raid5.c 178.8 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
static inline int raid5_bi_processed_stripes(struct bio *bio)
103
{
104 105
	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
	return (atomic_read(segments) >> 16) & 0xffff;
106 107
}

108
static inline int raid5_dec_bi_active_stripes(struct bio *bio)
109
{
110 111
	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
	return atomic_sub_return(1, segments) & 0xffff;
112 113
}

114
static inline void raid5_inc_bi_active_stripes(struct bio *bio)
115
{
116 117
	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
	atomic_inc(segments);
118 119
}

120 121
static inline void raid5_set_bi_processed_stripes(struct bio *bio,
	unsigned int cnt)
122
{
123 124
	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
	int old, new;
125

126 127 128 129
	do {
		old = atomic_read(segments);
		new = (old & 0xffff) | (cnt << 16);
	} while (atomic_cmpxchg(segments, old, new) != old);
130 131
}

132
static inline void raid5_set_bi_stripes(struct bio *bio, unsigned int cnt)
133
{
134 135
	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
	atomic_set(segments, cnt);
136 137
}

138 139 140
/* Find first data disk in a raid6 stripe */
static inline int raid6_d0(struct stripe_head *sh)
{
141 142 143 144
	if (sh->ddf_layout)
		/* ddf always start from first device */
		return 0;
	/* md starts just after Q block */
145 146 147 148 149
	if (sh->qd_idx == sh->disks - 1)
		return 0;
	else
		return sh->qd_idx + 1;
}
150 151 152 153 154
static inline int raid6_next_disk(int disk, int raid_disks)
{
	disk++;
	return (disk < raid_disks) ? disk : 0;
}
155

156 157 158 159 160
/* 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.
 */
161 162
static int raid6_idx_to_slot(int idx, struct stripe_head *sh,
			     int *count, int syndrome_disks)
163
{
164
	int slot = *count;
165

166
	if (sh->ddf_layout)
167
		(*count)++;
168
	if (idx == sh->pd_idx)
169
		return syndrome_disks;
170
	if (idx == sh->qd_idx)
171
		return syndrome_disks + 1;
172
	if (!sh->ddf_layout)
173
		(*count)++;
174 175 176
	return slot;
}

177 178 179 180 181 182 183 184
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;
185
		bio_endio(bi, 0);
186 187 188 189
		bi = return_bi;
	}
}

190
static void print_raid5_conf (struct r5conf *conf);
L
Linus Torvalds 已提交
191

192 193 194 195 196 197 198
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);
}

199
static void do_release_stripe(struct r5conf *conf, struct stripe_head *sh)
L
Linus Torvalds 已提交
200
{
201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227
	BUG_ON(!list_empty(&sh->lru));
	BUG_ON(atomic_read(&conf->active_stripes)==0);
	if (test_bit(STRIPE_HANDLE, &sh->state)) {
		if (test_bit(STRIPE_DELAYED, &sh->state) &&
		    !test_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
			list_add_tail(&sh->lru, &conf->delayed_list);
		else if (test_bit(STRIPE_BIT_DELAY, &sh->state) &&
			   sh->bm_seq - conf->seq_write > 0)
			list_add_tail(&sh->lru, &conf->bitmap_list);
		else {
			clear_bit(STRIPE_DELAYED, &sh->state);
			clear_bit(STRIPE_BIT_DELAY, &sh->state);
			list_add_tail(&sh->lru, &conf->handle_list);
		}
		md_wakeup_thread(conf->mddev->thread);
	} else {
		BUG_ON(stripe_operations_active(sh));
		if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
			if (atomic_dec_return(&conf->preread_active_stripes)
			    < IO_THRESHOLD)
				md_wakeup_thread(conf->mddev->thread);
		atomic_dec(&conf->active_stripes);
		if (!test_bit(STRIPE_EXPANDING, &sh->state)) {
			list_add_tail(&sh->lru, &conf->inactive_list);
			wake_up(&conf->wait_for_stripe);
			if (conf->retry_read_aligned)
				md_wakeup_thread(conf->mddev->thread);
L
Linus Torvalds 已提交
228 229 230
		}
	}
}
231

232 233 234 235 236 237
static void __release_stripe(struct r5conf *conf, struct stripe_head *sh)
{
	if (atomic_dec_and_test(&sh->count))
		do_release_stripe(conf, sh);
}

L
Linus Torvalds 已提交
238 239
static void release_stripe(struct stripe_head *sh)
{
240
	struct r5conf *conf = sh->raid_conf;
L
Linus Torvalds 已提交
241
	unsigned long flags;
242

243 244 245 246 247 248
	local_irq_save(flags);
	if (atomic_dec_and_lock(&sh->count, &conf->device_lock)) {
		do_release_stripe(conf, sh);
		spin_unlock(&conf->device_lock);
	}
	local_irq_restore(flags);
L
Linus Torvalds 已提交
249 250
}

251
static inline void remove_hash(struct stripe_head *sh)
L
Linus Torvalds 已提交
252
{
253 254
	pr_debug("remove_hash(), stripe %llu\n",
		(unsigned long long)sh->sector);
L
Linus Torvalds 已提交
255

256
	hlist_del_init(&sh->hash);
L
Linus Torvalds 已提交
257 258
}

259
static inline void insert_hash(struct r5conf *conf, struct stripe_head *sh)
L
Linus Torvalds 已提交
260
{
261
	struct hlist_head *hp = stripe_hash(conf, sh->sector);
L
Linus Torvalds 已提交
262

263 264
	pr_debug("insert_hash(), stripe %llu\n",
		(unsigned long long)sh->sector);
L
Linus Torvalds 已提交
265

266
	hlist_add_head(&sh->hash, hp);
L
Linus Torvalds 已提交
267 268 269 270
}


/* find an idle stripe, make sure it is unhashed, and return it. */
271
static struct stripe_head *get_free_stripe(struct r5conf *conf)
L
Linus Torvalds 已提交
272 273 274 275 276 277 278 279 280 281 282 283 284 285 286
{
	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;
}

287
static void shrink_buffers(struct stripe_head *sh)
L
Linus Torvalds 已提交
288 289 290
{
	struct page *p;
	int i;
291
	int num = sh->raid_conf->pool_size;
L
Linus Torvalds 已提交
292

293
	for (i = 0; i < num ; i++) {
L
Linus Torvalds 已提交
294 295 296 297
		p = sh->dev[i].page;
		if (!p)
			continue;
		sh->dev[i].page = NULL;
298
		put_page(p);
L
Linus Torvalds 已提交
299 300 301
	}
}

302
static int grow_buffers(struct stripe_head *sh)
L
Linus Torvalds 已提交
303 304
{
	int i;
305
	int num = sh->raid_conf->pool_size;
L
Linus Torvalds 已提交
306

307
	for (i = 0; i < num; i++) {
L
Linus Torvalds 已提交
308 309 310 311 312 313 314 315 316 317
		struct page *page;

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

318
static void raid5_build_block(struct stripe_head *sh, int i, int previous);
319
static void stripe_set_idx(sector_t stripe, struct r5conf *conf, int previous,
320
			    struct stripe_head *sh);
L
Linus Torvalds 已提交
321

322
static void init_stripe(struct stripe_head *sh, sector_t sector, int previous)
L
Linus Torvalds 已提交
323
{
324
	struct r5conf *conf = sh->raid_conf;
325
	int i;
L
Linus Torvalds 已提交
326

327 328
	BUG_ON(atomic_read(&sh->count) != 0);
	BUG_ON(test_bit(STRIPE_HANDLE, &sh->state));
329
	BUG_ON(stripe_operations_active(sh));
330

331
	pr_debug("init_stripe called, stripe %llu\n",
L
Linus Torvalds 已提交
332 333 334
		(unsigned long long)sh->sector);

	remove_hash(sh);
335

336
	sh->generation = conf->generation - previous;
337
	sh->disks = previous ? conf->previous_raid_disks : conf->raid_disks;
L
Linus Torvalds 已提交
338
	sh->sector = sector;
339
	stripe_set_idx(sector, conf, previous, sh);
L
Linus Torvalds 已提交
340 341
	sh->state = 0;

342 343

	for (i = sh->disks; i--; ) {
L
Linus Torvalds 已提交
344 345
		struct r5dev *dev = &sh->dev[i];

346
		if (dev->toread || dev->read || dev->towrite || dev->written ||
L
Linus Torvalds 已提交
347
		    test_bit(R5_LOCKED, &dev->flags)) {
348
			printk(KERN_ERR "sector=%llx i=%d %p %p %p %p %d\n",
L
Linus Torvalds 已提交
349
			       (unsigned long long)sh->sector, i, dev->toread,
350
			       dev->read, dev->towrite, dev->written,
L
Linus Torvalds 已提交
351
			       test_bit(R5_LOCKED, &dev->flags));
352
			WARN_ON(1);
L
Linus Torvalds 已提交
353 354
		}
		dev->flags = 0;
355
		raid5_build_block(sh, i, previous);
L
Linus Torvalds 已提交
356 357 358 359
	}
	insert_hash(conf, sh);
}

360
static struct stripe_head *__find_stripe(struct r5conf *conf, sector_t sector,
361
					 short generation)
L
Linus Torvalds 已提交
362 363
{
	struct stripe_head *sh;
364
	struct hlist_node *hn;
L
Linus Torvalds 已提交
365

366
	pr_debug("__find_stripe, sector %llu\n", (unsigned long long)sector);
367
	hlist_for_each_entry(sh, hn, stripe_hash(conf, sector), hash)
368
		if (sh->sector == sector && sh->generation == generation)
L
Linus Torvalds 已提交
369
			return sh;
370
	pr_debug("__stripe %llu not in cache\n", (unsigned long long)sector);
L
Linus Torvalds 已提交
371 372 373
	return NULL;
}

374 375 376 377 378 379 380 381 382 383 384 385 386
/*
 * 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.
 */
387
static int calc_degraded(struct r5conf *conf)
388
{
389
	int degraded, degraded2;
390 391 392 393 394
	int i;

	rcu_read_lock();
	degraded = 0;
	for (i = 0; i < conf->previous_raid_disks; i++) {
395
		struct md_rdev *rdev = rcu_dereference(conf->disks[i].rdev);
396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413
		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();
414 415
	if (conf->raid_disks == conf->previous_raid_disks)
		return degraded;
416
	rcu_read_lock();
417
	degraded2 = 0;
418
	for (i = 0; i < conf->raid_disks; i++) {
419
		struct md_rdev *rdev = rcu_dereference(conf->disks[i].rdev);
420
		if (!rdev || test_bit(Faulty, &rdev->flags))
421
			degraded2++;
422 423 424 425 426 427 428 429 430
		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)
431
				degraded2++;
432 433
	}
	rcu_read_unlock();
434 435 436 437 438 439 440 441 442 443 444 445 446
	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);
447 448 449 450 451
	if (degraded > conf->max_degraded)
		return 1;
	return 0;
}

452
static struct stripe_head *
453
get_active_stripe(struct r5conf *conf, sector_t sector,
454
		  int previous, int noblock, int noquiesce)
L
Linus Torvalds 已提交
455 456 457
{
	struct stripe_head *sh;

458
	pr_debug("get_stripe, sector %llu\n", (unsigned long long)sector);
L
Linus Torvalds 已提交
459 460 461 462

	spin_lock_irq(&conf->device_lock);

	do {
463
		wait_event_lock_irq(conf->wait_for_stripe,
464
				    conf->quiesce == 0 || noquiesce,
465
				    conf->device_lock, /* nothing */);
466
		sh = __find_stripe(conf, sector, conf->generation - previous);
L
Linus Torvalds 已提交
467 468 469 470 471 472 473 474 475
		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) &&
476 477
						    (atomic_read(&conf->active_stripes)
						     < (conf->max_nr_stripes *3/4)
L
Linus Torvalds 已提交
478 479
						     || !conf->inactive_blocked),
						    conf->device_lock,
480
						    );
L
Linus Torvalds 已提交
481 482
				conf->inactive_blocked = 0;
			} else
483
				init_stripe(sh, sector, previous);
L
Linus Torvalds 已提交
484 485
		} else {
			if (atomic_read(&sh->count)) {
486
				BUG_ON(!list_empty(&sh->lru)
487 488
				    && !test_bit(STRIPE_EXPANDING, &sh->state)
				    && !test_bit(STRIPE_ON_UNPLUG_LIST, &sh->state));
L
Linus Torvalds 已提交
489 490 491
			} else {
				if (!test_bit(STRIPE_HANDLE, &sh->state))
					atomic_inc(&conf->active_stripes);
492 493
				if (list_empty(&sh->lru) &&
				    !test_bit(STRIPE_EXPANDING, &sh->state))
494 495
					BUG();
				list_del_init(&sh->lru);
L
Linus Torvalds 已提交
496 497 498 499 500 501 502 503 504 505 506
			}
		}
	} while (sh == NULL);

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

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

507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527
/* 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;
}

528 529 530 531
static void
raid5_end_read_request(struct bio *bi, int error);
static void
raid5_end_write_request(struct bio *bi, int error);
532

533
static void ops_run_io(struct stripe_head *sh, struct stripe_head_state *s)
534
{
535
	struct r5conf *conf = sh->raid_conf;
536 537 538 539 540 541
	int i, disks = sh->disks;

	might_sleep();

	for (i = disks; i--; ) {
		int rw;
542
		int replace_only = 0;
543 544
		struct bio *bi, *rbi;
		struct md_rdev *rdev, *rrdev = NULL;
T
Tejun Heo 已提交
545 546 547 548 549
		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;
S
Shaohua Li 已提交
550 551
			if (test_and_clear_bit(R5_Discard, &sh->dev[i].flags))
				rw |= REQ_DISCARD;
T
Tejun Heo 已提交
552
		} else if (test_and_clear_bit(R5_Wantread, &sh->dev[i].flags))
553
			rw = READ;
554 555 556 557 558
		else if (test_and_clear_bit(R5_WantReplace,
					    &sh->dev[i].flags)) {
			rw = WRITE;
			replace_only = 1;
		} else
559
			continue;
S
Shaohua Li 已提交
560 561
		if (test_and_clear_bit(R5_SyncIO, &sh->dev[i].flags))
			rw |= REQ_SYNC;
562 563

		bi = &sh->dev[i].req;
564
		rbi = &sh->dev[i].rreq; /* For writing to replacement */
565 566

		bi->bi_rw = rw;
567 568
		rbi->bi_rw = rw;
		if (rw & WRITE) {
569
			bi->bi_end_io = raid5_end_write_request;
570 571
			rbi->bi_end_io = raid5_end_write_request;
		} else
572 573 574
			bi->bi_end_io = raid5_end_read_request;

		rcu_read_lock();
575
		rrdev = rcu_dereference(conf->disks[i].replacement);
576 577 578 579 580 581
		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;
		}
582 583 584
		if (rw & WRITE) {
			if (replace_only)
				rdev = NULL;
585 586 587
			if (rdev == rrdev)
				/* We raced and saw duplicates */
				rrdev = NULL;
588
		} else {
589
			if (test_bit(R5_ReadRepl, &sh->dev[i].flags) && rrdev)
590 591 592
				rdev = rrdev;
			rrdev = NULL;
		}
593

594 595 596 597
		if (rdev && test_bit(Faulty, &rdev->flags))
			rdev = NULL;
		if (rdev)
			atomic_inc(&rdev->nr_pending);
598 599 600 601
		if (rrdev && test_bit(Faulty, &rrdev->flags))
			rrdev = NULL;
		if (rrdev)
			atomic_inc(&rrdev->nr_pending);
602 603
		rcu_read_unlock();

604
		/* We have already checked bad blocks for reads.  Now
605 606
		 * need to check for writes.  We never accept write errors
		 * on the replacement, so we don't to check rrdev.
607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626
		 */
		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);
				}
627 628 629 630 631 632
				/*
				 * Because md_wait_for_blocked_rdev
				 * will dec nr_pending, we must
				 * increment it first.
				 */
				atomic_inc(&rdev->nr_pending);
633 634 635 636 637 638 639 640
				md_wait_for_blocked_rdev(rdev, conf->mddev);
			} else {
				/* Acknowledged bad block - skip the write */
				rdev_dec_pending(rdev, conf->mddev);
				rdev = NULL;
			}
		}

641
		if (rdev) {
642 643
			if (s->syncing || s->expanding || s->expanded
			    || s->replacing)
644 645
				md_sync_acct(rdev->bdev, STRIPE_SECTORS);

D
Dan Williams 已提交
646 647
			set_bit(STRIPE_IO_STARTED, &sh->state);

648 649
			bi->bi_bdev = rdev->bdev;
			pr_debug("%s: for %llu schedule op %ld on disc %d\n",
650
				__func__, (unsigned long long)sh->sector,
651 652
				bi->bi_rw, i);
			atomic_inc(&sh->count);
653 654 655 656 657 658
			if (use_new_offset(conf, sh))
				bi->bi_sector = (sh->sector
						 + rdev->new_data_offset);
			else
				bi->bi_sector = (sh->sector
						 + rdev->data_offset);
659 660 661
			if (test_bit(R5_ReadNoMerge, &sh->dev[i].flags))
				bi->bi_rw |= REQ_FLUSH;

662 663 664 665 666 667
			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;
668 669
			if (rrdev)
				set_bit(R5_DOUBLE_LOCKED, &sh->dev[i].flags);
670
			generic_make_request(bi);
671 672
		}
		if (rrdev) {
673 674
			if (s->syncing || s->expanding || s->expanded
			    || s->replacing)
675 676 677 678 679 680 681 682 683 684
				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);
685 686 687 688 689 690
			if (use_new_offset(conf, sh))
				rbi->bi_sector = (sh->sector
						  + rrdev->new_data_offset);
			else
				rbi->bi_sector = (sh->sector
						  + rrdev->data_offset);
691 692 693 694 695 696 697 698 699
			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) {
700
			if (rw & WRITE)
701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717
				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;
718
	struct async_submit_ctl submit;
D
Dan Williams 已提交
719
	enum async_tx_flags flags = 0;
720 721 722 723 724

	if (bio->bi_sector >= sector)
		page_offset = (signed)(bio->bi_sector - sector) * 512;
	else
		page_offset = (signed)(sector - bio->bi_sector) * -512;
725

D
Dan Williams 已提交
726 727 728 729
	if (frombio)
		flags |= ASYNC_TX_FENCE;
	init_async_submit(&submit, flags, tx, NULL, NULL, NULL);

730
	bio_for_each_segment(bvl, bio, i) {
731
		int len = bvl->bv_len;
732 733 734 735 736 737 738 739 740 741 742 743 744 745 746
		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) {
747 748
			b_offset += bvl->bv_offset;
			bio_page = bvl->bv_page;
749 750
			if (frombio)
				tx = async_memcpy(page, bio_page, page_offset,
751
						  b_offset, clen, &submit);
752 753
			else
				tx = async_memcpy(bio_page, page, b_offset,
754
						  page_offset, clen, &submit);
755
		}
756 757 758
		/* chain the operations */
		submit.depend_tx = tx;

759 760 761 762 763 764 765 766 767 768 769 770
		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;
771
	int i;
772

773
	pr_debug("%s: stripe %llu\n", __func__,
774 775 776 777 778 779 780
		(unsigned long long)sh->sector);

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

		/* acknowledge completion of a biofill operation */
781 782
		/* and check if we need to reply to a read request,
		 * new R5_Wantfill requests are held off until
783
		 * !STRIPE_BIOFILL_RUN
784 785
		 */
		if (test_and_clear_bit(R5_Wantfill, &dev->flags)) {
786 787 788 789 790 791 792 793
			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);
794
				if (!raid5_dec_bi_active_stripes(rbi)) {
795 796 797 798 799 800 801
					rbi->bi_next = return_bi;
					return_bi = rbi;
				}
				rbi = rbi2;
			}
		}
	}
802
	clear_bit(STRIPE_BIOFILL_RUN, &sh->state);
803 804 805

	return_io(return_bi);

806
	set_bit(STRIPE_HANDLE, &sh->state);
807 808 809 810 811 812
	release_stripe(sh);
}

static void ops_run_biofill(struct stripe_head *sh)
{
	struct dma_async_tx_descriptor *tx = NULL;
813
	struct async_submit_ctl submit;
814 815
	int i;

816
	pr_debug("%s: stripe %llu\n", __func__,
817 818 819 820 821 822
		(unsigned long long)sh->sector);

	for (i = sh->disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
		if (test_bit(R5_Wantfill, &dev->flags)) {
			struct bio *rbi;
S
Shaohua Li 已提交
823
			spin_lock_irq(&sh->stripe_lock);
824 825
			dev->read = rbi = dev->toread;
			dev->toread = NULL;
S
Shaohua Li 已提交
826
			spin_unlock_irq(&sh->stripe_lock);
827 828 829 830 831 832 833 834 835 836
			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);
837 838
	init_async_submit(&submit, ASYNC_TX_ACK, tx, ops_complete_biofill, sh, NULL);
	async_trigger_callback(&submit);
839 840
}

841
static void mark_target_uptodate(struct stripe_head *sh, int target)
842
{
843
	struct r5dev *tgt;
844

845 846
	if (target < 0)
		return;
847

848
	tgt = &sh->dev[target];
849 850 851
	set_bit(R5_UPTODATE, &tgt->flags);
	BUG_ON(!test_bit(R5_Wantcompute, &tgt->flags));
	clear_bit(R5_Wantcompute, &tgt->flags);
852 853
}

854
static void ops_complete_compute(void *stripe_head_ref)
855 856 857
{
	struct stripe_head *sh = stripe_head_ref;

858
	pr_debug("%s: stripe %llu\n", __func__,
859 860
		(unsigned long long)sh->sector);

861
	/* mark the computed target(s) as uptodate */
862
	mark_target_uptodate(sh, sh->ops.target);
863
	mark_target_uptodate(sh, sh->ops.target2);
864

865 866 867
	clear_bit(STRIPE_COMPUTE_RUN, &sh->state);
	if (sh->check_state == check_state_compute_run)
		sh->check_state = check_state_compute_result;
868 869 870 871
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

872 873 874 875 876 877 878 879 880
/* 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)
881 882
{
	int disks = sh->disks;
883
	struct page **xor_srcs = percpu->scribble;
884 885 886 887 888
	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;
889
	struct async_submit_ctl submit;
890 891 892
	int i;

	pr_debug("%s: stripe %llu block: %d\n",
893
		__func__, (unsigned long long)sh->sector, target);
894 895 896 897 898 899 900 901
	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
902
	init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST, NULL,
903
			  ops_complete_compute, sh, to_addr_conv(sh, percpu));
904
	if (unlikely(count == 1))
905
		tx = async_memcpy(xor_dest, xor_srcs[0], 0, 0, STRIPE_SIZE, &submit);
906
	else
907
		tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, &submit);
908 909 910 911

	return tx;
}

912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929
/* 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++)
930
		srcs[i] = NULL;
931 932 933 934 935 936 937 938 939 940

	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);

941
	return syndrome_disks;
942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961
}

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;
962
	else
963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978
		/* 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
979 980
		init_async_submit(&submit, ASYNC_TX_FENCE, NULL,
				  ops_complete_compute, sh,
981 982 983 984 985 986 987 988 989 990 991
				  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
992 993
		init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST,
				  NULL, ops_complete_compute, sh,
994 995 996
				  to_addr_conv(sh, percpu));
		tx = async_xor(dest, blocks, 0, count, STRIPE_SIZE, &submit);
	}
997 998 999 1000

	return tx;
}

1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021
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));

1022
	/* we need to open-code set_syndrome_sources to handle the
1023 1024 1025
	 * slot number conversion for 'faila' and 'failb'
	 */
	for (i = 0; i < disks ; i++)
1026
		blocks[i] = NULL;
1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052
	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
1053 1054 1055
			init_async_submit(&submit, ASYNC_TX_FENCE, NULL,
					  ops_complete_compute, sh,
					  to_addr_conv(sh, percpu));
1056
			return async_gen_syndrome(blocks, 0, syndrome_disks+2,
1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075
						  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
1076 1077 1078 1079
			init_async_submit(&submit,
					  ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST,
					  NULL, NULL, NULL,
					  to_addr_conv(sh, percpu));
1080 1081 1082 1083
			tx = async_xor(dest, blocks, 0, count, STRIPE_SIZE,
				       &submit);

			count = set_syndrome_sources(blocks, sh);
D
Dan Williams 已提交
1084 1085 1086
			init_async_submit(&submit, ASYNC_TX_FENCE, tx,
					  ops_complete_compute, sh,
					  to_addr_conv(sh, percpu));
1087 1088 1089 1090
			return async_gen_syndrome(blocks, 0, count+2,
						  STRIPE_SIZE, &submit);
		}
	} else {
1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104
		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);
		}
1105 1106 1107 1108
	}
}


1109 1110 1111 1112
static void ops_complete_prexor(void *stripe_head_ref)
{
	struct stripe_head *sh = stripe_head_ref;

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

static struct dma_async_tx_descriptor *
1118 1119
ops_run_prexor(struct stripe_head *sh, struct raid5_percpu *percpu,
	       struct dma_async_tx_descriptor *tx)
1120 1121
{
	int disks = sh->disks;
1122
	struct page **xor_srcs = percpu->scribble;
1123
	int count = 0, pd_idx = sh->pd_idx, i;
1124
	struct async_submit_ctl submit;
1125 1126 1127 1128

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

1129
	pr_debug("%s: stripe %llu\n", __func__,
1130 1131 1132 1133 1134
		(unsigned long long)sh->sector);

	for (i = disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
		/* Only process blocks that are known to be uptodate */
1135
		if (test_bit(R5_Wantdrain, &dev->flags))
1136 1137 1138
			xor_srcs[count++] = dev->page;
	}

D
Dan Williams 已提交
1139
	init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_DROP_DST, tx,
1140
			  ops_complete_prexor, sh, to_addr_conv(sh, percpu));
1141
	tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, &submit);
1142 1143 1144 1145 1146

	return tx;
}

static struct dma_async_tx_descriptor *
1147
ops_run_biodrain(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
1148 1149
{
	int disks = sh->disks;
1150
	int i;
1151

1152
	pr_debug("%s: stripe %llu\n", __func__,
1153 1154 1155 1156 1157 1158
		(unsigned long long)sh->sector);

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

1159
		if (test_and_clear_bit(R5_Wantdrain, &dev->flags)) {
1160 1161
			struct bio *wbi;

S
Shaohua Li 已提交
1162
			spin_lock_irq(&sh->stripe_lock);
1163 1164 1165 1166
			chosen = dev->towrite;
			dev->towrite = NULL;
			BUG_ON(dev->written);
			wbi = dev->written = chosen;
S
Shaohua Li 已提交
1167
			spin_unlock_irq(&sh->stripe_lock);
1168 1169 1170

			while (wbi && wbi->bi_sector <
				dev->sector + STRIPE_SECTORS) {
T
Tejun Heo 已提交
1171 1172
				if (wbi->bi_rw & REQ_FUA)
					set_bit(R5_WantFUA, &dev->flags);
S
Shaohua Li 已提交
1173 1174
				if (wbi->bi_rw & REQ_SYNC)
					set_bit(R5_SyncIO, &dev->flags);
S
Shaohua Li 已提交
1175 1176 1177 1178 1179 1180 1181
				if (wbi->bi_rw & REQ_DISCARD) {
					memset(page_address(dev->page), 0,
						STRIPE_SECTORS << 9);
					set_bit(R5_Discard, &dev->flags);
				} else
					tx = async_copy_data(1, wbi, dev->page,
						dev->sector, tx);
1182 1183 1184 1185 1186 1187 1188 1189
				wbi = r5_next_bio(wbi, dev->sector);
			}
		}
	}

	return tx;
}

1190
static void ops_complete_reconstruct(void *stripe_head_ref)
1191 1192
{
	struct stripe_head *sh = stripe_head_ref;
1193 1194 1195 1196
	int disks = sh->disks;
	int pd_idx = sh->pd_idx;
	int qd_idx = sh->qd_idx;
	int i;
S
Shaohua Li 已提交
1197
	bool fua = false, sync = false;
1198

1199
	pr_debug("%s: stripe %llu\n", __func__,
1200 1201
		(unsigned long long)sh->sector);

S
Shaohua Li 已提交
1202
	for (i = disks; i--; ) {
T
Tejun Heo 已提交
1203
		fua |= test_bit(R5_WantFUA, &sh->dev[i].flags);
S
Shaohua Li 已提交
1204 1205
		sync |= test_bit(R5_SyncIO, &sh->dev[i].flags);
	}
T
Tejun Heo 已提交
1206

1207 1208
	for (i = disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
1209

T
Tejun Heo 已提交
1210
		if (dev->written || i == pd_idx || i == qd_idx) {
1211
			set_bit(R5_UPTODATE, &dev->flags);
T
Tejun Heo 已提交
1212 1213
			if (fua)
				set_bit(R5_WantFUA, &dev->flags);
S
Shaohua Li 已提交
1214 1215
			if (sync)
				set_bit(R5_SyncIO, &dev->flags);
T
Tejun Heo 已提交
1216
		}
1217 1218
	}

1219 1220 1221 1222 1223 1224 1225 1226
	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;
	}
1227 1228 1229 1230 1231 1232

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

static void
1233 1234
ops_run_reconstruct5(struct stripe_head *sh, struct raid5_percpu *percpu,
		     struct dma_async_tx_descriptor *tx)
1235 1236
{
	int disks = sh->disks;
1237
	struct page **xor_srcs = percpu->scribble;
1238
	struct async_submit_ctl submit;
1239 1240
	int count = 0, pd_idx = sh->pd_idx, i;
	struct page *xor_dest;
1241
	int prexor = 0;
1242 1243
	unsigned long flags;

1244
	pr_debug("%s: stripe %llu\n", __func__,
1245 1246
		(unsigned long long)sh->sector);

S
Shaohua Li 已提交
1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260
	for (i = 0; i < sh->disks; i++) {
		if (pd_idx == i)
			continue;
		if (!test_bit(R5_Discard, &sh->dev[i].flags))
			break;
	}
	if (i >= sh->disks) {
		atomic_inc(&sh->count);
		memset(page_address(sh->dev[pd_idx].page), 0,
			STRIPE_SECTORS << 9);
		set_bit(R5_Discard, &sh->dev[pd_idx].flags);
		ops_complete_reconstruct(sh);
		return;
	}
1261 1262 1263
	/* check if prexor is active which means only process blocks
	 * that are part of a read-modify-write (written)
	 */
1264 1265
	if (sh->reconstruct_state == reconstruct_state_prexor_drain_run) {
		prexor = 1;
1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285
		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
	 */
1286
	flags = ASYNC_TX_ACK |
1287 1288 1289 1290
		(prexor ? ASYNC_TX_XOR_DROP_DST : ASYNC_TX_XOR_ZERO_DST);

	atomic_inc(&sh->count);

1291
	init_async_submit(&submit, flags, tx, ops_complete_reconstruct, sh,
1292
			  to_addr_conv(sh, percpu));
1293 1294 1295 1296
	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);
1297 1298
}

1299 1300 1301 1302 1303 1304
static void
ops_run_reconstruct6(struct stripe_head *sh, struct raid5_percpu *percpu,
		     struct dma_async_tx_descriptor *tx)
{
	struct async_submit_ctl submit;
	struct page **blocks = percpu->scribble;
S
Shaohua Li 已提交
1305
	int count, i;
1306 1307 1308

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

S
Shaohua Li 已提交
1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326
	for (i = 0; i < sh->disks; i++) {
		if (sh->pd_idx == i || sh->qd_idx == i)
			continue;
		if (!test_bit(R5_Discard, &sh->dev[i].flags))
			break;
	}
	if (i >= sh->disks) {
		atomic_inc(&sh->count);
		memset(page_address(sh->dev[sh->pd_idx].page), 0,
			STRIPE_SECTORS << 9);
		memset(page_address(sh->dev[sh->qd_idx].page), 0,
			STRIPE_SECTORS << 9);
		set_bit(R5_Discard, &sh->dev[sh->pd_idx].flags);
		set_bit(R5_Discard, &sh->dev[sh->qd_idx].flags);
		ops_complete_reconstruct(sh);
		return;
	}

1327 1328 1329 1330 1331 1332 1333
	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);
1334 1335 1336 1337 1338 1339
}

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

1340
	pr_debug("%s: stripe %llu\n", __func__,
1341 1342
		(unsigned long long)sh->sector);

1343
	sh->check_state = check_state_check_result;
1344 1345 1346 1347
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

1348
static void ops_run_check_p(struct stripe_head *sh, struct raid5_percpu *percpu)
1349 1350
{
	int disks = sh->disks;
1351 1352 1353
	int pd_idx = sh->pd_idx;
	int qd_idx = sh->qd_idx;
	struct page *xor_dest;
1354
	struct page **xor_srcs = percpu->scribble;
1355
	struct dma_async_tx_descriptor *tx;
1356
	struct async_submit_ctl submit;
1357 1358
	int count;
	int i;
1359

1360
	pr_debug("%s: stripe %llu\n", __func__,
1361 1362
		(unsigned long long)sh->sector);

1363 1364 1365
	count = 0;
	xor_dest = sh->dev[pd_idx].page;
	xor_srcs[count++] = xor_dest;
1366
	for (i = disks; i--; ) {
1367 1368 1369
		if (i == pd_idx || i == qd_idx)
			continue;
		xor_srcs[count++] = sh->dev[i].page;
1370 1371
	}

1372 1373
	init_async_submit(&submit, 0, NULL, NULL, NULL,
			  to_addr_conv(sh, percpu));
D
Dan Williams 已提交
1374
	tx = async_xor_val(xor_dest, xor_srcs, 0, count, STRIPE_SIZE,
1375
			   &sh->ops.zero_sum_result, &submit);
1376 1377

	atomic_inc(&sh->count);
1378 1379
	init_async_submit(&submit, ASYNC_TX_ACK, tx, ops_complete_check, sh, NULL);
	tx = async_trigger_callback(&submit);
1380 1381
}

1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393
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;
1394 1395

	atomic_inc(&sh->count);
1396 1397 1398 1399
	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);
1400 1401
}

1402
static void __raid_run_ops(struct stripe_head *sh, unsigned long ops_request)
1403 1404 1405
{
	int overlap_clear = 0, i, disks = sh->disks;
	struct dma_async_tx_descriptor *tx = NULL;
1406
	struct r5conf *conf = sh->raid_conf;
1407
	int level = conf->level;
1408 1409
	struct raid5_percpu *percpu;
	unsigned long cpu;
1410

1411 1412
	cpu = get_cpu();
	percpu = per_cpu_ptr(conf->percpu, cpu);
1413
	if (test_bit(STRIPE_OP_BIOFILL, &ops_request)) {
1414 1415 1416 1417
		ops_run_biofill(sh);
		overlap_clear++;
	}

1418
	if (test_bit(STRIPE_OP_COMPUTE_BLK, &ops_request)) {
1419 1420 1421 1422 1423 1424 1425 1426 1427 1428
		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))
1429 1430
			async_tx_ack(tx);
	}
1431

1432
	if (test_bit(STRIPE_OP_PREXOR, &ops_request))
1433
		tx = ops_run_prexor(sh, percpu, tx);
1434

1435
	if (test_bit(STRIPE_OP_BIODRAIN, &ops_request)) {
1436
		tx = ops_run_biodrain(sh, tx);
1437 1438 1439
		overlap_clear++;
	}

1440 1441 1442 1443 1444 1445
	if (test_bit(STRIPE_OP_RECONSTRUCT, &ops_request)) {
		if (level < 6)
			ops_run_reconstruct5(sh, percpu, tx);
		else
			ops_run_reconstruct6(sh, percpu, tx);
	}
1446

1447 1448 1449 1450 1451 1452 1453 1454 1455 1456
	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();
	}
1457 1458 1459 1460 1461 1462 1463

	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);
		}
1464
	put_cpu();
1465 1466
}

1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496
#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

1497
static int grow_one_stripe(struct r5conf *conf)
L
Linus Torvalds 已提交
1498 1499
{
	struct stripe_head *sh;
N
Namhyung Kim 已提交
1500
	sh = kmem_cache_zalloc(conf->slab_cache, GFP_KERNEL);
1501 1502
	if (!sh)
		return 0;
N
Namhyung Kim 已提交
1503

1504
	sh->raid_conf = conf;
1505 1506 1507
	#ifdef CONFIG_MULTICORE_RAID456
	init_waitqueue_head(&sh->ops.wait_for_ops);
	#endif
1508

S
Shaohua Li 已提交
1509 1510
	spin_lock_init(&sh->stripe_lock);

1511 1512
	if (grow_buffers(sh)) {
		shrink_buffers(sh);
1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523
		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;
}

1524
static int grow_stripes(struct r5conf *conf, int num)
1525
{
1526
	struct kmem_cache *sc;
1527
	int devs = max(conf->raid_disks, conf->previous_raid_disks);
L
Linus Torvalds 已提交
1528

1529 1530 1531 1532 1533 1534 1535 1536
	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]);

1537 1538
	conf->active_name = 0;
	sc = kmem_cache_create(conf->cache_name[conf->active_name],
L
Linus Torvalds 已提交
1539
			       sizeof(struct stripe_head)+(devs-1)*sizeof(struct r5dev),
1540
			       0, 0, NULL);
L
Linus Torvalds 已提交
1541 1542 1543
	if (!sc)
		return 1;
	conf->slab_cache = sc;
1544
	conf->pool_size = devs;
1545
	while (num--)
1546
		if (!grow_one_stripe(conf))
L
Linus Torvalds 已提交
1547 1548 1549
			return 1;
	return 0;
}
1550

1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572
/**
 * 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;
}

1573
static int resize_stripes(struct r5conf *conf, int newsize)
1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600
{
	/* 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;
1601
	unsigned long cpu;
1602
	int err;
1603
	struct kmem_cache *sc;
1604 1605 1606 1607 1608
	int i;

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

1609 1610 1611
	err = md_allow_write(conf->mddev);
	if (err)
		return err;
1612

1613 1614 1615
	/* Step 1 */
	sc = kmem_cache_create(conf->cache_name[1-conf->active_name],
			       sizeof(struct stripe_head)+(newsize-1)*sizeof(struct r5dev),
1616
			       0, 0, NULL);
1617 1618 1619 1620
	if (!sc)
		return -ENOMEM;

	for (i = conf->max_nr_stripes; i; i--) {
N
Namhyung Kim 已提交
1621
		nsh = kmem_cache_zalloc(sc, GFP_KERNEL);
1622 1623 1624 1625
		if (!nsh)
			break;

		nsh->raid_conf = conf;
1626 1627 1628
		#ifdef CONFIG_MULTICORE_RAID456
		init_waitqueue_head(&nsh->ops.wait_for_ops);
		#endif
1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650

		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
1651
				    );
1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665
		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
1666
	 * conf->disks and the scribble region
1667 1668 1669 1670 1671 1672 1673 1674 1675 1676
	 */
	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;

1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695
	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();

1696 1697 1698 1699
	/* 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);
1700

1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716
		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
1717

1718
static int drop_one_stripe(struct r5conf *conf)
L
Linus Torvalds 已提交
1719 1720 1721
{
	struct stripe_head *sh;

1722 1723 1724 1725 1726
	spin_lock_irq(&conf->device_lock);
	sh = get_free_stripe(conf);
	spin_unlock_irq(&conf->device_lock);
	if (!sh)
		return 0;
1727
	BUG_ON(atomic_read(&sh->count));
1728
	shrink_buffers(sh);
1729 1730 1731 1732 1733
	kmem_cache_free(conf->slab_cache, sh);
	atomic_dec(&conf->active_stripes);
	return 1;
}

1734
static void shrink_stripes(struct r5conf *conf)
1735 1736 1737 1738
{
	while (drop_one_stripe(conf))
		;

N
NeilBrown 已提交
1739 1740
	if (conf->slab_cache)
		kmem_cache_destroy(conf->slab_cache);
L
Linus Torvalds 已提交
1741 1742 1743
	conf->slab_cache = NULL;
}

1744
static void raid5_end_read_request(struct bio * bi, int error)
L
Linus Torvalds 已提交
1745
{
1746
	struct stripe_head *sh = bi->bi_private;
1747
	struct r5conf *conf = sh->raid_conf;
1748
	int disks = sh->disks, i;
L
Linus Torvalds 已提交
1749
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
1750
	char b[BDEVNAME_SIZE];
1751
	struct md_rdev *rdev = NULL;
1752
	sector_t s;
L
Linus Torvalds 已提交
1753 1754 1755 1756 1757

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

1758 1759
	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
1760 1761 1762
		uptodate);
	if (i == disks) {
		BUG();
1763
		return;
L
Linus Torvalds 已提交
1764
	}
1765
	if (test_bit(R5_ReadRepl, &sh->dev[i].flags))
1766 1767 1768 1769 1770
		/* 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.
		 */
1771
		rdev = conf->disks[i].replacement;
1772
	if (!rdev)
1773
		rdev = conf->disks[i].rdev;
L
Linus Torvalds 已提交
1774

1775 1776 1777 1778
	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
1779 1780
	if (uptodate) {
		set_bit(R5_UPTODATE, &sh->dev[i].flags);
1781
		if (test_bit(R5_ReadError, &sh->dev[i].flags)) {
1782 1783 1784 1785
			/* Note that this cannot happen on a
			 * replacement device.  We just fail those on
			 * any error
			 */
1786 1787 1788 1789 1790
			printk_ratelimited(
				KERN_INFO
				"md/raid:%s: read error corrected"
				" (%lu sectors at %llu on %s)\n",
				mdname(conf->mddev), STRIPE_SECTORS,
1791
				(unsigned long long)s,
1792
				bdevname(rdev->bdev, b));
1793
			atomic_add(STRIPE_SECTORS, &rdev->corrected_errors);
1794 1795
			clear_bit(R5_ReadError, &sh->dev[i].flags);
			clear_bit(R5_ReWrite, &sh->dev[i].flags);
1796 1797 1798
		} else if (test_bit(R5_ReadNoMerge, &sh->dev[i].flags))
			clear_bit(R5_ReadNoMerge, &sh->dev[i].flags);

1799 1800
		if (atomic_read(&rdev->read_errors))
			atomic_set(&rdev->read_errors, 0);
L
Linus Torvalds 已提交
1801
	} else {
1802
		const char *bdn = bdevname(rdev->bdev, b);
1803
		int retry = 0;
1804
		int set_bad = 0;
1805

L
Linus Torvalds 已提交
1806
		clear_bit(R5_UPTODATE, &sh->dev[i].flags);
1807
		atomic_inc(&rdev->read_errors);
1808 1809 1810 1811 1812 1813
		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),
1814
				(unsigned long long)s,
1815
				bdn);
1816 1817
		else if (conf->mddev->degraded >= conf->max_degraded) {
			set_bad = 1;
1818 1819 1820 1821 1822
			printk_ratelimited(
				KERN_WARNING
				"md/raid:%s: read error not correctable "
				"(sector %llu on %s).\n",
				mdname(conf->mddev),
1823
				(unsigned long long)s,
1824
				bdn);
1825
		} else if (test_bit(R5_ReWrite, &sh->dev[i].flags)) {
1826
			/* Oh, no!!! */
1827
			set_bad = 1;
1828 1829 1830 1831 1832
			printk_ratelimited(
				KERN_WARNING
				"md/raid:%s: read error NOT corrected!! "
				"(sector %llu on %s).\n",
				mdname(conf->mddev),
1833
				(unsigned long long)s,
1834
				bdn);
1835
		} else if (atomic_read(&rdev->read_errors)
1836
			 > conf->max_nr_stripes)
N
NeilBrown 已提交
1837
			printk(KERN_WARNING
1838
			       "md/raid:%s: Too many read errors, failing device %s.\n",
1839
			       mdname(conf->mddev), bdn);
1840 1841 1842
		else
			retry = 1;
		if (retry)
1843 1844 1845 1846 1847
			if (test_bit(R5_ReadNoMerge, &sh->dev[i].flags)) {
				set_bit(R5_ReadError, &sh->dev[i].flags);
				clear_bit(R5_ReadNoMerge, &sh->dev[i].flags);
			} else
				set_bit(R5_ReadNoMerge, &sh->dev[i].flags);
1848
		else {
1849 1850
			clear_bit(R5_ReadError, &sh->dev[i].flags);
			clear_bit(R5_ReWrite, &sh->dev[i].flags);
1851 1852 1853 1854 1855
			if (!(set_bad
			      && test_bit(In_sync, &rdev->flags)
			      && rdev_set_badblocks(
				      rdev, sh->sector, STRIPE_SECTORS, 0)))
				md_error(conf->mddev, rdev);
1856
		}
L
Linus Torvalds 已提交
1857
	}
1858
	rdev_dec_pending(rdev, conf->mddev);
L
Linus Torvalds 已提交
1859 1860 1861 1862 1863
	clear_bit(R5_LOCKED, &sh->dev[i].flags);
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

1864
static void raid5_end_write_request(struct bio *bi, int error)
L
Linus Torvalds 已提交
1865
{
1866
	struct stripe_head *sh = bi->bi_private;
1867
	struct r5conf *conf = sh->raid_conf;
1868
	int disks = sh->disks, i;
1869
	struct md_rdev *uninitialized_var(rdev);
L
Linus Torvalds 已提交
1870
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
1871 1872
	sector_t first_bad;
	int bad_sectors;
1873
	int replacement = 0;
L
Linus Torvalds 已提交
1874

1875 1876 1877
	for (i = 0 ; i < disks; i++) {
		if (bi == &sh->dev[i].req) {
			rdev = conf->disks[i].rdev;
L
Linus Torvalds 已提交
1878
			break;
1879 1880 1881
		}
		if (bi == &sh->dev[i].rreq) {
			rdev = conf->disks[i].replacement;
1882 1883 1884 1885 1886 1887 1888 1889
			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;
1890 1891 1892
			break;
		}
	}
1893
	pr_debug("end_write_request %llu/%d, count %d, uptodate: %d.\n",
L
Linus Torvalds 已提交
1894 1895 1896 1897
		(unsigned long long)sh->sector, i, atomic_read(&sh->count),
		uptodate);
	if (i == disks) {
		BUG();
1898
		return;
L
Linus Torvalds 已提交
1899 1900
	}

1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911
	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);
1912 1913 1914
			if (!test_and_set_bit(WantReplacement, &rdev->flags))
				set_bit(MD_RECOVERY_NEEDED,
					&rdev->mddev->recovery);
1915 1916 1917 1918 1919 1920
		} 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
1921

1922 1923
	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
1924
	set_bit(STRIPE_HANDLE, &sh->state);
1925
	release_stripe(sh);
L
Linus Torvalds 已提交
1926 1927
}

1928
static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous);
L
Linus Torvalds 已提交
1929
	
1930
static void raid5_build_block(struct stripe_head *sh, int i, int previous)
L
Linus Torvalds 已提交
1931 1932 1933 1934 1935 1936 1937 1938
{
	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;
1939
	dev->vec.bv_page = dev->page;
L
Linus Torvalds 已提交
1940

1941 1942 1943 1944 1945 1946 1947
	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
1948
	dev->flags = 0;
1949
	dev->sector = compute_blocknr(sh, i, previous);
L
Linus Torvalds 已提交
1950 1951
}

1952
static void error(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
1953 1954
{
	char b[BDEVNAME_SIZE];
1955
	struct r5conf *conf = mddev->private;
1956
	unsigned long flags;
1957
	pr_debug("raid456: error called\n");
L
Linus Torvalds 已提交
1958

1959 1960 1961 1962 1963 1964
	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);

1965
	set_bit(Blocked, &rdev->flags);
1966 1967 1968 1969 1970 1971 1972 1973 1974
	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);
1975
}
L
Linus Torvalds 已提交
1976 1977 1978 1979 1980

/*
 * Input: a 'big' sector number,
 * Output: index of the data and parity disk, and the sector # in them.
 */
1981
static sector_t raid5_compute_sector(struct r5conf *conf, sector_t r_sector,
1982 1983
				     int previous, int *dd_idx,
				     struct stripe_head *sh)
L
Linus Torvalds 已提交
1984
{
N
NeilBrown 已提交
1985
	sector_t stripe, stripe2;
1986
	sector_t chunk_number;
L
Linus Torvalds 已提交
1987
	unsigned int chunk_offset;
1988
	int pd_idx, qd_idx;
1989
	int ddf_layout = 0;
L
Linus Torvalds 已提交
1990
	sector_t new_sector;
1991 1992
	int algorithm = previous ? conf->prev_algo
				 : conf->algorithm;
1993 1994
	int sectors_per_chunk = previous ? conf->prev_chunk_sectors
					 : conf->chunk_sectors;
1995 1996 1997
	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
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

	/* 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
	 */
2010 2011
	stripe = chunk_number;
	*dd_idx = sector_div(stripe, data_disks);
N
NeilBrown 已提交
2012
	stripe2 = stripe;
L
Linus Torvalds 已提交
2013 2014 2015
	/*
	 * Select the parity disk based on the user selected algorithm.
	 */
2016
	pd_idx = qd_idx = -1;
2017 2018
	switch(conf->level) {
	case 4:
2019
		pd_idx = data_disks;
2020 2021
		break;
	case 5:
2022
		switch (algorithm) {
L
Linus Torvalds 已提交
2023
		case ALGORITHM_LEFT_ASYMMETRIC:
N
NeilBrown 已提交
2024
			pd_idx = data_disks - sector_div(stripe2, raid_disks);
2025
			if (*dd_idx >= pd_idx)
L
Linus Torvalds 已提交
2026 2027 2028
				(*dd_idx)++;
			break;
		case ALGORITHM_RIGHT_ASYMMETRIC:
N
NeilBrown 已提交
2029
			pd_idx = sector_div(stripe2, raid_disks);
2030
			if (*dd_idx >= pd_idx)
L
Linus Torvalds 已提交
2031 2032 2033
				(*dd_idx)++;
			break;
		case ALGORITHM_LEFT_SYMMETRIC:
N
NeilBrown 已提交
2034
			pd_idx = data_disks - sector_div(stripe2, raid_disks);
2035
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
L
Linus Torvalds 已提交
2036 2037
			break;
		case ALGORITHM_RIGHT_SYMMETRIC:
N
NeilBrown 已提交
2038
			pd_idx = sector_div(stripe2, raid_disks);
2039
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
L
Linus Torvalds 已提交
2040
			break;
2041 2042 2043 2044 2045 2046 2047
		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
2048
		default:
2049
			BUG();
2050 2051 2052 2053
		}
		break;
	case 6:

2054
		switch (algorithm) {
2055
		case ALGORITHM_LEFT_ASYMMETRIC:
N
NeilBrown 已提交
2056
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2057 2058
			qd_idx = pd_idx + 1;
			if (pd_idx == raid_disks-1) {
2059
				(*dd_idx)++;	/* Q D D D P */
2060 2061
				qd_idx = 0;
			} else if (*dd_idx >= pd_idx)
2062 2063 2064
				(*dd_idx) += 2; /* D D P Q D */
			break;
		case ALGORITHM_RIGHT_ASYMMETRIC:
N
NeilBrown 已提交
2065
			pd_idx = sector_div(stripe2, raid_disks);
2066 2067
			qd_idx = pd_idx + 1;
			if (pd_idx == raid_disks-1) {
2068
				(*dd_idx)++;	/* Q D D D P */
2069 2070
				qd_idx = 0;
			} else if (*dd_idx >= pd_idx)
2071 2072 2073
				(*dd_idx) += 2; /* D D P Q D */
			break;
		case ALGORITHM_LEFT_SYMMETRIC:
N
NeilBrown 已提交
2074
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2075 2076
			qd_idx = (pd_idx + 1) % raid_disks;
			*dd_idx = (pd_idx + 2 + *dd_idx) % raid_disks;
2077 2078
			break;
		case ALGORITHM_RIGHT_SYMMETRIC:
N
NeilBrown 已提交
2079
			pd_idx = sector_div(stripe2, raid_disks);
2080 2081
			qd_idx = (pd_idx + 1) % raid_disks;
			*dd_idx = (pd_idx + 2 + *dd_idx) % raid_disks;
2082
			break;
2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097

		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
2098
			pd_idx = sector_div(stripe2, raid_disks);
2099 2100 2101 2102 2103 2104
			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 */
2105
			ddf_layout = 1;
2106 2107 2108 2109 2110 2111 2112
			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
2113 2114
			stripe2 += 1;
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2115 2116 2117 2118 2119 2120
			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 */
2121
			ddf_layout = 1;
2122 2123 2124 2125
			break;

		case ALGORITHM_ROTATING_N_CONTINUE:
			/* Same as left_symmetric but Q is before P */
N
NeilBrown 已提交
2126
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2127 2128
			qd_idx = (pd_idx + raid_disks - 1) % raid_disks;
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
2129
			ddf_layout = 1;
2130 2131 2132 2133
			break;

		case ALGORITHM_LEFT_ASYMMETRIC_6:
			/* RAID5 left_asymmetric, with Q on last device */
N
NeilBrown 已提交
2134
			pd_idx = data_disks - sector_div(stripe2, raid_disks-1);
2135 2136 2137 2138 2139 2140
			if (*dd_idx >= pd_idx)
				(*dd_idx)++;
			qd_idx = raid_disks - 1;
			break;

		case ALGORITHM_RIGHT_ASYMMETRIC_6:
N
NeilBrown 已提交
2141
			pd_idx = sector_div(stripe2, raid_disks-1);
2142 2143 2144 2145 2146 2147
			if (*dd_idx >= pd_idx)
				(*dd_idx)++;
			qd_idx = raid_disks - 1;
			break;

		case ALGORITHM_LEFT_SYMMETRIC_6:
N
NeilBrown 已提交
2148
			pd_idx = data_disks - sector_div(stripe2, raid_disks-1);
2149 2150 2151 2152 2153
			*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
2154
			pd_idx = sector_div(stripe2, raid_disks-1);
2155 2156 2157 2158 2159 2160 2161 2162 2163 2164
			*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;

2165
		default:
2166
			BUG();
2167 2168
		}
		break;
L
Linus Torvalds 已提交
2169 2170
	}

2171 2172 2173
	if (sh) {
		sh->pd_idx = pd_idx;
		sh->qd_idx = qd_idx;
2174
		sh->ddf_layout = ddf_layout;
2175
	}
L
Linus Torvalds 已提交
2176 2177 2178 2179 2180 2181 2182 2183
	/*
	 * Finally, compute the new sector number
	 */
	new_sector = (sector_t)stripe * sectors_per_chunk + chunk_offset;
	return new_sector;
}


2184
static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous)
L
Linus Torvalds 已提交
2185
{
2186
	struct r5conf *conf = sh->raid_conf;
2187 2188
	int raid_disks = sh->disks;
	int data_disks = raid_disks - conf->max_degraded;
L
Linus Torvalds 已提交
2189
	sector_t new_sector = sh->sector, check;
2190 2191
	int sectors_per_chunk = previous ? conf->prev_chunk_sectors
					 : conf->chunk_sectors;
2192 2193
	int algorithm = previous ? conf->prev_algo
				 : conf->algorithm;
L
Linus Torvalds 已提交
2194 2195
	sector_t stripe;
	int chunk_offset;
2196 2197
	sector_t chunk_number;
	int dummy1, dd_idx = i;
L
Linus Torvalds 已提交
2198
	sector_t r_sector;
2199
	struct stripe_head sh2;
L
Linus Torvalds 已提交
2200

2201

L
Linus Torvalds 已提交
2202 2203 2204
	chunk_offset = sector_div(new_sector, sectors_per_chunk);
	stripe = new_sector;

2205 2206 2207 2208 2209
	if (i == sh->pd_idx)
		return 0;
	switch(conf->level) {
	case 4: break;
	case 5:
2210
		switch (algorithm) {
L
Linus Torvalds 已提交
2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221
		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;
2222 2223 2224 2225 2226
		case ALGORITHM_PARITY_0:
			i -= 1;
			break;
		case ALGORITHM_PARITY_N:
			break;
L
Linus Torvalds 已提交
2227
		default:
2228
			BUG();
2229 2230 2231
		}
		break;
	case 6:
2232
		if (i == sh->qd_idx)
2233
			return 0; /* It is the Q disk */
2234
		switch (algorithm) {
2235 2236
		case ALGORITHM_LEFT_ASYMMETRIC:
		case ALGORITHM_RIGHT_ASYMMETRIC:
2237 2238 2239 2240
		case ALGORITHM_ROTATING_ZERO_RESTART:
		case ALGORITHM_ROTATING_N_RESTART:
			if (sh->pd_idx == raid_disks-1)
				i--;	/* Q D D D P */
2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254
			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;
2255 2256 2257 2258 2259 2260
		case ALGORITHM_PARITY_0:
			i -= 2;
			break;
		case ALGORITHM_PARITY_N:
			break;
		case ALGORITHM_ROTATING_N_CONTINUE:
2261
			/* Like left_symmetric, but P is before Q */
2262 2263
			if (sh->pd_idx == 0)
				i--;	/* P D D D Q */
2264 2265 2266 2267 2268 2269
			else {
				/* D D Q P D */
				if (i < sh->pd_idx)
					i += raid_disks;
				i -= (sh->pd_idx + 1);
			}
2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284
			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;
2285
		default:
2286
			BUG();
2287 2288
		}
		break;
L
Linus Torvalds 已提交
2289 2290 2291
	}

	chunk_number = stripe * data_disks + i;
2292
	r_sector = chunk_number * sectors_per_chunk + chunk_offset;
L
Linus Torvalds 已提交
2293

2294
	check = raid5_compute_sector(conf, r_sector,
2295
				     previous, &dummy1, &sh2);
2296 2297
	if (check != sh->sector || dummy1 != dd_idx || sh2.pd_idx != sh->pd_idx
		|| sh2.qd_idx != sh->qd_idx) {
2298 2299
		printk(KERN_ERR "md/raid:%s: compute_blocknr: map not correct\n",
		       mdname(conf->mddev));
L
Linus Torvalds 已提交
2300 2301 2302 2303 2304 2305
		return 0;
	}
	return r_sector;
}


2306
static void
2307
schedule_reconstruction(struct stripe_head *sh, struct stripe_head_state *s,
2308
			 int rcw, int expand)
2309 2310
{
	int i, pd_idx = sh->pd_idx, disks = sh->disks;
2311
	struct r5conf *conf = sh->raid_conf;
2312
	int level = conf->level;
2313 2314 2315 2316 2317 2318 2319

	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) {
2320 2321 2322 2323
			sh->reconstruct_state = reconstruct_state_drain_run;
			set_bit(STRIPE_OP_BIODRAIN, &s->ops_request);
		} else
			sh->reconstruct_state = reconstruct_state_run;
2324

2325
		set_bit(STRIPE_OP_RECONSTRUCT, &s->ops_request);
2326 2327 2328 2329 2330 2331

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

			if (dev->towrite) {
				set_bit(R5_LOCKED, &dev->flags);
2332
				set_bit(R5_Wantdrain, &dev->flags);
2333 2334
				if (!expand)
					clear_bit(R5_UPTODATE, &dev->flags);
2335
				s->locked++;
2336 2337
			}
		}
2338
		if (s->locked + conf->max_degraded == disks)
2339
			if (!test_and_set_bit(STRIPE_FULL_WRITE, &sh->state))
2340
				atomic_inc(&conf->pending_full_writes);
2341
	} else {
2342
		BUG_ON(level == 6);
2343 2344 2345
		BUG_ON(!(test_bit(R5_UPTODATE, &sh->dev[pd_idx].flags) ||
			test_bit(R5_Wantcompute, &sh->dev[pd_idx].flags)));

2346
		sh->reconstruct_state = reconstruct_state_prexor_drain_run;
2347 2348
		set_bit(STRIPE_OP_PREXOR, &s->ops_request);
		set_bit(STRIPE_OP_BIODRAIN, &s->ops_request);
2349
		set_bit(STRIPE_OP_RECONSTRUCT, &s->ops_request);
2350 2351 2352 2353 2354 2355 2356 2357

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

			if (dev->towrite &&
			    (test_bit(R5_UPTODATE, &dev->flags) ||
2358 2359
			     test_bit(R5_Wantcompute, &dev->flags))) {
				set_bit(R5_Wantdrain, &dev->flags);
2360 2361
				set_bit(R5_LOCKED, &dev->flags);
				clear_bit(R5_UPTODATE, &dev->flags);
2362
				s->locked++;
2363 2364 2365 2366
			}
		}
	}

2367
	/* keep the parity disk(s) locked while asynchronous operations
2368 2369 2370 2371
	 * are in flight
	 */
	set_bit(R5_LOCKED, &sh->dev[pd_idx].flags);
	clear_bit(R5_UPTODATE, &sh->dev[pd_idx].flags);
2372
	s->locked++;
2373

2374 2375 2376 2377 2378 2379 2380 2381 2382
	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++;
	}

2383
	pr_debug("%s: stripe %llu locked: %d ops_request: %lx\n",
2384
		__func__, (unsigned long long)sh->sector,
2385
		s->locked, s->ops_request);
2386
}
2387

L
Linus Torvalds 已提交
2388 2389
/*
 * Each stripe/dev can have one or more bion attached.
2390
 * toread/towrite point to the first in a chain.
L
Linus Torvalds 已提交
2391 2392 2393 2394 2395
 * 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;
2396
	struct r5conf *conf = sh->raid_conf;
2397
	int firstwrite=0;
L
Linus Torvalds 已提交
2398

2399
	pr_debug("adding bi b#%llu to stripe s#%llu\n",
L
Linus Torvalds 已提交
2400 2401 2402
		(unsigned long long)bi->bi_sector,
		(unsigned long long)sh->sector);

S
Shaohua Li 已提交
2403 2404 2405 2406 2407 2408 2409 2410 2411
	/*
	 * If several bio share a stripe. The bio bi_phys_segments acts as a
	 * reference count to avoid race. The reference count should already be
	 * increased before this function is called (for example, in
	 * make_request()), so other bio sharing this stripe will not free the
	 * stripe. If a stripe is owned by one stripe, the stripe lock will
	 * protect it.
	 */
	spin_lock_irq(&sh->stripe_lock);
2412
	if (forwrite) {
L
Linus Torvalds 已提交
2413
		bip = &sh->dev[dd_idx].towrite;
2414
		if (*bip == NULL)
2415 2416
			firstwrite = 1;
	} else
L
Linus Torvalds 已提交
2417 2418 2419 2420 2421 2422 2423 2424 2425
		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;

2426
	BUG_ON(*bip && bi->bi_next && (*bip) != bi->bi_next);
L
Linus Torvalds 已提交
2427 2428 2429
	if (*bip)
		bi->bi_next = *bip;
	*bip = bi;
2430
	raid5_inc_bi_active_stripes(bi);
2431

L
Linus Torvalds 已提交
2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444
	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);
	}
S
Shaohua Li 已提交
2445
	spin_unlock_irq(&sh->stripe_lock);
2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456

	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
2457 2458 2459 2460
	return 1;

 overlap:
	set_bit(R5_Overlap, &sh->dev[dd_idx].flags);
S
Shaohua Li 已提交
2461
	spin_unlock_irq(&sh->stripe_lock);
L
Linus Torvalds 已提交
2462 2463 2464
	return 0;
}

2465
static void end_reshape(struct r5conf *conf);
2466

2467
static void stripe_set_idx(sector_t stripe, struct r5conf *conf, int previous,
2468
			    struct stripe_head *sh)
2469
{
2470
	int sectors_per_chunk =
2471
		previous ? conf->prev_chunk_sectors : conf->chunk_sectors;
2472
	int dd_idx;
2473
	int chunk_offset = sector_div(stripe, sectors_per_chunk);
2474
	int disks = previous ? conf->previous_raid_disks : conf->raid_disks;
2475

2476 2477
	raid5_compute_sector(conf,
			     stripe * (disks - conf->max_degraded)
2478
			     *sectors_per_chunk + chunk_offset,
2479
			     previous,
2480
			     &dd_idx, sh);
2481 2482
}

2483
static void
2484
handle_failed_stripe(struct r5conf *conf, struct stripe_head *sh,
2485 2486 2487 2488 2489 2490 2491 2492 2493
				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)) {
2494
			struct md_rdev *rdev;
2495 2496 2497
			rcu_read_lock();
			rdev = rcu_dereference(conf->disks[i].rdev);
			if (rdev && test_bit(In_sync, &rdev->flags))
2498 2499 2500
				atomic_inc(&rdev->nr_pending);
			else
				rdev = NULL;
2501
			rcu_read_unlock();
2502 2503 2504 2505 2506 2507 2508 2509
			if (rdev) {
				if (!rdev_set_badblocks(
					    rdev,
					    sh->sector,
					    STRIPE_SECTORS, 0))
					md_error(conf->mddev, rdev);
				rdev_dec_pending(rdev, conf->mddev);
			}
2510
		}
S
Shaohua Li 已提交
2511
		spin_lock_irq(&sh->stripe_lock);
2512 2513 2514
		/* fail all writes first */
		bi = sh->dev[i].towrite;
		sh->dev[i].towrite = NULL;
S
Shaohua Li 已提交
2515
		spin_unlock_irq(&sh->stripe_lock);
2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527
		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);
2528
			if (!raid5_dec_bi_active_stripes(bi)) {
2529 2530 2531 2532 2533 2534
				md_write_end(conf->mddev);
				bi->bi_next = *return_bi;
				*return_bi = bi;
			}
			bi = nextbi;
		}
2535 2536 2537 2538
		if (bitmap_end)
			bitmap_endwrite(conf->mddev->bitmap, sh->sector,
				STRIPE_SECTORS, 0, 0);
		bitmap_end = 0;
2539 2540 2541 2542 2543 2544 2545 2546
		/* 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);
2547
			if (!raid5_dec_bi_active_stripes(bi)) {
2548 2549 2550 2551 2552 2553 2554
				md_write_end(conf->mddev);
				bi->bi_next = *return_bi;
				*return_bi = bi;
			}
			bi = bi2;
		}

2555 2556 2557 2558 2559 2560
		/* 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))) {
2561 2562 2563 2564 2565 2566 2567 2568 2569 2570
			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);
2571
				if (!raid5_dec_bi_active_stripes(bi)) {
2572 2573 2574 2575 2576 2577 2578 2579 2580
					bi->bi_next = *return_bi;
					*return_bi = bi;
				}
				bi = nextbi;
			}
		}
		if (bitmap_end)
			bitmap_endwrite(conf->mddev->bitmap, sh->sector,
					STRIPE_SECTORS, 0, 0);
2581 2582 2583 2584
		/* 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);
2585 2586
	}

2587 2588 2589
	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);
2590 2591
}

2592
static void
2593
handle_failed_sync(struct r5conf *conf, struct stripe_head *sh,
2594 2595 2596 2597 2598 2599 2600
		   struct stripe_head_state *s)
{
	int abort = 0;
	int i;

	clear_bit(STRIPE_SYNCING, &sh->state);
	s->syncing = 0;
2601
	s->replacing = 0;
2602
	/* There is nothing more to do for sync/check/repair.
2603 2604 2605
	 * 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.
2606
	 * For recover/replace we need to record a bad block on all
2607 2608
	 * non-sync devices, or abort the recovery
	 */
2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631
	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;
2632
	}
2633
	md_done_sync(conf->mddev, STRIPE_SECTORS, !abort);
2634 2635
}

2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651
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;
}

2652
/* fetch_block - checks the given member device to see if its data needs
2653 2654 2655
 * to be read or computed to satisfy a request.
 *
 * Returns 1 when no more member devices need to be checked, otherwise returns
2656
 * 0 to tell the loop in handle_stripe_fill to continue
2657
 */
2658 2659
static int fetch_block(struct stripe_head *sh, struct stripe_head_state *s,
		       int disk_idx, int disks)
2660
{
2661
	struct r5dev *dev = &sh->dev[disk_idx];
2662 2663
	struct r5dev *fdev[2] = { &sh->dev[s->failed_num[0]],
				  &sh->dev[s->failed_num[1]] };
2664

2665
	/* is the data in this block needed, and can we get it? */
2666 2667 2668 2669 2670
	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 ||
2671
	     (s->replacing && want_replace(sh, disk_idx)) ||
2672 2673
	     (s->failed >= 1 && fdev[0]->toread) ||
	     (s->failed >= 2 && fdev[1]->toread) ||
2674 2675 2676
	     (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))) {
2677 2678 2679 2680 2681 2682
		/* 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) &&
2683 2684
		    (s->failed && (disk_idx == s->failed_num[0] ||
				   disk_idx == s->failed_num[1]))) {
2685 2686
			/* have disk failed, and we're requested to fetch it;
			 * do compute it
2687
			 */
2688 2689 2690 2691 2692 2693 2694 2695
			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;
2696 2697 2698 2699 2700 2701
			/* 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.
			 */
2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714
			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;
2715
			}
2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734
			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);
2735 2736
		}
	}
2737 2738 2739 2740 2741

	return 0;
}

/**
2742
 * handle_stripe_fill - read or compute data to satisfy pending requests.
2743
 */
2744 2745 2746
static void handle_stripe_fill(struct stripe_head *sh,
			       struct stripe_head_state *s,
			       int disks)
2747 2748 2749 2750 2751 2752 2753 2754 2755 2756
{
	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--; )
2757
			if (fetch_block(sh, s, i, disks))
2758
				break;
2759 2760 2761 2762
	set_bit(STRIPE_HANDLE, &sh->state);
}


2763
/* handle_stripe_clean_event
2764 2765 2766 2767
 * 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.
 */
2768
static void handle_stripe_clean_event(struct r5conf *conf,
2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780
	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;
2781
				pr_debug("Return write for disc %d\n", i);
2782 2783 2784 2785 2786
				wbi = dev->written;
				dev->written = NULL;
				while (wbi && wbi->bi_sector <
					dev->sector + STRIPE_SECTORS) {
					wbi2 = r5_next_bio(wbi, dev->sector);
2787
					if (!raid5_dec_bi_active_stripes(wbi)) {
2788 2789 2790 2791 2792 2793
						md_write_end(conf->mddev);
						wbi->bi_next = *return_bi;
						*return_bi = wbi;
					}
					wbi = wbi2;
				}
2794 2795
				bitmap_endwrite(conf->mddev->bitmap, sh->sector,
						STRIPE_SECTORS,
2796
					 !test_bit(STRIPE_DEGRADED, &sh->state),
2797
						0);
2798 2799
			}
		}
2800 2801 2802 2803

	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);
2804 2805
}

2806
static void handle_stripe_dirtying(struct r5conf *conf,
2807 2808 2809
				   struct stripe_head *sh,
				   struct stripe_head_state *s,
				   int disks)
2810 2811
{
	int rmw = 0, rcw = 0, i;
2812 2813 2814 2815 2816 2817 2818
	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--; ) {
2819 2820 2821 2822
		/* 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) &&
2823 2824
		    !(test_bit(R5_UPTODATE, &dev->flags) ||
		      test_bit(R5_Wantcompute, &dev->flags))) {
2825 2826 2827 2828 2829 2830 2831 2832
			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) &&
2833 2834 2835
		    !(test_bit(R5_UPTODATE, &dev->flags) ||
		    test_bit(R5_Wantcompute, &dev->flags))) {
			if (test_bit(R5_Insync, &dev->flags)) rcw++;
2836 2837 2838 2839
			else
				rcw += 2*disks;
		}
	}
2840
	pr_debug("for sector %llu, rmw=%d rcw=%d\n",
2841 2842 2843 2844 2845 2846 2847 2848
		(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) &&
2849 2850
			    !(test_bit(R5_UPTODATE, &dev->flags) ||
			    test_bit(R5_Wantcompute, &dev->flags)) &&
2851 2852 2853
			    test_bit(R5_Insync, &dev->flags)) {
				if (
				  test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
2854
					pr_debug("Read_old block "
2855 2856 2857 2858 2859 2860 2861 2862 2863 2864
						"%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);
				}
			}
		}
2865
	if (rcw <= rmw && rcw > 0) {
2866
		/* want reconstruct write, but need to get some data */
2867
		rcw = 0;
2868 2869 2870
		for (i = disks; i--; ) {
			struct r5dev *dev = &sh->dev[i];
			if (!test_bit(R5_OVERWRITE, &dev->flags) &&
2871
			    i != sh->pd_idx && i != sh->qd_idx &&
2872
			    !test_bit(R5_LOCKED, &dev->flags) &&
2873
			    !(test_bit(R5_UPTODATE, &dev->flags) ||
2874 2875 2876 2877
			      test_bit(R5_Wantcompute, &dev->flags))) {
				rcw++;
				if (!test_bit(R5_Insync, &dev->flags))
					continue; /* it's a failed drive */
2878 2879
				if (
				  test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
2880
					pr_debug("Read_old block "
2881 2882 2883 2884 2885 2886 2887 2888 2889 2890
						"%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);
				}
			}
		}
2891
	}
2892 2893 2894
	/* now if nothing is locked, and if we have enough data,
	 * we can start a write request
	 */
2895 2896
	/* 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
2897 2898
	 * subsequent call wants to start a write request.  raid_run_ops only
	 * handles the case where compute block and reconstruct are requested
2899 2900 2901
	 * simultaneously.  If this is not the case then new writes need to be
	 * held off until the compute completes.
	 */
2902 2903 2904
	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)))
2905
		schedule_reconstruction(sh, s, rcw == 0, 0);
2906 2907
}

2908
static void handle_parity_checks5(struct r5conf *conf, struct stripe_head *sh,
2909 2910
				struct stripe_head_state *s, int disks)
{
2911
	struct r5dev *dev = NULL;
2912

2913
	set_bit(STRIPE_HANDLE, &sh->state);
2914

2915 2916 2917
	switch (sh->check_state) {
	case check_state_idle:
		/* start a new check operation if there are no failures */
2918 2919
		if (s->failed == 0) {
			BUG_ON(s->uptodate != disks);
2920 2921
			sh->check_state = check_state_run;
			set_bit(STRIPE_OP_CHECK, &s->ops_request);
2922 2923
			clear_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags);
			s->uptodate--;
2924
			break;
2925
		}
2926
		dev = &sh->dev[s->failed_num[0]];
2927 2928 2929 2930 2931 2932 2933 2934 2935
		/* 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
2936

2937 2938 2939 2940 2941
		/* 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);
2942
		s->locked++;
2943
		set_bit(R5_Wantwrite, &dev->flags);
2944

2945 2946
		clear_bit(STRIPE_DEGRADED, &sh->state);
		set_bit(STRIPE_INSYNC, &sh->state);
2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962
		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
2963
		if ((sh->ops.zero_sum_result & SUM_CHECK_P_RESULT) == 0)
2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974
			/* 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;
2975
				set_bit(STRIPE_COMPUTE_RUN, &sh->state);
2976 2977 2978 2979
				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;
2980
				sh->ops.target2 = -1;
2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991
				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();
2992 2993 2994 2995
	}
}


2996
static void handle_parity_checks6(struct r5conf *conf, struct stripe_head *sh,
2997
				  struct stripe_head_state *s,
2998
				  int disks)
2999 3000
{
	int pd_idx = sh->pd_idx;
N
NeilBrown 已提交
3001
	int qd_idx = sh->qd_idx;
3002
	struct r5dev *dev;
3003 3004 3005 3006

	set_bit(STRIPE_HANDLE, &sh->state);

	BUG_ON(s->failed > 2);
3007

3008 3009 3010 3011 3012 3013
	/* 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
	 */

3014 3015 3016
	switch (sh->check_state) {
	case check_state_idle:
		/* start a new check operation if there are < 2 failures */
3017
		if (s->failed == s->q_failed) {
3018
			/* The only possible failed device holds Q, so it
3019 3020 3021
			 * makes sense to check P (If anything else were failed,
			 * we would have used P to recreate it).
			 */
3022
			sh->check_state = check_state_run;
3023
		}
3024
		if (!s->q_failed && s->failed < 2) {
3025
			/* Q is not failed, and we didn't use it to generate
3026 3027
			 * anything, so it makes sense to check it
			 */
3028 3029 3030 3031
			if (sh->check_state == check_state_run)
				sh->check_state = check_state_run_pq;
			else
				sh->check_state = check_state_run_q;
3032 3033
		}

3034 3035
		/* discard potentially stale zero_sum_result */
		sh->ops.zero_sum_result = 0;
3036

3037 3038 3039 3040
		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--;
3041
		}
3042 3043 3044 3045 3046 3047 3048
		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;
3049 3050
		}

3051 3052 3053 3054 3055
		/* we have 2-disk failure */
		BUG_ON(s->failed != 2);
		/* fall through */
	case check_state_compute_result:
		sh->check_state = check_state_idle;
3056

3057 3058 3059
		/* check that a write has not made the stripe insync */
		if (test_bit(STRIPE_INSYNC, &sh->state))
			break;
3060 3061

		/* now write out any block on a failed drive,
3062
		 * or P or Q if they were recomputed
3063
		 */
3064
		BUG_ON(s->uptodate < disks - 1); /* We don't need Q to recover */
3065
		if (s->failed == 2) {
3066
			dev = &sh->dev[s->failed_num[1]];
3067 3068 3069 3070 3071
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
		if (s->failed >= 1) {
3072
			dev = &sh->dev[s->failed_num[0]];
3073 3074 3075 3076
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
3077
		if (sh->ops.zero_sum_result & SUM_CHECK_P_RESULT) {
3078 3079 3080 3081 3082
			dev = &sh->dev[pd_idx];
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
3083
		if (sh->ops.zero_sum_result & SUM_CHECK_Q_RESULT) {
3084 3085 3086 3087 3088 3089 3090 3091
			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);
3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155
		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();
3156 3157 3158
	}
}

3159
static void handle_stripe_expansion(struct r5conf *conf, struct stripe_head *sh)
3160 3161 3162 3163 3164 3165
{
	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.
	 */
3166
	struct dma_async_tx_descriptor *tx = NULL;
3167 3168
	clear_bit(STRIPE_EXPAND_SOURCE, &sh->state);
	for (i = 0; i < sh->disks; i++)
N
NeilBrown 已提交
3169
		if (i != sh->pd_idx && i != sh->qd_idx) {
3170
			int dd_idx, j;
3171
			struct stripe_head *sh2;
3172
			struct async_submit_ctl submit;
3173

3174
			sector_t bn = compute_blocknr(sh, i, 1);
3175 3176
			sector_t s = raid5_compute_sector(conf, bn, 0,
							  &dd_idx, NULL);
3177
			sh2 = get_active_stripe(conf, s, 0, 1, 1);
3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189
			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;
			}
3190 3191

			/* place all the copies on one channel */
3192
			init_async_submit(&submit, 0, tx, NULL, NULL, NULL);
3193
			tx = async_memcpy(sh2->dev[dd_idx].page,
3194
					  sh->dev[i].page, 0, 0, STRIPE_SIZE,
3195
					  &submit);
3196

3197 3198 3199 3200
			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 &&
3201
				    j != sh2->qd_idx &&
3202 3203 3204 3205 3206 3207 3208
				    !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);
3209

3210
		}
3211 3212 3213 3214 3215
	/* done submitting copies, wait for them to complete */
	if (tx) {
		async_tx_ack(tx);
		dma_wait_for_async_tx(tx);
	}
3216
}
L
Linus Torvalds 已提交
3217 3218 3219 3220

/*
 * handle_stripe - do things to a stripe.
 *
3221 3222
 * 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
3223
 * Possible results:
3224 3225
 *    return some read requests which now have data
 *    return some write requests which are safely on storage
L
Linus Torvalds 已提交
3226 3227 3228 3229 3230
 *    schedule a read on some buffers
 *    schedule a write of some buffers
 *    return confirmation of parity correctness
 *
 */
3231

3232
static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
L
Linus Torvalds 已提交
3233
{
3234
	struct r5conf *conf = sh->raid_conf;
3235
	int disks = sh->disks;
3236 3237
	struct r5dev *dev;
	int i;
3238
	int do_recovery = 0;
L
Linus Torvalds 已提交
3239

3240 3241 3242 3243 3244 3245
	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
3246

3247
	/* Now to look around and see what can be done */
L
Linus Torvalds 已提交
3248
	rcu_read_lock();
3249
	for (i=disks; i--; ) {
3250
		struct md_rdev *rdev;
3251 3252 3253
		sector_t first_bad;
		int bad_sectors;
		int is_bad = 0;
3254

3255
		dev = &sh->dev[i];
L
Linus Torvalds 已提交
3256

3257
		pr_debug("check %d: state 0x%lx read %p write %p written %p\n",
3258 3259
			 i, dev->flags,
			 dev->toread, dev->towrite, dev->written);
3260 3261 3262 3263 3264 3265 3266 3267
		/* 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
3268

3269
		/* now count some things */
3270 3271 3272 3273
		if (test_bit(R5_LOCKED, &dev->flags))
			s->locked++;
		if (test_bit(R5_UPTODATE, &dev->flags))
			s->uptodate++;
3274
		if (test_bit(R5_Wantcompute, &dev->flags)) {
3275 3276
			s->compute++;
			BUG_ON(s->compute > 2);
3277
		}
L
Linus Torvalds 已提交
3278

3279
		if (test_bit(R5_Wantfill, &dev->flags))
3280
			s->to_fill++;
3281
		else if (dev->toread)
3282
			s->to_read++;
3283
		if (dev->towrite) {
3284
			s->to_write++;
3285
			if (!test_bit(R5_OVERWRITE, &dev->flags))
3286
				s->non_overwrite++;
3287
		}
3288
		if (dev->written)
3289
			s->written++;
3290 3291 3292 3293 3294 3295 3296 3297 3298 3299
		/* 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 {
3300 3301
			if (rdev)
				set_bit(R5_NeedReplace, &dev->flags);
3302 3303 3304
			rdev = rcu_dereference(conf->disks[i].rdev);
			clear_bit(R5_ReadRepl, &dev->flags);
		}
3305 3306
		if (rdev && test_bit(Faulty, &rdev->flags))
			rdev = NULL;
3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318
		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);
			}
3319
		}
3320 3321 3322
		clear_bit(R5_Insync, &dev->flags);
		if (!rdev)
			/* Not in-sync */;
3323 3324
		else if (is_bad) {
			/* also not in-sync */
3325 3326
			if (!test_bit(WriteErrorSeen, &rdev->flags) &&
			    test_bit(R5_UPTODATE, &dev->flags)) {
3327 3328 3329 3330 3331 3332 3333
				/* 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))
3334
			set_bit(R5_Insync, &dev->flags);
3335
		else if (sh->sector + STRIPE_SECTORS <= rdev->recovery_offset)
3336
			/* in sync if before recovery_offset */
3337 3338 3339 3340 3341 3342 3343 3344 3345
			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
3346
		if (rdev && test_bit(R5_WriteError, &dev->flags)) {
3347 3348 3349 3350 3351 3352 3353
			/* 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)) {
3354
				s->handle_bad_blocks = 1;
3355
				atomic_inc(&rdev2->nr_pending);
3356 3357 3358
			} else
				clear_bit(R5_WriteError, &dev->flags);
		}
A
Adam Kwolek 已提交
3359
		if (rdev && test_bit(R5_MadeGood, &dev->flags)) {
3360 3361 3362 3363 3364
			/* 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)) {
3365
				s->handle_bad_blocks = 1;
3366
				atomic_inc(&rdev2->nr_pending);
3367 3368 3369
			} else
				clear_bit(R5_MadeGood, &dev->flags);
		}
3370 3371 3372 3373 3374 3375 3376 3377 3378
		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);
		}
3379
		if (!test_bit(R5_Insync, &dev->flags)) {
3380 3381 3382
			/* 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
3383
		}
3384 3385 3386
		if (test_bit(R5_ReadError, &dev->flags))
			clear_bit(R5_Insync, &dev->flags);
		if (!test_bit(R5_Insync, &dev->flags)) {
3387 3388 3389
			if (s->failed < 2)
				s->failed_num[s->failed] = i;
			s->failed++;
3390 3391
			if (rdev && !test_bit(Faulty, &rdev->flags))
				do_recovery = 1;
3392
		}
L
Linus Torvalds 已提交
3393
	}
3394 3395 3396 3397
	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
3398
		 * else if MD_RECOVERY_REQUESTED is set, we also are syncing.
3399 3400 3401 3402 3403
		 * else we can only be replacing
		 * sync and recovery both need to read all devices, and so
		 * use the same flag.
		 */
		if (do_recovery ||
3404 3405
		    sh->sector >= conf->mddev->recovery_cp ||
		    test_bit(MD_RECOVERY_REQUESTED, &(conf->mddev->recovery)))
3406 3407 3408 3409
			s->syncing = 1;
		else
			s->replacing = 1;
	}
L
Linus Torvalds 已提交
3410
	rcu_read_unlock();
3411 3412 3413 3414 3415
}

static void handle_stripe(struct stripe_head *sh)
{
	struct stripe_head_state s;
3416
	struct r5conf *conf = sh->raid_conf;
3417
	int i;
3418 3419
	int prexor;
	int disks = sh->disks;
3420
	struct r5dev *pdev, *qdev;
3421 3422

	clear_bit(STRIPE_HANDLE, &sh->state);
3423
	if (test_and_set_bit_lock(STRIPE_ACTIVE, &sh->state)) {
3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440
		/* 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);
3441

3442
	analyse_stripe(sh, &s);
3443

3444 3445 3446 3447 3448
	if (s.handle_bad_blocks) {
		set_bit(STRIPE_HANDLE, &sh->state);
		goto finish;
	}

3449 3450
	if (unlikely(s.blocked_rdev)) {
		if (s.syncing || s.expanding || s.expanded ||
3451
		    s.replacing || s.to_write || s.written) {
3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471
			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.
	 */
3472 3473 3474 3475 3476
	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);
3477
		if (s.syncing + s.replacing)
3478 3479
			handle_failed_sync(conf, sh, &s);
	}
3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507

	/*
	 * 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)
3508 3509 3510
	    || (s.syncing && (s.uptodate + s.compute < disks))
	    || s.replacing
	    || s.expanding)
3511 3512
		handle_stripe_fill(sh, &s, disks);

3513 3514 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 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
	/* 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);
	}
3571

3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585
	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)) {
3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614
		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++;
				}
			}
		}


3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641
	/* 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++;
		}
	}
3642

3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658
	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);
3659

3660
finish:
3661
	/* wait for this device to become unblocked */
3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673
	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);
	}
3674

3675 3676
	if (s.handle_bad_blocks)
		for (i = disks; i--; ) {
3677
			struct md_rdev *rdev;
3678 3679 3680 3681 3682 3683 3684 3685 3686
			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);
			}
3687 3688 3689
			if (test_and_clear_bit(R5_MadeGood, &dev->flags)) {
				rdev = conf->disks[i].rdev;
				rdev_clear_badblocks(rdev, sh->sector,
3690
						     STRIPE_SECTORS, 0);
3691 3692
				rdev_dec_pending(rdev, conf->mddev);
			}
3693 3694
			if (test_and_clear_bit(R5_MadeGoodRepl, &dev->flags)) {
				rdev = conf->disks[i].replacement;
3695 3696 3697
				if (!rdev)
					/* rdev have been moved down */
					rdev = conf->disks[i].rdev;
3698
				rdev_clear_badblocks(rdev, sh->sector,
3699
						     STRIPE_SECTORS, 0);
3700 3701
				rdev_dec_pending(rdev, conf->mddev);
			}
3702 3703
		}

3704 3705 3706
	if (s.ops_request)
		raid_run_ops(sh, s.ops_request);

D
Dan Williams 已提交
3707
	ops_run_io(sh, &s);
3708

3709
	if (s.dec_preread_active) {
3710
		/* We delay this until after ops_run_io so that if make_request
T
Tejun Heo 已提交
3711
		 * is waiting on a flush, it won't continue until the writes
3712 3713 3714 3715 3716 3717 3718 3719
		 * 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);
	}

3720
	return_io(s.return_bi);
3721

3722
	clear_bit_unlock(STRIPE_ACTIVE, &sh->state);
3723 3724
}

3725
static void raid5_activate_delayed(struct r5conf *conf)
3726 3727 3728 3729 3730 3731 3732 3733 3734 3735
{
	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);
3736
			list_add_tail(&sh->lru, &conf->hold_list);
3737
		}
N
NeilBrown 已提交
3738
	}
3739 3740
}

3741
static void activate_bit_delay(struct r5conf *conf)
3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754
{
	/* 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);
	}
}

3755
int md_raid5_congested(struct mddev *mddev, int bits)
3756
{
3757
	struct r5conf *conf = mddev->private;
3758 3759 3760 3761

	/* No difference between reads and writes.  Just check
	 * how busy the stripe_cache is
	 */
3762

3763 3764 3765 3766 3767 3768 3769 3770 3771
	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
3772 3773 3774 3775
EXPORT_SYMBOL_GPL(md_raid5_congested);

static int raid5_congested(void *data, int bits)
{
3776
	struct mddev *mddev = data;
N
NeilBrown 已提交
3777 3778 3779 3780

	return mddev_congested(mddev, bits) ||
		md_raid5_congested(mddev, bits);
}
3781

3782 3783 3784
/* We want read requests to align with chunks where possible,
 * but write requests don't need to.
 */
3785 3786 3787
static int raid5_mergeable_bvec(struct request_queue *q,
				struct bvec_merge_data *bvm,
				struct bio_vec *biovec)
3788
{
3789
	struct mddev *mddev = q->queuedata;
3790
	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
3791
	int max;
3792
	unsigned int chunk_sectors = mddev->chunk_sectors;
3793
	unsigned int bio_sectors = bvm->bi_size >> 9;
3794

3795
	if ((bvm->bi_rw & 1) == WRITE)
3796 3797
		return biovec->bv_len; /* always allow writes to be mergeable */

3798 3799
	if (mddev->new_chunk_sectors < mddev->chunk_sectors)
		chunk_sectors = mddev->new_chunk_sectors;
3800 3801 3802 3803 3804 3805 3806 3807
	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;
}

3808

3809
static int in_chunk_boundary(struct mddev *mddev, struct bio *bio)
3810 3811
{
	sector_t sector = bio->bi_sector + get_start_sect(bio->bi_bdev);
3812
	unsigned int chunk_sectors = mddev->chunk_sectors;
3813 3814
	unsigned int bio_sectors = bio->bi_size >> 9;

3815 3816
	if (mddev->new_chunk_sectors < mddev->chunk_sectors)
		chunk_sectors = mddev->new_chunk_sectors;
3817 3818 3819 3820
	return  chunk_sectors >=
		((sector & (chunk_sectors - 1)) + bio_sectors);
}

3821 3822 3823 3824
/*
 *  add bio to the retry LIFO  ( in O(1) ... we are in interrupt )
 *  later sampled by raid5d.
 */
3825
static void add_bio_to_retry(struct bio *bi,struct r5conf *conf)
3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838
{
	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);
}


3839
static struct bio *remove_bio_from_retry(struct r5conf *conf)
3840 3841 3842 3843 3844 3845 3846 3847 3848 3849
{
	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) {
3850
		conf->retry_read_aligned_list = bi->bi_next;
3851
		bi->bi_next = NULL;
3852 3853 3854 3855
		/*
		 * this sets the active strip count to 1 and the processed
		 * strip count to zero (upper 8 bits)
		 */
3856
		raid5_set_bi_stripes(bi, 1); /* biased count of active stripes */
3857 3858 3859 3860 3861 3862
	}

	return bi;
}


3863 3864 3865 3866 3867 3868
/*
 *  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..
 */
3869
static void raid5_align_endio(struct bio *bi, int error)
3870 3871
{
	struct bio* raid_bi  = bi->bi_private;
3872
	struct mddev *mddev;
3873
	struct r5conf *conf;
3874
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
3875
	struct md_rdev *rdev;
3876

3877
	bio_put(bi);
3878 3879 3880

	rdev = (void*)raid_bi->bi_next;
	raid_bi->bi_next = NULL;
3881 3882
	mddev = rdev->mddev;
	conf = mddev->private;
3883 3884 3885 3886

	rdev_dec_pending(rdev, conf->mddev);

	if (!error && uptodate) {
3887
		bio_endio(raid_bi, 0);
3888 3889
		if (atomic_dec_and_test(&conf->active_aligned_reads))
			wake_up(&conf->wait_for_stripe);
3890
		return;
3891 3892 3893
	}


3894
	pr_debug("raid5_align_endio : io error...handing IO for a retry\n");
3895 3896

	add_bio_to_retry(raid_bi, conf);
3897 3898
}

3899 3900
static int bio_fits_rdev(struct bio *bi)
{
3901
	struct request_queue *q = bdev_get_queue(bi->bi_bdev);
3902

3903
	if ((bi->bi_size>>9) > queue_max_sectors(q))
3904 3905
		return 0;
	blk_recount_segments(q, bi);
3906
	if (bi->bi_phys_segments > queue_max_segments(q))
3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918
		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;
}


3919
static int chunk_aligned_read(struct mddev *mddev, struct bio * raid_bio)
3920
{
3921
	struct r5conf *conf = mddev->private;
N
NeilBrown 已提交
3922
	int dd_idx;
3923
	struct bio* align_bi;
3924
	struct md_rdev *rdev;
3925
	sector_t end_sector;
3926 3927

	if (!in_chunk_boundary(mddev, raid_bio)) {
3928
		pr_debug("chunk_aligned_read : non aligned\n");
3929 3930 3931
		return 0;
	}
	/*
3932
	 * use bio_clone_mddev to make a copy of the bio
3933
	 */
3934
	align_bi = bio_clone_mddev(raid_bio, GFP_NOIO, mddev);
3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945
	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
	 */
3946 3947
	align_bi->bi_sector =  raid5_compute_sector(conf, raid_bio->bi_sector,
						    0,
3948
						    &dd_idx, NULL);
3949

3950
	end_sector = align_bi->bi_sector + (align_bi->bi_size >> 9);
3951
	rcu_read_lock();
3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962
	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) {
3963 3964 3965
		sector_t first_bad;
		int bad_sectors;

3966 3967
		atomic_inc(&rdev->nr_pending);
		rcu_read_unlock();
3968 3969 3970 3971
		raid_bio->bi_next = (void*)rdev;
		align_bi->bi_bdev =  rdev->bdev;
		align_bi->bi_flags &= ~(1 << BIO_SEG_VALID);

3972 3973 3974 3975
		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 */
3976 3977 3978 3979 3980
			bio_put(align_bi);
			rdev_dec_pending(rdev, mddev);
			return 0;
		}

3981 3982 3983
		/* No reshape active, so we can trust rdev->data_offset */
		align_bi->bi_sector += rdev->data_offset;

3984 3985 3986 3987 3988 3989 3990
		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);

3991 3992 3993 3994
		generic_make_request(align_bi);
		return 1;
	} else {
		rcu_read_unlock();
3995
		bio_put(align_bi);
3996 3997 3998 3999
		return 0;
	}
}

4000 4001 4002 4003 4004 4005 4006 4007 4008 4009
/* __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.
 */
4010
static struct stripe_head *__get_priority_stripe(struct r5conf *conf)
4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051
{
	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;
}
4052

4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108
struct raid5_plug_cb {
	struct blk_plug_cb	cb;
	struct list_head	list;
};

static void raid5_unplug(struct blk_plug_cb *blk_cb, bool from_schedule)
{
	struct raid5_plug_cb *cb = container_of(
		blk_cb, struct raid5_plug_cb, cb);
	struct stripe_head *sh;
	struct mddev *mddev = cb->cb.data;
	struct r5conf *conf = mddev->private;

	if (cb->list.next && !list_empty(&cb->list)) {
		spin_lock_irq(&conf->device_lock);
		while (!list_empty(&cb->list)) {
			sh = list_first_entry(&cb->list, struct stripe_head, lru);
			list_del_init(&sh->lru);
			/*
			 * avoid race release_stripe_plug() sees
			 * STRIPE_ON_UNPLUG_LIST clear but the stripe
			 * is still in our list
			 */
			smp_mb__before_clear_bit();
			clear_bit(STRIPE_ON_UNPLUG_LIST, &sh->state);
			__release_stripe(conf, sh);
		}
		spin_unlock_irq(&conf->device_lock);
	}
	kfree(cb);
}

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

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

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

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

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

S
Shaohua Li 已提交
4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190
static void make_discard_request(struct mddev *mddev, struct bio *bi)
{
	struct r5conf *conf = mddev->private;
	sector_t logical_sector, last_sector;
	struct stripe_head *sh;
	int remaining;
	int stripe_sectors;

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

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

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

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

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

	for (; logical_sector < last_sector;
	     logical_sector += STRIPE_SECTORS) {
		DEFINE_WAIT(w);
		int d;
	again:
		sh = get_active_stripe(conf, logical_sector, 0, 0, 0);
		prepare_to_wait(&conf->wait_for_overlap, &w,
				TASK_UNINTERRUPTIBLE);
		spin_lock_irq(&sh->stripe_lock);
		for (d = 0; d < conf->raid_disks; d++) {
			if (d == sh->pd_idx || d == sh->qd_idx)
				continue;
			if (sh->dev[d].towrite || sh->dev[d].toread) {
				set_bit(R5_Overlap, &sh->dev[d].flags);
				spin_unlock_irq(&sh->stripe_lock);
				release_stripe(sh);
				schedule();
				goto again;
			}
		}
		finish_wait(&conf->wait_for_overlap, &w);
		for (d = 0; d < conf->raid_disks; d++) {
			if (d == sh->pd_idx || d == sh->qd_idx)
				continue;
			sh->dev[d].towrite = bi;
			set_bit(R5_OVERWRITE, &sh->dev[d].flags);
			raid5_inc_bi_active_stripes(bi);
		}
		spin_unlock_irq(&sh->stripe_lock);
		if (conf->mddev->bitmap) {
			for (d = 0;
			     d < conf->raid_disks - conf->max_degraded;
			     d++)
				bitmap_startwrite(mddev->bitmap,
						  sh->sector,
						  STRIPE_SECTORS,
						  0);
			sh->bm_seq = conf->seq_flush + 1;
			set_bit(STRIPE_BIT_DELAY, &sh->state);
		}

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

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

4191
static void make_request(struct mddev *mddev, struct bio * bi)
L
Linus Torvalds 已提交
4192
{
4193
	struct r5conf *conf = mddev->private;
4194
	int dd_idx;
L
Linus Torvalds 已提交
4195 4196 4197
	sector_t new_sector;
	sector_t logical_sector, last_sector;
	struct stripe_head *sh;
4198
	const int rw = bio_data_dir(bi);
4199
	int remaining;
L
Linus Torvalds 已提交
4200

T
Tejun Heo 已提交
4201 4202
	if (unlikely(bi->bi_rw & REQ_FLUSH)) {
		md_flush_request(mddev, bi);
4203
		return;
4204 4205
	}

4206
	md_write_start(mddev, bi);
4207

4208
	if (rw == READ &&
4209
	     mddev->reshape_position == MaxSector &&
4210
	     chunk_aligned_read(mddev,bi))
4211
		return;
4212

S
Shaohua Li 已提交
4213 4214 4215 4216 4217
	if (unlikely(bi->bi_rw & REQ_DISCARD)) {
		make_discard_request(mddev, bi);
		return;
	}

L
Linus Torvalds 已提交
4218 4219 4220 4221
	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 */
4222

L
Linus Torvalds 已提交
4223 4224
	for (;logical_sector < last_sector; logical_sector += STRIPE_SECTORS) {
		DEFINE_WAIT(w);
4225
		int previous;
4226

4227
	retry:
4228
		previous = 0;
4229
		prepare_to_wait(&conf->wait_for_overlap, &w, TASK_UNINTERRUPTIBLE);
4230
		if (unlikely(conf->reshape_progress != MaxSector)) {
4231
			/* spinlock is needed as reshape_progress may be
4232 4233
			 * 64bit on a 32bit platform, and so it might be
			 * possible to see a half-updated value
4234
			 * Of course reshape_progress could change after
4235 4236 4237 4238
			 * the lock is dropped, so once we get a reference
			 * to the stripe that we think it is, we will have
			 * to check again.
			 */
4239
			spin_lock_irq(&conf->device_lock);
4240
			if (mddev->reshape_backwards
4241 4242
			    ? logical_sector < conf->reshape_progress
			    : logical_sector >= conf->reshape_progress) {
4243 4244
				previous = 1;
			} else {
4245
				if (mddev->reshape_backwards
4246 4247
				    ? logical_sector < conf->reshape_safe
				    : logical_sector >= conf->reshape_safe) {
4248 4249 4250 4251 4252
					spin_unlock_irq(&conf->device_lock);
					schedule();
					goto retry;
				}
			}
4253 4254
			spin_unlock_irq(&conf->device_lock);
		}
4255

4256 4257
		new_sector = raid5_compute_sector(conf, logical_sector,
						  previous,
4258
						  &dd_idx, NULL);
4259
		pr_debug("raid456: make_request, sector %llu logical %llu\n",
L
Linus Torvalds 已提交
4260 4261 4262
			(unsigned long long)new_sector, 
			(unsigned long long)logical_sector);

4263
		sh = get_active_stripe(conf, new_sector, previous,
4264
				       (bi->bi_rw&RWA_MASK), 0);
L
Linus Torvalds 已提交
4265
		if (sh) {
4266
			if (unlikely(previous)) {
4267
				/* expansion might have moved on while waiting for a
4268 4269 4270 4271 4272 4273
				 * 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.
4274 4275 4276
				 */
				int must_retry = 0;
				spin_lock_irq(&conf->device_lock);
4277
				if (mddev->reshape_backwards
4278 4279
				    ? logical_sector >= conf->reshape_progress
				    : logical_sector < conf->reshape_progress)
4280 4281 4282 4283 4284
					/* mismatch, need to try again */
					must_retry = 1;
				spin_unlock_irq(&conf->device_lock);
				if (must_retry) {
					release_stripe(sh);
4285
					schedule();
4286 4287 4288
					goto retry;
				}
			}
4289

4290
			if (rw == WRITE &&
4291
			    logical_sector >= mddev->suspend_lo &&
4292 4293
			    logical_sector < mddev->suspend_hi) {
				release_stripe(sh);
4294 4295 4296 4297 4298 4299 4300 4301 4302 4303
				/* 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();
4304 4305
				goto retry;
			}
4306 4307

			if (test_bit(STRIPE_EXPANDING, &sh->state) ||
4308
			    !add_stripe_bio(sh, bi, dd_idx, rw)) {
4309 4310
				/* Stripe is busy expanding or
				 * add failed due to overlap.  Flush everything
L
Linus Torvalds 已提交
4311 4312
				 * and wait a while
				 */
N
NeilBrown 已提交
4313
				md_wakeup_thread(mddev->thread);
L
Linus Torvalds 已提交
4314 4315 4316 4317 4318
				release_stripe(sh);
				schedule();
				goto retry;
			}
			finish_wait(&conf->wait_for_overlap, &w);
4319 4320
			set_bit(STRIPE_HANDLE, &sh->state);
			clear_bit(STRIPE_DELAYED, &sh->state);
4321
			if ((bi->bi_rw & REQ_NOIDLE) &&
4322 4323
			    !test_and_set_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
				atomic_inc(&conf->preread_active_stripes);
4324
			release_stripe_plug(mddev, sh);
L
Linus Torvalds 已提交
4325 4326 4327 4328 4329 4330 4331
		} 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;
		}
	}
4332

4333
	remaining = raid5_dec_bi_active_stripes(bi);
4334
	if (remaining == 0) {
L
Linus Torvalds 已提交
4335

4336
		if ( rw == WRITE )
L
Linus Torvalds 已提交
4337
			md_write_end(mddev);
4338

4339
		bio_endio(bi, 0);
L
Linus Torvalds 已提交
4340 4341 4342
	}
}

4343
static sector_t raid5_size(struct mddev *mddev, sector_t sectors, int raid_disks);
D
Dan Williams 已提交
4344

4345
static sector_t reshape_request(struct mddev *mddev, sector_t sector_nr, int *skipped)
L
Linus Torvalds 已提交
4346
{
4347 4348 4349 4350 4351 4352 4353 4354 4355
	/* 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.
	 */
4356
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
4357
	struct stripe_head *sh;
4358
	sector_t first_sector, last_sector;
4359 4360 4361
	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;
4362 4363
	int i;
	int dd_idx;
4364
	sector_t writepos, readpos, safepos;
4365
	sector_t stripe_addr;
4366
	int reshape_sectors;
4367
	struct list_head stripes;
4368

4369 4370
	if (sector_nr == 0) {
		/* If restarting in the middle, skip the initial sectors */
4371
		if (mddev->reshape_backwards &&
4372 4373 4374
		    conf->reshape_progress < raid5_size(mddev, 0, 0)) {
			sector_nr = raid5_size(mddev, 0, 0)
				- conf->reshape_progress;
4375
		} else if (!mddev->reshape_backwards &&
4376 4377
			   conf->reshape_progress > 0)
			sector_nr = conf->reshape_progress;
4378
		sector_div(sector_nr, new_data_disks);
4379
		if (sector_nr) {
4380 4381
			mddev->curr_resync_completed = sector_nr;
			sysfs_notify(&mddev->kobj, NULL, "sync_completed");
4382 4383 4384
			*skipped = 1;
			return sector_nr;
		}
4385 4386
	}

4387 4388 4389 4390
	/* 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
	 */
4391 4392
	if (mddev->new_chunk_sectors > mddev->chunk_sectors)
		reshape_sectors = mddev->new_chunk_sectors;
4393
	else
4394
		reshape_sectors = mddev->chunk_sectors;
4395

4396 4397 4398 4399 4400
	/* 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
4401
	 */
4402
	writepos = conf->reshape_progress;
4403
	sector_div(writepos, new_data_disks);
4404 4405
	readpos = conf->reshape_progress;
	sector_div(readpos, data_disks);
4406
	safepos = conf->reshape_safe;
4407
	sector_div(safepos, data_disks);
4408
	if (mddev->reshape_backwards) {
4409
		writepos -= min_t(sector_t, reshape_sectors, writepos);
4410
		readpos += reshape_sectors;
4411
		safepos += reshape_sectors;
4412
	} else {
4413
		writepos += reshape_sectors;
4414 4415
		readpos -= min_t(sector_t, reshape_sectors, readpos);
		safepos -= min_t(sector_t, reshape_sectors, safepos);
4416
	}
4417

4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432
	/* 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;
	}

4433 4434 4435 4436
	/* '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.
4437 4438 4439 4440
	 * 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
4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452
	 * 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???
	 */
4453 4454 4455 4456 4457 4458
	if (conf->min_offset_diff < 0) {
		safepos += -conf->min_offset_diff;
		readpos += -conf->min_offset_diff;
	} else
		writepos += conf->min_offset_diff;

4459
	if ((mddev->reshape_backwards
4460 4461 4462
	     ? (safepos > writepos && readpos < writepos)
	     : (safepos < writepos && readpos > writepos)) ||
	    time_after(jiffies, conf->reshape_checkpoint + 10*HZ)) {
4463 4464 4465
		/* Cannot proceed until we've updated the superblock... */
		wait_event(conf->wait_for_overlap,
			   atomic_read(&conf->reshape_stripes)==0);
4466
		mddev->reshape_position = conf->reshape_progress;
4467
		mddev->curr_resync_completed = sector_nr;
4468
		conf->reshape_checkpoint = jiffies;
4469
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
4470
		md_wakeup_thread(mddev->thread);
4471
		wait_event(mddev->sb_wait, mddev->flags == 0 ||
4472 4473
			   kthread_should_stop());
		spin_lock_irq(&conf->device_lock);
4474
		conf->reshape_safe = mddev->reshape_position;
4475 4476
		spin_unlock_irq(&conf->device_lock);
		wake_up(&conf->wait_for_overlap);
4477
		sysfs_notify(&mddev->kobj, NULL, "sync_completed");
4478 4479
	}

4480
	INIT_LIST_HEAD(&stripes);
4481
	for (i = 0; i < reshape_sectors; i += STRIPE_SECTORS) {
4482
		int j;
4483
		int skipped_disk = 0;
4484
		sh = get_active_stripe(conf, stripe_addr+i, 0, 0, 1);
4485 4486 4487 4488 4489 4490 4491 4492 4493
		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;
4494
			if (conf->level == 6 &&
4495
			    j == sh->qd_idx)
4496
				continue;
4497
			s = compute_blocknr(sh, j, 0);
D
Dan Williams 已提交
4498
			if (s < raid5_size(mddev, 0, 0)) {
4499
				skipped_disk = 1;
4500 4501 4502 4503 4504 4505
				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);
		}
4506
		if (!skipped_disk) {
4507 4508 4509
			set_bit(STRIPE_EXPAND_READY, &sh->state);
			set_bit(STRIPE_HANDLE, &sh->state);
		}
4510
		list_add(&sh->lru, &stripes);
4511 4512
	}
	spin_lock_irq(&conf->device_lock);
4513
	if (mddev->reshape_backwards)
4514
		conf->reshape_progress -= reshape_sectors * new_data_disks;
4515
	else
4516
		conf->reshape_progress += reshape_sectors * new_data_disks;
4517 4518 4519 4520 4521 4522 4523
	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 =
4524
		raid5_compute_sector(conf, stripe_addr*(new_data_disks),
4525
				     1, &dd_idx, NULL);
4526
	last_sector =
4527
		raid5_compute_sector(conf, ((stripe_addr+reshape_sectors)
4528
					    * new_data_disks - 1),
4529
				     1, &dd_idx, NULL);
A
Andre Noll 已提交
4530 4531
	if (last_sector >= mddev->dev_sectors)
		last_sector = mddev->dev_sectors - 1;
4532
	while (first_sector <= last_sector) {
4533
		sh = get_active_stripe(conf, first_sector, 1, 0, 1);
4534 4535 4536 4537 4538
		set_bit(STRIPE_EXPAND_SOURCE, &sh->state);
		set_bit(STRIPE_HANDLE, &sh->state);
		release_stripe(sh);
		first_sector += STRIPE_SECTORS;
	}
4539 4540 4541 4542 4543 4544 4545 4546
	/* 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);
	}
4547 4548 4549
	/* If this takes us to the resync_max point where we have to pause,
	 * then we need to write out the superblock.
	 */
4550
	sector_nr += reshape_sectors;
4551 4552
	if ((sector_nr - mddev->curr_resync_completed) * 2
	    >= mddev->resync_max - mddev->curr_resync_completed) {
4553 4554 4555
		/* Cannot proceed until we've updated the superblock... */
		wait_event(conf->wait_for_overlap,
			   atomic_read(&conf->reshape_stripes) == 0);
4556
		mddev->reshape_position = conf->reshape_progress;
4557
		mddev->curr_resync_completed = sector_nr;
4558
		conf->reshape_checkpoint = jiffies;
4559 4560 4561 4562 4563 4564
		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);
4565
		conf->reshape_safe = mddev->reshape_position;
4566 4567
		spin_unlock_irq(&conf->device_lock);
		wake_up(&conf->wait_for_overlap);
4568
		sysfs_notify(&mddev->kobj, NULL, "sync_completed");
4569
	}
4570
	return reshape_sectors;
4571 4572 4573
}

/* FIXME go_faster isn't used */
4574
static inline sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipped, int go_faster)
4575
{
4576
	struct r5conf *conf = mddev->private;
4577
	struct stripe_head *sh;
A
Andre Noll 已提交
4578
	sector_t max_sector = mddev->dev_sectors;
N
NeilBrown 已提交
4579
	sector_t sync_blocks;
4580 4581
	int still_degraded = 0;
	int i;
L
Linus Torvalds 已提交
4582

4583
	if (sector_nr >= max_sector) {
L
Linus Torvalds 已提交
4584
		/* just being told to finish up .. nothing much to do */
4585

4586 4587 4588 4589
		if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) {
			end_reshape(conf);
			return 0;
		}
4590 4591 4592 4593

		if (mddev->curr_resync < max_sector) /* aborted */
			bitmap_end_sync(mddev->bitmap, mddev->curr_resync,
					&sync_blocks, 1);
4594
		else /* completed sync */
4595 4596 4597
			conf->fullsync = 0;
		bitmap_close_sync(mddev->bitmap);

L
Linus Torvalds 已提交
4598 4599
		return 0;
	}
4600

4601 4602 4603
	/* Allow raid5_quiesce to complete */
	wait_event(conf->wait_for_overlap, conf->quiesce != 2);

4604 4605
	if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
		return reshape_request(mddev, sector_nr, skipped);
4606

4607 4608 4609 4610 4611 4612
	/* 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
	 */

4613
	/* if there is too many failed drives and we are trying
L
Linus Torvalds 已提交
4614 4615 4616
	 * to resync, then assert that we are finished, because there is
	 * nothing we can do.
	 */
4617
	if (mddev->degraded >= conf->max_degraded &&
4618
	    test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
A
Andre Noll 已提交
4619
		sector_t rv = mddev->dev_sectors - sector_nr;
4620
		*skipped = 1;
L
Linus Torvalds 已提交
4621 4622
		return rv;
	}
4623
	if (!bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, 1) &&
4624
	    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
4625 4626 4627 4628 4629 4630
	    !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
4631

N
NeilBrown 已提交
4632 4633
	bitmap_cond_end_sync(mddev->bitmap, sector_nr);

4634
	sh = get_active_stripe(conf, sector_nr, 0, 1, 0);
L
Linus Torvalds 已提交
4635
	if (sh == NULL) {
4636
		sh = get_active_stripe(conf, sector_nr, 0, 0, 0);
L
Linus Torvalds 已提交
4637
		/* make sure we don't swamp the stripe cache if someone else
4638
		 * is trying to get access
L
Linus Torvalds 已提交
4639
		 */
4640
		schedule_timeout_uninterruptible(1);
L
Linus Torvalds 已提交
4641
	}
4642 4643 4644 4645
	/* 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.
	 */
4646
	for (i = 0; i < conf->raid_disks; i++)
4647 4648 4649 4650 4651
		if (conf->disks[i].rdev == NULL)
			still_degraded = 1;

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

4652
	set_bit(STRIPE_SYNC_REQUESTED, &sh->state);
L
Linus Torvalds 已提交
4653

4654
	handle_stripe(sh);
L
Linus Torvalds 已提交
4655 4656 4657 4658 4659
	release_stripe(sh);

	return STRIPE_SECTORS;
}

4660
static int  retry_aligned_read(struct r5conf *conf, struct bio *raid_bio)
4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672
{
	/* 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;
4673
	int dd_idx;
4674 4675 4676 4677 4678 4679
	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);
4680
	sector = raid5_compute_sector(conf, logical_sector,
4681
				      0, &dd_idx, NULL);
4682 4683 4684
	last_sector = raid_bio->bi_sector + (raid_bio->bi_size>>9);

	for (; logical_sector < last_sector;
4685 4686 4687
	     logical_sector += STRIPE_SECTORS,
		     sector += STRIPE_SECTORS,
		     scnt++) {
4688

4689
		if (scnt < raid5_bi_processed_stripes(raid_bio))
4690 4691 4692
			/* already done this stripe */
			continue;

4693
		sh = get_active_stripe(conf, sector, 0, 1, 0);
4694 4695 4696

		if (!sh) {
			/* failed to get a stripe - must wait */
4697
			raid5_set_bi_processed_stripes(raid_bio, scnt);
4698 4699 4700 4701
			conf->retry_read_aligned = raid_bio;
			return handled;
		}

4702 4703
		if (!add_stripe_bio(sh, raid_bio, dd_idx, 0)) {
			release_stripe(sh);
4704
			raid5_set_bi_processed_stripes(raid_bio, scnt);
4705 4706 4707 4708
			conf->retry_read_aligned = raid_bio;
			return handled;
		}

4709
		set_bit(R5_ReadNoMerge, &sh->dev[dd_idx].flags);
4710
		handle_stripe(sh);
4711 4712 4713
		release_stripe(sh);
		handled++;
	}
4714
	remaining = raid5_dec_bi_active_stripes(raid_bio);
4715 4716
	if (remaining == 0)
		bio_endio(raid_bio, 0);
4717 4718 4719 4720 4721
	if (atomic_dec_and_test(&conf->active_aligned_reads))
		wake_up(&conf->wait_for_stripe);
	return handled;
}

4722 4723 4724 4725 4726 4727 4728 4729 4730 4731 4732 4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743 4744 4745
#define MAX_STRIPE_BATCH 8
static int handle_active_stripes(struct r5conf *conf)
{
	struct stripe_head *batch[MAX_STRIPE_BATCH], *sh;
	int i, batch_size = 0;

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

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

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

	cond_resched();

	spin_lock_irq(&conf->device_lock);
	for (i = 0; i < batch_size; i++)
		__release_stripe(conf, batch[i]);
	return batch_size;
}
4746

L
Linus Torvalds 已提交
4747 4748 4749 4750 4751 4752 4753
/*
 * This is our raid5 kernel thread.
 *
 * We scan the hash table for stripes which can be handled now.
 * During the scan, completed stripes are saved for us by the interrupt
 * handler, so that they will not have to wait for our next wakeup.
 */
S
Shaohua Li 已提交
4754
static void raid5d(struct md_thread *thread)
L
Linus Torvalds 已提交
4755
{
S
Shaohua Li 已提交
4756
	struct mddev *mddev = thread->mddev;
4757
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
4758
	int handled;
4759
	struct blk_plug plug;
L
Linus Torvalds 已提交
4760

4761
	pr_debug("+++ raid5d active\n");
L
Linus Torvalds 已提交
4762 4763 4764

	md_check_recovery(mddev);

4765
	blk_start_plug(&plug);
L
Linus Torvalds 已提交
4766 4767 4768
	handled = 0;
	spin_lock_irq(&conf->device_lock);
	while (1) {
4769
		struct bio *bio;
4770
		int batch_size;
L
Linus Torvalds 已提交
4771

4772
		if (
4773 4774 4775
		    !list_empty(&conf->bitmap_list)) {
			/* Now is a good time to flush some bitmap updates */
			conf->seq_flush++;
4776
			spin_unlock_irq(&conf->device_lock);
4777
			bitmap_unplug(mddev->bitmap);
4778
			spin_lock_irq(&conf->device_lock);
4779
			conf->seq_write = conf->seq_flush;
4780 4781
			activate_bit_delay(conf);
		}
4782
		raid5_activate_delayed(conf);
4783

4784 4785 4786 4787 4788 4789 4790 4791 4792 4793
		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++;
		}

4794 4795
		batch_size = handle_active_stripes(conf);
		if (!batch_size)
L
Linus Torvalds 已提交
4796
			break;
4797
		handled += batch_size;
L
Linus Torvalds 已提交
4798

4799 4800
		if (mddev->flags & ~(1<<MD_CHANGE_PENDING)) {
			spin_unlock_irq(&conf->device_lock);
4801
			md_check_recovery(mddev);
4802 4803
			spin_lock_irq(&conf->device_lock);
		}
L
Linus Torvalds 已提交
4804
	}
4805
	pr_debug("%d stripes handled\n", handled);
L
Linus Torvalds 已提交
4806 4807 4808

	spin_unlock_irq(&conf->device_lock);

4809
	async_tx_issue_pending_all();
4810
	blk_finish_plug(&plug);
L
Linus Torvalds 已提交
4811

4812
	pr_debug("--- raid5d inactive\n");
L
Linus Torvalds 已提交
4813 4814
}

4815
static ssize_t
4816
raid5_show_stripe_cache_size(struct mddev *mddev, char *page)
4817
{
4818
	struct r5conf *conf = mddev->private;
4819 4820 4821 4822
	if (conf)
		return sprintf(page, "%d\n", conf->max_nr_stripes);
	else
		return 0;
4823 4824
}

4825
int
4826
raid5_set_cache_size(struct mddev *mddev, int size)
4827
{
4828
	struct r5conf *conf = mddev->private;
4829 4830
	int err;

4831
	if (size <= 16 || size > 32768)
4832
		return -EINVAL;
4833
	while (size < conf->max_nr_stripes) {
4834 4835 4836 4837 4838
		if (drop_one_stripe(conf))
			conf->max_nr_stripes--;
		else
			break;
	}
4839 4840 4841
	err = md_allow_write(mddev);
	if (err)
		return err;
4842
	while (size > conf->max_nr_stripes) {
4843 4844 4845 4846
		if (grow_one_stripe(conf))
			conf->max_nr_stripes++;
		else break;
	}
4847 4848 4849 4850 4851
	return 0;
}
EXPORT_SYMBOL(raid5_set_cache_size);

static ssize_t
4852
raid5_store_stripe_cache_size(struct mddev *mddev, const char *page, size_t len)
4853
{
4854
	struct r5conf *conf = mddev->private;
4855 4856 4857 4858 4859 4860 4861 4862 4863 4864 4865 4866 4867
	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;
4868 4869
	return len;
}
4870

4871 4872 4873 4874
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);
4875

4876
static ssize_t
4877
raid5_show_preread_threshold(struct mddev *mddev, char *page)
4878
{
4879
	struct r5conf *conf = mddev->private;
4880 4881 4882 4883 4884 4885 4886
	if (conf)
		return sprintf(page, "%d\n", conf->bypass_threshold);
	else
		return 0;
}

static ssize_t
4887
raid5_store_preread_threshold(struct mddev *mddev, const char *page, size_t len)
4888
{
4889
	struct r5conf *conf = mddev->private;
4890
	unsigned long new;
4891 4892 4893 4894 4895
	if (len >= PAGE_SIZE)
		return -EINVAL;
	if (!conf)
		return -ENODEV;

4896
	if (strict_strtoul(page, 10, &new))
4897
		return -EINVAL;
4898
	if (new > conf->max_nr_stripes)
4899 4900 4901 4902 4903 4904 4905 4906 4907 4908 4909
		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);

4910
static ssize_t
4911
stripe_cache_active_show(struct mddev *mddev, char *page)
4912
{
4913
	struct r5conf *conf = mddev->private;
4914 4915 4916 4917
	if (conf)
		return sprintf(page, "%d\n", atomic_read(&conf->active_stripes));
	else
		return 0;
4918 4919
}

4920 4921
static struct md_sysfs_entry
raid5_stripecache_active = __ATTR_RO(stripe_cache_active);
4922

4923
static struct attribute *raid5_attrs[] =  {
4924 4925
	&raid5_stripecache_size.attr,
	&raid5_stripecache_active.attr,
4926
	&raid5_preread_bypass_threshold.attr,
4927 4928
	NULL,
};
4929 4930 4931
static struct attribute_group raid5_attrs_group = {
	.name = NULL,
	.attrs = raid5_attrs,
4932 4933
};

4934
static sector_t
4935
raid5_size(struct mddev *mddev, sector_t sectors, int raid_disks)
4936
{
4937
	struct r5conf *conf = mddev->private;
4938 4939 4940

	if (!sectors)
		sectors = mddev->dev_sectors;
4941
	if (!raid_disks)
4942
		/* size is defined by the smallest of previous and new size */
4943
		raid_disks = min(conf->raid_disks, conf->previous_raid_disks);
4944

4945
	sectors &= ~((sector_t)mddev->chunk_sectors - 1);
4946
	sectors &= ~((sector_t)mddev->new_chunk_sectors - 1);
4947 4948 4949
	return sectors * (raid_disks - conf->max_degraded);
}

4950
static void raid5_free_percpu(struct r5conf *conf)
4951 4952 4953 4954 4955 4956 4957 4958 4959 4960 4961
{
	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);
4962
		kfree(percpu->scribble);
4963 4964 4965 4966 4967 4968 4969 4970 4971
	}
#ifdef CONFIG_HOTPLUG_CPU
	unregister_cpu_notifier(&conf->cpu_notify);
#endif
	put_online_cpus();

	free_percpu(conf->percpu);
}

4972
static void free_conf(struct r5conf *conf)
4973 4974
{
	shrink_stripes(conf);
4975
	raid5_free_percpu(conf);
4976 4977 4978 4979 4980
	kfree(conf->disks);
	kfree(conf->stripe_hashtbl);
	kfree(conf);
}

4981 4982 4983 4984
#ifdef CONFIG_HOTPLUG_CPU
static int raid456_cpu_notify(struct notifier_block *nfb, unsigned long action,
			      void *hcpu)
{
4985
	struct r5conf *conf = container_of(nfb, struct r5conf, cpu_notify);
4986 4987 4988 4989 4990 4991
	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:
4992
		if (conf->level == 6 && !percpu->spare_page)
4993
			percpu->spare_page = alloc_page(GFP_KERNEL);
4994 4995 4996 4997 4998 4999 5000
		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);
5001 5002
			pr_err("%s: failed memory allocation for cpu%ld\n",
			       __func__, cpu);
5003
			return notifier_from_errno(-ENOMEM);
5004 5005 5006 5007 5008
		}
		break;
	case CPU_DEAD:
	case CPU_DEAD_FROZEN:
		safe_put_page(percpu->spare_page);
5009
		kfree(percpu->scribble);
5010
		percpu->spare_page = NULL;
5011
		percpu->scribble = NULL;
5012 5013 5014 5015 5016 5017 5018 5019
		break;
	default:
		break;
	}
	return NOTIFY_OK;
}
#endif

5020
static int raid5_alloc_percpu(struct r5conf *conf)
5021 5022 5023
{
	unsigned long cpu;
	struct page *spare_page;
5024
	struct raid5_percpu __percpu *allcpus;
5025
	void *scribble;
5026 5027 5028 5029 5030 5031 5032 5033 5034 5035
	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) {
5036 5037 5038 5039 5040 5041 5042 5043
		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;
		}
5044
		scribble = kmalloc(conf->scribble_len, GFP_KERNEL);
5045
		if (!scribble) {
5046 5047 5048
			err = -ENOMEM;
			break;
		}
5049
		per_cpu_ptr(conf->percpu, cpu)->scribble = scribble;
5050 5051 5052 5053 5054 5055 5056 5057 5058 5059 5060 5061
	}
#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;
}

5062
static struct r5conf *setup_conf(struct mddev *mddev)
L
Linus Torvalds 已提交
5063
{
5064
	struct r5conf *conf;
5065
	int raid_disk, memory, max_disks;
5066
	struct md_rdev *rdev;
L
Linus Torvalds 已提交
5067
	struct disk_info *disk;
5068
	char pers_name[6];
L
Linus Torvalds 已提交
5069

N
NeilBrown 已提交
5070 5071 5072
	if (mddev->new_level != 5
	    && mddev->new_level != 4
	    && mddev->new_level != 6) {
5073
		printk(KERN_ERR "md/raid:%s: raid level not set to 4/5/6 (%d)\n",
N
NeilBrown 已提交
5074 5075
		       mdname(mddev), mddev->new_level);
		return ERR_PTR(-EIO);
L
Linus Torvalds 已提交
5076
	}
N
NeilBrown 已提交
5077 5078 5079 5080
	if ((mddev->new_level == 5
	     && !algorithm_valid_raid5(mddev->new_layout)) ||
	    (mddev->new_level == 6
	     && !algorithm_valid_raid6(mddev->new_layout))) {
5081
		printk(KERN_ERR "md/raid:%s: layout %d not supported\n",
N
NeilBrown 已提交
5082 5083
		       mdname(mddev), mddev->new_layout);
		return ERR_PTR(-EIO);
5084
	}
N
NeilBrown 已提交
5085
	if (mddev->new_level == 6 && mddev->raid_disks < 4) {
5086
		printk(KERN_ERR "md/raid:%s: not enough configured devices (%d, minimum 4)\n",
N
NeilBrown 已提交
5087 5088
		       mdname(mddev), mddev->raid_disks);
		return ERR_PTR(-EINVAL);
5089 5090
	}

5091 5092 5093
	if (!mddev->new_chunk_sectors ||
	    (mddev->new_chunk_sectors << 9) % PAGE_SIZE ||
	    !is_power_of_2(mddev->new_chunk_sectors)) {
5094 5095
		printk(KERN_ERR "md/raid:%s: invalid chunk size %d\n",
		       mdname(mddev), mddev->new_chunk_sectors << 9);
N
NeilBrown 已提交
5096
		return ERR_PTR(-EINVAL);
5097 5098
	}

5099
	conf = kzalloc(sizeof(struct r5conf), GFP_KERNEL);
N
NeilBrown 已提交
5100
	if (conf == NULL)
L
Linus Torvalds 已提交
5101
		goto abort;
5102 5103 5104 5105 5106 5107 5108 5109 5110 5111 5112 5113
	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;
5114
	conf->recovery_disabled = mddev->recovery_disabled - 1;
N
NeilBrown 已提交
5115 5116 5117 5118 5119

	conf->raid_disks = mddev->raid_disks;
	if (mddev->reshape_position == MaxSector)
		conf->previous_raid_disks = mddev->raid_disks;
	else
5120
		conf->previous_raid_disks = mddev->raid_disks - mddev->delta_disks;
5121 5122
	max_disks = max(conf->raid_disks, conf->previous_raid_disks);
	conf->scribble_len = scribble_len(max_disks);
5123

5124
	conf->disks = kzalloc(max_disks * sizeof(struct disk_info),
5125 5126 5127
			      GFP_KERNEL);
	if (!conf->disks)
		goto abort;
5128

L
Linus Torvalds 已提交
5129 5130
	conf->mddev = mddev;

5131
	if ((conf->stripe_hashtbl = kzalloc(PAGE_SIZE, GFP_KERNEL)) == NULL)
L
Linus Torvalds 已提交
5132 5133
		goto abort;

5134 5135 5136 5137
	conf->level = mddev->new_level;
	if (raid5_alloc_percpu(conf) != 0)
		goto abort;

5138
	pr_debug("raid456: run(%s) called.\n", mdname(mddev));
L
Linus Torvalds 已提交
5139

N
NeilBrown 已提交
5140
	rdev_for_each(rdev, mddev) {
L
Linus Torvalds 已提交
5141
		raid_disk = rdev->raid_disk;
5142
		if (raid_disk >= max_disks
L
Linus Torvalds 已提交
5143 5144 5145 5146
		    || raid_disk < 0)
			continue;
		disk = conf->disks + raid_disk;

5147 5148 5149 5150 5151 5152 5153 5154 5155
		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
5156

5157
		if (test_bit(In_sync, &rdev->flags)) {
L
Linus Torvalds 已提交
5158
			char b[BDEVNAME_SIZE];
5159 5160 5161
			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
5162
		} else if (rdev->saved_raid_disk != raid_disk)
5163 5164
			/* Cannot rely on bitmap to complete recovery */
			conf->fullsync = 1;
L
Linus Torvalds 已提交
5165 5166
	}

5167
	conf->chunk_sectors = mddev->new_chunk_sectors;
N
NeilBrown 已提交
5168
	conf->level = mddev->new_level;
5169 5170 5171 5172
	if (conf->level == 6)
		conf->max_degraded = 2;
	else
		conf->max_degraded = 1;
N
NeilBrown 已提交
5173
	conf->algorithm = mddev->new_layout;
L
Linus Torvalds 已提交
5174
	conf->max_nr_stripes = NR_STRIPES;
5175
	conf->reshape_progress = mddev->reshape_position;
5176
	if (conf->reshape_progress != MaxSector) {
5177
		conf->prev_chunk_sectors = mddev->chunk_sectors;
5178 5179
		conf->prev_algo = mddev->layout;
	}
L
Linus Torvalds 已提交
5180

N
NeilBrown 已提交
5181
	memory = conf->max_nr_stripes * (sizeof(struct stripe_head) +
5182
		 max_disks * ((sizeof(struct bio) + PAGE_SIZE))) / 1024;
N
NeilBrown 已提交
5183 5184
	if (grow_stripes(conf, conf->max_nr_stripes)) {
		printk(KERN_ERR
5185 5186
		       "md/raid:%s: couldn't allocate %dkB for buffers\n",
		       mdname(mddev), memory);
N
NeilBrown 已提交
5187 5188
		goto abort;
	} else
5189 5190
		printk(KERN_INFO "md/raid:%s: allocated %dkB\n",
		       mdname(mddev), memory);
L
Linus Torvalds 已提交
5191

5192 5193
	sprintf(pers_name, "raid%d", mddev->new_level);
	conf->thread = md_register_thread(raid5d, mddev, pers_name);
N
NeilBrown 已提交
5194 5195
	if (!conf->thread) {
		printk(KERN_ERR
5196
		       "md/raid:%s: couldn't allocate thread.\n",
N
NeilBrown 已提交
5197
		       mdname(mddev));
5198 5199
		goto abort;
	}
N
NeilBrown 已提交
5200 5201 5202 5203 5204

	return conf;

 abort:
	if (conf) {
5205
		free_conf(conf);
N
NeilBrown 已提交
5206 5207 5208 5209 5210
		return ERR_PTR(-EIO);
	} else
		return ERR_PTR(-ENOMEM);
}

5211 5212 5213 5214 5215 5216 5217 5218 5219 5220 5221 5222 5223 5224 5225 5226 5227 5228 5229 5230 5231 5232 5233 5234 5235 5236 5237

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;
}

5238
static int run(struct mddev *mddev)
N
NeilBrown 已提交
5239
{
5240
	struct r5conf *conf;
5241
	int working_disks = 0;
5242
	int dirty_parity_disks = 0;
5243
	struct md_rdev *rdev;
5244
	sector_t reshape_offset = 0;
5245
	int i;
5246 5247
	long long min_offset_diff = 0;
	int first = 1;
N
NeilBrown 已提交
5248

5249
	if (mddev->recovery_cp != MaxSector)
5250
		printk(KERN_NOTICE "md/raid:%s: not clean"
5251 5252
		       " -- starting background reconstruction\n",
		       mdname(mddev));
5253 5254 5255 5256 5257 5258 5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269

	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
5270 5271
	if (mddev->reshape_position != MaxSector) {
		/* Check that we can continue the reshape.
5272 5273 5274 5275 5276 5277 5278 5279 5280 5281
		 * 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
5282 5283 5284
		 */
		sector_t here_new, here_old;
		int old_disks;
5285
		int max_degraded = (mddev->level == 6 ? 2 : 1);
N
NeilBrown 已提交
5286

5287
		if (mddev->new_level != mddev->level) {
5288
			printk(KERN_ERR "md/raid:%s: unsupported reshape "
N
NeilBrown 已提交
5289 5290 5291 5292 5293 5294 5295 5296 5297 5298
			       "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;
5299
		if (sector_div(here_new, mddev->new_chunk_sectors *
N
NeilBrown 已提交
5300
			       (mddev->raid_disks - max_degraded))) {
5301 5302
			printk(KERN_ERR "md/raid:%s: reshape_position not "
			       "on a stripe boundary\n", mdname(mddev));
N
NeilBrown 已提交
5303 5304
			return -EINVAL;
		}
5305
		reshape_offset = here_new * mddev->new_chunk_sectors;
N
NeilBrown 已提交
5306 5307
		/* here_new is the stripe we will write to */
		here_old = mddev->reshape_position;
5308
		sector_div(here_old, mddev->chunk_sectors *
N
NeilBrown 已提交
5309 5310 5311
			   (old_disks-max_degraded));
		/* here_old is the first stripe that we might need to read
		 * from */
5312
		if (mddev->delta_disks == 0) {
5313 5314 5315 5316 5317 5318
			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;
			}
5319
			/* We cannot be sure it is safe to start an in-place
5320
			 * reshape.  It is only safe if user-space is monitoring
5321 5322 5323 5324 5325
			 * 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.
			 */
5326 5327 5328 5329 5330 5331 5332
			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",
5333
				       mdname(mddev));
5334 5335
				return -EINVAL;
			}
5336
		} else if (mddev->reshape_backwards
5337
		    ? (here_new * mddev->new_chunk_sectors + min_offset_diff <=
5338 5339
		       here_old * mddev->chunk_sectors)
		    : (here_new * mddev->new_chunk_sectors >=
5340
		       here_old * mddev->chunk_sectors + (-min_offset_diff))) {
N
NeilBrown 已提交
5341
			/* Reading from the same stripe as writing to - bad */
5342 5343 5344
			printk(KERN_ERR "md/raid:%s: reshape_position too early for "
			       "auto-recovery - aborting.\n",
			       mdname(mddev));
N
NeilBrown 已提交
5345 5346
			return -EINVAL;
		}
5347 5348
		printk(KERN_INFO "md/raid:%s: reshape will continue\n",
		       mdname(mddev));
N
NeilBrown 已提交
5349 5350 5351 5352
		/* OK, we should be able to continue; */
	} else {
		BUG_ON(mddev->level != mddev->new_level);
		BUG_ON(mddev->layout != mddev->new_layout);
5353
		BUG_ON(mddev->chunk_sectors != mddev->new_chunk_sectors);
N
NeilBrown 已提交
5354
		BUG_ON(mddev->delta_disks != 0);
L
Linus Torvalds 已提交
5355
	}
N
NeilBrown 已提交
5356

5357 5358 5359 5360 5361
	if (mddev->private == NULL)
		conf = setup_conf(mddev);
	else
		conf = mddev->private;

N
NeilBrown 已提交
5362 5363 5364
	if (IS_ERR(conf))
		return PTR_ERR(conf);

5365
	conf->min_offset_diff = min_offset_diff;
N
NeilBrown 已提交
5366 5367 5368 5369
	mddev->thread = conf->thread;
	conf->thread = NULL;
	mddev->private = conf;

5370 5371 5372 5373 5374 5375 5376 5377 5378 5379 5380
	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)
5381
			continue;
5382 5383 5384 5385 5386 5387 5388
		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;
		}
5389
		if (test_bit(In_sync, &rdev->flags)) {
N
NeilBrown 已提交
5390
			working_disks++;
5391 5392
			continue;
		}
5393 5394 5395 5396 5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410 5411 5412 5413 5414 5415 5416 5417 5418 5419 5420
		/* 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
5421

5422 5423 5424
	/*
	 * 0 for a fully functional array, 1 or 2 for a degraded array.
	 */
5425
	mddev->degraded = calc_degraded(conf);
N
NeilBrown 已提交
5426

5427
	if (has_failed(conf)) {
5428
		printk(KERN_ERR "md/raid:%s: not enough operational devices"
L
Linus Torvalds 已提交
5429
			" (%d/%d failed)\n",
5430
			mdname(mddev), mddev->degraded, conf->raid_disks);
L
Linus Torvalds 已提交
5431 5432 5433
		goto abort;
	}

N
NeilBrown 已提交
5434
	/* device size must be a multiple of chunk size */
5435
	mddev->dev_sectors &= ~(mddev->chunk_sectors - 1);
N
NeilBrown 已提交
5436 5437
	mddev->resync_max_sectors = mddev->dev_sectors;

5438
	if (mddev->degraded > dirty_parity_disks &&
L
Linus Torvalds 已提交
5439
	    mddev->recovery_cp != MaxSector) {
5440 5441
		if (mddev->ok_start_degraded)
			printk(KERN_WARNING
5442 5443
			       "md/raid:%s: starting dirty degraded array"
			       " - data corruption possible.\n",
5444 5445 5446
			       mdname(mddev));
		else {
			printk(KERN_ERR
5447
			       "md/raid:%s: cannot start dirty degraded array.\n",
5448 5449 5450
			       mdname(mddev));
			goto abort;
		}
L
Linus Torvalds 已提交
5451 5452 5453
	}

	if (mddev->degraded == 0)
5454 5455
		printk(KERN_INFO "md/raid:%s: raid level %d active with %d out of %d"
		       " devices, algorithm %d\n", mdname(mddev), conf->level,
5456 5457
		       mddev->raid_disks-mddev->degraded, mddev->raid_disks,
		       mddev->new_layout);
L
Linus Torvalds 已提交
5458
	else
5459 5460 5461 5462 5463
		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
5464 5465 5466

	print_raid5_conf(conf);

5467 5468
	if (conf->reshape_progress != MaxSector) {
		conf->reshape_safe = conf->reshape_progress;
5469 5470 5471 5472 5473 5474
		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,
5475
							"reshape");
5476 5477
	}

L
Linus Torvalds 已提交
5478 5479

	/* Ok, everything is just fine now */
5480 5481
	if (mddev->to_remove == &raid5_attrs_group)
		mddev->to_remove = NULL;
N
NeilBrown 已提交
5482 5483
	else if (mddev->kobj.sd &&
	    sysfs_create_group(&mddev->kobj, &raid5_attrs_group))
5484
		printk(KERN_WARNING
5485
		       "raid5: failed to create sysfs attributes for %s\n",
5486
		       mdname(mddev));
5487
	md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
5488

5489
	if (mddev->queue) {
5490
		int chunk_size;
S
Shaohua Li 已提交
5491
		bool discard_supported = true;
5492 5493 5494 5495 5496 5497 5498 5499 5500
		/* 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
5501

5502
		blk_queue_merge_bvec(mddev->queue, raid5_mergeable_bvec);
5503

N
NeilBrown 已提交
5504 5505
		mddev->queue->backing_dev_info.congested_data = mddev;
		mddev->queue->backing_dev_info.congested_fn = raid5_congested;
5506

5507 5508 5509 5510
		chunk_size = mddev->chunk_sectors << 9;
		blk_queue_io_min(mddev->queue, chunk_size);
		blk_queue_io_opt(mddev->queue, chunk_size *
				 (conf->raid_disks - conf->max_degraded));
S
Shaohua Li 已提交
5511 5512 5513 5514 5515 5516 5517 5518 5519 5520 5521 5522
		/*
		 * We can only discard a whole stripe. It doesn't make sense to
		 * discard data disk but write parity disk
		 */
		stripe = stripe * PAGE_SIZE;
		mddev->queue->limits.discard_alignment = stripe;
		mddev->queue->limits.discard_granularity = stripe;
		/*
		 * unaligned part of discard request will be ignored, so can't
		 * guarantee discard_zerors_data
		 */
		mddev->queue->limits.discard_zeroes_data = 0;
5523

5524
		rdev_for_each(rdev, mddev) {
5525 5526
			disk_stack_limits(mddev->gendisk, rdev->bdev,
					  rdev->data_offset << 9);
5527 5528
			disk_stack_limits(mddev->gendisk, rdev->bdev,
					  rdev->new_data_offset << 9);
S
Shaohua Li 已提交
5529 5530 5531 5532 5533 5534 5535 5536 5537 5538 5539 5540 5541 5542
			/*
			 * discard_zeroes_data is required, otherwise data
			 * could be lost. Consider a scenario: discard a stripe
			 * (the stripe could be inconsistent if
			 * discard_zeroes_data is 0); write one disk of the
			 * stripe (the stripe could be inconsistent again
			 * depending on which disks are used to calculate
			 * parity); the disk is broken; The stripe data of this
			 * disk is lost.
			 */
			if (!blk_queue_discard(bdev_get_queue(rdev->bdev)) ||
			    !bdev_get_queue(rdev->bdev)->
						limits.discard_zeroes_data)
				discard_supported = false;
5543
		}
S
Shaohua Li 已提交
5544 5545 5546 5547 5548 5549 5550 5551 5552

		if (discard_supported &&
		   mddev->queue->limits.max_discard_sectors >= stripe &&
		   mddev->queue->limits.discard_granularity >= stripe)
			queue_flag_set_unlocked(QUEUE_FLAG_DISCARD,
						mddev->queue);
		else
			queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD,
						mddev->queue);
5553
	}
5554

L
Linus Torvalds 已提交
5555 5556
	return 0;
abort:
5557
	md_unregister_thread(&mddev->thread);
N
NeilBrown 已提交
5558 5559
	print_raid5_conf(conf);
	free_conf(conf);
L
Linus Torvalds 已提交
5560
	mddev->private = NULL;
5561
	printk(KERN_ALERT "md/raid:%s: failed to run raid set.\n", mdname(mddev));
L
Linus Torvalds 已提交
5562 5563 5564
	return -EIO;
}

5565
static int stop(struct mddev *mddev)
L
Linus Torvalds 已提交
5566
{
5567
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5568

5569
	md_unregister_thread(&mddev->thread);
N
NeilBrown 已提交
5570 5571
	if (mddev->queue)
		mddev->queue->backing_dev_info.congested_fn = NULL;
5572
	free_conf(conf);
5573 5574
	mddev->private = NULL;
	mddev->to_remove = &raid5_attrs_group;
L
Linus Torvalds 已提交
5575 5576 5577
	return 0;
}

5578
static void status(struct seq_file *seq, struct mddev *mddev)
L
Linus Torvalds 已提交
5579
{
5580
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5581 5582
	int i;

5583 5584
	seq_printf(seq, " level %d, %dk chunk, algorithm %d", mddev->level,
		mddev->chunk_sectors / 2, mddev->layout);
5585
	seq_printf (seq, " [%d/%d] [", conf->raid_disks, conf->raid_disks - mddev->degraded);
L
Linus Torvalds 已提交
5586 5587 5588
	for (i = 0; i < conf->raid_disks; i++)
		seq_printf (seq, "%s",
			       conf->disks[i].rdev &&
5589
			       test_bit(In_sync, &conf->disks[i].rdev->flags) ? "U" : "_");
L
Linus Torvalds 已提交
5590 5591 5592
	seq_printf (seq, "]");
}

5593
static void print_raid5_conf (struct r5conf *conf)
L
Linus Torvalds 已提交
5594 5595 5596 5597
{
	int i;
	struct disk_info *tmp;

5598
	printk(KERN_DEBUG "RAID conf printout:\n");
L
Linus Torvalds 已提交
5599 5600 5601 5602
	if (!conf) {
		printk("(conf==NULL)\n");
		return;
	}
5603 5604 5605
	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
5606 5607 5608 5609 5610

	for (i = 0; i < conf->raid_disks; i++) {
		char b[BDEVNAME_SIZE];
		tmp = conf->disks + i;
		if (tmp->rdev)
5611 5612 5613
			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
5614 5615 5616
	}
}

5617
static int raid5_spare_active(struct mddev *mddev)
L
Linus Torvalds 已提交
5618 5619
{
	int i;
5620
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5621
	struct disk_info *tmp;
5622 5623
	int count = 0;
	unsigned long flags;
L
Linus Torvalds 已提交
5624 5625 5626

	for (i = 0; i < conf->raid_disks; i++) {
		tmp = conf->disks + i;
5627 5628 5629 5630 5631 5632 5633 5634 5635 5636 5637 5638 5639 5640 5641 5642 5643 5644 5645
		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
5646
		    && tmp->rdev->recovery_offset == MaxSector
5647
		    && !test_bit(Faulty, &tmp->rdev->flags)
5648
		    && !test_and_set_bit(In_sync, &tmp->rdev->flags)) {
5649
			count++;
5650
			sysfs_notify_dirent_safe(tmp->rdev->sysfs_state);
L
Linus Torvalds 已提交
5651 5652
		}
	}
5653
	spin_lock_irqsave(&conf->device_lock, flags);
5654
	mddev->degraded = calc_degraded(conf);
5655
	spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
5656
	print_raid5_conf(conf);
5657
	return count;
L
Linus Torvalds 已提交
5658 5659
}

5660
static int raid5_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
5661
{
5662
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5663
	int err = 0;
5664
	int number = rdev->raid_disk;
5665
	struct md_rdev **rdevp;
L
Linus Torvalds 已提交
5666 5667 5668
	struct disk_info *p = conf->disks + number;

	print_raid5_conf(conf);
5669 5670 5671 5672 5673 5674 5675 5676 5677 5678 5679 5680 5681 5682 5683 5684 5685 5686 5687 5688 5689 5690
	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) &&
5691
	    (!p->replacement || p->replacement == rdev) &&
5692 5693 5694 5695 5696 5697 5698 5699 5700 5701
	    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;
5702 5703 5704 5705 5706 5707 5708 5709 5710 5711 5712 5713 5714 5715
	} 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
5716 5717 5718 5719 5720 5721
abort:

	print_raid5_conf(conf);
	return err;
}

5722
static int raid5_add_disk(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
5723
{
5724
	struct r5conf *conf = mddev->private;
5725
	int err = -EEXIST;
L
Linus Torvalds 已提交
5726 5727
	int disk;
	struct disk_info *p;
5728 5729
	int first = 0;
	int last = conf->raid_disks - 1;
L
Linus Torvalds 已提交
5730

5731 5732 5733
	if (mddev->recovery_disabled == conf->recovery_disabled)
		return -EBUSY;

N
NeilBrown 已提交
5734
	if (rdev->saved_raid_disk < 0 && has_failed(conf))
L
Linus Torvalds 已提交
5735
		/* no point adding a device */
5736
		return -EINVAL;
L
Linus Torvalds 已提交
5737

5738 5739
	if (rdev->raid_disk >= 0)
		first = last = rdev->raid_disk;
L
Linus Torvalds 已提交
5740 5741

	/*
5742 5743
	 * find the disk ... but prefer rdev->saved_raid_disk
	 * if possible.
L
Linus Torvalds 已提交
5744
	 */
5745
	if (rdev->saved_raid_disk >= 0 &&
5746
	    rdev->saved_raid_disk >= first &&
5747
	    conf->disks[rdev->saved_raid_disk].rdev == NULL)
5748 5749 5750
		first = rdev->saved_raid_disk;

	for (disk = first; disk <= last; disk++) {
5751 5752
		p = conf->disks + disk;
		if (p->rdev == NULL) {
5753
			clear_bit(In_sync, &rdev->flags);
L
Linus Torvalds 已提交
5754
			rdev->raid_disk = disk;
5755
			err = 0;
5756 5757
			if (rdev->saved_raid_disk != disk)
				conf->fullsync = 1;
5758
			rcu_assign_pointer(p->rdev, rdev);
5759
			goto out;
L
Linus Torvalds 已提交
5760
		}
5761 5762 5763
	}
	for (disk = first; disk <= last; disk++) {
		p = conf->disks + disk;
5764 5765 5766 5767 5768 5769 5770 5771 5772 5773 5774
		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;
		}
	}
5775
out:
L
Linus Torvalds 已提交
5776
	print_raid5_conf(conf);
5777
	return err;
L
Linus Torvalds 已提交
5778 5779
}

5780
static int raid5_resize(struct mddev *mddev, sector_t sectors)
L
Linus Torvalds 已提交
5781 5782 5783 5784 5785 5786 5787 5788
{
	/* 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.
	 */
5789
	sector_t newsize;
5790
	sectors &= ~((sector_t)mddev->chunk_sectors - 1);
5791 5792 5793
	newsize = raid5_size(mddev, sectors, mddev->raid_disks);
	if (mddev->external_size &&
	    mddev->array_sectors > newsize)
D
Dan Williams 已提交
5794
		return -EINVAL;
5795 5796 5797 5798 5799 5800
	if (mddev->bitmap) {
		int ret = bitmap_resize(mddev->bitmap, sectors, 0, 0);
		if (ret)
			return ret;
	}
	md_set_array_sectors(mddev, newsize);
5801
	set_capacity(mddev->gendisk, mddev->array_sectors);
5802
	revalidate_disk(mddev->gendisk);
5803 5804
	if (sectors > mddev->dev_sectors &&
	    mddev->recovery_cp > mddev->dev_sectors) {
A
Andre Noll 已提交
5805
		mddev->recovery_cp = mddev->dev_sectors;
L
Linus Torvalds 已提交
5806 5807
		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
	}
A
Andre Noll 已提交
5808
	mddev->dev_sectors = sectors;
5809
	mddev->resync_max_sectors = sectors;
L
Linus Torvalds 已提交
5810 5811 5812
	return 0;
}

5813
static int check_stripe_cache(struct mddev *mddev)
5814 5815 5816 5817 5818 5819 5820 5821 5822
{
	/* 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.
	 */
5823
	struct r5conf *conf = mddev->private;
5824 5825 5826 5827
	if (((mddev->chunk_sectors << 9) / STRIPE_SIZE) * 4
	    > conf->max_nr_stripes ||
	    ((mddev->new_chunk_sectors << 9) / STRIPE_SIZE) * 4
	    > conf->max_nr_stripes) {
5828 5829
		printk(KERN_WARNING "md/raid:%s: reshape: not enough stripes.  Needed %lu\n",
		       mdname(mddev),
5830 5831 5832 5833 5834 5835 5836
		       ((max(mddev->chunk_sectors, mddev->new_chunk_sectors) << 9)
			/ STRIPE_SIZE)*4);
		return 0;
	}
	return 1;
}

5837
static int check_reshape(struct mddev *mddev)
5838
{
5839
	struct r5conf *conf = mddev->private;
5840

5841 5842
	if (mddev->delta_disks == 0 &&
	    mddev->new_layout == mddev->layout &&
5843
	    mddev->new_chunk_sectors == mddev->chunk_sectors)
5844
		return 0; /* nothing to do */
5845
	if (has_failed(conf))
5846 5847 5848 5849 5850 5851 5852 5853 5854 5855 5856 5857 5858
		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;
	}
5859

5860
	if (!check_stripe_cache(mddev))
5861 5862
		return -ENOSPC;

5863
	return resize_stripes(conf, conf->raid_disks + mddev->delta_disks);
5864 5865
}

5866
static int raid5_start_reshape(struct mddev *mddev)
5867
{
5868
	struct r5conf *conf = mddev->private;
5869
	struct md_rdev *rdev;
5870
	int spares = 0;
5871
	unsigned long flags;
5872

5873
	if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
5874 5875
		return -EBUSY;

5876 5877 5878
	if (!check_stripe_cache(mddev))
		return -ENOSPC;

5879 5880 5881
	if (has_failed(conf))
		return -EINVAL;

5882
	rdev_for_each(rdev, mddev) {
5883 5884
		if (!test_bit(In_sync, &rdev->flags)
		    && !test_bit(Faulty, &rdev->flags))
5885
			spares++;
5886
	}
5887

5888
	if (spares - mddev->degraded < mddev->delta_disks - conf->max_degraded)
5889 5890 5891 5892 5893
		/* Not enough devices even to make a degraded array
		 * of that size
		 */
		return -EINVAL;

5894 5895 5896 5897 5898 5899
	/* 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) {
5900
		printk(KERN_ERR "md/raid:%s: array size must be reduced "
5901 5902 5903 5904
		       "before number of disks\n", mdname(mddev));
		return -EINVAL;
	}

5905
	atomic_set(&conf->reshape_stripes, 0);
5906 5907
	spin_lock_irq(&conf->device_lock);
	conf->previous_raid_disks = conf->raid_disks;
5908
	conf->raid_disks += mddev->delta_disks;
5909 5910
	conf->prev_chunk_sectors = conf->chunk_sectors;
	conf->chunk_sectors = mddev->new_chunk_sectors;
5911 5912
	conf->prev_algo = conf->algorithm;
	conf->algorithm = mddev->new_layout;
5913 5914 5915 5916 5917
	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();
5918
	if (mddev->reshape_backwards)
5919 5920 5921 5922
		conf->reshape_progress = raid5_size(mddev, 0, 0);
	else
		conf->reshape_progress = 0;
	conf->reshape_safe = conf->reshape_progress;
5923 5924 5925 5926
	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.
5927 5928 5929 5930
	 * 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.
5931
	 */
5932
	if (mddev->delta_disks >= 0) {
N
NeilBrown 已提交
5933
		rdev_for_each(rdev, mddev)
5934 5935 5936 5937
			if (rdev->raid_disk < 0 &&
			    !test_bit(Faulty, &rdev->flags)) {
				if (raid5_add_disk(mddev, rdev) == 0) {
					if (rdev->raid_disk
5938
					    >= conf->previous_raid_disks)
5939
						set_bit(In_sync, &rdev->flags);
5940
					else
5941
						rdev->recovery_offset = 0;
5942 5943

					if (sysfs_link_rdev(mddev, rdev))
5944
						/* Failure here is OK */;
5945
				}
5946 5947 5948 5949 5950
			} 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);
			}
5951

5952 5953 5954 5955
		/* When a reshape changes the number of devices,
		 * ->degraded is measured against the larger of the
		 * pre and post number of devices.
		 */
5956
		spin_lock_irqsave(&conf->device_lock, flags);
5957
		mddev->degraded = calc_degraded(conf);
5958 5959
		spin_unlock_irqrestore(&conf->device_lock, flags);
	}
5960
	mddev->raid_disks = conf->raid_disks;
5961
	mddev->reshape_position = conf->reshape_progress;
5962
	set_bit(MD_CHANGE_DEVS, &mddev->flags);
5963

5964 5965 5966 5967 5968
	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,
5969
						"reshape");
5970 5971 5972 5973
	if (!mddev->sync_thread) {
		mddev->recovery = 0;
		spin_lock_irq(&conf->device_lock);
		mddev->raid_disks = conf->raid_disks = conf->previous_raid_disks;
5974 5975 5976
		rdev_for_each(rdev, mddev)
			rdev->new_data_offset = rdev->data_offset;
		smp_wmb();
5977
		conf->reshape_progress = MaxSector;
5978
		mddev->reshape_position = MaxSector;
5979 5980 5981
		spin_unlock_irq(&conf->device_lock);
		return -EAGAIN;
	}
5982
	conf->reshape_checkpoint = jiffies;
5983 5984 5985 5986 5987
	md_wakeup_thread(mddev->sync_thread);
	md_new_event(mddev);
	return 0;
}

5988 5989 5990
/* This is called from the reshape thread and should make any
 * changes needed in 'conf'
 */
5991
static void end_reshape(struct r5conf *conf)
5992 5993
{

5994
	if (!test_bit(MD_RECOVERY_INTR, &conf->mddev->recovery)) {
5995
		struct md_rdev *rdev;
5996 5997

		spin_lock_irq(&conf->device_lock);
5998
		conf->previous_raid_disks = conf->raid_disks;
5999 6000 6001
		rdev_for_each(rdev, conf->mddev)
			rdev->data_offset = rdev->new_data_offset;
		smp_wmb();
6002
		conf->reshape_progress = MaxSector;
6003
		spin_unlock_irq(&conf->device_lock);
6004
		wake_up(&conf->wait_for_overlap);
6005 6006 6007 6008

		/* read-ahead size must cover two whole stripes, which is
		 * 2 * (datadisks) * chunksize where 'n' is the number of raid devices
		 */
6009
		if (conf->mddev->queue) {
6010
			int data_disks = conf->raid_disks - conf->max_degraded;
6011
			int stripe = data_disks * ((conf->chunk_sectors << 9)
6012
						   / PAGE_SIZE);
6013 6014 6015
			if (conf->mddev->queue->backing_dev_info.ra_pages < 2 * stripe)
				conf->mddev->queue->backing_dev_info.ra_pages = 2 * stripe;
		}
6016 6017 6018
	}
}

6019 6020 6021
/* This is called from the raid5d thread with mddev_lock held.
 * It makes config changes to the device.
 */
6022
static void raid5_finish_reshape(struct mddev *mddev)
6023
{
6024
	struct r5conf *conf = mddev->private;
6025 6026 6027

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

6028 6029 6030
		if (mddev->delta_disks > 0) {
			md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
			set_capacity(mddev->gendisk, mddev->array_sectors);
6031
			revalidate_disk(mddev->gendisk);
6032 6033
		} else {
			int d;
6034 6035 6036
			spin_lock_irq(&conf->device_lock);
			mddev->degraded = calc_degraded(conf);
			spin_unlock_irq(&conf->device_lock);
6037 6038
			for (d = conf->raid_disks ;
			     d < conf->raid_disks - mddev->delta_disks;
6039
			     d++) {
6040
				struct md_rdev *rdev = conf->disks[d].rdev;
6041 6042 6043 6044 6045
				if (rdev)
					clear_bit(In_sync, &rdev->flags);
				rdev = conf->disks[d].replacement;
				if (rdev)
					clear_bit(In_sync, &rdev->flags);
6046
			}
6047
		}
6048
		mddev->layout = conf->algorithm;
6049
		mddev->chunk_sectors = conf->chunk_sectors;
6050 6051
		mddev->reshape_position = MaxSector;
		mddev->delta_disks = 0;
6052
		mddev->reshape_backwards = 0;
6053 6054 6055
	}
}

6056
static void raid5_quiesce(struct mddev *mddev, int state)
6057
{
6058
	struct r5conf *conf = mddev->private;
6059 6060

	switch(state) {
6061 6062 6063 6064
	case 2: /* resume for a suspend */
		wake_up(&conf->wait_for_overlap);
		break;

6065 6066
	case 1: /* stop all writes */
		spin_lock_irq(&conf->device_lock);
6067 6068 6069 6070
		/* '2' tells resync/reshape to pause so that all
		 * active stripes can drain
		 */
		conf->quiesce = 2;
6071
		wait_event_lock_irq(conf->wait_for_stripe,
6072 6073
				    atomic_read(&conf->active_stripes) == 0 &&
				    atomic_read(&conf->active_aligned_reads) == 0,
6074
				    conf->device_lock, /* nothing */);
6075
		conf->quiesce = 1;
6076
		spin_unlock_irq(&conf->device_lock);
6077 6078
		/* allow reshape to continue */
		wake_up(&conf->wait_for_overlap);
6079 6080 6081 6082 6083 6084
		break;

	case 0: /* re-enable writes */
		spin_lock_irq(&conf->device_lock);
		conf->quiesce = 0;
		wake_up(&conf->wait_for_stripe);
6085
		wake_up(&conf->wait_for_overlap);
6086 6087 6088 6089
		spin_unlock_irq(&conf->device_lock);
		break;
	}
}
6090

6091

6092
static void *raid45_takeover_raid0(struct mddev *mddev, int level)
6093
{
6094
	struct r0conf *raid0_conf = mddev->private;
6095
	sector_t sectors;
6096

D
Dan Williams 已提交
6097
	/* for raid0 takeover only one zone is supported */
6098
	if (raid0_conf->nr_strip_zones > 1) {
6099 6100
		printk(KERN_ERR "md/raid:%s: cannot takeover raid0 with more than one zone.\n",
		       mdname(mddev));
D
Dan Williams 已提交
6101 6102 6103
		return ERR_PTR(-EINVAL);
	}

6104 6105
	sectors = raid0_conf->strip_zone[0].zone_end;
	sector_div(sectors, raid0_conf->strip_zone[0].nb_dev);
6106
	mddev->dev_sectors = sectors;
D
Dan Williams 已提交
6107
	mddev->new_level = level;
6108 6109 6110 6111 6112 6113 6114 6115 6116 6117 6118
	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);
}


6119
static void *raid5_takeover_raid1(struct mddev *mddev)
6120 6121 6122 6123 6124 6125 6126 6127 6128 6129 6130 6131 6132 6133 6134 6135 6136 6137 6138 6139 6140
{
	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;
6141
	mddev->new_chunk_sectors = chunksect;
6142 6143 6144 6145

	return setup_conf(mddev);
}

6146
static void *raid5_takeover_raid6(struct mddev *mddev)
6147 6148 6149 6150 6151 6152 6153 6154 6155 6156 6157 6158 6159 6160 6161 6162 6163 6164 6165 6166 6167 6168 6169 6170 6171 6172 6173 6174 6175 6176 6177 6178
{
	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);
}

6179

6180
static int raid5_check_reshape(struct mddev *mddev)
6181
{
6182 6183 6184 6185
	/* 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.
6186
	 */
6187
	struct r5conf *conf = mddev->private;
6188
	int new_chunk = mddev->new_chunk_sectors;
6189

6190
	if (mddev->new_layout >= 0 && !algorithm_valid_raid5(mddev->new_layout))
6191 6192
		return -EINVAL;
	if (new_chunk > 0) {
6193
		if (!is_power_of_2(new_chunk))
6194
			return -EINVAL;
6195
		if (new_chunk < (PAGE_SIZE>>9))
6196
			return -EINVAL;
6197
		if (mddev->array_sectors & (new_chunk-1))
6198 6199 6200 6201 6202 6203
			/* not factor of array size */
			return -EINVAL;
	}

	/* They look valid */

6204
	if (mddev->raid_disks == 2) {
6205 6206 6207 6208
		/* can make the change immediately */
		if (mddev->new_layout >= 0) {
			conf->algorithm = mddev->new_layout;
			mddev->layout = mddev->new_layout;
6209 6210
		}
		if (new_chunk > 0) {
6211 6212
			conf->chunk_sectors = new_chunk ;
			mddev->chunk_sectors = new_chunk;
6213 6214 6215
		}
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
		md_wakeup_thread(mddev->thread);
6216
	}
6217
	return check_reshape(mddev);
6218 6219
}

6220
static int raid6_check_reshape(struct mddev *mddev)
6221
{
6222
	int new_chunk = mddev->new_chunk_sectors;
6223

6224
	if (mddev->new_layout >= 0 && !algorithm_valid_raid6(mddev->new_layout))
6225
		return -EINVAL;
6226
	if (new_chunk > 0) {
6227
		if (!is_power_of_2(new_chunk))
6228
			return -EINVAL;
6229
		if (new_chunk < (PAGE_SIZE >> 9))
6230
			return -EINVAL;
6231
		if (mddev->array_sectors & (new_chunk-1))
6232 6233
			/* not factor of array size */
			return -EINVAL;
6234
	}
6235 6236

	/* They look valid */
6237
	return check_reshape(mddev);
6238 6239
}

6240
static void *raid5_takeover(struct mddev *mddev)
6241 6242
{
	/* raid5 can take over:
D
Dan Williams 已提交
6243
	 *  raid0 - if there is only one strip zone - make it a raid4 layout
6244 6245 6246 6247
	 *  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
6248 6249
	if (mddev->level == 0)
		return raid45_takeover_raid0(mddev, 5);
6250 6251
	if (mddev->level == 1)
		return raid5_takeover_raid1(mddev);
6252 6253 6254 6255 6256
	if (mddev->level == 4) {
		mddev->new_layout = ALGORITHM_PARITY_N;
		mddev->new_level = 5;
		return setup_conf(mddev);
	}
6257 6258
	if (mddev->level == 6)
		return raid5_takeover_raid6(mddev);
6259 6260 6261 6262

	return ERR_PTR(-EINVAL);
}

6263
static void *raid4_takeover(struct mddev *mddev)
6264
{
D
Dan Williams 已提交
6265 6266 6267
	/* raid4 can take over:
	 *  raid0 - if there is only one strip zone
	 *  raid5 - if layout is right
6268
	 */
D
Dan Williams 已提交
6269 6270
	if (mddev->level == 0)
		return raid45_takeover_raid0(mddev, 4);
6271 6272 6273 6274 6275 6276 6277 6278
	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);
}
6279

6280
static struct md_personality raid5_personality;
6281

6282
static void *raid6_takeover(struct mddev *mddev)
6283 6284 6285 6286 6287 6288 6289 6290 6291 6292 6293 6294 6295 6296 6297 6298 6299 6300 6301 6302 6303 6304 6305 6306 6307 6308 6309 6310 6311 6312 6313 6314 6315 6316 6317 6318 6319 6320 6321 6322 6323 6324 6325 6326 6327 6328
{
	/* 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);
}


6329
static struct md_personality raid6_personality =
6330 6331 6332 6333 6334 6335 6336 6337 6338 6339 6340 6341 6342 6343
{
	.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,
6344
	.size		= raid5_size,
6345
	.check_reshape	= raid6_check_reshape,
6346
	.start_reshape  = raid5_start_reshape,
6347
	.finish_reshape = raid5_finish_reshape,
6348
	.quiesce	= raid5_quiesce,
6349
	.takeover	= raid6_takeover,
6350
};
6351
static struct md_personality raid5_personality =
L
Linus Torvalds 已提交
6352 6353
{
	.name		= "raid5",
6354
	.level		= 5,
L
Linus Torvalds 已提交
6355 6356 6357 6358 6359 6360 6361 6362 6363 6364 6365
	.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,
6366
	.size		= raid5_size,
6367 6368
	.check_reshape	= raid5_check_reshape,
	.start_reshape  = raid5_start_reshape,
6369
	.finish_reshape = raid5_finish_reshape,
6370
	.quiesce	= raid5_quiesce,
6371
	.takeover	= raid5_takeover,
L
Linus Torvalds 已提交
6372 6373
};

6374
static struct md_personality raid4_personality =
L
Linus Torvalds 已提交
6375
{
6376 6377 6378 6379 6380 6381 6382 6383 6384 6385 6386 6387 6388
	.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,
6389
	.size		= raid5_size,
6390 6391
	.check_reshape	= raid5_check_reshape,
	.start_reshape  = raid5_start_reshape,
6392
	.finish_reshape = raid5_finish_reshape,
6393
	.quiesce	= raid5_quiesce,
6394
	.takeover	= raid4_takeover,
6395 6396 6397 6398
};

static int __init raid5_init(void)
{
6399
	register_md_personality(&raid6_personality);
6400 6401 6402
	register_md_personality(&raid5_personality);
	register_md_personality(&raid4_personality);
	return 0;
L
Linus Torvalds 已提交
6403 6404
}

6405
static void raid5_exit(void)
L
Linus Torvalds 已提交
6406
{
6407
	unregister_md_personality(&raid6_personality);
6408 6409
	unregister_md_personality(&raid5_personality);
	unregister_md_personality(&raid4_personality);
L
Linus Torvalds 已提交
6410 6411 6412 6413 6414
}

module_init(raid5_init);
module_exit(raid5_exit);
MODULE_LICENSE("GPL");
6415
MODULE_DESCRIPTION("RAID4/5/6 (striping with parity) personality for MD");
L
Linus Torvalds 已提交
6416
MODULE_ALIAS("md-personality-4"); /* RAID5 */
6417 6418
MODULE_ALIAS("md-raid5");
MODULE_ALIAS("md-raid4");
6419 6420
MODULE_ALIAS("md-level-5");
MODULE_ALIAS("md-level-4");
6421 6422 6423 6424 6425 6426 6427
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");