Re: [PATCH 2/2] scsi: set timed out out mq requests to complete
From: Bart Van Assche <hidden>
Date: 2018-07-20 15:12:59
Also in:
linux-nvme
On Thu, 2018-07-19 at 15:26 -0600, Keith Busch wrote:
quoted hunk ↗ jump to hunk
diff --git a/drivers/scsi/scsi_error.c b/drivers/scsi/scsi_er=
ror.c
quoted hunk ↗ jump to hunk
index 8932ae81a15a..86ee10b2c775 100644--- a/drivers/scsi/scsi_error.c +++ b/drivers/scsi/scsi_error.c@@ -286,6 +286,9 @@ enum blk_eh_timer_return=
scsi_times_out(struct request *req)
enum blk_eh_timer_return rtn = BLK_EH_DONE�
ADs-
struct Scsi_Host *host = scmd->device->host;=
-
=20 + if (req->q->mq_ops && blk_mq_mark_co=
mplete(req))
quoted hunk ↗ jump to hunk
+ return rtn; + trace_scsi_dispatch_cmd_timeout(scmd); scsi_log_completion(scmd, TIMEOUT_ERROR); =20@@ -300,7 +303,8 @@ enum blk_eh_timer_return=
scsi_times_out(struct request *req)
set_host_byte(scmd, DID_TIME_OUT); scsi_eh_scmd_add(scmd); } - } + } else if (req->q->mq_ops) + WRITE_ONCE(req->state, MQ_RQ_IN_FLIGHT); =20 return rtn; }
Modifying the completion state and req->state from the SCSI core are la= yering violations. Have you considered to move the above changes into blk_mq� F8-rq_timed_out()? An additional benefit of that approach is that the req->q->mq_o= ps checks can be left out. Thanks, Bart.=