Re: [PATCHv2 05/11] libbpf: Detect uprobe syscall with new error

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
- [High] Missing kernel dependency causes immediate selftest failure and libbpf feature detection breakage.
- [High] libbpf feature detection breaks backward compatibility by not accepting the old ENXIO error code.
--

commit 674b83913f8f61b2f16ba8f8b86e6bf8e84da5a3
Author: Jiri Olsa <jolsa@kernel.org>

libbpf: Detect uprobe syscall with new error

This commit updates libbpf's feature detection and the uprobe selftests to
expect the uprobe syscall to return -EPROTO instead of -ENXIO. This is in
response to an optimized uprobe fix that changed the syscall's error code.

> diff --git a/tools/lib/bpf/features.c b/tools/lib/bpf/features.c
> index 4f19a0d79b0cf..a37aaea5253e5 100644
> --- a/tools/lib/bpf/features.c
> +++ b/tools/lib/bpf/features.c
> @@ -577,10 +577,10 @@ static int probe_ldimm64_full_range_off(int token_fd)
>  static int probe_uprobe_syscall(int token_fd)
>  {
>  	/*
> -	 * If kernel supports uprobe() syscall, it will return -ENXIO when called
> +	 * If kernel supports uprobe() syscall, it will return -EPROTO when called
>  	 * from the outside of a kernel-generated uprobe trampoline.
>  	 */
> -	return syscall(__NR_uprobe) < 0 && errno == ENXIO;
> +	return syscall(__NR_uprobe) < 0 && errno == EPROTO;

Is it possible this change breaks libbpf's backward compatibility with older
kernels? Since probe_uprobe_syscall() detects uprobe syscall support on
various kernel versions, replacing the check rather than allowing either
(e.g., errno == ENXIO || errno == EPROTO) might cause libbpf to silently
fail to detect uprobe support on kernels that predate the EPROTO change.

Also, looking at arch/x86/kernel/uprobes.c, the kernel-side change does
not appear to be present in the tree yet:

SYSCALL_DEFINE0(uprobe)
{
	...
	/* Allow execution only from uprobe trampolines. */
	if (!in_uprobe_trampoline(regs->ip))
		return -ENXIO;

Will replacing this check cause immediate regressions for both feature
detection and the selftests until the kernel dependency is merged?

[ ... ]

> diff --git a/tools/testing/selftests/bpf/prog_tests/uprobe_syscall.c b/tools/testing/selftests/bpf/prog_tests/uprobe_syscall.c
> index 955a37751b52d..c944136252c6d 100644
> --- a/tools/testing/selftests/bpf/prog_tests/uprobe_syscall.c
> +++ b/tools/testing/selftests/bpf/prog_tests/uprobe_syscall.c
> @@ -762,7 +762,7 @@ static void test_uprobe_error(void)
>  	long err = syscall(__NR_uprobe);
>  
>  	ASSERT_EQ(err, -1, "error");
> -	ASSERT_EQ(errno, ENXIO, "errno");
> +	ASSERT_EQ(errno, EPROTO, "errno");

Similarly, does this selftest change introduce an immediate regression if
the running kernel still returns -ENXIO?

>  }

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260518105957.123445-6-jolsa@kernel.org?part=1