On Wednesday 13 February 2008 20:01, Andrew Morton wrote:
> On Wed, 13 Feb 2008 08:26:27 +0100 Bart Dopheide <[EMAIL PROTECTED]> wrote:
> > On Wed, Feb 13, 2008 at 12:05:45PM +1100, Nick Piggin wrote:
> > :)On Wednesday 13 February 2008 08:50, Alan Cox wrote:
> > :)> Almost certainly a hardware fail of some sort.
> > :)
> > :)Right, but the kernel shouldn't go bug...
> >
> > Indeed, that's why I'm reporting.
> >
> > :)I don't have a copy of your exact source code... which condition in
> > :)__mpage_writepage went BUG?
> >
> > BUG_ON(buffer_locked(bh));
> >
> > In a bit of context:
> > 482: if (page_has_buffers(page)) {
> > 483: struct buffer_head *head = page_buffers(page);
> > 484: struct buffer_head *bh = head;
> > 485:
> > 486: /* If they're all mapped and dirty, do it */
> > 487: page_block = 0;
> > 488: do {
> > 489: BUG_ON(buffer_locked(bh));
> > 490: if (!buffer_mapped(bh)) {
> > 491: /*
> > 492: * unmapped dirty buffers are created by
> > 493: * __set_page_dirty_buffers -> mmapped
> > data 494: */
> > 495: if (buffer_dirty(bh))
> > 496: goto confused;
> > 497: if (first_unmapped == blocks_per_page)
> > 498: first_unmapped = page_block;
> > 499: continue;
> > 500: }
>
> Probably means that either fat, IDE, block or fs/buffer.c failed to unlock
> a buffer_head when the IO error happened. It's unlikely to be fat.
Yes that looks like it would be the problem. I can't really
see anything in buffer.c that would do it...
BTW is it really true that the buffer can never be locked by
anything else at this point? What about fsync_buffers_list?
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
