scsi: sd: Handle read/write CDL timeout failures
authorNiklas Cassel <niklas.cassel@wdc.com>
Thu, 11 May 2023 01:13:44 +0000 (03:13 +0200)
committerMartin K. Petersen <martin.petersen@oracle.com>
Mon, 22 May 2023 21:05:19 +0000 (17:05 -0400)
Commands using a duration limit descriptor that has limit policies set to a
value other than 0x0 may be failed by the device if one of the limits are
exceeded. For such commands, since the failure is the result of the user
duration limit configuration and workload, the commands should not be
retried and terminated immediately. Furthermore, to allow the user to
differentiate these "soft" failures from hard errors due to hardware
problem, a different error code than EIO should be returned.

There are 2 cases to consider:

(1) The failure is due to a limit policy failing the command with a check
condition sense key, that is, any limit policy other than 0xD.  For this
case, scsi_check_sense() is modified to detect failures with the ABORTED
COMMAND sense key and the COMMAND TIMEOUT BEFORE PROCESSING or COMMAND
TIMEOUT DURING PROCESSING or COMMAND TIMEOUT DURING PROCESSING DUE TO ERROR
RECOVERY additional sense code. For these failures, a SUCCESS disposition
is returned so that scsi_finish_command() is called to terminate the
command.

(2) The failure is due to a limit policy set to 0xD, which result in the
command being terminated with a GOOD status, COMPLETED sense key, and DATA
CURRENTLY UNAVAILABLE additional sense code. To handle this case, the
scsi_check_sense() is modified to return a SUCCESS disposition so that
scsi_finish_command() is called to terminate the command.  In addition,
scsi_decide_disposition() has to be modified to see if a command being
terminated with GOOD status has sense data.  This is as defined in SCSI
Primary Commands - 6 (SPC-6), so all according to spec, even if GOOD status
commands were not checked before.

If scsi_check_sense() detects sense data representing a duration limit,
scsi_check_sense() will set the newly introduced SCSI ML byte
SCSIML_STAT_DL_TIMEOUT. This SCSI ML byte is checked in scsi_noretry_cmd(),
so that a command that failed because of a CDL timeout cannot be
retried. The SCSI ML byte is also checked in scsi_result_to_blk_status() to
complete the command request with the BLK_STS_DURATION_LIMIT status, which
result in the user seeing ETIME errors for the failed commands.

Co-developed-by: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Link: https://lore.kernel.org/r/20230511011356.227789-12-nks@flawful.org
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
drivers/scsi/scsi_error.c
drivers/scsi/scsi_lib.c
drivers/scsi/scsi_priv.h

index 8b7d227..c67cdcd 100644 (file)
@@ -536,6 +536,7 @@ static inline void set_scsi_ml_byte(struct scsi_cmnd *cmd, u8 status)
  */
 enum scsi_disposition scsi_check_sense(struct scsi_cmnd *scmd)
 {
+       struct request *req = scsi_cmd_to_rq(scmd);
        struct scsi_device *sdev = scmd->device;
        struct scsi_sense_hdr sshdr;
 
@@ -595,6 +596,22 @@ enum scsi_disposition scsi_check_sense(struct scsi_cmnd *scmd)
                if (sshdr.asc == 0x10) /* DIF */
                        return SUCCESS;
 
+               /*
+                * Check aborts due to command duration limit policy:
+                * ABORTED COMMAND additional sense code with the
+                * COMMAND TIMEOUT BEFORE PROCESSING or
+                * COMMAND TIMEOUT DURING PROCESSING or
+                * COMMAND TIMEOUT DURING PROCESSING DUE TO ERROR RECOVERY
+                * additional sense code qualifiers.
+                */
+               if (sshdr.asc == 0x2e &&
+                   sshdr.ascq >= 0x01 && sshdr.ascq <= 0x03) {
+                       set_scsi_ml_byte(scmd, SCSIML_STAT_DL_TIMEOUT);
+                       req->cmd_flags |= REQ_FAILFAST_DEV;
+                       req->rq_flags |= RQF_QUIET;
+                       return SUCCESS;
+               }
+
                if (sshdr.asc == 0x44 && sdev->sdev_bflags & BLIST_RETRY_ITF)
                        return ADD_TO_MLQUEUE;
                if (sshdr.asc == 0xc1 && sshdr.ascq == 0x01 &&
@@ -691,6 +708,14 @@ enum scsi_disposition scsi_check_sense(struct scsi_cmnd *scmd)
                }
                return SUCCESS;
 
+       case COMPLETED:
+               if (sshdr.asc == 0x55 && sshdr.ascq == 0x0a) {
+                       set_scsi_ml_byte(scmd, SCSIML_STAT_DL_TIMEOUT);
+                       req->cmd_flags |= REQ_FAILFAST_DEV;
+                       req->rq_flags |= RQF_QUIET;
+               }
+               return SUCCESS;
+
        default:
                return SUCCESS;
        }
@@ -785,6 +810,14 @@ static enum scsi_disposition scsi_eh_completed_normally(struct scsi_cmnd *scmd)
        switch (get_status_byte(scmd)) {
        case SAM_STAT_GOOD:
                scsi_handle_queue_ramp_up(scmd->device);
+               if (scmd->sense_buffer && SCSI_SENSE_VALID(scmd))
+                       /*
+                        * If we have sense data, call scsi_check_sense() in
+                        * order to set the correct SCSI ML byte (if any).
+                        * No point in checking the return value, since the
+                        * command has already completed successfully.
+                        */
+                       scsi_check_sense(scmd);
                fallthrough;
        case SAM_STAT_COMMAND_TERMINATED:
                return SUCCESS;
@@ -1807,6 +1840,10 @@ bool scsi_noretry_cmd(struct scsi_cmnd *scmd)
                return !!(req->cmd_flags & REQ_FAILFAST_DRIVER);
        }
 
+       /* Never retry commands aborted due to a duration limit timeout */
+       if (scsi_ml_byte(scmd->result) == SCSIML_STAT_DL_TIMEOUT)
+               return true;
+
        if (!scsi_status_is_check_condition(scmd->result))
                return false;
 
@@ -1966,6 +2003,14 @@ enum scsi_disposition scsi_decide_disposition(struct scsi_cmnd *scmd)
                if (scmd->cmnd[0] == REPORT_LUNS)
                        scmd->device->sdev_target->expecting_lun_change = 0;
                scsi_handle_queue_ramp_up(scmd->device);
+               if (scmd->sense_buffer && SCSI_SENSE_VALID(scmd))
+                       /*
+                        * If we have sense data, call scsi_check_sense() in
+                        * order to set the correct SCSI ML byte (if any).
+                        * No point in checking the return value, since the
+                        * command has already completed successfully.
+                        */
+                       scsi_check_sense(scmd);
                fallthrough;
        case SAM_STAT_COMMAND_TERMINATED:
                return SUCCESS;
index 633c4e8..b894432 100644 (file)
@@ -601,6 +601,8 @@ static blk_status_t scsi_result_to_blk_status(int result)
                return BLK_STS_MEDIUM;
        case SCSIML_STAT_TGT_FAILURE:
                return BLK_STS_TARGET;
+       case SCSIML_STAT_DL_TIMEOUT:
+               return BLK_STS_DURATION_LIMIT;
        }
 
        switch (host_byte(result)) {
@@ -798,6 +800,8 @@ static void scsi_io_completion_action(struct scsi_cmnd *cmd, int result)
                                blk_stat = BLK_STS_ZONE_OPEN_RESOURCE;
                        }
                        break;
+               case COMPLETED:
+                       fallthrough;
                default:
                        action = ACTION_FAIL;
                        break;
index 74324fb..f42388e 100644 (file)
@@ -27,6 +27,7 @@ enum scsi_ml_status {
        SCSIML_STAT_NOSPC               = 0x02, /* Space allocation on the dev failed */
        SCSIML_STAT_MED_ERROR           = 0x03, /* Medium error */
        SCSIML_STAT_TGT_FAILURE         = 0x04, /* Permanent target failure */
+       SCSIML_STAT_DL_TIMEOUT          = 0x05, /* Command Duration Limit timeout */
 };
 
 static inline u8 scsi_ml_byte(int result)