From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id 87FD9C6FD1F for ; Fri, 10 Mar 2023 14:30:19 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S232312AbjCJOaS (ORCPT ); Fri, 10 Mar 2023 09:30:18 -0500 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:55752 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S232316AbjCJO37 (ORCPT ); Fri, 10 Mar 2023 09:29:59 -0500 Received: from ams.source.kernel.org (ams.source.kernel.org [145.40.68.75]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id BB03BB1A58 for ; Fri, 10 Mar 2023 06:28:41 -0800 (PST) Received: from smtp.kernel.org (relay.kernel.org [52.25.139.140]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ams.source.kernel.org (Postfix) with ESMTPS id 3D1F6B822DA for ; Fri, 10 Mar 2023 14:28:40 +0000 (UTC) Received: by smtp.kernel.org (Postfix) with ESMTPSA id 7B10AC433D2; Fri, 10 Mar 2023 14:28:38 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=linuxfoundation.org; s=korg; t=1678458518; bh=Aypx1r3foT6OwPO47HyI0kBOjtBpqPhKqDWI3/ML9iA=; h=From:To:Cc:Subject:Date:In-Reply-To:References:From; b=kYLc8G84WlglaR5iCRBNTyi9D2mdle/0OT6LE7XilbLkO9JSP1uZYw1juiwzEj8Vr p5bxTKAtNvn1mXNXC02qVNJuCtVSoMvmmxeuk0ly2tOYbzs8ISGb5N08aYeSYS7Azs 6Qym2tw8789KBAcHMImOXqVqPW0yaSSGT4Pd+a6k= From: Greg Kroah-Hartman To: stable@vger.kernel.org Cc: Greg Kroah-Hartman , patches@lists.linux.dev, Salman Qazi , Ming Lei , Jens Axboe , Sasha Levin Subject: [PATCH 5.4 023/357] block: Limit number of items taken from the I/O scheduler in one go Date: Fri, 10 Mar 2023 14:35:12 +0100 Message-Id: <20230310133735.020880330@linuxfoundation.org> X-Mailer: git-send-email 2.39.2 In-Reply-To: <20230310133733.973883071@linuxfoundation.org> References: <20230310133733.973883071@linuxfoundation.org> User-Agent: quilt/0.67 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Precedence: bulk List-ID: X-Mailing-List: stable@vger.kernel.org From: Salman Qazi [ Upstream commit 28d65729b050977d8a9125e6726871e83bd22124 ] Flushes bypass the I/O scheduler and get added to hctx->dispatch in blk_mq_sched_bypass_insert. This can happen while a kworker is running hctx->run_work work item and is past the point in blk_mq_sched_dispatch_requests where hctx->dispatch is checked. The blk_mq_do_dispatch_sched call is not guaranteed to end in bounded time, because the I/O scheduler can feed an arbitrary number of commands. Since we have only one hctx->run_work, the commands waiting in hctx->dispatch will wait an arbitrary length of time for run_work to be rerun. A similar phenomenon exists with dispatches from the software queue. The solution is to poll hctx->dispatch in blk_mq_do_dispatch_sched and blk_mq_do_dispatch_ctx and return from the run_work handler and let it rerun. Signed-off-by: Salman Qazi Reviewed-by: Ming Lei Signed-off-by: Jens Axboe Stable-dep-of: c31e76bcc379 ("blk-mq: remove stale comment for blk_mq_sched_mark_restart_hctx") Signed-off-by: Sasha Levin --- block/blk-mq-sched.c | 64 +++++++++++++++++++++++++++++++++++--------- 1 file changed, 51 insertions(+), 13 deletions(-) diff --git a/block/blk-mq-sched.c b/block/blk-mq-sched.c index f422c7feea7e0..2e1afd3559b28 100644 --- a/block/blk-mq-sched.c +++ b/block/blk-mq-sched.c @@ -93,12 +93,16 @@ 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 -EAGAIN if hctx->dispatch was found non-empty and run_work has to + * be run again. This is necessary to avoid starving flushes. */ -static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx) +static int blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx) { struct request_queue *q = hctx->queue; struct elevator_queue *e = q->elevator; LIST_HEAD(rq_list); + int ret = 0; do { struct request *rq; @@ -106,6 +110,11 @@ static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx) if (e->type->ops.has_work && !e->type->ops.has_work(hctx)) break; + if (!list_empty_careful(&hctx->dispatch)) { + ret = -EAGAIN; + break; + } + if (!blk_mq_get_dispatch_budget(hctx)) break; @@ -122,6 +131,8 @@ static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx) */ list_add(&rq->queuelist, &rq_list); } while (blk_mq_dispatch_rq_list(q, &rq_list, true)); + + return ret; } static struct blk_mq_ctx *blk_mq_next_ctx(struct blk_mq_hw_ctx *hctx, @@ -139,16 +150,25 @@ static struct blk_mq_ctx *blk_mq_next_ctx(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 -EAGAIN if hctx->dispatch was found non-empty and run_work has to + * to be run again. This is necessary to avoid starving flushes. */ -static void blk_mq_do_dispatch_ctx(struct blk_mq_hw_ctx *hctx) +static int blk_mq_do_dispatch_ctx(struct blk_mq_hw_ctx *hctx) { struct request_queue *q = hctx->queue; LIST_HEAD(rq_list); struct blk_mq_ctx *ctx = READ_ONCE(hctx->dispatch_from); + int ret = 0; do { struct request *rq; + if (!list_empty_careful(&hctx->dispatch)) { + ret = -EAGAIN; + break; + } + if (!sbitmap_any_bit_set(&hctx->ctx_map)) break; @@ -174,21 +194,17 @@ static void blk_mq_do_dispatch_ctx(struct blk_mq_hw_ctx *hctx) } while (blk_mq_dispatch_rq_list(q, &rq_list, true)); WRITE_ONCE(hctx->dispatch_from, ctx); + return ret; } -void blk_mq_sched_dispatch_requests(struct blk_mq_hw_ctx *hctx) +int __blk_mq_sched_dispatch_requests(struct blk_mq_hw_ctx *hctx) { struct request_queue *q = hctx->queue; struct elevator_queue *e = q->elevator; const bool has_sched_dispatch = e && e->type->ops.dispatch_request; + int ret = 0; LIST_HEAD(rq_list); - /* RCU or SRCU read lock is needed before checking quiesced flag */ - if (unlikely(blk_mq_hctx_stopped(hctx) || blk_queue_quiesced(q))) - return; - - hctx->run++; - /* * If we have previous entries on our dispatch list, grab them first for * more fair dispatch. @@ -217,19 +233,41 @@ 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); + ret = blk_mq_do_dispatch_sched(hctx); else - blk_mq_do_dispatch_ctx(hctx); + ret = blk_mq_do_dispatch_ctx(hctx); } } else if (has_sched_dispatch) { - blk_mq_do_dispatch_sched(hctx); + ret = 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); + ret = blk_mq_do_dispatch_ctx(hctx); } else { blk_mq_flush_busy_ctxs(hctx, &rq_list); blk_mq_dispatch_rq_list(q, &rq_list, false); } + + return ret; +} + +void blk_mq_sched_dispatch_requests(struct blk_mq_hw_ctx *hctx) +{ + struct request_queue *q = hctx->queue; + + /* RCU or SRCU read lock is needed before checking quiesced flag */ + if (unlikely(blk_mq_hctx_stopped(hctx) || blk_queue_quiesced(q))) + return; + + hctx->run++; + + /* + * A return of -EAGAIN is an indication that hctx->dispatch is not + * empty and we must run again in order to avoid starving flushes. + */ + if (__blk_mq_sched_dispatch_requests(hctx) == -EAGAIN) { + if (__blk_mq_sched_dispatch_requests(hctx) == -EAGAIN) + blk_mq_run_hw_queue(hctx, true); + } } bool blk_mq_sched_try_merge(struct request_queue *q, struct bio *bio, -- 2.39.2