From patchwork Thu Mar 9 08:05:31 2017 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Tahsin Erdogan X-Patchwork-Id: 9612557 Return-Path: Received: from mail.wl.linuxfoundation.org (pdx-wl-mail.web.codeaurora.org [172.30.200.125]) by pdx-korg-patchwork.web.codeaurora.org (Postfix) with ESMTP id 3E916604D9 for ; Thu, 9 Mar 2017 08:06:58 +0000 (UTC) Received: from mail.wl.linuxfoundation.org (localhost [127.0.0.1]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id 1E2371FFB2 for ; Thu, 9 Mar 2017 08:06:58 +0000 (UTC) Received: by mail.wl.linuxfoundation.org (Postfix, from userid 486) id 106EA285FF; Thu, 9 Mar 2017 08:06:58 +0000 (UTC) X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on pdx-wl-mail.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-6.5 required=2.0 tests=BAYES_00,DKIM_SIGNED, DKIM_VALID,DKIM_VALID_AU,RCVD_IN_DNSWL_HI,RCVD_IN_SORBS_SPAM autolearn=unavailable version=3.3.1 Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id 6222E27B13 for ; Thu, 9 Mar 2017 08:06:57 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753047AbdCIIG3 (ORCPT ); Thu, 9 Mar 2017 03:06:29 -0500 Received: from mail-pg0-f53.google.com ([74.125.83.53]:33554 "EHLO mail-pg0-f53.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752699AbdCIIG1 (ORCPT ); Thu, 9 Mar 2017 03:06:27 -0500 Received: by mail-pg0-f53.google.com with SMTP id 25so23666299pgy.0 for ; Thu, 09 Mar 2017 00:06:12 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20161025; h=from:to:cc:subject:date:message-id:in-reply-to:references; bh=f03kxtQBEtxwKjeHksbPPFQI/E59swr+lXF3Ys/WLRs=; b=XMqtpEgIzVg46gYmc/GhThep73LABIfVInJBFsy4jMJcliM53yfyhSEF8j85h3+PSY 9XPZSA2FwDh4W5AQjecu0kiZwVyNWLLXp2YcFyqB5GSHTvr3r5MKwc7M/LmPOmyr4yaH uGPMA4tDmDPwF5HRHKVev8VvDYQzeIRwTjSqYmhC/DSXpkejfXEc32C/KMTI5IgCUa5O VvMLfDO1oQ9ZQvQe9As1yGQWZeJy0S/UqkJfEpxzY0FtvDCqSrljdoqJaNtmda8AJae+ XH747+DcJ5ayMoaoZRUjemDnkgfSJniWiBDY9tNa09Uo/GEbiaMu2SqXrL3Kr74snqvB T/EQ== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:from:to:cc:subject:date:message-id:in-reply-to :references; bh=f03kxtQBEtxwKjeHksbPPFQI/E59swr+lXF3Ys/WLRs=; b=TNRMpgu3Leg5guAyq+sHvH0yj/0CJdSm950pRG15HY0NzswCEg1J+3Ugw8BKcZ1v5/ EfuWyQynX8MnoTp0QfKKgEjK9kbGpeNjuPPNtiBiwDP89qIDPdQOCDV5WF2Jfs9CzfZn sTroMCdGeDQp3agCI9cYsUHMJOAS/YOuU1zylv6ow96qdjiWYWDGzYstRHGqWADnvsAm V7crY0k5sUrqKi+2SPoOLlDwoROGoamNE6YOw2Pe1h401++mgmnzddbP0ID2oo+SZaR6 xWRNDeIk278l/GT0DPOFXXl6+xV3yjqjwSjUSnCRFvkNDCHhWIFbG9AEYRIQ1Jadoprx B1jQ== X-Gm-Message-State: AMke39l22+ITjqJrrQrEHSl++BFGWPcxL87wzmSu7VRlKXcHzuPcyhEPbECTiJBuJQqb4V9T X-Received: by 10.84.216.81 with SMTP id f17mr15211471plj.170.1489046761818; Thu, 09 Mar 2017 00:06:01 -0800 (PST) Received: from tahsin1.mtv.corp.google.com ([100.99.140.90]) by smtp.gmail.com with ESMTPSA id 129sm10528808pgj.59.2017.03.09.00.06.00 (version=TLS1_2 cipher=ECDHE-RSA-AES128-SHA bits=128/128); Thu, 09 Mar 2017 00:06:00 -0800 (PST) From: Tahsin Erdogan To: Tejun Heo , Jens Axboe Cc: linux-block@vger.kernel.org, David Rientjes , linux-kernel@vger.kernel.org, Tahsin Erdogan Subject: [PATCH v5] blkcg: allocate struct blkcg_gq outside request queue spinlock Date: Thu, 9 Mar 2017 00:05:31 -0800 Message-Id: <20170309080531.9048-1-tahsin@google.com> X-Mailer: git-send-email 2.12.0.246.ga2ecc84866-goog In-Reply-To: <20170306200319.GF19696@htj.duckdns.org> References: <20170306200319.GF19696@htj.duckdns.org> Sender: linux-block-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-block@vger.kernel.org X-Virus-Scanned: ClamAV using ClamSMTP blkg_conf_prep() currently calls blkg_lookup_create() while holding request queue spinlock. This means allocating memory for struct blkcg_gq has to be made non-blocking. This causes occasional -ENOMEM failures in call paths like below: pcpu_alloc+0x68f/0x710 __alloc_percpu_gfp+0xd/0x10 __percpu_counter_init+0x55/0xc0 cfq_pd_alloc+0x3b2/0x4e0 blkg_alloc+0x187/0x230 blkg_create+0x489/0x670 blkg_lookup_create+0x9a/0x230 blkg_conf_prep+0x1fb/0x240 __cfqg_set_weight_device.isra.105+0x5c/0x180 cfq_set_weight_on_dfl+0x69/0xc0 cgroup_file_write+0x39/0x1c0 kernfs_fop_write+0x13f/0x1d0 __vfs_write+0x23/0x120 vfs_write+0xc2/0x1f0 SyS_write+0x44/0xb0 entry_SYSCALL_64_fastpath+0x18/0xad In the code path above, percpu allocator cannot call vmalloc() due to queue spinlock. A failure in this call path gives grief to tools which are trying to configure io weights. We see occasional failures happen shortly after reboots even when system is not under any memory pressure. Machines with a lot of cpus are more vulnerable to this condition. Update blkg_create() function to temporarily drop the rcu and queue locks when it is allowed by gfp mask. Suggested-by: Tejun Heo Signed-off-by: Tahsin Erdogan Acked-by: Tejun Heo --- v5: Removed stale blkg_alloc() in blkcg_init_queue() Pushed down radix_tree_preload() into blkg_create() because it disables preemption on return and makes it unsafe to call blocking memory allocations. v4: Simplified error checking in blkg_create() Factored out __blkg_lookup_create() v3: Pushed down all blkg allocations into blkg_create() v2: Moved blkg creation into blkg_lookup_create() to avoid duplicating blkg_lookup_create() logic. block/blk-cgroup.c | 138 ++++++++++++++++++++++++++++----------------- include/linux/blk-cgroup.h | 6 +- 2 files changed, 91 insertions(+), 53 deletions(-) diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c index bbe7ee00bd3d..bdf87f0c1b1b 100644 --- a/block/blk-cgroup.c +++ b/block/blk-cgroup.c @@ -165,16 +165,18 @@ struct blkcg_gq *blkg_lookup_slowpath(struct blkcg *blkcg, EXPORT_SYMBOL_GPL(blkg_lookup_slowpath); /* - * If @new_blkg is %NULL, this function tries to allocate a new one as - * necessary using %GFP_NOWAIT. @new_blkg is always consumed on return. + * If gfp mask allows blocking, this function temporarily drops rcu and queue + * locks to allocate memory. */ static struct blkcg_gq *blkg_create(struct blkcg *blkcg, - struct request_queue *q, - struct blkcg_gq *new_blkg) + struct request_queue *q, gfp_t gfp, + const struct blkcg_policy *pol) { - struct blkcg_gq *blkg; + struct blkcg_gq *blkg = NULL; struct bdi_writeback_congested *wb_congested; int i, ret; + const bool drop_locks = gfpflags_allow_blocking(gfp); + bool preloaded = false; WARN_ON_ONCE(!rcu_read_lock_held()); lockdep_assert_held(q->queue_lock); @@ -185,31 +187,53 @@ static struct blkcg_gq *blkg_create(struct blkcg *blkcg, goto err_free_blkg; } + if (drop_locks) { + spin_unlock_irq(q->queue_lock); + rcu_read_unlock(); + } + wb_congested = wb_congested_get_create(q->backing_dev_info, - blkcg->css.id, - GFP_NOWAIT | __GFP_NOWARN); - if (!wb_congested) { + blkcg->css.id, gfp); + blkg = blkg_alloc(blkcg, q, gfp); + + if (drop_locks) { + preloaded = !radix_tree_preload(gfp); + rcu_read_lock(); + spin_lock_irq(q->queue_lock); + } + + if (unlikely(!wb_congested || !blkg)) { ret = -ENOMEM; - goto err_put_css; + goto err_put; } - /* allocate */ - if (!new_blkg) { - new_blkg = blkg_alloc(blkcg, q, GFP_NOWAIT | __GFP_NOWARN); - if (unlikely(!new_blkg)) { - ret = -ENOMEM; - goto err_put_congested; + blkg->wb_congested = wb_congested; + + if (pol) { + WARN_ON(!drop_locks); + + if (!blkcg_policy_enabled(q, pol)) { + ret = -EOPNOTSUPP; + goto err_put; + } + + /* + * This could be the first entry point of blkcg implementation + * and we shouldn't allow anything to go through for a bypassing + * queue. + */ + if (unlikely(blk_queue_bypass(q))) { + ret = blk_queue_dying(q) ? -ENODEV : -EBUSY; + goto err_put; } } - blkg = new_blkg; - blkg->wb_congested = wb_congested; /* link parent */ if (blkcg_parent(blkcg)) { blkg->parent = __blkg_lookup(blkcg_parent(blkcg), q, false); if (WARN_ON_ONCE(!blkg->parent)) { ret = -ENODEV; - goto err_put_congested; + goto err_put; } blkg_get(blkg->parent); } @@ -236,6 +260,9 @@ static struct blkcg_gq *blkg_create(struct blkcg *blkcg, pol->pd_online_fn(blkg->pd[i]); } } + + if (preloaded) + radix_tree_preload_end(); blkg->online = true; spin_unlock(&blkcg->lock); @@ -246,44 +273,45 @@ static struct blkcg_gq *blkg_create(struct blkcg *blkcg, blkg_put(blkg); return ERR_PTR(ret); -err_put_congested: - wb_congested_put(wb_congested); -err_put_css: +err_put: + if (preloaded) + radix_tree_preload_end(); + if (wb_congested) + wb_congested_put(wb_congested); css_put(&blkcg->css); err_free_blkg: - blkg_free(new_blkg); + blkg_free(blkg); return ERR_PTR(ret); } /** - * blkg_lookup_create - lookup blkg, try to create one if not there + * __blkg_lookup_create - lookup blkg, try to create one if not there * @blkcg: blkcg of interest * @q: request_queue of interest + * @gfp: gfp mask + * @pol: blkcg policy (optional) * * Lookup blkg for the @blkcg - @q pair. If it doesn't exist, try to * create one. blkg creation is performed recursively from blkcg_root such * that all non-root blkg's have access to the parent blkg. This function * should be called under RCU read lock and @q->queue_lock. * + * When gfp mask allows blocking, rcu and queue locks may be dropped for + * allocating memory. In this case, the locks will be reacquired on return. + * * Returns pointer to the looked up or created blkg on success, ERR_PTR() * value on error. If @q is dead, returns ERR_PTR(-EINVAL). If @q is not * dead and bypassing, returns ERR_PTR(-EBUSY). */ -struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg, - struct request_queue *q) +struct blkcg_gq *__blkg_lookup_create(struct blkcg *blkcg, + struct request_queue *q, gfp_t gfp, + const struct blkcg_policy *pol) { struct blkcg_gq *blkg; WARN_ON_ONCE(!rcu_read_lock_held()); lockdep_assert_held(q->queue_lock); - /* - * This could be the first entry point of blkcg implementation and - * we shouldn't allow anything to go through for a bypassing queue. - */ - if (unlikely(blk_queue_bypass(q))) - return ERR_PTR(blk_queue_dying(q) ? -ENODEV : -EBUSY); - blkg = __blkg_lookup(blkcg, q, true); if (blkg) return blkg; @@ -301,12 +329,35 @@ struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg, parent = blkcg_parent(parent); } - blkg = blkg_create(pos, q, NULL); + blkg = blkg_create(pos, q, gfp, pol); if (pos == blkcg || IS_ERR(blkg)) return blkg; } } +/** + * blkg_lookup_create - lookup blkg, try to create one if not there + * + * Performs an initial queue bypass check and then passes control to + * __blkg_lookup_create(). + */ +struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg, + struct request_queue *q, gfp_t gfp, + const struct blkcg_policy *pol) +{ + WARN_ON_ONCE(!rcu_read_lock_held()); + lockdep_assert_held(q->queue_lock); + + /* + * This could be the first entry point of blkcg implementation and + * we shouldn't allow anything to go through for a bypassing queue. + */ + if (unlikely(blk_queue_bypass(q))) + return ERR_PTR(blk_queue_dying(q) ? -ENODEV : -EBUSY); + + return __blkg_lookup_create(blkcg, q, gfp, pol); +} + static void blkg_destroy(struct blkcg_gq *blkg) { struct blkcg *blkcg = blkg->blkcg; @@ -817,7 +868,7 @@ int blkg_conf_prep(struct blkcg *blkcg, const struct blkcg_policy *pol, spin_lock_irq(disk->queue->queue_lock); if (blkcg_policy_enabled(disk->queue, pol)) - blkg = blkg_lookup_create(blkcg, disk->queue); + blkg = blkg_lookup_create(blkcg, disk->queue, GFP_KERNEL, pol); else blkg = ERR_PTR(-EOPNOTSUPP); @@ -1056,30 +1107,15 @@ blkcg_css_alloc(struct cgroup_subsys_state *parent_css) */ int blkcg_init_queue(struct request_queue *q) { - struct blkcg_gq *new_blkg, *blkg; - bool preloaded; + struct blkcg_gq *blkg; int ret; - new_blkg = blkg_alloc(&blkcg_root, q, GFP_KERNEL); - if (!new_blkg) - return -ENOMEM; - - preloaded = !radix_tree_preload(GFP_KERNEL); - - /* - * Make sure the root blkg exists and count the existing blkgs. As - * @q is bypassing at this point, blkg_lookup_create() can't be - * used. Open code insertion. - */ rcu_read_lock(); spin_lock_irq(q->queue_lock); - blkg = blkg_create(&blkcg_root, q, new_blkg); + blkg = __blkg_lookup_create(&blkcg_root, q, GFP_KERNEL, NULL); spin_unlock_irq(q->queue_lock); rcu_read_unlock(); - if (preloaded) - radix_tree_preload_end(); - if (IS_ERR(blkg)) return PTR_ERR(blkg); diff --git a/include/linux/blk-cgroup.h b/include/linux/blk-cgroup.h index 01b62e7bac74..955903a8f6cb 100644 --- a/include/linux/blk-cgroup.h +++ b/include/linux/blk-cgroup.h @@ -172,7 +172,8 @@ extern struct cgroup_subsys_state * const blkcg_root_css; struct blkcg_gq *blkg_lookup_slowpath(struct blkcg *blkcg, struct request_queue *q, bool update_hint); struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg, - struct request_queue *q); + struct request_queue *q, gfp_t gfp, + const struct blkcg_policy *pol); int blkcg_init_queue(struct request_queue *q); void blkcg_drain_queue(struct request_queue *q); void blkcg_exit_queue(struct request_queue *q); @@ -694,7 +695,8 @@ static inline bool blkcg_bio_issue_check(struct request_queue *q, blkg = blkg_lookup(blkcg, q); if (unlikely(!blkg)) { spin_lock_irq(q->queue_lock); - blkg = blkg_lookup_create(blkcg, q); + blkg = blkg_lookup_create(blkcg, q, GFP_NOWAIT | __GFP_NOWARN, + NULL); if (IS_ERR(blkg)) blkg = NULL; spin_unlock_irq(q->queue_lock);