lkml.org 
[lkml]   [2020]   [Apr]   [24]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH v2] block: Limit number of items taken from the I/O scheduler in one go
> index 74cedea56034..b69b780351c1 100644
> --- a/block/blk-mq-sched.c
> +++ b/block/blk-mq-sched.c
> @@ -84,12 +84,17 @@ void blk_mq_sched_restart(struct blk_mq_hw_ctx *hctx)
> * Only SCSI implements .get_budget and .put_budget, and SCSI restarts
> * its queue by itself in its completion handler, so we don't need to
> * restart queue if .get_budget() returns BLK_STS_NO_RESOURCE.
> + *
> + * Returns true if hctx->dispatch was found non-empty and
> + * run_work has to be run again. This is necessary to avoid
> + * starving flushes.
> */

Please use up all 80 chars for your comments (just like the existing
part of the comment).

> * Only SCSI implements .get_budget and .put_budget, and SCSI restarts
> * its queue by itself in its completion handler, so we don't need to
> * restart queue if .get_budget() returns BLK_STS_NO_RESOURCE.
> + *
> + * Returns true if hctx->dispatch was found non-empty and
> + * run_work has to be run again. This is necessary to avoid
> + * starving flushes.

Same here.

> +again:
> + run_again = false;
> +
> /*
> * If we have previous entries on our dispatch list, grab them first for
> * more fair dispatch.
> @@ -208,19 +236,28 @@ void blk_mq_sched_dispatch_requests(struct blk_mq_hw_ctx *hctx)
> blk_mq_sched_mark_restart_hctx(hctx);
> if (blk_mq_dispatch_rq_list(q, &rq_list, false)) {
> if (has_sched_dispatch)
> - blk_mq_do_dispatch_sched(hctx);
> + run_again = blk_mq_do_dispatch_sched(hctx);
> else
> - blk_mq_do_dispatch_ctx(hctx);
> + run_again = blk_mq_do_dispatch_ctx(hctx);
> }
> } else if (has_sched_dispatch) {
> - blk_mq_do_dispatch_sched(hctx);
> + run_again = blk_mq_do_dispatch_sched(hctx);
> } else if (hctx->dispatch_busy) {
> /* dequeue request one by one from sw queue if queue is busy */
> - blk_mq_do_dispatch_ctx(hctx);
> + run_again = blk_mq_do_dispatch_ctx(hctx);
> } else {
> blk_mq_flush_busy_ctxs(hctx, &rq_list);
> blk_mq_dispatch_rq_list(q, &rq_list, false);
> }
> +
> + if (run_again) {
> + if (!restarted) {
> + restarted = true;
> + goto again;
> + }
> +
> + blk_mq_run_hw_queue(hctx, true);
> + }

This is a weird loop. I'd split the code betweem the again label and
the run_again check here into a __blk_mq_sched_dispatch_requests
helper, and then you can do:

if (__blk_mq_sched_dispatch_requests()) {
if (__blk_mq_sched_dispatch_requests())
blk_mq_run_hw_queue(hctx, true);
}

here. Preferably with ha good comment explaining the logic.

\
 
 \ /
  Last update: 2020-04-24 08:16    [W:0.060 / U:0.332 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site