Merge git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-fixes

* git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-fixes:
  GFS2: Resolve inode eviction and ail list interaction bug
  GFS2: Fix race during filesystem mount
  GFS2: force a log flush when invalidating the rindex glock

fs/gfs2/aops.c

@@ -1069,6 +1069,7 @@ int gfs2_releasepage(struct page *page, gfp_t gfp_mask)
 		return 0;
 
 	gfs2_log_lock(sdp);
+	spin_lock(&sdp->sd_ail_lock);
 	head = bh = page_buffers(page);
 	do {
 		if (atomic_read(&bh->b_count))
@@ -1080,6 +1081,7 @@ int gfs2_releasepage(struct page *page, gfp_t gfp_mask)
 			goto not_possible;
 		bh = bh->b_this_page;
 	} while(bh != head);
+	spin_unlock(&sdp->sd_ail_lock);
 	gfs2_log_unlock(sdp);
 
 	head = bh = page_buffers(page);
@@ -1112,6 +1114,7 @@ int gfs2_releasepage(struct page *page, gfp_t gfp_mask)
 	WARN_ON(buffer_dirty(bh));
 	WARN_ON(buffer_pinned(bh));
 cannot_release:
+	spin_unlock(&sdp->sd_ail_lock);
 	gfs2_log_unlock(sdp);
 	return 0;
 }

fs/gfs2/glops.c

@@ -47,10 +47,10 @@ static void __gfs2_ail_flush(struct gfs2_glock *gl)
 				bd_ail_gl_list);
 		bh = bd->bd_bh;
 		gfs2_remove_from_ail(bd);
-		spin_unlock(&sdp->sd_ail_lock);
-
 		bd->bd_bh = NULL;
 		bh->b_private = NULL;
+		spin_unlock(&sdp->sd_ail_lock);
+
 		bd->bd_blkno = bh->b_blocknr;
 		gfs2_log_lock(sdp);
 		gfs2_assert_withdraw(sdp, !buffer_busy(bh));
@@ -221,8 +221,10 @@ static void inode_go_inval(struct gfs2_glock *gl, int flags)
 		}
 	}
 
-	if (ip == GFS2_I(gl->gl_sbd->sd_rindex))
+	if (ip == GFS2_I(gl->gl_sbd->sd_rindex)) {
+		gfs2_log_flush(gl->gl_sbd, NULL);
 		gl->gl_sbd->sd_rindex_uptodate = 0;
+	}
 	if (ip && S_ISREG(ip->i_inode.i_mode))
 		truncate_inode_pages(ip->i_inode.i_mapping, 0);
 }

fs/gfs2/incore.h

@@ -17,6 +17,7 @@
 #include <linux/buffer_head.h>
 #include <linux/rcupdate.h>
 #include <linux/rculist_bl.h>
+#include <linux/completion.h>
 
 #define DIO_WAIT	0x00000010
 #define DIO_METADATA	0x00000020
@@ -546,6 +547,7 @@ struct gfs2_sbd {
 	struct gfs2_glock *sd_trans_gl;
 	wait_queue_head_t sd_glock_wait;
 	atomic_t sd_glock_disposal;
+	struct completion sd_locking_init;
 
 	/* Inode Stuff */
 

fs/gfs2/log.c

@@ -903,6 +903,7 @@ void gfs2_meta_syncfs(struct gfs2_sbd *sdp)
 		if (gfs2_ail1_empty(sdp))
 			break;
 	}
+	gfs2_log_flush(sdp, NULL);
 }
 
 static inline int gfs2_jrnl_flush_reqd(struct gfs2_sbd *sdp)

fs/gfs2/ops_fstype.c

@@ -72,6 +72,7 @@ static struct gfs2_sbd *init_sbd(struct super_block *sb)
 
 	init_waitqueue_head(&sdp->sd_glock_wait);
 	atomic_set(&sdp->sd_glock_disposal, 0);
+	init_completion(&sdp->sd_locking_init);
 	spin_lock_init(&sdp->sd_statfs_spin);
 
 	spin_lock_init(&sdp->sd_rindex_spin);
@@ -1017,11 +1018,13 @@ static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent)
 		fsname++;
 	if (lm->lm_mount == NULL) {
 		fs_info(sdp, "Now mounting FS...\n");
+		complete(&sdp->sd_locking_init);
 		return 0;
 	}
 	ret = lm->lm_mount(sdp, fsname);
 	if (ret == 0)
 		fs_info(sdp, "Joined cluster. Now mounting FS...\n");
+	complete(&sdp->sd_locking_init);
 	return ret;
 }
 

fs/gfs2/super.c

@@ -757,13 +757,17 @@ static int gfs2_write_inode(struct inode *inode, struct writeback_control *wbc)
 	struct timespec atime;
 	struct gfs2_dinode *di;
 	int ret = -EAGAIN;
+	int unlock_required = 0;
 
 	/* Skip timestamp update, if this is from a memalloc */
 	if (current->flags & PF_MEMALLOC)
 		goto do_flush;
-	ret = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
-	if (ret)
-		goto do_flush;
+	if (!gfs2_glock_is_locked_by_me(ip->i_gl)) {
+		ret = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
+		if (ret)
+			goto do_flush;
+		unlock_required = 1;
+	}
 	ret = gfs2_trans_begin(sdp, RES_DINODE, 0);
 	if (ret)
 		goto do_unlock;
@@ -780,7 +784,8 @@ static int gfs2_write_inode(struct inode *inode, struct writeback_control *wbc)
 	}
 	gfs2_trans_end(sdp);
 do_unlock:
-	gfs2_glock_dq_uninit(&gh);
+	if (unlock_required)
+		gfs2_glock_dq_uninit(&gh);
 do_flush:
 	if (wbc->sync_mode == WB_SYNC_ALL)
 		gfs2_log_flush(GFS2_SB(inode), ip->i_gl);
@@ -1427,7 +1432,20 @@ static int gfs2_dinode_dealloc(struct gfs2_inode *ip)
 	return error;
 }
 
-/*
+/**
+ * gfs2_evict_inode - Remove an inode from cache
+ * @inode: The inode to evict
+ *
+ * There are three cases to consider:
+ * 1. i_nlink == 0, we are final opener (and must deallocate)
+ * 2. i_nlink == 0, we are not the final opener (and cannot deallocate)
+ * 3. i_nlink > 0
+ *
+ * If the fs is read only, then we have to treat all cases as per #3
+ * since we are unable to do any deallocation. The inode will be
+ * deallocated by the next read/write node to attempt an allocation
+ * in the same resource group
+ *
  * We have to (at the moment) hold the inodes main lock to cover
  * the gap between unlocking the shared lock on the iopen lock and
  * taking the exclusive lock. I'd rather do a shared -> exclusive
@@ -1470,6 +1488,8 @@ static void gfs2_evict_inode(struct inode *inode)
 	if (error)
 		goto out_truncate;
 
+	/* Case 1 starts here */
+
 	if (S_ISDIR(inode->i_mode) &&
 	    (ip->i_diskflags & GFS2_DIF_EXHASH)) {
 		error = gfs2_dir_exhash_dealloc(ip);
@@ -1493,13 +1513,16 @@ static void gfs2_evict_inode(struct inode *inode)
 	goto out_unlock;
 
 out_truncate:
+	/* Case 2 starts here */
 	error = gfs2_trans_begin(sdp, 0, sdp->sd_jdesc->jd_blocks);
 	if (error)
 		goto out_unlock;
-	gfs2_final_release_pages(ip);
+	/* Needs to be done before glock release & also in a transaction */
+	truncate_inode_pages(&inode->i_data, 0);
 	gfs2_trans_end(sdp);
 
 out_unlock:
+	/* Error path for case 1 */
 	if (test_bit(HIF_HOLDER, &ip->i_iopen_gh.gh_iflags))
 		gfs2_glock_dq(&ip->i_iopen_gh);
 	gfs2_holder_uninit(&ip->i_iopen_gh);
@@ -1507,6 +1530,7 @@ static void gfs2_evict_inode(struct inode *inode)
 	if (error && error != GLR_TRYFAILED && error != -EROFS)
 		fs_warn(sdp, "gfs2_evict_inode: %d\n", error);
 out:
+	/* Case 3 starts here */
 	truncate_inode_pages(&inode->i_data, 0);
 	end_writeback(inode);
 

fs/gfs2/sys.c

@@ -338,6 +338,9 @@ static ssize_t lkfirst_store(struct gfs2_sbd *sdp, const char *buf, size_t len)
 	rv = sscanf(buf, "%u", &first);
 	if (rv != 1 || first > 1)
 		return -EINVAL;
+	rv = wait_for_completion_killable(&sdp->sd_locking_init);
+	if (rv)
+		return rv;
 	spin_lock(&sdp->sd_jindex_spin);
 	rv = -EBUSY;
 	if (test_bit(SDF_NOJOURNALID, &sdp->sd_flags) == 0)
@@ -414,7 +417,9 @@ static ssize_t jid_store(struct gfs2_sbd *sdp, const char *buf, size_t len)
 	rv = sscanf(buf, "%d", &jid);
 	if (rv != 1)
 		return -EINVAL;
-
+	rv = wait_for_completion_killable(&sdp->sd_locking_init);
+	if (rv)
+		return rv;
 	spin_lock(&sdp->sd_jindex_spin);
 	rv = -EINVAL;
 	if (sdp->sd_lockstruct.ls_ops->lm_mount == NULL)