lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20241029002208.1947947-3-dolinux.peng@gmail.com>
Date: Tue, 29 Oct 2024 08:22:07 +0800
From: Donglin Peng <dolinux.peng@...il.com>
To: andrii@...nel.org,
	eddyz87@...il.com
Cc: ast@...nel.org,
	rostedt@...dmis.org,
	mhiramat@...nel.org,
	bpf@...r.kernel.org,
	linux-trace-kernel@...r.kernel.org,
	linux-kselftest@...r.kernel.org,
	linux-kernel@...r.kernel.org,
	Donglin Peng <dolinux.peng@...il.com>
Subject: [PATCH v4 2/3] bpf: Using binary search to improve the performance of btf_find_by_name_kind

Currently, we are only using the linear search method to find the type id
by the name, which has a time complexity of O(n). This change involves
sorting the names of btf types in ascending order and using binary search,
which has a time complexity of O(log(n)). This idea was inspired by the
following patch:

60443c88f3a8 ("kallsyms: Improve the performance of kallsyms_lookup_name()").

At present, this improvement is only for searching in vmlinux's and
module's BTFs.

Another change is the search direction, where we search the BTF first and
then its base, the type id of the first matched btf_type will be returned.

Here is a time-consuming result that finding 87590 type ids by their names in
vmlinux's BTF.

Before: 158426 ms
After:     114 ms

The average lookup performance has improved more than 1000x in the above scenario.

Tested-by: Masami Hiramatsu (Google) <mhiramat@...nel.org>
Signed-off-by: Donglin Peng <dolinux.peng@...il.com>
---
v4:
 - move the modification of libbpf to another patch

v3:
 - Link: https://lore.kernel.org/all/20240608140835.965949-1-dolinux.peng@gmail.com/
 - Sort btf_types during build process other than during boot, to reduce the
   overhead of memory and boot time.

v2:
 - Link: https://lore.kernel.org/all/20230909091646.420163-1-pengdonglin@sangfor.com.cn
---
 include/linux/btf.h |   1 +
 kernel/bpf/btf.c    | 157 ++++++++++++++++++++++++++++++++++++++++----
 2 files changed, 147 insertions(+), 11 deletions(-)

diff --git a/include/linux/btf.h b/include/linux/btf.h
index b8a583194c4a..64c35aaa22fa 100644
--- a/include/linux/btf.h
+++ b/include/linux/btf.h
@@ -216,6 +216,7 @@ bool btf_is_module(const struct btf *btf);
 bool btf_is_vmlinux(const struct btf *btf);
 struct module *btf_try_get_module(const struct btf *btf);
 u32 btf_nr_types(const struct btf *btf);
+u32 btf_type_cnt(const struct btf *btf);
 struct btf *btf_base_btf(const struct btf *btf);
 bool btf_member_is_reg_int(const struct btf *btf, const struct btf_type *s,
 			   const struct btf_member *m,
diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
index 5cd1c7a23848..6d0d58989640 100644
--- a/kernel/bpf/btf.c
+++ b/kernel/bpf/btf.c
@@ -262,6 +262,7 @@ struct btf {
 	u32 data_size;
 	refcount_t refcnt;
 	u32 id;
+	u32 nr_types_sorted;
 	struct rcu_head rcu;
 	struct btf_kfunc_set_tab *kfunc_set_tab;
 	struct btf_id_dtor_kfunc_tab *dtor_kfunc_tab;
@@ -548,23 +549,102 @@ u32 btf_nr_types(const struct btf *btf)
 	return total;
 }
 
-s32 btf_find_by_name_kind(const struct btf *btf, const char *name, u8 kind)
+u32 btf_type_cnt(const struct btf *btf)
+{
+	return btf->start_id + btf->nr_types;
+}
+
+static s32 btf_find_by_name_bsearch(const struct btf *btf, const char *name,
+				    int *start, int *end)
 {
 	const struct btf_type *t;
-	const char *tname;
-	u32 i, total;
+	const char *name_buf;
+	int low, low_start, mid, high, high_end;
+	int ret, start_id;
+
+	start_id = btf->base_btf ? btf->start_id : 1;
+	low_start = low = start_id;
+	high_end = high = start_id + btf->nr_types_sorted - 1;
+
+	while (low <= high) {
+		mid = low + (high - low) / 2;
+		t = btf_type_by_id(btf, mid);
+		name_buf = btf_name_by_offset(btf, t->name_off);
+		ret = strcmp(name, name_buf);
+		if (ret > 0)
+			low = mid + 1;
+		else if (ret < 0)
+			high = mid - 1;
+		else
+			break;
+	}
 
-	total = btf_nr_types(btf);
-	for (i = 1; i < total; i++) {
-		t = btf_type_by_id(btf, i);
-		if (BTF_INFO_KIND(t->info) != kind)
-			continue;
+	if (low > high)
+		return -ESRCH;
 
-		tname = btf_name_by_offset(btf, t->name_off);
-		if (!strcmp(tname, name))
-			return i;
+	if (start) {
+		low = mid;
+		while (low > low_start) {
+			t = btf_type_by_id(btf, low-1);
+			name_buf = btf_name_by_offset(btf, t->name_off);
+			if (strcmp(name, name_buf))
+				break;
+			low--;
+		}
+		*start = low;
+	}
+
+	if (end) {
+		high = mid;
+		while (high < high_end) {
+			t = btf_type_by_id(btf, high+1);
+			name_buf = btf_name_by_offset(btf, t->name_off);
+			if (strcmp(name, name_buf))
+				break;
+			high++;
+		}
+		*end = high;
 	}
 
+	return mid;
+}
+
+s32 btf_find_by_name_kind(const struct btf *btf, const char *name, u8 kind)
+{
+	const struct btf_type *t;
+	const char *tname;
+	int start, end;
+	s32 id, total;
+
+	do {
+		if (btf->nr_types_sorted) {
+			/* binary search */
+			id = btf_find_by_name_bsearch(btf, name, &start, &end);
+			if (id > 0) {
+				while (start <= end) {
+					t = btf_type_by_id(btf, start);
+					if (BTF_INFO_KIND(t->info) == kind)
+						return start;
+					start++;
+				}
+			}
+		} else {
+			/* linear search */
+			total = btf_type_cnt(btf);
+			for (id = btf->base_btf ? btf->start_id : 1;
+				id < total; id++) {
+				t = btf_type_by_id(btf, id);
+				if (BTF_INFO_KIND(t->info) != kind)
+					continue;
+
+				tname = btf_name_by_offset(btf, t->name_off);
+				if (!strcmp(tname, name))
+					return id;
+			}
+		}
+		btf = btf->base_btf;
+	} while (btf);
+
 	return -ENOENT;
 }
 
@@ -6141,6 +6221,53 @@ int get_kern_ctx_btf_id(struct bpf_verifier_log *log, enum bpf_prog_type prog_ty
 	return kctx_type_id;
 }
 
+static int btf_check_sort(struct btf *btf, int start_id)
+{
+	int i, n, nr_names = 0;
+
+	n = btf_nr_types(btf);
+	for (i = start_id; i < n; i++) {
+		const struct btf_type *t;
+		const char *name;
+
+		t = btf_type_by_id(btf, i);
+		if (!t)
+			return -EINVAL;
+
+		name = btf_str_by_offset(btf, t->name_off);
+		if (!str_is_empty(name))
+			nr_names++;
+	}
+
+	for (i = 0; i < nr_names - 1; i++) {
+		const struct btf_type *t1, *t2;
+		const char *s1, *s2;
+
+		t1 = btf_type_by_id(btf, start_id + i);
+		if (!t1)
+			return -EINVAL;
+
+		s1 = btf_str_by_offset(btf, t1->name_off);
+		if (str_is_empty(s1))
+			goto out;
+
+		t2 = btf_type_by_id(btf, start_id + i + 1);
+		if (!t2)
+			return -EINVAL;
+
+		s2 = btf_str_by_offset(btf, t2->name_off);
+		if (str_is_empty(s2))
+			goto out;
+
+		if (strcmp(s1, s2) > 0)
+			goto out;
+	}
+
+	btf->nr_types_sorted = nr_names;
+out:
+	return 0;
+}
+
 BTF_ID_LIST(bpf_ctx_convert_btf_id)
 BTF_ID(struct, bpf_ctx_convert)
 
@@ -6212,6 +6339,10 @@ struct btf *btf_parse_vmlinux(void)
 	if (IS_ERR(btf))
 		goto err_out;
 
+	err = btf_check_sort(btf, 1);
+	if (err)
+		goto err_out;
+
 	/* btf_parse_vmlinux() runs under bpf_verifier_lock */
 	bpf_ctx_convert.t = btf_type_by_id(btf, bpf_ctx_convert_btf_id[0]);
 	err = btf_alloc_id(btf);
@@ -6315,6 +6446,10 @@ static struct btf *btf_parse_module(const char *module_name, const void *data,
 		base_btf = vmlinux_btf;
 	}
 
+	err = btf_check_sort(btf, btf_nr_types(base_btf));
+	if (err)
+		goto errout;
+
 	btf_verifier_env_free(env);
 	refcount_set(&btf->refcnt, 1);
 	return btf;
-- 
2.34.1


Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ