Re: 2.6.15-mm3 bisection: git-xfs.patch makes reiserfs oops

[Nathan Scott <nathans@sgi.com>]

On Sun, Jan 15, 2006 at 11:14:58PM +0100, Mattia Dongili wrote:
> [CC-in relevant people/ML]
> 
> Hello!
> 
> second bisection result!
> 
> On Tue, Jan 10, 2006 at 05:00:37PM -0800, Andrew Morton wrote:
> > Mattia Dongili <malattia@linux.it> wrote:
> [...]
> > > 1- reiser3 oopsed[1] twice while suspending to ram. It seems
> > >    reproducible (have some activity on the fs and suspend)
> > 
> > No significant reiser3 changes in there, so I'd be suspecting something
> > else has gone haywire.
> 
> you're right: git-xfs.patch is the bad guy.
> 
> Unfortunately netconsole isn't helpful in capturing the oops (no serial
> ports here) but I have two more shots (more readable):
> http://oioio.altervista.org/linux/dsc03148.jpg
> http://oioio.altervista.org/linux/dsc03149.jpg

Hmm, thats odd.  It seems to be coming from:
reiserfs_commit_page -> reiserfs_add_ordered_list -> __add_jh(inline)

I guess XFS may have left a buffer_head in an unusual state (with some
private flag/b_private set), somehow, and perhaps that buffer_head has
later been allocated for a page in a reiserfs write.  Does this patch,
below, help at all?

I see one BUG check in __add_jh for non-NULL b_private, but can't see
the top of your console output from the photos - is there a preceding
line with "kernel BUG at ..." in it?

cheers.

-- 
Nathan


--- fs/buffer.c.orig	2006-01-16 10:15:01.332010750 +1100
+++ fs/buffer.c	2006-01-16 10:18:15.276131500 +1100
@@ -1027,7 +1027,7 @@ try_again:
 		/* Link the buffer to its page */
 		set_bh_page(bh, page, offset);
 
-		bh->b_end_io = NULL;
+		init_buffer(bh, NULL, NULL);
 	}
 	return head;
 /*
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/