Re: [PATCH 2/5] sched: always use blk_schedule_flush_plug in io_schedule_out

[Shaohua Li <shli@fb.com>]

On Fri, May 01, 2015 at 07:14:10PM +0200, Christoph Hellwig wrote:
> On Thu, Apr 30, 2015 at 10:45:15AM -0700, Shaohua Li wrote:
> >  long __sched io_schedule_timeout(long timeout)
> >  {
> > -	int old_iowait = current->in_iowait;
> >  	struct rq *rq;
> >  	long ret;
> >  
> >  	current->in_iowait = 1;
> > -	if (old_iowait)
> > -		blk_schedule_flush_plug(current);
> > -	else
> > -		blk_flush_plug(current);
> > +	blk_schedule_flush_plug(current);
> >  
> >  	delayacct_blkio_start();
> >  	rq = raw_rq();
> >  	atomic_inc(&rq->nr_iowait);
> >  	ret = schedule_timeout(timeout);
> > -	current->in_iowait = old_iowait;
> > +	current->in_iowait = 0;
> 
> Always clearing ->in_iowait is behavior change not mentioned in the
> changelog.  Even if it was intentional it would better be done in
> a separate patch.

9cff8adeaa34b5d28 changes to use blk_schedule_flush_plug for recursive
io_schedule with old_iowait for recurse detection. Since we always use
blk_schedule_flush_plug now, I thought we don't need the recurse
detection any more. I agree it's better to do it in another patch.

Thanks,
Shaohua