Re: [PATCH] as i/o hang with aacraid driver 2.6.0-test1

[Jens Axboe <axboe@suse.de>]

On Thu, Jul 17 2003, Nick Piggin wrote:
> 
> 
> Jens Axboe wrote:
> 
> >On Thu, Jul 17 2003, Nick Piggin wrote:
> >
> >>
> >>Jens Axboe wrote:
> >>
> >>
> >>>On Thu, Jul 17 2003, Nick Piggin wrote:
> >>>
> >>>
> >>>>Jens Axboe wrote:
> >>>>
> >>>>
> >>>>
> >>>>>On Thu, Jul 17 2003, Andrew Morton wrote:
> >>>>>
> >>>>>
> >>>>>
> >>>>>>So this is what I ended up with.  Could we please have confirmation 
> >>>>>>that it
> >>>>>>fixes the aacraid hang?
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>>It doesn't, it's just a pre-requisite to fixing the bug :-)
> >>>>>
> >>>>>Nick should chime in with how we wants it to be handled from
> >>>>>blk_requeue_request(), he needs to decrease dispatched from there. We
> >>>>>could always add some hook for it of course, but...
> >>>>>
> >>>>>
> >>>>>
> >>>>Well you could just put an elv_completed_request in there, but
> >>>>
> >>>>
> >>>Yeah, I told Andrew to add that for now.
> >>>
> >>>
> >>>
> >>>>I suppose it really wants an elv_requeue_request - which would
> >>>>just default to elv_add_request for other schedulers.
> >>>>
> >>>>
> >>>I'd rather keep it seperate, ie just a requeue notifier. How's this?
> >>>
> >>>
> >>Well it would be much nicer for AS if it were seperate. Basically
> >>AS wants the requeue implemented as as_add_request but without
> >>the accounting updates: there has been no request completed, and
> >>no really new request.
> >>
> >>If the requeue were seperate to the add, it would simply be a call
> >>to as_completed_request in as-iosched.c.
> >>
> >
> >Like this then? Nicer semantics, too.
> >
> 
> Jens, this is exactly right. Would you be OK with this? Everyone else?

I'm happy with it.

-- 
Jens Axboe