On Thu, May 23, 2024 at 4:00 PM Paul Moore <paul@xxxxxxxxxxxxxx> wrote: > > On Thu, May 23, 2024 at 8:59 AM Stephen Smalley > <stephen.smalley.work@xxxxxxxxx> wrote: > > On Wed, May 15, 2024 at 10:54 AM Stephen Smalley > > <stephen.smalley.work@xxxxxxxxx> wrote: > > > > > > On Mon, May 6, 2024 at 1:51 PM Stephen Smalley > > > <stephen.smalley.work@xxxxxxxxx> wrote: > > > > > > > > These tests currently fail on mount(2) calls due to the directory being > > > > unlabeled at the point where search access is checked. Until we can resolve > > > > the underlying issue, comment out these tests to allow the NFS tests to > > > > be run. > > > > > > With these two patches, I can run the nfs.sh script to completion with > > > all tests passing on not only the latest kernel w/ the fix but also > > > Linux v5.14 with no changes. So it is unclear to me that the tests > > > being disabled by these two patches ever worked... > > > > Last call - any objections to me applying these two patches? > > Do we have any hope of resolving the issues anytime soon? It might be > nice to see an in-depth discussion/brain-dump of the issues in the > commit description so that someone looking at this at a later date has > some hope of understanding the problem. I already put what I knew into the commit messages, unless I missed something. I am not sure the tests I am commenting out ever worked, and Ondrej said earlier that he never enabled the NFS tests as part of his automated testing because they still weren't working fully last he tried. Meanwhile, we've had two separate regressions in labeled NFS since that time, one from Neil Brown's patches and one from Ondrej's patches, and only just discovered the one long after it was introduced. Hence, to me the value of getting the tests running so we can turn them on in automated testing far outweighs any benefit we get from retaining these tests that may have never worked at all.
