isci: Terminate outstanding TCs on TX/RX RNC suspensions.

TCs must only be terminated when RNCs are suspended.

Signed-off-by: Jeff Skirvin <jeffrey.d.skirvin@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>

7 files changed, 188 insertions, 32 deletions

diff --git a/drivers/scsi/isci/host.c b/drivers/scsi/isci/host.c
index 5832b13e7b07..d241b5722eb3 100644
--- a/drivers/scsi/isci/host.c
+++ b/drivers/scsi/isci/host.c
@@ -2696,18 +2696,18 @@ enum sci_status sci_controller_terminate_request(struct isci_host *ihost,
                          __func__, ihost->sm.current_state_id);
                 return SCI_FAILURE_INVALID_STATE;
         }
-
         status = sci_io_request_terminate(ireq);
-        if (status != SCI_SUCCESS)
-                return status;
-
-        /*
-         * Utilize the original post context command and or in the POST_TC_ABORT
-         * request sub-type.
-         */
-        sci_controller_post_request(ihost,
-                                    ireq->post_context | SCU_CONTEXT_COMMAND_REQUEST_POST_TC_ABORT);
-        return SCI_SUCCESS;
+        if ((status == SCI_SUCCESS) &&
+            !test_bit(IREQ_PENDING_ABORT, &ireq->flags) &&
+            !test_and_set_bit(IREQ_TC_ABORT_POSTED, &ireq->flags)) {
+                /* Utilize the original post context command and or in the
+                 * POST_TC_ABORT request sub-type.
+                 */
+                sci_controller_post_request(
+                        ihost, ireq->post_context |
+                                SCU_CONTEXT_COMMAND_REQUEST_POST_TC_ABORT);
+        }
+        return status;
 }
 
 /**
diff --git a/drivers/scsi/isci/remote_device.c b/drivers/scsi/isci/remote_device.c
index b1a8000a5ef8..9f03877534d2 100644
--- a/drivers/scsi/isci/remote_device.c
+++ b/drivers/scsi/isci/remote_device.c
@@ -133,6 +133,50 @@ static void isci_remote_device_ready(struct isci_host *ihost, struct isci_remote
                 wake_up(&ihost->eventq);
 }
 
+static enum sci_status sci_remote_device_suspend(
+        struct isci_remote_device *idev)
+{
+        return sci_remote_node_context_suspend(
+                &idev->rnc,
+                SCI_SOFTWARE_SUSPENSION,
+                SCI_SOFTWARE_SUSPEND_EXPECTED_EVENT,
+                NULL, NULL);
+}
+
+enum sci_status isci_remote_device_suspend(
+        struct isci_host *ihost,
+        struct isci_remote_device *idev)
+{
+        enum sci_status status;
+        unsigned long flags;
+
+        spin_lock_irqsave(&ihost->scic_lock, flags);
+
+        if (isci_lookup_device(idev->domain_dev) == NULL) {
+                spin_unlock_irqrestore(&ihost->scic_lock, flags);
+                status = SCI_FAILURE;
+        } else {
+                status = sci_remote_device_suspend(idev);
+                spin_unlock_irqrestore(&ihost->scic_lock, flags);
+                if (status == SCI_SUCCESS) {
+                        wait_event(ihost->eventq,
+                                   test_bit(IDEV_TXRX_SUSPENDED, &idev->flags)
+                                || !test_bit(IDEV_ALLOCATED, &idev->flags));
+
+                        status = test_bit(IDEV_TXRX_SUSPENDED, &idev->flags)
+                                        ? SCI_SUCCESS : SCI_FAILURE;
+                        dev_dbg(&ihost->pdev->dev,
+                                "%s: idev=%p, wait done, device is %s\n",
+                                __func__, idev,
+                                test_bit(IDEV_TXRX_SUSPENDED, &idev->flags)
+                                        ? "<suspended>" : "<deallocated!>");
+
+                }
+                isci_put_device(idev);
+        }
+        return status;
+}
+
 /* called once the remote node context is ready to be freed.
  * The remote device can now report that its stop operation is complete. none
  */
@@ -144,7 +188,9 @@ static void rnc_destruct_done(void *_dev)
         sci_change_state(&idev->sm, SCI_DEV_STOPPED);
 }
 
-static enum sci_status sci_remote_device_terminate_requests(struct isci_remote_device *idev)
+static enum sci_status sci_remote_device_terminate_requests_checkabort(
+        struct isci_remote_device *idev,
+        int check_abort_pending)
 {
         struct isci_host *ihost = idev->owning_port->owning_controller;
         enum sci_status status  = SCI_SUCCESS;
@@ -155,7 +201,9 @@ static enum sci_status sci_remote_device_terminate_requests(struct isci_remote_d
                 enum sci_status s;
 
                 if (!test_bit(IREQ_ACTIVE, &ireq->flags) ||
-                    ireq->target_device != idev)
+                    (ireq->target_device != idev) ||
+                    (check_abort_pending && !test_bit(IREQ_PENDING_ABORT,
+                                                      &ireq->flags)))
                         continue;
 
                 s = sci_controller_terminate_request(ihost, idev, ireq);
@@ -166,6 +214,12 @@ static enum sci_status sci_remote_device_terminate_requests(struct isci_remote_d
         return status;
 }
 
+enum sci_status sci_remote_device_terminate_requests(
+        struct isci_remote_device *idev)
+{
+        return sci_remote_device_terminate_requests_checkabort(idev, 0);
+}
+
 enum sci_status sci_remote_device_stop(struct isci_remote_device *idev,
                                         u32 timeout)
 {
@@ -265,14 +319,6 @@ enum sci_status sci_remote_device_reset_complete(struct isci_remote_device *idev
         return SCI_SUCCESS;
 }
 
-enum sci_status sci_remote_device_suspend(struct isci_remote_device *idev)
-{
-        return sci_remote_node_context_suspend(&idev->rnc,
-                                               SCI_SOFTWARE_SUSPENSION,
-                                               SCI_SOFTWARE_SUSPEND_EXPECTED_EVENT,
-                                               NULL, NULL);
-}
-
 enum sci_status sci_remote_device_frame_handler(struct isci_remote_device *idev,
                                                      u32 frame_index)
 {
@@ -1186,7 +1232,7 @@ static enum sci_status sci_remote_device_ea_construct(struct isci_port *iport,
  * the device when there have been no phys added to it.
  */
 static enum sci_status sci_remote_device_start(struct isci_remote_device *idev,
-                                                u32 timeout)
+                                               u32 timeout)
 {
         struct sci_base_state_machine *sm = &idev->sm;
         enum sci_remote_device_states state = sm->current_state_id;
@@ -1413,3 +1459,41 @@ int isci_remote_device_found(struct domain_device *dev)
 
         return status == SCI_SUCCESS ? 0 : -ENODEV;
 }
+
+enum sci_status isci_remote_device_reset(
+        struct isci_remote_device *idev)
+{
+        struct isci_host *ihost = dev_to_ihost(idev->domain_dev);
+        unsigned long flags;
+        enum sci_status status;
+
+        /* Wait for the device suspend. */
+        status = isci_remote_device_suspend(ihost, idev);
+        if (status != SCI_SUCCESS) {
+                dev_dbg(&ihost->pdev->dev,
+                        "%s: isci_remote_device_suspend(%p) returned %d!\n",
+                        __func__, idev, status);
+                return status;
+        }
+        spin_lock_irqsave(&ihost->scic_lock, flags);
+        status = sci_remote_device_reset(idev);
+        spin_unlock_irqrestore(&ihost->scic_lock, flags);
+        if (status != SCI_SUCCESS) {
+                dev_dbg(&ihost->pdev->dev,
+                        "%s: sci_remote_device_reset(%p) returned %d!\n",
+                        __func__, idev, status);
+        }
+        return status;
+}
+
+int isci_remote_device_is_safe_to_abort(
+        struct isci_remote_device *idev)
+{
+        return sci_remote_node_context_is_safe_to_abort(&idev->rnc);
+}
+
+enum sci_status sci_remote_device_abort_requests_pending_abort(
+        struct isci_remote_device *idev)
+{
+        return sci_remote_device_terminate_requests_checkabort(idev, 1);
+}
diff --git a/drivers/scsi/isci/remote_device.h b/drivers/scsi/isci/remote_device.h
index 39159053b7ff..ae508ee33664 100644
--- a/drivers/scsi/isci/remote_device.h
+++ b/drivers/scsi/isci/remote_device.h
@@ -85,6 +85,7 @@ struct isci_remote_device {
         #define IDEV_GONE 3
         #define IDEV_IO_READY 4
         #define IDEV_IO_NCQERROR 5
+        #define IDEV_TXRX_SUSPENDED 6
         unsigned long flags;
         struct kref kref;
         struct isci_port *isci_port;
@@ -335,4 +336,13 @@ void sci_remote_device_post_request(
         struct isci_remote_device *idev,
         u32 request);
 
+enum sci_status sci_remote_device_terminate_requests(
+        struct isci_remote_device *idev);
+
+int isci_remote_device_is_safe_to_abort(
+        struct isci_remote_device *idev);
+
+enum sci_status
+sci_remote_device_abort_requests_pending_abort(
+        struct isci_remote_device *idev);
 #endif /* !defined(_ISCI_REMOTE_DEVICE_H_) */
diff --git a/drivers/scsi/isci/remote_node_context.c b/drivers/scsi/isci/remote_node_context.c
index f180c726c5bb..7a8347e51767 100644
--- a/drivers/scsi/isci/remote_node_context.c
+++ b/drivers/scsi/isci/remote_node_context.c
@@ -268,6 +268,8 @@ static void sci_remote_node_context_invalidating_state_enter(struct sci_base_sta
 {
         struct sci_remote_node_context *rnc = container_of(sm, typeof(*rnc), sm);
 
+        /* Terminate outstanding requests pending abort. */
+        sci_remote_device_abort_requests_pending_abort(rnc_to_dev(rnc));
         sci_remote_node_context_invalidate_context_buffer(rnc);
 }
 
@@ -312,10 +314,28 @@ static void sci_remote_node_context_tx_suspended_state_enter(struct sci_base_sta
 static void sci_remote_node_context_tx_rx_suspended_state_enter(struct sci_base_state_machine *sm)
 {
         struct sci_remote_node_context *rnc = container_of(sm, typeof(*rnc), sm);
+        struct isci_remote_device *idev = rnc_to_dev(rnc);
+        struct isci_host *ihost = idev->owning_port->owning_controller;
+
+        set_bit(IDEV_TXRX_SUSPENDED, &idev->flags);
 
+        /* Terminate outstanding requests pending abort. */
+        sci_remote_device_abort_requests_pending_abort(idev);
+
+        wake_up(&ihost->eventq);
         sci_remote_node_context_continue_state_transitions(rnc);
 }
 
+static void sci_remote_node_context_tx_rx_suspended_state_exit(
+        struct sci_base_state_machine *sm)
+{
+        struct sci_remote_node_context *rnc
+                = container_of(sm, typeof(*rnc), sm);
+        struct isci_remote_device *idev = rnc_to_dev(rnc);
+
+        clear_bit(IDEV_TXRX_SUSPENDED, &idev->flags);
+}
+
 static void sci_remote_node_context_await_suspend_state_exit(
         struct sci_base_state_machine *sm)
 {
@@ -346,6 +366,8 @@ static const struct sci_base_state sci_remote_node_context_state_table[] = {
         },
         [SCI_RNC_TX_RX_SUSPENDED] = {
                 .enter_state = sci_remote_node_context_tx_rx_suspended_state_enter,
+                .exit_state
+                        = sci_remote_node_context_tx_rx_suspended_state_exit,
         },
         [SCI_RNC_AWAIT_SUSPENSION] = {
                 .exit_state = sci_remote_node_context_await_suspend_state_exit,
@@ -515,6 +537,13 @@ enum sci_status sci_remote_node_context_suspend(
         struct isci_remote_device *idev = rnc_to_dev(sci_rnc);
         enum sci_status status = SCI_FAILURE_INVALID_STATE;
 
+        dev_dbg(scirdev_to_dev(idev),
+                "%s: current state %d, current suspend_type %x dest state %d,"
+                        " arg suspend_reason %d, arg suspend_type %x",
+                __func__, state, sci_rnc->suspend_type,
+                sci_rnc->destination_state, suspend_reason,
+                suspend_type);
+
         /* Disable automatic state continuations if explicitly suspending. */
         if (suspend_reason == SCI_SOFTWARE_SUSPENSION)
                 sci_rnc->destination_state
@@ -546,7 +575,10 @@ enum sci_status sci_remote_node_context_suspend(
         sci_rnc->suspend_type  = suspend_type;
 
         if (status == SCI_SUCCESS) { /* Already in the destination state? */
+                struct isci_host *ihost = idev->owning_port->owning_controller;
+
                 sci_remote_node_context_notify_user(sci_rnc);
+                wake_up_all(&ihost->eventq); /* Let observers look. */
                 return SCI_SUCCESS;
         }
         if (suspend_reason == SCI_SOFTWARE_SUSPENSION) {
@@ -661,3 +693,27 @@ enum sci_status sci_remote_node_context_start_task(struct sci_remote_node_contex
                 return SCI_FAILURE_INVALID_STATE;
         }
 }
+
+int sci_remote_node_context_is_safe_to_abort(
+        struct sci_remote_node_context *sci_rnc)
+{
+        enum scis_sds_remote_node_context_states state;
+
+        state = sci_rnc->sm.current_state_id;
+        switch (state) {
+        case SCI_RNC_INVALIDATING:
+        case SCI_RNC_TX_RX_SUSPENDED:
+                return 1;
+        case SCI_RNC_POSTING:
+        case SCI_RNC_RESUMING:
+        case SCI_RNC_READY:
+        case SCI_RNC_TX_SUSPENDED:
+        case SCI_RNC_AWAIT_SUSPENSION:
+        case SCI_RNC_INITIAL:
+                return 0;
+        default:
+                dev_warn(scirdev_to_dev(rnc_to_dev(sci_rnc)),
+                         "%s: invalid state %d\n", __func__, state);
+                return 0;
+        }
+}
diff --git a/drivers/scsi/isci/remote_node_context.h b/drivers/scsi/isci/remote_node_context.h
index 276fc491e8f9..5ddf88b53133 100644
--- a/drivers/scsi/isci/remote_node_context.h
+++ b/drivers/scsi/isci/remote_node_context.h
@@ -214,5 +214,7 @@ enum sci_status sci_remote_node_context_start_task(struct sci_remote_node_contex
                                                         struct isci_request *ireq);
 enum sci_status sci_remote_node_context_start_io(struct sci_remote_node_context *sci_rnc,
                                                       struct isci_request *ireq);
+int sci_remote_node_context_is_safe_to_abort(
+        struct sci_remote_node_context *sci_rnc);
 
 #endif  /* _SCIC_SDS_REMOTE_NODE_CONTEXT_H_ */
diff --git a/drivers/scsi/isci/request.c b/drivers/scsi/isci/request.c
index 605dc68cbf79..1f314d0d71d5 100644
--- a/drivers/scsi/isci/request.c
+++ b/drivers/scsi/isci/request.c
@@ -863,6 +863,8 @@ sci_io_request_terminate(struct isci_request *ireq)
 
         switch (state) {
         case SCI_REQ_CONSTRUCTED:
+                /* Set to make sure no HW terminate posting is done: */
+                set_bit(IREQ_TC_ABORT_POSTED, &ireq->flags);
                 ireq->scu_status = SCU_TASK_DONE_TASK_ABORT;
                 ireq->sci_status = SCI_FAILURE_IO_TERMINATED;
                 sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
@@ -883,8 +885,7 @@ sci_io_request_terminate(struct isci_request *ireq)
         case SCI_REQ_ATAPI_WAIT_PIO_SETUP:
         case SCI_REQ_ATAPI_WAIT_D2H:
         case SCI_REQ_ATAPI_WAIT_TC_COMP:
-                sci_change_state(&ireq->sm, SCI_REQ_ABORTING);
-                return SCI_SUCCESS;
+                /* Fall through and change state to ABORTING... */
         case SCI_REQ_TASK_WAIT_TC_RESP:
                 /* The task frame was already confirmed to have been
                  * sent by the SCU HW.  Since the state machine is
@@ -893,20 +894,21 @@ sci_io_request_terminate(struct isci_request *ireq)
                  * and don't wait for the task response.
                  */
                 sci_change_state(&ireq->sm, SCI_REQ_ABORTING);
-                sci_change_state(&ireq->sm, SCI_REQ_COMPLETED);
-                return SCI_SUCCESS;
+                /* Fall through and handle like ABORTING... */
         case SCI_REQ_ABORTING:
-                /* If a request has a termination requested twice, return
-                 * a failure indication, since HW confirmation of the first
-                 * abort is still outstanding.
+                if (!isci_remote_device_is_safe_to_abort(ireq->target_device))
+                        set_bit(IREQ_PENDING_ABORT, &ireq->flags);
+                else
+                        clear_bit(IREQ_PENDING_ABORT, &ireq->flags);
+                /* If the request is only waiting on the remote device
+                 * suspension, return SUCCESS so the caller will wait too.
                  */
+                return SCI_SUCCESS;
         case SCI_REQ_COMPLETED:
         default:
                 dev_warn(&ireq->owning_controller->pdev->dev,
                          "%s: SCIC IO Request requested to abort while in wrong "
-                         "state %d\n",
-                         __func__,
-                         ireq->sm.current_state_id);
+                         "state %d\n", __func__, ireq->sm.current_state_id);
                 break;
         }
 
diff --git a/drivers/scsi/isci/request.h b/drivers/scsi/isci/request.h
index e845a31ecebb..8d55f78010aa 100644
--- a/drivers/scsi/isci/request.h
+++ b/drivers/scsi/isci/request.h
@@ -102,6 +102,8 @@ struct isci_request {
         #define IREQ_TERMINATED 1
         #define IREQ_TMF 2
         #define IREQ_ACTIVE 3
+        #define IREQ_PENDING_ABORT 4 /* Set == device was not suspended yet */
+        #define IREQ_TC_ABORT_POSTED 5
         unsigned long flags;
         /* XXX kill ttype and ttype_ptr, allocate full sas_task */
         union ttype_ptr_union {