From axboe@suse.de Thu Mar 17 19:52:29 2005 Return-Path: X-Original-To: con@kolivas.org Delivered-To: con@kolivas.org Received: from bhhdoa.org.au (bhhdoa.org.au [216.17.101.199]) by mail.kolivas.org (Postfix) with ESMTP id F027CC5C97 for ; Thu, 17 Mar 2005 19:52:35 +1100 (EST) Received: from virtualhost.dk (ns.virtualhost.dk [195.184.98.160]) by bhhdoa.org.au (Postfix) with ESMTP id 44FFD5140F for ; Thu, 17 Mar 2005 20:19:51 +1100 (EST) Received: from [62.242.22.158] (helo=wiggum.home.kernel.dk) by virtualhost.dk with esmtp (Exim 3.36 #1) id 1DBqk3-0006DR-00 for con@kolivas.org; Thu, 17 Mar 2005 09:52:31 +0100 Received: from axboe by wiggum.home.kernel.dk with local (Exim 4.44) id 1DBqk1-0000Cx-So for con@kolivas.org; Thu, 17 Mar 2005 09:52:30 +0100 Date: Thu, 17 Mar 2005 09:52:29 +0100 From: Jens Axboe To: con@kolivas.org Subject: [axboe@suse.de: [PATCH] cfq-iosched.c: fix soft hang with non-fs requests] Message-ID: <20050317085229.GN7842@suse.de> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline X-UID: 803 X-Length: 3205 Status: R X-Status: NC X-KMail-EncryptionState: X-KMail-SignatureState: X-KMail-MDN-Sent: X-Spam-Checker-Version: SpamAssassin 3.0.0 (2004-09-13) on localhost X-Spam-Level: X-Spam-Status: No, score=-0.7 required=5.0 tests=AWL,BAYES_00, FORGED_RCVD_HELO autolearn=ham version=3.0.0 Con, Trying to keep you in sync with -mm cfq-ts, you should add this as well. The hang is not easy to encounter, it was found during beta testing of the suse kernel. ----- Forwarded message from Jens Axboe ----- From: Jens Axboe Date: Thu, 17 Mar 2005 09:51:25 +0100 To: Andrew Morton Subject: [PATCH] cfq-iosched.c: fix soft hang with non-fs requests Hi, In several places cfq checks cfqd->busy_queues to see if we have pending work, but that doesn't detect requests going directly to the dispatch list since they don't end up in a queue and thus get ->busy_queues incremented. Abstract the check into cfq_pending_work() and use that throughout the code. Signed-off-by: Jens Axboe Index: linux-2.6.11-ck7/drivers/block/cfq-iosched.c =================================================================== --- linux-2.6.11-ck7.orig/drivers/block/cfq-iosched.c 2005-05-01 10:36:05.000000000 +1000 +++ linux-2.6.11-ck7/drivers/block/cfq-iosched.c 2005-05-01 10:36:06.000000000 +1000 @@ -1723,11 +1723,16 @@ cfq_insert_request(request_queue_t *q, s } } +static inline int cfq_pending_requests(struct cfq_data *cfqd) +{ + return !list_empty(&cfqd->queue->queue_head) || cfqd->busy_queues; +} + static int cfq_queue_empty(request_queue_t *q) { struct cfq_data *cfqd = q->elevator->elevator_data; - return list_empty(&q->queue_head) && !cfqd->busy_queues; + return !cfq_pending_requests(cfqd); } static void cfq_completed_request(request_queue_t *q, struct request *rq) @@ -2048,7 +2053,7 @@ static void cfq_idle_slice_timer(unsigne * only expire and reinvoke request handler, if there are * other queues with pending requests */ - if (!cfqd->busy_queues) { + if (!cfq_pending_requests(cfqd)) { cfqd->idle_slice_timer.expires = min(now + cfqd->cfq_slice_idle, cfqq->slice_end); add_timer(&cfqd->idle_slice_timer); goto out_cont; @@ -2065,7 +2070,7 @@ static void cfq_idle_slice_timer(unsigne expire: cfq_slice_expired(cfqd, 0); out_kick: - if (cfqd->busy_queues) + if (cfq_pending_requests(cfqd)) kblockd_schedule_work(&cfqd->unplug_work); out_cont: spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);