isci: Check IDEV_GONE before performing abort path operations.

In the link fail path, set IDEV_GONE for every device on the domain when the last link in the port fails. In the abort path functions like isci_reset_device, make sure that there has not already been a detected domain failure with the device by checking IDEV_GONE, before performing any kind of hard reset, SMP phy control, or TMF operation. The check for IDEV_GONE makes sure that the device in the abort path really has control of the port with which it is associated. This prevents starting hard resets at incorrect times and scheduling unnecessary LUN resets for SATA devices. Signed-off-by: Jeff Skirvin <jeffrey.d.skirvin@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>

isci: Check IDEV_GONE before performing abort path operations.
In the link fail path, set IDEV_GONE for every device on the domain when the last link in the port fails. In the abort path functions like isci_reset_device, make sure that there has not already been a detected domain failure with the device by checking IDEV_GONE, before performing any kind of hard reset, SMP phy control, or TMF operation. The check for IDEV_GONE makes sure that the device in the abort path really has control of the port with which it is associated. This prevents starting hard resets at incorrect times and scheduling unnecessary LUN resets for SATA devices. Signed-off-by: Jeff Skirvin <jeffrey.d.skirvin@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Jeff Skirvin · Dan Williams
1 parent 87805162b6
Showing 2 changed files with 57 additions and 21 deletions Side-by-side Diff
drivers/scsi/isci/port.c
drivers/scsi/isci/task.c
@@ -240,8 +240,31 @@
 				struct isci_phy *isci_phy,
 				struct isci_port *isci_port)
 {
+	struct isci_remote_device *isci_device;
+
 	dev_dbg(&isci_host->pdev->dev,
 		"%s: isci_port = %p\n", __func__, isci_port);
+
+	if (isci_port) {
+
+		/* check to see if this is the last phy on this port. */
+		if (isci_phy->sas_phy.port &&
+		    isci_phy->sas_phy.port->num_phys == 1) {
+			/* change the state for all devices on this port.  The
+			* next task sent to this device will be returned as
+			* SAS_TASK_UNDELIVERED, and the scsi mid layer will
+			* remove the target
+			*/
+			list_for_each_entry(isci_device,
+					    &isci_port->remote_dev_list,
+					    node) {
+				dev_dbg(&isci_host->pdev->dev,
+					"%s: isci_device = %p\n",
+					__func__, isci_device);
+				set_bit(IDEV_GONE, &isci_device->flags);
+			}
+		}
+	}
  
 	/* Notify libsas of the borken link, this will trigger calls to our
 	 * isci_port_deformed and isci_dev_gone functions.
@@ -421,7 +421,7 @@
 	struct isci_host *ihost = dev_to_ihost(dev);
 	struct isci_remote_device *idev;
 	unsigned long flags;
-	int ret;
+	int ret = TMF_RESP_FUNC_COMPLETE;
  
 	spin_lock_irqsave(&ihost->scic_lock, flags);
 	idev = isci_get_device(dev->lldd_dev);
@@ -447,12 +447,12 @@
 		goto out;
 	}
 	/* All pending I/Os have been terminated and cleaned up. */
-	if (dev_is_sata(dev)) {
-		sas_ata_schedule_reset(dev);
-		ret = TMF_RESP_FUNC_COMPLETE;
-	} else {
-		/* Send the task management part of the reset. */
-		ret = isci_task_send_lu_reset_sas(ihost, idev, lun);
+	if (!test_bit(IDEV_GONE, &idev->flags)) {
+		if (dev_is_sata(dev))
+			sas_ata_schedule_reset(dev);
+		else
+			/* Send the task management part of the reset. */
+			ret = isci_task_send_lu_reset_sas(ihost, idev, lun);
 	}
  out:
 	isci_put_device(idev);
@@ -512,8 +512,17 @@
 	spin_unlock_irqrestore(&ihost->scic_lock, flags);
  
 	dev_warn(&ihost->pdev->dev,
-		 "%s: dev = %p, task = %p, old_request == %p\n",
-		 __func__, idev, task, old_request);
+		 "%s: dev = %p (%s%s), task = %p, old_request == %p\n",
+		 __func__, idev,
+		 (dev_is_sata(task->dev) ? "STP/SATA"
+					 : ((dev_is_expander(task->dev))
+						? "SMP"
+						: "SSP")),
+		 ((idev) ? ((test_bit(IDEV_GONE, &idev->flags))
+			   ? " IDEV_GONE"
+			   : "")
+			 : " <NULL>"),
+		 task, old_request);
  
 	/* Device reset conditions signalled in task_state_flags are the
 	 * responsbility of libsas to observe at the start of the error
@@ -552,7 +561,8 @@
  
 	if (task->task_proto == SAS_PROTOCOL_SMP ||
 	    sas_protocol_ata(task->task_proto) ||
-	    test_bit(IREQ_COMPLETE_IN_TARGET, &old_request->flags)) {
+	    test_bit(IREQ_COMPLETE_IN_TARGET, &old_request->flags) ||
+	    test_bit(IDEV_GONE, &idev->flags)) {
  
 		spin_unlock_irqrestore(&ihost->scic_lock, flags);
  
@@ -561,7 +571,8 @@
  
 		dev_warn(&ihost->pdev->dev,
 			 "%s: %s request"
-				 " or complete_in_target (%d), thus no TMF\n",
+				 " or complete_in_target (%d), "
+				 "or IDEV_GONE (%d), thus no TMF\n",
 			 __func__,
 			 ((task->task_proto == SAS_PROTOCOL_SMP)
 			  ? "SMP"
@@ -570,7 +581,8 @@
 				: "<other>")
 			  ),
 			 test_bit(IREQ_COMPLETE_IN_TARGET,
-				  &old_request->flags));
+				  &old_request->flags),
+			 test_bit(IDEV_GONE, &idev->flags));
  
 		spin_lock_irqsave(&task->task_state_lock, flags);
 		task->task_state_flags &= ~(SAS_TASK_AT_INITIATOR |
@@ -734,7 +746,7 @@
 			     struct domain_device *dev,
 			     struct isci_remote_device *idev)
 {
-	int rc = TMF_RESP_FUNC_COMPLETE, reset_stat;
+	int rc = TMF_RESP_FUNC_COMPLETE, reset_stat = -1;
 	struct sas_phy *phy = sas_get_local_phy(dev);
 	struct isci_port *iport = dev->port->lldd_port;
  
  
@@ -752,14 +764,15 @@
 	 * primary duty of this function is to cleanup tasks, so that is the
 	 * relevant status.
 	 */
+	if (!test_bit(IDEV_GONE, &idev->flags)) {
+		if (scsi_is_sas_phy_local(phy)) {
+			struct isci_phy *iphy = &ihost->phys[phy->number];
  
-	if (scsi_is_sas_phy_local(phy)) {
-		struct isci_phy *iphy = &ihost->phys[phy->number];
-
-		reset_stat = isci_port_perform_hard_reset(ihost, iport, iphy);
-	} else
-		reset_stat = sas_phy_reset(phy, !dev_is_sata(dev));
-
+			reset_stat = isci_port_perform_hard_reset(ihost, iport,
+								  iphy);
+		} else
+			reset_stat = sas_phy_reset(phy, !dev_is_sata(dev));
+	}
 	/* Explicitly resume the RNC here, since there was no task sent. */
 	isci_remote_device_resume_from_abort(ihost, idev);
...	...	@@ -240,8 +240,31 @@
240	240	struct isci_phy *isci_phy,
241	241	struct isci_port *isci_port)
242	242	{
	243	+ struct isci_remote_device *isci_device;
	244	+
243	245	dev_dbg(&isci_host->pdev->dev,
244	246	"%s: isci_port = %p\n", __func__, isci_port);
	247	+
	248	+ if (isci_port) {
	249	+
	250	+ /* check to see if this is the last phy on this port. */
	251	+ if (isci_phy->sas_phy.port &&
	252	+ isci_phy->sas_phy.port->num_phys == 1) {
	253	+ /* change the state for all devices on this port. The
	254	+ * next task sent to this device will be returned as
	255	+ * SAS_TASK_UNDELIVERED, and the scsi mid layer will
	256	+ * remove the target
	257	+ */
	258	+ list_for_each_entry(isci_device,
	259	+ &isci_port->remote_dev_list,
	260	+ node) {
	261	+ dev_dbg(&isci_host->pdev->dev,
	262	+ "%s: isci_device = %p\n",
	263	+ __func__, isci_device);
	264	+ set_bit(IDEV_GONE, &isci_device->flags);
	265	+ }
	266	+ }
	267	+ }
245	268
246	269	/* Notify libsas of the borken link, this will trigger calls to our
247	270	* isci_port_deformed and isci_dev_gone functions.
...	...	@@ -421,7 +421,7 @@
421	421	struct isci_host *ihost = dev_to_ihost(dev);
422	422	struct isci_remote_device *idev;
423	423	unsigned long flags;
424		- int ret;
	424	+ int ret = TMF_RESP_FUNC_COMPLETE;
425	425
426	426	spin_lock_irqsave(&ihost->scic_lock, flags);
427	427	idev = isci_get_device(dev->lldd_dev);
...	...	@@ -447,12 +447,12 @@
447	447	goto out;
448	448	}
449	449	/* All pending I/Os have been terminated and cleaned up. */
450		- if (dev_is_sata(dev)) {
451		- sas_ata_schedule_reset(dev);
452		- ret = TMF_RESP_FUNC_COMPLETE;
453		- } else {
454		- /* Send the task management part of the reset. */
455		- ret = isci_task_send_lu_reset_sas(ihost, idev, lun);
	450	+ if (!test_bit(IDEV_GONE, &idev->flags)) {
	451	+ if (dev_is_sata(dev))
	452	+ sas_ata_schedule_reset(dev);
	453	+ else
	454	+ /* Send the task management part of the reset. */
	455	+ ret = isci_task_send_lu_reset_sas(ihost, idev, lun);
456	456	}
457	457	out:
458	458	isci_put_device(idev);
...	...	@@ -512,8 +512,17 @@
512	512	spin_unlock_irqrestore(&ihost->scic_lock, flags);
513	513
514	514	dev_warn(&ihost->pdev->dev,
515		- "%s: dev = %p, task = %p, old_request == %p\n",
516		- __func__, idev, task, old_request);
	515	+ "%s: dev = %p (%s%s), task = %p, old_request == %p\n",
	516	+ __func__, idev,
	517	+ (dev_is_sata(task->dev) ? "STP/SATA"
	518	+ : ((dev_is_expander(task->dev))
	519	+ ? "SMP"
	520	+ : "SSP")),
	521	+ ((idev) ? ((test_bit(IDEV_GONE, &idev->flags))
	522	+ ? " IDEV_GONE"
	523	+ : "")
	524	+ : " <NULL>"),
	525	+ task, old_request);
517	526
518	527	/* Device reset conditions signalled in task_state_flags are the
519	528	* responsbility of libsas to observe at the start of the error
...	...	@@ -552,7 +561,8 @@
552	561
553	562	if (task->task_proto == SAS_PROTOCOL_SMP \|\|
554	563	sas_protocol_ata(task->task_proto) \|\|
555		- test_bit(IREQ_COMPLETE_IN_TARGET, &old_request->flags)) {
	564	+ test_bit(IREQ_COMPLETE_IN_TARGET, &old_request->flags) \|\|
	565	+ test_bit(IDEV_GONE, &idev->flags)) {
556	566
557	567	spin_unlock_irqrestore(&ihost->scic_lock, flags);
558	568
...	...	@@ -561,7 +571,8 @@
561	571
562	572	dev_warn(&ihost->pdev->dev,
563	573	"%s: %s request"
564		- " or complete_in_target (%d), thus no TMF\n",
	574	+ " or complete_in_target (%d), "
	575	+ "or IDEV_GONE (%d), thus no TMF\n",
565	576	__func__,
566	577	((task->task_proto == SAS_PROTOCOL_SMP)
567	578	? "SMP"
...	...	@@ -570,7 +581,8 @@
570	581	: "<other>")
571	582	),
572	583	test_bit(IREQ_COMPLETE_IN_TARGET,
573		- &old_request->flags));
	584	+ &old_request->flags),
	585	+ test_bit(IDEV_GONE, &idev->flags));
574	586
575	587	spin_lock_irqsave(&task->task_state_lock, flags);
576	588	task->task_state_flags &= ~(SAS_TASK_AT_INITIATOR \|
...	...	@@ -734,7 +746,7 @@
734	746	struct domain_device *dev,
735	747	struct isci_remote_device *idev)
736	748	{
737		- int rc = TMF_RESP_FUNC_COMPLETE, reset_stat;
	749	+ int rc = TMF_RESP_FUNC_COMPLETE, reset_stat = -1;
738	750	struct sas_phy *phy = sas_get_local_phy(dev);
739	751	struct isci_port *iport = dev->port->lldd_port;
740	752
741	753
...	...	@@ -752,14 +764,15 @@
752	764	* primary duty of this function is to cleanup tasks, so that is the
753	765	* relevant status.
754	766	*/
	767	+ if (!test_bit(IDEV_GONE, &idev->flags)) {
	768	+ if (scsi_is_sas_phy_local(phy)) {
	769	+ struct isci_phy *iphy = &ihost->phys[phy->number];
755	770
756		- if (scsi_is_sas_phy_local(phy)) {
757		- struct isci_phy *iphy = &ihost->phys[phy->number];
758		-
759		- reset_stat = isci_port_perform_hard_reset(ihost, iport, iphy);
760		- } else
761		- reset_stat = sas_phy_reset(phy, !dev_is_sata(dev));
762		-
	771	+ reset_stat = isci_port_perform_hard_reset(ihost, iport,
	772	+ iphy);
	773	+ } else
	774	+ reset_stat = sas_phy_reset(phy, !dev_is_sata(dev));
	775	+ }
763	776	/* Explicitly resume the RNC here, since there was no task sent. */
764	777	isci_remote_device_resume_from_abort(ihost, idev);
765	778