Re: [PATCHv3 05/12] libbpf: Change has_nop_combo to work on top of nop10

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Replacing the 5-byte nop check entirely with a 10-byte nop check silently degrades tracing performance for existing compiled binaries and causes BPF selftest failures.
--

commit 9ee5f26b2092c13d658ca6962b0ca9bbe4cfe44a
Author: Jiri Olsa <jolsa@kernel.org>

libbpf: Change has_nop_combo to work on top of nop10

This commit modifies the has_nop_combo() function to detect an 11-byte nop
sequence rather than a 6-byte nop sequence to optimize uprobe handling.

> diff --git a/tools/lib/bpf/usdt.c b/tools/lib/bpf/usdt.c
> index e3710933fd52a..484a4354e82ba 100644
> --- a/tools/lib/bpf/usdt.c
> +++ b/tools/lib/bpf/usdt.c
[ ... ]
> @@ -596,14 +596,14 @@ static int parse_usdt_spec(struct usdt_spec *spec, const struct usdt_note *note,
>  #if defined(__x86_64__)
>  static bool has_nop_combo(int fd, long off)
>  {
> -	unsigned char nop_combo[6] = {
> -		0x90, 0x0f, 0x1f, 0x44, 0x00, 0x00 /* nop,nop5 */
> +	unsigned char nop_combo[11] = {
> +		0x90, 0x66, 0x2e, 0x0f, 0x1f, 0x84, 0x00, 0x00, 0x00, 0x00, 0x00,
>  	};

[Severity: Medium]
Will this drop support for older compiled binaries and cause BPF selftests
to fail?

By completely replacing the 5-byte nop check rather than supporting both, it
looks like legacy binaries will silently fall back to unoptimized uprobes on
the first byte, which could degrade tracing performance.

Additionally, tools/testing/selftests/bpf/usdt.h still seems to define
USDT_NOP using the 6-byte sequence (0x90, 0x0f, 0x1f, 0x44, 0x00, 0x00).
Could this cause subtest_optimized_attach() in prog_tests/usdt.c to fail its
assertions, since the +1 offset for the nop5 instruction will no longer be
applied during testing?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260521124411.31133-6-jolsa@kernel.org?part=1