); if ((tmp = getenv("HOMEPATH"))) { strbuf_addstr(&buf, tmp); - if (is_directory(buf.buf)) + if (!is_system32_path(buf.buf) && + is_directory(buf.buf)) setenv("HOME", buf.buf, 1); else tmp = NULL; /* use $USERPROFILE */ From bc1a4221d4e164d5d6422ce06124847d2b06fedf Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Wed, 4 Mar 2020 21:55:28 +0100 Subject: [PATCH 109/312] http: use new "best effort" strategy for Secure Channel revoke checking MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The native Windows HTTPS backend is based on Secure Channel which lets the caller decide how to handle revocation checking problems caused by missing information in the certificate or offline CRL distribution points. Unfortunately, cURL chose to handle these problems differently than OpenSSL by default: while OpenSSL happily ignores those problems (essentially saying "¯\_(ツ)_/¯"), the Secure Channel backend will error out instead. As a remedy, the "no revoke" mode was introduced, which turns off revocation checking altogether. This is a bit heavy-handed. We support this via the `http.schannelCheckRevoke` setting. In https://github.com/curl/curl/pull/4981, we contributed an opt-in "best effort" strategy that emulates what OpenSSL seems to do. In Git for Windows, we actually want this to be the default. This patch makes it so, introducing it as a new value for the `http.schannelCheckRevoke" setting, which now becmes a tristate: it accepts the values "false", "true" or "best-effort" (defaulting to the last one). Signed-off-by: Johannes Schindelin --- Documentation/config/http.adoc | 12 +++++++----- http.c | 26 ++++++++++++++++++++++---- 2 files changed, 29 insertions(+), 9 deletions(-) diff --git a/Documentation/config/http.adoc b/Documentation/config/http.adoc index 22a8803deaf964..de52a1c44fbf14 100644 --- a/Documentation/config/http.adoc +++ b/Documentation/config/http.adoc @@ -233,11 +233,13 @@ http.sslKeyType:: http.schannelCheckRevoke:: Used to enforce or disable certificate revocation checks in cURL - when http.sslBackend is set to "schannel". Defaults to `true` if - unset. Only necessary to disable this if Git consistently errors - and the message is about checking the revocation status of a - certificate. This option is ignored if cURL lacks support for - setting the relevant SSL option at runtime. + when http.sslBackend is set to "schannel" via "true" and "false", + respectively. Another accepted value is "best-effort" (the default) + in which case revocation checks are performed, but errors due to + revocation list distribution points that are offline are silently + ignored, as well as errors due to certificates missing revocation + list distribution points. This option is ignored if cURL lacks + support for setting the relevant SSL option at runtime. http.schannelUseSSLCAInfo:: As of cURL v7.60.0, the Secure Channel backend can use the diff --git a/http.c b/http.c index 0c9a872809f872..657c3e1dc5e2c2 100644 --- a/http.c +++ b/http.c @@ -143,7 +143,13 @@ static char *cached_accept_language; static char *http_ssl_backend; -static int http_schannel_check_revoke = 1; +static int http_schannel_check_revoke_mode = +#ifdef CURLSSLOPT_REVOKE_BEST_EFFORT + CURLSSLOPT_REVOKE_BEST_EFFORT; +#else + CURLSSLOPT_NO_REVOKE; +#endif + /* * With the backend being set to `schannel`, setting sslCAinfo would override * the Certificate Store in cURL v7.60.0 and later, which is not what we want @@ -418,7 +424,19 @@ static int http_options(const char *var, const char *value, } if (!strcmp("http.schannelcheckrevoke", var)) { - http_schannel_check_revoke = git_config_bool(var, value); + if (value && !strcmp(value, "best-effort")) { + http_schannel_check_revoke_mode = +#ifdef CURLSSLOPT_REVOKE_BEST_EFFORT + CURLSSLOPT_REVOKE_BEST_EFFORT; +#else + CURLSSLOPT_NO_REVOKE; + warning(_("%s=%s unsupported by current cURL"), + var, value); +#endif + } else + http_schannel_check_revoke_mode = + (git_config_bool(var, value) ? + 0 : CURLSSLOPT_NO_REVOKE); return 0; } @@ -1044,8 +1062,8 @@ static CURL *get_curl_handle(void) #endif if (http_ssl_backend && !strcmp("schannel", http_ssl_backend) && - !http_schannel_check_revoke) { - curl_easy_setopt(result, CURLOPT_SSL_OPTIONS, CURLSSLOPT_NO_REVOKE); + http_schannel_check_revoke_mode) { + curl_easy_setopt(result, CURLOPT_SSL_OPTIONS, http_schannel_check_revoke_mode); } if (http_proactive_auth != PROACTIVE_AUTH_NONE) From 31dfabfa8400392c85526811d8c272fcb82870a2 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Sat, 9 May 2020 19:24:23 +0200 Subject: [PATCH 110/312] t5505/t5516: fix white-space around redirectors The convention in Git project's shell scripts is to have white-space _before_, but not _after_ the `>` (or `<`). Signed-off-by: Johannes Schindelin --- t/t5505-remote.sh | 6 +++--- t/t5516-fetch-push.sh | 10 +++++----- 2 files changed, 8 insertions(+), 8 deletions(-) diff --git a/t/t5505-remote.sh b/t/t5505-remote.sh index 75e2177e8ce254..a89b2883f35079 100755 --- a/t/t5505-remote.sh +++ b/t/t5505-remote.sh @@ -951,8 +951,8 @@ test_expect_success '"remote show" does not show symbolic refs' ' ( cd three && git remote show origin >output && - ! grep "^ *HEAD$" < output && - ! grep -i stale < output + ! grep "^ *HEAD$" .git/branches/origin && + echo "quux#foom" >.git/branches/origin && git remote rename origin origin && test_path_is_missing .git/branches/origin && test "$(git config remote.origin.url)" = "quux" && diff --git a/t/t5516-fetch-push.sh b/t/t5516-fetch-push.sh index 3aa7cc14c032a7..ecca661e58a44c 100755 --- a/t/t5516-fetch-push.sh +++ b/t/t5516-fetch-push.sh @@ -980,7 +980,7 @@ test_expect_success WITHOUT_BREAKING_CHANGES 'fetch with branches' ' git branch second $the_first_commit && git checkout second && mkdir -p testrepo/.git/branches && - echo ".." > testrepo/.git/branches/branch1 && + echo ".." >testrepo/.git/branches/branch1 && ( cd testrepo && git fetch branch1 && @@ -994,7 +994,7 @@ test_expect_success WITHOUT_BREAKING_CHANGES 'fetch with branches' ' test_expect_success WITHOUT_BREAKING_CHANGES 'fetch with branches containing #' ' mk_empty testrepo && mkdir -p testrepo/.git/branches && - echo "..#second" > testrepo/.git/branches/branch2 && + echo "..#second" >testrepo/.git/branches/branch2 && ( cd testrepo && git fetch branch2 && @@ -1011,7 +1011,7 @@ test_expect_success WITHOUT_BREAKING_CHANGES 'push with branches' ' test_when_finished "rm -rf .git/branches" && mkdir -p .git/branches && - echo "testrepo" > .git/branches/branch1 && + echo "testrepo" >.git/branches/branch1 && git push branch1 && ( @@ -1027,7 +1027,7 @@ test_expect_success WITHOUT_BREAKING_CHANGES 'push with branches containing #' ' test_when_finished "rm -rf .git/branches" && mkdir -p .git/branches && - echo "testrepo#branch3" > .git/branches/branch2 && + echo "testrepo#branch3" >.git/branches/branch2 && git push branch2 && ( @@ -1555,7 +1555,7 @@ EOF git init no-thin && git --git-dir=no-thin/.git config receive.unpacklimit 0 && git push no-thin/.git refs/heads/main:refs/heads/foo && - echo modified >> path1 && + echo modified >>path1 && git commit -am modified && git repack -adf && rcvpck="git receive-pack --reject-thin-pack-for-testing" && From 18ee827c236ae1834c34997d8d1546361c51922c Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Sat, 12 Sep 2015 12:25:47 +0200 Subject: [PATCH 111/312] t3701: verify that we can add *lots* of files interactively Signed-off-by: Johannes Schindelin --- t/t3701-add-interactive.sh | 21 +++++++++++++++++++++ 1 file changed, 21 insertions(+) diff --git a/t/t3701-add-interactive.sh b/t/t3701-add-interactive.sh index b8a05d95f3f130..81d0487b3b90eb 100755 --- a/t/t3701-add-interactive.sh +++ b/t/t3701-add-interactive.sh @@ -1166,6 +1166,27 @@ test_expect_success 'checkout -p patch editing of added file' ' ) ' +test_expect_success EXPENSIVE 'add -i with a lot of files' ' + git reset --hard && + x160=0123456789012345678901234567890123456789 && + x160=$x160$x160$x160$x160 && + y= && + i=0 && + while test $i -le 200 + do + name=$(printf "%s%03d" $x160 $i) && + echo $name >$name && + git add -N $name && + y="${y}y$LF" && + i=$(($i+1)) || + exit 1 + done && + echo "$y" | git add -p -- . && + git diff --cached >staged && + test_line_count = 1407 staged && + git reset --hard +' + test_expect_success 'show help from add--helper' ' git reset --hard && cat >expect <<-EOF && From efefd04bf52cff8577a8bee5a5f0a8d1c37deba1 Mon Sep 17 00:00:00 2001 From: Jeff Hostetler Date: Tue, 30 Mar 2021 14:25:31 -0400 Subject: [PATCH 112/312] clink.pl: fix libexpatd.lib link error when using MSVC When building with `make MSVC=1 DEBUG=1`, link to `libexpatd.lib` rather than `libexpat.lib`. It appears that the `vcpkg` package for "libexpat" has changed and now creates `libexpatd.lib` for debug mode builds. Previously, both debug and release builds created a ".lib" with the same basename. Signed-off-by: Jeff Hostetler --- compat/vcbuild/scripts/clink.pl | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/compat/vcbuild/scripts/clink.pl b/compat/vcbuild/scripts/clink.pl index 3bd824154be381..2768ae15f1879f 100755 --- a/compat/vcbuild/scripts/clink.pl +++ b/compat/vcbuild/scripts/clink.pl @@ -66,7 +66,11 @@ } push(@args, $lib); } elsif ("$arg" eq "-lexpat") { + if ($is_debug) { + push(@args, "libexpatd.lib"); + } else { push(@args, "libexpat.lib"); + } } elsif ("$arg" =~ /^-L/ && "$arg" ne "-LTCG") { $arg =~ s/^-L/-LIBPATH:/; push(@lflags, $arg); From 94d1bf61b0e202aad9c838487ec84ba816b993c9 Mon Sep 17 00:00:00 2001 From: Jeff Hostetler Date: Mon, 5 Apr 2021 15:27:38 -0400 Subject: [PATCH 113/312] Makefile: clean up .ilk files when MSVC=1 Signed-off-by: Jeff Hostetler --- Makefile | 3 +++ 1 file changed, 3 insertions(+) diff --git a/Makefile b/Makefile index 7315507381e146..e77a81be291f35 100644 --- a/Makefile +++ b/Makefile @@ -3781,12 +3781,15 @@ ifdef MSVC $(RM) $(patsubst %.o,%.o.pdb,$(OBJECTS)) $(RM) headless-git.o.pdb $(RM) $(patsubst %.exe,%.pdb,$(OTHER_PROGRAMS)) + $(RM) $(patsubst %.exe,%.ilk,$(OTHER_PROGRAMS)) $(RM) $(patsubst %.exe,%.iobj,$(OTHER_PROGRAMS)) $(RM) $(patsubst %.exe,%.ipdb,$(OTHER_PROGRAMS)) $(RM) $(patsubst %.exe,%.pdb,$(PROGRAMS)) + $(RM) $(patsubst %.exe,%.ilk,$(PROGRAMS)) $(RM) $(patsubst %.exe,%.iobj,$(PROGRAMS)) $(RM) $(patsubst %.exe,%.ipdb,$(PROGRAMS)) $(RM) $(patsubst %.exe,%.pdb,$(TEST_PROGRAMS)) + $(RM) $(patsubst %.exe,%.ilk,$(TEST_PROGRAMS)) $(RM) $(patsubst %.exe,%.iobj,$(TEST_PROGRAMS)) $(RM) $(patsubst %.exe,%.ipdb,$(TEST_PROGRAMS)) $(RM) compat/vcbuild/MSVC-DEFS-GEN From cc9b967bbb9f3b20a79b0ecb4298339bc530c802 Mon Sep 17 00:00:00 2001 From: Jeff Hostetler Date: Mon, 5 Apr 2021 14:08:22 -0400 Subject: [PATCH 114/312] vcbuild: add support for compiling Windows resource files Create a wrapper for the Windows Resource Compiler (RC.EXE) for use by the MSVC=1 builds. This is similar to the CL.EXE and LIB.EXE wrappers used for the MSVC=1 builds. Signed-off-by: Jeff Hostetler --- compat/vcbuild/find_vs_env.bat | 7 ++++++ compat/vcbuild/scripts/rc.pl | 46 ++++++++++++++++++++++++++++++++++ config.mak.uname | 3 ++- 3 files changed, 55 insertions(+), 1 deletion(-) create mode 100644 compat/vcbuild/scripts/rc.pl diff --git a/compat/vcbuild/find_vs_env.bat b/compat/vcbuild/find_vs_env.bat index b35d264c0e6bed..379b16296e09c2 100644 --- a/compat/vcbuild/find_vs_env.bat +++ b/compat/vcbuild/find_vs_env.bat @@ -99,6 +99,7 @@ REM ================================================================ SET sdk_dir=%WindowsSdkDir% SET sdk_ver=%WindowsSDKVersion% + SET sdk_ver_bin_dir=%WindowsSdkVerBinPath%%tgt% SET si=%sdk_dir%Include\%sdk_ver% SET sdk_includes=-I"%si%ucrt" -I"%si%um" -I"%si%shared" SET sl=%sdk_dir%lib\%sdk_ver% @@ -130,6 +131,7 @@ REM ================================================================ SET sdk_dir=%WindowsSdkDir% SET sdk_ver=%WindowsSDKVersion% + SET sdk_ver_bin_dir=%WindowsSdkVerBinPath%bin\amd64 SET si=%sdk_dir%Include\%sdk_ver% SET sdk_includes=-I"%si%ucrt" -I"%si%um" -I"%si%shared" -I"%si%winrt" SET sl=%sdk_dir%lib\%sdk_ver% @@ -160,6 +162,11 @@ REM ================================================================ echo msvc_includes=%msvc_includes% echo msvc_libs=%msvc_libs% + echo sdk_ver_bin_dir=%sdk_ver_bin_dir% + SET X1=%sdk_ver_bin_dir:C:=/C% + SET X2=%X1:\=/% + echo sdk_ver_bin_dir_msys=%X2% + echo sdk_includes=%sdk_includes% echo sdk_libs=%sdk_libs% diff --git a/compat/vcbuild/scripts/rc.pl b/compat/vcbuild/scripts/rc.pl new file mode 100644 index 00000000000000..7bca4cd81c6c63 --- /dev/null +++ b/compat/vcbuild/scripts/rc.pl @@ -0,0 +1,46 @@ +#!/usr/bin/perl -w +###################################################################### +# Compile Resources on Windows +# +# This is a wrapper to facilitate the compilation of Git with MSVC +# using GNU Make as the build system. So, instead of manipulating the +# Makefile into something nasty, just to support non-space arguments +# etc, we use this wrapper to fix the command line options +# +###################################################################### +use strict; +my @args = (); +my @input = (); + +while (@ARGV) { + my $arg = shift @ARGV; + if ("$arg" =~ /^-[dD]/) { + # GIT_VERSION gets passed with too many + # layers of dquote escaping. + $arg =~ s/\\"/"/g; + + push(@args, $arg); + + } elsif ("$arg" eq "-i") { + my $arg = shift @ARGV; + # TODO complain if NULL or is dashed ?? + push(@input, $arg); + + } elsif ("$arg" eq "-o") { + my $arg = shift @ARGV; + # TODO complain if NULL or is dashed ?? + push(@args, "-fo$arg"); + + } else { + push(@args, $arg); + } +} + +push(@args, "-nologo"); +push(@args, "-v"); +push(@args, @input); + +unshift(@args, "rc.exe"); +printf("**** @args\n"); + +exit (system(@args) != 0); diff --git a/config.mak.uname b/config.mak.uname index 8e30da5c411715..8e0ed259ab715b 100644 --- a/config.mak.uname +++ b/config.mak.uname @@ -440,7 +440,7 @@ ifeq ($(uname_S),Windows) # link.exe next to, and required by, cl.exe, we have to prepend this # onto the existing $PATH. # - SANE_TOOL_PATH ?= $(msvc_bin_dir_msys) + SANE_TOOL_PATH ?= $(msvc_bin_dir_msys):$(sdk_ver_bin_dir_msys) HAVE_ALLOCA_H = YesPlease NO_PREAD = YesPlease NEEDS_CRYPTO_WITH_SSL = YesPlease @@ -508,6 +508,7 @@ endif # See https://msdn.microsoft.com/en-us/library/ms235330.aspx EXTLIBS = user32.lib advapi32.lib shell32.lib wininet.lib ws2_32.lib invalidcontinue.obj kernel32.lib ntdll.lib PTHREAD_LIBS = + RC = compat/vcbuild/scripts/rc.pl lib = BASIC_CFLAGS += $(vcpkg_inc) $(sdk_includes) $(msvc_includes) ifndef DEBUG From 0c8c60283c415936e1a2a9046f9db564b5dc5ce7 Mon Sep 17 00:00:00 2001 From: Jeff Hostetler Date: Mon, 5 Apr 2021 14:12:14 -0400 Subject: [PATCH 115/312] config.mak.uname: add git.rc to MSVC builds Teach MSVC=1 builds to depend on the `git.rc` file so that the resulting executables have Windows-style resources and version number information within them. Signed-off-by: Jeff Hostetler --- config.mak.uname | 1 + 1 file changed, 1 insertion(+) diff --git a/config.mak.uname b/config.mak.uname index 8e0ed259ab715b..8ef58660aa3ecf 100644 --- a/config.mak.uname +++ b/config.mak.uname @@ -507,6 +507,7 @@ endif # handle twice, or to access the osfhandle of an already-closed stdout # See https://msdn.microsoft.com/en-us/library/ms235330.aspx EXTLIBS = user32.lib advapi32.lib shell32.lib wininet.lib ws2_32.lib invalidcontinue.obj kernel32.lib ntdll.lib + GITLIBS += git.res PTHREAD_LIBS = RC = compat/vcbuild/scripts/rc.pl lib = From a8efd0b086df97c27b1635490b5d408522f257c3 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Thu, 2 Jul 2020 16:35:05 +0200 Subject: [PATCH 116/312] git add -i: handle CR/LF line endings in the interactive input As of Git for Windows v2.27.0, there is an option to use Windows' newly-introduced Pseudo Console support. When running an interactive add operation with this support enabled, Git will receive CR/LF line endings. Therefore, let's not pretend that we are expecting Unix line endings. This fixes https://github.com/git-for-windows/git/issues/2729 Signed-off-by: Johannes Schindelin --- prompt.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/prompt.c b/prompt.c index f21c5bf1c7e42f..a07b135050b9c2 100644 --- a/prompt.c +++ b/prompt.c @@ -80,7 +80,7 @@ int git_read_line_interactively(struct strbuf *line) int ret; fflush(stdout); - ret = strbuf_getline_lf(line, stdin); + ret = strbuf_getline(line, stdin); if (ret != EOF) strbuf_trim_trailing_newline(line); From d246e80a602fde26701beabebeb93c760a28703f Mon Sep 17 00:00:00 2001 From: Luke Bonanomi Date: Wed, 24 Jun 2020 07:45:52 -0400 Subject: [PATCH 117/312] commit: accept "scissors" with CR/LF line endings This change enhances `git commit --cleanup=scissors` by detecting scissors lines ending in either LF (UNIX-style) or CR/LF (DOS-style). Regression tests are included to specifically test for trailing comments after a CR/LF-terminated scissors line. Signed-off-by: Luke Bonanomi Signed-off-by: Johannes Schindelin --- t/t7502-commit-porcelain.sh | 42 +++++++++++++++++++++++++++++++++++++ wt-status.c | 13 +++++++++--- 2 files changed, 52 insertions(+), 3 deletions(-) diff --git a/t/t7502-commit-porcelain.sh b/t/t7502-commit-porcelain.sh index b37e2018a74a7b..c38b96b66cd20a 100755 --- a/t/t7502-commit-porcelain.sh +++ b/t/t7502-commit-porcelain.sh @@ -623,6 +623,48 @@ test_expect_success 'cleanup commit messages (scissors option,-F,-e, scissors on test_must_be_empty actual ' +test_expect_success 'helper-editor' ' + + write_script lf-to-crlf.sh <<-\EOF + sed "s/\$/Q/" <"$1" | tr Q "\\015" >"$1".new && + mv -f "$1".new "$1" + EOF +' + +test_expect_success 'cleanup commit messages (scissors option,-F,-e, CR/LF line endings)' ' + + test_config core.editor "\"$PWD/lf-to-crlf.sh\"" && + scissors="# ------------------------ >8 ------------------------" && + + test_write_lines >text \ + "# Keep this comment" "" " $scissors" \ + "# Keep this comment, too" "$scissors" \ + "# Remove this comment" "$scissors" \ + "Remove this comment, too" && + + test_write_lines >expect \ + "# Keep this comment" "" " $scissors" \ + "# Keep this comment, too" && + + git commit --cleanup=scissors -e -F text --allow-empty && + git cat-file -p HEAD >raw && + sed -e "1,/^\$/d" raw >actual && + test_cmp expect actual +' + +test_expect_success 'cleanup commit messages (scissors option,-F,-e, scissors on first line, CR/LF line endings)' ' + + scissors="# ------------------------ >8 ------------------------" && + test_write_lines >text \ + "$scissors" \ + "# Remove this comment and any following lines" && + cp text /tmp/test2-text && + git commit --cleanup=scissors -e -F text --allow-empty --allow-empty-message && + git cat-file -p HEAD >raw && + sed -e "1,/^\$/d" raw >actual && + test_must_be_empty actual +' + test_expect_success 'cleanup commit messages (strip option,-F)' ' echo >>negative && diff --git a/wt-status.c b/wt-status.c index 1da5732f57b115..50bace8378bd15 100644 --- a/wt-status.c +++ b/wt-status.c @@ -40,7 +40,7 @@ #define UF_DELAY_WARNING_IN_MS (2 * 1000) static const char cut_line[] = -"------------------------ >8 ------------------------\n"; +"------------------------ >8 ------------------------"; static char default_wt_status_colors[][COLOR_MAXLEN] = { GIT_COLOR_NORMAL, /* WT_STATUS_HEADER */ @@ -1096,15 +1096,22 @@ static void wt_longstatus_print_other(struct wt_status *s, status_printf_ln(s, GIT_COLOR_NORMAL, "%s", ""); } +static inline int starts_with_newline(const char *p) +{ + return *p == '\n' || (*p == '\r' && p[1] == '\n'); +} + size_t wt_status_locate_end(const char *s, size_t len) { const char *p; struct strbuf pattern = STRBUF_INIT; strbuf_addf(&pattern, "\n%s %s", comment_line_str, cut_line); - if (starts_with(s, pattern.buf + 1)) + if (starts_with(s, pattern.buf + 1) && + starts_with_newline(s + pattern.len - 1)) len = 0; - else if ((p = strstr(s, pattern.buf))) { + else if ((p = strstr(s, pattern.buf)) && + starts_with_newline(p + pattern.len)) { size_t newlen = p - s + 1; if (newlen < len) len = newlen; From deed3892f77dc04cff27f8f98443b714ebc4b35c Mon Sep 17 00:00:00 2001 From: Jens Glathe Date: Tue, 2 Jun 2020 12:12:25 +0200 Subject: [PATCH 118/312] t0014: fix indentation For some reason, this test case was indented with 4 spaces instead of 1 horizontal tab. The other test cases in the same test script are fine. Signed-off-by: Jens Glathe Signed-off-by: Johannes Schindelin --- t/t0014-alias.sh | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/t/t0014-alias.sh b/t/t0014-alias.sh index 854d59ec58c25a..30708146887d19 100755 --- a/t/t0014-alias.sh +++ b/t/t0014-alias.sh @@ -38,10 +38,10 @@ test_expect_success 'looping aliases - internal execution' ' #' test_expect_success 'run-command formats empty args properly' ' - test_must_fail env GIT_TRACE=1 git frotz a "" b " " c 2>actual.raw && - sed -ne "/run_command:/s/.*trace: run_command: //p" actual.raw >actual && - echo "git-frotz a '\'''\'' b '\'' '\'' c" >expect && - test_cmp expect actual + test_must_fail env GIT_TRACE=1 git frotz a "" b " " c 2>actual.raw && + sed -ne "/run_command:/s/.*trace: run_command: //p" actual.raw >actual && + echo "git-frotz a '\'''\'' b '\'' '\'' c" >expect && + test_cmp expect actual ' test_expect_success 'tracing a shell alias with arguments shows trace of prepared command' ' From 79097ecf381028db5020324fd857e5d0434196b6 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Wed, 12 Aug 2020 15:06:17 +0000 Subject: [PATCH 119/312] git-gui: accommodate for intent-to-add files As of Git v2.28.0, the diff for files staged via `git add -N` marks them as new files. Git GUI was ill-prepared for that, and this patch teaches Git GUI about them. Please note that this will not even fix things with v2.28.0, as the `rp/apply-cached-with-i-t-a` patches are required on Git's side, too. This fixes https://github.com/git-for-windows/git/issues/2779 Signed-off-by: Johannes Schindelin Signed-off-by: Pratyush Yadav --- git-gui/git-gui.sh | 2 ++ git-gui/lib/diff.tcl | 12 ++++++++---- 2 files changed, 10 insertions(+), 4 deletions(-) diff --git a/git-gui/git-gui.sh b/git-gui/git-gui.sh index 887d6d596c16fc..3390dc15ccaffe 100755 --- a/git-gui/git-gui.sh +++ b/git-gui/git-gui.sh @@ -2078,6 +2078,7 @@ set all_icons(U$ui_index) file_merge set all_icons(T$ui_index) file_statechange set all_icons(_$ui_workdir) file_plain +set all_icons(A$ui_workdir) file_plain set all_icons(M$ui_workdir) file_mod set all_icons(D$ui_workdir) file_question set all_icons(U$ui_workdir) file_merge @@ -2104,6 +2105,7 @@ foreach i { {A_ {mc "Staged for commit"}} {AM {mc "Portions staged for commit"}} {AD {mc "Staged for commit, missing"}} + {AA {mc "Intended to be added"}} {_D {mc "Missing"}} {D_ {mc "Staged for removal"}} diff --git a/git-gui/lib/diff.tcl b/git-gui/lib/diff.tcl index d657bfec05b498..9a593a3f6e1547 100644 --- a/git-gui/lib/diff.tcl +++ b/git-gui/lib/diff.tcl @@ -568,7 +568,8 @@ proc apply_or_revert_hunk {x y revert} { if {$current_diff_side eq $ui_index} { set failed_msg [mc "Failed to unstage selected hunk."] lappend apply_cmd --reverse --cached - if {[string index $mi 0] ne {M}} { + set file_state [string index $mi 0] + if {$file_state ne {M} && $file_state ne {A}} { unlock_index return } @@ -581,7 +582,8 @@ proc apply_or_revert_hunk {x y revert} { lappend apply_cmd --cached } - if {[string index $mi 1] ne {M}} { + set file_state [string index $mi 1] + if {$file_state ne {M} && $file_state ne {A}} { unlock_index return } @@ -673,7 +675,8 @@ proc apply_or_revert_range_or_line {x y revert} { set failed_msg [mc "Failed to unstage selected line."] set to_context {+} lappend apply_cmd --reverse --cached - if {[string index $mi 0] ne {M}} { + set file_state [string index $mi 0] + if {$file_state ne {M} && $file_state ne {A}} { unlock_index return } @@ -688,7 +691,8 @@ proc apply_or_revert_range_or_line {x y revert} { lappend apply_cmd --cached } - if {[string index $mi 1] ne {M}} { + set file_state [string index $mi 1] + if {$file_state ne {M} && $file_state ne {A}} { unlock_index return } From c0e81c0a570321aaa5b2d3532fbd1feccffd4328 Mon Sep 17 00:00:00 2001 From: Ian Bearman Date: Fri, 31 Jan 2020 16:00:25 -0800 Subject: [PATCH 120/312] vcbuild: install ARM64 dependencies when building ARM64 binaries Co-authored-by: Dennis Ameling Signed-off-by: Ian Bearman Signed-off-by: Dennis Ameling Signed-off-by: Johannes Schindelin --- compat/vcbuild/README | 6 +++++- compat/vcbuild/vcpkg_copy_dlls.bat | 7 ++++++- compat/vcbuild/vcpkg_install.bat | 9 +++++++-- 3 files changed, 18 insertions(+), 4 deletions(-) diff --git a/compat/vcbuild/README b/compat/vcbuild/README index 29ec1d0f104b80..1df1cabb1ebbbd 100644 --- a/compat/vcbuild/README +++ b/compat/vcbuild/README @@ -6,7 +6,11 @@ The Steps to Build Git with VS2015 or VS2017 from the command line. Prompt or from an SDK bash window: $ cd - $ ./compat/vcbuild/vcpkg_install.bat + $ ./compat/vcbuild/vcpkg_install.bat x64-windows + + or + + $ ./compat/vcbuild/vcpkg_install.bat arm64-windows The vcpkg tools and all of the third-party sources will be installed in this folder: diff --git a/compat/vcbuild/vcpkg_copy_dlls.bat b/compat/vcbuild/vcpkg_copy_dlls.bat index 13661c14f8705c..8bea0cbf83b6cf 100644 --- a/compat/vcbuild/vcpkg_copy_dlls.bat +++ b/compat/vcbuild/vcpkg_copy_dlls.bat @@ -15,7 +15,12 @@ REM ================================================================ @FOR /F "delims=" %%D IN ("%~dp0") DO @SET cwd=%%~fD cd %cwd% - SET arch=x64-windows + SET arch=%2 + IF NOT DEFINED arch ( + echo defaulting to 'x64-windows`. Invoke %0 with 'x86-windows', 'x64-windows', or 'arm64-windows' + set arch=x64-windows + ) + SET inst=%cwd%vcpkg\installed\%arch% IF [%1]==[release] ( diff --git a/compat/vcbuild/vcpkg_install.bat b/compat/vcbuild/vcpkg_install.bat index 8330d8120fb511..cacef18c11dc79 100644 --- a/compat/vcbuild/vcpkg_install.bat +++ b/compat/vcbuild/vcpkg_install.bat @@ -31,6 +31,12 @@ REM ================================================================ SETLOCAL EnableDelayedExpansion + SET arch=%1 + IF NOT DEFINED arch ( + echo defaulting to 'x64-windows`. Invoke %0 with 'x86-windows', 'x64-windows', or 'arm64-windows' + set arch=x64-windows + ) + @FOR /F "delims=" %%D IN ("%~dp0") DO @SET cwd=%%~fD cd %cwd% @@ -55,9 +61,8 @@ REM ================================================================ echo Successfully installed %cwd%vcpkg\vcpkg.exe :install_libraries - SET arch=x64-windows - echo Installing third-party libraries... + echo Installing third-party libraries(%arch%)... FOR %%i IN (zlib expat libiconv openssl libssh2 curl) DO ( cd %cwd%vcpkg IF NOT EXIST "packages\%%i_%arch%" CALL :sub__install_one %%i From b73460883afcddbd4ed76d03e215f51e44d46ad0 Mon Sep 17 00:00:00 2001 From: Ian Bearman Date: Tue, 4 Feb 2020 10:34:40 -0800 Subject: [PATCH 121/312] vcbuild: add an option to install individual 'features' In this context, a "feature" is a dependency combined with its own dependencies. Signed-off-by: Ian Bearman Signed-off-by: Johannes Schindelin --- compat/vcbuild/vcpkg_install.bat | 35 +++++++++++++++++++++++++++++++- 1 file changed, 34 insertions(+), 1 deletion(-) diff --git a/compat/vcbuild/vcpkg_install.bat b/compat/vcbuild/vcpkg_install.bat index cacef18c11dc79..8da212487ae97d 100644 --- a/compat/vcbuild/vcpkg_install.bat +++ b/compat/vcbuild/vcpkg_install.bat @@ -85,14 +85,47 @@ REM ================================================================ :sub__install_one echo Installing package %1... + call :%1_features + REM vcpkg may not be reliable on slow, intermittent or proxy REM connections, see e.g. REM https://social.msdn.microsoft.com/Forums/windowsdesktop/en-US/4a8f7be5-5e15-4213-a7bb-ddf424a954e6/winhttpsendrequest-ends-with-12002-errorhttptimeout-after-21-seconds-no-matter-what-timeout?forum=windowssdk REM which explains the hidden 21 second timeout REM (last post by Dave : Microsoft - Windows Networking team) - .\vcpkg.exe install %1:%arch% + .\vcpkg.exe install %1%features%:%arch% IF ERRORLEVEL 1 ( EXIT /B 1 ) echo Finished %1 goto :EOF + +:: +:: features for each vcpkg to install +:: there should be an entry here for each package to install +:: 'set features=' means use the default otherwise +:: 'set features=[comma-delimited-feature-set]' is the syntax +:: + +:zlib_features +set features= +goto :EOF + +:expat_features +set features= +goto :EOF + +:libiconv_features +set features= +goto :EOF + +:openssl_features +set features= +goto :EOF + +:libssh2_features +set features= +goto :EOF + +:curl_features +set features=[core,openssl] +goto :EOF From 3efbb5050a4ef7cc15f475acf715faee9c6ace41 Mon Sep 17 00:00:00 2001 From: Dennis Ameling Date: Fri, 4 Dec 2020 14:11:34 +0100 Subject: [PATCH 122/312] cmake: allow building for Windows/ARM64 Signed-off-by: Dennis Ameling Signed-off-by: Johannes Schindelin --- contrib/buildsystems/CMakeLists.txt | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/contrib/buildsystems/CMakeLists.txt b/contrib/buildsystems/CMakeLists.txt index 25b495fa737d07..aaaee1bc4b7943 100644 --- a/contrib/buildsystems/CMakeLists.txt +++ b/contrib/buildsystems/CMakeLists.txt @@ -65,9 +65,9 @@ if(USE_VCPKG) set(VCPKG_DIR "${CMAKE_SOURCE_DIR}/compat/vcbuild/vcpkg") if(NOT EXISTS ${VCPKG_DIR}) message("Initializing vcpkg and building the Git's dependencies (this will take a while...)") - execute_process(COMMAND ${CMAKE_SOURCE_DIR}/compat/vcbuild/vcpkg_install.bat) + execute_process(COMMAND ${CMAKE_SOURCE_DIR}/compat/vcbuild/vcpkg_install.bat ${VCPKG_ARCH}) endif() - list(APPEND CMAKE_PREFIX_PATH "${VCPKG_DIR}/installed/x64-windows") + list(APPEND CMAKE_PREFIX_PATH "${VCPKG_DIR}/installed/${VCPKG_ARCH}") # In the vcpkg edition, we need this to be able to link to libcurl set(CURL_NO_CURL_CMAKE ON) @@ -1211,7 +1211,7 @@ string(REPLACE "@USE_LIBPCRE2@" "" git_build_options "${git_build_options}") string(REPLACE "@WITH_BREAKING_CHANGES@" "" git_build_options "${git_build_options}") string(REPLACE "@X@" "${EXE_EXTENSION}" git_build_options "${git_build_options}") if(USE_VCPKG) - string(APPEND git_build_options "PATH=\"$PATH:$TEST_DIRECTORY/../compat/vcbuild/vcpkg/installed/x64-windows/bin\"\n") + string(APPEND git_build_options "PATH=\"$PATH:$TEST_DIRECTORY/../compat/vcbuild/vcpkg/installed/${VCPKG_ARCH}/bin\"\n") endif() file(WRITE ${CMAKE_BINARY_DIR}/GIT-BUILD-OPTIONS ${git_build_options}) From 0911d8c3aa079e2044d8cc756fe8dee07cee380d Mon Sep 17 00:00:00 2001 From: Jeff Hostetler Date: Mon, 5 Apr 2021 14:24:52 -0400 Subject: [PATCH 123/312] clink.pl: ignore no-stack-protector arg on MSVC=1 builds Ignore the `-fno-stack-protector` compiler argument when building with MSVC. This will be used in a later commit that needs to build a Win32 GUI app. Signed-off-by: Jeff Hostetler --- compat/vcbuild/scripts/clink.pl | 2 ++ 1 file changed, 2 insertions(+) diff --git a/compat/vcbuild/scripts/clink.pl b/compat/vcbuild/scripts/clink.pl index 2768ae15f1879f..73c8a2b184f38b 100755 --- a/compat/vcbuild/scripts/clink.pl +++ b/compat/vcbuild/scripts/clink.pl @@ -122,6 +122,8 @@ push(@cflags, "-wd4996"); } elsif ("$arg" =~ /^-W[a-z]/) { # let's ignore those + } elsif ("$arg" eq "-fno-stack-protector") { + # eat this } else { push(@args, $arg); } From e3e58b69f2fa195955e00e9ad162ca1149eb7e47 Mon Sep 17 00:00:00 2001 From: Philip Oakley Date: Sun, 6 Oct 2019 18:40:55 +0100 Subject: [PATCH 124/312] vcpkg_install: detect lack of Git The vcpkg_install batch file depends on the availability of a working Git on the CMD path. This may not be present if the user has selected the 'bash only' option during Git-for-Windows install. Detect and tell the user about their lack of a working Git in the CMD window. Fixes #2348. A separate PR https://github.com/git-for-windows/build-extra/pull/258 now highlights the recommended path setting during install. Signed-off-by: Philip Oakley --- compat/vcbuild/vcpkg_install.bat | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/compat/vcbuild/vcpkg_install.bat b/compat/vcbuild/vcpkg_install.bat index ebd0bad242a8ca..bcbbf536af3141 100644 --- a/compat/vcbuild/vcpkg_install.bat +++ b/compat/vcbuild/vcpkg_install.bat @@ -36,6 +36,13 @@ REM ================================================================ dir vcpkg\vcpkg.exe >nul 2>nul && GOTO :install_libraries + git.exe version 2>nul + IF ERRORLEVEL 1 ( + echo "***" + echo "Git not found. Please adjust your CMD path or Git install option." + echo "***" + EXIT /B 1 ) + echo Fetching vcpkg in %cwd%vcpkg git.exe clone https://github.com/Microsoft/vcpkg vcpkg IF ERRORLEVEL 1 ( EXIT /B 1 ) From 52b0a9edeee5f5394e0153b2657388fa0deaa32b Mon Sep 17 00:00:00 2001 From: Dennis Ameling Date: Sun, 29 Nov 2020 00:12:26 +0100 Subject: [PATCH 125/312] ci(vs-build) also build Windows/ARM64 artifacts There are no Windows/ARM64 agents in GitHub Actions yet, therefore we just skip adjusting the `vs-test` job for now. Signed-off-by: Dennis Ameling Signed-off-by: Johannes Schindelin --- .github/workflows/main.yml | 17 ++++++++++------- 1 file changed, 10 insertions(+), 7 deletions(-) diff --git a/.github/workflows/main.yml b/.github/workflows/main.yml index 9959b61ece2bcc..6ee3c5b42b85c4 100644 --- a/.github/workflows/main.yml +++ b/.github/workflows/main.yml @@ -169,8 +169,11 @@ jobs: NO_PERL: 1 GIT_CONFIG_PARAMETERS: "'user.name=CI' 'user.email=ci@git'" runs-on: windows-latest + strategy: + matrix: + arch: [x64, arm64] concurrency: - group: vs-build-${{ github.ref }} + group: vs-build-${{ github.ref }}-${{ matrix.arch }} cancel-in-progress: ${{ needs.ci-config.outputs.skip_concurrent == 'yes' }} steps: - uses: actions/checkout@v4 @@ -189,14 +192,14 @@ jobs: uses: microsoft/setup-msbuild@v2 - name: copy dlls to root shell: cmd - run: compat\vcbuild\vcpkg_copy_dlls.bat release + run: compat\vcbuild\vcpkg_copy_dlls.bat release ${{ matrix.arch }}-windows - name: generate Visual Studio solution shell: bash run: | - cmake `pwd`/contrib/buildsystems/ -DCMAKE_PREFIX_PATH=`pwd`/compat/vcbuild/vcpkg/installed/x64-windows \ - -DNO_GETTEXT=YesPlease -DPERL_TESTS=OFF -DPYTHON_TESTS=OFF -DCURL_NO_CURL_CMAKE=ON + cmake `pwd`/contrib/buildsystems/ -DCMAKE_PREFIX_PATH=`pwd`/compat/vcbuild/vcpkg/installed/${{ matrix.arch }}-windows \ + -DNO_GETTEXT=YesPlease -DPERL_TESTS=OFF -DPYTHON_TESTS=OFF -DCURL_NO_CURL_CMAKE=ON -DCMAKE_GENERATOR_PLATFORM=${{ matrix.arch }} -DVCPKG_ARCH=${{ matrix.arch }}-windows - name: MSBuild - run: msbuild git.sln -property:Configuration=Release -property:Platform=x64 -maxCpuCount:4 -property:PlatformToolset=v142 + run: msbuild git.sln -property:Configuration=Release -property:Platform=${{ matrix.arch }} -maxCpuCount:4 -property:PlatformToolset=v142 - name: bundle artifact tar shell: bash env: @@ -210,7 +213,7 @@ jobs: - name: upload tracked files and build artifacts uses: actions/upload-artifact@v4 with: - name: vs-artifacts + name: vs-artifacts-${{ matrix.arch }} path: artifacts vs-test: name: win+VS test @@ -228,7 +231,7 @@ jobs: - name: download tracked files and build artifacts uses: actions/download-artifact@v4 with: - name: vs-artifacts + name: vs-artifacts-x64 path: ${{github.workspace}} - name: extract tracked files and build artifacts shell: bash From 3c5d4e3d6bd199edd884bfb3ae6c3d94d06ad28d Mon Sep 17 00:00:00 2001 From: Jeff Hostetler Date: Mon, 5 Apr 2021 14:39:33 -0400 Subject: [PATCH 126/312] clink.pl: move default linker options for MSVC=1 builds Move the default `-ENTRY` and `-SUBSYSTEM` arguments for MSVC=1 builds from `config.mak.uname` into `clink.pl`. These args are constant for console-mode executables. Add support to `clink.pl` for generating a Win32 GUI application using the `-mwindows` argument (to match how GCC does it). This changes the `-ENTRY` and `-SUBSYSTEM` arguments accordingly. Signed-off-by: Jeff Hostetler --- compat/vcbuild/scripts/clink.pl | 11 +++++++++++ config.mak.uname | 2 +- 2 files changed, 12 insertions(+), 1 deletion(-) diff --git a/compat/vcbuild/scripts/clink.pl b/compat/vcbuild/scripts/clink.pl index 73c8a2b184f38b..a38b360015ece9 100755 --- a/compat/vcbuild/scripts/clink.pl +++ b/compat/vcbuild/scripts/clink.pl @@ -15,6 +15,7 @@ my @lflags = (); my $is_linking = 0; my $is_debug = 0; +my $is_gui = 0; while (@ARGV) { my $arg = shift @ARGV; if ("$arg" eq "-DDEBUG") { @@ -124,11 +125,21 @@ # let's ignore those } elsif ("$arg" eq "-fno-stack-protector") { # eat this + } elsif ("$arg" eq "-mwindows") { + $is_gui = 1; } else { push(@args, $arg); } } if ($is_linking) { + if ($is_gui) { + push(@args, "-ENTRY:wWinMainCRTStartup"); + push(@args, "-SUBSYSTEM:WINDOWS"); + } else { + push(@args, "-ENTRY:wmainCRTStartup"); + push(@args, "-SUBSYSTEM:CONSOLE"); + } + push(@args, @lflags); unshift(@args, "link.exe"); } else { diff --git a/config.mak.uname b/config.mak.uname index 8ef58660aa3ecf..8839dc511c426f 100644 --- a/config.mak.uname +++ b/config.mak.uname @@ -502,7 +502,7 @@ endif compat/win32/trace2_win32_process_info.o \ compat/win32/dirent.o COMPAT_CFLAGS = -D__USE_MINGW_ACCESS -DDETECT_MSYS_TTY -DENSURE_MSYSTEM_IS_SET -DNOGDI -DHAVE_STRING_H -Icompat -Icompat/regex -Icompat/win32 -DSTRIP_EXTENSION=\".exe\" - BASIC_LDFLAGS = -IGNORE:4217 -IGNORE:4049 -NOLOGO -ENTRY:wmainCRTStartup -SUBSYSTEM:CONSOLE + BASIC_LDFLAGS = -IGNORE:4217 -IGNORE:4049 -NOLOGO # invalidcontinue.obj allows Git's source code to close the same file # handle twice, or to access the osfhandle of an already-closed stdout # See https://msdn.microsoft.com/en-us/library/ms235330.aspx From eb0ac4c9811c9935dab01da4dd6878303cdd6ec8 Mon Sep 17 00:00:00 2001 From: Yuyi Wang Date: Sat, 11 Mar 2023 17:51:18 +0800 Subject: [PATCH 127/312] cmake: install headless-git. headless-git is a git executable without opening a console window. It is useful when other GUI executables want to call git. We should install it together with git on Windows. Signed-off-by: Yuyi Wang --- contrib/buildsystems/CMakeLists.txt | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/contrib/buildsystems/CMakeLists.txt b/contrib/buildsystems/CMakeLists.txt index 25b495fa737d07..709f51120a683a 100644 --- a/contrib/buildsystems/CMakeLists.txt +++ b/contrib/buildsystems/CMakeLists.txt @@ -745,6 +745,7 @@ if(WIN32) endif() add_executable(headless-git ${CMAKE_SOURCE_DIR}/compat/win32/headless.c) + list(APPEND PROGRAMS_BUILT headless-git) if(CMAKE_C_COMPILER_ID STREQUAL "GNU" OR CMAKE_C_COMPILER_ID STREQUAL "Clang") target_link_options(headless-git PUBLIC -municode -Wl,-subsystem,windows) elseif(CMAKE_C_COMPILER_ID STREQUAL "MSVC") @@ -945,7 +946,7 @@ list(TRANSFORM git_perl_scripts PREPEND "${CMAKE_BINARY_DIR}/") #install foreach(program ${PROGRAMS_BUILT}) -if(program MATCHES "^(git|git-shell|scalar)$") +if(program MATCHES "^(git|git-shell|headless-git|scalar)$") install(TARGETS ${program} RUNTIME DESTINATION bin) else() From 42994a37ef61f9072b1e7eaf5f0358edcda1041d Mon Sep 17 00:00:00 2001 From: Philip Oakley Date: Sun, 6 Oct 2019 18:43:57 +0100 Subject: [PATCH 128/312] vcpkg_install: add comment regarding slow network connections The vcpkg downloads may not succeed. Warn careful readers of the time out. A simple retry will usually resolve the issue. Signed-off-by: Philip Oakley Signed-off-by: Johannes Schindelin --- compat/vcbuild/vcpkg_install.bat | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/compat/vcbuild/vcpkg_install.bat b/compat/vcbuild/vcpkg_install.bat index bcbbf536af3141..8330d8120fb511 100644 --- a/compat/vcbuild/vcpkg_install.bat +++ b/compat/vcbuild/vcpkg_install.bat @@ -80,6 +80,12 @@ REM ================================================================ :sub__install_one echo Installing package %1... + REM vcpkg may not be reliable on slow, intermittent or proxy + REM connections, see e.g. + REM https://social.msdn.microsoft.com/Forums/windowsdesktop/en-US/4a8f7be5-5e15-4213-a7bb-ddf424a954e6/winhttpsendrequest-ends-with-12002-errorhttptimeout-after-21-seconds-no-matter-what-timeout?forum=windowssdk + REM which explains the hidden 21 second timeout + REM (last post by Dave : Microsoft - Windows Networking team) + .\vcpkg.exe install %1:%arch% IF ERRORLEVEL 1 ( EXIT /B 1 ) From bbeca033a52817a030c273a7b66f111828dd0dab Mon Sep 17 00:00:00 2001 From: Dennis Ameling Date: Sun, 6 Dec 2020 18:39:26 +0100 Subject: [PATCH 129/312] Add schannel to curl installation Signed-off-by: Dennis Ameling --- compat/vcbuild/vcpkg_install.bat | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/compat/vcbuild/vcpkg_install.bat b/compat/vcbuild/vcpkg_install.bat index 8da212487ae97d..575c65c20ba307 100644 --- a/compat/vcbuild/vcpkg_install.bat +++ b/compat/vcbuild/vcpkg_install.bat @@ -127,5 +127,5 @@ set features= goto :EOF :curl_features -set features=[core,openssl] +set features=[core,openssl,schannel] goto :EOF From 7d2f291981ae4d5c3ee8dfe1dcd8673a8f61d176 Mon Sep 17 00:00:00 2001 From: Philip Oakley Date: Fri, 2 Jul 2021 00:30:24 +0100 Subject: [PATCH 130/312] CMake: default Visual Studio generator has changed Correct some wording and inform users regarding the Visual Studio changes (from V16.6) to the default generator. Subsequent commits ensure that Git for Windows can be directly opened in modern Visual Studio without needing special configuration of the CMakeLists settings. It appeares that internally Visual Studio creates it's own version of the .sln file (etc.) for extension tools that expect them. The large number of references below document the shifting of Visual Studio default and CMake setting options. refs: https://docs.microsoft.com/en-us/search/?scope=C%2B%2B&view=msvc-150&terms=Ninja 1. https://docs.microsoft.com/en-us/cpp/linux/cmake-linux-configure?view=msvc-160 (note the linux bit) "In Visual Studio 2019 version 16.6 or later ***, Ninja is the default generator for configurations targeting a remote system or WSL. For more information, see this post on the C++ Team Blog [https://devblogs.microsoft.com/cppblog/linux-development-with-visual-studio-first-class-support-for-gdbserver-improved-build-times-with-ninja-and-updates-to-the-connection-manager/]. For more information about these settings, see CMakeSettings.json reference [https://docs.microsoft.com/en-us/cpp/build/cmakesettings-reference?view=msvc-160]." 2. https://docs.microsoft.com/en-us/cpp/build/cmake-presets-vs?view=msvc-160 "CMake supports two files that allow users to specify common configure, build, and test options and share them with others: CMakePresets.json and CMakeUserPresets.json." " Both files are supported in Visual Studio 2019 version 16.10 or later. ***" 3. https://devblogs.microsoft.com/cppblog/linux-development-with-visual-studio-first-class-support-for-gdbserver-improved-build-times-with-ninja-and-updates-to-the-connection-manager/ " Ninja has been the default generator (underlying build system) for CMake configurations targeting Windows for some time***, but in Visual Studio 2019 version 16.6 Preview 3*** we added support for Ninja on Linux." 4. https://docs.microsoft.com/en-us/cpp/build/cmakesettings-reference?view=msvc-160 " `generator`: specifies CMake generator to use for this configuration. May be one of: Visual Studio 2019 only: Visual Studio 16 2019 Visual Studio 16 2019 Win64 Visual Studio 16 2019 ARM Visual Studio 2017 and later: Visual Studio 15 2017 Visual Studio 15 2017 Win64 Visual Studio 15 2017 ARM Visual Studio 14 2015 Visual Studio 14 2015 Win64 Visual Studio 14 2015 ARM Unix Makefiles Ninja Because Ninja is designed for fast build speeds instead of flexibility and function, it is set as the default. However, some CMake projects may be unable to correctly build using Ninja. If this occurs, you can instruct CMake to generate Visual Studio projects instead. To specify a Visual Studio generator in Visual Studio 2017, open the settings editor from the main menu by choosing CMake | Change CMake Settings. Delete "Ninja" and type "V". This activates IntelliSense, which enables you to choose the generator you want." "To specify a Visual Studio generator in Visual Studio 2019, right-click on the CMakeLists.txt file in Solution Explorer and choose CMake Settings for project > Show Advanced Settings > CMake Generator. When the active configuration specifies a Visual Studio generator, by default MSBuild.exe is invoked with` -m -v:minimal` arguments." 5. https://docs.microsoft.com/en-us/cpp/build/cmake-presets-vs?view=msvc-160#enable-cmakepresetsjson-integration-in-visual-studio-2019 "Enable CMakePresets.json integration in Visual Studio 2019 CMakePresets.json integration isn't enabled by default in Visual Studio 2019. You can enable it for all CMake projects in Tools > Options > CMake > General: (tick a box)" ... see more. 6. https://docs.microsoft.com/en-us/cpp/build/cmakesettings-reference?view=msvc-140 (whichever v140 is..) "CMake projects are supported in Visual Studio 2017 and later." 7. https://docs.microsoft.com/en-us/cpp/overview/what-s-new-for-cpp-2017?view=msvc-150 "Support added for the CMake Ninja generator." 8. https://docs.microsoft.com/en-us/cpp/overview/what-s-new-for-cpp-2017?view=msvc-150#cmake-support-via-open-folder "CMake support via Open Folder Visual Studio 2017 introduces support for using CMake projects without converting to MSBuild project files (.vcxproj). For more information, see CMake projects in Visual Studio[https://docs.microsoft.com/en-us/cpp/build/cmake-projects-in-visual-studio?view=msvc-150]. Opening CMake projects with Open Folder automatically configures the environment for C++ editing, building, and debugging." ... +more! 9. https://docs.microsoft.com/en-us/cpp/build/cmake-presets-vs?view=msvc-160#supported-cmake-and-cmakepresetsjson-versions "Visual Studio reads and evaluates CMakePresets.json and CMakeUserPresets.json itself and doesn't invoke CMake directly with the --preset option. So, CMake version 3.20 or later isn't strictly required when you're building with CMakePresets.json inside Visual Studio. We recommend using CMake version 3.14 or later." 10. https://docs.microsoft.com/en-us/cpp/build/cmake-presets-vs?view=msvc-160#enable-cmakepresetsjson-integration-in-visual-studio-2019 "If you don't want to enable CMakePresets.json integration for all CMake projects, you can enable CMakePresets.json integration for a single CMake project by adding a CMakePresets.json file to the root of the open folder. You must close and reopen the folder in Visual Studio to activate the integration. 11. https://docs.microsoft.com/en-us/cpp/build/cmake-presets-vs?view=msvc-160#default-configure-presets ***(doesn't actually say which version..) "Default Configure Presets If no CMakePresets.json or CMakeUserPresets.json file exists, or if CMakePresets.json or CMakeUserPresets.json is invalid, Visual Studio will fall back*** on the following default Configure Presets: Windows example JSON { "name": "windows-default", "displayName": "Windows x64 Debug", "description": "Sets Ninja generator, compilers, x64 architecture, build and install directory, debug build type", "generator": "Ninja", "binaryDir": "${sourceDir}/out/build/${presetName}", "architecture": { "value": "x64", "strategy": "external" }, "cacheVariables": { "CMAKE_BUILD_TYPE": "Debug", "CMAKE_INSTALL_PREFIX": "${sourceDir}/out/install/${presetName}" }, "vendor": { "microsoft.com/VisualStudioSettings/CMake/1.0": { "hostOS": [ "Windows" ] } } }, " Signed-off-by: Philip Oakley --- contrib/buildsystems/CMakeLists.txt | 16 ++++++++++------ 1 file changed, 10 insertions(+), 6 deletions(-) diff --git a/contrib/buildsystems/CMakeLists.txt b/contrib/buildsystems/CMakeLists.txt index d61e54a7f6dd4e..30fa3250c75bf1 100644 --- a/contrib/buildsystems/CMakeLists.txt +++ b/contrib/buildsystems/CMakeLists.txt @@ -14,6 +14,11 @@ Note: Visual Studio also has the option of opening `CMakeLists.txt` directly; Using this option, Visual Studio will not find the source code, though, therefore the `File>Open>Folder...` option is preferred. +Visual Studio does not produce a .sln solution file nor the .vcxproj files +that may be required by VS extension tools. + +To generate the .sln/.vcxproj files run CMake manually, as described below. + Instructions to run CMake manually: mkdir -p contrib/buildsystems/out @@ -22,7 +27,7 @@ Instructions to run CMake manually: This will build the git binaries in contrib/buildsystems/out directory (our top-level .gitignore file knows to ignore contents of -this directory). +this directory). The project .sln and .vcxproj files are also generated. Possible build configurations(-DCMAKE_BUILD_TYPE) with corresponding compiler flags @@ -35,17 +40,16 @@ empty(default) : NOTE: -DCMAKE_BUILD_TYPE is optional. For multi-config generators like Visual Studio this option is ignored -This process generates a Makefile(Linux/*BSD/MacOS) , Visual Studio solution(Windows) by default. +This process generates a Makefile(Linux/*BSD/MacOS), Visual Studio solution(Windows) by default. Run `make` to build Git on Linux/*BSD/MacOS. Open git.sln on Windows and build Git. -NOTE: By default CMake uses Makefile as the build tool on Linux and Visual Studio in Windows, -to use another tool say `ninja` add this to the command line when configuring. -`-G Ninja` - NOTE: By default CMake will install vcpkg locally to your source tree on configuration, to avoid this, add `-DNO_VCPKG=TRUE` to the command line when configuring. +The Visual Studio default generator changed in v16.6 from its Visual Studio +implemenation to `Ninja` This required changes to many CMake scripts. + ]] cmake_minimum_required(VERSION 3.14) From 8e87931c1ea014f9558f6e2d0984ace0f086a335 Mon Sep 17 00:00:00 2001 From: Philip Oakley Date: Sat, 24 Apr 2021 11:09:58 +0100 Subject: [PATCH 131/312] .gitignore: add Visual Studio CMakeSetting.json file The CMakeSettings.json file is tool generated. Developers may track it should they provide additional settings. Signed-off-by: Philip Oakley --- .gitignore | 1 + 1 file changed, 1 insertion(+) diff --git a/.gitignore b/.gitignore index 08a66ca50844c5..b633ef317214b8 100644 --- a/.gitignore +++ b/.gitignore @@ -251,5 +251,6 @@ Release/ /git.VC.db *.dSYM /contrib/buildsystems/out +CMakeSettings.json /contrib/libgit-rs/target /contrib/libgit-sys/target From f238f4ea6ee3f5509dbd1ad4d793d2587492102a Mon Sep 17 00:00:00 2001 From: Victoria Dye Date: Thu, 5 Aug 2021 19:04:13 -0400 Subject: [PATCH 132/312] subtree: update `contrib/subtree` `test` target The intention of this change is to align with how the top-level git `Makefile` defines its own test target (which also internally calls `$(MAKE) -C t/ all`). This change also ensures the consistency of `make -C contrib/subtree test` with other testing in CI executions (which rely on `$DEFAULT_TEST_TARGET` being defined as `prove`). Signed-off-by: Victoria Dye --- contrib/subtree/Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/contrib/subtree/Makefile b/contrib/subtree/Makefile index c0c9f21cb78022..dab2dfc08ee222 100644 --- a/contrib/subtree/Makefile +++ b/contrib/subtree/Makefile @@ -95,7 +95,7 @@ $(GIT_SUBTREE_TEST): $(GIT_SUBTREE) cp $< $@ test: $(GIT_SUBTREE_TEST) - $(MAKE) -C t/ test + $(MAKE) -C t/ all clean: $(RM) $(GIT_SUBTREE) From cf7c114407163b59c12fac6a198f6f245f937611 Mon Sep 17 00:00:00 2001 From: Philip Oakley Date: Thu, 22 Apr 2021 11:11:38 +0100 Subject: [PATCH 133/312] CMakeLists: add default "x64-windows" arch for Visual Studio In Git-for-Windows, work on using ARM64 has progressed. The commit 2d94b77b27 (cmake: allow building for Windows/ARM64, 2020-12-04) failed to notice that /compat/vcbuild/vcpkg_install.bat will default to using the "x64-windows" architecture for the vcpkg installation if not set, but CMake is not told of this default. Commit 635b6d99b3 (vcbuild: install ARM64 dependencies when building ARM64 binaries, 2020-01-31) later updated vcpkg_install.bat to accept an arch (%1) parameter, but retained the default. This default is neccessary for the use case where the project directory is opened directly in Visual Studio, which will find and build a CMakeLists.txt file without any parameters, thus expecting use of the default setting. Also Visual studio will generate internal .sln solution and .vcxproj project files needed for some extension tools. Inform users of the additional .sln/.vcxproj generation. ** How to test: rm -rf '.vs' # remove old visual studio settings rm -rf 'compat/vcbuild/vcpkg' # remove any vcpkg downloads rm -rf 'contrib/buildsystems/out' # remove builds & CMake artifacts with a fresh Visual Studio Community Edition, File>>Open>>(git *folder*) to load the project (which will take some time!). check for successful compilation. The implicit .sln (etc.) are in the hidden .vs directory created by Visual Studio. Signed-off-by: Philip Oakley --- contrib/buildsystems/CMakeLists.txt | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/contrib/buildsystems/CMakeLists.txt b/contrib/buildsystems/CMakeLists.txt index 30fa3250c75bf1..d21e22835643d0 100644 --- a/contrib/buildsystems/CMakeLists.txt +++ b/contrib/buildsystems/CMakeLists.txt @@ -71,6 +71,10 @@ if(USE_VCPKG) message("Initializing vcpkg and building the Git's dependencies (this will take a while...)") execute_process(COMMAND ${CMAKE_SOURCE_DIR}/compat/vcbuild/vcpkg_install.bat ${VCPKG_ARCH}) endif() + if(NOT EXISTS ${VCPKG_ARCH}) + message("VCPKG_ARCH: unset, using 'x64-windows'") + set(VCPKG_ARCH "x64-windows") # default from vcpkg_install.bat + endif() list(APPEND CMAKE_PREFIX_PATH "${VCPKG_DIR}/installed/${VCPKG_ARCH}") # In the vcpkg edition, we need this to be able to link to libcurl From 88e7bacdb17599928702493d15d71cef68dd6960 Mon Sep 17 00:00:00 2001 From: Dennis Ameling Date: Mon, 19 Jul 2021 13:02:16 +0200 Subject: [PATCH 134/312] cmake(): allow setting HOST_CPU for cross-compilation Git's regular Makefile mentions that HOST_CPU should be defined when cross-compiling Git: https://github.com/git-for-windows/git/blob/37796bca76ef4180c39ee508ca3e42c0777ba444/Makefile#L438-L439 This is then used to set the GIT_HOST_CPU variable when compiling Git: https://github.com/git-for-windows/git/blob/37796bca76ef4180c39ee508ca3e42c0777ba444/Makefile#L1337-L1341 Then, when the user runs `git version --build-options`, it returns that value: https://github.com/git-for-windows/git/blob/37796bca76ef4180c39ee508ca3e42c0777ba444/help.c#L658 This commit adds the same functionality to the CMake configuration. Users can now set -DHOST_CPU= to set the target architecture. Signed-off-by: Dennis Ameling --- .github/workflows/main.yml | 2 +- contrib/buildsystems/CMakeLists.txt | 9 ++++++++- 2 files changed, 9 insertions(+), 2 deletions(-) diff --git a/.github/workflows/main.yml b/.github/workflows/main.yml index 6ee3c5b42b85c4..4073c8b076eb24 100644 --- a/.github/workflows/main.yml +++ b/.github/workflows/main.yml @@ -197,7 +197,7 @@ jobs: shell: bash run: | cmake `pwd`/contrib/buildsystems/ -DCMAKE_PREFIX_PATH=`pwd`/compat/vcbuild/vcpkg/installed/${{ matrix.arch }}-windows \ - -DNO_GETTEXT=YesPlease -DPERL_TESTS=OFF -DPYTHON_TESTS=OFF -DCURL_NO_CURL_CMAKE=ON -DCMAKE_GENERATOR_PLATFORM=${{ matrix.arch }} -DVCPKG_ARCH=${{ matrix.arch }}-windows + -DNO_GETTEXT=YesPlease -DPERL_TESTS=OFF -DPYTHON_TESTS=OFF -DCURL_NO_CURL_CMAKE=ON -DCMAKE_GENERATOR_PLATFORM=${{ matrix.arch }} -DVCPKG_ARCH=${{ matrix.arch }}-windows -DHOST_CPU=${{ matrix.arch }} - name: MSBuild run: msbuild git.sln -property:Configuration=Release -property:Platform=${{ matrix.arch }} -maxCpuCount:4 -property:PlatformToolset=v142 - name: bundle artifact tar diff --git a/contrib/buildsystems/CMakeLists.txt b/contrib/buildsystems/CMakeLists.txt index aaaee1bc4b7943..d61e54a7f6dd4e 100644 --- a/contrib/buildsystems/CMakeLists.txt +++ b/contrib/buildsystems/CMakeLists.txt @@ -212,7 +212,14 @@ endif() #default behaviour include_directories(${CMAKE_SOURCE_DIR}) -add_compile_definitions(GIT_HOST_CPU="${CMAKE_SYSTEM_PROCESSOR}") + +# When cross-compiling, define HOST_CPU as the canonical name of the CPU on +# which the built Git will run (for instance "x86_64"). +if(NOT HOST_CPU) + add_compile_definitions(GIT_HOST_CPU="${CMAKE_SYSTEM_PROCESSOR}") +else() + add_compile_definitions(GIT_HOST_CPU="${HOST_CPU}") +endif() add_compile_definitions(SHA256_BLK INTERNAL_QSORT RUNTIME_PREFIX) add_compile_definitions(NO_OPENSSL SHA1_DC SHA1DC_NO_STANDARD_INCLUDES SHA1DC_INIT_SAFE_HASH_DEFAULT=0 From 3d2699ab1bd158ba35e329c3b9c053fe60195b27 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Fri, 2 Apr 2021 22:50:54 +0200 Subject: [PATCH 135/312] mingw: allow for longer paths in `parse_interpreter()` MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit As reported in https://github.com/newren/git-filter-repo/pull/225, it looks like 99 bytes is not really sufficient to represent e.g. the full path to Python when installed via Windows Store (and this path is used in the hasb bang line when installing scripts via `pip`). Let's increase it to what is probably the maximum sensible path size: MAX_PATH. This makes `parse_interpreter()` in line with what `lookup_prog()` handles. Signed-off-by: Johannes Schindelin Signed-off-by: Vilius Šumskas --- compat/mingw.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/compat/mingw.c b/compat/mingw.c index f524c54d06d965..f65d8e2bc0a576 100644 --- a/compat/mingw.c +++ b/compat/mingw.c @@ -1339,7 +1339,7 @@ static const char *quote_arg_msys2(const char *arg) static const char *parse_interpreter(const char *cmd) { - static char buf[100]; + static char buf[MAX_PATH]; char *p, *opt; ssize_t n; /* read() can return negative values */ int fd; From c49386854544fd5bc675c2c72fe72d964623c634 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Mon, 17 May 2021 10:46:52 +0200 Subject: [PATCH 136/312] compat/vcbuild: document preferred way to build in Visual Studio We used to have that `make vcxproj` hack, but a hack it is. In the meantime, we have a much cleaner solution: using CMake, either explicitly, or even more conveniently via Visual Studio's built-in CMake support (simply open Git's top-level directory via File>Open>Folder...). Let's let the `README` reflect this. Signed-off-by: Johannes Schindelin --- compat/vcbuild/README | 28 +++++++++------------------- 1 file changed, 9 insertions(+), 19 deletions(-) diff --git a/compat/vcbuild/README b/compat/vcbuild/README index 29ec1d0f104b80..5c71ea2daa4017 100644 --- a/compat/vcbuild/README +++ b/compat/vcbuild/README @@ -37,27 +37,17 @@ The Steps to Build Git with VS2015 or VS2017 from the command line. ================================================================ -Alternatively, run `make vcxproj` and then load the generated `git.sln` in -Visual Studio. The initial build will install the vcpkg system and build the +Alternatively, just open Git's top-level directory in Visual Studio, via +`File>Open>Folder...`. This will use CMake internally to generate the +project definitions. It will also install the vcpkg system and build the dependencies automatically. This will take a while. -Instead of generating the `git.sln` file yourself (which requires a full Git -for Windows SDK), you may want to consider fetching the `vs/master` branch of -https://github.com/git-for-windows/git instead (which is updated automatically -via CI running `make vcxproj`). The `vs/master` branch does not require a Git -for Windows to build, but you can run the test scripts in a regular Git Bash. - -Note that `make vcxproj` will automatically add and commit the generated `.sln` -and `.vcxproj` files to the repo. This is necessary to allow building a -fully-testable Git in Visual Studio, where a regular Git Bash can be used to -run the test scripts (as opposed to a full Git for Windows SDK): a number of -build targets, such as Git commands implemented as Unix shell scripts (where -`@@SHELL_PATH@@` and other placeholders are interpolated) require a full-blown -Git for Windows SDK (which is about 10x the size of a regular Git for Windows -installation). - -If your plan is to open a Pull Request with Git for Windows, it is a good idea -to drop this commit before submitting. +You can also generate the Visual Studio solution manually by downloading +and running CMake explicitly rather than letting Visual Studio doing +that implicitly. + +Another, deprecated option is to run `make vcxproj`. This option is +superseded by the CMake-based build, and will be removed at some point. ================================================================ The Steps of Build Git with VS2008 From 1f5dfe8eb17ac9dd4534b82d9797fc3669a13b99 Mon Sep 17 00:00:00 2001 From: Pascal Muller Date: Wed, 23 Jun 2021 21:21:10 +0200 Subject: [PATCH 137/312] http: optionally send SSL client certificate This adds support for a new http.sslAutoClientCert config value. In cURL 7.77 or later the schannel backend does not automatically send client certificates from the Windows Certificate Store anymore. This config value is only used if http.sslBackend is set to "schannel", and can be used to opt in to the old behavior and force cURL to send client certificates. This fixes https://github.com/git-for-windows/git/issues/3292 Signed-off-by: Pascal Muller --- Documentation/config/http.adoc | 5 +++++ git-curl-compat.h | 8 ++++++++ http.c | 24 +++++++++++++++++++++--- 3 files changed, 34 insertions(+), 3 deletions(-) diff --git a/Documentation/config/http.adoc b/Documentation/config/http.adoc index de52a1c44fbf14..484a6b574473e7 100644 --- a/Documentation/config/http.adoc +++ b/Documentation/config/http.adoc @@ -249,6 +249,11 @@ http.schannelUseSSLCAInfo:: when the `schannel` backend was configured via `http.sslBackend`, unless `http.schannelUseSSLCAInfo` overrides this behavior. +http.sslAutoClientCert:: + As of cURL v7.77.0, the Secure Channel backend won't automatically + send client certificates from the Windows Certificate Store anymore. + To opt in to the old behavior, http.sslAutoClientCert can be set. + http.pinnedPubkey:: Public key of the https service. It may either be the filename of a PEM or DER encoded public key file or a string starting with diff --git a/git-curl-compat.h b/git-curl-compat.h index 703756ba851d5b..8291755f50454f 100644 --- a/git-curl-compat.h +++ b/git-curl-compat.h @@ -45,4 +45,12 @@ #define GIT_CURL_HAVE_CURLOPT_PROTOCOLS_STR 1 #endif +/** + * CURLSSLOPT_AUTO_CLIENT_CERT was added in 7.77.0, released in May + * 2021. + */ +#if LIBCURL_VERSION_NUM >= 0x074d00 +#define GIT_CURL_HAVE_CURLSSLOPT_AUTO_CLIENT_CERT +#endif + #endif diff --git a/http.c b/http.c index 657c3e1dc5e2c2..d5396a3ce2ba34 100644 --- a/http.c +++ b/http.c @@ -157,6 +157,8 @@ static int http_schannel_check_revoke_mode = */ static int http_schannel_use_ssl_cainfo; +static int http_auto_client_cert; + static int always_auth_proactively(void) { return http_proactive_auth != PROACTIVE_AUTH_NONE && @@ -445,6 +447,11 @@ static int http_options(const char *var, const char *value, return 0; } + if (!strcmp("http.sslautoclientcert", var)) { + http_auto_client_cert = git_config_bool(var, value); + return 0; + } + if (!strcmp("http.minsessions", var)) { min_curl_sessions = git_config_int(var, value, ctx->kvi); if (min_curl_sessions > 1) @@ -1061,9 +1068,20 @@ static CURL *get_curl_handle(void) } #endif - if (http_ssl_backend && !strcmp("schannel", http_ssl_backend) && - http_schannel_check_revoke_mode) { - curl_easy_setopt(result, CURLOPT_SSL_OPTIONS, http_schannel_check_revoke_mode); + if (http_ssl_backend && !strcmp("schannel", http_ssl_backend)) { + long ssl_options = 0; + if (http_schannel_check_revoke_mode) { + ssl_options |= http_schannel_check_revoke_mode; + } + + if (http_auto_client_cert) { +#ifdef GIT_CURL_HAVE_CURLSSLOPT_AUTO_CLIENT_CERT + ssl_options |= CURLSSLOPT_AUTO_CLIENT_CERT; +#endif + } + + if (ssl_options) + curl_easy_setopt(result, CURLOPT_SSL_OPTIONS, ssl_options); } if (http_proactive_auth != PROACTIVE_AUTH_NONE) From c8f5ffc9a919215d001f28817a1cebdf77051fb4 Mon Sep 17 00:00:00 2001 From: Victoria Dye Date: Thu, 5 Aug 2021 19:11:59 -0400 Subject: [PATCH 138/312] ci: run `contrib/subtree` tests in CI builds Because `git subtree` (unlike most other `contrib` modules) is included as part of the standard release of Git for Windows, its stability should be verified as consistently as it is for the rest of git. By including the `git subtree` tests in the CI workflow, these tests are as much of a gate to merging and indicator of stability as the standard test suite. Signed-off-by: Victoria Dye --- ci/run-build-and-tests.sh | 4 ++++ ci/run-test-slice.sh | 3 +++ 2 files changed, 7 insertions(+) diff --git a/ci/run-build-and-tests.sh b/ci/run-build-and-tests.sh index 1c698467235943..a7b7f2b846a579 100755 --- a/ci/run-build-and-tests.sh +++ b/ci/run-build-and-tests.sh @@ -77,5 +77,9 @@ case "$jobname" in ;; esac +case " $MAKE_TARGETS " in +*" all "*) make -C contrib/subtree test;; +esac + check_unignored_build_artifacts save_good_tree diff --git a/ci/run-test-slice.sh b/ci/run-test-slice.sh index e167e646f79e3d..8cb0038d07ca64 100755 --- a/ci/run-test-slice.sh +++ b/ci/run-test-slice.sh @@ -20,4 +20,7 @@ if [ "$1" == "0" ] ; then group "Run unit tests" make --quiet -C t unit-tests-test-tool fi +# Run the git subtree tests only if main tests succeeded +test 0 != "$1" || make -C contrib/subtree test + check_unignored_build_artifacts From c733f85e006009baf340227671bc95e655483360 Mon Sep 17 00:00:00 2001 From: Philip Oakley Date: Mon, 10 May 2021 16:47:40 +0100 Subject: [PATCH 139/312] CMake: show Win32 and Generator_platform build-option values Ensure key CMake option values are part of the CMake output to facilitate user support when tool updates impact the wider CMake actions, particularly ongoing 'improvements' in Visual Studio. These CMake displays perform the same function as the build-options.txt provided in the main Git for Windows. CMake is already chatty. The setting of CMAKE_EXPORT_COMPILE_COMMANDS is also reported. Include the environment's CMAKE_EXPORT_COMPILE_COMMANDS value which may have been propogated to CMake's internal value. Testing the CMAKE_EXPORT_COMPILE_COMMANDS processing can be difficult in the Visual Studio environment, as it may be cached in many places. The 'environment' may include the OS, the user shell, CMake's own environment, along with the Visual Studio presets and caches. See previous commit for arefacts that need removing for a clean test. Signed-off-by: Philip Oakley --- contrib/buildsystems/CMakeLists.txt | 10 ++++++++++ 1 file changed, 10 insertions(+) diff --git a/contrib/buildsystems/CMakeLists.txt b/contrib/buildsystems/CMakeLists.txt index d21e22835643d0..fa3885b9bc99b0 100644 --- a/contrib/buildsystems/CMakeLists.txt +++ b/contrib/buildsystems/CMakeLists.txt @@ -63,10 +63,20 @@ endif() if(NOT DEFINED CMAKE_EXPORT_COMPILE_COMMANDS) set(CMAKE_EXPORT_COMPILE_COMMANDS TRUE) + message("settting CMAKE_EXPORT_COMPILE_COMMANDS: ${CMAKE_EXPORT_COMPILE_COMMANDS}") endif() if(USE_VCPKG) set(VCPKG_DIR "${CMAKE_SOURCE_DIR}/compat/vcbuild/vcpkg") + message("WIN32: ${WIN32}") # show its underlying text values + message("VCPKG_DIR: ${VCPKG_DIR}") + message("VCPKG_ARCH: ${VCPKG_ARCH}") # maybe unset + message("MSVC: ${MSVC}") + message("CMAKE_GENERATOR: ${CMAKE_GENERATOR}") + message("CMAKE_CXX_COMPILER_ID: ${CMAKE_CXX_COMPILER_ID}") + message("CMAKE_GENERATOR_PLATFORM: ${CMAKE_GENERATOR_PLATFORM}") + message("CMAKE_EXPORT_COMPILE_COMMANDS: ${CMAKE_EXPORT_COMPILE_COMMANDS}") + message("ENV(CMAKE_EXPORT_COMPILE_COMMANDS): $ENV{CMAKE_EXPORT_COMPILE_COMMANDS}") if(NOT EXISTS ${VCPKG_DIR}) message("Initializing vcpkg and building the Git's dependencies (this will take a while...)") execute_process(COMMAND ${CMAKE_SOURCE_DIR}/compat/vcbuild/vcpkg_install.bat ${VCPKG_ARCH}) From f05ffbe3d893260632fc9f96e807d3928952f088 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Wed, 8 Sep 2021 13:05:42 +0200 Subject: [PATCH 140/312] init: do parse _all_ core.* settings early In Git for Windows, `has_symlinks` is set to 0 by default. Therefore, we need to parse the config setting `core.symlinks` to know if it has been set to `true`. In `git init`, we must do that before copying the templates because they might contain symbolic links. Even if the support for symbolic links on Windows has not made it to upstream Git yet, we really should make sure that all the `core.*` settings are parsed before proceeding, as they might very well change the behavior of `git init` in a way the user intended. This fixes https://github.com/git-for-windows/git/issues/3414 Signed-off-by: Johannes Schindelin --- config.c | 4 ++-- config.h | 2 ++ setup.c | 2 +- 3 files changed, 5 insertions(+), 3 deletions(-) diff --git a/config.c b/config.c index 658569af084897..e1a9e9aa7aa6ec 100644 --- a/config.c +++ b/config.c @@ -1393,8 +1393,8 @@ int git_config_color(char *dest, const char *var, const char *value) return 0; } -static int git_default_core_config(const char *var, const char *value, - const struct config_context *ctx, void *cb) +int git_default_core_config(const char *var, const char *value, + const struct config_context *ctx, void *cb) { /* This needs a better name */ if (!strcmp(var, "core.filemode")) { diff --git a/config.h b/config.h index 5c730c4f899160..e4199bbdc07685 100644 --- a/config.h +++ b/config.h @@ -167,6 +167,8 @@ typedef int (*config_fn_t)(const char *, const char *, int git_default_config(const char *, const char *, const struct config_context *, void *); +int git_default_core_config(const char *var, const char *value, + const struct config_context *ctx, void *cb); /** * Read a specific file in git-config format. diff --git a/setup.c b/setup.c index f93bd6a24a5d9c..554b3e3751390d 100644 --- a/setup.c +++ b/setup.c @@ -2590,7 +2590,7 @@ int init_db(const char *git_dir, const char *real_git_dir, * have set up the repository format such that we can evaluate * includeIf conditions correctly in the case of re-initialization. */ - git_config(platform_core_config, NULL); + git_config(git_default_core_config, NULL); safe_create_dir(the_repository, git_dir, 0); From a76d967114a2e9506bbb9f3991db07d51a602cd2 Mon Sep 17 00:00:00 2001 From: Philip Oakley Date: Sun, 31 Oct 2021 23:15:13 +0000 Subject: [PATCH 141/312] hash-object: demonstrate a >4GB/LLP64 problem On LLP64 systems, such as Windows, the size of `long`, `int`, etc. is only 32 bits (for backward compatibility). Git's use of `unsigned long` for file memory sizes in many places, rather than size_t, limits the handling of large files on LLP64 systems (commonly given as `>4GB`). Provide a minimum test for handling a >4GB file. The `hash-object` command, with the `--literally` and without `-w` option avoids writing the object, either loose or packed. This avoids the code paths hitting the `bigFileThreshold` config test code, the zlib code, and the pack code. Subsequent patches will walk the test's call chain, converting types to `size_t` (which is larger in LLP64 data models) where appropriate. Signed-off-by: Philip Oakley Signed-off-by: Johannes Schindelin --- t/t1007-hash-object.sh | 11 +++++++++++ 1 file changed, 11 insertions(+) diff --git a/t/t1007-hash-object.sh b/t/t1007-hash-object.sh index a0481139de5519..51604797e98b35 100755 --- a/t/t1007-hash-object.sh +++ b/t/t1007-hash-object.sh @@ -49,6 +49,9 @@ test_expect_success 'setup' ' example sha1:ddd3f836d3e3fbb7ae289aa9ae83536f76956399 example sha256:b44fe1fe65589848253737db859bd490453510719d7424daab03daf0767b85ae + + large5GB sha1:0be2be10a4c8764f32c4bf372a98edc731a4b204 + large5GB sha256:dc18ca621300c8d3cfa505a275641ebab00de189859e022a975056882d313e64 EOF ' @@ -265,4 +268,12 @@ test_expect_success '--stdin outside of repository (uses SHA-1)' ' test_cmp expect actual ' +test_expect_failure EXPENSIVE,SIZE_T_IS_64BIT,!LONG_IS_64BIT \ + 'files over 4GB hash literally' ' + test-tool genzeros $((5*1024*1024*1024)) >big && + test_oid large5GB >expect && + git hash-object --stdin --literally actual && + test_cmp expect actual +' + test_done From bc532ace20e4ff0e79c17a1423314418a4dee79b Mon Sep 17 00:00:00 2001 From: Philip Oakley Date: Fri, 12 Nov 2021 21:07:03 +0000 Subject: [PATCH 142/312] write_object_file_literally(): use size_t The previous commit adds a test that demonstrates a problem in the `hash-object --literally` command, manifesting in an unnecessary file size limit on systems using the LLP64 data model (which includes Windows). Walking the affected code path is `cmd_hash_object()` >> `hash_fd()` >> `hash_literally()` >> `hash_object_file_literally()`. The function `hash_object_file_literally()` is the first with a file length parameter (via a mem buffer). This commit changes the type of that parameter to the LLP64 compatible `size_t` type. There are no other uses of the function. The `strbuf` type is already `size_t` compatible. Note: The hash-object test does not yet pass. Subsequent commits will continue to walk the call tree's lower level functions to identify further fixes. Signed-off-by: Philip Oakley Signed-off-by: Johannes Schindelin --- object-file.c | 4 ++-- object-store-ll.h | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/object-file.c b/object-file.c index 726e41a0475b43..d9c1b2baa5cfc9 100644 --- a/object-file.c +++ b/object-file.c @@ -1977,7 +1977,7 @@ static void write_object_file_prepare(const struct git_hash_algo *algo, } static void write_object_file_prepare_literally(const struct git_hash_algo *algo, - const void *buf, unsigned long len, + const void *buf, size_t len, const char *type, struct object_id *oid, char *hdr, int *hdrlen) { @@ -2531,7 +2531,7 @@ int write_object_file_flags(const void *buf, unsigned long len, return 0; } -int write_object_file_literally(const void *buf, unsigned long len, +int write_object_file_literally(const void *buf, size_t len, const char *type, struct object_id *oid, unsigned flags) { diff --git a/object-store-ll.h b/object-store-ll.h index cd3bd5bd99f78c..d54534742f97df 100644 --- a/object-store-ll.h +++ b/object-store-ll.h @@ -282,7 +282,7 @@ static inline int write_object_file(const void *buf, unsigned long len, return write_object_file_flags(buf, len, type, oid, NULL, 0); } -int write_object_file_literally(const void *buf, unsigned long len, +int write_object_file_literally(const void *buf, size_t len, const char *type, struct object_id *oid, unsigned flags); int stream_loose_object(struct input_stream *in_stream, size_t len, From a4821e717599b123415948f6e1b498e45e321330 Mon Sep 17 00:00:00 2001 From: Philip Oakley Date: Fri, 12 Nov 2021 21:14:50 +0000 Subject: [PATCH 143/312] object-file.c: use size_t for header lengths Continue walking the code path for the >4GB `hash-object --literally` test. The `hash_object_file_literally()` function internally uses both `hash_object_file()` and `write_object_file_prepare()`. Both function signatures use `unsigned long` rather than `size_t` for the mem buffer sizes. Use `size_t` instead, for LLP64 compatibility. While at it, convert those function's object's header buffer length to `size_t` for consistency. The value is already upcast to `uintmax_t` for print format compatibility. Note: The hash-object test still does not pass. A subsequent commit continues to walk the call tree's lower level hash functions to identify further fixes. Signed-off-by: Philip Oakley Signed-off-by: Johannes Schindelin --- object-file.c | 22 +++++++++++----------- object-store-ll.h | 4 ++-- 2 files changed, 13 insertions(+), 13 deletions(-) diff --git a/object-file.c b/object-file.c index d9c1b2baa5cfc9..5c8870c2a7d20a 100644 --- a/object-file.c +++ b/object-file.c @@ -1954,7 +1954,7 @@ void *read_object_with_reference(struct repository *r, static void hash_object_body(const struct git_hash_algo *algo, struct git_hash_ctx *c, const void *buf, unsigned long len, struct object_id *oid, - char *hdr, int *hdrlen) + char *hdr, size_t *hdrlen) { algo->init_fn(c); git_hash_update(c, hdr, *hdrlen); @@ -1963,9 +1963,9 @@ static void hash_object_body(const struct git_hash_algo *algo, struct git_hash_c } static void write_object_file_prepare(const struct git_hash_algo *algo, - const void *buf, unsigned long len, + const void *buf, size_t len, enum object_type type, struct object_id *oid, - char *hdr, int *hdrlen) + char *hdr, size_t *hdrlen) { struct git_hash_ctx c; @@ -1979,7 +1979,7 @@ static void write_object_file_prepare(const struct git_hash_algo *algo, static void write_object_file_prepare_literally(const struct git_hash_algo *algo, const void *buf, size_t len, const char *type, struct object_id *oid, - char *hdr, int *hdrlen) + char *hdr, size_t *hdrlen) { struct git_hash_ctx c; @@ -2117,17 +2117,17 @@ int finalize_object_file_flags(const char *tmpfile, const char *filename, } static void hash_object_file_literally(const struct git_hash_algo *algo, - const void *buf, unsigned long len, + const void *buf, size_t len, const char *type, struct object_id *oid) { char hdr[MAX_HEADER_LEN]; - int hdrlen = sizeof(hdr); + size_t hdrlen = sizeof(hdr); write_object_file_prepare_literally(algo, buf, len, type, oid, hdr, &hdrlen); } void hash_object_file(const struct git_hash_algo *algo, const void *buf, - unsigned long len, enum object_type type, + size_t len, enum object_type type, struct object_id *oid) { hash_object_file_literally(algo, buf, len, type_name(type), oid); @@ -2491,7 +2491,7 @@ int stream_loose_object(struct input_stream *in_stream, size_t len, return err; } -int write_object_file_flags(const void *buf, unsigned long len, +int write_object_file_flags(const void *buf, size_t len, enum object_type type, struct object_id *oid, struct object_id *compat_oid_in, unsigned flags) { @@ -2500,7 +2500,7 @@ int write_object_file_flags(const void *buf, unsigned long len, const struct git_hash_algo *compat = repo->compat_hash_algo; struct object_id compat_oid; char hdr[MAX_HEADER_LEN]; - int hdrlen = sizeof(hdr); + size_t hdrlen = sizeof(hdr); /* Generate compat_oid */ if (compat) { @@ -2540,8 +2540,8 @@ int write_object_file_literally(const void *buf, size_t len, const struct git_hash_algo *algo = repo->hash_algo; const struct git_hash_algo *compat = repo->compat_hash_algo; struct object_id compat_oid; - int hdrlen, status = 0; - int compat_type = -1; + size_t hdrlen; + int status = 0, compat_type = -1; if (compat) { compat_type = type_from_string_gently(type, -1, 1); diff --git a/object-store-ll.h b/object-store-ll.h index d54534742f97df..f9eee3433a112c 100644 --- a/object-store-ll.h +++ b/object-store-ll.h @@ -270,10 +270,10 @@ void *repo_read_object_file(struct repository *r, int oid_object_info(struct repository *r, const struct object_id *, unsigned long *); void hash_object_file(const struct git_hash_algo *algo, const void *buf, - unsigned long len, enum object_type type, + size_t len, enum object_type type, struct object_id *oid); -int write_object_file_flags(const void *buf, unsigned long len, +int write_object_file_flags(const void *buf, size_t len, enum object_type type, struct object_id *oid, struct object_id *comapt_oid_in, unsigned flags); static inline int write_object_file(const void *buf, unsigned long len, From 73f1c2ef7883426d1160e5232a486fd7876908b5 Mon Sep 17 00:00:00 2001 From: Philip Oakley Date: Fri, 12 Nov 2021 21:16:51 +0000 Subject: [PATCH 144/312] hash algorithms: use size_t for section lengths Continue walking the code path for the >4GB `hash-object --literally` test to the hash algorithm step for LLP64 systems. This patch lets the SHA1DC code use `size_t`, making it compatible with LLP64 data models (as used e.g. by Windows). The interested reader of this patch will note that we adjust the signature of the `git_SHA1DCUpdate()` function without updating _any_ call site. This certainly puzzled at least one reviewer already, so here is an explanation: This function is never called directly, but always via the macro `platform_SHA1_Update`, which is usually called via the macro `git_SHA1_Update`. However, we never call `git_SHA1_Update()` directly in `struct git_hash_algo`. Instead, we call `git_hash_sha1_update()`, which is defined thusly: static void git_hash_sha1_update(git_hash_ctx *ctx, const void *data, size_t len) { git_SHA1_Update(&ctx->sha1, data, len); } i.e. it contains an implicit downcast from `size_t` to `unsigned long` (before this here patch). With this patch, there is no downcast anymore. With this patch, finally, the t1007-hash-object.sh "files over 4GB hash literally" test case is fixed. Signed-off-by: Philip Oakley Signed-off-by: Johannes Schindelin --- object-file.c | 4 ++-- sha1dc_git.c | 3 +-- sha1dc_git.h | 2 +- t/t1007-hash-object.sh | 2 +- 4 files changed, 5 insertions(+), 6 deletions(-) diff --git a/object-file.c b/object-file.c index 5c8870c2a7d20a..4d29bf2e3d2415 100644 --- a/object-file.c +++ b/object-file.c @@ -1952,7 +1952,7 @@ void *read_object_with_reference(struct repository *r, } static void hash_object_body(const struct git_hash_algo *algo, struct git_hash_ctx *c, - const void *buf, unsigned long len, + const void *buf, size_t len, struct object_id *oid, char *hdr, size_t *hdrlen) { @@ -1972,7 +1972,7 @@ static void write_object_file_prepare(const struct git_hash_algo *algo, /* Generate the header */ *hdrlen = format_object_header(hdr, *hdrlen, type, len); - /* Sha1.. */ + /* Hash (function pointers) computation */ hash_object_body(algo, &c, buf, len, oid, hdr, hdrlen); } diff --git a/sha1dc_git.c b/sha1dc_git.c index 9b675a046ee699..fe58d7962a30c9 100644 --- a/sha1dc_git.c +++ b/sha1dc_git.c @@ -27,10 +27,9 @@ void git_SHA1DCFinal(unsigned char hash[20], SHA1_CTX *ctx) /* * Same as SHA1DCUpdate, but adjust types to match git's usual interface. */ -void git_SHA1DCUpdate(SHA1_CTX *ctx, const void *vdata, unsigned long len) +void git_SHA1DCUpdate(SHA1_CTX *ctx, const void *vdata, size_t len) { const char *data = vdata; - /* We expect an unsigned long, but sha1dc only takes an int */ while (len > INT_MAX) { SHA1DCUpdate(ctx, data, INT_MAX); data += INT_MAX; diff --git a/sha1dc_git.h b/sha1dc_git.h index f6f880cabea382..0bcf1aa84b7241 100644 --- a/sha1dc_git.h +++ b/sha1dc_git.h @@ -15,7 +15,7 @@ void git_SHA1DCInit(SHA1_CTX *); #endif void git_SHA1DCFinal(unsigned char [20], SHA1_CTX *); -void git_SHA1DCUpdate(SHA1_CTX *ctx, const void *data, unsigned long len); +void git_SHA1DCUpdate(SHA1_CTX *ctx, const void *data, size_t len); #define platform_SHA_IS_SHA1DC /* used by "test-tool sha1-is-sha1dc" */ diff --git a/t/t1007-hash-object.sh b/t/t1007-hash-object.sh index 51604797e98b35..710139837e7a91 100755 --- a/t/t1007-hash-object.sh +++ b/t/t1007-hash-object.sh @@ -268,7 +268,7 @@ test_expect_success '--stdin outside of repository (uses SHA-1)' ' test_cmp expect actual ' -test_expect_failure EXPENSIVE,SIZE_T_IS_64BIT,!LONG_IS_64BIT \ +test_expect_success EXPENSIVE,SIZE_T_IS_64BIT,!LONG_IS_64BIT \ 'files over 4GB hash literally' ' test-tool genzeros $((5*1024*1024*1024)) >big && test_oid large5GB >expect && From a0d9eba24028fd6c6715349be569101ac71d7519 Mon Sep 17 00:00:00 2001 From: Philip Oakley Date: Mon, 6 Dec 2021 22:26:50 +0000 Subject: [PATCH 145/312] hash-object --stdin: verify that it works with >4GB/LLP64 Just like the `hash-object --literally` code path, the `--stdin` code path also needs to use `size_t` instead of `unsigned long` to represent memory sizes, otherwise it would cause problems on platforms using the LLP64 data model (such as Windows). To limit the scope of the test case, the object is explicitly not written to the object store, nor are any filters applied. The `big` file from the previous test case is reused to save setup time; To avoid relying on that side effect, it is generated if it does not exist (e.g. when running via `sh t1007-*.sh --long --run=1,41`). Signed-off-by: Philip Oakley Signed-off-by: Johannes Schindelin --- t/t1007-hash-object.sh | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/t/t1007-hash-object.sh b/t/t1007-hash-object.sh index 710139837e7a91..aa1dd51709d5cc 100755 --- a/t/t1007-hash-object.sh +++ b/t/t1007-hash-object.sh @@ -276,4 +276,12 @@ test_expect_success EXPENSIVE,SIZE_T_IS_64BIT,!LONG_IS_64BIT \ test_cmp expect actual ' +test_expect_success EXPENSIVE,SIZE_T_IS_64BIT,!LONG_IS_64BIT \ + 'files over 4GB hash correctly via --stdin' ' + { test -f big || test-tool genzeros $((5*1024*1024*1024)) >big; } && + test_oid large5GB >expect && + git hash-object --stdin actual && + test_cmp expect actual +' + test_done From dc7a0aeba8634e129e42ecab09007d98c8fd293c Mon Sep 17 00:00:00 2001 From: Philip Oakley Date: Mon, 6 Dec 2021 22:42:46 +0000 Subject: [PATCH 146/312] hash-object: add another >4GB/LLP64 test case To complement the `--stdin` and `--literally` test cases that verify that we can hash files larger than 4GB on 64-bit platforms using the LLP64 data model, here is a test case that exercises `hash-object` _without_ any options. Just as before, we use the `big` file from the previous test case if it exists to save on setup time, otherwise generate it. Signed-off-by: Philip Oakley Signed-off-by: Johannes Schindelin --- t/t1007-hash-object.sh | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/t/t1007-hash-object.sh b/t/t1007-hash-object.sh index aa1dd51709d5cc..7c6b7913d4f23d 100755 --- a/t/t1007-hash-object.sh +++ b/t/t1007-hash-object.sh @@ -284,4 +284,12 @@ test_expect_success EXPENSIVE,SIZE_T_IS_64BIT,!LONG_IS_64BIT \ test_cmp expect actual ' +test_expect_success EXPENSIVE,SIZE_T_IS_64BIT,!LONG_IS_64BIT \ + 'files over 4GB hash correctly' ' + { test -f big || test-tool genzeros $((5*1024*1024*1024)) >big; } && + test_oid large5GB >expect && + git hash-object -- big >actual && + test_cmp expect actual +' + test_done From d72ac165b1a5461db6dedd271662f30613e1edf4 Mon Sep 17 00:00:00 2001 From: Derrick Stolee Date: Wed, 13 Apr 2022 14:49:17 -0400 Subject: [PATCH 147/312] setup: properly use "%(prefix)/" when in WSL Signed-off-by: Derrick Stolee --- setup.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) diff --git a/setup.c b/setup.c index f93bd6a24a5d9c..6791111992c5bb 100644 --- a/setup.c +++ b/setup.c @@ -1781,10 +1781,19 @@ const char *setup_git_directory_gently(int *nongit_ok) break; case GIT_DIR_INVALID_OWNERSHIP: if (!nongit_ok) { + struct strbuf prequoted = STRBUF_INIT; struct strbuf quoted = STRBUF_INIT; strbuf_complete(&report, '\n'); - sq_quote_buf_pretty("ed, dir.buf); + +#ifdef __MINGW32__ + if (dir.buf[0] == '/') + strbuf_addstr(&prequoted, "%(prefix)/"); +#endif + + strbuf_add(&prequoted, dir.buf, dir.len); + sq_quote_buf_pretty("ed, prequoted.buf); + die(_("detected dubious ownership in repository at '%s'\n" "%s" "To add an exception for this directory, call:\n" From 4e53e830e6d18c2899879886b8c08bf0905225b3 Mon Sep 17 00:00:00 2001 From: Philip Oakley Date: Tue, 7 Dec 2021 09:53:41 +0000 Subject: [PATCH 148/312] hash-object: add a >4GB/LLP64 test case using filtered input To verify that the `clean` side of the `clean`/`smudge` filter code is correct with regards to LLP64 (read: to ensure that `size_t` is used instead of `unsigned long`), here is a test case using a trivial filter, specifically _not_ writing anything to the object store to limit the scope of the test case. As in previous commits, the `big` file from previous test cases is reused if available, to save setup time, otherwise re-generated. Signed-off-by: Philip Oakley Signed-off-by: Johannes Schindelin --- t/t1007-hash-object.sh | 12 ++++++++++++ 1 file changed, 12 insertions(+) diff --git a/t/t1007-hash-object.sh b/t/t1007-hash-object.sh index 7c6b7913d4f23d..5850c27a9cc315 100755 --- a/t/t1007-hash-object.sh +++ b/t/t1007-hash-object.sh @@ -292,4 +292,16 @@ test_expect_success EXPENSIVE,SIZE_T_IS_64BIT,!LONG_IS_64BIT \ test_cmp expect actual ' +# This clean filter does nothing, other than excercising the interface. +# We ensure that cleaning doesn't mangle large files on 64-bit Windows. +test_expect_success EXPENSIVE,SIZE_T_IS_64BIT,!LONG_IS_64BIT \ + 'hash filtered files over 4GB correctly' ' + { test -f big || test-tool genzeros $((5*1024*1024*1024)) >big; } && + test_oid large5GB >expect && + test_config filter.null-filter.clean "cat" && + echo "big filter=null-filter" >.gitattributes && + git hash-object -- big >actual && + test_cmp expect actual +' + test_done From a37cf6e06f713b12210f2dfbf554fa0eefd23277 Mon Sep 17 00:00:00 2001 From: Derrick Stolee Date: Wed, 13 Apr 2022 14:54:43 -0400 Subject: [PATCH 149/312] compat/mingw.c: do not warn when failing to get owner In the case of Git for Windows (say, in a Git Bash window) running in a Windows Subsystem for Linux (WSL) directory, the GetNamedSecurityInfoW() call in is_path_owned_By_current_side() returns an error code other than ERROR_SUCCESS. This is consistent behavior across this boundary. In these cases, the owner would always be different because the WSL owner is a different entity than the Windows user. The change here is to suppress the error message that looks like this: error: failed to get owner for '//wsl.localhost/...' (1) Before this change, this warning happens for every Git command, regardless of whether the directory is marked with safe.directory. Signed-off-by: Derrick Stolee --- compat/mingw.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) diff --git a/compat/mingw.c b/compat/mingw.c index f524c54d06d965..4f0833b3523208 100644 --- a/compat/mingw.c +++ b/compat/mingw.c @@ -2926,9 +2926,7 @@ int is_path_owned_by_current_sid(const char *path, struct strbuf *report) DACL_SECURITY_INFORMATION, &sid, NULL, NULL, NULL, &descriptor); - if (err != ERROR_SUCCESS) - error(_("failed to get owner for '%s' (%ld)"), path, err); - else if (sid && IsValidSid(sid)) { + if (err == ERROR_SUCCESS && sid && IsValidSid(sid)) { /* Now, verify that the SID matches the current user's */ static PSID current_user_sid; BOOL is_member; From 7e994f9f38944eda33a5ed70657a1e70aae869c0 Mon Sep 17 00:00:00 2001 From: Rafael Kitover Date: Tue, 12 Apr 2022 19:53:33 +0000 Subject: [PATCH 150/312] mingw: $env:TERM="xterm-256color" for newer OSes For Windows builds >= 15063 set $env:TERM to "xterm-256color" instead of "cygwin" because they have a more capable console system that supports this. Also set $env:COLORTERM="truecolor" if unset. $env:TERM is initialized so that ANSI colors in color.c work, see 29a3963484 (Win32: patch Windows environment on startup, 2012-01-15). See git-for-windows/git#3629 regarding problems caused by always setting $env:TERM="cygwin". This is the same heuristic used by the Cygwin runtime. Signed-off-by: Rafael Kitover Signed-off-by: Johannes Schindelin --- compat/mingw.c | 17 ++++++++++++++--- 1 file changed, 14 insertions(+), 3 deletions(-) diff --git a/compat/mingw.c b/compat/mingw.c index f524c54d06d965..301d28b8d173a9 100644 --- a/compat/mingw.c +++ b/compat/mingw.c @@ -2797,9 +2797,20 @@ static void setup_windows_environment(void) convert_slashes(tmp); } - /* simulate TERM to enable auto-color (see color.c) */ - if (!getenv("TERM")) - setenv("TERM", "cygwin", 1); + + /* + * Make sure TERM is set up correctly to enable auto-color + * (see color.c .) Use "cygwin" for older OS releases which + * works correctly with MSYS2 utilities on older consoles. + */ + if (!getenv("TERM")) { + if ((GetVersion() >> 16) < 15063) + setenv("TERM", "cygwin", 0); + else { + setenv("TERM", "xterm-256color", 0); + setenv("COLORTERM", "truecolor", 0); + } + } /* calculate HOME if not set */ if (!getenv("HOME")) { From b7695062ad78850a156130083a8de40c99ca211b Mon Sep 17 00:00:00 2001 From: Christopher Degawa Date: Sat, 28 May 2022 14:53:54 -0500 Subject: [PATCH 151/312] winansi: check result and Buffer before using Name NtQueryObject under Wine can return a success but fill out no name. In those situations, Wine will set Buffer to NULL, and set result to the sizeof(OBJECT_NAME_INFORMATION). Running a command such as echo "$(git.exe --version 2>/dev/null)" will crash due to a NULL pointer dereference when the code attempts to null terminate the buffer, although, weirdly, removing the subshell or redirecting stdout to a file will not trigger the crash. Code has been added to also check Buffer and Length to ensure the check is as robust as possible due to the current behavior being fragile at best, and could potentially change in the future This code is based on the behavior of NtQueryObject under wine and reactos. Signed-off-by: Christopher Degawa --- compat/winansi.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/compat/winansi.c b/compat/winansi.c index ac2ffb78691a7d..d28137a20b0bcc 100644 --- a/compat/winansi.c +++ b/compat/winansi.c @@ -575,6 +575,9 @@ static void detect_msys_tty(int fd) if (!NT_SUCCESS(NtQueryObject(h, ObjectNameInformation, buffer, sizeof(buffer) - 2, &result))) return; + if (result < sizeof(*nameinfo) || !nameinfo->Name.Buffer || + !nameinfo->Name.Length) + return; name = nameinfo->Name.Buffer; name[nameinfo->Name.Length / sizeof(*name)] = 0; From 461b25607433da204158048f407f6cb1974f1e60 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E5=AD=99=E5=8D=93=E8=AF=86?= Date: Sun, 16 Jan 2022 03:38:33 +0800 Subject: [PATCH 152/312] Add config option `windows.appendAtomically` MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Atomic append on windows is only supported on local disk files, and it may cause errors in other situations, e.g. network file system. If that is the case, this config option should be used to turn atomic append off. Co-Authored-By: Johannes Schindelin Signed-off-by: 孙卓识 Signed-off-by: Johannes Schindelin --- Documentation/config.adoc | 2 ++ Documentation/config/windows.adoc | 4 ++++ compat/mingw.c | 36 ++++++++++++++++++++++++++++--- 3 files changed, 39 insertions(+), 3 deletions(-) create mode 100644 Documentation/config/windows.adoc diff --git a/Documentation/config.adoc b/Documentation/config.adoc index a8b04c4e5183d2..30e11203c87d1c 100644 --- a/Documentation/config.adoc +++ b/Documentation/config.adoc @@ -558,4 +558,6 @@ include::config/versionsort.adoc[] include::config/web.adoc[] +include::config/windows.adoc[] + include::config/worktree.adoc[] diff --git a/Documentation/config/windows.adoc b/Documentation/config/windows.adoc new file mode 100644 index 00000000000000..fdaaf1c65504f3 --- /dev/null +++ b/Documentation/config/windows.adoc @@ -0,0 +1,4 @@ +windows.appendAtomically:: + By default, append atomic API is used on windows. But it works only with + local disk files, if you're working on a network file system, you should + set it false to turn it off. diff --git a/compat/mingw.c b/compat/mingw.c index f524c54d06d965..0658cdfc3c733f 100644 --- a/compat/mingw.c +++ b/compat/mingw.c @@ -21,6 +21,7 @@ #include "gettext.h" #define SECURITY_WIN32 #include +#include "../repository.h" #define HCAST(type, handle) ((type)(intptr_t)handle) @@ -614,6 +615,7 @@ static int is_local_named_pipe_path(const char *filename) int mingw_open (const char *filename, int oflags, ...) { + static int append_atomically = -1; typedef int (*open_fn_t)(wchar_t const *wfilename, int oflags, ...); va_list args; unsigned mode; @@ -630,7 +632,16 @@ int mingw_open (const char *filename, int oflags, ...) return -1; } - if ((oflags & O_APPEND) && !is_local_named_pipe_path(filename)) + /* + * Only set append_atomically to default value(1) when repo is initialized + * and fail to get config value + */ + if (append_atomically < 0 && the_repository && the_repository->commondir && + git_config_get_bool("windows.appendatomically", &append_atomically)) + append_atomically = 1; + + if (append_atomically && (oflags & O_APPEND) && + !is_local_named_pipe_path(filename)) open_fn = mingw_open_append; else if (!(oflags & ~(O_ACCMODE | O_NOINHERIT))) open_fn = mingw_open_existing; @@ -781,9 +792,28 @@ ssize_t mingw_write(int fd, const void *buf, size_t len) /* check if fd is a pipe */ HANDLE h = (HANDLE) _get_osfhandle(fd); - if (GetFileType(h) != FILE_TYPE_PIPE) + if (GetFileType(h) != FILE_TYPE_PIPE) { + if (orig == EINVAL) { + wchar_t path[MAX_PATH]; + DWORD ret = GetFinalPathNameByHandleW(h, path, + ARRAY_SIZE(path), 0); + UINT drive_type = ret > 0 && ret < ARRAY_SIZE(path) ? + GetDriveTypeW(path) : DRIVE_UNKNOWN; + + /* + * The default atomic append causes such an error on + * network file systems, in such a case, it should be + * turned off via config. + * + * `drive_type` of UNC path: DRIVE_NO_ROOT_DIR + */ + if (DRIVE_NO_ROOT_DIR == drive_type || DRIVE_REMOTE == drive_type) + warning("invalid write operation detected; you may try:\n" + "\n\tgit config windows.appendAtomically false"); + } + errno = orig; - else if (orig == EINVAL) + } else if (orig == EINVAL) errno = EPIPE; else { DWORD buf_size; From 7fabd872c56cfca675c44830f60326b8c61f9ac5 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Mon, 4 Sep 2017 11:59:45 +0200 Subject: [PATCH 153/312] mingw: change core.fsyncObjectFiles = 1 by default MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit From the documentation of said setting: This boolean will enable fsync() when writing object files. This is a total waste of time and effort on a filesystem that orders data writes properly, but can be useful for filesystems that do not use journalling (traditional UNIX filesystems) or that only journal metadata and not file contents (OS X’s HFS+, or Linux ext3 with "data=writeback"). The most common file system on Windows (NTFS) does not guarantee that order, therefore a sudden loss of power (or any other event causing an unclean shutdown) would cause corrupt files (i.e. files filled with NULs). Therefore we need to change the default. Note that the documentation makes it sound as if this causes really bad performance. In reality, writing loose objects is something that is done only rarely, and only a handful of files at a time. Signed-off-by: Johannes Schindelin --- compat/mingw.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/compat/mingw.c b/compat/mingw.c index 0658cdfc3c733f..7a19f84e246f1a 100644 --- a/compat/mingw.c +++ b/compat/mingw.c @@ -21,6 +21,7 @@ #include "gettext.h" #define SECURITY_WIN32 #include +#include "../write-or-die.h" #include "../repository.h" #define HCAST(type, handle) ((type)(intptr_t)handle) @@ -3285,6 +3286,7 @@ int wmain(int argc, const wchar_t **wargv) #endif maybe_redirect_std_handles(); + fsync_object_files = 1; /* determine size of argv and environ conversion buffer */ maxlen = wcslen(wargv[0]); From 1d2d350cee0250ba48c68e545696e6c6eeef932f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Matthias=20A=C3=9Fhauer?= Date: Sun, 10 Jul 2022 11:27:25 +0200 Subject: [PATCH 154/312] MinGW: link as terminal server aware MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Whith Windows 2000, Microsoft introduced a flag to the PE header to mark executables as "terminal server aware". Windows terminal servers provide a redirected Windows directory and redirected registry hives when launching legacy applications without this flag set. Since we do not use any INI files in the Windows directory and don't write to the registry, we don't need this additional preparation. Telling the OS that we don't need this should provide slightly improved startup times in terminal server environments. When building for supported Windows Versions with MSVC the /TSAWARE linker flag is automatically set, but MinGW requires us to set the --tsaware flag manually. This partially addresses https://github.com/git-for-windows/git/issues/3935. Signed-off-by: Matthias Aßhauer --- config.mak.uname | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/config.mak.uname b/config.mak.uname index b12d4e168ae119..1736c4f0116e75 100644 --- a/config.mak.uname +++ b/config.mak.uname @@ -691,7 +691,7 @@ ifeq ($(uname_S),MINGW) DEFAULT_HELP_FORMAT = html HAVE_PLATFORM_PROCINFO = YesPlease CSPRNG_METHOD = rtlgenrandom - BASIC_LDFLAGS += -municode + BASIC_LDFLAGS += -municode -Wl,--tsaware COMPAT_CFLAGS += -DNOGDI -Icompat -Icompat/win32 COMPAT_CFLAGS += -DSTRIP_EXTENSION=\".exe\" COMPAT_OBJS += compat/mingw.o compat/winansi.o \ From da7eb94b5ba750963d9fdd1cdc4ba041f05f5dcf Mon Sep 17 00:00:00 2001 From: Dennis Ameling Date: Tue, 4 Oct 2022 09:59:32 +0200 Subject: [PATCH 155/312] config.mak.uname: add support for clangarm64 CLANGARM64 is a relatively new MSYSTEM added by the MSYS2 team. In order to have Git build correctly for this platform, let's add some configuration for it to config.mak.uname. Signed-off-by: Dennis Ameling --- config.mak.uname | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/config.mak.uname b/config.mak.uname index 2ebb82feb9f1b2..bf0247c3ecc841 100644 --- a/config.mak.uname +++ b/config.mak.uname @@ -724,6 +724,10 @@ ifeq ($(uname_S),MINGW) prefix = /mingw64 HOST_CPU = x86_64 BASIC_LDFLAGS += -Wl,--pic-executable,-e,mainCRTStartup + else ifeq (CLANGARM64,$(MSYSTEM)) + prefix = /clangarm64 + HOST_CPU = aarch64 + BASIC_LDFLAGS += -Wl,--pic-executable,-e,mainCRTStartup else COMPAT_CFLAGS += -D_USE_32BIT_TIME_T BASIC_LDFLAGS += -Wl,--large-address-aware From 3fd5dffe9a1ef3eeb377fe8559e70ebc5c8f2538 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Tue, 25 Feb 2025 10:17:12 +0000 Subject: [PATCH 156/312] msvc: do handle builds on Windows/ARM64 Git for Windows/ARM64 settled on using `clang` to compile `git.exe`, and hence needs to run in a system where `MSYSTEM` is set to `CLANGARM64` and the prefix to use is `/clangarm64`. We already did that in the `MINGW` arm, i.e. for regular Git for Windows builds using MINGW GCC (or `clang`'s shim pretending to be GCC), now it is time to do the same in the MS Visual C part. Signed-off-by: Johannes Schindelin --- config.mak.uname | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/config.mak.uname b/config.mak.uname index bf0247c3ecc841..a955b69527bffb 100644 --- a/config.mak.uname +++ b/config.mak.uname @@ -432,7 +432,11 @@ ifeq ($(uname_S),Windows) ifeq (MINGW32,$(MSYSTEM)) prefix = /mingw32 else - prefix = /mingw64 + ifeq (CLANGARM64,$(MSYSTEM)) + prefix = /clangarm64 + else + prefix = /mingw64 + endif endif # Prepend MSVC 64-bit tool-chain to PATH. # From e52c2736ff40a4058c7a9c87c46ea4af00a000df Mon Sep 17 00:00:00 2001 From: Kiel Hurley Date: Wed, 2 Nov 2022 22:56:16 +1300 Subject: [PATCH 157/312] Fix Windows version resources Add FileVersion, which is a required field As not all required fields were present, none were being included Fixes #4090 Signed-off-by: Kiel Hurley --- git.rc.in | 1 + 1 file changed, 1 insertion(+) diff --git a/git.rc.in b/git.rc.in index e69444eef3f0c5..460ea39561b87f 100644 --- a/git.rc.in +++ b/git.rc.in @@ -12,6 +12,7 @@ BEGIN VALUE "OriginalFilename", "git.exe\0" VALUE "ProductName", "Git\0" VALUE "ProductVersion", "@GIT_VERSION@\0" + VALUE "FileVersion", "@GIT_VERSION@\0" END END From dd00e9b164319fe5e3b3026d31c41aeb8932bebf Mon Sep 17 00:00:00 2001 From: Dennis Ameling Date: Tue, 4 Oct 2022 09:58:10 +0200 Subject: [PATCH 158/312] bswap.h: add support for built-in bswap functions Newer compiler versions, like GCC 10 and Clang 12, have built-in functions for bswap32 and bswap64. This comes in handy, for example, when targeting CLANGARM64 on Windows, which would not be supported without this logic. Signed-off-by: Dennis Ameling --- compat/bswap.h | 14 +++++++++++++- 1 file changed, 13 insertions(+), 1 deletion(-) diff --git a/compat/bswap.h b/compat/bswap.h index b34054f2bd7284..9e0f98e00b93a4 100644 --- a/compat/bswap.h +++ b/compat/bswap.h @@ -35,7 +35,19 @@ static inline uint64_t default_bswap64(uint64_t val) #undef bswap32 #undef bswap64 -#if defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__)) +/** + * __has_builtin is available since Clang 10 and GCC 10. + * Below is a fallback for older compilers. + */ +#ifndef __has_builtin + #define __has_builtin(x) 0 +#endif + +#if __has_builtin(__builtin_bswap32) && __has_builtin(__builtin_bswap64) +#define bswap32(x) __builtin_bswap32((x)) +#define bswap64(x) __builtin_bswap64((x)) + +#elif defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__)) #define bswap32 git_bswap32 static inline uint32_t git_bswap32(uint32_t x) From cba5b4bc2670f37d286eab11edf7c23aa95345dd Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Tue, 25 Feb 2025 10:21:52 +0000 Subject: [PATCH 159/312] mingw(arm64): do move the `/etc/git*` location In fb5e3378f8 (mingw: move Git for Windows' system config where users expect it, 2021-06-22), I moved the location of Git for Windows' system config and system Git attributes file to the top-level `/etc/` directory (because it is a much more obvious location than, say, `/mingw64/etc/`). The patch relied on a very specific scenario that the newly-supported Windows/ARM64 builds of `git.exe` fails to fall into. So let's broaden the condition a bit, so that Windows/ARM64 builds also use that location (instead of the even more obscure `/clangarm64/etc/` directory). This fixes https://github.com/git-for-windows/git/issues/5431. Signed-off-by: Johannes Schindelin --- config.mak.uname | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/config.mak.uname b/config.mak.uname index a955b69527bffb..0229eb3b8fe21f 100644 --- a/config.mak.uname +++ b/config.mak.uname @@ -489,7 +489,7 @@ ifeq ($(uname_S),Windows) NO_POSIX_GOODIES = UnfortunatelyYes NATIVE_CRLF = YesPlease DEFAULT_HELP_FORMAT = html -ifeq (/mingw64,$(subst 32,64,$(prefix))) +ifeq (/mingw64,$(subst 32,64,$(subst clangarm,mingw,$(prefix)))) # Move system config into top-level /etc/ ETC_GITCONFIG = ../etc/gitconfig ETC_GITATTRIBUTES = ../etc/gitattributes @@ -748,7 +748,7 @@ ifeq ($(uname_S),MINGW) USE_LIBPCRE = YesPlease USE_MIMALLOC = YesPlease NO_PYTHON = - ifeq (/mingw64,$(subst 32,64,$(prefix))) + ifeq (/mingw64,$(subst 32,64,$(subst clangarm,mingw,$(prefix)))) # Move system config into top-level /etc/ ETC_GITCONFIG = ../etc/gitconfig ETC_GITATTRIBUTES = ../etc/gitattributes From e7f0009d69817cc01f20be4f5ec354bc6b586569 Mon Sep 17 00:00:00 2001 From: Andrey Zabavnikov Date: Fri, 28 Oct 2022 17:12:06 +0300 Subject: [PATCH 160/312] status: fix for old-style submodules with commondir In f9b7573f6b00 (repository: free fields before overwriting them, 2017-09-05), Git was taught to release memory before overwriting it, but 357a03ebe9e0 (repository.c: move env-related setup code back to environment.c, 2018-03-03) changed the code so that it would not _always_ be overwritten. As a consequence, the `commondir` attribute would point to already-free()d memory. This seems not to cause problems in core Git, but there are add-on patches in Git for Windows where the `commondir` attribute is subsequently used and causing invalid memory accesses e.g. in setups containing old-style submodules (i.e. the ones with a `.git` directory within theirs worktrees) that have `commondir` configured. This fixes https://github.com/git-for-windows/git/pull/4083. Signed-off-by: Andrey Zabavnikov Signed-off-by: Johannes Schindelin --- repository.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/repository.c b/repository.c index 6cbaf2e3daa93a..6cf423da50adec 100644 --- a/repository.c +++ b/repository.c @@ -136,7 +136,7 @@ static void repo_set_commondir(struct repository *repo, { struct strbuf sb = STRBUF_INIT; - free(repo->commondir); + FREE_AND_NULL(repo->commondir); if (commondir) { repo->different_commondir = 1; From fd6be1d5ccebf0414882651ff393179ca422cec4 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Fri, 27 Jan 2023 08:55:21 +0100 Subject: [PATCH 161/312] windows: skip linking `git-` for built-ins It is merely a historical wart that, say, `git-commit` exists in the `libexec/git-core/` directory, a tribute to the original idea to let Git be essentially a bunch of Unix shell scripts revolving around very few "plumbing" (AKA low-level) commands. Git has evolved a lot from there. These days, most of Git's functionality is contained within the `git` executable, in the form of "built-in" commands. To accommodate for scripts that use the "dashed" form of Git commands, even today, Git provides hard-links that make the `git` executable available as, say, `git-commit`, just in case that an old script has not been updated to invoke `git commit`. Those hard-links do not come cheap: they take about half a minute for every build of Git on Windows, they are mistaken for taking up huge amounts of space by some Windows Explorer versions that do not understand hard-links, and therefore many a "bug" report had to be addressed. The "dashed form" has been officially deprecated in Git version 1.5.4, which was released on February 2nd, 2008, i.e. a very long time ago. This deprecation was never finalized by skipping these hard-links, but we can start the process now, in Git for Windows. Signed-off-by: Johannes Schindelin --- config.mak.uname | 2 ++ 1 file changed, 2 insertions(+) diff --git a/config.mak.uname b/config.mak.uname index 0229eb3b8fe21f..1f78fbad5f4067 100644 --- a/config.mak.uname +++ b/config.mak.uname @@ -489,6 +489,7 @@ ifeq ($(uname_S),Windows) NO_POSIX_GOODIES = UnfortunatelyYes NATIVE_CRLF = YesPlease DEFAULT_HELP_FORMAT = html + SKIP_DASHED_BUILT_INS = YabbaDabbaDoo ifeq (/mingw64,$(subst 32,64,$(subst clangarm,mingw,$(prefix)))) # Move system config into top-level /etc/ ETC_GITCONFIG = ../etc/gitconfig @@ -681,6 +682,7 @@ ifeq ($(uname_S),MINGW) FSMONITOR_DAEMON_BACKEND = win32 FSMONITOR_OS_SETTINGS = win32 + SKIP_DASHED_BUILT_INS = YabbaDabbaDoo RUNTIME_PREFIX = YesPlease HAVE_WPGMPTR = YesWeDo NO_ST_BLOCKS_IN_STRUCT_STAT = YesPlease From 1dd6dc800d935bc69c2a8168e9cbd2372fead007 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Sat, 6 May 2023 22:26:15 +0200 Subject: [PATCH 162/312] http: optionally load libcurl lazily This compile-time option allows to ask Git to load libcurl dynamically at runtime. Together with a follow-up patch that optionally overrides the file name depending on the `http.sslBackend` setting, this kicks open the door for installing multiple libcurl flavors side by side, and load the one corresponding to the (runtime-)configured SSL/TLS backend. Signed-off-by: Johannes Schindelin --- Makefile | 28 +++- compat/lazyload-curl.c | 354 +++++++++++++++++++++++++++++++++++++++++ 2 files changed, 375 insertions(+), 7 deletions(-) create mode 100644 compat/lazyload-curl.c diff --git a/Makefile b/Makefile index 7315507381e146..59689d92fd0261 100644 --- a/Makefile +++ b/Makefile @@ -475,6 +475,11 @@ include shared.mak # # CURL_LDFLAGS=-lcurl # +# Define LAZYLOAD_LIBCURL to dynamically load the libcurl; This can be useful +# if Multiple libcurl versions exist (with different file names) that link to +# various SSL/TLS backends, to support the `http.sslBackend` runtime switch in +# such a scenario. +# # === Optional library: libpcre2 === # # Define USE_LIBPCRE if you have and want to use libpcre. Various @@ -1661,10 +1666,19 @@ else CURL_LIBCURL = endif - ifndef CURL_LDFLAGS - CURL_LDFLAGS = $(eval CURL_LDFLAGS := $$(shell $$(CURL_CONFIG) --libs))$(CURL_LDFLAGS) + ifdef LAZYLOAD_LIBCURL + LAZYLOAD_LIBCURL_OBJ = compat/lazyload-curl.o + OBJECTS += $(LAZYLOAD_LIBCURL_OBJ) + # The `CURL_STATICLIB` constant must be defined to avoid seeing the functions + # declared as DLL imports + CURL_CFLAGS = -DCURL_STATICLIB + CURL_LIBCURL = -ldl + else + ifndef CURL_LDFLAGS + CURL_LDFLAGS = $(eval CURL_LDFLAGS := $$(shell $$(CURL_CONFIG) --libs))$(CURL_LDFLAGS) + endif + CURL_LIBCURL += $(CURL_LDFLAGS) endif - CURL_LIBCURL += $(CURL_LDFLAGS) ifndef CURL_CFLAGS CURL_CFLAGS = $(eval CURL_CFLAGS := $$(shell $$(CURL_CONFIG) --cflags))$(CURL_CFLAGS) @@ -1685,7 +1699,7 @@ else endif ifdef USE_CURL_FOR_IMAP_SEND BASIC_CFLAGS += -DUSE_CURL_FOR_IMAP_SEND - IMAP_SEND_BUILDDEPS = http.o + IMAP_SEND_BUILDDEPS = http.o $(LAZYLOAD_LIBCURL_OBJ) IMAP_SEND_LDFLAGS += $(CURL_LIBCURL) endif ifndef NO_EXPAT @@ -2886,10 +2900,10 @@ git-imap-send$X: imap-send.o $(IMAP_SEND_BUILDDEPS) GIT-LDFLAGS $(GITLIBS) $(QUIET_LINK)$(CC) $(ALL_CFLAGS) -o $@ $(ALL_LDFLAGS) $(filter %.o,$^) \ $(IMAP_SEND_LDFLAGS) $(LIBS) -git-http-fetch$X: http.o http-walker.o http-fetch.o GIT-LDFLAGS $(GITLIBS) +git-http-fetch$X: http.o http-walker.o http-fetch.o $(LAZYLOAD_LIBCURL_OBJ) GIT-LDFLAGS $(GITLIBS) $(QUIET_LINK)$(CC) $(ALL_CFLAGS) -o $@ $(ALL_LDFLAGS) $(filter %.o,$^) \ $(CURL_LIBCURL) $(LIBS) -git-http-push$X: http.o http-push.o GIT-LDFLAGS $(GITLIBS) +git-http-push$X: http.o http-push.o $(LAZYLOAD_LIBCURL_OBJ) GIT-LDFLAGS $(GITLIBS) $(QUIET_LINK)$(CC) $(ALL_CFLAGS) -o $@ $(ALL_LDFLAGS) $(filter %.o,$^) \ $(CURL_LIBCURL) $(EXPAT_LIBEXPAT) $(LIBS) @@ -2899,7 +2913,7 @@ $(REMOTE_CURL_ALIASES): $(REMOTE_CURL_PRIMARY) ln -s $< $@ 2>/dev/null || \ cp $< $@ -$(REMOTE_CURL_PRIMARY): remote-curl.o http.o http-walker.o GIT-LDFLAGS $(GITLIBS) +$(REMOTE_CURL_PRIMARY): remote-curl.o http.o http-walker.o $(LAZYLOAD_LIBCURL_OBJ) GIT-LDFLAGS $(GITLIBS) $(QUIET_LINK)$(CC) $(ALL_CFLAGS) -o $@ $(ALL_LDFLAGS) $(filter %.o,$^) \ $(CURL_LIBCURL) $(EXPAT_LIBEXPAT) $(LIBS) diff --git a/compat/lazyload-curl.c b/compat/lazyload-curl.c new file mode 100644 index 00000000000000..19aa2b6d4b6942 --- /dev/null +++ b/compat/lazyload-curl.c @@ -0,0 +1,354 @@ +#include "../git-compat-util.h" +#include "../git-curl-compat.h" +#include + +/* + * The ABI version of libcurl is encoded in its shared libraries' file names. + * This ABI version has not changed since October 2006 and is unlikely to be + * changed in the future. See https://curl.se/libcurl/abi.html for details. + */ +#define LIBCURL_ABI_VERSION "4" + +typedef void (*func_t)(void); + +#ifdef __APPLE__ +#define LIBCURL_FILE_NAME(base) base "." LIBCURL_ABI_VERSION ".dylib" +#else +#define LIBCURL_FILE_NAME(base) base ".so." LIBCURL_ABI_VERSION +#endif + +static void *load_library(const char *name) +{ + return dlopen(name, RTLD_LAZY); +} + +static func_t load_function(void *handle, const char *name) +{ + /* + * Casting the return value of `dlsym()` to a function pointer is + * explicitly allowed in recent POSIX standards, but GCC complains + * about this in pedantic mode nevertheless. For more about this issue, + * see https://stackoverflow.com/q/31526876/1860823 and + * http://stackoverflow.com/a/36385690/1905491. + */ + func_t f; + *(void **)&f = dlsym(handle, name); + return f; +} + +typedef struct curl_version_info_data *(*curl_version_info_type)(CURLversion version); +static curl_version_info_type curl_version_info_func; + +typedef char *(*curl_easy_escape_type)(CURL *handle, const char *string, int length); +static curl_easy_escape_type curl_easy_escape_func; + +typedef void (*curl_free_type)(void *p); +static curl_free_type curl_free_func; + +typedef CURLcode (*curl_global_init_type)(long flags); +static curl_global_init_type curl_global_init_func; + +typedef CURLsslset (*curl_global_sslset_type)(curl_sslbackend id, const char *name, const curl_ssl_backend ***avail); +static curl_global_sslset_type curl_global_sslset_func; + +typedef void (*curl_global_cleanup_type)(void); +static curl_global_cleanup_type curl_global_cleanup_func; + +typedef struct curl_slist *(*curl_slist_append_type)(struct curl_slist *list, const char *data); +static curl_slist_append_type curl_slist_append_func; + +typedef void (*curl_slist_free_all_type)(struct curl_slist *list); +static curl_slist_free_all_type curl_slist_free_all_func; + +typedef const char *(*curl_easy_strerror_type)(CURLcode error); +static curl_easy_strerror_type curl_easy_strerror_func; + +typedef CURLM *(*curl_multi_init_type)(void); +static curl_multi_init_type curl_multi_init_func; + +typedef CURLMcode (*curl_multi_add_handle_type)(CURLM *multi_handle, CURL *curl_handle); +static curl_multi_add_handle_type curl_multi_add_handle_func; + +typedef CURLMcode (*curl_multi_remove_handle_type)(CURLM *multi_handle, CURL *curl_handle); +static curl_multi_remove_handle_type curl_multi_remove_handle_func; + +typedef CURLMcode (*curl_multi_fdset_type)(CURLM *multi_handle, fd_set *read_fd_set, fd_set *write_fd_set, fd_set *exc_fd_set, int *max_fd); +static curl_multi_fdset_type curl_multi_fdset_func; + +typedef CURLMcode (*curl_multi_perform_type)(CURLM *multi_handle, int *running_handles); +static curl_multi_perform_type curl_multi_perform_func; + +typedef CURLMcode (*curl_multi_cleanup_type)(CURLM *multi_handle); +static curl_multi_cleanup_type curl_multi_cleanup_func; + +typedef CURLMsg *(*curl_multi_info_read_type)(CURLM *multi_handle, int *msgs_in_queue); +static curl_multi_info_read_type curl_multi_info_read_func; + +typedef const char *(*curl_multi_strerror_type)(CURLMcode error); +static curl_multi_strerror_type curl_multi_strerror_func; + +typedef CURLMcode (*curl_multi_timeout_type)(CURLM *multi_handle, long *milliseconds); +static curl_multi_timeout_type curl_multi_timeout_func; + +typedef CURL *(*curl_easy_init_type)(void); +static curl_easy_init_type curl_easy_init_func; + +typedef CURLcode (*curl_easy_perform_type)(CURL *curl); +static curl_easy_perform_type curl_easy_perform_func; + +typedef void (*curl_easy_cleanup_type)(CURL *curl); +static curl_easy_cleanup_type curl_easy_cleanup_func; + +typedef CURL *(*curl_easy_duphandle_type)(CURL *curl); +static curl_easy_duphandle_type curl_easy_duphandle_func; + +typedef CURLcode (*curl_easy_getinfo_long_type)(CURL *curl, CURLINFO info, long *value); +static curl_easy_getinfo_long_type curl_easy_getinfo_long_func; + +typedef CURLcode (*curl_easy_getinfo_pointer_type)(CURL *curl, CURLINFO info, void **value); +static curl_easy_getinfo_pointer_type curl_easy_getinfo_pointer_func; + +typedef CURLcode (*curl_easy_getinfo_off_t_type)(CURL *curl, CURLINFO info, curl_off_t *value); +static curl_easy_getinfo_off_t_type curl_easy_getinfo_off_t_func; + +typedef CURLcode (*curl_easy_setopt_long_type)(CURL *curl, CURLoption opt, long value); +static curl_easy_setopt_long_type curl_easy_setopt_long_func; + +typedef CURLcode (*curl_easy_setopt_pointer_type)(CURL *curl, CURLoption opt, void *value); +static curl_easy_setopt_pointer_type curl_easy_setopt_pointer_func; + +typedef CURLcode (*curl_easy_setopt_off_t_type)(CURL *curl, CURLoption opt, curl_off_t value); +static curl_easy_setopt_off_t_type curl_easy_setopt_off_t_func; + +static void lazy_load_curl(void) +{ + static int initialized; + void *libcurl; + func_t curl_easy_getinfo_func, curl_easy_setopt_func; + + if (initialized) + return; + + initialized = 1; + libcurl = load_library(LIBCURL_FILE_NAME("libcurl")); + if (!libcurl) + die("failed to load library '%s'", LIBCURL_FILE_NAME("libcurl")); + + curl_version_info_func = (curl_version_info_type)load_function(libcurl, "curl_version_info"); + curl_easy_escape_func = (curl_easy_escape_type)load_function(libcurl, "curl_easy_escape"); + curl_free_func = (curl_free_type)load_function(libcurl, "curl_free"); + curl_global_init_func = (curl_global_init_type)load_function(libcurl, "curl_global_init"); + curl_global_sslset_func = (curl_global_sslset_type)load_function(libcurl, "curl_global_sslset"); + curl_global_cleanup_func = (curl_global_cleanup_type)load_function(libcurl, "curl_global_cleanup"); + curl_slist_append_func = (curl_slist_append_type)load_function(libcurl, "curl_slist_append"); + curl_slist_free_all_func = (curl_slist_free_all_type)load_function(libcurl, "curl_slist_free_all"); + curl_easy_strerror_func = (curl_easy_strerror_type)load_function(libcurl, "curl_easy_strerror"); + curl_multi_init_func = (curl_multi_init_type)load_function(libcurl, "curl_multi_init"); + curl_multi_add_handle_func = (curl_multi_add_handle_type)load_function(libcurl, "curl_multi_add_handle"); + curl_multi_remove_handle_func = (curl_multi_remove_handle_type)load_function(libcurl, "curl_multi_remove_handle"); + curl_multi_fdset_func = (curl_multi_fdset_type)load_function(libcurl, "curl_multi_fdset"); + curl_multi_perform_func = (curl_multi_perform_type)load_function(libcurl, "curl_multi_perform"); + curl_multi_cleanup_func = (curl_multi_cleanup_type)load_function(libcurl, "curl_multi_cleanup"); + curl_multi_info_read_func = (curl_multi_info_read_type)load_function(libcurl, "curl_multi_info_read"); + curl_multi_strerror_func = (curl_multi_strerror_type)load_function(libcurl, "curl_multi_strerror"); + curl_multi_timeout_func = (curl_multi_timeout_type)load_function(libcurl, "curl_multi_timeout"); + curl_easy_init_func = (curl_easy_init_type)load_function(libcurl, "curl_easy_init"); + curl_easy_perform_func = (curl_easy_perform_type)load_function(libcurl, "curl_easy_perform"); + curl_easy_cleanup_func = (curl_easy_cleanup_type)load_function(libcurl, "curl_easy_cleanup"); + curl_easy_duphandle_func = (curl_easy_duphandle_type)load_function(libcurl, "curl_easy_duphandle"); + + curl_easy_getinfo_func = load_function(libcurl, "curl_easy_getinfo"); + curl_easy_getinfo_long_func = (curl_easy_getinfo_long_type)curl_easy_getinfo_func; + curl_easy_getinfo_pointer_func = (curl_easy_getinfo_pointer_type)curl_easy_getinfo_func; + curl_easy_getinfo_off_t_func = (curl_easy_getinfo_off_t_type)curl_easy_getinfo_func; + + curl_easy_setopt_func = load_function(libcurl, "curl_easy_setopt"); + curl_easy_setopt_long_func = (curl_easy_setopt_long_type)curl_easy_setopt_func; + curl_easy_setopt_pointer_func = (curl_easy_setopt_pointer_type)curl_easy_setopt_func; + curl_easy_setopt_off_t_func = (curl_easy_setopt_off_t_type)curl_easy_setopt_func; +} + +struct curl_version_info_data *curl_version_info(CURLversion version) +{ + lazy_load_curl(); + return curl_version_info_func(version); +} + +char *curl_easy_escape(CURL *handle, const char *string, int length) +{ + lazy_load_curl(); + return curl_easy_escape_func(handle, string, length); +} + +void curl_free(void *p) +{ + lazy_load_curl(); + curl_free_func(p); +} + +CURLcode curl_global_init(long flags) +{ + lazy_load_curl(); + return curl_global_init_func(flags); +} + +CURLsslset curl_global_sslset(curl_sslbackend id, const char *name, const curl_ssl_backend ***avail) +{ + lazy_load_curl(); + return curl_global_sslset_func(id, name, avail); +} + +void curl_global_cleanup(void) +{ + lazy_load_curl(); + curl_global_cleanup_func(); +} + +struct curl_slist *curl_slist_append(struct curl_slist *list, const char *data) +{ + lazy_load_curl(); + return curl_slist_append_func(list, data); +} + +void curl_slist_free_all(struct curl_slist *list) +{ + lazy_load_curl(); + curl_slist_free_all_func(list); +} + +const char *curl_easy_strerror(CURLcode error) +{ + lazy_load_curl(); + return curl_easy_strerror_func(error); +} + +CURLM *curl_multi_init(void) +{ + lazy_load_curl(); + return curl_multi_init_func(); +} + +CURLMcode curl_multi_add_handle(CURLM *multi_handle, CURL *curl_handle) +{ + lazy_load_curl(); + return curl_multi_add_handle_func(multi_handle, curl_handle); +} + +CURLMcode curl_multi_remove_handle(CURLM *multi_handle, CURL *curl_handle) +{ + lazy_load_curl(); + return curl_multi_remove_handle_func(multi_handle, curl_handle); +} + +CURLMcode curl_multi_fdset(CURLM *multi_handle, fd_set *read_fd_set, fd_set *write_fd_set, fd_set *exc_fd_set, int *max_fd) +{ + lazy_load_curl(); + return curl_multi_fdset_func(multi_handle, read_fd_set, write_fd_set, exc_fd_set, max_fd); +} + +CURLMcode curl_multi_perform(CURLM *multi_handle, int *running_handles) +{ + lazy_load_curl(); + return curl_multi_perform_func(multi_handle, running_handles); +} + +CURLMcode curl_multi_cleanup(CURLM *multi_handle) +{ + lazy_load_curl(); + return curl_multi_cleanup_func(multi_handle); +} + +CURLMsg *curl_multi_info_read(CURLM *multi_handle, int *msgs_in_queue) +{ + lazy_load_curl(); + return curl_multi_info_read_func(multi_handle, msgs_in_queue); +} + +const char *curl_multi_strerror(CURLMcode error) +{ + lazy_load_curl(); + return curl_multi_strerror_func(error); +} + +CURLMcode curl_multi_timeout(CURLM *multi_handle, long *milliseconds) +{ + lazy_load_curl(); + return curl_multi_timeout_func(multi_handle, milliseconds); +} + +CURL *curl_easy_init(void) +{ + lazy_load_curl(); + return curl_easy_init_func(); +} + +CURLcode curl_easy_perform(CURL *curl) +{ + lazy_load_curl(); + return curl_easy_perform_func(curl); +} + +void curl_easy_cleanup(CURL *curl) +{ + lazy_load_curl(); + curl_easy_cleanup_func(curl); +} + +CURL *curl_easy_duphandle(CURL *curl) +{ + lazy_load_curl(); + return curl_easy_duphandle_func(curl); +} + +#ifndef CURL_IGNORE_DEPRECATION +#define CURL_IGNORE_DEPRECATION(x) x +#endif + +#ifndef CURLOPTTYPE_BLOB +#define CURLOPTTYPE_BLOB 40000 +#endif + +#undef curl_easy_getinfo +CURLcode curl_easy_getinfo(CURL *curl, CURLINFO info, ...) +{ + va_list ap; + CURLcode res; + + va_start(ap, info); + lazy_load_curl(); + CURL_IGNORE_DEPRECATION( + if (info >= CURLINFO_LONG && info < CURLINFO_DOUBLE) + res = curl_easy_getinfo_long_func(curl, info, va_arg(ap, long *)); + else if ((info >= CURLINFO_STRING && info < CURLINFO_LONG) || + (info >= CURLINFO_SLIST && info < CURLINFO_SOCKET)) + res = curl_easy_getinfo_pointer_func(curl, info, va_arg(ap, void **)); + else if (info >= CURLINFO_OFF_T) + res = curl_easy_getinfo_off_t_func(curl, info, va_arg(ap, curl_off_t *)); + else + die("%s:%d: TODO (info: %d)!", __FILE__, __LINE__, info); + ) + va_end(ap); + return res; +} + +#undef curl_easy_setopt +CURLcode curl_easy_setopt(CURL *curl, CURLoption opt, ...) +{ + va_list ap; + CURLcode res; + + va_start(ap, opt); + lazy_load_curl(); + CURL_IGNORE_DEPRECATION( + if (opt >= CURLOPTTYPE_LONG && opt < CURLOPTTYPE_OBJECTPOINT) + res = curl_easy_setopt_long_func(curl, opt, va_arg(ap, long)); + else if (opt >= CURLOPTTYPE_OBJECTPOINT && opt < CURLOPTTYPE_OFF_T) + res = curl_easy_setopt_pointer_func(curl, opt, va_arg(ap, void *)); + else if (opt >= CURLOPTTYPE_OFF_T && opt < CURLOPTTYPE_BLOB) + res = curl_easy_setopt_off_t_func(curl, opt, va_arg(ap, curl_off_t)); + else + die("%s:%d: TODO (opt: %d)!", __FILE__, __LINE__, opt); + ) + va_end(ap); + return res; +} From dbc9b922faaffaef2e66fb0fca13f5679adc1445 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Sun, 7 May 2023 22:51:52 +0200 Subject: [PATCH 163/312] http: support lazy-loading libcurl also on Windows This implements the Windows-specific support code, because everything is slightly different on Windows, even loading shared libraries. Note: I specifically do _not_ use the code from `compat/win32/lazyload.h` here because that code is optimized for loading individual functions from various system DLLs, while we specifically want to load _many_ functions from _one_ DLL here, and distinctly not a system DLL (we expect libcurl to be located outside `C:\Windows\system32`, something `INIT_PROC_ADDR` refuses to work with). Also, the `curl_easy_getinfo()`/`curl_easy_setopt()` functions are declared as vararg functions, which `lazyload.h` cannot handle. Finally, we are about to optionally override the exact file name that is to be loaded, which is a goal contrary to `lazyload.h`'s design. Signed-off-by: Johannes Schindelin --- Makefile | 4 ++++ compat/lazyload-curl.c | 52 ++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 56 insertions(+) diff --git a/Makefile b/Makefile index 59689d92fd0261..234a709593c254 100644 --- a/Makefile +++ b/Makefile @@ -1672,7 +1672,11 @@ else # The `CURL_STATICLIB` constant must be defined to avoid seeing the functions # declared as DLL imports CURL_CFLAGS = -DCURL_STATICLIB +ifneq ($(uname_S),MINGW) +ifneq ($(uname_S),Windows) CURL_LIBCURL = -ldl +endif +endif else ifndef CURL_LDFLAGS CURL_LDFLAGS = $(eval CURL_LDFLAGS := $$(shell $$(CURL_CONFIG) --libs))$(CURL_LDFLAGS) diff --git a/compat/lazyload-curl.c b/compat/lazyload-curl.c index 19aa2b6d4b6942..98d73fb0f2a66f 100644 --- a/compat/lazyload-curl.c +++ b/compat/lazyload-curl.c @@ -1,6 +1,8 @@ #include "../git-compat-util.h" #include "../git-curl-compat.h" +#ifndef WIN32 #include +#endif /* * The ABI version of libcurl is encoded in its shared libraries' file names. @@ -11,6 +13,7 @@ typedef void (*func_t)(void); +#ifndef WIN32 #ifdef __APPLE__ #define LIBCURL_FILE_NAME(base) base "." LIBCURL_ABI_VERSION ".dylib" #else @@ -35,6 +38,55 @@ static func_t load_function(void *handle, const char *name) *(void **)&f = dlsym(handle, name); return f; } +#else +#define LIBCURL_FILE_NAME(base) base "-" LIBCURL_ABI_VERSION ".dll" + +static void *load_library(const char *name) +{ + size_t name_size = strlen(name) + 1; + const char *path = getenv("PATH"); + char dll_path[MAX_PATH]; + + while (path && *path) { + const char *sep = strchrnul(path, ';'); + size_t len = sep - path; + + if (len && len + name_size < sizeof(dll_path)) { + memcpy(dll_path, path, len); + dll_path[len] = '/'; + memcpy(dll_path + len + 1, name, name_size); + + if (!access(dll_path, R_OK)) { + wchar_t wpath[MAX_PATH]; + int wlen = MultiByteToWideChar(CP_UTF8, 0, dll_path, -1, wpath, ARRAY_SIZE(wpath)); + void *res = wlen ? (void *)LoadLibraryExW(wpath, NULL, 0) : NULL; + if (!res) { + DWORD err = GetLastError(); + char buf[1024]; + + if (!FormatMessageA(FORMAT_MESSAGE_FROM_SYSTEM | + FORMAT_MESSAGE_ARGUMENT_ARRAY | + FORMAT_MESSAGE_IGNORE_INSERTS, + NULL, err, LANG_NEUTRAL, + buf, sizeof(buf) - 1, NULL)) + xsnprintf(buf, sizeof(buf), "last error: %ld", err); + error("LoadLibraryExW() failed with: %s", buf); + } + return res; + } + } + + path = *sep ? sep + 1 : NULL; + } + + return NULL; +} + +static func_t load_function(void *handle, const char *name) +{ + return (func_t)GetProcAddress((HANDLE)handle, name); +} +#endif typedef struct curl_version_info_data *(*curl_version_info_type)(CURLversion version); static curl_version_info_type curl_version_info_func; From c937d816bcd92300dae755c6fcecbe6418a0901c Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Sun, 7 May 2023 22:05:33 +0200 Subject: [PATCH 164/312] http: when loading libcurl lazily, allow for multiple SSL backends The previous commits introduced a compile-time option to load libcurl lazily, but it uses the hard-coded name "libcurl-4.dll" (or equivalent on platforms other than Windows). To allow for installing multiple libcurl flavors side by side, where each supports one specific SSL/TLS backend, let's first look whether `libcurl--4.dll` exists, and only use `libcurl-4.dll` as a fall back. That will allow us to ship with a libcurl by default that only supports the Secure Channel backend for the `https://` protocol. This libcurl won't suffer from any dependency problem when upgrading OpenSSL to a new major version (which will change the DLL name, and hence break every program and library that depends on it). This is crucial because Git for Windows relies on libcurl to keep working when building and deploying a new OpenSSL package because that library is used by `git fetch` and `git clone`. Note that this feature is by no means specific to Windows. On Ubuntu, for example, a `git` built using `LAZY_LOAD_LIBCURL` will use `libcurl.so.4` for `http.sslbackend=openssl` and `libcurl-gnutls.so.4` for `http.sslbackend=gnutls`. Signed-off-by: Johannes Schindelin --- compat/lazyload-curl.c | 16 ++++++++++++++-- 1 file changed, 14 insertions(+), 2 deletions(-) diff --git a/compat/lazyload-curl.c b/compat/lazyload-curl.c index 98d73fb0f2a66f..2f2b2ebc27993e 100644 --- a/compat/lazyload-curl.c +++ b/compat/lazyload-curl.c @@ -172,17 +172,26 @@ static curl_easy_setopt_pointer_type curl_easy_setopt_pointer_func; typedef CURLcode (*curl_easy_setopt_off_t_type)(CURL *curl, CURLoption opt, curl_off_t value); static curl_easy_setopt_off_t_type curl_easy_setopt_off_t_func; +static char ssl_backend[64]; + static void lazy_load_curl(void) { static int initialized; - void *libcurl; + void *libcurl = NULL; func_t curl_easy_getinfo_func, curl_easy_setopt_func; if (initialized) return; initialized = 1; - libcurl = load_library(LIBCURL_FILE_NAME("libcurl")); + if (ssl_backend[0]) { + char dll_name[64 + 16]; + snprintf(dll_name, sizeof(dll_name) - 1, + LIBCURL_FILE_NAME("libcurl-%s"), ssl_backend); + libcurl = load_library(dll_name); + } + if (!libcurl) + libcurl = load_library(LIBCURL_FILE_NAME("libcurl")); if (!libcurl) die("failed to load library '%s'", LIBCURL_FILE_NAME("libcurl")); @@ -246,6 +255,9 @@ CURLcode curl_global_init(long flags) CURLsslset curl_global_sslset(curl_sslbackend id, const char *name, const curl_ssl_backend ***avail) { + if (name && strlen(name) < sizeof(ssl_backend)) + strlcpy(ssl_backend, name, sizeof(ssl_backend)); + lazy_load_curl(); return curl_global_sslset_func(id, name, avail); } From cbd87816c8516dbd13d1ba275f60668b29c0de79 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Tue, 21 Mar 2023 16:14:44 +0100 Subject: [PATCH 165/312] windows: fix Repository>Explore Working Copy Since Git v2.39.1, we are a bit more stringent in searching the PATH. In particular, we specifically require the `.exe` suffix. However, the `Repository>Explore Working Copy` command asks for `explorer.exe` to be found on the `PATH`, which _already_ has that suffix. Let's unstartle the PATH-finding logic about this scenario. This fixes https://github.com/git-for-windows/git/issues/4356 Signed-off-by: Johannes Schindelin --- git-gui/git-gui.sh | 3 +++ 1 file changed, 3 insertions(+) diff --git a/git-gui/git-gui.sh b/git-gui/git-gui.sh index 887d6d596c16fc..c913a08566542e 100755 --- a/git-gui/git-gui.sh +++ b/git-gui/git-gui.sh @@ -101,6 +101,9 @@ proc _which {what args} { if {[is_Windows] && [lsearch -exact $args -script] >= 0} { set suffix {} + } elseif {[string match *$_search_exe $what]} { + # The search string already has the file extension + set suffix {} } else { set suffix $_search_exe } From 1300c44cbb994b9d96494a831ec32cce9c5c345b Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Sun, 7 May 2023 22:43:37 +0200 Subject: [PATCH 166/312] mingw: do load libcurl dynamically by default This will help with Git for Windows' maintenance going forward: It allows Git for Windows to switch its primary libcurl to a variant without the OpenSSL backend, while still loading an alternate when setting `http.sslBackend = openssl`. This is necessary to avoid maintenance headaches with upgrading OpenSSL: its major version name is encoded in the shared library's file name and hence major version updates (temporarily) break libraries that are linked against the OpenSSL library. Signed-off-by: Johannes Schindelin --- config.mak.uname | 1 + 1 file changed, 1 insertion(+) diff --git a/config.mak.uname b/config.mak.uname index 1736c4f0116e75..76b60c2fba8d57 100644 --- a/config.mak.uname +++ b/config.mak.uname @@ -692,6 +692,7 @@ ifeq ($(uname_S),MINGW) HAVE_PLATFORM_PROCINFO = YesPlease CSPRNG_METHOD = rtlgenrandom BASIC_LDFLAGS += -municode -Wl,--tsaware + LAZYLOAD_LIBCURL = YesDoThatPlease COMPAT_CFLAGS += -DNOGDI -Icompat -Icompat/win32 COMPAT_CFLAGS += -DSTRIP_EXTENSION=\".exe\" COMPAT_OBJS += compat/mingw.o compat/winansi.o \ From a6e6c53f396231daa1b1d6121ea34ff5db0617b0 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Wed, 2 Nov 2022 16:23:58 +0100 Subject: [PATCH 167/312] Add a GitHub workflow to verify that Git/Scalar work in Nano Server In Git for Windows v2.39.0, we fixed a regression where `git.exe` would no longer work in Windows Nano Server (frequently used in Docker containers). This GitHub workflow can be used to verify manually that the Git/Scalar executables work in Nano Server. Signed-off-by: Johannes Schindelin --- .github/workflows/nano-server.yml | 76 +++++++++++++++++++++++++++++++ 1 file changed, 76 insertions(+) create mode 100644 .github/workflows/nano-server.yml diff --git a/.github/workflows/nano-server.yml b/.github/workflows/nano-server.yml new file mode 100644 index 00000000000000..3d943c23d2616d --- /dev/null +++ b/.github/workflows/nano-server.yml @@ -0,0 +1,76 @@ +name: Windows Nano Server tests + +on: + workflow_dispatch: + +env: + DEVELOPER: 1 + +jobs: + test-nano-server: + runs-on: windows-2022 + env: + WINDBG_DIR: "C:/Program Files (x86)/Windows Kits/10/Debuggers/x64" + IMAGE: mcr.microsoft.com/powershell:nanoserver-ltsc2022 + + steps: + - uses: actions/checkout@v4 + - uses: git-for-windows/setup-git-for-windows-sdk@v1 + - name: build Git + shell: bash + run: make -j15 + - name: pull nanoserver image + shell: bash + run: docker pull $IMAGE + - name: run nano-server test + shell: bash + run: | + docker run \ + --user "ContainerAdministrator" \ + -v "$WINDBG_DIR:C:/dbg" \ + -v "$(cygpath -aw /mingw64/bin):C:/mingw64-bin" \ + -v "$(cygpath -aw .):C:/test" \ + $IMAGE pwsh.exe -Command ' + # Extend the PATH to include the `.dll` files in /mingw64/bin/ + $env:PATH += ";C:\mingw64-bin" + + # For each executable to test pick some no-operation set of + # flags/subcommands or something that should quickly result in an + # error with known exit code that is not a negative 32-bit + # number, and set the expected return code appropriately. + # + # Only test executables that could be expected to run in a UI + # less environment. + # + # ( Executable path, arguments, expected return code ) + # also note space is required before close parenthesis (a + # powershell quirk when defining nested arrays like this) + + $executables_to_test = @( + ("C:\test\git.exe", "", 1 ), + ("C:\test\scalar.exe", "version", 0 ) + ) + + foreach ($executable in $executables_to_test) + { + Write-Output "Now testing $($executable[0])" + &$executable[0] $executable[1] + if ($LASTEXITCODE -ne $executable[2]) { + # if we failed, run the debugger to find out what function + # or DLL could not be found and then exit the script with + # failure The missing DLL or EXE will be referenced near + # the end of the output + + # Set a flag to have the debugger show loader stub + # diagnostics. This requires running as administrator, + # otherwise the flag will be ignored. + C:\dbg\gflags -i $executable[0] +SLS + + C:\dbg\cdb.exe -c "g" -c "q" $executable[0] $executable[1] + + exit 1 + } + } + + exit 0 + ' From b7300e3dbe554a337bd3f16ad2619f70fa959753 Mon Sep 17 00:00:00 2001 From: David Lomas Date: Fri, 28 Jul 2023 15:31:25 +0100 Subject: [PATCH 168/312] mingw: suggest `windows.appendAtomically` in more cases When running Git for Windows on a remote APFS filesystem, it would appear that the `mingw_open_append()`/`write()` combination would fail almost exactly like on some CIFS-mounted shares as had been reported in https://github.com/git-for-windows/git/issues/2753, albeit with a different `errno` value. Let's handle that `errno` value just the same, by suggesting to set `windows.appendAtomically=false`. Signed-off-by: David Lomas Signed-off-by: Johannes Schindelin --- compat/mingw.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/compat/mingw.c b/compat/mingw.c index 0658cdfc3c733f..adf2e423da539a 100644 --- a/compat/mingw.c +++ b/compat/mingw.c @@ -787,7 +787,7 @@ ssize_t mingw_write(int fd, const void *buf, size_t len) { ssize_t result = write(fd, buf, len); - if (result < 0 && (errno == EINVAL || errno == ENOSPC) && buf) { + if (result < 0 && (errno == EINVAL || errno == EBADF || errno == ENOSPC) && buf) { int orig = errno; /* check if fd is a pipe */ @@ -813,7 +813,7 @@ ssize_t mingw_write(int fd, const void *buf, size_t len) } errno = orig; - } else if (orig == EINVAL) + } else if (orig == EINVAL || errno == EBADF) errno = EPIPE; else { DWORD buf_size; From fbfeb1a987103ce5f4882b7560d491fb39482999 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Wed, 22 Nov 2023 22:57:38 +0100 Subject: [PATCH 169/312] win32: use native ANSI sequence processing, if possible Windows 10 version 1511 (also known as Anniversary Update), according to https://learn.microsoft.com/en-us/windows/console/console-virtual-terminal-sequences introduced native support for ANSI sequence processing. This allows using colors from the entire 24-bit color range. All we need to do is test whether the console's "virtual processing support" can be enabled. If it can, we do not even need to start the `console_thread` to handle ANSI sequences. Or, almost all we need to do: When `console_thread()` does its work, it uses the Unicode-aware `write_console()` function to write to the Win32 Console, which supports Git for Windows' implicit convention that all text that is written is encoded in UTF-8. The same is not necessarily true if native ANSI sequence processing is used, as the output is then subject to the current code page. Let's ensure that the code page is set to `CP_UTF8` as long as Git writes to it. Signed-off-by: Johannes Schindelin --- compat/winansi.c | 46 ++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 46 insertions(+) diff --git a/compat/winansi.c b/compat/winansi.c index ac2ffb78691a7d..a83a7f47ada9b2 100644 --- a/compat/winansi.c +++ b/compat/winansi.c @@ -593,6 +593,49 @@ static void detect_msys_tty(int fd) #endif +static HANDLE std_console_handle; +static DWORD std_console_mode = ENABLE_VIRTUAL_TERMINAL_PROCESSING; +static UINT std_console_code_page = CP_UTF8; + +static void reset_std_console(void) +{ + if (std_console_mode != ENABLE_VIRTUAL_TERMINAL_PROCESSING) + SetConsoleMode(std_console_handle, std_console_mode); + if (std_console_code_page != CP_UTF8) + SetConsoleOutputCP(std_console_code_page); +} + +static int enable_virtual_processing(void) +{ + std_console_handle = GetStdHandle(STD_OUTPUT_HANDLE); + if (std_console_handle == INVALID_HANDLE_VALUE || + !GetConsoleMode(std_console_handle, &std_console_mode)) { + std_console_handle = GetStdHandle(STD_ERROR_HANDLE); + if (std_console_handle == INVALID_HANDLE_VALUE || + !GetConsoleMode(std_console_handle, &std_console_mode)) + return 0; + } + + std_console_code_page = GetConsoleOutputCP(); + if (std_console_code_page != CP_UTF8) + SetConsoleOutputCP(CP_UTF8); + if (!std_console_code_page) + std_console_code_page = CP_UTF8; + + atexit(reset_std_console); + + if (std_console_mode & ENABLE_VIRTUAL_TERMINAL_PROCESSING) + return 1; + + if (!SetConsoleMode(std_console_handle, + std_console_mode | + ENABLE_PROCESSED_OUTPUT | + ENABLE_VIRTUAL_TERMINAL_PROCESSING)) + return 0; + + return 1; +} + /* * Wrapper for isatty(). Most calls in the main git code * call isatty(1 or 2) to see if the instance is interactive @@ -631,6 +674,9 @@ void winansi_init(void) return; } + if (enable_virtual_processing()) + return; + /* create a named pipe to communicate with the console thread */ if (swprintf(name, ARRAY_SIZE(name) - 1, L"\\\\.\\pipe\\winansi%lu", GetCurrentProcessId()) < 0) From 412c6b0e1e45b2095ca087e204f90914a759b480 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Matthias=20A=C3=9Fhauer?= Date: Sat, 2 Dec 2023 12:10:00 +0100 Subject: [PATCH 170/312] git.rc: include winuser.h MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit winuser.h contains the definition of RT_MANIFEST that our LLVM based toolchain needs to understand that we want to embed compat/win32/git.manifest as an application manifest. It currently just embeds it as additional data that Windows doesn't understand. This also helps our GCC based toolchain understand that we only want one copy embedded. It currently embeds one working assembly manifest and one nearly identical, but useless copy as additional data. This also teaches our Visual Studio based buildsystems to pick up the manifest file from git.rc. This means we don't have to explicitly specify it in contrib/buildsystems/Generators/Vcxproj.pm anymore. Slightly counter-intuitively this also means we have to explicitly tell Cmake not to embed a default manifest. This fixes https://github.com/git-for-windows/git/issues/4707 Signed-off-by: Matthias Aßhauer Signed-off-by: Johannes Schindelin --- contrib/buildsystems/CMakeLists.txt | 1 + git.rc.in | 1 + 2 files changed, 2 insertions(+) diff --git a/contrib/buildsystems/CMakeLists.txt b/contrib/buildsystems/CMakeLists.txt index 25b495fa737d07..5ac5850100033c 100644 --- a/contrib/buildsystems/CMakeLists.txt +++ b/contrib/buildsystems/CMakeLists.txt @@ -208,6 +208,7 @@ if(CMAKE_C_COMPILER_ID STREQUAL "MSVC") set(CMAKE_RUNTIME_OUTPUT_DIRECTORY_DEBUG ${CMAKE_BINARY_DIR}) set(CMAKE_RUNTIME_OUTPUT_DIRECTORY_RELEASE ${CMAKE_BINARY_DIR}) add_compile_options(/MP /std:c11) + add_link_options(/MANIFEST:NO) endif() #default behaviour diff --git a/git.rc.in b/git.rc.in index e69444eef3f0c5..1d5b627b610549 100644 --- a/git.rc.in +++ b/git.rc.in @@ -1,3 +1,4 @@ +#include 1 VERSIONINFO FILEVERSION @GIT_MAJOR_VERSION@,@GIT_MINOR_VERSION@,@GIT_MICRO_VERSION@,@GIT_PATCH_LEVEL@ PRODUCTVERSION @GIT_MAJOR_VERSION@,@GIT_MINOR_VERSION@,@GIT_MICRO_VERSION@,@GIT_PATCH_LEVEL@ From 38429d98e90ef97a0dedc48ff65d5e509f2ce4fa Mon Sep 17 00:00:00 2001 From: MinarKotonoha Date: Mon, 8 Apr 2024 16:41:10 +0800 Subject: [PATCH 171/312] common-main.c: fflush stdout buffer upon exit By default, the buffer type of Windows' `stdout` is unbuffered (_IONBF), and there is no need to manually fflush `stdout`. But some programs, such as the Windows Filtering Platform driver provided by the security software, may change the buffer type of `stdout` to full buffering. This nees `fflush(stdout)` to be called manually, otherwise there will be no output to `stdout`. Signed-off-by: MinarKotonoha Signed-off-by: Johannes Schindelin --- common-exit.c | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/common-exit.c b/common-exit.c index 1aaa538be3ed67..609f32abed8b53 100644 --- a/common-exit.c +++ b/common-exit.c @@ -11,6 +11,13 @@ static void check_bug_if_BUG(void) /* We wrap exit() to call common_exit() in git-compat-util.h */ int common_exit(const char *file, int line, int code) { + /* + * Windows Filtering Platform driver provided by the security software + * may change buffer type of stdout from _IONBF to _IOFBF. + * It will no output without fflush manually. + */ + fflush(stdout); + /* * For non-POSIX systems: Take the lowest 8 bits of the "code" * to e.g. turn -1 into 255. On a POSIX system this is From 07bb89238ef543cc10570cab816b8d1d427d6026 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Tue, 9 Apr 2024 16:50:56 +0200 Subject: [PATCH 172/312] t5601/t7406(mingw): do run tests with symlink support A long time ago, we decided to run tests in Git for Windows' SDK with the default `winsymlinks` mode: copying instead of linking. This is still the default mode of MSYS2 to this day. However, this is not how most users run Git for Windows: As the majority of Git for Windows' users seem to be on Windows 10 and newer, likely having enabled Developer Mode (which allows creating symbolic links without administrator privileges), they will run with symlink support enabled. This is the reason why it is crucial to get the fixes for CVE-2024-? to the users, and also why it is crucial to ensure that the test suite exercises the related test cases. This commit ensures the latter. Signed-off-by: Johannes Schindelin --- t/t5601-clone.sh | 10 ++++++++++ t/t7406-submodule-update.sh | 9 +++++++++ 2 files changed, 19 insertions(+) diff --git a/t/t5601-clone.sh b/t/t5601-clone.sh index d0c18660e33113..a529916e71a4de 100755 --- a/t/t5601-clone.sh +++ b/t/t5601-clone.sh @@ -7,6 +7,16 @@ export GIT_TEST_DEFAULT_INITIAL_BRANCH_NAME . ./test-lib.sh +# This test script contains test cases that need to create symbolic links. To +# make sure that these test cases are exercised in Git for Windows, where (for +# historical reasons) `ln -s` creates copies by default, let's specifically ask +# for `ln -s` to create symbolic links whenever possible. +if test_have_prereq MINGW +then + MSYS=${MSYS+$MSYS }winsymlinks:nativestrict + export MSYS +fi + X= test_have_prereq !MINGW || X=.exe diff --git a/t/t7406-submodule-update.sh b/t/t7406-submodule-update.sh index c562bad042ab2d..38bfaeb510a2fb 100755 --- a/t/t7406-submodule-update.sh +++ b/t/t7406-submodule-update.sh @@ -14,6 +14,15 @@ export GIT_TEST_DEFAULT_INITIAL_BRANCH_NAME . ./test-lib.sh +# This test script contains test cases that need to create symbolic links. To +# make sure that these test cases are exercised in Git for Windows, where (for +# historical reasons) `ln -s` creates copies by default, let's specifically ask +# for `ln -s` to create symbolic links whenever possible. +if test_have_prereq MINGW +then + MSYS=${MSYS+$MSYS }winsymlinks:nativestrict + export MSYS +fi compare_head() { From ac4bc068bae2d09538ceeafd06b9e7b97c1628e3 Mon Sep 17 00:00:00 2001 From: Johannes Schindelin Date: Tue, 21 May 2024 13:55:26 +0200 Subject: [PATCH 173/312] win32: ensure that `localtime_r()` is declared even in i686 builds The `__MINGW64__` constant is defined, surprise, surprise, only when building for a 64-bit CPU architecture. Therefore using it as a guard to define `_POSIX_C_SOURCE` (so that `localtime_r()` is declared, among other functions) is not enough, we also need to check `__MINGW32__`. Technically, the latter constant is defined even for 64-bit builds. But let's make things a bit easier to understand by testing for both constants. Making it so fixes this compile warning (turned error in GCC v14.1): archive-zip.c: In function 'dos_time': archive-zip.c:612:9: error: implicit declaration of function 'localtime_r'; did you mean 'localtime_s'? [-Wimplicit-function-declaration] 612 | localtime_r(&time, &tm); | ^~~~~~~~~~~ | localtime_s Signed-off-by: Johannes Schindelin --- git-compat-util.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/git-compat-util.h b/git-compat-util.h index e123288e8f1393..e63517630582ac 100644 --- a/git-compat-util.h +++ b/git-compat-util.h @@ -176,7 +176,7 @@ DISABLE_WARNING(-Wsign-compare) /* Approximation of the length of the decimal representation of this type. */ #define decimal_length(x) ((int)(sizeof(x) * 2.56 + 0.5) + 1) -#ifdef __MINGW64__ +#if defined(__MINGW32__) || defined(__MINGW64__) #define _POSIX_C_SOURCE 1 #elif defined(__sun__) /* From ceb530a5ebb5407fb1b877281a8f94946cac8609 Mon Sep 17 00:00:00 2001 From: Derrick Stolee Date: Fri, 6 Sep 2024 14:16:13 -0400 Subject: [PATCH 174/312] revision: create mark_trees_uninteresting_dense() The sparse tree walk algorithm was created in d5d2e93577e (revision: implement sparse algorithm, 2019-01-16) and involves using the mark_trees_uninteresting_sparse() method. This method takes a repository and an oidset of tree IDs, some of which have the UNINTERESTING flag and some of which do not. Create a method that has an equivalent set of preconditions but uses a "dense" walk (recursively visits all reachable trees, as long as they have not previously been marked UNINTERESTING). This is an important difference from mark_tree_uninteresting(), which short-circuits if the given tree has the UNINTERESTING flag. A use of this method will be added in a later change, with a condition set whether the sparse or dense approach should be used. Signed-off-by: Derrick Stolee --- revision.c | 15 +++++++++++++++ revision.h | 1 + 2 files changed, 16 insertions(+) diff --git a/revision.c b/revision.c index c4390f0938cbde..6c88841dcca7a4 100644 --- a/revision.c +++ b/revision.c @@ -220,6 +220,21 @@ static void add_children_by_path(struct repository *r, free_tree_buffer(tree); } +void mark_trees_uninteresting_dense(struct repository *r, + struct oidset *trees) +{ + struct object_id *oid; + struct oidset_iter iter; + + oidset_iter_init(trees, &iter); + while ((oid = oidset_iter_next(&iter))) { + struct tree *tree = lookup_tree(r, oid); + + if (tree->object.flags & UNINTERESTING) + mark_tree_contents_uninteresting(r, tree); + } +} + void mark_trees_uninteresting_sparse(struct repository *r, struct oidset *trees) { diff --git a/revision.h b/revision.h index 71e984c452b8d7..8938b2db112e2a 100644 --- a/revision.h +++ b/revision.h @@ -487,6 +487,7 @@ void put_revision_mark(const struct rev_info *revs, void mark_parents_uninteresting(struct rev_info *revs, struct commit *commit); void mark_tree_uninteresting(struct repository *r, struct tree *tree); +void mark_trees_uninteresting_dense(struct repository *r, struct oidset *trees); void mark_trees_uninteresting_sparse(struct repository *r, struct oidset *trees); void show_object_with_name(FILE *, struct object *, const char *); From 9e0b936a30e1eb36f581a38c8a5b97aec50c7e0a Mon Sep 17 00:00:00 2001 From: Derrick Stolee Date: Tue, 24 Sep 2024 13:45:19 -0400 Subject: [PATCH 175/312] pack-objects: extract should_attempt_deltas() This will be helpful in a future change. Signed-off-by: Derrick Stolee --- builtin/pack-objects.c | 53 +++++++++++++++++++++++------------------- 1 file changed, 29 insertions(+), 24 deletions(-) diff --git a/builtin/pack-objects.c b/builtin/pack-objects.c index 58a9b1612626e0..1d0992a8dac089 100644 --- a/builtin/pack-objects.c +++ b/builtin/pack-objects.c @@ -3196,6 +3196,33 @@ static int add_ref_tag(const char *tag UNUSED, const char *referent UNUSED, cons return 0; } +static int should_attempt_deltas(struct object_entry *entry) +{ + if (DELTA(entry)) + return 0; + + if (!entry->type_valid || + oe_size_less_than(&to_pack, entry, 50)) + return 0; + + if (entry->no_try_delta) + return 0; + + if (!entry->preferred_base) { + if (oe_type(entry) < 0) + die(_("unable to get type of object %s"), + oid_to_hex(&entry->idx.oid)); + } else if (oe_type(entry) < 0) { + /* + * This object is not found, but we + * don't have to include it anyway. + */ + return 0; + } + + return 1; +} + static void prepare_pack(int window, int depth) { struct object_entry **delta_list; @@ -3226,33 +3253,11 @@ static void prepare_pack(int window, int depth) for (i = 0; i < to_pack.nr_objects; i++) { struct object_entry *entry = to_pack.objects + i; - if (DELTA(entry)) - /* This happens if we decided to reuse existing - * delta from a pack. "reuse_delta &&" is implied. - */ - continue; - - if (!entry->type_valid || - oe_size_less_than(&to_pack, entry, 50)) + if (!should_attempt_deltas(entry)) continue; - if (entry->no_try_delta) - continue; - - if (!entry->preferred_base) { + if (!entry->preferred_base) nr_deltas++; - if (oe_type(entry) < 0) - die(_("unable to get type of object %s"), - oid_to_hex(&entry->idx.oid)); - } else { - if (oe_type(entry) < 0) { - /* - * This object is not found, but we - * don't have to include it anyway. - */ - continue; - } - } delta_list[n++] = entry; } From ff75aa0c52c5387b412481b7548ee6a144f9ada7 Mon Sep 17 00:00:00 2001 From: Derrick Stolee Date: Thu, 5 Sep 2024 10:04:51 -0400 Subject: [PATCH 176/312] pack-objects: add --path-walk option In order to more easily compute delta bases among objects that appear at the exact same path, add a --path-walk option to 'git pack-objects'. This option will use the path-walk API instead of the object walk given by the revision machinery. Since objects will be provided in batches representing a common path, those objects can be tested for delta bases immediately instead of waiting for a sort of the full object list by name-hash. This has multiple benefits, including avoiding collisions by name-hash. The objects marked as UNINTERESTING are included in these batches, so we are guaranteeing some locality to find good delta bases. After the individual passes are done on a per-path basis, the default name-hash is used to find other opportunistic delta bases that did not match exactly by the full path name. RFC TODO: It is important to note that this option is inherently incompatible with using a bitmap index. This walk probably also does not work with other advanced features, such as delta islands. Getting ahead of myself, this option compares well with --full-name-hash when the packfile is large enough, but also performs at least as well as the default in all cases that I've seen. RFC TODO: this should probably be recording the batch locations to another list so they could be processed in a second phase using threads. RFC TODO: list some examples of how this outperforms previous pack-objects strategies. (This is coming in later commits that include performance test changes.) Signed-off-by: Derrick Stolee --- Documentation/git-pack-objects.adoc | 12 +- Documentation/technical/api-path-walk.adoc | 1 + builtin/pack-objects.c | 147 +++++++++++++++++++-- t/t5300-pack-object.sh | 17 +++ 4 files changed, 167 insertions(+), 10 deletions(-) diff --git a/Documentation/git-pack-objects.adoc b/Documentation/git-pack-objects.adoc index 7f69ae4855f6ce..a91aa7989cbeff 100644 --- a/Documentation/git-pack-objects.adoc +++ b/Documentation/git-pack-objects.adoc @@ -16,7 +16,7 @@ SYNOPSIS [--cruft] [--cruft-expiration=