raid5.c 204.0 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>
57
#include <linux/flex_array.h>
N
NeilBrown 已提交
58 59
#include <trace/events/block.h>

60
#include "md.h"
61
#include "raid5.h"
62
#include "raid0.h"
63
#include "bitmap.h"
64

65 66 67
#define cpu_to_group(cpu) cpu_to_node(cpu)
#define ANY_GROUP NUMA_NO_NODE

68 69 70 71
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");
72
static struct workqueue_struct *raid5_wq;
L
Linus Torvalds 已提交
73 74 75 76 77 78 79 80 81
/*
 * 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
82
#define BYPASS_THRESHOLD	1
83
#define NR_HASH			(PAGE_SIZE / sizeof(struct hlist_head))
L
Linus Torvalds 已提交
84
#define HASH_MASK		(NR_HASH - 1)
85
#define MAX_STRIPE_BATCH	8
L
Linus Torvalds 已提交
86

87
static inline struct hlist_head *stripe_hash(struct r5conf *conf, sector_t sect)
88 89 90 91
{
	int hash = (sect >> STRIPE_SHIFT) & HASH_MASK;
	return &conf->stripe_hashtbl[hash];
}
L
Linus Torvalds 已提交
92

93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128
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
129 130 131 132 133 134
/* 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.
135
 * This function is used to determine the 'next' bio in the list, given the sector
L
Linus Torvalds 已提交
136 137
 * of the current stripe+device
 */
138 139
static inline struct bio *r5_next_bio(struct bio *bio, sector_t sector)
{
140
	int sectors = bio_sectors(bio);
141
	if (bio->bi_iter.bi_sector + sectors < sector + STRIPE_SECTORS)
142 143 144 145
		return bio->bi_next;
	else
		return NULL;
}
L
Linus Torvalds 已提交
146

147
/*
148 149
 * 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
150
 */
151
static inline int raid5_bi_processed_stripes(struct bio *bio)
152
{
153 154
	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
	return (atomic_read(segments) >> 16) & 0xffff;
155 156
}

157
static inline int raid5_dec_bi_active_stripes(struct bio *bio)
158
{
159 160
	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
	return atomic_sub_return(1, segments) & 0xffff;
161 162
}

163
static inline void raid5_inc_bi_active_stripes(struct bio *bio)
164
{
165 166
	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
	atomic_inc(segments);
167 168
}

169 170
static inline void raid5_set_bi_processed_stripes(struct bio *bio,
	unsigned int cnt)
171
{
172 173
	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
	int old, new;
174

175 176 177 178
	do {
		old = atomic_read(segments);
		new = (old & 0xffff) | (cnt << 16);
	} while (atomic_cmpxchg(segments, old, new) != old);
179 180
}

181
static inline void raid5_set_bi_stripes(struct bio *bio, unsigned int cnt)
182
{
183 184
	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
	atomic_set(segments, cnt);
185 186
}

187 188 189
/* Find first data disk in a raid6 stripe */
static inline int raid6_d0(struct stripe_head *sh)
{
190 191 192 193
	if (sh->ddf_layout)
		/* ddf always start from first device */
		return 0;
	/* md starts just after Q block */
194 195 196 197 198
	if (sh->qd_idx == sh->disks - 1)
		return 0;
	else
		return sh->qd_idx + 1;
}
199 200 201 202 203
static inline int raid6_next_disk(int disk, int raid_disks)
{
	disk++;
	return (disk < raid_disks) ? disk : 0;
}
204

205 206 207 208 209
/* 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.
 */
210 211
static int raid6_idx_to_slot(int idx, struct stripe_head *sh,
			     int *count, int syndrome_disks)
212
{
213
	int slot = *count;
214

215
	if (sh->ddf_layout)
216
		(*count)++;
217
	if (idx == sh->pd_idx)
218
		return syndrome_disks;
219
	if (idx == sh->qd_idx)
220
		return syndrome_disks + 1;
221
	if (!sh->ddf_layout)
222
		(*count)++;
223 224 225
	return slot;
}

226 227 228 229 230 231 232
static void return_io(struct bio *return_bi)
{
	struct bio *bi = return_bi;
	while (bi) {

		return_bi = bi->bi_next;
		bi->bi_next = NULL;
233
		bi->bi_iter.bi_size = 0;
234 235
		trace_block_bio_complete(bdev_get_queue(bi->bi_bdev),
					 bi, 0);
236
		bio_endio(bi, 0);
237 238 239 240
		bi = return_bi;
	}
}

241
static void print_raid5_conf (struct r5conf *conf);
L
Linus Torvalds 已提交
242

243 244 245 246 247 248 249
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);
}

250 251 252 253
static void raid5_wakeup_stripe_thread(struct stripe_head *sh)
{
	struct r5conf *conf = sh->raid_conf;
	struct r5worker_group *group;
254
	int thread_cnt;
255 256 257 258 259 260 261 262 263 264 265
	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);
266 267
		group->stripes_cnt++;
		sh->group = group;
268 269 270 271 272 273 274 275 276
	}

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

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

277 278 279 280 281 282 283 284 285 286 287 288 289 290
	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--;
		}
	}
291 292
}

293 294
static void do_release_stripe(struct r5conf *conf, struct stripe_head *sh,
			      struct list_head *temp_inactive_list)
L
Linus Torvalds 已提交
295
{
296 297 298 299
	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) &&
300
		    !test_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
301
			list_add_tail(&sh->lru, &conf->delayed_list);
302
		else if (test_bit(STRIPE_BIT_DELAY, &sh->state) &&
303 304 305 306 307
			   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);
308 309 310 311 312 313
			if (conf->worker_cnt_per_group == 0) {
				list_add_tail(&sh->lru, &conf->handle_list);
			} else {
				raid5_wakeup_stripe_thread(sh);
				return;
			}
314 315 316 317 318 319 320 321 322
		}
		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);
323 324
		if (!test_bit(STRIPE_EXPANDING, &sh->state))
			list_add_tail(&sh->lru, temp_inactive_list);
L
Linus Torvalds 已提交
325 326
	}
}
327

328 329
static void __release_stripe(struct r5conf *conf, struct stripe_head *sh,
			     struct list_head *temp_inactive_list)
330 331
{
	if (atomic_dec_and_test(&sh->count))
332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363
		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);
364 365 366
			if (list_empty(conf->inactive_list + hash) &&
			    !list_empty(list))
				atomic_dec(&conf->empty_inactive_list_nr);
367 368 369 370 371 372 373 374 375 376 377 378 379
			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);
	}
380 381
}

S
Shaohua Li 已提交
382
/* should hold conf->device_lock already */
383 384
static int release_stripe_list(struct r5conf *conf,
			       struct list_head *temp_inactive_list)
S
Shaohua Li 已提交
385 386 387 388 389 390
{
	struct stripe_head *sh;
	int count = 0;
	struct llist_node *head;

	head = llist_del_all(&conf->released_stripes);
S
Shaohua Li 已提交
391
	head = llist_reverse_order(head);
S
Shaohua Li 已提交
392
	while (head) {
393 394
		int hash;

S
Shaohua Li 已提交
395 396 397 398 399 400 401 402 403 404
		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.
		 */
405 406
		hash = sh->hash_lock_index;
		__release_stripe(conf, sh, &temp_inactive_list[hash]);
S
Shaohua Li 已提交
407 408 409 410 411 412
		count++;
	}

	return count;
}

L
Linus Torvalds 已提交
413 414
static void release_stripe(struct stripe_head *sh)
{
415
	struct r5conf *conf = sh->raid_conf;
L
Linus Torvalds 已提交
416
	unsigned long flags;
417 418
	struct list_head list;
	int hash;
S
Shaohua Li 已提交
419
	bool wakeup;
420

421 422 423 424 425
	/* Avoid release_list until the last reference.
	 */
	if (atomic_add_unless(&sh->count, -1, 1))
		return;

426 427
	if (unlikely(!conf->mddev->thread) ||
		test_and_set_bit(STRIPE_ON_RELEASE_LIST, &sh->state))
S
Shaohua Li 已提交
428 429 430 431 432 433
		goto slow_path;
	wakeup = llist_add(&sh->release_list, &conf->released_stripes);
	if (wakeup)
		md_wakeup_thread(conf->mddev->thread);
	return;
slow_path:
434
	local_irq_save(flags);
S
Shaohua Li 已提交
435
	/* we are ok here if STRIPE_ON_RELEASE_LIST is set or not */
436
	if (atomic_dec_and_lock(&sh->count, &conf->device_lock)) {
437 438 439
		INIT_LIST_HEAD(&list);
		hash = sh->hash_lock_index;
		do_release_stripe(conf, sh, &list);
440
		spin_unlock(&conf->device_lock);
441
		release_inactive_stripe_list(conf, &list, hash);
442 443
	}
	local_irq_restore(flags);
L
Linus Torvalds 已提交
444 445
}

446
static inline void remove_hash(struct stripe_head *sh)
L
Linus Torvalds 已提交
447
{
448 449
	pr_debug("remove_hash(), stripe %llu\n",
		(unsigned long long)sh->sector);
L
Linus Torvalds 已提交
450

451
	hlist_del_init(&sh->hash);
L
Linus Torvalds 已提交
452 453
}

454
static inline void insert_hash(struct r5conf *conf, struct stripe_head *sh)
L
Linus Torvalds 已提交
455
{
456
	struct hlist_head *hp = stripe_hash(conf, sh->sector);
L
Linus Torvalds 已提交
457

458 459
	pr_debug("insert_hash(), stripe %llu\n",
		(unsigned long long)sh->sector);
L
Linus Torvalds 已提交
460

461
	hlist_add_head(&sh->hash, hp);
L
Linus Torvalds 已提交
462 463 464
}

/* find an idle stripe, make sure it is unhashed, and return it. */
465
static struct stripe_head *get_free_stripe(struct r5conf *conf, int hash)
L
Linus Torvalds 已提交
466 467 468 469
{
	struct stripe_head *sh = NULL;
	struct list_head *first;

470
	if (list_empty(conf->inactive_list + hash))
L
Linus Torvalds 已提交
471
		goto out;
472
	first = (conf->inactive_list + hash)->next;
L
Linus Torvalds 已提交
473 474 475 476
	sh = list_entry(first, struct stripe_head, lru);
	list_del_init(first);
	remove_hash(sh);
	atomic_inc(&conf->active_stripes);
477
	BUG_ON(hash != sh->hash_lock_index);
478 479
	if (list_empty(conf->inactive_list + hash))
		atomic_inc(&conf->empty_inactive_list_nr);
L
Linus Torvalds 已提交
480 481 482 483
out:
	return sh;
}

484
static void shrink_buffers(struct stripe_head *sh)
L
Linus Torvalds 已提交
485 486 487
{
	struct page *p;
	int i;
488
	int num = sh->raid_conf->pool_size;
L
Linus Torvalds 已提交
489

490
	for (i = 0; i < num ; i++) {
491
		WARN_ON(sh->dev[i].page != sh->dev[i].orig_page);
L
Linus Torvalds 已提交
492 493 494 495
		p = sh->dev[i].page;
		if (!p)
			continue;
		sh->dev[i].page = NULL;
496
		put_page(p);
L
Linus Torvalds 已提交
497 498 499
	}
}

500
static int grow_buffers(struct stripe_head *sh)
L
Linus Torvalds 已提交
501 502
{
	int i;
503
	int num = sh->raid_conf->pool_size;
L
Linus Torvalds 已提交
504

505
	for (i = 0; i < num; i++) {
L
Linus Torvalds 已提交
506 507 508 509 510 511
		struct page *page;

		if (!(page = alloc_page(GFP_KERNEL))) {
			return 1;
		}
		sh->dev[i].page = page;
512
		sh->dev[i].orig_page = page;
L
Linus Torvalds 已提交
513 514 515 516
	}
	return 0;
}

517
static void raid5_build_block(struct stripe_head *sh, int i, int previous);
518
static void stripe_set_idx(sector_t stripe, struct r5conf *conf, int previous,
519
			    struct stripe_head *sh);
L
Linus Torvalds 已提交
520

521
static void init_stripe(struct stripe_head *sh, sector_t sector, int previous)
L
Linus Torvalds 已提交
522
{
523
	struct r5conf *conf = sh->raid_conf;
524
	int i, seq;
L
Linus Torvalds 已提交
525

526 527
	BUG_ON(atomic_read(&sh->count) != 0);
	BUG_ON(test_bit(STRIPE_HANDLE, &sh->state));
528
	BUG_ON(stripe_operations_active(sh));
529

530
	pr_debug("init_stripe called, stripe %llu\n",
531
		(unsigned long long)sector);
532 533
retry:
	seq = read_seqcount_begin(&conf->gen_lock);
534
	sh->generation = conf->generation - previous;
535
	sh->disks = previous ? conf->previous_raid_disks : conf->raid_disks;
L
Linus Torvalds 已提交
536
	sh->sector = sector;
537
	stripe_set_idx(sector, conf, previous, sh);
L
Linus Torvalds 已提交
538 539
	sh->state = 0;

540
	for (i = sh->disks; i--; ) {
L
Linus Torvalds 已提交
541 542
		struct r5dev *dev = &sh->dev[i];

543
		if (dev->toread || dev->read || dev->towrite || dev->written ||
L
Linus Torvalds 已提交
544
		    test_bit(R5_LOCKED, &dev->flags)) {
545
			printk(KERN_ERR "sector=%llx i=%d %p %p %p %p %d\n",
L
Linus Torvalds 已提交
546
			       (unsigned long long)sh->sector, i, dev->toread,
547
			       dev->read, dev->towrite, dev->written,
L
Linus Torvalds 已提交
548
			       test_bit(R5_LOCKED, &dev->flags));
549
			WARN_ON(1);
L
Linus Torvalds 已提交
550 551
		}
		dev->flags = 0;
552
		raid5_build_block(sh, i, previous);
L
Linus Torvalds 已提交
553
	}
554 555
	if (read_seqcount_retry(&conf->gen_lock, seq))
		goto retry;
556
	sh->overwrite_disks = 0;
L
Linus Torvalds 已提交
557
	insert_hash(conf, sh);
558
	sh->cpu = smp_processor_id();
559
	set_bit(STRIPE_BATCH_READY, &sh->state);
L
Linus Torvalds 已提交
560 561
}

562
static struct stripe_head *__find_stripe(struct r5conf *conf, sector_t sector,
563
					 short generation)
L
Linus Torvalds 已提交
564 565 566
{
	struct stripe_head *sh;

567
	pr_debug("__find_stripe, sector %llu\n", (unsigned long long)sector);
568
	hlist_for_each_entry(sh, stripe_hash(conf, sector), hash)
569
		if (sh->sector == sector && sh->generation == generation)
L
Linus Torvalds 已提交
570
			return sh;
571
	pr_debug("__stripe %llu not in cache\n", (unsigned long long)sector);
L
Linus Torvalds 已提交
572 573 574
	return NULL;
}

575 576 577 578 579 580 581 582 583 584 585 586 587
/*
 * 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.
 */
588
static int calc_degraded(struct r5conf *conf)
589
{
590
	int degraded, degraded2;
591 592 593 594 595
	int i;

	rcu_read_lock();
	degraded = 0;
	for (i = 0; i < conf->previous_raid_disks; i++) {
596
		struct md_rdev *rdev = rcu_dereference(conf->disks[i].rdev);
597 598
		if (rdev && test_bit(Faulty, &rdev->flags))
			rdev = rcu_dereference(conf->disks[i].replacement);
599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616
		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();
617 618
	if (conf->raid_disks == conf->previous_raid_disks)
		return degraded;
619
	rcu_read_lock();
620
	degraded2 = 0;
621
	for (i = 0; i < conf->raid_disks; i++) {
622
		struct md_rdev *rdev = rcu_dereference(conf->disks[i].rdev);
623 624
		if (rdev && test_bit(Faulty, &rdev->flags))
			rdev = rcu_dereference(conf->disks[i].replacement);
625
		if (!rdev || test_bit(Faulty, &rdev->flags))
626
			degraded2++;
627 628 629 630 631 632 633 634 635
		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)
636
				degraded2++;
637 638
	}
	rcu_read_unlock();
639 640 641 642 643 644 645 646 647 648 649 650 651
	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);
652 653 654 655 656
	if (degraded > conf->max_degraded)
		return 1;
	return 0;
}

657
static struct stripe_head *
658
get_active_stripe(struct r5conf *conf, sector_t sector,
659
		  int previous, int noblock, int noquiesce)
L
Linus Torvalds 已提交
660 661
{
	struct stripe_head *sh;
662
	int hash = stripe_hash_locks_hash(sector);
L
Linus Torvalds 已提交
663

664
	pr_debug("get_stripe, sector %llu\n", (unsigned long long)sector);
L
Linus Torvalds 已提交
665

666
	spin_lock_irq(conf->hash_locks + hash);
L
Linus Torvalds 已提交
667 668

	do {
669
		wait_event_lock_irq(conf->wait_for_stripe,
670
				    conf->quiesce == 0 || noquiesce,
671
				    *(conf->hash_locks + hash));
672
		sh = __find_stripe(conf, sector, conf->generation - previous);
L
Linus Torvalds 已提交
673 674
		if (!sh) {
			if (!conf->inactive_blocked)
675
				sh = get_free_stripe(conf, hash);
L
Linus Torvalds 已提交
676 677 678 679
			if (noblock && sh == NULL)
				break;
			if (!sh) {
				conf->inactive_blocked = 1;
680 681 682 683 684 685 686
				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
687
				conf->inactive_blocked = 0;
688
			} else {
689
				init_stripe(sh, sector, previous);
690 691
				atomic_inc(&sh->count);
			}
692
		} else if (!atomic_inc_not_zero(&sh->count)) {
693
			spin_lock(&conf->device_lock);
694
			if (!atomic_read(&sh->count)) {
L
Linus Torvalds 已提交
695 696
				if (!test_bit(STRIPE_HANDLE, &sh->state))
					atomic_inc(&conf->active_stripes);
697 698
				BUG_ON(list_empty(&sh->lru) &&
				       !test_bit(STRIPE_EXPANDING, &sh->state));
699
				list_del_init(&sh->lru);
700 701 702 703
				if (sh->group) {
					sh->group->stripes_cnt--;
					sh->group = NULL;
				}
L
Linus Torvalds 已提交
704
			}
705
			atomic_inc(&sh->count);
706
			spin_unlock(&conf->device_lock);
L
Linus Torvalds 已提交
707 708 709
		}
	} while (sh == NULL);

710
	spin_unlock_irq(conf->hash_locks + hash);
L
Linus Torvalds 已提交
711 712 713
	return sh;
}

714 715 716 717 718 719
static bool is_full_stripe_write(struct stripe_head *sh)
{
	BUG_ON(sh->overwrite_disks > (sh->disks - sh->raid_conf->max_degraded));
	return sh->overwrite_disks == (sh->disks - sh->raid_conf->max_degraded);
}

720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740
/* 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;
}

741 742 743 744
static void
raid5_end_read_request(struct bio *bi, int error);
static void
raid5_end_write_request(struct bio *bi, int error);
745

746
static void ops_run_io(struct stripe_head *sh, struct stripe_head_state *s)
747
{
748
	struct r5conf *conf = sh->raid_conf;
749 750 751 752 753 754
	int i, disks = sh->disks;

	might_sleep();

	for (i = disks; i--; ) {
		int rw;
755
		int replace_only = 0;
756 757
		struct bio *bi, *rbi;
		struct md_rdev *rdev, *rrdev = NULL;
T
Tejun Heo 已提交
758 759 760 761 762
		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;
763
			if (test_bit(R5_Discard, &sh->dev[i].flags))
S
Shaohua Li 已提交
764
				rw |= REQ_DISCARD;
T
Tejun Heo 已提交
765
		} else if (test_and_clear_bit(R5_Wantread, &sh->dev[i].flags))
766
			rw = READ;
767 768 769 770 771
		else if (test_and_clear_bit(R5_WantReplace,
					    &sh->dev[i].flags)) {
			rw = WRITE;
			replace_only = 1;
		} else
772
			continue;
S
Shaohua Li 已提交
773 774
		if (test_and_clear_bit(R5_SyncIO, &sh->dev[i].flags))
			rw |= REQ_SYNC;
775 776

		bi = &sh->dev[i].req;
777
		rbi = &sh->dev[i].rreq; /* For writing to replacement */
778 779

		rcu_read_lock();
780
		rrdev = rcu_dereference(conf->disks[i].replacement);
781 782 783 784 785 786
		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;
		}
787 788 789
		if (rw & WRITE) {
			if (replace_only)
				rdev = NULL;
790 791 792
			if (rdev == rrdev)
				/* We raced and saw duplicates */
				rrdev = NULL;
793
		} else {
794
			if (test_bit(R5_ReadRepl, &sh->dev[i].flags) && rrdev)
795 796 797
				rdev = rrdev;
			rrdev = NULL;
		}
798

799 800 801 802
		if (rdev && test_bit(Faulty, &rdev->flags))
			rdev = NULL;
		if (rdev)
			atomic_inc(&rdev->nr_pending);
803 804 805 806
		if (rrdev && test_bit(Faulty, &rrdev->flags))
			rrdev = NULL;
		if (rrdev)
			atomic_inc(&rrdev->nr_pending);
807 808
		rcu_read_unlock();

809
		/* We have already checked bad blocks for reads.  Now
810 811
		 * need to check for writes.  We never accept write errors
		 * on the replacement, so we don't to check rrdev.
812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831
		 */
		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);
				}
832 833 834 835 836 837
				/*
				 * Because md_wait_for_blocked_rdev
				 * will dec nr_pending, we must
				 * increment it first.
				 */
				atomic_inc(&rdev->nr_pending);
838 839 840 841 842 843 844 845
				md_wait_for_blocked_rdev(rdev, conf->mddev);
			} else {
				/* Acknowledged bad block - skip the write */
				rdev_dec_pending(rdev, conf->mddev);
				rdev = NULL;
			}
		}

846
		if (rdev) {
847 848
			if (s->syncing || s->expanding || s->expanded
			    || s->replacing)
849 850
				md_sync_acct(rdev->bdev, STRIPE_SECTORS);

D
Dan Williams 已提交
851 852
			set_bit(STRIPE_IO_STARTED, &sh->state);

K
Kent Overstreet 已提交
853
			bio_reset(bi);
854
			bi->bi_bdev = rdev->bdev;
K
Kent Overstreet 已提交
855 856 857 858 859 860
			bi->bi_rw = rw;
			bi->bi_end_io = (rw & WRITE)
				? raid5_end_write_request
				: raid5_end_read_request;
			bi->bi_private = sh;

861
			pr_debug("%s: for %llu schedule op %ld on disc %d\n",
862
				__func__, (unsigned long long)sh->sector,
863 864
				bi->bi_rw, i);
			atomic_inc(&sh->count);
865
			if (use_new_offset(conf, sh))
866
				bi->bi_iter.bi_sector = (sh->sector
867 868
						 + rdev->new_data_offset);
			else
869
				bi->bi_iter.bi_sector = (sh->sector
870
						 + rdev->data_offset);
871
			if (test_bit(R5_ReadNoMerge, &sh->dev[i].flags))
872
				bi->bi_rw |= REQ_NOMERGE;
873

874 875 876
			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
877
			bi->bi_vcnt = 1;
878 879
			bi->bi_io_vec[0].bv_len = STRIPE_SIZE;
			bi->bi_io_vec[0].bv_offset = 0;
880
			bi->bi_iter.bi_size = STRIPE_SIZE;
881 882 883 884 885 886
			/*
			 * 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;
887 888
			if (rrdev)
				set_bit(R5_DOUBLE_LOCKED, &sh->dev[i].flags);
889 890 891 892 893

			if (conf->mddev->gendisk)
				trace_block_bio_remap(bdev_get_queue(bi->bi_bdev),
						      bi, disk_devt(conf->mddev->gendisk),
						      sh->dev[i].sector);
894
			generic_make_request(bi);
895 896
		}
		if (rrdev) {
897 898
			if (s->syncing || s->expanding || s->expanded
			    || s->replacing)
899 900 901 902
				md_sync_acct(rrdev->bdev, STRIPE_SECTORS);

			set_bit(STRIPE_IO_STARTED, &sh->state);

K
Kent Overstreet 已提交
903
			bio_reset(rbi);
904
			rbi->bi_bdev = rrdev->bdev;
K
Kent Overstreet 已提交
905 906 907 908 909
			rbi->bi_rw = rw;
			BUG_ON(!(rw & WRITE));
			rbi->bi_end_io = raid5_end_write_request;
			rbi->bi_private = sh;

910 911 912 913 914
			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);
915
			if (use_new_offset(conf, sh))
916
				rbi->bi_iter.bi_sector = (sh->sector
917 918
						  + rrdev->new_data_offset);
			else
919
				rbi->bi_iter.bi_sector = (sh->sector
920
						  + rrdev->data_offset);
921 922 923
			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
924
			rbi->bi_vcnt = 1;
925 926
			rbi->bi_io_vec[0].bv_len = STRIPE_SIZE;
			rbi->bi_io_vec[0].bv_offset = 0;
927
			rbi->bi_iter.bi_size = STRIPE_SIZE;
928 929 930 931 932 933
			/*
			 * 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;
934 935 936 937
			if (conf->mddev->gendisk)
				trace_block_bio_remap(bdev_get_queue(rbi->bi_bdev),
						      rbi, disk_devt(conf->mddev->gendisk),
						      sh->dev[i].sector);
938 939 940
			generic_make_request(rbi);
		}
		if (!rdev && !rrdev) {
941
			if (rw & WRITE)
942 943 944 945 946 947 948 949 950 951
				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 *
952 953 954
async_copy_data(int frombio, struct bio *bio, struct page **page,
	sector_t sector, struct dma_async_tx_descriptor *tx,
	struct stripe_head *sh)
955
{
956 957
	struct bio_vec bvl;
	struct bvec_iter iter;
958 959
	struct page *bio_page;
	int page_offset;
960
	struct async_submit_ctl submit;
D
Dan Williams 已提交
961
	enum async_tx_flags flags = 0;
962

963 964
	if (bio->bi_iter.bi_sector >= sector)
		page_offset = (signed)(bio->bi_iter.bi_sector - sector) * 512;
965
	else
966
		page_offset = (signed)(sector - bio->bi_iter.bi_sector) * -512;
967

D
Dan Williams 已提交
968 969 970 971
	if (frombio)
		flags |= ASYNC_TX_FENCE;
	init_async_submit(&submit, flags, tx, NULL, NULL, NULL);

972 973
	bio_for_each_segment(bvl, bio, iter) {
		int len = bvl.bv_len;
974 975 976 977 978 979 980 981 982 983 984 985 986 987 988
		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) {
989 990
			b_offset += bvl.bv_offset;
			bio_page = bvl.bv_page;
991 992 993 994 995 996 997
			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,
998
						  b_offset, clen, &submit);
999 1000
			} else
				tx = async_memcpy(bio_page, *page, b_offset,
1001
						  page_offset, clen, &submit);
1002
		}
1003 1004 1005
		/* chain the operations */
		submit.depend_tx = tx;

1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017
		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;
1018
	int i;
1019

1020
	pr_debug("%s: stripe %llu\n", __func__,
1021 1022 1023 1024 1025 1026 1027
		(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 */
1028 1029
		/* and check if we need to reply to a read request,
		 * new R5_Wantfill requests are held off until
1030
		 * !STRIPE_BIOFILL_RUN
1031 1032
		 */
		if (test_and_clear_bit(R5_Wantfill, &dev->flags)) {
1033 1034 1035 1036 1037
			struct bio *rbi, *rbi2;

			BUG_ON(!dev->read);
			rbi = dev->read;
			dev->read = NULL;
1038
			while (rbi && rbi->bi_iter.bi_sector <
1039 1040
				dev->sector + STRIPE_SECTORS) {
				rbi2 = r5_next_bio(rbi, dev->sector);
1041
				if (!raid5_dec_bi_active_stripes(rbi)) {
1042 1043 1044 1045 1046 1047 1048
					rbi->bi_next = return_bi;
					return_bi = rbi;
				}
				rbi = rbi2;
			}
		}
	}
1049
	clear_bit(STRIPE_BIOFILL_RUN, &sh->state);
1050 1051 1052

	return_io(return_bi);

1053
	set_bit(STRIPE_HANDLE, &sh->state);
1054 1055 1056 1057 1058 1059
	release_stripe(sh);
}

static void ops_run_biofill(struct stripe_head *sh)
{
	struct dma_async_tx_descriptor *tx = NULL;
1060
	struct async_submit_ctl submit;
1061 1062
	int i;

1063
	pr_debug("%s: stripe %llu\n", __func__,
1064 1065 1066 1067 1068 1069
		(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
1070
			spin_lock_irq(&sh->stripe_lock);
1071 1072
			dev->read = rbi = dev->toread;
			dev->toread = NULL;
S
Shaohua Li 已提交
1073
			spin_unlock_irq(&sh->stripe_lock);
1074
			while (rbi && rbi->bi_iter.bi_sector <
1075
				dev->sector + STRIPE_SECTORS) {
1076 1077
				tx = async_copy_data(0, rbi, &dev->page,
					dev->sector, tx, sh);
1078 1079 1080 1081 1082 1083
				rbi = r5_next_bio(rbi, dev->sector);
			}
		}
	}

	atomic_inc(&sh->count);
1084 1085
	init_async_submit(&submit, ASYNC_TX_ACK, tx, ops_complete_biofill, sh, NULL);
	async_trigger_callback(&submit);
1086 1087
}

1088
static void mark_target_uptodate(struct stripe_head *sh, int target)
1089
{
1090
	struct r5dev *tgt;
1091

1092 1093
	if (target < 0)
		return;
1094

1095
	tgt = &sh->dev[target];
1096 1097 1098
	set_bit(R5_UPTODATE, &tgt->flags);
	BUG_ON(!test_bit(R5_Wantcompute, &tgt->flags));
	clear_bit(R5_Wantcompute, &tgt->flags);
1099 1100
}

1101
static void ops_complete_compute(void *stripe_head_ref)
1102 1103 1104
{
	struct stripe_head *sh = stripe_head_ref;

1105
	pr_debug("%s: stripe %llu\n", __func__,
1106 1107
		(unsigned long long)sh->sector);

1108
	/* mark the computed target(s) as uptodate */
1109
	mark_target_uptodate(sh, sh->ops.target);
1110
	mark_target_uptodate(sh, sh->ops.target2);
1111

1112 1113 1114
	clear_bit(STRIPE_COMPUTE_RUN, &sh->state);
	if (sh->check_state == check_state_compute_run)
		sh->check_state = check_state_compute_result;
1115 1116 1117 1118
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

1119 1120
/* return a pointer to the address conversion region of the scribble buffer */
static addr_conv_t *to_addr_conv(struct stripe_head *sh,
1121
				 struct raid5_percpu *percpu, int i)
1122
{
1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135
	void *addr;

	addr = flex_array_get(percpu->scribble, i);
	return addr + sizeof(struct page *) * (sh->disks + 2);
}

/* return a pointer to the address conversion region of the scribble buffer */
static struct page **to_addr_page(struct raid5_percpu *percpu, int i)
{
	void *addr;

	addr = flex_array_get(percpu->scribble, i);
	return addr;
1136 1137 1138 1139
}

static struct dma_async_tx_descriptor *
ops_run_compute5(struct stripe_head *sh, struct raid5_percpu *percpu)
1140 1141
{
	int disks = sh->disks;
1142
	struct page **xor_srcs = to_addr_page(percpu, 0);
1143 1144 1145 1146 1147
	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;
1148
	struct async_submit_ctl submit;
1149 1150 1151
	int i;

	pr_debug("%s: stripe %llu block: %d\n",
1152
		__func__, (unsigned long long)sh->sector, target);
1153 1154 1155 1156 1157 1158 1159 1160
	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
1161
	init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST, NULL,
1162
			  ops_complete_compute, sh, to_addr_conv(sh, percpu, 0));
1163
	if (unlikely(count == 1))
1164
		tx = async_memcpy(xor_dest, xor_srcs[0], 0, 0, STRIPE_SIZE, &submit);
1165
	else
1166
		tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, &submit);
1167 1168 1169 1170

	return tx;
}

1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188
/* 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++)
1189
		srcs[i] = NULL;
1190 1191 1192 1193 1194 1195 1196 1197 1198 1199

	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);

1200
	return syndrome_disks;
1201 1202 1203 1204 1205 1206
}

static struct dma_async_tx_descriptor *
ops_run_compute6_1(struct stripe_head *sh, struct raid5_percpu *percpu)
{
	int disks = sh->disks;
1207
	struct page **blocks = to_addr_page(percpu, 0);
1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220
	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;
1221
	else
1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237
		/* 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
1238 1239
		init_async_submit(&submit, ASYNC_TX_FENCE, NULL,
				  ops_complete_compute, sh,
1240
				  to_addr_conv(sh, percpu, 0));
1241 1242 1243 1244 1245 1246 1247 1248 1249 1250
		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
1251 1252
		init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST,
				  NULL, ops_complete_compute, sh,
1253
				  to_addr_conv(sh, percpu, 0));
1254 1255
		tx = async_xor(dest, blocks, 0, count, STRIPE_SIZE, &submit);
	}
1256 1257 1258 1259

	return tx;
}

1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271
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;
1272
	struct page **blocks = to_addr_page(percpu, 0);
1273 1274 1275 1276 1277 1278 1279 1280
	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));

1281
	/* we need to open-code set_syndrome_sources to handle the
1282 1283 1284
	 * slot number conversion for 'faila' and 'failb'
	 */
	for (i = 0; i < disks ; i++)
1285
		blocks[i] = NULL;
1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311
	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
1312 1313
			init_async_submit(&submit, ASYNC_TX_FENCE, NULL,
					  ops_complete_compute, sh,
1314
					  to_addr_conv(sh, percpu, 0));
1315
			return async_gen_syndrome(blocks, 0, syndrome_disks+2,
1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334
						  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
1335 1336 1337
			init_async_submit(&submit,
					  ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST,
					  NULL, NULL, NULL,
1338
					  to_addr_conv(sh, percpu, 0));
1339 1340 1341 1342
			tx = async_xor(dest, blocks, 0, count, STRIPE_SIZE,
				       &submit);

			count = set_syndrome_sources(blocks, sh);
D
Dan Williams 已提交
1343 1344
			init_async_submit(&submit, ASYNC_TX_FENCE, tx,
					  ops_complete_compute, sh,
1345
					  to_addr_conv(sh, percpu, 0));
1346 1347 1348 1349
			return async_gen_syndrome(blocks, 0, count+2,
						  STRIPE_SIZE, &submit);
		}
	} else {
1350 1351
		init_async_submit(&submit, ASYNC_TX_FENCE, NULL,
				  ops_complete_compute, sh,
1352
				  to_addr_conv(sh, percpu, 0));
1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363
		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);
		}
1364 1365 1366
	}
}

1367 1368 1369 1370
static void ops_complete_prexor(void *stripe_head_ref)
{
	struct stripe_head *sh = stripe_head_ref;

1371
	pr_debug("%s: stripe %llu\n", __func__,
1372 1373 1374 1375
		(unsigned long long)sh->sector);
}

static struct dma_async_tx_descriptor *
1376 1377
ops_run_prexor(struct stripe_head *sh, struct raid5_percpu *percpu,
	       struct dma_async_tx_descriptor *tx)
1378 1379
{
	int disks = sh->disks;
1380
	struct page **xor_srcs = to_addr_page(percpu, 0);
1381
	int count = 0, pd_idx = sh->pd_idx, i;
1382
	struct async_submit_ctl submit;
1383 1384 1385 1386

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

1387
	pr_debug("%s: stripe %llu\n", __func__,
1388 1389 1390 1391 1392
		(unsigned long long)sh->sector);

	for (i = disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
		/* Only process blocks that are known to be uptodate */
1393
		if (test_bit(R5_Wantdrain, &dev->flags))
1394 1395 1396
			xor_srcs[count++] = dev->page;
	}

D
Dan Williams 已提交
1397
	init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_DROP_DST, tx,
1398
			  ops_complete_prexor, sh, to_addr_conv(sh, percpu, 0));
1399
	tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, &submit);
1400 1401 1402 1403 1404

	return tx;
}

static struct dma_async_tx_descriptor *
1405
ops_run_biodrain(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
1406 1407
{
	int disks = sh->disks;
1408
	int i;
1409

1410
	pr_debug("%s: stripe %llu\n", __func__,
1411 1412 1413 1414 1415 1416
		(unsigned long long)sh->sector);

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

1417
		if (test_and_clear_bit(R5_Wantdrain, &dev->flags)) {
1418 1419
			struct bio *wbi;

S
Shaohua Li 已提交
1420
			spin_lock_irq(&sh->stripe_lock);
1421 1422
			chosen = dev->towrite;
			dev->towrite = NULL;
1423
			sh->overwrite_disks = 0;
1424 1425
			BUG_ON(dev->written);
			wbi = dev->written = chosen;
S
Shaohua Li 已提交
1426
			spin_unlock_irq(&sh->stripe_lock);
1427
			WARN_ON(dev->page != dev->orig_page);
1428

1429
			while (wbi && wbi->bi_iter.bi_sector <
1430
				dev->sector + STRIPE_SECTORS) {
T
Tejun Heo 已提交
1431 1432
				if (wbi->bi_rw & REQ_FUA)
					set_bit(R5_WantFUA, &dev->flags);
S
Shaohua Li 已提交
1433 1434
				if (wbi->bi_rw & REQ_SYNC)
					set_bit(R5_SyncIO, &dev->flags);
1435
				if (wbi->bi_rw & REQ_DISCARD)
S
Shaohua Li 已提交
1436
					set_bit(R5_Discard, &dev->flags);
1437 1438 1439 1440 1441 1442 1443 1444 1445
				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);
					}
				}
1446 1447 1448 1449 1450 1451 1452 1453
				wbi = r5_next_bio(wbi, dev->sector);
			}
		}
	}

	return tx;
}

1454
static void ops_complete_reconstruct(void *stripe_head_ref)
1455 1456
{
	struct stripe_head *sh = stripe_head_ref;
1457 1458 1459 1460
	int disks = sh->disks;
	int pd_idx = sh->pd_idx;
	int qd_idx = sh->qd_idx;
	int i;
1461
	bool fua = false, sync = false, discard = false;
1462

1463
	pr_debug("%s: stripe %llu\n", __func__,
1464 1465
		(unsigned long long)sh->sector);

S
Shaohua Li 已提交
1466
	for (i = disks; i--; ) {
T
Tejun Heo 已提交
1467
		fua |= test_bit(R5_WantFUA, &sh->dev[i].flags);
S
Shaohua Li 已提交
1468
		sync |= test_bit(R5_SyncIO, &sh->dev[i].flags);
1469
		discard |= test_bit(R5_Discard, &sh->dev[i].flags);
S
Shaohua Li 已提交
1470
	}
T
Tejun Heo 已提交
1471

1472 1473
	for (i = disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
1474

T
Tejun Heo 已提交
1475
		if (dev->written || i == pd_idx || i == qd_idx) {
1476
			if (!discard && !test_bit(R5_SkipCopy, &dev->flags))
1477
				set_bit(R5_UPTODATE, &dev->flags);
T
Tejun Heo 已提交
1478 1479
			if (fua)
				set_bit(R5_WantFUA, &dev->flags);
S
Shaohua Li 已提交
1480 1481
			if (sync)
				set_bit(R5_SyncIO, &dev->flags);
T
Tejun Heo 已提交
1482
		}
1483 1484
	}

1485 1486 1487 1488 1489 1490 1491 1492
	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;
	}
1493 1494 1495 1496 1497 1498

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

static void
1499 1500
ops_run_reconstruct5(struct stripe_head *sh, struct raid5_percpu *percpu,
		     struct dma_async_tx_descriptor *tx)
1501 1502
{
	int disks = sh->disks;
1503
	struct page **xor_srcs = to_addr_page(percpu, 0);
1504
	struct async_submit_ctl submit;
1505 1506
	int count = 0, pd_idx = sh->pd_idx, i;
	struct page *xor_dest;
1507
	int prexor = 0;
1508 1509
	unsigned long flags;

1510
	pr_debug("%s: stripe %llu\n", __func__,
1511 1512
		(unsigned long long)sh->sector);

S
Shaohua Li 已提交
1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524
	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;
	}
1525 1526 1527
	/* check if prexor is active which means only process blocks
	 * that are part of a read-modify-write (written)
	 */
1528 1529
	if (sh->reconstruct_state == reconstruct_state_prexor_drain_run) {
		prexor = 1;
1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549
		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
	 */
1550
	flags = ASYNC_TX_ACK |
1551 1552 1553 1554
		(prexor ? ASYNC_TX_XOR_DROP_DST : ASYNC_TX_XOR_ZERO_DST);

	atomic_inc(&sh->count);

1555
	init_async_submit(&submit, flags, tx, ops_complete_reconstruct, sh,
1556
			  to_addr_conv(sh, percpu, 0));
1557 1558 1559 1560
	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);
1561 1562
}

1563 1564 1565 1566 1567
static void
ops_run_reconstruct6(struct stripe_head *sh, struct raid5_percpu *percpu,
		     struct dma_async_tx_descriptor *tx)
{
	struct async_submit_ctl submit;
1568
	struct page **blocks = to_addr_page(percpu, 0);
S
Shaohua Li 已提交
1569
	int count, i;
1570 1571 1572

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

S
Shaohua Li 已提交
1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586
	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;
	}

1587 1588 1589 1590 1591
	count = set_syndrome_sources(blocks, sh);

	atomic_inc(&sh->count);

	init_async_submit(&submit, ASYNC_TX_ACK, tx, ops_complete_reconstruct,
1592
			  sh, to_addr_conv(sh, percpu, 0));
1593
	async_gen_syndrome(blocks, 0, count+2, STRIPE_SIZE,  &submit);
1594 1595 1596 1597 1598 1599
}

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

1600
	pr_debug("%s: stripe %llu\n", __func__,
1601 1602
		(unsigned long long)sh->sector);

1603
	sh->check_state = check_state_check_result;
1604 1605 1606 1607
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

1608
static void ops_run_check_p(struct stripe_head *sh, struct raid5_percpu *percpu)
1609 1610
{
	int disks = sh->disks;
1611 1612 1613
	int pd_idx = sh->pd_idx;
	int qd_idx = sh->qd_idx;
	struct page *xor_dest;
1614
	struct page **xor_srcs = to_addr_page(percpu, 0);
1615
	struct dma_async_tx_descriptor *tx;
1616
	struct async_submit_ctl submit;
1617 1618
	int count;
	int i;
1619

1620
	pr_debug("%s: stripe %llu\n", __func__,
1621 1622
		(unsigned long long)sh->sector);

1623 1624 1625
	count = 0;
	xor_dest = sh->dev[pd_idx].page;
	xor_srcs[count++] = xor_dest;
1626
	for (i = disks; i--; ) {
1627 1628 1629
		if (i == pd_idx || i == qd_idx)
			continue;
		xor_srcs[count++] = sh->dev[i].page;
1630 1631
	}

1632
	init_async_submit(&submit, 0, NULL, NULL, NULL,
1633
			  to_addr_conv(sh, percpu, 0));
D
Dan Williams 已提交
1634
	tx = async_xor_val(xor_dest, xor_srcs, 0, count, STRIPE_SIZE,
1635
			   &sh->ops.zero_sum_result, &submit);
1636 1637

	atomic_inc(&sh->count);
1638 1639
	init_async_submit(&submit, ASYNC_TX_ACK, tx, ops_complete_check, sh, NULL);
	tx = async_trigger_callback(&submit);
1640 1641
}

1642 1643
static void ops_run_check_pq(struct stripe_head *sh, struct raid5_percpu *percpu, int checkp)
{
1644
	struct page **srcs = to_addr_page(percpu, 0);
1645 1646 1647 1648 1649 1650 1651 1652 1653
	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;
1654 1655

	atomic_inc(&sh->count);
1656
	init_async_submit(&submit, ASYNC_TX_ACK, NULL, ops_complete_check,
1657
			  sh, to_addr_conv(sh, percpu, 0));
1658 1659
	async_syndrome_val(srcs, 0, count+2, STRIPE_SIZE,
			   &sh->ops.zero_sum_result, percpu->spare_page, &submit);
1660 1661
}

N
NeilBrown 已提交
1662
static void raid_run_ops(struct stripe_head *sh, unsigned long ops_request)
1663 1664 1665
{
	int overlap_clear = 0, i, disks = sh->disks;
	struct dma_async_tx_descriptor *tx = NULL;
1666
	struct r5conf *conf = sh->raid_conf;
1667
	int level = conf->level;
1668 1669
	struct raid5_percpu *percpu;
	unsigned long cpu;
1670

1671 1672
	cpu = get_cpu();
	percpu = per_cpu_ptr(conf->percpu, cpu);
1673
	if (test_bit(STRIPE_OP_BIOFILL, &ops_request)) {
1674 1675 1676 1677
		ops_run_biofill(sh);
		overlap_clear++;
	}

1678
	if (test_bit(STRIPE_OP_COMPUTE_BLK, &ops_request)) {
1679 1680 1681 1682 1683 1684 1685 1686 1687 1688
		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))
1689 1690
			async_tx_ack(tx);
	}
1691

1692
	if (test_bit(STRIPE_OP_PREXOR, &ops_request))
1693
		tx = ops_run_prexor(sh, percpu, tx);
1694

1695
	if (test_bit(STRIPE_OP_BIODRAIN, &ops_request)) {
1696
		tx = ops_run_biodrain(sh, tx);
1697 1698 1699
		overlap_clear++;
	}

1700 1701 1702 1703 1704 1705
	if (test_bit(STRIPE_OP_RECONSTRUCT, &ops_request)) {
		if (level < 6)
			ops_run_reconstruct5(sh, percpu, tx);
		else
			ops_run_reconstruct6(sh, percpu, tx);
	}
1706

1707 1708 1709 1710 1711 1712 1713 1714 1715 1716
	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();
	}
1717 1718 1719 1720 1721 1722 1723

	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);
		}
1724
	put_cpu();
1725 1726
}

1727
static int grow_one_stripe(struct r5conf *conf, int hash)
L
Linus Torvalds 已提交
1728 1729
{
	struct stripe_head *sh;
N
Namhyung Kim 已提交
1730
	sh = kmem_cache_zalloc(conf->slab_cache, GFP_KERNEL);
1731 1732
	if (!sh)
		return 0;
N
Namhyung Kim 已提交
1733

1734 1735
	sh->raid_conf = conf;

S
Shaohua Li 已提交
1736 1737
	spin_lock_init(&sh->stripe_lock);

1738 1739
	if (grow_buffers(sh)) {
		shrink_buffers(sh);
1740 1741 1742
		kmem_cache_free(conf->slab_cache, sh);
		return 0;
	}
1743
	sh->hash_lock_index = hash;
1744 1745 1746 1747 1748 1749 1750 1751
	/* 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;
}

1752
static int grow_stripes(struct r5conf *conf, int num)
1753
{
1754
	struct kmem_cache *sc;
1755
	int devs = max(conf->raid_disks, conf->previous_raid_disks);
1756
	int hash;
L
Linus Torvalds 已提交
1757

1758 1759 1760 1761 1762 1763 1764 1765
	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]);

1766 1767
	conf->active_name = 0;
	sc = kmem_cache_create(conf->cache_name[conf->active_name],
L
Linus Torvalds 已提交
1768
			       sizeof(struct stripe_head)+(devs-1)*sizeof(struct r5dev),
1769
			       0, 0, NULL);
L
Linus Torvalds 已提交
1770 1771 1772
	if (!sc)
		return 1;
	conf->slab_cache = sc;
1773
	conf->pool_size = devs;
1774 1775 1776
	hash = conf->max_nr_stripes % NR_STRIPE_HASH_LOCKS;
	while (num--) {
		if (!grow_one_stripe(conf, hash))
L
Linus Torvalds 已提交
1777
			return 1;
1778 1779 1780
		conf->max_nr_stripes++;
		hash = (hash + 1) % NR_STRIPE_HASH_LOCKS;
	}
L
Linus Torvalds 已提交
1781 1782
	return 0;
}
1783

1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796
/**
 * 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.
 */
1797
static struct flex_array *scribble_alloc(int num, int cnt, gfp_t flags)
1798
{
1799
	struct flex_array *ret;
1800 1801 1802
	size_t len;

	len = sizeof(struct page *) * (num+2) + sizeof(addr_conv_t) * (num+2);
1803 1804 1805 1806 1807 1808 1809 1810 1811
	ret = flex_array_alloc(len, cnt, flags);
	if (!ret)
		return NULL;
	/* always prealloc all elements, so no locking is required */
	if (flex_array_prealloc(ret, 0, cnt, flags)) {
		flex_array_free(ret);
		return NULL;
	}
	return ret;
1812 1813
}

1814
static int resize_stripes(struct r5conf *conf, int newsize)
1815 1816 1817 1818 1819 1820 1821
{
	/* 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
1822
	 * 2/ gather all the old stripe_heads and transfer the pages across
1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841
	 *    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;
1842
	unsigned long cpu;
1843
	int err;
1844
	struct kmem_cache *sc;
1845
	int i;
1846
	int hash, cnt;
1847 1848 1849 1850

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

1851 1852 1853
	err = md_allow_write(conf->mddev);
	if (err)
		return err;
1854

1855 1856 1857
	/* Step 1 */
	sc = kmem_cache_create(conf->cache_name[1-conf->active_name],
			       sizeof(struct stripe_head)+(newsize-1)*sizeof(struct r5dev),
1858
			       0, 0, NULL);
1859 1860 1861 1862
	if (!sc)
		return -ENOMEM;

	for (i = conf->max_nr_stripes; i; i--) {
N
Namhyung Kim 已提交
1863
		nsh = kmem_cache_zalloc(sc, GFP_KERNEL);
1864 1865 1866 1867
		if (!nsh)
			break;

		nsh->raid_conf = conf;
1868
		spin_lock_init(&nsh->stripe_lock);
1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885

		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
	 */
1886 1887
	hash = 0;
	cnt = 0;
1888
	list_for_each_entry(nsh, &newstripes, lru) {
1889 1890 1891 1892 1893 1894 1895
		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);
1896
		atomic_set(&nsh->count, 1);
1897
		for(i=0; i<conf->pool_size; i++) {
1898
			nsh->dev[i].page = osh->dev[i].page;
1899 1900
			nsh->dev[i].orig_page = osh->dev[i].page;
		}
1901 1902
		for( ; i<newsize; i++)
			nsh->dev[i].page = NULL;
1903
		nsh->hash_lock_index = hash;
1904
		kmem_cache_free(conf->slab_cache, osh);
1905 1906 1907 1908 1909 1910
		cnt++;
		if (cnt >= conf->max_nr_stripes / NR_STRIPE_HASH_LOCKS +
		    !!((conf->max_nr_stripes % NR_STRIPE_HASH_LOCKS) > hash)) {
			hash++;
			cnt = 0;
		}
1911 1912 1913 1914 1915 1916
	}
	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
1917
	 * conf->disks and the scribble region
1918 1919 1920 1921 1922 1923 1924 1925 1926 1927
	 */
	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;

1928 1929 1930
	get_online_cpus();
	for_each_present_cpu(cpu) {
		struct raid5_percpu *percpu;
1931
		struct flex_array *scribble;
1932 1933

		percpu = per_cpu_ptr(conf->percpu, cpu);
1934 1935
		scribble = scribble_alloc(newsize, conf->chunk_sectors /
			STRIPE_SECTORS, GFP_NOIO);
1936 1937

		if (scribble) {
1938
			flex_array_free(percpu->scribble);
1939 1940 1941 1942 1943 1944 1945 1946
			percpu->scribble = scribble;
		} else {
			err = -ENOMEM;
			break;
		}
	}
	put_online_cpus();

1947 1948 1949 1950
	/* 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);
1951

1952 1953 1954 1955
		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;
1956
				nsh->dev[i].orig_page = p;
1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968
				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
1969

1970
static int drop_one_stripe(struct r5conf *conf, int hash)
L
Linus Torvalds 已提交
1971 1972 1973
{
	struct stripe_head *sh;

1974 1975 1976
	spin_lock_irq(conf->hash_locks + hash);
	sh = get_free_stripe(conf, hash);
	spin_unlock_irq(conf->hash_locks + hash);
1977 1978
	if (!sh)
		return 0;
1979
	BUG_ON(atomic_read(&sh->count));
1980
	shrink_buffers(sh);
1981 1982 1983 1984 1985
	kmem_cache_free(conf->slab_cache, sh);
	atomic_dec(&conf->active_stripes);
	return 1;
}

1986
static void shrink_stripes(struct r5conf *conf)
1987
{
1988 1989 1990 1991
	int hash;
	for (hash = 0; hash < NR_STRIPE_HASH_LOCKS; hash++)
		while (drop_one_stripe(conf, hash))
			;
1992

N
NeilBrown 已提交
1993 1994
	if (conf->slab_cache)
		kmem_cache_destroy(conf->slab_cache);
L
Linus Torvalds 已提交
1995 1996 1997
	conf->slab_cache = NULL;
}

1998
static void raid5_end_read_request(struct bio * bi, int error)
L
Linus Torvalds 已提交
1999
{
2000
	struct stripe_head *sh = bi->bi_private;
2001
	struct r5conf *conf = sh->raid_conf;
2002
	int disks = sh->disks, i;
L
Linus Torvalds 已提交
2003
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
2004
	char b[BDEVNAME_SIZE];
2005
	struct md_rdev *rdev = NULL;
2006
	sector_t s;
L
Linus Torvalds 已提交
2007 2008 2009 2010 2011

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

2012 2013
	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
2014 2015 2016
		uptodate);
	if (i == disks) {
		BUG();
2017
		return;
L
Linus Torvalds 已提交
2018
	}
2019
	if (test_bit(R5_ReadRepl, &sh->dev[i].flags))
2020 2021 2022 2023 2024
		/* 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.
		 */
2025
		rdev = conf->disks[i].replacement;
2026
	if (!rdev)
2027
		rdev = conf->disks[i].rdev;
L
Linus Torvalds 已提交
2028

2029 2030 2031 2032
	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
2033 2034
	if (uptodate) {
		set_bit(R5_UPTODATE, &sh->dev[i].flags);
2035
		if (test_bit(R5_ReadError, &sh->dev[i].flags)) {
2036 2037 2038 2039
			/* Note that this cannot happen on a
			 * replacement device.  We just fail those on
			 * any error
			 */
2040 2041 2042 2043 2044
			printk_ratelimited(
				KERN_INFO
				"md/raid:%s: read error corrected"
				" (%lu sectors at %llu on %s)\n",
				mdname(conf->mddev), STRIPE_SECTORS,
2045
				(unsigned long long)s,
2046
				bdevname(rdev->bdev, b));
2047
			atomic_add(STRIPE_SECTORS, &rdev->corrected_errors);
2048 2049
			clear_bit(R5_ReadError, &sh->dev[i].flags);
			clear_bit(R5_ReWrite, &sh->dev[i].flags);
2050 2051 2052
		} else if (test_bit(R5_ReadNoMerge, &sh->dev[i].flags))
			clear_bit(R5_ReadNoMerge, &sh->dev[i].flags);

2053 2054
		if (atomic_read(&rdev->read_errors))
			atomic_set(&rdev->read_errors, 0);
L
Linus Torvalds 已提交
2055
	} else {
2056
		const char *bdn = bdevname(rdev->bdev, b);
2057
		int retry = 0;
2058
		int set_bad = 0;
2059

L
Linus Torvalds 已提交
2060
		clear_bit(R5_UPTODATE, &sh->dev[i].flags);
2061
		atomic_inc(&rdev->read_errors);
2062 2063 2064 2065 2066 2067
		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),
2068
				(unsigned long long)s,
2069
				bdn);
2070 2071
		else if (conf->mddev->degraded >= conf->max_degraded) {
			set_bad = 1;
2072 2073 2074 2075 2076
			printk_ratelimited(
				KERN_WARNING
				"md/raid:%s: read error not correctable "
				"(sector %llu on %s).\n",
				mdname(conf->mddev),
2077
				(unsigned long long)s,
2078
				bdn);
2079
		} else if (test_bit(R5_ReWrite, &sh->dev[i].flags)) {
2080
			/* Oh, no!!! */
2081
			set_bad = 1;
2082 2083 2084 2085 2086
			printk_ratelimited(
				KERN_WARNING
				"md/raid:%s: read error NOT corrected!! "
				"(sector %llu on %s).\n",
				mdname(conf->mddev),
2087
				(unsigned long long)s,
2088
				bdn);
2089
		} else if (atomic_read(&rdev->read_errors)
2090
			 > conf->max_nr_stripes)
N
NeilBrown 已提交
2091
			printk(KERN_WARNING
2092
			       "md/raid:%s: Too many read errors, failing device %s.\n",
2093
			       mdname(conf->mddev), bdn);
2094 2095
		else
			retry = 1;
2096 2097 2098
		if (set_bad && test_bit(In_sync, &rdev->flags)
		    && !test_bit(R5_ReadNoMerge, &sh->dev[i].flags))
			retry = 1;
2099
		if (retry)
2100 2101 2102 2103 2104
			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);
2105
		else {
2106 2107
			clear_bit(R5_ReadError, &sh->dev[i].flags);
			clear_bit(R5_ReWrite, &sh->dev[i].flags);
2108 2109 2110 2111 2112
			if (!(set_bad
			      && test_bit(In_sync, &rdev->flags)
			      && rdev_set_badblocks(
				      rdev, sh->sector, STRIPE_SECTORS, 0)))
				md_error(conf->mddev, rdev);
2113
		}
L
Linus Torvalds 已提交
2114
	}
2115
	rdev_dec_pending(rdev, conf->mddev);
L
Linus Torvalds 已提交
2116 2117 2118 2119 2120
	clear_bit(R5_LOCKED, &sh->dev[i].flags);
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

2121
static void raid5_end_write_request(struct bio *bi, int error)
L
Linus Torvalds 已提交
2122
{
2123
	struct stripe_head *sh = bi->bi_private;
2124
	struct r5conf *conf = sh->raid_conf;
2125
	int disks = sh->disks, i;
2126
	struct md_rdev *uninitialized_var(rdev);
L
Linus Torvalds 已提交
2127
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
2128 2129
	sector_t first_bad;
	int bad_sectors;
2130
	int replacement = 0;
L
Linus Torvalds 已提交
2131

2132 2133 2134
	for (i = 0 ; i < disks; i++) {
		if (bi == &sh->dev[i].req) {
			rdev = conf->disks[i].rdev;
L
Linus Torvalds 已提交
2135
			break;
2136 2137 2138
		}
		if (bi == &sh->dev[i].rreq) {
			rdev = conf->disks[i].replacement;
2139 2140 2141 2142 2143 2144 2145 2146
			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;
2147 2148 2149
			break;
		}
	}
2150
	pr_debug("end_write_request %llu/%d, count %d, uptodate: %d.\n",
L
Linus Torvalds 已提交
2151 2152 2153 2154
		(unsigned long long)sh->sector, i, atomic_read(&sh->count),
		uptodate);
	if (i == disks) {
		BUG();
2155
		return;
L
Linus Torvalds 已提交
2156 2157
	}

2158 2159 2160 2161 2162 2163 2164 2165 2166
	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) {
2167
			set_bit(STRIPE_DEGRADED, &sh->state);
2168 2169
			set_bit(WriteErrorSeen, &rdev->flags);
			set_bit(R5_WriteError, &sh->dev[i].flags);
2170 2171 2172
			if (!test_and_set_bit(WantReplacement, &rdev->flags))
				set_bit(MD_RECOVERY_NEEDED,
					&rdev->mddev->recovery);
2173 2174
		} else if (is_badblock(rdev, sh->sector,
				       STRIPE_SECTORS,
2175
				       &first_bad, &bad_sectors)) {
2176
			set_bit(R5_MadeGood, &sh->dev[i].flags);
2177 2178 2179 2180 2181 2182 2183
			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);
		}
2184 2185
	}
	rdev_dec_pending(rdev, conf->mddev);
L
Linus Torvalds 已提交
2186

2187 2188
	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
2189
	set_bit(STRIPE_HANDLE, &sh->state);
2190
	release_stripe(sh);
L
Linus Torvalds 已提交
2191 2192
}

2193
static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous);
2194

2195
static void raid5_build_block(struct stripe_head *sh, int i, int previous)
L
Linus Torvalds 已提交
2196 2197 2198 2199 2200
{
	struct r5dev *dev = &sh->dev[i];

	bio_init(&dev->req);
	dev->req.bi_io_vec = &dev->vec;
2201
	dev->req.bi_max_vecs = 1;
L
Linus Torvalds 已提交
2202 2203
	dev->req.bi_private = sh;

2204 2205
	bio_init(&dev->rreq);
	dev->rreq.bi_io_vec = &dev->rvec;
2206
	dev->rreq.bi_max_vecs = 1;
2207 2208
	dev->rreq.bi_private = sh;

L
Linus Torvalds 已提交
2209
	dev->flags = 0;
2210
	dev->sector = compute_blocknr(sh, i, previous);
L
Linus Torvalds 已提交
2211 2212
}

2213
static void error(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
2214 2215
{
	char b[BDEVNAME_SIZE];
2216
	struct r5conf *conf = mddev->private;
2217
	unsigned long flags;
2218
	pr_debug("raid456: error called\n");
L
Linus Torvalds 已提交
2219

2220 2221 2222 2223 2224 2225
	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);

2226
	set_bit(Blocked, &rdev->flags);
2227 2228 2229 2230 2231 2232 2233 2234 2235
	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);
2236
}
L
Linus Torvalds 已提交
2237 2238 2239 2240 2241

/*
 * Input: a 'big' sector number,
 * Output: index of the data and parity disk, and the sector # in them.
 */
2242
static sector_t raid5_compute_sector(struct r5conf *conf, sector_t r_sector,
2243 2244
				     int previous, int *dd_idx,
				     struct stripe_head *sh)
L
Linus Torvalds 已提交
2245
{
N
NeilBrown 已提交
2246
	sector_t stripe, stripe2;
2247
	sector_t chunk_number;
L
Linus Torvalds 已提交
2248
	unsigned int chunk_offset;
2249
	int pd_idx, qd_idx;
2250
	int ddf_layout = 0;
L
Linus Torvalds 已提交
2251
	sector_t new_sector;
2252 2253
	int algorithm = previous ? conf->prev_algo
				 : conf->algorithm;
2254 2255
	int sectors_per_chunk = previous ? conf->prev_chunk_sectors
					 : conf->chunk_sectors;
2256 2257 2258
	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
2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270

	/* 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
	 */
2271 2272
	stripe = chunk_number;
	*dd_idx = sector_div(stripe, data_disks);
N
NeilBrown 已提交
2273
	stripe2 = stripe;
L
Linus Torvalds 已提交
2274 2275 2276
	/*
	 * Select the parity disk based on the user selected algorithm.
	 */
2277
	pd_idx = qd_idx = -1;
2278 2279
	switch(conf->level) {
	case 4:
2280
		pd_idx = data_disks;
2281 2282
		break;
	case 5:
2283
		switch (algorithm) {
L
Linus Torvalds 已提交
2284
		case ALGORITHM_LEFT_ASYMMETRIC:
N
NeilBrown 已提交
2285
			pd_idx = data_disks - sector_div(stripe2, raid_disks);
2286
			if (*dd_idx >= pd_idx)
L
Linus Torvalds 已提交
2287 2288 2289
				(*dd_idx)++;
			break;
		case ALGORITHM_RIGHT_ASYMMETRIC:
N
NeilBrown 已提交
2290
			pd_idx = sector_div(stripe2, raid_disks);
2291
			if (*dd_idx >= pd_idx)
L
Linus Torvalds 已提交
2292 2293 2294
				(*dd_idx)++;
			break;
		case ALGORITHM_LEFT_SYMMETRIC:
N
NeilBrown 已提交
2295
			pd_idx = data_disks - sector_div(stripe2, raid_disks);
2296
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
L
Linus Torvalds 已提交
2297 2298
			break;
		case ALGORITHM_RIGHT_SYMMETRIC:
N
NeilBrown 已提交
2299
			pd_idx = sector_div(stripe2, raid_disks);
2300
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
L
Linus Torvalds 已提交
2301
			break;
2302 2303 2304 2305 2306 2307 2308
		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
2309
		default:
2310
			BUG();
2311 2312 2313 2314
		}
		break;
	case 6:

2315
		switch (algorithm) {
2316
		case ALGORITHM_LEFT_ASYMMETRIC:
N
NeilBrown 已提交
2317
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2318 2319
			qd_idx = pd_idx + 1;
			if (pd_idx == raid_disks-1) {
2320
				(*dd_idx)++;	/* Q D D D P */
2321 2322
				qd_idx = 0;
			} else if (*dd_idx >= pd_idx)
2323 2324 2325
				(*dd_idx) += 2; /* D D P Q D */
			break;
		case ALGORITHM_RIGHT_ASYMMETRIC:
N
NeilBrown 已提交
2326
			pd_idx = sector_div(stripe2, raid_disks);
2327 2328
			qd_idx = pd_idx + 1;
			if (pd_idx == raid_disks-1) {
2329
				(*dd_idx)++;	/* Q D D D P */
2330 2331
				qd_idx = 0;
			} else if (*dd_idx >= pd_idx)
2332 2333 2334
				(*dd_idx) += 2; /* D D P Q D */
			break;
		case ALGORITHM_LEFT_SYMMETRIC:
N
NeilBrown 已提交
2335
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2336 2337
			qd_idx = (pd_idx + 1) % raid_disks;
			*dd_idx = (pd_idx + 2 + *dd_idx) % raid_disks;
2338 2339
			break;
		case ALGORITHM_RIGHT_SYMMETRIC:
N
NeilBrown 已提交
2340
			pd_idx = sector_div(stripe2, raid_disks);
2341 2342
			qd_idx = (pd_idx + 1) % raid_disks;
			*dd_idx = (pd_idx + 2 + *dd_idx) % raid_disks;
2343
			break;
2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358

		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
2359
			pd_idx = sector_div(stripe2, raid_disks);
2360 2361 2362 2363 2364 2365
			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 */
2366
			ddf_layout = 1;
2367 2368 2369 2370 2371 2372 2373
			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
2374 2375
			stripe2 += 1;
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2376 2377 2378 2379 2380 2381
			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 */
2382
			ddf_layout = 1;
2383 2384 2385 2386
			break;

		case ALGORITHM_ROTATING_N_CONTINUE:
			/* Same as left_symmetric but Q is before P */
N
NeilBrown 已提交
2387
			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
2388 2389
			qd_idx = (pd_idx + raid_disks - 1) % raid_disks;
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
2390
			ddf_layout = 1;
2391 2392 2393 2394
			break;

		case ALGORITHM_LEFT_ASYMMETRIC_6:
			/* RAID5 left_asymmetric, with Q on last device */
N
NeilBrown 已提交
2395
			pd_idx = data_disks - sector_div(stripe2, raid_disks-1);
2396 2397 2398 2399 2400 2401
			if (*dd_idx >= pd_idx)
				(*dd_idx)++;
			qd_idx = raid_disks - 1;
			break;

		case ALGORITHM_RIGHT_ASYMMETRIC_6:
N
NeilBrown 已提交
2402
			pd_idx = sector_div(stripe2, raid_disks-1);
2403 2404 2405 2406 2407 2408
			if (*dd_idx >= pd_idx)
				(*dd_idx)++;
			qd_idx = raid_disks - 1;
			break;

		case ALGORITHM_LEFT_SYMMETRIC_6:
N
NeilBrown 已提交
2409
			pd_idx = data_disks - sector_div(stripe2, raid_disks-1);
2410 2411 2412 2413 2414
			*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
2415
			pd_idx = sector_div(stripe2, raid_disks-1);
2416 2417 2418 2419 2420 2421 2422 2423 2424 2425
			*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;

2426
		default:
2427
			BUG();
2428 2429
		}
		break;
L
Linus Torvalds 已提交
2430 2431
	}

2432 2433 2434
	if (sh) {
		sh->pd_idx = pd_idx;
		sh->qd_idx = qd_idx;
2435
		sh->ddf_layout = ddf_layout;
2436
	}
L
Linus Torvalds 已提交
2437 2438 2439 2440 2441 2442 2443
	/*
	 * Finally, compute the new sector number
	 */
	new_sector = (sector_t)stripe * sectors_per_chunk + chunk_offset;
	return new_sector;
}

2444
static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous)
L
Linus Torvalds 已提交
2445
{
2446
	struct r5conf *conf = sh->raid_conf;
2447 2448
	int raid_disks = sh->disks;
	int data_disks = raid_disks - conf->max_degraded;
L
Linus Torvalds 已提交
2449
	sector_t new_sector = sh->sector, check;
2450 2451
	int sectors_per_chunk = previous ? conf->prev_chunk_sectors
					 : conf->chunk_sectors;
2452 2453
	int algorithm = previous ? conf->prev_algo
				 : conf->algorithm;
L
Linus Torvalds 已提交
2454 2455
	sector_t stripe;
	int chunk_offset;
2456 2457
	sector_t chunk_number;
	int dummy1, dd_idx = i;
L
Linus Torvalds 已提交
2458
	sector_t r_sector;
2459
	struct stripe_head sh2;
L
Linus Torvalds 已提交
2460 2461 2462 2463

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

2464 2465 2466 2467 2468
	if (i == sh->pd_idx)
		return 0;
	switch(conf->level) {
	case 4: break;
	case 5:
2469
		switch (algorithm) {
L
Linus Torvalds 已提交
2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480
		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;
2481 2482 2483 2484 2485
		case ALGORITHM_PARITY_0:
			i -= 1;
			break;
		case ALGORITHM_PARITY_N:
			break;
L
Linus Torvalds 已提交
2486
		default:
2487
			BUG();
2488 2489 2490
		}
		break;
	case 6:
2491
		if (i == sh->qd_idx)
2492
			return 0; /* It is the Q disk */
2493
		switch (algorithm) {
2494 2495
		case ALGORITHM_LEFT_ASYMMETRIC:
		case ALGORITHM_RIGHT_ASYMMETRIC:
2496 2497 2498 2499
		case ALGORITHM_ROTATING_ZERO_RESTART:
		case ALGORITHM_ROTATING_N_RESTART:
			if (sh->pd_idx == raid_disks-1)
				i--;	/* Q D D D P */
2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513
			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;
2514 2515 2516 2517 2518 2519
		case ALGORITHM_PARITY_0:
			i -= 2;
			break;
		case ALGORITHM_PARITY_N:
			break;
		case ALGORITHM_ROTATING_N_CONTINUE:
2520
			/* Like left_symmetric, but P is before Q */
2521 2522
			if (sh->pd_idx == 0)
				i--;	/* P D D D Q */
2523 2524 2525 2526 2527 2528
			else {
				/* D D Q P D */
				if (i < sh->pd_idx)
					i += raid_disks;
				i -= (sh->pd_idx + 1);
			}
2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543
			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;
2544
		default:
2545
			BUG();
2546 2547
		}
		break;
L
Linus Torvalds 已提交
2548 2549 2550
	}

	chunk_number = stripe * data_disks + i;
2551
	r_sector = chunk_number * sectors_per_chunk + chunk_offset;
L
Linus Torvalds 已提交
2552

2553
	check = raid5_compute_sector(conf, r_sector,
2554
				     previous, &dummy1, &sh2);
2555 2556
	if (check != sh->sector || dummy1 != dd_idx || sh2.pd_idx != sh->pd_idx
		|| sh2.qd_idx != sh->qd_idx) {
2557 2558
		printk(KERN_ERR "md/raid:%s: compute_blocknr: map not correct\n",
		       mdname(conf->mddev));
L
Linus Torvalds 已提交
2559 2560 2561 2562 2563
		return 0;
	}
	return r_sector;
}

2564
static void
2565
schedule_reconstruction(struct stripe_head *sh, struct stripe_head_state *s,
2566
			 int rcw, int expand)
2567 2568
{
	int i, pd_idx = sh->pd_idx, disks = sh->disks;
2569
	struct r5conf *conf = sh->raid_conf;
2570
	int level = conf->level;
2571 2572 2573 2574 2575 2576 2577 2578

	if (rcw) {

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

			if (dev->towrite) {
				set_bit(R5_LOCKED, &dev->flags);
2579
				set_bit(R5_Wantdrain, &dev->flags);
2580 2581
				if (!expand)
					clear_bit(R5_UPTODATE, &dev->flags);
2582
				s->locked++;
2583 2584
			}
		}
2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599
		/* 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);

2600
		if (s->locked + conf->max_degraded == disks)
2601
			if (!test_and_set_bit(STRIPE_FULL_WRITE, &sh->state))
2602
				atomic_inc(&conf->pending_full_writes);
2603
	} else {
2604
		BUG_ON(level == 6);
2605 2606 2607 2608 2609 2610 2611 2612 2613 2614
		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) ||
2615 2616
			     test_bit(R5_Wantcompute, &dev->flags))) {
				set_bit(R5_Wantdrain, &dev->flags);
2617 2618
				set_bit(R5_LOCKED, &dev->flags);
				clear_bit(R5_UPTODATE, &dev->flags);
2619
				s->locked++;
2620 2621
			}
		}
2622 2623 2624 2625 2626 2627 2628
		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);
2629 2630
	}

2631
	/* keep the parity disk(s) locked while asynchronous operations
2632 2633 2634 2635
	 * are in flight
	 */
	set_bit(R5_LOCKED, &sh->dev[pd_idx].flags);
	clear_bit(R5_UPTODATE, &sh->dev[pd_idx].flags);
2636
	s->locked++;
2637

2638 2639 2640 2641 2642 2643 2644 2645 2646
	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++;
	}

2647
	pr_debug("%s: stripe %llu locked: %d ops_request: %lx\n",
2648
		__func__, (unsigned long long)sh->sector,
2649
		s->locked, s->ops_request);
2650
}
2651

L
Linus Torvalds 已提交
2652 2653
/*
 * Each stripe/dev can have one or more bion attached.
2654
 * toread/towrite point to the first in a chain.
L
Linus Torvalds 已提交
2655 2656
 * The bi_next chain must be in order.
 */
2657 2658
static int add_stripe_bio(struct stripe_head *sh, struct bio *bi, int dd_idx,
			  int forwrite, int previous)
L
Linus Torvalds 已提交
2659 2660
{
	struct bio **bip;
2661
	struct r5conf *conf = sh->raid_conf;
2662
	int firstwrite=0;
L
Linus Torvalds 已提交
2663

2664
	pr_debug("adding bi b#%llu to stripe s#%llu\n",
2665
		(unsigned long long)bi->bi_iter.bi_sector,
L
Linus Torvalds 已提交
2666 2667
		(unsigned long long)sh->sector);

S
Shaohua Li 已提交
2668 2669 2670 2671 2672 2673 2674 2675 2676
	/*
	 * 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);
2677
	if (forwrite) {
L
Linus Torvalds 已提交
2678
		bip = &sh->dev[dd_idx].towrite;
2679
		if (*bip == NULL)
2680 2681
			firstwrite = 1;
	} else
L
Linus Torvalds 已提交
2682
		bip = &sh->dev[dd_idx].toread;
2683 2684
	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
2685 2686 2687
			goto overlap;
		bip = & (*bip)->bi_next;
	}
2688
	if (*bip && (*bip)->bi_iter.bi_sector < bio_end_sector(bi))
L
Linus Torvalds 已提交
2689 2690
		goto overlap;

2691 2692 2693
	if (!forwrite || previous)
		clear_bit(STRIPE_BATCH_READY, &sh->state);

2694
	BUG_ON(*bip && bi->bi_next && (*bip) != bi->bi_next);
L
Linus Torvalds 已提交
2695 2696 2697
	if (*bip)
		bi->bi_next = *bip;
	*bip = bi;
2698
	raid5_inc_bi_active_stripes(bi);
2699

L
Linus Torvalds 已提交
2700 2701 2702 2703 2704
	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 &&
2705
			     bi && bi->bi_iter.bi_sector <= sector;
L
Linus Torvalds 已提交
2706
		     bi = r5_next_bio(bi, sh->dev[dd_idx].sector)) {
K
Kent Overstreet 已提交
2707 2708
			if (bio_end_sector(bi) >= sector)
				sector = bio_end_sector(bi);
L
Linus Torvalds 已提交
2709 2710
		}
		if (sector >= sh->dev[dd_idx].sector + STRIPE_SECTORS)
2711 2712
			if (!test_and_set_bit(R5_OVERWRITE, &sh->dev[dd_idx].flags))
				sh->overwrite_disks++;
L
Linus Torvalds 已提交
2713
	}
2714 2715

	pr_debug("added bi b#%llu to stripe s#%llu, disk %d.\n",
2716
		(unsigned long long)(*bip)->bi_iter.bi_sector,
2717
		(unsigned long long)sh->sector, dd_idx);
2718
	spin_unlock_irq(&sh->stripe_lock);
2719 2720 2721 2722 2723 2724 2725

	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
2726 2727 2728 2729
	return 1;

 overlap:
	set_bit(R5_Overlap, &sh->dev[dd_idx].flags);
S
Shaohua Li 已提交
2730
	spin_unlock_irq(&sh->stripe_lock);
L
Linus Torvalds 已提交
2731 2732 2733
	return 0;
}

2734
static void end_reshape(struct r5conf *conf);
2735

2736
static void stripe_set_idx(sector_t stripe, struct r5conf *conf, int previous,
2737
			    struct stripe_head *sh)
2738
{
2739
	int sectors_per_chunk =
2740
		previous ? conf->prev_chunk_sectors : conf->chunk_sectors;
2741
	int dd_idx;
2742
	int chunk_offset = sector_div(stripe, sectors_per_chunk);
2743
	int disks = previous ? conf->previous_raid_disks : conf->raid_disks;
2744

2745 2746
	raid5_compute_sector(conf,
			     stripe * (disks - conf->max_degraded)
2747
			     *sectors_per_chunk + chunk_offset,
2748
			     previous,
2749
			     &dd_idx, sh);
2750 2751
}

2752
static void
2753
handle_failed_stripe(struct r5conf *conf, struct stripe_head *sh,
2754 2755 2756 2757 2758 2759 2760 2761 2762
				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)) {
2763
			struct md_rdev *rdev;
2764 2765 2766
			rcu_read_lock();
			rdev = rcu_dereference(conf->disks[i].rdev);
			if (rdev && test_bit(In_sync, &rdev->flags))
2767 2768 2769
				atomic_inc(&rdev->nr_pending);
			else
				rdev = NULL;
2770
			rcu_read_unlock();
2771 2772 2773 2774 2775 2776 2777 2778
			if (rdev) {
				if (!rdev_set_badblocks(
					    rdev,
					    sh->sector,
					    STRIPE_SECTORS, 0))
					md_error(conf->mddev, rdev);
				rdev_dec_pending(rdev, conf->mddev);
			}
2779
		}
S
Shaohua Li 已提交
2780
		spin_lock_irq(&sh->stripe_lock);
2781 2782 2783
		/* fail all writes first */
		bi = sh->dev[i].towrite;
		sh->dev[i].towrite = NULL;
2784
		sh->overwrite_disks = 0;
S
Shaohua Li 已提交
2785
		spin_unlock_irq(&sh->stripe_lock);
2786
		if (bi)
2787 2788 2789 2790 2791
			bitmap_end = 1;

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

2792
		while (bi && bi->bi_iter.bi_sector <
2793 2794 2795
			sh->dev[i].sector + STRIPE_SECTORS) {
			struct bio *nextbi = r5_next_bio(bi, sh->dev[i].sector);
			clear_bit(BIO_UPTODATE, &bi->bi_flags);
2796
			if (!raid5_dec_bi_active_stripes(bi)) {
2797 2798 2799 2800 2801 2802
				md_write_end(conf->mddev);
				bi->bi_next = *return_bi;
				*return_bi = bi;
			}
			bi = nextbi;
		}
2803 2804 2805 2806
		if (bitmap_end)
			bitmap_endwrite(conf->mddev->bitmap, sh->sector,
				STRIPE_SECTORS, 0, 0);
		bitmap_end = 0;
2807 2808 2809
		/* and fail all 'written' */
		bi = sh->dev[i].written;
		sh->dev[i].written = NULL;
2810 2811 2812 2813 2814
		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;
		}

2815
		if (bi) bitmap_end = 1;
2816
		while (bi && bi->bi_iter.bi_sector <
2817 2818 2819
		       sh->dev[i].sector + STRIPE_SECTORS) {
			struct bio *bi2 = r5_next_bio(bi, sh->dev[i].sector);
			clear_bit(BIO_UPTODATE, &bi->bi_flags);
2820
			if (!raid5_dec_bi_active_stripes(bi)) {
2821 2822 2823 2824 2825 2826 2827
				md_write_end(conf->mddev);
				bi->bi_next = *return_bi;
				*return_bi = bi;
			}
			bi = bi2;
		}

2828 2829 2830 2831 2832 2833
		/* 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))) {
2834
			spin_lock_irq(&sh->stripe_lock);
2835 2836
			bi = sh->dev[i].toread;
			sh->dev[i].toread = NULL;
2837
			spin_unlock_irq(&sh->stripe_lock);
2838 2839
			if (test_and_clear_bit(R5_Overlap, &sh->dev[i].flags))
				wake_up(&conf->wait_for_overlap);
2840
			while (bi && bi->bi_iter.bi_sector <
2841 2842 2843 2844
			       sh->dev[i].sector + STRIPE_SECTORS) {
				struct bio *nextbi =
					r5_next_bio(bi, sh->dev[i].sector);
				clear_bit(BIO_UPTODATE, &bi->bi_flags);
2845
				if (!raid5_dec_bi_active_stripes(bi)) {
2846 2847 2848 2849 2850 2851 2852 2853 2854
					bi->bi_next = *return_bi;
					*return_bi = bi;
				}
				bi = nextbi;
			}
		}
		if (bitmap_end)
			bitmap_endwrite(conf->mddev->bitmap, sh->sector,
					STRIPE_SECTORS, 0, 0);
2855 2856 2857 2858
		/* 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);
2859 2860
	}

2861 2862 2863
	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);
2864 2865
}

2866
static void
2867
handle_failed_sync(struct r5conf *conf, struct stripe_head *sh,
2868 2869 2870 2871 2872 2873
		   struct stripe_head_state *s)
{
	int abort = 0;
	int i;

	clear_bit(STRIPE_SYNCING, &sh->state);
2874 2875
	if (test_and_clear_bit(R5_Overlap, &sh->dev[sh->pd_idx].flags))
		wake_up(&conf->wait_for_overlap);
2876
	s->syncing = 0;
2877
	s->replacing = 0;
2878
	/* There is nothing more to do for sync/check/repair.
2879 2880 2881
	 * 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.
2882
	 * For recover/replace we need to record a bad block on all
2883 2884
	 * non-sync devices, or abort the recovery
	 */
2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907
	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;
2908
	}
2909
	md_done_sync(conf->mddev, STRIPE_SECTORS, !abort);
2910 2911
}

2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927
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;
}

2928
/* fetch_block - checks the given member device to see if its data needs
2929 2930 2931
 * to be read or computed to satisfy a request.
 *
 * Returns 1 when no more member devices need to be checked, otherwise returns
2932
 * 0 to tell the loop in handle_stripe_fill to continue
2933
 */
2934 2935 2936

static int need_this_block(struct stripe_head *sh, struct stripe_head_state *s,
			   int disk_idx, int disks)
2937
{
2938
	struct r5dev *dev = &sh->dev[disk_idx];
2939 2940
	struct r5dev *fdev[2] = { &sh->dev[s->failed_num[0]],
				  &sh->dev[s->failed_num[1]] };
2941
	int i;
2942

2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969

	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;

2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988
	/* 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;
2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019

	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;
	}

3020 3021 3022 3023 3024 3025 3026 3027 3028 3029
	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)) {
3030 3031 3032 3033 3034 3035
		/* 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) &&
3036 3037
		    (s->failed && (disk_idx == s->failed_num[0] ||
				   disk_idx == s->failed_num[1]))) {
3038 3039
			/* have disk failed, and we're requested to fetch it;
			 * do compute it
3040
			 */
3041 3042 3043 3044 3045 3046 3047 3048
			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;
3049 3050 3051 3052 3053 3054
			/* 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.
			 */
3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067
			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;
3068
			}
3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087
			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);
3088 3089
		}
	}
3090 3091 3092 3093 3094

	return 0;
}

/**
3095
 * handle_stripe_fill - read or compute data to satisfy pending requests.
3096
 */
3097 3098 3099
static void handle_stripe_fill(struct stripe_head *sh,
			       struct stripe_head_state *s,
			       int disks)
3100 3101 3102 3103 3104 3105 3106 3107 3108 3109
{
	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--; )
3110
			if (fetch_block(sh, s, i, disks))
3111
				break;
3112 3113 3114
	set_bit(STRIPE_HANDLE, &sh->state);
}

3115
/* handle_stripe_clean_event
3116 3117 3118 3119
 * 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.
 */
3120
static void handle_stripe_clean_event(struct r5conf *conf,
3121 3122 3123 3124
	struct stripe_head *sh, int disks, struct bio **return_bi)
{
	int i;
	struct r5dev *dev;
3125
	int discard_pending = 0;
3126 3127 3128 3129 3130

	for (i = disks; i--; )
		if (sh->dev[i].written) {
			dev = &sh->dev[i];
			if (!test_bit(R5_LOCKED, &dev->flags) &&
3131
			    (test_bit(R5_UPTODATE, &dev->flags) ||
3132 3133
			     test_bit(R5_Discard, &dev->flags) ||
			     test_bit(R5_SkipCopy, &dev->flags))) {
3134 3135
				/* We can return any write requests */
				struct bio *wbi, *wbi2;
3136
				pr_debug("Return write for disc %d\n", i);
3137 3138
				if (test_and_clear_bit(R5_Discard, &dev->flags))
					clear_bit(R5_UPTODATE, &dev->flags);
3139 3140 3141 3142
				if (test_and_clear_bit(R5_SkipCopy, &dev->flags)) {
					WARN_ON(test_bit(R5_UPTODATE, &dev->flags));
					dev->page = dev->orig_page;
				}
3143 3144
				wbi = dev->written;
				dev->written = NULL;
3145
				while (wbi && wbi->bi_iter.bi_sector <
3146 3147
					dev->sector + STRIPE_SECTORS) {
					wbi2 = r5_next_bio(wbi, dev->sector);
3148
					if (!raid5_dec_bi_active_stripes(wbi)) {
3149 3150 3151 3152 3153 3154
						md_write_end(conf->mddev);
						wbi->bi_next = *return_bi;
						*return_bi = wbi;
					}
					wbi = wbi2;
				}
3155 3156
				bitmap_endwrite(conf->mddev->bitmap, sh->sector,
						STRIPE_SECTORS,
3157
					 !test_bit(STRIPE_DEGRADED, &sh->state),
3158
						0);
3159 3160
			} else if (test_bit(R5_Discard, &dev->flags))
				discard_pending = 1;
3161 3162
			WARN_ON(test_bit(R5_SkipCopy, &dev->flags));
			WARN_ON(dev->page != dev->orig_page);
3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173
		}
	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
3174 3175 3176 3177 3178 3179 3180 3181
		/*
		 * 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);
3182 3183 3184 3185
		if (test_bit(STRIPE_SYNC_REQUESTED, &sh->state))
			set_bit(STRIPE_HANDLE, &sh->state);

	}
3186 3187 3188 3189

	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);
3190 3191
}

3192
static void handle_stripe_dirtying(struct r5conf *conf,
3193 3194 3195
				   struct stripe_head *sh,
				   struct stripe_head_state *s,
				   int disks)
3196 3197
{
	int rmw = 0, rcw = 0, i;
3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208
	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 ||
3209 3210
	    (recovery_cp < MaxSector && sh->sector >= recovery_cp &&
	     s->failed == 0)) {
3211
		/* Calculate the real rcw later - for now make it
3212 3213 3214
		 * look like rcw is cheaper
		 */
		rcw = 1; rmw = 2;
3215 3216 3217
		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);
3218
	} else for (i = disks; i--; ) {
3219 3220 3221 3222
		/* 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) &&
3223 3224
		    !(test_bit(R5_UPTODATE, &dev->flags) ||
		      test_bit(R5_Wantcompute, &dev->flags))) {
3225 3226 3227 3228 3229 3230 3231 3232
			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) &&
3233 3234
		    !(test_bit(R5_UPTODATE, &dev->flags) ||
		    test_bit(R5_Wantcompute, &dev->flags))) {
3235 3236
			if (test_bit(R5_Insync, &dev->flags))
				rcw++;
3237 3238 3239 3240
			else
				rcw += 2*disks;
		}
	}
3241
	pr_debug("for sector %llu, rmw=%d rcw=%d\n",
3242 3243
		(unsigned long long)sh->sector, rmw, rcw);
	set_bit(STRIPE_HANDLE, &sh->state);
N
NeilBrown 已提交
3244
	if (rmw < rcw && rmw > 0) {
3245
		/* prefer read-modify-write, but need to get some data */
3246 3247 3248 3249
		if (conf->mddev->queue)
			blk_add_trace_msg(conf->mddev->queue,
					  "raid5 rmw %llu %d",
					  (unsigned long long)sh->sector, rmw);
3250 3251 3252 3253
		for (i = disks; i--; ) {
			struct r5dev *dev = &sh->dev[i];
			if ((dev->towrite || i == sh->pd_idx) &&
			    !test_bit(R5_LOCKED, &dev->flags) &&
3254 3255
			    !(test_bit(R5_UPTODATE, &dev->flags) ||
			    test_bit(R5_Wantcompute, &dev->flags)) &&
3256
			    test_bit(R5_Insync, &dev->flags)) {
3257 3258 3259 3260
				if (test_bit(STRIPE_PREREAD_ACTIVE,
					     &sh->state)) {
					pr_debug("Read_old block %d for r-m-w\n",
						 i);
3261 3262 3263 3264 3265 3266 3267 3268 3269
					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
3270
	}
3271
	if (rcw <= rmw && rcw > 0) {
3272
		/* want reconstruct write, but need to get some data */
N
NeilBrown 已提交
3273
		int qread =0;
3274
		rcw = 0;
3275 3276 3277
		for (i = disks; i--; ) {
			struct r5dev *dev = &sh->dev[i];
			if (!test_bit(R5_OVERWRITE, &dev->flags) &&
3278
			    i != sh->pd_idx && i != sh->qd_idx &&
3279
			    !test_bit(R5_LOCKED, &dev->flags) &&
3280
			    !(test_bit(R5_UPTODATE, &dev->flags) ||
3281 3282
			      test_bit(R5_Wantcompute, &dev->flags))) {
				rcw++;
3283 3284 3285
				if (test_bit(R5_Insync, &dev->flags) &&
				    test_bit(STRIPE_PREREAD_ACTIVE,
					     &sh->state)) {
3286
					pr_debug("Read_old block "
3287 3288 3289 3290
						"%d for Reconstruct\n", i);
					set_bit(R5_LOCKED, &dev->flags);
					set_bit(R5_Wantread, &dev->flags);
					s->locked++;
N
NeilBrown 已提交
3291
					qread++;
3292 3293 3294 3295 3296 3297
				} else {
					set_bit(STRIPE_DELAYED, &sh->state);
					set_bit(STRIPE_HANDLE, &sh->state);
				}
			}
		}
3298
		if (rcw && conf->mddev->queue)
N
NeilBrown 已提交
3299 3300 3301
			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));
3302
	}
3303 3304 3305 3306 3307

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

3308 3309 3310
	/* now if nothing is locked, and if we have enough data,
	 * we can start a write request
	 */
3311 3312
	/* 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
3313 3314
	 * subsequent call wants to start a write request.  raid_run_ops only
	 * handles the case where compute block and reconstruct are requested
3315 3316 3317
	 * simultaneously.  If this is not the case then new writes need to be
	 * held off until the compute completes.
	 */
3318 3319 3320
	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)))
3321
		schedule_reconstruction(sh, s, rcw == 0, 0);
3322 3323
}

3324
static void handle_parity_checks5(struct r5conf *conf, struct stripe_head *sh,
3325 3326
				struct stripe_head_state *s, int disks)
{
3327
	struct r5dev *dev = NULL;
3328

3329
	set_bit(STRIPE_HANDLE, &sh->state);
3330

3331 3332 3333
	switch (sh->check_state) {
	case check_state_idle:
		/* start a new check operation if there are no failures */
3334 3335
		if (s->failed == 0) {
			BUG_ON(s->uptodate != disks);
3336 3337
			sh->check_state = check_state_run;
			set_bit(STRIPE_OP_CHECK, &s->ops_request);
3338 3339
			clear_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags);
			s->uptodate--;
3340
			break;
3341
		}
3342
		dev = &sh->dev[s->failed_num[0]];
3343 3344 3345 3346 3347 3348 3349 3350 3351
		/* 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
3352

3353 3354 3355 3356 3357
		/* 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);
3358
		s->locked++;
3359
		set_bit(R5_Wantwrite, &dev->flags);
3360

3361 3362
		clear_bit(STRIPE_DEGRADED, &sh->state);
		set_bit(STRIPE_INSYNC, &sh->state);
3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378
		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
3379
		if ((sh->ops.zero_sum_result & SUM_CHECK_P_RESULT) == 0)
3380 3381 3382 3383 3384
			/* parity is correct (on disc,
			 * not in buffer any more)
			 */
			set_bit(STRIPE_INSYNC, &sh->state);
		else {
3385
			atomic64_add(STRIPE_SECTORS, &conf->mddev->resync_mismatches);
3386 3387 3388 3389 3390
			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;
3391
				set_bit(STRIPE_COMPUTE_RUN, &sh->state);
3392 3393 3394 3395
				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;
3396
				sh->ops.target2 = -1;
3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407
				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();
3408 3409 3410
	}
}

3411
static void handle_parity_checks6(struct r5conf *conf, struct stripe_head *sh,
3412
				  struct stripe_head_state *s,
3413
				  int disks)
3414 3415
{
	int pd_idx = sh->pd_idx;
N
NeilBrown 已提交
3416
	int qd_idx = sh->qd_idx;
3417
	struct r5dev *dev;
3418 3419 3420 3421

	set_bit(STRIPE_HANDLE, &sh->state);

	BUG_ON(s->failed > 2);
3422

3423 3424 3425 3426 3427 3428
	/* 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
	 */

3429 3430 3431
	switch (sh->check_state) {
	case check_state_idle:
		/* start a new check operation if there are < 2 failures */
3432
		if (s->failed == s->q_failed) {
3433
			/* The only possible failed device holds Q, so it
3434 3435 3436
			 * makes sense to check P (If anything else were failed,
			 * we would have used P to recreate it).
			 */
3437
			sh->check_state = check_state_run;
3438
		}
3439
		if (!s->q_failed && s->failed < 2) {
3440
			/* Q is not failed, and we didn't use it to generate
3441 3442
			 * anything, so it makes sense to check it
			 */
3443 3444 3445 3446
			if (sh->check_state == check_state_run)
				sh->check_state = check_state_run_pq;
			else
				sh->check_state = check_state_run_q;
3447 3448
		}

3449 3450
		/* discard potentially stale zero_sum_result */
		sh->ops.zero_sum_result = 0;
3451

3452 3453 3454 3455
		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--;
3456
		}
3457 3458 3459 3460 3461 3462 3463
		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;
3464 3465
		}

3466 3467 3468 3469 3470
		/* we have 2-disk failure */
		BUG_ON(s->failed != 2);
		/* fall through */
	case check_state_compute_result:
		sh->check_state = check_state_idle;
3471

3472 3473 3474
		/* check that a write has not made the stripe insync */
		if (test_bit(STRIPE_INSYNC, &sh->state))
			break;
3475 3476

		/* now write out any block on a failed drive,
3477
		 * or P or Q if they were recomputed
3478
		 */
3479
		BUG_ON(s->uptodate < disks - 1); /* We don't need Q to recover */
3480
		if (s->failed == 2) {
3481
			dev = &sh->dev[s->failed_num[1]];
3482 3483 3484 3485 3486
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
		if (s->failed >= 1) {
3487
			dev = &sh->dev[s->failed_num[0]];
3488 3489 3490 3491
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
3492
		if (sh->ops.zero_sum_result & SUM_CHECK_P_RESULT) {
3493 3494 3495 3496 3497
			dev = &sh->dev[pd_idx];
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
3498
		if (sh->ops.zero_sum_result & SUM_CHECK_Q_RESULT) {
3499 3500 3501 3502 3503 3504 3505 3506
			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);
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 3535
		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 {
3536
			atomic64_add(STRIPE_SECTORS, &conf->mddev->resync_mismatches);
3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570
			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();
3571 3572 3573
	}
}

3574
static void handle_stripe_expansion(struct r5conf *conf, struct stripe_head *sh)
3575 3576 3577 3578 3579 3580
{
	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.
	 */
3581
	struct dma_async_tx_descriptor *tx = NULL;
3582 3583
	clear_bit(STRIPE_EXPAND_SOURCE, &sh->state);
	for (i = 0; i < sh->disks; i++)
N
NeilBrown 已提交
3584
		if (i != sh->pd_idx && i != sh->qd_idx) {
3585
			int dd_idx, j;
3586
			struct stripe_head *sh2;
3587
			struct async_submit_ctl submit;
3588

3589
			sector_t bn = compute_blocknr(sh, i, 1);
3590 3591
			sector_t s = raid5_compute_sector(conf, bn, 0,
							  &dd_idx, NULL);
3592
			sh2 = get_active_stripe(conf, s, 0, 1, 1);
3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604
			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;
			}
3605 3606

			/* place all the copies on one channel */
3607
			init_async_submit(&submit, 0, tx, NULL, NULL, NULL);
3608
			tx = async_memcpy(sh2->dev[dd_idx].page,
3609
					  sh->dev[i].page, 0, 0, STRIPE_SIZE,
3610
					  &submit);
3611

3612 3613 3614 3615
			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 &&
3616
				    j != sh2->qd_idx &&
3617 3618 3619 3620 3621 3622 3623
				    !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);
3624

3625
		}
3626
	/* done submitting copies, wait for them to complete */
3627
	async_tx_quiesce(&tx);
3628
}
L
Linus Torvalds 已提交
3629 3630 3631 3632

/*
 * handle_stripe - do things to a stripe.
 *
3633 3634
 * 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
3635
 * Possible results:
3636 3637
 *    return some read requests which now have data
 *    return some write requests which are safely on storage
L
Linus Torvalds 已提交
3638 3639 3640 3641 3642
 *    schedule a read on some buffers
 *    schedule a write of some buffers
 *    return confirmation of parity correctness
 *
 */
3643

3644
static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
L
Linus Torvalds 已提交
3645
{
3646
	struct r5conf *conf = sh->raid_conf;
3647
	int disks = sh->disks;
3648 3649
	struct r5dev *dev;
	int i;
3650
	int do_recovery = 0;
L
Linus Torvalds 已提交
3651

3652 3653 3654 3655 3656 3657
	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
3658

3659
	/* Now to look around and see what can be done */
L
Linus Torvalds 已提交
3660
	rcu_read_lock();
3661
	for (i=disks; i--; ) {
3662
		struct md_rdev *rdev;
3663 3664 3665
		sector_t first_bad;
		int bad_sectors;
		int is_bad = 0;
3666

3667
		dev = &sh->dev[i];
L
Linus Torvalds 已提交
3668

3669
		pr_debug("check %d: state 0x%lx read %p write %p written %p\n",
3670 3671
			 i, dev->flags,
			 dev->toread, dev->towrite, dev->written);
3672 3673 3674 3675 3676 3677 3678 3679
		/* 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
3680

3681
		/* now count some things */
3682 3683 3684 3685
		if (test_bit(R5_LOCKED, &dev->flags))
			s->locked++;
		if (test_bit(R5_UPTODATE, &dev->flags))
			s->uptodate++;
3686
		if (test_bit(R5_Wantcompute, &dev->flags)) {
3687 3688
			s->compute++;
			BUG_ON(s->compute > 2);
3689
		}
L
Linus Torvalds 已提交
3690

3691
		if (test_bit(R5_Wantfill, &dev->flags))
3692
			s->to_fill++;
3693
		else if (dev->toread)
3694
			s->to_read++;
3695
		if (dev->towrite) {
3696
			s->to_write++;
3697
			if (!test_bit(R5_OVERWRITE, &dev->flags))
3698
				s->non_overwrite++;
3699
		}
3700
		if (dev->written)
3701
			s->written++;
3702 3703 3704 3705 3706 3707 3708 3709 3710 3711
		/* 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 {
3712 3713
			if (rdev)
				set_bit(R5_NeedReplace, &dev->flags);
3714 3715 3716
			rdev = rcu_dereference(conf->disks[i].rdev);
			clear_bit(R5_ReadRepl, &dev->flags);
		}
3717 3718
		if (rdev && test_bit(Faulty, &rdev->flags))
			rdev = NULL;
3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730
		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);
			}
3731
		}
3732 3733 3734
		clear_bit(R5_Insync, &dev->flags);
		if (!rdev)
			/* Not in-sync */;
3735 3736
		else if (is_bad) {
			/* also not in-sync */
3737 3738
			if (!test_bit(WriteErrorSeen, &rdev->flags) &&
			    test_bit(R5_UPTODATE, &dev->flags)) {
3739 3740 3741 3742 3743 3744 3745
				/* 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))
3746
			set_bit(R5_Insync, &dev->flags);
3747
		else if (sh->sector + STRIPE_SECTORS <= rdev->recovery_offset)
3748
			/* in sync if before recovery_offset */
3749 3750 3751 3752 3753 3754 3755 3756 3757
			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);

3758
		if (test_bit(R5_WriteError, &dev->flags)) {
3759 3760 3761 3762 3763 3764 3765
			/* 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)) {
3766
				s->handle_bad_blocks = 1;
3767
				atomic_inc(&rdev2->nr_pending);
3768 3769 3770
			} else
				clear_bit(R5_WriteError, &dev->flags);
		}
3771
		if (test_bit(R5_MadeGood, &dev->flags)) {
3772 3773 3774 3775 3776
			/* 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)) {
3777
				s->handle_bad_blocks = 1;
3778
				atomic_inc(&rdev2->nr_pending);
3779 3780 3781
			} else
				clear_bit(R5_MadeGood, &dev->flags);
		}
3782 3783 3784 3785 3786 3787 3788 3789 3790
		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);
		}
3791
		if (!test_bit(R5_Insync, &dev->flags)) {
3792 3793 3794
			/* 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
3795
		}
3796 3797 3798
		if (test_bit(R5_ReadError, &dev->flags))
			clear_bit(R5_Insync, &dev->flags);
		if (!test_bit(R5_Insync, &dev->flags)) {
3799 3800 3801
			if (s->failed < 2)
				s->failed_num[s->failed] = i;
			s->failed++;
3802 3803
			if (rdev && !test_bit(Faulty, &rdev->flags))
				do_recovery = 1;
3804
		}
L
Linus Torvalds 已提交
3805
	}
3806 3807 3808 3809
	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
3810
		 * else if MD_RECOVERY_REQUESTED is set, we also are syncing.
3811 3812 3813 3814 3815
		 * else we can only be replacing
		 * sync and recovery both need to read all devices, and so
		 * use the same flag.
		 */
		if (do_recovery ||
3816 3817
		    sh->sector >= conf->mddev->recovery_cp ||
		    test_bit(MD_RECOVERY_REQUESTED, &(conf->mddev->recovery)))
3818 3819 3820 3821
			s->syncing = 1;
		else
			s->replacing = 1;
	}
L
Linus Torvalds 已提交
3822
	rcu_read_unlock();
3823 3824 3825 3826 3827
}

static void handle_stripe(struct stripe_head *sh)
{
	struct stripe_head_state s;
3828
	struct r5conf *conf = sh->raid_conf;
3829
	int i;
3830 3831
	int prexor;
	int disks = sh->disks;
3832
	struct r5dev *pdev, *qdev;
3833 3834

	clear_bit(STRIPE_HANDLE, &sh->state);
3835
	if (test_and_set_bit_lock(STRIPE_ACTIVE, &sh->state)) {
3836 3837 3838 3839 3840 3841
		/* already being handled, ensure it gets handled
		 * again when current action finishes */
		set_bit(STRIPE_HANDLE, &sh->state);
		return;
	}

3842
	clear_bit(STRIPE_BATCH_READY, &sh->state);
3843 3844 3845 3846 3847 3848 3849
	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);
3850
			clear_bit(STRIPE_REPLACED, &sh->state);
3851 3852
		}
		spin_unlock(&sh->stripe_lock);
3853 3854 3855 3856 3857 3858 3859 3860
	}
	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);
3861

3862
	analyse_stripe(sh, &s);
3863

3864 3865 3866 3867 3868
	if (s.handle_bad_blocks) {
		set_bit(STRIPE_HANDLE, &sh->state);
		goto finish;
	}

3869 3870
	if (unlikely(s.blocked_rdev)) {
		if (s.syncing || s.expanding || s.expanded ||
3871
		    s.replacing || s.to_write || s.written) {
3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891
			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.
	 */
3892 3893 3894 3895 3896
	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);
3897
		if (s.syncing + s.replacing)
3898 3899
			handle_failed_sync(conf, sh, &s);
	}
3900

3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913
	/* 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
		 */
3914 3915
		BUG_ON(!test_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags) &&
		       !test_bit(R5_Discard, &sh->dev[sh->pd_idx].flags));
3916
		BUG_ON(sh->qd_idx >= 0 &&
3917 3918
		       !test_bit(R5_UPTODATE, &sh->dev[sh->qd_idx].flags) &&
		       !test_bit(R5_Discard, &sh->dev[sh->qd_idx].flags));
3919 3920 3921 3922 3923 3924 3925 3926 3927
		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;
3928 3929
				if (s.failed > 1)
					continue;
3930 3931 3932 3933 3934 3935 3936 3937 3938 3939
				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;
	}

3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973
	/*
	 * 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);

3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996
	/* 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);
	}
3997

3998 3999 4000
	if ((s.replacing || s.syncing) && s.locked == 0
	    && !test_bit(STRIPE_COMPUTE_RUN, &sh->state)
	    && !test_bit(STRIPE_REPLACED, &sh->state)) {
4001 4002
		/* Write out to replacement devices where possible */
		for (i = 0; i < conf->raid_disks; i++)
4003 4004
			if (test_bit(R5_NeedReplace, &sh->dev[i].flags)) {
				WARN_ON(!test_bit(R5_UPTODATE, &sh->dev[i].flags));
4005 4006 4007 4008
				set_bit(R5_WantReplace, &sh->dev[i].flags);
				set_bit(R5_LOCKED, &sh->dev[i].flags);
				s.locked++;
			}
4009 4010 4011
		if (s.replacing)
			set_bit(STRIPE_INSYNC, &sh->state);
		set_bit(STRIPE_REPLACED, &sh->state);
4012 4013
	}
	if ((s.syncing || s.replacing) && s.locked == 0 &&
4014
	    !test_bit(STRIPE_COMPUTE_RUN, &sh->state) &&
4015
	    test_bit(STRIPE_INSYNC, &sh->state)) {
4016 4017
		md_done_sync(conf->mddev, STRIPE_SECTORS, 1);
		clear_bit(STRIPE_SYNCING, &sh->state);
4018 4019
		if (test_and_clear_bit(R5_Overlap, &sh->dev[sh->pd_idx].flags))
			wake_up(&conf->wait_for_overlap);
4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045
	}

	/* 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++;
				}
			}
		}

4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072
	/* 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++;
		}
	}
4073

4074 4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089
	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);
4090

4091
finish:
4092
	/* wait for this device to become unblocked */
4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104
	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);
	}
4105

4106 4107
	if (s.handle_bad_blocks)
		for (i = disks; i--; ) {
4108
			struct md_rdev *rdev;
4109 4110 4111 4112 4113 4114 4115 4116 4117
			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);
			}
4118 4119 4120
			if (test_and_clear_bit(R5_MadeGood, &dev->flags)) {
				rdev = conf->disks[i].rdev;
				rdev_clear_badblocks(rdev, sh->sector,
4121
						     STRIPE_SECTORS, 0);
4122 4123
				rdev_dec_pending(rdev, conf->mddev);
			}
4124 4125
			if (test_and_clear_bit(R5_MadeGoodRepl, &dev->flags)) {
				rdev = conf->disks[i].replacement;
4126 4127 4128
				if (!rdev)
					/* rdev have been moved down */
					rdev = conf->disks[i].rdev;
4129
				rdev_clear_badblocks(rdev, sh->sector,
4130
						     STRIPE_SECTORS, 0);
4131 4132
				rdev_dec_pending(rdev, conf->mddev);
			}
4133 4134
		}

4135 4136 4137
	if (s.ops_request)
		raid_run_ops(sh, s.ops_request);

D
Dan Williams 已提交
4138
	ops_run_io(sh, &s);
4139

4140
	if (s.dec_preread_active) {
4141
		/* We delay this until after ops_run_io so that if make_request
T
Tejun Heo 已提交
4142
		 * is waiting on a flush, it won't continue until the writes
4143 4144 4145 4146 4147 4148 4149 4150
		 * 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);
	}

4151
	return_io(s.return_bi);
4152

4153
	clear_bit_unlock(STRIPE_ACTIVE, &sh->state);
4154 4155
}

4156
static void raid5_activate_delayed(struct r5conf *conf)
4157 4158 4159 4160 4161 4162 4163 4164 4165 4166
{
	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);
4167
			list_add_tail(&sh->lru, &conf->hold_list);
4168
			raid5_wakeup_stripe_thread(sh);
4169
		}
N
NeilBrown 已提交
4170
	}
4171 4172
}

4173 4174
static void activate_bit_delay(struct r5conf *conf,
	struct list_head *temp_inactive_list)
4175 4176 4177 4178 4179 4180 4181
{
	/* 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);
4182
		int hash;
4183 4184
		list_del_init(&sh->lru);
		atomic_inc(&sh->count);
4185 4186
		hash = sh->hash_lock_index;
		__release_stripe(conf, sh, &temp_inactive_list[hash]);
4187 4188 4189
	}
}

4190
static int raid5_congested(struct mddev *mddev, int bits)
4191
{
4192
	struct r5conf *conf = mddev->private;
4193 4194 4195 4196

	/* No difference between reads and writes.  Just check
	 * how busy the stripe_cache is
	 */
4197

4198 4199 4200 4201
	if (conf->inactive_blocked)
		return 1;
	if (conf->quiesce)
		return 1;
4202
	if (atomic_read(&conf->empty_inactive_list_nr))
4203 4204 4205 4206 4207
		return 1;

	return 0;
}

4208 4209 4210
/* We want read requests to align with chunks where possible,
 * but write requests don't need to.
 */
4211
static int raid5_mergeable_bvec(struct mddev *mddev,
4212 4213
				struct bvec_merge_data *bvm,
				struct bio_vec *biovec)
4214
{
4215
	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
4216
	int max;
4217
	unsigned int chunk_sectors = mddev->chunk_sectors;
4218
	unsigned int bio_sectors = bvm->bi_size >> 9;
4219

4220
	if ((bvm->bi_rw & 1) == WRITE)
4221 4222
		return biovec->bv_len; /* always allow writes to be mergeable */

4223 4224
	if (mddev->new_chunk_sectors < mddev->chunk_sectors)
		chunk_sectors = mddev->new_chunk_sectors;
4225 4226 4227 4228 4229 4230 4231 4232
	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;
}

4233
static int in_chunk_boundary(struct mddev *mddev, struct bio *bio)
4234
{
4235
	sector_t sector = bio->bi_iter.bi_sector + get_start_sect(bio->bi_bdev);
4236
	unsigned int chunk_sectors = mddev->chunk_sectors;
4237
	unsigned int bio_sectors = bio_sectors(bio);
4238

4239 4240
	if (mddev->new_chunk_sectors < mddev->chunk_sectors)
		chunk_sectors = mddev->new_chunk_sectors;
4241 4242 4243 4244
	return  chunk_sectors >=
		((sector & (chunk_sectors - 1)) + bio_sectors);
}

4245 4246 4247 4248
/*
 *  add bio to the retry LIFO  ( in O(1) ... we are in interrupt )
 *  later sampled by raid5d.
 */
4249
static void add_bio_to_retry(struct bio *bi,struct r5conf *conf)
4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261
{
	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);
}

4262
static struct bio *remove_bio_from_retry(struct r5conf *conf)
4263 4264 4265 4266 4267 4268 4269 4270 4271 4272
{
	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) {
4273
		conf->retry_read_aligned_list = bi->bi_next;
4274
		bi->bi_next = NULL;
4275 4276 4277 4278
		/*
		 * this sets the active strip count to 1 and the processed
		 * strip count to zero (upper 8 bits)
		 */
4279
		raid5_set_bi_stripes(bi, 1); /* biased count of active stripes */
4280 4281 4282 4283 4284
	}

	return bi;
}

4285 4286 4287 4288 4289 4290
/*
 *  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..
 */
4291
static void raid5_align_endio(struct bio *bi, int error)
4292 4293
{
	struct bio* raid_bi  = bi->bi_private;
4294
	struct mddev *mddev;
4295
	struct r5conf *conf;
4296
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
4297
	struct md_rdev *rdev;
4298

4299
	bio_put(bi);
4300 4301 4302

	rdev = (void*)raid_bi->bi_next;
	raid_bi->bi_next = NULL;
4303 4304
	mddev = rdev->mddev;
	conf = mddev->private;
4305 4306 4307 4308

	rdev_dec_pending(rdev, conf->mddev);

	if (!error && uptodate) {
4309 4310
		trace_block_bio_complete(bdev_get_queue(raid_bi->bi_bdev),
					 raid_bi, 0);
4311
		bio_endio(raid_bi, 0);
4312 4313
		if (atomic_dec_and_test(&conf->active_aligned_reads))
			wake_up(&conf->wait_for_stripe);
4314
		return;
4315 4316
	}

4317
	pr_debug("raid5_align_endio : io error...handing IO for a retry\n");
4318 4319

	add_bio_to_retry(raid_bi, conf);
4320 4321
}

4322 4323
static int bio_fits_rdev(struct bio *bi)
{
4324
	struct request_queue *q = bdev_get_queue(bi->bi_bdev);
4325

4326
	if (bio_sectors(bi) > queue_max_sectors(q))
4327 4328
		return 0;
	blk_recount_segments(q, bi);
4329
	if (bi->bi_phys_segments > queue_max_segments(q))
4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 4340
		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;
}

4341
static int chunk_aligned_read(struct mddev *mddev, struct bio * raid_bio)
4342
{
4343
	struct r5conf *conf = mddev->private;
N
NeilBrown 已提交
4344
	int dd_idx;
4345
	struct bio* align_bi;
4346
	struct md_rdev *rdev;
4347
	sector_t end_sector;
4348 4349

	if (!in_chunk_boundary(mddev, raid_bio)) {
4350
		pr_debug("chunk_aligned_read : non aligned\n");
4351 4352 4353
		return 0;
	}
	/*
4354
	 * use bio_clone_mddev to make a copy of the bio
4355
	 */
4356
	align_bi = bio_clone_mddev(raid_bio, GFP_NOIO, mddev);
4357 4358 4359 4360 4361 4362 4363 4364 4365 4366 4367
	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
	 */
4368 4369 4370
	align_bi->bi_iter.bi_sector =
		raid5_compute_sector(conf, raid_bio->bi_iter.bi_sector,
				     0, &dd_idx, NULL);
4371

K
Kent Overstreet 已提交
4372
	end_sector = bio_end_sector(align_bi);
4373
	rcu_read_lock();
4374 4375 4376 4377 4378 4379 4380 4381 4382 4383 4384
	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) {
4385 4386 4387
		sector_t first_bad;
		int bad_sectors;

4388 4389
		atomic_inc(&rdev->nr_pending);
		rcu_read_unlock();
4390 4391
		raid_bio->bi_next = (void*)rdev;
		align_bi->bi_bdev =  rdev->bdev;
4392
		__clear_bit(BIO_SEG_VALID, &align_bi->bi_flags);
4393

4394
		if (!bio_fits_rdev(align_bi) ||
4395 4396
		    is_badblock(rdev, align_bi->bi_iter.bi_sector,
				bio_sectors(align_bi),
4397 4398
				&first_bad, &bad_sectors)) {
			/* too big in some way, or has a known bad block */
4399 4400 4401 4402 4403
			bio_put(align_bi);
			rdev_dec_pending(rdev, mddev);
			return 0;
		}

4404
		/* No reshape active, so we can trust rdev->data_offset */
4405
		align_bi->bi_iter.bi_sector += rdev->data_offset;
4406

4407 4408 4409
		spin_lock_irq(&conf->device_lock);
		wait_event_lock_irq(conf->wait_for_stripe,
				    conf->quiesce == 0,
4410
				    conf->device_lock);
4411 4412 4413
		atomic_inc(&conf->active_aligned_reads);
		spin_unlock_irq(&conf->device_lock);

4414 4415 4416
		if (mddev->gendisk)
			trace_block_bio_remap(bdev_get_queue(align_bi->bi_bdev),
					      align_bi, disk_devt(mddev->gendisk),
4417
					      raid_bio->bi_iter.bi_sector);
4418 4419 4420 4421
		generic_make_request(align_bi);
		return 1;
	} else {
		rcu_read_unlock();
4422
		bio_put(align_bi);
4423 4424 4425 4426
		return 0;
	}
}

4427 4428 4429 4430 4431 4432 4433 4434 4435 4436
/* __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.
 */
4437
static struct stripe_head *__get_priority_stripe(struct r5conf *conf, int group)
4438
{
4439 4440
	struct stripe_head *sh = NULL, *tmp;
	struct list_head *handle_list = NULL;
4441
	struct r5worker_group *wg = NULL;
4442 4443 4444 4445 4446

	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;
4447
		wg = &conf->worker_groups[group];
4448 4449 4450 4451
	} else {
		int i;
		for (i = 0; i < conf->group_cnt; i++) {
			handle_list = &conf->worker_groups[i].handle_list;
4452
			wg = &conf->worker_groups[i];
4453 4454 4455 4456
			if (!list_empty(handle_list))
				break;
		}
	}
4457 4458 4459

	pr_debug("%s: handle: %s hold: %s full_writes: %d bypass_count: %d\n",
		  __func__,
4460
		  list_empty(handle_list) ? "empty" : "busy",
4461 4462 4463
		  list_empty(&conf->hold_list) ? "empty" : "busy",
		  atomic_read(&conf->pending_full_writes), conf->bypass_count);

4464 4465
	if (!list_empty(handle_list)) {
		sh = list_entry(handle_list->next, typeof(*sh), lru);
4466 4467 4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482

		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)) {
4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498

		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;
		}
4499
		wg = NULL;
4500 4501 4502
	}

	if (!sh)
4503 4504
		return NULL;

4505 4506 4507 4508
	if (wg) {
		wg->stripes_cnt--;
		sh->group = NULL;
	}
4509
	list_del_init(&sh->lru);
4510
	BUG_ON(atomic_inc_return(&sh->count) != 1);
4511 4512
	return sh;
}
4513

4514 4515 4516
struct raid5_plug_cb {
	struct blk_plug_cb	cb;
	struct list_head	list;
4517
	struct list_head	temp_inactive_list[NR_STRIPE_HASH_LOCKS];
4518 4519 4520 4521 4522 4523 4524 4525 4526
};

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
4527
	int cnt = 0;
4528
	int hash;
4529 4530 4531 4532 4533 4534 4535 4536 4537 4538 4539

	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
			 */
4540
			smp_mb__before_atomic();
4541
			clear_bit(STRIPE_ON_UNPLUG_LIST, &sh->state);
S
Shaohua Li 已提交
4542 4543 4544 4545
			/*
			 * STRIPE_ON_RELEASE_LIST could be set here. In that
			 * case, the count is always > 1 here
			 */
4546 4547
			hash = sh->hash_lock_index;
			__release_stripe(conf, sh, &cb->temp_inactive_list[hash]);
N
NeilBrown 已提交
4548
			cnt++;
4549 4550 4551
		}
		spin_unlock_irq(&conf->device_lock);
	}
4552 4553
	release_inactive_stripe_list(conf, cb->temp_inactive_list,
				     NR_STRIPE_HASH_LOCKS);
4554 4555
	if (mddev->queue)
		trace_block_unplug(mddev->queue, cnt, !from_schedule);
4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573
	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);

4574 4575
	if (cb->list.next == NULL) {
		int i;
4576
		INIT_LIST_HEAD(&cb->list);
4577 4578 4579
		for (i = 0; i < NR_STRIPE_HASH_LOCKS; i++)
			INIT_LIST_HEAD(cb->temp_inactive_list + i);
	}
4580 4581 4582 4583 4584 4585 4586

	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
4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598
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;

4599 4600
	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
4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621

	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);
4622 4623 4624 4625 4626 4627 4628
		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
4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640
		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;
			}
		}
4641
		set_bit(STRIPE_DISCARD, &sh->state);
S
Shaohua Li 已提交
4642
		finish_wait(&conf->wait_for_overlap, &w);
4643
		sh->overwrite_disks = 0;
S
Shaohua Li 已提交
4644 4645 4646 4647 4648 4649
		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);
4650
			sh->overwrite_disks++;
S
Shaohua Li 已提交
4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675 4676 4677 4678
		}
		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);
	}
}

4679
static void make_request(struct mddev *mddev, struct bio * bi)
L
Linus Torvalds 已提交
4680
{
4681
	struct r5conf *conf = mddev->private;
4682
	int dd_idx;
L
Linus Torvalds 已提交
4683 4684 4685
	sector_t new_sector;
	sector_t logical_sector, last_sector;
	struct stripe_head *sh;
4686
	const int rw = bio_data_dir(bi);
4687
	int remaining;
4688 4689
	DEFINE_WAIT(w);
	bool do_prepare;
L
Linus Torvalds 已提交
4690

T
Tejun Heo 已提交
4691 4692
	if (unlikely(bi->bi_rw & REQ_FLUSH)) {
		md_flush_request(mddev, bi);
4693
		return;
4694 4695
	}

4696
	md_write_start(mddev, bi);
4697

4698
	if (rw == READ &&
4699
	     mddev->reshape_position == MaxSector &&
4700
	     chunk_aligned_read(mddev,bi))
4701
		return;
4702

S
Shaohua Li 已提交
4703 4704 4705 4706 4707
	if (unlikely(bi->bi_rw & REQ_DISCARD)) {
		make_discard_request(mddev, bi);
		return;
	}

4708
	logical_sector = bi->bi_iter.bi_sector & ~((sector_t)STRIPE_SECTORS-1);
K
Kent Overstreet 已提交
4709
	last_sector = bio_end_sector(bi);
L
Linus Torvalds 已提交
4710 4711
	bi->bi_next = NULL;
	bi->bi_phys_segments = 1;	/* over-loaded to count active stripes */
4712

4713
	prepare_to_wait(&conf->wait_for_overlap, &w, TASK_UNINTERRUPTIBLE);
L
Linus Torvalds 已提交
4714
	for (;logical_sector < last_sector; logical_sector += STRIPE_SECTORS) {
4715
		int previous;
4716
		int seq;
4717

4718
		do_prepare = false;
4719
	retry:
4720
		seq = read_seqcount_begin(&conf->gen_lock);
4721
		previous = 0;
4722 4723 4724
		if (do_prepare)
			prepare_to_wait(&conf->wait_for_overlap, &w,
				TASK_UNINTERRUPTIBLE);
4725
		if (unlikely(conf->reshape_progress != MaxSector)) {
4726
			/* spinlock is needed as reshape_progress may be
4727 4728
			 * 64bit on a 32bit platform, and so it might be
			 * possible to see a half-updated value
4729
			 * Of course reshape_progress could change after
4730 4731 4732 4733
			 * the lock is dropped, so once we get a reference
			 * to the stripe that we think it is, we will have
			 * to check again.
			 */
4734
			spin_lock_irq(&conf->device_lock);
4735
			if (mddev->reshape_backwards
4736 4737
			    ? logical_sector < conf->reshape_progress
			    : logical_sector >= conf->reshape_progress) {
4738 4739
				previous = 1;
			} else {
4740
				if (mddev->reshape_backwards
4741 4742
				    ? logical_sector < conf->reshape_safe
				    : logical_sector >= conf->reshape_safe) {
4743 4744
					spin_unlock_irq(&conf->device_lock);
					schedule();
4745
					do_prepare = true;
4746 4747 4748
					goto retry;
				}
			}
4749 4750
			spin_unlock_irq(&conf->device_lock);
		}
4751

4752 4753
		new_sector = raid5_compute_sector(conf, logical_sector,
						  previous,
4754
						  &dd_idx, NULL);
4755
		pr_debug("raid456: make_request, sector %llu logical %llu\n",
4756
			(unsigned long long)new_sector,
L
Linus Torvalds 已提交
4757 4758
			(unsigned long long)logical_sector);

4759
		sh = get_active_stripe(conf, new_sector, previous,
4760
				       (bi->bi_rw&RWA_MASK), 0);
L
Linus Torvalds 已提交
4761
		if (sh) {
4762
			if (unlikely(previous)) {
4763
				/* expansion might have moved on while waiting for a
4764 4765 4766 4767 4768 4769
				 * 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.
4770 4771 4772
				 */
				int must_retry = 0;
				spin_lock_irq(&conf->device_lock);
4773
				if (mddev->reshape_backwards
4774 4775
				    ? logical_sector >= conf->reshape_progress
				    : logical_sector < conf->reshape_progress)
4776 4777 4778 4779 4780
					/* mismatch, need to try again */
					must_retry = 1;
				spin_unlock_irq(&conf->device_lock);
				if (must_retry) {
					release_stripe(sh);
4781
					schedule();
4782
					do_prepare = true;
4783 4784 4785
					goto retry;
				}
			}
4786 4787 4788 4789 4790 4791 4792
			if (read_seqcount_retry(&conf->gen_lock, seq)) {
				/* Might have got the wrong stripe_head
				 * by accident
				 */
				release_stripe(sh);
				goto retry;
			}
4793

4794
			if (rw == WRITE &&
4795
			    logical_sector >= mddev->suspend_lo &&
4796 4797
			    logical_sector < mddev->suspend_hi) {
				release_stripe(sh);
4798 4799 4800 4801 4802 4803 4804 4805
				/* 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 &&
4806
				    logical_sector < mddev->suspend_hi) {
4807
					schedule();
4808 4809
					do_prepare = true;
				}
4810 4811
				goto retry;
			}
4812 4813

			if (test_bit(STRIPE_EXPANDING, &sh->state) ||
4814
			    !add_stripe_bio(sh, bi, dd_idx, rw, previous)) {
4815 4816
				/* Stripe is busy expanding or
				 * add failed due to overlap.  Flush everything
L
Linus Torvalds 已提交
4817 4818
				 * and wait a while
				 */
N
NeilBrown 已提交
4819
				md_wakeup_thread(mddev->thread);
L
Linus Torvalds 已提交
4820 4821
				release_stripe(sh);
				schedule();
4822
				do_prepare = true;
L
Linus Torvalds 已提交
4823 4824
				goto retry;
			}
4825 4826
			set_bit(STRIPE_HANDLE, &sh->state);
			clear_bit(STRIPE_DELAYED, &sh->state);
4827
			if ((bi->bi_rw & REQ_SYNC) &&
4828 4829
			    !test_and_set_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
				atomic_inc(&conf->preread_active_stripes);
4830
			release_stripe_plug(mddev, sh);
L
Linus Torvalds 已提交
4831 4832 4833 4834 4835 4836
		} else {
			/* cannot get stripe for read-ahead, just give-up */
			clear_bit(BIO_UPTODATE, &bi->bi_flags);
			break;
		}
	}
4837
	finish_wait(&conf->wait_for_overlap, &w);
4838

4839
	remaining = raid5_dec_bi_active_stripes(bi);
4840
	if (remaining == 0) {
L
Linus Torvalds 已提交
4841

4842
		if ( rw == WRITE )
L
Linus Torvalds 已提交
4843
			md_write_end(mddev);
4844

4845 4846
		trace_block_bio_complete(bdev_get_queue(bi->bi_bdev),
					 bi, 0);
4847
		bio_endio(bi, 0);
L
Linus Torvalds 已提交
4848 4849 4850
	}
}

4851
static sector_t raid5_size(struct mddev *mddev, sector_t sectors, int raid_disks);
D
Dan Williams 已提交
4852

4853
static sector_t reshape_request(struct mddev *mddev, sector_t sector_nr, int *skipped)
L
Linus Torvalds 已提交
4854
{
4855 4856 4857 4858 4859 4860 4861 4862 4863
	/* 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.
	 */
4864
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
4865
	struct stripe_head *sh;
4866
	sector_t first_sector, last_sector;
4867 4868 4869
	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;
4870 4871
	int i;
	int dd_idx;
4872
	sector_t writepos, readpos, safepos;
4873
	sector_t stripe_addr;
4874
	int reshape_sectors;
4875
	struct list_head stripes;
4876

4877 4878
	if (sector_nr == 0) {
		/* If restarting in the middle, skip the initial sectors */
4879
		if (mddev->reshape_backwards &&
4880 4881 4882
		    conf->reshape_progress < raid5_size(mddev, 0, 0)) {
			sector_nr = raid5_size(mddev, 0, 0)
				- conf->reshape_progress;
4883
		} else if (!mddev->reshape_backwards &&
4884 4885
			   conf->reshape_progress > 0)
			sector_nr = conf->reshape_progress;
4886
		sector_div(sector_nr, new_data_disks);
4887
		if (sector_nr) {
4888 4889
			mddev->curr_resync_completed = sector_nr;
			sysfs_notify(&mddev->kobj, NULL, "sync_completed");
4890 4891 4892
			*skipped = 1;
			return sector_nr;
		}
4893 4894
	}

4895 4896 4897 4898
	/* 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
	 */
4899 4900
	if (mddev->new_chunk_sectors > mddev->chunk_sectors)
		reshape_sectors = mddev->new_chunk_sectors;
4901
	else
4902
		reshape_sectors = mddev->chunk_sectors;
4903

4904 4905 4906 4907 4908
	/* 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
4909
	 */
4910
	writepos = conf->reshape_progress;
4911
	sector_div(writepos, new_data_disks);
4912 4913
	readpos = conf->reshape_progress;
	sector_div(readpos, data_disks);
4914
	safepos = conf->reshape_safe;
4915
	sector_div(safepos, data_disks);
4916
	if (mddev->reshape_backwards) {
4917
		writepos -= min_t(sector_t, reshape_sectors, writepos);
4918
		readpos += reshape_sectors;
4919
		safepos += reshape_sectors;
4920
	} else {
4921
		writepos += reshape_sectors;
4922 4923
		readpos -= min_t(sector_t, reshape_sectors, readpos);
		safepos -= min_t(sector_t, reshape_sectors, safepos);
4924
	}
4925

4926 4927 4928 4929 4930 4931 4932 4933 4934 4935 4936 4937 4938 4939 4940
	/* 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;
	}

4941 4942 4943 4944
	/* '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.
4945 4946 4947 4948
	 * 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
4949 4950 4951 4952 4953 4954 4955 4956 4957 4958 4959 4960
	 * 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???
	 */
4961 4962 4963 4964 4965 4966
	if (conf->min_offset_diff < 0) {
		safepos += -conf->min_offset_diff;
		readpos += -conf->min_offset_diff;
	} else
		writepos += conf->min_offset_diff;

4967
	if ((mddev->reshape_backwards
4968 4969 4970
	     ? (safepos > writepos && readpos < writepos)
	     : (safepos < writepos && readpos > writepos)) ||
	    time_after(jiffies, conf->reshape_checkpoint + 10*HZ)) {
4971 4972
		/* Cannot proceed until we've updated the superblock... */
		wait_event(conf->wait_for_overlap,
4973 4974 4975 4976
			   atomic_read(&conf->reshape_stripes)==0
			   || test_bit(MD_RECOVERY_INTR, &mddev->recovery));
		if (atomic_read(&conf->reshape_stripes) != 0)
			return 0;
4977
		mddev->reshape_position = conf->reshape_progress;
4978
		mddev->curr_resync_completed = sector_nr;
4979
		conf->reshape_checkpoint = jiffies;
4980
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
4981
		md_wakeup_thread(mddev->thread);
4982
		wait_event(mddev->sb_wait, mddev->flags == 0 ||
4983 4984 4985
			   test_bit(MD_RECOVERY_INTR, &mddev->recovery));
		if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
			return 0;
4986
		spin_lock_irq(&conf->device_lock);
4987
		conf->reshape_safe = mddev->reshape_position;
4988 4989
		spin_unlock_irq(&conf->device_lock);
		wake_up(&conf->wait_for_overlap);
4990
		sysfs_notify(&mddev->kobj, NULL, "sync_completed");
4991 4992
	}

4993
	INIT_LIST_HEAD(&stripes);
4994
	for (i = 0; i < reshape_sectors; i += STRIPE_SECTORS) {
4995
		int j;
4996
		int skipped_disk = 0;
4997
		sh = get_active_stripe(conf, stripe_addr+i, 0, 0, 1);
4998 4999 5000 5001 5002 5003 5004 5005 5006
		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;
5007
			if (conf->level == 6 &&
5008
			    j == sh->qd_idx)
5009
				continue;
5010
			s = compute_blocknr(sh, j, 0);
D
Dan Williams 已提交
5011
			if (s < raid5_size(mddev, 0, 0)) {
5012
				skipped_disk = 1;
5013 5014 5015 5016 5017 5018
				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);
		}
5019
		if (!skipped_disk) {
5020 5021 5022
			set_bit(STRIPE_EXPAND_READY, &sh->state);
			set_bit(STRIPE_HANDLE, &sh->state);
		}
5023
		list_add(&sh->lru, &stripes);
5024 5025
	}
	spin_lock_irq(&conf->device_lock);
5026
	if (mddev->reshape_backwards)
5027
		conf->reshape_progress -= reshape_sectors * new_data_disks;
5028
	else
5029
		conf->reshape_progress += reshape_sectors * new_data_disks;
5030 5031 5032 5033 5034 5035 5036
	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 =
5037
		raid5_compute_sector(conf, stripe_addr*(new_data_disks),
5038
				     1, &dd_idx, NULL);
5039
	last_sector =
5040
		raid5_compute_sector(conf, ((stripe_addr+reshape_sectors)
5041
					    * new_data_disks - 1),
5042
				     1, &dd_idx, NULL);
A
Andre Noll 已提交
5043 5044
	if (last_sector >= mddev->dev_sectors)
		last_sector = mddev->dev_sectors - 1;
5045
	while (first_sector <= last_sector) {
5046
		sh = get_active_stripe(conf, first_sector, 1, 0, 1);
5047 5048 5049 5050 5051
		set_bit(STRIPE_EXPAND_SOURCE, &sh->state);
		set_bit(STRIPE_HANDLE, &sh->state);
		release_stripe(sh);
		first_sector += STRIPE_SECTORS;
	}
5052 5053 5054 5055 5056 5057 5058 5059
	/* 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);
	}
5060 5061 5062
	/* If this takes us to the resync_max point where we have to pause,
	 * then we need to write out the superblock.
	 */
5063
	sector_nr += reshape_sectors;
5064 5065
	if ((sector_nr - mddev->curr_resync_completed) * 2
	    >= mddev->resync_max - mddev->curr_resync_completed) {
5066 5067
		/* Cannot proceed until we've updated the superblock... */
		wait_event(conf->wait_for_overlap,
5068 5069 5070 5071
			   atomic_read(&conf->reshape_stripes) == 0
			   || test_bit(MD_RECOVERY_INTR, &mddev->recovery));
		if (atomic_read(&conf->reshape_stripes) != 0)
			goto ret;
5072
		mddev->reshape_position = conf->reshape_progress;
5073
		mddev->curr_resync_completed = sector_nr;
5074
		conf->reshape_checkpoint = jiffies;
5075 5076 5077 5078
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
		md_wakeup_thread(mddev->thread);
		wait_event(mddev->sb_wait,
			   !test_bit(MD_CHANGE_DEVS, &mddev->flags)
5079 5080 5081
			   || test_bit(MD_RECOVERY_INTR, &mddev->recovery));
		if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
			goto ret;
5082
		spin_lock_irq(&conf->device_lock);
5083
		conf->reshape_safe = mddev->reshape_position;
5084 5085
		spin_unlock_irq(&conf->device_lock);
		wake_up(&conf->wait_for_overlap);
5086
		sysfs_notify(&mddev->kobj, NULL, "sync_completed");
5087
	}
5088
ret:
5089
	return reshape_sectors;
5090 5091
}

5092
static inline sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipped)
5093
{
5094
	struct r5conf *conf = mddev->private;
5095
	struct stripe_head *sh;
A
Andre Noll 已提交
5096
	sector_t max_sector = mddev->dev_sectors;
N
NeilBrown 已提交
5097
	sector_t sync_blocks;
5098 5099
	int still_degraded = 0;
	int i;
L
Linus Torvalds 已提交
5100

5101
	if (sector_nr >= max_sector) {
L
Linus Torvalds 已提交
5102
		/* just being told to finish up .. nothing much to do */
5103

5104 5105 5106 5107
		if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) {
			end_reshape(conf);
			return 0;
		}
5108 5109 5110 5111

		if (mddev->curr_resync < max_sector) /* aborted */
			bitmap_end_sync(mddev->bitmap, mddev->curr_resync,
					&sync_blocks, 1);
5112
		else /* completed sync */
5113 5114 5115
			conf->fullsync = 0;
		bitmap_close_sync(mddev->bitmap);

L
Linus Torvalds 已提交
5116 5117
		return 0;
	}
5118

5119 5120 5121
	/* Allow raid5_quiesce to complete */
	wait_event(conf->wait_for_overlap, conf->quiesce != 2);

5122 5123
	if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
		return reshape_request(mddev, sector_nr, skipped);
5124

5125 5126 5127 5128 5129 5130
	/* 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
	 */

5131
	/* if there is too many failed drives and we are trying
L
Linus Torvalds 已提交
5132 5133 5134
	 * to resync, then assert that we are finished, because there is
	 * nothing we can do.
	 */
5135
	if (mddev->degraded >= conf->max_degraded &&
5136
	    test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
A
Andre Noll 已提交
5137
		sector_t rv = mddev->dev_sectors - sector_nr;
5138
		*skipped = 1;
L
Linus Torvalds 已提交
5139 5140
		return rv;
	}
5141 5142 5143 5144
	if (!test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
	    !conf->fullsync &&
	    !bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, 1) &&
	    sync_blocks >= STRIPE_SECTORS) {
5145 5146 5147 5148 5149
		/* 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
5150

N
NeilBrown 已提交
5151 5152
	bitmap_cond_end_sync(mddev->bitmap, sector_nr);

5153
	sh = get_active_stripe(conf, sector_nr, 0, 1, 0);
L
Linus Torvalds 已提交
5154
	if (sh == NULL) {
5155
		sh = get_active_stripe(conf, sector_nr, 0, 0, 0);
L
Linus Torvalds 已提交
5156
		/* make sure we don't swamp the stripe cache if someone else
5157
		 * is trying to get access
L
Linus Torvalds 已提交
5158
		 */
5159
		schedule_timeout_uninterruptible(1);
L
Linus Torvalds 已提交
5160
	}
5161
	/* Need to check if array will still be degraded after recovery/resync
5162 5163
	 * Note in case of > 1 drive failures it's possible we're rebuilding
	 * one drive while leaving another faulty drive in array.
5164
	 */
5165 5166 5167 5168 5169
	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))
5170
			still_degraded = 1;
5171 5172
	}
	rcu_read_unlock();
5173 5174 5175

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

5176
	set_bit(STRIPE_SYNC_REQUESTED, &sh->state);
5177
	set_bit(STRIPE_HANDLE, &sh->state);
L
Linus Torvalds 已提交
5178 5179 5180 5181 5182 5183

	release_stripe(sh);

	return STRIPE_SECTORS;
}

5184
static int  retry_aligned_read(struct r5conf *conf, struct bio *raid_bio)
5185 5186 5187 5188 5189 5190 5191 5192 5193 5194 5195 5196
{
	/* 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;
5197
	int dd_idx;
5198 5199 5200 5201 5202
	sector_t sector, logical_sector, last_sector;
	int scnt = 0;
	int remaining;
	int handled = 0;

5203 5204
	logical_sector = raid_bio->bi_iter.bi_sector &
		~((sector_t)STRIPE_SECTORS-1);
5205
	sector = raid5_compute_sector(conf, logical_sector,
5206
				      0, &dd_idx, NULL);
K
Kent Overstreet 已提交
5207
	last_sector = bio_end_sector(raid_bio);
5208 5209

	for (; logical_sector < last_sector;
5210 5211 5212
	     logical_sector += STRIPE_SECTORS,
		     sector += STRIPE_SECTORS,
		     scnt++) {
5213

5214
		if (scnt < raid5_bi_processed_stripes(raid_bio))
5215 5216 5217
			/* already done this stripe */
			continue;

5218
		sh = get_active_stripe(conf, sector, 0, 1, 1);
5219 5220 5221

		if (!sh) {
			/* failed to get a stripe - must wait */
5222
			raid5_set_bi_processed_stripes(raid_bio, scnt);
5223 5224 5225 5226
			conf->retry_read_aligned = raid_bio;
			return handled;
		}

5227
		if (!add_stripe_bio(sh, raid_bio, dd_idx, 0, 0)) {
5228
			release_stripe(sh);
5229
			raid5_set_bi_processed_stripes(raid_bio, scnt);
5230 5231 5232 5233
			conf->retry_read_aligned = raid_bio;
			return handled;
		}

5234
		set_bit(R5_ReadNoMerge, &sh->dev[dd_idx].flags);
5235
		handle_stripe(sh);
5236 5237 5238
		release_stripe(sh);
		handled++;
	}
5239
	remaining = raid5_dec_bi_active_stripes(raid_bio);
5240 5241 5242
	if (remaining == 0) {
		trace_block_bio_complete(bdev_get_queue(raid_bio->bi_bdev),
					 raid_bio, 0);
5243
		bio_endio(raid_bio, 0);
5244
	}
5245 5246 5247 5248 5249
	if (atomic_dec_and_test(&conf->active_aligned_reads))
		wake_up(&conf->wait_for_stripe);
	return handled;
}

5250
static int handle_active_stripes(struct r5conf *conf, int group,
5251 5252
				 struct r5worker *worker,
				 struct list_head *temp_inactive_list)
5253 5254
{
	struct stripe_head *batch[MAX_STRIPE_BATCH], *sh;
5255 5256
	int i, batch_size = 0, hash;
	bool release_inactive = false;
5257 5258

	while (batch_size < MAX_STRIPE_BATCH &&
5259
			(sh = __get_priority_stripe(conf, group)) != NULL)
5260 5261
		batch[batch_size++] = sh;

5262 5263 5264 5265 5266 5267 5268 5269
	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;
	}
5270 5271
	spin_unlock_irq(&conf->device_lock);

5272 5273 5274 5275 5276 5277 5278 5279
	release_inactive_stripe_list(conf, temp_inactive_list,
				     NR_STRIPE_HASH_LOCKS);

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

5280 5281 5282 5283 5284 5285
	for (i = 0; i < batch_size; i++)
		handle_stripe(batch[i]);

	cond_resched();

	spin_lock_irq(&conf->device_lock);
5286 5287 5288 5289
	for (i = 0; i < batch_size; i++) {
		hash = batch[i]->hash_lock_index;
		__release_stripe(conf, batch[i], &temp_inactive_list[hash]);
	}
5290 5291
	return batch_size;
}
5292

5293 5294 5295 5296 5297 5298 5299 5300 5301 5302 5303 5304 5305 5306 5307 5308 5309
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;

5310
		released = release_stripe_list(conf, worker->temp_inactive_list);
5311

5312 5313
		batch_size = handle_active_stripes(conf, group_id, worker,
						   worker->temp_inactive_list);
5314
		worker->working = false;
5315 5316 5317 5318 5319 5320 5321 5322 5323 5324 5325 5326
		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
5327 5328 5329 5330 5331 5332 5333
/*
 * 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
5334
static void raid5d(struct md_thread *thread)
L
Linus Torvalds 已提交
5335
{
S
Shaohua Li 已提交
5336
	struct mddev *mddev = thread->mddev;
5337
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
5338
	int handled;
5339
	struct blk_plug plug;
L
Linus Torvalds 已提交
5340

5341
	pr_debug("+++ raid5d active\n");
L
Linus Torvalds 已提交
5342 5343 5344

	md_check_recovery(mddev);

5345
	blk_start_plug(&plug);
L
Linus Torvalds 已提交
5346 5347 5348
	handled = 0;
	spin_lock_irq(&conf->device_lock);
	while (1) {
5349
		struct bio *bio;
S
Shaohua Li 已提交
5350 5351
		int batch_size, released;

5352
		released = release_stripe_list(conf, conf->temp_inactive_list);
L
Linus Torvalds 已提交
5353

5354
		if (
5355 5356 5357
		    !list_empty(&conf->bitmap_list)) {
			/* Now is a good time to flush some bitmap updates */
			conf->seq_flush++;
5358
			spin_unlock_irq(&conf->device_lock);
5359
			bitmap_unplug(mddev->bitmap);
5360
			spin_lock_irq(&conf->device_lock);
5361
			conf->seq_write = conf->seq_flush;
5362
			activate_bit_delay(conf, conf->temp_inactive_list);
5363
		}
5364
		raid5_activate_delayed(conf);
5365

5366 5367 5368 5369 5370 5371 5372 5373 5374 5375
		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++;
		}

5376 5377
		batch_size = handle_active_stripes(conf, ANY_GROUP, NULL,
						   conf->temp_inactive_list);
S
Shaohua Li 已提交
5378
		if (!batch_size && !released)
L
Linus Torvalds 已提交
5379
			break;
5380
		handled += batch_size;
L
Linus Torvalds 已提交
5381

5382 5383
		if (mddev->flags & ~(1<<MD_CHANGE_PENDING)) {
			spin_unlock_irq(&conf->device_lock);
5384
			md_check_recovery(mddev);
5385 5386
			spin_lock_irq(&conf->device_lock);
		}
L
Linus Torvalds 已提交
5387
	}
5388
	pr_debug("%d stripes handled\n", handled);
L
Linus Torvalds 已提交
5389 5390 5391

	spin_unlock_irq(&conf->device_lock);

5392
	async_tx_issue_pending_all();
5393
	blk_finish_plug(&plug);
L
Linus Torvalds 已提交
5394

5395
	pr_debug("--- raid5d inactive\n");
L
Linus Torvalds 已提交
5396 5397
}

5398
static ssize_t
5399
raid5_show_stripe_cache_size(struct mddev *mddev, char *page)
5400
{
5401 5402 5403 5404
	struct r5conf *conf;
	int ret = 0;
	spin_lock(&mddev->lock);
	conf = mddev->private;
5405
	if (conf)
5406 5407 5408
		ret = sprintf(page, "%d\n", conf->max_nr_stripes);
	spin_unlock(&mddev->lock);
	return ret;
5409 5410
}

5411
int
5412
raid5_set_cache_size(struct mddev *mddev, int size)
5413
{
5414
	struct r5conf *conf = mddev->private;
5415
	int err;
5416
	int hash;
5417

5418
	if (size <= 16 || size > 32768)
5419
		return -EINVAL;
5420
	hash = (conf->max_nr_stripes - 1) % NR_STRIPE_HASH_LOCKS;
5421
	while (size < conf->max_nr_stripes) {
5422
		if (drop_one_stripe(conf, hash))
5423 5424 5425
			conf->max_nr_stripes--;
		else
			break;
5426 5427 5428
		hash--;
		if (hash < 0)
			hash = NR_STRIPE_HASH_LOCKS - 1;
5429
	}
5430 5431 5432
	err = md_allow_write(mddev);
	if (err)
		return err;
5433
	hash = conf->max_nr_stripes % NR_STRIPE_HASH_LOCKS;
5434
	while (size > conf->max_nr_stripes) {
5435
		if (grow_one_stripe(conf, hash))
5436 5437
			conf->max_nr_stripes++;
		else break;
5438
		hash = (hash + 1) % NR_STRIPE_HASH_LOCKS;
5439
	}
5440 5441 5442 5443 5444
	return 0;
}
EXPORT_SYMBOL(raid5_set_cache_size);

static ssize_t
5445
raid5_store_stripe_cache_size(struct mddev *mddev, const char *page, size_t len)
5446
{
5447
	struct r5conf *conf;
5448 5449 5450 5451 5452
	unsigned long new;
	int err;

	if (len >= PAGE_SIZE)
		return -EINVAL;
5453
	if (kstrtoul(page, 10, &new))
5454
		return -EINVAL;
5455
	err = mddev_lock(mddev);
5456 5457
	if (err)
		return err;
5458 5459 5460 5461 5462 5463 5464 5465
	conf = mddev->private;
	if (!conf)
		err = -ENODEV;
	else
		err = raid5_set_cache_size(mddev, new);
	mddev_unlock(mddev);

	return err ?: len;
5466
}
5467

5468 5469 5470 5471
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);
5472

5473
static ssize_t
5474
raid5_show_preread_threshold(struct mddev *mddev, char *page)
5475
{
5476 5477 5478 5479
	struct r5conf *conf;
	int ret = 0;
	spin_lock(&mddev->lock);
	conf = mddev->private;
5480
	if (conf)
5481 5482 5483
		ret = sprintf(page, "%d\n", conf->bypass_threshold);
	spin_unlock(&mddev->lock);
	return ret;
5484 5485 5486
}

static ssize_t
5487
raid5_store_preread_threshold(struct mddev *mddev, const char *page, size_t len)
5488
{
5489
	struct r5conf *conf;
5490
	unsigned long new;
5491 5492
	int err;

5493 5494
	if (len >= PAGE_SIZE)
		return -EINVAL;
5495
	if (kstrtoul(page, 10, &new))
5496
		return -EINVAL;
5497 5498 5499 5500 5501 5502 5503 5504 5505 5506 5507 5508 5509

	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;
5510 5511 5512 5513 5514 5515 5516 5517
}

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);

5518 5519 5520
static ssize_t
raid5_show_skip_copy(struct mddev *mddev, char *page)
{
5521 5522 5523 5524
	struct r5conf *conf;
	int ret = 0;
	spin_lock(&mddev->lock);
	conf = mddev->private;
5525
	if (conf)
5526 5527 5528
		ret = sprintf(page, "%d\n", conf->skip_copy);
	spin_unlock(&mddev->lock);
	return ret;
5529 5530 5531 5532 5533
}

static ssize_t
raid5_store_skip_copy(struct mddev *mddev, const char *page, size_t len)
{
5534
	struct r5conf *conf;
5535
	unsigned long new;
5536 5537
	int err;

5538 5539 5540 5541 5542
	if (len >= PAGE_SIZE)
		return -EINVAL;
	if (kstrtoul(page, 10, &new))
		return -EINVAL;
	new = !!new;
5543 5544 5545 5546 5547 5548 5549 5550 5551 5552 5553 5554 5555 5556 5557 5558 5559 5560 5561 5562

	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;
5563 5564 5565 5566 5567 5568 5569
}

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

5570
static ssize_t
5571
stripe_cache_active_show(struct mddev *mddev, char *page)
5572
{
5573
	struct r5conf *conf = mddev->private;
5574 5575 5576 5577
	if (conf)
		return sprintf(page, "%d\n", atomic_read(&conf->active_stripes));
	else
		return 0;
5578 5579
}

5580 5581
static struct md_sysfs_entry
raid5_stripecache_active = __ATTR_RO(stripe_cache_active);
5582

5583 5584 5585
static ssize_t
raid5_show_group_thread_cnt(struct mddev *mddev, char *page)
{
5586 5587 5588 5589
	struct r5conf *conf;
	int ret = 0;
	spin_lock(&mddev->lock);
	conf = mddev->private;
5590
	if (conf)
5591 5592 5593
		ret = sprintf(page, "%d\n", conf->worker_cnt_per_group);
	spin_unlock(&mddev->lock);
	return ret;
5594 5595
}

5596 5597 5598 5599
static int alloc_thread_groups(struct r5conf *conf, int cnt,
			       int *group_cnt,
			       int *worker_cnt_per_group,
			       struct r5worker_group **worker_groups);
5600 5601 5602
static ssize_t
raid5_store_group_thread_cnt(struct mddev *mddev, const char *page, size_t len)
{
5603
	struct r5conf *conf;
5604 5605
	unsigned long new;
	int err;
5606 5607
	struct r5worker_group *new_groups, *old_groups;
	int group_cnt, worker_cnt_per_group;
5608 5609 5610 5611 5612 5613

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

5614 5615 5616 5617 5618 5619 5620 5621
	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);
5622

5623 5624 5625
		old_groups = conf->worker_groups;
		if (old_groups)
			flush_workqueue(raid5_wq);
5626

5627 5628 5629 5630 5631 5632 5633 5634 5635
		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);
5636

5637 5638 5639 5640 5641
			if (old_groups)
				kfree(old_groups[0].workers);
			kfree(old_groups);
		}
		mddev_resume(mddev);
5642
	}
5643
	mddev_unlock(mddev);
5644

5645
	return err ?: len;
5646 5647 5648 5649 5650 5651 5652
}

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);

5653
static struct attribute *raid5_attrs[] =  {
5654 5655
	&raid5_stripecache_size.attr,
	&raid5_stripecache_active.attr,
5656
	&raid5_preread_bypass_threshold.attr,
5657
	&raid5_group_thread_cnt.attr,
5658
	&raid5_skip_copy.attr,
5659 5660
	NULL,
};
5661 5662 5663
static struct attribute_group raid5_attrs_group = {
	.name = NULL,
	.attrs = raid5_attrs,
5664 5665
};

5666 5667 5668 5669
static int alloc_thread_groups(struct r5conf *conf, int cnt,
			       int *group_cnt,
			       int *worker_cnt_per_group,
			       struct r5worker_group **worker_groups)
5670
{
5671
	int i, j, k;
5672 5673 5674
	ssize_t size;
	struct r5worker *workers;

5675
	*worker_cnt_per_group = cnt;
5676
	if (cnt == 0) {
5677 5678
		*group_cnt = 0;
		*worker_groups = NULL;
5679 5680
		return 0;
	}
5681
	*group_cnt = num_possible_nodes();
5682
	size = sizeof(struct r5worker) * cnt;
5683 5684 5685 5686
	workers = kzalloc(size * *group_cnt, GFP_NOIO);
	*worker_groups = kzalloc(sizeof(struct r5worker_group) *
				*group_cnt, GFP_NOIO);
	if (!*worker_groups || !workers) {
5687
		kfree(workers);
5688
		kfree(*worker_groups);
5689 5690 5691
		return -ENOMEM;
	}

5692
	for (i = 0; i < *group_cnt; i++) {
5693 5694
		struct r5worker_group *group;

5695
		group = &(*worker_groups)[i];
5696 5697 5698 5699 5700
		INIT_LIST_HEAD(&group->handle_list);
		group->conf = conf;
		group->workers = workers + i * cnt;

		for (j = 0; j < cnt; j++) {
5701 5702 5703 5704 5705 5706
			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);
5707 5708 5709 5710 5711 5712 5713 5714 5715 5716 5717 5718 5719 5720
		}
	}

	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;
}

5721
static sector_t
5722
raid5_size(struct mddev *mddev, sector_t sectors, int raid_disks)
5723
{
5724
	struct r5conf *conf = mddev->private;
5725 5726 5727

	if (!sectors)
		sectors = mddev->dev_sectors;
5728
	if (!raid_disks)
5729
		/* size is defined by the smallest of previous and new size */
5730
		raid_disks = min(conf->raid_disks, conf->previous_raid_disks);
5731

5732
	sectors &= ~((sector_t)mddev->chunk_sectors - 1);
5733
	sectors &= ~((sector_t)mddev->new_chunk_sectors - 1);
5734 5735 5736
	return sectors * (raid_disks - conf->max_degraded);
}

5737 5738 5739
static void free_scratch_buffer(struct r5conf *conf, struct raid5_percpu *percpu)
{
	safe_put_page(percpu->spare_page);
5740 5741
	if (percpu->scribble)
		flex_array_free(percpu->scribble);
5742 5743 5744 5745 5746 5747 5748 5749 5750
	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)
5751 5752 5753
		percpu->scribble = scribble_alloc(max(conf->raid_disks,
			conf->previous_raid_disks), conf->chunk_sectors /
			STRIPE_SECTORS, GFP_KERNEL);
5754 5755 5756 5757 5758 5759 5760 5761 5762

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

	return 0;
}

5763
static void raid5_free_percpu(struct r5conf *conf)
5764 5765 5766 5767 5768 5769 5770 5771 5772
{
	unsigned long cpu;

	if (!conf->percpu)
		return;

#ifdef CONFIG_HOTPLUG_CPU
	unregister_cpu_notifier(&conf->cpu_notify);
#endif
5773 5774 5775 5776

	get_online_cpus();
	for_each_possible_cpu(cpu)
		free_scratch_buffer(conf, per_cpu_ptr(conf->percpu, cpu));
5777 5778 5779 5780 5781
	put_online_cpus();

	free_percpu(conf->percpu);
}

5782
static void free_conf(struct r5conf *conf)
5783
{
5784
	free_thread_groups(conf);
5785
	shrink_stripes(conf);
5786
	raid5_free_percpu(conf);
5787 5788 5789 5790 5791
	kfree(conf->disks);
	kfree(conf->stripe_hashtbl);
	kfree(conf);
}

5792 5793 5794 5795
#ifdef CONFIG_HOTPLUG_CPU
static int raid456_cpu_notify(struct notifier_block *nfb, unsigned long action,
			      void *hcpu)
{
5796
	struct r5conf *conf = container_of(nfb, struct r5conf, cpu_notify);
5797 5798 5799 5800 5801 5802
	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:
5803
		if (alloc_scratch_buffer(conf, percpu)) {
5804 5805
			pr_err("%s: failed memory allocation for cpu%ld\n",
			       __func__, cpu);
5806
			return notifier_from_errno(-ENOMEM);
5807 5808 5809 5810
		}
		break;
	case CPU_DEAD:
	case CPU_DEAD_FROZEN:
5811
		free_scratch_buffer(conf, per_cpu_ptr(conf->percpu, cpu));
5812 5813 5814 5815 5816 5817 5818 5819
		break;
	default:
		break;
	}
	return NOTIFY_OK;
}
#endif

5820
static int raid5_alloc_percpu(struct r5conf *conf)
5821 5822
{
	unsigned long cpu;
5823
	int err = 0;
5824

5825 5826
	conf->percpu = alloc_percpu(struct raid5_percpu);
	if (!conf->percpu)
5827
		return -ENOMEM;
5828 5829 5830 5831 5832 5833 5834 5835

#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
5836 5837 5838

	get_online_cpus();
	for_each_present_cpu(cpu) {
5839 5840 5841 5842
		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);
5843 5844 5845 5846 5847 5848 5849 5850
			break;
		}
	}
	put_online_cpus();

	return err;
}

5851
static struct r5conf *setup_conf(struct mddev *mddev)
L
Linus Torvalds 已提交
5852
{
5853
	struct r5conf *conf;
5854
	int raid_disk, memory, max_disks;
5855
	struct md_rdev *rdev;
L
Linus Torvalds 已提交
5856
	struct disk_info *disk;
5857
	char pers_name[6];
5858
	int i;
5859 5860
	int group_cnt, worker_cnt_per_group;
	struct r5worker_group *new_group;
L
Linus Torvalds 已提交
5861

N
NeilBrown 已提交
5862 5863 5864
	if (mddev->new_level != 5
	    && mddev->new_level != 4
	    && mddev->new_level != 6) {
5865
		printk(KERN_ERR "md/raid:%s: raid level not set to 4/5/6 (%d)\n",
N
NeilBrown 已提交
5866 5867
		       mdname(mddev), mddev->new_level);
		return ERR_PTR(-EIO);
L
Linus Torvalds 已提交
5868
	}
N
NeilBrown 已提交
5869 5870 5871 5872
	if ((mddev->new_level == 5
	     && !algorithm_valid_raid5(mddev->new_layout)) ||
	    (mddev->new_level == 6
	     && !algorithm_valid_raid6(mddev->new_layout))) {
5873
		printk(KERN_ERR "md/raid:%s: layout %d not supported\n",
N
NeilBrown 已提交
5874 5875
		       mdname(mddev), mddev->new_layout);
		return ERR_PTR(-EIO);
5876
	}
N
NeilBrown 已提交
5877
	if (mddev->new_level == 6 && mddev->raid_disks < 4) {
5878
		printk(KERN_ERR "md/raid:%s: not enough configured devices (%d, minimum 4)\n",
N
NeilBrown 已提交
5879 5880
		       mdname(mddev), mddev->raid_disks);
		return ERR_PTR(-EINVAL);
5881 5882
	}

5883 5884 5885
	if (!mddev->new_chunk_sectors ||
	    (mddev->new_chunk_sectors << 9) % PAGE_SIZE ||
	    !is_power_of_2(mddev->new_chunk_sectors)) {
5886 5887
		printk(KERN_ERR "md/raid:%s: invalid chunk size %d\n",
		       mdname(mddev), mddev->new_chunk_sectors << 9);
N
NeilBrown 已提交
5888
		return ERR_PTR(-EINVAL);
5889 5890
	}

5891
	conf = kzalloc(sizeof(struct r5conf), GFP_KERNEL);
N
NeilBrown 已提交
5892
	if (conf == NULL)
L
Linus Torvalds 已提交
5893
		goto abort;
5894
	/* Don't enable multi-threading by default*/
5895 5896 5897 5898 5899 5900
	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
5901
		goto abort;
5902
	spin_lock_init(&conf->device_lock);
5903
	seqcount_init(&conf->gen_lock);
5904 5905 5906 5907 5908 5909
	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
5910
	init_llist_head(&conf->released_stripes);
5911 5912 5913 5914
	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;
5915
	conf->recovery_disabled = mddev->recovery_disabled - 1;
N
NeilBrown 已提交
5916 5917 5918 5919 5920

	conf->raid_disks = mddev->raid_disks;
	if (mddev->reshape_position == MaxSector)
		conf->previous_raid_disks = mddev->raid_disks;
	else
5921
		conf->previous_raid_disks = mddev->raid_disks - mddev->delta_disks;
5922
	max_disks = max(conf->raid_disks, conf->previous_raid_disks);
5923

5924
	conf->disks = kzalloc(max_disks * sizeof(struct disk_info),
5925 5926 5927
			      GFP_KERNEL);
	if (!conf->disks)
		goto abort;
5928

L
Linus Torvalds 已提交
5929 5930
	conf->mddev = mddev;

5931
	if ((conf->stripe_hashtbl = kzalloc(PAGE_SIZE, GFP_KERNEL)) == NULL)
L
Linus Torvalds 已提交
5932 5933
		goto abort;

5934 5935 5936 5937 5938 5939 5940 5941 5942 5943 5944 5945 5946 5947 5948
	/* 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);

5949
	conf->level = mddev->new_level;
5950
	conf->chunk_sectors = mddev->new_chunk_sectors;
5951 5952 5953
	if (raid5_alloc_percpu(conf) != 0)
		goto abort;

5954
	pr_debug("raid456: run(%s) called.\n", mdname(mddev));
L
Linus Torvalds 已提交
5955

N
NeilBrown 已提交
5956
	rdev_for_each(rdev, mddev) {
L
Linus Torvalds 已提交
5957
		raid_disk = rdev->raid_disk;
5958
		if (raid_disk >= max_disks
L
Linus Torvalds 已提交
5959 5960 5961 5962
		    || raid_disk < 0)
			continue;
		disk = conf->disks + raid_disk;

5963 5964 5965 5966 5967 5968 5969 5970 5971
		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
5972

5973
		if (test_bit(In_sync, &rdev->flags)) {
L
Linus Torvalds 已提交
5974
			char b[BDEVNAME_SIZE];
5975 5976 5977
			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
5978
		} else if (rdev->saved_raid_disk != raid_disk)
5979 5980
			/* Cannot rely on bitmap to complete recovery */
			conf->fullsync = 1;
L
Linus Torvalds 已提交
5981 5982
	}

N
NeilBrown 已提交
5983
	conf->level = mddev->new_level;
5984 5985 5986 5987
	if (conf->level == 6)
		conf->max_degraded = 2;
	else
		conf->max_degraded = 1;
N
NeilBrown 已提交
5988
	conf->algorithm = mddev->new_layout;
5989
	conf->reshape_progress = mddev->reshape_position;
5990
	if (conf->reshape_progress != MaxSector) {
5991
		conf->prev_chunk_sectors = mddev->chunk_sectors;
5992 5993
		conf->prev_algo = mddev->layout;
	}
L
Linus Torvalds 已提交
5994

N
NeilBrown 已提交
5995
	memory = conf->max_nr_stripes * (sizeof(struct stripe_head) +
5996
		 max_disks * ((sizeof(struct bio) + PAGE_SIZE))) / 1024;
5997
	atomic_set(&conf->empty_inactive_list_nr, NR_STRIPE_HASH_LOCKS);
5998
	if (grow_stripes(conf, NR_STRIPES)) {
N
NeilBrown 已提交
5999
		printk(KERN_ERR
6000 6001
		       "md/raid:%s: couldn't allocate %dkB for buffers\n",
		       mdname(mddev), memory);
N
NeilBrown 已提交
6002 6003
		goto abort;
	} else
6004 6005
		printk(KERN_INFO "md/raid:%s: allocated %dkB\n",
		       mdname(mddev), memory);
L
Linus Torvalds 已提交
6006

6007 6008
	sprintf(pers_name, "raid%d", mddev->new_level);
	conf->thread = md_register_thread(raid5d, mddev, pers_name);
N
NeilBrown 已提交
6009 6010
	if (!conf->thread) {
		printk(KERN_ERR
6011
		       "md/raid:%s: couldn't allocate thread.\n",
N
NeilBrown 已提交
6012
		       mdname(mddev));
6013 6014
		goto abort;
	}
N
NeilBrown 已提交
6015 6016 6017 6018 6019

	return conf;

 abort:
	if (conf) {
6020
		free_conf(conf);
N
NeilBrown 已提交
6021 6022 6023 6024 6025
		return ERR_PTR(-EIO);
	} else
		return ERR_PTR(-ENOMEM);
}

6026 6027 6028 6029 6030 6031 6032 6033 6034 6035 6036 6037
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:
6038
		if (raid_disk == 0 ||
6039 6040 6041 6042 6043 6044 6045 6046 6047 6048 6049 6050 6051
		    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;
}

6052
static int run(struct mddev *mddev)
N
NeilBrown 已提交
6053
{
6054
	struct r5conf *conf;
6055
	int working_disks = 0;
6056
	int dirty_parity_disks = 0;
6057
	struct md_rdev *rdev;
6058
	sector_t reshape_offset = 0;
6059
	int i;
6060 6061
	long long min_offset_diff = 0;
	int first = 1;
N
NeilBrown 已提交
6062

6063
	if (mddev->recovery_cp != MaxSector)
6064
		printk(KERN_NOTICE "md/raid:%s: not clean"
6065 6066
		       " -- starting background reconstruction\n",
		       mdname(mddev));
6067 6068 6069 6070 6071 6072 6073 6074 6075 6076 6077 6078 6079 6080 6081 6082 6083

	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
6084 6085
	if (mddev->reshape_position != MaxSector) {
		/* Check that we can continue the reshape.
6086 6087 6088 6089 6090 6091 6092 6093 6094 6095
		 * 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
6096 6097 6098
		 */
		sector_t here_new, here_old;
		int old_disks;
6099
		int max_degraded = (mddev->level == 6 ? 2 : 1);
N
NeilBrown 已提交
6100

6101
		if (mddev->new_level != mddev->level) {
6102
			printk(KERN_ERR "md/raid:%s: unsupported reshape "
N
NeilBrown 已提交
6103 6104 6105 6106 6107 6108 6109 6110 6111 6112
			       "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;
6113
		if (sector_div(here_new, mddev->new_chunk_sectors *
N
NeilBrown 已提交
6114
			       (mddev->raid_disks - max_degraded))) {
6115 6116
			printk(KERN_ERR "md/raid:%s: reshape_position not "
			       "on a stripe boundary\n", mdname(mddev));
N
NeilBrown 已提交
6117 6118
			return -EINVAL;
		}
6119
		reshape_offset = here_new * mddev->new_chunk_sectors;
N
NeilBrown 已提交
6120 6121
		/* here_new is the stripe we will write to */
		here_old = mddev->reshape_position;
6122
		sector_div(here_old, mddev->chunk_sectors *
N
NeilBrown 已提交
6123 6124 6125
			   (old_disks-max_degraded));
		/* here_old is the first stripe that we might need to read
		 * from */
6126
		if (mddev->delta_disks == 0) {
6127 6128 6129 6130 6131 6132
			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;
			}
6133
			/* We cannot be sure it is safe to start an in-place
6134
			 * reshape.  It is only safe if user-space is monitoring
6135 6136 6137 6138 6139
			 * 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.
			 */
6140 6141 6142 6143 6144 6145 6146
			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",
6147
				       mdname(mddev));
6148 6149
				return -EINVAL;
			}
6150
		} else if (mddev->reshape_backwards
6151
		    ? (here_new * mddev->new_chunk_sectors + min_offset_diff <=
6152 6153
		       here_old * mddev->chunk_sectors)
		    : (here_new * mddev->new_chunk_sectors >=
6154
		       here_old * mddev->chunk_sectors + (-min_offset_diff))) {
N
NeilBrown 已提交
6155
			/* Reading from the same stripe as writing to - bad */
6156 6157 6158
			printk(KERN_ERR "md/raid:%s: reshape_position too early for "
			       "auto-recovery - aborting.\n",
			       mdname(mddev));
N
NeilBrown 已提交
6159 6160
			return -EINVAL;
		}
6161 6162
		printk(KERN_INFO "md/raid:%s: reshape will continue\n",
		       mdname(mddev));
N
NeilBrown 已提交
6163 6164 6165 6166
		/* OK, we should be able to continue; */
	} else {
		BUG_ON(mddev->level != mddev->new_level);
		BUG_ON(mddev->layout != mddev->new_layout);
6167
		BUG_ON(mddev->chunk_sectors != mddev->new_chunk_sectors);
N
NeilBrown 已提交
6168
		BUG_ON(mddev->delta_disks != 0);
L
Linus Torvalds 已提交
6169
	}
N
NeilBrown 已提交
6170

6171 6172 6173 6174 6175
	if (mddev->private == NULL)
		conf = setup_conf(mddev);
	else
		conf = mddev->private;

N
NeilBrown 已提交
6176 6177 6178
	if (IS_ERR(conf))
		return PTR_ERR(conf);

6179
	conf->min_offset_diff = min_offset_diff;
N
NeilBrown 已提交
6180 6181 6182 6183
	mddev->thread = conf->thread;
	conf->thread = NULL;
	mddev->private = conf;

6184 6185 6186 6187 6188 6189 6190 6191 6192 6193 6194
	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)
6195
			continue;
6196 6197 6198 6199 6200 6201 6202
		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;
		}
6203
		if (test_bit(In_sync, &rdev->flags)) {
N
NeilBrown 已提交
6204
			working_disks++;
6205 6206
			continue;
		}
6207 6208 6209 6210 6211 6212 6213 6214 6215 6216 6217 6218
		/* 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;
6219

6220 6221 6222 6223 6224 6225 6226 6227 6228 6229 6230 6231 6232 6233 6234
		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
6235

6236 6237 6238
	/*
	 * 0 for a fully functional array, 1 or 2 for a degraded array.
	 */
6239
	mddev->degraded = calc_degraded(conf);
N
NeilBrown 已提交
6240

6241
	if (has_failed(conf)) {
6242
		printk(KERN_ERR "md/raid:%s: not enough operational devices"
L
Linus Torvalds 已提交
6243
			" (%d/%d failed)\n",
6244
			mdname(mddev), mddev->degraded, conf->raid_disks);
L
Linus Torvalds 已提交
6245 6246 6247
		goto abort;
	}

N
NeilBrown 已提交
6248
	/* device size must be a multiple of chunk size */
6249
	mddev->dev_sectors &= ~(mddev->chunk_sectors - 1);
N
NeilBrown 已提交
6250 6251
	mddev->resync_max_sectors = mddev->dev_sectors;

6252
	if (mddev->degraded > dirty_parity_disks &&
L
Linus Torvalds 已提交
6253
	    mddev->recovery_cp != MaxSector) {
6254 6255
		if (mddev->ok_start_degraded)
			printk(KERN_WARNING
6256 6257
			       "md/raid:%s: starting dirty degraded array"
			       " - data corruption possible.\n",
6258 6259 6260
			       mdname(mddev));
		else {
			printk(KERN_ERR
6261
			       "md/raid:%s: cannot start dirty degraded array.\n",
6262 6263 6264
			       mdname(mddev));
			goto abort;
		}
L
Linus Torvalds 已提交
6265 6266 6267
	}

	if (mddev->degraded == 0)
6268 6269
		printk(KERN_INFO "md/raid:%s: raid level %d active with %d out of %d"
		       " devices, algorithm %d\n", mdname(mddev), conf->level,
6270 6271
		       mddev->raid_disks-mddev->degraded, mddev->raid_disks,
		       mddev->new_layout);
L
Linus Torvalds 已提交
6272
	else
6273 6274 6275 6276 6277
		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
6278 6279 6280

	print_raid5_conf(conf);

6281 6282
	if (conf->reshape_progress != MaxSector) {
		conf->reshape_safe = conf->reshape_progress;
6283 6284 6285 6286 6287 6288
		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,
6289
							"reshape");
6290 6291
	}

L
Linus Torvalds 已提交
6292
	/* Ok, everything is just fine now */
6293 6294
	if (mddev->to_remove == &raid5_attrs_group)
		mddev->to_remove = NULL;
N
NeilBrown 已提交
6295 6296
	else if (mddev->kobj.sd &&
	    sysfs_create_group(&mddev->kobj, &raid5_attrs_group))
6297
		printk(KERN_WARNING
6298
		       "raid5: failed to create sysfs attributes for %s\n",
6299
		       mdname(mddev));
6300
	md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
6301

6302
	if (mddev->queue) {
6303
		int chunk_size;
S
Shaohua Li 已提交
6304
		bool discard_supported = true;
6305 6306 6307 6308 6309 6310 6311 6312 6313
		/* 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
6314

6315 6316 6317 6318
		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));
6319
		mddev->queue->limits.raid_partial_stripes_expensive = 1;
S
Shaohua Li 已提交
6320 6321 6322 6323 6324
		/*
		 * We can only discard a whole stripe. It doesn't make sense to
		 * discard data disk but write parity disk
		 */
		stripe = stripe * PAGE_SIZE;
6325 6326 6327 6328
		/* 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
6329 6330 6331 6332
		mddev->queue->limits.discard_alignment = stripe;
		mddev->queue->limits.discard_granularity = stripe;
		/*
		 * unaligned part of discard request will be ignored, so can't
6333
		 * guarantee discard_zeroes_data
S
Shaohua Li 已提交
6334 6335
		 */
		mddev->queue->limits.discard_zeroes_data = 0;
6336

6337 6338
		blk_queue_max_write_same_sectors(mddev->queue, 0);

6339
		rdev_for_each(rdev, mddev) {
6340 6341
			disk_stack_limits(mddev->gendisk, rdev->bdev,
					  rdev->data_offset << 9);
6342 6343
			disk_stack_limits(mddev->gendisk, rdev->bdev,
					  rdev->new_data_offset << 9);
S
Shaohua Li 已提交
6344 6345 6346 6347 6348 6349 6350 6351 6352 6353 6354 6355 6356 6357
			/*
			 * 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;
6358 6359 6360 6361 6362 6363 6364 6365 6366 6367 6368 6369
			/* 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;
			}
6370
		}
S
Shaohua Li 已提交
6371 6372 6373 6374 6375 6376 6377 6378 6379

		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);
6380
	}
6381

L
Linus Torvalds 已提交
6382 6383
	return 0;
abort:
6384
	md_unregister_thread(&mddev->thread);
N
NeilBrown 已提交
6385 6386
	print_raid5_conf(conf);
	free_conf(conf);
L
Linus Torvalds 已提交
6387
	mddev->private = NULL;
6388
	printk(KERN_ALERT "md/raid:%s: failed to run raid set.\n", mdname(mddev));
L
Linus Torvalds 已提交
6389 6390 6391
	return -EIO;
}

N
NeilBrown 已提交
6392
static void raid5_free(struct mddev *mddev, void *priv)
L
Linus Torvalds 已提交
6393
{
N
NeilBrown 已提交
6394
	struct r5conf *conf = priv;
L
Linus Torvalds 已提交
6395

6396
	free_conf(conf);
6397
	mddev->to_remove = &raid5_attrs_group;
L
Linus Torvalds 已提交
6398 6399
}

6400
static void status(struct seq_file *seq, struct mddev *mddev)
L
Linus Torvalds 已提交
6401
{
6402
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
6403 6404
	int i;

6405 6406
	seq_printf(seq, " level %d, %dk chunk, algorithm %d", mddev->level,
		mddev->chunk_sectors / 2, mddev->layout);
6407
	seq_printf (seq, " [%d/%d] [", conf->raid_disks, conf->raid_disks - mddev->degraded);
L
Linus Torvalds 已提交
6408 6409 6410
	for (i = 0; i < conf->raid_disks; i++)
		seq_printf (seq, "%s",
			       conf->disks[i].rdev &&
6411
			       test_bit(In_sync, &conf->disks[i].rdev->flags) ? "U" : "_");
L
Linus Torvalds 已提交
6412 6413 6414
	seq_printf (seq, "]");
}

6415
static void print_raid5_conf (struct r5conf *conf)
L
Linus Torvalds 已提交
6416 6417 6418 6419
{
	int i;
	struct disk_info *tmp;

6420
	printk(KERN_DEBUG "RAID conf printout:\n");
L
Linus Torvalds 已提交
6421 6422 6423 6424
	if (!conf) {
		printk("(conf==NULL)\n");
		return;
	}
6425 6426 6427
	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
6428 6429 6430 6431 6432

	for (i = 0; i < conf->raid_disks; i++) {
		char b[BDEVNAME_SIZE];
		tmp = conf->disks + i;
		if (tmp->rdev)
6433 6434 6435
			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
6436 6437 6438
	}
}

6439
static int raid5_spare_active(struct mddev *mddev)
L
Linus Torvalds 已提交
6440 6441
{
	int i;
6442
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
6443
	struct disk_info *tmp;
6444 6445
	int count = 0;
	unsigned long flags;
L
Linus Torvalds 已提交
6446 6447 6448

	for (i = 0; i < conf->raid_disks; i++) {
		tmp = conf->disks + i;
6449 6450 6451 6452 6453 6454 6455 6456 6457 6458 6459 6460 6461 6462 6463 6464 6465 6466 6467
		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
6468
		    && tmp->rdev->recovery_offset == MaxSector
6469
		    && !test_bit(Faulty, &tmp->rdev->flags)
6470
		    && !test_and_set_bit(In_sync, &tmp->rdev->flags)) {
6471
			count++;
6472
			sysfs_notify_dirent_safe(tmp->rdev->sysfs_state);
L
Linus Torvalds 已提交
6473 6474
		}
	}
6475
	spin_lock_irqsave(&conf->device_lock, flags);
6476
	mddev->degraded = calc_degraded(conf);
6477
	spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
6478
	print_raid5_conf(conf);
6479
	return count;
L
Linus Torvalds 已提交
6480 6481
}

6482
static int raid5_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
6483
{
6484
	struct r5conf *conf = mddev->private;
L
Linus Torvalds 已提交
6485
	int err = 0;
6486
	int number = rdev->raid_disk;
6487
	struct md_rdev **rdevp;
L
Linus Torvalds 已提交
6488 6489 6490
	struct disk_info *p = conf->disks + number;

	print_raid5_conf(conf);
6491 6492 6493 6494 6495 6496 6497 6498 6499 6500 6501 6502 6503 6504 6505 6506 6507 6508 6509 6510 6511 6512
	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) &&
6513
	    (!p->replacement || p->replacement == rdev) &&
6514 6515 6516 6517 6518 6519 6520 6521 6522 6523
	    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;
6524 6525 6526 6527 6528 6529 6530 6531 6532 6533 6534 6535 6536 6537
	} 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
6538 6539 6540 6541 6542 6543
abort:

	print_raid5_conf(conf);
	return err;
}

6544
static int raid5_add_disk(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
6545
{
6546
	struct r5conf *conf = mddev->private;
6547
	int err = -EEXIST;
L
Linus Torvalds 已提交
6548 6549
	int disk;
	struct disk_info *p;
6550 6551
	int first = 0;
	int last = conf->raid_disks - 1;
L
Linus Torvalds 已提交
6552

6553 6554 6555
	if (mddev->recovery_disabled == conf->recovery_disabled)
		return -EBUSY;

N
NeilBrown 已提交
6556
	if (rdev->saved_raid_disk < 0 && has_failed(conf))
L
Linus Torvalds 已提交
6557
		/* no point adding a device */
6558
		return -EINVAL;
L
Linus Torvalds 已提交
6559

6560 6561
	if (rdev->raid_disk >= 0)
		first = last = rdev->raid_disk;
L
Linus Torvalds 已提交
6562 6563

	/*
6564 6565
	 * find the disk ... but prefer rdev->saved_raid_disk
	 * if possible.
L
Linus Torvalds 已提交
6566
	 */
6567
	if (rdev->saved_raid_disk >= 0 &&
6568
	    rdev->saved_raid_disk >= first &&
6569
	    conf->disks[rdev->saved_raid_disk].rdev == NULL)
6570 6571 6572
		first = rdev->saved_raid_disk;

	for (disk = first; disk <= last; disk++) {
6573 6574
		p = conf->disks + disk;
		if (p->rdev == NULL) {
6575
			clear_bit(In_sync, &rdev->flags);
L
Linus Torvalds 已提交
6576
			rdev->raid_disk = disk;
6577
			err = 0;
6578 6579
			if (rdev->saved_raid_disk != disk)
				conf->fullsync = 1;
6580
			rcu_assign_pointer(p->rdev, rdev);
6581
			goto out;
L
Linus Torvalds 已提交
6582
		}
6583 6584 6585
	}
	for (disk = first; disk <= last; disk++) {
		p = conf->disks + disk;
6586 6587 6588 6589 6590 6591 6592 6593 6594 6595 6596
		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;
		}
	}
6597
out:
L
Linus Torvalds 已提交
6598
	print_raid5_conf(conf);
6599
	return err;
L
Linus Torvalds 已提交
6600 6601
}

6602
static int raid5_resize(struct mddev *mddev, sector_t sectors)
L
Linus Torvalds 已提交
6603 6604 6605 6606 6607 6608 6609 6610
{
	/* 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.
	 */
6611
	sector_t newsize;
6612
	sectors &= ~((sector_t)mddev->chunk_sectors - 1);
6613 6614 6615
	newsize = raid5_size(mddev, sectors, mddev->raid_disks);
	if (mddev->external_size &&
	    mddev->array_sectors > newsize)
D
Dan Williams 已提交
6616
		return -EINVAL;
6617 6618 6619 6620 6621 6622
	if (mddev->bitmap) {
		int ret = bitmap_resize(mddev->bitmap, sectors, 0, 0);
		if (ret)
			return ret;
	}
	md_set_array_sectors(mddev, newsize);
6623
	set_capacity(mddev->gendisk, mddev->array_sectors);
6624
	revalidate_disk(mddev->gendisk);
6625 6626
	if (sectors > mddev->dev_sectors &&
	    mddev->recovery_cp > mddev->dev_sectors) {
A
Andre Noll 已提交
6627
		mddev->recovery_cp = mddev->dev_sectors;
L
Linus Torvalds 已提交
6628 6629
		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
	}
A
Andre Noll 已提交
6630
	mddev->dev_sectors = sectors;
6631
	mddev->resync_max_sectors = sectors;
L
Linus Torvalds 已提交
6632 6633 6634
	return 0;
}

6635
static int check_stripe_cache(struct mddev *mddev)
6636 6637 6638 6639 6640 6641 6642 6643 6644
{
	/* 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.
	 */
6645
	struct r5conf *conf = mddev->private;
6646 6647 6648 6649
	if (((mddev->chunk_sectors << 9) / STRIPE_SIZE) * 4
	    > conf->max_nr_stripes ||
	    ((mddev->new_chunk_sectors << 9) / STRIPE_SIZE) * 4
	    > conf->max_nr_stripes) {
6650 6651
		printk(KERN_WARNING "md/raid:%s: reshape: not enough stripes.  Needed %lu\n",
		       mdname(mddev),
6652 6653 6654 6655 6656 6657 6658
		       ((max(mddev->chunk_sectors, mddev->new_chunk_sectors) << 9)
			/ STRIPE_SIZE)*4);
		return 0;
	}
	return 1;
}

6659
static int check_reshape(struct mddev *mddev)
6660
{
6661
	struct r5conf *conf = mddev->private;
6662

6663 6664
	if (mddev->delta_disks == 0 &&
	    mddev->new_layout == mddev->layout &&
6665
	    mddev->new_chunk_sectors == mddev->chunk_sectors)
6666
		return 0; /* nothing to do */
6667
	if (has_failed(conf))
6668
		return -EINVAL;
6669
	if (mddev->delta_disks < 0 && mddev->reshape_position == MaxSector) {
6670 6671 6672 6673 6674 6675 6676 6677 6678 6679 6680
		/* 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;
	}
6681

6682
	if (!check_stripe_cache(mddev))
6683 6684
		return -ENOSPC;

6685 6686
	return resize_stripes(conf, (conf->previous_raid_disks
				     + mddev->delta_disks));
6687 6688
}

6689
static int raid5_start_reshape(struct mddev *mddev)
6690
{
6691
	struct r5conf *conf = mddev->private;
6692
	struct md_rdev *rdev;
6693
	int spares = 0;
6694
	unsigned long flags;
6695

6696
	if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
6697 6698
		return -EBUSY;

6699 6700 6701
	if (!check_stripe_cache(mddev))
		return -ENOSPC;

6702 6703 6704
	if (has_failed(conf))
		return -EINVAL;

6705
	rdev_for_each(rdev, mddev) {
6706 6707
		if (!test_bit(In_sync, &rdev->flags)
		    && !test_bit(Faulty, &rdev->flags))
6708
			spares++;
6709
	}
6710

6711
	if (spares - mddev->degraded < mddev->delta_disks - conf->max_degraded)
6712 6713 6714 6715 6716
		/* Not enough devices even to make a degraded array
		 * of that size
		 */
		return -EINVAL;

6717 6718 6719 6720 6721 6722
	/* 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) {
6723
		printk(KERN_ERR "md/raid:%s: array size must be reduced "
6724 6725 6726 6727
		       "before number of disks\n", mdname(mddev));
		return -EINVAL;
	}

6728
	atomic_set(&conf->reshape_stripes, 0);
6729
	spin_lock_irq(&conf->device_lock);
6730
	write_seqcount_begin(&conf->gen_lock);
6731
	conf->previous_raid_disks = conf->raid_disks;
6732
	conf->raid_disks += mddev->delta_disks;
6733 6734
	conf->prev_chunk_sectors = conf->chunk_sectors;
	conf->chunk_sectors = mddev->new_chunk_sectors;
6735 6736
	conf->prev_algo = conf->algorithm;
	conf->algorithm = mddev->new_layout;
6737 6738 6739 6740 6741
	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();
6742
	if (mddev->reshape_backwards)
6743 6744 6745 6746
		conf->reshape_progress = raid5_size(mddev, 0, 0);
	else
		conf->reshape_progress = 0;
	conf->reshape_safe = conf->reshape_progress;
6747
	write_seqcount_end(&conf->gen_lock);
6748 6749
	spin_unlock_irq(&conf->device_lock);

6750 6751 6752 6753 6754 6755 6756
	/* 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);

6757 6758
	/* Add some new drives, as many as will fit.
	 * We know there are enough to make the newly sized array work.
6759 6760 6761 6762
	 * 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.
6763
	 */
6764
	if (mddev->delta_disks >= 0) {
N
NeilBrown 已提交
6765
		rdev_for_each(rdev, mddev)
6766 6767 6768 6769
			if (rdev->raid_disk < 0 &&
			    !test_bit(Faulty, &rdev->flags)) {
				if (raid5_add_disk(mddev, rdev) == 0) {
					if (rdev->raid_disk
6770
					    >= conf->previous_raid_disks)
6771
						set_bit(In_sync, &rdev->flags);
6772
					else
6773
						rdev->recovery_offset = 0;
6774 6775

					if (sysfs_link_rdev(mddev, rdev))
6776
						/* Failure here is OK */;
6777
				}
6778 6779 6780 6781 6782
			} 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);
			}
6783

6784 6785 6786 6787
		/* When a reshape changes the number of devices,
		 * ->degraded is measured against the larger of the
		 * pre and post number of devices.
		 */
6788
		spin_lock_irqsave(&conf->device_lock, flags);
6789
		mddev->degraded = calc_degraded(conf);
6790 6791
		spin_unlock_irqrestore(&conf->device_lock, flags);
	}
6792
	mddev->raid_disks = conf->raid_disks;
6793
	mddev->reshape_position = conf->reshape_progress;
6794
	set_bit(MD_CHANGE_DEVS, &mddev->flags);
6795

6796 6797 6798 6799 6800
	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,
6801
						"reshape");
6802 6803 6804
	if (!mddev->sync_thread) {
		mddev->recovery = 0;
		spin_lock_irq(&conf->device_lock);
6805
		write_seqcount_begin(&conf->gen_lock);
6806
		mddev->raid_disks = conf->raid_disks = conf->previous_raid_disks;
6807 6808 6809
		mddev->new_chunk_sectors =
			conf->chunk_sectors = conf->prev_chunk_sectors;
		mddev->new_layout = conf->algorithm = conf->prev_algo;
6810 6811 6812
		rdev_for_each(rdev, mddev)
			rdev->new_data_offset = rdev->data_offset;
		smp_wmb();
6813
		conf->generation --;
6814
		conf->reshape_progress = MaxSector;
6815
		mddev->reshape_position = MaxSector;
6816
		write_seqcount_end(&conf->gen_lock);
6817 6818 6819
		spin_unlock_irq(&conf->device_lock);
		return -EAGAIN;
	}
6820
	conf->reshape_checkpoint = jiffies;
6821 6822 6823 6824 6825
	md_wakeup_thread(mddev->sync_thread);
	md_new_event(mddev);
	return 0;
}

6826 6827 6828
/* This is called from the reshape thread and should make any
 * changes needed in 'conf'
 */
6829
static void end_reshape(struct r5conf *conf)
6830 6831
{

6832
	if (!test_bit(MD_RECOVERY_INTR, &conf->mddev->recovery)) {
6833
		struct md_rdev *rdev;
6834 6835

		spin_lock_irq(&conf->device_lock);
6836
		conf->previous_raid_disks = conf->raid_disks;
6837 6838 6839
		rdev_for_each(rdev, conf->mddev)
			rdev->data_offset = rdev->new_data_offset;
		smp_wmb();
6840
		conf->reshape_progress = MaxSector;
6841
		spin_unlock_irq(&conf->device_lock);
6842
		wake_up(&conf->wait_for_overlap);
6843 6844 6845 6846

		/* read-ahead size must cover two whole stripes, which is
		 * 2 * (datadisks) * chunksize where 'n' is the number of raid devices
		 */
6847
		if (conf->mddev->queue) {
6848
			int data_disks = conf->raid_disks - conf->max_degraded;
6849
			int stripe = data_disks * ((conf->chunk_sectors << 9)
6850
						   / PAGE_SIZE);
6851 6852 6853
			if (conf->mddev->queue->backing_dev_info.ra_pages < 2 * stripe)
				conf->mddev->queue->backing_dev_info.ra_pages = 2 * stripe;
		}
6854 6855 6856
	}
}

6857 6858 6859
/* This is called from the raid5d thread with mddev_lock held.
 * It makes config changes to the device.
 */
6860
static void raid5_finish_reshape(struct mddev *mddev)
6861
{
6862
	struct r5conf *conf = mddev->private;
6863 6864 6865

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

6866 6867 6868
		if (mddev->delta_disks > 0) {
			md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
			set_capacity(mddev->gendisk, mddev->array_sectors);
6869
			revalidate_disk(mddev->gendisk);
6870 6871
		} else {
			int d;
6872 6873 6874
			spin_lock_irq(&conf->device_lock);
			mddev->degraded = calc_degraded(conf);
			spin_unlock_irq(&conf->device_lock);
6875 6876
			for (d = conf->raid_disks ;
			     d < conf->raid_disks - mddev->delta_disks;
6877
			     d++) {
6878
				struct md_rdev *rdev = conf->disks[d].rdev;
6879 6880 6881 6882 6883
				if (rdev)
					clear_bit(In_sync, &rdev->flags);
				rdev = conf->disks[d].replacement;
				if (rdev)
					clear_bit(In_sync, &rdev->flags);
6884
			}
6885
		}
6886
		mddev->layout = conf->algorithm;
6887
		mddev->chunk_sectors = conf->chunk_sectors;
6888 6889
		mddev->reshape_position = MaxSector;
		mddev->delta_disks = 0;
6890
		mddev->reshape_backwards = 0;
6891 6892 6893
	}
}

6894
static void raid5_quiesce(struct mddev *mddev, int state)
6895
{
6896
	struct r5conf *conf = mddev->private;
6897 6898

	switch(state) {
6899 6900 6901 6902
	case 2: /* resume for a suspend */
		wake_up(&conf->wait_for_overlap);
		break;

6903
	case 1: /* stop all writes */
6904
		lock_all_device_hash_locks_irq(conf);
6905 6906 6907 6908
		/* '2' tells resync/reshape to pause so that all
		 * active stripes can drain
		 */
		conf->quiesce = 2;
6909
		wait_event_cmd(conf->wait_for_stripe,
6910 6911
				    atomic_read(&conf->active_stripes) == 0 &&
				    atomic_read(&conf->active_aligned_reads) == 0,
6912 6913
				    unlock_all_device_hash_locks_irq(conf),
				    lock_all_device_hash_locks_irq(conf));
6914
		conf->quiesce = 1;
6915
		unlock_all_device_hash_locks_irq(conf);
6916 6917
		/* allow reshape to continue */
		wake_up(&conf->wait_for_overlap);
6918 6919 6920
		break;

	case 0: /* re-enable writes */
6921
		lock_all_device_hash_locks_irq(conf);
6922 6923
		conf->quiesce = 0;
		wake_up(&conf->wait_for_stripe);
6924
		wake_up(&conf->wait_for_overlap);
6925
		unlock_all_device_hash_locks_irq(conf);
6926 6927 6928
		break;
	}
}
6929

6930
static void *raid45_takeover_raid0(struct mddev *mddev, int level)
6931
{
6932
	struct r0conf *raid0_conf = mddev->private;
6933
	sector_t sectors;
6934

D
Dan Williams 已提交
6935
	/* for raid0 takeover only one zone is supported */
6936
	if (raid0_conf->nr_strip_zones > 1) {
6937 6938
		printk(KERN_ERR "md/raid:%s: cannot takeover raid0 with more than one zone.\n",
		       mdname(mddev));
D
Dan Williams 已提交
6939 6940 6941
		return ERR_PTR(-EINVAL);
	}

6942 6943
	sectors = raid0_conf->strip_zone[0].zone_end;
	sector_div(sectors, raid0_conf->strip_zone[0].nb_dev);
6944
	mddev->dev_sectors = sectors;
D
Dan Williams 已提交
6945
	mddev->new_level = level;
6946 6947 6948 6949 6950 6951 6952 6953 6954 6955
	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);
}

6956
static void *raid5_takeover_raid1(struct mddev *mddev)
6957 6958 6959 6960 6961 6962 6963 6964 6965 6966 6967 6968 6969 6970 6971 6972 6973 6974 6975 6976 6977
{
	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;
6978
	mddev->new_chunk_sectors = chunksect;
6979 6980 6981 6982

	return setup_conf(mddev);
}

6983
static void *raid5_takeover_raid6(struct mddev *mddev)
6984 6985 6986 6987 6988 6989 6990 6991 6992 6993 6994 6995 6996 6997 6998 6999 7000 7001 7002 7003 7004 7005 7006 7007 7008 7009 7010 7011 7012 7013 7014 7015
{
	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);
}

7016
static int raid5_check_reshape(struct mddev *mddev)
7017
{
7018 7019 7020 7021
	/* 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.
7022
	 */
7023
	struct r5conf *conf = mddev->private;
7024
	int new_chunk = mddev->new_chunk_sectors;
7025

7026
	if (mddev->new_layout >= 0 && !algorithm_valid_raid5(mddev->new_layout))
7027 7028
		return -EINVAL;
	if (new_chunk > 0) {
7029
		if (!is_power_of_2(new_chunk))
7030
			return -EINVAL;
7031
		if (new_chunk < (PAGE_SIZE>>9))
7032
			return -EINVAL;
7033
		if (mddev->array_sectors & (new_chunk-1))
7034 7035 7036 7037 7038 7039
			/* not factor of array size */
			return -EINVAL;
	}

	/* They look valid */

7040
	if (mddev->raid_disks == 2) {
7041 7042 7043 7044
		/* can make the change immediately */
		if (mddev->new_layout >= 0) {
			conf->algorithm = mddev->new_layout;
			mddev->layout = mddev->new_layout;
7045 7046
		}
		if (new_chunk > 0) {
7047 7048
			conf->chunk_sectors = new_chunk ;
			mddev->chunk_sectors = new_chunk;
7049 7050 7051
		}
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
		md_wakeup_thread(mddev->thread);
7052
	}
7053
	return check_reshape(mddev);
7054 7055
}

7056
static int raid6_check_reshape(struct mddev *mddev)
7057
{
7058
	int new_chunk = mddev->new_chunk_sectors;
7059

7060
	if (mddev->new_layout >= 0 && !algorithm_valid_raid6(mddev->new_layout))
7061
		return -EINVAL;
7062
	if (new_chunk > 0) {
7063
		if (!is_power_of_2(new_chunk))
7064
			return -EINVAL;
7065
		if (new_chunk < (PAGE_SIZE >> 9))
7066
			return -EINVAL;
7067
		if (mddev->array_sectors & (new_chunk-1))
7068 7069
			/* not factor of array size */
			return -EINVAL;
7070
	}
7071 7072

	/* They look valid */
7073
	return check_reshape(mddev);
7074 7075
}

7076
static void *raid5_takeover(struct mddev *mddev)
7077 7078
{
	/* raid5 can take over:
D
Dan Williams 已提交
7079
	 *  raid0 - if there is only one strip zone - make it a raid4 layout
7080 7081 7082 7083
	 *  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
7084 7085
	if (mddev->level == 0)
		return raid45_takeover_raid0(mddev, 5);
7086 7087
	if (mddev->level == 1)
		return raid5_takeover_raid1(mddev);
7088 7089 7090 7091 7092
	if (mddev->level == 4) {
		mddev->new_layout = ALGORITHM_PARITY_N;
		mddev->new_level = 5;
		return setup_conf(mddev);
	}
7093 7094
	if (mddev->level == 6)
		return raid5_takeover_raid6(mddev);
7095 7096 7097 7098

	return ERR_PTR(-EINVAL);
}

7099
static void *raid4_takeover(struct mddev *mddev)
7100
{
D
Dan Williams 已提交
7101 7102 7103
	/* raid4 can take over:
	 *  raid0 - if there is only one strip zone
	 *  raid5 - if layout is right
7104
	 */
D
Dan Williams 已提交
7105 7106
	if (mddev->level == 0)
		return raid45_takeover_raid0(mddev, 4);
7107 7108 7109 7110 7111 7112 7113 7114
	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);
}
7115

7116
static struct md_personality raid5_personality;
7117

7118
static void *raid6_takeover(struct mddev *mddev)
7119 7120 7121 7122 7123 7124 7125 7126 7127 7128 7129 7130 7131 7132 7133 7134 7135 7136 7137 7138 7139 7140 7141 7142 7143 7144 7145 7146 7147 7148 7149 7150 7151 7152 7153 7154 7155 7156 7157 7158 7159 7160 7161 7162 7163
{
	/* 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);
}

7164
static struct md_personality raid6_personality =
7165 7166 7167 7168 7169 7170
{
	.name		= "raid6",
	.level		= 6,
	.owner		= THIS_MODULE,
	.make_request	= make_request,
	.run		= run,
N
NeilBrown 已提交
7171
	.free		= raid5_free,
7172 7173 7174 7175 7176 7177 7178
	.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,
7179
	.size		= raid5_size,
7180
	.check_reshape	= raid6_check_reshape,
7181
	.start_reshape  = raid5_start_reshape,
7182
	.finish_reshape = raid5_finish_reshape,
7183
	.quiesce	= raid5_quiesce,
7184
	.takeover	= raid6_takeover,
7185
	.congested	= raid5_congested,
7186
	.mergeable_bvec	= raid5_mergeable_bvec,
7187
};
7188
static struct md_personality raid5_personality =
L
Linus Torvalds 已提交
7189 7190
{
	.name		= "raid5",
7191
	.level		= 5,
L
Linus Torvalds 已提交
7192 7193 7194
	.owner		= THIS_MODULE,
	.make_request	= make_request,
	.run		= run,
N
NeilBrown 已提交
7195
	.free		= raid5_free,
L
Linus Torvalds 已提交
7196 7197 7198 7199 7200 7201 7202
	.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,
7203
	.size		= raid5_size,
7204 7205
	.check_reshape	= raid5_check_reshape,
	.start_reshape  = raid5_start_reshape,
7206
	.finish_reshape = raid5_finish_reshape,
7207
	.quiesce	= raid5_quiesce,
7208
	.takeover	= raid5_takeover,
7209
	.congested	= raid5_congested,
7210
	.mergeable_bvec	= raid5_mergeable_bvec,
L
Linus Torvalds 已提交
7211 7212
};

7213
static struct md_personality raid4_personality =
L
Linus Torvalds 已提交
7214
{
7215 7216 7217 7218 7219
	.name		= "raid4",
	.level		= 4,
	.owner		= THIS_MODULE,
	.make_request	= make_request,
	.run		= run,
N
NeilBrown 已提交
7220
	.free		= raid5_free,
7221 7222 7223 7224 7225 7226 7227
	.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,
7228
	.size		= raid5_size,
7229 7230
	.check_reshape	= raid5_check_reshape,
	.start_reshape  = raid5_start_reshape,
7231
	.finish_reshape = raid5_finish_reshape,
7232
	.quiesce	= raid5_quiesce,
7233
	.takeover	= raid4_takeover,
7234
	.congested	= raid5_congested,
7235
	.mergeable_bvec	= raid5_mergeable_bvec,
7236 7237 7238 7239
};

static int __init raid5_init(void)
{
7240 7241 7242 7243
	raid5_wq = alloc_workqueue("raid5wq",
		WQ_UNBOUND|WQ_MEM_RECLAIM|WQ_CPU_INTENSIVE|WQ_SYSFS, 0);
	if (!raid5_wq)
		return -ENOMEM;
7244
	register_md_personality(&raid6_personality);
7245 7246 7247
	register_md_personality(&raid5_personality);
	register_md_personality(&raid4_personality);
	return 0;
L
Linus Torvalds 已提交
7248 7249
}

7250
static void raid5_exit(void)
L
Linus Torvalds 已提交
7251
{
7252
	unregister_md_personality(&raid6_personality);
7253 7254
	unregister_md_personality(&raid5_personality);
	unregister_md_personality(&raid4_personality);
7255
	destroy_workqueue(raid5_wq);
L
Linus Torvalds 已提交
7256 7257 7258 7259 7260
}

module_init(raid5_init);
module_exit(raid5_exit);
MODULE_LICENSE("GPL");
7261
MODULE_DESCRIPTION("RAID4/5/6 (striping with parity) personality for MD");
L
Linus Torvalds 已提交
7262
MODULE_ALIAS("md-personality-4"); /* RAID5 */
7263 7264
MODULE_ALIAS("md-raid5");
MODULE_ALIAS("md-raid4");
7265 7266
MODULE_ALIAS("md-level-5");
MODULE_ALIAS("md-level-4");
7267 7268 7269 7270 7271 7272 7273
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");