raid5.c 171.3 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));
211 212 213
			if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
				if (atomic_dec_return(&conf->preread_active_stripes)
				    < IO_THRESHOLD)
L
Linus Torvalds 已提交
214 215
					md_wakeup_thread(conf->mddev->thread);
			atomic_dec(&conf->active_stripes);
216 217
			if (!test_bit(STRIPE_EXPANDING, &sh->state)) {
				list_add_tail(&sh->lru, &conf->inactive_list);
L
Linus Torvalds 已提交
218
				wake_up(&conf->wait_for_stripe);
219 220
				if (conf->retry_read_aligned)
					md_wakeup_thread(conf->mddev->thread);
221
			}
L
Linus Torvalds 已提交
222 223 224
		}
	}
}
225

L
Linus Torvalds 已提交
226 227
static void release_stripe(struct stripe_head *sh)
{
228
	struct r5conf *conf = sh->raid_conf;
L
Linus Torvalds 已提交
229
	unsigned long flags;
230

L
Linus Torvalds 已提交
231 232 233 234 235
	spin_lock_irqsave(&conf->device_lock, flags);
	__release_stripe(conf, sh);
	spin_unlock_irqrestore(&conf->device_lock, flags);
}

236
static inline void remove_hash(struct stripe_head *sh)
L
Linus Torvalds 已提交
237
{
238 239
	pr_debug("remove_hash(), stripe %llu\n",
		(unsigned long long)sh->sector);
L
Linus Torvalds 已提交
240

241
	hlist_del_init(&sh->hash);
L
Linus Torvalds 已提交
242 243
}

244
static inline void insert_hash(struct r5conf *conf, struct stripe_head *sh)
L
Linus Torvalds 已提交
245
{
246
	struct hlist_head *hp = stripe_hash(conf, sh->sector);
L
Linus Torvalds 已提交
247

248 249
	pr_debug("insert_hash(), stripe %llu\n",
		(unsigned long long)sh->sector);
L
Linus Torvalds 已提交
250

251
	hlist_add_head(&sh->hash, hp);
L
Linus Torvalds 已提交
252 253 254 255
}


/* find an idle stripe, make sure it is unhashed, and return it. */
256
static struct stripe_head *get_free_stripe(struct r5conf *conf)
L
Linus Torvalds 已提交
257 258 259 260 261 262 263 264 265 266 267 268 269 270 271
{
	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;
}

272
static void shrink_buffers(struct stripe_head *sh)
L
Linus Torvalds 已提交
273 274 275
{
	struct page *p;
	int i;
276
	int num = sh->raid_conf->pool_size;
L
Linus Torvalds 已提交
277

278
	for (i = 0; i < num ; i++) {
L
Linus Torvalds 已提交
279 280 281 282
		p = sh->dev[i].page;
		if (!p)
			continue;
		sh->dev[i].page = NULL;
283
		put_page(p);
L
Linus Torvalds 已提交
284 285 286
	}
}

287
static int grow_buffers(struct stripe_head *sh)
L
Linus Torvalds 已提交
288 289
{
	int i;
290
	int num = sh->raid_conf->pool_size;
L
Linus Torvalds 已提交
291

292
	for (i = 0; i < num; i++) {
L
Linus Torvalds 已提交
293 294 295 296 297 298 299 300 301 302
		struct page *page;

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

303
static void raid5_build_block(struct stripe_head *sh, int i, int previous);
304
static void stripe_set_idx(sector_t stripe, struct r5conf *conf, int previous,
305
			    struct stripe_head *sh);
L
Linus Torvalds 已提交
306

307
static void init_stripe(struct stripe_head *sh, sector_t sector, int previous)
L
Linus Torvalds 已提交
308
{
309
	struct r5conf *conf = sh->raid_conf;
310
	int i;
L
Linus Torvalds 已提交
311

312 313
	BUG_ON(atomic_read(&sh->count) != 0);
	BUG_ON(test_bit(STRIPE_HANDLE, &sh->state));
314
	BUG_ON(stripe_operations_active(sh));
315

316
	pr_debug("init_stripe called, stripe %llu\n",
L
Linus Torvalds 已提交
317 318 319
		(unsigned long long)sh->sector);

	remove_hash(sh);
320

321
	sh->generation = conf->generation - previous;
322
	sh->disks = previous ? conf->previous_raid_disks : conf->raid_disks;
L
Linus Torvalds 已提交
323
	sh->sector = sector;
324
	stripe_set_idx(sector, conf, previous, sh);
L
Linus Torvalds 已提交
325 326
	sh->state = 0;

327 328

	for (i = sh->disks; i--; ) {
L
Linus Torvalds 已提交
329 330
		struct r5dev *dev = &sh->dev[i];

331
		if (dev->toread || dev->read || dev->towrite || dev->written ||
L
Linus Torvalds 已提交
332
		    test_bit(R5_LOCKED, &dev->flags)) {
333
			printk(KERN_ERR "sector=%llx i=%d %p %p %p %p %d\n",
L
Linus Torvalds 已提交
334
			       (unsigned long long)sh->sector, i, dev->toread,
335
			       dev->read, dev->towrite, dev->written,
L
Linus Torvalds 已提交
336
			       test_bit(R5_LOCKED, &dev->flags));
337
			WARN_ON(1);
L
Linus Torvalds 已提交
338 339
		}
		dev->flags = 0;
340
		raid5_build_block(sh, i, previous);
L
Linus Torvalds 已提交
341 342 343 344
	}
	insert_hash(conf, sh);
}

345
static struct stripe_head *__find_stripe(struct r5conf *conf, sector_t sector,
346
					 short generation)
L
Linus Torvalds 已提交
347 348
{
	struct stripe_head *sh;
349
	struct hlist_node *hn;
L
Linus Torvalds 已提交
350

351
	pr_debug("__find_stripe, sector %llu\n", (unsigned long long)sector);
352
	hlist_for_each_entry(sh, hn, stripe_hash(conf, sector), hash)
353
		if (sh->sector == sector && sh->generation == generation)
L
Linus Torvalds 已提交
354
			return sh;
355
	pr_debug("__stripe %llu not in cache\n", (unsigned long long)sector);
L
Linus Torvalds 已提交
356 357 358
	return NULL;
}

359 360 361 362 363 364 365 366 367 368 369 370 371
/*
 * 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.
 */
372
static int calc_degraded(struct r5conf *conf)
373
{
374
	int degraded, degraded2;
375 376 377 378 379
	int i;

	rcu_read_lock();
	degraded = 0;
	for (i = 0; i < conf->previous_raid_disks; i++) {
380
		struct md_rdev *rdev = rcu_dereference(conf->disks[i].rdev);
381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398
		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();
399 400
	if (conf->raid_disks == conf->previous_raid_disks)
		return degraded;
401
	rcu_read_lock();
402
	degraded2 = 0;
403
	for (i = 0; i < conf->raid_disks; i++) {
404
		struct md_rdev *rdev = rcu_dereference(conf->disks[i].rdev);
405
		if (!rdev || test_bit(Faulty, &rdev->flags))
406
			degraded2++;
407 408 409 410 411 412 413 414 415
		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)
416
				degraded2++;
417 418
	}
	rcu_read_unlock();
419 420 421 422 423 424 425 426 427 428 429 430 431
	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);
432 433 434 435 436
	if (degraded > conf->max_degraded)
		return 1;
	return 0;
}

437
static struct stripe_head *
438
get_active_stripe(struct r5conf *conf, sector_t sector,
439
		  int previous, int noblock, int noquiesce)
L
Linus Torvalds 已提交
440 441 442
{
	struct stripe_head *sh;

443
	pr_debug("get_stripe, sector %llu\n", (unsigned long long)sector);
L
Linus Torvalds 已提交
444 445 446 447

	spin_lock_irq(&conf->device_lock);

	do {
448
		wait_event_lock_irq(conf->wait_for_stripe,
449
				    conf->quiesce == 0 || noquiesce,
450
				    conf->device_lock, /* nothing */);
451
		sh = __find_stripe(conf, sector, conf->generation - previous);
L
Linus Torvalds 已提交
452 453 454 455 456 457 458 459 460
		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) &&
461 462
						    (atomic_read(&conf->active_stripes)
						     < (conf->max_nr_stripes *3/4)
L
Linus Torvalds 已提交
463 464
						     || !conf->inactive_blocked),
						    conf->device_lock,
465
						    );
L
Linus Torvalds 已提交
466 467
				conf->inactive_blocked = 0;
			} else
468
				init_stripe(sh, sector, previous);
L
Linus Torvalds 已提交
469 470
		} else {
			if (atomic_read(&sh->count)) {
471 472
				BUG_ON(!list_empty(&sh->lru)
				    && !test_bit(STRIPE_EXPANDING, &sh->state));
L
Linus Torvalds 已提交
473 474 475
			} else {
				if (!test_bit(STRIPE_HANDLE, &sh->state))
					atomic_inc(&conf->active_stripes);
476 477
				if (list_empty(&sh->lru) &&
				    !test_bit(STRIPE_EXPANDING, &sh->state))
478 479
					BUG();
				list_del_init(&sh->lru);
L
Linus Torvalds 已提交
480 481 482 483 484 485 486 487 488 489 490
			}
		}
	} while (sh == NULL);

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

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

491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511
/* Determine if 'data_offset' or 'new_data_offset' should be used
 * in this stripe_head.
 */
static int use_new_offset(struct r5conf *conf, struct stripe_head *sh)
{
	sector_t progress = conf->reshape_progress;
	/* Need a memory barrier to make sure we see the value
	 * of conf->generation, or ->data_offset that was set before
	 * reshape_progress was updated.
	 */
	smp_rmb();
	if (progress == MaxSector)
		return 0;
	if (sh->generation == conf->generation - 1)
		return 0;
	/* We are in a reshape, and this is a new-generation stripe,
	 * so use new_data_offset.
	 */
	return 1;
}

512 513 514 515
static void
raid5_end_read_request(struct bio *bi, int error);
static void
raid5_end_write_request(struct bio *bi, int error);
516

517
static void ops_run_io(struct stripe_head *sh, struct stripe_head_state *s)
518
{
519
	struct r5conf *conf = sh->raid_conf;
520 521 522 523 524 525
	int i, disks = sh->disks;

	might_sleep();

	for (i = disks; i--; ) {
		int rw;
526
		int replace_only = 0;
527 528
		struct bio *bi, *rbi;
		struct md_rdev *rdev, *rrdev = NULL;
T
Tejun Heo 已提交
529 530 531 532 533 534
		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))
535
			rw = READ;
536 537 538 539 540
		else if (test_and_clear_bit(R5_WantReplace,
					    &sh->dev[i].flags)) {
			rw = WRITE;
			replace_only = 1;
		} else
541
			continue;
S
Shaohua Li 已提交
542 543
		if (test_and_clear_bit(R5_SyncIO, &sh->dev[i].flags))
			rw |= REQ_SYNC;
544 545

		bi = &sh->dev[i].req;
546
		rbi = &sh->dev[i].rreq; /* For writing to replacement */
547 548

		bi->bi_rw = rw;
549 550
		rbi->bi_rw = rw;
		if (rw & WRITE) {
551
			bi->bi_end_io = raid5_end_write_request;
552 553
			rbi->bi_end_io = raid5_end_write_request;
		} else
554 555 556
			bi->bi_end_io = raid5_end_read_request;

		rcu_read_lock();
557
		rrdev = rcu_dereference(conf->disks[i].replacement);
558 559 560 561 562 563
		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;
		}
564 565 566
		if (rw & WRITE) {
			if (replace_only)
				rdev = NULL;
567 568 569
			if (rdev == rrdev)
				/* We raced and saw duplicates */
				rrdev = NULL;
570
		} else {
571
			if (test_bit(R5_ReadRepl, &sh->dev[i].flags) && rrdev)
572 573 574
				rdev = rrdev;
			rrdev = NULL;
		}
575

576 577 578 579
		if (rdev && test_bit(Faulty, &rdev->flags))
			rdev = NULL;
		if (rdev)
			atomic_inc(&rdev->nr_pending);
580 581 582 583
		if (rrdev && test_bit(Faulty, &rrdev->flags))
			rrdev = NULL;
		if (rrdev)
			atomic_inc(&rrdev->nr_pending);
584 585
		rcu_read_unlock();

586
		/* We have already checked bad blocks for reads.  Now
587 588
		 * need to check for writes.  We never accept write errors
		 * on the replacement, so we don't to check rrdev.
589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608
		 */
		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);
				}
609 610 611 612 613 614
				/*
				 * Because md_wait_for_blocked_rdev
				 * will dec nr_pending, we must
				 * increment it first.
				 */
				atomic_inc(&rdev->nr_pending);
615 616 617 618 619 620 621 622
				md_wait_for_blocked_rdev(rdev, conf->mddev);
			} else {
				/* Acknowledged bad block - skip the write */
				rdev_dec_pending(rdev, conf->mddev);
				rdev = NULL;
			}
		}

623
		if (rdev) {
624 625
			if (s->syncing || s->expanding || s->expanded
			    || s->replacing)
626 627
				md_sync_acct(rdev->bdev, STRIPE_SECTORS);

D
Dan Williams 已提交
628 629
			set_bit(STRIPE_IO_STARTED, &sh->state);

630 631
			bi->bi_bdev = rdev->bdev;
			pr_debug("%s: for %llu schedule op %ld on disc %d\n",
632
				__func__, (unsigned long long)sh->sector,
633 634
				bi->bi_rw, i);
			atomic_inc(&sh->count);
635 636 637 638 639 640
			if (use_new_offset(conf, sh))
				bi->bi_sector = (sh->sector
						 + rdev->new_data_offset);
			else
				bi->bi_sector = (sh->sector
						 + rdev->data_offset);
641 642 643 644 645 646
			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;
647 648
			if (rrdev)
				set_bit(R5_DOUBLE_LOCKED, &sh->dev[i].flags);
649
			generic_make_request(bi);
650 651
		}
		if (rrdev) {
652 653
			if (s->syncing || s->expanding || s->expanded
			    || s->replacing)
654 655 656 657 658 659 660 661 662 663
				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);
664 665 666 667 668 669
			if (use_new_offset(conf, sh))
				rbi->bi_sector = (sh->sector
						  + rrdev->new_data_offset);
			else
				rbi->bi_sector = (sh->sector
						  + rrdev->data_offset);
670 671 672 673 674 675 676 677 678
			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) {
679
			if (rw & WRITE)
680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696
				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;
697
	struct async_submit_ctl submit;
D
Dan Williams 已提交
698
	enum async_tx_flags flags = 0;
699 700 701 702 703

	if (bio->bi_sector >= sector)
		page_offset = (signed)(bio->bi_sector - sector) * 512;
	else
		page_offset = (signed)(sector - bio->bi_sector) * -512;
704

D
Dan Williams 已提交
705 706 707 708
	if (frombio)
		flags |= ASYNC_TX_FENCE;
	init_async_submit(&submit, flags, tx, NULL, NULL, NULL);

709
	bio_for_each_segment(bvl, bio, i) {
710
		int len = bvl->bv_len;
711 712 713 714 715 716 717 718 719 720 721 722 723 724 725
		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) {
726 727
			b_offset += bvl->bv_offset;
			bio_page = bvl->bv_page;
728 729
			if (frombio)
				tx = async_memcpy(page, bio_page, page_offset,
730
						  b_offset, clen, &submit);
731 732
			else
				tx = async_memcpy(bio_page, page, b_offset,
733
						  page_offset, clen, &submit);
734
		}
735 736 737
		/* chain the operations */
		submit.depend_tx = tx;

738 739 740 741 742 743 744 745 746 747 748 749
		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;
750
	struct r5conf *conf = sh->raid_conf;
751
	int i;
752

753
	pr_debug("%s: stripe %llu\n", __func__,
754 755 756
		(unsigned long long)sh->sector);

	/* clear completed biofills */
757
	spin_lock_irq(&conf->device_lock);
758 759 760 761
	for (i = sh->disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];

		/* acknowledge completion of a biofill operation */
762 763
		/* and check if we need to reply to a read request,
		 * new R5_Wantfill requests are held off until
764
		 * !STRIPE_BIOFILL_RUN
765 766
		 */
		if (test_and_clear_bit(R5_Wantfill, &dev->flags)) {
767 768 769 770 771 772 773 774
			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);
775
				if (!raid5_dec_bi_phys_segments(rbi)) {
776 777 778 779 780 781 782
					rbi->bi_next = return_bi;
					return_bi = rbi;
				}
				rbi = rbi2;
			}
		}
	}
783 784
	spin_unlock_irq(&conf->device_lock);
	clear_bit(STRIPE_BIOFILL_RUN, &sh->state);
785 786 787

	return_io(return_bi);

788
	set_bit(STRIPE_HANDLE, &sh->state);
789 790 791 792 793 794
	release_stripe(sh);
}

static void ops_run_biofill(struct stripe_head *sh)
{
	struct dma_async_tx_descriptor *tx = NULL;
795
	struct r5conf *conf = sh->raid_conf;
796
	struct async_submit_ctl submit;
797 798
	int i;

799
	pr_debug("%s: stripe %llu\n", __func__,
800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819
		(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);
820 821
	init_async_submit(&submit, ASYNC_TX_ACK, tx, ops_complete_biofill, sh, NULL);
	async_trigger_callback(&submit);
822 823
}

824
static void mark_target_uptodate(struct stripe_head *sh, int target)
825
{
826
	struct r5dev *tgt;
827

828 829
	if (target < 0)
		return;
830

831
	tgt = &sh->dev[target];
832 833 834
	set_bit(R5_UPTODATE, &tgt->flags);
	BUG_ON(!test_bit(R5_Wantcompute, &tgt->flags));
	clear_bit(R5_Wantcompute, &tgt->flags);
835 836
}

837
static void ops_complete_compute(void *stripe_head_ref)
838 839 840
{
	struct stripe_head *sh = stripe_head_ref;

841
	pr_debug("%s: stripe %llu\n", __func__,
842 843
		(unsigned long long)sh->sector);

844
	/* mark the computed target(s) as uptodate */
845
	mark_target_uptodate(sh, sh->ops.target);
846
	mark_target_uptodate(sh, sh->ops.target2);
847

848 849 850
	clear_bit(STRIPE_COMPUTE_RUN, &sh->state);
	if (sh->check_state == check_state_compute_run)
		sh->check_state = check_state_compute_result;
851 852 853 854
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

855 856 857 858 859 860 861 862 863
/* 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)
864 865
{
	int disks = sh->disks;
866
	struct page **xor_srcs = percpu->scribble;
867 868 869 870 871
	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;
872
	struct async_submit_ctl submit;
873 874 875
	int i;

	pr_debug("%s: stripe %llu block: %d\n",
876
		__func__, (unsigned long long)sh->sector, target);
877 878 879 880 881 882 883 884
	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
885
	init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST, NULL,
886
			  ops_complete_compute, sh, to_addr_conv(sh, percpu));
887
	if (unlikely(count == 1))
888
		tx = async_memcpy(xor_dest, xor_srcs[0], 0, 0, STRIPE_SIZE, &submit);
889
	else
890
		tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, &submit);
891 892 893 894

	return tx;
}

895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912
/* 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++)
913
		srcs[i] = NULL;
914 915 916 917 918 919 920 921 922 923

	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);

924
	return syndrome_disks;
925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944
}

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;
945
	else
946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961
		/* 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
962 963
		init_async_submit(&submit, ASYNC_TX_FENCE, NULL,
				  ops_complete_compute, sh,
964 965 966 967 968 969 970 971 972 973 974
				  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
975 976
		init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST,
				  NULL, ops_complete_compute, sh,
977 978 979
				  to_addr_conv(sh, percpu));
		tx = async_xor(dest, blocks, 0, count, STRIPE_SIZE, &submit);
	}
980 981 982 983

	return tx;
}

984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004
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));

1005
	/* we need to open-code set_syndrome_sources to handle the
1006 1007 1008
	 * slot number conversion for 'faila' and 'failb'
	 */
	for (i = 0; i < disks ; i++)
1009
		blocks[i] = NULL;
1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035
	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
1036 1037 1038
			init_async_submit(&submit, ASYNC_TX_FENCE, NULL,
					  ops_complete_compute, sh,
					  to_addr_conv(sh, percpu));
1039
			return async_gen_syndrome(blocks, 0, syndrome_disks+2,
1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058
						  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
1059 1060 1061 1062
			init_async_submit(&submit,
					  ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST,
					  NULL, NULL, NULL,
					  to_addr_conv(sh, percpu));
1063 1064 1065 1066
			tx = async_xor(dest, blocks, 0, count, STRIPE_SIZE,
				       &submit);

			count = set_syndrome_sources(blocks, sh);
D
Dan Williams 已提交
1067 1068 1069
			init_async_submit(&submit, ASYNC_TX_FENCE, tx,
					  ops_complete_compute, sh,
					  to_addr_conv(sh, percpu));
1070 1071 1072 1073
			return async_gen_syndrome(blocks, 0, count+2,
						  STRIPE_SIZE, &submit);
		}
	} else {
1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087
		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);
		}
1088 1089 1090 1091
	}
}


1092 1093 1094 1095
static void ops_complete_prexor(void *stripe_head_ref)
{
	struct stripe_head *sh = stripe_head_ref;

1096
	pr_debug("%s: stripe %llu\n", __func__,
1097 1098 1099 1100
		(unsigned long long)sh->sector);
}

static struct dma_async_tx_descriptor *
1101 1102
ops_run_prexor(struct stripe_head *sh, struct raid5_percpu *percpu,
	       struct dma_async_tx_descriptor *tx)
1103 1104
{
	int disks = sh->disks;
1105
	struct page **xor_srcs = percpu->scribble;
1106
	int count = 0, pd_idx = sh->pd_idx, i;
1107
	struct async_submit_ctl submit;
1108 1109 1110 1111

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

1112
	pr_debug("%s: stripe %llu\n", __func__,
1113 1114 1115 1116 1117
		(unsigned long long)sh->sector);

	for (i = disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
		/* Only process blocks that are known to be uptodate */
1118
		if (test_bit(R5_Wantdrain, &dev->flags))
1119 1120 1121
			xor_srcs[count++] = dev->page;
	}

D
Dan Williams 已提交
1122
	init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_DROP_DST, tx,
1123
			  ops_complete_prexor, sh, to_addr_conv(sh, percpu));
1124
	tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, &submit);
1125 1126 1127 1128 1129

	return tx;
}

static struct dma_async_tx_descriptor *
1130
ops_run_biodrain(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
1131 1132
{
	int disks = sh->disks;
1133
	int i;
1134

1135
	pr_debug("%s: stripe %llu\n", __func__,
1136 1137 1138 1139 1140 1141
		(unsigned long long)sh->sector);

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

1142
		if (test_and_clear_bit(R5_Wantdrain, &dev->flags)) {
1143 1144
			struct bio *wbi;

1145
			spin_lock_irq(&sh->raid_conf->device_lock);
1146 1147 1148 1149
			chosen = dev->towrite;
			dev->towrite = NULL;
			BUG_ON(dev->written);
			wbi = dev->written = chosen;
1150
			spin_unlock_irq(&sh->raid_conf->device_lock);
1151 1152 1153

			while (wbi && wbi->bi_sector <
				dev->sector + STRIPE_SECTORS) {
T
Tejun Heo 已提交
1154 1155
				if (wbi->bi_rw & REQ_FUA)
					set_bit(R5_WantFUA, &dev->flags);
S
Shaohua Li 已提交
1156 1157
				if (wbi->bi_rw & REQ_SYNC)
					set_bit(R5_SyncIO, &dev->flags);
1158 1159 1160 1161 1162 1163 1164 1165 1166 1167
				tx = async_copy_data(1, wbi, dev->page,
					dev->sector, tx);
				wbi = r5_next_bio(wbi, dev->sector);
			}
		}
	}

	return tx;
}

1168
static void ops_complete_reconstruct(void *stripe_head_ref)
1169 1170
{
	struct stripe_head *sh = stripe_head_ref;
1171 1172 1173 1174
	int disks = sh->disks;
	int pd_idx = sh->pd_idx;
	int qd_idx = sh->qd_idx;
	int i;
S
Shaohua Li 已提交
1175
	bool fua = false, sync = false;
1176

1177
	pr_debug("%s: stripe %llu\n", __func__,
1178 1179
		(unsigned long long)sh->sector);

S
Shaohua Li 已提交
1180
	for (i = disks; i--; ) {
T
Tejun Heo 已提交
1181
		fua |= test_bit(R5_WantFUA, &sh->dev[i].flags);
S
Shaohua Li 已提交
1182 1183
		sync |= test_bit(R5_SyncIO, &sh->dev[i].flags);
	}
T
Tejun Heo 已提交
1184

1185 1186
	for (i = disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
1187

T
Tejun Heo 已提交
1188
		if (dev->written || i == pd_idx || i == qd_idx) {
1189
			set_bit(R5_UPTODATE, &dev->flags);
T
Tejun Heo 已提交
1190 1191
			if (fua)
				set_bit(R5_WantFUA, &dev->flags);
S
Shaohua Li 已提交
1192 1193
			if (sync)
				set_bit(R5_SyncIO, &dev->flags);
T
Tejun Heo 已提交
1194
		}
1195 1196
	}

1197 1198 1199 1200 1201 1202 1203 1204
	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;
	}
1205 1206 1207 1208 1209 1210

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

static void
1211 1212
ops_run_reconstruct5(struct stripe_head *sh, struct raid5_percpu *percpu,
		     struct dma_async_tx_descriptor *tx)
1213 1214
{
	int disks = sh->disks;
1215
	struct page **xor_srcs = percpu->scribble;
1216
	struct async_submit_ctl submit;
1217 1218
	int count = 0, pd_idx = sh->pd_idx, i;
	struct page *xor_dest;
1219
	int prexor = 0;
1220 1221
	unsigned long flags;

1222
	pr_debug("%s: stripe %llu\n", __func__,
1223 1224 1225 1226 1227
		(unsigned long long)sh->sector);

	/* check if prexor is active which means only process blocks
	 * that are part of a read-modify-write (written)
	 */
1228 1229
	if (sh->reconstruct_state == reconstruct_state_prexor_drain_run) {
		prexor = 1;
1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249
		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
	 */
1250
	flags = ASYNC_TX_ACK |
1251 1252 1253 1254
		(prexor ? ASYNC_TX_XOR_DROP_DST : ASYNC_TX_XOR_ZERO_DST);

	atomic_inc(&sh->count);

1255
	init_async_submit(&submit, flags, tx, ops_complete_reconstruct, sh,
1256
			  to_addr_conv(sh, percpu));
1257 1258 1259 1260
	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);
1261 1262
}

1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279
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);
1280 1281 1282 1283 1284 1285
}

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

1286
	pr_debug("%s: stripe %llu\n", __func__,
1287 1288
		(unsigned long long)sh->sector);

1289
	sh->check_state = check_state_check_result;
1290 1291 1292 1293
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

1294
static void ops_run_check_p(struct stripe_head *sh, struct raid5_percpu *percpu)
1295 1296
{
	int disks = sh->disks;
1297 1298 1299
	int pd_idx = sh->pd_idx;
	int qd_idx = sh->qd_idx;
	struct page *xor_dest;
1300
	struct page **xor_srcs = percpu->scribble;
1301
	struct dma_async_tx_descriptor *tx;
1302
	struct async_submit_ctl submit;
1303 1304
	int count;
	int i;
1305

1306
	pr_debug("%s: stripe %llu\n", __func__,
1307 1308
		(unsigned long long)sh->sector);

1309 1310 1311
	count = 0;
	xor_dest = sh->dev[pd_idx].page;
	xor_srcs[count++] = xor_dest;
1312
	for (i = disks; i--; ) {
1313 1314 1315
		if (i == pd_idx || i == qd_idx)
			continue;
		xor_srcs[count++] = sh->dev[i].page;
1316 1317
	}

1318 1319
	init_async_submit(&submit, 0, NULL, NULL, NULL,
			  to_addr_conv(sh, percpu));
D
Dan Williams 已提交
1320
	tx = async_xor_val(xor_dest, xor_srcs, 0, count, STRIPE_SIZE,
1321
			   &sh->ops.zero_sum_result, &submit);
1322 1323

	atomic_inc(&sh->count);
1324 1325
	init_async_submit(&submit, ASYNC_TX_ACK, tx, ops_complete_check, sh, NULL);
	tx = async_trigger_callback(&submit);
1326 1327
}

1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339
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;
1340 1341

	atomic_inc(&sh->count);
1342 1343 1344 1345
	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);
1346 1347
}

1348
static void __raid_run_ops(struct stripe_head *sh, unsigned long ops_request)
1349 1350 1351
{
	int overlap_clear = 0, i, disks = sh->disks;
	struct dma_async_tx_descriptor *tx = NULL;
1352
	struct r5conf *conf = sh->raid_conf;
1353
	int level = conf->level;
1354 1355
	struct raid5_percpu *percpu;
	unsigned long cpu;
1356

1357 1358
	cpu = get_cpu();
	percpu = per_cpu_ptr(conf->percpu, cpu);
1359
	if (test_bit(STRIPE_OP_BIOFILL, &ops_request)) {
1360 1361 1362 1363
		ops_run_biofill(sh);
		overlap_clear++;
	}

1364
	if (test_bit(STRIPE_OP_COMPUTE_BLK, &ops_request)) {
1365 1366 1367 1368 1369 1370 1371 1372 1373 1374
		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))
1375 1376
			async_tx_ack(tx);
	}
1377

1378
	if (test_bit(STRIPE_OP_PREXOR, &ops_request))
1379
		tx = ops_run_prexor(sh, percpu, tx);
1380

1381
	if (test_bit(STRIPE_OP_BIODRAIN, &ops_request)) {
1382
		tx = ops_run_biodrain(sh, tx);
1383 1384 1385
		overlap_clear++;
	}

1386 1387 1388 1389 1390 1391
	if (test_bit(STRIPE_OP_RECONSTRUCT, &ops_request)) {
		if (level < 6)
			ops_run_reconstruct5(sh, percpu, tx);
		else
			ops_run_reconstruct6(sh, percpu, tx);
	}
1392

1393 1394 1395 1396 1397 1398 1399 1400 1401 1402
	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();
	}
1403 1404 1405 1406 1407 1408 1409

	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);
		}
1410
	put_cpu();
1411 1412
}

1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442
#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

1443
static int grow_one_stripe(struct r5conf *conf)
L
Linus Torvalds 已提交
1444 1445
{
	struct stripe_head *sh;
N
Namhyung Kim 已提交
1446
	sh = kmem_cache_zalloc(conf->slab_cache, GFP_KERNEL);
1447 1448
	if (!sh)
		return 0;
N
Namhyung Kim 已提交
1449

1450
	sh->raid_conf = conf;
1451 1452 1453
	#ifdef CONFIG_MULTICORE_RAID456
	init_waitqueue_head(&sh->ops.wait_for_ops);
	#endif
1454

1455 1456
	if (grow_buffers(sh)) {
		shrink_buffers(sh);
1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467
		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;
}

1468
static int grow_stripes(struct r5conf *conf, int num)
1469
{
1470
	struct kmem_cache *sc;
1471
	int devs = max(conf->raid_disks, conf->previous_raid_disks);
L
Linus Torvalds 已提交
1472

1473 1474 1475 1476 1477 1478 1479 1480
	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]);

1481 1482
	conf->active_name = 0;
	sc = kmem_cache_create(conf->cache_name[conf->active_name],
L
Linus Torvalds 已提交
1483
			       sizeof(struct stripe_head)+(devs-1)*sizeof(struct r5dev),
1484
			       0, 0, NULL);
L
Linus Torvalds 已提交
1485 1486 1487
	if (!sc)
		return 1;
	conf->slab_cache = sc;
1488
	conf->pool_size = devs;
1489
	while (num--)
1490
		if (!grow_one_stripe(conf))
L
Linus Torvalds 已提交
1491 1492 1493
			return 1;
	return 0;
}
1494

1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516
/**
 * 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;
}

1517
static int resize_stripes(struct r5conf *conf, int newsize)
1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544
{
	/* 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;
1545
	unsigned long cpu;
1546
	int err;
1547
	struct kmem_cache *sc;
1548 1549 1550 1551 1552
	int i;

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

1553 1554 1555
	err = md_allow_write(conf->mddev);
	if (err)
		return err;
1556

1557 1558 1559
	/* Step 1 */
	sc = kmem_cache_create(conf->cache_name[1-conf->active_name],
			       sizeof(struct stripe_head)+(newsize-1)*sizeof(struct r5dev),
1560
			       0, 0, NULL);
1561 1562 1563 1564
	if (!sc)
		return -ENOMEM;

	for (i = conf->max_nr_stripes; i; i--) {
N
Namhyung Kim 已提交
1565
		nsh = kmem_cache_zalloc(sc, GFP_KERNEL);
1566 1567 1568 1569
		if (!nsh)
			break;

		nsh->raid_conf = conf;
1570 1571 1572
		#ifdef CONFIG_MULTICORE_RAID456
		init_waitqueue_head(&nsh->ops.wait_for_ops);
		#endif
1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594

		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
1595
				    );
1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609
		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
1610
	 * conf->disks and the scribble region
1611 1612 1613 1614 1615 1616 1617 1618 1619 1620
	 */
	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;

1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639
	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();

1640 1641 1642 1643
	/* 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);
1644

1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660
		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
1661

1662
static int drop_one_stripe(struct r5conf *conf)
L
Linus Torvalds 已提交
1663 1664 1665
{
	struct stripe_head *sh;

1666 1667 1668 1669 1670
	spin_lock_irq(&conf->device_lock);
	sh = get_free_stripe(conf);
	spin_unlock_irq(&conf->device_lock);
	if (!sh)
		return 0;
1671
	BUG_ON(atomic_read(&sh->count));
1672
	shrink_buffers(sh);
1673 1674 1675 1676 1677
	kmem_cache_free(conf->slab_cache, sh);
	atomic_dec(&conf->active_stripes);
	return 1;
}

1678
static void shrink_stripes(struct r5conf *conf)
1679 1680 1681 1682
{
	while (drop_one_stripe(conf))
		;

N
NeilBrown 已提交
1683 1684
	if (conf->slab_cache)
		kmem_cache_destroy(conf->slab_cache);
L
Linus Torvalds 已提交
1685 1686 1687
	conf->slab_cache = NULL;
}

1688
static void raid5_end_read_request(struct bio * bi, int error)
L
Linus Torvalds 已提交
1689
{
1690
	struct stripe_head *sh = bi->bi_private;
1691
	struct r5conf *conf = sh->raid_conf;
1692
	int disks = sh->disks, i;
L
Linus Torvalds 已提交
1693
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
1694
	char b[BDEVNAME_SIZE];
1695
	struct md_rdev *rdev = NULL;
1696
	sector_t s;
L
Linus Torvalds 已提交
1697 1698 1699 1700 1701

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

1702 1703
	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
1704 1705 1706
		uptodate);
	if (i == disks) {
		BUG();
1707
		return;
L
Linus Torvalds 已提交
1708
	}
1709
	if (test_bit(R5_ReadRepl, &sh->dev[i].flags))
1710 1711 1712 1713 1714
		/* 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.
		 */
1715
		rdev = conf->disks[i].replacement;
1716
	if (!rdev)
1717
		rdev = conf->disks[i].rdev;
L
Linus Torvalds 已提交
1718

1719 1720 1721 1722
	if (use_new_offset(conf, sh))
		s = sh->sector + rdev->new_data_offset;
	else
		s = sh->sector + rdev->data_offset;
L
Linus Torvalds 已提交
1723 1724
	if (uptodate) {
		set_bit(R5_UPTODATE, &sh->dev[i].flags);
1725
		if (test_bit(R5_ReadError, &sh->dev[i].flags)) {
1726 1727 1728 1729
			/* Note that this cannot happen on a
			 * replacement device.  We just fail those on
			 * any error
			 */
1730 1731 1732 1733 1734
			printk_ratelimited(
				KERN_INFO
				"md/raid:%s: read error corrected"
				" (%lu sectors at %llu on %s)\n",
				mdname(conf->mddev), STRIPE_SECTORS,
1735
				(unsigned long long)s,
1736
				bdevname(rdev->bdev, b));
1737
			atomic_add(STRIPE_SECTORS, &rdev->corrected_errors);
1738 1739 1740
			clear_bit(R5_ReadError, &sh->dev[i].flags);
			clear_bit(R5_ReWrite, &sh->dev[i].flags);
		}
1741 1742
		if (atomic_read(&rdev->read_errors))
			atomic_set(&rdev->read_errors, 0);
L
Linus Torvalds 已提交
1743
	} else {
1744
		const char *bdn = bdevname(rdev->bdev, b);
1745
		int retry = 0;
1746

L
Linus Torvalds 已提交
1747
		clear_bit(R5_UPTODATE, &sh->dev[i].flags);
1748
		atomic_inc(&rdev->read_errors);
1749 1750 1751 1752 1753 1754
		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),
1755
				(unsigned long long)s,
1756 1757
				bdn);
		else if (conf->mddev->degraded >= conf->max_degraded)
1758 1759 1760 1761 1762
			printk_ratelimited(
				KERN_WARNING
				"md/raid:%s: read error not correctable "
				"(sector %llu on %s).\n",
				mdname(conf->mddev),
1763
				(unsigned long long)s,
1764
				bdn);
1765
		else if (test_bit(R5_ReWrite, &sh->dev[i].flags))
1766
			/* Oh, no!!! */
1767 1768 1769 1770 1771
			printk_ratelimited(
				KERN_WARNING
				"md/raid:%s: read error NOT corrected!! "
				"(sector %llu on %s).\n",
				mdname(conf->mddev),
1772
				(unsigned long long)s,
1773
				bdn);
1774
		else if (atomic_read(&rdev->read_errors)
1775
			 > conf->max_nr_stripes)
N
NeilBrown 已提交
1776
			printk(KERN_WARNING
1777
			       "md/raid:%s: Too many read errors, failing device %s.\n",
1778
			       mdname(conf->mddev), bdn);
1779 1780 1781 1782 1783
		else
			retry = 1;
		if (retry)
			set_bit(R5_ReadError, &sh->dev[i].flags);
		else {
1784 1785
			clear_bit(R5_ReadError, &sh->dev[i].flags);
			clear_bit(R5_ReWrite, &sh->dev[i].flags);
1786
			md_error(conf->mddev, rdev);
1787
		}
L
Linus Torvalds 已提交
1788
	}
1789
	rdev_dec_pending(rdev, conf->mddev);
L
Linus Torvalds 已提交
1790 1791 1792 1793 1794
	clear_bit(R5_LOCKED, &sh->dev[i].flags);
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

1795
static void raid5_end_write_request(struct bio *bi, int error)
L
Linus Torvalds 已提交
1796
{
1797
	struct stripe_head *sh = bi->bi_private;
1798
	struct r5conf *conf = sh->raid_conf;
1799
	int disks = sh->disks, i;
1800
	struct md_rdev *uninitialized_var(rdev);
L
Linus Torvalds 已提交
1801
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
1802 1803
	sector_t first_bad;
	int bad_sectors;
1804
	int replacement = 0;
L
Linus Torvalds 已提交
1805

1806 1807 1808
	for (i = 0 ; i < disks; i++) {
		if (bi == &sh->dev[i].req) {
			rdev = conf->disks[i].rdev;
L
Linus Torvalds 已提交
1809
			break;
1810 1811 1812
		}
		if (bi == &sh->dev[i].rreq) {
			rdev = conf->disks[i].replacement;
1813 1814 1815 1816 1817 1818 1819 1820
			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;
1821 1822 1823
			break;
		}
	}
1824
	pr_debug("end_write_request %llu/%d, count %d, uptodate: %d.\n",
L
Linus Torvalds 已提交
1825 1826 1827 1828
		(unsigned long long)sh->sector, i, atomic_read(&sh->count),
		uptodate);
	if (i == disks) {
		BUG();
1829
		return;
L
Linus Torvalds 已提交
1830 1831
	}

1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842
	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);
1843 1844 1845
			if (!test_and_set_bit(WantReplacement, &rdev->flags))
				set_bit(MD_RECOVERY_NEEDED,
					&rdev->mddev->recovery);
1846 1847 1848 1849 1850 1851
		} 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
1852

1853 1854
	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
1855
	set_bit(STRIPE_HANDLE, &sh->state);
1856
	release_stripe(sh);
L
Linus Torvalds 已提交
1857 1858
}

1859
static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous);
L
Linus Torvalds 已提交
1860
	
1861
static void raid5_build_block(struct stripe_head *sh, int i, int previous)
L
Linus Torvalds 已提交
1862 1863 1864 1865 1866 1867 1868 1869
{
	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;
1870
	dev->vec.bv_page = dev->page;
L
Linus Torvalds 已提交
1871

1872 1873 1874 1875 1876 1877 1878
	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
1879
	dev->flags = 0;
1880
	dev->sector = compute_blocknr(sh, i, previous);
L
Linus Torvalds 已提交
1881 1882
}

1883
static void error(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
1884 1885
{
	char b[BDEVNAME_SIZE];
1886
	struct r5conf *conf = mddev->private;
1887
	unsigned long flags;
1888
	pr_debug("raid456: error called\n");
L
Linus Torvalds 已提交
1889

1890 1891 1892 1893 1894 1895
	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);

1896
	set_bit(Blocked, &rdev->flags);
1897 1898 1899 1900 1901 1902 1903 1904 1905
	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);
1906
}
L
Linus Torvalds 已提交
1907 1908 1909 1910 1911

/*
 * Input: a 'big' sector number,
 * Output: index of the data and parity disk, and the sector # in them.
 */
1912
static sector_t raid5_compute_sector(struct r5conf *conf, sector_t r_sector,
1913 1914
				     int previous, int *dd_idx,
				     struct stripe_head *sh)
L
Linus Torvalds 已提交
1915
{
N
NeilBrown 已提交
1916
	sector_t stripe, stripe2;
1917
	sector_t chunk_number;
L
Linus Torvalds 已提交
1918
	unsigned int chunk_offset;
1919
	int pd_idx, qd_idx;
1920
	int ddf_layout = 0;
L
Linus Torvalds 已提交
1921
	sector_t new_sector;
1922 1923
	int algorithm = previous ? conf->prev_algo
				 : conf->algorithm;
1924 1925
	int sectors_per_chunk = previous ? conf->prev_chunk_sectors
					 : conf->chunk_sectors;
1926 1927 1928
	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
1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940

	/* 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
	 */
1941 1942
	stripe = chunk_number;
	*dd_idx = sector_div(stripe, data_disks);
N
NeilBrown 已提交
1943
	stripe2 = stripe;
L
Linus Torvalds 已提交
1944 1945 1946
	/*
	 * Select the parity disk based on the user selected algorithm.
	 */
1947
	pd_idx = qd_idx = -1;
1948 1949
	switch(conf->level) {
	case 4:
1950
		pd_idx = data_disks;
1951 1952
		break;
	case 5:
1953
		switch (algorithm) {
L
Linus Torvalds 已提交
1954
		case ALGORITHM_LEFT_ASYMMETRIC:
N
NeilBrown 已提交
1955
			pd_idx = data_disks - sector_div(stripe2, raid_disks);
1956
			if (*dd_idx >= pd_idx)
L
Linus Torvalds 已提交
1957 1958 1959
				(*dd_idx)++;
			break;
		case ALGORITHM_RIGHT_ASYMMETRIC:
N
NeilBrown 已提交
1960
			pd_idx = sector_div(stripe2, raid_disks);
1961
			if (*dd_idx >= pd_idx)
L
Linus Torvalds 已提交
1962 1963 1964
				(*dd_idx)++;
			break;
		case ALGORITHM_LEFT_SYMMETRIC:
N
NeilBrown 已提交
1965
			pd_idx = data_disks - sector_div(stripe2, raid_disks);
1966
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
L
Linus Torvalds 已提交
1967 1968
			break;
		case ALGORITHM_RIGHT_SYMMETRIC:
N
NeilBrown 已提交
1969
			pd_idx = sector_div(stripe2, raid_disks);
1970
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
L
Linus Torvalds 已提交
1971
			break;
1972 1973 1974 1975 1976 1977 1978
		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
1979
		default:
1980
			BUG();
1981 1982 1983 1984
		}
		break;
	case 6:

1985
		switch (algorithm) {
1986
		case ALGORITHM_LEFT_ASYMMETRIC:
N
NeilBrown 已提交
1987
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
1988 1989
			qd_idx = pd_idx + 1;
			if (pd_idx == raid_disks-1) {
1990
				(*dd_idx)++;	/* Q D D D P */
1991 1992
				qd_idx = 0;
			} else if (*dd_idx >= pd_idx)
1993 1994 1995
				(*dd_idx) += 2; /* D D P Q D */
			break;
		case ALGORITHM_RIGHT_ASYMMETRIC:
N
NeilBrown 已提交
1996
			pd_idx = sector_div(stripe2, raid_disks);
1997 1998
			qd_idx = pd_idx + 1;
			if (pd_idx == raid_disks-1) {
1999
				(*dd_idx)++;	/* Q D D D P */
2000 2001
				qd_idx = 0;
			} else if (*dd_idx >= pd_idx)
2002 2003 2004
				(*dd_idx) += 2; /* D D P Q D */
			break;
		case ALGORITHM_LEFT_SYMMETRIC:
N
NeilBrown 已提交
2005
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2006 2007
			qd_idx = (pd_idx + 1) % raid_disks;
			*dd_idx = (pd_idx + 2 + *dd_idx) % raid_disks;
2008 2009
			break;
		case ALGORITHM_RIGHT_SYMMETRIC:
N
NeilBrown 已提交
2010
			pd_idx = sector_div(stripe2, raid_disks);
2011 2012
			qd_idx = (pd_idx + 1) % raid_disks;
			*dd_idx = (pd_idx + 2 + *dd_idx) % raid_disks;
2013
			break;
2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028

		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
2029
			pd_idx = sector_div(stripe2, raid_disks);
2030 2031 2032 2033 2034 2035
			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 */
2036
			ddf_layout = 1;
2037 2038 2039 2040 2041 2042 2043
			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
2044 2045
			stripe2 += 1;
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2046 2047 2048 2049 2050 2051
			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 */
2052
			ddf_layout = 1;
2053 2054 2055 2056
			break;

		case ALGORITHM_ROTATING_N_CONTINUE:
			/* Same as left_symmetric but Q is before P */
N
NeilBrown 已提交
2057
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2058 2059
			qd_idx = (pd_idx + raid_disks - 1) % raid_disks;
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
2060
			ddf_layout = 1;
2061 2062 2063 2064
			break;

		case ALGORITHM_LEFT_ASYMMETRIC_6:
			/* RAID5 left_asymmetric, with Q on last device */
N
NeilBrown 已提交
2065
			pd_idx = data_disks - sector_div(stripe2, raid_disks-1);
2066 2067 2068 2069 2070 2071
			if (*dd_idx >= pd_idx)
				(*dd_idx)++;
			qd_idx = raid_disks - 1;
			break;

		case ALGORITHM_RIGHT_ASYMMETRIC_6:
N
NeilBrown 已提交
2072
			pd_idx = sector_div(stripe2, raid_disks-1);
2073 2074 2075 2076 2077 2078
			if (*dd_idx >= pd_idx)
				(*dd_idx)++;
			qd_idx = raid_disks - 1;
			break;

		case ALGORITHM_LEFT_SYMMETRIC_6:
N
NeilBrown 已提交
2079
			pd_idx = data_disks - sector_div(stripe2, raid_disks-1);
2080 2081 2082 2083 2084
			*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
2085
			pd_idx = sector_div(stripe2, raid_disks-1);
2086 2087 2088 2089 2090 2091 2092 2093 2094 2095
			*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;

2096
		default:
2097
			BUG();
2098 2099
		}
		break;
L
Linus Torvalds 已提交
2100 2101
	}

2102 2103 2104
	if (sh) {
		sh->pd_idx = pd_idx;
		sh->qd_idx = qd_idx;
2105
		sh->ddf_layout = ddf_layout;
2106
	}
L
Linus Torvalds 已提交
2107 2108 2109 2110 2111 2112 2113 2114
	/*
	 * Finally, compute the new sector number
	 */
	new_sector = (sector_t)stripe * sectors_per_chunk + chunk_offset;
	return new_sector;
}


2115
static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous)
L
Linus Torvalds 已提交
2116
{
2117
	struct r5conf *conf = sh->raid_conf;
2118 2119
	int raid_disks = sh->disks;
	int data_disks = raid_disks - conf->max_degraded;
L
Linus Torvalds 已提交
2120
	sector_t new_sector = sh->sector, check;
2121 2122
	int sectors_per_chunk = previous ? conf->prev_chunk_sectors
					 : conf->chunk_sectors;
2123 2124
	int algorithm = previous ? conf->prev_algo
				 : conf->algorithm;
L
Linus Torvalds 已提交
2125 2126
	sector_t stripe;
	int chunk_offset;
2127 2128
	sector_t chunk_number;
	int dummy1, dd_idx = i;
L
Linus Torvalds 已提交
2129
	sector_t r_sector;
2130
	struct stripe_head sh2;
L
Linus Torvalds 已提交
2131

2132

L
Linus Torvalds 已提交
2133 2134 2135
	chunk_offset = sector_div(new_sector, sectors_per_chunk);
	stripe = new_sector;

2136 2137 2138 2139 2140
	if (i == sh->pd_idx)
		return 0;
	switch(conf->level) {
	case 4: break;
	case 5:
2141
		switch (algorithm) {
L
Linus Torvalds 已提交
2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152
		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;
2153 2154 2155 2156 2157
		case ALGORITHM_PARITY_0:
			i -= 1;
			break;
		case ALGORITHM_PARITY_N:
			break;
L
Linus Torvalds 已提交
2158
		default:
2159
			BUG();
2160 2161 2162
		}
		break;
	case 6:
2163
		if (i == sh->qd_idx)
2164
			return 0; /* It is the Q disk */
2165
		switch (algorithm) {
2166 2167
		case ALGORITHM_LEFT_ASYMMETRIC:
		case ALGORITHM_RIGHT_ASYMMETRIC:
2168 2169 2170 2171
		case ALGORITHM_ROTATING_ZERO_RESTART:
		case ALGORITHM_ROTATING_N_RESTART:
			if (sh->pd_idx == raid_disks-1)
				i--;	/* Q D D D P */
2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185
			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;
2186 2187 2188 2189 2190 2191
		case ALGORITHM_PARITY_0:
			i -= 2;
			break;
		case ALGORITHM_PARITY_N:
			break;
		case ALGORITHM_ROTATING_N_CONTINUE:
2192
			/* Like left_symmetric, but P is before Q */
2193 2194
			if (sh->pd_idx == 0)
				i--;	/* P D D D Q */
2195 2196 2197 2198 2199 2200
			else {
				/* D D Q P D */
				if (i < sh->pd_idx)
					i += raid_disks;
				i -= (sh->pd_idx + 1);
			}
2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215
			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;
2216
		default:
2217
			BUG();
2218 2219
		}
		break;
L
Linus Torvalds 已提交
2220 2221 2222
	}

	chunk_number = stripe * data_disks + i;
2223
	r_sector = chunk_number * sectors_per_chunk + chunk_offset;
L
Linus Torvalds 已提交
2224

2225
	check = raid5_compute_sector(conf, r_sector,
2226
				     previous, &dummy1, &sh2);
2227 2228
	if (check != sh->sector || dummy1 != dd_idx || sh2.pd_idx != sh->pd_idx
		|| sh2.qd_idx != sh->qd_idx) {
2229 2230
		printk(KERN_ERR "md/raid:%s: compute_blocknr: map not correct\n",
		       mdname(conf->mddev));
L
Linus Torvalds 已提交
2231 2232 2233 2234 2235 2236
		return 0;
	}
	return r_sector;
}


2237
static void
2238
schedule_reconstruction(struct stripe_head *sh, struct stripe_head_state *s,
2239
			 int rcw, int expand)
2240 2241
{
	int i, pd_idx = sh->pd_idx, disks = sh->disks;
2242
	struct r5conf *conf = sh->raid_conf;
2243
	int level = conf->level;
2244 2245 2246 2247 2248 2249 2250

	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) {
2251 2252 2253 2254
			sh->reconstruct_state = reconstruct_state_drain_run;
			set_bit(STRIPE_OP_BIODRAIN, &s->ops_request);
		} else
			sh->reconstruct_state = reconstruct_state_run;
2255

2256
		set_bit(STRIPE_OP_RECONSTRUCT, &s->ops_request);
2257 2258 2259 2260 2261 2262

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

			if (dev->towrite) {
				set_bit(R5_LOCKED, &dev->flags);
2263
				set_bit(R5_Wantdrain, &dev->flags);
2264 2265
				if (!expand)
					clear_bit(R5_UPTODATE, &dev->flags);
2266
				s->locked++;
2267 2268
			}
		}
2269
		if (s->locked + conf->max_degraded == disks)
2270
			if (!test_and_set_bit(STRIPE_FULL_WRITE, &sh->state))
2271
				atomic_inc(&conf->pending_full_writes);
2272
	} else {
2273
		BUG_ON(level == 6);
2274 2275 2276
		BUG_ON(!(test_bit(R5_UPTODATE, &sh->dev[pd_idx].flags) ||
			test_bit(R5_Wantcompute, &sh->dev[pd_idx].flags)));

2277
		sh->reconstruct_state = reconstruct_state_prexor_drain_run;
2278 2279
		set_bit(STRIPE_OP_PREXOR, &s->ops_request);
		set_bit(STRIPE_OP_BIODRAIN, &s->ops_request);
2280
		set_bit(STRIPE_OP_RECONSTRUCT, &s->ops_request);
2281 2282 2283 2284 2285 2286 2287 2288

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

			if (dev->towrite &&
			    (test_bit(R5_UPTODATE, &dev->flags) ||
2289 2290
			     test_bit(R5_Wantcompute, &dev->flags))) {
				set_bit(R5_Wantdrain, &dev->flags);
2291 2292
				set_bit(R5_LOCKED, &dev->flags);
				clear_bit(R5_UPTODATE, &dev->flags);
2293
				s->locked++;
2294 2295 2296 2297
			}
		}
	}

2298
	/* keep the parity disk(s) locked while asynchronous operations
2299 2300 2301 2302
	 * are in flight
	 */
	set_bit(R5_LOCKED, &sh->dev[pd_idx].flags);
	clear_bit(R5_UPTODATE, &sh->dev[pd_idx].flags);
2303
	s->locked++;
2304

2305 2306 2307 2308 2309 2310 2311 2312 2313
	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++;
	}

2314
	pr_debug("%s: stripe %llu locked: %d ops_request: %lx\n",
2315
		__func__, (unsigned long long)sh->sector,
2316
		s->locked, s->ops_request);
2317
}
2318

L
Linus Torvalds 已提交
2319 2320
/*
 * Each stripe/dev can have one or more bion attached.
2321
 * toread/towrite point to the first in a chain.
L
Linus Torvalds 已提交
2322 2323 2324 2325 2326
 * 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;
2327
	struct r5conf *conf = sh->raid_conf;
2328
	int firstwrite=0;
L
Linus Torvalds 已提交
2329

2330
	pr_debug("adding bi b#%llu to stripe s#%llu\n",
L
Linus Torvalds 已提交
2331 2332 2333 2334 2335
		(unsigned long long)bi->bi_sector,
		(unsigned long long)sh->sector);


	spin_lock_irq(&conf->device_lock);
2336
	if (forwrite) {
L
Linus Torvalds 已提交
2337
		bip = &sh->dev[dd_idx].towrite;
2338 2339 2340
		if (*bip == NULL && sh->dev[dd_idx].written == NULL)
			firstwrite = 1;
	} else
L
Linus Torvalds 已提交
2341 2342 2343 2344 2345 2346 2347 2348 2349
		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;

2350
	BUG_ON(*bip && bi->bi_next && (*bip) != bi->bi_next);
L
Linus Torvalds 已提交
2351 2352 2353
	if (*bip)
		bi->bi_next = *bip;
	*bip = bi;
2354
	bi->bi_phys_segments++;
2355

L
Linus Torvalds 已提交
2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368
	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);
	}
2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380
	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
2381 2382 2383 2384 2385 2386 2387 2388
	return 1;

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

2389
static void end_reshape(struct r5conf *conf);
2390

2391
static void stripe_set_idx(sector_t stripe, struct r5conf *conf, int previous,
2392
			    struct stripe_head *sh)
2393
{
2394
	int sectors_per_chunk =
2395
		previous ? conf->prev_chunk_sectors : conf->chunk_sectors;
2396
	int dd_idx;
2397
	int chunk_offset = sector_div(stripe, sectors_per_chunk);
2398
	int disks = previous ? conf->previous_raid_disks : conf->raid_disks;
2399

2400 2401
	raid5_compute_sector(conf,
			     stripe * (disks - conf->max_degraded)
2402
			     *sectors_per_chunk + chunk_offset,
2403
			     previous,
2404
			     &dd_idx, sh);
2405 2406
}

2407
static void
2408
handle_failed_stripe(struct r5conf *conf, struct stripe_head *sh,
2409 2410 2411 2412 2413 2414 2415 2416 2417
				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)) {
2418
			struct md_rdev *rdev;
2419 2420 2421
			rcu_read_lock();
			rdev = rcu_dereference(conf->disks[i].rdev);
			if (rdev && test_bit(In_sync, &rdev->flags))
2422 2423 2424
				atomic_inc(&rdev->nr_pending);
			else
				rdev = NULL;
2425
			rcu_read_unlock();
2426 2427 2428 2429 2430 2431 2432 2433
			if (rdev) {
				if (!rdev_set_badblocks(
					    rdev,
					    sh->sector,
					    STRIPE_SECTORS, 0))
					md_error(conf->mddev, rdev);
				rdev_dec_pending(rdev, conf->mddev);
			}
2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450
		}
		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);
2451
			if (!raid5_dec_bi_phys_segments(bi)) {
2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465
				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);
2466
			if (!raid5_dec_bi_phys_segments(bi)) {
2467 2468 2469 2470 2471 2472 2473
				md_write_end(conf->mddev);
				bi->bi_next = *return_bi;
				*return_bi = bi;
			}
			bi = bi2;
		}

2474 2475 2476 2477 2478 2479
		/* 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))) {
2480 2481 2482 2483 2484 2485 2486 2487 2488 2489
			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);
2490
				if (!raid5_dec_bi_phys_segments(bi)) {
2491 2492 2493 2494 2495 2496 2497 2498 2499 2500
					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);
2501 2502 2503 2504
		/* 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);
2505 2506
	}

2507 2508 2509
	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);
2510 2511
}

2512
static void
2513
handle_failed_sync(struct r5conf *conf, struct stripe_head *sh,
2514 2515 2516 2517 2518 2519 2520
		   struct stripe_head_state *s)
{
	int abort = 0;
	int i;

	clear_bit(STRIPE_SYNCING, &sh->state);
	s->syncing = 0;
2521
	s->replacing = 0;
2522
	/* There is nothing more to do for sync/check/repair.
2523 2524 2525
	 * Don't even need to abort as that is handled elsewhere
	 * if needed, and not always wanted e.g. if there is a known
	 * bad block here.
2526
	 * For recover/replace we need to record a bad block on all
2527 2528
	 * non-sync devices, or abort the recovery
	 */
2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551
	if (test_bit(MD_RECOVERY_RECOVER, &conf->mddev->recovery)) {
		/* During recovery devices cannot be removed, so
		 * locking and refcounting of rdevs is not needed
		 */
		for (i = 0; i < conf->raid_disks; i++) {
			struct md_rdev *rdev = conf->disks[i].rdev;
			if (rdev
			    && !test_bit(Faulty, &rdev->flags)
			    && !test_bit(In_sync, &rdev->flags)
			    && !rdev_set_badblocks(rdev, sh->sector,
						   STRIPE_SECTORS, 0))
				abort = 1;
			rdev = conf->disks[i].replacement;
			if (rdev
			    && !test_bit(Faulty, &rdev->flags)
			    && !test_bit(In_sync, &rdev->flags)
			    && !rdev_set_badblocks(rdev, sh->sector,
						   STRIPE_SECTORS, 0))
				abort = 1;
		}
		if (abort)
			conf->recovery_disabled =
				conf->mddev->recovery_disabled;
2552
	}
2553
	md_done_sync(conf->mddev, STRIPE_SECTORS, !abort);
2554 2555
}

2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571
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;
}

2572
/* fetch_block - checks the given member device to see if its data needs
2573 2574 2575
 * to be read or computed to satisfy a request.
 *
 * Returns 1 when no more member devices need to be checked, otherwise returns
2576
 * 0 to tell the loop in handle_stripe_fill to continue
2577
 */
2578 2579
static int fetch_block(struct stripe_head *sh, struct stripe_head_state *s,
		       int disk_idx, int disks)
2580
{
2581
	struct r5dev *dev = &sh->dev[disk_idx];
2582 2583
	struct r5dev *fdev[2] = { &sh->dev[s->failed_num[0]],
				  &sh->dev[s->failed_num[1]] };
2584

2585
	/* is the data in this block needed, and can we get it? */
2586 2587 2588 2589 2590
	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 ||
2591
	     (s->replacing && want_replace(sh, disk_idx)) ||
2592 2593
	     (s->failed >= 1 && fdev[0]->toread) ||
	     (s->failed >= 2 && fdev[1]->toread) ||
2594 2595 2596
	     (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))) {
2597 2598 2599 2600 2601 2602
		/* 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) &&
2603 2604
		    (s->failed && (disk_idx == s->failed_num[0] ||
				   disk_idx == s->failed_num[1]))) {
2605 2606
			/* have disk failed, and we're requested to fetch it;
			 * do compute it
2607
			 */
2608 2609 2610 2611 2612 2613 2614 2615
			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;
2616 2617 2618 2619 2620 2621
			/* 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.
			 */
2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634
			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;
2635
			}
2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654
			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);
2655 2656
		}
	}
2657 2658 2659 2660 2661

	return 0;
}

/**
2662
 * handle_stripe_fill - read or compute data to satisfy pending requests.
2663
 */
2664 2665 2666
static void handle_stripe_fill(struct stripe_head *sh,
			       struct stripe_head_state *s,
			       int disks)
2667 2668 2669 2670 2671 2672 2673 2674 2675 2676
{
	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--; )
2677
			if (fetch_block(sh, s, i, disks))
2678
				break;
2679 2680 2681 2682
	set_bit(STRIPE_HANDLE, &sh->state);
}


2683
/* handle_stripe_clean_event
2684 2685 2686 2687
 * 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.
 */
2688
static void handle_stripe_clean_event(struct r5conf *conf,
2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701
	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;
2702
				pr_debug("Return write for disc %d\n", i);
2703 2704 2705 2706 2707 2708
				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);
2709
					if (!raid5_dec_bi_phys_segments(wbi)) {
2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726
						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);
			}
		}
2727 2728 2729 2730

	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);
2731 2732
}

2733
static void handle_stripe_dirtying(struct r5conf *conf,
2734 2735 2736
				   struct stripe_head *sh,
				   struct stripe_head_state *s,
				   int disks)
2737 2738
{
	int rmw = 0, rcw = 0, i;
2739 2740 2741 2742 2743 2744 2745
	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--; ) {
2746 2747 2748 2749
		/* 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) &&
2750 2751
		    !(test_bit(R5_UPTODATE, &dev->flags) ||
		      test_bit(R5_Wantcompute, &dev->flags))) {
2752 2753 2754 2755 2756 2757 2758 2759
			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) &&
2760 2761 2762
		    !(test_bit(R5_UPTODATE, &dev->flags) ||
		    test_bit(R5_Wantcompute, &dev->flags))) {
			if (test_bit(R5_Insync, &dev->flags)) rcw++;
2763 2764 2765 2766
			else
				rcw += 2*disks;
		}
	}
2767
	pr_debug("for sector %llu, rmw=%d rcw=%d\n",
2768 2769 2770 2771 2772 2773 2774 2775
		(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) &&
2776 2777
			    !(test_bit(R5_UPTODATE, &dev->flags) ||
			    test_bit(R5_Wantcompute, &dev->flags)) &&
2778 2779 2780
			    test_bit(R5_Insync, &dev->flags)) {
				if (
				  test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
2781
					pr_debug("Read_old block "
2782 2783 2784 2785 2786 2787 2788 2789 2790 2791
						"%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);
				}
			}
		}
2792
	if (rcw <= rmw && rcw > 0) {
2793
		/* want reconstruct write, but need to get some data */
2794
		rcw = 0;
2795 2796 2797
		for (i = disks; i--; ) {
			struct r5dev *dev = &sh->dev[i];
			if (!test_bit(R5_OVERWRITE, &dev->flags) &&
2798
			    i != sh->pd_idx && i != sh->qd_idx &&
2799
			    !test_bit(R5_LOCKED, &dev->flags) &&
2800
			    !(test_bit(R5_UPTODATE, &dev->flags) ||
2801 2802 2803 2804
			      test_bit(R5_Wantcompute, &dev->flags))) {
				rcw++;
				if (!test_bit(R5_Insync, &dev->flags))
					continue; /* it's a failed drive */
2805 2806
				if (
				  test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
2807
					pr_debug("Read_old block "
2808 2809 2810 2811 2812 2813 2814 2815 2816 2817
						"%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);
				}
			}
		}
2818
	}
2819 2820 2821
	/* now if nothing is locked, and if we have enough data,
	 * we can start a write request
	 */
2822 2823
	/* 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
2824 2825
	 * subsequent call wants to start a write request.  raid_run_ops only
	 * handles the case where compute block and reconstruct are requested
2826 2827 2828
	 * simultaneously.  If this is not the case then new writes need to be
	 * held off until the compute completes.
	 */
2829 2830 2831
	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)))
2832
		schedule_reconstruction(sh, s, rcw == 0, 0);
2833 2834
}

2835
static void handle_parity_checks5(struct r5conf *conf, struct stripe_head *sh,
2836 2837
				struct stripe_head_state *s, int disks)
{
2838
	struct r5dev *dev = NULL;
2839

2840
	set_bit(STRIPE_HANDLE, &sh->state);
2841

2842 2843 2844
	switch (sh->check_state) {
	case check_state_idle:
		/* start a new check operation if there are no failures */
2845 2846
		if (s->failed == 0) {
			BUG_ON(s->uptodate != disks);
2847 2848
			sh->check_state = check_state_run;
			set_bit(STRIPE_OP_CHECK, &s->ops_request);
2849 2850
			clear_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags);
			s->uptodate--;
2851
			break;
2852
		}
2853
		dev = &sh->dev[s->failed_num[0]];
2854 2855 2856 2857 2858 2859 2860 2861 2862
		/* 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
2863

2864 2865 2866 2867 2868
		/* 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);
2869
		s->locked++;
2870
		set_bit(R5_Wantwrite, &dev->flags);
2871

2872 2873
		clear_bit(STRIPE_DEGRADED, &sh->state);
		set_bit(STRIPE_INSYNC, &sh->state);
2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889
		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
2890
		if ((sh->ops.zero_sum_result & SUM_CHECK_P_RESULT) == 0)
2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901
			/* 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;
2902
				set_bit(STRIPE_COMPUTE_RUN, &sh->state);
2903 2904 2905 2906
				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;
2907
				sh->ops.target2 = -1;
2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918
				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();
2919 2920 2921 2922
	}
}


2923
static void handle_parity_checks6(struct r5conf *conf, struct stripe_head *sh,
2924
				  struct stripe_head_state *s,
2925
				  int disks)
2926 2927
{
	int pd_idx = sh->pd_idx;
N
NeilBrown 已提交
2928
	int qd_idx = sh->qd_idx;
2929
	struct r5dev *dev;
2930 2931 2932 2933

	set_bit(STRIPE_HANDLE, &sh->state);

	BUG_ON(s->failed > 2);
2934

2935 2936 2937 2938 2939 2940
	/* 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
	 */

2941 2942 2943
	switch (sh->check_state) {
	case check_state_idle:
		/* start a new check operation if there are < 2 failures */
2944
		if (s->failed == s->q_failed) {
2945
			/* The only possible failed device holds Q, so it
2946 2947 2948
			 * makes sense to check P (If anything else were failed,
			 * we would have used P to recreate it).
			 */
2949
			sh->check_state = check_state_run;
2950
		}
2951
		if (!s->q_failed && s->failed < 2) {
2952
			/* Q is not failed, and we didn't use it to generate
2953 2954
			 * anything, so it makes sense to check it
			 */
2955 2956 2957 2958
			if (sh->check_state == check_state_run)
				sh->check_state = check_state_run_pq;
			else
				sh->check_state = check_state_run_q;
2959 2960
		}

2961 2962
		/* discard potentially stale zero_sum_result */
		sh->ops.zero_sum_result = 0;
2963

2964 2965 2966 2967
		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--;
2968
		}
2969 2970 2971 2972 2973 2974 2975
		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;
2976 2977
		}

2978 2979 2980 2981 2982
		/* we have 2-disk failure */
		BUG_ON(s->failed != 2);
		/* fall through */
	case check_state_compute_result:
		sh->check_state = check_state_idle;
2983

2984 2985 2986
		/* check that a write has not made the stripe insync */
		if (test_bit(STRIPE_INSYNC, &sh->state))
			break;
2987 2988

		/* now write out any block on a failed drive,
2989
		 * or P or Q if they were recomputed
2990
		 */
2991
		BUG_ON(s->uptodate < disks - 1); /* We don't need Q to recover */
2992
		if (s->failed == 2) {
2993
			dev = &sh->dev[s->failed_num[1]];
2994 2995 2996 2997 2998
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
		if (s->failed >= 1) {
2999
			dev = &sh->dev[s->failed_num[0]];
3000 3001 3002 3003
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
3004
		if (sh->ops.zero_sum_result & SUM_CHECK_P_RESULT) {
3005 3006 3007 3008 3009
			dev = &sh->dev[pd_idx];
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
3010
		if (sh->ops.zero_sum_result & SUM_CHECK_Q_RESULT) {
3011 3012 3013 3014 3015 3016 3017 3018
			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);
3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082
		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();
3083 3084 3085
	}
}

3086
static void handle_stripe_expansion(struct r5conf *conf, struct stripe_head *sh)
3087 3088 3089 3090 3091 3092
{
	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.
	 */
3093
	struct dma_async_tx_descriptor *tx = NULL;
3094 3095
	clear_bit(STRIPE_EXPAND_SOURCE, &sh->state);
	for (i = 0; i < sh->disks; i++)
N
NeilBrown 已提交
3096
		if (i != sh->pd_idx && i != sh->qd_idx) {
3097
			int dd_idx, j;
3098
			struct stripe_head *sh2;
3099
			struct async_submit_ctl submit;
3100

3101
			sector_t bn = compute_blocknr(sh, i, 1);
3102 3103
			sector_t s = raid5_compute_sector(conf, bn, 0,
							  &dd_idx, NULL);
3104
			sh2 = get_active_stripe(conf, s, 0, 1, 1);
3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116
			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;
			}
3117 3118

			/* place all the copies on one channel */
3119
			init_async_submit(&submit, 0, tx, NULL, NULL, NULL);
3120
			tx = async_memcpy(sh2->dev[dd_idx].page,
3121
					  sh->dev[i].page, 0, 0, STRIPE_SIZE,
3122
					  &submit);
3123

3124 3125 3126 3127
			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 &&
3128
				    j != sh2->qd_idx &&
3129 3130 3131 3132 3133 3134 3135
				    !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);
3136

3137
		}
3138 3139 3140 3141 3142
	/* done submitting copies, wait for them to complete */
	if (tx) {
		async_tx_ack(tx);
		dma_wait_for_async_tx(tx);
	}
3143
}
L
Linus Torvalds 已提交
3144 3145 3146 3147

/*
 * handle_stripe - do things to a stripe.
 *
3148 3149
 * 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
3150
 * Possible results:
3151 3152
 *    return some read requests which now have data
 *    return some write requests which are safely on storage
L
Linus Torvalds 已提交
3153 3154 3155 3156 3157
 *    schedule a read on some buffers
 *    schedule a write of some buffers
 *    return confirmation of parity correctness
 *
 */
3158

3159
static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
L
Linus Torvalds 已提交
3160
{
3161
	struct r5conf *conf = sh->raid_conf;
3162
	int disks = sh->disks;
3163 3164
	struct r5dev *dev;
	int i;
3165
	int do_recovery = 0;
L
Linus Torvalds 已提交
3166

3167 3168 3169 3170 3171 3172
	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
3173

3174
	/* Now to look around and see what can be done */
L
Linus Torvalds 已提交
3175
	rcu_read_lock();
3176
	spin_lock_irq(&conf->device_lock);
3177
	for (i=disks; i--; ) {
3178
		struct md_rdev *rdev;
3179 3180 3181
		sector_t first_bad;
		int bad_sectors;
		int is_bad = 0;
3182

3183
		dev = &sh->dev[i];
L
Linus Torvalds 已提交
3184

3185
		pr_debug("check %d: state 0x%lx read %p write %p written %p\n",
3186 3187
			 i, dev->flags,
			 dev->toread, dev->towrite, dev->written);
3188 3189 3190 3191 3192 3193 3194 3195
		/* 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
3196

3197
		/* now count some things */
3198 3199 3200 3201
		if (test_bit(R5_LOCKED, &dev->flags))
			s->locked++;
		if (test_bit(R5_UPTODATE, &dev->flags))
			s->uptodate++;
3202
		if (test_bit(R5_Wantcompute, &dev->flags)) {
3203 3204
			s->compute++;
			BUG_ON(s->compute > 2);
3205
		}
L
Linus Torvalds 已提交
3206

3207
		if (test_bit(R5_Wantfill, &dev->flags))
3208
			s->to_fill++;
3209
		else if (dev->toread)
3210
			s->to_read++;
3211
		if (dev->towrite) {
3212
			s->to_write++;
3213
			if (!test_bit(R5_OVERWRITE, &dev->flags))
3214
				s->non_overwrite++;
3215
		}
3216
		if (dev->written)
3217
			s->written++;
3218 3219 3220 3221 3222 3223 3224 3225 3226 3227
		/* 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 {
3228 3229
			if (rdev)
				set_bit(R5_NeedReplace, &dev->flags);
3230 3231 3232
			rdev = rcu_dereference(conf->disks[i].rdev);
			clear_bit(R5_ReadRepl, &dev->flags);
		}
3233 3234
		if (rdev && test_bit(Faulty, &rdev->flags))
			rdev = NULL;
3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246
		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);
			}
3247
		}
3248 3249 3250
		clear_bit(R5_Insync, &dev->flags);
		if (!rdev)
			/* Not in-sync */;
3251 3252
		else if (is_bad) {
			/* also not in-sync */
3253 3254
			if (!test_bit(WriteErrorSeen, &rdev->flags) &&
			    test_bit(R5_UPTODATE, &dev->flags)) {
3255 3256 3257 3258 3259 3260 3261
				/* 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))
3262
			set_bit(R5_Insync, &dev->flags);
3263
		else if (sh->sector + STRIPE_SECTORS <= rdev->recovery_offset)
3264
			/* in sync if before recovery_offset */
3265 3266 3267 3268 3269 3270 3271 3272 3273
			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
3274
		if (rdev && test_bit(R5_WriteError, &dev->flags)) {
3275 3276 3277 3278 3279 3280 3281
			/* 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)) {
3282
				s->handle_bad_blocks = 1;
3283
				atomic_inc(&rdev2->nr_pending);
3284 3285 3286
			} else
				clear_bit(R5_WriteError, &dev->flags);
		}
A
Adam Kwolek 已提交
3287
		if (rdev && test_bit(R5_MadeGood, &dev->flags)) {
3288 3289 3290 3291 3292
			/* 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)) {
3293
				s->handle_bad_blocks = 1;
3294
				atomic_inc(&rdev2->nr_pending);
3295 3296 3297
			} else
				clear_bit(R5_MadeGood, &dev->flags);
		}
3298 3299 3300 3301 3302 3303 3304 3305 3306
		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);
		}
3307
		if (!test_bit(R5_Insync, &dev->flags)) {
3308 3309 3310
			/* 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
3311
		}
3312 3313 3314
		if (test_bit(R5_ReadError, &dev->flags))
			clear_bit(R5_Insync, &dev->flags);
		if (!test_bit(R5_Insync, &dev->flags)) {
3315 3316 3317
			if (s->failed < 2)
				s->failed_num[s->failed] = i;
			s->failed++;
3318 3319
			if (rdev && !test_bit(Faulty, &rdev->flags))
				do_recovery = 1;
3320
		}
L
Linus Torvalds 已提交
3321
	}
3322
	spin_unlock_irq(&conf->device_lock);
3323 3324 3325 3326
	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
3327
		 * else if MD_RECOVERY_REQUESTED is set, we also are syncing.
3328 3329 3330 3331 3332
		 * else we can only be replacing
		 * sync and recovery both need to read all devices, and so
		 * use the same flag.
		 */
		if (do_recovery ||
3333 3334
		    sh->sector >= conf->mddev->recovery_cp ||
		    test_bit(MD_RECOVERY_REQUESTED, &(conf->mddev->recovery)))
3335 3336 3337 3338
			s->syncing = 1;
		else
			s->replacing = 1;
	}
L
Linus Torvalds 已提交
3339
	rcu_read_unlock();
3340 3341 3342 3343 3344
}

static void handle_stripe(struct stripe_head *sh)
{
	struct stripe_head_state s;
3345
	struct r5conf *conf = sh->raid_conf;
3346
	int i;
3347 3348
	int prexor;
	int disks = sh->disks;
3349
	struct r5dev *pdev, *qdev;
3350 3351

	clear_bit(STRIPE_HANDLE, &sh->state);
3352
	if (test_and_set_bit_lock(STRIPE_ACTIVE, &sh->state)) {
3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369
		/* 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);
3370

3371
	analyse_stripe(sh, &s);
3372

3373 3374 3375 3376 3377
	if (s.handle_bad_blocks) {
		set_bit(STRIPE_HANDLE, &sh->state);
		goto finish;
	}

3378 3379
	if (unlikely(s.blocked_rdev)) {
		if (s.syncing || s.expanding || s.expanded ||
3380
		    s.replacing || s.to_write || s.written) {
3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400
			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.
	 */
3401 3402 3403 3404 3405
	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);
3406
		if (s.syncing + s.replacing)
3407 3408
			handle_failed_sync(conf, sh, &s);
	}
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

	/*
	 * 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)
3437 3438 3439
	    || (s.syncing && (s.uptodate + s.compute < disks))
	    || s.replacing
	    || s.expanding)
3440 3441
		handle_stripe_fill(sh, &s, disks);

3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 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 3492 3493 3494 3495 3496 3497 3498 3499
	/* 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);
	}
3500

3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514
	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)) {
3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543
		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++;
				}
			}
		}


3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570
	/* 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++;
		}
	}
3571

3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587
	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);
3588

3589
finish:
3590
	/* wait for this device to become unblocked */
3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602
	if (unlikely(s.blocked_rdev)) {
		if (conf->mddev->external)
			md_wait_for_blocked_rdev(s.blocked_rdev,
						 conf->mddev);
		else
			/* Internal metadata will immediately
			 * be written by raid5d, so we don't
			 * need to wait here.
			 */
			rdev_dec_pending(s.blocked_rdev,
					 conf->mddev);
	}
3603

3604 3605
	if (s.handle_bad_blocks)
		for (i = disks; i--; ) {
3606
			struct md_rdev *rdev;
3607 3608 3609 3610 3611 3612 3613 3614 3615
			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);
			}
3616 3617 3618
			if (test_and_clear_bit(R5_MadeGood, &dev->flags)) {
				rdev = conf->disks[i].rdev;
				rdev_clear_badblocks(rdev, sh->sector,
3619
						     STRIPE_SECTORS, 0);
3620 3621
				rdev_dec_pending(rdev, conf->mddev);
			}
3622 3623
			if (test_and_clear_bit(R5_MadeGoodRepl, &dev->flags)) {
				rdev = conf->disks[i].replacement;
3624 3625 3626
				if (!rdev)
					/* rdev have been moved down */
					rdev = conf->disks[i].rdev;
3627
				rdev_clear_badblocks(rdev, sh->sector,
3628
						     STRIPE_SECTORS, 0);
3629 3630
				rdev_dec_pending(rdev, conf->mddev);
			}
3631 3632
		}

3633 3634 3635
	if (s.ops_request)
		raid_run_ops(sh, s.ops_request);

D
Dan Williams 已提交
3636
	ops_run_io(sh, &s);
3637

3638
	if (s.dec_preread_active) {
3639
		/* We delay this until after ops_run_io so that if make_request
T
Tejun Heo 已提交
3640
		 * is waiting on a flush, it won't continue until the writes
3641 3642 3643 3644 3645 3646 3647 3648
		 * 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);
	}

3649
	return_io(s.return_bi);
3650

3651
	clear_bit_unlock(STRIPE_ACTIVE, &sh->state);
3652 3653
}

3654
static void raid5_activate_delayed(struct r5conf *conf)
3655 3656 3657 3658 3659 3660 3661 3662 3663 3664
{
	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);
3665
			list_add_tail(&sh->lru, &conf->hold_list);
3666
		}
N
NeilBrown 已提交
3667
	}
3668 3669
}

3670
static void activate_bit_delay(struct r5conf *conf)
3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683
{
	/* 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);
	}
}

3684
int md_raid5_congested(struct mddev *mddev, int bits)
3685
{
3686
	struct r5conf *conf = mddev->private;
3687 3688 3689 3690

	/* No difference between reads and writes.  Just check
	 * how busy the stripe_cache is
	 */
3691

3692 3693 3694 3695 3696 3697 3698 3699 3700
	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
3701 3702 3703 3704
EXPORT_SYMBOL_GPL(md_raid5_congested);

static int raid5_congested(void *data, int bits)
{
3705
	struct mddev *mddev = data;
N
NeilBrown 已提交
3706 3707 3708 3709

	return mddev_congested(mddev, bits) ||
		md_raid5_congested(mddev, bits);
}
3710

3711 3712 3713
/* We want read requests to align with chunks where possible,
 * but write requests don't need to.
 */
3714 3715 3716
static int raid5_mergeable_bvec(struct request_queue *q,
				struct bvec_merge_data *bvm,
				struct bio_vec *biovec)
3717
{
3718
	struct mddev *mddev = q->queuedata;
3719
	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
3720
	int max;
3721
	unsigned int chunk_sectors = mddev->chunk_sectors;
3722
	unsigned int bio_sectors = bvm->bi_size >> 9;
3723

3724
	if ((bvm->bi_rw & 1) == WRITE)
3725 3726
		return biovec->bv_len; /* always allow writes to be mergeable */

3727 3728
	if (mddev->new_chunk_sectors < mddev->chunk_sectors)
		chunk_sectors = mddev->new_chunk_sectors;
3729 3730 3731 3732 3733 3734 3735 3736
	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;
}

3737

3738
static int in_chunk_boundary(struct mddev *mddev, struct bio *bio)
3739 3740
{
	sector_t sector = bio->bi_sector + get_start_sect(bio->bi_bdev);
3741
	unsigned int chunk_sectors = mddev->chunk_sectors;
3742 3743
	unsigned int bio_sectors = bio->bi_size >> 9;

3744 3745
	if (mddev->new_chunk_sectors < mddev->chunk_sectors)
		chunk_sectors = mddev->new_chunk_sectors;
3746 3747 3748 3749
	return  chunk_sectors >=
		((sector & (chunk_sectors - 1)) + bio_sectors);
}

3750 3751 3752 3753
/*
 *  add bio to the retry LIFO  ( in O(1) ... we are in interrupt )
 *  later sampled by raid5d.
 */
3754
static void add_bio_to_retry(struct bio *bi,struct r5conf *conf)
3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767
{
	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);
}


3768
static struct bio *remove_bio_from_retry(struct r5conf *conf)
3769 3770 3771 3772 3773 3774 3775 3776 3777 3778
{
	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) {
3779
		conf->retry_read_aligned_list = bi->bi_next;
3780
		bi->bi_next = NULL;
3781 3782 3783 3784
		/*
		 * this sets the active strip count to 1 and the processed
		 * strip count to zero (upper 8 bits)
		 */
3785 3786 3787 3788 3789 3790 3791
		bi->bi_phys_segments = 1; /* biased count of active stripes */
	}

	return bi;
}


3792 3793 3794 3795 3796 3797
/*
 *  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..
 */
3798
static void raid5_align_endio(struct bio *bi, int error)
3799 3800
{
	struct bio* raid_bi  = bi->bi_private;
3801
	struct mddev *mddev;
3802
	struct r5conf *conf;
3803
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
3804
	struct md_rdev *rdev;
3805

3806
	bio_put(bi);
3807 3808 3809

	rdev = (void*)raid_bi->bi_next;
	raid_bi->bi_next = NULL;
3810 3811
	mddev = rdev->mddev;
	conf = mddev->private;
3812 3813 3814 3815

	rdev_dec_pending(rdev, conf->mddev);

	if (!error && uptodate) {
3816
		bio_endio(raid_bi, 0);
3817 3818
		if (atomic_dec_and_test(&conf->active_aligned_reads))
			wake_up(&conf->wait_for_stripe);
3819
		return;
3820 3821 3822
	}


3823
	pr_debug("raid5_align_endio : io error...handing IO for a retry\n");
3824 3825

	add_bio_to_retry(raid_bi, conf);
3826 3827
}

3828 3829
static int bio_fits_rdev(struct bio *bi)
{
3830
	struct request_queue *q = bdev_get_queue(bi->bi_bdev);
3831

3832
	if ((bi->bi_size>>9) > queue_max_sectors(q))
3833 3834
		return 0;
	blk_recount_segments(q, bi);
3835
	if (bi->bi_phys_segments > queue_max_segments(q))
3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846 3847
		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;
}


3848
static int chunk_aligned_read(struct mddev *mddev, struct bio * raid_bio)
3849
{
3850
	struct r5conf *conf = mddev->private;
N
NeilBrown 已提交
3851
	int dd_idx;
3852
	struct bio* align_bi;
3853
	struct md_rdev *rdev;
3854
	sector_t end_sector;
3855 3856

	if (!in_chunk_boundary(mddev, raid_bio)) {
3857
		pr_debug("chunk_aligned_read : non aligned\n");
3858 3859 3860
		return 0;
	}
	/*
3861
	 * use bio_clone_mddev to make a copy of the bio
3862
	 */
3863
	align_bi = bio_clone_mddev(raid_bio, GFP_NOIO, mddev);
3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874
	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
	 */
3875 3876
	align_bi->bi_sector =  raid5_compute_sector(conf, raid_bio->bi_sector,
						    0,
3877
						    &dd_idx, NULL);
3878

3879
	end_sector = align_bi->bi_sector + (align_bi->bi_size >> 9);
3880
	rcu_read_lock();
3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891
	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) {
3892 3893 3894
		sector_t first_bad;
		int bad_sectors;

3895 3896
		atomic_inc(&rdev->nr_pending);
		rcu_read_unlock();
3897 3898 3899 3900
		raid_bio->bi_next = (void*)rdev;
		align_bi->bi_bdev =  rdev->bdev;
		align_bi->bi_flags &= ~(1 << BIO_SEG_VALID);

3901 3902 3903 3904
		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 */
3905 3906 3907 3908 3909
			bio_put(align_bi);
			rdev_dec_pending(rdev, mddev);
			return 0;
		}

3910 3911 3912
		/* No reshape active, so we can trust rdev->data_offset */
		align_bi->bi_sector += rdev->data_offset;

3913 3914 3915 3916 3917 3918 3919
		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);

3920 3921 3922 3923
		generic_make_request(align_bi);
		return 1;
	} else {
		rcu_read_unlock();
3924
		bio_put(align_bi);
3925 3926 3927 3928
		return 0;
	}
}

3929 3930 3931 3932 3933 3934 3935 3936 3937 3938
/* __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.
 */
3939
static struct stripe_head *__get_priority_stripe(struct r5conf *conf)
3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980
{
	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;
}
3981

3982
static void make_request(struct mddev *mddev, struct bio * bi)
L
Linus Torvalds 已提交
3983
{
3984
	struct r5conf *conf = mddev->private;
3985
	int dd_idx;
L
Linus Torvalds 已提交
3986 3987 3988
	sector_t new_sector;
	sector_t logical_sector, last_sector;
	struct stripe_head *sh;
3989
	const int rw = bio_data_dir(bi);
3990
	int remaining;
3991
	int plugged;
L
Linus Torvalds 已提交
3992

T
Tejun Heo 已提交
3993 3994
	if (unlikely(bi->bi_rw & REQ_FLUSH)) {
		md_flush_request(mddev, bi);
3995
		return;
3996 3997
	}

3998
	md_write_start(mddev, bi);
3999

4000
	if (rw == READ &&
4001
	     mddev->reshape_position == MaxSector &&
4002
	     chunk_aligned_read(mddev,bi))
4003
		return;
4004

L
Linus Torvalds 已提交
4005 4006 4007 4008
	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 */
4009

4010
	plugged = mddev_check_plugged(mddev);
L
Linus Torvalds 已提交
4011 4012
	for (;logical_sector < last_sector; logical_sector += STRIPE_SECTORS) {
		DEFINE_WAIT(w);
4013
		int previous;
4014

4015
	retry:
4016
		previous = 0;
4017
		prepare_to_wait(&conf->wait_for_overlap, &w, TASK_UNINTERRUPTIBLE);
4018
		if (unlikely(conf->reshape_progress != MaxSector)) {
4019
			/* spinlock is needed as reshape_progress may be
4020 4021
			 * 64bit on a 32bit platform, and so it might be
			 * possible to see a half-updated value
4022
			 * Of course reshape_progress could change after
4023 4024 4025 4026
			 * the lock is dropped, so once we get a reference
			 * to the stripe that we think it is, we will have
			 * to check again.
			 */
4027
			spin_lock_irq(&conf->device_lock);
4028
			if (mddev->reshape_backwards
4029 4030
			    ? logical_sector < conf->reshape_progress
			    : logical_sector >= conf->reshape_progress) {
4031 4032
				previous = 1;
			} else {
4033
				if (mddev->reshape_backwards
4034 4035
				    ? logical_sector < conf->reshape_safe
				    : logical_sector >= conf->reshape_safe) {
4036 4037 4038 4039 4040
					spin_unlock_irq(&conf->device_lock);
					schedule();
					goto retry;
				}
			}
4041 4042
			spin_unlock_irq(&conf->device_lock);
		}
4043

4044 4045
		new_sector = raid5_compute_sector(conf, logical_sector,
						  previous,
4046
						  &dd_idx, NULL);
4047
		pr_debug("raid456: make_request, sector %llu logical %llu\n",
L
Linus Torvalds 已提交
4048 4049 4050
			(unsigned long long)new_sector, 
			(unsigned long long)logical_sector);

4051
		sh = get_active_stripe(conf, new_sector, previous,
4052
				       (bi->bi_rw&RWA_MASK), 0);
L
Linus Torvalds 已提交
4053
		if (sh) {
4054
			if (unlikely(previous)) {
4055
				/* expansion might have moved on while waiting for a
4056 4057 4058 4059 4060 4061
				 * 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.
4062 4063 4064
				 */
				int must_retry = 0;
				spin_lock_irq(&conf->device_lock);
4065
				if (mddev->reshape_backwards
4066 4067
				    ? logical_sector >= conf->reshape_progress
				    : logical_sector < conf->reshape_progress)
4068 4069 4070 4071 4072
					/* mismatch, need to try again */
					must_retry = 1;
				spin_unlock_irq(&conf->device_lock);
				if (must_retry) {
					release_stripe(sh);
4073
					schedule();
4074 4075 4076
					goto retry;
				}
			}
4077

4078
			if (rw == WRITE &&
4079
			    logical_sector >= mddev->suspend_lo &&
4080 4081
			    logical_sector < mddev->suspend_hi) {
				release_stripe(sh);
4082 4083 4084 4085 4086 4087 4088 4089 4090 4091
				/* 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();
4092 4093
				goto retry;
			}
4094 4095

			if (test_bit(STRIPE_EXPANDING, &sh->state) ||
4096
			    !add_stripe_bio(sh, bi, dd_idx, rw)) {
4097 4098
				/* Stripe is busy expanding or
				 * add failed due to overlap.  Flush everything
L
Linus Torvalds 已提交
4099 4100
				 * and wait a while
				 */
N
NeilBrown 已提交
4101
				md_wakeup_thread(mddev->thread);
L
Linus Torvalds 已提交
4102 4103 4104 4105 4106
				release_stripe(sh);
				schedule();
				goto retry;
			}
			finish_wait(&conf->wait_for_overlap, &w);
4107 4108
			set_bit(STRIPE_HANDLE, &sh->state);
			clear_bit(STRIPE_DELAYED, &sh->state);
T
Tejun Heo 已提交
4109
			if ((bi->bi_rw & REQ_SYNC) &&
4110 4111
			    !test_and_set_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
				atomic_inc(&conf->preread_active_stripes);
L
Linus Torvalds 已提交
4112 4113 4114 4115 4116 4117 4118 4119 4120
			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;
		}
			
	}
4121 4122 4123
	if (!plugged)
		md_wakeup_thread(mddev->thread);

L
Linus Torvalds 已提交
4124
	spin_lock_irq(&conf->device_lock);
4125
	remaining = raid5_dec_bi_phys_segments(bi);
4126 4127
	spin_unlock_irq(&conf->device_lock);
	if (remaining == 0) {
L
Linus Torvalds 已提交
4128

4129
		if ( rw == WRITE )
L
Linus Torvalds 已提交
4130
			md_write_end(mddev);
4131

4132
		bio_endio(bi, 0);
L
Linus Torvalds 已提交
4133 4134 4135
	}
}

4136
static sector_t raid5_size(struct mddev *mddev, sector_t sectors, int raid_disks);
D
Dan Williams 已提交
4137

4138
static sector_t reshape_request(struct mddev *mddev, sector_t sector_nr, int *skipped)
L
Linus Torvalds 已提交
4139
{
4140 4141 4142 4143 4144 4145 4146 4147 4148
	/* 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.
	 */
4149
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
4150
	struct stripe_head *sh;
4151
	sector_t first_sector, last_sector;
4152 4153 4154
	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;
4155 4156
	int i;
	int dd_idx;
4157
	sector_t writepos, readpos, safepos;
4158
	sector_t stripe_addr;
4159
	int reshape_sectors;
4160
	struct list_head stripes;
4161

4162 4163
	if (sector_nr == 0) {
		/* If restarting in the middle, skip the initial sectors */
4164
		if (mddev->reshape_backwards &&
4165 4166 4167
		    conf->reshape_progress < raid5_size(mddev, 0, 0)) {
			sector_nr = raid5_size(mddev, 0, 0)
				- conf->reshape_progress;
4168
		} else if (!mddev->reshape_backwards &&
4169 4170
			   conf->reshape_progress > 0)
			sector_nr = conf->reshape_progress;
4171
		sector_div(sector_nr, new_data_disks);
4172
		if (sector_nr) {
4173 4174
			mddev->curr_resync_completed = sector_nr;
			sysfs_notify(&mddev->kobj, NULL, "sync_completed");
4175 4176 4177
			*skipped = 1;
			return sector_nr;
		}
4178 4179
	}

4180 4181 4182 4183
	/* 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
	 */
4184 4185
	if (mddev->new_chunk_sectors > mddev->chunk_sectors)
		reshape_sectors = mddev->new_chunk_sectors;
4186
	else
4187
		reshape_sectors = mddev->chunk_sectors;
4188

4189 4190 4191 4192 4193
	/* We update the metadata at least every 10 seconds, or when
	 * the data about to be copied would over-write the source of
	 * the data at the front of the range.  i.e. one new_stripe
	 * along from reshape_progress new_maps to after where
	 * reshape_safe old_maps to
4194
	 */
4195
	writepos = conf->reshape_progress;
4196
	sector_div(writepos, new_data_disks);
4197 4198
	readpos = conf->reshape_progress;
	sector_div(readpos, data_disks);
4199
	safepos = conf->reshape_safe;
4200
	sector_div(safepos, data_disks);
4201
	if (mddev->reshape_backwards) {
4202
		writepos -= min_t(sector_t, reshape_sectors, writepos);
4203
		readpos += reshape_sectors;
4204
		safepos += reshape_sectors;
4205
	} else {
4206
		writepos += reshape_sectors;
4207 4208
		readpos -= min_t(sector_t, reshape_sectors, readpos);
		safepos -= min_t(sector_t, reshape_sectors, safepos);
4209
	}
4210

4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225
	/* Having calculated the 'writepos' possibly use it
	 * to set 'stripe_addr' which is where we will write to.
	 */
	if (mddev->reshape_backwards) {
		BUG_ON(conf->reshape_progress == 0);
		stripe_addr = writepos;
		BUG_ON((mddev->dev_sectors &
			~((sector_t)reshape_sectors - 1))
		       - reshape_sectors - stripe_addr
		       != sector_nr);
	} else {
		BUG_ON(writepos != sector_nr + reshape_sectors);
		stripe_addr = sector_nr;
	}

4226 4227 4228 4229
	/* '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.
4230 4231 4232 4233
	 * If there is a min_offset_diff, these are adjusted either by
	 * increasing the safepos/readpos if diff is negative, or
	 * increasing writepos if diff is positive.
	 * If 'readpos' is then behind 'writepos', there is no way that we can
4234 4235 4236 4237 4238 4239 4240 4241 4242 4243 4244 4245
	 * 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???
	 */
4246 4247 4248 4249 4250 4251
	if (conf->min_offset_diff < 0) {
		safepos += -conf->min_offset_diff;
		readpos += -conf->min_offset_diff;
	} else
		writepos += conf->min_offset_diff;

4252
	if ((mddev->reshape_backwards
4253 4254 4255
	     ? (safepos > writepos && readpos < writepos)
	     : (safepos < writepos && readpos > writepos)) ||
	    time_after(jiffies, conf->reshape_checkpoint + 10*HZ)) {
4256 4257 4258
		/* Cannot proceed until we've updated the superblock... */
		wait_event(conf->wait_for_overlap,
			   atomic_read(&conf->reshape_stripes)==0);
4259
		mddev->reshape_position = conf->reshape_progress;
4260
		mddev->curr_resync_completed = sector_nr;
4261
		conf->reshape_checkpoint = jiffies;
4262
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
4263
		md_wakeup_thread(mddev->thread);
4264
		wait_event(mddev->sb_wait, mddev->flags == 0 ||
4265 4266
			   kthread_should_stop());
		spin_lock_irq(&conf->device_lock);
4267
		conf->reshape_safe = mddev->reshape_position;
4268 4269
		spin_unlock_irq(&conf->device_lock);
		wake_up(&conf->wait_for_overlap);
4270
		sysfs_notify(&mddev->kobj, NULL, "sync_completed");
4271 4272
	}

4273
	INIT_LIST_HEAD(&stripes);
4274
	for (i = 0; i < reshape_sectors; i += STRIPE_SECTORS) {
4275
		int j;
4276
		int skipped_disk = 0;
4277
		sh = get_active_stripe(conf, stripe_addr+i, 0, 0, 1);
4278 4279 4280 4281 4282 4283 4284 4285 4286
		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;
4287
			if (conf->level == 6 &&
4288
			    j == sh->qd_idx)
4289
				continue;
4290
			s = compute_blocknr(sh, j, 0);
D
Dan Williams 已提交
4291
			if (s < raid5_size(mddev, 0, 0)) {
4292
				skipped_disk = 1;
4293 4294 4295 4296 4297 4298
				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);
		}
4299
		if (!skipped_disk) {
4300 4301 4302
			set_bit(STRIPE_EXPAND_READY, &sh->state);
			set_bit(STRIPE_HANDLE, &sh->state);
		}
4303
		list_add(&sh->lru, &stripes);
4304 4305
	}
	spin_lock_irq(&conf->device_lock);
4306
	if (mddev->reshape_backwards)
4307
		conf->reshape_progress -= reshape_sectors * new_data_disks;
4308
	else
4309
		conf->reshape_progress += reshape_sectors * new_data_disks;
4310 4311 4312 4313 4314 4315 4316
	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 =
4317
		raid5_compute_sector(conf, stripe_addr*(new_data_disks),
4318
				     1, &dd_idx, NULL);
4319
	last_sector =
4320
		raid5_compute_sector(conf, ((stripe_addr+reshape_sectors)
4321
					    * new_data_disks - 1),
4322
				     1, &dd_idx, NULL);
A
Andre Noll 已提交
4323 4324
	if (last_sector >= mddev->dev_sectors)
		last_sector = mddev->dev_sectors - 1;
4325
	while (first_sector <= last_sector) {
4326
		sh = get_active_stripe(conf, first_sector, 1, 0, 1);
4327 4328 4329 4330 4331
		set_bit(STRIPE_EXPAND_SOURCE, &sh->state);
		set_bit(STRIPE_HANDLE, &sh->state);
		release_stripe(sh);
		first_sector += STRIPE_SECTORS;
	}
4332 4333 4334 4335 4336 4337 4338 4339
	/* 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);
	}
4340 4341 4342
	/* If this takes us to the resync_max point where we have to pause,
	 * then we need to write out the superblock.
	 */
4343
	sector_nr += reshape_sectors;
4344 4345
	if ((sector_nr - mddev->curr_resync_completed) * 2
	    >= mddev->resync_max - mddev->curr_resync_completed) {
4346 4347 4348
		/* Cannot proceed until we've updated the superblock... */
		wait_event(conf->wait_for_overlap,
			   atomic_read(&conf->reshape_stripes) == 0);
4349
		mddev->reshape_position = conf->reshape_progress;
4350
		mddev->curr_resync_completed = sector_nr;
4351
		conf->reshape_checkpoint = jiffies;
4352 4353 4354 4355 4356 4357
		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);
4358
		conf->reshape_safe = mddev->reshape_position;
4359 4360
		spin_unlock_irq(&conf->device_lock);
		wake_up(&conf->wait_for_overlap);
4361
		sysfs_notify(&mddev->kobj, NULL, "sync_completed");
4362
	}
4363
	return reshape_sectors;
4364 4365 4366
}

/* FIXME go_faster isn't used */
4367
static inline sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipped, int go_faster)
4368
{
4369
	struct r5conf *conf = mddev->private;
4370
	struct stripe_head *sh;
A
Andre Noll 已提交
4371
	sector_t max_sector = mddev->dev_sectors;
N
NeilBrown 已提交
4372
	sector_t sync_blocks;
4373 4374
	int still_degraded = 0;
	int i;
L
Linus Torvalds 已提交
4375

4376
	if (sector_nr >= max_sector) {
L
Linus Torvalds 已提交
4377
		/* just being told to finish up .. nothing much to do */
4378

4379 4380 4381 4382
		if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) {
			end_reshape(conf);
			return 0;
		}
4383 4384 4385 4386

		if (mddev->curr_resync < max_sector) /* aborted */
			bitmap_end_sync(mddev->bitmap, mddev->curr_resync,
					&sync_blocks, 1);
4387
		else /* completed sync */
4388 4389 4390
			conf->fullsync = 0;
		bitmap_close_sync(mddev->bitmap);

L
Linus Torvalds 已提交
4391 4392
		return 0;
	}
4393

4394 4395 4396
	/* Allow raid5_quiesce to complete */
	wait_event(conf->wait_for_overlap, conf->quiesce != 2);

4397 4398
	if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
		return reshape_request(mddev, sector_nr, skipped);
4399

4400 4401 4402 4403 4404 4405
	/* 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
	 */

4406
	/* if there is too many failed drives and we are trying
L
Linus Torvalds 已提交
4407 4408 4409
	 * to resync, then assert that we are finished, because there is
	 * nothing we can do.
	 */
4410
	if (mddev->degraded >= conf->max_degraded &&
4411
	    test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
A
Andre Noll 已提交
4412
		sector_t rv = mddev->dev_sectors - sector_nr;
4413
		*skipped = 1;
L
Linus Torvalds 已提交
4414 4415
		return rv;
	}
4416
	if (!bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, 1) &&
4417
	    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
4418 4419 4420 4421 4422 4423
	    !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
4424

N
NeilBrown 已提交
4425 4426
	bitmap_cond_end_sync(mddev->bitmap, sector_nr);

4427
	sh = get_active_stripe(conf, sector_nr, 0, 1, 0);
L
Linus Torvalds 已提交
4428
	if (sh == NULL) {
4429
		sh = get_active_stripe(conf, sector_nr, 0, 0, 0);
L
Linus Torvalds 已提交
4430
		/* make sure we don't swamp the stripe cache if someone else
4431
		 * is trying to get access
L
Linus Torvalds 已提交
4432
		 */
4433
		schedule_timeout_uninterruptible(1);
L
Linus Torvalds 已提交
4434
	}
4435 4436 4437 4438
	/* 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.
	 */
4439
	for (i = 0; i < conf->raid_disks; i++)
4440 4441 4442 4443 4444
		if (conf->disks[i].rdev == NULL)
			still_degraded = 1;

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

4445
	set_bit(STRIPE_SYNC_REQUESTED, &sh->state);
L
Linus Torvalds 已提交
4446

4447
	handle_stripe(sh);
L
Linus Torvalds 已提交
4448 4449 4450 4451 4452
	release_stripe(sh);

	return STRIPE_SECTORS;
}

4453
static int  retry_aligned_read(struct r5conf *conf, struct bio *raid_bio)
4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465
{
	/* 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;
4466
	int dd_idx;
4467 4468 4469 4470 4471 4472
	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);
4473
	sector = raid5_compute_sector(conf, logical_sector,
4474
				      0, &dd_idx, NULL);
4475 4476 4477
	last_sector = raid_bio->bi_sector + (raid_bio->bi_size>>9);

	for (; logical_sector < last_sector;
4478 4479 4480
	     logical_sector += STRIPE_SECTORS,
		     sector += STRIPE_SECTORS,
		     scnt++) {
4481

4482
		if (scnt < raid5_bi_hw_segments(raid_bio))
4483 4484 4485
			/* already done this stripe */
			continue;

4486
		sh = get_active_stripe(conf, sector, 0, 1, 0);
4487 4488 4489

		if (!sh) {
			/* failed to get a stripe - must wait */
4490
			raid5_set_bi_hw_segments(raid_bio, scnt);
4491 4492 4493 4494
			conf->retry_read_aligned = raid_bio;
			return handled;
		}

4495 4496
		if (!add_stripe_bio(sh, raid_bio, dd_idx, 0)) {
			release_stripe(sh);
4497
			raid5_set_bi_hw_segments(raid_bio, scnt);
4498 4499 4500 4501
			conf->retry_read_aligned = raid_bio;
			return handled;
		}

4502
		handle_stripe(sh);
4503 4504 4505 4506
		release_stripe(sh);
		handled++;
	}
	spin_lock_irq(&conf->device_lock);
4507
	remaining = raid5_dec_bi_phys_segments(raid_bio);
4508
	spin_unlock_irq(&conf->device_lock);
4509 4510
	if (remaining == 0)
		bio_endio(raid_bio, 0);
4511 4512 4513 4514 4515 4516
	if (atomic_dec_and_test(&conf->active_aligned_reads))
		wake_up(&conf->wait_for_stripe);
	return handled;
}


L
Linus Torvalds 已提交
4517 4518 4519 4520 4521 4522 4523
/*
 * 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.
 */
4524
static void raid5d(struct mddev *mddev)
L
Linus Torvalds 已提交
4525 4526
{
	struct stripe_head *sh;
4527
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
4528
	int handled;
4529
	struct blk_plug plug;
L
Linus Torvalds 已提交
4530

4531
	pr_debug("+++ raid5d active\n");
L
Linus Torvalds 已提交
4532 4533 4534

	md_check_recovery(mddev);

4535
	blk_start_plug(&plug);
L
Linus Torvalds 已提交
4536 4537 4538
	handled = 0;
	spin_lock_irq(&conf->device_lock);
	while (1) {
4539
		struct bio *bio;
L
Linus Torvalds 已提交
4540

4541 4542 4543 4544
		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++;
4545
			spin_unlock_irq(&conf->device_lock);
4546
			bitmap_unplug(mddev->bitmap);
4547
			spin_lock_irq(&conf->device_lock);
4548
			conf->seq_write = conf->seq_flush;
4549 4550
			activate_bit_delay(conf);
		}
4551 4552
		if (atomic_read(&mddev->plug_cnt) == 0)
			raid5_activate_delayed(conf);
4553

4554 4555 4556 4557 4558 4559 4560 4561 4562 4563
		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++;
		}

4564 4565
		sh = __get_priority_stripe(conf);

4566
		if (!sh)
L
Linus Torvalds 已提交
4567 4568 4569 4570
			break;
		spin_unlock_irq(&conf->device_lock);
		
		handled++;
4571 4572 4573
		handle_stripe(sh);
		release_stripe(sh);
		cond_resched();
L
Linus Torvalds 已提交
4574

4575 4576 4577
		if (mddev->flags & ~(1<<MD_CHANGE_PENDING))
			md_check_recovery(mddev);

L
Linus Torvalds 已提交
4578 4579
		spin_lock_irq(&conf->device_lock);
	}
4580
	pr_debug("%d stripes handled\n", handled);
L
Linus Torvalds 已提交
4581 4582 4583

	spin_unlock_irq(&conf->device_lock);

4584
	async_tx_issue_pending_all();
4585
	blk_finish_plug(&plug);
L
Linus Torvalds 已提交
4586

4587
	pr_debug("--- raid5d inactive\n");
L
Linus Torvalds 已提交
4588 4589
}

4590
static ssize_t
4591
raid5_show_stripe_cache_size(struct mddev *mddev, char *page)
4592
{
4593
	struct r5conf *conf = mddev->private;
4594 4595 4596 4597
	if (conf)
		return sprintf(page, "%d\n", conf->max_nr_stripes);
	else
		return 0;
4598 4599
}

4600
int
4601
raid5_set_cache_size(struct mddev *mddev, int size)
4602
{
4603
	struct r5conf *conf = mddev->private;
4604 4605
	int err;

4606
	if (size <= 16 || size > 32768)
4607
		return -EINVAL;
4608
	while (size < conf->max_nr_stripes) {
4609 4610 4611 4612 4613
		if (drop_one_stripe(conf))
			conf->max_nr_stripes--;
		else
			break;
	}
4614 4615 4616
	err = md_allow_write(mddev);
	if (err)
		return err;
4617
	while (size > conf->max_nr_stripes) {
4618 4619 4620 4621
		if (grow_one_stripe(conf))
			conf->max_nr_stripes++;
		else break;
	}
4622 4623 4624 4625 4626
	return 0;
}
EXPORT_SYMBOL(raid5_set_cache_size);

static ssize_t
4627
raid5_store_stripe_cache_size(struct mddev *mddev, const char *page, size_t len)
4628
{
4629
	struct r5conf *conf = mddev->private;
4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642
	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;
4643 4644
	return len;
}
4645

4646 4647 4648 4649
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);
4650

4651
static ssize_t
4652
raid5_show_preread_threshold(struct mddev *mddev, char *page)
4653
{
4654
	struct r5conf *conf = mddev->private;
4655 4656 4657 4658 4659 4660 4661
	if (conf)
		return sprintf(page, "%d\n", conf->bypass_threshold);
	else
		return 0;
}

static ssize_t
4662
raid5_store_preread_threshold(struct mddev *mddev, const char *page, size_t len)
4663
{
4664
	struct r5conf *conf = mddev->private;
4665
	unsigned long new;
4666 4667 4668 4669 4670
	if (len >= PAGE_SIZE)
		return -EINVAL;
	if (!conf)
		return -ENODEV;

4671
	if (strict_strtoul(page, 10, &new))
4672
		return -EINVAL;
4673
	if (new > conf->max_nr_stripes)
4674 4675 4676 4677 4678 4679 4680 4681 4682 4683 4684
		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);

4685
static ssize_t
4686
stripe_cache_active_show(struct mddev *mddev, char *page)
4687
{
4688
	struct r5conf *conf = mddev->private;
4689 4690 4691 4692
	if (conf)
		return sprintf(page, "%d\n", atomic_read(&conf->active_stripes));
	else
		return 0;
4693 4694
}

4695 4696
static struct md_sysfs_entry
raid5_stripecache_active = __ATTR_RO(stripe_cache_active);
4697

4698
static struct attribute *raid5_attrs[] =  {
4699 4700
	&raid5_stripecache_size.attr,
	&raid5_stripecache_active.attr,
4701
	&raid5_preread_bypass_threshold.attr,
4702 4703
	NULL,
};
4704 4705 4706
static struct attribute_group raid5_attrs_group = {
	.name = NULL,
	.attrs = raid5_attrs,
4707 4708
};

4709
static sector_t
4710
raid5_size(struct mddev *mddev, sector_t sectors, int raid_disks)
4711
{
4712
	struct r5conf *conf = mddev->private;
4713 4714 4715

	if (!sectors)
		sectors = mddev->dev_sectors;
4716
	if (!raid_disks)
4717
		/* size is defined by the smallest of previous and new size */
4718
		raid_disks = min(conf->raid_disks, conf->previous_raid_disks);
4719

4720
	sectors &= ~((sector_t)mddev->chunk_sectors - 1);
4721
	sectors &= ~((sector_t)mddev->new_chunk_sectors - 1);
4722 4723 4724
	return sectors * (raid_disks - conf->max_degraded);
}

4725
static void raid5_free_percpu(struct r5conf *conf)
4726 4727 4728 4729 4730 4731 4732 4733 4734 4735 4736
{
	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);
4737
		kfree(percpu->scribble);
4738 4739 4740 4741 4742 4743 4744 4745 4746
	}
#ifdef CONFIG_HOTPLUG_CPU
	unregister_cpu_notifier(&conf->cpu_notify);
#endif
	put_online_cpus();

	free_percpu(conf->percpu);
}

4747
static void free_conf(struct r5conf *conf)
4748 4749
{
	shrink_stripes(conf);
4750
	raid5_free_percpu(conf);
4751 4752 4753 4754 4755
	kfree(conf->disks);
	kfree(conf->stripe_hashtbl);
	kfree(conf);
}

4756 4757 4758 4759
#ifdef CONFIG_HOTPLUG_CPU
static int raid456_cpu_notify(struct notifier_block *nfb, unsigned long action,
			      void *hcpu)
{
4760
	struct r5conf *conf = container_of(nfb, struct r5conf, cpu_notify);
4761 4762 4763 4764 4765 4766
	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:
4767
		if (conf->level == 6 && !percpu->spare_page)
4768
			percpu->spare_page = alloc_page(GFP_KERNEL);
4769 4770 4771 4772 4773 4774 4775
		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);
4776 4777
			pr_err("%s: failed memory allocation for cpu%ld\n",
			       __func__, cpu);
4778
			return notifier_from_errno(-ENOMEM);
4779 4780 4781 4782 4783
		}
		break;
	case CPU_DEAD:
	case CPU_DEAD_FROZEN:
		safe_put_page(percpu->spare_page);
4784
		kfree(percpu->scribble);
4785
		percpu->spare_page = NULL;
4786
		percpu->scribble = NULL;
4787 4788 4789 4790 4791 4792 4793 4794
		break;
	default:
		break;
	}
	return NOTIFY_OK;
}
#endif

4795
static int raid5_alloc_percpu(struct r5conf *conf)
4796 4797 4798
{
	unsigned long cpu;
	struct page *spare_page;
4799
	struct raid5_percpu __percpu *allcpus;
4800
	void *scribble;
4801 4802 4803 4804 4805 4806 4807 4808 4809 4810
	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) {
4811 4812 4813 4814 4815 4816 4817 4818
		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;
		}
4819
		scribble = kmalloc(conf->scribble_len, GFP_KERNEL);
4820
		if (!scribble) {
4821 4822 4823
			err = -ENOMEM;
			break;
		}
4824
		per_cpu_ptr(conf->percpu, cpu)->scribble = scribble;
4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836
	}
#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;
}

4837
static struct r5conf *setup_conf(struct mddev *mddev)
L
Linus Torvalds 已提交
4838
{
4839
	struct r5conf *conf;
4840
	int raid_disk, memory, max_disks;
4841
	struct md_rdev *rdev;
L
Linus Torvalds 已提交
4842
	struct disk_info *disk;
4843
	char pers_name[6];
L
Linus Torvalds 已提交
4844

N
NeilBrown 已提交
4845 4846 4847
	if (mddev->new_level != 5
	    && mddev->new_level != 4
	    && mddev->new_level != 6) {
4848
		printk(KERN_ERR "md/raid:%s: raid level not set to 4/5/6 (%d)\n",
N
NeilBrown 已提交
4849 4850
		       mdname(mddev), mddev->new_level);
		return ERR_PTR(-EIO);
L
Linus Torvalds 已提交
4851
	}
N
NeilBrown 已提交
4852 4853 4854 4855
	if ((mddev->new_level == 5
	     && !algorithm_valid_raid5(mddev->new_layout)) ||
	    (mddev->new_level == 6
	     && !algorithm_valid_raid6(mddev->new_layout))) {
4856
		printk(KERN_ERR "md/raid:%s: layout %d not supported\n",
N
NeilBrown 已提交
4857 4858
		       mdname(mddev), mddev->new_layout);
		return ERR_PTR(-EIO);
4859
	}
N
NeilBrown 已提交
4860
	if (mddev->new_level == 6 && mddev->raid_disks < 4) {
4861
		printk(KERN_ERR "md/raid:%s: not enough configured devices (%d, minimum 4)\n",
N
NeilBrown 已提交
4862 4863
		       mdname(mddev), mddev->raid_disks);
		return ERR_PTR(-EINVAL);
4864 4865
	}

4866 4867 4868
	if (!mddev->new_chunk_sectors ||
	    (mddev->new_chunk_sectors << 9) % PAGE_SIZE ||
	    !is_power_of_2(mddev->new_chunk_sectors)) {
4869 4870
		printk(KERN_ERR "md/raid:%s: invalid chunk size %d\n",
		       mdname(mddev), mddev->new_chunk_sectors << 9);
N
NeilBrown 已提交
4871
		return ERR_PTR(-EINVAL);
4872 4873
	}

4874
	conf = kzalloc(sizeof(struct r5conf), GFP_KERNEL);
N
NeilBrown 已提交
4875
	if (conf == NULL)
L
Linus Torvalds 已提交
4876
		goto abort;
4877 4878 4879 4880 4881 4882 4883 4884 4885 4886 4887 4888
	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;
4889
	conf->recovery_disabled = mddev->recovery_disabled - 1;
N
NeilBrown 已提交
4890 4891 4892 4893 4894

	conf->raid_disks = mddev->raid_disks;
	if (mddev->reshape_position == MaxSector)
		conf->previous_raid_disks = mddev->raid_disks;
	else
4895
		conf->previous_raid_disks = mddev->raid_disks - mddev->delta_disks;
4896 4897
	max_disks = max(conf->raid_disks, conf->previous_raid_disks);
	conf->scribble_len = scribble_len(max_disks);
4898

4899
	conf->disks = kzalloc(max_disks * sizeof(struct disk_info),
4900 4901 4902
			      GFP_KERNEL);
	if (!conf->disks)
		goto abort;
4903

L
Linus Torvalds 已提交
4904 4905
	conf->mddev = mddev;

4906
	if ((conf->stripe_hashtbl = kzalloc(PAGE_SIZE, GFP_KERNEL)) == NULL)
L
Linus Torvalds 已提交
4907 4908
		goto abort;

4909 4910 4911 4912
	conf->level = mddev->new_level;
	if (raid5_alloc_percpu(conf) != 0)
		goto abort;

4913
	pr_debug("raid456: run(%s) called.\n", mdname(mddev));
L
Linus Torvalds 已提交
4914

N
NeilBrown 已提交
4915
	rdev_for_each(rdev, mddev) {
L
Linus Torvalds 已提交
4916
		raid_disk = rdev->raid_disk;
4917
		if (raid_disk >= max_disks
L
Linus Torvalds 已提交
4918 4919 4920 4921
		    || raid_disk < 0)
			continue;
		disk = conf->disks + raid_disk;

4922 4923 4924 4925 4926 4927 4928 4929 4930
		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
4931

4932
		if (test_bit(In_sync, &rdev->flags)) {
L
Linus Torvalds 已提交
4933
			char b[BDEVNAME_SIZE];
4934 4935 4936
			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
4937
		} else if (rdev->saved_raid_disk != raid_disk)
4938 4939
			/* Cannot rely on bitmap to complete recovery */
			conf->fullsync = 1;
L
Linus Torvalds 已提交
4940 4941
	}

4942
	conf->chunk_sectors = mddev->new_chunk_sectors;
N
NeilBrown 已提交
4943
	conf->level = mddev->new_level;
4944 4945 4946 4947
	if (conf->level == 6)
		conf->max_degraded = 2;
	else
		conf->max_degraded = 1;
N
NeilBrown 已提交
4948
	conf->algorithm = mddev->new_layout;
L
Linus Torvalds 已提交
4949
	conf->max_nr_stripes = NR_STRIPES;
4950
	conf->reshape_progress = mddev->reshape_position;
4951
	if (conf->reshape_progress != MaxSector) {
4952
		conf->prev_chunk_sectors = mddev->chunk_sectors;
4953 4954
		conf->prev_algo = mddev->layout;
	}
L
Linus Torvalds 已提交
4955

N
NeilBrown 已提交
4956
	memory = conf->max_nr_stripes * (sizeof(struct stripe_head) +
4957
		 max_disks * ((sizeof(struct bio) + PAGE_SIZE))) / 1024;
N
NeilBrown 已提交
4958 4959
	if (grow_stripes(conf, conf->max_nr_stripes)) {
		printk(KERN_ERR
4960 4961
		       "md/raid:%s: couldn't allocate %dkB for buffers\n",
		       mdname(mddev), memory);
N
NeilBrown 已提交
4962 4963
		goto abort;
	} else
4964 4965
		printk(KERN_INFO "md/raid:%s: allocated %dkB\n",
		       mdname(mddev), memory);
L
Linus Torvalds 已提交
4966

4967 4968
	sprintf(pers_name, "raid%d", mddev->new_level);
	conf->thread = md_register_thread(raid5d, mddev, pers_name);
N
NeilBrown 已提交
4969 4970
	if (!conf->thread) {
		printk(KERN_ERR
4971
		       "md/raid:%s: couldn't allocate thread.\n",
N
NeilBrown 已提交
4972
		       mdname(mddev));
4973 4974
		goto abort;
	}
N
NeilBrown 已提交
4975 4976 4977 4978 4979

	return conf;

 abort:
	if (conf) {
4980
		free_conf(conf);
N
NeilBrown 已提交
4981 4982 4983 4984 4985
		return ERR_PTR(-EIO);
	} else
		return ERR_PTR(-ENOMEM);
}

4986 4987 4988 4989 4990 4991 4992 4993 4994 4995 4996 4997 4998 4999 5000 5001 5002 5003 5004 5005 5006 5007 5008 5009 5010 5011 5012

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;
}

5013
static int run(struct mddev *mddev)
N
NeilBrown 已提交
5014
{
5015
	struct r5conf *conf;
5016
	int working_disks = 0;
5017
	int dirty_parity_disks = 0;
5018
	struct md_rdev *rdev;
5019
	sector_t reshape_offset = 0;
5020
	int i;
5021 5022
	long long min_offset_diff = 0;
	int first = 1;
N
NeilBrown 已提交
5023

5024
	if (mddev->recovery_cp != MaxSector)
5025
		printk(KERN_NOTICE "md/raid:%s: not clean"
5026 5027
		       " -- starting background reconstruction\n",
		       mdname(mddev));
5028 5029 5030 5031 5032 5033 5034 5035 5036 5037 5038 5039 5040 5041 5042 5043 5044

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

N
NeilBrown 已提交
5045 5046
	if (mddev->reshape_position != MaxSector) {
		/* Check that we can continue the reshape.
5047 5048 5049 5050 5051 5052 5053 5054 5055 5056
		 * Difficulties arise if the stripe we would write to
		 * next is at or after the stripe we would read from next.
		 * For a reshape that changes the number of devices, this
		 * is only possible for a very short time, and mdadm makes
		 * sure that time appears to have past before assembling
		 * the array.  So we fail if that time hasn't passed.
		 * For a reshape that keeps the number of devices the same
		 * mdadm must be monitoring the reshape can keeping the
		 * critical areas read-only and backed up.  It will start
		 * the array in read-only mode, so we check for that.
N
NeilBrown 已提交
5057 5058 5059
		 */
		sector_t here_new, here_old;
		int old_disks;
5060
		int max_degraded = (mddev->level == 6 ? 2 : 1);
N
NeilBrown 已提交
5061

5062
		if (mddev->new_level != mddev->level) {
5063
			printk(KERN_ERR "md/raid:%s: unsupported reshape "
N
NeilBrown 已提交
5064 5065 5066 5067 5068 5069 5070 5071 5072 5073
			       "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;
5074
		if (sector_div(here_new, mddev->new_chunk_sectors *
N
NeilBrown 已提交
5075
			       (mddev->raid_disks - max_degraded))) {
5076 5077
			printk(KERN_ERR "md/raid:%s: reshape_position not "
			       "on a stripe boundary\n", mdname(mddev));
N
NeilBrown 已提交
5078 5079
			return -EINVAL;
		}
5080
		reshape_offset = here_new * mddev->new_chunk_sectors;
N
NeilBrown 已提交
5081 5082
		/* here_new is the stripe we will write to */
		here_old = mddev->reshape_position;
5083
		sector_div(here_old, mddev->chunk_sectors *
N
NeilBrown 已提交
5084 5085 5086
			   (old_disks-max_degraded));
		/* here_old is the first stripe that we might need to read
		 * from */
5087
		if (mddev->delta_disks == 0) {
5088 5089 5090 5091 5092 5093
			if ((here_new * mddev->new_chunk_sectors !=
			     here_old * mddev->chunk_sectors)) {
				printk(KERN_ERR "md/raid:%s: reshape position is"
				       " confused - aborting\n", mdname(mddev));
				return -EINVAL;
			}
5094
			/* We cannot be sure it is safe to start an in-place
5095
			 * reshape.  It is only safe if user-space is monitoring
5096 5097 5098 5099 5100
			 * 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.
			 */
5101 5102 5103 5104 5105 5106 5107
			if (abs(min_offset_diff) >= mddev->chunk_sectors &&
			    abs(min_offset_diff) >= mddev->new_chunk_sectors)
				/* not really in-place - so OK */;
			else if (mddev->ro == 0) {
				printk(KERN_ERR "md/raid:%s: in-place reshape "
				       "must be started in read-only mode "
				       "- aborting\n",
5108
				       mdname(mddev));
5109 5110
				return -EINVAL;
			}
5111
		} else if (mddev->reshape_backwards
5112
		    ? (here_new * mddev->new_chunk_sectors + min_offset_diff <=
5113 5114
		       here_old * mddev->chunk_sectors)
		    : (here_new * mddev->new_chunk_sectors >=
5115
		       here_old * mddev->chunk_sectors + (-min_offset_diff))) {
N
NeilBrown 已提交
5116
			/* Reading from the same stripe as writing to - bad */
5117 5118 5119
			printk(KERN_ERR "md/raid:%s: reshape_position too early for "
			       "auto-recovery - aborting.\n",
			       mdname(mddev));
N
NeilBrown 已提交
5120 5121
			return -EINVAL;
		}
5122 5123
		printk(KERN_INFO "md/raid:%s: reshape will continue\n",
		       mdname(mddev));
N
NeilBrown 已提交
5124 5125 5126 5127
		/* OK, we should be able to continue; */
	} else {
		BUG_ON(mddev->level != mddev->new_level);
		BUG_ON(mddev->layout != mddev->new_layout);
5128
		BUG_ON(mddev->chunk_sectors != mddev->new_chunk_sectors);
N
NeilBrown 已提交
5129
		BUG_ON(mddev->delta_disks != 0);
L
Linus Torvalds 已提交
5130
	}
N
NeilBrown 已提交
5131

5132 5133 5134 5135 5136
	if (mddev->private == NULL)
		conf = setup_conf(mddev);
	else
		conf = mddev->private;

N
NeilBrown 已提交
5137 5138 5139
	if (IS_ERR(conf))
		return PTR_ERR(conf);

5140
	conf->min_offset_diff = min_offset_diff;
N
NeilBrown 已提交
5141 5142 5143 5144
	mddev->thread = conf->thread;
	conf->thread = NULL;
	mddev->private = conf;

5145 5146 5147 5148 5149 5150 5151 5152 5153 5154 5155
	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)
5156
			continue;
5157 5158 5159 5160 5161 5162 5163
		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;
		}
5164
		if (test_bit(In_sync, &rdev->flags)) {
N
NeilBrown 已提交
5165
			working_disks++;
5166 5167
			continue;
		}
5168 5169 5170 5171 5172 5173 5174 5175 5176 5177 5178 5179 5180 5181 5182 5183 5184 5185 5186 5187 5188 5189 5190 5191 5192 5193 5194 5195
		/* 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
5196

5197 5198 5199
	/*
	 * 0 for a fully functional array, 1 or 2 for a degraded array.
	 */
5200
	mddev->degraded = calc_degraded(conf);
N
NeilBrown 已提交
5201

5202
	if (has_failed(conf)) {
5203
		printk(KERN_ERR "md/raid:%s: not enough operational devices"
L
Linus Torvalds 已提交
5204
			" (%d/%d failed)\n",
5205
			mdname(mddev), mddev->degraded, conf->raid_disks);
L
Linus Torvalds 已提交
5206 5207 5208
		goto abort;
	}

N
NeilBrown 已提交
5209
	/* device size must be a multiple of chunk size */
5210
	mddev->dev_sectors &= ~(mddev->chunk_sectors - 1);
N
NeilBrown 已提交
5211 5212
	mddev->resync_max_sectors = mddev->dev_sectors;

5213
	if (mddev->degraded > dirty_parity_disks &&
L
Linus Torvalds 已提交
5214
	    mddev->recovery_cp != MaxSector) {
5215 5216
		if (mddev->ok_start_degraded)
			printk(KERN_WARNING
5217 5218
			       "md/raid:%s: starting dirty degraded array"
			       " - data corruption possible.\n",
5219 5220 5221
			       mdname(mddev));
		else {
			printk(KERN_ERR
5222
			       "md/raid:%s: cannot start dirty degraded array.\n",
5223 5224 5225
			       mdname(mddev));
			goto abort;
		}
L
Linus Torvalds 已提交
5226 5227 5228
	}

	if (mddev->degraded == 0)
5229 5230
		printk(KERN_INFO "md/raid:%s: raid level %d active with %d out of %d"
		       " devices, algorithm %d\n", mdname(mddev), conf->level,
5231 5232
		       mddev->raid_disks-mddev->degraded, mddev->raid_disks,
		       mddev->new_layout);
L
Linus Torvalds 已提交
5233
	else
5234 5235 5236 5237 5238
		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
5239 5240 5241

	print_raid5_conf(conf);

5242 5243
	if (conf->reshape_progress != MaxSector) {
		conf->reshape_safe = conf->reshape_progress;
5244 5245 5246 5247 5248 5249
		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,
5250
							"reshape");
5251 5252
	}

L
Linus Torvalds 已提交
5253 5254

	/* Ok, everything is just fine now */
5255 5256
	if (mddev->to_remove == &raid5_attrs_group)
		mddev->to_remove = NULL;
N
NeilBrown 已提交
5257 5258
	else if (mddev->kobj.sd &&
	    sysfs_create_group(&mddev->kobj, &raid5_attrs_group))
5259
		printk(KERN_WARNING
5260
		       "raid5: failed to create sysfs attributes for %s\n",
5261
		       mdname(mddev));
5262
	md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
5263

5264
	if (mddev->queue) {
5265
		int chunk_size;
5266 5267 5268 5269 5270 5271 5272 5273 5274
		/* 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
5275

5276
		blk_queue_merge_bvec(mddev->queue, raid5_mergeable_bvec);
5277

N
NeilBrown 已提交
5278 5279
		mddev->queue->backing_dev_info.congested_data = mddev;
		mddev->queue->backing_dev_info.congested_fn = raid5_congested;
5280

5281 5282 5283 5284
		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));
5285

5286
		rdev_for_each(rdev, mddev) {
5287 5288
			disk_stack_limits(mddev->gendisk, rdev->bdev,
					  rdev->data_offset << 9);
5289 5290 5291
			disk_stack_limits(mddev->gendisk, rdev->bdev,
					  rdev->new_data_offset << 9);
		}
5292
	}
5293

L
Linus Torvalds 已提交
5294 5295
	return 0;
abort:
5296
	md_unregister_thread(&mddev->thread);
N
NeilBrown 已提交
5297 5298
	print_raid5_conf(conf);
	free_conf(conf);
L
Linus Torvalds 已提交
5299
	mddev->private = NULL;
5300
	printk(KERN_ALERT "md/raid:%s: failed to run raid set.\n", mdname(mddev));
L
Linus Torvalds 已提交
5301 5302 5303
	return -EIO;
}

5304
static int stop(struct mddev *mddev)
L
Linus Torvalds 已提交
5305
{
5306
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5307

5308
	md_unregister_thread(&mddev->thread);
N
NeilBrown 已提交
5309 5310
	if (mddev->queue)
		mddev->queue->backing_dev_info.congested_fn = NULL;
5311
	free_conf(conf);
5312 5313
	mddev->private = NULL;
	mddev->to_remove = &raid5_attrs_group;
L
Linus Torvalds 已提交
5314 5315 5316
	return 0;
}

5317
static void status(struct seq_file *seq, struct mddev *mddev)
L
Linus Torvalds 已提交
5318
{
5319
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5320 5321
	int i;

5322 5323
	seq_printf(seq, " level %d, %dk chunk, algorithm %d", mddev->level,
		mddev->chunk_sectors / 2, mddev->layout);
5324
	seq_printf (seq, " [%d/%d] [", conf->raid_disks, conf->raid_disks - mddev->degraded);
L
Linus Torvalds 已提交
5325 5326 5327
	for (i = 0; i < conf->raid_disks; i++)
		seq_printf (seq, "%s",
			       conf->disks[i].rdev &&
5328
			       test_bit(In_sync, &conf->disks[i].rdev->flags) ? "U" : "_");
L
Linus Torvalds 已提交
5329 5330 5331
	seq_printf (seq, "]");
}

5332
static void print_raid5_conf (struct r5conf *conf)
L
Linus Torvalds 已提交
5333 5334 5335 5336
{
	int i;
	struct disk_info *tmp;

5337
	printk(KERN_DEBUG "RAID conf printout:\n");
L
Linus Torvalds 已提交
5338 5339 5340 5341
	if (!conf) {
		printk("(conf==NULL)\n");
		return;
	}
5342 5343 5344
	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
5345 5346 5347 5348 5349

	for (i = 0; i < conf->raid_disks; i++) {
		char b[BDEVNAME_SIZE];
		tmp = conf->disks + i;
		if (tmp->rdev)
5350 5351 5352
			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
5353 5354 5355
	}
}

5356
static int raid5_spare_active(struct mddev *mddev)
L
Linus Torvalds 已提交
5357 5358
{
	int i;
5359
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5360
	struct disk_info *tmp;
5361 5362
	int count = 0;
	unsigned long flags;
L
Linus Torvalds 已提交
5363 5364 5365

	for (i = 0; i < conf->raid_disks; i++) {
		tmp = conf->disks + i;
5366 5367 5368 5369 5370 5371 5372 5373 5374 5375 5376 5377 5378 5379 5380 5381 5382 5383 5384
		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
5385
		    && tmp->rdev->recovery_offset == MaxSector
5386
		    && !test_bit(Faulty, &tmp->rdev->flags)
5387
		    && !test_and_set_bit(In_sync, &tmp->rdev->flags)) {
5388
			count++;
5389
			sysfs_notify_dirent_safe(tmp->rdev->sysfs_state);
L
Linus Torvalds 已提交
5390 5391
		}
	}
5392
	spin_lock_irqsave(&conf->device_lock, flags);
5393
	mddev->degraded = calc_degraded(conf);
5394
	spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
5395
	print_raid5_conf(conf);
5396
	return count;
L
Linus Torvalds 已提交
5397 5398
}

5399
static int raid5_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
5400
{
5401
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5402
	int err = 0;
5403
	int number = rdev->raid_disk;
5404
	struct md_rdev **rdevp;
L
Linus Torvalds 已提交
5405 5406 5407
	struct disk_info *p = conf->disks + number;

	print_raid5_conf(conf);
5408 5409 5410 5411 5412 5413 5414 5415 5416 5417 5418 5419 5420 5421 5422 5423 5424 5425 5426 5427 5428 5429
	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) &&
5430
	    (!p->replacement || p->replacement == rdev) &&
5431 5432 5433 5434 5435 5436 5437 5438 5439 5440
	    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;
5441 5442 5443 5444 5445 5446 5447 5448 5449 5450 5451 5452 5453 5454
	} 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
5455 5456 5457 5458 5459 5460
abort:

	print_raid5_conf(conf);
	return err;
}

5461
static int raid5_add_disk(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
5462
{
5463
	struct r5conf *conf = mddev->private;
5464
	int err = -EEXIST;
L
Linus Torvalds 已提交
5465 5466
	int disk;
	struct disk_info *p;
5467 5468
	int first = 0;
	int last = conf->raid_disks - 1;
L
Linus Torvalds 已提交
5469

5470 5471 5472
	if (mddev->recovery_disabled == conf->recovery_disabled)
		return -EBUSY;

N
NeilBrown 已提交
5473
	if (rdev->saved_raid_disk < 0 && has_failed(conf))
L
Linus Torvalds 已提交
5474
		/* no point adding a device */
5475
		return -EINVAL;
L
Linus Torvalds 已提交
5476

5477 5478
	if (rdev->raid_disk >= 0)
		first = last = rdev->raid_disk;
L
Linus Torvalds 已提交
5479 5480

	/*
5481 5482
	 * find the disk ... but prefer rdev->saved_raid_disk
	 * if possible.
L
Linus Torvalds 已提交
5483
	 */
5484
	if (rdev->saved_raid_disk >= 0 &&
5485
	    rdev->saved_raid_disk >= first &&
5486
	    conf->disks[rdev->saved_raid_disk].rdev == NULL)
5487 5488 5489
		first = rdev->saved_raid_disk;

	for (disk = first; disk <= last; disk++) {
5490 5491
		p = conf->disks + disk;
		if (p->rdev == NULL) {
5492
			clear_bit(In_sync, &rdev->flags);
L
Linus Torvalds 已提交
5493
			rdev->raid_disk = disk;
5494
			err = 0;
5495 5496
			if (rdev->saved_raid_disk != disk)
				conf->fullsync = 1;
5497
			rcu_assign_pointer(p->rdev, rdev);
5498
			goto out;
L
Linus Torvalds 已提交
5499
		}
5500 5501 5502
	}
	for (disk = first; disk <= last; disk++) {
		p = conf->disks + disk;
5503 5504 5505 5506 5507 5508 5509 5510 5511 5512 5513
		if (test_bit(WantReplacement, &p->rdev->flags) &&
		    p->replacement == NULL) {
			clear_bit(In_sync, &rdev->flags);
			set_bit(Replacement, &rdev->flags);
			rdev->raid_disk = disk;
			err = 0;
			conf->fullsync = 1;
			rcu_assign_pointer(p->replacement, rdev);
			break;
		}
	}
5514
out:
L
Linus Torvalds 已提交
5515
	print_raid5_conf(conf);
5516
	return err;
L
Linus Torvalds 已提交
5517 5518
}

5519
static int raid5_resize(struct mddev *mddev, sector_t sectors)
L
Linus Torvalds 已提交
5520 5521 5522 5523 5524 5525 5526 5527
{
	/* 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.
	 */
5528
	sector_t newsize;
5529
	sectors &= ~((sector_t)mddev->chunk_sectors - 1);
5530 5531 5532
	newsize = raid5_size(mddev, sectors, mddev->raid_disks);
	if (mddev->external_size &&
	    mddev->array_sectors > newsize)
D
Dan Williams 已提交
5533
		return -EINVAL;
5534 5535 5536 5537 5538 5539
	if (mddev->bitmap) {
		int ret = bitmap_resize(mddev->bitmap, sectors, 0, 0);
		if (ret)
			return ret;
	}
	md_set_array_sectors(mddev, newsize);
5540
	set_capacity(mddev->gendisk, mddev->array_sectors);
5541
	revalidate_disk(mddev->gendisk);
5542 5543
	if (sectors > mddev->dev_sectors &&
	    mddev->recovery_cp > mddev->dev_sectors) {
A
Andre Noll 已提交
5544
		mddev->recovery_cp = mddev->dev_sectors;
L
Linus Torvalds 已提交
5545 5546
		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
	}
A
Andre Noll 已提交
5547
	mddev->dev_sectors = sectors;
5548
	mddev->resync_max_sectors = sectors;
L
Linus Torvalds 已提交
5549 5550 5551
	return 0;
}

5552
static int check_stripe_cache(struct mddev *mddev)
5553 5554 5555 5556 5557 5558 5559 5560 5561
{
	/* 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.
	 */
5562
	struct r5conf *conf = mddev->private;
5563 5564 5565 5566
	if (((mddev->chunk_sectors << 9) / STRIPE_SIZE) * 4
	    > conf->max_nr_stripes ||
	    ((mddev->new_chunk_sectors << 9) / STRIPE_SIZE) * 4
	    > conf->max_nr_stripes) {
5567 5568
		printk(KERN_WARNING "md/raid:%s: reshape: not enough stripes.  Needed %lu\n",
		       mdname(mddev),
5569 5570 5571 5572 5573 5574 5575
		       ((max(mddev->chunk_sectors, mddev->new_chunk_sectors) << 9)
			/ STRIPE_SIZE)*4);
		return 0;
	}
	return 1;
}

5576
static int check_reshape(struct mddev *mddev)
5577
{
5578
	struct r5conf *conf = mddev->private;
5579

5580 5581
	if (mddev->delta_disks == 0 &&
	    mddev->new_layout == mddev->layout &&
5582
	    mddev->new_chunk_sectors == mddev->chunk_sectors)
5583
		return 0; /* nothing to do */
5584
	if (has_failed(conf))
5585 5586 5587 5588 5589 5590 5591 5592 5593 5594 5595 5596 5597
		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;
	}
5598

5599
	if (!check_stripe_cache(mddev))
5600 5601
		return -ENOSPC;

5602
	return resize_stripes(conf, conf->raid_disks + mddev->delta_disks);
5603 5604
}

5605
static int raid5_start_reshape(struct mddev *mddev)
5606
{
5607
	struct r5conf *conf = mddev->private;
5608
	struct md_rdev *rdev;
5609
	int spares = 0;
5610
	unsigned long flags;
5611

5612
	if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
5613 5614
		return -EBUSY;

5615 5616 5617
	if (!check_stripe_cache(mddev))
		return -ENOSPC;

5618 5619 5620
	if (has_failed(conf))
		return -EINVAL;

5621
	rdev_for_each(rdev, mddev) {
5622 5623
		if (!test_bit(In_sync, &rdev->flags)
		    && !test_bit(Faulty, &rdev->flags))
5624
			spares++;
5625
	}
5626

5627
	if (spares - mddev->degraded < mddev->delta_disks - conf->max_degraded)
5628 5629 5630 5631 5632
		/* Not enough devices even to make a degraded array
		 * of that size
		 */
		return -EINVAL;

5633 5634 5635 5636 5637 5638
	/* 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) {
5639
		printk(KERN_ERR "md/raid:%s: array size must be reduced "
5640 5641 5642 5643
		       "before number of disks\n", mdname(mddev));
		return -EINVAL;
	}

5644
	atomic_set(&conf->reshape_stripes, 0);
5645 5646
	spin_lock_irq(&conf->device_lock);
	conf->previous_raid_disks = conf->raid_disks;
5647
	conf->raid_disks += mddev->delta_disks;
5648 5649
	conf->prev_chunk_sectors = conf->chunk_sectors;
	conf->chunk_sectors = mddev->new_chunk_sectors;
5650 5651
	conf->prev_algo = conf->algorithm;
	conf->algorithm = mddev->new_layout;
5652 5653 5654 5655 5656
	conf->generation++;
	/* Code that selects data_offset needs to see the generation update
	 * if reshape_progress has been set - so a memory barrier needed.
	 */
	smp_mb();
5657
	if (mddev->reshape_backwards)
5658 5659 5660 5661
		conf->reshape_progress = raid5_size(mddev, 0, 0);
	else
		conf->reshape_progress = 0;
	conf->reshape_safe = conf->reshape_progress;
5662 5663 5664 5665
	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.
5666 5667 5668 5669
	 * 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.
5670
	 */
5671
	if (mddev->delta_disks >= 0) {
N
NeilBrown 已提交
5672
		rdev_for_each(rdev, mddev)
5673 5674 5675 5676
			if (rdev->raid_disk < 0 &&
			    !test_bit(Faulty, &rdev->flags)) {
				if (raid5_add_disk(mddev, rdev) == 0) {
					if (rdev->raid_disk
5677
					    >= conf->previous_raid_disks)
5678
						set_bit(In_sync, &rdev->flags);
5679
					else
5680
						rdev->recovery_offset = 0;
5681 5682

					if (sysfs_link_rdev(mddev, rdev))
5683
						/* Failure here is OK */;
5684
				}
5685 5686 5687 5688 5689
			} 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);
			}
5690

5691 5692 5693 5694
		/* When a reshape changes the number of devices,
		 * ->degraded is measured against the larger of the
		 * pre and post number of devices.
		 */
5695
		spin_lock_irqsave(&conf->device_lock, flags);
5696
		mddev->degraded = calc_degraded(conf);
5697 5698
		spin_unlock_irqrestore(&conf->device_lock, flags);
	}
5699
	mddev->raid_disks = conf->raid_disks;
5700
	mddev->reshape_position = conf->reshape_progress;
5701
	set_bit(MD_CHANGE_DEVS, &mddev->flags);
5702

5703 5704 5705 5706 5707
	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,
5708
						"reshape");
5709 5710 5711 5712
	if (!mddev->sync_thread) {
		mddev->recovery = 0;
		spin_lock_irq(&conf->device_lock);
		mddev->raid_disks = conf->raid_disks = conf->previous_raid_disks;
5713 5714 5715
		rdev_for_each(rdev, mddev)
			rdev->new_data_offset = rdev->data_offset;
		smp_wmb();
5716
		conf->reshape_progress = MaxSector;
5717
		mddev->reshape_position = MaxSector;
5718 5719 5720
		spin_unlock_irq(&conf->device_lock);
		return -EAGAIN;
	}
5721
	conf->reshape_checkpoint = jiffies;
5722 5723 5724 5725 5726
	md_wakeup_thread(mddev->sync_thread);
	md_new_event(mddev);
	return 0;
}

5727 5728 5729
/* This is called from the reshape thread and should make any
 * changes needed in 'conf'
 */
5730
static void end_reshape(struct r5conf *conf)
5731 5732
{

5733
	if (!test_bit(MD_RECOVERY_INTR, &conf->mddev->recovery)) {
5734
		struct md_rdev *rdev;
5735 5736

		spin_lock_irq(&conf->device_lock);
5737
		conf->previous_raid_disks = conf->raid_disks;
5738 5739 5740
		rdev_for_each(rdev, conf->mddev)
			rdev->data_offset = rdev->new_data_offset;
		smp_wmb();
5741
		conf->reshape_progress = MaxSector;
5742
		spin_unlock_irq(&conf->device_lock);
5743
		wake_up(&conf->wait_for_overlap);
5744 5745 5746 5747

		/* read-ahead size must cover two whole stripes, which is
		 * 2 * (datadisks) * chunksize where 'n' is the number of raid devices
		 */
5748
		if (conf->mddev->queue) {
5749
			int data_disks = conf->raid_disks - conf->max_degraded;
5750
			int stripe = data_disks * ((conf->chunk_sectors << 9)
5751
						   / PAGE_SIZE);
5752 5753 5754
			if (conf->mddev->queue->backing_dev_info.ra_pages < 2 * stripe)
				conf->mddev->queue->backing_dev_info.ra_pages = 2 * stripe;
		}
5755 5756 5757
	}
}

5758 5759 5760
/* This is called from the raid5d thread with mddev_lock held.
 * It makes config changes to the device.
 */
5761
static void raid5_finish_reshape(struct mddev *mddev)
5762
{
5763
	struct r5conf *conf = mddev->private;
5764 5765 5766

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

5767 5768 5769
		if (mddev->delta_disks > 0) {
			md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
			set_capacity(mddev->gendisk, mddev->array_sectors);
5770
			revalidate_disk(mddev->gendisk);
5771 5772
		} else {
			int d;
5773 5774 5775
			spin_lock_irq(&conf->device_lock);
			mddev->degraded = calc_degraded(conf);
			spin_unlock_irq(&conf->device_lock);
5776 5777
			for (d = conf->raid_disks ;
			     d < conf->raid_disks - mddev->delta_disks;
5778
			     d++) {
5779
				struct md_rdev *rdev = conf->disks[d].rdev;
5780 5781 5782 5783 5784
				if (rdev)
					clear_bit(In_sync, &rdev->flags);
				rdev = conf->disks[d].replacement;
				if (rdev)
					clear_bit(In_sync, &rdev->flags);
5785
			}
5786
		}
5787
		mddev->layout = conf->algorithm;
5788
		mddev->chunk_sectors = conf->chunk_sectors;
5789 5790
		mddev->reshape_position = MaxSector;
		mddev->delta_disks = 0;
5791
		mddev->reshape_backwards = 0;
5792 5793 5794
	}
}

5795
static void raid5_quiesce(struct mddev *mddev, int state)
5796
{
5797
	struct r5conf *conf = mddev->private;
5798 5799

	switch(state) {
5800 5801 5802 5803
	case 2: /* resume for a suspend */
		wake_up(&conf->wait_for_overlap);
		break;

5804 5805
	case 1: /* stop all writes */
		spin_lock_irq(&conf->device_lock);
5806 5807 5808 5809
		/* '2' tells resync/reshape to pause so that all
		 * active stripes can drain
		 */
		conf->quiesce = 2;
5810
		wait_event_lock_irq(conf->wait_for_stripe,
5811 5812
				    atomic_read(&conf->active_stripes) == 0 &&
				    atomic_read(&conf->active_aligned_reads) == 0,
5813
				    conf->device_lock, /* nothing */);
5814
		conf->quiesce = 1;
5815
		spin_unlock_irq(&conf->device_lock);
5816 5817
		/* allow reshape to continue */
		wake_up(&conf->wait_for_overlap);
5818 5819 5820 5821 5822 5823
		break;

	case 0: /* re-enable writes */
		spin_lock_irq(&conf->device_lock);
		conf->quiesce = 0;
		wake_up(&conf->wait_for_stripe);
5824
		wake_up(&conf->wait_for_overlap);
5825 5826 5827 5828
		spin_unlock_irq(&conf->device_lock);
		break;
	}
}
5829

5830

5831
static void *raid45_takeover_raid0(struct mddev *mddev, int level)
5832
{
5833
	struct r0conf *raid0_conf = mddev->private;
5834
	sector_t sectors;
5835

D
Dan Williams 已提交
5836
	/* for raid0 takeover only one zone is supported */
5837
	if (raid0_conf->nr_strip_zones > 1) {
5838 5839
		printk(KERN_ERR "md/raid:%s: cannot takeover raid0 with more than one zone.\n",
		       mdname(mddev));
D
Dan Williams 已提交
5840 5841 5842
		return ERR_PTR(-EINVAL);
	}

5843 5844
	sectors = raid0_conf->strip_zone[0].zone_end;
	sector_div(sectors, raid0_conf->strip_zone[0].nb_dev);
5845
	mddev->dev_sectors = sectors;
D
Dan Williams 已提交
5846
	mddev->new_level = level;
5847 5848 5849 5850 5851 5852 5853 5854 5855 5856 5857
	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);
}


5858
static void *raid5_takeover_raid1(struct mddev *mddev)
5859 5860 5861 5862 5863 5864 5865 5866 5867 5868 5869 5870 5871 5872 5873 5874 5875 5876 5877 5878 5879
{
	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;
5880
	mddev->new_chunk_sectors = chunksect;
5881 5882 5883 5884

	return setup_conf(mddev);
}

5885
static void *raid5_takeover_raid6(struct mddev *mddev)
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
{
	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);
}

5918

5919
static int raid5_check_reshape(struct mddev *mddev)
5920
{
5921 5922 5923 5924
	/* 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.
5925
	 */
5926
	struct r5conf *conf = mddev->private;
5927
	int new_chunk = mddev->new_chunk_sectors;
5928

5929
	if (mddev->new_layout >= 0 && !algorithm_valid_raid5(mddev->new_layout))
5930 5931
		return -EINVAL;
	if (new_chunk > 0) {
5932
		if (!is_power_of_2(new_chunk))
5933
			return -EINVAL;
5934
		if (new_chunk < (PAGE_SIZE>>9))
5935
			return -EINVAL;
5936
		if (mddev->array_sectors & (new_chunk-1))
5937 5938 5939 5940 5941 5942
			/* not factor of array size */
			return -EINVAL;
	}

	/* They look valid */

5943
	if (mddev->raid_disks == 2) {
5944 5945 5946 5947
		/* can make the change immediately */
		if (mddev->new_layout >= 0) {
			conf->algorithm = mddev->new_layout;
			mddev->layout = mddev->new_layout;
5948 5949
		}
		if (new_chunk > 0) {
5950 5951
			conf->chunk_sectors = new_chunk ;
			mddev->chunk_sectors = new_chunk;
5952 5953 5954
		}
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
		md_wakeup_thread(mddev->thread);
5955
	}
5956
	return check_reshape(mddev);
5957 5958
}

5959
static int raid6_check_reshape(struct mddev *mddev)
5960
{
5961
	int new_chunk = mddev->new_chunk_sectors;
5962

5963
	if (mddev->new_layout >= 0 && !algorithm_valid_raid6(mddev->new_layout))
5964
		return -EINVAL;
5965
	if (new_chunk > 0) {
5966
		if (!is_power_of_2(new_chunk))
5967
			return -EINVAL;
5968
		if (new_chunk < (PAGE_SIZE >> 9))
5969
			return -EINVAL;
5970
		if (mddev->array_sectors & (new_chunk-1))
5971 5972
			/* not factor of array size */
			return -EINVAL;
5973
	}
5974 5975

	/* They look valid */
5976
	return check_reshape(mddev);
5977 5978
}

5979
static void *raid5_takeover(struct mddev *mddev)
5980 5981
{
	/* raid5 can take over:
D
Dan Williams 已提交
5982
	 *  raid0 - if there is only one strip zone - make it a raid4 layout
5983 5984 5985 5986
	 *  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
5987 5988
	if (mddev->level == 0)
		return raid45_takeover_raid0(mddev, 5);
5989 5990
	if (mddev->level == 1)
		return raid5_takeover_raid1(mddev);
5991 5992 5993 5994 5995
	if (mddev->level == 4) {
		mddev->new_layout = ALGORITHM_PARITY_N;
		mddev->new_level = 5;
		return setup_conf(mddev);
	}
5996 5997
	if (mddev->level == 6)
		return raid5_takeover_raid6(mddev);
5998 5999 6000 6001

	return ERR_PTR(-EINVAL);
}

6002
static void *raid4_takeover(struct mddev *mddev)
6003
{
D
Dan Williams 已提交
6004 6005 6006
	/* raid4 can take over:
	 *  raid0 - if there is only one strip zone
	 *  raid5 - if layout is right
6007
	 */
D
Dan Williams 已提交
6008 6009
	if (mddev->level == 0)
		return raid45_takeover_raid0(mddev, 4);
6010 6011 6012 6013 6014 6015 6016 6017
	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);
}
6018

6019
static struct md_personality raid5_personality;
6020

6021
static void *raid6_takeover(struct mddev *mddev)
6022 6023 6024 6025 6026 6027 6028 6029 6030 6031 6032 6033 6034 6035 6036 6037 6038 6039 6040 6041 6042 6043 6044 6045 6046 6047 6048 6049 6050 6051 6052 6053 6054 6055 6056 6057 6058 6059 6060 6061 6062 6063 6064 6065 6066 6067
{
	/* 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);
}


6068
static struct md_personality raid6_personality =
6069 6070 6071 6072 6073 6074 6075 6076 6077 6078 6079 6080 6081 6082
{
	.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,
6083
	.size		= raid5_size,
6084
	.check_reshape	= raid6_check_reshape,
6085
	.start_reshape  = raid5_start_reshape,
6086
	.finish_reshape = raid5_finish_reshape,
6087
	.quiesce	= raid5_quiesce,
6088
	.takeover	= raid6_takeover,
6089
};
6090
static struct md_personality raid5_personality =
L
Linus Torvalds 已提交
6091 6092
{
	.name		= "raid5",
6093
	.level		= 5,
L
Linus Torvalds 已提交
6094 6095 6096 6097 6098 6099 6100 6101 6102 6103 6104
	.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,
6105
	.size		= raid5_size,
6106 6107
	.check_reshape	= raid5_check_reshape,
	.start_reshape  = raid5_start_reshape,
6108
	.finish_reshape = raid5_finish_reshape,
6109
	.quiesce	= raid5_quiesce,
6110
	.takeover	= raid5_takeover,
L
Linus Torvalds 已提交
6111 6112
};

6113
static struct md_personality raid4_personality =
L
Linus Torvalds 已提交
6114
{
6115 6116 6117 6118 6119 6120 6121 6122 6123 6124 6125 6126 6127
	.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,
6128
	.size		= raid5_size,
6129 6130
	.check_reshape	= raid5_check_reshape,
	.start_reshape  = raid5_start_reshape,
6131
	.finish_reshape = raid5_finish_reshape,
6132
	.quiesce	= raid5_quiesce,
6133
	.takeover	= raid4_takeover,
6134 6135 6136 6137
};

static int __init raid5_init(void)
{
6138
	register_md_personality(&raid6_personality);
6139 6140 6141
	register_md_personality(&raid5_personality);
	register_md_personality(&raid4_personality);
	return 0;
L
Linus Torvalds 已提交
6142 6143
}

6144
static void raid5_exit(void)
L
Linus Torvalds 已提交
6145
{
6146
	unregister_md_personality(&raid6_personality);
6147 6148
	unregister_md_personality(&raid5_personality);
	unregister_md_personality(&raid4_personality);
L
Linus Torvalds 已提交
6149 6150 6151 6152 6153
}

module_init(raid5_init);
module_exit(raid5_exit);
MODULE_LICENSE("GPL");
6154
MODULE_DESCRIPTION("RAID4/5/6 (striping with parity) personality for MD");
L
Linus Torvalds 已提交
6155
MODULE_ALIAS("md-personality-4"); /* RAID5 */
6156 6157
MODULE_ALIAS("md-raid5");
MODULE_ALIAS("md-raid4");
6158 6159
MODULE_ALIAS("md-level-5");
MODULE_ALIAS("md-level-4");
6160 6161 6162 6163 6164 6165 6166
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");