renegade-project/linux-next
2
0
Fork 0
mirror of https://github.com/edk2-porting/linux-next.git synced 2024-12-29 07:34:06 +08:00
Code

xfs: correct the narrative around misaligned rtinherit/extszinherit dirs

Browse Source 
While auditing the realtime growfs code, I realized that the GROWFSRT
ioctl (and by extension xfs_growfs) has always allowed sysadmins to
change the realtime extent size when adding a realtime section to the
filesystem.  Since we also have always allowed sysadmins to set
RTINHERIT and EXTSZINHERIT on directories even if there is no realtime
device, this invalidates the premise laid out in the comments added in
commit 603f000b15.

In other words, this is not a case of inadequate metadata validation.
This is a case of nearly forgotten (and apparently untested) but
supported functionality.  Update the comments to reflect what we've
learned, and remove the log message about correcting the misalignment.

Fixes: 603f000b15 ("xfs: validate extsz hints against rt extent size when rtinherit is set")
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Carlos Maiolino <cmaiolino@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
This commit is contained in:
Darrick J. Wong 2021-07-12 12:58:50 -07:00
parent 5838d0356b
commit 83193e5ebb
3 changed files with 24 additions and 22 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

28
fs/xfs/libxfs/xfs_inode_buf.c
View File

@ -592,23 +592,27 @@ xfs_inode_validate_extsize(
/*
* This comment describes a historic gap in this verifier function.
*
* On older kernels, the extent size hint verifier doesn't check that
* the extent size hint is an integer multiple of the realtime extent
* size on a directory with both RTINHERIT and EXTSZINHERIT flags set.
* The verifier has always enforced the alignment rule for regular
* files with the REALTIME flag set.
* For a directory with both RTINHERIT and EXTSZINHERIT flags set, this
* function has never checked that the extent size hint is an integer
* multiple of the realtime extent size. Since we allow users to set
* this combination on non-rt filesystems /and/ to change the rt
* extent size when adding a rt device to a filesystem, the net effect
* is that users can configure a filesystem anticipating one rt
* geometry and change their minds later. Directories do not use the
* extent size hint, so this is harmless for them.
*
* If a directory with a misaligned extent size hint is allowed to
* propagate that hint into a new regular realtime file, the result
* is that the inode cluster buffer verifier will trigger a corruption
* shutdown the next time it is run.
* shutdown the next time it is run, because the verifier has always
* enforced the alignment rule for regular files.
*
* Unfortunately, there could be filesystems with these misconfigured
* directories in the wild, so we cannot add a check to this verifier
* at this time because that will result a new source of directory
* corruption errors when reading an existing filesystem. Instead, we
* permit the misconfiguration to pass through the verifiers so that
* callers of this function can correct and mitigate externally.
* Because we allow administrators to set a new rt extent size when
* adding a rt section, we cannot add a check to this verifier because
* that will result a new source of directory corruption errors when
* reading an existing filesystem. Instead, we rely on callers to
* decide when alignment checks are appropriate, and fix things up as
* needed.
*/
if (rt_flag)

10
fs/xfs/libxfs/xfs_trans_inode.c
View File

@ -143,16 +143,14 @@ xfs_trans_log_inode(
}
/*
* Inode verifiers on older kernels don't check that the extent size
* hint is an integer multiple of the rt extent size on a directory
* with both rtinherit and extszinherit flags set. If we're logging a
* directory that is misconfigured in this way, clear the hint.
* Inode verifiers do not check that the extent size hint is an integer
* multiple of the rt extent size on a directory with both rtinherit
* and extszinherit flags set. If we're logging a directory that is
* misconfigured in this way, clear the hint.
*/
if ((ip->i_diflags & XFS_DIFLAG_RTINHERIT) &&
(ip->i_diflags & XFS_DIFLAG_EXTSZINHERIT) &&
(ip->i_extsize % ip->i_mount->m_sb.sb_rextsize) > 0) {
xfs_info_once(ip->i_mount,
"Correcting misaligned extent size hint in inode 0x%llx.", ip->i_ino);
ip->i_diflags &= ~(XFS_DIFLAG_EXTSIZE |
XFS_DIFLAG_EXTSZINHERIT);
ip->i_extsize = 0;

8
fs/xfs/xfs_ioctl.c
View File

@ -1292,10 +1292,10 @@ xfs_ioctl_setattr_check_extsize(
new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
/*
* Inode verifiers on older kernels don't check that the extent size
* hint is an integer multiple of the rt extent size on a directory
* with both rtinherit and extszinherit flags set. Don't let sysadmins
* misconfigure directories.
* Inode verifiers do not check that the extent size hint is an integer
* multiple of the rt extent size on a directory with both rtinherit
* and extszinherit flags set. Don't let sysadmins misconfigure
* directories.
*/
if ((new_diflags & XFS_DIFLAG_RTINHERIT) &&
(new_diflags & XFS_DIFLAG_EXTSZINHERIT)) {