Re: [PATCH bpf-next v2 09/11] selftests/bpf: Test trampolines on more BPF functions with stack args

[Yonghong Song <yonghong.song@linux.dev>]

On 10/13/26 3:28 AM, bot+bpf-ci@kernel.org wrote:
>> selftests/bpf: Test trampolines on more BPF functions with stack args
> TL;DR: Nit only: the callee_split() comment in the selftest says h is split
> between register and stack, but that is x86-64 specific; on arm64 nothing is
> split. Suggests saying so.
>
>> diff --git a/tools/testing/selftests/bpf/progs/tracing_struct_int128_tgt.c b/tools/testing/selftests/bpf/progs/tracing_struct_int128_tgt.c
>> --- a/tools/testing/selftests/bpf/progs/tracing_struct_int128_tgt.c
>> +++ b/tools/testing/selftests/bpf/progs/tracing_struct_int128_tgt.c
> [ ... ]
>
>> @@ -44,6 +44,25 @@ static __noinline __u64 callee_stack(__u64 a, __u64 b, __u64 c, __u64 d,
>>   	       i * 10;
>>   }
>>
>> +/* BPF code splits h between the last argument register and the stack. */
>> +static __noinline __u64 callee_split(__u64 a, __u64 b, __u64 c, __u64 d,
>> +				     __u64 e, struct int128_arg h, __u64 i)
>> +{
>> +	return a + b * 2 + c * 3 + d * 4 + e * 5 + (__u64)h.v * 6 +
>> +	       (__u64)((unsigned __int128)h.v >> 64) * 7 + i * 8;
>> +}
> This isn't a bug, but is this comment accurate for every architecture this
> test runs on?
>
> test_int128_tgt_prog() is built for x86-64 and for arm64, and a..e already
> occupy R1-R5 in BPF terms, so h and i are each stack arguments as far as
> the BPF ISA is concerned.
>
> The split between a register and the stack only comes from the x86-64 JIT,
> which maps the first stack argument slot to r9:
>
> arch/x86/net/bpf_jit_comp.c:x86_arg_reg[] = { BPF_REG_1, ..., BPF_REG_5,
> X86_REG_R9 }
>
> On arm64 the first three stack argument slots are mapped to x5-x7, so h
> would land in x5/x6 and i in x7, with nothing split.
>
> The commit message already says "on x86-64 it is passed split between r9
> and the stack", and the comment on the neighbouring callee_stack() names
> x86-64 and arm64 separately.
>
> Could this comment say that the split is x86-64 specific, for example "x86-64
> BPF code passes h in r9 and its first stack slot"?

You are right. The comment is not good. Will fix.

>
>
> ---
> AI reviewed your patch. Please fix the bug or email reply why it's not a bug.
> See: https://github.com/kernel-patches/vmtest/blob/master/ci/claude/README.md
>
> CI run summary: https://github.com/kernel-patches/bpf/actions/runs/38329598343