raid5.c 182.0 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4
/*
 * raid5.c : Multiple Devices driver for Linux
 *	   Copyright (C) 1996, 1997 Ingo Molnar, Miguel de Icaza, Gadi Oxman
 *	   Copyright (C) 1999, 2000 Ingo Molnar
5
 *	   Copyright (C) 2002, 2003 H. Peter Anvin
L
Linus Torvalds 已提交
6
 *
7 8 9
 * RAID-4/5/6 management functions.
 * Thanks to Penguin Computing for making the RAID-6 development possible
 * by donating a test server!
L
Linus Torvalds 已提交
10 11 12 13 14 15 16 17 18 19 20
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2, or (at your option)
 * any later version.
 *
 * You should have received a copy of the GNU General Public License
 * (for example /usr/src/linux/COPYING); if not, write to the Free
 * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 */

21 22 23 24 25 26 27 28 29
/*
 * BITMAP UNPLUGGING:
 *
 * The sequencing for updating the bitmap reliably is a little
 * subtle (and I got it wrong the first time) so it deserves some
 * explanation.
 *
 * We group bitmap updates into batches.  Each batch has a number.
 * We may write out several batches at once, but that isn't very important.
30 31
 * conf->seq_write is the number of the last batch successfully written.
 * conf->seq_flush is the number of the last batch that was closed to
32 33 34
 *    new additions.
 * When we discover that we will need to write to any block in a stripe
 * (in add_stripe_bio) we update the in-memory bitmap and record in sh->bm_seq
35
 * the number of the batch it will be in. This is seq_flush+1.
36 37 38 39 40 41 42 43 44
 * When we are ready to do a write, if that batch hasn't been written yet,
 *   we plug the array and queue the stripe for later.
 * When an unplug happens, we increment bm_flush, thus closing the current
 *   batch.
 * When we notice that bm_flush > bm_write, we write out all pending updates
 * to the bitmap, and advance bm_write to where bm_flush was.
 * This may occasionally write a bit out twice, but is sure never to
 * miss any bits.
 */
L
Linus Torvalds 已提交
45

46
#include <linux/blkdev.h>
47
#include <linux/kthread.h>
48
#include <linux/raid/pq.h>
49
#include <linux/async_tx.h>
50
#include <linux/module.h>
51
#include <linux/async.h>
52
#include <linux/seq_file.h>
53
#include <linux/cpu.h>
54
#include <linux/slab.h>
55
#include <linux/ratelimit.h>
N
NeilBrown 已提交
56 57
#include <trace/events/block.h>

58
#include "md.h"
59
#include "raid5.h"
60
#include "raid0.h"
61
#include "bitmap.h"
62

L
Linus Torvalds 已提交
63 64 65 66 67 68 69 70 71
/*
 * Stripe cache
 */

#define NR_STRIPES		256
#define STRIPE_SIZE		PAGE_SIZE
#define STRIPE_SHIFT		(PAGE_SHIFT - 9)
#define STRIPE_SECTORS		(STRIPE_SIZE>>9)
#define	IO_THRESHOLD		1
72
#define BYPASS_THRESHOLD	1
73
#define NR_HASH			(PAGE_SIZE / sizeof(struct hlist_head))
L
Linus Torvalds 已提交
74 75
#define HASH_MASK		(NR_HASH - 1)

76
static inline struct hlist_head *stripe_hash(struct r5conf *conf, sector_t sect)
77 78 79 80
{
	int hash = (sect >> STRIPE_SHIFT) & HASH_MASK;
	return &conf->stripe_hashtbl[hash];
}
L
Linus Torvalds 已提交
81 82 83 84 85 86 87

/* bio's attached to a stripe+device for I/O are linked together in bi_sector
 * order without overlap.  There may be several bio's per stripe+device, and
 * a bio could span several devices.
 * When walking this list for a particular stripe+device, we must never proceed
 * beyond a bio that extends past this device, as the next bio might no longer
 * be valid.
88
 * This function is used to determine the 'next' bio in the list, given the sector
L
Linus Torvalds 已提交
89 90
 * of the current stripe+device
 */
91 92
static inline struct bio *r5_next_bio(struct bio *bio, sector_t sector)
{
93
	int sectors = bio_sectors(bio);
94 95 96 97 98
	if (bio->bi_sector + sectors < sector + STRIPE_SECTORS)
		return bio->bi_next;
	else
		return NULL;
}
L
Linus Torvalds 已提交
99

100
/*
101 102
 * We maintain a biased count of active stripes in the bottom 16 bits of
 * bi_phys_segments, and a count of processed stripes in the upper 16 bits
103
 */
104
static inline int raid5_bi_processed_stripes(struct bio *bio)
105
{
106 107
	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
	return (atomic_read(segments) >> 16) & 0xffff;
108 109
}

110
static inline int raid5_dec_bi_active_stripes(struct bio *bio)
111
{
112 113
	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
	return atomic_sub_return(1, segments) & 0xffff;
114 115
}

116
static inline void raid5_inc_bi_active_stripes(struct bio *bio)
117
{
118 119
	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
	atomic_inc(segments);
120 121
}

122 123
static inline void raid5_set_bi_processed_stripes(struct bio *bio,
	unsigned int cnt)
124
{
125 126
	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
	int old, new;
127

128 129 130 131
	do {
		old = atomic_read(segments);
		new = (old & 0xffff) | (cnt << 16);
	} while (atomic_cmpxchg(segments, old, new) != old);
132 133
}

134
static inline void raid5_set_bi_stripes(struct bio *bio, unsigned int cnt)
135
{
136 137
	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
	atomic_set(segments, cnt);
138 139
}

140 141 142
/* Find first data disk in a raid6 stripe */
static inline int raid6_d0(struct stripe_head *sh)
{
143 144 145 146
	if (sh->ddf_layout)
		/* ddf always start from first device */
		return 0;
	/* md starts just after Q block */
147 148 149 150 151
	if (sh->qd_idx == sh->disks - 1)
		return 0;
	else
		return sh->qd_idx + 1;
}
152 153 154 155 156
static inline int raid6_next_disk(int disk, int raid_disks)
{
	disk++;
	return (disk < raid_disks) ? disk : 0;
}
157

158 159 160 161 162
/* When walking through the disks in a raid5, starting at raid6_d0,
 * We need to map each disk to a 'slot', where the data disks are slot
 * 0 .. raid_disks-3, the parity disk is raid_disks-2 and the Q disk
 * is raid_disks-1.  This help does that mapping.
 */
163 164
static int raid6_idx_to_slot(int idx, struct stripe_head *sh,
			     int *count, int syndrome_disks)
165
{
166
	int slot = *count;
167

168
	if (sh->ddf_layout)
169
		(*count)++;
170
	if (idx == sh->pd_idx)
171
		return syndrome_disks;
172
	if (idx == sh->qd_idx)
173
		return syndrome_disks + 1;
174
	if (!sh->ddf_layout)
175
		(*count)++;
176 177 178
	return slot;
}

179 180 181 182 183 184 185 186
static void return_io(struct bio *return_bi)
{
	struct bio *bi = return_bi;
	while (bi) {

		return_bi = bi->bi_next;
		bi->bi_next = NULL;
		bi->bi_size = 0;
187 188
		trace_block_bio_complete(bdev_get_queue(bi->bi_bdev),
					 bi, 0);
189
		bio_endio(bi, 0);
190 191 192 193
		bi = return_bi;
	}
}

194
static void print_raid5_conf (struct r5conf *conf);
L
Linus Torvalds 已提交
195

196 197 198 199 200 201 202
static int stripe_operations_active(struct stripe_head *sh)
{
	return sh->check_state || sh->reconstruct_state ||
	       test_bit(STRIPE_BIOFILL_RUN, &sh->state) ||
	       test_bit(STRIPE_COMPUTE_RUN, &sh->state);
}

203
static void do_release_stripe(struct r5conf *conf, struct stripe_head *sh)
L
Linus Torvalds 已提交
204
{
205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231
	BUG_ON(!list_empty(&sh->lru));
	BUG_ON(atomic_read(&conf->active_stripes)==0);
	if (test_bit(STRIPE_HANDLE, &sh->state)) {
		if (test_bit(STRIPE_DELAYED, &sh->state) &&
		    !test_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
			list_add_tail(&sh->lru, &conf->delayed_list);
		else if (test_bit(STRIPE_BIT_DELAY, &sh->state) &&
			   sh->bm_seq - conf->seq_write > 0)
			list_add_tail(&sh->lru, &conf->bitmap_list);
		else {
			clear_bit(STRIPE_DELAYED, &sh->state);
			clear_bit(STRIPE_BIT_DELAY, &sh->state);
			list_add_tail(&sh->lru, &conf->handle_list);
		}
		md_wakeup_thread(conf->mddev->thread);
	} else {
		BUG_ON(stripe_operations_active(sh));
		if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
			if (atomic_dec_return(&conf->preread_active_stripes)
			    < IO_THRESHOLD)
				md_wakeup_thread(conf->mddev->thread);
		atomic_dec(&conf->active_stripes);
		if (!test_bit(STRIPE_EXPANDING, &sh->state)) {
			list_add_tail(&sh->lru, &conf->inactive_list);
			wake_up(&conf->wait_for_stripe);
			if (conf->retry_read_aligned)
				md_wakeup_thread(conf->mddev->thread);
L
Linus Torvalds 已提交
232 233 234
		}
	}
}
235

236 237 238 239 240 241
static void __release_stripe(struct r5conf *conf, struct stripe_head *sh)
{
	if (atomic_dec_and_test(&sh->count))
		do_release_stripe(conf, sh);
}

L
Linus Torvalds 已提交
242 243
static void release_stripe(struct stripe_head *sh)
{
244
	struct r5conf *conf = sh->raid_conf;
L
Linus Torvalds 已提交
245
	unsigned long flags;
246

247 248 249 250 251 252
	local_irq_save(flags);
	if (atomic_dec_and_lock(&sh->count, &conf->device_lock)) {
		do_release_stripe(conf, sh);
		spin_unlock(&conf->device_lock);
	}
	local_irq_restore(flags);
L
Linus Torvalds 已提交
253 254
}

255
static inline void remove_hash(struct stripe_head *sh)
L
Linus Torvalds 已提交
256
{
257 258
	pr_debug("remove_hash(), stripe %llu\n",
		(unsigned long long)sh->sector);
L
Linus Torvalds 已提交
259

260
	hlist_del_init(&sh->hash);
L
Linus Torvalds 已提交
261 262
}

263
static inline void insert_hash(struct r5conf *conf, struct stripe_head *sh)
L
Linus Torvalds 已提交
264
{
265
	struct hlist_head *hp = stripe_hash(conf, sh->sector);
L
Linus Torvalds 已提交
266

267 268
	pr_debug("insert_hash(), stripe %llu\n",
		(unsigned long long)sh->sector);
L
Linus Torvalds 已提交
269

270
	hlist_add_head(&sh->hash, hp);
L
Linus Torvalds 已提交
271 272 273 274
}


/* find an idle stripe, make sure it is unhashed, and return it. */
275
static struct stripe_head *get_free_stripe(struct r5conf *conf)
L
Linus Torvalds 已提交
276 277 278 279 280 281 282 283 284 285 286 287 288 289 290
{
	struct stripe_head *sh = NULL;
	struct list_head *first;

	if (list_empty(&conf->inactive_list))
		goto out;
	first = conf->inactive_list.next;
	sh = list_entry(first, struct stripe_head, lru);
	list_del_init(first);
	remove_hash(sh);
	atomic_inc(&conf->active_stripes);
out:
	return sh;
}

291
static void shrink_buffers(struct stripe_head *sh)
L
Linus Torvalds 已提交
292 293 294
{
	struct page *p;
	int i;
295
	int num = sh->raid_conf->pool_size;
L
Linus Torvalds 已提交
296

297
	for (i = 0; i < num ; i++) {
L
Linus Torvalds 已提交
298 299 300 301
		p = sh->dev[i].page;
		if (!p)
			continue;
		sh->dev[i].page = NULL;
302
		put_page(p);
L
Linus Torvalds 已提交
303 304 305
	}
}

306
static int grow_buffers(struct stripe_head *sh)
L
Linus Torvalds 已提交
307 308
{
	int i;
309
	int num = sh->raid_conf->pool_size;
L
Linus Torvalds 已提交
310

311
	for (i = 0; i < num; i++) {
L
Linus Torvalds 已提交
312 313 314 315 316 317 318 319 320 321
		struct page *page;

		if (!(page = alloc_page(GFP_KERNEL))) {
			return 1;
		}
		sh->dev[i].page = page;
	}
	return 0;
}

322
static void raid5_build_block(struct stripe_head *sh, int i, int previous);
323
static void stripe_set_idx(sector_t stripe, struct r5conf *conf, int previous,
324
			    struct stripe_head *sh);
L
Linus Torvalds 已提交
325

326
static void init_stripe(struct stripe_head *sh, sector_t sector, int previous)
L
Linus Torvalds 已提交
327
{
328
	struct r5conf *conf = sh->raid_conf;
329
	int i;
L
Linus Torvalds 已提交
330

331 332
	BUG_ON(atomic_read(&sh->count) != 0);
	BUG_ON(test_bit(STRIPE_HANDLE, &sh->state));
333
	BUG_ON(stripe_operations_active(sh));
334

335
	pr_debug("init_stripe called, stripe %llu\n",
L
Linus Torvalds 已提交
336 337 338
		(unsigned long long)sh->sector);

	remove_hash(sh);
339

340
	sh->generation = conf->generation - previous;
341
	sh->disks = previous ? conf->previous_raid_disks : conf->raid_disks;
L
Linus Torvalds 已提交
342
	sh->sector = sector;
343
	stripe_set_idx(sector, conf, previous, sh);
L
Linus Torvalds 已提交
344 345
	sh->state = 0;

346 347

	for (i = sh->disks; i--; ) {
L
Linus Torvalds 已提交
348 349
		struct r5dev *dev = &sh->dev[i];

350
		if (dev->toread || dev->read || dev->towrite || dev->written ||
L
Linus Torvalds 已提交
351
		    test_bit(R5_LOCKED, &dev->flags)) {
352
			printk(KERN_ERR "sector=%llx i=%d %p %p %p %p %d\n",
L
Linus Torvalds 已提交
353
			       (unsigned long long)sh->sector, i, dev->toread,
354
			       dev->read, dev->towrite, dev->written,
L
Linus Torvalds 已提交
355
			       test_bit(R5_LOCKED, &dev->flags));
356
			WARN_ON(1);
L
Linus Torvalds 已提交
357 358
		}
		dev->flags = 0;
359
		raid5_build_block(sh, i, previous);
L
Linus Torvalds 已提交
360 361 362 363
	}
	insert_hash(conf, sh);
}

364
static struct stripe_head *__find_stripe(struct r5conf *conf, sector_t sector,
365
					 short generation)
L
Linus Torvalds 已提交
366 367 368
{
	struct stripe_head *sh;

369
	pr_debug("__find_stripe, sector %llu\n", (unsigned long long)sector);
370
	hlist_for_each_entry(sh, stripe_hash(conf, sector), hash)
371
		if (sh->sector == sector && sh->generation == generation)
L
Linus Torvalds 已提交
372
			return sh;
373
	pr_debug("__stripe %llu not in cache\n", (unsigned long long)sector);
L
Linus Torvalds 已提交
374 375 376
	return NULL;
}

377 378 379 380 381 382 383 384 385 386 387 388 389
/*
 * Need to check if array has failed when deciding whether to:
 *  - start an array
 *  - remove non-faulty devices
 *  - add a spare
 *  - allow a reshape
 * This determination is simple when no reshape is happening.
 * However if there is a reshape, we need to carefully check
 * both the before and after sections.
 * This is because some failed devices may only affect one
 * of the two sections, and some non-in_sync devices may
 * be insync in the section most affected by failed devices.
 */
390
static int calc_degraded(struct r5conf *conf)
391
{
392
	int degraded, degraded2;
393 394 395 396 397
	int i;

	rcu_read_lock();
	degraded = 0;
	for (i = 0; i < conf->previous_raid_disks; i++) {
398
		struct md_rdev *rdev = rcu_dereference(conf->disks[i].rdev);
399 400
		if (rdev && test_bit(Faulty, &rdev->flags))
			rdev = rcu_dereference(conf->disks[i].replacement);
401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418
		if (!rdev || test_bit(Faulty, &rdev->flags))
			degraded++;
		else if (test_bit(In_sync, &rdev->flags))
			;
		else
			/* not in-sync or faulty.
			 * If the reshape increases the number of devices,
			 * this is being recovered by the reshape, so
			 * this 'previous' section is not in_sync.
			 * If the number of devices is being reduced however,
			 * the device can only be part of the array if
			 * we are reverting a reshape, so this section will
			 * be in-sync.
			 */
			if (conf->raid_disks >= conf->previous_raid_disks)
				degraded++;
	}
	rcu_read_unlock();
419 420
	if (conf->raid_disks == conf->previous_raid_disks)
		return degraded;
421
	rcu_read_lock();
422
	degraded2 = 0;
423
	for (i = 0; i < conf->raid_disks; i++) {
424
		struct md_rdev *rdev = rcu_dereference(conf->disks[i].rdev);
425 426
		if (rdev && test_bit(Faulty, &rdev->flags))
			rdev = rcu_dereference(conf->disks[i].replacement);
427
		if (!rdev || test_bit(Faulty, &rdev->flags))
428
			degraded2++;
429 430 431 432 433 434 435 436 437
		else if (test_bit(In_sync, &rdev->flags))
			;
		else
			/* not in-sync or faulty.
			 * If reshape increases the number of devices, this
			 * section has already been recovered, else it
			 * almost certainly hasn't.
			 */
			if (conf->raid_disks <= conf->previous_raid_disks)
438
				degraded2++;
439 440
	}
	rcu_read_unlock();
441 442 443 444 445 446 447 448 449 450 451 452 453
	if (degraded2 > degraded)
		return degraded2;
	return degraded;
}

static int has_failed(struct r5conf *conf)
{
	int degraded;

	if (conf->mddev->reshape_position == MaxSector)
		return conf->mddev->degraded > conf->max_degraded;

	degraded = calc_degraded(conf);
454 455 456 457 458
	if (degraded > conf->max_degraded)
		return 1;
	return 0;
}

459
static struct stripe_head *
460
get_active_stripe(struct r5conf *conf, sector_t sector,
461
		  int previous, int noblock, int noquiesce)
L
Linus Torvalds 已提交
462 463 464
{
	struct stripe_head *sh;

465
	pr_debug("get_stripe, sector %llu\n", (unsigned long long)sector);
L
Linus Torvalds 已提交
466 467 468 469

	spin_lock_irq(&conf->device_lock);

	do {
470
		wait_event_lock_irq(conf->wait_for_stripe,
471
				    conf->quiesce == 0 || noquiesce,
472
				    conf->device_lock);
473
		sh = __find_stripe(conf, sector, conf->generation - previous);
L
Linus Torvalds 已提交
474 475 476 477 478 479 480 481 482
		if (!sh) {
			if (!conf->inactive_blocked)
				sh = get_free_stripe(conf);
			if (noblock && sh == NULL)
				break;
			if (!sh) {
				conf->inactive_blocked = 1;
				wait_event_lock_irq(conf->wait_for_stripe,
						    !list_empty(&conf->inactive_list) &&
483 484
						    (atomic_read(&conf->active_stripes)
						     < (conf->max_nr_stripes *3/4)
L
Linus Torvalds 已提交
485
						     || !conf->inactive_blocked),
486
						    conf->device_lock);
L
Linus Torvalds 已提交
487 488
				conf->inactive_blocked = 0;
			} else
489
				init_stripe(sh, sector, previous);
L
Linus Torvalds 已提交
490 491
		} else {
			if (atomic_read(&sh->count)) {
492
				BUG_ON(!list_empty(&sh->lru)
493 494
				    && !test_bit(STRIPE_EXPANDING, &sh->state)
				    && !test_bit(STRIPE_ON_UNPLUG_LIST, &sh->state));
L
Linus Torvalds 已提交
495 496 497
			} else {
				if (!test_bit(STRIPE_HANDLE, &sh->state))
					atomic_inc(&conf->active_stripes);
498 499
				if (list_empty(&sh->lru) &&
				    !test_bit(STRIPE_EXPANDING, &sh->state))
500 501
					BUG();
				list_del_init(&sh->lru);
L
Linus Torvalds 已提交
502 503 504 505 506 507 508 509 510 511 512
			}
		}
	} while (sh == NULL);

	if (sh)
		atomic_inc(&sh->count);

	spin_unlock_irq(&conf->device_lock);
	return sh;
}

513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533
/* Determine if 'data_offset' or 'new_data_offset' should be used
 * in this stripe_head.
 */
static int use_new_offset(struct r5conf *conf, struct stripe_head *sh)
{
	sector_t progress = conf->reshape_progress;
	/* Need a memory barrier to make sure we see the value
	 * of conf->generation, or ->data_offset that was set before
	 * reshape_progress was updated.
	 */
	smp_rmb();
	if (progress == MaxSector)
		return 0;
	if (sh->generation == conf->generation - 1)
		return 0;
	/* We are in a reshape, and this is a new-generation stripe,
	 * so use new_data_offset.
	 */
	return 1;
}

534 535 536 537
static void
raid5_end_read_request(struct bio *bi, int error);
static void
raid5_end_write_request(struct bio *bi, int error);
538

539
static void ops_run_io(struct stripe_head *sh, struct stripe_head_state *s)
540
{
541
	struct r5conf *conf = sh->raid_conf;
542 543 544 545 546 547
	int i, disks = sh->disks;

	might_sleep();

	for (i = disks; i--; ) {
		int rw;
548
		int replace_only = 0;
549 550
		struct bio *bi, *rbi;
		struct md_rdev *rdev, *rrdev = NULL;
T
Tejun Heo 已提交
551 552 553 554 555
		if (test_and_clear_bit(R5_Wantwrite, &sh->dev[i].flags)) {
			if (test_and_clear_bit(R5_WantFUA, &sh->dev[i].flags))
				rw = WRITE_FUA;
			else
				rw = WRITE;
556
			if (test_bit(R5_Discard, &sh->dev[i].flags))
S
Shaohua Li 已提交
557
				rw |= REQ_DISCARD;
T
Tejun Heo 已提交
558
		} else if (test_and_clear_bit(R5_Wantread, &sh->dev[i].flags))
559
			rw = READ;
560 561 562 563 564
		else if (test_and_clear_bit(R5_WantReplace,
					    &sh->dev[i].flags)) {
			rw = WRITE;
			replace_only = 1;
		} else
565
			continue;
S
Shaohua Li 已提交
566 567
		if (test_and_clear_bit(R5_SyncIO, &sh->dev[i].flags))
			rw |= REQ_SYNC;
568 569

		bi = &sh->dev[i].req;
570
		rbi = &sh->dev[i].rreq; /* For writing to replacement */
571 572

		rcu_read_lock();
573
		rrdev = rcu_dereference(conf->disks[i].replacement);
574 575 576 577 578 579
		smp_mb(); /* Ensure that if rrdev is NULL, rdev won't be */
		rdev = rcu_dereference(conf->disks[i].rdev);
		if (!rdev) {
			rdev = rrdev;
			rrdev = NULL;
		}
580 581 582
		if (rw & WRITE) {
			if (replace_only)
				rdev = NULL;
583 584 585
			if (rdev == rrdev)
				/* We raced and saw duplicates */
				rrdev = NULL;
586
		} else {
587
			if (test_bit(R5_ReadRepl, &sh->dev[i].flags) && rrdev)
588 589 590
				rdev = rrdev;
			rrdev = NULL;
		}
591

592 593 594 595
		if (rdev && test_bit(Faulty, &rdev->flags))
			rdev = NULL;
		if (rdev)
			atomic_inc(&rdev->nr_pending);
596 597 598 599
		if (rrdev && test_bit(Faulty, &rrdev->flags))
			rrdev = NULL;
		if (rrdev)
			atomic_inc(&rrdev->nr_pending);
600 601
		rcu_read_unlock();

602
		/* We have already checked bad blocks for reads.  Now
603 604
		 * need to check for writes.  We never accept write errors
		 * on the replacement, so we don't to check rrdev.
605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624
		 */
		while ((rw & WRITE) && rdev &&
		       test_bit(WriteErrorSeen, &rdev->flags)) {
			sector_t first_bad;
			int bad_sectors;
			int bad = is_badblock(rdev, sh->sector, STRIPE_SECTORS,
					      &first_bad, &bad_sectors);
			if (!bad)
				break;

			if (bad < 0) {
				set_bit(BlockedBadBlocks, &rdev->flags);
				if (!conf->mddev->external &&
				    conf->mddev->flags) {
					/* It is very unlikely, but we might
					 * still need to write out the
					 * bad block log - better give it
					 * a chance*/
					md_check_recovery(conf->mddev);
				}
625 626 627 628 629 630
				/*
				 * Because md_wait_for_blocked_rdev
				 * will dec nr_pending, we must
				 * increment it first.
				 */
				atomic_inc(&rdev->nr_pending);
631 632 633 634 635 636 637 638
				md_wait_for_blocked_rdev(rdev, conf->mddev);
			} else {
				/* Acknowledged bad block - skip the write */
				rdev_dec_pending(rdev, conf->mddev);
				rdev = NULL;
			}
		}

639
		if (rdev) {
640 641
			if (s->syncing || s->expanding || s->expanded
			    || s->replacing)
642 643
				md_sync_acct(rdev->bdev, STRIPE_SECTORS);

D
Dan Williams 已提交
644 645
			set_bit(STRIPE_IO_STARTED, &sh->state);

K
Kent Overstreet 已提交
646
			bio_reset(bi);
647
			bi->bi_bdev = rdev->bdev;
K
Kent Overstreet 已提交
648 649 650 651 652 653
			bi->bi_rw = rw;
			bi->bi_end_io = (rw & WRITE)
				? raid5_end_write_request
				: raid5_end_read_request;
			bi->bi_private = sh;

654
			pr_debug("%s: for %llu schedule op %ld on disc %d\n",
655
				__func__, (unsigned long long)sh->sector,
656 657
				bi->bi_rw, i);
			atomic_inc(&sh->count);
658 659 660 661 662 663
			if (use_new_offset(conf, sh))
				bi->bi_sector = (sh->sector
						 + rdev->new_data_offset);
			else
				bi->bi_sector = (sh->sector
						 + rdev->data_offset);
664 665 666
			if (test_bit(R5_ReadNoMerge, &sh->dev[i].flags))
				bi->bi_rw |= REQ_FLUSH;

K
Kent Overstreet 已提交
667
			bi->bi_vcnt = 1;
668 669 670
			bi->bi_io_vec[0].bv_len = STRIPE_SIZE;
			bi->bi_io_vec[0].bv_offset = 0;
			bi->bi_size = STRIPE_SIZE;
671 672
			if (rrdev)
				set_bit(R5_DOUBLE_LOCKED, &sh->dev[i].flags);
673 674 675 676 677

			if (conf->mddev->gendisk)
				trace_block_bio_remap(bdev_get_queue(bi->bi_bdev),
						      bi, disk_devt(conf->mddev->gendisk),
						      sh->dev[i].sector);
678
			generic_make_request(bi);
679 680
		}
		if (rrdev) {
681 682
			if (s->syncing || s->expanding || s->expanded
			    || s->replacing)
683 684 685 686
				md_sync_acct(rrdev->bdev, STRIPE_SECTORS);

			set_bit(STRIPE_IO_STARTED, &sh->state);

K
Kent Overstreet 已提交
687
			bio_reset(rbi);
688
			rbi->bi_bdev = rrdev->bdev;
K
Kent Overstreet 已提交
689 690 691 692 693
			rbi->bi_rw = rw;
			BUG_ON(!(rw & WRITE));
			rbi->bi_end_io = raid5_end_write_request;
			rbi->bi_private = sh;

694 695 696 697 698
			pr_debug("%s: for %llu schedule op %ld on "
				 "replacement disc %d\n",
				__func__, (unsigned long long)sh->sector,
				rbi->bi_rw, i);
			atomic_inc(&sh->count);
699 700 701 702 703 704
			if (use_new_offset(conf, sh))
				rbi->bi_sector = (sh->sector
						  + rrdev->new_data_offset);
			else
				rbi->bi_sector = (sh->sector
						  + rrdev->data_offset);
K
Kent Overstreet 已提交
705
			rbi->bi_vcnt = 1;
706 707 708
			rbi->bi_io_vec[0].bv_len = STRIPE_SIZE;
			rbi->bi_io_vec[0].bv_offset = 0;
			rbi->bi_size = STRIPE_SIZE;
709 710 711 712
			if (conf->mddev->gendisk)
				trace_block_bio_remap(bdev_get_queue(rbi->bi_bdev),
						      rbi, disk_devt(conf->mddev->gendisk),
						      sh->dev[i].sector);
713 714 715
			generic_make_request(rbi);
		}
		if (!rdev && !rrdev) {
716
			if (rw & WRITE)
717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733
				set_bit(STRIPE_DEGRADED, &sh->state);
			pr_debug("skip op %ld on disc %d for sector %llu\n",
				bi->bi_rw, i, (unsigned long long)sh->sector);
			clear_bit(R5_LOCKED, &sh->dev[i].flags);
			set_bit(STRIPE_HANDLE, &sh->state);
		}
	}
}

static struct dma_async_tx_descriptor *
async_copy_data(int frombio, struct bio *bio, struct page *page,
	sector_t sector, struct dma_async_tx_descriptor *tx)
{
	struct bio_vec *bvl;
	struct page *bio_page;
	int i;
	int page_offset;
734
	struct async_submit_ctl submit;
D
Dan Williams 已提交
735
	enum async_tx_flags flags = 0;
736 737 738 739 740

	if (bio->bi_sector >= sector)
		page_offset = (signed)(bio->bi_sector - sector) * 512;
	else
		page_offset = (signed)(sector - bio->bi_sector) * -512;
741

D
Dan Williams 已提交
742 743 744 745
	if (frombio)
		flags |= ASYNC_TX_FENCE;
	init_async_submit(&submit, flags, tx, NULL, NULL, NULL);

746
	bio_for_each_segment(bvl, bio, i) {
747
		int len = bvl->bv_len;
748 749 750 751 752 753 754 755 756 757 758 759 760 761 762
		int clen;
		int b_offset = 0;

		if (page_offset < 0) {
			b_offset = -page_offset;
			page_offset += b_offset;
			len -= b_offset;
		}

		if (len > 0 && page_offset + len > STRIPE_SIZE)
			clen = STRIPE_SIZE - page_offset;
		else
			clen = len;

		if (clen > 0) {
763 764
			b_offset += bvl->bv_offset;
			bio_page = bvl->bv_page;
765 766
			if (frombio)
				tx = async_memcpy(page, bio_page, page_offset,
767
						  b_offset, clen, &submit);
768 769
			else
				tx = async_memcpy(bio_page, page, b_offset,
770
						  page_offset, clen, &submit);
771
		}
772 773 774
		/* chain the operations */
		submit.depend_tx = tx;

775 776 777 778 779 780 781 782 783 784 785 786
		if (clen < len) /* hit end of page */
			break;
		page_offset +=  len;
	}

	return tx;
}

static void ops_complete_biofill(void *stripe_head_ref)
{
	struct stripe_head *sh = stripe_head_ref;
	struct bio *return_bi = NULL;
787
	int i;
788

789
	pr_debug("%s: stripe %llu\n", __func__,
790 791 792 793 794 795 796
		(unsigned long long)sh->sector);

	/* clear completed biofills */
	for (i = sh->disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];

		/* acknowledge completion of a biofill operation */
797 798
		/* and check if we need to reply to a read request,
		 * new R5_Wantfill requests are held off until
799
		 * !STRIPE_BIOFILL_RUN
800 801
		 */
		if (test_and_clear_bit(R5_Wantfill, &dev->flags)) {
802 803 804 805 806 807 808 809
			struct bio *rbi, *rbi2;

			BUG_ON(!dev->read);
			rbi = dev->read;
			dev->read = NULL;
			while (rbi && rbi->bi_sector <
				dev->sector + STRIPE_SECTORS) {
				rbi2 = r5_next_bio(rbi, dev->sector);
810
				if (!raid5_dec_bi_active_stripes(rbi)) {
811 812 813 814 815 816 817
					rbi->bi_next = return_bi;
					return_bi = rbi;
				}
				rbi = rbi2;
			}
		}
	}
818
	clear_bit(STRIPE_BIOFILL_RUN, &sh->state);
819 820 821

	return_io(return_bi);

822
	set_bit(STRIPE_HANDLE, &sh->state);
823 824 825 826 827 828
	release_stripe(sh);
}

static void ops_run_biofill(struct stripe_head *sh)
{
	struct dma_async_tx_descriptor *tx = NULL;
829
	struct async_submit_ctl submit;
830 831
	int i;

832
	pr_debug("%s: stripe %llu\n", __func__,
833 834 835 836 837 838
		(unsigned long long)sh->sector);

	for (i = sh->disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
		if (test_bit(R5_Wantfill, &dev->flags)) {
			struct bio *rbi;
S
Shaohua Li 已提交
839
			spin_lock_irq(&sh->stripe_lock);
840 841
			dev->read = rbi = dev->toread;
			dev->toread = NULL;
S
Shaohua Li 已提交
842
			spin_unlock_irq(&sh->stripe_lock);
843 844 845 846 847 848 849 850 851 852
			while (rbi && rbi->bi_sector <
				dev->sector + STRIPE_SECTORS) {
				tx = async_copy_data(0, rbi, dev->page,
					dev->sector, tx);
				rbi = r5_next_bio(rbi, dev->sector);
			}
		}
	}

	atomic_inc(&sh->count);
853 854
	init_async_submit(&submit, ASYNC_TX_ACK, tx, ops_complete_biofill, sh, NULL);
	async_trigger_callback(&submit);
855 856
}

857
static void mark_target_uptodate(struct stripe_head *sh, int target)
858
{
859
	struct r5dev *tgt;
860

861 862
	if (target < 0)
		return;
863

864
	tgt = &sh->dev[target];
865 866 867
	set_bit(R5_UPTODATE, &tgt->flags);
	BUG_ON(!test_bit(R5_Wantcompute, &tgt->flags));
	clear_bit(R5_Wantcompute, &tgt->flags);
868 869
}

870
static void ops_complete_compute(void *stripe_head_ref)
871 872 873
{
	struct stripe_head *sh = stripe_head_ref;

874
	pr_debug("%s: stripe %llu\n", __func__,
875 876
		(unsigned long long)sh->sector);

877
	/* mark the computed target(s) as uptodate */
878
	mark_target_uptodate(sh, sh->ops.target);
879
	mark_target_uptodate(sh, sh->ops.target2);
880

881 882 883
	clear_bit(STRIPE_COMPUTE_RUN, &sh->state);
	if (sh->check_state == check_state_compute_run)
		sh->check_state = check_state_compute_result;
884 885 886 887
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

888 889 890 891 892 893 894 895 896
/* return a pointer to the address conversion region of the scribble buffer */
static addr_conv_t *to_addr_conv(struct stripe_head *sh,
				 struct raid5_percpu *percpu)
{
	return percpu->scribble + sizeof(struct page *) * (sh->disks + 2);
}

static struct dma_async_tx_descriptor *
ops_run_compute5(struct stripe_head *sh, struct raid5_percpu *percpu)
897 898
{
	int disks = sh->disks;
899
	struct page **xor_srcs = percpu->scribble;
900 901 902 903 904
	int target = sh->ops.target;
	struct r5dev *tgt = &sh->dev[target];
	struct page *xor_dest = tgt->page;
	int count = 0;
	struct dma_async_tx_descriptor *tx;
905
	struct async_submit_ctl submit;
906 907 908
	int i;

	pr_debug("%s: stripe %llu block: %d\n",
909
		__func__, (unsigned long long)sh->sector, target);
910 911 912 913 914 915 916 917
	BUG_ON(!test_bit(R5_Wantcompute, &tgt->flags));

	for (i = disks; i--; )
		if (i != target)
			xor_srcs[count++] = sh->dev[i].page;

	atomic_inc(&sh->count);

D
Dan Williams 已提交
918
	init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST, NULL,
919
			  ops_complete_compute, sh, to_addr_conv(sh, percpu));
920
	if (unlikely(count == 1))
921
		tx = async_memcpy(xor_dest, xor_srcs[0], 0, 0, STRIPE_SIZE, &submit);
922
	else
923
		tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, &submit);
924 925 926 927

	return tx;
}

928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945
/* set_syndrome_sources - populate source buffers for gen_syndrome
 * @srcs - (struct page *) array of size sh->disks
 * @sh - stripe_head to parse
 *
 * Populates srcs in proper layout order for the stripe and returns the
 * 'count' of sources to be used in a call to async_gen_syndrome.  The P
 * destination buffer is recorded in srcs[count] and the Q destination
 * is recorded in srcs[count+1]].
 */
static int set_syndrome_sources(struct page **srcs, struct stripe_head *sh)
{
	int disks = sh->disks;
	int syndrome_disks = sh->ddf_layout ? disks : (disks - 2);
	int d0_idx = raid6_d0(sh);
	int count;
	int i;

	for (i = 0; i < disks; i++)
946
		srcs[i] = NULL;
947 948 949 950 951 952 953 954 955 956

	count = 0;
	i = d0_idx;
	do {
		int slot = raid6_idx_to_slot(i, sh, &count, syndrome_disks);

		srcs[slot] = sh->dev[i].page;
		i = raid6_next_disk(i, disks);
	} while (i != d0_idx);

957
	return syndrome_disks;
958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977
}

static struct dma_async_tx_descriptor *
ops_run_compute6_1(struct stripe_head *sh, struct raid5_percpu *percpu)
{
	int disks = sh->disks;
	struct page **blocks = percpu->scribble;
	int target;
	int qd_idx = sh->qd_idx;
	struct dma_async_tx_descriptor *tx;
	struct async_submit_ctl submit;
	struct r5dev *tgt;
	struct page *dest;
	int i;
	int count;

	if (sh->ops.target < 0)
		target = sh->ops.target2;
	else if (sh->ops.target2 < 0)
		target = sh->ops.target;
978
	else
979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994
		/* we should only have one valid target */
		BUG();
	BUG_ON(target < 0);
	pr_debug("%s: stripe %llu block: %d\n",
		__func__, (unsigned long long)sh->sector, target);

	tgt = &sh->dev[target];
	BUG_ON(!test_bit(R5_Wantcompute, &tgt->flags));
	dest = tgt->page;

	atomic_inc(&sh->count);

	if (target == qd_idx) {
		count = set_syndrome_sources(blocks, sh);
		blocks[count] = NULL; /* regenerating p is not necessary */
		BUG_ON(blocks[count+1] != dest); /* q should already be set */
D
Dan Williams 已提交
995 996
		init_async_submit(&submit, ASYNC_TX_FENCE, NULL,
				  ops_complete_compute, sh,
997 998 999 1000 1001 1002 1003 1004 1005 1006 1007
				  to_addr_conv(sh, percpu));
		tx = async_gen_syndrome(blocks, 0, count+2, STRIPE_SIZE, &submit);
	} else {
		/* Compute any data- or p-drive using XOR */
		count = 0;
		for (i = disks; i-- ; ) {
			if (i == target || i == qd_idx)
				continue;
			blocks[count++] = sh->dev[i].page;
		}

D
Dan Williams 已提交
1008 1009
		init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST,
				  NULL, ops_complete_compute, sh,
1010 1011 1012
				  to_addr_conv(sh, percpu));
		tx = async_xor(dest, blocks, 0, count, STRIPE_SIZE, &submit);
	}
1013 1014 1015 1016

	return tx;
}

1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037
static struct dma_async_tx_descriptor *
ops_run_compute6_2(struct stripe_head *sh, struct raid5_percpu *percpu)
{
	int i, count, disks = sh->disks;
	int syndrome_disks = sh->ddf_layout ? disks : disks-2;
	int d0_idx = raid6_d0(sh);
	int faila = -1, failb = -1;
	int target = sh->ops.target;
	int target2 = sh->ops.target2;
	struct r5dev *tgt = &sh->dev[target];
	struct r5dev *tgt2 = &sh->dev[target2];
	struct dma_async_tx_descriptor *tx;
	struct page **blocks = percpu->scribble;
	struct async_submit_ctl submit;

	pr_debug("%s: stripe %llu block1: %d block2: %d\n",
		 __func__, (unsigned long long)sh->sector, target, target2);
	BUG_ON(target < 0 || target2 < 0);
	BUG_ON(!test_bit(R5_Wantcompute, &tgt->flags));
	BUG_ON(!test_bit(R5_Wantcompute, &tgt2->flags));

1038
	/* we need to open-code set_syndrome_sources to handle the
1039 1040 1041
	 * slot number conversion for 'faila' and 'failb'
	 */
	for (i = 0; i < disks ; i++)
1042
		blocks[i] = NULL;
1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068
	count = 0;
	i = d0_idx;
	do {
		int slot = raid6_idx_to_slot(i, sh, &count, syndrome_disks);

		blocks[slot] = sh->dev[i].page;

		if (i == target)
			faila = slot;
		if (i == target2)
			failb = slot;
		i = raid6_next_disk(i, disks);
	} while (i != d0_idx);

	BUG_ON(faila == failb);
	if (failb < faila)
		swap(faila, failb);
	pr_debug("%s: stripe: %llu faila: %d failb: %d\n",
		 __func__, (unsigned long long)sh->sector, faila, failb);

	atomic_inc(&sh->count);

	if (failb == syndrome_disks+1) {
		/* Q disk is one of the missing disks */
		if (faila == syndrome_disks) {
			/* Missing P+Q, just recompute */
D
Dan Williams 已提交
1069 1070 1071
			init_async_submit(&submit, ASYNC_TX_FENCE, NULL,
					  ops_complete_compute, sh,
					  to_addr_conv(sh, percpu));
1072
			return async_gen_syndrome(blocks, 0, syndrome_disks+2,
1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091
						  STRIPE_SIZE, &submit);
		} else {
			struct page *dest;
			int data_target;
			int qd_idx = sh->qd_idx;

			/* Missing D+Q: recompute D from P, then recompute Q */
			if (target == qd_idx)
				data_target = target2;
			else
				data_target = target;

			count = 0;
			for (i = disks; i-- ; ) {
				if (i == data_target || i == qd_idx)
					continue;
				blocks[count++] = sh->dev[i].page;
			}
			dest = sh->dev[data_target].page;
D
Dan Williams 已提交
1092 1093 1094 1095
			init_async_submit(&submit,
					  ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST,
					  NULL, NULL, NULL,
					  to_addr_conv(sh, percpu));
1096 1097 1098 1099
			tx = async_xor(dest, blocks, 0, count, STRIPE_SIZE,
				       &submit);

			count = set_syndrome_sources(blocks, sh);
D
Dan Williams 已提交
1100 1101 1102
			init_async_submit(&submit, ASYNC_TX_FENCE, tx,
					  ops_complete_compute, sh,
					  to_addr_conv(sh, percpu));
1103 1104 1105 1106
			return async_gen_syndrome(blocks, 0, count+2,
						  STRIPE_SIZE, &submit);
		}
	} else {
1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120
		init_async_submit(&submit, ASYNC_TX_FENCE, NULL,
				  ops_complete_compute, sh,
				  to_addr_conv(sh, percpu));
		if (failb == syndrome_disks) {
			/* We're missing D+P. */
			return async_raid6_datap_recov(syndrome_disks+2,
						       STRIPE_SIZE, faila,
						       blocks, &submit);
		} else {
			/* We're missing D+D. */
			return async_raid6_2data_recov(syndrome_disks+2,
						       STRIPE_SIZE, faila, failb,
						       blocks, &submit);
		}
1121 1122 1123 1124
	}
}


1125 1126 1127 1128
static void ops_complete_prexor(void *stripe_head_ref)
{
	struct stripe_head *sh = stripe_head_ref;

1129
	pr_debug("%s: stripe %llu\n", __func__,
1130 1131 1132 1133
		(unsigned long long)sh->sector);
}

static struct dma_async_tx_descriptor *
1134 1135
ops_run_prexor(struct stripe_head *sh, struct raid5_percpu *percpu,
	       struct dma_async_tx_descriptor *tx)
1136 1137
{
	int disks = sh->disks;
1138
	struct page **xor_srcs = percpu->scribble;
1139
	int count = 0, pd_idx = sh->pd_idx, i;
1140
	struct async_submit_ctl submit;
1141 1142 1143 1144

	/* existing parity data subtracted */
	struct page *xor_dest = xor_srcs[count++] = sh->dev[pd_idx].page;

1145
	pr_debug("%s: stripe %llu\n", __func__,
1146 1147 1148 1149 1150
		(unsigned long long)sh->sector);

	for (i = disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
		/* Only process blocks that are known to be uptodate */
1151
		if (test_bit(R5_Wantdrain, &dev->flags))
1152 1153 1154
			xor_srcs[count++] = dev->page;
	}

D
Dan Williams 已提交
1155
	init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_DROP_DST, tx,
1156
			  ops_complete_prexor, sh, to_addr_conv(sh, percpu));
1157
	tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, &submit);
1158 1159 1160 1161 1162

	return tx;
}

static struct dma_async_tx_descriptor *
1163
ops_run_biodrain(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
1164 1165
{
	int disks = sh->disks;
1166
	int i;
1167

1168
	pr_debug("%s: stripe %llu\n", __func__,
1169 1170 1171 1172 1173 1174
		(unsigned long long)sh->sector);

	for (i = disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
		struct bio *chosen;

1175
		if (test_and_clear_bit(R5_Wantdrain, &dev->flags)) {
1176 1177
			struct bio *wbi;

S
Shaohua Li 已提交
1178
			spin_lock_irq(&sh->stripe_lock);
1179 1180 1181 1182
			chosen = dev->towrite;
			dev->towrite = NULL;
			BUG_ON(dev->written);
			wbi = dev->written = chosen;
S
Shaohua Li 已提交
1183
			spin_unlock_irq(&sh->stripe_lock);
1184 1185 1186

			while (wbi && wbi->bi_sector <
				dev->sector + STRIPE_SECTORS) {
T
Tejun Heo 已提交
1187 1188
				if (wbi->bi_rw & REQ_FUA)
					set_bit(R5_WantFUA, &dev->flags);
S
Shaohua Li 已提交
1189 1190
				if (wbi->bi_rw & REQ_SYNC)
					set_bit(R5_SyncIO, &dev->flags);
1191
				if (wbi->bi_rw & REQ_DISCARD)
S
Shaohua Li 已提交
1192
					set_bit(R5_Discard, &dev->flags);
1193
				else
S
Shaohua Li 已提交
1194 1195
					tx = async_copy_data(1, wbi, dev->page,
						dev->sector, tx);
1196 1197 1198 1199 1200 1201 1202 1203
				wbi = r5_next_bio(wbi, dev->sector);
			}
		}
	}

	return tx;
}

1204
static void ops_complete_reconstruct(void *stripe_head_ref)
1205 1206
{
	struct stripe_head *sh = stripe_head_ref;
1207 1208 1209 1210
	int disks = sh->disks;
	int pd_idx = sh->pd_idx;
	int qd_idx = sh->qd_idx;
	int i;
1211
	bool fua = false, sync = false, discard = false;
1212

1213
	pr_debug("%s: stripe %llu\n", __func__,
1214 1215
		(unsigned long long)sh->sector);

S
Shaohua Li 已提交
1216
	for (i = disks; i--; ) {
T
Tejun Heo 已提交
1217
		fua |= test_bit(R5_WantFUA, &sh->dev[i].flags);
S
Shaohua Li 已提交
1218
		sync |= test_bit(R5_SyncIO, &sh->dev[i].flags);
1219
		discard |= test_bit(R5_Discard, &sh->dev[i].flags);
S
Shaohua Li 已提交
1220
	}
T
Tejun Heo 已提交
1221

1222 1223
	for (i = disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
1224

T
Tejun Heo 已提交
1225
		if (dev->written || i == pd_idx || i == qd_idx) {
1226 1227
			if (!discard)
				set_bit(R5_UPTODATE, &dev->flags);
T
Tejun Heo 已提交
1228 1229
			if (fua)
				set_bit(R5_WantFUA, &dev->flags);
S
Shaohua Li 已提交
1230 1231
			if (sync)
				set_bit(R5_SyncIO, &dev->flags);
T
Tejun Heo 已提交
1232
		}
1233 1234
	}

1235 1236 1237 1238 1239 1240 1241 1242
	if (sh->reconstruct_state == reconstruct_state_drain_run)
		sh->reconstruct_state = reconstruct_state_drain_result;
	else if (sh->reconstruct_state == reconstruct_state_prexor_drain_run)
		sh->reconstruct_state = reconstruct_state_prexor_drain_result;
	else {
		BUG_ON(sh->reconstruct_state != reconstruct_state_run);
		sh->reconstruct_state = reconstruct_state_result;
	}
1243 1244 1245 1246 1247 1248

	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

static void
1249 1250
ops_run_reconstruct5(struct stripe_head *sh, struct raid5_percpu *percpu,
		     struct dma_async_tx_descriptor *tx)
1251 1252
{
	int disks = sh->disks;
1253
	struct page **xor_srcs = percpu->scribble;
1254
	struct async_submit_ctl submit;
1255 1256
	int count = 0, pd_idx = sh->pd_idx, i;
	struct page *xor_dest;
1257
	int prexor = 0;
1258 1259
	unsigned long flags;

1260
	pr_debug("%s: stripe %llu\n", __func__,
1261 1262
		(unsigned long long)sh->sector);

S
Shaohua Li 已提交
1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274
	for (i = 0; i < sh->disks; i++) {
		if (pd_idx == i)
			continue;
		if (!test_bit(R5_Discard, &sh->dev[i].flags))
			break;
	}
	if (i >= sh->disks) {
		atomic_inc(&sh->count);
		set_bit(R5_Discard, &sh->dev[pd_idx].flags);
		ops_complete_reconstruct(sh);
		return;
	}
1275 1276 1277
	/* check if prexor is active which means only process blocks
	 * that are part of a read-modify-write (written)
	 */
1278 1279
	if (sh->reconstruct_state == reconstruct_state_prexor_drain_run) {
		prexor = 1;
1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299
		xor_dest = xor_srcs[count++] = sh->dev[pd_idx].page;
		for (i = disks; i--; ) {
			struct r5dev *dev = &sh->dev[i];
			if (dev->written)
				xor_srcs[count++] = dev->page;
		}
	} else {
		xor_dest = sh->dev[pd_idx].page;
		for (i = disks; i--; ) {
			struct r5dev *dev = &sh->dev[i];
			if (i != pd_idx)
				xor_srcs[count++] = dev->page;
		}
	}

	/* 1/ if we prexor'd then the dest is reused as a source
	 * 2/ if we did not prexor then we are redoing the parity
	 * set ASYNC_TX_XOR_DROP_DST and ASYNC_TX_XOR_ZERO_DST
	 * for the synchronous xor case
	 */
1300
	flags = ASYNC_TX_ACK |
1301 1302 1303 1304
		(prexor ? ASYNC_TX_XOR_DROP_DST : ASYNC_TX_XOR_ZERO_DST);

	atomic_inc(&sh->count);

1305
	init_async_submit(&submit, flags, tx, ops_complete_reconstruct, sh,
1306
			  to_addr_conv(sh, percpu));
1307 1308 1309 1310
	if (unlikely(count == 1))
		tx = async_memcpy(xor_dest, xor_srcs[0], 0, 0, STRIPE_SIZE, &submit);
	else
		tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, &submit);
1311 1312
}

1313 1314 1315 1316 1317 1318
static void
ops_run_reconstruct6(struct stripe_head *sh, struct raid5_percpu *percpu,
		     struct dma_async_tx_descriptor *tx)
{
	struct async_submit_ctl submit;
	struct page **blocks = percpu->scribble;
S
Shaohua Li 已提交
1319
	int count, i;
1320 1321 1322

	pr_debug("%s: stripe %llu\n", __func__, (unsigned long long)sh->sector);

S
Shaohua Li 已提交
1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336
	for (i = 0; i < sh->disks; i++) {
		if (sh->pd_idx == i || sh->qd_idx == i)
			continue;
		if (!test_bit(R5_Discard, &sh->dev[i].flags))
			break;
	}
	if (i >= sh->disks) {
		atomic_inc(&sh->count);
		set_bit(R5_Discard, &sh->dev[sh->pd_idx].flags);
		set_bit(R5_Discard, &sh->dev[sh->qd_idx].flags);
		ops_complete_reconstruct(sh);
		return;
	}

1337 1338 1339 1340 1341 1342 1343
	count = set_syndrome_sources(blocks, sh);

	atomic_inc(&sh->count);

	init_async_submit(&submit, ASYNC_TX_ACK, tx, ops_complete_reconstruct,
			  sh, to_addr_conv(sh, percpu));
	async_gen_syndrome(blocks, 0, count+2, STRIPE_SIZE,  &submit);
1344 1345 1346 1347 1348 1349
}

static void ops_complete_check(void *stripe_head_ref)
{
	struct stripe_head *sh = stripe_head_ref;

1350
	pr_debug("%s: stripe %llu\n", __func__,
1351 1352
		(unsigned long long)sh->sector);

1353
	sh->check_state = check_state_check_result;
1354 1355 1356 1357
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

1358
static void ops_run_check_p(struct stripe_head *sh, struct raid5_percpu *percpu)
1359 1360
{
	int disks = sh->disks;
1361 1362 1363
	int pd_idx = sh->pd_idx;
	int qd_idx = sh->qd_idx;
	struct page *xor_dest;
1364
	struct page **xor_srcs = percpu->scribble;
1365
	struct dma_async_tx_descriptor *tx;
1366
	struct async_submit_ctl submit;
1367 1368
	int count;
	int i;
1369

1370
	pr_debug("%s: stripe %llu\n", __func__,
1371 1372
		(unsigned long long)sh->sector);

1373 1374 1375
	count = 0;
	xor_dest = sh->dev[pd_idx].page;
	xor_srcs[count++] = xor_dest;
1376
	for (i = disks; i--; ) {
1377 1378 1379
		if (i == pd_idx || i == qd_idx)
			continue;
		xor_srcs[count++] = sh->dev[i].page;
1380 1381
	}

1382 1383
	init_async_submit(&submit, 0, NULL, NULL, NULL,
			  to_addr_conv(sh, percpu));
D
Dan Williams 已提交
1384
	tx = async_xor_val(xor_dest, xor_srcs, 0, count, STRIPE_SIZE,
1385
			   &sh->ops.zero_sum_result, &submit);
1386 1387

	atomic_inc(&sh->count);
1388 1389
	init_async_submit(&submit, ASYNC_TX_ACK, tx, ops_complete_check, sh, NULL);
	tx = async_trigger_callback(&submit);
1390 1391
}

1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403
static void ops_run_check_pq(struct stripe_head *sh, struct raid5_percpu *percpu, int checkp)
{
	struct page **srcs = percpu->scribble;
	struct async_submit_ctl submit;
	int count;

	pr_debug("%s: stripe %llu checkp: %d\n", __func__,
		(unsigned long long)sh->sector, checkp);

	count = set_syndrome_sources(srcs, sh);
	if (!checkp)
		srcs[count] = NULL;
1404 1405

	atomic_inc(&sh->count);
1406 1407 1408 1409
	init_async_submit(&submit, ASYNC_TX_ACK, NULL, ops_complete_check,
			  sh, to_addr_conv(sh, percpu));
	async_syndrome_val(srcs, 0, count+2, STRIPE_SIZE,
			   &sh->ops.zero_sum_result, percpu->spare_page, &submit);
1410 1411
}

N
NeilBrown 已提交
1412
static void raid_run_ops(struct stripe_head *sh, unsigned long ops_request)
1413 1414 1415
{
	int overlap_clear = 0, i, disks = sh->disks;
	struct dma_async_tx_descriptor *tx = NULL;
1416
	struct r5conf *conf = sh->raid_conf;
1417
	int level = conf->level;
1418 1419
	struct raid5_percpu *percpu;
	unsigned long cpu;
1420

1421 1422
	cpu = get_cpu();
	percpu = per_cpu_ptr(conf->percpu, cpu);
1423
	if (test_bit(STRIPE_OP_BIOFILL, &ops_request)) {
1424 1425 1426 1427
		ops_run_biofill(sh);
		overlap_clear++;
	}

1428
	if (test_bit(STRIPE_OP_COMPUTE_BLK, &ops_request)) {
1429 1430 1431 1432 1433 1434 1435 1436 1437 1438
		if (level < 6)
			tx = ops_run_compute5(sh, percpu);
		else {
			if (sh->ops.target2 < 0 || sh->ops.target < 0)
				tx = ops_run_compute6_1(sh, percpu);
			else
				tx = ops_run_compute6_2(sh, percpu);
		}
		/* terminate the chain if reconstruct is not set to be run */
		if (tx && !test_bit(STRIPE_OP_RECONSTRUCT, &ops_request))
1439 1440
			async_tx_ack(tx);
	}
1441

1442
	if (test_bit(STRIPE_OP_PREXOR, &ops_request))
1443
		tx = ops_run_prexor(sh, percpu, tx);
1444

1445
	if (test_bit(STRIPE_OP_BIODRAIN, &ops_request)) {
1446
		tx = ops_run_biodrain(sh, tx);
1447 1448 1449
		overlap_clear++;
	}

1450 1451 1452 1453 1454 1455
	if (test_bit(STRIPE_OP_RECONSTRUCT, &ops_request)) {
		if (level < 6)
			ops_run_reconstruct5(sh, percpu, tx);
		else
			ops_run_reconstruct6(sh, percpu, tx);
	}
1456

1457 1458 1459 1460 1461 1462 1463 1464 1465 1466
	if (test_bit(STRIPE_OP_CHECK, &ops_request)) {
		if (sh->check_state == check_state_run)
			ops_run_check_p(sh, percpu);
		else if (sh->check_state == check_state_run_q)
			ops_run_check_pq(sh, percpu, 0);
		else if (sh->check_state == check_state_run_pq)
			ops_run_check_pq(sh, percpu, 1);
		else
			BUG();
	}
1467 1468 1469 1470 1471 1472 1473

	if (overlap_clear)
		for (i = disks; i--; ) {
			struct r5dev *dev = &sh->dev[i];
			if (test_and_clear_bit(R5_Overlap, &dev->flags))
				wake_up(&sh->raid_conf->wait_for_overlap);
		}
1474
	put_cpu();
1475 1476
}

1477
static int grow_one_stripe(struct r5conf *conf)
L
Linus Torvalds 已提交
1478 1479
{
	struct stripe_head *sh;
N
Namhyung Kim 已提交
1480
	sh = kmem_cache_zalloc(conf->slab_cache, GFP_KERNEL);
1481 1482
	if (!sh)
		return 0;
N
Namhyung Kim 已提交
1483

1484 1485
	sh->raid_conf = conf;

S
Shaohua Li 已提交
1486 1487
	spin_lock_init(&sh->stripe_lock);

1488 1489
	if (grow_buffers(sh)) {
		shrink_buffers(sh);
1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500
		kmem_cache_free(conf->slab_cache, sh);
		return 0;
	}
	/* we just created an active stripe so... */
	atomic_set(&sh->count, 1);
	atomic_inc(&conf->active_stripes);
	INIT_LIST_HEAD(&sh->lru);
	release_stripe(sh);
	return 1;
}

1501
static int grow_stripes(struct r5conf *conf, int num)
1502
{
1503
	struct kmem_cache *sc;
1504
	int devs = max(conf->raid_disks, conf->previous_raid_disks);
L
Linus Torvalds 已提交
1505

1506 1507 1508 1509 1510 1511 1512 1513
	if (conf->mddev->gendisk)
		sprintf(conf->cache_name[0],
			"raid%d-%s", conf->level, mdname(conf->mddev));
	else
		sprintf(conf->cache_name[0],
			"raid%d-%p", conf->level, conf->mddev);
	sprintf(conf->cache_name[1], "%s-alt", conf->cache_name[0]);

1514 1515
	conf->active_name = 0;
	sc = kmem_cache_create(conf->cache_name[conf->active_name],
L
Linus Torvalds 已提交
1516
			       sizeof(struct stripe_head)+(devs-1)*sizeof(struct r5dev),
1517
			       0, 0, NULL);
L
Linus Torvalds 已提交
1518 1519 1520
	if (!sc)
		return 1;
	conf->slab_cache = sc;
1521
	conf->pool_size = devs;
1522
	while (num--)
1523
		if (!grow_one_stripe(conf))
L
Linus Torvalds 已提交
1524 1525 1526
			return 1;
	return 0;
}
1527

1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549
/**
 * scribble_len - return the required size of the scribble region
 * @num - total number of disks in the array
 *
 * The size must be enough to contain:
 * 1/ a struct page pointer for each device in the array +2
 * 2/ room to convert each entry in (1) to its corresponding dma
 *    (dma_map_page()) or page (page_address()) address.
 *
 * Note: the +2 is for the destination buffers of the ddf/raid6 case where we
 * calculate over all devices (not just the data blocks), using zeros in place
 * of the P and Q blocks.
 */
static size_t scribble_len(int num)
{
	size_t len;

	len = sizeof(struct page *) * (num+2) + sizeof(addr_conv_t) * (num+2);

	return len;
}

1550
static int resize_stripes(struct r5conf *conf, int newsize)
1551 1552 1553 1554 1555 1556 1557
{
	/* Make all the stripes able to hold 'newsize' devices.
	 * New slots in each stripe get 'page' set to a new page.
	 *
	 * This happens in stages:
	 * 1/ create a new kmem_cache and allocate the required number of
	 *    stripe_heads.
M
Masanari Iida 已提交
1558
	 * 2/ gather all the old stripe_heads and transfer the pages across
1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577
	 *    to the new stripe_heads.  This will have the side effect of
	 *    freezing the array as once all stripe_heads have been collected,
	 *    no IO will be possible.  Old stripe heads are freed once their
	 *    pages have been transferred over, and the old kmem_cache is
	 *    freed when all stripes are done.
	 * 3/ reallocate conf->disks to be suitable bigger.  If this fails,
	 *    we simple return a failre status - no need to clean anything up.
	 * 4/ allocate new pages for the new slots in the new stripe_heads.
	 *    If this fails, we don't bother trying the shrink the
	 *    stripe_heads down again, we just leave them as they are.
	 *    As each stripe_head is processed the new one is released into
	 *    active service.
	 *
	 * Once step2 is started, we cannot afford to wait for a write,
	 * so we use GFP_NOIO allocations.
	 */
	struct stripe_head *osh, *nsh;
	LIST_HEAD(newstripes);
	struct disk_info *ndisks;
1578
	unsigned long cpu;
1579
	int err;
1580
	struct kmem_cache *sc;
1581 1582 1583 1584 1585
	int i;

	if (newsize <= conf->pool_size)
		return 0; /* never bother to shrink */

1586 1587 1588
	err = md_allow_write(conf->mddev);
	if (err)
		return err;
1589

1590 1591 1592
	/* Step 1 */
	sc = kmem_cache_create(conf->cache_name[1-conf->active_name],
			       sizeof(struct stripe_head)+(newsize-1)*sizeof(struct r5dev),
1593
			       0, 0, NULL);
1594 1595 1596 1597
	if (!sc)
		return -ENOMEM;

	for (i = conf->max_nr_stripes; i; i--) {
N
Namhyung Kim 已提交
1598
		nsh = kmem_cache_zalloc(sc, GFP_KERNEL);
1599 1600 1601 1602
		if (!nsh)
			break;

		nsh->raid_conf = conf;
1603
		spin_lock_init(&nsh->stripe_lock);
1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624

		list_add(&nsh->lru, &newstripes);
	}
	if (i) {
		/* didn't get enough, give up */
		while (!list_empty(&newstripes)) {
			nsh = list_entry(newstripes.next, struct stripe_head, lru);
			list_del(&nsh->lru);
			kmem_cache_free(sc, nsh);
		}
		kmem_cache_destroy(sc);
		return -ENOMEM;
	}
	/* Step 2 - Must use GFP_NOIO now.
	 * OK, we have enough stripes, start collecting inactive
	 * stripes and copying them over
	 */
	list_for_each_entry(nsh, &newstripes, lru) {
		spin_lock_irq(&conf->device_lock);
		wait_event_lock_irq(conf->wait_for_stripe,
				    !list_empty(&conf->inactive_list),
1625
				    conf->device_lock);
1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639
		osh = get_free_stripe(conf);
		spin_unlock_irq(&conf->device_lock);
		atomic_set(&nsh->count, 1);
		for(i=0; i<conf->pool_size; i++)
			nsh->dev[i].page = osh->dev[i].page;
		for( ; i<newsize; i++)
			nsh->dev[i].page = NULL;
		kmem_cache_free(conf->slab_cache, osh);
	}
	kmem_cache_destroy(conf->slab_cache);

	/* Step 3.
	 * At this point, we are holding all the stripes so the array
	 * is completely stalled, so now is a good time to resize
1640
	 * conf->disks and the scribble region
1641 1642 1643 1644 1645 1646 1647 1648 1649 1650
	 */
	ndisks = kzalloc(newsize * sizeof(struct disk_info), GFP_NOIO);
	if (ndisks) {
		for (i=0; i<conf->raid_disks; i++)
			ndisks[i] = conf->disks[i];
		kfree(conf->disks);
		conf->disks = ndisks;
	} else
		err = -ENOMEM;

1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669
	get_online_cpus();
	conf->scribble_len = scribble_len(newsize);
	for_each_present_cpu(cpu) {
		struct raid5_percpu *percpu;
		void *scribble;

		percpu = per_cpu_ptr(conf->percpu, cpu);
		scribble = kmalloc(conf->scribble_len, GFP_NOIO);

		if (scribble) {
			kfree(percpu->scribble);
			percpu->scribble = scribble;
		} else {
			err = -ENOMEM;
			break;
		}
	}
	put_online_cpus();

1670 1671 1672 1673
	/* Step 4, return new stripes to service */
	while(!list_empty(&newstripes)) {
		nsh = list_entry(newstripes.next, struct stripe_head, lru);
		list_del_init(&nsh->lru);
1674

1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690
		for (i=conf->raid_disks; i < newsize; i++)
			if (nsh->dev[i].page == NULL) {
				struct page *p = alloc_page(GFP_NOIO);
				nsh->dev[i].page = p;
				if (!p)
					err = -ENOMEM;
			}
		release_stripe(nsh);
	}
	/* critical section pass, GFP_NOIO no longer needed */

	conf->slab_cache = sc;
	conf->active_name = 1-conf->active_name;
	conf->pool_size = newsize;
	return err;
}
L
Linus Torvalds 已提交
1691

1692
static int drop_one_stripe(struct r5conf *conf)
L
Linus Torvalds 已提交
1693 1694 1695
{
	struct stripe_head *sh;

1696 1697 1698 1699 1700
	spin_lock_irq(&conf->device_lock);
	sh = get_free_stripe(conf);
	spin_unlock_irq(&conf->device_lock);
	if (!sh)
		return 0;
1701
	BUG_ON(atomic_read(&sh->count));
1702
	shrink_buffers(sh);
1703 1704 1705 1706 1707
	kmem_cache_free(conf->slab_cache, sh);
	atomic_dec(&conf->active_stripes);
	return 1;
}

1708
static void shrink_stripes(struct r5conf *conf)
1709 1710 1711 1712
{
	while (drop_one_stripe(conf))
		;

N
NeilBrown 已提交
1713 1714
	if (conf->slab_cache)
		kmem_cache_destroy(conf->slab_cache);
L
Linus Torvalds 已提交
1715 1716 1717
	conf->slab_cache = NULL;
}

1718
static void raid5_end_read_request(struct bio * bi, int error)
L
Linus Torvalds 已提交
1719
{
1720
	struct stripe_head *sh = bi->bi_private;
1721
	struct r5conf *conf = sh->raid_conf;
1722
	int disks = sh->disks, i;
L
Linus Torvalds 已提交
1723
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
1724
	char b[BDEVNAME_SIZE];
1725
	struct md_rdev *rdev = NULL;
1726
	sector_t s;
L
Linus Torvalds 已提交
1727 1728 1729 1730 1731

	for (i=0 ; i<disks; i++)
		if (bi == &sh->dev[i].req)
			break;

1732 1733
	pr_debug("end_read_request %llu/%d, count: %d, uptodate %d.\n",
		(unsigned long long)sh->sector, i, atomic_read(&sh->count),
L
Linus Torvalds 已提交
1734 1735 1736
		uptodate);
	if (i == disks) {
		BUG();
1737
		return;
L
Linus Torvalds 已提交
1738
	}
1739
	if (test_bit(R5_ReadRepl, &sh->dev[i].flags))
1740 1741 1742 1743 1744
		/* If replacement finished while this request was outstanding,
		 * 'replacement' might be NULL already.
		 * In that case it moved down to 'rdev'.
		 * rdev is not removed until all requests are finished.
		 */
1745
		rdev = conf->disks[i].replacement;
1746
	if (!rdev)
1747
		rdev = conf->disks[i].rdev;
L
Linus Torvalds 已提交
1748

1749 1750 1751 1752
	if (use_new_offset(conf, sh))
		s = sh->sector + rdev->new_data_offset;
	else
		s = sh->sector + rdev->data_offset;
L
Linus Torvalds 已提交
1753 1754
	if (uptodate) {
		set_bit(R5_UPTODATE, &sh->dev[i].flags);
1755
		if (test_bit(R5_ReadError, &sh->dev[i].flags)) {
1756 1757 1758 1759
			/* Note that this cannot happen on a
			 * replacement device.  We just fail those on
			 * any error
			 */
1760 1761 1762 1763 1764
			printk_ratelimited(
				KERN_INFO
				"md/raid:%s: read error corrected"
				" (%lu sectors at %llu on %s)\n",
				mdname(conf->mddev), STRIPE_SECTORS,
1765
				(unsigned long long)s,
1766
				bdevname(rdev->bdev, b));
1767
			atomic_add(STRIPE_SECTORS, &rdev->corrected_errors);
1768 1769
			clear_bit(R5_ReadError, &sh->dev[i].flags);
			clear_bit(R5_ReWrite, &sh->dev[i].flags);
1770 1771 1772
		} else if (test_bit(R5_ReadNoMerge, &sh->dev[i].flags))
			clear_bit(R5_ReadNoMerge, &sh->dev[i].flags);

1773 1774
		if (atomic_read(&rdev->read_errors))
			atomic_set(&rdev->read_errors, 0);
L
Linus Torvalds 已提交
1775
	} else {
1776
		const char *bdn = bdevname(rdev->bdev, b);
1777
		int retry = 0;
1778
		int set_bad = 0;
1779

L
Linus Torvalds 已提交
1780
		clear_bit(R5_UPTODATE, &sh->dev[i].flags);
1781
		atomic_inc(&rdev->read_errors);
1782 1783 1784 1785 1786 1787
		if (test_bit(R5_ReadRepl, &sh->dev[i].flags))
			printk_ratelimited(
				KERN_WARNING
				"md/raid:%s: read error on replacement device "
				"(sector %llu on %s).\n",
				mdname(conf->mddev),
1788
				(unsigned long long)s,
1789
				bdn);
1790 1791
		else if (conf->mddev->degraded >= conf->max_degraded) {
			set_bad = 1;
1792 1793 1794 1795 1796
			printk_ratelimited(
				KERN_WARNING
				"md/raid:%s: read error not correctable "
				"(sector %llu on %s).\n",
				mdname(conf->mddev),
1797
				(unsigned long long)s,
1798
				bdn);
1799
		} else if (test_bit(R5_ReWrite, &sh->dev[i].flags)) {
1800
			/* Oh, no!!! */
1801
			set_bad = 1;
1802 1803 1804 1805 1806
			printk_ratelimited(
				KERN_WARNING
				"md/raid:%s: read error NOT corrected!! "
				"(sector %llu on %s).\n",
				mdname(conf->mddev),
1807
				(unsigned long long)s,
1808
				bdn);
1809
		} else if (atomic_read(&rdev->read_errors)
1810
			 > conf->max_nr_stripes)
N
NeilBrown 已提交
1811
			printk(KERN_WARNING
1812
			       "md/raid:%s: Too many read errors, failing device %s.\n",
1813
			       mdname(conf->mddev), bdn);
1814 1815 1816
		else
			retry = 1;
		if (retry)
1817 1818 1819 1820 1821
			if (test_bit(R5_ReadNoMerge, &sh->dev[i].flags)) {
				set_bit(R5_ReadError, &sh->dev[i].flags);
				clear_bit(R5_ReadNoMerge, &sh->dev[i].flags);
			} else
				set_bit(R5_ReadNoMerge, &sh->dev[i].flags);
1822
		else {
1823 1824
			clear_bit(R5_ReadError, &sh->dev[i].flags);
			clear_bit(R5_ReWrite, &sh->dev[i].flags);
1825 1826 1827 1828 1829
			if (!(set_bad
			      && test_bit(In_sync, &rdev->flags)
			      && rdev_set_badblocks(
				      rdev, sh->sector, STRIPE_SECTORS, 0)))
				md_error(conf->mddev, rdev);
1830
		}
L
Linus Torvalds 已提交
1831
	}
1832
	rdev_dec_pending(rdev, conf->mddev);
L
Linus Torvalds 已提交
1833 1834 1835 1836 1837
	clear_bit(R5_LOCKED, &sh->dev[i].flags);
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

1838
static void raid5_end_write_request(struct bio *bi, int error)
L
Linus Torvalds 已提交
1839
{
1840
	struct stripe_head *sh = bi->bi_private;
1841
	struct r5conf *conf = sh->raid_conf;
1842
	int disks = sh->disks, i;
1843
	struct md_rdev *uninitialized_var(rdev);
L
Linus Torvalds 已提交
1844
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
1845 1846
	sector_t first_bad;
	int bad_sectors;
1847
	int replacement = 0;
L
Linus Torvalds 已提交
1848

1849 1850 1851
	for (i = 0 ; i < disks; i++) {
		if (bi == &sh->dev[i].req) {
			rdev = conf->disks[i].rdev;
L
Linus Torvalds 已提交
1852
			break;
1853 1854 1855
		}
		if (bi == &sh->dev[i].rreq) {
			rdev = conf->disks[i].replacement;
1856 1857 1858 1859 1860 1861 1862 1863
			if (rdev)
				replacement = 1;
			else
				/* rdev was removed and 'replacement'
				 * replaced it.  rdev is not removed
				 * until all requests are finished.
				 */
				rdev = conf->disks[i].rdev;
1864 1865 1866
			break;
		}
	}
1867
	pr_debug("end_write_request %llu/%d, count %d, uptodate: %d.\n",
L
Linus Torvalds 已提交
1868 1869 1870 1871
		(unsigned long long)sh->sector, i, atomic_read(&sh->count),
		uptodate);
	if (i == disks) {
		BUG();
1872
		return;
L
Linus Torvalds 已提交
1873 1874
	}

1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885
	if (replacement) {
		if (!uptodate)
			md_error(conf->mddev, rdev);
		else if (is_badblock(rdev, sh->sector,
				     STRIPE_SECTORS,
				     &first_bad, &bad_sectors))
			set_bit(R5_MadeGoodRepl, &sh->dev[i].flags);
	} else {
		if (!uptodate) {
			set_bit(WriteErrorSeen, &rdev->flags);
			set_bit(R5_WriteError, &sh->dev[i].flags);
1886 1887 1888
			if (!test_and_set_bit(WantReplacement, &rdev->flags))
				set_bit(MD_RECOVERY_NEEDED,
					&rdev->mddev->recovery);
1889 1890
		} else if (is_badblock(rdev, sh->sector,
				       STRIPE_SECTORS,
1891
				       &first_bad, &bad_sectors)) {
1892
			set_bit(R5_MadeGood, &sh->dev[i].flags);
1893 1894 1895 1896 1897 1898 1899
			if (test_bit(R5_ReadError, &sh->dev[i].flags))
				/* That was a successful write so make
				 * sure it looks like we already did
				 * a re-write.
				 */
				set_bit(R5_ReWrite, &sh->dev[i].flags);
		}
1900 1901
	}
	rdev_dec_pending(rdev, conf->mddev);
L
Linus Torvalds 已提交
1902

1903 1904
	if (!test_and_clear_bit(R5_DOUBLE_LOCKED, &sh->dev[i].flags))
		clear_bit(R5_LOCKED, &sh->dev[i].flags);
L
Linus Torvalds 已提交
1905
	set_bit(STRIPE_HANDLE, &sh->state);
1906
	release_stripe(sh);
L
Linus Torvalds 已提交
1907 1908
}

1909
static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous);
L
Linus Torvalds 已提交
1910
	
1911
static void raid5_build_block(struct stripe_head *sh, int i, int previous)
L
Linus Torvalds 已提交
1912 1913 1914 1915 1916 1917 1918 1919
{
	struct r5dev *dev = &sh->dev[i];

	bio_init(&dev->req);
	dev->req.bi_io_vec = &dev->vec;
	dev->req.bi_vcnt++;
	dev->req.bi_max_vecs++;
	dev->req.bi_private = sh;
1920
	dev->vec.bv_page = dev->page;
L
Linus Torvalds 已提交
1921

1922 1923 1924 1925 1926 1927 1928
	bio_init(&dev->rreq);
	dev->rreq.bi_io_vec = &dev->rvec;
	dev->rreq.bi_vcnt++;
	dev->rreq.bi_max_vecs++;
	dev->rreq.bi_private = sh;
	dev->rvec.bv_page = dev->page;

L
Linus Torvalds 已提交
1929
	dev->flags = 0;
1930
	dev->sector = compute_blocknr(sh, i, previous);
L
Linus Torvalds 已提交
1931 1932
}

1933
static void error(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
1934 1935
{
	char b[BDEVNAME_SIZE];
1936
	struct r5conf *conf = mddev->private;
1937
	unsigned long flags;
1938
	pr_debug("raid456: error called\n");
L
Linus Torvalds 已提交
1939

1940 1941 1942 1943 1944 1945
	spin_lock_irqsave(&conf->device_lock, flags);
	clear_bit(In_sync, &rdev->flags);
	mddev->degraded = calc_degraded(conf);
	spin_unlock_irqrestore(&conf->device_lock, flags);
	set_bit(MD_RECOVERY_INTR, &mddev->recovery);

1946
	set_bit(Blocked, &rdev->flags);
1947 1948 1949 1950 1951 1952 1953 1954 1955
	set_bit(Faulty, &rdev->flags);
	set_bit(MD_CHANGE_DEVS, &mddev->flags);
	printk(KERN_ALERT
	       "md/raid:%s: Disk failure on %s, disabling device.\n"
	       "md/raid:%s: Operation continuing on %d devices.\n",
	       mdname(mddev),
	       bdevname(rdev->bdev, b),
	       mdname(mddev),
	       conf->raid_disks - mddev->degraded);
1956
}
L
Linus Torvalds 已提交
1957 1958 1959 1960 1961

/*
 * Input: a 'big' sector number,
 * Output: index of the data and parity disk, and the sector # in them.
 */
1962
static sector_t raid5_compute_sector(struct r5conf *conf, sector_t r_sector,
1963 1964
				     int previous, int *dd_idx,
				     struct stripe_head *sh)
L
Linus Torvalds 已提交
1965
{
N
NeilBrown 已提交
1966
	sector_t stripe, stripe2;
1967
	sector_t chunk_number;
L
Linus Torvalds 已提交
1968
	unsigned int chunk_offset;
1969
	int pd_idx, qd_idx;
1970
	int ddf_layout = 0;
L
Linus Torvalds 已提交
1971
	sector_t new_sector;
1972 1973
	int algorithm = previous ? conf->prev_algo
				 : conf->algorithm;
1974 1975
	int sectors_per_chunk = previous ? conf->prev_chunk_sectors
					 : conf->chunk_sectors;
1976 1977 1978
	int raid_disks = previous ? conf->previous_raid_disks
				  : conf->raid_disks;
	int data_disks = raid_disks - conf->max_degraded;
L
Linus Torvalds 已提交
1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990

	/* First compute the information on this sector */

	/*
	 * Compute the chunk number and the sector offset inside the chunk
	 */
	chunk_offset = sector_div(r_sector, sectors_per_chunk);
	chunk_number = r_sector;

	/*
	 * Compute the stripe number
	 */
1991 1992
	stripe = chunk_number;
	*dd_idx = sector_div(stripe, data_disks);
N
NeilBrown 已提交
1993
	stripe2 = stripe;
L
Linus Torvalds 已提交
1994 1995 1996
	/*
	 * Select the parity disk based on the user selected algorithm.
	 */
1997
	pd_idx = qd_idx = -1;
1998 1999
	switch(conf->level) {
	case 4:
2000
		pd_idx = data_disks;
2001 2002
		break;
	case 5:
2003
		switch (algorithm) {
L
Linus Torvalds 已提交
2004
		case ALGORITHM_LEFT_ASYMMETRIC:
N
NeilBrown 已提交
2005
			pd_idx = data_disks - sector_div(stripe2, raid_disks);
2006
			if (*dd_idx >= pd_idx)
L
Linus Torvalds 已提交
2007 2008 2009
				(*dd_idx)++;
			break;
		case ALGORITHM_RIGHT_ASYMMETRIC:
N
NeilBrown 已提交
2010
			pd_idx = sector_div(stripe2, raid_disks);
2011
			if (*dd_idx >= pd_idx)
L
Linus Torvalds 已提交
2012 2013 2014
				(*dd_idx)++;
			break;
		case ALGORITHM_LEFT_SYMMETRIC:
N
NeilBrown 已提交
2015
			pd_idx = data_disks - sector_div(stripe2, raid_disks);
2016
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
L
Linus Torvalds 已提交
2017 2018
			break;
		case ALGORITHM_RIGHT_SYMMETRIC:
N
NeilBrown 已提交
2019
			pd_idx = sector_div(stripe2, raid_disks);
2020
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
L
Linus Torvalds 已提交
2021
			break;
2022 2023 2024 2025 2026 2027 2028
		case ALGORITHM_PARITY_0:
			pd_idx = 0;
			(*dd_idx)++;
			break;
		case ALGORITHM_PARITY_N:
			pd_idx = data_disks;
			break;
L
Linus Torvalds 已提交
2029
		default:
2030
			BUG();
2031 2032 2033 2034
		}
		break;
	case 6:

2035
		switch (algorithm) {
2036
		case ALGORITHM_LEFT_ASYMMETRIC:
N
NeilBrown 已提交
2037
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2038 2039
			qd_idx = pd_idx + 1;
			if (pd_idx == raid_disks-1) {
2040
				(*dd_idx)++;	/* Q D D D P */
2041 2042
				qd_idx = 0;
			} else if (*dd_idx >= pd_idx)
2043 2044 2045
				(*dd_idx) += 2; /* D D P Q D */
			break;
		case ALGORITHM_RIGHT_ASYMMETRIC:
N
NeilBrown 已提交
2046
			pd_idx = sector_div(stripe2, raid_disks);
2047 2048
			qd_idx = pd_idx + 1;
			if (pd_idx == raid_disks-1) {
2049
				(*dd_idx)++;	/* Q D D D P */
2050 2051
				qd_idx = 0;
			} else if (*dd_idx >= pd_idx)
2052 2053 2054
				(*dd_idx) += 2; /* D D P Q D */
			break;
		case ALGORITHM_LEFT_SYMMETRIC:
N
NeilBrown 已提交
2055
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2056 2057
			qd_idx = (pd_idx + 1) % raid_disks;
			*dd_idx = (pd_idx + 2 + *dd_idx) % raid_disks;
2058 2059
			break;
		case ALGORITHM_RIGHT_SYMMETRIC:
N
NeilBrown 已提交
2060
			pd_idx = sector_div(stripe2, raid_disks);
2061 2062
			qd_idx = (pd_idx + 1) % raid_disks;
			*dd_idx = (pd_idx + 2 + *dd_idx) % raid_disks;
2063
			break;
2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078

		case ALGORITHM_PARITY_0:
			pd_idx = 0;
			qd_idx = 1;
			(*dd_idx) += 2;
			break;
		case ALGORITHM_PARITY_N:
			pd_idx = data_disks;
			qd_idx = data_disks + 1;
			break;

		case ALGORITHM_ROTATING_ZERO_RESTART:
			/* Exactly the same as RIGHT_ASYMMETRIC, but or
			 * of blocks for computing Q is different.
			 */
N
NeilBrown 已提交
2079
			pd_idx = sector_div(stripe2, raid_disks);
2080 2081 2082 2083 2084 2085
			qd_idx = pd_idx + 1;
			if (pd_idx == raid_disks-1) {
				(*dd_idx)++;	/* Q D D D P */
				qd_idx = 0;
			} else if (*dd_idx >= pd_idx)
				(*dd_idx) += 2; /* D D P Q D */
2086
			ddf_layout = 1;
2087 2088 2089 2090 2091 2092 2093
			break;

		case ALGORITHM_ROTATING_N_RESTART:
			/* Same a left_asymmetric, by first stripe is
			 * D D D P Q  rather than
			 * Q D D D P
			 */
N
NeilBrown 已提交
2094 2095
			stripe2 += 1;
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2096 2097 2098 2099 2100 2101
			qd_idx = pd_idx + 1;
			if (pd_idx == raid_disks-1) {
				(*dd_idx)++;	/* Q D D D P */
				qd_idx = 0;
			} else if (*dd_idx >= pd_idx)
				(*dd_idx) += 2; /* D D P Q D */
2102
			ddf_layout = 1;
2103 2104 2105 2106
			break;

		case ALGORITHM_ROTATING_N_CONTINUE:
			/* Same as left_symmetric but Q is before P */
N
NeilBrown 已提交
2107
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2108 2109
			qd_idx = (pd_idx + raid_disks - 1) % raid_disks;
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
2110
			ddf_layout = 1;
2111 2112 2113 2114
			break;

		case ALGORITHM_LEFT_ASYMMETRIC_6:
			/* RAID5 left_asymmetric, with Q on last device */
N
NeilBrown 已提交
2115
			pd_idx = data_disks - sector_div(stripe2, raid_disks-1);
2116 2117 2118 2119 2120 2121
			if (*dd_idx >= pd_idx)
				(*dd_idx)++;
			qd_idx = raid_disks - 1;
			break;

		case ALGORITHM_RIGHT_ASYMMETRIC_6:
N
NeilBrown 已提交
2122
			pd_idx = sector_div(stripe2, raid_disks-1);
2123 2124 2125 2126 2127 2128
			if (*dd_idx >= pd_idx)
				(*dd_idx)++;
			qd_idx = raid_disks - 1;
			break;

		case ALGORITHM_LEFT_SYMMETRIC_6:
N
NeilBrown 已提交
2129
			pd_idx = data_disks - sector_div(stripe2, raid_disks-1);
2130 2131 2132 2133 2134
			*dd_idx = (pd_idx + 1 + *dd_idx) % (raid_disks-1);
			qd_idx = raid_disks - 1;
			break;

		case ALGORITHM_RIGHT_SYMMETRIC_6:
N
NeilBrown 已提交
2135
			pd_idx = sector_div(stripe2, raid_disks-1);
2136 2137 2138 2139 2140 2141 2142 2143 2144 2145
			*dd_idx = (pd_idx + 1 + *dd_idx) % (raid_disks-1);
			qd_idx = raid_disks - 1;
			break;

		case ALGORITHM_PARITY_0_6:
			pd_idx = 0;
			(*dd_idx)++;
			qd_idx = raid_disks - 1;
			break;

2146
		default:
2147
			BUG();
2148 2149
		}
		break;
L
Linus Torvalds 已提交
2150 2151
	}

2152 2153 2154
	if (sh) {
		sh->pd_idx = pd_idx;
		sh->qd_idx = qd_idx;
2155
		sh->ddf_layout = ddf_layout;
2156
	}
L
Linus Torvalds 已提交
2157 2158 2159 2160 2161 2162 2163 2164
	/*
	 * Finally, compute the new sector number
	 */
	new_sector = (sector_t)stripe * sectors_per_chunk + chunk_offset;
	return new_sector;
}


2165
static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous)
L
Linus Torvalds 已提交
2166
{
2167
	struct r5conf *conf = sh->raid_conf;
2168 2169
	int raid_disks = sh->disks;
	int data_disks = raid_disks - conf->max_degraded;
L
Linus Torvalds 已提交
2170
	sector_t new_sector = sh->sector, check;
2171 2172
	int sectors_per_chunk = previous ? conf->prev_chunk_sectors
					 : conf->chunk_sectors;
2173 2174
	int algorithm = previous ? conf->prev_algo
				 : conf->algorithm;
L
Linus Torvalds 已提交
2175 2176
	sector_t stripe;
	int chunk_offset;
2177 2178
	sector_t chunk_number;
	int dummy1, dd_idx = i;
L
Linus Torvalds 已提交
2179
	sector_t r_sector;
2180
	struct stripe_head sh2;
L
Linus Torvalds 已提交
2181

2182

L
Linus Torvalds 已提交
2183 2184 2185
	chunk_offset = sector_div(new_sector, sectors_per_chunk);
	stripe = new_sector;

2186 2187 2188 2189 2190
	if (i == sh->pd_idx)
		return 0;
	switch(conf->level) {
	case 4: break;
	case 5:
2191
		switch (algorithm) {
L
Linus Torvalds 已提交
2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202
		case ALGORITHM_LEFT_ASYMMETRIC:
		case ALGORITHM_RIGHT_ASYMMETRIC:
			if (i > sh->pd_idx)
				i--;
			break;
		case ALGORITHM_LEFT_SYMMETRIC:
		case ALGORITHM_RIGHT_SYMMETRIC:
			if (i < sh->pd_idx)
				i += raid_disks;
			i -= (sh->pd_idx + 1);
			break;
2203 2204 2205 2206 2207
		case ALGORITHM_PARITY_0:
			i -= 1;
			break;
		case ALGORITHM_PARITY_N:
			break;
L
Linus Torvalds 已提交
2208
		default:
2209
			BUG();
2210 2211 2212
		}
		break;
	case 6:
2213
		if (i == sh->qd_idx)
2214
			return 0; /* It is the Q disk */
2215
		switch (algorithm) {
2216 2217
		case ALGORITHM_LEFT_ASYMMETRIC:
		case ALGORITHM_RIGHT_ASYMMETRIC:
2218 2219 2220 2221
		case ALGORITHM_ROTATING_ZERO_RESTART:
		case ALGORITHM_ROTATING_N_RESTART:
			if (sh->pd_idx == raid_disks-1)
				i--;	/* Q D D D P */
2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235
			else if (i > sh->pd_idx)
				i -= 2; /* D D P Q D */
			break;
		case ALGORITHM_LEFT_SYMMETRIC:
		case ALGORITHM_RIGHT_SYMMETRIC:
			if (sh->pd_idx == raid_disks-1)
				i--; /* Q D D D P */
			else {
				/* D D P Q D */
				if (i < sh->pd_idx)
					i += raid_disks;
				i -= (sh->pd_idx + 2);
			}
			break;
2236 2237 2238 2239 2240 2241
		case ALGORITHM_PARITY_0:
			i -= 2;
			break;
		case ALGORITHM_PARITY_N:
			break;
		case ALGORITHM_ROTATING_N_CONTINUE:
2242
			/* Like left_symmetric, but P is before Q */
2243 2244
			if (sh->pd_idx == 0)
				i--;	/* P D D D Q */
2245 2246 2247 2248 2249 2250
			else {
				/* D D Q P D */
				if (i < sh->pd_idx)
					i += raid_disks;
				i -= (sh->pd_idx + 1);
			}
2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265
			break;
		case ALGORITHM_LEFT_ASYMMETRIC_6:
		case ALGORITHM_RIGHT_ASYMMETRIC_6:
			if (i > sh->pd_idx)
				i--;
			break;
		case ALGORITHM_LEFT_SYMMETRIC_6:
		case ALGORITHM_RIGHT_SYMMETRIC_6:
			if (i < sh->pd_idx)
				i += data_disks + 1;
			i -= (sh->pd_idx + 1);
			break;
		case ALGORITHM_PARITY_0_6:
			i -= 1;
			break;
2266
		default:
2267
			BUG();
2268 2269
		}
		break;
L
Linus Torvalds 已提交
2270 2271 2272
	}

	chunk_number = stripe * data_disks + i;
2273
	r_sector = chunk_number * sectors_per_chunk + chunk_offset;
L
Linus Torvalds 已提交
2274

2275
	check = raid5_compute_sector(conf, r_sector,
2276
				     previous, &dummy1, &sh2);
2277 2278
	if (check != sh->sector || dummy1 != dd_idx || sh2.pd_idx != sh->pd_idx
		|| sh2.qd_idx != sh->qd_idx) {
2279 2280
		printk(KERN_ERR "md/raid:%s: compute_blocknr: map not correct\n",
		       mdname(conf->mddev));
L
Linus Torvalds 已提交
2281 2282 2283 2284 2285 2286
		return 0;
	}
	return r_sector;
}


2287
static void
2288
schedule_reconstruction(struct stripe_head *sh, struct stripe_head_state *s,
2289
			 int rcw, int expand)
2290 2291
{
	int i, pd_idx = sh->pd_idx, disks = sh->disks;
2292
	struct r5conf *conf = sh->raid_conf;
2293
	int level = conf->level;
2294 2295 2296 2297 2298 2299 2300 2301

	if (rcw) {

		for (i = disks; i--; ) {
			struct r5dev *dev = &sh->dev[i];

			if (dev->towrite) {
				set_bit(R5_LOCKED, &dev->flags);
2302
				set_bit(R5_Wantdrain, &dev->flags);
2303 2304
				if (!expand)
					clear_bit(R5_UPTODATE, &dev->flags);
2305
				s->locked++;
2306 2307
			}
		}
2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322
		/* if we are not expanding this is a proper write request, and
		 * there will be bios with new data to be drained into the
		 * stripe cache
		 */
		if (!expand) {
			if (!s->locked)
				/* False alarm, nothing to do */
				return;
			sh->reconstruct_state = reconstruct_state_drain_run;
			set_bit(STRIPE_OP_BIODRAIN, &s->ops_request);
		} else
			sh->reconstruct_state = reconstruct_state_run;

		set_bit(STRIPE_OP_RECONSTRUCT, &s->ops_request);

2323
		if (s->locked + conf->max_degraded == disks)
2324
			if (!test_and_set_bit(STRIPE_FULL_WRITE, &sh->state))
2325
				atomic_inc(&conf->pending_full_writes);
2326
	} else {
2327
		BUG_ON(level == 6);
2328 2329 2330 2331 2332 2333 2334 2335 2336 2337
		BUG_ON(!(test_bit(R5_UPTODATE, &sh->dev[pd_idx].flags) ||
			test_bit(R5_Wantcompute, &sh->dev[pd_idx].flags)));

		for (i = disks; i--; ) {
			struct r5dev *dev = &sh->dev[i];
			if (i == pd_idx)
				continue;

			if (dev->towrite &&
			    (test_bit(R5_UPTODATE, &dev->flags) ||
2338 2339
			     test_bit(R5_Wantcompute, &dev->flags))) {
				set_bit(R5_Wantdrain, &dev->flags);
2340 2341
				set_bit(R5_LOCKED, &dev->flags);
				clear_bit(R5_UPTODATE, &dev->flags);
2342
				s->locked++;
2343 2344
			}
		}
2345 2346 2347 2348 2349 2350 2351
		if (!s->locked)
			/* False alarm - nothing to do */
			return;
		sh->reconstruct_state = reconstruct_state_prexor_drain_run;
		set_bit(STRIPE_OP_PREXOR, &s->ops_request);
		set_bit(STRIPE_OP_BIODRAIN, &s->ops_request);
		set_bit(STRIPE_OP_RECONSTRUCT, &s->ops_request);
2352 2353
	}

2354
	/* keep the parity disk(s) locked while asynchronous operations
2355 2356 2357 2358
	 * are in flight
	 */
	set_bit(R5_LOCKED, &sh->dev[pd_idx].flags);
	clear_bit(R5_UPTODATE, &sh->dev[pd_idx].flags);
2359
	s->locked++;
2360

2361 2362 2363 2364 2365 2366 2367 2368 2369
	if (level == 6) {
		int qd_idx = sh->qd_idx;
		struct r5dev *dev = &sh->dev[qd_idx];

		set_bit(R5_LOCKED, &dev->flags);
		clear_bit(R5_UPTODATE, &dev->flags);
		s->locked++;
	}

2370
	pr_debug("%s: stripe %llu locked: %d ops_request: %lx\n",
2371
		__func__, (unsigned long long)sh->sector,
2372
		s->locked, s->ops_request);
2373
}
2374

L
Linus Torvalds 已提交
2375 2376
/*
 * Each stripe/dev can have one or more bion attached.
2377
 * toread/towrite point to the first in a chain.
L
Linus Torvalds 已提交
2378 2379 2380 2381 2382
 * The bi_next chain must be in order.
 */
static int add_stripe_bio(struct stripe_head *sh, struct bio *bi, int dd_idx, int forwrite)
{
	struct bio **bip;
2383
	struct r5conf *conf = sh->raid_conf;
2384
	int firstwrite=0;
L
Linus Torvalds 已提交
2385

2386
	pr_debug("adding bi b#%llu to stripe s#%llu\n",
L
Linus Torvalds 已提交
2387 2388 2389
		(unsigned long long)bi->bi_sector,
		(unsigned long long)sh->sector);

S
Shaohua Li 已提交
2390 2391 2392 2393 2394 2395 2396 2397 2398
	/*
	 * If several bio share a stripe. The bio bi_phys_segments acts as a
	 * reference count to avoid race. The reference count should already be
	 * increased before this function is called (for example, in
	 * make_request()), so other bio sharing this stripe will not free the
	 * stripe. If a stripe is owned by one stripe, the stripe lock will
	 * protect it.
	 */
	spin_lock_irq(&sh->stripe_lock);
2399
	if (forwrite) {
L
Linus Torvalds 已提交
2400
		bip = &sh->dev[dd_idx].towrite;
2401
		if (*bip == NULL)
2402 2403
			firstwrite = 1;
	} else
L
Linus Torvalds 已提交
2404 2405
		bip = &sh->dev[dd_idx].toread;
	while (*bip && (*bip)->bi_sector < bi->bi_sector) {
K
Kent Overstreet 已提交
2406
		if (bio_end_sector(*bip) > bi->bi_sector)
L
Linus Torvalds 已提交
2407 2408 2409
			goto overlap;
		bip = & (*bip)->bi_next;
	}
K
Kent Overstreet 已提交
2410
	if (*bip && (*bip)->bi_sector < bio_end_sector(bi))
L
Linus Torvalds 已提交
2411 2412
		goto overlap;

2413
	BUG_ON(*bip && bi->bi_next && (*bip) != bi->bi_next);
L
Linus Torvalds 已提交
2414 2415 2416
	if (*bip)
		bi->bi_next = *bip;
	*bip = bi;
2417
	raid5_inc_bi_active_stripes(bi);
2418

L
Linus Torvalds 已提交
2419 2420 2421 2422 2423 2424 2425
	if (forwrite) {
		/* check if page is covered */
		sector_t sector = sh->dev[dd_idx].sector;
		for (bi=sh->dev[dd_idx].towrite;
		     sector < sh->dev[dd_idx].sector + STRIPE_SECTORS &&
			     bi && bi->bi_sector <= sector;
		     bi = r5_next_bio(bi, sh->dev[dd_idx].sector)) {
K
Kent Overstreet 已提交
2426 2427
			if (bio_end_sector(bi) >= sector)
				sector = bio_end_sector(bi);
L
Linus Torvalds 已提交
2428 2429 2430 2431
		}
		if (sector >= sh->dev[dd_idx].sector + STRIPE_SECTORS)
			set_bit(R5_OVERWRITE, &sh->dev[dd_idx].flags);
	}
2432 2433 2434 2435

	pr_debug("added bi b#%llu to stripe s#%llu, disk %d.\n",
		(unsigned long long)(*bip)->bi_sector,
		(unsigned long long)sh->sector, dd_idx);
2436
	spin_unlock_irq(&sh->stripe_lock);
2437 2438 2439 2440 2441 2442 2443

	if (conf->mddev->bitmap && firstwrite) {
		bitmap_startwrite(conf->mddev->bitmap, sh->sector,
				  STRIPE_SECTORS, 0);
		sh->bm_seq = conf->seq_flush+1;
		set_bit(STRIPE_BIT_DELAY, &sh->state);
	}
L
Linus Torvalds 已提交
2444 2445 2446 2447
	return 1;

 overlap:
	set_bit(R5_Overlap, &sh->dev[dd_idx].flags);
S
Shaohua Li 已提交
2448
	spin_unlock_irq(&sh->stripe_lock);
L
Linus Torvalds 已提交
2449 2450 2451
	return 0;
}

2452
static void end_reshape(struct r5conf *conf);
2453

2454
static void stripe_set_idx(sector_t stripe, struct r5conf *conf, int previous,
2455
			    struct stripe_head *sh)
2456
{
2457
	int sectors_per_chunk =
2458
		previous ? conf->prev_chunk_sectors : conf->chunk_sectors;
2459
	int dd_idx;
2460
	int chunk_offset = sector_div(stripe, sectors_per_chunk);
2461
	int disks = previous ? conf->previous_raid_disks : conf->raid_disks;
2462

2463 2464
	raid5_compute_sector(conf,
			     stripe * (disks - conf->max_degraded)
2465
			     *sectors_per_chunk + chunk_offset,
2466
			     previous,
2467
			     &dd_idx, sh);
2468 2469
}

2470
static void
2471
handle_failed_stripe(struct r5conf *conf, struct stripe_head *sh,
2472 2473 2474 2475 2476 2477 2478 2479 2480
				struct stripe_head_state *s, int disks,
				struct bio **return_bi)
{
	int i;
	for (i = disks; i--; ) {
		struct bio *bi;
		int bitmap_end = 0;

		if (test_bit(R5_ReadError, &sh->dev[i].flags)) {
2481
			struct md_rdev *rdev;
2482 2483 2484
			rcu_read_lock();
			rdev = rcu_dereference(conf->disks[i].rdev);
			if (rdev && test_bit(In_sync, &rdev->flags))
2485 2486 2487
				atomic_inc(&rdev->nr_pending);
			else
				rdev = NULL;
2488
			rcu_read_unlock();
2489 2490 2491 2492 2493 2494 2495 2496
			if (rdev) {
				if (!rdev_set_badblocks(
					    rdev,
					    sh->sector,
					    STRIPE_SECTORS, 0))
					md_error(conf->mddev, rdev);
				rdev_dec_pending(rdev, conf->mddev);
			}
2497
		}
S
Shaohua Li 已提交
2498
		spin_lock_irq(&sh->stripe_lock);
2499 2500 2501
		/* fail all writes first */
		bi = sh->dev[i].towrite;
		sh->dev[i].towrite = NULL;
S
Shaohua Li 已提交
2502
		spin_unlock_irq(&sh->stripe_lock);
2503
		if (bi)
2504 2505 2506 2507 2508 2509 2510 2511 2512
			bitmap_end = 1;

		if (test_and_clear_bit(R5_Overlap, &sh->dev[i].flags))
			wake_up(&conf->wait_for_overlap);

		while (bi && bi->bi_sector <
			sh->dev[i].sector + STRIPE_SECTORS) {
			struct bio *nextbi = r5_next_bio(bi, sh->dev[i].sector);
			clear_bit(BIO_UPTODATE, &bi->bi_flags);
2513
			if (!raid5_dec_bi_active_stripes(bi)) {
2514 2515 2516 2517 2518 2519
				md_write_end(conf->mddev);
				bi->bi_next = *return_bi;
				*return_bi = bi;
			}
			bi = nextbi;
		}
2520 2521 2522 2523
		if (bitmap_end)
			bitmap_endwrite(conf->mddev->bitmap, sh->sector,
				STRIPE_SECTORS, 0, 0);
		bitmap_end = 0;
2524 2525 2526 2527 2528 2529 2530 2531
		/* and fail all 'written' */
		bi = sh->dev[i].written;
		sh->dev[i].written = NULL;
		if (bi) bitmap_end = 1;
		while (bi && bi->bi_sector <
		       sh->dev[i].sector + STRIPE_SECTORS) {
			struct bio *bi2 = r5_next_bio(bi, sh->dev[i].sector);
			clear_bit(BIO_UPTODATE, &bi->bi_flags);
2532
			if (!raid5_dec_bi_active_stripes(bi)) {
2533 2534 2535 2536 2537 2538 2539
				md_write_end(conf->mddev);
				bi->bi_next = *return_bi;
				*return_bi = bi;
			}
			bi = bi2;
		}

2540 2541 2542 2543 2544 2545
		/* fail any reads if this device is non-operational and
		 * the data has not reached the cache yet.
		 */
		if (!test_bit(R5_Wantfill, &sh->dev[i].flags) &&
		    (!test_bit(R5_Insync, &sh->dev[i].flags) ||
		      test_bit(R5_ReadError, &sh->dev[i].flags))) {
2546
			spin_lock_irq(&sh->stripe_lock);
2547 2548
			bi = sh->dev[i].toread;
			sh->dev[i].toread = NULL;
2549
			spin_unlock_irq(&sh->stripe_lock);
2550 2551 2552 2553 2554 2555 2556
			if (test_and_clear_bit(R5_Overlap, &sh->dev[i].flags))
				wake_up(&conf->wait_for_overlap);
			while (bi && bi->bi_sector <
			       sh->dev[i].sector + STRIPE_SECTORS) {
				struct bio *nextbi =
					r5_next_bio(bi, sh->dev[i].sector);
				clear_bit(BIO_UPTODATE, &bi->bi_flags);
2557
				if (!raid5_dec_bi_active_stripes(bi)) {
2558 2559 2560 2561 2562 2563 2564 2565 2566
					bi->bi_next = *return_bi;
					*return_bi = bi;
				}
				bi = nextbi;
			}
		}
		if (bitmap_end)
			bitmap_endwrite(conf->mddev->bitmap, sh->sector,
					STRIPE_SECTORS, 0, 0);
2567 2568 2569 2570
		/* If we were in the middle of a write the parity block might
		 * still be locked - so just clear all R5_LOCKED flags
		 */
		clear_bit(R5_LOCKED, &sh->dev[i].flags);
2571 2572
	}

2573 2574 2575
	if (test_and_clear_bit(STRIPE_FULL_WRITE, &sh->state))
		if (atomic_dec_and_test(&conf->pending_full_writes))
			md_wakeup_thread(conf->mddev->thread);
2576 2577
}

2578
static void
2579
handle_failed_sync(struct r5conf *conf, struct stripe_head *sh,
2580 2581 2582 2583 2584 2585
		   struct stripe_head_state *s)
{
	int abort = 0;
	int i;

	clear_bit(STRIPE_SYNCING, &sh->state);
2586 2587
	if (test_and_clear_bit(R5_Overlap, &sh->dev[sh->pd_idx].flags))
		wake_up(&conf->wait_for_overlap);
2588
	s->syncing = 0;
2589
	s->replacing = 0;
2590
	/* There is nothing more to do for sync/check/repair.
2591 2592 2593
	 * Don't even need to abort as that is handled elsewhere
	 * if needed, and not always wanted e.g. if there is a known
	 * bad block here.
2594
	 * For recover/replace we need to record a bad block on all
2595 2596
	 * non-sync devices, or abort the recovery
	 */
2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619
	if (test_bit(MD_RECOVERY_RECOVER, &conf->mddev->recovery)) {
		/* During recovery devices cannot be removed, so
		 * locking and refcounting of rdevs is not needed
		 */
		for (i = 0; i < conf->raid_disks; i++) {
			struct md_rdev *rdev = conf->disks[i].rdev;
			if (rdev
			    && !test_bit(Faulty, &rdev->flags)
			    && !test_bit(In_sync, &rdev->flags)
			    && !rdev_set_badblocks(rdev, sh->sector,
						   STRIPE_SECTORS, 0))
				abort = 1;
			rdev = conf->disks[i].replacement;
			if (rdev
			    && !test_bit(Faulty, &rdev->flags)
			    && !test_bit(In_sync, &rdev->flags)
			    && !rdev_set_badblocks(rdev, sh->sector,
						   STRIPE_SECTORS, 0))
				abort = 1;
		}
		if (abort)
			conf->recovery_disabled =
				conf->mddev->recovery_disabled;
2620
	}
2621
	md_done_sync(conf->mddev, STRIPE_SECTORS, !abort);
2622 2623
}

2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639
static int want_replace(struct stripe_head *sh, int disk_idx)
{
	struct md_rdev *rdev;
	int rv = 0;
	/* Doing recovery so rcu locking not required */
	rdev = sh->raid_conf->disks[disk_idx].replacement;
	if (rdev
	    && !test_bit(Faulty, &rdev->flags)
	    && !test_bit(In_sync, &rdev->flags)
	    && (rdev->recovery_offset <= sh->sector
		|| rdev->mddev->recovery_cp <= sh->sector))
		rv = 1;

	return rv;
}

2640
/* fetch_block - checks the given member device to see if its data needs
2641 2642 2643
 * to be read or computed to satisfy a request.
 *
 * Returns 1 when no more member devices need to be checked, otherwise returns
2644
 * 0 to tell the loop in handle_stripe_fill to continue
2645
 */
2646 2647
static int fetch_block(struct stripe_head *sh, struct stripe_head_state *s,
		       int disk_idx, int disks)
2648
{
2649
	struct r5dev *dev = &sh->dev[disk_idx];
2650 2651
	struct r5dev *fdev[2] = { &sh->dev[s->failed_num[0]],
				  &sh->dev[s->failed_num[1]] };
2652

2653
	/* is the data in this block needed, and can we get it? */
2654 2655 2656 2657 2658
	if (!test_bit(R5_LOCKED, &dev->flags) &&
	    !test_bit(R5_UPTODATE, &dev->flags) &&
	    (dev->toread ||
	     (dev->towrite && !test_bit(R5_OVERWRITE, &dev->flags)) ||
	     s->syncing || s->expanding ||
2659
	     (s->replacing && want_replace(sh, disk_idx)) ||
2660 2661
	     (s->failed >= 1 && fdev[0]->toread) ||
	     (s->failed >= 2 && fdev[1]->toread) ||
2662 2663 2664
	     (sh->raid_conf->level <= 5 && s->failed && fdev[0]->towrite &&
	      !test_bit(R5_OVERWRITE, &fdev[0]->flags)) ||
	     (sh->raid_conf->level == 6 && s->failed && s->to_write))) {
2665 2666 2667 2668 2669 2670
		/* we would like to get this block, possibly by computing it,
		 * otherwise read it if the backing disk is insync
		 */
		BUG_ON(test_bit(R5_Wantcompute, &dev->flags));
		BUG_ON(test_bit(R5_Wantread, &dev->flags));
		if ((s->uptodate == disks - 1) &&
2671 2672
		    (s->failed && (disk_idx == s->failed_num[0] ||
				   disk_idx == s->failed_num[1]))) {
2673 2674
			/* have disk failed, and we're requested to fetch it;
			 * do compute it
2675
			 */
2676 2677 2678 2679 2680 2681 2682 2683
			pr_debug("Computing stripe %llu block %d\n",
			       (unsigned long long)sh->sector, disk_idx);
			set_bit(STRIPE_COMPUTE_RUN, &sh->state);
			set_bit(STRIPE_OP_COMPUTE_BLK, &s->ops_request);
			set_bit(R5_Wantcompute, &dev->flags);
			sh->ops.target = disk_idx;
			sh->ops.target2 = -1; /* no 2nd target */
			s->req_compute = 1;
2684 2685 2686 2687 2688 2689
			/* Careful: from this point on 'uptodate' is in the eye
			 * of raid_run_ops which services 'compute' operations
			 * before writes. R5_Wantcompute flags a block that will
			 * be R5_UPTODATE by the time it is needed for a
			 * subsequent operation.
			 */
2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702
			s->uptodate++;
			return 1;
		} else if (s->uptodate == disks-2 && s->failed >= 2) {
			/* Computing 2-failure is *very* expensive; only
			 * do it if failed >= 2
			 */
			int other;
			for (other = disks; other--; ) {
				if (other == disk_idx)
					continue;
				if (!test_bit(R5_UPTODATE,
				      &sh->dev[other].flags))
					break;
2703
			}
2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722
			BUG_ON(other < 0);
			pr_debug("Computing stripe %llu blocks %d,%d\n",
			       (unsigned long long)sh->sector,
			       disk_idx, other);
			set_bit(STRIPE_COMPUTE_RUN, &sh->state);
			set_bit(STRIPE_OP_COMPUTE_BLK, &s->ops_request);
			set_bit(R5_Wantcompute, &sh->dev[disk_idx].flags);
			set_bit(R5_Wantcompute, &sh->dev[other].flags);
			sh->ops.target = disk_idx;
			sh->ops.target2 = other;
			s->uptodate += 2;
			s->req_compute = 1;
			return 1;
		} else if (test_bit(R5_Insync, &dev->flags)) {
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantread, &dev->flags);
			s->locked++;
			pr_debug("Reading block %d (sync=%d)\n",
				disk_idx, s->syncing);
2723 2724
		}
	}
2725 2726 2727 2728 2729

	return 0;
}

/**
2730
 * handle_stripe_fill - read or compute data to satisfy pending requests.
2731
 */
2732 2733 2734
static void handle_stripe_fill(struct stripe_head *sh,
			       struct stripe_head_state *s,
			       int disks)
2735 2736 2737 2738 2739 2740 2741 2742 2743 2744
{
	int i;

	/* look for blocks to read/compute, skip this if a compute
	 * is already in flight, or if the stripe contents are in the
	 * midst of changing due to a write
	 */
	if (!test_bit(STRIPE_COMPUTE_RUN, &sh->state) && !sh->check_state &&
	    !sh->reconstruct_state)
		for (i = disks; i--; )
2745
			if (fetch_block(sh, s, i, disks))
2746
				break;
2747 2748 2749 2750
	set_bit(STRIPE_HANDLE, &sh->state);
}


2751
/* handle_stripe_clean_event
2752 2753 2754 2755
 * any written block on an uptodate or failed drive can be returned.
 * Note that if we 'wrote' to a failed drive, it will be UPTODATE, but
 * never LOCKED, so we don't need to test 'failed' directly.
 */
2756
static void handle_stripe_clean_event(struct r5conf *conf,
2757 2758 2759 2760
	struct stripe_head *sh, int disks, struct bio **return_bi)
{
	int i;
	struct r5dev *dev;
2761
	int discard_pending = 0;
2762 2763 2764 2765 2766

	for (i = disks; i--; )
		if (sh->dev[i].written) {
			dev = &sh->dev[i];
			if (!test_bit(R5_LOCKED, &dev->flags) &&
2767
			    (test_bit(R5_UPTODATE, &dev->flags) ||
2768
			     test_bit(R5_Discard, &dev->flags))) {
2769 2770
				/* We can return any write requests */
				struct bio *wbi, *wbi2;
2771
				pr_debug("Return write for disc %d\n", i);
2772 2773
				if (test_and_clear_bit(R5_Discard, &dev->flags))
					clear_bit(R5_UPTODATE, &dev->flags);
2774 2775 2776 2777 2778
				wbi = dev->written;
				dev->written = NULL;
				while (wbi && wbi->bi_sector <
					dev->sector + STRIPE_SECTORS) {
					wbi2 = r5_next_bio(wbi, dev->sector);
2779
					if (!raid5_dec_bi_active_stripes(wbi)) {
2780 2781 2782 2783 2784 2785
						md_write_end(conf->mddev);
						wbi->bi_next = *return_bi;
						*return_bi = wbi;
					}
					wbi = wbi2;
				}
2786 2787
				bitmap_endwrite(conf->mddev->bitmap, sh->sector,
						STRIPE_SECTORS,
2788
					 !test_bit(STRIPE_DEGRADED, &sh->state),
2789
						0);
2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806
			} else if (test_bit(R5_Discard, &dev->flags))
				discard_pending = 1;
		}
	if (!discard_pending &&
	    test_bit(R5_Discard, &sh->dev[sh->pd_idx].flags)) {
		clear_bit(R5_Discard, &sh->dev[sh->pd_idx].flags);
		clear_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags);
		if (sh->qd_idx >= 0) {
			clear_bit(R5_Discard, &sh->dev[sh->qd_idx].flags);
			clear_bit(R5_UPTODATE, &sh->dev[sh->qd_idx].flags);
		}
		/* now that discard is done we can proceed with any sync */
		clear_bit(STRIPE_DISCARD, &sh->state);
		if (test_bit(STRIPE_SYNC_REQUESTED, &sh->state))
			set_bit(STRIPE_HANDLE, &sh->state);

	}
2807 2808 2809 2810

	if (test_and_clear_bit(STRIPE_FULL_WRITE, &sh->state))
		if (atomic_dec_and_test(&conf->pending_full_writes))
			md_wakeup_thread(conf->mddev->thread);
2811 2812
}

2813
static void handle_stripe_dirtying(struct r5conf *conf,
2814 2815 2816
				   struct stripe_head *sh,
				   struct stripe_head_state *s,
				   int disks)
2817 2818
{
	int rmw = 0, rcw = 0, i;
2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831
	sector_t recovery_cp = conf->mddev->recovery_cp;

	/* RAID6 requires 'rcw' in current implementation.
	 * Otherwise, check whether resync is now happening or should start.
	 * If yes, then the array is dirty (after unclean shutdown or
	 * initial creation), so parity in some stripes might be inconsistent.
	 * In this case, we need to always do reconstruct-write, to ensure
	 * that in case of drive failure or read-error correction, we
	 * generate correct data from the parity.
	 */
	if (conf->max_degraded == 2 ||
	    (recovery_cp < MaxSector && sh->sector >= recovery_cp)) {
		/* Calculate the real rcw later - for now make it
2832 2833 2834
		 * look like rcw is cheaper
		 */
		rcw = 1; rmw = 2;
2835 2836 2837
		pr_debug("force RCW max_degraded=%u, recovery_cp=%llu sh->sector=%llu\n",
			 conf->max_degraded, (unsigned long long)recovery_cp,
			 (unsigned long long)sh->sector);
2838
	} else for (i = disks; i--; ) {
2839 2840 2841 2842
		/* would I have to read this buffer for read_modify_write */
		struct r5dev *dev = &sh->dev[i];
		if ((dev->towrite || i == sh->pd_idx) &&
		    !test_bit(R5_LOCKED, &dev->flags) &&
2843 2844
		    !(test_bit(R5_UPTODATE, &dev->flags) ||
		      test_bit(R5_Wantcompute, &dev->flags))) {
2845 2846 2847 2848 2849 2850 2851 2852
			if (test_bit(R5_Insync, &dev->flags))
				rmw++;
			else
				rmw += 2*disks;  /* cannot read it */
		}
		/* Would I have to read this buffer for reconstruct_write */
		if (!test_bit(R5_OVERWRITE, &dev->flags) && i != sh->pd_idx &&
		    !test_bit(R5_LOCKED, &dev->flags) &&
2853 2854 2855
		    !(test_bit(R5_UPTODATE, &dev->flags) ||
		    test_bit(R5_Wantcompute, &dev->flags))) {
			if (test_bit(R5_Insync, &dev->flags)) rcw++;
2856 2857 2858 2859
			else
				rcw += 2*disks;
		}
	}
2860
	pr_debug("for sector %llu, rmw=%d rcw=%d\n",
2861 2862
		(unsigned long long)sh->sector, rmw, rcw);
	set_bit(STRIPE_HANDLE, &sh->state);
N
NeilBrown 已提交
2863
	if (rmw < rcw && rmw > 0) {
2864
		/* prefer read-modify-write, but need to get some data */
2865 2866 2867 2868
		if (conf->mddev->queue)
			blk_add_trace_msg(conf->mddev->queue,
					  "raid5 rmw %llu %d",
					  (unsigned long long)sh->sector, rmw);
2869 2870 2871 2872
		for (i = disks; i--; ) {
			struct r5dev *dev = &sh->dev[i];
			if ((dev->towrite || i == sh->pd_idx) &&
			    !test_bit(R5_LOCKED, &dev->flags) &&
2873 2874
			    !(test_bit(R5_UPTODATE, &dev->flags) ||
			    test_bit(R5_Wantcompute, &dev->flags)) &&
2875 2876 2877
			    test_bit(R5_Insync, &dev->flags)) {
				if (
				  test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
2878
					pr_debug("Read_old block "
N
NeilBrown 已提交
2879
						 "%d for r-m-w\n", i);
2880 2881 2882 2883 2884 2885 2886 2887 2888
					set_bit(R5_LOCKED, &dev->flags);
					set_bit(R5_Wantread, &dev->flags);
					s->locked++;
				} else {
					set_bit(STRIPE_DELAYED, &sh->state);
					set_bit(STRIPE_HANDLE, &sh->state);
				}
			}
		}
N
NeilBrown 已提交
2889
	}
2890
	if (rcw <= rmw && rcw > 0) {
2891
		/* want reconstruct write, but need to get some data */
N
NeilBrown 已提交
2892
		int qread =0;
2893
		rcw = 0;
2894 2895 2896
		for (i = disks; i--; ) {
			struct r5dev *dev = &sh->dev[i];
			if (!test_bit(R5_OVERWRITE, &dev->flags) &&
2897
			    i != sh->pd_idx && i != sh->qd_idx &&
2898
			    !test_bit(R5_LOCKED, &dev->flags) &&
2899
			    !(test_bit(R5_UPTODATE, &dev->flags) ||
2900 2901 2902 2903
			      test_bit(R5_Wantcompute, &dev->flags))) {
				rcw++;
				if (!test_bit(R5_Insync, &dev->flags))
					continue; /* it's a failed drive */
2904 2905
				if (
				  test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
2906
					pr_debug("Read_old block "
2907 2908 2909 2910
						"%d for Reconstruct\n", i);
					set_bit(R5_LOCKED, &dev->flags);
					set_bit(R5_Wantread, &dev->flags);
					s->locked++;
N
NeilBrown 已提交
2911
					qread++;
2912 2913 2914 2915 2916 2917
				} else {
					set_bit(STRIPE_DELAYED, &sh->state);
					set_bit(STRIPE_HANDLE, &sh->state);
				}
			}
		}
2918
		if (rcw && conf->mddev->queue)
N
NeilBrown 已提交
2919 2920 2921
			blk_add_trace_msg(conf->mddev->queue, "raid5 rcw %llu %d %d %d",
					  (unsigned long long)sh->sector,
					  rcw, qread, test_bit(STRIPE_DELAYED, &sh->state));
2922
	}
2923 2924 2925
	/* now if nothing is locked, and if we have enough data,
	 * we can start a write request
	 */
2926 2927
	/* since handle_stripe can be called at any time we need to handle the
	 * case where a compute block operation has been submitted and then a
2928 2929
	 * subsequent call wants to start a write request.  raid_run_ops only
	 * handles the case where compute block and reconstruct are requested
2930 2931 2932
	 * simultaneously.  If this is not the case then new writes need to be
	 * held off until the compute completes.
	 */
2933 2934 2935
	if ((s->req_compute || !test_bit(STRIPE_COMPUTE_RUN, &sh->state)) &&
	    (s->locked == 0 && (rcw == 0 || rmw == 0) &&
	    !test_bit(STRIPE_BIT_DELAY, &sh->state)))
2936
		schedule_reconstruction(sh, s, rcw == 0, 0);
2937 2938
}

2939
static void handle_parity_checks5(struct r5conf *conf, struct stripe_head *sh,
2940 2941
				struct stripe_head_state *s, int disks)
{
2942
	struct r5dev *dev = NULL;
2943

2944
	set_bit(STRIPE_HANDLE, &sh->state);
2945

2946 2947 2948
	switch (sh->check_state) {
	case check_state_idle:
		/* start a new check operation if there are no failures */
2949 2950
		if (s->failed == 0) {
			BUG_ON(s->uptodate != disks);
2951 2952
			sh->check_state = check_state_run;
			set_bit(STRIPE_OP_CHECK, &s->ops_request);
2953 2954
			clear_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags);
			s->uptodate--;
2955
			break;
2956
		}
2957
		dev = &sh->dev[s->failed_num[0]];
2958 2959 2960 2961 2962 2963 2964 2965 2966
		/* fall through */
	case check_state_compute_result:
		sh->check_state = check_state_idle;
		if (!dev)
			dev = &sh->dev[sh->pd_idx];

		/* check that a write has not made the stripe insync */
		if (test_bit(STRIPE_INSYNC, &sh->state))
			break;
D
Dan Williams 已提交
2967

2968 2969 2970 2971 2972
		/* either failed parity check, or recovery is happening */
		BUG_ON(!test_bit(R5_UPTODATE, &dev->flags));
		BUG_ON(s->uptodate != disks);

		set_bit(R5_LOCKED, &dev->flags);
2973
		s->locked++;
2974
		set_bit(R5_Wantwrite, &dev->flags);
2975

2976 2977
		clear_bit(STRIPE_DEGRADED, &sh->state);
		set_bit(STRIPE_INSYNC, &sh->state);
2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993
		break;
	case check_state_run:
		break; /* we will be called again upon completion */
	case check_state_check_result:
		sh->check_state = check_state_idle;

		/* if a failure occurred during the check operation, leave
		 * STRIPE_INSYNC not set and let the stripe be handled again
		 */
		if (s->failed)
			break;

		/* handle a successful check operation, if parity is correct
		 * we are done.  Otherwise update the mismatch count and repair
		 * parity if !MD_RECOVERY_CHECK
		 */
D
Dan Williams 已提交
2994
		if ((sh->ops.zero_sum_result & SUM_CHECK_P_RESULT) == 0)
2995 2996 2997 2998 2999
			/* parity is correct (on disc,
			 * not in buffer any more)
			 */
			set_bit(STRIPE_INSYNC, &sh->state);
		else {
3000
			atomic64_add(STRIPE_SECTORS, &conf->mddev->resync_mismatches);
3001 3002 3003 3004 3005
			if (test_bit(MD_RECOVERY_CHECK, &conf->mddev->recovery))
				/* don't try to repair!! */
				set_bit(STRIPE_INSYNC, &sh->state);
			else {
				sh->check_state = check_state_compute_run;
3006
				set_bit(STRIPE_COMPUTE_RUN, &sh->state);
3007 3008 3009 3010
				set_bit(STRIPE_OP_COMPUTE_BLK, &s->ops_request);
				set_bit(R5_Wantcompute,
					&sh->dev[sh->pd_idx].flags);
				sh->ops.target = sh->pd_idx;
3011
				sh->ops.target2 = -1;
3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022
				s->uptodate++;
			}
		}
		break;
	case check_state_compute_run:
		break;
	default:
		printk(KERN_ERR "%s: unknown check_state: %d sector: %llu\n",
		       __func__, sh->check_state,
		       (unsigned long long) sh->sector);
		BUG();
3023 3024 3025 3026
	}
}


3027
static void handle_parity_checks6(struct r5conf *conf, struct stripe_head *sh,
3028
				  struct stripe_head_state *s,
3029
				  int disks)
3030 3031
{
	int pd_idx = sh->pd_idx;
N
NeilBrown 已提交
3032
	int qd_idx = sh->qd_idx;
3033
	struct r5dev *dev;
3034 3035 3036 3037

	set_bit(STRIPE_HANDLE, &sh->state);

	BUG_ON(s->failed > 2);
3038

3039 3040 3041 3042 3043 3044
	/* Want to check and possibly repair P and Q.
	 * However there could be one 'failed' device, in which
	 * case we can only check one of them, possibly using the
	 * other to generate missing data
	 */

3045 3046 3047
	switch (sh->check_state) {
	case check_state_idle:
		/* start a new check operation if there are < 2 failures */
3048
		if (s->failed == s->q_failed) {
3049
			/* The only possible failed device holds Q, so it
3050 3051 3052
			 * makes sense to check P (If anything else were failed,
			 * we would have used P to recreate it).
			 */
3053
			sh->check_state = check_state_run;
3054
		}
3055
		if (!s->q_failed && s->failed < 2) {
3056
			/* Q is not failed, and we didn't use it to generate
3057 3058
			 * anything, so it makes sense to check it
			 */
3059 3060 3061 3062
			if (sh->check_state == check_state_run)
				sh->check_state = check_state_run_pq;
			else
				sh->check_state = check_state_run_q;
3063 3064
		}

3065 3066
		/* discard potentially stale zero_sum_result */
		sh->ops.zero_sum_result = 0;
3067

3068 3069 3070 3071
		if (sh->check_state == check_state_run) {
			/* async_xor_zero_sum destroys the contents of P */
			clear_bit(R5_UPTODATE, &sh->dev[pd_idx].flags);
			s->uptodate--;
3072
		}
3073 3074 3075 3076 3077 3078 3079
		if (sh->check_state >= check_state_run &&
		    sh->check_state <= check_state_run_pq) {
			/* async_syndrome_zero_sum preserves P and Q, so
			 * no need to mark them !uptodate here
			 */
			set_bit(STRIPE_OP_CHECK, &s->ops_request);
			break;
3080 3081
		}

3082 3083 3084 3085 3086
		/* we have 2-disk failure */
		BUG_ON(s->failed != 2);
		/* fall through */
	case check_state_compute_result:
		sh->check_state = check_state_idle;
3087

3088 3089 3090
		/* check that a write has not made the stripe insync */
		if (test_bit(STRIPE_INSYNC, &sh->state))
			break;
3091 3092

		/* now write out any block on a failed drive,
3093
		 * or P or Q if they were recomputed
3094
		 */
3095
		BUG_ON(s->uptodate < disks - 1); /* We don't need Q to recover */
3096
		if (s->failed == 2) {
3097
			dev = &sh->dev[s->failed_num[1]];
3098 3099 3100 3101 3102
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
		if (s->failed >= 1) {
3103
			dev = &sh->dev[s->failed_num[0]];
3104 3105 3106 3107
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
3108
		if (sh->ops.zero_sum_result & SUM_CHECK_P_RESULT) {
3109 3110 3111 3112 3113
			dev = &sh->dev[pd_idx];
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
3114
		if (sh->ops.zero_sum_result & SUM_CHECK_Q_RESULT) {
3115 3116 3117 3118 3119 3120 3121 3122
			dev = &sh->dev[qd_idx];
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
		clear_bit(STRIPE_DEGRADED, &sh->state);

		set_bit(STRIPE_INSYNC, &sh->state);
3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151
		break;
	case check_state_run:
	case check_state_run_q:
	case check_state_run_pq:
		break; /* we will be called again upon completion */
	case check_state_check_result:
		sh->check_state = check_state_idle;

		/* handle a successful check operation, if parity is correct
		 * we are done.  Otherwise update the mismatch count and repair
		 * parity if !MD_RECOVERY_CHECK
		 */
		if (sh->ops.zero_sum_result == 0) {
			/* both parities are correct */
			if (!s->failed)
				set_bit(STRIPE_INSYNC, &sh->state);
			else {
				/* in contrast to the raid5 case we can validate
				 * parity, but still have a failure to write
				 * back
				 */
				sh->check_state = check_state_compute_result;
				/* Returning at this point means that we may go
				 * off and bring p and/or q uptodate again so
				 * we make sure to check zero_sum_result again
				 * to verify if p or q need writeback
				 */
			}
		} else {
3152
			atomic64_add(STRIPE_SECTORS, &conf->mddev->resync_mismatches);
3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186
			if (test_bit(MD_RECOVERY_CHECK, &conf->mddev->recovery))
				/* don't try to repair!! */
				set_bit(STRIPE_INSYNC, &sh->state);
			else {
				int *target = &sh->ops.target;

				sh->ops.target = -1;
				sh->ops.target2 = -1;
				sh->check_state = check_state_compute_run;
				set_bit(STRIPE_COMPUTE_RUN, &sh->state);
				set_bit(STRIPE_OP_COMPUTE_BLK, &s->ops_request);
				if (sh->ops.zero_sum_result & SUM_CHECK_P_RESULT) {
					set_bit(R5_Wantcompute,
						&sh->dev[pd_idx].flags);
					*target = pd_idx;
					target = &sh->ops.target2;
					s->uptodate++;
				}
				if (sh->ops.zero_sum_result & SUM_CHECK_Q_RESULT) {
					set_bit(R5_Wantcompute,
						&sh->dev[qd_idx].flags);
					*target = qd_idx;
					s->uptodate++;
				}
			}
		}
		break;
	case check_state_compute_run:
		break;
	default:
		printk(KERN_ERR "%s: unknown check_state: %d sector: %llu\n",
		       __func__, sh->check_state,
		       (unsigned long long) sh->sector);
		BUG();
3187 3188 3189
	}
}

3190
static void handle_stripe_expansion(struct r5conf *conf, struct stripe_head *sh)
3191 3192 3193 3194 3195 3196
{
	int i;

	/* We have read all the blocks in this stripe and now we need to
	 * copy some of them into a target stripe for expand.
	 */
3197
	struct dma_async_tx_descriptor *tx = NULL;
3198 3199
	clear_bit(STRIPE_EXPAND_SOURCE, &sh->state);
	for (i = 0; i < sh->disks; i++)
N
NeilBrown 已提交
3200
		if (i != sh->pd_idx && i != sh->qd_idx) {
3201
			int dd_idx, j;
3202
			struct stripe_head *sh2;
3203
			struct async_submit_ctl submit;
3204

3205
			sector_t bn = compute_blocknr(sh, i, 1);
3206 3207
			sector_t s = raid5_compute_sector(conf, bn, 0,
							  &dd_idx, NULL);
3208
			sh2 = get_active_stripe(conf, s, 0, 1, 1);
3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220
			if (sh2 == NULL)
				/* so far only the early blocks of this stripe
				 * have been requested.  When later blocks
				 * get requested, we will try again
				 */
				continue;
			if (!test_bit(STRIPE_EXPANDING, &sh2->state) ||
			   test_bit(R5_Expanded, &sh2->dev[dd_idx].flags)) {
				/* must have already done this block */
				release_stripe(sh2);
				continue;
			}
3221 3222

			/* place all the copies on one channel */
3223
			init_async_submit(&submit, 0, tx, NULL, NULL, NULL);
3224
			tx = async_memcpy(sh2->dev[dd_idx].page,
3225
					  sh->dev[i].page, 0, 0, STRIPE_SIZE,
3226
					  &submit);
3227

3228 3229 3230 3231
			set_bit(R5_Expanded, &sh2->dev[dd_idx].flags);
			set_bit(R5_UPTODATE, &sh2->dev[dd_idx].flags);
			for (j = 0; j < conf->raid_disks; j++)
				if (j != sh2->pd_idx &&
3232
				    j != sh2->qd_idx &&
3233 3234 3235 3236 3237 3238 3239
				    !test_bit(R5_Expanded, &sh2->dev[j].flags))
					break;
			if (j == conf->raid_disks) {
				set_bit(STRIPE_EXPAND_READY, &sh2->state);
				set_bit(STRIPE_HANDLE, &sh2->state);
			}
			release_stripe(sh2);
3240

3241
		}
3242
	/* done submitting copies, wait for them to complete */
3243
	async_tx_quiesce(&tx);
3244
}
L
Linus Torvalds 已提交
3245 3246 3247 3248

/*
 * handle_stripe - do things to a stripe.
 *
3249 3250
 * We lock the stripe by setting STRIPE_ACTIVE and then examine the
 * state of various bits to see what needs to be done.
L
Linus Torvalds 已提交
3251
 * Possible results:
3252 3253
 *    return some read requests which now have data
 *    return some write requests which are safely on storage
L
Linus Torvalds 已提交
3254 3255 3256 3257 3258
 *    schedule a read on some buffers
 *    schedule a write of some buffers
 *    return confirmation of parity correctness
 *
 */
3259

3260
static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
L
Linus Torvalds 已提交
3261
{
3262
	struct r5conf *conf = sh->raid_conf;
3263
	int disks = sh->disks;
3264 3265
	struct r5dev *dev;
	int i;
3266
	int do_recovery = 0;
L
Linus Torvalds 已提交
3267

3268 3269 3270 3271 3272 3273
	memset(s, 0, sizeof(*s));

	s->expanding = test_bit(STRIPE_EXPAND_SOURCE, &sh->state);
	s->expanded = test_bit(STRIPE_EXPAND_READY, &sh->state);
	s->failed_num[0] = -1;
	s->failed_num[1] = -1;
L
Linus Torvalds 已提交
3274

3275
	/* Now to look around and see what can be done */
L
Linus Torvalds 已提交
3276
	rcu_read_lock();
3277
	for (i=disks; i--; ) {
3278
		struct md_rdev *rdev;
3279 3280 3281
		sector_t first_bad;
		int bad_sectors;
		int is_bad = 0;
3282

3283
		dev = &sh->dev[i];
L
Linus Torvalds 已提交
3284

3285
		pr_debug("check %d: state 0x%lx read %p write %p written %p\n",
3286 3287
			 i, dev->flags,
			 dev->toread, dev->towrite, dev->written);
3288 3289 3290 3291 3292 3293 3294 3295
		/* maybe we can reply to a read
		 *
		 * new wantfill requests are only permitted while
		 * ops_complete_biofill is guaranteed to be inactive
		 */
		if (test_bit(R5_UPTODATE, &dev->flags) && dev->toread &&
		    !test_bit(STRIPE_BIOFILL_RUN, &sh->state))
			set_bit(R5_Wantfill, &dev->flags);
L
Linus Torvalds 已提交
3296

3297
		/* now count some things */
3298 3299 3300 3301
		if (test_bit(R5_LOCKED, &dev->flags))
			s->locked++;
		if (test_bit(R5_UPTODATE, &dev->flags))
			s->uptodate++;
3302
		if (test_bit(R5_Wantcompute, &dev->flags)) {
3303 3304
			s->compute++;
			BUG_ON(s->compute > 2);
3305
		}
L
Linus Torvalds 已提交
3306

3307
		if (test_bit(R5_Wantfill, &dev->flags))
3308
			s->to_fill++;
3309
		else if (dev->toread)
3310
			s->to_read++;
3311
		if (dev->towrite) {
3312
			s->to_write++;
3313
			if (!test_bit(R5_OVERWRITE, &dev->flags))
3314
				s->non_overwrite++;
3315
		}
3316
		if (dev->written)
3317
			s->written++;
3318 3319 3320 3321 3322 3323 3324 3325 3326 3327
		/* Prefer to use the replacement for reads, but only
		 * if it is recovered enough and has no bad blocks.
		 */
		rdev = rcu_dereference(conf->disks[i].replacement);
		if (rdev && !test_bit(Faulty, &rdev->flags) &&
		    rdev->recovery_offset >= sh->sector + STRIPE_SECTORS &&
		    !is_badblock(rdev, sh->sector, STRIPE_SECTORS,
				 &first_bad, &bad_sectors))
			set_bit(R5_ReadRepl, &dev->flags);
		else {
3328 3329
			if (rdev)
				set_bit(R5_NeedReplace, &dev->flags);
3330 3331 3332
			rdev = rcu_dereference(conf->disks[i].rdev);
			clear_bit(R5_ReadRepl, &dev->flags);
		}
3333 3334
		if (rdev && test_bit(Faulty, &rdev->flags))
			rdev = NULL;
3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346
		if (rdev) {
			is_bad = is_badblock(rdev, sh->sector, STRIPE_SECTORS,
					     &first_bad, &bad_sectors);
			if (s->blocked_rdev == NULL
			    && (test_bit(Blocked, &rdev->flags)
				|| is_bad < 0)) {
				if (is_bad < 0)
					set_bit(BlockedBadBlocks,
						&rdev->flags);
				s->blocked_rdev = rdev;
				atomic_inc(&rdev->nr_pending);
			}
3347
		}
3348 3349 3350
		clear_bit(R5_Insync, &dev->flags);
		if (!rdev)
			/* Not in-sync */;
3351 3352
		else if (is_bad) {
			/* also not in-sync */
3353 3354
			if (!test_bit(WriteErrorSeen, &rdev->flags) &&
			    test_bit(R5_UPTODATE, &dev->flags)) {
3355 3356 3357 3358 3359 3360 3361
				/* treat as in-sync, but with a read error
				 * which we can now try to correct
				 */
				set_bit(R5_Insync, &dev->flags);
				set_bit(R5_ReadError, &dev->flags);
			}
		} else if (test_bit(In_sync, &rdev->flags))
3362
			set_bit(R5_Insync, &dev->flags);
3363
		else if (sh->sector + STRIPE_SECTORS <= rdev->recovery_offset)
3364
			/* in sync if before recovery_offset */
3365 3366 3367 3368 3369 3370 3371 3372 3373
			set_bit(R5_Insync, &dev->flags);
		else if (test_bit(R5_UPTODATE, &dev->flags) &&
			 test_bit(R5_Expanded, &dev->flags))
			/* If we've reshaped into here, we assume it is Insync.
			 * We will shortly update recovery_offset to make
			 * it official.
			 */
			set_bit(R5_Insync, &dev->flags);

A
Adam Kwolek 已提交
3374
		if (rdev && test_bit(R5_WriteError, &dev->flags)) {
3375 3376 3377 3378 3379 3380 3381
			/* This flag does not apply to '.replacement'
			 * only to .rdev, so make sure to check that*/
			struct md_rdev *rdev2 = rcu_dereference(
				conf->disks[i].rdev);
			if (rdev2 == rdev)
				clear_bit(R5_Insync, &dev->flags);
			if (rdev2 && !test_bit(Faulty, &rdev2->flags)) {
3382
				s->handle_bad_blocks = 1;
3383
				atomic_inc(&rdev2->nr_pending);
3384 3385 3386
			} else
				clear_bit(R5_WriteError, &dev->flags);
		}
A
Adam Kwolek 已提交
3387
		if (rdev && test_bit(R5_MadeGood, &dev->flags)) {
3388 3389 3390 3391 3392
			/* This flag does not apply to '.replacement'
			 * only to .rdev, so make sure to check that*/
			struct md_rdev *rdev2 = rcu_dereference(
				conf->disks[i].rdev);
			if (rdev2 && !test_bit(Faulty, &rdev2->flags)) {
3393
				s->handle_bad_blocks = 1;
3394
				atomic_inc(&rdev2->nr_pending);
3395 3396 3397
			} else
				clear_bit(R5_MadeGood, &dev->flags);
		}
3398 3399 3400 3401 3402 3403 3404 3405 3406
		if (test_bit(R5_MadeGoodRepl, &dev->flags)) {
			struct md_rdev *rdev2 = rcu_dereference(
				conf->disks[i].replacement);
			if (rdev2 && !test_bit(Faulty, &rdev2->flags)) {
				s->handle_bad_blocks = 1;
				atomic_inc(&rdev2->nr_pending);
			} else
				clear_bit(R5_MadeGoodRepl, &dev->flags);
		}
3407
		if (!test_bit(R5_Insync, &dev->flags)) {
3408 3409 3410
			/* The ReadError flag will just be confusing now */
			clear_bit(R5_ReadError, &dev->flags);
			clear_bit(R5_ReWrite, &dev->flags);
L
Linus Torvalds 已提交
3411
		}
3412 3413 3414
		if (test_bit(R5_ReadError, &dev->flags))
			clear_bit(R5_Insync, &dev->flags);
		if (!test_bit(R5_Insync, &dev->flags)) {
3415 3416 3417
			if (s->failed < 2)
				s->failed_num[s->failed] = i;
			s->failed++;
3418 3419
			if (rdev && !test_bit(Faulty, &rdev->flags))
				do_recovery = 1;
3420
		}
L
Linus Torvalds 已提交
3421
	}
3422 3423 3424 3425
	if (test_bit(STRIPE_SYNCING, &sh->state)) {
		/* If there is a failed device being replaced,
		 *     we must be recovering.
		 * else if we are after recovery_cp, we must be syncing
3426
		 * else if MD_RECOVERY_REQUESTED is set, we also are syncing.
3427 3428 3429 3430 3431
		 * else we can only be replacing
		 * sync and recovery both need to read all devices, and so
		 * use the same flag.
		 */
		if (do_recovery ||
3432 3433
		    sh->sector >= conf->mddev->recovery_cp ||
		    test_bit(MD_RECOVERY_REQUESTED, &(conf->mddev->recovery)))
3434 3435 3436 3437
			s->syncing = 1;
		else
			s->replacing = 1;
	}
L
Linus Torvalds 已提交
3438
	rcu_read_unlock();
3439 3440 3441 3442 3443
}

static void handle_stripe(struct stripe_head *sh)
{
	struct stripe_head_state s;
3444
	struct r5conf *conf = sh->raid_conf;
3445
	int i;
3446 3447
	int prexor;
	int disks = sh->disks;
3448
	struct r5dev *pdev, *qdev;
3449 3450

	clear_bit(STRIPE_HANDLE, &sh->state);
3451
	if (test_and_set_bit_lock(STRIPE_ACTIVE, &sh->state)) {
3452 3453 3454 3455 3456 3457
		/* already being handled, ensure it gets handled
		 * again when current action finishes */
		set_bit(STRIPE_HANDLE, &sh->state);
		return;
	}

3458 3459 3460 3461 3462 3463 3464
	if (test_bit(STRIPE_SYNC_REQUESTED, &sh->state)) {
		spin_lock(&sh->stripe_lock);
		/* Cannot process 'sync' concurrently with 'discard' */
		if (!test_bit(STRIPE_DISCARD, &sh->state) &&
		    test_and_clear_bit(STRIPE_SYNC_REQUESTED, &sh->state)) {
			set_bit(STRIPE_SYNCING, &sh->state);
			clear_bit(STRIPE_INSYNC, &sh->state);
3465
			clear_bit(STRIPE_REPLACED, &sh->state);
3466 3467
		}
		spin_unlock(&sh->stripe_lock);
3468 3469 3470 3471 3472 3473 3474 3475
	}
	clear_bit(STRIPE_DELAYED, &sh->state);

	pr_debug("handling stripe %llu, state=%#lx cnt=%d, "
		"pd_idx=%d, qd_idx=%d\n, check:%d, reconstruct:%d\n",
	       (unsigned long long)sh->sector, sh->state,
	       atomic_read(&sh->count), sh->pd_idx, sh->qd_idx,
	       sh->check_state, sh->reconstruct_state);
3476

3477
	analyse_stripe(sh, &s);
3478

3479 3480 3481 3482 3483
	if (s.handle_bad_blocks) {
		set_bit(STRIPE_HANDLE, &sh->state);
		goto finish;
	}

3484 3485
	if (unlikely(s.blocked_rdev)) {
		if (s.syncing || s.expanding || s.expanded ||
3486
		    s.replacing || s.to_write || s.written) {
3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506
			set_bit(STRIPE_HANDLE, &sh->state);
			goto finish;
		}
		/* There is nothing for the blocked_rdev to block */
		rdev_dec_pending(s.blocked_rdev, conf->mddev);
		s.blocked_rdev = NULL;
	}

	if (s.to_fill && !test_bit(STRIPE_BIOFILL_RUN, &sh->state)) {
		set_bit(STRIPE_OP_BIOFILL, &s.ops_request);
		set_bit(STRIPE_BIOFILL_RUN, &sh->state);
	}

	pr_debug("locked=%d uptodate=%d to_read=%d"
	       " to_write=%d failed=%d failed_num=%d,%d\n",
	       s.locked, s.uptodate, s.to_read, s.to_write, s.failed,
	       s.failed_num[0], s.failed_num[1]);
	/* check if the array has lost more than max_degraded devices and,
	 * if so, some requests might need to be failed.
	 */
3507 3508 3509 3510 3511
	if (s.failed > conf->max_degraded) {
		sh->check_state = 0;
		sh->reconstruct_state = 0;
		if (s.to_read+s.to_write+s.written)
			handle_failed_stripe(conf, sh, &s, disks, &s.return_bi);
3512
		if (s.syncing + s.replacing)
3513 3514
			handle_failed_sync(conf, sh, &s);
	}
3515

3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528
	/* Now we check to see if any write operations have recently
	 * completed
	 */
	prexor = 0;
	if (sh->reconstruct_state == reconstruct_state_prexor_drain_result)
		prexor = 1;
	if (sh->reconstruct_state == reconstruct_state_drain_result ||
	    sh->reconstruct_state == reconstruct_state_prexor_drain_result) {
		sh->reconstruct_state = reconstruct_state_idle;

		/* All the 'written' buffers and the parity block are ready to
		 * be written back to disk
		 */
3529 3530
		BUG_ON(!test_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags) &&
		       !test_bit(R5_Discard, &sh->dev[sh->pd_idx].flags));
3531
		BUG_ON(sh->qd_idx >= 0 &&
3532 3533
		       !test_bit(R5_UPTODATE, &sh->dev[sh->qd_idx].flags) &&
		       !test_bit(R5_Discard, &sh->dev[sh->qd_idx].flags));
3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552
		for (i = disks; i--; ) {
			struct r5dev *dev = &sh->dev[i];
			if (test_bit(R5_LOCKED, &dev->flags) &&
				(i == sh->pd_idx || i == sh->qd_idx ||
				 dev->written)) {
				pr_debug("Writing block %d\n", i);
				set_bit(R5_Wantwrite, &dev->flags);
				if (prexor)
					continue;
				if (!test_bit(R5_Insync, &dev->flags) ||
				    ((i == sh->pd_idx || i == sh->qd_idx)  &&
				     s.failed == 0))
					set_bit(STRIPE_INSYNC, &sh->state);
			}
		}
		if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
			s.dec_preread_active = 1;
	}

3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586
	/*
	 * might be able to return some write requests if the parity blocks
	 * are safe, or on a failed drive
	 */
	pdev = &sh->dev[sh->pd_idx];
	s.p_failed = (s.failed >= 1 && s.failed_num[0] == sh->pd_idx)
		|| (s.failed >= 2 && s.failed_num[1] == sh->pd_idx);
	qdev = &sh->dev[sh->qd_idx];
	s.q_failed = (s.failed >= 1 && s.failed_num[0] == sh->qd_idx)
		|| (s.failed >= 2 && s.failed_num[1] == sh->qd_idx)
		|| conf->level < 6;

	if (s.written &&
	    (s.p_failed || ((test_bit(R5_Insync, &pdev->flags)
			     && !test_bit(R5_LOCKED, &pdev->flags)
			     && (test_bit(R5_UPTODATE, &pdev->flags) ||
				 test_bit(R5_Discard, &pdev->flags))))) &&
	    (s.q_failed || ((test_bit(R5_Insync, &qdev->flags)
			     && !test_bit(R5_LOCKED, &qdev->flags)
			     && (test_bit(R5_UPTODATE, &qdev->flags) ||
				 test_bit(R5_Discard, &qdev->flags))))))
		handle_stripe_clean_event(conf, sh, disks, &s.return_bi);

	/* Now we might consider reading some blocks, either to check/generate
	 * parity, or to satisfy requests
	 * or to load a block that is being partially written.
	 */
	if (s.to_read || s.non_overwrite
	    || (conf->level == 6 && s.to_write && s.failed)
	    || (s.syncing && (s.uptodate + s.compute < disks))
	    || s.replacing
	    || s.expanding)
		handle_stripe_fill(sh, &s, disks);

3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609
	/* Now to consider new write requests and what else, if anything
	 * should be read.  We do not handle new writes when:
	 * 1/ A 'write' operation (copy+xor) is already in flight.
	 * 2/ A 'check' operation is in flight, as it may clobber the parity
	 *    block.
	 */
	if (s.to_write && !sh->reconstruct_state && !sh->check_state)
		handle_stripe_dirtying(conf, sh, &s, disks);

	/* maybe we need to check and possibly fix the parity for this stripe
	 * Any reads will already have been scheduled, so we just see if enough
	 * data is available.  The parity check is held off while parity
	 * dependent operations are in flight.
	 */
	if (sh->check_state ||
	    (s.syncing && s.locked == 0 &&
	     !test_bit(STRIPE_COMPUTE_RUN, &sh->state) &&
	     !test_bit(STRIPE_INSYNC, &sh->state))) {
		if (conf->level == 6)
			handle_parity_checks6(conf, sh, &s, disks);
		else
			handle_parity_checks5(conf, sh, &s, disks);
	}
3610

3611 3612 3613
	if ((s.replacing || s.syncing) && s.locked == 0
	    && !test_bit(STRIPE_COMPUTE_RUN, &sh->state)
	    && !test_bit(STRIPE_REPLACED, &sh->state)) {
3614 3615
		/* Write out to replacement devices where possible */
		for (i = 0; i < conf->raid_disks; i++)
3616 3617
			if (test_bit(R5_NeedReplace, &sh->dev[i].flags)) {
				WARN_ON(!test_bit(R5_UPTODATE, &sh->dev[i].flags));
3618 3619 3620 3621
				set_bit(R5_WantReplace, &sh->dev[i].flags);
				set_bit(R5_LOCKED, &sh->dev[i].flags);
				s.locked++;
			}
3622 3623 3624
		if (s.replacing)
			set_bit(STRIPE_INSYNC, &sh->state);
		set_bit(STRIPE_REPLACED, &sh->state);
3625 3626
	}
	if ((s.syncing || s.replacing) && s.locked == 0 &&
3627
	    !test_bit(STRIPE_COMPUTE_RUN, &sh->state) &&
3628
	    test_bit(STRIPE_INSYNC, &sh->state)) {
3629 3630
		md_done_sync(conf->mddev, STRIPE_SECTORS, 1);
		clear_bit(STRIPE_SYNCING, &sh->state);
3631 3632
		if (test_and_clear_bit(R5_Overlap, &sh->dev[sh->pd_idx].flags))
			wake_up(&conf->wait_for_overlap);
3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659
	}

	/* If the failed drives are just a ReadError, then we might need
	 * to progress the repair/check process
	 */
	if (s.failed <= conf->max_degraded && !conf->mddev->ro)
		for (i = 0; i < s.failed; i++) {
			struct r5dev *dev = &sh->dev[s.failed_num[i]];
			if (test_bit(R5_ReadError, &dev->flags)
			    && !test_bit(R5_LOCKED, &dev->flags)
			    && test_bit(R5_UPTODATE, &dev->flags)
				) {
				if (!test_bit(R5_ReWrite, &dev->flags)) {
					set_bit(R5_Wantwrite, &dev->flags);
					set_bit(R5_ReWrite, &dev->flags);
					set_bit(R5_LOCKED, &dev->flags);
					s.locked++;
				} else {
					/* let's read it back */
					set_bit(R5_Wantread, &dev->flags);
					set_bit(R5_LOCKED, &dev->flags);
					s.locked++;
				}
			}
		}


3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686
	/* Finish reconstruct operations initiated by the expansion process */
	if (sh->reconstruct_state == reconstruct_state_result) {
		struct stripe_head *sh_src
			= get_active_stripe(conf, sh->sector, 1, 1, 1);
		if (sh_src && test_bit(STRIPE_EXPAND_SOURCE, &sh_src->state)) {
			/* sh cannot be written until sh_src has been read.
			 * so arrange for sh to be delayed a little
			 */
			set_bit(STRIPE_DELAYED, &sh->state);
			set_bit(STRIPE_HANDLE, &sh->state);
			if (!test_and_set_bit(STRIPE_PREREAD_ACTIVE,
					      &sh_src->state))
				atomic_inc(&conf->preread_active_stripes);
			release_stripe(sh_src);
			goto finish;
		}
		if (sh_src)
			release_stripe(sh_src);

		sh->reconstruct_state = reconstruct_state_idle;
		clear_bit(STRIPE_EXPANDING, &sh->state);
		for (i = conf->raid_disks; i--; ) {
			set_bit(R5_Wantwrite, &sh->dev[i].flags);
			set_bit(R5_LOCKED, &sh->dev[i].flags);
			s.locked++;
		}
	}
3687

3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703
	if (s.expanded && test_bit(STRIPE_EXPANDING, &sh->state) &&
	    !sh->reconstruct_state) {
		/* Need to write out all blocks after computing parity */
		sh->disks = conf->raid_disks;
		stripe_set_idx(sh->sector, conf, 0, sh);
		schedule_reconstruction(sh, &s, 1, 1);
	} else if (s.expanded && !sh->reconstruct_state && s.locked == 0) {
		clear_bit(STRIPE_EXPAND_READY, &sh->state);
		atomic_dec(&conf->reshape_stripes);
		wake_up(&conf->wait_for_overlap);
		md_done_sync(conf->mddev, STRIPE_SECTORS, 1);
	}

	if (s.expanding && s.locked == 0 &&
	    !test_bit(STRIPE_COMPUTE_RUN, &sh->state))
		handle_stripe_expansion(conf, sh);
3704

3705
finish:
3706
	/* wait for this device to become unblocked */
3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718
	if (unlikely(s.blocked_rdev)) {
		if (conf->mddev->external)
			md_wait_for_blocked_rdev(s.blocked_rdev,
						 conf->mddev);
		else
			/* Internal metadata will immediately
			 * be written by raid5d, so we don't
			 * need to wait here.
			 */
			rdev_dec_pending(s.blocked_rdev,
					 conf->mddev);
	}
3719

3720 3721
	if (s.handle_bad_blocks)
		for (i = disks; i--; ) {
3722
			struct md_rdev *rdev;
3723 3724 3725 3726 3727 3728 3729 3730 3731
			struct r5dev *dev = &sh->dev[i];
			if (test_and_clear_bit(R5_WriteError, &dev->flags)) {
				/* We own a safe reference to the rdev */
				rdev = conf->disks[i].rdev;
				if (!rdev_set_badblocks(rdev, sh->sector,
							STRIPE_SECTORS, 0))
					md_error(conf->mddev, rdev);
				rdev_dec_pending(rdev, conf->mddev);
			}
3732 3733 3734
			if (test_and_clear_bit(R5_MadeGood, &dev->flags)) {
				rdev = conf->disks[i].rdev;
				rdev_clear_badblocks(rdev, sh->sector,
3735
						     STRIPE_SECTORS, 0);
3736 3737
				rdev_dec_pending(rdev, conf->mddev);
			}
3738 3739
			if (test_and_clear_bit(R5_MadeGoodRepl, &dev->flags)) {
				rdev = conf->disks[i].replacement;
3740 3741 3742
				if (!rdev)
					/* rdev have been moved down */
					rdev = conf->disks[i].rdev;
3743
				rdev_clear_badblocks(rdev, sh->sector,
3744
						     STRIPE_SECTORS, 0);
3745 3746
				rdev_dec_pending(rdev, conf->mddev);
			}
3747 3748
		}

3749 3750 3751
	if (s.ops_request)
		raid_run_ops(sh, s.ops_request);

D
Dan Williams 已提交
3752
	ops_run_io(sh, &s);
3753

3754
	if (s.dec_preread_active) {
3755
		/* We delay this until after ops_run_io so that if make_request
T
Tejun Heo 已提交
3756
		 * is waiting on a flush, it won't continue until the writes
3757 3758 3759 3760 3761 3762 3763 3764
		 * have actually been submitted.
		 */
		atomic_dec(&conf->preread_active_stripes);
		if (atomic_read(&conf->preread_active_stripes) <
		    IO_THRESHOLD)
			md_wakeup_thread(conf->mddev->thread);
	}

3765
	return_io(s.return_bi);
3766

3767
	clear_bit_unlock(STRIPE_ACTIVE, &sh->state);
3768 3769
}

3770
static void raid5_activate_delayed(struct r5conf *conf)
3771 3772 3773 3774 3775 3776 3777 3778 3779 3780
{
	if (atomic_read(&conf->preread_active_stripes) < IO_THRESHOLD) {
		while (!list_empty(&conf->delayed_list)) {
			struct list_head *l = conf->delayed_list.next;
			struct stripe_head *sh;
			sh = list_entry(l, struct stripe_head, lru);
			list_del_init(l);
			clear_bit(STRIPE_DELAYED, &sh->state);
			if (!test_and_set_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
				atomic_inc(&conf->preread_active_stripes);
3781
			list_add_tail(&sh->lru, &conf->hold_list);
3782
		}
N
NeilBrown 已提交
3783
	}
3784 3785
}

3786
static void activate_bit_delay(struct r5conf *conf)
3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799
{
	/* device_lock is held */
	struct list_head head;
	list_add(&head, &conf->bitmap_list);
	list_del_init(&conf->bitmap_list);
	while (!list_empty(&head)) {
		struct stripe_head *sh = list_entry(head.next, struct stripe_head, lru);
		list_del_init(&sh->lru);
		atomic_inc(&sh->count);
		__release_stripe(conf, sh);
	}
}

3800
int md_raid5_congested(struct mddev *mddev, int bits)
3801
{
3802
	struct r5conf *conf = mddev->private;
3803 3804 3805 3806

	/* No difference between reads and writes.  Just check
	 * how busy the stripe_cache is
	 */
3807

3808 3809 3810 3811 3812 3813 3814 3815 3816
	if (conf->inactive_blocked)
		return 1;
	if (conf->quiesce)
		return 1;
	if (list_empty_careful(&conf->inactive_list))
		return 1;

	return 0;
}
N
NeilBrown 已提交
3817 3818 3819 3820
EXPORT_SYMBOL_GPL(md_raid5_congested);

static int raid5_congested(void *data, int bits)
{
3821
	struct mddev *mddev = data;
N
NeilBrown 已提交
3822 3823 3824 3825

	return mddev_congested(mddev, bits) ||
		md_raid5_congested(mddev, bits);
}
3826

3827 3828 3829
/* We want read requests to align with chunks where possible,
 * but write requests don't need to.
 */
3830 3831 3832
static int raid5_mergeable_bvec(struct request_queue *q,
				struct bvec_merge_data *bvm,
				struct bio_vec *biovec)
3833
{
3834
	struct mddev *mddev = q->queuedata;
3835
	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
3836
	int max;
3837
	unsigned int chunk_sectors = mddev->chunk_sectors;
3838
	unsigned int bio_sectors = bvm->bi_size >> 9;
3839

3840
	if ((bvm->bi_rw & 1) == WRITE)
3841 3842
		return biovec->bv_len; /* always allow writes to be mergeable */

3843 3844
	if (mddev->new_chunk_sectors < mddev->chunk_sectors)
		chunk_sectors = mddev->new_chunk_sectors;
3845 3846 3847 3848 3849 3850 3851 3852
	max =  (chunk_sectors - ((sector & (chunk_sectors - 1)) + bio_sectors)) << 9;
	if (max < 0) max = 0;
	if (max <= biovec->bv_len && bio_sectors == 0)
		return biovec->bv_len;
	else
		return max;
}

3853

3854
static int in_chunk_boundary(struct mddev *mddev, struct bio *bio)
3855 3856
{
	sector_t sector = bio->bi_sector + get_start_sect(bio->bi_bdev);
3857
	unsigned int chunk_sectors = mddev->chunk_sectors;
3858
	unsigned int bio_sectors = bio_sectors(bio);
3859

3860 3861
	if (mddev->new_chunk_sectors < mddev->chunk_sectors)
		chunk_sectors = mddev->new_chunk_sectors;
3862 3863 3864 3865
	return  chunk_sectors >=
		((sector & (chunk_sectors - 1)) + bio_sectors);
}

3866 3867 3868 3869
/*
 *  add bio to the retry LIFO  ( in O(1) ... we are in interrupt )
 *  later sampled by raid5d.
 */
3870
static void add_bio_to_retry(struct bio *bi,struct r5conf *conf)
3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883
{
	unsigned long flags;

	spin_lock_irqsave(&conf->device_lock, flags);

	bi->bi_next = conf->retry_read_aligned_list;
	conf->retry_read_aligned_list = bi;

	spin_unlock_irqrestore(&conf->device_lock, flags);
	md_wakeup_thread(conf->mddev->thread);
}


3884
static struct bio *remove_bio_from_retry(struct r5conf *conf)
3885 3886 3887 3888 3889 3890 3891 3892 3893 3894
{
	struct bio *bi;

	bi = conf->retry_read_aligned;
	if (bi) {
		conf->retry_read_aligned = NULL;
		return bi;
	}
	bi = conf->retry_read_aligned_list;
	if(bi) {
3895
		conf->retry_read_aligned_list = bi->bi_next;
3896
		bi->bi_next = NULL;
3897 3898 3899 3900
		/*
		 * this sets the active strip count to 1 and the processed
		 * strip count to zero (upper 8 bits)
		 */
3901
		raid5_set_bi_stripes(bi, 1); /* biased count of active stripes */
3902 3903 3904 3905 3906 3907
	}

	return bi;
}


3908 3909 3910 3911 3912 3913
/*
 *  The "raid5_align_endio" should check if the read succeeded and if it
 *  did, call bio_endio on the original bio (having bio_put the new bio
 *  first).
 *  If the read failed..
 */
3914
static void raid5_align_endio(struct bio *bi, int error)
3915 3916
{
	struct bio* raid_bi  = bi->bi_private;
3917
	struct mddev *mddev;
3918
	struct r5conf *conf;
3919
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
3920
	struct md_rdev *rdev;
3921

3922
	bio_put(bi);
3923 3924 3925

	rdev = (void*)raid_bi->bi_next;
	raid_bi->bi_next = NULL;
3926 3927
	mddev = rdev->mddev;
	conf = mddev->private;
3928 3929 3930 3931

	rdev_dec_pending(rdev, conf->mddev);

	if (!error && uptodate) {
3932 3933
		trace_block_bio_complete(bdev_get_queue(raid_bi->bi_bdev),
					 raid_bi, 0);
3934
		bio_endio(raid_bi, 0);
3935 3936
		if (atomic_dec_and_test(&conf->active_aligned_reads))
			wake_up(&conf->wait_for_stripe);
3937
		return;
3938 3939 3940
	}


3941
	pr_debug("raid5_align_endio : io error...handing IO for a retry\n");
3942 3943

	add_bio_to_retry(raid_bi, conf);
3944 3945
}

3946 3947
static int bio_fits_rdev(struct bio *bi)
{
3948
	struct request_queue *q = bdev_get_queue(bi->bi_bdev);
3949

3950
	if (bio_sectors(bi) > queue_max_sectors(q))
3951 3952
		return 0;
	blk_recount_segments(q, bi);
3953
	if (bi->bi_phys_segments > queue_max_segments(q))
3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965
		return 0;

	if (q->merge_bvec_fn)
		/* it's too hard to apply the merge_bvec_fn at this stage,
		 * just just give up
		 */
		return 0;

	return 1;
}


3966
static int chunk_aligned_read(struct mddev *mddev, struct bio * raid_bio)
3967
{
3968
	struct r5conf *conf = mddev->private;
N
NeilBrown 已提交
3969
	int dd_idx;
3970
	struct bio* align_bi;
3971
	struct md_rdev *rdev;
3972
	sector_t end_sector;
3973 3974

	if (!in_chunk_boundary(mddev, raid_bio)) {
3975
		pr_debug("chunk_aligned_read : non aligned\n");
3976 3977 3978
		return 0;
	}
	/*
3979
	 * use bio_clone_mddev to make a copy of the bio
3980
	 */
3981
	align_bi = bio_clone_mddev(raid_bio, GFP_NOIO, mddev);
3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992
	if (!align_bi)
		return 0;
	/*
	 *   set bi_end_io to a new function, and set bi_private to the
	 *     original bio.
	 */
	align_bi->bi_end_io  = raid5_align_endio;
	align_bi->bi_private = raid_bio;
	/*
	 *	compute position
	 */
3993 3994
	align_bi->bi_sector =  raid5_compute_sector(conf, raid_bio->bi_sector,
						    0,
3995
						    &dd_idx, NULL);
3996

K
Kent Overstreet 已提交
3997
	end_sector = bio_end_sector(align_bi);
3998
	rcu_read_lock();
3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009
	rdev = rcu_dereference(conf->disks[dd_idx].replacement);
	if (!rdev || test_bit(Faulty, &rdev->flags) ||
	    rdev->recovery_offset < end_sector) {
		rdev = rcu_dereference(conf->disks[dd_idx].rdev);
		if (rdev &&
		    (test_bit(Faulty, &rdev->flags) ||
		    !(test_bit(In_sync, &rdev->flags) ||
		      rdev->recovery_offset >= end_sector)))
			rdev = NULL;
	}
	if (rdev) {
4010 4011 4012
		sector_t first_bad;
		int bad_sectors;

4013 4014
		atomic_inc(&rdev->nr_pending);
		rcu_read_unlock();
4015 4016 4017 4018
		raid_bio->bi_next = (void*)rdev;
		align_bi->bi_bdev =  rdev->bdev;
		align_bi->bi_flags &= ~(1 << BIO_SEG_VALID);

4019
		if (!bio_fits_rdev(align_bi) ||
4020
		    is_badblock(rdev, align_bi->bi_sector, bio_sectors(align_bi),
4021 4022
				&first_bad, &bad_sectors)) {
			/* too big in some way, or has a known bad block */
4023 4024 4025 4026 4027
			bio_put(align_bi);
			rdev_dec_pending(rdev, mddev);
			return 0;
		}

4028 4029 4030
		/* No reshape active, so we can trust rdev->data_offset */
		align_bi->bi_sector += rdev->data_offset;

4031 4032 4033
		spin_lock_irq(&conf->device_lock);
		wait_event_lock_irq(conf->wait_for_stripe,
				    conf->quiesce == 0,
4034
				    conf->device_lock);
4035 4036 4037
		atomic_inc(&conf->active_aligned_reads);
		spin_unlock_irq(&conf->device_lock);

4038 4039 4040 4041
		if (mddev->gendisk)
			trace_block_bio_remap(bdev_get_queue(align_bi->bi_bdev),
					      align_bi, disk_devt(mddev->gendisk),
					      raid_bio->bi_sector);
4042 4043 4044 4045
		generic_make_request(align_bi);
		return 1;
	} else {
		rcu_read_unlock();
4046
		bio_put(align_bi);
4047 4048 4049 4050
		return 0;
	}
}

4051 4052 4053 4054 4055 4056 4057 4058 4059 4060
/* __get_priority_stripe - get the next stripe to process
 *
 * Full stripe writes are allowed to pass preread active stripes up until
 * the bypass_threshold is exceeded.  In general the bypass_count
 * increments when the handle_list is handled before the hold_list; however, it
 * will not be incremented when STRIPE_IO_STARTED is sampled set signifying a
 * stripe with in flight i/o.  The bypass_count will be reset when the
 * head of the hold_list has changed, i.e. the head was promoted to the
 * handle_list.
 */
4061
static struct stripe_head *__get_priority_stripe(struct r5conf *conf)
4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102
{
	struct stripe_head *sh;

	pr_debug("%s: handle: %s hold: %s full_writes: %d bypass_count: %d\n",
		  __func__,
		  list_empty(&conf->handle_list) ? "empty" : "busy",
		  list_empty(&conf->hold_list) ? "empty" : "busy",
		  atomic_read(&conf->pending_full_writes), conf->bypass_count);

	if (!list_empty(&conf->handle_list)) {
		sh = list_entry(conf->handle_list.next, typeof(*sh), lru);

		if (list_empty(&conf->hold_list))
			conf->bypass_count = 0;
		else if (!test_bit(STRIPE_IO_STARTED, &sh->state)) {
			if (conf->hold_list.next == conf->last_hold)
				conf->bypass_count++;
			else {
				conf->last_hold = conf->hold_list.next;
				conf->bypass_count -= conf->bypass_threshold;
				if (conf->bypass_count < 0)
					conf->bypass_count = 0;
			}
		}
	} else if (!list_empty(&conf->hold_list) &&
		   ((conf->bypass_threshold &&
		     conf->bypass_count > conf->bypass_threshold) ||
		    atomic_read(&conf->pending_full_writes) == 0)) {
		sh = list_entry(conf->hold_list.next,
				typeof(*sh), lru);
		conf->bypass_count -= conf->bypass_threshold;
		if (conf->bypass_count < 0)
			conf->bypass_count = 0;
	} else
		return NULL;

	list_del_init(&sh->lru);
	atomic_inc(&sh->count);
	BUG_ON(atomic_read(&sh->count) != 1);
	return sh;
}
4103

4104 4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115
struct raid5_plug_cb {
	struct blk_plug_cb	cb;
	struct list_head	list;
};

static void raid5_unplug(struct blk_plug_cb *blk_cb, bool from_schedule)
{
	struct raid5_plug_cb *cb = container_of(
		blk_cb, struct raid5_plug_cb, cb);
	struct stripe_head *sh;
	struct mddev *mddev = cb->cb.data;
	struct r5conf *conf = mddev->private;
N
NeilBrown 已提交
4116
	int cnt = 0;
4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130

	if (cb->list.next && !list_empty(&cb->list)) {
		spin_lock_irq(&conf->device_lock);
		while (!list_empty(&cb->list)) {
			sh = list_first_entry(&cb->list, struct stripe_head, lru);
			list_del_init(&sh->lru);
			/*
			 * avoid race release_stripe_plug() sees
			 * STRIPE_ON_UNPLUG_LIST clear but the stripe
			 * is still in our list
			 */
			smp_mb__before_clear_bit();
			clear_bit(STRIPE_ON_UNPLUG_LIST, &sh->state);
			__release_stripe(conf, sh);
N
NeilBrown 已提交
4131
			cnt++;
4132 4133 4134
		}
		spin_unlock_irq(&conf->device_lock);
	}
4135 4136
	if (mddev->queue)
		trace_block_unplug(mddev->queue, cnt, !from_schedule);
4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163
	kfree(cb);
}

static void release_stripe_plug(struct mddev *mddev,
				struct stripe_head *sh)
{
	struct blk_plug_cb *blk_cb = blk_check_plugged(
		raid5_unplug, mddev,
		sizeof(struct raid5_plug_cb));
	struct raid5_plug_cb *cb;

	if (!blk_cb) {
		release_stripe(sh);
		return;
	}

	cb = container_of(blk_cb, struct raid5_plug_cb, cb);

	if (cb->list.next == NULL)
		INIT_LIST_HEAD(&cb->list);

	if (!test_and_set_bit(STRIPE_ON_UNPLUG_LIST, &sh->state))
		list_add_tail(&sh->lru, &cb->list);
	else
		release_stripe(sh);
}

S
Shaohua Li 已提交
4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198
static void make_discard_request(struct mddev *mddev, struct bio *bi)
{
	struct r5conf *conf = mddev->private;
	sector_t logical_sector, last_sector;
	struct stripe_head *sh;
	int remaining;
	int stripe_sectors;

	if (mddev->reshape_position != MaxSector)
		/* Skip discard while reshape is happening */
		return;

	logical_sector = bi->bi_sector & ~((sector_t)STRIPE_SECTORS-1);
	last_sector = bi->bi_sector + (bi->bi_size>>9);

	bi->bi_next = NULL;
	bi->bi_phys_segments = 1; /* over-loaded to count active stripes */

	stripe_sectors = conf->chunk_sectors *
		(conf->raid_disks - conf->max_degraded);
	logical_sector = DIV_ROUND_UP_SECTOR_T(logical_sector,
					       stripe_sectors);
	sector_div(last_sector, stripe_sectors);

	logical_sector *= conf->chunk_sectors;
	last_sector *= conf->chunk_sectors;

	for (; logical_sector < last_sector;
	     logical_sector += STRIPE_SECTORS) {
		DEFINE_WAIT(w);
		int d;
	again:
		sh = get_active_stripe(conf, logical_sector, 0, 0, 0);
		prepare_to_wait(&conf->wait_for_overlap, &w,
				TASK_UNINTERRUPTIBLE);
4199 4200 4201 4202 4203 4204 4205
		set_bit(R5_Overlap, &sh->dev[sh->pd_idx].flags);
		if (test_bit(STRIPE_SYNCING, &sh->state)) {
			release_stripe(sh);
			schedule();
			goto again;
		}
		clear_bit(R5_Overlap, &sh->dev[sh->pd_idx].flags);
S
Shaohua Li 已提交
4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217
		spin_lock_irq(&sh->stripe_lock);
		for (d = 0; d < conf->raid_disks; d++) {
			if (d == sh->pd_idx || d == sh->qd_idx)
				continue;
			if (sh->dev[d].towrite || sh->dev[d].toread) {
				set_bit(R5_Overlap, &sh->dev[d].flags);
				spin_unlock_irq(&sh->stripe_lock);
				release_stripe(sh);
				schedule();
				goto again;
			}
		}
4218
		set_bit(STRIPE_DISCARD, &sh->state);
S
Shaohua Li 已提交
4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252 4253
		finish_wait(&conf->wait_for_overlap, &w);
		for (d = 0; d < conf->raid_disks; d++) {
			if (d == sh->pd_idx || d == sh->qd_idx)
				continue;
			sh->dev[d].towrite = bi;
			set_bit(R5_OVERWRITE, &sh->dev[d].flags);
			raid5_inc_bi_active_stripes(bi);
		}
		spin_unlock_irq(&sh->stripe_lock);
		if (conf->mddev->bitmap) {
			for (d = 0;
			     d < conf->raid_disks - conf->max_degraded;
			     d++)
				bitmap_startwrite(mddev->bitmap,
						  sh->sector,
						  STRIPE_SECTORS,
						  0);
			sh->bm_seq = conf->seq_flush + 1;
			set_bit(STRIPE_BIT_DELAY, &sh->state);
		}

		set_bit(STRIPE_HANDLE, &sh->state);
		clear_bit(STRIPE_DELAYED, &sh->state);
		if (!test_and_set_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
			atomic_inc(&conf->preread_active_stripes);
		release_stripe_plug(mddev, sh);
	}

	remaining = raid5_dec_bi_active_stripes(bi);
	if (remaining == 0) {
		md_write_end(mddev);
		bio_endio(bi, 0);
	}
}

4254
static void make_request(struct mddev *mddev, struct bio * bi)
L
Linus Torvalds 已提交
4255
{
4256
	struct r5conf *conf = mddev->private;
4257
	int dd_idx;
L
Linus Torvalds 已提交
4258 4259 4260
	sector_t new_sector;
	sector_t logical_sector, last_sector;
	struct stripe_head *sh;
4261
	const int rw = bio_data_dir(bi);
4262
	int remaining;
L
Linus Torvalds 已提交
4263

T
Tejun Heo 已提交
4264 4265
	if (unlikely(bi->bi_rw & REQ_FLUSH)) {
		md_flush_request(mddev, bi);
4266
		return;
4267 4268
	}

4269
	md_write_start(mddev, bi);
4270

4271
	if (rw == READ &&
4272
	     mddev->reshape_position == MaxSector &&
4273
	     chunk_aligned_read(mddev,bi))
4274
		return;
4275

S
Shaohua Li 已提交
4276 4277 4278 4279 4280
	if (unlikely(bi->bi_rw & REQ_DISCARD)) {
		make_discard_request(mddev, bi);
		return;
	}

L
Linus Torvalds 已提交
4281
	logical_sector = bi->bi_sector & ~((sector_t)STRIPE_SECTORS-1);
K
Kent Overstreet 已提交
4282
	last_sector = bio_end_sector(bi);
L
Linus Torvalds 已提交
4283 4284
	bi->bi_next = NULL;
	bi->bi_phys_segments = 1;	/* over-loaded to count active stripes */
4285

L
Linus Torvalds 已提交
4286 4287
	for (;logical_sector < last_sector; logical_sector += STRIPE_SECTORS) {
		DEFINE_WAIT(w);
4288
		int previous;
4289

4290
	retry:
4291
		previous = 0;
4292
		prepare_to_wait(&conf->wait_for_overlap, &w, TASK_UNINTERRUPTIBLE);
4293
		if (unlikely(conf->reshape_progress != MaxSector)) {
4294
			/* spinlock is needed as reshape_progress may be
4295 4296
			 * 64bit on a 32bit platform, and so it might be
			 * possible to see a half-updated value
4297
			 * Of course reshape_progress could change after
4298 4299 4300 4301
			 * the lock is dropped, so once we get a reference
			 * to the stripe that we think it is, we will have
			 * to check again.
			 */
4302
			spin_lock_irq(&conf->device_lock);
4303
			if (mddev->reshape_backwards
4304 4305
			    ? logical_sector < conf->reshape_progress
			    : logical_sector >= conf->reshape_progress) {
4306 4307
				previous = 1;
			} else {
4308
				if (mddev->reshape_backwards
4309 4310
				    ? logical_sector < conf->reshape_safe
				    : logical_sector >= conf->reshape_safe) {
4311 4312 4313 4314 4315
					spin_unlock_irq(&conf->device_lock);
					schedule();
					goto retry;
				}
			}
4316 4317
			spin_unlock_irq(&conf->device_lock);
		}
4318

4319 4320
		new_sector = raid5_compute_sector(conf, logical_sector,
						  previous,
4321
						  &dd_idx, NULL);
4322
		pr_debug("raid456: make_request, sector %llu logical %llu\n",
L
Linus Torvalds 已提交
4323 4324 4325
			(unsigned long long)new_sector, 
			(unsigned long long)logical_sector);

4326
		sh = get_active_stripe(conf, new_sector, previous,
4327
				       (bi->bi_rw&RWA_MASK), 0);
L
Linus Torvalds 已提交
4328
		if (sh) {
4329
			if (unlikely(previous)) {
4330
				/* expansion might have moved on while waiting for a
4331 4332 4333 4334 4335 4336
				 * stripe, so we must do the range check again.
				 * Expansion could still move past after this
				 * test, but as we are holding a reference to
				 * 'sh', we know that if that happens,
				 *  STRIPE_EXPANDING will get set and the expansion
				 * won't proceed until we finish with the stripe.
4337 4338 4339
				 */
				int must_retry = 0;
				spin_lock_irq(&conf->device_lock);
4340
				if (mddev->reshape_backwards
4341 4342
				    ? logical_sector >= conf->reshape_progress
				    : logical_sector < conf->reshape_progress)
4343 4344 4345 4346 4347
					/* mismatch, need to try again */
					must_retry = 1;
				spin_unlock_irq(&conf->device_lock);
				if (must_retry) {
					release_stripe(sh);
4348
					schedule();
4349 4350 4351
					goto retry;
				}
			}
4352

4353
			if (rw == WRITE &&
4354
			    logical_sector >= mddev->suspend_lo &&
4355 4356
			    logical_sector < mddev->suspend_hi) {
				release_stripe(sh);
4357 4358 4359 4360 4361 4362 4363 4364 4365 4366
				/* As the suspend_* range is controlled by
				 * userspace, we want an interruptible
				 * wait.
				 */
				flush_signals(current);
				prepare_to_wait(&conf->wait_for_overlap,
						&w, TASK_INTERRUPTIBLE);
				if (logical_sector >= mddev->suspend_lo &&
				    logical_sector < mddev->suspend_hi)
					schedule();
4367 4368
				goto retry;
			}
4369 4370

			if (test_bit(STRIPE_EXPANDING, &sh->state) ||
4371
			    !add_stripe_bio(sh, bi, dd_idx, rw)) {
4372 4373
				/* Stripe is busy expanding or
				 * add failed due to overlap.  Flush everything
L
Linus Torvalds 已提交
4374 4375
				 * and wait a while
				 */
N
NeilBrown 已提交
4376
				md_wakeup_thread(mddev->thread);
L
Linus Torvalds 已提交
4377 4378 4379 4380 4381
				release_stripe(sh);
				schedule();
				goto retry;
			}
			finish_wait(&conf->wait_for_overlap, &w);
4382 4383
			set_bit(STRIPE_HANDLE, &sh->state);
			clear_bit(STRIPE_DELAYED, &sh->state);
4384
			if ((bi->bi_rw & REQ_SYNC) &&
4385 4386
			    !test_and_set_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
				atomic_inc(&conf->preread_active_stripes);
4387
			release_stripe_plug(mddev, sh);
L
Linus Torvalds 已提交
4388 4389 4390 4391 4392 4393 4394
		} else {
			/* cannot get stripe for read-ahead, just give-up */
			clear_bit(BIO_UPTODATE, &bi->bi_flags);
			finish_wait(&conf->wait_for_overlap, &w);
			break;
		}
	}
4395

4396
	remaining = raid5_dec_bi_active_stripes(bi);
4397
	if (remaining == 0) {
L
Linus Torvalds 已提交
4398

4399
		if ( rw == WRITE )
L
Linus Torvalds 已提交
4400
			md_write_end(mddev);
4401

4402 4403
		trace_block_bio_complete(bdev_get_queue(bi->bi_bdev),
					 bi, 0);
4404
		bio_endio(bi, 0);
L
Linus Torvalds 已提交
4405 4406 4407
	}
}

4408
static sector_t raid5_size(struct mddev *mddev, sector_t sectors, int raid_disks);
D
Dan Williams 已提交
4409

4410
static sector_t reshape_request(struct mddev *mddev, sector_t sector_nr, int *skipped)
L
Linus Torvalds 已提交
4411
{
4412 4413 4414 4415 4416 4417 4418 4419 4420
	/* reshaping is quite different to recovery/resync so it is
	 * handled quite separately ... here.
	 *
	 * On each call to sync_request, we gather one chunk worth of
	 * destination stripes and flag them as expanding.
	 * Then we find all the source stripes and request reads.
	 * As the reads complete, handle_stripe will copy the data
	 * into the destination stripe and release that stripe.
	 */
4421
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
4422
	struct stripe_head *sh;
4423
	sector_t first_sector, last_sector;
4424 4425 4426
	int raid_disks = conf->previous_raid_disks;
	int data_disks = raid_disks - conf->max_degraded;
	int new_data_disks = conf->raid_disks - conf->max_degraded;
4427 4428
	int i;
	int dd_idx;
4429
	sector_t writepos, readpos, safepos;
4430
	sector_t stripe_addr;
4431
	int reshape_sectors;
4432
	struct list_head stripes;
4433

4434 4435
	if (sector_nr == 0) {
		/* If restarting in the middle, skip the initial sectors */
4436
		if (mddev->reshape_backwards &&
4437 4438 4439
		    conf->reshape_progress < raid5_size(mddev, 0, 0)) {
			sector_nr = raid5_size(mddev, 0, 0)
				- conf->reshape_progress;
4440
		} else if (!mddev->reshape_backwards &&
4441 4442
			   conf->reshape_progress > 0)
			sector_nr = conf->reshape_progress;
4443
		sector_div(sector_nr, new_data_disks);
4444
		if (sector_nr) {
4445 4446
			mddev->curr_resync_completed = sector_nr;
			sysfs_notify(&mddev->kobj, NULL, "sync_completed");
4447 4448 4449
			*skipped = 1;
			return sector_nr;
		}
4450 4451
	}

4452 4453 4454 4455
	/* We need to process a full chunk at a time.
	 * If old and new chunk sizes differ, we need to process the
	 * largest of these
	 */
4456 4457
	if (mddev->new_chunk_sectors > mddev->chunk_sectors)
		reshape_sectors = mddev->new_chunk_sectors;
4458
	else
4459
		reshape_sectors = mddev->chunk_sectors;
4460

4461 4462 4463 4464 4465
	/* We update the metadata at least every 10 seconds, or when
	 * the data about to be copied would over-write the source of
	 * the data at the front of the range.  i.e. one new_stripe
	 * along from reshape_progress new_maps to after where
	 * reshape_safe old_maps to
4466
	 */
4467
	writepos = conf->reshape_progress;
4468
	sector_div(writepos, new_data_disks);
4469 4470
	readpos = conf->reshape_progress;
	sector_div(readpos, data_disks);
4471
	safepos = conf->reshape_safe;
4472
	sector_div(safepos, data_disks);
4473
	if (mddev->reshape_backwards) {
4474
		writepos -= min_t(sector_t, reshape_sectors, writepos);
4475
		readpos += reshape_sectors;
4476
		safepos += reshape_sectors;
4477
	} else {
4478
		writepos += reshape_sectors;
4479 4480
		readpos -= min_t(sector_t, reshape_sectors, readpos);
		safepos -= min_t(sector_t, reshape_sectors, safepos);
4481
	}
4482

4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497
	/* Having calculated the 'writepos' possibly use it
	 * to set 'stripe_addr' which is where we will write to.
	 */
	if (mddev->reshape_backwards) {
		BUG_ON(conf->reshape_progress == 0);
		stripe_addr = writepos;
		BUG_ON((mddev->dev_sectors &
			~((sector_t)reshape_sectors - 1))
		       - reshape_sectors - stripe_addr
		       != sector_nr);
	} else {
		BUG_ON(writepos != sector_nr + reshape_sectors);
		stripe_addr = sector_nr;
	}

4498 4499 4500 4501
	/* 'writepos' is the most advanced device address we might write.
	 * 'readpos' is the least advanced device address we might read.
	 * 'safepos' is the least address recorded in the metadata as having
	 *     been reshaped.
4502 4503 4504 4505
	 * If there is a min_offset_diff, these are adjusted either by
	 * increasing the safepos/readpos if diff is negative, or
	 * increasing writepos if diff is positive.
	 * If 'readpos' is then behind 'writepos', there is no way that we can
4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517
	 * ensure safety in the face of a crash - that must be done by userspace
	 * making a backup of the data.  So in that case there is no particular
	 * rush to update metadata.
	 * Otherwise if 'safepos' is behind 'writepos', then we really need to
	 * update the metadata to advance 'safepos' to match 'readpos' so that
	 * we can be safe in the event of a crash.
	 * So we insist on updating metadata if safepos is behind writepos and
	 * readpos is beyond writepos.
	 * In any case, update the metadata every 10 seconds.
	 * Maybe that number should be configurable, but I'm not sure it is
	 * worth it.... maybe it could be a multiple of safemode_delay???
	 */
4518 4519 4520 4521 4522 4523
	if (conf->min_offset_diff < 0) {
		safepos += -conf->min_offset_diff;
		readpos += -conf->min_offset_diff;
	} else
		writepos += conf->min_offset_diff;

4524
	if ((mddev->reshape_backwards
4525 4526 4527
	     ? (safepos > writepos && readpos < writepos)
	     : (safepos < writepos && readpos > writepos)) ||
	    time_after(jiffies, conf->reshape_checkpoint + 10*HZ)) {
4528 4529 4530
		/* Cannot proceed until we've updated the superblock... */
		wait_event(conf->wait_for_overlap,
			   atomic_read(&conf->reshape_stripes)==0);
4531
		mddev->reshape_position = conf->reshape_progress;
4532
		mddev->curr_resync_completed = sector_nr;
4533
		conf->reshape_checkpoint = jiffies;
4534
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
4535
		md_wakeup_thread(mddev->thread);
4536
		wait_event(mddev->sb_wait, mddev->flags == 0 ||
4537 4538
			   kthread_should_stop());
		spin_lock_irq(&conf->device_lock);
4539
		conf->reshape_safe = mddev->reshape_position;
4540 4541
		spin_unlock_irq(&conf->device_lock);
		wake_up(&conf->wait_for_overlap);
4542
		sysfs_notify(&mddev->kobj, NULL, "sync_completed");
4543 4544
	}

4545
	INIT_LIST_HEAD(&stripes);
4546
	for (i = 0; i < reshape_sectors; i += STRIPE_SECTORS) {
4547
		int j;
4548
		int skipped_disk = 0;
4549
		sh = get_active_stripe(conf, stripe_addr+i, 0, 0, 1);
4550 4551 4552 4553 4554 4555 4556 4557 4558
		set_bit(STRIPE_EXPANDING, &sh->state);
		atomic_inc(&conf->reshape_stripes);
		/* If any of this stripe is beyond the end of the old
		 * array, then we need to zero those blocks
		 */
		for (j=sh->disks; j--;) {
			sector_t s;
			if (j == sh->pd_idx)
				continue;
4559
			if (conf->level == 6 &&
4560
			    j == sh->qd_idx)
4561
				continue;
4562
			s = compute_blocknr(sh, j, 0);
D
Dan Williams 已提交
4563
			if (s < raid5_size(mddev, 0, 0)) {
4564
				skipped_disk = 1;
4565 4566 4567 4568 4569 4570
				continue;
			}
			memset(page_address(sh->dev[j].page), 0, STRIPE_SIZE);
			set_bit(R5_Expanded, &sh->dev[j].flags);
			set_bit(R5_UPTODATE, &sh->dev[j].flags);
		}
4571
		if (!skipped_disk) {
4572 4573 4574
			set_bit(STRIPE_EXPAND_READY, &sh->state);
			set_bit(STRIPE_HANDLE, &sh->state);
		}
4575
		list_add(&sh->lru, &stripes);
4576 4577
	}
	spin_lock_irq(&conf->device_lock);
4578
	if (mddev->reshape_backwards)
4579
		conf->reshape_progress -= reshape_sectors * new_data_disks;
4580
	else
4581
		conf->reshape_progress += reshape_sectors * new_data_disks;
4582 4583 4584 4585 4586 4587 4588
	spin_unlock_irq(&conf->device_lock);
	/* Ok, those stripe are ready. We can start scheduling
	 * reads on the source stripes.
	 * The source stripes are determined by mapping the first and last
	 * block on the destination stripes.
	 */
	first_sector =
4589
		raid5_compute_sector(conf, stripe_addr*(new_data_disks),
4590
				     1, &dd_idx, NULL);
4591
	last_sector =
4592
		raid5_compute_sector(conf, ((stripe_addr+reshape_sectors)
4593
					    * new_data_disks - 1),
4594
				     1, &dd_idx, NULL);
A
Andre Noll 已提交
4595 4596
	if (last_sector >= mddev->dev_sectors)
		last_sector = mddev->dev_sectors - 1;
4597
	while (first_sector <= last_sector) {
4598
		sh = get_active_stripe(conf, first_sector, 1, 0, 1);
4599 4600 4601 4602 4603
		set_bit(STRIPE_EXPAND_SOURCE, &sh->state);
		set_bit(STRIPE_HANDLE, &sh->state);
		release_stripe(sh);
		first_sector += STRIPE_SECTORS;
	}
4604 4605 4606 4607 4608 4609 4610 4611
	/* Now that the sources are clearly marked, we can release
	 * the destination stripes
	 */
	while (!list_empty(&stripes)) {
		sh = list_entry(stripes.next, struct stripe_head, lru);
		list_del_init(&sh->lru);
		release_stripe(sh);
	}
4612 4613 4614
	/* If this takes us to the resync_max point where we have to pause,
	 * then we need to write out the superblock.
	 */
4615
	sector_nr += reshape_sectors;
4616 4617
	if ((sector_nr - mddev->curr_resync_completed) * 2
	    >= mddev->resync_max - mddev->curr_resync_completed) {
4618 4619 4620
		/* Cannot proceed until we've updated the superblock... */
		wait_event(conf->wait_for_overlap,
			   atomic_read(&conf->reshape_stripes) == 0);
4621
		mddev->reshape_position = conf->reshape_progress;
4622
		mddev->curr_resync_completed = sector_nr;
4623
		conf->reshape_checkpoint = jiffies;
4624 4625 4626 4627 4628 4629
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
		md_wakeup_thread(mddev->thread);
		wait_event(mddev->sb_wait,
			   !test_bit(MD_CHANGE_DEVS, &mddev->flags)
			   || kthread_should_stop());
		spin_lock_irq(&conf->device_lock);
4630
		conf->reshape_safe = mddev->reshape_position;
4631 4632
		spin_unlock_irq(&conf->device_lock);
		wake_up(&conf->wait_for_overlap);
4633
		sysfs_notify(&mddev->kobj, NULL, "sync_completed");
4634
	}
4635
	return reshape_sectors;
4636 4637 4638
}

/* FIXME go_faster isn't used */
4639
static inline sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipped, int go_faster)
4640
{
4641
	struct r5conf *conf = mddev->private;
4642
	struct stripe_head *sh;
A
Andre Noll 已提交
4643
	sector_t max_sector = mddev->dev_sectors;
N
NeilBrown 已提交
4644
	sector_t sync_blocks;
4645 4646
	int still_degraded = 0;
	int i;
L
Linus Torvalds 已提交
4647

4648
	if (sector_nr >= max_sector) {
L
Linus Torvalds 已提交
4649
		/* just being told to finish up .. nothing much to do */
4650

4651 4652 4653 4654
		if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) {
			end_reshape(conf);
			return 0;
		}
4655 4656 4657 4658

		if (mddev->curr_resync < max_sector) /* aborted */
			bitmap_end_sync(mddev->bitmap, mddev->curr_resync,
					&sync_blocks, 1);
4659
		else /* completed sync */
4660 4661 4662
			conf->fullsync = 0;
		bitmap_close_sync(mddev->bitmap);

L
Linus Torvalds 已提交
4663 4664
		return 0;
	}
4665

4666 4667 4668
	/* Allow raid5_quiesce to complete */
	wait_event(conf->wait_for_overlap, conf->quiesce != 2);

4669 4670
	if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
		return reshape_request(mddev, sector_nr, skipped);
4671

4672 4673 4674 4675 4676 4677
	/* No need to check resync_max as we never do more than one
	 * stripe, and as resync_max will always be on a chunk boundary,
	 * if the check in md_do_sync didn't fire, there is no chance
	 * of overstepping resync_max here
	 */

4678
	/* if there is too many failed drives and we are trying
L
Linus Torvalds 已提交
4679 4680 4681
	 * to resync, then assert that we are finished, because there is
	 * nothing we can do.
	 */
4682
	if (mddev->degraded >= conf->max_degraded &&
4683
	    test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
A
Andre Noll 已提交
4684
		sector_t rv = mddev->dev_sectors - sector_nr;
4685
		*skipped = 1;
L
Linus Torvalds 已提交
4686 4687
		return rv;
	}
4688 4689 4690 4691
	if (!test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
	    !conf->fullsync &&
	    !bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, 1) &&
	    sync_blocks >= STRIPE_SECTORS) {
4692 4693 4694 4695 4696
		/* we can skip this block, and probably more */
		sync_blocks /= STRIPE_SECTORS;
		*skipped = 1;
		return sync_blocks * STRIPE_SECTORS; /* keep things rounded to whole stripes */
	}
L
Linus Torvalds 已提交
4697

N
NeilBrown 已提交
4698 4699
	bitmap_cond_end_sync(mddev->bitmap, sector_nr);

4700
	sh = get_active_stripe(conf, sector_nr, 0, 1, 0);
L
Linus Torvalds 已提交
4701
	if (sh == NULL) {
4702
		sh = get_active_stripe(conf, sector_nr, 0, 0, 0);
L
Linus Torvalds 已提交
4703
		/* make sure we don't swamp the stripe cache if someone else
4704
		 * is trying to get access
L
Linus Torvalds 已提交
4705
		 */
4706
		schedule_timeout_uninterruptible(1);
L
Linus Torvalds 已提交
4707
	}
4708 4709 4710 4711
	/* Need to check if array will still be degraded after recovery/resync
	 * We don't need to check the 'failed' flag as when that gets set,
	 * recovery aborts.
	 */
4712
	for (i = 0; i < conf->raid_disks; i++)
4713 4714 4715 4716 4717
		if (conf->disks[i].rdev == NULL)
			still_degraded = 1;

	bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, still_degraded);

4718
	set_bit(STRIPE_SYNC_REQUESTED, &sh->state);
L
Linus Torvalds 已提交
4719

4720
	handle_stripe(sh);
L
Linus Torvalds 已提交
4721 4722 4723 4724 4725
	release_stripe(sh);

	return STRIPE_SECTORS;
}

4726
static int  retry_aligned_read(struct r5conf *conf, struct bio *raid_bio)
4727 4728 4729 4730 4731 4732 4733 4734 4735 4736 4737 4738
{
	/* We may not be able to submit a whole bio at once as there
	 * may not be enough stripe_heads available.
	 * We cannot pre-allocate enough stripe_heads as we may need
	 * more than exist in the cache (if we allow ever large chunks).
	 * So we do one stripe head at a time and record in
	 * ->bi_hw_segments how many have been done.
	 *
	 * We *know* that this entire raid_bio is in one chunk, so
	 * it will be only one 'dd_idx' and only need one call to raid5_compute_sector.
	 */
	struct stripe_head *sh;
4739
	int dd_idx;
4740 4741 4742 4743 4744 4745
	sector_t sector, logical_sector, last_sector;
	int scnt = 0;
	int remaining;
	int handled = 0;

	logical_sector = raid_bio->bi_sector & ~((sector_t)STRIPE_SECTORS-1);
4746
	sector = raid5_compute_sector(conf, logical_sector,
4747
				      0, &dd_idx, NULL);
K
Kent Overstreet 已提交
4748
	last_sector = bio_end_sector(raid_bio);
4749 4750

	for (; logical_sector < last_sector;
4751 4752 4753
	     logical_sector += STRIPE_SECTORS,
		     sector += STRIPE_SECTORS,
		     scnt++) {
4754

4755
		if (scnt < raid5_bi_processed_stripes(raid_bio))
4756 4757 4758
			/* already done this stripe */
			continue;

4759
		sh = get_active_stripe(conf, sector, 0, 1, 0);
4760 4761 4762

		if (!sh) {
			/* failed to get a stripe - must wait */
4763
			raid5_set_bi_processed_stripes(raid_bio, scnt);
4764 4765 4766 4767
			conf->retry_read_aligned = raid_bio;
			return handled;
		}

4768 4769
		if (!add_stripe_bio(sh, raid_bio, dd_idx, 0)) {
			release_stripe(sh);
4770
			raid5_set_bi_processed_stripes(raid_bio, scnt);
4771 4772 4773 4774
			conf->retry_read_aligned = raid_bio;
			return handled;
		}

4775
		set_bit(R5_ReadNoMerge, &sh->dev[dd_idx].flags);
4776
		handle_stripe(sh);
4777 4778 4779
		release_stripe(sh);
		handled++;
	}
4780
	remaining = raid5_dec_bi_active_stripes(raid_bio);
4781 4782 4783
	if (remaining == 0) {
		trace_block_bio_complete(bdev_get_queue(raid_bio->bi_bdev),
					 raid_bio, 0);
4784
		bio_endio(raid_bio, 0);
4785
	}
4786 4787 4788 4789 4790
	if (atomic_dec_and_test(&conf->active_aligned_reads))
		wake_up(&conf->wait_for_stripe);
	return handled;
}

4791 4792 4793 4794 4795 4796 4797 4798 4799 4800 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810 4811 4812 4813 4814
#define MAX_STRIPE_BATCH 8
static int handle_active_stripes(struct r5conf *conf)
{
	struct stripe_head *batch[MAX_STRIPE_BATCH], *sh;
	int i, batch_size = 0;

	while (batch_size < MAX_STRIPE_BATCH &&
			(sh = __get_priority_stripe(conf)) != NULL)
		batch[batch_size++] = sh;

	if (batch_size == 0)
		return batch_size;
	spin_unlock_irq(&conf->device_lock);

	for (i = 0; i < batch_size; i++)
		handle_stripe(batch[i]);

	cond_resched();

	spin_lock_irq(&conf->device_lock);
	for (i = 0; i < batch_size; i++)
		__release_stripe(conf, batch[i]);
	return batch_size;
}
4815

L
Linus Torvalds 已提交
4816 4817 4818 4819 4820 4821 4822
/*
 * This is our raid5 kernel thread.
 *
 * We scan the hash table for stripes which can be handled now.
 * During the scan, completed stripes are saved for us by the interrupt
 * handler, so that they will not have to wait for our next wakeup.
 */
S
Shaohua Li 已提交
4823
static void raid5d(struct md_thread *thread)
L
Linus Torvalds 已提交
4824
{
S
Shaohua Li 已提交
4825
	struct mddev *mddev = thread->mddev;
4826
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
4827
	int handled;
4828
	struct blk_plug plug;
L
Linus Torvalds 已提交
4829

4830
	pr_debug("+++ raid5d active\n");
L
Linus Torvalds 已提交
4831 4832 4833

	md_check_recovery(mddev);

4834
	blk_start_plug(&plug);
L
Linus Torvalds 已提交
4835 4836 4837
	handled = 0;
	spin_lock_irq(&conf->device_lock);
	while (1) {
4838
		struct bio *bio;
4839
		int batch_size;
L
Linus Torvalds 已提交
4840

4841
		if (
4842 4843 4844
		    !list_empty(&conf->bitmap_list)) {
			/* Now is a good time to flush some bitmap updates */
			conf->seq_flush++;
4845
			spin_unlock_irq(&conf->device_lock);
4846
			bitmap_unplug(mddev->bitmap);
4847
			spin_lock_irq(&conf->device_lock);
4848
			conf->seq_write = conf->seq_flush;
4849 4850
			activate_bit_delay(conf);
		}
4851
		raid5_activate_delayed(conf);
4852

4853 4854 4855 4856 4857 4858 4859 4860 4861 4862
		while ((bio = remove_bio_from_retry(conf))) {
			int ok;
			spin_unlock_irq(&conf->device_lock);
			ok = retry_aligned_read(conf, bio);
			spin_lock_irq(&conf->device_lock);
			if (!ok)
				break;
			handled++;
		}

4863 4864
		batch_size = handle_active_stripes(conf);
		if (!batch_size)
L
Linus Torvalds 已提交
4865
			break;
4866
		handled += batch_size;
L
Linus Torvalds 已提交
4867

4868 4869
		if (mddev->flags & ~(1<<MD_CHANGE_PENDING)) {
			spin_unlock_irq(&conf->device_lock);
4870
			md_check_recovery(mddev);
4871 4872
			spin_lock_irq(&conf->device_lock);
		}
L
Linus Torvalds 已提交
4873
	}
4874
	pr_debug("%d stripes handled\n", handled);
L
Linus Torvalds 已提交
4875 4876 4877

	spin_unlock_irq(&conf->device_lock);

4878
	async_tx_issue_pending_all();
4879
	blk_finish_plug(&plug);
L
Linus Torvalds 已提交
4880

4881
	pr_debug("--- raid5d inactive\n");
L
Linus Torvalds 已提交
4882 4883
}

4884
static ssize_t
4885
raid5_show_stripe_cache_size(struct mddev *mddev, char *page)
4886
{
4887
	struct r5conf *conf = mddev->private;
4888 4889 4890 4891
	if (conf)
		return sprintf(page, "%d\n", conf->max_nr_stripes);
	else
		return 0;
4892 4893
}

4894
int
4895
raid5_set_cache_size(struct mddev *mddev, int size)
4896
{
4897
	struct r5conf *conf = mddev->private;
4898 4899
	int err;

4900
	if (size <= 16 || size > 32768)
4901
		return -EINVAL;
4902
	while (size < conf->max_nr_stripes) {
4903 4904 4905 4906 4907
		if (drop_one_stripe(conf))
			conf->max_nr_stripes--;
		else
			break;
	}
4908 4909 4910
	err = md_allow_write(mddev);
	if (err)
		return err;
4911
	while (size > conf->max_nr_stripes) {
4912 4913 4914 4915
		if (grow_one_stripe(conf))
			conf->max_nr_stripes++;
		else break;
	}
4916 4917 4918 4919 4920
	return 0;
}
EXPORT_SYMBOL(raid5_set_cache_size);

static ssize_t
4921
raid5_store_stripe_cache_size(struct mddev *mddev, const char *page, size_t len)
4922
{
4923
	struct r5conf *conf = mddev->private;
4924 4925 4926 4927 4928 4929 4930 4931
	unsigned long new;
	int err;

	if (len >= PAGE_SIZE)
		return -EINVAL;
	if (!conf)
		return -ENODEV;

4932
	if (kstrtoul(page, 10, &new))
4933 4934 4935 4936
		return -EINVAL;
	err = raid5_set_cache_size(mddev, new);
	if (err)
		return err;
4937 4938
	return len;
}
4939

4940 4941 4942 4943
static struct md_sysfs_entry
raid5_stripecache_size = __ATTR(stripe_cache_size, S_IRUGO | S_IWUSR,
				raid5_show_stripe_cache_size,
				raid5_store_stripe_cache_size);
4944

4945
static ssize_t
4946
raid5_show_preread_threshold(struct mddev *mddev, char *page)
4947
{
4948
	struct r5conf *conf = mddev->private;
4949 4950 4951 4952 4953 4954 4955
	if (conf)
		return sprintf(page, "%d\n", conf->bypass_threshold);
	else
		return 0;
}

static ssize_t
4956
raid5_store_preread_threshold(struct mddev *mddev, const char *page, size_t len)
4957
{
4958
	struct r5conf *conf = mddev->private;
4959
	unsigned long new;
4960 4961 4962 4963 4964
	if (len >= PAGE_SIZE)
		return -EINVAL;
	if (!conf)
		return -ENODEV;

4965
	if (kstrtoul(page, 10, &new))
4966
		return -EINVAL;
4967
	if (new > conf->max_nr_stripes)
4968 4969 4970 4971 4972 4973 4974 4975 4976 4977 4978
		return -EINVAL;
	conf->bypass_threshold = new;
	return len;
}

static struct md_sysfs_entry
raid5_preread_bypass_threshold = __ATTR(preread_bypass_threshold,
					S_IRUGO | S_IWUSR,
					raid5_show_preread_threshold,
					raid5_store_preread_threshold);

4979
static ssize_t
4980
stripe_cache_active_show(struct mddev *mddev, char *page)
4981
{
4982
	struct r5conf *conf = mddev->private;
4983 4984 4985 4986
	if (conf)
		return sprintf(page, "%d\n", atomic_read(&conf->active_stripes));
	else
		return 0;
4987 4988
}

4989 4990
static struct md_sysfs_entry
raid5_stripecache_active = __ATTR_RO(stripe_cache_active);
4991

4992
static struct attribute *raid5_attrs[] =  {
4993 4994
	&raid5_stripecache_size.attr,
	&raid5_stripecache_active.attr,
4995
	&raid5_preread_bypass_threshold.attr,
4996 4997
	NULL,
};
4998 4999 5000
static struct attribute_group raid5_attrs_group = {
	.name = NULL,
	.attrs = raid5_attrs,
5001 5002
};

5003
static sector_t
5004
raid5_size(struct mddev *mddev, sector_t sectors, int raid_disks)
5005
{
5006
	struct r5conf *conf = mddev->private;
5007 5008 5009

	if (!sectors)
		sectors = mddev->dev_sectors;
5010
	if (!raid_disks)
5011
		/* size is defined by the smallest of previous and new size */
5012
		raid_disks = min(conf->raid_disks, conf->previous_raid_disks);
5013

5014
	sectors &= ~((sector_t)mddev->chunk_sectors - 1);
5015
	sectors &= ~((sector_t)mddev->new_chunk_sectors - 1);
5016 5017 5018
	return sectors * (raid_disks - conf->max_degraded);
}

5019
static void raid5_free_percpu(struct r5conf *conf)
5020 5021 5022 5023 5024 5025 5026 5027 5028 5029 5030
{
	struct raid5_percpu *percpu;
	unsigned long cpu;

	if (!conf->percpu)
		return;

	get_online_cpus();
	for_each_possible_cpu(cpu) {
		percpu = per_cpu_ptr(conf->percpu, cpu);
		safe_put_page(percpu->spare_page);
5031
		kfree(percpu->scribble);
5032 5033 5034 5035 5036 5037 5038 5039 5040
	}
#ifdef CONFIG_HOTPLUG_CPU
	unregister_cpu_notifier(&conf->cpu_notify);
#endif
	put_online_cpus();

	free_percpu(conf->percpu);
}

5041
static void free_conf(struct r5conf *conf)
5042 5043
{
	shrink_stripes(conf);
5044
	raid5_free_percpu(conf);
5045 5046 5047 5048 5049
	kfree(conf->disks);
	kfree(conf->stripe_hashtbl);
	kfree(conf);
}

5050 5051 5052 5053
#ifdef CONFIG_HOTPLUG_CPU
static int raid456_cpu_notify(struct notifier_block *nfb, unsigned long action,
			      void *hcpu)
{
5054
	struct r5conf *conf = container_of(nfb, struct r5conf, cpu_notify);
5055 5056 5057 5058 5059 5060
	long cpu = (long)hcpu;
	struct raid5_percpu *percpu = per_cpu_ptr(conf->percpu, cpu);

	switch (action) {
	case CPU_UP_PREPARE:
	case CPU_UP_PREPARE_FROZEN:
5061
		if (conf->level == 6 && !percpu->spare_page)
5062
			percpu->spare_page = alloc_page(GFP_KERNEL);
5063 5064 5065 5066 5067 5068 5069
		if (!percpu->scribble)
			percpu->scribble = kmalloc(conf->scribble_len, GFP_KERNEL);

		if (!percpu->scribble ||
		    (conf->level == 6 && !percpu->spare_page)) {
			safe_put_page(percpu->spare_page);
			kfree(percpu->scribble);
5070 5071
			pr_err("%s: failed memory allocation for cpu%ld\n",
			       __func__, cpu);
5072
			return notifier_from_errno(-ENOMEM);
5073 5074 5075 5076 5077
		}
		break;
	case CPU_DEAD:
	case CPU_DEAD_FROZEN:
		safe_put_page(percpu->spare_page);
5078
		kfree(percpu->scribble);
5079
		percpu->spare_page = NULL;
5080
		percpu->scribble = NULL;
5081 5082 5083 5084 5085 5086 5087 5088
		break;
	default:
		break;
	}
	return NOTIFY_OK;
}
#endif

5089
static int raid5_alloc_percpu(struct r5conf *conf)
5090 5091 5092
{
	unsigned long cpu;
	struct page *spare_page;
5093
	struct raid5_percpu __percpu *allcpus;
5094
	void *scribble;
5095 5096 5097 5098 5099 5100 5101 5102 5103 5104
	int err;

	allcpus = alloc_percpu(struct raid5_percpu);
	if (!allcpus)
		return -ENOMEM;
	conf->percpu = allcpus;

	get_online_cpus();
	err = 0;
	for_each_present_cpu(cpu) {
5105 5106 5107 5108 5109 5110 5111 5112
		if (conf->level == 6) {
			spare_page = alloc_page(GFP_KERNEL);
			if (!spare_page) {
				err = -ENOMEM;
				break;
			}
			per_cpu_ptr(conf->percpu, cpu)->spare_page = spare_page;
		}
5113
		scribble = kmalloc(conf->scribble_len, GFP_KERNEL);
5114
		if (!scribble) {
5115 5116 5117
			err = -ENOMEM;
			break;
		}
5118
		per_cpu_ptr(conf->percpu, cpu)->scribble = scribble;
5119 5120 5121 5122 5123 5124 5125 5126 5127 5128 5129 5130
	}
#ifdef CONFIG_HOTPLUG_CPU
	conf->cpu_notify.notifier_call = raid456_cpu_notify;
	conf->cpu_notify.priority = 0;
	if (err == 0)
		err = register_cpu_notifier(&conf->cpu_notify);
#endif
	put_online_cpus();

	return err;
}

5131
static struct r5conf *setup_conf(struct mddev *mddev)
L
Linus Torvalds 已提交
5132
{
5133
	struct r5conf *conf;
5134
	int raid_disk, memory, max_disks;
5135
	struct md_rdev *rdev;
L
Linus Torvalds 已提交
5136
	struct disk_info *disk;
5137
	char pers_name[6];
L
Linus Torvalds 已提交
5138

N
NeilBrown 已提交
5139 5140 5141
	if (mddev->new_level != 5
	    && mddev->new_level != 4
	    && mddev->new_level != 6) {
5142
		printk(KERN_ERR "md/raid:%s: raid level not set to 4/5/6 (%d)\n",
N
NeilBrown 已提交
5143 5144
		       mdname(mddev), mddev->new_level);
		return ERR_PTR(-EIO);
L
Linus Torvalds 已提交
5145
	}
N
NeilBrown 已提交
5146 5147 5148 5149
	if ((mddev->new_level == 5
	     && !algorithm_valid_raid5(mddev->new_layout)) ||
	    (mddev->new_level == 6
	     && !algorithm_valid_raid6(mddev->new_layout))) {
5150
		printk(KERN_ERR "md/raid:%s: layout %d not supported\n",
N
NeilBrown 已提交
5151 5152
		       mdname(mddev), mddev->new_layout);
		return ERR_PTR(-EIO);
5153
	}
N
NeilBrown 已提交
5154
	if (mddev->new_level == 6 && mddev->raid_disks < 4) {
5155
		printk(KERN_ERR "md/raid:%s: not enough configured devices (%d, minimum 4)\n",
N
NeilBrown 已提交
5156 5157
		       mdname(mddev), mddev->raid_disks);
		return ERR_PTR(-EINVAL);
5158 5159
	}

5160 5161 5162
	if (!mddev->new_chunk_sectors ||
	    (mddev->new_chunk_sectors << 9) % PAGE_SIZE ||
	    !is_power_of_2(mddev->new_chunk_sectors)) {
5163 5164
		printk(KERN_ERR "md/raid:%s: invalid chunk size %d\n",
		       mdname(mddev), mddev->new_chunk_sectors << 9);
N
NeilBrown 已提交
5165
		return ERR_PTR(-EINVAL);
5166 5167
	}

5168
	conf = kzalloc(sizeof(struct r5conf), GFP_KERNEL);
N
NeilBrown 已提交
5169
	if (conf == NULL)
L
Linus Torvalds 已提交
5170
		goto abort;
5171 5172 5173 5174 5175 5176 5177 5178 5179 5180 5181 5182
	spin_lock_init(&conf->device_lock);
	init_waitqueue_head(&conf->wait_for_stripe);
	init_waitqueue_head(&conf->wait_for_overlap);
	INIT_LIST_HEAD(&conf->handle_list);
	INIT_LIST_HEAD(&conf->hold_list);
	INIT_LIST_HEAD(&conf->delayed_list);
	INIT_LIST_HEAD(&conf->bitmap_list);
	INIT_LIST_HEAD(&conf->inactive_list);
	atomic_set(&conf->active_stripes, 0);
	atomic_set(&conf->preread_active_stripes, 0);
	atomic_set(&conf->active_aligned_reads, 0);
	conf->bypass_threshold = BYPASS_THRESHOLD;
5183
	conf->recovery_disabled = mddev->recovery_disabled - 1;
N
NeilBrown 已提交
5184 5185 5186 5187 5188

	conf->raid_disks = mddev->raid_disks;
	if (mddev->reshape_position == MaxSector)
		conf->previous_raid_disks = mddev->raid_disks;
	else
5189
		conf->previous_raid_disks = mddev->raid_disks - mddev->delta_disks;
5190 5191
	max_disks = max(conf->raid_disks, conf->previous_raid_disks);
	conf->scribble_len = scribble_len(max_disks);
5192

5193
	conf->disks = kzalloc(max_disks * sizeof(struct disk_info),
5194 5195 5196
			      GFP_KERNEL);
	if (!conf->disks)
		goto abort;
5197

L
Linus Torvalds 已提交
5198 5199
	conf->mddev = mddev;

5200
	if ((conf->stripe_hashtbl = kzalloc(PAGE_SIZE, GFP_KERNEL)) == NULL)
L
Linus Torvalds 已提交
5201 5202
		goto abort;

5203 5204 5205 5206
	conf->level = mddev->new_level;
	if (raid5_alloc_percpu(conf) != 0)
		goto abort;

5207
	pr_debug("raid456: run(%s) called.\n", mdname(mddev));
L
Linus Torvalds 已提交
5208

N
NeilBrown 已提交
5209
	rdev_for_each(rdev, mddev) {
L
Linus Torvalds 已提交
5210
		raid_disk = rdev->raid_disk;
5211
		if (raid_disk >= max_disks
L
Linus Torvalds 已提交
5212 5213 5214 5215
		    || raid_disk < 0)
			continue;
		disk = conf->disks + raid_disk;

5216 5217 5218 5219 5220 5221 5222 5223 5224
		if (test_bit(Replacement, &rdev->flags)) {
			if (disk->replacement)
				goto abort;
			disk->replacement = rdev;
		} else {
			if (disk->rdev)
				goto abort;
			disk->rdev = rdev;
		}
L
Linus Torvalds 已提交
5225

5226
		if (test_bit(In_sync, &rdev->flags)) {
L
Linus Torvalds 已提交
5227
			char b[BDEVNAME_SIZE];
5228 5229 5230
			printk(KERN_INFO "md/raid:%s: device %s operational as raid"
			       " disk %d\n",
			       mdname(mddev), bdevname(rdev->bdev, b), raid_disk);
J
Jonathan Brassow 已提交
5231
		} else if (rdev->saved_raid_disk != raid_disk)
5232 5233
			/* Cannot rely on bitmap to complete recovery */
			conf->fullsync = 1;
L
Linus Torvalds 已提交
5234 5235
	}

5236
	conf->chunk_sectors = mddev->new_chunk_sectors;
N
NeilBrown 已提交
5237
	conf->level = mddev->new_level;
5238 5239 5240 5241
	if (conf->level == 6)
		conf->max_degraded = 2;
	else
		conf->max_degraded = 1;
N
NeilBrown 已提交
5242
	conf->algorithm = mddev->new_layout;
L
Linus Torvalds 已提交
5243
	conf->max_nr_stripes = NR_STRIPES;
5244
	conf->reshape_progress = mddev->reshape_position;
5245
	if (conf->reshape_progress != MaxSector) {
5246
		conf->prev_chunk_sectors = mddev->chunk_sectors;
5247 5248
		conf->prev_algo = mddev->layout;
	}
L
Linus Torvalds 已提交
5249

N
NeilBrown 已提交
5250
	memory = conf->max_nr_stripes * (sizeof(struct stripe_head) +
5251
		 max_disks * ((sizeof(struct bio) + PAGE_SIZE))) / 1024;
N
NeilBrown 已提交
5252 5253
	if (grow_stripes(conf, conf->max_nr_stripes)) {
		printk(KERN_ERR
5254 5255
		       "md/raid:%s: couldn't allocate %dkB for buffers\n",
		       mdname(mddev), memory);
N
NeilBrown 已提交
5256 5257
		goto abort;
	} else
5258 5259
		printk(KERN_INFO "md/raid:%s: allocated %dkB\n",
		       mdname(mddev), memory);
L
Linus Torvalds 已提交
5260

5261 5262
	sprintf(pers_name, "raid%d", mddev->new_level);
	conf->thread = md_register_thread(raid5d, mddev, pers_name);
N
NeilBrown 已提交
5263 5264
	if (!conf->thread) {
		printk(KERN_ERR
5265
		       "md/raid:%s: couldn't allocate thread.\n",
N
NeilBrown 已提交
5266
		       mdname(mddev));
5267 5268
		goto abort;
	}
N
NeilBrown 已提交
5269 5270 5271 5272 5273

	return conf;

 abort:
	if (conf) {
5274
		free_conf(conf);
N
NeilBrown 已提交
5275 5276 5277 5278 5279
		return ERR_PTR(-EIO);
	} else
		return ERR_PTR(-ENOMEM);
}

5280 5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291 5292 5293 5294 5295 5296 5297 5298 5299 5300 5301 5302 5303 5304 5305 5306

static int only_parity(int raid_disk, int algo, int raid_disks, int max_degraded)
{
	switch (algo) {
	case ALGORITHM_PARITY_0:
		if (raid_disk < max_degraded)
			return 1;
		break;
	case ALGORITHM_PARITY_N:
		if (raid_disk >= raid_disks - max_degraded)
			return 1;
		break;
	case ALGORITHM_PARITY_0_6:
		if (raid_disk == 0 || 
		    raid_disk == raid_disks - 1)
			return 1;
		break;
	case ALGORITHM_LEFT_ASYMMETRIC_6:
	case ALGORITHM_RIGHT_ASYMMETRIC_6:
	case ALGORITHM_LEFT_SYMMETRIC_6:
	case ALGORITHM_RIGHT_SYMMETRIC_6:
		if (raid_disk == raid_disks - 1)
			return 1;
	}
	return 0;
}

5307
static int run(struct mddev *mddev)
N
NeilBrown 已提交
5308
{
5309
	struct r5conf *conf;
5310
	int working_disks = 0;
5311
	int dirty_parity_disks = 0;
5312
	struct md_rdev *rdev;
5313
	sector_t reshape_offset = 0;
5314
	int i;
5315 5316
	long long min_offset_diff = 0;
	int first = 1;
N
NeilBrown 已提交
5317

5318
	if (mddev->recovery_cp != MaxSector)
5319
		printk(KERN_NOTICE "md/raid:%s: not clean"
5320 5321
		       " -- starting background reconstruction\n",
		       mdname(mddev));
5322 5323 5324 5325 5326 5327 5328 5329 5330 5331 5332 5333 5334 5335 5336 5337 5338

	rdev_for_each(rdev, mddev) {
		long long diff;
		if (rdev->raid_disk < 0)
			continue;
		diff = (rdev->new_data_offset - rdev->data_offset);
		if (first) {
			min_offset_diff = diff;
			first = 0;
		} else if (mddev->reshape_backwards &&
			 diff < min_offset_diff)
			min_offset_diff = diff;
		else if (!mddev->reshape_backwards &&
			 diff > min_offset_diff)
			min_offset_diff = diff;
	}

N
NeilBrown 已提交
5339 5340
	if (mddev->reshape_position != MaxSector) {
		/* Check that we can continue the reshape.
5341 5342 5343 5344 5345 5346 5347 5348 5349 5350
		 * Difficulties arise if the stripe we would write to
		 * next is at or after the stripe we would read from next.
		 * For a reshape that changes the number of devices, this
		 * is only possible for a very short time, and mdadm makes
		 * sure that time appears to have past before assembling
		 * the array.  So we fail if that time hasn't passed.
		 * For a reshape that keeps the number of devices the same
		 * mdadm must be monitoring the reshape can keeping the
		 * critical areas read-only and backed up.  It will start
		 * the array in read-only mode, so we check for that.
N
NeilBrown 已提交
5351 5352 5353
		 */
		sector_t here_new, here_old;
		int old_disks;
5354
		int max_degraded = (mddev->level == 6 ? 2 : 1);
N
NeilBrown 已提交
5355

5356
		if (mddev->new_level != mddev->level) {
5357
			printk(KERN_ERR "md/raid:%s: unsupported reshape "
N
NeilBrown 已提交
5358 5359 5360 5361 5362 5363 5364 5365 5366 5367
			       "required - aborting.\n",
			       mdname(mddev));
			return -EINVAL;
		}
		old_disks = mddev->raid_disks - mddev->delta_disks;
		/* reshape_position must be on a new-stripe boundary, and one
		 * further up in new geometry must map after here in old
		 * geometry.
		 */
		here_new = mddev->reshape_position;
5368
		if (sector_div(here_new, mddev->new_chunk_sectors *
N
NeilBrown 已提交
5369
			       (mddev->raid_disks - max_degraded))) {
5370 5371
			printk(KERN_ERR "md/raid:%s: reshape_position not "
			       "on a stripe boundary\n", mdname(mddev));
N
NeilBrown 已提交
5372 5373
			return -EINVAL;
		}
5374
		reshape_offset = here_new * mddev->new_chunk_sectors;
N
NeilBrown 已提交
5375 5376
		/* here_new is the stripe we will write to */
		here_old = mddev->reshape_position;
5377
		sector_div(here_old, mddev->chunk_sectors *
N
NeilBrown 已提交
5378 5379 5380
			   (old_disks-max_degraded));
		/* here_old is the first stripe that we might need to read
		 * from */
5381
		if (mddev->delta_disks == 0) {
5382 5383 5384 5385 5386 5387
			if ((here_new * mddev->new_chunk_sectors !=
			     here_old * mddev->chunk_sectors)) {
				printk(KERN_ERR "md/raid:%s: reshape position is"
				       " confused - aborting\n", mdname(mddev));
				return -EINVAL;
			}
5388
			/* We cannot be sure it is safe to start an in-place
5389
			 * reshape.  It is only safe if user-space is monitoring
5390 5391 5392 5393 5394
			 * and taking constant backups.
			 * mdadm always starts a situation like this in
			 * readonly mode so it can take control before
			 * allowing any writes.  So just check for that.
			 */
5395 5396 5397 5398 5399 5400 5401
			if (abs(min_offset_diff) >= mddev->chunk_sectors &&
			    abs(min_offset_diff) >= mddev->new_chunk_sectors)
				/* not really in-place - so OK */;
			else if (mddev->ro == 0) {
				printk(KERN_ERR "md/raid:%s: in-place reshape "
				       "must be started in read-only mode "
				       "- aborting\n",
5402
				       mdname(mddev));
5403 5404
				return -EINVAL;
			}
5405
		} else if (mddev->reshape_backwards
5406
		    ? (here_new * mddev->new_chunk_sectors + min_offset_diff <=
5407 5408
		       here_old * mddev->chunk_sectors)
		    : (here_new * mddev->new_chunk_sectors >=
5409
		       here_old * mddev->chunk_sectors + (-min_offset_diff))) {
N
NeilBrown 已提交
5410
			/* Reading from the same stripe as writing to - bad */
5411 5412 5413
			printk(KERN_ERR "md/raid:%s: reshape_position too early for "
			       "auto-recovery - aborting.\n",
			       mdname(mddev));
N
NeilBrown 已提交
5414 5415
			return -EINVAL;
		}
5416 5417
		printk(KERN_INFO "md/raid:%s: reshape will continue\n",
		       mdname(mddev));
N
NeilBrown 已提交
5418 5419 5420 5421
		/* OK, we should be able to continue; */
	} else {
		BUG_ON(mddev->level != mddev->new_level);
		BUG_ON(mddev->layout != mddev->new_layout);
5422
		BUG_ON(mddev->chunk_sectors != mddev->new_chunk_sectors);
N
NeilBrown 已提交
5423
		BUG_ON(mddev->delta_disks != 0);
L
Linus Torvalds 已提交
5424
	}
N
NeilBrown 已提交
5425

5426 5427 5428 5429 5430
	if (mddev->private == NULL)
		conf = setup_conf(mddev);
	else
		conf = mddev->private;

N
NeilBrown 已提交
5431 5432 5433
	if (IS_ERR(conf))
		return PTR_ERR(conf);

5434
	conf->min_offset_diff = min_offset_diff;
N
NeilBrown 已提交
5435 5436 5437 5438
	mddev->thread = conf->thread;
	conf->thread = NULL;
	mddev->private = conf;

5439 5440 5441 5442 5443 5444 5445 5446 5447 5448 5449
	for (i = 0; i < conf->raid_disks && conf->previous_raid_disks;
	     i++) {
		rdev = conf->disks[i].rdev;
		if (!rdev && conf->disks[i].replacement) {
			/* The replacement is all we have yet */
			rdev = conf->disks[i].replacement;
			conf->disks[i].replacement = NULL;
			clear_bit(Replacement, &rdev->flags);
			conf->disks[i].rdev = rdev;
		}
		if (!rdev)
5450
			continue;
5451 5452 5453 5454 5455 5456 5457
		if (conf->disks[i].replacement &&
		    conf->reshape_progress != MaxSector) {
			/* replacements and reshape simply do not mix. */
			printk(KERN_ERR "md: cannot handle concurrent "
			       "replacement and reshape.\n");
			goto abort;
		}
5458
		if (test_bit(In_sync, &rdev->flags)) {
N
NeilBrown 已提交
5459
			working_disks++;
5460 5461
			continue;
		}
5462 5463 5464 5465 5466 5467 5468 5469 5470 5471 5472 5473
		/* This disc is not fully in-sync.  However if it
		 * just stored parity (beyond the recovery_offset),
		 * when we don't need to be concerned about the
		 * array being dirty.
		 * When reshape goes 'backwards', we never have
		 * partially completed devices, so we only need
		 * to worry about reshape going forwards.
		 */
		/* Hack because v0.91 doesn't store recovery_offset properly. */
		if (mddev->major_version == 0 &&
		    mddev->minor_version > 90)
			rdev->recovery_offset = reshape_offset;
5474

5475 5476 5477 5478 5479 5480 5481 5482 5483 5484 5485 5486 5487 5488 5489
		if (rdev->recovery_offset < reshape_offset) {
			/* We need to check old and new layout */
			if (!only_parity(rdev->raid_disk,
					 conf->algorithm,
					 conf->raid_disks,
					 conf->max_degraded))
				continue;
		}
		if (!only_parity(rdev->raid_disk,
				 conf->prev_algo,
				 conf->previous_raid_disks,
				 conf->max_degraded))
			continue;
		dirty_parity_disks++;
	}
N
NeilBrown 已提交
5490

5491 5492 5493
	/*
	 * 0 for a fully functional array, 1 or 2 for a degraded array.
	 */
5494
	mddev->degraded = calc_degraded(conf);
N
NeilBrown 已提交
5495

5496
	if (has_failed(conf)) {
5497
		printk(KERN_ERR "md/raid:%s: not enough operational devices"
L
Linus Torvalds 已提交
5498
			" (%d/%d failed)\n",
5499
			mdname(mddev), mddev->degraded, conf->raid_disks);
L
Linus Torvalds 已提交
5500 5501 5502
		goto abort;
	}

N
NeilBrown 已提交
5503
	/* device size must be a multiple of chunk size */
5504
	mddev->dev_sectors &= ~(mddev->chunk_sectors - 1);
N
NeilBrown 已提交
5505 5506
	mddev->resync_max_sectors = mddev->dev_sectors;

5507
	if (mddev->degraded > dirty_parity_disks &&
L
Linus Torvalds 已提交
5508
	    mddev->recovery_cp != MaxSector) {
5509 5510
		if (mddev->ok_start_degraded)
			printk(KERN_WARNING
5511 5512
			       "md/raid:%s: starting dirty degraded array"
			       " - data corruption possible.\n",
5513 5514 5515
			       mdname(mddev));
		else {
			printk(KERN_ERR
5516
			       "md/raid:%s: cannot start dirty degraded array.\n",
5517 5518 5519
			       mdname(mddev));
			goto abort;
		}
L
Linus Torvalds 已提交
5520 5521 5522
	}

	if (mddev->degraded == 0)
5523 5524
		printk(KERN_INFO "md/raid:%s: raid level %d active with %d out of %d"
		       " devices, algorithm %d\n", mdname(mddev), conf->level,
5525 5526
		       mddev->raid_disks-mddev->degraded, mddev->raid_disks,
		       mddev->new_layout);
L
Linus Torvalds 已提交
5527
	else
5528 5529 5530 5531 5532
		printk(KERN_ALERT "md/raid:%s: raid level %d active with %d"
		       " out of %d devices, algorithm %d\n",
		       mdname(mddev), conf->level,
		       mddev->raid_disks - mddev->degraded,
		       mddev->raid_disks, mddev->new_layout);
L
Linus Torvalds 已提交
5533 5534 5535

	print_raid5_conf(conf);

5536 5537
	if (conf->reshape_progress != MaxSector) {
		conf->reshape_safe = conf->reshape_progress;
5538 5539 5540 5541 5542 5543
		atomic_set(&conf->reshape_stripes, 0);
		clear_bit(MD_RECOVERY_SYNC, &mddev->recovery);
		clear_bit(MD_RECOVERY_CHECK, &mddev->recovery);
		set_bit(MD_RECOVERY_RESHAPE, &mddev->recovery);
		set_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
		mddev->sync_thread = md_register_thread(md_do_sync, mddev,
5544
							"reshape");
5545 5546
	}

L
Linus Torvalds 已提交
5547 5548

	/* Ok, everything is just fine now */
5549 5550
	if (mddev->to_remove == &raid5_attrs_group)
		mddev->to_remove = NULL;
N
NeilBrown 已提交
5551 5552
	else if (mddev->kobj.sd &&
	    sysfs_create_group(&mddev->kobj, &raid5_attrs_group))
5553
		printk(KERN_WARNING
5554
		       "raid5: failed to create sysfs attributes for %s\n",
5555
		       mdname(mddev));
5556
	md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
5557

5558
	if (mddev->queue) {
5559
		int chunk_size;
S
Shaohua Li 已提交
5560
		bool discard_supported = true;
5561 5562 5563 5564 5565 5566 5567 5568 5569
		/* read-ahead size must cover two whole stripes, which
		 * is 2 * (datadisks) * chunksize where 'n' is the
		 * number of raid devices
		 */
		int data_disks = conf->previous_raid_disks - conf->max_degraded;
		int stripe = data_disks *
			((mddev->chunk_sectors << 9) / PAGE_SIZE);
		if (mddev->queue->backing_dev_info.ra_pages < 2 * stripe)
			mddev->queue->backing_dev_info.ra_pages = 2 * stripe;
N
NeilBrown 已提交
5570

5571
		blk_queue_merge_bvec(mddev->queue, raid5_mergeable_bvec);
5572

N
NeilBrown 已提交
5573 5574
		mddev->queue->backing_dev_info.congested_data = mddev;
		mddev->queue->backing_dev_info.congested_fn = raid5_congested;
5575

5576 5577 5578 5579
		chunk_size = mddev->chunk_sectors << 9;
		blk_queue_io_min(mddev->queue, chunk_size);
		blk_queue_io_opt(mddev->queue, chunk_size *
				 (conf->raid_disks - conf->max_degraded));
S
Shaohua Li 已提交
5580 5581 5582 5583 5584
		/*
		 * We can only discard a whole stripe. It doesn't make sense to
		 * discard data disk but write parity disk
		 */
		stripe = stripe * PAGE_SIZE;
5585 5586 5587 5588
		/* Round up to power of 2, as discard handling
		 * currently assumes that */
		while ((stripe-1) & stripe)
			stripe = (stripe | (stripe-1)) + 1;
S
Shaohua Li 已提交
5589 5590 5591 5592 5593 5594 5595
		mddev->queue->limits.discard_alignment = stripe;
		mddev->queue->limits.discard_granularity = stripe;
		/*
		 * unaligned part of discard request will be ignored, so can't
		 * guarantee discard_zerors_data
		 */
		mddev->queue->limits.discard_zeroes_data = 0;
5596

5597 5598
		blk_queue_max_write_same_sectors(mddev->queue, 0);

5599
		rdev_for_each(rdev, mddev) {
5600 5601
			disk_stack_limits(mddev->gendisk, rdev->bdev,
					  rdev->data_offset << 9);
5602 5603
			disk_stack_limits(mddev->gendisk, rdev->bdev,
					  rdev->new_data_offset << 9);
S
Shaohua Li 已提交
5604 5605 5606 5607 5608 5609 5610 5611 5612 5613 5614 5615 5616 5617
			/*
			 * discard_zeroes_data is required, otherwise data
			 * could be lost. Consider a scenario: discard a stripe
			 * (the stripe could be inconsistent if
			 * discard_zeroes_data is 0); write one disk of the
			 * stripe (the stripe could be inconsistent again
			 * depending on which disks are used to calculate
			 * parity); the disk is broken; The stripe data of this
			 * disk is lost.
			 */
			if (!blk_queue_discard(bdev_get_queue(rdev->bdev)) ||
			    !bdev_get_queue(rdev->bdev)->
						limits.discard_zeroes_data)
				discard_supported = false;
5618
		}
S
Shaohua Li 已提交
5619 5620 5621 5622 5623 5624 5625 5626 5627

		if (discard_supported &&
		   mddev->queue->limits.max_discard_sectors >= stripe &&
		   mddev->queue->limits.discard_granularity >= stripe)
			queue_flag_set_unlocked(QUEUE_FLAG_DISCARD,
						mddev->queue);
		else
			queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD,
						mddev->queue);
5628
	}
5629

L
Linus Torvalds 已提交
5630 5631
	return 0;
abort:
5632
	md_unregister_thread(&mddev->thread);
N
NeilBrown 已提交
5633 5634
	print_raid5_conf(conf);
	free_conf(conf);
L
Linus Torvalds 已提交
5635
	mddev->private = NULL;
5636
	printk(KERN_ALERT "md/raid:%s: failed to run raid set.\n", mdname(mddev));
L
Linus Torvalds 已提交
5637 5638 5639
	return -EIO;
}

5640
static int stop(struct mddev *mddev)
L
Linus Torvalds 已提交
5641
{
5642
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5643

5644
	md_unregister_thread(&mddev->thread);
N
NeilBrown 已提交
5645 5646
	if (mddev->queue)
		mddev->queue->backing_dev_info.congested_fn = NULL;
5647
	free_conf(conf);
5648 5649
	mddev->private = NULL;
	mddev->to_remove = &raid5_attrs_group;
L
Linus Torvalds 已提交
5650 5651 5652
	return 0;
}

5653
static void status(struct seq_file *seq, struct mddev *mddev)
L
Linus Torvalds 已提交
5654
{
5655
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5656 5657
	int i;

5658 5659
	seq_printf(seq, " level %d, %dk chunk, algorithm %d", mddev->level,
		mddev->chunk_sectors / 2, mddev->layout);
5660
	seq_printf (seq, " [%d/%d] [", conf->raid_disks, conf->raid_disks - mddev->degraded);
L
Linus Torvalds 已提交
5661 5662 5663
	for (i = 0; i < conf->raid_disks; i++)
		seq_printf (seq, "%s",
			       conf->disks[i].rdev &&
5664
			       test_bit(In_sync, &conf->disks[i].rdev->flags) ? "U" : "_");
L
Linus Torvalds 已提交
5665 5666 5667
	seq_printf (seq, "]");
}

5668
static void print_raid5_conf (struct r5conf *conf)
L
Linus Torvalds 已提交
5669 5670 5671 5672
{
	int i;
	struct disk_info *tmp;

5673
	printk(KERN_DEBUG "RAID conf printout:\n");
L
Linus Torvalds 已提交
5674 5675 5676 5677
	if (!conf) {
		printk("(conf==NULL)\n");
		return;
	}
5678 5679 5680
	printk(KERN_DEBUG " --- level:%d rd:%d wd:%d\n", conf->level,
	       conf->raid_disks,
	       conf->raid_disks - conf->mddev->degraded);
L
Linus Torvalds 已提交
5681 5682 5683 5684 5685

	for (i = 0; i < conf->raid_disks; i++) {
		char b[BDEVNAME_SIZE];
		tmp = conf->disks + i;
		if (tmp->rdev)
5686 5687 5688
			printk(KERN_DEBUG " disk %d, o:%d, dev:%s\n",
			       i, !test_bit(Faulty, &tmp->rdev->flags),
			       bdevname(tmp->rdev->bdev, b));
L
Linus Torvalds 已提交
5689 5690 5691
	}
}

5692
static int raid5_spare_active(struct mddev *mddev)
L
Linus Torvalds 已提交
5693 5694
{
	int i;
5695
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5696
	struct disk_info *tmp;
5697 5698
	int count = 0;
	unsigned long flags;
L
Linus Torvalds 已提交
5699 5700 5701

	for (i = 0; i < conf->raid_disks; i++) {
		tmp = conf->disks + i;
5702 5703 5704 5705 5706 5707 5708 5709 5710 5711 5712 5713 5714 5715 5716 5717 5718 5719 5720
		if (tmp->replacement
		    && tmp->replacement->recovery_offset == MaxSector
		    && !test_bit(Faulty, &tmp->replacement->flags)
		    && !test_and_set_bit(In_sync, &tmp->replacement->flags)) {
			/* Replacement has just become active. */
			if (!tmp->rdev
			    || !test_and_clear_bit(In_sync, &tmp->rdev->flags))
				count++;
			if (tmp->rdev) {
				/* Replaced device not technically faulty,
				 * but we need to be sure it gets removed
				 * and never re-added.
				 */
				set_bit(Faulty, &tmp->rdev->flags);
				sysfs_notify_dirent_safe(
					tmp->rdev->sysfs_state);
			}
			sysfs_notify_dirent_safe(tmp->replacement->sysfs_state);
		} else if (tmp->rdev
5721
		    && tmp->rdev->recovery_offset == MaxSector
5722
		    && !test_bit(Faulty, &tmp->rdev->flags)
5723
		    && !test_and_set_bit(In_sync, &tmp->rdev->flags)) {
5724
			count++;
5725
			sysfs_notify_dirent_safe(tmp->rdev->sysfs_state);
L
Linus Torvalds 已提交
5726 5727
		}
	}
5728
	spin_lock_irqsave(&conf->device_lock, flags);
5729
	mddev->degraded = calc_degraded(conf);
5730
	spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
5731
	print_raid5_conf(conf);
5732
	return count;
L
Linus Torvalds 已提交
5733 5734
}

5735
static int raid5_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
5736
{
5737
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5738
	int err = 0;
5739
	int number = rdev->raid_disk;
5740
	struct md_rdev **rdevp;
L
Linus Torvalds 已提交
5741 5742 5743
	struct disk_info *p = conf->disks + number;

	print_raid5_conf(conf);
5744 5745 5746 5747 5748 5749 5750 5751 5752 5753 5754 5755 5756 5757 5758 5759 5760 5761 5762 5763 5764 5765
	if (rdev == p->rdev)
		rdevp = &p->rdev;
	else if (rdev == p->replacement)
		rdevp = &p->replacement;
	else
		return 0;

	if (number >= conf->raid_disks &&
	    conf->reshape_progress == MaxSector)
		clear_bit(In_sync, &rdev->flags);

	if (test_bit(In_sync, &rdev->flags) ||
	    atomic_read(&rdev->nr_pending)) {
		err = -EBUSY;
		goto abort;
	}
	/* Only remove non-faulty devices if recovery
	 * isn't possible.
	 */
	if (!test_bit(Faulty, &rdev->flags) &&
	    mddev->recovery_disabled != conf->recovery_disabled &&
	    !has_failed(conf) &&
5766
	    (!p->replacement || p->replacement == rdev) &&
5767 5768 5769 5770 5771 5772 5773 5774 5775 5776
	    number < conf->raid_disks) {
		err = -EBUSY;
		goto abort;
	}
	*rdevp = NULL;
	synchronize_rcu();
	if (atomic_read(&rdev->nr_pending)) {
		/* lost the race, try later */
		err = -EBUSY;
		*rdevp = rdev;
5777 5778 5779 5780 5781 5782 5783 5784 5785 5786 5787 5788 5789 5790
	} else if (p->replacement) {
		/* We must have just cleared 'rdev' */
		p->rdev = p->replacement;
		clear_bit(Replacement, &p->replacement->flags);
		smp_mb(); /* Make sure other CPUs may see both as identical
			   * but will never see neither - if they are careful
			   */
		p->replacement = NULL;
		clear_bit(WantReplacement, &rdev->flags);
	} else
		/* We might have just removed the Replacement as faulty-
		 * clear the bit just in case
		 */
		clear_bit(WantReplacement, &rdev->flags);
L
Linus Torvalds 已提交
5791 5792 5793 5794 5795 5796
abort:

	print_raid5_conf(conf);
	return err;
}

5797
static int raid5_add_disk(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
5798
{
5799
	struct r5conf *conf = mddev->private;
5800
	int err = -EEXIST;
L
Linus Torvalds 已提交
5801 5802
	int disk;
	struct disk_info *p;
5803 5804
	int first = 0;
	int last = conf->raid_disks - 1;
L
Linus Torvalds 已提交
5805

5806 5807 5808
	if (mddev->recovery_disabled == conf->recovery_disabled)
		return -EBUSY;

N
NeilBrown 已提交
5809
	if (rdev->saved_raid_disk < 0 && has_failed(conf))
L
Linus Torvalds 已提交
5810
		/* no point adding a device */
5811
		return -EINVAL;
L
Linus Torvalds 已提交
5812

5813 5814
	if (rdev->raid_disk >= 0)
		first = last = rdev->raid_disk;
L
Linus Torvalds 已提交
5815 5816

	/*
5817 5818
	 * find the disk ... but prefer rdev->saved_raid_disk
	 * if possible.
L
Linus Torvalds 已提交
5819
	 */
5820
	if (rdev->saved_raid_disk >= 0 &&
5821
	    rdev->saved_raid_disk >= first &&
5822
	    conf->disks[rdev->saved_raid_disk].rdev == NULL)
5823 5824 5825
		first = rdev->saved_raid_disk;

	for (disk = first; disk <= last; disk++) {
5826 5827
		p = conf->disks + disk;
		if (p->rdev == NULL) {
5828
			clear_bit(In_sync, &rdev->flags);
L
Linus Torvalds 已提交
5829
			rdev->raid_disk = disk;
5830
			err = 0;
5831 5832
			if (rdev->saved_raid_disk != disk)
				conf->fullsync = 1;
5833
			rcu_assign_pointer(p->rdev, rdev);
5834
			goto out;
L
Linus Torvalds 已提交
5835
		}
5836 5837 5838
	}
	for (disk = first; disk <= last; disk++) {
		p = conf->disks + disk;
5839 5840 5841 5842 5843 5844 5845 5846 5847 5848 5849
		if (test_bit(WantReplacement, &p->rdev->flags) &&
		    p->replacement == NULL) {
			clear_bit(In_sync, &rdev->flags);
			set_bit(Replacement, &rdev->flags);
			rdev->raid_disk = disk;
			err = 0;
			conf->fullsync = 1;
			rcu_assign_pointer(p->replacement, rdev);
			break;
		}
	}
5850
out:
L
Linus Torvalds 已提交
5851
	print_raid5_conf(conf);
5852
	return err;
L
Linus Torvalds 已提交
5853 5854
}

5855
static int raid5_resize(struct mddev *mddev, sector_t sectors)
L
Linus Torvalds 已提交
5856 5857 5858 5859 5860 5861 5862 5863
{
	/* no resync is happening, and there is enough space
	 * on all devices, so we can resize.
	 * We need to make sure resync covers any new space.
	 * If the array is shrinking we should possibly wait until
	 * any io in the removed space completes, but it hardly seems
	 * worth it.
	 */
5864
	sector_t newsize;
5865
	sectors &= ~((sector_t)mddev->chunk_sectors - 1);
5866 5867 5868
	newsize = raid5_size(mddev, sectors, mddev->raid_disks);
	if (mddev->external_size &&
	    mddev->array_sectors > newsize)
D
Dan Williams 已提交
5869
		return -EINVAL;
5870 5871 5872 5873 5874 5875
	if (mddev->bitmap) {
		int ret = bitmap_resize(mddev->bitmap, sectors, 0, 0);
		if (ret)
			return ret;
	}
	md_set_array_sectors(mddev, newsize);
5876
	set_capacity(mddev->gendisk, mddev->array_sectors);
5877
	revalidate_disk(mddev->gendisk);
5878 5879
	if (sectors > mddev->dev_sectors &&
	    mddev->recovery_cp > mddev->dev_sectors) {
A
Andre Noll 已提交
5880
		mddev->recovery_cp = mddev->dev_sectors;
L
Linus Torvalds 已提交
5881 5882
		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
	}
A
Andre Noll 已提交
5883
	mddev->dev_sectors = sectors;
5884
	mddev->resync_max_sectors = sectors;
L
Linus Torvalds 已提交
5885 5886 5887
	return 0;
}

5888
static int check_stripe_cache(struct mddev *mddev)
5889 5890 5891 5892 5893 5894 5895 5896 5897
{
	/* Can only proceed if there are plenty of stripe_heads.
	 * We need a minimum of one full stripe,, and for sensible progress
	 * it is best to have about 4 times that.
	 * If we require 4 times, then the default 256 4K stripe_heads will
	 * allow for chunk sizes up to 256K, which is probably OK.
	 * If the chunk size is greater, user-space should request more
	 * stripe_heads first.
	 */
5898
	struct r5conf *conf = mddev->private;
5899 5900 5901 5902
	if (((mddev->chunk_sectors << 9) / STRIPE_SIZE) * 4
	    > conf->max_nr_stripes ||
	    ((mddev->new_chunk_sectors << 9) / STRIPE_SIZE) * 4
	    > conf->max_nr_stripes) {
5903 5904
		printk(KERN_WARNING "md/raid:%s: reshape: not enough stripes.  Needed %lu\n",
		       mdname(mddev),
5905 5906 5907 5908 5909 5910 5911
		       ((max(mddev->chunk_sectors, mddev->new_chunk_sectors) << 9)
			/ STRIPE_SIZE)*4);
		return 0;
	}
	return 1;
}

5912
static int check_reshape(struct mddev *mddev)
5913
{
5914
	struct r5conf *conf = mddev->private;
5915

5916 5917
	if (mddev->delta_disks == 0 &&
	    mddev->new_layout == mddev->layout &&
5918
	    mddev->new_chunk_sectors == mddev->chunk_sectors)
5919
		return 0; /* nothing to do */
5920
	if (has_failed(conf))
5921
		return -EINVAL;
5922
	if (mddev->delta_disks < 0 && mddev->reshape_position == MaxSector) {
5923 5924 5925 5926 5927 5928 5929 5930 5931 5932 5933
		/* We might be able to shrink, but the devices must
		 * be made bigger first.
		 * For raid6, 4 is the minimum size.
		 * Otherwise 2 is the minimum
		 */
		int min = 2;
		if (mddev->level == 6)
			min = 4;
		if (mddev->raid_disks + mddev->delta_disks < min)
			return -EINVAL;
	}
5934

5935
	if (!check_stripe_cache(mddev))
5936 5937
		return -ENOSPC;

5938 5939
	return resize_stripes(conf, (conf->previous_raid_disks
				     + mddev->delta_disks));
5940 5941
}

5942
static int raid5_start_reshape(struct mddev *mddev)
5943
{
5944
	struct r5conf *conf = mddev->private;
5945
	struct md_rdev *rdev;
5946
	int spares = 0;
5947
	unsigned long flags;
5948

5949
	if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
5950 5951
		return -EBUSY;

5952 5953 5954
	if (!check_stripe_cache(mddev))
		return -ENOSPC;

5955 5956 5957
	if (has_failed(conf))
		return -EINVAL;

5958
	rdev_for_each(rdev, mddev) {
5959 5960
		if (!test_bit(In_sync, &rdev->flags)
		    && !test_bit(Faulty, &rdev->flags))
5961
			spares++;
5962
	}
5963

5964
	if (spares - mddev->degraded < mddev->delta_disks - conf->max_degraded)
5965 5966 5967 5968 5969
		/* Not enough devices even to make a degraded array
		 * of that size
		 */
		return -EINVAL;

5970 5971 5972 5973 5974 5975
	/* Refuse to reduce size of the array.  Any reductions in
	 * array size must be through explicit setting of array_size
	 * attribute.
	 */
	if (raid5_size(mddev, 0, conf->raid_disks + mddev->delta_disks)
	    < mddev->array_sectors) {
5976
		printk(KERN_ERR "md/raid:%s: array size must be reduced "
5977 5978 5979 5980
		       "before number of disks\n", mdname(mddev));
		return -EINVAL;
	}

5981
	atomic_set(&conf->reshape_stripes, 0);
5982 5983
	spin_lock_irq(&conf->device_lock);
	conf->previous_raid_disks = conf->raid_disks;
5984
	conf->raid_disks += mddev->delta_disks;
5985 5986
	conf->prev_chunk_sectors = conf->chunk_sectors;
	conf->chunk_sectors = mddev->new_chunk_sectors;
5987 5988
	conf->prev_algo = conf->algorithm;
	conf->algorithm = mddev->new_layout;
5989 5990 5991 5992 5993
	conf->generation++;
	/* Code that selects data_offset needs to see the generation update
	 * if reshape_progress has been set - so a memory barrier needed.
	 */
	smp_mb();
5994
	if (mddev->reshape_backwards)
5995 5996 5997 5998
		conf->reshape_progress = raid5_size(mddev, 0, 0);
	else
		conf->reshape_progress = 0;
	conf->reshape_safe = conf->reshape_progress;
5999 6000 6001 6002
	spin_unlock_irq(&conf->device_lock);

	/* Add some new drives, as many as will fit.
	 * We know there are enough to make the newly sized array work.
6003 6004 6005 6006
	 * Don't add devices if we are reducing the number of
	 * devices in the array.  This is because it is not possible
	 * to correctly record the "partially reconstructed" state of
	 * such devices during the reshape and confusion could result.
6007
	 */
6008
	if (mddev->delta_disks >= 0) {
N
NeilBrown 已提交
6009
		rdev_for_each(rdev, mddev)
6010 6011 6012 6013
			if (rdev->raid_disk < 0 &&
			    !test_bit(Faulty, &rdev->flags)) {
				if (raid5_add_disk(mddev, rdev) == 0) {
					if (rdev->raid_disk
6014
					    >= conf->previous_raid_disks)
6015
						set_bit(In_sync, &rdev->flags);
6016
					else
6017
						rdev->recovery_offset = 0;
6018 6019

					if (sysfs_link_rdev(mddev, rdev))
6020
						/* Failure here is OK */;
6021
				}
6022 6023 6024 6025 6026
			} else if (rdev->raid_disk >= conf->previous_raid_disks
				   && !test_bit(Faulty, &rdev->flags)) {
				/* This is a spare that was manually added */
				set_bit(In_sync, &rdev->flags);
			}
6027

6028 6029 6030 6031
		/* When a reshape changes the number of devices,
		 * ->degraded is measured against the larger of the
		 * pre and post number of devices.
		 */
6032
		spin_lock_irqsave(&conf->device_lock, flags);
6033
		mddev->degraded = calc_degraded(conf);
6034 6035
		spin_unlock_irqrestore(&conf->device_lock, flags);
	}
6036
	mddev->raid_disks = conf->raid_disks;
6037
	mddev->reshape_position = conf->reshape_progress;
6038
	set_bit(MD_CHANGE_DEVS, &mddev->flags);
6039

6040 6041 6042 6043 6044
	clear_bit(MD_RECOVERY_SYNC, &mddev->recovery);
	clear_bit(MD_RECOVERY_CHECK, &mddev->recovery);
	set_bit(MD_RECOVERY_RESHAPE, &mddev->recovery);
	set_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
	mddev->sync_thread = md_register_thread(md_do_sync, mddev,
6045
						"reshape");
6046 6047 6048 6049
	if (!mddev->sync_thread) {
		mddev->recovery = 0;
		spin_lock_irq(&conf->device_lock);
		mddev->raid_disks = conf->raid_disks = conf->previous_raid_disks;
6050 6051 6052
		rdev_for_each(rdev, mddev)
			rdev->new_data_offset = rdev->data_offset;
		smp_wmb();
6053
		conf->reshape_progress = MaxSector;
6054
		mddev->reshape_position = MaxSector;
6055 6056 6057
		spin_unlock_irq(&conf->device_lock);
		return -EAGAIN;
	}
6058
	conf->reshape_checkpoint = jiffies;
6059 6060 6061 6062 6063
	md_wakeup_thread(mddev->sync_thread);
	md_new_event(mddev);
	return 0;
}

6064 6065 6066
/* This is called from the reshape thread and should make any
 * changes needed in 'conf'
 */
6067
static void end_reshape(struct r5conf *conf)
6068 6069
{

6070
	if (!test_bit(MD_RECOVERY_INTR, &conf->mddev->recovery)) {
6071
		struct md_rdev *rdev;
6072 6073

		spin_lock_irq(&conf->device_lock);
6074
		conf->previous_raid_disks = conf->raid_disks;
6075 6076 6077
		rdev_for_each(rdev, conf->mddev)
			rdev->data_offset = rdev->new_data_offset;
		smp_wmb();
6078
		conf->reshape_progress = MaxSector;
6079
		spin_unlock_irq(&conf->device_lock);
6080
		wake_up(&conf->wait_for_overlap);
6081 6082 6083 6084

		/* read-ahead size must cover two whole stripes, which is
		 * 2 * (datadisks) * chunksize where 'n' is the number of raid devices
		 */
6085
		if (conf->mddev->queue) {
6086
			int data_disks = conf->raid_disks - conf->max_degraded;
6087
			int stripe = data_disks * ((conf->chunk_sectors << 9)
6088
						   / PAGE_SIZE);
6089 6090 6091
			if (conf->mddev->queue->backing_dev_info.ra_pages < 2 * stripe)
				conf->mddev->queue->backing_dev_info.ra_pages = 2 * stripe;
		}
6092 6093 6094
	}
}

6095 6096 6097
/* This is called from the raid5d thread with mddev_lock held.
 * It makes config changes to the device.
 */
6098
static void raid5_finish_reshape(struct mddev *mddev)
6099
{
6100
	struct r5conf *conf = mddev->private;
6101 6102 6103

	if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery)) {

6104 6105 6106
		if (mddev->delta_disks > 0) {
			md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
			set_capacity(mddev->gendisk, mddev->array_sectors);
6107
			revalidate_disk(mddev->gendisk);
6108 6109
		} else {
			int d;
6110 6111 6112
			spin_lock_irq(&conf->device_lock);
			mddev->degraded = calc_degraded(conf);
			spin_unlock_irq(&conf->device_lock);
6113 6114
			for (d = conf->raid_disks ;
			     d < conf->raid_disks - mddev->delta_disks;
6115
			     d++) {
6116
				struct md_rdev *rdev = conf->disks[d].rdev;
6117 6118 6119 6120 6121
				if (rdev)
					clear_bit(In_sync, &rdev->flags);
				rdev = conf->disks[d].replacement;
				if (rdev)
					clear_bit(In_sync, &rdev->flags);
6122
			}
6123
		}
6124
		mddev->layout = conf->algorithm;
6125
		mddev->chunk_sectors = conf->chunk_sectors;
6126 6127
		mddev->reshape_position = MaxSector;
		mddev->delta_disks = 0;
6128
		mddev->reshape_backwards = 0;
6129 6130 6131
	}
}

6132
static void raid5_quiesce(struct mddev *mddev, int state)
6133
{
6134
	struct r5conf *conf = mddev->private;
6135 6136

	switch(state) {
6137 6138 6139 6140
	case 2: /* resume for a suspend */
		wake_up(&conf->wait_for_overlap);
		break;

6141 6142
	case 1: /* stop all writes */
		spin_lock_irq(&conf->device_lock);
6143 6144 6145 6146
		/* '2' tells resync/reshape to pause so that all
		 * active stripes can drain
		 */
		conf->quiesce = 2;
6147
		wait_event_lock_irq(conf->wait_for_stripe,
6148 6149
				    atomic_read(&conf->active_stripes) == 0 &&
				    atomic_read(&conf->active_aligned_reads) == 0,
6150
				    conf->device_lock);
6151
		conf->quiesce = 1;
6152
		spin_unlock_irq(&conf->device_lock);
6153 6154
		/* allow reshape to continue */
		wake_up(&conf->wait_for_overlap);
6155 6156 6157 6158 6159 6160
		break;

	case 0: /* re-enable writes */
		spin_lock_irq(&conf->device_lock);
		conf->quiesce = 0;
		wake_up(&conf->wait_for_stripe);
6161
		wake_up(&conf->wait_for_overlap);
6162 6163 6164 6165
		spin_unlock_irq(&conf->device_lock);
		break;
	}
}
6166

6167

6168
static void *raid45_takeover_raid0(struct mddev *mddev, int level)
6169
{
6170
	struct r0conf *raid0_conf = mddev->private;
6171
	sector_t sectors;
6172

D
Dan Williams 已提交
6173
	/* for raid0 takeover only one zone is supported */
6174
	if (raid0_conf->nr_strip_zones > 1) {
6175 6176
		printk(KERN_ERR "md/raid:%s: cannot takeover raid0 with more than one zone.\n",
		       mdname(mddev));
D
Dan Williams 已提交
6177 6178 6179
		return ERR_PTR(-EINVAL);
	}

6180 6181
	sectors = raid0_conf->strip_zone[0].zone_end;
	sector_div(sectors, raid0_conf->strip_zone[0].nb_dev);
6182
	mddev->dev_sectors = sectors;
D
Dan Williams 已提交
6183
	mddev->new_level = level;
6184 6185 6186 6187 6188 6189 6190 6191 6192 6193 6194
	mddev->new_layout = ALGORITHM_PARITY_N;
	mddev->new_chunk_sectors = mddev->chunk_sectors;
	mddev->raid_disks += 1;
	mddev->delta_disks = 1;
	/* make sure it will be not marked as dirty */
	mddev->recovery_cp = MaxSector;

	return setup_conf(mddev);
}


6195
static void *raid5_takeover_raid1(struct mddev *mddev)
6196 6197 6198 6199 6200 6201 6202 6203 6204 6205 6206 6207 6208 6209 6210 6211 6212 6213 6214 6215 6216
{
	int chunksect;

	if (mddev->raid_disks != 2 ||
	    mddev->degraded > 1)
		return ERR_PTR(-EINVAL);

	/* Should check if there are write-behind devices? */

	chunksect = 64*2; /* 64K by default */

	/* The array must be an exact multiple of chunksize */
	while (chunksect && (mddev->array_sectors & (chunksect-1)))
		chunksect >>= 1;

	if ((chunksect<<9) < STRIPE_SIZE)
		/* array size does not allow a suitable chunk size */
		return ERR_PTR(-EINVAL);

	mddev->new_level = 5;
	mddev->new_layout = ALGORITHM_LEFT_SYMMETRIC;
6217
	mddev->new_chunk_sectors = chunksect;
6218 6219 6220 6221

	return setup_conf(mddev);
}

6222
static void *raid5_takeover_raid6(struct mddev *mddev)
6223 6224 6225 6226 6227 6228 6229 6230 6231 6232 6233 6234 6235 6236 6237 6238 6239 6240 6241 6242 6243 6244 6245 6246 6247 6248 6249 6250 6251 6252 6253 6254
{
	int new_layout;

	switch (mddev->layout) {
	case ALGORITHM_LEFT_ASYMMETRIC_6:
		new_layout = ALGORITHM_LEFT_ASYMMETRIC;
		break;
	case ALGORITHM_RIGHT_ASYMMETRIC_6:
		new_layout = ALGORITHM_RIGHT_ASYMMETRIC;
		break;
	case ALGORITHM_LEFT_SYMMETRIC_6:
		new_layout = ALGORITHM_LEFT_SYMMETRIC;
		break;
	case ALGORITHM_RIGHT_SYMMETRIC_6:
		new_layout = ALGORITHM_RIGHT_SYMMETRIC;
		break;
	case ALGORITHM_PARITY_0_6:
		new_layout = ALGORITHM_PARITY_0;
		break;
	case ALGORITHM_PARITY_N:
		new_layout = ALGORITHM_PARITY_N;
		break;
	default:
		return ERR_PTR(-EINVAL);
	}
	mddev->new_level = 5;
	mddev->new_layout = new_layout;
	mddev->delta_disks = -1;
	mddev->raid_disks -= 1;
	return setup_conf(mddev);
}

6255

6256
static int raid5_check_reshape(struct mddev *mddev)
6257
{
6258 6259 6260 6261
	/* For a 2-drive array, the layout and chunk size can be changed
	 * immediately as not restriping is needed.
	 * For larger arrays we record the new value - after validation
	 * to be used by a reshape pass.
6262
	 */
6263
	struct r5conf *conf = mddev->private;
6264
	int new_chunk = mddev->new_chunk_sectors;
6265

6266
	if (mddev->new_layout >= 0 && !algorithm_valid_raid5(mddev->new_layout))
6267 6268
		return -EINVAL;
	if (new_chunk > 0) {
6269
		if (!is_power_of_2(new_chunk))
6270
			return -EINVAL;
6271
		if (new_chunk < (PAGE_SIZE>>9))
6272
			return -EINVAL;
6273
		if (mddev->array_sectors & (new_chunk-1))
6274 6275 6276 6277 6278 6279
			/* not factor of array size */
			return -EINVAL;
	}

	/* They look valid */

6280
	if (mddev->raid_disks == 2) {
6281 6282 6283 6284
		/* can make the change immediately */
		if (mddev->new_layout >= 0) {
			conf->algorithm = mddev->new_layout;
			mddev->layout = mddev->new_layout;
6285 6286
		}
		if (new_chunk > 0) {
6287 6288
			conf->chunk_sectors = new_chunk ;
			mddev->chunk_sectors = new_chunk;
6289 6290 6291
		}
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
		md_wakeup_thread(mddev->thread);
6292
	}
6293
	return check_reshape(mddev);
6294 6295
}

6296
static int raid6_check_reshape(struct mddev *mddev)
6297
{
6298
	int new_chunk = mddev->new_chunk_sectors;
6299

6300
	if (mddev->new_layout >= 0 && !algorithm_valid_raid6(mddev->new_layout))
6301
		return -EINVAL;
6302
	if (new_chunk > 0) {
6303
		if (!is_power_of_2(new_chunk))
6304
			return -EINVAL;
6305
		if (new_chunk < (PAGE_SIZE >> 9))
6306
			return -EINVAL;
6307
		if (mddev->array_sectors & (new_chunk-1))
6308 6309
			/* not factor of array size */
			return -EINVAL;
6310
	}
6311 6312

	/* They look valid */
6313
	return check_reshape(mddev);
6314 6315
}

6316
static void *raid5_takeover(struct mddev *mddev)
6317 6318
{
	/* raid5 can take over:
D
Dan Williams 已提交
6319
	 *  raid0 - if there is only one strip zone - make it a raid4 layout
6320 6321 6322 6323
	 *  raid1 - if there are two drives.  We need to know the chunk size
	 *  raid4 - trivial - just use a raid4 layout.
	 *  raid6 - Providing it is a *_6 layout
	 */
D
Dan Williams 已提交
6324 6325
	if (mddev->level == 0)
		return raid45_takeover_raid0(mddev, 5);
6326 6327
	if (mddev->level == 1)
		return raid5_takeover_raid1(mddev);
6328 6329 6330 6331 6332
	if (mddev->level == 4) {
		mddev->new_layout = ALGORITHM_PARITY_N;
		mddev->new_level = 5;
		return setup_conf(mddev);
	}
6333 6334
	if (mddev->level == 6)
		return raid5_takeover_raid6(mddev);
6335 6336 6337 6338

	return ERR_PTR(-EINVAL);
}

6339
static void *raid4_takeover(struct mddev *mddev)
6340
{
D
Dan Williams 已提交
6341 6342 6343
	/* raid4 can take over:
	 *  raid0 - if there is only one strip zone
	 *  raid5 - if layout is right
6344
	 */
D
Dan Williams 已提交
6345 6346
	if (mddev->level == 0)
		return raid45_takeover_raid0(mddev, 4);
6347 6348 6349 6350 6351 6352 6353 6354
	if (mddev->level == 5 &&
	    mddev->layout == ALGORITHM_PARITY_N) {
		mddev->new_layout = 0;
		mddev->new_level = 4;
		return setup_conf(mddev);
	}
	return ERR_PTR(-EINVAL);
}
6355

6356
static struct md_personality raid5_personality;
6357

6358
static void *raid6_takeover(struct mddev *mddev)
6359 6360 6361 6362 6363 6364 6365 6366 6367 6368 6369 6370 6371 6372 6373 6374 6375 6376 6377 6378 6379 6380 6381 6382 6383 6384 6385 6386 6387 6388 6389 6390 6391 6392 6393 6394 6395 6396 6397 6398 6399 6400 6401 6402 6403 6404
{
	/* Currently can only take over a raid5.  We map the
	 * personality to an equivalent raid6 personality
	 * with the Q block at the end.
	 */
	int new_layout;

	if (mddev->pers != &raid5_personality)
		return ERR_PTR(-EINVAL);
	if (mddev->degraded > 1)
		return ERR_PTR(-EINVAL);
	if (mddev->raid_disks > 253)
		return ERR_PTR(-EINVAL);
	if (mddev->raid_disks < 3)
		return ERR_PTR(-EINVAL);

	switch (mddev->layout) {
	case ALGORITHM_LEFT_ASYMMETRIC:
		new_layout = ALGORITHM_LEFT_ASYMMETRIC_6;
		break;
	case ALGORITHM_RIGHT_ASYMMETRIC:
		new_layout = ALGORITHM_RIGHT_ASYMMETRIC_6;
		break;
	case ALGORITHM_LEFT_SYMMETRIC:
		new_layout = ALGORITHM_LEFT_SYMMETRIC_6;
		break;
	case ALGORITHM_RIGHT_SYMMETRIC:
		new_layout = ALGORITHM_RIGHT_SYMMETRIC_6;
		break;
	case ALGORITHM_PARITY_0:
		new_layout = ALGORITHM_PARITY_0_6;
		break;
	case ALGORITHM_PARITY_N:
		new_layout = ALGORITHM_PARITY_N;
		break;
	default:
		return ERR_PTR(-EINVAL);
	}
	mddev->new_level = 6;
	mddev->new_layout = new_layout;
	mddev->delta_disks = 1;
	mddev->raid_disks += 1;
	return setup_conf(mddev);
}


6405
static struct md_personality raid6_personality =
6406 6407 6408 6409 6410 6411 6412 6413 6414 6415 6416 6417 6418 6419
{
	.name		= "raid6",
	.level		= 6,
	.owner		= THIS_MODULE,
	.make_request	= make_request,
	.run		= run,
	.stop		= stop,
	.status		= status,
	.error_handler	= error,
	.hot_add_disk	= raid5_add_disk,
	.hot_remove_disk= raid5_remove_disk,
	.spare_active	= raid5_spare_active,
	.sync_request	= sync_request,
	.resize		= raid5_resize,
6420
	.size		= raid5_size,
6421
	.check_reshape	= raid6_check_reshape,
6422
	.start_reshape  = raid5_start_reshape,
6423
	.finish_reshape = raid5_finish_reshape,
6424
	.quiesce	= raid5_quiesce,
6425
	.takeover	= raid6_takeover,
6426
};
6427
static struct md_personality raid5_personality =
L
Linus Torvalds 已提交
6428 6429
{
	.name		= "raid5",
6430
	.level		= 5,
L
Linus Torvalds 已提交
6431 6432 6433 6434 6435 6436 6437 6438 6439 6440 6441
	.owner		= THIS_MODULE,
	.make_request	= make_request,
	.run		= run,
	.stop		= stop,
	.status		= status,
	.error_handler	= error,
	.hot_add_disk	= raid5_add_disk,
	.hot_remove_disk= raid5_remove_disk,
	.spare_active	= raid5_spare_active,
	.sync_request	= sync_request,
	.resize		= raid5_resize,
6442
	.size		= raid5_size,
6443 6444
	.check_reshape	= raid5_check_reshape,
	.start_reshape  = raid5_start_reshape,
6445
	.finish_reshape = raid5_finish_reshape,
6446
	.quiesce	= raid5_quiesce,
6447
	.takeover	= raid5_takeover,
L
Linus Torvalds 已提交
6448 6449
};

6450
static struct md_personality raid4_personality =
L
Linus Torvalds 已提交
6451
{
6452 6453 6454 6455 6456 6457 6458 6459 6460 6461 6462 6463 6464
	.name		= "raid4",
	.level		= 4,
	.owner		= THIS_MODULE,
	.make_request	= make_request,
	.run		= run,
	.stop		= stop,
	.status		= status,
	.error_handler	= error,
	.hot_add_disk	= raid5_add_disk,
	.hot_remove_disk= raid5_remove_disk,
	.spare_active	= raid5_spare_active,
	.sync_request	= sync_request,
	.resize		= raid5_resize,
6465
	.size		= raid5_size,
6466 6467
	.check_reshape	= raid5_check_reshape,
	.start_reshape  = raid5_start_reshape,
6468
	.finish_reshape = raid5_finish_reshape,
6469
	.quiesce	= raid5_quiesce,
6470
	.takeover	= raid4_takeover,
6471 6472 6473 6474
};

static int __init raid5_init(void)
{
6475
	register_md_personality(&raid6_personality);
6476 6477 6478
	register_md_personality(&raid5_personality);
	register_md_personality(&raid4_personality);
	return 0;
L
Linus Torvalds 已提交
6479 6480
}

6481
static void raid5_exit(void)
L
Linus Torvalds 已提交
6482
{
6483
	unregister_md_personality(&raid6_personality);
6484 6485
	unregister_md_personality(&raid5_personality);
	unregister_md_personality(&raid4_personality);
L
Linus Torvalds 已提交
6486 6487 6488 6489 6490
}

module_init(raid5_init);
module_exit(raid5_exit);
MODULE_LICENSE("GPL");
6491
MODULE_DESCRIPTION("RAID4/5/6 (striping with parity) personality for MD");
L
Linus Torvalds 已提交
6492
MODULE_ALIAS("md-personality-4"); /* RAID5 */
6493 6494
MODULE_ALIAS("md-raid5");
MODULE_ALIAS("md-raid4");
6495 6496
MODULE_ALIAS("md-level-5");
MODULE_ALIAS("md-level-4");
6497 6498 6499 6500 6501 6502 6503
MODULE_ALIAS("md-personality-8"); /* RAID6 */
MODULE_ALIAS("md-raid6");
MODULE_ALIAS("md-level-6");

/* This used to be two separate modules, they were: */
MODULE_ALIAS("raid5");
MODULE_ALIAS("raid6");