Here is a patch that removes the XFS-internal writing of data in sync
and instead trusts the VFS code to get it right, and also does the
ioend_wait under a lock, to protect against new pending I/Os showing up.
It passed a single xfsqa run so far, so at least it won't blow up
completely.
Can you check if that helps with your livelock test case? (And yes, I
should go and try it myself. Will do as soon as I'll get a bit more
time).
Index: xfs/fs/xfs/linux-2.6/xfs_super.c
===================================================================
--- xfs.orig/fs/xfs/linux-2.6/xfs_super.c 2011-06-09 14:17:06.952532698 +0200
+++ xfs/fs/xfs/linux-2.6/xfs_super.c 2011-06-09 14:17:36.742531190 +0200
@@ -991,7 +991,7 @@ xfs_fs_write_inode(
* ->sync_fs call do that for thus, which reduces the number
* of synchronous log foces dramatically.
*/
- xfs_ioend_wait(ip);
+// xfs_ioend_wait(ip);
xfs_ilock(ip, XFS_ILOCK_SHARED);
if (ip->i_update_core) {
error = xfs_log_inode(ip);
Index: xfs/fs/xfs/linux-2.6/xfs_sync.c
===================================================================
--- xfs.orig/fs/xfs/linux-2.6/xfs_sync.c 2011-06-09 14:16:59.575866405 +0200
+++ xfs/fs/xfs/linux-2.6/xfs_sync.c 2011-06-09 14:22:31.675849591 +0200
@@ -215,6 +215,19 @@ xfs_inode_ag_iterator(
}
STATIC int
+xfs_wait_ioend_cb(
+ struct xfs_inode *ip,
+ struct xfs_perag *pag,
+ int flags)
+{
+ xfs_ilock(ip, XFS_IOLOCK_SHARED);
+ xfs_ioend_wait(ip);
+ xfs_iunlock(ip, XFS_IOLOCK_SHARED);
+
+ return 0;
+}
+
+STATIC int
xfs_sync_inode_data(
struct xfs_inode *ip,
struct xfs_perag *pag,
@@ -359,14 +372,14 @@ xfs_quiesce_data(
{
int error, error2 = 0;
- /* push non-blocking */
- xfs_sync_data(mp, 0);
- xfs_qm_sync(mp, SYNC_TRYLOCK);
-
- /* push and block till complete */
- xfs_sync_data(mp, SYNC_WAIT);
xfs_qm_sync(mp, SYNC_WAIT);
+ error = xfs_inode_ag_iterator(mp, xfs_wait_ioend_cb, 0);
+ if (error)
+ return XFS_ERROR(error);
+
+ xfs_log_force(mp, XFS_LOG_SYNC);
+
/* write superblock and hoover up shutdown errors */
error = xfs_sync_fsdata(mp);
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
