xfs: replace i_flock with a sleeping bitlock

We almost never block on i_flock, the exception is synchronous inode
flushing.  Instead of bloating the inode with a 16/24-byte completion
that we abuse as a semaphore just implement it as a bitlock that uses
a bit waitqueue for the rare sleeping path.  This primarily is a
tradeoff between a much smaller inode and a faster non-blocking
path vs faster wakeups, and we are much better off with the former.

A small downside is that we will lose lockdep checking for i_flock, but
given that it's always taken inside the ilock that should be acceptable.

Note that for example the inode writeback locking is implemented in a
very similar way.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Alex Elder <aelder@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>

1 files changed, 4 insertions, 5 deletions

diff --git a/fs/xfs/xfs_sync.c b/fs/xfs/xfs_sync.c
index 72c01a1c16e7..40b75eecd2b4 100644
--- a/fs/xfs/xfs_sync.c
+++ b/fs/xfs/xfs_sync.c
@@ -707,14 +707,13 @@ xfs_reclaim_inode_grab(
 		return 1;
 
 	/*
-	 * do some unlocked checks first to avoid unnecessary lock traffic.
-	 * The first is a flush lock check, the second is a already in reclaim
-	 * check. Only do these checks if we are not going to block on locks.
+	 * If we are asked for non-blocking operation, do unlocked checks to
+	 * see if the inode already is being flushed or in reclaim to avoid
+	 * lock traffic.
 	 */
 	if ((flags & SYNC_TRYLOCK) &&
-	    (!ip->i_flush.done || __xfs_iflags_test(ip, XFS_IRECLAIM))) {
+	    __xfs_iflags_test(ip, XFS_IFLOCK | XFS_IRECLAIM))
 		return 1;
-	}
 
 	/*
 	 * The radix tree lock here protects a thread in xfs_iget from racing