xref: /openbmc/linux/drivers/s390/block/dasd.c (revision e4258d55)
1 /*
2  * File...........: linux/drivers/s390/block/dasd.c
3  * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com>
4  *		    Horst Hummel <Horst.Hummel@de.ibm.com>
5  *		    Carsten Otte <Cotte@de.ibm.com>
6  *		    Martin Schwidefsky <schwidefsky@de.ibm.com>
7  * Bugreports.to..: <Linux390@de.ibm.com>
8  * Copyright IBM Corp. 1999, 2009
9  */
10 
11 #define KMSG_COMPONENT "dasd"
12 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
13 
14 #include <linux/kernel_stat.h>
15 #include <linux/kmod.h>
16 #include <linux/init.h>
17 #include <linux/interrupt.h>
18 #include <linux/ctype.h>
19 #include <linux/major.h>
20 #include <linux/slab.h>
21 #include <linux/buffer_head.h>
22 #include <linux/hdreg.h>
23 #include <linux/async.h>
24 #include <linux/mutex.h>
25 #include <linux/debugfs.h>
26 #include <linux/seq_file.h>
27 #include <linux/vmalloc.h>
28 
29 #include <asm/ccwdev.h>
30 #include <asm/ebcdic.h>
31 #include <asm/idals.h>
32 #include <asm/itcw.h>
33 #include <asm/diag.h>
34 
35 /* This is ugly... */
36 #define PRINTK_HEADER "dasd:"
37 
38 #include "dasd_int.h"
39 /*
40  * SECTION: Constant definitions to be used within this file
41  */
42 #define DASD_CHANQ_MAX_SIZE 4
43 
44 #define DASD_SLEEPON_START_TAG	(void *) 1
45 #define DASD_SLEEPON_END_TAG	(void *) 2
46 
47 /*
48  * SECTION: exported variables of dasd.c
49  */
50 debug_info_t *dasd_debug_area;
51 static struct dentry *dasd_debugfs_root_entry;
52 struct dasd_discipline *dasd_diag_discipline_pointer;
53 void dasd_int_handler(struct ccw_device *, unsigned long, struct irb *);
54 
55 MODULE_AUTHOR("Holger Smolinski <Holger.Smolinski@de.ibm.com>");
56 MODULE_DESCRIPTION("Linux on S/390 DASD device driver,"
57 		   " Copyright 2000 IBM Corporation");
58 MODULE_SUPPORTED_DEVICE("dasd");
59 MODULE_LICENSE("GPL");
60 
61 /*
62  * SECTION: prototypes for static functions of dasd.c
63  */
64 static int  dasd_alloc_queue(struct dasd_block *);
65 static void dasd_setup_queue(struct dasd_block *);
66 static void dasd_free_queue(struct dasd_block *);
67 static void dasd_flush_request_queue(struct dasd_block *);
68 static int dasd_flush_block_queue(struct dasd_block *);
69 static void dasd_device_tasklet(struct dasd_device *);
70 static void dasd_block_tasklet(struct dasd_block *);
71 static void do_kick_device(struct work_struct *);
72 static void do_restore_device(struct work_struct *);
73 static void do_reload_device(struct work_struct *);
74 static void dasd_return_cqr_cb(struct dasd_ccw_req *, void *);
75 static void dasd_device_timeout(unsigned long);
76 static void dasd_block_timeout(unsigned long);
77 static void __dasd_process_erp(struct dasd_device *, struct dasd_ccw_req *);
78 static void dasd_profile_init(struct dasd_profile *, struct dentry *);
79 static void dasd_profile_exit(struct dasd_profile *);
80 
81 /*
82  * SECTION: Operations on the device structure.
83  */
84 static wait_queue_head_t dasd_init_waitq;
85 static wait_queue_head_t dasd_flush_wq;
86 static wait_queue_head_t generic_waitq;
87 
88 /*
89  * Allocate memory for a new device structure.
90  */
91 struct dasd_device *dasd_alloc_device(void)
92 {
93 	struct dasd_device *device;
94 
95 	device = kzalloc(sizeof(struct dasd_device), GFP_ATOMIC);
96 	if (!device)
97 		return ERR_PTR(-ENOMEM);
98 
99 	/* Get two pages for normal block device operations. */
100 	device->ccw_mem = (void *) __get_free_pages(GFP_ATOMIC | GFP_DMA, 1);
101 	if (!device->ccw_mem) {
102 		kfree(device);
103 		return ERR_PTR(-ENOMEM);
104 	}
105 	/* Get one page for error recovery. */
106 	device->erp_mem = (void *) get_zeroed_page(GFP_ATOMIC | GFP_DMA);
107 	if (!device->erp_mem) {
108 		free_pages((unsigned long) device->ccw_mem, 1);
109 		kfree(device);
110 		return ERR_PTR(-ENOMEM);
111 	}
112 
113 	dasd_init_chunklist(&device->ccw_chunks, device->ccw_mem, PAGE_SIZE*2);
114 	dasd_init_chunklist(&device->erp_chunks, device->erp_mem, PAGE_SIZE);
115 	spin_lock_init(&device->mem_lock);
116 	atomic_set(&device->tasklet_scheduled, 0);
117 	tasklet_init(&device->tasklet,
118 		     (void (*)(unsigned long)) dasd_device_tasklet,
119 		     (unsigned long) device);
120 	INIT_LIST_HEAD(&device->ccw_queue);
121 	init_timer(&device->timer);
122 	device->timer.function = dasd_device_timeout;
123 	device->timer.data = (unsigned long) device;
124 	INIT_WORK(&device->kick_work, do_kick_device);
125 	INIT_WORK(&device->restore_device, do_restore_device);
126 	INIT_WORK(&device->reload_device, do_reload_device);
127 	device->state = DASD_STATE_NEW;
128 	device->target = DASD_STATE_NEW;
129 	mutex_init(&device->state_mutex);
130 	spin_lock_init(&device->profile.lock);
131 	return device;
132 }
133 
134 /*
135  * Free memory of a device structure.
136  */
137 void dasd_free_device(struct dasd_device *device)
138 {
139 	kfree(device->private);
140 	free_page((unsigned long) device->erp_mem);
141 	free_pages((unsigned long) device->ccw_mem, 1);
142 	kfree(device);
143 }
144 
145 /*
146  * Allocate memory for a new device structure.
147  */
148 struct dasd_block *dasd_alloc_block(void)
149 {
150 	struct dasd_block *block;
151 
152 	block = kzalloc(sizeof(*block), GFP_ATOMIC);
153 	if (!block)
154 		return ERR_PTR(-ENOMEM);
155 	/* open_count = 0 means device online but not in use */
156 	atomic_set(&block->open_count, -1);
157 
158 	spin_lock_init(&block->request_queue_lock);
159 	atomic_set(&block->tasklet_scheduled, 0);
160 	tasklet_init(&block->tasklet,
161 		     (void (*)(unsigned long)) dasd_block_tasklet,
162 		     (unsigned long) block);
163 	INIT_LIST_HEAD(&block->ccw_queue);
164 	spin_lock_init(&block->queue_lock);
165 	init_timer(&block->timer);
166 	block->timer.function = dasd_block_timeout;
167 	block->timer.data = (unsigned long) block;
168 	spin_lock_init(&block->profile.lock);
169 
170 	return block;
171 }
172 
173 /*
174  * Free memory of a device structure.
175  */
176 void dasd_free_block(struct dasd_block *block)
177 {
178 	kfree(block);
179 }
180 
181 /*
182  * Make a new device known to the system.
183  */
184 static int dasd_state_new_to_known(struct dasd_device *device)
185 {
186 	int rc;
187 
188 	/*
189 	 * As long as the device is not in state DASD_STATE_NEW we want to
190 	 * keep the reference count > 0.
191 	 */
192 	dasd_get_device(device);
193 
194 	if (device->block) {
195 		rc = dasd_alloc_queue(device->block);
196 		if (rc) {
197 			dasd_put_device(device);
198 			return rc;
199 		}
200 	}
201 	device->state = DASD_STATE_KNOWN;
202 	return 0;
203 }
204 
205 /*
206  * Let the system forget about a device.
207  */
208 static int dasd_state_known_to_new(struct dasd_device *device)
209 {
210 	/* Disable extended error reporting for this device. */
211 	dasd_eer_disable(device);
212 	/* Forget the discipline information. */
213 	if (device->discipline) {
214 		if (device->discipline->uncheck_device)
215 			device->discipline->uncheck_device(device);
216 		module_put(device->discipline->owner);
217 	}
218 	device->discipline = NULL;
219 	if (device->base_discipline)
220 		module_put(device->base_discipline->owner);
221 	device->base_discipline = NULL;
222 	device->state = DASD_STATE_NEW;
223 
224 	if (device->block)
225 		dasd_free_queue(device->block);
226 
227 	/* Give up reference we took in dasd_state_new_to_known. */
228 	dasd_put_device(device);
229 	return 0;
230 }
231 
232 static struct dentry *dasd_debugfs_setup(const char *name,
233 					 struct dentry *base_dentry)
234 {
235 	struct dentry *pde;
236 
237 	if (!base_dentry)
238 		return NULL;
239 	pde = debugfs_create_dir(name, base_dentry);
240 	if (!pde || IS_ERR(pde))
241 		return NULL;
242 	return pde;
243 }
244 
245 /*
246  * Request the irq line for the device.
247  */
248 static int dasd_state_known_to_basic(struct dasd_device *device)
249 {
250 	struct dasd_block *block = device->block;
251 	int rc;
252 
253 	/* Allocate and register gendisk structure. */
254 	if (block) {
255 		rc = dasd_gendisk_alloc(block);
256 		if (rc)
257 			return rc;
258 		block->debugfs_dentry =
259 			dasd_debugfs_setup(block->gdp->disk_name,
260 					   dasd_debugfs_root_entry);
261 		dasd_profile_init(&block->profile, block->debugfs_dentry);
262 		if (dasd_global_profile_level == DASD_PROFILE_ON)
263 			dasd_profile_on(&device->block->profile);
264 	}
265 	device->debugfs_dentry =
266 		dasd_debugfs_setup(dev_name(&device->cdev->dev),
267 				   dasd_debugfs_root_entry);
268 	dasd_profile_init(&device->profile, device->debugfs_dentry);
269 
270 	/* register 'device' debug area, used for all DBF_DEV_XXX calls */
271 	device->debug_area = debug_register(dev_name(&device->cdev->dev), 4, 1,
272 					    8 * sizeof(long));
273 	debug_register_view(device->debug_area, &debug_sprintf_view);
274 	debug_set_level(device->debug_area, DBF_WARNING);
275 	DBF_DEV_EVENT(DBF_EMERG, device, "%s", "debug area created");
276 
277 	device->state = DASD_STATE_BASIC;
278 	return 0;
279 }
280 
281 /*
282  * Release the irq line for the device. Terminate any running i/o.
283  */
284 static int dasd_state_basic_to_known(struct dasd_device *device)
285 {
286 	int rc;
287 	if (device->block) {
288 		dasd_profile_exit(&device->block->profile);
289 		if (device->block->debugfs_dentry)
290 			debugfs_remove(device->block->debugfs_dentry);
291 		dasd_gendisk_free(device->block);
292 		dasd_block_clear_timer(device->block);
293 	}
294 	rc = dasd_flush_device_queue(device);
295 	if (rc)
296 		return rc;
297 	dasd_device_clear_timer(device);
298 	dasd_profile_exit(&device->profile);
299 	if (device->debugfs_dentry)
300 		debugfs_remove(device->debugfs_dentry);
301 
302 	DBF_DEV_EVENT(DBF_EMERG, device, "%p debug area deleted", device);
303 	if (device->debug_area != NULL) {
304 		debug_unregister(device->debug_area);
305 		device->debug_area = NULL;
306 	}
307 	device->state = DASD_STATE_KNOWN;
308 	return 0;
309 }
310 
311 /*
312  * Do the initial analysis. The do_analysis function may return
313  * -EAGAIN in which case the device keeps the state DASD_STATE_BASIC
314  * until the discipline decides to continue the startup sequence
315  * by calling the function dasd_change_state. The eckd disciplines
316  * uses this to start a ccw that detects the format. The completion
317  * interrupt for this detection ccw uses the kernel event daemon to
318  * trigger the call to dasd_change_state. All this is done in the
319  * discipline code, see dasd_eckd.c.
320  * After the analysis ccw is done (do_analysis returned 0) the block
321  * device is setup.
322  * In case the analysis returns an error, the device setup is stopped
323  * (a fake disk was already added to allow formatting).
324  */
325 static int dasd_state_basic_to_ready(struct dasd_device *device)
326 {
327 	int rc;
328 	struct dasd_block *block;
329 
330 	rc = 0;
331 	block = device->block;
332 	/* make disk known with correct capacity */
333 	if (block) {
334 		if (block->base->discipline->do_analysis != NULL)
335 			rc = block->base->discipline->do_analysis(block);
336 		if (rc) {
337 			if (rc != -EAGAIN)
338 				device->state = DASD_STATE_UNFMT;
339 			return rc;
340 		}
341 		dasd_setup_queue(block);
342 		set_capacity(block->gdp,
343 			     block->blocks << block->s2b_shift);
344 		device->state = DASD_STATE_READY;
345 		rc = dasd_scan_partitions(block);
346 		if (rc)
347 			device->state = DASD_STATE_BASIC;
348 	} else {
349 		device->state = DASD_STATE_READY;
350 	}
351 	return rc;
352 }
353 
354 /*
355  * Remove device from block device layer. Destroy dirty buffers.
356  * Forget format information. Check if the target level is basic
357  * and if it is create fake disk for formatting.
358  */
359 static int dasd_state_ready_to_basic(struct dasd_device *device)
360 {
361 	int rc;
362 
363 	device->state = DASD_STATE_BASIC;
364 	if (device->block) {
365 		struct dasd_block *block = device->block;
366 		rc = dasd_flush_block_queue(block);
367 		if (rc) {
368 			device->state = DASD_STATE_READY;
369 			return rc;
370 		}
371 		dasd_flush_request_queue(block);
372 		dasd_destroy_partitions(block);
373 		block->blocks = 0;
374 		block->bp_block = 0;
375 		block->s2b_shift = 0;
376 	}
377 	return 0;
378 }
379 
380 /*
381  * Back to basic.
382  */
383 static int dasd_state_unfmt_to_basic(struct dasd_device *device)
384 {
385 	device->state = DASD_STATE_BASIC;
386 	return 0;
387 }
388 
389 /*
390  * Make the device online and schedule the bottom half to start
391  * the requeueing of requests from the linux request queue to the
392  * ccw queue.
393  */
394 static int
395 dasd_state_ready_to_online(struct dasd_device * device)
396 {
397 	int rc;
398 	struct gendisk *disk;
399 	struct disk_part_iter piter;
400 	struct hd_struct *part;
401 
402 	if (device->discipline->ready_to_online) {
403 		rc = device->discipline->ready_to_online(device);
404 		if (rc)
405 			return rc;
406 	}
407 	device->state = DASD_STATE_ONLINE;
408 	if (device->block) {
409 		dasd_schedule_block_bh(device->block);
410 		if ((device->features & DASD_FEATURE_USERAW)) {
411 			disk = device->block->gdp;
412 			kobject_uevent(&disk_to_dev(disk)->kobj, KOBJ_CHANGE);
413 			return 0;
414 		}
415 		disk = device->block->bdev->bd_disk;
416 		disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
417 		while ((part = disk_part_iter_next(&piter)))
418 			kobject_uevent(&part_to_dev(part)->kobj, KOBJ_CHANGE);
419 		disk_part_iter_exit(&piter);
420 	}
421 	return 0;
422 }
423 
424 /*
425  * Stop the requeueing of requests again.
426  */
427 static int dasd_state_online_to_ready(struct dasd_device *device)
428 {
429 	int rc;
430 	struct gendisk *disk;
431 	struct disk_part_iter piter;
432 	struct hd_struct *part;
433 
434 	if (device->discipline->online_to_ready) {
435 		rc = device->discipline->online_to_ready(device);
436 		if (rc)
437 			return rc;
438 	}
439 	device->state = DASD_STATE_READY;
440 	if (device->block && !(device->features & DASD_FEATURE_USERAW)) {
441 		disk = device->block->bdev->bd_disk;
442 		disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
443 		while ((part = disk_part_iter_next(&piter)))
444 			kobject_uevent(&part_to_dev(part)->kobj, KOBJ_CHANGE);
445 		disk_part_iter_exit(&piter);
446 	}
447 	return 0;
448 }
449 
450 /*
451  * Device startup state changes.
452  */
453 static int dasd_increase_state(struct dasd_device *device)
454 {
455 	int rc;
456 
457 	rc = 0;
458 	if (device->state == DASD_STATE_NEW &&
459 	    device->target >= DASD_STATE_KNOWN)
460 		rc = dasd_state_new_to_known(device);
461 
462 	if (!rc &&
463 	    device->state == DASD_STATE_KNOWN &&
464 	    device->target >= DASD_STATE_BASIC)
465 		rc = dasd_state_known_to_basic(device);
466 
467 	if (!rc &&
468 	    device->state == DASD_STATE_BASIC &&
469 	    device->target >= DASD_STATE_READY)
470 		rc = dasd_state_basic_to_ready(device);
471 
472 	if (!rc &&
473 	    device->state == DASD_STATE_UNFMT &&
474 	    device->target > DASD_STATE_UNFMT)
475 		rc = -EPERM;
476 
477 	if (!rc &&
478 	    device->state == DASD_STATE_READY &&
479 	    device->target >= DASD_STATE_ONLINE)
480 		rc = dasd_state_ready_to_online(device);
481 
482 	return rc;
483 }
484 
485 /*
486  * Device shutdown state changes.
487  */
488 static int dasd_decrease_state(struct dasd_device *device)
489 {
490 	int rc;
491 
492 	rc = 0;
493 	if (device->state == DASD_STATE_ONLINE &&
494 	    device->target <= DASD_STATE_READY)
495 		rc = dasd_state_online_to_ready(device);
496 
497 	if (!rc &&
498 	    device->state == DASD_STATE_READY &&
499 	    device->target <= DASD_STATE_BASIC)
500 		rc = dasd_state_ready_to_basic(device);
501 
502 	if (!rc &&
503 	    device->state == DASD_STATE_UNFMT &&
504 	    device->target <= DASD_STATE_BASIC)
505 		rc = dasd_state_unfmt_to_basic(device);
506 
507 	if (!rc &&
508 	    device->state == DASD_STATE_BASIC &&
509 	    device->target <= DASD_STATE_KNOWN)
510 		rc = dasd_state_basic_to_known(device);
511 
512 	if (!rc &&
513 	    device->state == DASD_STATE_KNOWN &&
514 	    device->target <= DASD_STATE_NEW)
515 		rc = dasd_state_known_to_new(device);
516 
517 	return rc;
518 }
519 
520 /*
521  * This is the main startup/shutdown routine.
522  */
523 static void dasd_change_state(struct dasd_device *device)
524 {
525 	int rc;
526 
527 	if (device->state == device->target)
528 		/* Already where we want to go today... */
529 		return;
530 	if (device->state < device->target)
531 		rc = dasd_increase_state(device);
532 	else
533 		rc = dasd_decrease_state(device);
534 	if (rc == -EAGAIN)
535 		return;
536 	if (rc)
537 		device->target = device->state;
538 
539 	if (device->state == device->target)
540 		wake_up(&dasd_init_waitq);
541 
542 	/* let user-space know that the device status changed */
543 	kobject_uevent(&device->cdev->dev.kobj, KOBJ_CHANGE);
544 }
545 
546 /*
547  * Kick starter for devices that did not complete the startup/shutdown
548  * procedure or were sleeping because of a pending state.
549  * dasd_kick_device will schedule a call do do_kick_device to the kernel
550  * event daemon.
551  */
552 static void do_kick_device(struct work_struct *work)
553 {
554 	struct dasd_device *device = container_of(work, struct dasd_device, kick_work);
555 	mutex_lock(&device->state_mutex);
556 	dasd_change_state(device);
557 	mutex_unlock(&device->state_mutex);
558 	dasd_schedule_device_bh(device);
559 	dasd_put_device(device);
560 }
561 
562 void dasd_kick_device(struct dasd_device *device)
563 {
564 	dasd_get_device(device);
565 	/* queue call to dasd_kick_device to the kernel event daemon. */
566 	schedule_work(&device->kick_work);
567 }
568 
569 /*
570  * dasd_reload_device will schedule a call do do_reload_device to the kernel
571  * event daemon.
572  */
573 static void do_reload_device(struct work_struct *work)
574 {
575 	struct dasd_device *device = container_of(work, struct dasd_device,
576 						  reload_device);
577 	device->discipline->reload(device);
578 	dasd_put_device(device);
579 }
580 
581 void dasd_reload_device(struct dasd_device *device)
582 {
583 	dasd_get_device(device);
584 	/* queue call to dasd_reload_device to the kernel event daemon. */
585 	schedule_work(&device->reload_device);
586 }
587 EXPORT_SYMBOL(dasd_reload_device);
588 
589 /*
590  * dasd_restore_device will schedule a call do do_restore_device to the kernel
591  * event daemon.
592  */
593 static void do_restore_device(struct work_struct *work)
594 {
595 	struct dasd_device *device = container_of(work, struct dasd_device,
596 						  restore_device);
597 	device->cdev->drv->restore(device->cdev);
598 	dasd_put_device(device);
599 }
600 
601 void dasd_restore_device(struct dasd_device *device)
602 {
603 	dasd_get_device(device);
604 	/* queue call to dasd_restore_device to the kernel event daemon. */
605 	schedule_work(&device->restore_device);
606 }
607 
608 /*
609  * Set the target state for a device and starts the state change.
610  */
611 void dasd_set_target_state(struct dasd_device *device, int target)
612 {
613 	dasd_get_device(device);
614 	mutex_lock(&device->state_mutex);
615 	/* If we are in probeonly mode stop at DASD_STATE_READY. */
616 	if (dasd_probeonly && target > DASD_STATE_READY)
617 		target = DASD_STATE_READY;
618 	if (device->target != target) {
619 		if (device->state == target)
620 			wake_up(&dasd_init_waitq);
621 		device->target = target;
622 	}
623 	if (device->state != device->target)
624 		dasd_change_state(device);
625 	mutex_unlock(&device->state_mutex);
626 	dasd_put_device(device);
627 }
628 
629 /*
630  * Enable devices with device numbers in [from..to].
631  */
632 static inline int _wait_for_device(struct dasd_device *device)
633 {
634 	return (device->state == device->target);
635 }
636 
637 void dasd_enable_device(struct dasd_device *device)
638 {
639 	dasd_set_target_state(device, DASD_STATE_ONLINE);
640 	if (device->state <= DASD_STATE_KNOWN)
641 		/* No discipline for device found. */
642 		dasd_set_target_state(device, DASD_STATE_NEW);
643 	/* Now wait for the devices to come up. */
644 	wait_event(dasd_init_waitq, _wait_for_device(device));
645 }
646 
647 /*
648  * SECTION: device operation (interrupt handler, start i/o, term i/o ...)
649  */
650 
651 unsigned int dasd_global_profile_level = DASD_PROFILE_OFF;
652 
653 #ifdef CONFIG_DASD_PROFILE
654 struct dasd_profile_info dasd_global_profile_data;
655 static struct dentry *dasd_global_profile_dentry;
656 static struct dentry *dasd_debugfs_global_entry;
657 
658 /*
659  * Add profiling information for cqr before execution.
660  */
661 static void dasd_profile_start(struct dasd_block *block,
662 			       struct dasd_ccw_req *cqr,
663 			       struct request *req)
664 {
665 	struct list_head *l;
666 	unsigned int counter;
667 	struct dasd_device *device;
668 
669 	/* count the length of the chanq for statistics */
670 	counter = 0;
671 	if (dasd_global_profile_level || block->profile.data)
672 		list_for_each(l, &block->ccw_queue)
673 			if (++counter >= 31)
674 				break;
675 
676 	if (dasd_global_profile_level) {
677 		dasd_global_profile_data.dasd_io_nr_req[counter]++;
678 		if (rq_data_dir(req) == READ)
679 			dasd_global_profile_data.dasd_read_nr_req[counter]++;
680 	}
681 
682 	spin_lock(&block->profile.lock);
683 	if (block->profile.data)
684 		block->profile.data->dasd_io_nr_req[counter]++;
685 		if (rq_data_dir(req) == READ)
686 			block->profile.data->dasd_read_nr_req[counter]++;
687 	spin_unlock(&block->profile.lock);
688 
689 	/*
690 	 * We count the request for the start device, even though it may run on
691 	 * some other device due to error recovery. This way we make sure that
692 	 * we count each request only once.
693 	 */
694 	device = cqr->startdev;
695 	if (device->profile.data) {
696 		counter = 1; /* request is not yet queued on the start device */
697 		list_for_each(l, &device->ccw_queue)
698 			if (++counter >= 31)
699 				break;
700 	}
701 	spin_lock(&device->profile.lock);
702 	if (device->profile.data) {
703 		device->profile.data->dasd_io_nr_req[counter]++;
704 		if (rq_data_dir(req) == READ)
705 			device->profile.data->dasd_read_nr_req[counter]++;
706 	}
707 	spin_unlock(&device->profile.lock);
708 }
709 
710 /*
711  * Add profiling information for cqr after execution.
712  */
713 
714 #define dasd_profile_counter(value, index)			   \
715 {								   \
716 	for (index = 0; index < 31 && value >> (2+index); index++) \
717 		;						   \
718 }
719 
720 static void dasd_profile_end_add_data(struct dasd_profile_info *data,
721 				      int is_alias,
722 				      int is_tpm,
723 				      int is_read,
724 				      long sectors,
725 				      int sectors_ind,
726 				      int tottime_ind,
727 				      int tottimeps_ind,
728 				      int strtime_ind,
729 				      int irqtime_ind,
730 				      int irqtimeps_ind,
731 				      int endtime_ind)
732 {
733 	/* in case of an overflow, reset the whole profile */
734 	if (data->dasd_io_reqs == UINT_MAX) {
735 			memset(data, 0, sizeof(*data));
736 			getnstimeofday(&data->starttod);
737 	}
738 	data->dasd_io_reqs++;
739 	data->dasd_io_sects += sectors;
740 	if (is_alias)
741 		data->dasd_io_alias++;
742 	if (is_tpm)
743 		data->dasd_io_tpm++;
744 
745 	data->dasd_io_secs[sectors_ind]++;
746 	data->dasd_io_times[tottime_ind]++;
747 	data->dasd_io_timps[tottimeps_ind]++;
748 	data->dasd_io_time1[strtime_ind]++;
749 	data->dasd_io_time2[irqtime_ind]++;
750 	data->dasd_io_time2ps[irqtimeps_ind]++;
751 	data->dasd_io_time3[endtime_ind]++;
752 
753 	if (is_read) {
754 		data->dasd_read_reqs++;
755 		data->dasd_read_sects += sectors;
756 		if (is_alias)
757 			data->dasd_read_alias++;
758 		if (is_tpm)
759 			data->dasd_read_tpm++;
760 		data->dasd_read_secs[sectors_ind]++;
761 		data->dasd_read_times[tottime_ind]++;
762 		data->dasd_read_time1[strtime_ind]++;
763 		data->dasd_read_time2[irqtime_ind]++;
764 		data->dasd_read_time3[endtime_ind]++;
765 	}
766 }
767 
768 static void dasd_profile_end(struct dasd_block *block,
769 			     struct dasd_ccw_req *cqr,
770 			     struct request *req)
771 {
772 	long strtime, irqtime, endtime, tottime;	/* in microseconds */
773 	long tottimeps, sectors;
774 	struct dasd_device *device;
775 	int sectors_ind, tottime_ind, tottimeps_ind, strtime_ind;
776 	int irqtime_ind, irqtimeps_ind, endtime_ind;
777 
778 	device = cqr->startdev;
779 	if (!(dasd_global_profile_level ||
780 	      block->profile.data ||
781 	      device->profile.data))
782 		return;
783 
784 	sectors = blk_rq_sectors(req);
785 	if (!cqr->buildclk || !cqr->startclk ||
786 	    !cqr->stopclk || !cqr->endclk ||
787 	    !sectors)
788 		return;
789 
790 	strtime = ((cqr->startclk - cqr->buildclk) >> 12);
791 	irqtime = ((cqr->stopclk - cqr->startclk) >> 12);
792 	endtime = ((cqr->endclk - cqr->stopclk) >> 12);
793 	tottime = ((cqr->endclk - cqr->buildclk) >> 12);
794 	tottimeps = tottime / sectors;
795 
796 	dasd_profile_counter(sectors, sectors_ind);
797 	dasd_profile_counter(tottime, tottime_ind);
798 	dasd_profile_counter(tottimeps, tottimeps_ind);
799 	dasd_profile_counter(strtime, strtime_ind);
800 	dasd_profile_counter(irqtime, irqtime_ind);
801 	dasd_profile_counter(irqtime / sectors, irqtimeps_ind);
802 	dasd_profile_counter(endtime, endtime_ind);
803 
804 	if (dasd_global_profile_level) {
805 		dasd_profile_end_add_data(&dasd_global_profile_data,
806 					  cqr->startdev != block->base,
807 					  cqr->cpmode == 1,
808 					  rq_data_dir(req) == READ,
809 					  sectors, sectors_ind, tottime_ind,
810 					  tottimeps_ind, strtime_ind,
811 					  irqtime_ind, irqtimeps_ind,
812 					  endtime_ind);
813 	}
814 
815 	spin_lock(&block->profile.lock);
816 	if (block->profile.data)
817 		dasd_profile_end_add_data(block->profile.data,
818 					  cqr->startdev != block->base,
819 					  cqr->cpmode == 1,
820 					  rq_data_dir(req) == READ,
821 					  sectors, sectors_ind, tottime_ind,
822 					  tottimeps_ind, strtime_ind,
823 					  irqtime_ind, irqtimeps_ind,
824 					  endtime_ind);
825 	spin_unlock(&block->profile.lock);
826 
827 	spin_lock(&device->profile.lock);
828 	if (device->profile.data)
829 		dasd_profile_end_add_data(device->profile.data,
830 					  cqr->startdev != block->base,
831 					  cqr->cpmode == 1,
832 					  rq_data_dir(req) == READ,
833 					  sectors, sectors_ind, tottime_ind,
834 					  tottimeps_ind, strtime_ind,
835 					  irqtime_ind, irqtimeps_ind,
836 					  endtime_ind);
837 	spin_unlock(&device->profile.lock);
838 }
839 
840 void dasd_profile_reset(struct dasd_profile *profile)
841 {
842 	struct dasd_profile_info *data;
843 
844 	spin_lock_bh(&profile->lock);
845 	data = profile->data;
846 	if (!data) {
847 		spin_unlock_bh(&profile->lock);
848 		return;
849 	}
850 	memset(data, 0, sizeof(*data));
851 	getnstimeofday(&data->starttod);
852 	spin_unlock_bh(&profile->lock);
853 }
854 
855 void dasd_global_profile_reset(void)
856 {
857 	memset(&dasd_global_profile_data, 0, sizeof(dasd_global_profile_data));
858 	getnstimeofday(&dasd_global_profile_data.starttod);
859 }
860 
861 int dasd_profile_on(struct dasd_profile *profile)
862 {
863 	struct dasd_profile_info *data;
864 
865 	data = kzalloc(sizeof(*data), GFP_KERNEL);
866 	if (!data)
867 		return -ENOMEM;
868 	spin_lock_bh(&profile->lock);
869 	if (profile->data) {
870 		spin_unlock_bh(&profile->lock);
871 		kfree(data);
872 		return 0;
873 	}
874 	getnstimeofday(&data->starttod);
875 	profile->data = data;
876 	spin_unlock_bh(&profile->lock);
877 	return 0;
878 }
879 
880 void dasd_profile_off(struct dasd_profile *profile)
881 {
882 	spin_lock_bh(&profile->lock);
883 	kfree(profile->data);
884 	profile->data = NULL;
885 	spin_unlock_bh(&profile->lock);
886 }
887 
888 char *dasd_get_user_string(const char __user *user_buf, size_t user_len)
889 {
890 	char *buffer;
891 
892 	buffer = vmalloc(user_len + 1);
893 	if (buffer == NULL)
894 		return ERR_PTR(-ENOMEM);
895 	if (copy_from_user(buffer, user_buf, user_len) != 0) {
896 		vfree(buffer);
897 		return ERR_PTR(-EFAULT);
898 	}
899 	/* got the string, now strip linefeed. */
900 	if (buffer[user_len - 1] == '\n')
901 		buffer[user_len - 1] = 0;
902 	else
903 		buffer[user_len] = 0;
904 	return buffer;
905 }
906 
907 static ssize_t dasd_stats_write(struct file *file,
908 				const char __user *user_buf,
909 				size_t user_len, loff_t *pos)
910 {
911 	char *buffer, *str;
912 	int rc;
913 	struct seq_file *m = (struct seq_file *)file->private_data;
914 	struct dasd_profile *prof = m->private;
915 
916 	if (user_len > 65536)
917 		user_len = 65536;
918 	buffer = dasd_get_user_string(user_buf, user_len);
919 	if (IS_ERR(buffer))
920 		return PTR_ERR(buffer);
921 
922 	str = skip_spaces(buffer);
923 	rc = user_len;
924 	if (strncmp(str, "reset", 5) == 0) {
925 		dasd_profile_reset(prof);
926 	} else if (strncmp(str, "on", 2) == 0) {
927 		rc = dasd_profile_on(prof);
928 		if (!rc)
929 			rc = user_len;
930 	} else if (strncmp(str, "off", 3) == 0) {
931 		dasd_profile_off(prof);
932 	} else
933 		rc = -EINVAL;
934 	vfree(buffer);
935 	return rc;
936 }
937 
938 static void dasd_stats_array(struct seq_file *m, unsigned int *array)
939 {
940 	int i;
941 
942 	for (i = 0; i < 32; i++)
943 		seq_printf(m, "%u ", array[i]);
944 	seq_putc(m, '\n');
945 }
946 
947 static void dasd_stats_seq_print(struct seq_file *m,
948 				 struct dasd_profile_info *data)
949 {
950 	seq_printf(m, "start_time %ld.%09ld\n",
951 		   data->starttod.tv_sec, data->starttod.tv_nsec);
952 	seq_printf(m, "total_requests %u\n", data->dasd_io_reqs);
953 	seq_printf(m, "total_sectors %u\n", data->dasd_io_sects);
954 	seq_printf(m, "total_pav %u\n", data->dasd_io_alias);
955 	seq_printf(m, "total_hpf %u\n", data->dasd_io_tpm);
956 	seq_printf(m, "histogram_sectors ");
957 	dasd_stats_array(m, data->dasd_io_secs);
958 	seq_printf(m, "histogram_io_times ");
959 	dasd_stats_array(m, data->dasd_io_times);
960 	seq_printf(m, "histogram_io_times_weighted ");
961 	dasd_stats_array(m, data->dasd_io_timps);
962 	seq_printf(m, "histogram_time_build_to_ssch ");
963 	dasd_stats_array(m, data->dasd_io_time1);
964 	seq_printf(m, "histogram_time_ssch_to_irq ");
965 	dasd_stats_array(m, data->dasd_io_time2);
966 	seq_printf(m, "histogram_time_ssch_to_irq_weighted ");
967 	dasd_stats_array(m, data->dasd_io_time2ps);
968 	seq_printf(m, "histogram_time_irq_to_end ");
969 	dasd_stats_array(m, data->dasd_io_time3);
970 	seq_printf(m, "histogram_ccw_queue_length ");
971 	dasd_stats_array(m, data->dasd_io_nr_req);
972 	seq_printf(m, "total_read_requests %u\n", data->dasd_read_reqs);
973 	seq_printf(m, "total_read_sectors %u\n", data->dasd_read_sects);
974 	seq_printf(m, "total_read_pav %u\n", data->dasd_read_alias);
975 	seq_printf(m, "total_read_hpf %u\n", data->dasd_read_tpm);
976 	seq_printf(m, "histogram_read_sectors ");
977 	dasd_stats_array(m, data->dasd_read_secs);
978 	seq_printf(m, "histogram_read_times ");
979 	dasd_stats_array(m, data->dasd_read_times);
980 	seq_printf(m, "histogram_read_time_build_to_ssch ");
981 	dasd_stats_array(m, data->dasd_read_time1);
982 	seq_printf(m, "histogram_read_time_ssch_to_irq ");
983 	dasd_stats_array(m, data->dasd_read_time2);
984 	seq_printf(m, "histogram_read_time_irq_to_end ");
985 	dasd_stats_array(m, data->dasd_read_time3);
986 	seq_printf(m, "histogram_read_ccw_queue_length ");
987 	dasd_stats_array(m, data->dasd_read_nr_req);
988 }
989 
990 static int dasd_stats_show(struct seq_file *m, void *v)
991 {
992 	struct dasd_profile *profile;
993 	struct dasd_profile_info *data;
994 
995 	profile = m->private;
996 	spin_lock_bh(&profile->lock);
997 	data = profile->data;
998 	if (!data) {
999 		spin_unlock_bh(&profile->lock);
1000 		seq_printf(m, "disabled\n");
1001 		return 0;
1002 	}
1003 	dasd_stats_seq_print(m, data);
1004 	spin_unlock_bh(&profile->lock);
1005 	return 0;
1006 }
1007 
1008 static int dasd_stats_open(struct inode *inode, struct file *file)
1009 {
1010 	struct dasd_profile *profile = inode->i_private;
1011 	return single_open(file, dasd_stats_show, profile);
1012 }
1013 
1014 static const struct file_operations dasd_stats_raw_fops = {
1015 	.owner		= THIS_MODULE,
1016 	.open		= dasd_stats_open,
1017 	.read		= seq_read,
1018 	.llseek		= seq_lseek,
1019 	.release	= single_release,
1020 	.write		= dasd_stats_write,
1021 };
1022 
1023 static ssize_t dasd_stats_global_write(struct file *file,
1024 				       const char __user *user_buf,
1025 				       size_t user_len, loff_t *pos)
1026 {
1027 	char *buffer, *str;
1028 	ssize_t rc;
1029 
1030 	if (user_len > 65536)
1031 		user_len = 65536;
1032 	buffer = dasd_get_user_string(user_buf, user_len);
1033 	if (IS_ERR(buffer))
1034 		return PTR_ERR(buffer);
1035 	str = skip_spaces(buffer);
1036 	rc = user_len;
1037 	if (strncmp(str, "reset", 5) == 0) {
1038 		dasd_global_profile_reset();
1039 	} else if (strncmp(str, "on", 2) == 0) {
1040 		dasd_global_profile_reset();
1041 		dasd_global_profile_level = DASD_PROFILE_GLOBAL_ONLY;
1042 	} else if (strncmp(str, "off", 3) == 0) {
1043 		dasd_global_profile_level = DASD_PROFILE_OFF;
1044 	} else
1045 		rc = -EINVAL;
1046 	vfree(buffer);
1047 	return rc;
1048 }
1049 
1050 static int dasd_stats_global_show(struct seq_file *m, void *v)
1051 {
1052 	if (!dasd_global_profile_level) {
1053 		seq_printf(m, "disabled\n");
1054 		return 0;
1055 	}
1056 	dasd_stats_seq_print(m, &dasd_global_profile_data);
1057 	return 0;
1058 }
1059 
1060 static int dasd_stats_global_open(struct inode *inode, struct file *file)
1061 {
1062 	return single_open(file, dasd_stats_global_show, NULL);
1063 }
1064 
1065 static const struct file_operations dasd_stats_global_fops = {
1066 	.owner		= THIS_MODULE,
1067 	.open		= dasd_stats_global_open,
1068 	.read		= seq_read,
1069 	.llseek		= seq_lseek,
1070 	.release	= single_release,
1071 	.write		= dasd_stats_global_write,
1072 };
1073 
1074 static void dasd_profile_init(struct dasd_profile *profile,
1075 			      struct dentry *base_dentry)
1076 {
1077 	mode_t mode;
1078 	struct dentry *pde;
1079 
1080 	if (!base_dentry)
1081 		return;
1082 	profile->dentry = NULL;
1083 	profile->data = NULL;
1084 	mode = (S_IRUSR | S_IWUSR | S_IFREG);
1085 	pde = debugfs_create_file("statistics", mode, base_dentry,
1086 				  profile, &dasd_stats_raw_fops);
1087 	if (pde && !IS_ERR(pde))
1088 		profile->dentry = pde;
1089 	return;
1090 }
1091 
1092 static void dasd_profile_exit(struct dasd_profile *profile)
1093 {
1094 	dasd_profile_off(profile);
1095 	if (profile->dentry) {
1096 		debugfs_remove(profile->dentry);
1097 		profile->dentry = NULL;
1098 	}
1099 }
1100 
1101 static void dasd_statistics_removeroot(void)
1102 {
1103 	dasd_global_profile_level = DASD_PROFILE_OFF;
1104 	if (dasd_global_profile_dentry) {
1105 		debugfs_remove(dasd_global_profile_dentry);
1106 		dasd_global_profile_dentry = NULL;
1107 	}
1108 	if (dasd_debugfs_global_entry)
1109 		debugfs_remove(dasd_debugfs_global_entry);
1110 	if (dasd_debugfs_root_entry)
1111 		debugfs_remove(dasd_debugfs_root_entry);
1112 }
1113 
1114 static void dasd_statistics_createroot(void)
1115 {
1116 	mode_t mode;
1117 	struct dentry *pde;
1118 
1119 	dasd_debugfs_root_entry = NULL;
1120 	dasd_debugfs_global_entry = NULL;
1121 	dasd_global_profile_dentry = NULL;
1122 	pde = debugfs_create_dir("dasd", NULL);
1123 	if (!pde || IS_ERR(pde))
1124 		goto error;
1125 	dasd_debugfs_root_entry = pde;
1126 	pde = debugfs_create_dir("global", dasd_debugfs_root_entry);
1127 	if (!pde || IS_ERR(pde))
1128 		goto error;
1129 	dasd_debugfs_global_entry = pde;
1130 
1131 	mode = (S_IRUSR | S_IWUSR | S_IFREG);
1132 	pde = debugfs_create_file("statistics", mode, dasd_debugfs_global_entry,
1133 				  NULL, &dasd_stats_global_fops);
1134 	if (!pde || IS_ERR(pde))
1135 		goto error;
1136 	dasd_global_profile_dentry = pde;
1137 	return;
1138 
1139 error:
1140 	DBF_EVENT(DBF_ERR, "%s",
1141 		  "Creation of the dasd debugfs interface failed");
1142 	dasd_statistics_removeroot();
1143 	return;
1144 }
1145 
1146 #else
1147 #define dasd_profile_start(block, cqr, req) do {} while (0)
1148 #define dasd_profile_end(block, cqr, req) do {} while (0)
1149 
1150 static void dasd_statistics_createroot(void)
1151 {
1152 	return;
1153 }
1154 
1155 static void dasd_statistics_removeroot(void)
1156 {
1157 	return;
1158 }
1159 
1160 int dasd_stats_generic_show(struct seq_file *m, void *v)
1161 {
1162 	seq_printf(m, "Statistics are not activated in this kernel\n");
1163 	return 0;
1164 }
1165 
1166 static void dasd_profile_init(struct dasd_profile *profile,
1167 			      struct dentry *base_dentry)
1168 {
1169 	return;
1170 }
1171 
1172 static void dasd_profile_exit(struct dasd_profile *profile)
1173 {
1174 	return;
1175 }
1176 
1177 int dasd_profile_on(struct dasd_profile *profile)
1178 {
1179 	return 0;
1180 }
1181 
1182 #endif				/* CONFIG_DASD_PROFILE */
1183 
1184 /*
1185  * Allocate memory for a channel program with 'cplength' channel
1186  * command words and 'datasize' additional space. There are two
1187  * variantes: 1) dasd_kmalloc_request uses kmalloc to get the needed
1188  * memory and 2) dasd_smalloc_request uses the static ccw memory
1189  * that gets allocated for each device.
1190  */
1191 struct dasd_ccw_req *dasd_kmalloc_request(int magic, int cplength,
1192 					  int datasize,
1193 					  struct dasd_device *device)
1194 {
1195 	struct dasd_ccw_req *cqr;
1196 
1197 	/* Sanity checks */
1198 	BUG_ON(datasize > PAGE_SIZE ||
1199 	     (cplength*sizeof(struct ccw1)) > PAGE_SIZE);
1200 
1201 	cqr = kzalloc(sizeof(struct dasd_ccw_req), GFP_ATOMIC);
1202 	if (cqr == NULL)
1203 		return ERR_PTR(-ENOMEM);
1204 	cqr->cpaddr = NULL;
1205 	if (cplength > 0) {
1206 		cqr->cpaddr = kcalloc(cplength, sizeof(struct ccw1),
1207 				      GFP_ATOMIC | GFP_DMA);
1208 		if (cqr->cpaddr == NULL) {
1209 			kfree(cqr);
1210 			return ERR_PTR(-ENOMEM);
1211 		}
1212 	}
1213 	cqr->data = NULL;
1214 	if (datasize > 0) {
1215 		cqr->data = kzalloc(datasize, GFP_ATOMIC | GFP_DMA);
1216 		if (cqr->data == NULL) {
1217 			kfree(cqr->cpaddr);
1218 			kfree(cqr);
1219 			return ERR_PTR(-ENOMEM);
1220 		}
1221 	}
1222 	cqr->magic =  magic;
1223 	set_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags);
1224 	dasd_get_device(device);
1225 	return cqr;
1226 }
1227 
1228 struct dasd_ccw_req *dasd_smalloc_request(int magic, int cplength,
1229 					  int datasize,
1230 					  struct dasd_device *device)
1231 {
1232 	unsigned long flags;
1233 	struct dasd_ccw_req *cqr;
1234 	char *data;
1235 	int size;
1236 
1237 	size = (sizeof(struct dasd_ccw_req) + 7L) & -8L;
1238 	if (cplength > 0)
1239 		size += cplength * sizeof(struct ccw1);
1240 	if (datasize > 0)
1241 		size += datasize;
1242 	spin_lock_irqsave(&device->mem_lock, flags);
1243 	cqr = (struct dasd_ccw_req *)
1244 		dasd_alloc_chunk(&device->ccw_chunks, size);
1245 	spin_unlock_irqrestore(&device->mem_lock, flags);
1246 	if (cqr == NULL)
1247 		return ERR_PTR(-ENOMEM);
1248 	memset(cqr, 0, sizeof(struct dasd_ccw_req));
1249 	data = (char *) cqr + ((sizeof(struct dasd_ccw_req) + 7L) & -8L);
1250 	cqr->cpaddr = NULL;
1251 	if (cplength > 0) {
1252 		cqr->cpaddr = (struct ccw1 *) data;
1253 		data += cplength*sizeof(struct ccw1);
1254 		memset(cqr->cpaddr, 0, cplength*sizeof(struct ccw1));
1255 	}
1256 	cqr->data = NULL;
1257 	if (datasize > 0) {
1258 		cqr->data = data;
1259  		memset(cqr->data, 0, datasize);
1260 	}
1261 	cqr->magic = magic;
1262 	set_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags);
1263 	dasd_get_device(device);
1264 	return cqr;
1265 }
1266 
1267 /*
1268  * Free memory of a channel program. This function needs to free all the
1269  * idal lists that might have been created by dasd_set_cda and the
1270  * struct dasd_ccw_req itself.
1271  */
1272 void dasd_kfree_request(struct dasd_ccw_req *cqr, struct dasd_device *device)
1273 {
1274 #ifdef CONFIG_64BIT
1275 	struct ccw1 *ccw;
1276 
1277 	/* Clear any idals used for the request. */
1278 	ccw = cqr->cpaddr;
1279 	do {
1280 		clear_normalized_cda(ccw);
1281 	} while (ccw++->flags & (CCW_FLAG_CC | CCW_FLAG_DC));
1282 #endif
1283 	kfree(cqr->cpaddr);
1284 	kfree(cqr->data);
1285 	kfree(cqr);
1286 	dasd_put_device(device);
1287 }
1288 
1289 void dasd_sfree_request(struct dasd_ccw_req *cqr, struct dasd_device *device)
1290 {
1291 	unsigned long flags;
1292 
1293 	spin_lock_irqsave(&device->mem_lock, flags);
1294 	dasd_free_chunk(&device->ccw_chunks, cqr);
1295 	spin_unlock_irqrestore(&device->mem_lock, flags);
1296 	dasd_put_device(device);
1297 }
1298 
1299 /*
1300  * Check discipline magic in cqr.
1301  */
1302 static inline int dasd_check_cqr(struct dasd_ccw_req *cqr)
1303 {
1304 	struct dasd_device *device;
1305 
1306 	if (cqr == NULL)
1307 		return -EINVAL;
1308 	device = cqr->startdev;
1309 	if (strncmp((char *) &cqr->magic, device->discipline->ebcname, 4)) {
1310 		DBF_DEV_EVENT(DBF_WARNING, device,
1311 			    " dasd_ccw_req 0x%08x magic doesn't match"
1312 			    " discipline 0x%08x",
1313 			    cqr->magic,
1314 			    *(unsigned int *) device->discipline->name);
1315 		return -EINVAL;
1316 	}
1317 	return 0;
1318 }
1319 
1320 /*
1321  * Terminate the current i/o and set the request to clear_pending.
1322  * Timer keeps device runnig.
1323  * ccw_device_clear can fail if the i/o subsystem
1324  * is in a bad mood.
1325  */
1326 int dasd_term_IO(struct dasd_ccw_req *cqr)
1327 {
1328 	struct dasd_device *device;
1329 	int retries, rc;
1330 	char errorstring[ERRORLENGTH];
1331 
1332 	/* Check the cqr */
1333 	rc = dasd_check_cqr(cqr);
1334 	if (rc)
1335 		return rc;
1336 	retries = 0;
1337 	device = (struct dasd_device *) cqr->startdev;
1338 	while ((retries < 5) && (cqr->status == DASD_CQR_IN_IO)) {
1339 		rc = ccw_device_clear(device->cdev, (long) cqr);
1340 		switch (rc) {
1341 		case 0:	/* termination successful */
1342 			cqr->status = DASD_CQR_CLEAR_PENDING;
1343 			cqr->stopclk = get_clock();
1344 			cqr->starttime = 0;
1345 			DBF_DEV_EVENT(DBF_DEBUG, device,
1346 				      "terminate cqr %p successful",
1347 				      cqr);
1348 			break;
1349 		case -ENODEV:
1350 			DBF_DEV_EVENT(DBF_ERR, device, "%s",
1351 				      "device gone, retry");
1352 			break;
1353 		case -EIO:
1354 			DBF_DEV_EVENT(DBF_ERR, device, "%s",
1355 				      "I/O error, retry");
1356 			break;
1357 		case -EINVAL:
1358 		case -EBUSY:
1359 			DBF_DEV_EVENT(DBF_ERR, device, "%s",
1360 				      "device busy, retry later");
1361 			break;
1362 		default:
1363 			/* internal error 10 - unknown rc*/
1364 			snprintf(errorstring, ERRORLENGTH, "10 %d", rc);
1365 			dev_err(&device->cdev->dev, "An error occurred in the "
1366 				"DASD device driver, reason=%s\n", errorstring);
1367 			BUG();
1368 			break;
1369 		}
1370 		retries++;
1371 	}
1372 	dasd_schedule_device_bh(device);
1373 	return rc;
1374 }
1375 
1376 /*
1377  * Start the i/o. This start_IO can fail if the channel is really busy.
1378  * In that case set up a timer to start the request later.
1379  */
1380 int dasd_start_IO(struct dasd_ccw_req *cqr)
1381 {
1382 	struct dasd_device *device;
1383 	int rc;
1384 	char errorstring[ERRORLENGTH];
1385 
1386 	/* Check the cqr */
1387 	rc = dasd_check_cqr(cqr);
1388 	if (rc) {
1389 		cqr->intrc = rc;
1390 		return rc;
1391 	}
1392 	device = (struct dasd_device *) cqr->startdev;
1393 	if (((cqr->block &&
1394 	      test_bit(DASD_FLAG_LOCK_STOLEN, &cqr->block->base->flags)) ||
1395 	     test_bit(DASD_FLAG_LOCK_STOLEN, &device->flags)) &&
1396 	    !test_bit(DASD_CQR_ALLOW_SLOCK, &cqr->flags)) {
1397 		DBF_DEV_EVENT(DBF_DEBUG, device, "start_IO: return request %p "
1398 			      "because of stolen lock", cqr);
1399 		cqr->status = DASD_CQR_ERROR;
1400 		cqr->intrc = -EPERM;
1401 		return -EPERM;
1402 	}
1403 	if (cqr->retries < 0) {
1404 		/* internal error 14 - start_IO run out of retries */
1405 		sprintf(errorstring, "14 %p", cqr);
1406 		dev_err(&device->cdev->dev, "An error occurred in the DASD "
1407 			"device driver, reason=%s\n", errorstring);
1408 		cqr->status = DASD_CQR_ERROR;
1409 		return -EIO;
1410 	}
1411 	cqr->startclk = get_clock();
1412 	cqr->starttime = jiffies;
1413 	cqr->retries--;
1414 	if (!test_bit(DASD_CQR_VERIFY_PATH, &cqr->flags)) {
1415 		cqr->lpm &= device->path_data.opm;
1416 		if (!cqr->lpm)
1417 			cqr->lpm = device->path_data.opm;
1418 	}
1419 	if (cqr->cpmode == 1) {
1420 		rc = ccw_device_tm_start(device->cdev, cqr->cpaddr,
1421 					 (long) cqr, cqr->lpm);
1422 	} else {
1423 		rc = ccw_device_start(device->cdev, cqr->cpaddr,
1424 				      (long) cqr, cqr->lpm, 0);
1425 	}
1426 	switch (rc) {
1427 	case 0:
1428 		cqr->status = DASD_CQR_IN_IO;
1429 		break;
1430 	case -EBUSY:
1431 		DBF_DEV_EVENT(DBF_WARNING, device, "%s",
1432 			      "start_IO: device busy, retry later");
1433 		break;
1434 	case -ETIMEDOUT:
1435 		DBF_DEV_EVENT(DBF_WARNING, device, "%s",
1436 			      "start_IO: request timeout, retry later");
1437 		break;
1438 	case -EACCES:
1439 		/* -EACCES indicates that the request used only a subset of the
1440 		 * available paths and all these paths are gone. If the lpm of
1441 		 * this request was only a subset of the opm (e.g. the ppm) then
1442 		 * we just do a retry with all available paths.
1443 		 * If we already use the full opm, something is amiss, and we
1444 		 * need a full path verification.
1445 		 */
1446 		if (test_bit(DASD_CQR_VERIFY_PATH, &cqr->flags)) {
1447 			DBF_DEV_EVENT(DBF_WARNING, device,
1448 				      "start_IO: selected paths gone (%x)",
1449 				      cqr->lpm);
1450 		} else if (cqr->lpm != device->path_data.opm) {
1451 			cqr->lpm = device->path_data.opm;
1452 			DBF_DEV_EVENT(DBF_DEBUG, device, "%s",
1453 				      "start_IO: selected paths gone,"
1454 				      " retry on all paths");
1455 		} else {
1456 			DBF_DEV_EVENT(DBF_WARNING, device, "%s",
1457 				      "start_IO: all paths in opm gone,"
1458 				      " do path verification");
1459 			dasd_generic_last_path_gone(device);
1460 			device->path_data.opm = 0;
1461 			device->path_data.ppm = 0;
1462 			device->path_data.npm = 0;
1463 			device->path_data.tbvpm =
1464 				ccw_device_get_path_mask(device->cdev);
1465 		}
1466 		break;
1467 	case -ENODEV:
1468 		DBF_DEV_EVENT(DBF_WARNING, device, "%s",
1469 			      "start_IO: -ENODEV device gone, retry");
1470 		break;
1471 	case -EIO:
1472 		DBF_DEV_EVENT(DBF_WARNING, device, "%s",
1473 			      "start_IO: -EIO device gone, retry");
1474 		break;
1475 	case -EINVAL:
1476 		/* most likely caused in power management context */
1477 		DBF_DEV_EVENT(DBF_WARNING, device, "%s",
1478 			      "start_IO: -EINVAL device currently "
1479 			      "not accessible");
1480 		break;
1481 	default:
1482 		/* internal error 11 - unknown rc */
1483 		snprintf(errorstring, ERRORLENGTH, "11 %d", rc);
1484 		dev_err(&device->cdev->dev,
1485 			"An error occurred in the DASD device driver, "
1486 			"reason=%s\n", errorstring);
1487 		BUG();
1488 		break;
1489 	}
1490 	cqr->intrc = rc;
1491 	return rc;
1492 }
1493 
1494 /*
1495  * Timeout function for dasd devices. This is used for different purposes
1496  *  1) missing interrupt handler for normal operation
1497  *  2) delayed start of request where start_IO failed with -EBUSY
1498  *  3) timeout for missing state change interrupts
1499  * The head of the ccw queue will have status DASD_CQR_IN_IO for 1),
1500  * DASD_CQR_QUEUED for 2) and 3).
1501  */
1502 static void dasd_device_timeout(unsigned long ptr)
1503 {
1504 	unsigned long flags;
1505 	struct dasd_device *device;
1506 
1507 	device = (struct dasd_device *) ptr;
1508 	spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
1509 	/* re-activate request queue */
1510 	dasd_device_remove_stop_bits(device, DASD_STOPPED_PENDING);
1511 	spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
1512 	dasd_schedule_device_bh(device);
1513 }
1514 
1515 /*
1516  * Setup timeout for a device in jiffies.
1517  */
1518 void dasd_device_set_timer(struct dasd_device *device, int expires)
1519 {
1520 	if (expires == 0)
1521 		del_timer(&device->timer);
1522 	else
1523 		mod_timer(&device->timer, jiffies + expires);
1524 }
1525 
1526 /*
1527  * Clear timeout for a device.
1528  */
1529 void dasd_device_clear_timer(struct dasd_device *device)
1530 {
1531 	del_timer(&device->timer);
1532 }
1533 
1534 static void dasd_handle_killed_request(struct ccw_device *cdev,
1535 				       unsigned long intparm)
1536 {
1537 	struct dasd_ccw_req *cqr;
1538 	struct dasd_device *device;
1539 
1540 	if (!intparm)
1541 		return;
1542 	cqr = (struct dasd_ccw_req *) intparm;
1543 	if (cqr->status != DASD_CQR_IN_IO) {
1544 		DBF_EVENT_DEVID(DBF_DEBUG, cdev,
1545 				"invalid status in handle_killed_request: "
1546 				"%02x", cqr->status);
1547 		return;
1548 	}
1549 
1550 	device = dasd_device_from_cdev_locked(cdev);
1551 	if (IS_ERR(device)) {
1552 		DBF_EVENT_DEVID(DBF_DEBUG, cdev, "%s",
1553 				"unable to get device from cdev");
1554 		return;
1555 	}
1556 
1557 	if (!cqr->startdev ||
1558 	    device != cqr->startdev ||
1559 	    strncmp(cqr->startdev->discipline->ebcname,
1560 		    (char *) &cqr->magic, 4)) {
1561 		DBF_EVENT_DEVID(DBF_DEBUG, cdev, "%s",
1562 				"invalid device in request");
1563 		dasd_put_device(device);
1564 		return;
1565 	}
1566 
1567 	/* Schedule request to be retried. */
1568 	cqr->status = DASD_CQR_QUEUED;
1569 
1570 	dasd_device_clear_timer(device);
1571 	dasd_schedule_device_bh(device);
1572 	dasd_put_device(device);
1573 }
1574 
1575 void dasd_generic_handle_state_change(struct dasd_device *device)
1576 {
1577 	/* First of all start sense subsystem status request. */
1578 	dasd_eer_snss(device);
1579 
1580 	dasd_device_remove_stop_bits(device, DASD_STOPPED_PENDING);
1581 	dasd_schedule_device_bh(device);
1582 	if (device->block)
1583 		dasd_schedule_block_bh(device->block);
1584 }
1585 
1586 /*
1587  * Interrupt handler for "normal" ssch-io based dasd devices.
1588  */
1589 void dasd_int_handler(struct ccw_device *cdev, unsigned long intparm,
1590 		      struct irb *irb)
1591 {
1592 	struct dasd_ccw_req *cqr, *next;
1593 	struct dasd_device *device;
1594 	unsigned long long now;
1595 	int expires;
1596 
1597 	kstat_cpu(smp_processor_id()).irqs[IOINT_DAS]++;
1598 	if (IS_ERR(irb)) {
1599 		switch (PTR_ERR(irb)) {
1600 		case -EIO:
1601 			break;
1602 		case -ETIMEDOUT:
1603 			DBF_EVENT_DEVID(DBF_WARNING, cdev, "%s: "
1604 					"request timed out\n", __func__);
1605 			break;
1606 		default:
1607 			DBF_EVENT_DEVID(DBF_WARNING, cdev, "%s: "
1608 					"unknown error %ld\n", __func__,
1609 					PTR_ERR(irb));
1610 		}
1611 		dasd_handle_killed_request(cdev, intparm);
1612 		return;
1613 	}
1614 
1615 	now = get_clock();
1616 	cqr = (struct dasd_ccw_req *) intparm;
1617 	/* check for conditions that should be handled immediately */
1618 	if (!cqr ||
1619 	    !(scsw_dstat(&irb->scsw) == (DEV_STAT_CHN_END | DEV_STAT_DEV_END) &&
1620 	      scsw_cstat(&irb->scsw) == 0)) {
1621 		if (cqr)
1622 			memcpy(&cqr->irb, irb, sizeof(*irb));
1623 		device = dasd_device_from_cdev_locked(cdev);
1624 		if (IS_ERR(device))
1625 			return;
1626 		/* ignore unsolicited interrupts for DIAG discipline */
1627 		if (device->discipline == dasd_diag_discipline_pointer) {
1628 			dasd_put_device(device);
1629 			return;
1630 		}
1631 		device->discipline->dump_sense_dbf(device, irb, "int");
1632 		if (device->features & DASD_FEATURE_ERPLOG)
1633 			device->discipline->dump_sense(device, cqr, irb);
1634 		device->discipline->check_for_device_change(device, cqr, irb);
1635 		dasd_put_device(device);
1636 	}
1637 	if (!cqr)
1638 		return;
1639 
1640 	device = (struct dasd_device *) cqr->startdev;
1641 	if (!device ||
1642 	    strncmp(device->discipline->ebcname, (char *) &cqr->magic, 4)) {
1643 		DBF_EVENT_DEVID(DBF_DEBUG, cdev, "%s",
1644 				"invalid device in request");
1645 		return;
1646 	}
1647 
1648 	/* Check for clear pending */
1649 	if (cqr->status == DASD_CQR_CLEAR_PENDING &&
1650 	    scsw_fctl(&irb->scsw) & SCSW_FCTL_CLEAR_FUNC) {
1651 		cqr->status = DASD_CQR_CLEARED;
1652 		dasd_device_clear_timer(device);
1653 		wake_up(&dasd_flush_wq);
1654 		dasd_schedule_device_bh(device);
1655 		return;
1656 	}
1657 
1658 	/* check status - the request might have been killed by dyn detach */
1659 	if (cqr->status != DASD_CQR_IN_IO) {
1660 		DBF_DEV_EVENT(DBF_DEBUG, device, "invalid status: bus_id %s, "
1661 			      "status %02x", dev_name(&cdev->dev), cqr->status);
1662 		return;
1663 	}
1664 
1665 	next = NULL;
1666 	expires = 0;
1667 	if (scsw_dstat(&irb->scsw) == (DEV_STAT_CHN_END | DEV_STAT_DEV_END) &&
1668 	    scsw_cstat(&irb->scsw) == 0) {
1669 		/* request was completed successfully */
1670 		cqr->status = DASD_CQR_SUCCESS;
1671 		cqr->stopclk = now;
1672 		/* Start first request on queue if possible -> fast_io. */
1673 		if (cqr->devlist.next != &device->ccw_queue) {
1674 			next = list_entry(cqr->devlist.next,
1675 					  struct dasd_ccw_req, devlist);
1676 		}
1677 	} else {  /* error */
1678 		/*
1679 		 * If we don't want complex ERP for this request, then just
1680 		 * reset this and retry it in the fastpath
1681 		 */
1682 		if (!test_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags) &&
1683 		    cqr->retries > 0) {
1684 			if (cqr->lpm == device->path_data.opm)
1685 				DBF_DEV_EVENT(DBF_DEBUG, device,
1686 					      "default ERP in fastpath "
1687 					      "(%i retries left)",
1688 					      cqr->retries);
1689 			if (!test_bit(DASD_CQR_VERIFY_PATH, &cqr->flags))
1690 				cqr->lpm = device->path_data.opm;
1691 			cqr->status = DASD_CQR_QUEUED;
1692 			next = cqr;
1693 		} else
1694 			cqr->status = DASD_CQR_ERROR;
1695 	}
1696 	if (next && (next->status == DASD_CQR_QUEUED) &&
1697 	    (!device->stopped)) {
1698 		if (device->discipline->start_IO(next) == 0)
1699 			expires = next->expires;
1700 	}
1701 	if (expires != 0)
1702 		dasd_device_set_timer(device, expires);
1703 	else
1704 		dasd_device_clear_timer(device);
1705 	dasd_schedule_device_bh(device);
1706 }
1707 
1708 enum uc_todo dasd_generic_uc_handler(struct ccw_device *cdev, struct irb *irb)
1709 {
1710 	struct dasd_device *device;
1711 
1712 	device = dasd_device_from_cdev_locked(cdev);
1713 
1714 	if (IS_ERR(device))
1715 		goto out;
1716 	if (test_bit(DASD_FLAG_OFFLINE, &device->flags) ||
1717 	   device->state != device->target ||
1718 	   !device->discipline->check_for_device_change){
1719 		dasd_put_device(device);
1720 		goto out;
1721 	}
1722 	if (device->discipline->dump_sense_dbf)
1723 		device->discipline->dump_sense_dbf(device, irb, "uc");
1724 	device->discipline->check_for_device_change(device, NULL, irb);
1725 	dasd_put_device(device);
1726 out:
1727 	return UC_TODO_RETRY;
1728 }
1729 EXPORT_SYMBOL_GPL(dasd_generic_uc_handler);
1730 
1731 /*
1732  * If we have an error on a dasd_block layer request then we cancel
1733  * and return all further requests from the same dasd_block as well.
1734  */
1735 static void __dasd_device_recovery(struct dasd_device *device,
1736 				   struct dasd_ccw_req *ref_cqr)
1737 {
1738 	struct list_head *l, *n;
1739 	struct dasd_ccw_req *cqr;
1740 
1741 	/*
1742 	 * only requeue request that came from the dasd_block layer
1743 	 */
1744 	if (!ref_cqr->block)
1745 		return;
1746 
1747 	list_for_each_safe(l, n, &device->ccw_queue) {
1748 		cqr = list_entry(l, struct dasd_ccw_req, devlist);
1749 		if (cqr->status == DASD_CQR_QUEUED &&
1750 		    ref_cqr->block == cqr->block) {
1751 			cqr->status = DASD_CQR_CLEARED;
1752 		}
1753 	}
1754 };
1755 
1756 /*
1757  * Remove those ccw requests from the queue that need to be returned
1758  * to the upper layer.
1759  */
1760 static void __dasd_device_process_ccw_queue(struct dasd_device *device,
1761 					    struct list_head *final_queue)
1762 {
1763 	struct list_head *l, *n;
1764 	struct dasd_ccw_req *cqr;
1765 
1766 	/* Process request with final status. */
1767 	list_for_each_safe(l, n, &device->ccw_queue) {
1768 		cqr = list_entry(l, struct dasd_ccw_req, devlist);
1769 
1770 		/* Stop list processing at the first non-final request. */
1771 		if (cqr->status == DASD_CQR_QUEUED ||
1772 		    cqr->status == DASD_CQR_IN_IO ||
1773 		    cqr->status == DASD_CQR_CLEAR_PENDING)
1774 			break;
1775 		if (cqr->status == DASD_CQR_ERROR) {
1776 			__dasd_device_recovery(device, cqr);
1777 		}
1778 		/* Rechain finished requests to final queue */
1779 		list_move_tail(&cqr->devlist, final_queue);
1780 	}
1781 }
1782 
1783 /*
1784  * the cqrs from the final queue are returned to the upper layer
1785  * by setting a dasd_block state and calling the callback function
1786  */
1787 static void __dasd_device_process_final_queue(struct dasd_device *device,
1788 					      struct list_head *final_queue)
1789 {
1790 	struct list_head *l, *n;
1791 	struct dasd_ccw_req *cqr;
1792 	struct dasd_block *block;
1793 	void (*callback)(struct dasd_ccw_req *, void *data);
1794 	void *callback_data;
1795 	char errorstring[ERRORLENGTH];
1796 
1797 	list_for_each_safe(l, n, final_queue) {
1798 		cqr = list_entry(l, struct dasd_ccw_req, devlist);
1799 		list_del_init(&cqr->devlist);
1800 		block = cqr->block;
1801 		callback = cqr->callback;
1802 		callback_data = cqr->callback_data;
1803 		if (block)
1804 			spin_lock_bh(&block->queue_lock);
1805 		switch (cqr->status) {
1806 		case DASD_CQR_SUCCESS:
1807 			cqr->status = DASD_CQR_DONE;
1808 			break;
1809 		case DASD_CQR_ERROR:
1810 			cqr->status = DASD_CQR_NEED_ERP;
1811 			break;
1812 		case DASD_CQR_CLEARED:
1813 			cqr->status = DASD_CQR_TERMINATED;
1814 			break;
1815 		default:
1816 			/* internal error 12 - wrong cqr status*/
1817 			snprintf(errorstring, ERRORLENGTH, "12 %p %x02", cqr, cqr->status);
1818 			dev_err(&device->cdev->dev,
1819 				"An error occurred in the DASD device driver, "
1820 				"reason=%s\n", errorstring);
1821 			BUG();
1822 		}
1823 		if (cqr->callback != NULL)
1824 			(callback)(cqr, callback_data);
1825 		if (block)
1826 			spin_unlock_bh(&block->queue_lock);
1827 	}
1828 }
1829 
1830 /*
1831  * Take a look at the first request on the ccw queue and check
1832  * if it reached its expire time. If so, terminate the IO.
1833  */
1834 static void __dasd_device_check_expire(struct dasd_device *device)
1835 {
1836 	struct dasd_ccw_req *cqr;
1837 
1838 	if (list_empty(&device->ccw_queue))
1839 		return;
1840 	cqr = list_entry(device->ccw_queue.next, struct dasd_ccw_req, devlist);
1841 	if ((cqr->status == DASD_CQR_IN_IO && cqr->expires != 0) &&
1842 	    (time_after_eq(jiffies, cqr->expires + cqr->starttime))) {
1843 		if (device->discipline->term_IO(cqr) != 0) {
1844 			/* Hmpf, try again in 5 sec */
1845 			dev_err(&device->cdev->dev,
1846 				"cqr %p timed out (%lus) but cannot be "
1847 				"ended, retrying in 5 s\n",
1848 				cqr, (cqr->expires/HZ));
1849 			cqr->expires += 5*HZ;
1850 			dasd_device_set_timer(device, 5*HZ);
1851 		} else {
1852 			dev_err(&device->cdev->dev,
1853 				"cqr %p timed out (%lus), %i retries "
1854 				"remaining\n", cqr, (cqr->expires/HZ),
1855 				cqr->retries);
1856 		}
1857 	}
1858 }
1859 
1860 /*
1861  * Take a look at the first request on the ccw queue and check
1862  * if it needs to be started.
1863  */
1864 static void __dasd_device_start_head(struct dasd_device *device)
1865 {
1866 	struct dasd_ccw_req *cqr;
1867 	int rc;
1868 
1869 	if (list_empty(&device->ccw_queue))
1870 		return;
1871 	cqr = list_entry(device->ccw_queue.next, struct dasd_ccw_req, devlist);
1872 	if (cqr->status != DASD_CQR_QUEUED)
1873 		return;
1874 	/* when device is stopped, return request to previous layer
1875 	 * exception: only the disconnect or unresumed bits are set and the
1876 	 * cqr is a path verification request
1877 	 */
1878 	if (device->stopped &&
1879 	    !(!(device->stopped & ~(DASD_STOPPED_DC_WAIT | DASD_UNRESUMED_PM))
1880 	      && test_bit(DASD_CQR_VERIFY_PATH, &cqr->flags))) {
1881 		cqr->intrc = -EAGAIN;
1882 		cqr->status = DASD_CQR_CLEARED;
1883 		dasd_schedule_device_bh(device);
1884 		return;
1885 	}
1886 
1887 	rc = device->discipline->start_IO(cqr);
1888 	if (rc == 0)
1889 		dasd_device_set_timer(device, cqr->expires);
1890 	else if (rc == -EACCES) {
1891 		dasd_schedule_device_bh(device);
1892 	} else
1893 		/* Hmpf, try again in 1/2 sec */
1894 		dasd_device_set_timer(device, 50);
1895 }
1896 
1897 static void __dasd_device_check_path_events(struct dasd_device *device)
1898 {
1899 	int rc;
1900 
1901 	if (device->path_data.tbvpm) {
1902 		if (device->stopped & ~(DASD_STOPPED_DC_WAIT |
1903 					DASD_UNRESUMED_PM))
1904 			return;
1905 		rc = device->discipline->verify_path(
1906 			device, device->path_data.tbvpm);
1907 		if (rc)
1908 			dasd_device_set_timer(device, 50);
1909 		else
1910 			device->path_data.tbvpm = 0;
1911 	}
1912 };
1913 
1914 /*
1915  * Go through all request on the dasd_device request queue,
1916  * terminate them on the cdev if necessary, and return them to the
1917  * submitting layer via callback.
1918  * Note:
1919  * Make sure that all 'submitting layers' still exist when
1920  * this function is called!. In other words, when 'device' is a base
1921  * device then all block layer requests must have been removed before
1922  * via dasd_flush_block_queue.
1923  */
1924 int dasd_flush_device_queue(struct dasd_device *device)
1925 {
1926 	struct dasd_ccw_req *cqr, *n;
1927 	int rc;
1928 	struct list_head flush_queue;
1929 
1930 	INIT_LIST_HEAD(&flush_queue);
1931 	spin_lock_irq(get_ccwdev_lock(device->cdev));
1932 	rc = 0;
1933 	list_for_each_entry_safe(cqr, n, &device->ccw_queue, devlist) {
1934 		/* Check status and move request to flush_queue */
1935 		switch (cqr->status) {
1936 		case DASD_CQR_IN_IO:
1937 			rc = device->discipline->term_IO(cqr);
1938 			if (rc) {
1939 				/* unable to terminate requeust */
1940 				dev_err(&device->cdev->dev,
1941 					"Flushing the DASD request queue "
1942 					"failed for request %p\n", cqr);
1943 				/* stop flush processing */
1944 				goto finished;
1945 			}
1946 			break;
1947 		case DASD_CQR_QUEUED:
1948 			cqr->stopclk = get_clock();
1949 			cqr->status = DASD_CQR_CLEARED;
1950 			break;
1951 		default: /* no need to modify the others */
1952 			break;
1953 		}
1954 		list_move_tail(&cqr->devlist, &flush_queue);
1955 	}
1956 finished:
1957 	spin_unlock_irq(get_ccwdev_lock(device->cdev));
1958 	/*
1959 	 * After this point all requests must be in state CLEAR_PENDING,
1960 	 * CLEARED, SUCCESS or ERROR. Now wait for CLEAR_PENDING to become
1961 	 * one of the others.
1962 	 */
1963 	list_for_each_entry_safe(cqr, n, &flush_queue, devlist)
1964 		wait_event(dasd_flush_wq,
1965 			   (cqr->status != DASD_CQR_CLEAR_PENDING));
1966 	/*
1967 	 * Now set each request back to TERMINATED, DONE or NEED_ERP
1968 	 * and call the callback function of flushed requests
1969 	 */
1970 	__dasd_device_process_final_queue(device, &flush_queue);
1971 	return rc;
1972 }
1973 
1974 /*
1975  * Acquire the device lock and process queues for the device.
1976  */
1977 static void dasd_device_tasklet(struct dasd_device *device)
1978 {
1979 	struct list_head final_queue;
1980 
1981 	atomic_set (&device->tasklet_scheduled, 0);
1982 	INIT_LIST_HEAD(&final_queue);
1983 	spin_lock_irq(get_ccwdev_lock(device->cdev));
1984 	/* Check expire time of first request on the ccw queue. */
1985 	__dasd_device_check_expire(device);
1986 	/* find final requests on ccw queue */
1987 	__dasd_device_process_ccw_queue(device, &final_queue);
1988 	__dasd_device_check_path_events(device);
1989 	spin_unlock_irq(get_ccwdev_lock(device->cdev));
1990 	/* Now call the callback function of requests with final status */
1991 	__dasd_device_process_final_queue(device, &final_queue);
1992 	spin_lock_irq(get_ccwdev_lock(device->cdev));
1993 	/* Now check if the head of the ccw queue needs to be started. */
1994 	__dasd_device_start_head(device);
1995 	spin_unlock_irq(get_ccwdev_lock(device->cdev));
1996 	dasd_put_device(device);
1997 }
1998 
1999 /*
2000  * Schedules a call to dasd_tasklet over the device tasklet.
2001  */
2002 void dasd_schedule_device_bh(struct dasd_device *device)
2003 {
2004 	/* Protect against rescheduling. */
2005 	if (atomic_cmpxchg (&device->tasklet_scheduled, 0, 1) != 0)
2006 		return;
2007 	dasd_get_device(device);
2008 	tasklet_hi_schedule(&device->tasklet);
2009 }
2010 
2011 void dasd_device_set_stop_bits(struct dasd_device *device, int bits)
2012 {
2013 	device->stopped |= bits;
2014 }
2015 EXPORT_SYMBOL_GPL(dasd_device_set_stop_bits);
2016 
2017 void dasd_device_remove_stop_bits(struct dasd_device *device, int bits)
2018 {
2019 	device->stopped &= ~bits;
2020 	if (!device->stopped)
2021 		wake_up(&generic_waitq);
2022 }
2023 EXPORT_SYMBOL_GPL(dasd_device_remove_stop_bits);
2024 
2025 /*
2026  * Queue a request to the head of the device ccw_queue.
2027  * Start the I/O if possible.
2028  */
2029 void dasd_add_request_head(struct dasd_ccw_req *cqr)
2030 {
2031 	struct dasd_device *device;
2032 	unsigned long flags;
2033 
2034 	device = cqr->startdev;
2035 	spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
2036 	cqr->status = DASD_CQR_QUEUED;
2037 	list_add(&cqr->devlist, &device->ccw_queue);
2038 	/* let the bh start the request to keep them in order */
2039 	dasd_schedule_device_bh(device);
2040 	spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
2041 }
2042 
2043 /*
2044  * Queue a request to the tail of the device ccw_queue.
2045  * Start the I/O if possible.
2046  */
2047 void dasd_add_request_tail(struct dasd_ccw_req *cqr)
2048 {
2049 	struct dasd_device *device;
2050 	unsigned long flags;
2051 
2052 	device = cqr->startdev;
2053 	spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
2054 	cqr->status = DASD_CQR_QUEUED;
2055 	list_add_tail(&cqr->devlist, &device->ccw_queue);
2056 	/* let the bh start the request to keep them in order */
2057 	dasd_schedule_device_bh(device);
2058 	spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
2059 }
2060 
2061 /*
2062  * Wakeup helper for the 'sleep_on' functions.
2063  */
2064 static void dasd_wakeup_cb(struct dasd_ccw_req *cqr, void *data)
2065 {
2066 	spin_lock_irq(get_ccwdev_lock(cqr->startdev->cdev));
2067 	cqr->callback_data = DASD_SLEEPON_END_TAG;
2068 	spin_unlock_irq(get_ccwdev_lock(cqr->startdev->cdev));
2069 	wake_up(&generic_waitq);
2070 }
2071 
2072 static inline int _wait_for_wakeup(struct dasd_ccw_req *cqr)
2073 {
2074 	struct dasd_device *device;
2075 	int rc;
2076 
2077 	device = cqr->startdev;
2078 	spin_lock_irq(get_ccwdev_lock(device->cdev));
2079 	rc = (cqr->callback_data == DASD_SLEEPON_END_TAG);
2080 	spin_unlock_irq(get_ccwdev_lock(device->cdev));
2081 	return rc;
2082 }
2083 
2084 /*
2085  * checks if error recovery is necessary, returns 1 if yes, 0 otherwise.
2086  */
2087 static int __dasd_sleep_on_erp(struct dasd_ccw_req *cqr)
2088 {
2089 	struct dasd_device *device;
2090 	dasd_erp_fn_t erp_fn;
2091 
2092 	if (cqr->status == DASD_CQR_FILLED)
2093 		return 0;
2094 	device = cqr->startdev;
2095 	if (test_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags)) {
2096 		if (cqr->status == DASD_CQR_TERMINATED) {
2097 			device->discipline->handle_terminated_request(cqr);
2098 			return 1;
2099 		}
2100 		if (cqr->status == DASD_CQR_NEED_ERP) {
2101 			erp_fn = device->discipline->erp_action(cqr);
2102 			erp_fn(cqr);
2103 			return 1;
2104 		}
2105 		if (cqr->status == DASD_CQR_FAILED)
2106 			dasd_log_sense(cqr, &cqr->irb);
2107 		if (cqr->refers) {
2108 			__dasd_process_erp(device, cqr);
2109 			return 1;
2110 		}
2111 	}
2112 	return 0;
2113 }
2114 
2115 static int __dasd_sleep_on_loop_condition(struct dasd_ccw_req *cqr)
2116 {
2117 	if (test_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags)) {
2118 		if (cqr->refers) /* erp is not done yet */
2119 			return 1;
2120 		return ((cqr->status != DASD_CQR_DONE) &&
2121 			(cqr->status != DASD_CQR_FAILED));
2122 	} else
2123 		return (cqr->status == DASD_CQR_FILLED);
2124 }
2125 
2126 static int _dasd_sleep_on(struct dasd_ccw_req *maincqr, int interruptible)
2127 {
2128 	struct dasd_device *device;
2129 	int rc;
2130 	struct list_head ccw_queue;
2131 	struct dasd_ccw_req *cqr;
2132 
2133 	INIT_LIST_HEAD(&ccw_queue);
2134 	maincqr->status = DASD_CQR_FILLED;
2135 	device = maincqr->startdev;
2136 	list_add(&maincqr->blocklist, &ccw_queue);
2137 	for (cqr = maincqr;  __dasd_sleep_on_loop_condition(cqr);
2138 	     cqr = list_first_entry(&ccw_queue,
2139 				    struct dasd_ccw_req, blocklist)) {
2140 
2141 		if (__dasd_sleep_on_erp(cqr))
2142 			continue;
2143 		if (cqr->status != DASD_CQR_FILLED) /* could be failed */
2144 			continue;
2145 		if (test_bit(DASD_FLAG_LOCK_STOLEN, &device->flags) &&
2146 		    !test_bit(DASD_CQR_ALLOW_SLOCK, &cqr->flags)) {
2147 			cqr->status = DASD_CQR_FAILED;
2148 			cqr->intrc = -EPERM;
2149 			continue;
2150 		}
2151 		/* Non-temporary stop condition will trigger fail fast */
2152 		if (device->stopped & ~DASD_STOPPED_PENDING &&
2153 		    test_bit(DASD_CQR_FLAGS_FAILFAST, &cqr->flags) &&
2154 		    (!dasd_eer_enabled(device))) {
2155 			cqr->status = DASD_CQR_FAILED;
2156 			continue;
2157 		}
2158 		/* Don't try to start requests if device is stopped */
2159 		if (interruptible) {
2160 			rc = wait_event_interruptible(
2161 				generic_waitq, !(device->stopped));
2162 			if (rc == -ERESTARTSYS) {
2163 				cqr->status = DASD_CQR_FAILED;
2164 				maincqr->intrc = rc;
2165 				continue;
2166 			}
2167 		} else
2168 			wait_event(generic_waitq, !(device->stopped));
2169 
2170 		cqr->callback = dasd_wakeup_cb;
2171 		cqr->callback_data = DASD_SLEEPON_START_TAG;
2172 		dasd_add_request_tail(cqr);
2173 		if (interruptible) {
2174 			rc = wait_event_interruptible(
2175 				generic_waitq, _wait_for_wakeup(cqr));
2176 			if (rc == -ERESTARTSYS) {
2177 				dasd_cancel_req(cqr);
2178 				/* wait (non-interruptible) for final status */
2179 				wait_event(generic_waitq,
2180 					   _wait_for_wakeup(cqr));
2181 				cqr->status = DASD_CQR_FAILED;
2182 				maincqr->intrc = rc;
2183 				continue;
2184 			}
2185 		} else
2186 			wait_event(generic_waitq, _wait_for_wakeup(cqr));
2187 	}
2188 
2189 	maincqr->endclk = get_clock();
2190 	if ((maincqr->status != DASD_CQR_DONE) &&
2191 	    (maincqr->intrc != -ERESTARTSYS))
2192 		dasd_log_sense(maincqr, &maincqr->irb);
2193 	if (maincqr->status == DASD_CQR_DONE)
2194 		rc = 0;
2195 	else if (maincqr->intrc)
2196 		rc = maincqr->intrc;
2197 	else
2198 		rc = -EIO;
2199 	return rc;
2200 }
2201 
2202 /*
2203  * Queue a request to the tail of the device ccw_queue and wait for
2204  * it's completion.
2205  */
2206 int dasd_sleep_on(struct dasd_ccw_req *cqr)
2207 {
2208 	return _dasd_sleep_on(cqr, 0);
2209 }
2210 
2211 /*
2212  * Queue a request to the tail of the device ccw_queue and wait
2213  * interruptible for it's completion.
2214  */
2215 int dasd_sleep_on_interruptible(struct dasd_ccw_req *cqr)
2216 {
2217 	return _dasd_sleep_on(cqr, 1);
2218 }
2219 
2220 /*
2221  * Whoa nelly now it gets really hairy. For some functions (e.g. steal lock
2222  * for eckd devices) the currently running request has to be terminated
2223  * and be put back to status queued, before the special request is added
2224  * to the head of the queue. Then the special request is waited on normally.
2225  */
2226 static inline int _dasd_term_running_cqr(struct dasd_device *device)
2227 {
2228 	struct dasd_ccw_req *cqr;
2229 	int rc;
2230 
2231 	if (list_empty(&device->ccw_queue))
2232 		return 0;
2233 	cqr = list_entry(device->ccw_queue.next, struct dasd_ccw_req, devlist);
2234 	rc = device->discipline->term_IO(cqr);
2235 	if (!rc)
2236 		/*
2237 		 * CQR terminated because a more important request is pending.
2238 		 * Undo decreasing of retry counter because this is
2239 		 * not an error case.
2240 		 */
2241 		cqr->retries++;
2242 	return rc;
2243 }
2244 
2245 int dasd_sleep_on_immediatly(struct dasd_ccw_req *cqr)
2246 {
2247 	struct dasd_device *device;
2248 	int rc;
2249 
2250 	device = cqr->startdev;
2251 	if (test_bit(DASD_FLAG_LOCK_STOLEN, &device->flags) &&
2252 	    !test_bit(DASD_CQR_ALLOW_SLOCK, &cqr->flags)) {
2253 		cqr->status = DASD_CQR_FAILED;
2254 		cqr->intrc = -EPERM;
2255 		return -EIO;
2256 	}
2257 	spin_lock_irq(get_ccwdev_lock(device->cdev));
2258 	rc = _dasd_term_running_cqr(device);
2259 	if (rc) {
2260 		spin_unlock_irq(get_ccwdev_lock(device->cdev));
2261 		return rc;
2262 	}
2263 	cqr->callback = dasd_wakeup_cb;
2264 	cqr->callback_data = DASD_SLEEPON_START_TAG;
2265 	cqr->status = DASD_CQR_QUEUED;
2266 	list_add(&cqr->devlist, &device->ccw_queue);
2267 
2268 	/* let the bh start the request to keep them in order */
2269 	dasd_schedule_device_bh(device);
2270 
2271 	spin_unlock_irq(get_ccwdev_lock(device->cdev));
2272 
2273 	wait_event(generic_waitq, _wait_for_wakeup(cqr));
2274 
2275 	if (cqr->status == DASD_CQR_DONE)
2276 		rc = 0;
2277 	else if (cqr->intrc)
2278 		rc = cqr->intrc;
2279 	else
2280 		rc = -EIO;
2281 	return rc;
2282 }
2283 
2284 /*
2285  * Cancels a request that was started with dasd_sleep_on_req.
2286  * This is useful to timeout requests. The request will be
2287  * terminated if it is currently in i/o.
2288  * Returns 1 if the request has been terminated.
2289  *	   0 if there was no need to terminate the request (not started yet)
2290  *	   negative error code if termination failed
2291  * Cancellation of a request is an asynchronous operation! The calling
2292  * function has to wait until the request is properly returned via callback.
2293  */
2294 int dasd_cancel_req(struct dasd_ccw_req *cqr)
2295 {
2296 	struct dasd_device *device = cqr->startdev;
2297 	unsigned long flags;
2298 	int rc;
2299 
2300 	rc = 0;
2301 	spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
2302 	switch (cqr->status) {
2303 	case DASD_CQR_QUEUED:
2304 		/* request was not started - just set to cleared */
2305 		cqr->status = DASD_CQR_CLEARED;
2306 		break;
2307 	case DASD_CQR_IN_IO:
2308 		/* request in IO - terminate IO and release again */
2309 		rc = device->discipline->term_IO(cqr);
2310 		if (rc) {
2311 			dev_err(&device->cdev->dev,
2312 				"Cancelling request %p failed with rc=%d\n",
2313 				cqr, rc);
2314 		} else {
2315 			cqr->stopclk = get_clock();
2316 		}
2317 		break;
2318 	default: /* already finished or clear pending - do nothing */
2319 		break;
2320 	}
2321 	spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
2322 	dasd_schedule_device_bh(device);
2323 	return rc;
2324 }
2325 
2326 
2327 /*
2328  * SECTION: Operations of the dasd_block layer.
2329  */
2330 
2331 /*
2332  * Timeout function for dasd_block. This is used when the block layer
2333  * is waiting for something that may not come reliably, (e.g. a state
2334  * change interrupt)
2335  */
2336 static void dasd_block_timeout(unsigned long ptr)
2337 {
2338 	unsigned long flags;
2339 	struct dasd_block *block;
2340 
2341 	block = (struct dasd_block *) ptr;
2342 	spin_lock_irqsave(get_ccwdev_lock(block->base->cdev), flags);
2343 	/* re-activate request queue */
2344 	dasd_device_remove_stop_bits(block->base, DASD_STOPPED_PENDING);
2345 	spin_unlock_irqrestore(get_ccwdev_lock(block->base->cdev), flags);
2346 	dasd_schedule_block_bh(block);
2347 }
2348 
2349 /*
2350  * Setup timeout for a dasd_block in jiffies.
2351  */
2352 void dasd_block_set_timer(struct dasd_block *block, int expires)
2353 {
2354 	if (expires == 0)
2355 		del_timer(&block->timer);
2356 	else
2357 		mod_timer(&block->timer, jiffies + expires);
2358 }
2359 
2360 /*
2361  * Clear timeout for a dasd_block.
2362  */
2363 void dasd_block_clear_timer(struct dasd_block *block)
2364 {
2365 	del_timer(&block->timer);
2366 }
2367 
2368 /*
2369  * Process finished error recovery ccw.
2370  */
2371 static void __dasd_process_erp(struct dasd_device *device,
2372 			       struct dasd_ccw_req *cqr)
2373 {
2374 	dasd_erp_fn_t erp_fn;
2375 
2376 	if (cqr->status == DASD_CQR_DONE)
2377 		DBF_DEV_EVENT(DBF_NOTICE, device, "%s", "ERP successful");
2378 	else
2379 		dev_err(&device->cdev->dev, "ERP failed for the DASD\n");
2380 	erp_fn = device->discipline->erp_postaction(cqr);
2381 	erp_fn(cqr);
2382 }
2383 
2384 /*
2385  * Fetch requests from the block device queue.
2386  */
2387 static void __dasd_process_request_queue(struct dasd_block *block)
2388 {
2389 	struct request_queue *queue;
2390 	struct request *req;
2391 	struct dasd_ccw_req *cqr;
2392 	struct dasd_device *basedev;
2393 	unsigned long flags;
2394 	queue = block->request_queue;
2395 	basedev = block->base;
2396 	/* No queue ? Then there is nothing to do. */
2397 	if (queue == NULL)
2398 		return;
2399 
2400 	/*
2401 	 * We requeue request from the block device queue to the ccw
2402 	 * queue only in two states. In state DASD_STATE_READY the
2403 	 * partition detection is done and we need to requeue requests
2404 	 * for that. State DASD_STATE_ONLINE is normal block device
2405 	 * operation.
2406 	 */
2407 	if (basedev->state < DASD_STATE_READY) {
2408 		while ((req = blk_fetch_request(block->request_queue)))
2409 			__blk_end_request_all(req, -EIO);
2410 		return;
2411 	}
2412 	/* Now we try to fetch requests from the request queue */
2413 	while ((req = blk_peek_request(queue))) {
2414 		if (basedev->features & DASD_FEATURE_READONLY &&
2415 		    rq_data_dir(req) == WRITE) {
2416 			DBF_DEV_EVENT(DBF_ERR, basedev,
2417 				      "Rejecting write request %p",
2418 				      req);
2419 			blk_start_request(req);
2420 			__blk_end_request_all(req, -EIO);
2421 			continue;
2422 		}
2423 		cqr = basedev->discipline->build_cp(basedev, block, req);
2424 		if (IS_ERR(cqr)) {
2425 			if (PTR_ERR(cqr) == -EBUSY)
2426 				break;	/* normal end condition */
2427 			if (PTR_ERR(cqr) == -ENOMEM)
2428 				break;	/* terminate request queue loop */
2429 			if (PTR_ERR(cqr) == -EAGAIN) {
2430 				/*
2431 				 * The current request cannot be build right
2432 				 * now, we have to try later. If this request
2433 				 * is the head-of-queue we stop the device
2434 				 * for 1/2 second.
2435 				 */
2436 				if (!list_empty(&block->ccw_queue))
2437 					break;
2438 				spin_lock_irqsave(
2439 					get_ccwdev_lock(basedev->cdev), flags);
2440 				dasd_device_set_stop_bits(basedev,
2441 							  DASD_STOPPED_PENDING);
2442 				spin_unlock_irqrestore(
2443 					get_ccwdev_lock(basedev->cdev), flags);
2444 				dasd_block_set_timer(block, HZ/2);
2445 				break;
2446 			}
2447 			DBF_DEV_EVENT(DBF_ERR, basedev,
2448 				      "CCW creation failed (rc=%ld) "
2449 				      "on request %p",
2450 				      PTR_ERR(cqr), req);
2451 			blk_start_request(req);
2452 			__blk_end_request_all(req, -EIO);
2453 			continue;
2454 		}
2455 		/*
2456 		 *  Note: callback is set to dasd_return_cqr_cb in
2457 		 * __dasd_block_start_head to cover erp requests as well
2458 		 */
2459 		cqr->callback_data = (void *) req;
2460 		cqr->status = DASD_CQR_FILLED;
2461 		blk_start_request(req);
2462 		list_add_tail(&cqr->blocklist, &block->ccw_queue);
2463 		dasd_profile_start(block, cqr, req);
2464 	}
2465 }
2466 
2467 static void __dasd_cleanup_cqr(struct dasd_ccw_req *cqr)
2468 {
2469 	struct request *req;
2470 	int status;
2471 	int error = 0;
2472 
2473 	req = (struct request *) cqr->callback_data;
2474 	dasd_profile_end(cqr->block, cqr, req);
2475 	status = cqr->block->base->discipline->free_cp(cqr, req);
2476 	if (status <= 0)
2477 		error = status ? status : -EIO;
2478 	__blk_end_request_all(req, error);
2479 }
2480 
2481 /*
2482  * Process ccw request queue.
2483  */
2484 static void __dasd_process_block_ccw_queue(struct dasd_block *block,
2485 					   struct list_head *final_queue)
2486 {
2487 	struct list_head *l, *n;
2488 	struct dasd_ccw_req *cqr;
2489 	dasd_erp_fn_t erp_fn;
2490 	unsigned long flags;
2491 	struct dasd_device *base = block->base;
2492 
2493 restart:
2494 	/* Process request with final status. */
2495 	list_for_each_safe(l, n, &block->ccw_queue) {
2496 		cqr = list_entry(l, struct dasd_ccw_req, blocklist);
2497 		if (cqr->status != DASD_CQR_DONE &&
2498 		    cqr->status != DASD_CQR_FAILED &&
2499 		    cqr->status != DASD_CQR_NEED_ERP &&
2500 		    cqr->status != DASD_CQR_TERMINATED)
2501 			continue;
2502 
2503 		if (cqr->status == DASD_CQR_TERMINATED) {
2504 			base->discipline->handle_terminated_request(cqr);
2505 			goto restart;
2506 		}
2507 
2508 		/*  Process requests that may be recovered */
2509 		if (cqr->status == DASD_CQR_NEED_ERP) {
2510 			erp_fn = base->discipline->erp_action(cqr);
2511 			if (IS_ERR(erp_fn(cqr)))
2512 				continue;
2513 			goto restart;
2514 		}
2515 
2516 		/* log sense for fatal error */
2517 		if (cqr->status == DASD_CQR_FAILED) {
2518 			dasd_log_sense(cqr, &cqr->irb);
2519 		}
2520 
2521 		/* First of all call extended error reporting. */
2522 		if (dasd_eer_enabled(base) &&
2523 		    cqr->status == DASD_CQR_FAILED) {
2524 			dasd_eer_write(base, cqr, DASD_EER_FATALERROR);
2525 
2526 			/* restart request  */
2527 			cqr->status = DASD_CQR_FILLED;
2528 			cqr->retries = 255;
2529 			spin_lock_irqsave(get_ccwdev_lock(base->cdev), flags);
2530 			dasd_device_set_stop_bits(base, DASD_STOPPED_QUIESCE);
2531 			spin_unlock_irqrestore(get_ccwdev_lock(base->cdev),
2532 					       flags);
2533 			goto restart;
2534 		}
2535 
2536 		/* Process finished ERP request. */
2537 		if (cqr->refers) {
2538 			__dasd_process_erp(base, cqr);
2539 			goto restart;
2540 		}
2541 
2542 		/* Rechain finished requests to final queue */
2543 		cqr->endclk = get_clock();
2544 		list_move_tail(&cqr->blocklist, final_queue);
2545 	}
2546 }
2547 
2548 static void dasd_return_cqr_cb(struct dasd_ccw_req *cqr, void *data)
2549 {
2550 	dasd_schedule_block_bh(cqr->block);
2551 }
2552 
2553 static void __dasd_block_start_head(struct dasd_block *block)
2554 {
2555 	struct dasd_ccw_req *cqr;
2556 
2557 	if (list_empty(&block->ccw_queue))
2558 		return;
2559 	/* We allways begin with the first requests on the queue, as some
2560 	 * of previously started requests have to be enqueued on a
2561 	 * dasd_device again for error recovery.
2562 	 */
2563 	list_for_each_entry(cqr, &block->ccw_queue, blocklist) {
2564 		if (cqr->status != DASD_CQR_FILLED)
2565 			continue;
2566 		if (test_bit(DASD_FLAG_LOCK_STOLEN, &block->base->flags) &&
2567 		    !test_bit(DASD_CQR_ALLOW_SLOCK, &cqr->flags)) {
2568 			cqr->status = DASD_CQR_FAILED;
2569 			cqr->intrc = -EPERM;
2570 			dasd_schedule_block_bh(block);
2571 			continue;
2572 		}
2573 		/* Non-temporary stop condition will trigger fail fast */
2574 		if (block->base->stopped & ~DASD_STOPPED_PENDING &&
2575 		    test_bit(DASD_CQR_FLAGS_FAILFAST, &cqr->flags) &&
2576 		    (!dasd_eer_enabled(block->base))) {
2577 			cqr->status = DASD_CQR_FAILED;
2578 			dasd_schedule_block_bh(block);
2579 			continue;
2580 		}
2581 		/* Don't try to start requests if device is stopped */
2582 		if (block->base->stopped)
2583 			return;
2584 
2585 		/* just a fail safe check, should not happen */
2586 		if (!cqr->startdev)
2587 			cqr->startdev = block->base;
2588 
2589 		/* make sure that the requests we submit find their way back */
2590 		cqr->callback = dasd_return_cqr_cb;
2591 
2592 		dasd_add_request_tail(cqr);
2593 	}
2594 }
2595 
2596 /*
2597  * Central dasd_block layer routine. Takes requests from the generic
2598  * block layer request queue, creates ccw requests, enqueues them on
2599  * a dasd_device and processes ccw requests that have been returned.
2600  */
2601 static void dasd_block_tasklet(struct dasd_block *block)
2602 {
2603 	struct list_head final_queue;
2604 	struct list_head *l, *n;
2605 	struct dasd_ccw_req *cqr;
2606 
2607 	atomic_set(&block->tasklet_scheduled, 0);
2608 	INIT_LIST_HEAD(&final_queue);
2609 	spin_lock(&block->queue_lock);
2610 	/* Finish off requests on ccw queue */
2611 	__dasd_process_block_ccw_queue(block, &final_queue);
2612 	spin_unlock(&block->queue_lock);
2613 	/* Now call the callback function of requests with final status */
2614 	spin_lock_irq(&block->request_queue_lock);
2615 	list_for_each_safe(l, n, &final_queue) {
2616 		cqr = list_entry(l, struct dasd_ccw_req, blocklist);
2617 		list_del_init(&cqr->blocklist);
2618 		__dasd_cleanup_cqr(cqr);
2619 	}
2620 	spin_lock(&block->queue_lock);
2621 	/* Get new request from the block device request queue */
2622 	__dasd_process_request_queue(block);
2623 	/* Now check if the head of the ccw queue needs to be started. */
2624 	__dasd_block_start_head(block);
2625 	spin_unlock(&block->queue_lock);
2626 	spin_unlock_irq(&block->request_queue_lock);
2627 	dasd_put_device(block->base);
2628 }
2629 
2630 static void _dasd_wake_block_flush_cb(struct dasd_ccw_req *cqr, void *data)
2631 {
2632 	wake_up(&dasd_flush_wq);
2633 }
2634 
2635 /*
2636  * Go through all request on the dasd_block request queue, cancel them
2637  * on the respective dasd_device, and return them to the generic
2638  * block layer.
2639  */
2640 static int dasd_flush_block_queue(struct dasd_block *block)
2641 {
2642 	struct dasd_ccw_req *cqr, *n;
2643 	int rc, i;
2644 	struct list_head flush_queue;
2645 
2646 	INIT_LIST_HEAD(&flush_queue);
2647 	spin_lock_bh(&block->queue_lock);
2648 	rc = 0;
2649 restart:
2650 	list_for_each_entry_safe(cqr, n, &block->ccw_queue, blocklist) {
2651 		/* if this request currently owned by a dasd_device cancel it */
2652 		if (cqr->status >= DASD_CQR_QUEUED)
2653 			rc = dasd_cancel_req(cqr);
2654 		if (rc < 0)
2655 			break;
2656 		/* Rechain request (including erp chain) so it won't be
2657 		 * touched by the dasd_block_tasklet anymore.
2658 		 * Replace the callback so we notice when the request
2659 		 * is returned from the dasd_device layer.
2660 		 */
2661 		cqr->callback = _dasd_wake_block_flush_cb;
2662 		for (i = 0; cqr != NULL; cqr = cqr->refers, i++)
2663 			list_move_tail(&cqr->blocklist, &flush_queue);
2664 		if (i > 1)
2665 			/* moved more than one request - need to restart */
2666 			goto restart;
2667 	}
2668 	spin_unlock_bh(&block->queue_lock);
2669 	/* Now call the callback function of flushed requests */
2670 restart_cb:
2671 	list_for_each_entry_safe(cqr, n, &flush_queue, blocklist) {
2672 		wait_event(dasd_flush_wq, (cqr->status < DASD_CQR_QUEUED));
2673 		/* Process finished ERP request. */
2674 		if (cqr->refers) {
2675 			spin_lock_bh(&block->queue_lock);
2676 			__dasd_process_erp(block->base, cqr);
2677 			spin_unlock_bh(&block->queue_lock);
2678 			/* restart list_for_xx loop since dasd_process_erp
2679 			 * might remove multiple elements */
2680 			goto restart_cb;
2681 		}
2682 		/* call the callback function */
2683 		spin_lock_irq(&block->request_queue_lock);
2684 		cqr->endclk = get_clock();
2685 		list_del_init(&cqr->blocklist);
2686 		__dasd_cleanup_cqr(cqr);
2687 		spin_unlock_irq(&block->request_queue_lock);
2688 	}
2689 	return rc;
2690 }
2691 
2692 /*
2693  * Schedules a call to dasd_tasklet over the device tasklet.
2694  */
2695 void dasd_schedule_block_bh(struct dasd_block *block)
2696 {
2697 	/* Protect against rescheduling. */
2698 	if (atomic_cmpxchg(&block->tasklet_scheduled, 0, 1) != 0)
2699 		return;
2700 	/* life cycle of block is bound to it's base device */
2701 	dasd_get_device(block->base);
2702 	tasklet_hi_schedule(&block->tasklet);
2703 }
2704 
2705 
2706 /*
2707  * SECTION: external block device operations
2708  * (request queue handling, open, release, etc.)
2709  */
2710 
2711 /*
2712  * Dasd request queue function. Called from ll_rw_blk.c
2713  */
2714 static void do_dasd_request(struct request_queue *queue)
2715 {
2716 	struct dasd_block *block;
2717 
2718 	block = queue->queuedata;
2719 	spin_lock(&block->queue_lock);
2720 	/* Get new request from the block device request queue */
2721 	__dasd_process_request_queue(block);
2722 	/* Now check if the head of the ccw queue needs to be started. */
2723 	__dasd_block_start_head(block);
2724 	spin_unlock(&block->queue_lock);
2725 }
2726 
2727 /*
2728  * Allocate and initialize request queue and default I/O scheduler.
2729  */
2730 static int dasd_alloc_queue(struct dasd_block *block)
2731 {
2732 	int rc;
2733 
2734 	block->request_queue = blk_init_queue(do_dasd_request,
2735 					       &block->request_queue_lock);
2736 	if (block->request_queue == NULL)
2737 		return -ENOMEM;
2738 
2739 	block->request_queue->queuedata = block;
2740 
2741 	elevator_exit(block->request_queue->elevator);
2742 	block->request_queue->elevator = NULL;
2743 	rc = elevator_init(block->request_queue, "deadline");
2744 	if (rc) {
2745 		blk_cleanup_queue(block->request_queue);
2746 		return rc;
2747 	}
2748 	return 0;
2749 }
2750 
2751 /*
2752  * Allocate and initialize request queue.
2753  */
2754 static void dasd_setup_queue(struct dasd_block *block)
2755 {
2756 	int max;
2757 
2758 	if (block->base->features & DASD_FEATURE_USERAW) {
2759 		/*
2760 		 * the max_blocks value for raw_track access is 256
2761 		 * it is higher than the native ECKD value because we
2762 		 * only need one ccw per track
2763 		 * so the max_hw_sectors are
2764 		 * 2048 x 512B = 1024kB = 16 tracks
2765 		 */
2766 		max = 2048;
2767 	} else {
2768 		max = block->base->discipline->max_blocks << block->s2b_shift;
2769 	}
2770 	blk_queue_logical_block_size(block->request_queue,
2771 				     block->bp_block);
2772 	blk_queue_max_hw_sectors(block->request_queue, max);
2773 	blk_queue_max_segments(block->request_queue, -1L);
2774 	/* with page sized segments we can translate each segement into
2775 	 * one idaw/tidaw
2776 	 */
2777 	blk_queue_max_segment_size(block->request_queue, PAGE_SIZE);
2778 	blk_queue_segment_boundary(block->request_queue, PAGE_SIZE - 1);
2779 }
2780 
2781 /*
2782  * Deactivate and free request queue.
2783  */
2784 static void dasd_free_queue(struct dasd_block *block)
2785 {
2786 	if (block->request_queue) {
2787 		blk_cleanup_queue(block->request_queue);
2788 		block->request_queue = NULL;
2789 	}
2790 }
2791 
2792 /*
2793  * Flush request on the request queue.
2794  */
2795 static void dasd_flush_request_queue(struct dasd_block *block)
2796 {
2797 	struct request *req;
2798 
2799 	if (!block->request_queue)
2800 		return;
2801 
2802 	spin_lock_irq(&block->request_queue_lock);
2803 	while ((req = blk_fetch_request(block->request_queue)))
2804 		__blk_end_request_all(req, -EIO);
2805 	spin_unlock_irq(&block->request_queue_lock);
2806 }
2807 
2808 static int dasd_open(struct block_device *bdev, fmode_t mode)
2809 {
2810 	struct dasd_device *base;
2811 	int rc;
2812 
2813 	base = dasd_device_from_gendisk(bdev->bd_disk);
2814 	if (!base)
2815 		return -ENODEV;
2816 
2817 	atomic_inc(&base->block->open_count);
2818 	if (test_bit(DASD_FLAG_OFFLINE, &base->flags)) {
2819 		rc = -ENODEV;
2820 		goto unlock;
2821 	}
2822 
2823 	if (!try_module_get(base->discipline->owner)) {
2824 		rc = -EINVAL;
2825 		goto unlock;
2826 	}
2827 
2828 	if (dasd_probeonly) {
2829 		dev_info(&base->cdev->dev,
2830 			 "Accessing the DASD failed because it is in "
2831 			 "probeonly mode\n");
2832 		rc = -EPERM;
2833 		goto out;
2834 	}
2835 
2836 	if (base->state <= DASD_STATE_BASIC) {
2837 		DBF_DEV_EVENT(DBF_ERR, base, " %s",
2838 			      " Cannot open unrecognized device");
2839 		rc = -ENODEV;
2840 		goto out;
2841 	}
2842 
2843 	if ((mode & FMODE_WRITE) &&
2844 	    (test_bit(DASD_FLAG_DEVICE_RO, &base->flags) ||
2845 	     (base->features & DASD_FEATURE_READONLY))) {
2846 		rc = -EROFS;
2847 		goto out;
2848 	}
2849 
2850 	dasd_put_device(base);
2851 	return 0;
2852 
2853 out:
2854 	module_put(base->discipline->owner);
2855 unlock:
2856 	atomic_dec(&base->block->open_count);
2857 	dasd_put_device(base);
2858 	return rc;
2859 }
2860 
2861 static int dasd_release(struct gendisk *disk, fmode_t mode)
2862 {
2863 	struct dasd_device *base;
2864 
2865 	base = dasd_device_from_gendisk(disk);
2866 	if (!base)
2867 		return -ENODEV;
2868 
2869 	atomic_dec(&base->block->open_count);
2870 	module_put(base->discipline->owner);
2871 	dasd_put_device(base);
2872 	return 0;
2873 }
2874 
2875 /*
2876  * Return disk geometry.
2877  */
2878 static int dasd_getgeo(struct block_device *bdev, struct hd_geometry *geo)
2879 {
2880 	struct dasd_device *base;
2881 
2882 	base = dasd_device_from_gendisk(bdev->bd_disk);
2883 	if (!base)
2884 		return -ENODEV;
2885 
2886 	if (!base->discipline ||
2887 	    !base->discipline->fill_geometry) {
2888 		dasd_put_device(base);
2889 		return -EINVAL;
2890 	}
2891 	base->discipline->fill_geometry(base->block, geo);
2892 	geo->start = get_start_sect(bdev) >> base->block->s2b_shift;
2893 	dasd_put_device(base);
2894 	return 0;
2895 }
2896 
2897 const struct block_device_operations
2898 dasd_device_operations = {
2899 	.owner		= THIS_MODULE,
2900 	.open		= dasd_open,
2901 	.release	= dasd_release,
2902 	.ioctl		= dasd_ioctl,
2903 	.compat_ioctl	= dasd_ioctl,
2904 	.getgeo		= dasd_getgeo,
2905 };
2906 
2907 /*******************************************************************************
2908  * end of block device operations
2909  */
2910 
2911 static void
2912 dasd_exit(void)
2913 {
2914 #ifdef CONFIG_PROC_FS
2915 	dasd_proc_exit();
2916 #endif
2917 	dasd_eer_exit();
2918         if (dasd_page_cache != NULL) {
2919 		kmem_cache_destroy(dasd_page_cache);
2920 		dasd_page_cache = NULL;
2921 	}
2922 	dasd_gendisk_exit();
2923 	dasd_devmap_exit();
2924 	if (dasd_debug_area != NULL) {
2925 		debug_unregister(dasd_debug_area);
2926 		dasd_debug_area = NULL;
2927 	}
2928 	dasd_statistics_removeroot();
2929 }
2930 
2931 /*
2932  * SECTION: common functions for ccw_driver use
2933  */
2934 
2935 /*
2936  * Is the device read-only?
2937  * Note that this function does not report the setting of the
2938  * readonly device attribute, but how it is configured in z/VM.
2939  */
2940 int dasd_device_is_ro(struct dasd_device *device)
2941 {
2942 	struct ccw_dev_id dev_id;
2943 	struct diag210 diag_data;
2944 	int rc;
2945 
2946 	if (!MACHINE_IS_VM)
2947 		return 0;
2948 	ccw_device_get_id(device->cdev, &dev_id);
2949 	memset(&diag_data, 0, sizeof(diag_data));
2950 	diag_data.vrdcdvno = dev_id.devno;
2951 	diag_data.vrdclen = sizeof(diag_data);
2952 	rc = diag210(&diag_data);
2953 	if (rc == 0 || rc == 2) {
2954 		return diag_data.vrdcvfla & 0x80;
2955 	} else {
2956 		DBF_EVENT(DBF_WARNING, "diag210 failed for dev=%04x with rc=%d",
2957 			  dev_id.devno, rc);
2958 		return 0;
2959 	}
2960 }
2961 EXPORT_SYMBOL_GPL(dasd_device_is_ro);
2962 
2963 static void dasd_generic_auto_online(void *data, async_cookie_t cookie)
2964 {
2965 	struct ccw_device *cdev = data;
2966 	int ret;
2967 
2968 	ret = ccw_device_set_online(cdev);
2969 	if (ret)
2970 		pr_warning("%s: Setting the DASD online failed with rc=%d\n",
2971 			   dev_name(&cdev->dev), ret);
2972 }
2973 
2974 /*
2975  * Initial attempt at a probe function. this can be simplified once
2976  * the other detection code is gone.
2977  */
2978 int dasd_generic_probe(struct ccw_device *cdev,
2979 		       struct dasd_discipline *discipline)
2980 {
2981 	int ret;
2982 
2983 	ret = dasd_add_sysfs_files(cdev);
2984 	if (ret) {
2985 		DBF_EVENT_DEVID(DBF_WARNING, cdev, "%s",
2986 				"dasd_generic_probe: could not add "
2987 				"sysfs entries");
2988 		return ret;
2989 	}
2990 	cdev->handler = &dasd_int_handler;
2991 
2992 	/*
2993 	 * Automatically online either all dasd devices (dasd_autodetect)
2994 	 * or all devices specified with dasd= parameters during
2995 	 * initial probe.
2996 	 */
2997 	if ((dasd_get_feature(cdev, DASD_FEATURE_INITIAL_ONLINE) > 0 ) ||
2998 	    (dasd_autodetect && dasd_busid_known(dev_name(&cdev->dev)) != 0))
2999 		async_schedule(dasd_generic_auto_online, cdev);
3000 	return 0;
3001 }
3002 
3003 /*
3004  * This will one day be called from a global not_oper handler.
3005  * It is also used by driver_unregister during module unload.
3006  */
3007 void dasd_generic_remove(struct ccw_device *cdev)
3008 {
3009 	struct dasd_device *device;
3010 	struct dasd_block *block;
3011 
3012 	cdev->handler = NULL;
3013 
3014 	dasd_remove_sysfs_files(cdev);
3015 	device = dasd_device_from_cdev(cdev);
3016 	if (IS_ERR(device))
3017 		return;
3018 	if (test_and_set_bit(DASD_FLAG_OFFLINE, &device->flags)) {
3019 		/* Already doing offline processing */
3020 		dasd_put_device(device);
3021 		return;
3022 	}
3023 	/*
3024 	 * This device is removed unconditionally. Set offline
3025 	 * flag to prevent dasd_open from opening it while it is
3026 	 * no quite down yet.
3027 	 */
3028 	dasd_set_target_state(device, DASD_STATE_NEW);
3029 	/* dasd_delete_device destroys the device reference. */
3030 	block = device->block;
3031 	dasd_delete_device(device);
3032 	/*
3033 	 * life cycle of block is bound to device, so delete it after
3034 	 * device was safely removed
3035 	 */
3036 	if (block)
3037 		dasd_free_block(block);
3038 }
3039 
3040 /*
3041  * Activate a device. This is called from dasd_{eckd,fba}_probe() when either
3042  * the device is detected for the first time and is supposed to be used
3043  * or the user has started activation through sysfs.
3044  */
3045 int dasd_generic_set_online(struct ccw_device *cdev,
3046 			    struct dasd_discipline *base_discipline)
3047 {
3048 	struct dasd_discipline *discipline;
3049 	struct dasd_device *device;
3050 	int rc;
3051 
3052 	/* first online clears initial online feature flag */
3053 	dasd_set_feature(cdev, DASD_FEATURE_INITIAL_ONLINE, 0);
3054 	device = dasd_create_device(cdev);
3055 	if (IS_ERR(device))
3056 		return PTR_ERR(device);
3057 
3058 	discipline = base_discipline;
3059 	if (device->features & DASD_FEATURE_USEDIAG) {
3060 	  	if (!dasd_diag_discipline_pointer) {
3061 			pr_warning("%s Setting the DASD online failed because "
3062 				   "of missing DIAG discipline\n",
3063 				   dev_name(&cdev->dev));
3064 			dasd_delete_device(device);
3065 			return -ENODEV;
3066 		}
3067 		discipline = dasd_diag_discipline_pointer;
3068 	}
3069 	if (!try_module_get(base_discipline->owner)) {
3070 		dasd_delete_device(device);
3071 		return -EINVAL;
3072 	}
3073 	if (!try_module_get(discipline->owner)) {
3074 		module_put(base_discipline->owner);
3075 		dasd_delete_device(device);
3076 		return -EINVAL;
3077 	}
3078 	device->base_discipline = base_discipline;
3079 	device->discipline = discipline;
3080 
3081 	/* check_device will allocate block device if necessary */
3082 	rc = discipline->check_device(device);
3083 	if (rc) {
3084 		pr_warning("%s Setting the DASD online with discipline %s "
3085 			   "failed with rc=%i\n",
3086 			   dev_name(&cdev->dev), discipline->name, rc);
3087 		module_put(discipline->owner);
3088 		module_put(base_discipline->owner);
3089 		dasd_delete_device(device);
3090 		return rc;
3091 	}
3092 
3093 	dasd_set_target_state(device, DASD_STATE_ONLINE);
3094 	if (device->state <= DASD_STATE_KNOWN) {
3095 		pr_warning("%s Setting the DASD online failed because of a "
3096 			   "missing discipline\n", dev_name(&cdev->dev));
3097 		rc = -ENODEV;
3098 		dasd_set_target_state(device, DASD_STATE_NEW);
3099 		if (device->block)
3100 			dasd_free_block(device->block);
3101 		dasd_delete_device(device);
3102 	} else
3103 		pr_debug("dasd_generic device %s found\n",
3104 				dev_name(&cdev->dev));
3105 
3106 	wait_event(dasd_init_waitq, _wait_for_device(device));
3107 
3108 	dasd_put_device(device);
3109 	return rc;
3110 }
3111 
3112 int dasd_generic_set_offline(struct ccw_device *cdev)
3113 {
3114 	struct dasd_device *device;
3115 	struct dasd_block *block;
3116 	int max_count, open_count;
3117 
3118 	device = dasd_device_from_cdev(cdev);
3119 	if (IS_ERR(device))
3120 		return PTR_ERR(device);
3121 	if (test_and_set_bit(DASD_FLAG_OFFLINE, &device->flags)) {
3122 		/* Already doing offline processing */
3123 		dasd_put_device(device);
3124 		return 0;
3125 	}
3126 	/*
3127 	 * We must make sure that this device is currently not in use.
3128 	 * The open_count is increased for every opener, that includes
3129 	 * the blkdev_get in dasd_scan_partitions. We are only interested
3130 	 * in the other openers.
3131 	 */
3132 	if (device->block) {
3133 		max_count = device->block->bdev ? 0 : -1;
3134 		open_count = atomic_read(&device->block->open_count);
3135 		if (open_count > max_count) {
3136 			if (open_count > 0)
3137 				pr_warning("%s: The DASD cannot be set offline "
3138 					   "with open count %i\n",
3139 					   dev_name(&cdev->dev), open_count);
3140 			else
3141 				pr_warning("%s: The DASD cannot be set offline "
3142 					   "while it is in use\n",
3143 					   dev_name(&cdev->dev));
3144 			clear_bit(DASD_FLAG_OFFLINE, &device->flags);
3145 			dasd_put_device(device);
3146 			return -EBUSY;
3147 		}
3148 	}
3149 	dasd_set_target_state(device, DASD_STATE_NEW);
3150 	/* dasd_delete_device destroys the device reference. */
3151 	block = device->block;
3152 	dasd_delete_device(device);
3153 	/*
3154 	 * life cycle of block is bound to device, so delete it after
3155 	 * device was safely removed
3156 	 */
3157 	if (block)
3158 		dasd_free_block(block);
3159 	return 0;
3160 }
3161 
3162 int dasd_generic_last_path_gone(struct dasd_device *device)
3163 {
3164 	struct dasd_ccw_req *cqr;
3165 
3166 	dev_warn(&device->cdev->dev, "No operational channel path is left "
3167 		 "for the device\n");
3168 	DBF_DEV_EVENT(DBF_WARNING, device, "%s", "last path gone");
3169 	/* First of all call extended error reporting. */
3170 	dasd_eer_write(device, NULL, DASD_EER_NOPATH);
3171 
3172 	if (device->state < DASD_STATE_BASIC)
3173 		return 0;
3174 	/* Device is active. We want to keep it. */
3175 	list_for_each_entry(cqr, &device->ccw_queue, devlist)
3176 		if ((cqr->status == DASD_CQR_IN_IO) ||
3177 		    (cqr->status == DASD_CQR_CLEAR_PENDING)) {
3178 			cqr->status = DASD_CQR_QUEUED;
3179 			cqr->retries++;
3180 		}
3181 	dasd_device_set_stop_bits(device, DASD_STOPPED_DC_WAIT);
3182 	dasd_device_clear_timer(device);
3183 	dasd_schedule_device_bh(device);
3184 	return 1;
3185 }
3186 EXPORT_SYMBOL_GPL(dasd_generic_last_path_gone);
3187 
3188 int dasd_generic_path_operational(struct dasd_device *device)
3189 {
3190 	dev_info(&device->cdev->dev, "A channel path to the device has become "
3191 		 "operational\n");
3192 	DBF_DEV_EVENT(DBF_WARNING, device, "%s", "path operational");
3193 	dasd_device_remove_stop_bits(device, DASD_STOPPED_DC_WAIT);
3194 	if (device->stopped & DASD_UNRESUMED_PM) {
3195 		dasd_device_remove_stop_bits(device, DASD_UNRESUMED_PM);
3196 		dasd_restore_device(device);
3197 		return 1;
3198 	}
3199 	dasd_schedule_device_bh(device);
3200 	if (device->block)
3201 		dasd_schedule_block_bh(device->block);
3202 	return 1;
3203 }
3204 EXPORT_SYMBOL_GPL(dasd_generic_path_operational);
3205 
3206 int dasd_generic_notify(struct ccw_device *cdev, int event)
3207 {
3208 	struct dasd_device *device;
3209 	int ret;
3210 
3211 	device = dasd_device_from_cdev_locked(cdev);
3212 	if (IS_ERR(device))
3213 		return 0;
3214 	ret = 0;
3215 	switch (event) {
3216 	case CIO_GONE:
3217 	case CIO_BOXED:
3218 	case CIO_NO_PATH:
3219 		device->path_data.opm = 0;
3220 		device->path_data.ppm = 0;
3221 		device->path_data.npm = 0;
3222 		ret = dasd_generic_last_path_gone(device);
3223 		break;
3224 	case CIO_OPER:
3225 		ret = 1;
3226 		if (device->path_data.opm)
3227 			ret = dasd_generic_path_operational(device);
3228 		break;
3229 	}
3230 	dasd_put_device(device);
3231 	return ret;
3232 }
3233 
3234 void dasd_generic_path_event(struct ccw_device *cdev, int *path_event)
3235 {
3236 	int chp;
3237 	__u8 oldopm, eventlpm;
3238 	struct dasd_device *device;
3239 
3240 	device = dasd_device_from_cdev_locked(cdev);
3241 	if (IS_ERR(device))
3242 		return;
3243 	for (chp = 0; chp < 8; chp++) {
3244 		eventlpm = 0x80 >> chp;
3245 		if (path_event[chp] & PE_PATH_GONE) {
3246 			oldopm = device->path_data.opm;
3247 			device->path_data.opm &= ~eventlpm;
3248 			device->path_data.ppm &= ~eventlpm;
3249 			device->path_data.npm &= ~eventlpm;
3250 			if (oldopm && !device->path_data.opm)
3251 				dasd_generic_last_path_gone(device);
3252 		}
3253 		if (path_event[chp] & PE_PATH_AVAILABLE) {
3254 			device->path_data.opm &= ~eventlpm;
3255 			device->path_data.ppm &= ~eventlpm;
3256 			device->path_data.npm &= ~eventlpm;
3257 			device->path_data.tbvpm |= eventlpm;
3258 			dasd_schedule_device_bh(device);
3259 		}
3260 	}
3261 	dasd_put_device(device);
3262 }
3263 EXPORT_SYMBOL_GPL(dasd_generic_path_event);
3264 
3265 int dasd_generic_verify_path(struct dasd_device *device, __u8 lpm)
3266 {
3267 	if (!device->path_data.opm && lpm) {
3268 		device->path_data.opm = lpm;
3269 		dasd_generic_path_operational(device);
3270 	} else
3271 		device->path_data.opm |= lpm;
3272 	return 0;
3273 }
3274 EXPORT_SYMBOL_GPL(dasd_generic_verify_path);
3275 
3276 
3277 int dasd_generic_pm_freeze(struct ccw_device *cdev)
3278 {
3279 	struct dasd_ccw_req *cqr, *n;
3280 	int rc;
3281 	struct list_head freeze_queue;
3282 	struct dasd_device *device = dasd_device_from_cdev(cdev);
3283 
3284 	if (IS_ERR(device))
3285 		return PTR_ERR(device);
3286 
3287 	if (device->discipline->freeze)
3288 		rc = device->discipline->freeze(device);
3289 
3290 	/* disallow new I/O  */
3291 	dasd_device_set_stop_bits(device, DASD_STOPPED_PM);
3292 	/* clear active requests */
3293 	INIT_LIST_HEAD(&freeze_queue);
3294 	spin_lock_irq(get_ccwdev_lock(cdev));
3295 	rc = 0;
3296 	list_for_each_entry_safe(cqr, n, &device->ccw_queue, devlist) {
3297 		/* Check status and move request to flush_queue */
3298 		if (cqr->status == DASD_CQR_IN_IO) {
3299 			rc = device->discipline->term_IO(cqr);
3300 			if (rc) {
3301 				/* unable to terminate requeust */
3302 				dev_err(&device->cdev->dev,
3303 					"Unable to terminate request %p "
3304 					"on suspend\n", cqr);
3305 				spin_unlock_irq(get_ccwdev_lock(cdev));
3306 				dasd_put_device(device);
3307 				return rc;
3308 			}
3309 		}
3310 		list_move_tail(&cqr->devlist, &freeze_queue);
3311 	}
3312 
3313 	spin_unlock_irq(get_ccwdev_lock(cdev));
3314 
3315 	list_for_each_entry_safe(cqr, n, &freeze_queue, devlist) {
3316 		wait_event(dasd_flush_wq,
3317 			   (cqr->status != DASD_CQR_CLEAR_PENDING));
3318 		if (cqr->status == DASD_CQR_CLEARED)
3319 			cqr->status = DASD_CQR_QUEUED;
3320 	}
3321 	/* move freeze_queue to start of the ccw_queue */
3322 	spin_lock_irq(get_ccwdev_lock(cdev));
3323 	list_splice_tail(&freeze_queue, &device->ccw_queue);
3324 	spin_unlock_irq(get_ccwdev_lock(cdev));
3325 
3326 	dasd_put_device(device);
3327 	return rc;
3328 }
3329 EXPORT_SYMBOL_GPL(dasd_generic_pm_freeze);
3330 
3331 int dasd_generic_restore_device(struct ccw_device *cdev)
3332 {
3333 	struct dasd_device *device = dasd_device_from_cdev(cdev);
3334 	int rc = 0;
3335 
3336 	if (IS_ERR(device))
3337 		return PTR_ERR(device);
3338 
3339 	/* allow new IO again */
3340 	dasd_device_remove_stop_bits(device,
3341 				     (DASD_STOPPED_PM | DASD_UNRESUMED_PM));
3342 
3343 	dasd_schedule_device_bh(device);
3344 
3345 	/*
3346 	 * call discipline restore function
3347 	 * if device is stopped do nothing e.g. for disconnected devices
3348 	 */
3349 	if (device->discipline->restore && !(device->stopped))
3350 		rc = device->discipline->restore(device);
3351 	if (rc || device->stopped)
3352 		/*
3353 		 * if the resume failed for the DASD we put it in
3354 		 * an UNRESUMED stop state
3355 		 */
3356 		device->stopped |= DASD_UNRESUMED_PM;
3357 
3358 	if (device->block)
3359 		dasd_schedule_block_bh(device->block);
3360 
3361 	dasd_put_device(device);
3362 	return 0;
3363 }
3364 EXPORT_SYMBOL_GPL(dasd_generic_restore_device);
3365 
3366 static struct dasd_ccw_req *dasd_generic_build_rdc(struct dasd_device *device,
3367 						   void *rdc_buffer,
3368 						   int rdc_buffer_size,
3369 						   int magic)
3370 {
3371 	struct dasd_ccw_req *cqr;
3372 	struct ccw1 *ccw;
3373 	unsigned long *idaw;
3374 
3375 	cqr = dasd_smalloc_request(magic, 1 /* RDC */, rdc_buffer_size, device);
3376 
3377 	if (IS_ERR(cqr)) {
3378 		/* internal error 13 - Allocating the RDC request failed*/
3379 		dev_err(&device->cdev->dev,
3380 			 "An error occurred in the DASD device driver, "
3381 			 "reason=%s\n", "13");
3382 		return cqr;
3383 	}
3384 
3385 	ccw = cqr->cpaddr;
3386 	ccw->cmd_code = CCW_CMD_RDC;
3387 	if (idal_is_needed(rdc_buffer, rdc_buffer_size)) {
3388 		idaw = (unsigned long *) (cqr->data);
3389 		ccw->cda = (__u32)(addr_t) idaw;
3390 		ccw->flags = CCW_FLAG_IDA;
3391 		idaw = idal_create_words(idaw, rdc_buffer, rdc_buffer_size);
3392 	} else {
3393 		ccw->cda = (__u32)(addr_t) rdc_buffer;
3394 		ccw->flags = 0;
3395 	}
3396 
3397 	ccw->count = rdc_buffer_size;
3398 	cqr->startdev = device;
3399 	cqr->memdev = device;
3400 	cqr->expires = 10*HZ;
3401 	cqr->retries = 256;
3402 	cqr->buildclk = get_clock();
3403 	cqr->status = DASD_CQR_FILLED;
3404 	return cqr;
3405 }
3406 
3407 
3408 int dasd_generic_read_dev_chars(struct dasd_device *device, int magic,
3409 				void *rdc_buffer, int rdc_buffer_size)
3410 {
3411 	int ret;
3412 	struct dasd_ccw_req *cqr;
3413 
3414 	cqr = dasd_generic_build_rdc(device, rdc_buffer, rdc_buffer_size,
3415 				     magic);
3416 	if (IS_ERR(cqr))
3417 		return PTR_ERR(cqr);
3418 
3419 	ret = dasd_sleep_on(cqr);
3420 	dasd_sfree_request(cqr, cqr->memdev);
3421 	return ret;
3422 }
3423 EXPORT_SYMBOL_GPL(dasd_generic_read_dev_chars);
3424 
3425 /*
3426  *   In command mode and transport mode we need to look for sense
3427  *   data in different places. The sense data itself is allways
3428  *   an array of 32 bytes, so we can unify the sense data access
3429  *   for both modes.
3430  */
3431 char *dasd_get_sense(struct irb *irb)
3432 {
3433 	struct tsb *tsb = NULL;
3434 	char *sense = NULL;
3435 
3436 	if (scsw_is_tm(&irb->scsw) && (irb->scsw.tm.fcxs == 0x01)) {
3437 		if (irb->scsw.tm.tcw)
3438 			tsb = tcw_get_tsb((struct tcw *)(unsigned long)
3439 					  irb->scsw.tm.tcw);
3440 		if (tsb && tsb->length == 64 && tsb->flags)
3441 			switch (tsb->flags & 0x07) {
3442 			case 1:	/* tsa_iostat */
3443 				sense = tsb->tsa.iostat.sense;
3444 				break;
3445 			case 2: /* tsa_ddpc */
3446 				sense = tsb->tsa.ddpc.sense;
3447 				break;
3448 			default:
3449 				/* currently we don't use interrogate data */
3450 				break;
3451 			}
3452 	} else if (irb->esw.esw0.erw.cons) {
3453 		sense = irb->ecw;
3454 	}
3455 	return sense;
3456 }
3457 EXPORT_SYMBOL_GPL(dasd_get_sense);
3458 
3459 static int __init dasd_init(void)
3460 {
3461 	int rc;
3462 
3463 	init_waitqueue_head(&dasd_init_waitq);
3464 	init_waitqueue_head(&dasd_flush_wq);
3465 	init_waitqueue_head(&generic_waitq);
3466 
3467 	/* register 'common' DASD debug area, used for all DBF_XXX calls */
3468 	dasd_debug_area = debug_register("dasd", 1, 1, 8 * sizeof(long));
3469 	if (dasd_debug_area == NULL) {
3470 		rc = -ENOMEM;
3471 		goto failed;
3472 	}
3473 	debug_register_view(dasd_debug_area, &debug_sprintf_view);
3474 	debug_set_level(dasd_debug_area, DBF_WARNING);
3475 
3476 	DBF_EVENT(DBF_EMERG, "%s", "debug area created");
3477 
3478 	dasd_diag_discipline_pointer = NULL;
3479 
3480 	dasd_statistics_createroot();
3481 
3482 	rc = dasd_devmap_init();
3483 	if (rc)
3484 		goto failed;
3485 	rc = dasd_gendisk_init();
3486 	if (rc)
3487 		goto failed;
3488 	rc = dasd_parse();
3489 	if (rc)
3490 		goto failed;
3491 	rc = dasd_eer_init();
3492 	if (rc)
3493 		goto failed;
3494 #ifdef CONFIG_PROC_FS
3495 	rc = dasd_proc_init();
3496 	if (rc)
3497 		goto failed;
3498 #endif
3499 
3500 	return 0;
3501 failed:
3502 	pr_info("The DASD device driver could not be initialized\n");
3503 	dasd_exit();
3504 	return rc;
3505 }
3506 
3507 module_init(dasd_init);
3508 module_exit(dasd_exit);
3509 
3510 EXPORT_SYMBOL(dasd_debug_area);
3511 EXPORT_SYMBOL(dasd_diag_discipline_pointer);
3512 
3513 EXPORT_SYMBOL(dasd_add_request_head);
3514 EXPORT_SYMBOL(dasd_add_request_tail);
3515 EXPORT_SYMBOL(dasd_cancel_req);
3516 EXPORT_SYMBOL(dasd_device_clear_timer);
3517 EXPORT_SYMBOL(dasd_block_clear_timer);
3518 EXPORT_SYMBOL(dasd_enable_device);
3519 EXPORT_SYMBOL(dasd_int_handler);
3520 EXPORT_SYMBOL(dasd_kfree_request);
3521 EXPORT_SYMBOL(dasd_kick_device);
3522 EXPORT_SYMBOL(dasd_kmalloc_request);
3523 EXPORT_SYMBOL(dasd_schedule_device_bh);
3524 EXPORT_SYMBOL(dasd_schedule_block_bh);
3525 EXPORT_SYMBOL(dasd_set_target_state);
3526 EXPORT_SYMBOL(dasd_device_set_timer);
3527 EXPORT_SYMBOL(dasd_block_set_timer);
3528 EXPORT_SYMBOL(dasd_sfree_request);
3529 EXPORT_SYMBOL(dasd_sleep_on);
3530 EXPORT_SYMBOL(dasd_sleep_on_immediatly);
3531 EXPORT_SYMBOL(dasd_sleep_on_interruptible);
3532 EXPORT_SYMBOL(dasd_smalloc_request);
3533 EXPORT_SYMBOL(dasd_start_IO);
3534 EXPORT_SYMBOL(dasd_term_IO);
3535 
3536 EXPORT_SYMBOL_GPL(dasd_generic_probe);
3537 EXPORT_SYMBOL_GPL(dasd_generic_remove);
3538 EXPORT_SYMBOL_GPL(dasd_generic_notify);
3539 EXPORT_SYMBOL_GPL(dasd_generic_set_online);
3540 EXPORT_SYMBOL_GPL(dasd_generic_set_offline);
3541 EXPORT_SYMBOL_GPL(dasd_generic_handle_state_change);
3542 EXPORT_SYMBOL_GPL(dasd_flush_device_queue);
3543 EXPORT_SYMBOL_GPL(dasd_alloc_block);
3544 EXPORT_SYMBOL_GPL(dasd_free_block);
3545