raid5.c 167.2 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>
56
#include "md.h"
57
#include "raid5.h"
58
#include "raid0.h"
59
#include "bitmap.h"
60

L
Linus Torvalds 已提交
61 62 63 64 65 66 67 68 69
/*
 * 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
70
#define BYPASS_THRESHOLD	1
71
#define NR_HASH			(PAGE_SIZE / sizeof(struct hlist_head))
L
Linus Torvalds 已提交
72 73
#define HASH_MASK		(NR_HASH - 1)

74
static inline struct hlist_head *stripe_hash(struct r5conf *conf, sector_t sect)
75 76 77 78
{
	int hash = (sect >> STRIPE_SHIFT) & HASH_MASK;
	return &conf->stripe_hashtbl[hash];
}
L
Linus Torvalds 已提交
79 80 81 82 83 84 85

/* 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.
86
 * This function is used to determine the 'next' bio in the list, given the sector
L
Linus Torvalds 已提交
87 88
 * of the current stripe+device
 */
89 90 91 92 93 94 95 96
static inline struct bio *r5_next_bio(struct bio *bio, sector_t sector)
{
	int sectors = bio->bi_size >> 9;
	if (bio->bi_sector + sectors < sector + STRIPE_SECTORS)
		return bio->bi_next;
	else
		return NULL;
}
L
Linus Torvalds 已提交
97

98
/*
99 100
 * 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
101 102 103
 */
static inline int raid5_bi_phys_segments(struct bio *bio)
{
104
	return bio->bi_phys_segments & 0xffff;
105 106 107 108
}

static inline int raid5_bi_hw_segments(struct bio *bio)
{
109
	return (bio->bi_phys_segments >> 16) & 0xffff;
110 111 112 113 114 115 116 117 118 119 120 121 122
}

static inline int raid5_dec_bi_phys_segments(struct bio *bio)
{
	--bio->bi_phys_segments;
	return raid5_bi_phys_segments(bio);
}

static inline int raid5_dec_bi_hw_segments(struct bio *bio)
{
	unsigned short val = raid5_bi_hw_segments(bio);

	--val;
123
	bio->bi_phys_segments = (val << 16) | raid5_bi_phys_segments(bio);
124 125 126 127 128
	return val;
}

static inline void raid5_set_bi_hw_segments(struct bio *bio, unsigned int cnt)
{
129
	bio->bi_phys_segments = raid5_bi_phys_segments(bio) | (cnt << 16);
130 131
}

132 133 134
/* Find first data disk in a raid6 stripe */
static inline int raid6_d0(struct stripe_head *sh)
{
135 136 137 138
	if (sh->ddf_layout)
		/* ddf always start from first device */
		return 0;
	/* md starts just after Q block */
139 140 141 142 143
	if (sh->qd_idx == sh->disks - 1)
		return 0;
	else
		return sh->qd_idx + 1;
}
144 145 146 147 148
static inline int raid6_next_disk(int disk, int raid_disks)
{
	disk++;
	return (disk < raid_disks) ? disk : 0;
}
149

150 151 152 153 154
/* 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.
 */
155 156
static int raid6_idx_to_slot(int idx, struct stripe_head *sh,
			     int *count, int syndrome_disks)
157
{
158
	int slot = *count;
159

160
	if (sh->ddf_layout)
161
		(*count)++;
162
	if (idx == sh->pd_idx)
163
		return syndrome_disks;
164
	if (idx == sh->qd_idx)
165
		return syndrome_disks + 1;
166
	if (!sh->ddf_layout)
167
		(*count)++;
168 169 170
	return slot;
}

171 172 173 174 175 176 177 178
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;
179
		bio_endio(bi, 0);
180 181 182 183
		bi = return_bi;
	}
}

184
static void print_raid5_conf (struct r5conf *conf);
L
Linus Torvalds 已提交
185

186 187 188 189 190 191 192
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);
}

193
static void __release_stripe(struct r5conf *conf, struct stripe_head *sh)
L
Linus Torvalds 已提交
194 195
{
	if (atomic_dec_and_test(&sh->count)) {
196 197
		BUG_ON(!list_empty(&sh->lru));
		BUG_ON(atomic_read(&conf->active_stripes)==0);
L
Linus Torvalds 已提交
198
		if (test_bit(STRIPE_HANDLE, &sh->state)) {
N
NeilBrown 已提交
199
			if (test_bit(STRIPE_DELAYED, &sh->state))
L
Linus Torvalds 已提交
200
				list_add_tail(&sh->lru, &conf->delayed_list);
N
NeilBrown 已提交
201 202
			else if (test_bit(STRIPE_BIT_DELAY, &sh->state) &&
				   sh->bm_seq - conf->seq_write > 0)
203
				list_add_tail(&sh->lru, &conf->bitmap_list);
N
NeilBrown 已提交
204
			else {
205
				clear_bit(STRIPE_BIT_DELAY, &sh->state);
L
Linus Torvalds 已提交
206
				list_add_tail(&sh->lru, &conf->handle_list);
207
			}
L
Linus Torvalds 已提交
208 209
			md_wakeup_thread(conf->mddev->thread);
		} else {
210
			BUG_ON(stripe_operations_active(sh));
L
Linus Torvalds 已提交
211 212 213 214 215 216
			if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
				atomic_dec(&conf->preread_active_stripes);
				if (atomic_read(&conf->preread_active_stripes) < IO_THRESHOLD)
					md_wakeup_thread(conf->mddev->thread);
			}
			atomic_dec(&conf->active_stripes);
217 218
			if (!test_bit(STRIPE_EXPANDING, &sh->state)) {
				list_add_tail(&sh->lru, &conf->inactive_list);
L
Linus Torvalds 已提交
219
				wake_up(&conf->wait_for_stripe);
220 221
				if (conf->retry_read_aligned)
					md_wakeup_thread(conf->mddev->thread);
222
			}
L
Linus Torvalds 已提交
223 224 225
		}
	}
}
226

L
Linus Torvalds 已提交
227 228
static void release_stripe(struct stripe_head *sh)
{
229
	struct r5conf *conf = sh->raid_conf;
L
Linus Torvalds 已提交
230
	unsigned long flags;
231

L
Linus Torvalds 已提交
232 233 234 235 236
	spin_lock_irqsave(&conf->device_lock, flags);
	__release_stripe(conf, sh);
	spin_unlock_irqrestore(&conf->device_lock, flags);
}

237
static inline void remove_hash(struct stripe_head *sh)
L
Linus Torvalds 已提交
238
{
239 240
	pr_debug("remove_hash(), stripe %llu\n",
		(unsigned long long)sh->sector);
L
Linus Torvalds 已提交
241

242
	hlist_del_init(&sh->hash);
L
Linus Torvalds 已提交
243 244
}

245
static inline void insert_hash(struct r5conf *conf, struct stripe_head *sh)
L
Linus Torvalds 已提交
246
{
247
	struct hlist_head *hp = stripe_hash(conf, sh->sector);
L
Linus Torvalds 已提交
248

249 250
	pr_debug("insert_hash(), stripe %llu\n",
		(unsigned long long)sh->sector);
L
Linus Torvalds 已提交
251

252
	hlist_add_head(&sh->hash, hp);
L
Linus Torvalds 已提交
253 254 255 256
}


/* find an idle stripe, make sure it is unhashed, and return it. */
257
static struct stripe_head *get_free_stripe(struct r5conf *conf)
L
Linus Torvalds 已提交
258 259 260 261 262 263 264 265 266 267 268 269 270 271 272
{
	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;
}

273
static void shrink_buffers(struct stripe_head *sh)
L
Linus Torvalds 已提交
274 275 276
{
	struct page *p;
	int i;
277
	int num = sh->raid_conf->pool_size;
L
Linus Torvalds 已提交
278

279
	for (i = 0; i < num ; i++) {
L
Linus Torvalds 已提交
280 281 282 283
		p = sh->dev[i].page;
		if (!p)
			continue;
		sh->dev[i].page = NULL;
284
		put_page(p);
L
Linus Torvalds 已提交
285 286 287
	}
}

288
static int grow_buffers(struct stripe_head *sh)
L
Linus Torvalds 已提交
289 290
{
	int i;
291
	int num = sh->raid_conf->pool_size;
L
Linus Torvalds 已提交
292

293
	for (i = 0; i < num; i++) {
L
Linus Torvalds 已提交
294 295 296 297 298 299 300 301 302 303
		struct page *page;

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

304
static void raid5_build_block(struct stripe_head *sh, int i, int previous);
305
static void stripe_set_idx(sector_t stripe, struct r5conf *conf, int previous,
306
			    struct stripe_head *sh);
L
Linus Torvalds 已提交
307

308
static void init_stripe(struct stripe_head *sh, sector_t sector, int previous)
L
Linus Torvalds 已提交
309
{
310
	struct r5conf *conf = sh->raid_conf;
311
	int i;
L
Linus Torvalds 已提交
312

313 314
	BUG_ON(atomic_read(&sh->count) != 0);
	BUG_ON(test_bit(STRIPE_HANDLE, &sh->state));
315
	BUG_ON(stripe_operations_active(sh));
316

317
	pr_debug("init_stripe called, stripe %llu\n",
L
Linus Torvalds 已提交
318 319 320
		(unsigned long long)sh->sector);

	remove_hash(sh);
321

322
	sh->generation = conf->generation - previous;
323
	sh->disks = previous ? conf->previous_raid_disks : conf->raid_disks;
L
Linus Torvalds 已提交
324
	sh->sector = sector;
325
	stripe_set_idx(sector, conf, previous, sh);
L
Linus Torvalds 已提交
326 327
	sh->state = 0;

328 329

	for (i = sh->disks; i--; ) {
L
Linus Torvalds 已提交
330 331
		struct r5dev *dev = &sh->dev[i];

332
		if (dev->toread || dev->read || dev->towrite || dev->written ||
L
Linus Torvalds 已提交
333
		    test_bit(R5_LOCKED, &dev->flags)) {
334
			printk(KERN_ERR "sector=%llx i=%d %p %p %p %p %d\n",
L
Linus Torvalds 已提交
335
			       (unsigned long long)sh->sector, i, dev->toread,
336
			       dev->read, dev->towrite, dev->written,
L
Linus Torvalds 已提交
337
			       test_bit(R5_LOCKED, &dev->flags));
338
			WARN_ON(1);
L
Linus Torvalds 已提交
339 340
		}
		dev->flags = 0;
341
		raid5_build_block(sh, i, previous);
L
Linus Torvalds 已提交
342 343 344 345
	}
	insert_hash(conf, sh);
}

346
static struct stripe_head *__find_stripe(struct r5conf *conf, sector_t sector,
347
					 short generation)
L
Linus Torvalds 已提交
348 349
{
	struct stripe_head *sh;
350
	struct hlist_node *hn;
L
Linus Torvalds 已提交
351

352
	pr_debug("__find_stripe, sector %llu\n", (unsigned long long)sector);
353
	hlist_for_each_entry(sh, hn, stripe_hash(conf, sector), hash)
354
		if (sh->sector == sector && sh->generation == generation)
L
Linus Torvalds 已提交
355
			return sh;
356
	pr_debug("__stripe %llu not in cache\n", (unsigned long long)sector);
L
Linus Torvalds 已提交
357 358 359
	return NULL;
}

360 361 362 363 364 365 366 367 368 369 370 371 372
/*
 * 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.
 */
373
static int calc_degraded(struct r5conf *conf)
374
{
375
	int degraded, degraded2;
376 377 378 379 380
	int i;

	rcu_read_lock();
	degraded = 0;
	for (i = 0; i < conf->previous_raid_disks; i++) {
381
		struct md_rdev *rdev = rcu_dereference(conf->disks[i].rdev);
382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399
		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();
400 401
	if (conf->raid_disks == conf->previous_raid_disks)
		return degraded;
402
	rcu_read_lock();
403
	degraded2 = 0;
404
	for (i = 0; i < conf->raid_disks; i++) {
405
		struct md_rdev *rdev = rcu_dereference(conf->disks[i].rdev);
406
		if (!rdev || test_bit(Faulty, &rdev->flags))
407
			degraded2++;
408 409 410 411 412 413 414 415 416
		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)
417
				degraded2++;
418 419
	}
	rcu_read_unlock();
420 421 422 423 424 425 426 427 428 429 430 431 432
	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);
433 434 435 436 437
	if (degraded > conf->max_degraded)
		return 1;
	return 0;
}

438
static struct stripe_head *
439
get_active_stripe(struct r5conf *conf, sector_t sector,
440
		  int previous, int noblock, int noquiesce)
L
Linus Torvalds 已提交
441 442 443
{
	struct stripe_head *sh;

444
	pr_debug("get_stripe, sector %llu\n", (unsigned long long)sector);
L
Linus Torvalds 已提交
445 446 447 448

	spin_lock_irq(&conf->device_lock);

	do {
449
		wait_event_lock_irq(conf->wait_for_stripe,
450
				    conf->quiesce == 0 || noquiesce,
451
				    conf->device_lock, /* nothing */);
452
		sh = __find_stripe(conf, sector, conf->generation - previous);
L
Linus Torvalds 已提交
453 454 455 456 457 458 459 460 461
		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) &&
462 463
						    (atomic_read(&conf->active_stripes)
						     < (conf->max_nr_stripes *3/4)
L
Linus Torvalds 已提交
464 465
						     || !conf->inactive_blocked),
						    conf->device_lock,
466
						    );
L
Linus Torvalds 已提交
467 468
				conf->inactive_blocked = 0;
			} else
469
				init_stripe(sh, sector, previous);
L
Linus Torvalds 已提交
470 471
		} else {
			if (atomic_read(&sh->count)) {
472 473
				BUG_ON(!list_empty(&sh->lru)
				    && !test_bit(STRIPE_EXPANDING, &sh->state));
L
Linus Torvalds 已提交
474 475 476
			} else {
				if (!test_bit(STRIPE_HANDLE, &sh->state))
					atomic_inc(&conf->active_stripes);
477 478
				if (list_empty(&sh->lru) &&
				    !test_bit(STRIPE_EXPANDING, &sh->state))
479 480
					BUG();
				list_del_init(&sh->lru);
L
Linus Torvalds 已提交
481 482 483 484 485 486 487 488 489 490 491
			}
		}
	} while (sh == NULL);

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

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

492 493 494 495
static void
raid5_end_read_request(struct bio *bi, int error);
static void
raid5_end_write_request(struct bio *bi, int error);
496

497
static void ops_run_io(struct stripe_head *sh, struct stripe_head_state *s)
498
{
499
	struct r5conf *conf = sh->raid_conf;
500 501 502 503 504 505
	int i, disks = sh->disks;

	might_sleep();

	for (i = disks; i--; ) {
		int rw;
506
		int replace_only = 0;
507 508
		struct bio *bi, *rbi;
		struct md_rdev *rdev, *rrdev = NULL;
T
Tejun Heo 已提交
509 510 511 512 513 514
		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;
		} else if (test_and_clear_bit(R5_Wantread, &sh->dev[i].flags))
515
			rw = READ;
516 517 518 519 520
		else if (test_and_clear_bit(R5_WantReplace,
					    &sh->dev[i].flags)) {
			rw = WRITE;
			replace_only = 1;
		} else
521 522 523
			continue;

		bi = &sh->dev[i].req;
524
		rbi = &sh->dev[i].rreq; /* For writing to replacement */
525 526

		bi->bi_rw = rw;
527 528
		rbi->bi_rw = rw;
		if (rw & WRITE) {
529
			bi->bi_end_io = raid5_end_write_request;
530 531
			rbi->bi_end_io = raid5_end_write_request;
		} else
532 533 534
			bi->bi_end_io = raid5_end_read_request;

		rcu_read_lock();
535
		rrdev = rcu_dereference(conf->disks[i].replacement);
536 537 538 539 540 541
		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;
		}
542 543 544
		if (rw & WRITE) {
			if (replace_only)
				rdev = NULL;
545 546 547
			if (rdev == rrdev)
				/* We raced and saw duplicates */
				rrdev = NULL;
548
		} else {
549
			if (test_bit(R5_ReadRepl, &sh->dev[i].flags) && rrdev)
550 551 552
				rdev = rrdev;
			rrdev = NULL;
		}
553

554 555 556 557
		if (rdev && test_bit(Faulty, &rdev->flags))
			rdev = NULL;
		if (rdev)
			atomic_inc(&rdev->nr_pending);
558 559 560 561
		if (rrdev && test_bit(Faulty, &rrdev->flags))
			rrdev = NULL;
		if (rrdev)
			atomic_inc(&rrdev->nr_pending);
562 563
		rcu_read_unlock();

564
		/* We have already checked bad blocks for reads.  Now
565 566
		 * need to check for writes.  We never accept write errors
		 * on the replacement, so we don't to check rrdev.
567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594
		 */
		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);
				}
				md_wait_for_blocked_rdev(rdev, conf->mddev);
			} else {
				/* Acknowledged bad block - skip the write */
				rdev_dec_pending(rdev, conf->mddev);
				rdev = NULL;
			}
		}

595
		if (rdev) {
596 597
			if (s->syncing || s->expanding || s->expanded
			    || s->replacing)
598 599
				md_sync_acct(rdev->bdev, STRIPE_SECTORS);

D
Dan Williams 已提交
600 601
			set_bit(STRIPE_IO_STARTED, &sh->state);

602 603
			bi->bi_bdev = rdev->bdev;
			pr_debug("%s: for %llu schedule op %ld on disc %d\n",
604
				__func__, (unsigned long long)sh->sector,
605 606 607 608 609 610 611 612 613
				bi->bi_rw, i);
			atomic_inc(&sh->count);
			bi->bi_sector = sh->sector + rdev->data_offset;
			bi->bi_flags = 1 << BIO_UPTODATE;
			bi->bi_idx = 0;
			bi->bi_io_vec[0].bv_len = STRIPE_SIZE;
			bi->bi_io_vec[0].bv_offset = 0;
			bi->bi_size = STRIPE_SIZE;
			bi->bi_next = NULL;
614 615
			if (rrdev)
				set_bit(R5_DOUBLE_LOCKED, &sh->dev[i].flags);
616
			generic_make_request(bi);
617 618
		}
		if (rrdev) {
619 620
			if (s->syncing || s->expanding || s->expanded
			    || s->replacing)
621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640
				md_sync_acct(rrdev->bdev, STRIPE_SECTORS);

			set_bit(STRIPE_IO_STARTED, &sh->state);

			rbi->bi_bdev = rrdev->bdev;
			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);
			rbi->bi_sector = sh->sector + rrdev->data_offset;
			rbi->bi_flags = 1 << BIO_UPTODATE;
			rbi->bi_idx = 0;
			rbi->bi_io_vec[0].bv_len = STRIPE_SIZE;
			rbi->bi_io_vec[0].bv_offset = 0;
			rbi->bi_size = STRIPE_SIZE;
			rbi->bi_next = NULL;
			generic_make_request(rbi);
		}
		if (!rdev && !rrdev) {
641
			if (rw & WRITE)
642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658
				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;
659
	struct async_submit_ctl submit;
D
Dan Williams 已提交
660
	enum async_tx_flags flags = 0;
661 662 663 664 665

	if (bio->bi_sector >= sector)
		page_offset = (signed)(bio->bi_sector - sector) * 512;
	else
		page_offset = (signed)(sector - bio->bi_sector) * -512;
666

D
Dan Williams 已提交
667 668 669 670
	if (frombio)
		flags |= ASYNC_TX_FENCE;
	init_async_submit(&submit, flags, tx, NULL, NULL, NULL);

671
	bio_for_each_segment(bvl, bio, i) {
672
		int len = bvl->bv_len;
673 674 675 676 677 678 679 680 681 682 683 684 685 686 687
		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) {
688 689
			b_offset += bvl->bv_offset;
			bio_page = bvl->bv_page;
690 691
			if (frombio)
				tx = async_memcpy(page, bio_page, page_offset,
692
						  b_offset, clen, &submit);
693 694
			else
				tx = async_memcpy(bio_page, page, b_offset,
695
						  page_offset, clen, &submit);
696
		}
697 698 699
		/* chain the operations */
		submit.depend_tx = tx;

700 701 702 703 704 705 706 707 708 709 710 711
		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;
712
	struct r5conf *conf = sh->raid_conf;
713
	int i;
714

715
	pr_debug("%s: stripe %llu\n", __func__,
716 717 718
		(unsigned long long)sh->sector);

	/* clear completed biofills */
719
	spin_lock_irq(&conf->device_lock);
720 721 722 723
	for (i = sh->disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];

		/* acknowledge completion of a biofill operation */
724 725
		/* and check if we need to reply to a read request,
		 * new R5_Wantfill requests are held off until
726
		 * !STRIPE_BIOFILL_RUN
727 728
		 */
		if (test_and_clear_bit(R5_Wantfill, &dev->flags)) {
729 730 731 732 733 734 735 736
			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);
737
				if (!raid5_dec_bi_phys_segments(rbi)) {
738 739 740 741 742 743 744
					rbi->bi_next = return_bi;
					return_bi = rbi;
				}
				rbi = rbi2;
			}
		}
	}
745 746
	spin_unlock_irq(&conf->device_lock);
	clear_bit(STRIPE_BIOFILL_RUN, &sh->state);
747 748 749

	return_io(return_bi);

750
	set_bit(STRIPE_HANDLE, &sh->state);
751 752 753 754 755 756
	release_stripe(sh);
}

static void ops_run_biofill(struct stripe_head *sh)
{
	struct dma_async_tx_descriptor *tx = NULL;
757
	struct r5conf *conf = sh->raid_conf;
758
	struct async_submit_ctl submit;
759 760
	int i;

761
	pr_debug("%s: stripe %llu\n", __func__,
762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781
		(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;
			spin_lock_irq(&conf->device_lock);
			dev->read = rbi = dev->toread;
			dev->toread = NULL;
			spin_unlock_irq(&conf->device_lock);
			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);
782 783
	init_async_submit(&submit, ASYNC_TX_ACK, tx, ops_complete_biofill, sh, NULL);
	async_trigger_callback(&submit);
784 785
}

786
static void mark_target_uptodate(struct stripe_head *sh, int target)
787
{
788
	struct r5dev *tgt;
789

790 791
	if (target < 0)
		return;
792

793
	tgt = &sh->dev[target];
794 795 796
	set_bit(R5_UPTODATE, &tgt->flags);
	BUG_ON(!test_bit(R5_Wantcompute, &tgt->flags));
	clear_bit(R5_Wantcompute, &tgt->flags);
797 798
}

799
static void ops_complete_compute(void *stripe_head_ref)
800 801 802
{
	struct stripe_head *sh = stripe_head_ref;

803
	pr_debug("%s: stripe %llu\n", __func__,
804 805
		(unsigned long long)sh->sector);

806
	/* mark the computed target(s) as uptodate */
807
	mark_target_uptodate(sh, sh->ops.target);
808
	mark_target_uptodate(sh, sh->ops.target2);
809

810 811 812
	clear_bit(STRIPE_COMPUTE_RUN, &sh->state);
	if (sh->check_state == check_state_compute_run)
		sh->check_state = check_state_compute_result;
813 814 815 816
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

817 818 819 820 821 822 823 824 825
/* 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)
826 827
{
	int disks = sh->disks;
828
	struct page **xor_srcs = percpu->scribble;
829 830 831 832 833
	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;
834
	struct async_submit_ctl submit;
835 836 837
	int i;

	pr_debug("%s: stripe %llu block: %d\n",
838
		__func__, (unsigned long long)sh->sector, target);
839 840 841 842 843 844 845 846
	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
847
	init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST, NULL,
848
			  ops_complete_compute, sh, to_addr_conv(sh, percpu));
849
	if (unlikely(count == 1))
850
		tx = async_memcpy(xor_dest, xor_srcs[0], 0, 0, STRIPE_SIZE, &submit);
851
	else
852
		tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, &submit);
853 854 855 856

	return tx;
}

857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874
/* 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++)
875
		srcs[i] = NULL;
876 877 878 879 880 881 882 883 884 885

	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);

886
	return syndrome_disks;
887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906
}

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;
907
	else
908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923
		/* 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
924 925
		init_async_submit(&submit, ASYNC_TX_FENCE, NULL,
				  ops_complete_compute, sh,
926 927 928 929 930 931 932 933 934 935 936
				  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
937 938
		init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST,
				  NULL, ops_complete_compute, sh,
939 940 941
				  to_addr_conv(sh, percpu));
		tx = async_xor(dest, blocks, 0, count, STRIPE_SIZE, &submit);
	}
942 943 944 945

	return tx;
}

946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966
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));

967
	/* we need to open-code set_syndrome_sources to handle the
968 969 970
	 * slot number conversion for 'faila' and 'failb'
	 */
	for (i = 0; i < disks ; i++)
971
		blocks[i] = NULL;
972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997
	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
998 999 1000
			init_async_submit(&submit, ASYNC_TX_FENCE, NULL,
					  ops_complete_compute, sh,
					  to_addr_conv(sh, percpu));
1001
			return async_gen_syndrome(blocks, 0, syndrome_disks+2,
1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020
						  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
1021 1022 1023 1024
			init_async_submit(&submit,
					  ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST,
					  NULL, NULL, NULL,
					  to_addr_conv(sh, percpu));
1025 1026 1027 1028
			tx = async_xor(dest, blocks, 0, count, STRIPE_SIZE,
				       &submit);

			count = set_syndrome_sources(blocks, sh);
D
Dan Williams 已提交
1029 1030 1031
			init_async_submit(&submit, ASYNC_TX_FENCE, tx,
					  ops_complete_compute, sh,
					  to_addr_conv(sh, percpu));
1032 1033 1034 1035
			return async_gen_syndrome(blocks, 0, count+2,
						  STRIPE_SIZE, &submit);
		}
	} else {
1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049
		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);
		}
1050 1051 1052 1053
	}
}


1054 1055 1056 1057
static void ops_complete_prexor(void *stripe_head_ref)
{
	struct stripe_head *sh = stripe_head_ref;

1058
	pr_debug("%s: stripe %llu\n", __func__,
1059 1060 1061 1062
		(unsigned long long)sh->sector);
}

static struct dma_async_tx_descriptor *
1063 1064
ops_run_prexor(struct stripe_head *sh, struct raid5_percpu *percpu,
	       struct dma_async_tx_descriptor *tx)
1065 1066
{
	int disks = sh->disks;
1067
	struct page **xor_srcs = percpu->scribble;
1068
	int count = 0, pd_idx = sh->pd_idx, i;
1069
	struct async_submit_ctl submit;
1070 1071 1072 1073

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

1074
	pr_debug("%s: stripe %llu\n", __func__,
1075 1076 1077 1078 1079
		(unsigned long long)sh->sector);

	for (i = disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
		/* Only process blocks that are known to be uptodate */
1080
		if (test_bit(R5_Wantdrain, &dev->flags))
1081 1082 1083
			xor_srcs[count++] = dev->page;
	}

D
Dan Williams 已提交
1084
	init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_DROP_DST, tx,
1085
			  ops_complete_prexor, sh, to_addr_conv(sh, percpu));
1086
	tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, &submit);
1087 1088 1089 1090 1091

	return tx;
}

static struct dma_async_tx_descriptor *
1092
ops_run_biodrain(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
1093 1094
{
	int disks = sh->disks;
1095
	int i;
1096

1097
	pr_debug("%s: stripe %llu\n", __func__,
1098 1099 1100 1101 1102 1103
		(unsigned long long)sh->sector);

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

1104
		if (test_and_clear_bit(R5_Wantdrain, &dev->flags)) {
1105 1106
			struct bio *wbi;

1107
			spin_lock_irq(&sh->raid_conf->device_lock);
1108 1109 1110 1111
			chosen = dev->towrite;
			dev->towrite = NULL;
			BUG_ON(dev->written);
			wbi = dev->written = chosen;
1112
			spin_unlock_irq(&sh->raid_conf->device_lock);
1113 1114 1115

			while (wbi && wbi->bi_sector <
				dev->sector + STRIPE_SECTORS) {
T
Tejun Heo 已提交
1116 1117
				if (wbi->bi_rw & REQ_FUA)
					set_bit(R5_WantFUA, &dev->flags);
1118 1119 1120 1121 1122 1123 1124 1125 1126 1127
				tx = async_copy_data(1, wbi, dev->page,
					dev->sector, tx);
				wbi = r5_next_bio(wbi, dev->sector);
			}
		}
	}

	return tx;
}

1128
static void ops_complete_reconstruct(void *stripe_head_ref)
1129 1130
{
	struct stripe_head *sh = stripe_head_ref;
1131 1132 1133 1134
	int disks = sh->disks;
	int pd_idx = sh->pd_idx;
	int qd_idx = sh->qd_idx;
	int i;
T
Tejun Heo 已提交
1135
	bool fua = false;
1136

1137
	pr_debug("%s: stripe %llu\n", __func__,
1138 1139
		(unsigned long long)sh->sector);

T
Tejun Heo 已提交
1140 1141 1142
	for (i = disks; i--; )
		fua |= test_bit(R5_WantFUA, &sh->dev[i].flags);

1143 1144
	for (i = disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
1145

T
Tejun Heo 已提交
1146
		if (dev->written || i == pd_idx || i == qd_idx) {
1147
			set_bit(R5_UPTODATE, &dev->flags);
T
Tejun Heo 已提交
1148 1149 1150
			if (fua)
				set_bit(R5_WantFUA, &dev->flags);
		}
1151 1152
	}

1153 1154 1155 1156 1157 1158 1159 1160
	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;
	}
1161 1162 1163 1164 1165 1166

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

static void
1167 1168
ops_run_reconstruct5(struct stripe_head *sh, struct raid5_percpu *percpu,
		     struct dma_async_tx_descriptor *tx)
1169 1170
{
	int disks = sh->disks;
1171
	struct page **xor_srcs = percpu->scribble;
1172
	struct async_submit_ctl submit;
1173 1174
	int count = 0, pd_idx = sh->pd_idx, i;
	struct page *xor_dest;
1175
	int prexor = 0;
1176 1177
	unsigned long flags;

1178
	pr_debug("%s: stripe %llu\n", __func__,
1179 1180 1181 1182 1183
		(unsigned long long)sh->sector);

	/* check if prexor is active which means only process blocks
	 * that are part of a read-modify-write (written)
	 */
1184 1185
	if (sh->reconstruct_state == reconstruct_state_prexor_drain_run) {
		prexor = 1;
1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205
		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
	 */
1206
	flags = ASYNC_TX_ACK |
1207 1208 1209 1210
		(prexor ? ASYNC_TX_XOR_DROP_DST : ASYNC_TX_XOR_ZERO_DST);

	atomic_inc(&sh->count);

1211
	init_async_submit(&submit, flags, tx, ops_complete_reconstruct, sh,
1212
			  to_addr_conv(sh, percpu));
1213 1214 1215 1216
	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);
1217 1218
}

1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235
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;
	int count;

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

	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);
1236 1237 1238 1239 1240 1241
}

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

1242
	pr_debug("%s: stripe %llu\n", __func__,
1243 1244
		(unsigned long long)sh->sector);

1245
	sh->check_state = check_state_check_result;
1246 1247 1248 1249
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

1250
static void ops_run_check_p(struct stripe_head *sh, struct raid5_percpu *percpu)
1251 1252
{
	int disks = sh->disks;
1253 1254 1255
	int pd_idx = sh->pd_idx;
	int qd_idx = sh->qd_idx;
	struct page *xor_dest;
1256
	struct page **xor_srcs = percpu->scribble;
1257
	struct dma_async_tx_descriptor *tx;
1258
	struct async_submit_ctl submit;
1259 1260
	int count;
	int i;
1261

1262
	pr_debug("%s: stripe %llu\n", __func__,
1263 1264
		(unsigned long long)sh->sector);

1265 1266 1267
	count = 0;
	xor_dest = sh->dev[pd_idx].page;
	xor_srcs[count++] = xor_dest;
1268
	for (i = disks; i--; ) {
1269 1270 1271
		if (i == pd_idx || i == qd_idx)
			continue;
		xor_srcs[count++] = sh->dev[i].page;
1272 1273
	}

1274 1275
	init_async_submit(&submit, 0, NULL, NULL, NULL,
			  to_addr_conv(sh, percpu));
D
Dan Williams 已提交
1276
	tx = async_xor_val(xor_dest, xor_srcs, 0, count, STRIPE_SIZE,
1277
			   &sh->ops.zero_sum_result, &submit);
1278 1279

	atomic_inc(&sh->count);
1280 1281
	init_async_submit(&submit, ASYNC_TX_ACK, tx, ops_complete_check, sh, NULL);
	tx = async_trigger_callback(&submit);
1282 1283
}

1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295
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;
1296 1297

	atomic_inc(&sh->count);
1298 1299 1300 1301
	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);
1302 1303
}

1304
static void __raid_run_ops(struct stripe_head *sh, unsigned long ops_request)
1305 1306 1307
{
	int overlap_clear = 0, i, disks = sh->disks;
	struct dma_async_tx_descriptor *tx = NULL;
1308
	struct r5conf *conf = sh->raid_conf;
1309
	int level = conf->level;
1310 1311
	struct raid5_percpu *percpu;
	unsigned long cpu;
1312

1313 1314
	cpu = get_cpu();
	percpu = per_cpu_ptr(conf->percpu, cpu);
1315
	if (test_bit(STRIPE_OP_BIOFILL, &ops_request)) {
1316 1317 1318 1319
		ops_run_biofill(sh);
		overlap_clear++;
	}

1320
	if (test_bit(STRIPE_OP_COMPUTE_BLK, &ops_request)) {
1321 1322 1323 1324 1325 1326 1327 1328 1329 1330
		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))
1331 1332
			async_tx_ack(tx);
	}
1333

1334
	if (test_bit(STRIPE_OP_PREXOR, &ops_request))
1335
		tx = ops_run_prexor(sh, percpu, tx);
1336

1337
	if (test_bit(STRIPE_OP_BIODRAIN, &ops_request)) {
1338
		tx = ops_run_biodrain(sh, tx);
1339 1340 1341
		overlap_clear++;
	}

1342 1343 1344 1345 1346 1347
	if (test_bit(STRIPE_OP_RECONSTRUCT, &ops_request)) {
		if (level < 6)
			ops_run_reconstruct5(sh, percpu, tx);
		else
			ops_run_reconstruct6(sh, percpu, tx);
	}
1348

1349 1350 1351 1352 1353 1354 1355 1356 1357 1358
	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();
	}
1359 1360 1361 1362 1363 1364 1365

	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);
		}
1366
	put_cpu();
1367 1368
}

1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398
#ifdef CONFIG_MULTICORE_RAID456
static void async_run_ops(void *param, async_cookie_t cookie)
{
	struct stripe_head *sh = param;
	unsigned long ops_request = sh->ops.request;

	clear_bit_unlock(STRIPE_OPS_REQ_PENDING, &sh->state);
	wake_up(&sh->ops.wait_for_ops);

	__raid_run_ops(sh, ops_request);
	release_stripe(sh);
}

static void raid_run_ops(struct stripe_head *sh, unsigned long ops_request)
{
	/* since handle_stripe can be called outside of raid5d context
	 * we need to ensure sh->ops.request is de-staged before another
	 * request arrives
	 */
	wait_event(sh->ops.wait_for_ops,
		   !test_and_set_bit_lock(STRIPE_OPS_REQ_PENDING, &sh->state));
	sh->ops.request = ops_request;

	atomic_inc(&sh->count);
	async_schedule(async_run_ops, sh);
}
#else
#define raid_run_ops __raid_run_ops
#endif

1399
static int grow_one_stripe(struct r5conf *conf)
L
Linus Torvalds 已提交
1400 1401
{
	struct stripe_head *sh;
N
Namhyung Kim 已提交
1402
	sh = kmem_cache_zalloc(conf->slab_cache, GFP_KERNEL);
1403 1404
	if (!sh)
		return 0;
N
Namhyung Kim 已提交
1405

1406
	sh->raid_conf = conf;
1407 1408 1409
	#ifdef CONFIG_MULTICORE_RAID456
	init_waitqueue_head(&sh->ops.wait_for_ops);
	#endif
1410

1411 1412
	if (grow_buffers(sh)) {
		shrink_buffers(sh);
1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423
		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;
}

1424
static int grow_stripes(struct r5conf *conf, int num)
1425
{
1426
	struct kmem_cache *sc;
1427
	int devs = max(conf->raid_disks, conf->previous_raid_disks);
L
Linus Torvalds 已提交
1428

1429 1430 1431 1432 1433 1434 1435 1436
	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]);

1437 1438
	conf->active_name = 0;
	sc = kmem_cache_create(conf->cache_name[conf->active_name],
L
Linus Torvalds 已提交
1439
			       sizeof(struct stripe_head)+(devs-1)*sizeof(struct r5dev),
1440
			       0, 0, NULL);
L
Linus Torvalds 已提交
1441 1442 1443
	if (!sc)
		return 1;
	conf->slab_cache = sc;
1444
	conf->pool_size = devs;
1445
	while (num--)
1446
		if (!grow_one_stripe(conf))
L
Linus Torvalds 已提交
1447 1448 1449
			return 1;
	return 0;
}
1450

1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472
/**
 * 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;
}

1473
static int resize_stripes(struct r5conf *conf, int newsize)
1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500
{
	/* 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.
	 * 2/ gather all the old stripe_heads and tranfer the pages across
	 *    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;
1501
	unsigned long cpu;
1502
	int err;
1503
	struct kmem_cache *sc;
1504 1505 1506 1507 1508
	int i;

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

1509 1510 1511
	err = md_allow_write(conf->mddev);
	if (err)
		return err;
1512

1513 1514 1515
	/* Step 1 */
	sc = kmem_cache_create(conf->cache_name[1-conf->active_name],
			       sizeof(struct stripe_head)+(newsize-1)*sizeof(struct r5dev),
1516
			       0, 0, NULL);
1517 1518 1519 1520
	if (!sc)
		return -ENOMEM;

	for (i = conf->max_nr_stripes; i; i--) {
N
Namhyung Kim 已提交
1521
		nsh = kmem_cache_zalloc(sc, GFP_KERNEL);
1522 1523 1524 1525
		if (!nsh)
			break;

		nsh->raid_conf = conf;
1526 1527 1528
		#ifdef CONFIG_MULTICORE_RAID456
		init_waitqueue_head(&nsh->ops.wait_for_ops);
		#endif
1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550

		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),
				    conf->device_lock,
N
NeilBrown 已提交
1551
				    );
1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565
		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
1566
	 * conf->disks and the scribble region
1567 1568 1569 1570 1571 1572 1573 1574 1575 1576
	 */
	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;

1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595
	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();

1596 1597 1598 1599
	/* 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);
1600

1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616
		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
1617

1618
static int drop_one_stripe(struct r5conf *conf)
L
Linus Torvalds 已提交
1619 1620 1621
{
	struct stripe_head *sh;

1622 1623 1624 1625 1626
	spin_lock_irq(&conf->device_lock);
	sh = get_free_stripe(conf);
	spin_unlock_irq(&conf->device_lock);
	if (!sh)
		return 0;
1627
	BUG_ON(atomic_read(&sh->count));
1628
	shrink_buffers(sh);
1629 1630 1631 1632 1633
	kmem_cache_free(conf->slab_cache, sh);
	atomic_dec(&conf->active_stripes);
	return 1;
}

1634
static void shrink_stripes(struct r5conf *conf)
1635 1636 1637 1638
{
	while (drop_one_stripe(conf))
		;

N
NeilBrown 已提交
1639 1640
	if (conf->slab_cache)
		kmem_cache_destroy(conf->slab_cache);
L
Linus Torvalds 已提交
1641 1642 1643
	conf->slab_cache = NULL;
}

1644
static void raid5_end_read_request(struct bio * bi, int error)
L
Linus Torvalds 已提交
1645
{
1646
	struct stripe_head *sh = bi->bi_private;
1647
	struct r5conf *conf = sh->raid_conf;
1648
	int disks = sh->disks, i;
L
Linus Torvalds 已提交
1649
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
1650
	char b[BDEVNAME_SIZE];
1651
	struct md_rdev *rdev = NULL;
L
Linus Torvalds 已提交
1652 1653 1654 1655 1656 1657


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

1658 1659
	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
1660 1661 1662
		uptodate);
	if (i == disks) {
		BUG();
1663
		return;
L
Linus Torvalds 已提交
1664
	}
1665
	if (test_bit(R5_ReadRepl, &sh->dev[i].flags))
1666 1667 1668 1669 1670
		/* 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.
		 */
1671
		rdev = conf->disks[i].replacement;
1672
	if (!rdev)
1673
		rdev = conf->disks[i].rdev;
L
Linus Torvalds 已提交
1674 1675 1676

	if (uptodate) {
		set_bit(R5_UPTODATE, &sh->dev[i].flags);
1677
		if (test_bit(R5_ReadError, &sh->dev[i].flags)) {
1678 1679 1680 1681
			/* Note that this cannot happen on a
			 * replacement device.  We just fail those on
			 * any error
			 */
1682 1683 1684 1685 1686 1687 1688 1689
			printk_ratelimited(
				KERN_INFO
				"md/raid:%s: read error corrected"
				" (%lu sectors at %llu on %s)\n",
				mdname(conf->mddev), STRIPE_SECTORS,
				(unsigned long long)(sh->sector
						     + rdev->data_offset),
				bdevname(rdev->bdev, b));
1690
			atomic_add(STRIPE_SECTORS, &rdev->corrected_errors);
1691 1692 1693
			clear_bit(R5_ReadError, &sh->dev[i].flags);
			clear_bit(R5_ReWrite, &sh->dev[i].flags);
		}
1694 1695
		if (atomic_read(&rdev->read_errors))
			atomic_set(&rdev->read_errors, 0);
L
Linus Torvalds 已提交
1696
	} else {
1697
		const char *bdn = bdevname(rdev->bdev, b);
1698
		int retry = 0;
1699

L
Linus Torvalds 已提交
1700
		clear_bit(R5_UPTODATE, &sh->dev[i].flags);
1701
		atomic_inc(&rdev->read_errors);
1702 1703 1704 1705 1706 1707 1708 1709 1710 1711
		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),
				(unsigned long long)(sh->sector
						     + rdev->data_offset),
				bdn);
		else if (conf->mddev->degraded >= conf->max_degraded)
1712 1713 1714 1715 1716 1717 1718 1719
			printk_ratelimited(
				KERN_WARNING
				"md/raid:%s: read error not correctable "
				"(sector %llu on %s).\n",
				mdname(conf->mddev),
				(unsigned long long)(sh->sector
						     + rdev->data_offset),
				bdn);
1720
		else if (test_bit(R5_ReWrite, &sh->dev[i].flags))
1721
			/* Oh, no!!! */
1722 1723 1724 1725 1726 1727 1728 1729
			printk_ratelimited(
				KERN_WARNING
				"md/raid:%s: read error NOT corrected!! "
				"(sector %llu on %s).\n",
				mdname(conf->mddev),
				(unsigned long long)(sh->sector
						     + rdev->data_offset),
				bdn);
1730
		else if (atomic_read(&rdev->read_errors)
1731
			 > conf->max_nr_stripes)
N
NeilBrown 已提交
1732
			printk(KERN_WARNING
1733
			       "md/raid:%s: Too many read errors, failing device %s.\n",
1734
			       mdname(conf->mddev), bdn);
1735 1736 1737 1738 1739
		else
			retry = 1;
		if (retry)
			set_bit(R5_ReadError, &sh->dev[i].flags);
		else {
1740 1741
			clear_bit(R5_ReadError, &sh->dev[i].flags);
			clear_bit(R5_ReWrite, &sh->dev[i].flags);
1742
			md_error(conf->mddev, rdev);
1743
		}
L
Linus Torvalds 已提交
1744
	}
1745
	rdev_dec_pending(rdev, conf->mddev);
L
Linus Torvalds 已提交
1746 1747 1748 1749 1750
	clear_bit(R5_LOCKED, &sh->dev[i].flags);
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

1751
static void raid5_end_write_request(struct bio *bi, int error)
L
Linus Torvalds 已提交
1752
{
1753
	struct stripe_head *sh = bi->bi_private;
1754
	struct r5conf *conf = sh->raid_conf;
1755
	int disks = sh->disks, i;
1756
	struct md_rdev *uninitialized_var(rdev);
L
Linus Torvalds 已提交
1757
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
1758 1759
	sector_t first_bad;
	int bad_sectors;
1760
	int replacement = 0;
L
Linus Torvalds 已提交
1761

1762 1763 1764
	for (i = 0 ; i < disks; i++) {
		if (bi == &sh->dev[i].req) {
			rdev = conf->disks[i].rdev;
L
Linus Torvalds 已提交
1765
			break;
1766 1767 1768
		}
		if (bi == &sh->dev[i].rreq) {
			rdev = conf->disks[i].replacement;
1769 1770 1771 1772 1773 1774 1775 1776
			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;
1777 1778 1779
			break;
		}
	}
1780
	pr_debug("end_write_request %llu/%d, count %d, uptodate: %d.\n",
L
Linus Torvalds 已提交
1781 1782 1783 1784
		(unsigned long long)sh->sector, i, atomic_read(&sh->count),
		uptodate);
	if (i == disks) {
		BUG();
1785
		return;
L
Linus Torvalds 已提交
1786 1787
	}

1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804
	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);
		} else if (is_badblock(rdev, sh->sector,
				       STRIPE_SECTORS,
				       &first_bad, &bad_sectors))
			set_bit(R5_MadeGood, &sh->dev[i].flags);
	}
	rdev_dec_pending(rdev, conf->mddev);
L
Linus Torvalds 已提交
1805

1806 1807
	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
1808
	set_bit(STRIPE_HANDLE, &sh->state);
1809
	release_stripe(sh);
L
Linus Torvalds 已提交
1810 1811
}

1812
static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous);
L
Linus Torvalds 已提交
1813
	
1814
static void raid5_build_block(struct stripe_head *sh, int i, int previous)
L
Linus Torvalds 已提交
1815 1816 1817 1818 1819 1820 1821 1822
{
	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;
1823
	dev->vec.bv_page = dev->page;
L
Linus Torvalds 已提交
1824

1825 1826 1827 1828 1829 1830 1831
	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
1832
	dev->flags = 0;
1833
	dev->sector = compute_blocknr(sh, i, previous);
L
Linus Torvalds 已提交
1834 1835
}

1836
static void error(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
1837 1838
{
	char b[BDEVNAME_SIZE];
1839
	struct r5conf *conf = mddev->private;
1840
	unsigned long flags;
1841
	pr_debug("raid456: error called\n");
L
Linus Torvalds 已提交
1842

1843 1844 1845 1846 1847 1848
	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);

1849
	set_bit(Blocked, &rdev->flags);
1850 1851 1852 1853 1854 1855 1856 1857 1858
	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);
1859
}
L
Linus Torvalds 已提交
1860 1861 1862 1863 1864

/*
 * Input: a 'big' sector number,
 * Output: index of the data and parity disk, and the sector # in them.
 */
1865
static sector_t raid5_compute_sector(struct r5conf *conf, sector_t r_sector,
1866 1867
				     int previous, int *dd_idx,
				     struct stripe_head *sh)
L
Linus Torvalds 已提交
1868
{
N
NeilBrown 已提交
1869
	sector_t stripe, stripe2;
1870
	sector_t chunk_number;
L
Linus Torvalds 已提交
1871
	unsigned int chunk_offset;
1872
	int pd_idx, qd_idx;
1873
	int ddf_layout = 0;
L
Linus Torvalds 已提交
1874
	sector_t new_sector;
1875 1876
	int algorithm = previous ? conf->prev_algo
				 : conf->algorithm;
1877 1878
	int sectors_per_chunk = previous ? conf->prev_chunk_sectors
					 : conf->chunk_sectors;
1879 1880 1881
	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
1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893

	/* 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
	 */
1894 1895
	stripe = chunk_number;
	*dd_idx = sector_div(stripe, data_disks);
N
NeilBrown 已提交
1896
	stripe2 = stripe;
L
Linus Torvalds 已提交
1897 1898 1899
	/*
	 * Select the parity disk based on the user selected algorithm.
	 */
1900
	pd_idx = qd_idx = -1;
1901 1902
	switch(conf->level) {
	case 4:
1903
		pd_idx = data_disks;
1904 1905
		break;
	case 5:
1906
		switch (algorithm) {
L
Linus Torvalds 已提交
1907
		case ALGORITHM_LEFT_ASYMMETRIC:
N
NeilBrown 已提交
1908
			pd_idx = data_disks - sector_div(stripe2, raid_disks);
1909
			if (*dd_idx >= pd_idx)
L
Linus Torvalds 已提交
1910 1911 1912
				(*dd_idx)++;
			break;
		case ALGORITHM_RIGHT_ASYMMETRIC:
N
NeilBrown 已提交
1913
			pd_idx = sector_div(stripe2, raid_disks);
1914
			if (*dd_idx >= pd_idx)
L
Linus Torvalds 已提交
1915 1916 1917
				(*dd_idx)++;
			break;
		case ALGORITHM_LEFT_SYMMETRIC:
N
NeilBrown 已提交
1918
			pd_idx = data_disks - sector_div(stripe2, raid_disks);
1919
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
L
Linus Torvalds 已提交
1920 1921
			break;
		case ALGORITHM_RIGHT_SYMMETRIC:
N
NeilBrown 已提交
1922
			pd_idx = sector_div(stripe2, raid_disks);
1923
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
L
Linus Torvalds 已提交
1924
			break;
1925 1926 1927 1928 1929 1930 1931
		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
1932
		default:
1933
			BUG();
1934 1935 1936 1937
		}
		break;
	case 6:

1938
		switch (algorithm) {
1939
		case ALGORITHM_LEFT_ASYMMETRIC:
N
NeilBrown 已提交
1940
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
1941 1942
			qd_idx = pd_idx + 1;
			if (pd_idx == raid_disks-1) {
1943
				(*dd_idx)++;	/* Q D D D P */
1944 1945
				qd_idx = 0;
			} else if (*dd_idx >= pd_idx)
1946 1947 1948
				(*dd_idx) += 2; /* D D P Q D */
			break;
		case ALGORITHM_RIGHT_ASYMMETRIC:
N
NeilBrown 已提交
1949
			pd_idx = sector_div(stripe2, raid_disks);
1950 1951
			qd_idx = pd_idx + 1;
			if (pd_idx == raid_disks-1) {
1952
				(*dd_idx)++;	/* Q D D D P */
1953 1954
				qd_idx = 0;
			} else if (*dd_idx >= pd_idx)
1955 1956 1957
				(*dd_idx) += 2; /* D D P Q D */
			break;
		case ALGORITHM_LEFT_SYMMETRIC:
N
NeilBrown 已提交
1958
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
1959 1960
			qd_idx = (pd_idx + 1) % raid_disks;
			*dd_idx = (pd_idx + 2 + *dd_idx) % raid_disks;
1961 1962
			break;
		case ALGORITHM_RIGHT_SYMMETRIC:
N
NeilBrown 已提交
1963
			pd_idx = sector_div(stripe2, raid_disks);
1964 1965
			qd_idx = (pd_idx + 1) % raid_disks;
			*dd_idx = (pd_idx + 2 + *dd_idx) % raid_disks;
1966
			break;
1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981

		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
1982
			pd_idx = sector_div(stripe2, raid_disks);
1983 1984 1985 1986 1987 1988
			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 */
1989
			ddf_layout = 1;
1990 1991 1992 1993 1994 1995 1996
			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
1997 1998
			stripe2 += 1;
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
1999 2000 2001 2002 2003 2004
			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 */
2005
			ddf_layout = 1;
2006 2007 2008 2009
			break;

		case ALGORITHM_ROTATING_N_CONTINUE:
			/* Same as left_symmetric but Q is before P */
N
NeilBrown 已提交
2010
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2011 2012
			qd_idx = (pd_idx + raid_disks - 1) % raid_disks;
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
2013
			ddf_layout = 1;
2014 2015 2016 2017
			break;

		case ALGORITHM_LEFT_ASYMMETRIC_6:
			/* RAID5 left_asymmetric, with Q on last device */
N
NeilBrown 已提交
2018
			pd_idx = data_disks - sector_div(stripe2, raid_disks-1);
2019 2020 2021 2022 2023 2024
			if (*dd_idx >= pd_idx)
				(*dd_idx)++;
			qd_idx = raid_disks - 1;
			break;

		case ALGORITHM_RIGHT_ASYMMETRIC_6:
N
NeilBrown 已提交
2025
			pd_idx = sector_div(stripe2, raid_disks-1);
2026 2027 2028 2029 2030 2031
			if (*dd_idx >= pd_idx)
				(*dd_idx)++;
			qd_idx = raid_disks - 1;
			break;

		case ALGORITHM_LEFT_SYMMETRIC_6:
N
NeilBrown 已提交
2032
			pd_idx = data_disks - sector_div(stripe2, raid_disks-1);
2033 2034 2035 2036 2037
			*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
2038
			pd_idx = sector_div(stripe2, raid_disks-1);
2039 2040 2041 2042 2043 2044 2045 2046 2047 2048
			*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;

2049
		default:
2050
			BUG();
2051 2052
		}
		break;
L
Linus Torvalds 已提交
2053 2054
	}

2055 2056 2057
	if (sh) {
		sh->pd_idx = pd_idx;
		sh->qd_idx = qd_idx;
2058
		sh->ddf_layout = ddf_layout;
2059
	}
L
Linus Torvalds 已提交
2060 2061 2062 2063 2064 2065 2066 2067
	/*
	 * Finally, compute the new sector number
	 */
	new_sector = (sector_t)stripe * sectors_per_chunk + chunk_offset;
	return new_sector;
}


2068
static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous)
L
Linus Torvalds 已提交
2069
{
2070
	struct r5conf *conf = sh->raid_conf;
2071 2072
	int raid_disks = sh->disks;
	int data_disks = raid_disks - conf->max_degraded;
L
Linus Torvalds 已提交
2073
	sector_t new_sector = sh->sector, check;
2074 2075
	int sectors_per_chunk = previous ? conf->prev_chunk_sectors
					 : conf->chunk_sectors;
2076 2077
	int algorithm = previous ? conf->prev_algo
				 : conf->algorithm;
L
Linus Torvalds 已提交
2078 2079
	sector_t stripe;
	int chunk_offset;
2080 2081
	sector_t chunk_number;
	int dummy1, dd_idx = i;
L
Linus Torvalds 已提交
2082
	sector_t r_sector;
2083
	struct stripe_head sh2;
L
Linus Torvalds 已提交
2084

2085

L
Linus Torvalds 已提交
2086 2087 2088
	chunk_offset = sector_div(new_sector, sectors_per_chunk);
	stripe = new_sector;

2089 2090 2091 2092 2093
	if (i == sh->pd_idx)
		return 0;
	switch(conf->level) {
	case 4: break;
	case 5:
2094
		switch (algorithm) {
L
Linus Torvalds 已提交
2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105
		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;
2106 2107 2108 2109 2110
		case ALGORITHM_PARITY_0:
			i -= 1;
			break;
		case ALGORITHM_PARITY_N:
			break;
L
Linus Torvalds 已提交
2111
		default:
2112
			BUG();
2113 2114 2115
		}
		break;
	case 6:
2116
		if (i == sh->qd_idx)
2117
			return 0; /* It is the Q disk */
2118
		switch (algorithm) {
2119 2120
		case ALGORITHM_LEFT_ASYMMETRIC:
		case ALGORITHM_RIGHT_ASYMMETRIC:
2121 2122 2123 2124
		case ALGORITHM_ROTATING_ZERO_RESTART:
		case ALGORITHM_ROTATING_N_RESTART:
			if (sh->pd_idx == raid_disks-1)
				i--;	/* Q D D D P */
2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138
			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;
2139 2140 2141 2142 2143 2144
		case ALGORITHM_PARITY_0:
			i -= 2;
			break;
		case ALGORITHM_PARITY_N:
			break;
		case ALGORITHM_ROTATING_N_CONTINUE:
2145
			/* Like left_symmetric, but P is before Q */
2146 2147
			if (sh->pd_idx == 0)
				i--;	/* P D D D Q */
2148 2149 2150 2151 2152 2153
			else {
				/* D D Q P D */
				if (i < sh->pd_idx)
					i += raid_disks;
				i -= (sh->pd_idx + 1);
			}
2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168
			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;
2169
		default:
2170
			BUG();
2171 2172
		}
		break;
L
Linus Torvalds 已提交
2173 2174 2175
	}

	chunk_number = stripe * data_disks + i;
2176
	r_sector = chunk_number * sectors_per_chunk + chunk_offset;
L
Linus Torvalds 已提交
2177

2178
	check = raid5_compute_sector(conf, r_sector,
2179
				     previous, &dummy1, &sh2);
2180 2181
	if (check != sh->sector || dummy1 != dd_idx || sh2.pd_idx != sh->pd_idx
		|| sh2.qd_idx != sh->qd_idx) {
2182 2183
		printk(KERN_ERR "md/raid:%s: compute_blocknr: map not correct\n",
		       mdname(conf->mddev));
L
Linus Torvalds 已提交
2184 2185 2186 2187 2188 2189
		return 0;
	}
	return r_sector;
}


2190
static void
2191
schedule_reconstruction(struct stripe_head *sh, struct stripe_head_state *s,
2192
			 int rcw, int expand)
2193 2194
{
	int i, pd_idx = sh->pd_idx, disks = sh->disks;
2195
	struct r5conf *conf = sh->raid_conf;
2196
	int level = conf->level;
2197 2198 2199 2200 2201 2202 2203

	if (rcw) {
		/* 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) {
2204 2205 2206 2207
			sh->reconstruct_state = reconstruct_state_drain_run;
			set_bit(STRIPE_OP_BIODRAIN, &s->ops_request);
		} else
			sh->reconstruct_state = reconstruct_state_run;
2208

2209
		set_bit(STRIPE_OP_RECONSTRUCT, &s->ops_request);
2210 2211 2212 2213 2214 2215

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

			if (dev->towrite) {
				set_bit(R5_LOCKED, &dev->flags);
2216
				set_bit(R5_Wantdrain, &dev->flags);
2217 2218
				if (!expand)
					clear_bit(R5_UPTODATE, &dev->flags);
2219
				s->locked++;
2220 2221
			}
		}
2222
		if (s->locked + conf->max_degraded == disks)
2223
			if (!test_and_set_bit(STRIPE_FULL_WRITE, &sh->state))
2224
				atomic_inc(&conf->pending_full_writes);
2225
	} else {
2226
		BUG_ON(level == 6);
2227 2228 2229
		BUG_ON(!(test_bit(R5_UPTODATE, &sh->dev[pd_idx].flags) ||
			test_bit(R5_Wantcompute, &sh->dev[pd_idx].flags)));

2230
		sh->reconstruct_state = reconstruct_state_prexor_drain_run;
2231 2232
		set_bit(STRIPE_OP_PREXOR, &s->ops_request);
		set_bit(STRIPE_OP_BIODRAIN, &s->ops_request);
2233
		set_bit(STRIPE_OP_RECONSTRUCT, &s->ops_request);
2234 2235 2236 2237 2238 2239 2240 2241

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

			if (dev->towrite &&
			    (test_bit(R5_UPTODATE, &dev->flags) ||
2242 2243
			     test_bit(R5_Wantcompute, &dev->flags))) {
				set_bit(R5_Wantdrain, &dev->flags);
2244 2245
				set_bit(R5_LOCKED, &dev->flags);
				clear_bit(R5_UPTODATE, &dev->flags);
2246
				s->locked++;
2247 2248 2249 2250
			}
		}
	}

2251
	/* keep the parity disk(s) locked while asynchronous operations
2252 2253 2254 2255
	 * are in flight
	 */
	set_bit(R5_LOCKED, &sh->dev[pd_idx].flags);
	clear_bit(R5_UPTODATE, &sh->dev[pd_idx].flags);
2256
	s->locked++;
2257

2258 2259 2260 2261 2262 2263 2264 2265 2266
	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++;
	}

2267
	pr_debug("%s: stripe %llu locked: %d ops_request: %lx\n",
2268
		__func__, (unsigned long long)sh->sector,
2269
		s->locked, s->ops_request);
2270
}
2271

L
Linus Torvalds 已提交
2272 2273
/*
 * Each stripe/dev can have one or more bion attached.
2274
 * toread/towrite point to the first in a chain.
L
Linus Torvalds 已提交
2275 2276 2277 2278 2279
 * 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;
2280
	struct r5conf *conf = sh->raid_conf;
2281
	int firstwrite=0;
L
Linus Torvalds 已提交
2282

2283
	pr_debug("adding bi b#%llu to stripe s#%llu\n",
L
Linus Torvalds 已提交
2284 2285 2286 2287 2288
		(unsigned long long)bi->bi_sector,
		(unsigned long long)sh->sector);


	spin_lock_irq(&conf->device_lock);
2289
	if (forwrite) {
L
Linus Torvalds 已提交
2290
		bip = &sh->dev[dd_idx].towrite;
2291 2292 2293
		if (*bip == NULL && sh->dev[dd_idx].written == NULL)
			firstwrite = 1;
	} else
L
Linus Torvalds 已提交
2294 2295 2296 2297 2298 2299 2300 2301 2302
		bip = &sh->dev[dd_idx].toread;
	while (*bip && (*bip)->bi_sector < bi->bi_sector) {
		if ((*bip)->bi_sector + ((*bip)->bi_size >> 9) > bi->bi_sector)
			goto overlap;
		bip = & (*bip)->bi_next;
	}
	if (*bip && (*bip)->bi_sector < bi->bi_sector + ((bi->bi_size)>>9))
		goto overlap;

2303
	BUG_ON(*bip && bi->bi_next && (*bip) != bi->bi_next);
L
Linus Torvalds 已提交
2304 2305 2306
	if (*bip)
		bi->bi_next = *bip;
	*bip = bi;
2307
	bi->bi_phys_segments++;
2308

L
Linus Torvalds 已提交
2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321
	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)) {
			if (bi->bi_sector + (bi->bi_size>>9) >= sector)
				sector = bi->bi_sector + (bi->bi_size>>9);
		}
		if (sector >= sh->dev[dd_idx].sector + STRIPE_SECTORS)
			set_bit(R5_OVERWRITE, &sh->dev[dd_idx].flags);
	}
2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333
	spin_unlock_irq(&conf->device_lock);

	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);

	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
2334 2335 2336 2337 2338 2339 2340 2341
	return 1;

 overlap:
	set_bit(R5_Overlap, &sh->dev[dd_idx].flags);
	spin_unlock_irq(&conf->device_lock);
	return 0;
}

2342
static void end_reshape(struct r5conf *conf);
2343

2344
static void stripe_set_idx(sector_t stripe, struct r5conf *conf, int previous,
2345
			    struct stripe_head *sh)
2346
{
2347
	int sectors_per_chunk =
2348
		previous ? conf->prev_chunk_sectors : conf->chunk_sectors;
2349
	int dd_idx;
2350
	int chunk_offset = sector_div(stripe, sectors_per_chunk);
2351
	int disks = previous ? conf->previous_raid_disks : conf->raid_disks;
2352

2353 2354
	raid5_compute_sector(conf,
			     stripe * (disks - conf->max_degraded)
2355
			     *sectors_per_chunk + chunk_offset,
2356
			     previous,
2357
			     &dd_idx, sh);
2358 2359
}

2360
static void
2361
handle_failed_stripe(struct r5conf *conf, struct stripe_head *sh,
2362 2363 2364 2365 2366 2367 2368 2369 2370
				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)) {
2371
			struct md_rdev *rdev;
2372 2373 2374
			rcu_read_lock();
			rdev = rcu_dereference(conf->disks[i].rdev);
			if (rdev && test_bit(In_sync, &rdev->flags))
2375 2376 2377
				atomic_inc(&rdev->nr_pending);
			else
				rdev = NULL;
2378
			rcu_read_unlock();
2379 2380 2381 2382 2383 2384 2385 2386
			if (rdev) {
				if (!rdev_set_badblocks(
					    rdev,
					    sh->sector,
					    STRIPE_SECTORS, 0))
					md_error(conf->mddev, rdev);
				rdev_dec_pending(rdev, conf->mddev);
			}
2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403
		}
		spin_lock_irq(&conf->device_lock);
		/* fail all writes first */
		bi = sh->dev[i].towrite;
		sh->dev[i].towrite = NULL;
		if (bi) {
			s->to_write--;
			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);
2404
			if (!raid5_dec_bi_phys_segments(bi)) {
2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418
				md_write_end(conf->mddev);
				bi->bi_next = *return_bi;
				*return_bi = bi;
			}
			bi = nextbi;
		}
		/* 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);
2419
			if (!raid5_dec_bi_phys_segments(bi)) {
2420 2421 2422 2423 2424 2425 2426
				md_write_end(conf->mddev);
				bi->bi_next = *return_bi;
				*return_bi = bi;
			}
			bi = bi2;
		}

2427 2428 2429 2430 2431 2432
		/* 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))) {
2433 2434 2435 2436 2437 2438 2439 2440 2441 2442
			bi = sh->dev[i].toread;
			sh->dev[i].toread = NULL;
			if (test_and_clear_bit(R5_Overlap, &sh->dev[i].flags))
				wake_up(&conf->wait_for_overlap);
			if (bi) s->to_read--;
			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);
2443
				if (!raid5_dec_bi_phys_segments(bi)) {
2444 2445 2446 2447 2448 2449 2450 2451 2452 2453
					bi->bi_next = *return_bi;
					*return_bi = bi;
				}
				bi = nextbi;
			}
		}
		spin_unlock_irq(&conf->device_lock);
		if (bitmap_end)
			bitmap_endwrite(conf->mddev->bitmap, sh->sector,
					STRIPE_SECTORS, 0, 0);
2454 2455 2456 2457
		/* 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);
2458 2459
	}

2460 2461 2462
	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);
2463 2464
}

2465
static void
2466
handle_failed_sync(struct r5conf *conf, struct stripe_head *sh,
2467 2468 2469 2470 2471 2472 2473 2474
		   struct stripe_head_state *s)
{
	int abort = 0;
	int i;

	md_done_sync(conf->mddev, STRIPE_SECTORS, 0);
	clear_bit(STRIPE_SYNCING, &sh->state);
	s->syncing = 0;
2475
	s->replacing = 0;
2476
	/* There is nothing more to do for sync/check/repair.
2477
	 * For recover/replace we need to record a bad block on all
2478 2479 2480 2481 2482 2483 2484 2485
	 * non-sync devices, or abort the recovery
	 */
	if (!test_bit(MD_RECOVERY_RECOVER, &conf->mddev->recovery))
		return;
	/* During recovery devices cannot be removed, so locking and
	 * refcounting of rdevs is not needed
	 */
	for (i = 0; i < conf->raid_disks; i++) {
2486
		struct md_rdev *rdev = conf->disks[i].rdev;
2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498
		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))
2499 2500 2501 2502 2503 2504 2505 2506
			abort = 1;
	}
	if (abort) {
		conf->recovery_disabled = conf->mddev->recovery_disabled;
		set_bit(MD_RECOVERY_INTR, &conf->mddev->recovery);
	}
}

2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522
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;
}

2523
/* fetch_block - checks the given member device to see if its data needs
2524 2525 2526
 * to be read or computed to satisfy a request.
 *
 * Returns 1 when no more member devices need to be checked, otherwise returns
2527
 * 0 to tell the loop in handle_stripe_fill to continue
2528
 */
2529 2530
static int fetch_block(struct stripe_head *sh, struct stripe_head_state *s,
		       int disk_idx, int disks)
2531
{
2532
	struct r5dev *dev = &sh->dev[disk_idx];
2533 2534
	struct r5dev *fdev[2] = { &sh->dev[s->failed_num[0]],
				  &sh->dev[s->failed_num[1]] };
2535

2536
	/* is the data in this block needed, and can we get it? */
2537 2538 2539 2540 2541
	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 ||
2542
	     (s->replacing && want_replace(sh, disk_idx)) ||
2543 2544
	     (s->failed >= 1 && fdev[0]->toread) ||
	     (s->failed >= 2 && fdev[1]->toread) ||
2545 2546 2547
	     (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))) {
2548 2549 2550 2551 2552 2553
		/* 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) &&
2554 2555
		    (s->failed && (disk_idx == s->failed_num[0] ||
				   disk_idx == s->failed_num[1]))) {
2556 2557
			/* have disk failed, and we're requested to fetch it;
			 * do compute it
2558
			 */
2559 2560 2561 2562 2563 2564 2565 2566
			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;
2567 2568 2569 2570 2571 2572
			/* 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.
			 */
2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585
			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;
2586
			}
2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605
			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);
2606 2607
		}
	}
2608 2609 2610 2611 2612

	return 0;
}

/**
2613
 * handle_stripe_fill - read or compute data to satisfy pending requests.
2614
 */
2615 2616 2617
static void handle_stripe_fill(struct stripe_head *sh,
			       struct stripe_head_state *s,
			       int disks)
2618 2619 2620 2621 2622 2623 2624 2625 2626 2627
{
	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--; )
2628
			if (fetch_block(sh, s, i, disks))
2629
				break;
2630 2631 2632 2633
	set_bit(STRIPE_HANDLE, &sh->state);
}


2634
/* handle_stripe_clean_event
2635 2636 2637 2638
 * 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.
 */
2639
static void handle_stripe_clean_event(struct r5conf *conf,
2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652
	struct stripe_head *sh, int disks, struct bio **return_bi)
{
	int i;
	struct r5dev *dev;

	for (i = disks; i--; )
		if (sh->dev[i].written) {
			dev = &sh->dev[i];
			if (!test_bit(R5_LOCKED, &dev->flags) &&
				test_bit(R5_UPTODATE, &dev->flags)) {
				/* We can return any write requests */
				struct bio *wbi, *wbi2;
				int bitmap_end = 0;
2653
				pr_debug("Return write for disc %d\n", i);
2654 2655 2656 2657 2658 2659
				spin_lock_irq(&conf->device_lock);
				wbi = dev->written;
				dev->written = NULL;
				while (wbi && wbi->bi_sector <
					dev->sector + STRIPE_SECTORS) {
					wbi2 = r5_next_bio(wbi, dev->sector);
2660
					if (!raid5_dec_bi_phys_segments(wbi)) {
2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677
						md_write_end(conf->mddev);
						wbi->bi_next = *return_bi;
						*return_bi = wbi;
					}
					wbi = wbi2;
				}
				if (dev->towrite == NULL)
					bitmap_end = 1;
				spin_unlock_irq(&conf->device_lock);
				if (bitmap_end)
					bitmap_endwrite(conf->mddev->bitmap,
							sh->sector,
							STRIPE_SECTORS,
					 !test_bit(STRIPE_DEGRADED, &sh->state),
							0);
			}
		}
2678 2679 2680 2681

	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);
2682 2683
}

2684
static void handle_stripe_dirtying(struct r5conf *conf,
2685 2686 2687
				   struct stripe_head *sh,
				   struct stripe_head_state *s,
				   int disks)
2688 2689
{
	int rmw = 0, rcw = 0, i;
2690 2691 2692 2693 2694 2695 2696
	if (conf->max_degraded == 2) {
		/* RAID6 requires 'rcw' in current implementation
		 * Calculate the real rcw later - for now fake it
		 * look like rcw is cheaper
		 */
		rcw = 1; rmw = 2;
	} else for (i = disks; i--; ) {
2697 2698 2699 2700
		/* 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) &&
2701 2702
		    !(test_bit(R5_UPTODATE, &dev->flags) ||
		      test_bit(R5_Wantcompute, &dev->flags))) {
2703 2704 2705 2706 2707 2708 2709 2710
			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) &&
2711 2712 2713
		    !(test_bit(R5_UPTODATE, &dev->flags) ||
		    test_bit(R5_Wantcompute, &dev->flags))) {
			if (test_bit(R5_Insync, &dev->flags)) rcw++;
2714 2715 2716 2717
			else
				rcw += 2*disks;
		}
	}
2718
	pr_debug("for sector %llu, rmw=%d rcw=%d\n",
2719 2720 2721 2722 2723 2724 2725 2726
		(unsigned long long)sh->sector, rmw, rcw);
	set_bit(STRIPE_HANDLE, &sh->state);
	if (rmw < rcw && rmw > 0)
		/* prefer read-modify-write, but need to get some data */
		for (i = disks; i--; ) {
			struct r5dev *dev = &sh->dev[i];
			if ((dev->towrite || i == sh->pd_idx) &&
			    !test_bit(R5_LOCKED, &dev->flags) &&
2727 2728
			    !(test_bit(R5_UPTODATE, &dev->flags) ||
			    test_bit(R5_Wantcompute, &dev->flags)) &&
2729 2730 2731
			    test_bit(R5_Insync, &dev->flags)) {
				if (
				  test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
2732
					pr_debug("Read_old block "
2733 2734 2735 2736 2737 2738 2739 2740 2741 2742
						"%d for r-m-w\n", i);
					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);
				}
			}
		}
2743
	if (rcw <= rmw && rcw > 0) {
2744
		/* want reconstruct write, but need to get some data */
2745
		rcw = 0;
2746 2747 2748
		for (i = disks; i--; ) {
			struct r5dev *dev = &sh->dev[i];
			if (!test_bit(R5_OVERWRITE, &dev->flags) &&
2749
			    i != sh->pd_idx && i != sh->qd_idx &&
2750
			    !test_bit(R5_LOCKED, &dev->flags) &&
2751
			    !(test_bit(R5_UPTODATE, &dev->flags) ||
2752 2753 2754 2755
			      test_bit(R5_Wantcompute, &dev->flags))) {
				rcw++;
				if (!test_bit(R5_Insync, &dev->flags))
					continue; /* it's a failed drive */
2756 2757
				if (
				  test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
2758
					pr_debug("Read_old block "
2759 2760 2761 2762 2763 2764 2765 2766 2767 2768
						"%d for Reconstruct\n", i);
					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);
				}
			}
		}
2769
	}
2770 2771 2772
	/* now if nothing is locked, and if we have enough data,
	 * we can start a write request
	 */
2773 2774
	/* 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
2775 2776
	 * subsequent call wants to start a write request.  raid_run_ops only
	 * handles the case where compute block and reconstruct are requested
2777 2778 2779
	 * simultaneously.  If this is not the case then new writes need to be
	 * held off until the compute completes.
	 */
2780 2781 2782
	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)))
2783
		schedule_reconstruction(sh, s, rcw == 0, 0);
2784 2785
}

2786
static void handle_parity_checks5(struct r5conf *conf, struct stripe_head *sh,
2787 2788
				struct stripe_head_state *s, int disks)
{
2789
	struct r5dev *dev = NULL;
2790

2791
	set_bit(STRIPE_HANDLE, &sh->state);
2792

2793 2794 2795
	switch (sh->check_state) {
	case check_state_idle:
		/* start a new check operation if there are no failures */
2796 2797
		if (s->failed == 0) {
			BUG_ON(s->uptodate != disks);
2798 2799
			sh->check_state = check_state_run;
			set_bit(STRIPE_OP_CHECK, &s->ops_request);
2800 2801
			clear_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags);
			s->uptodate--;
2802
			break;
2803
		}
2804
		dev = &sh->dev[s->failed_num[0]];
2805 2806 2807 2808 2809 2810 2811 2812 2813
		/* 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
2814

2815 2816 2817 2818 2819
		/* 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);
2820
		s->locked++;
2821
		set_bit(R5_Wantwrite, &dev->flags);
2822

2823 2824
		clear_bit(STRIPE_DEGRADED, &sh->state);
		set_bit(STRIPE_INSYNC, &sh->state);
2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840
		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
2841
		if ((sh->ops.zero_sum_result & SUM_CHECK_P_RESULT) == 0)
2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852
			/* parity is correct (on disc,
			 * not in buffer any more)
			 */
			set_bit(STRIPE_INSYNC, &sh->state);
		else {
			conf->mddev->resync_mismatches += STRIPE_SECTORS;
			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;
2853
				set_bit(STRIPE_COMPUTE_RUN, &sh->state);
2854 2855 2856 2857
				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;
2858
				sh->ops.target2 = -1;
2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869
				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();
2870 2871 2872 2873
	}
}


2874
static void handle_parity_checks6(struct r5conf *conf, struct stripe_head *sh,
2875
				  struct stripe_head_state *s,
2876
				  int disks)
2877 2878
{
	int pd_idx = sh->pd_idx;
N
NeilBrown 已提交
2879
	int qd_idx = sh->qd_idx;
2880
	struct r5dev *dev;
2881 2882 2883 2884

	set_bit(STRIPE_HANDLE, &sh->state);

	BUG_ON(s->failed > 2);
2885

2886 2887 2888 2889 2890 2891
	/* 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
	 */

2892 2893 2894
	switch (sh->check_state) {
	case check_state_idle:
		/* start a new check operation if there are < 2 failures */
2895
		if (s->failed == s->q_failed) {
2896
			/* The only possible failed device holds Q, so it
2897 2898 2899
			 * makes sense to check P (If anything else were failed,
			 * we would have used P to recreate it).
			 */
2900
			sh->check_state = check_state_run;
2901
		}
2902
		if (!s->q_failed && s->failed < 2) {
2903
			/* Q is not failed, and we didn't use it to generate
2904 2905
			 * anything, so it makes sense to check it
			 */
2906 2907 2908 2909
			if (sh->check_state == check_state_run)
				sh->check_state = check_state_run_pq;
			else
				sh->check_state = check_state_run_q;
2910 2911
		}

2912 2913
		/* discard potentially stale zero_sum_result */
		sh->ops.zero_sum_result = 0;
2914

2915 2916 2917 2918
		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--;
2919
		}
2920 2921 2922 2923 2924 2925 2926
		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;
2927 2928
		}

2929 2930 2931 2932 2933
		/* we have 2-disk failure */
		BUG_ON(s->failed != 2);
		/* fall through */
	case check_state_compute_result:
		sh->check_state = check_state_idle;
2934

2935 2936 2937
		/* check that a write has not made the stripe insync */
		if (test_bit(STRIPE_INSYNC, &sh->state))
			break;
2938 2939

		/* now write out any block on a failed drive,
2940
		 * or P or Q if they were recomputed
2941
		 */
2942
		BUG_ON(s->uptodate < disks - 1); /* We don't need Q to recover */
2943
		if (s->failed == 2) {
2944
			dev = &sh->dev[s->failed_num[1]];
2945 2946 2947 2948 2949
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
		if (s->failed >= 1) {
2950
			dev = &sh->dev[s->failed_num[0]];
2951 2952 2953 2954
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
2955
		if (sh->ops.zero_sum_result & SUM_CHECK_P_RESULT) {
2956 2957 2958 2959 2960
			dev = &sh->dev[pd_idx];
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
2961
		if (sh->ops.zero_sum_result & SUM_CHECK_Q_RESULT) {
2962 2963 2964 2965 2966 2967 2968 2969
			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);
2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033
		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 {
			conf->mddev->resync_mismatches += STRIPE_SECTORS;
			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();
3034 3035 3036
	}
}

3037
static void handle_stripe_expansion(struct r5conf *conf, struct stripe_head *sh)
3038 3039 3040 3041 3042 3043
{
	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.
	 */
3044
	struct dma_async_tx_descriptor *tx = NULL;
3045 3046
	clear_bit(STRIPE_EXPAND_SOURCE, &sh->state);
	for (i = 0; i < sh->disks; i++)
N
NeilBrown 已提交
3047
		if (i != sh->pd_idx && i != sh->qd_idx) {
3048
			int dd_idx, j;
3049
			struct stripe_head *sh2;
3050
			struct async_submit_ctl submit;
3051

3052
			sector_t bn = compute_blocknr(sh, i, 1);
3053 3054
			sector_t s = raid5_compute_sector(conf, bn, 0,
							  &dd_idx, NULL);
3055
			sh2 = get_active_stripe(conf, s, 0, 1, 1);
3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067
			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;
			}
3068 3069

			/* place all the copies on one channel */
3070
			init_async_submit(&submit, 0, tx, NULL, NULL, NULL);
3071
			tx = async_memcpy(sh2->dev[dd_idx].page,
3072
					  sh->dev[i].page, 0, 0, STRIPE_SIZE,
3073
					  &submit);
3074

3075 3076 3077 3078
			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 &&
3079
				    j != sh2->qd_idx &&
3080 3081 3082 3083 3084 3085 3086
				    !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);
3087

3088
		}
3089 3090 3091 3092 3093
	/* done submitting copies, wait for them to complete */
	if (tx) {
		async_tx_ack(tx);
		dma_wait_for_async_tx(tx);
	}
3094
}
L
Linus Torvalds 已提交
3095 3096 3097 3098

/*
 * handle_stripe - do things to a stripe.
 *
3099 3100
 * 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
3101
 * Possible results:
3102 3103
 *    return some read requests which now have data
 *    return some write requests which are safely on storage
L
Linus Torvalds 已提交
3104 3105 3106 3107 3108
 *    schedule a read on some buffers
 *    schedule a write of some buffers
 *    return confirmation of parity correctness
 *
 */
3109

3110
static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
L
Linus Torvalds 已提交
3111
{
3112
	struct r5conf *conf = sh->raid_conf;
3113
	int disks = sh->disks;
3114 3115
	struct r5dev *dev;
	int i;
3116
	int do_recovery = 0;
L
Linus Torvalds 已提交
3117

3118 3119 3120 3121 3122 3123
	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
3124

3125
	/* Now to look around and see what can be done */
L
Linus Torvalds 已提交
3126
	rcu_read_lock();
3127
	spin_lock_irq(&conf->device_lock);
3128
	for (i=disks; i--; ) {
3129
		struct md_rdev *rdev;
3130 3131 3132
		sector_t first_bad;
		int bad_sectors;
		int is_bad = 0;
3133

3134
		dev = &sh->dev[i];
L
Linus Torvalds 已提交
3135

3136
		pr_debug("check %d: state 0x%lx read %p write %p written %p\n",
3137 3138
			 i, dev->flags,
			 dev->toread, dev->towrite, dev->written);
3139 3140 3141 3142 3143 3144 3145 3146
		/* 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
3147

3148
		/* now count some things */
3149 3150 3151 3152
		if (test_bit(R5_LOCKED, &dev->flags))
			s->locked++;
		if (test_bit(R5_UPTODATE, &dev->flags))
			s->uptodate++;
3153
		if (test_bit(R5_Wantcompute, &dev->flags)) {
3154 3155
			s->compute++;
			BUG_ON(s->compute > 2);
3156
		}
L
Linus Torvalds 已提交
3157

3158
		if (test_bit(R5_Wantfill, &dev->flags))
3159
			s->to_fill++;
3160
		else if (dev->toread)
3161
			s->to_read++;
3162
		if (dev->towrite) {
3163
			s->to_write++;
3164
			if (!test_bit(R5_OVERWRITE, &dev->flags))
3165
				s->non_overwrite++;
3166
		}
3167
		if (dev->written)
3168
			s->written++;
3169 3170 3171 3172 3173 3174 3175 3176 3177 3178
		/* 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 {
3179 3180
			if (rdev)
				set_bit(R5_NeedReplace, &dev->flags);
3181 3182 3183
			rdev = rcu_dereference(conf->disks[i].rdev);
			clear_bit(R5_ReadRepl, &dev->flags);
		}
3184 3185
		if (rdev && test_bit(Faulty, &rdev->flags))
			rdev = NULL;
3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197
		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);
			}
3198
		}
3199 3200 3201
		clear_bit(R5_Insync, &dev->flags);
		if (!rdev)
			/* Not in-sync */;
3202 3203 3204 3205 3206 3207 3208 3209 3210 3211
		else if (is_bad) {
			/* also not in-sync */
			if (!test_bit(WriteErrorSeen, &rdev->flags)) {
				/* 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))
3212
			set_bit(R5_Insync, &dev->flags);
3213
		else if (sh->sector + STRIPE_SECTORS <= rdev->recovery_offset)
3214
			/* in sync if before recovery_offset */
3215 3216 3217 3218 3219 3220 3221 3222 3223
			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
3224
		if (rdev && test_bit(R5_WriteError, &dev->flags)) {
3225 3226 3227 3228 3229 3230 3231
			/* 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)) {
3232
				s->handle_bad_blocks = 1;
3233
				atomic_inc(&rdev2->nr_pending);
3234 3235 3236
			} else
				clear_bit(R5_WriteError, &dev->flags);
		}
A
Adam Kwolek 已提交
3237
		if (rdev && test_bit(R5_MadeGood, &dev->flags)) {
3238 3239 3240 3241 3242
			/* 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)) {
3243
				s->handle_bad_blocks = 1;
3244
				atomic_inc(&rdev2->nr_pending);
3245 3246 3247
			} else
				clear_bit(R5_MadeGood, &dev->flags);
		}
3248 3249 3250 3251 3252 3253 3254 3255 3256
		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);
		}
3257
		if (!test_bit(R5_Insync, &dev->flags)) {
3258 3259 3260
			/* 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
3261
		}
3262 3263 3264
		if (test_bit(R5_ReadError, &dev->flags))
			clear_bit(R5_Insync, &dev->flags);
		if (!test_bit(R5_Insync, &dev->flags)) {
3265 3266 3267
			if (s->failed < 2)
				s->failed_num[s->failed] = i;
			s->failed++;
3268 3269
			if (rdev && !test_bit(Faulty, &rdev->flags))
				do_recovery = 1;
3270
		}
L
Linus Torvalds 已提交
3271
	}
3272
	spin_unlock_irq(&conf->device_lock);
3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286
	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
		 * else we can only be replacing
		 * sync and recovery both need to read all devices, and so
		 * use the same flag.
		 */
		if (do_recovery ||
		    sh->sector >= conf->mddev->recovery_cp)
			s->syncing = 1;
		else
			s->replacing = 1;
	}
L
Linus Torvalds 已提交
3287
	rcu_read_unlock();
3288 3289 3290 3291 3292
}

static void handle_stripe(struct stripe_head *sh)
{
	struct stripe_head_state s;
3293
	struct r5conf *conf = sh->raid_conf;
3294
	int i;
3295 3296
	int prexor;
	int disks = sh->disks;
3297
	struct r5dev *pdev, *qdev;
3298 3299

	clear_bit(STRIPE_HANDLE, &sh->state);
3300
	if (test_and_set_bit_lock(STRIPE_ACTIVE, &sh->state)) {
3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317
		/* already being handled, ensure it gets handled
		 * again when current action finishes */
		set_bit(STRIPE_HANDLE, &sh->state);
		return;
	}

	if (test_and_clear_bit(STRIPE_SYNC_REQUESTED, &sh->state)) {
		set_bit(STRIPE_SYNCING, &sh->state);
		clear_bit(STRIPE_INSYNC, &sh->state);
	}
	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);
3318

3319
	analyse_stripe(sh, &s);
3320

3321 3322 3323 3324 3325
	if (s.handle_bad_blocks) {
		set_bit(STRIPE_HANDLE, &sh->state);
		goto finish;
	}

3326 3327
	if (unlikely(s.blocked_rdev)) {
		if (s.syncing || s.expanding || s.expanded ||
3328
		    s.replacing || s.to_write || s.written) {
3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348
			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.
	 */
3349 3350 3351 3352 3353
	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);
3354
		if (s.syncing + s.replacing)
3355 3356
			handle_failed_sync(conf, sh, &s);
	}
3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384

	/*
	 * 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)))) &&
	    (s.q_failed || ((test_bit(R5_Insync, &qdev->flags)
			     && !test_bit(R5_LOCKED, &qdev->flags)
			     && test_bit(R5_UPTODATE, &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)
3385 3386 3387
	    || (s.syncing && (s.uptodate + s.compute < disks))
	    || s.replacing
	    || s.expanding)
3388 3389
		handle_stripe_fill(sh, &s, disks);

3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447
	/* 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
		 */
		BUG_ON(!test_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags));
		BUG_ON(sh->qd_idx >= 0 &&
		       !test_bit(R5_UPTODATE, &sh->dev[sh->qd_idx].flags));
		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;
	}

	/* 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);
	}
3448

3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462
	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)) {
3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491
		md_done_sync(conf->mddev, STRIPE_SECTORS, 1);
		clear_bit(STRIPE_SYNCING, &sh->state);
	}

	/* 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++;
				}
			}
		}


3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518
	/* 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++;
		}
	}
3519

3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535
	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);
3536

3537
finish:
3538
	/* wait for this device to become unblocked */
3539
	if (conf->mddev->external && unlikely(s.blocked_rdev))
3540
		md_wait_for_blocked_rdev(s.blocked_rdev, conf->mddev);
3541

3542 3543
	if (s.handle_bad_blocks)
		for (i = disks; i--; ) {
3544
			struct md_rdev *rdev;
3545 3546 3547 3548 3549 3550 3551 3552 3553
			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);
			}
3554 3555 3556 3557 3558 3559
			if (test_and_clear_bit(R5_MadeGood, &dev->flags)) {
				rdev = conf->disks[i].rdev;
				rdev_clear_badblocks(rdev, sh->sector,
						     STRIPE_SECTORS);
				rdev_dec_pending(rdev, conf->mddev);
			}
3560 3561
			if (test_and_clear_bit(R5_MadeGoodRepl, &dev->flags)) {
				rdev = conf->disks[i].replacement;
3562 3563 3564
				if (!rdev)
					/* rdev have been moved down */
					rdev = conf->disks[i].rdev;
3565 3566 3567 3568
				rdev_clear_badblocks(rdev, sh->sector,
						     STRIPE_SECTORS);
				rdev_dec_pending(rdev, conf->mddev);
			}
3569 3570
		}

3571 3572 3573
	if (s.ops_request)
		raid_run_ops(sh, s.ops_request);

D
Dan Williams 已提交
3574
	ops_run_io(sh, &s);
3575

3576
	if (s.dec_preread_active) {
3577
		/* We delay this until after ops_run_io so that if make_request
T
Tejun Heo 已提交
3578
		 * is waiting on a flush, it won't continue until the writes
3579 3580 3581 3582 3583 3584 3585 3586
		 * 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);
	}

3587
	return_io(s.return_bi);
3588

3589
	clear_bit_unlock(STRIPE_ACTIVE, &sh->state);
3590 3591
}

3592
static void raid5_activate_delayed(struct r5conf *conf)
3593 3594 3595 3596 3597 3598 3599 3600 3601 3602
{
	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);
3603
			list_add_tail(&sh->lru, &conf->hold_list);
3604
		}
N
NeilBrown 已提交
3605
	}
3606 3607
}

3608
static void activate_bit_delay(struct r5conf *conf)
3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621
{
	/* 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);
	}
}

3622
int md_raid5_congested(struct mddev *mddev, int bits)
3623
{
3624
	struct r5conf *conf = mddev->private;
3625 3626 3627 3628

	/* No difference between reads and writes.  Just check
	 * how busy the stripe_cache is
	 */
3629

3630 3631 3632 3633 3634 3635 3636 3637 3638
	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
3639 3640 3641 3642
EXPORT_SYMBOL_GPL(md_raid5_congested);

static int raid5_congested(void *data, int bits)
{
3643
	struct mddev *mddev = data;
N
NeilBrown 已提交
3644 3645 3646 3647

	return mddev_congested(mddev, bits) ||
		md_raid5_congested(mddev, bits);
}
3648

3649 3650 3651
/* We want read requests to align with chunks where possible,
 * but write requests don't need to.
 */
3652 3653 3654
static int raid5_mergeable_bvec(struct request_queue *q,
				struct bvec_merge_data *bvm,
				struct bio_vec *biovec)
3655
{
3656
	struct mddev *mddev = q->queuedata;
3657
	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
3658
	int max;
3659
	unsigned int chunk_sectors = mddev->chunk_sectors;
3660
	unsigned int bio_sectors = bvm->bi_size >> 9;
3661

3662
	if ((bvm->bi_rw & 1) == WRITE)
3663 3664
		return biovec->bv_len; /* always allow writes to be mergeable */

3665 3666
	if (mddev->new_chunk_sectors < mddev->chunk_sectors)
		chunk_sectors = mddev->new_chunk_sectors;
3667 3668 3669 3670 3671 3672 3673 3674
	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;
}

3675

3676
static int in_chunk_boundary(struct mddev *mddev, struct bio *bio)
3677 3678
{
	sector_t sector = bio->bi_sector + get_start_sect(bio->bi_bdev);
3679
	unsigned int chunk_sectors = mddev->chunk_sectors;
3680 3681
	unsigned int bio_sectors = bio->bi_size >> 9;

3682 3683
	if (mddev->new_chunk_sectors < mddev->chunk_sectors)
		chunk_sectors = mddev->new_chunk_sectors;
3684 3685 3686 3687
	return  chunk_sectors >=
		((sector & (chunk_sectors - 1)) + bio_sectors);
}

3688 3689 3690 3691
/*
 *  add bio to the retry LIFO  ( in O(1) ... we are in interrupt )
 *  later sampled by raid5d.
 */
3692
static void add_bio_to_retry(struct bio *bi,struct r5conf *conf)
3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705
{
	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);
}


3706
static struct bio *remove_bio_from_retry(struct r5conf *conf)
3707 3708 3709 3710 3711 3712 3713 3714 3715 3716
{
	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) {
3717
		conf->retry_read_aligned_list = bi->bi_next;
3718
		bi->bi_next = NULL;
3719 3720 3721 3722
		/*
		 * this sets the active strip count to 1 and the processed
		 * strip count to zero (upper 8 bits)
		 */
3723 3724 3725 3726 3727 3728 3729
		bi->bi_phys_segments = 1; /* biased count of active stripes */
	}

	return bi;
}


3730 3731 3732 3733 3734 3735
/*
 *  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..
 */
3736
static void raid5_align_endio(struct bio *bi, int error)
3737 3738
{
	struct bio* raid_bi  = bi->bi_private;
3739
	struct mddev *mddev;
3740
	struct r5conf *conf;
3741
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
3742
	struct md_rdev *rdev;
3743

3744
	bio_put(bi);
3745 3746 3747

	rdev = (void*)raid_bi->bi_next;
	raid_bi->bi_next = NULL;
3748 3749
	mddev = rdev->mddev;
	conf = mddev->private;
3750 3751 3752 3753

	rdev_dec_pending(rdev, conf->mddev);

	if (!error && uptodate) {
3754
		bio_endio(raid_bi, 0);
3755 3756
		if (atomic_dec_and_test(&conf->active_aligned_reads))
			wake_up(&conf->wait_for_stripe);
3757
		return;
3758 3759 3760
	}


3761
	pr_debug("raid5_align_endio : io error...handing IO for a retry\n");
3762 3763

	add_bio_to_retry(raid_bi, conf);
3764 3765
}

3766 3767
static int bio_fits_rdev(struct bio *bi)
{
3768
	struct request_queue *q = bdev_get_queue(bi->bi_bdev);
3769

3770
	if ((bi->bi_size>>9) > queue_max_sectors(q))
3771 3772
		return 0;
	blk_recount_segments(q, bi);
3773
	if (bi->bi_phys_segments > queue_max_segments(q))
3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785
		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;
}


3786
static int chunk_aligned_read(struct mddev *mddev, struct bio * raid_bio)
3787
{
3788
	struct r5conf *conf = mddev->private;
N
NeilBrown 已提交
3789
	int dd_idx;
3790
	struct bio* align_bi;
3791
	struct md_rdev *rdev;
3792
	sector_t end_sector;
3793 3794

	if (!in_chunk_boundary(mddev, raid_bio)) {
3795
		pr_debug("chunk_aligned_read : non aligned\n");
3796 3797 3798
		return 0;
	}
	/*
3799
	 * use bio_clone_mddev to make a copy of the bio
3800
	 */
3801
	align_bi = bio_clone_mddev(raid_bio, GFP_NOIO, mddev);
3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812
	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
	 */
3813 3814
	align_bi->bi_sector =  raid5_compute_sector(conf, raid_bio->bi_sector,
						    0,
3815
						    &dd_idx, NULL);
3816

3817
	end_sector = align_bi->bi_sector + (align_bi->bi_size >> 9);
3818
	rcu_read_lock();
3819 3820 3821 3822 3823 3824 3825 3826 3827 3828 3829
	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) {
3830 3831 3832
		sector_t first_bad;
		int bad_sectors;

3833 3834
		atomic_inc(&rdev->nr_pending);
		rcu_read_unlock();
3835 3836 3837 3838 3839
		raid_bio->bi_next = (void*)rdev;
		align_bi->bi_bdev =  rdev->bdev;
		align_bi->bi_flags &= ~(1 << BIO_SEG_VALID);
		align_bi->bi_sector += rdev->data_offset;

3840 3841 3842 3843
		if (!bio_fits_rdev(align_bi) ||
		    is_badblock(rdev, align_bi->bi_sector, align_bi->bi_size>>9,
				&first_bad, &bad_sectors)) {
			/* too big in some way, or has a known bad block */
3844 3845 3846 3847 3848
			bio_put(align_bi);
			rdev_dec_pending(rdev, mddev);
			return 0;
		}

3849 3850 3851 3852 3853 3854 3855
		spin_lock_irq(&conf->device_lock);
		wait_event_lock_irq(conf->wait_for_stripe,
				    conf->quiesce == 0,
				    conf->device_lock, /* nothing */);
		atomic_inc(&conf->active_aligned_reads);
		spin_unlock_irq(&conf->device_lock);

3856 3857 3858 3859
		generic_make_request(align_bi);
		return 1;
	} else {
		rcu_read_unlock();
3860
		bio_put(align_bi);
3861 3862 3863 3864
		return 0;
	}
}

3865 3866 3867 3868 3869 3870 3871 3872 3873 3874
/* __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.
 */
3875
static struct stripe_head *__get_priority_stripe(struct r5conf *conf)
3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916
{
	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;
}
3917

3918
static void make_request(struct mddev *mddev, struct bio * bi)
L
Linus Torvalds 已提交
3919
{
3920
	struct r5conf *conf = mddev->private;
3921
	int dd_idx;
L
Linus Torvalds 已提交
3922 3923 3924
	sector_t new_sector;
	sector_t logical_sector, last_sector;
	struct stripe_head *sh;
3925
	const int rw = bio_data_dir(bi);
3926
	int remaining;
3927
	int plugged;
L
Linus Torvalds 已提交
3928

T
Tejun Heo 已提交
3929 3930
	if (unlikely(bi->bi_rw & REQ_FLUSH)) {
		md_flush_request(mddev, bi);
3931
		return;
3932 3933
	}

3934
	md_write_start(mddev, bi);
3935

3936
	if (rw == READ &&
3937
	     mddev->reshape_position == MaxSector &&
3938
	     chunk_aligned_read(mddev,bi))
3939
		return;
3940

L
Linus Torvalds 已提交
3941 3942 3943 3944
	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 */
3945

3946
	plugged = mddev_check_plugged(mddev);
L
Linus Torvalds 已提交
3947 3948
	for (;logical_sector < last_sector; logical_sector += STRIPE_SECTORS) {
		DEFINE_WAIT(w);
3949
		int disks, data_disks;
3950
		int previous;
3951

3952
	retry:
3953
		previous = 0;
3954
		disks = conf->raid_disks;
3955
		prepare_to_wait(&conf->wait_for_overlap, &w, TASK_UNINTERRUPTIBLE);
3956
		if (unlikely(conf->reshape_progress != MaxSector)) {
3957
			/* spinlock is needed as reshape_progress may be
3958 3959
			 * 64bit on a 32bit platform, and so it might be
			 * possible to see a half-updated value
3960
			 * Of course reshape_progress could change after
3961 3962 3963 3964
			 * the lock is dropped, so once we get a reference
			 * to the stripe that we think it is, we will have
			 * to check again.
			 */
3965
			spin_lock_irq(&conf->device_lock);
3966 3967 3968
			if (mddev->delta_disks < 0
			    ? logical_sector < conf->reshape_progress
			    : logical_sector >= conf->reshape_progress) {
3969
				disks = conf->previous_raid_disks;
3970 3971
				previous = 1;
			} else {
3972 3973 3974
				if (mddev->delta_disks < 0
				    ? logical_sector < conf->reshape_safe
				    : logical_sector >= conf->reshape_safe) {
3975 3976 3977 3978 3979
					spin_unlock_irq(&conf->device_lock);
					schedule();
					goto retry;
				}
			}
3980 3981
			spin_unlock_irq(&conf->device_lock);
		}
3982 3983
		data_disks = disks - conf->max_degraded;

3984 3985
		new_sector = raid5_compute_sector(conf, logical_sector,
						  previous,
3986
						  &dd_idx, NULL);
3987
		pr_debug("raid456: make_request, sector %llu logical %llu\n",
L
Linus Torvalds 已提交
3988 3989 3990
			(unsigned long long)new_sector, 
			(unsigned long long)logical_sector);

3991
		sh = get_active_stripe(conf, new_sector, previous,
3992
				       (bi->bi_rw&RWA_MASK), 0);
L
Linus Torvalds 已提交
3993
		if (sh) {
3994
			if (unlikely(previous)) {
3995
				/* expansion might have moved on while waiting for a
3996 3997 3998 3999 4000 4001
				 * 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.
4002 4003 4004
				 */
				int must_retry = 0;
				spin_lock_irq(&conf->device_lock);
4005 4006 4007
				if (mddev->delta_disks < 0
				    ? logical_sector >= conf->reshape_progress
				    : logical_sector < conf->reshape_progress)
4008 4009 4010 4011 4012
					/* mismatch, need to try again */
					must_retry = 1;
				spin_unlock_irq(&conf->device_lock);
				if (must_retry) {
					release_stripe(sh);
4013
					schedule();
4014 4015 4016
					goto retry;
				}
			}
4017

4018
			if (rw == WRITE &&
4019
			    logical_sector >= mddev->suspend_lo &&
4020 4021
			    logical_sector < mddev->suspend_hi) {
				release_stripe(sh);
4022 4023 4024 4025 4026 4027 4028 4029 4030 4031
				/* 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();
4032 4033
				goto retry;
			}
4034 4035

			if (test_bit(STRIPE_EXPANDING, &sh->state) ||
4036
			    !add_stripe_bio(sh, bi, dd_idx, rw)) {
4037 4038
				/* Stripe is busy expanding or
				 * add failed due to overlap.  Flush everything
L
Linus Torvalds 已提交
4039 4040
				 * and wait a while
				 */
N
NeilBrown 已提交
4041
				md_wakeup_thread(mddev->thread);
L
Linus Torvalds 已提交
4042 4043 4044 4045 4046
				release_stripe(sh);
				schedule();
				goto retry;
			}
			finish_wait(&conf->wait_for_overlap, &w);
4047 4048
			set_bit(STRIPE_HANDLE, &sh->state);
			clear_bit(STRIPE_DELAYED, &sh->state);
T
Tejun Heo 已提交
4049
			if ((bi->bi_rw & REQ_SYNC) &&
4050 4051
			    !test_and_set_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
				atomic_inc(&conf->preread_active_stripes);
L
Linus Torvalds 已提交
4052 4053 4054 4055 4056 4057 4058 4059 4060
			release_stripe(sh);
		} 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;
		}
			
	}
4061 4062 4063
	if (!plugged)
		md_wakeup_thread(mddev->thread);

L
Linus Torvalds 已提交
4064
	spin_lock_irq(&conf->device_lock);
4065
	remaining = raid5_dec_bi_phys_segments(bi);
4066 4067
	spin_unlock_irq(&conf->device_lock);
	if (remaining == 0) {
L
Linus Torvalds 已提交
4068

4069
		if ( rw == WRITE )
L
Linus Torvalds 已提交
4070
			md_write_end(mddev);
4071

4072
		bio_endio(bi, 0);
L
Linus Torvalds 已提交
4073 4074 4075
	}
}

4076
static sector_t raid5_size(struct mddev *mddev, sector_t sectors, int raid_disks);
D
Dan Williams 已提交
4077

4078
static sector_t reshape_request(struct mddev *mddev, sector_t sector_nr, int *skipped)
L
Linus Torvalds 已提交
4079
{
4080 4081 4082 4083 4084 4085 4086 4087 4088
	/* 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.
	 */
4089
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
4090
	struct stripe_head *sh;
4091
	sector_t first_sector, last_sector;
4092 4093 4094
	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;
4095 4096
	int i;
	int dd_idx;
4097
	sector_t writepos, readpos, safepos;
4098
	sector_t stripe_addr;
4099
	int reshape_sectors;
4100
	struct list_head stripes;
4101

4102 4103 4104 4105 4106 4107
	if (sector_nr == 0) {
		/* If restarting in the middle, skip the initial sectors */
		if (mddev->delta_disks < 0 &&
		    conf->reshape_progress < raid5_size(mddev, 0, 0)) {
			sector_nr = raid5_size(mddev, 0, 0)
				- conf->reshape_progress;
4108
		} else if (mddev->delta_disks >= 0 &&
4109 4110
			   conf->reshape_progress > 0)
			sector_nr = conf->reshape_progress;
4111
		sector_div(sector_nr, new_data_disks);
4112
		if (sector_nr) {
4113 4114
			mddev->curr_resync_completed = sector_nr;
			sysfs_notify(&mddev->kobj, NULL, "sync_completed");
4115 4116 4117
			*skipped = 1;
			return sector_nr;
		}
4118 4119
	}

4120 4121 4122 4123
	/* 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
	 */
4124 4125
	if (mddev->new_chunk_sectors > mddev->chunk_sectors)
		reshape_sectors = mddev->new_chunk_sectors;
4126
	else
4127
		reshape_sectors = mddev->chunk_sectors;
4128

4129 4130 4131 4132 4133
	/* we update the metadata when there is more than 3Meg
	 * in the block range (that is rather arbitrary, should
	 * probably be time based) or when the data about to be
	 * copied would over-write the source of the data at
	 * the front of the range.
4134 4135
	 * i.e. one new_stripe along from reshape_progress new_maps
	 * to after where reshape_safe old_maps to
4136
	 */
4137
	writepos = conf->reshape_progress;
4138
	sector_div(writepos, new_data_disks);
4139 4140
	readpos = conf->reshape_progress;
	sector_div(readpos, data_disks);
4141
	safepos = conf->reshape_safe;
4142
	sector_div(safepos, data_disks);
4143
	if (mddev->delta_disks < 0) {
4144
		writepos -= min_t(sector_t, reshape_sectors, writepos);
4145
		readpos += reshape_sectors;
4146
		safepos += reshape_sectors;
4147
	} else {
4148
		writepos += reshape_sectors;
4149 4150
		readpos -= min_t(sector_t, reshape_sectors, readpos);
		safepos -= min_t(sector_t, reshape_sectors, safepos);
4151
	}
4152

4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169
	/* '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.
	 * If 'readpos' is behind 'writepos', then there is no way that we can
	 * 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???
	 */
4170
	if ((mddev->delta_disks < 0
4171 4172 4173
	     ? (safepos > writepos && readpos < writepos)
	     : (safepos < writepos && readpos > writepos)) ||
	    time_after(jiffies, conf->reshape_checkpoint + 10*HZ)) {
4174 4175 4176
		/* Cannot proceed until we've updated the superblock... */
		wait_event(conf->wait_for_overlap,
			   atomic_read(&conf->reshape_stripes)==0);
4177
		mddev->reshape_position = conf->reshape_progress;
4178
		mddev->curr_resync_completed = sector_nr;
4179
		conf->reshape_checkpoint = jiffies;
4180
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
4181
		md_wakeup_thread(mddev->thread);
4182
		wait_event(mddev->sb_wait, mddev->flags == 0 ||
4183 4184
			   kthread_should_stop());
		spin_lock_irq(&conf->device_lock);
4185
		conf->reshape_safe = mddev->reshape_position;
4186 4187
		spin_unlock_irq(&conf->device_lock);
		wake_up(&conf->wait_for_overlap);
4188
		sysfs_notify(&mddev->kobj, NULL, "sync_completed");
4189 4190
	}

4191 4192 4193 4194
	if (mddev->delta_disks < 0) {
		BUG_ON(conf->reshape_progress == 0);
		stripe_addr = writepos;
		BUG_ON((mddev->dev_sectors &
4195 4196
			~((sector_t)reshape_sectors - 1))
		       - reshape_sectors - stripe_addr
4197 4198
		       != sector_nr);
	} else {
4199
		BUG_ON(writepos != sector_nr + reshape_sectors);
4200 4201
		stripe_addr = sector_nr;
	}
4202
	INIT_LIST_HEAD(&stripes);
4203
	for (i = 0; i < reshape_sectors; i += STRIPE_SECTORS) {
4204
		int j;
4205
		int skipped_disk = 0;
4206
		sh = get_active_stripe(conf, stripe_addr+i, 0, 0, 1);
4207 4208 4209 4210 4211 4212 4213 4214 4215
		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;
4216
			if (conf->level == 6 &&
4217
			    j == sh->qd_idx)
4218
				continue;
4219
			s = compute_blocknr(sh, j, 0);
D
Dan Williams 已提交
4220
			if (s < raid5_size(mddev, 0, 0)) {
4221
				skipped_disk = 1;
4222 4223 4224 4225 4226 4227
				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);
		}
4228
		if (!skipped_disk) {
4229 4230 4231
			set_bit(STRIPE_EXPAND_READY, &sh->state);
			set_bit(STRIPE_HANDLE, &sh->state);
		}
4232
		list_add(&sh->lru, &stripes);
4233 4234
	}
	spin_lock_irq(&conf->device_lock);
4235
	if (mddev->delta_disks < 0)
4236
		conf->reshape_progress -= reshape_sectors * new_data_disks;
4237
	else
4238
		conf->reshape_progress += reshape_sectors * new_data_disks;
4239 4240 4241 4242 4243 4244 4245
	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 =
4246
		raid5_compute_sector(conf, stripe_addr*(new_data_disks),
4247
				     1, &dd_idx, NULL);
4248
	last_sector =
4249
		raid5_compute_sector(conf, ((stripe_addr+reshape_sectors)
4250
					    * new_data_disks - 1),
4251
				     1, &dd_idx, NULL);
A
Andre Noll 已提交
4252 4253
	if (last_sector >= mddev->dev_sectors)
		last_sector = mddev->dev_sectors - 1;
4254
	while (first_sector <= last_sector) {
4255
		sh = get_active_stripe(conf, first_sector, 1, 0, 1);
4256 4257 4258 4259 4260
		set_bit(STRIPE_EXPAND_SOURCE, &sh->state);
		set_bit(STRIPE_HANDLE, &sh->state);
		release_stripe(sh);
		first_sector += STRIPE_SECTORS;
	}
4261 4262 4263 4264 4265 4266 4267 4268
	/* 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);
	}
4269 4270 4271
	/* If this takes us to the resync_max point where we have to pause,
	 * then we need to write out the superblock.
	 */
4272
	sector_nr += reshape_sectors;
4273 4274
	if ((sector_nr - mddev->curr_resync_completed) * 2
	    >= mddev->resync_max - mddev->curr_resync_completed) {
4275 4276 4277
		/* Cannot proceed until we've updated the superblock... */
		wait_event(conf->wait_for_overlap,
			   atomic_read(&conf->reshape_stripes) == 0);
4278
		mddev->reshape_position = conf->reshape_progress;
4279
		mddev->curr_resync_completed = sector_nr;
4280
		conf->reshape_checkpoint = jiffies;
4281 4282 4283 4284 4285 4286
		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);
4287
		conf->reshape_safe = mddev->reshape_position;
4288 4289
		spin_unlock_irq(&conf->device_lock);
		wake_up(&conf->wait_for_overlap);
4290
		sysfs_notify(&mddev->kobj, NULL, "sync_completed");
4291
	}
4292
	return reshape_sectors;
4293 4294 4295
}

/* FIXME go_faster isn't used */
4296
static inline sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipped, int go_faster)
4297
{
4298
	struct r5conf *conf = mddev->private;
4299
	struct stripe_head *sh;
A
Andre Noll 已提交
4300
	sector_t max_sector = mddev->dev_sectors;
N
NeilBrown 已提交
4301
	sector_t sync_blocks;
4302 4303
	int still_degraded = 0;
	int i;
L
Linus Torvalds 已提交
4304

4305
	if (sector_nr >= max_sector) {
L
Linus Torvalds 已提交
4306
		/* just being told to finish up .. nothing much to do */
4307

4308 4309 4310 4311
		if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) {
			end_reshape(conf);
			return 0;
		}
4312 4313 4314 4315

		if (mddev->curr_resync < max_sector) /* aborted */
			bitmap_end_sync(mddev->bitmap, mddev->curr_resync,
					&sync_blocks, 1);
4316
		else /* completed sync */
4317 4318 4319
			conf->fullsync = 0;
		bitmap_close_sync(mddev->bitmap);

L
Linus Torvalds 已提交
4320 4321
		return 0;
	}
4322

4323 4324 4325
	/* Allow raid5_quiesce to complete */
	wait_event(conf->wait_for_overlap, conf->quiesce != 2);

4326 4327
	if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
		return reshape_request(mddev, sector_nr, skipped);
4328

4329 4330 4331 4332 4333 4334
	/* 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
	 */

4335
	/* if there is too many failed drives and we are trying
L
Linus Torvalds 已提交
4336 4337 4338
	 * to resync, then assert that we are finished, because there is
	 * nothing we can do.
	 */
4339
	if (mddev->degraded >= conf->max_degraded &&
4340
	    test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
A
Andre Noll 已提交
4341
		sector_t rv = mddev->dev_sectors - sector_nr;
4342
		*skipped = 1;
L
Linus Torvalds 已提交
4343 4344
		return rv;
	}
4345
	if (!bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, 1) &&
4346
	    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
4347 4348 4349 4350 4351 4352
	    !conf->fullsync && sync_blocks >= STRIPE_SECTORS) {
		/* 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
4353

N
NeilBrown 已提交
4354 4355
	bitmap_cond_end_sync(mddev->bitmap, sector_nr);

4356
	sh = get_active_stripe(conf, sector_nr, 0, 1, 0);
L
Linus Torvalds 已提交
4357
	if (sh == NULL) {
4358
		sh = get_active_stripe(conf, sector_nr, 0, 0, 0);
L
Linus Torvalds 已提交
4359
		/* make sure we don't swamp the stripe cache if someone else
4360
		 * is trying to get access
L
Linus Torvalds 已提交
4361
		 */
4362
		schedule_timeout_uninterruptible(1);
L
Linus Torvalds 已提交
4363
	}
4364 4365 4366 4367
	/* 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.
	 */
4368
	for (i = 0; i < conf->raid_disks; i++)
4369 4370 4371 4372 4373
		if (conf->disks[i].rdev == NULL)
			still_degraded = 1;

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

4374
	set_bit(STRIPE_SYNC_REQUESTED, &sh->state);
L
Linus Torvalds 已提交
4375

4376
	handle_stripe(sh);
L
Linus Torvalds 已提交
4377 4378 4379 4380 4381
	release_stripe(sh);

	return STRIPE_SECTORS;
}

4382
static int  retry_aligned_read(struct r5conf *conf, struct bio *raid_bio)
4383 4384 4385 4386 4387 4388 4389 4390 4391 4392 4393 4394
{
	/* 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;
4395
	int dd_idx;
4396 4397 4398 4399 4400 4401
	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);
4402
	sector = raid5_compute_sector(conf, logical_sector,
4403
				      0, &dd_idx, NULL);
4404 4405 4406
	last_sector = raid_bio->bi_sector + (raid_bio->bi_size>>9);

	for (; logical_sector < last_sector;
4407 4408 4409
	     logical_sector += STRIPE_SECTORS,
		     sector += STRIPE_SECTORS,
		     scnt++) {
4410

4411
		if (scnt < raid5_bi_hw_segments(raid_bio))
4412 4413 4414
			/* already done this stripe */
			continue;

4415
		sh = get_active_stripe(conf, sector, 0, 1, 0);
4416 4417 4418

		if (!sh) {
			/* failed to get a stripe - must wait */
4419
			raid5_set_bi_hw_segments(raid_bio, scnt);
4420 4421 4422 4423
			conf->retry_read_aligned = raid_bio;
			return handled;
		}

4424 4425
		if (!add_stripe_bio(sh, raid_bio, dd_idx, 0)) {
			release_stripe(sh);
4426
			raid5_set_bi_hw_segments(raid_bio, scnt);
4427 4428 4429 4430
			conf->retry_read_aligned = raid_bio;
			return handled;
		}

4431
		handle_stripe(sh);
4432 4433 4434 4435
		release_stripe(sh);
		handled++;
	}
	spin_lock_irq(&conf->device_lock);
4436
	remaining = raid5_dec_bi_phys_segments(raid_bio);
4437
	spin_unlock_irq(&conf->device_lock);
4438 4439
	if (remaining == 0)
		bio_endio(raid_bio, 0);
4440 4441 4442 4443 4444 4445
	if (atomic_dec_and_test(&conf->active_aligned_reads))
		wake_up(&conf->wait_for_stripe);
	return handled;
}


L
Linus Torvalds 已提交
4446 4447 4448 4449 4450 4451 4452
/*
 * 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.
 */
4453
static void raid5d(struct mddev *mddev)
L
Linus Torvalds 已提交
4454 4455
{
	struct stripe_head *sh;
4456
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
4457
	int handled;
4458
	struct blk_plug plug;
L
Linus Torvalds 已提交
4459

4460
	pr_debug("+++ raid5d active\n");
L
Linus Torvalds 已提交
4461 4462 4463

	md_check_recovery(mddev);

4464
	blk_start_plug(&plug);
L
Linus Torvalds 已提交
4465 4466 4467
	handled = 0;
	spin_lock_irq(&conf->device_lock);
	while (1) {
4468
		struct bio *bio;
L
Linus Torvalds 已提交
4469

4470 4471 4472 4473
		if (atomic_read(&mddev->plug_cnt) == 0 &&
		    !list_empty(&conf->bitmap_list)) {
			/* Now is a good time to flush some bitmap updates */
			conf->seq_flush++;
4474
			spin_unlock_irq(&conf->device_lock);
4475
			bitmap_unplug(mddev->bitmap);
4476
			spin_lock_irq(&conf->device_lock);
4477
			conf->seq_write = conf->seq_flush;
4478 4479
			activate_bit_delay(conf);
		}
4480 4481
		if (atomic_read(&mddev->plug_cnt) == 0)
			raid5_activate_delayed(conf);
4482

4483 4484 4485 4486 4487 4488 4489 4490 4491 4492
		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++;
		}

4493 4494
		sh = __get_priority_stripe(conf);

4495
		if (!sh)
L
Linus Torvalds 已提交
4496 4497 4498 4499
			break;
		spin_unlock_irq(&conf->device_lock);
		
		handled++;
4500 4501 4502
		handle_stripe(sh);
		release_stripe(sh);
		cond_resched();
L
Linus Torvalds 已提交
4503

4504 4505 4506
		if (mddev->flags & ~(1<<MD_CHANGE_PENDING))
			md_check_recovery(mddev);

L
Linus Torvalds 已提交
4507 4508
		spin_lock_irq(&conf->device_lock);
	}
4509
	pr_debug("%d stripes handled\n", handled);
L
Linus Torvalds 已提交
4510 4511 4512

	spin_unlock_irq(&conf->device_lock);

4513
	async_tx_issue_pending_all();
4514
	blk_finish_plug(&plug);
L
Linus Torvalds 已提交
4515

4516
	pr_debug("--- raid5d inactive\n");
L
Linus Torvalds 已提交
4517 4518
}

4519
static ssize_t
4520
raid5_show_stripe_cache_size(struct mddev *mddev, char *page)
4521
{
4522
	struct r5conf *conf = mddev->private;
4523 4524 4525 4526
	if (conf)
		return sprintf(page, "%d\n", conf->max_nr_stripes);
	else
		return 0;
4527 4528
}

4529
int
4530
raid5_set_cache_size(struct mddev *mddev, int size)
4531
{
4532
	struct r5conf *conf = mddev->private;
4533 4534
	int err;

4535
	if (size <= 16 || size > 32768)
4536
		return -EINVAL;
4537
	while (size < conf->max_nr_stripes) {
4538 4539 4540 4541 4542
		if (drop_one_stripe(conf))
			conf->max_nr_stripes--;
		else
			break;
	}
4543 4544 4545
	err = md_allow_write(mddev);
	if (err)
		return err;
4546
	while (size > conf->max_nr_stripes) {
4547 4548 4549 4550
		if (grow_one_stripe(conf))
			conf->max_nr_stripes++;
		else break;
	}
4551 4552 4553 4554 4555
	return 0;
}
EXPORT_SYMBOL(raid5_set_cache_size);

static ssize_t
4556
raid5_store_stripe_cache_size(struct mddev *mddev, const char *page, size_t len)
4557
{
4558
	struct r5conf *conf = mddev->private;
4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571
	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;
4572 4573
	return len;
}
4574

4575 4576 4577 4578
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);
4579

4580
static ssize_t
4581
raid5_show_preread_threshold(struct mddev *mddev, char *page)
4582
{
4583
	struct r5conf *conf = mddev->private;
4584 4585 4586 4587 4588 4589 4590
	if (conf)
		return sprintf(page, "%d\n", conf->bypass_threshold);
	else
		return 0;
}

static ssize_t
4591
raid5_store_preread_threshold(struct mddev *mddev, const char *page, size_t len)
4592
{
4593
	struct r5conf *conf = mddev->private;
4594
	unsigned long new;
4595 4596 4597 4598 4599
	if (len >= PAGE_SIZE)
		return -EINVAL;
	if (!conf)
		return -ENODEV;

4600
	if (strict_strtoul(page, 10, &new))
4601
		return -EINVAL;
4602
	if (new > conf->max_nr_stripes)
4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613
		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);

4614
static ssize_t
4615
stripe_cache_active_show(struct mddev *mddev, char *page)
4616
{
4617
	struct r5conf *conf = mddev->private;
4618 4619 4620 4621
	if (conf)
		return sprintf(page, "%d\n", atomic_read(&conf->active_stripes));
	else
		return 0;
4622 4623
}

4624 4625
static struct md_sysfs_entry
raid5_stripecache_active = __ATTR_RO(stripe_cache_active);
4626

4627
static struct attribute *raid5_attrs[] =  {
4628 4629
	&raid5_stripecache_size.attr,
	&raid5_stripecache_active.attr,
4630
	&raid5_preread_bypass_threshold.attr,
4631 4632
	NULL,
};
4633 4634 4635
static struct attribute_group raid5_attrs_group = {
	.name = NULL,
	.attrs = raid5_attrs,
4636 4637
};

4638
static sector_t
4639
raid5_size(struct mddev *mddev, sector_t sectors, int raid_disks)
4640
{
4641
	struct r5conf *conf = mddev->private;
4642 4643 4644

	if (!sectors)
		sectors = mddev->dev_sectors;
4645
	if (!raid_disks)
4646
		/* size is defined by the smallest of previous and new size */
4647
		raid_disks = min(conf->raid_disks, conf->previous_raid_disks);
4648

4649
	sectors &= ~((sector_t)mddev->chunk_sectors - 1);
4650
	sectors &= ~((sector_t)mddev->new_chunk_sectors - 1);
4651 4652 4653
	return sectors * (raid_disks - conf->max_degraded);
}

4654
static void raid5_free_percpu(struct r5conf *conf)
4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665
{
	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);
4666
		kfree(percpu->scribble);
4667 4668 4669 4670 4671 4672 4673 4674 4675
	}
#ifdef CONFIG_HOTPLUG_CPU
	unregister_cpu_notifier(&conf->cpu_notify);
#endif
	put_online_cpus();

	free_percpu(conf->percpu);
}

4676
static void free_conf(struct r5conf *conf)
4677 4678
{
	shrink_stripes(conf);
4679
	raid5_free_percpu(conf);
4680 4681 4682 4683 4684
	kfree(conf->disks);
	kfree(conf->stripe_hashtbl);
	kfree(conf);
}

4685 4686 4687 4688
#ifdef CONFIG_HOTPLUG_CPU
static int raid456_cpu_notify(struct notifier_block *nfb, unsigned long action,
			      void *hcpu)
{
4689
	struct r5conf *conf = container_of(nfb, struct r5conf, cpu_notify);
4690 4691 4692 4693 4694 4695
	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:
4696
		if (conf->level == 6 && !percpu->spare_page)
4697
			percpu->spare_page = alloc_page(GFP_KERNEL);
4698 4699 4700 4701 4702 4703 4704
		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);
4705 4706
			pr_err("%s: failed memory allocation for cpu%ld\n",
			       __func__, cpu);
4707
			return notifier_from_errno(-ENOMEM);
4708 4709 4710 4711 4712
		}
		break;
	case CPU_DEAD:
	case CPU_DEAD_FROZEN:
		safe_put_page(percpu->spare_page);
4713
		kfree(percpu->scribble);
4714
		percpu->spare_page = NULL;
4715
		percpu->scribble = NULL;
4716 4717 4718 4719 4720 4721 4722 4723
		break;
	default:
		break;
	}
	return NOTIFY_OK;
}
#endif

4724
static int raid5_alloc_percpu(struct r5conf *conf)
4725 4726 4727
{
	unsigned long cpu;
	struct page *spare_page;
4728
	struct raid5_percpu __percpu *allcpus;
4729
	void *scribble;
4730 4731 4732 4733 4734 4735 4736 4737 4738 4739
	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) {
4740 4741 4742 4743 4744 4745 4746 4747
		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;
		}
4748
		scribble = kmalloc(conf->scribble_len, GFP_KERNEL);
4749
		if (!scribble) {
4750 4751 4752
			err = -ENOMEM;
			break;
		}
4753
		per_cpu_ptr(conf->percpu, cpu)->scribble = scribble;
4754 4755 4756 4757 4758 4759 4760 4761 4762 4763 4764 4765
	}
#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;
}

4766
static struct r5conf *setup_conf(struct mddev *mddev)
L
Linus Torvalds 已提交
4767
{
4768
	struct r5conf *conf;
4769
	int raid_disk, memory, max_disks;
4770
	struct md_rdev *rdev;
L
Linus Torvalds 已提交
4771 4772
	struct disk_info *disk;

N
NeilBrown 已提交
4773 4774 4775
	if (mddev->new_level != 5
	    && mddev->new_level != 4
	    && mddev->new_level != 6) {
4776
		printk(KERN_ERR "md/raid:%s: raid level not set to 4/5/6 (%d)\n",
N
NeilBrown 已提交
4777 4778
		       mdname(mddev), mddev->new_level);
		return ERR_PTR(-EIO);
L
Linus Torvalds 已提交
4779
	}
N
NeilBrown 已提交
4780 4781 4782 4783
	if ((mddev->new_level == 5
	     && !algorithm_valid_raid5(mddev->new_layout)) ||
	    (mddev->new_level == 6
	     && !algorithm_valid_raid6(mddev->new_layout))) {
4784
		printk(KERN_ERR "md/raid:%s: layout %d not supported\n",
N
NeilBrown 已提交
4785 4786
		       mdname(mddev), mddev->new_layout);
		return ERR_PTR(-EIO);
4787
	}
N
NeilBrown 已提交
4788
	if (mddev->new_level == 6 && mddev->raid_disks < 4) {
4789
		printk(KERN_ERR "md/raid:%s: not enough configured devices (%d, minimum 4)\n",
N
NeilBrown 已提交
4790 4791
		       mdname(mddev), mddev->raid_disks);
		return ERR_PTR(-EINVAL);
4792 4793
	}

4794 4795 4796
	if (!mddev->new_chunk_sectors ||
	    (mddev->new_chunk_sectors << 9) % PAGE_SIZE ||
	    !is_power_of_2(mddev->new_chunk_sectors)) {
4797 4798
		printk(KERN_ERR "md/raid:%s: invalid chunk size %d\n",
		       mdname(mddev), mddev->new_chunk_sectors << 9);
N
NeilBrown 已提交
4799
		return ERR_PTR(-EINVAL);
4800 4801
	}

4802
	conf = kzalloc(sizeof(struct r5conf), GFP_KERNEL);
N
NeilBrown 已提交
4803
	if (conf == NULL)
L
Linus Torvalds 已提交
4804
		goto abort;
4805 4806 4807 4808 4809 4810 4811 4812 4813 4814 4815 4816
	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;
4817
	conf->recovery_disabled = mddev->recovery_disabled - 1;
N
NeilBrown 已提交
4818 4819 4820 4821 4822

	conf->raid_disks = mddev->raid_disks;
	if (mddev->reshape_position == MaxSector)
		conf->previous_raid_disks = mddev->raid_disks;
	else
4823
		conf->previous_raid_disks = mddev->raid_disks - mddev->delta_disks;
4824 4825
	max_disks = max(conf->raid_disks, conf->previous_raid_disks);
	conf->scribble_len = scribble_len(max_disks);
4826

4827
	conf->disks = kzalloc(max_disks * sizeof(struct disk_info),
4828 4829 4830
			      GFP_KERNEL);
	if (!conf->disks)
		goto abort;
4831

L
Linus Torvalds 已提交
4832 4833
	conf->mddev = mddev;

4834
	if ((conf->stripe_hashtbl = kzalloc(PAGE_SIZE, GFP_KERNEL)) == NULL)
L
Linus Torvalds 已提交
4835 4836
		goto abort;

4837 4838 4839 4840
	conf->level = mddev->new_level;
	if (raid5_alloc_percpu(conf) != 0)
		goto abort;

4841
	pr_debug("raid456: run(%s) called.\n", mdname(mddev));
L
Linus Torvalds 已提交
4842

4843
	list_for_each_entry(rdev, &mddev->disks, same_set) {
L
Linus Torvalds 已提交
4844
		raid_disk = rdev->raid_disk;
4845
		if (raid_disk >= max_disks
L
Linus Torvalds 已提交
4846 4847 4848 4849
		    || raid_disk < 0)
			continue;
		disk = conf->disks + raid_disk;

4850 4851 4852 4853 4854 4855 4856 4857 4858
		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
4859

4860
		if (test_bit(In_sync, &rdev->flags)) {
L
Linus Torvalds 已提交
4861
			char b[BDEVNAME_SIZE];
4862 4863 4864
			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
4865
		} else if (rdev->saved_raid_disk != raid_disk)
4866 4867
			/* Cannot rely on bitmap to complete recovery */
			conf->fullsync = 1;
L
Linus Torvalds 已提交
4868 4869
	}

4870
	conf->chunk_sectors = mddev->new_chunk_sectors;
N
NeilBrown 已提交
4871
	conf->level = mddev->new_level;
4872 4873 4874 4875
	if (conf->level == 6)
		conf->max_degraded = 2;
	else
		conf->max_degraded = 1;
N
NeilBrown 已提交
4876
	conf->algorithm = mddev->new_layout;
L
Linus Torvalds 已提交
4877
	conf->max_nr_stripes = NR_STRIPES;
4878
	conf->reshape_progress = mddev->reshape_position;
4879
	if (conf->reshape_progress != MaxSector) {
4880
		conf->prev_chunk_sectors = mddev->chunk_sectors;
4881 4882
		conf->prev_algo = mddev->layout;
	}
L
Linus Torvalds 已提交
4883

N
NeilBrown 已提交
4884
	memory = conf->max_nr_stripes * (sizeof(struct stripe_head) +
4885
		 max_disks * ((sizeof(struct bio) + PAGE_SIZE))) / 1024;
N
NeilBrown 已提交
4886 4887
	if (grow_stripes(conf, conf->max_nr_stripes)) {
		printk(KERN_ERR
4888 4889
		       "md/raid:%s: couldn't allocate %dkB for buffers\n",
		       mdname(mddev), memory);
N
NeilBrown 已提交
4890 4891
		goto abort;
	} else
4892 4893
		printk(KERN_INFO "md/raid:%s: allocated %dkB\n",
		       mdname(mddev), memory);
L
Linus Torvalds 已提交
4894

4895
	conf->thread = md_register_thread(raid5d, mddev, NULL);
N
NeilBrown 已提交
4896 4897
	if (!conf->thread) {
		printk(KERN_ERR
4898
		       "md/raid:%s: couldn't allocate thread.\n",
N
NeilBrown 已提交
4899
		       mdname(mddev));
4900 4901
		goto abort;
	}
N
NeilBrown 已提交
4902 4903 4904 4905 4906

	return conf;

 abort:
	if (conf) {
4907
		free_conf(conf);
N
NeilBrown 已提交
4908 4909 4910 4911 4912
		return ERR_PTR(-EIO);
	} else
		return ERR_PTR(-ENOMEM);
}

4913 4914 4915 4916 4917 4918 4919 4920 4921 4922 4923 4924 4925 4926 4927 4928 4929 4930 4931 4932 4933 4934 4935 4936 4937 4938 4939

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;
}

4940
static int run(struct mddev *mddev)
N
NeilBrown 已提交
4941
{
4942
	struct r5conf *conf;
4943
	int working_disks = 0;
4944
	int dirty_parity_disks = 0;
4945
	struct md_rdev *rdev;
4946
	sector_t reshape_offset = 0;
4947
	int i;
N
NeilBrown 已提交
4948

4949
	if (mddev->recovery_cp != MaxSector)
4950
		printk(KERN_NOTICE "md/raid:%s: not clean"
4951 4952
		       " -- starting background reconstruction\n",
		       mdname(mddev));
N
NeilBrown 已提交
4953 4954 4955 4956 4957 4958 4959 4960
	if (mddev->reshape_position != MaxSector) {
		/* Check that we can continue the reshape.
		 * Currently only disks can change, it must
		 * increase, and we must be past the point where
		 * a stripe over-writes itself
		 */
		sector_t here_new, here_old;
		int old_disks;
4961
		int max_degraded = (mddev->level == 6 ? 2 : 1);
N
NeilBrown 已提交
4962

4963
		if (mddev->new_level != mddev->level) {
4964
			printk(KERN_ERR "md/raid:%s: unsupported reshape "
N
NeilBrown 已提交
4965 4966 4967 4968 4969 4970 4971 4972 4973 4974
			       "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;
4975
		if (sector_div(here_new, mddev->new_chunk_sectors *
N
NeilBrown 已提交
4976
			       (mddev->raid_disks - max_degraded))) {
4977 4978
			printk(KERN_ERR "md/raid:%s: reshape_position not "
			       "on a stripe boundary\n", mdname(mddev));
N
NeilBrown 已提交
4979 4980
			return -EINVAL;
		}
4981
		reshape_offset = here_new * mddev->new_chunk_sectors;
N
NeilBrown 已提交
4982 4983
		/* here_new is the stripe we will write to */
		here_old = mddev->reshape_position;
4984
		sector_div(here_old, mddev->chunk_sectors *
N
NeilBrown 已提交
4985 4986 4987
			   (old_disks-max_degraded));
		/* here_old is the first stripe that we might need to read
		 * from */
4988 4989 4990 4991 4992 4993 4994 4995 4996 4997 4998
		if (mddev->delta_disks == 0) {
			/* We cannot be sure it is safe to start an in-place
			 * reshape.  It is only safe if user-space if monitoring
			 * 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.
			 */
			if ((here_new * mddev->new_chunk_sectors != 
			     here_old * mddev->chunk_sectors) ||
			    mddev->ro == 0) {
4999 5000 5001
				printk(KERN_ERR "md/raid:%s: in-place reshape must be started"
				       " in read-only mode - aborting\n",
				       mdname(mddev));
5002 5003 5004 5005 5006 5007 5008
				return -EINVAL;
			}
		} else if (mddev->delta_disks < 0
		    ? (here_new * mddev->new_chunk_sectors <=
		       here_old * mddev->chunk_sectors)
		    : (here_new * mddev->new_chunk_sectors >=
		       here_old * mddev->chunk_sectors)) {
N
NeilBrown 已提交
5009
			/* Reading from the same stripe as writing to - bad */
5010 5011 5012
			printk(KERN_ERR "md/raid:%s: reshape_position too early for "
			       "auto-recovery - aborting.\n",
			       mdname(mddev));
N
NeilBrown 已提交
5013 5014
			return -EINVAL;
		}
5015 5016
		printk(KERN_INFO "md/raid:%s: reshape will continue\n",
		       mdname(mddev));
N
NeilBrown 已提交
5017 5018 5019 5020
		/* OK, we should be able to continue; */
	} else {
		BUG_ON(mddev->level != mddev->new_level);
		BUG_ON(mddev->layout != mddev->new_layout);
5021
		BUG_ON(mddev->chunk_sectors != mddev->new_chunk_sectors);
N
NeilBrown 已提交
5022
		BUG_ON(mddev->delta_disks != 0);
L
Linus Torvalds 已提交
5023
	}
N
NeilBrown 已提交
5024

5025 5026 5027 5028 5029
	if (mddev->private == NULL)
		conf = setup_conf(mddev);
	else
		conf = mddev->private;

N
NeilBrown 已提交
5030 5031 5032 5033 5034 5035 5036
	if (IS_ERR(conf))
		return PTR_ERR(conf);

	mddev->thread = conf->thread;
	conf->thread = NULL;
	mddev->private = conf;

5037 5038 5039 5040 5041 5042 5043 5044 5045 5046 5047
	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)
5048
			continue;
5049 5050 5051 5052 5053 5054 5055
		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;
		}
5056
		if (test_bit(In_sync, &rdev->flags)) {
N
NeilBrown 已提交
5057
			working_disks++;
5058 5059
			continue;
		}
5060 5061 5062 5063 5064 5065 5066 5067 5068 5069 5070 5071 5072 5073 5074 5075 5076 5077 5078 5079 5080 5081 5082 5083 5084 5085 5086 5087
		/* 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
5088

5089 5090 5091
	/*
	 * 0 for a fully functional array, 1 or 2 for a degraded array.
	 */
5092
	mddev->degraded = calc_degraded(conf);
N
NeilBrown 已提交
5093

5094
	if (has_failed(conf)) {
5095
		printk(KERN_ERR "md/raid:%s: not enough operational devices"
L
Linus Torvalds 已提交
5096
			" (%d/%d failed)\n",
5097
			mdname(mddev), mddev->degraded, conf->raid_disks);
L
Linus Torvalds 已提交
5098 5099 5100
		goto abort;
	}

N
NeilBrown 已提交
5101
	/* device size must be a multiple of chunk size */
5102
	mddev->dev_sectors &= ~(mddev->chunk_sectors - 1);
N
NeilBrown 已提交
5103 5104
	mddev->resync_max_sectors = mddev->dev_sectors;

5105
	if (mddev->degraded > dirty_parity_disks &&
L
Linus Torvalds 已提交
5106
	    mddev->recovery_cp != MaxSector) {
5107 5108
		if (mddev->ok_start_degraded)
			printk(KERN_WARNING
5109 5110
			       "md/raid:%s: starting dirty degraded array"
			       " - data corruption possible.\n",
5111 5112 5113
			       mdname(mddev));
		else {
			printk(KERN_ERR
5114
			       "md/raid:%s: cannot start dirty degraded array.\n",
5115 5116 5117
			       mdname(mddev));
			goto abort;
		}
L
Linus Torvalds 已提交
5118 5119 5120
	}

	if (mddev->degraded == 0)
5121 5122
		printk(KERN_INFO "md/raid:%s: raid level %d active with %d out of %d"
		       " devices, algorithm %d\n", mdname(mddev), conf->level,
5123 5124
		       mddev->raid_disks-mddev->degraded, mddev->raid_disks,
		       mddev->new_layout);
L
Linus Torvalds 已提交
5125
	else
5126 5127 5128 5129 5130
		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
5131 5132 5133

	print_raid5_conf(conf);

5134 5135
	if (conf->reshape_progress != MaxSector) {
		conf->reshape_safe = conf->reshape_progress;
5136 5137 5138 5139 5140 5141
		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,
5142
							"reshape");
5143 5144
	}

L
Linus Torvalds 已提交
5145 5146

	/* Ok, everything is just fine now */
5147 5148
	if (mddev->to_remove == &raid5_attrs_group)
		mddev->to_remove = NULL;
N
NeilBrown 已提交
5149 5150
	else if (mddev->kobj.sd &&
	    sysfs_create_group(&mddev->kobj, &raid5_attrs_group))
5151
		printk(KERN_WARNING
5152
		       "raid5: failed to create sysfs attributes for %s\n",
5153
		       mdname(mddev));
5154
	md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
5155

5156
	if (mddev->queue) {
5157
		int chunk_size;
5158 5159 5160 5161 5162 5163 5164 5165 5166
		/* 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
5167

5168
		blk_queue_merge_bvec(mddev->queue, raid5_mergeable_bvec);
5169

N
NeilBrown 已提交
5170 5171
		mddev->queue->backing_dev_info.congested_data = mddev;
		mddev->queue->backing_dev_info.congested_fn = raid5_congested;
5172

5173 5174 5175 5176
		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));
5177

5178 5179 5180 5181
		list_for_each_entry(rdev, &mddev->disks, same_set)
			disk_stack_limits(mddev->gendisk, rdev->bdev,
					  rdev->data_offset << 9);
	}
5182

L
Linus Torvalds 已提交
5183 5184
	return 0;
abort:
5185
	md_unregister_thread(&mddev->thread);
N
NeilBrown 已提交
5186 5187
	print_raid5_conf(conf);
	free_conf(conf);
L
Linus Torvalds 已提交
5188
	mddev->private = NULL;
5189
	printk(KERN_ALERT "md/raid:%s: failed to run raid set.\n", mdname(mddev));
L
Linus Torvalds 已提交
5190 5191 5192
	return -EIO;
}

5193
static int stop(struct mddev *mddev)
L
Linus Torvalds 已提交
5194
{
5195
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5196

5197
	md_unregister_thread(&mddev->thread);
N
NeilBrown 已提交
5198 5199
	if (mddev->queue)
		mddev->queue->backing_dev_info.congested_fn = NULL;
5200
	free_conf(conf);
5201 5202
	mddev->private = NULL;
	mddev->to_remove = &raid5_attrs_group;
L
Linus Torvalds 已提交
5203 5204 5205
	return 0;
}

5206
static void status(struct seq_file *seq, struct mddev *mddev)
L
Linus Torvalds 已提交
5207
{
5208
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5209 5210
	int i;

5211 5212
	seq_printf(seq, " level %d, %dk chunk, algorithm %d", mddev->level,
		mddev->chunk_sectors / 2, mddev->layout);
5213
	seq_printf (seq, " [%d/%d] [", conf->raid_disks, conf->raid_disks - mddev->degraded);
L
Linus Torvalds 已提交
5214 5215 5216
	for (i = 0; i < conf->raid_disks; i++)
		seq_printf (seq, "%s",
			       conf->disks[i].rdev &&
5217
			       test_bit(In_sync, &conf->disks[i].rdev->flags) ? "U" : "_");
L
Linus Torvalds 已提交
5218 5219 5220
	seq_printf (seq, "]");
}

5221
static void print_raid5_conf (struct r5conf *conf)
L
Linus Torvalds 已提交
5222 5223 5224 5225
{
	int i;
	struct disk_info *tmp;

5226
	printk(KERN_DEBUG "RAID conf printout:\n");
L
Linus Torvalds 已提交
5227 5228 5229 5230
	if (!conf) {
		printk("(conf==NULL)\n");
		return;
	}
5231 5232 5233
	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
5234 5235 5236 5237 5238

	for (i = 0; i < conf->raid_disks; i++) {
		char b[BDEVNAME_SIZE];
		tmp = conf->disks + i;
		if (tmp->rdev)
5239 5240 5241
			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
5242 5243 5244
	}
}

5245
static int raid5_spare_active(struct mddev *mddev)
L
Linus Torvalds 已提交
5246 5247
{
	int i;
5248
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5249
	struct disk_info *tmp;
5250 5251
	int count = 0;
	unsigned long flags;
L
Linus Torvalds 已提交
5252 5253 5254

	for (i = 0; i < conf->raid_disks; i++) {
		tmp = conf->disks + i;
5255 5256 5257 5258 5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269 5270 5271 5272 5273
		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
5274
		    && tmp->rdev->recovery_offset == MaxSector
5275
		    && !test_bit(Faulty, &tmp->rdev->flags)
5276
		    && !test_and_set_bit(In_sync, &tmp->rdev->flags)) {
5277
			count++;
5278
			sysfs_notify_dirent_safe(tmp->rdev->sysfs_state);
L
Linus Torvalds 已提交
5279 5280
		}
	}
5281
	spin_lock_irqsave(&conf->device_lock, flags);
5282
	mddev->degraded = calc_degraded(conf);
5283
	spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
5284
	print_raid5_conf(conf);
5285
	return count;
L
Linus Torvalds 已提交
5286 5287
}

5288
static int raid5_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
5289
{
5290
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5291
	int err = 0;
5292
	int number = rdev->raid_disk;
5293
	struct md_rdev **rdevp;
L
Linus Torvalds 已提交
5294 5295 5296
	struct disk_info *p = conf->disks + number;

	print_raid5_conf(conf);
5297 5298 5299 5300 5301 5302 5303 5304 5305 5306 5307 5308 5309 5310 5311 5312 5313 5314 5315 5316 5317 5318
	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) &&
5319
	    (!p->replacement || p->replacement == rdev) &&
5320 5321 5322 5323 5324 5325 5326 5327 5328 5329
	    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;
5330 5331 5332 5333 5334 5335 5336 5337 5338 5339 5340 5341 5342 5343
	} 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
5344 5345 5346 5347 5348 5349
abort:

	print_raid5_conf(conf);
	return err;
}

5350
static int raid5_add_disk(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
5351
{
5352
	struct r5conf *conf = mddev->private;
5353
	int err = -EEXIST;
L
Linus Torvalds 已提交
5354 5355
	int disk;
	struct disk_info *p;
5356 5357
	int first = 0;
	int last = conf->raid_disks - 1;
L
Linus Torvalds 已提交
5358

5359 5360 5361
	if (mddev->recovery_disabled == conf->recovery_disabled)
		return -EBUSY;

5362
	if (has_failed(conf))
L
Linus Torvalds 已提交
5363
		/* no point adding a device */
5364
		return -EINVAL;
L
Linus Torvalds 已提交
5365

5366 5367
	if (rdev->raid_disk >= 0)
		first = last = rdev->raid_disk;
L
Linus Torvalds 已提交
5368 5369

	/*
5370 5371
	 * find the disk ... but prefer rdev->saved_raid_disk
	 * if possible.
L
Linus Torvalds 已提交
5372
	 */
5373
	if (rdev->saved_raid_disk >= 0 &&
5374
	    rdev->saved_raid_disk >= first &&
5375 5376 5377
	    conf->disks[rdev->saved_raid_disk].rdev == NULL)
		disk = rdev->saved_raid_disk;
	else
5378 5379
		disk = first;
	for ( ; disk <= last ; disk++)
L
Linus Torvalds 已提交
5380
		if ((p=conf->disks + disk)->rdev == NULL) {
5381
			clear_bit(In_sync, &rdev->flags);
L
Linus Torvalds 已提交
5382
			rdev->raid_disk = disk;
5383
			err = 0;
5384 5385
			if (rdev->saved_raid_disk != disk)
				conf->fullsync = 1;
5386
			rcu_assign_pointer(p->rdev, rdev);
L
Linus Torvalds 已提交
5387 5388 5389
			break;
		}
	print_raid5_conf(conf);
5390
	return err;
L
Linus Torvalds 已提交
5391 5392
}

5393
static int raid5_resize(struct mddev *mddev, sector_t sectors)
L
Linus Torvalds 已提交
5394 5395 5396 5397 5398 5399 5400 5401
{
	/* 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.
	 */
5402
	sectors &= ~((sector_t)mddev->chunk_sectors - 1);
5403 5404
	md_set_array_sectors(mddev, raid5_size(mddev, sectors,
					       mddev->raid_disks));
D
Dan Williams 已提交
5405 5406 5407
	if (mddev->array_sectors >
	    raid5_size(mddev, sectors, mddev->raid_disks))
		return -EINVAL;
5408
	set_capacity(mddev->gendisk, mddev->array_sectors);
5409
	revalidate_disk(mddev->gendisk);
5410 5411
	if (sectors > mddev->dev_sectors &&
	    mddev->recovery_cp > mddev->dev_sectors) {
A
Andre Noll 已提交
5412
		mddev->recovery_cp = mddev->dev_sectors;
L
Linus Torvalds 已提交
5413 5414
		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
	}
A
Andre Noll 已提交
5415
	mddev->dev_sectors = sectors;
5416
	mddev->resync_max_sectors = sectors;
L
Linus Torvalds 已提交
5417 5418 5419
	return 0;
}

5420
static int check_stripe_cache(struct mddev *mddev)
5421 5422 5423 5424 5425 5426 5427 5428 5429
{
	/* 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.
	 */
5430
	struct r5conf *conf = mddev->private;
5431 5432 5433 5434
	if (((mddev->chunk_sectors << 9) / STRIPE_SIZE) * 4
	    > conf->max_nr_stripes ||
	    ((mddev->new_chunk_sectors << 9) / STRIPE_SIZE) * 4
	    > conf->max_nr_stripes) {
5435 5436
		printk(KERN_WARNING "md/raid:%s: reshape: not enough stripes.  Needed %lu\n",
		       mdname(mddev),
5437 5438 5439 5440 5441 5442 5443
		       ((max(mddev->chunk_sectors, mddev->new_chunk_sectors) << 9)
			/ STRIPE_SIZE)*4);
		return 0;
	}
	return 1;
}

5444
static int check_reshape(struct mddev *mddev)
5445
{
5446
	struct r5conf *conf = mddev->private;
5447

5448 5449
	if (mddev->delta_disks == 0 &&
	    mddev->new_layout == mddev->layout &&
5450
	    mddev->new_chunk_sectors == mddev->chunk_sectors)
5451
		return 0; /* nothing to do */
5452 5453 5454
	if (mddev->bitmap)
		/* Cannot grow a bitmap yet */
		return -EBUSY;
5455
	if (has_failed(conf))
5456 5457 5458 5459 5460 5461 5462 5463 5464 5465 5466 5467 5468
		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;
	}
5469

5470
	if (!check_stripe_cache(mddev))
5471 5472
		return -ENOSPC;

5473
	return resize_stripes(conf, conf->raid_disks + mddev->delta_disks);
5474 5475
}

5476
static int raid5_start_reshape(struct mddev *mddev)
5477
{
5478
	struct r5conf *conf = mddev->private;
5479
	struct md_rdev *rdev;
5480
	int spares = 0;
5481
	unsigned long flags;
5482

5483
	if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
5484 5485
		return -EBUSY;

5486 5487 5488
	if (!check_stripe_cache(mddev))
		return -ENOSPC;

5489
	list_for_each_entry(rdev, &mddev->disks, same_set)
5490 5491
		if (!test_bit(In_sync, &rdev->flags)
		    && !test_bit(Faulty, &rdev->flags))
5492
			spares++;
5493

5494
	if (spares - mddev->degraded < mddev->delta_disks - conf->max_degraded)
5495 5496 5497 5498 5499
		/* Not enough devices even to make a degraded array
		 * of that size
		 */
		return -EINVAL;

5500 5501 5502 5503 5504 5505
	/* 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) {
5506
		printk(KERN_ERR "md/raid:%s: array size must be reduced "
5507 5508 5509 5510
		       "before number of disks\n", mdname(mddev));
		return -EINVAL;
	}

5511
	atomic_set(&conf->reshape_stripes, 0);
5512 5513
	spin_lock_irq(&conf->device_lock);
	conf->previous_raid_disks = conf->raid_disks;
5514
	conf->raid_disks += mddev->delta_disks;
5515 5516
	conf->prev_chunk_sectors = conf->chunk_sectors;
	conf->chunk_sectors = mddev->new_chunk_sectors;
5517 5518
	conf->prev_algo = conf->algorithm;
	conf->algorithm = mddev->new_layout;
5519 5520 5521 5522 5523
	if (mddev->delta_disks < 0)
		conf->reshape_progress = raid5_size(mddev, 0, 0);
	else
		conf->reshape_progress = 0;
	conf->reshape_safe = conf->reshape_progress;
5524
	conf->generation++;
5525 5526 5527 5528
	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.
5529 5530 5531 5532
	 * 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.
5533
	 */
5534 5535 5536 5537 5538 5539 5540 5541 5542 5543 5544 5545
	if (mddev->delta_disks >= 0) {
		int added_devices = 0;
		list_for_each_entry(rdev, &mddev->disks, same_set)
			if (rdev->raid_disk < 0 &&
			    !test_bit(Faulty, &rdev->flags)) {
				if (raid5_add_disk(mddev, rdev) == 0) {
					if (rdev->raid_disk
					    >= conf->previous_raid_disks) {
						set_bit(In_sync, &rdev->flags);
						added_devices++;
					} else
						rdev->recovery_offset = 0;
5546 5547

					if (sysfs_link_rdev(mddev, rdev))
5548
						/* Failure here is OK */;
5549
				}
5550 5551 5552 5553 5554 5555
			} 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);
				added_devices++;
			}
5556

5557 5558 5559 5560
		/* When a reshape changes the number of devices,
		 * ->degraded is measured against the larger of the
		 * pre and post number of devices.
		 */
5561
		spin_lock_irqsave(&conf->device_lock, flags);
5562
		mddev->degraded = calc_degraded(conf);
5563 5564
		spin_unlock_irqrestore(&conf->device_lock, flags);
	}
5565
	mddev->raid_disks = conf->raid_disks;
5566
	mddev->reshape_position = conf->reshape_progress;
5567
	set_bit(MD_CHANGE_DEVS, &mddev->flags);
5568

5569 5570 5571 5572 5573
	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,
5574
						"reshape");
5575 5576 5577 5578
	if (!mddev->sync_thread) {
		mddev->recovery = 0;
		spin_lock_irq(&conf->device_lock);
		mddev->raid_disks = conf->raid_disks = conf->previous_raid_disks;
5579
		conf->reshape_progress = MaxSector;
5580 5581 5582
		spin_unlock_irq(&conf->device_lock);
		return -EAGAIN;
	}
5583
	conf->reshape_checkpoint = jiffies;
5584 5585 5586 5587 5588
	md_wakeup_thread(mddev->sync_thread);
	md_new_event(mddev);
	return 0;
}

5589 5590 5591
/* This is called from the reshape thread and should make any
 * changes needed in 'conf'
 */
5592
static void end_reshape(struct r5conf *conf)
5593 5594
{

5595 5596 5597
	if (!test_bit(MD_RECOVERY_INTR, &conf->mddev->recovery)) {

		spin_lock_irq(&conf->device_lock);
5598
		conf->previous_raid_disks = conf->raid_disks;
5599
		conf->reshape_progress = MaxSector;
5600
		spin_unlock_irq(&conf->device_lock);
5601
		wake_up(&conf->wait_for_overlap);
5602 5603 5604 5605

		/* read-ahead size must cover two whole stripes, which is
		 * 2 * (datadisks) * chunksize where 'n' is the number of raid devices
		 */
5606
		if (conf->mddev->queue) {
5607
			int data_disks = conf->raid_disks - conf->max_degraded;
5608
			int stripe = data_disks * ((conf->chunk_sectors << 9)
5609
						   / PAGE_SIZE);
5610 5611 5612
			if (conf->mddev->queue->backing_dev_info.ra_pages < 2 * stripe)
				conf->mddev->queue->backing_dev_info.ra_pages = 2 * stripe;
		}
5613 5614 5615
	}
}

5616 5617 5618
/* This is called from the raid5d thread with mddev_lock held.
 * It makes config changes to the device.
 */
5619
static void raid5_finish_reshape(struct mddev *mddev)
5620
{
5621
	struct r5conf *conf = mddev->private;
5622 5623 5624

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

5625 5626 5627
		if (mddev->delta_disks > 0) {
			md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
			set_capacity(mddev->gendisk, mddev->array_sectors);
5628
			revalidate_disk(mddev->gendisk);
5629 5630
		} else {
			int d;
5631 5632 5633
			spin_lock_irq(&conf->device_lock);
			mddev->degraded = calc_degraded(conf);
			spin_unlock_irq(&conf->device_lock);
5634 5635
			for (d = conf->raid_disks ;
			     d < conf->raid_disks - mddev->delta_disks;
5636
			     d++) {
5637
				struct md_rdev *rdev = conf->disks[d].rdev;
5638 5639
				if (rdev &&
				    raid5_remove_disk(mddev, rdev) == 0) {
5640
					sysfs_unlink_rdev(mddev, rdev);
5641 5642 5643
					rdev->raid_disk = -1;
				}
			}
5644
		}
5645
		mddev->layout = conf->algorithm;
5646
		mddev->chunk_sectors = conf->chunk_sectors;
5647 5648
		mddev->reshape_position = MaxSector;
		mddev->delta_disks = 0;
5649 5650 5651
	}
}

5652
static void raid5_quiesce(struct mddev *mddev, int state)
5653
{
5654
	struct r5conf *conf = mddev->private;
5655 5656

	switch(state) {
5657 5658 5659 5660
	case 2: /* resume for a suspend */
		wake_up(&conf->wait_for_overlap);
		break;

5661 5662
	case 1: /* stop all writes */
		spin_lock_irq(&conf->device_lock);
5663 5664 5665 5666
		/* '2' tells resync/reshape to pause so that all
		 * active stripes can drain
		 */
		conf->quiesce = 2;
5667
		wait_event_lock_irq(conf->wait_for_stripe,
5668 5669
				    atomic_read(&conf->active_stripes) == 0 &&
				    atomic_read(&conf->active_aligned_reads) == 0,
5670
				    conf->device_lock, /* nothing */);
5671
		conf->quiesce = 1;
5672
		spin_unlock_irq(&conf->device_lock);
5673 5674
		/* allow reshape to continue */
		wake_up(&conf->wait_for_overlap);
5675 5676 5677 5678 5679 5680
		break;

	case 0: /* re-enable writes */
		spin_lock_irq(&conf->device_lock);
		conf->quiesce = 0;
		wake_up(&conf->wait_for_stripe);
5681
		wake_up(&conf->wait_for_overlap);
5682 5683 5684 5685
		spin_unlock_irq(&conf->device_lock);
		break;
	}
}
5686

5687

5688
static void *raid45_takeover_raid0(struct mddev *mddev, int level)
5689
{
5690
	struct r0conf *raid0_conf = mddev->private;
5691
	sector_t sectors;
5692

D
Dan Williams 已提交
5693
	/* for raid0 takeover only one zone is supported */
5694
	if (raid0_conf->nr_strip_zones > 1) {
5695 5696
		printk(KERN_ERR "md/raid:%s: cannot takeover raid0 with more than one zone.\n",
		       mdname(mddev));
D
Dan Williams 已提交
5697 5698 5699
		return ERR_PTR(-EINVAL);
	}

5700 5701
	sectors = raid0_conf->strip_zone[0].zone_end;
	sector_div(sectors, raid0_conf->strip_zone[0].nb_dev);
5702
	mddev->dev_sectors = sectors;
D
Dan Williams 已提交
5703
	mddev->new_level = level;
5704 5705 5706 5707 5708 5709 5710 5711 5712 5713 5714
	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);
}


5715
static void *raid5_takeover_raid1(struct mddev *mddev)
5716 5717 5718 5719 5720 5721 5722 5723 5724 5725 5726 5727 5728 5729 5730 5731 5732 5733 5734 5735 5736
{
	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;
5737
	mddev->new_chunk_sectors = chunksect;
5738 5739 5740 5741

	return setup_conf(mddev);
}

5742
static void *raid5_takeover_raid6(struct mddev *mddev)
5743 5744 5745 5746 5747 5748 5749 5750 5751 5752 5753 5754 5755 5756 5757 5758 5759 5760 5761 5762 5763 5764 5765 5766 5767 5768 5769 5770 5771 5772 5773 5774
{
	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);
}

5775

5776
static int raid5_check_reshape(struct mddev *mddev)
5777
{
5778 5779 5780 5781
	/* 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.
5782
	 */
5783
	struct r5conf *conf = mddev->private;
5784
	int new_chunk = mddev->new_chunk_sectors;
5785

5786
	if (mddev->new_layout >= 0 && !algorithm_valid_raid5(mddev->new_layout))
5787 5788
		return -EINVAL;
	if (new_chunk > 0) {
5789
		if (!is_power_of_2(new_chunk))
5790
			return -EINVAL;
5791
		if (new_chunk < (PAGE_SIZE>>9))
5792
			return -EINVAL;
5793
		if (mddev->array_sectors & (new_chunk-1))
5794 5795 5796 5797 5798 5799
			/* not factor of array size */
			return -EINVAL;
	}

	/* They look valid */

5800
	if (mddev->raid_disks == 2) {
5801 5802 5803 5804
		/* can make the change immediately */
		if (mddev->new_layout >= 0) {
			conf->algorithm = mddev->new_layout;
			mddev->layout = mddev->new_layout;
5805 5806
		}
		if (new_chunk > 0) {
5807 5808
			conf->chunk_sectors = new_chunk ;
			mddev->chunk_sectors = new_chunk;
5809 5810 5811
		}
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
		md_wakeup_thread(mddev->thread);
5812
	}
5813
	return check_reshape(mddev);
5814 5815
}

5816
static int raid6_check_reshape(struct mddev *mddev)
5817
{
5818
	int new_chunk = mddev->new_chunk_sectors;
5819

5820
	if (mddev->new_layout >= 0 && !algorithm_valid_raid6(mddev->new_layout))
5821
		return -EINVAL;
5822
	if (new_chunk > 0) {
5823
		if (!is_power_of_2(new_chunk))
5824
			return -EINVAL;
5825
		if (new_chunk < (PAGE_SIZE >> 9))
5826
			return -EINVAL;
5827
		if (mddev->array_sectors & (new_chunk-1))
5828 5829
			/* not factor of array size */
			return -EINVAL;
5830
	}
5831 5832

	/* They look valid */
5833
	return check_reshape(mddev);
5834 5835
}

5836
static void *raid5_takeover(struct mddev *mddev)
5837 5838
{
	/* raid5 can take over:
D
Dan Williams 已提交
5839
	 *  raid0 - if there is only one strip zone - make it a raid4 layout
5840 5841 5842 5843
	 *  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
5844 5845
	if (mddev->level == 0)
		return raid45_takeover_raid0(mddev, 5);
5846 5847
	if (mddev->level == 1)
		return raid5_takeover_raid1(mddev);
5848 5849 5850 5851 5852
	if (mddev->level == 4) {
		mddev->new_layout = ALGORITHM_PARITY_N;
		mddev->new_level = 5;
		return setup_conf(mddev);
	}
5853 5854
	if (mddev->level == 6)
		return raid5_takeover_raid6(mddev);
5855 5856 5857 5858

	return ERR_PTR(-EINVAL);
}

5859
static void *raid4_takeover(struct mddev *mddev)
5860
{
D
Dan Williams 已提交
5861 5862 5863
	/* raid4 can take over:
	 *  raid0 - if there is only one strip zone
	 *  raid5 - if layout is right
5864
	 */
D
Dan Williams 已提交
5865 5866
	if (mddev->level == 0)
		return raid45_takeover_raid0(mddev, 4);
5867 5868 5869 5870 5871 5872 5873 5874
	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);
}
5875

5876
static struct md_personality raid5_personality;
5877

5878
static void *raid6_takeover(struct mddev *mddev)
5879 5880 5881 5882 5883 5884 5885 5886 5887 5888 5889 5890 5891 5892 5893 5894 5895 5896 5897 5898 5899 5900 5901 5902 5903 5904 5905 5906 5907 5908 5909 5910 5911 5912 5913 5914 5915 5916 5917 5918 5919 5920 5921 5922 5923 5924
{
	/* 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);
}


5925
static struct md_personality raid6_personality =
5926 5927 5928 5929 5930 5931 5932 5933 5934 5935 5936 5937 5938 5939
{
	.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,
5940
	.size		= raid5_size,
5941
	.check_reshape	= raid6_check_reshape,
5942
	.start_reshape  = raid5_start_reshape,
5943
	.finish_reshape = raid5_finish_reshape,
5944
	.quiesce	= raid5_quiesce,
5945
	.takeover	= raid6_takeover,
5946
};
5947
static struct md_personality raid5_personality =
L
Linus Torvalds 已提交
5948 5949
{
	.name		= "raid5",
5950
	.level		= 5,
L
Linus Torvalds 已提交
5951 5952 5953 5954 5955 5956 5957 5958 5959 5960 5961
	.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,
5962
	.size		= raid5_size,
5963 5964
	.check_reshape	= raid5_check_reshape,
	.start_reshape  = raid5_start_reshape,
5965
	.finish_reshape = raid5_finish_reshape,
5966
	.quiesce	= raid5_quiesce,
5967
	.takeover	= raid5_takeover,
L
Linus Torvalds 已提交
5968 5969
};

5970
static struct md_personality raid4_personality =
L
Linus Torvalds 已提交
5971
{
5972 5973 5974 5975 5976 5977 5978 5979 5980 5981 5982 5983 5984
	.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,
5985
	.size		= raid5_size,
5986 5987
	.check_reshape	= raid5_check_reshape,
	.start_reshape  = raid5_start_reshape,
5988
	.finish_reshape = raid5_finish_reshape,
5989
	.quiesce	= raid5_quiesce,
5990
	.takeover	= raid4_takeover,
5991 5992 5993 5994
};

static int __init raid5_init(void)
{
5995
	register_md_personality(&raid6_personality);
5996 5997 5998
	register_md_personality(&raid5_personality);
	register_md_personality(&raid4_personality);
	return 0;
L
Linus Torvalds 已提交
5999 6000
}

6001
static void raid5_exit(void)
L
Linus Torvalds 已提交
6002
{
6003
	unregister_md_personality(&raid6_personality);
6004 6005
	unregister_md_personality(&raid5_personality);
	unregister_md_personality(&raid4_personality);
L
Linus Torvalds 已提交
6006 6007 6008 6009 6010
}

module_init(raid5_init);
module_exit(raid5_exit);
MODULE_LICENSE("GPL");
6011
MODULE_DESCRIPTION("RAID4/5/6 (striping with parity) personality for MD");
L
Linus Torvalds 已提交
6012
MODULE_ALIAS("md-personality-4"); /* RAID5 */
6013 6014
MODULE_ALIAS("md-raid5");
MODULE_ALIAS("md-raid4");
6015 6016
MODULE_ALIAS("md-level-5");
MODULE_ALIAS("md-level-4");
6017 6018 6019 6020 6021 6022 6023
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");