scsi: hpsa: Correct dev cmds outstanding for retried cmds

Prevent incrementing device->commands_outstanding for ioaccel command
retries that are driver initiated.  If the command goes through the retry
path, the device->commands_outstanding counter has already accounted for
the number of commands outstanding to the device.  Only commands going
through function hpsa_cmd_resolve_events decrement this counter.

 - ioaccel commands go to either HBA disks or to logical volumes comprised
   of SSDs.

The extra increment is causing device resets to hang.

 - Resets wait for all device outstanding commands to complete before
   returning.

Replace unused field abort_pending with retry_pending. This is a
maintenance driver so these changes have the least impact/risk.

Link: https://lore.kernel.org/r/161342801747.29388.13045495968308188518.stgit@brunhilda
Tested-by: Joe Szczypek <[email protected]>
Reviewed-by: Scott Benesh <[email protected]>
Reviewed-by: Scott Teel <[email protected]>
Reviewed-by: Tomas Henzl <[email protected]>
Signed-off-by: Don Brace <[email protected]>
Signed-off-by: Martin K. Petersen <[email protected]>

Author: Don Brace

drivers/scsi/hpsa.c

@@ -1151,7 +1151,10 @@ static void __enqueue_cmd_and_start_io(struct ctlr_info *h,
 {
 	dial_down_lockup_detection_during_fw_flash(h, c);
 	atomic_inc(&h->commands_outstanding);
-	if (c->device)
+	/*
+	 * Check to see if the command is being retried.
+	 */
+	if (c->device && !c->retry_pending)
 		atomic_inc(&c->device->commands_outstanding);
 
 	reply_queue = h->reply_map[raw_smp_processor_id()];
@@ -5567,7 +5570,8 @@ static inline void hpsa_cmd_partial_init(struct ctlr_info *h, int index,
 }
 
 static int hpsa_ioaccel_submit(struct ctlr_info *h,
-		struct CommandList *c, struct scsi_cmnd *cmd)
+		struct CommandList *c, struct scsi_cmnd *cmd,
+		bool retry)
 {
 	struct hpsa_scsi_dev_t *dev = cmd->device->hostdata;
 	int rc = IO_ACCEL_INELIGIBLE;
@@ -5584,18 +5588,22 @@ static int hpsa_ioaccel_submit(struct ctlr_info *h,
 	cmd->host_scribble = (unsigned char *) c;
 
 	if (dev->offload_enabled) {
-		hpsa_cmd_init(h, c->cmdindex, c);
+		hpsa_cmd_init(h, c->cmdindex, c); /* Zeroes out all fields */
 		c->cmd_type = CMD_SCSI;
 		c->scsi_cmd = cmd;
 		c->device = dev;
+		if (retry) /* Resubmit but do not increment device->commands_outstanding. */
+			c->retry_pending = true;
 		rc = hpsa_scsi_ioaccel_raid_map(h, c);
 		if (rc < 0)     /* scsi_dma_map failed. */
 			rc = SCSI_MLQUEUE_HOST_BUSY;
 	} else if (dev->hba_ioaccel_enabled) {
-		hpsa_cmd_init(h, c->cmdindex, c);
+		hpsa_cmd_init(h, c->cmdindex, c); /* Zeroes out all fields */
 		c->cmd_type = CMD_SCSI;
 		c->scsi_cmd = cmd;
 		c->device = dev;
+		if (retry) /* Resubmit but do not increment device->commands_outstanding. */
+			c->retry_pending = true;
 		rc = hpsa_scsi_ioaccel_direct_map(h, c);
 		if (rc < 0)     /* scsi_dma_map failed. */
 			rc = SCSI_MLQUEUE_HOST_BUSY;
@@ -5628,7 +5636,8 @@ static void hpsa_command_resubmit_worker(struct work_struct *work)
 
 		if (c2->error_data.serv_response ==
 				IOACCEL2_STATUS_SR_TASK_COMP_SET_FULL) {
-			rc = hpsa_ioaccel_submit(h, c, cmd);
+			/* Resubmit with the retry_pending flag set. */
+			rc = hpsa_ioaccel_submit(h, c, cmd, true);
 			if (rc == 0)
 				return;
 			if (rc == SCSI_MLQUEUE_HOST_BUSY) {
@@ -5644,6 +5653,15 @@ static void hpsa_command_resubmit_worker(struct work_struct *work)
 		}
 	}
 	hpsa_cmd_partial_init(c->h, c->cmdindex, c);
+	/*
+	 * Here we have not come in though queue_command, so we
+	 * can set the retry_pending flag to true for a driver initiated
+	 * retry attempt (I.E. not a SML retry).
+	 * I.E. We are submitting a driver initiated retry.
+	 * Note: hpsa_ciss_submit does not zero out the command fields like
+	 *       ioaccel submit does.
+	 */
+	c->retry_pending = true;
 	if (hpsa_ciss_submit(c->h, c, cmd, dev)) {
 		/*
 		 * If we get here, it means dma mapping failed. Try
@@ -5706,11 +5724,16 @@ static int hpsa_scsi_queue_command(struct Scsi_Host *sh, struct scsi_cmnd *cmd)
 	/*
 	 * Call alternate submit routine for I/O accelerated commands.
 	 * Retries always go down the normal I/O path.
+	 * Note: If cmd->retries is non-zero, then this is a SML
+	 *       initiated retry and not a driver initiated retry.
+	 *       This command has been obtained from cmd_tagged_alloc
+	 *       and is therefore a brand-new command.
 	 */
 	if (likely(cmd->retries == 0 &&
 			!blk_rq_is_passthrough(cmd->request) &&
 			h->acciopath_status)) {
-		rc = hpsa_ioaccel_submit(h, c, cmd);
+		/* Submit with the retry_pending flag unset. */
+		rc = hpsa_ioaccel_submit(h, c, cmd, false);
 		if (rc == 0)
 			return 0;
 		if (rc == SCSI_MLQUEUE_HOST_BUSY) {
@@ -6105,6 +6128,7 @@ static int hpsa_eh_device_reset_handler(struct scsi_cmnd *scsicmd)
  * at init, and managed by cmd_tagged_alloc() and cmd_tagged_free() using the
  * block request tag as an index into a table of entries.  cmd_tagged_free() is
  * the complement, although cmd_free() may be called instead.
+ * This function is only called for new requests from queue_command.
  */
 static struct CommandList *cmd_tagged_alloc(struct ctlr_info *h,
 					    struct scsi_cmnd *scmd)
@@ -6139,8 +6163,14 @@ static struct CommandList *cmd_tagged_alloc(struct ctlr_info *h,
 	}
 
 	atomic_inc(&c->refcount);
-
 	hpsa_cmd_partial_init(h, idx, c);
+
+	/*
+	 * This is a new command obtained from queue_command so
+	 * there have not been any driver initiated retry attempts.
+	 */
+	c->retry_pending = false;
+
 	return c;
 }
 
@@ -6208,6 +6238,13 @@ static struct CommandList *cmd_alloc(struct ctlr_info *h)
 	}
 	hpsa_cmd_partial_init(h, i, c);
 	c->device = NULL;
+
+	/*
+	 * cmd_alloc is for "internal" commands and they are never
+	 * retried.
+	 */
+	c->retry_pending = false;
+
 	return c;
 }
 

drivers/scsi/hpsa_cmd.h

@@ -448,7 +448,7 @@ struct CommandList {
 	 */
 	struct hpsa_scsi_dev_t *phys_disk;
 
-	int abort_pending;
+	bool retry_pending;
 	struct hpsa_scsi_dev_t *device;
 	atomic_t refcount; /* Must be last to avoid memset in hpsa_cmd_init() */
 } __aligned(COMMANDLIST_ALIGNMENT);