[PATCH bpf-next v2 0/2] selftests/bpf: Handle some unexpected failures

[PATCH bpf-next v2 0/2] selftests/bpf: Handle some unexpected failures

[Tiezhu Yang]

v2:
  - Use libbpf_probe_bpf_map_type() in prog_tests instead of
     handling EOPNOTSUPP globally in test_loader.
  - Fix unaligned access at the source by applying a bitmask
     'off &= ~1' instead of skipping these tests entirely.
  - Modify the patch subjects and update the commit messages.

Tiezhu Yang (2):
  selftests/bpf: Skip jit_insns_arena tests if not supported
  selftests/bpf: Fix unaligned access failure in fexit_bpf2bpf

 tools/testing/selftests/bpf/prog_tests/jit_insns.c  | 9 ++++++++-
 tools/testing/selftests/bpf/progs/fexit_bpf2bpf.c   | 1 +
 tools/testing/selftests/bpf/progs/test_pkt_access.c | 1 +
 3 files changed, 10 insertions(+), 1 deletion(-)

-- 
2.42.0

[PATCH bpf-next v2 1/2] selftests/bpf: Skip jit_insns_arena tests if not supported

[Tiezhu Yang]

When loading a BPF skeleton program via test_loader, certain test cases
such as jit_insns/cast_to_user may fail if the underlying hardware lacks
specific execution extensions required by newer features like BPF Arena.

Use libbpf_probe_bpf_map_type() in prog_tests/jit_insns.c to explicitly
check for BPF_MAP_TYPE_ARENA support before loading the jit_insns_arena
test skeleton. If the JIT backend or hardware does not support Arena,
cleanly report the subtest status as SKIPPED via test__skip().

This prevents false-alarm test failures on unsupported hardware while
preserving the core verification coverage for the rest of the standard
JIT instruction tests (such as alu_imm, alu_reg).

Before:
  $ sudo ./test_progs -a jit_insns/cast_to_user
  ...
  libbpf: map 'arena': failed to create: -EOPNOTSUPP
  run_subtest:FAIL:unexpected_load_failure unexpected error: -95 (errno 95)
  ...
  Summary: 0/0 PASSED, 0 SKIPPED, 1/1 FAILED

After:
  $ sudo ./test_progs -a jit_insns/cast_to_user
  ...
  Summary: 1/0 PASSED, 1 SKIPPED, 0/0 FAILED

Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
---
 tools/testing/selftests/bpf/prog_tests/jit_insns.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/tools/testing/selftests/bpf/prog_tests/jit_insns.c b/tools/testing/selftests/bpf/prog_tests/jit_insns.c
index 0aaab23a2eef..e8a0e5ec2d6b 100644
--- a/tools/testing/selftests/bpf/prog_tests/jit_insns.c
+++ b/tools/testing/selftests/bpf/prog_tests/jit_insns.c
@@ -3688,7 +3688,14 @@ static void test_tail_call(void)
 
 void test_jit_insns(void)
 {
-	RUN_TESTS(jit_insns_arena);
+	if (libbpf_probe_bpf_map_type(BPF_MAP_TYPE_ARENA, NULL)) {
+		RUN_TESTS(jit_insns_arena);
+	} else {
+		if (test__start_subtest("arena_zero_off"))
+			test__skip();
+		if (test__start_subtest("cast_to_user"))
+			test__skip();
+	}
 	if (test__start_subtest("alu_imm"))
 		test_alu_imm();
 	if (test__start_subtest("alu_reg"))
-- 
2.42.0

[PATCH bpf-next v2 2/2] selftests/bpf: Fix unaligned access failure in fexit_bpf2bpf

[Tiezhu Yang]

When executing "./test_progs -t fexit_bpf2bpf", there are 5 subtest
failures on platforms where CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
is not set, because these platforms lack hardware unaligned access
support:

  ; tcp->check++; @ test_pkt_access.c:98
  226: (69) r1 = *(u16 *)(r2 +16)
  misaligned packet access off 2+(0x0; 0x3f)+16 size 2
  ...
  libbpf: prog 'test_pkt_access': failed to load: -EACCES
  libbpf: failed to load object './test_pkt_access.bpf.o'

Instead of skipping these tests entirely, fix this at the source by
applying a bitmask 'off &= ~1' right after the initial range check.
This explicitly communicates the 2-byte alignment state to pass the
verifier and run the full fexit_bpf2bpf test matrix.

Before:
  $ sudo ./test_progs -t fexit_bpf2bpf | tail -1
  Summary: 0/10 PASSED, 1 SKIPPED, 1/5 FAILED

After:
  $ sudo ./test_progs -t fexit_bpf2bpf | tail -1
  Summary: 1/15 PASSED, 1 SKIPPED, 0/0 FAILED

Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
---
 tools/testing/selftests/bpf/progs/fexit_bpf2bpf.c   | 1 +
 tools/testing/selftests/bpf/progs/test_pkt_access.c | 1 +
 2 files changed, 2 insertions(+)

diff --git a/tools/testing/selftests/bpf/progs/fexit_bpf2bpf.c b/tools/testing/selftests/bpf/progs/fexit_bpf2bpf.c
index f4bbf87b82dd..ed06345825e2 100644
--- a/tools/testing/selftests/bpf/progs/fexit_bpf2bpf.c
+++ b/tools/testing/selftests/bpf/progs/fexit_bpf2bpf.c
@@ -173,6 +173,7 @@ int new_test_pkt_write_access_subprog(struct __sk_buff *skb, __u32 off)
 	if (off > sizeof(struct ethhdr) + sizeof(struct ipv6hdr))
 		return -1;
 
+	off &= ~1;
 	tcp = data + off;
 	if (tcp + 1 > data_end)
 		return -1;
diff --git a/tools/testing/selftests/bpf/progs/test_pkt_access.c b/tools/testing/selftests/bpf/progs/test_pkt_access.c
index bce7173152c6..21f28a762d60 100644
--- a/tools/testing/selftests/bpf/progs/test_pkt_access.c
+++ b/tools/testing/selftests/bpf/progs/test_pkt_access.c
@@ -91,6 +91,7 @@ int test_pkt_write_access_subprog(struct __sk_buff *skb, __u32 off)
 	if (off > sizeof(struct ethhdr) + sizeof(struct ipv6hdr))
 		return -1;
 
+	off &= ~1;
 	tcp = data + off;
 	if (tcp + 1 > data_end)
 		return -1;
-- 
2.42.0

Re: [PATCH bpf-next v2 2/2] selftests/bpf: Fix unaligned access failure in fexit_bpf2bpf

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 1 potential
issue to consider:
- [Low] selftests/bpf: missing Fixes trailer for commit 6dc03dc71387
--

commit 507288d3c1afad0bf83be8f5f5a571a1da547ccc
Author: Tiezhu Yang <yangtiezhu@loongson.cn>

> selftests/bpf: Fix unaligned access failure in fexit_bpf2bpf
>
> When executing "./test_progs -t fexit_bpf2bpf", there are 5 subtest
> failures on platforms where CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
> is not set, because these platforms lack hardware unaligned access
> support:
>
>   ; tcp->check++; @ test_pkt_access.c:98
>   226: (69) r1 = *(u16 *)(r2 +16)
>   misaligned packet access off 2+(0x0; 0x3f)+16 size 2
>   ...
>   libbpf: prog 'test_pkt_access': failed to load: -EACCES
>   libbpf: failed to load object './test_pkt_access.bpf.o'
>
> Instead of skipping these tests entirely, fix this at the source by
> applying a bitmask 'off &= ~1' right after the initial range check.
> This explicitly communicates the 2-byte alignment state to pass the
> verifier and run the full fexit_bpf2bpf test matrix.
>
> Before:
>   $ sudo ./test_progs -t fexit_bpf2bpf | tail -1
>   Summary: 0/10 PASSED, 1 SKIPPED, 1/5 FAILED
>
> After:
>   $ sudo ./test_progs -t fexit_bpf2bpf | tail -1
>   Summary: 1/15 PASSED, 1 SKIPPED, 0/0 FAILED
>
> Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>

[Severity: Low]
This fixes the unaligned packet offset in test_pkt_write_access_subprog()
and new_test_pkt_write_access_subprog().  Should this include:

  Fixes: 6dc03dc71387 ("selftests/bpf: Add test for freplace program with write access")

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261109041207.15221-1-yangtiezhu@loongson.cn?part=2

Re: [PATCH bpf-next v2 1/2] selftests/bpf: Skip jit_insns_arena tests if not supported

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 1 potential
issue to consider:
- [Low] selftests/bpf: incomplete jit_insns_arena skip list in
  test_jit_insns()
--

commit 5301df7df06ac60c33f099fbdfa02a082cfa0c99
Author: Tiezhu Yang <yangtiezhu@loongson.cn>

selftests/bpf: Skip jit_insns_arena tests if not supported

This commit probes for BPF_MAP_TYPE_ARENA support in test_jit_insns() and
skips the jit_insns_arena subtests when arena maps are not supported.

> diff --git a/tools/testing/selftests/bpf/prog_tests/jit_insns.c b/tools/testing/selftests/bpf/prog_tests/jit_insns.c
> index 0aaab23a2eef8..e8a0e5ec2d6bf 100644
> --- a/tools/testing/selftests/bpf/prog_tests/jit_insns.c
> +++ b/tools/testing/selftests/bpf/prog_tests/jit_insns.c
> @@ -3688,7 +3688,14 @@ static void test_tail_call(void)
>  
>  void test_jit_insns(void)
>  {
> -	RUN_TESTS(jit_insns_arena);
> +	if (libbpf_probe_bpf_map_type(BPF_MAP_TYPE_ARENA, NULL)) {
> +		RUN_TESTS(jit_insns_arena);
> +	} else {
> +		if (test__start_subtest("arena_zero_off"))
> +			test__skip();
> +		if (test__start_subtest("cast_to_user"))
> +			test__skip();
> +	}

[Severity: Low]
Does this else branch miss the remaining subtests in jit_insns_arena?

Commit a9fee553d3db ("selftests/bpf: Extend the JIT instruction tests for
arm64 encodings") added arena_offsets (built on all architectures) as well as
arena_offsets_ldsx, arena_offsets_atomics, and arena_acq_rel to
tools/testing/selftests/bpf/progs/jit_insns_arena.c.  When BPF_MAP_TYPE_ARENA
is unsupported, those subtests are silently omitted rather than registered and
marked skipped, so targeted runs such as ./test_progs -a
jit_insns/arena_offsets match zero subtests.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261109041207.15221-1-yangtiezhu@loongson.cn?part=1

Re: [PATCH bpf-next v2 1/2] selftests/bpf: Skip jit_insns_arena tests if not supported

[Tiezhu Yang]

On 11/9/26 12:25 PM, sashiko-bot@kernel.org wrote:
> Thank you for your contribution! Sashiko AI review found 1 potential
> issue to consider:
> - [Low] selftests/bpf: incomplete jit_insns_arena skip list in
>    test_jit_insns()

...

> [Severity: Low]
> Does this else branch miss the remaining subtests in jit_insns_arena?
> 
> Commit a9fee553d3db ("selftests/bpf: Extend the JIT instruction tests for
> arm64 encodings") added arena_offsets (built on all architectures) as well as
> arena_offsets_ldsx, arena_offsets_atomics, and arena_acq_rel to
> tools/testing/selftests/bpf/progs/jit_insns_arena.c.  When BPF_MAP_TYPE_ARENA
> is unsupported, those subtests are silently omitted rather than registered and
> marked skipped, so targeted runs such as ./test_progs -a
> jit_insns/arena_offsets match zero subtests.

I will introduce test_loader__skip_all_subtests() helper as a generic
utility to avoid hardcoded string arrays in test runners.

Thanks,
Tiezhu

Re: [PATCH bpf-next v2 2/2] selftests/bpf: Fix unaligned access failure in fexit_bpf2bpf

[Tiezhu Yang]

On 11/9/26 12:17 PM, sashiko-bot@kernel.org wrote:
> Thank you for your contribution! Sashiko AI review found 1 potential
> issue to consider:
> - [Low] selftests/bpf: missing Fixes trailer for commit 6dc03dc71387

...

> [Severity: Low]
> This fixes the unaligned packet offset in test_pkt_write_access_subprog()
> and new_test_pkt_write_access_subprog().  Should this include:
> 
>    Fixes: 6dc03dc71387 ("selftests/bpf: Add test for freplace program with write access")

I will add Fixes tag in v3.

Thanks,
Tiezhu