Creation of Cybook 2416 (actually Gen4) repository

drivers/s390/block/Kconfig

@@ -0,0 +1,66 @@
+if S390 && BLOCK
+
+comment "S/390 block device drivers"
+	depends on S390
+
+config BLK_DEV_XPRAM
+	tristate "XPRAM disk support"
+	depends on S390
+	help
+	  Select this option if you want to use your expanded storage on S/390
+	  or zSeries as a disk.  This is useful as a _fast_ swap device if you
+	  want to access more than 2G of memory when running in 31 bit mode.
+	  This option is also available as a module which will be called
+	  xpram.  If unsure, say "N".
+
+config DCSSBLK
+	tristate "DCSSBLK support"
+	help
+	  Support for dcss block device
+
+config DASD
+	tristate "Support for DASD devices"
+	depends on CCW
+	help
+	  Enable this option if you want to access DASDs directly utilizing
+	  S/390s channel subsystem commands. This is necessary for running
+	  natively on a single image or an LPAR.
+
+config DASD_PROFILE
+	bool "Profiling support for dasd devices"
+	depends on DASD
+	help
+	  Enable this option if you want to see profiling information
+          in /proc/dasd/statistics.
+
+config DASD_ECKD
+	tristate "Support for ECKD Disks"
+	depends on DASD
+	help
+	  ECKD devices are the most commonly used devices. You should enable
+	  this option unless you are very sure to have no ECKD device.
+
+config DASD_FBA
+	tristate "Support for FBA  Disks"
+	depends on DASD
+	help
+	  Select this option to be able to access FBA devices. It is safe to
+	  say "Y".
+
+config DASD_DIAG
+	tristate "Support for DIAG access to Disks"
+	depends on DASD
+	help
+	  Select this option if you want to use Diagnose250 command to access
+	  Disks under VM.  If you are not running under VM or unsure what it is,
+	  say "N".
+
+config DASD_EER
+	bool "Extended error reporting (EER)"
+	depends on DASD
+	help
+	  This driver provides a character device interface to the
+	  DASD extended error reporting. This is only needed if you want to
+	  use applications written for the EER facility.
+
+endif

drivers/s390/block/Makefile

@@ -0,0 +1,19 @@
+#
+# S/390 block devices
+#
+
+dasd_eckd_mod-objs := dasd_eckd.o dasd_3990_erp.o dasd_9343_erp.o
+dasd_fba_mod-objs  := dasd_fba.o dasd_3370_erp.o dasd_9336_erp.o
+dasd_diag_mod-objs := dasd_diag.o
+dasd_mod-objs      := dasd.o dasd_ioctl.o dasd_proc.o dasd_devmap.o \
+			dasd_genhd.o dasd_erp.o
+ifdef CONFIG_DASD_EER
+dasd_mod-objs      += dasd_eer.o
+endif
+
+obj-$(CONFIG_DASD) += dasd_mod.o
+obj-$(CONFIG_DASD_DIAG) += dasd_diag_mod.o
+obj-$(CONFIG_DASD_ECKD) += dasd_eckd_mod.o
+obj-$(CONFIG_DASD_FBA)  += dasd_fba_mod.o
+obj-$(CONFIG_BLK_DEV_XPRAM) += xpram.o
+obj-$(CONFIG_DCSSBLK) += dcssblk.o

drivers/s390/block/dasd_3370_erp.c

@@ -0,0 +1,84 @@
+/*
+ * File...........: linux/drivers/s390/block/dasd_3370_erp.c
+ * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com>
+ * Bugreports.to..: <Linux390@de.ibm.com>
+ * (C) IBM Corporation, IBM Deutschland Entwicklung GmbH, 2000
+ *
+ */
+
+#define PRINTK_HEADER "dasd_erp(3370)"
+
+#include "dasd_int.h"
+
+
+/*
+ * DASD_3370_ERP_EXAMINE
+ *
+ * DESCRIPTION
+ *   Checks only for fatal/no/recover error.
+ *   A detailed examination of the sense data is done later outside
+ *   the interrupt handler.
+ *
+ *   The logic is based on the 'IBM 3880 Storage Control Reference' manual
+ *   'Chapter 7. 3370 Sense Data'.
+ *
+ * RETURN VALUES
+ *   dasd_era_none	no error
+ *   dasd_era_fatal	for all fatal (unrecoverable errors)
+ *   dasd_era_recover	for all others.
+ */
+dasd_era_t
+dasd_3370_erp_examine(struct dasd_ccw_req * cqr, struct irb * irb)
+{
+	char *sense = irb->ecw;
+
+	/* check for successful execution first */
+	if (irb->scsw.cstat == 0x00 &&
+	    irb->scsw.dstat == (DEV_STAT_CHN_END | DEV_STAT_DEV_END))
+		return dasd_era_none;
+	if (sense[0] & 0x80) {	/* CMD reject */
+		return dasd_era_fatal;
+	}
+	if (sense[0] & 0x40) {	/* Drive offline */
+		return dasd_era_recover;
+	}
+	if (sense[0] & 0x20) {	/* Bus out parity */
+		return dasd_era_recover;
+	}
+	if (sense[0] & 0x10) {	/* equipment check */
+		if (sense[1] & 0x80) {
+			return dasd_era_fatal;
+		}
+		return dasd_era_recover;
+	}
+	if (sense[0] & 0x08) {	/* data check */
+		if (sense[1] & 0x80) {
+			return dasd_era_fatal;
+		}
+		return dasd_era_recover;
+	}
+	if (sense[0] & 0x04) {	/* overrun */
+		if (sense[1] & 0x80) {
+			return dasd_era_fatal;
+		}
+		return dasd_era_recover;
+	}
+	if (sense[1] & 0x40) {	/* invalid blocksize */
+		return dasd_era_fatal;
+	}
+	if (sense[1] & 0x04) {	/* file protected */
+		return dasd_era_recover;
+	}
+	if (sense[1] & 0x01) {	/* operation incomplete */
+		return dasd_era_recover;
+	}
+	if (sense[2] & 0x80) {	/* check data erroor */
+		return dasd_era_recover;
+	}
+	if (sense[2] & 0x10) {	/* Env. data present */
+		return dasd_era_recover;
+	}
+	/* examine the 24 byte sense data */
+	return dasd_era_recover;
+
+}				/* END dasd_3370_erp_examine */

drivers/s390/block/dasd_9336_erp.c

@@ -0,0 +1,41 @@
+/*
+ * File...........: linux/drivers/s390/block/dasd_9336_erp.c
+ * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com>
+ * Bugreports.to..: <Linux390@de.ibm.com>
+ * (C) IBM Corporation, IBM Deutschland Entwicklung GmbH, 2000
+ *
+ */
+
+#define PRINTK_HEADER "dasd_erp(9336)"
+
+#include "dasd_int.h"
+
+
+/*
+ * DASD_9336_ERP_EXAMINE
+ *
+ * DESCRIPTION
+ *   Checks only for fatal/no/recover error.
+ *   A detailed examination of the sense data is done later outside
+ *   the interrupt handler.
+ *
+ *   The logic is based on the 'IBM 3880 Storage Control Reference' manual
+ *   'Chapter 7. 9336 Sense Data'.
+ *
+ * RETURN VALUES
+ *   dasd_era_none	no error
+ *   dasd_era_fatal	for all fatal (unrecoverable errors)
+ *   dasd_era_recover	for all others.
+ */
+dasd_era_t
+dasd_9336_erp_examine(struct dasd_ccw_req * cqr, struct irb * irb)
+{
+	/* check for successful execution first */
+	if (irb->scsw.cstat == 0x00 &&
+	    irb->scsw.dstat == (DEV_STAT_CHN_END | DEV_STAT_DEV_END))
+		return dasd_era_none;
+
+	/* examine the 24 byte sense data */
+	return dasd_era_recover;
+
+}				/* END dasd_9336_erp_examine */

drivers/s390/block/dasd_9343_erp.c

@@ -0,0 +1,21 @@
+/*
+ * File...........: linux/drivers/s390/block/dasd_9345_erp.c
+ * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com>
+ * Bugreports.to..: <Linux390@de.ibm.com>
+ * (C) IBM Corporation, IBM Deutschland Entwicklung GmbH, 2000
+ *
+ */
+
+#define PRINTK_HEADER "dasd_erp(9343)"
+
+#include "dasd_int.h"
+
+dasd_era_t
+dasd_9343_erp_examine(struct dasd_ccw_req * cqr, struct irb * irb)
+{
+	if (irb->scsw.cstat == 0x00 &&
+	    irb->scsw.dstat == (DEV_STAT_CHN_END | DEV_STAT_DEV_END))
+		return dasd_era_none;
+
+	return dasd_era_recover;
+}

drivers/s390/block/dasd_diag.c

@@ -0,0 +1,624 @@
+/*
+ * File...........: linux/drivers/s390/block/dasd_diag.c
+ * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com>
+ * Based on.......: linux/drivers/s390/block/mdisk.c
+ * ...............: by Hartmunt Penner <hpenner@de.ibm.com>
+ * Bugreports.to..: <Linux390@de.ibm.com>
+ * (C) IBM Corporation, IBM Deutschland Entwicklung GmbH, 1999,2000
+ *
+ */
+
+#include <linux/stddef.h>
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/hdreg.h>
+#include <linux/bio.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/jiffies.h>
+
+#include <asm/dasd.h>
+#include <asm/debug.h>
+#include <asm/ebcdic.h>
+#include <asm/io.h>
+#include <asm/s390_ext.h>
+#include <asm/todclk.h>
+#include <asm/vtoc.h>
+
+#include "dasd_int.h"
+#include "dasd_diag.h"
+
+#define PRINTK_HEADER "dasd(diag):"
+
+MODULE_LICENSE("GPL");
+
+/* The maximum number of blocks per request (max_blocks) is dependent on the
+ * amount of storage that is available in the static I/O buffer for each
+ * device. Currently each device gets 2 pages. We want to fit two requests
+ * into the available memory so that we can immediately start the next if one
+ * finishes. */
+#define DIAG_MAX_BLOCKS	(((2 * PAGE_SIZE - sizeof(struct dasd_ccw_req) - \
+			   sizeof(struct dasd_diag_req)) / \
+		           sizeof(struct dasd_diag_bio)) / 2)
+#define DIAG_MAX_RETRIES	32
+#define DIAG_TIMEOUT		50 * HZ
+
+static struct dasd_discipline dasd_diag_discipline;
+
+struct dasd_diag_private {
+	struct dasd_diag_characteristics rdc_data;
+	struct dasd_diag_rw_io iob;
+	struct dasd_diag_init_io iib;
+	blocknum_t pt_block;
+};
+
+struct dasd_diag_req {
+	unsigned int block_count;
+	struct dasd_diag_bio bio[0];
+};
+
+static const u8 DASD_DIAG_CMS1[] = { 0xc3, 0xd4, 0xe2, 0xf1 };/* EBCDIC CMS1 */
+
+/* Perform DIAG250 call with block I/O parameter list iob (input and output)
+ * and function code cmd.
+ * In case of an exception return 3. Otherwise return result of bitwise OR of
+ * resulting condition code and DIAG return code. */
+static inline int dia250(void *iob, int cmd)
+{
+	register unsigned long reg2 asm ("2") = (unsigned long) iob;
+	typedef union {
+		struct dasd_diag_init_io init_io;
+		struct dasd_diag_rw_io rw_io;
+	} addr_type;
+	int rc;
+
+	rc = 3;
+	asm volatile(
+		"	diag	2,%2,0x250\n"
+		"0:	ipm	%0\n"
+		"	srl	%0,28\n"
+		"	or	%0,3\n"
+		"1:\n"
+		EX_TABLE(0b,1b)
+		: "+d" (rc), "=m" (*(addr_type *) iob)
+		: "d" (cmd), "d" (reg2), "m" (*(addr_type *) iob)
+		: "3", "cc");
+	return rc;
+}
+
+/* Initialize block I/O to DIAG device using the specified blocksize and
+ * block offset. On success, return zero and set end_block to contain the
+ * number of blocks on the device minus the specified offset. Return non-zero
+ * otherwise. */
+static inline int
+mdsk_init_io(struct dasd_device *device, unsigned int blocksize,
+	     blocknum_t offset, blocknum_t *end_block)
+{
+	struct dasd_diag_private *private;
+	struct dasd_diag_init_io *iib;
+	int rc;
+
+	private = (struct dasd_diag_private *) device->private;
+	iib = &private->iib;
+	memset(iib, 0, sizeof (struct dasd_diag_init_io));
+
+	iib->dev_nr = _ccw_device_get_device_number(device->cdev);
+	iib->block_size = blocksize;
+	iib->offset = offset;
+	iib->flaga = DASD_DIAG_FLAGA_DEFAULT;
+
+	rc = dia250(iib, INIT_BIO);
+
+	if ((rc & 3) == 0 && end_block)
+		*end_block = iib->end_block;
+
+	return rc;
+}
+
+/* Remove block I/O environment for device. Return zero on success, non-zero
+ * otherwise. */
+static inline int
+mdsk_term_io(struct dasd_device * device)
+{
+	struct dasd_diag_private *private;
+	struct dasd_diag_init_io *iib;
+	int rc;
+
+	private = (struct dasd_diag_private *) device->private;
+	iib = &private->iib;
+	memset(iib, 0, sizeof (struct dasd_diag_init_io));
+	iib->dev_nr = _ccw_device_get_device_number(device->cdev);
+	rc = dia250(iib, TERM_BIO);
+	return rc;
+}
+
+/* Error recovery for failed DIAG requests - try to reestablish the DIAG
+ * environment. */
+static void
+dasd_diag_erp(struct dasd_device *device)
+{
+	int rc;
+
+	mdsk_term_io(device);
+	rc = mdsk_init_io(device, device->bp_block, 0, NULL);
+	if (rc)
+		DEV_MESSAGE(KERN_WARNING, device, "DIAG ERP unsuccessful, "
+			    "rc=%d", rc);
+}
+
+/* Start a given request at the device. Return zero on success, non-zero
+ * otherwise. */
+static int
+dasd_start_diag(struct dasd_ccw_req * cqr)
+{
+	struct dasd_device *device;
+	struct dasd_diag_private *private;
+	struct dasd_diag_req *dreq;
+	int rc;
+
+	device = cqr->device;
+	if (cqr->retries < 0) {
+		DEV_MESSAGE(KERN_WARNING, device, "DIAG start_IO: request %p "
+			    "- no retry left)", cqr);
+		cqr->status = DASD_CQR_FAILED;
+		return -EIO;
+	}
+	private = (struct dasd_diag_private *) device->private;
+	dreq = (struct dasd_diag_req *) cqr->data;
+
+	private->iob.dev_nr = _ccw_device_get_device_number(device->cdev);
+	private->iob.key = 0;
+	private->iob.flags = DASD_DIAG_RWFLAG_ASYNC;
+	private->iob.block_count = dreq->block_count;
+	private->iob.interrupt_params = (addr_t) cqr;
+	private->iob.bio_list = dreq->bio;
+	private->iob.flaga = DASD_DIAG_FLAGA_DEFAULT;
+
+	cqr->startclk = get_clock();
+	cqr->starttime = jiffies;
+	cqr->retries--;
+
+	rc = dia250(&private->iob, RW_BIO);
+	switch (rc) {
+	case 0: /* Synchronous I/O finished successfully */
+		cqr->stopclk = get_clock();
+		cqr->status = DASD_CQR_DONE;
+		/* Indicate to calling function that only a dasd_schedule_bh()
+		   and no timer is needed */
+                rc = -EACCES;
+		break;
+	case 8: /* Asynchronous I/O was started */
+		cqr->status = DASD_CQR_IN_IO;
+		rc = 0;
+		break;
+	default: /* Error condition */
+		cqr->status = DASD_CQR_QUEUED;
+		DEV_MESSAGE(KERN_WARNING, device, "dia250 returned rc=%d", rc);
+		dasd_diag_erp(device);
+		rc = -EIO;
+		break;
+	}
+	return rc;
+}
+
+/* Terminate given request at the device. */
+static int
+dasd_diag_term_IO(struct dasd_ccw_req * cqr)
+{
+	struct dasd_device *device;
+
+	device = cqr->device;
+	mdsk_term_io(device);
+	mdsk_init_io(device, device->bp_block, 0, NULL);
+	cqr->status = DASD_CQR_CLEAR;
+	cqr->stopclk = get_clock();
+	dasd_schedule_bh(device);
+	return 0;
+}
+
+/* Handle external interruption. */
+static void
+dasd_ext_handler(__u16 code)
+{
+	struct dasd_ccw_req *cqr, *next;
+	struct dasd_device *device;
+	unsigned long long expires;
+	unsigned long flags;
+	u8 int_code, status;
+	addr_t ip;
+	int rc;
+
+	int_code = *((u8 *) DASD_DIAG_LC_INT_CODE);
+	status = *((u8 *) DASD_DIAG_LC_INT_STATUS);
+	switch (int_code) {
+	case DASD_DIAG_CODE_31BIT:
+		ip = (addr_t) *((u32 *) DASD_DIAG_LC_INT_PARM_31BIT);
+		break;
+	case DASD_DIAG_CODE_64BIT:
+		ip = (addr_t) *((u64 *) DASD_DIAG_LC_INT_PARM_64BIT);
+		break;
+	default:
+		return;
+	}
+	if (!ip) {		/* no intparm: unsolicited interrupt */
+		MESSAGE(KERN_DEBUG, "%s", "caught unsolicited interrupt");
+		return;
+	}
+	cqr = (struct dasd_ccw_req *) ip;
+	device = (struct dasd_device *) cqr->device;
+	if (strncmp(device->discipline->ebcname, (char *) &cqr->magic, 4)) {
+		DEV_MESSAGE(KERN_WARNING, device,
+			    " magic number of dasd_ccw_req 0x%08X doesn't"
+			    " match discipline 0x%08X",
+			    cqr->magic, *(int *) (&device->discipline->name));
+		return;
+	}
+
+	/* get irq lock to modify request queue */
+	spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
+
+	/* Check for a pending clear operation */
+	if (cqr->status == DASD_CQR_CLEAR) {
+		cqr->status = DASD_CQR_QUEUED;
+		dasd_clear_timer(device);
+		dasd_schedule_bh(device);
+		spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
+		return;
+	}
+
+	cqr->stopclk = get_clock();
+
+	expires = 0;
+	if (status == 0) {
+		cqr->status = DASD_CQR_DONE;
+		/* Start first request on queue if possible -> fast_io. */
+		if (!list_empty(&device->ccw_queue)) {
+			next = list_entry(device->ccw_queue.next,
+					  struct dasd_ccw_req, list);
+			if (next->status == DASD_CQR_QUEUED) {
+				rc = dasd_start_diag(next);
+				if (rc == 0)
+					expires = next->expires;
+				else if (rc != -EACCES)
+					DEV_MESSAGE(KERN_WARNING, device, "%s",
+						    "Interrupt fastpath "
+						    "failed!");
+			}
+		}
+	} else {
+		cqr->status = DASD_CQR_QUEUED;
+		DEV_MESSAGE(KERN_WARNING, device, "interrupt status for "
+			    "request %p was %d (%d retries left)", cqr, status,
+			    cqr->retries);
+		dasd_diag_erp(device);
+	}
+
+	if (expires != 0)
+		dasd_set_timer(device, expires);
+	else
+		dasd_clear_timer(device);
+	dasd_schedule_bh(device);
+
+	spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
+}
+
+/* Check whether device can be controlled by DIAG discipline. Return zero on
+ * success, non-zero otherwise. */
+static int
+dasd_diag_check_device(struct dasd_device *device)
+{
+	struct dasd_diag_private *private;
+	struct dasd_diag_characteristics *rdc_data;
+	struct dasd_diag_bio bio;
+	struct vtoc_cms_label *label;
+	blocknum_t end_block;
+	unsigned int sb, bsize;
+	int rc;
+
+	private = (struct dasd_diag_private *) device->private;
+	if (private == NULL) {
+		private = kzalloc(sizeof(struct dasd_diag_private),GFP_KERNEL);
+		if (private == NULL) {
+			DEV_MESSAGE(KERN_WARNING, device, "%s",
+				"memory allocation failed for private data");
+			return -ENOMEM;
+		}
+		device->private = (void *) private;
+	}
+	/* Read Device Characteristics */
+	rdc_data = (void *) &(private->rdc_data);
+	rdc_data->dev_nr = _ccw_device_get_device_number(device->cdev);
+	rdc_data->rdc_len = sizeof (struct dasd_diag_characteristics);
+
+	rc = diag210((struct diag210 *) rdc_data);
+	if (rc) {
+		DEV_MESSAGE(KERN_WARNING, device, "failed to retrieve device "
+			    "information (rc=%d)", rc);
+		return -ENOTSUPP;
+	}
+
+	/* Figure out position of label block */
+	switch (private->rdc_data.vdev_class) {
+	case DEV_CLASS_FBA:
+		private->pt_block = 1;
+		break;
+	case DEV_CLASS_ECKD:
+		private->pt_block = 2;
+		break;
+	default:
+		DEV_MESSAGE(KERN_WARNING, device, "unsupported device class "
+			    "(class=%d)", private->rdc_data.vdev_class);
+		return -ENOTSUPP;
+	}
+
+	DBF_DEV_EVENT(DBF_INFO, device,
+		      "%04X: %04X on real %04X/%02X",
+		      rdc_data->dev_nr,
+		      rdc_data->vdev_type,
+		      rdc_data->rdev_type, rdc_data->rdev_model);
+
+	/* terminate all outstanding operations */
+	mdsk_term_io(device);
+
+	/* figure out blocksize of device */
+	label = (struct vtoc_cms_label *) get_zeroed_page(GFP_KERNEL);
+	if (label == NULL)  {
+		DEV_MESSAGE(KERN_WARNING, device, "%s",
+			    "No memory to allocate initialization request");
+		return -ENOMEM;
+	}
+	rc = 0;
+	end_block = 0;
+	/* try all sizes - needed for ECKD devices */
+	for (bsize = 512; bsize <= PAGE_SIZE; bsize <<= 1) {
+		mdsk_init_io(device, bsize, 0, &end_block);
+		memset(&bio, 0, sizeof (struct dasd_diag_bio));
+		bio.type = MDSK_READ_REQ;
+		bio.block_number = private->pt_block + 1;
+		bio.buffer = label;
+		memset(&private->iob, 0, sizeof (struct dasd_diag_rw_io));
+		private->iob.dev_nr = rdc_data->dev_nr;
+		private->iob.key = 0;
+		private->iob.flags = 0;	/* do synchronous io */
+		private->iob.block_count = 1;
+		private->iob.interrupt_params = 0;
+		private->iob.bio_list = &bio;
+		private->iob.flaga = DASD_DIAG_FLAGA_DEFAULT;
+		rc = dia250(&private->iob, RW_BIO);
+		if (rc == 3) {
+			DEV_MESSAGE(KERN_WARNING, device, "%s",
+				"DIAG call failed");
+			rc = -EOPNOTSUPP;
+			goto out;
+		}
+		mdsk_term_io(device);
+		if (rc == 0)
+			break;
+	}
+	if (bsize > PAGE_SIZE) {
+		DEV_MESSAGE(KERN_WARNING, device, "device access failed "
+			    "(rc=%d)", rc);
+		rc = -EIO;
+		goto out;
+	}
+	/* check for label block */
+	if (memcmp(label->label_id, DASD_DIAG_CMS1,
+		  sizeof(DASD_DIAG_CMS1)) == 0) {
+		/* get formatted blocksize from label block */
+		bsize = (unsigned int) label->block_size;
+		device->blocks = (unsigned long) label->block_count;
+	} else
+		device->blocks = end_block;
+	device->bp_block = bsize;
+	device->s2b_shift = 0;	/* bits to shift 512 to get a block */
+	for (sb = 512; sb < bsize; sb = sb << 1)
+		device->s2b_shift++;
+	rc = mdsk_init_io(device, device->bp_block, 0, NULL);
+	if (rc) {
+		DEV_MESSAGE(KERN_WARNING, device, "DIAG initialization "
+			"failed (rc=%d)", rc);
+		rc = -EIO;
+	} else {
+		DEV_MESSAGE(KERN_INFO, device,
+			    "(%ld B/blk): %ldkB",
+			    (unsigned long) device->bp_block,
+			    (unsigned long) (device->blocks <<
+				device->s2b_shift) >> 1);
+	}
+out:
+	free_page((long) label);
+	return rc;
+}
+
+/* Fill in virtual disk geometry for device. Return zero on success, non-zero
+ * otherwise. */
+static int
+dasd_diag_fill_geometry(struct dasd_device *device, struct hd_geometry *geo)
+{
+	if (dasd_check_blocksize(device->bp_block) != 0)
+		return -EINVAL;
+	geo->cylinders = (device->blocks << device->s2b_shift) >> 10;
+	geo->heads = 16;
+	geo->sectors = 128 >> device->s2b_shift;
+	return 0;
+}
+
+static dasd_era_t
+dasd_diag_examine_error(struct dasd_ccw_req * cqr, struct irb * stat)
+{
+	return dasd_era_fatal;
+}
+
+static dasd_erp_fn_t
+dasd_diag_erp_action(struct dasd_ccw_req * cqr)
+{
+	return dasd_default_erp_action;
+}
+
+static dasd_erp_fn_t
+dasd_diag_erp_postaction(struct dasd_ccw_req * cqr)
+{
+	return dasd_default_erp_postaction;
+}
+
+/* Create DASD request from block device request. Return pointer to new
+ * request on success, ERR_PTR otherwise. */
+static struct dasd_ccw_req *
+dasd_diag_build_cp(struct dasd_device * device, struct request *req)
+{
+	struct dasd_ccw_req *cqr;
+	struct dasd_diag_req *dreq;
+	struct dasd_diag_bio *dbio;
+	struct bio *bio;
+	struct bio_vec *bv;
+	char *dst;
+	unsigned int count, datasize;
+	sector_t recid, first_rec, last_rec;
+	unsigned int blksize, off;
+	unsigned char rw_cmd;
+	int i;
+
+	if (rq_data_dir(req) == READ)
+		rw_cmd = MDSK_READ_REQ;
+	else if (rq_data_dir(req) == WRITE)
+		rw_cmd = MDSK_WRITE_REQ;
+	else
+		return ERR_PTR(-EINVAL);
+	blksize = device->bp_block;
+	/* Calculate record id of first and last block. */
+	first_rec = req->sector >> device->s2b_shift;
+	last_rec = (req->sector + req->nr_sectors - 1) >> device->s2b_shift;
+	/* Check struct bio and count the number of blocks for the request. */
+	count = 0;
+	rq_for_each_bio(bio, req) {
+		bio_for_each_segment(bv, bio, i) {
+			if (bv->bv_len & (blksize - 1))
+				/* Fba can only do full blocks. */
+				return ERR_PTR(-EINVAL);
+			count += bv->bv_len >> (device->s2b_shift + 9);
+		}
+	}
+	/* Paranoia. */
+	if (count != last_rec - first_rec + 1)
+		return ERR_PTR(-EINVAL);
+	/* Build the request */
+	datasize = sizeof(struct dasd_diag_req) +
+		count*sizeof(struct dasd_diag_bio);
+	cqr = dasd_smalloc_request(dasd_diag_discipline.name, 0,
+				   datasize, device);
+	if (IS_ERR(cqr))
+		return cqr;
+
+	dreq = (struct dasd_diag_req *) cqr->data;
+	dreq->block_count = count;
+	dbio = dreq->bio;
+	recid = first_rec;
+	rq_for_each_bio(bio, req) {
+		bio_for_each_segment(bv, bio, i) {
+			dst = page_address(bv->bv_page) + bv->bv_offset;
+			for (off = 0; off < bv->bv_len; off += blksize) {
+				memset(dbio, 0, sizeof (struct dasd_diag_bio));
+				dbio->type = rw_cmd;
+				dbio->block_number = recid + 1;
+				dbio->buffer = dst;
+				dbio++;
+				dst += blksize;
+				recid++;
+			}
+		}
+	}
+	cqr->retries = DIAG_MAX_RETRIES;
+	cqr->buildclk = get_clock();
+	if (req->cmd_flags & REQ_FAILFAST)
+		set_bit(DASD_CQR_FLAGS_FAILFAST, &cqr->flags);
+	cqr->device = device;
+	cqr->expires = DIAG_TIMEOUT;
+	cqr->status = DASD_CQR_FILLED;
+	return cqr;
+}
+
+/* Release DASD request. Return non-zero if request was successful, zero
+ * otherwise. */
+static int
+dasd_diag_free_cp(struct dasd_ccw_req *cqr, struct request *req)
+{
+	int status;
+
+	status = cqr->status == DASD_CQR_DONE;
+	dasd_sfree_request(cqr, cqr->device);
+	return status;
+}
+
+/* Fill in IOCTL data for device. */
+static int
+dasd_diag_fill_info(struct dasd_device * device,
+		    struct dasd_information2_t * info)
+{
+	struct dasd_diag_private *private;
+
+	private = (struct dasd_diag_private *) device->private;
+	info->label_block = (unsigned int) private->pt_block;
+	info->FBA_layout = 1;
+	info->format = DASD_FORMAT_LDL;
+	info->characteristics_size = sizeof (struct dasd_diag_characteristics);
+	memcpy(info->characteristics,
+	       &((struct dasd_diag_private *) device->private)->rdc_data,
+	       sizeof (struct dasd_diag_characteristics));
+	info->confdata_size = 0;
+	return 0;
+}
+
+static void
+dasd_diag_dump_sense(struct dasd_device *device, struct dasd_ccw_req * req,
+		     struct irb *stat)
+{
+	DEV_MESSAGE(KERN_ERR, device, "%s",
+		    "dump sense not available for DIAG data");
+}
+
+static struct dasd_discipline dasd_diag_discipline = {
+	.owner = THIS_MODULE,
+	.name = "DIAG",
+	.ebcname = "DIAG",
+	.max_blocks = DIAG_MAX_BLOCKS,
+	.check_device = dasd_diag_check_device,
+	.fill_geometry = dasd_diag_fill_geometry,
+	.start_IO = dasd_start_diag,
+	.term_IO = dasd_diag_term_IO,
+	.examine_error = dasd_diag_examine_error,
+	.erp_action = dasd_diag_erp_action,
+	.erp_postaction = dasd_diag_erp_postaction,
+	.build_cp = dasd_diag_build_cp,
+	.free_cp = dasd_diag_free_cp,
+	.dump_sense = dasd_diag_dump_sense,
+	.fill_info = dasd_diag_fill_info,
+};
+
+static int __init
+dasd_diag_init(void)
+{
+	if (!MACHINE_IS_VM) {
+		MESSAGE_LOG(KERN_INFO,
+			    "Machine is not VM: %s "
+			    "discipline not initializing",
+			    dasd_diag_discipline.name);
+		return -ENODEV;
+	}
+	ASCEBC(dasd_diag_discipline.ebcname, 4);
+
+	ctl_set_bit(0, 9);
+	register_external_interrupt(0x2603, dasd_ext_handler);
+	dasd_diag_discipline_pointer = &dasd_diag_discipline;
+	return 0;
+}
+
+static void __exit
+dasd_diag_cleanup(void)
+{
+	unregister_external_interrupt(0x2603, dasd_ext_handler);
+	ctl_clear_bit(0, 9);
+	dasd_diag_discipline_pointer = NULL;
+}
+
+module_init(dasd_diag_init);
+module_exit(dasd_diag_cleanup);

drivers/s390/block/dasd_diag.h

@@ -0,0 +1,127 @@
+/*
+ * File...........: linux/drivers/s390/block/dasd_diag.h
+ * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com>
+ * Based on.......: linux/drivers/s390/block/mdisk.h
+ * ...............: by Hartmunt Penner <hpenner@de.ibm.com>
+ * Bugreports.to..: <Linux390@de.ibm.com>
+ * (C) IBM Corporation, IBM Deutschland Entwicklung GmbH, 1999,2000
+ *
+ */
+
+#define MDSK_WRITE_REQ 0x01
+#define MDSK_READ_REQ  0x02
+
+#define INIT_BIO	0x00
+#define RW_BIO		0x01
+#define TERM_BIO	0x02
+
+#define DEV_CLASS_FBA	0x01
+#define DEV_CLASS_ECKD	0x04
+
+#define DASD_DIAG_LC_INT_CODE		132
+#define DASD_DIAG_LC_INT_STATUS		133
+#define DASD_DIAG_LC_INT_PARM_31BIT	128
+#define DASD_DIAG_LC_INT_PARM_64BIT	4536
+#define DASD_DIAG_CODE_31BIT		0x03
+#define DASD_DIAG_CODE_64BIT		0x07
+
+#define DASD_DIAG_RWFLAG_ASYNC		0x02
+#define DASD_DIAG_RWFLAG_NOCACHE	0x01
+
+#define DASD_DIAG_FLAGA_FORMAT_64BIT	0x80
+
+struct dasd_diag_characteristics {
+	u16 dev_nr;
+	u16 rdc_len;
+	u8 vdev_class;
+	u8 vdev_type;
+	u8 vdev_status;
+	u8 vdev_flags;
+	u8 rdev_class;
+	u8 rdev_type;
+	u8 rdev_model;
+	u8 rdev_features;
+} __attribute__ ((packed, aligned(4)));
+
+
+#ifdef CONFIG_64BIT
+#define DASD_DIAG_FLAGA_DEFAULT		DASD_DIAG_FLAGA_FORMAT_64BIT
+
+typedef u64 blocknum_t;
+typedef s64 sblocknum_t;
+
+struct dasd_diag_bio {
+	u8 type;
+	u8 status;
+	u8 spare1[2];
+	u32 alet;
+	blocknum_t block_number;
+	void *buffer;
+} __attribute__ ((packed, aligned(8)));
+
+struct dasd_diag_init_io {
+	u16 dev_nr;
+	u8 flaga;
+	u8 spare1[21];
+	u32 block_size;
+	u8 spare2[4];
+	blocknum_t offset;
+	sblocknum_t start_block;
+	blocknum_t end_block;
+	u8  spare3[8];
+} __attribute__ ((packed, aligned(8)));
+
+struct dasd_diag_rw_io {
+	u16 dev_nr;
+	u8  flaga;
+	u8  spare1[21];
+	u8  key;
+	u8  flags;
+	u8  spare2[2];
+	u32 block_count;
+	u32 alet;
+	u8  spare3[4];
+	u64 interrupt_params;
+	struct dasd_diag_bio *bio_list;
+	u8  spare4[8];
+} __attribute__ ((packed, aligned(8)));
+#else /* CONFIG_64BIT */
+#define DASD_DIAG_FLAGA_DEFAULT		0x0
+
+typedef u32 blocknum_t;
+typedef s32 sblocknum_t;
+
+struct dasd_diag_bio {
+	u8 type;
+	u8 status;
+	u16 spare1;
+	blocknum_t block_number;
+	u32 alet;
+	void *buffer;
+} __attribute__ ((packed, aligned(8)));
+
+struct dasd_diag_init_io {
+	u16 dev_nr;
+	u8 flaga;
+	u8 spare1[21];
+	u32 block_size;
+	blocknum_t offset;
+	sblocknum_t start_block;
+	blocknum_t end_block;
+	u8 spare2[24];
+} __attribute__ ((packed, aligned(8)));
+
+struct dasd_diag_rw_io {
+	u16 dev_nr;
+	u8 flaga;
+	u8 spare1[21];
+	u8 key;
+	u8 flags;
+	u8 spare2[2];
+	u32 block_count;
+	u32 alet;
+	struct dasd_diag_bio *bio_list;
+	u32 interrupt_params;
+	u8 spare3[20];
+} __attribute__ ((packed, aligned(8)));
+#endif /* CONFIG_64BIT */

drivers/s390/block/dasd_eckd.h

@@ -0,0 +1,368 @@
+/*
+ * File...........: linux/drivers/s390/block/dasd_eckd.h
+ * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com>
+ *		    Horst Hummel <Horst.Hummel@de.ibm.com>
+ * Bugreports.to..: <Linux390@de.ibm.com>
+ * (C) IBM Corporation, IBM Deutschland Entwicklung GmbH, 1999,2000
+ *
+ */
+
+#ifndef DASD_ECKD_H
+#define DASD_ECKD_H
+
+/*****************************************************************************
+ * SECTION: CCW Definitions
+ ****************************************************************************/
+#define DASD_ECKD_CCW_WRITE		 0x05
+#define DASD_ECKD_CCW_READ		 0x06
+#define DASD_ECKD_CCW_WRITE_HOME_ADDRESS 0x09
+#define DASD_ECKD_CCW_READ_HOME_ADDRESS	 0x0a
+#define DASD_ECKD_CCW_WRITE_KD		 0x0d
+#define DASD_ECKD_CCW_READ_KD		 0x0e
+#define DASD_ECKD_CCW_ERASE		 0x11
+#define DASD_ECKD_CCW_READ_COUNT	 0x12
+#define DASD_ECKD_CCW_SLCK		 0x14
+#define DASD_ECKD_CCW_WRITE_RECORD_ZERO	 0x15
+#define DASD_ECKD_CCW_READ_RECORD_ZERO	 0x16
+#define DASD_ECKD_CCW_WRITE_CKD		 0x1d
+#define DASD_ECKD_CCW_READ_CKD		 0x1e
+#define DASD_ECKD_CCW_PSF		 0x27
+#define DASD_ECKD_CCW_RSSD		 0x3e
+#define DASD_ECKD_CCW_LOCATE_RECORD	 0x47
+#define DASD_ECKD_CCW_SNSS		 0x54
+#define DASD_ECKD_CCW_DEFINE_EXTENT	 0x63
+#define DASD_ECKD_CCW_WRITE_MT		 0x85
+#define DASD_ECKD_CCW_READ_MT		 0x86
+#define DASD_ECKD_CCW_WRITE_KD_MT	 0x8d
+#define DASD_ECKD_CCW_READ_KD_MT	 0x8e
+#define DASD_ECKD_CCW_RELEASE		 0x94
+#define DASD_ECKD_CCW_READ_CKD_MT	 0x9e
+#define DASD_ECKD_CCW_WRITE_CKD_MT	 0x9d
+#define DASD_ECKD_CCW_RESERVE		 0xB4
+
+/*
+ * Perform Subsystem Function / Sub-Orders
+ */
+#define PSF_ORDER_PRSSD 0x18
+#define PSF_ORDER_SSC	0x1D
+
+/*****************************************************************************
+ * SECTION: Type Definitions
+ ****************************************************************************/
+
+struct eckd_count {
+	__u16 cyl;
+	__u16 head;
+	__u8 record;
+	__u8 kl;
+	__u16 dl;
+} __attribute__ ((packed));
+
+struct ch_t {
+	__u16 cyl;
+	__u16 head;
+} __attribute__ ((packed));
+
+struct chs_t {
+	__u16 cyl;
+	__u16 head;
+	__u32 sector;
+} __attribute__ ((packed));
+
+struct chr_t {
+	__u16 cyl;
+	__u16 head;
+	__u8 record;
+} __attribute__ ((packed));
+
+struct geom_t {
+	__u16 cyl;
+	__u16 head;
+	__u32 sector;
+} __attribute__ ((packed));
+
+struct eckd_home {
+	__u8 skip_control[14];
+	__u16 cell_number;
+	__u8 physical_addr[3];
+	__u8 flag;
+	struct ch_t track_addr;
+	__u8 reserved;
+	__u8 key_length;
+	__u8 reserved2[2];
+} __attribute__ ((packed));
+
+struct DE_eckd_data {
+	struct {
+		unsigned char perm:2;	/* Permissions on this extent */
+		unsigned char reserved:1;
+		unsigned char seek:2;	/* Seek control */
+		unsigned char auth:2;	/* Access authorization */
+		unsigned char pci:1;	/* PCI Fetch mode */
+	} __attribute__ ((packed)) mask;
+	struct {
+		unsigned char mode:2;	/* Architecture mode */
+		unsigned char ckd:1;	/* CKD Conversion */
+		unsigned char operation:3;	/* Operation mode */
+		unsigned char cfw:1;	/* Cache fast write */
+		unsigned char dfw:1;	/* DASD fast write */
+	} __attribute__ ((packed)) attributes;
+	__u16 blk_size;		/* Blocksize */
+	__u16 fast_write_id;
+	__u8 ga_additional;	/* Global Attributes Additional */
+	__u8 ga_extended;	/* Global Attributes Extended	*/
+	struct ch_t beg_ext;
+	struct ch_t end_ext;
+	unsigned long long ep_sys_time; /* Ext Parameter - System Time Stamp */
+	__u8 ep_format;        /* Extended Parameter format byte       */
+	__u8 ep_prio;          /* Extended Parameter priority I/O byte */
+	__u8 ep_reserved[6];   /* Extended Parameter Reserved          */
+} __attribute__ ((packed));
+
+struct LO_eckd_data {
+	struct {
+		unsigned char orientation:2;
+		unsigned char operation:6;
+	} __attribute__ ((packed)) operation;
+	struct {
+		unsigned char last_bytes_used:1;
+		unsigned char reserved:6;
+		unsigned char read_count_suffix:1;
+	} __attribute__ ((packed)) auxiliary;
+	__u8 unused;
+	__u8 count;
+	struct ch_t seek_addr;
+	struct chr_t search_arg;
+	__u8 sector;
+	__u16 length;
+} __attribute__ ((packed));
+
+struct dasd_eckd_characteristics {
+	__u16 cu_type;
+	struct {
+		unsigned char support:2;
+		unsigned char async:1;
+		unsigned char reserved:1;
+		unsigned char cache_info:1;
+		unsigned char model:3;
+	} __attribute__ ((packed)) cu_model;
+	__u16 dev_type;
+	__u8 dev_model;
+	struct {
+		unsigned char mult_burst:1;
+		unsigned char RT_in_LR:1;
+		unsigned char reserved1:1;
+		unsigned char RD_IN_LR:1;
+		unsigned char reserved2:4;
+		unsigned char reserved3:8;
+		unsigned char defect_wr:1;
+		unsigned char XRC_supported:1;
+		unsigned char reserved4:1;
+		unsigned char striping:1;
+		unsigned char reserved5:4;
+		unsigned char cfw:1;
+		unsigned char reserved6:2;
+		unsigned char cache:1;
+		unsigned char dual_copy:1;
+		unsigned char dfw:1;
+		unsigned char reset_alleg:1;
+		unsigned char sense_down:1;
+	} __attribute__ ((packed)) facilities;
+	__u8 dev_class;
+	__u8 unit_type;
+	__u16 no_cyl;
+	__u16 trk_per_cyl;
+	__u8 sec_per_trk;
+	__u8 byte_per_track[3];
+	__u16 home_bytes;
+	__u8 formula;
+	union {
+		struct {
+			__u8 f1;
+			__u16 f2;
+			__u16 f3;
+		} __attribute__ ((packed)) f_0x01;
+		struct {
+			__u8 f1;
+			__u8 f2;
+			__u8 f3;
+			__u8 f4;
+			__u8 f5;
+		} __attribute__ ((packed)) f_0x02;
+	} __attribute__ ((packed)) factors;
+	__u16 first_alt_trk;
+	__u16 no_alt_trk;
+	__u16 first_dia_trk;
+	__u16 no_dia_trk;
+	__u16 first_sup_trk;
+	__u16 no_sup_trk;
+	__u8 MDR_ID;
+	__u8 OBR_ID;
+	__u8 director;
+	__u8 rd_trk_set;
+	__u16 max_rec_zero;
+	__u8 reserved1;
+	__u8 RWANY_in_LR;
+	__u8 factor6;
+	__u8 factor7;
+	__u8 factor8;
+	__u8 reserved2[3];
+	__u8 reserved3[10];
+} __attribute__ ((packed));
+
+struct dasd_eckd_confdata {
+	struct {
+		struct {
+			unsigned char identifier:2;
+			unsigned char token_id:1;
+			unsigned char sno_valid:1;
+			unsigned char subst_sno:1;
+			unsigned char recNED:1;
+			unsigned char emuNED:1;
+			unsigned char reserved:1;
+		} __attribute__ ((packed)) flags;
+		__u8 descriptor;
+		__u8 dev_class;
+		__u8 reserved;
+		unsigned char dev_type[6];
+		unsigned char dev_model[3];
+		unsigned char HDA_manufacturer[3];
+		unsigned char HDA_location[2];
+		unsigned char HDA_seqno[12];
+		__u8 ID;
+		__u8 unit_addr;
+	} __attribute__ ((packed)) ned1;
+	union {
+		struct {
+			struct {
+				unsigned char identifier:2;
+				unsigned char token_id:1;
+				unsigned char sno_valid:1;
+				unsigned char subst_sno:1;
+				unsigned char recNED:1;
+				unsigned char emuNED:1;
+				unsigned char reserved:1;
+			} __attribute__ ((packed)) flags;
+			__u8 descriptor;
+			__u8 reserved[2];
+			unsigned char dev_type[6];
+			unsigned char dev_model[3];
+			unsigned char DASD_manufacturer[3];
+			unsigned char DASD_location[2];
+			unsigned char DASD_seqno[12];
+			__u16 ID;
+		} __attribute__ ((packed)) ned;
+		struct {
+			unsigned char flags;            /* byte  0    */
+			unsigned char res2[7];          /* byte  1- 7 */
+			unsigned char sua_flags;	/* byte  8    */
+			__u8 base_unit_addr;            /* byte  9    */
+			unsigned char res3[22];	        /* byte 10-31 */
+		} __attribute__ ((packed)) sneq;
+	} __attribute__ ((packed)) ned2;
+	struct {
+		struct {
+			unsigned char identifier:2;
+			unsigned char token_id:1;
+			unsigned char sno_valid:1;
+			unsigned char subst_sno:1;
+			unsigned char recNED:1;
+			unsigned char emuNED:1;
+			unsigned char reserved:1;
+		} __attribute__ ((packed)) flags;
+		__u8 descriptor;
+		__u8 reserved[2];
+		unsigned char cont_type[6];
+		unsigned char cont_model[3];
+		unsigned char cont_manufacturer[3];
+		unsigned char cont_location[2];
+		unsigned char cont_seqno[12];
+		__u16 ID;
+	} __attribute__ ((packed)) ned3;
+	struct {
+		struct {
+			unsigned char identifier:2;
+			unsigned char token_id:1;
+			unsigned char sno_valid:1;
+			unsigned char subst_sno:1;
+			unsigned char recNED:1;
+			unsigned char emuNED:1;
+			unsigned char reserved:1;
+		} __attribute__ ((packed)) flags;
+		__u8 descriptor;
+		__u8 reserved[2];
+		unsigned char cont_type[6];
+		unsigned char empty[3];
+		unsigned char cont_manufacturer[3];
+		unsigned char cont_location[2];
+		unsigned char cont_seqno[12];
+		__u16 ID;
+	} __attribute__ ((packed)) ned4;
+	unsigned char ned5[32];
+	unsigned char ned6[32];
+	unsigned char ned7[32];
+	struct {
+		struct {
+			unsigned char identifier:2;
+			unsigned char reserved:6;
+		} __attribute__ ((packed)) flags;
+		__u8 selector;
+		__u16 interfaceID;
+		__u32 reserved;
+		__u16 subsystemID;
+		struct {
+			unsigned char sp0:1;
+			unsigned char sp1:1;
+			unsigned char reserved:5;
+			unsigned char scluster:1;
+		} __attribute__ ((packed)) spathID;
+		__u8 unit_address;
+		__u8 dev_ID;
+		__u8 dev_address;
+		__u8 adapterID;
+		__u16 link_address;
+		struct {
+			unsigned char parallel:1;
+			unsigned char escon:1;
+			unsigned char reserved:1;
+			unsigned char ficon:1;
+			unsigned char reserved2:4;
+		} __attribute__ ((packed)) protocol_type;
+		struct {
+			unsigned char PID_in_236:1;
+			unsigned char reserved:7;
+		} __attribute__ ((packed)) format_flags;
+		__u8 log_dev_address;
+		unsigned char reserved2[12];
+	} __attribute__ ((packed)) neq;
+} __attribute__ ((packed));
+
+struct dasd_eckd_path {
+	__u8 opm;
+	__u8 ppm;
+	__u8 npm;
+};
+
+/*
+ * Perform Subsystem Function - Prepare for Read Subsystem Data
+ */
+struct dasd_psf_prssd_data {
+	unsigned char order;
+	unsigned char flags;
+	unsigned char reserved[4];
+	unsigned char suborder;
+	unsigned char varies[9];
+} __attribute__ ((packed));
+
+/*
+ * Perform Subsystem Function - Set Subsystem Characteristics
+ */
+struct dasd_psf_ssc_data {
+	unsigned char order;
+	unsigned char flags;
+	unsigned char cu_type[4];
+	unsigned char suborder;
+	unsigned char reserved[59];
+} __attribute__((packed));
+
+#endif				/* DASD_ECKD_H */

drivers/s390/block/dasd_eer.c

@@ -0,0 +1,695 @@
+/*
+ *  Character device driver for extended error reporting.
+ *
+ *  Copyright (C) 2005 IBM Corporation
+ *  extended error reporting for DASD ECKD devices
+ *  Author(s): Stefan Weinhuber <wein@de.ibm.com>
+ */
+
+#include <linux/init.h>
+#include <linux/fs.h>
+#include <linux/kernel.h>
+#include <linux/miscdevice.h>
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/device.h>
+#include <linux/poll.h>
+
+#include <asm/uaccess.h>
+#include <asm/semaphore.h>
+#include <asm/atomic.h>
+#include <asm/ebcdic.h>
+
+#include "dasd_int.h"
+#include "dasd_eckd.h"
+
+#ifdef PRINTK_HEADER
+#undef PRINTK_HEADER
+#endif				/* PRINTK_HEADER */
+#define PRINTK_HEADER "dasd(eer):"
+
+/*
+ * SECTION: the internal buffer
+ */
+
+/*
+ * The internal buffer is meant to store obaque blobs of data, so it does
+ * not know of higher level concepts like triggers.
+ * It consists of a number of pages that are used as a ringbuffer. Each data
+ * blob is stored in a simple record that consists of an integer, which
+ * contains the size of the following data, and the data bytes themselfes.
+ *
+ * To allow for multiple independent readers we create one internal buffer
+ * each time the device is opened and destroy the buffer when the file is
+ * closed again. The number of pages used for this buffer is determined by
+ * the module parmeter eer_pages.
+ *
+ * One record can be written to a buffer by using the functions
+ * - dasd_eer_start_record (one time per record to write the size to the
+ *                          buffer and reserve the space for the data)
+ * - dasd_eer_write_buffer (one or more times per record to write the data)
+ * The data can be written in several steps but you will have to compute
+ * the total size up front for the invocation of dasd_eer_start_record.
+ * If the ringbuffer is full, dasd_eer_start_record will remove the required
+ * number of old records.
+ *
+ * A record is typically read in two steps, first read the integer that
+ * specifies the size of the following data, then read the data.
+ * Both can be done by
+ * - dasd_eer_read_buffer
+ *
+ * For all mentioned functions you need to get the bufferlock first and keep
+ * it until a complete record is written or read.
+ *
+ * All information necessary to keep track of an internal buffer is kept in
+ * a struct eerbuffer. The buffer specific to a file pointer is strored in
+ * the private_data field of that file. To be able to write data to all
+ * existing buffers, each buffer is also added to the bufferlist.
+ * If the user does not want to read a complete record in one go, we have to
+ * keep track of the rest of the record. residual stores the number of bytes
+ * that are still to deliver. If the rest of the record is invalidated between
+ * two reads then residual will be set to -1 so that the next read will fail.
+ * All entries in the eerbuffer structure are protected with the bufferlock.
+ * To avoid races between writing to a buffer on the one side and creating
+ * and destroying buffers on the other side, the bufferlock must also be used
+ * to protect the bufferlist.
+ */
+
+static int eer_pages = 5;
+module_param(eer_pages, int, S_IRUGO|S_IWUSR);
+
+struct eerbuffer {
+	struct list_head list;
+	char **buffer;
+	int buffersize;
+	int buffer_page_count;
+	int head;
+        int tail;
+	int residual;
+};
+
+static LIST_HEAD(bufferlist);
+static DEFINE_SPINLOCK(bufferlock);
+static DECLARE_WAIT_QUEUE_HEAD(dasd_eer_read_wait_queue);
+
+/*
+ * How many free bytes are available on the buffer.
+ * Needs to be called with bufferlock held.
+ */
+static int dasd_eer_get_free_bytes(struct eerbuffer *eerb)
+{
+	if (eerb->head < eerb->tail)
+		return eerb->tail - eerb->head - 1;
+	return eerb->buffersize - eerb->head + eerb->tail -1;
+}
+
+/*
+ * How many bytes of buffer space are used.
+ * Needs to be called with bufferlock held.
+ */
+static int dasd_eer_get_filled_bytes(struct eerbuffer *eerb)
+{
+
+	if (eerb->head >= eerb->tail)
+		return eerb->head - eerb->tail;
+	return eerb->buffersize - eerb->tail + eerb->head;
+}
+
+/*
+ * The dasd_eer_write_buffer function just copies count bytes of data
+ * to the buffer. Make sure to call dasd_eer_start_record first, to
+ * make sure that enough free space is available.
+ * Needs to be called with bufferlock held.
+ */
+static void dasd_eer_write_buffer(struct eerbuffer *eerb,
+				  char *data, int count)
+{
+
+	unsigned long headindex,localhead;
+	unsigned long rest, len;
+	char *nextdata;
+
+	nextdata = data;
+	rest = count;
+	while (rest > 0) {
+ 		headindex = eerb->head / PAGE_SIZE;
+ 		localhead = eerb->head % PAGE_SIZE;
+		len = min(rest, PAGE_SIZE - localhead);
+		memcpy(eerb->buffer[headindex]+localhead, nextdata, len);
+		nextdata += len;
+		rest -= len;
+		eerb->head += len;
+		if (eerb->head == eerb->buffersize)
+			eerb->head = 0; /* wrap around */
+		BUG_ON(eerb->head > eerb->buffersize);
+	}
+}
+
+/*
+ * Needs to be called with bufferlock held.
+ */
+static int dasd_eer_read_buffer(struct eerbuffer *eerb, char *data, int count)
+{
+
+	unsigned long tailindex,localtail;
+	unsigned long rest, len, finalcount;
+	char *nextdata;
+
+	finalcount = min(count, dasd_eer_get_filled_bytes(eerb));
+	nextdata = data;
+	rest = finalcount;
+	while (rest > 0) {
+ 		tailindex = eerb->tail / PAGE_SIZE;
+ 		localtail = eerb->tail % PAGE_SIZE;
+		len = min(rest, PAGE_SIZE - localtail);
+		memcpy(nextdata, eerb->buffer[tailindex] + localtail, len);
+		nextdata += len;
+		rest -= len;
+		eerb->tail += len;
+		if (eerb->tail == eerb->buffersize)
+			eerb->tail = 0; /* wrap around */
+		BUG_ON(eerb->tail > eerb->buffersize);
+	}
+	return finalcount;
+}
+
+/*
+ * Whenever you want to write a blob of data to the internal buffer you
+ * have to start by using this function first. It will write the number
+ * of bytes that will be written to the buffer. If necessary it will remove
+ * old records to make room for the new one.
+ * Needs to be called with bufferlock held.
+ */
+static int dasd_eer_start_record(struct eerbuffer *eerb, int count)
+{
+	int tailcount;
+
+	if (count + sizeof(count) > eerb->buffersize)
+		return -ENOMEM;
+	while (dasd_eer_get_free_bytes(eerb) < count + sizeof(count)) {
+		if (eerb->residual > 0) {
+			eerb->tail += eerb->residual;
+			if (eerb->tail >= eerb->buffersize)
+				eerb->tail -= eerb->buffersize;
+			eerb->residual = -1;
+		}
+		dasd_eer_read_buffer(eerb, (char *) &tailcount,
+				     sizeof(tailcount));
+		eerb->tail += tailcount;
+		if (eerb->tail >= eerb->buffersize)
+			eerb->tail -= eerb->buffersize;
+	}
+	dasd_eer_write_buffer(eerb, (char*) &count, sizeof(count));
+
+	return 0;
+};
+
+/*
+ * Release pages that are not used anymore.
+ */
+static void dasd_eer_free_buffer_pages(char **buf, int no_pages)
+{
+	int i;
+
+	for (i = 0; i < no_pages; i++)
+		free_page((unsigned long) buf[i]);
+}
+
+/*
+ * Allocate a new set of memory pages.
+ */
+static int dasd_eer_allocate_buffer_pages(char **buf, int no_pages)
+{
+	int i;
+
+	for (i = 0; i < no_pages; i++) {
+		buf[i] = (char *) get_zeroed_page(GFP_KERNEL);
+		if (!buf[i]) {
+			dasd_eer_free_buffer_pages(buf, i);
+			return -ENOMEM;
+		}
+	}
+	return 0;
+}
+
+/*
+ * SECTION: The extended error reporting functionality
+ */
+
+/*
+ * When a DASD device driver wants to report an error, it calls the
+ * function dasd_eer_write and gives the respective trigger ID as
+ * parameter. Currently there are four kinds of triggers:
+ *
+ * DASD_EER_FATALERROR:  all kinds of unrecoverable I/O problems
+ * DASD_EER_PPRCSUSPEND: PPRC was suspended
+ * DASD_EER_NOPATH:      There is no path to the device left.
+ * DASD_EER_STATECHANGE: The state of the device has changed.
+ *
+ * For the first three triggers all required information can be supplied by
+ * the caller. For these triggers a record is written by the function
+ * dasd_eer_write_standard_trigger.
+ *
+ * The DASD_EER_STATECHANGE trigger is special since a sense subsystem
+ * status ccw need to be executed to gather the necessary sense data first.
+ * The dasd_eer_snss function will queue the SNSS request and the request
+ * callback will then call dasd_eer_write with the DASD_EER_STATCHANGE
+ * trigger.
+ *
+ * To avoid memory allocations at runtime, the necessary memory is allocated
+ * when the extended error reporting is enabled for a device (by
+ * dasd_eer_probe). There is one sense subsystem status request for each
+ * eer enabled DASD device. The presence of the cqr in device->eer_cqr
+ * indicates that eer is enable for the device. The use of the snss request
+ * is protected by the DASD_FLAG_EER_IN_USE bit. When this flag indicates
+ * that the cqr is currently in use, dasd_eer_snss cannot start a second
+ * request but sets the DASD_FLAG_EER_SNSS flag instead. The callback of
+ * the SNSS request will check the bit and call dasd_eer_snss again.
+ */
+
+#define SNSS_DATA_SIZE 44
+
+#define DASD_EER_BUSID_SIZE 10
+struct dasd_eer_header {
+	__u32 total_size;
+	__u32 trigger;
+	__u64 tv_sec;
+	__u64 tv_usec;
+	char busid[DASD_EER_BUSID_SIZE];
+} __attribute__ ((packed));
+
+/*
+ * The following function can be used for those triggers that have
+ * all necessary data available when the function is called.
+ * If the parameter cqr is not NULL, the chain of requests will be searched
+ * for valid sense data, and all valid sense data sets will be added to
+ * the triggers data.
+ */
+static void dasd_eer_write_standard_trigger(struct dasd_device *device,
+					    struct dasd_ccw_req *cqr,
+					    int trigger)
+{
+	struct dasd_ccw_req *temp_cqr;
+	int data_size;
+	struct timeval tv;
+	struct dasd_eer_header header;
+	unsigned long flags;
+	struct eerbuffer *eerb;
+
+	/* go through cqr chain and count the valid sense data sets */
+	data_size = 0;
+	for (temp_cqr = cqr; temp_cqr; temp_cqr = temp_cqr->refers)
+		if (temp_cqr->irb.esw.esw0.erw.cons)
+			data_size += 32;
+
+	header.total_size = sizeof(header) + data_size + 4; /* "EOR" */
+	header.trigger = trigger;
+	do_gettimeofday(&tv);
+	header.tv_sec = tv.tv_sec;
+	header.tv_usec = tv.tv_usec;
+	strncpy(header.busid, device->cdev->dev.bus_id, DASD_EER_BUSID_SIZE);
+
+	spin_lock_irqsave(&bufferlock, flags);
+	list_for_each_entry(eerb, &bufferlist, list) {
+		dasd_eer_start_record(eerb, header.total_size);
+		dasd_eer_write_buffer(eerb, (char *) &header, sizeof(header));
+		for (temp_cqr = cqr; temp_cqr; temp_cqr = temp_cqr->refers)
+			if (temp_cqr->irb.esw.esw0.erw.cons)
+				dasd_eer_write_buffer(eerb, cqr->irb.ecw, 32);
+		dasd_eer_write_buffer(eerb, "EOR", 4);
+	}
+	spin_unlock_irqrestore(&bufferlock, flags);
+	wake_up_interruptible(&dasd_eer_read_wait_queue);
+}
+
+/*
+ * This function writes a DASD_EER_STATECHANGE trigger.
+ */
+static void dasd_eer_write_snss_trigger(struct dasd_device *device,
+					struct dasd_ccw_req *cqr,
+					int trigger)
+{
+	int data_size;
+	int snss_rc;
+	struct timeval tv;
+	struct dasd_eer_header header;
+	unsigned long flags;
+	struct eerbuffer *eerb;
+
+	snss_rc = (cqr->status == DASD_CQR_FAILED) ? -EIO : 0;
+	if (snss_rc)
+		data_size = 0;
+	else
+		data_size = SNSS_DATA_SIZE;
+
+	header.total_size = sizeof(header) + data_size + 4; /* "EOR" */
+	header.trigger = DASD_EER_STATECHANGE;
+	do_gettimeofday(&tv);
+	header.tv_sec = tv.tv_sec;
+	header.tv_usec = tv.tv_usec;
+	strncpy(header.busid, device->cdev->dev.bus_id, DASD_EER_BUSID_SIZE);
+
+	spin_lock_irqsave(&bufferlock, flags);
+	list_for_each_entry(eerb, &bufferlist, list) {
+		dasd_eer_start_record(eerb, header.total_size);
+		dasd_eer_write_buffer(eerb, (char *) &header , sizeof(header));
+		if (!snss_rc)
+			dasd_eer_write_buffer(eerb, cqr->data, SNSS_DATA_SIZE);
+		dasd_eer_write_buffer(eerb, "EOR", 4);
+	}
+	spin_unlock_irqrestore(&bufferlock, flags);
+	wake_up_interruptible(&dasd_eer_read_wait_queue);
+}
+
+/*
+ * This function is called for all triggers. It calls the appropriate
+ * function that writes the actual trigger records.
+ */
+void dasd_eer_write(struct dasd_device *device, struct dasd_ccw_req *cqr,
+		    unsigned int id)
+{
+	if (!device->eer_cqr)
+		return;
+	switch (id) {
+	case DASD_EER_FATALERROR:
+	case DASD_EER_PPRCSUSPEND:
+		dasd_eer_write_standard_trigger(device, cqr, id);
+		break;
+	case DASD_EER_NOPATH:
+		dasd_eer_write_standard_trigger(device, NULL, id);
+		break;
+	case DASD_EER_STATECHANGE:
+		dasd_eer_write_snss_trigger(device, cqr, id);
+		break;
+	default: /* unknown trigger, so we write it without any sense data */
+		dasd_eer_write_standard_trigger(device, NULL, id);
+		break;
+	}
+}
+EXPORT_SYMBOL(dasd_eer_write);
+
+/*
+ * Start a sense subsystem status request.
+ * Needs to be called with the device held.
+ */
+void dasd_eer_snss(struct dasd_device *device)
+{
+	struct dasd_ccw_req *cqr;
+
+	cqr = device->eer_cqr;
+	if (!cqr)	/* Device not eer enabled. */
+		return;
+	if (test_and_set_bit(DASD_FLAG_EER_IN_USE, &device->flags)) {
+		/* Sense subsystem status request in use. */
+		set_bit(DASD_FLAG_EER_SNSS, &device->flags);
+		return;
+	}
+	clear_bit(DASD_FLAG_EER_SNSS, &device->flags);
+	cqr->status = DASD_CQR_QUEUED;
+	list_add(&cqr->list, &device->ccw_queue);
+	dasd_schedule_bh(device);
+}
+
+/*
+ * Callback function for use with sense subsystem status request.
+ */
+static void dasd_eer_snss_cb(struct dasd_ccw_req *cqr, void *data)
+{
+        struct dasd_device *device = cqr->device;
+	unsigned long flags;
+
+	dasd_eer_write(device, cqr, DASD_EER_STATECHANGE);
+	spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
+	if (device->eer_cqr == cqr) {
+		clear_bit(DASD_FLAG_EER_IN_USE, &device->flags);
+		if (test_bit(DASD_FLAG_EER_SNSS, &device->flags))
+			/* Another SNSS has been requested in the meantime. */
+			dasd_eer_snss(device);
+		cqr = NULL;
+	}
+	spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
+	if (cqr)
+		/*
+		 * Extended error recovery has been switched off while
+		 * the SNSS request was running. It could even have
+		 * been switched off and on again in which case there
+		 * is a new ccw in device->eer_cqr. Free the "old"
+		 * snss request now.
+		 */
+		dasd_kfree_request(cqr, device);
+}
+
+/*
+ * Enable error reporting on a given device.
+ */
+int dasd_eer_enable(struct dasd_device *device)
+{
+	struct dasd_ccw_req *cqr;
+	unsigned long flags;
+
+	if (device->eer_cqr)
+		return 0;
+
+	if (!device->discipline || strcmp(device->discipline->name, "ECKD"))
+		return -EPERM;	/* FIXME: -EMEDIUMTYPE ? */
+
+	cqr = dasd_kmalloc_request("ECKD", 1 /* SNSS */,
+				   SNSS_DATA_SIZE, device);
+	if (!cqr)
+		return -ENOMEM;
+
+	cqr->device = device;
+	cqr->retries = 255;
+	cqr->expires = 10 * HZ;
+	clear_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags);
+
+	cqr->cpaddr->cmd_code = DASD_ECKD_CCW_SNSS;
+	cqr->cpaddr->count = SNSS_DATA_SIZE;
+	cqr->cpaddr->flags = 0;
+	cqr->cpaddr->cda = (__u32)(addr_t) cqr->data;
+
+	cqr->buildclk = get_clock();
+	cqr->status = DASD_CQR_FILLED;
+	cqr->callback = dasd_eer_snss_cb;
+
+	spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
+	if (!device->eer_cqr) {
+		device->eer_cqr = cqr;
+		cqr = NULL;
+	}
+	spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
+	if (cqr)
+		dasd_kfree_request(cqr, device);
+	return 0;
+}
+
+/*
+ * Disable error reporting on a given device.
+ */
+void dasd_eer_disable(struct dasd_device *device)
+{
+	struct dasd_ccw_req *cqr;
+	unsigned long flags;
+	int in_use;
+
+	if (!device->eer_cqr)
+		return;
+	spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
+	cqr = device->eer_cqr;
+	device->eer_cqr = NULL;
+	clear_bit(DASD_FLAG_EER_SNSS, &device->flags);
+	in_use = test_and_clear_bit(DASD_FLAG_EER_IN_USE, &device->flags);
+	spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
+	if (cqr && !in_use)
+		dasd_kfree_request(cqr, device);
+}
+
+/*
+ * SECTION: the device operations
+ */
+
+/*
+ * On the one side we need a lock to access our internal buffer, on the
+ * other side a copy_to_user can sleep. So we need to copy the data we have
+ * to transfer in a readbuffer, which is protected by the readbuffer_mutex.
+ */
+static char readbuffer[PAGE_SIZE];
+static DECLARE_MUTEX(readbuffer_mutex);
+
+static int dasd_eer_open(struct inode *inp, struct file *filp)
+{
+	struct eerbuffer *eerb;
+	unsigned long flags;
+
+	eerb = kzalloc(sizeof(struct eerbuffer), GFP_KERNEL);
+	if (!eerb)
+		return -ENOMEM;
+	eerb->buffer_page_count = eer_pages;
+	if (eerb->buffer_page_count < 1 ||
+	    eerb->buffer_page_count > INT_MAX / PAGE_SIZE) {
+		kfree(eerb);
+		MESSAGE(KERN_WARNING, "can't open device since module "
+			"parameter eer_pages is smaller then 1 or"
+			" bigger then %d", (int)(INT_MAX / PAGE_SIZE));
+		return -EINVAL;
+	}
+	eerb->buffersize = eerb->buffer_page_count * PAGE_SIZE;
+	eerb->buffer = kmalloc(eerb->buffer_page_count * sizeof(char *),
+			       GFP_KERNEL);
+        if (!eerb->buffer) {
+		kfree(eerb);
+                return -ENOMEM;
+	}
+	if (dasd_eer_allocate_buffer_pages(eerb->buffer,
+					   eerb->buffer_page_count)) {
+		kfree(eerb->buffer);
+		kfree(eerb);
+		return -ENOMEM;
+	}
+	filp->private_data = eerb;
+	spin_lock_irqsave(&bufferlock, flags);
+	list_add(&eerb->list, &bufferlist);
+	spin_unlock_irqrestore(&bufferlock, flags);
+
+	return nonseekable_open(inp,filp);
+}
+
+static int dasd_eer_close(struct inode *inp, struct file *filp)
+{
+	struct eerbuffer *eerb;
+	unsigned long flags;
+
+	eerb = (struct eerbuffer *) filp->private_data;
+	spin_lock_irqsave(&bufferlock, flags);
+	list_del(&eerb->list);
+	spin_unlock_irqrestore(&bufferlock, flags);
+	dasd_eer_free_buffer_pages(eerb->buffer, eerb->buffer_page_count);
+	kfree(eerb->buffer);
+	kfree(eerb);
+
+	return 0;
+}
+
+static ssize_t dasd_eer_read(struct file *filp, char __user *buf,
+			     size_t count, loff_t *ppos)
+{
+	int tc,rc;
+	int tailcount,effective_count;
+        unsigned long flags;
+	struct eerbuffer *eerb;
+
+	eerb = (struct eerbuffer *) filp->private_data;
+	if (down_interruptible(&readbuffer_mutex))
+		return -ERESTARTSYS;
+
+	spin_lock_irqsave(&bufferlock, flags);
+
+	if (eerb->residual < 0) { /* the remainder of this record */
+		                  /* has been deleted             */
+		eerb->residual = 0;
+		spin_unlock_irqrestore(&bufferlock, flags);
+		up(&readbuffer_mutex);
+		return -EIO;
+	} else if (eerb->residual > 0) {
+		/* OK we still have a second half of a record to deliver */
+		effective_count = min(eerb->residual, (int) count);
+		eerb->residual -= effective_count;
+	} else {
+		tc = 0;
+		while (!tc) {
+			tc = dasd_eer_read_buffer(eerb, (char *) &tailcount,
+						  sizeof(tailcount));
+			if (!tc) {
+				/* no data available */
+				spin_unlock_irqrestore(&bufferlock, flags);
+				up(&readbuffer_mutex);
+				if (filp->f_flags & O_NONBLOCK)
+					return -EAGAIN;
+				rc = wait_event_interruptible(
+					dasd_eer_read_wait_queue,
+					eerb->head != eerb->tail);
+				if (rc)
+					return rc;
+				if (down_interruptible(&readbuffer_mutex))
+					return -ERESTARTSYS;
+				spin_lock_irqsave(&bufferlock, flags);
+			}
+		}
+		WARN_ON(tc != sizeof(tailcount));
+		effective_count = min(tailcount,(int)count);
+		eerb->residual = tailcount - effective_count;
+	}
+
+	tc = dasd_eer_read_buffer(eerb, readbuffer, effective_count);
+	WARN_ON(tc != effective_count);
+
+	spin_unlock_irqrestore(&bufferlock, flags);
+
+	if (copy_to_user(buf, readbuffer, effective_count)) {
+		up(&readbuffer_mutex);
+		return -EFAULT;
+	}
+
+	up(&readbuffer_mutex);
+	return effective_count;
+}
+
+static unsigned int dasd_eer_poll(struct file *filp, poll_table *ptable)
+{
+	unsigned int mask;
+	unsigned long flags;
+	struct eerbuffer *eerb;
+
+	eerb = (struct eerbuffer *) filp->private_data;
+	poll_wait(filp, &dasd_eer_read_wait_queue, ptable);
+	spin_lock_irqsave(&bufferlock, flags);
+	if (eerb->head != eerb->tail)
+		mask = POLLIN | POLLRDNORM ;
+	else
+		mask = 0;
+	spin_unlock_irqrestore(&bufferlock, flags);
+	return mask;
+}
+
+static const struct file_operations dasd_eer_fops = {
+	.open		= &dasd_eer_open,
+	.release	= &dasd_eer_close,
+	.read		= &dasd_eer_read,
+	.poll		= &dasd_eer_poll,
+	.owner		= THIS_MODULE,
+};
+
+static struct miscdevice *dasd_eer_dev = NULL;
+
+int __init dasd_eer_init(void)
+{
+	int rc;
+
+	dasd_eer_dev = kzalloc(sizeof(*dasd_eer_dev), GFP_KERNEL);
+	if (!dasd_eer_dev)
+		return -ENOMEM;
+
+	dasd_eer_dev->minor = MISC_DYNAMIC_MINOR;
+	dasd_eer_dev->name  = "dasd_eer";
+	dasd_eer_dev->fops  = &dasd_eer_fops;
+
+	rc = misc_register(dasd_eer_dev);
+	if (rc) {
+		kfree(dasd_eer_dev);
+		dasd_eer_dev = NULL;
+		MESSAGE(KERN_ERR, "%s", "dasd_eer_init could not "
+		       "register misc device");
+		return rc;
+	}
+
+	return 0;
+}
+
+void dasd_eer_exit(void)
+{
+	if (dasd_eer_dev) {
+		WARN_ON(misc_deregister(dasd_eer_dev) != 0);
+		kfree(dasd_eer_dev);
+		dasd_eer_dev = NULL;
+	}
+}

drivers/s390/block/dasd_erp.c

@@ -0,0 +1,170 @@
+/*
+ * File...........: linux/drivers/s390/block/dasd.c
+ * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com>
+ *		    Horst Hummel <Horst.Hummel@de.ibm.com>
+ *		    Carsten Otte <Cotte@de.ibm.com>
+ *		    Martin Schwidefsky <schwidefsky@de.ibm.com>
+ * Bugreports.to..: <Linux390@de.ibm.com>
+ * (C) IBM Corporation, IBM Deutschland Entwicklung GmbH, 1999-2001
+ *
+ */
+
+#include <linux/ctype.h>
+#include <linux/init.h>
+
+#include <asm/debug.h>
+#include <asm/ebcdic.h>
+#include <asm/uaccess.h>
+
+/* This is ugly... */
+#define PRINTK_HEADER "dasd_erp:"
+
+#include "dasd_int.h"
+
+struct dasd_ccw_req *
+dasd_alloc_erp_request(char *magic, int cplength, int datasize,
+		       struct dasd_device * device)
+{
+	unsigned long flags;
+	struct dasd_ccw_req *cqr;
+	char *data;
+	int size;
+
+	/* Sanity checks */
+	BUG_ON( magic == NULL || datasize > PAGE_SIZE ||
+	     (cplength*sizeof(struct ccw1)) > PAGE_SIZE);
+
+	size = (sizeof(struct dasd_ccw_req) + 7L) & -8L;
+	if (cplength > 0)
+		size += cplength * sizeof(struct ccw1);
+	if (datasize > 0)
+		size += datasize;
+	spin_lock_irqsave(&device->mem_lock, flags);
+	cqr = (struct dasd_ccw_req *)
+		dasd_alloc_chunk(&device->erp_chunks, size);
+	spin_unlock_irqrestore(&device->mem_lock, flags);
+	if (cqr == NULL)
+		return ERR_PTR(-ENOMEM);
+	memset(cqr, 0, sizeof(struct dasd_ccw_req));
+	data = (char *) cqr + ((sizeof(struct dasd_ccw_req) + 7L) & -8L);
+	cqr->cpaddr = NULL;
+	if (cplength > 0) {
+		cqr->cpaddr = (struct ccw1 *) data;
+		data += cplength*sizeof(struct ccw1);
+		memset(cqr->cpaddr, 0, cplength*sizeof(struct ccw1));
+	}
+	cqr->data = NULL;
+	if (datasize > 0) {
+		cqr->data = data;
+ 		memset(cqr->data, 0, datasize);
+	}
+	strncpy((char *) &cqr->magic, magic, 4);
+	ASCEBC((char *) &cqr->magic, 4);
+	set_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags);
+	dasd_get_device(device);
+	return cqr;
+}
+
+void
+dasd_free_erp_request(struct dasd_ccw_req * cqr, struct dasd_device * device)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&device->mem_lock, flags);
+	dasd_free_chunk(&device->erp_chunks, cqr);
+	spin_unlock_irqrestore(&device->mem_lock, flags);
+	atomic_dec(&device->ref_count);
+}
+
+
+/*
+ * dasd_default_erp_action just retries the current cqr
+ */
+struct dasd_ccw_req *
+dasd_default_erp_action(struct dasd_ccw_req * cqr)
+{
+	struct dasd_device *device;
+
+	device = cqr->device;
+
+        /* just retry - there is nothing to save ... I got no sense data.... */
+        if (cqr->retries > 0) {
+		DEV_MESSAGE (KERN_DEBUG, device,
+                             "default ERP called (%i retries left)",
+                             cqr->retries);
+		cqr->lpm    = LPM_ANYPATH;
+		cqr->status = DASD_CQR_QUEUED;
+        } else {
+                DEV_MESSAGE (KERN_WARNING, device, "%s",
+			     "default ERP called (NO retry left)");
+		cqr->status = DASD_CQR_FAILED;
+		cqr->stopclk = get_clock ();
+        }
+        return cqr;
+}				/* end dasd_default_erp_action */
+
+/*
+ * DESCRIPTION
+ *   Frees all ERPs of the current ERP Chain and set the status
+ *   of the original CQR either to DASD_CQR_DONE if ERP was successful
+ *   or to DASD_CQR_FAILED if ERP was NOT successful.
+ *   NOTE: This function is only called if no discipline postaction
+ *	   is available
+ *
+ * PARAMETER
+ *   erp		current erp_head
+ *
+ * RETURN VALUES
+ *   cqr		pointer to the original CQR
+ */
+struct dasd_ccw_req *
+dasd_default_erp_postaction(struct dasd_ccw_req * cqr)
+{
+	struct dasd_device *device;
+	int success;
+
+	BUG_ON(cqr->refers == NULL || cqr->function == NULL);
+
+	device = cqr->device;
+	success = cqr->status == DASD_CQR_DONE;
+
+	/* free all ERPs - but NOT the original cqr */
+	while (cqr->refers != NULL) {
+		struct dasd_ccw_req *refers;
+
+		refers = cqr->refers;
+		/* remove the request from the device queue */
+		list_del(&cqr->list);
+		/* free the finished erp request */
+		dasd_free_erp_request(cqr, device);
+		cqr = refers;
+	}
+
+	/* set corresponding status to original cqr */
+	if (success)
+		cqr->status = DASD_CQR_DONE;
+	else {
+		cqr->status = DASD_CQR_FAILED;
+		cqr->stopclk = get_clock();
+	}
+
+	return cqr;
+
+}				/* end default_erp_postaction */
+
+void
+dasd_log_sense(struct dasd_ccw_req *cqr, struct irb *irb)
+{
+	struct dasd_device *device;
+
+	device = cqr->device;
+	/* dump sense data */
+	if (device->discipline && device->discipline->dump_sense)
+		device->discipline->dump_sense(device, cqr, irb);
+}
+
+EXPORT_SYMBOL(dasd_default_erp_action);
+EXPORT_SYMBOL(dasd_default_erp_postaction);
+EXPORT_SYMBOL(dasd_alloc_erp_request);
+EXPORT_SYMBOL(dasd_free_erp_request);
+EXPORT_SYMBOL(dasd_log_sense);

drivers/s390/block/dasd_fba.c

@@ -0,0 +1,576 @@
+/*
+ * File...........: linux/drivers/s390/block/dasd_fba.c
+ * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com>
+ * Bugreports.to..: <Linux390@de.ibm.com>
+ * (C) IBM Corporation, IBM Deutschland Entwicklung GmbH, 1999,2000
+ *
+ */
+
+#include <linux/stddef.h>
+#include <linux/kernel.h>
+#include <asm/debug.h>
+
+#include <linux/slab.h>
+#include <linux/hdreg.h>	/* HDIO_GETGEO			    */
+#include <linux/bio.h>
+#include <linux/module.h>
+#include <linux/init.h>
+
+#include <asm/idals.h>
+#include <asm/ebcdic.h>
+#include <asm/io.h>
+#include <asm/todclk.h>
+#include <asm/ccwdev.h>
+
+#include "dasd_int.h"
+#include "dasd_fba.h"
+
+#ifdef PRINTK_HEADER
+#undef PRINTK_HEADER
+#endif				/* PRINTK_HEADER */
+#define PRINTK_HEADER "dasd(fba):"
+
+#define DASD_FBA_CCW_WRITE 0x41
+#define DASD_FBA_CCW_READ 0x42
+#define DASD_FBA_CCW_LOCATE 0x43
+#define DASD_FBA_CCW_DEFINE_EXTENT 0x63
+
+MODULE_LICENSE("GPL");
+
+static struct dasd_discipline dasd_fba_discipline;
+
+struct dasd_fba_private {
+	struct dasd_fba_characteristics rdc_data;
+};
+
+static struct ccw_device_id dasd_fba_ids[] = {
+	{ CCW_DEVICE_DEVTYPE (0x6310, 0, 0x9336, 0), .driver_info = 0x1},
+	{ CCW_DEVICE_DEVTYPE (0x3880, 0, 0x3370, 0), .driver_info = 0x2},
+	{ /* end of list */ },
+};
+
+MODULE_DEVICE_TABLE(ccw, dasd_fba_ids);
+
+static struct ccw_driver dasd_fba_driver; /* see below */
+static int
+dasd_fba_probe(struct ccw_device *cdev)
+{
+	return dasd_generic_probe(cdev, &dasd_fba_discipline);
+}
+
+static int
+dasd_fba_set_online(struct ccw_device *cdev)
+{
+	return dasd_generic_set_online(cdev, &dasd_fba_discipline);
+}
+
+static struct ccw_driver dasd_fba_driver = {
+	.name        = "dasd-fba",
+	.owner       = THIS_MODULE,
+	.ids         = dasd_fba_ids,
+	.probe       = dasd_fba_probe,
+	.remove      = dasd_generic_remove,
+	.set_offline = dasd_generic_set_offline,
+	.set_online  = dasd_fba_set_online,
+	.notify      = dasd_generic_notify,
+};
+
+static void
+define_extent(struct ccw1 * ccw, struct DE_fba_data *data, int rw,
+	      int blksize, int beg, int nr)
+{
+	ccw->cmd_code = DASD_FBA_CCW_DEFINE_EXTENT;
+	ccw->flags = 0;
+	ccw->count = 16;
+	ccw->cda = (__u32) __pa(data);
+	memset(data, 0, sizeof (struct DE_fba_data));
+	if (rw == WRITE)
+		(data->mask).perm = 0x0;
+	else if (rw == READ)
+		(data->mask).perm = 0x1;
+	else
+		data->mask.perm = 0x2;
+	data->blk_size = blksize;
+	data->ext_loc = beg;
+	data->ext_end = nr - 1;
+}
+
+static void
+locate_record(struct ccw1 * ccw, struct LO_fba_data *data, int rw,
+	      int block_nr, int block_ct)
+{
+	ccw->cmd_code = DASD_FBA_CCW_LOCATE;
+	ccw->flags = 0;
+	ccw->count = 8;
+	ccw->cda = (__u32) __pa(data);
+	memset(data, 0, sizeof (struct LO_fba_data));
+	if (rw == WRITE)
+		data->operation.cmd = 0x5;
+	else if (rw == READ)
+		data->operation.cmd = 0x6;
+	else
+		data->operation.cmd = 0x8;
+	data->blk_nr = block_nr;
+	data->blk_ct = block_ct;
+}
+
+static int
+dasd_fba_check_characteristics(struct dasd_device *device)
+{
+	struct dasd_fba_private *private;
+	struct ccw_device *cdev = device->cdev;
+	void *rdc_data;
+	int rc;
+
+	private = (struct dasd_fba_private *) device->private;
+	if (private == NULL) {
+		private = kzalloc(sizeof(struct dasd_fba_private), GFP_KERNEL);
+		if (private == NULL) {
+			DEV_MESSAGE(KERN_WARNING, device, "%s",
+				    "memory allocation failed for private "
+				    "data");
+			return -ENOMEM;
+		}
+		device->private = (void *) private;
+	}
+	/* Read Device Characteristics */
+	rdc_data = (void *) &(private->rdc_data);
+	rc = read_dev_chars(device->cdev, &rdc_data, 32);
+	if (rc) {
+		DEV_MESSAGE(KERN_WARNING, device,
+			    "Read device characteristics returned error %d",
+			    rc);
+		return rc;
+	}
+
+	DEV_MESSAGE(KERN_INFO, device,
+		    "%04X/%02X(CU:%04X/%02X) %dMB at(%d B/blk)",
+		    cdev->id.dev_type,
+		    cdev->id.dev_model,
+		    cdev->id.cu_type,
+		    cdev->id.cu_model,
+		    ((private->rdc_data.blk_bdsa *
+		      (private->rdc_data.blk_size >> 9)) >> 11),
+		    private->rdc_data.blk_size);
+	return 0;
+}
+
+static int
+dasd_fba_do_analysis(struct dasd_device *device)
+{
+	struct dasd_fba_private *private;
+	int sb, rc;
+
+	private = (struct dasd_fba_private *) device->private;
+	rc = dasd_check_blocksize(private->rdc_data.blk_size);
+	if (rc) {
+		DEV_MESSAGE(KERN_INFO, device, "unknown blocksize %d",
+			    private->rdc_data.blk_size);
+		return rc;
+	}
+	device->blocks = private->rdc_data.blk_bdsa;
+	device->bp_block = private->rdc_data.blk_size;
+	device->s2b_shift = 0;	/* bits to shift 512 to get a block */
+	for (sb = 512; sb < private->rdc_data.blk_size; sb = sb << 1)
+		device->s2b_shift++;
+	return 0;
+}
+
+static int
+dasd_fba_fill_geometry(struct dasd_device *device, struct hd_geometry *geo)
+{
+	if (dasd_check_blocksize(device->bp_block) != 0)
+		return -EINVAL;
+	geo->cylinders = (device->blocks << device->s2b_shift) >> 10;
+	geo->heads = 16;
+	geo->sectors = 128 >> device->s2b_shift;
+	return 0;
+}
+
+static dasd_era_t
+dasd_fba_examine_error(struct dasd_ccw_req * cqr, struct irb * irb)
+{
+	struct dasd_device *device;
+	struct ccw_device *cdev;
+
+	device = (struct dasd_device *) cqr->device;
+	if (irb->scsw.cstat == 0x00 &&
+	    irb->scsw.dstat == (DEV_STAT_CHN_END | DEV_STAT_DEV_END))
+		return dasd_era_none;
+
+	cdev = device->cdev;
+	switch (cdev->id.dev_type) {
+	case 0x3370:
+		return dasd_3370_erp_examine(cqr, irb);
+	case 0x9336:
+		return dasd_9336_erp_examine(cqr, irb);
+	default:
+		return dasd_era_recover;
+	}
+}
+
+static dasd_erp_fn_t
+dasd_fba_erp_action(struct dasd_ccw_req * cqr)
+{
+	return dasd_default_erp_action;
+}
+
+static dasd_erp_fn_t
+dasd_fba_erp_postaction(struct dasd_ccw_req * cqr)
+{
+	if (cqr->function == dasd_default_erp_action)
+		return dasd_default_erp_postaction;
+
+	DEV_MESSAGE(KERN_WARNING, cqr->device, "unknown ERP action %p",
+		    cqr->function);
+	return NULL;
+}
+
+static struct dasd_ccw_req *
+dasd_fba_build_cp(struct dasd_device * device, struct request *req)
+{
+	struct dasd_fba_private *private;
+	unsigned long *idaws;
+	struct LO_fba_data *LO_data;
+	struct dasd_ccw_req *cqr;
+	struct ccw1 *ccw;
+	struct bio *bio;
+	struct bio_vec *bv;
+	char *dst;
+	int count, cidaw, cplength, datasize;
+	sector_t recid, first_rec, last_rec;
+	unsigned int blksize, off;
+	unsigned char cmd;
+	int i;
+
+	private = (struct dasd_fba_private *) device->private;
+	if (rq_data_dir(req) == READ) {
+		cmd = DASD_FBA_CCW_READ;
+	} else if (rq_data_dir(req) == WRITE) {
+		cmd = DASD_FBA_CCW_WRITE;
+	} else
+		return ERR_PTR(-EINVAL);
+	blksize = device->bp_block;
+	/* Calculate record id of first and last block. */
+	first_rec = req->sector >> device->s2b_shift;
+	last_rec = (req->sector + req->nr_sectors - 1) >> device->s2b_shift;
+	/* Check struct bio and count the number of blocks for the request. */
+	count = 0;
+	cidaw = 0;
+	rq_for_each_bio(bio, req) {
+		bio_for_each_segment(bv, bio, i) {
+			if (bv->bv_len & (blksize - 1))
+				/* Fba can only do full blocks. */
+				return ERR_PTR(-EINVAL);
+			count += bv->bv_len >> (device->s2b_shift + 9);
+#if defined(CONFIG_64BIT)
+			if (idal_is_needed (page_address(bv->bv_page),
+					    bv->bv_len))
+				cidaw += bv->bv_len / blksize;
+#endif
+		}
+	}
+	/* Paranoia. */
+	if (count != last_rec - first_rec + 1)
+		return ERR_PTR(-EINVAL);
+	/* 1x define extent + 1x locate record + number of blocks */
+	cplength = 2 + count;
+	/* 1x define extent + 1x locate record */
+	datasize = sizeof(struct DE_fba_data) + sizeof(struct LO_fba_data) +
+		cidaw * sizeof(unsigned long);
+	/*
+	 * Find out number of additional locate record ccws if the device
+	 * can't do data chaining.
+	 */
+	if (private->rdc_data.mode.bits.data_chain == 0) {
+		cplength += count - 1;
+		datasize += (count - 1)*sizeof(struct LO_fba_data);
+	}
+	/* Allocate the ccw request. */
+	cqr = dasd_smalloc_request(dasd_fba_discipline.name,
+				   cplength, datasize, device);
+	if (IS_ERR(cqr))
+		return cqr;
+	ccw = cqr->cpaddr;
+	/* First ccw is define extent. */
+	define_extent(ccw++, cqr->data, rq_data_dir(req),
+		      device->bp_block, req->sector, req->nr_sectors);
+	/* Build locate_record + read/write ccws. */
+	idaws = (unsigned long *) (cqr->data + sizeof(struct DE_fba_data));
+	LO_data = (struct LO_fba_data *) (idaws + cidaw);
+	/* Locate record for all blocks for smart devices. */
+	if (private->rdc_data.mode.bits.data_chain != 0) {
+		ccw[-1].flags |= CCW_FLAG_CC;
+		locate_record(ccw++, LO_data++, rq_data_dir(req), 0, count);
+	}
+	recid = first_rec;
+	rq_for_each_bio(bio, req) bio_for_each_segment(bv, bio, i) {
+		dst = page_address(bv->bv_page) + bv->bv_offset;
+		if (dasd_page_cache) {
+			char *copy = kmem_cache_alloc(dasd_page_cache,
+						      GFP_DMA | __GFP_NOWARN);
+			if (copy && rq_data_dir(req) == WRITE)
+				memcpy(copy + bv->bv_offset, dst, bv->bv_len);
+			if (copy)
+				dst = copy + bv->bv_offset;
+		}
+		for (off = 0; off < bv->bv_len; off += blksize) {
+			/* Locate record for stupid devices. */
+			if (private->rdc_data.mode.bits.data_chain == 0) {
+				ccw[-1].flags |= CCW_FLAG_CC;
+				locate_record(ccw, LO_data++,
+					      rq_data_dir(req),
+					      recid - first_rec, 1);
+				ccw->flags = CCW_FLAG_CC;
+				ccw++;
+			} else {
+				if (recid > first_rec)
+					ccw[-1].flags |= CCW_FLAG_DC;
+				else
+					ccw[-1].flags |= CCW_FLAG_CC;
+			}
+			ccw->cmd_code = cmd;
+			ccw->count = device->bp_block;
+			if (idal_is_needed(dst, blksize)) {
+				ccw->cda = (__u32)(addr_t) idaws;
+				ccw->flags = CCW_FLAG_IDA;
+				idaws = idal_create_words(idaws, dst, blksize);
+			} else {
+				ccw->cda = (__u32)(addr_t) dst;
+				ccw->flags = 0;
+			}
+			ccw++;
+			dst += blksize;
+			recid++;
+		}
+	}
+	if (req->cmd_flags & REQ_FAILFAST)
+		set_bit(DASD_CQR_FLAGS_FAILFAST, &cqr->flags);
+	cqr->device = device;
+	cqr->expires = 5 * 60 * HZ;	/* 5 minutes */
+	cqr->retries = 32;
+	cqr->buildclk = get_clock();
+	cqr->status = DASD_CQR_FILLED;
+	return cqr;
+}
+
+static int
+dasd_fba_free_cp(struct dasd_ccw_req *cqr, struct request *req)
+{
+	struct dasd_fba_private *private;
+	struct ccw1 *ccw;
+	struct bio *bio;
+	struct bio_vec *bv;
+	char *dst, *cda;
+	unsigned int blksize, off;
+	int i, status;
+
+	if (!dasd_page_cache)
+		goto out;
+	private = (struct dasd_fba_private *) cqr->device->private;
+	blksize = cqr->device->bp_block;
+	ccw = cqr->cpaddr;
+	/* Skip over define extent & locate record. */
+	ccw++;
+	if (private->rdc_data.mode.bits.data_chain != 0)
+		ccw++;
+	rq_for_each_bio(bio, req) bio_for_each_segment(bv, bio, i) {
+		dst = page_address(bv->bv_page) + bv->bv_offset;
+		for (off = 0; off < bv->bv_len; off += blksize) {
+			/* Skip locate record. */
+			if (private->rdc_data.mode.bits.data_chain == 0)
+				ccw++;
+			if (dst) {
+				if (ccw->flags & CCW_FLAG_IDA)
+					cda = *((char **)((addr_t) ccw->cda));
+				else
+					cda = (char *)((addr_t) ccw->cda);
+				if (dst != cda) {
+					if (rq_data_dir(req) == READ)
+						memcpy(dst, cda, bv->bv_len);
+					kmem_cache_free(dasd_page_cache,
+					    (void *)((addr_t)cda & PAGE_MASK));
+				}
+				dst = NULL;
+			}
+			ccw++;
+		}
+	}
+out:
+	status = cqr->status == DASD_CQR_DONE;
+	dasd_sfree_request(cqr, cqr->device);
+	return status;
+}
+
+static int
+dasd_fba_fill_info(struct dasd_device * device,
+		   struct dasd_information2_t * info)
+{
+	info->label_block = 1;
+	info->FBA_layout = 1;
+	info->format = DASD_FORMAT_LDL;
+	info->characteristics_size = sizeof(struct dasd_fba_characteristics);
+	memcpy(info->characteristics,
+	       &((struct dasd_fba_private *) device->private)->rdc_data,
+	       sizeof (struct dasd_fba_characteristics));
+	info->confdata_size = 0;
+	return 0;
+}
+
+static void
+dasd_fba_dump_sense(struct dasd_device *device, struct dasd_ccw_req * req,
+		    struct irb *irb)
+{
+	char *page;
+	struct ccw1 *act, *end, *last;
+	int len, sl, sct, count;
+
+	page = (char *) get_zeroed_page(GFP_ATOMIC);
+	if (page == NULL) {
+		DEV_MESSAGE(KERN_ERR, device, " %s",
+			    "No memory to dump sense data");
+		return;
+	}
+	len = sprintf(page, KERN_ERR PRINTK_HEADER
+		      " I/O status report for device %s:\n",
+		      device->cdev->dev.bus_id);
+	len += sprintf(page + len, KERN_ERR PRINTK_HEADER
+		       " in req: %p CS: 0x%02X DS: 0x%02X\n", req,
+		       irb->scsw.cstat, irb->scsw.dstat);
+	len += sprintf(page + len, KERN_ERR PRINTK_HEADER
+		       " device %s: Failing CCW: %p\n",
+		       device->cdev->dev.bus_id,
+		       (void *) (addr_t) irb->scsw.cpa);
+	if (irb->esw.esw0.erw.cons) {
+		for (sl = 0; sl < 4; sl++) {
+			len += sprintf(page + len, KERN_ERR PRINTK_HEADER
+				       " Sense(hex) %2d-%2d:",
+				       (8 * sl), ((8 * sl) + 7));
+
+			for (sct = 0; sct < 8; sct++) {
+				len += sprintf(page + len, " %02x",
+					       irb->ecw[8 * sl + sct]);
+			}
+			len += sprintf(page + len, "\n");
+		}
+	} else {
+	        len += sprintf(page + len, KERN_ERR PRINTK_HEADER
+			       " SORRY - NO VALID SENSE AVAILABLE\n");
+	}
+	MESSAGE_LOG(KERN_ERR, "%s",
+		    page + sizeof(KERN_ERR PRINTK_HEADER));
+
+	/* dump the Channel Program */
+	/* print first CCWs (maximum 8) */
+	act = req->cpaddr;
+        for (last = act; last->flags & (CCW_FLAG_CC | CCW_FLAG_DC); last++);
+	end = min(act + 8, last);
+	len = sprintf(page, KERN_ERR PRINTK_HEADER
+		      " Related CP in req: %p\n", req);
+	while (act <= end) {
+		len += sprintf(page + len, KERN_ERR PRINTK_HEADER
+			       " CCW %p: %08X %08X DAT:",
+			       act, ((int *) act)[0], ((int *) act)[1]);
+		for (count = 0; count < 32 && count < act->count;
+		     count += sizeof(int))
+			len += sprintf(page + len, " %08X",
+				       ((int *) (addr_t) act->cda)
+				       [(count>>2)]);
+		len += sprintf(page + len, "\n");
+		act++;
+	}
+	MESSAGE_LOG(KERN_ERR, "%s",
+		    page + sizeof(KERN_ERR PRINTK_HEADER));
+
+
+	/* print failing CCW area */
+	len = 0;
+	if (act <  ((struct ccw1 *)(addr_t) irb->scsw.cpa) - 2) {
+		act = ((struct ccw1 *)(addr_t) irb->scsw.cpa) - 2;
+		len += sprintf(page + len, KERN_ERR PRINTK_HEADER "......\n");
+	}
+	end = min((struct ccw1 *)(addr_t) irb->scsw.cpa + 2, last);
+	while (act <= end) {
+		len += sprintf(page + len, KERN_ERR PRINTK_HEADER
+			       " CCW %p: %08X %08X DAT:",
+			       act, ((int *) act)[0], ((int *) act)[1]);
+		for (count = 0; count < 32 && count < act->count;
+		     count += sizeof(int))
+			len += sprintf(page + len, " %08X",
+				       ((int *) (addr_t) act->cda)
+				       [(count>>2)]);
+		len += sprintf(page + len, "\n");
+		act++;
+	}
+
+	/* print last CCWs */
+	if (act <  last - 2) {
+		act = last - 2;
+		len += sprintf(page + len, KERN_ERR PRINTK_HEADER "......\n");
+	}
+	while (act <= last) {
+		len += sprintf(page + len, KERN_ERR PRINTK_HEADER
+			       " CCW %p: %08X %08X DAT:",
+			       act, ((int *) act)[0], ((int *) act)[1]);
+		for (count = 0; count < 32 && count < act->count;
+		     count += sizeof(int))
+			len += sprintf(page + len, " %08X",
+				       ((int *) (addr_t) act->cda)
+				       [(count>>2)]);
+		len += sprintf(page + len, "\n");
+		act++;
+	}
+	if (len > 0)
+		MESSAGE_LOG(KERN_ERR, "%s",
+			    page + sizeof(KERN_ERR PRINTK_HEADER));
+	free_page((unsigned long) page);
+}
+
+/*
+ * max_blocks is dependent on the amount of storage that is available
+ * in the static io buffer for each device. Currently each device has
+ * 8192 bytes (=2 pages). For 64 bit one dasd_mchunkt_t structure has
+ * 24 bytes, the struct dasd_ccw_req has 136 bytes and each block can use
+ * up to 16 bytes (8 for the ccw and 8 for the idal pointer). In
+ * addition we have one define extent ccw + 16 bytes of data and a
+ * locate record ccw for each block (stupid devices!) + 16 bytes of data.
+ * That makes:
+ * (8192 - 24 - 136 - 8 - 16) / 40 = 200.2 blocks at maximum.
+ * We want to fit two into the available memory so that we can immediately
+ * start the next request if one finishes off. That makes 100.1 blocks
+ * for one request. Give a little safety and the result is 96.
+ */
+static struct dasd_discipline dasd_fba_discipline = {
+	.owner = THIS_MODULE,
+	.name = "FBA ",
+	.ebcname = "FBA ",
+	.max_blocks = 96,
+	.check_device = dasd_fba_check_characteristics,
+	.do_analysis = dasd_fba_do_analysis,
+	.fill_geometry = dasd_fba_fill_geometry,
+	.start_IO = dasd_start_IO,
+	.term_IO = dasd_term_IO,
+	.examine_error = dasd_fba_examine_error,
+	.erp_action = dasd_fba_erp_action,
+	.erp_postaction = dasd_fba_erp_postaction,
+	.build_cp = dasd_fba_build_cp,
+	.free_cp = dasd_fba_free_cp,
+	.dump_sense = dasd_fba_dump_sense,
+	.fill_info = dasd_fba_fill_info,
+};
+
+static int __init
+dasd_fba_init(void)
+{
+	ASCEBC(dasd_fba_discipline.ebcname, 4);
+	return ccw_driver_register(&dasd_fba_driver);
+}
+
+static void __exit
+dasd_fba_cleanup(void)
+{
+	ccw_driver_unregister(&dasd_fba_driver);
+}
+
+module_init(dasd_fba_init);
+module_exit(dasd_fba_cleanup);

drivers/s390/block/dasd_fba.h

@@ -0,0 +1,72 @@
+/*
+ * File...........: linux/drivers/s390/block/dasd_fba.h
+ * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com>
+ * Bugreports.to..: <Linux390@de.ibm.com>
+ * (C) IBM Corporation, IBM Deutschland Entwicklung GmbH, 1999,2000
+ *
+ */
+
+#ifndef DASD_FBA_H
+#define DASD_FBA_H
+
+struct DE_fba_data {
+	struct {
+		unsigned char perm:2;	/* Permissions on this extent */
+		unsigned char zero:2;	/* Must be zero */
+		unsigned char da:1;	/* usually zero */
+		unsigned char diag:1;	/* allow diagnose */
+		unsigned char zero2:2;	/* zero */
+	} __attribute__ ((packed)) mask;
+	__u8 zero;		/* Must be zero */
+	__u16 blk_size;		/* Blocksize */
+	__u32 ext_loc;		/* Extent locator */
+	__u32 ext_beg;		/* logical number of block 0 in extent */
+	__u32 ext_end;		/* logocal number of last block in extent */
+} __attribute__ ((packed));
+
+struct LO_fba_data {
+	struct {
+		unsigned char zero:4;
+		unsigned char cmd:4;
+	} __attribute__ ((packed)) operation;
+	__u8 auxiliary;
+	__u16 blk_ct;
+	__u32 blk_nr;
+} __attribute__ ((packed));
+
+struct dasd_fba_characteristics {
+	union {
+		__u8 c;
+		struct {
+			unsigned char reserved:1;
+			unsigned char overrunnable:1;
+			unsigned char burst_byte:1;
+			unsigned char data_chain:1;
+			unsigned char zeros:4;
+		} __attribute__ ((packed)) bits;
+	} __attribute__ ((packed)) mode;
+	union {
+		__u8 c;
+		struct {
+			unsigned char zero0:1;
+			unsigned char removable:1;
+			unsigned char shared:1;
+			unsigned char zero1:1;
+			unsigned char mam:1;
+			unsigned char zeros:3;
+		} __attribute__ ((packed)) bits;
+	} __attribute__ ((packed)) features;
+	__u8 dev_class;
+	__u8 unit_type;
+	__u16 blk_size;
+	__u32 blk_per_cycl;
+	__u32 blk_per_bound;
+	__u32 blk_bdsa;
+	__u32 reserved0;
+	__u16 reserved1;
+	__u16 blk_ce;
+	__u32 reserved2;
+	__u16 reserved3;
+} __attribute__ ((packed));
+
+#endif				/* DASD_FBA_H */

drivers/s390/block/dasd_genhd.c

@@ -0,0 +1,181 @@
+/*
+ * File...........: linux/drivers/s390/block/dasd_genhd.c
+ * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com>
+ *		    Horst Hummel <Horst.Hummel@de.ibm.com>
+ *		    Carsten Otte <Cotte@de.ibm.com>
+ *		    Martin Schwidefsky <schwidefsky@de.ibm.com>
+ * Bugreports.to..: <Linux390@de.ibm.com>
+ * (C) IBM Corporation, IBM Deutschland Entwicklung GmbH, 1999-2001
+ *
+ * gendisk related functions for the dasd driver.
+ *
+ */
+
+#include <linux/interrupt.h>
+#include <linux/fs.h>
+#include <linux/blkpg.h>
+
+#include <asm/uaccess.h>
+
+/* This is ugly... */
+#define PRINTK_HEADER "dasd_gendisk:"
+
+#include "dasd_int.h"
+
+/*
+ * Allocate and register gendisk structure for device.
+ */
+int
+dasd_gendisk_alloc(struct dasd_device *device)
+{
+	struct gendisk *gdp;
+	int len;
+
+	/* Make sure the minor for this device exists. */
+	if (device->devindex >= DASD_PER_MAJOR)
+		return -EBUSY;
+
+	gdp = alloc_disk(1 << DASD_PARTN_BITS);
+	if (!gdp)
+		return -ENOMEM;
+
+	/* Initialize gendisk structure. */
+	gdp->major = DASD_MAJOR;
+	gdp->first_minor = device->devindex << DASD_PARTN_BITS;
+	gdp->fops = &dasd_device_operations;
+	gdp->driverfs_dev = &device->cdev->dev;
+
+	/*
+	 * Set device name.
+	 *   dasda - dasdz : 26 devices
+	 *   dasdaa - dasdzz : 676 devices, added up = 702
+	 *   dasdaaa - dasdzzz : 17576 devices, added up = 18278
+	 *   dasdaaaa - dasdzzzz : 456976 devices, added up = 475252
+	 */
+	len = sprintf(gdp->disk_name, "dasd");
+	if (device->devindex > 25) {
+	        if (device->devindex > 701) {
+		        if (device->devindex > 18277)
+			        len += sprintf(gdp->disk_name + len, "%c",
+					       'a'+(((device->devindex-18278)
+						     /17576)%26));
+			len += sprintf(gdp->disk_name + len, "%c",
+				       'a'+(((device->devindex-702)/676)%26));
+		}
+		len += sprintf(gdp->disk_name + len, "%c",
+			       'a'+(((device->devindex-26)/26)%26));
+	}
+	len += sprintf(gdp->disk_name + len, "%c", 'a'+(device->devindex%26));
+
+	if (device->features & DASD_FEATURE_READONLY)
+		set_disk_ro(gdp, 1);
+	gdp->private_data = device;
+	gdp->queue = device->request_queue;
+	device->gdp = gdp;
+	set_capacity(device->gdp, 0);
+	add_disk(device->gdp);
+	return 0;
+}
+
+/*
+ * Unregister and free gendisk structure for device.
+ */
+void
+dasd_gendisk_free(struct dasd_device *device)
+{
+	if (device->gdp) {
+		del_gendisk(device->gdp);
+		device->gdp->queue = NULL;
+		put_disk(device->gdp);
+		device->gdp = NULL;
+	}
+}
+
+/*
+ * Trigger a partition detection.
+ */
+int
+dasd_scan_partitions(struct dasd_device * device)
+{
+	struct block_device *bdev;
+
+	bdev = bdget_disk(device->gdp, 0);
+	if (!bdev || blkdev_get(bdev, FMODE_READ, 1) < 0)
+		return -ENODEV;
+	/*
+	 * See fs/partition/check.c:register_disk,rescan_partitions
+	 * Can't call rescan_partitions directly. Use ioctl.
+	 */
+	ioctl_by_bdev(bdev, BLKRRPART, 0);
+	/*
+	 * Since the matching blkdev_put call to the blkdev_get in
+	 * this function is not called before dasd_destroy_partitions
+	 * the offline open_count limit needs to be increased from
+	 * 0 to 1. This is done by setting device->bdev (see
+	 * dasd_generic_set_offline). As long as the partition
+	 * detection is running no offline should be allowed. That
+	 * is why the assignment to device->bdev is done AFTER
+	 * the BLKRRPART ioctl.
+	 */
+	device->bdev = bdev;
+	return 0;
+}
+
+/*
+ * Remove all inodes in the system for a device, delete the
+ * partitions and make device unusable by setting its size to zero.
+ */
+void
+dasd_destroy_partitions(struct dasd_device * device)
+{
+	/* The two structs have 168/176 byte on 31/64 bit. */
+	struct blkpg_partition bpart;
+	struct blkpg_ioctl_arg barg;
+	struct block_device *bdev;
+
+	/*
+	 * Get the bdev pointer from the device structure and clear
+	 * device->bdev to lower the offline open_count limit again.
+	 */
+	bdev = device->bdev;
+	device->bdev = NULL;
+
+	/*
+	 * See fs/partition/check.c:delete_partition
+	 * Can't call delete_partitions directly. Use ioctl.
+	 * The ioctl also does locking and invalidation.
+	 */
+	memset(&bpart, 0, sizeof(struct blkpg_partition));
+	memset(&barg, 0, sizeof(struct blkpg_ioctl_arg));
+	barg.data = (void __force __user *) &bpart;
+	barg.op = BLKPG_DEL_PARTITION;
+	for (bpart.pno = device->gdp->minors - 1; bpart.pno > 0; bpart.pno--)
+		ioctl_by_bdev(bdev, BLKPG, (unsigned long) &barg);
+
+	invalidate_partition(device->gdp, 0);
+	/* Matching blkdev_put to the blkdev_get in dasd_scan_partitions. */
+	blkdev_put(bdev);
+	set_capacity(device->gdp, 0);
+}
+
+int
+dasd_gendisk_init(void)
+{
+	int rc;
+
+	/* Register to static dasd major 94 */
+	rc = register_blkdev(DASD_MAJOR, "dasd");
+	if (rc != 0) {
+		MESSAGE(KERN_WARNING,
+			"Couldn't register successfully to "
+			"major no %d", DASD_MAJOR);
+		return rc;
+	}
+	return 0;
+}
+
+void
+dasd_gendisk_exit(void)
+{
+	unregister_blkdev(DASD_MAJOR, "dasd");
+}

drivers/s390/block/dasd_int.h

@@ -0,0 +1,603 @@
+/*
+ * File...........: linux/drivers/s390/block/dasd_int.h
+ * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com>
+ *		    Horst Hummel <Horst.Hummel@de.ibm.com>
+ *		    Martin Schwidefsky <schwidefsky@de.ibm.com>
+ * Bugreports.to..: <Linux390@de.ibm.com>
+ * (C) IBM Corporation, IBM Deutschland Entwicklung GmbH, 1999,2000
+ *
+ */
+
+#ifndef DASD_INT_H
+#define DASD_INT_H
+
+#ifdef __KERNEL__
+
+/* we keep old device allocation scheme; IOW, minors are still in 0..255 */
+#define DASD_PER_MAJOR (1U << (MINORBITS - DASD_PARTN_BITS))
+#define DASD_PARTN_MASK ((1 << DASD_PARTN_BITS) - 1)
+
+/*
+ * States a dasd device can have:
+ *   new: the dasd_device structure is allocated.
+ *   known: the discipline for the device is identified.
+ *   basic: the device can do basic i/o.
+ *   unfmt: the device could not be analyzed (format is unknown).
+ *   ready: partition detection is done and the device is can do block io.
+ *   online: the device accepts requests from the block device queue.
+ *
+ * Things to do for startup state transitions:
+ *   new -> known: find discipline for the device and create devfs entries.
+ *   known -> basic: request irq line for the device.
+ *   basic -> ready: do the initial analysis, e.g. format detection,
+ *                   do block device setup and detect partitions.
+ *   ready -> online: schedule the device tasklet.
+ * Things to do for shutdown state transitions:
+ *   online -> ready: just set the new device state.
+ *   ready -> basic: flush requests from the block device layer, clear
+ *                   partition information and reset format information.
+ *   basic -> known: terminate all requests and free irq.
+ *   known -> new: remove devfs entries and forget discipline.
+ */
+
+#define DASD_STATE_NEW	  0
+#define DASD_STATE_KNOWN  1
+#define DASD_STATE_BASIC  2
+#define DASD_STATE_UNFMT  3
+#define DASD_STATE_READY  4
+#define DASD_STATE_ONLINE 5
+
+#include <linux/module.h>
+#include <linux/wait.h>
+#include <linux/blkdev.h>
+#include <linux/genhd.h>
+#include <linux/hdreg.h>
+#include <linux/interrupt.h>
+#include <asm/ccwdev.h>
+#include <linux/workqueue.h>
+#include <asm/debug.h>
+#include <asm/dasd.h>
+#include <asm/idals.h>
+
+/*
+ * SECTION: Type definitions
+ */
+struct dasd_device;
+
+typedef enum {
+	dasd_era_fatal = -1,	/* no chance to recover		     */
+	dasd_era_none = 0,	/* don't recover, everything alright */
+	dasd_era_msg = 1,	/* don't recover, just report...     */
+	dasd_era_recover = 2	/* recovery action recommended	     */
+} dasd_era_t;
+
+/* BIT DEFINITIONS FOR SENSE DATA */
+#define DASD_SENSE_BIT_0 0x80
+#define DASD_SENSE_BIT_1 0x40
+#define DASD_SENSE_BIT_2 0x20
+#define DASD_SENSE_BIT_3 0x10
+
+/*
+ * SECTION: MACROs for klogd and s390 debug feature (dbf)
+ */
+#define DBF_DEV_EVENT(d_level, d_device, d_str, d_data...) \
+do { \
+	debug_sprintf_event(d_device->debug_area, \
+			    d_level, \
+			    d_str "\n", \
+			    d_data); \
+} while(0)
+
+#define DBF_DEV_EXC(d_level, d_device, d_str, d_data...) \
+do { \
+	debug_sprintf_exception(d_device->debug_area, \
+				d_level, \
+				d_str "\n", \
+				d_data); \
+} while(0)
+
+#define DBF_EVENT(d_level, d_str, d_data...)\
+do { \
+	debug_sprintf_event(dasd_debug_area, \
+			    d_level,\
+			    d_str "\n", \
+			    d_data); \
+} while(0)
+
+#define DBF_EXC(d_level, d_str, d_data...)\
+do { \
+	debug_sprintf_exception(dasd_debug_area, \
+				d_level,\
+				d_str "\n", \
+				d_data); \
+} while(0)
+
+/* definition of dbf debug levels */
+#define	DBF_EMERG	0	/* system is unusable			*/
+#define	DBF_ALERT	1	/* action must be taken immediately	*/
+#define	DBF_CRIT	2	/* critical conditions			*/
+#define	DBF_ERR		3	/* error conditions			*/
+#define	DBF_WARNING	4	/* warning conditions			*/
+#define	DBF_NOTICE	5	/* normal but significant condition	*/
+#define	DBF_INFO	6	/* informational			*/
+#define	DBF_DEBUG	6	/* debug-level messages			*/
+
+/* messages to be written via klogd and dbf */
+#define DEV_MESSAGE(d_loglevel,d_device,d_string,d_args...)\
+do { \
+	printk(d_loglevel PRINTK_HEADER " %s: " d_string "\n", \
+	       d_device->cdev->dev.bus_id, d_args); \
+	DBF_DEV_EVENT(DBF_ALERT, d_device, d_string, d_args); \
+} while(0)
+
+#define MESSAGE(d_loglevel,d_string,d_args...)\
+do { \
+	printk(d_loglevel PRINTK_HEADER " " d_string "\n", d_args); \
+	DBF_EVENT(DBF_ALERT, d_string, d_args); \
+} while(0)
+
+/* messages to be written via klogd only */
+#define DEV_MESSAGE_LOG(d_loglevel,d_device,d_string,d_args...)\
+do { \
+	printk(d_loglevel PRINTK_HEADER " %s: " d_string "\n", \
+	       d_device->cdev->dev.bus_id, d_args); \
+} while(0)
+
+#define MESSAGE_LOG(d_loglevel,d_string,d_args...)\
+do { \
+	printk(d_loglevel PRINTK_HEADER " " d_string "\n", d_args); \
+} while(0)
+
+struct dasd_ccw_req {
+	unsigned int magic;		/* Eye catcher */
+        struct list_head list;		/* list_head for request queueing. */
+
+	/* Where to execute what... */
+	struct dasd_device *device;	/* device the request is for */
+	struct ccw1 *cpaddr;		/* address of channel program */
+	char status;	        	/* status of this request */
+	short retries;			/* A retry counter */
+	unsigned long flags;        	/* flags of this request */
+
+	/* ... and how */
+	unsigned long starttime;	/* jiffies time of request start */
+	int expires;			/* expiration period in jiffies */
+	char lpm;               	/* logical path mask */
+	void *data;			/* pointer to data area */
+
+	/* these are important for recovering erroneous requests          */
+	struct irb irb;			/* device status in case of an error */
+	struct dasd_ccw_req *refers;	/* ERP-chain queueing. */
+	void *function; 		/* originating ERP action */
+
+	/* these are for statistics only */
+	unsigned long long buildclk;	/* TOD-clock of request generation */
+	unsigned long long startclk;	/* TOD-clock of request start */
+	unsigned long long stopclk;	/* TOD-clock of request interrupt */
+	unsigned long long endclk;	/* TOD-clock of request termination */
+
+        /* Callback that is called after reaching final status. */
+        void (*callback)(struct dasd_ccw_req *, void *data);
+        void *callback_data;
+};
+
+/*
+ * dasd_ccw_req -> status can be:
+ */
+#define DASD_CQR_FILLED   0x00	/* request is ready to be processed */
+#define DASD_CQR_QUEUED   0x01	/* request is queued to be processed */
+#define DASD_CQR_IN_IO    0x02	/* request is currently in IO */
+#define DASD_CQR_DONE     0x03	/* request is completed successfully */
+#define DASD_CQR_ERROR    0x04	/* request is completed with error */
+#define DASD_CQR_FAILED   0x05	/* request is finally failed */
+#define DASD_CQR_CLEAR    0x06	/* request is clear pending */
+
+/* per dasd_ccw_req flags */
+#define DASD_CQR_FLAGS_USE_ERP   0	/* use ERP for this request */
+#define DASD_CQR_FLAGS_FAILFAST  1	/* FAILFAST */
+
+/* Signature for error recovery functions. */
+typedef struct dasd_ccw_req *(*dasd_erp_fn_t) (struct dasd_ccw_req *);
+
+/*
+ * the struct dasd_discipline is
+ * sth like a table of virtual functions, if you think of dasd_eckd
+ * inheriting dasd...
+ * no, currently we are not planning to reimplement the driver in C++
+ */
+struct dasd_discipline {
+	struct module *owner;
+	char ebcname[8];	/* a name used for tagging and printks */
+	char name[8];		/* a name used for tagging and printks */
+	int max_blocks;		/* maximum number of blocks to be chained */
+
+	struct list_head list;	/* used for list of disciplines */
+
+        /*
+         * Device recognition functions. check_device is used to verify
+         * the sense data and the information returned by read device
+         * characteristics. It returns 0 if the discipline can be used
+         * for the device in question.
+         * do_analysis is used in the step from device state "basic" to
+         * state "accept". It returns 0 if the device can be made ready,
+         * it returns -EMEDIUMTYPE if the device can't be made ready or
+         * -EAGAIN if do_analysis started a ccw that needs to complete
+         * before the analysis may be repeated.
+         */
+        int (*check_device)(struct dasd_device *);
+	int (*do_analysis) (struct dasd_device *);
+
+        /*
+         * Device operation functions. build_cp creates a ccw chain for
+         * a block device request, start_io starts the request and
+         * term_IO cancels it (e.g. in case of a timeout). format_device
+         * returns a ccw chain to be used to format the device.
+         */
+	struct dasd_ccw_req *(*build_cp) (struct dasd_device *,
+					  struct request *);
+	int (*start_IO) (struct dasd_ccw_req *);
+	int (*term_IO) (struct dasd_ccw_req *);
+	struct dasd_ccw_req *(*format_device) (struct dasd_device *,
+					       struct format_data_t *);
+	int (*free_cp) (struct dasd_ccw_req *, struct request *);
+        /*
+         * Error recovery functions. examine_error() returns a value that
+         * indicates what to do for an error condition. If examine_error()
+	 * returns 'dasd_era_recover' erp_action() is called to create a
+         * special error recovery ccw. erp_postaction() is called after
+         * an error recovery ccw has finished its execution. dump_sense
+         * is called for every error condition to print the sense data
+         * to the console.
+         */
+	dasd_era_t(*examine_error) (struct dasd_ccw_req *, struct irb *);
+	dasd_erp_fn_t(*erp_action) (struct dasd_ccw_req *);
+	dasd_erp_fn_t(*erp_postaction) (struct dasd_ccw_req *);
+	void (*dump_sense) (struct dasd_device *, struct dasd_ccw_req *,
+			    struct irb *);
+
+        /* i/o control functions. */
+	int (*fill_geometry) (struct dasd_device *, struct hd_geometry *);
+	int (*fill_info) (struct dasd_device *, struct dasd_information2_t *);
+	int (*ioctl) (struct dasd_device *, unsigned int, void __user *);
+};
+
+extern struct dasd_discipline *dasd_diag_discipline_pointer;
+
+/*
+ * Unique identifier for dasd device.
+ */
+struct dasd_uid {
+	__u8 alias;
+	char vendor[4];
+	char serial[15];
+	__u16 ssid;
+	__u8 unit_addr;
+};
+
+/*
+ * Notification numbers for extended error reporting notifications:
+ * The DASD_EER_DISABLE notification is sent before a dasd_device (and it's
+ * eer pointer) is freed. The error reporting module needs to do all necessary
+ * cleanup steps.
+ * The DASD_EER_TRIGGER notification sends the actual error reports (triggers).
+ */
+#define DASD_EER_DISABLE 0
+#define DASD_EER_TRIGGER 1
+
+/* Trigger IDs for extended error reporting DASD_EER_TRIGGER notification */
+#define DASD_EER_FATALERROR  1
+#define DASD_EER_NOPATH      2
+#define DASD_EER_STATECHANGE 3
+#define DASD_EER_PPRCSUSPEND 4
+
+struct dasd_device {
+	/* Block device stuff. */
+	struct gendisk *gdp;
+	request_queue_t *request_queue;
+	spinlock_t request_queue_lock;
+	struct block_device *bdev;
+        unsigned int devindex;
+	unsigned long blocks;	   /* size of volume in blocks */
+	unsigned int bp_block;	   /* bytes per block */
+	unsigned int s2b_shift;	   /* log2 (bp_block/512) */
+	unsigned long flags;	   /* per device flags */
+	unsigned short features;   /* copy of devmap-features (read-only!) */
+
+	/* extended error reporting stuff (eer) */
+	struct dasd_ccw_req *eer_cqr;
+
+	/* Device discipline stuff. */
+	struct dasd_discipline *discipline;
+	struct dasd_discipline *base_discipline;
+	char *private;
+
+	/* Device state and target state. */
+	int state, target;
+	int stopped;		/* device (ccw_device_start) was stopped */
+
+	/* Open and reference count. */
+        atomic_t ref_count;
+	atomic_t open_count;
+
+	/* ccw queue and memory for static ccw/erp buffers. */
+	struct list_head ccw_queue;
+	spinlock_t mem_lock;
+	void *ccw_mem;
+	void *erp_mem;
+	struct list_head ccw_chunks;
+	struct list_head erp_chunks;
+
+	atomic_t tasklet_scheduled;
+        struct tasklet_struct tasklet;
+	struct work_struct kick_work;
+	struct timer_list timer;
+
+	debug_info_t *debug_area;
+
+	struct ccw_device *cdev;
+
+#ifdef CONFIG_DASD_PROFILE
+	struct dasd_profile_info_t profile;
+#endif
+};
+
+/* reasons why device (ccw_device_start) was stopped */
+#define DASD_STOPPED_NOT_ACC 1         /* not accessible */
+#define DASD_STOPPED_QUIESCE 2         /* Quiesced */
+#define DASD_STOPPED_PENDING 4         /* long busy */
+#define DASD_STOPPED_DC_WAIT 8         /* disconnected, wait */
+#define DASD_STOPPED_DC_EIO  16        /* disconnected, return -EIO */
+
+/* per device flags */
+#define DASD_FLAG_DSC_ERROR	2	/* return -EIO when disconnected */
+#define DASD_FLAG_OFFLINE	3	/* device is in offline processing */
+#define DASD_FLAG_EER_SNSS	4	/* A SNSS is required */
+#define DASD_FLAG_EER_IN_USE	5	/* A SNSS request is running */
+
+void dasd_put_device_wake(struct dasd_device *);
+
+/*
+ * Reference count inliners
+ */
+static inline void
+dasd_get_device(struct dasd_device *device)
+{
+	atomic_inc(&device->ref_count);
+}
+
+static inline void
+dasd_put_device(struct dasd_device *device)
+{
+	if (atomic_dec_return(&device->ref_count) == 0)
+		dasd_put_device_wake(device);
+}
+
+/*
+ * The static memory in ccw_mem and erp_mem is managed by a sorted
+ * list of free memory chunks.
+ */
+struct dasd_mchunk
+{
+	struct list_head list;
+	unsigned long size;
+} __attribute__ ((aligned(8)));
+
+static inline void
+dasd_init_chunklist(struct list_head *chunk_list, void *mem,
+		    unsigned long size)
+{
+	struct dasd_mchunk *chunk;
+
+	INIT_LIST_HEAD(chunk_list);
+	chunk = (struct dasd_mchunk *) mem;
+	chunk->size = size - sizeof(struct dasd_mchunk);
+	list_add(&chunk->list, chunk_list);
+}
+
+static inline void *
+dasd_alloc_chunk(struct list_head *chunk_list, unsigned long size)
+{
+	struct dasd_mchunk *chunk, *tmp;
+
+	size = (size + 7L) & -8L;
+	list_for_each_entry(chunk, chunk_list, list) {
+		if (chunk->size < size)
+			continue;
+		if (chunk->size > size + sizeof(struct dasd_mchunk)) {
+			char *endaddr = (char *) (chunk + 1) + chunk->size;
+			tmp = (struct dasd_mchunk *) (endaddr - size) - 1;
+			tmp->size = size;
+			chunk->size -= size + sizeof(struct dasd_mchunk);
+			chunk = tmp;
+		} else
+			list_del(&chunk->list);
+		return (void *) (chunk + 1);
+	}
+	return NULL;
+}
+
+static inline void
+dasd_free_chunk(struct list_head *chunk_list, void *mem)
+{
+	struct dasd_mchunk *chunk, *tmp;
+	struct list_head *p, *left;
+
+	chunk = (struct dasd_mchunk *)
+		((char *) mem - sizeof(struct dasd_mchunk));
+	/* Find out the left neighbour in chunk_list. */
+	left = chunk_list;
+	list_for_each(p, chunk_list) {
+		if (list_entry(p, struct dasd_mchunk, list) > chunk)
+			break;
+		left = p;
+	}
+	/* Try to merge with right neighbour = next element from left. */
+	if (left->next != chunk_list) {
+		tmp = list_entry(left->next, struct dasd_mchunk, list);
+		if ((char *) (chunk + 1) + chunk->size == (char *) tmp) {
+			list_del(&tmp->list);
+			chunk->size += tmp->size + sizeof(struct dasd_mchunk);
+		}
+	}
+	/* Try to merge with left neighbour. */
+	if (left != chunk_list) {
+		tmp = list_entry(left, struct dasd_mchunk, list);
+		if ((char *) (tmp + 1) + tmp->size == (char *) chunk) {
+			tmp->size += chunk->size + sizeof(struct dasd_mchunk);
+			return;
+		}
+	}
+	__list_add(&chunk->list, left, left->next);
+}
+
+/*
+ * Check if bsize is in { 512, 1024, 2048, 4096 }
+ */
+static inline int
+dasd_check_blocksize(int bsize)
+{
+	if (bsize < 512 || bsize > 4096 || (bsize & (bsize - 1)) != 0)
+		return -EMEDIUMTYPE;
+	return 0;
+}
+
+/* externals in dasd.c */
+#define DASD_PROFILE_ON	 1
+#define DASD_PROFILE_OFF 0
+
+extern debug_info_t *dasd_debug_area;
+extern struct dasd_profile_info_t dasd_global_profile;
+extern unsigned int dasd_profile_level;
+extern struct block_device_operations dasd_device_operations;
+
+extern struct kmem_cache *dasd_page_cache;
+
+struct dasd_ccw_req *
+dasd_kmalloc_request(char *, int, int, struct dasd_device *);
+struct dasd_ccw_req *
+dasd_smalloc_request(char *, int, int, struct dasd_device *);
+void dasd_kfree_request(struct dasd_ccw_req *, struct dasd_device *);
+void dasd_sfree_request(struct dasd_ccw_req *, struct dasd_device *);
+
+static inline int
+dasd_kmalloc_set_cda(struct ccw1 *ccw, void *cda, struct dasd_device *device)
+{
+	return set_normalized_cda(ccw, cda);
+}
+
+struct dasd_device *dasd_alloc_device(void);
+void dasd_free_device(struct dasd_device *);
+
+void dasd_enable_device(struct dasd_device *);
+void dasd_set_target_state(struct dasd_device *, int);
+void dasd_kick_device(struct dasd_device *);
+
+void dasd_add_request_head(struct dasd_ccw_req *);
+void dasd_add_request_tail(struct dasd_ccw_req *);
+int  dasd_start_IO(struct dasd_ccw_req *);
+int  dasd_term_IO(struct dasd_ccw_req *);
+void dasd_schedule_bh(struct dasd_device *);
+int  dasd_sleep_on(struct dasd_ccw_req *);
+int  dasd_sleep_on_immediatly(struct dasd_ccw_req *);
+int  dasd_sleep_on_interruptible(struct dasd_ccw_req *);
+void dasd_set_timer(struct dasd_device *, int);
+void dasd_clear_timer(struct dasd_device *);
+int  dasd_cancel_req(struct dasd_ccw_req *);
+int dasd_generic_probe (struct ccw_device *, struct dasd_discipline *);
+void dasd_generic_remove (struct ccw_device *cdev);
+int dasd_generic_set_online(struct ccw_device *, struct dasd_discipline *);
+int dasd_generic_set_offline (struct ccw_device *cdev);
+int dasd_generic_notify(struct ccw_device *, int);
+
+/* externals in dasd_devmap.c */
+extern int dasd_max_devindex;
+extern int dasd_probeonly;
+extern int dasd_autodetect;
+extern int dasd_nopav;
+
+int dasd_devmap_init(void);
+void dasd_devmap_exit(void);
+
+struct dasd_device *dasd_create_device(struct ccw_device *);
+void dasd_delete_device(struct dasd_device *);
+
+int dasd_get_uid(struct ccw_device *, struct dasd_uid *);
+int dasd_set_uid(struct ccw_device *, struct dasd_uid *);
+int dasd_get_feature(struct ccw_device *, int);
+int dasd_set_feature(struct ccw_device *, int, int);
+
+int dasd_add_sysfs_files(struct ccw_device *);
+void dasd_remove_sysfs_files(struct ccw_device *);
+
+struct dasd_device *dasd_device_from_cdev(struct ccw_device *);
+struct dasd_device *dasd_device_from_cdev_locked(struct ccw_device *);
+struct dasd_device *dasd_device_from_devindex(int);
+
+int dasd_parse(void);
+int dasd_busid_known(char *);
+
+/* externals in dasd_gendisk.c */
+int  dasd_gendisk_init(void);
+void dasd_gendisk_exit(void);
+int dasd_gendisk_alloc(struct dasd_device *);
+void dasd_gendisk_free(struct dasd_device *);
+int dasd_scan_partitions(struct dasd_device *);
+void dasd_destroy_partitions(struct dasd_device *);
+
+/* externals in dasd_ioctl.c */
+int  dasd_ioctl(struct inode *, struct file *, unsigned int, unsigned long);
+long dasd_compat_ioctl(struct file *, unsigned int, unsigned long);
+
+/* externals in dasd_proc.c */
+int dasd_proc_init(void);
+void dasd_proc_exit(void);
+
+/* externals in dasd_erp.c */
+struct dasd_ccw_req *dasd_default_erp_action(struct dasd_ccw_req *);
+struct dasd_ccw_req *dasd_default_erp_postaction(struct dasd_ccw_req *);
+struct dasd_ccw_req *dasd_alloc_erp_request(char *, int, int,
+					    struct dasd_device *);
+void dasd_free_erp_request(struct dasd_ccw_req *, struct dasd_device *);
+void dasd_log_sense(struct dasd_ccw_req *, struct irb *);
+
+/* externals in dasd_3370_erp.c */
+dasd_era_t dasd_3370_erp_examine(struct dasd_ccw_req *, struct irb *);
+
+/* externals in dasd_3990_erp.c */
+dasd_era_t dasd_3990_erp_examine(struct dasd_ccw_req *, struct irb *);
+struct dasd_ccw_req *dasd_3990_erp_action(struct dasd_ccw_req *);
+
+/* externals in dasd_9336_erp.c */
+dasd_era_t dasd_9336_erp_examine(struct dasd_ccw_req *, struct irb *);
+
+/* externals in dasd_9336_erp.c */
+dasd_era_t dasd_9343_erp_examine(struct dasd_ccw_req *, struct irb *);
+struct dasd_ccw_req *dasd_9343_erp_action(struct dasd_ccw_req *);
+
+/* externals in dasd_eer.c */
+#ifdef CONFIG_DASD_EER
+int dasd_eer_init(void);
+void dasd_eer_exit(void);
+int dasd_eer_enable(struct dasd_device *);
+void dasd_eer_disable(struct dasd_device *);
+void dasd_eer_write(struct dasd_device *, struct dasd_ccw_req *cqr,
+		    unsigned int id);
+void dasd_eer_snss(struct dasd_device *);
+
+static inline int dasd_eer_enabled(struct dasd_device *device)
+{
+	return device->eer_cqr != NULL;
+}
+#else
+#define dasd_eer_init()		(0)
+#define dasd_eer_exit()		do { } while (0)
+#define dasd_eer_enable(d)	(0)
+#define dasd_eer_disable(d)	do { } while (0)
+#define dasd_eer_write(d,c,i)	do { } while (0)
+#define dasd_eer_snss(d)	do { } while (0)
+#define dasd_eer_enabled(d)	(0)
+#endif	/* CONFIG_DASD_ERR */
+
+#endif				/* __KERNEL__ */
+
+#endif				/* DASD_H */

drivers/s390/block/dasd_ioctl.c

@@ -0,0 +1,437 @@
+/*
+ * File...........: linux/drivers/s390/block/dasd_ioctl.c
+ * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com>
+ *		    Horst Hummel <Horst.Hummel@de.ibm.com>
+ *		    Carsten Otte <Cotte@de.ibm.com>
+ *		    Martin Schwidefsky <schwidefsky@de.ibm.com>
+ * Bugreports.to..: <Linux390@de.ibm.com>
+ * (C) IBM Corporation, IBM Deutschland Entwicklung GmbH, 1999-2001
+ *
+ * i/o controls for the dasd driver.
+ */
+#include <linux/interrupt.h>
+#include <linux/major.h>
+#include <linux/fs.h>
+#include <linux/blkpg.h>
+
+#include <asm/ccwdev.h>
+#include <asm/cmb.h>
+#include <asm/uaccess.h>
+
+/* This is ugly... */
+#define PRINTK_HEADER "dasd_ioctl:"
+
+#include "dasd_int.h"
+
+
+static int
+dasd_ioctl_api_version(void __user *argp)
+{
+	int ver = DASD_API_VERSION;
+	return put_user(ver, (int __user *)argp);
+}
+
+/*
+ * Enable device.
+ * used by dasdfmt after BIODASDDISABLE to retrigger blocksize detection
+ */
+static int
+dasd_ioctl_enable(struct block_device *bdev)
+{
+	struct dasd_device *device = bdev->bd_disk->private_data;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EACCES;
+
+	dasd_enable_device(device);
+	/* Formatting the dasd device can change the capacity. */
+	mutex_lock(&bdev->bd_mutex);
+	i_size_write(bdev->bd_inode, (loff_t)get_capacity(device->gdp) << 9);
+	mutex_unlock(&bdev->bd_mutex);
+	return 0;
+}
+
+/*
+ * Disable device.
+ * Used by dasdfmt. Disable I/O operations but allow ioctls.
+ */
+static int
+dasd_ioctl_disable(struct block_device *bdev)
+{
+	struct dasd_device *device = bdev->bd_disk->private_data;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EACCES;
+
+	/*
+	 * Man this is sick. We don't do a real disable but only downgrade
+	 * the device to DASD_STATE_BASIC. The reason is that dasdfmt uses
+	 * BIODASDDISABLE to disable accesses to the device via the block
+	 * device layer but it still wants to do i/o on the device by
+	 * using the BIODASDFMT ioctl. Therefore the correct state for the
+	 * device is DASD_STATE_BASIC that allows to do basic i/o.
+	 */
+	dasd_set_target_state(device, DASD_STATE_BASIC);
+	/*
+	 * Set i_size to zero, since read, write, etc. check against this
+	 * value.
+	 */
+	mutex_lock(&bdev->bd_mutex);
+	i_size_write(bdev->bd_inode, 0);
+	mutex_unlock(&bdev->bd_mutex);
+	return 0;
+}
+
+/*
+ * Quiesce device.
+ */
+static int
+dasd_ioctl_quiesce(struct dasd_device *device)
+{
+	unsigned long flags;
+
+	if (!capable (CAP_SYS_ADMIN))
+		return -EACCES;
+
+	DEV_MESSAGE (KERN_DEBUG, device, "%s",
+		     "Quiesce IO on device");
+	spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
+	device->stopped |= DASD_STOPPED_QUIESCE;
+	spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
+	return 0;
+}
+
+
+/*
+ * Quiesce device.
+ */
+static int
+dasd_ioctl_resume(struct dasd_device *device)
+{
+	unsigned long flags;
+
+	if (!capable (CAP_SYS_ADMIN))
+		return -EACCES;
+
+	DEV_MESSAGE (KERN_DEBUG, device, "%s",
+		     "resume IO on device");
+
+	spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
+	device->stopped &= ~DASD_STOPPED_QUIESCE;
+	spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
+
+	dasd_schedule_bh (device);
+	return 0;
+}
+
+/*
+ * performs formatting of _device_ according to _fdata_
+ * Note: The discipline's format_function is assumed to deliver formatting
+ * commands to format a single unit of the device. In terms of the ECKD
+ * devices this means CCWs are generated to format a single track.
+ */
+static int
+dasd_format(struct dasd_device * device, struct format_data_t * fdata)
+{
+	struct dasd_ccw_req *cqr;
+	int rc;
+
+	if (device->discipline->format_device == NULL)
+		return -EPERM;
+
+	if (device->state != DASD_STATE_BASIC) {
+		DEV_MESSAGE(KERN_WARNING, device, "%s",
+			    "dasd_format: device is not disabled! ");
+		return -EBUSY;
+	}
+
+	DBF_DEV_EVENT(DBF_NOTICE, device,
+		      "formatting units %d to %d (%d B blocks) flags %d",
+		      fdata->start_unit,
+		      fdata->stop_unit, fdata->blksize, fdata->intensity);
+
+	/* Since dasdfmt keeps the device open after it was disabled,
+	 * there still exists an inode for this device.
+	 * We must update i_blkbits, otherwise we might get errors when
+	 * enabling the device later.
+	 */
+	if (fdata->start_unit == 0) {
+		struct block_device *bdev = bdget_disk(device->gdp, 0);
+		bdev->bd_inode->i_blkbits = blksize_bits(fdata->blksize);
+		bdput(bdev);
+	}
+
+	while (fdata->start_unit <= fdata->stop_unit) {
+		cqr = device->discipline->format_device(device, fdata);
+		if (IS_ERR(cqr))
+			return PTR_ERR(cqr);
+		rc = dasd_sleep_on_interruptible(cqr);
+		dasd_sfree_request(cqr, cqr->device);
+		if (rc) {
+			if (rc != -ERESTARTSYS)
+				DEV_MESSAGE(KERN_ERR, device,
+					    " Formatting of unit %d failed "
+					    "with rc = %d",
+					    fdata->start_unit, rc);
+			return rc;
+		}
+		fdata->start_unit++;
+	}
+	return 0;
+}
+
+/*
+ * Format device.
+ */
+static int
+dasd_ioctl_format(struct block_device *bdev, void __user *argp)
+{
+	struct dasd_device *device = bdev->bd_disk->private_data;
+	struct format_data_t fdata;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EACCES;
+	if (!argp)
+		return -EINVAL;
+
+	if (device->features & DASD_FEATURE_READONLY)
+		return -EROFS;
+	if (copy_from_user(&fdata, argp, sizeof(struct format_data_t)))
+		return -EFAULT;
+	if (bdev != bdev->bd_contains) {
+		DEV_MESSAGE(KERN_WARNING, device, "%s",
+			    "Cannot low-level format a partition");
+		return -EINVAL;
+	}
+	return dasd_format(device, &fdata);
+}
+
+#ifdef CONFIG_DASD_PROFILE
+/*
+ * Reset device profile information
+ */
+static int
+dasd_ioctl_reset_profile(struct dasd_device *device)
+{
+	memset(&device->profile, 0, sizeof (struct dasd_profile_info_t));
+	return 0;
+}
+
+/*
+ * Return device profile information
+ */
+static int
+dasd_ioctl_read_profile(struct dasd_device *device, void __user *argp)
+{
+	if (dasd_profile_level == DASD_PROFILE_OFF)
+		return -EIO;
+	if (copy_to_user(argp, &device->profile,
+			 sizeof (struct dasd_profile_info_t)))
+		return -EFAULT;
+	return 0;
+}
+#else
+static int
+dasd_ioctl_reset_profile(struct dasd_device *device)
+{
+	return -ENOSYS;
+}
+
+static int
+dasd_ioctl_read_profile(struct dasd_device *device, void __user *argp)
+{
+	return -ENOSYS;
+}
+#endif
+
+/*
+ * Return dasd information. Used for BIODASDINFO and BIODASDINFO2.
+ */
+static int
+dasd_ioctl_information(struct dasd_device *device,
+		unsigned int cmd, void __user *argp)
+{
+	struct dasd_information2_t *dasd_info;
+	unsigned long flags;
+	int rc;
+	struct ccw_device *cdev;
+
+	if (!device->discipline->fill_info)
+		return -EINVAL;
+
+	dasd_info = kzalloc(sizeof(struct dasd_information2_t), GFP_KERNEL);
+	if (dasd_info == NULL)
+		return -ENOMEM;
+
+	rc = device->discipline->fill_info(device, dasd_info);
+	if (rc) {
+		kfree(dasd_info);
+		return rc;
+	}
+
+	cdev = device->cdev;
+
+	dasd_info->devno = _ccw_device_get_device_number(device->cdev);
+	dasd_info->schid = _ccw_device_get_subchannel_number(device->cdev);
+	dasd_info->cu_type = cdev->id.cu_type;
+	dasd_info->cu_model = cdev->id.cu_model;
+	dasd_info->dev_type = cdev->id.dev_type;
+	dasd_info->dev_model = cdev->id.dev_model;
+	dasd_info->status = device->state;
+	/*
+	 * The open_count is increased for every opener, that includes
+	 * the blkdev_get in dasd_scan_partitions.
+	 * This must be hidden from user-space.
+	 */
+	dasd_info->open_count = atomic_read(&device->open_count);
+	if (!device->bdev)
+		dasd_info->open_count++;
+
+	/*
+	 * check if device is really formatted
+	 * LDL / CDL was returned by 'fill_info'
+	 */
+	if ((device->state < DASD_STATE_READY) ||
+	    (dasd_check_blocksize(device->bp_block)))
+		dasd_info->format = DASD_FORMAT_NONE;
+
+	dasd_info->features |=
+		((device->features & DASD_FEATURE_READONLY) != 0);
+
+	if (device->discipline)
+		memcpy(dasd_info->type, device->discipline->name, 4);
+	else
+		memcpy(dasd_info->type, "none", 4);
+
+	if (device->request_queue->request_fn) {
+		struct list_head *l;
+#ifdef DASD_EXTENDED_PROFILING
+		{
+			struct list_head *l;
+			spin_lock_irqsave(&device->lock, flags);
+			list_for_each(l, &device->request_queue->queue_head)
+				dasd_info->req_queue_len++;
+			spin_unlock_irqrestore(&device->lock, flags);
+		}
+#endif				/* DASD_EXTENDED_PROFILING */
+		spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
+		list_for_each(l, &device->ccw_queue)
+			dasd_info->chanq_len++;
+		spin_unlock_irqrestore(get_ccwdev_lock(device->cdev),
+				       flags);
+	}
+
+	rc = 0;
+	if (copy_to_user(argp, dasd_info,
+			 ((cmd == (unsigned int) BIODASDINFO2) ?
+			  sizeof (struct dasd_information2_t) :
+			  sizeof (struct dasd_information_t))))
+		rc = -EFAULT;
+	kfree(dasd_info);
+	return rc;
+}
+
+/*
+ * Set read only
+ */
+static int
+dasd_ioctl_set_ro(struct block_device *bdev, void __user *argp)
+{
+	struct dasd_device *device =  bdev->bd_disk->private_data;
+	int intval;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EACCES;
+	if (bdev != bdev->bd_contains)
+		// ro setting is not allowed for partitions
+		return -EINVAL;
+	if (get_user(intval, (int __user *)argp))
+		return -EFAULT;
+
+	set_disk_ro(bdev->bd_disk, intval);
+	return dasd_set_feature(device->cdev, DASD_FEATURE_READONLY, intval);
+}
+
+static int
+dasd_ioctl_readall_cmb(struct dasd_device *device, unsigned int cmd,
+		unsigned long arg)
+{
+	struct cmbdata __user *argp = (void __user *) arg;
+	size_t size = _IOC_SIZE(cmd);
+	struct cmbdata data;
+	int ret;
+
+	ret = cmf_readall(device->cdev, &data);
+	if (!ret && copy_to_user(argp, &data, min(size, sizeof(*argp))))
+		return -EFAULT;
+	return ret;
+}
+
+int
+dasd_ioctl(struct inode *inode, struct file *file,
+	   unsigned int cmd, unsigned long arg)
+{
+	struct block_device *bdev = inode->i_bdev;
+	struct dasd_device *device = bdev->bd_disk->private_data;
+	void __user *argp = (void __user *)arg;
+
+	if (!device)
+                return -ENODEV;
+
+	if ((_IOC_DIR(cmd) != _IOC_NONE) && !arg) {
+		PRINT_DEBUG("empty data ptr");
+		return -EINVAL;
+	}
+
+	switch (cmd) {
+	case BIODASDDISABLE:
+		return dasd_ioctl_disable(bdev);
+	case BIODASDENABLE:
+		return dasd_ioctl_enable(bdev);
+	case BIODASDQUIESCE:
+		return dasd_ioctl_quiesce(device);
+	case BIODASDRESUME:
+		return dasd_ioctl_resume(device);
+	case BIODASDFMT:
+		return dasd_ioctl_format(bdev, argp);
+	case BIODASDINFO:
+		return dasd_ioctl_information(device, cmd, argp);
+	case BIODASDINFO2:
+		return dasd_ioctl_information(device, cmd, argp);
+	case BIODASDPRRD:
+		return dasd_ioctl_read_profile(device, argp);
+	case BIODASDPRRST:
+		return dasd_ioctl_reset_profile(device);
+	case BLKROSET:
+		return dasd_ioctl_set_ro(bdev, argp);
+	case DASDAPIVER:
+		return dasd_ioctl_api_version(argp);
+	case BIODASDCMFENABLE:
+		return enable_cmf(device->cdev);
+	case BIODASDCMFDISABLE:
+		return disable_cmf(device->cdev);
+	case BIODASDREADALLCMB:
+		return dasd_ioctl_readall_cmb(device, cmd, arg);
+	default:
+		/* if the discipline has an ioctl method try it. */
+		if (device->discipline->ioctl) {
+			int rval = device->discipline->ioctl(device, cmd, argp);
+			if (rval != -ENOIOCTLCMD)
+				return rval;
+		}
+
+		return -EINVAL;
+	}
+}
+
+long
+dasd_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
+{
+	int rval;
+
+	lock_kernel();
+	rval = dasd_ioctl(filp->f_path.dentry->d_inode, filp, cmd, arg);
+	unlock_kernel();
+
+	return (rval == -EINVAL) ? -ENOIOCTLCMD : rval;
+}

drivers/s390/block/dasd_proc.c

@@ -0,0 +1,338 @@
+/*
+ * File...........: linux/drivers/s390/block/dasd_proc.c
+ * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com>
+ *		    Horst Hummel <Horst.Hummel@de.ibm.com>
+ *		    Carsten Otte <Cotte@de.ibm.com>
+ *		    Martin Schwidefsky <schwidefsky@de.ibm.com>
+ * Bugreports.to..: <Linux390@de.ibm.com>
+ * (C) IBM Corporation, IBM Deutschland Entwicklung GmbH, 1999-2002
+ *
+ * /proc interface for the dasd driver.
+ *
+ */
+
+#include <linux/ctype.h>
+#include <linux/seq_file.h>
+#include <linux/vmalloc.h>
+#include <linux/proc_fs.h>
+
+#include <asm/debug.h>
+#include <asm/uaccess.h>
+
+/* This is ugly... */
+#define PRINTK_HEADER "dasd_proc:"
+
+#include "dasd_int.h"
+
+static struct proc_dir_entry *dasd_proc_root_entry = NULL;
+static struct proc_dir_entry *dasd_devices_entry = NULL;
+static struct proc_dir_entry *dasd_statistics_entry = NULL;
+
+static char *
+dasd_get_user_string(const char __user *user_buf, size_t user_len)
+{
+	char *buffer;
+
+	buffer = kmalloc(user_len + 1, GFP_KERNEL);
+	if (buffer == NULL)
+		return ERR_PTR(-ENOMEM);
+	if (copy_from_user(buffer, user_buf, user_len) != 0) {
+		kfree(buffer);
+		return ERR_PTR(-EFAULT);
+	}
+	/* got the string, now strip linefeed. */
+	if (buffer[user_len - 1] == '\n')
+		buffer[user_len - 1] = 0;
+	else
+		buffer[user_len] = 0;
+	return buffer;
+}
+
+static int
+dasd_devices_show(struct seq_file *m, void *v)
+{
+	struct dasd_device *device;
+	char *substr;
+
+	device = dasd_device_from_devindex((unsigned long) v - 1);
+	if (IS_ERR(device))
+		return 0;
+	/* Print device number. */
+	seq_printf(m, "%s", device->cdev->dev.bus_id);
+	/* Print discipline string. */
+	if (device != NULL && device->discipline != NULL)
+		seq_printf(m, "(%s)", device->discipline->name);
+	else
+		seq_printf(m, "(none)");
+	/* Print kdev. */
+	if (device->gdp)
+		seq_printf(m, " at (%3d:%6d)",
+			   device->gdp->major, device->gdp->first_minor);
+	else
+		seq_printf(m, "  at (???:??????)");
+	/* Print device name. */
+	if (device->gdp)
+		seq_printf(m, " is %-8s", device->gdp->disk_name);
+	else
+		seq_printf(m, " is ????????");
+	/* Print devices features. */
+	substr = (device->features & DASD_FEATURE_READONLY) ? "(ro)" : " ";
+	seq_printf(m, "%4s: ", substr);
+	/* Print device status information. */
+	switch ((device != NULL) ? device->state : -1) {
+	case -1:
+		seq_printf(m, "unknown");
+		break;
+	case DASD_STATE_NEW:
+		seq_printf(m, "new");
+		break;
+	case DASD_STATE_KNOWN:
+		seq_printf(m, "detected");
+		break;
+	case DASD_STATE_BASIC:
+		seq_printf(m, "basic");
+		break;
+	case DASD_STATE_UNFMT:
+		seq_printf(m, "unformatted");
+		break;
+	case DASD_STATE_READY:
+	case DASD_STATE_ONLINE:
+		seq_printf(m, "active ");
+		if (dasd_check_blocksize(device->bp_block))
+			seq_printf(m, "n/f	 ");
+		else
+			seq_printf(m,
+				   "at blocksize: %d, %ld blocks, %ld MB",
+				   device->bp_block, device->blocks,
+				   ((device->bp_block >> 9) *
+				    device->blocks) >> 11);
+		break;
+	default:
+		seq_printf(m, "no stat");
+		break;
+	}
+	dasd_put_device(device);
+	if (dasd_probeonly)
+		seq_printf(m, "(probeonly)");
+	seq_printf(m, "\n");
+	return 0;
+}
+
+static void *dasd_devices_start(struct seq_file *m, loff_t *pos)
+{
+	if (*pos >= dasd_max_devindex)
+		return NULL;
+	return (void *)((unsigned long) *pos + 1);
+}
+
+static void *dasd_devices_next(struct seq_file *m, void *v, loff_t *pos)
+{
+	++*pos;
+	return dasd_devices_start(m, pos);
+}
+
+static void dasd_devices_stop(struct seq_file *m, void *v)
+{
+}
+
+static struct seq_operations dasd_devices_seq_ops = {
+	.start		= dasd_devices_start,
+	.next		= dasd_devices_next,
+	.stop		= dasd_devices_stop,
+	.show		= dasd_devices_show,
+};
+
+static int dasd_devices_open(struct inode *inode, struct file *file)
+{
+	return seq_open(file, &dasd_devices_seq_ops);
+}
+
+static const struct file_operations dasd_devices_file_ops = {
+	.open		= dasd_devices_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= seq_release,
+};
+
+static int
+dasd_calc_metrics(char *page, char **start, off_t off,
+		  int count, int *eof, int len)
+{
+	len = (len > off) ? len - off : 0;
+	if (len > count)
+		len = count;
+	if (len < count)
+		*eof = 1;
+	*start = page + off;
+	return len;
+}
+
+static char *
+dasd_statistics_array(char *str, unsigned int *array, int shift)
+{
+	int i;
+
+	for (i = 0; i < 32; i++) {
+		str += sprintf(str, "%7d ", array[i] >> shift);
+		if (i == 15)
+			str += sprintf(str, "\n");
+	}
+	str += sprintf(str,"\n");
+	return str;
+}
+
+static int
+dasd_statistics_read(char *page, char **start, off_t off,
+		     int count, int *eof, void *data)
+{
+	unsigned long len;
+#ifdef CONFIG_DASD_PROFILE
+	struct dasd_profile_info_t *prof;
+	char *str;
+	int shift;
+
+	/* check for active profiling */
+	if (dasd_profile_level == DASD_PROFILE_OFF) {
+		len = sprintf(page, "Statistics are off - they might be "
+				    "switched on using 'echo set on > "
+				    "/proc/dasd/statistics'\n");
+		return dasd_calc_metrics(page, start, off, count, eof, len);
+	}
+
+	prof = &dasd_global_profile;
+	/* prevent couter 'overflow' on output */
+	for (shift = 0; (prof->dasd_io_reqs >> shift) > 9999999; shift++);
+
+	str = page;
+	str += sprintf(str, "%d dasd I/O requests\n", prof->dasd_io_reqs);
+	str += sprintf(str, "with %d sectors(512B each)\n",
+		       prof->dasd_io_sects);
+	str += sprintf(str,
+		       "   __<4	   ___8	   __16	   __32	   __64	   _128	"
+		       "   _256	   _512	   __1k	   __2k	   __4k	   __8k	"
+		       "   _16k	   _32k	   _64k	   128k\n");
+	str += sprintf(str,
+		       "   _256	   _512	   __1M	   __2M	   __4M	   __8M	"
+		       "   _16M	   _32M	   _64M	   128M	   256M	   512M	"
+		       "   __1G	   __2G	   __4G " "   _>4G\n");
+
+	str += sprintf(str, "Histogram of sizes (512B secs)\n");
+	str = dasd_statistics_array(str, prof->dasd_io_secs, shift);
+	str += sprintf(str, "Histogram of I/O times (microseconds)\n");
+	str = dasd_statistics_array(str, prof->dasd_io_times, shift);
+	str += sprintf(str, "Histogram of I/O times per sector\n");
+	str = dasd_statistics_array(str, prof->dasd_io_timps, shift);
+	str += sprintf(str, "Histogram of I/O time till ssch\n");
+	str = dasd_statistics_array(str, prof->dasd_io_time1, shift);
+	str += sprintf(str, "Histogram of I/O time between ssch and irq\n");
+	str = dasd_statistics_array(str, prof->dasd_io_time2, shift);
+	str += sprintf(str, "Histogram of I/O time between ssch "
+			    "and irq per sector\n");
+	str = dasd_statistics_array(str, prof->dasd_io_time2ps, shift);
+	str += sprintf(str, "Histogram of I/O time between irq and end\n");
+	str = dasd_statistics_array(str, prof->dasd_io_time3, shift);
+	str += sprintf(str, "# of req in chanq at enqueuing (1..32) \n");
+	str = dasd_statistics_array(str, prof->dasd_io_nr_req, shift);
+	len = str - page;
+#else
+	len = sprintf(page, "Statistics are not activated in this kernel\n");
+#endif
+	return dasd_calc_metrics(page, start, off, count, eof, len);
+}
+
+static int
+dasd_statistics_write(struct file *file, const char __user *user_buf,
+		      unsigned long user_len, void *data)
+{
+#ifdef CONFIG_DASD_PROFILE
+	char *buffer, *str;
+
+	if (user_len > 65536)
+		user_len = 65536;
+	buffer = dasd_get_user_string(user_buf, user_len);
+	if (IS_ERR(buffer))
+		return PTR_ERR(buffer);
+	MESSAGE_LOG(KERN_INFO, "/proc/dasd/statictics: '%s'", buffer);
+
+	/* check for valid verbs */
+	for (str = buffer; isspace(*str); str++);
+	if (strncmp(str, "set", 3) == 0 && isspace(str[3])) {
+		/* 'set xxx' was given */
+		for (str = str + 4; isspace(*str); str++);
+		if (strcmp(str, "on") == 0) {
+			/* switch on statistics profiling */
+			dasd_profile_level = DASD_PROFILE_ON;
+			MESSAGE(KERN_INFO, "%s", "Statistics switched on");
+		} else if (strcmp(str, "off") == 0) {
+			/* switch off and reset statistics profiling */
+			memset(&dasd_global_profile,
+			       0, sizeof (struct dasd_profile_info_t));
+			dasd_profile_level = DASD_PROFILE_OFF;
+			MESSAGE(KERN_INFO, "%s", "Statistics switched off");
+		} else
+			goto out_error;
+	} else if (strncmp(str, "reset", 5) == 0) {
+		/* reset the statistics */
+		memset(&dasd_global_profile, 0,
+		       sizeof (struct dasd_profile_info_t));
+		MESSAGE(KERN_INFO, "%s", "Statistics reset");
+	} else
+		goto out_error;
+	kfree(buffer);
+	return user_len;
+out_error:
+	MESSAGE(KERN_WARNING, "%s",
+		"/proc/dasd/statistics: only 'set on', 'set off' "
+		"and 'reset' are supported verbs");
+	kfree(buffer);
+	return -EINVAL;
+#else
+	MESSAGE(KERN_WARNING, "%s",
+		"/proc/dasd/statistics: is not activated in this kernel");
+	return user_len;
+#endif				/* CONFIG_DASD_PROFILE */
+}
+
+/*
+ * Create dasd proc-fs entries.
+ * In case creation failed, cleanup and return -ENOENT.
+ */
+int
+dasd_proc_init(void)
+{
+	dasd_proc_root_entry = proc_mkdir("dasd", &proc_root);
+	if (!dasd_proc_root_entry)
+		goto out_nodasd;
+	dasd_proc_root_entry->owner = THIS_MODULE;
+	dasd_devices_entry = create_proc_entry("devices",
+					       S_IFREG | S_IRUGO | S_IWUSR,
+					       dasd_proc_root_entry);
+	if (!dasd_devices_entry)
+		goto out_nodevices;
+	dasd_devices_entry->proc_fops = &dasd_devices_file_ops;
+	dasd_devices_entry->owner = THIS_MODULE;
+	dasd_statistics_entry = create_proc_entry("statistics",
+						  S_IFREG | S_IRUGO | S_IWUSR,
+						  dasd_proc_root_entry);
+	if (!dasd_statistics_entry)
+		goto out_nostatistics;
+	dasd_statistics_entry->read_proc = dasd_statistics_read;
+	dasd_statistics_entry->write_proc = dasd_statistics_write;
+	dasd_statistics_entry->owner = THIS_MODULE;
+	return 0;
+
+ out_nostatistics:
+	remove_proc_entry("devices", dasd_proc_root_entry);
+ out_nodevices:
+	remove_proc_entry("dasd", &proc_root);
+ out_nodasd:
+	return -ENOENT;
+}
+
+void
+dasd_proc_exit(void)
+{
+	remove_proc_entry("devices", dasd_proc_root_entry);
+	remove_proc_entry("statistics", dasd_proc_root_entry);
+	remove_proc_entry("dasd", &proc_root);
+}

drivers/s390/block/dcssblk.c

@@ -0,0 +1,810 @@
+/*
+ * dcssblk.c -- the S/390 block driver for dcss memory
+ *
+ * Authors: Carsten Otte, Stefan Weinhuber, Gerald Schaefer
+ */
+
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/ctype.h>
+#include <linux/errno.h>
+#include <linux/init.h>
+#include <linux/slab.h>
+#include <linux/blkdev.h>
+#include <asm/extmem.h>
+#include <asm/io.h>
+#include <linux/completion.h>
+#include <linux/interrupt.h>
+#include <asm/s390_rdev.h>
+
+//#define DCSSBLK_DEBUG		/* Debug messages on/off */
+#define DCSSBLK_NAME "dcssblk"
+#define DCSSBLK_MINORS_PER_DISK 1
+#define DCSSBLK_PARM_LEN 400
+
+#ifdef DCSSBLK_DEBUG
+#define PRINT_DEBUG(x...) printk(KERN_DEBUG DCSSBLK_NAME " debug: " x)
+#else
+#define PRINT_DEBUG(x...) do {} while (0)
+#endif
+#define PRINT_INFO(x...)  printk(KERN_INFO DCSSBLK_NAME " info: " x)
+#define PRINT_WARN(x...)  printk(KERN_WARNING DCSSBLK_NAME " warning: " x)
+#define PRINT_ERR(x...)	  printk(KERN_ERR DCSSBLK_NAME " error: " x)
+
+
+static int dcssblk_open(struct inode *inode, struct file *filp);
+static int dcssblk_release(struct inode *inode, struct file *filp);
+static int dcssblk_make_request(struct request_queue *q, struct bio *bio);
+static int dcssblk_direct_access(struct block_device *bdev, sector_t secnum,
+				 unsigned long *data);
+
+static char dcssblk_segments[DCSSBLK_PARM_LEN] = "\0";
+
+static int dcssblk_major;
+static struct block_device_operations dcssblk_devops = {
+	.owner   	= THIS_MODULE,
+	.open    	= dcssblk_open,
+	.release 	= dcssblk_release,
+	.direct_access 	= dcssblk_direct_access,
+};
+
+static ssize_t dcssblk_add_store(struct device * dev, struct device_attribute *attr, const char * buf,
+				  size_t count);
+static ssize_t dcssblk_remove_store(struct device * dev, struct device_attribute *attr, const char * buf,
+				  size_t count);
+static ssize_t dcssblk_save_store(struct device * dev, struct device_attribute *attr, const char * buf,
+				  size_t count);
+static ssize_t dcssblk_save_show(struct device *dev, struct device_attribute *attr, char *buf);
+static ssize_t dcssblk_shared_store(struct device * dev, struct device_attribute *attr, const char * buf,
+				  size_t count);
+static ssize_t dcssblk_shared_show(struct device *dev, struct device_attribute *attr, char *buf);
+
+static DEVICE_ATTR(add, S_IWUSR, NULL, dcssblk_add_store);
+static DEVICE_ATTR(remove, S_IWUSR, NULL, dcssblk_remove_store);
+static DEVICE_ATTR(save, S_IWUSR | S_IRUGO, dcssblk_save_show,
+		   dcssblk_save_store);
+static DEVICE_ATTR(shared, S_IWUSR | S_IRUGO, dcssblk_shared_show,
+		   dcssblk_shared_store);
+
+static struct device *dcssblk_root_dev;
+
+struct dcssblk_dev_info {
+	struct list_head lh;
+	struct device dev;
+	char segment_name[BUS_ID_SIZE];
+	atomic_t use_count;
+	struct gendisk *gd;
+	unsigned long start;
+	unsigned long end;
+	int segment_type;
+	unsigned char save_pending;
+	unsigned char is_shared;
+	struct request_queue *dcssblk_queue;
+};
+
+static struct list_head dcssblk_devices = LIST_HEAD_INIT(dcssblk_devices);
+static struct rw_semaphore dcssblk_devices_sem;
+
+/*
+ * release function for segment device.
+ */
+static void
+dcssblk_release_segment(struct device *dev)
+{
+	PRINT_DEBUG("segment release fn called for %s\n", dev->bus_id);
+	kfree(container_of(dev, struct dcssblk_dev_info, dev));
+	module_put(THIS_MODULE);
+}
+
+/*
+ * get a minor number. needs to be called with
+ * down_write(&dcssblk_devices_sem) and the
+ * device needs to be enqueued before the semaphore is
+ * freed.
+ */
+static int
+dcssblk_assign_free_minor(struct dcssblk_dev_info *dev_info)
+{
+	int minor, found;
+	struct dcssblk_dev_info *entry;
+
+	if (dev_info == NULL)
+		return -EINVAL;
+	for (minor = 0; minor < (1<<MINORBITS); minor++) {
+		found = 0;
+		// test if minor available
+		list_for_each_entry(entry, &dcssblk_devices, lh)
+			if (minor == entry->gd->first_minor)
+				found++;
+		if (!found) break; // got unused minor
+	}
+	if (found)
+		return -EBUSY;
+	dev_info->gd->first_minor = minor;
+	return 0;
+}
+
+/*
+ * get the struct dcssblk_dev_info from dcssblk_devices
+ * for the given name.
+ * down_read(&dcssblk_devices_sem) must be held.
+ */
+static struct dcssblk_dev_info *
+dcssblk_get_device_by_name(char *name)
+{
+	struct dcssblk_dev_info *entry;
+
+	list_for_each_entry(entry, &dcssblk_devices, lh) {
+		if (!strcmp(name, entry->segment_name)) {
+			return entry;
+		}
+	}
+	return NULL;
+}
+
+/*
+ * print appropriate error message for segment_load()/segment_type()
+ * return code
+ */
+static void
+dcssblk_segment_warn(int rc, char* seg_name)
+{
+	switch (rc) {
+	case -ENOENT:
+		PRINT_WARN("cannot load/query segment %s, does not exist\n",
+			   seg_name);
+		break;
+	case -ENOSYS:
+		PRINT_WARN("cannot load/query segment %s, not running on VM\n",
+			   seg_name);
+		break;
+	case -EIO:
+		PRINT_WARN("cannot load/query segment %s, hardware error\n",
+			   seg_name);
+		break;
+	case -ENOTSUPP:
+		PRINT_WARN("cannot load/query segment %s, is a multi-part "
+			   "segment\n", seg_name);
+		break;
+	case -ENOSPC:
+		PRINT_WARN("cannot load/query segment %s, overlaps with "
+			   "storage\n", seg_name);
+		break;
+	case -EBUSY:
+		PRINT_WARN("cannot load/query segment %s, overlaps with "
+			   "already loaded dcss\n", seg_name);
+		break;
+	case -EPERM:
+		PRINT_WARN("cannot load/query segment %s, already loaded in "
+			   "incompatible mode\n", seg_name);
+		break;
+	case -ENOMEM:
+		PRINT_WARN("cannot load/query segment %s, out of memory\n",
+			   seg_name);
+		break;
+	case -ERANGE:
+		PRINT_WARN("cannot load/query segment %s, exceeds kernel "
+			   "mapping range\n", seg_name);
+		break;
+	default:
+		PRINT_WARN("cannot load/query segment %s, return value %i\n",
+			   seg_name, rc);
+		break;
+	}
+}
+
+/*
+ * device attribute for switching shared/nonshared (exclusive)
+ * operation (show + store)
+ */
+static ssize_t
+dcssblk_shared_show(struct device *dev, struct device_attribute *attr, char *buf)
+{
+	struct dcssblk_dev_info *dev_info;
+
+	dev_info = container_of(dev, struct dcssblk_dev_info, dev);
+	return sprintf(buf, dev_info->is_shared ? "1\n" : "0\n");
+}
+
+static ssize_t
+dcssblk_shared_store(struct device *dev, struct device_attribute *attr, const char *inbuf, size_t count)
+{
+	struct dcssblk_dev_info *dev_info;
+	int rc;
+
+	if ((count > 1) && (inbuf[1] != '\n') && (inbuf[1] != '\0')) {
+		PRINT_WARN("Invalid value, must be 0 or 1\n");
+		return -EINVAL;
+	}
+	down_write(&dcssblk_devices_sem);
+	dev_info = container_of(dev, struct dcssblk_dev_info, dev);
+	if (atomic_read(&dev_info->use_count)) {
+		PRINT_ERR("share: segment %s is busy!\n",
+			  dev_info->segment_name);
+		rc = -EBUSY;
+		goto out;
+	}
+	if (inbuf[0] == '1') {
+		// reload segment in shared mode
+		rc = segment_modify_shared(dev_info->segment_name,
+					   SEGMENT_SHARED);
+		if (rc < 0) {
+			BUG_ON(rc == -EINVAL);
+			if (rc != -EAGAIN)
+				goto removeseg;
+		} else {
+			dev_info->is_shared = 1;
+			switch (dev_info->segment_type) {
+				case SEG_TYPE_SR:
+				case SEG_TYPE_ER:
+				case SEG_TYPE_SC:
+					set_disk_ro(dev_info->gd,1);
+			}
+		}
+	} else if (inbuf[0] == '0') {
+		// reload segment in exclusive mode
+		if (dev_info->segment_type == SEG_TYPE_SC) {
+			PRINT_ERR("Segment type SC (%s) cannot be loaded in "
+				  "non-shared mode\n", dev_info->segment_name);
+			rc = -EINVAL;
+			goto out;
+		}
+		rc = segment_modify_shared(dev_info->segment_name,
+					   SEGMENT_EXCLUSIVE);
+		if (rc < 0) {
+			BUG_ON(rc == -EINVAL);
+			if (rc != -EAGAIN)
+				goto removeseg;
+		} else {
+			dev_info->is_shared = 0;
+			set_disk_ro(dev_info->gd, 0);
+		}
+	} else {
+		PRINT_WARN("Invalid value, must be 0 or 1\n");
+		rc = -EINVAL;
+		goto out;
+	}
+	rc = count;
+	goto out;
+
+removeseg:
+	PRINT_ERR("Could not reload segment %s, removing it now!\n",
+			dev_info->segment_name);
+	list_del(&dev_info->lh);
+
+	del_gendisk(dev_info->gd);
+	blk_cleanup_queue(dev_info->dcssblk_queue);
+	dev_info->gd->queue = NULL;
+	put_disk(dev_info->gd);
+	device_unregister(dev);
+	put_device(dev);
+out:
+	up_write(&dcssblk_devices_sem);
+	return rc;
+}
+
+/*
+ * device attribute for save operation on current copy
+ * of the segment. If the segment is busy, saving will
+ * become pending until it gets released, which can be
+ * undone by storing a non-true value to this entry.
+ * (show + store)
+ */
+static ssize_t
+dcssblk_save_show(struct device *dev, struct device_attribute *attr, char *buf)
+{
+	struct dcssblk_dev_info *dev_info;
+
+	dev_info = container_of(dev, struct dcssblk_dev_info, dev);
+	return sprintf(buf, dev_info->save_pending ? "1\n" : "0\n");
+}
+
+static ssize_t
+dcssblk_save_store(struct device *dev, struct device_attribute *attr, const char *inbuf, size_t count)
+{
+	struct dcssblk_dev_info *dev_info;
+
+	if ((count > 1) && (inbuf[1] != '\n') && (inbuf[1] != '\0')) {
+		PRINT_WARN("Invalid value, must be 0 or 1\n");
+		return -EINVAL;
+	}
+	dev_info = container_of(dev, struct dcssblk_dev_info, dev);
+
+	down_write(&dcssblk_devices_sem);
+	if (inbuf[0] == '1') {
+		if (atomic_read(&dev_info->use_count) == 0) {
+			// device is idle => we save immediately
+			PRINT_INFO("Saving segment %s\n",
+				   dev_info->segment_name);
+			segment_save(dev_info->segment_name);
+		}  else {
+			// device is busy => we save it when it becomes
+			// idle in dcssblk_release
+			PRINT_INFO("Segment %s is currently busy, it will "
+				   "be saved when it becomes idle...\n",
+				   dev_info->segment_name);
+			dev_info->save_pending = 1;
+		}
+	} else if (inbuf[0] == '0') {
+		if (dev_info->save_pending) {
+			// device is busy & the user wants to undo his save
+			// request
+			dev_info->save_pending = 0;
+			PRINT_INFO("Pending save for segment %s deactivated\n",
+					dev_info->segment_name);
+		}
+	} else {
+		up_write(&dcssblk_devices_sem);
+		PRINT_WARN("Invalid value, must be 0 or 1\n");
+		return -EINVAL;
+	}
+	up_write(&dcssblk_devices_sem);
+	return count;
+}
+
+/*
+ * device attribute for adding devices
+ */
+static ssize_t
+dcssblk_add_store(struct device *dev, struct device_attribute *attr, const char *buf, size_t count)
+{
+	int rc, i;
+	struct dcssblk_dev_info *dev_info;
+	char *local_buf;
+	unsigned long seg_byte_size;
+
+	dev_info = NULL;
+	if (dev != dcssblk_root_dev) {
+		rc = -EINVAL;
+		goto out_nobuf;
+	}
+	local_buf = kmalloc(count + 1, GFP_KERNEL);
+	if (local_buf == NULL) {
+		rc = -ENOMEM;
+		goto out_nobuf;
+	}
+	/*
+	 * parse input
+	 */
+	for (i = 0; ((buf[i] != '\0') && (buf[i] != '\n') && i < count); i++) {
+		local_buf[i] = toupper(buf[i]);
+	}
+	local_buf[i] = '\0';
+	if ((i == 0) || (i > 8)) {
+		rc = -ENAMETOOLONG;
+		goto out;
+	}
+	/*
+	 * already loaded?
+	 */
+	down_read(&dcssblk_devices_sem);
+	dev_info = dcssblk_get_device_by_name(local_buf);
+	up_read(&dcssblk_devices_sem);
+	if (dev_info != NULL) {
+		PRINT_WARN("Segment %s already loaded!\n", local_buf);
+		rc = -EEXIST;
+		goto out;
+	}
+	/*
+	 * get a struct dcssblk_dev_info
+	 */
+	dev_info = kzalloc(sizeof(struct dcssblk_dev_info), GFP_KERNEL);
+	if (dev_info == NULL) {
+		rc = -ENOMEM;
+		goto out;
+	}
+
+	strcpy(dev_info->segment_name, local_buf);
+	strlcpy(dev_info->dev.bus_id, local_buf, BUS_ID_SIZE);
+	dev_info->dev.release = dcssblk_release_segment;
+	INIT_LIST_HEAD(&dev_info->lh);
+
+	dev_info->gd = alloc_disk(DCSSBLK_MINORS_PER_DISK);
+	if (dev_info->gd == NULL) {
+		rc = -ENOMEM;
+		goto free_dev_info;
+	}
+	dev_info->gd->major = dcssblk_major;
+	dev_info->gd->fops = &dcssblk_devops;
+	dev_info->dcssblk_queue = blk_alloc_queue(GFP_KERNEL);
+	dev_info->gd->queue = dev_info->dcssblk_queue;
+	dev_info->gd->private_data = dev_info;
+	dev_info->gd->driverfs_dev = &dev_info->dev;
+	/*
+	 * load the segment
+	 */
+	rc = segment_load(local_buf, SEGMENT_SHARED,
+				&dev_info->start, &dev_info->end);
+	if (rc < 0) {
+		dcssblk_segment_warn(rc, dev_info->segment_name);
+		goto dealloc_gendisk;
+	}
+	seg_byte_size = (dev_info->end - dev_info->start + 1);
+	set_capacity(dev_info->gd, seg_byte_size >> 9); // size in sectors
+	PRINT_INFO("Loaded segment %s, size = %lu Byte, "
+		   "capacity = %lu (512 Byte) sectors\n", local_buf,
+		   seg_byte_size, seg_byte_size >> 9);
+
+	dev_info->segment_type = rc;
+	dev_info->save_pending = 0;
+	dev_info->is_shared = 1;
+	dev_info->dev.parent = dcssblk_root_dev;
+
+	/*
+	 * get minor, add to list
+	 */
+	down_write(&dcssblk_devices_sem);
+	rc = dcssblk_assign_free_minor(dev_info);
+	if (rc) {
+		up_write(&dcssblk_devices_sem);
+		PRINT_ERR("No free minor number available! "
+			  "Unloading segment...\n");
+		goto unload_seg;
+	}
+	sprintf(dev_info->gd->disk_name, "dcssblk%d",
+		dev_info->gd->first_minor);
+	list_add_tail(&dev_info->lh, &dcssblk_devices);
+
+	if (!try_module_get(THIS_MODULE)) {
+		rc = -ENODEV;
+		goto list_del;
+	}
+	/*
+	 * register the device
+	 */
+	rc = device_register(&dev_info->dev);
+	if (rc) {
+		PRINT_ERR("Segment %s could not be registered RC=%d\n",
+				local_buf, rc);
+		module_put(THIS_MODULE);
+		goto list_del;
+	}
+	get_device(&dev_info->dev);
+	rc = device_create_file(&dev_info->dev, &dev_attr_shared);
+	if (rc)
+		goto unregister_dev;
+	rc = device_create_file(&dev_info->dev, &dev_attr_save);
+	if (rc)
+		goto unregister_dev;
+
+	add_disk(dev_info->gd);
+
+	blk_queue_make_request(dev_info->dcssblk_queue, dcssblk_make_request);
+	blk_queue_hardsect_size(dev_info->dcssblk_queue, 4096);
+
+	switch (dev_info->segment_type) {
+		case SEG_TYPE_SR:
+		case SEG_TYPE_ER:
+		case SEG_TYPE_SC:
+			set_disk_ro(dev_info->gd,1);
+			break;
+		default:
+			set_disk_ro(dev_info->gd,0);
+			break;
+	}
+	PRINT_DEBUG("Segment %s loaded successfully\n", local_buf);
+	up_write(&dcssblk_devices_sem);
+	rc = count;
+	goto out;
+
+unregister_dev:
+	PRINT_ERR("device_create_file() failed!\n");
+	list_del(&dev_info->lh);
+	blk_cleanup_queue(dev_info->dcssblk_queue);
+	dev_info->gd->queue = NULL;
+	put_disk(dev_info->gd);
+	device_unregister(&dev_info->dev);
+	segment_unload(dev_info->segment_name);
+	put_device(&dev_info->dev);
+	up_write(&dcssblk_devices_sem);
+	goto out;
+list_del:
+	list_del(&dev_info->lh);
+	up_write(&dcssblk_devices_sem);
+unload_seg:
+	segment_unload(local_buf);
+dealloc_gendisk:
+	blk_cleanup_queue(dev_info->dcssblk_queue);
+	dev_info->gd->queue = NULL;
+	put_disk(dev_info->gd);
+free_dev_info:
+	kfree(dev_info);
+out:
+	kfree(local_buf);
+out_nobuf:
+	return rc;
+}
+
+/*
+ * device attribute for removing devices
+ */
+static ssize_t
+dcssblk_remove_store(struct device *dev, struct device_attribute *attr, const char *buf, size_t count)
+{
+	struct dcssblk_dev_info *dev_info;
+	int rc, i;
+	char *local_buf;
+
+	if (dev != dcssblk_root_dev) {
+		return -EINVAL;
+	}
+	local_buf = kmalloc(count + 1, GFP_KERNEL);
+	if (local_buf == NULL) {
+		return -ENOMEM;
+	}
+	/*
+	 * parse input
+	 */
+	for (i = 0; ((*(buf+i)!='\0') && (*(buf+i)!='\n') && i < count); i++) {
+		local_buf[i] = toupper(buf[i]);
+	}
+	local_buf[i] = '\0';
+	if ((i == 0) || (i > 8)) {
+		rc = -ENAMETOOLONG;
+		goto out_buf;
+	}
+
+	down_write(&dcssblk_devices_sem);
+	dev_info = dcssblk_get_device_by_name(local_buf);
+	if (dev_info == NULL) {
+		up_write(&dcssblk_devices_sem);
+		PRINT_WARN("Segment %s is not loaded!\n", local_buf);
+		rc = -ENODEV;
+		goto out_buf;
+	}
+	if (atomic_read(&dev_info->use_count) != 0) {
+		up_write(&dcssblk_devices_sem);
+		PRINT_WARN("Segment %s is in use!\n", local_buf);
+		rc = -EBUSY;
+		goto out_buf;
+	}
+	list_del(&dev_info->lh);
+
+	del_gendisk(dev_info->gd);
+	blk_cleanup_queue(dev_info->dcssblk_queue);
+	dev_info->gd->queue = NULL;
+	put_disk(dev_info->gd);
+	device_unregister(&dev_info->dev);
+	segment_unload(dev_info->segment_name);
+	PRINT_DEBUG("Segment %s unloaded successfully\n",
+			dev_info->segment_name);
+	put_device(&dev_info->dev);
+	up_write(&dcssblk_devices_sem);
+
+	rc = count;
+out_buf:
+	kfree(local_buf);
+	return rc;
+}
+
+static int
+dcssblk_open(struct inode *inode, struct file *filp)
+{
+	struct dcssblk_dev_info *dev_info;
+	int rc;
+
+	dev_info = inode->i_bdev->bd_disk->private_data;
+	if (NULL == dev_info) {
+		rc = -ENODEV;
+		goto out;
+	}
+	atomic_inc(&dev_info->use_count);
+	inode->i_bdev->bd_block_size = 4096;
+	rc = 0;
+out:
+	return rc;
+}
+
+static int
+dcssblk_release(struct inode *inode, struct file *filp)
+{
+	struct dcssblk_dev_info *dev_info;
+	int rc;
+
+	dev_info = inode->i_bdev->bd_disk->private_data;
+	if (NULL == dev_info) {
+		rc = -ENODEV;
+		goto out;
+	}
+	down_write(&dcssblk_devices_sem);
+	if (atomic_dec_and_test(&dev_info->use_count)
+	    && (dev_info->save_pending)) {
+		PRINT_INFO("Segment %s became idle and is being saved now\n",
+			    dev_info->segment_name);
+		segment_save(dev_info->segment_name);
+		dev_info->save_pending = 0;
+	}
+	up_write(&dcssblk_devices_sem);
+	rc = 0;
+out:
+	return rc;
+}
+
+static int
+dcssblk_make_request(request_queue_t *q, struct bio *bio)
+{
+	struct dcssblk_dev_info *dev_info;
+	struct bio_vec *bvec;
+	unsigned long index;
+	unsigned long page_addr;
+	unsigned long source_addr;
+	unsigned long bytes_done;
+	int i;
+
+	bytes_done = 0;
+	dev_info = bio->bi_bdev->bd_disk->private_data;
+	if (dev_info == NULL)
+		goto fail;
+	if ((bio->bi_sector & 7) != 0 || (bio->bi_size & 4095) != 0)
+		/* Request is not page-aligned. */
+		goto fail;
+	if (((bio->bi_size >> 9) + bio->bi_sector)
+			> get_capacity(bio->bi_bdev->bd_disk)) {
+		/* Request beyond end of DCSS segment. */
+		goto fail;
+	}
+	/* verify data transfer direction */
+	if (dev_info->is_shared) {
+		switch (dev_info->segment_type) {
+		case SEG_TYPE_SR:
+		case SEG_TYPE_ER:
+		case SEG_TYPE_SC:
+			/* cannot write to these segments */
+			if (bio_data_dir(bio) == WRITE) {
+				PRINT_WARN("rejecting write to ro segment %s\n", dev_info->dev.bus_id);
+				goto fail;
+			}
+		}
+	}
+
+	index = (bio->bi_sector >> 3);
+	bio_for_each_segment(bvec, bio, i) {
+		page_addr = (unsigned long)
+			page_address(bvec->bv_page) + bvec->bv_offset;
+		source_addr = dev_info->start + (index<<12) + bytes_done;
+		if (unlikely(page_addr & 4095) != 0 || (bvec->bv_len & 4095) != 0)
+			// More paranoia.
+			goto fail;
+		if (bio_data_dir(bio) == READ) {
+			memcpy((void*)page_addr, (void*)source_addr,
+				bvec->bv_len);
+		} else {
+			memcpy((void*)source_addr, (void*)page_addr,
+				bvec->bv_len);
+		}
+		bytes_done += bvec->bv_len;
+	}
+	bio_endio(bio, bytes_done, 0);
+	return 0;
+fail:
+	bio_io_error(bio, bio->bi_size);
+	return 0;
+}
+
+static int
+dcssblk_direct_access (struct block_device *bdev, sector_t secnum,
+			unsigned long *data)
+{
+	struct dcssblk_dev_info *dev_info;
+	unsigned long pgoff;
+
+	dev_info = bdev->bd_disk->private_data;
+	if (!dev_info)
+		return -ENODEV;
+	if (secnum % (PAGE_SIZE/512))
+		return -EINVAL;
+	pgoff = secnum / (PAGE_SIZE / 512);
+	if ((pgoff+1)*PAGE_SIZE-1 > dev_info->end - dev_info->start)
+		return -ERANGE;
+	*data = (unsigned long) (dev_info->start+pgoff*PAGE_SIZE);
+	return 0;
+}
+
+static void
+dcssblk_check_params(void)
+{
+	int rc, i, j, k;
+	char buf[9];
+	struct dcssblk_dev_info *dev_info;
+
+	for (i = 0; (i < DCSSBLK_PARM_LEN) && (dcssblk_segments[i] != '\0');
+	     i++) {
+		for (j = i; (dcssblk_segments[j] != ',')  &&
+			    (dcssblk_segments[j] != '\0') &&
+			    (dcssblk_segments[j] != '(')  &&
+			    (j - i) < 8; j++)
+		{
+			buf[j-i] = dcssblk_segments[j];
+		}
+		buf[j-i] = '\0';
+		rc = dcssblk_add_store(dcssblk_root_dev, NULL, buf, j-i);
+		if ((rc >= 0) && (dcssblk_segments[j] == '(')) {
+			for (k = 0; buf[k] != '\0'; k++)
+				buf[k] = toupper(buf[k]);
+			if (!strncmp(&dcssblk_segments[j], "(local)", 7)) {
+				down_read(&dcssblk_devices_sem);
+				dev_info = dcssblk_get_device_by_name(buf);
+				up_read(&dcssblk_devices_sem);
+				if (dev_info)
+					dcssblk_shared_store(&dev_info->dev,
+							     NULL, "0\n", 2);
+			}
+		}
+		while ((dcssblk_segments[j] != ',') &&
+		       (dcssblk_segments[j] != '\0'))
+		{
+			j++;
+		}
+		if (dcssblk_segments[j] == '\0')
+			break;
+		i = j;
+	}
+}
+
+/*
+ * The init/exit functions.
+ */
+static void __exit
+dcssblk_exit(void)
+{
+	int rc;
+
+	PRINT_DEBUG("DCSSBLOCK EXIT...\n");
+	s390_root_dev_unregister(dcssblk_root_dev);
+	rc = unregister_blkdev(dcssblk_major, DCSSBLK_NAME);
+	if (rc) {
+		PRINT_ERR("unregister_blkdev() failed!\n");
+	}
+	PRINT_DEBUG("...finished!\n");
+}
+
+static int __init
+dcssblk_init(void)
+{
+	int rc;
+
+	PRINT_DEBUG("DCSSBLOCK INIT...\n");
+	dcssblk_root_dev = s390_root_dev_register("dcssblk");
+	if (IS_ERR(dcssblk_root_dev)) {
+		PRINT_ERR("device_register() failed!\n");
+		return PTR_ERR(dcssblk_root_dev);
+	}
+	rc = device_create_file(dcssblk_root_dev, &dev_attr_add);
+	if (rc) {
+		PRINT_ERR("device_create_file(add) failed!\n");
+		s390_root_dev_unregister(dcssblk_root_dev);
+		return rc;
+	}
+	rc = device_create_file(dcssblk_root_dev, &dev_attr_remove);
+	if (rc) {
+		PRINT_ERR("device_create_file(remove) failed!\n");
+		s390_root_dev_unregister(dcssblk_root_dev);
+		return rc;
+	}
+	rc = register_blkdev(0, DCSSBLK_NAME);
+	if (rc < 0) {
+		PRINT_ERR("Can't get dynamic major!\n");
+		s390_root_dev_unregister(dcssblk_root_dev);
+		return rc;
+	}
+	dcssblk_major = rc;
+	init_rwsem(&dcssblk_devices_sem);
+
+	dcssblk_check_params();
+
+	PRINT_DEBUG("...finished!\n");
+	return 0;
+}
+
+module_init(dcssblk_init);
+module_exit(dcssblk_exit);
+
+module_param_string(segments, dcssblk_segments, DCSSBLK_PARM_LEN, 0444);
+MODULE_PARM_DESC(segments, "Name of DCSS segment(s) to be loaded, "
+		 "comma-separated list, each name max. 8 chars.\n"
+		 "Adding \"(local)\" to segment name equals echoing 0 to "
+		 "/sys/devices/dcssblk/<segment name>/shared after loading "
+		 "the segment - \n"
+		 "e.g. segments=\"mydcss1,mydcss2,mydcss3(local)\"");
+
+MODULE_LICENSE("GPL");

drivers/s390/block/xpram.c

@@ -0,0 +1,440 @@
+/*
+ * Xpram.c -- the S/390 expanded memory RAM-disk
+ *           
+ * significant parts of this code are based on
+ * the sbull device driver presented in
+ * A. Rubini: Linux Device Drivers
+ *
+ * Author of XPRAM specific coding: Reinhard Buendgen
+ *                                  buendgen@de.ibm.com
+ * Rewrite for 2.5: Martin Schwidefsky <schwidefsky@de.ibm.com>
+ *
+ * External interfaces:
+ *   Interfaces to linux kernel
+ *        xpram_setup: read kernel parameters
+ *   Device specific file operations
+ *        xpram_iotcl
+ *        xpram_open
+ *
+ * "ad-hoc" partitioning:
+ *    the expanded memory can be partitioned among several devices 
+ *    (with different minors). The partitioning set up can be
+ *    set by kernel or module parameters (int devs & int sizes[])
+ *
+ * Potential future improvements:
+ *   generic hard disk support to replace ad-hoc partitioning
+ */
+
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/ctype.h>  /* isdigit, isxdigit */
+#include <linux/errno.h>
+#include <linux/init.h>
+#include <linux/slab.h>
+#include <linux/blkdev.h>
+#include <linux/blkpg.h>
+#include <linux/hdreg.h>  /* HDIO_GETGEO */
+#include <linux/sysdev.h>
+#include <linux/bio.h>
+#include <asm/uaccess.h>
+
+#define XPRAM_NAME	"xpram"
+#define XPRAM_DEVS	1	/* one partition */
+#define XPRAM_MAX_DEVS	32	/* maximal number of devices (partitions) */
+
+#define PRINT_DEBUG(x...)	printk(KERN_DEBUG XPRAM_NAME " debug:" x)
+#define PRINT_INFO(x...)	printk(KERN_INFO XPRAM_NAME " info:" x)
+#define PRINT_WARN(x...)	printk(KERN_WARNING XPRAM_NAME " warning:" x)
+#define PRINT_ERR(x...)		printk(KERN_ERR XPRAM_NAME " error:" x)
+
+
+typedef struct {
+	unsigned int	size;		/* size of xpram segment in pages */
+	unsigned int	offset;		/* start page of xpram segment */
+} xpram_device_t;
+
+static xpram_device_t xpram_devices[XPRAM_MAX_DEVS];
+static unsigned int xpram_sizes[XPRAM_MAX_DEVS];
+static struct gendisk *xpram_disks[XPRAM_MAX_DEVS];
+static unsigned int xpram_pages;
+static int xpram_devs;
+
+/*
+ * Parameter parsing functions.
+ */
+static int __initdata devs = XPRAM_DEVS;
+static char __initdata *sizes[XPRAM_MAX_DEVS];
+
+module_param(devs, int, 0);
+module_param_array(sizes, charp, NULL, 0);
+
+MODULE_PARM_DESC(devs, "number of devices (\"partitions\"), " \
+		 "the default is " __MODULE_STRING(XPRAM_DEVS) "\n");
+MODULE_PARM_DESC(sizes, "list of device (partition) sizes " \
+		 "the defaults are 0s \n" \
+		 "All devices with size 0 equally partition the "
+		 "remaining space on the expanded strorage not "
+		 "claimed by explicit sizes\n");
+MODULE_LICENSE("GPL");
+
+/*
+ * Copy expanded memory page (4kB) into main memory                  
+ * Arguments                                                         
+ *           page_addr:    address of target page                    
+ *           xpage_index:  index of expandeded memory page           
+ * Return value                                                      
+ *           0:            if operation succeeds
+ *           -EIO:         if pgin failed
+ *           -ENXIO:       if xpram has vanished
+ */
+static int xpram_page_in (unsigned long page_addr, unsigned int xpage_index)
+{
+	int cc = 2;	/* return unused cc 2 if pgin traps */
+
+	asm volatile(
+		"	.insn	rre,0xb22e0000,%1,%2\n"  /* pgin %1,%2 */
+		"0:	ipm	%0\n"
+		"	srl	%0,28\n"
+		"1:\n"
+		EX_TABLE(0b,1b)
+		: "+d" (cc) : "a" (__pa(page_addr)), "d" (xpage_index) : "cc");
+	if (cc == 3)
+		return -ENXIO;
+	if (cc == 2) {
+		PRINT_ERR("expanded storage lost!\n");
+		return -ENXIO;
+	}
+	if (cc == 1) {
+		PRINT_ERR("page in failed for page index %u.\n",
+			  xpage_index);
+		return -EIO;
+	}
+	return 0;
+}
+
+/*
+ * Copy a 4kB page of main memory to an expanded memory page          
+ * Arguments                                                          
+ *           page_addr:    address of source page                     
+ *           xpage_index:  index of expandeded memory page            
+ * Return value                                                       
+ *           0:            if operation succeeds
+ *           -EIO:         if pgout failed
+ *           -ENXIO:       if xpram has vanished
+ */
+static long xpram_page_out (unsigned long page_addr, unsigned int xpage_index)
+{
+	int cc = 2;	/* return unused cc 2 if pgin traps */
+
+	asm volatile(
+		"	.insn	rre,0xb22f0000,%1,%2\n"  /* pgout %1,%2 */
+		"0:	ipm	%0\n"
+		"	srl	%0,28\n"
+		"1:\n"
+		EX_TABLE(0b,1b)
+		: "+d" (cc) : "a" (__pa(page_addr)), "d" (xpage_index) : "cc");
+	if (cc == 3)
+		return -ENXIO;
+	if (cc == 2) {
+		PRINT_ERR("expanded storage lost!\n");
+		return -ENXIO;
+	}
+	if (cc == 1) {
+		PRINT_ERR("page out failed for page index %u.\n",
+			  xpage_index);
+		return -EIO;
+	}
+	return 0;
+}
+
+/*
+ * Check if xpram is available.
+ */
+static int __init xpram_present(void)
+{
+	unsigned long mem_page;
+	int rc;
+
+	mem_page = (unsigned long) __get_free_page(GFP_KERNEL);
+	if (!mem_page)
+		return -ENOMEM;
+	rc = xpram_page_in(mem_page, 0);
+	free_page(mem_page);
+	return rc ? -ENXIO : 0;
+}
+
+/*
+ * Return index of the last available xpram page.
+ */
+static unsigned long __init xpram_highest_page_index(void)
+{
+	unsigned int page_index, add_bit;
+	unsigned long mem_page;
+
+	mem_page = (unsigned long) __get_free_page(GFP_KERNEL);
+	if (!mem_page)
+		return 0;
+
+	page_index = 0;
+	add_bit = 1ULL << (sizeof(unsigned int)*8 - 1);
+	while (add_bit > 0) {
+		if (xpram_page_in(mem_page, page_index | add_bit) == 0)
+			page_index |= add_bit;
+		add_bit >>= 1;
+	}
+
+	free_page (mem_page);
+
+	return page_index;
+}
+
+/*
+ * Block device make request function.
+ */
+static int xpram_make_request(request_queue_t *q, struct bio *bio)
+{
+	xpram_device_t *xdev = bio->bi_bdev->bd_disk->private_data;
+	struct bio_vec *bvec;
+	unsigned int index;
+	unsigned long page_addr;
+	unsigned long bytes;
+	int i;
+
+	if ((bio->bi_sector & 7) != 0 || (bio->bi_size & 4095) != 0)
+		/* Request is not page-aligned. */
+		goto fail;
+	if ((bio->bi_size >> 12) > xdev->size)
+		/* Request size is no page-aligned. */
+		goto fail;
+	if ((bio->bi_sector >> 3) > 0xffffffffU - xdev->offset)
+		goto fail;
+	index = (bio->bi_sector >> 3) + xdev->offset;
+	bio_for_each_segment(bvec, bio, i) {
+		page_addr = (unsigned long)
+			kmap(bvec->bv_page) + bvec->bv_offset;
+		bytes = bvec->bv_len;
+		if ((page_addr & 4095) != 0 || (bytes & 4095) != 0)
+			/* More paranoia. */
+			goto fail;
+		while (bytes > 0) {
+			if (bio_data_dir(bio) == READ) {
+				if (xpram_page_in(page_addr, index) != 0)
+					goto fail;
+			} else {
+				if (xpram_page_out(page_addr, index) != 0)
+					goto fail;
+			}
+			page_addr += 4096;
+			bytes -= 4096;
+			index++;
+		}
+	}
+	set_bit(BIO_UPTODATE, &bio->bi_flags);
+	bytes = bio->bi_size;
+	bio->bi_size = 0;
+	bio->bi_end_io(bio, bytes, 0);
+	return 0;
+fail:
+	bio_io_error(bio, bio->bi_size);
+	return 0;
+}
+
+static int xpram_getgeo(struct block_device *bdev, struct hd_geometry *geo)
+{
+	unsigned long size;
+
+	/*
+	 * get geometry: we have to fake one...  trim the size to a
+	 * multiple of 64 (32k): tell we have 16 sectors, 4 heads,
+	 * whatever cylinders. Tell also that data starts at sector. 4.
+	 */
+	size = (xpram_pages * 8) & ~0x3f;
+	geo->cylinders = size >> 6;
+	geo->heads = 4;
+	geo->sectors = 16;
+	geo->start = 4;
+	return 0;
+}
+
+static struct block_device_operations xpram_devops =
+{
+	.owner	= THIS_MODULE,
+	.getgeo	= xpram_getgeo,
+};
+
+/*
+ * Setup xpram_sizes array.
+ */
+static int __init xpram_setup_sizes(unsigned long pages)
+{
+	unsigned long mem_needed;
+	unsigned long mem_auto;
+	unsigned long long size;
+	int mem_auto_no;
+	int i;
+
+	/* Check number of devices. */
+	if (devs <= 0 || devs > XPRAM_MAX_DEVS) {
+		PRINT_ERR("invalid number %d of devices\n",devs);
+		return -EINVAL;
+	}
+	xpram_devs = devs;
+
+	/*
+	 * Copy sizes array to xpram_sizes and align partition
+	 * sizes to page boundary.
+	 */
+	mem_needed = 0;
+	mem_auto_no = 0;
+	for (i = 0; i < xpram_devs; i++) {
+		if (sizes[i]) {
+			size = simple_strtoull(sizes[i], &sizes[i], 0);
+			switch (sizes[i][0]) {
+			case 'g':
+			case 'G':
+				size <<= 20;
+				break;
+			case 'm':
+			case 'M':
+				size <<= 10;
+			}
+			xpram_sizes[i] = (size + 3) & -4UL;
+		}
+		if (xpram_sizes[i])
+			mem_needed += xpram_sizes[i];
+		else
+			mem_auto_no++;
+	}
+	
+	PRINT_INFO("  number of devices (partitions): %d \n", xpram_devs);
+	for (i = 0; i < xpram_devs; i++) {
+		if (xpram_sizes[i])
+			PRINT_INFO("  size of partition %d: %u kB\n",
+				   i, xpram_sizes[i]);
+		else
+			PRINT_INFO("  size of partition %d to be set "
+				   "automatically\n",i);
+	}
+	PRINT_DEBUG("  memory needed (for sized partitions): %lu kB\n",
+		    mem_needed);
+	PRINT_DEBUG("  partitions to be sized automatically: %d\n",
+		    mem_auto_no);
+
+	if (mem_needed > pages * 4) {
+		PRINT_ERR("Not enough expanded memory available\n");
+		return -EINVAL;
+	}
+
+	/*
+	 * partitioning:
+	 * xpram_sizes[i] != 0; partition i has size xpram_sizes[i] kB
+	 * else:             ; all partitions with zero xpram_sizes[i]
+	 *                     partition equally the remaining space
+	 */
+	if (mem_auto_no) {
+		mem_auto = ((pages - mem_needed / 4) / mem_auto_no) * 4;
+		PRINT_INFO("  automatically determined "
+			   "partition size: %lu kB\n", mem_auto);
+		for (i = 0; i < xpram_devs; i++)
+			if (xpram_sizes[i] == 0)
+				xpram_sizes[i] = mem_auto;
+	}
+	return 0;
+}
+
+static struct request_queue *xpram_queue;
+
+static int __init xpram_setup_blkdev(void)
+{
+	unsigned long offset;
+	int i, rc = -ENOMEM;
+
+	for (i = 0; i < xpram_devs; i++) {
+		struct gendisk *disk = alloc_disk(1);
+		if (!disk)
+			goto out;
+		xpram_disks[i] = disk;
+	}
+
+	/*
+	 * Register xpram major.
+	 */
+	rc = register_blkdev(XPRAM_MAJOR, XPRAM_NAME);
+	if (rc < 0)
+		goto out;
+
+	/*
+	 * Assign the other needed values: make request function, sizes and
+	 * hardsect size. All the minor devices feature the same value.
+	 */
+	xpram_queue = blk_alloc_queue(GFP_KERNEL);
+	if (!xpram_queue) {
+		rc = -ENOMEM;
+		goto out_unreg;
+	}
+	blk_queue_make_request(xpram_queue, xpram_make_request);
+	blk_queue_hardsect_size(xpram_queue, 4096);
+
+	/*
+	 * Setup device structures.
+	 */
+	offset = 0;
+	for (i = 0; i < xpram_devs; i++) {
+		struct gendisk *disk = xpram_disks[i];
+
+		xpram_devices[i].size = xpram_sizes[i] / 4;
+		xpram_devices[i].offset = offset;
+		offset += xpram_devices[i].size;
+		disk->major = XPRAM_MAJOR;
+		disk->first_minor = i;
+		disk->fops = &xpram_devops;
+		disk->private_data = &xpram_devices[i];
+		disk->queue = xpram_queue;
+		sprintf(disk->disk_name, "slram%d", i);
+		set_capacity(disk, xpram_sizes[i] << 1);
+		add_disk(disk);
+	}
+
+	return 0;
+out_unreg:
+	unregister_blkdev(XPRAM_MAJOR, XPRAM_NAME);
+out:
+	while (i--)
+		put_disk(xpram_disks[i]);
+	return rc;
+}
+
+/*
+ * Finally, the init/exit functions.
+ */
+static void __exit xpram_exit(void)
+{
+	int i;
+	for (i = 0; i < xpram_devs; i++) {
+		del_gendisk(xpram_disks[i]);
+		put_disk(xpram_disks[i]);
+	}
+	unregister_blkdev(XPRAM_MAJOR, XPRAM_NAME);
+	blk_cleanup_queue(xpram_queue);
+}
+
+static int __init xpram_init(void)
+{
+	int rc;
+
+	/* Find out size of expanded memory. */
+	if (xpram_present() != 0) {
+		PRINT_WARN("No expanded memory available\n");
+		return -ENODEV;
+	}
+	xpram_pages = xpram_highest_page_index() + 1;
+	PRINT_INFO("  %u pages expanded memory found (%lu KB).\n",
+		   xpram_pages, (unsigned long) xpram_pages*4);
+	rc = xpram_setup_sizes(xpram_pages);
+	if (rc)
+		return rc;
+	return xpram_setup_blkdev();
+}
+
+module_init(xpram_init);
+module_exit(xpram_exit);