| Message ID | 20250228-mm-selftests-v3-0-958e3b6f0203@google.com |
|---|---|
| Headers | show |
| Series | selftests/mm: Some cleanups from trying to run them | expand |
On Fri, Feb 28, 2025 at 10:55:00PM +0530, Dev Jain wrote: > > > On 28/02/25 10:24 pm, Brendan Jackman wrote: > > It's obvious that this should fail in that case, but still, save the > > reader the effort of figuring out that they've run into this by just > > SKIPping > > > > Signed-off-by: Brendan Jackman <jackmanb@google.com> > > --- > > tools/testing/selftests/mm/uffd-wp-mremap.c | 5 ++++- > > 1 file changed, 4 insertions(+), 1 deletion(-) > > > > diff --git a/tools/testing/selftests/mm/uffd-wp-mremap.c b/tools/testing/selftests/mm/uffd-wp-mremap.c > > index 2c4f984bd73caa17e12b9f4a5bb71e7fdf5d8554..c2ba7d46c7b4581a3c32a6b6acd148e3e89c2172 100644 > > --- a/tools/testing/selftests/mm/uffd-wp-mremap.c > > +++ b/tools/testing/selftests/mm/uffd-wp-mremap.c > > @@ -182,7 +182,10 @@ static void test_one_folio(size_t size, bool private, bool swapout, bool hugetlb > > /* Register range for uffd-wp. */ > > if (userfaultfd_open(&features)) { > > - ksft_test_result_fail("userfaultfd_open() failed\n"); > > + if (errno == ENOENT) > > + ksft_test_result_skip("userfaultfd not available\n"); > > + else > > + ksft_test_result_fail("userfaultfd_open() failed\n"); > > goto out; > > } > > if (uffd_register(uffd, mem, size, false, true, false)) { > > > > I think you are correct, just want to confirm whether "uffd not available" > if and only if "errno == ENOENT" is true. That is, > is it possible that errno can be something else and uffd is still not > available, Yeah, I strongly suspect this can happen. This is an attempt to improve things but I don't think it's a full solution. I've been pondering this a bit and I think it's impractical to solve problems like this in the code of individual testst. I think the right thing to do is either: 1. Have a centralised facility for detecting conditions like "userfaultfd not available" that tests can just query it, so they say something like: ksft_test_requires("userfaultfd"); Which would do some sort of actual principled check for presence and then skip the test with an informative message when it's not there. There would be a list of these "system requirements" in the code so you can easily see in one place what things might be needed to successfully run all the tests. or 2. Specify out of band that there's a fixed set of requirements for running the tests and document that you shouldn't run them without satisfying them. Then just don't bother with SKIPs and call it user error. This would require some reasonably usable tooling for actually getting a system that satisfies the requirements. But both of them require a deeper investment. I would quite like to explore option 1 a bit but that's for a future Brendan. In the meantime I'm just trying to get these tests running on virtme-ng. (I'm not even gonna add all of them, because e.g. once I noticed this one I added a `scripts/config -e USERFAULTFD` to my script, so I won't notice if anything else is missing the check). > or errno can be ENOENT even if uffd is available. I think it's probably posible for this to happen too, e.g. if the system has a perverted /dev or something. But again I think that can only be solved with the kinda stuff I mentioned above. Sorry for the essay :D
I never had much luck running mm selftests so I spent a few hours digging into why. Looks like most of the reason is missing SKIP checks, so this series is just adding a bunch of those that I found. I did not do anything like all of them, just the ones I spotted in gup_longterm, gup_test, mmap, userfaultfd and memfd_secret. It's a bit unfortunate to have to skip those tests when ftruncate() fails, but I don't have time to dig deep enough into it to actually make them pass. I have observed the issue on 9pfs and heard rumours that NFS has a similar problem. I'm now able to run these test groups successfully: - mmap - gup_test - compaction - migration - page_frag - userfaultfd Signed-off-by: Brendan Jackman <jackmanb@google.com> --- Changes in v3: - Added fix for userfaultfd tests. - Dropped attempts to use sudo. - Fixed garbage printf in uffd-stress. (Added EXTRA_CFLAGS=-Werror FORCE_TARGETS=1 to my scripts to prevent such errors happening again). - Fixed missing newlines in ksft_test_result_skip() calls. - Link to v2: https://lore.kernel.org/r/20250221-mm-selftests-v2-0-28c4d66383c5@google.com Changes in v2 (Thanks to Dev for the reviews): - Improve and cleanup some error messages - Add some extra SKIPs - Fix misnaming of nr_cpus variable in uffd tests - Link to v1: https://lore.kernel.org/r/20250220-mm-selftests-v1-0-9bbf57d64463@google.com --- Brendan Jackman (10): selftests/mm: Report errno when things fail in gup_longterm selftests/mm: Skip uffd-stress if userfaultfd not available selftests/mm: Skip uffd-wp-mremap if userfaultfd not available selftests/mm/uffd: Rename nr_cpus -> nr_threads selftests/mm: Print some details when uffd-stress gets bad params selftests/mm: Don't fail uffd-stress if too many CPUs selftests/mm: Skip map_populate on weird filesystems selftests/mm: Skip gup_longerm tests on weird filesystems selftests/mm: Drop unnecessary sudo usage selftests/mm: Ensure uffd-wp-mremap gets pages of each size tools/testing/selftests/mm/gup_longterm.c | 45 ++++++++++++++++++---------- tools/testing/selftests/mm/map_populate.c | 7 +++++ tools/testing/selftests/mm/run_vmtests.sh | 25 ++++++++++++++-- tools/testing/selftests/mm/uffd-common.c | 8 ++--- tools/testing/selftests/mm/uffd-common.h | 2 +- tools/testing/selftests/mm/uffd-stress.c | 42 ++++++++++++++++---------- tools/testing/selftests/mm/uffd-unit-tests.c | 2 +- tools/testing/selftests/mm/uffd-wp-mremap.c | 5 +++- 8 files changed, 95 insertions(+), 41 deletions(-) --- base-commit: 76544811c850a1f4c055aa182b513b7a843868ea change-id: 20250220-mm-selftests-2d7d0542face Best regards,