]> git.ipfire.org Git - thirdparty/xfsprogs-dev.git/commitdiff
xfs: don't trip over uninitialized buffer on extent read of corrupted inode
authorBrian Foster <bfoster@redhat.com>
Mon, 6 May 2019 22:00:29 +0000 (18:00 -0400)
committerEric Sandeen <sandeen@redhat.com>
Mon, 6 May 2019 22:00:29 +0000 (18:00 -0400)
Source kernel commit: 6958d11f77d45db80f7e22a21a74d4d5f44dc667

We've had rather rare reports of bmap btree block corruption where
the bmap root block has a level count of zero. The root cause of the
corruption is so far unknown. We do have verifier checks to detect
this form of on-disk corruption, but this doesn't cover a memory
corruption variant of the problem. The latter is a reasonable
possibility because the root block is part of the inode fork and can
reside in-core for some time before inode extents are read.

If this occurs, it leads to a system crash such as the following:

BUG: unable to handle kernel paging request at ffffffff00000221
PF error: [normal kernel read fault]
...
RIP: 0010:xfs_trans_brelse+0xf/0x200 [xfs]
...
Call Trace:
xfs_iread_extents+0x379/0x540 [xfs]
xfs_file_iomap_begin_delay+0x11a/0xb40 [xfs]
? xfs_attr_get+0xd1/0x120 [xfs]
? iomap_write_begin.constprop.40+0x2d0/0x2d0
xfs_file_iomap_begin+0x4c4/0x6d0 [xfs]
? __vfs_getxattr+0x53/0x70
? iomap_write_begin.constprop.40+0x2d0/0x2d0
iomap_apply+0x63/0x130
? iomap_write_begin.constprop.40+0x2d0/0x2d0
iomap_file_buffered_write+0x62/0x90
? iomap_write_begin.constprop.40+0x2d0/0x2d0
xfs_file_buffered_aio_write+0xe4/0x3b0 [xfs]
__vfs_write+0x150/0x1b0
vfs_write+0xba/0x1c0
ksys_pwrite64+0x64/0xa0
do_syscall_64+0x5a/0x1d0
entry_SYSCALL_64_after_hwframe+0x49/0xbe

The crash occurs because xfs_iread_extents() attempts to release an
uninitialized buffer pointer as the level == 0 value prevented the
buffer from ever being allocated or read. Change the level > 0
assert to an explicit error check in xfs_iread_extents() to avoid
crashing the kernel in the event of localized, in-core inode
corruption.

Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
libxfs/xfs_bmap.c

index 3ef11e8979e35f71b39c66a3cc542c8706f54f52..9b38d5df85e50b05a9af7e988d36d751db9d0214 100644 (file)
@@ -1182,7 +1182,10 @@ xfs_iread_extents(
         * Root level must use BMAP_BROOT_PTR_ADDR macro to get ptr out.
         */
        level = be16_to_cpu(block->bb_level);
-       ASSERT(level > 0);
+       if (unlikely(level == 0)) {
+               XFS_ERROR_REPORT(__func__, XFS_ERRLEVEL_LOW, mp);
+               return -EFSCORRUPTED;
+       }
        pp = XFS_BMAP_BROOT_PTR_ADDR(mp, block, 1, ifp->if_broot_bytes);
        bno = be64_to_cpu(*pp);