Sorry, I screwed up when sending it out. It's not [1/2], it's supposed
to be a stand-alone patch. IOW, don't expect a [2/2] followup :)
On Fri, Dec 29, 2023 at 6:56 PM Suren Baghdasaryan <surenb@xxxxxxxxxx> wrote:
>
> Add a test for UFFDIO_MOVE ioctl operating on a hugepage which has to
> be split because destination is marked with MADV_NOHUGEPAGE. With this
> we cover all 3 cases: normal page move, hugepage move, hugepage splitting
> before move.
>
> Signed-off-by: Suren Baghdasaryan <surenb@xxxxxxxxxx>
> ---
> Patch applies over mm-unstable.
>
> tools/testing/selftests/mm/uffd-unit-tests.c | 17 +++++++++++++++++
> 1 file changed, 17 insertions(+)
>
> diff --git a/tools/testing/selftests/mm/uffd-unit-tests.c b/tools/testing/selftests/mm/uffd-unit-tests.c
> index d8091523c2df..cce90a10515a 100644
> --- a/tools/testing/selftests/mm/uffd-unit-tests.c
> +++ b/tools/testing/selftests/mm/uffd-unit-tests.c
> @@ -1199,6 +1199,16 @@ static void uffd_move_test(uffd_test_args_t *targs)
>
> static void uffd_move_pmd_test(uffd_test_args_t *targs)
> {
> + if (madvise(area_dst, nr_pages * page_size, MADV_HUGEPAGE))
> + err("madvise(MADV_HUGEPAGE) failure");
> + uffd_move_test_common(targs, read_pmd_pagesize(),
> + uffd_move_pmd_handle_fault);
> +}
> +
> +static void uffd_move_pmd_split_test(uffd_test_args_t *targs)
> +{
> + if (madvise(area_dst, nr_pages * page_size, MADV_NOHUGEPAGE))
> + err("madvise(MADV_NOHUGEPAGE) failure");
> uffd_move_test_common(targs, read_pmd_pagesize(),
> uffd_move_pmd_handle_fault);
> }
> @@ -1330,6 +1340,13 @@ uffd_test_case_t uffd_tests[] = {
> .uffd_feature_required = UFFD_FEATURE_MOVE,
> .test_case_ops = &uffd_move_test_pmd_case_ops,
> },
> + {
> + .name = "move-pmd-split",
> + .uffd_fn = uffd_move_pmd_split_test,
> + .mem_targets = MEM_ANON,
> + .uffd_feature_required = UFFD_FEATURE_MOVE,
> + .test_case_ops = &uffd_move_test_pmd_case_ops,
> + },
> {
> .name = "wp-fork",
> .uffd_fn = uffd_wp_fork_test,
> --
> 2.34.1
>
