raid5.c 202.7 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 <linux/nodemask.h>
N
NeilBrown 已提交
57 58
#include <trace/events/block.h>

59
#include "md.h"
60
#include "raid5.h"
61
#include "raid0.h"
62
#include "bitmap.h"
63

64 65 66
#define cpu_to_group(cpu) cpu_to_node(cpu)
#define ANY_GROUP NUMA_NO_NODE

67 68 69 70
static bool devices_handle_discard_safely = false;
module_param(devices_handle_discard_safely, bool, 0644);
MODULE_PARM_DESC(devices_handle_discard_safely,
		 "Set to Y if all devices in each array reliably return zeroes on reads from discarded regions");
71
static struct workqueue_struct *raid5_wq;
L
Linus Torvalds 已提交
72 73 74 75 76 77 78 79 80
/*
 * 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
81
#define BYPASS_THRESHOLD	1
82
#define NR_HASH			(PAGE_SIZE / sizeof(struct hlist_head))
L
Linus Torvalds 已提交
83
#define HASH_MASK		(NR_HASH - 1)
84
#define MAX_STRIPE_BATCH	8
L
Linus Torvalds 已提交
85

86
static inline struct hlist_head *stripe_hash(struct r5conf *conf, sector_t sect)
87 88 89 90
{
	int hash = (sect >> STRIPE_SHIFT) & HASH_MASK;
	return &conf->stripe_hashtbl[hash];
}
L
Linus Torvalds 已提交
91

92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127
static inline int stripe_hash_locks_hash(sector_t sect)
{
	return (sect >> STRIPE_SHIFT) & STRIPE_HASH_LOCKS_MASK;
}

static inline void lock_device_hash_lock(struct r5conf *conf, int hash)
{
	spin_lock_irq(conf->hash_locks + hash);
	spin_lock(&conf->device_lock);
}

static inline void unlock_device_hash_lock(struct r5conf *conf, int hash)
{
	spin_unlock(&conf->device_lock);
	spin_unlock_irq(conf->hash_locks + hash);
}

static inline void lock_all_device_hash_locks_irq(struct r5conf *conf)
{
	int i;
	local_irq_disable();
	spin_lock(conf->hash_locks);
	for (i = 1; i < NR_STRIPE_HASH_LOCKS; i++)
		spin_lock_nest_lock(conf->hash_locks + i, conf->hash_locks);
	spin_lock(&conf->device_lock);
}

static inline void unlock_all_device_hash_locks_irq(struct r5conf *conf)
{
	int i;
	spin_unlock(&conf->device_lock);
	for (i = NR_STRIPE_HASH_LOCKS; i; i--)
		spin_unlock(conf->hash_locks + i - 1);
	local_irq_enable();
}

L
Linus Torvalds 已提交
128 129 130 131 132 133
/* 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.
134
 * This function is used to determine the 'next' bio in the list, given the sector
L
Linus Torvalds 已提交
135 136
 * of the current stripe+device
 */
137 138
static inline struct bio *r5_next_bio(struct bio *bio, sector_t sector)
{
139
	int sectors = bio_sectors(bio);
140
	if (bio->bi_iter.bi_sector + sectors < sector + STRIPE_SECTORS)
141 142 143 144
		return bio->bi_next;
	else
		return NULL;
}
L
Linus Torvalds 已提交
145

146
/*
147 148
 * 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
149
 */
150
static inline int raid5_bi_processed_stripes(struct bio *bio)
151
{
152 153
	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
	return (atomic_read(segments) >> 16) & 0xffff;
154 155
}

156
static inline int raid5_dec_bi_active_stripes(struct bio *bio)
157
{
158 159
	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
	return atomic_sub_return(1, segments) & 0xffff;
160 161
}

162
static inline void raid5_inc_bi_active_stripes(struct bio *bio)
163
{
164 165
	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
	atomic_inc(segments);
166 167
}

168 169
static inline void raid5_set_bi_processed_stripes(struct bio *bio,
	unsigned int cnt)
170
{
171 172
	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
	int old, new;
173

174 175 176 177
	do {
		old = atomic_read(segments);
		new = (old & 0xffff) | (cnt << 16);
	} while (atomic_cmpxchg(segments, old, new) != old);
178 179
}

180
static inline void raid5_set_bi_stripes(struct bio *bio, unsigned int cnt)
181
{
182 183
	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
	atomic_set(segments, cnt);
184 185
}

186 187 188
/* Find first data disk in a raid6 stripe */
static inline int raid6_d0(struct stripe_head *sh)
{
189 190 191 192
	if (sh->ddf_layout)
		/* ddf always start from first device */
		return 0;
	/* md starts just after Q block */
193 194 195 196 197
	if (sh->qd_idx == sh->disks - 1)
		return 0;
	else
		return sh->qd_idx + 1;
}
198 199 200 201 202
static inline int raid6_next_disk(int disk, int raid_disks)
{
	disk++;
	return (disk < raid_disks) ? disk : 0;
}
203

204 205 206 207 208
/* 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.
 */
209 210
static int raid6_idx_to_slot(int idx, struct stripe_head *sh,
			     int *count, int syndrome_disks)
211
{
212
	int slot = *count;
213

214
	if (sh->ddf_layout)
215
		(*count)++;
216
	if (idx == sh->pd_idx)
217
		return syndrome_disks;
218
	if (idx == sh->qd_idx)
219
		return syndrome_disks + 1;
220
	if (!sh->ddf_layout)
221
		(*count)++;
222 223 224
	return slot;
}

225 226 227 228 229 230 231
static void return_io(struct bio *return_bi)
{
	struct bio *bi = return_bi;
	while (bi) {

		return_bi = bi->bi_next;
		bi->bi_next = NULL;
232
		bi->bi_iter.bi_size = 0;
233 234
		trace_block_bio_complete(bdev_get_queue(bi->bi_bdev),
					 bi, 0);
235
		bio_endio(bi, 0);
236 237 238 239
		bi = return_bi;
	}
}

240
static void print_raid5_conf (struct r5conf *conf);
L
Linus Torvalds 已提交
241

242 243 244 245 246 247 248
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);
}

249 250 251 252
static void raid5_wakeup_stripe_thread(struct stripe_head *sh)
{
	struct r5conf *conf = sh->raid_conf;
	struct r5worker_group *group;
253
	int thread_cnt;
254 255 256 257 258 259 260 261 262 263 264
	int i, cpu = sh->cpu;

	if (!cpu_online(cpu)) {
		cpu = cpumask_any(cpu_online_mask);
		sh->cpu = cpu;
	}

	if (list_empty(&sh->lru)) {
		struct r5worker_group *group;
		group = conf->worker_groups + cpu_to_group(cpu);
		list_add_tail(&sh->lru, &group->handle_list);
265 266
		group->stripes_cnt++;
		sh->group = group;
267 268 269 270 271 272 273 274 275
	}

	if (conf->worker_cnt_per_group == 0) {
		md_wakeup_thread(conf->mddev->thread);
		return;
	}

	group = conf->worker_groups + cpu_to_group(sh->cpu);

276 277 278 279 280 281 282 283 284 285 286 287 288 289
	group->workers[0].working = true;
	/* at least one worker should run to avoid race */
	queue_work_on(sh->cpu, raid5_wq, &group->workers[0].work);

	thread_cnt = group->stripes_cnt / MAX_STRIPE_BATCH - 1;
	/* wakeup more workers */
	for (i = 1; i < conf->worker_cnt_per_group && thread_cnt > 0; i++) {
		if (group->workers[i].working == false) {
			group->workers[i].working = true;
			queue_work_on(sh->cpu, raid5_wq,
				      &group->workers[i].work);
			thread_cnt--;
		}
	}
290 291
}

292 293
static void do_release_stripe(struct r5conf *conf, struct stripe_head *sh,
			      struct list_head *temp_inactive_list)
L
Linus Torvalds 已提交
294
{
295 296 297 298
	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) &&
299
		    !test_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
300
			list_add_tail(&sh->lru, &conf->delayed_list);
301
		else if (test_bit(STRIPE_BIT_DELAY, &sh->state) &&
302 303 304 305 306
			   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);
307 308 309 310 311 312
			if (conf->worker_cnt_per_group == 0) {
				list_add_tail(&sh->lru, &conf->handle_list);
			} else {
				raid5_wakeup_stripe_thread(sh);
				return;
			}
313 314 315 316 317 318 319 320 321
		}
		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);
322 323
		if (!test_bit(STRIPE_EXPANDING, &sh->state))
			list_add_tail(&sh->lru, temp_inactive_list);
L
Linus Torvalds 已提交
324 325
	}
}
326

327 328
static void __release_stripe(struct r5conf *conf, struct stripe_head *sh,
			     struct list_head *temp_inactive_list)
329 330
{
	if (atomic_dec_and_test(&sh->count))
331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362
		do_release_stripe(conf, sh, temp_inactive_list);
}

/*
 * @hash could be NR_STRIPE_HASH_LOCKS, then we have a list of inactive_list
 *
 * Be careful: Only one task can add/delete stripes from temp_inactive_list at
 * given time. Adding stripes only takes device lock, while deleting stripes
 * only takes hash lock.
 */
static void release_inactive_stripe_list(struct r5conf *conf,
					 struct list_head *temp_inactive_list,
					 int hash)
{
	int size;
	bool do_wakeup = false;
	unsigned long flags;

	if (hash == NR_STRIPE_HASH_LOCKS) {
		size = NR_STRIPE_HASH_LOCKS;
		hash = NR_STRIPE_HASH_LOCKS - 1;
	} else
		size = 1;
	while (size) {
		struct list_head *list = &temp_inactive_list[size - 1];

		/*
		 * We don't hold any lock here yet, get_active_stripe() might
		 * remove stripes from the list
		 */
		if (!list_empty_careful(list)) {
			spin_lock_irqsave(conf->hash_locks + hash, flags);
363 364 365
			if (list_empty(conf->inactive_list + hash) &&
			    !list_empty(list))
				atomic_dec(&conf->empty_inactive_list_nr);
366 367 368 369 370 371 372 373 374 375 376 377 378
			list_splice_tail_init(list, conf->inactive_list + hash);
			do_wakeup = true;
			spin_unlock_irqrestore(conf->hash_locks + hash, flags);
		}
		size--;
		hash--;
	}

	if (do_wakeup) {
		wake_up(&conf->wait_for_stripe);
		if (conf->retry_read_aligned)
			md_wakeup_thread(conf->mddev->thread);
	}
379 380
}

S
Shaohua Li 已提交
381
/* should hold conf->device_lock already */
382 383
static int release_stripe_list(struct r5conf *conf,
			       struct list_head *temp_inactive_list)
S
Shaohua Li 已提交
384 385 386 387 388 389
{
	struct stripe_head *sh;
	int count = 0;
	struct llist_node *head;

	head = llist_del_all(&conf->released_stripes);
S
Shaohua Li 已提交
390
	head = llist_reverse_order(head);
S
Shaohua Li 已提交
391
	while (head) {
392 393
		int hash;

S
Shaohua Li 已提交
394 395 396 397 398 399 400 401 402 403
		sh = llist_entry(head, struct stripe_head, release_list);
		head = llist_next(head);
		/* sh could be readded after STRIPE_ON_RELEASE_LIST is cleard */
		smp_mb();
		clear_bit(STRIPE_ON_RELEASE_LIST, &sh->state);
		/*
		 * Don't worry the bit is set here, because if the bit is set
		 * again, the count is always > 1. This is true for
		 * STRIPE_ON_UNPLUG_LIST bit too.
		 */
404 405
		hash = sh->hash_lock_index;
		__release_stripe(conf, sh, &temp_inactive_list[hash]);
S
Shaohua Li 已提交
406 407 408 409 410 411
		count++;
	}

	return count;
}

L
Linus Torvalds 已提交
412 413
static void release_stripe(struct stripe_head *sh)
{
414
	struct r5conf *conf = sh->raid_conf;
L
Linus Torvalds 已提交
415
	unsigned long flags;
416 417
	struct list_head list;
	int hash;
S
Shaohua Li 已提交
418
	bool wakeup;
419

420 421 422 423 424
	/* Avoid release_list until the last reference.
	 */
	if (atomic_add_unless(&sh->count, -1, 1))
		return;

425 426
	if (unlikely(!conf->mddev->thread) ||
		test_and_set_bit(STRIPE_ON_RELEASE_LIST, &sh->state))
S
Shaohua Li 已提交
427 428 429 430 431 432
		goto slow_path;
	wakeup = llist_add(&sh->release_list, &conf->released_stripes);
	if (wakeup)
		md_wakeup_thread(conf->mddev->thread);
	return;
slow_path:
433
	local_irq_save(flags);
S
Shaohua Li 已提交
434
	/* we are ok here if STRIPE_ON_RELEASE_LIST is set or not */
435
	if (atomic_dec_and_lock(&sh->count, &conf->device_lock)) {
436 437 438
		INIT_LIST_HEAD(&list);
		hash = sh->hash_lock_index;
		do_release_stripe(conf, sh, &list);
439
		spin_unlock(&conf->device_lock);
440
		release_inactive_stripe_list(conf, &list, hash);
441 442
	}
	local_irq_restore(flags);
L
Linus Torvalds 已提交
443 444
}

445
static inline void remove_hash(struct stripe_head *sh)
L
Linus Torvalds 已提交
446
{
447 448
	pr_debug("remove_hash(), stripe %llu\n",
		(unsigned long long)sh->sector);
L
Linus Torvalds 已提交
449

450
	hlist_del_init(&sh->hash);
L
Linus Torvalds 已提交
451 452
}

453
static inline void insert_hash(struct r5conf *conf, struct stripe_head *sh)
L
Linus Torvalds 已提交
454
{
455
	struct hlist_head *hp = stripe_hash(conf, sh->sector);
L
Linus Torvalds 已提交
456

457 458
	pr_debug("insert_hash(), stripe %llu\n",
		(unsigned long long)sh->sector);
L
Linus Torvalds 已提交
459

460
	hlist_add_head(&sh->hash, hp);
L
Linus Torvalds 已提交
461 462 463
}

/* find an idle stripe, make sure it is unhashed, and return it. */
464
static struct stripe_head *get_free_stripe(struct r5conf *conf, int hash)
L
Linus Torvalds 已提交
465 466 467 468
{
	struct stripe_head *sh = NULL;
	struct list_head *first;

469
	if (list_empty(conf->inactive_list + hash))
L
Linus Torvalds 已提交
470
		goto out;
471
	first = (conf->inactive_list + hash)->next;
L
Linus Torvalds 已提交
472 473 474 475
	sh = list_entry(first, struct stripe_head, lru);
	list_del_init(first);
	remove_hash(sh);
	atomic_inc(&conf->active_stripes);
476
	BUG_ON(hash != sh->hash_lock_index);
477 478
	if (list_empty(conf->inactive_list + hash))
		atomic_inc(&conf->empty_inactive_list_nr);
L
Linus Torvalds 已提交
479 480 481 482
out:
	return sh;
}

483
static void shrink_buffers(struct stripe_head *sh)
L
Linus Torvalds 已提交
484 485 486
{
	struct page *p;
	int i;
487
	int num = sh->raid_conf->pool_size;
L
Linus Torvalds 已提交
488

489
	for (i = 0; i < num ; i++) {
490
		WARN_ON(sh->dev[i].page != sh->dev[i].orig_page);
L
Linus Torvalds 已提交
491 492 493 494
		p = sh->dev[i].page;
		if (!p)
			continue;
		sh->dev[i].page = NULL;
495
		put_page(p);
L
Linus Torvalds 已提交
496 497 498
	}
}

499
static int grow_buffers(struct stripe_head *sh)
L
Linus Torvalds 已提交
500 501
{
	int i;
502
	int num = sh->raid_conf->pool_size;
L
Linus Torvalds 已提交
503

504
	for (i = 0; i < num; i++) {
L
Linus Torvalds 已提交
505 506 507 508 509 510
		struct page *page;

		if (!(page = alloc_page(GFP_KERNEL))) {
			return 1;
		}
		sh->dev[i].page = page;
511
		sh->dev[i].orig_page = page;
L
Linus Torvalds 已提交
512 513 514 515
	}
	return 0;
}

516
static void raid5_build_block(struct stripe_head *sh, int i, int previous);
517
static void stripe_set_idx(sector_t stripe, struct r5conf *conf, int previous,
518
			    struct stripe_head *sh);
L
Linus Torvalds 已提交
519

520
static void init_stripe(struct stripe_head *sh, sector_t sector, int previous)
L
Linus Torvalds 已提交
521
{
522
	struct r5conf *conf = sh->raid_conf;
523
	int i, seq;
L
Linus Torvalds 已提交
524

525 526
	BUG_ON(atomic_read(&sh->count) != 0);
	BUG_ON(test_bit(STRIPE_HANDLE, &sh->state));
527
	BUG_ON(stripe_operations_active(sh));
528

529
	pr_debug("init_stripe called, stripe %llu\n",
530
		(unsigned long long)sector);
531 532
retry:
	seq = read_seqcount_begin(&conf->gen_lock);
533
	sh->generation = conf->generation - previous;
534
	sh->disks = previous ? conf->previous_raid_disks : conf->raid_disks;
L
Linus Torvalds 已提交
535
	sh->sector = sector;
536
	stripe_set_idx(sector, conf, previous, sh);
L
Linus Torvalds 已提交
537 538
	sh->state = 0;

539
	for (i = sh->disks; i--; ) {
L
Linus Torvalds 已提交
540 541
		struct r5dev *dev = &sh->dev[i];

542
		if (dev->toread || dev->read || dev->towrite || dev->written ||
L
Linus Torvalds 已提交
543
		    test_bit(R5_LOCKED, &dev->flags)) {
544
			printk(KERN_ERR "sector=%llx i=%d %p %p %p %p %d\n",
L
Linus Torvalds 已提交
545
			       (unsigned long long)sh->sector, i, dev->toread,
546
			       dev->read, dev->towrite, dev->written,
L
Linus Torvalds 已提交
547
			       test_bit(R5_LOCKED, &dev->flags));
548
			WARN_ON(1);
L
Linus Torvalds 已提交
549 550
		}
		dev->flags = 0;
551
		raid5_build_block(sh, i, previous);
L
Linus Torvalds 已提交
552
	}
553 554
	if (read_seqcount_retry(&conf->gen_lock, seq))
		goto retry;
L
Linus Torvalds 已提交
555
	insert_hash(conf, sh);
556
	sh->cpu = smp_processor_id();
L
Linus Torvalds 已提交
557 558
}

559
static struct stripe_head *__find_stripe(struct r5conf *conf, sector_t sector,
560
					 short generation)
L
Linus Torvalds 已提交
561 562 563
{
	struct stripe_head *sh;

564
	pr_debug("__find_stripe, sector %llu\n", (unsigned long long)sector);
565
	hlist_for_each_entry(sh, stripe_hash(conf, sector), hash)
566
		if (sh->sector == sector && sh->generation == generation)
L
Linus Torvalds 已提交
567
			return sh;
568
	pr_debug("__stripe %llu not in cache\n", (unsigned long long)sector);
L
Linus Torvalds 已提交
569 570 571
	return NULL;
}

572 573 574 575 576 577 578 579 580 581 582 583 584
/*
 * 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.
 */
585
static int calc_degraded(struct r5conf *conf)
586
{
587
	int degraded, degraded2;
588 589 590 591 592
	int i;

	rcu_read_lock();
	degraded = 0;
	for (i = 0; i < conf->previous_raid_disks; i++) {
593
		struct md_rdev *rdev = rcu_dereference(conf->disks[i].rdev);
594 595
		if (rdev && test_bit(Faulty, &rdev->flags))
			rdev = rcu_dereference(conf->disks[i].replacement);
596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613
		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();
614 615
	if (conf->raid_disks == conf->previous_raid_disks)
		return degraded;
616
	rcu_read_lock();
617
	degraded2 = 0;
618
	for (i = 0; i < conf->raid_disks; i++) {
619
		struct md_rdev *rdev = rcu_dereference(conf->disks[i].rdev);
620 621
		if (rdev && test_bit(Faulty, &rdev->flags))
			rdev = rcu_dereference(conf->disks[i].replacement);
622
		if (!rdev || test_bit(Faulty, &rdev->flags))
623
			degraded2++;
624 625 626 627 628 629 630 631 632
		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)
633
				degraded2++;
634 635
	}
	rcu_read_unlock();
636 637 638 639 640 641 642 643 644 645 646 647 648
	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);
649 650 651 652 653
	if (degraded > conf->max_degraded)
		return 1;
	return 0;
}

654
static struct stripe_head *
655
get_active_stripe(struct r5conf *conf, sector_t sector,
656
		  int previous, int noblock, int noquiesce)
L
Linus Torvalds 已提交
657 658
{
	struct stripe_head *sh;
659
	int hash = stripe_hash_locks_hash(sector);
L
Linus Torvalds 已提交
660

661
	pr_debug("get_stripe, sector %llu\n", (unsigned long long)sector);
L
Linus Torvalds 已提交
662

663
	spin_lock_irq(conf->hash_locks + hash);
L
Linus Torvalds 已提交
664 665

	do {
666
		wait_event_lock_irq(conf->wait_for_stripe,
667
				    conf->quiesce == 0 || noquiesce,
668
				    *(conf->hash_locks + hash));
669
		sh = __find_stripe(conf, sector, conf->generation - previous);
L
Linus Torvalds 已提交
670 671
		if (!sh) {
			if (!conf->inactive_blocked)
672
				sh = get_free_stripe(conf, hash);
L
Linus Torvalds 已提交
673 674 675 676
			if (noblock && sh == NULL)
				break;
			if (!sh) {
				conf->inactive_blocked = 1;
677 678 679 680 681 682 683
				wait_event_lock_irq(
					conf->wait_for_stripe,
					!list_empty(conf->inactive_list + hash) &&
					(atomic_read(&conf->active_stripes)
					 < (conf->max_nr_stripes * 3 / 4)
					 || !conf->inactive_blocked),
					*(conf->hash_locks + hash));
L
Linus Torvalds 已提交
684
				conf->inactive_blocked = 0;
685
			} else {
686
				init_stripe(sh, sector, previous);
687 688
				atomic_inc(&sh->count);
			}
689
		} else if (!atomic_inc_not_zero(&sh->count)) {
690
			spin_lock(&conf->device_lock);
691
			if (!atomic_read(&sh->count)) {
L
Linus Torvalds 已提交
692 693
				if (!test_bit(STRIPE_HANDLE, &sh->state))
					atomic_inc(&conf->active_stripes);
694 695
				BUG_ON(list_empty(&sh->lru) &&
				       !test_bit(STRIPE_EXPANDING, &sh->state));
696
				list_del_init(&sh->lru);
697 698 699 700
				if (sh->group) {
					sh->group->stripes_cnt--;
					sh->group = NULL;
				}
L
Linus Torvalds 已提交
701
			}
702
			atomic_inc(&sh->count);
703
			spin_unlock(&conf->device_lock);
L
Linus Torvalds 已提交
704 705 706
		}
	} while (sh == NULL);

707
	spin_unlock_irq(conf->hash_locks + hash);
L
Linus Torvalds 已提交
708 709 710
	return sh;
}

711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731
/* 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;
}

732 733 734 735
static void
raid5_end_read_request(struct bio *bi, int error);
static void
raid5_end_write_request(struct bio *bi, int error);
736

737
static void ops_run_io(struct stripe_head *sh, struct stripe_head_state *s)
738
{
739
	struct r5conf *conf = sh->raid_conf;
740 741 742 743 744 745
	int i, disks = sh->disks;

	might_sleep();

	for (i = disks; i--; ) {
		int rw;
746
		int replace_only = 0;
747 748
		struct bio *bi, *rbi;
		struct md_rdev *rdev, *rrdev = NULL;
T
Tejun Heo 已提交
749 750 751 752 753
		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;
754
			if (test_bit(R5_Discard, &sh->dev[i].flags))
S
Shaohua Li 已提交
755
				rw |= REQ_DISCARD;
T
Tejun Heo 已提交
756
		} else if (test_and_clear_bit(R5_Wantread, &sh->dev[i].flags))
757
			rw = READ;
758 759 760 761 762
		else if (test_and_clear_bit(R5_WantReplace,
					    &sh->dev[i].flags)) {
			rw = WRITE;
			replace_only = 1;
		} else
763
			continue;
S
Shaohua Li 已提交
764 765
		if (test_and_clear_bit(R5_SyncIO, &sh->dev[i].flags))
			rw |= REQ_SYNC;
766 767

		bi = &sh->dev[i].req;
768
		rbi = &sh->dev[i].rreq; /* For writing to replacement */
769 770

		rcu_read_lock();
771
		rrdev = rcu_dereference(conf->disks[i].replacement);
772 773 774 775 776 777
		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;
		}
778 779 780
		if (rw & WRITE) {
			if (replace_only)
				rdev = NULL;
781 782 783
			if (rdev == rrdev)
				/* We raced and saw duplicates */
				rrdev = NULL;
784
		} else {
785
			if (test_bit(R5_ReadRepl, &sh->dev[i].flags) && rrdev)
786 787 788
				rdev = rrdev;
			rrdev = NULL;
		}
789

790 791 792 793
		if (rdev && test_bit(Faulty, &rdev->flags))
			rdev = NULL;
		if (rdev)
			atomic_inc(&rdev->nr_pending);
794 795 796 797
		if (rrdev && test_bit(Faulty, &rrdev->flags))
			rrdev = NULL;
		if (rrdev)
			atomic_inc(&rrdev->nr_pending);
798 799
		rcu_read_unlock();

800
		/* We have already checked bad blocks for reads.  Now
801 802
		 * need to check for writes.  We never accept write errors
		 * on the replacement, so we don't to check rrdev.
803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822
		 */
		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);
				}
823 824 825 826 827 828
				/*
				 * Because md_wait_for_blocked_rdev
				 * will dec nr_pending, we must
				 * increment it first.
				 */
				atomic_inc(&rdev->nr_pending);
829 830 831 832 833 834 835 836
				md_wait_for_blocked_rdev(rdev, conf->mddev);
			} else {
				/* Acknowledged bad block - skip the write */
				rdev_dec_pending(rdev, conf->mddev);
				rdev = NULL;
			}
		}

837
		if (rdev) {
838 839
			if (s->syncing || s->expanding || s->expanded
			    || s->replacing)
840 841
				md_sync_acct(rdev->bdev, STRIPE_SECTORS);

D
Dan Williams 已提交
842 843
			set_bit(STRIPE_IO_STARTED, &sh->state);

K
Kent Overstreet 已提交
844
			bio_reset(bi);
845
			bi->bi_bdev = rdev->bdev;
K
Kent Overstreet 已提交
846 847 848 849 850 851
			bi->bi_rw = rw;
			bi->bi_end_io = (rw & WRITE)
				? raid5_end_write_request
				: raid5_end_read_request;
			bi->bi_private = sh;

852
			pr_debug("%s: for %llu schedule op %ld on disc %d\n",
853
				__func__, (unsigned long long)sh->sector,
854 855
				bi->bi_rw, i);
			atomic_inc(&sh->count);
856
			if (use_new_offset(conf, sh))
857
				bi->bi_iter.bi_sector = (sh->sector
858 859
						 + rdev->new_data_offset);
			else
860
				bi->bi_iter.bi_sector = (sh->sector
861
						 + rdev->data_offset);
862
			if (test_bit(R5_ReadNoMerge, &sh->dev[i].flags))
863
				bi->bi_rw |= REQ_NOMERGE;
864

865 866 867
			if (test_bit(R5_SkipCopy, &sh->dev[i].flags))
				WARN_ON(test_bit(R5_UPTODATE, &sh->dev[i].flags));
			sh->dev[i].vec.bv_page = sh->dev[i].page;
K
Kent Overstreet 已提交
868
			bi->bi_vcnt = 1;
869 870
			bi->bi_io_vec[0].bv_len = STRIPE_SIZE;
			bi->bi_io_vec[0].bv_offset = 0;
871
			bi->bi_iter.bi_size = STRIPE_SIZE;
872 873 874 875 876 877
			/*
			 * If this is discard request, set bi_vcnt 0. We don't
			 * want to confuse SCSI because SCSI will replace payload
			 */
			if (rw & REQ_DISCARD)
				bi->bi_vcnt = 0;
878 879
			if (rrdev)
				set_bit(R5_DOUBLE_LOCKED, &sh->dev[i].flags);
880 881 882 883 884

			if (conf->mddev->gendisk)
				trace_block_bio_remap(bdev_get_queue(bi->bi_bdev),
						      bi, disk_devt(conf->mddev->gendisk),
						      sh->dev[i].sector);
885
			generic_make_request(bi);
886 887
		}
		if (rrdev) {
888 889
			if (s->syncing || s->expanding || s->expanded
			    || s->replacing)
890 891 892 893
				md_sync_acct(rrdev->bdev, STRIPE_SECTORS);

			set_bit(STRIPE_IO_STARTED, &sh->state);

K
Kent Overstreet 已提交
894
			bio_reset(rbi);
895
			rbi->bi_bdev = rrdev->bdev;
K
Kent Overstreet 已提交
896 897 898 899 900
			rbi->bi_rw = rw;
			BUG_ON(!(rw & WRITE));
			rbi->bi_end_io = raid5_end_write_request;
			rbi->bi_private = sh;

901 902 903 904 905
			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);
906
			if (use_new_offset(conf, sh))
907
				rbi->bi_iter.bi_sector = (sh->sector
908 909
						  + rrdev->new_data_offset);
			else
910
				rbi->bi_iter.bi_sector = (sh->sector
911
						  + rrdev->data_offset);
912 913 914
			if (test_bit(R5_SkipCopy, &sh->dev[i].flags))
				WARN_ON(test_bit(R5_UPTODATE, &sh->dev[i].flags));
			sh->dev[i].rvec.bv_page = sh->dev[i].page;
K
Kent Overstreet 已提交
915
			rbi->bi_vcnt = 1;
916 917
			rbi->bi_io_vec[0].bv_len = STRIPE_SIZE;
			rbi->bi_io_vec[0].bv_offset = 0;
918
			rbi->bi_iter.bi_size = STRIPE_SIZE;
919 920 921 922 923 924
			/*
			 * If this is discard request, set bi_vcnt 0. We don't
			 * want to confuse SCSI because SCSI will replace payload
			 */
			if (rw & REQ_DISCARD)
				rbi->bi_vcnt = 0;
925 926 927 928
			if (conf->mddev->gendisk)
				trace_block_bio_remap(bdev_get_queue(rbi->bi_bdev),
						      rbi, disk_devt(conf->mddev->gendisk),
						      sh->dev[i].sector);
929 930 931
			generic_make_request(rbi);
		}
		if (!rdev && !rrdev) {
932
			if (rw & WRITE)
933 934 935 936 937 938 939 940 941 942
				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 *
943 944 945
async_copy_data(int frombio, struct bio *bio, struct page **page,
	sector_t sector, struct dma_async_tx_descriptor *tx,
	struct stripe_head *sh)
946
{
947 948
	struct bio_vec bvl;
	struct bvec_iter iter;
949 950
	struct page *bio_page;
	int page_offset;
951
	struct async_submit_ctl submit;
D
Dan Williams 已提交
952
	enum async_tx_flags flags = 0;
953

954 955
	if (bio->bi_iter.bi_sector >= sector)
		page_offset = (signed)(bio->bi_iter.bi_sector - sector) * 512;
956
	else
957
		page_offset = (signed)(sector - bio->bi_iter.bi_sector) * -512;
958

D
Dan Williams 已提交
959 960 961 962
	if (frombio)
		flags |= ASYNC_TX_FENCE;
	init_async_submit(&submit, flags, tx, NULL, NULL, NULL);

963 964
	bio_for_each_segment(bvl, bio, iter) {
		int len = bvl.bv_len;
965 966 967 968 969 970 971 972 973 974 975 976 977 978 979
		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) {
980 981
			b_offset += bvl.bv_offset;
			bio_page = bvl.bv_page;
982 983 984 985 986 987 988
			if (frombio) {
				if (sh->raid_conf->skip_copy &&
				    b_offset == 0 && page_offset == 0 &&
				    clen == STRIPE_SIZE)
					*page = bio_page;
				else
					tx = async_memcpy(*page, bio_page, page_offset,
989
						  b_offset, clen, &submit);
990 991
			} else
				tx = async_memcpy(bio_page, *page, b_offset,
992
						  page_offset, clen, &submit);
993
		}
994 995 996
		/* chain the operations */
		submit.depend_tx = tx;

997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008
		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;
1009
	int i;
1010

1011
	pr_debug("%s: stripe %llu\n", __func__,
1012 1013 1014 1015 1016 1017 1018
		(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 */
1019 1020
		/* and check if we need to reply to a read request,
		 * new R5_Wantfill requests are held off until
1021
		 * !STRIPE_BIOFILL_RUN
1022 1023
		 */
		if (test_and_clear_bit(R5_Wantfill, &dev->flags)) {
1024 1025 1026 1027 1028
			struct bio *rbi, *rbi2;

			BUG_ON(!dev->read);
			rbi = dev->read;
			dev->read = NULL;
1029
			while (rbi && rbi->bi_iter.bi_sector <
1030 1031
				dev->sector + STRIPE_SECTORS) {
				rbi2 = r5_next_bio(rbi, dev->sector);
1032
				if (!raid5_dec_bi_active_stripes(rbi)) {
1033 1034 1035 1036 1037 1038 1039
					rbi->bi_next = return_bi;
					return_bi = rbi;
				}
				rbi = rbi2;
			}
		}
	}
1040
	clear_bit(STRIPE_BIOFILL_RUN, &sh->state);
1041 1042 1043

	return_io(return_bi);

1044
	set_bit(STRIPE_HANDLE, &sh->state);
1045 1046 1047 1048 1049 1050
	release_stripe(sh);
}

static void ops_run_biofill(struct stripe_head *sh)
{
	struct dma_async_tx_descriptor *tx = NULL;
1051
	struct async_submit_ctl submit;
1052 1053
	int i;

1054
	pr_debug("%s: stripe %llu\n", __func__,
1055 1056 1057 1058 1059 1060
		(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
1061
			spin_lock_irq(&sh->stripe_lock);
1062 1063
			dev->read = rbi = dev->toread;
			dev->toread = NULL;
S
Shaohua Li 已提交
1064
			spin_unlock_irq(&sh->stripe_lock);
1065
			while (rbi && rbi->bi_iter.bi_sector <
1066
				dev->sector + STRIPE_SECTORS) {
1067 1068
				tx = async_copy_data(0, rbi, &dev->page,
					dev->sector, tx, sh);
1069 1070 1071 1072 1073 1074
				rbi = r5_next_bio(rbi, dev->sector);
			}
		}
	}

	atomic_inc(&sh->count);
1075 1076
	init_async_submit(&submit, ASYNC_TX_ACK, tx, ops_complete_biofill, sh, NULL);
	async_trigger_callback(&submit);
1077 1078
}

1079
static void mark_target_uptodate(struct stripe_head *sh, int target)
1080
{
1081
	struct r5dev *tgt;
1082

1083 1084
	if (target < 0)
		return;
1085

1086
	tgt = &sh->dev[target];
1087 1088 1089
	set_bit(R5_UPTODATE, &tgt->flags);
	BUG_ON(!test_bit(R5_Wantcompute, &tgt->flags));
	clear_bit(R5_Wantcompute, &tgt->flags);
1090 1091
}

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

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

1099
	/* mark the computed target(s) as uptodate */
1100
	mark_target_uptodate(sh, sh->ops.target);
1101
	mark_target_uptodate(sh, sh->ops.target2);
1102

1103 1104 1105
	clear_bit(STRIPE_COMPUTE_RUN, &sh->state);
	if (sh->check_state == check_state_compute_run)
		sh->check_state = check_state_compute_result;
1106 1107 1108 1109
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

1110 1111 1112 1113 1114 1115 1116 1117 1118
/* 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)
1119 1120
{
	int disks = sh->disks;
1121
	struct page **xor_srcs = percpu->scribble;
1122 1123 1124 1125 1126
	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;
1127
	struct async_submit_ctl submit;
1128 1129 1130
	int i;

	pr_debug("%s: stripe %llu block: %d\n",
1131
		__func__, (unsigned long long)sh->sector, target);
1132 1133 1134 1135 1136 1137 1138 1139
	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
1140
	init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST, NULL,
1141
			  ops_complete_compute, sh, to_addr_conv(sh, percpu));
1142
	if (unlikely(count == 1))
1143
		tx = async_memcpy(xor_dest, xor_srcs[0], 0, 0, STRIPE_SIZE, &submit);
1144
	else
1145
		tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, &submit);
1146 1147 1148 1149

	return tx;
}

1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167
/* 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++)
1168
		srcs[i] = NULL;
1169 1170 1171 1172 1173 1174 1175 1176 1177 1178

	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);

1179
	return syndrome_disks;
1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199
}

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;
1200
	else
1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216
		/* 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
1217 1218
		init_async_submit(&submit, ASYNC_TX_FENCE, NULL,
				  ops_complete_compute, sh,
1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229
				  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
1230 1231
		init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST,
				  NULL, ops_complete_compute, sh,
1232 1233 1234
				  to_addr_conv(sh, percpu));
		tx = async_xor(dest, blocks, 0, count, STRIPE_SIZE, &submit);
	}
1235 1236 1237 1238

	return tx;
}

1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259
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));

1260
	/* we need to open-code set_syndrome_sources to handle the
1261 1262 1263
	 * slot number conversion for 'faila' and 'failb'
	 */
	for (i = 0; i < disks ; i++)
1264
		blocks[i] = NULL;
1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290
	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
1291 1292 1293
			init_async_submit(&submit, ASYNC_TX_FENCE, NULL,
					  ops_complete_compute, sh,
					  to_addr_conv(sh, percpu));
1294
			return async_gen_syndrome(blocks, 0, syndrome_disks+2,
1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313
						  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
1314 1315 1316 1317
			init_async_submit(&submit,
					  ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST,
					  NULL, NULL, NULL,
					  to_addr_conv(sh, percpu));
1318 1319 1320 1321
			tx = async_xor(dest, blocks, 0, count, STRIPE_SIZE,
				       &submit);

			count = set_syndrome_sources(blocks, sh);
D
Dan Williams 已提交
1322 1323 1324
			init_async_submit(&submit, ASYNC_TX_FENCE, tx,
					  ops_complete_compute, sh,
					  to_addr_conv(sh, percpu));
1325 1326 1327 1328
			return async_gen_syndrome(blocks, 0, count+2,
						  STRIPE_SIZE, &submit);
		}
	} else {
1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342
		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);
		}
1343 1344 1345
	}
}

1346 1347 1348 1349
static void ops_complete_prexor(void *stripe_head_ref)
{
	struct stripe_head *sh = stripe_head_ref;

1350
	pr_debug("%s: stripe %llu\n", __func__,
1351 1352 1353 1354
		(unsigned long long)sh->sector);
}

static struct dma_async_tx_descriptor *
1355 1356
ops_run_prexor(struct stripe_head *sh, struct raid5_percpu *percpu,
	       struct dma_async_tx_descriptor *tx)
1357 1358
{
	int disks = sh->disks;
1359
	struct page **xor_srcs = percpu->scribble;
1360
	int count = 0, pd_idx = sh->pd_idx, i;
1361
	struct async_submit_ctl submit;
1362 1363 1364 1365

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

1366
	pr_debug("%s: stripe %llu\n", __func__,
1367 1368 1369 1370 1371
		(unsigned long long)sh->sector);

	for (i = disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
		/* Only process blocks that are known to be uptodate */
1372
		if (test_bit(R5_Wantdrain, &dev->flags))
1373 1374 1375
			xor_srcs[count++] = dev->page;
	}

D
Dan Williams 已提交
1376
	init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_DROP_DST, tx,
1377
			  ops_complete_prexor, sh, to_addr_conv(sh, percpu));
1378
	tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, &submit);
1379 1380 1381 1382 1383

	return tx;
}

static struct dma_async_tx_descriptor *
1384
ops_run_biodrain(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
1385 1386
{
	int disks = sh->disks;
1387
	int i;
1388

1389
	pr_debug("%s: stripe %llu\n", __func__,
1390 1391 1392 1393 1394 1395
		(unsigned long long)sh->sector);

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

1396
		if (test_and_clear_bit(R5_Wantdrain, &dev->flags)) {
1397 1398
			struct bio *wbi;

S
Shaohua Li 已提交
1399
			spin_lock_irq(&sh->stripe_lock);
1400 1401 1402 1403
			chosen = dev->towrite;
			dev->towrite = NULL;
			BUG_ON(dev->written);
			wbi = dev->written = chosen;
S
Shaohua Li 已提交
1404
			spin_unlock_irq(&sh->stripe_lock);
1405
			WARN_ON(dev->page != dev->orig_page);
1406

1407
			while (wbi && wbi->bi_iter.bi_sector <
1408
				dev->sector + STRIPE_SECTORS) {
T
Tejun Heo 已提交
1409 1410
				if (wbi->bi_rw & REQ_FUA)
					set_bit(R5_WantFUA, &dev->flags);
S
Shaohua Li 已提交
1411 1412
				if (wbi->bi_rw & REQ_SYNC)
					set_bit(R5_SyncIO, &dev->flags);
1413
				if (wbi->bi_rw & REQ_DISCARD)
S
Shaohua Li 已提交
1414
					set_bit(R5_Discard, &dev->flags);
1415 1416 1417 1418 1419 1420 1421 1422 1423
				else {
					tx = async_copy_data(1, wbi, &dev->page,
						dev->sector, tx, sh);
					if (dev->page != dev->orig_page) {
						set_bit(R5_SkipCopy, &dev->flags);
						clear_bit(R5_UPTODATE, &dev->flags);
						clear_bit(R5_OVERWRITE, &dev->flags);
					}
				}
1424 1425 1426 1427 1428 1429 1430 1431
				wbi = r5_next_bio(wbi, dev->sector);
			}
		}
	}

	return tx;
}

1432
static void ops_complete_reconstruct(void *stripe_head_ref)
1433 1434
{
	struct stripe_head *sh = stripe_head_ref;
1435 1436 1437 1438
	int disks = sh->disks;
	int pd_idx = sh->pd_idx;
	int qd_idx = sh->qd_idx;
	int i;
1439
	bool fua = false, sync = false, discard = false;
1440

1441
	pr_debug("%s: stripe %llu\n", __func__,
1442 1443
		(unsigned long long)sh->sector);

S
Shaohua Li 已提交
1444
	for (i = disks; i--; ) {
T
Tejun Heo 已提交
1445
		fua |= test_bit(R5_WantFUA, &sh->dev[i].flags);
S
Shaohua Li 已提交
1446
		sync |= test_bit(R5_SyncIO, &sh->dev[i].flags);
1447
		discard |= test_bit(R5_Discard, &sh->dev[i].flags);
S
Shaohua Li 已提交
1448
	}
T
Tejun Heo 已提交
1449

1450 1451
	for (i = disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
1452

T
Tejun Heo 已提交
1453
		if (dev->written || i == pd_idx || i == qd_idx) {
1454
			if (!discard && !test_bit(R5_SkipCopy, &dev->flags))
1455
				set_bit(R5_UPTODATE, &dev->flags);
T
Tejun Heo 已提交
1456 1457
			if (fua)
				set_bit(R5_WantFUA, &dev->flags);
S
Shaohua Li 已提交
1458 1459
			if (sync)
				set_bit(R5_SyncIO, &dev->flags);
T
Tejun Heo 已提交
1460
		}
1461 1462
	}

1463 1464 1465 1466 1467 1468 1469 1470
	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;
	}
1471 1472 1473 1474 1475 1476

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

static void
1477 1478
ops_run_reconstruct5(struct stripe_head *sh, struct raid5_percpu *percpu,
		     struct dma_async_tx_descriptor *tx)
1479 1480
{
	int disks = sh->disks;
1481
	struct page **xor_srcs = percpu->scribble;
1482
	struct async_submit_ctl submit;
1483 1484
	int count = 0, pd_idx = sh->pd_idx, i;
	struct page *xor_dest;
1485
	int prexor = 0;
1486 1487
	unsigned long flags;

1488
	pr_debug("%s: stripe %llu\n", __func__,
1489 1490
		(unsigned long long)sh->sector);

S
Shaohua Li 已提交
1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502
	for (i = 0; i < sh->disks; i++) {
		if (pd_idx == i)
			continue;
		if (!test_bit(R5_Discard, &sh->dev[i].flags))
			break;
	}
	if (i >= sh->disks) {
		atomic_inc(&sh->count);
		set_bit(R5_Discard, &sh->dev[pd_idx].flags);
		ops_complete_reconstruct(sh);
		return;
	}
1503 1504 1505
	/* check if prexor is active which means only process blocks
	 * that are part of a read-modify-write (written)
	 */
1506 1507
	if (sh->reconstruct_state == reconstruct_state_prexor_drain_run) {
		prexor = 1;
1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527
		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
	 */
1528
	flags = ASYNC_TX_ACK |
1529 1530 1531 1532
		(prexor ? ASYNC_TX_XOR_DROP_DST : ASYNC_TX_XOR_ZERO_DST);

	atomic_inc(&sh->count);

1533
	init_async_submit(&submit, flags, tx, ops_complete_reconstruct, sh,
1534
			  to_addr_conv(sh, percpu));
1535 1536 1537 1538
	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);
1539 1540
}

1541 1542 1543 1544 1545 1546
static void
ops_run_reconstruct6(struct stripe_head *sh, struct raid5_percpu *percpu,
		     struct dma_async_tx_descriptor *tx)
{
	struct async_submit_ctl submit;
	struct page **blocks = percpu->scribble;
S
Shaohua Li 已提交
1547
	int count, i;
1548 1549 1550

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

S
Shaohua Li 已提交
1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564
	for (i = 0; i < sh->disks; i++) {
		if (sh->pd_idx == i || sh->qd_idx == i)
			continue;
		if (!test_bit(R5_Discard, &sh->dev[i].flags))
			break;
	}
	if (i >= sh->disks) {
		atomic_inc(&sh->count);
		set_bit(R5_Discard, &sh->dev[sh->pd_idx].flags);
		set_bit(R5_Discard, &sh->dev[sh->qd_idx].flags);
		ops_complete_reconstruct(sh);
		return;
	}

1565 1566 1567 1568 1569 1570 1571
	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);
1572 1573 1574 1575 1576 1577
}

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

1578
	pr_debug("%s: stripe %llu\n", __func__,
1579 1580
		(unsigned long long)sh->sector);

1581
	sh->check_state = check_state_check_result;
1582 1583 1584 1585
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

1586
static void ops_run_check_p(struct stripe_head *sh, struct raid5_percpu *percpu)
1587 1588
{
	int disks = sh->disks;
1589 1590 1591
	int pd_idx = sh->pd_idx;
	int qd_idx = sh->qd_idx;
	struct page *xor_dest;
1592
	struct page **xor_srcs = percpu->scribble;
1593
	struct dma_async_tx_descriptor *tx;
1594
	struct async_submit_ctl submit;
1595 1596
	int count;
	int i;
1597

1598
	pr_debug("%s: stripe %llu\n", __func__,
1599 1600
		(unsigned long long)sh->sector);

1601 1602 1603
	count = 0;
	xor_dest = sh->dev[pd_idx].page;
	xor_srcs[count++] = xor_dest;
1604
	for (i = disks; i--; ) {
1605 1606 1607
		if (i == pd_idx || i == qd_idx)
			continue;
		xor_srcs[count++] = sh->dev[i].page;
1608 1609
	}

1610 1611
	init_async_submit(&submit, 0, NULL, NULL, NULL,
			  to_addr_conv(sh, percpu));
D
Dan Williams 已提交
1612
	tx = async_xor_val(xor_dest, xor_srcs, 0, count, STRIPE_SIZE,
1613
			   &sh->ops.zero_sum_result, &submit);
1614 1615

	atomic_inc(&sh->count);
1616 1617
	init_async_submit(&submit, ASYNC_TX_ACK, tx, ops_complete_check, sh, NULL);
	tx = async_trigger_callback(&submit);
1618 1619
}

1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631
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;
1632 1633

	atomic_inc(&sh->count);
1634 1635 1636 1637
	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);
1638 1639
}

N
NeilBrown 已提交
1640
static void raid_run_ops(struct stripe_head *sh, unsigned long ops_request)
1641 1642 1643
{
	int overlap_clear = 0, i, disks = sh->disks;
	struct dma_async_tx_descriptor *tx = NULL;
1644
	struct r5conf *conf = sh->raid_conf;
1645
	int level = conf->level;
1646 1647
	struct raid5_percpu *percpu;
	unsigned long cpu;
1648

1649 1650
	cpu = get_cpu();
	percpu = per_cpu_ptr(conf->percpu, cpu);
1651
	if (test_bit(STRIPE_OP_BIOFILL, &ops_request)) {
1652 1653 1654 1655
		ops_run_biofill(sh);
		overlap_clear++;
	}

1656
	if (test_bit(STRIPE_OP_COMPUTE_BLK, &ops_request)) {
1657 1658 1659 1660 1661 1662 1663 1664 1665 1666
		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))
1667 1668
			async_tx_ack(tx);
	}
1669

1670
	if (test_bit(STRIPE_OP_PREXOR, &ops_request))
1671
		tx = ops_run_prexor(sh, percpu, tx);
1672

1673
	if (test_bit(STRIPE_OP_BIODRAIN, &ops_request)) {
1674
		tx = ops_run_biodrain(sh, tx);
1675 1676 1677
		overlap_clear++;
	}

1678 1679 1680 1681 1682 1683
	if (test_bit(STRIPE_OP_RECONSTRUCT, &ops_request)) {
		if (level < 6)
			ops_run_reconstruct5(sh, percpu, tx);
		else
			ops_run_reconstruct6(sh, percpu, tx);
	}
1684

1685 1686 1687 1688 1689 1690 1691 1692 1693 1694
	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();
	}
1695 1696 1697 1698 1699 1700 1701

	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);
		}
1702
	put_cpu();
1703 1704
}

1705
static int grow_one_stripe(struct r5conf *conf, int hash)
L
Linus Torvalds 已提交
1706 1707
{
	struct stripe_head *sh;
N
Namhyung Kim 已提交
1708
	sh = kmem_cache_zalloc(conf->slab_cache, GFP_KERNEL);
1709 1710
	if (!sh)
		return 0;
N
Namhyung Kim 已提交
1711

1712 1713
	sh->raid_conf = conf;

S
Shaohua Li 已提交
1714 1715
	spin_lock_init(&sh->stripe_lock);

1716 1717
	if (grow_buffers(sh)) {
		shrink_buffers(sh);
1718 1719 1720
		kmem_cache_free(conf->slab_cache, sh);
		return 0;
	}
1721
	sh->hash_lock_index = hash;
1722 1723 1724 1725 1726 1727 1728 1729
	/* 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;
}

1730
static int grow_stripes(struct r5conf *conf, int num)
1731
{
1732
	struct kmem_cache *sc;
1733
	int devs = max(conf->raid_disks, conf->previous_raid_disks);
1734
	int hash;
L
Linus Torvalds 已提交
1735

1736 1737 1738 1739 1740 1741 1742 1743
	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]);

1744 1745
	conf->active_name = 0;
	sc = kmem_cache_create(conf->cache_name[conf->active_name],
L
Linus Torvalds 已提交
1746
			       sizeof(struct stripe_head)+(devs-1)*sizeof(struct r5dev),
1747
			       0, 0, NULL);
L
Linus Torvalds 已提交
1748 1749 1750
	if (!sc)
		return 1;
	conf->slab_cache = sc;
1751
	conf->pool_size = devs;
1752 1753 1754
	hash = conf->max_nr_stripes % NR_STRIPE_HASH_LOCKS;
	while (num--) {
		if (!grow_one_stripe(conf, hash))
L
Linus Torvalds 已提交
1755
			return 1;
1756 1757 1758
		conf->max_nr_stripes++;
		hash = (hash + 1) % NR_STRIPE_HASH_LOCKS;
	}
L
Linus Torvalds 已提交
1759 1760
	return 0;
}
1761

1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783
/**
 * 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;
}

1784
static int resize_stripes(struct r5conf *conf, int newsize)
1785 1786 1787 1788 1789 1790 1791
{
	/* Make all the stripes able to hold 'newsize' devices.
	 * New slots in each stripe get 'page' set to a new page.
	 *
	 * This happens in stages:
	 * 1/ create a new kmem_cache and allocate the required number of
	 *    stripe_heads.
M
Masanari Iida 已提交
1792
	 * 2/ gather all the old stripe_heads and transfer the pages across
1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811
	 *    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;
1812
	unsigned long cpu;
1813
	int err;
1814
	struct kmem_cache *sc;
1815
	int i;
1816
	int hash, cnt;
1817 1818 1819 1820

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

1821 1822 1823
	err = md_allow_write(conf->mddev);
	if (err)
		return err;
1824

1825 1826 1827
	/* Step 1 */
	sc = kmem_cache_create(conf->cache_name[1-conf->active_name],
			       sizeof(struct stripe_head)+(newsize-1)*sizeof(struct r5dev),
1828
			       0, 0, NULL);
1829 1830 1831 1832
	if (!sc)
		return -ENOMEM;

	for (i = conf->max_nr_stripes; i; i--) {
N
Namhyung Kim 已提交
1833
		nsh = kmem_cache_zalloc(sc, GFP_KERNEL);
1834 1835 1836 1837
		if (!nsh)
			break;

		nsh->raid_conf = conf;
1838
		spin_lock_init(&nsh->stripe_lock);
1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855

		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
	 */
1856 1857
	hash = 0;
	cnt = 0;
1858
	list_for_each_entry(nsh, &newstripes, lru) {
1859 1860 1861 1862 1863 1864 1865
		lock_device_hash_lock(conf, hash);
		wait_event_cmd(conf->wait_for_stripe,
				    !list_empty(conf->inactive_list + hash),
				    unlock_device_hash_lock(conf, hash),
				    lock_device_hash_lock(conf, hash));
		osh = get_free_stripe(conf, hash);
		unlock_device_hash_lock(conf, hash);
1866
		atomic_set(&nsh->count, 1);
1867
		for(i=0; i<conf->pool_size; i++) {
1868
			nsh->dev[i].page = osh->dev[i].page;
1869 1870
			nsh->dev[i].orig_page = osh->dev[i].page;
		}
1871 1872
		for( ; i<newsize; i++)
			nsh->dev[i].page = NULL;
1873
		nsh->hash_lock_index = hash;
1874
		kmem_cache_free(conf->slab_cache, osh);
1875 1876 1877 1878 1879 1880
		cnt++;
		if (cnt >= conf->max_nr_stripes / NR_STRIPE_HASH_LOCKS +
		    !!((conf->max_nr_stripes % NR_STRIPE_HASH_LOCKS) > hash)) {
			hash++;
			cnt = 0;
		}
1881 1882 1883 1884 1885 1886
	}
	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
1887
	 * conf->disks and the scribble region
1888 1889 1890 1891 1892 1893 1894 1895 1896 1897
	 */
	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;

1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916
	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();

1917 1918 1919 1920
	/* 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);
1921

1922 1923 1924 1925
		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;
1926
				nsh->dev[i].orig_page = p;
1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938
				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
1939

1940
static int drop_one_stripe(struct r5conf *conf, int hash)
L
Linus Torvalds 已提交
1941 1942 1943
{
	struct stripe_head *sh;

1944 1945 1946
	spin_lock_irq(conf->hash_locks + hash);
	sh = get_free_stripe(conf, hash);
	spin_unlock_irq(conf->hash_locks + hash);
1947 1948
	if (!sh)
		return 0;
1949
	BUG_ON(atomic_read(&sh->count));
1950
	shrink_buffers(sh);
1951 1952 1953 1954 1955
	kmem_cache_free(conf->slab_cache, sh);
	atomic_dec(&conf->active_stripes);
	return 1;
}

1956
static void shrink_stripes(struct r5conf *conf)
1957
{
1958 1959 1960 1961
	int hash;
	for (hash = 0; hash < NR_STRIPE_HASH_LOCKS; hash++)
		while (drop_one_stripe(conf, hash))
			;
1962

N
NeilBrown 已提交
1963 1964
	if (conf->slab_cache)
		kmem_cache_destroy(conf->slab_cache);
L
Linus Torvalds 已提交
1965 1966 1967
	conf->slab_cache = NULL;
}

1968
static void raid5_end_read_request(struct bio * bi, int error)
L
Linus Torvalds 已提交
1969
{
1970
	struct stripe_head *sh = bi->bi_private;
1971
	struct r5conf *conf = sh->raid_conf;
1972
	int disks = sh->disks, i;
L
Linus Torvalds 已提交
1973
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
1974
	char b[BDEVNAME_SIZE];
1975
	struct md_rdev *rdev = NULL;
1976
	sector_t s;
L
Linus Torvalds 已提交
1977 1978 1979 1980 1981

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

1982 1983
	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
1984 1985 1986
		uptodate);
	if (i == disks) {
		BUG();
1987
		return;
L
Linus Torvalds 已提交
1988
	}
1989
	if (test_bit(R5_ReadRepl, &sh->dev[i].flags))
1990 1991 1992 1993 1994
		/* 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.
		 */
1995
		rdev = conf->disks[i].replacement;
1996
	if (!rdev)
1997
		rdev = conf->disks[i].rdev;
L
Linus Torvalds 已提交
1998

1999 2000 2001 2002
	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
2003 2004
	if (uptodate) {
		set_bit(R5_UPTODATE, &sh->dev[i].flags);
2005
		if (test_bit(R5_ReadError, &sh->dev[i].flags)) {
2006 2007 2008 2009
			/* Note that this cannot happen on a
			 * replacement device.  We just fail those on
			 * any error
			 */
2010 2011 2012 2013 2014
			printk_ratelimited(
				KERN_INFO
				"md/raid:%s: read error corrected"
				" (%lu sectors at %llu on %s)\n",
				mdname(conf->mddev), STRIPE_SECTORS,
2015
				(unsigned long long)s,
2016
				bdevname(rdev->bdev, b));
2017
			atomic_add(STRIPE_SECTORS, &rdev->corrected_errors);
2018 2019
			clear_bit(R5_ReadError, &sh->dev[i].flags);
			clear_bit(R5_ReWrite, &sh->dev[i].flags);
2020 2021 2022
		} else if (test_bit(R5_ReadNoMerge, &sh->dev[i].flags))
			clear_bit(R5_ReadNoMerge, &sh->dev[i].flags);

2023 2024
		if (atomic_read(&rdev->read_errors))
			atomic_set(&rdev->read_errors, 0);
L
Linus Torvalds 已提交
2025
	} else {
2026
		const char *bdn = bdevname(rdev->bdev, b);
2027
		int retry = 0;
2028
		int set_bad = 0;
2029

L
Linus Torvalds 已提交
2030
		clear_bit(R5_UPTODATE, &sh->dev[i].flags);
2031
		atomic_inc(&rdev->read_errors);
2032 2033 2034 2035 2036 2037
		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),
2038
				(unsigned long long)s,
2039
				bdn);
2040 2041
		else if (conf->mddev->degraded >= conf->max_degraded) {
			set_bad = 1;
2042 2043 2044 2045 2046
			printk_ratelimited(
				KERN_WARNING
				"md/raid:%s: read error not correctable "
				"(sector %llu on %s).\n",
				mdname(conf->mddev),
2047
				(unsigned long long)s,
2048
				bdn);
2049
		} else if (test_bit(R5_ReWrite, &sh->dev[i].flags)) {
2050
			/* Oh, no!!! */
2051
			set_bad = 1;
2052 2053 2054 2055 2056
			printk_ratelimited(
				KERN_WARNING
				"md/raid:%s: read error NOT corrected!! "
				"(sector %llu on %s).\n",
				mdname(conf->mddev),
2057
				(unsigned long long)s,
2058
				bdn);
2059
		} else if (atomic_read(&rdev->read_errors)
2060
			 > conf->max_nr_stripes)
N
NeilBrown 已提交
2061
			printk(KERN_WARNING
2062
			       "md/raid:%s: Too many read errors, failing device %s.\n",
2063
			       mdname(conf->mddev), bdn);
2064 2065
		else
			retry = 1;
2066 2067 2068
		if (set_bad && test_bit(In_sync, &rdev->flags)
		    && !test_bit(R5_ReadNoMerge, &sh->dev[i].flags))
			retry = 1;
2069
		if (retry)
2070 2071 2072 2073 2074
			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);
2075
		else {
2076 2077
			clear_bit(R5_ReadError, &sh->dev[i].flags);
			clear_bit(R5_ReWrite, &sh->dev[i].flags);
2078 2079 2080 2081 2082
			if (!(set_bad
			      && test_bit(In_sync, &rdev->flags)
			      && rdev_set_badblocks(
				      rdev, sh->sector, STRIPE_SECTORS, 0)))
				md_error(conf->mddev, rdev);
2083
		}
L
Linus Torvalds 已提交
2084
	}
2085
	rdev_dec_pending(rdev, conf->mddev);
L
Linus Torvalds 已提交
2086 2087 2088 2089 2090
	clear_bit(R5_LOCKED, &sh->dev[i].flags);
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

2091
static void raid5_end_write_request(struct bio *bi, int error)
L
Linus Torvalds 已提交
2092
{
2093
	struct stripe_head *sh = bi->bi_private;
2094
	struct r5conf *conf = sh->raid_conf;
2095
	int disks = sh->disks, i;
2096
	struct md_rdev *uninitialized_var(rdev);
L
Linus Torvalds 已提交
2097
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
2098 2099
	sector_t first_bad;
	int bad_sectors;
2100
	int replacement = 0;
L
Linus Torvalds 已提交
2101

2102 2103 2104
	for (i = 0 ; i < disks; i++) {
		if (bi == &sh->dev[i].req) {
			rdev = conf->disks[i].rdev;
L
Linus Torvalds 已提交
2105
			break;
2106 2107 2108
		}
		if (bi == &sh->dev[i].rreq) {
			rdev = conf->disks[i].replacement;
2109 2110 2111 2112 2113 2114 2115 2116
			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;
2117 2118 2119
			break;
		}
	}
2120
	pr_debug("end_write_request %llu/%d, count %d, uptodate: %d.\n",
L
Linus Torvalds 已提交
2121 2122 2123 2124
		(unsigned long long)sh->sector, i, atomic_read(&sh->count),
		uptodate);
	if (i == disks) {
		BUG();
2125
		return;
L
Linus Torvalds 已提交
2126 2127
	}

2128 2129 2130 2131 2132 2133 2134 2135 2136
	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) {
2137
			set_bit(STRIPE_DEGRADED, &sh->state);
2138 2139
			set_bit(WriteErrorSeen, &rdev->flags);
			set_bit(R5_WriteError, &sh->dev[i].flags);
2140 2141 2142
			if (!test_and_set_bit(WantReplacement, &rdev->flags))
				set_bit(MD_RECOVERY_NEEDED,
					&rdev->mddev->recovery);
2143 2144
		} else if (is_badblock(rdev, sh->sector,
				       STRIPE_SECTORS,
2145
				       &first_bad, &bad_sectors)) {
2146
			set_bit(R5_MadeGood, &sh->dev[i].flags);
2147 2148 2149 2150 2151 2152 2153
			if (test_bit(R5_ReadError, &sh->dev[i].flags))
				/* That was a successful write so make
				 * sure it looks like we already did
				 * a re-write.
				 */
				set_bit(R5_ReWrite, &sh->dev[i].flags);
		}
2154 2155
	}
	rdev_dec_pending(rdev, conf->mddev);
L
Linus Torvalds 已提交
2156

2157 2158
	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
2159
	set_bit(STRIPE_HANDLE, &sh->state);
2160
	release_stripe(sh);
L
Linus Torvalds 已提交
2161 2162
}

2163
static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous);
2164

2165
static void raid5_build_block(struct stripe_head *sh, int i, int previous)
L
Linus Torvalds 已提交
2166 2167 2168 2169 2170
{
	struct r5dev *dev = &sh->dev[i];

	bio_init(&dev->req);
	dev->req.bi_io_vec = &dev->vec;
2171
	dev->req.bi_max_vecs = 1;
L
Linus Torvalds 已提交
2172 2173
	dev->req.bi_private = sh;

2174 2175
	bio_init(&dev->rreq);
	dev->rreq.bi_io_vec = &dev->rvec;
2176
	dev->rreq.bi_max_vecs = 1;
2177 2178
	dev->rreq.bi_private = sh;

L
Linus Torvalds 已提交
2179
	dev->flags = 0;
2180
	dev->sector = compute_blocknr(sh, i, previous);
L
Linus Torvalds 已提交
2181 2182
}

2183
static void error(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
2184 2185
{
	char b[BDEVNAME_SIZE];
2186
	struct r5conf *conf = mddev->private;
2187
	unsigned long flags;
2188
	pr_debug("raid456: error called\n");
L
Linus Torvalds 已提交
2189

2190 2191 2192 2193 2194 2195
	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);

2196
	set_bit(Blocked, &rdev->flags);
2197 2198 2199 2200 2201 2202 2203 2204 2205
	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);
2206
}
L
Linus Torvalds 已提交
2207 2208 2209 2210 2211

/*
 * Input: a 'big' sector number,
 * Output: index of the data and parity disk, and the sector # in them.
 */
2212
static sector_t raid5_compute_sector(struct r5conf *conf, sector_t r_sector,
2213 2214
				     int previous, int *dd_idx,
				     struct stripe_head *sh)
L
Linus Torvalds 已提交
2215
{
N
NeilBrown 已提交
2216
	sector_t stripe, stripe2;
2217
	sector_t chunk_number;
L
Linus Torvalds 已提交
2218
	unsigned int chunk_offset;
2219
	int pd_idx, qd_idx;
2220
	int ddf_layout = 0;
L
Linus Torvalds 已提交
2221
	sector_t new_sector;
2222 2223
	int algorithm = previous ? conf->prev_algo
				 : conf->algorithm;
2224 2225
	int sectors_per_chunk = previous ? conf->prev_chunk_sectors
					 : conf->chunk_sectors;
2226 2227 2228
	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
2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240

	/* 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
	 */
2241 2242
	stripe = chunk_number;
	*dd_idx = sector_div(stripe, data_disks);
N
NeilBrown 已提交
2243
	stripe2 = stripe;
L
Linus Torvalds 已提交
2244 2245 2246
	/*
	 * Select the parity disk based on the user selected algorithm.
	 */
2247
	pd_idx = qd_idx = -1;
2248 2249
	switch(conf->level) {
	case 4:
2250
		pd_idx = data_disks;
2251 2252
		break;
	case 5:
2253
		switch (algorithm) {
L
Linus Torvalds 已提交
2254
		case ALGORITHM_LEFT_ASYMMETRIC:
N
NeilBrown 已提交
2255
			pd_idx = data_disks - sector_div(stripe2, raid_disks);
2256
			if (*dd_idx >= pd_idx)
L
Linus Torvalds 已提交
2257 2258 2259
				(*dd_idx)++;
			break;
		case ALGORITHM_RIGHT_ASYMMETRIC:
N
NeilBrown 已提交
2260
			pd_idx = sector_div(stripe2, raid_disks);
2261
			if (*dd_idx >= pd_idx)
L
Linus Torvalds 已提交
2262 2263 2264
				(*dd_idx)++;
			break;
		case ALGORITHM_LEFT_SYMMETRIC:
N
NeilBrown 已提交
2265
			pd_idx = data_disks - sector_div(stripe2, raid_disks);
2266
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
L
Linus Torvalds 已提交
2267 2268
			break;
		case ALGORITHM_RIGHT_SYMMETRIC:
N
NeilBrown 已提交
2269
			pd_idx = sector_div(stripe2, raid_disks);
2270
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
L
Linus Torvalds 已提交
2271
			break;
2272 2273 2274 2275 2276 2277 2278
		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
2279
		default:
2280
			BUG();
2281 2282 2283 2284
		}
		break;
	case 6:

2285
		switch (algorithm) {
2286
		case ALGORITHM_LEFT_ASYMMETRIC:
N
NeilBrown 已提交
2287
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2288 2289
			qd_idx = pd_idx + 1;
			if (pd_idx == raid_disks-1) {
2290
				(*dd_idx)++;	/* Q D D D P */
2291 2292
				qd_idx = 0;
			} else if (*dd_idx >= pd_idx)
2293 2294 2295
				(*dd_idx) += 2; /* D D P Q D */
			break;
		case ALGORITHM_RIGHT_ASYMMETRIC:
N
NeilBrown 已提交
2296
			pd_idx = sector_div(stripe2, raid_disks);
2297 2298
			qd_idx = pd_idx + 1;
			if (pd_idx == raid_disks-1) {
2299
				(*dd_idx)++;	/* Q D D D P */
2300 2301
				qd_idx = 0;
			} else if (*dd_idx >= pd_idx)
2302 2303 2304
				(*dd_idx) += 2; /* D D P Q D */
			break;
		case ALGORITHM_LEFT_SYMMETRIC:
N
NeilBrown 已提交
2305
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2306 2307
			qd_idx = (pd_idx + 1) % raid_disks;
			*dd_idx = (pd_idx + 2 + *dd_idx) % raid_disks;
2308 2309
			break;
		case ALGORITHM_RIGHT_SYMMETRIC:
N
NeilBrown 已提交
2310
			pd_idx = sector_div(stripe2, raid_disks);
2311 2312
			qd_idx = (pd_idx + 1) % raid_disks;
			*dd_idx = (pd_idx + 2 + *dd_idx) % raid_disks;
2313
			break;
2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328

		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
2329
			pd_idx = sector_div(stripe2, raid_disks);
2330 2331 2332 2333 2334 2335
			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 */
2336
			ddf_layout = 1;
2337 2338 2339 2340 2341 2342 2343
			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
2344 2345
			stripe2 += 1;
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2346 2347 2348 2349 2350 2351
			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 */
2352
			ddf_layout = 1;
2353 2354 2355 2356
			break;

		case ALGORITHM_ROTATING_N_CONTINUE:
			/* Same as left_symmetric but Q is before P */
N
NeilBrown 已提交
2357
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2358 2359
			qd_idx = (pd_idx + raid_disks - 1) % raid_disks;
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
2360
			ddf_layout = 1;
2361 2362 2363 2364
			break;

		case ALGORITHM_LEFT_ASYMMETRIC_6:
			/* RAID5 left_asymmetric, with Q on last device */
N
NeilBrown 已提交
2365
			pd_idx = data_disks - sector_div(stripe2, raid_disks-1);
2366 2367 2368 2369 2370 2371
			if (*dd_idx >= pd_idx)
				(*dd_idx)++;
			qd_idx = raid_disks - 1;
			break;

		case ALGORITHM_RIGHT_ASYMMETRIC_6:
N
NeilBrown 已提交
2372
			pd_idx = sector_div(stripe2, raid_disks-1);
2373 2374 2375 2376 2377 2378
			if (*dd_idx >= pd_idx)
				(*dd_idx)++;
			qd_idx = raid_disks - 1;
			break;

		case ALGORITHM_LEFT_SYMMETRIC_6:
N
NeilBrown 已提交
2379
			pd_idx = data_disks - sector_div(stripe2, raid_disks-1);
2380 2381 2382 2383 2384
			*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
2385
			pd_idx = sector_div(stripe2, raid_disks-1);
2386 2387 2388 2389 2390 2391 2392 2393 2394 2395
			*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;

2396
		default:
2397
			BUG();
2398 2399
		}
		break;
L
Linus Torvalds 已提交
2400 2401
	}

2402 2403 2404
	if (sh) {
		sh->pd_idx = pd_idx;
		sh->qd_idx = qd_idx;
2405
		sh->ddf_layout = ddf_layout;
2406
	}
L
Linus Torvalds 已提交
2407 2408 2409 2410 2411 2412 2413
	/*
	 * Finally, compute the new sector number
	 */
	new_sector = (sector_t)stripe * sectors_per_chunk + chunk_offset;
	return new_sector;
}

2414
static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous)
L
Linus Torvalds 已提交
2415
{
2416
	struct r5conf *conf = sh->raid_conf;
2417 2418
	int raid_disks = sh->disks;
	int data_disks = raid_disks - conf->max_degraded;
L
Linus Torvalds 已提交
2419
	sector_t new_sector = sh->sector, check;
2420 2421
	int sectors_per_chunk = previous ? conf->prev_chunk_sectors
					 : conf->chunk_sectors;
2422 2423
	int algorithm = previous ? conf->prev_algo
				 : conf->algorithm;
L
Linus Torvalds 已提交
2424 2425
	sector_t stripe;
	int chunk_offset;
2426 2427
	sector_t chunk_number;
	int dummy1, dd_idx = i;
L
Linus Torvalds 已提交
2428
	sector_t r_sector;
2429
	struct stripe_head sh2;
L
Linus Torvalds 已提交
2430 2431 2432 2433

	chunk_offset = sector_div(new_sector, sectors_per_chunk);
	stripe = new_sector;

2434 2435 2436 2437 2438
	if (i == sh->pd_idx)
		return 0;
	switch(conf->level) {
	case 4: break;
	case 5:
2439
		switch (algorithm) {
L
Linus Torvalds 已提交
2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450
		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;
2451 2452 2453 2454 2455
		case ALGORITHM_PARITY_0:
			i -= 1;
			break;
		case ALGORITHM_PARITY_N:
			break;
L
Linus Torvalds 已提交
2456
		default:
2457
			BUG();
2458 2459 2460
		}
		break;
	case 6:
2461
		if (i == sh->qd_idx)
2462
			return 0; /* It is the Q disk */
2463
		switch (algorithm) {
2464 2465
		case ALGORITHM_LEFT_ASYMMETRIC:
		case ALGORITHM_RIGHT_ASYMMETRIC:
2466 2467 2468 2469
		case ALGORITHM_ROTATING_ZERO_RESTART:
		case ALGORITHM_ROTATING_N_RESTART:
			if (sh->pd_idx == raid_disks-1)
				i--;	/* Q D D D P */
2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483
			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;
2484 2485 2486 2487 2488 2489
		case ALGORITHM_PARITY_0:
			i -= 2;
			break;
		case ALGORITHM_PARITY_N:
			break;
		case ALGORITHM_ROTATING_N_CONTINUE:
2490
			/* Like left_symmetric, but P is before Q */
2491 2492
			if (sh->pd_idx == 0)
				i--;	/* P D D D Q */
2493 2494 2495 2496 2497 2498
			else {
				/* D D Q P D */
				if (i < sh->pd_idx)
					i += raid_disks;
				i -= (sh->pd_idx + 1);
			}
2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513
			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;
2514
		default:
2515
			BUG();
2516 2517
		}
		break;
L
Linus Torvalds 已提交
2518 2519 2520
	}

	chunk_number = stripe * data_disks + i;
2521
	r_sector = chunk_number * sectors_per_chunk + chunk_offset;
L
Linus Torvalds 已提交
2522

2523
	check = raid5_compute_sector(conf, r_sector,
2524
				     previous, &dummy1, &sh2);
2525 2526
	if (check != sh->sector || dummy1 != dd_idx || sh2.pd_idx != sh->pd_idx
		|| sh2.qd_idx != sh->qd_idx) {
2527 2528
		printk(KERN_ERR "md/raid:%s: compute_blocknr: map not correct\n",
		       mdname(conf->mddev));
L
Linus Torvalds 已提交
2529 2530 2531 2532 2533
		return 0;
	}
	return r_sector;
}

2534
static void
2535
schedule_reconstruction(struct stripe_head *sh, struct stripe_head_state *s,
2536
			 int rcw, int expand)
2537 2538
{
	int i, pd_idx = sh->pd_idx, disks = sh->disks;
2539
	struct r5conf *conf = sh->raid_conf;
2540
	int level = conf->level;
2541 2542 2543 2544 2545 2546 2547 2548

	if (rcw) {

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

			if (dev->towrite) {
				set_bit(R5_LOCKED, &dev->flags);
2549
				set_bit(R5_Wantdrain, &dev->flags);
2550 2551
				if (!expand)
					clear_bit(R5_UPTODATE, &dev->flags);
2552
				s->locked++;
2553 2554
			}
		}
2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569
		/* if we are not expanding this is a proper write request, and
		 * there will be bios with new data to be drained into the
		 * stripe cache
		 */
		if (!expand) {
			if (!s->locked)
				/* False alarm, nothing to do */
				return;
			sh->reconstruct_state = reconstruct_state_drain_run;
			set_bit(STRIPE_OP_BIODRAIN, &s->ops_request);
		} else
			sh->reconstruct_state = reconstruct_state_run;

		set_bit(STRIPE_OP_RECONSTRUCT, &s->ops_request);

2570
		if (s->locked + conf->max_degraded == disks)
2571
			if (!test_and_set_bit(STRIPE_FULL_WRITE, &sh->state))
2572
				atomic_inc(&conf->pending_full_writes);
2573
	} else {
2574
		BUG_ON(level == 6);
2575 2576 2577 2578 2579 2580 2581 2582 2583 2584
		BUG_ON(!(test_bit(R5_UPTODATE, &sh->dev[pd_idx].flags) ||
			test_bit(R5_Wantcompute, &sh->dev[pd_idx].flags)));

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

			if (dev->towrite &&
			    (test_bit(R5_UPTODATE, &dev->flags) ||
2585 2586
			     test_bit(R5_Wantcompute, &dev->flags))) {
				set_bit(R5_Wantdrain, &dev->flags);
2587 2588
				set_bit(R5_LOCKED, &dev->flags);
				clear_bit(R5_UPTODATE, &dev->flags);
2589
				s->locked++;
2590 2591
			}
		}
2592 2593 2594 2595 2596 2597 2598
		if (!s->locked)
			/* False alarm - nothing to do */
			return;
		sh->reconstruct_state = reconstruct_state_prexor_drain_run;
		set_bit(STRIPE_OP_PREXOR, &s->ops_request);
		set_bit(STRIPE_OP_BIODRAIN, &s->ops_request);
		set_bit(STRIPE_OP_RECONSTRUCT, &s->ops_request);
2599 2600
	}

2601
	/* keep the parity disk(s) locked while asynchronous operations
2602 2603 2604 2605
	 * are in flight
	 */
	set_bit(R5_LOCKED, &sh->dev[pd_idx].flags);
	clear_bit(R5_UPTODATE, &sh->dev[pd_idx].flags);
2606
	s->locked++;
2607

2608 2609 2610 2611 2612 2613 2614 2615 2616
	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++;
	}

2617
	pr_debug("%s: stripe %llu locked: %d ops_request: %lx\n",
2618
		__func__, (unsigned long long)sh->sector,
2619
		s->locked, s->ops_request);
2620
}
2621

L
Linus Torvalds 已提交
2622 2623
/*
 * Each stripe/dev can have one or more bion attached.
2624
 * toread/towrite point to the first in a chain.
L
Linus Torvalds 已提交
2625 2626 2627 2628 2629
 * 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;
2630
	struct r5conf *conf = sh->raid_conf;
2631
	int firstwrite=0;
L
Linus Torvalds 已提交
2632

2633
	pr_debug("adding bi b#%llu to stripe s#%llu\n",
2634
		(unsigned long long)bi->bi_iter.bi_sector,
L
Linus Torvalds 已提交
2635 2636
		(unsigned long long)sh->sector);

S
Shaohua Li 已提交
2637 2638 2639 2640 2641 2642 2643 2644 2645
	/*
	 * 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);
2646
	if (forwrite) {
L
Linus Torvalds 已提交
2647
		bip = &sh->dev[dd_idx].towrite;
2648
		if (*bip == NULL)
2649 2650
			firstwrite = 1;
	} else
L
Linus Torvalds 已提交
2651
		bip = &sh->dev[dd_idx].toread;
2652 2653
	while (*bip && (*bip)->bi_iter.bi_sector < bi->bi_iter.bi_sector) {
		if (bio_end_sector(*bip) > bi->bi_iter.bi_sector)
L
Linus Torvalds 已提交
2654 2655 2656
			goto overlap;
		bip = & (*bip)->bi_next;
	}
2657
	if (*bip && (*bip)->bi_iter.bi_sector < bio_end_sector(bi))
L
Linus Torvalds 已提交
2658 2659
		goto overlap;

2660
	BUG_ON(*bip && bi->bi_next && (*bip) != bi->bi_next);
L
Linus Torvalds 已提交
2661 2662 2663
	if (*bip)
		bi->bi_next = *bip;
	*bip = bi;
2664
	raid5_inc_bi_active_stripes(bi);
2665

L
Linus Torvalds 已提交
2666 2667 2668 2669 2670
	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 &&
2671
			     bi && bi->bi_iter.bi_sector <= sector;
L
Linus Torvalds 已提交
2672
		     bi = r5_next_bio(bi, sh->dev[dd_idx].sector)) {
K
Kent Overstreet 已提交
2673 2674
			if (bio_end_sector(bi) >= sector)
				sector = bio_end_sector(bi);
L
Linus Torvalds 已提交
2675 2676 2677 2678
		}
		if (sector >= sh->dev[dd_idx].sector + STRIPE_SECTORS)
			set_bit(R5_OVERWRITE, &sh->dev[dd_idx].flags);
	}
2679 2680

	pr_debug("added bi b#%llu to stripe s#%llu, disk %d.\n",
2681
		(unsigned long long)(*bip)->bi_iter.bi_sector,
2682
		(unsigned long long)sh->sector, dd_idx);
2683
	spin_unlock_irq(&sh->stripe_lock);
2684 2685 2686 2687 2688 2689 2690

	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
2691 2692 2693 2694
	return 1;

 overlap:
	set_bit(R5_Overlap, &sh->dev[dd_idx].flags);
S
Shaohua Li 已提交
2695
	spin_unlock_irq(&sh->stripe_lock);
L
Linus Torvalds 已提交
2696 2697 2698
	return 0;
}

2699
static void end_reshape(struct r5conf *conf);
2700

2701
static void stripe_set_idx(sector_t stripe, struct r5conf *conf, int previous,
2702
			    struct stripe_head *sh)
2703
{
2704
	int sectors_per_chunk =
2705
		previous ? conf->prev_chunk_sectors : conf->chunk_sectors;
2706
	int dd_idx;
2707
	int chunk_offset = sector_div(stripe, sectors_per_chunk);
2708
	int disks = previous ? conf->previous_raid_disks : conf->raid_disks;
2709

2710 2711
	raid5_compute_sector(conf,
			     stripe * (disks - conf->max_degraded)
2712
			     *sectors_per_chunk + chunk_offset,
2713
			     previous,
2714
			     &dd_idx, sh);
2715 2716
}

2717
static void
2718
handle_failed_stripe(struct r5conf *conf, struct stripe_head *sh,
2719 2720 2721 2722 2723 2724 2725 2726 2727
				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)) {
2728
			struct md_rdev *rdev;
2729 2730 2731
			rcu_read_lock();
			rdev = rcu_dereference(conf->disks[i].rdev);
			if (rdev && test_bit(In_sync, &rdev->flags))
2732 2733 2734
				atomic_inc(&rdev->nr_pending);
			else
				rdev = NULL;
2735
			rcu_read_unlock();
2736 2737 2738 2739 2740 2741 2742 2743
			if (rdev) {
				if (!rdev_set_badblocks(
					    rdev,
					    sh->sector,
					    STRIPE_SECTORS, 0))
					md_error(conf->mddev, rdev);
				rdev_dec_pending(rdev, conf->mddev);
			}
2744
		}
S
Shaohua Li 已提交
2745
		spin_lock_irq(&sh->stripe_lock);
2746 2747 2748
		/* fail all writes first */
		bi = sh->dev[i].towrite;
		sh->dev[i].towrite = NULL;
S
Shaohua Li 已提交
2749
		spin_unlock_irq(&sh->stripe_lock);
2750
		if (bi)
2751 2752 2753 2754 2755
			bitmap_end = 1;

		if (test_and_clear_bit(R5_Overlap, &sh->dev[i].flags))
			wake_up(&conf->wait_for_overlap);

2756
		while (bi && bi->bi_iter.bi_sector <
2757 2758 2759
			sh->dev[i].sector + STRIPE_SECTORS) {
			struct bio *nextbi = r5_next_bio(bi, sh->dev[i].sector);
			clear_bit(BIO_UPTODATE, &bi->bi_flags);
2760
			if (!raid5_dec_bi_active_stripes(bi)) {
2761 2762 2763 2764 2765 2766
				md_write_end(conf->mddev);
				bi->bi_next = *return_bi;
				*return_bi = bi;
			}
			bi = nextbi;
		}
2767 2768 2769 2770
		if (bitmap_end)
			bitmap_endwrite(conf->mddev->bitmap, sh->sector,
				STRIPE_SECTORS, 0, 0);
		bitmap_end = 0;
2771 2772 2773
		/* and fail all 'written' */
		bi = sh->dev[i].written;
		sh->dev[i].written = NULL;
2774 2775 2776 2777 2778
		if (test_and_clear_bit(R5_SkipCopy, &sh->dev[i].flags)) {
			WARN_ON(test_bit(R5_UPTODATE, &sh->dev[i].flags));
			sh->dev[i].page = sh->dev[i].orig_page;
		}

2779
		if (bi) bitmap_end = 1;
2780
		while (bi && bi->bi_iter.bi_sector <
2781 2782 2783
		       sh->dev[i].sector + STRIPE_SECTORS) {
			struct bio *bi2 = r5_next_bio(bi, sh->dev[i].sector);
			clear_bit(BIO_UPTODATE, &bi->bi_flags);
2784
			if (!raid5_dec_bi_active_stripes(bi)) {
2785 2786 2787 2788 2789 2790 2791
				md_write_end(conf->mddev);
				bi->bi_next = *return_bi;
				*return_bi = bi;
			}
			bi = bi2;
		}

2792 2793 2794 2795 2796 2797
		/* 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))) {
2798
			spin_lock_irq(&sh->stripe_lock);
2799 2800
			bi = sh->dev[i].toread;
			sh->dev[i].toread = NULL;
2801
			spin_unlock_irq(&sh->stripe_lock);
2802 2803
			if (test_and_clear_bit(R5_Overlap, &sh->dev[i].flags))
				wake_up(&conf->wait_for_overlap);
2804
			while (bi && bi->bi_iter.bi_sector <
2805 2806 2807 2808
			       sh->dev[i].sector + STRIPE_SECTORS) {
				struct bio *nextbi =
					r5_next_bio(bi, sh->dev[i].sector);
				clear_bit(BIO_UPTODATE, &bi->bi_flags);
2809
				if (!raid5_dec_bi_active_stripes(bi)) {
2810 2811 2812 2813 2814 2815 2816 2817 2818
					bi->bi_next = *return_bi;
					*return_bi = bi;
				}
				bi = nextbi;
			}
		}
		if (bitmap_end)
			bitmap_endwrite(conf->mddev->bitmap, sh->sector,
					STRIPE_SECTORS, 0, 0);
2819 2820 2821 2822
		/* 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);
2823 2824
	}

2825 2826 2827
	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);
2828 2829
}

2830
static void
2831
handle_failed_sync(struct r5conf *conf, struct stripe_head *sh,
2832 2833 2834 2835 2836 2837
		   struct stripe_head_state *s)
{
	int abort = 0;
	int i;

	clear_bit(STRIPE_SYNCING, &sh->state);
2838 2839
	if (test_and_clear_bit(R5_Overlap, &sh->dev[sh->pd_idx].flags))
		wake_up(&conf->wait_for_overlap);
2840
	s->syncing = 0;
2841
	s->replacing = 0;
2842
	/* There is nothing more to do for sync/check/repair.
2843 2844 2845
	 * 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.
2846
	 * For recover/replace we need to record a bad block on all
2847 2848
	 * non-sync devices, or abort the recovery
	 */
2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871
	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;
2872
	}
2873
	md_done_sync(conf->mddev, STRIPE_SECTORS, !abort);
2874 2875
}

2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891
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;
}

2892
/* fetch_block - checks the given member device to see if its data needs
2893 2894 2895
 * to be read or computed to satisfy a request.
 *
 * Returns 1 when no more member devices need to be checked, otherwise returns
2896
 * 0 to tell the loop in handle_stripe_fill to continue
2897
 */
2898 2899 2900

static int need_this_block(struct stripe_head *sh, struct stripe_head_state *s,
			   int disk_idx, int disks)
2901
{
2902
	struct r5dev *dev = &sh->dev[disk_idx];
2903 2904
	struct r5dev *fdev[2] = { &sh->dev[s->failed_num[0]],
				  &sh->dev[s->failed_num[1]] };
2905
	int i;
2906

2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933

	if (test_bit(R5_LOCKED, &dev->flags) ||
	    test_bit(R5_UPTODATE, &dev->flags))
		/* No point reading this as we already have it or have
		 * decided to get it.
		 */
		return 0;

	if (dev->toread ||
	    (dev->towrite && !test_bit(R5_OVERWRITE, &dev->flags)))
		/* We need this block to directly satisfy a request */
		return 1;

	if (s->syncing || s->expanding ||
	    (s->replacing && want_replace(sh, disk_idx)))
		/* When syncing, or expanding we read everything.
		 * When replacing, we need the replaced block.
		 */
		return 1;

	if ((s->failed >= 1 && fdev[0]->toread) ||
	    (s->failed >= 2 && fdev[1]->toread))
		/* If we want to read from a failed device, then
		 * we need to actually read every other device.
		 */
		return 1;

2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952
	/* Sometimes neither read-modify-write nor reconstruct-write
	 * cycles can work.  In those cases we read every block we
	 * can.  Then the parity-update is certain to have enough to
	 * work with.
	 * This can only be a problem when we need to write something,
	 * and some device has failed.  If either of those tests
	 * fail we need look no further.
	 */
	if (!s->failed || !s->to_write)
		return 0;

	if (test_bit(R5_Insync, &dev->flags) &&
	    !test_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
		/* Pre-reads at not permitted until after short delay
		 * to gather multiple requests.  However if this
		 * device is no Insync, the block could only be be computed
		 * and there is no need to delay that.
		 */
		return 0;
2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983

	for (i = 0; i < s->failed; i++) {
		if (fdev[i]->towrite &&
		    !test_bit(R5_UPTODATE, &fdev[i]->flags) &&
		    !test_bit(R5_OVERWRITE, &fdev[i]->flags))
			/* If we have a partial write to a failed
			 * device, then we will need to reconstruct
			 * the content of that device, so all other
			 * devices must be read.
			 */
			return 1;
	}

	/* If we are forced to do a reconstruct-write, either because
	 * the current RAID6 implementation only supports that, or
	 * or because parity cannot be trusted and we are currently
	 * recovering it, there is extra need to be careful.
	 * If one of the devices that we would need to read, because
	 * it is not being overwritten (and maybe not written at all)
	 * is missing/faulty, then we need to read everything we can.
	 */
	if (sh->raid_conf->level != 6 &&
	    sh->sector < sh->raid_conf->mddev->recovery_cp)
		/* reconstruct-write isn't being forced */
		return 0;
	for (i = 0; i < s->failed; i++) {
		if (!test_bit(R5_UPTODATE, &fdev[i]->flags) &&
		    !test_bit(R5_OVERWRITE, &fdev[i]->flags))
			return 1;
	}

2984 2985 2986 2987 2988 2989 2990 2991 2992 2993
	return 0;
}

static int fetch_block(struct stripe_head *sh, struct stripe_head_state *s,
		       int disk_idx, int disks)
{
	struct r5dev *dev = &sh->dev[disk_idx];

	/* is the data in this block needed, and can we get it? */
	if (need_this_block(sh, s, disk_idx, disks)) {
2994 2995 2996 2997 2998 2999
		/* 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) &&
3000 3001
		    (s->failed && (disk_idx == s->failed_num[0] ||
				   disk_idx == s->failed_num[1]))) {
3002 3003
			/* have disk failed, and we're requested to fetch it;
			 * do compute it
3004
			 */
3005 3006 3007 3008 3009 3010 3011 3012
			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;
3013 3014 3015 3016 3017 3018
			/* 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.
			 */
3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031
			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;
3032
			}
3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051
			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);
3052 3053
		}
	}
3054 3055 3056 3057 3058

	return 0;
}

/**
3059
 * handle_stripe_fill - read or compute data to satisfy pending requests.
3060
 */
3061 3062 3063
static void handle_stripe_fill(struct stripe_head *sh,
			       struct stripe_head_state *s,
			       int disks)
3064 3065 3066 3067 3068 3069 3070 3071 3072 3073
{
	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--; )
3074
			if (fetch_block(sh, s, i, disks))
3075
				break;
3076 3077 3078
	set_bit(STRIPE_HANDLE, &sh->state);
}

3079
/* handle_stripe_clean_event
3080 3081 3082 3083
 * 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.
 */
3084
static void handle_stripe_clean_event(struct r5conf *conf,
3085 3086 3087 3088
	struct stripe_head *sh, int disks, struct bio **return_bi)
{
	int i;
	struct r5dev *dev;
3089
	int discard_pending = 0;
3090 3091 3092 3093 3094

	for (i = disks; i--; )
		if (sh->dev[i].written) {
			dev = &sh->dev[i];
			if (!test_bit(R5_LOCKED, &dev->flags) &&
3095
			    (test_bit(R5_UPTODATE, &dev->flags) ||
3096 3097
			     test_bit(R5_Discard, &dev->flags) ||
			     test_bit(R5_SkipCopy, &dev->flags))) {
3098 3099
				/* We can return any write requests */
				struct bio *wbi, *wbi2;
3100
				pr_debug("Return write for disc %d\n", i);
3101 3102
				if (test_and_clear_bit(R5_Discard, &dev->flags))
					clear_bit(R5_UPTODATE, &dev->flags);
3103 3104 3105 3106
				if (test_and_clear_bit(R5_SkipCopy, &dev->flags)) {
					WARN_ON(test_bit(R5_UPTODATE, &dev->flags));
					dev->page = dev->orig_page;
				}
3107 3108
				wbi = dev->written;
				dev->written = NULL;
3109
				while (wbi && wbi->bi_iter.bi_sector <
3110 3111
					dev->sector + STRIPE_SECTORS) {
					wbi2 = r5_next_bio(wbi, dev->sector);
3112
					if (!raid5_dec_bi_active_stripes(wbi)) {
3113 3114 3115 3116 3117 3118
						md_write_end(conf->mddev);
						wbi->bi_next = *return_bi;
						*return_bi = wbi;
					}
					wbi = wbi2;
				}
3119 3120
				bitmap_endwrite(conf->mddev->bitmap, sh->sector,
						STRIPE_SECTORS,
3121
					 !test_bit(STRIPE_DEGRADED, &sh->state),
3122
						0);
3123 3124
			} else if (test_bit(R5_Discard, &dev->flags))
				discard_pending = 1;
3125 3126
			WARN_ON(test_bit(R5_SkipCopy, &dev->flags));
			WARN_ON(dev->page != dev->orig_page);
3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137
		}
	if (!discard_pending &&
	    test_bit(R5_Discard, &sh->dev[sh->pd_idx].flags)) {
		clear_bit(R5_Discard, &sh->dev[sh->pd_idx].flags);
		clear_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags);
		if (sh->qd_idx >= 0) {
			clear_bit(R5_Discard, &sh->dev[sh->qd_idx].flags);
			clear_bit(R5_UPTODATE, &sh->dev[sh->qd_idx].flags);
		}
		/* now that discard is done we can proceed with any sync */
		clear_bit(STRIPE_DISCARD, &sh->state);
S
Shaohua Li 已提交
3138 3139 3140 3141 3142 3143 3144 3145
		/*
		 * SCSI discard will change some bio fields and the stripe has
		 * no updated data, so remove it from hash list and the stripe
		 * will be reinitialized
		 */
		spin_lock_irq(&conf->device_lock);
		remove_hash(sh);
		spin_unlock_irq(&conf->device_lock);
3146 3147 3148 3149
		if (test_bit(STRIPE_SYNC_REQUESTED, &sh->state))
			set_bit(STRIPE_HANDLE, &sh->state);

	}
3150 3151 3152 3153

	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);
3154 3155
}

3156
static void handle_stripe_dirtying(struct r5conf *conf,
3157 3158 3159
				   struct stripe_head *sh,
				   struct stripe_head_state *s,
				   int disks)
3160 3161
{
	int rmw = 0, rcw = 0, i;
3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172
	sector_t recovery_cp = conf->mddev->recovery_cp;

	/* RAID6 requires 'rcw' in current implementation.
	 * Otherwise, check whether resync is now happening or should start.
	 * If yes, then the array is dirty (after unclean shutdown or
	 * initial creation), so parity in some stripes might be inconsistent.
	 * In this case, we need to always do reconstruct-write, to ensure
	 * that in case of drive failure or read-error correction, we
	 * generate correct data from the parity.
	 */
	if (conf->max_degraded == 2 ||
3173 3174
	    (recovery_cp < MaxSector && sh->sector >= recovery_cp &&
	     s->failed == 0)) {
3175
		/* Calculate the real rcw later - for now make it
3176 3177 3178
		 * look like rcw is cheaper
		 */
		rcw = 1; rmw = 2;
3179 3180 3181
		pr_debug("force RCW max_degraded=%u, recovery_cp=%llu sh->sector=%llu\n",
			 conf->max_degraded, (unsigned long long)recovery_cp,
			 (unsigned long long)sh->sector);
3182
	} else for (i = disks; i--; ) {
3183 3184 3185 3186
		/* 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) &&
3187 3188
		    !(test_bit(R5_UPTODATE, &dev->flags) ||
		      test_bit(R5_Wantcompute, &dev->flags))) {
3189 3190 3191 3192 3193 3194 3195 3196
			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) &&
3197 3198
		    !(test_bit(R5_UPTODATE, &dev->flags) ||
		    test_bit(R5_Wantcompute, &dev->flags))) {
3199 3200
			if (test_bit(R5_Insync, &dev->flags))
				rcw++;
3201 3202 3203 3204
			else
				rcw += 2*disks;
		}
	}
3205
	pr_debug("for sector %llu, rmw=%d rcw=%d\n",
3206 3207
		(unsigned long long)sh->sector, rmw, rcw);
	set_bit(STRIPE_HANDLE, &sh->state);
N
NeilBrown 已提交
3208
	if (rmw < rcw && rmw > 0) {
3209
		/* prefer read-modify-write, but need to get some data */
3210 3211 3212 3213
		if (conf->mddev->queue)
			blk_add_trace_msg(conf->mddev->queue,
					  "raid5 rmw %llu %d",
					  (unsigned long long)sh->sector, rmw);
3214 3215 3216 3217
		for (i = disks; i--; ) {
			struct r5dev *dev = &sh->dev[i];
			if ((dev->towrite || i == sh->pd_idx) &&
			    !test_bit(R5_LOCKED, &dev->flags) &&
3218 3219
			    !(test_bit(R5_UPTODATE, &dev->flags) ||
			    test_bit(R5_Wantcompute, &dev->flags)) &&
3220
			    test_bit(R5_Insync, &dev->flags)) {
3221 3222 3223 3224
				if (test_bit(STRIPE_PREREAD_ACTIVE,
					     &sh->state)) {
					pr_debug("Read_old block %d for r-m-w\n",
						 i);
3225 3226 3227 3228 3229 3230 3231 3232 3233
					set_bit(R5_LOCKED, &dev->flags);
					set_bit(R5_Wantread, &dev->flags);
					s->locked++;
				} else {
					set_bit(STRIPE_DELAYED, &sh->state);
					set_bit(STRIPE_HANDLE, &sh->state);
				}
			}
		}
N
NeilBrown 已提交
3234
	}
3235
	if (rcw <= rmw && rcw > 0) {
3236
		/* want reconstruct write, but need to get some data */
N
NeilBrown 已提交
3237
		int qread =0;
3238
		rcw = 0;
3239 3240 3241
		for (i = disks; i--; ) {
			struct r5dev *dev = &sh->dev[i];
			if (!test_bit(R5_OVERWRITE, &dev->flags) &&
3242
			    i != sh->pd_idx && i != sh->qd_idx &&
3243
			    !test_bit(R5_LOCKED, &dev->flags) &&
3244
			    !(test_bit(R5_UPTODATE, &dev->flags) ||
3245 3246
			      test_bit(R5_Wantcompute, &dev->flags))) {
				rcw++;
3247 3248 3249
				if (test_bit(R5_Insync, &dev->flags) &&
				    test_bit(STRIPE_PREREAD_ACTIVE,
					     &sh->state)) {
3250
					pr_debug("Read_old block "
3251 3252 3253 3254
						"%d for Reconstruct\n", i);
					set_bit(R5_LOCKED, &dev->flags);
					set_bit(R5_Wantread, &dev->flags);
					s->locked++;
N
NeilBrown 已提交
3255
					qread++;
3256 3257 3258 3259 3260 3261
				} else {
					set_bit(STRIPE_DELAYED, &sh->state);
					set_bit(STRIPE_HANDLE, &sh->state);
				}
			}
		}
3262
		if (rcw && conf->mddev->queue)
N
NeilBrown 已提交
3263 3264 3265
			blk_add_trace_msg(conf->mddev->queue, "raid5 rcw %llu %d %d %d",
					  (unsigned long long)sh->sector,
					  rcw, qread, test_bit(STRIPE_DELAYED, &sh->state));
3266
	}
3267 3268 3269 3270 3271

	if (rcw > disks && rmw > disks &&
	    !test_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
		set_bit(STRIPE_DELAYED, &sh->state);

3272 3273 3274
	/* now if nothing is locked, and if we have enough data,
	 * we can start a write request
	 */
3275 3276
	/* 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
3277 3278
	 * subsequent call wants to start a write request.  raid_run_ops only
	 * handles the case where compute block and reconstruct are requested
3279 3280 3281
	 * simultaneously.  If this is not the case then new writes need to be
	 * held off until the compute completes.
	 */
3282 3283 3284
	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)))
3285
		schedule_reconstruction(sh, s, rcw == 0, 0);
3286 3287
}

3288
static void handle_parity_checks5(struct r5conf *conf, struct stripe_head *sh,
3289 3290
				struct stripe_head_state *s, int disks)
{
3291
	struct r5dev *dev = NULL;
3292

3293
	set_bit(STRIPE_HANDLE, &sh->state);
3294

3295 3296 3297
	switch (sh->check_state) {
	case check_state_idle:
		/* start a new check operation if there are no failures */
3298 3299
		if (s->failed == 0) {
			BUG_ON(s->uptodate != disks);
3300 3301
			sh->check_state = check_state_run;
			set_bit(STRIPE_OP_CHECK, &s->ops_request);
3302 3303
			clear_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags);
			s->uptodate--;
3304
			break;
3305
		}
3306
		dev = &sh->dev[s->failed_num[0]];
3307 3308 3309 3310 3311 3312 3313 3314 3315
		/* 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
3316

3317 3318 3319 3320 3321
		/* 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);
3322
		s->locked++;
3323
		set_bit(R5_Wantwrite, &dev->flags);
3324

3325 3326
		clear_bit(STRIPE_DEGRADED, &sh->state);
		set_bit(STRIPE_INSYNC, &sh->state);
3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342
		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
3343
		if ((sh->ops.zero_sum_result & SUM_CHECK_P_RESULT) == 0)
3344 3345 3346 3347 3348
			/* parity is correct (on disc,
			 * not in buffer any more)
			 */
			set_bit(STRIPE_INSYNC, &sh->state);
		else {
3349
			atomic64_add(STRIPE_SECTORS, &conf->mddev->resync_mismatches);
3350 3351 3352 3353 3354
			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;
3355
				set_bit(STRIPE_COMPUTE_RUN, &sh->state);
3356 3357 3358 3359
				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;
3360
				sh->ops.target2 = -1;
3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371
				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();
3372 3373 3374
	}
}

3375
static void handle_parity_checks6(struct r5conf *conf, struct stripe_head *sh,
3376
				  struct stripe_head_state *s,
3377
				  int disks)
3378 3379
{
	int pd_idx = sh->pd_idx;
N
NeilBrown 已提交
3380
	int qd_idx = sh->qd_idx;
3381
	struct r5dev *dev;
3382 3383 3384 3385

	set_bit(STRIPE_HANDLE, &sh->state);

	BUG_ON(s->failed > 2);
3386

3387 3388 3389 3390 3391 3392
	/* 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
	 */

3393 3394 3395
	switch (sh->check_state) {
	case check_state_idle:
		/* start a new check operation if there are < 2 failures */
3396
		if (s->failed == s->q_failed) {
3397
			/* The only possible failed device holds Q, so it
3398 3399 3400
			 * makes sense to check P (If anything else were failed,
			 * we would have used P to recreate it).
			 */
3401
			sh->check_state = check_state_run;
3402
		}
3403
		if (!s->q_failed && s->failed < 2) {
3404
			/* Q is not failed, and we didn't use it to generate
3405 3406
			 * anything, so it makes sense to check it
			 */
3407 3408 3409 3410
			if (sh->check_state == check_state_run)
				sh->check_state = check_state_run_pq;
			else
				sh->check_state = check_state_run_q;
3411 3412
		}

3413 3414
		/* discard potentially stale zero_sum_result */
		sh->ops.zero_sum_result = 0;
3415

3416 3417 3418 3419
		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--;
3420
		}
3421 3422 3423 3424 3425 3426 3427
		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;
3428 3429
		}

3430 3431 3432 3433 3434
		/* we have 2-disk failure */
		BUG_ON(s->failed != 2);
		/* fall through */
	case check_state_compute_result:
		sh->check_state = check_state_idle;
3435

3436 3437 3438
		/* check that a write has not made the stripe insync */
		if (test_bit(STRIPE_INSYNC, &sh->state))
			break;
3439 3440

		/* now write out any block on a failed drive,
3441
		 * or P or Q if they were recomputed
3442
		 */
3443
		BUG_ON(s->uptodate < disks - 1); /* We don't need Q to recover */
3444
		if (s->failed == 2) {
3445
			dev = &sh->dev[s->failed_num[1]];
3446 3447 3448 3449 3450
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
		if (s->failed >= 1) {
3451
			dev = &sh->dev[s->failed_num[0]];
3452 3453 3454 3455
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
3456
		if (sh->ops.zero_sum_result & SUM_CHECK_P_RESULT) {
3457 3458 3459 3460 3461
			dev = &sh->dev[pd_idx];
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
3462
		if (sh->ops.zero_sum_result & SUM_CHECK_Q_RESULT) {
3463 3464 3465 3466 3467 3468 3469 3470
			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);
3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499
		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 {
3500
			atomic64_add(STRIPE_SECTORS, &conf->mddev->resync_mismatches);
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 3532 3533 3534
			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();
3535 3536 3537
	}
}

3538
static void handle_stripe_expansion(struct r5conf *conf, struct stripe_head *sh)
3539 3540 3541 3542 3543 3544
{
	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.
	 */
3545
	struct dma_async_tx_descriptor *tx = NULL;
3546 3547
	clear_bit(STRIPE_EXPAND_SOURCE, &sh->state);
	for (i = 0; i < sh->disks; i++)
N
NeilBrown 已提交
3548
		if (i != sh->pd_idx && i != sh->qd_idx) {
3549
			int dd_idx, j;
3550
			struct stripe_head *sh2;
3551
			struct async_submit_ctl submit;
3552

3553
			sector_t bn = compute_blocknr(sh, i, 1);
3554 3555
			sector_t s = raid5_compute_sector(conf, bn, 0,
							  &dd_idx, NULL);
3556
			sh2 = get_active_stripe(conf, s, 0, 1, 1);
3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568
			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;
			}
3569 3570

			/* place all the copies on one channel */
3571
			init_async_submit(&submit, 0, tx, NULL, NULL, NULL);
3572
			tx = async_memcpy(sh2->dev[dd_idx].page,
3573
					  sh->dev[i].page, 0, 0, STRIPE_SIZE,
3574
					  &submit);
3575

3576 3577 3578 3579
			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 &&
3580
				    j != sh2->qd_idx &&
3581 3582 3583 3584 3585 3586 3587
				    !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);
3588

3589
		}
3590
	/* done submitting copies, wait for them to complete */
3591
	async_tx_quiesce(&tx);
3592
}
L
Linus Torvalds 已提交
3593 3594 3595 3596

/*
 * handle_stripe - do things to a stripe.
 *
3597 3598
 * 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
3599
 * Possible results:
3600 3601
 *    return some read requests which now have data
 *    return some write requests which are safely on storage
L
Linus Torvalds 已提交
3602 3603 3604 3605 3606
 *    schedule a read on some buffers
 *    schedule a write of some buffers
 *    return confirmation of parity correctness
 *
 */
3607

3608
static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
L
Linus Torvalds 已提交
3609
{
3610
	struct r5conf *conf = sh->raid_conf;
3611
	int disks = sh->disks;
3612 3613
	struct r5dev *dev;
	int i;
3614
	int do_recovery = 0;
L
Linus Torvalds 已提交
3615

3616 3617 3618 3619 3620 3621
	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
3622

3623
	/* Now to look around and see what can be done */
L
Linus Torvalds 已提交
3624
	rcu_read_lock();
3625
	for (i=disks; i--; ) {
3626
		struct md_rdev *rdev;
3627 3628 3629
		sector_t first_bad;
		int bad_sectors;
		int is_bad = 0;
3630

3631
		dev = &sh->dev[i];
L
Linus Torvalds 已提交
3632

3633
		pr_debug("check %d: state 0x%lx read %p write %p written %p\n",
3634 3635
			 i, dev->flags,
			 dev->toread, dev->towrite, dev->written);
3636 3637 3638 3639 3640 3641 3642 3643
		/* 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
3644

3645
		/* now count some things */
3646 3647 3648 3649
		if (test_bit(R5_LOCKED, &dev->flags))
			s->locked++;
		if (test_bit(R5_UPTODATE, &dev->flags))
			s->uptodate++;
3650
		if (test_bit(R5_Wantcompute, &dev->flags)) {
3651 3652
			s->compute++;
			BUG_ON(s->compute > 2);
3653
		}
L
Linus Torvalds 已提交
3654

3655
		if (test_bit(R5_Wantfill, &dev->flags))
3656
			s->to_fill++;
3657
		else if (dev->toread)
3658
			s->to_read++;
3659
		if (dev->towrite) {
3660
			s->to_write++;
3661
			if (!test_bit(R5_OVERWRITE, &dev->flags))
3662
				s->non_overwrite++;
3663
		}
3664
		if (dev->written)
3665
			s->written++;
3666 3667 3668 3669 3670 3671 3672 3673 3674 3675
		/* 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 {
3676 3677
			if (rdev)
				set_bit(R5_NeedReplace, &dev->flags);
3678 3679 3680
			rdev = rcu_dereference(conf->disks[i].rdev);
			clear_bit(R5_ReadRepl, &dev->flags);
		}
3681 3682
		if (rdev && test_bit(Faulty, &rdev->flags))
			rdev = NULL;
3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694
		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);
			}
3695
		}
3696 3697 3698
		clear_bit(R5_Insync, &dev->flags);
		if (!rdev)
			/* Not in-sync */;
3699 3700
		else if (is_bad) {
			/* also not in-sync */
3701 3702
			if (!test_bit(WriteErrorSeen, &rdev->flags) &&
			    test_bit(R5_UPTODATE, &dev->flags)) {
3703 3704 3705 3706 3707 3708 3709
				/* 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))
3710
			set_bit(R5_Insync, &dev->flags);
3711
		else if (sh->sector + STRIPE_SECTORS <= rdev->recovery_offset)
3712
			/* in sync if before recovery_offset */
3713 3714 3715 3716 3717 3718 3719 3720 3721
			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);

3722
		if (test_bit(R5_WriteError, &dev->flags)) {
3723 3724 3725 3726 3727 3728 3729
			/* 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)) {
3730
				s->handle_bad_blocks = 1;
3731
				atomic_inc(&rdev2->nr_pending);
3732 3733 3734
			} else
				clear_bit(R5_WriteError, &dev->flags);
		}
3735
		if (test_bit(R5_MadeGood, &dev->flags)) {
3736 3737 3738 3739 3740
			/* 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)) {
3741
				s->handle_bad_blocks = 1;
3742
				atomic_inc(&rdev2->nr_pending);
3743 3744 3745
			} else
				clear_bit(R5_MadeGood, &dev->flags);
		}
3746 3747 3748 3749 3750 3751 3752 3753 3754
		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);
		}
3755
		if (!test_bit(R5_Insync, &dev->flags)) {
3756 3757 3758
			/* 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
3759
		}
3760 3761 3762
		if (test_bit(R5_ReadError, &dev->flags))
			clear_bit(R5_Insync, &dev->flags);
		if (!test_bit(R5_Insync, &dev->flags)) {
3763 3764 3765
			if (s->failed < 2)
				s->failed_num[s->failed] = i;
			s->failed++;
3766 3767
			if (rdev && !test_bit(Faulty, &rdev->flags))
				do_recovery = 1;
3768
		}
L
Linus Torvalds 已提交
3769
	}
3770 3771 3772 3773
	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
3774
		 * else if MD_RECOVERY_REQUESTED is set, we also are syncing.
3775 3776 3777 3778 3779
		 * else we can only be replacing
		 * sync and recovery both need to read all devices, and so
		 * use the same flag.
		 */
		if (do_recovery ||
3780 3781
		    sh->sector >= conf->mddev->recovery_cp ||
		    test_bit(MD_RECOVERY_REQUESTED, &(conf->mddev->recovery)))
3782 3783 3784 3785
			s->syncing = 1;
		else
			s->replacing = 1;
	}
L
Linus Torvalds 已提交
3786
	rcu_read_unlock();
3787 3788 3789 3790 3791
}

static void handle_stripe(struct stripe_head *sh)
{
	struct stripe_head_state s;
3792
	struct r5conf *conf = sh->raid_conf;
3793
	int i;
3794 3795
	int prexor;
	int disks = sh->disks;
3796
	struct r5dev *pdev, *qdev;
3797 3798

	clear_bit(STRIPE_HANDLE, &sh->state);
3799
	if (test_and_set_bit_lock(STRIPE_ACTIVE, &sh->state)) {
3800 3801 3802 3803 3804 3805
		/* already being handled, ensure it gets handled
		 * again when current action finishes */
		set_bit(STRIPE_HANDLE, &sh->state);
		return;
	}

3806 3807 3808 3809 3810 3811 3812
	if (test_bit(STRIPE_SYNC_REQUESTED, &sh->state)) {
		spin_lock(&sh->stripe_lock);
		/* Cannot process 'sync' concurrently with 'discard' */
		if (!test_bit(STRIPE_DISCARD, &sh->state) &&
		    test_and_clear_bit(STRIPE_SYNC_REQUESTED, &sh->state)) {
			set_bit(STRIPE_SYNCING, &sh->state);
			clear_bit(STRIPE_INSYNC, &sh->state);
3813
			clear_bit(STRIPE_REPLACED, &sh->state);
3814 3815
		}
		spin_unlock(&sh->stripe_lock);
3816 3817 3818 3819 3820 3821 3822 3823
	}
	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);
3824

3825
	analyse_stripe(sh, &s);
3826

3827 3828 3829 3830 3831
	if (s.handle_bad_blocks) {
		set_bit(STRIPE_HANDLE, &sh->state);
		goto finish;
	}

3832 3833
	if (unlikely(s.blocked_rdev)) {
		if (s.syncing || s.expanding || s.expanded ||
3834
		    s.replacing || s.to_write || s.written) {
3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854
			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.
	 */
3855 3856 3857 3858 3859
	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);
3860
		if (s.syncing + s.replacing)
3861 3862
			handle_failed_sync(conf, sh, &s);
	}
3863

3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876
	/* 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
		 */
3877 3878
		BUG_ON(!test_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags) &&
		       !test_bit(R5_Discard, &sh->dev[sh->pd_idx].flags));
3879
		BUG_ON(sh->qd_idx >= 0 &&
3880 3881
		       !test_bit(R5_UPTODATE, &sh->dev[sh->qd_idx].flags) &&
		       !test_bit(R5_Discard, &sh->dev[sh->qd_idx].flags));
3882 3883 3884 3885 3886 3887 3888 3889 3890
		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;
3891 3892
				if (s.failed > 1)
					continue;
3893 3894 3895 3896 3897 3898 3899 3900 3901 3902
				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;
	}

3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936
	/*
	 * might be able to return some write requests if the parity blocks
	 * are safe, or on a failed drive
	 */
	pdev = &sh->dev[sh->pd_idx];
	s.p_failed = (s.failed >= 1 && s.failed_num[0] == sh->pd_idx)
		|| (s.failed >= 2 && s.failed_num[1] == sh->pd_idx);
	qdev = &sh->dev[sh->qd_idx];
	s.q_failed = (s.failed >= 1 && s.failed_num[0] == sh->qd_idx)
		|| (s.failed >= 2 && s.failed_num[1] == sh->qd_idx)
		|| conf->level < 6;

	if (s.written &&
	    (s.p_failed || ((test_bit(R5_Insync, &pdev->flags)
			     && !test_bit(R5_LOCKED, &pdev->flags)
			     && (test_bit(R5_UPTODATE, &pdev->flags) ||
				 test_bit(R5_Discard, &pdev->flags))))) &&
	    (s.q_failed || ((test_bit(R5_Insync, &qdev->flags)
			     && !test_bit(R5_LOCKED, &qdev->flags)
			     && (test_bit(R5_UPTODATE, &qdev->flags) ||
				 test_bit(R5_Discard, &qdev->flags))))))
		handle_stripe_clean_event(conf, sh, disks, &s.return_bi);

	/* Now we might consider reading some blocks, either to check/generate
	 * parity, or to satisfy requests
	 * or to load a block that is being partially written.
	 */
	if (s.to_read || s.non_overwrite
	    || (conf->level == 6 && s.to_write && s.failed)
	    || (s.syncing && (s.uptodate + s.compute < disks))
	    || s.replacing
	    || s.expanding)
		handle_stripe_fill(sh, &s, disks);

3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959
	/* 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);
	}
3960

3961 3962 3963
	if ((s.replacing || s.syncing) && s.locked == 0
	    && !test_bit(STRIPE_COMPUTE_RUN, &sh->state)
	    && !test_bit(STRIPE_REPLACED, &sh->state)) {
3964 3965
		/* Write out to replacement devices where possible */
		for (i = 0; i < conf->raid_disks; i++)
3966 3967
			if (test_bit(R5_NeedReplace, &sh->dev[i].flags)) {
				WARN_ON(!test_bit(R5_UPTODATE, &sh->dev[i].flags));
3968 3969 3970 3971
				set_bit(R5_WantReplace, &sh->dev[i].flags);
				set_bit(R5_LOCKED, &sh->dev[i].flags);
				s.locked++;
			}
3972 3973 3974
		if (s.replacing)
			set_bit(STRIPE_INSYNC, &sh->state);
		set_bit(STRIPE_REPLACED, &sh->state);
3975 3976
	}
	if ((s.syncing || s.replacing) && s.locked == 0 &&
3977
	    !test_bit(STRIPE_COMPUTE_RUN, &sh->state) &&
3978
	    test_bit(STRIPE_INSYNC, &sh->state)) {
3979 3980
		md_done_sync(conf->mddev, STRIPE_SECTORS, 1);
		clear_bit(STRIPE_SYNCING, &sh->state);
3981 3982
		if (test_and_clear_bit(R5_Overlap, &sh->dev[sh->pd_idx].flags))
			wake_up(&conf->wait_for_overlap);
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
	}

	/* 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++;
				}
			}
		}

4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035
	/* 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++;
		}
	}
4036

4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052
	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);
4053

4054
finish:
4055
	/* wait for this device to become unblocked */
4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067
	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);
	}
4068

4069 4070
	if (s.handle_bad_blocks)
		for (i = disks; i--; ) {
4071
			struct md_rdev *rdev;
4072 4073 4074 4075 4076 4077 4078 4079 4080
			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);
			}
4081 4082 4083
			if (test_and_clear_bit(R5_MadeGood, &dev->flags)) {
				rdev = conf->disks[i].rdev;
				rdev_clear_badblocks(rdev, sh->sector,
4084
						     STRIPE_SECTORS, 0);
4085 4086
				rdev_dec_pending(rdev, conf->mddev);
			}
4087 4088
			if (test_and_clear_bit(R5_MadeGoodRepl, &dev->flags)) {
				rdev = conf->disks[i].replacement;
4089 4090 4091
				if (!rdev)
					/* rdev have been moved down */
					rdev = conf->disks[i].rdev;
4092
				rdev_clear_badblocks(rdev, sh->sector,
4093
						     STRIPE_SECTORS, 0);
4094 4095
				rdev_dec_pending(rdev, conf->mddev);
			}
4096 4097
		}

4098 4099 4100
	if (s.ops_request)
		raid_run_ops(sh, s.ops_request);

D
Dan Williams 已提交
4101
	ops_run_io(sh, &s);
4102

4103
	if (s.dec_preread_active) {
4104
		/* We delay this until after ops_run_io so that if make_request
T
Tejun Heo 已提交
4105
		 * is waiting on a flush, it won't continue until the writes
4106 4107 4108 4109 4110 4111 4112 4113
		 * 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);
	}

4114
	return_io(s.return_bi);
4115

4116
	clear_bit_unlock(STRIPE_ACTIVE, &sh->state);
4117 4118
}

4119
static void raid5_activate_delayed(struct r5conf *conf)
4120 4121 4122 4123 4124 4125 4126 4127 4128 4129
{
	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);
4130
			list_add_tail(&sh->lru, &conf->hold_list);
4131
			raid5_wakeup_stripe_thread(sh);
4132
		}
N
NeilBrown 已提交
4133
	}
4134 4135
}

4136 4137
static void activate_bit_delay(struct r5conf *conf,
	struct list_head *temp_inactive_list)
4138 4139 4140 4141 4142 4143 4144
{
	/* 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);
4145
		int hash;
4146 4147
		list_del_init(&sh->lru);
		atomic_inc(&sh->count);
4148 4149
		hash = sh->hash_lock_index;
		__release_stripe(conf, sh, &temp_inactive_list[hash]);
4150 4151 4152
	}
}

4153
static int raid5_congested(struct mddev *mddev, int bits)
4154
{
4155
	struct r5conf *conf = mddev->private;
4156 4157 4158 4159

	/* No difference between reads and writes.  Just check
	 * how busy the stripe_cache is
	 */
4160

4161 4162 4163 4164
	if (conf->inactive_blocked)
		return 1;
	if (conf->quiesce)
		return 1;
4165
	if (atomic_read(&conf->empty_inactive_list_nr))
4166 4167 4168 4169 4170
		return 1;

	return 0;
}

4171 4172 4173
/* We want read requests to align with chunks where possible,
 * but write requests don't need to.
 */
4174
static int raid5_mergeable_bvec(struct mddev *mddev,
4175 4176
				struct bvec_merge_data *bvm,
				struct bio_vec *biovec)
4177
{
4178
	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
4179
	int max;
4180
	unsigned int chunk_sectors = mddev->chunk_sectors;
4181
	unsigned int bio_sectors = bvm->bi_size >> 9;
4182

4183
	if ((bvm->bi_rw & 1) == WRITE)
4184 4185
		return biovec->bv_len; /* always allow writes to be mergeable */

4186 4187
	if (mddev->new_chunk_sectors < mddev->chunk_sectors)
		chunk_sectors = mddev->new_chunk_sectors;
4188 4189 4190 4191 4192 4193 4194 4195
	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;
}

4196
static int in_chunk_boundary(struct mddev *mddev, struct bio *bio)
4197
{
4198
	sector_t sector = bio->bi_iter.bi_sector + get_start_sect(bio->bi_bdev);
4199
	unsigned int chunk_sectors = mddev->chunk_sectors;
4200
	unsigned int bio_sectors = bio_sectors(bio);
4201

4202 4203
	if (mddev->new_chunk_sectors < mddev->chunk_sectors)
		chunk_sectors = mddev->new_chunk_sectors;
4204 4205 4206 4207
	return  chunk_sectors >=
		((sector & (chunk_sectors - 1)) + bio_sectors);
}

4208 4209 4210 4211
/*
 *  add bio to the retry LIFO  ( in O(1) ... we are in interrupt )
 *  later sampled by raid5d.
 */
4212
static void add_bio_to_retry(struct bio *bi,struct r5conf *conf)
4213 4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224
{
	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);
}

4225
static struct bio *remove_bio_from_retry(struct r5conf *conf)
4226 4227 4228 4229 4230 4231 4232 4233 4234 4235
{
	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) {
4236
		conf->retry_read_aligned_list = bi->bi_next;
4237
		bi->bi_next = NULL;
4238 4239 4240 4241
		/*
		 * this sets the active strip count to 1 and the processed
		 * strip count to zero (upper 8 bits)
		 */
4242
		raid5_set_bi_stripes(bi, 1); /* biased count of active stripes */
4243 4244 4245 4246 4247
	}

	return bi;
}

4248 4249 4250 4251 4252 4253
/*
 *  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..
 */
4254
static void raid5_align_endio(struct bio *bi, int error)
4255 4256
{
	struct bio* raid_bi  = bi->bi_private;
4257
	struct mddev *mddev;
4258
	struct r5conf *conf;
4259
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
4260
	struct md_rdev *rdev;
4261

4262
	bio_put(bi);
4263 4264 4265

	rdev = (void*)raid_bi->bi_next;
	raid_bi->bi_next = NULL;
4266 4267
	mddev = rdev->mddev;
	conf = mddev->private;
4268 4269 4270 4271

	rdev_dec_pending(rdev, conf->mddev);

	if (!error && uptodate) {
4272 4273
		trace_block_bio_complete(bdev_get_queue(raid_bi->bi_bdev),
					 raid_bi, 0);
4274
		bio_endio(raid_bi, 0);
4275 4276
		if (atomic_dec_and_test(&conf->active_aligned_reads))
			wake_up(&conf->wait_for_stripe);
4277
		return;
4278 4279
	}

4280
	pr_debug("raid5_align_endio : io error...handing IO for a retry\n");
4281 4282

	add_bio_to_retry(raid_bi, conf);
4283 4284
}

4285 4286
static int bio_fits_rdev(struct bio *bi)
{
4287
	struct request_queue *q = bdev_get_queue(bi->bi_bdev);
4288

4289
	if (bio_sectors(bi) > queue_max_sectors(q))
4290 4291
		return 0;
	blk_recount_segments(q, bi);
4292
	if (bi->bi_phys_segments > queue_max_segments(q))
4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303
		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;
}

4304
static int chunk_aligned_read(struct mddev *mddev, struct bio * raid_bio)
4305
{
4306
	struct r5conf *conf = mddev->private;
N
NeilBrown 已提交
4307
	int dd_idx;
4308
	struct bio* align_bi;
4309
	struct md_rdev *rdev;
4310
	sector_t end_sector;
4311 4312

	if (!in_chunk_boundary(mddev, raid_bio)) {
4313
		pr_debug("chunk_aligned_read : non aligned\n");
4314 4315 4316
		return 0;
	}
	/*
4317
	 * use bio_clone_mddev to make a copy of the bio
4318
	 */
4319
	align_bi = bio_clone_mddev(raid_bio, GFP_NOIO, mddev);
4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330
	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
	 */
4331 4332 4333
	align_bi->bi_iter.bi_sector =
		raid5_compute_sector(conf, raid_bio->bi_iter.bi_sector,
				     0, &dd_idx, NULL);
4334

K
Kent Overstreet 已提交
4335
	end_sector = bio_end_sector(align_bi);
4336
	rcu_read_lock();
4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347
	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) {
4348 4349 4350
		sector_t first_bad;
		int bad_sectors;

4351 4352
		atomic_inc(&rdev->nr_pending);
		rcu_read_unlock();
4353 4354
		raid_bio->bi_next = (void*)rdev;
		align_bi->bi_bdev =  rdev->bdev;
4355
		__clear_bit(BIO_SEG_VALID, &align_bi->bi_flags);
4356

4357
		if (!bio_fits_rdev(align_bi) ||
4358 4359
		    is_badblock(rdev, align_bi->bi_iter.bi_sector,
				bio_sectors(align_bi),
4360 4361
				&first_bad, &bad_sectors)) {
			/* too big in some way, or has a known bad block */
4362 4363 4364 4365 4366
			bio_put(align_bi);
			rdev_dec_pending(rdev, mddev);
			return 0;
		}

4367
		/* No reshape active, so we can trust rdev->data_offset */
4368
		align_bi->bi_iter.bi_sector += rdev->data_offset;
4369

4370 4371 4372
		spin_lock_irq(&conf->device_lock);
		wait_event_lock_irq(conf->wait_for_stripe,
				    conf->quiesce == 0,
4373
				    conf->device_lock);
4374 4375 4376
		atomic_inc(&conf->active_aligned_reads);
		spin_unlock_irq(&conf->device_lock);

4377 4378 4379
		if (mddev->gendisk)
			trace_block_bio_remap(bdev_get_queue(align_bi->bi_bdev),
					      align_bi, disk_devt(mddev->gendisk),
4380
					      raid_bio->bi_iter.bi_sector);
4381 4382 4383 4384
		generic_make_request(align_bi);
		return 1;
	} else {
		rcu_read_unlock();
4385
		bio_put(align_bi);
4386 4387 4388 4389
		return 0;
	}
}

4390 4391 4392 4393 4394 4395 4396 4397 4398 4399
/* __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.
 */
4400
static struct stripe_head *__get_priority_stripe(struct r5conf *conf, int group)
4401
{
4402 4403
	struct stripe_head *sh = NULL, *tmp;
	struct list_head *handle_list = NULL;
4404
	struct r5worker_group *wg = NULL;
4405 4406 4407 4408 4409

	if (conf->worker_cnt_per_group == 0) {
		handle_list = &conf->handle_list;
	} else if (group != ANY_GROUP) {
		handle_list = &conf->worker_groups[group].handle_list;
4410
		wg = &conf->worker_groups[group];
4411 4412 4413 4414
	} else {
		int i;
		for (i = 0; i < conf->group_cnt; i++) {
			handle_list = &conf->worker_groups[i].handle_list;
4415
			wg = &conf->worker_groups[i];
4416 4417 4418 4419
			if (!list_empty(handle_list))
				break;
		}
	}
4420 4421 4422

	pr_debug("%s: handle: %s hold: %s full_writes: %d bypass_count: %d\n",
		  __func__,
4423
		  list_empty(handle_list) ? "empty" : "busy",
4424 4425 4426
		  list_empty(&conf->hold_list) ? "empty" : "busy",
		  atomic_read(&conf->pending_full_writes), conf->bypass_count);

4427 4428
	if (!list_empty(handle_list)) {
		sh = list_entry(handle_list->next, typeof(*sh), lru);
4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445

		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)) {
4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461

		list_for_each_entry(tmp, &conf->hold_list,  lru) {
			if (conf->worker_cnt_per_group == 0 ||
			    group == ANY_GROUP ||
			    !cpu_online(tmp->cpu) ||
			    cpu_to_group(tmp->cpu) == group) {
				sh = tmp;
				break;
			}
		}

		if (sh) {
			conf->bypass_count -= conf->bypass_threshold;
			if (conf->bypass_count < 0)
				conf->bypass_count = 0;
		}
4462
		wg = NULL;
4463 4464 4465
	}

	if (!sh)
4466 4467
		return NULL;

4468 4469 4470 4471
	if (wg) {
		wg->stripes_cnt--;
		sh->group = NULL;
	}
4472
	list_del_init(&sh->lru);
4473
	BUG_ON(atomic_inc_return(&sh->count) != 1);
4474 4475
	return sh;
}
4476

4477 4478 4479
struct raid5_plug_cb {
	struct blk_plug_cb	cb;
	struct list_head	list;
4480
	struct list_head	temp_inactive_list[NR_STRIPE_HASH_LOCKS];
4481 4482 4483 4484 4485 4486 4487 4488 4489
};

static void raid5_unplug(struct blk_plug_cb *blk_cb, bool from_schedule)
{
	struct raid5_plug_cb *cb = container_of(
		blk_cb, struct raid5_plug_cb, cb);
	struct stripe_head *sh;
	struct mddev *mddev = cb->cb.data;
	struct r5conf *conf = mddev->private;
N
NeilBrown 已提交
4490
	int cnt = 0;
4491
	int hash;
4492 4493 4494 4495 4496 4497 4498 4499 4500 4501 4502

	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
			 */
4503
			smp_mb__before_atomic();
4504
			clear_bit(STRIPE_ON_UNPLUG_LIST, &sh->state);
S
Shaohua Li 已提交
4505 4506 4507 4508
			/*
			 * STRIPE_ON_RELEASE_LIST could be set here. In that
			 * case, the count is always > 1 here
			 */
4509 4510
			hash = sh->hash_lock_index;
			__release_stripe(conf, sh, &cb->temp_inactive_list[hash]);
N
NeilBrown 已提交
4511
			cnt++;
4512 4513 4514
		}
		spin_unlock_irq(&conf->device_lock);
	}
4515 4516
	release_inactive_stripe_list(conf, cb->temp_inactive_list,
				     NR_STRIPE_HASH_LOCKS);
4517 4518
	if (mddev->queue)
		trace_block_unplug(mddev->queue, cnt, !from_schedule);
4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536
	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);

4537 4538
	if (cb->list.next == NULL) {
		int i;
4539
		INIT_LIST_HEAD(&cb->list);
4540 4541 4542
		for (i = 0; i < NR_STRIPE_HASH_LOCKS; i++)
			INIT_LIST_HEAD(cb->temp_inactive_list + i);
	}
4543 4544 4545 4546 4547 4548 4549

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

S
Shaohua Li 已提交
4550 4551 4552 4553 4554 4555 4556 4557 4558 4559 4560 4561
static void make_discard_request(struct mddev *mddev, struct bio *bi)
{
	struct r5conf *conf = mddev->private;
	sector_t logical_sector, last_sector;
	struct stripe_head *sh;
	int remaining;
	int stripe_sectors;

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

4562 4563
	logical_sector = bi->bi_iter.bi_sector & ~((sector_t)STRIPE_SECTORS-1);
	last_sector = bi->bi_iter.bi_sector + (bi->bi_iter.bi_size>>9);
S
Shaohua Li 已提交
4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584

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

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

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

	for (; logical_sector < last_sector;
	     logical_sector += STRIPE_SECTORS) {
		DEFINE_WAIT(w);
		int d;
	again:
		sh = get_active_stripe(conf, logical_sector, 0, 0, 0);
		prepare_to_wait(&conf->wait_for_overlap, &w,
				TASK_UNINTERRUPTIBLE);
4585 4586 4587 4588 4589 4590 4591
		set_bit(R5_Overlap, &sh->dev[sh->pd_idx].flags);
		if (test_bit(STRIPE_SYNCING, &sh->state)) {
			release_stripe(sh);
			schedule();
			goto again;
		}
		clear_bit(R5_Overlap, &sh->dev[sh->pd_idx].flags);
S
Shaohua Li 已提交
4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603
		spin_lock_irq(&sh->stripe_lock);
		for (d = 0; d < conf->raid_disks; d++) {
			if (d == sh->pd_idx || d == sh->qd_idx)
				continue;
			if (sh->dev[d].towrite || sh->dev[d].toread) {
				set_bit(R5_Overlap, &sh->dev[d].flags);
				spin_unlock_irq(&sh->stripe_lock);
				release_stripe(sh);
				schedule();
				goto again;
			}
		}
4604
		set_bit(STRIPE_DISCARD, &sh->state);
S
Shaohua Li 已提交
4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639
		finish_wait(&conf->wait_for_overlap, &w);
		for (d = 0; d < conf->raid_disks; d++) {
			if (d == sh->pd_idx || d == sh->qd_idx)
				continue;
			sh->dev[d].towrite = bi;
			set_bit(R5_OVERWRITE, &sh->dev[d].flags);
			raid5_inc_bi_active_stripes(bi);
		}
		spin_unlock_irq(&sh->stripe_lock);
		if (conf->mddev->bitmap) {
			for (d = 0;
			     d < conf->raid_disks - conf->max_degraded;
			     d++)
				bitmap_startwrite(mddev->bitmap,
						  sh->sector,
						  STRIPE_SECTORS,
						  0);
			sh->bm_seq = conf->seq_flush + 1;
			set_bit(STRIPE_BIT_DELAY, &sh->state);
		}

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

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

4640
static void make_request(struct mddev *mddev, struct bio * bi)
L
Linus Torvalds 已提交
4641
{
4642
	struct r5conf *conf = mddev->private;
4643
	int dd_idx;
L
Linus Torvalds 已提交
4644 4645 4646
	sector_t new_sector;
	sector_t logical_sector, last_sector;
	struct stripe_head *sh;
4647
	const int rw = bio_data_dir(bi);
4648
	int remaining;
4649 4650
	DEFINE_WAIT(w);
	bool do_prepare;
L
Linus Torvalds 已提交
4651

T
Tejun Heo 已提交
4652 4653
	if (unlikely(bi->bi_rw & REQ_FLUSH)) {
		md_flush_request(mddev, bi);
4654
		return;
4655 4656
	}

4657
	md_write_start(mddev, bi);
4658

4659
	if (rw == READ &&
4660
	     mddev->reshape_position == MaxSector &&
4661
	     chunk_aligned_read(mddev,bi))
4662
		return;
4663

S
Shaohua Li 已提交
4664 4665 4666 4667 4668
	if (unlikely(bi->bi_rw & REQ_DISCARD)) {
		make_discard_request(mddev, bi);
		return;
	}

4669
	logical_sector = bi->bi_iter.bi_sector & ~((sector_t)STRIPE_SECTORS-1);
K
Kent Overstreet 已提交
4670
	last_sector = bio_end_sector(bi);
L
Linus Torvalds 已提交
4671 4672
	bi->bi_next = NULL;
	bi->bi_phys_segments = 1;	/* over-loaded to count active stripes */
4673

4674
	prepare_to_wait(&conf->wait_for_overlap, &w, TASK_UNINTERRUPTIBLE);
L
Linus Torvalds 已提交
4675
	for (;logical_sector < last_sector; logical_sector += STRIPE_SECTORS) {
4676
		int previous;
4677
		int seq;
4678

4679
		do_prepare = false;
4680
	retry:
4681
		seq = read_seqcount_begin(&conf->gen_lock);
4682
		previous = 0;
4683 4684 4685
		if (do_prepare)
			prepare_to_wait(&conf->wait_for_overlap, &w,
				TASK_UNINTERRUPTIBLE);
4686
		if (unlikely(conf->reshape_progress != MaxSector)) {
4687
			/* spinlock is needed as reshape_progress may be
4688 4689
			 * 64bit on a 32bit platform, and so it might be
			 * possible to see a half-updated value
4690
			 * Of course reshape_progress could change after
4691 4692 4693 4694
			 * the lock is dropped, so once we get a reference
			 * to the stripe that we think it is, we will have
			 * to check again.
			 */
4695
			spin_lock_irq(&conf->device_lock);
4696
			if (mddev->reshape_backwards
4697 4698
			    ? logical_sector < conf->reshape_progress
			    : logical_sector >= conf->reshape_progress) {
4699 4700
				previous = 1;
			} else {
4701
				if (mddev->reshape_backwards
4702 4703
				    ? logical_sector < conf->reshape_safe
				    : logical_sector >= conf->reshape_safe) {
4704 4705
					spin_unlock_irq(&conf->device_lock);
					schedule();
4706
					do_prepare = true;
4707 4708 4709
					goto retry;
				}
			}
4710 4711
			spin_unlock_irq(&conf->device_lock);
		}
4712

4713 4714
		new_sector = raid5_compute_sector(conf, logical_sector,
						  previous,
4715
						  &dd_idx, NULL);
4716
		pr_debug("raid456: make_request, sector %llu logical %llu\n",
4717
			(unsigned long long)new_sector,
L
Linus Torvalds 已提交
4718 4719
			(unsigned long long)logical_sector);

4720
		sh = get_active_stripe(conf, new_sector, previous,
4721
				       (bi->bi_rw&RWA_MASK), 0);
L
Linus Torvalds 已提交
4722
		if (sh) {
4723
			if (unlikely(previous)) {
4724
				/* expansion might have moved on while waiting for a
4725 4726 4727 4728 4729 4730
				 * 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.
4731 4732 4733
				 */
				int must_retry = 0;
				spin_lock_irq(&conf->device_lock);
4734
				if (mddev->reshape_backwards
4735 4736
				    ? logical_sector >= conf->reshape_progress
				    : logical_sector < conf->reshape_progress)
4737 4738 4739 4740 4741
					/* mismatch, need to try again */
					must_retry = 1;
				spin_unlock_irq(&conf->device_lock);
				if (must_retry) {
					release_stripe(sh);
4742
					schedule();
4743
					do_prepare = true;
4744 4745 4746
					goto retry;
				}
			}
4747 4748 4749 4750 4751 4752 4753
			if (read_seqcount_retry(&conf->gen_lock, seq)) {
				/* Might have got the wrong stripe_head
				 * by accident
				 */
				release_stripe(sh);
				goto retry;
			}
4754

4755
			if (rw == WRITE &&
4756
			    logical_sector >= mddev->suspend_lo &&
4757 4758
			    logical_sector < mddev->suspend_hi) {
				release_stripe(sh);
4759 4760 4761 4762 4763 4764 4765 4766
				/* 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 &&
4767
				    logical_sector < mddev->suspend_hi) {
4768
					schedule();
4769 4770
					do_prepare = true;
				}
4771 4772
				goto retry;
			}
4773 4774

			if (test_bit(STRIPE_EXPANDING, &sh->state) ||
4775
			    !add_stripe_bio(sh, bi, dd_idx, rw)) {
4776 4777
				/* Stripe is busy expanding or
				 * add failed due to overlap.  Flush everything
L
Linus Torvalds 已提交
4778 4779
				 * and wait a while
				 */
N
NeilBrown 已提交
4780
				md_wakeup_thread(mddev->thread);
L
Linus Torvalds 已提交
4781 4782
				release_stripe(sh);
				schedule();
4783
				do_prepare = true;
L
Linus Torvalds 已提交
4784 4785
				goto retry;
			}
4786 4787
			set_bit(STRIPE_HANDLE, &sh->state);
			clear_bit(STRIPE_DELAYED, &sh->state);
4788
			if ((bi->bi_rw & REQ_SYNC) &&
4789 4790
			    !test_and_set_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
				atomic_inc(&conf->preread_active_stripes);
4791
			release_stripe_plug(mddev, sh);
L
Linus Torvalds 已提交
4792 4793 4794 4795 4796 4797
		} else {
			/* cannot get stripe for read-ahead, just give-up */
			clear_bit(BIO_UPTODATE, &bi->bi_flags);
			break;
		}
	}
4798
	finish_wait(&conf->wait_for_overlap, &w);
4799

4800
	remaining = raid5_dec_bi_active_stripes(bi);
4801
	if (remaining == 0) {
L
Linus Torvalds 已提交
4802

4803
		if ( rw == WRITE )
L
Linus Torvalds 已提交
4804
			md_write_end(mddev);
4805

4806 4807
		trace_block_bio_complete(bdev_get_queue(bi->bi_bdev),
					 bi, 0);
4808
		bio_endio(bi, 0);
L
Linus Torvalds 已提交
4809 4810 4811
	}
}

4812
static sector_t raid5_size(struct mddev *mddev, sector_t sectors, int raid_disks);
D
Dan Williams 已提交
4813

4814
static sector_t reshape_request(struct mddev *mddev, sector_t sector_nr, int *skipped)
L
Linus Torvalds 已提交
4815
{
4816 4817 4818 4819 4820 4821 4822 4823 4824
	/* 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.
	 */
4825
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
4826
	struct stripe_head *sh;
4827
	sector_t first_sector, last_sector;
4828 4829 4830
	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;
4831 4832
	int i;
	int dd_idx;
4833
	sector_t writepos, readpos, safepos;
4834
	sector_t stripe_addr;
4835
	int reshape_sectors;
4836
	struct list_head stripes;
4837

4838 4839
	if (sector_nr == 0) {
		/* If restarting in the middle, skip the initial sectors */
4840
		if (mddev->reshape_backwards &&
4841 4842 4843
		    conf->reshape_progress < raid5_size(mddev, 0, 0)) {
			sector_nr = raid5_size(mddev, 0, 0)
				- conf->reshape_progress;
4844
		} else if (!mddev->reshape_backwards &&
4845 4846
			   conf->reshape_progress > 0)
			sector_nr = conf->reshape_progress;
4847
		sector_div(sector_nr, new_data_disks);
4848
		if (sector_nr) {
4849 4850
			mddev->curr_resync_completed = sector_nr;
			sysfs_notify(&mddev->kobj, NULL, "sync_completed");
4851 4852 4853
			*skipped = 1;
			return sector_nr;
		}
4854 4855
	}

4856 4857 4858 4859
	/* 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
	 */
4860 4861
	if (mddev->new_chunk_sectors > mddev->chunk_sectors)
		reshape_sectors = mddev->new_chunk_sectors;
4862
	else
4863
		reshape_sectors = mddev->chunk_sectors;
4864

4865 4866 4867 4868 4869
	/* 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
4870
	 */
4871
	writepos = conf->reshape_progress;
4872
	sector_div(writepos, new_data_disks);
4873 4874
	readpos = conf->reshape_progress;
	sector_div(readpos, data_disks);
4875
	safepos = conf->reshape_safe;
4876
	sector_div(safepos, data_disks);
4877
	if (mddev->reshape_backwards) {
4878
		writepos -= min_t(sector_t, reshape_sectors, writepos);
4879
		readpos += reshape_sectors;
4880
		safepos += reshape_sectors;
4881
	} else {
4882
		writepos += reshape_sectors;
4883 4884
		readpos -= min_t(sector_t, reshape_sectors, readpos);
		safepos -= min_t(sector_t, reshape_sectors, safepos);
4885
	}
4886

4887 4888 4889 4890 4891 4892 4893 4894 4895 4896 4897 4898 4899 4900 4901
	/* 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;
	}

4902 4903 4904 4905
	/* '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.
4906 4907 4908 4909
	 * 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
4910 4911 4912 4913 4914 4915 4916 4917 4918 4919 4920 4921
	 * 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???
	 */
4922 4923 4924 4925 4926 4927
	if (conf->min_offset_diff < 0) {
		safepos += -conf->min_offset_diff;
		readpos += -conf->min_offset_diff;
	} else
		writepos += conf->min_offset_diff;

4928
	if ((mddev->reshape_backwards
4929 4930 4931
	     ? (safepos > writepos && readpos < writepos)
	     : (safepos < writepos && readpos > writepos)) ||
	    time_after(jiffies, conf->reshape_checkpoint + 10*HZ)) {
4932 4933
		/* Cannot proceed until we've updated the superblock... */
		wait_event(conf->wait_for_overlap,
4934 4935 4936 4937
			   atomic_read(&conf->reshape_stripes)==0
			   || test_bit(MD_RECOVERY_INTR, &mddev->recovery));
		if (atomic_read(&conf->reshape_stripes) != 0)
			return 0;
4938
		mddev->reshape_position = conf->reshape_progress;
4939
		mddev->curr_resync_completed = sector_nr;
4940
		conf->reshape_checkpoint = jiffies;
4941
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
4942
		md_wakeup_thread(mddev->thread);
4943
		wait_event(mddev->sb_wait, mddev->flags == 0 ||
4944 4945 4946
			   test_bit(MD_RECOVERY_INTR, &mddev->recovery));
		if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
			return 0;
4947
		spin_lock_irq(&conf->device_lock);
4948
		conf->reshape_safe = mddev->reshape_position;
4949 4950
		spin_unlock_irq(&conf->device_lock);
		wake_up(&conf->wait_for_overlap);
4951
		sysfs_notify(&mddev->kobj, NULL, "sync_completed");
4952 4953
	}

4954
	INIT_LIST_HEAD(&stripes);
4955
	for (i = 0; i < reshape_sectors; i += STRIPE_SECTORS) {
4956
		int j;
4957
		int skipped_disk = 0;
4958
		sh = get_active_stripe(conf, stripe_addr+i, 0, 0, 1);
4959 4960 4961 4962 4963 4964 4965 4966 4967
		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;
4968
			if (conf->level == 6 &&
4969
			    j == sh->qd_idx)
4970
				continue;
4971
			s = compute_blocknr(sh, j, 0);
D
Dan Williams 已提交
4972
			if (s < raid5_size(mddev, 0, 0)) {
4973
				skipped_disk = 1;
4974 4975 4976 4977 4978 4979
				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);
		}
4980
		if (!skipped_disk) {
4981 4982 4983
			set_bit(STRIPE_EXPAND_READY, &sh->state);
			set_bit(STRIPE_HANDLE, &sh->state);
		}
4984
		list_add(&sh->lru, &stripes);
4985 4986
	}
	spin_lock_irq(&conf->device_lock);
4987
	if (mddev->reshape_backwards)
4988
		conf->reshape_progress -= reshape_sectors * new_data_disks;
4989
	else
4990
		conf->reshape_progress += reshape_sectors * new_data_disks;
4991 4992 4993 4994 4995 4996 4997
	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 =
4998
		raid5_compute_sector(conf, stripe_addr*(new_data_disks),
4999
				     1, &dd_idx, NULL);
5000
	last_sector =
5001
		raid5_compute_sector(conf, ((stripe_addr+reshape_sectors)
5002
					    * new_data_disks - 1),
5003
				     1, &dd_idx, NULL);
A
Andre Noll 已提交
5004 5005
	if (last_sector >= mddev->dev_sectors)
		last_sector = mddev->dev_sectors - 1;
5006
	while (first_sector <= last_sector) {
5007
		sh = get_active_stripe(conf, first_sector, 1, 0, 1);
5008 5009 5010 5011 5012
		set_bit(STRIPE_EXPAND_SOURCE, &sh->state);
		set_bit(STRIPE_HANDLE, &sh->state);
		release_stripe(sh);
		first_sector += STRIPE_SECTORS;
	}
5013 5014 5015 5016 5017 5018 5019 5020
	/* 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);
	}
5021 5022 5023
	/* If this takes us to the resync_max point where we have to pause,
	 * then we need to write out the superblock.
	 */
5024
	sector_nr += reshape_sectors;
5025 5026
	if ((sector_nr - mddev->curr_resync_completed) * 2
	    >= mddev->resync_max - mddev->curr_resync_completed) {
5027 5028
		/* Cannot proceed until we've updated the superblock... */
		wait_event(conf->wait_for_overlap,
5029 5030 5031 5032
			   atomic_read(&conf->reshape_stripes) == 0
			   || test_bit(MD_RECOVERY_INTR, &mddev->recovery));
		if (atomic_read(&conf->reshape_stripes) != 0)
			goto ret;
5033
		mddev->reshape_position = conf->reshape_progress;
5034
		mddev->curr_resync_completed = sector_nr;
5035
		conf->reshape_checkpoint = jiffies;
5036 5037 5038 5039
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
		md_wakeup_thread(mddev->thread);
		wait_event(mddev->sb_wait,
			   !test_bit(MD_CHANGE_DEVS, &mddev->flags)
5040 5041 5042
			   || test_bit(MD_RECOVERY_INTR, &mddev->recovery));
		if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
			goto ret;
5043
		spin_lock_irq(&conf->device_lock);
5044
		conf->reshape_safe = mddev->reshape_position;
5045 5046
		spin_unlock_irq(&conf->device_lock);
		wake_up(&conf->wait_for_overlap);
5047
		sysfs_notify(&mddev->kobj, NULL, "sync_completed");
5048
	}
5049
ret:
5050
	return reshape_sectors;
5051 5052
}

5053
static inline sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipped)
5054
{
5055
	struct r5conf *conf = mddev->private;
5056
	struct stripe_head *sh;
A
Andre Noll 已提交
5057
	sector_t max_sector = mddev->dev_sectors;
N
NeilBrown 已提交
5058
	sector_t sync_blocks;
5059 5060
	int still_degraded = 0;
	int i;
L
Linus Torvalds 已提交
5061

5062
	if (sector_nr >= max_sector) {
L
Linus Torvalds 已提交
5063
		/* just being told to finish up .. nothing much to do */
5064

5065 5066 5067 5068
		if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) {
			end_reshape(conf);
			return 0;
		}
5069 5070 5071 5072

		if (mddev->curr_resync < max_sector) /* aborted */
			bitmap_end_sync(mddev->bitmap, mddev->curr_resync,
					&sync_blocks, 1);
5073
		else /* completed sync */
5074 5075 5076
			conf->fullsync = 0;
		bitmap_close_sync(mddev->bitmap);

L
Linus Torvalds 已提交
5077 5078
		return 0;
	}
5079

5080 5081 5082
	/* Allow raid5_quiesce to complete */
	wait_event(conf->wait_for_overlap, conf->quiesce != 2);

5083 5084
	if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
		return reshape_request(mddev, sector_nr, skipped);
5085

5086 5087 5088 5089 5090 5091
	/* 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
	 */

5092
	/* if there is too many failed drives and we are trying
L
Linus Torvalds 已提交
5093 5094 5095
	 * to resync, then assert that we are finished, because there is
	 * nothing we can do.
	 */
5096
	if (mddev->degraded >= conf->max_degraded &&
5097
	    test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
A
Andre Noll 已提交
5098
		sector_t rv = mddev->dev_sectors - sector_nr;
5099
		*skipped = 1;
L
Linus Torvalds 已提交
5100 5101
		return rv;
	}
5102 5103 5104 5105
	if (!test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
	    !conf->fullsync &&
	    !bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, 1) &&
	    sync_blocks >= STRIPE_SECTORS) {
5106 5107 5108 5109 5110
		/* 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
5111

N
NeilBrown 已提交
5112 5113
	bitmap_cond_end_sync(mddev->bitmap, sector_nr);

5114
	sh = get_active_stripe(conf, sector_nr, 0, 1, 0);
L
Linus Torvalds 已提交
5115
	if (sh == NULL) {
5116
		sh = get_active_stripe(conf, sector_nr, 0, 0, 0);
L
Linus Torvalds 已提交
5117
		/* make sure we don't swamp the stripe cache if someone else
5118
		 * is trying to get access
L
Linus Torvalds 已提交
5119
		 */
5120
		schedule_timeout_uninterruptible(1);
L
Linus Torvalds 已提交
5121
	}
5122
	/* Need to check if array will still be degraded after recovery/resync
5123 5124
	 * Note in case of > 1 drive failures it's possible we're rebuilding
	 * one drive while leaving another faulty drive in array.
5125
	 */
5126 5127 5128 5129 5130
	rcu_read_lock();
	for (i = 0; i < conf->raid_disks; i++) {
		struct md_rdev *rdev = ACCESS_ONCE(conf->disks[i].rdev);

		if (rdev == NULL || test_bit(Faulty, &rdev->flags))
5131
			still_degraded = 1;
5132 5133
	}
	rcu_read_unlock();
5134 5135 5136

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

5137
	set_bit(STRIPE_SYNC_REQUESTED, &sh->state);
5138
	set_bit(STRIPE_HANDLE, &sh->state);
L
Linus Torvalds 已提交
5139 5140 5141 5142 5143 5144

	release_stripe(sh);

	return STRIPE_SECTORS;
}

5145
static int  retry_aligned_read(struct r5conf *conf, struct bio *raid_bio)
5146 5147 5148 5149 5150 5151 5152 5153 5154 5155 5156 5157
{
	/* 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;
5158
	int dd_idx;
5159 5160 5161 5162 5163
	sector_t sector, logical_sector, last_sector;
	int scnt = 0;
	int remaining;
	int handled = 0;

5164 5165
	logical_sector = raid_bio->bi_iter.bi_sector &
		~((sector_t)STRIPE_SECTORS-1);
5166
	sector = raid5_compute_sector(conf, logical_sector,
5167
				      0, &dd_idx, NULL);
K
Kent Overstreet 已提交
5168
	last_sector = bio_end_sector(raid_bio);
5169 5170

	for (; logical_sector < last_sector;
5171 5172 5173
	     logical_sector += STRIPE_SECTORS,
		     sector += STRIPE_SECTORS,
		     scnt++) {
5174

5175
		if (scnt < raid5_bi_processed_stripes(raid_bio))
5176 5177 5178
			/* already done this stripe */
			continue;

5179
		sh = get_active_stripe(conf, sector, 0, 1, 1);
5180 5181 5182

		if (!sh) {
			/* failed to get a stripe - must wait */
5183
			raid5_set_bi_processed_stripes(raid_bio, scnt);
5184 5185 5186 5187
			conf->retry_read_aligned = raid_bio;
			return handled;
		}

5188 5189
		if (!add_stripe_bio(sh, raid_bio, dd_idx, 0)) {
			release_stripe(sh);
5190
			raid5_set_bi_processed_stripes(raid_bio, scnt);
5191 5192 5193 5194
			conf->retry_read_aligned = raid_bio;
			return handled;
		}

5195
		set_bit(R5_ReadNoMerge, &sh->dev[dd_idx].flags);
5196
		handle_stripe(sh);
5197 5198 5199
		release_stripe(sh);
		handled++;
	}
5200
	remaining = raid5_dec_bi_active_stripes(raid_bio);
5201 5202 5203
	if (remaining == 0) {
		trace_block_bio_complete(bdev_get_queue(raid_bio->bi_bdev),
					 raid_bio, 0);
5204
		bio_endio(raid_bio, 0);
5205
	}
5206 5207 5208 5209 5210
	if (atomic_dec_and_test(&conf->active_aligned_reads))
		wake_up(&conf->wait_for_stripe);
	return handled;
}

5211
static int handle_active_stripes(struct r5conf *conf, int group,
5212 5213
				 struct r5worker *worker,
				 struct list_head *temp_inactive_list)
5214 5215
{
	struct stripe_head *batch[MAX_STRIPE_BATCH], *sh;
5216 5217
	int i, batch_size = 0, hash;
	bool release_inactive = false;
5218 5219

	while (batch_size < MAX_STRIPE_BATCH &&
5220
			(sh = __get_priority_stripe(conf, group)) != NULL)
5221 5222
		batch[batch_size++] = sh;

5223 5224 5225 5226 5227 5228 5229 5230
	if (batch_size == 0) {
		for (i = 0; i < NR_STRIPE_HASH_LOCKS; i++)
			if (!list_empty(temp_inactive_list + i))
				break;
		if (i == NR_STRIPE_HASH_LOCKS)
			return batch_size;
		release_inactive = true;
	}
5231 5232
	spin_unlock_irq(&conf->device_lock);

5233 5234 5235 5236 5237 5238 5239 5240
	release_inactive_stripe_list(conf, temp_inactive_list,
				     NR_STRIPE_HASH_LOCKS);

	if (release_inactive) {
		spin_lock_irq(&conf->device_lock);
		return 0;
	}

5241 5242 5243 5244 5245 5246
	for (i = 0; i < batch_size; i++)
		handle_stripe(batch[i]);

	cond_resched();

	spin_lock_irq(&conf->device_lock);
5247 5248 5249 5250
	for (i = 0; i < batch_size; i++) {
		hash = batch[i]->hash_lock_index;
		__release_stripe(conf, batch[i], &temp_inactive_list[hash]);
	}
5251 5252
	return batch_size;
}
5253

5254 5255 5256 5257 5258 5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269 5270
static void raid5_do_work(struct work_struct *work)
{
	struct r5worker *worker = container_of(work, struct r5worker, work);
	struct r5worker_group *group = worker->group;
	struct r5conf *conf = group->conf;
	int group_id = group - conf->worker_groups;
	int handled;
	struct blk_plug plug;

	pr_debug("+++ raid5worker active\n");

	blk_start_plug(&plug);
	handled = 0;
	spin_lock_irq(&conf->device_lock);
	while (1) {
		int batch_size, released;

5271
		released = release_stripe_list(conf, worker->temp_inactive_list);
5272

5273 5274
		batch_size = handle_active_stripes(conf, group_id, worker,
						   worker->temp_inactive_list);
5275
		worker->working = false;
5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286 5287
		if (!batch_size && !released)
			break;
		handled += batch_size;
	}
	pr_debug("%d stripes handled\n", handled);

	spin_unlock_irq(&conf->device_lock);
	blk_finish_plug(&plug);

	pr_debug("--- raid5worker inactive\n");
}

L
Linus Torvalds 已提交
5288 5289 5290 5291 5292 5293 5294
/*
 * This is our raid5 kernel thread.
 *
 * We scan the hash table for stripes which can be handled now.
 * During the scan, completed stripes are saved for us by the interrupt
 * handler, so that they will not have to wait for our next wakeup.
 */
S
Shaohua Li 已提交
5295
static void raid5d(struct md_thread *thread)
L
Linus Torvalds 已提交
5296
{
S
Shaohua Li 已提交
5297
	struct mddev *mddev = thread->mddev;
5298
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5299
	int handled;
5300
	struct blk_plug plug;
L
Linus Torvalds 已提交
5301

5302
	pr_debug("+++ raid5d active\n");
L
Linus Torvalds 已提交
5303 5304 5305

	md_check_recovery(mddev);

5306
	blk_start_plug(&plug);
L
Linus Torvalds 已提交
5307 5308 5309
	handled = 0;
	spin_lock_irq(&conf->device_lock);
	while (1) {
5310
		struct bio *bio;
S
Shaohua Li 已提交
5311 5312
		int batch_size, released;

5313
		released = release_stripe_list(conf, conf->temp_inactive_list);
L
Linus Torvalds 已提交
5314

5315
		if (
5316 5317 5318
		    !list_empty(&conf->bitmap_list)) {
			/* Now is a good time to flush some bitmap updates */
			conf->seq_flush++;
5319
			spin_unlock_irq(&conf->device_lock);
5320
			bitmap_unplug(mddev->bitmap);
5321
			spin_lock_irq(&conf->device_lock);
5322
			conf->seq_write = conf->seq_flush;
5323
			activate_bit_delay(conf, conf->temp_inactive_list);
5324
		}
5325
		raid5_activate_delayed(conf);
5326

5327 5328 5329 5330 5331 5332 5333 5334 5335 5336
		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++;
		}

5337 5338
		batch_size = handle_active_stripes(conf, ANY_GROUP, NULL,
						   conf->temp_inactive_list);
S
Shaohua Li 已提交
5339
		if (!batch_size && !released)
L
Linus Torvalds 已提交
5340
			break;
5341
		handled += batch_size;
L
Linus Torvalds 已提交
5342

5343 5344
		if (mddev->flags & ~(1<<MD_CHANGE_PENDING)) {
			spin_unlock_irq(&conf->device_lock);
5345
			md_check_recovery(mddev);
5346 5347
			spin_lock_irq(&conf->device_lock);
		}
L
Linus Torvalds 已提交
5348
	}
5349
	pr_debug("%d stripes handled\n", handled);
L
Linus Torvalds 已提交
5350 5351 5352

	spin_unlock_irq(&conf->device_lock);

5353
	async_tx_issue_pending_all();
5354
	blk_finish_plug(&plug);
L
Linus Torvalds 已提交
5355

5356
	pr_debug("--- raid5d inactive\n");
L
Linus Torvalds 已提交
5357 5358
}

5359
static ssize_t
5360
raid5_show_stripe_cache_size(struct mddev *mddev, char *page)
5361
{
5362 5363 5364 5365
	struct r5conf *conf;
	int ret = 0;
	spin_lock(&mddev->lock);
	conf = mddev->private;
5366
	if (conf)
5367 5368 5369
		ret = sprintf(page, "%d\n", conf->max_nr_stripes);
	spin_unlock(&mddev->lock);
	return ret;
5370 5371
}

5372
int
5373
raid5_set_cache_size(struct mddev *mddev, int size)
5374
{
5375
	struct r5conf *conf = mddev->private;
5376
	int err;
5377
	int hash;
5378

5379
	if (size <= 16 || size > 32768)
5380
		return -EINVAL;
5381
	hash = (conf->max_nr_stripes - 1) % NR_STRIPE_HASH_LOCKS;
5382
	while (size < conf->max_nr_stripes) {
5383
		if (drop_one_stripe(conf, hash))
5384 5385 5386
			conf->max_nr_stripes--;
		else
			break;
5387 5388 5389
		hash--;
		if (hash < 0)
			hash = NR_STRIPE_HASH_LOCKS - 1;
5390
	}
5391 5392 5393
	err = md_allow_write(mddev);
	if (err)
		return err;
5394
	hash = conf->max_nr_stripes % NR_STRIPE_HASH_LOCKS;
5395
	while (size > conf->max_nr_stripes) {
5396
		if (grow_one_stripe(conf, hash))
5397 5398
			conf->max_nr_stripes++;
		else break;
5399
		hash = (hash + 1) % NR_STRIPE_HASH_LOCKS;
5400
	}
5401 5402 5403 5404 5405
	return 0;
}
EXPORT_SYMBOL(raid5_set_cache_size);

static ssize_t
5406
raid5_store_stripe_cache_size(struct mddev *mddev, const char *page, size_t len)
5407
{
5408
	struct r5conf *conf;
5409 5410 5411 5412 5413
	unsigned long new;
	int err;

	if (len >= PAGE_SIZE)
		return -EINVAL;
5414
	if (kstrtoul(page, 10, &new))
5415
		return -EINVAL;
5416
	err = mddev_lock(mddev);
5417 5418
	if (err)
		return err;
5419 5420 5421 5422 5423 5424 5425 5426
	conf = mddev->private;
	if (!conf)
		err = -ENODEV;
	else
		err = raid5_set_cache_size(mddev, new);
	mddev_unlock(mddev);

	return err ?: len;
5427
}
5428

5429 5430 5431 5432
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);
5433

5434
static ssize_t
5435
raid5_show_preread_threshold(struct mddev *mddev, char *page)
5436
{
5437 5438 5439 5440
	struct r5conf *conf;
	int ret = 0;
	spin_lock(&mddev->lock);
	conf = mddev->private;
5441
	if (conf)
5442 5443 5444
		ret = sprintf(page, "%d\n", conf->bypass_threshold);
	spin_unlock(&mddev->lock);
	return ret;
5445 5446 5447
}

static ssize_t
5448
raid5_store_preread_threshold(struct mddev *mddev, const char *page, size_t len)
5449
{
5450
	struct r5conf *conf;
5451
	unsigned long new;
5452 5453
	int err;

5454 5455
	if (len >= PAGE_SIZE)
		return -EINVAL;
5456
	if (kstrtoul(page, 10, &new))
5457
		return -EINVAL;
5458 5459 5460 5461 5462 5463 5464 5465 5466 5467 5468 5469 5470

	err = mddev_lock(mddev);
	if (err)
		return err;
	conf = mddev->private;
	if (!conf)
		err = -ENODEV;
	else if (new > conf->max_nr_stripes)
		err = -EINVAL;
	else
		conf->bypass_threshold = new;
	mddev_unlock(mddev);
	return err ?: len;
5471 5472 5473 5474 5475 5476 5477 5478
}

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);

5479 5480 5481
static ssize_t
raid5_show_skip_copy(struct mddev *mddev, char *page)
{
5482 5483 5484 5485
	struct r5conf *conf;
	int ret = 0;
	spin_lock(&mddev->lock);
	conf = mddev->private;
5486
	if (conf)
5487 5488 5489
		ret = sprintf(page, "%d\n", conf->skip_copy);
	spin_unlock(&mddev->lock);
	return ret;
5490 5491 5492 5493 5494
}

static ssize_t
raid5_store_skip_copy(struct mddev *mddev, const char *page, size_t len)
{
5495
	struct r5conf *conf;
5496
	unsigned long new;
5497 5498
	int err;

5499 5500 5501 5502 5503
	if (len >= PAGE_SIZE)
		return -EINVAL;
	if (kstrtoul(page, 10, &new))
		return -EINVAL;
	new = !!new;
5504 5505 5506 5507 5508 5509 5510 5511 5512 5513 5514 5515 5516 5517 5518 5519 5520 5521 5522 5523

	err = mddev_lock(mddev);
	if (err)
		return err;
	conf = mddev->private;
	if (!conf)
		err = -ENODEV;
	else if (new != conf->skip_copy) {
		mddev_suspend(mddev);
		conf->skip_copy = new;
		if (new)
			mddev->queue->backing_dev_info.capabilities |=
				BDI_CAP_STABLE_WRITES;
		else
			mddev->queue->backing_dev_info.capabilities &=
				~BDI_CAP_STABLE_WRITES;
		mddev_resume(mddev);
	}
	mddev_unlock(mddev);
	return err ?: len;
5524 5525 5526 5527 5528 5529 5530
}

static struct md_sysfs_entry
raid5_skip_copy = __ATTR(skip_copy, S_IRUGO | S_IWUSR,
					raid5_show_skip_copy,
					raid5_store_skip_copy);

5531
static ssize_t
5532
stripe_cache_active_show(struct mddev *mddev, char *page)
5533
{
5534
	struct r5conf *conf = mddev->private;
5535 5536 5537 5538
	if (conf)
		return sprintf(page, "%d\n", atomic_read(&conf->active_stripes));
	else
		return 0;
5539 5540
}

5541 5542
static struct md_sysfs_entry
raid5_stripecache_active = __ATTR_RO(stripe_cache_active);
5543

5544 5545 5546
static ssize_t
raid5_show_group_thread_cnt(struct mddev *mddev, char *page)
{
5547 5548 5549 5550
	struct r5conf *conf;
	int ret = 0;
	spin_lock(&mddev->lock);
	conf = mddev->private;
5551
	if (conf)
5552 5553 5554
		ret = sprintf(page, "%d\n", conf->worker_cnt_per_group);
	spin_unlock(&mddev->lock);
	return ret;
5555 5556
}

5557 5558 5559 5560
static int alloc_thread_groups(struct r5conf *conf, int cnt,
			       int *group_cnt,
			       int *worker_cnt_per_group,
			       struct r5worker_group **worker_groups);
5561 5562 5563
static ssize_t
raid5_store_group_thread_cnt(struct mddev *mddev, const char *page, size_t len)
{
5564
	struct r5conf *conf;
5565 5566
	unsigned long new;
	int err;
5567 5568
	struct r5worker_group *new_groups, *old_groups;
	int group_cnt, worker_cnt_per_group;
5569 5570 5571 5572 5573 5574

	if (len >= PAGE_SIZE)
		return -EINVAL;
	if (kstrtoul(page, 10, &new))
		return -EINVAL;

5575 5576 5577 5578 5579 5580 5581 5582
	err = mddev_lock(mddev);
	if (err)
		return err;
	conf = mddev->private;
	if (!conf)
		err = -ENODEV;
	else if (new != conf->worker_cnt_per_group) {
		mddev_suspend(mddev);
5583

5584 5585 5586
		old_groups = conf->worker_groups;
		if (old_groups)
			flush_workqueue(raid5_wq);
5587

5588 5589 5590 5591 5592 5593 5594 5595 5596
		err = alloc_thread_groups(conf, new,
					  &group_cnt, &worker_cnt_per_group,
					  &new_groups);
		if (!err) {
			spin_lock_irq(&conf->device_lock);
			conf->group_cnt = group_cnt;
			conf->worker_cnt_per_group = worker_cnt_per_group;
			conf->worker_groups = new_groups;
			spin_unlock_irq(&conf->device_lock);
5597

5598 5599 5600 5601 5602
			if (old_groups)
				kfree(old_groups[0].workers);
			kfree(old_groups);
		}
		mddev_resume(mddev);
5603
	}
5604
	mddev_unlock(mddev);
5605

5606
	return err ?: len;
5607 5608 5609 5610 5611 5612 5613
}

static struct md_sysfs_entry
raid5_group_thread_cnt = __ATTR(group_thread_cnt, S_IRUGO | S_IWUSR,
				raid5_show_group_thread_cnt,
				raid5_store_group_thread_cnt);

5614
static struct attribute *raid5_attrs[] =  {
5615 5616
	&raid5_stripecache_size.attr,
	&raid5_stripecache_active.attr,
5617
	&raid5_preread_bypass_threshold.attr,
5618
	&raid5_group_thread_cnt.attr,
5619
	&raid5_skip_copy.attr,
5620 5621
	NULL,
};
5622 5623 5624
static struct attribute_group raid5_attrs_group = {
	.name = NULL,
	.attrs = raid5_attrs,
5625 5626
};

5627 5628 5629 5630
static int alloc_thread_groups(struct r5conf *conf, int cnt,
			       int *group_cnt,
			       int *worker_cnt_per_group,
			       struct r5worker_group **worker_groups)
5631
{
5632
	int i, j, k;
5633 5634 5635
	ssize_t size;
	struct r5worker *workers;

5636
	*worker_cnt_per_group = cnt;
5637
	if (cnt == 0) {
5638 5639
		*group_cnt = 0;
		*worker_groups = NULL;
5640 5641
		return 0;
	}
5642
	*group_cnt = num_possible_nodes();
5643
	size = sizeof(struct r5worker) * cnt;
5644 5645 5646 5647
	workers = kzalloc(size * *group_cnt, GFP_NOIO);
	*worker_groups = kzalloc(sizeof(struct r5worker_group) *
				*group_cnt, GFP_NOIO);
	if (!*worker_groups || !workers) {
5648
		kfree(workers);
5649
		kfree(*worker_groups);
5650 5651 5652
		return -ENOMEM;
	}

5653
	for (i = 0; i < *group_cnt; i++) {
5654 5655
		struct r5worker_group *group;

5656
		group = &(*worker_groups)[i];
5657 5658 5659 5660 5661
		INIT_LIST_HEAD(&group->handle_list);
		group->conf = conf;
		group->workers = workers + i * cnt;

		for (j = 0; j < cnt; j++) {
5662 5663 5664 5665 5666 5667
			struct r5worker *worker = group->workers + j;
			worker->group = group;
			INIT_WORK(&worker->work, raid5_do_work);

			for (k = 0; k < NR_STRIPE_HASH_LOCKS; k++)
				INIT_LIST_HEAD(worker->temp_inactive_list + k);
5668 5669 5670 5671 5672 5673 5674 5675 5676 5677 5678 5679 5680 5681
		}
	}

	return 0;
}

static void free_thread_groups(struct r5conf *conf)
{
	if (conf->worker_groups)
		kfree(conf->worker_groups[0].workers);
	kfree(conf->worker_groups);
	conf->worker_groups = NULL;
}

5682
static sector_t
5683
raid5_size(struct mddev *mddev, sector_t sectors, int raid_disks)
5684
{
5685
	struct r5conf *conf = mddev->private;
5686 5687 5688

	if (!sectors)
		sectors = mddev->dev_sectors;
5689
	if (!raid_disks)
5690
		/* size is defined by the smallest of previous and new size */
5691
		raid_disks = min(conf->raid_disks, conf->previous_raid_disks);
5692

5693
	sectors &= ~((sector_t)mddev->chunk_sectors - 1);
5694
	sectors &= ~((sector_t)mddev->new_chunk_sectors - 1);
5695 5696 5697
	return sectors * (raid_disks - conf->max_degraded);
}

5698 5699 5700 5701 5702 5703 5704 5705 5706 5707 5708 5709 5710 5711 5712 5713 5714 5715 5716 5717 5718 5719 5720
static void free_scratch_buffer(struct r5conf *conf, struct raid5_percpu *percpu)
{
	safe_put_page(percpu->spare_page);
	kfree(percpu->scribble);
	percpu->spare_page = NULL;
	percpu->scribble = NULL;
}

static int alloc_scratch_buffer(struct r5conf *conf, struct raid5_percpu *percpu)
{
	if (conf->level == 6 && !percpu->spare_page)
		percpu->spare_page = alloc_page(GFP_KERNEL);
	if (!percpu->scribble)
		percpu->scribble = kmalloc(conf->scribble_len, GFP_KERNEL);

	if (!percpu->scribble || (conf->level == 6 && !percpu->spare_page)) {
		free_scratch_buffer(conf, percpu);
		return -ENOMEM;
	}

	return 0;
}

5721
static void raid5_free_percpu(struct r5conf *conf)
5722 5723 5724 5725 5726 5727 5728 5729 5730
{
	unsigned long cpu;

	if (!conf->percpu)
		return;

#ifdef CONFIG_HOTPLUG_CPU
	unregister_cpu_notifier(&conf->cpu_notify);
#endif
5731 5732 5733 5734

	get_online_cpus();
	for_each_possible_cpu(cpu)
		free_scratch_buffer(conf, per_cpu_ptr(conf->percpu, cpu));
5735 5736 5737 5738 5739
	put_online_cpus();

	free_percpu(conf->percpu);
}

5740
static void free_conf(struct r5conf *conf)
5741
{
5742
	free_thread_groups(conf);
5743
	shrink_stripes(conf);
5744
	raid5_free_percpu(conf);
5745 5746 5747 5748 5749
	kfree(conf->disks);
	kfree(conf->stripe_hashtbl);
	kfree(conf);
}

5750 5751 5752 5753
#ifdef CONFIG_HOTPLUG_CPU
static int raid456_cpu_notify(struct notifier_block *nfb, unsigned long action,
			      void *hcpu)
{
5754
	struct r5conf *conf = container_of(nfb, struct r5conf, cpu_notify);
5755 5756 5757 5758 5759 5760
	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:
5761
		if (alloc_scratch_buffer(conf, percpu)) {
5762 5763
			pr_err("%s: failed memory allocation for cpu%ld\n",
			       __func__, cpu);
5764
			return notifier_from_errno(-ENOMEM);
5765 5766 5767 5768
		}
		break;
	case CPU_DEAD:
	case CPU_DEAD_FROZEN:
5769
		free_scratch_buffer(conf, per_cpu_ptr(conf->percpu, cpu));
5770 5771 5772 5773 5774 5775 5776 5777
		break;
	default:
		break;
	}
	return NOTIFY_OK;
}
#endif

5778
static int raid5_alloc_percpu(struct r5conf *conf)
5779 5780
{
	unsigned long cpu;
5781
	int err = 0;
5782

5783 5784
	conf->percpu = alloc_percpu(struct raid5_percpu);
	if (!conf->percpu)
5785
		return -ENOMEM;
5786 5787 5788 5789 5790 5791 5792 5793

#ifdef CONFIG_HOTPLUG_CPU
	conf->cpu_notify.notifier_call = raid456_cpu_notify;
	conf->cpu_notify.priority = 0;
	err = register_cpu_notifier(&conf->cpu_notify);
	if (err)
		return err;
#endif
5794 5795 5796

	get_online_cpus();
	for_each_present_cpu(cpu) {
5797 5798 5799 5800
		err = alloc_scratch_buffer(conf, per_cpu_ptr(conf->percpu, cpu));
		if (err) {
			pr_err("%s: failed memory allocation for cpu%ld\n",
			       __func__, cpu);
5801 5802 5803 5804 5805 5806 5807 5808
			break;
		}
	}
	put_online_cpus();

	return err;
}

5809
static struct r5conf *setup_conf(struct mddev *mddev)
L
Linus Torvalds 已提交
5810
{
5811
	struct r5conf *conf;
5812
	int raid_disk, memory, max_disks;
5813
	struct md_rdev *rdev;
L
Linus Torvalds 已提交
5814
	struct disk_info *disk;
5815
	char pers_name[6];
5816
	int i;
5817 5818
	int group_cnt, worker_cnt_per_group;
	struct r5worker_group *new_group;
L
Linus Torvalds 已提交
5819

N
NeilBrown 已提交
5820 5821 5822
	if (mddev->new_level != 5
	    && mddev->new_level != 4
	    && mddev->new_level != 6) {
5823
		printk(KERN_ERR "md/raid:%s: raid level not set to 4/5/6 (%d)\n",
N
NeilBrown 已提交
5824 5825
		       mdname(mddev), mddev->new_level);
		return ERR_PTR(-EIO);
L
Linus Torvalds 已提交
5826
	}
N
NeilBrown 已提交
5827 5828 5829 5830
	if ((mddev->new_level == 5
	     && !algorithm_valid_raid5(mddev->new_layout)) ||
	    (mddev->new_level == 6
	     && !algorithm_valid_raid6(mddev->new_layout))) {
5831
		printk(KERN_ERR "md/raid:%s: layout %d not supported\n",
N
NeilBrown 已提交
5832 5833
		       mdname(mddev), mddev->new_layout);
		return ERR_PTR(-EIO);
5834
	}
N
NeilBrown 已提交
5835
	if (mddev->new_level == 6 && mddev->raid_disks < 4) {
5836
		printk(KERN_ERR "md/raid:%s: not enough configured devices (%d, minimum 4)\n",
N
NeilBrown 已提交
5837 5838
		       mdname(mddev), mddev->raid_disks);
		return ERR_PTR(-EINVAL);
5839 5840
	}

5841 5842 5843
	if (!mddev->new_chunk_sectors ||
	    (mddev->new_chunk_sectors << 9) % PAGE_SIZE ||
	    !is_power_of_2(mddev->new_chunk_sectors)) {
5844 5845
		printk(KERN_ERR "md/raid:%s: invalid chunk size %d\n",
		       mdname(mddev), mddev->new_chunk_sectors << 9);
N
NeilBrown 已提交
5846
		return ERR_PTR(-EINVAL);
5847 5848
	}

5849
	conf = kzalloc(sizeof(struct r5conf), GFP_KERNEL);
N
NeilBrown 已提交
5850
	if (conf == NULL)
L
Linus Torvalds 已提交
5851
		goto abort;
5852
	/* Don't enable multi-threading by default*/
5853 5854 5855 5856 5857 5858
	if (!alloc_thread_groups(conf, 0, &group_cnt, &worker_cnt_per_group,
				 &new_group)) {
		conf->group_cnt = group_cnt;
		conf->worker_cnt_per_group = worker_cnt_per_group;
		conf->worker_groups = new_group;
	} else
5859
		goto abort;
5860
	spin_lock_init(&conf->device_lock);
5861
	seqcount_init(&conf->gen_lock);
5862 5863 5864 5865 5866 5867
	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);
S
Shaohua Li 已提交
5868
	init_llist_head(&conf->released_stripes);
5869 5870 5871 5872
	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;
5873
	conf->recovery_disabled = mddev->recovery_disabled - 1;
N
NeilBrown 已提交
5874 5875 5876 5877 5878

	conf->raid_disks = mddev->raid_disks;
	if (mddev->reshape_position == MaxSector)
		conf->previous_raid_disks = mddev->raid_disks;
	else
5879
		conf->previous_raid_disks = mddev->raid_disks - mddev->delta_disks;
5880 5881
	max_disks = max(conf->raid_disks, conf->previous_raid_disks);
	conf->scribble_len = scribble_len(max_disks);
5882

5883
	conf->disks = kzalloc(max_disks * sizeof(struct disk_info),
5884 5885 5886
			      GFP_KERNEL);
	if (!conf->disks)
		goto abort;
5887

L
Linus Torvalds 已提交
5888 5889
	conf->mddev = mddev;

5890
	if ((conf->stripe_hashtbl = kzalloc(PAGE_SIZE, GFP_KERNEL)) == NULL)
L
Linus Torvalds 已提交
5891 5892
		goto abort;

5893 5894 5895 5896 5897 5898 5899 5900 5901 5902 5903 5904 5905 5906 5907
	/* We init hash_locks[0] separately to that it can be used
	 * as the reference lock in the spin_lock_nest_lock() call
	 * in lock_all_device_hash_locks_irq in order to convince
	 * lockdep that we know what we are doing.
	 */
	spin_lock_init(conf->hash_locks);
	for (i = 1; i < NR_STRIPE_HASH_LOCKS; i++)
		spin_lock_init(conf->hash_locks + i);

	for (i = 0; i < NR_STRIPE_HASH_LOCKS; i++)
		INIT_LIST_HEAD(conf->inactive_list + i);

	for (i = 0; i < NR_STRIPE_HASH_LOCKS; i++)
		INIT_LIST_HEAD(conf->temp_inactive_list + i);

5908 5909 5910 5911
	conf->level = mddev->new_level;
	if (raid5_alloc_percpu(conf) != 0)
		goto abort;

5912
	pr_debug("raid456: run(%s) called.\n", mdname(mddev));
L
Linus Torvalds 已提交
5913

N
NeilBrown 已提交
5914
	rdev_for_each(rdev, mddev) {
L
Linus Torvalds 已提交
5915
		raid_disk = rdev->raid_disk;
5916
		if (raid_disk >= max_disks
L
Linus Torvalds 已提交
5917 5918 5919 5920
		    || raid_disk < 0)
			continue;
		disk = conf->disks + raid_disk;

5921 5922 5923 5924 5925 5926 5927 5928 5929
		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
5930

5931
		if (test_bit(In_sync, &rdev->flags)) {
L
Linus Torvalds 已提交
5932
			char b[BDEVNAME_SIZE];
5933 5934 5935
			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
5936
		} else if (rdev->saved_raid_disk != raid_disk)
5937 5938
			/* Cannot rely on bitmap to complete recovery */
			conf->fullsync = 1;
L
Linus Torvalds 已提交
5939 5940
	}

5941
	conf->chunk_sectors = mddev->new_chunk_sectors;
N
NeilBrown 已提交
5942
	conf->level = mddev->new_level;
5943 5944 5945 5946
	if (conf->level == 6)
		conf->max_degraded = 2;
	else
		conf->max_degraded = 1;
N
NeilBrown 已提交
5947
	conf->algorithm = mddev->new_layout;
5948
	conf->reshape_progress = mddev->reshape_position;
5949
	if (conf->reshape_progress != MaxSector) {
5950
		conf->prev_chunk_sectors = mddev->chunk_sectors;
5951 5952
		conf->prev_algo = mddev->layout;
	}
L
Linus Torvalds 已提交
5953

N
NeilBrown 已提交
5954
	memory = conf->max_nr_stripes * (sizeof(struct stripe_head) +
5955
		 max_disks * ((sizeof(struct bio) + PAGE_SIZE))) / 1024;
5956
	atomic_set(&conf->empty_inactive_list_nr, NR_STRIPE_HASH_LOCKS);
5957
	if (grow_stripes(conf, NR_STRIPES)) {
N
NeilBrown 已提交
5958
		printk(KERN_ERR
5959 5960
		       "md/raid:%s: couldn't allocate %dkB for buffers\n",
		       mdname(mddev), memory);
N
NeilBrown 已提交
5961 5962
		goto abort;
	} else
5963 5964
		printk(KERN_INFO "md/raid:%s: allocated %dkB\n",
		       mdname(mddev), memory);
L
Linus Torvalds 已提交
5965

5966 5967
	sprintf(pers_name, "raid%d", mddev->new_level);
	conf->thread = md_register_thread(raid5d, mddev, pers_name);
N
NeilBrown 已提交
5968 5969
	if (!conf->thread) {
		printk(KERN_ERR
5970
		       "md/raid:%s: couldn't allocate thread.\n",
N
NeilBrown 已提交
5971
		       mdname(mddev));
5972 5973
		goto abort;
	}
N
NeilBrown 已提交
5974 5975 5976 5977 5978

	return conf;

 abort:
	if (conf) {
5979
		free_conf(conf);
N
NeilBrown 已提交
5980 5981 5982 5983 5984
		return ERR_PTR(-EIO);
	} else
		return ERR_PTR(-ENOMEM);
}

5985 5986 5987 5988 5989 5990 5991 5992 5993 5994 5995 5996
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:
5997
		if (raid_disk == 0 ||
5998 5999 6000 6001 6002 6003 6004 6005 6006 6007 6008 6009 6010
		    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;
}

6011
static int run(struct mddev *mddev)
N
NeilBrown 已提交
6012
{
6013
	struct r5conf *conf;
6014
	int working_disks = 0;
6015
	int dirty_parity_disks = 0;
6016
	struct md_rdev *rdev;
6017
	sector_t reshape_offset = 0;
6018
	int i;
6019 6020
	long long min_offset_diff = 0;
	int first = 1;
N
NeilBrown 已提交
6021

6022
	if (mddev->recovery_cp != MaxSector)
6023
		printk(KERN_NOTICE "md/raid:%s: not clean"
6024 6025
		       " -- starting background reconstruction\n",
		       mdname(mddev));
6026 6027 6028 6029 6030 6031 6032 6033 6034 6035 6036 6037 6038 6039 6040 6041 6042

	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
6043 6044
	if (mddev->reshape_position != MaxSector) {
		/* Check that we can continue the reshape.
6045 6046 6047 6048 6049 6050 6051 6052 6053 6054
		 * 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
6055 6056 6057
		 */
		sector_t here_new, here_old;
		int old_disks;
6058
		int max_degraded = (mddev->level == 6 ? 2 : 1);
N
NeilBrown 已提交
6059

6060
		if (mddev->new_level != mddev->level) {
6061
			printk(KERN_ERR "md/raid:%s: unsupported reshape "
N
NeilBrown 已提交
6062 6063 6064 6065 6066 6067 6068 6069 6070 6071
			       "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;
6072
		if (sector_div(here_new, mddev->new_chunk_sectors *
N
NeilBrown 已提交
6073
			       (mddev->raid_disks - max_degraded))) {
6074 6075
			printk(KERN_ERR "md/raid:%s: reshape_position not "
			       "on a stripe boundary\n", mdname(mddev));
N
NeilBrown 已提交
6076 6077
			return -EINVAL;
		}
6078
		reshape_offset = here_new * mddev->new_chunk_sectors;
N
NeilBrown 已提交
6079 6080
		/* here_new is the stripe we will write to */
		here_old = mddev->reshape_position;
6081
		sector_div(here_old, mddev->chunk_sectors *
N
NeilBrown 已提交
6082 6083 6084
			   (old_disks-max_degraded));
		/* here_old is the first stripe that we might need to read
		 * from */
6085
		if (mddev->delta_disks == 0) {
6086 6087 6088 6089 6090 6091
			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;
			}
6092
			/* We cannot be sure it is safe to start an in-place
6093
			 * reshape.  It is only safe if user-space is monitoring
6094 6095 6096 6097 6098
			 * 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.
			 */
6099 6100 6101 6102 6103 6104 6105
			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",
6106
				       mdname(mddev));
6107 6108
				return -EINVAL;
			}
6109
		} else if (mddev->reshape_backwards
6110
		    ? (here_new * mddev->new_chunk_sectors + min_offset_diff <=
6111 6112
		       here_old * mddev->chunk_sectors)
		    : (here_new * mddev->new_chunk_sectors >=
6113
		       here_old * mddev->chunk_sectors + (-min_offset_diff))) {
N
NeilBrown 已提交
6114
			/* Reading from the same stripe as writing to - bad */
6115 6116 6117
			printk(KERN_ERR "md/raid:%s: reshape_position too early for "
			       "auto-recovery - aborting.\n",
			       mdname(mddev));
N
NeilBrown 已提交
6118 6119
			return -EINVAL;
		}
6120 6121
		printk(KERN_INFO "md/raid:%s: reshape will continue\n",
		       mdname(mddev));
N
NeilBrown 已提交
6122 6123 6124 6125
		/* OK, we should be able to continue; */
	} else {
		BUG_ON(mddev->level != mddev->new_level);
		BUG_ON(mddev->layout != mddev->new_layout);
6126
		BUG_ON(mddev->chunk_sectors != mddev->new_chunk_sectors);
N
NeilBrown 已提交
6127
		BUG_ON(mddev->delta_disks != 0);
L
Linus Torvalds 已提交
6128
	}
N
NeilBrown 已提交
6129

6130 6131 6132 6133 6134
	if (mddev->private == NULL)
		conf = setup_conf(mddev);
	else
		conf = mddev->private;

N
NeilBrown 已提交
6135 6136 6137
	if (IS_ERR(conf))
		return PTR_ERR(conf);

6138
	conf->min_offset_diff = min_offset_diff;
N
NeilBrown 已提交
6139 6140 6141 6142
	mddev->thread = conf->thread;
	conf->thread = NULL;
	mddev->private = conf;

6143 6144 6145 6146 6147 6148 6149 6150 6151 6152 6153
	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)
6154
			continue;
6155 6156 6157 6158 6159 6160 6161
		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;
		}
6162
		if (test_bit(In_sync, &rdev->flags)) {
N
NeilBrown 已提交
6163
			working_disks++;
6164 6165
			continue;
		}
6166 6167 6168 6169 6170 6171 6172 6173 6174 6175 6176 6177
		/* 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;
6178

6179 6180 6181 6182 6183 6184 6185 6186 6187 6188 6189 6190 6191 6192 6193
		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
6194

6195 6196 6197
	/*
	 * 0 for a fully functional array, 1 or 2 for a degraded array.
	 */
6198
	mddev->degraded = calc_degraded(conf);
N
NeilBrown 已提交
6199

6200
	if (has_failed(conf)) {
6201
		printk(KERN_ERR "md/raid:%s: not enough operational devices"
L
Linus Torvalds 已提交
6202
			" (%d/%d failed)\n",
6203
			mdname(mddev), mddev->degraded, conf->raid_disks);
L
Linus Torvalds 已提交
6204 6205 6206
		goto abort;
	}

N
NeilBrown 已提交
6207
	/* device size must be a multiple of chunk size */
6208
	mddev->dev_sectors &= ~(mddev->chunk_sectors - 1);
N
NeilBrown 已提交
6209 6210
	mddev->resync_max_sectors = mddev->dev_sectors;

6211
	if (mddev->degraded > dirty_parity_disks &&
L
Linus Torvalds 已提交
6212
	    mddev->recovery_cp != MaxSector) {
6213 6214
		if (mddev->ok_start_degraded)
			printk(KERN_WARNING
6215 6216
			       "md/raid:%s: starting dirty degraded array"
			       " - data corruption possible.\n",
6217 6218 6219
			       mdname(mddev));
		else {
			printk(KERN_ERR
6220
			       "md/raid:%s: cannot start dirty degraded array.\n",
6221 6222 6223
			       mdname(mddev));
			goto abort;
		}
L
Linus Torvalds 已提交
6224 6225 6226
	}

	if (mddev->degraded == 0)
6227 6228
		printk(KERN_INFO "md/raid:%s: raid level %d active with %d out of %d"
		       " devices, algorithm %d\n", mdname(mddev), conf->level,
6229 6230
		       mddev->raid_disks-mddev->degraded, mddev->raid_disks,
		       mddev->new_layout);
L
Linus Torvalds 已提交
6231
	else
6232 6233 6234 6235 6236
		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
6237 6238 6239

	print_raid5_conf(conf);

6240 6241
	if (conf->reshape_progress != MaxSector) {
		conf->reshape_safe = conf->reshape_progress;
6242 6243 6244 6245 6246 6247
		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,
6248
							"reshape");
6249 6250
	}

L
Linus Torvalds 已提交
6251
	/* Ok, everything is just fine now */
6252 6253
	if (mddev->to_remove == &raid5_attrs_group)
		mddev->to_remove = NULL;
N
NeilBrown 已提交
6254 6255
	else if (mddev->kobj.sd &&
	    sysfs_create_group(&mddev->kobj, &raid5_attrs_group))
6256
		printk(KERN_WARNING
6257
		       "raid5: failed to create sysfs attributes for %s\n",
6258
		       mdname(mddev));
6259
	md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
6260

6261
	if (mddev->queue) {
6262
		int chunk_size;
S
Shaohua Li 已提交
6263
		bool discard_supported = true;
6264 6265 6266 6267 6268 6269 6270 6271 6272
		/* 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
6273

6274 6275 6276 6277
		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));
6278
		mddev->queue->limits.raid_partial_stripes_expensive = 1;
S
Shaohua Li 已提交
6279 6280 6281 6282 6283
		/*
		 * We can only discard a whole stripe. It doesn't make sense to
		 * discard data disk but write parity disk
		 */
		stripe = stripe * PAGE_SIZE;
6284 6285 6286 6287
		/* Round up to power of 2, as discard handling
		 * currently assumes that */
		while ((stripe-1) & stripe)
			stripe = (stripe | (stripe-1)) + 1;
S
Shaohua Li 已提交
6288 6289 6290 6291
		mddev->queue->limits.discard_alignment = stripe;
		mddev->queue->limits.discard_granularity = stripe;
		/*
		 * unaligned part of discard request will be ignored, so can't
6292
		 * guarantee discard_zeroes_data
S
Shaohua Li 已提交
6293 6294
		 */
		mddev->queue->limits.discard_zeroes_data = 0;
6295

6296 6297
		blk_queue_max_write_same_sectors(mddev->queue, 0);

6298
		rdev_for_each(rdev, mddev) {
6299 6300
			disk_stack_limits(mddev->gendisk, rdev->bdev,
					  rdev->data_offset << 9);
6301 6302
			disk_stack_limits(mddev->gendisk, rdev->bdev,
					  rdev->new_data_offset << 9);
S
Shaohua Li 已提交
6303 6304 6305 6306 6307 6308 6309 6310 6311 6312 6313 6314 6315 6316
			/*
			 * discard_zeroes_data is required, otherwise data
			 * could be lost. Consider a scenario: discard a stripe
			 * (the stripe could be inconsistent if
			 * discard_zeroes_data is 0); write one disk of the
			 * stripe (the stripe could be inconsistent again
			 * depending on which disks are used to calculate
			 * parity); the disk is broken; The stripe data of this
			 * disk is lost.
			 */
			if (!blk_queue_discard(bdev_get_queue(rdev->bdev)) ||
			    !bdev_get_queue(rdev->bdev)->
						limits.discard_zeroes_data)
				discard_supported = false;
6317 6318 6319 6320 6321 6322 6323 6324 6325 6326 6327 6328
			/* Unfortunately, discard_zeroes_data is not currently
			 * a guarantee - just a hint.  So we only allow DISCARD
			 * if the sysadmin has confirmed that only safe devices
			 * are in use by setting a module parameter.
			 */
			if (!devices_handle_discard_safely) {
				if (discard_supported) {
					pr_info("md/raid456: discard support disabled due to uncertainty.\n");
					pr_info("Set raid456.devices_handle_discard_safely=Y to override.\n");
				}
				discard_supported = false;
			}
6329
		}
S
Shaohua Li 已提交
6330 6331 6332 6333 6334 6335 6336 6337 6338

		if (discard_supported &&
		   mddev->queue->limits.max_discard_sectors >= stripe &&
		   mddev->queue->limits.discard_granularity >= stripe)
			queue_flag_set_unlocked(QUEUE_FLAG_DISCARD,
						mddev->queue);
		else
			queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD,
						mddev->queue);
6339
	}
6340

L
Linus Torvalds 已提交
6341 6342
	return 0;
abort:
6343
	md_unregister_thread(&mddev->thread);
N
NeilBrown 已提交
6344 6345
	print_raid5_conf(conf);
	free_conf(conf);
L
Linus Torvalds 已提交
6346
	mddev->private = NULL;
6347
	printk(KERN_ALERT "md/raid:%s: failed to run raid set.\n", mdname(mddev));
L
Linus Torvalds 已提交
6348 6349 6350
	return -EIO;
}

N
NeilBrown 已提交
6351
static void raid5_free(struct mddev *mddev, void *priv)
L
Linus Torvalds 已提交
6352
{
N
NeilBrown 已提交
6353
	struct r5conf *conf = priv;
L
Linus Torvalds 已提交
6354

6355
	free_conf(conf);
6356
	mddev->to_remove = &raid5_attrs_group;
L
Linus Torvalds 已提交
6357 6358
}

6359
static void status(struct seq_file *seq, struct mddev *mddev)
L
Linus Torvalds 已提交
6360
{
6361
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
6362 6363
	int i;

6364 6365
	seq_printf(seq, " level %d, %dk chunk, algorithm %d", mddev->level,
		mddev->chunk_sectors / 2, mddev->layout);
6366
	seq_printf (seq, " [%d/%d] [", conf->raid_disks, conf->raid_disks - mddev->degraded);
L
Linus Torvalds 已提交
6367 6368 6369
	for (i = 0; i < conf->raid_disks; i++)
		seq_printf (seq, "%s",
			       conf->disks[i].rdev &&
6370
			       test_bit(In_sync, &conf->disks[i].rdev->flags) ? "U" : "_");
L
Linus Torvalds 已提交
6371 6372 6373
	seq_printf (seq, "]");
}

6374
static void print_raid5_conf (struct r5conf *conf)
L
Linus Torvalds 已提交
6375 6376 6377 6378
{
	int i;
	struct disk_info *tmp;

6379
	printk(KERN_DEBUG "RAID conf printout:\n");
L
Linus Torvalds 已提交
6380 6381 6382 6383
	if (!conf) {
		printk("(conf==NULL)\n");
		return;
	}
6384 6385 6386
	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
6387 6388 6389 6390 6391

	for (i = 0; i < conf->raid_disks; i++) {
		char b[BDEVNAME_SIZE];
		tmp = conf->disks + i;
		if (tmp->rdev)
6392 6393 6394
			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
6395 6396 6397
	}
}

6398
static int raid5_spare_active(struct mddev *mddev)
L
Linus Torvalds 已提交
6399 6400
{
	int i;
6401
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
6402
	struct disk_info *tmp;
6403 6404
	int count = 0;
	unsigned long flags;
L
Linus Torvalds 已提交
6405 6406 6407

	for (i = 0; i < conf->raid_disks; i++) {
		tmp = conf->disks + i;
6408 6409 6410 6411 6412 6413 6414 6415 6416 6417 6418 6419 6420 6421 6422 6423 6424 6425 6426
		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
6427
		    && tmp->rdev->recovery_offset == MaxSector
6428
		    && !test_bit(Faulty, &tmp->rdev->flags)
6429
		    && !test_and_set_bit(In_sync, &tmp->rdev->flags)) {
6430
			count++;
6431
			sysfs_notify_dirent_safe(tmp->rdev->sysfs_state);
L
Linus Torvalds 已提交
6432 6433
		}
	}
6434
	spin_lock_irqsave(&conf->device_lock, flags);
6435
	mddev->degraded = calc_degraded(conf);
6436
	spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
6437
	print_raid5_conf(conf);
6438
	return count;
L
Linus Torvalds 已提交
6439 6440
}

6441
static int raid5_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
6442
{
6443
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
6444
	int err = 0;
6445
	int number = rdev->raid_disk;
6446
	struct md_rdev **rdevp;
L
Linus Torvalds 已提交
6447 6448 6449
	struct disk_info *p = conf->disks + number;

	print_raid5_conf(conf);
6450 6451 6452 6453 6454 6455 6456 6457 6458 6459 6460 6461 6462 6463 6464 6465 6466 6467 6468 6469 6470 6471
	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) &&
6472
	    (!p->replacement || p->replacement == rdev) &&
6473 6474 6475 6476 6477 6478 6479 6480 6481 6482
	    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;
6483 6484 6485 6486 6487 6488 6489 6490 6491 6492 6493 6494 6495 6496
	} 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
6497 6498 6499 6500 6501 6502
abort:

	print_raid5_conf(conf);
	return err;
}

6503
static int raid5_add_disk(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
6504
{
6505
	struct r5conf *conf = mddev->private;
6506
	int err = -EEXIST;
L
Linus Torvalds 已提交
6507 6508
	int disk;
	struct disk_info *p;
6509 6510
	int first = 0;
	int last = conf->raid_disks - 1;
L
Linus Torvalds 已提交
6511

6512 6513 6514
	if (mddev->recovery_disabled == conf->recovery_disabled)
		return -EBUSY;

N
NeilBrown 已提交
6515
	if (rdev->saved_raid_disk < 0 && has_failed(conf))
L
Linus Torvalds 已提交
6516
		/* no point adding a device */
6517
		return -EINVAL;
L
Linus Torvalds 已提交
6518

6519 6520
	if (rdev->raid_disk >= 0)
		first = last = rdev->raid_disk;
L
Linus Torvalds 已提交
6521 6522

	/*
6523 6524
	 * find the disk ... but prefer rdev->saved_raid_disk
	 * if possible.
L
Linus Torvalds 已提交
6525
	 */
6526
	if (rdev->saved_raid_disk >= 0 &&
6527
	    rdev->saved_raid_disk >= first &&
6528
	    conf->disks[rdev->saved_raid_disk].rdev == NULL)
6529 6530 6531
		first = rdev->saved_raid_disk;

	for (disk = first; disk <= last; disk++) {
6532 6533
		p = conf->disks + disk;
		if (p->rdev == NULL) {
6534
			clear_bit(In_sync, &rdev->flags);
L
Linus Torvalds 已提交
6535
			rdev->raid_disk = disk;
6536
			err = 0;
6537 6538
			if (rdev->saved_raid_disk != disk)
				conf->fullsync = 1;
6539
			rcu_assign_pointer(p->rdev, rdev);
6540
			goto out;
L
Linus Torvalds 已提交
6541
		}
6542 6543 6544
	}
	for (disk = first; disk <= last; disk++) {
		p = conf->disks + disk;
6545 6546 6547 6548 6549 6550 6551 6552 6553 6554 6555
		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;
		}
	}
6556
out:
L
Linus Torvalds 已提交
6557
	print_raid5_conf(conf);
6558
	return err;
L
Linus Torvalds 已提交
6559 6560
}

6561
static int raid5_resize(struct mddev *mddev, sector_t sectors)
L
Linus Torvalds 已提交
6562 6563 6564 6565 6566 6567 6568 6569
{
	/* 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.
	 */
6570
	sector_t newsize;
6571
	sectors &= ~((sector_t)mddev->chunk_sectors - 1);
6572 6573 6574
	newsize = raid5_size(mddev, sectors, mddev->raid_disks);
	if (mddev->external_size &&
	    mddev->array_sectors > newsize)
D
Dan Williams 已提交
6575
		return -EINVAL;
6576 6577 6578 6579 6580 6581
	if (mddev->bitmap) {
		int ret = bitmap_resize(mddev->bitmap, sectors, 0, 0);
		if (ret)
			return ret;
	}
	md_set_array_sectors(mddev, newsize);
6582
	set_capacity(mddev->gendisk, mddev->array_sectors);
6583
	revalidate_disk(mddev->gendisk);
6584 6585
	if (sectors > mddev->dev_sectors &&
	    mddev->recovery_cp > mddev->dev_sectors) {
A
Andre Noll 已提交
6586
		mddev->recovery_cp = mddev->dev_sectors;
L
Linus Torvalds 已提交
6587 6588
		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
	}
A
Andre Noll 已提交
6589
	mddev->dev_sectors = sectors;
6590
	mddev->resync_max_sectors = sectors;
L
Linus Torvalds 已提交
6591 6592 6593
	return 0;
}

6594
static int check_stripe_cache(struct mddev *mddev)
6595 6596 6597 6598 6599 6600 6601 6602 6603
{
	/* 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.
	 */
6604
	struct r5conf *conf = mddev->private;
6605 6606 6607 6608
	if (((mddev->chunk_sectors << 9) / STRIPE_SIZE) * 4
	    > conf->max_nr_stripes ||
	    ((mddev->new_chunk_sectors << 9) / STRIPE_SIZE) * 4
	    > conf->max_nr_stripes) {
6609 6610
		printk(KERN_WARNING "md/raid:%s: reshape: not enough stripes.  Needed %lu\n",
		       mdname(mddev),
6611 6612 6613 6614 6615 6616 6617
		       ((max(mddev->chunk_sectors, mddev->new_chunk_sectors) << 9)
			/ STRIPE_SIZE)*4);
		return 0;
	}
	return 1;
}

6618
static int check_reshape(struct mddev *mddev)
6619
{
6620
	struct r5conf *conf = mddev->private;
6621

6622 6623
	if (mddev->delta_disks == 0 &&
	    mddev->new_layout == mddev->layout &&
6624
	    mddev->new_chunk_sectors == mddev->chunk_sectors)
6625
		return 0; /* nothing to do */
6626
	if (has_failed(conf))
6627
		return -EINVAL;
6628
	if (mddev->delta_disks < 0 && mddev->reshape_position == MaxSector) {
6629 6630 6631 6632 6633 6634 6635 6636 6637 6638 6639
		/* 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;
	}
6640

6641
	if (!check_stripe_cache(mddev))
6642 6643
		return -ENOSPC;

6644 6645
	return resize_stripes(conf, (conf->previous_raid_disks
				     + mddev->delta_disks));
6646 6647
}

6648
static int raid5_start_reshape(struct mddev *mddev)
6649
{
6650
	struct r5conf *conf = mddev->private;
6651
	struct md_rdev *rdev;
6652
	int spares = 0;
6653
	unsigned long flags;
6654

6655
	if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
6656 6657
		return -EBUSY;

6658 6659 6660
	if (!check_stripe_cache(mddev))
		return -ENOSPC;

6661 6662 6663
	if (has_failed(conf))
		return -EINVAL;

6664
	rdev_for_each(rdev, mddev) {
6665 6666
		if (!test_bit(In_sync, &rdev->flags)
		    && !test_bit(Faulty, &rdev->flags))
6667
			spares++;
6668
	}
6669

6670
	if (spares - mddev->degraded < mddev->delta_disks - conf->max_degraded)
6671 6672 6673 6674 6675
		/* Not enough devices even to make a degraded array
		 * of that size
		 */
		return -EINVAL;

6676 6677 6678 6679 6680 6681
	/* 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) {
6682
		printk(KERN_ERR "md/raid:%s: array size must be reduced "
6683 6684 6685 6686
		       "before number of disks\n", mdname(mddev));
		return -EINVAL;
	}

6687
	atomic_set(&conf->reshape_stripes, 0);
6688
	spin_lock_irq(&conf->device_lock);
6689
	write_seqcount_begin(&conf->gen_lock);
6690
	conf->previous_raid_disks = conf->raid_disks;
6691
	conf->raid_disks += mddev->delta_disks;
6692 6693
	conf->prev_chunk_sectors = conf->chunk_sectors;
	conf->chunk_sectors = mddev->new_chunk_sectors;
6694 6695
	conf->prev_algo = conf->algorithm;
	conf->algorithm = mddev->new_layout;
6696 6697 6698 6699 6700
	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();
6701
	if (mddev->reshape_backwards)
6702 6703 6704 6705
		conf->reshape_progress = raid5_size(mddev, 0, 0);
	else
		conf->reshape_progress = 0;
	conf->reshape_safe = conf->reshape_progress;
6706
	write_seqcount_end(&conf->gen_lock);
6707 6708
	spin_unlock_irq(&conf->device_lock);

6709 6710 6711 6712 6713 6714 6715
	/* Now make sure any requests that proceeded on the assumption
	 * the reshape wasn't running - like Discard or Read - have
	 * completed.
	 */
	mddev_suspend(mddev);
	mddev_resume(mddev);

6716 6717
	/* Add some new drives, as many as will fit.
	 * We know there are enough to make the newly sized array work.
6718 6719 6720 6721
	 * 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.
6722
	 */
6723
	if (mddev->delta_disks >= 0) {
N
NeilBrown 已提交
6724
		rdev_for_each(rdev, mddev)
6725 6726 6727 6728
			if (rdev->raid_disk < 0 &&
			    !test_bit(Faulty, &rdev->flags)) {
				if (raid5_add_disk(mddev, rdev) == 0) {
					if (rdev->raid_disk
6729
					    >= conf->previous_raid_disks)
6730
						set_bit(In_sync, &rdev->flags);
6731
					else
6732
						rdev->recovery_offset = 0;
6733 6734

					if (sysfs_link_rdev(mddev, rdev))
6735
						/* Failure here is OK */;
6736
				}
6737 6738 6739 6740 6741
			} 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);
			}
6742

6743 6744 6745 6746
		/* When a reshape changes the number of devices,
		 * ->degraded is measured against the larger of the
		 * pre and post number of devices.
		 */
6747
		spin_lock_irqsave(&conf->device_lock, flags);
6748
		mddev->degraded = calc_degraded(conf);
6749 6750
		spin_unlock_irqrestore(&conf->device_lock, flags);
	}
6751
	mddev->raid_disks = conf->raid_disks;
6752
	mddev->reshape_position = conf->reshape_progress;
6753
	set_bit(MD_CHANGE_DEVS, &mddev->flags);
6754

6755 6756 6757 6758 6759
	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,
6760
						"reshape");
6761 6762 6763
	if (!mddev->sync_thread) {
		mddev->recovery = 0;
		spin_lock_irq(&conf->device_lock);
6764
		write_seqcount_begin(&conf->gen_lock);
6765
		mddev->raid_disks = conf->raid_disks = conf->previous_raid_disks;
6766 6767 6768
		mddev->new_chunk_sectors =
			conf->chunk_sectors = conf->prev_chunk_sectors;
		mddev->new_layout = conf->algorithm = conf->prev_algo;
6769 6770 6771
		rdev_for_each(rdev, mddev)
			rdev->new_data_offset = rdev->data_offset;
		smp_wmb();
6772
		conf->generation --;
6773
		conf->reshape_progress = MaxSector;
6774
		mddev->reshape_position = MaxSector;
6775
		write_seqcount_end(&conf->gen_lock);
6776 6777 6778
		spin_unlock_irq(&conf->device_lock);
		return -EAGAIN;
	}
6779
	conf->reshape_checkpoint = jiffies;
6780 6781 6782 6783 6784
	md_wakeup_thread(mddev->sync_thread);
	md_new_event(mddev);
	return 0;
}

6785 6786 6787
/* This is called from the reshape thread and should make any
 * changes needed in 'conf'
 */
6788
static void end_reshape(struct r5conf *conf)
6789 6790
{

6791
	if (!test_bit(MD_RECOVERY_INTR, &conf->mddev->recovery)) {
6792
		struct md_rdev *rdev;
6793 6794

		spin_lock_irq(&conf->device_lock);
6795
		conf->previous_raid_disks = conf->raid_disks;
6796 6797 6798
		rdev_for_each(rdev, conf->mddev)
			rdev->data_offset = rdev->new_data_offset;
		smp_wmb();
6799
		conf->reshape_progress = MaxSector;
6800
		spin_unlock_irq(&conf->device_lock);
6801
		wake_up(&conf->wait_for_overlap);
6802 6803 6804 6805

		/* read-ahead size must cover two whole stripes, which is
		 * 2 * (datadisks) * chunksize where 'n' is the number of raid devices
		 */
6806
		if (conf->mddev->queue) {
6807
			int data_disks = conf->raid_disks - conf->max_degraded;
6808
			int stripe = data_disks * ((conf->chunk_sectors << 9)
6809
						   / PAGE_SIZE);
6810 6811 6812
			if (conf->mddev->queue->backing_dev_info.ra_pages < 2 * stripe)
				conf->mddev->queue->backing_dev_info.ra_pages = 2 * stripe;
		}
6813 6814 6815
	}
}

6816 6817 6818
/* This is called from the raid5d thread with mddev_lock held.
 * It makes config changes to the device.
 */
6819
static void raid5_finish_reshape(struct mddev *mddev)
6820
{
6821
	struct r5conf *conf = mddev->private;
6822 6823 6824

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

6825 6826 6827
		if (mddev->delta_disks > 0) {
			md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
			set_capacity(mddev->gendisk, mddev->array_sectors);
6828
			revalidate_disk(mddev->gendisk);
6829 6830
		} else {
			int d;
6831 6832 6833
			spin_lock_irq(&conf->device_lock);
			mddev->degraded = calc_degraded(conf);
			spin_unlock_irq(&conf->device_lock);
6834 6835
			for (d = conf->raid_disks ;
			     d < conf->raid_disks - mddev->delta_disks;
6836
			     d++) {
6837
				struct md_rdev *rdev = conf->disks[d].rdev;
6838 6839 6840 6841 6842
				if (rdev)
					clear_bit(In_sync, &rdev->flags);
				rdev = conf->disks[d].replacement;
				if (rdev)
					clear_bit(In_sync, &rdev->flags);
6843
			}
6844
		}
6845
		mddev->layout = conf->algorithm;
6846
		mddev->chunk_sectors = conf->chunk_sectors;
6847 6848
		mddev->reshape_position = MaxSector;
		mddev->delta_disks = 0;
6849
		mddev->reshape_backwards = 0;
6850 6851 6852
	}
}

6853
static void raid5_quiesce(struct mddev *mddev, int state)
6854
{
6855
	struct r5conf *conf = mddev->private;
6856 6857

	switch(state) {
6858 6859 6860 6861
	case 2: /* resume for a suspend */
		wake_up(&conf->wait_for_overlap);
		break;

6862
	case 1: /* stop all writes */
6863
		lock_all_device_hash_locks_irq(conf);
6864 6865 6866 6867
		/* '2' tells resync/reshape to pause so that all
		 * active stripes can drain
		 */
		conf->quiesce = 2;
6868
		wait_event_cmd(conf->wait_for_stripe,
6869 6870
				    atomic_read(&conf->active_stripes) == 0 &&
				    atomic_read(&conf->active_aligned_reads) == 0,
6871 6872
				    unlock_all_device_hash_locks_irq(conf),
				    lock_all_device_hash_locks_irq(conf));
6873
		conf->quiesce = 1;
6874
		unlock_all_device_hash_locks_irq(conf);
6875 6876
		/* allow reshape to continue */
		wake_up(&conf->wait_for_overlap);
6877 6878 6879
		break;

	case 0: /* re-enable writes */
6880
		lock_all_device_hash_locks_irq(conf);
6881 6882
		conf->quiesce = 0;
		wake_up(&conf->wait_for_stripe);
6883
		wake_up(&conf->wait_for_overlap);
6884
		unlock_all_device_hash_locks_irq(conf);
6885 6886 6887
		break;
	}
}
6888

6889
static void *raid45_takeover_raid0(struct mddev *mddev, int level)
6890
{
6891
	struct r0conf *raid0_conf = mddev->private;
6892
	sector_t sectors;
6893

D
Dan Williams 已提交
6894
	/* for raid0 takeover only one zone is supported */
6895
	if (raid0_conf->nr_strip_zones > 1) {
6896 6897
		printk(KERN_ERR "md/raid:%s: cannot takeover raid0 with more than one zone.\n",
		       mdname(mddev));
D
Dan Williams 已提交
6898 6899 6900
		return ERR_PTR(-EINVAL);
	}

6901 6902
	sectors = raid0_conf->strip_zone[0].zone_end;
	sector_div(sectors, raid0_conf->strip_zone[0].nb_dev);
6903
	mddev->dev_sectors = sectors;
D
Dan Williams 已提交
6904
	mddev->new_level = level;
6905 6906 6907 6908 6909 6910 6911 6912 6913 6914
	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);
}

6915
static void *raid5_takeover_raid1(struct mddev *mddev)
6916 6917 6918 6919 6920 6921 6922 6923 6924 6925 6926 6927 6928 6929 6930 6931 6932 6933 6934 6935 6936
{
	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;
6937
	mddev->new_chunk_sectors = chunksect;
6938 6939 6940 6941

	return setup_conf(mddev);
}

6942
static void *raid5_takeover_raid6(struct mddev *mddev)
6943 6944 6945 6946 6947 6948 6949 6950 6951 6952 6953 6954 6955 6956 6957 6958 6959 6960 6961 6962 6963 6964 6965 6966 6967 6968 6969 6970 6971 6972 6973 6974
{
	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);
}

6975
static int raid5_check_reshape(struct mddev *mddev)
6976
{
6977 6978 6979 6980
	/* 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.
6981
	 */
6982
	struct r5conf *conf = mddev->private;
6983
	int new_chunk = mddev->new_chunk_sectors;
6984

6985
	if (mddev->new_layout >= 0 && !algorithm_valid_raid5(mddev->new_layout))
6986 6987
		return -EINVAL;
	if (new_chunk > 0) {
6988
		if (!is_power_of_2(new_chunk))
6989
			return -EINVAL;
6990
		if (new_chunk < (PAGE_SIZE>>9))
6991
			return -EINVAL;
6992
		if (mddev->array_sectors & (new_chunk-1))
6993 6994 6995 6996 6997 6998
			/* not factor of array size */
			return -EINVAL;
	}

	/* They look valid */

6999
	if (mddev->raid_disks == 2) {
7000 7001 7002 7003
		/* can make the change immediately */
		if (mddev->new_layout >= 0) {
			conf->algorithm = mddev->new_layout;
			mddev->layout = mddev->new_layout;
7004 7005
		}
		if (new_chunk > 0) {
7006 7007
			conf->chunk_sectors = new_chunk ;
			mddev->chunk_sectors = new_chunk;
7008 7009 7010
		}
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
		md_wakeup_thread(mddev->thread);
7011
	}
7012
	return check_reshape(mddev);
7013 7014
}

7015
static int raid6_check_reshape(struct mddev *mddev)
7016
{
7017
	int new_chunk = mddev->new_chunk_sectors;
7018

7019
	if (mddev->new_layout >= 0 && !algorithm_valid_raid6(mddev->new_layout))
7020
		return -EINVAL;
7021
	if (new_chunk > 0) {
7022
		if (!is_power_of_2(new_chunk))
7023
			return -EINVAL;
7024
		if (new_chunk < (PAGE_SIZE >> 9))
7025
			return -EINVAL;
7026
		if (mddev->array_sectors & (new_chunk-1))
7027 7028
			/* not factor of array size */
			return -EINVAL;
7029
	}
7030 7031

	/* They look valid */
7032
	return check_reshape(mddev);
7033 7034
}

7035
static void *raid5_takeover(struct mddev *mddev)
7036 7037
{
	/* raid5 can take over:
D
Dan Williams 已提交
7038
	 *  raid0 - if there is only one strip zone - make it a raid4 layout
7039 7040 7041 7042
	 *  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
7043 7044
	if (mddev->level == 0)
		return raid45_takeover_raid0(mddev, 5);
7045 7046
	if (mddev->level == 1)
		return raid5_takeover_raid1(mddev);
7047 7048 7049 7050 7051
	if (mddev->level == 4) {
		mddev->new_layout = ALGORITHM_PARITY_N;
		mddev->new_level = 5;
		return setup_conf(mddev);
	}
7052 7053
	if (mddev->level == 6)
		return raid5_takeover_raid6(mddev);
7054 7055 7056 7057

	return ERR_PTR(-EINVAL);
}

7058
static void *raid4_takeover(struct mddev *mddev)
7059
{
D
Dan Williams 已提交
7060 7061 7062
	/* raid4 can take over:
	 *  raid0 - if there is only one strip zone
	 *  raid5 - if layout is right
7063
	 */
D
Dan Williams 已提交
7064 7065
	if (mddev->level == 0)
		return raid45_takeover_raid0(mddev, 4);
7066 7067 7068 7069 7070 7071 7072 7073
	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);
}
7074

7075
static struct md_personality raid5_personality;
7076

7077
static void *raid6_takeover(struct mddev *mddev)
7078 7079 7080 7081 7082 7083 7084 7085 7086 7087 7088 7089 7090 7091 7092 7093 7094 7095 7096 7097 7098 7099 7100 7101 7102 7103 7104 7105 7106 7107 7108 7109 7110 7111 7112 7113 7114 7115 7116 7117 7118 7119 7120 7121 7122
{
	/* 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);
}

7123
static struct md_personality raid6_personality =
7124 7125 7126 7127 7128 7129
{
	.name		= "raid6",
	.level		= 6,
	.owner		= THIS_MODULE,
	.make_request	= make_request,
	.run		= run,
N
NeilBrown 已提交
7130
	.free		= raid5_free,
7131 7132 7133 7134 7135 7136 7137
	.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,
7138
	.size		= raid5_size,
7139
	.check_reshape	= raid6_check_reshape,
7140
	.start_reshape  = raid5_start_reshape,
7141
	.finish_reshape = raid5_finish_reshape,
7142
	.quiesce	= raid5_quiesce,
7143
	.takeover	= raid6_takeover,
7144
	.congested	= raid5_congested,
7145
	.mergeable_bvec	= raid5_mergeable_bvec,
7146
};
7147
static struct md_personality raid5_personality =
L
Linus Torvalds 已提交
7148 7149
{
	.name		= "raid5",
7150
	.level		= 5,
L
Linus Torvalds 已提交
7151 7152 7153
	.owner		= THIS_MODULE,
	.make_request	= make_request,
	.run		= run,
N
NeilBrown 已提交
7154
	.free		= raid5_free,
L
Linus Torvalds 已提交
7155 7156 7157 7158 7159 7160 7161
	.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,
7162
	.size		= raid5_size,
7163 7164
	.check_reshape	= raid5_check_reshape,
	.start_reshape  = raid5_start_reshape,
7165
	.finish_reshape = raid5_finish_reshape,
7166
	.quiesce	= raid5_quiesce,
7167
	.takeover	= raid5_takeover,
7168
	.congested	= raid5_congested,
7169
	.mergeable_bvec	= raid5_mergeable_bvec,
L
Linus Torvalds 已提交
7170 7171
};

7172
static struct md_personality raid4_personality =
L
Linus Torvalds 已提交
7173
{
7174 7175 7176 7177 7178
	.name		= "raid4",
	.level		= 4,
	.owner		= THIS_MODULE,
	.make_request	= make_request,
	.run		= run,
N
NeilBrown 已提交
7179
	.free		= raid5_free,
7180 7181 7182 7183 7184 7185 7186
	.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,
7187
	.size		= raid5_size,
7188 7189
	.check_reshape	= raid5_check_reshape,
	.start_reshape  = raid5_start_reshape,
7190
	.finish_reshape = raid5_finish_reshape,
7191
	.quiesce	= raid5_quiesce,
7192
	.takeover	= raid4_takeover,
7193
	.congested	= raid5_congested,
7194
	.mergeable_bvec	= raid5_mergeable_bvec,
7195 7196 7197 7198
};

static int __init raid5_init(void)
{
7199 7200 7201 7202
	raid5_wq = alloc_workqueue("raid5wq",
		WQ_UNBOUND|WQ_MEM_RECLAIM|WQ_CPU_INTENSIVE|WQ_SYSFS, 0);
	if (!raid5_wq)
		return -ENOMEM;
7203
	register_md_personality(&raid6_personality);
7204 7205 7206
	register_md_personality(&raid5_personality);
	register_md_personality(&raid4_personality);
	return 0;
L
Linus Torvalds 已提交
7207 7208
}

7209
static void raid5_exit(void)
L
Linus Torvalds 已提交
7210
{
7211
	unregister_md_personality(&raid6_personality);
7212 7213
	unregister_md_personality(&raid5_personality);
	unregister_md_personality(&raid4_personality);
7214
	destroy_workqueue(raid5_wq);
L
Linus Torvalds 已提交
7215 7216 7217 7218 7219
}

module_init(raid5_init);
module_exit(raid5_exit);
MODULE_LICENSE("GPL");
7220
MODULE_DESCRIPTION("RAID4/5/6 (striping with parity) personality for MD");
L
Linus Torvalds 已提交
7221
MODULE_ALIAS("md-personality-4"); /* RAID5 */
7222 7223
MODULE_ALIAS("md-raid5");
MODULE_ALIAS("md-raid4");
7224 7225
MODULE_ALIAS("md-level-5");
MODULE_ALIAS("md-level-4");
7226 7227 7228 7229 7230 7231 7232
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");