[Cluster-devel] GFS2: Pre-pull patch posting (merge window)

[Cluster-devel] GFS2: Pre-pull patch posting (merge window)

[Steven Whitehouse]

This is one of the smallest collections of patches for the merge
window for some time. There are some clean ups relating to the
transaction code and the shrinker, which are mostly in preparation
for further development, but also make the code much easier to
follow in these areas.

There is a patch which allows the use of ->writepages even in the
default ordered write mode for all writebacks. This results in
sending larger i/os to the block layer, and a subsequent increase
in performance. It also reduces the number of different i/o paths
by one.

There is also a bug fix reinstating the withdraw ack system which
somehow got lost when the lock modules were merged into GFS2.

And thats all this time around,

Steve.

[Cluster-devel] [PATCH 01/10] GFS2: Separate LRU scanning from shrinker

[Steven Whitehouse]

This breaks out the LRU scanning function from the shrinker in
preparation for adding other callers to the LRU scanner.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index 992c5c0..3ad8fd3 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -1376,23 +1376,19 @@ void gfs2_glock_complete(struct gfs2_glock *gl, int ret)
 		gfs2_glock_put(gl);
 }
 
+/**
+ * gfs2_scan_glock_lru - Scan the LRU looking for locks to demote
+ * @nr: The number of entries to scan
+ *
+ */
 
-static int gfs2_shrink_glock_memory(struct shrinker *shrink,
-				    struct shrink_control *sc)
+static void gfs2_scan_glock_lru(int nr)
 {
 	struct gfs2_glock *gl;
 	int may_demote;
 	int nr_skipped = 0;
-	int nr = sc->nr_to_scan;
-	gfp_t gfp_mask = sc->gfp_mask;
 	LIST_HEAD(skipped);
 
-	if (nr == 0)
-		goto out;
-
-	if (!(gfp_mask & __GFP_FS))
-		return -1;
-
 	spin_lock(&lru_lock);
 	while(nr && !list_empty(&lru_list)) {
 		gl = list_entry(lru_list.next, struct gfs2_glock, gl_lru);
@@ -1425,7 +1421,17 @@ static int gfs2_shrink_glock_memory(struct shrinker *shrink,
 	list_splice(&skipped, &lru_list);
 	atomic_add(nr_skipped, &lru_count);
 	spin_unlock(&lru_lock);
-out:
+}
+
+static int gfs2_shrink_glock_memory(struct shrinker *shrink,
+				    struct shrink_control *sc)
+{
+	if (sc->nr_to_scan) {
+		if (!(sc->gfp_mask & __GFP_FS))
+			return -1;
+		gfs2_scan_glock_lru(sc->nr_to_scan);
+	}
+
 	return (atomic_read(&lru_count) / 100) * sysctl_vfs_cache_pressure;
 }
 
-- 
1.7.4

[Cluster-devel] [PATCH 02/10] GFS2: Merge revoke adding functions

[Steven Whitehouse]

This moves the lo_add function for revokes into trans.c, removing
a function call and making the code easier to read.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/lops.c b/fs/gfs2/lops.c
index 9ceccb1..9c80742 100644
--- a/fs/gfs2/lops.c
+++ b/fs/gfs2/lops.c
@@ -600,20 +600,6 @@ static void buf_lo_after_scan(struct gfs2_jdesc *jd, int error, int pass)
 	        jd->jd_jid, sdp->sd_replayed_blocks, sdp->sd_found_blocks);
 }
 
-static void revoke_lo_add(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd)
-{
-	struct gfs2_glock *gl = bd->bd_gl;
-	struct gfs2_trans *tr;
-
-	tr = current->journal_info;
-	tr->tr_touched = 1;
-	tr->tr_num_revoke++;
-	sdp->sd_log_num_revoke++;
-	atomic_inc(&gl->gl_revokes);
-	set_bit(GLF_LFLUSH, &gl->gl_flags);
-	list_add(&bd->bd_list, &sdp->sd_log_le_revoke);
-}
-
 static void revoke_lo_before_commit(struct gfs2_sbd *sdp)
 {
 	struct gfs2_meta_header *mh;
@@ -895,7 +881,6 @@ const struct gfs2_log_operations gfs2_buf_lops = {
 };
 
 const struct gfs2_log_operations gfs2_revoke_lops = {
-	.lo_add = revoke_lo_add,
 	.lo_before_commit = revoke_lo_before_commit,
 	.lo_after_commit = revoke_lo_after_commit,
 	.lo_before_scan = revoke_lo_before_scan,
diff --git a/fs/gfs2/trans.c b/fs/gfs2/trans.c
index 4136270..6f3ddbc 100644
--- a/fs/gfs2/trans.c
+++ b/fs/gfs2/trans.c
@@ -175,11 +175,19 @@ void gfs2_trans_add_bh(struct gfs2_glock *gl, struct buffer_head *bh, int meta)
 
 void gfs2_trans_add_revoke(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd)
 {
+	struct gfs2_glock *gl = bd->bd_gl;
+	struct gfs2_trans *tr = current->journal_info;
+
 	BUG_ON(!list_empty(&bd->bd_list));
 	BUG_ON(!list_empty(&bd->bd_ail_st_list));
 	BUG_ON(!list_empty(&bd->bd_ail_gl_list));
 	lops_init_le(bd, &gfs2_revoke_lops);
-	lops_add(sdp, bd);
+	tr->tr_touched = 1;
+	tr->tr_num_revoke++;
+	sdp->sd_log_num_revoke++;
+	atomic_inc(&gl->gl_revokes);
+	set_bit(GLF_LFLUSH, &gl->gl_flags);
+	list_add(&bd->bd_list, &sdp->sd_log_le_revoke);
 }
 
 void gfs2_trans_add_unrevoke(struct gfs2_sbd *sdp, u64 blkno, unsigned int len)
-- 
1.7.4

[Cluster-devel] [PATCH 03/10] GFS2: Split gfs2_trans_add_bh() into two

[Steven Whitehouse]

There is little common content in gfs2_trans_add_bh() between the data
and meta classes by the time that the functions which it calls are
taken into account. The intent here is to split this into two
separate functions. Stage one is to introduce gfs2_trans_add_data()
and gfs2_trans_add_meta() and update the callers accordingly.

Later patches will then pull in the content of gfs2_trans_add_bh()
and its dependent functions in order to clean up the code in this
area.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/aops.c b/fs/gfs2/aops.c
index 30de4f2..92340dd 100644
--- a/fs/gfs2/aops.c
+++ b/fs/gfs2/aops.c
@@ -51,7 +51,7 @@ static void gfs2_page_add_databufs(struct gfs2_inode *ip, struct page *page,
 			continue;
 		if (gfs2_is_jdata(ip))
 			set_buffer_uptodate(bh);
-		gfs2_trans_add_bh(ip->i_gl, bh, 0);
+		gfs2_trans_add_data(ip->i_gl, bh);
 	}
 }
 
@@ -852,7 +852,7 @@ static int gfs2_write_end(struct file *file, struct address_space *mapping,
 		goto failed;
 	}
 
-	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
+	gfs2_trans_add_meta(ip->i_gl, dibh);
 
 	if (gfs2_is_stuffed(ip))
 		return gfs2_stuffed_write_end(inode, dibh, pos, len, copied, page);
diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index a68e91b..7a86275 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -93,7 +93,7 @@ static int gfs2_unstuffer_page(struct gfs2_inode *ip, struct buffer_head *dibh,
 	if (!gfs2_is_jdata(ip))
 		mark_buffer_dirty(bh);
 	if (!gfs2_is_writeback(ip))
-		gfs2_trans_add_bh(ip->i_gl, bh, 0);
+		gfs2_trans_add_data(ip->i_gl, bh);
 
 	if (release) {
 		unlock_page(page);
@@ -153,7 +153,7 @@ int gfs2_unstuff_dinode(struct gfs2_inode *ip, struct page *page)
 
 	/*  Set up the pointer to the new block  */
 
-	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
+	gfs2_trans_add_meta(ip->i_gl, dibh);
 	di = (struct gfs2_dinode *)dibh->b_data;
 	gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
 
@@ -405,7 +405,7 @@ static inline __be64 *gfs2_indirect_init(struct metapath *mp,
 	BUG_ON(i < 1);
 	BUG_ON(mp->mp_bh[i] != NULL);
 	mp->mp_bh[i] = gfs2_meta_new(gl, bn);
-	gfs2_trans_add_bh(gl, mp->mp_bh[i], 1);
+	gfs2_trans_add_meta(gl, mp->mp_bh[i]);
 	gfs2_metatype_set(mp->mp_bh[i], GFS2_METATYPE_IN, GFS2_FORMAT_IN);
 	gfs2_buffer_clear_tail(mp->mp_bh[i], sizeof(struct gfs2_meta_header));
 	ptr += offset;
@@ -468,7 +468,7 @@ static int gfs2_bmap_alloc(struct inode *inode, const sector_t lblock,
 	BUG_ON(sheight < 1);
 	BUG_ON(dibh == NULL);
 
-	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
+	gfs2_trans_add_meta(ip->i_gl, dibh);
 
 	if (height == sheight) {
 		struct buffer_head *bh;
@@ -544,7 +544,7 @@ static int gfs2_bmap_alloc(struct inode *inode, const sector_t lblock,
 		/* Branching from existing tree */
 		case ALLOC_GROW_DEPTH:
 			if (i > 1 && i < height)
-				gfs2_trans_add_bh(ip->i_gl, mp->mp_bh[i-1], 1);
+				gfs2_trans_add_meta(ip->i_gl, mp->mp_bh[i-1]);
 			for (; i < height && n > 0; i++, n--)
 				gfs2_indirect_init(mp, ip->i_gl, i,
 						   mp->mp_list[i-1], bn++);
@@ -556,7 +556,7 @@ static int gfs2_bmap_alloc(struct inode *inode, const sector_t lblock,
 		case ALLOC_DATA:
 			BUG_ON(n > dblks);
 			BUG_ON(mp->mp_bh[end_of_metadata] == NULL);
-			gfs2_trans_add_bh(ip->i_gl, mp->mp_bh[end_of_metadata], 1);
+			gfs2_trans_add_meta(ip->i_gl, mp->mp_bh[end_of_metadata]);
 			dblks = n;
 			ptr = metapointer(end_of_metadata, mp);
 			dblock = bn;
@@ -796,8 +796,8 @@ static int do_strip(struct gfs2_inode *ip, struct buffer_head *dibh,
 
 	down_write(&ip->i_rw_mutex);
 
-	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
-	gfs2_trans_add_bh(ip->i_gl, bh, 1);
+	gfs2_trans_add_meta(ip->i_gl, dibh);
+	gfs2_trans_add_meta(ip->i_gl, bh);
 
 	bstart = 0;
 	blen = 0;
@@ -981,7 +981,7 @@ static int gfs2_block_truncate_page(struct address_space *mapping, loff_t from)
 	}
 
 	if (!gfs2_is_writeback(ip))
-		gfs2_trans_add_bh(ip->i_gl, bh, 0);
+		gfs2_trans_add_data(ip->i_gl, bh);
 
 	zero_user(page, offset, length);
 	mark_buffer_dirty(bh);
@@ -1046,7 +1046,7 @@ static int trunc_start(struct inode *inode, u64 oldsize, u64 newsize)
 	if (error)
 		goto out;
 
-	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
+	gfs2_trans_add_meta(ip->i_gl, dibh);
 
 	if (gfs2_is_stuffed(ip)) {
 		gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode) + newsize);
@@ -1141,7 +1141,7 @@ static int trunc_end(struct gfs2_inode *ip)
 	ip->i_inode.i_mtime = ip->i_inode.i_ctime = CURRENT_TIME;
 	ip->i_diskflags &= ~GFS2_DIF_TRUNC_IN_PROG;
 
-	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
+	gfs2_trans_add_meta(ip->i_gl, dibh);
 	gfs2_dinode_out(ip, dibh->b_data);
 	brelse(dibh);
 
@@ -1246,7 +1246,7 @@ static int do_grow(struct inode *inode, u64 size)
 
 	i_size_write(inode, size);
 	ip->i_inode.i_mtime = ip->i_inode.i_ctime = CURRENT_TIME;
-	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
+	gfs2_trans_add_meta(ip->i_gl, dibh);
 	gfs2_dinode_out(ip, dibh->b_data);
 	brelse(dibh);
 
diff --git a/fs/gfs2/dir.c b/fs/gfs2/dir.c
index 9a35670..7179478 100644
--- a/fs/gfs2/dir.c
+++ b/fs/gfs2/dir.c
@@ -93,7 +93,7 @@ int gfs2_dir_get_new_buffer(struct gfs2_inode *ip, u64 block,
 	struct buffer_head *bh;
 
 	bh = gfs2_meta_new(ip->i_gl, block);
-	gfs2_trans_add_bh(ip->i_gl, bh, 1);
+	gfs2_trans_add_meta(ip->i_gl, bh);
 	gfs2_metatype_set(bh, GFS2_METATYPE_JD, GFS2_FORMAT_JD);
 	gfs2_buffer_clear_tail(bh, sizeof(struct gfs2_meta_header));
 	*bhp = bh;
@@ -127,7 +127,7 @@ static int gfs2_dir_write_stuffed(struct gfs2_inode *ip, const char *buf,
 	if (error)
 		return error;
 
-	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
+	gfs2_trans_add_meta(ip->i_gl, dibh);
 	memcpy(dibh->b_data + offset + sizeof(struct gfs2_dinode), buf, size);
 	if (ip->i_inode.i_size < offset + size)
 		i_size_write(&ip->i_inode, offset + size);
@@ -209,7 +209,7 @@ static int gfs2_dir_write_data(struct gfs2_inode *ip, const char *buf,
 		if (error)
 			goto fail;
 
-		gfs2_trans_add_bh(ip->i_gl, bh, 1);
+		gfs2_trans_add_meta(ip->i_gl, bh);
 		memcpy(bh->b_data + o, buf, amount);
 		brelse(bh);
 
@@ -231,7 +231,7 @@ out:
 		i_size_write(&ip->i_inode, offset + copied);
 	ip->i_inode.i_mtime = ip->i_inode.i_ctime = CURRENT_TIME;
 
-	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
+	gfs2_trans_add_meta(ip->i_gl, dibh);
 	gfs2_dinode_out(ip, dibh->b_data);
 	brelse(dibh);
 
@@ -647,7 +647,7 @@ static void dirent_del(struct gfs2_inode *dip, struct buffer_head *bh,
 		return;
 	}
 
-	gfs2_trans_add_bh(dip->i_gl, bh, 1);
+	gfs2_trans_add_meta(dip->i_gl, bh);
 
 	/* If there is no prev entry, this is the first entry in the block.
 	   The de_rec_len is already as big as it needs to be.  Just zero
@@ -690,7 +690,7 @@ static struct gfs2_dirent *gfs2_init_dirent(struct inode *inode,
 		offset = GFS2_DIRENT_SIZE(be16_to_cpu(dent->de_name_len));
 	totlen = be16_to_cpu(dent->de_rec_len);
 	BUG_ON(offset + name->len > totlen);
-	gfs2_trans_add_bh(ip->i_gl, bh, 1);
+	gfs2_trans_add_meta(ip->i_gl, bh);
 	ndent = (struct gfs2_dirent *)((char *)dent + offset);
 	dent->de_rec_len = cpu_to_be16(offset);
 	gfs2_qstr2dirent(name, totlen - offset, ndent);
@@ -831,7 +831,7 @@ static struct gfs2_leaf *new_leaf(struct inode *inode, struct buffer_head **pbh,
 		return NULL;
 
 	gfs2_trans_add_unrevoke(GFS2_SB(inode), bn, 1);
-	gfs2_trans_add_bh(ip->i_gl, bh, 1);
+	gfs2_trans_add_meta(ip->i_gl, bh);
 	gfs2_metatype_set(bh, GFS2_METATYPE_LF, GFS2_FORMAT_LF);
 	leaf = (struct gfs2_leaf *)bh->b_data;
 	leaf->lf_depth = cpu_to_be16(depth);
@@ -916,7 +916,7 @@ static int dir_make_exhash(struct inode *inode)
 	/*  We're done with the new leaf block, now setup the new
 	    hash table.  */
 
-	gfs2_trans_add_bh(dip->i_gl, dibh, 1);
+	gfs2_trans_add_meta(dip->i_gl, dibh);
 	gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
 
 	lp = (__be64 *)(dibh->b_data + sizeof(struct gfs2_dinode));
@@ -976,7 +976,7 @@ static int dir_split_leaf(struct inode *inode, const struct qstr *name)
 		return 1; /* can't split */
 	}
 
-	gfs2_trans_add_bh(dip->i_gl, obh, 1);
+	gfs2_trans_add_meta(dip->i_gl, obh);
 
 	nleaf = new_leaf(inode, &nbh, be16_to_cpu(oleaf->lf_depth) + 1);
 	if (!nleaf) {
@@ -1069,7 +1069,7 @@ static int dir_split_leaf(struct inode *inode, const struct qstr *name)
 
 	error = gfs2_meta_inode_buffer(dip, &dibh);
 	if (!gfs2_assert_withdraw(GFS2_SB(&dip->i_inode), !error)) {
-		gfs2_trans_add_bh(dip->i_gl, dibh, 1);
+		gfs2_trans_add_meta(dip->i_gl, dibh);
 		gfs2_add_inode_blocks(&dip->i_inode, 1);
 		gfs2_dinode_out(dip, dibh->b_data);
 		brelse(dibh);
@@ -1622,7 +1622,7 @@ static int dir_new_leaf(struct inode *inode, const struct qstr *name)
 			return error;
 	} while(1);
 
-	gfs2_trans_add_bh(ip->i_gl, obh, 1);
+	gfs2_trans_add_meta(ip->i_gl, obh);
 
 	leaf = new_leaf(inode, &bh, be16_to_cpu(oleaf->lf_depth));
 	if (!leaf) {
@@ -1636,7 +1636,7 @@ static int dir_new_leaf(struct inode *inode, const struct qstr *name)
 	error = gfs2_meta_inode_buffer(ip, &bh);
 	if (error)
 		return error;
-	gfs2_trans_add_bh(ip->i_gl, bh, 1);
+	gfs2_trans_add_meta(ip->i_gl, bh);
 	gfs2_add_inode_blocks(&ip->i_inode, 1);
 	gfs2_dinode_out(ip, bh->b_data);
 	brelse(bh);
@@ -1795,7 +1795,7 @@ int gfs2_dir_mvino(struct gfs2_inode *dip, const struct qstr *filename,
 	if (IS_ERR(dent))
 		return PTR_ERR(dent);
 
-	gfs2_trans_add_bh(dip->i_gl, bh, 1);
+	gfs2_trans_add_meta(dip->i_gl, bh);
 	gfs2_inum_out(nip, dent);
 	dent->de_type = cpu_to_be16(new_type);
 
@@ -1804,7 +1804,7 @@ int gfs2_dir_mvino(struct gfs2_inode *dip, const struct qstr *filename,
 		error = gfs2_meta_inode_buffer(dip, &bh);
 		if (error)
 			return error;
-		gfs2_trans_add_bh(dip->i_gl, bh, 1);
+		gfs2_trans_add_meta(dip->i_gl, bh);
 	}
 
 	dip->i_inode.i_mtime = dip->i_inode.i_ctime = CURRENT_TIME;
@@ -1917,7 +1917,7 @@ static int leaf_dealloc(struct gfs2_inode *dip, u32 index, u32 len,
 	if (error)
 		goto out_end_trans;
 
-	gfs2_trans_add_bh(dip->i_gl, dibh, 1);
+	gfs2_trans_add_meta(dip->i_gl, dibh);
 	/* On the last dealloc, make this a regular file in case we crash.
 	   (We don't want to free these blocks a second time.)  */
 	if (last_dealloc)
diff --git a/fs/gfs2/file.c b/fs/gfs2/file.c
index 991ab2d..06b7092 100644
--- a/fs/gfs2/file.c
+++ b/fs/gfs2/file.c
@@ -276,7 +276,7 @@ static int do_gfs2_set_flags(struct file *filp, u32 reqflags, u32 mask)
 	error = gfs2_meta_inode_buffer(ip, &bh);
 	if (error)
 		goto out_trans_end;
-	gfs2_trans_add_bh(ip->i_gl, bh, 1);
+	gfs2_trans_add_meta(ip->i_gl, bh);
 	ip->i_diskflags = new_flags;
 	gfs2_dinode_out(ip, bh->b_data);
 	brelse(bh);
@@ -709,7 +709,7 @@ static int fallocate_chunk(struct inode *inode, loff_t offset, loff_t len,
 	if (unlikely(error))
 		return error;
 
-	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
+	gfs2_trans_add_meta(ip->i_gl, dibh);
 
 	if (gfs2_is_stuffed(ip)) {
 		error = gfs2_unstuff_dinode(ip, NULL);
diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index 2b6f569..db048a8 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -447,7 +447,7 @@ static void init_dinode(struct gfs2_inode *dip, struct gfs2_inode *ip,
 	struct timespec tv = CURRENT_TIME;
 
 	dibh = gfs2_meta_new(ip->i_gl, ip->i_no_addr);
-	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
+	gfs2_trans_add_meta(ip->i_gl, dibh);
 	gfs2_metatype_set(dibh, GFS2_METATYPE_DI, GFS2_FORMAT_DI);
 	gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
 	di = (struct gfs2_dinode *)dibh->b_data;
@@ -584,7 +584,7 @@ static int link_dinode(struct gfs2_inode *dip, const struct qstr *name,
 	if (error)
 		goto fail_end_trans;
 	set_nlink(&ip->i_inode, S_ISDIR(ip->i_inode.i_mode) ? 2 : 1);
-	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
+	gfs2_trans_add_meta(ip->i_gl, dibh);
 	gfs2_dinode_out(ip, dibh->b_data);
 	brelse(dibh);
 	return 0;
@@ -931,7 +931,7 @@ static int gfs2_link(struct dentry *old_dentry, struct inode *dir,
 	if (error)
 		goto out_brelse;
 
-	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
+	gfs2_trans_add_meta(ip->i_gl, dibh);
 	inc_nlink(&ip->i_inode);
 	ip->i_inode.i_ctime = CURRENT_TIME;
 	ihold(inode);
@@ -1412,7 +1412,7 @@ static int gfs2_rename(struct inode *odir, struct dentry *odentry,
 		if (error)
 			goto out_end_trans;
 		ip->i_inode.i_ctime = CURRENT_TIME;
-		gfs2_trans_add_bh(ip->i_gl, dibh, 1);
+		gfs2_trans_add_meta(ip->i_gl, dibh);
 		gfs2_dinode_out(ip, dibh->b_data);
 		brelse(dibh);
 	}
diff --git a/fs/gfs2/quota.c b/fs/gfs2/quota.c
index ae55e24..06122d0 100644
--- a/fs/gfs2/quota.c
+++ b/fs/gfs2/quota.c
@@ -590,7 +590,7 @@ static void do_qc(struct gfs2_quota_data *qd, s64 change)
 	s64 x;
 
 	mutex_lock(&sdp->sd_quota_mutex);
-	gfs2_trans_add_bh(ip->i_gl, qd->qd_bh, 1);
+	gfs2_trans_add_meta(ip->i_gl, qd->qd_bh);
 
 	if (!test_bit(QDF_CHANGE, &qd->qd_flags)) {
 		qc->qc_change = 0;
@@ -726,7 +726,7 @@ get_a_page:
 			goto unlock_out;
 	}
 
-	gfs2_trans_add_bh(ip->i_gl, bh, 0);
+	gfs2_trans_add_meta(ip->i_gl, bh);
 
 	kaddr = kmap_atomic(page);
 	if (offset + sizeof(struct gfs2_quota) > PAGE_CACHE_SIZE)
diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index b7eff07..52c2aea 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -1323,7 +1323,7 @@ int gfs2_fitrim(struct file *filp, void __user *argp)
 			if (ret == 0) {
 				bh = rgd->rd_bits[0].bi_bh;
 				rgd->rd_flags |= GFS2_RGF_TRIMMED;
-				gfs2_trans_add_bh(rgd->rd_gl, bh, 1);
+				gfs2_trans_add_meta(rgd->rd_gl, bh);
 				gfs2_rgrp_out(rgd, bh->b_data);
 				gfs2_rgrp_ondisk2lvb(rgd->rd_rgl, bh->b_data);
 				gfs2_trans_end(sdp);
@@ -1968,14 +1968,14 @@ static void gfs2_alloc_extent(const struct gfs2_rbm *rbm, bool dinode,
 
 	*n = 1;
 	block = gfs2_rbm_to_block(rbm);
-	gfs2_trans_add_bh(rbm->rgd->rd_gl, rbm->bi->bi_bh, 1);
+	gfs2_trans_add_meta(rbm->rgd->rd_gl, rbm->bi->bi_bh);
 	gfs2_setbit(rbm, true, dinode ? GFS2_BLKST_DINODE : GFS2_BLKST_USED);
 	block++;
 	while (*n < elen) {
 		ret = gfs2_rbm_from_block(&pos, block);
 		if (ret || gfs2_testbit(&pos) != GFS2_BLKST_FREE)
 			break;
-		gfs2_trans_add_bh(pos.rgd->rd_gl, pos.bi->bi_bh, 1);
+		gfs2_trans_add_meta(pos.rgd->rd_gl, pos.bi->bi_bh);
 		gfs2_setbit(&pos, true, GFS2_BLKST_USED);
 		(*n)++;
 		block++;
@@ -2014,7 +2014,7 @@ static struct gfs2_rgrpd *rgblk_free(struct gfs2_sbd *sdp, u64 bstart,
 			       rbm.bi->bi_bh->b_data + rbm.bi->bi_offset,
 			       rbm.bi->bi_len);
 		}
-		gfs2_trans_add_bh(rbm.rgd->rd_gl, rbm.bi->bi_bh, 1);
+		gfs2_trans_add_meta(rbm.rgd->rd_gl, rbm.bi->bi_bh);
 		gfs2_setbit(&rbm, false, new_state);
 	}
 
@@ -2157,7 +2157,7 @@ int gfs2_alloc_blocks(struct gfs2_inode *ip, u64 *bn, unsigned int *nblocks,
 		if (error == 0) {
 			struct gfs2_dinode *di =
 				(struct gfs2_dinode *)dibh->b_data;
-			gfs2_trans_add_bh(ip->i_gl, dibh, 1);
+			gfs2_trans_add_meta(ip->i_gl, dibh);
 			di->di_goal_meta = di->di_goal_data =
 				cpu_to_be64(ip->i_goal);
 			brelse(dibh);
@@ -2176,7 +2176,7 @@ int gfs2_alloc_blocks(struct gfs2_inode *ip, u64 *bn, unsigned int *nblocks,
 			*generation = rbm.rgd->rd_igeneration++;
 	}
 
-	gfs2_trans_add_bh(rbm.rgd->rd_gl, rbm.rgd->rd_bits[0].bi_bh, 1);
+	gfs2_trans_add_meta(rbm.rgd->rd_gl, rbm.rgd->rd_bits[0].bi_bh);
 	gfs2_rgrp_out(rbm.rgd, rbm.rgd->rd_bits[0].bi_bh->b_data);
 	gfs2_rgrp_ondisk2lvb(rbm.rgd->rd_rgl, rbm.rgd->rd_bits[0].bi_bh->b_data);
 
@@ -2223,7 +2223,7 @@ void __gfs2_free_blocks(struct gfs2_inode *ip, u64 bstart, u32 blen, int meta)
 	trace_gfs2_block_alloc(ip, rgd, bstart, blen, GFS2_BLKST_FREE);
 	rgd->rd_free += blen;
 	rgd->rd_flags &= ~GFS2_RGF_TRIMMED;
-	gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1);
+	gfs2_trans_add_meta(rgd->rd_gl, rgd->rd_bits[0].bi_bh);
 	gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data);
 	gfs2_rgrp_ondisk2lvb(rgd->rd_rgl, rgd->rd_bits[0].bi_bh->b_data);
 
@@ -2260,7 +2260,7 @@ void gfs2_unlink_di(struct inode *inode)
 	if (!rgd)
 		return;
 	trace_gfs2_block_alloc(ip, rgd, blkno, 1, GFS2_BLKST_UNLINKED);
-	gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1);
+	gfs2_trans_add_meta(rgd->rd_gl, rgd->rd_bits[0].bi_bh);
 	gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data);
 	gfs2_rgrp_ondisk2lvb(rgd->rd_rgl, rgd->rd_bits[0].bi_bh->b_data);
 	update_rgrp_lvb_unlinked(rgd, 1);
@@ -2281,7 +2281,7 @@ static void gfs2_free_uninit_di(struct gfs2_rgrpd *rgd, u64 blkno)
 	rgd->rd_dinodes--;
 	rgd->rd_free++;
 
-	gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1);
+	gfs2_trans_add_meta(rgd->rd_gl, rgd->rd_bits[0].bi_bh);
 	gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data);
 	gfs2_rgrp_ondisk2lvb(rgd->rd_rgl, rgd->rd_bits[0].bi_bh->b_data);
 	update_rgrp_lvb_unlinked(rgd, -1);
diff --git a/fs/gfs2/super.c b/fs/gfs2/super.c
index d648867..4dfda4c 100644
--- a/fs/gfs2/super.c
+++ b/fs/gfs2/super.c
@@ -500,7 +500,7 @@ void gfs2_statfs_change(struct gfs2_sbd *sdp, s64 total, s64 free,
 	if (error)
 		return;
 
-	gfs2_trans_add_bh(l_ip->i_gl, l_bh, 1);
+	gfs2_trans_add_meta(l_ip->i_gl, l_bh);
 
 	spin_lock(&sdp->sd_statfs_spin);
 	l_sc->sc_total += total;
@@ -528,7 +528,7 @@ void update_statfs(struct gfs2_sbd *sdp, struct buffer_head *m_bh,
 	struct gfs2_statfs_change_host *m_sc = &sdp->sd_statfs_master;
 	struct gfs2_statfs_change_host *l_sc = &sdp->sd_statfs_local;
 
-	gfs2_trans_add_bh(l_ip->i_gl, l_bh, 1);
+	gfs2_trans_add_meta(l_ip->i_gl, l_bh);
 
 	spin_lock(&sdp->sd_statfs_spin);
 	m_sc->sc_total += l_sc->sc_total;
@@ -539,7 +539,7 @@ void update_statfs(struct gfs2_sbd *sdp, struct buffer_head *m_bh,
 	       0, sizeof(struct gfs2_statfs_change));
 	spin_unlock(&sdp->sd_statfs_spin);
 
-	gfs2_trans_add_bh(m_ip->i_gl, m_bh, 1);
+	gfs2_trans_add_meta(m_ip->i_gl, m_bh);
 	gfs2_statfs_change_out(m_sc, m_bh->b_data + sizeof(struct gfs2_dinode));
 }
 
@@ -824,7 +824,7 @@ static void gfs2_dirty_inode(struct inode *inode, int flags)
 
 	ret = gfs2_meta_inode_buffer(ip, &bh);
 	if (ret == 0) {
-		gfs2_trans_add_bh(ip->i_gl, bh, 1);
+		gfs2_trans_add_meta(ip->i_gl, bh);
 		gfs2_dinode_out(ip, bh->b_data);
 		brelse(bh);
 	}
diff --git a/fs/gfs2/trans.c b/fs/gfs2/trans.c
index 6f3ddbc..cb795ac 100644
--- a/fs/gfs2/trans.c
+++ b/fs/gfs2/trans.c
@@ -150,7 +150,7 @@ void gfs2_trans_end(struct gfs2_sbd *sdp)
  *
  */
 
-void gfs2_trans_add_bh(struct gfs2_glock *gl, struct buffer_head *bh, int meta)
+static void gfs2_trans_add_bh(struct gfs2_glock *gl, struct buffer_head *bh, int meta)
 {
 	struct gfs2_sbd *sdp = gl->gl_sbd;
 	struct gfs2_bufdata *bd;
@@ -173,6 +173,16 @@ void gfs2_trans_add_bh(struct gfs2_glock *gl, struct buffer_head *bh, int meta)
 	unlock_buffer(bh);
 }
 
+void gfs2_trans_add_data(struct gfs2_glock *gl, struct buffer_head *bh)
+{
+	return gfs2_trans_add_bh(gl, bh, 0);
+}
+
+void gfs2_trans_add_meta(struct gfs2_glock *gl, struct buffer_head *bh)
+{
+	return gfs2_trans_add_bh(gl, bh, 1);
+}
+
 void gfs2_trans_add_revoke(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd)
 {
 	struct gfs2_glock *gl = bd->bd_gl;
diff --git a/fs/gfs2/trans.h b/fs/gfs2/trans.h
index bf2ae9a..1e6e7da 100644
--- a/fs/gfs2/trans.h
+++ b/fs/gfs2/trans.h
@@ -39,7 +39,8 @@ extern int gfs2_trans_begin(struct gfs2_sbd *sdp, unsigned int blocks,
 			    unsigned int revokes);
 
 extern void gfs2_trans_end(struct gfs2_sbd *sdp);
-extern void gfs2_trans_add_bh(struct gfs2_glock *gl, struct buffer_head *bh, int meta);
+extern void gfs2_trans_add_data(struct gfs2_glock *gl, struct buffer_head *bh);
+extern void gfs2_trans_add_meta(struct gfs2_glock *gl, struct buffer_head *bh);
 extern void gfs2_trans_add_revoke(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd);
 extern void gfs2_trans_add_unrevoke(struct gfs2_sbd *sdp, u64 blkno, unsigned int len);
 
diff --git a/fs/gfs2/xattr.c b/fs/gfs2/xattr.c
index 76c144b..cbb46c2 100644
--- a/fs/gfs2/xattr.c
+++ b/fs/gfs2/xattr.c
@@ -270,7 +270,7 @@ static int ea_dealloc_unstuffed(struct gfs2_inode *ip, struct buffer_head *bh,
 	if (error)
 		goto out_gunlock;
 
-	gfs2_trans_add_bh(ip->i_gl, bh, 1);
+	gfs2_trans_add_meta(ip->i_gl, bh);
 
 	dataptrs = GFS2_EA2DATAPTRS(ea);
 	for (x = 0; x < ea->ea_num_ptrs; x++, dataptrs++) {
@@ -309,7 +309,7 @@ static int ea_dealloc_unstuffed(struct gfs2_inode *ip, struct buffer_head *bh,
 	error = gfs2_meta_inode_buffer(ip, &dibh);
 	if (!error) {
 		ip->i_inode.i_ctime = CURRENT_TIME;
-		gfs2_trans_add_bh(ip->i_gl, dibh, 1);
+		gfs2_trans_add_meta(ip->i_gl, dibh);
 		gfs2_dinode_out(ip, dibh->b_data);
 		brelse(dibh);
 	}
@@ -509,7 +509,7 @@ static int gfs2_iter_unstuffed(struct gfs2_inode *ip, struct gfs2_ea_header *ea,
 		}
 
 		if (din) {
-			gfs2_trans_add_bh(ip->i_gl, bh[x], 1);
+			gfs2_trans_add_meta(ip->i_gl, bh[x]);
 			memcpy(pos, din, cp_size);
 			din += sdp->sd_jbsize;
 		}
@@ -629,7 +629,7 @@ static int ea_alloc_blk(struct gfs2_inode *ip, struct buffer_head **bhp)
 		return error;
 	gfs2_trans_add_unrevoke(sdp, block, 1);
 	*bhp = gfs2_meta_new(ip->i_gl, block);
-	gfs2_trans_add_bh(ip->i_gl, *bhp, 1);
+	gfs2_trans_add_meta(ip->i_gl, *bhp);
 	gfs2_metatype_set(*bhp, GFS2_METATYPE_EA, GFS2_FORMAT_EA);
 	gfs2_buffer_clear_tail(*bhp, sizeof(struct gfs2_meta_header));
 
@@ -691,7 +691,7 @@ static int ea_write(struct gfs2_inode *ip, struct gfs2_ea_header *ea,
 				return error;
 			gfs2_trans_add_unrevoke(sdp, block, 1);
 			bh = gfs2_meta_new(ip->i_gl, block);
-			gfs2_trans_add_bh(ip->i_gl, bh, 1);
+			gfs2_trans_add_meta(ip->i_gl, bh);
 			gfs2_metatype_set(bh, GFS2_METATYPE_ED, GFS2_FORMAT_ED);
 
 			gfs2_add_inode_blocks(&ip->i_inode, 1);
@@ -751,7 +751,7 @@ static int ea_alloc_skeleton(struct gfs2_inode *ip, struct gfs2_ea_request *er,
 	error = gfs2_meta_inode_buffer(ip, &dibh);
 	if (!error) {
 		ip->i_inode.i_ctime = CURRENT_TIME;
-		gfs2_trans_add_bh(ip->i_gl, dibh, 1);
+		gfs2_trans_add_meta(ip->i_gl, dibh);
 		gfs2_dinode_out(ip, dibh->b_data);
 		brelse(dibh);
 	}
@@ -834,7 +834,7 @@ static void ea_set_remove_stuffed(struct gfs2_inode *ip,
 	struct gfs2_ea_header *prev = el->el_prev;
 	u32 len;
 
-	gfs2_trans_add_bh(ip->i_gl, el->el_bh, 1);
+	gfs2_trans_add_meta(ip->i_gl, el->el_bh);
 
 	if (!prev || !GFS2_EA_IS_STUFFED(ea)) {
 		ea->ea_type = GFS2_EATYPE_UNUSED;
@@ -872,7 +872,7 @@ static int ea_set_simple_noalloc(struct gfs2_inode *ip, struct buffer_head *bh,
 	if (error)
 		return error;
 
-	gfs2_trans_add_bh(ip->i_gl, bh, 1);
+	gfs2_trans_add_meta(ip->i_gl, bh);
 
 	if (es->ea_split)
 		ea = ea_split_ea(ea);
@@ -886,7 +886,7 @@ static int ea_set_simple_noalloc(struct gfs2_inode *ip, struct buffer_head *bh,
 	if (error)
 		goto out;
 	ip->i_inode.i_ctime = CURRENT_TIME;
-	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
+	gfs2_trans_add_meta(ip->i_gl, dibh);
 	gfs2_dinode_out(ip, dibh->b_data);
 	brelse(dibh);
 out:
@@ -901,7 +901,7 @@ static int ea_set_simple_alloc(struct gfs2_inode *ip,
 	struct gfs2_ea_header *ea = es->es_ea;
 	int error;
 
-	gfs2_trans_add_bh(ip->i_gl, es->es_bh, 1);
+	gfs2_trans_add_meta(ip->i_gl, es->es_bh);
 
 	if (es->ea_split)
 		ea = ea_split_ea(ea);
@@ -997,7 +997,7 @@ static int ea_set_block(struct gfs2_inode *ip, struct gfs2_ea_request *er,
 			goto out;
 		}
 
-		gfs2_trans_add_bh(ip->i_gl, indbh, 1);
+		gfs2_trans_add_meta(ip->i_gl, indbh);
 	} else {
 		u64 blk;
 		unsigned int n = 1;
@@ -1006,7 +1006,7 @@ static int ea_set_block(struct gfs2_inode *ip, struct gfs2_ea_request *er,
 			return error;
 		gfs2_trans_add_unrevoke(sdp, blk, 1);
 		indbh = gfs2_meta_new(ip->i_gl, blk);
-		gfs2_trans_add_bh(ip->i_gl, indbh, 1);
+		gfs2_trans_add_meta(ip->i_gl, indbh);
 		gfs2_metatype_set(indbh, GFS2_METATYPE_IN, GFS2_FORMAT_IN);
 		gfs2_buffer_clear_tail(indbh, mh_size);
 
@@ -1092,7 +1092,7 @@ static int ea_remove_stuffed(struct gfs2_inode *ip, struct gfs2_ea_location *el)
 	if (error)
 		return error;
 
-	gfs2_trans_add_bh(ip->i_gl, el->el_bh, 1);
+	gfs2_trans_add_meta(ip->i_gl, el->el_bh);
 
 	if (prev) {
 		u32 len;
@@ -1109,7 +1109,7 @@ static int ea_remove_stuffed(struct gfs2_inode *ip, struct gfs2_ea_location *el)
 	error = gfs2_meta_inode_buffer(ip, &dibh);
 	if (!error) {
 		ip->i_inode.i_ctime = CURRENT_TIME;
-		gfs2_trans_add_bh(ip->i_gl, dibh, 1);
+		gfs2_trans_add_meta(ip->i_gl, dibh);
 		gfs2_dinode_out(ip, dibh->b_data);
 		brelse(dibh);
 	}
@@ -1265,7 +1265,7 @@ int gfs2_xattr_acl_chmod(struct gfs2_inode *ip, struct iattr *attr, char *data)
 	if (GFS2_EA_IS_STUFFED(el.el_ea)) {
 		error = gfs2_trans_begin(sdp, RES_DINODE + RES_EATTR, 0);
 		if (error == 0) {
-			gfs2_trans_add_bh(ip->i_gl, el.el_bh, 1);
+			gfs2_trans_add_meta(ip->i_gl, el.el_bh);
 			memcpy(GFS2_EA2DATA(el.el_ea), data,
 			       GFS2_EA_DATA_LEN(el.el_ea));
 		}
@@ -1352,7 +1352,7 @@ static int ea_dealloc_indirect(struct gfs2_inode *ip)
 	if (error)
 		goto out_gunlock;
 
-	gfs2_trans_add_bh(ip->i_gl, indbh, 1);
+	gfs2_trans_add_meta(ip->i_gl, indbh);
 
 	eablk = (__be64 *)(indbh->b_data + sizeof(struct gfs2_meta_header));
 	bstart = 0;
@@ -1384,7 +1384,7 @@ static int ea_dealloc_indirect(struct gfs2_inode *ip)
 
 	error = gfs2_meta_inode_buffer(ip, &dibh);
 	if (!error) {
-		gfs2_trans_add_bh(ip->i_gl, dibh, 1);
+		gfs2_trans_add_meta(ip->i_gl, dibh);
 		gfs2_dinode_out(ip, dibh->b_data);
 		brelse(dibh);
 	}
@@ -1434,7 +1434,7 @@ static int ea_dealloc_block(struct gfs2_inode *ip)
 
 	error = gfs2_meta_inode_buffer(ip, &dibh);
 	if (!error) {
-		gfs2_trans_add_bh(ip->i_gl, dibh, 1);
+		gfs2_trans_add_meta(ip->i_gl, dibh);
 		gfs2_dinode_out(ip, dibh->b_data);
 		brelse(dibh);
 	}
-- 
1.7.4

[Cluster-devel] [PATCH 04/10] GFS2: Copy gfs2_trans_add_bh into new data/meta functions

[Steven Whitehouse]

This patch copies the body of gfs2_trans_add_bh into the two newly
added gfs2_trans_add_data and gfs2_trans_add_meta functions. We can
then move the .lo_add functions from lops.c into trans.c and call
them directly.

As a result of this, we no longer need to use the .lo_add functions
at all, so that is removed from the log operations structure.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index c373a24..5d129ab 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -52,7 +52,6 @@ struct gfs2_log_header_host {
  */
 
 struct gfs2_log_operations {
-	void (*lo_add) (struct gfs2_sbd *sdp, struct gfs2_bufdata *bd);
 	void (*lo_before_commit) (struct gfs2_sbd *sdp);
 	void (*lo_after_commit) (struct gfs2_sbd *sdp, struct gfs2_ail *ai);
 	void (*lo_before_scan) (struct gfs2_jdesc *jd,
diff --git a/fs/gfs2/lops.c b/fs/gfs2/lops.c
index 9c80742..a505597 100644
--- a/fs/gfs2/lops.c
+++ b/fs/gfs2/lops.c
@@ -37,7 +37,7 @@
  *
  * The log lock must be held when calling this function
  */
-static void gfs2_pin(struct gfs2_sbd *sdp, struct buffer_head *bh)
+void gfs2_pin(struct gfs2_sbd *sdp, struct buffer_head *bh)
 {
 	struct gfs2_bufdata *bd;
 
@@ -388,32 +388,6 @@ static struct page *gfs2_get_log_desc(struct gfs2_sbd *sdp, u32 ld_type,
 	return page;
 }
 
-static void buf_lo_add(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd)
-{
-	struct gfs2_meta_header *mh;
-	struct gfs2_trans *tr;
-
-	tr = current->journal_info;
-	tr->tr_touched = 1;
-	if (!list_empty(&bd->bd_list))
-		return;
-	set_bit(GLF_LFLUSH, &bd->bd_gl->gl_flags);
-	set_bit(GLF_DIRTY, &bd->bd_gl->gl_flags);
-	mh = (struct gfs2_meta_header *)bd->bd_bh->b_data;
-	if (unlikely(mh->mh_magic != cpu_to_be32(GFS2_MAGIC))) {
-		printk(KERN_ERR
-		       "Attempting to add uninitialised block to journal (inplace block=%lld)\n",
-		       (unsigned long long)bd->bd_bh->b_blocknr);
-		BUG();
-	}
-	gfs2_pin(sdp, bd->bd_bh);
-	mh->__pad0 = cpu_to_be64(0);
-	mh->mh_jid = cpu_to_be32(sdp->sd_jdesc->jd_jid);
-	sdp->sd_log_num_buf++;
-	list_add(&bd->bd_list, &sdp->sd_log_le_buf);
-	tr->tr_num_buf_new++;
-}
-
 static void gfs2_check_magic(struct buffer_head *bh)
 {
 	void *kaddr;
@@ -735,44 +709,6 @@ static void revoke_lo_after_scan(struct gfs2_jdesc *jd, int error, int pass)
 }
 
 /**
- * databuf_lo_add - Add a databuf to the transaction.
- *
- * This is used in two distinct cases:
- * i) In ordered write mode
- *    We put the data buffer on a list so that we can ensure that its
- *    synced to disk at the right time
- * ii) In journaled data mode
- *    We need to journal the data block in the same way as metadata in
- *    the functions above. The difference is that here we have a tag
- *    which is two __be64's being the block number (as per meta data)
- *    and a flag which says whether the data block needs escaping or
- *    not. This means we need a new log entry for each 251 or so data
- *    blocks, which isn't an enormous overhead but twice as much as
- *    for normal metadata blocks.
- */
-static void databuf_lo_add(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd)
-{
-	struct gfs2_trans *tr = current->journal_info;
-	struct address_space *mapping = bd->bd_bh->b_page->mapping;
-	struct gfs2_inode *ip = GFS2_I(mapping->host);
-
-	if (tr)
-		tr->tr_touched = 1;
-	if (!list_empty(&bd->bd_list))
-		return;
-	set_bit(GLF_LFLUSH, &bd->bd_gl->gl_flags);
-	set_bit(GLF_DIRTY, &bd->bd_gl->gl_flags);
-	if (gfs2_is_jdata(ip)) {
-		gfs2_pin(sdp, bd->bd_bh);
-		tr->tr_num_databuf_new++;
-		sdp->sd_log_num_databuf++;
-		list_add_tail(&bd->bd_list, &sdp->sd_log_le_databuf);
-	} else {
-		list_add_tail(&bd->bd_list, &sdp->sd_log_le_ordered);
-	}
-}
-
-/**
  * databuf_lo_before_commit - Scan the data buffers, writing as we go
  *
  */
@@ -871,7 +807,6 @@ static void databuf_lo_after_commit(struct gfs2_sbd *sdp, struct gfs2_ail *ai)
 
 
 const struct gfs2_log_operations gfs2_buf_lops = {
-	.lo_add = buf_lo_add,
 	.lo_before_commit = buf_lo_before_commit,
 	.lo_after_commit = buf_lo_after_commit,
 	.lo_before_scan = buf_lo_before_scan,
@@ -894,7 +829,6 @@ const struct gfs2_log_operations gfs2_rg_lops = {
 };
 
 const struct gfs2_log_operations gfs2_databuf_lops = {
-	.lo_add = databuf_lo_add,
 	.lo_before_commit = databuf_lo_before_commit,
 	.lo_after_commit = databuf_lo_after_commit,
 	.lo_scan_elements = databuf_lo_scan_elements,
diff --git a/fs/gfs2/lops.h b/fs/gfs2/lops.h
index 954a330..d85b376 100644
--- a/fs/gfs2/lops.h
+++ b/fs/gfs2/lops.h
@@ -29,6 +29,7 @@ extern const struct gfs2_log_operations gfs2_databuf_lops;
 extern const struct gfs2_log_operations *gfs2_log_ops[];
 extern void gfs2_log_write_page(struct gfs2_sbd *sdp, struct page *page);
 extern void gfs2_log_flush_bio(struct gfs2_sbd *sdp, int rw);
+extern void gfs2_pin(struct gfs2_sbd *sdp, struct buffer_head *bh);
 
 static inline unsigned int buf_limit(struct gfs2_sbd *sdp)
 {
@@ -53,12 +54,6 @@ static inline void lops_init_le(struct gfs2_bufdata *bd,
 	bd->bd_ops = lops;
 }
 
-static inline void lops_add(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd)
-{
-	if (bd->bd_ops->lo_add)
-		bd->bd_ops->lo_add(sdp, bd);
-}
-
 static inline void lops_before_commit(struct gfs2_sbd *sdp)
 {
 	int x;
diff --git a/fs/gfs2/trans.c b/fs/gfs2/trans.c
index cb795ac..1fbd57e 100644
--- a/fs/gfs2/trans.c
+++ b/fs/gfs2/trans.c
@@ -18,6 +18,7 @@
 #include "gfs2.h"
 #include "incore.h"
 #include "glock.h"
+#include "inode.h"
 #include "log.h"
 #include "lops.h"
 #include "meta_io.h"
@@ -143,15 +144,46 @@ void gfs2_trans_end(struct gfs2_sbd *sdp)
 }
 
 /**
- * gfs2_trans_add_bh - Add a to-be-modified buffer to the current transaction
- * @gl: the glock the buffer belongs to
- * @bh: The buffer to add
- * @meta: True in the case of adding metadata
+ * databuf_lo_add - Add a databuf to the transaction.
  *
+ * This is used in two distinct cases:
+ * i) In ordered write mode
+ *    We put the data buffer on a list so that we can ensure that its
+ *    synced to disk at the right time
+ * ii) In journaled data mode
+ *    We need to journal the data block in the same way as metadata in
+ *    the functions above. The difference is that here we have a tag
+ *    which is two __be64's being the block number (as per meta data)
+ *    and a flag which says whether the data block needs escaping or
+ *    not. This means we need a new log entry for each 251 or so data
+ *    blocks, which isn't an enormous overhead but twice as much as
+ *    for normal metadata blocks.
  */
+static void databuf_lo_add(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd)
+{
+	struct gfs2_trans *tr = current->journal_info;
+	struct address_space *mapping = bd->bd_bh->b_page->mapping;
+	struct gfs2_inode *ip = GFS2_I(mapping->host);
+
+	if (tr)
+		tr->tr_touched = 1;
+	if (!list_empty(&bd->bd_list))
+		return;
+	set_bit(GLF_LFLUSH, &bd->bd_gl->gl_flags);
+	set_bit(GLF_DIRTY, &bd->bd_gl->gl_flags);
+	if (gfs2_is_jdata(ip)) {
+		gfs2_pin(sdp, bd->bd_bh);
+		tr->tr_num_databuf_new++;
+		sdp->sd_log_num_databuf++;
+		list_add_tail(&bd->bd_list, &sdp->sd_log_le_databuf);
+	} else {
+		list_add_tail(&bd->bd_list, &sdp->sd_log_le_ordered);
+	}
+}
 
-static void gfs2_trans_add_bh(struct gfs2_glock *gl, struct buffer_head *bh, int meta)
+void gfs2_trans_add_data(struct gfs2_glock *gl, struct buffer_head *bh)
 {
+
 	struct gfs2_sbd *sdp = gl->gl_sbd;
 	struct gfs2_bufdata *bd;
 
@@ -163,24 +195,64 @@ static void gfs2_trans_add_bh(struct gfs2_glock *gl, struct buffer_head *bh, int
 	else {
 		gfs2_log_unlock(sdp);
 		unlock_buffer(bh);
-		gfs2_attach_bufdata(gl, bh, meta);
+		gfs2_attach_bufdata(gl, bh, 0);
 		bd = bh->b_private;
 		lock_buffer(bh);
 		gfs2_log_lock(sdp);
 	}
-	lops_add(sdp, bd);
+	databuf_lo_add(sdp, bd);
 	gfs2_log_unlock(sdp);
 	unlock_buffer(bh);
 }
 
-void gfs2_trans_add_data(struct gfs2_glock *gl, struct buffer_head *bh)
+static void meta_lo_add(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd)
 {
-	return gfs2_trans_add_bh(gl, bh, 0);
+	struct gfs2_meta_header *mh;
+	struct gfs2_trans *tr;
+
+	tr = current->journal_info;
+	tr->tr_touched = 1;
+	if (!list_empty(&bd->bd_list))
+		return;
+	set_bit(GLF_LFLUSH, &bd->bd_gl->gl_flags);
+	set_bit(GLF_DIRTY, &bd->bd_gl->gl_flags);
+	mh = (struct gfs2_meta_header *)bd->bd_bh->b_data;
+	if (unlikely(mh->mh_magic != cpu_to_be32(GFS2_MAGIC))) {
+		printk(KERN_ERR
+		       "Attempting to add uninitialised block to journal (inplace block=%lld)\n",
+		       (unsigned long long)bd->bd_bh->b_blocknr);
+		BUG();
+	}
+	gfs2_pin(sdp, bd->bd_bh);
+	mh->__pad0 = cpu_to_be64(0);
+	mh->mh_jid = cpu_to_be32(sdp->sd_jdesc->jd_jid);
+	sdp->sd_log_num_buf++;
+	list_add(&bd->bd_list, &sdp->sd_log_le_buf);
+	tr->tr_num_buf_new++;
 }
 
 void gfs2_trans_add_meta(struct gfs2_glock *gl, struct buffer_head *bh)
 {
-	return gfs2_trans_add_bh(gl, bh, 1);
+
+	struct gfs2_sbd *sdp = gl->gl_sbd;
+	struct gfs2_bufdata *bd;
+
+	lock_buffer(bh);
+	gfs2_log_lock(sdp);
+	bd = bh->b_private;
+	if (bd)
+		gfs2_assert(sdp, bd->bd_gl == gl);
+	else {
+		gfs2_log_unlock(sdp);
+		unlock_buffer(bh);
+		gfs2_attach_bufdata(gl, bh, 1);
+		bd = bh->b_private;
+		lock_buffer(bh);
+		gfs2_log_lock(sdp);
+	}
+	meta_lo_add(sdp, bd);
+	gfs2_log_unlock(sdp);
+	unlock_buffer(bh);
 }
 
 void gfs2_trans_add_revoke(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd)
-- 
1.7.4

[Cluster-devel] [PATCH 05/10] GFS2: Merge gfs2_attach_bufdata() into trans.c

[Steven Whitehouse]

The locking in gfs2_attach_bufdata() was type specific (data/meta)
which made the function rather confusing. This patch moves the core
of gfs2_attach_bufdata() into trans.c renaming it gfs2_alloc_bufdata()
and moving the locking into gfs2_trans_add_data()/gfs2_trans_add_meta()

As a result all of the locking related to adding data and metadata to
the journal is now in these two functions. This should help to clarify
what is going on, and give us some opportunities to simplify in
some cases.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/lops.h b/fs/gfs2/lops.h
index d85b376..ba77b7d 100644
--- a/fs/gfs2/lops.h
+++ b/fs/gfs2/lops.h
@@ -47,13 +47,6 @@ static inline unsigned int databuf_limit(struct gfs2_sbd *sdp)
 	return limit;
 }
 
-static inline void lops_init_le(struct gfs2_bufdata *bd,
-				const struct gfs2_log_operations *lops)
-{
-	INIT_LIST_HEAD(&bd->bd_list);
-	bd->bd_ops = lops;
-}
-
 static inline void lops_before_commit(struct gfs2_sbd *sdp)
 {
 	int x;
diff --git a/fs/gfs2/meta_io.c b/fs/gfs2/meta_io.c
index 22255d9..b059bbb 100644
--- a/fs/gfs2/meta_io.c
+++ b/fs/gfs2/meta_io.c
@@ -271,41 +271,6 @@ int gfs2_meta_wait(struct gfs2_sbd *sdp, struct buffer_head *bh)
 	return 0;
 }
 
-/**
- * gfs2_attach_bufdata - attach a struct gfs2_bufdata structure to a buffer
- * @gl: the glock the buffer belongs to
- * @bh: The buffer to be attached to
- * @meta: Flag to indicate whether its metadata or not
- */
-
-void gfs2_attach_bufdata(struct gfs2_glock *gl, struct buffer_head *bh,
-			 int meta)
-{
-	struct gfs2_bufdata *bd;
-
-	if (meta)
-		lock_page(bh->b_page);
-
-	if (bh->b_private) {
-		if (meta)
-			unlock_page(bh->b_page);
-		return;
-	}
-
-	bd = kmem_cache_zalloc(gfs2_bufdata_cachep, GFP_NOFS | __GFP_NOFAIL);
-	bd->bd_bh = bh;
-	bd->bd_gl = gl;
-
-	if (meta)
-		lops_init_le(bd, &gfs2_buf_lops);
-	else
-		lops_init_le(bd, &gfs2_databuf_lops);
-	bh->b_private = bd;
-
-	if (meta)
-		unlock_page(bh->b_page);
-}
-
 void gfs2_remove_from_journal(struct buffer_head *bh, struct gfs2_trans *tr, int meta)
 {
 	struct address_space *mapping = bh->b_page->mapping;
diff --git a/fs/gfs2/meta_io.h b/fs/gfs2/meta_io.h
index c30973b..0d4c843 100644
--- a/fs/gfs2/meta_io.h
+++ b/fs/gfs2/meta_io.h
@@ -56,9 +56,6 @@ int gfs2_meta_read(struct gfs2_glock *gl, u64 blkno,
 int gfs2_meta_wait(struct gfs2_sbd *sdp, struct buffer_head *bh);
 struct buffer_head *gfs2_getbuf(struct gfs2_glock *gl, u64 blkno, int create);
 
-void gfs2_attach_bufdata(struct gfs2_glock *gl, struct buffer_head *bh,
-			 int meta);
-
 void gfs2_remove_from_journal(struct buffer_head *bh, struct gfs2_trans *tr,
 			      int meta);
 
diff --git a/fs/gfs2/trans.c b/fs/gfs2/trans.c
index 1fbd57e..14dbf6d 100644
--- a/fs/gfs2/trans.c
+++ b/fs/gfs2/trans.c
@@ -143,6 +143,21 @@ void gfs2_trans_end(struct gfs2_sbd *sdp)
 	sb_end_intwrite(sdp->sd_vfs);
 }
 
+static struct gfs2_bufdata *gfs2_alloc_bufdata(struct gfs2_glock *gl,
+					       struct buffer_head *bh,
+					       const struct gfs2_log_operations *lops)
+{
+	struct gfs2_bufdata *bd;
+
+	bd = kmem_cache_zalloc(gfs2_bufdata_cachep, GFP_NOFS | __GFP_NOFAIL);
+	bd->bd_bh = bh;
+	bd->bd_gl = gl;
+	bd->bd_ops = lops;
+	INIT_LIST_HEAD(&bd->bd_list);
+	bh->b_private = bd;
+	return bd;
+}
+
 /**
  * databuf_lo_add - Add a databuf to the transaction.
  *
@@ -190,16 +205,15 @@ void gfs2_trans_add_data(struct gfs2_glock *gl, struct buffer_head *bh)
 	lock_buffer(bh);
 	gfs2_log_lock(sdp);
 	bd = bh->b_private;
-	if (bd)
-		gfs2_assert(sdp, bd->bd_gl == gl);
-	else {
+	if (bd == NULL) {
 		gfs2_log_unlock(sdp);
 		unlock_buffer(bh);
-		gfs2_attach_bufdata(gl, bh, 0);
-		bd = bh->b_private;
+		if (bh->b_private == NULL)
+			bd = gfs2_alloc_bufdata(gl, bh, &gfs2_databuf_lops);
 		lock_buffer(bh);
 		gfs2_log_lock(sdp);
 	}
+	gfs2_assert(sdp, bd->bd_gl == gl);
 	databuf_lo_add(sdp, bd);
 	gfs2_log_unlock(sdp);
 	unlock_buffer(bh);
@@ -240,16 +254,17 @@ void gfs2_trans_add_meta(struct gfs2_glock *gl, struct buffer_head *bh)
 	lock_buffer(bh);
 	gfs2_log_lock(sdp);
 	bd = bh->b_private;
-	if (bd)
-		gfs2_assert(sdp, bd->bd_gl == gl);
-	else {
+	if (bd == NULL) {
 		gfs2_log_unlock(sdp);
 		unlock_buffer(bh);
-		gfs2_attach_bufdata(gl, bh, 1);
-		bd = bh->b_private;
+		lock_page(bh->b_page);
+		if (bh->b_private == NULL)
+			bd = gfs2_alloc_bufdata(gl, bh, &gfs2_buf_lops);
+		unlock_page(bh->b_page);
 		lock_buffer(bh);
 		gfs2_log_lock(sdp);
 	}
+	gfs2_assert(sdp, bd->bd_gl == gl);
 	meta_lo_add(sdp, bd);
 	gfs2_log_unlock(sdp);
 	unlock_buffer(bh);
@@ -263,7 +278,7 @@ void gfs2_trans_add_revoke(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd)
 	BUG_ON(!list_empty(&bd->bd_list));
 	BUG_ON(!list_empty(&bd->bd_ail_st_list));
 	BUG_ON(!list_empty(&bd->bd_ail_gl_list));
-	lops_init_le(bd, &gfs2_revoke_lops);
+	bd->bd_ops = &gfs2_revoke_lops;
 	tr->tr_touched = 1;
 	tr->tr_num_revoke++;
 	sdp->sd_log_num_revoke++;
-- 
1.7.4

[Cluster-devel] [PATCH 06/10] GFS2: Clean up freeze code

[Steven Whitehouse]

The freeze code has not been looked at a lot recently. Upstream has
moved on, and this is an attempt to catch us back up again. There
is a vfs level interface for the freeze code which can be called
from our (obsolete, but kept for backward compatibility purposes)
sysfs freeze interface. This means freezing this way vs. doing it
from the ioctl should now work in identical fashion.

As a result of this, the freeze function is only called once
and we can drop our own special purpose code for counting the
number of freezes.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index 5d129ab..19750bc 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -757,10 +757,7 @@ struct gfs2_sbd {
 	unsigned int sd_replayed_blocks;
 
 	/* For quiescing the filesystem */
-
 	struct gfs2_holder sd_freeze_gh;
-	struct mutex sd_freeze_lock;
-	unsigned int sd_freeze_count;
 
 	char sd_fsname[GFS2_FSNAME_LEN];
 	char sd_table_name[GFS2_FSNAME_LEN];
diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c
index 0e3554e..5f5aba5 100644
--- a/fs/gfs2/ops_fstype.c
+++ b/fs/gfs2/ops_fstype.c
@@ -115,8 +115,6 @@ static struct gfs2_sbd *init_sbd(struct super_block *sb)
 
 	INIT_LIST_HEAD(&sdp->sd_revoke_list);
 
-	mutex_init(&sdp->sd_freeze_lock);
-
 	return sdp;
 }
 
diff --git a/fs/gfs2/super.c b/fs/gfs2/super.c
index 4dfda4c..c075b62 100644
--- a/fs/gfs2/super.c
+++ b/fs/gfs2/super.c
@@ -663,54 +663,6 @@ out:
 	return error;
 }
 
-/**
- * gfs2_freeze_fs - freezes the file system
- * @sdp: the file system
- *
- * This function flushes data and meta data for all machines by
- * acquiring the transaction log exclusively.  All journals are
- * ensured to be in a clean state as well.
- *
- * Returns: errno
- */
-
-int gfs2_freeze_fs(struct gfs2_sbd *sdp)
-{
-	int error = 0;
-
-	mutex_lock(&sdp->sd_freeze_lock);
-
-	if (!sdp->sd_freeze_count++) {
-		error = gfs2_lock_fs_check_clean(sdp, &sdp->sd_freeze_gh);
-		if (error)
-			sdp->sd_freeze_count--;
-	}
-
-	mutex_unlock(&sdp->sd_freeze_lock);
-
-	return error;
-}
-
-/**
- * gfs2_unfreeze_fs - unfreezes the file system
- * @sdp: the file system
- *
- * This function allows the file system to proceed by unlocking
- * the exclusively held transaction lock.  Other GFS2 nodes are
- * now free to acquire the lock shared and go on with their lives.
- *
- */
-
-void gfs2_unfreeze_fs(struct gfs2_sbd *sdp)
-{
-	mutex_lock(&sdp->sd_freeze_lock);
-
-	if (sdp->sd_freeze_count && !--sdp->sd_freeze_count)
-		gfs2_glock_dq_uninit(&sdp->sd_freeze_gh);
-
-	mutex_unlock(&sdp->sd_freeze_lock);
-}
-
 void gfs2_dinode_out(const struct gfs2_inode *ip, void *buf)
 {
 	struct gfs2_dinode *str = buf;
@@ -888,13 +840,6 @@ static void gfs2_put_super(struct super_block *sb)
 	int error;
 	struct gfs2_jdesc *jd;
 
-	/*  Unfreeze the filesystem, if we need to  */
-
-	mutex_lock(&sdp->sd_freeze_lock);
-	if (sdp->sd_freeze_count)
-		gfs2_glock_dq_uninit(&sdp->sd_freeze_gh);
-	mutex_unlock(&sdp->sd_freeze_lock);
-
 	/* No more recovery requests */
 	set_bit(SDF_NORECOVERY, &sdp->sd_flags);
 	smp_mb();
@@ -985,7 +930,7 @@ static int gfs2_freeze(struct super_block *sb)
 		return -EINVAL;
 
 	for (;;) {
-		error = gfs2_freeze_fs(sdp);
+		error = gfs2_lock_fs_check_clean(sdp, &sdp->sd_freeze_gh);
 		if (!error)
 			break;
 
@@ -1013,7 +958,9 @@ static int gfs2_freeze(struct super_block *sb)
 
 static int gfs2_unfreeze(struct super_block *sb)
 {
-	gfs2_unfreeze_fs(sb->s_fs_info);
+	struct gfs2_sbd *sdp = sb->s_fs_info;
+
+	gfs2_glock_dq_uninit(&sdp->sd_freeze_gh);
 	return 0;
 }
 
diff --git a/fs/gfs2/super.h b/fs/gfs2/super.h
index a046468..90e3322 100644
--- a/fs/gfs2/super.h
+++ b/fs/gfs2/super.h
@@ -46,9 +46,6 @@ extern void update_statfs(struct gfs2_sbd *sdp, struct buffer_head *m_bh,
 			  struct buffer_head *l_bh);
 extern int gfs2_statfs_sync(struct super_block *sb, int type);
 
-extern int gfs2_freeze_fs(struct gfs2_sbd *sdp);
-extern void gfs2_unfreeze_fs(struct gfs2_sbd *sdp);
-
 extern struct file_system_type gfs2_fs_type;
 extern struct file_system_type gfs2meta_fs_type;
 extern const struct export_operations gfs2_export_ops;
diff --git a/fs/gfs2/sys.c b/fs/gfs2/sys.c
index 8056b7b..462e841 100644
--- a/fs/gfs2/sys.c
+++ b/fs/gfs2/sys.c
@@ -91,19 +91,15 @@ static ssize_t uuid_show(struct gfs2_sbd *sdp, char *buf)
 
 static ssize_t freeze_show(struct gfs2_sbd *sdp, char *buf)
 {
-	unsigned int count;
-
-	mutex_lock(&sdp->sd_freeze_lock);
-	count = sdp->sd_freeze_count;
-	mutex_unlock(&sdp->sd_freeze_lock);
+	struct super_block *sb = sdp->sd_vfs;
+	int frozen = (sb->s_writers.frozen == SB_UNFROZEN) ? 0 : 1;
 
-	return snprintf(buf, PAGE_SIZE, "%u\n", count);
+	return snprintf(buf, PAGE_SIZE, "%u\n", frozen);
 }
 
 static ssize_t freeze_store(struct gfs2_sbd *sdp, const char *buf, size_t len)
 {
-	ssize_t ret = len;
-	int error = 0;
+	int error;
 	int n = simple_strtol(buf, NULL, 0);
 
 	if (!capable(CAP_SYS_ADMIN))
@@ -111,19 +107,21 @@ static ssize_t freeze_store(struct gfs2_sbd *sdp, const char *buf, size_t len)
 
 	switch (n) {
 	case 0:
-		gfs2_unfreeze_fs(sdp);
+		error = thaw_super(sdp->sd_vfs);
 		break;
 	case 1:
-		error = gfs2_freeze_fs(sdp);
+		error = freeze_super(sdp->sd_vfs);
 		break;
 	default:
-		ret = -EINVAL;
+		return -EINVAL;
 	}
 
-	if (error)
+	if (error) {
 		fs_warn(sdp, "freeze %d error %d", n, error);
+		return error;
+	}
 
-	return ret;
+	return len;
 }
 
 static ssize_t withdraw_show(struct gfs2_sbd *sdp, char *buf)
-- 
1.7.4

[Cluster-devel] [PATCH 07/10] GFS2: Use ->writepages for ordered writes

[Steven Whitehouse]

Instead of using a list of buffers to write ahead of the journal
flush, this now uses a list of inodes and calls ->writepages
via filemap_fdatawrite() in order to achieve the same thing. For
most use cases this results in a shorter ordered write list,
as well as much larger i/os being issued.

The ordered write list is sorted by inode number before writing
in order to retain the disk block ordering between inodes as
per the previous code.

The previous ordered write code used to conflict in its assumptions
about how to write out the disk blocks with mpage_writepages()
so that with this updated version we can also use mpage_writepages()
for GFS2's ordered write, writepages implementation. So we will
also send larger i/os from writeback too.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/aops.c b/fs/gfs2/aops.c
index 92340dd..24f414f 100644
--- a/fs/gfs2/aops.c
+++ b/fs/gfs2/aops.c
@@ -230,16 +230,14 @@ out_ignore:
 }
 
 /**
- * gfs2_writeback_writepages - Write a bunch of dirty pages back to disk
+ * gfs2_writepages - Write a bunch of dirty pages back to disk
  * @mapping: The mapping to write
  * @wbc: Write-back control
  *
- * For the data=writeback case we can already ignore buffer heads
- * and write whole extents at once. This is a big reduction in the
- * number of I/O requests we send and the bmap calls we make in this case.
+ * Used for both ordered and writeback modes.
  */
-static int gfs2_writeback_writepages(struct address_space *mapping,
-				     struct writeback_control *wbc)
+static int gfs2_writepages(struct address_space *mapping,
+			   struct writeback_control *wbc)
 {
 	return mpage_writepages(mapping, wbc, gfs2_get_block_noalloc);
 }
@@ -1102,7 +1100,7 @@ cannot_release:
 
 static const struct address_space_operations gfs2_writeback_aops = {
 	.writepage = gfs2_writeback_writepage,
-	.writepages = gfs2_writeback_writepages,
+	.writepages = gfs2_writepages,
 	.readpage = gfs2_readpage,
 	.readpages = gfs2_readpages,
 	.write_begin = gfs2_write_begin,
@@ -1118,6 +1116,7 @@ static const struct address_space_operations gfs2_writeback_aops = {
 
 static const struct address_space_operations gfs2_ordered_aops = {
 	.writepage = gfs2_ordered_writepage,
+	.writepages = gfs2_writepages,
 	.readpage = gfs2_readpage,
 	.readpages = gfs2_readpages,
 	.write_begin = gfs2_write_begin,
diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index 7a86275..d29d779 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -22,6 +22,7 @@
 #include "meta_io.h"
 #include "quota.h"
 #include "rgrp.h"
+#include "log.h"
 #include "super.h"
 #include "trans.h"
 #include "dir.h"
@@ -1137,6 +1138,7 @@ static int trunc_end(struct gfs2_inode *ip)
 		ip->i_height = 0;
 		ip->i_goal = ip->i_no_addr;
 		gfs2_buffer_clear_tail(dibh, sizeof(struct gfs2_dinode));
+		gfs2_ordered_del_inode(ip);
 	}
 	ip->i_inode.i_mtime = ip->i_inode.i_ctime = CURRENT_TIME;
 	ip->i_diskflags &= ~GFS2_DIF_TRUNC_IN_PROG;
diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index 19750bc..1533cf8 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -340,6 +340,7 @@ enum {
 	GIF_QD_LOCKED		= 1,
 	GIF_ALLOC_FAILED	= 2,
 	GIF_SW_PAGED		= 3,
+	GIF_ORDERED		= 4,
 };
 
 struct gfs2_inode {
@@ -356,6 +357,7 @@ struct gfs2_inode {
 	struct gfs2_rgrpd *i_rgd;
 	u64 i_goal;	/* goal block for allocations */
 	struct rw_semaphore i_rw_mutex;
+	struct list_head i_ordered;
 	struct list_head i_trunc_list;
 	__be64 *i_hash_cache;
 	u32 i_entries;
@@ -722,6 +724,7 @@ struct gfs2_sbd {
 	struct list_head sd_log_le_revoke;
 	struct list_head sd_log_le_databuf;
 	struct list_head sd_log_le_ordered;
+	spinlock_t sd_ordered_lock;
 
 	atomic_t sd_log_thresh1;
 	atomic_t sd_log_thresh2;
diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c
index f4beeb9..9a2ca8b 100644
--- a/fs/gfs2/log.c
+++ b/fs/gfs2/log.c
@@ -482,70 +482,66 @@ static void log_flush_wait(struct gfs2_sbd *sdp)
 	}
 }
 
-static int bd_cmp(void *priv, struct list_head *a, struct list_head *b)
+static int ip_cmp(void *priv, struct list_head *a, struct list_head *b)
 {
-	struct gfs2_bufdata *bda, *bdb;
+	struct gfs2_inode *ipa, *ipb;
 
-	bda = list_entry(a, struct gfs2_bufdata, bd_list);
-	bdb = list_entry(b, struct gfs2_bufdata, bd_list);
+	ipa = list_entry(a, struct gfs2_inode, i_ordered);
+	ipb = list_entry(b, struct gfs2_inode, i_ordered);
 
-	if (bda->bd_bh->b_blocknr < bdb->bd_bh->b_blocknr)
+	if (ipa->i_no_addr < ipb->i_no_addr)
 		return -1;
-	if (bda->bd_bh->b_blocknr > bdb->bd_bh->b_blocknr)
+	if (ipa->i_no_addr > ipb->i_no_addr)
 		return 1;
 	return 0;
 }
 
 static void gfs2_ordered_write(struct gfs2_sbd *sdp)
 {
-	struct gfs2_bufdata *bd;
-	struct buffer_head *bh;
+	struct gfs2_inode *ip;
 	LIST_HEAD(written);
 
-	gfs2_log_lock(sdp);
-	list_sort(NULL, &sdp->sd_log_le_ordered, &bd_cmp);
+	spin_lock(&sdp->sd_ordered_lock);
+	list_sort(NULL, &sdp->sd_log_le_ordered, &ip_cmp);
 	while (!list_empty(&sdp->sd_log_le_ordered)) {
-		bd = list_entry(sdp->sd_log_le_ordered.next, struct gfs2_bufdata, bd_list);
-		list_move(&bd->bd_list, &written);
-		bh = bd->bd_bh;
-		if (!buffer_dirty(bh))
+		ip = list_entry(sdp->sd_log_le_ordered.next, struct gfs2_inode, i_ordered);
+		list_move(&ip->i_ordered, &written);
+		if (ip->i_inode.i_mapping->nrpages == 0)
 			continue;
-		get_bh(bh);
-		gfs2_log_unlock(sdp);
-		lock_buffer(bh);
-		if (buffer_mapped(bh) && test_clear_buffer_dirty(bh)) {
-			bh->b_end_io = end_buffer_write_sync;
-			submit_bh(WRITE_SYNC, bh);
-		} else {
-			unlock_buffer(bh);
-			brelse(bh);
-		}
-		gfs2_log_lock(sdp);
+		spin_unlock(&sdp->sd_ordered_lock);
+		filemap_fdatawrite(ip->i_inode.i_mapping);
+		spin_lock(&sdp->sd_ordered_lock);
 	}
 	list_splice(&written, &sdp->sd_log_le_ordered);
-	gfs2_log_unlock(sdp);
+	spin_unlock(&sdp->sd_ordered_lock);
 }
 
 static void gfs2_ordered_wait(struct gfs2_sbd *sdp)
 {
-	struct gfs2_bufdata *bd;
-	struct buffer_head *bh;
+	struct gfs2_inode *ip;
 
-	gfs2_log_lock(sdp);
+	spin_lock(&sdp->sd_ordered_lock);
 	while (!list_empty(&sdp->sd_log_le_ordered)) {
-		bd = list_entry(sdp->sd_log_le_ordered.prev, struct gfs2_bufdata, bd_list);
-		bh = bd->bd_bh;
-		if (buffer_locked(bh)) {
-			get_bh(bh);
-			gfs2_log_unlock(sdp);
-			wait_on_buffer(bh);
-			brelse(bh);
-			gfs2_log_lock(sdp);
+		ip = list_entry(sdp->sd_log_le_ordered.next, struct gfs2_inode, i_ordered);
+		list_del(&ip->i_ordered);
+		WARN_ON(!test_and_clear_bit(GIF_ORDERED, &ip->i_flags));
+		if (ip->i_inode.i_mapping->nrpages == 0)
 			continue;
-		}
-		list_del_init(&bd->bd_list);
+		spin_unlock(&sdp->sd_ordered_lock);
+		filemap_fdatawait(ip->i_inode.i_mapping);
+		spin_lock(&sdp->sd_ordered_lock);
 	}
-	gfs2_log_unlock(sdp);
+	spin_unlock(&sdp->sd_ordered_lock);
+}
+
+void gfs2_ordered_del_inode(struct gfs2_inode *ip)
+{
+	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
+
+	spin_lock(&sdp->sd_ordered_lock);
+	if (test_and_clear_bit(GIF_ORDERED, &ip->i_flags))
+		list_del(&ip->i_ordered);
+	spin_unlock(&sdp->sd_ordered_lock);
 }
 
 /**
diff --git a/fs/gfs2/log.h b/fs/gfs2/log.h
index 3fd5215..3566f35 100644
--- a/fs/gfs2/log.h
+++ b/fs/gfs2/log.h
@@ -48,6 +48,18 @@ static inline void gfs2_log_pointers_init(struct gfs2_sbd *sdp,
 	sdp->sd_log_head = sdp->sd_log_tail = value;
 }
 
+static inline void gfs2_ordered_add_inode(struct gfs2_inode *ip)
+{
+	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
+
+	if (!test_bit(GIF_ORDERED, &ip->i_flags)) {
+		spin_lock(&sdp->sd_ordered_lock);
+		if (!test_and_set_bit(GIF_ORDERED, &ip->i_flags))
+			list_add(&ip->i_ordered, &sdp->sd_log_le_ordered);
+		spin_unlock(&sdp->sd_ordered_lock);
+	}
+}
+extern void gfs2_ordered_del_inode(struct gfs2_inode *ip);
 extern unsigned int gfs2_struct2blk(struct gfs2_sbd *sdp, unsigned int nstruct,
 			    unsigned int ssize);
 
diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c
index 5f5aba5..e063f22 100644
--- a/fs/gfs2/ops_fstype.c
+++ b/fs/gfs2/ops_fstype.c
@@ -102,6 +102,7 @@ static struct gfs2_sbd *init_sbd(struct super_block *sb)
 	INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
 	INIT_LIST_HEAD(&sdp->sd_log_le_databuf);
 	INIT_LIST_HEAD(&sdp->sd_log_le_ordered);
+	spin_lock_init(&sdp->sd_ordered_lock);
 
 	init_waitqueue_head(&sdp->sd_log_waitq);
 	init_waitqueue_head(&sdp->sd_logd_waitq);
diff --git a/fs/gfs2/super.c b/fs/gfs2/super.c
index c075b62..a3b40ee 100644
--- a/fs/gfs2/super.c
+++ b/fs/gfs2/super.c
@@ -1524,6 +1524,7 @@ out:
 	/* Case 3 starts here */
 	truncate_inode_pages(&inode->i_data, 0);
 	gfs2_rs_delete(ip);
+	gfs2_ordered_del_inode(ip);
 	clear_inode(inode);
 	gfs2_dir_hash_inval(ip);
 	ip->i_gl->gl_object = NULL;
diff --git a/fs/gfs2/trans.c b/fs/gfs2/trans.c
index 14dbf6d..88162fa 100644
--- a/fs/gfs2/trans.c
+++ b/fs/gfs2/trans.c
@@ -159,7 +159,9 @@ static struct gfs2_bufdata *gfs2_alloc_bufdata(struct gfs2_glock *gl,
 }
 
 /**
- * databuf_lo_add - Add a databuf to the transaction.
+ * gfs2_trans_add_data - Add a databuf to the transaction.
+ * @gl: The inode glock associated with the buffer
+ * @bh: The buffer to add
  *
  * This is used in two distinct cases:
  * i) In ordered write mode
@@ -174,33 +176,18 @@ static struct gfs2_bufdata *gfs2_alloc_bufdata(struct gfs2_glock *gl,
  *    blocks, which isn't an enormous overhead but twice as much as
  *    for normal metadata blocks.
  */
-static void databuf_lo_add(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd)
+void gfs2_trans_add_data(struct gfs2_glock *gl, struct buffer_head *bh)
 {
 	struct gfs2_trans *tr = current->journal_info;
-	struct address_space *mapping = bd->bd_bh->b_page->mapping;
+	struct gfs2_sbd *sdp = gl->gl_sbd;
+	struct address_space *mapping = bh->b_page->mapping;
 	struct gfs2_inode *ip = GFS2_I(mapping->host);
+	struct gfs2_bufdata *bd;
 
-	if (tr)
-		tr->tr_touched = 1;
-	if (!list_empty(&bd->bd_list))
+	if (!gfs2_is_jdata(ip)) {
+		gfs2_ordered_add_inode(ip);
 		return;
-	set_bit(GLF_LFLUSH, &bd->bd_gl->gl_flags);
-	set_bit(GLF_DIRTY, &bd->bd_gl->gl_flags);
-	if (gfs2_is_jdata(ip)) {
-		gfs2_pin(sdp, bd->bd_bh);
-		tr->tr_num_databuf_new++;
-		sdp->sd_log_num_databuf++;
-		list_add_tail(&bd->bd_list, &sdp->sd_log_le_databuf);
-	} else {
-		list_add_tail(&bd->bd_list, &sdp->sd_log_le_ordered);
 	}
-}
-
-void gfs2_trans_add_data(struct gfs2_glock *gl, struct buffer_head *bh)
-{
-
-	struct gfs2_sbd *sdp = gl->gl_sbd;
-	struct gfs2_bufdata *bd;
 
 	lock_buffer(bh);
 	gfs2_log_lock(sdp);
@@ -214,7 +201,15 @@ void gfs2_trans_add_data(struct gfs2_glock *gl, struct buffer_head *bh)
 		gfs2_log_lock(sdp);
 	}
 	gfs2_assert(sdp, bd->bd_gl == gl);
-	databuf_lo_add(sdp, bd);
+	tr->tr_touched = 1;
+	if (list_empty(&bd->bd_list)) {
+		set_bit(GLF_LFLUSH, &bd->bd_gl->gl_flags);
+		set_bit(GLF_DIRTY, &bd->bd_gl->gl_flags);
+		gfs2_pin(sdp, bd->bd_bh);
+		tr->tr_num_databuf_new++;
+		sdp->sd_log_num_databuf++;
+		list_add_tail(&bd->bd_list, &sdp->sd_log_le_databuf);
+	}
 	gfs2_log_unlock(sdp);
 	unlock_buffer(bh);
 }
-- 
1.7.4

[Cluster-devel] [PATCH 08/10] GFS2: Split glock lru processing into two parts

[Steven Whitehouse]

The intent here is to split the processing of the glock lru
list into two parts, so that the selection of glocks and the
disposal are separate functions. The plan is then, that further
updates can then be made to these functions in the future
to improve the selection of glocks and also the efficiency of
glock disposal.

The new feature which this patch brings is sorting the
glocks to be disposed of into glock number (and thus also
disk block number) order. Not all glocks will need i/o in
order to dispose of them, but some will, and at least we'll
generate mostly disk block order i/o now.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index 3ad8fd3..cf35155 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -30,6 +30,7 @@
 #include <linux/rculist_bl.h>
 #include <linux/bit_spinlock.h>
 #include <linux/percpu.h>
+#include <linux/list_sort.h>
 
 #include "gfs2.h"
 #include "incore.h"
@@ -1376,50 +1377,93 @@ void gfs2_glock_complete(struct gfs2_glock *gl, int ret)
 		gfs2_glock_put(gl);
 }
 
+static int glock_cmp(void *priv, struct list_head *a, struct list_head *b)
+{
+	struct gfs2_glock *gla, *glb;
+
+	gla = list_entry(a, struct gfs2_glock, gl_lru);
+	glb = list_entry(b, struct gfs2_glock, gl_lru);
+
+	if (gla->gl_name.ln_number > glb->gl_name.ln_number)
+		return 1;
+	if (gla->gl_name.ln_number < glb->gl_name.ln_number)
+		return -1;
+
+	return 0;
+}
+
+/**
+ * gfs2_dispose_glock_lru - Demote a list of glocks
+ * @list: The list to dispose of
+ *
+ * Disposing of glocks may involve disk accesses, so that here we sort
+ * the glocks by number (i.e. disk location of the inodes) so that if
+ * there are any such accesses, they'll be sent in order (mostly).
+ *
+ * Must be called under the lru_lock, but may drop and retake this
+ * lock. While the lru_lock is dropped, entries may vanish from the
+ * list, but no new entries will appear on the list (since it is
+ * private)
+ */
+
+static void gfs2_dispose_glock_lru(struct list_head *list)
+__releases(&lru_lock)
+__acquires(&lru_lock)
+{
+	struct gfs2_glock *gl;
+
+	list_sort(NULL, list, glock_cmp);
+
+	while(!list_empty(list)) {
+		gl = list_entry(list->next, struct gfs2_glock, gl_lru);
+		list_del_init(&gl->gl_lru);
+		clear_bit(GLF_LRU, &gl->gl_flags);
+		gfs2_glock_hold(gl);
+		spin_unlock(&lru_lock);
+		spin_lock(&gl->gl_spin);
+		if (demote_ok(gl))
+			handle_callback(gl, LM_ST_UNLOCKED, 0);
+		WARN_ON(!test_and_clear_bit(GLF_LOCK, &gl->gl_flags));
+		smp_mb__after_clear_bit();
+		if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0)
+			gfs2_glock_put_nolock(gl);
+		spin_unlock(&gl->gl_spin);
+		spin_lock(&lru_lock);
+	}
+}
+
 /**
  * gfs2_scan_glock_lru - Scan the LRU looking for locks to demote
  * @nr: The number of entries to scan
  *
+ * This function selects the entries on the LRU which are able to
+ * be demoted, and then kicks off the process by calling
+ * gfs2_dispose_glock_lru() above.
  */
 
 static void gfs2_scan_glock_lru(int nr)
 {
 	struct gfs2_glock *gl;
-	int may_demote;
-	int nr_skipped = 0;
 	LIST_HEAD(skipped);
+	LIST_HEAD(dispose);
 
 	spin_lock(&lru_lock);
 	while(nr && !list_empty(&lru_list)) {
 		gl = list_entry(lru_list.next, struct gfs2_glock, gl_lru);
-		list_del_init(&gl->gl_lru);
-		clear_bit(GLF_LRU, &gl->gl_flags);
-		atomic_dec(&lru_count);
 
 		/* Test for being demotable */
 		if (!test_and_set_bit(GLF_LOCK, &gl->gl_flags)) {
-			gfs2_glock_hold(gl);
-			spin_unlock(&lru_lock);
-			spin_lock(&gl->gl_spin);
-			may_demote = demote_ok(gl);
-			if (may_demote) {
-				handle_callback(gl, LM_ST_UNLOCKED, 0);
-				nr--;
-			}
-			clear_bit(GLF_LOCK, &gl->gl_flags);
-			smp_mb__after_clear_bit();
-			if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0)
-				gfs2_glock_put_nolock(gl);
-			spin_unlock(&gl->gl_spin);
-			spin_lock(&lru_lock);
+			list_move(&gl->gl_lru, &dispose);
+			atomic_dec(&lru_count);
+			nr--;
 			continue;
 		}
-		nr_skipped++;
-		list_add(&gl->gl_lru, &skipped);
-		set_bit(GLF_LRU, &gl->gl_flags);
+
+		list_move(&gl->gl_lru, &skipped);
 	}
 	list_splice(&skipped, &lru_list);
-	atomic_add(nr_skipped, &lru_count);
+	if (!list_empty(&dispose))
+		gfs2_dispose_glock_lru(&dispose);
 	spin_unlock(&lru_lock);
 }
 
-- 
1.7.4

[Cluster-devel] [PATCH 09/10] GFS2: Get a block reservation before resizing a file

[Steven Whitehouse]

From: Bob Peterson <rpeterso@redhat.com>

This patch allocates a block reservation structure before growing
or shrinking a file. Without this structure, the grow or shink code
can reference the bad pointer.

Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index d29d779..df686d1 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -1288,6 +1288,10 @@ int gfs2_setattr_size(struct inode *inode, u64 newsize)
 
 	inode_dio_wait(inode);
 
+	ret = gfs2_rs_alloc(GFS2_I(inode));
+	if (ret)
+		return ret;
+
 	oldsize = inode->i_size;
 	if (newsize >= oldsize)
 		return do_grow(inode, newsize);
-- 
1.7.4

[Cluster-devel] [PATCH 10/10] GFS2: Reinstate withdraw ack system

[Steven Whitehouse]

This patch reinstates the ack system which withdraw should be using. It
appears to have been accidentally forgotten when the lock module was
merged into GFS2, due to two different sysfs files having the same name.

Reported-by: David Teigland <teigland@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index 1533cf8..e2601ba 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -642,6 +642,7 @@ struct gfs2_sbd {
 	wait_queue_head_t sd_glock_wait;
 	atomic_t sd_glock_disposal;
 	struct completion sd_locking_init;
+	struct completion sd_wdack;
 	struct delayed_work sd_control_work;
 
 	/* Inode Stuff */
diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c
index e063f22..1b612be 100644
--- a/fs/gfs2/ops_fstype.c
+++ b/fs/gfs2/ops_fstype.c
@@ -81,6 +81,7 @@ 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);
+	init_completion(&sdp->sd_wdack);
 	spin_lock_init(&sdp->sd_statfs_spin);
 
 	spin_lock_init(&sdp->sd_rindex_spin);
diff --git a/fs/gfs2/sys.c b/fs/gfs2/sys.c
index 462e841..4fb9ad8 100644
--- a/fs/gfs2/sys.c
+++ b/fs/gfs2/sys.c
@@ -330,6 +330,28 @@ static ssize_t block_store(struct gfs2_sbd *sdp, const char *buf, size_t len)
 	return ret;
 }
 
+static ssize_t wdack_show(struct gfs2_sbd *sdp, char *buf)
+{
+	int val = completion_done(&sdp->sd_wdack) ? 1 : 0;
+
+	return sprintf(buf, "%d\n", val);
+}
+
+static ssize_t wdack_store(struct gfs2_sbd *sdp, const char *buf, size_t len)
+{
+	ssize_t ret = len;
+	int val;
+
+	val = simple_strtol(buf, NULL, 0);
+
+	if ((val == 1) &&
+	    !strcmp(sdp->sd_lockstruct.ls_ops->lm_proto_name, "lock_dlm"))
+		complete(&sdp->sd_wdack);
+	else
+		ret = -EINVAL;
+	return ret;
+}
+
 static ssize_t lkfirst_show(struct gfs2_sbd *sdp, char *buf)
 {
 	struct lm_lockstruct *ls = &sdp->sd_lockstruct;
@@ -461,7 +483,7 @@ static struct gfs2_attr gdlm_attr_##_name = __ATTR(_name,_mode,_show,_store)
 
 GDLM_ATTR(proto_name,		0444, proto_name_show,		NULL);
 GDLM_ATTR(block,		0644, block_show,		block_store);
-GDLM_ATTR(withdraw,		0644, withdraw_show,		withdraw_store);
+GDLM_ATTR(withdraw,		0644, wdack_show,		wdack_store);
 GDLM_ATTR(jid,			0644, jid_show,			jid_store);
 GDLM_ATTR(first,		0644, lkfirst_show,		lkfirst_store);
 GDLM_ATTR(first_done,		0444, first_done_show,		NULL);
diff --git a/fs/gfs2/util.c b/fs/gfs2/util.c
index f00d7c5..6402fb6 100644
--- a/fs/gfs2/util.c
+++ b/fs/gfs2/util.c
@@ -54,6 +54,9 @@ int gfs2_lm_withdraw(struct gfs2_sbd *sdp, char *fmt, ...)
 
 		kobject_uevent(&sdp->sd_kobj, KOBJ_OFFLINE);
 
+		if (!strcmp(sdp->sd_lockstruct.ls_ops->lm_proto_name, "lock_dlm"))
+			wait_for_completion(&sdp->sd_wdack);
+
 		if (lm->lm_unmount) {
 			fs_err(sdp, "telling LM to unmount\n");
 			lm->lm_unmount(sdp);
-- 
1.7.4