| Message ID | 20240314170700.352845-3-aalbersh@redhat.com (mailing list archive) |
|---|---|
| State | Superseded, archived |
| Headers | show |
| Series | [v2] xfs: allow cross-linking special files without project quota | expand |
On Thu, Mar 14, 2024 at 06:07:02PM +0100, Andrey Albershteyn wrote: > There's an issue that if special files is created before quota > project is enabled, then it's not possible to link this file. This > works fine for normal files. This happens because xfs_quota skips > special files (no ioctls to set necessary flags). The check for > having the same project ID for source and destination then fails as > source file doesn't have any ID. > > mkfs.xfs -f /dev/sda > mount -o prjquota /dev/sda /mnt/test > > mkdir /mnt/test/foo > mkfifo /mnt/test/foo/fifo1 > > xfs_quota -xc "project -sp /mnt/test/foo 9" /mnt/test > > Setting up project 9 (path /mnt/test/foo)... > > xfs_quota: skipping special file /mnt/test/foo/fifo1 > > Processed 1 (/etc/projects and cmdline) paths for project 9 with recursion depth infinite (-1). > > ln /mnt/test/foo/fifo1 /mnt/test/foo/fifo1_link > > ln: failed to create hard link '/mnt/test/testdir/fifo1_link' => '/mnt/test/testdir/fifo1': Invalid cross-device link Aha. So hardlinking special files within a directory subtree that all have the same nonzero project quota ID fails if that special file happened to have been created before the subtree was assigned that pqid. And there's nothing we can do about that, because there's no way to call XFS_IOC_SETFSXATTR on a special file because opening those gets you a different inode from the special block/fifo/chardev filesystem... > mkfifo /mnt/test/foo/fifo2 > ln /mnt/test/foo/fifo2 /mnt/test/foo/fifo2_link > > Fix this by allowing linking of special files to the project quota > if special files doesn't have any ID set (ID = 0). ...and that's the workaround for this situation. The project quota accounting here will be weird because there will be (more) files in a directory subtree than is reported by xfs_quota, but the subtree was already messed up in that manner. Question: Should we have a XFS_IOC_SETFSXATTRAT where we can pass in relative directory paths and actually query/update special files? > Signed-off-by: Andrey Albershteyn <aalbersh@redhat.com> Reviewed-by: Darrick J. Wong <djwong@kernel.org> --D > --- > fs/xfs/xfs_inode.c | 15 +++++++++++++-- > 1 file changed, 13 insertions(+), 2 deletions(-) > > diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c > index 1fd94958aa97..b7be19be0132 100644 > --- a/fs/xfs/xfs_inode.c > +++ b/fs/xfs/xfs_inode.c > @@ -1240,8 +1240,19 @@ xfs_link( > */ > if (unlikely((tdp->i_diflags & XFS_DIFLAG_PROJINHERIT) && > tdp->i_projid != sip->i_projid)) { > - error = -EXDEV; > - goto error_return; > + /* > + * Project quota setup skips special files which can > + * leave inodes in a PROJINHERIT directory without a > + * project ID set. We need to allow links to be made > + * to these "project-less" inodes because userspace > + * expects them to succeed after project ID setup, > + * but everything else should be rejected. > + */ > + if (!special_file(VFS_I(sip)->i_mode) || > + sip->i_projid != 0) { > + error = -EXDEV; > + goto error_return; > + } > } > > if (!resblks) { > -- > 2.42.0 > >
On 2024-03-14 19:48:26, Darrick J. Wong wrote: > On Thu, Mar 14, 2024 at 06:07:02PM +0100, Andrey Albershteyn wrote: > > There's an issue that if special files is created before quota > > project is enabled, then it's not possible to link this file. This > > works fine for normal files. This happens because xfs_quota skips > > special files (no ioctls to set necessary flags). The check for > > having the same project ID for source and destination then fails as > > source file doesn't have any ID. > > > > mkfs.xfs -f /dev/sda > > mount -o prjquota /dev/sda /mnt/test > > > > mkdir /mnt/test/foo > > mkfifo /mnt/test/foo/fifo1 > > > > xfs_quota -xc "project -sp /mnt/test/foo 9" /mnt/test > > > Setting up project 9 (path /mnt/test/foo)... > > > xfs_quota: skipping special file /mnt/test/foo/fifo1 > > > Processed 1 (/etc/projects and cmdline) paths for project 9 with recursion depth infinite (-1). > > > > ln /mnt/test/foo/fifo1 /mnt/test/foo/fifo1_link > > > ln: failed to create hard link '/mnt/test/testdir/fifo1_link' => '/mnt/test/testdir/fifo1': Invalid cross-device link > > Aha. So hardlinking special files within a directory subtree that all > have the same nonzero project quota ID fails if that special file > happened to have been created before the subtree was assigned that pqid. > And there's nothing we can do about that, because there's no way to call > XFS_IOC_SETFSXATTR on a special file because opening those gets you a > different inode from the special block/fifo/chardev filesystem... > > > mkfifo /mnt/test/foo/fifo2 > > ln /mnt/test/foo/fifo2 /mnt/test/foo/fifo2_link > > > > Fix this by allowing linking of special files to the project quota > > if special files doesn't have any ID set (ID = 0). > > ...and that's the workaround for this situation. The project quota > accounting here will be weird because there will be (more) files in a > directory subtree than is reported by xfs_quota, but the subtree was > already messed up in that manner. Yeah, there's already that prj ID = 0 file, so nothing changes regarding accounting. > Question: Should we have a XFS_IOC_SETFSXATTRAT where we can pass in > relative directory paths and actually query/update special files? Added to xfs_quota to not skip them? It would probably solve the issue, but for existing filesystems with projects this will require to go through all of special files > > > Signed-off-by: Andrey Albershteyn <aalbersh@redhat.com> > > Reviewed-by: Darrick J. Wong <djwong@kernel.org> > > --D > > > --- > > fs/xfs/xfs_inode.c | 15 +++++++++++++-- > > 1 file changed, 13 insertions(+), 2 deletions(-) > > > > diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c > > index 1fd94958aa97..b7be19be0132 100644 > > --- a/fs/xfs/xfs_inode.c > > +++ b/fs/xfs/xfs_inode.c > > @@ -1240,8 +1240,19 @@ xfs_link( > > */ > > if (unlikely((tdp->i_diflags & XFS_DIFLAG_PROJINHERIT) && > > tdp->i_projid != sip->i_projid)) { > > - error = -EXDEV; > > - goto error_return; > > + /* > > + * Project quota setup skips special files which can > > + * leave inodes in a PROJINHERIT directory without a > > + * project ID set. We need to allow links to be made > > + * to these "project-less" inodes because userspace > > + * expects them to succeed after project ID setup, > > + * but everything else should be rejected. > > + */ > > + if (!special_file(VFS_I(sip)->i_mode) || > > + sip->i_projid != 0) { > > + error = -EXDEV; > > + goto error_return; > > + } > > } > > > > if (!resblks) { > > -- > > 2.42.0 > > > > >
On 2024-03-14 19:48:26, Darrick J. Wong wrote: > On Thu, Mar 14, 2024 at 06:07:02PM +0100, Andrey Albershteyn wrote: > > There's an issue that if special files is created before quota > > project is enabled, then it's not possible to link this file. This > > works fine for normal files. This happens because xfs_quota skips > > special files (no ioctls to set necessary flags). The check for > > having the same project ID for source and destination then fails as > > source file doesn't have any ID. > > > > mkfs.xfs -f /dev/sda > > mount -o prjquota /dev/sda /mnt/test > > > > mkdir /mnt/test/foo > > mkfifo /mnt/test/foo/fifo1 > > > > xfs_quota -xc "project -sp /mnt/test/foo 9" /mnt/test > > > Setting up project 9 (path /mnt/test/foo)... > > > xfs_quota: skipping special file /mnt/test/foo/fifo1 > > > Processed 1 (/etc/projects and cmdline) paths for project 9 with recursion depth infinite (-1). > > > > ln /mnt/test/foo/fifo1 /mnt/test/foo/fifo1_link > > > ln: failed to create hard link '/mnt/test/testdir/fifo1_link' => '/mnt/test/testdir/fifo1': Invalid cross-device link > > Aha. So hardlinking special files within a directory subtree that all > have the same nonzero project quota ID fails if that special file > happened to have been created before the subtree was assigned that pqid. > And there's nothing we can do about that, because there's no way to call > XFS_IOC_SETFSXATTR on a special file because opening those gets you a > different inode from the special block/fifo/chardev filesystem... > > > mkfifo /mnt/test/foo/fifo2 > > ln /mnt/test/foo/fifo2 /mnt/test/foo/fifo2_link > > > > Fix this by allowing linking of special files to the project quota > > if special files doesn't have any ID set (ID = 0). > > ...and that's the workaround for this situation. The project quota > accounting here will be weird because there will be (more) files in a > directory subtree than is reported by xfs_quota, but the subtree was > already messed up in that manner. > > Question: Should we have a XFS_IOC_SETFSXATTRAT where we can pass in > relative directory paths and actually query/update special files? After some more thinking/looking into the code this is probably the only way to make it work the same for special files. Also, I've noticed that this workaround can be applied to xfs_rename then. So, I will start with implementing XFS_IOC_SETFSXATTRAT
diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c index 1fd94958aa97..b7be19be0132 100644 --- a/fs/xfs/xfs_inode.c +++ b/fs/xfs/xfs_inode.c @@ -1240,8 +1240,19 @@ xfs_link( */ if (unlikely((tdp->i_diflags & XFS_DIFLAG_PROJINHERIT) && tdp->i_projid != sip->i_projid)) { - error = -EXDEV; - goto error_return; + /* + * Project quota setup skips special files which can + * leave inodes in a PROJINHERIT directory without a + * project ID set. We need to allow links to be made + * to these "project-less" inodes because userspace + * expects them to succeed after project ID setup, + * but everything else should be rejected. + */ + if (!special_file(VFS_I(sip)->i_mode) || + sip->i_projid != 0) { + error = -EXDEV; + goto error_return; + } } if (!resblks) {