[PATCH v3 bpf-next 0/2] bpf: keep track of verifier insn_processed

[PATCH v3 bpf-next 0/2] bpf: keep track of verifier insn_processed

[Dave Marchevsky]

This is a followup to discussion around RFC patchset "bpf: keep track of
prog verification stats" [0]. The RFC elaborates on my usecase, but to
summarize: keeping track of verifier stats for programs as they - and
the kernels they run on - change over time can help developers of
individual programs and BPF kernel folks.

The RFC added a verif_stats to the uapi which contained most of the info
which verifier prints currently. Feedback here was to avoid polluting
uapi with stats that might be meaningless after major changes to the
verifier, but that insn_processed or conceptually similar number would
exist in the long term and was safe to expose.

So let's expose just insn_processed via bpf_prog_info and fdinfo for now
and explore good ways of getting more complicated stats in the future.

[0] https://lore.kernel.org/bpf/20210920151112.3770991-1-davemarchevsky@fb.com/

v2->v3:
  * Remove unnecessary check in patch 2's test [Andrii]
  * Go back to adding new u32 in bpf_prog_info (vs using spare bits) [Andrii]
	* Rebase + add acks [Andrii, John]

v1->v2:
  * Rename uapi field from insn_processed to verified_insns [Daniel]
  * use 31 bits of existing bitfield space in bpf_prog_info [Daniel]
  * change underlying type from 64-> 32 bits [Daniel]

Dave Marchevsky (2):
  bpf: add verified_insns to bpf_prog_info and fdinfo
  selftests/bpf: add verif_stats test

 include/linux/bpf.h                           |  1 +
 include/uapi/linux/bpf.h                      |  1 +
 kernel/bpf/syscall.c                          |  8 ++++--
 kernel/bpf/verifier.c                         |  1 +
 tools/include/uapi/linux/bpf.h                |  1 +
 .../selftests/bpf/prog_tests/verif_stats.c    | 28 +++++++++++++++++++
 6 files changed, 38 insertions(+), 2 deletions(-)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/verif_stats.c

-- 
2.30.2

[PATCH v3 bpf-next 1/2] bpf: add verified_insns to bpf_prog_info and fdinfo

[Dave Marchevsky]

This stat is currently printed in the verifier log and not stored
anywhere. To ease consumption of this data, add a field to bpf_prog_aux
so it can be exposed via BPF_OBJ_GET_INFO_BY_FD and fdinfo.

Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
---
 include/linux/bpf.h            | 1 +
 include/uapi/linux/bpf.h       | 1 +
 kernel/bpf/syscall.c           | 8 ++++++--
 kernel/bpf/verifier.c          | 1 +
 tools/include/uapi/linux/bpf.h | 1 +
 5 files changed, 10 insertions(+), 2 deletions(-)

diff --git a/include/linux/bpf.h b/include/linux/bpf.h
index d604c8251d88..c93fd845a758 100644
--- a/include/linux/bpf.h
+++ b/include/linux/bpf.h
@@ -887,6 +887,7 @@ struct bpf_prog_aux {
 	struct bpf_prog *prog;
 	struct user_struct *user;
 	u64 load_time; /* ns since boottime */
+	u32 verified_insns;
 	struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
 	char name[BPF_OBJ_NAME_LEN];
 #ifdef CONFIG_SECURITY
diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
index 6fc59d61937a..35571a9878ac 100644
--- a/include/uapi/linux/bpf.h
+++ b/include/uapi/linux/bpf.h
@@ -5613,6 +5613,7 @@ struct bpf_prog_info {
 	__u64 run_time_ns;
 	__u64 run_cnt;
 	__u64 recursion_misses;
+	__u32 verified_insns;
 } __attribute__((aligned(8)));
 
 struct bpf_map_info {
diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
index 4e50c0bfdb7d..5beb321b3b3b 100644
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c
@@ -1848,7 +1848,8 @@ static void bpf_prog_show_fdinfo(struct seq_file *m, struct file *filp)
 		   "prog_id:\t%u\n"
 		   "run_time_ns:\t%llu\n"
 		   "run_cnt:\t%llu\n"
-		   "recursion_misses:\t%llu\n",
+		   "recursion_misses:\t%llu\n"
+		   "verified_insns:\t%u\n",
 		   prog->type,
 		   prog->jited,
 		   prog_tag,
@@ -1856,7 +1857,8 @@ static void bpf_prog_show_fdinfo(struct seq_file *m, struct file *filp)
 		   prog->aux->id,
 		   stats.nsecs,
 		   stats.cnt,
-		   stats.misses);
+		   stats.misses,
+		   prog->aux->verified_insns);
 }
 #endif
 
@@ -3625,6 +3627,8 @@ static int bpf_prog_get_info_by_fd(struct file *file,
 	info.run_cnt = stats.cnt;
 	info.recursion_misses = stats.misses;
 
+	info.verified_insns = prog->aux->verified_insns;
+
 	if (!bpf_capable()) {
 		info.jited_prog_len = 0;
 		info.xlated_prog_len = 0;
diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index 21cdff35a2f9..c6616e325803 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c
@@ -14033,6 +14033,7 @@ int bpf_check(struct bpf_prog **prog, union bpf_attr *attr, bpfptr_t uattr)
 
 	env->verification_time = ktime_get_ns() - start_time;
 	print_verification_stats(env);
+	env->prog->aux->verified_insns = env->insn_processed;
 
 	if (log->level && bpf_verifier_log_full(log))
 		ret = -ENOSPC;
diff --git a/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h
index 6fc59d61937a..35571a9878ac 100644
--- a/tools/include/uapi/linux/bpf.h
+++ b/tools/include/uapi/linux/bpf.h
@@ -5613,6 +5613,7 @@ struct bpf_prog_info {
 	__u64 run_time_ns;
 	__u64 run_cnt;
 	__u64 recursion_misses;
+	__u32 verified_insns;
 } __attribute__((aligned(8)));
 
 struct bpf_map_info {
-- 
2.30.2

[PATCH v3 bpf-next 2/2] selftests/bpf: add verif_stats test

[Dave Marchevsky]

verified_insns field was added to response of bpf_obj_get_info_by_fd
call on a prog. Confirm that it's being populated by loading a simple
program and asking for its info.

Signed-off-by: Dave Marchevsky <davemarchevsky@fb.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
---
 .../selftests/bpf/prog_tests/verif_stats.c    | 28 +++++++++++++++++++
 1 file changed, 28 insertions(+)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/verif_stats.c

diff --git a/tools/testing/selftests/bpf/prog_tests/verif_stats.c b/tools/testing/selftests/bpf/prog_tests/verif_stats.c
new file mode 100644
index 000000000000..b4bae1340cf1
--- /dev/null
+++ b/tools/testing/selftests/bpf/prog_tests/verif_stats.c
@@ -0,0 +1,28 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2021 Facebook */
+
+#include <test_progs.h>
+
+#include "trace_vprintk.lskel.h"
+
+void test_verif_stats(void)
+{
+	__u32 len = sizeof(struct bpf_prog_info);
+	struct bpf_prog_info info = {};
+	struct trace_vprintk *skel;
+	int err;
+
+	skel = trace_vprintk__open_and_load();
+	if (!ASSERT_OK_PTR(skel, "trace_vprintk__open_and_load"))
+		goto cleanup;
+
+	err = bpf_obj_get_info_by_fd(skel->progs.sys_enter.prog_fd, &info, &len);
+	if (!ASSERT_OK(err, "bpf_obj_get_info_by_fd"))
+		goto cleanup;
+
+	if (!ASSERT_GT(info.verified_insns, 0, "verified_insns"))
+		goto cleanup;
+
+cleanup:
+	trace_vprintk__destroy(skel);
+}
-- 
2.30.2

Re: [PATCH v3 bpf-next 0/2] bpf: keep track of verifier insn_processed

[patchwork-bot+netdevbpf]

Hello:

This series was applied to bpf/bpf-next.git (master)
by Andrii Nakryiko <andrii@kernel.org>:

On Wed, 20 Oct 2021 00:48:16 -0700 you wrote:
> This is a followup to discussion around RFC patchset "bpf: keep track of
> prog verification stats" [0]. The RFC elaborates on my usecase, but to
> summarize: keeping track of verifier stats for programs as they - and
> the kernels they run on - change over time can help developers of
> individual programs and BPF kernel folks.
> 
> The RFC added a verif_stats to the uapi which contained most of the info
> which verifier prints currently. Feedback here was to avoid polluting
> uapi with stats that might be meaningless after major changes to the
> verifier, but that insn_processed or conceptually similar number would
> exist in the long term and was safe to expose.
> 
> [...]

Here is the summary with links:
  - [v3,bpf-next,1/2] bpf: add verified_insns to bpf_prog_info and fdinfo
    https://git.kernel.org/bpf/bpf-next/c/aba64c7da983
  - [v3,bpf-next,2/2] selftests/bpf: add verif_stats test
    https://git.kernel.org/bpf/bpf-next/c/e1b9023fc7ab

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html