[dm-devel] Re: [PATCH 2.6.19 5/5] fs: freeze_bdev with semaphore not mutex

["Rafael J. Wysocki" <rjw sisk pl>]

On Wednesday, 8 November 2006 15:25, Alasdair G Kergon wrote:
> On Wed, Nov 08, 2006 at 07:27:22PM +1100, David Chinner wrote:
> > But it's trivial to detect this condition - if (sb->s_frozen != SB_UNFROZEN)
> > then the filesystem is already frozen and you shouldn't try to freeze
> > it again. It's simple to do, and the whole problem then just goes away....
>  
> So is that another vote in support of explicitly supporting multiple concurrent
> freeze requests, letting them all succeed, and only thawing after the last one
> has requested its thaw?  (It's not enough just to check SB_UNFROZEN - also need
> to track whether any other outstanding requests to avoid risk of it getting
> unfrozen while something independent believes it still to be frozen.)

So, I think, we need the following patch to fix freeze_filesystems().

Will it be enough to cover the interactions with dm?

Rafael


Signed-off-by: Rafael J. Wysocki <rjw sisk pl>
---
 fs/buffer.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Index: linux-2.6.19-rc5-mm1/fs/buffer.c
===================================================================
--- linux-2.6.19-rc5-mm1.orig/fs/buffer.c
+++ linux-2.6.19-rc5-mm1/fs/buffer.c
@@ -264,7 +264,7 @@ void freeze_filesystems(void)
 	 */
 	list_for_each_entry_reverse(sb, &super_blocks, s_list) {
 		if (!sb->s_root || !sb->s_bdev ||
-		    (sb->s_frozen == SB_FREEZE_TRANS) ||
+		    (sb->s_frozen != SB_UNFROZEN) ||
 		    (sb->s_flags & MS_RDONLY) ||
 		    (sb->s_flags & MS_FROZEN))
 			continue;