WSL2-Linux-Kernel/drivers/s390/block/dasd_diag.c

636 строки
17 KiB
C
Исходник Обычный вид История

/*
* 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 <asm/diag.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 ccw_dev_id dev_id;
};
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 = private->dev_id.devno;
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 = private->dev_id.devno;
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);
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
rc = mdsk_init_io(device, device->block->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;
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
device = cqr->startdev;
if (cqr->retries < 0) {
DEV_MESSAGE(KERN_WARNING, device, "DIAG start_IO: request %p "
"- no retry left)", cqr);
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
cqr->status = DASD_CQR_ERROR;
return -EIO;
}
private = (struct dasd_diag_private *) device->private;
dreq = (struct dasd_diag_req *) cqr->data;
private->iob.dev_nr = private->dev_id.devno;
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();
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
cqr->status = DASD_CQR_SUCCESS;
/* 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;
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
device = cqr->startdev;
mdsk_term_io(device);
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
mdsk_init_io(device, device->block->bp_block, 0, NULL);
cqr->status = DASD_CQR_CLEAR_PENDING;
cqr->stopclk = get_clock();
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
dasd_schedule_device_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;
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
device = (struct dasd_device *) cqr->startdev;
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 */
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
if (cqr->status == DASD_CQR_CLEAR_PENDING) {
cqr->status = DASD_CQR_CLEARED;
dasd_device_clear_timer(device);
dasd_schedule_device_bh(device);
spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
return;
}
cqr->stopclk = get_clock();
expires = 0;
if (status == 0) {
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
cqr->status = DASD_CQR_SUCCESS;
/* Start first request on queue if possible -> fast_io. */
if (!list_empty(&device->ccw_queue)) {
next = list_entry(device->ccw_queue.next,
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
struct dasd_ccw_req, devlist);
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)
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
dasd_device_set_timer(device, expires);
else
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
dasd_device_clear_timer(device);
dasd_schedule_device_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)
{
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
struct dasd_block *block;
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;
}
ccw_device_get_id(device->cdev, &private->dev_id);
device->private = (void *) private;
}
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
block = dasd_alloc_block();
if (IS_ERR(block)) {
DEV_MESSAGE(KERN_WARNING, device, "%s",
"could not allocate dasd block structure");
kfree(device->private);
return PTR_ERR(block);
}
device->block = block;
block->base = device;
/* Read Device Characteristics */
rdc_data = (void *) &(private->rdc_data);
rdc_data->dev_nr = private->dev_id.devno;
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;
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
block->blocks = (unsigned long) label->block_count;
} else
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
block->blocks = end_block;
block->bp_block = bsize;
block->s2b_shift = 0; /* bits to shift 512 to get a block */
for (sb = 512; sb < bsize; sb = sb << 1)
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
block->s2b_shift++;
rc = mdsk_init_io(device, block->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",
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
(unsigned long) block->bp_block,
(unsigned long) (block->blocks <<
block->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
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
dasd_diag_fill_geometry(struct dasd_block *block, struct hd_geometry *geo)
{
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
if (dasd_check_blocksize(block->bp_block) != 0)
return -EINVAL;
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
geo->cylinders = (block->blocks << block->s2b_shift) >> 10;
geo->heads = 16;
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
geo->sectors = 128 >> block->s2b_shift;
return 0;
}
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. */
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
static struct dasd_ccw_req *dasd_diag_build_cp(struct dasd_device *memdev,
struct dasd_block *block,
struct request *req)
{
struct dasd_ccw_req *cqr;
struct dasd_diag_req *dreq;
struct dasd_diag_bio *dbio;
struct req_iterator iter;
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;
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);
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
blksize = block->bp_block;
/* Calculate record id of first and last block. */
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
first_rec = req->sector >> block->s2b_shift;
last_rec = (req->sector + req->nr_sectors - 1) >> block->s2b_shift;
/* Check struct bio and count the number of blocks for the request. */
count = 0;
rq_for_each_segment(bv, req, iter) {
if (bv->bv_len & (blksize - 1))
/* Fba can only do full blocks. */
return ERR_PTR(-EINVAL);
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
count += bv->bv_len >> (block->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,
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
datasize, memdev);
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_segment(bv, req, iter) {
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);
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
cqr->startdev = memdev;
cqr->memdev = memdev;
cqr->block = block;
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;
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
dasd_sfree_request(cqr, cqr->memdev);
return status;
}
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
static void dasd_diag_handle_terminated_request(struct dasd_ccw_req *cqr)
{
cqr->status = DASD_CQR_FILLED;
};
/* 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,
[S390] dasd: add hyper PAV support to DASD device driver, part 1 Parallel access volumes (PAV) is a storage server feature, that allows to start multiple channel programs on the same DASD in parallel. It defines alias devices which can be used as alternative paths to the same disk. With the old base PAV support we only needed rudimentary functionality in the DASD device driver. As the mapping between base and alias devices was static, we just had to export an identifier (uid) and could leave the combining of devices to external layers like a device mapper multipath. Now hyper PAV removes the requirement to dedicate alias devices to specific base devices. Instead each alias devices can be combined with multiple base device on a per request basis. This requires full support by the DASD device driver as now each channel program itself has to identify the target base device. The changes to the dasd device driver and the ECKD discipline are: - Separate subchannel device representation (dasd_device) from block device representation (dasd_block). Only base devices are block devices. - Gather information about base and alias devices and possible combinations. - For each request decide which dasd_device should be used (base or alias) and build specific channel program. - Support summary unit checks, which allow the storage server to upgrade / downgrade between base and hyper PAV at runtime (support is mandatory). Signed-off-by: Stefan Weinhuber <wein@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2008-01-26 16:11:23 +03:00
.handle_terminated_request = dasd_diag_handle_terminated_request,
.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);