[Cluster-devel] [GFS2 PATCH 2/2] GFS2: Submit all ordered writes in bulk, wait after that

[Bob Peterson <rpeterso@redhat.com>]

Before this patch, the ordered_write function would submit all
the ordered writes with filemap_fdatawrite, which waited for each
one to complete before moving on to the next. This patch allows it
to submit them all, then wait for them after they're submitted.

Signed-off-by: Bob Peterson <rpeterso@redhat.com>
---
 fs/gfs2/log.c | 34 ++++++++++++++++++++++++----------
 1 file changed, 24 insertions(+), 10 deletions(-)

diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c
index bedd1c056ca2..b05d0fbc3d05 100644
--- a/fs/gfs2/log.c
+++ b/fs/gfs2/log.c
@@ -22,6 +22,7 @@
 #include <linux/blkdev.h>
 #include <linux/writeback.h>
 #include <linux/list_sort.h>
+#include <linux/fs.h>
 
 #include "gfs2.h"
 #include "incore.h"
@@ -530,15 +531,20 @@ static int ip_cmp(void *priv, struct list_head *a, struct list_head *b)
 	return 0;
 }
 
-static void gfs2_ordered_wait(struct gfs2_sbd *sdp)
+static void gfs2_ordered_wait(struct gfs2_sbd *sdp, struct list_head *src,
+			      struct list_head *dst)
 {
 	struct gfs2_inode *ip;
 
 	spin_lock(&sdp->sd_ordered_lock);
-	while (!list_empty(&sdp->sd_log_le_ordered)) {
-		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));
+	while (!list_empty(src)) {
+		ip = list_entry(src->next, struct gfs2_inode, i_ordered);
+		if (dst)
+			list_move_tail(&ip->i_ordered, dst);
+		else {
+			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;
 		spin_unlock(&sdp->sd_ordered_lock);
@@ -551,24 +557,32 @@ static void gfs2_ordered_wait(struct gfs2_sbd *sdp)
 static void gfs2_ordered_write(struct gfs2_sbd *sdp)
 {
 	struct gfs2_inode *ip;
+	struct address_space *mapping;
 	LIST_HEAD(written);
+	struct writeback_control wbc = {
+		.sync_mode = WB_SYNC_NONE,
+		.nr_to_write = LONG_MAX,
+		.range_start = 0,
+		.range_end = LLONG_MAX,
+	};
 
 	spin_lock(&sdp->sd_ordered_lock);
 	list_sort(NULL, &sdp->sd_log_le_ordered, &ip_cmp);
 	while (!list_empty(&sdp->sd_log_le_ordered)) {
 		ip = list_entry(sdp->sd_log_le_ordered.next, struct gfs2_inode, i_ordered);
-		if (ip->i_inode.i_mapping->nrpages == 0) {
+		mapping = ip->i_inode.i_mapping;
+		if (mapping->nrpages == 0) {
 			test_and_clear_bit(GIF_ORDERED, &ip->i_flags);
 			list_del(&ip->i_ordered);
 			continue;
 		}
-		list_move(&ip->i_ordered, &written);
+		list_move_tail(&ip->i_ordered, &written);
 		spin_unlock(&sdp->sd_ordered_lock);
-		filemap_fdatawrite(ip->i_inode.i_mapping);
+		mapping->a_ops->writepages(mapping, &wbc);
 		spin_lock(&sdp->sd_ordered_lock);
 	}
-	list_splice(&written, &sdp->sd_log_le_ordered);
 	spin_unlock(&sdp->sd_ordered_lock);
+	gfs2_ordered_wait(sdp, &written, &sdp->sd_log_le_ordered);
 }
 
 void gfs2_ordered_del_inode(struct gfs2_inode *ip)
@@ -741,7 +755,7 @@ static void log_write_header(struct gfs2_sbd *sdp, u32 flags)
 	tail = current_tail(sdp);
 
 	if (test_bit(SDF_NOBARRIERS, &sdp->sd_flags)) {
-		gfs2_ordered_wait(sdp);
+		gfs2_ordered_wait(sdp, &sdp->sd_log_le_ordered, NULL);
 		log_flush_wait(sdp);
 		op_flags = REQ_SYNC | REQ_META | REQ_PRIO;
 	}
-- 
2.17.0