XFS reports lchmod failure, but changes file system contents

[Darrick J. Wong]

On Wed, Feb 12, 2020 at 10:11:28AM -0800, Christoph Hellwig wrote:
> On Wed, Feb 12, 2020 at 08:16:04AM -0800, Darrick J. Wong wrote:
> > xfs_setattr_nonsize calls posix_acl_chmod which returns EOPNOTSUPP
> > because the xfs symlink inode_operations do not include a ->set_acl
> > pointer.
> > 
> > I /think/ that posix_acl_chmod code exists to enforce that the file mode
> > reflects any acl that might be set on the inode, but in this case the
> > inode is a symbolic link.
> > 
> > I don't remember off the top of my head if ACLs are supposed to apply to
> > symlinks, but what do you think about adding get_acl/set_acl pointers to
> > xfs_symlink_inode_operations and xfs_inline_symlink_inode_operations ?
> 
> Symlinks don't have permissions or ACLs, so adding them makes no
> sense.

Ahh, I thought so!

> xfs doesn't seem all that different from the other file systems,
> so I suspect you'll also see it with other on-disk file systems.

Yeah, I noticed that btrfs seems to exhibit the same behavior.

I also noticed that ext4 actually /does/ implement [gs]et_acl for
symlinks.

> We probably need a check high up in the chmod and co code to reject
> the operation early for O_PATH file descriptors pointing to symlinks.

<nod>

--D