Re: RFC: Re: PATCH: udf fs corruption on linux-2.6

[Jan Kara <jack@suse.cz>]

On Thu 14-06-07 15:12:58, Linus Torvalds wrote:
> 
> 
> On Thu, 14 Jun 2007, Rich Coe wrote:
> >
> > I've updated the patch below to use drop_inode rather than put_inode.
> > 
> > drop_inode is only called when the last iput() reference to the inode is
> > released, where put_inode is called for every iput().
> 
> Patch looks fine, but this late in the -rc series, I'd really like to get 
> an ACK from Jan or somebody else, just to make sure there are no other 
> issues with it.
> 
> Jan? 
  My fix for this problem is already sitting in Andrew's patch queue
(http://lkml.org/lkml/2007/6/11/79). Rich's patch still has a problem - you
cannot call udf_discard_prealloc() from drop_inode() because it is called
under inode_lock and thus you cannot call e.g. mark_inode_dirty(). I've done
that mistake too ;). So please don't apply the patch.

								Honza
-- 
Jan Kara <jack@suse.cz>
SuSE CR Labs