lkml.org 
[lkml]   [2009]   [Jun]   [17]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 07/11] vfs: Unmap underlying metadata of new data buffers only when buffer is mapped
On Wed, Jun 17, 2009 at 02:05:20PM +0200, Jan Kara wrote:
> On Wed 17-06-09 12:35:43, Nick Piggin wrote:
> > On Mon, Jun 15, 2009 at 07:59:54PM +0200, Jan Kara wrote:
> > > When we do delayed allocation of some buffer, we want to signal to VFS that
> > > the buffer is new (set buffer_new) so that it properly zeros out everything.
> > > But we don't have the buffer mapped yet so we cannot really unmap underlying
> > > metadata in this state. Make VFS avoid doing unmapping of metadata when the
> > > buffer is not yet mapped.
> >
> > Is this a seperate bugfix for delalloc filesystems? What is the error
> > case of attempting to unmap underlying metadata of non mapped buffer?
> > Won't translate to a serious bug will it?
> If you do unmap_underlying_metadata on !mapped buffer, the kernel will
> oops because it will try to dereference bh->b_bdev which is NULL. Ext4 or
> XFS workaround this issue by setting b_bdev to the real device and b_blocknr
> to ~0 so unmap_underlying_metadata does not oops. As I didn't want to do
> the same hack in ext3, I need this patch...

OK, just trying to understand the patchset. It would be nice to
merge this ASAP as well and remove the ext4 and xfs hacks.


> You're right it's not directly connected with the mkwrite problem and
> can go in separately. Given how late it is, I'd like to get patch number 2
> reviewed (generic mkwrite changes), so that it can go together with patch
> number 4 (ext4 fixes) in the current merge window. The rest is not that
> urgent since it's not oopsable and you can hit it only when running out
> of space (or hitting quota limit)...

Sorry I was so late with looking at it. I am reading it now though
(especially #2) ;)



\
 
 \ /
  Last update: 2009-06-17 15:57    [from the cache]
©2003-2011 Jasper Spaans