[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>

drivers/s390/block/dasd_3990_erp.c

@@ -69,8 +69,7 @@ dasd_3990_erp_cleanup(struct dasd_ccw_req * erp, char final_status)
  *   processing until the started timer has expired or an related
  *   interrupt was received.
  */
-static void
-dasd_3990_erp_block_queue(struct dasd_ccw_req * erp, int expires)
+static void dasd_3990_erp_block_queue(struct dasd_ccw_req *erp, int expires)
 {
 
 	struct dasd_device *device = erp->startdev;
@@ -80,10 +79,13 @@ dasd_3990_erp_block_queue(struct dasd_ccw_req * erp, int expires)
 		    "blocking request queue for %is", expires/HZ);
 
 	spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
-	device->stopped |= DASD_STOPPED_PENDING;
+	dasd_device_set_stop_bits(device, DASD_STOPPED_PENDING);
 	spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
 	erp->status = DASD_CQR_FILLED;
-	dasd_block_set_timer(device->block, expires);
+	if (erp->block)
+		dasd_block_set_timer(erp->block, expires);
+	else
+		dasd_device_set_timer(device, expires);
 }
 
 /*
@@ -242,9 +244,13 @@ dasd_3990_erp_DCTL(struct dasd_ccw_req * erp, char modifier)
  * DESCRIPTION
  *   Setup ERP to do the ERP action 1 (see Reference manual).
  *   Repeat the operation on a different channel path.
- *   If all alternate paths have been tried, the request is posted with a
- *   permanent error.
- *   Note: duplex handling is not implemented (yet).
+ *   As deviation from the recommended recovery action, we reset the path mask
+ *   after we have tried each path and go through all paths a second time.
+ *   This will cover situations where only one path at a time is actually down,
+ *   but all paths fail and recover just with the same sequence and timing as
+ *   we try to use them (flapping links).
+ *   If all alternate paths have been tried twice, the request is posted with
+ *   a permanent error.
  *
  *  PARAMETER
  *   erp		pointer to the current ERP
@@ -253,17 +259,25 @@ dasd_3990_erp_DCTL(struct dasd_ccw_req * erp, char modifier)
  *   erp		pointer to the ERP
  *
  */
-static struct dasd_ccw_req *
-dasd_3990_erp_action_1(struct dasd_ccw_req * erp)
+static struct dasd_ccw_req *dasd_3990_erp_action_1_sec(struct dasd_ccw_req *erp)
 {
+	erp->function = dasd_3990_erp_action_1_sec;
+	dasd_3990_erp_alternate_path(erp);
+	return erp;
+}
 
+static struct dasd_ccw_req *dasd_3990_erp_action_1(struct dasd_ccw_req *erp)
+{
 	erp->function = dasd_3990_erp_action_1;
-
 	dasd_3990_erp_alternate_path(erp);
-
+	if (erp->status == DASD_CQR_FAILED) {
+		erp->status = DASD_CQR_FILLED;
+		erp->retries = 10;
+		erp->lpm = LPM_ANYPATH;
+		erp->function = dasd_3990_erp_action_1_sec;
+	}
 	return erp;
-
-}				/* end dasd_3990_erp_action_1 */
+}				/* end dasd_3990_erp_action_1(b) */
 
 /*
  * DASD_3990_ERP_ACTION_4
@@ -2294,6 +2308,7 @@ static struct dasd_ccw_req *dasd_3990_erp_add_erp(struct dasd_ccw_req *cqr)
 		return cqr;
 	}
 
+	ccw = cqr->cpaddr;
 	if (cqr->cpmode == 1) {
 		/* make a shallow copy of the original tcw but set new tsb */
 		erp->cpmode = 1;
@@ -2302,6 +2317,9 @@ static struct dasd_ccw_req *dasd_3990_erp_add_erp(struct dasd_ccw_req *cqr)
 		tsb = (struct tsb *) &tcw[1];
 		*tcw = *((struct tcw *)cqr->cpaddr);
 		tcw->tsb = (long)tsb;
+	} else if (ccw->cmd_code == DASD_ECKD_CCW_PSF) {
+		/* PSF cannot be chained from NOOP/TIC */
+		erp->cpaddr = cqr->cpaddr;
 	} else {
 		/* initialize request with default TIC to current ERP/CQR */
 		ccw = erp->cpaddr;
@@ -2486,6 +2504,8 @@ dasd_3990_erp_further_erp(struct dasd_ccw_req *erp)
 
 		erp = dasd_3990_erp_action_1(erp);
 
+	} else if (erp->function == dasd_3990_erp_action_1_sec) {
+		erp = dasd_3990_erp_action_1_sec(erp);
 	} else if (erp->function == dasd_3990_erp_action_5) {
 
 		/* retries have not been successful */

drivers/s390/block/dasd_alias.c

@@ -755,11 +755,11 @@ static void __stop_device_on_lcu(struct dasd_device *device,
 {
 	/* If pos == device then device is already locked! */
 	if (pos == device) {
-		pos->stopped |= DASD_STOPPED_SU;
+		dasd_device_set_stop_bits(pos, DASD_STOPPED_SU);
 		return;
 	}
 	spin_lock(get_ccwdev_lock(pos->cdev));
-	pos->stopped |= DASD_STOPPED_SU;
+	dasd_device_set_stop_bits(pos, DASD_STOPPED_SU);
 	spin_unlock(get_ccwdev_lock(pos->cdev));
 }
 
@@ -793,26 +793,26 @@ static void _unstop_all_devices_on_lcu(struct alias_lcu *lcu)
 
 	list_for_each_entry(device, &lcu->active_devices, alias_list) {
 		spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
-		device->stopped &= ~DASD_STOPPED_SU;
+		dasd_device_remove_stop_bits(device, DASD_STOPPED_SU);
 		spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
 	}
 
 	list_for_each_entry(device, &lcu->inactive_devices, alias_list) {
 		spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
-		device->stopped &= ~DASD_STOPPED_SU;
+		dasd_device_remove_stop_bits(device, DASD_STOPPED_SU);
 		spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
 	}
 
 	list_for_each_entry(pavgroup, &lcu->grouplist, group) {
 		list_for_each_entry(device, &pavgroup->baselist, alias_list) {
 			spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
-			device->stopped &= ~DASD_STOPPED_SU;
+			dasd_device_remove_stop_bits(device, DASD_STOPPED_SU);
 			spin_unlock_irqrestore(get_ccwdev_lock(device->cdev),
 					       flags);
 		}
 		list_for_each_entry(device, &pavgroup->aliaslist, alias_list) {
 			spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
-			device->stopped &= ~DASD_STOPPED_SU;
+			dasd_device_remove_stop_bits(device, DASD_STOPPED_SU);
 			spin_unlock_irqrestore(get_ccwdev_lock(device->cdev),
 					       flags);
 		}
@@ -836,7 +836,8 @@ static void summary_unit_check_handling_work(struct work_struct *work)
 
 	/* 2. reset summary unit check */
 	spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
-	device->stopped &= ~(DASD_STOPPED_SU | DASD_STOPPED_PENDING);
+	dasd_device_remove_stop_bits(device,
+				     (DASD_STOPPED_SU | DASD_STOPPED_PENDING));
 	spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
 	reset_summary_unit_check(lcu, device, suc_data->reason);
 

drivers/s390/block/dasd_eckd.c

@@ -77,6 +77,11 @@ MODULE_DEVICE_TABLE(ccw, dasd_eckd_ids);
 
 static struct ccw_driver dasd_eckd_driver; /* see below */
 
+#define INIT_CQR_OK 0
+#define INIT_CQR_UNFORMATTED 1
+#define INIT_CQR_ERROR 2
+
+
 /* initial attempt at a probe function. this can be simplified once
  * the other detection code is gone */
 static int
@@ -749,8 +754,7 @@ static struct dasd_ccw_req *dasd_eckd_build_rcd_lpm(struct dasd_device *device,
 	cqr->block = NULL;
 	cqr->expires = 10*HZ;
 	cqr->lpm = lpm;
-	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;
@@ -949,8 +953,7 @@ static int dasd_eckd_read_features(struct dasd_device *device)
 	cqr->startdev = device;
 	cqr->memdev = device;
 	cqr->block = NULL;
-	clear_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags);
-	cqr->retries = 5;
+	cqr->retries = 256;
 	cqr->expires = 10 * HZ;
 
 	/* Prepare for Read Subsystem Data */
@@ -1025,6 +1028,7 @@ static struct dasd_ccw_req *dasd_eckd_build_psf_ssc(struct dasd_device *device,
 	cqr->startdev = device;
 	cqr->memdev = device;
 	cqr->block = NULL;
+	cqr->retries = 256;
 	cqr->expires = 10*HZ;
 	cqr->buildclk = get_clock();
 	cqr->status = DASD_CQR_FILLED;
@@ -1068,6 +1072,7 @@ static int dasd_eckd_validate_server(struct dasd_device *device)
 	else
 		enable_pav = 1;
 	rc = dasd_eckd_psf_ssc(device, enable_pav);
+
 	/* may be requested feature is not available on server,
 	 * therefore just report error and go ahead */
 	private = (struct dasd_eckd_private *) device->private;
@@ -1265,12 +1270,29 @@ dasd_eckd_analysis_ccw(struct dasd_device *device)
 	cqr->block = NULL;
 	cqr->startdev = device;
 	cqr->memdev = device;
-	cqr->retries = 0;
+	cqr->retries = 255;
 	cqr->buildclk = get_clock();
 	cqr->status = DASD_CQR_FILLED;
 	return cqr;
 }
 
+/* differentiate between 'no record found' and any other error */
+static int dasd_eckd_analysis_evaluation(struct dasd_ccw_req *init_cqr)
+{
+	char *sense;
+	if (init_cqr->status == DASD_CQR_DONE)
+		return INIT_CQR_OK;
+	else if (init_cqr->status == DASD_CQR_NEED_ERP ||
+		 init_cqr->status == DASD_CQR_FAILED) {
+		sense = dasd_get_sense(&init_cqr->irb);
+		if (sense && (sense[1] & SNS1_NO_REC_FOUND))
+			return INIT_CQR_UNFORMATTED;
+		else
+			return INIT_CQR_ERROR;
+	} else
+		return INIT_CQR_ERROR;
+}
+
 /*
  * This is the callback function for the init_analysis cqr. It saves
  * the status of the initial analysis ccw before it frees it and kicks
@@ -1278,21 +1300,20 @@ dasd_eckd_analysis_ccw(struct dasd_device *device)
  * dasd_eckd_do_analysis again (if the devices has not been marked
  * for deletion in the meantime).
  */
-static void
-dasd_eckd_analysis_callback(struct dasd_ccw_req *init_cqr, void *data)
+static void dasd_eckd_analysis_callback(struct dasd_ccw_req *init_cqr,
+					void *data)
 {
 	struct dasd_eckd_private *private;
 	struct dasd_device *device;
 
 	device = init_cqr->startdev;
 	private = (struct dasd_eckd_private *) device->private;
-	private->init_cqr_status = init_cqr->status;
+	private->init_cqr_status = dasd_eckd_analysis_evaluation(init_cqr);
 	dasd_sfree_request(init_cqr, device);
 	dasd_kick_device(device);
 }
 
-static int
-dasd_eckd_start_analysis(struct dasd_block *block)
+static int dasd_eckd_start_analysis(struct dasd_block *block)
 {
 	struct dasd_eckd_private *private;
 	struct dasd_ccw_req *init_cqr;
@@ -1304,27 +1325,44 @@ dasd_eckd_start_analysis(struct dasd_block *block)
 	init_cqr->callback = dasd_eckd_analysis_callback;
 	init_cqr->callback_data = NULL;
 	init_cqr->expires = 5*HZ;
+	/* first try without ERP, so we can later handle unformatted
+	 * devices as special case
+	 */
+	clear_bit(DASD_CQR_FLAGS_USE_ERP, &init_cqr->flags);
+	init_cqr->retries = 0;
 	dasd_add_request_head(init_cqr);
 	return -EAGAIN;
 }
 
-static int
-dasd_eckd_end_analysis(struct dasd_block *block)
+static int dasd_eckd_end_analysis(struct dasd_block *block)
 {
 	struct dasd_device *device;
 	struct dasd_eckd_private *private;
 	struct eckd_count *count_area;
 	unsigned int sb, blk_per_trk;
 	int status, i;
+	struct dasd_ccw_req *init_cqr;
 
 	device = block->base;
 	private = (struct dasd_eckd_private *) device->private;
 	status = private->init_cqr_status;
 	private->init_cqr_status = -1;
-	if (status != DASD_CQR_DONE) {
-		dev_warn(&device->cdev->dev,
-			    "The DASD is not formatted\n");
+	if (status == INIT_CQR_ERROR) {
+		/* try again, this time with full ERP */
+		init_cqr = dasd_eckd_analysis_ccw(device);
+		dasd_sleep_on(init_cqr);
+		status = dasd_eckd_analysis_evaluation(init_cqr);
+		dasd_sfree_request(init_cqr, device);
+	}
+
+	if (status == INIT_CQR_UNFORMATTED) {
+		dev_warn(&device->cdev->dev, "The DASD is not formatted\n");
 		return -EMEDIUMTYPE;
+	} else if (status == INIT_CQR_ERROR) {
+		dev_err(&device->cdev->dev,
+			"Detecting the DASD disk layout failed because "
+			"of an I/O error\n");
+		return -EIO;
 	}
 
 	private->uses_cdl = 1;
@@ -1616,8 +1654,7 @@ dasd_eckd_format_device(struct dasd_device * device,
 	}
 	fcp->startdev = device;
 	fcp->memdev = device;
-	clear_bit(DASD_CQR_FLAGS_USE_ERP, &fcp->flags);
-	fcp->retries = 5;	/* set retry counter to enable default ERP */
+	fcp->retries = 256;
 	fcp->buildclk = get_clock();
 	fcp->status = DASD_CQR_FILLED;
 	return fcp;
@@ -2699,6 +2736,7 @@ dasd_eckd_performance(struct dasd_device *device, void __user *argp)
 	cqr->startdev = device;
 	cqr->memdev = device;
 	cqr->retries = 0;
+	clear_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags);
 	cqr->expires = 10 * HZ;
 
 	/* Prepare for Read Subsystem Data */

drivers/s390/block/dasd_int.h

@@ -595,6 +595,9 @@ int dasd_generic_restore_device(struct ccw_device *);
 int dasd_generic_read_dev_chars(struct dasd_device *, int, void *, int);
 char *dasd_get_sense(struct irb *);
 
+void dasd_device_set_stop_bits(struct dasd_device *, int);
+void dasd_device_remove_stop_bits(struct dasd_device *, int);
+
 /* externals in dasd_devmap.c */
 extern int dasd_max_devindex;
 extern int dasd_probeonly;

drivers/s390/block/dasd_ioctl.c

@@ -101,7 +101,7 @@ static int dasd_ioctl_quiesce(struct dasd_block *block)
 	pr_info("%s: The DASD has been put in the quiesce "
 		"state\n", dev_name(&base->cdev->dev));
 	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);
 	return 0;
 }
@@ -122,7 +122,7 @@ static int dasd_ioctl_resume(struct dasd_block *block)
 	pr_info("%s: I/O operations have been resumed "
 		"on the DASD\n", dev_name(&base->cdev->dev));
 	spin_lock_irqsave(get_ccwdev_lock(base->cdev), flags);
-	base->stopped &= ~DASD_STOPPED_QUIESCE;
+	dasd_device_remove_stop_bits(base, DASD_STOPPED_QUIESCE);
 	spin_unlock_irqrestore(get_ccwdev_lock(base->cdev), flags);
 
 	dasd_schedule_block_bh(block);