raid5.c 181.6 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;

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

			if (conf->mddev->gendisk)
				trace_block_bio_remap(bdev_get_queue(bi->bi_bdev),
						      bi, disk_devt(conf->mddev->gendisk),
						      sh->dev[i].sector);
677
			generic_make_request(bi);
678 679
		}
		if (rrdev) {
680 681
			if (s->syncing || s->expanding || s->expanded
			    || s->replacing)
682 683 684 685
				md_sync_acct(rrdev->bdev, STRIPE_SECTORS);

			set_bit(STRIPE_IO_STARTED, &sh->state);

K
Kent Overstreet 已提交
686
			bio_reset(rbi);
687
			rbi->bi_bdev = rrdev->bdev;
K
Kent Overstreet 已提交
688 689 690 691 692
			rbi->bi_rw = rw;
			BUG_ON(!(rw & WRITE));
			rbi->bi_end_io = raid5_end_write_request;
			rbi->bi_private = sh;

693 694 695 696 697
			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);
698 699 700 701 702 703
			if (use_new_offset(conf, sh))
				rbi->bi_sector = (sh->sector
						  + rrdev->new_data_offset);
			else
				rbi->bi_sector = (sh->sector
						  + rrdev->data_offset);
704 705 706
			rbi->bi_io_vec[0].bv_len = STRIPE_SIZE;
			rbi->bi_io_vec[0].bv_offset = 0;
			rbi->bi_size = STRIPE_SIZE;
707 708 709 710
			if (conf->mddev->gendisk)
				trace_block_bio_remap(bdev_get_queue(rbi->bi_bdev),
						      rbi, disk_devt(conf->mddev->gendisk),
						      sh->dev[i].sector);
711 712 713
			generic_make_request(rbi);
		}
		if (!rdev && !rrdev) {
714
			if (rw & WRITE)
715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731
				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;
732
	struct async_submit_ctl submit;
D
Dan Williams 已提交
733
	enum async_tx_flags flags = 0;
734 735 736 737 738

	if (bio->bi_sector >= sector)
		page_offset = (signed)(bio->bi_sector - sector) * 512;
	else
		page_offset = (signed)(sector - bio->bi_sector) * -512;
739

D
Dan Williams 已提交
740 741 742 743
	if (frombio)
		flags |= ASYNC_TX_FENCE;
	init_async_submit(&submit, flags, tx, NULL, NULL, NULL);

744
	bio_for_each_segment(bvl, bio, i) {
745
		int len = bvl->bv_len;
746 747 748 749 750 751 752 753 754 755 756 757 758 759 760
		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) {
761 762
			b_offset += bvl->bv_offset;
			bio_page = bvl->bv_page;
763 764
			if (frombio)
				tx = async_memcpy(page, bio_page, page_offset,
765
						  b_offset, clen, &submit);
766 767
			else
				tx = async_memcpy(bio_page, page, b_offset,
768
						  page_offset, clen, &submit);
769
		}
770 771 772
		/* chain the operations */
		submit.depend_tx = tx;

773 774 775 776 777 778 779 780 781 782 783 784
		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;
785
	int i;
786

787
	pr_debug("%s: stripe %llu\n", __func__,
788 789 790 791 792 793 794
		(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 */
795 796
		/* and check if we need to reply to a read request,
		 * new R5_Wantfill requests are held off until
797
		 * !STRIPE_BIOFILL_RUN
798 799
		 */
		if (test_and_clear_bit(R5_Wantfill, &dev->flags)) {
800 801 802 803 804 805 806 807
			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);
808
				if (!raid5_dec_bi_active_stripes(rbi)) {
809 810 811 812 813 814 815
					rbi->bi_next = return_bi;
					return_bi = rbi;
				}
				rbi = rbi2;
			}
		}
	}
816
	clear_bit(STRIPE_BIOFILL_RUN, &sh->state);
817 818 819

	return_io(return_bi);

820
	set_bit(STRIPE_HANDLE, &sh->state);
821 822 823 824 825 826
	release_stripe(sh);
}

static void ops_run_biofill(struct stripe_head *sh)
{
	struct dma_async_tx_descriptor *tx = NULL;
827
	struct async_submit_ctl submit;
828 829
	int i;

830
	pr_debug("%s: stripe %llu\n", __func__,
831 832 833 834 835 836
		(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
837
			spin_lock_irq(&sh->stripe_lock);
838 839
			dev->read = rbi = dev->toread;
			dev->toread = NULL;
S
Shaohua Li 已提交
840
			spin_unlock_irq(&sh->stripe_lock);
841 842 843 844 845 846 847 848 849 850
			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);
851 852
	init_async_submit(&submit, ASYNC_TX_ACK, tx, ops_complete_biofill, sh, NULL);
	async_trigger_callback(&submit);
853 854
}

855
static void mark_target_uptodate(struct stripe_head *sh, int target)
856
{
857
	struct r5dev *tgt;
858

859 860
	if (target < 0)
		return;
861

862
	tgt = &sh->dev[target];
863 864 865
	set_bit(R5_UPTODATE, &tgt->flags);
	BUG_ON(!test_bit(R5_Wantcompute, &tgt->flags));
	clear_bit(R5_Wantcompute, &tgt->flags);
866 867
}

868
static void ops_complete_compute(void *stripe_head_ref)
869 870 871
{
	struct stripe_head *sh = stripe_head_ref;

872
	pr_debug("%s: stripe %llu\n", __func__,
873 874
		(unsigned long long)sh->sector);

875
	/* mark the computed target(s) as uptodate */
876
	mark_target_uptodate(sh, sh->ops.target);
877
	mark_target_uptodate(sh, sh->ops.target2);
878

879 880 881
	clear_bit(STRIPE_COMPUTE_RUN, &sh->state);
	if (sh->check_state == check_state_compute_run)
		sh->check_state = check_state_compute_result;
882 883 884 885
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

886 887 888 889 890 891 892 893 894
/* 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)
895 896
{
	int disks = sh->disks;
897
	struct page **xor_srcs = percpu->scribble;
898 899 900 901 902
	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;
903
	struct async_submit_ctl submit;
904 905 906
	int i;

	pr_debug("%s: stripe %llu block: %d\n",
907
		__func__, (unsigned long long)sh->sector, target);
908 909 910 911 912 913 914 915
	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
916
	init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST, NULL,
917
			  ops_complete_compute, sh, to_addr_conv(sh, percpu));
918
	if (unlikely(count == 1))
919
		tx = async_memcpy(xor_dest, xor_srcs[0], 0, 0, STRIPE_SIZE, &submit);
920
	else
921
		tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, &submit);
922 923 924 925

	return tx;
}

926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943
/* 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++)
944
		srcs[i] = NULL;
945 946 947 948 949 950 951 952 953 954

	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);

955
	return syndrome_disks;
956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975
}

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;
976
	else
977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992
		/* 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
993 994
		init_async_submit(&submit, ASYNC_TX_FENCE, NULL,
				  ops_complete_compute, sh,
995 996 997 998 999 1000 1001 1002 1003 1004 1005
				  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
1006 1007
		init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST,
				  NULL, ops_complete_compute, sh,
1008 1009 1010
				  to_addr_conv(sh, percpu));
		tx = async_xor(dest, blocks, 0, count, STRIPE_SIZE, &submit);
	}
1011 1012 1013 1014

	return tx;
}

1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035
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));

1036
	/* we need to open-code set_syndrome_sources to handle the
1037 1038 1039
	 * slot number conversion for 'faila' and 'failb'
	 */
	for (i = 0; i < disks ; i++)
1040
		blocks[i] = NULL;
1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066
	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
1067 1068 1069
			init_async_submit(&submit, ASYNC_TX_FENCE, NULL,
					  ops_complete_compute, sh,
					  to_addr_conv(sh, percpu));
1070
			return async_gen_syndrome(blocks, 0, syndrome_disks+2,
1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089
						  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
1090 1091 1092 1093
			init_async_submit(&submit,
					  ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST,
					  NULL, NULL, NULL,
					  to_addr_conv(sh, percpu));
1094 1095 1096 1097
			tx = async_xor(dest, blocks, 0, count, STRIPE_SIZE,
				       &submit);

			count = set_syndrome_sources(blocks, sh);
D
Dan Williams 已提交
1098 1099 1100
			init_async_submit(&submit, ASYNC_TX_FENCE, tx,
					  ops_complete_compute, sh,
					  to_addr_conv(sh, percpu));
1101 1102 1103 1104
			return async_gen_syndrome(blocks, 0, count+2,
						  STRIPE_SIZE, &submit);
		}
	} else {
1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118
		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);
		}
1119 1120 1121 1122
	}
}


1123 1124 1125 1126
static void ops_complete_prexor(void *stripe_head_ref)
{
	struct stripe_head *sh = stripe_head_ref;

1127
	pr_debug("%s: stripe %llu\n", __func__,
1128 1129 1130 1131
		(unsigned long long)sh->sector);
}

static struct dma_async_tx_descriptor *
1132 1133
ops_run_prexor(struct stripe_head *sh, struct raid5_percpu *percpu,
	       struct dma_async_tx_descriptor *tx)
1134 1135
{
	int disks = sh->disks;
1136
	struct page **xor_srcs = percpu->scribble;
1137
	int count = 0, pd_idx = sh->pd_idx, i;
1138
	struct async_submit_ctl submit;
1139 1140 1141 1142

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

1143
	pr_debug("%s: stripe %llu\n", __func__,
1144 1145 1146 1147 1148
		(unsigned long long)sh->sector);

	for (i = disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
		/* Only process blocks that are known to be uptodate */
1149
		if (test_bit(R5_Wantdrain, &dev->flags))
1150 1151 1152
			xor_srcs[count++] = dev->page;
	}

D
Dan Williams 已提交
1153
	init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_DROP_DST, tx,
1154
			  ops_complete_prexor, sh, to_addr_conv(sh, percpu));
1155
	tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, &submit);
1156 1157 1158 1159 1160

	return tx;
}

static struct dma_async_tx_descriptor *
1161
ops_run_biodrain(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
1162 1163
{
	int disks = sh->disks;
1164
	int i;
1165

1166
	pr_debug("%s: stripe %llu\n", __func__,
1167 1168 1169 1170 1171 1172
		(unsigned long long)sh->sector);

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

1173
		if (test_and_clear_bit(R5_Wantdrain, &dev->flags)) {
1174 1175
			struct bio *wbi;

S
Shaohua Li 已提交
1176
			spin_lock_irq(&sh->stripe_lock);
1177 1178 1179 1180
			chosen = dev->towrite;
			dev->towrite = NULL;
			BUG_ON(dev->written);
			wbi = dev->written = chosen;
S
Shaohua Li 已提交
1181
			spin_unlock_irq(&sh->stripe_lock);
1182 1183 1184

			while (wbi && wbi->bi_sector <
				dev->sector + STRIPE_SECTORS) {
T
Tejun Heo 已提交
1185 1186
				if (wbi->bi_rw & REQ_FUA)
					set_bit(R5_WantFUA, &dev->flags);
S
Shaohua Li 已提交
1187 1188
				if (wbi->bi_rw & REQ_SYNC)
					set_bit(R5_SyncIO, &dev->flags);
1189
				if (wbi->bi_rw & REQ_DISCARD)
S
Shaohua Li 已提交
1190
					set_bit(R5_Discard, &dev->flags);
1191
				else
S
Shaohua Li 已提交
1192 1193
					tx = async_copy_data(1, wbi, dev->page,
						dev->sector, tx);
1194 1195 1196 1197 1198 1199 1200 1201
				wbi = r5_next_bio(wbi, dev->sector);
			}
		}
	}

	return tx;
}

1202
static void ops_complete_reconstruct(void *stripe_head_ref)
1203 1204
{
	struct stripe_head *sh = stripe_head_ref;
1205 1206 1207 1208
	int disks = sh->disks;
	int pd_idx = sh->pd_idx;
	int qd_idx = sh->qd_idx;
	int i;
1209
	bool fua = false, sync = false, discard = false;
1210

1211
	pr_debug("%s: stripe %llu\n", __func__,
1212 1213
		(unsigned long long)sh->sector);

S
Shaohua Li 已提交
1214
	for (i = disks; i--; ) {
T
Tejun Heo 已提交
1215
		fua |= test_bit(R5_WantFUA, &sh->dev[i].flags);
S
Shaohua Li 已提交
1216
		sync |= test_bit(R5_SyncIO, &sh->dev[i].flags);
1217
		discard |= test_bit(R5_Discard, &sh->dev[i].flags);
S
Shaohua Li 已提交
1218
	}
T
Tejun Heo 已提交
1219

1220 1221
	for (i = disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
1222

T
Tejun Heo 已提交
1223
		if (dev->written || i == pd_idx || i == qd_idx) {
1224 1225
			if (!discard)
				set_bit(R5_UPTODATE, &dev->flags);
T
Tejun Heo 已提交
1226 1227
			if (fua)
				set_bit(R5_WantFUA, &dev->flags);
S
Shaohua Li 已提交
1228 1229
			if (sync)
				set_bit(R5_SyncIO, &dev->flags);
T
Tejun Heo 已提交
1230
		}
1231 1232
	}

1233 1234 1235 1236 1237 1238 1239 1240
	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;
	}
1241 1242 1243 1244 1245 1246

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

static void
1247 1248
ops_run_reconstruct5(struct stripe_head *sh, struct raid5_percpu *percpu,
		     struct dma_async_tx_descriptor *tx)
1249 1250
{
	int disks = sh->disks;
1251
	struct page **xor_srcs = percpu->scribble;
1252
	struct async_submit_ctl submit;
1253 1254
	int count = 0, pd_idx = sh->pd_idx, i;
	struct page *xor_dest;
1255
	int prexor = 0;
1256 1257
	unsigned long flags;

1258
	pr_debug("%s: stripe %llu\n", __func__,
1259 1260
		(unsigned long long)sh->sector);

S
Shaohua Li 已提交
1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272
	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;
	}
1273 1274 1275
	/* check if prexor is active which means only process blocks
	 * that are part of a read-modify-write (written)
	 */
1276 1277
	if (sh->reconstruct_state == reconstruct_state_prexor_drain_run) {
		prexor = 1;
1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297
		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
	 */
1298
	flags = ASYNC_TX_ACK |
1299 1300 1301 1302
		(prexor ? ASYNC_TX_XOR_DROP_DST : ASYNC_TX_XOR_ZERO_DST);

	atomic_inc(&sh->count);

1303
	init_async_submit(&submit, flags, tx, ops_complete_reconstruct, sh,
1304
			  to_addr_conv(sh, percpu));
1305 1306 1307 1308
	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);
1309 1310
}

1311 1312 1313 1314 1315 1316
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
1317
	int count, i;
1318 1319 1320

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

S
Shaohua Li 已提交
1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334
	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;
	}

1335 1336 1337 1338 1339 1340 1341
	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);
1342 1343 1344 1345 1346 1347
}

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

1348
	pr_debug("%s: stripe %llu\n", __func__,
1349 1350
		(unsigned long long)sh->sector);

1351
	sh->check_state = check_state_check_result;
1352 1353 1354 1355
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

1356
static void ops_run_check_p(struct stripe_head *sh, struct raid5_percpu *percpu)
1357 1358
{
	int disks = sh->disks;
1359 1360 1361
	int pd_idx = sh->pd_idx;
	int qd_idx = sh->qd_idx;
	struct page *xor_dest;
1362
	struct page **xor_srcs = percpu->scribble;
1363
	struct dma_async_tx_descriptor *tx;
1364
	struct async_submit_ctl submit;
1365 1366
	int count;
	int i;
1367

1368
	pr_debug("%s: stripe %llu\n", __func__,
1369 1370
		(unsigned long long)sh->sector);

1371 1372 1373
	count = 0;
	xor_dest = sh->dev[pd_idx].page;
	xor_srcs[count++] = xor_dest;
1374
	for (i = disks; i--; ) {
1375 1376 1377
		if (i == pd_idx || i == qd_idx)
			continue;
		xor_srcs[count++] = sh->dev[i].page;
1378 1379
	}

1380 1381
	init_async_submit(&submit, 0, NULL, NULL, NULL,
			  to_addr_conv(sh, percpu));
D
Dan Williams 已提交
1382
	tx = async_xor_val(xor_dest, xor_srcs, 0, count, STRIPE_SIZE,
1383
			   &sh->ops.zero_sum_result, &submit);
1384 1385

	atomic_inc(&sh->count);
1386 1387
	init_async_submit(&submit, ASYNC_TX_ACK, tx, ops_complete_check, sh, NULL);
	tx = async_trigger_callback(&submit);
1388 1389
}

1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401
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;
1402 1403

	atomic_inc(&sh->count);
1404 1405 1406 1407
	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);
1408 1409
}

N
NeilBrown 已提交
1410
static void raid_run_ops(struct stripe_head *sh, unsigned long ops_request)
1411 1412 1413
{
	int overlap_clear = 0, i, disks = sh->disks;
	struct dma_async_tx_descriptor *tx = NULL;
1414
	struct r5conf *conf = sh->raid_conf;
1415
	int level = conf->level;
1416 1417
	struct raid5_percpu *percpu;
	unsigned long cpu;
1418

1419 1420
	cpu = get_cpu();
	percpu = per_cpu_ptr(conf->percpu, cpu);
1421
	if (test_bit(STRIPE_OP_BIOFILL, &ops_request)) {
1422 1423 1424 1425
		ops_run_biofill(sh);
		overlap_clear++;
	}

1426
	if (test_bit(STRIPE_OP_COMPUTE_BLK, &ops_request)) {
1427 1428 1429 1430 1431 1432 1433 1434 1435 1436
		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))
1437 1438
			async_tx_ack(tx);
	}
1439

1440
	if (test_bit(STRIPE_OP_PREXOR, &ops_request))
1441
		tx = ops_run_prexor(sh, percpu, tx);
1442

1443
	if (test_bit(STRIPE_OP_BIODRAIN, &ops_request)) {
1444
		tx = ops_run_biodrain(sh, tx);
1445 1446 1447
		overlap_clear++;
	}

1448 1449 1450 1451 1452 1453
	if (test_bit(STRIPE_OP_RECONSTRUCT, &ops_request)) {
		if (level < 6)
			ops_run_reconstruct5(sh, percpu, tx);
		else
			ops_run_reconstruct6(sh, percpu, tx);
	}
1454

1455 1456 1457 1458 1459 1460 1461 1462 1463 1464
	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();
	}
1465 1466 1467 1468 1469 1470 1471

	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);
		}
1472
	put_cpu();
1473 1474
}

1475
static int grow_one_stripe(struct r5conf *conf)
L
Linus Torvalds 已提交
1476 1477
{
	struct stripe_head *sh;
N
Namhyung Kim 已提交
1478
	sh = kmem_cache_zalloc(conf->slab_cache, GFP_KERNEL);
1479 1480
	if (!sh)
		return 0;
N
Namhyung Kim 已提交
1481

1482 1483
	sh->raid_conf = conf;

S
Shaohua Li 已提交
1484 1485
	spin_lock_init(&sh->stripe_lock);

1486 1487
	if (grow_buffers(sh)) {
		shrink_buffers(sh);
1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498
		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;
}

1499
static int grow_stripes(struct r5conf *conf, int num)
1500
{
1501
	struct kmem_cache *sc;
1502
	int devs = max(conf->raid_disks, conf->previous_raid_disks);
L
Linus Torvalds 已提交
1503

1504 1505 1506 1507 1508 1509 1510 1511
	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]);

1512 1513
	conf->active_name = 0;
	sc = kmem_cache_create(conf->cache_name[conf->active_name],
L
Linus Torvalds 已提交
1514
			       sizeof(struct stripe_head)+(devs-1)*sizeof(struct r5dev),
1515
			       0, 0, NULL);
L
Linus Torvalds 已提交
1516 1517 1518
	if (!sc)
		return 1;
	conf->slab_cache = sc;
1519
	conf->pool_size = devs;
1520
	while (num--)
1521
		if (!grow_one_stripe(conf))
L
Linus Torvalds 已提交
1522 1523 1524
			return 1;
	return 0;
}
1525

1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547
/**
 * 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;
}

1548
static int resize_stripes(struct r5conf *conf, int newsize)
1549 1550 1551 1552 1553 1554 1555
{
	/* 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
1556
	 * 2/ gather all the old stripe_heads and transfer the pages across
1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575
	 *    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;
1576
	unsigned long cpu;
1577
	int err;
1578
	struct kmem_cache *sc;
1579 1580 1581 1582 1583
	int i;

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

1584 1585 1586
	err = md_allow_write(conf->mddev);
	if (err)
		return err;
1587

1588 1589 1590
	/* Step 1 */
	sc = kmem_cache_create(conf->cache_name[1-conf->active_name],
			       sizeof(struct stripe_head)+(newsize-1)*sizeof(struct r5dev),
1591
			       0, 0, NULL);
1592 1593 1594 1595
	if (!sc)
		return -ENOMEM;

	for (i = conf->max_nr_stripes; i; i--) {
N
Namhyung Kim 已提交
1596
		nsh = kmem_cache_zalloc(sc, GFP_KERNEL);
1597 1598 1599 1600
		if (!nsh)
			break;

		nsh->raid_conf = conf;
1601
		spin_lock_init(&nsh->stripe_lock);
1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622

		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),
1623
				    conf->device_lock);
1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637
		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
1638
	 * conf->disks and the scribble region
1639 1640 1641 1642 1643 1644 1645 1646 1647 1648
	 */
	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;

1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667
	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();

1668 1669 1670 1671
	/* 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);
1672

1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688
		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
1689

1690
static int drop_one_stripe(struct r5conf *conf)
L
Linus Torvalds 已提交
1691 1692 1693
{
	struct stripe_head *sh;

1694 1695 1696 1697 1698
	spin_lock_irq(&conf->device_lock);
	sh = get_free_stripe(conf);
	spin_unlock_irq(&conf->device_lock);
	if (!sh)
		return 0;
1699
	BUG_ON(atomic_read(&sh->count));
1700
	shrink_buffers(sh);
1701 1702 1703 1704 1705
	kmem_cache_free(conf->slab_cache, sh);
	atomic_dec(&conf->active_stripes);
	return 1;
}

1706
static void shrink_stripes(struct r5conf *conf)
1707 1708 1709 1710
{
	while (drop_one_stripe(conf))
		;

N
NeilBrown 已提交
1711 1712
	if (conf->slab_cache)
		kmem_cache_destroy(conf->slab_cache);
L
Linus Torvalds 已提交
1713 1714 1715
	conf->slab_cache = NULL;
}

1716
static void raid5_end_read_request(struct bio * bi, int error)
L
Linus Torvalds 已提交
1717
{
1718
	struct stripe_head *sh = bi->bi_private;
1719
	struct r5conf *conf = sh->raid_conf;
1720
	int disks = sh->disks, i;
L
Linus Torvalds 已提交
1721
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
1722
	char b[BDEVNAME_SIZE];
1723
	struct md_rdev *rdev = NULL;
1724
	sector_t s;
L
Linus Torvalds 已提交
1725 1726 1727 1728 1729

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

1730 1731
	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
1732 1733 1734
		uptodate);
	if (i == disks) {
		BUG();
1735
		return;
L
Linus Torvalds 已提交
1736
	}
1737
	if (test_bit(R5_ReadRepl, &sh->dev[i].flags))
1738 1739 1740 1741 1742
		/* 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.
		 */
1743
		rdev = conf->disks[i].replacement;
1744
	if (!rdev)
1745
		rdev = conf->disks[i].rdev;
L
Linus Torvalds 已提交
1746

1747 1748 1749 1750
	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
1751 1752
	if (uptodate) {
		set_bit(R5_UPTODATE, &sh->dev[i].flags);
1753
		if (test_bit(R5_ReadError, &sh->dev[i].flags)) {
1754 1755 1756 1757
			/* Note that this cannot happen on a
			 * replacement device.  We just fail those on
			 * any error
			 */
1758 1759 1760 1761 1762
			printk_ratelimited(
				KERN_INFO
				"md/raid:%s: read error corrected"
				" (%lu sectors at %llu on %s)\n",
				mdname(conf->mddev), STRIPE_SECTORS,
1763
				(unsigned long long)s,
1764
				bdevname(rdev->bdev, b));
1765
			atomic_add(STRIPE_SECTORS, &rdev->corrected_errors);
1766 1767
			clear_bit(R5_ReadError, &sh->dev[i].flags);
			clear_bit(R5_ReWrite, &sh->dev[i].flags);
1768 1769 1770
		} else if (test_bit(R5_ReadNoMerge, &sh->dev[i].flags))
			clear_bit(R5_ReadNoMerge, &sh->dev[i].flags);

1771 1772
		if (atomic_read(&rdev->read_errors))
			atomic_set(&rdev->read_errors, 0);
L
Linus Torvalds 已提交
1773
	} else {
1774
		const char *bdn = bdevname(rdev->bdev, b);
1775
		int retry = 0;
1776
		int set_bad = 0;
1777

L
Linus Torvalds 已提交
1778
		clear_bit(R5_UPTODATE, &sh->dev[i].flags);
1779
		atomic_inc(&rdev->read_errors);
1780 1781 1782 1783 1784 1785
		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),
1786
				(unsigned long long)s,
1787
				bdn);
1788 1789
		else if (conf->mddev->degraded >= conf->max_degraded) {
			set_bad = 1;
1790 1791 1792 1793 1794
			printk_ratelimited(
				KERN_WARNING
				"md/raid:%s: read error not correctable "
				"(sector %llu on %s).\n",
				mdname(conf->mddev),
1795
				(unsigned long long)s,
1796
				bdn);
1797
		} else if (test_bit(R5_ReWrite, &sh->dev[i].flags)) {
1798
			/* Oh, no!!! */
1799
			set_bad = 1;
1800 1801 1802 1803 1804
			printk_ratelimited(
				KERN_WARNING
				"md/raid:%s: read error NOT corrected!! "
				"(sector %llu on %s).\n",
				mdname(conf->mddev),
1805
				(unsigned long long)s,
1806
				bdn);
1807
		} else if (atomic_read(&rdev->read_errors)
1808
			 > conf->max_nr_stripes)
N
NeilBrown 已提交
1809
			printk(KERN_WARNING
1810
			       "md/raid:%s: Too many read errors, failing device %s.\n",
1811
			       mdname(conf->mddev), bdn);
1812 1813 1814
		else
			retry = 1;
		if (retry)
1815 1816 1817 1818 1819
			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);
1820
		else {
1821 1822
			clear_bit(R5_ReadError, &sh->dev[i].flags);
			clear_bit(R5_ReWrite, &sh->dev[i].flags);
1823 1824 1825 1826 1827
			if (!(set_bad
			      && test_bit(In_sync, &rdev->flags)
			      && rdev_set_badblocks(
				      rdev, sh->sector, STRIPE_SECTORS, 0)))
				md_error(conf->mddev, rdev);
1828
		}
L
Linus Torvalds 已提交
1829
	}
1830
	rdev_dec_pending(rdev, conf->mddev);
L
Linus Torvalds 已提交
1831 1832 1833 1834 1835
	clear_bit(R5_LOCKED, &sh->dev[i].flags);
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

1836
static void raid5_end_write_request(struct bio *bi, int error)
L
Linus Torvalds 已提交
1837
{
1838
	struct stripe_head *sh = bi->bi_private;
1839
	struct r5conf *conf = sh->raid_conf;
1840
	int disks = sh->disks, i;
1841
	struct md_rdev *uninitialized_var(rdev);
L
Linus Torvalds 已提交
1842
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
1843 1844
	sector_t first_bad;
	int bad_sectors;
1845
	int replacement = 0;
L
Linus Torvalds 已提交
1846

1847 1848 1849
	for (i = 0 ; i < disks; i++) {
		if (bi == &sh->dev[i].req) {
			rdev = conf->disks[i].rdev;
L
Linus Torvalds 已提交
1850
			break;
1851 1852 1853
		}
		if (bi == &sh->dev[i].rreq) {
			rdev = conf->disks[i].replacement;
1854 1855 1856 1857 1858 1859 1860 1861
			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;
1862 1863 1864
			break;
		}
	}
1865
	pr_debug("end_write_request %llu/%d, count %d, uptodate: %d.\n",
L
Linus Torvalds 已提交
1866 1867 1868 1869
		(unsigned long long)sh->sector, i, atomic_read(&sh->count),
		uptodate);
	if (i == disks) {
		BUG();
1870
		return;
L
Linus Torvalds 已提交
1871 1872
	}

1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883
	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);
1884 1885 1886
			if (!test_and_set_bit(WantReplacement, &rdev->flags))
				set_bit(MD_RECOVERY_NEEDED,
					&rdev->mddev->recovery);
1887 1888
		} else if (is_badblock(rdev, sh->sector,
				       STRIPE_SECTORS,
1889
				       &first_bad, &bad_sectors)) {
1890
			set_bit(R5_MadeGood, &sh->dev[i].flags);
1891 1892 1893 1894 1895 1896 1897
			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);
		}
1898 1899
	}
	rdev_dec_pending(rdev, conf->mddev);
L
Linus Torvalds 已提交
1900

1901 1902
	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
1903
	set_bit(STRIPE_HANDLE, &sh->state);
1904
	release_stripe(sh);
L
Linus Torvalds 已提交
1905 1906
}

1907
static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous);
L
Linus Torvalds 已提交
1908
	
1909
static void raid5_build_block(struct stripe_head *sh, int i, int previous)
L
Linus Torvalds 已提交
1910 1911 1912 1913 1914 1915 1916 1917
{
	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;
1918
	dev->vec.bv_page = dev->page;
L
Linus Torvalds 已提交
1919

1920 1921 1922 1923 1924 1925 1926
	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
1927
	dev->flags = 0;
1928
	dev->sector = compute_blocknr(sh, i, previous);
L
Linus Torvalds 已提交
1929 1930
}

1931
static void error(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
1932 1933
{
	char b[BDEVNAME_SIZE];
1934
	struct r5conf *conf = mddev->private;
1935
	unsigned long flags;
1936
	pr_debug("raid456: error called\n");
L
Linus Torvalds 已提交
1937

1938 1939 1940 1941 1942 1943
	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);

1944
	set_bit(Blocked, &rdev->flags);
1945 1946 1947 1948 1949 1950 1951 1952 1953
	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);
1954
}
L
Linus Torvalds 已提交
1955 1956 1957 1958 1959

/*
 * Input: a 'big' sector number,
 * Output: index of the data and parity disk, and the sector # in them.
 */
1960
static sector_t raid5_compute_sector(struct r5conf *conf, sector_t r_sector,
1961 1962
				     int previous, int *dd_idx,
				     struct stripe_head *sh)
L
Linus Torvalds 已提交
1963
{
N
NeilBrown 已提交
1964
	sector_t stripe, stripe2;
1965
	sector_t chunk_number;
L
Linus Torvalds 已提交
1966
	unsigned int chunk_offset;
1967
	int pd_idx, qd_idx;
1968
	int ddf_layout = 0;
L
Linus Torvalds 已提交
1969
	sector_t new_sector;
1970 1971
	int algorithm = previous ? conf->prev_algo
				 : conf->algorithm;
1972 1973
	int sectors_per_chunk = previous ? conf->prev_chunk_sectors
					 : conf->chunk_sectors;
1974 1975 1976
	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
1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988

	/* 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
	 */
1989 1990
	stripe = chunk_number;
	*dd_idx = sector_div(stripe, data_disks);
N
NeilBrown 已提交
1991
	stripe2 = stripe;
L
Linus Torvalds 已提交
1992 1993 1994
	/*
	 * Select the parity disk based on the user selected algorithm.
	 */
1995
	pd_idx = qd_idx = -1;
1996 1997
	switch(conf->level) {
	case 4:
1998
		pd_idx = data_disks;
1999 2000
		break;
	case 5:
2001
		switch (algorithm) {
L
Linus Torvalds 已提交
2002
		case ALGORITHM_LEFT_ASYMMETRIC:
N
NeilBrown 已提交
2003
			pd_idx = data_disks - sector_div(stripe2, raid_disks);
2004
			if (*dd_idx >= pd_idx)
L
Linus Torvalds 已提交
2005 2006 2007
				(*dd_idx)++;
			break;
		case ALGORITHM_RIGHT_ASYMMETRIC:
N
NeilBrown 已提交
2008
			pd_idx = sector_div(stripe2, raid_disks);
2009
			if (*dd_idx >= pd_idx)
L
Linus Torvalds 已提交
2010 2011 2012
				(*dd_idx)++;
			break;
		case ALGORITHM_LEFT_SYMMETRIC:
N
NeilBrown 已提交
2013
			pd_idx = data_disks - sector_div(stripe2, raid_disks);
2014
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
L
Linus Torvalds 已提交
2015 2016
			break;
		case ALGORITHM_RIGHT_SYMMETRIC:
N
NeilBrown 已提交
2017
			pd_idx = sector_div(stripe2, raid_disks);
2018
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
L
Linus Torvalds 已提交
2019
			break;
2020 2021 2022 2023 2024 2025 2026
		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
2027
		default:
2028
			BUG();
2029 2030 2031 2032
		}
		break;
	case 6:

2033
		switch (algorithm) {
2034
		case ALGORITHM_LEFT_ASYMMETRIC:
N
NeilBrown 已提交
2035
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2036 2037
			qd_idx = pd_idx + 1;
			if (pd_idx == raid_disks-1) {
2038
				(*dd_idx)++;	/* Q D D D P */
2039 2040
				qd_idx = 0;
			} else if (*dd_idx >= pd_idx)
2041 2042 2043
				(*dd_idx) += 2; /* D D P Q D */
			break;
		case ALGORITHM_RIGHT_ASYMMETRIC:
N
NeilBrown 已提交
2044
			pd_idx = sector_div(stripe2, raid_disks);
2045 2046
			qd_idx = pd_idx + 1;
			if (pd_idx == raid_disks-1) {
2047
				(*dd_idx)++;	/* Q D D D P */
2048 2049
				qd_idx = 0;
			} else if (*dd_idx >= pd_idx)
2050 2051 2052
				(*dd_idx) += 2; /* D D P Q D */
			break;
		case ALGORITHM_LEFT_SYMMETRIC:
N
NeilBrown 已提交
2053
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2054 2055
			qd_idx = (pd_idx + 1) % raid_disks;
			*dd_idx = (pd_idx + 2 + *dd_idx) % raid_disks;
2056 2057
			break;
		case ALGORITHM_RIGHT_SYMMETRIC:
N
NeilBrown 已提交
2058
			pd_idx = sector_div(stripe2, raid_disks);
2059 2060
			qd_idx = (pd_idx + 1) % raid_disks;
			*dd_idx = (pd_idx + 2 + *dd_idx) % raid_disks;
2061
			break;
2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076

		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
2077
			pd_idx = sector_div(stripe2, raid_disks);
2078 2079 2080 2081 2082 2083
			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 */
2084
			ddf_layout = 1;
2085 2086 2087 2088 2089 2090 2091
			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
2092 2093
			stripe2 += 1;
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2094 2095 2096 2097 2098 2099
			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 */
2100
			ddf_layout = 1;
2101 2102 2103 2104
			break;

		case ALGORITHM_ROTATING_N_CONTINUE:
			/* Same as left_symmetric but Q is before P */
N
NeilBrown 已提交
2105
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2106 2107
			qd_idx = (pd_idx + raid_disks - 1) % raid_disks;
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
2108
			ddf_layout = 1;
2109 2110 2111 2112
			break;

		case ALGORITHM_LEFT_ASYMMETRIC_6:
			/* RAID5 left_asymmetric, with Q on last device */
N
NeilBrown 已提交
2113
			pd_idx = data_disks - sector_div(stripe2, raid_disks-1);
2114 2115 2116 2117 2118 2119
			if (*dd_idx >= pd_idx)
				(*dd_idx)++;
			qd_idx = raid_disks - 1;
			break;

		case ALGORITHM_RIGHT_ASYMMETRIC_6:
N
NeilBrown 已提交
2120
			pd_idx = sector_div(stripe2, raid_disks-1);
2121 2122 2123 2124 2125 2126
			if (*dd_idx >= pd_idx)
				(*dd_idx)++;
			qd_idx = raid_disks - 1;
			break;

		case ALGORITHM_LEFT_SYMMETRIC_6:
N
NeilBrown 已提交
2127
			pd_idx = data_disks - sector_div(stripe2, raid_disks-1);
2128 2129 2130 2131 2132
			*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
2133
			pd_idx = sector_div(stripe2, raid_disks-1);
2134 2135 2136 2137 2138 2139 2140 2141 2142 2143
			*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;

2144
		default:
2145
			BUG();
2146 2147
		}
		break;
L
Linus Torvalds 已提交
2148 2149
	}

2150 2151 2152
	if (sh) {
		sh->pd_idx = pd_idx;
		sh->qd_idx = qd_idx;
2153
		sh->ddf_layout = ddf_layout;
2154
	}
L
Linus Torvalds 已提交
2155 2156 2157 2158 2159 2160 2161 2162
	/*
	 * Finally, compute the new sector number
	 */
	new_sector = (sector_t)stripe * sectors_per_chunk + chunk_offset;
	return new_sector;
}


2163
static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous)
L
Linus Torvalds 已提交
2164
{
2165
	struct r5conf *conf = sh->raid_conf;
2166 2167
	int raid_disks = sh->disks;
	int data_disks = raid_disks - conf->max_degraded;
L
Linus Torvalds 已提交
2168
	sector_t new_sector = sh->sector, check;
2169 2170
	int sectors_per_chunk = previous ? conf->prev_chunk_sectors
					 : conf->chunk_sectors;
2171 2172
	int algorithm = previous ? conf->prev_algo
				 : conf->algorithm;
L
Linus Torvalds 已提交
2173 2174
	sector_t stripe;
	int chunk_offset;
2175 2176
	sector_t chunk_number;
	int dummy1, dd_idx = i;
L
Linus Torvalds 已提交
2177
	sector_t r_sector;
2178
	struct stripe_head sh2;
L
Linus Torvalds 已提交
2179

2180

L
Linus Torvalds 已提交
2181 2182 2183
	chunk_offset = sector_div(new_sector, sectors_per_chunk);
	stripe = new_sector;

2184 2185 2186 2187 2188
	if (i == sh->pd_idx)
		return 0;
	switch(conf->level) {
	case 4: break;
	case 5:
2189
		switch (algorithm) {
L
Linus Torvalds 已提交
2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200
		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;
2201 2202 2203 2204 2205
		case ALGORITHM_PARITY_0:
			i -= 1;
			break;
		case ALGORITHM_PARITY_N:
			break;
L
Linus Torvalds 已提交
2206
		default:
2207
			BUG();
2208 2209 2210
		}
		break;
	case 6:
2211
		if (i == sh->qd_idx)
2212
			return 0; /* It is the Q disk */
2213
		switch (algorithm) {
2214 2215
		case ALGORITHM_LEFT_ASYMMETRIC:
		case ALGORITHM_RIGHT_ASYMMETRIC:
2216 2217 2218 2219
		case ALGORITHM_ROTATING_ZERO_RESTART:
		case ALGORITHM_ROTATING_N_RESTART:
			if (sh->pd_idx == raid_disks-1)
				i--;	/* Q D D D P */
2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233
			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;
2234 2235 2236 2237 2238 2239
		case ALGORITHM_PARITY_0:
			i -= 2;
			break;
		case ALGORITHM_PARITY_N:
			break;
		case ALGORITHM_ROTATING_N_CONTINUE:
2240
			/* Like left_symmetric, but P is before Q */
2241 2242
			if (sh->pd_idx == 0)
				i--;	/* P D D D Q */
2243 2244 2245 2246 2247 2248
			else {
				/* D D Q P D */
				if (i < sh->pd_idx)
					i += raid_disks;
				i -= (sh->pd_idx + 1);
			}
2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263
			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;
2264
		default:
2265
			BUG();
2266 2267
		}
		break;
L
Linus Torvalds 已提交
2268 2269 2270
	}

	chunk_number = stripe * data_disks + i;
2271
	r_sector = chunk_number * sectors_per_chunk + chunk_offset;
L
Linus Torvalds 已提交
2272

2273
	check = raid5_compute_sector(conf, r_sector,
2274
				     previous, &dummy1, &sh2);
2275 2276
	if (check != sh->sector || dummy1 != dd_idx || sh2.pd_idx != sh->pd_idx
		|| sh2.qd_idx != sh->qd_idx) {
2277 2278
		printk(KERN_ERR "md/raid:%s: compute_blocknr: map not correct\n",
		       mdname(conf->mddev));
L
Linus Torvalds 已提交
2279 2280 2281 2282 2283 2284
		return 0;
	}
	return r_sector;
}


2285
static void
2286
schedule_reconstruction(struct stripe_head *sh, struct stripe_head_state *s,
2287
			 int rcw, int expand)
2288 2289
{
	int i, pd_idx = sh->pd_idx, disks = sh->disks;
2290
	struct r5conf *conf = sh->raid_conf;
2291
	int level = conf->level;
2292 2293 2294 2295 2296 2297 2298 2299

	if (rcw) {

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

			if (dev->towrite) {
				set_bit(R5_LOCKED, &dev->flags);
2300
				set_bit(R5_Wantdrain, &dev->flags);
2301 2302
				if (!expand)
					clear_bit(R5_UPTODATE, &dev->flags);
2303
				s->locked++;
2304 2305
			}
		}
2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320
		/* 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);

2321
		if (s->locked + conf->max_degraded == disks)
2322
			if (!test_and_set_bit(STRIPE_FULL_WRITE, &sh->state))
2323
				atomic_inc(&conf->pending_full_writes);
2324
	} else {
2325
		BUG_ON(level == 6);
2326 2327 2328 2329 2330 2331 2332 2333 2334 2335
		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) ||
2336 2337
			     test_bit(R5_Wantcompute, &dev->flags))) {
				set_bit(R5_Wantdrain, &dev->flags);
2338 2339
				set_bit(R5_LOCKED, &dev->flags);
				clear_bit(R5_UPTODATE, &dev->flags);
2340
				s->locked++;
2341 2342
			}
		}
2343 2344 2345 2346 2347 2348 2349
		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);
2350 2351
	}

2352
	/* keep the parity disk(s) locked while asynchronous operations
2353 2354 2355 2356
	 * are in flight
	 */
	set_bit(R5_LOCKED, &sh->dev[pd_idx].flags);
	clear_bit(R5_UPTODATE, &sh->dev[pd_idx].flags);
2357
	s->locked++;
2358

2359 2360 2361 2362 2363 2364 2365 2366 2367
	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++;
	}

2368
	pr_debug("%s: stripe %llu locked: %d ops_request: %lx\n",
2369
		__func__, (unsigned long long)sh->sector,
2370
		s->locked, s->ops_request);
2371
}
2372

L
Linus Torvalds 已提交
2373 2374
/*
 * Each stripe/dev can have one or more bion attached.
2375
 * toread/towrite point to the first in a chain.
L
Linus Torvalds 已提交
2376 2377 2378 2379 2380
 * 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;
2381
	struct r5conf *conf = sh->raid_conf;
2382
	int firstwrite=0;
L
Linus Torvalds 已提交
2383

2384
	pr_debug("adding bi b#%llu to stripe s#%llu\n",
L
Linus Torvalds 已提交
2385 2386 2387
		(unsigned long long)bi->bi_sector,
		(unsigned long long)sh->sector);

S
Shaohua Li 已提交
2388 2389 2390 2391 2392 2393 2394 2395 2396
	/*
	 * 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);
2397
	if (forwrite) {
L
Linus Torvalds 已提交
2398
		bip = &sh->dev[dd_idx].towrite;
2399
		if (*bip == NULL)
2400 2401
			firstwrite = 1;
	} else
L
Linus Torvalds 已提交
2402 2403
		bip = &sh->dev[dd_idx].toread;
	while (*bip && (*bip)->bi_sector < bi->bi_sector) {
K
Kent Overstreet 已提交
2404
		if (bio_end_sector(*bip) > bi->bi_sector)
L
Linus Torvalds 已提交
2405 2406 2407
			goto overlap;
		bip = & (*bip)->bi_next;
	}
K
Kent Overstreet 已提交
2408
	if (*bip && (*bip)->bi_sector < bio_end_sector(bi))
L
Linus Torvalds 已提交
2409 2410
		goto overlap;

2411
	BUG_ON(*bip && bi->bi_next && (*bip) != bi->bi_next);
L
Linus Torvalds 已提交
2412 2413 2414
	if (*bip)
		bi->bi_next = *bip;
	*bip = bi;
2415
	raid5_inc_bi_active_stripes(bi);
2416

L
Linus Torvalds 已提交
2417 2418 2419 2420 2421 2422 2423
	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
2424 2425
			if (bio_end_sector(bi) >= sector)
				sector = bio_end_sector(bi);
L
Linus Torvalds 已提交
2426 2427 2428 2429
		}
		if (sector >= sh->dev[dd_idx].sector + STRIPE_SECTORS)
			set_bit(R5_OVERWRITE, &sh->dev[dd_idx].flags);
	}
2430 2431 2432 2433

	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);
2434
	spin_unlock_irq(&sh->stripe_lock);
2435 2436 2437 2438 2439 2440 2441

	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
2442 2443 2444 2445
	return 1;

 overlap:
	set_bit(R5_Overlap, &sh->dev[dd_idx].flags);
S
Shaohua Li 已提交
2446
	spin_unlock_irq(&sh->stripe_lock);
L
Linus Torvalds 已提交
2447 2448 2449
	return 0;
}

2450
static void end_reshape(struct r5conf *conf);
2451

2452
static void stripe_set_idx(sector_t stripe, struct r5conf *conf, int previous,
2453
			    struct stripe_head *sh)
2454
{
2455
	int sectors_per_chunk =
2456
		previous ? conf->prev_chunk_sectors : conf->chunk_sectors;
2457
	int dd_idx;
2458
	int chunk_offset = sector_div(stripe, sectors_per_chunk);
2459
	int disks = previous ? conf->previous_raid_disks : conf->raid_disks;
2460

2461 2462
	raid5_compute_sector(conf,
			     stripe * (disks - conf->max_degraded)
2463
			     *sectors_per_chunk + chunk_offset,
2464
			     previous,
2465
			     &dd_idx, sh);
2466 2467
}

2468
static void
2469
handle_failed_stripe(struct r5conf *conf, struct stripe_head *sh,
2470 2471 2472 2473 2474 2475 2476 2477 2478
				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)) {
2479
			struct md_rdev *rdev;
2480 2481 2482
			rcu_read_lock();
			rdev = rcu_dereference(conf->disks[i].rdev);
			if (rdev && test_bit(In_sync, &rdev->flags))
2483 2484 2485
				atomic_inc(&rdev->nr_pending);
			else
				rdev = NULL;
2486
			rcu_read_unlock();
2487 2488 2489 2490 2491 2492 2493 2494
			if (rdev) {
				if (!rdev_set_badblocks(
					    rdev,
					    sh->sector,
					    STRIPE_SECTORS, 0))
					md_error(conf->mddev, rdev);
				rdev_dec_pending(rdev, conf->mddev);
			}
2495
		}
S
Shaohua Li 已提交
2496
		spin_lock_irq(&sh->stripe_lock);
2497 2498 2499
		/* fail all writes first */
		bi = sh->dev[i].towrite;
		sh->dev[i].towrite = NULL;
S
Shaohua Li 已提交
2500
		spin_unlock_irq(&sh->stripe_lock);
2501
		if (bi)
2502 2503 2504 2505 2506 2507 2508 2509 2510
			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);
2511
			if (!raid5_dec_bi_active_stripes(bi)) {
2512 2513 2514 2515 2516 2517
				md_write_end(conf->mddev);
				bi->bi_next = *return_bi;
				*return_bi = bi;
			}
			bi = nextbi;
		}
2518 2519 2520 2521
		if (bitmap_end)
			bitmap_endwrite(conf->mddev->bitmap, sh->sector,
				STRIPE_SECTORS, 0, 0);
		bitmap_end = 0;
2522 2523 2524 2525 2526 2527 2528 2529
		/* 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);
2530
			if (!raid5_dec_bi_active_stripes(bi)) {
2531 2532 2533 2534 2535 2536 2537
				md_write_end(conf->mddev);
				bi->bi_next = *return_bi;
				*return_bi = bi;
			}
			bi = bi2;
		}

2538 2539 2540 2541 2542 2543
		/* 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))) {
2544
			spin_lock_irq(&sh->stripe_lock);
2545 2546
			bi = sh->dev[i].toread;
			sh->dev[i].toread = NULL;
2547
			spin_unlock_irq(&sh->stripe_lock);
2548 2549 2550 2551 2552 2553 2554
			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);
2555
				if (!raid5_dec_bi_active_stripes(bi)) {
2556 2557 2558 2559 2560 2561 2562 2563 2564
					bi->bi_next = *return_bi;
					*return_bi = bi;
				}
				bi = nextbi;
			}
		}
		if (bitmap_end)
			bitmap_endwrite(conf->mddev->bitmap, sh->sector,
					STRIPE_SECTORS, 0, 0);
2565 2566 2567 2568
		/* 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);
2569 2570
	}

2571 2572 2573
	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);
2574 2575
}

2576
static void
2577
handle_failed_sync(struct r5conf *conf, struct stripe_head *sh,
2578 2579 2580 2581 2582 2583
		   struct stripe_head_state *s)
{
	int abort = 0;
	int i;

	clear_bit(STRIPE_SYNCING, &sh->state);
2584 2585
	if (test_and_clear_bit(R5_Overlap, &sh->dev[sh->pd_idx].flags))
		wake_up(&conf->wait_for_overlap);
2586
	s->syncing = 0;
2587
	s->replacing = 0;
2588
	/* There is nothing more to do for sync/check/repair.
2589 2590 2591
	 * 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.
2592
	 * For recover/replace we need to record a bad block on all
2593 2594
	 * non-sync devices, or abort the recovery
	 */
2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617
	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;
2618
	}
2619
	md_done_sync(conf->mddev, STRIPE_SECTORS, !abort);
2620 2621
}

2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637
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;
}

2638
/* fetch_block - checks the given member device to see if its data needs
2639 2640 2641
 * to be read or computed to satisfy a request.
 *
 * Returns 1 when no more member devices need to be checked, otherwise returns
2642
 * 0 to tell the loop in handle_stripe_fill to continue
2643
 */
2644 2645
static int fetch_block(struct stripe_head *sh, struct stripe_head_state *s,
		       int disk_idx, int disks)
2646
{
2647
	struct r5dev *dev = &sh->dev[disk_idx];
2648 2649
	struct r5dev *fdev[2] = { &sh->dev[s->failed_num[0]],
				  &sh->dev[s->failed_num[1]] };
2650

2651
	/* is the data in this block needed, and can we get it? */
2652 2653 2654 2655 2656
	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 ||
2657
	     (s->replacing && want_replace(sh, disk_idx)) ||
2658 2659
	     (s->failed >= 1 && fdev[0]->toread) ||
	     (s->failed >= 2 && fdev[1]->toread) ||
2660 2661 2662
	     (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))) {
2663 2664 2665 2666 2667 2668
		/* 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) &&
2669 2670
		    (s->failed && (disk_idx == s->failed_num[0] ||
				   disk_idx == s->failed_num[1]))) {
2671 2672
			/* have disk failed, and we're requested to fetch it;
			 * do compute it
2673
			 */
2674 2675 2676 2677 2678 2679 2680 2681
			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;
2682 2683 2684 2685 2686 2687
			/* 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.
			 */
2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700
			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;
2701
			}
2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720
			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);
2721 2722
		}
	}
2723 2724 2725 2726 2727

	return 0;
}

/**
2728
 * handle_stripe_fill - read or compute data to satisfy pending requests.
2729
 */
2730 2731 2732
static void handle_stripe_fill(struct stripe_head *sh,
			       struct stripe_head_state *s,
			       int disks)
2733 2734 2735 2736 2737 2738 2739 2740 2741 2742
{
	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--; )
2743
			if (fetch_block(sh, s, i, disks))
2744
				break;
2745 2746 2747 2748
	set_bit(STRIPE_HANDLE, &sh->state);
}


2749
/* handle_stripe_clean_event
2750 2751 2752 2753
 * 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.
 */
2754
static void handle_stripe_clean_event(struct r5conf *conf,
2755 2756 2757 2758
	struct stripe_head *sh, int disks, struct bio **return_bi)
{
	int i;
	struct r5dev *dev;
2759
	int discard_pending = 0;
2760 2761 2762 2763 2764

	for (i = disks; i--; )
		if (sh->dev[i].written) {
			dev = &sh->dev[i];
			if (!test_bit(R5_LOCKED, &dev->flags) &&
2765
			    (test_bit(R5_UPTODATE, &dev->flags) ||
2766
			     test_bit(R5_Discard, &dev->flags))) {
2767 2768
				/* We can return any write requests */
				struct bio *wbi, *wbi2;
2769
				pr_debug("Return write for disc %d\n", i);
2770 2771
				if (test_and_clear_bit(R5_Discard, &dev->flags))
					clear_bit(R5_UPTODATE, &dev->flags);
2772 2773 2774 2775 2776
				wbi = dev->written;
				dev->written = NULL;
				while (wbi && wbi->bi_sector <
					dev->sector + STRIPE_SECTORS) {
					wbi2 = r5_next_bio(wbi, dev->sector);
2777
					if (!raid5_dec_bi_active_stripes(wbi)) {
2778 2779 2780 2781 2782 2783
						md_write_end(conf->mddev);
						wbi->bi_next = *return_bi;
						*return_bi = wbi;
					}
					wbi = wbi2;
				}
2784 2785
				bitmap_endwrite(conf->mddev->bitmap, sh->sector,
						STRIPE_SECTORS,
2786
					 !test_bit(STRIPE_DEGRADED, &sh->state),
2787
						0);
2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804
			} 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);

	}
2805 2806 2807 2808

	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);
2809 2810
}

2811
static void handle_stripe_dirtying(struct r5conf *conf,
2812 2813 2814
				   struct stripe_head *sh,
				   struct stripe_head_state *s,
				   int disks)
2815 2816
{
	int rmw = 0, rcw = 0, i;
2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829
	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
2830 2831 2832
		 * look like rcw is cheaper
		 */
		rcw = 1; rmw = 2;
2833 2834 2835
		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);
2836
	} else for (i = disks; i--; ) {
2837 2838 2839 2840
		/* 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) &&
2841 2842
		    !(test_bit(R5_UPTODATE, &dev->flags) ||
		      test_bit(R5_Wantcompute, &dev->flags))) {
2843 2844 2845 2846 2847 2848 2849 2850
			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) &&
2851 2852 2853
		    !(test_bit(R5_UPTODATE, &dev->flags) ||
		    test_bit(R5_Wantcompute, &dev->flags))) {
			if (test_bit(R5_Insync, &dev->flags)) rcw++;
2854 2855 2856 2857
			else
				rcw += 2*disks;
		}
	}
2858
	pr_debug("for sector %llu, rmw=%d rcw=%d\n",
2859 2860
		(unsigned long long)sh->sector, rmw, rcw);
	set_bit(STRIPE_HANDLE, &sh->state);
N
NeilBrown 已提交
2861
	if (rmw < rcw && rmw > 0) {
2862
		/* prefer read-modify-write, but need to get some data */
2863 2864 2865 2866
		if (conf->mddev->queue)
			blk_add_trace_msg(conf->mddev->queue,
					  "raid5 rmw %llu %d",
					  (unsigned long long)sh->sector, rmw);
2867 2868 2869 2870
		for (i = disks; i--; ) {
			struct r5dev *dev = &sh->dev[i];
			if ((dev->towrite || i == sh->pd_idx) &&
			    !test_bit(R5_LOCKED, &dev->flags) &&
2871 2872
			    !(test_bit(R5_UPTODATE, &dev->flags) ||
			    test_bit(R5_Wantcompute, &dev->flags)) &&
2873 2874 2875
			    test_bit(R5_Insync, &dev->flags)) {
				if (
				  test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
2876
					pr_debug("Read_old block "
N
NeilBrown 已提交
2877
						 "%d for r-m-w\n", i);
2878 2879 2880 2881 2882 2883 2884 2885 2886
					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
2887
	}
2888
	if (rcw <= rmw && rcw > 0) {
2889
		/* want reconstruct write, but need to get some data */
N
NeilBrown 已提交
2890
		int qread =0;
2891
		rcw = 0;
2892 2893 2894
		for (i = disks; i--; ) {
			struct r5dev *dev = &sh->dev[i];
			if (!test_bit(R5_OVERWRITE, &dev->flags) &&
2895
			    i != sh->pd_idx && i != sh->qd_idx &&
2896
			    !test_bit(R5_LOCKED, &dev->flags) &&
2897
			    !(test_bit(R5_UPTODATE, &dev->flags) ||
2898 2899 2900 2901
			      test_bit(R5_Wantcompute, &dev->flags))) {
				rcw++;
				if (!test_bit(R5_Insync, &dev->flags))
					continue; /* it's a failed drive */
2902 2903
				if (
				  test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
2904
					pr_debug("Read_old block "
2905 2906 2907 2908
						"%d for Reconstruct\n", i);
					set_bit(R5_LOCKED, &dev->flags);
					set_bit(R5_Wantread, &dev->flags);
					s->locked++;
N
NeilBrown 已提交
2909
					qread++;
2910 2911 2912 2913 2914 2915
				} else {
					set_bit(STRIPE_DELAYED, &sh->state);
					set_bit(STRIPE_HANDLE, &sh->state);
				}
			}
		}
2916
		if (rcw && conf->mddev->queue)
N
NeilBrown 已提交
2917 2918 2919
			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));
2920
	}
2921 2922 2923
	/* now if nothing is locked, and if we have enough data,
	 * we can start a write request
	 */
2924 2925
	/* 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
2926 2927
	 * subsequent call wants to start a write request.  raid_run_ops only
	 * handles the case where compute block and reconstruct are requested
2928 2929 2930
	 * simultaneously.  If this is not the case then new writes need to be
	 * held off until the compute completes.
	 */
2931 2932 2933
	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)))
2934
		schedule_reconstruction(sh, s, rcw == 0, 0);
2935 2936
}

2937
static void handle_parity_checks5(struct r5conf *conf, struct stripe_head *sh,
2938 2939
				struct stripe_head_state *s, int disks)
{
2940
	struct r5dev *dev = NULL;
2941

2942
	set_bit(STRIPE_HANDLE, &sh->state);
2943

2944 2945 2946
	switch (sh->check_state) {
	case check_state_idle:
		/* start a new check operation if there are no failures */
2947 2948
		if (s->failed == 0) {
			BUG_ON(s->uptodate != disks);
2949 2950
			sh->check_state = check_state_run;
			set_bit(STRIPE_OP_CHECK, &s->ops_request);
2951 2952
			clear_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags);
			s->uptodate--;
2953
			break;
2954
		}
2955
		dev = &sh->dev[s->failed_num[0]];
2956 2957 2958 2959 2960 2961 2962 2963 2964
		/* 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
2965

2966 2967 2968 2969 2970
		/* 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);
2971
		s->locked++;
2972
		set_bit(R5_Wantwrite, &dev->flags);
2973

2974 2975
		clear_bit(STRIPE_DEGRADED, &sh->state);
		set_bit(STRIPE_INSYNC, &sh->state);
2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991
		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
2992
		if ((sh->ops.zero_sum_result & SUM_CHECK_P_RESULT) == 0)
2993 2994 2995 2996 2997
			/* parity is correct (on disc,
			 * not in buffer any more)
			 */
			set_bit(STRIPE_INSYNC, &sh->state);
		else {
2998
			atomic64_add(STRIPE_SECTORS, &conf->mddev->resync_mismatches);
2999 3000 3001 3002 3003
			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;
3004
				set_bit(STRIPE_COMPUTE_RUN, &sh->state);
3005 3006 3007 3008
				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;
3009
				sh->ops.target2 = -1;
3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020
				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();
3021 3022 3023 3024
	}
}


3025
static void handle_parity_checks6(struct r5conf *conf, struct stripe_head *sh,
3026
				  struct stripe_head_state *s,
3027
				  int disks)
3028 3029
{
	int pd_idx = sh->pd_idx;
N
NeilBrown 已提交
3030
	int qd_idx = sh->qd_idx;
3031
	struct r5dev *dev;
3032 3033 3034 3035

	set_bit(STRIPE_HANDLE, &sh->state);

	BUG_ON(s->failed > 2);
3036

3037 3038 3039 3040 3041 3042
	/* 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
	 */

3043 3044 3045
	switch (sh->check_state) {
	case check_state_idle:
		/* start a new check operation if there are < 2 failures */
3046
		if (s->failed == s->q_failed) {
3047
			/* The only possible failed device holds Q, so it
3048 3049 3050
			 * makes sense to check P (If anything else were failed,
			 * we would have used P to recreate it).
			 */
3051
			sh->check_state = check_state_run;
3052
		}
3053
		if (!s->q_failed && s->failed < 2) {
3054
			/* Q is not failed, and we didn't use it to generate
3055 3056
			 * anything, so it makes sense to check it
			 */
3057 3058 3059 3060
			if (sh->check_state == check_state_run)
				sh->check_state = check_state_run_pq;
			else
				sh->check_state = check_state_run_q;
3061 3062
		}

3063 3064
		/* discard potentially stale zero_sum_result */
		sh->ops.zero_sum_result = 0;
3065

3066 3067 3068 3069
		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--;
3070
		}
3071 3072 3073 3074 3075 3076 3077
		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;
3078 3079
		}

3080 3081 3082 3083 3084
		/* we have 2-disk failure */
		BUG_ON(s->failed != 2);
		/* fall through */
	case check_state_compute_result:
		sh->check_state = check_state_idle;
3085

3086 3087 3088
		/* check that a write has not made the stripe insync */
		if (test_bit(STRIPE_INSYNC, &sh->state))
			break;
3089 3090

		/* now write out any block on a failed drive,
3091
		 * or P or Q if they were recomputed
3092
		 */
3093
		BUG_ON(s->uptodate < disks - 1); /* We don't need Q to recover */
3094
		if (s->failed == 2) {
3095
			dev = &sh->dev[s->failed_num[1]];
3096 3097 3098 3099 3100
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
		if (s->failed >= 1) {
3101
			dev = &sh->dev[s->failed_num[0]];
3102 3103 3104 3105
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
3106
		if (sh->ops.zero_sum_result & SUM_CHECK_P_RESULT) {
3107 3108 3109 3110 3111
			dev = &sh->dev[pd_idx];
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
3112
		if (sh->ops.zero_sum_result & SUM_CHECK_Q_RESULT) {
3113 3114 3115 3116 3117 3118 3119 3120
			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);
3121 3122 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
		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 {
3150
			atomic64_add(STRIPE_SECTORS, &conf->mddev->resync_mismatches);
3151 3152 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
			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();
3185 3186 3187
	}
}

3188
static void handle_stripe_expansion(struct r5conf *conf, struct stripe_head *sh)
3189 3190 3191 3192 3193 3194
{
	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.
	 */
3195
	struct dma_async_tx_descriptor *tx = NULL;
3196 3197
	clear_bit(STRIPE_EXPAND_SOURCE, &sh->state);
	for (i = 0; i < sh->disks; i++)
N
NeilBrown 已提交
3198
		if (i != sh->pd_idx && i != sh->qd_idx) {
3199
			int dd_idx, j;
3200
			struct stripe_head *sh2;
3201
			struct async_submit_ctl submit;
3202

3203
			sector_t bn = compute_blocknr(sh, i, 1);
3204 3205
			sector_t s = raid5_compute_sector(conf, bn, 0,
							  &dd_idx, NULL);
3206
			sh2 = get_active_stripe(conf, s, 0, 1, 1);
3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218
			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;
			}
3219 3220

			/* place all the copies on one channel */
3221
			init_async_submit(&submit, 0, tx, NULL, NULL, NULL);
3222
			tx = async_memcpy(sh2->dev[dd_idx].page,
3223
					  sh->dev[i].page, 0, 0, STRIPE_SIZE,
3224
					  &submit);
3225

3226 3227 3228 3229
			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 &&
3230
				    j != sh2->qd_idx &&
3231 3232 3233 3234 3235 3236 3237
				    !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);
3238

3239
		}
3240
	/* done submitting copies, wait for them to complete */
3241
	async_tx_quiesce(&tx);
3242
}
L
Linus Torvalds 已提交
3243 3244 3245 3246

/*
 * handle_stripe - do things to a stripe.
 *
3247 3248
 * 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
3249
 * Possible results:
3250 3251
 *    return some read requests which now have data
 *    return some write requests which are safely on storage
L
Linus Torvalds 已提交
3252 3253 3254 3255 3256
 *    schedule a read on some buffers
 *    schedule a write of some buffers
 *    return confirmation of parity correctness
 *
 */
3257

3258
static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
L
Linus Torvalds 已提交
3259
{
3260
	struct r5conf *conf = sh->raid_conf;
3261
	int disks = sh->disks;
3262 3263
	struct r5dev *dev;
	int i;
3264
	int do_recovery = 0;
L
Linus Torvalds 已提交
3265

3266 3267 3268 3269 3270 3271
	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
3272

3273
	/* Now to look around and see what can be done */
L
Linus Torvalds 已提交
3274
	rcu_read_lock();
3275
	for (i=disks; i--; ) {
3276
		struct md_rdev *rdev;
3277 3278 3279
		sector_t first_bad;
		int bad_sectors;
		int is_bad = 0;
3280

3281
		dev = &sh->dev[i];
L
Linus Torvalds 已提交
3282

3283
		pr_debug("check %d: state 0x%lx read %p write %p written %p\n",
3284 3285
			 i, dev->flags,
			 dev->toread, dev->towrite, dev->written);
3286 3287 3288 3289 3290 3291 3292 3293
		/* 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
3294

3295
		/* now count some things */
3296 3297 3298 3299
		if (test_bit(R5_LOCKED, &dev->flags))
			s->locked++;
		if (test_bit(R5_UPTODATE, &dev->flags))
			s->uptodate++;
3300
		if (test_bit(R5_Wantcompute, &dev->flags)) {
3301 3302
			s->compute++;
			BUG_ON(s->compute > 2);
3303
		}
L
Linus Torvalds 已提交
3304

3305
		if (test_bit(R5_Wantfill, &dev->flags))
3306
			s->to_fill++;
3307
		else if (dev->toread)
3308
			s->to_read++;
3309
		if (dev->towrite) {
3310
			s->to_write++;
3311
			if (!test_bit(R5_OVERWRITE, &dev->flags))
3312
				s->non_overwrite++;
3313
		}
3314
		if (dev->written)
3315
			s->written++;
3316 3317 3318 3319 3320 3321 3322 3323 3324 3325
		/* 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 {
3326 3327
			if (rdev)
				set_bit(R5_NeedReplace, &dev->flags);
3328 3329 3330
			rdev = rcu_dereference(conf->disks[i].rdev);
			clear_bit(R5_ReadRepl, &dev->flags);
		}
3331 3332
		if (rdev && test_bit(Faulty, &rdev->flags))
			rdev = NULL;
3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344
		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);
			}
3345
		}
3346 3347 3348
		clear_bit(R5_Insync, &dev->flags);
		if (!rdev)
			/* Not in-sync */;
3349 3350
		else if (is_bad) {
			/* also not in-sync */
3351 3352
			if (!test_bit(WriteErrorSeen, &rdev->flags) &&
			    test_bit(R5_UPTODATE, &dev->flags)) {
3353 3354 3355 3356 3357 3358 3359
				/* 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))
3360
			set_bit(R5_Insync, &dev->flags);
3361
		else if (sh->sector + STRIPE_SECTORS <= rdev->recovery_offset)
3362
			/* in sync if before recovery_offset */
3363 3364 3365 3366 3367 3368 3369 3370 3371
			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
3372
		if (rdev && test_bit(R5_WriteError, &dev->flags)) {
3373 3374 3375 3376 3377 3378 3379
			/* 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)) {
3380
				s->handle_bad_blocks = 1;
3381
				atomic_inc(&rdev2->nr_pending);
3382 3383 3384
			} else
				clear_bit(R5_WriteError, &dev->flags);
		}
A
Adam Kwolek 已提交
3385
		if (rdev && test_bit(R5_MadeGood, &dev->flags)) {
3386 3387 3388 3389 3390
			/* 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)) {
3391
				s->handle_bad_blocks = 1;
3392
				atomic_inc(&rdev2->nr_pending);
3393 3394 3395
			} else
				clear_bit(R5_MadeGood, &dev->flags);
		}
3396 3397 3398 3399 3400 3401 3402 3403 3404
		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);
		}
3405
		if (!test_bit(R5_Insync, &dev->flags)) {
3406 3407 3408
			/* 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
3409
		}
3410 3411 3412
		if (test_bit(R5_ReadError, &dev->flags))
			clear_bit(R5_Insync, &dev->flags);
		if (!test_bit(R5_Insync, &dev->flags)) {
3413 3414 3415
			if (s->failed < 2)
				s->failed_num[s->failed] = i;
			s->failed++;
3416 3417
			if (rdev && !test_bit(Faulty, &rdev->flags))
				do_recovery = 1;
3418
		}
L
Linus Torvalds 已提交
3419
	}
3420 3421 3422 3423
	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
3424
		 * else if MD_RECOVERY_REQUESTED is set, we also are syncing.
3425 3426 3427 3428 3429
		 * else we can only be replacing
		 * sync and recovery both need to read all devices, and so
		 * use the same flag.
		 */
		if (do_recovery ||
3430 3431
		    sh->sector >= conf->mddev->recovery_cp ||
		    test_bit(MD_RECOVERY_REQUESTED, &(conf->mddev->recovery)))
3432 3433 3434 3435
			s->syncing = 1;
		else
			s->replacing = 1;
	}
L
Linus Torvalds 已提交
3436
	rcu_read_unlock();
3437 3438 3439 3440 3441
}

static void handle_stripe(struct stripe_head *sh)
{
	struct stripe_head_state s;
3442
	struct r5conf *conf = sh->raid_conf;
3443
	int i;
3444 3445
	int prexor;
	int disks = sh->disks;
3446
	struct r5dev *pdev, *qdev;
3447 3448

	clear_bit(STRIPE_HANDLE, &sh->state);
3449
	if (test_and_set_bit_lock(STRIPE_ACTIVE, &sh->state)) {
3450 3451 3452 3453 3454 3455
		/* already being handled, ensure it gets handled
		 * again when current action finishes */
		set_bit(STRIPE_HANDLE, &sh->state);
		return;
	}

3456 3457 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);
		}
		spin_unlock(&sh->stripe_lock);
3465 3466 3467 3468 3469 3470 3471 3472
	}
	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);
3473

3474
	analyse_stripe(sh, &s);
3475

3476 3477 3478 3479 3480
	if (s.handle_bad_blocks) {
		set_bit(STRIPE_HANDLE, &sh->state);
		goto finish;
	}

3481 3482
	if (unlikely(s.blocked_rdev)) {
		if (s.syncing || s.expanding || s.expanded ||
3483
		    s.replacing || s.to_write || s.written) {
3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503
			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.
	 */
3504 3505 3506 3507 3508
	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);
3509
		if (s.syncing + s.replacing)
3510 3511
			handle_failed_sync(conf, sh, &s);
	}
3512

3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525
	/* 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
		 */
3526 3527
		BUG_ON(!test_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags) &&
		       !test_bit(R5_Discard, &sh->dev[sh->pd_idx].flags));
3528
		BUG_ON(sh->qd_idx >= 0 &&
3529 3530
		       !test_bit(R5_UPTODATE, &sh->dev[sh->qd_idx].flags) &&
		       !test_bit(R5_Discard, &sh->dev[sh->qd_idx].flags));
3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549
		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;
	}

3550 3551 3552 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
	/*
	 * 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);

3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606
	/* 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);
	}
3607

3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621
	if (s.replacing && s.locked == 0
	    && !test_bit(STRIPE_INSYNC, &sh->state)) {
		/* Write out to replacement devices where possible */
		for (i = 0; i < conf->raid_disks; i++)
			if (test_bit(R5_UPTODATE, &sh->dev[i].flags) &&
			    test_bit(R5_NeedReplace, &sh->dev[i].flags)) {
				set_bit(R5_WantReplace, &sh->dev[i].flags);
				set_bit(R5_LOCKED, &sh->dev[i].flags);
				s.locked++;
			}
		set_bit(STRIPE_INSYNC, &sh->state);
	}
	if ((s.syncing || s.replacing) && s.locked == 0 &&
	    test_bit(STRIPE_INSYNC, &sh->state)) {
3622 3623
		md_done_sync(conf->mddev, STRIPE_SECTORS, 1);
		clear_bit(STRIPE_SYNCING, &sh->state);
3624 3625
		if (test_and_clear_bit(R5_Overlap, &sh->dev[sh->pd_idx].flags))
			wake_up(&conf->wait_for_overlap);
3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652
	}

	/* 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++;
				}
			}
		}


3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679
	/* 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++;
		}
	}
3680

3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695 3696
	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);
3697

3698
finish:
3699
	/* wait for this device to become unblocked */
3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711
	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);
	}
3712

3713 3714
	if (s.handle_bad_blocks)
		for (i = disks; i--; ) {
3715
			struct md_rdev *rdev;
3716 3717 3718 3719 3720 3721 3722 3723 3724
			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);
			}
3725 3726 3727
			if (test_and_clear_bit(R5_MadeGood, &dev->flags)) {
				rdev = conf->disks[i].rdev;
				rdev_clear_badblocks(rdev, sh->sector,
3728
						     STRIPE_SECTORS, 0);
3729 3730
				rdev_dec_pending(rdev, conf->mddev);
			}
3731 3732
			if (test_and_clear_bit(R5_MadeGoodRepl, &dev->flags)) {
				rdev = conf->disks[i].replacement;
3733 3734 3735
				if (!rdev)
					/* rdev have been moved down */
					rdev = conf->disks[i].rdev;
3736
				rdev_clear_badblocks(rdev, sh->sector,
3737
						     STRIPE_SECTORS, 0);
3738 3739
				rdev_dec_pending(rdev, conf->mddev);
			}
3740 3741
		}

3742 3743 3744
	if (s.ops_request)
		raid_run_ops(sh, s.ops_request);

D
Dan Williams 已提交
3745
	ops_run_io(sh, &s);
3746

3747
	if (s.dec_preread_active) {
3748
		/* We delay this until after ops_run_io so that if make_request
T
Tejun Heo 已提交
3749
		 * is waiting on a flush, it won't continue until the writes
3750 3751 3752 3753 3754 3755 3756 3757
		 * 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);
	}

3758
	return_io(s.return_bi);
3759

3760
	clear_bit_unlock(STRIPE_ACTIVE, &sh->state);
3761 3762
}

3763
static void raid5_activate_delayed(struct r5conf *conf)
3764 3765 3766 3767 3768 3769 3770 3771 3772 3773
{
	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);
3774
			list_add_tail(&sh->lru, &conf->hold_list);
3775
		}
N
NeilBrown 已提交
3776
	}
3777 3778
}

3779
static void activate_bit_delay(struct r5conf *conf)
3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792
{
	/* 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);
	}
}

3793
int md_raid5_congested(struct mddev *mddev, int bits)
3794
{
3795
	struct r5conf *conf = mddev->private;
3796 3797 3798 3799

	/* No difference between reads and writes.  Just check
	 * how busy the stripe_cache is
	 */
3800

3801 3802 3803 3804 3805 3806 3807 3808 3809
	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
3810 3811 3812 3813
EXPORT_SYMBOL_GPL(md_raid5_congested);

static int raid5_congested(void *data, int bits)
{
3814
	struct mddev *mddev = data;
N
NeilBrown 已提交
3815 3816 3817 3818

	return mddev_congested(mddev, bits) ||
		md_raid5_congested(mddev, bits);
}
3819

3820 3821 3822
/* We want read requests to align with chunks where possible,
 * but write requests don't need to.
 */
3823 3824 3825
static int raid5_mergeable_bvec(struct request_queue *q,
				struct bvec_merge_data *bvm,
				struct bio_vec *biovec)
3826
{
3827
	struct mddev *mddev = q->queuedata;
3828
	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
3829
	int max;
3830
	unsigned int chunk_sectors = mddev->chunk_sectors;
3831
	unsigned int bio_sectors = bvm->bi_size >> 9;
3832

3833
	if ((bvm->bi_rw & 1) == WRITE)
3834 3835
		return biovec->bv_len; /* always allow writes to be mergeable */

3836 3837
	if (mddev->new_chunk_sectors < mddev->chunk_sectors)
		chunk_sectors = mddev->new_chunk_sectors;
3838 3839 3840 3841 3842 3843 3844 3845
	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;
}

3846

3847
static int in_chunk_boundary(struct mddev *mddev, struct bio *bio)
3848 3849
{
	sector_t sector = bio->bi_sector + get_start_sect(bio->bi_bdev);
3850
	unsigned int chunk_sectors = mddev->chunk_sectors;
3851
	unsigned int bio_sectors = bio_sectors(bio);
3852

3853 3854
	if (mddev->new_chunk_sectors < mddev->chunk_sectors)
		chunk_sectors = mddev->new_chunk_sectors;
3855 3856 3857 3858
	return  chunk_sectors >=
		((sector & (chunk_sectors - 1)) + bio_sectors);
}

3859 3860 3861 3862
/*
 *  add bio to the retry LIFO  ( in O(1) ... we are in interrupt )
 *  later sampled by raid5d.
 */
3863
static void add_bio_to_retry(struct bio *bi,struct r5conf *conf)
3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876
{
	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);
}


3877
static struct bio *remove_bio_from_retry(struct r5conf *conf)
3878 3879 3880 3881 3882 3883 3884 3885 3886 3887
{
	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) {
3888
		conf->retry_read_aligned_list = bi->bi_next;
3889
		bi->bi_next = NULL;
3890 3891 3892 3893
		/*
		 * this sets the active strip count to 1 and the processed
		 * strip count to zero (upper 8 bits)
		 */
3894
		raid5_set_bi_stripes(bi, 1); /* biased count of active stripes */
3895 3896 3897 3898 3899 3900
	}

	return bi;
}


3901 3902 3903 3904 3905 3906
/*
 *  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..
 */
3907
static void raid5_align_endio(struct bio *bi, int error)
3908 3909
{
	struct bio* raid_bi  = bi->bi_private;
3910
	struct mddev *mddev;
3911
	struct r5conf *conf;
3912
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
3913
	struct md_rdev *rdev;
3914

3915
	bio_put(bi);
3916 3917 3918

	rdev = (void*)raid_bi->bi_next;
	raid_bi->bi_next = NULL;
3919 3920
	mddev = rdev->mddev;
	conf = mddev->private;
3921 3922 3923 3924

	rdev_dec_pending(rdev, conf->mddev);

	if (!error && uptodate) {
3925 3926
		trace_block_bio_complete(bdev_get_queue(raid_bi->bi_bdev),
					 raid_bi, 0);
3927
		bio_endio(raid_bi, 0);
3928 3929
		if (atomic_dec_and_test(&conf->active_aligned_reads))
			wake_up(&conf->wait_for_stripe);
3930
		return;
3931 3932 3933
	}


3934
	pr_debug("raid5_align_endio : io error...handing IO for a retry\n");
3935 3936

	add_bio_to_retry(raid_bi, conf);
3937 3938
}

3939 3940
static int bio_fits_rdev(struct bio *bi)
{
3941
	struct request_queue *q = bdev_get_queue(bi->bi_bdev);
3942

3943
	if (bio_sectors(bi) > queue_max_sectors(q))
3944 3945
		return 0;
	blk_recount_segments(q, bi);
3946
	if (bi->bi_phys_segments > queue_max_segments(q))
3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958
		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;
}


3959
static int chunk_aligned_read(struct mddev *mddev, struct bio * raid_bio)
3960
{
3961
	struct r5conf *conf = mddev->private;
N
NeilBrown 已提交
3962
	int dd_idx;
3963
	struct bio* align_bi;
3964
	struct md_rdev *rdev;
3965
	sector_t end_sector;
3966 3967

	if (!in_chunk_boundary(mddev, raid_bio)) {
3968
		pr_debug("chunk_aligned_read : non aligned\n");
3969 3970 3971
		return 0;
	}
	/*
3972
	 * use bio_clone_mddev to make a copy of the bio
3973
	 */
3974
	align_bi = bio_clone_mddev(raid_bio, GFP_NOIO, mddev);
3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985
	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
	 */
3986 3987
	align_bi->bi_sector =  raid5_compute_sector(conf, raid_bio->bi_sector,
						    0,
3988
						    &dd_idx, NULL);
3989

K
Kent Overstreet 已提交
3990
	end_sector = bio_end_sector(align_bi);
3991
	rcu_read_lock();
3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002
	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) {
4003 4004 4005
		sector_t first_bad;
		int bad_sectors;

4006 4007
		atomic_inc(&rdev->nr_pending);
		rcu_read_unlock();
4008 4009 4010 4011
		raid_bio->bi_next = (void*)rdev;
		align_bi->bi_bdev =  rdev->bdev;
		align_bi->bi_flags &= ~(1 << BIO_SEG_VALID);

4012
		if (!bio_fits_rdev(align_bi) ||
4013
		    is_badblock(rdev, align_bi->bi_sector, bio_sectors(align_bi),
4014 4015
				&first_bad, &bad_sectors)) {
			/* too big in some way, or has a known bad block */
4016 4017 4018 4019 4020
			bio_put(align_bi);
			rdev_dec_pending(rdev, mddev);
			return 0;
		}

4021 4022 4023
		/* No reshape active, so we can trust rdev->data_offset */
		align_bi->bi_sector += rdev->data_offset;

4024 4025 4026
		spin_lock_irq(&conf->device_lock);
		wait_event_lock_irq(conf->wait_for_stripe,
				    conf->quiesce == 0,
4027
				    conf->device_lock);
4028 4029 4030
		atomic_inc(&conf->active_aligned_reads);
		spin_unlock_irq(&conf->device_lock);

4031 4032 4033 4034
		if (mddev->gendisk)
			trace_block_bio_remap(bdev_get_queue(align_bi->bi_bdev),
					      align_bi, disk_devt(mddev->gendisk),
					      raid_bio->bi_sector);
4035 4036 4037 4038
		generic_make_request(align_bi);
		return 1;
	} else {
		rcu_read_unlock();
4039
		bio_put(align_bi);
4040 4041 4042 4043
		return 0;
	}
}

4044 4045 4046 4047 4048 4049 4050 4051 4052 4053
/* __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.
 */
4054
static struct stripe_head *__get_priority_stripe(struct r5conf *conf)
4055 4056 4057 4058 4059 4060 4061 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
{
	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;
}
4096

4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108
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
4109
	int cnt = 0;
4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123

	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
4124
			cnt++;
4125 4126 4127
		}
		spin_unlock_irq(&conf->device_lock);
	}
4128 4129
	if (mddev->queue)
		trace_block_unplug(mddev->queue, cnt, !from_schedule);
4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 4156
	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
4157 4158 4159 4160 4161 4162 4163 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
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);
4192 4193 4194 4195 4196 4197 4198
		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
4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210
		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;
			}
		}
4211
		set_bit(STRIPE_DISCARD, &sh->state);
S
Shaohua Li 已提交
4212 4213 4214 4215 4216 4217 4218 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
		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);
	}
}

4247
static void make_request(struct mddev *mddev, struct bio * bi)
L
Linus Torvalds 已提交
4248
{
4249
	struct r5conf *conf = mddev->private;
4250
	int dd_idx;
L
Linus Torvalds 已提交
4251 4252 4253
	sector_t new_sector;
	sector_t logical_sector, last_sector;
	struct stripe_head *sh;
4254
	const int rw = bio_data_dir(bi);
4255
	int remaining;
L
Linus Torvalds 已提交
4256

T
Tejun Heo 已提交
4257 4258
	if (unlikely(bi->bi_rw & REQ_FLUSH)) {
		md_flush_request(mddev, bi);
4259
		return;
4260 4261
	}

4262
	md_write_start(mddev, bi);
4263

4264
	if (rw == READ &&
4265
	     mddev->reshape_position == MaxSector &&
4266
	     chunk_aligned_read(mddev,bi))
4267
		return;
4268

S
Shaohua Li 已提交
4269 4270 4271 4272 4273
	if (unlikely(bi->bi_rw & REQ_DISCARD)) {
		make_discard_request(mddev, bi);
		return;
	}

L
Linus Torvalds 已提交
4274
	logical_sector = bi->bi_sector & ~((sector_t)STRIPE_SECTORS-1);
K
Kent Overstreet 已提交
4275
	last_sector = bio_end_sector(bi);
L
Linus Torvalds 已提交
4276 4277
	bi->bi_next = NULL;
	bi->bi_phys_segments = 1;	/* over-loaded to count active stripes */
4278

L
Linus Torvalds 已提交
4279 4280
	for (;logical_sector < last_sector; logical_sector += STRIPE_SECTORS) {
		DEFINE_WAIT(w);
4281
		int previous;
4282

4283
	retry:
4284
		previous = 0;
4285
		prepare_to_wait(&conf->wait_for_overlap, &w, TASK_UNINTERRUPTIBLE);
4286
		if (unlikely(conf->reshape_progress != MaxSector)) {
4287
			/* spinlock is needed as reshape_progress may be
4288 4289
			 * 64bit on a 32bit platform, and so it might be
			 * possible to see a half-updated value
4290
			 * Of course reshape_progress could change after
4291 4292 4293 4294
			 * the lock is dropped, so once we get a reference
			 * to the stripe that we think it is, we will have
			 * to check again.
			 */
4295
			spin_lock_irq(&conf->device_lock);
4296
			if (mddev->reshape_backwards
4297 4298
			    ? logical_sector < conf->reshape_progress
			    : logical_sector >= conf->reshape_progress) {
4299 4300
				previous = 1;
			} else {
4301
				if (mddev->reshape_backwards
4302 4303
				    ? logical_sector < conf->reshape_safe
				    : logical_sector >= conf->reshape_safe) {
4304 4305 4306 4307 4308
					spin_unlock_irq(&conf->device_lock);
					schedule();
					goto retry;
				}
			}
4309 4310
			spin_unlock_irq(&conf->device_lock);
		}
4311

4312 4313
		new_sector = raid5_compute_sector(conf, logical_sector,
						  previous,
4314
						  &dd_idx, NULL);
4315
		pr_debug("raid456: make_request, sector %llu logical %llu\n",
L
Linus Torvalds 已提交
4316 4317 4318
			(unsigned long long)new_sector, 
			(unsigned long long)logical_sector);

4319
		sh = get_active_stripe(conf, new_sector, previous,
4320
				       (bi->bi_rw&RWA_MASK), 0);
L
Linus Torvalds 已提交
4321
		if (sh) {
4322
			if (unlikely(previous)) {
4323
				/* expansion might have moved on while waiting for a
4324 4325 4326 4327 4328 4329
				 * 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.
4330 4331 4332
				 */
				int must_retry = 0;
				spin_lock_irq(&conf->device_lock);
4333
				if (mddev->reshape_backwards
4334 4335
				    ? logical_sector >= conf->reshape_progress
				    : logical_sector < conf->reshape_progress)
4336 4337 4338 4339 4340
					/* mismatch, need to try again */
					must_retry = 1;
				spin_unlock_irq(&conf->device_lock);
				if (must_retry) {
					release_stripe(sh);
4341
					schedule();
4342 4343 4344
					goto retry;
				}
			}
4345

4346
			if (rw == WRITE &&
4347
			    logical_sector >= mddev->suspend_lo &&
4348 4349
			    logical_sector < mddev->suspend_hi) {
				release_stripe(sh);
4350 4351 4352 4353 4354 4355 4356 4357 4358 4359
				/* 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();
4360 4361
				goto retry;
			}
4362 4363

			if (test_bit(STRIPE_EXPANDING, &sh->state) ||
4364
			    !add_stripe_bio(sh, bi, dd_idx, rw)) {
4365 4366
				/* Stripe is busy expanding or
				 * add failed due to overlap.  Flush everything
L
Linus Torvalds 已提交
4367 4368
				 * and wait a while
				 */
N
NeilBrown 已提交
4369
				md_wakeup_thread(mddev->thread);
L
Linus Torvalds 已提交
4370 4371 4372 4373 4374
				release_stripe(sh);
				schedule();
				goto retry;
			}
			finish_wait(&conf->wait_for_overlap, &w);
4375 4376
			set_bit(STRIPE_HANDLE, &sh->state);
			clear_bit(STRIPE_DELAYED, &sh->state);
4377
			if ((bi->bi_rw & REQ_SYNC) &&
4378 4379
			    !test_and_set_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
				atomic_inc(&conf->preread_active_stripes);
4380
			release_stripe_plug(mddev, sh);
L
Linus Torvalds 已提交
4381 4382 4383 4384 4385 4386 4387
		} 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;
		}
	}
4388

4389
	remaining = raid5_dec_bi_active_stripes(bi);
4390
	if (remaining == 0) {
L
Linus Torvalds 已提交
4391

4392
		if ( rw == WRITE )
L
Linus Torvalds 已提交
4393
			md_write_end(mddev);
4394

4395 4396
		trace_block_bio_complete(bdev_get_queue(bi->bi_bdev),
					 bi, 0);
4397
		bio_endio(bi, 0);
L
Linus Torvalds 已提交
4398 4399 4400
	}
}

4401
static sector_t raid5_size(struct mddev *mddev, sector_t sectors, int raid_disks);
D
Dan Williams 已提交
4402

4403
static sector_t reshape_request(struct mddev *mddev, sector_t sector_nr, int *skipped)
L
Linus Torvalds 已提交
4404
{
4405 4406 4407 4408 4409 4410 4411 4412 4413
	/* 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.
	 */
4414
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
4415
	struct stripe_head *sh;
4416
	sector_t first_sector, last_sector;
4417 4418 4419
	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;
4420 4421
	int i;
	int dd_idx;
4422
	sector_t writepos, readpos, safepos;
4423
	sector_t stripe_addr;
4424
	int reshape_sectors;
4425
	struct list_head stripes;
4426

4427 4428
	if (sector_nr == 0) {
		/* If restarting in the middle, skip the initial sectors */
4429
		if (mddev->reshape_backwards &&
4430 4431 4432
		    conf->reshape_progress < raid5_size(mddev, 0, 0)) {
			sector_nr = raid5_size(mddev, 0, 0)
				- conf->reshape_progress;
4433
		} else if (!mddev->reshape_backwards &&
4434 4435
			   conf->reshape_progress > 0)
			sector_nr = conf->reshape_progress;
4436
		sector_div(sector_nr, new_data_disks);
4437
		if (sector_nr) {
4438 4439
			mddev->curr_resync_completed = sector_nr;
			sysfs_notify(&mddev->kobj, NULL, "sync_completed");
4440 4441 4442
			*skipped = 1;
			return sector_nr;
		}
4443 4444
	}

4445 4446 4447 4448
	/* 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
	 */
4449 4450
	if (mddev->new_chunk_sectors > mddev->chunk_sectors)
		reshape_sectors = mddev->new_chunk_sectors;
4451
	else
4452
		reshape_sectors = mddev->chunk_sectors;
4453

4454 4455 4456 4457 4458
	/* 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
4459
	 */
4460
	writepos = conf->reshape_progress;
4461
	sector_div(writepos, new_data_disks);
4462 4463
	readpos = conf->reshape_progress;
	sector_div(readpos, data_disks);
4464
	safepos = conf->reshape_safe;
4465
	sector_div(safepos, data_disks);
4466
	if (mddev->reshape_backwards) {
4467
		writepos -= min_t(sector_t, reshape_sectors, writepos);
4468
		readpos += reshape_sectors;
4469
		safepos += reshape_sectors;
4470
	} else {
4471
		writepos += reshape_sectors;
4472 4473
		readpos -= min_t(sector_t, reshape_sectors, readpos);
		safepos -= min_t(sector_t, reshape_sectors, safepos);
4474
	}
4475

4476 4477 4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490
	/* 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;
	}

4491 4492 4493 4494
	/* '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.
4495 4496 4497 4498
	 * 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
4499 4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 4510
	 * 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???
	 */
4511 4512 4513 4514 4515 4516
	if (conf->min_offset_diff < 0) {
		safepos += -conf->min_offset_diff;
		readpos += -conf->min_offset_diff;
	} else
		writepos += conf->min_offset_diff;

4517
	if ((mddev->reshape_backwards
4518 4519 4520
	     ? (safepos > writepos && readpos < writepos)
	     : (safepos < writepos && readpos > writepos)) ||
	    time_after(jiffies, conf->reshape_checkpoint + 10*HZ)) {
4521 4522 4523
		/* Cannot proceed until we've updated the superblock... */
		wait_event(conf->wait_for_overlap,
			   atomic_read(&conf->reshape_stripes)==0);
4524
		mddev->reshape_position = conf->reshape_progress;
4525
		mddev->curr_resync_completed = sector_nr;
4526
		conf->reshape_checkpoint = jiffies;
4527
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
4528
		md_wakeup_thread(mddev->thread);
4529
		wait_event(mddev->sb_wait, mddev->flags == 0 ||
4530 4531
			   kthread_should_stop());
		spin_lock_irq(&conf->device_lock);
4532
		conf->reshape_safe = mddev->reshape_position;
4533 4534
		spin_unlock_irq(&conf->device_lock);
		wake_up(&conf->wait_for_overlap);
4535
		sysfs_notify(&mddev->kobj, NULL, "sync_completed");
4536 4537
	}

4538
	INIT_LIST_HEAD(&stripes);
4539
	for (i = 0; i < reshape_sectors; i += STRIPE_SECTORS) {
4540
		int j;
4541
		int skipped_disk = 0;
4542
		sh = get_active_stripe(conf, stripe_addr+i, 0, 0, 1);
4543 4544 4545 4546 4547 4548 4549 4550 4551
		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;
4552
			if (conf->level == 6 &&
4553
			    j == sh->qd_idx)
4554
				continue;
4555
			s = compute_blocknr(sh, j, 0);
D
Dan Williams 已提交
4556
			if (s < raid5_size(mddev, 0, 0)) {
4557
				skipped_disk = 1;
4558 4559 4560 4561 4562 4563
				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);
		}
4564
		if (!skipped_disk) {
4565 4566 4567
			set_bit(STRIPE_EXPAND_READY, &sh->state);
			set_bit(STRIPE_HANDLE, &sh->state);
		}
4568
		list_add(&sh->lru, &stripes);
4569 4570
	}
	spin_lock_irq(&conf->device_lock);
4571
	if (mddev->reshape_backwards)
4572
		conf->reshape_progress -= reshape_sectors * new_data_disks;
4573
	else
4574
		conf->reshape_progress += reshape_sectors * new_data_disks;
4575 4576 4577 4578 4579 4580 4581
	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 =
4582
		raid5_compute_sector(conf, stripe_addr*(new_data_disks),
4583
				     1, &dd_idx, NULL);
4584
	last_sector =
4585
		raid5_compute_sector(conf, ((stripe_addr+reshape_sectors)
4586
					    * new_data_disks - 1),
4587
				     1, &dd_idx, NULL);
A
Andre Noll 已提交
4588 4589
	if (last_sector >= mddev->dev_sectors)
		last_sector = mddev->dev_sectors - 1;
4590
	while (first_sector <= last_sector) {
4591
		sh = get_active_stripe(conf, first_sector, 1, 0, 1);
4592 4593 4594 4595 4596
		set_bit(STRIPE_EXPAND_SOURCE, &sh->state);
		set_bit(STRIPE_HANDLE, &sh->state);
		release_stripe(sh);
		first_sector += STRIPE_SECTORS;
	}
4597 4598 4599 4600 4601 4602 4603 4604
	/* 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);
	}
4605 4606 4607
	/* If this takes us to the resync_max point where we have to pause,
	 * then we need to write out the superblock.
	 */
4608
	sector_nr += reshape_sectors;
4609 4610
	if ((sector_nr - mddev->curr_resync_completed) * 2
	    >= mddev->resync_max - mddev->curr_resync_completed) {
4611 4612 4613
		/* Cannot proceed until we've updated the superblock... */
		wait_event(conf->wait_for_overlap,
			   atomic_read(&conf->reshape_stripes) == 0);
4614
		mddev->reshape_position = conf->reshape_progress;
4615
		mddev->curr_resync_completed = sector_nr;
4616
		conf->reshape_checkpoint = jiffies;
4617 4618 4619 4620 4621 4622
		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);
4623
		conf->reshape_safe = mddev->reshape_position;
4624 4625
		spin_unlock_irq(&conf->device_lock);
		wake_up(&conf->wait_for_overlap);
4626
		sysfs_notify(&mddev->kobj, NULL, "sync_completed");
4627
	}
4628
	return reshape_sectors;
4629 4630 4631
}

/* FIXME go_faster isn't used */
4632
static inline sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipped, int go_faster)
4633
{
4634
	struct r5conf *conf = mddev->private;
4635
	struct stripe_head *sh;
A
Andre Noll 已提交
4636
	sector_t max_sector = mddev->dev_sectors;
N
NeilBrown 已提交
4637
	sector_t sync_blocks;
4638 4639
	int still_degraded = 0;
	int i;
L
Linus Torvalds 已提交
4640

4641
	if (sector_nr >= max_sector) {
L
Linus Torvalds 已提交
4642
		/* just being told to finish up .. nothing much to do */
4643

4644 4645 4646 4647
		if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) {
			end_reshape(conf);
			return 0;
		}
4648 4649 4650 4651

		if (mddev->curr_resync < max_sector) /* aborted */
			bitmap_end_sync(mddev->bitmap, mddev->curr_resync,
					&sync_blocks, 1);
4652
		else /* completed sync */
4653 4654 4655
			conf->fullsync = 0;
		bitmap_close_sync(mddev->bitmap);

L
Linus Torvalds 已提交
4656 4657
		return 0;
	}
4658

4659 4660 4661
	/* Allow raid5_quiesce to complete */
	wait_event(conf->wait_for_overlap, conf->quiesce != 2);

4662 4663
	if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
		return reshape_request(mddev, sector_nr, skipped);
4664

4665 4666 4667 4668 4669 4670
	/* 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
	 */

4671
	/* if there is too many failed drives and we are trying
L
Linus Torvalds 已提交
4672 4673 4674
	 * to resync, then assert that we are finished, because there is
	 * nothing we can do.
	 */
4675
	if (mddev->degraded >= conf->max_degraded &&
4676
	    test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
A
Andre Noll 已提交
4677
		sector_t rv = mddev->dev_sectors - sector_nr;
4678
		*skipped = 1;
L
Linus Torvalds 已提交
4679 4680
		return rv;
	}
4681 4682 4683 4684
	if (!test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
	    !conf->fullsync &&
	    !bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, 1) &&
	    sync_blocks >= STRIPE_SECTORS) {
4685 4686 4687 4688 4689
		/* 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
4690

N
NeilBrown 已提交
4691 4692
	bitmap_cond_end_sync(mddev->bitmap, sector_nr);

4693
	sh = get_active_stripe(conf, sector_nr, 0, 1, 0);
L
Linus Torvalds 已提交
4694
	if (sh == NULL) {
4695
		sh = get_active_stripe(conf, sector_nr, 0, 0, 0);
L
Linus Torvalds 已提交
4696
		/* make sure we don't swamp the stripe cache if someone else
4697
		 * is trying to get access
L
Linus Torvalds 已提交
4698
		 */
4699
		schedule_timeout_uninterruptible(1);
L
Linus Torvalds 已提交
4700
	}
4701 4702 4703 4704
	/* 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.
	 */
4705
	for (i = 0; i < conf->raid_disks; i++)
4706 4707 4708 4709 4710
		if (conf->disks[i].rdev == NULL)
			still_degraded = 1;

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

4711
	set_bit(STRIPE_SYNC_REQUESTED, &sh->state);
L
Linus Torvalds 已提交
4712

4713
	handle_stripe(sh);
L
Linus Torvalds 已提交
4714 4715 4716 4717 4718
	release_stripe(sh);

	return STRIPE_SECTORS;
}

4719
static int  retry_aligned_read(struct r5conf *conf, struct bio *raid_bio)
4720 4721 4722 4723 4724 4725 4726 4727 4728 4729 4730 4731
{
	/* 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;
4732
	int dd_idx;
4733 4734 4735 4736 4737 4738
	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);
4739
	sector = raid5_compute_sector(conf, logical_sector,
4740
				      0, &dd_idx, NULL);
K
Kent Overstreet 已提交
4741
	last_sector = bio_end_sector(raid_bio);
4742 4743

	for (; logical_sector < last_sector;
4744 4745 4746
	     logical_sector += STRIPE_SECTORS,
		     sector += STRIPE_SECTORS,
		     scnt++) {
4747

4748
		if (scnt < raid5_bi_processed_stripes(raid_bio))
4749 4750 4751
			/* already done this stripe */
			continue;

4752
		sh = get_active_stripe(conf, sector, 0, 1, 0);
4753 4754 4755

		if (!sh) {
			/* failed to get a stripe - must wait */
4756
			raid5_set_bi_processed_stripes(raid_bio, scnt);
4757 4758 4759 4760
			conf->retry_read_aligned = raid_bio;
			return handled;
		}

4761 4762
		if (!add_stripe_bio(sh, raid_bio, dd_idx, 0)) {
			release_stripe(sh);
4763
			raid5_set_bi_processed_stripes(raid_bio, scnt);
4764 4765 4766 4767
			conf->retry_read_aligned = raid_bio;
			return handled;
		}

4768
		set_bit(R5_ReadNoMerge, &sh->dev[dd_idx].flags);
4769
		handle_stripe(sh);
4770 4771 4772
		release_stripe(sh);
		handled++;
	}
4773
	remaining = raid5_dec_bi_active_stripes(raid_bio);
4774 4775 4776
	if (remaining == 0) {
		trace_block_bio_complete(bdev_get_queue(raid_bio->bi_bdev),
					 raid_bio, 0);
4777
		bio_endio(raid_bio, 0);
4778
	}
4779 4780 4781 4782 4783
	if (atomic_dec_and_test(&conf->active_aligned_reads))
		wake_up(&conf->wait_for_stripe);
	return handled;
}

4784 4785 4786 4787 4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 4798 4799 4800 4801 4802 4803 4804 4805 4806 4807
#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;
}
4808

L
Linus Torvalds 已提交
4809 4810 4811 4812 4813 4814 4815
/*
 * 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
4816
static void raid5d(struct md_thread *thread)
L
Linus Torvalds 已提交
4817
{
S
Shaohua Li 已提交
4818
	struct mddev *mddev = thread->mddev;
4819
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
4820
	int handled;
4821
	struct blk_plug plug;
L
Linus Torvalds 已提交
4822

4823
	pr_debug("+++ raid5d active\n");
L
Linus Torvalds 已提交
4824 4825 4826

	md_check_recovery(mddev);

4827
	blk_start_plug(&plug);
L
Linus Torvalds 已提交
4828 4829 4830
	handled = 0;
	spin_lock_irq(&conf->device_lock);
	while (1) {
4831
		struct bio *bio;
4832
		int batch_size;
L
Linus Torvalds 已提交
4833

4834
		if (
4835 4836 4837
		    !list_empty(&conf->bitmap_list)) {
			/* Now is a good time to flush some bitmap updates */
			conf->seq_flush++;
4838
			spin_unlock_irq(&conf->device_lock);
4839
			bitmap_unplug(mddev->bitmap);
4840
			spin_lock_irq(&conf->device_lock);
4841
			conf->seq_write = conf->seq_flush;
4842 4843
			activate_bit_delay(conf);
		}
4844
		raid5_activate_delayed(conf);
4845

4846 4847 4848 4849 4850 4851 4852 4853 4854 4855
		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++;
		}

4856 4857
		batch_size = handle_active_stripes(conf);
		if (!batch_size)
L
Linus Torvalds 已提交
4858
			break;
4859
		handled += batch_size;
L
Linus Torvalds 已提交
4860

4861 4862
		if (mddev->flags & ~(1<<MD_CHANGE_PENDING)) {
			spin_unlock_irq(&conf->device_lock);
4863
			md_check_recovery(mddev);
4864 4865
			spin_lock_irq(&conf->device_lock);
		}
L
Linus Torvalds 已提交
4866
	}
4867
	pr_debug("%d stripes handled\n", handled);
L
Linus Torvalds 已提交
4868 4869 4870

	spin_unlock_irq(&conf->device_lock);

4871
	async_tx_issue_pending_all();
4872
	blk_finish_plug(&plug);
L
Linus Torvalds 已提交
4873

4874
	pr_debug("--- raid5d inactive\n");
L
Linus Torvalds 已提交
4875 4876
}

4877
static ssize_t
4878
raid5_show_stripe_cache_size(struct mddev *mddev, char *page)
4879
{
4880
	struct r5conf *conf = mddev->private;
4881 4882 4883 4884
	if (conf)
		return sprintf(page, "%d\n", conf->max_nr_stripes);
	else
		return 0;
4885 4886
}

4887
int
4888
raid5_set_cache_size(struct mddev *mddev, int size)
4889
{
4890
	struct r5conf *conf = mddev->private;
4891 4892
	int err;

4893
	if (size <= 16 || size > 32768)
4894
		return -EINVAL;
4895
	while (size < conf->max_nr_stripes) {
4896 4897 4898 4899 4900
		if (drop_one_stripe(conf))
			conf->max_nr_stripes--;
		else
			break;
	}
4901 4902 4903
	err = md_allow_write(mddev);
	if (err)
		return err;
4904
	while (size > conf->max_nr_stripes) {
4905 4906 4907 4908
		if (grow_one_stripe(conf))
			conf->max_nr_stripes++;
		else break;
	}
4909 4910 4911 4912 4913
	return 0;
}
EXPORT_SYMBOL(raid5_set_cache_size);

static ssize_t
4914
raid5_store_stripe_cache_size(struct mddev *mddev, const char *page, size_t len)
4915
{
4916
	struct r5conf *conf = mddev->private;
4917 4918 4919 4920 4921 4922 4923 4924 4925 4926 4927 4928 4929
	unsigned long new;
	int err;

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

	if (strict_strtoul(page, 10, &new))
		return -EINVAL;
	err = raid5_set_cache_size(mddev, new);
	if (err)
		return err;
4930 4931
	return len;
}
4932

4933 4934 4935 4936
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);
4937

4938
static ssize_t
4939
raid5_show_preread_threshold(struct mddev *mddev, char *page)
4940
{
4941
	struct r5conf *conf = mddev->private;
4942 4943 4944 4945 4946 4947 4948
	if (conf)
		return sprintf(page, "%d\n", conf->bypass_threshold);
	else
		return 0;
}

static ssize_t
4949
raid5_store_preread_threshold(struct mddev *mddev, const char *page, size_t len)
4950
{
4951
	struct r5conf *conf = mddev->private;
4952
	unsigned long new;
4953 4954 4955 4956 4957
	if (len >= PAGE_SIZE)
		return -EINVAL;
	if (!conf)
		return -ENODEV;

4958
	if (strict_strtoul(page, 10, &new))
4959
		return -EINVAL;
4960
	if (new > conf->max_nr_stripes)
4961 4962 4963 4964 4965 4966 4967 4968 4969 4970 4971
		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);

4972
static ssize_t
4973
stripe_cache_active_show(struct mddev *mddev, char *page)
4974
{
4975
	struct r5conf *conf = mddev->private;
4976 4977 4978 4979
	if (conf)
		return sprintf(page, "%d\n", atomic_read(&conf->active_stripes));
	else
		return 0;
4980 4981
}

4982 4983
static struct md_sysfs_entry
raid5_stripecache_active = __ATTR_RO(stripe_cache_active);
4984

4985
static struct attribute *raid5_attrs[] =  {
4986 4987
	&raid5_stripecache_size.attr,
	&raid5_stripecache_active.attr,
4988
	&raid5_preread_bypass_threshold.attr,
4989 4990
	NULL,
};
4991 4992 4993
static struct attribute_group raid5_attrs_group = {
	.name = NULL,
	.attrs = raid5_attrs,
4994 4995
};

4996
static sector_t
4997
raid5_size(struct mddev *mddev, sector_t sectors, int raid_disks)
4998
{
4999
	struct r5conf *conf = mddev->private;
5000 5001 5002

	if (!sectors)
		sectors = mddev->dev_sectors;
5003
	if (!raid_disks)
5004
		/* size is defined by the smallest of previous and new size */
5005
		raid_disks = min(conf->raid_disks, conf->previous_raid_disks);
5006

5007
	sectors &= ~((sector_t)mddev->chunk_sectors - 1);
5008
	sectors &= ~((sector_t)mddev->new_chunk_sectors - 1);
5009 5010 5011
	return sectors * (raid_disks - conf->max_degraded);
}

5012
static void raid5_free_percpu(struct r5conf *conf)
5013 5014 5015 5016 5017 5018 5019 5020 5021 5022 5023
{
	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);
5024
		kfree(percpu->scribble);
5025 5026 5027 5028 5029 5030 5031 5032 5033
	}
#ifdef CONFIG_HOTPLUG_CPU
	unregister_cpu_notifier(&conf->cpu_notify);
#endif
	put_online_cpus();

	free_percpu(conf->percpu);
}

5034
static void free_conf(struct r5conf *conf)
5035 5036
{
	shrink_stripes(conf);
5037
	raid5_free_percpu(conf);
5038 5039 5040 5041 5042
	kfree(conf->disks);
	kfree(conf->stripe_hashtbl);
	kfree(conf);
}

5043 5044 5045 5046
#ifdef CONFIG_HOTPLUG_CPU
static int raid456_cpu_notify(struct notifier_block *nfb, unsigned long action,
			      void *hcpu)
{
5047
	struct r5conf *conf = container_of(nfb, struct r5conf, cpu_notify);
5048 5049 5050 5051 5052 5053
	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:
5054
		if (conf->level == 6 && !percpu->spare_page)
5055
			percpu->spare_page = alloc_page(GFP_KERNEL);
5056 5057 5058 5059 5060 5061 5062
		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);
5063 5064
			pr_err("%s: failed memory allocation for cpu%ld\n",
			       __func__, cpu);
5065
			return notifier_from_errno(-ENOMEM);
5066 5067 5068 5069 5070
		}
		break;
	case CPU_DEAD:
	case CPU_DEAD_FROZEN:
		safe_put_page(percpu->spare_page);
5071
		kfree(percpu->scribble);
5072
		percpu->spare_page = NULL;
5073
		percpu->scribble = NULL;
5074 5075 5076 5077 5078 5079 5080 5081
		break;
	default:
		break;
	}
	return NOTIFY_OK;
}
#endif

5082
static int raid5_alloc_percpu(struct r5conf *conf)
5083 5084 5085
{
	unsigned long cpu;
	struct page *spare_page;
5086
	struct raid5_percpu __percpu *allcpus;
5087
	void *scribble;
5088 5089 5090 5091 5092 5093 5094 5095 5096 5097
	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) {
5098 5099 5100 5101 5102 5103 5104 5105
		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;
		}
5106
		scribble = kmalloc(conf->scribble_len, GFP_KERNEL);
5107
		if (!scribble) {
5108 5109 5110
			err = -ENOMEM;
			break;
		}
5111
		per_cpu_ptr(conf->percpu, cpu)->scribble = scribble;
5112 5113 5114 5115 5116 5117 5118 5119 5120 5121 5122 5123
	}
#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;
}

5124
static struct r5conf *setup_conf(struct mddev *mddev)
L
Linus Torvalds 已提交
5125
{
5126
	struct r5conf *conf;
5127
	int raid_disk, memory, max_disks;
5128
	struct md_rdev *rdev;
L
Linus Torvalds 已提交
5129
	struct disk_info *disk;
5130
	char pers_name[6];
L
Linus Torvalds 已提交
5131

N
NeilBrown 已提交
5132 5133 5134
	if (mddev->new_level != 5
	    && mddev->new_level != 4
	    && mddev->new_level != 6) {
5135
		printk(KERN_ERR "md/raid:%s: raid level not set to 4/5/6 (%d)\n",
N
NeilBrown 已提交
5136 5137
		       mdname(mddev), mddev->new_level);
		return ERR_PTR(-EIO);
L
Linus Torvalds 已提交
5138
	}
N
NeilBrown 已提交
5139 5140 5141 5142
	if ((mddev->new_level == 5
	     && !algorithm_valid_raid5(mddev->new_layout)) ||
	    (mddev->new_level == 6
	     && !algorithm_valid_raid6(mddev->new_layout))) {
5143
		printk(KERN_ERR "md/raid:%s: layout %d not supported\n",
N
NeilBrown 已提交
5144 5145
		       mdname(mddev), mddev->new_layout);
		return ERR_PTR(-EIO);
5146
	}
N
NeilBrown 已提交
5147
	if (mddev->new_level == 6 && mddev->raid_disks < 4) {
5148
		printk(KERN_ERR "md/raid:%s: not enough configured devices (%d, minimum 4)\n",
N
NeilBrown 已提交
5149 5150
		       mdname(mddev), mddev->raid_disks);
		return ERR_PTR(-EINVAL);
5151 5152
	}

5153 5154 5155
	if (!mddev->new_chunk_sectors ||
	    (mddev->new_chunk_sectors << 9) % PAGE_SIZE ||
	    !is_power_of_2(mddev->new_chunk_sectors)) {
5156 5157
		printk(KERN_ERR "md/raid:%s: invalid chunk size %d\n",
		       mdname(mddev), mddev->new_chunk_sectors << 9);
N
NeilBrown 已提交
5158
		return ERR_PTR(-EINVAL);
5159 5160
	}

5161
	conf = kzalloc(sizeof(struct r5conf), GFP_KERNEL);
N
NeilBrown 已提交
5162
	if (conf == NULL)
L
Linus Torvalds 已提交
5163
		goto abort;
5164 5165 5166 5167 5168 5169 5170 5171 5172 5173 5174 5175
	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;
5176
	conf->recovery_disabled = mddev->recovery_disabled - 1;
N
NeilBrown 已提交
5177 5178 5179 5180 5181

	conf->raid_disks = mddev->raid_disks;
	if (mddev->reshape_position == MaxSector)
		conf->previous_raid_disks = mddev->raid_disks;
	else
5182
		conf->previous_raid_disks = mddev->raid_disks - mddev->delta_disks;
5183 5184
	max_disks = max(conf->raid_disks, conf->previous_raid_disks);
	conf->scribble_len = scribble_len(max_disks);
5185

5186
	conf->disks = kzalloc(max_disks * sizeof(struct disk_info),
5187 5188 5189
			      GFP_KERNEL);
	if (!conf->disks)
		goto abort;
5190

L
Linus Torvalds 已提交
5191 5192
	conf->mddev = mddev;

5193
	if ((conf->stripe_hashtbl = kzalloc(PAGE_SIZE, GFP_KERNEL)) == NULL)
L
Linus Torvalds 已提交
5194 5195
		goto abort;

5196 5197 5198 5199
	conf->level = mddev->new_level;
	if (raid5_alloc_percpu(conf) != 0)
		goto abort;

5200
	pr_debug("raid456: run(%s) called.\n", mdname(mddev));
L
Linus Torvalds 已提交
5201

N
NeilBrown 已提交
5202
	rdev_for_each(rdev, mddev) {
L
Linus Torvalds 已提交
5203
		raid_disk = rdev->raid_disk;
5204
		if (raid_disk >= max_disks
L
Linus Torvalds 已提交
5205 5206 5207 5208
		    || raid_disk < 0)
			continue;
		disk = conf->disks + raid_disk;

5209 5210 5211 5212 5213 5214 5215 5216 5217
		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
5218

5219
		if (test_bit(In_sync, &rdev->flags)) {
L
Linus Torvalds 已提交
5220
			char b[BDEVNAME_SIZE];
5221 5222 5223
			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
5224
		} else if (rdev->saved_raid_disk != raid_disk)
5225 5226
			/* Cannot rely on bitmap to complete recovery */
			conf->fullsync = 1;
L
Linus Torvalds 已提交
5227 5228
	}

5229
	conf->chunk_sectors = mddev->new_chunk_sectors;
N
NeilBrown 已提交
5230
	conf->level = mddev->new_level;
5231 5232 5233 5234
	if (conf->level == 6)
		conf->max_degraded = 2;
	else
		conf->max_degraded = 1;
N
NeilBrown 已提交
5235
	conf->algorithm = mddev->new_layout;
L
Linus Torvalds 已提交
5236
	conf->max_nr_stripes = NR_STRIPES;
5237
	conf->reshape_progress = mddev->reshape_position;
5238
	if (conf->reshape_progress != MaxSector) {
5239
		conf->prev_chunk_sectors = mddev->chunk_sectors;
5240 5241
		conf->prev_algo = mddev->layout;
	}
L
Linus Torvalds 已提交
5242

N
NeilBrown 已提交
5243
	memory = conf->max_nr_stripes * (sizeof(struct stripe_head) +
5244
		 max_disks * ((sizeof(struct bio) + PAGE_SIZE))) / 1024;
N
NeilBrown 已提交
5245 5246
	if (grow_stripes(conf, conf->max_nr_stripes)) {
		printk(KERN_ERR
5247 5248
		       "md/raid:%s: couldn't allocate %dkB for buffers\n",
		       mdname(mddev), memory);
N
NeilBrown 已提交
5249 5250
		goto abort;
	} else
5251 5252
		printk(KERN_INFO "md/raid:%s: allocated %dkB\n",
		       mdname(mddev), memory);
L
Linus Torvalds 已提交
5253

5254 5255
	sprintf(pers_name, "raid%d", mddev->new_level);
	conf->thread = md_register_thread(raid5d, mddev, pers_name);
N
NeilBrown 已提交
5256 5257
	if (!conf->thread) {
		printk(KERN_ERR
5258
		       "md/raid:%s: couldn't allocate thread.\n",
N
NeilBrown 已提交
5259
		       mdname(mddev));
5260 5261
		goto abort;
	}
N
NeilBrown 已提交
5262 5263 5264 5265 5266

	return conf;

 abort:
	if (conf) {
5267
		free_conf(conf);
N
NeilBrown 已提交
5268 5269 5270 5271 5272
		return ERR_PTR(-EIO);
	} else
		return ERR_PTR(-ENOMEM);
}

5273 5274 5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291 5292 5293 5294 5295 5296 5297 5298 5299

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;
}

5300
static int run(struct mddev *mddev)
N
NeilBrown 已提交
5301
{
5302
	struct r5conf *conf;
5303
	int working_disks = 0;
5304
	int dirty_parity_disks = 0;
5305
	struct md_rdev *rdev;
5306
	sector_t reshape_offset = 0;
5307
	int i;
5308 5309
	long long min_offset_diff = 0;
	int first = 1;
N
NeilBrown 已提交
5310

5311
	if (mddev->recovery_cp != MaxSector)
5312
		printk(KERN_NOTICE "md/raid:%s: not clean"
5313 5314
		       " -- starting background reconstruction\n",
		       mdname(mddev));
5315 5316 5317 5318 5319 5320 5321 5322 5323 5324 5325 5326 5327 5328 5329 5330 5331

	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
5332 5333
	if (mddev->reshape_position != MaxSector) {
		/* Check that we can continue the reshape.
5334 5335 5336 5337 5338 5339 5340 5341 5342 5343
		 * 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
5344 5345 5346
		 */
		sector_t here_new, here_old;
		int old_disks;
5347
		int max_degraded = (mddev->level == 6 ? 2 : 1);
N
NeilBrown 已提交
5348

5349
		if (mddev->new_level != mddev->level) {
5350
			printk(KERN_ERR "md/raid:%s: unsupported reshape "
N
NeilBrown 已提交
5351 5352 5353 5354 5355 5356 5357 5358 5359 5360
			       "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;
5361
		if (sector_div(here_new, mddev->new_chunk_sectors *
N
NeilBrown 已提交
5362
			       (mddev->raid_disks - max_degraded))) {
5363 5364
			printk(KERN_ERR "md/raid:%s: reshape_position not "
			       "on a stripe boundary\n", mdname(mddev));
N
NeilBrown 已提交
5365 5366
			return -EINVAL;
		}
5367
		reshape_offset = here_new * mddev->new_chunk_sectors;
N
NeilBrown 已提交
5368 5369
		/* here_new is the stripe we will write to */
		here_old = mddev->reshape_position;
5370
		sector_div(here_old, mddev->chunk_sectors *
N
NeilBrown 已提交
5371 5372 5373
			   (old_disks-max_degraded));
		/* here_old is the first stripe that we might need to read
		 * from */
5374
		if (mddev->delta_disks == 0) {
5375 5376 5377 5378 5379 5380
			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;
			}
5381
			/* We cannot be sure it is safe to start an in-place
5382
			 * reshape.  It is only safe if user-space is monitoring
5383 5384 5385 5386 5387
			 * 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.
			 */
5388 5389 5390 5391 5392 5393 5394
			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",
5395
				       mdname(mddev));
5396 5397
				return -EINVAL;
			}
5398
		} else if (mddev->reshape_backwards
5399
		    ? (here_new * mddev->new_chunk_sectors + min_offset_diff <=
5400 5401
		       here_old * mddev->chunk_sectors)
		    : (here_new * mddev->new_chunk_sectors >=
5402
		       here_old * mddev->chunk_sectors + (-min_offset_diff))) {
N
NeilBrown 已提交
5403
			/* Reading from the same stripe as writing to - bad */
5404 5405 5406
			printk(KERN_ERR "md/raid:%s: reshape_position too early for "
			       "auto-recovery - aborting.\n",
			       mdname(mddev));
N
NeilBrown 已提交
5407 5408
			return -EINVAL;
		}
5409 5410
		printk(KERN_INFO "md/raid:%s: reshape will continue\n",
		       mdname(mddev));
N
NeilBrown 已提交
5411 5412 5413 5414
		/* OK, we should be able to continue; */
	} else {
		BUG_ON(mddev->level != mddev->new_level);
		BUG_ON(mddev->layout != mddev->new_layout);
5415
		BUG_ON(mddev->chunk_sectors != mddev->new_chunk_sectors);
N
NeilBrown 已提交
5416
		BUG_ON(mddev->delta_disks != 0);
L
Linus Torvalds 已提交
5417
	}
N
NeilBrown 已提交
5418

5419 5420 5421 5422 5423
	if (mddev->private == NULL)
		conf = setup_conf(mddev);
	else
		conf = mddev->private;

N
NeilBrown 已提交
5424 5425 5426
	if (IS_ERR(conf))
		return PTR_ERR(conf);

5427
	conf->min_offset_diff = min_offset_diff;
N
NeilBrown 已提交
5428 5429 5430 5431
	mddev->thread = conf->thread;
	conf->thread = NULL;
	mddev->private = conf;

5432 5433 5434 5435 5436 5437 5438 5439 5440 5441 5442
	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)
5443
			continue;
5444 5445 5446 5447 5448 5449 5450
		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;
		}
5451
		if (test_bit(In_sync, &rdev->flags)) {
N
NeilBrown 已提交
5452
			working_disks++;
5453 5454
			continue;
		}
5455 5456 5457 5458 5459 5460 5461 5462 5463 5464 5465 5466 5467 5468 5469 5470 5471 5472 5473 5474 5475 5476 5477 5478 5479 5480 5481 5482
		/* 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;
			
		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
5483

5484 5485 5486
	/*
	 * 0 for a fully functional array, 1 or 2 for a degraded array.
	 */
5487
	mddev->degraded = calc_degraded(conf);
N
NeilBrown 已提交
5488

5489
	if (has_failed(conf)) {
5490
		printk(KERN_ERR "md/raid:%s: not enough operational devices"
L
Linus Torvalds 已提交
5491
			" (%d/%d failed)\n",
5492
			mdname(mddev), mddev->degraded, conf->raid_disks);
L
Linus Torvalds 已提交
5493 5494 5495
		goto abort;
	}

N
NeilBrown 已提交
5496
	/* device size must be a multiple of chunk size */
5497
	mddev->dev_sectors &= ~(mddev->chunk_sectors - 1);
N
NeilBrown 已提交
5498 5499
	mddev->resync_max_sectors = mddev->dev_sectors;

5500
	if (mddev->degraded > dirty_parity_disks &&
L
Linus Torvalds 已提交
5501
	    mddev->recovery_cp != MaxSector) {
5502 5503
		if (mddev->ok_start_degraded)
			printk(KERN_WARNING
5504 5505
			       "md/raid:%s: starting dirty degraded array"
			       " - data corruption possible.\n",
5506 5507 5508
			       mdname(mddev));
		else {
			printk(KERN_ERR
5509
			       "md/raid:%s: cannot start dirty degraded array.\n",
5510 5511 5512
			       mdname(mddev));
			goto abort;
		}
L
Linus Torvalds 已提交
5513 5514 5515
	}

	if (mddev->degraded == 0)
5516 5517
		printk(KERN_INFO "md/raid:%s: raid level %d active with %d out of %d"
		       " devices, algorithm %d\n", mdname(mddev), conf->level,
5518 5519
		       mddev->raid_disks-mddev->degraded, mddev->raid_disks,
		       mddev->new_layout);
L
Linus Torvalds 已提交
5520
	else
5521 5522 5523 5524 5525
		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
5526 5527 5528

	print_raid5_conf(conf);

5529 5530
	if (conf->reshape_progress != MaxSector) {
		conf->reshape_safe = conf->reshape_progress;
5531 5532 5533 5534 5535 5536
		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,
5537
							"reshape");
5538 5539
	}

L
Linus Torvalds 已提交
5540 5541

	/* Ok, everything is just fine now */
5542 5543
	if (mddev->to_remove == &raid5_attrs_group)
		mddev->to_remove = NULL;
N
NeilBrown 已提交
5544 5545
	else if (mddev->kobj.sd &&
	    sysfs_create_group(&mddev->kobj, &raid5_attrs_group))
5546
		printk(KERN_WARNING
5547
		       "raid5: failed to create sysfs attributes for %s\n",
5548
		       mdname(mddev));
5549
	md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
5550

5551
	if (mddev->queue) {
5552
		int chunk_size;
S
Shaohua Li 已提交
5553
		bool discard_supported = true;
5554 5555 5556 5557 5558 5559 5560 5561 5562
		/* 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
5563

5564
		blk_queue_merge_bvec(mddev->queue, raid5_mergeable_bvec);
5565

N
NeilBrown 已提交
5566 5567
		mddev->queue->backing_dev_info.congested_data = mddev;
		mddev->queue->backing_dev_info.congested_fn = raid5_congested;
5568

5569 5570 5571 5572
		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
5573 5574 5575 5576 5577
		/*
		 * We can only discard a whole stripe. It doesn't make sense to
		 * discard data disk but write parity disk
		 */
		stripe = stripe * PAGE_SIZE;
5578 5579 5580 5581
		/* 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
5582 5583 5584 5585 5586 5587 5588
		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;
5589

5590
		rdev_for_each(rdev, mddev) {
5591 5592
			disk_stack_limits(mddev->gendisk, rdev->bdev,
					  rdev->data_offset << 9);
5593 5594
			disk_stack_limits(mddev->gendisk, rdev->bdev,
					  rdev->new_data_offset << 9);
S
Shaohua Li 已提交
5595 5596 5597 5598 5599 5600 5601 5602 5603 5604 5605 5606 5607 5608
			/*
			 * 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;
5609
		}
S
Shaohua Li 已提交
5610 5611 5612 5613 5614 5615 5616 5617 5618

		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);
5619
	}
5620

L
Linus Torvalds 已提交
5621 5622
	return 0;
abort:
5623
	md_unregister_thread(&mddev->thread);
N
NeilBrown 已提交
5624 5625
	print_raid5_conf(conf);
	free_conf(conf);
L
Linus Torvalds 已提交
5626
	mddev->private = NULL;
5627
	printk(KERN_ALERT "md/raid:%s: failed to run raid set.\n", mdname(mddev));
L
Linus Torvalds 已提交
5628 5629 5630
	return -EIO;
}

5631
static int stop(struct mddev *mddev)
L
Linus Torvalds 已提交
5632
{
5633
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5634

5635
	md_unregister_thread(&mddev->thread);
N
NeilBrown 已提交
5636 5637
	if (mddev->queue)
		mddev->queue->backing_dev_info.congested_fn = NULL;
5638
	free_conf(conf);
5639 5640
	mddev->private = NULL;
	mddev->to_remove = &raid5_attrs_group;
L
Linus Torvalds 已提交
5641 5642 5643
	return 0;
}

5644
static void status(struct seq_file *seq, struct mddev *mddev)
L
Linus Torvalds 已提交
5645
{
5646
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5647 5648
	int i;

5649 5650
	seq_printf(seq, " level %d, %dk chunk, algorithm %d", mddev->level,
		mddev->chunk_sectors / 2, mddev->layout);
5651
	seq_printf (seq, " [%d/%d] [", conf->raid_disks, conf->raid_disks - mddev->degraded);
L
Linus Torvalds 已提交
5652 5653 5654
	for (i = 0; i < conf->raid_disks; i++)
		seq_printf (seq, "%s",
			       conf->disks[i].rdev &&
5655
			       test_bit(In_sync, &conf->disks[i].rdev->flags) ? "U" : "_");
L
Linus Torvalds 已提交
5656 5657 5658
	seq_printf (seq, "]");
}

5659
static void print_raid5_conf (struct r5conf *conf)
L
Linus Torvalds 已提交
5660 5661 5662 5663
{
	int i;
	struct disk_info *tmp;

5664
	printk(KERN_DEBUG "RAID conf printout:\n");
L
Linus Torvalds 已提交
5665 5666 5667 5668
	if (!conf) {
		printk("(conf==NULL)\n");
		return;
	}
5669 5670 5671
	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
5672 5673 5674 5675 5676

	for (i = 0; i < conf->raid_disks; i++) {
		char b[BDEVNAME_SIZE];
		tmp = conf->disks + i;
		if (tmp->rdev)
5677 5678 5679
			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
5680 5681 5682
	}
}

5683
static int raid5_spare_active(struct mddev *mddev)
L
Linus Torvalds 已提交
5684 5685
{
	int i;
5686
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5687
	struct disk_info *tmp;
5688 5689
	int count = 0;
	unsigned long flags;
L
Linus Torvalds 已提交
5690 5691 5692

	for (i = 0; i < conf->raid_disks; i++) {
		tmp = conf->disks + i;
5693 5694 5695 5696 5697 5698 5699 5700 5701 5702 5703 5704 5705 5706 5707 5708 5709 5710 5711
		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
5712
		    && tmp->rdev->recovery_offset == MaxSector
5713
		    && !test_bit(Faulty, &tmp->rdev->flags)
5714
		    && !test_and_set_bit(In_sync, &tmp->rdev->flags)) {
5715
			count++;
5716
			sysfs_notify_dirent_safe(tmp->rdev->sysfs_state);
L
Linus Torvalds 已提交
5717 5718
		}
	}
5719
	spin_lock_irqsave(&conf->device_lock, flags);
5720
	mddev->degraded = calc_degraded(conf);
5721
	spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
5722
	print_raid5_conf(conf);
5723
	return count;
L
Linus Torvalds 已提交
5724 5725
}

5726
static int raid5_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
5727
{
5728
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5729
	int err = 0;
5730
	int number = rdev->raid_disk;
5731
	struct md_rdev **rdevp;
L
Linus Torvalds 已提交
5732 5733 5734
	struct disk_info *p = conf->disks + number;

	print_raid5_conf(conf);
5735 5736 5737 5738 5739 5740 5741 5742 5743 5744 5745 5746 5747 5748 5749 5750 5751 5752 5753 5754 5755 5756
	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) &&
5757
	    (!p->replacement || p->replacement == rdev) &&
5758 5759 5760 5761 5762 5763 5764 5765 5766 5767
	    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;
5768 5769 5770 5771 5772 5773 5774 5775 5776 5777 5778 5779 5780 5781
	} 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
5782 5783 5784 5785 5786 5787
abort:

	print_raid5_conf(conf);
	return err;
}

5788
static int raid5_add_disk(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
5789
{
5790
	struct r5conf *conf = mddev->private;
5791
	int err = -EEXIST;
L
Linus Torvalds 已提交
5792 5793
	int disk;
	struct disk_info *p;
5794 5795
	int first = 0;
	int last = conf->raid_disks - 1;
L
Linus Torvalds 已提交
5796

5797 5798 5799
	if (mddev->recovery_disabled == conf->recovery_disabled)
		return -EBUSY;

N
NeilBrown 已提交
5800
	if (rdev->saved_raid_disk < 0 && has_failed(conf))
L
Linus Torvalds 已提交
5801
		/* no point adding a device */
5802
		return -EINVAL;
L
Linus Torvalds 已提交
5803

5804 5805
	if (rdev->raid_disk >= 0)
		first = last = rdev->raid_disk;
L
Linus Torvalds 已提交
5806 5807

	/*
5808 5809
	 * find the disk ... but prefer rdev->saved_raid_disk
	 * if possible.
L
Linus Torvalds 已提交
5810
	 */
5811
	if (rdev->saved_raid_disk >= 0 &&
5812
	    rdev->saved_raid_disk >= first &&
5813
	    conf->disks[rdev->saved_raid_disk].rdev == NULL)
5814 5815 5816
		first = rdev->saved_raid_disk;

	for (disk = first; disk <= last; disk++) {
5817 5818
		p = conf->disks + disk;
		if (p->rdev == NULL) {
5819
			clear_bit(In_sync, &rdev->flags);
L
Linus Torvalds 已提交
5820
			rdev->raid_disk = disk;
5821
			err = 0;
5822 5823
			if (rdev->saved_raid_disk != disk)
				conf->fullsync = 1;
5824
			rcu_assign_pointer(p->rdev, rdev);
5825
			goto out;
L
Linus Torvalds 已提交
5826
		}
5827 5828 5829
	}
	for (disk = first; disk <= last; disk++) {
		p = conf->disks + disk;
5830 5831 5832 5833 5834 5835 5836 5837 5838 5839 5840
		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;
		}
	}
5841
out:
L
Linus Torvalds 已提交
5842
	print_raid5_conf(conf);
5843
	return err;
L
Linus Torvalds 已提交
5844 5845
}

5846
static int raid5_resize(struct mddev *mddev, sector_t sectors)
L
Linus Torvalds 已提交
5847 5848 5849 5850 5851 5852 5853 5854
{
	/* 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.
	 */
5855
	sector_t newsize;
5856
	sectors &= ~((sector_t)mddev->chunk_sectors - 1);
5857 5858 5859
	newsize = raid5_size(mddev, sectors, mddev->raid_disks);
	if (mddev->external_size &&
	    mddev->array_sectors > newsize)
D
Dan Williams 已提交
5860
		return -EINVAL;
5861 5862 5863 5864 5865 5866
	if (mddev->bitmap) {
		int ret = bitmap_resize(mddev->bitmap, sectors, 0, 0);
		if (ret)
			return ret;
	}
	md_set_array_sectors(mddev, newsize);
5867
	set_capacity(mddev->gendisk, mddev->array_sectors);
5868
	revalidate_disk(mddev->gendisk);
5869 5870
	if (sectors > mddev->dev_sectors &&
	    mddev->recovery_cp > mddev->dev_sectors) {
A
Andre Noll 已提交
5871
		mddev->recovery_cp = mddev->dev_sectors;
L
Linus Torvalds 已提交
5872 5873
		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
	}
A
Andre Noll 已提交
5874
	mddev->dev_sectors = sectors;
5875
	mddev->resync_max_sectors = sectors;
L
Linus Torvalds 已提交
5876 5877 5878
	return 0;
}

5879
static int check_stripe_cache(struct mddev *mddev)
5880 5881 5882 5883 5884 5885 5886 5887 5888
{
	/* 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.
	 */
5889
	struct r5conf *conf = mddev->private;
5890 5891 5892 5893
	if (((mddev->chunk_sectors << 9) / STRIPE_SIZE) * 4
	    > conf->max_nr_stripes ||
	    ((mddev->new_chunk_sectors << 9) / STRIPE_SIZE) * 4
	    > conf->max_nr_stripes) {
5894 5895
		printk(KERN_WARNING "md/raid:%s: reshape: not enough stripes.  Needed %lu\n",
		       mdname(mddev),
5896 5897 5898 5899 5900 5901 5902
		       ((max(mddev->chunk_sectors, mddev->new_chunk_sectors) << 9)
			/ STRIPE_SIZE)*4);
		return 0;
	}
	return 1;
}

5903
static int check_reshape(struct mddev *mddev)
5904
{
5905
	struct r5conf *conf = mddev->private;
5906

5907 5908
	if (mddev->delta_disks == 0 &&
	    mddev->new_layout == mddev->layout &&
5909
	    mddev->new_chunk_sectors == mddev->chunk_sectors)
5910
		return 0; /* nothing to do */
5911
	if (has_failed(conf))
5912 5913 5914 5915 5916 5917 5918 5919 5920 5921 5922 5923 5924
		return -EINVAL;
	if (mddev->delta_disks < 0) {
		/* 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;
	}
5925

5926
	if (!check_stripe_cache(mddev))
5927 5928
		return -ENOSPC;

5929 5930
	return resize_stripes(conf, (conf->previous_raid_disks
				     + mddev->delta_disks));
5931 5932
}

5933
static int raid5_start_reshape(struct mddev *mddev)
5934
{
5935
	struct r5conf *conf = mddev->private;
5936
	struct md_rdev *rdev;
5937
	int spares = 0;
5938
	unsigned long flags;
5939

5940
	if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
5941 5942
		return -EBUSY;

5943 5944 5945
	if (!check_stripe_cache(mddev))
		return -ENOSPC;

5946 5947 5948
	if (has_failed(conf))
		return -EINVAL;

5949
	rdev_for_each(rdev, mddev) {
5950 5951
		if (!test_bit(In_sync, &rdev->flags)
		    && !test_bit(Faulty, &rdev->flags))
5952
			spares++;
5953
	}
5954

5955
	if (spares - mddev->degraded < mddev->delta_disks - conf->max_degraded)
5956 5957 5958 5959 5960
		/* Not enough devices even to make a degraded array
		 * of that size
		 */
		return -EINVAL;

5961 5962 5963 5964 5965 5966
	/* 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) {
5967
		printk(KERN_ERR "md/raid:%s: array size must be reduced "
5968 5969 5970 5971
		       "before number of disks\n", mdname(mddev));
		return -EINVAL;
	}

5972
	atomic_set(&conf->reshape_stripes, 0);
5973 5974
	spin_lock_irq(&conf->device_lock);
	conf->previous_raid_disks = conf->raid_disks;
5975
	conf->raid_disks += mddev->delta_disks;
5976 5977
	conf->prev_chunk_sectors = conf->chunk_sectors;
	conf->chunk_sectors = mddev->new_chunk_sectors;
5978 5979
	conf->prev_algo = conf->algorithm;
	conf->algorithm = mddev->new_layout;
5980 5981 5982 5983 5984
	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();
5985
	if (mddev->reshape_backwards)
5986 5987 5988 5989
		conf->reshape_progress = raid5_size(mddev, 0, 0);
	else
		conf->reshape_progress = 0;
	conf->reshape_safe = conf->reshape_progress;
5990 5991 5992 5993
	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.
5994 5995 5996 5997
	 * 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.
5998
	 */
5999
	if (mddev->delta_disks >= 0) {
N
NeilBrown 已提交
6000
		rdev_for_each(rdev, mddev)
6001 6002 6003 6004
			if (rdev->raid_disk < 0 &&
			    !test_bit(Faulty, &rdev->flags)) {
				if (raid5_add_disk(mddev, rdev) == 0) {
					if (rdev->raid_disk
6005
					    >= conf->previous_raid_disks)
6006
						set_bit(In_sync, &rdev->flags);
6007
					else
6008
						rdev->recovery_offset = 0;
6009 6010

					if (sysfs_link_rdev(mddev, rdev))
6011
						/* Failure here is OK */;
6012
				}
6013 6014 6015 6016 6017
			} 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);
			}
6018

6019 6020 6021 6022
		/* When a reshape changes the number of devices,
		 * ->degraded is measured against the larger of the
		 * pre and post number of devices.
		 */
6023
		spin_lock_irqsave(&conf->device_lock, flags);
6024
		mddev->degraded = calc_degraded(conf);
6025 6026
		spin_unlock_irqrestore(&conf->device_lock, flags);
	}
6027
	mddev->raid_disks = conf->raid_disks;
6028
	mddev->reshape_position = conf->reshape_progress;
6029
	set_bit(MD_CHANGE_DEVS, &mddev->flags);
6030

6031 6032 6033 6034 6035
	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,
6036
						"reshape");
6037 6038 6039 6040
	if (!mddev->sync_thread) {
		mddev->recovery = 0;
		spin_lock_irq(&conf->device_lock);
		mddev->raid_disks = conf->raid_disks = conf->previous_raid_disks;
6041 6042 6043
		rdev_for_each(rdev, mddev)
			rdev->new_data_offset = rdev->data_offset;
		smp_wmb();
6044
		conf->reshape_progress = MaxSector;
6045
		mddev->reshape_position = MaxSector;
6046 6047 6048
		spin_unlock_irq(&conf->device_lock);
		return -EAGAIN;
	}
6049
	conf->reshape_checkpoint = jiffies;
6050 6051 6052 6053 6054
	md_wakeup_thread(mddev->sync_thread);
	md_new_event(mddev);
	return 0;
}

6055 6056 6057
/* This is called from the reshape thread and should make any
 * changes needed in 'conf'
 */
6058
static void end_reshape(struct r5conf *conf)
6059 6060
{

6061
	if (!test_bit(MD_RECOVERY_INTR, &conf->mddev->recovery)) {
6062
		struct md_rdev *rdev;
6063 6064

		spin_lock_irq(&conf->device_lock);
6065
		conf->previous_raid_disks = conf->raid_disks;
6066 6067 6068
		rdev_for_each(rdev, conf->mddev)
			rdev->data_offset = rdev->new_data_offset;
		smp_wmb();
6069
		conf->reshape_progress = MaxSector;
6070
		spin_unlock_irq(&conf->device_lock);
6071
		wake_up(&conf->wait_for_overlap);
6072 6073 6074 6075

		/* read-ahead size must cover two whole stripes, which is
		 * 2 * (datadisks) * chunksize where 'n' is the number of raid devices
		 */
6076
		if (conf->mddev->queue) {
6077
			int data_disks = conf->raid_disks - conf->max_degraded;
6078
			int stripe = data_disks * ((conf->chunk_sectors << 9)
6079
						   / PAGE_SIZE);
6080 6081 6082
			if (conf->mddev->queue->backing_dev_info.ra_pages < 2 * stripe)
				conf->mddev->queue->backing_dev_info.ra_pages = 2 * stripe;
		}
6083 6084 6085
	}
}

6086 6087 6088
/* This is called from the raid5d thread with mddev_lock held.
 * It makes config changes to the device.
 */
6089
static void raid5_finish_reshape(struct mddev *mddev)
6090
{
6091
	struct r5conf *conf = mddev->private;
6092 6093 6094

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

6095 6096 6097
		if (mddev->delta_disks > 0) {
			md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
			set_capacity(mddev->gendisk, mddev->array_sectors);
6098
			revalidate_disk(mddev->gendisk);
6099 6100
		} else {
			int d;
6101 6102 6103
			spin_lock_irq(&conf->device_lock);
			mddev->degraded = calc_degraded(conf);
			spin_unlock_irq(&conf->device_lock);
6104 6105
			for (d = conf->raid_disks ;
			     d < conf->raid_disks - mddev->delta_disks;
6106
			     d++) {
6107
				struct md_rdev *rdev = conf->disks[d].rdev;
6108 6109 6110 6111 6112
				if (rdev)
					clear_bit(In_sync, &rdev->flags);
				rdev = conf->disks[d].replacement;
				if (rdev)
					clear_bit(In_sync, &rdev->flags);
6113
			}
6114
		}
6115
		mddev->layout = conf->algorithm;
6116
		mddev->chunk_sectors = conf->chunk_sectors;
6117 6118
		mddev->reshape_position = MaxSector;
		mddev->delta_disks = 0;
6119
		mddev->reshape_backwards = 0;
6120 6121 6122
	}
}

6123
static void raid5_quiesce(struct mddev *mddev, int state)
6124
{
6125
	struct r5conf *conf = mddev->private;
6126 6127

	switch(state) {
6128 6129 6130 6131
	case 2: /* resume for a suspend */
		wake_up(&conf->wait_for_overlap);
		break;

6132 6133
	case 1: /* stop all writes */
		spin_lock_irq(&conf->device_lock);
6134 6135 6136 6137
		/* '2' tells resync/reshape to pause so that all
		 * active stripes can drain
		 */
		conf->quiesce = 2;
6138
		wait_event_lock_irq(conf->wait_for_stripe,
6139 6140
				    atomic_read(&conf->active_stripes) == 0 &&
				    atomic_read(&conf->active_aligned_reads) == 0,
6141
				    conf->device_lock);
6142
		conf->quiesce = 1;
6143
		spin_unlock_irq(&conf->device_lock);
6144 6145
		/* allow reshape to continue */
		wake_up(&conf->wait_for_overlap);
6146 6147 6148 6149 6150 6151
		break;

	case 0: /* re-enable writes */
		spin_lock_irq(&conf->device_lock);
		conf->quiesce = 0;
		wake_up(&conf->wait_for_stripe);
6152
		wake_up(&conf->wait_for_overlap);
6153 6154 6155 6156
		spin_unlock_irq(&conf->device_lock);
		break;
	}
}
6157

6158

6159
static void *raid45_takeover_raid0(struct mddev *mddev, int level)
6160
{
6161
	struct r0conf *raid0_conf = mddev->private;
6162
	sector_t sectors;
6163

D
Dan Williams 已提交
6164
	/* for raid0 takeover only one zone is supported */
6165
	if (raid0_conf->nr_strip_zones > 1) {
6166 6167
		printk(KERN_ERR "md/raid:%s: cannot takeover raid0 with more than one zone.\n",
		       mdname(mddev));
D
Dan Williams 已提交
6168 6169 6170
		return ERR_PTR(-EINVAL);
	}

6171 6172
	sectors = raid0_conf->strip_zone[0].zone_end;
	sector_div(sectors, raid0_conf->strip_zone[0].nb_dev);
6173
	mddev->dev_sectors = sectors;
D
Dan Williams 已提交
6174
	mddev->new_level = level;
6175 6176 6177 6178 6179 6180 6181 6182 6183 6184 6185
	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);
}


6186
static void *raid5_takeover_raid1(struct mddev *mddev)
6187 6188 6189 6190 6191 6192 6193 6194 6195 6196 6197 6198 6199 6200 6201 6202 6203 6204 6205 6206 6207
{
	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;
6208
	mddev->new_chunk_sectors = chunksect;
6209 6210 6211 6212

	return setup_conf(mddev);
}

6213
static void *raid5_takeover_raid6(struct mddev *mddev)
6214 6215 6216 6217 6218 6219 6220 6221 6222 6223 6224 6225 6226 6227 6228 6229 6230 6231 6232 6233 6234 6235 6236 6237 6238 6239 6240 6241 6242 6243 6244 6245
{
	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);
}

6246

6247
static int raid5_check_reshape(struct mddev *mddev)
6248
{
6249 6250 6251 6252
	/* 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.
6253
	 */
6254
	struct r5conf *conf = mddev->private;
6255
	int new_chunk = mddev->new_chunk_sectors;
6256

6257
	if (mddev->new_layout >= 0 && !algorithm_valid_raid5(mddev->new_layout))
6258 6259
		return -EINVAL;
	if (new_chunk > 0) {
6260
		if (!is_power_of_2(new_chunk))
6261
			return -EINVAL;
6262
		if (new_chunk < (PAGE_SIZE>>9))
6263
			return -EINVAL;
6264
		if (mddev->array_sectors & (new_chunk-1))
6265 6266 6267 6268 6269 6270
			/* not factor of array size */
			return -EINVAL;
	}

	/* They look valid */

6271
	if (mddev->raid_disks == 2) {
6272 6273 6274 6275
		/* can make the change immediately */
		if (mddev->new_layout >= 0) {
			conf->algorithm = mddev->new_layout;
			mddev->layout = mddev->new_layout;
6276 6277
		}
		if (new_chunk > 0) {
6278 6279
			conf->chunk_sectors = new_chunk ;
			mddev->chunk_sectors = new_chunk;
6280 6281 6282
		}
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
		md_wakeup_thread(mddev->thread);
6283
	}
6284
	return check_reshape(mddev);
6285 6286
}

6287
static int raid6_check_reshape(struct mddev *mddev)
6288
{
6289
	int new_chunk = mddev->new_chunk_sectors;
6290

6291
	if (mddev->new_layout >= 0 && !algorithm_valid_raid6(mddev->new_layout))
6292
		return -EINVAL;
6293
	if (new_chunk > 0) {
6294
		if (!is_power_of_2(new_chunk))
6295
			return -EINVAL;
6296
		if (new_chunk < (PAGE_SIZE >> 9))
6297
			return -EINVAL;
6298
		if (mddev->array_sectors & (new_chunk-1))
6299 6300
			/* not factor of array size */
			return -EINVAL;
6301
	}
6302 6303

	/* They look valid */
6304
	return check_reshape(mddev);
6305 6306
}

6307
static void *raid5_takeover(struct mddev *mddev)
6308 6309
{
	/* raid5 can take over:
D
Dan Williams 已提交
6310
	 *  raid0 - if there is only one strip zone - make it a raid4 layout
6311 6312 6313 6314
	 *  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
6315 6316
	if (mddev->level == 0)
		return raid45_takeover_raid0(mddev, 5);
6317 6318
	if (mddev->level == 1)
		return raid5_takeover_raid1(mddev);
6319 6320 6321 6322 6323
	if (mddev->level == 4) {
		mddev->new_layout = ALGORITHM_PARITY_N;
		mddev->new_level = 5;
		return setup_conf(mddev);
	}
6324 6325
	if (mddev->level == 6)
		return raid5_takeover_raid6(mddev);
6326 6327 6328 6329

	return ERR_PTR(-EINVAL);
}

6330
static void *raid4_takeover(struct mddev *mddev)
6331
{
D
Dan Williams 已提交
6332 6333 6334
	/* raid4 can take over:
	 *  raid0 - if there is only one strip zone
	 *  raid5 - if layout is right
6335
	 */
D
Dan Williams 已提交
6336 6337
	if (mddev->level == 0)
		return raid45_takeover_raid0(mddev, 4);
6338 6339 6340 6341 6342 6343 6344 6345
	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);
}
6346

6347
static struct md_personality raid5_personality;
6348

6349
static void *raid6_takeover(struct mddev *mddev)
6350 6351 6352 6353 6354 6355 6356 6357 6358 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
{
	/* 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);
}


6396
static struct md_personality raid6_personality =
6397 6398 6399 6400 6401 6402 6403 6404 6405 6406 6407 6408 6409 6410
{
	.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,
6411
	.size		= raid5_size,
6412
	.check_reshape	= raid6_check_reshape,
6413
	.start_reshape  = raid5_start_reshape,
6414
	.finish_reshape = raid5_finish_reshape,
6415
	.quiesce	= raid5_quiesce,
6416
	.takeover	= raid6_takeover,
6417
};
6418
static struct md_personality raid5_personality =
L
Linus Torvalds 已提交
6419 6420
{
	.name		= "raid5",
6421
	.level		= 5,
L
Linus Torvalds 已提交
6422 6423 6424 6425 6426 6427 6428 6429 6430 6431 6432
	.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,
6433
	.size		= raid5_size,
6434 6435
	.check_reshape	= raid5_check_reshape,
	.start_reshape  = raid5_start_reshape,
6436
	.finish_reshape = raid5_finish_reshape,
6437
	.quiesce	= raid5_quiesce,
6438
	.takeover	= raid5_takeover,
L
Linus Torvalds 已提交
6439 6440
};

6441
static struct md_personality raid4_personality =
L
Linus Torvalds 已提交
6442
{
6443 6444 6445 6446 6447 6448 6449 6450 6451 6452 6453 6454 6455
	.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,
6456
	.size		= raid5_size,
6457 6458
	.check_reshape	= raid5_check_reshape,
	.start_reshape  = raid5_start_reshape,
6459
	.finish_reshape = raid5_finish_reshape,
6460
	.quiesce	= raid5_quiesce,
6461
	.takeover	= raid4_takeover,
6462 6463 6464 6465
};

static int __init raid5_init(void)
{
6466
	register_md_personality(&raid6_personality);
6467 6468 6469
	register_md_personality(&raid5_personality);
	register_md_personality(&raid4_personality);
	return 0;
L
Linus Torvalds 已提交
6470 6471
}

6472
static void raid5_exit(void)
L
Linus Torvalds 已提交
6473
{
6474
	unregister_md_personality(&raid6_personality);
6475 6476
	unregister_md_personality(&raid5_personality);
	unregister_md_personality(&raid4_personality);
L
Linus Torvalds 已提交
6477 6478 6479 6480 6481
}

module_init(raid5_init);
module_exit(raid5_exit);
MODULE_LICENSE("GPL");
6482
MODULE_DESCRIPTION("RAID4/5/6 (striping with parity) personality for MD");
L
Linus Torvalds 已提交
6483
MODULE_ALIAS("md-personality-4"); /* RAID5 */
6484 6485
MODULE_ALIAS("md-raid5");
MODULE_ALIAS("md-raid4");
6486 6487
MODULE_ALIAS("md-level-5");
MODULE_ALIAS("md-level-4");
6488 6489 6490 6491 6492 6493 6494
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");