[S390] dasd: improve error recovery for internal I/O

Most of the error conditions reported by a FICON storage server
indicate situations which can be recovered. Sometimes the host just
needs to retry an I/O request, but sometimes the recovery
is more complex and requires the device driver to wait, choose
a different path, etc.

The DASD device driver has a fully featured error recovery
for normal block layer I/O, but not for internal I/O request which
are for example used during the device bring up.
This can lead to situations where the IPL of a system fails because
DASD devices are not properly recognized.
This patch will extend the internal I/O handling to use the existing
error recovery procedures.

Signed-off-by: Stefan Weinhuber <wein@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>

1 files changed, 152 insertions, 55 deletions

diff --git a/drivers/s390/block/dasd.c b/drivers/s390/block/dasd.c
index 329115a4d4b3..4f211c175b55 100644
--- a/drivers/s390/block/dasd.c
+++ b/drivers/s390/block/dasd.c
@@ -63,6 +63,7 @@ static void do_restore_device(struct work_struct *);
 static void dasd_return_cqr_cb(struct dasd_ccw_req *, void *);
 static void dasd_device_timeout(unsigned long);
 static void dasd_block_timeout(unsigned long);
+static void __dasd_process_erp(struct dasd_device *, struct dasd_ccw_req *);
 
 /*
  * SECTION: Operations on the device structure.
@@ -959,7 +960,7 @@ static void dasd_device_timeout(unsigned long ptr)
 	device = (struct dasd_device *) ptr;
 	spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
 	/* re-activate request queue */
-        device->stopped &= ~DASD_STOPPED_PENDING;
+	dasd_device_remove_stop_bits(device, DASD_STOPPED_PENDING);
 	spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
 	dasd_schedule_device_bh(device);
 }
@@ -1022,7 +1023,7 @@ void dasd_generic_handle_state_change(struct dasd_device *device)
 	/* First of all start sense subsystem status request. */
 	dasd_eer_snss(device);
 
-	device->stopped &= ~DASD_STOPPED_PENDING;
+	dasd_device_remove_stop_bits(device, DASD_STOPPED_PENDING);
 	dasd_schedule_device_bh(device);
 	if (device->block)
 		dasd_schedule_block_bh(device->block);
@@ -1404,6 +1405,20 @@ void dasd_schedule_device_bh(struct dasd_device *device)
 	tasklet_hi_schedule(&device->tasklet);
 }
 
+void dasd_device_set_stop_bits(struct dasd_device *device, int bits)
+{
+	device->stopped |= bits;
+}
+EXPORT_SYMBOL_GPL(dasd_device_set_stop_bits);
+
+void dasd_device_remove_stop_bits(struct dasd_device *device, int bits)
+{
+	device->stopped &= ~bits;
+	if (!device->stopped)
+		wake_up(&generic_waitq);
+}
+EXPORT_SYMBOL_GPL(dasd_device_remove_stop_bits);
+
 /*
  * Queue a request to the head of the device ccw_queue.
  * Start the I/O if possible.
@@ -1464,58 +1479,135 @@ static inline int _wait_for_wakeup(struct dasd_ccw_req *cqr)
 }
 
 /*
- * Queue a request to the tail of the device ccw_queue and wait for
- * it's completion.
+ * checks if error recovery is necessary, returns 1 if yes, 0 otherwise.
  */
-int dasd_sleep_on(struct dasd_ccw_req *cqr)
+static int __dasd_sleep_on_erp(struct dasd_ccw_req *cqr)
 {
 	struct dasd_device *device;
-	int rc;
+	dasd_erp_fn_t erp_fn;
 
+	if (cqr->status == DASD_CQR_FILLED)
+		return 0;
 	device = cqr->startdev;
+	if (test_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags)) {
+		if (cqr->status == DASD_CQR_TERMINATED) {
+			device->discipline->handle_terminated_request(cqr);
+			return 1;
+		}
+		if (cqr->status == DASD_CQR_NEED_ERP) {
+			erp_fn = device->discipline->erp_action(cqr);
+			erp_fn(cqr);
+			return 1;
+		}
+		if (cqr->status == DASD_CQR_FAILED)
+			dasd_log_sense(cqr, &cqr->irb);
+		if (cqr->refers) {
+			__dasd_process_erp(device, cqr);
+			return 1;
+		}
+	}
+	return 0;
+}
 
-	cqr->callback = dasd_wakeup_cb;
-	cqr->callback_data = (void *) &generic_waitq;
-	dasd_add_request_tail(cqr);
-	wait_event(generic_waitq, _wait_for_wakeup(cqr));
+static int __dasd_sleep_on_loop_condition(struct dasd_ccw_req *cqr)
+{
+	if (test_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags)) {
+		if (cqr->refers) /* erp is not done yet */
+			return 1;
+		return ((cqr->status != DASD_CQR_DONE) &&
+			(cqr->status != DASD_CQR_FAILED));
+	} else
+		return (cqr->status == DASD_CQR_FILLED);
+}
 
-	if (cqr->status == DASD_CQR_DONE)
+static int _dasd_sleep_on(struct dasd_ccw_req *maincqr, int interruptible)
+{
+	struct dasd_device *device;
+	int rc;
+	struct list_head ccw_queue;
+	struct dasd_ccw_req *cqr;
+
+	INIT_LIST_HEAD(&ccw_queue);
+	maincqr->status = DASD_CQR_FILLED;
+	device = maincqr->startdev;
+	list_add(&maincqr->blocklist, &ccw_queue);
+	for (cqr = maincqr;  __dasd_sleep_on_loop_condition(cqr);
+	     cqr = list_first_entry(&ccw_queue,
+				    struct dasd_ccw_req, blocklist)) {
+
+		if (__dasd_sleep_on_erp(cqr))
+			continue;
+		if (cqr->status != DASD_CQR_FILLED) /* could be failed */
+			continue;
+
+		/* Non-temporary stop condition will trigger fail fast */
+		if (device->stopped & ~DASD_STOPPED_PENDING &&
+		    test_bit(DASD_CQR_FLAGS_FAILFAST, &cqr->flags) &&
+		    (!dasd_eer_enabled(device))) {
+			cqr->status = DASD_CQR_FAILED;
+			continue;
+		}
+
+		/* Don't try to start requests if device is stopped */
+		if (interruptible) {
+			rc = wait_event_interruptible(
+				generic_waitq, !(device->stopped));
+			if (rc == -ERESTARTSYS) {
+				cqr->status = DASD_CQR_FAILED;
+				maincqr->intrc = rc;
+				continue;
+			}
+		} else
+			wait_event(generic_waitq, !(device->stopped));
+
+		cqr->callback = dasd_wakeup_cb;
+		cqr->callback_data = (void *) &generic_waitq;
+		dasd_add_request_tail(cqr);
+		if (interruptible) {
+			rc = wait_event_interruptible(
+				generic_waitq, _wait_for_wakeup(cqr));
+			if (rc == -ERESTARTSYS) {
+				dasd_cancel_req(cqr);
+				/* wait (non-interruptible) for final status */
+				wait_event(generic_waitq,
+					   _wait_for_wakeup(cqr));
+				cqr->status = DASD_CQR_FAILED;
+				maincqr->intrc = rc;
+				continue;
+			}
+		} else
+			wait_event(generic_waitq, _wait_for_wakeup(cqr));
+	}
+
+	maincqr->endclk = get_clock();
+	if ((maincqr->status != DASD_CQR_DONE) &&
+	    (maincqr->intrc != -ERESTARTSYS))
+		dasd_log_sense(maincqr, &maincqr->irb);
+	if (maincqr->status == DASD_CQR_DONE)
 		rc = 0;
-	else if (cqr->intrc)
-		rc = cqr->intrc;
+	else if (maincqr->intrc)
+		rc = maincqr->intrc;
 	else
 		rc = -EIO;
 	return rc;
 }
 
 /*
+ * Queue a request to the tail of the device ccw_queue and wait for
+ * it's completion.
+ */
+int dasd_sleep_on(struct dasd_ccw_req *cqr)
+{
+	return _dasd_sleep_on(cqr, 0);
+}
+
+/*
  * Queue a request to the tail of the device ccw_queue and wait
  * interruptible for it's completion.
  */
 int dasd_sleep_on_interruptible(struct dasd_ccw_req *cqr)
 {
-	struct dasd_device *device;
-	int rc;
-
-	device = cqr->startdev;
-	cqr->callback = dasd_wakeup_cb;
-	cqr->callback_data = (void *) &generic_waitq;
-	dasd_add_request_tail(cqr);
-	rc = wait_event_interruptible(generic_waitq, _wait_for_wakeup(cqr));
-	if (rc == -ERESTARTSYS) {
-		dasd_cancel_req(cqr);
-		/* wait (non-interruptible) for final status */
-		wait_event(generic_waitq, _wait_for_wakeup(cqr));
-		cqr->intrc = rc;
-	}
-
-	if (cqr->status == DASD_CQR_DONE)
-		rc = 0;
-	else if (cqr->intrc)
-		rc = cqr->intrc;
-	else
-		rc = -EIO;
-	return rc;
+	return _dasd_sleep_on(cqr, 1);
 }
 
 /*
@@ -1629,7 +1721,7 @@ static void dasd_block_timeout(unsigned long ptr)
 	block = (struct dasd_block *) ptr;
 	spin_lock_irqsave(get_ccwdev_lock(block->base->cdev), flags);
 	/* re-activate request queue */
-	block->base->stopped &= ~DASD_STOPPED_PENDING;
+	dasd_device_remove_stop_bits(block->base, DASD_STOPPED_PENDING);
 	spin_unlock_irqrestore(get_ccwdev_lock(block->base->cdev), flags);
 	dasd_schedule_block_bh(block);
 }
@@ -1656,11 +1748,10 @@ void dasd_block_clear_timer(struct dasd_block *block)
 /*
  * Process finished error recovery ccw.
  */
-static inline void __dasd_block_process_erp(struct dasd_block *block,
-					    struct dasd_ccw_req *cqr)
+static void __dasd_process_erp(struct dasd_device *device,
+			       struct dasd_ccw_req *cqr)
 {
 	dasd_erp_fn_t erp_fn;
-	struct dasd_device *device = block->base;
 
 	if (cqr->status == DASD_CQR_DONE)
 		DBF_DEV_EVENT(DBF_NOTICE, device, "%s", "ERP successful");
@@ -1724,9 +1815,12 @@ static void __dasd_process_request_queue(struct dasd_block *block)
 				 */
 				if (!list_empty(&block->ccw_queue))
 					break;
-				spin_lock_irqsave(get_ccwdev_lock(basedev->cdev), flags);
-				basedev->stopped |= DASD_STOPPED_PENDING;
-				spin_unlock_irqrestore(get_ccwdev_lock(basedev->cdev), flags);
+				spin_lock_irqsave(
+					get_ccwdev_lock(basedev->cdev), flags);
+				dasd_device_set_stop_bits(basedev,
+							  DASD_STOPPED_PENDING);
+				spin_unlock_irqrestore(
+					get_ccwdev_lock(basedev->cdev), flags);
 				dasd_block_set_timer(block, HZ/2);
 				break;
 			}
@@ -1812,7 +1906,7 @@ restart:
 			cqr->status = DASD_CQR_FILLED;
 			cqr->retries = 255;
 			spin_lock_irqsave(get_ccwdev_lock(base->cdev), flags);
-			base->stopped |= DASD_STOPPED_QUIESCE;
+			dasd_device_set_stop_bits(base, DASD_STOPPED_QUIESCE);
 			spin_unlock_irqrestore(get_ccwdev_lock(base->cdev),
 					       flags);
 			goto restart;
@@ -1820,7 +1914,7 @@ restart:
 
 		/* Process finished ERP request. */
 		if (cqr->refers) {
-			__dasd_block_process_erp(block, cqr);
+			__dasd_process_erp(base, cqr);
 			goto restart;
 		}
 
@@ -1951,7 +2045,7 @@ restart_cb:
 		/* Process finished ERP request. */
 		if (cqr->refers) {
 			spin_lock_bh(&block->queue_lock);
-			__dasd_block_process_erp(block, cqr);
+			__dasd_process_erp(block->base, cqr);
 			spin_unlock_bh(&block->queue_lock);
 			/* restart list_for_xx loop since dasd_process_erp
 			 * might remove multiple elements */
@@ -2410,16 +2504,16 @@ int dasd_generic_notify(struct ccw_device *cdev, int event)
 				cqr->status = DASD_CQR_QUEUED;
 				cqr->retries++;
 			}
-		device->stopped |= DASD_STOPPED_DC_WAIT;
+		dasd_device_set_stop_bits(device, DASD_STOPPED_DC_WAIT);
 		dasd_device_clear_timer(device);
 		dasd_schedule_device_bh(device);
 		ret = 1;
 		break;
 	case CIO_OPER:
 		/* FIXME: add a sanity check. */
-		device->stopped &= ~DASD_STOPPED_DC_WAIT;
+		dasd_device_remove_stop_bits(device, DASD_STOPPED_DC_WAIT);
 		if (device->stopped & DASD_UNRESUMED_PM) {
-			device->stopped &= ~DASD_UNRESUMED_PM;
+			dasd_device_remove_stop_bits(device, DASD_UNRESUMED_PM);
 			dasd_restore_device(device);
 			ret = 1;
 			break;
@@ -2444,7 +2538,7 @@ int dasd_generic_pm_freeze(struct ccw_device *cdev)
 	if (IS_ERR(device))
 		return PTR_ERR(device);
 	/* disallow new I/O  */
-	device->stopped |= DASD_STOPPED_PM;
+	dasd_device_set_stop_bits(device, DASD_STOPPED_PM);
 	/* clear active requests */
 	INIT_LIST_HEAD(&freeze_queue);
 	spin_lock_irq(get_ccwdev_lock(cdev));
@@ -2496,14 +2590,18 @@ int dasd_generic_restore_device(struct ccw_device *cdev)
 		return PTR_ERR(device);
 
 	/* allow new IO again */
-	device->stopped &= ~DASD_STOPPED_PM;
-	device->stopped &= ~DASD_UNRESUMED_PM;
+	dasd_device_remove_stop_bits(device,
+				     (DASD_STOPPED_PM | DASD_UNRESUMED_PM));
 
 	dasd_schedule_device_bh(device);
 
-	if (device->discipline->restore)
+	/*
+	 * call discipline restore function
+	 * if device is stopped do nothing e.g. for disconnected devices
+	 */
+	if (device->discipline->restore && !(device->stopped))
 		rc = device->discipline->restore(device);
-	if (rc)
+	if (rc || device->stopped)
 		/*
 		 * if the resume failed for the DASD we put it in
 		 * an UNRESUMED stop state
@@ -2553,8 +2651,7 @@ static struct dasd_ccw_req *dasd_generic_build_rdc(struct dasd_device *device,
 	cqr->startdev = device;
 	cqr->memdev = device;
 	cqr->expires = 10*HZ;
-	clear_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags);
-	cqr->retries = 2;
+	cqr->retries = 256;
 	cqr->buildclk = get_clock();
 	cqr->status = DASD_CQR_FILLED;
 	return cqr;