Skip to contentSkip to conversation

move sync_supers to end of sync functions

PATCH New Andreas Dilger 1 participant 1 message Aug 1, 2001 VFS

lore
Share & tools
lore

Andreas Dilger

author
#0 lore raw

PATCH 1/1 move sync_supers to end of sync functions

Alan, Linus,
the following patch ensures that write_super() or sync_supers() is called _after_ other filesystem I/O is completed. The current order means that a superblock could be dirtied by other filesystem I/O after the superblock has been "synced". This has been part of the ext3 patches for a long time, but should really be done in the stock kernel sources as well.

Cheers, Andreas
========================================================================

fs/buffer.c +3 -3
diff -ru linux-2.4.7.orig/fs/buffer.c linux-2.4.7-aed/fs/buffer.c
@@ -310,11 +312,11 @@
 
 	lock_kernel();
 	sync_inodes_sb(sb);
+	DQUOT_SYNC(dev);
 	lock_super(sb);
 	if (sb->s_dirt && sb->s_op && sb->s_op->write_super)
 		sb->s_op->write_super(sb);
 	unlock_super(sb);
-	DQUOT_SYNC(dev);
 	unlock_kernel();
 
 	return sync_buffers(dev, 1);

@@ -325,9 +327,9 @@
 	sync_buffers(dev, 0);
 
 	lock_kernel();
-	sync_supers(dev);
 	sync_inodes(dev);
 	DQUOT_SYNC(dev);
+	sync_supers(dev);
 	unlock_kernel();
 
 	return sync_buffers(dev, 1);

@@ -2608,8 +2693,8 @@
 static int sync_old_buffers(void)
 {
 	lock_kernel();
-	sync_supers(0);
 	sync_unlocked_inodes();
+	sync_supers(0);
 	unlock_kernel();
 
 	flush_dirty_buffers(1);
Signature · 3 lines
Andreas Dilger  \ "If a man ate a pound of pasta and a pound of antipasto,
                 \  would they cancel out, leaving him still hungry?"
http://www-mddsp.enel.ucalgary.ca/People/adilger/               -- Dogbert