[XFS] Fix inode allocation latency

The log force added in xfs_iget_core() has been a performance issue since
it was introduced for tight loops that allocate then unlink a single file.
under heavy writeback, this can introduce unnecessary latency due tothe
log I/o getting stuck behind bulk data writes.

Fix this latency problem by avoinding the need for the log force by moving
the place we mark linux inode dirty to the transaction commit rather than
on transaction completion.

This also closes a potential hole in the sync code where a linux inode is
not dirty between the time it is modified and the time the log buffer has
been written to disk.

SGI-PV: 972753
SGI-Modid: xfs-linux-melb:xfs-kern:30007a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>

1 files changed, 1 insertions, 33 deletions

diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
index 597e0ed4d2b..805cab7b277 100644
--- a/fs/xfs/xfs_inode.c
+++ b/fs/xfs/xfs_inode.c
@@ -2814,40 +2814,8 @@ xfs_iunpin(
 {
 	ASSERT(atomic_read(&ip->i_pincount) > 0);
 
-	if (atomic_dec_and_lock(&ip->i_pincount, &ip->i_flags_lock)) {
-
-		/*
-		 * If the inode is currently being reclaimed, the link between
-		 * the bhv_vnode and the xfs_inode will be broken after the
-		 * XFS_IRECLAIM* flag is set. Hence, if these flags are not
-		 * set, then we can move forward and mark the linux inode dirty
-		 * knowing that it is still valid as it won't freed until after
-		 * the bhv_vnode<->xfs_inode link is broken in xfs_reclaim. The
-		 * i_flags_lock is used to synchronise the setting of the
-		 * XFS_IRECLAIM* flags and the breaking of the link, and so we
-		 * can execute atomically w.r.t to reclaim by holding this lock
-		 * here.
-		 *
-		 * However, we still need to issue the unpin wakeup call as the
-		 * inode reclaim may be blocked waiting for the inode to become
-		 * unpinned.
-		 */
-
-		if (!__xfs_iflags_test(ip, XFS_IRECLAIM|XFS_IRECLAIMABLE)) {
-			bhv_vnode_t	*vp = XFS_ITOV_NULL(ip);
-			struct inode *inode = NULL;
-
-			BUG_ON(vp == NULL);
-			inode = vn_to_inode(vp);
-			BUG_ON(inode->i_state & I_CLEAR);
-
-			/* make sync come back and flush this inode */
-			if (!(inode->i_state & (I_NEW|I_FREEING)))
-				mark_inode_dirty_sync(inode);
-		}
-		spin_unlock(&ip->i_flags_lock);
+	if (atomic_dec_and_test(&ip->i_pincount))
 		wake_up(&ip->i_ipin_wait);
-	}
 }
 
 /*