raid5.c 174.1 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
545 546 547 548 549 550
		if (test_and_clear_bit(R5_Wantwrite, &sh->dev[i].flags)) {
			if (test_and_clear_bit(R5_WantFUA, &sh->dev[i].flags))
				rw = WRITE_FUA;
			else
				rw = WRITE;
		} else if (test_and_clear_bit(R5_Wantread, &sh->dev[i].flags))
551
			rw = READ;
552 553 554 555 556
		else if (test_and_clear_bit(R5_WantReplace,
					    &sh->dev[i].flags)) {
			rw = WRITE;
			replace_only = 1;
		} else
557
			continue;
S
Shaohua Li 已提交
558 559
		if (test_and_clear_bit(R5_SyncIO, &sh->dev[i].flags))
			rw |= REQ_SYNC;
560 561

		bi = &sh->dev[i].req;
562
		rbi = &sh->dev[i].rreq; /* For writing to replacement */
563 564

		bi->bi_rw = rw;
565 566
		rbi->bi_rw = rw;
		if (rw & WRITE) {
567
			bi->bi_end_io = raid5_end_write_request;
568 569
			rbi->bi_end_io = raid5_end_write_request;
		} else
570 571 572
			bi->bi_end_io = raid5_end_read_request;

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

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

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

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

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

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

646 647
			bi->bi_bdev = rdev->bdev;
			pr_debug("%s: for %llu schedule op %ld on disc %d\n",
648
				__func__, (unsigned long long)sh->sector,
649 650
				bi->bi_rw, i);
			atomic_inc(&sh->count);
651 652 653 654 655 656
			if (use_new_offset(conf, sh))
				bi->bi_sector = (sh->sector
						 + rdev->new_data_offset);
			else
				bi->bi_sector = (sh->sector
						 + rdev->data_offset);
657 658 659
			if (test_bit(R5_ReadNoMerge, &sh->dev[i].flags))
				bi->bi_rw |= REQ_FLUSH;

660 661 662 663 664 665
			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;
666 667
			if (rrdev)
				set_bit(R5_DOUBLE_LOCKED, &sh->dev[i].flags);
668
			generic_make_request(bi);
669 670
		}
		if (rrdev) {
671 672
			if (s->syncing || s->expanding || s->expanded
			    || s->replacing)
673 674 675 676 677 678 679 680 681 682
				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);
683 684 685 686 687 688
			if (use_new_offset(conf, sh))
				rbi->bi_sector = (sh->sector
						  + rrdev->new_data_offset);
			else
				rbi->bi_sector = (sh->sector
						  + rrdev->data_offset);
689 690 691 692 693 694 695 696 697
			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) {
698
			if (rw & WRITE)
699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715
				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;
716
	struct async_submit_ctl submit;
D
Dan Williams 已提交
717
	enum async_tx_flags flags = 0;
718 719 720 721 722

	if (bio->bi_sector >= sector)
		page_offset = (signed)(bio->bi_sector - sector) * 512;
	else
		page_offset = (signed)(sector - bio->bi_sector) * -512;
723

D
Dan Williams 已提交
724 725 726 727
	if (frombio)
		flags |= ASYNC_TX_FENCE;
	init_async_submit(&submit, flags, tx, NULL, NULL, NULL);

728
	bio_for_each_segment(bvl, bio, i) {
729
		int len = bvl->bv_len;
730 731 732 733 734 735 736 737 738 739 740 741 742 743 744
		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) {
745 746
			b_offset += bvl->bv_offset;
			bio_page = bvl->bv_page;
747 748
			if (frombio)
				tx = async_memcpy(page, bio_page, page_offset,
749
						  b_offset, clen, &submit);
750 751
			else
				tx = async_memcpy(bio_page, page, b_offset,
752
						  page_offset, clen, &submit);
753
		}
754 755 756
		/* chain the operations */
		submit.depend_tx = tx;

757 758 759 760 761 762 763 764 765 766 767 768
		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;
769
	int i;
770

771
	pr_debug("%s: stripe %llu\n", __func__,
772 773 774 775 776 777 778
		(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 */
779 780
		/* and check if we need to reply to a read request,
		 * new R5_Wantfill requests are held off until
781
		 * !STRIPE_BIOFILL_RUN
782 783
		 */
		if (test_and_clear_bit(R5_Wantfill, &dev->flags)) {
784 785 786 787 788 789 790 791
			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);
792
				if (!raid5_dec_bi_active_stripes(rbi)) {
793 794 795 796 797 798 799
					rbi->bi_next = return_bi;
					return_bi = rbi;
				}
				rbi = rbi2;
			}
		}
	}
800
	clear_bit(STRIPE_BIOFILL_RUN, &sh->state);
801 802 803

	return_io(return_bi);

804
	set_bit(STRIPE_HANDLE, &sh->state);
805 806 807 808 809 810
	release_stripe(sh);
}

static void ops_run_biofill(struct stripe_head *sh)
{
	struct dma_async_tx_descriptor *tx = NULL;
811
	struct async_submit_ctl submit;
812 813
	int i;

814
	pr_debug("%s: stripe %llu\n", __func__,
815 816 817 818 819 820
		(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
821
			spin_lock_irq(&sh->stripe_lock);
822 823
			dev->read = rbi = dev->toread;
			dev->toread = NULL;
S
Shaohua Li 已提交
824
			spin_unlock_irq(&sh->stripe_lock);
825 826 827 828 829 830 831 832 833 834
			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);
835 836
	init_async_submit(&submit, ASYNC_TX_ACK, tx, ops_complete_biofill, sh, NULL);
	async_trigger_callback(&submit);
837 838
}

839
static void mark_target_uptodate(struct stripe_head *sh, int target)
840
{
841
	struct r5dev *tgt;
842

843 844
	if (target < 0)
		return;
845

846
	tgt = &sh->dev[target];
847 848 849
	set_bit(R5_UPTODATE, &tgt->flags);
	BUG_ON(!test_bit(R5_Wantcompute, &tgt->flags));
	clear_bit(R5_Wantcompute, &tgt->flags);
850 851
}

852
static void ops_complete_compute(void *stripe_head_ref)
853 854 855
{
	struct stripe_head *sh = stripe_head_ref;

856
	pr_debug("%s: stripe %llu\n", __func__,
857 858
		(unsigned long long)sh->sector);

859
	/* mark the computed target(s) as uptodate */
860
	mark_target_uptodate(sh, sh->ops.target);
861
	mark_target_uptodate(sh, sh->ops.target2);
862

863 864 865
	clear_bit(STRIPE_COMPUTE_RUN, &sh->state);
	if (sh->check_state == check_state_compute_run)
		sh->check_state = check_state_compute_result;
866 867 868 869
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

870 871 872 873 874 875 876 877 878
/* 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)
879 880
{
	int disks = sh->disks;
881
	struct page **xor_srcs = percpu->scribble;
882 883 884 885 886
	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;
887
	struct async_submit_ctl submit;
888 889 890
	int i;

	pr_debug("%s: stripe %llu block: %d\n",
891
		__func__, (unsigned long long)sh->sector, target);
892 893 894 895 896 897 898 899
	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
900
	init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST, NULL,
901
			  ops_complete_compute, sh, to_addr_conv(sh, percpu));
902
	if (unlikely(count == 1))
903
		tx = async_memcpy(xor_dest, xor_srcs[0], 0, 0, STRIPE_SIZE, &submit);
904
	else
905
		tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, &submit);
906 907 908 909

	return tx;
}

910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927
/* 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++)
928
		srcs[i] = NULL;
929 930 931 932 933 934 935 936 937 938

	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);

939
	return syndrome_disks;
940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959
}

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;
960
	else
961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976
		/* 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
977 978
		init_async_submit(&submit, ASYNC_TX_FENCE, NULL,
				  ops_complete_compute, sh,
979 980 981 982 983 984 985 986 987 988 989
				  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
990 991
		init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST,
				  NULL, ops_complete_compute, sh,
992 993 994
				  to_addr_conv(sh, percpu));
		tx = async_xor(dest, blocks, 0, count, STRIPE_SIZE, &submit);
	}
995 996 997 998

	return tx;
}

999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019
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));

1020
	/* we need to open-code set_syndrome_sources to handle the
1021 1022 1023
	 * slot number conversion for 'faila' and 'failb'
	 */
	for (i = 0; i < disks ; i++)
1024
		blocks[i] = NULL;
1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050
	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
1051 1052 1053
			init_async_submit(&submit, ASYNC_TX_FENCE, NULL,
					  ops_complete_compute, sh,
					  to_addr_conv(sh, percpu));
1054
			return async_gen_syndrome(blocks, 0, syndrome_disks+2,
1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073
						  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
1074 1075 1076 1077
			init_async_submit(&submit,
					  ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST,
					  NULL, NULL, NULL,
					  to_addr_conv(sh, percpu));
1078 1079 1080 1081
			tx = async_xor(dest, blocks, 0, count, STRIPE_SIZE,
				       &submit);

			count = set_syndrome_sources(blocks, sh);
D
Dan Williams 已提交
1082 1083 1084
			init_async_submit(&submit, ASYNC_TX_FENCE, tx,
					  ops_complete_compute, sh,
					  to_addr_conv(sh, percpu));
1085 1086 1087 1088
			return async_gen_syndrome(blocks, 0, count+2,
						  STRIPE_SIZE, &submit);
		}
	} else {
1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102
		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);
		}
1103 1104 1105 1106
	}
}


1107 1108 1109 1110
static void ops_complete_prexor(void *stripe_head_ref)
{
	struct stripe_head *sh = stripe_head_ref;

1111
	pr_debug("%s: stripe %llu\n", __func__,
1112 1113 1114 1115
		(unsigned long long)sh->sector);
}

static struct dma_async_tx_descriptor *
1116 1117
ops_run_prexor(struct stripe_head *sh, struct raid5_percpu *percpu,
	       struct dma_async_tx_descriptor *tx)
1118 1119
{
	int disks = sh->disks;
1120
	struct page **xor_srcs = percpu->scribble;
1121
	int count = 0, pd_idx = sh->pd_idx, i;
1122
	struct async_submit_ctl submit;
1123 1124 1125 1126

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

1127
	pr_debug("%s: stripe %llu\n", __func__,
1128 1129 1130 1131 1132
		(unsigned long long)sh->sector);

	for (i = disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
		/* Only process blocks that are known to be uptodate */
1133
		if (test_bit(R5_Wantdrain, &dev->flags))
1134 1135 1136
			xor_srcs[count++] = dev->page;
	}

D
Dan Williams 已提交
1137
	init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_DROP_DST, tx,
1138
			  ops_complete_prexor, sh, to_addr_conv(sh, percpu));
1139
	tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, &submit);
1140 1141 1142 1143 1144

	return tx;
}

static struct dma_async_tx_descriptor *
1145
ops_run_biodrain(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
1146 1147
{
	int disks = sh->disks;
1148
	int i;
1149

1150
	pr_debug("%s: stripe %llu\n", __func__,
1151 1152 1153 1154 1155 1156
		(unsigned long long)sh->sector);

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

1157
		if (test_and_clear_bit(R5_Wantdrain, &dev->flags)) {
1158 1159
			struct bio *wbi;

S
Shaohua Li 已提交
1160
			spin_lock_irq(&sh->stripe_lock);
1161 1162 1163 1164
			chosen = dev->towrite;
			dev->towrite = NULL;
			BUG_ON(dev->written);
			wbi = dev->written = chosen;
S
Shaohua Li 已提交
1165
			spin_unlock_irq(&sh->stripe_lock);
1166 1167 1168

			while (wbi && wbi->bi_sector <
				dev->sector + STRIPE_SECTORS) {
T
Tejun Heo 已提交
1169 1170
				if (wbi->bi_rw & REQ_FUA)
					set_bit(R5_WantFUA, &dev->flags);
S
Shaohua Li 已提交
1171 1172
				if (wbi->bi_rw & REQ_SYNC)
					set_bit(R5_SyncIO, &dev->flags);
1173 1174 1175 1176 1177 1178 1179 1180 1181 1182
				tx = async_copy_data(1, wbi, dev->page,
					dev->sector, tx);
				wbi = r5_next_bio(wbi, dev->sector);
			}
		}
	}

	return tx;
}

1183
static void ops_complete_reconstruct(void *stripe_head_ref)
1184 1185
{
	struct stripe_head *sh = stripe_head_ref;
1186 1187 1188 1189
	int disks = sh->disks;
	int pd_idx = sh->pd_idx;
	int qd_idx = sh->qd_idx;
	int i;
S
Shaohua Li 已提交
1190
	bool fua = false, sync = false;
1191

1192
	pr_debug("%s: stripe %llu\n", __func__,
1193 1194
		(unsigned long long)sh->sector);

S
Shaohua Li 已提交
1195
	for (i = disks; i--; ) {
T
Tejun Heo 已提交
1196
		fua |= test_bit(R5_WantFUA, &sh->dev[i].flags);
S
Shaohua Li 已提交
1197 1198
		sync |= test_bit(R5_SyncIO, &sh->dev[i].flags);
	}
T
Tejun Heo 已提交
1199

1200 1201
	for (i = disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
1202

T
Tejun Heo 已提交
1203
		if (dev->written || i == pd_idx || i == qd_idx) {
1204
			set_bit(R5_UPTODATE, &dev->flags);
T
Tejun Heo 已提交
1205 1206
			if (fua)
				set_bit(R5_WantFUA, &dev->flags);
S
Shaohua Li 已提交
1207 1208
			if (sync)
				set_bit(R5_SyncIO, &dev->flags);
T
Tejun Heo 已提交
1209
		}
1210 1211
	}

1212 1213 1214 1215 1216 1217 1218 1219
	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;
	}
1220 1221 1222 1223 1224 1225

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

static void
1226 1227
ops_run_reconstruct5(struct stripe_head *sh, struct raid5_percpu *percpu,
		     struct dma_async_tx_descriptor *tx)
1228 1229
{
	int disks = sh->disks;
1230
	struct page **xor_srcs = percpu->scribble;
1231
	struct async_submit_ctl submit;
1232 1233
	int count = 0, pd_idx = sh->pd_idx, i;
	struct page *xor_dest;
1234
	int prexor = 0;
1235 1236
	unsigned long flags;

1237
	pr_debug("%s: stripe %llu\n", __func__,
1238 1239 1240 1241 1242
		(unsigned long long)sh->sector);

	/* check if prexor is active which means only process blocks
	 * that are part of a read-modify-write (written)
	 */
1243 1244
	if (sh->reconstruct_state == reconstruct_state_prexor_drain_run) {
		prexor = 1;
1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264
		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
	 */
1265
	flags = ASYNC_TX_ACK |
1266 1267 1268 1269
		(prexor ? ASYNC_TX_XOR_DROP_DST : ASYNC_TX_XOR_ZERO_DST);

	atomic_inc(&sh->count);

1270
	init_async_submit(&submit, flags, tx, ops_complete_reconstruct, sh,
1271
			  to_addr_conv(sh, percpu));
1272 1273 1274 1275
	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);
1276 1277
}

1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294
static void
ops_run_reconstruct6(struct stripe_head *sh, struct raid5_percpu *percpu,
		     struct dma_async_tx_descriptor *tx)
{
	struct async_submit_ctl submit;
	struct page **blocks = percpu->scribble;
	int count;

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

	count = set_syndrome_sources(blocks, sh);

	atomic_inc(&sh->count);

	init_async_submit(&submit, ASYNC_TX_ACK, tx, ops_complete_reconstruct,
			  sh, to_addr_conv(sh, percpu));
	async_gen_syndrome(blocks, 0, count+2, STRIPE_SIZE,  &submit);
1295 1296 1297 1298 1299 1300
}

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

1301
	pr_debug("%s: stripe %llu\n", __func__,
1302 1303
		(unsigned long long)sh->sector);

1304
	sh->check_state = check_state_check_result;
1305 1306 1307 1308
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

1309
static void ops_run_check_p(struct stripe_head *sh, struct raid5_percpu *percpu)
1310 1311
{
	int disks = sh->disks;
1312 1313 1314
	int pd_idx = sh->pd_idx;
	int qd_idx = sh->qd_idx;
	struct page *xor_dest;
1315
	struct page **xor_srcs = percpu->scribble;
1316
	struct dma_async_tx_descriptor *tx;
1317
	struct async_submit_ctl submit;
1318 1319
	int count;
	int i;
1320

1321
	pr_debug("%s: stripe %llu\n", __func__,
1322 1323
		(unsigned long long)sh->sector);

1324 1325 1326
	count = 0;
	xor_dest = sh->dev[pd_idx].page;
	xor_srcs[count++] = xor_dest;
1327
	for (i = disks; i--; ) {
1328 1329 1330
		if (i == pd_idx || i == qd_idx)
			continue;
		xor_srcs[count++] = sh->dev[i].page;
1331 1332
	}

1333 1334
	init_async_submit(&submit, 0, NULL, NULL, NULL,
			  to_addr_conv(sh, percpu));
D
Dan Williams 已提交
1335
	tx = async_xor_val(xor_dest, xor_srcs, 0, count, STRIPE_SIZE,
1336
			   &sh->ops.zero_sum_result, &submit);
1337 1338

	atomic_inc(&sh->count);
1339 1340
	init_async_submit(&submit, ASYNC_TX_ACK, tx, ops_complete_check, sh, NULL);
	tx = async_trigger_callback(&submit);
1341 1342
}

1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354
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;
1355 1356

	atomic_inc(&sh->count);
1357 1358 1359 1360
	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);
1361 1362
}

1363
static void __raid_run_ops(struct stripe_head *sh, unsigned long ops_request)
1364 1365 1366
{
	int overlap_clear = 0, i, disks = sh->disks;
	struct dma_async_tx_descriptor *tx = NULL;
1367
	struct r5conf *conf = sh->raid_conf;
1368
	int level = conf->level;
1369 1370
	struct raid5_percpu *percpu;
	unsigned long cpu;
1371

1372 1373
	cpu = get_cpu();
	percpu = per_cpu_ptr(conf->percpu, cpu);
1374
	if (test_bit(STRIPE_OP_BIOFILL, &ops_request)) {
1375 1376 1377 1378
		ops_run_biofill(sh);
		overlap_clear++;
	}

1379
	if (test_bit(STRIPE_OP_COMPUTE_BLK, &ops_request)) {
1380 1381 1382 1383 1384 1385 1386 1387 1388 1389
		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))
1390 1391
			async_tx_ack(tx);
	}
1392

1393
	if (test_bit(STRIPE_OP_PREXOR, &ops_request))
1394
		tx = ops_run_prexor(sh, percpu, tx);
1395

1396
	if (test_bit(STRIPE_OP_BIODRAIN, &ops_request)) {
1397
		tx = ops_run_biodrain(sh, tx);
1398 1399 1400
		overlap_clear++;
	}

1401 1402 1403 1404 1405 1406
	if (test_bit(STRIPE_OP_RECONSTRUCT, &ops_request)) {
		if (level < 6)
			ops_run_reconstruct5(sh, percpu, tx);
		else
			ops_run_reconstruct6(sh, percpu, tx);
	}
1407

1408 1409 1410 1411 1412 1413 1414 1415 1416 1417
	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();
	}
1418 1419 1420 1421 1422 1423 1424

	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);
		}
1425
	put_cpu();
1426 1427
}

1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457
#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

1458
static int grow_one_stripe(struct r5conf *conf)
L
Linus Torvalds 已提交
1459 1460
{
	struct stripe_head *sh;
N
Namhyung Kim 已提交
1461
	sh = kmem_cache_zalloc(conf->slab_cache, GFP_KERNEL);
1462 1463
	if (!sh)
		return 0;
N
Namhyung Kim 已提交
1464

1465
	sh->raid_conf = conf;
1466 1467 1468
	#ifdef CONFIG_MULTICORE_RAID456
	init_waitqueue_head(&sh->ops.wait_for_ops);
	#endif
1469

S
Shaohua Li 已提交
1470 1471
	spin_lock_init(&sh->stripe_lock);

1472 1473
	if (grow_buffers(sh)) {
		shrink_buffers(sh);
1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484
		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;
}

1485
static int grow_stripes(struct r5conf *conf, int num)
1486
{
1487
	struct kmem_cache *sc;
1488
	int devs = max(conf->raid_disks, conf->previous_raid_disks);
L
Linus Torvalds 已提交
1489

1490 1491 1492 1493 1494 1495 1496 1497
	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]);

1498 1499
	conf->active_name = 0;
	sc = kmem_cache_create(conf->cache_name[conf->active_name],
L
Linus Torvalds 已提交
1500
			       sizeof(struct stripe_head)+(devs-1)*sizeof(struct r5dev),
1501
			       0, 0, NULL);
L
Linus Torvalds 已提交
1502 1503 1504
	if (!sc)
		return 1;
	conf->slab_cache = sc;
1505
	conf->pool_size = devs;
1506
	while (num--)
1507
		if (!grow_one_stripe(conf))
L
Linus Torvalds 已提交
1508 1509 1510
			return 1;
	return 0;
}
1511

1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533
/**
 * 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;
}

1534
static int resize_stripes(struct r5conf *conf, int newsize)
1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561
{
	/* 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;
1562
	unsigned long cpu;
1563
	int err;
1564
	struct kmem_cache *sc;
1565 1566 1567 1568 1569
	int i;

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

1570 1571 1572
	err = md_allow_write(conf->mddev);
	if (err)
		return err;
1573

1574 1575 1576
	/* Step 1 */
	sc = kmem_cache_create(conf->cache_name[1-conf->active_name],
			       sizeof(struct stripe_head)+(newsize-1)*sizeof(struct r5dev),
1577
			       0, 0, NULL);
1578 1579 1580 1581
	if (!sc)
		return -ENOMEM;

	for (i = conf->max_nr_stripes; i; i--) {
N
Namhyung Kim 已提交
1582
		nsh = kmem_cache_zalloc(sc, GFP_KERNEL);
1583 1584 1585 1586
		if (!nsh)
			break;

		nsh->raid_conf = conf;
1587 1588 1589
		#ifdef CONFIG_MULTICORE_RAID456
		init_waitqueue_head(&nsh->ops.wait_for_ops);
		#endif
1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611

		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
1612
				    );
1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626
		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
1627
	 * conf->disks and the scribble region
1628 1629 1630 1631 1632 1633 1634 1635 1636 1637
	 */
	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;

1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656
	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();

1657 1658 1659 1660
	/* 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);
1661

1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677
		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
1678

1679
static int drop_one_stripe(struct r5conf *conf)
L
Linus Torvalds 已提交
1680 1681 1682
{
	struct stripe_head *sh;

1683 1684 1685 1686 1687
	spin_lock_irq(&conf->device_lock);
	sh = get_free_stripe(conf);
	spin_unlock_irq(&conf->device_lock);
	if (!sh)
		return 0;
1688
	BUG_ON(atomic_read(&sh->count));
1689
	shrink_buffers(sh);
1690 1691 1692 1693 1694
	kmem_cache_free(conf->slab_cache, sh);
	atomic_dec(&conf->active_stripes);
	return 1;
}

1695
static void shrink_stripes(struct r5conf *conf)
1696 1697 1698 1699
{
	while (drop_one_stripe(conf))
		;

N
NeilBrown 已提交
1700 1701
	if (conf->slab_cache)
		kmem_cache_destroy(conf->slab_cache);
L
Linus Torvalds 已提交
1702 1703 1704
	conf->slab_cache = NULL;
}

1705
static void raid5_end_read_request(struct bio * bi, int error)
L
Linus Torvalds 已提交
1706
{
1707
	struct stripe_head *sh = bi->bi_private;
1708
	struct r5conf *conf = sh->raid_conf;
1709
	int disks = sh->disks, i;
L
Linus Torvalds 已提交
1710
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
1711
	char b[BDEVNAME_SIZE];
1712
	struct md_rdev *rdev = NULL;
1713
	sector_t s;
L
Linus Torvalds 已提交
1714 1715 1716 1717 1718

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

1719 1720
	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
1721 1722 1723
		uptodate);
	if (i == disks) {
		BUG();
1724
		return;
L
Linus Torvalds 已提交
1725
	}
1726
	if (test_bit(R5_ReadRepl, &sh->dev[i].flags))
1727 1728 1729 1730 1731
		/* 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.
		 */
1732
		rdev = conf->disks[i].replacement;
1733
	if (!rdev)
1734
		rdev = conf->disks[i].rdev;
L
Linus Torvalds 已提交
1735

1736 1737 1738 1739
	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
1740 1741
	if (uptodate) {
		set_bit(R5_UPTODATE, &sh->dev[i].flags);
1742
		if (test_bit(R5_ReadError, &sh->dev[i].flags)) {
1743 1744 1745 1746
			/* Note that this cannot happen on a
			 * replacement device.  We just fail those on
			 * any error
			 */
1747 1748 1749 1750 1751
			printk_ratelimited(
				KERN_INFO
				"md/raid:%s: read error corrected"
				" (%lu sectors at %llu on %s)\n",
				mdname(conf->mddev), STRIPE_SECTORS,
1752
				(unsigned long long)s,
1753
				bdevname(rdev->bdev, b));
1754
			atomic_add(STRIPE_SECTORS, &rdev->corrected_errors);
1755 1756
			clear_bit(R5_ReadError, &sh->dev[i].flags);
			clear_bit(R5_ReWrite, &sh->dev[i].flags);
1757 1758 1759
		} else if (test_bit(R5_ReadNoMerge, &sh->dev[i].flags))
			clear_bit(R5_ReadNoMerge, &sh->dev[i].flags);

1760 1761
		if (atomic_read(&rdev->read_errors))
			atomic_set(&rdev->read_errors, 0);
L
Linus Torvalds 已提交
1762
	} else {
1763
		const char *bdn = bdevname(rdev->bdev, b);
1764
		int retry = 0;
1765
		int set_bad = 0;
1766

L
Linus Torvalds 已提交
1767
		clear_bit(R5_UPTODATE, &sh->dev[i].flags);
1768
		atomic_inc(&rdev->read_errors);
1769 1770 1771 1772 1773 1774
		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),
1775
				(unsigned long long)s,
1776
				bdn);
1777 1778
		else if (conf->mddev->degraded >= conf->max_degraded) {
			set_bad = 1;
1779 1780 1781 1782 1783
			printk_ratelimited(
				KERN_WARNING
				"md/raid:%s: read error not correctable "
				"(sector %llu on %s).\n",
				mdname(conf->mddev),
1784
				(unsigned long long)s,
1785
				bdn);
1786
		} else if (test_bit(R5_ReWrite, &sh->dev[i].flags)) {
1787
			/* Oh, no!!! */
1788
			set_bad = 1;
1789 1790 1791 1792 1793
			printk_ratelimited(
				KERN_WARNING
				"md/raid:%s: read error NOT corrected!! "
				"(sector %llu on %s).\n",
				mdname(conf->mddev),
1794
				(unsigned long long)s,
1795
				bdn);
1796
		} else if (atomic_read(&rdev->read_errors)
1797
			 > conf->max_nr_stripes)
N
NeilBrown 已提交
1798
			printk(KERN_WARNING
1799
			       "md/raid:%s: Too many read errors, failing device %s.\n",
1800
			       mdname(conf->mddev), bdn);
1801 1802 1803
		else
			retry = 1;
		if (retry)
1804 1805 1806 1807 1808
			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);
1809
		else {
1810 1811
			clear_bit(R5_ReadError, &sh->dev[i].flags);
			clear_bit(R5_ReWrite, &sh->dev[i].flags);
1812 1813 1814 1815 1816
			if (!(set_bad
			      && test_bit(In_sync, &rdev->flags)
			      && rdev_set_badblocks(
				      rdev, sh->sector, STRIPE_SECTORS, 0)))
				md_error(conf->mddev, rdev);
1817
		}
L
Linus Torvalds 已提交
1818
	}
1819
	rdev_dec_pending(rdev, conf->mddev);
L
Linus Torvalds 已提交
1820 1821 1822 1823 1824
	clear_bit(R5_LOCKED, &sh->dev[i].flags);
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

1825
static void raid5_end_write_request(struct bio *bi, int error)
L
Linus Torvalds 已提交
1826
{
1827
	struct stripe_head *sh = bi->bi_private;
1828
	struct r5conf *conf = sh->raid_conf;
1829
	int disks = sh->disks, i;
1830
	struct md_rdev *uninitialized_var(rdev);
L
Linus Torvalds 已提交
1831
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
1832 1833
	sector_t first_bad;
	int bad_sectors;
1834
	int replacement = 0;
L
Linus Torvalds 已提交
1835

1836 1837 1838
	for (i = 0 ; i < disks; i++) {
		if (bi == &sh->dev[i].req) {
			rdev = conf->disks[i].rdev;
L
Linus Torvalds 已提交
1839
			break;
1840 1841 1842
		}
		if (bi == &sh->dev[i].rreq) {
			rdev = conf->disks[i].replacement;
1843 1844 1845 1846 1847 1848 1849 1850
			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;
1851 1852 1853
			break;
		}
	}
1854
	pr_debug("end_write_request %llu/%d, count %d, uptodate: %d.\n",
L
Linus Torvalds 已提交
1855 1856 1857 1858
		(unsigned long long)sh->sector, i, atomic_read(&sh->count),
		uptodate);
	if (i == disks) {
		BUG();
1859
		return;
L
Linus Torvalds 已提交
1860 1861
	}

1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872
	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);
1873 1874 1875
			if (!test_and_set_bit(WantReplacement, &rdev->flags))
				set_bit(MD_RECOVERY_NEEDED,
					&rdev->mddev->recovery);
1876 1877 1878 1879 1880 1881
		} 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
1882

1883 1884
	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
1885
	set_bit(STRIPE_HANDLE, &sh->state);
1886
	release_stripe(sh);
L
Linus Torvalds 已提交
1887 1888
}

1889
static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous);
L
Linus Torvalds 已提交
1890
	
1891
static void raid5_build_block(struct stripe_head *sh, int i, int previous)
L
Linus Torvalds 已提交
1892 1893 1894 1895 1896 1897 1898 1899
{
	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;
1900
	dev->vec.bv_page = dev->page;
L
Linus Torvalds 已提交
1901

1902 1903 1904 1905 1906 1907 1908
	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
1909
	dev->flags = 0;
1910
	dev->sector = compute_blocknr(sh, i, previous);
L
Linus Torvalds 已提交
1911 1912
}

1913
static void error(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
1914 1915
{
	char b[BDEVNAME_SIZE];
1916
	struct r5conf *conf = mddev->private;
1917
	unsigned long flags;
1918
	pr_debug("raid456: error called\n");
L
Linus Torvalds 已提交
1919

1920 1921 1922 1923 1924 1925
	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);

1926
	set_bit(Blocked, &rdev->flags);
1927 1928 1929 1930 1931 1932 1933 1934 1935
	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);
1936
}
L
Linus Torvalds 已提交
1937 1938 1939 1940 1941

/*
 * Input: a 'big' sector number,
 * Output: index of the data and parity disk, and the sector # in them.
 */
1942
static sector_t raid5_compute_sector(struct r5conf *conf, sector_t r_sector,
1943 1944
				     int previous, int *dd_idx,
				     struct stripe_head *sh)
L
Linus Torvalds 已提交
1945
{
N
NeilBrown 已提交
1946
	sector_t stripe, stripe2;
1947
	sector_t chunk_number;
L
Linus Torvalds 已提交
1948
	unsigned int chunk_offset;
1949
	int pd_idx, qd_idx;
1950
	int ddf_layout = 0;
L
Linus Torvalds 已提交
1951
	sector_t new_sector;
1952 1953
	int algorithm = previous ? conf->prev_algo
				 : conf->algorithm;
1954 1955
	int sectors_per_chunk = previous ? conf->prev_chunk_sectors
					 : conf->chunk_sectors;
1956 1957 1958
	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
1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970

	/* 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
	 */
1971 1972
	stripe = chunk_number;
	*dd_idx = sector_div(stripe, data_disks);
N
NeilBrown 已提交
1973
	stripe2 = stripe;
L
Linus Torvalds 已提交
1974 1975 1976
	/*
	 * Select the parity disk based on the user selected algorithm.
	 */
1977
	pd_idx = qd_idx = -1;
1978 1979
	switch(conf->level) {
	case 4:
1980
		pd_idx = data_disks;
1981 1982
		break;
	case 5:
1983
		switch (algorithm) {
L
Linus Torvalds 已提交
1984
		case ALGORITHM_LEFT_ASYMMETRIC:
N
NeilBrown 已提交
1985
			pd_idx = data_disks - sector_div(stripe2, raid_disks);
1986
			if (*dd_idx >= pd_idx)
L
Linus Torvalds 已提交
1987 1988 1989
				(*dd_idx)++;
			break;
		case ALGORITHM_RIGHT_ASYMMETRIC:
N
NeilBrown 已提交
1990
			pd_idx = sector_div(stripe2, raid_disks);
1991
			if (*dd_idx >= pd_idx)
L
Linus Torvalds 已提交
1992 1993 1994
				(*dd_idx)++;
			break;
		case ALGORITHM_LEFT_SYMMETRIC:
N
NeilBrown 已提交
1995
			pd_idx = data_disks - sector_div(stripe2, raid_disks);
1996
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
L
Linus Torvalds 已提交
1997 1998
			break;
		case ALGORITHM_RIGHT_SYMMETRIC:
N
NeilBrown 已提交
1999
			pd_idx = sector_div(stripe2, raid_disks);
2000
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
L
Linus Torvalds 已提交
2001
			break;
2002 2003 2004 2005 2006 2007 2008
		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
2009
		default:
2010
			BUG();
2011 2012 2013 2014
		}
		break;
	case 6:

2015
		switch (algorithm) {
2016
		case ALGORITHM_LEFT_ASYMMETRIC:
N
NeilBrown 已提交
2017
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2018 2019
			qd_idx = pd_idx + 1;
			if (pd_idx == raid_disks-1) {
2020
				(*dd_idx)++;	/* Q D D D P */
2021 2022
				qd_idx = 0;
			} else if (*dd_idx >= pd_idx)
2023 2024 2025
				(*dd_idx) += 2; /* D D P Q D */
			break;
		case ALGORITHM_RIGHT_ASYMMETRIC:
N
NeilBrown 已提交
2026
			pd_idx = sector_div(stripe2, raid_disks);
2027 2028
			qd_idx = pd_idx + 1;
			if (pd_idx == raid_disks-1) {
2029
				(*dd_idx)++;	/* Q D D D P */
2030 2031
				qd_idx = 0;
			} else if (*dd_idx >= pd_idx)
2032 2033 2034
				(*dd_idx) += 2; /* D D P Q D */
			break;
		case ALGORITHM_LEFT_SYMMETRIC:
N
NeilBrown 已提交
2035
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2036 2037
			qd_idx = (pd_idx + 1) % raid_disks;
			*dd_idx = (pd_idx + 2 + *dd_idx) % raid_disks;
2038 2039
			break;
		case ALGORITHM_RIGHT_SYMMETRIC:
N
NeilBrown 已提交
2040
			pd_idx = sector_div(stripe2, raid_disks);
2041 2042
			qd_idx = (pd_idx + 1) % raid_disks;
			*dd_idx = (pd_idx + 2 + *dd_idx) % raid_disks;
2043
			break;
2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058

		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
2059
			pd_idx = sector_div(stripe2, raid_disks);
2060 2061 2062 2063 2064 2065
			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 */
2066
			ddf_layout = 1;
2067 2068 2069 2070 2071 2072 2073
			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
2074 2075
			stripe2 += 1;
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2076 2077 2078 2079 2080 2081
			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 */
2082
			ddf_layout = 1;
2083 2084 2085 2086
			break;

		case ALGORITHM_ROTATING_N_CONTINUE:
			/* Same as left_symmetric but Q is before P */
N
NeilBrown 已提交
2087
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2088 2089
			qd_idx = (pd_idx + raid_disks - 1) % raid_disks;
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
2090
			ddf_layout = 1;
2091 2092 2093 2094
			break;

		case ALGORITHM_LEFT_ASYMMETRIC_6:
			/* RAID5 left_asymmetric, with Q on last device */
N
NeilBrown 已提交
2095
			pd_idx = data_disks - sector_div(stripe2, raid_disks-1);
2096 2097 2098 2099 2100 2101
			if (*dd_idx >= pd_idx)
				(*dd_idx)++;
			qd_idx = raid_disks - 1;
			break;

		case ALGORITHM_RIGHT_ASYMMETRIC_6:
N
NeilBrown 已提交
2102
			pd_idx = sector_div(stripe2, raid_disks-1);
2103 2104 2105 2106 2107 2108
			if (*dd_idx >= pd_idx)
				(*dd_idx)++;
			qd_idx = raid_disks - 1;
			break;

		case ALGORITHM_LEFT_SYMMETRIC_6:
N
NeilBrown 已提交
2109
			pd_idx = data_disks - sector_div(stripe2, raid_disks-1);
2110 2111 2112 2113 2114
			*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
2115
			pd_idx = sector_div(stripe2, raid_disks-1);
2116 2117 2118 2119 2120 2121 2122 2123 2124 2125
			*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;

2126
		default:
2127
			BUG();
2128 2129
		}
		break;
L
Linus Torvalds 已提交
2130 2131
	}

2132 2133 2134
	if (sh) {
		sh->pd_idx = pd_idx;
		sh->qd_idx = qd_idx;
2135
		sh->ddf_layout = ddf_layout;
2136
	}
L
Linus Torvalds 已提交
2137 2138 2139 2140 2141 2142 2143 2144
	/*
	 * Finally, compute the new sector number
	 */
	new_sector = (sector_t)stripe * sectors_per_chunk + chunk_offset;
	return new_sector;
}


2145
static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous)
L
Linus Torvalds 已提交
2146
{
2147
	struct r5conf *conf = sh->raid_conf;
2148 2149
	int raid_disks = sh->disks;
	int data_disks = raid_disks - conf->max_degraded;
L
Linus Torvalds 已提交
2150
	sector_t new_sector = sh->sector, check;
2151 2152
	int sectors_per_chunk = previous ? conf->prev_chunk_sectors
					 : conf->chunk_sectors;
2153 2154
	int algorithm = previous ? conf->prev_algo
				 : conf->algorithm;
L
Linus Torvalds 已提交
2155 2156
	sector_t stripe;
	int chunk_offset;
2157 2158
	sector_t chunk_number;
	int dummy1, dd_idx = i;
L
Linus Torvalds 已提交
2159
	sector_t r_sector;
2160
	struct stripe_head sh2;
L
Linus Torvalds 已提交
2161

2162

L
Linus Torvalds 已提交
2163 2164 2165
	chunk_offset = sector_div(new_sector, sectors_per_chunk);
	stripe = new_sector;

2166 2167 2168 2169 2170
	if (i == sh->pd_idx)
		return 0;
	switch(conf->level) {
	case 4: break;
	case 5:
2171
		switch (algorithm) {
L
Linus Torvalds 已提交
2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182
		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;
2183 2184 2185 2186 2187
		case ALGORITHM_PARITY_0:
			i -= 1;
			break;
		case ALGORITHM_PARITY_N:
			break;
L
Linus Torvalds 已提交
2188
		default:
2189
			BUG();
2190 2191 2192
		}
		break;
	case 6:
2193
		if (i == sh->qd_idx)
2194
			return 0; /* It is the Q disk */
2195
		switch (algorithm) {
2196 2197
		case ALGORITHM_LEFT_ASYMMETRIC:
		case ALGORITHM_RIGHT_ASYMMETRIC:
2198 2199 2200 2201
		case ALGORITHM_ROTATING_ZERO_RESTART:
		case ALGORITHM_ROTATING_N_RESTART:
			if (sh->pd_idx == raid_disks-1)
				i--;	/* Q D D D P */
2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215
			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;
2216 2217 2218 2219 2220 2221
		case ALGORITHM_PARITY_0:
			i -= 2;
			break;
		case ALGORITHM_PARITY_N:
			break;
		case ALGORITHM_ROTATING_N_CONTINUE:
2222
			/* Like left_symmetric, but P is before Q */
2223 2224
			if (sh->pd_idx == 0)
				i--;	/* P D D D Q */
2225 2226 2227 2228 2229 2230
			else {
				/* D D Q P D */
				if (i < sh->pd_idx)
					i += raid_disks;
				i -= (sh->pd_idx + 1);
			}
2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245
			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;
2246
		default:
2247
			BUG();
2248 2249
		}
		break;
L
Linus Torvalds 已提交
2250 2251 2252
	}

	chunk_number = stripe * data_disks + i;
2253
	r_sector = chunk_number * sectors_per_chunk + chunk_offset;
L
Linus Torvalds 已提交
2254

2255
	check = raid5_compute_sector(conf, r_sector,
2256
				     previous, &dummy1, &sh2);
2257 2258
	if (check != sh->sector || dummy1 != dd_idx || sh2.pd_idx != sh->pd_idx
		|| sh2.qd_idx != sh->qd_idx) {
2259 2260
		printk(KERN_ERR "md/raid:%s: compute_blocknr: map not correct\n",
		       mdname(conf->mddev));
L
Linus Torvalds 已提交
2261 2262 2263 2264 2265 2266
		return 0;
	}
	return r_sector;
}


2267
static void
2268
schedule_reconstruction(struct stripe_head *sh, struct stripe_head_state *s,
2269
			 int rcw, int expand)
2270 2271
{
	int i, pd_idx = sh->pd_idx, disks = sh->disks;
2272
	struct r5conf *conf = sh->raid_conf;
2273
	int level = conf->level;
2274 2275 2276 2277 2278 2279 2280

	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) {
2281 2282 2283 2284
			sh->reconstruct_state = reconstruct_state_drain_run;
			set_bit(STRIPE_OP_BIODRAIN, &s->ops_request);
		} else
			sh->reconstruct_state = reconstruct_state_run;
2285

2286
		set_bit(STRIPE_OP_RECONSTRUCT, &s->ops_request);
2287 2288 2289 2290 2291 2292

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

			if (dev->towrite) {
				set_bit(R5_LOCKED, &dev->flags);
2293
				set_bit(R5_Wantdrain, &dev->flags);
2294 2295
				if (!expand)
					clear_bit(R5_UPTODATE, &dev->flags);
2296
				s->locked++;
2297 2298
			}
		}
2299
		if (s->locked + conf->max_degraded == disks)
2300
			if (!test_and_set_bit(STRIPE_FULL_WRITE, &sh->state))
2301
				atomic_inc(&conf->pending_full_writes);
2302
	} else {
2303
		BUG_ON(level == 6);
2304 2305 2306
		BUG_ON(!(test_bit(R5_UPTODATE, &sh->dev[pd_idx].flags) ||
			test_bit(R5_Wantcompute, &sh->dev[pd_idx].flags)));

2307
		sh->reconstruct_state = reconstruct_state_prexor_drain_run;
2308 2309
		set_bit(STRIPE_OP_PREXOR, &s->ops_request);
		set_bit(STRIPE_OP_BIODRAIN, &s->ops_request);
2310
		set_bit(STRIPE_OP_RECONSTRUCT, &s->ops_request);
2311 2312 2313 2314 2315 2316 2317 2318

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

			if (dev->towrite &&
			    (test_bit(R5_UPTODATE, &dev->flags) ||
2319 2320
			     test_bit(R5_Wantcompute, &dev->flags))) {
				set_bit(R5_Wantdrain, &dev->flags);
2321 2322
				set_bit(R5_LOCKED, &dev->flags);
				clear_bit(R5_UPTODATE, &dev->flags);
2323
				s->locked++;
2324 2325 2326 2327
			}
		}
	}

2328
	/* keep the parity disk(s) locked while asynchronous operations
2329 2330 2331 2332
	 * are in flight
	 */
	set_bit(R5_LOCKED, &sh->dev[pd_idx].flags);
	clear_bit(R5_UPTODATE, &sh->dev[pd_idx].flags);
2333
	s->locked++;
2334

2335 2336 2337 2338 2339 2340 2341 2342 2343
	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++;
	}

2344
	pr_debug("%s: stripe %llu locked: %d ops_request: %lx\n",
2345
		__func__, (unsigned long long)sh->sector,
2346
		s->locked, s->ops_request);
2347
}
2348

L
Linus Torvalds 已提交
2349 2350
/*
 * Each stripe/dev can have one or more bion attached.
2351
 * toread/towrite point to the first in a chain.
L
Linus Torvalds 已提交
2352 2353 2354 2355 2356
 * 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;
2357
	struct r5conf *conf = sh->raid_conf;
2358
	int firstwrite=0;
L
Linus Torvalds 已提交
2359

2360
	pr_debug("adding bi b#%llu to stripe s#%llu\n",
L
Linus Torvalds 已提交
2361 2362 2363
		(unsigned long long)bi->bi_sector,
		(unsigned long long)sh->sector);

S
Shaohua Li 已提交
2364 2365 2366 2367 2368 2369 2370 2371 2372
	/*
	 * 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);
2373
	if (forwrite) {
L
Linus Torvalds 已提交
2374
		bip = &sh->dev[dd_idx].towrite;
2375
		if (*bip == NULL)
2376 2377
			firstwrite = 1;
	} else
L
Linus Torvalds 已提交
2378 2379 2380 2381 2382 2383 2384 2385 2386
		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;

2387
	BUG_ON(*bip && bi->bi_next && (*bip) != bi->bi_next);
L
Linus Torvalds 已提交
2388 2389 2390
	if (*bip)
		bi->bi_next = *bip;
	*bip = bi;
2391
	raid5_inc_bi_active_stripes(bi);
2392

L
Linus Torvalds 已提交
2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405
	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
2406
	spin_unlock_irq(&sh->stripe_lock);
2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417

	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
2418 2419 2420 2421
	return 1;

 overlap:
	set_bit(R5_Overlap, &sh->dev[dd_idx].flags);
S
Shaohua Li 已提交
2422
	spin_unlock_irq(&sh->stripe_lock);
L
Linus Torvalds 已提交
2423 2424 2425
	return 0;
}

2426
static void end_reshape(struct r5conf *conf);
2427

2428
static void stripe_set_idx(sector_t stripe, struct r5conf *conf, int previous,
2429
			    struct stripe_head *sh)
2430
{
2431
	int sectors_per_chunk =
2432
		previous ? conf->prev_chunk_sectors : conf->chunk_sectors;
2433
	int dd_idx;
2434
	int chunk_offset = sector_div(stripe, sectors_per_chunk);
2435
	int disks = previous ? conf->previous_raid_disks : conf->raid_disks;
2436

2437 2438
	raid5_compute_sector(conf,
			     stripe * (disks - conf->max_degraded)
2439
			     *sectors_per_chunk + chunk_offset,
2440
			     previous,
2441
			     &dd_idx, sh);
2442 2443
}

2444
static void
2445
handle_failed_stripe(struct r5conf *conf, struct stripe_head *sh,
2446 2447 2448 2449 2450 2451 2452 2453 2454
				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)) {
2455
			struct md_rdev *rdev;
2456 2457 2458
			rcu_read_lock();
			rdev = rcu_dereference(conf->disks[i].rdev);
			if (rdev && test_bit(In_sync, &rdev->flags))
2459 2460 2461
				atomic_inc(&rdev->nr_pending);
			else
				rdev = NULL;
2462
			rcu_read_unlock();
2463 2464 2465 2466 2467 2468 2469 2470
			if (rdev) {
				if (!rdev_set_badblocks(
					    rdev,
					    sh->sector,
					    STRIPE_SECTORS, 0))
					md_error(conf->mddev, rdev);
				rdev_dec_pending(rdev, conf->mddev);
			}
2471
		}
S
Shaohua Li 已提交
2472
		spin_lock_irq(&sh->stripe_lock);
2473 2474 2475
		/* fail all writes first */
		bi = sh->dev[i].towrite;
		sh->dev[i].towrite = NULL;
S
Shaohua Li 已提交
2476
		spin_unlock_irq(&sh->stripe_lock);
2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488
		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);
2489
			if (!raid5_dec_bi_active_stripes(bi)) {
2490 2491 2492 2493 2494 2495
				md_write_end(conf->mddev);
				bi->bi_next = *return_bi;
				*return_bi = bi;
			}
			bi = nextbi;
		}
2496 2497 2498 2499
		if (bitmap_end)
			bitmap_endwrite(conf->mddev->bitmap, sh->sector,
				STRIPE_SECTORS, 0, 0);
		bitmap_end = 0;
2500 2501 2502 2503 2504 2505 2506 2507
		/* 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);
2508
			if (!raid5_dec_bi_active_stripes(bi)) {
2509 2510 2511 2512 2513 2514 2515
				md_write_end(conf->mddev);
				bi->bi_next = *return_bi;
				*return_bi = bi;
			}
			bi = bi2;
		}

2516 2517 2518 2519 2520 2521
		/* 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))) {
2522 2523 2524 2525 2526 2527 2528 2529 2530 2531
			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);
2532
				if (!raid5_dec_bi_active_stripes(bi)) {
2533 2534 2535 2536 2537 2538 2539 2540 2541
					bi->bi_next = *return_bi;
					*return_bi = bi;
				}
				bi = nextbi;
			}
		}
		if (bitmap_end)
			bitmap_endwrite(conf->mddev->bitmap, sh->sector,
					STRIPE_SECTORS, 0, 0);
2542 2543 2544 2545
		/* 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);
2546 2547
	}

2548 2549 2550
	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);
2551 2552
}

2553
static void
2554
handle_failed_sync(struct r5conf *conf, struct stripe_head *sh,
2555 2556 2557 2558 2559 2560 2561
		   struct stripe_head_state *s)
{
	int abort = 0;
	int i;

	clear_bit(STRIPE_SYNCING, &sh->state);
	s->syncing = 0;
2562
	s->replacing = 0;
2563
	/* There is nothing more to do for sync/check/repair.
2564 2565 2566
	 * 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.
2567
	 * For recover/replace we need to record a bad block on all
2568 2569
	 * non-sync devices, or abort the recovery
	 */
2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592
	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;
2593
	}
2594
	md_done_sync(conf->mddev, STRIPE_SECTORS, !abort);
2595 2596
}

2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612
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;
}

2613
/* fetch_block - checks the given member device to see if its data needs
2614 2615 2616
 * to be read or computed to satisfy a request.
 *
 * Returns 1 when no more member devices need to be checked, otherwise returns
2617
 * 0 to tell the loop in handle_stripe_fill to continue
2618
 */
2619 2620
static int fetch_block(struct stripe_head *sh, struct stripe_head_state *s,
		       int disk_idx, int disks)
2621
{
2622
	struct r5dev *dev = &sh->dev[disk_idx];
2623 2624
	struct r5dev *fdev[2] = { &sh->dev[s->failed_num[0]],
				  &sh->dev[s->failed_num[1]] };
2625

2626
	/* is the data in this block needed, and can we get it? */
2627 2628 2629 2630 2631
	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 ||
2632
	     (s->replacing && want_replace(sh, disk_idx)) ||
2633 2634
	     (s->failed >= 1 && fdev[0]->toread) ||
	     (s->failed >= 2 && fdev[1]->toread) ||
2635 2636 2637
	     (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))) {
2638 2639 2640 2641 2642 2643
		/* 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) &&
2644 2645
		    (s->failed && (disk_idx == s->failed_num[0] ||
				   disk_idx == s->failed_num[1]))) {
2646 2647
			/* have disk failed, and we're requested to fetch it;
			 * do compute it
2648
			 */
2649 2650 2651 2652 2653 2654 2655 2656
			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;
2657 2658 2659 2660 2661 2662
			/* 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.
			 */
2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675
			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;
2676
			}
2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695
			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);
2696 2697
		}
	}
2698 2699 2700 2701 2702

	return 0;
}

/**
2703
 * handle_stripe_fill - read or compute data to satisfy pending requests.
2704
 */
2705 2706 2707
static void handle_stripe_fill(struct stripe_head *sh,
			       struct stripe_head_state *s,
			       int disks)
2708 2709 2710 2711 2712 2713 2714 2715 2716 2717
{
	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--; )
2718
			if (fetch_block(sh, s, i, disks))
2719
				break;
2720 2721 2722 2723
	set_bit(STRIPE_HANDLE, &sh->state);
}


2724
/* handle_stripe_clean_event
2725 2726 2727 2728
 * 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.
 */
2729
static void handle_stripe_clean_event(struct r5conf *conf,
2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741
	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;
2742
				pr_debug("Return write for disc %d\n", i);
2743 2744 2745 2746 2747
				wbi = dev->written;
				dev->written = NULL;
				while (wbi && wbi->bi_sector <
					dev->sector + STRIPE_SECTORS) {
					wbi2 = r5_next_bio(wbi, dev->sector);
2748
					if (!raid5_dec_bi_active_stripes(wbi)) {
2749 2750 2751 2752 2753 2754
						md_write_end(conf->mddev);
						wbi->bi_next = *return_bi;
						*return_bi = wbi;
					}
					wbi = wbi2;
				}
2755 2756
				bitmap_endwrite(conf->mddev->bitmap, sh->sector,
						STRIPE_SECTORS,
2757
					 !test_bit(STRIPE_DEGRADED, &sh->state),
2758
						0);
2759 2760
			}
		}
2761 2762 2763 2764

	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);
2765 2766
}

2767
static void handle_stripe_dirtying(struct r5conf *conf,
2768 2769 2770
				   struct stripe_head *sh,
				   struct stripe_head_state *s,
				   int disks)
2771 2772
{
	int rmw = 0, rcw = 0, i;
2773 2774 2775 2776 2777 2778 2779
	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--; ) {
2780 2781 2782 2783
		/* 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) &&
2784 2785
		    !(test_bit(R5_UPTODATE, &dev->flags) ||
		      test_bit(R5_Wantcompute, &dev->flags))) {
2786 2787 2788 2789 2790 2791 2792 2793
			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) &&
2794 2795 2796
		    !(test_bit(R5_UPTODATE, &dev->flags) ||
		    test_bit(R5_Wantcompute, &dev->flags))) {
			if (test_bit(R5_Insync, &dev->flags)) rcw++;
2797 2798 2799 2800
			else
				rcw += 2*disks;
		}
	}
2801
	pr_debug("for sector %llu, rmw=%d rcw=%d\n",
2802 2803 2804 2805 2806 2807 2808 2809
		(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) &&
2810 2811
			    !(test_bit(R5_UPTODATE, &dev->flags) ||
			    test_bit(R5_Wantcompute, &dev->flags)) &&
2812 2813 2814
			    test_bit(R5_Insync, &dev->flags)) {
				if (
				  test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
2815
					pr_debug("Read_old block "
2816 2817 2818 2819 2820 2821 2822 2823 2824 2825
						"%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);
				}
			}
		}
2826
	if (rcw <= rmw && rcw > 0) {
2827
		/* want reconstruct write, but need to get some data */
2828
		rcw = 0;
2829 2830 2831
		for (i = disks; i--; ) {
			struct r5dev *dev = &sh->dev[i];
			if (!test_bit(R5_OVERWRITE, &dev->flags) &&
2832
			    i != sh->pd_idx && i != sh->qd_idx &&
2833
			    !test_bit(R5_LOCKED, &dev->flags) &&
2834
			    !(test_bit(R5_UPTODATE, &dev->flags) ||
2835 2836 2837 2838
			      test_bit(R5_Wantcompute, &dev->flags))) {
				rcw++;
				if (!test_bit(R5_Insync, &dev->flags))
					continue; /* it's a failed drive */
2839 2840
				if (
				  test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
2841
					pr_debug("Read_old block "
2842 2843 2844 2845 2846 2847 2848 2849 2850 2851
						"%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);
				}
			}
		}
2852
	}
2853 2854 2855
	/* now if nothing is locked, and if we have enough data,
	 * we can start a write request
	 */
2856 2857
	/* 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
2858 2859
	 * subsequent call wants to start a write request.  raid_run_ops only
	 * handles the case where compute block and reconstruct are requested
2860 2861 2862
	 * simultaneously.  If this is not the case then new writes need to be
	 * held off until the compute completes.
	 */
2863 2864 2865
	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)))
2866
		schedule_reconstruction(sh, s, rcw == 0, 0);
2867 2868
}

2869
static void handle_parity_checks5(struct r5conf *conf, struct stripe_head *sh,
2870 2871
				struct stripe_head_state *s, int disks)
{
2872
	struct r5dev *dev = NULL;
2873

2874
	set_bit(STRIPE_HANDLE, &sh->state);
2875

2876 2877 2878
	switch (sh->check_state) {
	case check_state_idle:
		/* start a new check operation if there are no failures */
2879 2880
		if (s->failed == 0) {
			BUG_ON(s->uptodate != disks);
2881 2882
			sh->check_state = check_state_run;
			set_bit(STRIPE_OP_CHECK, &s->ops_request);
2883 2884
			clear_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags);
			s->uptodate--;
2885
			break;
2886
		}
2887
		dev = &sh->dev[s->failed_num[0]];
2888 2889 2890 2891 2892 2893 2894 2895 2896
		/* 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
2897

2898 2899 2900 2901 2902
		/* 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);
2903
		s->locked++;
2904
		set_bit(R5_Wantwrite, &dev->flags);
2905

2906 2907
		clear_bit(STRIPE_DEGRADED, &sh->state);
		set_bit(STRIPE_INSYNC, &sh->state);
2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923
		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
2924
		if ((sh->ops.zero_sum_result & SUM_CHECK_P_RESULT) == 0)
2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935
			/* 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;
2936
				set_bit(STRIPE_COMPUTE_RUN, &sh->state);
2937 2938 2939 2940
				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;
2941
				sh->ops.target2 = -1;
2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952
				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();
2953 2954 2955 2956
	}
}


2957
static void handle_parity_checks6(struct r5conf *conf, struct stripe_head *sh,
2958
				  struct stripe_head_state *s,
2959
				  int disks)
2960 2961
{
	int pd_idx = sh->pd_idx;
N
NeilBrown 已提交
2962
	int qd_idx = sh->qd_idx;
2963
	struct r5dev *dev;
2964 2965 2966 2967

	set_bit(STRIPE_HANDLE, &sh->state);

	BUG_ON(s->failed > 2);
2968

2969 2970 2971 2972 2973 2974
	/* 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
	 */

2975 2976 2977
	switch (sh->check_state) {
	case check_state_idle:
		/* start a new check operation if there are < 2 failures */
2978
		if (s->failed == s->q_failed) {
2979
			/* The only possible failed device holds Q, so it
2980 2981 2982
			 * makes sense to check P (If anything else were failed,
			 * we would have used P to recreate it).
			 */
2983
			sh->check_state = check_state_run;
2984
		}
2985
		if (!s->q_failed && s->failed < 2) {
2986
			/* Q is not failed, and we didn't use it to generate
2987 2988
			 * anything, so it makes sense to check it
			 */
2989 2990 2991 2992
			if (sh->check_state == check_state_run)
				sh->check_state = check_state_run_pq;
			else
				sh->check_state = check_state_run_q;
2993 2994
		}

2995 2996
		/* discard potentially stale zero_sum_result */
		sh->ops.zero_sum_result = 0;
2997

2998 2999 3000 3001
		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--;
3002
		}
3003 3004 3005 3006 3007 3008 3009
		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;
3010 3011
		}

3012 3013 3014 3015 3016
		/* we have 2-disk failure */
		BUG_ON(s->failed != 2);
		/* fall through */
	case check_state_compute_result:
		sh->check_state = check_state_idle;
3017

3018 3019 3020
		/* check that a write has not made the stripe insync */
		if (test_bit(STRIPE_INSYNC, &sh->state))
			break;
3021 3022

		/* now write out any block on a failed drive,
3023
		 * or P or Q if they were recomputed
3024
		 */
3025
		BUG_ON(s->uptodate < disks - 1); /* We don't need Q to recover */
3026
		if (s->failed == 2) {
3027
			dev = &sh->dev[s->failed_num[1]];
3028 3029 3030 3031 3032
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
		if (s->failed >= 1) {
3033
			dev = &sh->dev[s->failed_num[0]];
3034 3035 3036 3037
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
3038
		if (sh->ops.zero_sum_result & SUM_CHECK_P_RESULT) {
3039 3040 3041 3042 3043
			dev = &sh->dev[pd_idx];
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
3044
		if (sh->ops.zero_sum_result & SUM_CHECK_Q_RESULT) {
3045 3046 3047 3048 3049 3050 3051 3052
			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);
3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 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
		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();
3117 3118 3119
	}
}

3120
static void handle_stripe_expansion(struct r5conf *conf, struct stripe_head *sh)
3121 3122 3123 3124 3125 3126
{
	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.
	 */
3127
	struct dma_async_tx_descriptor *tx = NULL;
3128 3129
	clear_bit(STRIPE_EXPAND_SOURCE, &sh->state);
	for (i = 0; i < sh->disks; i++)
N
NeilBrown 已提交
3130
		if (i != sh->pd_idx && i != sh->qd_idx) {
3131
			int dd_idx, j;
3132
			struct stripe_head *sh2;
3133
			struct async_submit_ctl submit;
3134

3135
			sector_t bn = compute_blocknr(sh, i, 1);
3136 3137
			sector_t s = raid5_compute_sector(conf, bn, 0,
							  &dd_idx, NULL);
3138
			sh2 = get_active_stripe(conf, s, 0, 1, 1);
3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150
			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;
			}
3151 3152

			/* place all the copies on one channel */
3153
			init_async_submit(&submit, 0, tx, NULL, NULL, NULL);
3154
			tx = async_memcpy(sh2->dev[dd_idx].page,
3155
					  sh->dev[i].page, 0, 0, STRIPE_SIZE,
3156
					  &submit);
3157

3158 3159 3160 3161
			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 &&
3162
				    j != sh2->qd_idx &&
3163 3164 3165 3166 3167 3168 3169
				    !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);
3170

3171
		}
3172 3173 3174 3175 3176
	/* done submitting copies, wait for them to complete */
	if (tx) {
		async_tx_ack(tx);
		dma_wait_for_async_tx(tx);
	}
3177
}
L
Linus Torvalds 已提交
3178 3179 3180 3181

/*
 * handle_stripe - do things to a stripe.
 *
3182 3183
 * 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
3184
 * Possible results:
3185 3186
 *    return some read requests which now have data
 *    return some write requests which are safely on storage
L
Linus Torvalds 已提交
3187 3188 3189 3190 3191
 *    schedule a read on some buffers
 *    schedule a write of some buffers
 *    return confirmation of parity correctness
 *
 */
3192

3193
static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
L
Linus Torvalds 已提交
3194
{
3195
	struct r5conf *conf = sh->raid_conf;
3196
	int disks = sh->disks;
3197 3198
	struct r5dev *dev;
	int i;
3199
	int do_recovery = 0;
L
Linus Torvalds 已提交
3200

3201 3202 3203 3204 3205 3206
	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
3207

3208
	/* Now to look around and see what can be done */
L
Linus Torvalds 已提交
3209
	rcu_read_lock();
3210
	for (i=disks; i--; ) {
3211
		struct md_rdev *rdev;
3212 3213 3214
		sector_t first_bad;
		int bad_sectors;
		int is_bad = 0;
3215

3216
		dev = &sh->dev[i];
L
Linus Torvalds 已提交
3217

3218
		pr_debug("check %d: state 0x%lx read %p write %p written %p\n",
3219 3220
			 i, dev->flags,
			 dev->toread, dev->towrite, dev->written);
3221 3222 3223 3224 3225 3226 3227 3228
		/* 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
3229

3230
		/* now count some things */
3231 3232 3233 3234
		if (test_bit(R5_LOCKED, &dev->flags))
			s->locked++;
		if (test_bit(R5_UPTODATE, &dev->flags))
			s->uptodate++;
3235
		if (test_bit(R5_Wantcompute, &dev->flags)) {
3236 3237
			s->compute++;
			BUG_ON(s->compute > 2);
3238
		}
L
Linus Torvalds 已提交
3239

3240
		if (test_bit(R5_Wantfill, &dev->flags))
3241
			s->to_fill++;
3242
		else if (dev->toread)
3243
			s->to_read++;
3244
		if (dev->towrite) {
3245
			s->to_write++;
3246
			if (!test_bit(R5_OVERWRITE, &dev->flags))
3247
				s->non_overwrite++;
3248
		}
3249
		if (dev->written)
3250
			s->written++;
3251 3252 3253 3254 3255 3256 3257 3258 3259 3260
		/* 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 {
3261 3262
			if (rdev)
				set_bit(R5_NeedReplace, &dev->flags);
3263 3264 3265
			rdev = rcu_dereference(conf->disks[i].rdev);
			clear_bit(R5_ReadRepl, &dev->flags);
		}
3266 3267
		if (rdev && test_bit(Faulty, &rdev->flags))
			rdev = NULL;
3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279
		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);
			}
3280
		}
3281 3282 3283
		clear_bit(R5_Insync, &dev->flags);
		if (!rdev)
			/* Not in-sync */;
3284 3285
		else if (is_bad) {
			/* also not in-sync */
3286 3287
			if (!test_bit(WriteErrorSeen, &rdev->flags) &&
			    test_bit(R5_UPTODATE, &dev->flags)) {
3288 3289 3290 3291 3292 3293 3294
				/* 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))
3295
			set_bit(R5_Insync, &dev->flags);
3296
		else if (sh->sector + STRIPE_SECTORS <= rdev->recovery_offset)
3297
			/* in sync if before recovery_offset */
3298 3299 3300 3301 3302 3303 3304 3305 3306
			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
3307
		if (rdev && test_bit(R5_WriteError, &dev->flags)) {
3308 3309 3310 3311 3312 3313 3314
			/* 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)) {
3315
				s->handle_bad_blocks = 1;
3316
				atomic_inc(&rdev2->nr_pending);
3317 3318 3319
			} else
				clear_bit(R5_WriteError, &dev->flags);
		}
A
Adam Kwolek 已提交
3320
		if (rdev && test_bit(R5_MadeGood, &dev->flags)) {
3321 3322 3323 3324 3325
			/* 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)) {
3326
				s->handle_bad_blocks = 1;
3327
				atomic_inc(&rdev2->nr_pending);
3328 3329 3330
			} else
				clear_bit(R5_MadeGood, &dev->flags);
		}
3331 3332 3333 3334 3335 3336 3337 3338 3339
		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);
		}
3340
		if (!test_bit(R5_Insync, &dev->flags)) {
3341 3342 3343
			/* 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
3344
		}
3345 3346 3347
		if (test_bit(R5_ReadError, &dev->flags))
			clear_bit(R5_Insync, &dev->flags);
		if (!test_bit(R5_Insync, &dev->flags)) {
3348 3349 3350
			if (s->failed < 2)
				s->failed_num[s->failed] = i;
			s->failed++;
3351 3352
			if (rdev && !test_bit(Faulty, &rdev->flags))
				do_recovery = 1;
3353
		}
L
Linus Torvalds 已提交
3354
	}
3355 3356 3357 3358
	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
3359
		 * else if MD_RECOVERY_REQUESTED is set, we also are syncing.
3360 3361 3362 3363 3364
		 * else we can only be replacing
		 * sync and recovery both need to read all devices, and so
		 * use the same flag.
		 */
		if (do_recovery ||
3365 3366
		    sh->sector >= conf->mddev->recovery_cp ||
		    test_bit(MD_RECOVERY_REQUESTED, &(conf->mddev->recovery)))
3367 3368 3369 3370
			s->syncing = 1;
		else
			s->replacing = 1;
	}
L
Linus Torvalds 已提交
3371
	rcu_read_unlock();
3372 3373 3374 3375 3376
}

static void handle_stripe(struct stripe_head *sh)
{
	struct stripe_head_state s;
3377
	struct r5conf *conf = sh->raid_conf;
3378
	int i;
3379 3380
	int prexor;
	int disks = sh->disks;
3381
	struct r5dev *pdev, *qdev;
3382 3383

	clear_bit(STRIPE_HANDLE, &sh->state);
3384
	if (test_and_set_bit_lock(STRIPE_ACTIVE, &sh->state)) {
3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401
		/* 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);
3402

3403
	analyse_stripe(sh, &s);
3404

3405 3406 3407 3408 3409
	if (s.handle_bad_blocks) {
		set_bit(STRIPE_HANDLE, &sh->state);
		goto finish;
	}

3410 3411
	if (unlikely(s.blocked_rdev)) {
		if (s.syncing || s.expanding || s.expanded ||
3412
		    s.replacing || s.to_write || s.written) {
3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432
			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.
	 */
3433 3434 3435 3436 3437
	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);
3438
		if (s.syncing + s.replacing)
3439 3440
			handle_failed_sync(conf, sh, &s);
	}
3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468

	/*
	 * 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)
3469 3470 3471
	    || (s.syncing && (s.uptodate + s.compute < disks))
	    || s.replacing
	    || s.expanding)
3472 3473
		handle_stripe_fill(sh, &s, disks);

3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531
	/* 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);
	}
3532

3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546
	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)) {
3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575
		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++;
				}
			}
		}


3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602
	/* 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++;
		}
	}
3603

3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619
	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);
3620

3621
finish:
3622
	/* wait for this device to become unblocked */
3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634
	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);
	}
3635

3636 3637
	if (s.handle_bad_blocks)
		for (i = disks; i--; ) {
3638
			struct md_rdev *rdev;
3639 3640 3641 3642 3643 3644 3645 3646 3647
			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);
			}
3648 3649 3650
			if (test_and_clear_bit(R5_MadeGood, &dev->flags)) {
				rdev = conf->disks[i].rdev;
				rdev_clear_badblocks(rdev, sh->sector,
3651
						     STRIPE_SECTORS, 0);
3652 3653
				rdev_dec_pending(rdev, conf->mddev);
			}
3654 3655
			if (test_and_clear_bit(R5_MadeGoodRepl, &dev->flags)) {
				rdev = conf->disks[i].replacement;
3656 3657 3658
				if (!rdev)
					/* rdev have been moved down */
					rdev = conf->disks[i].rdev;
3659
				rdev_clear_badblocks(rdev, sh->sector,
3660
						     STRIPE_SECTORS, 0);
3661 3662
				rdev_dec_pending(rdev, conf->mddev);
			}
3663 3664
		}

3665 3666 3667
	if (s.ops_request)
		raid_run_ops(sh, s.ops_request);

D
Dan Williams 已提交
3668
	ops_run_io(sh, &s);
3669

3670
	if (s.dec_preread_active) {
3671
		/* We delay this until after ops_run_io so that if make_request
T
Tejun Heo 已提交
3672
		 * is waiting on a flush, it won't continue until the writes
3673 3674 3675 3676 3677 3678 3679 3680
		 * 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);
	}

3681
	return_io(s.return_bi);
3682

3683
	clear_bit_unlock(STRIPE_ACTIVE, &sh->state);
3684 3685
}

3686
static void raid5_activate_delayed(struct r5conf *conf)
3687 3688 3689 3690 3691 3692 3693 3694 3695 3696
{
	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);
3697
			list_add_tail(&sh->lru, &conf->hold_list);
3698
		}
N
NeilBrown 已提交
3699
	}
3700 3701
}

3702
static void activate_bit_delay(struct r5conf *conf)
3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715
{
	/* 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);
	}
}

3716
int md_raid5_congested(struct mddev *mddev, int bits)
3717
{
3718
	struct r5conf *conf = mddev->private;
3719 3720 3721 3722

	/* No difference between reads and writes.  Just check
	 * how busy the stripe_cache is
	 */
3723

3724 3725 3726 3727 3728 3729 3730 3731 3732
	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
3733 3734 3735 3736
EXPORT_SYMBOL_GPL(md_raid5_congested);

static int raid5_congested(void *data, int bits)
{
3737
	struct mddev *mddev = data;
N
NeilBrown 已提交
3738 3739 3740 3741

	return mddev_congested(mddev, bits) ||
		md_raid5_congested(mddev, bits);
}
3742

3743 3744 3745
/* We want read requests to align with chunks where possible,
 * but write requests don't need to.
 */
3746 3747 3748
static int raid5_mergeable_bvec(struct request_queue *q,
				struct bvec_merge_data *bvm,
				struct bio_vec *biovec)
3749
{
3750
	struct mddev *mddev = q->queuedata;
3751
	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
3752
	int max;
3753
	unsigned int chunk_sectors = mddev->chunk_sectors;
3754
	unsigned int bio_sectors = bvm->bi_size >> 9;
3755

3756
	if ((bvm->bi_rw & 1) == WRITE)
3757 3758
		return biovec->bv_len; /* always allow writes to be mergeable */

3759 3760
	if (mddev->new_chunk_sectors < mddev->chunk_sectors)
		chunk_sectors = mddev->new_chunk_sectors;
3761 3762 3763 3764 3765 3766 3767 3768
	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;
}

3769

3770
static int in_chunk_boundary(struct mddev *mddev, struct bio *bio)
3771 3772
{
	sector_t sector = bio->bi_sector + get_start_sect(bio->bi_bdev);
3773
	unsigned int chunk_sectors = mddev->chunk_sectors;
3774 3775
	unsigned int bio_sectors = bio->bi_size >> 9;

3776 3777
	if (mddev->new_chunk_sectors < mddev->chunk_sectors)
		chunk_sectors = mddev->new_chunk_sectors;
3778 3779 3780 3781
	return  chunk_sectors >=
		((sector & (chunk_sectors - 1)) + bio_sectors);
}

3782 3783 3784 3785
/*
 *  add bio to the retry LIFO  ( in O(1) ... we are in interrupt )
 *  later sampled by raid5d.
 */
3786
static void add_bio_to_retry(struct bio *bi,struct r5conf *conf)
3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799
{
	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);
}


3800
static struct bio *remove_bio_from_retry(struct r5conf *conf)
3801 3802 3803 3804 3805 3806 3807 3808 3809 3810
{
	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) {
3811
		conf->retry_read_aligned_list = bi->bi_next;
3812
		bi->bi_next = NULL;
3813 3814 3815 3816
		/*
		 * this sets the active strip count to 1 and the processed
		 * strip count to zero (upper 8 bits)
		 */
3817
		raid5_set_bi_stripes(bi, 1); /* biased count of active stripes */
3818 3819 3820 3821 3822 3823
	}

	return bi;
}


3824 3825 3826 3827 3828 3829
/*
 *  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..
 */
3830
static void raid5_align_endio(struct bio *bi, int error)
3831 3832
{
	struct bio* raid_bi  = bi->bi_private;
3833
	struct mddev *mddev;
3834
	struct r5conf *conf;
3835
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
3836
	struct md_rdev *rdev;
3837

3838
	bio_put(bi);
3839 3840 3841

	rdev = (void*)raid_bi->bi_next;
	raid_bi->bi_next = NULL;
3842 3843
	mddev = rdev->mddev;
	conf = mddev->private;
3844 3845 3846 3847

	rdev_dec_pending(rdev, conf->mddev);

	if (!error && uptodate) {
3848
		bio_endio(raid_bi, 0);
3849 3850
		if (atomic_dec_and_test(&conf->active_aligned_reads))
			wake_up(&conf->wait_for_stripe);
3851
		return;
3852 3853 3854
	}


3855
	pr_debug("raid5_align_endio : io error...handing IO for a retry\n");
3856 3857

	add_bio_to_retry(raid_bi, conf);
3858 3859
}

3860 3861
static int bio_fits_rdev(struct bio *bi)
{
3862
	struct request_queue *q = bdev_get_queue(bi->bi_bdev);
3863

3864
	if ((bi->bi_size>>9) > queue_max_sectors(q))
3865 3866
		return 0;
	blk_recount_segments(q, bi);
3867
	if (bi->bi_phys_segments > queue_max_segments(q))
3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879
		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;
}


3880
static int chunk_aligned_read(struct mddev *mddev, struct bio * raid_bio)
3881
{
3882
	struct r5conf *conf = mddev->private;
N
NeilBrown 已提交
3883
	int dd_idx;
3884
	struct bio* align_bi;
3885
	struct md_rdev *rdev;
3886
	sector_t end_sector;
3887 3888

	if (!in_chunk_boundary(mddev, raid_bio)) {
3889
		pr_debug("chunk_aligned_read : non aligned\n");
3890 3891 3892
		return 0;
	}
	/*
3893
	 * use bio_clone_mddev to make a copy of the bio
3894
	 */
3895
	align_bi = bio_clone_mddev(raid_bio, GFP_NOIO, mddev);
3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906
	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
	 */
3907 3908
	align_bi->bi_sector =  raid5_compute_sector(conf, raid_bio->bi_sector,
						    0,
3909
						    &dd_idx, NULL);
3910

3911
	end_sector = align_bi->bi_sector + (align_bi->bi_size >> 9);
3912
	rcu_read_lock();
3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923
	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) {
3924 3925 3926
		sector_t first_bad;
		int bad_sectors;

3927 3928
		atomic_inc(&rdev->nr_pending);
		rcu_read_unlock();
3929 3930 3931 3932
		raid_bio->bi_next = (void*)rdev;
		align_bi->bi_bdev =  rdev->bdev;
		align_bi->bi_flags &= ~(1 << BIO_SEG_VALID);

3933 3934 3935 3936
		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 */
3937 3938 3939 3940 3941
			bio_put(align_bi);
			rdev_dec_pending(rdev, mddev);
			return 0;
		}

3942 3943 3944
		/* No reshape active, so we can trust rdev->data_offset */
		align_bi->bi_sector += rdev->data_offset;

3945 3946 3947 3948 3949 3950 3951
		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);

3952 3953 3954 3955
		generic_make_request(align_bi);
		return 1;
	} else {
		rcu_read_unlock();
3956
		bio_put(align_bi);
3957 3958 3959 3960
		return 0;
	}
}

3961 3962 3963 3964 3965 3966 3967 3968 3969 3970
/* __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.
 */
3971
static struct stripe_head *__get_priority_stripe(struct r5conf *conf)
3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012
{
	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;
}
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 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069
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);
}

4070
static void make_request(struct mddev *mddev, struct bio * bi)
L
Linus Torvalds 已提交
4071
{
4072
	struct r5conf *conf = mddev->private;
4073
	int dd_idx;
L
Linus Torvalds 已提交
4074 4075 4076
	sector_t new_sector;
	sector_t logical_sector, last_sector;
	struct stripe_head *sh;
4077
	const int rw = bio_data_dir(bi);
4078
	int remaining;
L
Linus Torvalds 已提交
4079

T
Tejun Heo 已提交
4080 4081
	if (unlikely(bi->bi_rw & REQ_FLUSH)) {
		md_flush_request(mddev, bi);
4082
		return;
4083 4084
	}

4085
	md_write_start(mddev, bi);
4086

4087
	if (rw == READ &&
4088
	     mddev->reshape_position == MaxSector &&
4089
	     chunk_aligned_read(mddev,bi))
4090
		return;
4091

L
Linus Torvalds 已提交
4092 4093 4094 4095
	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 */
4096

L
Linus Torvalds 已提交
4097 4098
	for (;logical_sector < last_sector; logical_sector += STRIPE_SECTORS) {
		DEFINE_WAIT(w);
4099
		int previous;
4100

4101
	retry:
4102
		previous = 0;
4103
		prepare_to_wait(&conf->wait_for_overlap, &w, TASK_UNINTERRUPTIBLE);
4104
		if (unlikely(conf->reshape_progress != MaxSector)) {
4105
			/* spinlock is needed as reshape_progress may be
4106 4107
			 * 64bit on a 32bit platform, and so it might be
			 * possible to see a half-updated value
4108
			 * Of course reshape_progress could change after
4109 4110 4111 4112
			 * the lock is dropped, so once we get a reference
			 * to the stripe that we think it is, we will have
			 * to check again.
			 */
4113
			spin_lock_irq(&conf->device_lock);
4114
			if (mddev->reshape_backwards
4115 4116
			    ? logical_sector < conf->reshape_progress
			    : logical_sector >= conf->reshape_progress) {
4117 4118
				previous = 1;
			} else {
4119
				if (mddev->reshape_backwards
4120 4121
				    ? logical_sector < conf->reshape_safe
				    : logical_sector >= conf->reshape_safe) {
4122 4123 4124 4125 4126
					spin_unlock_irq(&conf->device_lock);
					schedule();
					goto retry;
				}
			}
4127 4128
			spin_unlock_irq(&conf->device_lock);
		}
4129

4130 4131
		new_sector = raid5_compute_sector(conf, logical_sector,
						  previous,
4132
						  &dd_idx, NULL);
4133
		pr_debug("raid456: make_request, sector %llu logical %llu\n",
L
Linus Torvalds 已提交
4134 4135 4136
			(unsigned long long)new_sector, 
			(unsigned long long)logical_sector);

4137
		sh = get_active_stripe(conf, new_sector, previous,
4138
				       (bi->bi_rw&RWA_MASK), 0);
L
Linus Torvalds 已提交
4139
		if (sh) {
4140
			if (unlikely(previous)) {
4141
				/* expansion might have moved on while waiting for a
4142 4143 4144 4145 4146 4147
				 * 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.
4148 4149 4150
				 */
				int must_retry = 0;
				spin_lock_irq(&conf->device_lock);
4151
				if (mddev->reshape_backwards
4152 4153
				    ? logical_sector >= conf->reshape_progress
				    : logical_sector < conf->reshape_progress)
4154 4155 4156 4157 4158
					/* mismatch, need to try again */
					must_retry = 1;
				spin_unlock_irq(&conf->device_lock);
				if (must_retry) {
					release_stripe(sh);
4159
					schedule();
4160 4161 4162
					goto retry;
				}
			}
4163

4164
			if (rw == WRITE &&
4165
			    logical_sector >= mddev->suspend_lo &&
4166 4167
			    logical_sector < mddev->suspend_hi) {
				release_stripe(sh);
4168 4169 4170 4171 4172 4173 4174 4175 4176 4177
				/* 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();
4178 4179
				goto retry;
			}
4180 4181

			if (test_bit(STRIPE_EXPANDING, &sh->state) ||
4182
			    !add_stripe_bio(sh, bi, dd_idx, rw)) {
4183 4184
				/* Stripe is busy expanding or
				 * add failed due to overlap.  Flush everything
L
Linus Torvalds 已提交
4185 4186
				 * and wait a while
				 */
N
NeilBrown 已提交
4187
				md_wakeup_thread(mddev->thread);
L
Linus Torvalds 已提交
4188 4189 4190 4191 4192
				release_stripe(sh);
				schedule();
				goto retry;
			}
			finish_wait(&conf->wait_for_overlap, &w);
4193 4194
			set_bit(STRIPE_HANDLE, &sh->state);
			clear_bit(STRIPE_DELAYED, &sh->state);
4195
			if ((bi->bi_rw & REQ_NOIDLE) &&
4196 4197
			    !test_and_set_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
				atomic_inc(&conf->preread_active_stripes);
4198
			release_stripe_plug(mddev, sh);
L
Linus Torvalds 已提交
4199 4200 4201 4202 4203 4204 4205
		} 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;
		}
	}
4206

4207
	remaining = raid5_dec_bi_active_stripes(bi);
4208
	if (remaining == 0) {
L
Linus Torvalds 已提交
4209

4210
		if ( rw == WRITE )
L
Linus Torvalds 已提交
4211
			md_write_end(mddev);
4212

4213
		bio_endio(bi, 0);
L
Linus Torvalds 已提交
4214 4215 4216
	}
}

4217
static sector_t raid5_size(struct mddev *mddev, sector_t sectors, int raid_disks);
D
Dan Williams 已提交
4218

4219
static sector_t reshape_request(struct mddev *mddev, sector_t sector_nr, int *skipped)
L
Linus Torvalds 已提交
4220
{
4221 4222 4223 4224 4225 4226 4227 4228 4229
	/* 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.
	 */
4230
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
4231
	struct stripe_head *sh;
4232
	sector_t first_sector, last_sector;
4233 4234 4235
	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;
4236 4237
	int i;
	int dd_idx;
4238
	sector_t writepos, readpos, safepos;
4239
	sector_t stripe_addr;
4240
	int reshape_sectors;
4241
	struct list_head stripes;
4242

4243 4244
	if (sector_nr == 0) {
		/* If restarting in the middle, skip the initial sectors */
4245
		if (mddev->reshape_backwards &&
4246 4247 4248
		    conf->reshape_progress < raid5_size(mddev, 0, 0)) {
			sector_nr = raid5_size(mddev, 0, 0)
				- conf->reshape_progress;
4249
		} else if (!mddev->reshape_backwards &&
4250 4251
			   conf->reshape_progress > 0)
			sector_nr = conf->reshape_progress;
4252
		sector_div(sector_nr, new_data_disks);
4253
		if (sector_nr) {
4254 4255
			mddev->curr_resync_completed = sector_nr;
			sysfs_notify(&mddev->kobj, NULL, "sync_completed");
4256 4257 4258
			*skipped = 1;
			return sector_nr;
		}
4259 4260
	}

4261 4262 4263 4264
	/* 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
	 */
4265 4266
	if (mddev->new_chunk_sectors > mddev->chunk_sectors)
		reshape_sectors = mddev->new_chunk_sectors;
4267
	else
4268
		reshape_sectors = mddev->chunk_sectors;
4269

4270 4271 4272 4273 4274
	/* 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
4275
	 */
4276
	writepos = conf->reshape_progress;
4277
	sector_div(writepos, new_data_disks);
4278 4279
	readpos = conf->reshape_progress;
	sector_div(readpos, data_disks);
4280
	safepos = conf->reshape_safe;
4281
	sector_div(safepos, data_disks);
4282
	if (mddev->reshape_backwards) {
4283
		writepos -= min_t(sector_t, reshape_sectors, writepos);
4284
		readpos += reshape_sectors;
4285
		safepos += reshape_sectors;
4286
	} else {
4287
		writepos += reshape_sectors;
4288 4289
		readpos -= min_t(sector_t, reshape_sectors, readpos);
		safepos -= min_t(sector_t, reshape_sectors, safepos);
4290
	}
4291

4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306
	/* 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;
	}

4307 4308 4309 4310
	/* '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.
4311 4312 4313 4314
	 * 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
4315 4316 4317 4318 4319 4320 4321 4322 4323 4324 4325 4326
	 * 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???
	 */
4327 4328 4329 4330 4331 4332
	if (conf->min_offset_diff < 0) {
		safepos += -conf->min_offset_diff;
		readpos += -conf->min_offset_diff;
	} else
		writepos += conf->min_offset_diff;

4333
	if ((mddev->reshape_backwards
4334 4335 4336
	     ? (safepos > writepos && readpos < writepos)
	     : (safepos < writepos && readpos > writepos)) ||
	    time_after(jiffies, conf->reshape_checkpoint + 10*HZ)) {
4337 4338 4339
		/* Cannot proceed until we've updated the superblock... */
		wait_event(conf->wait_for_overlap,
			   atomic_read(&conf->reshape_stripes)==0);
4340
		mddev->reshape_position = conf->reshape_progress;
4341
		mddev->curr_resync_completed = sector_nr;
4342
		conf->reshape_checkpoint = jiffies;
4343
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
4344
		md_wakeup_thread(mddev->thread);
4345
		wait_event(mddev->sb_wait, mddev->flags == 0 ||
4346 4347
			   kthread_should_stop());
		spin_lock_irq(&conf->device_lock);
4348
		conf->reshape_safe = mddev->reshape_position;
4349 4350
		spin_unlock_irq(&conf->device_lock);
		wake_up(&conf->wait_for_overlap);
4351
		sysfs_notify(&mddev->kobj, NULL, "sync_completed");
4352 4353
	}

4354
	INIT_LIST_HEAD(&stripes);
4355
	for (i = 0; i < reshape_sectors; i += STRIPE_SECTORS) {
4356
		int j;
4357
		int skipped_disk = 0;
4358
		sh = get_active_stripe(conf, stripe_addr+i, 0, 0, 1);
4359 4360 4361 4362 4363 4364 4365 4366 4367
		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;
4368
			if (conf->level == 6 &&
4369
			    j == sh->qd_idx)
4370
				continue;
4371
			s = compute_blocknr(sh, j, 0);
D
Dan Williams 已提交
4372
			if (s < raid5_size(mddev, 0, 0)) {
4373
				skipped_disk = 1;
4374 4375 4376 4377 4378 4379
				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);
		}
4380
		if (!skipped_disk) {
4381 4382 4383
			set_bit(STRIPE_EXPAND_READY, &sh->state);
			set_bit(STRIPE_HANDLE, &sh->state);
		}
4384
		list_add(&sh->lru, &stripes);
4385 4386
	}
	spin_lock_irq(&conf->device_lock);
4387
	if (mddev->reshape_backwards)
4388
		conf->reshape_progress -= reshape_sectors * new_data_disks;
4389
	else
4390
		conf->reshape_progress += reshape_sectors * new_data_disks;
4391 4392 4393 4394 4395 4396 4397
	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 =
4398
		raid5_compute_sector(conf, stripe_addr*(new_data_disks),
4399
				     1, &dd_idx, NULL);
4400
	last_sector =
4401
		raid5_compute_sector(conf, ((stripe_addr+reshape_sectors)
4402
					    * new_data_disks - 1),
4403
				     1, &dd_idx, NULL);
A
Andre Noll 已提交
4404 4405
	if (last_sector >= mddev->dev_sectors)
		last_sector = mddev->dev_sectors - 1;
4406
	while (first_sector <= last_sector) {
4407
		sh = get_active_stripe(conf, first_sector, 1, 0, 1);
4408 4409 4410 4411 4412
		set_bit(STRIPE_EXPAND_SOURCE, &sh->state);
		set_bit(STRIPE_HANDLE, &sh->state);
		release_stripe(sh);
		first_sector += STRIPE_SECTORS;
	}
4413 4414 4415 4416 4417 4418 4419 4420
	/* 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);
	}
4421 4422 4423
	/* If this takes us to the resync_max point where we have to pause,
	 * then we need to write out the superblock.
	 */
4424
	sector_nr += reshape_sectors;
4425 4426
	if ((sector_nr - mddev->curr_resync_completed) * 2
	    >= mddev->resync_max - mddev->curr_resync_completed) {
4427 4428 4429
		/* Cannot proceed until we've updated the superblock... */
		wait_event(conf->wait_for_overlap,
			   atomic_read(&conf->reshape_stripes) == 0);
4430
		mddev->reshape_position = conf->reshape_progress;
4431
		mddev->curr_resync_completed = sector_nr;
4432
		conf->reshape_checkpoint = jiffies;
4433 4434 4435 4436 4437 4438
		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);
4439
		conf->reshape_safe = mddev->reshape_position;
4440 4441
		spin_unlock_irq(&conf->device_lock);
		wake_up(&conf->wait_for_overlap);
4442
		sysfs_notify(&mddev->kobj, NULL, "sync_completed");
4443
	}
4444
	return reshape_sectors;
4445 4446 4447
}

/* FIXME go_faster isn't used */
4448
static inline sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipped, int go_faster)
4449
{
4450
	struct r5conf *conf = mddev->private;
4451
	struct stripe_head *sh;
A
Andre Noll 已提交
4452
	sector_t max_sector = mddev->dev_sectors;
N
NeilBrown 已提交
4453
	sector_t sync_blocks;
4454 4455
	int still_degraded = 0;
	int i;
L
Linus Torvalds 已提交
4456

4457
	if (sector_nr >= max_sector) {
L
Linus Torvalds 已提交
4458
		/* just being told to finish up .. nothing much to do */
4459

4460 4461 4462 4463
		if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) {
			end_reshape(conf);
			return 0;
		}
4464 4465 4466 4467

		if (mddev->curr_resync < max_sector) /* aborted */
			bitmap_end_sync(mddev->bitmap, mddev->curr_resync,
					&sync_blocks, 1);
4468
		else /* completed sync */
4469 4470 4471
			conf->fullsync = 0;
		bitmap_close_sync(mddev->bitmap);

L
Linus Torvalds 已提交
4472 4473
		return 0;
	}
4474

4475 4476 4477
	/* Allow raid5_quiesce to complete */
	wait_event(conf->wait_for_overlap, conf->quiesce != 2);

4478 4479
	if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
		return reshape_request(mddev, sector_nr, skipped);
4480

4481 4482 4483 4484 4485 4486
	/* 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
	 */

4487
	/* if there is too many failed drives and we are trying
L
Linus Torvalds 已提交
4488 4489 4490
	 * to resync, then assert that we are finished, because there is
	 * nothing we can do.
	 */
4491
	if (mddev->degraded >= conf->max_degraded &&
4492
	    test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
A
Andre Noll 已提交
4493
		sector_t rv = mddev->dev_sectors - sector_nr;
4494
		*skipped = 1;
L
Linus Torvalds 已提交
4495 4496
		return rv;
	}
4497
	if (!bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, 1) &&
4498
	    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
4499 4500 4501 4502 4503 4504
	    !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
4505

N
NeilBrown 已提交
4506 4507
	bitmap_cond_end_sync(mddev->bitmap, sector_nr);

4508
	sh = get_active_stripe(conf, sector_nr, 0, 1, 0);
L
Linus Torvalds 已提交
4509
	if (sh == NULL) {
4510
		sh = get_active_stripe(conf, sector_nr, 0, 0, 0);
L
Linus Torvalds 已提交
4511
		/* make sure we don't swamp the stripe cache if someone else
4512
		 * is trying to get access
L
Linus Torvalds 已提交
4513
		 */
4514
		schedule_timeout_uninterruptible(1);
L
Linus Torvalds 已提交
4515
	}
4516 4517 4518 4519
	/* 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.
	 */
4520
	for (i = 0; i < conf->raid_disks; i++)
4521 4522 4523 4524 4525
		if (conf->disks[i].rdev == NULL)
			still_degraded = 1;

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

4526
	set_bit(STRIPE_SYNC_REQUESTED, &sh->state);
L
Linus Torvalds 已提交
4527

4528
	handle_stripe(sh);
L
Linus Torvalds 已提交
4529 4530 4531 4532 4533
	release_stripe(sh);

	return STRIPE_SECTORS;
}

4534
static int  retry_aligned_read(struct r5conf *conf, struct bio *raid_bio)
4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546
{
	/* 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;
4547
	int dd_idx;
4548 4549 4550 4551 4552 4553
	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);
4554
	sector = raid5_compute_sector(conf, logical_sector,
4555
				      0, &dd_idx, NULL);
4556 4557 4558
	last_sector = raid_bio->bi_sector + (raid_bio->bi_size>>9);

	for (; logical_sector < last_sector;
4559 4560 4561
	     logical_sector += STRIPE_SECTORS,
		     sector += STRIPE_SECTORS,
		     scnt++) {
4562

4563
		if (scnt < raid5_bi_processed_stripes(raid_bio))
4564 4565 4566
			/* already done this stripe */
			continue;

4567
		sh = get_active_stripe(conf, sector, 0, 1, 0);
4568 4569 4570

		if (!sh) {
			/* failed to get a stripe - must wait */
4571
			raid5_set_bi_processed_stripes(raid_bio, scnt);
4572 4573 4574 4575
			conf->retry_read_aligned = raid_bio;
			return handled;
		}

4576 4577
		if (!add_stripe_bio(sh, raid_bio, dd_idx, 0)) {
			release_stripe(sh);
4578
			raid5_set_bi_processed_stripes(raid_bio, scnt);
4579 4580 4581 4582
			conf->retry_read_aligned = raid_bio;
			return handled;
		}

4583
		set_bit(R5_ReadNoMerge, &sh->dev[dd_idx].flags);
4584
		handle_stripe(sh);
4585 4586 4587
		release_stripe(sh);
		handled++;
	}
4588
	remaining = raid5_dec_bi_active_stripes(raid_bio);
4589 4590
	if (remaining == 0)
		bio_endio(raid_bio, 0);
4591 4592 4593 4594 4595
	if (atomic_dec_and_test(&conf->active_aligned_reads))
		wake_up(&conf->wait_for_stripe);
	return handled;
}

4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619
#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;
}
4620

L
Linus Torvalds 已提交
4621 4622 4623 4624 4625 4626 4627
/*
 * 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.
 */
4628
static void raid5d(struct mddev *mddev)
L
Linus Torvalds 已提交
4629
{
4630
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
4631
	int handled;
4632
	struct blk_plug plug;
L
Linus Torvalds 已提交
4633

4634
	pr_debug("+++ raid5d active\n");
L
Linus Torvalds 已提交
4635 4636 4637

	md_check_recovery(mddev);

4638
	blk_start_plug(&plug);
L
Linus Torvalds 已提交
4639 4640 4641
	handled = 0;
	spin_lock_irq(&conf->device_lock);
	while (1) {
4642
		struct bio *bio;
4643
		int batch_size;
L
Linus Torvalds 已提交
4644

4645
		if (
4646 4647 4648
		    !list_empty(&conf->bitmap_list)) {
			/* Now is a good time to flush some bitmap updates */
			conf->seq_flush++;
4649
			spin_unlock_irq(&conf->device_lock);
4650
			bitmap_unplug(mddev->bitmap);
4651
			spin_lock_irq(&conf->device_lock);
4652
			conf->seq_write = conf->seq_flush;
4653 4654
			activate_bit_delay(conf);
		}
4655
		raid5_activate_delayed(conf);
4656

4657 4658 4659 4660 4661 4662 4663 4664 4665 4666
		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++;
		}

4667 4668
		batch_size = handle_active_stripes(conf);
		if (!batch_size)
L
Linus Torvalds 已提交
4669
			break;
4670
		handled += batch_size;
L
Linus Torvalds 已提交
4671

4672 4673
		if (mddev->flags & ~(1<<MD_CHANGE_PENDING)) {
			spin_unlock_irq(&conf->device_lock);
4674
			md_check_recovery(mddev);
4675 4676
			spin_lock_irq(&conf->device_lock);
		}
L
Linus Torvalds 已提交
4677
	}
4678
	pr_debug("%d stripes handled\n", handled);
L
Linus Torvalds 已提交
4679 4680 4681

	spin_unlock_irq(&conf->device_lock);

4682
	async_tx_issue_pending_all();
4683
	blk_finish_plug(&plug);
L
Linus Torvalds 已提交
4684

4685
	pr_debug("--- raid5d inactive\n");
L
Linus Torvalds 已提交
4686 4687
}

4688
static ssize_t
4689
raid5_show_stripe_cache_size(struct mddev *mddev, char *page)
4690
{
4691
	struct r5conf *conf = mddev->private;
4692 4693 4694 4695
	if (conf)
		return sprintf(page, "%d\n", conf->max_nr_stripes);
	else
		return 0;
4696 4697
}

4698
int
4699
raid5_set_cache_size(struct mddev *mddev, int size)
4700
{
4701
	struct r5conf *conf = mddev->private;
4702 4703
	int err;

4704
	if (size <= 16 || size > 32768)
4705
		return -EINVAL;
4706
	while (size < conf->max_nr_stripes) {
4707 4708 4709 4710 4711
		if (drop_one_stripe(conf))
			conf->max_nr_stripes--;
		else
			break;
	}
4712 4713 4714
	err = md_allow_write(mddev);
	if (err)
		return err;
4715
	while (size > conf->max_nr_stripes) {
4716 4717 4718 4719
		if (grow_one_stripe(conf))
			conf->max_nr_stripes++;
		else break;
	}
4720 4721 4722 4723 4724
	return 0;
}
EXPORT_SYMBOL(raid5_set_cache_size);

static ssize_t
4725
raid5_store_stripe_cache_size(struct mddev *mddev, const char *page, size_t len)
4726
{
4727
	struct r5conf *conf = mddev->private;
4728 4729 4730 4731 4732 4733 4734 4735 4736 4737 4738 4739 4740
	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;
4741 4742
	return len;
}
4743

4744 4745 4746 4747
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);
4748

4749
static ssize_t
4750
raid5_show_preread_threshold(struct mddev *mddev, char *page)
4751
{
4752
	struct r5conf *conf = mddev->private;
4753 4754 4755 4756 4757 4758 4759
	if (conf)
		return sprintf(page, "%d\n", conf->bypass_threshold);
	else
		return 0;
}

static ssize_t
4760
raid5_store_preread_threshold(struct mddev *mddev, const char *page, size_t len)
4761
{
4762
	struct r5conf *conf = mddev->private;
4763
	unsigned long new;
4764 4765 4766 4767 4768
	if (len >= PAGE_SIZE)
		return -EINVAL;
	if (!conf)
		return -ENODEV;

4769
	if (strict_strtoul(page, 10, &new))
4770
		return -EINVAL;
4771
	if (new > conf->max_nr_stripes)
4772 4773 4774 4775 4776 4777 4778 4779 4780 4781 4782
		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);

4783
static ssize_t
4784
stripe_cache_active_show(struct mddev *mddev, char *page)
4785
{
4786
	struct r5conf *conf = mddev->private;
4787 4788 4789 4790
	if (conf)
		return sprintf(page, "%d\n", atomic_read(&conf->active_stripes));
	else
		return 0;
4791 4792
}

4793 4794
static struct md_sysfs_entry
raid5_stripecache_active = __ATTR_RO(stripe_cache_active);
4795

4796
static struct attribute *raid5_attrs[] =  {
4797 4798
	&raid5_stripecache_size.attr,
	&raid5_stripecache_active.attr,
4799
	&raid5_preread_bypass_threshold.attr,
4800 4801
	NULL,
};
4802 4803 4804
static struct attribute_group raid5_attrs_group = {
	.name = NULL,
	.attrs = raid5_attrs,
4805 4806
};

4807
static sector_t
4808
raid5_size(struct mddev *mddev, sector_t sectors, int raid_disks)
4809
{
4810
	struct r5conf *conf = mddev->private;
4811 4812 4813

	if (!sectors)
		sectors = mddev->dev_sectors;
4814
	if (!raid_disks)
4815
		/* size is defined by the smallest of previous and new size */
4816
		raid_disks = min(conf->raid_disks, conf->previous_raid_disks);
4817

4818
	sectors &= ~((sector_t)mddev->chunk_sectors - 1);
4819
	sectors &= ~((sector_t)mddev->new_chunk_sectors - 1);
4820 4821 4822
	return sectors * (raid_disks - conf->max_degraded);
}

4823
static void raid5_free_percpu(struct r5conf *conf)
4824 4825 4826 4827 4828 4829 4830 4831 4832 4833 4834
{
	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);
4835
		kfree(percpu->scribble);
4836 4837 4838 4839 4840 4841 4842 4843 4844
	}
#ifdef CONFIG_HOTPLUG_CPU
	unregister_cpu_notifier(&conf->cpu_notify);
#endif
	put_online_cpus();

	free_percpu(conf->percpu);
}

4845
static void free_conf(struct r5conf *conf)
4846 4847
{
	shrink_stripes(conf);
4848
	raid5_free_percpu(conf);
4849 4850 4851 4852 4853
	kfree(conf->disks);
	kfree(conf->stripe_hashtbl);
	kfree(conf);
}

4854 4855 4856 4857
#ifdef CONFIG_HOTPLUG_CPU
static int raid456_cpu_notify(struct notifier_block *nfb, unsigned long action,
			      void *hcpu)
{
4858
	struct r5conf *conf = container_of(nfb, struct r5conf, cpu_notify);
4859 4860 4861 4862 4863 4864
	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:
4865
		if (conf->level == 6 && !percpu->spare_page)
4866
			percpu->spare_page = alloc_page(GFP_KERNEL);
4867 4868 4869 4870 4871 4872 4873
		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);
4874 4875
			pr_err("%s: failed memory allocation for cpu%ld\n",
			       __func__, cpu);
4876
			return notifier_from_errno(-ENOMEM);
4877 4878 4879 4880 4881
		}
		break;
	case CPU_DEAD:
	case CPU_DEAD_FROZEN:
		safe_put_page(percpu->spare_page);
4882
		kfree(percpu->scribble);
4883
		percpu->spare_page = NULL;
4884
		percpu->scribble = NULL;
4885 4886 4887 4888 4889 4890 4891 4892
		break;
	default:
		break;
	}
	return NOTIFY_OK;
}
#endif

4893
static int raid5_alloc_percpu(struct r5conf *conf)
4894 4895 4896
{
	unsigned long cpu;
	struct page *spare_page;
4897
	struct raid5_percpu __percpu *allcpus;
4898
	void *scribble;
4899 4900 4901 4902 4903 4904 4905 4906 4907 4908
	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) {
4909 4910 4911 4912 4913 4914 4915 4916
		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;
		}
4917
		scribble = kmalloc(conf->scribble_len, GFP_KERNEL);
4918
		if (!scribble) {
4919 4920 4921
			err = -ENOMEM;
			break;
		}
4922
		per_cpu_ptr(conf->percpu, cpu)->scribble = scribble;
4923 4924 4925 4926 4927 4928 4929 4930 4931 4932 4933 4934
	}
#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;
}

4935
static struct r5conf *setup_conf(struct mddev *mddev)
L
Linus Torvalds 已提交
4936
{
4937
	struct r5conf *conf;
4938
	int raid_disk, memory, max_disks;
4939
	struct md_rdev *rdev;
L
Linus Torvalds 已提交
4940
	struct disk_info *disk;
4941
	char pers_name[6];
L
Linus Torvalds 已提交
4942

N
NeilBrown 已提交
4943 4944 4945
	if (mddev->new_level != 5
	    && mddev->new_level != 4
	    && mddev->new_level != 6) {
4946
		printk(KERN_ERR "md/raid:%s: raid level not set to 4/5/6 (%d)\n",
N
NeilBrown 已提交
4947 4948
		       mdname(mddev), mddev->new_level);
		return ERR_PTR(-EIO);
L
Linus Torvalds 已提交
4949
	}
N
NeilBrown 已提交
4950 4951 4952 4953
	if ((mddev->new_level == 5
	     && !algorithm_valid_raid5(mddev->new_layout)) ||
	    (mddev->new_level == 6
	     && !algorithm_valid_raid6(mddev->new_layout))) {
4954
		printk(KERN_ERR "md/raid:%s: layout %d not supported\n",
N
NeilBrown 已提交
4955 4956
		       mdname(mddev), mddev->new_layout);
		return ERR_PTR(-EIO);
4957
	}
N
NeilBrown 已提交
4958
	if (mddev->new_level == 6 && mddev->raid_disks < 4) {
4959
		printk(KERN_ERR "md/raid:%s: not enough configured devices (%d, minimum 4)\n",
N
NeilBrown 已提交
4960 4961
		       mdname(mddev), mddev->raid_disks);
		return ERR_PTR(-EINVAL);
4962 4963
	}

4964 4965 4966
	if (!mddev->new_chunk_sectors ||
	    (mddev->new_chunk_sectors << 9) % PAGE_SIZE ||
	    !is_power_of_2(mddev->new_chunk_sectors)) {
4967 4968
		printk(KERN_ERR "md/raid:%s: invalid chunk size %d\n",
		       mdname(mddev), mddev->new_chunk_sectors << 9);
N
NeilBrown 已提交
4969
		return ERR_PTR(-EINVAL);
4970 4971
	}

4972
	conf = kzalloc(sizeof(struct r5conf), GFP_KERNEL);
N
NeilBrown 已提交
4973
	if (conf == NULL)
L
Linus Torvalds 已提交
4974
		goto abort;
4975 4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986
	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;
4987
	conf->recovery_disabled = mddev->recovery_disabled - 1;
N
NeilBrown 已提交
4988 4989 4990 4991 4992

	conf->raid_disks = mddev->raid_disks;
	if (mddev->reshape_position == MaxSector)
		conf->previous_raid_disks = mddev->raid_disks;
	else
4993
		conf->previous_raid_disks = mddev->raid_disks - mddev->delta_disks;
4994 4995
	max_disks = max(conf->raid_disks, conf->previous_raid_disks);
	conf->scribble_len = scribble_len(max_disks);
4996

4997
	conf->disks = kzalloc(max_disks * sizeof(struct disk_info),
4998 4999 5000
			      GFP_KERNEL);
	if (!conf->disks)
		goto abort;
5001

L
Linus Torvalds 已提交
5002 5003
	conf->mddev = mddev;

5004
	if ((conf->stripe_hashtbl = kzalloc(PAGE_SIZE, GFP_KERNEL)) == NULL)
L
Linus Torvalds 已提交
5005 5006
		goto abort;

5007 5008 5009 5010
	conf->level = mddev->new_level;
	if (raid5_alloc_percpu(conf) != 0)
		goto abort;

5011
	pr_debug("raid456: run(%s) called.\n", mdname(mddev));
L
Linus Torvalds 已提交
5012

N
NeilBrown 已提交
5013
	rdev_for_each(rdev, mddev) {
L
Linus Torvalds 已提交
5014
		raid_disk = rdev->raid_disk;
5015
		if (raid_disk >= max_disks
L
Linus Torvalds 已提交
5016 5017 5018 5019
		    || raid_disk < 0)
			continue;
		disk = conf->disks + raid_disk;

5020 5021 5022 5023 5024 5025 5026 5027 5028
		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
5029

5030
		if (test_bit(In_sync, &rdev->flags)) {
L
Linus Torvalds 已提交
5031
			char b[BDEVNAME_SIZE];
5032 5033 5034
			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
5035
		} else if (rdev->saved_raid_disk != raid_disk)
5036 5037
			/* Cannot rely on bitmap to complete recovery */
			conf->fullsync = 1;
L
Linus Torvalds 已提交
5038 5039
	}

5040
	conf->chunk_sectors = mddev->new_chunk_sectors;
N
NeilBrown 已提交
5041
	conf->level = mddev->new_level;
5042 5043 5044 5045
	if (conf->level == 6)
		conf->max_degraded = 2;
	else
		conf->max_degraded = 1;
N
NeilBrown 已提交
5046
	conf->algorithm = mddev->new_layout;
L
Linus Torvalds 已提交
5047
	conf->max_nr_stripes = NR_STRIPES;
5048
	conf->reshape_progress = mddev->reshape_position;
5049
	if (conf->reshape_progress != MaxSector) {
5050
		conf->prev_chunk_sectors = mddev->chunk_sectors;
5051 5052
		conf->prev_algo = mddev->layout;
	}
L
Linus Torvalds 已提交
5053

N
NeilBrown 已提交
5054
	memory = conf->max_nr_stripes * (sizeof(struct stripe_head) +
5055
		 max_disks * ((sizeof(struct bio) + PAGE_SIZE))) / 1024;
N
NeilBrown 已提交
5056 5057
	if (grow_stripes(conf, conf->max_nr_stripes)) {
		printk(KERN_ERR
5058 5059
		       "md/raid:%s: couldn't allocate %dkB for buffers\n",
		       mdname(mddev), memory);
N
NeilBrown 已提交
5060 5061
		goto abort;
	} else
5062 5063
		printk(KERN_INFO "md/raid:%s: allocated %dkB\n",
		       mdname(mddev), memory);
L
Linus Torvalds 已提交
5064

5065 5066
	sprintf(pers_name, "raid%d", mddev->new_level);
	conf->thread = md_register_thread(raid5d, mddev, pers_name);
N
NeilBrown 已提交
5067 5068
	if (!conf->thread) {
		printk(KERN_ERR
5069
		       "md/raid:%s: couldn't allocate thread.\n",
N
NeilBrown 已提交
5070
		       mdname(mddev));
5071 5072
		goto abort;
	}
N
NeilBrown 已提交
5073 5074 5075 5076 5077

	return conf;

 abort:
	if (conf) {
5078
		free_conf(conf);
N
NeilBrown 已提交
5079 5080 5081 5082 5083
		return ERR_PTR(-EIO);
	} else
		return ERR_PTR(-ENOMEM);
}

5084 5085 5086 5087 5088 5089 5090 5091 5092 5093 5094 5095 5096 5097 5098 5099 5100 5101 5102 5103 5104 5105 5106 5107 5108 5109 5110

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;
}

5111
static int run(struct mddev *mddev)
N
NeilBrown 已提交
5112
{
5113
	struct r5conf *conf;
5114
	int working_disks = 0;
5115
	int dirty_parity_disks = 0;
5116
	struct md_rdev *rdev;
5117
	sector_t reshape_offset = 0;
5118
	int i;
5119 5120
	long long min_offset_diff = 0;
	int first = 1;
N
NeilBrown 已提交
5121

5122
	if (mddev->recovery_cp != MaxSector)
5123
		printk(KERN_NOTICE "md/raid:%s: not clean"
5124 5125
		       " -- starting background reconstruction\n",
		       mdname(mddev));
5126 5127 5128 5129 5130 5131 5132 5133 5134 5135 5136 5137 5138 5139 5140 5141 5142

	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
5143 5144
	if (mddev->reshape_position != MaxSector) {
		/* Check that we can continue the reshape.
5145 5146 5147 5148 5149 5150 5151 5152 5153 5154
		 * 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
5155 5156 5157
		 */
		sector_t here_new, here_old;
		int old_disks;
5158
		int max_degraded = (mddev->level == 6 ? 2 : 1);
N
NeilBrown 已提交
5159

5160
		if (mddev->new_level != mddev->level) {
5161
			printk(KERN_ERR "md/raid:%s: unsupported reshape "
N
NeilBrown 已提交
5162 5163 5164 5165 5166 5167 5168 5169 5170 5171
			       "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;
5172
		if (sector_div(here_new, mddev->new_chunk_sectors *
N
NeilBrown 已提交
5173
			       (mddev->raid_disks - max_degraded))) {
5174 5175
			printk(KERN_ERR "md/raid:%s: reshape_position not "
			       "on a stripe boundary\n", mdname(mddev));
N
NeilBrown 已提交
5176 5177
			return -EINVAL;
		}
5178
		reshape_offset = here_new * mddev->new_chunk_sectors;
N
NeilBrown 已提交
5179 5180
		/* here_new is the stripe we will write to */
		here_old = mddev->reshape_position;
5181
		sector_div(here_old, mddev->chunk_sectors *
N
NeilBrown 已提交
5182 5183 5184
			   (old_disks-max_degraded));
		/* here_old is the first stripe that we might need to read
		 * from */
5185
		if (mddev->delta_disks == 0) {
5186 5187 5188 5189 5190 5191
			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;
			}
5192
			/* We cannot be sure it is safe to start an in-place
5193
			 * reshape.  It is only safe if user-space is monitoring
5194 5195 5196 5197 5198
			 * 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.
			 */
5199 5200 5201 5202 5203 5204 5205
			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",
5206
				       mdname(mddev));
5207 5208
				return -EINVAL;
			}
5209
		} else if (mddev->reshape_backwards
5210
		    ? (here_new * mddev->new_chunk_sectors + min_offset_diff <=
5211 5212
		       here_old * mddev->chunk_sectors)
		    : (here_new * mddev->new_chunk_sectors >=
5213
		       here_old * mddev->chunk_sectors + (-min_offset_diff))) {
N
NeilBrown 已提交
5214
			/* Reading from the same stripe as writing to - bad */
5215 5216 5217
			printk(KERN_ERR "md/raid:%s: reshape_position too early for "
			       "auto-recovery - aborting.\n",
			       mdname(mddev));
N
NeilBrown 已提交
5218 5219
			return -EINVAL;
		}
5220 5221
		printk(KERN_INFO "md/raid:%s: reshape will continue\n",
		       mdname(mddev));
N
NeilBrown 已提交
5222 5223 5224 5225
		/* OK, we should be able to continue; */
	} else {
		BUG_ON(mddev->level != mddev->new_level);
		BUG_ON(mddev->layout != mddev->new_layout);
5226
		BUG_ON(mddev->chunk_sectors != mddev->new_chunk_sectors);
N
NeilBrown 已提交
5227
		BUG_ON(mddev->delta_disks != 0);
L
Linus Torvalds 已提交
5228
	}
N
NeilBrown 已提交
5229

5230 5231 5232 5233 5234
	if (mddev->private == NULL)
		conf = setup_conf(mddev);
	else
		conf = mddev->private;

N
NeilBrown 已提交
5235 5236 5237
	if (IS_ERR(conf))
		return PTR_ERR(conf);

5238
	conf->min_offset_diff = min_offset_diff;
N
NeilBrown 已提交
5239 5240 5241 5242
	mddev->thread = conf->thread;
	conf->thread = NULL;
	mddev->private = conf;

5243 5244 5245 5246 5247 5248 5249 5250 5251 5252 5253
	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)
5254
			continue;
5255 5256 5257 5258 5259 5260 5261
		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;
		}
5262
		if (test_bit(In_sync, &rdev->flags)) {
N
NeilBrown 已提交
5263
			working_disks++;
5264 5265
			continue;
		}
5266 5267 5268 5269 5270 5271 5272 5273 5274 5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291 5292 5293
		/* 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
5294

5295 5296 5297
	/*
	 * 0 for a fully functional array, 1 or 2 for a degraded array.
	 */
5298
	mddev->degraded = calc_degraded(conf);
N
NeilBrown 已提交
5299

5300
	if (has_failed(conf)) {
5301
		printk(KERN_ERR "md/raid:%s: not enough operational devices"
L
Linus Torvalds 已提交
5302
			" (%d/%d failed)\n",
5303
			mdname(mddev), mddev->degraded, conf->raid_disks);
L
Linus Torvalds 已提交
5304 5305 5306
		goto abort;
	}

N
NeilBrown 已提交
5307
	/* device size must be a multiple of chunk size */
5308
	mddev->dev_sectors &= ~(mddev->chunk_sectors - 1);
N
NeilBrown 已提交
5309 5310
	mddev->resync_max_sectors = mddev->dev_sectors;

5311
	if (mddev->degraded > dirty_parity_disks &&
L
Linus Torvalds 已提交
5312
	    mddev->recovery_cp != MaxSector) {
5313 5314
		if (mddev->ok_start_degraded)
			printk(KERN_WARNING
5315 5316
			       "md/raid:%s: starting dirty degraded array"
			       " - data corruption possible.\n",
5317 5318 5319
			       mdname(mddev));
		else {
			printk(KERN_ERR
5320
			       "md/raid:%s: cannot start dirty degraded array.\n",
5321 5322 5323
			       mdname(mddev));
			goto abort;
		}
L
Linus Torvalds 已提交
5324 5325 5326
	}

	if (mddev->degraded == 0)
5327 5328
		printk(KERN_INFO "md/raid:%s: raid level %d active with %d out of %d"
		       " devices, algorithm %d\n", mdname(mddev), conf->level,
5329 5330
		       mddev->raid_disks-mddev->degraded, mddev->raid_disks,
		       mddev->new_layout);
L
Linus Torvalds 已提交
5331
	else
5332 5333 5334 5335 5336
		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
5337 5338 5339

	print_raid5_conf(conf);

5340 5341
	if (conf->reshape_progress != MaxSector) {
		conf->reshape_safe = conf->reshape_progress;
5342 5343 5344 5345 5346 5347
		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,
5348
							"reshape");
5349 5350
	}

L
Linus Torvalds 已提交
5351 5352

	/* Ok, everything is just fine now */
5353 5354
	if (mddev->to_remove == &raid5_attrs_group)
		mddev->to_remove = NULL;
N
NeilBrown 已提交
5355 5356
	else if (mddev->kobj.sd &&
	    sysfs_create_group(&mddev->kobj, &raid5_attrs_group))
5357
		printk(KERN_WARNING
5358
		       "raid5: failed to create sysfs attributes for %s\n",
5359
		       mdname(mddev));
5360
	md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
5361

5362
	if (mddev->queue) {
5363
		int chunk_size;
5364 5365 5366 5367 5368 5369 5370 5371 5372
		/* 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
5373

5374
		blk_queue_merge_bvec(mddev->queue, raid5_mergeable_bvec);
5375

N
NeilBrown 已提交
5376 5377
		mddev->queue->backing_dev_info.congested_data = mddev;
		mddev->queue->backing_dev_info.congested_fn = raid5_congested;
5378

5379 5380 5381 5382
		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));
5383

5384
		rdev_for_each(rdev, mddev) {
5385 5386
			disk_stack_limits(mddev->gendisk, rdev->bdev,
					  rdev->data_offset << 9);
5387 5388 5389
			disk_stack_limits(mddev->gendisk, rdev->bdev,
					  rdev->new_data_offset << 9);
		}
5390
	}
5391

L
Linus Torvalds 已提交
5392 5393
	return 0;
abort:
5394
	md_unregister_thread(&mddev->thread);
N
NeilBrown 已提交
5395 5396
	print_raid5_conf(conf);
	free_conf(conf);
L
Linus Torvalds 已提交
5397
	mddev->private = NULL;
5398
	printk(KERN_ALERT "md/raid:%s: failed to run raid set.\n", mdname(mddev));
L
Linus Torvalds 已提交
5399 5400 5401
	return -EIO;
}

5402
static int stop(struct mddev *mddev)
L
Linus Torvalds 已提交
5403
{
5404
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5405

5406
	md_unregister_thread(&mddev->thread);
N
NeilBrown 已提交
5407 5408
	if (mddev->queue)
		mddev->queue->backing_dev_info.congested_fn = NULL;
5409
	free_conf(conf);
5410 5411
	mddev->private = NULL;
	mddev->to_remove = &raid5_attrs_group;
L
Linus Torvalds 已提交
5412 5413 5414
	return 0;
}

5415
static void status(struct seq_file *seq, struct mddev *mddev)
L
Linus Torvalds 已提交
5416
{
5417
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5418 5419
	int i;

5420 5421
	seq_printf(seq, " level %d, %dk chunk, algorithm %d", mddev->level,
		mddev->chunk_sectors / 2, mddev->layout);
5422
	seq_printf (seq, " [%d/%d] [", conf->raid_disks, conf->raid_disks - mddev->degraded);
L
Linus Torvalds 已提交
5423 5424 5425
	for (i = 0; i < conf->raid_disks; i++)
		seq_printf (seq, "%s",
			       conf->disks[i].rdev &&
5426
			       test_bit(In_sync, &conf->disks[i].rdev->flags) ? "U" : "_");
L
Linus Torvalds 已提交
5427 5428 5429
	seq_printf (seq, "]");
}

5430
static void print_raid5_conf (struct r5conf *conf)
L
Linus Torvalds 已提交
5431 5432 5433 5434
{
	int i;
	struct disk_info *tmp;

5435
	printk(KERN_DEBUG "RAID conf printout:\n");
L
Linus Torvalds 已提交
5436 5437 5438 5439
	if (!conf) {
		printk("(conf==NULL)\n");
		return;
	}
5440 5441 5442
	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
5443 5444 5445 5446 5447

	for (i = 0; i < conf->raid_disks; i++) {
		char b[BDEVNAME_SIZE];
		tmp = conf->disks + i;
		if (tmp->rdev)
5448 5449 5450
			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
5451 5452 5453
	}
}

5454
static int raid5_spare_active(struct mddev *mddev)
L
Linus Torvalds 已提交
5455 5456
{
	int i;
5457
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5458
	struct disk_info *tmp;
5459 5460
	int count = 0;
	unsigned long flags;
L
Linus Torvalds 已提交
5461 5462 5463

	for (i = 0; i < conf->raid_disks; i++) {
		tmp = conf->disks + i;
5464 5465 5466 5467 5468 5469 5470 5471 5472 5473 5474 5475 5476 5477 5478 5479 5480 5481 5482
		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
5483
		    && tmp->rdev->recovery_offset == MaxSector
5484
		    && !test_bit(Faulty, &tmp->rdev->flags)
5485
		    && !test_and_set_bit(In_sync, &tmp->rdev->flags)) {
5486
			count++;
5487
			sysfs_notify_dirent_safe(tmp->rdev->sysfs_state);
L
Linus Torvalds 已提交
5488 5489
		}
	}
5490
	spin_lock_irqsave(&conf->device_lock, flags);
5491
	mddev->degraded = calc_degraded(conf);
5492
	spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
5493
	print_raid5_conf(conf);
5494
	return count;
L
Linus Torvalds 已提交
5495 5496
}

5497
static int raid5_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
5498
{
5499
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5500
	int err = 0;
5501
	int number = rdev->raid_disk;
5502
	struct md_rdev **rdevp;
L
Linus Torvalds 已提交
5503 5504 5505
	struct disk_info *p = conf->disks + number;

	print_raid5_conf(conf);
5506 5507 5508 5509 5510 5511 5512 5513 5514 5515 5516 5517 5518 5519 5520 5521 5522 5523 5524 5525 5526 5527
	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) &&
5528
	    (!p->replacement || p->replacement == rdev) &&
5529 5530 5531 5532 5533 5534 5535 5536 5537 5538
	    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;
5539 5540 5541 5542 5543 5544 5545 5546 5547 5548 5549 5550 5551 5552
	} 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
5553 5554 5555 5556 5557 5558
abort:

	print_raid5_conf(conf);
	return err;
}

5559
static int raid5_add_disk(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
5560
{
5561
	struct r5conf *conf = mddev->private;
5562
	int err = -EEXIST;
L
Linus Torvalds 已提交
5563 5564
	int disk;
	struct disk_info *p;
5565 5566
	int first = 0;
	int last = conf->raid_disks - 1;
L
Linus Torvalds 已提交
5567

5568 5569 5570
	if (mddev->recovery_disabled == conf->recovery_disabled)
		return -EBUSY;

N
NeilBrown 已提交
5571
	if (rdev->saved_raid_disk < 0 && has_failed(conf))
L
Linus Torvalds 已提交
5572
		/* no point adding a device */
5573
		return -EINVAL;
L
Linus Torvalds 已提交
5574

5575 5576
	if (rdev->raid_disk >= 0)
		first = last = rdev->raid_disk;
L
Linus Torvalds 已提交
5577 5578

	/*
5579 5580
	 * find the disk ... but prefer rdev->saved_raid_disk
	 * if possible.
L
Linus Torvalds 已提交
5581
	 */
5582
	if (rdev->saved_raid_disk >= 0 &&
5583
	    rdev->saved_raid_disk >= first &&
5584
	    conf->disks[rdev->saved_raid_disk].rdev == NULL)
5585 5586 5587
		first = rdev->saved_raid_disk;

	for (disk = first; disk <= last; disk++) {
5588 5589
		p = conf->disks + disk;
		if (p->rdev == NULL) {
5590
			clear_bit(In_sync, &rdev->flags);
L
Linus Torvalds 已提交
5591
			rdev->raid_disk = disk;
5592
			err = 0;
5593 5594
			if (rdev->saved_raid_disk != disk)
				conf->fullsync = 1;
5595
			rcu_assign_pointer(p->rdev, rdev);
5596
			goto out;
L
Linus Torvalds 已提交
5597
		}
5598 5599 5600
	}
	for (disk = first; disk <= last; disk++) {
		p = conf->disks + disk;
5601 5602 5603 5604 5605 5606 5607 5608 5609 5610 5611
		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;
		}
	}
5612
out:
L
Linus Torvalds 已提交
5613
	print_raid5_conf(conf);
5614
	return err;
L
Linus Torvalds 已提交
5615 5616
}

5617
static int raid5_resize(struct mddev *mddev, sector_t sectors)
L
Linus Torvalds 已提交
5618 5619 5620 5621 5622 5623 5624 5625
{
	/* 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.
	 */
5626
	sector_t newsize;
5627
	sectors &= ~((sector_t)mddev->chunk_sectors - 1);
5628 5629 5630
	newsize = raid5_size(mddev, sectors, mddev->raid_disks);
	if (mddev->external_size &&
	    mddev->array_sectors > newsize)
D
Dan Williams 已提交
5631
		return -EINVAL;
5632 5633 5634 5635 5636 5637
	if (mddev->bitmap) {
		int ret = bitmap_resize(mddev->bitmap, sectors, 0, 0);
		if (ret)
			return ret;
	}
	md_set_array_sectors(mddev, newsize);
5638
	set_capacity(mddev->gendisk, mddev->array_sectors);
5639
	revalidate_disk(mddev->gendisk);
5640 5641
	if (sectors > mddev->dev_sectors &&
	    mddev->recovery_cp > mddev->dev_sectors) {
A
Andre Noll 已提交
5642
		mddev->recovery_cp = mddev->dev_sectors;
L
Linus Torvalds 已提交
5643 5644
		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
	}
A
Andre Noll 已提交
5645
	mddev->dev_sectors = sectors;
5646
	mddev->resync_max_sectors = sectors;
L
Linus Torvalds 已提交
5647 5648 5649
	return 0;
}

5650
static int check_stripe_cache(struct mddev *mddev)
5651 5652 5653 5654 5655 5656 5657 5658 5659
{
	/* 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.
	 */
5660
	struct r5conf *conf = mddev->private;
5661 5662 5663 5664
	if (((mddev->chunk_sectors << 9) / STRIPE_SIZE) * 4
	    > conf->max_nr_stripes ||
	    ((mddev->new_chunk_sectors << 9) / STRIPE_SIZE) * 4
	    > conf->max_nr_stripes) {
5665 5666
		printk(KERN_WARNING "md/raid:%s: reshape: not enough stripes.  Needed %lu\n",
		       mdname(mddev),
5667 5668 5669 5670 5671 5672 5673
		       ((max(mddev->chunk_sectors, mddev->new_chunk_sectors) << 9)
			/ STRIPE_SIZE)*4);
		return 0;
	}
	return 1;
}

5674
static int check_reshape(struct mddev *mddev)
5675
{
5676
	struct r5conf *conf = mddev->private;
5677

5678 5679
	if (mddev->delta_disks == 0 &&
	    mddev->new_layout == mddev->layout &&
5680
	    mddev->new_chunk_sectors == mddev->chunk_sectors)
5681
		return 0; /* nothing to do */
5682
	if (has_failed(conf))
5683 5684 5685 5686 5687 5688 5689 5690 5691 5692 5693 5694 5695
		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;
	}
5696

5697
	if (!check_stripe_cache(mddev))
5698 5699
		return -ENOSPC;

5700
	return resize_stripes(conf, conf->raid_disks + mddev->delta_disks);
5701 5702
}

5703
static int raid5_start_reshape(struct mddev *mddev)
5704
{
5705
	struct r5conf *conf = mddev->private;
5706
	struct md_rdev *rdev;
5707
	int spares = 0;
5708
	unsigned long flags;
5709

5710
	if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
5711 5712
		return -EBUSY;

5713 5714 5715
	if (!check_stripe_cache(mddev))
		return -ENOSPC;

5716 5717 5718
	if (has_failed(conf))
		return -EINVAL;

5719
	rdev_for_each(rdev, mddev) {
5720 5721
		if (!test_bit(In_sync, &rdev->flags)
		    && !test_bit(Faulty, &rdev->flags))
5722
			spares++;
5723
	}
5724

5725
	if (spares - mddev->degraded < mddev->delta_disks - conf->max_degraded)
5726 5727 5728 5729 5730
		/* Not enough devices even to make a degraded array
		 * of that size
		 */
		return -EINVAL;

5731 5732 5733 5734 5735 5736
	/* 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) {
5737
		printk(KERN_ERR "md/raid:%s: array size must be reduced "
5738 5739 5740 5741
		       "before number of disks\n", mdname(mddev));
		return -EINVAL;
	}

5742
	atomic_set(&conf->reshape_stripes, 0);
5743 5744
	spin_lock_irq(&conf->device_lock);
	conf->previous_raid_disks = conf->raid_disks;
5745
	conf->raid_disks += mddev->delta_disks;
5746 5747
	conf->prev_chunk_sectors = conf->chunk_sectors;
	conf->chunk_sectors = mddev->new_chunk_sectors;
5748 5749
	conf->prev_algo = conf->algorithm;
	conf->algorithm = mddev->new_layout;
5750 5751 5752 5753 5754
	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();
5755
	if (mddev->reshape_backwards)
5756 5757 5758 5759
		conf->reshape_progress = raid5_size(mddev, 0, 0);
	else
		conf->reshape_progress = 0;
	conf->reshape_safe = conf->reshape_progress;
5760 5761 5762 5763
	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.
5764 5765 5766 5767
	 * 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.
5768
	 */
5769
	if (mddev->delta_disks >= 0) {
N
NeilBrown 已提交
5770
		rdev_for_each(rdev, mddev)
5771 5772 5773 5774
			if (rdev->raid_disk < 0 &&
			    !test_bit(Faulty, &rdev->flags)) {
				if (raid5_add_disk(mddev, rdev) == 0) {
					if (rdev->raid_disk
5775
					    >= conf->previous_raid_disks)
5776
						set_bit(In_sync, &rdev->flags);
5777
					else
5778
						rdev->recovery_offset = 0;
5779 5780

					if (sysfs_link_rdev(mddev, rdev))
5781
						/* Failure here is OK */;
5782
				}
5783 5784 5785 5786 5787
			} 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);
			}
5788

5789 5790 5791 5792
		/* When a reshape changes the number of devices,
		 * ->degraded is measured against the larger of the
		 * pre and post number of devices.
		 */
5793
		spin_lock_irqsave(&conf->device_lock, flags);
5794
		mddev->degraded = calc_degraded(conf);
5795 5796
		spin_unlock_irqrestore(&conf->device_lock, flags);
	}
5797
	mddev->raid_disks = conf->raid_disks;
5798
	mddev->reshape_position = conf->reshape_progress;
5799
	set_bit(MD_CHANGE_DEVS, &mddev->flags);
5800

5801 5802 5803 5804 5805
	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,
5806
						"reshape");
5807 5808 5809 5810
	if (!mddev->sync_thread) {
		mddev->recovery = 0;
		spin_lock_irq(&conf->device_lock);
		mddev->raid_disks = conf->raid_disks = conf->previous_raid_disks;
5811 5812 5813
		rdev_for_each(rdev, mddev)
			rdev->new_data_offset = rdev->data_offset;
		smp_wmb();
5814
		conf->reshape_progress = MaxSector;
5815
		mddev->reshape_position = MaxSector;
5816 5817 5818
		spin_unlock_irq(&conf->device_lock);
		return -EAGAIN;
	}
5819
	conf->reshape_checkpoint = jiffies;
5820 5821 5822 5823 5824
	md_wakeup_thread(mddev->sync_thread);
	md_new_event(mddev);
	return 0;
}

5825 5826 5827
/* This is called from the reshape thread and should make any
 * changes needed in 'conf'
 */
5828
static void end_reshape(struct r5conf *conf)
5829 5830
{

5831
	if (!test_bit(MD_RECOVERY_INTR, &conf->mddev->recovery)) {
5832
		struct md_rdev *rdev;
5833 5834

		spin_lock_irq(&conf->device_lock);
5835
		conf->previous_raid_disks = conf->raid_disks;
5836 5837 5838
		rdev_for_each(rdev, conf->mddev)
			rdev->data_offset = rdev->new_data_offset;
		smp_wmb();
5839
		conf->reshape_progress = MaxSector;
5840
		spin_unlock_irq(&conf->device_lock);
5841
		wake_up(&conf->wait_for_overlap);
5842 5843 5844 5845

		/* read-ahead size must cover two whole stripes, which is
		 * 2 * (datadisks) * chunksize where 'n' is the number of raid devices
		 */
5846
		if (conf->mddev->queue) {
5847
			int data_disks = conf->raid_disks - conf->max_degraded;
5848
			int stripe = data_disks * ((conf->chunk_sectors << 9)
5849
						   / PAGE_SIZE);
5850 5851 5852
			if (conf->mddev->queue->backing_dev_info.ra_pages < 2 * stripe)
				conf->mddev->queue->backing_dev_info.ra_pages = 2 * stripe;
		}
5853 5854 5855
	}
}

5856 5857 5858
/* This is called from the raid5d thread with mddev_lock held.
 * It makes config changes to the device.
 */
5859
static void raid5_finish_reshape(struct mddev *mddev)
5860
{
5861
	struct r5conf *conf = mddev->private;
5862 5863 5864

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

5865 5866 5867
		if (mddev->delta_disks > 0) {
			md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
			set_capacity(mddev->gendisk, mddev->array_sectors);
5868
			revalidate_disk(mddev->gendisk);
5869 5870
		} else {
			int d;
5871 5872 5873
			spin_lock_irq(&conf->device_lock);
			mddev->degraded = calc_degraded(conf);
			spin_unlock_irq(&conf->device_lock);
5874 5875
			for (d = conf->raid_disks ;
			     d < conf->raid_disks - mddev->delta_disks;
5876
			     d++) {
5877
				struct md_rdev *rdev = conf->disks[d].rdev;
5878 5879 5880 5881 5882
				if (rdev)
					clear_bit(In_sync, &rdev->flags);
				rdev = conf->disks[d].replacement;
				if (rdev)
					clear_bit(In_sync, &rdev->flags);
5883
			}
5884
		}
5885
		mddev->layout = conf->algorithm;
5886
		mddev->chunk_sectors = conf->chunk_sectors;
5887 5888
		mddev->reshape_position = MaxSector;
		mddev->delta_disks = 0;
5889
		mddev->reshape_backwards = 0;
5890 5891 5892
	}
}

5893
static void raid5_quiesce(struct mddev *mddev, int state)
5894
{
5895
	struct r5conf *conf = mddev->private;
5896 5897

	switch(state) {
5898 5899 5900 5901
	case 2: /* resume for a suspend */
		wake_up(&conf->wait_for_overlap);
		break;

5902 5903
	case 1: /* stop all writes */
		spin_lock_irq(&conf->device_lock);
5904 5905 5906 5907
		/* '2' tells resync/reshape to pause so that all
		 * active stripes can drain
		 */
		conf->quiesce = 2;
5908
		wait_event_lock_irq(conf->wait_for_stripe,
5909 5910
				    atomic_read(&conf->active_stripes) == 0 &&
				    atomic_read(&conf->active_aligned_reads) == 0,
5911
				    conf->device_lock, /* nothing */);
5912
		conf->quiesce = 1;
5913
		spin_unlock_irq(&conf->device_lock);
5914 5915
		/* allow reshape to continue */
		wake_up(&conf->wait_for_overlap);
5916 5917 5918 5919 5920 5921
		break;

	case 0: /* re-enable writes */
		spin_lock_irq(&conf->device_lock);
		conf->quiesce = 0;
		wake_up(&conf->wait_for_stripe);
5922
		wake_up(&conf->wait_for_overlap);
5923 5924 5925 5926
		spin_unlock_irq(&conf->device_lock);
		break;
	}
}
5927

5928

5929
static void *raid45_takeover_raid0(struct mddev *mddev, int level)
5930
{
5931
	struct r0conf *raid0_conf = mddev->private;
5932
	sector_t sectors;
5933

D
Dan Williams 已提交
5934
	/* for raid0 takeover only one zone is supported */
5935
	if (raid0_conf->nr_strip_zones > 1) {
5936 5937
		printk(KERN_ERR "md/raid:%s: cannot takeover raid0 with more than one zone.\n",
		       mdname(mddev));
D
Dan Williams 已提交
5938 5939 5940
		return ERR_PTR(-EINVAL);
	}

5941 5942
	sectors = raid0_conf->strip_zone[0].zone_end;
	sector_div(sectors, raid0_conf->strip_zone[0].nb_dev);
5943
	mddev->dev_sectors = sectors;
D
Dan Williams 已提交
5944
	mddev->new_level = level;
5945 5946 5947 5948 5949 5950 5951 5952 5953 5954 5955
	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);
}


5956
static void *raid5_takeover_raid1(struct mddev *mddev)
5957 5958 5959 5960 5961 5962 5963 5964 5965 5966 5967 5968 5969 5970 5971 5972 5973 5974 5975 5976 5977
{
	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;
5978
	mddev->new_chunk_sectors = chunksect;
5979 5980 5981 5982

	return setup_conf(mddev);
}

5983
static void *raid5_takeover_raid6(struct mddev *mddev)
5984 5985 5986 5987 5988 5989 5990 5991 5992 5993 5994 5995 5996 5997 5998 5999 6000 6001 6002 6003 6004 6005 6006 6007 6008 6009 6010 6011 6012 6013 6014 6015
{
	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);
}

6016

6017
static int raid5_check_reshape(struct mddev *mddev)
6018
{
6019 6020 6021 6022
	/* 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.
6023
	 */
6024
	struct r5conf *conf = mddev->private;
6025
	int new_chunk = mddev->new_chunk_sectors;
6026

6027
	if (mddev->new_layout >= 0 && !algorithm_valid_raid5(mddev->new_layout))
6028 6029
		return -EINVAL;
	if (new_chunk > 0) {
6030
		if (!is_power_of_2(new_chunk))
6031
			return -EINVAL;
6032
		if (new_chunk < (PAGE_SIZE>>9))
6033
			return -EINVAL;
6034
		if (mddev->array_sectors & (new_chunk-1))
6035 6036 6037 6038 6039 6040
			/* not factor of array size */
			return -EINVAL;
	}

	/* They look valid */

6041
	if (mddev->raid_disks == 2) {
6042 6043 6044 6045
		/* can make the change immediately */
		if (mddev->new_layout >= 0) {
			conf->algorithm = mddev->new_layout;
			mddev->layout = mddev->new_layout;
6046 6047
		}
		if (new_chunk > 0) {
6048 6049
			conf->chunk_sectors = new_chunk ;
			mddev->chunk_sectors = new_chunk;
6050 6051 6052
		}
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
		md_wakeup_thread(mddev->thread);
6053
	}
6054
	return check_reshape(mddev);
6055 6056
}

6057
static int raid6_check_reshape(struct mddev *mddev)
6058
{
6059
	int new_chunk = mddev->new_chunk_sectors;
6060

6061
	if (mddev->new_layout >= 0 && !algorithm_valid_raid6(mddev->new_layout))
6062
		return -EINVAL;
6063
	if (new_chunk > 0) {
6064
		if (!is_power_of_2(new_chunk))
6065
			return -EINVAL;
6066
		if (new_chunk < (PAGE_SIZE >> 9))
6067
			return -EINVAL;
6068
		if (mddev->array_sectors & (new_chunk-1))
6069 6070
			/* not factor of array size */
			return -EINVAL;
6071
	}
6072 6073

	/* They look valid */
6074
	return check_reshape(mddev);
6075 6076
}

6077
static void *raid5_takeover(struct mddev *mddev)
6078 6079
{
	/* raid5 can take over:
D
Dan Williams 已提交
6080
	 *  raid0 - if there is only one strip zone - make it a raid4 layout
6081 6082 6083 6084
	 *  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
6085 6086
	if (mddev->level == 0)
		return raid45_takeover_raid0(mddev, 5);
6087 6088
	if (mddev->level == 1)
		return raid5_takeover_raid1(mddev);
6089 6090 6091 6092 6093
	if (mddev->level == 4) {
		mddev->new_layout = ALGORITHM_PARITY_N;
		mddev->new_level = 5;
		return setup_conf(mddev);
	}
6094 6095
	if (mddev->level == 6)
		return raid5_takeover_raid6(mddev);
6096 6097 6098 6099

	return ERR_PTR(-EINVAL);
}

6100
static void *raid4_takeover(struct mddev *mddev)
6101
{
D
Dan Williams 已提交
6102 6103 6104
	/* raid4 can take over:
	 *  raid0 - if there is only one strip zone
	 *  raid5 - if layout is right
6105
	 */
D
Dan Williams 已提交
6106 6107
	if (mddev->level == 0)
		return raid45_takeover_raid0(mddev, 4);
6108 6109 6110 6111 6112 6113 6114 6115
	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);
}
6116

6117
static struct md_personality raid5_personality;
6118

6119
static void *raid6_takeover(struct mddev *mddev)
6120 6121 6122 6123 6124 6125 6126 6127 6128 6129 6130 6131 6132 6133 6134 6135 6136 6137 6138 6139 6140 6141 6142 6143 6144 6145 6146 6147 6148 6149 6150 6151 6152 6153 6154 6155 6156 6157 6158 6159 6160 6161 6162 6163 6164 6165
{
	/* 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);
}


6166
static struct md_personality raid6_personality =
6167 6168 6169 6170 6171 6172 6173 6174 6175 6176 6177 6178 6179 6180
{
	.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,
6181
	.size		= raid5_size,
6182
	.check_reshape	= raid6_check_reshape,
6183
	.start_reshape  = raid5_start_reshape,
6184
	.finish_reshape = raid5_finish_reshape,
6185
	.quiesce	= raid5_quiesce,
6186
	.takeover	= raid6_takeover,
6187
};
6188
static struct md_personality raid5_personality =
L
Linus Torvalds 已提交
6189 6190
{
	.name		= "raid5",
6191
	.level		= 5,
L
Linus Torvalds 已提交
6192 6193 6194 6195 6196 6197 6198 6199 6200 6201 6202
	.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,
6203
	.size		= raid5_size,
6204 6205
	.check_reshape	= raid5_check_reshape,
	.start_reshape  = raid5_start_reshape,
6206
	.finish_reshape = raid5_finish_reshape,
6207
	.quiesce	= raid5_quiesce,
6208
	.takeover	= raid5_takeover,
L
Linus Torvalds 已提交
6209 6210
};

6211
static struct md_personality raid4_personality =
L
Linus Torvalds 已提交
6212
{
6213 6214 6215 6216 6217 6218 6219 6220 6221 6222 6223 6224 6225
	.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,
6226
	.size		= raid5_size,
6227 6228
	.check_reshape	= raid5_check_reshape,
	.start_reshape  = raid5_start_reshape,
6229
	.finish_reshape = raid5_finish_reshape,
6230
	.quiesce	= raid5_quiesce,
6231
	.takeover	= raid4_takeover,
6232 6233 6234 6235
};

static int __init raid5_init(void)
{
6236
	register_md_personality(&raid6_personality);
6237 6238 6239
	register_md_personality(&raid5_personality);
	register_md_personality(&raid4_personality);
	return 0;
L
Linus Torvalds 已提交
6240 6241
}

6242
static void raid5_exit(void)
L
Linus Torvalds 已提交
6243
{
6244
	unregister_md_personality(&raid6_personality);
6245 6246
	unregister_md_personality(&raid5_personality);
	unregister_md_personality(&raid4_personality);
L
Linus Torvalds 已提交
6247 6248 6249 6250 6251
}

module_init(raid5_init);
module_exit(raid5_exit);
MODULE_LICENSE("GPL");
6252
MODULE_DESCRIPTION("RAID4/5/6 (striping with parity) personality for MD");
L
Linus Torvalds 已提交
6253
MODULE_ALIAS("md-personality-4"); /* RAID5 */
6254 6255
MODULE_ALIAS("md-raid5");
MODULE_ALIAS("md-raid4");
6256 6257
MODULE_ALIAS("md-level-5");
MODULE_ALIAS("md-level-4");
6258 6259 6260 6261 6262 6263 6264
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");