Since project quota returns ENOSPC rather than EDQUOT, the distinction
between those two error conditions in xfs_file_buffered_aio_write() is
incomplete. If project quota is on, and we get ENOSPC, it seems that
we have no good way to know if it was due to quota, or due to actual
out of space conditions, and we may need to run both remediation options...
This is completely untested and not even built, because I'm really not sure
what the best way to go here is. True ENOSPC is hopefully rare, pquota
ENOSPC is probably less so, so I'm not sure how far we should go digging
to figure out what the root cause of the ENOSPC condition is, when pquota
is on (on this inode).
If project quota is on on this inode and pquota enforced, should we look
to the super to make a determination about low free blocks in the fs?
Should we crack open the dquot and see if it's over limit?
Should we just run both conditions and hope for the best?
Is this all best effort anyway, so we just simply care if we take the
improper action for pquota+ENOSPC?
Probably-shouldn't-merge-this-sez: Eric Sandeen <sandeen@xxxxxxxxxx>
Signed-off-by: Eric Sandeen <sandeen@xxxxxxxxxx>
---
diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index 4b8bdecc3863..8cec826046ce 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -647,27 +647,31 @@ xfs_file_buffered_aio_write(
* waits on dirty mappings. Since xfs_flush_inodes() is serialized, this
* also behaves as a filter to prevent too many eofblocks scans from
* running at the same time.
+ *
+ * Fun fact: Project quota returns -ENOSPC, so... complexity here.
*/
- if (ret == -EDQUOT && !enospc) {
- xfs_iunlock(ip, iolock);
- enospc = xfs_inode_free_quota_eofblocks(ip);
- if (enospc)
- goto write_retry;
- enospc = xfs_inode_free_quota_cowblocks(ip);
+ if (!enospc) {
+ if (ret == -ENOSPC) {
+ struct xfs_eofblocks eofb = {0};
+
+ enospc = 1;
+ xfs_flush_inodes(ip->i_mount);
+
+ xfs_iunlock(ip, iolock);
+ eofb.eof_flags = XFS_EOF_FLAGS_SYNC;
+ xfs_icache_free_eofblocks(ip->i_mount, &eofb);
+ xfs_icache_free_cowblocks(ip->i_mount, &eofb);
+ }
+ if (ret == -EDQUOT ||
+ (ret == -ENOSPC && ip->i_pdquot &&
+ XFS_IS_PQUOTA_ENFORCED(mp) && ip->i_pdquot)) {
+ xfs_iunlock(ip, iolock);
+ enospc |= xfs_inode_free_quota_eofblocks(ip);
+ enospc |= xfs_inode_free_quota_cowblocks(ip);
+ iolock = 0;
+ }
if (enospc)
goto write_retry;
- iolock = 0;
- } else if (ret == -ENOSPC && !enospc) {
- struct xfs_eofblocks eofb = {0};
-
- enospc = 1;
- xfs_flush_inodes(ip->i_mount);
-
- xfs_iunlock(ip, iolock);
- eofb.eof_flags = XFS_EOF_FLAGS_SYNC;
- xfs_icache_free_eofblocks(ip->i_mount, &eofb);
- xfs_icache_free_cowblocks(ip->i_mount, &eofb);
- goto write_retry;
}
current->backing_dev_info = NULL;
