From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932604AbcHUPdf (ORCPT ); Sun, 21 Aug 2016 11:33:35 -0400 Received: from wtarreau.pck.nerim.net ([62.212.114.60]:55574 "EHLO 1wt.eu" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932559AbcHUPdc (ORCPT ); Sun, 21 Aug 2016 11:33:32 -0400 From: Willy Tarreau To: linux-kernel@vger.kernel.org, stable@vger.kernel.org Cc: Wei Fang , "Martin K . Petersen" , Willy Tarreau Subject: [PATCH 3.10 135/180] scsi: fix race between simultaneous decrements of ->host_failed Date: Sun, 21 Aug 2016 17:31:05 +0200 Message-Id: <1471793510-13022-136-git-send-email-w@1wt.eu> X-Mailer: git-send-email 2.8.0.rc2.1.gbe9624a In-Reply-To: <1471793510-13022-1-git-send-email-w@1wt.eu> References: <1471793510-13022-1-git-send-email-w@1wt.eu> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org From: Wei Fang commit 72d8c36ec364c82bf1bf0c64dfa1041cfaf139f7 upstream. sas_ata_strategy_handler() adds the works of the ata error handler to system_unbound_wq. This workqueue asynchronously runs work items, so the ata error handler will be performed concurrently on different CPUs. In this case, ->host_failed will be decreased simultaneously in scsi_eh_finish_cmd() on different CPUs, and become abnormal. It will lead to permanently inequality between ->host_failed and ->host_busy, and scsi error handler thread won't start running. IO errors after that won't be handled. Since all scmds must have been handled in the strategy handler, just remove the decrement in scsi_eh_finish_cmd() and zero ->host_busy after the strategy handler to fix this race. Fixes: 50824d6c5657 ("[SCSI] libsas: async ata-eh") Cc: stable@vger.kernel.org Signed-off-by: Wei Fang Reviewed-by: James Bottomley Signed-off-by: Martin K. Petersen Signed-off-by: Willy Tarreau --- Documentation/scsi/scsi_eh.txt | 8 ++++++-- drivers/ata/libata-eh.c | 2 +- drivers/scsi/scsi_error.c | 4 +++- 3 files changed, 10 insertions(+), 4 deletions(-) diff --git a/Documentation/scsi/scsi_eh.txt b/Documentation/scsi/scsi_eh.txt index 6ff16b6..c08b62d 100644 --- a/Documentation/scsi/scsi_eh.txt +++ b/Documentation/scsi/scsi_eh.txt @@ -255,19 +255,23 @@ scmd->allowed. 3. scmd recovered ACTION: scsi_eh_finish_cmd() is invoked to EH-finish scmd - - shost->host_failed-- - clear scmd->eh_eflags - scsi_setup_cmd_retry() - move from local eh_work_q to local eh_done_q LOCKING: none + CONCURRENCY: at most one thread per separate eh_work_q to + keep queue manipulation lockless 4. EH completes ACTION: scsi_eh_flush_done_q() retries scmds or notifies upper - layer of failure. + layer of failure. May be called concurrently but must have + a no more than one thread per separate eh_work_q to + manipulate the queue locklessly - scmd is removed from eh_done_q and scmd->eh_entry is cleared - if retry is necessary, scmd is requeued using scsi_queue_insert() - otherwise, scsi_finish_command() is invoked for scmd + - zero shost->host_failed LOCKING: queue or finish function performs appropriate locking diff --git a/drivers/ata/libata-eh.c b/drivers/ata/libata-eh.c index 063036d..126eb86 100644 --- a/drivers/ata/libata-eh.c +++ b/drivers/ata/libata-eh.c @@ -604,7 +604,7 @@ void ata_scsi_error(struct Scsi_Host *host) ata_scsi_port_error_handler(host, ap); /* finish or retry handled scmd's and clean up */ - WARN_ON(host->host_failed || !list_empty(&eh_work_q)); + WARN_ON(!list_empty(&eh_work_q)); DPRINTK("EXIT\n"); } diff --git a/drivers/scsi/scsi_error.c b/drivers/scsi/scsi_error.c index 9acbc88..5ba69ea 100644 --- a/drivers/scsi/scsi_error.c +++ b/drivers/scsi/scsi_error.c @@ -898,7 +898,6 @@ static int scsi_request_sense(struct scsi_cmnd *scmd) */ void scsi_eh_finish_cmd(struct scsi_cmnd *scmd, struct list_head *done_q) { - scmd->device->host->host_failed--; scmd->eh_eflags = 0; list_move_tail(&scmd->eh_entry, done_q); } @@ -1892,6 +1891,9 @@ int scsi_error_handler(void *data) else scsi_unjam_host(shost); + /* All scmds have been handled */ + shost->host_failed = 0; + /* * Note - if the above fails completely, the action is to take * individual devices offline and flush the queue of any -- 2.8.0.rc2.1.gbe9624a