Skip to contentSkip to conversation

bpf: Fix WARNING in bpf_trampoline_multi_detach

PATCH Superseded Leon Hwang 3 participants 7 messages Jul 10 – Jul 11

Subsystems (4)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Leon Hwang

author
#0 lore raw

cover bpf: Fix WARNING in bpf_trampoline_multi_detach

Since tracing_multi link does not set ftrace_managed, it would fail to release the tracing_multi link when attaching tracing_multi link and then attaching fentry link.

[    3.714215] WARNING: kernel/bpf/trampoline.c:1727 at bpf_trampoline_multi_detach+0x20b/0x240, CPU#1: test_progs/97
...
[    3.733170]  bpf_tracing_multi_link_release+0x14/0x30
[    3.733890]  bpf_link_free+0x58/0x130
[    3.734414]  bpf_link_release+0x23/0x30

Fix it by setting 'ftrace_managed = true' in register_fentry_multi().

Note: the reason of targeting bpf-next tree is I'm planning to add tracing_multi link support for bpf progs, that would rely on the fix and have code conflict with the test.

Leon Hwang (2):
bpf: Mark tracing_multi trampolines as ftrace managed selftests/bpf: Test fentry link after tracing_multi link

 kernel/bpf/trampoline.c                       |  1 +
 .../selftests/bpf/prog_tests/tracing_multi.c  | 69 +++++++++++++++++++
 .../progs/tracing_multi_intersect_attach.c    |  8 +++
 3 files changed, 78 insertions(+)
Signature · 1 line
2.55.0
author
#1 lore raw

PATCH 1/2 bpf: Mark tracing_multi trampolines as ftrace managed

Since tracing_multi link does not set ftrace_managed, it would fail to release the tracing_multi link when attaching tracing_multi link and then attaching fentry link.

[    3.714215] WARNING: kernel/bpf/trampoline.c:1727 at bpf_trampoline_multi_detach+0x20b/0x240, CPU#1: test_progs/97
...
[    3.733170]  bpf_tracing_multi_link_release+0x14/0x30
[    3.733890]  bpf_link_free+0x58/0x130
[    3.734414]  bpf_link_release+0x23/0x30

Fix it by setting 'ftrace_managed = true' in register_fentry_multi().

Fixes aef4dfa790b2 ("bpf: Add bpf_trampoline_multi_attach/detach functions")
---
 kernel/bpf/trampoline.c | 1 +
 1 file changed, 1 insertion(+)
kernel/bpf/trampoline.c +1 -0
index 1a721fc4bef5..6eadf64f7ec9 100644
@@ -1536,6 +1536,7 @@ static int register_fentry_multi(struct bpf_trampoline *tr, struct bpf_tramp_ima
 	if (bpf_trampoline_use_jmp(tr->flags))
 		addr = ftrace_jmp_set(addr);
 
+	tr->func.ftrace_managed = true;
 	ftrace_hash_add(data->reg, data->entry, ip, addr);
 	tr->cur_image = im;
 	return 0;
Signature · 1 line
2.55.0
Acked-by on v1 1/2
#2 lore raw

Re: [PATCH bpf-next 1/2] bpf: Mark tracing_multi trampolines as ftrace managed

Quoted 23 lines · Leon Hwang

Since tracing_multi link does not set ftrace_managed, it would fail to release the tracing_multi link when attaching tracing_multi link and then attaching fentry link.

[    3.714215] WARNING: kernel/bpf/trampoline.c:1727 at bpf_trampoline_multi_detach+0x20b/0x240, CPU#1: test_progs/97
...
[    3.733170]  bpf_tracing_multi_link_release+0x14/0x30
[    3.733890]  bpf_link_free+0x58/0x130
[    3.734414]  bpf_link_release+0x23/0x30

Fix it by setting 'ftrace_managed = true' in register_fentry_multi().

Fixes aef4dfa790b2 ("bpf: Add bpf_trampoline_multi_attach/detach functions")
---
 kernel/bpf/trampoline.c | 1 +
 1 file changed, 1 insertion(+)
kernel/bpf/trampoline.c +0 -0
index 1a721fc4bef5..6eadf64f7ec9 100644

kernel/bpf/trampoline.c static int register_fentry_multi(struct bpf_trampoline *tr, struct bpf_tramp_ima · line 1539

Quoted 4 lines
kernel/bpf/trampoline.c +1 -0
@@ -1536,6 +1536,7 @@ static int register_fentry_multi(struct bpf_trampoline *tr, struct bpf_tramp_ima
 	if (bpf_trampoline_use_jmp(tr->flags))
 		addr = ftrace_jmp_set(addr);
 
+	tr->func.ftrace_managed = true;

I wonder we could set tr->func.ftrace_managed early in bpf_trampoline_get, but we could do that as follow up

thanks,
jirka

Quoted 6 lines
kernel/bpf/trampoline.c +0 -0
 	ftrace_hash_add(data->reg, data->entry, ip, addr);
 	tr->cur_image = im;
 	return 0;
Signature · 1 line
2.55.0
author
#3 lore raw

Re: [PATCH bpf-next 1/2] bpf: Mark tracing_multi trampolines as ftrace managed

Quoted 27 lines · Jiri Olsa
Quoted 26 lines · Leon Hwang

Since tracing_multi link does not set ftrace_managed, it would fail to release the tracing_multi link when attaching tracing_multi link and then attaching fentry link.

[    3.714215] WARNING: kernel/bpf/trampoline.c:1727 at bpf_trampoline_multi_detach+0x20b/0x240, CPU#1: test_progs/97
...
[    3.733170]  bpf_tracing_multi_link_release+0x14/0x30
[    3.733890]  bpf_link_free+0x58/0x130
[    3.734414]  bpf_link_release+0x23/0x30

Fix it by setting 'ftrace_managed = true' in register_fentry_multi().

Fixes aef4dfa790b2 ("bpf: Add bpf_trampoline_multi_attach/detach functions")
---
 kernel/bpf/trampoline.c | 1 +
 1 file changed, 1 insertion(+)
kernel/bpf/trampoline.c +0 -0
index 1a721fc4bef5..6eadf64f7ec9 100644
@@ -1536,6 +1536,7 @@ static int register_fentry_multi(struct bpf_trampoline *tr, struct bpf_tramp_ima
 	if (bpf_trampoline_use_jmp(tr->flags))
 		addr = ftrace_jmp_set(addr);
 
Quoted 4 lines

kernel/bpf/trampoline.c static int register_fentry_multi(struct bpf_trampoline *tr, struct bpf_tramp_ima · line 1539

Quoted 1 line
kernel/bpf/trampoline.c +1 -0
+	tr->func.ftrace_managed = true;

I wonder we could set tr->func.ftrace_managed early in bpf_trampoline_get, but we could do that as follow up

Makes sense. Let me try it when adding tracing_multi link support for bpf progs.

Quoted 2 lines

Thanks for your review.

Leon

Quoted 11 lines

thanks,
jirka

Quoted 6 lines
kernel/bpf/trampoline.c +0 -0
 	ftrace_hash_add(data->reg, data->entry, ip, addr);
 	tr->cur_image = im;
 	return 0;
Signature · 1 line
2.55.0
author
#4 lore raw

PATCH 2/2 selftests/bpf: Test fentry link after tracing_multi link

In reply to Leon Hwang

Verify that there's no any issue to attach tracing_multi link, then attach fentry link.

Assisted-by Codex:gpt-5.6-sol
---
 .../selftests/bpf/prog_tests/tracing_multi.c  | 69 +++++++++++++++++++
 .../progs/tracing_multi_intersect_attach.c    |  8 +++
 2 files changed, 77 insertions(+)
tools/testing/selftests/bpf/prog_tests/tracing_multi.c +69 -0
index f02ffc7f41d7..0aa9532a05cf 100644
@@ -460,6 +460,73 @@ static void test_intersect(void)
 	tracing_multi_intersect__destroy(skel);
 }
 
+static void test_fentry_after_multi(void)
+{
+	static const char * const funcs[] = {
+		"bpf_fentry_test1",
+	};
+	struct bpf_link *fentry_link = NULL, *multi_link = NULL;
+	struct tracing_multi_intersect *skel = NULL;
+	LIBBPF_OPTS(bpf_tracing_multi_opts, opts);
+	LIBBPF_OPTS(bpf_test_run_opts, topts);
+	__u32 *ids = NULL;
+	int err;
+
+	skel = tracing_multi_intersect__open_and_load();
+	if (!ASSERT_OK_PTR(skel, "tracing_multi_intersect__open_and_load"))
+		return;
+
+	skel->bss->pid = getpid();
+
+	ids = get_ids(funcs, ARRAY_SIZE(funcs), NULL);
+	if (!ASSERT_OK_PTR(ids, "get_ids"))
+		goto cleanup;
+
+	opts.ids = ids;
+	opts.cnt = ARRAY_SIZE(funcs);
+	multi_link = bpf_program__attach_tracing_multi(skel->progs.fentry_1, NULL, &opts);
+	if (!ASSERT_OK_PTR(multi_link, "attach_multi"))
+		goto cleanup;
+
+	fentry_link = bpf_program__attach(skel->progs.fentry);
+	if (!ASSERT_OK_PTR(fentry_link, "attach_fentry"))
+		goto cleanup;
+
+	err = bpf_prog_test_run_opts(bpf_program__fd(skel->progs.fentry_1), &topts);
+	if (!ASSERT_OK(err, "test_run"))
+		goto cleanup;
+	ASSERT_EQ(skel->bss->test_result_fentry_1, 1, "multi_fentry");
+	ASSERT_EQ(skel->bss->test_result_fentry, 1, "fentry");
+
+	err = bpf_link__destroy(fentry_link);
+	fentry_link = NULL;
+	if (!ASSERT_OK(err, "destroy_fentry"))
+		goto cleanup;
+
+	err = bpf_prog_test_run_opts(bpf_program__fd(skel->progs.fentry_1), &topts);
+	if (!ASSERT_OK(err, "test_run_multi"))
+		goto cleanup;
+	ASSERT_EQ(skel->bss->test_result_fentry_1, 2, "multi_fentry_only");
+	ASSERT_EQ(skel->bss->test_result_fentry, 1, "fentry_detached");
+
+	err = bpf_link__destroy(multi_link);
+	multi_link = NULL;
+	if (!ASSERT_OK(err, "destroy_multi"))
+		goto cleanup;
+
+	err = bpf_prog_test_run_opts(bpf_program__fd(skel->progs.fentry_1), &topts);
+	if (!ASSERT_OK(err, "test_run_detached"))
+		goto cleanup;
+	ASSERT_EQ(skel->bss->test_result_fentry_1, 2, "multi_fentry_detached");
+	ASSERT_EQ(skel->bss->test_result_fentry, 1, "fentry_still_detached");
+
+cleanup:
+	bpf_link__destroy(fentry_link);
+	bpf_link__destroy(multi_link);
+	free(ids);
+	tracing_multi_intersect__destroy(skel);
+}
+
 static void test_session(void)
 {
 	LIBBPF_OPTS(bpf_test_run_opts, topts);
@@ -957,4 +1024,6 @@ void test_tracing_multi_test(void)
 	if (test__start_subtest("attach_api_fails"))
 		test_attach_api_fails();
 	RUN_TESTS(tracing_multi_verifier);
+	if (test__start_subtest("fentry_after_multi"))
+		test_fentry_after_multi();
 }
tools/testing/selftests/bpf/progs/tracing_multi_intersect_attach.c +8 -0
index cd5be0bb6ffd..5ebbe3b602a4 100644
@@ -11,6 +11,14 @@ __u64 test_result_fentry_1 = 0;
 __u64 test_result_fentry_2 = 0;
 __u64 test_result_fexit_1 = 0;
 __u64 test_result_fexit_2 = 0;
+__u64 test_result_fentry = 0;
+
+SEC("fentry/bpf_fentry_test1")
+int BPF_PROG(fentry, int a)
+{
+	test_result_fentry += a == 1;
+	return 0;
+}
 
 SEC("fentry.multi")
 int BPF_PROG(fentry_1)
Signature · 1 line
2.55.0
#5 lore raw

Re: [PATCH bpf-next 2/2] selftests/bpf: Test fentry link after tracing_multi link

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider: - [Medium] Test Flakiness due to Missing PID Filtering in BPF `fentry` Program --

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Test Flakiness due to Missing PID Filtering in BPF fentry Program --

--- Patch [2]: [PATCH bpf-next 2/2] selftests/bpf: Test fentry link after tracing_multi link --- commit 7dc95a7a3d3a0ca96e45e5ad2ebebe5157ae17ed
Author: Leon Hwang <[email protected]>

selftests/bpf: Test fentry link after tracing_multi link

This commit adds a selftest to verify that an fentry link can be attached safely after a tracing_multi link. It introduces a new fentry BPF program and a corresponding user-space test case.

Quoted 4 lines
tools/testing/selftests/bpf/progs/tracing_multi_intersect_attach.c +0 -0
index cd5be0bb6ffdd..5ebbe3b602a4f 100644

[ ... ]

tools/testing/selftests/bpf/progs/tracing_multi_intersect_attach.c __u64 test_result_fentry_1 = 0; · line 19

Quoted 10 lines
tools/testing/selftests/bpf/progs/tracing_multi_intersect_attach.c +6 -0
@@ -11,6 +11,14 @@ __u64 test_result_fentry_1 = 0;
 __u64 test_result_fentry_2 = 0;
 __u64 test_result_fexit_1 = 0;
 __u64 test_result_fexit_2 = 0;
+__u64 test_result_fentry = 0;
+
+SEC("fentry/bpf_fentry_test1")
+int BPF_PROG(fentry, int a)
+{
+	test_result_fentry += a == 1;

[Severity: Medium]
Can this cause test flakiness during parallel selftest runs?

Because bpf_fentry_test1 is a global kernel function widely used across various BPF selftests, parallel test executions might trigger this program. Since the counter is incremented blindly without filtering by the executing process PID, could parallel executions increase test_result_fentry above 1 and cause the userland assertion ASSERT_EQ(skel->bss->test_result_fentry, 1) to sporadically fail?

Quoted 5 lines
tools/testing/selftests/bpf/progs/tracing_multi_intersect_attach.c +2 -0
+	return 0;
+}
 
 SEC("fentry.multi")
 int BPF_PROG(fentry_1)
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=2
author
#6 lore raw

Re: [PATCH bpf-next 2/2] selftests/bpf: Test fentry link after tracing_multi link

Quoted 35 lines · On 2026/7/10 23:04, [email protected] wrote:

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Test Flakiness due to Missing PID Filtering in BPF fentry Program --

--- Patch [2]: [PATCH bpf-next 2/2] selftests/bpf: Test fentry link after tracing_multi link --- commit 7dc95a7a3d3a0ca96e45e5ad2ebebe5157ae17ed
Author: Leon Hwang <[email protected]>

selftests/bpf: Test fentry link after tracing_multi link

This commit adds a selftest to verify that an fentry link can be attached safely after a tracing_multi link. It introduces a new fentry BPF program and a corresponding user-space test case.

Quoted 4 lines
tools/testing/selftests/bpf/progs/tracing_multi_intersect_attach.c +0 -0
index cd5be0bb6ffdd..5ebbe3b602a4f 100644

[ ... ]

tools/testing/selftests/bpf/progs/tracing_multi_intersect_attach.c __u64 test_result_fentry_1 = 0; · line 19

Quoted 10 lines
tools/testing/selftests/bpf/progs/tracing_multi_intersect_attach.c +6 -0
@@ -11,6 +11,14 @@ __u64 test_result_fentry_1 = 0;
 __u64 test_result_fentry_2 = 0;
 __u64 test_result_fexit_1 = 0;
 __u64 test_result_fexit_2 = 0;
+__u64 test_result_fentry = 0;
+
+SEC("fentry/bpf_fentry_test1")
+int BPF_PROG(fentry, int a)
+{
+	test_result_fentry += a == 1;

[Severity: Medium]
Can this cause test flakiness during parallel selftest runs?

Quoted 7 lines

Because bpf_fentry_test1 is a global kernel function widely used across various BPF selftests, parallel test executions might trigger this program. Since the counter is incremented blindly without filtering by the executing process PID, could parallel executions increase test_result_fentry above 1 and cause the userland assertion ASSERT_EQ(skel->bss->test_result_fentry, 1) to sporadically fail?

Correct.

I should use tracing_multi_arg_check() here, which would filter the PID.

Thanks,
Leon

Quoted 6 lines
Quoted 5 lines
tools/testing/selftests/bpf/progs/tracing_multi_intersect_attach.c +2 -0
+	return 0;
+}
 
 SEC("fentry.multi")
 int BPF_PROG(fentry_1)