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=-10.2 required=3.0 tests=DKIMWL_WL_HIGH,DKIM_SIGNED, DKIM_VALID,DKIM_VALID_AU,HEADER_FROM_DIFFERENT_DOMAINS,INCLUDES_PATCH, MAILING_LIST_MULTI,MENTIONS_GIT_HOSTING,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 4137FC282C0 for ; Sat, 26 Jan 2019 01:23:47 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id 02D8321902 for ; Sat, 26 Jan 2019 01:23:47 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=pass (2048-bit key) header.d=oracle.com header.i=@oracle.com header.b="3P8VYIaB" Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1726991AbfAZBXq (ORCPT ); Fri, 25 Jan 2019 20:23:46 -0500 Received: from aserp2130.oracle.com ([141.146.126.79]:51740 "EHLO aserp2130.oracle.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1725550AbfAZBXq (ORCPT ); Fri, 25 Jan 2019 20:23:46 -0500 Received: from pps.filterd (aserp2130.oracle.com [127.0.0.1]) by aserp2130.oracle.com (8.16.0.22/8.16.0.22) with SMTP id x0Q1JVex045652; Sat, 26 Jan 2019 01:23:39 GMT DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=oracle.com; h=subject : to : cc : references : from : message-id : date : mime-version : in-reply-to : content-type : content-transfer-encoding; s=corp-2018-07-02; bh=qu9vxDdpFmRBgmwHL4JHREpqE+YrIWp+qSm3IsHXkeg=; b=3P8VYIaBgPZTdHvMQxZeSFhQI1Lv+02Eiiz0bq52apOD+PC6mM8Rb62oCD6MYZp5NdjU 7F8WZ5w206RIRRHevJnNKcQgjCiRwHQdZ4ARjTiWYOiwwdP1zquwgZiZvCTF5PsqbV4E Qyb3yp3ZlowCwH3B3nZ9Th25BuZRGwSS2PPjzniPZkvo/tXK+fi7m9zETMaylLPFVOoO 8d8byfGcKtjKjlILzi36lzRdbIMowuIKMP8Utv9dpplBoqR+Nlv2Un/kDLH97hBx3C9q NFwnvlEcyuOalDB4VIYaGdIUSq6hbb04QLMgPjF25FW0ShvKUe80eAu9LCDuRVwz6ND1 Fw== Received: from userv0021.oracle.com (userv0021.oracle.com [156.151.31.71]) by aserp2130.oracle.com with ESMTP id 2q8d2dr2tj-1 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-GCM-SHA384 bits=256 verify=OK); Sat, 26 Jan 2019 01:23:39 +0000 Received: from userv0122.oracle.com (userv0122.oracle.com [156.151.31.75]) by userv0021.oracle.com (8.14.4/8.14.4) with ESMTP id x0Q1Nchw031528 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-GCM-SHA384 bits=256 verify=OK); Sat, 26 Jan 2019 01:23:38 GMT Received: from abhmp0017.oracle.com (abhmp0017.oracle.com [141.146.116.23]) by userv0122.oracle.com (8.14.4/8.14.4) with ESMTP id x0Q1Nbll009885; Sat, 26 Jan 2019 01:23:37 GMT Received: from [10.191.30.196] (/10.191.30.196) by default (Oracle Beehive Gateway v4.0) with ESMTP ; Fri, 25 Jan 2019 17:23:37 -0800 Subject: Re: fsync hangs after scsi rejected a request To: Florian Stecker , Ming Lei Cc: Linux SCSI List , "Martin K. Petersen" , Bart Van Assche , linux-block References: <70df6d90-31eb-96f4-908d-462ea58dea6e@florianstecker.de> <4b631218-49ee-c3ed-46be-17f9d30b663f@florianstecker.de> <3d921990-06c4-4662-2853-7667b4e9a462@oracle.com> From: "jianchao.wang" Message-ID: <4fe48ebc-43db-7ada-087b-5ccbdb64cc66@oracle.com> Date: Sat, 26 Jan 2019 09:25:23 +0800 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101 Thunderbird/60.2.1 MIME-Version: 1.0 In-Reply-To: <3d921990-06c4-4662-2853-7667b4e9a462@oracle.com> Content-Type: text/plain; charset=utf-8 Content-Language: en-US Content-Transfer-Encoding: 8bit X-Proofpoint-Virus-Version: vendor=nai engine=5900 definitions=9147 signatures=668682 X-Proofpoint-Spam-Details: rule=notspam policy=default score=0 suspectscore=0 malwarescore=0 phishscore=0 bulkscore=0 spamscore=0 mlxscore=0 mlxlogscore=999 adultscore=0 classifier=spam adjust=0 reason=mlx scancount=1 engine=8.0.1-1810050000 definitions=main-1901260005 Sender: linux-block-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-block@vger.kernel.org On 1/26/19 9:18 AM, jianchao.wang wrote: > > > On 1/26/19 6:10 AM, Florian Stecker wrote: >> >> >> On 1/25/19 10:05 AM, jianchao.wang wrote: >>> >>> >>> On 1/25/19 4:56 PM, Florian Stecker wrote: >>>> On 1/25/19 4:49 AM, jianchao.wang wrote: >>>>> >>>>> It sounds like not so easy to trigger. >>>>> >>>>> blk_mq_dispatch_rq_list >>>>>     scsi_queue_rq >>>>>        if (atomic_read(&sdev->device_busy) || >>>>>          scsi_device_blocked(sdev)) >>>>>          ret = BLK_STS_DEV_RESOURCE;             scsi_end_request >>>>>                                                    __blk_mq_end_request >>>>>                                                      blk_mq_sched_restart // clear RESTART >>>>>                                                        blk_mq_run_hw_queue >>>>>                                                    blk_mq_run_hw_queues >>>>>       list_splice_init(list, &hctx->dispatch) >>>>>     needs_restart = blk_mq_sched_needs_restart(hctx) >>>>> >>>>> The 'needs_restart' will be false, so the queue would be rerun. >>>>> >>>>> Thanks >>>>> Jianchao >>>> >>>> Good point. So the RESTART flag is supposed to protect against this? Now I see, this is also sort of what the lengthy comment in blk_mq_dispatch_rq_list is saying. >>>> >>>> May I complain that this is very unintuitive (the queue gets rerun when the RESTART flag is _not_ set) and also unreliable, as not every caller of blk_mq_dispatch_rq_list seems to set the flag, and also it does not always get cleared in __blk_mq_end_request? >>>> >>>> __blk_mq_end_request does the following: >>>> >>>>      if (rq->end_io) { >>>>          rq_qos_done(rq->q, rq); >>>>          rq->end_io(rq, error); >>>>      } else { >>>>          if (unlikely(blk_bidi_rq(rq))) >>>>              blk_mq_free_request(rq->next_rq); >>>>          blk_mq_free_request(rq); >>>>      } >>>> >>>> and blk_mq_free_request then calls blk_mq_sched_restart, which clears the flag. But in my case, rq->end_io != 0, so blk_mq_free_request is never called. >>> >>> So what is your rq->end_io ? >>> flush_end_io ? or mq_flush_data_end_io ? or other ? >>> In normal case, the blk_mq_end_request should be finally invoked. >>> >>> Did you ever try the bfq io scheduler instead of mq-deadline ? >>> >>> Can you share your dmesg and config file here ? >> >> Sure. >> >> dmesg: https://urldefense.proofpoint.com/v2/url?u=https-3A__florianstecker.de_dmesg-2D2019-2D01-2D25.txt&d=DwIDaQ&c=RoP1YumCXCgaWHvlZYR8PZh8Bv7qIrMUB65eapI_JnE&r=7WdAxUBeiTUTCy8v-7zXyr4qk7sx26ATvfo6QSTvZyQ&m=t_udGGsH8ivtjXHin4oQ4LxELuPoXbl_ARP7oSxiQ6k&s=DC2fhRw2g4haGhhHxOYyX4zD6FWQQ5YoERdfptVRHM4&e= >> Note that no I/O scheduler is in use here, I deactivated them in a udev rule because I still want to be able to use my laptop. When I test this stuff I just reactivate mq-deadline manually. >> >> Config is the default in Arch Linux: https://urldefense.proofpoint.com/v2/url?u=https-3A__git.archlinux.org_svntogit_packages.git_tree_trunk_config-3Fh-3Dpackages_linux&d=DwIDaQ&c=RoP1YumCXCgaWHvlZYR8PZh8Bv7qIrMUB65eapI_JnE&r=7WdAxUBeiTUTCy8v-7zXyr4qk7sx26ATvfo6QSTvZyQ&m=t_udGGsH8ivtjXHin4oQ4LxELuPoXbl_ARP7oSxiQ6k&s=GjV_fyXaKZ8za8zdB_gqmKW7yBlbla2OHRaAT0iGZkc&e= >> >> The problem also appears with BFQ. >> >> And rq->end_io is set to mq_flush_data_end_io when this happens. The only point I see where this function could invoke blk_mq_end_request is via blk_flush_complete_seq, but it does so only if seq == REQ_FSEQ_DONE. However, seq is REQ_FSEQ_POSTFLUSH for me (i.e. it just transitioned from REQ_FSEQ_DATA to REQ_FSEQ_POSTFLUSH). > > > Could it be this scenario ? > > data + post flush > > blk_flush_complete_seq a flush > case REQ_FSEQ_DATA > blk_flush_queue_rq > > finally issued to driver blk_mq_dispatch_rq_list > try to issue a flush req > failed due to NON-NCQ command > due to RESTART is set, do nothing > req complete > req->end_io（） // doesn't check RESTART > mq_flush_data_end_io > 　 blk_flush_complete_seq > case REQ_FSEQ_POSTFLUSH > blk_kick_flush > do nothing because previous flush > has not been completed, so > flush_pending_idx != flush_running_idx > If it is right, maybe we could try following diff --git a/block/blk-mq.c b/block/blk-mq.c index 3ba37b9..3411b6a 100644 --- a/block/blk-mq.c +++ b/block/blk-mq.c @@ -549,6 +549,7 @@ inline void __blk_mq_end_request(struct request *rq, blk_status_t error) if (rq->end_io) { rq_qos_done(rq->q, rq); rq->end_io(rq, error); + blk_mq_sched_restart(hctx); } else { if (unlikely(blk_bidi_rq(rq))) blk_mq_free_request(rq->next_rq); > > > >>> >>