dasd: fix error recovery for alias devices during format

Kernel panic or a hanging device during format if an alias device is
set offline or I/O errors occur.

Omit the error recovery procedure for alias devices and do retries on
the base device with full erp.
Signed-off-by: Stefan Haberland <stefan.haberland@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>

drivers/s390/block/dasd.c

@@ -2307,17 +2307,27 @@ static int _dasd_sleep_on_queue(struct list_head *ccw_queue, int interruptible)
 
 	rc = 0;
 	list_for_each_entry_safe(cqr, n, ccw_queue, blocklist) {
-		if (__dasd_sleep_on_erp(cqr)) {
-			if (!cqr->status == DASD_CQR_TERMINATED &&
-			    !cqr->status == DASD_CQR_NEED_ERP)
-				break;
-			rc = 1;
+		/*
+		 * for alias devices simplify error recovery and
+		 * return to upper layer
+		 */
+		if (cqr->startdev != cqr->basedev &&
+		    (cqr->status == DASD_CQR_TERMINATED ||
+		     cqr->status == DASD_CQR_NEED_ERP))
+			return -EAGAIN;
+		else {
+			/* normal recovery for basedev IO */
+			if (__dasd_sleep_on_erp(cqr)) {
+				if (!cqr->status == DASD_CQR_TERMINATED &&
+				    !cqr->status == DASD_CQR_NEED_ERP)
+					break;
+				rc = 1;
+			}
 		}
 	}
 	if (rc)
 		goto retry;
 
-
 	return 0;
 }
 

drivers/s390/block/dasd_eckd.c

@@ -2061,11 +2061,12 @@ dasd_eckd_fill_geometry(struct dasd_block *block, struct hd_geometry *geo)
 
 static struct dasd_ccw_req *
 dasd_eckd_build_format(struct dasd_device *base,
-		       struct format_data_t *fdata)
+		       struct format_data_t *fdata,
+		       int enable_PAV)
 {
 	struct dasd_eckd_private *base_priv;
 	struct dasd_eckd_private *start_priv;
-	struct dasd_device *startdev;
+	struct dasd_device *startdev = NULL;
 	struct dasd_ccw_req *fcp;
 	struct eckd_count *ect;
 	struct ch_t address;
@@ -2079,7 +2080,9 @@ dasd_eckd_build_format(struct dasd_device *base,
 	int nr_tracks;
 	int use_prefix;
 
-	startdev = dasd_alias_get_start_dev(base);
+	if (enable_PAV)
+		startdev = dasd_alias_get_start_dev(base);
+
 	if (!startdev)
 		startdev = base;
 
@@ -2309,6 +2312,7 @@ dasd_eckd_build_format(struct dasd_device *base,
 
 	fcp->startdev = startdev;
 	fcp->memdev = startdev;
+	fcp->basedev = base;
 	fcp->retries = 256;
 	fcp->expires = startdev->default_expires * HZ;
 	fcp->buildclk = get_tod_clock();
@@ -2319,7 +2323,8 @@ dasd_eckd_build_format(struct dasd_device *base,
 
 static int
 dasd_eckd_format_device(struct dasd_device *base,
-			struct format_data_t *fdata)
+			struct format_data_t *fdata,
+			int enable_PAV)
 {
 	struct dasd_ccw_req *cqr, *n;
 	struct dasd_block *block;
@@ -2327,7 +2332,7 @@ dasd_eckd_format_device(struct dasd_device *base,
 	struct list_head format_queue;
 	struct dasd_device *device;
 	int old_stop, format_step;
-	int step, rc = 0;
+	int step, rc = 0, sleep_rc;
 
 	block = base->block;
 	private = (struct dasd_eckd_private *) base->private;
@@ -2361,11 +2366,11 @@ dasd_eckd_format_device(struct dasd_device *base,
 	}
 
 	INIT_LIST_HEAD(&format_queue);
-	old_stop = fdata->stop_unit;
 
+	old_stop = fdata->stop_unit;
 	while (fdata->start_unit <= 1) {
 		fdata->stop_unit = fdata->start_unit;
-		cqr = dasd_eckd_build_format(base, fdata);
+		cqr = dasd_eckd_build_format(base, fdata, enable_PAV);
 		list_add(&cqr->blocklist, &format_queue);
 
 		fdata->stop_unit = old_stop;
@@ -2383,7 +2388,7 @@ dasd_eckd_format_device(struct dasd_device *base,
 		if (step > format_step)
 			fdata->stop_unit = fdata->start_unit + format_step - 1;
 
-		cqr = dasd_eckd_build_format(base, fdata);
+		cqr = dasd_eckd_build_format(base, fdata, enable_PAV);
 		if (IS_ERR(cqr)) {
 			if (PTR_ERR(cqr) == -ENOMEM) {
 				/*
@@ -2403,7 +2408,7 @@ dasd_eckd_format_device(struct dasd_device *base,
 	}
 
 sleep:
-	dasd_sleep_on_queue(&format_queue);
+	sleep_rc = dasd_sleep_on_queue(&format_queue);
 
 	list_for_each_entry_safe(cqr, n, &format_queue, blocklist) {
 		device = cqr->startdev;
@@ -2415,6 +2420,9 @@ dasd_eckd_format_device(struct dasd_device *base,
 		private->count--;
 	}
 
+	if (sleep_rc)
+		return sleep_rc;
+
 	/*
 	 * in case of ENOMEM we need to retry after
 	 * first requests are finished

drivers/s390/block/dasd_int.h

@@ -175,6 +175,7 @@ struct dasd_ccw_req {
 	struct dasd_block *block;	/* the originating block device */
 	struct dasd_device *memdev;	/* the device used to allocate this */
 	struct dasd_device *startdev;	/* device the request is started on */
+	struct dasd_device *basedev;	/* base device if no block->base */
 	void *cpaddr;			/* address of ccw or tcw */
 	unsigned char cpmode;		/* 0 = cmd mode, 1 = itcw */
 	char status;			/* status of this request */
@@ -321,7 +322,7 @@ struct dasd_discipline {
 	int (*term_IO) (struct dasd_ccw_req *);
 	void (*handle_terminated_request) (struct dasd_ccw_req *);
 	int (*format_device) (struct dasd_device *,
-			      struct format_data_t *);
+			      struct format_data_t *, int enable_PAV);
 	int (*free_cp) (struct dasd_ccw_req *, struct request *);
 
 	/*

drivers/s390/block/dasd_ioctl.c

@@ -203,7 +203,9 @@ static int
 dasd_format(struct dasd_block *block, struct format_data_t *fdata)
 {
 	struct dasd_device *base;
-	int rc;
+	int enable_PAV = 1;
+	int rc, retries;
+	int start, stop;
 
 	base = block->base;
 	if (base->discipline->format_device == NULL)
@@ -231,11 +233,30 @@ dasd_format(struct dasd_block *block, struct format_data_t *fdata)
 		bdput(bdev);
 	}
 
-	rc = base->discipline->format_device(base, fdata);
-	if (rc)
-		return rc;
-
-	return 0;
+	retries = 255;
+	/* backup start- and endtrack for retries */
+	start = fdata->start_unit;
+	stop = fdata->stop_unit;
+	do {
+		rc = base->discipline->format_device(base, fdata, enable_PAV);
+		if (rc) {
+			if (rc == -EAGAIN) {
+				retries--;
+				/* disable PAV in case of errors */
+				enable_PAV = 0;
+				fdata->start_unit = start;
+				fdata->stop_unit = stop;
+			} else
+				return rc;
+		} else
+			/* success */
+			break;
+	} while (retries);
+
+	if (!retries)
+		return -EIO;
+	else
+		return 0;
 }
 
 /*