xref: /linux/drivers/md/dm-raid1.c (revision 7034228792cc561e79ff8600f02884bd4c80e287)
1 /*
2  * Copyright (C) 2003 Sistina Software Limited.
3  * Copyright (C) 2005-2008 Red Hat, Inc. All rights reserved.
4  *
5  * This file is released under the GPL.
6  */
7 
8 #include "dm-bio-record.h"
9 
10 #include <linux/init.h>
11 #include <linux/mempool.h>
12 #include <linux/module.h>
13 #include <linux/pagemap.h>
14 #include <linux/slab.h>
15 #include <linux/workqueue.h>
16 #include <linux/device-mapper.h>
17 #include <linux/dm-io.h>
18 #include <linux/dm-dirty-log.h>
19 #include <linux/dm-kcopyd.h>
20 #include <linux/dm-region-hash.h>
21 
22 #define DM_MSG_PREFIX "raid1"
23 
24 #define MAX_RECOVERY 1	/* Maximum number of regions recovered in parallel. */
25 
26 #define DM_RAID1_HANDLE_ERRORS 0x01
27 #define errors_handled(p)	((p)->features & DM_RAID1_HANDLE_ERRORS)
28 
29 static DECLARE_WAIT_QUEUE_HEAD(_kmirrord_recovery_stopped);
30 
31 /*-----------------------------------------------------------------
32  * Mirror set structures.
33  *---------------------------------------------------------------*/
34 enum dm_raid1_error {
35 	DM_RAID1_WRITE_ERROR,
36 	DM_RAID1_FLUSH_ERROR,
37 	DM_RAID1_SYNC_ERROR,
38 	DM_RAID1_READ_ERROR
39 };
40 
41 struct mirror {
42 	struct mirror_set *ms;
43 	atomic_t error_count;
44 	unsigned long error_type;
45 	struct dm_dev *dev;
46 	sector_t offset;
47 };
48 
49 struct mirror_set {
50 	struct dm_target *ti;
51 	struct list_head list;
52 
53 	uint64_t features;
54 
55 	spinlock_t lock;	/* protects the lists */
56 	struct bio_list reads;
57 	struct bio_list writes;
58 	struct bio_list failures;
59 	struct bio_list holds;	/* bios are waiting until suspend */
60 
61 	struct dm_region_hash *rh;
62 	struct dm_kcopyd_client *kcopyd_client;
63 	struct dm_io_client *io_client;
64 
65 	/* recovery */
66 	region_t nr_regions;
67 	int in_sync;
68 	int log_failure;
69 	int leg_failure;
70 	atomic_t suspend;
71 
72 	atomic_t default_mirror;	/* Default mirror */
73 
74 	struct workqueue_struct *kmirrord_wq;
75 	struct work_struct kmirrord_work;
76 	struct timer_list timer;
77 	unsigned long timer_pending;
78 
79 	struct work_struct trigger_event;
80 
81 	unsigned nr_mirrors;
82 	struct mirror mirror[0];
83 };
84 
85 static void wakeup_mirrord(void *context)
86 {
87 	struct mirror_set *ms = context;
88 
89 	queue_work(ms->kmirrord_wq, &ms->kmirrord_work);
90 }
91 
92 static void delayed_wake_fn(unsigned long data)
93 {
94 	struct mirror_set *ms = (struct mirror_set *) data;
95 
96 	clear_bit(0, &ms->timer_pending);
97 	wakeup_mirrord(ms);
98 }
99 
100 static void delayed_wake(struct mirror_set *ms)
101 {
102 	if (test_and_set_bit(0, &ms->timer_pending))
103 		return;
104 
105 	ms->timer.expires = jiffies + HZ / 5;
106 	ms->timer.data = (unsigned long) ms;
107 	ms->timer.function = delayed_wake_fn;
108 	add_timer(&ms->timer);
109 }
110 
111 static void wakeup_all_recovery_waiters(void *context)
112 {
113 	wake_up_all(&_kmirrord_recovery_stopped);
114 }
115 
116 static void queue_bio(struct mirror_set *ms, struct bio *bio, int rw)
117 {
118 	unsigned long flags;
119 	int should_wake = 0;
120 	struct bio_list *bl;
121 
122 	bl = (rw == WRITE) ? &ms->writes : &ms->reads;
123 	spin_lock_irqsave(&ms->lock, flags);
124 	should_wake = !(bl->head);
125 	bio_list_add(bl, bio);
126 	spin_unlock_irqrestore(&ms->lock, flags);
127 
128 	if (should_wake)
129 		wakeup_mirrord(ms);
130 }
131 
132 static void dispatch_bios(void *context, struct bio_list *bio_list)
133 {
134 	struct mirror_set *ms = context;
135 	struct bio *bio;
136 
137 	while ((bio = bio_list_pop(bio_list)))
138 		queue_bio(ms, bio, WRITE);
139 }
140 
141 struct dm_raid1_bio_record {
142 	struct mirror *m;
143 	/* if details->bi_bdev == NULL, details were not saved */
144 	struct dm_bio_details details;
145 	region_t write_region;
146 };
147 
148 /*
149  * Every mirror should look like this one.
150  */
151 #define DEFAULT_MIRROR 0
152 
153 /*
154  * This is yucky.  We squirrel the mirror struct away inside
155  * bi_next for read/write buffers.  This is safe since the bh
156  * doesn't get submitted to the lower levels of block layer.
157  */
158 static struct mirror *bio_get_m(struct bio *bio)
159 {
160 	return (struct mirror *) bio->bi_next;
161 }
162 
163 static void bio_set_m(struct bio *bio, struct mirror *m)
164 {
165 	bio->bi_next = (struct bio *) m;
166 }
167 
168 static struct mirror *get_default_mirror(struct mirror_set *ms)
169 {
170 	return &ms->mirror[atomic_read(&ms->default_mirror)];
171 }
172 
173 static void set_default_mirror(struct mirror *m)
174 {
175 	struct mirror_set *ms = m->ms;
176 	struct mirror *m0 = &(ms->mirror[0]);
177 
178 	atomic_set(&ms->default_mirror, m - m0);
179 }
180 
181 static struct mirror *get_valid_mirror(struct mirror_set *ms)
182 {
183 	struct mirror *m;
184 
185 	for (m = ms->mirror; m < ms->mirror + ms->nr_mirrors; m++)
186 		if (!atomic_read(&m->error_count))
187 			return m;
188 
189 	return NULL;
190 }
191 
192 /* fail_mirror
193  * @m: mirror device to fail
194  * @error_type: one of the enum's, DM_RAID1_*_ERROR
195  *
196  * If errors are being handled, record the type of
197  * error encountered for this device.  If this type
198  * of error has already been recorded, we can return;
199  * otherwise, we must signal userspace by triggering
200  * an event.  Additionally, if the device is the
201  * primary device, we must choose a new primary, but
202  * only if the mirror is in-sync.
203  *
204  * This function must not block.
205  */
206 static void fail_mirror(struct mirror *m, enum dm_raid1_error error_type)
207 {
208 	struct mirror_set *ms = m->ms;
209 	struct mirror *new;
210 
211 	ms->leg_failure = 1;
212 
213 	/*
214 	 * error_count is used for nothing more than a
215 	 * simple way to tell if a device has encountered
216 	 * errors.
217 	 */
218 	atomic_inc(&m->error_count);
219 
220 	if (test_and_set_bit(error_type, &m->error_type))
221 		return;
222 
223 	if (!errors_handled(ms))
224 		return;
225 
226 	if (m != get_default_mirror(ms))
227 		goto out;
228 
229 	if (!ms->in_sync) {
230 		/*
231 		 * Better to issue requests to same failing device
232 		 * than to risk returning corrupt data.
233 		 */
234 		DMERR("Primary mirror (%s) failed while out-of-sync: "
235 		      "Reads may fail.", m->dev->name);
236 		goto out;
237 	}
238 
239 	new = get_valid_mirror(ms);
240 	if (new)
241 		set_default_mirror(new);
242 	else
243 		DMWARN("All sides of mirror have failed.");
244 
245 out:
246 	schedule_work(&ms->trigger_event);
247 }
248 
249 static int mirror_flush(struct dm_target *ti)
250 {
251 	struct mirror_set *ms = ti->private;
252 	unsigned long error_bits;
253 
254 	unsigned int i;
255 	struct dm_io_region io[ms->nr_mirrors];
256 	struct mirror *m;
257 	struct dm_io_request io_req = {
258 		.bi_rw = WRITE_FLUSH,
259 		.mem.type = DM_IO_KMEM,
260 		.mem.ptr.addr = NULL,
261 		.client = ms->io_client,
262 	};
263 
264 	for (i = 0, m = ms->mirror; i < ms->nr_mirrors; i++, m++) {
265 		io[i].bdev = m->dev->bdev;
266 		io[i].sector = 0;
267 		io[i].count = 0;
268 	}
269 
270 	error_bits = -1;
271 	dm_io(&io_req, ms->nr_mirrors, io, &error_bits);
272 	if (unlikely(error_bits != 0)) {
273 		for (i = 0; i < ms->nr_mirrors; i++)
274 			if (test_bit(i, &error_bits))
275 				fail_mirror(ms->mirror + i,
276 					    DM_RAID1_FLUSH_ERROR);
277 		return -EIO;
278 	}
279 
280 	return 0;
281 }
282 
283 /*-----------------------------------------------------------------
284  * Recovery.
285  *
286  * When a mirror is first activated we may find that some regions
287  * are in the no-sync state.  We have to recover these by
288  * recopying from the default mirror to all the others.
289  *---------------------------------------------------------------*/
290 static void recovery_complete(int read_err, unsigned long write_err,
291 			      void *context)
292 {
293 	struct dm_region *reg = context;
294 	struct mirror_set *ms = dm_rh_region_context(reg);
295 	int m, bit = 0;
296 
297 	if (read_err) {
298 		/* Read error means the failure of default mirror. */
299 		DMERR_LIMIT("Unable to read primary mirror during recovery");
300 		fail_mirror(get_default_mirror(ms), DM_RAID1_SYNC_ERROR);
301 	}
302 
303 	if (write_err) {
304 		DMERR_LIMIT("Write error during recovery (error = 0x%lx)",
305 			    write_err);
306 		/*
307 		 * Bits correspond to devices (excluding default mirror).
308 		 * The default mirror cannot change during recovery.
309 		 */
310 		for (m = 0; m < ms->nr_mirrors; m++) {
311 			if (&ms->mirror[m] == get_default_mirror(ms))
312 				continue;
313 			if (test_bit(bit, &write_err))
314 				fail_mirror(ms->mirror + m,
315 					    DM_RAID1_SYNC_ERROR);
316 			bit++;
317 		}
318 	}
319 
320 	dm_rh_recovery_end(reg, !(read_err || write_err));
321 }
322 
323 static int recover(struct mirror_set *ms, struct dm_region *reg)
324 {
325 	int r;
326 	unsigned i;
327 	struct dm_io_region from, to[DM_KCOPYD_MAX_REGIONS], *dest;
328 	struct mirror *m;
329 	unsigned long flags = 0;
330 	region_t key = dm_rh_get_region_key(reg);
331 	sector_t region_size = dm_rh_get_region_size(ms->rh);
332 
333 	/* fill in the source */
334 	m = get_default_mirror(ms);
335 	from.bdev = m->dev->bdev;
336 	from.sector = m->offset + dm_rh_region_to_sector(ms->rh, key);
337 	if (key == (ms->nr_regions - 1)) {
338 		/*
339 		 * The final region may be smaller than
340 		 * region_size.
341 		 */
342 		from.count = ms->ti->len & (region_size - 1);
343 		if (!from.count)
344 			from.count = region_size;
345 	} else
346 		from.count = region_size;
347 
348 	/* fill in the destinations */
349 	for (i = 0, dest = to; i < ms->nr_mirrors; i++) {
350 		if (&ms->mirror[i] == get_default_mirror(ms))
351 			continue;
352 
353 		m = ms->mirror + i;
354 		dest->bdev = m->dev->bdev;
355 		dest->sector = m->offset + dm_rh_region_to_sector(ms->rh, key);
356 		dest->count = from.count;
357 		dest++;
358 	}
359 
360 	/* hand to kcopyd */
361 	if (!errors_handled(ms))
362 		set_bit(DM_KCOPYD_IGNORE_ERROR, &flags);
363 
364 	r = dm_kcopyd_copy(ms->kcopyd_client, &from, ms->nr_mirrors - 1, to,
365 			   flags, recovery_complete, reg);
366 
367 	return r;
368 }
369 
370 static void do_recovery(struct mirror_set *ms)
371 {
372 	struct dm_region *reg;
373 	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);
374 	int r;
375 
376 	/*
377 	 * Start quiescing some regions.
378 	 */
379 	dm_rh_recovery_prepare(ms->rh);
380 
381 	/*
382 	 * Copy any already quiesced regions.
383 	 */
384 	while ((reg = dm_rh_recovery_start(ms->rh))) {
385 		r = recover(ms, reg);
386 		if (r)
387 			dm_rh_recovery_end(reg, 0);
388 	}
389 
390 	/*
391 	 * Update the in sync flag.
392 	 */
393 	if (!ms->in_sync &&
394 	    (log->type->get_sync_count(log) == ms->nr_regions)) {
395 		/* the sync is complete */
396 		dm_table_event(ms->ti->table);
397 		ms->in_sync = 1;
398 	}
399 }
400 
401 /*-----------------------------------------------------------------
402  * Reads
403  *---------------------------------------------------------------*/
404 static struct mirror *choose_mirror(struct mirror_set *ms, sector_t sector)
405 {
406 	struct mirror *m = get_default_mirror(ms);
407 
408 	do {
409 		if (likely(!atomic_read(&m->error_count)))
410 			return m;
411 
412 		if (m-- == ms->mirror)
413 			m += ms->nr_mirrors;
414 	} while (m != get_default_mirror(ms));
415 
416 	return NULL;
417 }
418 
419 static int default_ok(struct mirror *m)
420 {
421 	struct mirror *default_mirror = get_default_mirror(m->ms);
422 
423 	return !atomic_read(&default_mirror->error_count);
424 }
425 
426 static int mirror_available(struct mirror_set *ms, struct bio *bio)
427 {
428 	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);
429 	region_t region = dm_rh_bio_to_region(ms->rh, bio);
430 
431 	if (log->type->in_sync(log, region, 0))
432 		return choose_mirror(ms,  bio->bi_sector) ? 1 : 0;
433 
434 	return 0;
435 }
436 
437 /*
438  * remap a buffer to a particular mirror.
439  */
440 static sector_t map_sector(struct mirror *m, struct bio *bio)
441 {
442 	if (unlikely(!bio->bi_size))
443 		return 0;
444 	return m->offset + dm_target_offset(m->ms->ti, bio->bi_sector);
445 }
446 
447 static void map_bio(struct mirror *m, struct bio *bio)
448 {
449 	bio->bi_bdev = m->dev->bdev;
450 	bio->bi_sector = map_sector(m, bio);
451 }
452 
453 static void map_region(struct dm_io_region *io, struct mirror *m,
454 		       struct bio *bio)
455 {
456 	io->bdev = m->dev->bdev;
457 	io->sector = map_sector(m, bio);
458 	io->count = bio->bi_size >> 9;
459 }
460 
461 static void hold_bio(struct mirror_set *ms, struct bio *bio)
462 {
463 	/*
464 	 * Lock is required to avoid race condition during suspend
465 	 * process.
466 	 */
467 	spin_lock_irq(&ms->lock);
468 
469 	if (atomic_read(&ms->suspend)) {
470 		spin_unlock_irq(&ms->lock);
471 
472 		/*
473 		 * If device is suspended, complete the bio.
474 		 */
475 		if (dm_noflush_suspending(ms->ti))
476 			bio_endio(bio, DM_ENDIO_REQUEUE);
477 		else
478 			bio_endio(bio, -EIO);
479 		return;
480 	}
481 
482 	/*
483 	 * Hold bio until the suspend is complete.
484 	 */
485 	bio_list_add(&ms->holds, bio);
486 	spin_unlock_irq(&ms->lock);
487 }
488 
489 /*-----------------------------------------------------------------
490  * Reads
491  *---------------------------------------------------------------*/
492 static void read_callback(unsigned long error, void *context)
493 {
494 	struct bio *bio = context;
495 	struct mirror *m;
496 
497 	m = bio_get_m(bio);
498 	bio_set_m(bio, NULL);
499 
500 	if (likely(!error)) {
501 		bio_endio(bio, 0);
502 		return;
503 	}
504 
505 	fail_mirror(m, DM_RAID1_READ_ERROR);
506 
507 	if (likely(default_ok(m)) || mirror_available(m->ms, bio)) {
508 		DMWARN_LIMIT("Read failure on mirror device %s.  "
509 			     "Trying alternative device.",
510 			     m->dev->name);
511 		queue_bio(m->ms, bio, bio_rw(bio));
512 		return;
513 	}
514 
515 	DMERR_LIMIT("Read failure on mirror device %s.  Failing I/O.",
516 		    m->dev->name);
517 	bio_endio(bio, -EIO);
518 }
519 
520 /* Asynchronous read. */
521 static void read_async_bio(struct mirror *m, struct bio *bio)
522 {
523 	struct dm_io_region io;
524 	struct dm_io_request io_req = {
525 		.bi_rw = READ,
526 		.mem.type = DM_IO_BVEC,
527 		.mem.ptr.bvec = bio->bi_io_vec + bio->bi_idx,
528 		.notify.fn = read_callback,
529 		.notify.context = bio,
530 		.client = m->ms->io_client,
531 	};
532 
533 	map_region(&io, m, bio);
534 	bio_set_m(bio, m);
535 	BUG_ON(dm_io(&io_req, 1, &io, NULL));
536 }
537 
538 static inline int region_in_sync(struct mirror_set *ms, region_t region,
539 				 int may_block)
540 {
541 	int state = dm_rh_get_state(ms->rh, region, may_block);
542 	return state == DM_RH_CLEAN || state == DM_RH_DIRTY;
543 }
544 
545 static void do_reads(struct mirror_set *ms, struct bio_list *reads)
546 {
547 	region_t region;
548 	struct bio *bio;
549 	struct mirror *m;
550 
551 	while ((bio = bio_list_pop(reads))) {
552 		region = dm_rh_bio_to_region(ms->rh, bio);
553 		m = get_default_mirror(ms);
554 
555 		/*
556 		 * We can only read balance if the region is in sync.
557 		 */
558 		if (likely(region_in_sync(ms, region, 1)))
559 			m = choose_mirror(ms, bio->bi_sector);
560 		else if (m && atomic_read(&m->error_count))
561 			m = NULL;
562 
563 		if (likely(m))
564 			read_async_bio(m, bio);
565 		else
566 			bio_endio(bio, -EIO);
567 	}
568 }
569 
570 /*-----------------------------------------------------------------
571  * Writes.
572  *
573  * We do different things with the write io depending on the
574  * state of the region that it's in:
575  *
576  * SYNC: 	increment pending, use kcopyd to write to *all* mirrors
577  * RECOVERING:	delay the io until recovery completes
578  * NOSYNC:	increment pending, just write to the default mirror
579  *---------------------------------------------------------------*/
580 
581 
582 static void write_callback(unsigned long error, void *context)
583 {
584 	unsigned i, ret = 0;
585 	struct bio *bio = (struct bio *) context;
586 	struct mirror_set *ms;
587 	int should_wake = 0;
588 	unsigned long flags;
589 
590 	ms = bio_get_m(bio)->ms;
591 	bio_set_m(bio, NULL);
592 
593 	/*
594 	 * NOTE: We don't decrement the pending count here,
595 	 * instead it is done by the targets endio function.
596 	 * This way we handle both writes to SYNC and NOSYNC
597 	 * regions with the same code.
598 	 */
599 	if (likely(!error)) {
600 		bio_endio(bio, ret);
601 		return;
602 	}
603 
604 	for (i = 0; i < ms->nr_mirrors; i++)
605 		if (test_bit(i, &error))
606 			fail_mirror(ms->mirror + i, DM_RAID1_WRITE_ERROR);
607 
608 	/*
609 	 * Need to raise event.  Since raising
610 	 * events can block, we need to do it in
611 	 * the main thread.
612 	 */
613 	spin_lock_irqsave(&ms->lock, flags);
614 	if (!ms->failures.head)
615 		should_wake = 1;
616 	bio_list_add(&ms->failures, bio);
617 	spin_unlock_irqrestore(&ms->lock, flags);
618 	if (should_wake)
619 		wakeup_mirrord(ms);
620 }
621 
622 static void do_write(struct mirror_set *ms, struct bio *bio)
623 {
624 	unsigned int i;
625 	struct dm_io_region io[ms->nr_mirrors], *dest = io;
626 	struct mirror *m;
627 	struct dm_io_request io_req = {
628 		.bi_rw = WRITE | (bio->bi_rw & WRITE_FLUSH_FUA),
629 		.mem.type = DM_IO_BVEC,
630 		.mem.ptr.bvec = bio->bi_io_vec + bio->bi_idx,
631 		.notify.fn = write_callback,
632 		.notify.context = bio,
633 		.client = ms->io_client,
634 	};
635 
636 	if (bio->bi_rw & REQ_DISCARD) {
637 		io_req.bi_rw |= REQ_DISCARD;
638 		io_req.mem.type = DM_IO_KMEM;
639 		io_req.mem.ptr.addr = NULL;
640 	}
641 
642 	for (i = 0, m = ms->mirror; i < ms->nr_mirrors; i++, m++)
643 		map_region(dest++, m, bio);
644 
645 	/*
646 	 * Use default mirror because we only need it to retrieve the reference
647 	 * to the mirror set in write_callback().
648 	 */
649 	bio_set_m(bio, get_default_mirror(ms));
650 
651 	BUG_ON(dm_io(&io_req, ms->nr_mirrors, io, NULL));
652 }
653 
654 static void do_writes(struct mirror_set *ms, struct bio_list *writes)
655 {
656 	int state;
657 	struct bio *bio;
658 	struct bio_list sync, nosync, recover, *this_list = NULL;
659 	struct bio_list requeue;
660 	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);
661 	region_t region;
662 
663 	if (!writes->head)
664 		return;
665 
666 	/*
667 	 * Classify each write.
668 	 */
669 	bio_list_init(&sync);
670 	bio_list_init(&nosync);
671 	bio_list_init(&recover);
672 	bio_list_init(&requeue);
673 
674 	while ((bio = bio_list_pop(writes))) {
675 		if ((bio->bi_rw & REQ_FLUSH) ||
676 		    (bio->bi_rw & REQ_DISCARD)) {
677 			bio_list_add(&sync, bio);
678 			continue;
679 		}
680 
681 		region = dm_rh_bio_to_region(ms->rh, bio);
682 
683 		if (log->type->is_remote_recovering &&
684 		    log->type->is_remote_recovering(log, region)) {
685 			bio_list_add(&requeue, bio);
686 			continue;
687 		}
688 
689 		state = dm_rh_get_state(ms->rh, region, 1);
690 		switch (state) {
691 		case DM_RH_CLEAN:
692 		case DM_RH_DIRTY:
693 			this_list = &sync;
694 			break;
695 
696 		case DM_RH_NOSYNC:
697 			this_list = &nosync;
698 			break;
699 
700 		case DM_RH_RECOVERING:
701 			this_list = &recover;
702 			break;
703 		}
704 
705 		bio_list_add(this_list, bio);
706 	}
707 
708 	/*
709 	 * Add bios that are delayed due to remote recovery
710 	 * back on to the write queue
711 	 */
712 	if (unlikely(requeue.head)) {
713 		spin_lock_irq(&ms->lock);
714 		bio_list_merge(&ms->writes, &requeue);
715 		spin_unlock_irq(&ms->lock);
716 		delayed_wake(ms);
717 	}
718 
719 	/*
720 	 * Increment the pending counts for any regions that will
721 	 * be written to (writes to recover regions are going to
722 	 * be delayed).
723 	 */
724 	dm_rh_inc_pending(ms->rh, &sync);
725 	dm_rh_inc_pending(ms->rh, &nosync);
726 
727 	/*
728 	 * If the flush fails on a previous call and succeeds here,
729 	 * we must not reset the log_failure variable.  We need
730 	 * userspace interaction to do that.
731 	 */
732 	ms->log_failure = dm_rh_flush(ms->rh) ? 1 : ms->log_failure;
733 
734 	/*
735 	 * Dispatch io.
736 	 */
737 	if (unlikely(ms->log_failure) && errors_handled(ms)) {
738 		spin_lock_irq(&ms->lock);
739 		bio_list_merge(&ms->failures, &sync);
740 		spin_unlock_irq(&ms->lock);
741 		wakeup_mirrord(ms);
742 	} else
743 		while ((bio = bio_list_pop(&sync)))
744 			do_write(ms, bio);
745 
746 	while ((bio = bio_list_pop(&recover)))
747 		dm_rh_delay(ms->rh, bio);
748 
749 	while ((bio = bio_list_pop(&nosync))) {
750 		if (unlikely(ms->leg_failure) && errors_handled(ms)) {
751 			spin_lock_irq(&ms->lock);
752 			bio_list_add(&ms->failures, bio);
753 			spin_unlock_irq(&ms->lock);
754 			wakeup_mirrord(ms);
755 		} else {
756 			map_bio(get_default_mirror(ms), bio);
757 			generic_make_request(bio);
758 		}
759 	}
760 }
761 
762 static void do_failures(struct mirror_set *ms, struct bio_list *failures)
763 {
764 	struct bio *bio;
765 
766 	if (likely(!failures->head))
767 		return;
768 
769 	/*
770 	 * If the log has failed, unattempted writes are being
771 	 * put on the holds list.  We can't issue those writes
772 	 * until a log has been marked, so we must store them.
773 	 *
774 	 * If a 'noflush' suspend is in progress, we can requeue
775 	 * the I/O's to the core.  This give userspace a chance
776 	 * to reconfigure the mirror, at which point the core
777 	 * will reissue the writes.  If the 'noflush' flag is
778 	 * not set, we have no choice but to return errors.
779 	 *
780 	 * Some writes on the failures list may have been
781 	 * submitted before the log failure and represent a
782 	 * failure to write to one of the devices.  It is ok
783 	 * for us to treat them the same and requeue them
784 	 * as well.
785 	 */
786 	while ((bio = bio_list_pop(failures))) {
787 		if (!ms->log_failure) {
788 			ms->in_sync = 0;
789 			dm_rh_mark_nosync(ms->rh, bio);
790 		}
791 
792 		/*
793 		 * If all the legs are dead, fail the I/O.
794 		 * If we have been told to handle errors, hold the bio
795 		 * and wait for userspace to deal with the problem.
796 		 * Otherwise pretend that the I/O succeeded. (This would
797 		 * be wrong if the failed leg returned after reboot and
798 		 * got replicated back to the good legs.)
799 		 */
800 		if (!get_valid_mirror(ms))
801 			bio_endio(bio, -EIO);
802 		else if (errors_handled(ms))
803 			hold_bio(ms, bio);
804 		else
805 			bio_endio(bio, 0);
806 	}
807 }
808 
809 static void trigger_event(struct work_struct *work)
810 {
811 	struct mirror_set *ms =
812 		container_of(work, struct mirror_set, trigger_event);
813 
814 	dm_table_event(ms->ti->table);
815 }
816 
817 /*-----------------------------------------------------------------
818  * kmirrord
819  *---------------------------------------------------------------*/
820 static void do_mirror(struct work_struct *work)
821 {
822 	struct mirror_set *ms = container_of(work, struct mirror_set,
823 					     kmirrord_work);
824 	struct bio_list reads, writes, failures;
825 	unsigned long flags;
826 
827 	spin_lock_irqsave(&ms->lock, flags);
828 	reads = ms->reads;
829 	writes = ms->writes;
830 	failures = ms->failures;
831 	bio_list_init(&ms->reads);
832 	bio_list_init(&ms->writes);
833 	bio_list_init(&ms->failures);
834 	spin_unlock_irqrestore(&ms->lock, flags);
835 
836 	dm_rh_update_states(ms->rh, errors_handled(ms));
837 	do_recovery(ms);
838 	do_reads(ms, &reads);
839 	do_writes(ms, &writes);
840 	do_failures(ms, &failures);
841 }
842 
843 /*-----------------------------------------------------------------
844  * Target functions
845  *---------------------------------------------------------------*/
846 static struct mirror_set *alloc_context(unsigned int nr_mirrors,
847 					uint32_t region_size,
848 					struct dm_target *ti,
849 					struct dm_dirty_log *dl)
850 {
851 	size_t len;
852 	struct mirror_set *ms = NULL;
853 
854 	len = sizeof(*ms) + (sizeof(ms->mirror[0]) * nr_mirrors);
855 
856 	ms = kzalloc(len, GFP_KERNEL);
857 	if (!ms) {
858 		ti->error = "Cannot allocate mirror context";
859 		return NULL;
860 	}
861 
862 	spin_lock_init(&ms->lock);
863 	bio_list_init(&ms->reads);
864 	bio_list_init(&ms->writes);
865 	bio_list_init(&ms->failures);
866 	bio_list_init(&ms->holds);
867 
868 	ms->ti = ti;
869 	ms->nr_mirrors = nr_mirrors;
870 	ms->nr_regions = dm_sector_div_up(ti->len, region_size);
871 	ms->in_sync = 0;
872 	ms->log_failure = 0;
873 	ms->leg_failure = 0;
874 	atomic_set(&ms->suspend, 0);
875 	atomic_set(&ms->default_mirror, DEFAULT_MIRROR);
876 
877 	ms->io_client = dm_io_client_create();
878 	if (IS_ERR(ms->io_client)) {
879 		ti->error = "Error creating dm_io client";
880 		kfree(ms);
881  		return NULL;
882 	}
883 
884 	ms->rh = dm_region_hash_create(ms, dispatch_bios, wakeup_mirrord,
885 				       wakeup_all_recovery_waiters,
886 				       ms->ti->begin, MAX_RECOVERY,
887 				       dl, region_size, ms->nr_regions);
888 	if (IS_ERR(ms->rh)) {
889 		ti->error = "Error creating dirty region hash";
890 		dm_io_client_destroy(ms->io_client);
891 		kfree(ms);
892 		return NULL;
893 	}
894 
895 	return ms;
896 }
897 
898 static void free_context(struct mirror_set *ms, struct dm_target *ti,
899 			 unsigned int m)
900 {
901 	while (m--)
902 		dm_put_device(ti, ms->mirror[m].dev);
903 
904 	dm_io_client_destroy(ms->io_client);
905 	dm_region_hash_destroy(ms->rh);
906 	kfree(ms);
907 }
908 
909 static int get_mirror(struct mirror_set *ms, struct dm_target *ti,
910 		      unsigned int mirror, char **argv)
911 {
912 	unsigned long long offset;
913 	char dummy;
914 
915 	if (sscanf(argv[1], "%llu%c", &offset, &dummy) != 1) {
916 		ti->error = "Invalid offset";
917 		return -EINVAL;
918 	}
919 
920 	if (dm_get_device(ti, argv[0], dm_table_get_mode(ti->table),
921 			  &ms->mirror[mirror].dev)) {
922 		ti->error = "Device lookup failure";
923 		return -ENXIO;
924 	}
925 
926 	ms->mirror[mirror].ms = ms;
927 	atomic_set(&(ms->mirror[mirror].error_count), 0);
928 	ms->mirror[mirror].error_type = 0;
929 	ms->mirror[mirror].offset = offset;
930 
931 	return 0;
932 }
933 
934 /*
935  * Create dirty log: log_type #log_params <log_params>
936  */
937 static struct dm_dirty_log *create_dirty_log(struct dm_target *ti,
938 					     unsigned argc, char **argv,
939 					     unsigned *args_used)
940 {
941 	unsigned param_count;
942 	struct dm_dirty_log *dl;
943 	char dummy;
944 
945 	if (argc < 2) {
946 		ti->error = "Insufficient mirror log arguments";
947 		return NULL;
948 	}
949 
950 	if (sscanf(argv[1], "%u%c", &param_count, &dummy) != 1) {
951 		ti->error = "Invalid mirror log argument count";
952 		return NULL;
953 	}
954 
955 	*args_used = 2 + param_count;
956 
957 	if (argc < *args_used) {
958 		ti->error = "Insufficient mirror log arguments";
959 		return NULL;
960 	}
961 
962 	dl = dm_dirty_log_create(argv[0], ti, mirror_flush, param_count,
963 				 argv + 2);
964 	if (!dl) {
965 		ti->error = "Error creating mirror dirty log";
966 		return NULL;
967 	}
968 
969 	return dl;
970 }
971 
972 static int parse_features(struct mirror_set *ms, unsigned argc, char **argv,
973 			  unsigned *args_used)
974 {
975 	unsigned num_features;
976 	struct dm_target *ti = ms->ti;
977 	char dummy;
978 
979 	*args_used = 0;
980 
981 	if (!argc)
982 		return 0;
983 
984 	if (sscanf(argv[0], "%u%c", &num_features, &dummy) != 1) {
985 		ti->error = "Invalid number of features";
986 		return -EINVAL;
987 	}
988 
989 	argc--;
990 	argv++;
991 	(*args_used)++;
992 
993 	if (num_features > argc) {
994 		ti->error = "Not enough arguments to support feature count";
995 		return -EINVAL;
996 	}
997 
998 	if (!strcmp("handle_errors", argv[0]))
999 		ms->features |= DM_RAID1_HANDLE_ERRORS;
1000 	else {
1001 		ti->error = "Unrecognised feature requested";
1002 		return -EINVAL;
1003 	}
1004 
1005 	(*args_used)++;
1006 
1007 	return 0;
1008 }
1009 
1010 /*
1011  * Construct a mirror mapping:
1012  *
1013  * log_type #log_params <log_params>
1014  * #mirrors [mirror_path offset]{2,}
1015  * [#features <features>]
1016  *
1017  * log_type is "core" or "disk"
1018  * #log_params is between 1 and 3
1019  *
1020  * If present, features must be "handle_errors".
1021  */
1022 static int mirror_ctr(struct dm_target *ti, unsigned int argc, char **argv)
1023 {
1024 	int r;
1025 	unsigned int nr_mirrors, m, args_used;
1026 	struct mirror_set *ms;
1027 	struct dm_dirty_log *dl;
1028 	char dummy;
1029 
1030 	dl = create_dirty_log(ti, argc, argv, &args_used);
1031 	if (!dl)
1032 		return -EINVAL;
1033 
1034 	argv += args_used;
1035 	argc -= args_used;
1036 
1037 	if (!argc || sscanf(argv[0], "%u%c", &nr_mirrors, &dummy) != 1 ||
1038 	    nr_mirrors < 2 || nr_mirrors > DM_KCOPYD_MAX_REGIONS + 1) {
1039 		ti->error = "Invalid number of mirrors";
1040 		dm_dirty_log_destroy(dl);
1041 		return -EINVAL;
1042 	}
1043 
1044 	argv++, argc--;
1045 
1046 	if (argc < nr_mirrors * 2) {
1047 		ti->error = "Too few mirror arguments";
1048 		dm_dirty_log_destroy(dl);
1049 		return -EINVAL;
1050 	}
1051 
1052 	ms = alloc_context(nr_mirrors, dl->type->get_region_size(dl), ti, dl);
1053 	if (!ms) {
1054 		dm_dirty_log_destroy(dl);
1055 		return -ENOMEM;
1056 	}
1057 
1058 	/* Get the mirror parameter sets */
1059 	for (m = 0; m < nr_mirrors; m++) {
1060 		r = get_mirror(ms, ti, m, argv);
1061 		if (r) {
1062 			free_context(ms, ti, m);
1063 			return r;
1064 		}
1065 		argv += 2;
1066 		argc -= 2;
1067 	}
1068 
1069 	ti->private = ms;
1070 
1071 	r = dm_set_target_max_io_len(ti, dm_rh_get_region_size(ms->rh));
1072 	if (r)
1073 		goto err_free_context;
1074 
1075 	ti->num_flush_requests = 1;
1076 	ti->num_discard_requests = 1;
1077 	ti->per_bio_data_size = sizeof(struct dm_raid1_bio_record);
1078 	ti->discard_zeroes_data_unsupported = true;
1079 
1080 	ms->kmirrord_wq = alloc_workqueue("kmirrord",
1081 					  WQ_NON_REENTRANT | WQ_MEM_RECLAIM, 0);
1082 	if (!ms->kmirrord_wq) {
1083 		DMERR("couldn't start kmirrord");
1084 		r = -ENOMEM;
1085 		goto err_free_context;
1086 	}
1087 	INIT_WORK(&ms->kmirrord_work, do_mirror);
1088 	init_timer(&ms->timer);
1089 	ms->timer_pending = 0;
1090 	INIT_WORK(&ms->trigger_event, trigger_event);
1091 
1092 	r = parse_features(ms, argc, argv, &args_used);
1093 	if (r)
1094 		goto err_destroy_wq;
1095 
1096 	argv += args_used;
1097 	argc -= args_used;
1098 
1099 	/*
1100 	 * Any read-balancing addition depends on the
1101 	 * DM_RAID1_HANDLE_ERRORS flag being present.
1102 	 * This is because the decision to balance depends
1103 	 * on the sync state of a region.  If the above
1104 	 * flag is not present, we ignore errors; and
1105 	 * the sync state may be inaccurate.
1106 	 */
1107 
1108 	if (argc) {
1109 		ti->error = "Too many mirror arguments";
1110 		r = -EINVAL;
1111 		goto err_destroy_wq;
1112 	}
1113 
1114 	ms->kcopyd_client = dm_kcopyd_client_create();
1115 	if (IS_ERR(ms->kcopyd_client)) {
1116 		r = PTR_ERR(ms->kcopyd_client);
1117 		goto err_destroy_wq;
1118 	}
1119 
1120 	wakeup_mirrord(ms);
1121 	return 0;
1122 
1123 err_destroy_wq:
1124 	destroy_workqueue(ms->kmirrord_wq);
1125 err_free_context:
1126 	free_context(ms, ti, ms->nr_mirrors);
1127 	return r;
1128 }
1129 
1130 static void mirror_dtr(struct dm_target *ti)
1131 {
1132 	struct mirror_set *ms = (struct mirror_set *) ti->private;
1133 
1134 	del_timer_sync(&ms->timer);
1135 	flush_workqueue(ms->kmirrord_wq);
1136 	flush_work(&ms->trigger_event);
1137 	dm_kcopyd_client_destroy(ms->kcopyd_client);
1138 	destroy_workqueue(ms->kmirrord_wq);
1139 	free_context(ms, ti, ms->nr_mirrors);
1140 }
1141 
1142 /*
1143  * Mirror mapping function
1144  */
1145 static int mirror_map(struct dm_target *ti, struct bio *bio)
1146 {
1147 	int r, rw = bio_rw(bio);
1148 	struct mirror *m;
1149 	struct mirror_set *ms = ti->private;
1150 	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);
1151 	struct dm_raid1_bio_record *bio_record =
1152 	  dm_per_bio_data(bio, sizeof(struct dm_raid1_bio_record));
1153 
1154 	bio_record->details.bi_bdev = NULL;
1155 
1156 	if (rw == WRITE) {
1157 		/* Save region for mirror_end_io() handler */
1158 		bio_record->write_region = dm_rh_bio_to_region(ms->rh, bio);
1159 		queue_bio(ms, bio, rw);
1160 		return DM_MAPIO_SUBMITTED;
1161 	}
1162 
1163 	r = log->type->in_sync(log, dm_rh_bio_to_region(ms->rh, bio), 0);
1164 	if (r < 0 && r != -EWOULDBLOCK)
1165 		return r;
1166 
1167 	/*
1168 	 * If region is not in-sync queue the bio.
1169 	 */
1170 	if (!r || (r == -EWOULDBLOCK)) {
1171 		if (rw == READA)
1172 			return -EWOULDBLOCK;
1173 
1174 		queue_bio(ms, bio, rw);
1175 		return DM_MAPIO_SUBMITTED;
1176 	}
1177 
1178 	/*
1179 	 * The region is in-sync and we can perform reads directly.
1180 	 * Store enough information so we can retry if it fails.
1181 	 */
1182 	m = choose_mirror(ms, bio->bi_sector);
1183 	if (unlikely(!m))
1184 		return -EIO;
1185 
1186 	dm_bio_record(&bio_record->details, bio);
1187 	bio_record->m = m;
1188 
1189 	map_bio(m, bio);
1190 
1191 	return DM_MAPIO_REMAPPED;
1192 }
1193 
1194 static int mirror_end_io(struct dm_target *ti, struct bio *bio, int error)
1195 {
1196 	int rw = bio_rw(bio);
1197 	struct mirror_set *ms = (struct mirror_set *) ti->private;
1198 	struct mirror *m = NULL;
1199 	struct dm_bio_details *bd = NULL;
1200 	struct dm_raid1_bio_record *bio_record =
1201 	  dm_per_bio_data(bio, sizeof(struct dm_raid1_bio_record));
1202 
1203 	/*
1204 	 * We need to dec pending if this was a write.
1205 	 */
1206 	if (rw == WRITE) {
1207 		if (!(bio->bi_rw & (REQ_FLUSH | REQ_DISCARD)))
1208 			dm_rh_dec(ms->rh, bio_record->write_region);
1209 		return error;
1210 	}
1211 
1212 	if (error == -EOPNOTSUPP)
1213 		goto out;
1214 
1215 	if ((error == -EWOULDBLOCK) && (bio->bi_rw & REQ_RAHEAD))
1216 		goto out;
1217 
1218 	if (unlikely(error)) {
1219 		if (!bio_record->details.bi_bdev) {
1220 			/*
1221 			 * There wasn't enough memory to record necessary
1222 			 * information for a retry or there was no other
1223 			 * mirror in-sync.
1224 			 */
1225 			DMERR_LIMIT("Mirror read failed.");
1226 			return -EIO;
1227 		}
1228 
1229 		m = bio_record->m;
1230 
1231 		DMERR("Mirror read failed from %s. Trying alternative device.",
1232 		      m->dev->name);
1233 
1234 		fail_mirror(m, DM_RAID1_READ_ERROR);
1235 
1236 		/*
1237 		 * A failed read is requeued for another attempt using an intact
1238 		 * mirror.
1239 		 */
1240 		if (default_ok(m) || mirror_available(ms, bio)) {
1241 			bd = &bio_record->details;
1242 
1243 			dm_bio_restore(bd, bio);
1244 			bio_record->details.bi_bdev = NULL;
1245 			queue_bio(ms, bio, rw);
1246 			return DM_ENDIO_INCOMPLETE;
1247 		}
1248 		DMERR("All replicated volumes dead, failing I/O");
1249 	}
1250 
1251 out:
1252 	bio_record->details.bi_bdev = NULL;
1253 
1254 	return error;
1255 }
1256 
1257 static void mirror_presuspend(struct dm_target *ti)
1258 {
1259 	struct mirror_set *ms = (struct mirror_set *) ti->private;
1260 	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);
1261 
1262 	struct bio_list holds;
1263 	struct bio *bio;
1264 
1265 	atomic_set(&ms->suspend, 1);
1266 
1267 	/*
1268 	 * Process bios in the hold list to start recovery waiting
1269 	 * for bios in the hold list. After the process, no bio has
1270 	 * a chance to be added in the hold list because ms->suspend
1271 	 * is set.
1272 	 */
1273 	spin_lock_irq(&ms->lock);
1274 	holds = ms->holds;
1275 	bio_list_init(&ms->holds);
1276 	spin_unlock_irq(&ms->lock);
1277 
1278 	while ((bio = bio_list_pop(&holds)))
1279 		hold_bio(ms, bio);
1280 
1281 	/*
1282 	 * We must finish up all the work that we've
1283 	 * generated (i.e. recovery work).
1284 	 */
1285 	dm_rh_stop_recovery(ms->rh);
1286 
1287 	wait_event(_kmirrord_recovery_stopped,
1288 		   !dm_rh_recovery_in_flight(ms->rh));
1289 
1290 	if (log->type->presuspend && log->type->presuspend(log))
1291 		/* FIXME: need better error handling */
1292 		DMWARN("log presuspend failed");
1293 
1294 	/*
1295 	 * Now that recovery is complete/stopped and the
1296 	 * delayed bios are queued, we need to wait for
1297 	 * the worker thread to complete.  This way,
1298 	 * we know that all of our I/O has been pushed.
1299 	 */
1300 	flush_workqueue(ms->kmirrord_wq);
1301 }
1302 
1303 static void mirror_postsuspend(struct dm_target *ti)
1304 {
1305 	struct mirror_set *ms = ti->private;
1306 	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);
1307 
1308 	if (log->type->postsuspend && log->type->postsuspend(log))
1309 		/* FIXME: need better error handling */
1310 		DMWARN("log postsuspend failed");
1311 }
1312 
1313 static void mirror_resume(struct dm_target *ti)
1314 {
1315 	struct mirror_set *ms = ti->private;
1316 	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);
1317 
1318 	atomic_set(&ms->suspend, 0);
1319 	if (log->type->resume && log->type->resume(log))
1320 		/* FIXME: need better error handling */
1321 		DMWARN("log resume failed");
1322 	dm_rh_start_recovery(ms->rh);
1323 }
1324 
1325 /*
1326  * device_status_char
1327  * @m: mirror device/leg we want the status of
1328  *
1329  * We return one character representing the most severe error
1330  * we have encountered.
1331  *    A => Alive - No failures
1332  *    D => Dead - A write failure occurred leaving mirror out-of-sync
1333  *    S => Sync - A sychronization failure occurred, mirror out-of-sync
1334  *    R => Read - A read failure occurred, mirror data unaffected
1335  *
1336  * Returns: <char>
1337  */
1338 static char device_status_char(struct mirror *m)
1339 {
1340 	if (!atomic_read(&(m->error_count)))
1341 		return 'A';
1342 
1343 	return (test_bit(DM_RAID1_FLUSH_ERROR, &(m->error_type))) ? 'F' :
1344 		(test_bit(DM_RAID1_WRITE_ERROR, &(m->error_type))) ? 'D' :
1345 		(test_bit(DM_RAID1_SYNC_ERROR, &(m->error_type))) ? 'S' :
1346 		(test_bit(DM_RAID1_READ_ERROR, &(m->error_type))) ? 'R' : 'U';
1347 }
1348 
1349 
1350 static int mirror_status(struct dm_target *ti, status_type_t type,
1351 			 unsigned status_flags, char *result, unsigned maxlen)
1352 {
1353 	unsigned int m, sz = 0;
1354 	struct mirror_set *ms = (struct mirror_set *) ti->private;
1355 	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);
1356 	char buffer[ms->nr_mirrors + 1];
1357 
1358 	switch (type) {
1359 	case STATUSTYPE_INFO:
1360 		DMEMIT("%d ", ms->nr_mirrors);
1361 		for (m = 0; m < ms->nr_mirrors; m++) {
1362 			DMEMIT("%s ", ms->mirror[m].dev->name);
1363 			buffer[m] = device_status_char(&(ms->mirror[m]));
1364 		}
1365 		buffer[m] = '\0';
1366 
1367 		DMEMIT("%llu/%llu 1 %s ",
1368 		      (unsigned long long)log->type->get_sync_count(log),
1369 		      (unsigned long long)ms->nr_regions, buffer);
1370 
1371 		sz += log->type->status(log, type, result+sz, maxlen-sz);
1372 
1373 		break;
1374 
1375 	case STATUSTYPE_TABLE:
1376 		sz = log->type->status(log, type, result, maxlen);
1377 
1378 		DMEMIT("%d", ms->nr_mirrors);
1379 		for (m = 0; m < ms->nr_mirrors; m++)
1380 			DMEMIT(" %s %llu", ms->mirror[m].dev->name,
1381 			       (unsigned long long)ms->mirror[m].offset);
1382 
1383 		if (ms->features & DM_RAID1_HANDLE_ERRORS)
1384 			DMEMIT(" 1 handle_errors");
1385 	}
1386 
1387 	return 0;
1388 }
1389 
1390 static int mirror_iterate_devices(struct dm_target *ti,
1391 				  iterate_devices_callout_fn fn, void *data)
1392 {
1393 	struct mirror_set *ms = ti->private;
1394 	int ret = 0;
1395 	unsigned i;
1396 
1397 	for (i = 0; !ret && i < ms->nr_mirrors; i++)
1398 		ret = fn(ti, ms->mirror[i].dev,
1399 			 ms->mirror[i].offset, ti->len, data);
1400 
1401 	return ret;
1402 }
1403 
1404 static struct target_type mirror_target = {
1405 	.name	 = "mirror",
1406 	.version = {1, 13, 1},
1407 	.module	 = THIS_MODULE,
1408 	.ctr	 = mirror_ctr,
1409 	.dtr	 = mirror_dtr,
1410 	.map	 = mirror_map,
1411 	.end_io	 = mirror_end_io,
1412 	.presuspend = mirror_presuspend,
1413 	.postsuspend = mirror_postsuspend,
1414 	.resume	 = mirror_resume,
1415 	.status	 = mirror_status,
1416 	.iterate_devices = mirror_iterate_devices,
1417 };
1418 
1419 static int __init dm_mirror_init(void)
1420 {
1421 	int r;
1422 
1423 	r = dm_register_target(&mirror_target);
1424 	if (r < 0) {
1425 		DMERR("Failed to register mirror target");
1426 		goto bad_target;
1427 	}
1428 
1429 	return 0;
1430 
1431 bad_target:
1432 	return r;
1433 }
1434 
1435 static void __exit dm_mirror_exit(void)
1436 {
1437 	dm_unregister_target(&mirror_target);
1438 }
1439 
1440 /* Module hooks */
1441 module_init(dm_mirror_init);
1442 module_exit(dm_mirror_exit);
1443 
1444 MODULE_DESCRIPTION(DM_NAME " mirror target");
1445 MODULE_AUTHOR("Joe Thornber");
1446 MODULE_LICENSE("GPL");
1447