Re: [PATCH 0/2] Make task doing heavy writing killable

[Jan Kara <jack@suse.cz>]

On Mon 14-11-11 10:32:26, Christoph Hellwig wrote:
> On Mon, Nov 14, 2011 at 04:28:32PM +0100, Jan Kara wrote:
> > On Mon 14-11-11 21:01:53, Wu Fengguang wrote:
> > > On Mon, Nov 14, 2011 at 08:41:45PM +0800, Christoph Hellwig wrote:
> > > > On Mon, Nov 14, 2011 at 08:29:32PM +0800, Wu Fengguang wrote:
> > > > > > Can you make balance_dirty_pages(_ratelimited) return an error instead
> > > > > > of opencoding the fatal signal che
> > > > > 
> > > > > Hmm that means a bigger change to the prototype of the _exported_
> > > > > balance_dirty_pages_ratelimited_nr() which impacts out of tree FS... 
> > > > 
> > > > So what?  out of tree fses have always been entirely supported on their
> > > > own, and not a matter for keeping progress back.
> > > 
> > > OK. Jan, would you do the mm/*.c part in v2? I'll do the various FS
> > > parts.
> >   Sure, I'll update my patch to balance_dirty_pages(). Should I also update
> > my patch to generic_perform_write() or does that fall under "various FS
> > parts" cathegory?
> 
> I'd be much happier if all bits go in together.  IMHO this is a change
> worthwhile for 3.2 still, so we should probably try to get it in ASAP.
  OK, I've resent my patches with requested changes. I think those two
patches make sense on it's own and fix the biggest single pain point.
Fengguang can add other fs bits to these two patches (although personally
I'd be happier to have respective fs maintainers acks on them before
pushing).

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