fprobe: use rhashtable for fprobe_ip_table

include/linux/fprobe.h

@@ -7,6 +7,7 @@
 #include <linux/ftrace.h>
 #include <linux/rcupdate.h>
 #include <linux/refcount.h>
+#include <linux/rhashtable.h>
 #include <linux/slab.h>
 
 struct fprobe;
@@ -26,7 +27,7 @@ typedef void (*fprobe_exit_cb)(struct fprobe *fp, unsigned long entry_ip,
  * @fp: The fprobe which owns this.
  */
 struct fprobe_hlist_node {
-	struct hlist_node	hlist;
+	struct rhlist_head	hlist;
 	unsigned long		addr;
 	struct fprobe		*fp;
 };

kernel/trace/fprobe.c

@@ -10,6 +10,7 @@
 #include <linux/kprobes.h>
 #include <linux/list.h>
 #include <linux/mutex.h>
+#include <linux/rhashtable.h>
 #include <linux/slab.h>
 #include <linux/sort.h>
 
@@ -41,47 +42,46 @@
  *  - RCU hlist traversal under disabling preempt
  */
 static struct hlist_head fprobe_table[FPROBE_TABLE_SIZE];
-static struct hlist_head fprobe_ip_table[FPROBE_IP_TABLE_SIZE];
+static struct rhltable fprobe_ip_table;
 static DEFINE_MUTEX(fprobe_mutex);
 
-/*
- * Find first fprobe in the hlist. It will be iterated twice in the entry
- * probe, once for correcting the total required size, the second time is
- * calling back the user handlers.
- * Thus the hlist in the fprobe_table must be sorted and new probe needs to
- * be added *before* the first fprobe.
- */
-static struct fprobe_hlist_node *find_first_fprobe_node(unsigned long ip)
+static u32 fprobe_node_hashfn(const void *data, u32 len, u32 seed)
 {
-	struct fprobe_hlist_node *node;
-	struct hlist_head *head;
+	return hash_ptr(*(unsigned long **)data, 32);
+}
 
-	head = &fprobe_ip_table[hash_ptr((void *)ip, FPROBE_IP_HASH_BITS)];
-	hlist_for_each_entry_rcu(node, head, hlist,
-				 lockdep_is_held(&fprobe_mutex)) {
-		if (node->addr == ip)
-			return node;
-	}
-	return NULL;
+static int fprobe_node_cmp(struct rhashtable_compare_arg *arg,
+			   const void *ptr)
+{
+	unsigned long key = *(unsigned long *)arg->key;
+	const struct fprobe_hlist_node *n = ptr;
+
+	return n->addr != key;
 }
-NOKPROBE_SYMBOL(find_first_fprobe_node);
 
-/* Node insertion and deletion requires the fprobe_mutex */
-static void insert_fprobe_node(struct fprobe_hlist_node *node)
+static u32 fprobe_node_obj_hashfn(const void *data, u32 len, u32 seed)
 {
-	unsigned long ip = node->addr;
-	struct fprobe_hlist_node *next;
-	struct hlist_head *head;
+	const struct fprobe_hlist_node *n = data;
+
+	return hash_ptr((void *)n->addr, 32);
+}
+
+static const struct rhashtable_params fprobe_rht_params = {
+	.head_offset		= offsetof(struct fprobe_hlist_node, hlist),
+	.key_offset		= offsetof(struct fprobe_hlist_node, addr),
+	.key_len		= sizeof_field(struct fprobe_hlist_node, addr),
+	.hashfn			= fprobe_node_hashfn,
+	.obj_hashfn		= fprobe_node_obj_hashfn,
+	.obj_cmpfn		= fprobe_node_cmp,
+	.automatic_shrinking	= true,
+};
 
+/* Node insertion and deletion requires the fprobe_mutex */
+static int insert_fprobe_node(struct fprobe_hlist_node *node)
+{
 	lockdep_assert_held(&fprobe_mutex);
 
-	next = find_first_fprobe_node(ip);
-	if (next) {
-		hlist_add_before_rcu(&node->hlist, &next->hlist);
-		return;
-	}
-	head = &fprobe_ip_table[hash_ptr((void *)ip, FPROBE_IP_HASH_BITS)];
-	hlist_add_head_rcu(&node->hlist, head);
+	return rhltable_insert(&fprobe_ip_table, &node->hlist, fprobe_rht_params);
 }
 
 /* Return true if there are synonims */
@@ -92,9 +92,11 @@ static bool delete_fprobe_node(struct fprobe_hlist_node *node)
 	/* Avoid double deleting */
 	if (READ_ONCE(node->fp) != NULL) {
 		WRITE_ONCE(node->fp, NULL);
-		hlist_del_rcu(&node->hlist);
+		rhltable_remove(&fprobe_ip_table, &node->hlist,
+				fprobe_rht_params);
 	}
-	return !!find_first_fprobe_node(node->addr);
+	return !!rhltable_lookup(&fprobe_ip_table, &node->addr,
+				 fprobe_rht_params);
 }
 
 /* Check existence of the fprobe */
@@ -249,9 +251,10 @@ static inline int __fprobe_kprobe_handler(unsigned long ip, unsigned long parent
 static int fprobe_entry(struct ftrace_graph_ent *trace, struct fgraph_ops *gops,
 			struct ftrace_regs *fregs)
 {
-	struct fprobe_hlist_node *node, *first;
 	unsigned long *fgraph_data = NULL;
 	unsigned long func = trace->func;
+	struct fprobe_hlist_node *node;
+	struct rhlist_head *head, *pos;
 	unsigned long ret_ip;
 	int reserved_words;
 	struct fprobe *fp;
@@ -260,14 +263,11 @@ static int fprobe_entry(struct ftrace_graph_ent *trace, struct fgraph_ops *gops,
 	if (WARN_ON_ONCE(!fregs))
 		return 0;
 
-	first = node = find_first_fprobe_node(func);
-	if (unlikely(!first))
-		return 0;
-
+	head = rhltable_lookup(&fprobe_ip_table, &func, fprobe_rht_params);
 	reserved_words = 0;
-	hlist_for_each_entry_from_rcu(node, hlist) {
+	rhl_for_each_entry_rcu(node, pos, head, hlist) {
 		if (node->addr != func)
-			break;
+			continue;
 		fp = READ_ONCE(node->fp);
 		if (!fp || !fp->exit_handler)
 			continue;
@@ -278,13 +278,12 @@ static int fprobe_entry(struct ftrace_graph_ent *trace, struct fgraph_ops *gops,
 		reserved_words +=
 			FPROBE_HEADER_SIZE_IN_LONG + SIZE_IN_LONG(fp->entry_data_size);
 	}
-	node = first;
 	if (reserved_words) {
 		fgraph_data = fgraph_reserve_data(gops->idx, reserved_words * sizeof(long));
 		if (unlikely(!fgraph_data)) {
-			hlist_for_each_entry_from_rcu(node, hlist) {
+			rhl_for_each_entry_rcu(node, pos, head, hlist) {
 				if (node->addr != func)
-					break;
+					continue;
 				fp = READ_ONCE(node->fp);
 				if (fp && !fprobe_disabled(fp))
 					fp->nmissed++;
@@ -299,12 +298,12 @@ static int fprobe_entry(struct ftrace_graph_ent *trace, struct fgraph_ops *gops,
 	 */
 	ret_ip = ftrace_regs_get_return_address(fregs);
 	used = 0;
-	hlist_for_each_entry_from_rcu(node, hlist) {
+	rhl_for_each_entry_rcu(node, pos, head, hlist) {
 		int data_size;
 		void *data;
 
 		if (node->addr != func)
-			break;
+			continue;
 		fp = READ_ONCE(node->fp);
 		if (!fp || fprobe_disabled(fp))
 			continue;
@@ -448,34 +447,31 @@ static int fprobe_addr_list_add(struct fprobe_addr_list *alist, unsigned long ad
 	return 0;
 }
 
-static void fprobe_remove_node_in_module(struct module *mod, struct hlist_head *head,
-					struct fprobe_addr_list *alist)
+static void fprobe_remove_node_in_module(struct module *mod, struct fprobe_hlist_node *node,
+					 struct fprobe_addr_list *alist)
 {
-	struct fprobe_hlist_node *node;
 	int ret = 0;
 
-	hlist_for_each_entry_rcu(node, head, hlist,
-				 lockdep_is_held(&fprobe_mutex)) {
-		if (!within_module(node->addr, mod))
-			continue;
-		if (delete_fprobe_node(node))
-			continue;
-		/*
-		 * If failed to update alist, just continue to update hlist.
-		 * Therefore, at list user handler will not hit anymore.
-		 */
-		if (!ret)
-			ret = fprobe_addr_list_add(alist, node->addr);
-	}
+	if (!within_module(node->addr, mod))
+		return;
+	if (delete_fprobe_node(node))
+		return;
+	/*
+	 * If failed to update alist, just continue to update hlist.
+	 * Therefore, at list user handler will not hit anymore.
+	 */
+	if (!ret)
+		ret = fprobe_addr_list_add(alist, node->addr);
 }
 
 /* Handle module unloading to manage fprobe_ip_table. */
 static int fprobe_module_callback(struct notifier_block *nb,
 				  unsigned long val, void *data)
 {
 	struct fprobe_addr_list alist = {.size = FPROBE_IPS_BATCH_INIT};
+	struct fprobe_hlist_node *node;
+	struct rhashtable_iter iter;
 	struct module *mod = data;
-	int i;
 
 	if (val != MODULE_STATE_GOING)
 		return NOTIFY_DONE;
@@ -486,8 +482,16 @@ static int fprobe_module_callback(struct notifier_block *nb,
 		return NOTIFY_DONE;
 
 	mutex_lock(&fprobe_mutex);
-	for (i = 0; i < FPROBE_IP_TABLE_SIZE; i++)
-		fprobe_remove_node_in_module(mod, &fprobe_ip_table[i], &alist);
+	rhltable_walk_enter(&fprobe_ip_table, &iter);
+	do {
+		rhashtable_walk_start(&iter);
+
+		while ((node = rhashtable_walk_next(&iter)) && !IS_ERR(node))
+			fprobe_remove_node_in_module(mod, node, &alist);
+
+		rhashtable_walk_stop(&iter);
+	} while (node == ERR_PTR(-EAGAIN));
+	rhashtable_walk_exit(&iter);
 
 	if (alist.index < alist.size && alist.index > 0)
 		ftrace_set_filter_ips(&fprobe_graph_ops.ops,
@@ -727,8 +731,16 @@ int register_fprobe_ips(struct fprobe *fp, unsigned long *addrs, int num)
 	ret = fprobe_graph_add_ips(addrs, num);
 	if (!ret) {
 		add_fprobe_hash(fp);
-		for (i = 0; i < hlist_array->size; i++)
-			insert_fprobe_node(&hlist_array->array[i]);
+		for (i = 0; i < hlist_array->size; i++) {
+			ret = insert_fprobe_node(&hlist_array->array[i]);
+			if (ret)
+				break;
+		}
+		/* fallback on insert error */
+		if (ret) {
+			for (i--; i >= 0; i--)
+				delete_fprobe_node(&hlist_array->array[i]);
+		}
 	}
 	mutex_unlock(&fprobe_mutex);
 
@@ -824,3 +836,10 @@ int unregister_fprobe(struct fprobe *fp)
 	return ret;
 }
 EXPORT_SYMBOL_GPL(unregister_fprobe);
+
+static int __init fprobe_initcall(void)
+{
+	rhltable_init(&fprobe_ip_table, &fprobe_rht_params);
+	return 0;
+}
+late_initcall(fprobe_initcall);

tools/testing/selftests/bpf/bench.c

@@ -510,6 +510,8 @@ extern const struct bench bench_trig_kretprobe;
 extern const struct bench bench_trig_kprobe_multi;
 extern const struct bench bench_trig_kretprobe_multi;
 extern const struct bench bench_trig_fentry;
+extern const struct bench bench_trig_kprobe_multi_all;
+extern const struct bench bench_trig_kretprobe_multi_all;
 extern const struct bench bench_trig_fexit;
 extern const struct bench bench_trig_fmodret;
 extern const struct bench bench_trig_tp;
@@ -578,6 +580,8 @@ static const struct bench *benchs[] = {
 	&bench_trig_kprobe_multi,
 	&bench_trig_kretprobe_multi,
 	&bench_trig_fentry,
+	&bench_trig_kprobe_multi_all,
+	&bench_trig_kretprobe_multi_all,
 	&bench_trig_fexit,
 	&bench_trig_fmodret,
 	&bench_trig_tp,

tools/testing/selftests/bpf/benchs/bench_trigger.c

@@ -226,6 +226,58 @@ static void trigger_fentry_setup(void)
 	attach_bpf(ctx.skel->progs.bench_trigger_fentry);
 }
 
+static void attach_ksyms_all(struct bpf_program *empty, bool kretprobe)
+{
+	LIBBPF_OPTS(bpf_kprobe_multi_opts, opts);
+	char **syms = NULL;
+	size_t cnt = 0;
+
+	if (bpf_get_ksyms(&syms, &cnt, true)) {
+		printf("failed to get ksyms\n");
+		exit(1);
+	}
+
+	printf("found %zu ksyms\n", cnt);
+	opts.syms = (const char **) syms;
+	opts.cnt = cnt;
+	opts.retprobe = kretprobe;
+	/* attach empty to all the kernel functions except bpf_get_numa_node_id. */
+	if (!bpf_program__attach_kprobe_multi_opts(empty, NULL, &opts)) {
+		printf("failed to attach bpf_program__attach_kprobe_multi_opts to all\n");
+		exit(1);
+	}
+}
+
+static void trigger_kprobe_multi_all_setup(void)
+{
+	struct bpf_program *prog, *empty;
+
+	setup_ctx();
+	empty = ctx.skel->progs.bench_kprobe_multi_empty;
+	prog = ctx.skel->progs.bench_trigger_kprobe_multi;
+	bpf_program__set_autoload(empty, true);
+	bpf_program__set_autoload(prog, true);
+	load_ctx();
+
+	attach_ksyms_all(empty, false);
+	attach_bpf(prog);
+}
+
+static void trigger_kretprobe_multi_all_setup(void)
+{
+	struct bpf_program *prog, *empty;
+
+	setup_ctx();
+	empty = ctx.skel->progs.bench_kretprobe_multi_empty;
+	prog = ctx.skel->progs.bench_trigger_kretprobe_multi;
+	bpf_program__set_autoload(empty, true);
+	bpf_program__set_autoload(prog, true);
+	load_ctx();
+
+	attach_ksyms_all(empty, true);
+	attach_bpf(prog);
+}
+
 static void trigger_fexit_setup(void)
 {
 	setup_ctx();
@@ -512,6 +564,8 @@ BENCH_TRIG_KERNEL(kretprobe, "kretprobe");
 BENCH_TRIG_KERNEL(kprobe_multi, "kprobe-multi");
 BENCH_TRIG_KERNEL(kretprobe_multi, "kretprobe-multi");
 BENCH_TRIG_KERNEL(fentry, "fentry");
+BENCH_TRIG_KERNEL(kprobe_multi_all, "kprobe-multi-all");
+BENCH_TRIG_KERNEL(kretprobe_multi_all, "kretprobe-multi-all");
 BENCH_TRIG_KERNEL(fexit, "fexit");
 BENCH_TRIG_KERNEL(fmodret, "fmodret");
 BENCH_TRIG_KERNEL(tp, "tp");

tools/testing/selftests/bpf/benchs/run_bench_trigger.sh

@@ -6,8 +6,8 @@ def_tests=( \
 	usermode-count kernel-count syscall-count \
 	fentry fexit fmodret \
 	rawtp tp \
-	kprobe kprobe-multi \
-	kretprobe kretprobe-multi \
+	kprobe kprobe-multi kprobe-multi-all \
+	kretprobe kretprobe-multi kretprobe-multi-all \
 )
 
 tests=("$@")