Re: [PATCH bpf-next v2 06/11] selftests/bpf: Test trampoline args around a 16-byte aligned stack arg

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Low] bpf: endianness assumption causes selftest failure on aarch64 big-endian
--

commit df1efd6b9c0e934bad453d07e6efe674bda8e1e6
Author: Yonghong Song <yonghong.song@linux.dev>

selftests/bpf: Test trampoline args around a 16-byte aligned stack arg

This commit adds tests for bpf_testmod functions taking a 16-byte aligned
struct containing an __int128. It checks that the BPF trampoline packs
these arguments correctly according to the BPF calling convention.

> diff --git a/tools/testing/selftests/bpf/prog_tests/tracing_struct.c b/tools/testing/selftests/bpf/prog_tests/tracing_struct.c
> index 8086586d3e196..a2da219e14769 100644
> --- a/tools/testing/selftests/bpf/prog_tests/tracing_struct.c
> +++ b/tools/testing/selftests/bpf/prog_tests/tracing_struct.c
[ ... ]
> @@ -141,6 +143,20 @@ static void test_int128_args(void)
>  	ASSERT_EQ(skel->bss->t_c, 3, "t:c");
>  	ASSERT_EQ(skel->bss->t_ret, 6, "t ret");
>  
> +	ASSERT_EQ(skel->bss->s_g, 7, "s:g");
> +	ASSERT_EQ(skel->bss->s_h_lo, 8, "s:h.lo");
> +	ASSERT_EQ(skel->bss->s_h_hi, 9, "s:h.hi");

[Severity: Low]
Does this code assume a little-endian memory layout for the __int128
argument?

The test passes a 16-byte struct containing ((__int128)9 << 64) | 8.
On big-endian architectures like aarch64 (when configured with
CONFIG_CPU_BIG_ENDIAN), the most significant 64 bits (9) are placed at the
lowest address, and the least significant bits (8) at the higher address.
This hardcoded expectation for the little-endian layout appears to cause the
test_int128_args selftest to fail its assertions on big-endian systems.

> +	ASSERT_EQ(skel->bss->s_i, 10, "s:i");
> +	ASSERT_EQ(skel->bss->s_ret, 385, "s ret");
[ ... ]
> diff --git a/tools/testing/selftests/bpf/progs/tracing_struct_int128.c b/tools/testing/selftests/bpf/progs/tracing_struct_int128.c
> index 4638dfec1f38e..5520ee500c01d 100644
> --- a/tools/testing/selftests/bpf/progs/tracing_struct_int128.c
> +++ b/tools/testing/selftests/bpf/progs/tracing_struct_int128.c
[ ... ]
> @@ -15,4 +15,49 @@ int test_int128_arg_fexit(unsigned long long *ctx)
>  	return 0;
>  }
>  
> +/*
> + * The arguments come packed in the BPF convention whatever holes the
> + * kernel convention left between them: an eightbyte each, and two for the
> + * struct holding an __int128.
> + */
> +long s_g, s_h_lo, s_h_hi, s_i, s_ret;
> +
> +SEC("fentry/bpf_testmod_test_int128_stack")
> +int test_int128_stack_fentry(unsigned long long *ctx)
> +{
> +	s_g = ctx[6];
> +	s_h_lo = ctx[7];
> +	s_h_hi = ctx[8];

[Severity: Low]
Is this direct array indexing in test_int128_stack_fentry() subject to the
same endianness assumption?

When the BPF trampoline projects the struct onto the u64 ctx[] array on a
big-endian system, ctx[7] receives 9 and ctx[8] receives 8, rather than the
little-endian projection expected here.

> +	s_i = ctx[9];
> +	return 0;
> +}

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261013003429.1853052-1-yonghong.song@linux.dev?part=6