On Fri, Feb 10, 2023 at 9:51 PM Leah Rumancik <leah.rumancik@xxxxxxxxx> wrote:
>
[...]
> > > I did not push on that until now because SGID test expectations were
> > > a moving target, but since xfstests commit 81e6f628 ("generic: update
> > > setgid tests") in this week's xfstests release, I think that tests should be
> > > stable and we can finally start backporting all relevant SGID fixes to
> > > align the SGID behavior of LTS kernels with that of upstream.
>
> Ooo goody, ok, will do this next.
>
> The following patches are on my radar to look into for this set. I have
> yet to look into dependencies, so the set may grow. If the sgid tests
> still fail after these ptaches, I will continue hunting for more fixes
> to include in this set.
>
> e014f37db1a2 xfs: use setattr_copy to set vfs inode attributes
> 472c6e46f589 xfs: remove XFS_PREALLOC_SYNC
> fbe7e5200365 xfs: fallocate() should call file_modified()
> 0b02c8c0d75a xfs: set prealloc flag in xfs_alloc_file_space()
> 2b3416ceff5e fs: add mode_strip_sgid() helper
> 1639a49ccdce fs: move S_ISGID stripping into the vfs_*() helpers
> ed5a7047d201 attr: use consistent sgid stripping checks
> 8d84e39d76bd fs: use consistent setgid checks in is_sxid()
>
> In addition to the normal regression testing, I will specifically look
> at the following tests for the sgid changes:
>
> generic/673
> generic/68[3-7]
> generic/69[6-7]
>
> I will also do some extra runs on the entire perms group.
>
> Let me know if you think something should be dropped or added.
>
I reckon you will need those dependency prep commits from
Christian's PR [1]:
11c2a8700cdc attr: add in_group_or_capable()
e243e3f94c80 fs: move should_remove_suid()
72ae017c5451 attr: add setattr_should_drop_sgid()
FYI, the ovl commits from this PR are independent fixes that were
already applied to 5.15.y.
Thanks,
Amir.
[1] https://lore.kernel.org/linux-fsdevel/20221212112053.99208-1-brauner@xxxxxxxxxx/
