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 X-Spam-Level: X-Spam-Status: No, score=-7.0 required=3.0 tests=HEADER_FROM_DIFFERENT_DOMAINS, INCLUDES_PATCH,MAILING_LIST_MULTI,SIGNED_OFF_BY,SPF_PASS autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 03BD2C43387 for ; Mon, 17 Dec 2018 10:43:33 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id BBBE52145D for ; Mon, 17 Dec 2018 10:43:32 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1732188AbeLQKnc (ORCPT ); Mon, 17 Dec 2018 05:43:32 -0500 Received: from mx1.redhat.com ([209.132.183.28]:52654 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1726706AbeLQKnb (ORCPT ); Mon, 17 Dec 2018 05:43:31 -0500 Received: from smtp.corp.redhat.com (int-mx04.intmail.prod.int.phx2.redhat.com [10.5.11.14]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by mx1.redhat.com (Postfix) with ESMTPS id 3313F8B10C; Mon, 17 Dec 2018 10:43:31 +0000 (UTC) Received: from localhost (ovpn-8-31.pek2.redhat.com [10.72.8.31]) by smtp.corp.redhat.com (Postfix) with ESMTP id 308CD5DA2A; Mon, 17 Dec 2018 10:43:20 +0000 (UTC) From: Ming Lei To: Jens Axboe Cc: linux-block@vger.kernel.org, Ming Lei , Jeff Moyer , Mike Snitzer , Christoph Hellwig Subject: [PATCH V2 3/4] blk-mq: fix dispatch from sw queue Date: Mon, 17 Dec 2018 18:42:47 +0800 Message-Id: <20181217104248.5828-4-ming.lei@redhat.com> In-Reply-To: <20181217104248.5828-1-ming.lei@redhat.com> References: <20181217104248.5828-1-ming.lei@redhat.com> X-Scanned-By: MIMEDefang 2.79 on 10.5.11.14 X-Greylist: Sender IP whitelisted, not delayed by milter-greylist-4.5.16 (mx1.redhat.com [10.5.110.28]); Mon, 17 Dec 2018 10:43:31 +0000 (UTC) Sender: linux-block-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-block@vger.kernel.org When requst is added to rq list of sw queue(ctx), the rq may be from different hctx, after multi queue mapping is introduced. So we have to put the request into one per-queue-type list inside sw queue, otherwise the request may be dispatched to wrong hw queue. Cc: Jeff Moyer Cc: Mike Snitzer Cc: Christoph Hellwig Signed-off-by: Ming Lei --- block/blk-mq-debugfs.c | 69 ++++++++++++++++++++++++++++---------------------- block/blk-mq-sched.c | 23 +++++++++++------ block/blk-mq.c | 52 ++++++++++++++++++++++--------------- block/blk-mq.h | 10 +++++--- 4 files changed, 91 insertions(+), 63 deletions(-) diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c index 2793e91bc7a4..7021d44cef6d 100644 --- a/block/blk-mq-debugfs.c +++ b/block/blk-mq-debugfs.c @@ -637,36 +637,43 @@ static int hctx_dispatch_busy_show(void *data, struct seq_file *m) return 0; } -static void *ctx_rq_list_start(struct seq_file *m, loff_t *pos) - __acquires(&ctx->lock) -{ - struct blk_mq_ctx *ctx = m->private; - - spin_lock(&ctx->lock); - return seq_list_start(&ctx->rq_list, *pos); -} - -static void *ctx_rq_list_next(struct seq_file *m, void *v, loff_t *pos) -{ - struct blk_mq_ctx *ctx = m->private; - - return seq_list_next(v, &ctx->rq_list, pos); -} +#define CTX_RQ_SEQ_OPS(name, type) \ +static void *ctx_##name##_rq_list_start(struct seq_file *m, loff_t *pos) \ + __acquires(&ctx->lock) \ +{ \ + struct blk_mq_ctx *ctx = m->private; \ + \ + spin_lock(&ctx->list[type].lock); \ + return seq_list_start(&ctx->list[type].rq_list, *pos); \ +} \ + \ +static void *ctx_##name##_rq_list_next(struct seq_file *m, void *v, \ + loff_t *pos) \ +{ \ + struct blk_mq_ctx *ctx = m->private; \ + \ + return seq_list_next(v, &ctx->list[type].rq_list, pos); \ +} \ + \ +static void ctx_##name##_rq_list_stop(struct seq_file *m, void *v) \ + __releases(&ctx->lock) \ +{ \ + struct blk_mq_ctx *ctx = m->private; \ + \ + spin_unlock(&ctx->list[type].lock); \ +} \ + \ +static const struct seq_operations ctx_##name##_rq_list_seq_ops = { \ + .start = ctx_##name##_rq_list_start, \ + .next = ctx_##name##_rq_list_next, \ + .stop = ctx_##name##_rq_list_stop, \ + .show = blk_mq_debugfs_rq_show, \ +} + +CTX_RQ_SEQ_OPS(default, HCTX_TYPE_DEFAULT); +CTX_RQ_SEQ_OPS(read, HCTX_TYPE_READ); +CTX_RQ_SEQ_OPS(poll, HCTX_TYPE_POLL); -static void ctx_rq_list_stop(struct seq_file *m, void *v) - __releases(&ctx->lock) -{ - struct blk_mq_ctx *ctx = m->private; - - spin_unlock(&ctx->lock); -} - -static const struct seq_operations ctx_rq_list_seq_ops = { - .start = ctx_rq_list_start, - .next = ctx_rq_list_next, - .stop = ctx_rq_list_stop, - .show = blk_mq_debugfs_rq_show, -}; static int ctx_dispatched_show(void *data, struct seq_file *m) { struct blk_mq_ctx *ctx = data; @@ -803,7 +810,9 @@ static const struct blk_mq_debugfs_attr blk_mq_debugfs_hctx_attrs[] = { }; static const struct blk_mq_debugfs_attr blk_mq_debugfs_ctx_attrs[] = { - {"rq_list", 0400, .seq_ops = &ctx_rq_list_seq_ops}, + {"default_rq_list", 0400, .seq_ops = &ctx_default_rq_list_seq_ops}, + {"read_rq_list", 0400, .seq_ops = &ctx_read_rq_list_seq_ops}, + {"poll_rq_list", 0400, .seq_ops = &ctx_poll_rq_list_seq_ops}, {"dispatched", 0600, ctx_dispatched_show, ctx_dispatched_write}, {"merged", 0600, ctx_merged_show, ctx_merged_write}, {"completed", 0600, ctx_completed_show, ctx_completed_write}, diff --git a/block/blk-mq-sched.c b/block/blk-mq-sched.c index 5b4d52d9cba2..09594947933e 100644 --- a/block/blk-mq-sched.c +++ b/block/blk-mq-sched.c @@ -301,11 +301,14 @@ EXPORT_SYMBOL_GPL(blk_mq_bio_list_merge); * too much time checking for merges. */ static bool blk_mq_attempt_merge(struct request_queue *q, + struct blk_mq_hw_ctx *hctx, struct blk_mq_ctx *ctx, struct bio *bio) { - lockdep_assert_held(&ctx->lock); + enum hctx_type type = hctx->type; - if (blk_mq_bio_list_merge(q, &ctx->rq_list, bio)) { + lockdep_assert_held(&ctx->list[type].lock); + + if (blk_mq_bio_list_merge(q, &ctx->list[type].rq_list, bio)) { ctx->rq_merged++; return true; } @@ -319,18 +322,20 @@ bool __blk_mq_sched_bio_merge(struct request_queue *q, struct bio *bio) struct blk_mq_ctx *ctx = blk_mq_get_ctx(q); struct blk_mq_hw_ctx *hctx = blk_mq_map_queue(q, bio->bi_opf, ctx->cpu); bool ret = false; + enum hctx_type type; if (e && e->type->ops.bio_merge) { blk_mq_put_ctx(ctx); return e->type->ops.bio_merge(hctx, bio); } + type = hctx->type; if ((hctx->flags & BLK_MQ_F_SHOULD_MERGE) && - !list_empty_careful(&ctx->rq_list)) { + !list_empty_careful(&ctx->list[type].rq_list)) { /* default per sw-queue merge */ - spin_lock(&ctx->lock); - ret = blk_mq_attempt_merge(q, ctx, bio); - spin_unlock(&ctx->lock); + spin_lock(&ctx->list[type].lock); + ret = blk_mq_attempt_merge(q, hctx, ctx, bio); + spin_unlock(&ctx->list[type].lock); } blk_mq_put_ctx(ctx); @@ -392,9 +397,11 @@ void blk_mq_sched_insert_request(struct request *rq, bool at_head, list_add(&rq->queuelist, &list); e->type->ops.insert_requests(hctx, &list, at_head); } else { - spin_lock(&ctx->lock); + enum hctx_type type = hctx->type; + + spin_lock(&ctx->list[type].lock); __blk_mq_insert_request(hctx, rq, at_head); - spin_unlock(&ctx->lock); + spin_unlock(&ctx->list[type].lock); } run: diff --git a/block/blk-mq.c b/block/blk-mq.c index e843f23843c8..27303951a752 100644 --- a/block/blk-mq.c +++ b/block/blk-mq.c @@ -958,11 +958,12 @@ static bool flush_busy_ctx(struct sbitmap *sb, unsigned int bitnr, void *data) struct flush_busy_ctx_data *flush_data = data; struct blk_mq_hw_ctx *hctx = flush_data->hctx; struct blk_mq_ctx *ctx = hctx->ctxs[bitnr]; + enum hctx_type type = hctx->type; - spin_lock(&ctx->lock); - list_splice_tail_init(&ctx->rq_list, flush_data->list); + spin_lock(&ctx->list[type].lock); + list_splice_tail_init(&ctx->list[type].rq_list, flush_data->list); sbitmap_clear_bit(sb, bitnr); - spin_unlock(&ctx->lock); + spin_unlock(&ctx->list[type].lock); return true; } @@ -992,15 +993,16 @@ static bool dispatch_rq_from_ctx(struct sbitmap *sb, unsigned int bitnr, struct dispatch_rq_data *dispatch_data = data; struct blk_mq_hw_ctx *hctx = dispatch_data->hctx; struct blk_mq_ctx *ctx = hctx->ctxs[bitnr]; + enum hctx_type type = hctx->type; - spin_lock(&ctx->lock); - if (!list_empty(&ctx->rq_list)) { - dispatch_data->rq = list_entry_rq(ctx->rq_list.next); + spin_lock(&ctx->list[type].lock); + if (!list_empty(&ctx->list[type].rq_list)) { + dispatch_data->rq = list_entry_rq(ctx->list[type].rq_list.next); list_del_init(&dispatch_data->rq->queuelist); - if (list_empty(&ctx->rq_list)) + if (list_empty(&ctx->list[type].rq_list)) sbitmap_clear_bit(sb, bitnr); } - spin_unlock(&ctx->lock); + spin_unlock(&ctx->list[type].lock); return !dispatch_data->rq; } @@ -1608,15 +1610,16 @@ static inline void __blk_mq_insert_req_list(struct blk_mq_hw_ctx *hctx, bool at_head) { struct blk_mq_ctx *ctx = rq->mq_ctx; + enum hctx_type type = hctx->type; - lockdep_assert_held(&ctx->lock); + lockdep_assert_held(&ctx->list[type].lock); trace_block_rq_insert(hctx->queue, rq); if (at_head) - list_add(&rq->queuelist, &ctx->rq_list); + list_add(&rq->queuelist, &ctx->list[type].rq_list); else - list_add_tail(&rq->queuelist, &ctx->rq_list); + list_add_tail(&rq->queuelist, &ctx->list[type].rq_list); } void __blk_mq_insert_request(struct blk_mq_hw_ctx *hctx, struct request *rq, @@ -1624,7 +1627,7 @@ void __blk_mq_insert_request(struct blk_mq_hw_ctx *hctx, struct request *rq, { struct blk_mq_ctx *ctx = rq->mq_ctx; - lockdep_assert_held(&ctx->lock); + lockdep_assert_held(&ctx->list[hctx->type].lock); __blk_mq_insert_req_list(hctx, rq, at_head); blk_mq_hctx_mark_pending(hctx, ctx); @@ -1651,6 +1654,7 @@ void blk_mq_insert_requests(struct blk_mq_hw_ctx *hctx, struct blk_mq_ctx *ctx, { struct request *rq; + enum hctx_type type = hctx->type; /* * preemption doesn't flush plug list, so it's possible ctx->cpu is @@ -1661,10 +1665,10 @@ void blk_mq_insert_requests(struct blk_mq_hw_ctx *hctx, struct blk_mq_ctx *ctx, trace_block_rq_insert(hctx->queue, rq); } - spin_lock(&ctx->lock); - list_splice_tail_init(list, &ctx->rq_list); + spin_lock(&ctx->list[type].lock); + list_splice_tail_init(list, &ctx->list[type].rq_list); blk_mq_hctx_mark_pending(hctx, ctx); - spin_unlock(&ctx->lock); + spin_unlock(&ctx->list[type].lock); } static int plug_rq_cmp(void *priv, struct list_head *a, struct list_head *b) @@ -2200,16 +2204,18 @@ static int blk_mq_hctx_notify_dead(unsigned int cpu, struct hlist_node *node) struct blk_mq_hw_ctx *hctx; struct blk_mq_ctx *ctx; LIST_HEAD(tmp); + enum hctx_type type; hctx = hlist_entry_safe(node, struct blk_mq_hw_ctx, cpuhp_dead); ctx = __blk_mq_get_ctx(hctx->queue, cpu); + type = hctx->type; - spin_lock(&ctx->lock); - if (!list_empty(&ctx->rq_list)) { - list_splice_init(&ctx->rq_list, &tmp); + spin_lock(&ctx->list[type].lock); + if (!list_empty(&ctx->list[type].rq_list)) { + list_splice_init(&ctx->list[type].rq_list, &tmp); blk_mq_hctx_clear_pending(hctx, ctx); } - spin_unlock(&ctx->lock); + spin_unlock(&ctx->list[type].lock); if (list_empty(&tmp)) return 0; @@ -2343,10 +2349,14 @@ static void blk_mq_init_cpu_queues(struct request_queue *q, for_each_possible_cpu(i) { struct blk_mq_ctx *__ctx = per_cpu_ptr(q->queue_ctx, i); struct blk_mq_hw_ctx *hctx; + int k; __ctx->cpu = i; - spin_lock_init(&__ctx->lock); - INIT_LIST_HEAD(&__ctx->rq_list); + + for (k = HCTX_TYPE_DEFAULT; k < HCTX_MAX_TYPES; k++) { + spin_lock_init(&__ctx->list[k].lock); + INIT_LIST_HEAD(&__ctx->list[k].rq_list); + } __ctx->queue = q; /* diff --git a/block/blk-mq.h b/block/blk-mq.h index f50c73d559d7..39064144326e 100644 --- a/block/blk-mq.h +++ b/block/blk-mq.h @@ -12,14 +12,16 @@ struct blk_mq_ctxs { struct blk_mq_ctx __percpu *queue_ctx; }; +struct blk_mq_ctx_list { + spinlock_t lock; + struct list_head rq_list; +} ____cacheline_aligned_in_smp; + /** * struct blk_mq_ctx - State for a software queue facing the submitting CPUs */ struct blk_mq_ctx { - struct { - spinlock_t lock; - struct list_head rq_list; - } ____cacheline_aligned_in_smp; + struct blk_mq_ctx_list list[HCTX_MAX_TYPES]; unsigned int cpu; unsigned short index_hw[HCTX_MAX_TYPES]; -- 2.9.5