[PATCH] xfs: force buffer writeback before blocking on the ilock in inode reclaim

[Christoph Hellwig]

If we are doing synchronous inode reclaim we block the VM from making
progress in memory reclaim.  So if we encouter a flush locked inode
make sure we force out all delayed buffers ASAP to speed up the wait
for it to be unlocked.  Without this we can get hangs of up to 30
seconds during workloads hitting synchronous inode reclaim.

Reported-by: Simon Kirby <sim at hostway.ca>
Tested-by: Simon Kirby <sim at hostway.ca>
Signed-off-by: Christoph Hellwig <hch at lst.de>

Index: xfs/fs/xfs/xfs_sync.c
===================================================================
--- xfs.orig/fs/xfs/xfs_sync.c	2011-11-19 20:14:52.110141228 +0100
+++ xfs/fs/xfs/xfs_sync.c	2011-11-19 20:40:17.381878121 +0100
@@ -762,7 +762,8 @@ xfs_reclaim_inode(
 	struct xfs_perag	*pag,
 	int			sync_mode)
 {
-	int	error;
+	struct xfs_mount	*mp = ip->i_mount;
+	int			error;
 
 restart:
 	error = 0;
@@ -770,12 +771,25 @@ restart:
 	if (!xfs_iflock_nowait(ip)) {
 		if (!(sync_mode & SYNC_WAIT))
 			goto out;
+
+		/*
+		 * If we only have a single dirty inode in a cluster there is
+		 * a fair chance that the AIL push may have pushed it into
+		 * the buffer, but xfsbufd won't touch it until 30 seconds
+		 * from now, and thus we will lock up here.
+		 *
+		 * Wakeup xfsbufd now, and force it to write back even
+		 * recently dirtied buffers.
+		 */
+		set_bit(XBT_FORCE_FLUSH, &mp->m_ddev_targp->bt_flags);
+		wake_up_process(mp->m_ddev_targp->bt_task);
+
 		xfs_iflock(ip);
 	}
 
 	if (is_bad_inode(VFS_I(ip)))
 		goto reclaim;
-	if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
+	if (XFS_FORCED_SHUTDOWN(mp)) {
 		xfs_iunpin_wait(ip);
 		goto reclaim;
 	}
@@ -829,8 +843,8 @@ restart:
 	 * is permanent then the next sync reclaim will reclaim the inode and
 	 * pass on the error.
 	 */
-	if (error && error != EAGAIN && !XFS_FORCED_SHUTDOWN(ip->i_mount)) {
-		xfs_warn(ip->i_mount,
+	if (error && error != EAGAIN && !XFS_FORCED_SHUTDOWN(mp)) {
+		xfs_warn(mp,
 			"inode 0x%llx background reclaim flush failed with %d",
 			(long long)ip->i_ino, error);
 	}
@@ -860,7 +874,7 @@ reclaim:
 	 */
 	spin_lock(&pag->pag_ici_lock);
 	if (!radix_tree_delete(&pag->pag_ici_root,
-				XFS_INO_TO_AGINO(ip->i_mount, ip->i_ino)))
+				XFS_INO_TO_AGINO(mp, ip->i_ino)))
 		ASSERT(0);
 	__xfs_inode_clear_reclaim(pag, ip);
 	spin_unlock(&pag->pag_ici_lock);