Queue CCBs internally instead of using CAM_REQUEUE_REQ status. This fixes

problem where userspace apps such as smartctl fail due to CAM_REQUEUE_REQ
status getting returned when tagged commands are outstanding when smartctl
sends its I/O using the pass(4) interface.

Sponsored by: Intel
Found and tested by: Ravi Pokala <rpokala at panasas dot com>
Reviewed by: scottl
Approved by: scottl
MFC after: 1 week

4 files changed, 56 insertions, 29 deletions

diff --git a/sys/dev/isci/isci.h b/sys/dev/isci/isci.h
index 50f0236..d6ad93c 100644
--- a/sys/dev/isci/isci.h
+++ b/sys/dev/isci/isci.h
@@ -86,6 +86,7 @@ struct ISCI_REMOTE_DEVICE {
 	BOOL				is_resetting;
 	uint32_t			frozen_lun_mask;
 	SCI_FAST_LIST_ELEMENT_T		pending_device_reset_element;
+	TAILQ_HEAD(,ccb_hdr)		queued_ccbs;
 };
 
 struct ISCI_DOMAIN {
diff --git a/sys/dev/isci/isci_controller.c b/sys/dev/isci/isci_controller.c
index 8511400..a4b7cfe 100644
--- a/sys/dev/isci/isci_controller.c
+++ b/sys/dev/isci/isci_controller.c
@@ -430,6 +430,7 @@ int isci_controller_allocate_memory(struct ISCI_CONTROLLER *controller)
 		remote_device->frozen_lun_mask = 0;
 		sci_fast_list_element_init(remote_device,
 		    &remote_device->pending_device_reset_element);
+		TAILQ_INIT(&remote_device->queued_ccbs);
 
 		/*
 		 * For the first SCI_MAX_DOMAINS device objects, do not put
diff --git a/sys/dev/isci/isci_io_request.c b/sys/dev/isci/isci_io_request.c
index b671e67..985a2e4 100644
--- a/sys/dev/isci/isci_io_request.c
+++ b/sys/dev/isci/isci_io_request.c
@@ -85,7 +85,9 @@ isci_io_request_complete(SCI_CONTROLLER_HANDLE_T scif_controller,
 	struct ISCI_CONTROLLER *isci_controller;
 	struct ISCI_REMOTE_DEVICE *isci_remote_device;
 	union ccb *ccb;
+	BOOL complete_ccb;
 
+	complete_ccb = TRUE;
 	isci_controller = (struct ISCI_CONTROLLER *) sci_object_get_association(scif_controller);
 	isci_remote_device =
 		(struct ISCI_REMOTE_DEVICE *) sci_object_get_association(remote_device);
@@ -163,9 +165,7 @@ isci_io_request_complete(SCI_CONTROLLER_HANDLE_T scif_controller,
 
 	case SCI_IO_FAILURE_INVALID_STATE:
 	case SCI_IO_FAILURE_INSUFFICIENT_RESOURCES:
-		ccb->ccb_h.status |= CAM_REQUEUE_REQ;
-		isci_remote_device_freeze_lun_queue(isci_remote_device,
-		    ccb->ccb_h.target_lun);
+		complete_ccb = FALSE;
 		break;
 
 	case SCI_IO_FAILURE_INVALID_REMOTE_DEVICE:
@@ -189,7 +189,7 @@ isci_io_request_complete(SCI_CONTROLLER_HANDLE_T scif_controller,
 			    scif_remote_device_get_max_queue_depth(remote_device);
 			xpt_action((union ccb *)&ccb_relsim);
 			xpt_free_path(path);
-			ccb->ccb_h.status |= CAM_REQUEUE_REQ;
+			complete_ccb = FALSE;
 		}
 		break;
 
@@ -209,17 +209,6 @@ isci_io_request_complete(SCI_CONTROLLER_HANDLE_T scif_controller,
 		break;
 	}
 
-	if (ccb->ccb_h.status != CAM_REQ_CMP) {
-		/* ccb will be completed with some type of non-success
-		 *  status.  So temporarily freeze the queue until the
-		 *  upper layers can act on the status.  The CAM_DEV_QFRZN
-		 *  flag will then release the queue after the status is
-		 *  acted upon.
-		 */
-		ccb->ccb_h.status |= CAM_DEV_QFRZN;
-		xpt_freeze_devq(ccb->ccb_h.path, 1);
-	}
-
 	callout_stop(&isci_request->parent.timer);
 	bus_dmamap_sync(isci_request->parent.dma_tag,
 	    isci_request->parent.dma_map,
@@ -228,20 +217,43 @@ isci_io_request_complete(SCI_CONTROLLER_HANDLE_T scif_controller,
 	bus_dmamap_unload(isci_request->parent.dma_tag,
 	    isci_request->parent.dma_map);
 
-	if (isci_remote_device->frozen_lun_mask != 0 &&
-	    !(ccb->ccb_h.status & CAM_REQUEUE_REQ))
-		isci_remote_device_release_device_queue(isci_remote_device);
-
-	xpt_done(ccb);
 	isci_request->ccb = NULL;
 
-	if (isci_controller->is_frozen == TRUE) {
-		isci_controller->is_frozen = FALSE;
-		xpt_release_simq(isci_controller->sim, TRUE);
-	}
-
 	sci_pool_put(isci_controller->request_pool,
 	    (struct ISCI_REQUEST *)isci_request);
+
+	if (complete_ccb) {
+		if (ccb->ccb_h.status != CAM_REQ_CMP) {
+			/* ccb will be completed with some type of non-success
+			 *  status.  So temporarily freeze the queue until the
+			 *  upper layers can act on the status.  The
+			 *  CAM_DEV_QFRZN flag will then release the queue
+			 *  after the status is acted upon.
+			 */
+			ccb->ccb_h.status |= CAM_DEV_QFRZN;
+			xpt_freeze_devq(ccb->ccb_h.path, 1);
+		}
+
+		if (isci_remote_device->frozen_lun_mask != 0) {
+			isci_remote_device_release_device_queue(isci_remote_device);
+		}
+
+		xpt_done(ccb);
+
+		if (isci_controller->is_frozen == TRUE) {
+			isci_controller->is_frozen = FALSE;
+			xpt_release_simq(isci_controller->sim, TRUE);
+		}
+	} else {
+		isci_remote_device_freeze_lun_queue(isci_remote_device,
+		    ccb->ccb_h.target_lun);
+
+		isci_log_message(1, "ISCI", "queue %p %x\n", ccb,
+		    ccb->csio.cdb_io.cdb_bytes[0]);
+		ccb->ccb_h.status |= CAM_SIM_QUEUED;
+		TAILQ_INSERT_TAIL(&isci_remote_device->queued_ccbs,
+		    &ccb->ccb_h, sim_links.tqe);
+	}
 }
 
 /**
diff --git a/sys/dev/isci/isci_remote_device.c b/sys/dev/isci/isci_remote_device.c
index d22b4d4..c9434f8 100644
--- a/sys/dev/isci/isci_remote_device.c
+++ b/sys/dev/isci/isci_remote_device.c
@@ -289,9 +289,22 @@ isci_remote_device_release_lun_queue(struct ISCI_REMOTE_DEVICE *remote_device,
 
 void
 isci_remote_device_release_device_queue(
-    struct ISCI_REMOTE_DEVICE *remote_device)
+    struct ISCI_REMOTE_DEVICE *device)
 {
-	lun_id_t lun;
-	for (lun = 0; lun < ISCI_MAX_LUN; lun++)
-		isci_remote_device_release_lun_queue(remote_device, lun);
+	if (TAILQ_EMPTY(&device->queued_ccbs)) {
+		lun_id_t lun;
+
+		for (lun = 0; lun < ISCI_MAX_LUN; lun++)
+			isci_remote_device_release_lun_queue(device, lun);
+	} else {
+		struct ccb_hdr *ccb_h;
+
+		ccb_h = TAILQ_FIRST(&device->queued_ccbs);
+		TAILQ_REMOVE(&device->queued_ccbs, ccb_h, sim_links.tqe);
+		ccb_h->status &= ~CAM_SIM_QUEUED;
+		isci_log_message(1, "ISCI", "release %p %x\n", ccb_h,
+		    ((union ccb*)ccb_h)->csio.cdb_io.cdb_bytes[0]);
+		isci_io_request_execute_scsi_io((union ccb *)ccb_h,
+		    device->domain->controller);
+	}
 }