Re: [PATCHv2 06/11] selftests/bpf: Emit nop,nop10 instructions combo for x86_64 arch

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
- [Medium] The patch modifies `USDT_NOP` to emit an 11-byte `nop, nop10` sequence, breaking existing BPF selftests that explicitly expect `nop, nop5`.
- [Medium] Libbpf is not updated to recognize the new `nop, nop10` sequence, causing the intended uprobe optimization to silently fail.
--

commit 1ccfa6afa421e13f88c377be82ec249ff187bf51
Author: Jiri Olsa <jolsa@kernel.org>

selftests/bpf: Emit nop,nop10 instructions combo for x86_64 arch

This commit updates the USDT_NOP macro in the BPF selftests to use a nop,nop10
instruction sequence instead of nop,nop5 for x86_64. This is intended to
support nop10 optimization testing.

> diff --git a/tools/testing/selftests/bpf/usdt.h b/tools/testing/selftests/bpf/usdt.h
> index c71e21df38b32..d359663b9c329 100644
> --- a/tools/testing/selftests/bpf/usdt.h
> +++ b/tools/testing/selftests/bpf/usdt.h
> @@ -313,7 +313,7 @@ struct usdt_sema { volatile unsigned short active; };
>  #if defined(__ia64__) || defined(__s390__) || defined(__s390x__)
>  #define USDT_NOP			nop 0
>  #elif defined(__x86_64__)
> -#define USDT_NOP                       .byte 0x90, 0x0f, 0x1f, 0x44, 0x00, 0x0 /* nop, nop5 */
> +#define USDT_NOP                       .byte 0x90, 0x66, 0x66, 0x0f, 0x1f, 0x84, 0x00, 0x00, 0x00, 0x00, 0x00 /* nop, nop10 */

Will this change break the existing BPF selftests that explicitly expect the
nop,nop5 sequence?

Looking at tools/testing/selftests/bpf/prog_tests/usdt.c,
subtest_optimized_attach() uses find_instr() to search for a hardcoded 6-byte
sequence:

    const char nop1_nop5_combo[] = { 0x90, 0x0f, 0x1f, 0x44, 0x00, 0x00 };
    ...
    addr_2 = find_instr(usdt_2, nop1_nop5_combo, sizeof(nop1_nop5_combo));

Because usdt_2.c is compiled with the new 11-byte nop,nop10 sequence,
find_instr() will return NULL and cause the ASSERT_OK_PTR() check to fail.

Additionally, does libbpf need to be updated to recognize this new sequence?

In tools/lib/bpf/usdt.c, has_nop_combo() appears to be hardcoded to check only
for the 6-byte nop,nop5 sequence.

If libbpf doesn't recognize the new 11-byte sequence, it won't increment the
instruction pointer to skip the initial 1-byte nop. This would cause the
uprobe to attach to the unoptimized 1-byte nop instead of utilizing the
intended optimization.

>  #else
>  #define USDT_NOP			nop
>  #endif

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260518105957.123445-7-jolsa@kernel.org?part=1