[GFS2] Fix bz 224480 and cleanup glock demotion code

This patch prevents the printing of a warning message in cases where
the fs is functioning normally by handing off responsibility for
unlinked, but still open inodes, to another node for eventual deallocation.
Also, there is now an improved system for ensuring that such requests
to other nodes do not get lost. The callback on the iopen lock is
only ever called when i_nlink == 0 and when a node is unable to deallocate
it due to it still being in use on another node. When a node receives
the callback therefore, it knows that i_nlink must be zero, so we mark
it as such (in gfs2_drop_inode) in order that it will then attempt
deallocation of the inode itself.

As an additional benefit, queuing a demote request no longer requires
a memory allocation. This simplifies the code for dealing with gfs2_holders
as it removes one special case.

There are two new fields in struct gfs2_glock. gl_demote_state is the
state which the remote node has requested and gl_demote_time is the
time when the request came in. Both fields are only valid when the
GLF_DEMOTE flag is set in gl_flags.
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

fs/gfs2/glock.h

@@ -67,7 +67,7 @@ static inline int gfs2_glock_is_blocking(struct gfs2_glock *gl)
 {
 	int ret;
 	spin_lock(&gl->gl_spin);
-	ret = !list_empty(&gl->gl_waiters2) || !list_empty(&gl->gl_waiters3);
+	ret = test_bit(GLF_DEMOTE, &gl->gl_flags) || !list_empty(&gl->gl_waiters3);
 	spin_unlock(&gl->gl_spin);
 	return ret;
 }

fs/gfs2/incore.h

@@ -115,11 +115,8 @@ enum {
 	/* Actions */
 	HIF_MUTEX		= 0,
 	HIF_PROMOTE		= 1,
-	HIF_DEMOTE		= 2,
 
 	/* States */
-	HIF_ALLOCED		= 4,
-	HIF_DEALLOC		= 5,
 	HIF_HOLDER		= 6,
 	HIF_FIRST		= 7,
 	HIF_ABORTED		= 9,
@@ -142,8 +139,8 @@ struct gfs2_holder {
 enum {
 	GLF_LOCK		= 1,
 	GLF_STICKY		= 2,
+	GLF_DEMOTE		= 3,
 	GLF_DIRTY		= 5,
-	GLF_SKIP_WAITERS2	= 6,
 };
 
 struct gfs2_glock {
@@ -156,11 +153,12 @@ struct gfs2_glock {
 
 	unsigned int gl_state;
 	unsigned int gl_hash;
+	unsigned int gl_demote_state; /* state requested by remote node */
+	unsigned long gl_demote_time; /* time of first demote request */
 	struct task_struct *gl_owner;
 	unsigned long gl_ip;
 	struct list_head gl_holders;
 	struct list_head gl_waiters1;	/* HIF_MUTEX */
-	struct list_head gl_waiters2;	/* HIF_DEMOTE */
 	struct list_head gl_waiters3;	/* HIF_PROMOTE */
 
 	const struct gfs2_glock_operations *gl_ops;

fs/gfs2/main.c

@@ -45,7 +45,6 @@ static void gfs2_init_glock_once(void *foo, struct kmem_cache *cachep, unsigned
 		spin_lock_init(&gl->gl_spin);
 		INIT_LIST_HEAD(&gl->gl_holders);
 		INIT_LIST_HEAD(&gl->gl_waiters1);
-		INIT_LIST_HEAD(&gl->gl_waiters2);
 		INIT_LIST_HEAD(&gl->gl_waiters3);
 		gl->gl_lvb = NULL;
 		atomic_set(&gl->gl_lvb_count, 0);

fs/gfs2/ops_super.c

@@ -283,6 +283,31 @@ static int gfs2_remount_fs(struct super_block *sb, int *flags, char *data)
 	return error;
 }
 
+/**
+ * gfs2_drop_inode - Drop an inode (test for remote unlink)
+ * @inode: The inode to drop
+ *
+ * If we've received a callback on an iopen lock then its because a
+ * remote node tried to deallocate the inode but failed due to this node
+ * still having the inode open. Here we mark the link count zero
+ * since we know that it must have reached zero if the GLF_DEMOTE flag
+ * is set on the iopen glock. If we didn't do a disk read since the
+ * remote node removed the final link then we might otherwise miss
+ * this event. This check ensures that this node will deallocate the
+ * inode's blocks, or alternatively pass the baton on to another
+ * node for later deallocation.
+ */
+static void gfs2_drop_inode(struct inode *inode)
+{
+	if (inode->i_private && inode->i_nlink) {
+		struct gfs2_inode *ip = GFS2_I(inode);
+		struct gfs2_glock *gl = ip->i_iopen_gh.gh_gl;
+		if (gl && test_bit(GLF_DEMOTE, &gl->gl_flags))
+			clear_nlink(inode);
+	}
+	generic_drop_inode(inode);
+}
+
 /**
  * gfs2_clear_inode - Deallocate an inode when VFS is done with it
  * @inode: The VFS inode
@@ -441,7 +466,7 @@ out_unlock:
 out_uninit:
 	gfs2_holder_uninit(&ip->i_iopen_gh);
 	gfs2_glock_dq_uninit(&gh);
-	if (error)
+	if (error && error != GLR_TRYFAILED)
 		fs_warn(sdp, "gfs2_delete_inode: %d\n", error);
 out:
 	truncate_inode_pages(&inode->i_data, 0);
@@ -481,6 +506,7 @@ const struct super_operations gfs2_super_ops = {
 	.statfs			= gfs2_statfs,
 	.remount_fs		= gfs2_remount_fs,
 	.clear_inode		= gfs2_clear_inode,
+	.drop_inode		= gfs2_drop_inode,
 	.show_options		= gfs2_show_options,
 };