4 files changed, 39 insertions, 40 deletions

diff --git a/fs/xfs/xfs_iget.c b/fs/xfs/xfs_iget.c
index 539c2dd8eae8..e229e9e001c2 100644
--- a/fs/xfs/xfs_iget.c
+++ b/fs/xfs/xfs_iget.c
@@ -216,7 +216,14 @@ finish_inode:
         mrlock_init(&ip->i_iolock, MRLOCK_BARRIER, "xfsio", ip->i_ino);
         init_waitqueue_head(&ip->i_ipin_wait);
         atomic_set(&ip->i_pincount, 0);
-        initnsema(&ip->i_flock, 1, "xfsfino");
+
+        /*
+         * Because we want to use a counting completion, complete
+         * the flush completion once to allow a single access to
+         * the flush completion without blocking.
+         */
+        init_completion(&ip->i_flush);
+        complete(&ip->i_flush);
 
         if (lock_flags)
                 xfs_ilock(ip, lock_flags);
@@ -783,26 +790,3 @@ xfs_isilocked(
 }
 #endif
 
-/*
- * The following three routines simply manage the i_flock
- * semaphore embedded in the inode.  This semaphore synchronizes
- * processes attempting to flush the in-core inode back to disk.
- */
-void
-xfs_iflock(xfs_inode_t *ip)
-{
-        psema(&(ip->i_flock), PINOD|PLTWAIT);
-}
-
-int
-xfs_iflock_nowait(xfs_inode_t *ip)
-{
-        return (cpsema(&(ip->i_flock)));
-}
-
-void
-xfs_ifunlock(xfs_inode_t *ip)
-{
-        ASSERT(issemalocked(&(ip->i_flock)));
-        vsema(&(ip->i_flock));
-}
diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
index b8444ee4dc95..aea62222b3a1 100644
--- a/fs/xfs/xfs_inode.c
+++ b/fs/xfs/xfs_inode.c
@@ -2626,7 +2626,6 @@ xfs_idestroy(
                 xfs_idestroy_fork(ip, XFS_ATTR_FORK);
         mrfree(&ip->i_lock);
         mrfree(&ip->i_iolock);
-        freesema(&ip->i_flock);
 
 #ifdef XFS_INODE_TRACE
         ktrace_free(ip->i_trace);
@@ -3044,10 +3043,10 @@ cluster_corrupt_out:
 /*
  * xfs_iflush() will write a modified inode's changes out to the
  * inode's on disk home.  The caller must have the inode lock held
- * in at least shared mode and the inode flush semaphore must be
- * held as well.  The inode lock will still be held upon return from
+ * in at least shared mode and the inode flush completion must be
+ * active as well.  The inode lock will still be held upon return from
  * the call and the caller is free to unlock it.
- * The inode flush lock will be unlocked when the inode reaches the disk.
+ * The inode flush will be completed when the inode reaches the disk.
  * The flags indicate how the inode's buffer should be written out.
  */
 int
@@ -3066,7 +3065,7 @@ xfs_iflush(
         XFS_STATS_INC(xs_iflush_count);
 
         ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL|XFS_ILOCK_SHARED));
-        ASSERT(issemalocked(&(ip->i_flock)));
+        ASSERT(!completion_done(&ip->i_flush));
         ASSERT(ip->i_d.di_format != XFS_DINODE_FMT_BTREE ||
                ip->i_d.di_nextents > ip->i_df.if_ext_max);
 
@@ -3229,7 +3228,7 @@ xfs_iflush_int(
 #endif
 
         ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL|XFS_ILOCK_SHARED));
-        ASSERT(issemalocked(&(ip->i_flock)));
+        ASSERT(!completion_done(&ip->i_flush));
         ASSERT(ip->i_d.di_format != XFS_DINODE_FMT_BTREE ||
                ip->i_d.di_nextents > ip->i_df.if_ext_max);
 
diff --git a/fs/xfs/xfs_inode.h b/fs/xfs/xfs_inode.h
index 571724404355..f771df6bfa6f 100644
--- a/fs/xfs/xfs_inode.h
+++ b/fs/xfs/xfs_inode.h
@@ -223,7 +223,7 @@ typedef struct xfs_inode {
         struct xfs_inode_log_item *i_itemp;     /* logging information */
         mrlock_t                i_lock;         /* inode lock */
         mrlock_t                i_iolock;       /* inode IO lock */
-        sema_t                  i_flock;        /* inode flush lock */
+        struct completion       i_flush;        /* inode flush completion q */
         atomic_t                i_pincount;     /* inode pin count */
         wait_queue_head_t       i_ipin_wait;    /* inode pinning wait queue */
         spinlock_t              i_flags_lock;   /* inode i_flags lock */
@@ -482,11 +482,8 @@ int		xfs_ilock_nowait(xfs_inode_t *, uint);
 void            xfs_iunlock(xfs_inode_t *, uint);
 void            xfs_ilock_demote(xfs_inode_t *, uint);
 int             xfs_isilocked(xfs_inode_t *, uint);
-void            xfs_iflock(xfs_inode_t *);
-int             xfs_iflock_nowait(xfs_inode_t *);
 uint            xfs_ilock_map_shared(xfs_inode_t *);
 void            xfs_iunlock_map_shared(xfs_inode_t *, uint);
-void            xfs_ifunlock(xfs_inode_t *);
 void            xfs_ireclaim(xfs_inode_t *);
 int             xfs_finish_reclaim(xfs_inode_t *, int, int);
 int             xfs_finish_reclaim_all(struct xfs_mount *, int);
@@ -580,6 +577,26 @@ extern struct kmem_zone	*xfs_ifork_zone;
 extern struct kmem_zone *xfs_inode_zone;
 extern struct kmem_zone *xfs_ili_zone;
 
+/*
+ * Manage the i_flush queue embedded in the inode.  This completion
+ * queue synchronizes processes attempting to flush the in-core
+ * inode back to disk.
+ */
+static inline void xfs_iflock(xfs_inode_t *ip)
+{
+        wait_for_completion(&ip->i_flush);
+}
+
+static inline int xfs_iflock_nowait(xfs_inode_t *ip)
+{
+        return try_wait_for_completion(&ip->i_flush);
+}
+
+static inline void xfs_ifunlock(xfs_inode_t *ip)
+{
+        complete(&ip->i_flush);
+}
+
 #endif  /* __KERNEL__ */
 
 #endif  /* __XFS_INODE_H__ */
diff --git a/fs/xfs/xfs_inode_item.c b/fs/xfs/xfs_inode_item.c
index 0eee08a32c26..97c7452e2620 100644
--- a/fs/xfs/xfs_inode_item.c
+++ b/fs/xfs/xfs_inode_item.c
@@ -779,11 +779,10 @@ xfs_inode_item_pushbuf(
         ASSERT(iip->ili_push_owner == current_pid());
 
         /*
-         * If flushlock isn't locked anymore, chances are that the
-         * inode flush completed and the inode was taken off the AIL.
-         * So, just get out.
+         * If a flush is not in progress anymore, chances are that the
+         * inode was taken off the AIL. So, just get out.
          */
-        if (!issemalocked(&(ip->i_flock)) ||
+        if (completion_done(&ip->i_flush) ||
             ((iip->ili_item.li_flags & XFS_LI_IN_AIL) == 0)) {
                 iip->ili_pushbuf_flag = 0;
                 xfs_iunlock(ip, XFS_ILOCK_SHARED);
@@ -805,7 +804,7 @@ xfs_inode_item_pushbuf(
                          * If not, we can flush it async.
                          */
                         dopush = ((iip->ili_item.li_flags & XFS_LI_IN_AIL) &&
-                                  issemalocked(&(ip->i_flock)));
+                                  !completion_done(&ip->i_flush));
                         iip->ili_pushbuf_flag = 0;
                         xfs_iunlock(ip, XFS_ILOCK_SHARED);
                         xfs_buftrace("INODE ITEM PUSH", bp);
@@ -858,7 +857,7 @@ xfs_inode_item_push(
         ip = iip->ili_inode;
 
         ASSERT(xfs_isilocked(ip, XFS_ILOCK_SHARED));
-        ASSERT(issemalocked(&(ip->i_flock)));
+        ASSERT(!completion_done(&ip->i_flush));
         /*
          * Since we were able to lock the inode's flush lock and
          * we found it on the AIL, the inode must be dirty.  This