[S390] tape: cleanup reference counting

[net-next-2.6.git] / drivers / s390 / char / tape_core.c

/*
 *  drivers/s390/char/tape_core.c
 *    basic function of the tape device driver
 *
 *  S390 and zSeries version
 *    Copyright IBM Corp. 2001, 2009
 *    Author(s): Carsten Otte <cotte@de.ibm.com>
 *		 Michael Holzheu <holzheu@de.ibm.com>
 *		 Tuan Ngo-Anh <ngoanh@de.ibm.com>
 *		 Martin Schwidefsky <schwidefsky@de.ibm.com>
 *		 Stefan Bader <shbader@de.ibm.com>
 */

#define KMSG_COMPONENT "tape"
#include <linux/module.h>
#include <linux/init.h>	     // for kernel parameters
#include <linux/kmod.h>	     // for requesting modules
#include <linux/spinlock.h>  // for locks
#include <linux/vmalloc.h>
#include <linux/list.h>

#include <asm/types.h>	     // for variable types

#define TAPE_DBF_AREA	tape_core_dbf

#include "tape.h"
#include "tape_std.h"

#define LONG_BUSY_TIMEOUT 180 /* seconds */

static void __tape_do_irq (struct ccw_device *, unsigned long, struct irb *);
static void tape_delayed_next_request(struct work_struct *);
static void tape_long_busy_timeout(unsigned long data);

/*
 * One list to contain all tape devices of all disciplines, so
 * we can assign the devices to minor numbers of the same major
 * The list is protected by the rwlock
 */
static LIST_HEAD(tape_device_list);
static DEFINE_RWLOCK(tape_device_lock);

/*
 * Pointer to debug area.
 */
debug_info_t *TAPE_DBF_AREA = NULL;
EXPORT_SYMBOL(TAPE_DBF_AREA);

/*
 * Printable strings for tape enumerations.
 */
const char *tape_state_verbose[TS_SIZE] =
{
	[TS_UNUSED]   = "UNUSED",
	[TS_IN_USE]   = "IN_USE",
	[TS_BLKUSE]   = "BLKUSE",
	[TS_INIT]     = "INIT  ",
	[TS_NOT_OPER] = "NOT_OP"
};

const char *tape_op_verbose[TO_SIZE] =
{
	[TO_BLOCK] = "BLK",	[TO_BSB] = "BSB",
	[TO_BSF] = "BSF",	[TO_DSE] = "DSE",
	[TO_FSB] = "FSB",	[TO_FSF] = "FSF",
	[TO_LBL] = "LBL",	[TO_NOP] = "NOP",
	[TO_RBA] = "RBA",	[TO_RBI] = "RBI",
	[TO_RFO] = "RFO",	[TO_REW] = "REW",
	[TO_RUN] = "RUN",	[TO_WRI] = "WRI",
	[TO_WTM] = "WTM",	[TO_MSEN] = "MSN",
	[TO_LOAD] = "LOA",	[TO_READ_CONFIG] = "RCF",
	[TO_READ_ATTMSG] = "RAT",
	[TO_DIS] = "DIS",	[TO_ASSIGN] = "ASS",
	[TO_UNASSIGN] = "UAS",  [TO_CRYPT_ON] = "CON",
	[TO_CRYPT_OFF] = "COF",	[TO_KEKL_SET] = "KLS",
	[TO_KEKL_QUERY] = "KLQ",[TO_RDC] = "RDC",
};

static int devid_to_int(struct ccw_dev_id *dev_id)
{
	return dev_id->devno + (dev_id->ssid << 16);
}

/*
 * Some channel attached tape specific attributes.
 *
 * FIXME: In the future the first_minor and blocksize attribute should be
 *        replaced by a link to the cdev tree.
 */
static ssize_t
tape_medium_state_show(struct device *dev, struct device_attribute *attr, char *buf)
{
	struct tape_device *tdev;

	tdev = dev_get_drvdata(dev);
	return scnprintf(buf, PAGE_SIZE, "%i\n", tdev->medium_state);
}

static
DEVICE_ATTR(medium_state, 0444, tape_medium_state_show, NULL);

static ssize_t
tape_first_minor_show(struct device *dev, struct device_attribute *attr, char *buf)
{
	struct tape_device *tdev;

	tdev = dev_get_drvdata(dev);
	return scnprintf(buf, PAGE_SIZE, "%i\n", tdev->first_minor);
}

static
DEVICE_ATTR(first_minor, 0444, tape_first_minor_show, NULL);

static ssize_t
tape_state_show(struct device *dev, struct device_attribute *attr, char *buf)
{
	struct tape_device *tdev;

	tdev = dev_get_drvdata(dev);
	return scnprintf(buf, PAGE_SIZE, "%s\n", (tdev->first_minor < 0) ?
		"OFFLINE" : tape_state_verbose[tdev->tape_state]);
}

static
DEVICE_ATTR(state, 0444, tape_state_show, NULL);

static ssize_t
tape_operation_show(struct device *dev, struct device_attribute *attr, char *buf)
{
	struct tape_device *tdev;
	ssize_t rc;

	tdev = dev_get_drvdata(dev);
	if (tdev->first_minor < 0)
		return scnprintf(buf, PAGE_SIZE, "N/A\n");

	spin_lock_irq(get_ccwdev_lock(tdev->cdev));
	if (list_empty(&tdev->req_queue))
		rc = scnprintf(buf, PAGE_SIZE, "---\n");
	else {
		struct tape_request *req;

		req = list_entry(tdev->req_queue.next, struct tape_request,
			list);
		rc = scnprintf(buf,PAGE_SIZE, "%s\n", tape_op_verbose[req->op]);
	}
	spin_unlock_irq(get_ccwdev_lock(tdev->cdev));
	return rc;
}

static
DEVICE_ATTR(operation, 0444, tape_operation_show, NULL);

static ssize_t
tape_blocksize_show(struct device *dev, struct device_attribute *attr, char *buf)
{
	struct tape_device *tdev;

	tdev = dev_get_drvdata(dev);

	return scnprintf(buf, PAGE_SIZE, "%i\n", tdev->char_data.block_size);
}

static
DEVICE_ATTR(blocksize, 0444, tape_blocksize_show, NULL);

static struct attribute *tape_attrs[] = {
	&dev_attr_medium_state.attr,
	&dev_attr_first_minor.attr,
	&dev_attr_state.attr,
	&dev_attr_operation.attr,
	&dev_attr_blocksize.attr,
	NULL
};

static struct attribute_group tape_attr_group = {
	.attrs = tape_attrs,
};

/*
 * Tape state functions
 */
void
tape_state_set(struct tape_device *device, enum tape_state newstate)
{
	const char *str;

	if (device->tape_state == TS_NOT_OPER) {
		DBF_EVENT(3, "ts_set err: not oper\n");
		return;
	}
	DBF_EVENT(4, "ts. dev:	%x\n", device->first_minor);
	DBF_EVENT(4, "old ts:\t\n");
	if (device->tape_state < TS_SIZE && device->tape_state >=0 )
		str = tape_state_verbose[device->tape_state];
	else
		str = "UNKNOWN TS";
	DBF_EVENT(4, "%s\n", str);
	DBF_EVENT(4, "new ts:\t\n");
	if (newstate < TS_SIZE && newstate >= 0)
		str = tape_state_verbose[newstate];
	else
		str = "UNKNOWN TS";
	DBF_EVENT(4, "%s\n", str);
	device->tape_state = newstate;
	wake_up(&device->state_change_wq);
}

void
tape_med_state_set(struct tape_device *device, enum tape_medium_state newstate)
{
	if (device->medium_state == newstate)
		return;
	switch(newstate){
	case MS_UNLOADED:
		device->tape_generic_status |= GMT_DR_OPEN(~0);
		if (device->medium_state == MS_LOADED)
			pr_info("%s: The tape cartridge has been successfully "
				"unloaded\n", dev_name(&device->cdev->dev));
		break;
	case MS_LOADED:
		device->tape_generic_status &= ~GMT_DR_OPEN(~0);
		if (device->medium_state == MS_UNLOADED)
			pr_info("%s: A tape cartridge has been mounted\n",
				dev_name(&device->cdev->dev));
		break;
	default:
		// print nothing
		break;
	}
	device->medium_state = newstate;
	wake_up(&device->state_change_wq);
}

/*
 * Stop running ccw. Has to be called with the device lock held.
 */
static int
__tape_cancel_io(struct tape_device *device, struct tape_request *request)
{
	int retries;
	int rc;

	/* Check if interrupt has already been processed */
	if (request->callback == NULL)
		return 0;

	rc = 0;
	for (retries = 0; retries < 5; retries++) {
		rc = ccw_device_clear(device->cdev, (long) request);

		switch (rc) {
			case 0:
				request->status	= TAPE_REQUEST_DONE;
				return 0;
			case -EBUSY:
				request->status	= TAPE_REQUEST_CANCEL;
				schedule_delayed_work(&device->tape_dnr, 0);
				return 0;
			case -ENODEV:
				DBF_EXCEPTION(2, "device gone, retry\n");
				break;
			case -EIO:
				DBF_EXCEPTION(2, "I/O error, retry\n");
				break;
			default:
				BUG();
		}
	}

	return rc;
}

/*
 * Add device into the sorted list, giving it the first
 * available minor number.
 */
static int
tape_assign_minor(struct tape_device *device)
{
	struct tape_device *tmp;
	int minor;

	minor = 0;
	write_lock(&tape_device_lock);
	list_for_each_entry(tmp, &tape_device_list, node) {
		if (minor < tmp->first_minor)
			break;
		minor += TAPE_MINORS_PER_DEV;
	}
	if (minor >= 256) {
		write_unlock(&tape_device_lock);
		return -ENODEV;
	}
	device->first_minor = minor;
	list_add_tail(&device->node, &tmp->node);
	write_unlock(&tape_device_lock);
	return 0;
}

/* remove device from the list */
static void
tape_remove_minor(struct tape_device *device)
{
	write_lock(&tape_device_lock);
	list_del_init(&device->node);
	device->first_minor = -1;
	write_unlock(&tape_device_lock);
}

/*
 * Set a device online.
 *
 * This function is called by the common I/O layer to move a device from the
 * detected but offline into the online state.
 * If we return an error (RC < 0) the device remains in the offline state. This
 * can happen if the device is assigned somewhere else, for example.
 */
int
tape_generic_online(struct tape_device *device,
		   struct tape_discipline *discipline)
{
	int rc;

	DBF_LH(6, "tape_enable_device(%p, %p)\n", device, discipline);

	if (device->tape_state != TS_INIT) {
		DBF_LH(3, "Tapestate not INIT (%d)\n", device->tape_state);
		return -EINVAL;
	}

	init_timer(&device->lb_timeout);
	device->lb_timeout.function = tape_long_busy_timeout;

	/* Let the discipline have a go at the device. */
	device->discipline = discipline;
	if (!try_module_get(discipline->owner)) {
		return -EINVAL;
	}

	rc = discipline->setup_device(device);
	if (rc)
		goto out;
	rc = tape_assign_minor(device);
	if (rc)
		goto out_discipline;

	rc = tapechar_setup_device(device);
	if (rc)
		goto out_minor;
	rc = tapeblock_setup_device(device);
	if (rc)
		goto out_char;

	tape_state_set(device, TS_UNUSED);

	DBF_LH(3, "(%08x): Drive set online\n", device->cdev_id);

	return 0;

out_char:
	tapechar_cleanup_device(device);
out_minor:
	tape_remove_minor(device);
out_discipline:
	device->discipline->cleanup_device(device);
	device->discipline = NULL;
out:
	module_put(discipline->owner);
	return rc;
}

static void
tape_cleanup_device(struct tape_device *device)
{
	tapeblock_cleanup_device(device);
	tapechar_cleanup_device(device);
	device->discipline->cleanup_device(device);
	module_put(device->discipline->owner);
	tape_remove_minor(device);
	tape_med_state_set(device, MS_UNKNOWN);
}

/*
 * Suspend device.
 *
 * Called by the common I/O layer if the drive should be suspended on user
 * request. We refuse to suspend if the device is loaded or in use for the
 * following reason:
 * While the Linux guest is suspended, it might be logged off which causes
 * devices to be detached. Tape devices are automatically rewound and unloaded
 * during DETACH processing (unless the tape device was attached with the
 * NOASSIGN or MULTIUSER option). After rewind/unload, there is no way to
 * resume the original state of the tape device, since we would need to
 * manually re-load the cartridge which was active at suspend time.
 */
int tape_generic_pm_suspend(struct ccw_device *cdev)
{
	struct tape_device *device;

	device = dev_get_drvdata(&cdev->dev);
	if (!device) {
		return -ENODEV;
	}

	DBF_LH(3, "(%08x): tape_generic_pm_suspend(%p)\n",
		device->cdev_id, device);

	if (device->medium_state != MS_UNLOADED) {
		pr_err("A cartridge is loaded in tape device %s, "
		       "refusing to suspend\n", dev_name(&cdev->dev));
		return -EBUSY;
	}

	spin_lock_irq(get_ccwdev_lock(device->cdev));
	switch (device->tape_state) {
		case TS_INIT:
		case TS_NOT_OPER:
		case TS_UNUSED:
			spin_unlock_irq(get_ccwdev_lock(device->cdev));
			break;
		default:
			pr_err("Tape device %s is busy, refusing to "
			       "suspend\n", dev_name(&cdev->dev));
			spin_unlock_irq(get_ccwdev_lock(device->cdev));
			return -EBUSY;
	}

	DBF_LH(3, "(%08x): Drive suspended.\n", device->cdev_id);
	return 0;
}

/*
 * Set device offline.
 *
 * Called by the common I/O layer if the drive should set offline on user
 * request. We may prevent this by returning an error.
 * Manual offline is only allowed while the drive is not in use.
 */
int
tape_generic_offline(struct ccw_device *cdev)
{
	struct tape_device *device;

	device = dev_get_drvdata(&cdev->dev);
	if (!device) {
		return -ENODEV;
	}

	DBF_LH(3, "(%08x): tape_generic_offline(%p)\n",
		device->cdev_id, device);

	spin_lock_irq(get_ccwdev_lock(device->cdev));
	switch (device->tape_state) {
		case TS_INIT:
		case TS_NOT_OPER:
			spin_unlock_irq(get_ccwdev_lock(device->cdev));
			break;
		case TS_UNUSED:
			tape_state_set(device, TS_INIT);
			spin_unlock_irq(get_ccwdev_lock(device->cdev));
			tape_cleanup_device(device);
			break;
		default:
			DBF_EVENT(3, "(%08x): Set offline failed "
				"- drive in use.\n",
				device->cdev_id);
			spin_unlock_irq(get_ccwdev_lock(device->cdev));
			return -EBUSY;
	}

	DBF_LH(3, "(%08x): Drive set offline.\n", device->cdev_id);
	return 0;
}

/*
 * Allocate memory for a new device structure.
 */
static struct tape_device *
tape_alloc_device(void)
{
	struct tape_device *device;

	device = kzalloc(sizeof(struct tape_device), GFP_KERNEL);
	if (device == NULL) {
		DBF_EXCEPTION(2, "ti:no mem\n");
		return ERR_PTR(-ENOMEM);
	}
	device->modeset_byte = kmalloc(1, GFP_KERNEL | GFP_DMA);
	if (device->modeset_byte == NULL) {
		DBF_EXCEPTION(2, "ti:no mem\n");
		kfree(device);
		return ERR_PTR(-ENOMEM);
	}
	INIT_LIST_HEAD(&device->req_queue);
	INIT_LIST_HEAD(&device->node);
	init_waitqueue_head(&device->state_change_wq);
	init_waitqueue_head(&device->wait_queue);
	device->tape_state = TS_INIT;
	device->medium_state = MS_UNKNOWN;
	*device->modeset_byte = 0;
	device->first_minor = -1;
	atomic_set(&device->ref_count, 1);
	INIT_DELAYED_WORK(&device->tape_dnr, tape_delayed_next_request);

	return device;
}

/*
 * Get a reference to an existing device structure. This will automatically
 * increment the reference count.
 */
struct tape_device *
tape_get_device(struct tape_device *device)
{
	int count;

	count = atomic_inc_return(&device->ref_count);
	DBF_EVENT(4, "tape_get_device(%p) = %i\n", device, count);
	return device;
}

/*
 * Decrease the reference counter of a devices structure. If the
 * reference counter reaches zero free the device structure.
 * The function returns a NULL pointer to be used by the caller
 * for clearing reference pointers.
 */
void
tape_put_device(struct tape_device *device)
{
	int count;

	count = atomic_dec_return(&device->ref_count);
	DBF_EVENT(4, "tape_put_device(%p) -> %i\n", device, count);
	BUG_ON(count < 0);
	if (count == 0) {
		kfree(device->modeset_byte);
		kfree(device);
	}
}

/*
 * Find tape device by a device index.
 */
struct tape_device *
tape_find_device(int devindex)
{
	struct tape_device *device, *tmp;

	device = ERR_PTR(-ENODEV);
	read_lock(&tape_device_lock);
	list_for_each_entry(tmp, &tape_device_list, node) {
		if (tmp->first_minor / TAPE_MINORS_PER_DEV == devindex) {
			device = tape_get_device(tmp);
			break;
		}
	}
	read_unlock(&tape_device_lock);
	return device;
}

/*
 * Driverfs tape probe function.
 */
int
tape_generic_probe(struct ccw_device *cdev)
{
	struct tape_device *device;
	int ret;
	struct ccw_dev_id dev_id;

	device = tape_alloc_device();
	if (IS_ERR(device))
		return -ENODEV;
	ccw_device_set_options(cdev, CCWDEV_DO_PATHGROUP |
				     CCWDEV_DO_MULTIPATH);
	ret = sysfs_create_group(&cdev->dev.kobj, &tape_attr_group);
	if (ret) {
		tape_put_device(device);
		return ret;
	}
	dev_set_drvdata(&cdev->dev, device);
	cdev->handler = __tape_do_irq;
	device->cdev = cdev;
	ccw_device_get_id(cdev, &dev_id);
	device->cdev_id = devid_to_int(&dev_id);
	return ret;
}

static void
__tape_discard_requests(struct tape_device *device)
{
	struct tape_request *	request;
	struct list_head *	l, *n;

	list_for_each_safe(l, n, &device->req_queue) {
		request = list_entry(l, struct tape_request, list);
		if (request->status == TAPE_REQUEST_IN_IO)
			request->status = TAPE_REQUEST_DONE;
		list_del(&request->list);

		/* Decrease ref_count for removed request. */
		request->device = NULL;
		tape_put_device(device);
		request->rc = -EIO;
		if (request->callback != NULL)
			request->callback(request, request->callback_data);
	}
}

/*
 * Driverfs tape remove function.
 *
 * This function is called whenever the common I/O layer detects the device
 * gone. This can happen at any time and we cannot refuse.
 */
void
tape_generic_remove(struct ccw_device *cdev)
{
	struct tape_device *	device;

	device = dev_get_drvdata(&cdev->dev);
	if (!device) {
		return;
	}
	DBF_LH(3, "(%08x): tape_generic_remove(%p)\n", device->cdev_id, cdev);

	spin_lock_irq(get_ccwdev_lock(device->cdev));
	switch (device->tape_state) {
		case TS_INIT:
			tape_state_set(device, TS_NOT_OPER);
		case TS_NOT_OPER:
			/*
			 * Nothing to do.
			 */
			spin_unlock_irq(get_ccwdev_lock(device->cdev));
			break;
		case TS_UNUSED:
			/*
			 * Need only to release the device.
			 */
			tape_state_set(device, TS_NOT_OPER);
			spin_unlock_irq(get_ccwdev_lock(device->cdev));
			tape_cleanup_device(device);
			break;
		default:
			/*
			 * There may be requests on the queue. We will not get
			 * an interrupt for a request that was running. So we
			 * just post them all as I/O errors.
			 */
			DBF_EVENT(3, "(%08x): Drive in use vanished!\n",
				device->cdev_id);
			pr_warning("%s: A tape unit was detached while in "
				   "use\n", dev_name(&device->cdev->dev));
			tape_state_set(device, TS_NOT_OPER);
			__tape_discard_requests(device);
			spin_unlock_irq(get_ccwdev_lock(device->cdev));
			tape_cleanup_device(device);
	}

	device = dev_get_drvdata(&cdev->dev);
	if (device) {
		sysfs_remove_group(&cdev->dev.kobj, &tape_attr_group);
		dev_set_drvdata(&cdev->dev, NULL);
		tape_put_device(device);
	}
}

/*
 * Allocate a new tape ccw request
 */
struct tape_request *
tape_alloc_request(int cplength, int datasize)
{
	struct tape_request *request;

	BUG_ON(datasize > PAGE_SIZE || (cplength*sizeof(struct ccw1)) > PAGE_SIZE);

	DBF_LH(6, "tape_alloc_request(%d, %d)\n", cplength, datasize);

	request = kzalloc(sizeof(struct tape_request), GFP_KERNEL);
	if (request == NULL) {
		DBF_EXCEPTION(1, "cqra nomem\n");
		return ERR_PTR(-ENOMEM);
	}
	/* allocate channel program */
	if (cplength > 0) {
		request->cpaddr = kcalloc(cplength, sizeof(struct ccw1),
					  GFP_ATOMIC | GFP_DMA);
		if (request->cpaddr == NULL) {
			DBF_EXCEPTION(1, "cqra nomem\n");
			kfree(request);
			return ERR_PTR(-ENOMEM);
		}
	}
	/* alloc small kernel buffer */
	if (datasize > 0) {
		request->cpdata = kzalloc(datasize, GFP_KERNEL | GFP_DMA);
		if (request->cpdata == NULL) {
			DBF_EXCEPTION(1, "cqra nomem\n");
			kfree(request->cpaddr);
			kfree(request);
			return ERR_PTR(-ENOMEM);
		}
	}
	DBF_LH(6, "New request %p(%p/%p)\n", request, request->cpaddr,
		request->cpdata);

	return request;
}

/*
 * Free tape ccw request
 */
void
tape_free_request (struct tape_request * request)
{
	DBF_LH(6, "Free request %p\n", request);

	if (request->device)
		tape_put_device(request->device);
	kfree(request->cpdata);
	kfree(request->cpaddr);
	kfree(request);
}

static int
__tape_start_io(struct tape_device *device, struct tape_request *request)
{
	int rc;

#ifdef CONFIG_S390_TAPE_BLOCK
	if (request->op == TO_BLOCK)
		device->discipline->check_locate(device, request);
#endif
	rc = ccw_device_start(
		device->cdev,
		request->cpaddr,
		(unsigned long) request,
		0x00,
		request->options
	);
	if (rc == 0) {
		request->status = TAPE_REQUEST_IN_IO;
	} else if (rc == -EBUSY) {
		/* The common I/O subsystem is currently busy. Retry later. */
		request->status = TAPE_REQUEST_QUEUED;
		schedule_delayed_work(&device->tape_dnr, 0);
		rc = 0;
	} else {
		/* Start failed. Remove request and indicate failure. */
		DBF_EVENT(1, "tape: start request failed with RC = %i\n", rc);
	}
	return rc;
}

static void
__tape_start_next_request(struct tape_device *device)
{
	struct list_head *l, *n;
	struct tape_request *request;
	int rc;

	DBF_LH(6, "__tape_start_next_request(%p)\n", device);
	/*
	 * Try to start each request on request queue until one is
	 * started successful.
	 */
	list_for_each_safe(l, n, &device->req_queue) {
		request = list_entry(l, struct tape_request, list);

		/*
		 * Avoid race condition if bottom-half was triggered more than
		 * once.
		 */
		if (request->status == TAPE_REQUEST_IN_IO)
			return;
		/*
		 * Request has already been stopped. We have to wait until
		 * the request is removed from the queue in the interrupt
		 * handling.
		 */
		if (request->status == TAPE_REQUEST_DONE)
			return;

		/*
		 * We wanted to cancel the request but the common I/O layer
		 * was busy at that time. This can only happen if this
		 * function is called by delayed_next_request.
		 * Otherwise we start the next request on the queue.
		 */
		if (request->status == TAPE_REQUEST_CANCEL) {
			rc = __tape_cancel_io(device, request);
		} else {
			rc = __tape_start_io(device, request);
		}
		if (rc == 0)
			return;

		/* Set ending status. */
		request->rc = rc;
		request->status = TAPE_REQUEST_DONE;

		/* Remove from request queue. */
		list_del(&request->list);

		/* Do callback. */
		if (request->callback != NULL)
			request->callback(request, request->callback_data);
	}
}

static void
tape_delayed_next_request(struct work_struct *work)
{
	struct tape_device *device =
		container_of(work, struct tape_device, tape_dnr.work);

	DBF_LH(6, "tape_delayed_next_request(%p)\n", device);
	spin_lock_irq(get_ccwdev_lock(device->cdev));
	__tape_start_next_request(device);
	spin_unlock_irq(get_ccwdev_lock(device->cdev));
}

static void tape_long_busy_timeout(unsigned long data)
{
	struct tape_request *request;
	struct tape_device *device;

	device = (struct tape_device *) data;
	spin_lock_irq(get_ccwdev_lock(device->cdev));
	request = list_entry(device->req_queue.next, struct tape_request, list);
	BUG_ON(request->status != TAPE_REQUEST_LONG_BUSY);
	DBF_LH(6, "%08x: Long busy timeout.\n", device->cdev_id);
	__tape_start_next_request(device);
	device->lb_timeout.data = 0UL;
	tape_put_device(device);
	spin_unlock_irq(get_ccwdev_lock(device->cdev));
}

static void
__tape_end_request(
	struct tape_device *	device,
	struct tape_request *	request,
	int			rc)
{
	DBF_LH(6, "__tape_end_request(%p, %p, %i)\n", device, request, rc);
	if (request) {
		request->rc = rc;
		request->status = TAPE_REQUEST_DONE;

		/* Remove from request queue. */
		list_del(&request->list);

		/* Do callback. */
		if (request->callback != NULL)
			request->callback(request, request->callback_data);
	}

	/* Start next request. */
	if (!list_empty(&device->req_queue))
		__tape_start_next_request(device);
}

/*
 * Write sense data to dbf
 */
void
tape_dump_sense_dbf(struct tape_device *device, struct tape_request *request,
		    struct irb *irb)
{
	unsigned int *sptr;
	const char* op;

	if (request != NULL)
		op = tape_op_verbose[request->op];
	else
		op = "---";
	DBF_EVENT(3, "DSTAT : %02x   CSTAT: %02x\n",
		  irb->scsw.cmd.dstat, irb->scsw.cmd.cstat);
	DBF_EVENT(3, "DEVICE: %08x OP\t: %s\n", device->cdev_id, op);
	sptr = (unsigned int *) irb->ecw;
	DBF_EVENT(3, "%08x %08x\n", sptr[0], sptr[1]);
	DBF_EVENT(3, "%08x %08x\n", sptr[2], sptr[3]);
	DBF_EVENT(3, "%08x %08x\n", sptr[4], sptr[5]);
	DBF_EVENT(3, "%08x %08x\n", sptr[6], sptr[7]);
}

/*
 * I/O helper function. Adds the request to the request queue
 * and starts it if the tape is idle. Has to be called with
 * the device lock held.
 */
static int
__tape_start_request(struct tape_device *device, struct tape_request *request)
{
	int rc;

	switch (request->op) {
		case TO_MSEN:
		case TO_ASSIGN:
		case TO_UNASSIGN:
		case TO_READ_ATTMSG:
		case TO_RDC:
			if (device->tape_state == TS_INIT)
				break;
			if (device->tape_state == TS_UNUSED)
				break;
		default:
			if (device->tape_state == TS_BLKUSE)
				break;
			if (device->tape_state != TS_IN_USE)
				return -ENODEV;
	}

	/* Increase use count of device for the added request. */
	request->device = tape_get_device(device);

	if (list_empty(&device->req_queue)) {
		/* No other requests are on the queue. Start this one. */
		rc = __tape_start_io(device, request);
		if (rc)
			return rc;

		DBF_LH(5, "Request %p added for execution.\n", request);
		list_add(&request->list, &device->req_queue);
	} else {
		DBF_LH(5, "Request %p add to queue.\n", request);
		request->status = TAPE_REQUEST_QUEUED;
		list_add_tail(&request->list, &device->req_queue);
	}
	return 0;
}

/*
 * Add the request to the request queue, try to start it if the
 * tape is idle. Return without waiting for end of i/o.
 */
int
tape_do_io_async(struct tape_device *device, struct tape_request *request)
{
	int rc;

	DBF_LH(6, "tape_do_io_async(%p, %p)\n", device, request);

	spin_lock_irq(get_ccwdev_lock(device->cdev));
	/* Add request to request queue and try to start it. */
	rc = __tape_start_request(device, request);
	spin_unlock_irq(get_ccwdev_lock(device->cdev));
	return rc;
}

/*
 * tape_do_io/__tape_wake_up
 * Add the request to the request queue, try to start it if the
 * tape is idle and wait uninterruptible for its completion.
 */
static void
__tape_wake_up(struct tape_request *request, void *data)
{
	request->callback = NULL;
	wake_up((wait_queue_head_t *) data);
}

int
tape_do_io(struct tape_device *device, struct tape_request *request)
{
	int rc;

	spin_lock_irq(get_ccwdev_lock(device->cdev));
	/* Setup callback */
	request->callback = __tape_wake_up;
	request->callback_data = &device->wait_queue;
	/* Add request to request queue and try to start it. */
	rc = __tape_start_request(device, request);
	spin_unlock_irq(get_ccwdev_lock(device->cdev));
	if (rc)
		return rc;
	/* Request added to the queue. Wait for its completion. */
	wait_event(device->wait_queue, (request->callback == NULL));
	/* Get rc from request */
	return request->rc;
}

/*
 * tape_do_io_interruptible/__tape_wake_up_interruptible
 * Add the request to the request queue, try to start it if the
 * tape is idle and wait uninterruptible for its completion.
 */
static void
__tape_wake_up_interruptible(struct tape_request *request, void *data)
{
	request->callback = NULL;
	wake_up_interruptible((wait_queue_head_t *) data);
}

int
tape_do_io_interruptible(struct tape_device *device,
			 struct tape_request *request)
{
	int rc;

	spin_lock_irq(get_ccwdev_lock(device->cdev));
	/* Setup callback */
	request->callback = __tape_wake_up_interruptible;
	request->callback_data = &device->wait_queue;
	rc = __tape_start_request(device, request);
	spin_unlock_irq(get_ccwdev_lock(device->cdev));
	if (rc)
		return rc;
	/* Request added to the queue. Wait for its completion. */
	rc = wait_event_interruptible(device->wait_queue,
				      (request->callback == NULL));
	if (rc != -ERESTARTSYS)
		/* Request finished normally. */
		return request->rc;

	/* Interrupted by a signal. We have to stop the current request. */
	spin_lock_irq(get_ccwdev_lock(device->cdev));
	rc = __tape_cancel_io(device, request);
	spin_unlock_irq(get_ccwdev_lock(device->cdev));
	if (rc == 0) {
		/* Wait for the interrupt that acknowledges the halt. */
		do {
			rc = wait_event_interruptible(
				device->wait_queue,
				(request->callback == NULL)
			);
		} while (rc == -ERESTARTSYS);

		DBF_EVENT(3, "IO stopped on %08x\n", device->cdev_id);
		rc = -ERESTARTSYS;
	}
	return rc;
}

/*
 * Stop running ccw.
 */
int
tape_cancel_io(struct tape_device *device, struct tape_request *request)
{
	int rc;

	spin_lock_irq(get_ccwdev_lock(device->cdev));
	rc = __tape_cancel_io(device, request);
	spin_unlock_irq(get_ccwdev_lock(device->cdev));
	return rc;
}

/*
 * Tape interrupt routine, called from the ccw_device layer
 */
static void
__tape_do_irq (struct ccw_device *cdev, unsigned long intparm, struct irb *irb)
{
	struct tape_device *device;
	struct tape_request *request;
	int rc;

	device = dev_get_drvdata(&cdev->dev);
	if (device == NULL) {
		return;
	}
	request = (struct tape_request *) intparm;

	DBF_LH(6, "__tape_do_irq(device=%p, request=%p)\n", device, request);

	/* On special conditions irb is an error pointer */
	if (IS_ERR(irb)) {
		/* FIXME: What to do with the request? */
		switch (PTR_ERR(irb)) {
			case -ETIMEDOUT:
				DBF_LH(1, "(%s): Request timed out\n",
					dev_name(&cdev->dev));
			case -EIO:
				__tape_end_request(device, request, -EIO);
				break;
			default:
				DBF_LH(1, "(%s): Unexpected i/o error %li\n",
					dev_name(&cdev->dev),
					PTR_ERR(irb));
		}
		return;
	}

	/*
	 * If the condition code is not zero and the start function bit is
	 * still set, this is an deferred error and the last start I/O did
	 * not succeed. At this point the condition that caused the deferred
	 * error might still apply. So we just schedule the request to be
	 * started later.
	 */
	if (irb->scsw.cmd.cc != 0 &&
	    (irb->scsw.cmd.fctl & SCSW_FCTL_START_FUNC) &&
	    (request->status == TAPE_REQUEST_IN_IO)) {
		DBF_EVENT(3,"(%08x): deferred cc=%i, fctl=%i. restarting\n",
			device->cdev_id, irb->scsw.cmd.cc, irb->scsw.cmd.fctl);
		request->status = TAPE_REQUEST_QUEUED;
		schedule_delayed_work(&device->tape_dnr, HZ);
		return;
	}

	/* May be an unsolicited irq */
	if(request != NULL)
		request->rescnt = irb->scsw.cmd.count;
	else if ((irb->scsw.cmd.dstat == 0x85 || irb->scsw.cmd.dstat == 0x80) &&
		 !list_empty(&device->req_queue)) {
		/* Not Ready to Ready after long busy ? */
		struct tape_request *req;
		req = list_entry(device->req_queue.next,
				 struct tape_request, list);
		if (req->status == TAPE_REQUEST_LONG_BUSY) {
			DBF_EVENT(3, "(%08x): del timer\n", device->cdev_id);
			if (del_timer(&device->lb_timeout)) {
				device->lb_timeout.data = 0UL;
				tape_put_device(device);
				__tape_start_next_request(device);
			}
			return;
		}
	}
	if (irb->scsw.cmd.dstat != 0x0c) {
		/* Set the 'ONLINE' flag depending on sense byte 1 */
		if(*(((__u8 *) irb->ecw) + 1) & SENSE_DRIVE_ONLINE)
			device->tape_generic_status |= GMT_ONLINE(~0);
		else
			device->tape_generic_status &= ~GMT_ONLINE(~0);

		/*
		 * Any request that does not come back with channel end
		 * and device end is unusual. Log the sense data.
		 */
		DBF_EVENT(3,"-- Tape Interrupthandler --\n");
		tape_dump_sense_dbf(device, request, irb);
	} else {
		/* Upon normal completion the device _is_ online */
		device->tape_generic_status |= GMT_ONLINE(~0);
	}
	if (device->tape_state == TS_NOT_OPER) {
		DBF_EVENT(6, "tape:device is not operational\n");
		return;
	}

	/*
	 * Request that were canceled still come back with an interrupt.
	 * To detect these request the state will be set to TAPE_REQUEST_DONE.
	 */
	if(request != NULL && request->status == TAPE_REQUEST_DONE) {
		__tape_end_request(device, request, -EIO);
		return;
	}

	rc = device->discipline->irq(device, request, irb);
	/*
	 * rc < 0 : request finished unsuccessfully.
	 * rc == TAPE_IO_SUCCESS: request finished successfully.
	 * rc == TAPE_IO_PENDING: request is still running. Ignore rc.
	 * rc == TAPE_IO_RETRY: request finished but needs another go.
	 * rc == TAPE_IO_STOP: request needs to get terminated.
	 */
	switch (rc) {
		case TAPE_IO_SUCCESS:
			/* Upon normal completion the device _is_ online */
			device->tape_generic_status |= GMT_ONLINE(~0);
			__tape_end_request(device, request, rc);
			break;
		case TAPE_IO_PENDING:
			break;
		case TAPE_IO_LONG_BUSY:
			device->lb_timeout.data =
				(unsigned long) tape_get_device(device);
			device->lb_timeout.expires = jiffies +
				LONG_BUSY_TIMEOUT * HZ;
			DBF_EVENT(3, "(%08x): add timer\n", device->cdev_id);
			add_timer(&device->lb_timeout);
			request->status = TAPE_REQUEST_LONG_BUSY;
			break;
		case TAPE_IO_RETRY:
			rc = __tape_start_io(device, request);
			if (rc)
				__tape_end_request(device, request, rc);
			break;
		case TAPE_IO_STOP:
			rc = __tape_cancel_io(device, request);
			if (rc)
				__tape_end_request(device, request, rc);
			break;
		default:
			if (rc > 0) {
				DBF_EVENT(6, "xunknownrc\n");
				__tape_end_request(device, request, -EIO);
			} else {
				__tape_end_request(device, request, rc);
			}
			break;
	}
}

/*
 * Tape device open function used by tape_char & tape_block frontends.
 */
int
tape_open(struct tape_device *device)
{
	int rc;

	spin_lock_irq(get_ccwdev_lock(device->cdev));
	if (device->tape_state == TS_NOT_OPER) {
		DBF_EVENT(6, "TAPE:nodev\n");
		rc = -ENODEV;
	} else if (device->tape_state == TS_IN_USE) {
		DBF_EVENT(6, "TAPE:dbusy\n");
		rc = -EBUSY;
	} else if (device->tape_state == TS_BLKUSE) {
		DBF_EVENT(6, "TAPE:dbusy\n");
		rc = -EBUSY;
	} else if (device->discipline != NULL &&
		   !try_module_get(device->discipline->owner)) {
		DBF_EVENT(6, "TAPE:nodisc\n");
		rc = -ENODEV;
	} else {
		tape_state_set(device, TS_IN_USE);
		rc = 0;
	}
	spin_unlock_irq(get_ccwdev_lock(device->cdev));
	return rc;
}

/*
 * Tape device release function used by tape_char & tape_block frontends.
 */
int
tape_release(struct tape_device *device)
{
	spin_lock_irq(get_ccwdev_lock(device->cdev));
	if (device->tape_state == TS_IN_USE)
		tape_state_set(device, TS_UNUSED);
	module_put(device->discipline->owner);
	spin_unlock_irq(get_ccwdev_lock(device->cdev));
	return 0;
}

/*
 * Execute a magnetic tape command a number of times.
 */
int
tape_mtop(struct tape_device *device, int mt_op, int mt_count)
{
	tape_mtop_fn fn;
	int rc;

	DBF_EVENT(6, "TAPE:mtio\n");
	DBF_EVENT(6, "TAPE:ioop: %x\n", mt_op);
	DBF_EVENT(6, "TAPE:arg:	 %x\n", mt_count);

	if (mt_op < 0 || mt_op >= TAPE_NR_MTOPS)
		return -EINVAL;
	fn = device->discipline->mtop_array[mt_op];
	if (fn == NULL)
		return -EINVAL;

	/* We assume that the backends can handle count up to 500. */
	if (mt_op == MTBSR  || mt_op == MTFSR  || mt_op == MTFSF  ||
	    mt_op == MTBSF  || mt_op == MTFSFM || mt_op == MTBSFM) {
		rc = 0;
		for (; mt_count > 500; mt_count -= 500)
			if ((rc = fn(device, 500)) != 0)
				break;
		if (rc == 0)
			rc = fn(device, mt_count);
	} else
		rc = fn(device, mt_count);
	return rc;

}

/*
 * Tape init function.
 */
static int
tape_init (void)
{
	TAPE_DBF_AREA = debug_register ( "tape", 2, 2, 4*sizeof(long));
	debug_register_view(TAPE_DBF_AREA, &debug_sprintf_view);
#ifdef DBF_LIKE_HELL
	debug_set_level(TAPE_DBF_AREA, 6);
#endif
	DBF_EVENT(3, "tape init\n");
	tape_proc_init();
	tapechar_init ();
	tapeblock_init ();
	return 0;
}

/*
 * Tape exit function.
 */
static void
tape_exit(void)
{
	DBF_EVENT(6, "tape exit\n");

	/* Get rid of the frontends */
	tapechar_exit();
	tapeblock_exit();
	tape_proc_cleanup();
	debug_unregister (TAPE_DBF_AREA);
}

MODULE_AUTHOR("(C) 2001 IBM Deutschland Entwicklung GmbH by Carsten Otte and "
	      "Michael Holzheu (cotte@de.ibm.com,holzheu@de.ibm.com)");
MODULE_DESCRIPTION("Linux on zSeries channel attached tape device driver");
MODULE_LICENSE("GPL");

module_init(tape_init);
module_exit(tape_exit);

EXPORT_SYMBOL(tape_generic_remove);
EXPORT_SYMBOL(tape_generic_probe);
EXPORT_SYMBOL(tape_generic_online);
EXPORT_SYMBOL(tape_generic_offline);
EXPORT_SYMBOL(tape_generic_pm_suspend);
EXPORT_SYMBOL(tape_put_device);
EXPORT_SYMBOL(tape_get_device);
EXPORT_SYMBOL(tape_state_verbose);
EXPORT_SYMBOL(tape_op_verbose);
EXPORT_SYMBOL(tape_state_set);
EXPORT_SYMBOL(tape_med_state_set);
EXPORT_SYMBOL(tape_alloc_request);
EXPORT_SYMBOL(tape_free_request);
EXPORT_SYMBOL(tape_dump_sense_dbf);
EXPORT_SYMBOL(tape_do_io);
EXPORT_SYMBOL(tape_do_io_async);
EXPORT_SYMBOL(tape_do_io_interruptible);
EXPORT_SYMBOL(tape_cancel_io);
EXPORT_SYMBOL(tape_mtop);