On 9/24/26 07:00, Sarthak Sharma wrote:
> The userfaultfd syscall can return ENOSYS even when __NR_userfaultfd is
> defined by the architecture headers, for example when the kernel is
> built without CONFIG_USERFAULTFD. The test currently skips for EPERM
> when the caller lacks permission to use userfaultfd but treats
> ENOSYS as a failure.
>
> Handle ENOSYS like EPERM and skip the mremap_move_multi_invalid_vmas()
> test when userfaultfd syscall is not implemented.
>
> Fixes: 742d3663a577 ("selftests/mm: add test for invalid multi VMA
> operations")
> Signed-off-by: Sarthak Sharma <[email protected]>
> ---
> tools/testing/selftests/mm/mremap_test.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/tools/testing/selftests/mm/mremap_test.c
> b/tools/testing/selftests/mm/mremap_test.c
> index 5532d3151ed0..a5813a5e8f5b 100644
> --- a/tools/testing/selftests/mm/mremap_test.c
> +++ b/tools/testing/selftests/mm/mremap_test.c
> @@ -721,7 +721,7 @@ static void mremap_move_multi_invalid_vmas(FILE *maps_fp,
> if (uffd == -1) {
> err = errno;
> ksft_perror("userfaultfd");
Should we rather move that below the if? Doesn't quite make sense to print the
error if we are skipping and printing a proper message.
Maybe that should go into the previous patch?
> - if (err == EPERM) {
> + if (err == EPERM || err == ENOSYS) {
> ksft_test_result_skip("%s - missing uffd\n", test_name);
> return;
> }
Apart from that LGTM.
Acked-by: David Hildenbrand (Arm) <[email protected]>
--
Cheers,
David
