Merge pull request #30291 from keszybz/seccomp-unknown-syscall

Backwardscompatibly handle syscalls unknown to us or libseccomp

Author: bluca

src/core/exec-invoke.c

@@ -5119,7 +5119,7 @@ int exec_invoke(
 #endif
 
 #if HAVE_SECCOMP
-                /* This really should remain as close to the execve() as possible, to make sure our own code is unaffected
+                /* This really should remain as close to the execve() as possible, to make sure our own code is affected
                  * by the filter as little as possible. */
                 r = apply_syscall_filter(context, params, needs_ambient_hack);
                 if (r < 0) {

src/nspawn/nspawn-seccomp.c

@@ -169,7 +169,7 @@ static int add_syscall_filters(
         /* We have a large filter here, so let's turn on the binary tree mode if possible. */
         r = seccomp_attr_set(ctx, SCMP_FLTATR_CTL_OPTIMIZE, 2);
         if (r < 0)
-                return r;
+                log_warning_errno(r, "Failed to set SCMP_FLTATR_CTL_OPTIMIZE, ignoring: %m");
 #endif
 
         return 0;

src/shared/seccomp-util.c

@@ -1129,7 +1129,9 @@ int seccomp_load_syscall_filter_set_raw(uint32_t default_action, Hashmap* filter
 
                 log_trace("Operating on architecture: %s", seccomp_arch_to_string(arch));
 
-                r = seccomp_init_for_arch(&seccomp, arch, default_action);
+                /* We install ENOSYS as the default action, but it will only apply to syscalls which are not
+                 * in the @known set. */
+                r = seccomp_init_for_arch(&seccomp, arch, SCMP_ACT_ERRNO(ENOSYS));
                 if (r < 0)
                         return r;
 
@@ -1164,6 +1166,30 @@ int seccomp_load_syscall_filter_set_raw(uint32_t default_action, Hashmap* filter
                         }
                 }
 
+                NULSTR_FOREACH(name, syscall_filter_sets[SYSCALL_FILTER_SET_KNOWN].value) {
+                        int id;
+
+                        id = seccomp_syscall_resolve_name(name);
+                        if (id < 0)
+                                continue;
+
+                        /* Ignore the syscall if it was already handled above */
+                        if (hashmap_contains(filter, INT_TO_PTR(id + 1)))
+                                continue;
+
+                        r = seccomp_rule_add_exact(seccomp, default_action, id, 0);
+                        if (r < 0 && r != -EDOM)  /* EDOM means that the syscall is not available for arch */
+                                return log_debug_errno(r, "Failed to add rule for system call %s() / %d: %m",
+                                                       name, id);
+                }
+
+#if (SCMP_VER_MAJOR == 2 && SCMP_VER_MINOR >= 5) || SCMP_VER_MAJOR > 2
+                /* We have a large filter here, so let's turn on the binary tree mode if possible. */
+                r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_OPTIMIZE, 2);
+                if (r < 0)
+                        log_warning_errno(r, "Failed to set SCMP_FLTATR_CTL_OPTIMIZE, ignoring: %m");
+#endif
+
                 r = seccomp_load(seccomp);
                 if (ERRNO_IS_NEG_SECCOMP_FATAL(r))
                         return r;
@@ -1223,7 +1249,7 @@ int seccomp_parse_syscall_filter(
                                 return -EINVAL;
 
                         log_syntax(unit, FLAGS_SET(flags, SECCOMP_PARSE_LOG) ? LOG_WARNING : LOG_DEBUG, filename, line, 0,
-                                   "Failed to parse system call, ignoring: %s", name);
+                                   "System call %s is not known, ignoring.", name);
                         return 0;
                 }
 
@@ -1981,7 +2007,7 @@ int seccomp_filter_set_add(Hashmap *filter, bool add, const SyscallFilterSet *se
 
                         id = seccomp_syscall_resolve_name(i);
                         if (id == __NR_SCMP_ERROR) {
-                                log_debug("Couldn't resolve system call, ignoring: %s", i);
+                                log_debug("System call %s is not known, ignoring.", i);
                                 continue;
                         }