[PATCH 7/7] libxfs: reset dirty buffer priority on lookup

Dave Chinner david at fromorbit.com
Thu Feb 4 17:05:08 CST 2016


From: Dave Chinner <dchinner at redhat.com>

When a buffer on the dirty MRU is looked up and found, we remove the
buffer from the MRU. However, we've already set the priority ofthe
buffer to "dirty" so when we are donw with it it will go back on the
dirty buffer MRU regardless of whether it needs to or not.

Hence when we move a buffer to a the dirty MRU, reocrd the old
priority and restore it when we remove the buffer from the MRU on
lookup. This will prevent us from putting fixed, now writeable
buffers back on the dirty MRU and allow the cache routine to write,
shake and reclaim the buffers once they are clean.

Signed-off-by: Dave Chinner <dchinner at redhat.com>
---
 include/cache.h | 1 +
 libxfs/cache.c  | 7 +++++++
 2 files changed, 8 insertions(+)

diff --git a/include/cache.h b/include/cache.h
index 55761d3..3a989d7 100644
--- a/include/cache.h
+++ b/include/cache.h
@@ -99,6 +99,7 @@ struct cache_node {
 	unsigned int		cn_count;	/* reference count */
 	unsigned int		cn_hashidx;	/* hash chain index */
 	int			cn_priority;	/* priority, -1 = free list */
+	int			cn_old_priority;/* saved pre-dirty prio */
 	pthread_mutex_t		cn_mutex;	/* node mutex */
 };
 
diff --git a/libxfs/cache.c b/libxfs/cache.c
index d4b4a4e..0398be3 100644
--- a/libxfs/cache.c
+++ b/libxfs/cache.c
@@ -197,6 +197,7 @@ cache_move_to_dirty_mru(
 	mru = &cache->c_mrus[CACHE_DIRTY_PRIORITY];
 
 	pthread_mutex_lock(&mru->cm_mutex);
+	node->cn_old_priority = node->cn_priority;
 	node->cn_priority = CACHE_DIRTY_PRIORITY;
 	list_move(&node->cn_mru, &mru->cm_list);
 	mru->cm_count++;
@@ -325,6 +326,7 @@ cache_node_allocate(
 	list_head_init(&node->cn_mru);
 	node->cn_count = 1;
 	node->cn_priority = 0;
+	node->cn_old_priority = -1;
 	return node;
 }
 
@@ -434,6 +436,10 @@ cache_node_get(
 				mru->cm_count--;
 				list_del_init(&node->cn_mru);
 				pthread_mutex_unlock(&mru->cm_mutex);
+				if (node->cn_old_priority != -1) {
+					node->cn_priority = node->cn_old_priority;
+					node->cn_old_priority = -1;
+				}
 			}
 			node->cn_count++;
 
@@ -534,6 +540,7 @@ cache_node_set_priority(
 	pthread_mutex_lock(&node->cn_mutex);
 	ASSERT(node->cn_count > 0);
 	node->cn_priority = priority;
+	node->cn_old_priority = -1;
 	pthread_mutex_unlock(&node->cn_mutex);
 }
 
-- 
2.5.0



More information about the xfs mailing list