Re: blk_throtl_exit taking q->queue_lock is problematic

[Vivek Goyal <vgoyal@redhat.com>]

On Mon, Feb 21, 2011 at 06:24:19PM +1100, NeilBrown wrote:
> On Fri, 18 Feb 2011 10:04:29 -0500 Vivek Goyal <vgoyal@redhat.com> wrote:
> 
> > On Fri, Feb 18, 2011 at 02:33:25PM +1100, NeilBrown wrote:
> > > On Thu, 17 Feb 2011 22:19:52 -0500 Mike Snitzer <snitzer@redhat.com> wrote:
> 
> > > > Noticed an inconsistency, raid10.c's additional locking also protects
> > > > the bio_list_add() whereas raid1.c's doesn't.  Seems the additional
> > > > protection in raid10 isn't needed?
> > > 
> > > Correct - not needed at all.
> > > I put it there because it felt a little cleaner keeping the two 'lock's
> > > together like the two 'unlock's.  Probably confusing though...
> > 
> > I guess you could use blk_plug_device_unlocked() to get rid of ugliness
> > and this routine will take care of taking queue lock.
> > 
> 
> Yep, that gets rid of some ugliness.
> I've made that change and will submit it in due course.
> So blk_throtl doesn't need any change to avoid the problem with md - that
> changes are made in md instead.

Thanks Neil. I might still end up moving blk_throtl_exit() to
blk_cleanup_queue() once I have sorted out blk_sync_queue(). Because
at the end of blk_cleanup_queue() driver is free to release the spin
lock and there are no gurantees that in blk_release_queue() lock is 
still there.

Thanks
Vivek