Re: virtio ring cleanups, which save stack on older gcc

[Dave Chinner <david@fromorbit.com>]

On Thu, May 29, 2014 at 04:41:17PM +0900, Minchan Kim wrote:
> Hello Rusty,
> 
> On Thu, May 29, 2014 at 04:56:41PM +0930, Rusty Russell wrote:
> > They don't make much difference: the easier fix is use gcc 4.8
> > which drops stack required across virtio block's virtio_queue_rq
> > down to that kmalloc in virtio_ring from 528 to 392 bytes.
> > 
> > Still, these (*lightly tested*) patches reduce to 432 bytes,
> > even for gcc 4.6.4.  Posted here FYI.
> 
> I am testing with below which was hack for Dave's idea so don't have
> a machine to test your patches until tomorrow.
> So, I will queue your patches into testing machine tomorrow morning.
> 
> Thanks!
> 
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index f5c6635b806c..95f169e85dbe 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -4241,10 +4241,13 @@ EXPORT_SYMBOL_GPL(yield_to);
>  void __sched io_schedule(void)
>  {
>  	struct rq *rq = raw_rq();
> +	struct blk_plug *plug = current->plug;
>  
>  	delayacct_blkio_start();
>  	atomic_inc(&rq->nr_iowait);
> -	blk_flush_plug(current);
> +	if (plug)
> +		blk_flush_plug_list(plug, true);
> +

Could simply be

-	blk_flush_plug(current);
+	blk_schedule_flush_plug(current);

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>