1 files changed, 23 insertions, 7 deletions

diff --git a/fs/buffer.c b/fs/buffer.c
index 826baf4f04bc..3ebccf4aa7e3 100644
--- a/fs/buffer.c
+++ b/fs/buffer.c
@@ -67,14 +67,14 @@ static int sync_buffer(void *word)
         return 0;
 }
 
-void fastcall __lock_buffer(struct buffer_head *bh)
+void __lock_buffer(struct buffer_head *bh)
 {
         wait_on_bit_lock(&bh->b_state, BH_Lock, sync_buffer,
                                                         TASK_UNINTERRUPTIBLE);
 }
 EXPORT_SYMBOL(__lock_buffer);
 
-void fastcall unlock_buffer(struct buffer_head *bh)
+void unlock_buffer(struct buffer_head *bh)
 {
         smp_mb__before_clear_bit();
         clear_buffer_locked(bh);
@@ -678,7 +678,7 @@ void mark_buffer_dirty_inode(struct buffer_head *bh, struct inode *inode)
         } else {
                 BUG_ON(mapping->assoc_mapping != buffer_mapping);
         }
-        if (list_empty(&bh->b_assoc_buffers)) {
+        if (!bh->b_assoc_map) {
                 spin_lock(&buffer_mapping->private_lock);
                 list_move_tail(&bh->b_assoc_buffers,
                                 &mapping->private_list);
@@ -794,6 +794,7 @@ static int fsync_buffers_list(spinlock_t *lock, struct list_head *list)
 {
         struct buffer_head *bh;
         struct list_head tmp;
+        struct address_space *mapping;
         int err = 0, err2;
 
         INIT_LIST_HEAD(&tmp);
@@ -801,9 +802,14 @@ static int fsync_buffers_list(spinlock_t *lock, struct list_head *list)
         spin_lock(lock);
         while (!list_empty(list)) {
                 bh = BH_ENTRY(list->next);
+                mapping = bh->b_assoc_map;
                 __remove_assoc_queue(bh);
+                /* Avoid race with mark_buffer_dirty_inode() which does
+                 * a lockless check and we rely on seeing the dirty bit */
+                smp_mb();
                 if (buffer_dirty(bh) || buffer_locked(bh)) {
                         list_add(&bh->b_assoc_buffers, &tmp);
+                        bh->b_assoc_map = mapping;
                         if (buffer_dirty(bh)) {
                                 get_bh(bh);
                                 spin_unlock(lock);
@@ -822,8 +828,17 @@ static int fsync_buffers_list(spinlock_t *lock, struct list_head *list)
 
         while (!list_empty(&tmp)) {
                 bh = BH_ENTRY(tmp.prev);
-                list_del_init(&bh->b_assoc_buffers);
                 get_bh(bh);
+                mapping = bh->b_assoc_map;
+                __remove_assoc_queue(bh);
+                /* Avoid race with mark_buffer_dirty_inode() which does
+                 * a lockless check and we rely on seeing the dirty bit */
+                smp_mb();
+                if (buffer_dirty(bh)) {
+                        list_add(&bh->b_assoc_buffers,
+                                 &bh->b_assoc_map->private_list);
+                        bh->b_assoc_map = mapping;
+                }
                 spin_unlock(lock);
                 wait_on_buffer(bh);
                 if (!buffer_uptodate(bh))
@@ -1164,7 +1179,7 @@ __getblk_slow(struct block_device *bdev, sector_t block, int size)
  * mark_buffer_dirty() is atomic.  It takes bh->b_page->mapping->private_lock,
  * mapping->tree_lock and the global inode_lock.
  */
-void fastcall mark_buffer_dirty(struct buffer_head *bh)
+void mark_buffer_dirty(struct buffer_head *bh)
 {
         WARN_ON_ONCE(!buffer_uptodate(bh));
         if (!buffer_dirty(bh) && !test_set_buffer_dirty(bh))
@@ -1195,7 +1210,7 @@ void __brelse(struct buffer_head * buf)
 void __bforget(struct buffer_head *bh)
 {
         clear_buffer_dirty(bh);
-        if (!list_empty(&bh->b_assoc_buffers)) {
+        if (bh->b_assoc_map) {
                 struct address_space *buffer_mapping = bh->b_page->mapping;
 
                 spin_lock(&buffer_mapping->private_lock);
@@ -1436,6 +1451,7 @@ void invalidate_bh_lrus(void)
 {
         on_each_cpu(invalidate_bh_lru, NULL, 1, 1);
 }
+EXPORT_SYMBOL_GPL(invalidate_bh_lrus);
 
 void set_bh_page(struct buffer_head *bh,
                 struct page *page, unsigned long offset)
@@ -3021,7 +3037,7 @@ drop_buffers(struct page *page, struct buffer_head **buffers_to_free)
         do {
                 struct buffer_head *next = bh->b_this_page;
 
-                if (!list_empty(&bh->b_assoc_buffers))
+                if (bh->b_assoc_map)
                         __remove_assoc_queue(bh);
                 bh = next;
         } while (bh != head);