1/*
2 * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com>
3 *		    Horst Hummel <Horst.Hummel@de.ibm.com>
4 *		    Carsten Otte <Cotte@de.ibm.com>
5 *		    Martin Schwidefsky <schwidefsky@de.ibm.com>
6 * Bugreports.to..: <Linux390@de.ibm.com>
7 * Copyright IBM Corp. 1999, 2009
8 */
9
10#define KMSG_COMPONENT "dasd"
11#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
12
13#include <linux/kmod.h>
14#include <linux/init.h>
15#include <linux/interrupt.h>
16#include <linux/ctype.h>
17#include <linux/major.h>
18#include <linux/slab.h>
19#include <linux/hdreg.h>
20#include <linux/async.h>
21#include <linux/mutex.h>
22#include <linux/debugfs.h>
23#include <linux/seq_file.h>
24#include <linux/vmalloc.h>
25
26#include <asm/ccwdev.h>
27#include <asm/ebcdic.h>
28#include <asm/idals.h>
29#include <asm/itcw.h>
30#include <asm/diag.h>
31
32/* This is ugly... */
33#define PRINTK_HEADER "dasd:"
34
35#include "dasd_int.h"
36/*
37 * SECTION: Constant definitions to be used within this file
38 */
39#define DASD_CHANQ_MAX_SIZE 4
40
41#define DASD_DIAG_MOD		"dasd_diag_mod"
42
43/*
44 * SECTION: exported variables of dasd.c
45 */
46debug_info_t *dasd_debug_area;
47EXPORT_SYMBOL(dasd_debug_area);
48static struct dentry *dasd_debugfs_root_entry;
49struct dasd_discipline *dasd_diag_discipline_pointer;
50EXPORT_SYMBOL(dasd_diag_discipline_pointer);
51void dasd_int_handler(struct ccw_device *, unsigned long, struct irb *);
52
53MODULE_AUTHOR("Holger Smolinski <Holger.Smolinski@de.ibm.com>");
54MODULE_DESCRIPTION("Linux on S/390 DASD device driver,"
55		   " Copyright IBM Corp. 2000");
56MODULE_SUPPORTED_DEVICE("dasd");
57MODULE_LICENSE("GPL");
58
59/*
60 * SECTION: prototypes for static functions of dasd.c
61 */
62static int  dasd_alloc_queue(struct dasd_block *);
63static void dasd_setup_queue(struct dasd_block *);
64static void dasd_free_queue(struct dasd_block *);
65static void dasd_flush_request_queue(struct dasd_block *);
66static int dasd_flush_block_queue(struct dasd_block *);
67static void dasd_device_tasklet(struct dasd_device *);
68static void dasd_block_tasklet(struct dasd_block *);
69static void do_kick_device(struct work_struct *);
70static void do_restore_device(struct work_struct *);
71static void do_reload_device(struct work_struct *);
72static void dasd_return_cqr_cb(struct dasd_ccw_req *, void *);
73static void dasd_device_timeout(unsigned long);
74static void dasd_block_timeout(unsigned long);
75static void __dasd_process_erp(struct dasd_device *, struct dasd_ccw_req *);
76static void dasd_profile_init(struct dasd_profile *, struct dentry *);
77static void dasd_profile_exit(struct dasd_profile *);
78
79/*
80 * SECTION: Operations on the device structure.
81 */
82static wait_queue_head_t dasd_init_waitq;
83static wait_queue_head_t dasd_flush_wq;
84static wait_queue_head_t generic_waitq;
85static wait_queue_head_t shutdown_waitq;
86
87/*
88 * Allocate memory for a new device structure.
89 */
90struct dasd_device *dasd_alloc_device(void)
91{
92	struct dasd_device *device;
93
94	device = kzalloc(sizeof(struct dasd_device), GFP_ATOMIC);
95	if (!device)
96		return ERR_PTR(-ENOMEM);
97
98	/* Get two pages for normal block device operations. */
99	device->ccw_mem = (void *) __get_free_pages(GFP_ATOMIC | GFP_DMA, 1);
100	if (!device->ccw_mem) {
101		kfree(device);
102		return ERR_PTR(-ENOMEM);
103	}
104	/* Get one page for error recovery. */
105	device->erp_mem = (void *) get_zeroed_page(GFP_ATOMIC | GFP_DMA);
106	if (!device->erp_mem) {
107		free_pages((unsigned long) device->ccw_mem, 1);
108		kfree(device);
109		return ERR_PTR(-ENOMEM);
110	}
111
112	dasd_init_chunklist(&device->ccw_chunks, device->ccw_mem, PAGE_SIZE*2);
113	dasd_init_chunklist(&device->erp_chunks, device->erp_mem, PAGE_SIZE);
114	spin_lock_init(&device->mem_lock);
115	atomic_set(&device->tasklet_scheduled, 0);
116	tasklet_init(&device->tasklet,
117		     (void (*)(unsigned long)) dasd_device_tasklet,
118		     (unsigned long) device);
119	INIT_LIST_HEAD(&device->ccw_queue);
120	init_timer(&device->timer);
121	device->timer.function = dasd_device_timeout;
122	device->timer.data = (unsigned long) device;
123	INIT_WORK(&device->kick_work, do_kick_device);
124	INIT_WORK(&device->restore_device, do_restore_device);
125	INIT_WORK(&device->reload_device, do_reload_device);
126	device->state = DASD_STATE_NEW;
127	device->target = DASD_STATE_NEW;
128	mutex_init(&device->state_mutex);
129	spin_lock_init(&device->profile.lock);
130	return device;
131}
132
133/*
134 * Free memory of a device structure.
135 */
136void dasd_free_device(struct dasd_device *device)
137{
138	kfree(device->private);
139	free_page((unsigned long) device->erp_mem);
140	free_pages((unsigned long) device->ccw_mem, 1);
141	kfree(device);
142}
143
144/*
145 * Allocate memory for a new device structure.
146 */
147struct dasd_block *dasd_alloc_block(void)
148{
149	struct dasd_block *block;
150
151	block = kzalloc(sizeof(*block), GFP_ATOMIC);
152	if (!block)
153		return ERR_PTR(-ENOMEM);
154	/* open_count = 0 means device online but not in use */
155	atomic_set(&block->open_count, -1);
156
157	spin_lock_init(&block->request_queue_lock);
158	atomic_set(&block->tasklet_scheduled, 0);
159	tasklet_init(&block->tasklet,
160		     (void (*)(unsigned long)) dasd_block_tasklet,
161		     (unsigned long) block);
162	INIT_LIST_HEAD(&block->ccw_queue);
163	spin_lock_init(&block->queue_lock);
164	init_timer(&block->timer);
165	block->timer.function = dasd_block_timeout;
166	block->timer.data = (unsigned long) block;
167	spin_lock_init(&block->profile.lock);
168
169	return block;
170}
171EXPORT_SYMBOL_GPL(dasd_alloc_block);
172
173/*
174 * Free memory of a device structure.
175 */
176void dasd_free_block(struct dasd_block *block)
177{
178	kfree(block);
179}
180EXPORT_SYMBOL_GPL(dasd_free_block);
181
182/*
183 * Make a new device known to the system.
184 */
185static int dasd_state_new_to_known(struct dasd_device *device)
186{
187	int rc;
188
189	/*
190	 * As long as the device is not in state DASD_STATE_NEW we want to
191	 * keep the reference count > 0.
192	 */
193	dasd_get_device(device);
194
195	if (device->block) {
196		rc = dasd_alloc_queue(device->block);
197		if (rc) {
198			dasd_put_device(device);
199			return rc;
200		}
201	}
202	device->state = DASD_STATE_KNOWN;
203	return 0;
204}
205
206/*
207 * Let the system forget about a device.
208 */
209static int dasd_state_known_to_new(struct dasd_device *device)
210{
211	/* Disable extended error reporting for this device. */
212	dasd_eer_disable(device);
213	/* Forget the discipline information. */
214	if (device->discipline) {
215		if (device->discipline->uncheck_device)
216			device->discipline->uncheck_device(device);
217		module_put(device->discipline->owner);
218	}
219	device->discipline = NULL;
220	if (device->base_discipline)
221		module_put(device->base_discipline->owner);
222	device->base_discipline = NULL;
223	device->state = DASD_STATE_NEW;
224
225	if (device->block)
226		dasd_free_queue(device->block);
227
228	/* Give up reference we took in dasd_state_new_to_known. */
229	dasd_put_device(device);
230	return 0;
231}
232
233static struct dentry *dasd_debugfs_setup(const char *name,
234					 struct dentry *base_dentry)
235{
236	struct dentry *pde;
237
238	if (!base_dentry)
239		return NULL;
240	pde = debugfs_create_dir(name, base_dentry);
241	if (!pde || IS_ERR(pde))
242		return NULL;
243	return pde;
244}
245
246/*
247 * Request the irq line for the device.
248 */
249static int dasd_state_known_to_basic(struct dasd_device *device)
250{
251	struct dasd_block *block = device->block;
252	int rc = 0;
253
254	/* Allocate and register gendisk structure. */
255	if (block) {
256		rc = dasd_gendisk_alloc(block);
257		if (rc)
258			return rc;
259		block->debugfs_dentry =
260			dasd_debugfs_setup(block->gdp->disk_name,
261					   dasd_debugfs_root_entry);
262		dasd_profile_init(&block->profile, block->debugfs_dentry);
263		if (dasd_global_profile_level == DASD_PROFILE_ON)
264			dasd_profile_on(&device->block->profile);
265	}
266	device->debugfs_dentry =
267		dasd_debugfs_setup(dev_name(&device->cdev->dev),
268				   dasd_debugfs_root_entry);
269	dasd_profile_init(&device->profile, device->debugfs_dentry);
270
271	/* register 'device' debug area, used for all DBF_DEV_XXX calls */
272	device->debug_area = debug_register(dev_name(&device->cdev->dev), 4, 1,
273					    8 * sizeof(long));
274	debug_register_view(device->debug_area, &debug_sprintf_view);
275	debug_set_level(device->debug_area, DBF_WARNING);
276	DBF_DEV_EVENT(DBF_EMERG, device, "%s", "debug area created");
277
278	device->state = DASD_STATE_BASIC;
279
280	return rc;
281}
282
283/*
284 * Release the irq line for the device. Terminate any running i/o.
285 */
286static int dasd_state_basic_to_known(struct dasd_device *device)
287{
288	int rc;
289
290	if (device->discipline->basic_to_known) {
291		rc = device->discipline->basic_to_known(device);
292		if (rc)
293			return rc;
294	}
295
296	if (device->block) {
297		dasd_profile_exit(&device->block->profile);
298		debugfs_remove(device->block->debugfs_dentry);
299		dasd_gendisk_free(device->block);
300		dasd_block_clear_timer(device->block);
301	}
302	rc = dasd_flush_device_queue(device);
303	if (rc)
304		return rc;
305	dasd_device_clear_timer(device);
306	dasd_profile_exit(&device->profile);
307	debugfs_remove(device->debugfs_dentry);
308	DBF_DEV_EVENT(DBF_EMERG, device, "%p debug area deleted", device);
309	if (device->debug_area != NULL) {
310		debug_unregister(device->debug_area);
311		device->debug_area = NULL;
312	}
313	device->state = DASD_STATE_KNOWN;
314	return 0;
315}
316
317/*
318 * Do the initial analysis. The do_analysis function may return
319 * -EAGAIN in which case the device keeps the state DASD_STATE_BASIC
320 * until the discipline decides to continue the startup sequence
321 * by calling the function dasd_change_state. The eckd disciplines
322 * uses this to start a ccw that detects the format. The completion
323 * interrupt for this detection ccw uses the kernel event daemon to
324 * trigger the call to dasd_change_state. All this is done in the
325 * discipline code, see dasd_eckd.c.
326 * After the analysis ccw is done (do_analysis returned 0) the block
327 * device is setup.
328 * In case the analysis returns an error, the device setup is stopped
329 * (a fake disk was already added to allow formatting).
330 */
331static int dasd_state_basic_to_ready(struct dasd_device *device)
332{
333	int rc;
334	struct dasd_block *block;
335
336	rc = 0;
337	block = device->block;
338	/* make disk known with correct capacity */
339	if (block) {
340		if (block->base->discipline->do_analysis != NULL)
341			rc = block->base->discipline->do_analysis(block);
342		if (rc) {
343			if (rc != -EAGAIN) {
344				device->state = DASD_STATE_UNFMT;
345				goto out;
346			}
347			return rc;
348		}
349		dasd_setup_queue(block);
350		set_capacity(block->gdp,
351			     block->blocks << block->s2b_shift);
352		device->state = DASD_STATE_READY;
353		rc = dasd_scan_partitions(block);
354		if (rc) {
355			device->state = DASD_STATE_BASIC;
356			return rc;
357		}
358	} else {
359		device->state = DASD_STATE_READY;
360	}
361out:
362	if (device->discipline->basic_to_ready)
363		rc = device->discipline->basic_to_ready(device);
364	return rc;
365}
366
367static inline
368int _wait_for_empty_queues(struct dasd_device *device)
369{
370	if (device->block)
371		return list_empty(&device->ccw_queue) &&
372			list_empty(&device->block->ccw_queue);
373	else
374		return list_empty(&device->ccw_queue);
375}
376
377/*
378 * Remove device from block device layer. Destroy dirty buffers.
379 * Forget format information. Check if the target level is basic
380 * and if it is create fake disk for formatting.
381 */
382static int dasd_state_ready_to_basic(struct dasd_device *device)
383{
384	int rc;
385
386	device->state = DASD_STATE_BASIC;
387	if (device->block) {
388		struct dasd_block *block = device->block;
389		rc = dasd_flush_block_queue(block);
390		if (rc) {
391			device->state = DASD_STATE_READY;
392			return rc;
393		}
394		dasd_flush_request_queue(block);
395		dasd_destroy_partitions(block);
396		block->blocks = 0;
397		block->bp_block = 0;
398		block->s2b_shift = 0;
399	}
400	return 0;
401}
402
403/*
404 * Back to basic.
405 */
406static int dasd_state_unfmt_to_basic(struct dasd_device *device)
407{
408	device->state = DASD_STATE_BASIC;
409	return 0;
410}
411
412/*
413 * Make the device online and schedule the bottom half to start
414 * the requeueing of requests from the linux request queue to the
415 * ccw queue.
416 */
417static int
418dasd_state_ready_to_online(struct dasd_device * device)
419{
420	struct gendisk *disk;
421	struct disk_part_iter piter;
422	struct hd_struct *part;
423
424	device->state = DASD_STATE_ONLINE;
425	if (device->block) {
426		dasd_schedule_block_bh(device->block);
427		if ((device->features & DASD_FEATURE_USERAW)) {
428			disk = device->block->gdp;
429			kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE);
430			return 0;
431		}
432		disk = device->block->bdev->bd_disk;
433		disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
434		while ((part = disk_part_iter_next(&piter)))
435			kobject_uevent(&part_to_dev(part)->kobj, KOBJ_CHANGE);
436		disk_part_iter_exit(&piter);
437	}
438	return 0;
439}
440
441/*
442 * Stop the requeueing of requests again.
443 */
444static int dasd_state_online_to_ready(struct dasd_device *device)
445{
446	int rc;
447	struct gendisk *disk;
448	struct disk_part_iter piter;
449	struct hd_struct *part;
450
451	if (device->discipline->online_to_ready) {
452		rc = device->discipline->online_to_ready(device);
453		if (rc)
454			return rc;
455	}
456
457	device->state = DASD_STATE_READY;
458	if (device->block && !(device->features & DASD_FEATURE_USERAW)) {
459		disk = device->block->bdev->bd_disk;
460		disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
461		while ((part = disk_part_iter_next(&piter)))
462			kobject_uevent(&part_to_dev(part)->kobj, KOBJ_CHANGE);
463		disk_part_iter_exit(&piter);
464	}
465	return 0;
466}
467
468/*
469 * Device startup state changes.
470 */
471static int dasd_increase_state(struct dasd_device *device)
472{
473	int rc;
474
475	rc = 0;
476	if (device->state == DASD_STATE_NEW &&
477	    device->target >= DASD_STATE_KNOWN)
478		rc = dasd_state_new_to_known(device);
479
480	if (!rc &&
481	    device->state == DASD_STATE_KNOWN &&
482	    device->target >= DASD_STATE_BASIC)
483		rc = dasd_state_known_to_basic(device);
484
485	if (!rc &&
486	    device->state == DASD_STATE_BASIC &&
487	    device->target >= DASD_STATE_READY)
488		rc = dasd_state_basic_to_ready(device);
489
490	if (!rc &&
491	    device->state == DASD_STATE_UNFMT &&
492	    device->target > DASD_STATE_UNFMT)
493		rc = -EPERM;
494
495	if (!rc &&
496	    device->state == DASD_STATE_READY &&
497	    device->target >= DASD_STATE_ONLINE)
498		rc = dasd_state_ready_to_online(device);
499
500	return rc;
501}
502
503/*
504 * Device shutdown state changes.
505 */
506static int dasd_decrease_state(struct dasd_device *device)
507{
508	int rc;
509
510	rc = 0;
511	if (device->state == DASD_STATE_ONLINE &&
512	    device->target <= DASD_STATE_READY)
513		rc = dasd_state_online_to_ready(device);
514
515	if (!rc &&
516	    device->state == DASD_STATE_READY &&
517	    device->target <= DASD_STATE_BASIC)
518		rc = dasd_state_ready_to_basic(device);
519
520	if (!rc &&
521	    device->state == DASD_STATE_UNFMT &&
522	    device->target <= DASD_STATE_BASIC)
523		rc = dasd_state_unfmt_to_basic(device);
524
525	if (!rc &&
526	    device->state == DASD_STATE_BASIC &&
527	    device->target <= DASD_STATE_KNOWN)
528		rc = dasd_state_basic_to_known(device);
529
530	if (!rc &&
531	    device->state == DASD_STATE_KNOWN &&
532	    device->target <= DASD_STATE_NEW)
533		rc = dasd_state_known_to_new(device);
534
535	return rc;
536}
537
538/*
539 * This is the main startup/shutdown routine.
540 */
541static void dasd_change_state(struct dasd_device *device)
542{
543	int rc;
544
545	if (device->state == device->target)
546		/* Already where we want to go today... */
547		return;
548	if (device->state < device->target)
549		rc = dasd_increase_state(device);
550	else
551		rc = dasd_decrease_state(device);
552	if (rc == -EAGAIN)
553		return;
554	if (rc)
555		device->target = device->state;
556
557	/* let user-space know that the device status changed */
558	kobject_uevent(&device->cdev->dev.kobj, KOBJ_CHANGE);
559
560	if (device->state == device->target)
561		wake_up(&dasd_init_waitq);
562}
563
564/*
565 * Kick starter for devices that did not complete the startup/shutdown
566 * procedure or were sleeping because of a pending state.
567 * dasd_kick_device will schedule a call do do_kick_device to the kernel
568 * event daemon.
569 */
570static void do_kick_device(struct work_struct *work)
571{
572	struct dasd_device *device = container_of(work, struct dasd_device, kick_work);
573	mutex_lock(&device->state_mutex);
574	dasd_change_state(device);
575	mutex_unlock(&device->state_mutex);
576	dasd_schedule_device_bh(device);
577	dasd_put_device(device);
578}
579
580void dasd_kick_device(struct dasd_device *device)
581{
582	dasd_get_device(device);
583	/* queue call to dasd_kick_device to the kernel event daemon. */
584	if (!schedule_work(&device->kick_work))
585		dasd_put_device(device);
586}
587EXPORT_SYMBOL(dasd_kick_device);
588
589/*
590 * dasd_reload_device will schedule a call do do_reload_device to the kernel
591 * event daemon.
592 */
593static void do_reload_device(struct work_struct *work)
594{
595	struct dasd_device *device = container_of(work, struct dasd_device,
596						  reload_device);
597	device->discipline->reload(device);
598	dasd_put_device(device);
599}
600
601void dasd_reload_device(struct dasd_device *device)
602{
603	dasd_get_device(device);
604	/* queue call to dasd_reload_device to the kernel event daemon. */
605	if (!schedule_work(&device->reload_device))
606		dasd_put_device(device);
607}
608EXPORT_SYMBOL(dasd_reload_device);
609
610/*
611 * dasd_restore_device will schedule a call do do_restore_device to the kernel
612 * event daemon.
613 */
614static void do_restore_device(struct work_struct *work)
615{
616	struct dasd_device *device = container_of(work, struct dasd_device,
617						  restore_device);
618	device->cdev->drv->restore(device->cdev);
619	dasd_put_device(device);
620}
621
622void dasd_restore_device(struct dasd_device *device)
623{
624	dasd_get_device(device);
625	/* queue call to dasd_restore_device to the kernel event daemon. */
626	if (!schedule_work(&device->restore_device))
627		dasd_put_device(device);
628}
629
630/*
631 * Set the target state for a device and starts the state change.
632 */
633void dasd_set_target_state(struct dasd_device *device, int target)
634{
635	dasd_get_device(device);
636	mutex_lock(&device->state_mutex);
637	/* If we are in probeonly mode stop at DASD_STATE_READY. */
638	if (dasd_probeonly && target > DASD_STATE_READY)
639		target = DASD_STATE_READY;
640	if (device->target != target) {
641		if (device->state == target)
642			wake_up(&dasd_init_waitq);
643		device->target = target;
644	}
645	if (device->state != device->target)
646		dasd_change_state(device);
647	mutex_unlock(&device->state_mutex);
648	dasd_put_device(device);
649}
650EXPORT_SYMBOL(dasd_set_target_state);
651
652/*
653 * Enable devices with device numbers in [from..to].
654 */
655static inline int _wait_for_device(struct dasd_device *device)
656{
657	return (device->state == device->target);
658}
659
660void dasd_enable_device(struct dasd_device *device)
661{
662	dasd_set_target_state(device, DASD_STATE_ONLINE);
663	if (device->state <= DASD_STATE_KNOWN)
664		/* No discipline for device found. */
665		dasd_set_target_state(device, DASD_STATE_NEW);
666	/* Now wait for the devices to come up. */
667	wait_event(dasd_init_waitq, _wait_for_device(device));
668
669	dasd_reload_device(device);
670	if (device->discipline->kick_validate)
671		device->discipline->kick_validate(device);
672}
673EXPORT_SYMBOL(dasd_enable_device);
674
675/*
676 * SECTION: device operation (interrupt handler, start i/o, term i/o ...)
677 */
678
679unsigned int dasd_global_profile_level = DASD_PROFILE_OFF;
680
681#ifdef CONFIG_DASD_PROFILE
682struct dasd_profile dasd_global_profile = {
683	.lock = __SPIN_LOCK_UNLOCKED(dasd_global_profile.lock),
684};
685static struct dentry *dasd_debugfs_global_entry;
686
687/*
688 * Add profiling information for cqr before execution.
689 */
690static void dasd_profile_start(struct dasd_block *block,
691			       struct dasd_ccw_req *cqr,
692			       struct request *req)
693{
694	struct list_head *l;
695	unsigned int counter;
696	struct dasd_device *device;
697
698	/* count the length of the chanq for statistics */
699	counter = 0;
700	if (dasd_global_profile_level || block->profile.data)
701		list_for_each(l, &block->ccw_queue)
702			if (++counter >= 31)
703				break;
704
705	spin_lock(&dasd_global_profile.lock);
706	if (dasd_global_profile.data) {
707		dasd_global_profile.data->dasd_io_nr_req[counter]++;
708		if (rq_data_dir(req) == READ)
709			dasd_global_profile.data->dasd_read_nr_req[counter]++;
710	}
711	spin_unlock(&dasd_global_profile.lock);
712
713	spin_lock(&block->profile.lock);
714	if (block->profile.data) {
715		block->profile.data->dasd_io_nr_req[counter]++;
716		if (rq_data_dir(req) == READ)
717			block->profile.data->dasd_read_nr_req[counter]++;
718	}
719	spin_unlock(&block->profile.lock);
720
721	/*
722	 * We count the request for the start device, even though it may run on
723	 * some other device due to error recovery. This way we make sure that
724	 * we count each request only once.
725	 */
726	device = cqr->startdev;
727	if (device->profile.data) {
728		counter = 1; /* request is not yet queued on the start device */
729		list_for_each(l, &device->ccw_queue)
730			if (++counter >= 31)
731				break;
732	}
733	spin_lock(&device->profile.lock);
734	if (device->profile.data) {
735		device->profile.data->dasd_io_nr_req[counter]++;
736		if (rq_data_dir(req) == READ)
737			device->profile.data->dasd_read_nr_req[counter]++;
738	}
739	spin_unlock(&device->profile.lock);
740}
741
742/*
743 * Add profiling information for cqr after execution.
744 */
745
746#define dasd_profile_counter(value, index)			   \
747{								   \
748	for (index = 0; index < 31 && value >> (2+index); index++) \
749		;						   \
750}
751
752static void dasd_profile_end_add_data(struct dasd_profile_info *data,
753				      int is_alias,
754				      int is_tpm,
755				      int is_read,
756				      long sectors,
757				      int sectors_ind,
758				      int tottime_ind,
759				      int tottimeps_ind,
760				      int strtime_ind,
761				      int irqtime_ind,
762				      int irqtimeps_ind,
763				      int endtime_ind)
764{
765	/* in case of an overflow, reset the whole profile */
766	if (data->dasd_io_reqs == UINT_MAX) {
767			memset(data, 0, sizeof(*data));
768			getnstimeofday(&data->starttod);
769	}
770	data->dasd_io_reqs++;
771	data->dasd_io_sects += sectors;
772	if (is_alias)
773		data->dasd_io_alias++;
774	if (is_tpm)
775		data->dasd_io_tpm++;
776
777	data->dasd_io_secs[sectors_ind]++;
778	data->dasd_io_times[tottime_ind]++;
779	data->dasd_io_timps[tottimeps_ind]++;
780	data->dasd_io_time1[strtime_ind]++;
781	data->dasd_io_time2[irqtime_ind]++;
782	data->dasd_io_time2ps[irqtimeps_ind]++;
783	data->dasd_io_time3[endtime_ind]++;
784
785	if (is_read) {
786		data->dasd_read_reqs++;
787		data->dasd_read_sects += sectors;
788		if (is_alias)
789			data->dasd_read_alias++;
790		if (is_tpm)
791			data->dasd_read_tpm++;
792		data->dasd_read_secs[sectors_ind]++;
793		data->dasd_read_times[tottime_ind]++;
794		data->dasd_read_time1[strtime_ind]++;
795		data->dasd_read_time2[irqtime_ind]++;
796		data->dasd_read_time3[endtime_ind]++;
797	}
798}
799
800static void dasd_profile_end(struct dasd_block *block,
801			     struct dasd_ccw_req *cqr,
802			     struct request *req)
803{
804	long strtime, irqtime, endtime, tottime;	/* in microseconds */
805	long tottimeps, sectors;
806	struct dasd_device *device;
807	int sectors_ind, tottime_ind, tottimeps_ind, strtime_ind;
808	int irqtime_ind, irqtimeps_ind, endtime_ind;
809
810	device = cqr->startdev;
811	if (!(dasd_global_profile_level ||
812	      block->profile.data ||
813	      device->profile.data))
814		return;
815
816	sectors = blk_rq_sectors(req);
817	if (!cqr->buildclk || !cqr->startclk ||
818	    !cqr->stopclk || !cqr->endclk ||
819	    !sectors)
820		return;
821
822	strtime = ((cqr->startclk - cqr->buildclk) >> 12);
823	irqtime = ((cqr->stopclk - cqr->startclk) >> 12);
824	endtime = ((cqr->endclk - cqr->stopclk) >> 12);
825	tottime = ((cqr->endclk - cqr->buildclk) >> 12);
826	tottimeps = tottime / sectors;
827
828	dasd_profile_counter(sectors, sectors_ind);
829	dasd_profile_counter(tottime, tottime_ind);
830	dasd_profile_counter(tottimeps, tottimeps_ind);
831	dasd_profile_counter(strtime, strtime_ind);
832	dasd_profile_counter(irqtime, irqtime_ind);
833	dasd_profile_counter(irqtime / sectors, irqtimeps_ind);
834	dasd_profile_counter(endtime, endtime_ind);
835
836	spin_lock(&dasd_global_profile.lock);
837	if (dasd_global_profile.data) {
838		dasd_profile_end_add_data(dasd_global_profile.data,
839					  cqr->startdev != block->base,
840					  cqr->cpmode == 1,
841					  rq_data_dir(req) == READ,
842					  sectors, sectors_ind, tottime_ind,
843					  tottimeps_ind, strtime_ind,
844					  irqtime_ind, irqtimeps_ind,
845					  endtime_ind);
846	}
847	spin_unlock(&dasd_global_profile.lock);
848
849	spin_lock(&block->profile.lock);
850	if (block->profile.data)
851		dasd_profile_end_add_data(block->profile.data,
852					  cqr->startdev != block->base,
853					  cqr->cpmode == 1,
854					  rq_data_dir(req) == READ,
855					  sectors, sectors_ind, tottime_ind,
856					  tottimeps_ind, strtime_ind,
857					  irqtime_ind, irqtimeps_ind,
858					  endtime_ind);
859	spin_unlock(&block->profile.lock);
860
861	spin_lock(&device->profile.lock);
862	if (device->profile.data)
863		dasd_profile_end_add_data(device->profile.data,
864					  cqr->startdev != block->base,
865					  cqr->cpmode == 1,
866					  rq_data_dir(req) == READ,
867					  sectors, sectors_ind, tottime_ind,
868					  tottimeps_ind, strtime_ind,
869					  irqtime_ind, irqtimeps_ind,
870					  endtime_ind);
871	spin_unlock(&device->profile.lock);
872}
873
874void dasd_profile_reset(struct dasd_profile *profile)
875{
876	struct dasd_profile_info *data;
877
878	spin_lock_bh(&profile->lock);
879	data = profile->data;
880	if (!data) {
881		spin_unlock_bh(&profile->lock);
882		return;
883	}
884	memset(data, 0, sizeof(*data));
885	getnstimeofday(&data->starttod);
886	spin_unlock_bh(&profile->lock);
887}
888
889int dasd_profile_on(struct dasd_profile *profile)
890{
891	struct dasd_profile_info *data;
892
893	data = kzalloc(sizeof(*data), GFP_KERNEL);
894	if (!data)
895		return -ENOMEM;
896	spin_lock_bh(&profile->lock);
897	if (profile->data) {
898		spin_unlock_bh(&profile->lock);
899		kfree(data);
900		return 0;
901	}
902	getnstimeofday(&data->starttod);
903	profile->data = data;
904	spin_unlock_bh(&profile->lock);
905	return 0;
906}
907
908void dasd_profile_off(struct dasd_profile *profile)
909{
910	spin_lock_bh(&profile->lock);
911	kfree(profile->data);
912	profile->data = NULL;
913	spin_unlock_bh(&profile->lock);
914}
915
916char *dasd_get_user_string(const char __user *user_buf, size_t user_len)
917{
918	char *buffer;
919
920	buffer = vmalloc(user_len + 1);
921	if (buffer == NULL)
922		return ERR_PTR(-ENOMEM);
923	if (copy_from_user(buffer, user_buf, user_len) != 0) {
924		vfree(buffer);
925		return ERR_PTR(-EFAULT);
926	}
927	/* got the string, now strip linefeed. */
928	if (buffer[user_len - 1] == '\n')
929		buffer[user_len - 1] = 0;
930	else
931		buffer[user_len] = 0;
932	return buffer;
933}
934
935static ssize_t dasd_stats_write(struct file *file,
936				const char __user *user_buf,
937				size_t user_len, loff_t *pos)
938{
939	char *buffer, *str;
940	int rc;
941	struct seq_file *m = (struct seq_file *)file->private_data;
942	struct dasd_profile *prof = m->private;
943
944	if (user_len > 65536)
945		user_len = 65536;
946	buffer = dasd_get_user_string(user_buf, user_len);
947	if (IS_ERR(buffer))
948		return PTR_ERR(buffer);
949
950	str = skip_spaces(buffer);
951	rc = user_len;
952	if (strncmp(str, "reset", 5) == 0) {
953		dasd_profile_reset(prof);
954	} else if (strncmp(str, "on", 2) == 0) {
955		rc = dasd_profile_on(prof);
956		if (rc)
957			goto out;
958		rc = user_len;
959		if (prof == &dasd_global_profile) {
960			dasd_profile_reset(prof);
961			dasd_global_profile_level = DASD_PROFILE_GLOBAL_ONLY;
962		}
963	} else if (strncmp(str, "off", 3) == 0) {
964		if (prof == &dasd_global_profile)
965			dasd_global_profile_level = DASD_PROFILE_OFF;
966		dasd_profile_off(prof);
967	} else
968		rc = -EINVAL;
969out:
970	vfree(buffer);
971	return rc;
972}
973
974static void dasd_stats_array(struct seq_file *m, unsigned int *array)
975{
976	int i;
977
978	for (i = 0; i < 32; i++)
979		seq_printf(m, "%u ", array[i]);
980	seq_putc(m, '\n');
981}
982
983static void dasd_stats_seq_print(struct seq_file *m,
984				 struct dasd_profile_info *data)
985{
986	seq_printf(m, "start_time %ld.%09ld\n",
987		   data->starttod.tv_sec, data->starttod.tv_nsec);
988	seq_printf(m, "total_requests %u\n", data->dasd_io_reqs);
989	seq_printf(m, "total_sectors %u\n", data->dasd_io_sects);
990	seq_printf(m, "total_pav %u\n", data->dasd_io_alias);
991	seq_printf(m, "total_hpf %u\n", data->dasd_io_tpm);
992	seq_puts(m, "histogram_sectors ");
993	dasd_stats_array(m, data->dasd_io_secs);
994	seq_puts(m, "histogram_io_times ");
995	dasd_stats_array(m, data->dasd_io_times);
996	seq_puts(m, "histogram_io_times_weighted ");
997	dasd_stats_array(m, data->dasd_io_timps);
998	seq_puts(m, "histogram_time_build_to_ssch ");
999	dasd_stats_array(m, data->dasd_io_time1);
1000	seq_puts(m, "histogram_time_ssch_to_irq ");
1001	dasd_stats_array(m, data->dasd_io_time2);
1002	seq_puts(m, "histogram_time_ssch_to_irq_weighted ");
1003	dasd_stats_array(m, data->dasd_io_time2ps);
1004	seq_puts(m, "histogram_time_irq_to_end ");
1005	dasd_stats_array(m, data->dasd_io_time3);
1006	seq_puts(m, "histogram_ccw_queue_length ");
1007	dasd_stats_array(m, data->dasd_io_nr_req);
1008	seq_printf(m, "total_read_requests %u\n", data->dasd_read_reqs);
1009	seq_printf(m, "total_read_sectors %u\n", data->dasd_read_sects);
1010	seq_printf(m, "total_read_pav %u\n", data->dasd_read_alias);
1011	seq_printf(m, "total_read_hpf %u\n", data->dasd_read_tpm);
1012	seq_puts(m, "histogram_read_sectors ");
1013	dasd_stats_array(m, data->dasd_read_secs);
1014	seq_puts(m, "histogram_read_times ");
1015	dasd_stats_array(m, data->dasd_read_times);
1016	seq_puts(m, "histogram_read_time_build_to_ssch ");
1017	dasd_stats_array(m, data->dasd_read_time1);
1018	seq_puts(m, "histogram_read_time_ssch_to_irq ");
1019	dasd_stats_array(m, data->dasd_read_time2);
1020	seq_puts(m, "histogram_read_time_irq_to_end ");
1021	dasd_stats_array(m, data->dasd_read_time3);
1022	seq_puts(m, "histogram_read_ccw_queue_length ");
1023	dasd_stats_array(m, data->dasd_read_nr_req);
1024}
1025
1026static int dasd_stats_show(struct seq_file *m, void *v)
1027{
1028	struct dasd_profile *profile;
1029	struct dasd_profile_info *data;
1030
1031	profile = m->private;
1032	spin_lock_bh(&profile->lock);
1033	data = profile->data;
1034	if (!data) {
1035		spin_unlock_bh(&profile->lock);
1036		seq_puts(m, "disabled\n");
1037		return 0;
1038	}
1039	dasd_stats_seq_print(m, data);
1040	spin_unlock_bh(&profile->lock);
1041	return 0;
1042}
1043
1044static int dasd_stats_open(struct inode *inode, struct file *file)
1045{
1046	struct dasd_profile *profile = inode->i_private;
1047	return single_open(file, dasd_stats_show, profile);
1048}
1049
1050static const struct file_operations dasd_stats_raw_fops = {
1051	.owner		= THIS_MODULE,
1052	.open		= dasd_stats_open,
1053	.read		= seq_read,
1054	.llseek		= seq_lseek,
1055	.release	= single_release,
1056	.write		= dasd_stats_write,
1057};
1058
1059static void dasd_profile_init(struct dasd_profile *profile,
1060			      struct dentry *base_dentry)
1061{
1062	umode_t mode;
1063	struct dentry *pde;
1064
1065	if (!base_dentry)
1066		return;
1067	profile->dentry = NULL;
1068	profile->data = NULL;
1069	mode = (S_IRUSR | S_IWUSR | S_IFREG);
1070	pde = debugfs_create_file("statistics", mode, base_dentry,
1071				  profile, &dasd_stats_raw_fops);
1072	if (pde && !IS_ERR(pde))
1073		profile->dentry = pde;
1074	return;
1075}
1076
1077static void dasd_profile_exit(struct dasd_profile *profile)
1078{
1079	dasd_profile_off(profile);
1080	debugfs_remove(profile->dentry);
1081	profile->dentry = NULL;
1082}
1083
1084static void dasd_statistics_removeroot(void)
1085{
1086	dasd_global_profile_level = DASD_PROFILE_OFF;
1087	dasd_profile_exit(&dasd_global_profile);
1088	debugfs_remove(dasd_debugfs_global_entry);
1089	debugfs_remove(dasd_debugfs_root_entry);
1090}
1091
1092static void dasd_statistics_createroot(void)
1093{
1094	struct dentry *pde;
1095
1096	dasd_debugfs_root_entry = NULL;
1097	pde = debugfs_create_dir("dasd", NULL);
1098	if (!pde || IS_ERR(pde))
1099		goto error;
1100	dasd_debugfs_root_entry = pde;
1101	pde = debugfs_create_dir("global", dasd_debugfs_root_entry);
1102	if (!pde || IS_ERR(pde))
1103		goto error;
1104	dasd_debugfs_global_entry = pde;
1105	dasd_profile_init(&dasd_global_profile, dasd_debugfs_global_entry);
1106	return;
1107
1108error:
1109	DBF_EVENT(DBF_ERR, "%s",
1110		  "Creation of the dasd debugfs interface failed");
1111	dasd_statistics_removeroot();
1112	return;
1113}
1114
1115#else
1116#define dasd_profile_start(block, cqr, req) do {} while (0)
1117#define dasd_profile_end(block, cqr, req) do {} while (0)
1118
1119static void dasd_statistics_createroot(void)
1120{
1121	return;
1122}
1123
1124static void dasd_statistics_removeroot(void)
1125{
1126	return;
1127}
1128
1129int dasd_stats_generic_show(struct seq_file *m, void *v)
1130{
1131	seq_puts(m, "Statistics are not activated in this kernel\n");
1132	return 0;
1133}
1134
1135static void dasd_profile_init(struct dasd_profile *profile,
1136			      struct dentry *base_dentry)
1137{
1138	return;
1139}
1140
1141static void dasd_profile_exit(struct dasd_profile *profile)
1142{
1143	return;
1144}
1145
1146int dasd_profile_on(struct dasd_profile *profile)
1147{
1148	return 0;
1149}
1150
1151#endif				/* CONFIG_DASD_PROFILE */
1152
1153/*
1154 * Allocate memory for a channel program with 'cplength' channel
1155 * command words and 'datasize' additional space. There are two
1156 * variantes: 1) dasd_kmalloc_request uses kmalloc to get the needed
1157 * memory and 2) dasd_smalloc_request uses the static ccw memory
1158 * that gets allocated for each device.
1159 */
1160struct dasd_ccw_req *dasd_kmalloc_request(int magic, int cplength,
1161					  int datasize,
1162					  struct dasd_device *device)
1163{
1164	struct dasd_ccw_req *cqr;
1165
1166	/* Sanity checks */
1167	BUG_ON(datasize > PAGE_SIZE ||
1168	     (cplength*sizeof(struct ccw1)) > PAGE_SIZE);
1169
1170	cqr = kzalloc(sizeof(struct dasd_ccw_req), GFP_ATOMIC);
1171	if (cqr == NULL)
1172		return ERR_PTR(-ENOMEM);
1173	cqr->cpaddr = NULL;
1174	if (cplength > 0) {
1175		cqr->cpaddr = kcalloc(cplength, sizeof(struct ccw1),
1176				      GFP_ATOMIC | GFP_DMA);
1177		if (cqr->cpaddr == NULL) {
1178			kfree(cqr);
1179			return ERR_PTR(-ENOMEM);
1180		}
1181	}
1182	cqr->data = NULL;
1183	if (datasize > 0) {
1184		cqr->data = kzalloc(datasize, GFP_ATOMIC | GFP_DMA);
1185		if (cqr->data == NULL) {
1186			kfree(cqr->cpaddr);
1187			kfree(cqr);
1188			return ERR_PTR(-ENOMEM);
1189		}
1190	}
1191	cqr->magic =  magic;
1192	set_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags);
1193	dasd_get_device(device);
1194	return cqr;
1195}
1196EXPORT_SYMBOL(dasd_kmalloc_request);
1197
1198struct dasd_ccw_req *dasd_smalloc_request(int magic, int cplength,
1199					  int datasize,
1200					  struct dasd_device *device)
1201{
1202	unsigned long flags;
1203	struct dasd_ccw_req *cqr;
1204	char *data;
1205	int size;
1206
1207	size = (sizeof(struct dasd_ccw_req) + 7L) & -8L;
1208	if (cplength > 0)
1209		size += cplength * sizeof(struct ccw1);
1210	if (datasize > 0)
1211		size += datasize;
1212	spin_lock_irqsave(&device->mem_lock, flags);
1213	cqr = (struct dasd_ccw_req *)
1214		dasd_alloc_chunk(&device->ccw_chunks, size);
1215	spin_unlock_irqrestore(&device->mem_lock, flags);
1216	if (cqr == NULL)
1217		return ERR_PTR(-ENOMEM);
1218	memset(cqr, 0, sizeof(struct dasd_ccw_req));
1219	data = (char *) cqr + ((sizeof(struct dasd_ccw_req) + 7L) & -8L);
1220	cqr->cpaddr = NULL;
1221	if (cplength > 0) {
1222		cqr->cpaddr = (struct ccw1 *) data;
1223		data += cplength*sizeof(struct ccw1);
1224		memset(cqr->cpaddr, 0, cplength*sizeof(struct ccw1));
1225	}
1226	cqr->data = NULL;
1227	if (datasize > 0) {
1228		cqr->data = data;
1229 		memset(cqr->data, 0, datasize);
1230	}
1231	cqr->magic = magic;
1232	set_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags);
1233	dasd_get_device(device);
1234	return cqr;
1235}
1236EXPORT_SYMBOL(dasd_smalloc_request);
1237
1238/*
1239 * Free memory of a channel program. This function needs to free all the
1240 * idal lists that might have been created by dasd_set_cda and the
1241 * struct dasd_ccw_req itself.
1242 */
1243void dasd_kfree_request(struct dasd_ccw_req *cqr, struct dasd_device *device)
1244{
1245	struct ccw1 *ccw;
1246
1247	/* Clear any idals used for the request. */
1248	ccw = cqr->cpaddr;
1249	do {
1250		clear_normalized_cda(ccw);
1251	} while (ccw++->flags & (CCW_FLAG_CC | CCW_FLAG_DC));
1252	kfree(cqr->cpaddr);
1253	kfree(cqr->data);
1254	kfree(cqr);
1255	dasd_put_device(device);
1256}
1257EXPORT_SYMBOL(dasd_kfree_request);
1258
1259void dasd_sfree_request(struct dasd_ccw_req *cqr, struct dasd_device *device)
1260{
1261	unsigned long flags;
1262
1263	spin_lock_irqsave(&device->mem_lock, flags);
1264	dasd_free_chunk(&device->ccw_chunks, cqr);
1265	spin_unlock_irqrestore(&device->mem_lock, flags);
1266	dasd_put_device(device);
1267}
1268EXPORT_SYMBOL(dasd_sfree_request);
1269
1270/*
1271 * Check discipline magic in cqr.
1272 */
1273static inline int dasd_check_cqr(struct dasd_ccw_req *cqr)
1274{
1275	struct dasd_device *device;
1276
1277	if (cqr == NULL)
1278		return -EINVAL;
1279	device = cqr->startdev;
1280	if (strncmp((char *) &cqr->magic, device->discipline->ebcname, 4)) {
1281		DBF_DEV_EVENT(DBF_WARNING, device,
1282			    " dasd_ccw_req 0x%08x magic doesn't match"
1283			    " discipline 0x%08x",
1284			    cqr->magic,
1285			    *(unsigned int *) device->discipline->name);
1286		return -EINVAL;
1287	}
1288	return 0;
1289}
1290
1291/*
1292 * Terminate the current i/o and set the request to clear_pending.
1293 * Timer keeps device runnig.
1294 * ccw_device_clear can fail if the i/o subsystem
1295 * is in a bad mood.
1296 */
1297int dasd_term_IO(struct dasd_ccw_req *cqr)
1298{
1299	struct dasd_device *device;
1300	int retries, rc;
1301	char errorstring[ERRORLENGTH];
1302
1303	/* Check the cqr */
1304	rc = dasd_check_cqr(cqr);
1305	if (rc)
1306		return rc;
1307	retries = 0;
1308	device = (struct dasd_device *) cqr->startdev;
1309	while ((retries < 5) && (cqr->status == DASD_CQR_IN_IO)) {
1310		rc = ccw_device_clear(device->cdev, (long) cqr);
1311		switch (rc) {
1312		case 0:	/* termination successful */
1313			cqr->status = DASD_CQR_CLEAR_PENDING;
1314			cqr->stopclk = get_tod_clock();
1315			cqr->starttime = 0;
1316			DBF_DEV_EVENT(DBF_DEBUG, device,
1317				      "terminate cqr %p successful",
1318				      cqr);
1319			break;
1320		case -ENODEV:
1321			DBF_DEV_EVENT(DBF_ERR, device, "%s",
1322				      "device gone, retry");
1323			break;
1324		case -EIO:
1325			DBF_DEV_EVENT(DBF_ERR, device, "%s",
1326				      "I/O error, retry");
1327			break;
1328		case -EINVAL:
1329			/*
1330			 * device not valid so no I/O could be running
1331			 * handle CQR as termination successful
1332			 */
1333			cqr->status = DASD_CQR_CLEARED;
1334			cqr->stopclk = get_tod_clock();
1335			cqr->starttime = 0;
1336			/* no retries for invalid devices */
1337			cqr->retries = -1;
1338			DBF_DEV_EVENT(DBF_ERR, device, "%s",
1339				      "EINVAL, handle as terminated");
1340			/* fake rc to success */
1341			rc = 0;
1342			break;
1343		case -EBUSY:
1344			DBF_DEV_EVENT(DBF_ERR, device, "%s",
1345				      "device busy, retry later");
1346			break;
1347		default:
1348			/* internal error 10 - unknown rc*/
1349			snprintf(errorstring, ERRORLENGTH, "10 %d", rc);
1350			dev_err(&device->cdev->dev, "An error occurred in the "
1351				"DASD device driver, reason=%s\n", errorstring);
1352			BUG();
1353			break;
1354		}
1355		retries++;
1356	}
1357	dasd_schedule_device_bh(device);
1358	return rc;
1359}
1360EXPORT_SYMBOL(dasd_term_IO);
1361
1362/*
1363 * Start the i/o. This start_IO can fail if the channel is really busy.
1364 * In that case set up a timer to start the request later.
1365 */
1366int dasd_start_IO(struct dasd_ccw_req *cqr)
1367{
1368	struct dasd_device *device;
1369	int rc;
1370	char errorstring[ERRORLENGTH];
1371
1372	/* Check the cqr */
1373	rc = dasd_check_cqr(cqr);
1374	if (rc) {
1375		cqr->intrc = rc;
1376		return rc;
1377	}
1378	device = (struct dasd_device *) cqr->startdev;
1379	if (((cqr->block &&
1380	      test_bit(DASD_FLAG_LOCK_STOLEN, &cqr->block->base->flags)) ||
1381	     test_bit(DASD_FLAG_LOCK_STOLEN, &device->flags)) &&
1382	    !test_bit(DASD_CQR_ALLOW_SLOCK, &cqr->flags)) {
1383		DBF_DEV_EVENT(DBF_DEBUG, device, "start_IO: return request %p "
1384			      "because of stolen lock", cqr);
1385		cqr->status = DASD_CQR_ERROR;
1386		cqr->intrc = -EPERM;
1387		return -EPERM;
1388	}
1389	if (cqr->retries < 0) {
1390		/* internal error 14 - start_IO run out of retries */
1391		sprintf(errorstring, "14 %p", cqr);
1392		dev_err(&device->cdev->dev, "An error occurred in the DASD "
1393			"device driver, reason=%s\n", errorstring);
1394		cqr->status = DASD_CQR_ERROR;
1395		return -EIO;
1396	}
1397	cqr->startclk = get_tod_clock();
1398	cqr->starttime = jiffies;
1399	cqr->retries--;
1400	if (!test_bit(DASD_CQR_VERIFY_PATH, &cqr->flags)) {
1401		cqr->lpm &= device->path_data.opm;
1402		if (!cqr->lpm)
1403			cqr->lpm = device->path_data.opm;
1404	}
1405	if (cqr->cpmode == 1) {
1406		rc = ccw_device_tm_start(device->cdev, cqr->cpaddr,
1407					 (long) cqr, cqr->lpm);
1408	} else {
1409		rc = ccw_device_start(device->cdev, cqr->cpaddr,
1410				      (long) cqr, cqr->lpm, 0);
1411	}
1412	switch (rc) {
1413	case 0:
1414		cqr->status = DASD_CQR_IN_IO;
1415		break;
1416	case -EBUSY:
1417		DBF_DEV_EVENT(DBF_WARNING, device, "%s",
1418			      "start_IO: device busy, retry later");
1419		break;
1420	case -ETIMEDOUT:
1421		DBF_DEV_EVENT(DBF_WARNING, device, "%s",
1422			      "start_IO: request timeout, retry later");
1423		break;
1424	case -EACCES:
1425		/* -EACCES indicates that the request used only a subset of the
1426		 * available paths and all these paths are gone. If the lpm of
1427		 * this request was only a subset of the opm (e.g. the ppm) then
1428		 * we just do a retry with all available paths.
1429		 * If we already use the full opm, something is amiss, and we
1430		 * need a full path verification.
1431		 */
1432		if (test_bit(DASD_CQR_VERIFY_PATH, &cqr->flags)) {
1433			DBF_DEV_EVENT(DBF_WARNING, device,
1434				      "start_IO: selected paths gone (%x)",
1435				      cqr->lpm);
1436		} else if (cqr->lpm != device->path_data.opm) {
1437			cqr->lpm = device->path_data.opm;
1438			DBF_DEV_EVENT(DBF_DEBUG, device, "%s",
1439				      "start_IO: selected paths gone,"
1440				      " retry on all paths");
1441		} else {
1442			DBF_DEV_EVENT(DBF_WARNING, device, "%s",
1443				      "start_IO: all paths in opm gone,"
1444				      " do path verification");
1445			dasd_generic_last_path_gone(device);
1446			device->path_data.opm = 0;
1447			device->path_data.ppm = 0;
1448			device->path_data.npm = 0;
1449			device->path_data.tbvpm =
1450				ccw_device_get_path_mask(device->cdev);
1451		}
1452		break;
1453	case -ENODEV:
1454		DBF_DEV_EVENT(DBF_WARNING, device, "%s",
1455			      "start_IO: -ENODEV device gone, retry");
1456		break;
1457	case -EIO:
1458		DBF_DEV_EVENT(DBF_WARNING, device, "%s",
1459			      "start_IO: -EIO device gone, retry");
1460		break;
1461	case -EINVAL:
1462		/* most likely caused in power management context */
1463		DBF_DEV_EVENT(DBF_WARNING, device, "%s",
1464			      "start_IO: -EINVAL device currently "
1465			      "not accessible");
1466		break;
1467	default:
1468		/* internal error 11 - unknown rc */
1469		snprintf(errorstring, ERRORLENGTH, "11 %d", rc);
1470		dev_err(&device->cdev->dev,
1471			"An error occurred in the DASD device driver, "
1472			"reason=%s\n", errorstring);
1473		BUG();
1474		break;
1475	}
1476	cqr->intrc = rc;
1477	return rc;
1478}
1479EXPORT_SYMBOL(dasd_start_IO);
1480
1481/*
1482 * Timeout function for dasd devices. This is used for different purposes
1483 *  1) missing interrupt handler for normal operation
1484 *  2) delayed start of request where start_IO failed with -EBUSY
1485 *  3) timeout for missing state change interrupts
1486 * The head of the ccw queue will have status DASD_CQR_IN_IO for 1),
1487 * DASD_CQR_QUEUED for 2) and 3).
1488 */
1489static void dasd_device_timeout(unsigned long ptr)
1490{
1491	unsigned long flags;
1492	struct dasd_device *device;
1493
1494	device = (struct dasd_device *) ptr;
1495	spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
1496	/* re-activate request queue */
1497	dasd_device_remove_stop_bits(device, DASD_STOPPED_PENDING);
1498	spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
1499	dasd_schedule_device_bh(device);
1500}
1501
1502/*
1503 * Setup timeout for a device in jiffies.
1504 */
1505void dasd_device_set_timer(struct dasd_device *device, int expires)
1506{
1507	if (expires == 0)
1508		del_timer(&device->timer);
1509	else
1510		mod_timer(&device->timer, jiffies + expires);
1511}
1512EXPORT_SYMBOL(dasd_device_set_timer);
1513
1514/*
1515 * Clear timeout for a device.
1516 */
1517void dasd_device_clear_timer(struct dasd_device *device)
1518{
1519	del_timer(&device->timer);
1520}
1521EXPORT_SYMBOL(dasd_device_clear_timer);
1522
1523static void dasd_handle_killed_request(struct ccw_device *cdev,
1524				       unsigned long intparm)
1525{
1526	struct dasd_ccw_req *cqr;
1527	struct dasd_device *device;
1528
1529	if (!intparm)
1530		return;
1531	cqr = (struct dasd_ccw_req *) intparm;
1532	if (cqr->status != DASD_CQR_IN_IO) {
1533		DBF_EVENT_DEVID(DBF_DEBUG, cdev,
1534				"invalid status in handle_killed_request: "
1535				"%02x", cqr->status);
1536		return;
1537	}
1538
1539	device = dasd_device_from_cdev_locked(cdev);
1540	if (IS_ERR(device)) {
1541		DBF_EVENT_DEVID(DBF_DEBUG, cdev, "%s",
1542				"unable to get device from cdev");
1543		return;
1544	}
1545
1546	if (!cqr->startdev ||
1547	    device != cqr->startdev ||
1548	    strncmp(cqr->startdev->discipline->ebcname,
1549		    (char *) &cqr->magic, 4)) {
1550		DBF_EVENT_DEVID(DBF_DEBUG, cdev, "%s",
1551				"invalid device in request");
1552		dasd_put_device(device);
1553		return;
1554	}
1555
1556	/* Schedule request to be retried. */
1557	cqr->status = DASD_CQR_QUEUED;
1558
1559	dasd_device_clear_timer(device);
1560	dasd_schedule_device_bh(device);
1561	dasd_put_device(device);
1562}
1563
1564void dasd_generic_handle_state_change(struct dasd_device *device)
1565{
1566	/* First of all start sense subsystem status request. */
1567	dasd_eer_snss(device);
1568
1569	dasd_device_remove_stop_bits(device, DASD_STOPPED_PENDING);
1570	dasd_schedule_device_bh(device);
1571	if (device->block)
1572		dasd_schedule_block_bh(device->block);
1573}
1574EXPORT_SYMBOL_GPL(dasd_generic_handle_state_change);
1575
1576/*
1577 * Interrupt handler for "normal" ssch-io based dasd devices.
1578 */
1579void dasd_int_handler(struct ccw_device *cdev, unsigned long intparm,
1580		      struct irb *irb)
1581{
1582	struct dasd_ccw_req *cqr, *next;
1583	struct dasd_device *device;
1584	unsigned long long now;
1585	int expires;
1586
1587	if (IS_ERR(irb)) {
1588		switch (PTR_ERR(irb)) {
1589		case -EIO:
1590			break;
1591		case -ETIMEDOUT:
1592			DBF_EVENT_DEVID(DBF_WARNING, cdev, "%s: "
1593					"request timed out\n", __func__);
1594			break;
1595		default:
1596			DBF_EVENT_DEVID(DBF_WARNING, cdev, "%s: "
1597					"unknown error %ld\n", __func__,
1598					PTR_ERR(irb));
1599		}
1600		dasd_handle_killed_request(cdev, intparm);
1601		return;
1602	}
1603
1604	now = get_tod_clock();
1605	cqr = (struct dasd_ccw_req *) intparm;
1606	/* check for conditions that should be handled immediately */
1607	if (!cqr ||
1608	    !(scsw_dstat(&irb->scsw) == (DEV_STAT_CHN_END | DEV_STAT_DEV_END) &&
1609	      scsw_cstat(&irb->scsw) == 0)) {
1610		if (cqr)
1611			memcpy(&cqr->irb, irb, sizeof(*irb));
1612		device = dasd_device_from_cdev_locked(cdev);
1613		if (IS_ERR(device))
1614			return;
1615		/* ignore unsolicited interrupts for DIAG discipline */
1616		if (device->discipline == dasd_diag_discipline_pointer) {
1617			dasd_put_device(device);
1618			return;
1619		}
1620		device->discipline->dump_sense_dbf(device, irb, "int");
1621		if (device->features & DASD_FEATURE_ERPLOG)
1622			device->discipline->dump_sense(device, cqr, irb);
1623		device->discipline->check_for_device_change(device, cqr, irb);
1624		dasd_put_device(device);
1625	}
1626
1627	/* check for for attention message */
1628	if (scsw_dstat(&irb->scsw) & DEV_STAT_ATTENTION) {
1629		device = dasd_device_from_cdev_locked(cdev);
1630		device->discipline->check_attention(device, irb->esw.esw1.lpum);
1631		dasd_put_device(device);
1632	}
1633
1634	if (!cqr)
1635		return;
1636
1637	device = (struct dasd_device *) cqr->startdev;
1638	if (!device ||
1639	    strncmp(device->discipline->ebcname, (char *) &cqr->magic, 4)) {
1640		DBF_EVENT_DEVID(DBF_DEBUG, cdev, "%s",
1641				"invalid device in request");
1642		return;
1643	}
1644
1645	/* Check for clear pending */
1646	if (cqr->status == DASD_CQR_CLEAR_PENDING &&
1647	    scsw_fctl(&irb->scsw) & SCSW_FCTL_CLEAR_FUNC) {
1648		cqr->status = DASD_CQR_CLEARED;
1649		dasd_device_clear_timer(device);
1650		wake_up(&dasd_flush_wq);
1651		dasd_schedule_device_bh(device);
1652		return;
1653	}
1654
1655	/* check status - the request might have been killed by dyn detach */
1656	if (cqr->status != DASD_CQR_IN_IO) {
1657		DBF_DEV_EVENT(DBF_DEBUG, device, "invalid status: bus_id %s, "
1658			      "status %02x", dev_name(&cdev->dev), cqr->status);
1659		return;
1660	}
1661
1662	next = NULL;
1663	expires = 0;
1664	if (scsw_dstat(&irb->scsw) == (DEV_STAT_CHN_END | DEV_STAT_DEV_END) &&
1665	    scsw_cstat(&irb->scsw) == 0) {
1666		/* request was completed successfully */
1667		cqr->status = DASD_CQR_SUCCESS;
1668		cqr->stopclk = now;
1669		/* Start first request on queue if possible -> fast_io. */
1670		if (cqr->devlist.next != &device->ccw_queue) {
1671			next = list_entry(cqr->devlist.next,
1672					  struct dasd_ccw_req, devlist);
1673		}
1674	} else {  /* error */
1675		/*
1676		 * If we don't want complex ERP for this request, then just
1677		 * reset this and retry it in the fastpath
1678		 */
1679		if (!test_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags) &&
1680		    cqr->retries > 0) {
1681			if (cqr->lpm == device->path_data.opm)
1682				DBF_DEV_EVENT(DBF_DEBUG, device,
1683					      "default ERP in fastpath "
1684					      "(%i retries left)",
1685					      cqr->retries);
1686			if (!test_bit(DASD_CQR_VERIFY_PATH, &cqr->flags))
1687				cqr->lpm = device->path_data.opm;
1688			cqr->status = DASD_CQR_QUEUED;
1689			next = cqr;
1690		} else
1691			cqr->status = DASD_CQR_ERROR;
1692	}
1693	if (next && (next->status == DASD_CQR_QUEUED) &&
1694	    (!device->stopped)) {
1695		if (device->discipline->start_IO(next) == 0)
1696			expires = next->expires;
1697	}
1698	if (expires != 0)
1699		dasd_device_set_timer(device, expires);
1700	else
1701		dasd_device_clear_timer(device);
1702	dasd_schedule_device_bh(device);
1703}
1704EXPORT_SYMBOL(dasd_int_handler);
1705
1706enum uc_todo dasd_generic_uc_handler(struct ccw_device *cdev, struct irb *irb)
1707{
1708	struct dasd_device *device;
1709
1710	device = dasd_device_from_cdev_locked(cdev);
1711
1712	if (IS_ERR(device))
1713		goto out;
1714	if (test_bit(DASD_FLAG_OFFLINE, &device->flags) ||
1715	   device->state != device->target ||
1716	   !device->discipline->check_for_device_change){
1717		dasd_put_device(device);
1718		goto out;
1719	}
1720	if (device->discipline->dump_sense_dbf)
1721		device->discipline->dump_sense_dbf(device, irb, "uc");
1722	device->discipline->check_for_device_change(device, NULL, irb);
1723	dasd_put_device(device);
1724out:
1725	return UC_TODO_RETRY;
1726}
1727EXPORT_SYMBOL_GPL(dasd_generic_uc_handler);
1728
1729/*
1730 * If we have an error on a dasd_block layer request then we cancel
1731 * and return all further requests from the same dasd_block as well.
1732 */
1733static void __dasd_device_recovery(struct dasd_device *device,
1734				   struct dasd_ccw_req *ref_cqr)
1735{
1736	struct list_head *l, *n;
1737	struct dasd_ccw_req *cqr;
1738
1739	/*
1740	 * only requeue request that came from the dasd_block layer
1741	 */
1742	if (!ref_cqr->block)
1743		return;
1744
1745	list_for_each_safe(l, n, &device->ccw_queue) {
1746		cqr = list_entry(l, struct dasd_ccw_req, devlist);
1747		if (cqr->status == DASD_CQR_QUEUED &&
1748		    ref_cqr->block == cqr->block) {
1749			cqr->status = DASD_CQR_CLEARED;
1750		}
1751	}
1752};
1753
1754/*
1755 * Remove those ccw requests from the queue that need to be returned
1756 * to the upper layer.
1757 */
1758static void __dasd_device_process_ccw_queue(struct dasd_device *device,
1759					    struct list_head *final_queue)
1760{
1761	struct list_head *l, *n;
1762	struct dasd_ccw_req *cqr;
1763
1764	/* Process request with final status. */
1765	list_for_each_safe(l, n, &device->ccw_queue) {
1766		cqr = list_entry(l, struct dasd_ccw_req, devlist);
1767
1768		/* Skip any non-final request. */
1769		if (cqr->status == DASD_CQR_QUEUED ||
1770		    cqr->status == DASD_CQR_IN_IO ||
1771		    cqr->status == DASD_CQR_CLEAR_PENDING)
1772			continue;
1773		if (cqr->status == DASD_CQR_ERROR) {
1774			__dasd_device_recovery(device, cqr);
1775		}
1776		/* Rechain finished requests to final queue */
1777		list_move_tail(&cqr->devlist, final_queue);
1778	}
1779}
1780
1781/*
1782 * the cqrs from the final queue are returned to the upper layer
1783 * by setting a dasd_block state and calling the callback function
1784 */
1785static void __dasd_device_process_final_queue(struct dasd_device *device,
1786					      struct list_head *final_queue)
1787{
1788	struct list_head *l, *n;
1789	struct dasd_ccw_req *cqr;
1790	struct dasd_block *block;
1791	void (*callback)(struct dasd_ccw_req *, void *data);
1792	void *callback_data;
1793	char errorstring[ERRORLENGTH];
1794
1795	list_for_each_safe(l, n, final_queue) {
1796		cqr = list_entry(l, struct dasd_ccw_req, devlist);
1797		list_del_init(&cqr->devlist);
1798		block = cqr->block;
1799		callback = cqr->callback;
1800		callback_data = cqr->callback_data;
1801		if (block)
1802			spin_lock_bh(&block->queue_lock);
1803		switch (cqr->status) {
1804		case DASD_CQR_SUCCESS:
1805			cqr->status = DASD_CQR_DONE;
1806			break;
1807		case DASD_CQR_ERROR:
1808			cqr->status = DASD_CQR_NEED_ERP;
1809			break;
1810		case DASD_CQR_CLEARED:
1811			cqr->status = DASD_CQR_TERMINATED;
1812			break;
1813		default:
1814			/* internal error 12 - wrong cqr status*/
1815			snprintf(errorstring, ERRORLENGTH, "12 %p %x02", cqr, cqr->status);
1816			dev_err(&device->cdev->dev,
1817				"An error occurred in the DASD device driver, "
1818				"reason=%s\n", errorstring);
1819			BUG();
1820		}
1821		if (cqr->callback != NULL)
1822			(callback)(cqr, callback_data);
1823		if (block)
1824			spin_unlock_bh(&block->queue_lock);
1825	}
1826}
1827
1828/*
1829 * Take a look at the first request on the ccw queue and check
1830 * if it reached its expire time. If so, terminate the IO.
1831 */
1832static void __dasd_device_check_expire(struct dasd_device *device)
1833{
1834	struct dasd_ccw_req *cqr;
1835
1836	if (list_empty(&device->ccw_queue))
1837		return;
1838	cqr = list_entry(device->ccw_queue.next, struct dasd_ccw_req, devlist);
1839	if ((cqr->status == DASD_CQR_IN_IO && cqr->expires != 0) &&
1840	    (time_after_eq(jiffies, cqr->expires + cqr->starttime))) {
1841		if (test_bit(DASD_FLAG_SAFE_OFFLINE_RUNNING, &device->flags)) {
1842			/*
1843			 * IO in safe offline processing should not
1844			 * run out of retries
1845			 */
1846			cqr->retries++;
1847		}
1848		if (device->discipline->term_IO(cqr) != 0) {
1849			/* Hmpf, try again in 5 sec */
1850			dev_err(&device->cdev->dev,
1851				"cqr %p timed out (%lus) but cannot be "
1852				"ended, retrying in 5 s\n",
1853				cqr, (cqr->expires/HZ));
1854			cqr->expires += 5*HZ;
1855			dasd_device_set_timer(device, 5*HZ);
1856		} else {
1857			dev_err(&device->cdev->dev,
1858				"cqr %p timed out (%lus), %i retries "
1859				"remaining\n", cqr, (cqr->expires/HZ),
1860				cqr->retries);
1861		}
1862	}
1863}
1864
1865/*
1866 * return 1 when device is not eligible for IO
1867 */
1868static int __dasd_device_is_unusable(struct dasd_device *device,
1869				     struct dasd_ccw_req *cqr)
1870{
1871	int mask = ~(DASD_STOPPED_DC_WAIT | DASD_UNRESUMED_PM);
1872
1873	if (test_bit(DASD_FLAG_OFFLINE, &device->flags)) {
1874		/* dasd is being set offline. */
1875		return 1;
1876	}
1877	if (device->stopped) {
1878		if (device->stopped & mask) {
1879			/* stopped and CQR will not change that. */
1880			return 1;
1881		}
1882		if (!test_bit(DASD_CQR_VERIFY_PATH, &cqr->flags)) {
1883			/* CQR is not able to change device to
1884			 * operational. */
1885			return 1;
1886		}
1887		/* CQR required to get device operational. */
1888	}
1889	return 0;
1890}
1891
1892/*
1893 * Take a look at the first request on the ccw queue and check
1894 * if it needs to be started.
1895 */
1896static void __dasd_device_start_head(struct dasd_device *device)
1897{
1898	struct dasd_ccw_req *cqr;
1899	int rc;
1900
1901	if (list_empty(&device->ccw_queue))
1902		return;
1903	cqr = list_entry(device->ccw_queue.next, struct dasd_ccw_req, devlist);
1904	if (cqr->status != DASD_CQR_QUEUED)
1905		return;
1906	/* if device is not usable return request to upper layer */
1907	if (__dasd_device_is_unusable(device, cqr)) {
1908		cqr->intrc = -EAGAIN;
1909		cqr->status = DASD_CQR_CLEARED;
1910		dasd_schedule_device_bh(device);
1911		return;
1912	}
1913
1914	rc = device->discipline->start_IO(cqr);
1915	if (rc == 0)
1916		dasd_device_set_timer(device, cqr->expires);
1917	else if (rc == -EACCES) {
1918		dasd_schedule_device_bh(device);
1919	} else
1920		/* Hmpf, try again in 1/2 sec */
1921		dasd_device_set_timer(device, 50);
1922}
1923
1924static void __dasd_device_check_path_events(struct dasd_device *device)
1925{
1926	int rc;
1927
1928	if (device->path_data.tbvpm) {
1929		if (device->stopped & ~(DASD_STOPPED_DC_WAIT |
1930					DASD_UNRESUMED_PM))
1931			return;
1932		rc = device->discipline->verify_path(
1933			device, device->path_data.tbvpm);
1934		if (rc)
1935			dasd_device_set_timer(device, 50);
1936		else
1937			device->path_data.tbvpm = 0;
1938	}
1939};
1940
1941/*
1942 * Go through all request on the dasd_device request queue,
1943 * terminate them on the cdev if necessary, and return them to the
1944 * submitting layer via callback.
1945 * Note:
1946 * Make sure that all 'submitting layers' still exist when
1947 * this function is called!. In other words, when 'device' is a base
1948 * device then all block layer requests must have been removed before
1949 * via dasd_flush_block_queue.
1950 */
1951int dasd_flush_device_queue(struct dasd_device *device)
1952{
1953	struct dasd_ccw_req *cqr, *n;
1954	int rc;
1955	struct list_head flush_queue;
1956
1957	INIT_LIST_HEAD(&flush_queue);
1958	spin_lock_irq(get_ccwdev_lock(device->cdev));
1959	rc = 0;
1960	list_for_each_entry_safe(cqr, n, &device->ccw_queue, devlist) {
1961		/* Check status and move request to flush_queue */
1962		switch (cqr->status) {
1963		case DASD_CQR_IN_IO:
1964			rc = device->discipline->term_IO(cqr);
1965			if (rc) {
1966				/* unable to terminate requeust */
1967				dev_err(&device->cdev->dev,
1968					"Flushing the DASD request queue "
1969					"failed for request %p\n", cqr);
1970				/* stop flush processing */
1971				goto finished;
1972			}
1973			break;
1974		case DASD_CQR_QUEUED:
1975			cqr->stopclk = get_tod_clock();
1976			cqr->status = DASD_CQR_CLEARED;
1977			break;
1978		default: /* no need to modify the others */
1979			break;
1980		}
1981		list_move_tail(&cqr->devlist, &flush_queue);
1982	}
1983finished:
1984	spin_unlock_irq(get_ccwdev_lock(device->cdev));
1985	/*
1986	 * After this point all requests must be in state CLEAR_PENDING,
1987	 * CLEARED, SUCCESS or ERROR. Now wait for CLEAR_PENDING to become
1988	 * one of the others.
1989	 */
1990	list_for_each_entry_safe(cqr, n, &flush_queue, devlist)
1991		wait_event(dasd_flush_wq,
1992			   (cqr->status != DASD_CQR_CLEAR_PENDING));
1993	/*
1994	 * Now set each request back to TERMINATED, DONE or NEED_ERP
1995	 * and call the callback function of flushed requests
1996	 */
1997	__dasd_device_process_final_queue(device, &flush_queue);
1998	return rc;
1999}
2000EXPORT_SYMBOL_GPL(dasd_flush_device_queue);
2001
2002/*
2003 * Acquire the device lock and process queues for the device.
2004 */
2005static void dasd_device_tasklet(struct dasd_device *device)
2006{
2007	struct list_head final_queue;
2008
2009	atomic_set (&device->tasklet_scheduled, 0);
2010	INIT_LIST_HEAD(&final_queue);
2011	spin_lock_irq(get_ccwdev_lock(device->cdev));
2012	/* Check expire time of first request on the ccw queue. */
2013	__dasd_device_check_expire(device);
2014	/* find final requests on ccw queue */
2015	__dasd_device_process_ccw_queue(device, &final_queue);
2016	__dasd_device_check_path_events(device);
2017	spin_unlock_irq(get_ccwdev_lock(device->cdev));
2018	/* Now call the callback function of requests with final status */
2019	__dasd_device_process_final_queue(device, &final_queue);
2020	spin_lock_irq(get_ccwdev_lock(device->cdev));
2021	/* Now check if the head of the ccw queue needs to be started. */
2022	__dasd_device_start_head(device);
2023	spin_unlock_irq(get_ccwdev_lock(device->cdev));
2024	if (waitqueue_active(&shutdown_waitq))
2025		wake_up(&shutdown_waitq);
2026	dasd_put_device(device);
2027}
2028
2029/*
2030 * Schedules a call to dasd_tasklet over the device tasklet.
2031 */
2032void dasd_schedule_device_bh(struct dasd_device *device)
2033{
2034	/* Protect against rescheduling. */
2035	if (atomic_cmpxchg (&device->tasklet_scheduled, 0, 1) != 0)
2036		return;
2037	dasd_get_device(device);
2038	tasklet_hi_schedule(&device->tasklet);
2039}
2040EXPORT_SYMBOL(dasd_schedule_device_bh);
2041
2042void dasd_device_set_stop_bits(struct dasd_device *device, int bits)
2043{
2044	device->stopped |= bits;
2045}
2046EXPORT_SYMBOL_GPL(dasd_device_set_stop_bits);
2047
2048void dasd_device_remove_stop_bits(struct dasd_device *device, int bits)
2049{
2050	device->stopped &= ~bits;
2051	if (!device->stopped)
2052		wake_up(&generic_waitq);
2053}
2054EXPORT_SYMBOL_GPL(dasd_device_remove_stop_bits);
2055
2056/*
2057 * Queue a request to the head of the device ccw_queue.
2058 * Start the I/O if possible.
2059 */
2060void dasd_add_request_head(struct dasd_ccw_req *cqr)
2061{
2062	struct dasd_device *device;
2063	unsigned long flags;
2064
2065	device = cqr->startdev;
2066	spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
2067	cqr->status = DASD_CQR_QUEUED;
2068	list_add(&cqr->devlist, &device->ccw_queue);
2069	/* let the bh start the request to keep them in order */
2070	dasd_schedule_device_bh(device);
2071	spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
2072}
2073EXPORT_SYMBOL(dasd_add_request_head);
2074
2075/*
2076 * Queue a request to the tail of the device ccw_queue.
2077 * Start the I/O if possible.
2078 */
2079void dasd_add_request_tail(struct dasd_ccw_req *cqr)
2080{
2081	struct dasd_device *device;
2082	unsigned long flags;
2083
2084	device = cqr->startdev;
2085	spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
2086	cqr->status = DASD_CQR_QUEUED;
2087	list_add_tail(&cqr->devlist, &device->ccw_queue);
2088	/* let the bh start the request to keep them in order */
2089	dasd_schedule_device_bh(device);
2090	spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
2091}
2092EXPORT_SYMBOL(dasd_add_request_tail);
2093
2094/*
2095 * Wakeup helper for the 'sleep_on' functions.
2096 */
2097void dasd_wakeup_cb(struct dasd_ccw_req *cqr, void *data)
2098{
2099	spin_lock_irq(get_ccwdev_lock(cqr->startdev->cdev));
2100	cqr->callback_data = DASD_SLEEPON_END_TAG;
2101	spin_unlock_irq(get_ccwdev_lock(cqr->startdev->cdev));
2102	wake_up(&generic_waitq);
2103}
2104EXPORT_SYMBOL_GPL(dasd_wakeup_cb);
2105
2106static inline int _wait_for_wakeup(struct dasd_ccw_req *cqr)
2107{
2108	struct dasd_device *device;
2109	int rc;
2110
2111	device = cqr->startdev;
2112	spin_lock_irq(get_ccwdev_lock(device->cdev));
2113	rc = (cqr->callback_data == DASD_SLEEPON_END_TAG);
2114	spin_unlock_irq(get_ccwdev_lock(device->cdev));
2115	return rc;
2116}
2117
2118/*
2119 * checks if error recovery is necessary, returns 1 if yes, 0 otherwise.
2120 */
2121static int __dasd_sleep_on_erp(struct dasd_ccw_req *cqr)
2122{
2123	struct dasd_device *device;
2124	dasd_erp_fn_t erp_fn;
2125
2126	if (cqr->status == DASD_CQR_FILLED)
2127		return 0;
2128	device = cqr->startdev;
2129	if (test_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags)) {
2130		if (cqr->status == DASD_CQR_TERMINATED) {
2131			device->discipline->handle_terminated_request(cqr);
2132			return 1;
2133		}
2134		if (cqr->status == DASD_CQR_NEED_ERP) {
2135			erp_fn = device->discipline->erp_action(cqr);
2136			erp_fn(cqr);
2137			return 1;
2138		}
2139		if (cqr->status == DASD_CQR_FAILED)
2140			dasd_log_sense(cqr, &cqr->irb);
2141		if (cqr->refers) {
2142			__dasd_process_erp(device, cqr);
2143			return 1;
2144		}
2145	}
2146	return 0;
2147}
2148
2149static int __dasd_sleep_on_loop_condition(struct dasd_ccw_req *cqr)
2150{
2151	if (test_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags)) {
2152		if (cqr->refers) /* erp is not done yet */
2153			return 1;
2154		return ((cqr->status != DASD_CQR_DONE) &&
2155			(cqr->status != DASD_CQR_FAILED));
2156	} else
2157		return (cqr->status == DASD_CQR_FILLED);
2158}
2159
2160static int _dasd_sleep_on(struct dasd_ccw_req *maincqr, int interruptible)
2161{
2162	struct dasd_device *device;
2163	int rc;
2164	struct list_head ccw_queue;
2165	struct dasd_ccw_req *cqr;
2166
2167	INIT_LIST_HEAD(&ccw_queue);
2168	maincqr->status = DASD_CQR_FILLED;
2169	device = maincqr->startdev;
2170	list_add(&maincqr->blocklist, &ccw_queue);
2171	for (cqr = maincqr;  __dasd_sleep_on_loop_condition(cqr);
2172	     cqr = list_first_entry(&ccw_queue,
2173				    struct dasd_ccw_req, blocklist)) {
2174
2175		if (__dasd_sleep_on_erp(cqr))
2176			continue;
2177		if (cqr->status != DASD_CQR_FILLED) /* could be failed */
2178			continue;
2179		if (test_bit(DASD_FLAG_LOCK_STOLEN, &device->flags) &&
2180		    !test_bit(DASD_CQR_ALLOW_SLOCK, &cqr->flags)) {
2181			cqr->status = DASD_CQR_FAILED;
2182			cqr->intrc = -EPERM;
2183			continue;
2184		}
2185		/* Non-temporary stop condition will trigger fail fast */
2186		if (device->stopped & ~DASD_STOPPED_PENDING &&
2187		    test_bit(DASD_CQR_FLAGS_FAILFAST, &cqr->flags) &&
2188		    (!dasd_eer_enabled(device))) {
2189			cqr->status = DASD_CQR_FAILED;
2190			cqr->intrc = -ENOLINK;
2191			continue;
2192		}
2193		/*
2194		 * Don't try to start requests if device is stopped
2195		 * except path verification requests
2196		 */
2197		if (!test_bit(DASD_CQR_VERIFY_PATH, &cqr->flags)) {
2198			if (interruptible) {
2199				rc = wait_event_interruptible(
2200					generic_waitq, !(device->stopped));
2201				if (rc == -ERESTARTSYS) {
2202					cqr->status = DASD_CQR_FAILED;
2203					maincqr->intrc = rc;
2204					continue;
2205				}
2206			} else
2207				wait_event(generic_waitq, !(device->stopped));
2208		}
2209		if (!cqr->callback)
2210			cqr->callback = dasd_wakeup_cb;
2211
2212		cqr->callback_data = DASD_SLEEPON_START_TAG;
2213		dasd_add_request_tail(cqr);
2214		if (interruptible) {
2215			rc = wait_event_interruptible(
2216				generic_waitq, _wait_for_wakeup(cqr));
2217			if (rc == -ERESTARTSYS) {
2218				dasd_cancel_req(cqr);
2219				/* wait (non-interruptible) for final status */
2220				wait_event(generic_waitq,
2221					   _wait_for_wakeup(cqr));
2222				cqr->status = DASD_CQR_FAILED;
2223				maincqr->intrc = rc;
2224				continue;
2225			}
2226		} else
2227			wait_event(generic_waitq, _wait_for_wakeup(cqr));
2228	}
2229
2230	maincqr->endclk = get_tod_clock();
2231	if ((maincqr->status != DASD_CQR_DONE) &&
2232	    (maincqr->intrc != -ERESTARTSYS))
2233		dasd_log_sense(maincqr, &maincqr->irb);
2234	if (maincqr->status == DASD_CQR_DONE)
2235		rc = 0;
2236	else if (maincqr->intrc)
2237		rc = maincqr->intrc;
2238	else
2239		rc = -EIO;
2240	return rc;
2241}
2242
2243static inline int _wait_for_wakeup_queue(struct list_head *ccw_queue)
2244{
2245	struct dasd_ccw_req *cqr;
2246
2247	list_for_each_entry(cqr, ccw_queue, blocklist) {
2248		if (cqr->callback_data != DASD_SLEEPON_END_TAG)
2249			return 0;
2250	}
2251
2252	return 1;
2253}
2254
2255static int _dasd_sleep_on_queue(struct list_head *ccw_queue, int interruptible)
2256{
2257	struct dasd_device *device;
2258	struct dasd_ccw_req *cqr, *n;
2259	int rc;
2260
2261retry:
2262	list_for_each_entry_safe(cqr, n, ccw_queue, blocklist) {
2263		device = cqr->startdev;
2264		if (cqr->status != DASD_CQR_FILLED) /*could be failed*/
2265			continue;
2266
2267		if (test_bit(DASD_FLAG_LOCK_STOLEN, &device->flags) &&
2268		    !test_bit(DASD_CQR_ALLOW_SLOCK, &cqr->flags)) {
2269			cqr->status = DASD_CQR_FAILED;
2270			cqr->intrc = -EPERM;
2271			continue;
2272		}
2273		/*Non-temporary stop condition will trigger fail fast*/
2274		if (device->stopped & ~DASD_STOPPED_PENDING &&
2275		    test_bit(DASD_CQR_FLAGS_FAILFAST, &cqr->flags) &&
2276		    !dasd_eer_enabled(device)) {
2277			cqr->status = DASD_CQR_FAILED;
2278			cqr->intrc = -EAGAIN;
2279			continue;
2280		}
2281
2282		/*Don't try to start requests if device is stopped*/
2283		if (interruptible) {
2284			rc = wait_event_interruptible(
2285				generic_waitq, !device->stopped);
2286			if (rc == -ERESTARTSYS) {
2287				cqr->status = DASD_CQR_FAILED;
2288				cqr->intrc = rc;
2289				continue;
2290			}
2291		} else
2292			wait_event(generic_waitq, !(device->stopped));
2293
2294		if (!cqr->callback)
2295			cqr->callback = dasd_wakeup_cb;
2296		cqr->callback_data = DASD_SLEEPON_START_TAG;
2297		dasd_add_request_tail(cqr);
2298	}
2299
2300	wait_event(generic_waitq, _wait_for_wakeup_queue(ccw_queue));
2301
2302	rc = 0;
2303	list_for_each_entry_safe(cqr, n, ccw_queue, blocklist) {
2304		/*
2305		 * for alias devices simplify error recovery and
2306		 * return to upper layer
2307		 * do not skip ERP requests
2308		 */
2309		if (cqr->startdev != cqr->basedev && !cqr->refers &&
2310		    (cqr->status == DASD_CQR_TERMINATED ||
2311		     cqr->status == DASD_CQR_NEED_ERP))
2312			return -EAGAIN;
2313
2314		/* normal recovery for basedev IO */
2315		if (__dasd_sleep_on_erp(cqr))
2316			/* handle erp first */
2317			goto retry;
2318	}
2319
2320	return 0;
2321}
2322
2323/*
2324 * Queue a request to the tail of the device ccw_queue and wait for
2325 * it's completion.
2326 */
2327int dasd_sleep_on(struct dasd_ccw_req *cqr)
2328{
2329	return _dasd_sleep_on(cqr, 0);
2330}
2331EXPORT_SYMBOL(dasd_sleep_on);
2332
2333/*
2334 * Start requests from a ccw_queue and wait for their completion.
2335 */
2336int dasd_sleep_on_queue(struct list_head *ccw_queue)
2337{
2338	return _dasd_sleep_on_queue(ccw_queue, 0);
2339}
2340EXPORT_SYMBOL(dasd_sleep_on_queue);
2341
2342/*
2343 * Queue a request to the tail of the device ccw_queue and wait
2344 * interruptible for it's completion.
2345 */
2346int dasd_sleep_on_interruptible(struct dasd_ccw_req *cqr)
2347{
2348	return _dasd_sleep_on(cqr, 1);
2349}
2350EXPORT_SYMBOL(dasd_sleep_on_interruptible);
2351
2352/*
2353 * Whoa nelly now it gets really hairy. For some functions (e.g. steal lock
2354 * for eckd devices) the currently running request has to be terminated
2355 * and be put back to status queued, before the special request is added
2356 * to the head of the queue. Then the special request is waited on normally.
2357 */
2358static inline int _dasd_term_running_cqr(struct dasd_device *device)
2359{
2360	struct dasd_ccw_req *cqr;
2361	int rc;
2362
2363	if (list_empty(&device->ccw_queue))
2364		return 0;
2365	cqr = list_entry(device->ccw_queue.next, struct dasd_ccw_req, devlist);
2366	rc = device->discipline->term_IO(cqr);
2367	if (!rc)
2368		/*
2369		 * CQR terminated because a more important request is pending.
2370		 * Undo decreasing of retry counter because this is
2371		 * not an error case.
2372		 */
2373		cqr->retries++;
2374	return rc;
2375}
2376
2377int dasd_sleep_on_immediatly(struct dasd_ccw_req *cqr)
2378{
2379	struct dasd_device *device;
2380	int rc;
2381
2382	device = cqr->startdev;
2383	if (test_bit(DASD_FLAG_LOCK_STOLEN, &device->flags) &&
2384	    !test_bit(DASD_CQR_ALLOW_SLOCK, &cqr->flags)) {
2385		cqr->status = DASD_CQR_FAILED;
2386		cqr->intrc = -EPERM;
2387		return -EIO;
2388	}
2389	spin_lock_irq(get_ccwdev_lock(device->cdev));
2390	rc = _dasd_term_running_cqr(device);
2391	if (rc) {
2392		spin_unlock_irq(get_ccwdev_lock(device->cdev));
2393		return rc;
2394	}
2395	cqr->callback = dasd_wakeup_cb;
2396	cqr->callback_data = DASD_SLEEPON_START_TAG;
2397	cqr->status = DASD_CQR_QUEUED;
2398	/*
2399	 * add new request as second
2400	 * first the terminated cqr needs to be finished
2401	 */
2402	list_add(&cqr->devlist, device->ccw_queue.next);
2403
2404	/* let the bh start the request to keep them in order */
2405	dasd_schedule_device_bh(device);
2406
2407	spin_unlock_irq(get_ccwdev_lock(device->cdev));
2408
2409	wait_event(generic_waitq, _wait_for_wakeup(cqr));
2410
2411	if (cqr->status == DASD_CQR_DONE)
2412		rc = 0;
2413	else if (cqr->intrc)
2414		rc = cqr->intrc;
2415	else
2416		rc = -EIO;
2417
2418	/* kick tasklets */
2419	dasd_schedule_device_bh(device);
2420	if (device->block)
2421		dasd_schedule_block_bh(device->block);
2422
2423	return rc;
2424}
2425EXPORT_SYMBOL(dasd_sleep_on_immediatly);
2426
2427/*
2428 * Cancels a request that was started with dasd_sleep_on_req.
2429 * This is useful to timeout requests. The request will be
2430 * terminated if it is currently in i/o.
2431 * Returns 0 if request termination was successful
2432 *	   negative error code if termination failed
2433 * Cancellation of a request is an asynchronous operation! The calling
2434 * function has to wait until the request is properly returned via callback.
2435 */
2436int dasd_cancel_req(struct dasd_ccw_req *cqr)
2437{
2438	struct dasd_device *device = cqr->startdev;
2439	unsigned long flags;
2440	int rc;
2441
2442	rc = 0;
2443	spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
2444	switch (cqr->status) {
2445	case DASD_CQR_QUEUED:
2446		/* request was not started - just set to cleared */
2447		cqr->status = DASD_CQR_CLEARED;
2448		if (cqr->callback_data == DASD_SLEEPON_START_TAG)
2449			cqr->callback_data = DASD_SLEEPON_END_TAG;
2450		break;
2451	case DASD_CQR_IN_IO:
2452		/* request in IO - terminate IO and release again */
2453		rc = device->discipline->term_IO(cqr);
2454		if (rc) {
2455			dev_err(&device->cdev->dev,
2456				"Cancelling request %p failed with rc=%d\n",
2457				cqr, rc);
2458		} else {
2459			cqr->stopclk = get_tod_clock();
2460		}
2461		break;
2462	default: /* already finished or clear pending - do nothing */
2463		break;
2464	}
2465	spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
2466	dasd_schedule_device_bh(device);
2467	return rc;
2468}
2469EXPORT_SYMBOL(dasd_cancel_req);
2470
2471/*
2472 * SECTION: Operations of the dasd_block layer.
2473 */
2474
2475/*
2476 * Timeout function for dasd_block. This is used when the block layer
2477 * is waiting for something that may not come reliably, (e.g. a state
2478 * change interrupt)
2479 */
2480static void dasd_block_timeout(unsigned long ptr)
2481{
2482	unsigned long flags;
2483	struct dasd_block *block;
2484
2485	block = (struct dasd_block *) ptr;
2486	spin_lock_irqsave(get_ccwdev_lock(block->base->cdev), flags);
2487	/* re-activate request queue */
2488	dasd_device_remove_stop_bits(block->base, DASD_STOPPED_PENDING);
2489	spin_unlock_irqrestore(get_ccwdev_lock(block->base->cdev), flags);
2490	dasd_schedule_block_bh(block);
2491}
2492
2493/*
2494 * Setup timeout for a dasd_block in jiffies.
2495 */
2496void dasd_block_set_timer(struct dasd_block *block, int expires)
2497{
2498	if (expires == 0)
2499		del_timer(&block->timer);
2500	else
2501		mod_timer(&block->timer, jiffies + expires);
2502}
2503EXPORT_SYMBOL(dasd_block_set_timer);
2504
2505/*
2506 * Clear timeout for a dasd_block.
2507 */
2508void dasd_block_clear_timer(struct dasd_block *block)
2509{
2510	del_timer(&block->timer);
2511}
2512EXPORT_SYMBOL(dasd_block_clear_timer);
2513
2514/*
2515 * Process finished error recovery ccw.
2516 */
2517static void __dasd_process_erp(struct dasd_device *device,
2518			       struct dasd_ccw_req *cqr)
2519{
2520	dasd_erp_fn_t erp_fn;
2521
2522	if (cqr->status == DASD_CQR_DONE)
2523		DBF_DEV_EVENT(DBF_NOTICE, device, "%s", "ERP successful");
2524	else
2525		dev_err(&device->cdev->dev, "ERP failed for the DASD\n");
2526	erp_fn = device->discipline->erp_postaction(cqr);
2527	erp_fn(cqr);
2528}
2529
2530/*
2531 * Fetch requests from the block device queue.
2532 */
2533static void __dasd_process_request_queue(struct dasd_block *block)
2534{
2535	struct request_queue *queue;
2536	struct request *req;
2537	struct dasd_ccw_req *cqr;
2538	struct dasd_device *basedev;
2539	unsigned long flags;
2540	queue = block->request_queue;
2541	basedev = block->base;
2542	/* No queue ? Then there is nothing to do. */
2543	if (queue == NULL)
2544		return;
2545
2546	/*
2547	 * We requeue request from the block device queue to the ccw
2548	 * queue only in two states. In state DASD_STATE_READY the
2549	 * partition detection is done and we need to requeue requests
2550	 * for that. State DASD_STATE_ONLINE is normal block device
2551	 * operation.
2552	 */
2553	if (basedev->state < DASD_STATE_READY) {
2554		while ((req = blk_fetch_request(block->request_queue)))
2555			__blk_end_request_all(req, -EIO);
2556		return;
2557	}
2558
2559	/* if device ist stopped do not fetch new requests */
2560	if (basedev->stopped)
2561		return;
2562
2563	/* Now we try to fetch requests from the request queue */
2564	while ((req = blk_peek_request(queue))) {
2565		if (basedev->features & DASD_FEATURE_READONLY &&
2566		    rq_data_dir(req) == WRITE) {
2567			DBF_DEV_EVENT(DBF_ERR, basedev,
2568				      "Rejecting write request %p",
2569				      req);
2570			blk_start_request(req);
2571			__blk_end_request_all(req, -EIO);
2572			continue;
2573		}
2574		if (test_bit(DASD_FLAG_ABORTALL, &basedev->flags) &&
2575		    (basedev->features & DASD_FEATURE_FAILFAST ||
2576		     blk_noretry_request(req))) {
2577			DBF_DEV_EVENT(DBF_ERR, basedev,
2578				      "Rejecting failfast request %p",
2579				      req);
2580			blk_start_request(req);
2581			__blk_end_request_all(req, -ETIMEDOUT);
2582			continue;
2583		}
2584		cqr = basedev->discipline->build_cp(basedev, block, req);
2585		if (IS_ERR(cqr)) {
2586			if (PTR_ERR(cqr) == -EBUSY)
2587				break;	/* normal end condition */
2588			if (PTR_ERR(cqr) == -ENOMEM)
2589				break;	/* terminate request queue loop */
2590			if (PTR_ERR(cqr) == -EAGAIN) {
2591				/*
2592				 * The current request cannot be build right
2593				 * now, we have to try later. If this request
2594				 * is the head-of-queue we stop the device
2595				 * for 1/2 second.
2596				 */
2597				if (!list_empty(&block->ccw_queue))
2598					break;
2599				spin_lock_irqsave(
2600					get_ccwdev_lock(basedev->cdev), flags);
2601				dasd_device_set_stop_bits(basedev,
2602							  DASD_STOPPED_PENDING);
2603				spin_unlock_irqrestore(
2604					get_ccwdev_lock(basedev->cdev), flags);
2605				dasd_block_set_timer(block, HZ/2);
2606				break;
2607			}
2608			DBF_DEV_EVENT(DBF_ERR, basedev,
2609				      "CCW creation failed (rc=%ld) "
2610				      "on request %p",
2611				      PTR_ERR(cqr), req);
2612			blk_start_request(req);
2613			__blk_end_request_all(req, -EIO);
2614			continue;
2615		}
2616		/*
2617		 *  Note: callback is set to dasd_return_cqr_cb in
2618		 * __dasd_block_start_head to cover erp requests as well
2619		 */
2620		cqr->callback_data = (void *) req;
2621		cqr->status = DASD_CQR_FILLED;
2622		req->completion_data = cqr;
2623		blk_start_request(req);
2624		list_add_tail(&cqr->blocklist, &block->ccw_queue);
2625		INIT_LIST_HEAD(&cqr->devlist);
2626		dasd_profile_start(block, cqr, req);
2627	}
2628}
2629
2630static void __dasd_cleanup_cqr(struct dasd_ccw_req *cqr)
2631{
2632	struct request *req;
2633	int status;
2634	int error = 0;
2635
2636	req = (struct request *) cqr->callback_data;
2637	dasd_profile_end(cqr->block, cqr, req);
2638	status = cqr->block->base->discipline->free_cp(cqr, req);
2639	if (status < 0)
2640		error = status;
2641	else if (status == 0) {
2642		if (cqr->intrc == -EPERM)
2643			error = -EBADE;
2644		else if (cqr->intrc == -ENOLINK ||
2645			 cqr->intrc == -ETIMEDOUT)
2646			error = cqr->intrc;
2647		else
2648			error = -EIO;
2649	}
2650	__blk_end_request_all(req, error);
2651}
2652
2653/*
2654 * Process ccw request queue.
2655 */
2656static void __dasd_process_block_ccw_queue(struct dasd_block *block,
2657					   struct list_head *final_queue)
2658{
2659	struct list_head *l, *n;
2660	struct dasd_ccw_req *cqr;
2661	dasd_erp_fn_t erp_fn;
2662	unsigned long flags;
2663	struct dasd_device *base = block->base;
2664
2665restart:
2666	/* Process request with final status. */
2667	list_for_each_safe(l, n, &block->ccw_queue) {
2668		cqr = list_entry(l, struct dasd_ccw_req, blocklist);
2669		if (cqr->status != DASD_CQR_DONE &&
2670		    cqr->status != DASD_CQR_FAILED &&
2671		    cqr->status != DASD_CQR_NEED_ERP &&
2672		    cqr->status != DASD_CQR_TERMINATED)
2673			continue;
2674
2675		if (cqr->status == DASD_CQR_TERMINATED) {
2676			base->discipline->handle_terminated_request(cqr);
2677			goto restart;
2678		}
2679
2680		/*  Process requests that may be recovered */
2681		if (cqr->status == DASD_CQR_NEED_ERP) {
2682			erp_fn = base->discipline->erp_action(cqr);
2683			if (IS_ERR(erp_fn(cqr)))
2684				continue;
2685			goto restart;
2686		}
2687
2688		/* log sense for fatal error */
2689		if (cqr->status == DASD_CQR_FAILED) {
2690			dasd_log_sense(cqr, &cqr->irb);
2691		}
2692
2693		/* First of all call extended error reporting. */
2694		if (dasd_eer_enabled(base) &&
2695		    cqr->status == DASD_CQR_FAILED) {
2696			dasd_eer_write(base, cqr, DASD_EER_FATALERROR);
2697
2698			/* restart request  */
2699			cqr->status = DASD_CQR_FILLED;
2700			cqr->retries = 255;
2701			spin_lock_irqsave(get_ccwdev_lock(base->cdev), flags);
2702			dasd_device_set_stop_bits(base, DASD_STOPPED_QUIESCE);
2703			spin_unlock_irqrestore(get_ccwdev_lock(base->cdev),
2704					       flags);
2705			goto restart;
2706		}
2707
2708		/* Process finished ERP request. */
2709		if (cqr->refers) {
2710			__dasd_process_erp(base, cqr);
2711			goto restart;
2712		}
2713
2714		/* Rechain finished requests to final queue */
2715		cqr->endclk = get_tod_clock();
2716		list_move_tail(&cqr->blocklist, final_queue);
2717	}
2718}
2719
2720static void dasd_return_cqr_cb(struct dasd_ccw_req *cqr, void *data)
2721{
2722	dasd_schedule_block_bh(cqr->block);
2723}
2724
2725static void __dasd_block_start_head(struct dasd_block *block)
2726{
2727	struct dasd_ccw_req *cqr;
2728
2729	if (list_empty(&block->ccw_queue))
2730		return;
2731	/* We allways begin with the first requests on the queue, as some
2732	 * of previously started requests have to be enqueued on a
2733	 * dasd_device again for error recovery.
2734	 */
2735	list_for_each_entry(cqr, &block->ccw_queue, blocklist) {
2736		if (cqr->status != DASD_CQR_FILLED)
2737			continue;
2738		if (test_bit(DASD_FLAG_LOCK_STOLEN, &block->base->flags) &&
2739		    !test_bit(DASD_CQR_ALLOW_SLOCK, &cqr->flags)) {
2740			cqr->status = DASD_CQR_FAILED;
2741			cqr->intrc = -EPERM;
2742			dasd_schedule_block_bh(block);
2743			continue;
2744		}
2745		/* Non-temporary stop condition will trigger fail fast */
2746		if (block->base->stopped & ~DASD_STOPPED_PENDING &&
2747		    test_bit(DASD_CQR_FLAGS_FAILFAST, &cqr->flags) &&
2748		    (!dasd_eer_enabled(block->base))) {
2749			cqr->status = DASD_CQR_FAILED;
2750			cqr->intrc = -ENOLINK;
2751			dasd_schedule_block_bh(block);
2752			continue;
2753		}
2754		/* Don't try to start requests if device is stopped */
2755		if (block->base->stopped)
2756			return;
2757
2758		/* just a fail safe check, should not happen */
2759		if (!cqr->startdev)
2760			cqr->startdev = block->base;
2761
2762		/* make sure that the requests we submit find their way back */
2763		cqr->callback = dasd_return_cqr_cb;
2764
2765		dasd_add_request_tail(cqr);
2766	}
2767}
2768
2769/*
2770 * Central dasd_block layer routine. Takes requests from the generic
2771 * block layer request queue, creates ccw requests, enqueues them on
2772 * a dasd_device and processes ccw requests that have been returned.
2773 */
2774static void dasd_block_tasklet(struct dasd_block *block)
2775{
2776	struct list_head final_queue;
2777	struct list_head *l, *n;
2778	struct dasd_ccw_req *cqr;
2779
2780	atomic_set(&block->tasklet_scheduled, 0);
2781	INIT_LIST_HEAD(&final_queue);
2782	spin_lock(&block->queue_lock);
2783	/* Finish off requests on ccw queue */
2784	__dasd_process_block_ccw_queue(block, &final_queue);
2785	spin_unlock(&block->queue_lock);
2786	/* Now call the callback function of requests with final status */
2787	spin_lock_irq(&block->request_queue_lock);
2788	list_for_each_safe(l, n, &final_queue) {
2789		cqr = list_entry(l, struct dasd_ccw_req, blocklist);
2790		list_del_init(&cqr->blocklist);
2791		__dasd_cleanup_cqr(cqr);
2792	}
2793	spin_lock(&block->queue_lock);
2794	/* Get new request from the block device request queue */
2795	__dasd_process_request_queue(block);
2796	/* Now check if the head of the ccw queue needs to be started. */
2797	__dasd_block_start_head(block);
2798	spin_unlock(&block->queue_lock);
2799	spin_unlock_irq(&block->request_queue_lock);
2800	if (waitqueue_active(&shutdown_waitq))
2801		wake_up(&shutdown_waitq);
2802	dasd_put_device(block->base);
2803}
2804
2805static void _dasd_wake_block_flush_cb(struct dasd_ccw_req *cqr, void *data)
2806{
2807	wake_up(&dasd_flush_wq);
2808}
2809
2810/*
2811 * Requeue a request back to the block request queue
2812 * only works for block requests
2813 */
2814static int _dasd_requeue_request(struct dasd_ccw_req *cqr)
2815{
2816	struct dasd_block *block = cqr->block;
2817	struct request *req;
2818	unsigned long flags;
2819
2820	if (!block)
2821		return -EINVAL;
2822	spin_lock_irqsave(&block->queue_lock, flags);
2823	req = (struct request *) cqr->callback_data;
2824	blk_requeue_request(block->request_queue, req);
2825	spin_unlock_irqrestore(&block->queue_lock, flags);
2826
2827	return 0;
2828}
2829
2830/*
2831 * Go through all request on the dasd_block request queue, cancel them
2832 * on the respective dasd_device, and return them to the generic
2833 * block layer.
2834 */
2835static int dasd_flush_block_queue(struct dasd_block *block)
2836{
2837	struct dasd_ccw_req *cqr, *n;
2838	int rc, i;
2839	struct list_head flush_queue;
2840
2841	INIT_LIST_HEAD(&flush_queue);
2842	spin_lock_bh(&block->queue_lock);
2843	rc = 0;
2844restart:
2845	list_for_each_entry_safe(cqr, n, &block->ccw_queue, blocklist) {
2846		/* if this request currently owned by a dasd_device cancel it */
2847		if (cqr->status >= DASD_CQR_QUEUED)
2848			rc = dasd_cancel_req(cqr);
2849		if (rc < 0)
2850			break;
2851		/* Rechain request (including erp chain) so it won't be
2852		 * touched by the dasd_block_tasklet anymore.
2853		 * Replace the callback so we notice when the request
2854		 * is returned from the dasd_device layer.
2855		 */
2856		cqr->callback = _dasd_wake_block_flush_cb;
2857		for (i = 0; cqr != NULL; cqr = cqr->refers, i++)
2858			list_move_tail(&cqr->blocklist, &flush_queue);
2859		if (i > 1)
2860			/* moved more than one request - need to restart */
2861			goto restart;
2862	}
2863	spin_unlock_bh(&block->queue_lock);
2864	/* Now call the callback function of flushed requests */
2865restart_cb:
2866	list_for_each_entry_safe(cqr, n, &flush_queue, blocklist) {
2867		wait_event(dasd_flush_wq, (cqr->status < DASD_CQR_QUEUED));
2868		/* Process finished ERP request. */
2869		if (cqr->refers) {
2870			spin_lock_bh(&block->queue_lock);
2871			__dasd_process_erp(block->base, cqr);
2872			spin_unlock_bh(&block->queue_lock);
2873			/* restart list_for_xx loop since dasd_process_erp
2874			 * might remove multiple elements */
2875			goto restart_cb;
2876		}
2877		/* call the callback function */
2878		spin_lock_irq(&block->request_queue_lock);
2879		cqr->endclk = get_tod_clock();
2880		list_del_init(&cqr->blocklist);
2881		__dasd_cleanup_cqr(cqr);
2882		spin_unlock_irq(&block->request_queue_lock);
2883	}
2884	return rc;
2885}
2886
2887/*
2888 * Schedules a call to dasd_tasklet over the device tasklet.
2889 */
2890void dasd_schedule_block_bh(struct dasd_block *block)
2891{
2892	/* Protect against rescheduling. */
2893	if (atomic_cmpxchg(&block->tasklet_scheduled, 0, 1) != 0)
2894		return;
2895	/* life cycle of block is bound to it's base device */
2896	dasd_get_device(block->base);
2897	tasklet_hi_schedule(&block->tasklet);
2898}
2899EXPORT_SYMBOL(dasd_schedule_block_bh);
2900
2901
2902/*
2903 * SECTION: external block device operations
2904 * (request queue handling, open, release, etc.)
2905 */
2906
2907/*
2908 * Dasd request queue function. Called from ll_rw_blk.c
2909 */
2910static void do_dasd_request(struct request_queue *queue)
2911{
2912	struct dasd_block *block;
2913
2914	block = queue->queuedata;
2915	spin_lock(&block->queue_lock);
2916	/* Get new request from the block device request queue */
2917	__dasd_process_request_queue(block);
2918	/* Now check if the head of the ccw queue needs to be started. */
2919	__dasd_block_start_head(block);
2920	spin_unlock(&block->queue_lock);
2921}
2922
2923/*
2924 * Block timeout callback, called from the block layer
2925 *
2926 * request_queue lock is held on entry.
2927 *
2928 * Return values:
2929 * BLK_EH_RESET_TIMER if the request should be left running
2930 * BLK_EH_NOT_HANDLED if the request is handled or terminated
2931 *		      by the driver.
2932 */
2933enum blk_eh_timer_return dasd_times_out(struct request *req)
2934{
2935	struct dasd_ccw_req *cqr = req->completion_data;
2936	struct dasd_block *block = req->q->queuedata;
2937	struct dasd_device *device;
2938	int rc = 0;
2939
2940	if (!cqr)
2941		return BLK_EH_NOT_HANDLED;
2942
2943	device = cqr->startdev ? cqr->startdev : block->base;
2944	if (!device->blk_timeout)
2945		return BLK_EH_RESET_TIMER;
2946	DBF_DEV_EVENT(DBF_WARNING, device,
2947		      " dasd_times_out cqr %p status %x",
2948		      cqr, cqr->status);
2949
2950	spin_lock(&block->queue_lock);
2951	spin_lock(get_ccwdev_lock(device->cdev));
2952	cqr->retries = -1;
2953	cqr->intrc = -ETIMEDOUT;
2954	if (cqr->status >= DASD_CQR_QUEUED) {
2955		spin_unlock(get_ccwdev_lock(device->cdev));
2956		rc = dasd_cancel_req(cqr);
2957	} else if (cqr->status == DASD_CQR_FILLED ||
2958		   cqr->status == DASD_CQR_NEED_ERP) {
2959		cqr->status = DASD_CQR_TERMINATED;
2960		spin_unlock(get_ccwdev_lock(device->cdev));
2961	} else if (cqr->status == DASD_CQR_IN_ERP) {
2962		struct dasd_ccw_req *searchcqr, *nextcqr, *tmpcqr;
2963
2964		list_for_each_entry_safe(searchcqr, nextcqr,
2965					 &block->ccw_queue, blocklist) {
2966			tmpcqr = searchcqr;
2967			while (tmpcqr->refers)
2968				tmpcqr = tmpcqr->refers;
2969			if (tmpcqr != cqr)
2970				continue;
2971			/* searchcqr is an ERP request for cqr */
2972			searchcqr->retries = -1;
2973			searchcqr->intrc = -ETIMEDOUT;
2974			if (searchcqr->status >= DASD_CQR_QUEUED) {
2975				spin_unlock(get_ccwdev_lock(device->cdev));
2976				rc = dasd_cancel_req(searchcqr);
2977				spin_lock(get_ccwdev_lock(device->cdev));
2978			} else if ((searchcqr->status == DASD_CQR_FILLED) ||
2979				   (searchcqr->status == DASD_CQR_NEED_ERP)) {
2980				searchcqr->status = DASD_CQR_TERMINATED;
2981				rc = 0;
2982			} else if (searchcqr->status == DASD_CQR_IN_ERP) {
2983				/*
2984				 * Shouldn't happen; most recent ERP
2985				 * request is at the front of queue
2986				 */
2987				continue;
2988			}
2989			break;
2990		}
2991		spin_unlock(get_ccwdev_lock(device->cdev));
2992	}
2993	dasd_schedule_block_bh(block);
2994	spin_unlock(&block->queue_lock);
2995
2996	return rc ? BLK_EH_RESET_TIMER : BLK_EH_NOT_HANDLED;
2997}
2998
2999/*
3000 * Allocate and initialize request queue and default I/O scheduler.
3001 */
3002static int dasd_alloc_queue(struct dasd_block *block)
3003{
3004	block->request_queue = blk_init_queue(do_dasd_request,
3005					       &block->request_queue_lock);
3006	if (block->request_queue == NULL)
3007		return -ENOMEM;
3008
3009	block->request_queue->queuedata = block;
3010
3011	return 0;
3012}
3013
3014/*
3015 * Allocate and initialize request queue.
3016 */
3017static void dasd_setup_queue(struct dasd_block *block)
3018{
3019	int max;
3020
3021	if (block->base->features & DASD_FEATURE_USERAW) {
3022		/*
3023		 * the max_blocks value for raw_track access is 256
3024		 * it is higher than the native ECKD value because we
3025		 * only need one ccw per track
3026		 * so the max_hw_sectors are
3027		 * 2048 x 512B = 1024kB = 16 tracks
3028		 */
3029		max = 2048;
3030	} else {
3031		max = block->base->discipline->max_blocks << block->s2b_shift;
3032	}
3033	queue_flag_set_unlocked(QUEUE_FLAG_NONROT, block->request_queue);
3034	block->request_queue->limits.max_dev_sectors = max;
3035	blk_queue_logical_block_size(block->request_queue,
3036				     block->bp_block);
3037	blk_queue_max_hw_sectors(block->request_queue, max);
3038	blk_queue_max_segments(block->request_queue, -1L);
3039	/* with page sized segments we can translate each segement into
3040	 * one idaw/tidaw
3041	 */
3042	blk_queue_max_segment_size(block->request_queue, PAGE_SIZE);
3043	blk_queue_segment_boundary(block->request_queue, PAGE_SIZE - 1);
3044}
3045
3046/*
3047 * Deactivate and free request queue.
3048 */
3049static void dasd_free_queue(struct dasd_block *block)
3050{
3051	if (block->request_queue) {
3052		blk_cleanup_queue(block->request_queue);
3053		block->request_queue = NULL;
3054	}
3055}
3056
3057/*
3058 * Flush request on the request queue.
3059 */
3060static void dasd_flush_request_queue(struct dasd_block *block)
3061{
3062	struct request *req;
3063
3064	if (!block->request_queue)
3065		return;
3066
3067	spin_lock_irq(&block->request_queue_lock);
3068	while ((req = blk_fetch_request(block->request_queue)))
3069		__blk_end_request_all(req, -EIO);
3070	spin_unlock_irq(&block->request_queue_lock);
3071}
3072
3073static int dasd_open(struct block_device *bdev, fmode_t mode)
3074{
3075	struct dasd_device *base;
3076	int rc;
3077
3078	base = dasd_device_from_gendisk(bdev->bd_disk);
3079	if (!base)
3080		return -ENODEV;
3081
3082	atomic_inc(&base->block->open_count);
3083	if (test_bit(DASD_FLAG_OFFLINE, &base->flags)) {
3084		rc = -ENODEV;
3085		goto unlock;
3086	}
3087
3088	if (!try_module_get(base->discipline->owner)) {
3089		rc = -EINVAL;
3090		goto unlock;
3091	}
3092
3093	if (dasd_probeonly) {
3094		dev_info(&base->cdev->dev,
3095			 "Accessing the DASD failed because it is in "
3096			 "probeonly mode\n");
3097		rc = -EPERM;
3098		goto out;
3099	}
3100
3101	if (base->state <= DASD_STATE_BASIC) {
3102		DBF_DEV_EVENT(DBF_ERR, base, " %s",
3103			      " Cannot open unrecognized device");
3104		rc = -ENODEV;
3105		goto out;
3106	}
3107
3108	if ((mode & FMODE_WRITE) &&
3109	    (test_bit(DASD_FLAG_DEVICE_RO, &base->flags) ||
3110	     (base->features & DASD_FEATURE_READONLY))) {
3111		rc = -EROFS;
3112		goto out;
3113	}
3114
3115	dasd_put_device(base);
3116	return 0;
3117
3118out:
3119	module_put(base->discipline->owner);
3120unlock:
3121	atomic_dec(&base->block->open_count);
3122	dasd_put_device(base);
3123	return rc;
3124}
3125
3126static void dasd_release(struct gendisk *disk, fmode_t mode)
3127{
3128	struct dasd_device *base = dasd_device_from_gendisk(disk);
3129	if (base) {
3130		atomic_dec(&base->block->open_count);
3131		module_put(base->discipline->owner);
3132		dasd_put_device(base);
3133	}
3134}
3135
3136/*
3137 * Return disk geometry.
3138 */
3139static int dasd_getgeo(struct block_device *bdev, struct hd_geometry *geo)
3140{
3141	struct dasd_device *base;
3142
3143	base = dasd_device_from_gendisk(bdev->bd_disk);
3144	if (!base)
3145		return -ENODEV;
3146
3147	if (!base->discipline ||
3148	    !base->discipline->fill_geometry) {
3149		dasd_put_device(base);
3150		return -EINVAL;
3151	}
3152	base->discipline->fill_geometry(base->block, geo);
3153	geo->start = get_start_sect(bdev) >> base->block->s2b_shift;
3154	dasd_put_device(base);
3155	return 0;
3156}
3157
3158const struct block_device_operations
3159dasd_device_operations = {
3160	.owner		= THIS_MODULE,
3161	.open		= dasd_open,
3162	.release	= dasd_release,
3163	.ioctl		= dasd_ioctl,
3164	.compat_ioctl	= dasd_ioctl,
3165	.getgeo		= dasd_getgeo,
3166};
3167
3168/*******************************************************************************
3169 * end of block device operations
3170 */
3171
3172static void
3173dasd_exit(void)
3174{
3175#ifdef CONFIG_PROC_FS
3176	dasd_proc_exit();
3177#endif
3178	dasd_eer_exit();
3179        if (dasd_page_cache != NULL) {
3180		kmem_cache_destroy(dasd_page_cache);
3181		dasd_page_cache = NULL;
3182	}
3183	dasd_gendisk_exit();
3184	dasd_devmap_exit();
3185	if (dasd_debug_area != NULL) {
3186		debug_unregister(dasd_debug_area);
3187		dasd_debug_area = NULL;
3188	}
3189	dasd_statistics_removeroot();
3190}
3191
3192/*
3193 * SECTION: common functions for ccw_driver use
3194 */
3195
3196/*
3197 * Is the device read-only?
3198 * Note that this function does not report the setting of the
3199 * readonly device attribute, but how it is configured in z/VM.
3200 */
3201int dasd_device_is_ro(struct dasd_device *device)
3202{
3203	struct ccw_dev_id dev_id;
3204	struct diag210 diag_data;
3205	int rc;
3206
3207	if (!MACHINE_IS_VM)
3208		return 0;
3209	ccw_device_get_id(device->cdev, &dev_id);
3210	memset(&diag_data, 0, sizeof(diag_data));
3211	diag_data.vrdcdvno = dev_id.devno;
3212	diag_data.vrdclen = sizeof(diag_data);
3213	rc = diag210(&diag_data);
3214	if (rc == 0 || rc == 2) {
3215		return diag_data.vrdcvfla & 0x80;
3216	} else {
3217		DBF_EVENT(DBF_WARNING, "diag210 failed for dev=%04x with rc=%d",
3218			  dev_id.devno, rc);
3219		return 0;
3220	}
3221}
3222EXPORT_SYMBOL_GPL(dasd_device_is_ro);
3223
3224static void dasd_generic_auto_online(void *data, async_cookie_t cookie)
3225{
3226	struct ccw_device *cdev = data;
3227	int ret;
3228
3229	ret = ccw_device_set_online(cdev);
3230	if (ret)
3231		pr_warn("%s: Setting the DASD online failed with rc=%d\n",
3232			dev_name(&cdev->dev), ret);
3233}
3234
3235/*
3236 * Initial attempt at a probe function. this can be simplified once
3237 * the other detection code is gone.
3238 */
3239int dasd_generic_probe(struct ccw_device *cdev,
3240		       struct dasd_discipline *discipline)
3241{
3242	int ret;
3243
3244	ret = dasd_add_sysfs_files(cdev);
3245	if (ret) {
3246		DBF_EVENT_DEVID(DBF_WARNING, cdev, "%s",
3247				"dasd_generic_probe: could not add "
3248				"sysfs entries");
3249		return ret;
3250	}
3251	cdev->handler = &dasd_int_handler;
3252
3253	/*
3254	 * Automatically online either all dasd devices (dasd_autodetect)
3255	 * or all devices specified with dasd= parameters during
3256	 * initial probe.
3257	 */
3258	if ((dasd_get_feature(cdev, DASD_FEATURE_INITIAL_ONLINE) > 0 ) ||
3259	    (dasd_autodetect && dasd_busid_known(dev_name(&cdev->dev)) != 0))
3260		async_schedule(dasd_generic_auto_online, cdev);
3261	return 0;
3262}
3263EXPORT_SYMBOL_GPL(dasd_generic_probe);
3264
3265/*
3266 * This will one day be called from a global not_oper handler.
3267 * It is also used by driver_unregister during module unload.
3268 */
3269void dasd_generic_remove(struct ccw_device *cdev)
3270{
3271	struct dasd_device *device;
3272	struct dasd_block *block;
3273
3274	cdev->handler = NULL;
3275
3276	device = dasd_device_from_cdev(cdev);
3277	if (IS_ERR(device)) {
3278		dasd_remove_sysfs_files(cdev);
3279		return;
3280	}
3281	if (test_and_set_bit(DASD_FLAG_OFFLINE, &device->flags) &&
3282	    !test_bit(DASD_FLAG_SAFE_OFFLINE_RUNNING, &device->flags)) {
3283		/* Already doing offline processing */
3284		dasd_put_device(device);
3285		dasd_remove_sysfs_files(cdev);
3286		return;
3287	}
3288	/*
3289	 * This device is removed unconditionally. Set offline
3290	 * flag to prevent dasd_open from opening it while it is
3291	 * no quite down yet.
3292	 */
3293	dasd_set_target_state(device, DASD_STATE_NEW);
3294	/* dasd_delete_device destroys the device reference. */
3295	block = device->block;
3296	dasd_delete_device(device);
3297	/*
3298	 * life cycle of block is bound to device, so delete it after
3299	 * device was safely removed
3300	 */
3301	if (block)
3302		dasd_free_block(block);
3303
3304	dasd_remove_sysfs_files(cdev);
3305}
3306EXPORT_SYMBOL_GPL(dasd_generic_remove);
3307
3308/*
3309 * Activate a device. This is called from dasd_{eckd,fba}_probe() when either
3310 * the device is detected for the first time and is supposed to be used
3311 * or the user has started activation through sysfs.
3312 */
3313int dasd_generic_set_online(struct ccw_device *cdev,
3314			    struct dasd_discipline *base_discipline)
3315{
3316	struct dasd_discipline *discipline;
3317	struct dasd_device *device;
3318	int rc;
3319
3320	/* first online clears initial online feature flag */
3321	dasd_set_feature(cdev, DASD_FEATURE_INITIAL_ONLINE, 0);
3322	device = dasd_create_device(cdev);
3323	if (IS_ERR(device))
3324		return PTR_ERR(device);
3325
3326	discipline = base_discipline;
3327	if (device->features & DASD_FEATURE_USEDIAG) {
3328	  	if (!dasd_diag_discipline_pointer) {
3329			/* Try to load the required module. */
3330			rc = request_module(DASD_DIAG_MOD);
3331			if (rc) {
3332				pr_warn("%s Setting the DASD online failed "
3333					"because the required module %s "
3334					"could not be loaded (rc=%d)\n",
3335					dev_name(&cdev->dev), DASD_DIAG_MOD,
3336					rc);
3337				dasd_delete_device(device);
3338				return -ENODEV;
3339			}
3340		}
3341		/* Module init could have failed, so check again here after
3342		 * request_module(). */
3343		if (!dasd_diag_discipline_pointer) {
3344			pr_warn("%s Setting the DASD online failed because of missing DIAG discipline\n",
3345				dev_name(&cdev->dev));
3346			dasd_delete_device(device);
3347			return -ENODEV;
3348		}
3349		discipline = dasd_diag_discipline_pointer;
3350	}
3351	if (!try_module_get(base_discipline->owner)) {
3352		dasd_delete_device(device);
3353		return -EINVAL;
3354	}
3355	if (!try_module_get(discipline->owner)) {
3356		module_put(base_discipline->owner);
3357		dasd_delete_device(device);
3358		return -EINVAL;
3359	}
3360	device->base_discipline = base_discipline;
3361	device->discipline = discipline;
3362
3363	/* check_device will allocate block device if necessary */
3364	rc = discipline->check_device(device);
3365	if (rc) {
3366		pr_warn("%s Setting the DASD online with discipline %s failed with rc=%i\n",
3367			dev_name(&cdev->dev), discipline->name, rc);
3368		module_put(discipline->owner);
3369		module_put(base_discipline->owner);
3370		dasd_delete_device(device);
3371		return rc;
3372	}
3373
3374	dasd_set_target_state(device, DASD_STATE_ONLINE);
3375	if (device->state <= DASD_STATE_KNOWN) {
3376		pr_warn("%s Setting the DASD online failed because of a missing discipline\n",
3377			dev_name(&cdev->dev));
3378		rc = -ENODEV;
3379		dasd_set_target_state(device, DASD_STATE_NEW);
3380		if (device->block)
3381			dasd_free_block(device->block);
3382		dasd_delete_device(device);
3383	} else
3384		pr_debug("dasd_generic device %s found\n",
3385				dev_name(&cdev->dev));
3386
3387	wait_event(dasd_init_waitq, _wait_for_device(device));
3388
3389	dasd_put_device(device);
3390	return rc;
3391}
3392EXPORT_SYMBOL_GPL(dasd_generic_set_online);
3393
3394int dasd_generic_set_offline(struct ccw_device *cdev)
3395{
3396	struct dasd_device *device;
3397	struct dasd_block *block;
3398	int max_count, open_count, rc;
3399
3400	rc = 0;
3401	device = dasd_device_from_cdev(cdev);
3402	if (IS_ERR(device))
3403		return PTR_ERR(device);
3404
3405	/*
3406	 * We must make sure that this device is currently not in use.
3407	 * The open_count is increased for every opener, that includes
3408	 * the blkdev_get in dasd_scan_partitions. We are only interested
3409	 * in the other openers.
3410	 */
3411	if (device->block) {
3412		max_count = device->block->bdev ? 0 : -1;
3413		open_count = atomic_read(&device->block->open_count);
3414		if (open_count > max_count) {
3415			if (open_count > 0)
3416				pr_warn("%s: The DASD cannot be set offline with open count %i\n",
3417					dev_name(&cdev->dev), open_count);
3418			else
3419				pr_warn("%s: The DASD cannot be set offline while it is in use\n",
3420					dev_name(&cdev->dev));
3421			clear_bit(DASD_FLAG_OFFLINE, &device->flags);
3422			dasd_put_device(device);
3423			return -EBUSY;
3424		}
3425	}
3426
3427	if (test_bit(DASD_FLAG_SAFE_OFFLINE_RUNNING, &device->flags)) {
3428		/*
3429		 * safe offline already running
3430		 * could only be called by normal offline so safe_offline flag
3431		 * needs to be removed to run normal offline and kill all I/O
3432		 */
3433		if (test_and_set_bit(DASD_FLAG_OFFLINE, &device->flags)) {
3434			/* Already doing normal offline processing */
3435			dasd_put_device(device);
3436			return -EBUSY;
3437		} else
3438			clear_bit(DASD_FLAG_SAFE_OFFLINE, &device->flags);
3439
3440	} else
3441		if (test_bit(DASD_FLAG_OFFLINE, &device->flags)) {
3442			/* Already doing offline processing */
3443			dasd_put_device(device);
3444			return -EBUSY;
3445		}
3446
3447	/*
3448	 * if safe_offline called set safe_offline_running flag and
3449	 * clear safe_offline so that a call to normal offline
3450	 * can overrun safe_offline processing
3451	 */
3452	if (test_and_clear_bit(DASD_FLAG_SAFE_OFFLINE, &device->flags) &&
3453	    !test_and_set_bit(DASD_FLAG_SAFE_OFFLINE_RUNNING, &device->flags)) {
3454		/*
3455		 * If we want to set the device safe offline all IO operations
3456		 * should be finished before continuing the offline process
3457		 * so sync bdev first and then wait for our queues to become
3458		 * empty
3459		 */
3460		/* sync blockdev and partitions */
3461		rc = fsync_bdev(device->block->bdev);
3462		if (rc != 0)
3463			goto interrupted;
3464
3465		/* schedule device tasklet and wait for completion */
3466		dasd_schedule_device_bh(device);
3467		rc = wait_event_interruptible(shutdown_waitq,
3468					      _wait_for_empty_queues(device));
3469		if (rc != 0)
3470			goto interrupted;
3471	}
3472
3473	set_bit(DASD_FLAG_OFFLINE, &device->flags);
3474	dasd_set_target_state(device, DASD_STATE_NEW);
3475	/* dasd_delete_device destroys the device reference. */
3476	block = device->block;
3477	dasd_delete_device(device);
3478	/*
3479	 * life cycle of block is bound to device, so delete it after
3480	 * device was safely removed
3481	 */
3482	if (block)
3483		dasd_free_block(block);
3484	return 0;
3485
3486interrupted:
3487	/* interrupted by signal */
3488	clear_bit(DASD_FLAG_SAFE_OFFLINE, &device->flags);
3489	clear_bit(DASD_FLAG_SAFE_OFFLINE_RUNNING, &device->flags);
3490	clear_bit(DASD_FLAG_OFFLINE, &device->flags);
3491	dasd_put_device(device);
3492	return rc;
3493}
3494EXPORT_SYMBOL_GPL(dasd_generic_set_offline);
3495
3496int dasd_generic_last_path_gone(struct dasd_device *device)
3497{
3498	struct dasd_ccw_req *cqr;
3499
3500	dev_warn(&device->cdev->dev, "No operational channel path is left "
3501		 "for the device\n");
3502	DBF_DEV_EVENT(DBF_WARNING, device, "%s", "last path gone");
3503	/* First of all call extended error reporting. */
3504	dasd_eer_write(device, NULL, DASD_EER_NOPATH);
3505
3506	if (device->state < DASD_STATE_BASIC)
3507		return 0;
3508	/* Device is active. We want to keep it. */
3509	list_for_each_entry(cqr, &device->ccw_queue, devlist)
3510		if ((cqr->status == DASD_CQR_IN_IO) ||
3511		    (cqr->status == DASD_CQR_CLEAR_PENDING)) {
3512			cqr->status = DASD_CQR_QUEUED;
3513			cqr->retries++;
3514		}
3515	dasd_device_set_stop_bits(device, DASD_STOPPED_DC_WAIT);
3516	dasd_device_clear_timer(device);
3517	dasd_schedule_device_bh(device);
3518	return 1;
3519}
3520EXPORT_SYMBOL_GPL(dasd_generic_last_path_gone);
3521
3522int dasd_generic_path_operational(struct dasd_device *device)
3523{
3524	dev_info(&device->cdev->dev, "A channel path to the device has become "
3525		 "operational\n");
3526	DBF_DEV_EVENT(DBF_WARNING, device, "%s", "path operational");
3527	dasd_device_remove_stop_bits(device, DASD_STOPPED_DC_WAIT);
3528	if (device->stopped & DASD_UNRESUMED_PM) {
3529		dasd_device_remove_stop_bits(device, DASD_UNRESUMED_PM);
3530		dasd_restore_device(device);
3531		return 1;
3532	}
3533	dasd_schedule_device_bh(device);
3534	if (device->block)
3535		dasd_schedule_block_bh(device->block);
3536
3537	if (!device->stopped)
3538		wake_up(&generic_waitq);
3539
3540	return 1;
3541}
3542EXPORT_SYMBOL_GPL(dasd_generic_path_operational);
3543
3544int dasd_generic_notify(struct ccw_device *cdev, int event)
3545{
3546	struct dasd_device *device;
3547	int ret;
3548
3549	device = dasd_device_from_cdev_locked(cdev);
3550	if (IS_ERR(device))
3551		return 0;
3552	ret = 0;
3553	switch (event) {
3554	case CIO_GONE:
3555	case CIO_BOXED:
3556	case CIO_NO_PATH:
3557		device->path_data.opm = 0;
3558		device->path_data.ppm = 0;
3559		device->path_data.npm = 0;
3560		ret = dasd_generic_last_path_gone(device);
3561		break;
3562	case CIO_OPER:
3563		ret = 1;
3564		if (device->path_data.opm)
3565			ret = dasd_generic_path_operational(device);
3566		break;
3567	}
3568	dasd_put_device(device);
3569	return ret;
3570}
3571EXPORT_SYMBOL_GPL(dasd_generic_notify);
3572
3573void dasd_generic_path_event(struct ccw_device *cdev, int *path_event)
3574{
3575	int chp;
3576	__u8 oldopm, eventlpm;
3577	struct dasd_device *device;
3578
3579	device = dasd_device_from_cdev_locked(cdev);
3580	if (IS_ERR(device))
3581		return;
3582	for (chp = 0; chp < 8; chp++) {
3583		eventlpm = 0x80 >> chp;
3584		if (path_event[chp] & PE_PATH_GONE) {
3585			oldopm = device->path_data.opm;
3586			device->path_data.opm &= ~eventlpm;
3587			device->path_data.ppm &= ~eventlpm;
3588			device->path_data.npm &= ~eventlpm;
3589			if (oldopm && !device->path_data.opm) {
3590				dev_warn(&device->cdev->dev,
3591					 "No verified channel paths remain "
3592					 "for the device\n");
3593				DBF_DEV_EVENT(DBF_WARNING, device,
3594					      "%s", "last verified path gone");
3595				dasd_eer_write(device, NULL, DASD_EER_NOPATH);
3596				dasd_device_set_stop_bits(device,
3597							  DASD_STOPPED_DC_WAIT);
3598			}
3599		}
3600		if (path_event[chp] & PE_PATH_AVAILABLE) {
3601			device->path_data.opm &= ~eventlpm;
3602			device->path_data.ppm &= ~eventlpm;
3603			device->path_data.npm &= ~eventlpm;
3604			device->path_data.tbvpm |= eventlpm;
3605			dasd_schedule_device_bh(device);
3606		}
3607		if (path_event[chp] & PE_PATHGROUP_ESTABLISHED) {
3608			if (!(device->path_data.opm & eventlpm) &&
3609			    !(device->path_data.tbvpm & eventlpm)) {
3610				/*
3611				 * we can not establish a pathgroup on an
3612				 * unavailable path, so trigger a path
3613				 * verification first
3614				 */
3615				device->path_data.tbvpm |= eventlpm;
3616				dasd_schedule_device_bh(device);
3617			}
3618			DBF_DEV_EVENT(DBF_WARNING, device, "%s",
3619				      "Pathgroup re-established\n");
3620			if (device->discipline->kick_validate)
3621				device->discipline->kick_validate(device);
3622		}
3623	}
3624	dasd_put_device(device);
3625}
3626EXPORT_SYMBOL_GPL(dasd_generic_path_event);
3627
3628int dasd_generic_verify_path(struct dasd_device *device, __u8 lpm)
3629{
3630	if (!device->path_data.opm && lpm) {
3631		device->path_data.opm = lpm;
3632		dasd_generic_path_operational(device);
3633	} else
3634		device->path_data.opm |= lpm;
3635	return 0;
3636}
3637EXPORT_SYMBOL_GPL(dasd_generic_verify_path);
3638
3639
3640int dasd_generic_pm_freeze(struct ccw_device *cdev)
3641{
3642	struct dasd_device *device = dasd_device_from_cdev(cdev);
3643	struct list_head freeze_queue;
3644	struct dasd_ccw_req *cqr, *n;
3645	struct dasd_ccw_req *refers;
3646	int rc;
3647
3648	if (IS_ERR(device))
3649		return PTR_ERR(device);
3650
3651	/* mark device as suspended */
3652	set_bit(DASD_FLAG_SUSPENDED, &device->flags);
3653
3654	if (device->discipline->freeze)
3655		rc = device->discipline->freeze(device);
3656
3657	/* disallow new I/O  */
3658	dasd_device_set_stop_bits(device, DASD_STOPPED_PM);
3659
3660	/* clear active requests and requeue them to block layer if possible */
3661	INIT_LIST_HEAD(&freeze_queue);
3662	spin_lock_irq(get_ccwdev_lock(cdev));
3663	rc = 0;
3664	list_for_each_entry_safe(cqr, n, &device->ccw_queue, devlist) {
3665		/* Check status and move request to flush_queue */
3666		if (cqr->status == DASD_CQR_IN_IO) {
3667			rc = device->discipline->term_IO(cqr);
3668			if (rc) {
3669				/* unable to terminate requeust */
3670				dev_err(&device->cdev->dev,
3671					"Unable to terminate request %p "
3672					"on suspend\n", cqr);
3673				spin_unlock_irq(get_ccwdev_lock(cdev));
3674				dasd_put_device(device);
3675				return rc;
3676			}
3677		}
3678		list_move_tail(&cqr->devlist, &freeze_queue);
3679	}
3680	spin_unlock_irq(get_ccwdev_lock(cdev));
3681
3682	list_for_each_entry_safe(cqr, n, &freeze_queue, devlist) {
3683		wait_event(dasd_flush_wq,
3684			   (cqr->status != DASD_CQR_CLEAR_PENDING));
3685		if (cqr->status == DASD_CQR_CLEARED)
3686			cqr->status = DASD_CQR_QUEUED;
3687
3688		/* requeue requests to blocklayer will only work for
3689		   block device requests */
3690		if (_dasd_requeue_request(cqr))
3691			continue;
3692
3693		/* remove requests from device and block queue */
3694		list_del_init(&cqr->devlist);
3695		while (cqr->refers != NULL) {
3696			refers = cqr->refers;
3697			/* remove the request from the block queue */
3698			list_del(&cqr->blocklist);
3699			/* free the finished erp request */
3700			dasd_free_erp_request(cqr, cqr->memdev);
3701			cqr = refers;
3702		}
3703		if (cqr->block)
3704			list_del_init(&cqr->blocklist);
3705		cqr->block->base->discipline->free_cp(
3706			cqr, (struct request *) cqr->callback_data);
3707	}
3708
3709	/*
3710	 * if requests remain then they are internal request
3711	 * and go back to the device queue
3712	 */
3713	if (!list_empty(&freeze_queue)) {
3714		/* move freeze_queue to start of the ccw_queue */
3715		spin_lock_irq(get_ccwdev_lock(cdev));
3716		list_splice_tail(&freeze_queue, &device->ccw_queue);
3717		spin_unlock_irq(get_ccwdev_lock(cdev));
3718	}
3719	dasd_put_device(device);
3720	return rc;
3721}
3722EXPORT_SYMBOL_GPL(dasd_generic_pm_freeze);
3723
3724int dasd_generic_restore_device(struct ccw_device *cdev)
3725{
3726	struct dasd_device *device = dasd_device_from_cdev(cdev);
3727	int rc = 0;
3728
3729	if (IS_ERR(device))
3730		return PTR_ERR(device);
3731
3732	/* allow new IO again */
3733	dasd_device_remove_stop_bits(device,
3734				     (DASD_STOPPED_PM | DASD_UNRESUMED_PM));
3735
3736	dasd_schedule_device_bh(device);
3737
3738	/*
3739	 * call discipline restore function
3740	 * if device is stopped do nothing e.g. for disconnected devices
3741	 */
3742	if (device->discipline->restore && !(device->stopped))
3743		rc = device->discipline->restore(device);
3744	if (rc || device->stopped)
3745		/*
3746		 * if the resume failed for the DASD we put it in
3747		 * an UNRESUMED stop state
3748		 */
3749		device->stopped |= DASD_UNRESUMED_PM;
3750
3751	if (device->block)
3752		dasd_schedule_block_bh(device->block);
3753
3754	clear_bit(DASD_FLAG_SUSPENDED, &device->flags);
3755	dasd_put_device(device);
3756	return 0;
3757}
3758EXPORT_SYMBOL_GPL(dasd_generic_restore_device);
3759
3760static struct dasd_ccw_req *dasd_generic_build_rdc(struct dasd_device *device,
3761						   void *rdc_buffer,
3762						   int rdc_buffer_size,
3763						   int magic)
3764{
3765	struct dasd_ccw_req *cqr;
3766	struct ccw1 *ccw;
3767	unsigned long *idaw;
3768
3769	cqr = dasd_smalloc_request(magic, 1 /* RDC */, rdc_buffer_size, device);
3770
3771	if (IS_ERR(cqr)) {
3772		/* internal error 13 - Allocating the RDC request failed*/
3773		dev_err(&device->cdev->dev,
3774			 "An error occurred in the DASD device driver, "
3775			 "reason=%s\n", "13");
3776		return cqr;
3777	}
3778
3779	ccw = cqr->cpaddr;
3780	ccw->cmd_code = CCW_CMD_RDC;
3781	if (idal_is_needed(rdc_buffer, rdc_buffer_size)) {
3782		idaw = (unsigned long *) (cqr->data);
3783		ccw->cda = (__u32)(addr_t) idaw;
3784		ccw->flags = CCW_FLAG_IDA;
3785		idaw = idal_create_words(idaw, rdc_buffer, rdc_buffer_size);
3786	} else {
3787		ccw->cda = (__u32)(addr_t) rdc_buffer;
3788		ccw->flags = 0;
3789	}
3790
3791	ccw->count = rdc_buffer_size;
3792	cqr->startdev = device;
3793	cqr->memdev = device;
3794	cqr->expires = 10*HZ;
3795	cqr->retries = 256;
3796	cqr->buildclk = get_tod_clock();
3797	cqr->status = DASD_CQR_FILLED;
3798	return cqr;
3799}
3800
3801
3802int dasd_generic_read_dev_chars(struct dasd_device *device, int magic,
3803				void *rdc_buffer, int rdc_buffer_size)
3804{
3805	int ret;
3806	struct dasd_ccw_req *cqr;
3807
3808	cqr = dasd_generic_build_rdc(device, rdc_buffer, rdc_buffer_size,
3809				     magic);
3810	if (IS_ERR(cqr))
3811		return PTR_ERR(cqr);
3812
3813	ret = dasd_sleep_on(cqr);
3814	dasd_sfree_request(cqr, cqr->memdev);
3815	return ret;
3816}
3817EXPORT_SYMBOL_GPL(dasd_generic_read_dev_chars);
3818
3819/*
3820 *   In command mode and transport mode we need to look for sense
3821 *   data in different places. The sense data itself is allways
3822 *   an array of 32 bytes, so we can unify the sense data access
3823 *   for both modes.
3824 */
3825char *dasd_get_sense(struct irb *irb)
3826{
3827	struct tsb *tsb = NULL;
3828	char *sense = NULL;
3829
3830	if (scsw_is_tm(&irb->scsw) && (irb->scsw.tm.fcxs == 0x01)) {
3831		if (irb->scsw.tm.tcw)
3832			tsb = tcw_get_tsb((struct tcw *)(unsigned long)
3833					  irb->scsw.tm.tcw);
3834		if (tsb && tsb->length == 64 && tsb->flags)
3835			switch (tsb->flags & 0x07) {
3836			case 1:	/* tsa_iostat */
3837				sense = tsb->tsa.iostat.sense;
3838				break;
3839			case 2: /* tsa_ddpc */
3840				sense = tsb->tsa.ddpc.sense;
3841				break;
3842			default:
3843				/* currently we don't use interrogate data */
3844				break;
3845			}
3846	} else if (irb->esw.esw0.erw.cons) {
3847		sense = irb->ecw;
3848	}
3849	return sense;
3850}
3851EXPORT_SYMBOL_GPL(dasd_get_sense);
3852
3853void dasd_generic_shutdown(struct ccw_device *cdev)
3854{
3855	struct dasd_device *device;
3856
3857	device = dasd_device_from_cdev(cdev);
3858	if (IS_ERR(device))
3859		return;
3860
3861	if (device->block)
3862		dasd_schedule_block_bh(device->block);
3863
3864	dasd_schedule_device_bh(device);
3865
3866	wait_event(shutdown_waitq, _wait_for_empty_queues(device));
3867}
3868EXPORT_SYMBOL_GPL(dasd_generic_shutdown);
3869
3870static int __init dasd_init(void)
3871{
3872	int rc;
3873
3874	init_waitqueue_head(&dasd_init_waitq);
3875	init_waitqueue_head(&dasd_flush_wq);
3876	init_waitqueue_head(&generic_waitq);
3877	init_waitqueue_head(&shutdown_waitq);
3878
3879	/* register 'common' DASD debug area, used for all DBF_XXX calls */
3880	dasd_debug_area = debug_register("dasd", 1, 1, 8 * sizeof(long));
3881	if (dasd_debug_area == NULL) {
3882		rc = -ENOMEM;
3883		goto failed;
3884	}
3885	debug_register_view(dasd_debug_area, &debug_sprintf_view);
3886	debug_set_level(dasd_debug_area, DBF_WARNING);
3887
3888	DBF_EVENT(DBF_EMERG, "%s", "debug area created");
3889
3890	dasd_diag_discipline_pointer = NULL;
3891
3892	dasd_statistics_createroot();
3893
3894	rc = dasd_devmap_init();
3895	if (rc)
3896		goto failed;
3897	rc = dasd_gendisk_init();
3898	if (rc)
3899		goto failed;
3900	rc = dasd_parse();
3901	if (rc)
3902		goto failed;
3903	rc = dasd_eer_init();
3904	if (rc)
3905		goto failed;
3906#ifdef CONFIG_PROC_FS
3907	rc = dasd_proc_init();
3908	if (rc)
3909		goto failed;
3910#endif
3911
3912	return 0;
3913failed:
3914	pr_info("The DASD device driver could not be initialized\n");
3915	dasd_exit();
3916	return rc;
3917}
3918
3919module_init(dasd_init);
3920module_exit(dasd_exit);
3921