Re: [PATCH bpf-next v2 0/5] bpf: add bpf_memcmp and bpf_strpbrk kfuncs

["Alexei Starovoitov" <alexei.starovoitov@gmail.com>]

On Wed Jun 24, 2026 at 2:36 AM PDT, Longjun Tang wrote:
> From: Longjun Tang <tanglongjun@kylinos.cn>
>
> This series adds two new BPF kfuncs to kernel/bpf/helpers.c.
>
> === bpf_memcmp() ===
>
> BPF programs frequently need to compare raw binary buffers, such as:
> - MAC addresses (6 bytes) in XDP/TC for L2 filtering and forwarding;
> - UUIDs (16 bytes) and hash digests for allowlist/denylist policies;
>
> Currently there is no safe way to do this — bpf_strncmp (helper) stops
> at null bytes which corrupts binary data, and hand-rolled loops are
> fragile with the verifier (as evidenced by the local_memcmp() workaround
> in selftests/bpf/progs/dynptr_success.c). __builtin_memcmp may work in
> some cases but is not verifier-safe for arbitrary kernel memory.

Hand rolled loops were fragile in the past, but not any more.
With bpf_for() and may_goto/can_loop the loops can be large.
For MAC and UUIDs even old verifiers can deal just with with "bounded loop".

Sorry, but I don't see a human driven motivation in this patch.
Looks like AI flagged something and AI sloped a patch in response.

pw-bot: cr