Re: [PATCH 1/2] src/vfs: probe O_TMPFILE support on the base mount in the idmapped tests

[Zorro Lang <zlang@kernel.org>]

On Mon, Jul 20, 2026 at 06:15:31PM +0930, Qu Wenruo wrote:
> 
> 
> 在 2026/6/16 01:03, Christian Brauner 写道:
> > openat_tmpfile_supported() in the idmapped mount tests probes O_TMPFILE
> > on the idmapped mount (open_tree_fd) as the test's fsuid -- but that
> > fsuid is not mapped by the idmapped mount under test, so it really asks
> > "can this unmapped caller create a tmpfile through this idmap?" instead
> > of "does the filesystem support O_TMPFILE?".
> > 
> > On overlayfs this is a false negative: the backing tmpfile ends up owned
> > by an unmapped id, overlayfs opens it with O_NOATIME and may_open()
> > returns -EPERM, so the tmpfile sub-tests are wrongly skipped.
> > 
> > O_TMPFILE support is a property of the filesystem, not of the idmapped
> > mount or the caller's mapping. Probe info->t_dir1_fd, the base mount
> > that the caller owns, exactly as the non-idmapped setgid tests already
> > do. The probe then succeeds on every filesystem and the tmpfile
> > sub-tests run as the mapped fsuid.
> > 
> > Signed-off-by: Christian Brauner (Amutable) <brauner@kernel.org>
> 
> Reviewed-by: Qu Wenruo <wqu@suse.com>
> 
> Furthermore, since kernel commit 539dce114465 ("fs: refuse O_TMPFILE
> creation with an unmapped fsuid or fsgid"), even without overlayfs the
> involved tests will fail due to the extra checks.
> 
> So this is not only affecting overlayfs but all regular fses now.
> 
> I hope we can at least merge this one asap.

Hi Wenruo,

Before merging this patch, I've merged your:
  commit 1cca2bc0d7a9462bb1b59edb108bac7c9e7c424d
  Author: Qu Wenruo <wqu@suse.com>
  Date:   Fri Jul 17 16:37:17 2026 +0930

      fstests: vfstest: skip rejected tmpfile creation

Since (Christian's) patch seems to replace yours (I think), so do you mind
if I revert your patch and merge this one instead?

Thanks,
Zorro

> 
> Thanks,
> Qu
> 
> > ---
> >   src/vfs/idmapped-mounts.c | 12 ++++++------
> >   1 file changed, 6 insertions(+), 6 deletions(-)
> > 
> > diff --git a/src/vfs/idmapped-mounts.c b/src/vfs/idmapped-mounts.c
> > index ed9992f9..8f8441c9 100644
> > --- a/src/vfs/idmapped-mounts.c
> > +++ b/src/vfs/idmapped-mounts.c
> > @@ -3838,7 +3838,7 @@ int tcore_setgid_create_idmapped(const struct vfstest_info *info)
> >   		goto out;
> >   	}
> > -	supported = openat_tmpfile_supported(open_tree_fd);
> > +	supported = openat_tmpfile_supported(info->t_dir1_fd);
> >   	pid = fork();
> >   	if (pid < 0) {
> > @@ -4014,7 +4014,7 @@ int tcore_setgid_create_idmapped_in_userns(const struct vfstest_info *info)
> >   		goto out;
> >   	}
> > -	supported = openat_tmpfile_supported(open_tree_fd);
> > +	supported = openat_tmpfile_supported(info->t_dir1_fd);
> >   	pid = fork();
> >   	if (pid < 0) {
> > @@ -7733,7 +7733,7 @@ static int setgid_create_umask_idmapped(const struct vfstest_info *info)
> >   		goto out;
> >   	}
> > -	supported = openat_tmpfile_supported(open_tree_fd);
> > +	supported = openat_tmpfile_supported(info->t_dir1_fd);
> >   	pid = fork();
> >   	if (pid < 0) {
> > @@ -7947,7 +7947,7 @@ static int setgid_create_umask_idmapped_in_userns(const struct vfstest_info *inf
> >   		goto out;
> >   	}
> > -	supported = openat_tmpfile_supported(open_tree_fd);
> > +	supported = openat_tmpfile_supported(info->t_dir1_fd);
> >   	/*
> >   	 * Below we verify that setgid inheritance for a newly created file or
> > @@ -8181,7 +8181,7 @@ static int setgid_create_acl_idmapped(const struct vfstest_info *info)
> >   		goto out;
> >   	}
> > -	supported = openat_tmpfile_supported(open_tree_fd);
> > +	supported = openat_tmpfile_supported(info->t_dir1_fd);
> >   	pid = fork();
> >   	if (pid < 0) {
> > @@ -8536,7 +8536,7 @@ static int setgid_create_acl_idmapped_in_userns(const struct vfstest_info *info)
> >   		goto out;
> >   	}
> > -	supported = openat_tmpfile_supported(open_tree_fd);
> > +	supported = openat_tmpfile_supported(info->t_dir1_fd);
> >   	/*
> >   	 * Below we verify that setgid inheritance for a newly created file or
> > 
>