From: Steven Whitehouse <swhiteho@redhat.com>

The following patch removes the ordered write processing from
databuf_lo_before_commit() and moves it to log.c. This has the effect of
greatly simplyfying databuf_lo_before_commit() and well as potentially
making the ordered write code more efficient.

As a side effect of this, its now possible to remove ordered buffers
from the ordered buffer list at any time, so we now make use of this in
invalidatepage and releasepage to ensure timely release of these
buffers.

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

diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index 23b611a..388dc1b 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -612,13 +612,13 @@ struct gfs2_sbd {
 	unsigned int sd_log_num_revoke;
 	unsigned int sd_log_num_rg;
 	unsigned int sd_log_num_databuf;
-	unsigned int sd_log_num_jdata;
 
 	struct list_head sd_log_le_gl;
 	struct list_head sd_log_le_buf;
 	struct list_head sd_log_le_revoke;
 	struct list_head sd_log_le_rg;
 	struct list_head sd_log_le_databuf;
+	struct list_head sd_log_le_ordered;
 
 	unsigned int sd_log_blks_free;
 	struct mutex sd_log_reserve_mutex;
diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c
index d8232ec..20fa528 100644
--- a/fs/gfs2/log.c
+++ b/fs/gfs2/log.c
@@ -620,6 +620,57 @@ static void log_flush_commit(struct gfs2_sbd *sdp)
 	}
 }
 
+static void gfs2_ordered_write(struct gfs2_sbd *sdp)
+{
+	struct gfs2_bufdata *bd;
+	struct buffer_head *bh;
+	LIST_HEAD(written);
+
+	gfs2_log_lock(sdp);
+	while (!list_empty(&sdp->sd_log_le_ordered)) {
+		bd = list_entry(sdp->sd_log_le_ordered.next, struct gfs2_bufdata, bd_le.le_list);
+		list_move(&bd->bd_le.le_list, &written);
+		bh = bd->bd_bh;
+		if (!buffer_dirty(bh))
+			continue;
+		get_bh(bh);
+		gfs2_log_unlock(sdp);
+		lock_buffer(bh);
+		if (test_clear_buffer_dirty(bh)) {
+			bh->b_end_io = end_buffer_write_sync;
+			submit_bh(WRITE, bh);
+		} else ...
[ message continues ]

From: Bob Peterson <rpeterso@redhat.com>

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 1e56f4d..93fa427 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -93,10 +93,10 @@ static int gfs2_unstuffer_page(struct gfs2_inode *ip, struct buffer_head *dibh,
 		map_bh(bh, inode->i_sb, block);
 
 	set_buffer_uptodate(bh);
-	if (sdp->sd_args.ar_data == GFS2_DATA_ORDERED || gfs2_is_jdata(ip))
-		gfs2_trans_add_bh(ip->i_gl, bh, 0);
 	if (!gfs2_is_jdata(ip))
 		mark_buffer_dirty(bh);
+	if (sdp->sd_args.ar_data == GFS2_DATA_ORDERED || gfs2_is_jdata(ip))
+		gfs2_trans_add_bh(ip->i_gl, bh, 0);
 
 	if (release) {
 		unlock_page(page);
-- 
1.5.1.2

-

[ message continues ]

From: Steven Whitehouse <swhiteho@redhat.com>

Both the revoke structure and the bufdata structure are quite similar.
They are basically small tags which are put on lists. In addition to
which the revoke structure is always allocated when there is a bufdata
structure which is (or can be) freed. As such it should be possible to
reduce the number of frees and allocations by using the same structure
for both purposes.

This patch is the first step along that path. It replaces existing uses
of the revoke structure with the bufdata structure.

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

diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index 388dc1b..8aa5780 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -114,7 +114,13 @@ struct gfs2_bufdata {
 	struct buffer_head *bd_bh;
 	struct gfs2_glock *bd_gl;
 
-	struct list_head bd_list_tr;
+	union {
+		struct list_head list_tr;
+		u64 blkno;
+	} u;
+#define bd_list_tr u.list_tr
+#define bd_blkno u.blkno
+
 	struct gfs2_log_element bd_le;
 
 	struct gfs2_ail *bd_ail;
@@ -298,11 +304,6 @@ struct gfs2_file {
 	struct gfs2_holder f_fl_gh;
 };
 
-struct gfs2_revoke {
-	struct gfs2_log_element rv_le;
-	u64 rv_blkno;
-};
-
 struct gfs2_revoke_replay {
 	struct list_head rr_list;
 	u64 rr_blkno;
diff --git a/fs/gfs2/lops.c b/fs/gfs2/lops.c
index 7e2d4e6..cf6fe36 100644
--- a/fs/gfs2/lops.c
+++ b/fs/gfs2/lops.c
@@ -357,7 +357,7 @@ static void revoke_lo_before_commit(struct gfs2_sbd *sdp)
 	struct buffer_head *bh;
 	unsigned int offset;
 	struct list_head *head = &sdp->sd_log_le_revoke;
-	struct gfs2_revoke *rv;
+	struct gfs2_bufdata *bd;
 
 	if (!sdp->sd_log_num_revoke)
 		return;
@@ -376,8 +376,8 @@ static void revoke_lo_before_commit(struct gfs2_sbd *sdp)
 	offset = sizeof(struct gfs2_log_descriptor);
 
 	while (!list_empty(head)) {
-		rv = list_entry(head->next, struct gfs2_revoke, rv_le.le_list);
-		list_del_init(&rv->rv_le.le_list);
+		bd = list_entry(head->next, struct gfs2_bufdata, ...
[ message continues ]

From: Steven Whitehouse <swhiteho@redhat.com>

The old revoke structure was allocated using kalloc/kfree but
there is a slab cache for gfs2_bufdata, so we should use that
now that the structures have been converted.

This is part two of the patch series to merge the revoke
and gfs2_bufdata structures.

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

diff --git a/fs/gfs2/lops.c b/fs/gfs2/lops.c
index cf6fe36..4cbef4c 100644
--- a/fs/gfs2/lops.c
+++ b/fs/gfs2/lops.c
@@ -393,7 +393,7 @@ static void revoke_lo_before_commit(struct gfs2_sbd *sdp)
 		}
 
 		*(__be64 *)(bh->b_data + offset) = cpu_to_be64(bd->bd_blkno);
-		kfree(bd);
+		kmem_cache_free(gfs2_bufdata_cachep, bd);
 
 		offset += sizeof(u64);
 	}
diff --git a/fs/gfs2/trans.c b/fs/gfs2/trans.c
index eadf96e..01cc27f 100644
--- a/fs/gfs2/trans.c
+++ b/fs/gfs2/trans.c
@@ -144,7 +144,7 @@ void gfs2_trans_add_bh(struct gfs2_glock *gl, struct buffer_head *bh, int meta)
 
 void gfs2_trans_add_revoke(struct gfs2_sbd *sdp, u64 blkno)
 {
-	struct gfs2_bufdata *bd = kmalloc(sizeof(struct gfs2_bufdata),
+	struct gfs2_bufdata *bd = kmem_cache_alloc(gfs2_bufdata_cachep,
 					  GFP_NOFS | __GFP_NOFAIL);
 	lops_init_le(&bd->bd_le, &gfs2_revoke_lops);
 	bd->bd_blkno = blkno;
@@ -172,7 +172,7 @@ void gfs2_trans_add_unrevoke(struct gfs2_sbd *sdp, u64 blkno)
 
 	if (found) {
 		struct gfs2_trans *tr = current->journal_info;
-		kfree(bd);
+		kmem_cache_free(gfs2_bufdata_cachep, bd);
 		tr->tr_num_revoke_rm++;
 	}
 }
-- 
1.5.1.2

-

[ message continues ]

From: Steven Whitehouse <swhiteho@redhat.com>

The following alters gfs2_trans_add_revoke() to take a struct
gfs2_bufdata as an argument. This eliminates the memory allocation which
was previously required by making use of the already existing struct
gfs2_bufdata. It makes some sanity checks to ensure that the
gfs2_bufdata has been removed from all the lists before its recycled as
a revoke structure. This saves one memory allocation and one free per
revoke structure.

Also as a result, and to simplify the locking, since there is no longer
any blocking code in gfs2_trans_add_revoke() we must hold the log lock
whenever this function is called. This reduces the amount of times we
take and unlock the log lock.

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

diff --git a/fs/gfs2/glops.c b/fs/gfs2/glops.c
index b17346a..4670dcb 100644
--- a/fs/gfs2/glops.c
+++ b/fs/gfs2/glops.c
@@ -41,7 +41,6 @@ static void gfs2_ail_empty_gl(struct gfs2_glock *gl)
 	struct list_head *head = &gl->gl_ail_list;
 	struct gfs2_bufdata *bd;
 	struct buffer_head *bh;
-	u64 blkno;
 	int error;
 
 	blocks = atomic_read(&gl->gl_ail_count);
@@ -57,15 +56,12 @@ static void gfs2_ail_empty_gl(struct gfs2_glock *gl)
 		bd = list_entry(head->next, struct gfs2_bufdata,
 				bd_ail_gl_list);
 		bh = bd->bd_bh;
-		blkno = bh->b_blocknr;
-		gfs2_assert_withdraw(sdp, !buffer_busy(bh));
-
 		gfs2_remove_from_ail(NULL, bd);
-		gfs2_log_unlock(sdp);
-
-		gfs2_trans_add_revoke(sdp, blkno);
-
-		gfs2_log_lock(sdp);
+		bd->bd_bh = NULL;
+		bh->b_private = NULL;
+		bd->bd_blkno = bh->b_blocknr;
+		gfs2_assert_withdraw(sdp, !buffer_busy(bh));
+		gfs2_trans_add_revoke(sdp, bd);
 	}
 	gfs2_assert_withdraw(sdp, !atomic_read(&gl->gl_ail_count));
 	gfs2_log_unlock(sdp);
diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c
index 20fa528..4d04e6f 100644
--- a/fs/gfs2/log.c
+++ b/fs/gfs2/log.c
@@ -71,8 +71,8 @@ unsigned int gfs2_struct2blk(struct gfs2_sbd *sdp, unsigned int nstruct,
 void ...
[ message continues ]