raid5.c 149.0 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 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44
/*
 * 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.
 * conf->bm_write is the number of the last batch successfully written.
 * conf->bm_flush is the number of the last batch that was closed to
 *    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
 * the number of the batch it will be in. This is bm_flush+1.
 * 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/seq_file.h>
51
#include "md.h"
52
#include "raid5.h"
53
#include "bitmap.h"
54

L
Linus Torvalds 已提交
55 56 57 58 59 60 61 62 63
/*
 * 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
64
#define BYPASS_THRESHOLD	1
65
#define NR_HASH			(PAGE_SIZE / sizeof(struct hlist_head))
L
Linus Torvalds 已提交
66 67
#define HASH_MASK		(NR_HASH - 1)

68
#define stripe_hash(conf, sect)	(&((conf)->stripe_hashtbl[((sect) >> STRIPE_SHIFT) & HASH_MASK]))
L
Linus Torvalds 已提交
69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89

/* 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.
 * This macro is used to determine the 'next' bio in the list, given the sector
 * of the current stripe+device
 */
#define r5_next_bio(bio, sect) ( ( (bio)->bi_sector + ((bio)->bi_size>>9) < sect + STRIPE_SECTORS) ? (bio)->bi_next : NULL)
/*
 * The following can be used to debug the driver
 */
#define RAID5_PARANOIA	1
#if RAID5_PARANOIA && defined(CONFIG_SMP)
# define CHECK_DEVLOCK() assert_spin_locked(&conf->device_lock)
#else
# define CHECK_DEVLOCK()
#endif

90
#ifdef DEBUG
L
Linus Torvalds 已提交
91 92 93 94
#define inline
#define __inline__
#endif

95 96
#define printk_rl(args...) ((void) (printk_ratelimit() && printk(args)))

97
/*
98 99
 * 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
100 101 102
 */
static inline int raid5_bi_phys_segments(struct bio *bio)
{
103
	return bio->bi_phys_segments & 0xffff;
104 105 106 107
}

static inline int raid5_bi_hw_segments(struct bio *bio)
{
108
	return (bio->bi_phys_segments >> 16) & 0xffff;
109 110 111 112 113 114 115 116 117 118 119 120 121
}

static inline int raid5_dec_bi_phys_segments(struct bio *bio)
{
	--bio->bi_phys_segments;
	return raid5_bi_phys_segments(bio);
}

static inline int raid5_dec_bi_hw_segments(struct bio *bio)
{
	unsigned short val = raid5_bi_hw_segments(bio);

	--val;
122
	bio->bi_phys_segments = (val << 16) | raid5_bi_phys_segments(bio);
123 124 125 126 127
	return val;
}

static inline void raid5_set_bi_hw_segments(struct bio *bio, unsigned int cnt)
{
128
	bio->bi_phys_segments = raid5_bi_phys_segments(bio) || (cnt << 16);
129 130
}

131 132 133
/* Find first data disk in a raid6 stripe */
static inline int raid6_d0(struct stripe_head *sh)
{
134 135 136 137
	if (sh->ddf_layout)
		/* ddf always start from first device */
		return 0;
	/* md starts just after Q block */
138 139 140 141 142
	if (sh->qd_idx == sh->disks - 1)
		return 0;
	else
		return sh->qd_idx + 1;
}
143 144 145 146 147
static inline int raid6_next_disk(int disk, int raid_disks)
{
	disk++;
	return (disk < raid_disks) ? disk : 0;
}
148

149 150 151 152 153
/* 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.
 */
154 155
static int raid6_idx_to_slot(int idx, struct stripe_head *sh,
			     int *count, int syndrome_disks)
156 157
{
	int slot;
158

159
	if (idx == sh->pd_idx)
160
		return syndrome_disks;
161
	if (idx == sh->qd_idx)
162
		return syndrome_disks + 1;
163 164 165 166
	slot = (*count)++;
	return slot;
}

167 168 169 170 171 172 173 174
static void return_io(struct bio *return_bi)
{
	struct bio *bi = return_bi;
	while (bi) {

		return_bi = bi->bi_next;
		bi->bi_next = NULL;
		bi->bi_size = 0;
175
		bio_endio(bi, 0);
176 177 178 179
		bi = return_bi;
	}
}

L
Linus Torvalds 已提交
180 181
static void print_raid5_conf (raid5_conf_t *conf);

182 183 184 185 186 187 188
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);
}

189
static void __release_stripe(raid5_conf_t *conf, struct stripe_head *sh)
L
Linus Torvalds 已提交
190 191
{
	if (atomic_dec_and_test(&sh->count)) {
192 193
		BUG_ON(!list_empty(&sh->lru));
		BUG_ON(atomic_read(&conf->active_stripes)==0);
L
Linus Torvalds 已提交
194
		if (test_bit(STRIPE_HANDLE, &sh->state)) {
195
			if (test_bit(STRIPE_DELAYED, &sh->state)) {
L
Linus Torvalds 已提交
196
				list_add_tail(&sh->lru, &conf->delayed_list);
197 198
				blk_plug_device(conf->mddev->queue);
			} else if (test_bit(STRIPE_BIT_DELAY, &sh->state) &&
199
				   sh->bm_seq - conf->seq_write > 0) {
200
				list_add_tail(&sh->lru, &conf->bitmap_list);
201 202
				blk_plug_device(conf->mddev->queue);
			} else {
203
				clear_bit(STRIPE_BIT_DELAY, &sh->state);
L
Linus Torvalds 已提交
204
				list_add_tail(&sh->lru, &conf->handle_list);
205
			}
L
Linus Torvalds 已提交
206 207
			md_wakeup_thread(conf->mddev->thread);
		} else {
208
			BUG_ON(stripe_operations_active(sh));
L
Linus Torvalds 已提交
209 210 211 212 213 214
			if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
				atomic_dec(&conf->preread_active_stripes);
				if (atomic_read(&conf->preread_active_stripes) < IO_THRESHOLD)
					md_wakeup_thread(conf->mddev->thread);
			}
			atomic_dec(&conf->active_stripes);
215 216
			if (!test_bit(STRIPE_EXPANDING, &sh->state)) {
				list_add_tail(&sh->lru, &conf->inactive_list);
L
Linus Torvalds 已提交
217
				wake_up(&conf->wait_for_stripe);
218 219
				if (conf->retry_read_aligned)
					md_wakeup_thread(conf->mddev->thread);
220
			}
L
Linus Torvalds 已提交
221 222 223
		}
	}
}
224

L
Linus Torvalds 已提交
225 226 227 228
static void release_stripe(struct stripe_head *sh)
{
	raid5_conf_t *conf = sh->raid_conf;
	unsigned long flags;
229

L
Linus Torvalds 已提交
230 231 232 233 234
	spin_lock_irqsave(&conf->device_lock, flags);
	__release_stripe(conf, sh);
	spin_unlock_irqrestore(&conf->device_lock, flags);
}

235
static inline void remove_hash(struct stripe_head *sh)
L
Linus Torvalds 已提交
236
{
237 238
	pr_debug("remove_hash(), stripe %llu\n",
		(unsigned long long)sh->sector);
L
Linus Torvalds 已提交
239

240
	hlist_del_init(&sh->hash);
L
Linus Torvalds 已提交
241 242
}

243
static inline void insert_hash(raid5_conf_t *conf, struct stripe_head *sh)
L
Linus Torvalds 已提交
244
{
245
	struct hlist_head *hp = stripe_hash(conf, sh->sector);
L
Linus Torvalds 已提交
246

247 248
	pr_debug("insert_hash(), stripe %llu\n",
		(unsigned long long)sh->sector);
L
Linus Torvalds 已提交
249 250

	CHECK_DEVLOCK();
251
	hlist_add_head(&sh->hash, hp);
L
Linus Torvalds 已提交
252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282
}


/* find an idle stripe, make sure it is unhashed, and return it. */
static struct stripe_head *get_free_stripe(raid5_conf_t *conf)
{
	struct stripe_head *sh = NULL;
	struct list_head *first;

	CHECK_DEVLOCK();
	if (list_empty(&conf->inactive_list))
		goto out;
	first = conf->inactive_list.next;
	sh = list_entry(first, struct stripe_head, lru);
	list_del_init(first);
	remove_hash(sh);
	atomic_inc(&conf->active_stripes);
out:
	return sh;
}

static void shrink_buffers(struct stripe_head *sh, int num)
{
	struct page *p;
	int i;

	for (i=0; i<num ; i++) {
		p = sh->dev[i].page;
		if (!p)
			continue;
		sh->dev[i].page = NULL;
283
		put_page(p);
L
Linus Torvalds 已提交
284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301
	}
}

static int grow_buffers(struct stripe_head *sh, int num)
{
	int i;

	for (i=0; i<num; i++) {
		struct page *page;

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

302
static void raid5_build_block(struct stripe_head *sh, int i, int previous);
303 304
static void stripe_set_idx(sector_t stripe, raid5_conf_t *conf, int previous,
			    struct stripe_head *sh);
L
Linus Torvalds 已提交
305

306
static void init_stripe(struct stripe_head *sh, sector_t sector, int previous)
L
Linus Torvalds 已提交
307 308
{
	raid5_conf_t *conf = sh->raid_conf;
309
	int i;
L
Linus Torvalds 已提交
310

311 312
	BUG_ON(atomic_read(&sh->count) != 0);
	BUG_ON(test_bit(STRIPE_HANDLE, &sh->state));
313
	BUG_ON(stripe_operations_active(sh));
314

L
Linus Torvalds 已提交
315
	CHECK_DEVLOCK();
316
	pr_debug("init_stripe called, stripe %llu\n",
L
Linus Torvalds 已提交
317 318 319
		(unsigned long long)sh->sector);

	remove_hash(sh);
320

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

327 328

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

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

345 346
static struct stripe_head *__find_stripe(raid5_conf_t *conf, sector_t sector,
					 short generation)
L
Linus Torvalds 已提交
347 348
{
	struct stripe_head *sh;
349
	struct hlist_node *hn;
L
Linus Torvalds 已提交
350 351

	CHECK_DEVLOCK();
352
	pr_debug("__find_stripe, sector %llu\n", (unsigned long long)sector);
353
	hlist_for_each_entry(sh, hn, stripe_hash(conf, sector), hash)
354
		if (sh->sector == sector && sh->generation == generation)
L
Linus Torvalds 已提交
355
			return sh;
356
	pr_debug("__stripe %llu not in cache\n", (unsigned long long)sector);
L
Linus Torvalds 已提交
357 358 359 360
	return NULL;
}

static void unplug_slaves(mddev_t *mddev);
361
static void raid5_unplug_device(struct request_queue *q);
L
Linus Torvalds 已提交
362

363 364
static struct stripe_head *
get_active_stripe(raid5_conf_t *conf, sector_t sector,
365
		  int previous, int noblock, int noquiesce)
L
Linus Torvalds 已提交
366 367 368
{
	struct stripe_head *sh;

369
	pr_debug("get_stripe, sector %llu\n", (unsigned long long)sector);
L
Linus Torvalds 已提交
370 371 372 373

	spin_lock_irq(&conf->device_lock);

	do {
374
		wait_event_lock_irq(conf->wait_for_stripe,
375
				    conf->quiesce == 0 || noquiesce,
376
				    conf->device_lock, /* nothing */);
377
		sh = __find_stripe(conf, sector, conf->generation - previous);
L
Linus Torvalds 已提交
378 379 380 381 382 383 384 385 386
		if (!sh) {
			if (!conf->inactive_blocked)
				sh = get_free_stripe(conf);
			if (noblock && sh == NULL)
				break;
			if (!sh) {
				conf->inactive_blocked = 1;
				wait_event_lock_irq(conf->wait_for_stripe,
						    !list_empty(&conf->inactive_list) &&
387 388
						    (atomic_read(&conf->active_stripes)
						     < (conf->max_nr_stripes *3/4)
L
Linus Torvalds 已提交
389 390
						     || !conf->inactive_blocked),
						    conf->device_lock,
391
						    raid5_unplug_device(conf->mddev->queue)
L
Linus Torvalds 已提交
392 393 394
					);
				conf->inactive_blocked = 0;
			} else
395
				init_stripe(sh, sector, previous);
L
Linus Torvalds 已提交
396 397
		} else {
			if (atomic_read(&sh->count)) {
398 399
				BUG_ON(!list_empty(&sh->lru)
				    && !test_bit(STRIPE_EXPANDING, &sh->state));
L
Linus Torvalds 已提交
400 401 402
			} else {
				if (!test_bit(STRIPE_HANDLE, &sh->state))
					atomic_inc(&conf->active_stripes);
403 404
				if (list_empty(&sh->lru) &&
				    !test_bit(STRIPE_EXPANDING, &sh->state))
405 406
					BUG();
				list_del_init(&sh->lru);
L
Linus Torvalds 已提交
407 408 409 410 411 412 413 414 415 416 417
			}
		}
	} while (sh == NULL);

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

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

418 419 420 421
static void
raid5_end_read_request(struct bio *bi, int error);
static void
raid5_end_write_request(struct bio *bi, int error);
422

423
static void ops_run_io(struct stripe_head *sh, struct stripe_head_state *s)
424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457
{
	raid5_conf_t *conf = sh->raid_conf;
	int i, disks = sh->disks;

	might_sleep();

	for (i = disks; i--; ) {
		int rw;
		struct bio *bi;
		mdk_rdev_t *rdev;
		if (test_and_clear_bit(R5_Wantwrite, &sh->dev[i].flags))
			rw = WRITE;
		else if (test_and_clear_bit(R5_Wantread, &sh->dev[i].flags))
			rw = READ;
		else
			continue;

		bi = &sh->dev[i].req;

		bi->bi_rw = rw;
		if (rw == WRITE)
			bi->bi_end_io = raid5_end_write_request;
		else
			bi->bi_end_io = raid5_end_read_request;

		rcu_read_lock();
		rdev = rcu_dereference(conf->disks[i].rdev);
		if (rdev && test_bit(Faulty, &rdev->flags))
			rdev = NULL;
		if (rdev)
			atomic_inc(&rdev->nr_pending);
		rcu_read_unlock();

		if (rdev) {
458
			if (s->syncing || s->expanding || s->expanded)
459 460
				md_sync_acct(rdev->bdev, STRIPE_SECTORS);

D
Dan Williams 已提交
461 462
			set_bit(STRIPE_IO_STARTED, &sh->state);

463 464
			bi->bi_bdev = rdev->bdev;
			pr_debug("%s: for %llu schedule op %ld on disc %d\n",
465
				__func__, (unsigned long long)sh->sector,
466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528
				bi->bi_rw, i);
			atomic_inc(&sh->count);
			bi->bi_sector = sh->sector + rdev->data_offset;
			bi->bi_flags = 1 << BIO_UPTODATE;
			bi->bi_vcnt = 1;
			bi->bi_max_vecs = 1;
			bi->bi_idx = 0;
			bi->bi_io_vec = &sh->dev[i].vec;
			bi->bi_io_vec[0].bv_len = STRIPE_SIZE;
			bi->bi_io_vec[0].bv_offset = 0;
			bi->bi_size = STRIPE_SIZE;
			bi->bi_next = NULL;
			if (rw == WRITE &&
			    test_bit(R5_ReWrite, &sh->dev[i].flags))
				atomic_add(STRIPE_SECTORS,
					&rdev->corrected_errors);
			generic_make_request(bi);
		} else {
			if (rw == WRITE)
				set_bit(STRIPE_DEGRADED, &sh->state);
			pr_debug("skip op %ld on disc %d for sector %llu\n",
				bi->bi_rw, i, (unsigned long long)sh->sector);
			clear_bit(R5_LOCKED, &sh->dev[i].flags);
			set_bit(STRIPE_HANDLE, &sh->state);
		}
	}
}

static struct dma_async_tx_descriptor *
async_copy_data(int frombio, struct bio *bio, struct page *page,
	sector_t sector, struct dma_async_tx_descriptor *tx)
{
	struct bio_vec *bvl;
	struct page *bio_page;
	int i;
	int page_offset;

	if (bio->bi_sector >= sector)
		page_offset = (signed)(bio->bi_sector - sector) * 512;
	else
		page_offset = (signed)(sector - bio->bi_sector) * -512;
	bio_for_each_segment(bvl, bio, i) {
		int len = bio_iovec_idx(bio, i)->bv_len;
		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) {
			b_offset += bio_iovec_idx(bio, i)->bv_offset;
			bio_page = bio_iovec_idx(bio, i)->bv_page;
			if (frombio)
				tx = async_memcpy(page, bio_page, page_offset,
					b_offset, clen,
529
					ASYNC_TX_DEP_ACK,
530 531 532 533
					tx, NULL, NULL);
			else
				tx = async_memcpy(bio_page, page, b_offset,
					page_offset, clen,
534
					ASYNC_TX_DEP_ACK,
535 536 537 538 539 540 541 542 543 544 545 546 547 548 549
					tx, NULL, NULL);
		}
		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;
	raid5_conf_t *conf = sh->raid_conf;
550
	int i;
551

552
	pr_debug("%s: stripe %llu\n", __func__,
553 554 555
		(unsigned long long)sh->sector);

	/* clear completed biofills */
556
	spin_lock_irq(&conf->device_lock);
557 558 559 560
	for (i = sh->disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];

		/* acknowledge completion of a biofill operation */
561 562
		/* and check if we need to reply to a read request,
		 * new R5_Wantfill requests are held off until
563
		 * !STRIPE_BIOFILL_RUN
564 565
		 */
		if (test_and_clear_bit(R5_Wantfill, &dev->flags)) {
566 567 568 569 570 571 572 573
			struct bio *rbi, *rbi2;

			BUG_ON(!dev->read);
			rbi = dev->read;
			dev->read = NULL;
			while (rbi && rbi->bi_sector <
				dev->sector + STRIPE_SECTORS) {
				rbi2 = r5_next_bio(rbi, dev->sector);
574
				if (!raid5_dec_bi_phys_segments(rbi)) {
575 576 577 578 579 580 581
					rbi->bi_next = return_bi;
					return_bi = rbi;
				}
				rbi = rbi2;
			}
		}
	}
582 583
	spin_unlock_irq(&conf->device_lock);
	clear_bit(STRIPE_BIOFILL_RUN, &sh->state);
584 585 586

	return_io(return_bi);

587
	set_bit(STRIPE_HANDLE, &sh->state);
588 589 590 591 592 593 594 595 596
	release_stripe(sh);
}

static void ops_run_biofill(struct stripe_head *sh)
{
	struct dma_async_tx_descriptor *tx = NULL;
	raid5_conf_t *conf = sh->raid_conf;
	int i;

597
	pr_debug("%s: stripe %llu\n", __func__,
598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627
		(unsigned long long)sh->sector);

	for (i = sh->disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
		if (test_bit(R5_Wantfill, &dev->flags)) {
			struct bio *rbi;
			spin_lock_irq(&conf->device_lock);
			dev->read = rbi = dev->toread;
			dev->toread = NULL;
			spin_unlock_irq(&conf->device_lock);
			while (rbi && rbi->bi_sector <
				dev->sector + STRIPE_SECTORS) {
				tx = async_copy_data(0, rbi, dev->page,
					dev->sector, tx);
				rbi = r5_next_bio(rbi, dev->sector);
			}
		}
	}

	atomic_inc(&sh->count);
	async_trigger_callback(ASYNC_TX_DEP_ACK | ASYNC_TX_ACK, tx,
		ops_complete_biofill, sh);
}

static void ops_complete_compute5(void *stripe_head_ref)
{
	struct stripe_head *sh = stripe_head_ref;
	int target = sh->ops.target;
	struct r5dev *tgt = &sh->dev[target];

628
	pr_debug("%s: stripe %llu\n", __func__,
629 630 631 632 633
		(unsigned long long)sh->sector);

	set_bit(R5_UPTODATE, &tgt->flags);
	BUG_ON(!test_bit(R5_Wantcompute, &tgt->flags));
	clear_bit(R5_Wantcompute, &tgt->flags);
634 635 636
	clear_bit(STRIPE_COMPUTE_RUN, &sh->state);
	if (sh->check_state == check_state_compute_run)
		sh->check_state = check_state_compute_result;
637 638 639 640
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

641
static struct dma_async_tx_descriptor *ops_run_compute5(struct stripe_head *sh)
642 643 644 645 646 647 648 649 650 651 652 653
{
	/* kernel stack size limits the total number of disks */
	int disks = sh->disks;
	struct page *xor_srcs[disks];
	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;
	int i;

	pr_debug("%s: stripe %llu block: %d\n",
654
		__func__, (unsigned long long)sh->sector, target);
655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677
	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);

	if (unlikely(count == 1))
		tx = async_memcpy(xor_dest, xor_srcs[0], 0, 0, STRIPE_SIZE,
			0, NULL, ops_complete_compute5, sh);
	else
		tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE,
			ASYNC_TX_XOR_ZERO_DST, NULL,
			ops_complete_compute5, sh);

	return tx;
}

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

678
	pr_debug("%s: stripe %llu\n", __func__,
679 680 681 682 683 684 685 686 687 688 689 690 691 692
		(unsigned long long)sh->sector);
}

static struct dma_async_tx_descriptor *
ops_run_prexor(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
{
	/* kernel stack size limits the total number of disks */
	int disks = sh->disks;
	struct page *xor_srcs[disks];
	int count = 0, pd_idx = sh->pd_idx, i;

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

693
	pr_debug("%s: stripe %llu\n", __func__,
694 695 696 697 698
		(unsigned long long)sh->sector);

	for (i = disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
		/* Only process blocks that are known to be uptodate */
699
		if (test_bit(R5_Wantdrain, &dev->flags))
700 701 702 703 704 705 706 707 708 709 710
			xor_srcs[count++] = dev->page;
	}

	tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE,
		ASYNC_TX_DEP_ACK | ASYNC_TX_XOR_DROP_DST, tx,
		ops_complete_prexor, sh);

	return tx;
}

static struct dma_async_tx_descriptor *
711
ops_run_biodrain(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
712 713
{
	int disks = sh->disks;
714
	int i;
715

716
	pr_debug("%s: stripe %llu\n", __func__,
717 718 719 720 721 722
		(unsigned long long)sh->sector);

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

723
		if (test_and_clear_bit(R5_Wantdrain, &dev->flags)) {
724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749
			struct bio *wbi;

			spin_lock(&sh->lock);
			chosen = dev->towrite;
			dev->towrite = NULL;
			BUG_ON(dev->written);
			wbi = dev->written = chosen;
			spin_unlock(&sh->lock);

			while (wbi && wbi->bi_sector <
				dev->sector + STRIPE_SECTORS) {
				tx = async_copy_data(1, wbi, dev->page,
					dev->sector, tx);
				wbi = r5_next_bio(wbi, dev->sector);
			}
		}
	}

	return tx;
}

static void ops_complete_postxor(void *stripe_head_ref)
{
	struct stripe_head *sh = stripe_head_ref;
	int disks = sh->disks, i, pd_idx = sh->pd_idx;

750
	pr_debug("%s: stripe %llu\n", __func__,
751 752 753 754 755 756 757 758
		(unsigned long long)sh->sector);

	for (i = disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
		if (dev->written || i == pd_idx)
			set_bit(R5_UPTODATE, &dev->flags);
	}

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

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

static void
773
ops_run_postxor(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
774 775 776 777 778 779 780
{
	/* kernel stack size limits the total number of disks */
	int disks = sh->disks;
	struct page *xor_srcs[disks];

	int count = 0, pd_idx = sh->pd_idx, i;
	struct page *xor_dest;
781
	int prexor = 0;
782 783
	unsigned long flags;

784
	pr_debug("%s: stripe %llu\n", __func__,
785 786 787 788 789
		(unsigned long long)sh->sector);

	/* check if prexor is active which means only process blocks
	 * that are part of a read-modify-write (written)
	 */
790 791
	if (sh->reconstruct_state == reconstruct_state_prexor_drain_run) {
		prexor = 1;
792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819
		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
	 */
	flags = ASYNC_TX_DEP_ACK | ASYNC_TX_ACK |
		(prexor ? ASYNC_TX_XOR_DROP_DST : ASYNC_TX_XOR_ZERO_DST);

	atomic_inc(&sh->count);

	if (unlikely(count == 1)) {
		flags &= ~(ASYNC_TX_XOR_DROP_DST | ASYNC_TX_XOR_ZERO_DST);
		tx = async_memcpy(xor_dest, xor_srcs[0], 0, 0, STRIPE_SIZE,
820
			flags, tx, ops_complete_postxor, sh);
821 822
	} else
		tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE,
823
			flags, tx, ops_complete_postxor, sh);
824 825 826 827 828 829
}

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

830
	pr_debug("%s: stripe %llu\n", __func__,
831 832
		(unsigned long long)sh->sector);

833
	sh->check_state = check_state_check_result;
834 835 836 837 838 839 840 841 842 843 844 845 846 847
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

static void ops_run_check(struct stripe_head *sh)
{
	/* kernel stack size limits the total number of disks */
	int disks = sh->disks;
	struct page *xor_srcs[disks];
	struct dma_async_tx_descriptor *tx;

	int count = 0, pd_idx = sh->pd_idx, i;
	struct page *xor_dest = xor_srcs[count++] = sh->dev[pd_idx].page;

848
	pr_debug("%s: stripe %llu\n", __func__,
849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864
		(unsigned long long)sh->sector);

	for (i = disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
		if (i != pd_idx)
			xor_srcs[count++] = dev->page;
	}

	tx = async_xor_zero_sum(xor_dest, xor_srcs, 0, count, STRIPE_SIZE,
		&sh->ops.zero_sum_result, 0, NULL, NULL, NULL);

	atomic_inc(&sh->count);
	tx = async_trigger_callback(ASYNC_TX_DEP_ACK | ASYNC_TX_ACK, tx,
		ops_complete_check, sh);
}

865
static void raid5_run_ops(struct stripe_head *sh, unsigned long ops_request)
866 867 868 869
{
	int overlap_clear = 0, i, disks = sh->disks;
	struct dma_async_tx_descriptor *tx = NULL;

870
	if (test_bit(STRIPE_OP_BIOFILL, &ops_request)) {
871 872 873 874
		ops_run_biofill(sh);
		overlap_clear++;
	}

875 876 877 878 879 880
	if (test_bit(STRIPE_OP_COMPUTE_BLK, &ops_request)) {
		tx = ops_run_compute5(sh);
		/* terminate the chain if postxor is not set to be run */
		if (tx && !test_bit(STRIPE_OP_POSTXOR, &ops_request))
			async_tx_ack(tx);
	}
881

882
	if (test_bit(STRIPE_OP_PREXOR, &ops_request))
883 884
		tx = ops_run_prexor(sh, tx);

885
	if (test_bit(STRIPE_OP_BIODRAIN, &ops_request)) {
886
		tx = ops_run_biodrain(sh, tx);
887 888 889
		overlap_clear++;
	}

890
	if (test_bit(STRIPE_OP_POSTXOR, &ops_request))
891
		ops_run_postxor(sh, tx);
892

893
	if (test_bit(STRIPE_OP_CHECK, &ops_request))
894 895 896 897 898 899 900 901 902 903
		ops_run_check(sh);

	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);
		}
}

904
static int grow_one_stripe(raid5_conf_t *conf)
L
Linus Torvalds 已提交
905 906
{
	struct stripe_head *sh;
907 908 909 910 911 912 913 914 915 916 917 918
	sh = kmem_cache_alloc(conf->slab_cache, GFP_KERNEL);
	if (!sh)
		return 0;
	memset(sh, 0, sizeof(*sh) + (conf->raid_disks-1)*sizeof(struct r5dev));
	sh->raid_conf = conf;
	spin_lock_init(&sh->lock);

	if (grow_buffers(sh, conf->raid_disks)) {
		shrink_buffers(sh, conf->raid_disks);
		kmem_cache_free(conf->slab_cache, sh);
		return 0;
	}
919
	sh->disks = conf->raid_disks;
920 921 922 923 924 925 926 927 928 929
	/* 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;
}

static int grow_stripes(raid5_conf_t *conf, int num)
{
930
	struct kmem_cache *sc;
L
Linus Torvalds 已提交
931 932
	int devs = conf->raid_disks;

933 934 935 936
	sprintf(conf->cache_name[0],
		"raid%d-%s", conf->level, mdname(conf->mddev));
	sprintf(conf->cache_name[1],
		"raid%d-%s-alt", conf->level, mdname(conf->mddev));
937 938
	conf->active_name = 0;
	sc = kmem_cache_create(conf->cache_name[conf->active_name],
L
Linus Torvalds 已提交
939
			       sizeof(struct stripe_head)+(devs-1)*sizeof(struct r5dev),
940
			       0, 0, NULL);
L
Linus Torvalds 已提交
941 942 943
	if (!sc)
		return 1;
	conf->slab_cache = sc;
944
	conf->pool_size = devs;
945
	while (num--)
946
		if (!grow_one_stripe(conf))
L
Linus Torvalds 已提交
947 948 949
			return 1;
	return 0;
}
950

951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978
static int resize_stripes(raid5_conf_t *conf, int newsize)
{
	/* Make all the stripes able to hold 'newsize' devices.
	 * New slots in each stripe get 'page' set to a new page.
	 *
	 * This happens in stages:
	 * 1/ create a new kmem_cache and allocate the required number of
	 *    stripe_heads.
	 * 2/ gather all the old stripe_heads and tranfer the pages across
	 *    to the new stripe_heads.  This will have the side effect of
	 *    freezing the array as once all stripe_heads have been collected,
	 *    no IO will be possible.  Old stripe heads are freed once their
	 *    pages have been transferred over, and the old kmem_cache is
	 *    freed when all stripes are done.
	 * 3/ reallocate conf->disks to be suitable bigger.  If this fails,
	 *    we simple return a failre status - no need to clean anything up.
	 * 4/ allocate new pages for the new slots in the new stripe_heads.
	 *    If this fails, we don't bother trying the shrink the
	 *    stripe_heads down again, we just leave them as they are.
	 *    As each stripe_head is processed the new one is released into
	 *    active service.
	 *
	 * Once step2 is started, we cannot afford to wait for a write,
	 * so we use GFP_NOIO allocations.
	 */
	struct stripe_head *osh, *nsh;
	LIST_HEAD(newstripes);
	struct disk_info *ndisks;
979
	int err;
980
	struct kmem_cache *sc;
981 982 983 984 985
	int i;

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

986 987 988
	err = md_allow_write(conf->mddev);
	if (err)
		return err;
989

990 991 992
	/* Step 1 */
	sc = kmem_cache_create(conf->cache_name[1-conf->active_name],
			       sizeof(struct stripe_head)+(newsize-1)*sizeof(struct r5dev),
993
			       0, 0, NULL);
994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027
	if (!sc)
		return -ENOMEM;

	for (i = conf->max_nr_stripes; i; i--) {
		nsh = kmem_cache_alloc(sc, GFP_KERNEL);
		if (!nsh)
			break;

		memset(nsh, 0, sizeof(*nsh) + (newsize-1)*sizeof(struct r5dev));

		nsh->raid_conf = conf;
		spin_lock_init(&nsh->lock);

		list_add(&nsh->lru, &newstripes);
	}
	if (i) {
		/* didn't get enough, give up */
		while (!list_empty(&newstripes)) {
			nsh = list_entry(newstripes.next, struct stripe_head, lru);
			list_del(&nsh->lru);
			kmem_cache_free(sc, nsh);
		}
		kmem_cache_destroy(sc);
		return -ENOMEM;
	}
	/* Step 2 - Must use GFP_NOIO now.
	 * OK, we have enough stripes, start collecting inactive
	 * stripes and copying them over
	 */
	list_for_each_entry(nsh, &newstripes, lru) {
		spin_lock_irq(&conf->device_lock);
		wait_event_lock_irq(conf->wait_for_stripe,
				    !list_empty(&conf->inactive_list),
				    conf->device_lock,
1028
				    unplug_slaves(conf->mddev)
1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074
			);
		osh = get_free_stripe(conf);
		spin_unlock_irq(&conf->device_lock);
		atomic_set(&nsh->count, 1);
		for(i=0; i<conf->pool_size; i++)
			nsh->dev[i].page = osh->dev[i].page;
		for( ; i<newsize; i++)
			nsh->dev[i].page = NULL;
		kmem_cache_free(conf->slab_cache, osh);
	}
	kmem_cache_destroy(conf->slab_cache);

	/* Step 3.
	 * At this point, we are holding all the stripes so the array
	 * is completely stalled, so now is a good time to resize
	 * conf->disks.
	 */
	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;

	/* 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);
		for (i=conf->raid_disks; i < newsize; i++)
			if (nsh->dev[i].page == NULL) {
				struct page *p = alloc_page(GFP_NOIO);
				nsh->dev[i].page = p;
				if (!p)
					err = -ENOMEM;
			}
		release_stripe(nsh);
	}
	/* critical section pass, GFP_NOIO no longer needed */

	conf->slab_cache = sc;
	conf->active_name = 1-conf->active_name;
	conf->pool_size = newsize;
	return err;
}
L
Linus Torvalds 已提交
1075

1076
static int drop_one_stripe(raid5_conf_t *conf)
L
Linus Torvalds 已提交
1077 1078 1079
{
	struct stripe_head *sh;

1080 1081 1082 1083 1084
	spin_lock_irq(&conf->device_lock);
	sh = get_free_stripe(conf);
	spin_unlock_irq(&conf->device_lock);
	if (!sh)
		return 0;
1085
	BUG_ON(atomic_read(&sh->count));
1086
	shrink_buffers(sh, conf->pool_size);
1087 1088 1089 1090 1091 1092 1093 1094 1095 1096
	kmem_cache_free(conf->slab_cache, sh);
	atomic_dec(&conf->active_stripes);
	return 1;
}

static void shrink_stripes(raid5_conf_t *conf)
{
	while (drop_one_stripe(conf))
		;

N
NeilBrown 已提交
1097 1098
	if (conf->slab_cache)
		kmem_cache_destroy(conf->slab_cache);
L
Linus Torvalds 已提交
1099 1100 1101
	conf->slab_cache = NULL;
}

1102
static void raid5_end_read_request(struct bio * bi, int error)
L
Linus Torvalds 已提交
1103
{
1104
	struct stripe_head *sh = bi->bi_private;
L
Linus Torvalds 已提交
1105
	raid5_conf_t *conf = sh->raid_conf;
1106
	int disks = sh->disks, i;
L
Linus Torvalds 已提交
1107
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
1108 1109
	char b[BDEVNAME_SIZE];
	mdk_rdev_t *rdev;
L
Linus Torvalds 已提交
1110 1111 1112 1113 1114 1115


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

1116 1117
	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
1118 1119 1120
		uptodate);
	if (i == disks) {
		BUG();
1121
		return;
L
Linus Torvalds 已提交
1122 1123 1124 1125
	}

	if (uptodate) {
		set_bit(R5_UPTODATE, &sh->dev[i].flags);
1126
		if (test_bit(R5_ReadError, &sh->dev[i].flags)) {
1127
			rdev = conf->disks[i].rdev;
1128 1129 1130 1131 1132 1133
			printk_rl(KERN_INFO "raid5:%s: read error corrected"
				  " (%lu sectors at %llu on %s)\n",
				  mdname(conf->mddev), STRIPE_SECTORS,
				  (unsigned long long)(sh->sector
						       + rdev->data_offset),
				  bdevname(rdev->bdev, b));
1134 1135 1136
			clear_bit(R5_ReadError, &sh->dev[i].flags);
			clear_bit(R5_ReWrite, &sh->dev[i].flags);
		}
1137 1138
		if (atomic_read(&conf->disks[i].rdev->read_errors))
			atomic_set(&conf->disks[i].rdev->read_errors, 0);
L
Linus Torvalds 已提交
1139
	} else {
1140
		const char *bdn = bdevname(conf->disks[i].rdev->bdev, b);
1141
		int retry = 0;
1142 1143
		rdev = conf->disks[i].rdev;

L
Linus Torvalds 已提交
1144
		clear_bit(R5_UPTODATE, &sh->dev[i].flags);
1145
		atomic_inc(&rdev->read_errors);
1146
		if (conf->mddev->degraded)
1147 1148 1149 1150 1151 1152 1153
			printk_rl(KERN_WARNING
				  "raid5:%s: read error not correctable "
				  "(sector %llu on %s).\n",
				  mdname(conf->mddev),
				  (unsigned long long)(sh->sector
						       + rdev->data_offset),
				  bdn);
1154
		else if (test_bit(R5_ReWrite, &sh->dev[i].flags))
1155
			/* Oh, no!!! */
1156 1157 1158 1159 1160 1161 1162
			printk_rl(KERN_WARNING
				  "raid5:%s: read error NOT corrected!! "
				  "(sector %llu on %s).\n",
				  mdname(conf->mddev),
				  (unsigned long long)(sh->sector
						       + rdev->data_offset),
				  bdn);
1163
		else if (atomic_read(&rdev->read_errors)
1164
			 > conf->max_nr_stripes)
N
NeilBrown 已提交
1165
			printk(KERN_WARNING
1166 1167
			       "raid5:%s: Too many read errors, failing device %s.\n",
			       mdname(conf->mddev), bdn);
1168 1169 1170 1171 1172
		else
			retry = 1;
		if (retry)
			set_bit(R5_ReadError, &sh->dev[i].flags);
		else {
1173 1174
			clear_bit(R5_ReadError, &sh->dev[i].flags);
			clear_bit(R5_ReWrite, &sh->dev[i].flags);
1175
			md_error(conf->mddev, rdev);
1176
		}
L
Linus Torvalds 已提交
1177 1178 1179 1180 1181 1182 1183
	}
	rdev_dec_pending(conf->disks[i].rdev, conf->mddev);
	clear_bit(R5_LOCKED, &sh->dev[i].flags);
	set_bit(STRIPE_HANDLE, &sh->state);
	release_stripe(sh);
}

1184
static void raid5_end_write_request(struct bio *bi, int error)
L
Linus Torvalds 已提交
1185
{
1186
	struct stripe_head *sh = bi->bi_private;
L
Linus Torvalds 已提交
1187
	raid5_conf_t *conf = sh->raid_conf;
1188
	int disks = sh->disks, i;
L
Linus Torvalds 已提交
1189 1190 1191 1192 1193 1194
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);

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

1195
	pr_debug("end_write_request %llu/%d, count %d, uptodate: %d.\n",
L
Linus Torvalds 已提交
1196 1197 1198 1199
		(unsigned long long)sh->sector, i, atomic_read(&sh->count),
		uptodate);
	if (i == disks) {
		BUG();
1200
		return;
L
Linus Torvalds 已提交
1201 1202 1203 1204 1205 1206 1207 1208 1209
	}

	if (!uptodate)
		md_error(conf->mddev, conf->disks[i].rdev);

	rdev_dec_pending(conf->disks[i].rdev, conf->mddev);
	
	clear_bit(R5_LOCKED, &sh->dev[i].flags);
	set_bit(STRIPE_HANDLE, &sh->state);
1210
	release_stripe(sh);
L
Linus Torvalds 已提交
1211 1212 1213
}


1214
static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous);
L
Linus Torvalds 已提交
1215
	
1216
static void raid5_build_block(struct stripe_head *sh, int i, int previous)
L
Linus Torvalds 已提交
1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231
{
	struct r5dev *dev = &sh->dev[i];

	bio_init(&dev->req);
	dev->req.bi_io_vec = &dev->vec;
	dev->req.bi_vcnt++;
	dev->req.bi_max_vecs++;
	dev->vec.bv_page = dev->page;
	dev->vec.bv_len = STRIPE_SIZE;
	dev->vec.bv_offset = 0;

	dev->req.bi_sector = sh->sector;
	dev->req.bi_private = sh;

	dev->flags = 0;
1232
	dev->sector = compute_blocknr(sh, i, previous);
L
Linus Torvalds 已提交
1233 1234 1235 1236 1237 1238
}

static void error(mddev_t *mddev, mdk_rdev_t *rdev)
{
	char b[BDEVNAME_SIZE];
	raid5_conf_t *conf = (raid5_conf_t *) mddev->private;
1239
	pr_debug("raid5: error called\n");
L
Linus Torvalds 已提交
1240

1241
	if (!test_bit(Faulty, &rdev->flags)) {
1242
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
1243 1244 1245
		if (test_and_clear_bit(In_sync, &rdev->flags)) {
			unsigned long flags;
			spin_lock_irqsave(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1246
			mddev->degraded++;
1247
			spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1248 1249 1250
			/*
			 * if recovery was running, make sure it aborts.
			 */
1251
			set_bit(MD_RECOVERY_INTR, &mddev->recovery);
L
Linus Torvalds 已提交
1252
		}
1253
		set_bit(Faulty, &rdev->flags);
1254 1255 1256 1257
		printk(KERN_ALERT
		       "raid5: Disk failure on %s, disabling device.\n"
		       "raid5: Operation continuing on %d devices.\n",
		       bdevname(rdev->bdev,b), conf->raid_disks - mddev->degraded);
L
Linus Torvalds 已提交
1258
	}
1259
}
L
Linus Torvalds 已提交
1260 1261 1262 1263 1264

/*
 * Input: a 'big' sector number,
 * Output: index of the data and parity disk, and the sector # in them.
 */
1265
static sector_t raid5_compute_sector(raid5_conf_t *conf, sector_t r_sector,
1266 1267
				     int previous, int *dd_idx,
				     struct stripe_head *sh)
L
Linus Torvalds 已提交
1268 1269 1270 1271
{
	long stripe;
	unsigned long chunk_number;
	unsigned int chunk_offset;
1272
	int pd_idx, qd_idx;
1273
	int ddf_layout = 0;
L
Linus Torvalds 已提交
1274
	sector_t new_sector;
1275 1276
	int algorithm = previous ? conf->prev_algo
				 : conf->algorithm;
1277 1278
	int sectors_per_chunk = previous ? (conf->prev_chunk >> 9)
					 : (conf->chunk_size >> 9);
1279 1280 1281
	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
1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304

	/* 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;
	BUG_ON(r_sector != chunk_number);

	/*
	 * Compute the stripe number
	 */
	stripe = chunk_number / data_disks;

	/*
	 * Compute the data disk and parity disk indexes inside the stripe
	 */
	*dd_idx = chunk_number % data_disks;

	/*
	 * Select the parity disk based on the user selected algorithm.
	 */
1305
	pd_idx = qd_idx = ~0;
1306 1307
	switch(conf->level) {
	case 4:
1308
		pd_idx = data_disks;
1309 1310
		break;
	case 5:
1311
		switch (algorithm) {
L
Linus Torvalds 已提交
1312
		case ALGORITHM_LEFT_ASYMMETRIC:
1313 1314
			pd_idx = data_disks - stripe % raid_disks;
			if (*dd_idx >= pd_idx)
L
Linus Torvalds 已提交
1315 1316 1317
				(*dd_idx)++;
			break;
		case ALGORITHM_RIGHT_ASYMMETRIC:
1318 1319
			pd_idx = stripe % raid_disks;
			if (*dd_idx >= pd_idx)
L
Linus Torvalds 已提交
1320 1321 1322
				(*dd_idx)++;
			break;
		case ALGORITHM_LEFT_SYMMETRIC:
1323 1324
			pd_idx = data_disks - stripe % raid_disks;
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
L
Linus Torvalds 已提交
1325 1326
			break;
		case ALGORITHM_RIGHT_SYMMETRIC:
1327 1328
			pd_idx = stripe % raid_disks;
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
L
Linus Torvalds 已提交
1329
			break;
1330 1331 1332 1333 1334 1335 1336
		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
1337
		default:
N
NeilBrown 已提交
1338
			printk(KERN_ERR "raid5: unsupported algorithm %d\n",
1339
				algorithm);
1340
			BUG();
1341 1342 1343 1344
		}
		break;
	case 6:

1345
		switch (algorithm) {
1346
		case ALGORITHM_LEFT_ASYMMETRIC:
1347 1348 1349
			pd_idx = raid_disks - 1 - (stripe % raid_disks);
			qd_idx = pd_idx + 1;
			if (pd_idx == raid_disks-1) {
1350
				(*dd_idx)++;	/* Q D D D P */
1351 1352
				qd_idx = 0;
			} else if (*dd_idx >= pd_idx)
1353 1354 1355
				(*dd_idx) += 2; /* D D P Q D */
			break;
		case ALGORITHM_RIGHT_ASYMMETRIC:
1356 1357 1358
			pd_idx = stripe % raid_disks;
			qd_idx = pd_idx + 1;
			if (pd_idx == raid_disks-1) {
1359
				(*dd_idx)++;	/* Q D D D P */
1360 1361
				qd_idx = 0;
			} else if (*dd_idx >= pd_idx)
1362 1363 1364
				(*dd_idx) += 2; /* D D P Q D */
			break;
		case ALGORITHM_LEFT_SYMMETRIC:
1365 1366 1367
			pd_idx = raid_disks - 1 - (stripe % raid_disks);
			qd_idx = (pd_idx + 1) % raid_disks;
			*dd_idx = (pd_idx + 2 + *dd_idx) % raid_disks;
1368 1369
			break;
		case ALGORITHM_RIGHT_SYMMETRIC:
1370 1371 1372
			pd_idx = stripe % raid_disks;
			qd_idx = (pd_idx + 1) % raid_disks;
			*dd_idx = (pd_idx + 2 + *dd_idx) % raid_disks;
1373
			break;
1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395

		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.
			 */
			pd_idx = stripe % raid_disks;
			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 */
1396
			ddf_layout = 1;
1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410
			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
			 */
			pd_idx = raid_disks - 1 - ((stripe + 1) % raid_disks);
			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 */
1411
			ddf_layout = 1;
1412 1413 1414 1415 1416 1417 1418
			break;

		case ALGORITHM_ROTATING_N_CONTINUE:
			/* Same as left_symmetric but Q is before P */
			pd_idx = raid_disks - 1 - (stripe % raid_disks);
			qd_idx = (pd_idx + raid_disks - 1) % raid_disks;
			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
1419
			ddf_layout = 1;
1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455
			break;

		case ALGORITHM_LEFT_ASYMMETRIC_6:
			/* RAID5 left_asymmetric, with Q on last device */
			pd_idx = data_disks - stripe % (raid_disks-1);
			if (*dd_idx >= pd_idx)
				(*dd_idx)++;
			qd_idx = raid_disks - 1;
			break;

		case ALGORITHM_RIGHT_ASYMMETRIC_6:
			pd_idx = stripe % (raid_disks-1);
			if (*dd_idx >= pd_idx)
				(*dd_idx)++;
			qd_idx = raid_disks - 1;
			break;

		case ALGORITHM_LEFT_SYMMETRIC_6:
			pd_idx = data_disks - stripe % (raid_disks-1);
			*dd_idx = (pd_idx + 1 + *dd_idx) % (raid_disks-1);
			qd_idx = raid_disks - 1;
			break;

		case ALGORITHM_RIGHT_SYMMETRIC_6:
			pd_idx = stripe % (raid_disks-1);
			*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;


1456
		default:
1457
			printk(KERN_CRIT "raid6: unsupported algorithm %d\n",
1458
			       algorithm);
1459
			BUG();
1460 1461
		}
		break;
L
Linus Torvalds 已提交
1462 1463
	}

1464 1465 1466
	if (sh) {
		sh->pd_idx = pd_idx;
		sh->qd_idx = qd_idx;
1467
		sh->ddf_layout = ddf_layout;
1468
	}
L
Linus Torvalds 已提交
1469 1470 1471 1472 1473 1474 1475 1476
	/*
	 * Finally, compute the new sector number
	 */
	new_sector = (sector_t)stripe * sectors_per_chunk + chunk_offset;
	return new_sector;
}


1477
static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous)
L
Linus Torvalds 已提交
1478 1479
{
	raid5_conf_t *conf = sh->raid_conf;
1480 1481
	int raid_disks = sh->disks;
	int data_disks = raid_disks - conf->max_degraded;
L
Linus Torvalds 已提交
1482
	sector_t new_sector = sh->sector, check;
1483 1484
	int sectors_per_chunk = previous ? (conf->prev_chunk >> 9)
					 : (conf->chunk_size >> 9);
1485 1486
	int algorithm = previous ? conf->prev_algo
				 : conf->algorithm;
L
Linus Torvalds 已提交
1487 1488
	sector_t stripe;
	int chunk_offset;
1489
	int chunk_number, dummy1, dd_idx = i;
L
Linus Torvalds 已提交
1490
	sector_t r_sector;
1491
	struct stripe_head sh2;
L
Linus Torvalds 已提交
1492

1493

L
Linus Torvalds 已提交
1494 1495 1496 1497
	chunk_offset = sector_div(new_sector, sectors_per_chunk);
	stripe = new_sector;
	BUG_ON(new_sector != stripe);

1498 1499 1500 1501 1502
	if (i == sh->pd_idx)
		return 0;
	switch(conf->level) {
	case 4: break;
	case 5:
1503
		switch (algorithm) {
L
Linus Torvalds 已提交
1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514
		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;
1515 1516 1517 1518 1519
		case ALGORITHM_PARITY_0:
			i -= 1;
			break;
		case ALGORITHM_PARITY_N:
			break;
L
Linus Torvalds 已提交
1520
		default:
N
NeilBrown 已提交
1521
			printk(KERN_ERR "raid5: unsupported algorithm %d\n",
1522
			       algorithm);
1523
			BUG();
1524 1525 1526
		}
		break;
	case 6:
1527
		if (i == sh->qd_idx)
1528
			return 0; /* It is the Q disk */
1529
		switch (algorithm) {
1530 1531
		case ALGORITHM_LEFT_ASYMMETRIC:
		case ALGORITHM_RIGHT_ASYMMETRIC:
1532 1533 1534 1535
		case ALGORITHM_ROTATING_ZERO_RESTART:
		case ALGORITHM_ROTATING_N_RESTART:
			if (sh->pd_idx == raid_disks-1)
				i--;	/* Q D D D P */
1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549
			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;
1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574
		case ALGORITHM_PARITY_0:
			i -= 2;
			break;
		case ALGORITHM_PARITY_N:
			break;
		case ALGORITHM_ROTATING_N_CONTINUE:
			if (sh->pd_idx == 0)
				i--;	/* P D D D Q */
			else if (i > sh->pd_idx)
				i -= 2; /* D D Q P D */
			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;
1575
		default:
1576
			printk(KERN_CRIT "raid6: unsupported algorithm %d\n",
1577
			       algorithm);
1578
			BUG();
1579 1580
		}
		break;
L
Linus Torvalds 已提交
1581 1582 1583 1584 1585
	}

	chunk_number = stripe * data_disks + i;
	r_sector = (sector_t)chunk_number * sectors_per_chunk + chunk_offset;

1586
	check = raid5_compute_sector(conf, r_sector,
1587
				     previous, &dummy1, &sh2);
1588 1589
	if (check != sh->sector || dummy1 != dd_idx || sh2.pd_idx != sh->pd_idx
		|| sh2.qd_idx != sh->qd_idx) {
N
NeilBrown 已提交
1590
		printk(KERN_ERR "compute_blocknr: map not correct\n");
L
Linus Torvalds 已提交
1591 1592 1593 1594 1595 1596 1597 1598
		return 0;
	}
	return r_sector;
}



/*
1599 1600 1601 1602 1603
 * Copy data between a page in the stripe cache, and one or more bion
 * The page could align with the middle of the bio, or there could be
 * several bion, each with several bio_vecs, which cover part of the page
 * Multiple bion are linked together on bi_next.  There may be extras
 * at the end of this list.  We ignore them.
L
Linus Torvalds 已提交
1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631
 */
static void copy_data(int frombio, struct bio *bio,
		     struct page *page,
		     sector_t sector)
{
	char *pa = page_address(page);
	struct bio_vec *bvl;
	int i;
	int page_offset;

	if (bio->bi_sector >= sector)
		page_offset = (signed)(bio->bi_sector - sector) * 512;
	else
		page_offset = (signed)(sector - bio->bi_sector) * -512;
	bio_for_each_segment(bvl, bio, i) {
		int len = bio_iovec_idx(bio,i)->bv_len;
		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;
1632

L
Linus Torvalds 已提交
1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646
		if (clen > 0) {
			char *ba = __bio_kmap_atomic(bio, i, KM_USER0);
			if (frombio)
				memcpy(pa+page_offset, ba+b_offset, clen);
			else
				memcpy(ba+b_offset, pa+page_offset, clen);
			__bio_kunmap_atomic(ba, KM_USER0);
		}
		if (clen < len) /* hit end of page */
			break;
		page_offset +=  len;
	}
}

D
Dan Williams 已提交
1647 1648 1649 1650 1651
#define check_xor()	do {						  \
				if (count == MAX_XOR_BLOCKS) {		  \
				xor_blocks(count, STRIPE_SIZE, dest, ptr);\
				count = 0;				  \
			   }						  \
L
Linus Torvalds 已提交
1652 1653
			} while(0)

1654 1655
static void compute_parity6(struct stripe_head *sh, int method)
{
1656
	raid5_conf_t *conf = sh->raid_conf;
1657
	int i, pd_idx, qd_idx, d0_idx, disks = sh->disks, count;
1658
	int syndrome_disks = sh->ddf_layout ? disks : (disks - 2);
1659 1660
	struct bio *chosen;
	/**** FIX THIS: This could be very bad if disks is close to 256 ****/
1661
	void *ptrs[syndrome_disks+2];
1662

1663 1664 1665
	pd_idx = sh->pd_idx;
	qd_idx = sh->qd_idx;
	d0_idx = raid6_d0(sh);
1666

1667
	pr_debug("compute_parity, stripe %llu, method %d\n",
1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681
		(unsigned long long)sh->sector, method);

	switch(method) {
	case READ_MODIFY_WRITE:
		BUG();		/* READ_MODIFY_WRITE N/A for RAID-6 */
	case RECONSTRUCT_WRITE:
		for (i= disks; i-- ;)
			if ( i != pd_idx && i != qd_idx && sh->dev[i].towrite ) {
				chosen = sh->dev[i].towrite;
				sh->dev[i].towrite = NULL;

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

E
Eric Sesterhenn 已提交
1682
				BUG_ON(sh->dev[i].written);
1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702
				sh->dev[i].written = chosen;
			}
		break;
	case CHECK_PARITY:
		BUG();		/* Not implemented yet */
	}

	for (i = disks; i--;)
		if (sh->dev[i].written) {
			sector_t sector = sh->dev[i].sector;
			struct bio *wbi = sh->dev[i].written;
			while (wbi && wbi->bi_sector < sector + STRIPE_SECTORS) {
				copy_data(1, wbi, sh->dev[i].page, sector);
				wbi = r5_next_bio(wbi, sector);
			}

			set_bit(R5_LOCKED, &sh->dev[i].flags);
			set_bit(R5_UPTODATE, &sh->dev[i].flags);
		}

1703
	/* Note that unlike RAID-5, the ordering of the disks matters greatly.*/
1704 1705 1706 1707

	for (i = 0; i < disks; i++)
		ptrs[i] = (void *)raid6_empty_zero_page;

1708 1709 1710
	count = 0;
	i = d0_idx;
	do {
1711 1712
		int slot = raid6_idx_to_slot(i, sh, &count, syndrome_disks);

1713
		ptrs[slot] = page_address(sh->dev[i].page);
1714
		if (slot < syndrome_disks &&
1715 1716 1717 1718 1719
		    !test_bit(R5_UPTODATE, &sh->dev[i].flags)) {
			printk(KERN_ERR "block %d/%d not uptodate "
			       "on parity calc\n", i, count);
			BUG();
		}
1720

1721 1722
		i = raid6_next_disk(i, disks);
	} while (i != d0_idx);
1723
	BUG_ON(count != syndrome_disks);
1724

1725
	raid6_call.gen_syndrome(syndrome_disks+2, STRIPE_SIZE, ptrs);
1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744

	switch(method) {
	case RECONSTRUCT_WRITE:
		set_bit(R5_UPTODATE, &sh->dev[pd_idx].flags);
		set_bit(R5_UPTODATE, &sh->dev[qd_idx].flags);
		set_bit(R5_LOCKED,   &sh->dev[pd_idx].flags);
		set_bit(R5_LOCKED,   &sh->dev[qd_idx].flags);
		break;
	case UPDATE_PARITY:
		set_bit(R5_UPTODATE, &sh->dev[pd_idx].flags);
		set_bit(R5_UPTODATE, &sh->dev[qd_idx].flags);
		break;
	}
}


/* Compute one missing block */
static void compute_block_1(struct stripe_head *sh, int dd_idx, int nozero)
{
1745
	int i, count, disks = sh->disks;
D
Dan Williams 已提交
1746
	void *ptr[MAX_XOR_BLOCKS], *dest, *p;
1747
	int qd_idx = sh->qd_idx;
1748

1749
	pr_debug("compute_block_1, stripe %llu, idx %d\n",
1750 1751 1752 1753 1754 1755
		(unsigned long long)sh->sector, dd_idx);

	if ( dd_idx == qd_idx ) {
		/* We're actually computing the Q drive */
		compute_parity6(sh, UPDATE_PARITY);
	} else {
D
Dan Williams 已提交
1756 1757 1758
		dest = page_address(sh->dev[dd_idx].page);
		if (!nozero) memset(dest, 0, STRIPE_SIZE);
		count = 0;
1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771
		for (i = disks ; i--; ) {
			if (i == dd_idx || i == qd_idx)
				continue;
			p = page_address(sh->dev[i].page);
			if (test_bit(R5_UPTODATE, &sh->dev[i].flags))
				ptr[count++] = p;
			else
				printk("compute_block() %d, stripe %llu, %d"
				       " not present\n", dd_idx,
				       (unsigned long long)sh->sector, i);

			check_xor();
		}
D
Dan Williams 已提交
1772 1773
		if (count)
			xor_blocks(count, STRIPE_SIZE, dest, ptr);
1774 1775 1776 1777 1778 1779 1780 1781
		if (!nozero) set_bit(R5_UPTODATE, &sh->dev[dd_idx].flags);
		else clear_bit(R5_UPTODATE, &sh->dev[dd_idx].flags);
	}
}

/* Compute two missing blocks */
static void compute_block_2(struct stripe_head *sh, int dd_idx1, int dd_idx2)
{
1782
	int i, count, disks = sh->disks;
1783
	int syndrome_disks = sh->ddf_layout ? disks : disks-2;
1784 1785 1786
	int d0_idx = raid6_d0(sh);
	int faila = -1, failb = -1;
	/**** FIX THIS: This could be very bad if disks is close to 256 ****/
1787
	void *ptrs[syndrome_disks+2];
1788

1789 1790
	for (i = 0; i < disks ; i++)
		ptrs[i] = (void *)raid6_empty_zero_page;
1791 1792 1793
	count = 0;
	i = d0_idx;
	do {
1794 1795
		int slot = raid6_idx_to_slot(i, sh, &count, syndrome_disks);

1796
		ptrs[slot] = page_address(sh->dev[i].page);
1797

1798 1799 1800 1801 1802 1803
		if (i == dd_idx1)
			faila = slot;
		if (i == dd_idx2)
			failb = slot;
		i = raid6_next_disk(i, disks);
	} while (i != d0_idx);
1804
	BUG_ON(count != syndrome_disks);
1805 1806 1807 1808

	BUG_ON(faila == failb);
	if ( failb < faila ) { int tmp = faila; faila = failb; failb = tmp; }

1809
	pr_debug("compute_block_2, stripe %llu, idx %d,%d (%d,%d)\n",
1810 1811
		 (unsigned long long)sh->sector, dd_idx1, dd_idx2,
		 faila, failb);
1812

1813
	if (failb == syndrome_disks+1) {
1814
		/* Q disk is one of the missing disks */
1815
		if (faila == syndrome_disks) {
1816 1817 1818 1819 1820
			/* Missing P+Q, just recompute */
			compute_parity6(sh, UPDATE_PARITY);
			return;
		} else {
			/* We're missing D+Q; recompute D from P */
1821 1822 1823
			compute_block_1(sh, ((dd_idx1 == sh->qd_idx) ?
					     dd_idx2 : dd_idx1),
					0);
1824 1825 1826 1827 1828
			compute_parity6(sh, UPDATE_PARITY); /* Is this necessary? */
			return;
		}
	}

1829
	/* We're missing D+P or D+D; */
1830
	if (failb == syndrome_disks) {
1831
		/* We're missing D+P. */
1832
		raid6_datap_recov(syndrome_disks+2, STRIPE_SIZE, faila, ptrs);
1833 1834
	} else {
		/* We're missing D+D. */
1835 1836
		raid6_2data_recov(syndrome_disks+2, STRIPE_SIZE, faila, failb,
				  ptrs);
1837
	}
1838 1839 1840 1841

	/* Both the above update both missing blocks */
	set_bit(R5_UPTODATE, &sh->dev[dd_idx1].flags);
	set_bit(R5_UPTODATE, &sh->dev[dd_idx2].flags);
1842 1843
}

1844
static void
1845
schedule_reconstruction5(struct stripe_head *sh, struct stripe_head_state *s,
1846
			 int rcw, int expand)
1847 1848 1849 1850 1851 1852 1853 1854 1855
{
	int i, pd_idx = sh->pd_idx, disks = sh->disks;

	if (rcw) {
		/* if we are not expanding this is a proper write request, and
		 * there will be bios with new data to be drained into the
		 * stripe cache
		 */
		if (!expand) {
1856 1857 1858 1859
			sh->reconstruct_state = reconstruct_state_drain_run;
			set_bit(STRIPE_OP_BIODRAIN, &s->ops_request);
		} else
			sh->reconstruct_state = reconstruct_state_run;
1860

1861
		set_bit(STRIPE_OP_POSTXOR, &s->ops_request);
1862 1863 1864 1865 1866 1867

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

			if (dev->towrite) {
				set_bit(R5_LOCKED, &dev->flags);
1868
				set_bit(R5_Wantdrain, &dev->flags);
1869 1870
				if (!expand)
					clear_bit(R5_UPTODATE, &dev->flags);
1871
				s->locked++;
1872 1873
			}
		}
1874
		if (s->locked + 1 == disks)
1875 1876
			if (!test_and_set_bit(STRIPE_FULL_WRITE, &sh->state))
				atomic_inc(&sh->raid_conf->pending_full_writes);
1877 1878 1879 1880
	} else {
		BUG_ON(!(test_bit(R5_UPTODATE, &sh->dev[pd_idx].flags) ||
			test_bit(R5_Wantcompute, &sh->dev[pd_idx].flags)));

1881
		sh->reconstruct_state = reconstruct_state_prexor_drain_run;
1882 1883 1884
		set_bit(STRIPE_OP_PREXOR, &s->ops_request);
		set_bit(STRIPE_OP_BIODRAIN, &s->ops_request);
		set_bit(STRIPE_OP_POSTXOR, &s->ops_request);
1885 1886 1887 1888 1889 1890 1891 1892

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

			if (dev->towrite &&
			    (test_bit(R5_UPTODATE, &dev->flags) ||
1893 1894
			     test_bit(R5_Wantcompute, &dev->flags))) {
				set_bit(R5_Wantdrain, &dev->flags);
1895 1896
				set_bit(R5_LOCKED, &dev->flags);
				clear_bit(R5_UPTODATE, &dev->flags);
1897
				s->locked++;
1898 1899 1900 1901 1902 1903 1904 1905 1906
			}
		}
	}

	/* keep the parity disk locked while asynchronous operations
	 * are in flight
	 */
	set_bit(R5_LOCKED, &sh->dev[pd_idx].flags);
	clear_bit(R5_UPTODATE, &sh->dev[pd_idx].flags);
1907
	s->locked++;
1908

1909
	pr_debug("%s: stripe %llu locked: %d ops_request: %lx\n",
1910
		__func__, (unsigned long long)sh->sector,
1911
		s->locked, s->ops_request);
1912
}
1913

L
Linus Torvalds 已提交
1914 1915
/*
 * Each stripe/dev can have one or more bion attached.
1916
 * toread/towrite point to the first in a chain.
L
Linus Torvalds 已提交
1917 1918 1919 1920 1921 1922
 * The bi_next chain must be in order.
 */
static int add_stripe_bio(struct stripe_head *sh, struct bio *bi, int dd_idx, int forwrite)
{
	struct bio **bip;
	raid5_conf_t *conf = sh->raid_conf;
1923
	int firstwrite=0;
L
Linus Torvalds 已提交
1924

1925
	pr_debug("adding bh b#%llu to stripe s#%llu\n",
L
Linus Torvalds 已提交
1926 1927 1928 1929 1930 1931
		(unsigned long long)bi->bi_sector,
		(unsigned long long)sh->sector);


	spin_lock(&sh->lock);
	spin_lock_irq(&conf->device_lock);
1932
	if (forwrite) {
L
Linus Torvalds 已提交
1933
		bip = &sh->dev[dd_idx].towrite;
1934 1935 1936
		if (*bip == NULL && sh->dev[dd_idx].written == NULL)
			firstwrite = 1;
	} else
L
Linus Torvalds 已提交
1937 1938 1939 1940 1941 1942 1943 1944 1945
		bip = &sh->dev[dd_idx].toread;
	while (*bip && (*bip)->bi_sector < bi->bi_sector) {
		if ((*bip)->bi_sector + ((*bip)->bi_size >> 9) > bi->bi_sector)
			goto overlap;
		bip = & (*bip)->bi_next;
	}
	if (*bip && (*bip)->bi_sector < bi->bi_sector + ((bi->bi_size)>>9))
		goto overlap;

1946
	BUG_ON(*bip && bi->bi_next && (*bip) != bi->bi_next);
L
Linus Torvalds 已提交
1947 1948 1949
	if (*bip)
		bi->bi_next = *bip;
	*bip = bi;
1950
	bi->bi_phys_segments++;
L
Linus Torvalds 已提交
1951 1952 1953
	spin_unlock_irq(&conf->device_lock);
	spin_unlock(&sh->lock);

1954
	pr_debug("added bi b#%llu to stripe s#%llu, disk %d.\n",
L
Linus Torvalds 已提交
1955 1956 1957
		(unsigned long long)bi->bi_sector,
		(unsigned long long)sh->sector, dd_idx);

1958 1959 1960
	if (conf->mddev->bitmap && firstwrite) {
		bitmap_startwrite(conf->mddev->bitmap, sh->sector,
				  STRIPE_SECTORS, 0);
1961
		sh->bm_seq = conf->seq_flush+1;
1962 1963 1964
		set_bit(STRIPE_BIT_DELAY, &sh->state);
	}

L
Linus Torvalds 已提交
1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
	if (forwrite) {
		/* check if page is covered */
		sector_t sector = sh->dev[dd_idx].sector;
		for (bi=sh->dev[dd_idx].towrite;
		     sector < sh->dev[dd_idx].sector + STRIPE_SECTORS &&
			     bi && bi->bi_sector <= sector;
		     bi = r5_next_bio(bi, sh->dev[dd_idx].sector)) {
			if (bi->bi_sector + (bi->bi_size>>9) >= sector)
				sector = bi->bi_sector + (bi->bi_size>>9);
		}
		if (sector >= sh->dev[dd_idx].sector + STRIPE_SECTORS)
			set_bit(R5_OVERWRITE, &sh->dev[dd_idx].flags);
	}
	return 1;

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

1987 1988
static void end_reshape(raid5_conf_t *conf);

1989 1990 1991 1992 1993 1994 1995
static int page_is_zero(struct page *p)
{
	char *a = page_address(p);
	return ((*(u32*)a) == 0 &&
		memcmp(a, a+4, STRIPE_SIZE-4)==0);
}

1996 1997
static void stripe_set_idx(sector_t stripe, raid5_conf_t *conf, int previous,
			    struct stripe_head *sh)
1998
{
1999 2000 2001
	int sectors_per_chunk =
		previous ? (conf->prev_chunk >> 9)
			 : (conf->chunk_size >> 9);
2002
	int dd_idx;
2003
	int chunk_offset = sector_div(stripe, sectors_per_chunk);
2004
	int disks = previous ? conf->previous_raid_disks : conf->raid_disks;
2005

2006 2007
	raid5_compute_sector(conf,
			     stripe * (disks - conf->max_degraded)
2008
			     *sectors_per_chunk + chunk_offset,
2009
			     previous,
2010
			     &dd_idx, sh);
2011 2012
}

2013
static void
2014
handle_failed_stripe(raid5_conf_t *conf, struct stripe_head *sh,
2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047
				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)) {
			mdk_rdev_t *rdev;
			rcu_read_lock();
			rdev = rcu_dereference(conf->disks[i].rdev);
			if (rdev && test_bit(In_sync, &rdev->flags))
				/* multiple read failures in one stripe */
				md_error(conf->mddev, rdev);
			rcu_read_unlock();
		}
		spin_lock_irq(&conf->device_lock);
		/* fail all writes first */
		bi = sh->dev[i].towrite;
		sh->dev[i].towrite = NULL;
		if (bi) {
			s->to_write--;
			bitmap_end = 1;
		}

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

		while (bi && bi->bi_sector <
			sh->dev[i].sector + STRIPE_SECTORS) {
			struct bio *nextbi = r5_next_bio(bi, sh->dev[i].sector);
			clear_bit(BIO_UPTODATE, &bi->bi_flags);
2048
			if (!raid5_dec_bi_phys_segments(bi)) {
2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062
				md_write_end(conf->mddev);
				bi->bi_next = *return_bi;
				*return_bi = bi;
			}
			bi = nextbi;
		}
		/* and fail all 'written' */
		bi = sh->dev[i].written;
		sh->dev[i].written = NULL;
		if (bi) bitmap_end = 1;
		while (bi && bi->bi_sector <
		       sh->dev[i].sector + STRIPE_SECTORS) {
			struct bio *bi2 = r5_next_bio(bi, sh->dev[i].sector);
			clear_bit(BIO_UPTODATE, &bi->bi_flags);
2063
			if (!raid5_dec_bi_phys_segments(bi)) {
2064 2065 2066 2067 2068 2069 2070
				md_write_end(conf->mddev);
				bi->bi_next = *return_bi;
				*return_bi = bi;
			}
			bi = bi2;
		}

2071 2072 2073 2074 2075 2076
		/* 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))) {
2077 2078 2079 2080 2081 2082 2083 2084 2085 2086
			bi = sh->dev[i].toread;
			sh->dev[i].toread = NULL;
			if (test_and_clear_bit(R5_Overlap, &sh->dev[i].flags))
				wake_up(&conf->wait_for_overlap);
			if (bi) s->to_read--;
			while (bi && bi->bi_sector <
			       sh->dev[i].sector + STRIPE_SECTORS) {
				struct bio *nextbi =
					r5_next_bio(bi, sh->dev[i].sector);
				clear_bit(BIO_UPTODATE, &bi->bi_flags);
2087
				if (!raid5_dec_bi_phys_segments(bi)) {
2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099
					bi->bi_next = *return_bi;
					*return_bi = bi;
				}
				bi = nextbi;
			}
		}
		spin_unlock_irq(&conf->device_lock);
		if (bitmap_end)
			bitmap_endwrite(conf->mddev->bitmap, sh->sector,
					STRIPE_SECTORS, 0, 0);
	}

2100 2101 2102
	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);
2103 2104
}

2105 2106 2107 2108 2109
/* fetch_block5 - checks the given member device to see if its data needs
 * to be read or computed to satisfy a request.
 *
 * Returns 1 when no more member devices need to be checked, otherwise returns
 * 0 to tell the loop in handle_stripe_fill5 to continue
2110
 */
2111 2112
static int fetch_block5(struct stripe_head *sh, struct stripe_head_state *s,
			int disk_idx, int disks)
2113 2114 2115 2116 2117 2118
{
	struct r5dev *dev = &sh->dev[disk_idx];
	struct r5dev *failed_dev = &sh->dev[s->failed_num];

	/* is the data in this block needed, and can we get it? */
	if (!test_bit(R5_LOCKED, &dev->flags) &&
2119 2120 2121 2122 2123 2124 2125 2126
	    !test_bit(R5_UPTODATE, &dev->flags) &&
	    (dev->toread ||
	     (dev->towrite && !test_bit(R5_OVERWRITE, &dev->flags)) ||
	     s->syncing || s->expanding ||
	     (s->failed &&
	      (failed_dev->toread ||
	       (failed_dev->towrite &&
		!test_bit(R5_OVERWRITE, &failed_dev->flags)))))) {
2127 2128
		/* We would like to get this block, possibly by computing it,
		 * otherwise read it if the backing disk is insync
2129 2130
		 */
		if ((s->uptodate == disks - 1) &&
2131
		    (s->failed && disk_idx == s->failed_num)) {
2132 2133
			set_bit(STRIPE_COMPUTE_RUN, &sh->state);
			set_bit(STRIPE_OP_COMPUTE_BLK, &s->ops_request);
2134 2135 2136 2137 2138 2139 2140 2141 2142 2143
			set_bit(R5_Wantcompute, &dev->flags);
			sh->ops.target = disk_idx;
			s->req_compute = 1;
			/* Careful: from this point on 'uptodate' is in the eye
			 * of raid5_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.
			 */
			s->uptodate++;
2144
			return 1; /* uptodate + compute == disks */
2145
		} else if (test_bit(R5_Insync, &dev->flags)) {
2146 2147 2148 2149 2150 2151 2152 2153
			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);
		}
	}

2154
	return 0;
2155 2156
}

2157 2158 2159 2160
/**
 * handle_stripe_fill5 - read or compute data to satisfy pending requests.
 */
static void handle_stripe_fill5(struct stripe_head *sh,
2161 2162 2163
			struct stripe_head_state *s, int disks)
{
	int i;
2164 2165 2166 2167 2168

	/* 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
	 */
2169
	if (!test_bit(STRIPE_COMPUTE_RUN, &sh->state) && !sh->check_state &&
2170
	    !sh->reconstruct_state)
2171
		for (i = disks; i--; )
2172
			if (fetch_block5(sh, s, i, disks))
2173
				break;
2174 2175 2176
	set_bit(STRIPE_HANDLE, &sh->state);
}

2177
static void handle_stripe_fill6(struct stripe_head *sh,
2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197
			struct stripe_head_state *s, struct r6_state *r6s,
			int disks)
{
	int i;
	for (i = disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
		if (!test_bit(R5_LOCKED, &dev->flags) &&
		    !test_bit(R5_UPTODATE, &dev->flags) &&
		    (dev->toread || (dev->towrite &&
		     !test_bit(R5_OVERWRITE, &dev->flags)) ||
		     s->syncing || s->expanding ||
		     (s->failed >= 1 &&
		      (sh->dev[r6s->failed_num[0]].toread ||
		       s->to_write)) ||
		     (s->failed >= 2 &&
		      (sh->dev[r6s->failed_num[1]].toread ||
		       s->to_write)))) {
			/* we would like to get this block, possibly
			 * by computing it, but we might not be able to
			 */
2198 2199 2200
			if ((s->uptodate == disks - 1) &&
			    (s->failed && (i == r6s->failed_num[0] ||
					   i == r6s->failed_num[1]))) {
2201
				pr_debug("Computing stripe %llu block %d\n",
2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217
				       (unsigned long long)sh->sector, i);
				compute_block_1(sh, i, 0);
				s->uptodate++;
			} 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 == i)
						continue;
					if (!test_bit(R5_UPTODATE,
					      &sh->dev[other].flags))
						break;
				}
				BUG_ON(other < 0);
2218
				pr_debug("Computing stripe %llu blocks %d,%d\n",
2219 2220 2221 2222 2223 2224 2225 2226
				       (unsigned long long)sh->sector,
				       i, other);
				compute_block_2(sh, i, other);
				s->uptodate += 2;
			} else if (test_bit(R5_Insync, &dev->flags)) {
				set_bit(R5_LOCKED, &dev->flags);
				set_bit(R5_Wantread, &dev->flags);
				s->locked++;
2227
				pr_debug("Reading block %d (sync=%d)\n",
2228 2229 2230 2231 2232 2233 2234 2235
					i, s->syncing);
			}
		}
	}
	set_bit(STRIPE_HANDLE, &sh->state);
}


2236
/* handle_stripe_clean_event
2237 2238 2239 2240
 * 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.
 */
2241
static void handle_stripe_clean_event(raid5_conf_t *conf,
2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254
	struct stripe_head *sh, int disks, struct bio **return_bi)
{
	int i;
	struct r5dev *dev;

	for (i = disks; i--; )
		if (sh->dev[i].written) {
			dev = &sh->dev[i];
			if (!test_bit(R5_LOCKED, &dev->flags) &&
				test_bit(R5_UPTODATE, &dev->flags)) {
				/* We can return any write requests */
				struct bio *wbi, *wbi2;
				int bitmap_end = 0;
2255
				pr_debug("Return write for disc %d\n", i);
2256 2257 2258 2259 2260 2261
				spin_lock_irq(&conf->device_lock);
				wbi = dev->written;
				dev->written = NULL;
				while (wbi && wbi->bi_sector <
					dev->sector + STRIPE_SECTORS) {
					wbi2 = r5_next_bio(wbi, dev->sector);
2262
					if (!raid5_dec_bi_phys_segments(wbi)) {
2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279
						md_write_end(conf->mddev);
						wbi->bi_next = *return_bi;
						*return_bi = wbi;
					}
					wbi = wbi2;
				}
				if (dev->towrite == NULL)
					bitmap_end = 1;
				spin_unlock_irq(&conf->device_lock);
				if (bitmap_end)
					bitmap_endwrite(conf->mddev->bitmap,
							sh->sector,
							STRIPE_SECTORS,
					 !test_bit(STRIPE_DEGRADED, &sh->state),
							0);
			}
		}
2280 2281 2282 2283

	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);
2284 2285
}

2286
static void handle_stripe_dirtying5(raid5_conf_t *conf,
2287 2288 2289 2290 2291 2292 2293 2294
		struct stripe_head *sh,	struct stripe_head_state *s, int disks)
{
	int rmw = 0, rcw = 0, i;
	for (i = disks; i--; ) {
		/* 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) &&
2295 2296
		    !(test_bit(R5_UPTODATE, &dev->flags) ||
		      test_bit(R5_Wantcompute, &dev->flags))) {
2297 2298 2299 2300 2301 2302 2303 2304
			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) &&
2305 2306 2307
		    !(test_bit(R5_UPTODATE, &dev->flags) ||
		    test_bit(R5_Wantcompute, &dev->flags))) {
			if (test_bit(R5_Insync, &dev->flags)) rcw++;
2308 2309 2310 2311
			else
				rcw += 2*disks;
		}
	}
2312
	pr_debug("for sector %llu, rmw=%d rcw=%d\n",
2313 2314 2315 2316 2317 2318 2319 2320
		(unsigned long long)sh->sector, rmw, rcw);
	set_bit(STRIPE_HANDLE, &sh->state);
	if (rmw < rcw && rmw > 0)
		/* prefer read-modify-write, but need to get some data */
		for (i = disks; i--; ) {
			struct r5dev *dev = &sh->dev[i];
			if ((dev->towrite || i == sh->pd_idx) &&
			    !test_bit(R5_LOCKED, &dev->flags) &&
2321 2322
			    !(test_bit(R5_UPTODATE, &dev->flags) ||
			    test_bit(R5_Wantcompute, &dev->flags)) &&
2323 2324 2325
			    test_bit(R5_Insync, &dev->flags)) {
				if (
				  test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
2326
					pr_debug("Read_old block "
2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343
						"%d for r-m-w\n", i);
					set_bit(R5_LOCKED, &dev->flags);
					set_bit(R5_Wantread, &dev->flags);
					s->locked++;
				} else {
					set_bit(STRIPE_DELAYED, &sh->state);
					set_bit(STRIPE_HANDLE, &sh->state);
				}
			}
		}
	if (rcw <= rmw && rcw > 0)
		/* want reconstruct write, but need to get some data */
		for (i = disks; i--; ) {
			struct r5dev *dev = &sh->dev[i];
			if (!test_bit(R5_OVERWRITE, &dev->flags) &&
			    i != sh->pd_idx &&
			    !test_bit(R5_LOCKED, &dev->flags) &&
2344 2345
			    !(test_bit(R5_UPTODATE, &dev->flags) ||
			    test_bit(R5_Wantcompute, &dev->flags)) &&
2346 2347 2348
			    test_bit(R5_Insync, &dev->flags)) {
				if (
				  test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
2349
					pr_debug("Read_old block "
2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362
						"%d for Reconstruct\n", i);
					set_bit(R5_LOCKED, &dev->flags);
					set_bit(R5_Wantread, &dev->flags);
					s->locked++;
				} else {
					set_bit(STRIPE_DELAYED, &sh->state);
					set_bit(STRIPE_HANDLE, &sh->state);
				}
			}
		}
	/* now if nothing is locked, and if we have enough data,
	 * we can start a write request
	 */
2363 2364 2365 2366 2367 2368 2369
	/* 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
	 * subsequent call wants to start a write request.  raid5_run_ops only
	 * handles the case where compute block and postxor are requested
	 * simultaneously.  If this is not the case then new writes need to be
	 * held off until the compute completes.
	 */
2370 2371 2372
	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)))
2373
		schedule_reconstruction5(sh, s, rcw == 0, 0);
2374 2375
}

2376
static void handle_stripe_dirtying6(raid5_conf_t *conf,
2377 2378 2379 2380
		struct stripe_head *sh,	struct stripe_head_state *s,
		struct r6_state *r6s, int disks)
{
	int rcw = 0, must_compute = 0, pd_idx = sh->pd_idx, i;
N
NeilBrown 已提交
2381
	int qd_idx = sh->qd_idx;
2382 2383 2384 2385 2386 2387 2388 2389 2390 2391
	for (i = disks; i--; ) {
		struct r5dev *dev = &sh->dev[i];
		/* Would I have to read this buffer for reconstruct_write */
		if (!test_bit(R5_OVERWRITE, &dev->flags)
		    && i != pd_idx && i != qd_idx
		    && (!test_bit(R5_LOCKED, &dev->flags)
			    ) &&
		    !test_bit(R5_UPTODATE, &dev->flags)) {
			if (test_bit(R5_Insync, &dev->flags)) rcw++;
			else {
2392
				pr_debug("raid6: must_compute: "
2393 2394 2395 2396 2397
					"disk %d flags=%#lx\n", i, dev->flags);
				must_compute++;
			}
		}
	}
2398
	pr_debug("for sector %llu, rcw=%d, must_compute=%d\n",
2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412
	       (unsigned long long)sh->sector, rcw, must_compute);
	set_bit(STRIPE_HANDLE, &sh->state);

	if (rcw > 0)
		/* want reconstruct write, but need to get some data */
		for (i = disks; i--; ) {
			struct r5dev *dev = &sh->dev[i];
			if (!test_bit(R5_OVERWRITE, &dev->flags)
			    && !(s->failed == 0 && (i == pd_idx || i == qd_idx))
			    && !test_bit(R5_LOCKED, &dev->flags) &&
			    !test_bit(R5_UPTODATE, &dev->flags) &&
			    test_bit(R5_Insync, &dev->flags)) {
				if (
				  test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
2413
					pr_debug("Read_old stripe %llu "
2414 2415 2416 2417 2418 2419
						"block %d for Reconstruct\n",
					     (unsigned long long)sh->sector, i);
					set_bit(R5_LOCKED, &dev->flags);
					set_bit(R5_Wantread, &dev->flags);
					s->locked++;
				} else {
2420
					pr_debug("Request delayed stripe %llu "
2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449
						"block %d for Reconstruct\n",
					     (unsigned long long)sh->sector, i);
					set_bit(STRIPE_DELAYED, &sh->state);
					set_bit(STRIPE_HANDLE, &sh->state);
				}
			}
		}
	/* now if nothing is locked, and if we have enough data, we can start a
	 * write request
	 */
	if (s->locked == 0 && rcw == 0 &&
	    !test_bit(STRIPE_BIT_DELAY, &sh->state)) {
		if (must_compute > 0) {
			/* We have failed blocks and need to compute them */
			switch (s->failed) {
			case 0:
				BUG();
			case 1:
				compute_block_1(sh, r6s->failed_num[0], 0);
				break;
			case 2:
				compute_block_2(sh, r6s->failed_num[0],
						r6s->failed_num[1]);
				break;
			default: /* This request should have been failed? */
				BUG();
			}
		}

2450
		pr_debug("Computing parity for stripe %llu\n",
2451 2452 2453 2454 2455
			(unsigned long long)sh->sector);
		compute_parity6(sh, RECONSTRUCT_WRITE);
		/* now every locked buffer is ready to be written */
		for (i = disks; i--; )
			if (test_bit(R5_LOCKED, &sh->dev[i].flags)) {
2456
				pr_debug("Writing stripe %llu block %d\n",
2457 2458 2459 2460
				       (unsigned long long)sh->sector, i);
				s->locked++;
				set_bit(R5_Wantwrite, &sh->dev[i].flags);
			}
2461 2462 2463
		if (s->locked == disks)
			if (!test_and_set_bit(STRIPE_FULL_WRITE, &sh->state))
				atomic_inc(&conf->pending_full_writes);
2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478
		/* after a RECONSTRUCT_WRITE, the stripe MUST be in-sync */
		set_bit(STRIPE_INSYNC, &sh->state);

		if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
			atomic_dec(&conf->preread_active_stripes);
			if (atomic_read(&conf->preread_active_stripes) <
			    IO_THRESHOLD)
				md_wakeup_thread(conf->mddev->thread);
		}
	}
}

static void handle_parity_checks5(raid5_conf_t *conf, struct stripe_head *sh,
				struct stripe_head_state *s, int disks)
{
2479
	struct r5dev *dev = NULL;
2480

2481
	set_bit(STRIPE_HANDLE, &sh->state);
2482

2483 2484 2485
	switch (sh->check_state) {
	case check_state_idle:
		/* start a new check operation if there are no failures */
2486 2487
		if (s->failed == 0) {
			BUG_ON(s->uptodate != disks);
2488 2489
			sh->check_state = check_state_run;
			set_bit(STRIPE_OP_CHECK, &s->ops_request);
2490 2491
			clear_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags);
			s->uptodate--;
2492
			break;
2493
		}
2494 2495 2496 2497 2498 2499 2500 2501 2502 2503
		dev = &sh->dev[s->failed_num];
		/* 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
2504

2505 2506 2507 2508 2509
		/* 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);
2510
		s->locked++;
2511
		set_bit(R5_Wantwrite, &dev->flags);
2512

2513 2514
		clear_bit(STRIPE_DEGRADED, &sh->state);
		set_bit(STRIPE_INSYNC, &sh->state);
2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542
		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
		 */
		if (sh->ops.zero_sum_result == 0)
			/* parity is correct (on disc,
			 * not in buffer any more)
			 */
			set_bit(STRIPE_INSYNC, &sh->state);
		else {
			conf->mddev->resync_mismatches += STRIPE_SECTORS;
			if (test_bit(MD_RECOVERY_CHECK, &conf->mddev->recovery))
				/* don't try to repair!! */
				set_bit(STRIPE_INSYNC, &sh->state);
			else {
				sh->check_state = check_state_compute_run;
2543
				set_bit(STRIPE_COMPUTE_RUN, &sh->state);
2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558
				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;
				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();
2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570
	}
}


static void handle_parity_checks6(raid5_conf_t *conf, struct stripe_head *sh,
				struct stripe_head_state *s,
				struct r6_state *r6s, struct page *tmp_page,
				int disks)
{
	int update_p = 0, update_q = 0;
	struct r5dev *dev;
	int pd_idx = sh->pd_idx;
N
NeilBrown 已提交
2571
	int qd_idx = sh->qd_idx;
2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663

	set_bit(STRIPE_HANDLE, &sh->state);

	BUG_ON(s->failed > 2);
	BUG_ON(s->uptodate < disks);
	/* 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
	 */

	/* If !tmp_page, we cannot do the calculations,
	 * but as we have set STRIPE_HANDLE, we will soon be called
	 * by stripe_handle with a tmp_page - just wait until then.
	 */
	if (tmp_page) {
		if (s->failed == r6s->q_failed) {
			/* The only possible failed device holds 'Q', so it
			 * makes sense to check P (If anything else were failed,
			 * we would have used P to recreate it).
			 */
			compute_block_1(sh, pd_idx, 1);
			if (!page_is_zero(sh->dev[pd_idx].page)) {
				compute_block_1(sh, pd_idx, 0);
				update_p = 1;
			}
		}
		if (!r6s->q_failed && s->failed < 2) {
			/* q is not failed, and we didn't use it to generate
			 * anything, so it makes sense to check it
			 */
			memcpy(page_address(tmp_page),
			       page_address(sh->dev[qd_idx].page),
			       STRIPE_SIZE);
			compute_parity6(sh, UPDATE_PARITY);
			if (memcmp(page_address(tmp_page),
				   page_address(sh->dev[qd_idx].page),
				   STRIPE_SIZE) != 0) {
				clear_bit(STRIPE_INSYNC, &sh->state);
				update_q = 1;
			}
		}
		if (update_p || update_q) {
			conf->mddev->resync_mismatches += STRIPE_SECTORS;
			if (test_bit(MD_RECOVERY_CHECK, &conf->mddev->recovery))
				/* don't try to repair!! */
				update_p = update_q = 0;
		}

		/* now write out any block on a failed drive,
		 * or P or Q if they need it
		 */

		if (s->failed == 2) {
			dev = &sh->dev[r6s->failed_num[1]];
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
		if (s->failed >= 1) {
			dev = &sh->dev[r6s->failed_num[0]];
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}

		if (update_p) {
			dev = &sh->dev[pd_idx];
			s->locked++;
			set_bit(R5_LOCKED, &dev->flags);
			set_bit(R5_Wantwrite, &dev->flags);
		}
		if (update_q) {
			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);
	}
}

static void handle_stripe_expansion(raid5_conf_t *conf, struct stripe_head *sh,
				struct r6_state *r6s)
{
	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.
	 */
2664
	struct dma_async_tx_descriptor *tx = NULL;
2665 2666
	clear_bit(STRIPE_EXPAND_SOURCE, &sh->state);
	for (i = 0; i < sh->disks; i++)
N
NeilBrown 已提交
2667
		if (i != sh->pd_idx && i != sh->qd_idx) {
2668
			int dd_idx, j;
2669 2670
			struct stripe_head *sh2;

2671
			sector_t bn = compute_blocknr(sh, i, 1);
2672 2673
			sector_t s = raid5_compute_sector(conf, bn, 0,
							  &dd_idx, NULL);
2674
			sh2 = get_active_stripe(conf, s, 0, 1, 1);
2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686
			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;
			}
2687 2688 2689 2690 2691 2692

			/* place all the copies on one channel */
			tx = async_memcpy(sh2->dev[dd_idx].page,
				sh->dev[i].page, 0, 0, STRIPE_SIZE,
				ASYNC_TX_DEP_ACK, tx, NULL, NULL);

2693 2694 2695 2696
			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 &&
2697
				    (!r6s || j != sh2->qd_idx) &&
2698 2699 2700 2701 2702 2703 2704
				    !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);
2705

2706
		}
2707 2708 2709 2710 2711
	/* done submitting copies, wait for them to complete */
	if (tx) {
		async_tx_ack(tx);
		dma_wait_for_async_tx(tx);
	}
2712
}
L
Linus Torvalds 已提交
2713

2714

L
Linus Torvalds 已提交
2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730
/*
 * handle_stripe - do things to a stripe.
 *
 * We lock the stripe and then examine the state of various bits
 * to see what needs to be done.
 * Possible results:
 *    return some read request which now have data
 *    return some write requests which are safely on disc
 *    schedule a read on some buffers
 *    schedule a write of some buffers
 *    return confirmation of parity correctness
 *
 * buffers are taken off read_list or write_list, and bh_cache buffers
 * get BH_Lock set before the stripe lock is released.
 *
 */
2731

2732
static bool handle_stripe5(struct stripe_head *sh)
L
Linus Torvalds 已提交
2733 2734
{
	raid5_conf_t *conf = sh->raid_conf;
2735 2736 2737
	int disks = sh->disks, i;
	struct bio *return_bi = NULL;
	struct stripe_head_state s;
L
Linus Torvalds 已提交
2738
	struct r5dev *dev;
2739
	mdk_rdev_t *blocked_rdev = NULL;
2740
	int prexor;
L
Linus Torvalds 已提交
2741

2742
	memset(&s, 0, sizeof(s));
2743 2744 2745 2746
	pr_debug("handling stripe %llu, state=%#lx cnt=%d, pd_idx=%d check:%d "
		 "reconstruct:%d\n", (unsigned long long)sh->sector, sh->state,
		 atomic_read(&sh->count), sh->pd_idx, sh->check_state,
		 sh->reconstruct_state);
L
Linus Torvalds 已提交
2747 2748 2749 2750 2751

	spin_lock(&sh->lock);
	clear_bit(STRIPE_HANDLE, &sh->state);
	clear_bit(STRIPE_DELAYED, &sh->state);

2752 2753 2754
	s.syncing = test_bit(STRIPE_SYNCING, &sh->state);
	s.expanding = test_bit(STRIPE_EXPAND_SOURCE, &sh->state);
	s.expanded = test_bit(STRIPE_EXPAND_READY, &sh->state);
N
Neil Brown 已提交
2755

2756
	/* Now to look around and see what can be done */
2757
	rcu_read_lock();
L
Linus Torvalds 已提交
2758 2759
	for (i=disks; i--; ) {
		mdk_rdev_t *rdev;
2760
		struct r5dev *dev = &sh->dev[i];
L
Linus Torvalds 已提交
2761 2762
		clear_bit(R5_Insync, &dev->flags);

2763 2764 2765 2766 2767 2768 2769
		pr_debug("check %d: state 0x%lx toread %p read %p write %p "
			"written %p\n",	i, dev->flags, dev->toread, dev->read,
			dev->towrite, dev->written);

		/* maybe we can request a biofill operation
		 *
		 * new wantfill requests are only permitted while
2770
		 * ops_complete_biofill is guaranteed to be inactive
2771 2772
		 */
		if (test_bit(R5_UPTODATE, &dev->flags) && dev->toread &&
2773
		    !test_bit(STRIPE_BIOFILL_RUN, &sh->state))
2774
			set_bit(R5_Wantfill, &dev->flags);
L
Linus Torvalds 已提交
2775 2776

		/* now count some things */
2777 2778
		if (test_bit(R5_LOCKED, &dev->flags)) s.locked++;
		if (test_bit(R5_UPTODATE, &dev->flags)) s.uptodate++;
2779
		if (test_bit(R5_Wantcompute, &dev->flags)) s.compute++;
L
Linus Torvalds 已提交
2780

2781 2782 2783
		if (test_bit(R5_Wantfill, &dev->flags))
			s.to_fill++;
		else if (dev->toread)
2784
			s.to_read++;
L
Linus Torvalds 已提交
2785
		if (dev->towrite) {
2786
			s.to_write++;
L
Linus Torvalds 已提交
2787
			if (!test_bit(R5_OVERWRITE, &dev->flags))
2788
				s.non_overwrite++;
L
Linus Torvalds 已提交
2789
		}
2790 2791
		if (dev->written)
			s.written++;
2792
		rdev = rcu_dereference(conf->disks[i].rdev);
2793 2794
		if (blocked_rdev == NULL &&
		    rdev && unlikely(test_bit(Blocked, &rdev->flags))) {
2795 2796 2797
			blocked_rdev = rdev;
			atomic_inc(&rdev->nr_pending);
		}
2798
		if (!rdev || !test_bit(In_sync, &rdev->flags)) {
N
NeilBrown 已提交
2799
			/* The ReadError flag will just be confusing now */
2800 2801 2802
			clear_bit(R5_ReadError, &dev->flags);
			clear_bit(R5_ReWrite, &dev->flags);
		}
2803
		if (!rdev || !test_bit(In_sync, &rdev->flags)
2804
		    || test_bit(R5_ReadError, &dev->flags)) {
2805 2806
			s.failed++;
			s.failed_num = i;
L
Linus Torvalds 已提交
2807 2808 2809
		} else
			set_bit(R5_Insync, &dev->flags);
	}
2810
	rcu_read_unlock();
2811

2812
	if (unlikely(blocked_rdev)) {
2813 2814 2815 2816 2817 2818 2819 2820
		if (s.syncing || s.expanding || s.expanded ||
		    s.to_write || s.written) {
			set_bit(STRIPE_HANDLE, &sh->state);
			goto unlock;
		}
		/* There is nothing for the blocked_rdev to block */
		rdev_dec_pending(blocked_rdev, conf->mddev);
		blocked_rdev = NULL;
2821 2822
	}

2823 2824 2825 2826
	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);
	}
2827

2828
	pr_debug("locked=%d uptodate=%d to_read=%d"
L
Linus Torvalds 已提交
2829
		" to_write=%d failed=%d failed_num=%d\n",
2830 2831
		s.locked, s.uptodate, s.to_read, s.to_write,
		s.failed, s.failed_num);
L
Linus Torvalds 已提交
2832 2833 2834
	/* check if the array has lost two devices and, if so, some requests might
	 * need to be failed
	 */
2835
	if (s.failed > 1 && s.to_read+s.to_write+s.written)
2836
		handle_failed_stripe(conf, sh, &s, disks, &return_bi);
2837
	if (s.failed > 1 && s.syncing) {
L
Linus Torvalds 已提交
2838 2839
		md_done_sync(conf->mddev, STRIPE_SECTORS,0);
		clear_bit(STRIPE_SYNCING, &sh->state);
2840
		s.syncing = 0;
L
Linus Torvalds 已提交
2841 2842 2843 2844 2845 2846
	}

	/* might be able to return some write requests if the parity block
	 * is safe, or on a failed drive
	 */
	dev = &sh->dev[sh->pd_idx];
2847 2848 2849 2850 2851
	if ( s.written &&
	     ((test_bit(R5_Insync, &dev->flags) &&
	       !test_bit(R5_LOCKED, &dev->flags) &&
	       test_bit(R5_UPTODATE, &dev->flags)) ||
	       (s.failed == 1 && s.failed_num == sh->pd_idx)))
2852
		handle_stripe_clean_event(conf, sh, disks, &return_bi);
L
Linus Torvalds 已提交
2853 2854 2855 2856 2857

	/* 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.
	 */
2858
	if (s.to_read || s.non_overwrite ||
2859
	    (s.syncing && (s.uptodate + s.compute < disks)) || s.expanding)
2860
		handle_stripe_fill5(sh, &s, disks);
L
Linus Torvalds 已提交
2861

2862 2863 2864
	/* Now we check to see if any write operations have recently
	 * completed
	 */
2865
	prexor = 0;
2866
	if (sh->reconstruct_state == reconstruct_state_prexor_drain_result)
2867
		prexor = 1;
2868 2869
	if (sh->reconstruct_state == reconstruct_state_drain_result ||
	    sh->reconstruct_state == reconstruct_state_prexor_drain_result) {
2870
		sh->reconstruct_state = reconstruct_state_idle;
2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881

		/* All the 'written' buffers and the parity block are ready to
		 * be written back to disk
		 */
		BUG_ON(!test_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags));
		for (i = disks; i--; ) {
			dev = &sh->dev[i];
			if (test_bit(R5_LOCKED, &dev->flags) &&
				(i == sh->pd_idx || dev->written)) {
				pr_debug("Writing block %d\n", i);
				set_bit(R5_Wantwrite, &dev->flags);
2882 2883
				if (prexor)
					continue;
2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902
				if (!test_bit(R5_Insync, &dev->flags) ||
				    (i == sh->pd_idx && s.failed == 0))
					set_bit(STRIPE_INSYNC, &sh->state);
			}
		}
		if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
			atomic_dec(&conf->preread_active_stripes);
			if (atomic_read(&conf->preread_active_stripes) <
				IO_THRESHOLD)
				md_wakeup_thread(conf->mddev->thread);
		}
	}

	/* 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.
	 */
2903
	if (s.to_write && !sh->reconstruct_state && !sh->check_state)
2904
		handle_stripe_dirtying5(conf, sh, &s, disks);
L
Linus Torvalds 已提交
2905 2906

	/* maybe we need to check and possibly fix the parity for this stripe
2907 2908 2909
	 * 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.
L
Linus Torvalds 已提交
2910
	 */
2911 2912
	if (sh->check_state ||
	    (s.syncing && s.locked == 0 &&
2913
	     !test_bit(STRIPE_COMPUTE_RUN, &sh->state) &&
2914
	     !test_bit(STRIPE_INSYNC, &sh->state)))
2915
		handle_parity_checks5(conf, sh, &s, disks);
2916

2917
	if (s.syncing && s.locked == 0 && test_bit(STRIPE_INSYNC, &sh->state)) {
L
Linus Torvalds 已提交
2918 2919 2920
		md_done_sync(conf->mddev, STRIPE_SECTORS,1);
		clear_bit(STRIPE_SYNCING, &sh->state);
	}
2921 2922 2923 2924

	/* If the failed drive is just a ReadError, then we might need to progress
	 * the repair/check process
	 */
2925 2926 2927 2928
	if (s.failed == 1 && !conf->mddev->ro &&
	    test_bit(R5_ReadError, &sh->dev[s.failed_num].flags)
	    && !test_bit(R5_LOCKED, &sh->dev[s.failed_num].flags)
	    && test_bit(R5_UPTODATE, &sh->dev[s.failed_num].flags)
2929
		) {
2930
		dev = &sh->dev[s.failed_num];
2931 2932 2933 2934
		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);
2935
			s.locked++;
2936 2937 2938 2939
		} else {
			/* let's read it back */
			set_bit(R5_Wantread, &dev->flags);
			set_bit(R5_LOCKED, &dev->flags);
2940
			s.locked++;
2941 2942 2943
		}
	}

2944 2945
	/* Finish reconstruct operations initiated by the expansion process */
	if (sh->reconstruct_state == reconstruct_state_result) {
2946
		struct stripe_head *sh2
2947
			= get_active_stripe(conf, sh->sector, 1, 1, 1);
2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962
		if (sh2 && test_bit(STRIPE_EXPAND_SOURCE, &sh2->state)) {
			/* sh cannot be written until sh2 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,
					      &sh2->state))
				atomic_inc(&conf->preread_active_stripes);
			release_stripe(sh2);
			goto unlock;
		}
		if (sh2)
			release_stripe(sh2);

2963
		sh->reconstruct_state = reconstruct_state_idle;
2964
		clear_bit(STRIPE_EXPANDING, &sh->state);
D
Dan Williams 已提交
2965
		for (i = conf->raid_disks; i--; ) {
2966
			set_bit(R5_Wantwrite, &sh->dev[i].flags);
D
Dan Williams 已提交
2967
			set_bit(R5_LOCKED, &sh->dev[i].flags);
2968
			s.locked++;
D
Dan Williams 已提交
2969
		}
2970 2971 2972
	}

	if (s.expanded && test_bit(STRIPE_EXPANDING, &sh->state) &&
2973
	    !sh->reconstruct_state) {
2974 2975
		/* Need to write out all blocks after computing parity */
		sh->disks = conf->raid_disks;
2976
		stripe_set_idx(sh->sector, conf, 0, sh);
2977
		schedule_reconstruction5(sh, &s, 1, 1);
2978
	} else if (s.expanded && !sh->reconstruct_state && s.locked == 0) {
2979
		clear_bit(STRIPE_EXPAND_READY, &sh->state);
2980
		atomic_dec(&conf->reshape_stripes);
2981 2982 2983 2984
		wake_up(&conf->wait_for_overlap);
		md_done_sync(conf->mddev, STRIPE_SECTORS, 1);
	}

2985
	if (s.expanding && s.locked == 0 &&
2986
	    !test_bit(STRIPE_COMPUTE_RUN, &sh->state))
2987
		handle_stripe_expansion(conf, sh, NULL);
2988

2989
 unlock:
L
Linus Torvalds 已提交
2990 2991
	spin_unlock(&sh->lock);

2992 2993 2994 2995
	/* wait for this device to become unblocked */
	if (unlikely(blocked_rdev))
		md_wait_for_blocked_rdev(blocked_rdev, conf->mddev);

2996 2997
	if (s.ops_request)
		raid5_run_ops(sh, s.ops_request);
2998

2999
	ops_run_io(sh, &s);
L
Linus Torvalds 已提交
3000

3001
	return_io(return_bi);
3002 3003

	return blocked_rdev == NULL;
L
Linus Torvalds 已提交
3004 3005
}

3006
static bool handle_stripe6(struct stripe_head *sh, struct page *tmp_page)
L
Linus Torvalds 已提交
3007
{
3008
	raid5_conf_t *conf = sh->raid_conf;
3009
	int disks = sh->disks;
3010
	struct bio *return_bi = NULL;
N
NeilBrown 已提交
3011
	int i, pd_idx = sh->pd_idx, qd_idx = sh->qd_idx;
3012 3013
	struct stripe_head_state s;
	struct r6_state r6s;
3014
	struct r5dev *dev, *pdev, *qdev;
3015
	mdk_rdev_t *blocked_rdev = NULL;
L
Linus Torvalds 已提交
3016

3017
	pr_debug("handling stripe %llu, state=%#lx cnt=%d, "
3018 3019
		"pd_idx=%d, qd_idx=%d\n",
	       (unsigned long long)sh->sector, sh->state,
N
NeilBrown 已提交
3020
	       atomic_read(&sh->count), pd_idx, qd_idx);
3021
	memset(&s, 0, sizeof(s));
3022

3023 3024 3025 3026
	spin_lock(&sh->lock);
	clear_bit(STRIPE_HANDLE, &sh->state);
	clear_bit(STRIPE_DELAYED, &sh->state);

3027 3028 3029
	s.syncing = test_bit(STRIPE_SYNCING, &sh->state);
	s.expanding = test_bit(STRIPE_EXPAND_SOURCE, &sh->state);
	s.expanded = test_bit(STRIPE_EXPAND_READY, &sh->state);
3030
	/* Now to look around and see what can be done */
L
Linus Torvalds 已提交
3031 3032

	rcu_read_lock();
3033 3034 3035 3036
	for (i=disks; i--; ) {
		mdk_rdev_t *rdev;
		dev = &sh->dev[i];
		clear_bit(R5_Insync, &dev->flags);
L
Linus Torvalds 已提交
3037

3038
		pr_debug("check %d: state 0x%lx read %p write %p written %p\n",
3039 3040 3041 3042
			i, dev->flags, dev->toread, dev->towrite, dev->written);
		/* maybe we can reply to a read */
		if (test_bit(R5_UPTODATE, &dev->flags) && dev->toread) {
			struct bio *rbi, *rbi2;
3043
			pr_debug("Return read for disc %d\n", i);
3044 3045 3046 3047 3048 3049 3050 3051 3052 3053
			spin_lock_irq(&conf->device_lock);
			rbi = dev->toread;
			dev->toread = NULL;
			if (test_and_clear_bit(R5_Overlap, &dev->flags))
				wake_up(&conf->wait_for_overlap);
			spin_unlock_irq(&conf->device_lock);
			while (rbi && rbi->bi_sector < dev->sector + STRIPE_SECTORS) {
				copy_data(0, rbi, dev->page, dev->sector);
				rbi2 = r5_next_bio(rbi, dev->sector);
				spin_lock_irq(&conf->device_lock);
3054
				if (!raid5_dec_bi_phys_segments(rbi)) {
3055 3056 3057 3058 3059 3060 3061
					rbi->bi_next = return_bi;
					return_bi = rbi;
				}
				spin_unlock_irq(&conf->device_lock);
				rbi = rbi2;
			}
		}
L
Linus Torvalds 已提交
3062

3063
		/* now count some things */
3064 3065
		if (test_bit(R5_LOCKED, &dev->flags)) s.locked++;
		if (test_bit(R5_UPTODATE, &dev->flags)) s.uptodate++;
L
Linus Torvalds 已提交
3066

3067

3068 3069
		if (dev->toread)
			s.to_read++;
3070
		if (dev->towrite) {
3071
			s.to_write++;
3072
			if (!test_bit(R5_OVERWRITE, &dev->flags))
3073
				s.non_overwrite++;
3074
		}
3075 3076
		if (dev->written)
			s.written++;
3077
		rdev = rcu_dereference(conf->disks[i].rdev);
3078 3079
		if (blocked_rdev == NULL &&
		    rdev && unlikely(test_bit(Blocked, &rdev->flags))) {
3080 3081 3082
			blocked_rdev = rdev;
			atomic_inc(&rdev->nr_pending);
		}
3083 3084 3085 3086
		if (!rdev || !test_bit(In_sync, &rdev->flags)) {
			/* 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
3087
		}
3088 3089
		if (!rdev || !test_bit(In_sync, &rdev->flags)
		    || test_bit(R5_ReadError, &dev->flags)) {
3090 3091 3092
			if (s.failed < 2)
				r6s.failed_num[s.failed] = i;
			s.failed++;
3093 3094
		} else
			set_bit(R5_Insync, &dev->flags);
L
Linus Torvalds 已提交
3095 3096
	}
	rcu_read_unlock();
3097 3098

	if (unlikely(blocked_rdev)) {
3099 3100 3101 3102 3103 3104 3105 3106
		if (s.syncing || s.expanding || s.expanded ||
		    s.to_write || s.written) {
			set_bit(STRIPE_HANDLE, &sh->state);
			goto unlock;
		}
		/* There is nothing for the blocked_rdev to block */
		rdev_dec_pending(blocked_rdev, conf->mddev);
		blocked_rdev = NULL;
3107
	}
3108

3109
	pr_debug("locked=%d uptodate=%d to_read=%d"
3110
	       " to_write=%d failed=%d failed_num=%d,%d\n",
3111 3112 3113 3114
	       s.locked, s.uptodate, s.to_read, s.to_write, s.failed,
	       r6s.failed_num[0], r6s.failed_num[1]);
	/* check if the array has lost >2 devices and, if so, some requests
	 * might need to be failed
3115
	 */
3116
	if (s.failed > 2 && s.to_read+s.to_write+s.written)
3117
		handle_failed_stripe(conf, sh, &s, disks, &return_bi);
3118
	if (s.failed > 2 && s.syncing) {
3119 3120
		md_done_sync(conf->mddev, STRIPE_SECTORS,0);
		clear_bit(STRIPE_SYNCING, &sh->state);
3121
		s.syncing = 0;
3122 3123 3124 3125 3126 3127 3128
	}

	/*
	 * might be able to return some write requests if the parity blocks
	 * are safe, or on a failed drive
	 */
	pdev = &sh->dev[pd_idx];
3129 3130
	r6s.p_failed = (s.failed >= 1 && r6s.failed_num[0] == pd_idx)
		|| (s.failed >= 2 && r6s.failed_num[1] == pd_idx);
N
NeilBrown 已提交
3131 3132 3133
	qdev = &sh->dev[qd_idx];
	r6s.q_failed = (s.failed >= 1 && r6s.failed_num[0] == qd_idx)
		|| (s.failed >= 2 && r6s.failed_num[1] == qd_idx);
3134 3135 3136

	if ( s.written &&
	     ( r6s.p_failed || ((test_bit(R5_Insync, &pdev->flags)
3137
			     && !test_bit(R5_LOCKED, &pdev->flags)
3138 3139
			     && test_bit(R5_UPTODATE, &pdev->flags)))) &&
	     ( r6s.q_failed || ((test_bit(R5_Insync, &qdev->flags)
3140
			     && !test_bit(R5_LOCKED, &qdev->flags)
3141
			     && test_bit(R5_UPTODATE, &qdev->flags)))))
3142
		handle_stripe_clean_event(conf, sh, disks, &return_bi);
3143 3144 3145 3146 3147

	/* 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.
	 */
3148 3149
	if (s.to_read || s.non_overwrite || (s.to_write && s.failed) ||
	    (s.syncing && (s.uptodate < disks)) || s.expanding)
3150
		handle_stripe_fill6(sh, &s, &r6s, disks);
3151 3152

	/* now to consider writing and what else, if anything should be read */
3153
	if (s.to_write)
3154
		handle_stripe_dirtying6(conf, sh, &s, &r6s, disks);
3155 3156

	/* maybe we need to check and possibly fix the parity for this stripe
3157 3158
	 * Any reads will already have been scheduled, so we just see if enough
	 * data is available
3159
	 */
3160 3161
	if (s.syncing && s.locked == 0 && !test_bit(STRIPE_INSYNC, &sh->state))
		handle_parity_checks6(conf, sh, &s, &r6s, tmp_page, disks);
3162

3163
	if (s.syncing && s.locked == 0 && test_bit(STRIPE_INSYNC, &sh->state)) {
3164 3165 3166 3167 3168 3169 3170
		md_done_sync(conf->mddev, STRIPE_SECTORS,1);
		clear_bit(STRIPE_SYNCING, &sh->state);
	}

	/* If the failed drives are just a ReadError, then we might need
	 * to progress the repair/check process
	 */
3171 3172 3173
	if (s.failed <= 2 && !conf->mddev->ro)
		for (i = 0; i < s.failed; i++) {
			dev = &sh->dev[r6s.failed_num[i]];
3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188
			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);
				} else {
					/* let's read it back */
					set_bit(R5_Wantread, &dev->flags);
					set_bit(R5_LOCKED, &dev->flags);
				}
			}
		}
3189

3190
	if (s.expanded && test_bit(STRIPE_EXPANDING, &sh->state)) {
3191
		struct stripe_head *sh2
3192
			= get_active_stripe(conf, sh->sector, 1, 1, 1);
3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207
		if (sh2 && test_bit(STRIPE_EXPAND_SOURCE, &sh2->state)) {
			/* sh cannot be written until sh2 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,
					      &sh2->state))
				atomic_inc(&conf->preread_active_stripes);
			release_stripe(sh2);
			goto unlock;
		}
		if (sh2)
			release_stripe(sh2);

3208 3209
		/* Need to write out all blocks after computing P&Q */
		sh->disks = conf->raid_disks;
3210
		stripe_set_idx(sh->sector, conf, 0, sh);
3211 3212 3213
		compute_parity6(sh, RECONSTRUCT_WRITE);
		for (i = conf->raid_disks ; i-- ;  ) {
			set_bit(R5_LOCKED, &sh->dev[i].flags);
3214
			s.locked++;
3215 3216 3217
			set_bit(R5_Wantwrite, &sh->dev[i].flags);
		}
		clear_bit(STRIPE_EXPANDING, &sh->state);
3218
	} else if (s.expanded) {
3219 3220 3221 3222 3223 3224
		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);
	}

3225
	if (s.expanding && s.locked == 0 &&
3226
	    !test_bit(STRIPE_COMPUTE_RUN, &sh->state))
3227
		handle_stripe_expansion(conf, sh, &r6s);
3228

3229
 unlock:
3230 3231
	spin_unlock(&sh->lock);

3232 3233 3234 3235
	/* wait for this device to become unblocked */
	if (unlikely(blocked_rdev))
		md_wait_for_blocked_rdev(blocked_rdev, conf->mddev);

D
Dan Williams 已提交
3236
	ops_run_io(sh, &s);
3237

D
Dan Williams 已提交
3238
	return_io(return_bi);
3239 3240

	return blocked_rdev == NULL;
3241 3242
}

3243 3244
/* returns true if the stripe was handled */
static bool handle_stripe(struct stripe_head *sh, struct page *tmp_page)
3245 3246
{
	if (sh->raid_conf->level == 6)
3247
		return handle_stripe6(sh, tmp_page);
3248
	else
3249
		return handle_stripe5(sh);
3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264
}



static void raid5_activate_delayed(raid5_conf_t *conf)
{
	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);
3265
			list_add_tail(&sh->lru, &conf->hold_list);
3266
		}
3267 3268
	} else
		blk_plug_device(conf->mddev->queue);
3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286
}

static void activate_bit_delay(raid5_conf_t *conf)
{
	/* device_lock is held */
	struct list_head head;
	list_add(&head, &conf->bitmap_list);
	list_del_init(&conf->bitmap_list);
	while (!list_empty(&head)) {
		struct stripe_head *sh = list_entry(head.next, struct stripe_head, lru);
		list_del_init(&sh->lru);
		atomic_inc(&sh->count);
		__release_stripe(conf, sh);
	}
}

static void unplug_slaves(mddev_t *mddev)
{
3287
	raid5_conf_t *conf = mddev->private;
3288 3289 3290
	int i;

	rcu_read_lock();
3291
	for (i = 0; i < conf->raid_disks; i++) {
3292 3293
		mdk_rdev_t *rdev = rcu_dereference(conf->disks[i].rdev);
		if (rdev && !test_bit(Faulty, &rdev->flags) && atomic_read(&rdev->nr_pending)) {
3294
			struct request_queue *r_queue = bdev_get_queue(rdev->bdev);
3295 3296 3297 3298

			atomic_inc(&rdev->nr_pending);
			rcu_read_unlock();

3299
			blk_unplug(r_queue);
3300 3301 3302 3303 3304 3305 3306 3307

			rdev_dec_pending(rdev, mddev);
			rcu_read_lock();
		}
	}
	rcu_read_unlock();
}

3308
static void raid5_unplug_device(struct request_queue *q)
3309 3310
{
	mddev_t *mddev = q->queuedata;
3311
	raid5_conf_t *conf = mddev->private;
3312 3313 3314 3315 3316 3317 3318
	unsigned long flags;

	spin_lock_irqsave(&conf->device_lock, flags);

	if (blk_remove_plug(q)) {
		conf->seq_flush++;
		raid5_activate_delayed(conf);
3319
	}
L
Linus Torvalds 已提交
3320 3321 3322 3323 3324 3325 3326
	md_wakeup_thread(mddev->thread);

	spin_unlock_irqrestore(&conf->device_lock, flags);

	unplug_slaves(mddev);
}

3327 3328 3329
static int raid5_congested(void *data, int bits)
{
	mddev_t *mddev = data;
3330
	raid5_conf_t *conf = mddev->private;
3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344

	/* No difference between reads and writes.  Just check
	 * how busy the stripe_cache is
	 */
	if (conf->inactive_blocked)
		return 1;
	if (conf->quiesce)
		return 1;
	if (list_empty_careful(&conf->inactive_list))
		return 1;

	return 0;
}

3345 3346 3347
/* We want read requests to align with chunks where possible,
 * but write requests don't need to.
 */
3348 3349 3350
static int raid5_mergeable_bvec(struct request_queue *q,
				struct bvec_merge_data *bvm,
				struct bio_vec *biovec)
3351 3352
{
	mddev_t *mddev = q->queuedata;
3353
	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
3354 3355
	int max;
	unsigned int chunk_sectors = mddev->chunk_size >> 9;
3356
	unsigned int bio_sectors = bvm->bi_size >> 9;
3357

3358
	if ((bvm->bi_rw & 1) == WRITE)
3359 3360
		return biovec->bv_len; /* always allow writes to be mergeable */

3361 3362
	if (mddev->new_chunk < mddev->chunk_size)
		chunk_sectors = mddev->new_chunk >> 9;
3363 3364 3365 3366 3367 3368 3369 3370
	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;
}

3371 3372 3373 3374 3375 3376 3377

static int in_chunk_boundary(mddev_t *mddev, struct bio *bio)
{
	sector_t sector = bio->bi_sector + get_start_sect(bio->bi_bdev);
	unsigned int chunk_sectors = mddev->chunk_size >> 9;
	unsigned int bio_sectors = bio->bi_size >> 9;

3378 3379
	if (mddev->new_chunk < mddev->chunk_size)
		chunk_sectors = mddev->new_chunk >> 9;
3380 3381 3382 3383
	return  chunk_sectors >=
		((sector & (chunk_sectors - 1)) + bio_sectors);
}

3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412
/*
 *  add bio to the retry LIFO  ( in O(1) ... we are in interrupt )
 *  later sampled by raid5d.
 */
static void add_bio_to_retry(struct bio *bi,raid5_conf_t *conf)
{
	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);
}


static struct bio *remove_bio_from_retry(raid5_conf_t *conf)
{
	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) {
3413
		conf->retry_read_aligned_list = bi->bi_next;
3414
		bi->bi_next = NULL;
3415 3416 3417 3418
		/*
		 * this sets the active strip count to 1 and the processed
		 * strip count to zero (upper 8 bits)
		 */
3419 3420 3421 3422 3423 3424 3425
		bi->bi_phys_segments = 1; /* biased count of active stripes */
	}

	return bi;
}


3426 3427 3428 3429 3430 3431
/*
 *  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..
 */
3432
static void raid5_align_endio(struct bio *bi, int error)
3433 3434
{
	struct bio* raid_bi  = bi->bi_private;
3435 3436 3437 3438 3439
	mddev_t *mddev;
	raid5_conf_t *conf;
	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
	mdk_rdev_t *rdev;

3440
	bio_put(bi);
3441 3442

	mddev = raid_bi->bi_bdev->bd_disk->queue->queuedata;
3443
	conf = mddev->private;
3444 3445 3446 3447 3448 3449
	rdev = (void*)raid_bi->bi_next;
	raid_bi->bi_next = NULL;

	rdev_dec_pending(rdev, conf->mddev);

	if (!error && uptodate) {
3450
		bio_endio(raid_bi, 0);
3451 3452
		if (atomic_dec_and_test(&conf->active_aligned_reads))
			wake_up(&conf->wait_for_stripe);
3453
		return;
3454 3455 3456
	}


3457
	pr_debug("raid5_align_endio : io error...handing IO for a retry\n");
3458 3459

	add_bio_to_retry(raid_bi, conf);
3460 3461
}

3462 3463
static int bio_fits_rdev(struct bio *bi)
{
3464
	struct request_queue *q = bdev_get_queue(bi->bi_bdev);
3465

3466
	if ((bi->bi_size>>9) > queue_max_sectors(q))
3467 3468
		return 0;
	blk_recount_segments(q, bi);
3469
	if (bi->bi_phys_segments > queue_max_phys_segments(q))
3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481
		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;
}


3482
static int chunk_aligned_read(struct request_queue *q, struct bio * raid_bio)
3483 3484
{
	mddev_t *mddev = q->queuedata;
3485
	raid5_conf_t *conf = mddev->private;
3486
	unsigned int dd_idx;
3487 3488 3489 3490
	struct bio* align_bi;
	mdk_rdev_t *rdev;

	if (!in_chunk_boundary(mddev, raid_bio)) {
3491
		pr_debug("chunk_aligned_read : non aligned\n");
3492 3493 3494
		return 0;
	}
	/*
3495
	 * use bio_clone to make a copy of the bio
3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508
	 */
	align_bi = bio_clone(raid_bio, GFP_NOIO);
	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
	 */
3509 3510
	align_bi->bi_sector =  raid5_compute_sector(conf, raid_bio->bi_sector,
						    0,
3511
						    &dd_idx, NULL);
3512 3513 3514 3515 3516 3517

	rcu_read_lock();
	rdev = rcu_dereference(conf->disks[dd_idx].rdev);
	if (rdev && test_bit(In_sync, &rdev->flags)) {
		atomic_inc(&rdev->nr_pending);
		rcu_read_unlock();
3518 3519 3520 3521 3522
		raid_bio->bi_next = (void*)rdev;
		align_bi->bi_bdev =  rdev->bdev;
		align_bi->bi_flags &= ~(1 << BIO_SEG_VALID);
		align_bi->bi_sector += rdev->data_offset;

3523 3524 3525 3526 3527 3528 3529
		if (!bio_fits_rdev(align_bi)) {
			/* too big in some way */
			bio_put(align_bi);
			rdev_dec_pending(rdev, mddev);
			return 0;
		}

3530 3531 3532 3533 3534 3535 3536
		spin_lock_irq(&conf->device_lock);
		wait_event_lock_irq(conf->wait_for_stripe,
				    conf->quiesce == 0,
				    conf->device_lock, /* nothing */);
		atomic_inc(&conf->active_aligned_reads);
		spin_unlock_irq(&conf->device_lock);

3537 3538 3539 3540
		generic_make_request(align_bi);
		return 1;
	} else {
		rcu_read_unlock();
3541
		bio_put(align_bi);
3542 3543 3544 3545
		return 0;
	}
}

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 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597
/* __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.
 */
static struct stripe_head *__get_priority_stripe(raid5_conf_t *conf)
{
	struct stripe_head *sh;

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

	if (!list_empty(&conf->handle_list)) {
		sh = list_entry(conf->handle_list.next, typeof(*sh), lru);

		if (list_empty(&conf->hold_list))
			conf->bypass_count = 0;
		else if (!test_bit(STRIPE_IO_STARTED, &sh->state)) {
			if (conf->hold_list.next == conf->last_hold)
				conf->bypass_count++;
			else {
				conf->last_hold = conf->hold_list.next;
				conf->bypass_count -= conf->bypass_threshold;
				if (conf->bypass_count < 0)
					conf->bypass_count = 0;
			}
		}
	} else if (!list_empty(&conf->hold_list) &&
		   ((conf->bypass_threshold &&
		     conf->bypass_count > conf->bypass_threshold) ||
		    atomic_read(&conf->pending_full_writes) == 0)) {
		sh = list_entry(conf->hold_list.next,
				typeof(*sh), lru);
		conf->bypass_count -= conf->bypass_threshold;
		if (conf->bypass_count < 0)
			conf->bypass_count = 0;
	} else
		return NULL;

	list_del_init(&sh->lru);
	atomic_inc(&sh->count);
	BUG_ON(atomic_read(&sh->count) != 1);
	return sh;
}
3598

3599
static int make_request(struct request_queue *q, struct bio * bi)
L
Linus Torvalds 已提交
3600 3601
{
	mddev_t *mddev = q->queuedata;
3602
	raid5_conf_t *conf = mddev->private;
3603
	int dd_idx;
L
Linus Torvalds 已提交
3604 3605 3606
	sector_t new_sector;
	sector_t logical_sector, last_sector;
	struct stripe_head *sh;
3607
	const int rw = bio_data_dir(bi);
T
Tejun Heo 已提交
3608
	int cpu, remaining;
L
Linus Torvalds 已提交
3609

3610
	if (unlikely(bio_barrier(bi))) {
3611
		bio_endio(bi, -EOPNOTSUPP);
3612 3613 3614
		return 0;
	}

3615
	md_write_start(mddev, bi);
3616

T
Tejun Heo 已提交
3617 3618 3619 3620 3621
	cpu = part_stat_lock();
	part_stat_inc(cpu, &mddev->gendisk->part0, ios[rw]);
	part_stat_add(cpu, &mddev->gendisk->part0, sectors[rw],
		      bio_sectors(bi));
	part_stat_unlock();
L
Linus Torvalds 已提交
3622

3623
	if (rw == READ &&
3624 3625
	     mddev->reshape_position == MaxSector &&
	     chunk_aligned_read(q,bi))
3626
		return 0;
3627

L
Linus Torvalds 已提交
3628 3629 3630 3631
	logical_sector = bi->bi_sector & ~((sector_t)STRIPE_SECTORS-1);
	last_sector = bi->bi_sector + (bi->bi_size>>9);
	bi->bi_next = NULL;
	bi->bi_phys_segments = 1;	/* over-loaded to count active stripes */
3632

L
Linus Torvalds 已提交
3633 3634
	for (;logical_sector < last_sector; logical_sector += STRIPE_SECTORS) {
		DEFINE_WAIT(w);
3635
		int disks, data_disks;
3636
		int previous;
3637

3638
	retry:
3639
		previous = 0;
3640
		disks = conf->raid_disks;
3641
		prepare_to_wait(&conf->wait_for_overlap, &w, TASK_UNINTERRUPTIBLE);
3642
		if (unlikely(conf->reshape_progress != MaxSector)) {
3643
			/* spinlock is needed as reshape_progress may be
3644 3645
			 * 64bit on a 32bit platform, and so it might be
			 * possible to see a half-updated value
3646
			 * Ofcourse reshape_progress could change after
3647 3648 3649 3650
			 * the lock is dropped, so once we get a reference
			 * to the stripe that we think it is, we will have
			 * to check again.
			 */
3651
			spin_lock_irq(&conf->device_lock);
3652 3653 3654
			if (mddev->delta_disks < 0
			    ? logical_sector < conf->reshape_progress
			    : logical_sector >= conf->reshape_progress) {
3655
				disks = conf->previous_raid_disks;
3656 3657
				previous = 1;
			} else {
3658 3659 3660
				if (mddev->delta_disks < 0
				    ? logical_sector < conf->reshape_safe
				    : logical_sector >= conf->reshape_safe) {
3661 3662 3663 3664 3665
					spin_unlock_irq(&conf->device_lock);
					schedule();
					goto retry;
				}
			}
3666 3667
			spin_unlock_irq(&conf->device_lock);
		}
3668 3669
		data_disks = disks - conf->max_degraded;

3670 3671
		new_sector = raid5_compute_sector(conf, logical_sector,
						  previous,
3672
						  &dd_idx, NULL);
3673
		pr_debug("raid5: make_request, sector %llu logical %llu\n",
L
Linus Torvalds 已提交
3674 3675 3676
			(unsigned long long)new_sector, 
			(unsigned long long)logical_sector);

3677
		sh = get_active_stripe(conf, new_sector, previous,
3678
				       (bi->bi_rw&RWA_MASK), 0);
L
Linus Torvalds 已提交
3679
		if (sh) {
3680
			if (unlikely(previous)) {
3681
				/* expansion might have moved on while waiting for a
3682 3683 3684 3685 3686 3687
				 * 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.
3688 3689 3690
				 */
				int must_retry = 0;
				spin_lock_irq(&conf->device_lock);
3691 3692 3693
				if (mddev->delta_disks < 0
				    ? logical_sector >= conf->reshape_progress
				    : logical_sector < conf->reshape_progress)
3694 3695 3696 3697 3698 3699 3700 3701
					/* mismatch, need to try again */
					must_retry = 1;
				spin_unlock_irq(&conf->device_lock);
				if (must_retry) {
					release_stripe(sh);
					goto retry;
				}
			}
3702 3703 3704 3705 3706 3707 3708 3709 3710
			/* FIXME what if we get a false positive because these
			 * are being updated.
			 */
			if (logical_sector >= mddev->suspend_lo &&
			    logical_sector < mddev->suspend_hi) {
				release_stripe(sh);
				schedule();
				goto retry;
			}
3711 3712 3713 3714 3715

			if (test_bit(STRIPE_EXPANDING, &sh->state) ||
			    !add_stripe_bio(sh, bi, dd_idx, (bi->bi_rw&RW_MASK))) {
				/* Stripe is busy expanding or
				 * add failed due to overlap.  Flush everything
L
Linus Torvalds 已提交
3716 3717 3718 3719 3720 3721 3722 3723
				 * and wait a while
				 */
				raid5_unplug_device(mddev->queue);
				release_stripe(sh);
				schedule();
				goto retry;
			}
			finish_wait(&conf->wait_for_overlap, &w);
3724 3725
			set_bit(STRIPE_HANDLE, &sh->state);
			clear_bit(STRIPE_DELAYED, &sh->state);
L
Linus Torvalds 已提交
3726 3727 3728 3729 3730 3731 3732 3733 3734 3735
			release_stripe(sh);
		} else {
			/* cannot get stripe for read-ahead, just give-up */
			clear_bit(BIO_UPTODATE, &bi->bi_flags);
			finish_wait(&conf->wait_for_overlap, &w);
			break;
		}
			
	}
	spin_lock_irq(&conf->device_lock);
3736
	remaining = raid5_dec_bi_phys_segments(bi);
3737 3738
	spin_unlock_irq(&conf->device_lock);
	if (remaining == 0) {
L
Linus Torvalds 已提交
3739

3740
		if ( rw == WRITE )
L
Linus Torvalds 已提交
3741
			md_write_end(mddev);
3742

3743
		bio_endio(bi, 0);
L
Linus Torvalds 已提交
3744 3745 3746 3747
	}
	return 0;
}

D
Dan Williams 已提交
3748 3749
static sector_t raid5_size(mddev_t *mddev, sector_t sectors, int raid_disks);

3750
static sector_t reshape_request(mddev_t *mddev, sector_t sector_nr, int *skipped)
L
Linus Torvalds 已提交
3751
{
3752 3753 3754 3755 3756 3757 3758 3759 3760
	/* 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.
	 */
L
Linus Torvalds 已提交
3761 3762
	raid5_conf_t *conf = (raid5_conf_t *) mddev->private;
	struct stripe_head *sh;
3763
	sector_t first_sector, last_sector;
3764 3765 3766
	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;
3767 3768
	int i;
	int dd_idx;
3769
	sector_t writepos, readpos, safepos;
3770
	sector_t stripe_addr;
3771
	int reshape_sectors;
3772
	struct list_head stripes;
3773

3774 3775 3776 3777 3778 3779 3780 3781 3782
	if (sector_nr == 0) {
		/* If restarting in the middle, skip the initial sectors */
		if (mddev->delta_disks < 0 &&
		    conf->reshape_progress < raid5_size(mddev, 0, 0)) {
			sector_nr = raid5_size(mddev, 0, 0)
				- conf->reshape_progress;
		} else if (mddev->delta_disks > 0 &&
			   conf->reshape_progress > 0)
			sector_nr = conf->reshape_progress;
3783
		sector_div(sector_nr, new_data_disks);
3784 3785 3786 3787
		if (sector_nr) {
			*skipped = 1;
			return sector_nr;
		}
3788 3789
	}

3790 3791 3792 3793 3794 3795 3796 3797 3798
	/* 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
	 */
	if (mddev->new_chunk > mddev->chunk_size)
		reshape_sectors = mddev->new_chunk / 512;
	else
		reshape_sectors = mddev->chunk_size / 512;

3799 3800 3801 3802 3803
	/* we update the metadata when there is more than 3Meg
	 * in the block range (that is rather arbitrary, should
	 * probably be time based) or when the data about to be
	 * copied would over-write the source of the data at
	 * the front of the range.
3804 3805
	 * i.e. one new_stripe along from reshape_progress new_maps
	 * to after where reshape_safe old_maps to
3806
	 */
3807
	writepos = conf->reshape_progress;
3808
	sector_div(writepos, new_data_disks);
3809 3810
	readpos = conf->reshape_progress;
	sector_div(readpos, data_disks);
3811
	safepos = conf->reshape_safe;
3812
	sector_div(safepos, data_disks);
3813
	if (mddev->delta_disks < 0) {
3814
		writepos -= min_t(sector_t, reshape_sectors, writepos);
3815
		readpos += reshape_sectors;
3816
		safepos += reshape_sectors;
3817
	} else {
3818
		writepos += reshape_sectors;
3819 3820
		readpos -= min_t(sector_t, reshape_sectors, readpos);
		safepos -= min_t(sector_t, reshape_sectors, safepos);
3821
	}
3822

3823 3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839
	/* '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.
	 * If 'readpos' is behind 'writepos', then there is no way that we can
	 * 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???
	 */
3840
	if ((mddev->delta_disks < 0
3841 3842 3843
	     ? (safepos > writepos && readpos < writepos)
	     : (safepos < writepos && readpos > writepos)) ||
	    time_after(jiffies, conf->reshape_checkpoint + 10*HZ)) {
3844 3845 3846
		/* Cannot proceed until we've updated the superblock... */
		wait_event(conf->wait_for_overlap,
			   atomic_read(&conf->reshape_stripes)==0);
3847
		mddev->reshape_position = conf->reshape_progress;
3848
		mddev->curr_resync_completed = mddev->curr_resync;
3849
		conf->reshape_checkpoint = jiffies;
3850
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
3851
		md_wakeup_thread(mddev->thread);
3852
		wait_event(mddev->sb_wait, mddev->flags == 0 ||
3853 3854
			   kthread_should_stop());
		spin_lock_irq(&conf->device_lock);
3855
		conf->reshape_safe = mddev->reshape_position;
3856 3857
		spin_unlock_irq(&conf->device_lock);
		wake_up(&conf->wait_for_overlap);
3858
		sysfs_notify(&mddev->kobj, NULL, "sync_completed");
3859 3860
	}

3861 3862 3863 3864
	if (mddev->delta_disks < 0) {
		BUG_ON(conf->reshape_progress == 0);
		stripe_addr = writepos;
		BUG_ON((mddev->dev_sectors &
3865 3866
			~((sector_t)reshape_sectors - 1))
		       - reshape_sectors - stripe_addr
3867 3868
		       != sector_nr);
	} else {
3869
		BUG_ON(writepos != sector_nr + reshape_sectors);
3870 3871
		stripe_addr = sector_nr;
	}
3872
	INIT_LIST_HEAD(&stripes);
3873
	for (i = 0; i < reshape_sectors; i += STRIPE_SECTORS) {
3874 3875
		int j;
		int skipped = 0;
3876
		sh = get_active_stripe(conf, stripe_addr+i, 0, 0, 1);
3877 3878 3879 3880 3881 3882 3883 3884 3885
		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;
3886
			if (conf->level == 6 &&
3887
			    j == sh->qd_idx)
3888
				continue;
3889
			s = compute_blocknr(sh, j, 0);
D
Dan Williams 已提交
3890
			if (s < raid5_size(mddev, 0, 0)) {
3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901
				skipped = 1;
				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);
		}
		if (!skipped) {
			set_bit(STRIPE_EXPAND_READY, &sh->state);
			set_bit(STRIPE_HANDLE, &sh->state);
		}
3902
		list_add(&sh->lru, &stripes);
3903 3904
	}
	spin_lock_irq(&conf->device_lock);
3905
	if (mddev->delta_disks < 0)
3906
		conf->reshape_progress -= reshape_sectors * new_data_disks;
3907
	else
3908
		conf->reshape_progress += reshape_sectors * new_data_disks;
3909 3910 3911 3912 3913 3914 3915
	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 =
3916
		raid5_compute_sector(conf, stripe_addr*(new_data_disks),
3917
				     1, &dd_idx, NULL);
3918
	last_sector =
3919
		raid5_compute_sector(conf, ((stripe_addr+reshape_sectors)
3920
					    *(new_data_disks) - 1),
3921
				     1, &dd_idx, NULL);
A
Andre Noll 已提交
3922 3923
	if (last_sector >= mddev->dev_sectors)
		last_sector = mddev->dev_sectors - 1;
3924
	while (first_sector <= last_sector) {
3925
		sh = get_active_stripe(conf, first_sector, 1, 0, 1);
3926 3927 3928 3929 3930
		set_bit(STRIPE_EXPAND_SOURCE, &sh->state);
		set_bit(STRIPE_HANDLE, &sh->state);
		release_stripe(sh);
		first_sector += STRIPE_SECTORS;
	}
3931 3932 3933 3934 3935 3936 3937 3938
	/* 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);
	}
3939 3940 3941
	/* If this takes us to the resync_max point where we have to pause,
	 * then we need to write out the superblock.
	 */
3942
	sector_nr += reshape_sectors;
3943 3944
	if ((sector_nr - mddev->curr_resync_completed) * 2
	    >= mddev->resync_max - mddev->curr_resync_completed) {
3945 3946 3947
		/* Cannot proceed until we've updated the superblock... */
		wait_event(conf->wait_for_overlap,
			   atomic_read(&conf->reshape_stripes) == 0);
3948
		mddev->reshape_position = conf->reshape_progress;
3949
		mddev->curr_resync_completed = mddev->curr_resync;
3950
		conf->reshape_checkpoint = jiffies;
3951 3952 3953 3954 3955 3956
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
		md_wakeup_thread(mddev->thread);
		wait_event(mddev->sb_wait,
			   !test_bit(MD_CHANGE_DEVS, &mddev->flags)
			   || kthread_should_stop());
		spin_lock_irq(&conf->device_lock);
3957
		conf->reshape_safe = mddev->reshape_position;
3958 3959
		spin_unlock_irq(&conf->device_lock);
		wake_up(&conf->wait_for_overlap);
3960
		sysfs_notify(&mddev->kobj, NULL, "sync_completed");
3961
	}
3962
	return reshape_sectors;
3963 3964 3965 3966 3967 3968 3969
}

/* FIXME go_faster isn't used */
static inline sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster)
{
	raid5_conf_t *conf = (raid5_conf_t *) mddev->private;
	struct stripe_head *sh;
A
Andre Noll 已提交
3970
	sector_t max_sector = mddev->dev_sectors;
3971
	int sync_blocks;
3972 3973
	int still_degraded = 0;
	int i;
L
Linus Torvalds 已提交
3974

3975
	if (sector_nr >= max_sector) {
L
Linus Torvalds 已提交
3976 3977
		/* just being told to finish up .. nothing much to do */
		unplug_slaves(mddev);
3978

3979 3980 3981 3982
		if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) {
			end_reshape(conf);
			return 0;
		}
3983 3984 3985 3986

		if (mddev->curr_resync < max_sector) /* aborted */
			bitmap_end_sync(mddev->bitmap, mddev->curr_resync,
					&sync_blocks, 1);
3987
		else /* completed sync */
3988 3989 3990
			conf->fullsync = 0;
		bitmap_close_sync(mddev->bitmap);

L
Linus Torvalds 已提交
3991 3992
		return 0;
	}
3993

3994 3995
	if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
		return reshape_request(mddev, sector_nr, skipped);
3996

3997 3998 3999 4000 4001 4002
	/* 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
	 */

4003
	/* if there is too many failed drives and we are trying
L
Linus Torvalds 已提交
4004 4005 4006
	 * to resync, then assert that we are finished, because there is
	 * nothing we can do.
	 */
4007
	if (mddev->degraded >= conf->max_degraded &&
4008
	    test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
A
Andre Noll 已提交
4009
		sector_t rv = mddev->dev_sectors - sector_nr;
4010
		*skipped = 1;
L
Linus Torvalds 已提交
4011 4012
		return rv;
	}
4013
	if (!bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, 1) &&
4014
	    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
4015 4016 4017 4018 4019 4020
	    !conf->fullsync && sync_blocks >= STRIPE_SECTORS) {
		/* we can skip this block, and probably more */
		sync_blocks /= STRIPE_SECTORS;
		*skipped = 1;
		return sync_blocks * STRIPE_SECTORS; /* keep things rounded to whole stripes */
	}
L
Linus Torvalds 已提交
4021

N
NeilBrown 已提交
4022 4023 4024

	bitmap_cond_end_sync(mddev->bitmap, sector_nr);

4025
	sh = get_active_stripe(conf, sector_nr, 0, 1, 0);
L
Linus Torvalds 已提交
4026
	if (sh == NULL) {
4027
		sh = get_active_stripe(conf, sector_nr, 0, 0, 0);
L
Linus Torvalds 已提交
4028
		/* make sure we don't swamp the stripe cache if someone else
4029
		 * is trying to get access
L
Linus Torvalds 已提交
4030
		 */
4031
		schedule_timeout_uninterruptible(1);
L
Linus Torvalds 已提交
4032
	}
4033 4034 4035 4036
	/* Need to check if array will still be degraded after recovery/resync
	 * We don't need to check the 'failed' flag as when that gets set,
	 * recovery aborts.
	 */
4037
	for (i = 0; i < conf->raid_disks; i++)
4038 4039 4040 4041 4042 4043
		if (conf->disks[i].rdev == NULL)
			still_degraded = 1;

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

	spin_lock(&sh->lock);
L
Linus Torvalds 已提交
4044 4045 4046 4047
	set_bit(STRIPE_SYNCING, &sh->state);
	clear_bit(STRIPE_INSYNC, &sh->state);
	spin_unlock(&sh->lock);

4048 4049 4050
	/* wait for any blocked device to be handled */
	while(unlikely(!handle_stripe(sh, NULL)))
		;
L
Linus Torvalds 已提交
4051 4052 4053 4054 4055
	release_stripe(sh);

	return STRIPE_SECTORS;
}

4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068
static int  retry_aligned_read(raid5_conf_t *conf, struct bio *raid_bio)
{
	/* 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;
4069
	int dd_idx;
4070 4071 4072 4073 4074 4075
	sector_t sector, logical_sector, last_sector;
	int scnt = 0;
	int remaining;
	int handled = 0;

	logical_sector = raid_bio->bi_sector & ~((sector_t)STRIPE_SECTORS-1);
4076
	sector = raid5_compute_sector(conf, logical_sector,
4077
				      0, &dd_idx, NULL);
4078 4079 4080
	last_sector = raid_bio->bi_sector + (raid_bio->bi_size>>9);

	for (; logical_sector < last_sector;
4081 4082 4083
	     logical_sector += STRIPE_SECTORS,
		     sector += STRIPE_SECTORS,
		     scnt++) {
4084

4085
		if (scnt < raid5_bi_hw_segments(raid_bio))
4086 4087 4088
			/* already done this stripe */
			continue;

4089
		sh = get_active_stripe(conf, sector, 0, 1, 0);
4090 4091 4092

		if (!sh) {
			/* failed to get a stripe - must wait */
4093
			raid5_set_bi_hw_segments(raid_bio, scnt);
4094 4095 4096 4097 4098
			conf->retry_read_aligned = raid_bio;
			return handled;
		}

		set_bit(R5_ReadError, &sh->dev[dd_idx].flags);
4099 4100
		if (!add_stripe_bio(sh, raid_bio, dd_idx, 0)) {
			release_stripe(sh);
4101
			raid5_set_bi_hw_segments(raid_bio, scnt);
4102 4103 4104 4105
			conf->retry_read_aligned = raid_bio;
			return handled;
		}

4106 4107 4108 4109 4110
		handle_stripe(sh, NULL);
		release_stripe(sh);
		handled++;
	}
	spin_lock_irq(&conf->device_lock);
4111
	remaining = raid5_dec_bi_phys_segments(raid_bio);
4112
	spin_unlock_irq(&conf->device_lock);
4113 4114
	if (remaining == 0)
		bio_endio(raid_bio, 0);
4115 4116 4117 4118 4119 4120 4121
	if (atomic_dec_and_test(&conf->active_aligned_reads))
		wake_up(&conf->wait_for_stripe);
	return handled;
}



L
Linus Torvalds 已提交
4122 4123 4124 4125 4126 4127 4128
/*
 * 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.
 */
4129
static void raid5d(mddev_t *mddev)
L
Linus Torvalds 已提交
4130 4131
{
	struct stripe_head *sh;
4132
	raid5_conf_t *conf = mddev->private;
L
Linus Torvalds 已提交
4133 4134
	int handled;

4135
	pr_debug("+++ raid5d active\n");
L
Linus Torvalds 已提交
4136 4137 4138 4139 4140 4141

	md_check_recovery(mddev);

	handled = 0;
	spin_lock_irq(&conf->device_lock);
	while (1) {
4142
		struct bio *bio;
L
Linus Torvalds 已提交
4143

4144
		if (conf->seq_flush != conf->seq_write) {
4145
			int seq = conf->seq_flush;
4146
			spin_unlock_irq(&conf->device_lock);
4147
			bitmap_unplug(mddev->bitmap);
4148
			spin_lock_irq(&conf->device_lock);
4149 4150 4151 4152
			conf->seq_write = seq;
			activate_bit_delay(conf);
		}

4153 4154 4155 4156 4157 4158 4159 4160 4161 4162
		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++;
		}

4163 4164
		sh = __get_priority_stripe(conf);

4165
		if (!sh)
L
Linus Torvalds 已提交
4166 4167 4168 4169
			break;
		spin_unlock_irq(&conf->device_lock);
		
		handled++;
4170
		handle_stripe(sh, conf->spare_page);
L
Linus Torvalds 已提交
4171 4172 4173 4174
		release_stripe(sh);

		spin_lock_irq(&conf->device_lock);
	}
4175
	pr_debug("%d stripes handled\n", handled);
L
Linus Torvalds 已提交
4176 4177 4178

	spin_unlock_irq(&conf->device_lock);

4179
	async_tx_issue_pending_all();
L
Linus Torvalds 已提交
4180 4181
	unplug_slaves(mddev);

4182
	pr_debug("--- raid5d inactive\n");
L
Linus Torvalds 已提交
4183 4184
}

4185
static ssize_t
4186
raid5_show_stripe_cache_size(mddev_t *mddev, char *page)
4187
{
4188
	raid5_conf_t *conf = mddev->private;
4189 4190 4191 4192
	if (conf)
		return sprintf(page, "%d\n", conf->max_nr_stripes);
	else
		return 0;
4193 4194 4195
}

static ssize_t
4196
raid5_store_stripe_cache_size(mddev_t *mddev, const char *page, size_t len)
4197
{
4198
	raid5_conf_t *conf = mddev->private;
4199
	unsigned long new;
4200 4201
	int err;

4202 4203
	if (len >= PAGE_SIZE)
		return -EINVAL;
4204 4205
	if (!conf)
		return -ENODEV;
4206

4207
	if (strict_strtoul(page, 10, &new))
4208 4209 4210 4211 4212 4213 4214 4215 4216
		return -EINVAL;
	if (new <= 16 || new > 32768)
		return -EINVAL;
	while (new < conf->max_nr_stripes) {
		if (drop_one_stripe(conf))
			conf->max_nr_stripes--;
		else
			break;
	}
4217 4218 4219
	err = md_allow_write(mddev);
	if (err)
		return err;
4220 4221 4222 4223 4224 4225 4226
	while (new > conf->max_nr_stripes) {
		if (grow_one_stripe(conf))
			conf->max_nr_stripes++;
		else break;
	}
	return len;
}
4227

4228 4229 4230 4231
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);
4232

4233 4234 4235
static ssize_t
raid5_show_preread_threshold(mddev_t *mddev, char *page)
{
4236
	raid5_conf_t *conf = mddev->private;
4237 4238 4239 4240 4241 4242 4243 4244 4245
	if (conf)
		return sprintf(page, "%d\n", conf->bypass_threshold);
	else
		return 0;
}

static ssize_t
raid5_store_preread_threshold(mddev_t *mddev, const char *page, size_t len)
{
4246
	raid5_conf_t *conf = mddev->private;
4247
	unsigned long new;
4248 4249 4250 4251 4252
	if (len >= PAGE_SIZE)
		return -EINVAL;
	if (!conf)
		return -ENODEV;

4253
	if (strict_strtoul(page, 10, &new))
4254
		return -EINVAL;
4255
	if (new > conf->max_nr_stripes)
4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266
		return -EINVAL;
	conf->bypass_threshold = new;
	return len;
}

static struct md_sysfs_entry
raid5_preread_bypass_threshold = __ATTR(preread_bypass_threshold,
					S_IRUGO | S_IWUSR,
					raid5_show_preread_threshold,
					raid5_store_preread_threshold);

4267
static ssize_t
4268
stripe_cache_active_show(mddev_t *mddev, char *page)
4269
{
4270
	raid5_conf_t *conf = mddev->private;
4271 4272 4273 4274
	if (conf)
		return sprintf(page, "%d\n", atomic_read(&conf->active_stripes));
	else
		return 0;
4275 4276
}

4277 4278
static struct md_sysfs_entry
raid5_stripecache_active = __ATTR_RO(stripe_cache_active);
4279

4280
static struct attribute *raid5_attrs[] =  {
4281 4282
	&raid5_stripecache_size.attr,
	&raid5_stripecache_active.attr,
4283
	&raid5_preread_bypass_threshold.attr,
4284 4285
	NULL,
};
4286 4287 4288
static struct attribute_group raid5_attrs_group = {
	.name = NULL,
	.attrs = raid5_attrs,
4289 4290
};

4291 4292 4293
static sector_t
raid5_size(mddev_t *mddev, sector_t sectors, int raid_disks)
{
4294
	raid5_conf_t *conf = mddev->private;
4295 4296 4297

	if (!sectors)
		sectors = mddev->dev_sectors;
4298 4299 4300 4301 4302 4303 4304
	if (!raid_disks) {
		/* size is defined by the smallest of previous and new size */
		if (conf->raid_disks < conf->previous_raid_disks)
			raid_disks = conf->raid_disks;
		else
			raid_disks = conf->previous_raid_disks;
	}
4305 4306

	sectors &= ~((sector_t)mddev->chunk_size/512 - 1);
4307
	sectors &= ~((sector_t)mddev->new_chunk/512 - 1);
4308 4309 4310
	return sectors * (raid_disks - conf->max_degraded);
}

N
NeilBrown 已提交
4311
static raid5_conf_t *setup_conf(mddev_t *mddev)
L
Linus Torvalds 已提交
4312 4313 4314 4315 4316 4317
{
	raid5_conf_t *conf;
	int raid_disk, memory;
	mdk_rdev_t *rdev;
	struct disk_info *disk;

N
NeilBrown 已提交
4318 4319 4320
	if (mddev->new_level != 5
	    && mddev->new_level != 4
	    && mddev->new_level != 6) {
4321
		printk(KERN_ERR "raid5: %s: raid level not set to 4/5/6 (%d)\n",
N
NeilBrown 已提交
4322 4323
		       mdname(mddev), mddev->new_level);
		return ERR_PTR(-EIO);
L
Linus Torvalds 已提交
4324
	}
N
NeilBrown 已提交
4325 4326 4327 4328
	if ((mddev->new_level == 5
	     && !algorithm_valid_raid5(mddev->new_layout)) ||
	    (mddev->new_level == 6
	     && !algorithm_valid_raid6(mddev->new_layout))) {
4329
		printk(KERN_ERR "raid5: %s: layout %d not supported\n",
N
NeilBrown 已提交
4330 4331
		       mdname(mddev), mddev->new_layout);
		return ERR_PTR(-EIO);
4332
	}
N
NeilBrown 已提交
4333 4334 4335 4336
	if (mddev->new_level == 6 && mddev->raid_disks < 4) {
		printk(KERN_ERR "raid6: not enough configured devices for %s (%d, minimum 4)\n",
		       mdname(mddev), mddev->raid_disks);
		return ERR_PTR(-EINVAL);
4337 4338
	}

4339 4340
	if (!mddev->new_chunk || mddev->new_chunk % PAGE_SIZE ||
	    !is_power_of_2(mddev->new_chunk)) {
N
NeilBrown 已提交
4341 4342 4343
		printk(KERN_ERR "raid5: invalid chunk size %d for %s\n",
			mddev->new_chunk, mdname(mddev));
		return ERR_PTR(-EINVAL);
4344 4345
	}

N
NeilBrown 已提交
4346 4347
	conf = kzalloc(sizeof(raid5_conf_t), GFP_KERNEL);
	if (conf == NULL)
L
Linus Torvalds 已提交
4348
		goto abort;
N
NeilBrown 已提交
4349 4350 4351 4352 4353

	conf->raid_disks = mddev->raid_disks;
	if (mddev->reshape_position == MaxSector)
		conf->previous_raid_disks = mddev->raid_disks;
	else
4354 4355 4356
		conf->previous_raid_disks = mddev->raid_disks - mddev->delta_disks;

	conf->disks = kzalloc(conf->raid_disks * sizeof(struct disk_info),
4357 4358 4359
			      GFP_KERNEL);
	if (!conf->disks)
		goto abort;
4360

L
Linus Torvalds 已提交
4361 4362
	conf->mddev = mddev;

4363
	if ((conf->stripe_hashtbl = kzalloc(PAGE_SIZE, GFP_KERNEL)) == NULL)
L
Linus Torvalds 已提交
4364 4365
		goto abort;

N
NeilBrown 已提交
4366
	if (mddev->new_level == 6) {
4367 4368 4369 4370
		conf->spare_page = alloc_page(GFP_KERNEL);
		if (!conf->spare_page)
			goto abort;
	}
L
Linus Torvalds 已提交
4371 4372 4373 4374
	spin_lock_init(&conf->device_lock);
	init_waitqueue_head(&conf->wait_for_stripe);
	init_waitqueue_head(&conf->wait_for_overlap);
	INIT_LIST_HEAD(&conf->handle_list);
4375
	INIT_LIST_HEAD(&conf->hold_list);
L
Linus Torvalds 已提交
4376
	INIT_LIST_HEAD(&conf->delayed_list);
4377
	INIT_LIST_HEAD(&conf->bitmap_list);
L
Linus Torvalds 已提交
4378 4379 4380
	INIT_LIST_HEAD(&conf->inactive_list);
	atomic_set(&conf->active_stripes, 0);
	atomic_set(&conf->preread_active_stripes, 0);
4381
	atomic_set(&conf->active_aligned_reads, 0);
4382
	conf->bypass_threshold = BYPASS_THRESHOLD;
L
Linus Torvalds 已提交
4383

4384
	pr_debug("raid5: run(%s) called.\n", mdname(mddev));
L
Linus Torvalds 已提交
4385

4386
	list_for_each_entry(rdev, &mddev->disks, same_set) {
L
Linus Torvalds 已提交
4387
		raid_disk = rdev->raid_disk;
4388
		if (raid_disk >= conf->raid_disks
L
Linus Torvalds 已提交
4389 4390 4391 4392 4393 4394
		    || raid_disk < 0)
			continue;
		disk = conf->disks + raid_disk;

		disk->rdev = rdev;

4395
		if (test_bit(In_sync, &rdev->flags)) {
L
Linus Torvalds 已提交
4396 4397 4398 4399
			char b[BDEVNAME_SIZE];
			printk(KERN_INFO "raid5: device %s operational as raid"
				" disk %d\n", bdevname(rdev->bdev,b),
				raid_disk);
4400 4401 4402
		} else
			/* Cannot rely on bitmap to complete recovery */
			conf->fullsync = 1;
L
Linus Torvalds 已提交
4403 4404
	}

N
NeilBrown 已提交
4405 4406
	conf->chunk_size = mddev->new_chunk;
	conf->level = mddev->new_level;
4407 4408 4409 4410
	if (conf->level == 6)
		conf->max_degraded = 2;
	else
		conf->max_degraded = 1;
N
NeilBrown 已提交
4411
	conf->algorithm = mddev->new_layout;
L
Linus Torvalds 已提交
4412
	conf->max_nr_stripes = NR_STRIPES;
4413
	conf->reshape_progress = mddev->reshape_position;
4414
	if (conf->reshape_progress != MaxSector) {
4415
		conf->prev_chunk = mddev->chunk_size;
4416 4417
		conf->prev_algo = mddev->layout;
	}
L
Linus Torvalds 已提交
4418

N
NeilBrown 已提交
4419 4420 4421 4422 4423 4424 4425 4426 4427
	memory = conf->max_nr_stripes * (sizeof(struct stripe_head) +
		 conf->raid_disks * ((sizeof(struct bio) + PAGE_SIZE))) / 1024;
	if (grow_stripes(conf, conf->max_nr_stripes)) {
		printk(KERN_ERR
			"raid5: couldn't allocate %dkB for buffers\n", memory);
		goto abort;
	} else
		printk(KERN_INFO "raid5: allocated %dkB for %s\n",
			memory, mdname(mddev));
L
Linus Torvalds 已提交
4428

N
NeilBrown 已提交
4429 4430 4431 4432 4433
	conf->thread = md_register_thread(raid5d, mddev, "%s_raid5");
	if (!conf->thread) {
		printk(KERN_ERR
		       "raid5: couldn't allocate thread for %s\n",
		       mdname(mddev));
4434 4435
		goto abort;
	}
N
NeilBrown 已提交
4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464

	return conf;

 abort:
	if (conf) {
		shrink_stripes(conf);
		safe_put_page(conf->spare_page);
		kfree(conf->disks);
		kfree(conf->stripe_hashtbl);
		kfree(conf);
		return ERR_PTR(-EIO);
	} else
		return ERR_PTR(-ENOMEM);
}

static int run(mddev_t *mddev)
{
	raid5_conf_t *conf;
	int working_disks = 0;
	mdk_rdev_t *rdev;

	if (mddev->reshape_position != MaxSector) {
		/* Check that we can continue the reshape.
		 * Currently only disks can change, it must
		 * increase, and we must be past the point where
		 * a stripe over-writes itself
		 */
		sector_t here_new, here_old;
		int old_disks;
4465
		int max_degraded = (mddev->level == 6 ? 2 : 1);
N
NeilBrown 已提交
4466

4467
		if (mddev->new_level != mddev->level) {
N
NeilBrown 已提交
4468 4469 4470 4471 4472 4473 4474 4475 4476 4477 4478
			printk(KERN_ERR "raid5: %s: unsupported reshape "
			       "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;
4479
		if (sector_div(here_new, (mddev->new_chunk>>9)*
N
NeilBrown 已提交
4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501 4502 4503
			       (mddev->raid_disks - max_degraded))) {
			printk(KERN_ERR "raid5: reshape_position not "
			       "on a stripe boundary\n");
			return -EINVAL;
		}
		/* here_new is the stripe we will write to */
		here_old = mddev->reshape_position;
		sector_div(here_old, (mddev->chunk_size>>9)*
			   (old_disks-max_degraded));
		/* here_old is the first stripe that we might need to read
		 * from */
		if (here_new >= here_old) {
			/* Reading from the same stripe as writing to - bad */
			printk(KERN_ERR "raid5: reshape_position too early for "
			       "auto-recovery - aborting.\n");
			return -EINVAL;
		}
		printk(KERN_INFO "raid5: reshape will continue\n");
		/* OK, we should be able to continue; */
	} else {
		BUG_ON(mddev->level != mddev->new_level);
		BUG_ON(mddev->layout != mddev->new_layout);
		BUG_ON(mddev->chunk_size != mddev->new_chunk);
		BUG_ON(mddev->delta_disks != 0);
L
Linus Torvalds 已提交
4504
	}
N
NeilBrown 已提交
4505

4506 4507 4508 4509 4510
	if (mddev->private == NULL)
		conf = setup_conf(mddev);
	else
		conf = mddev->private;

N
NeilBrown 已提交
4511 4512 4513 4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527
	if (IS_ERR(conf))
		return PTR_ERR(conf);

	mddev->thread = conf->thread;
	conf->thread = NULL;
	mddev->private = conf;

	/*
	 * 0 for a fully functional array, 1 or 2 for a degraded array.
	 */
	list_for_each_entry(rdev, &mddev->disks, same_set)
		if (rdev->raid_disk >= 0 &&
		    test_bit(In_sync, &rdev->flags))
			working_disks++;

	mddev->degraded = conf->raid_disks - working_disks;

4528
	if (mddev->degraded > conf->max_degraded) {
L
Linus Torvalds 已提交
4529 4530
		printk(KERN_ERR "raid5: not enough operational devices for %s"
			" (%d/%d failed)\n",
4531
			mdname(mddev), mddev->degraded, conf->raid_disks);
L
Linus Torvalds 已提交
4532 4533 4534
		goto abort;
	}

N
NeilBrown 已提交
4535 4536 4537 4538
	/* device size must be a multiple of chunk size */
	mddev->dev_sectors &= ~(mddev->chunk_size / 512 - 1);
	mddev->resync_max_sectors = mddev->dev_sectors;

4539
	if (mddev->degraded > 0 &&
L
Linus Torvalds 已提交
4540
	    mddev->recovery_cp != MaxSector) {
4541 4542 4543 4544 4545 4546 4547 4548 4549 4550 4551
		if (mddev->ok_start_degraded)
			printk(KERN_WARNING
			       "raid5: starting dirty degraded array: %s"
			       "- data corruption possible.\n",
			       mdname(mddev));
		else {
			printk(KERN_ERR
			       "raid5: cannot start dirty degraded array for %s\n",
			       mdname(mddev));
			goto abort;
		}
L
Linus Torvalds 已提交
4552 4553 4554 4555
	}

	if (mddev->degraded == 0)
		printk("raid5: raid level %d set %s active with %d out of %d"
4556 4557 4558
		       " devices, algorithm %d\n", conf->level, mdname(mddev),
		       mddev->raid_disks-mddev->degraded, mddev->raid_disks,
		       mddev->new_layout);
L
Linus Torvalds 已提交
4559 4560 4561 4562
	else
		printk(KERN_ALERT "raid5: raid level %d set %s active with %d"
			" out of %d devices, algorithm %d\n", conf->level,
			mdname(mddev), mddev->raid_disks - mddev->degraded,
4563
			mddev->raid_disks, mddev->new_layout);
L
Linus Torvalds 已提交
4564 4565 4566

	print_raid5_conf(conf);

4567
	if (conf->reshape_progress != MaxSector) {
4568
		printk("...ok start reshape thread\n");
4569
		conf->reshape_safe = conf->reshape_progress;
4570 4571 4572 4573 4574 4575 4576 4577 4578
		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,
							"%s_reshape");
	}

L
Linus Torvalds 已提交
4579
	/* read-ahead size must cover two whole stripes, which is
4580
	 * 2 * (datadisks) * chunksize where 'n' is the number of raid devices
L
Linus Torvalds 已提交
4581 4582
	 */
	{
4583 4584
		int data_disks = conf->previous_raid_disks - conf->max_degraded;
		int stripe = data_disks *
4585
			(mddev->chunk_size / PAGE_SIZE);
L
Linus Torvalds 已提交
4586 4587 4588 4589 4590
		if (mddev->queue->backing_dev_info.ra_pages < 2 * stripe)
			mddev->queue->backing_dev_info.ra_pages = 2 * stripe;
	}

	/* Ok, everything is just fine now */
4591 4592 4593 4594
	if (sysfs_create_group(&mddev->kobj, &raid5_attrs_group))
		printk(KERN_WARNING
		       "raid5: failed to create sysfs attributes for %s\n",
		       mdname(mddev));
4595

N
NeilBrown 已提交
4596 4597
	mddev->queue->queue_lock = &conf->device_lock;

4598
	mddev->queue->unplug_fn = raid5_unplug_device;
4599
	mddev->queue->backing_dev_info.congested_data = mddev;
4600
	mddev->queue->backing_dev_info.congested_fn = raid5_congested;
4601

4602
	md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
4603

4604 4605
	blk_queue_merge_bvec(mddev->queue, raid5_mergeable_bvec);

L
Linus Torvalds 已提交
4606 4607
	return 0;
abort:
4608
	md_unregister_thread(mddev->thread);
N
NeilBrown 已提交
4609
	mddev->thread = NULL;
L
Linus Torvalds 已提交
4610
	if (conf) {
N
NeilBrown 已提交
4611
		shrink_stripes(conf);
L
Linus Torvalds 已提交
4612
		print_raid5_conf(conf);
4613
		safe_put_page(conf->spare_page);
4614
		kfree(conf->disks);
4615
		kfree(conf->stripe_hashtbl);
L
Linus Torvalds 已提交
4616 4617 4618 4619 4620 4621 4622 4623 4624
		kfree(conf);
	}
	mddev->private = NULL;
	printk(KERN_ALERT "raid5: failed to run raid set %s\n", mdname(mddev));
	return -EIO;
}



4625
static int stop(mddev_t *mddev)
L
Linus Torvalds 已提交
4626 4627 4628 4629 4630 4631
{
	raid5_conf_t *conf = (raid5_conf_t *) mddev->private;

	md_unregister_thread(mddev->thread);
	mddev->thread = NULL;
	shrink_stripes(conf);
4632
	kfree(conf->stripe_hashtbl);
4633
	mddev->queue->backing_dev_info.congested_fn = NULL;
L
Linus Torvalds 已提交
4634
	blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
4635
	sysfs_remove_group(&mddev->kobj, &raid5_attrs_group);
4636
	kfree(conf->disks);
4637
	kfree(conf);
L
Linus Torvalds 已提交
4638 4639 4640 4641
	mddev->private = NULL;
	return 0;
}

4642
#ifdef DEBUG
4643
static void print_sh(struct seq_file *seq, struct stripe_head *sh)
L
Linus Torvalds 已提交
4644 4645 4646
{
	int i;

4647 4648 4649 4650 4651
	seq_printf(seq, "sh %llu, pd_idx %d, state %ld.\n",
		   (unsigned long long)sh->sector, sh->pd_idx, sh->state);
	seq_printf(seq, "sh %llu,  count %d.\n",
		   (unsigned long long)sh->sector, atomic_read(&sh->count));
	seq_printf(seq, "sh %llu, ", (unsigned long long)sh->sector);
4652
	for (i = 0; i < sh->disks; i++) {
4653 4654
		seq_printf(seq, "(cache%d: %p %ld) ",
			   i, sh->dev[i].page, sh->dev[i].flags);
L
Linus Torvalds 已提交
4655
	}
4656
	seq_printf(seq, "\n");
L
Linus Torvalds 已提交
4657 4658
}

4659
static void printall(struct seq_file *seq, raid5_conf_t *conf)
L
Linus Torvalds 已提交
4660 4661
{
	struct stripe_head *sh;
4662
	struct hlist_node *hn;
L
Linus Torvalds 已提交
4663 4664 4665 4666
	int i;

	spin_lock_irq(&conf->device_lock);
	for (i = 0; i < NR_HASH; i++) {
4667
		hlist_for_each_entry(sh, hn, &conf->stripe_hashtbl[i], hash) {
L
Linus Torvalds 已提交
4668 4669
			if (sh->raid_conf != conf)
				continue;
4670
			print_sh(seq, sh);
L
Linus Torvalds 已提交
4671 4672 4673 4674 4675 4676
		}
	}
	spin_unlock_irq(&conf->device_lock);
}
#endif

4677
static void status(struct seq_file *seq, mddev_t *mddev)
L
Linus Torvalds 已提交
4678 4679 4680 4681 4682
{
	raid5_conf_t *conf = (raid5_conf_t *) mddev->private;
	int i;

	seq_printf (seq, " level %d, %dk chunk, algorithm %d", mddev->level, mddev->chunk_size >> 10, mddev->layout);
4683
	seq_printf (seq, " [%d/%d] [", conf->raid_disks, conf->raid_disks - mddev->degraded);
L
Linus Torvalds 已提交
4684 4685 4686
	for (i = 0; i < conf->raid_disks; i++)
		seq_printf (seq, "%s",
			       conf->disks[i].rdev &&
4687
			       test_bit(In_sync, &conf->disks[i].rdev->flags) ? "U" : "_");
L
Linus Torvalds 已提交
4688
	seq_printf (seq, "]");
4689
#ifdef DEBUG
4690 4691
	seq_printf (seq, "\n");
	printall(seq, conf);
L
Linus Torvalds 已提交
4692 4693 4694 4695 4696 4697 4698 4699 4700 4701 4702 4703 4704
#endif
}

static void print_raid5_conf (raid5_conf_t *conf)
{
	int i;
	struct disk_info *tmp;

	printk("RAID5 conf printout:\n");
	if (!conf) {
		printk("(conf==NULL)\n");
		return;
	}
4705 4706
	printk(" --- rd:%d wd:%d\n", conf->raid_disks,
		 conf->raid_disks - conf->mddev->degraded);
L
Linus Torvalds 已提交
4707 4708 4709 4710 4711 4712

	for (i = 0; i < conf->raid_disks; i++) {
		char b[BDEVNAME_SIZE];
		tmp = conf->disks + i;
		if (tmp->rdev)
		printk(" disk %d, o:%d, dev:%s\n",
4713
			i, !test_bit(Faulty, &tmp->rdev->flags),
L
Linus Torvalds 已提交
4714 4715 4716 4717 4718 4719 4720 4721 4722 4723 4724 4725 4726
			bdevname(tmp->rdev->bdev,b));
	}
}

static int raid5_spare_active(mddev_t *mddev)
{
	int i;
	raid5_conf_t *conf = mddev->private;
	struct disk_info *tmp;

	for (i = 0; i < conf->raid_disks; i++) {
		tmp = conf->disks + i;
		if (tmp->rdev
4727
		    && !test_bit(Faulty, &tmp->rdev->flags)
4728 4729 4730
		    && !test_and_set_bit(In_sync, &tmp->rdev->flags)) {
			unsigned long flags;
			spin_lock_irqsave(&conf->device_lock, flags);
L
Linus Torvalds 已提交
4731
			mddev->degraded--;
4732
			spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743 4744 4745 4746 4747 4748
		}
	}
	print_raid5_conf(conf);
	return 0;
}

static int raid5_remove_disk(mddev_t *mddev, int number)
{
	raid5_conf_t *conf = mddev->private;
	int err = 0;
	mdk_rdev_t *rdev;
	struct disk_info *p = conf->disks + number;

	print_raid5_conf(conf);
	rdev = p->rdev;
	if (rdev) {
4749 4750 4751 4752
		if (number >= conf->raid_disks &&
		    conf->reshape_progress == MaxSector)
			clear_bit(In_sync, &rdev->flags);

4753
		if (test_bit(In_sync, &rdev->flags) ||
L
Linus Torvalds 已提交
4754 4755 4756 4757
		    atomic_read(&rdev->nr_pending)) {
			err = -EBUSY;
			goto abort;
		}
4758 4759 4760 4761
		/* Only remove non-faulty devices if recovery
		 * isn't possible.
		 */
		if (!test_bit(Faulty, &rdev->flags) &&
4762 4763
		    mddev->degraded <= conf->max_degraded &&
		    number < conf->raid_disks) {
4764 4765 4766
			err = -EBUSY;
			goto abort;
		}
L
Linus Torvalds 已提交
4767
		p->rdev = NULL;
4768
		synchronize_rcu();
L
Linus Torvalds 已提交
4769 4770 4771 4772 4773 4774 4775 4776 4777 4778 4779 4780 4781 4782 4783
		if (atomic_read(&rdev->nr_pending)) {
			/* lost the race, try later */
			err = -EBUSY;
			p->rdev = rdev;
		}
	}
abort:

	print_raid5_conf(conf);
	return err;
}

static int raid5_add_disk(mddev_t *mddev, mdk_rdev_t *rdev)
{
	raid5_conf_t *conf = mddev->private;
4784
	int err = -EEXIST;
L
Linus Torvalds 已提交
4785 4786
	int disk;
	struct disk_info *p;
4787 4788
	int first = 0;
	int last = conf->raid_disks - 1;
L
Linus Torvalds 已提交
4789

4790
	if (mddev->degraded > conf->max_degraded)
L
Linus Torvalds 已提交
4791
		/* no point adding a device */
4792
		return -EINVAL;
L
Linus Torvalds 已提交
4793

4794 4795
	if (rdev->raid_disk >= 0)
		first = last = rdev->raid_disk;
L
Linus Torvalds 已提交
4796 4797

	/*
4798 4799
	 * find the disk ... but prefer rdev->saved_raid_disk
	 * if possible.
L
Linus Torvalds 已提交
4800
	 */
4801
	if (rdev->saved_raid_disk >= 0 &&
4802
	    rdev->saved_raid_disk >= first &&
4803 4804 4805
	    conf->disks[rdev->saved_raid_disk].rdev == NULL)
		disk = rdev->saved_raid_disk;
	else
4806 4807
		disk = first;
	for ( ; disk <= last ; disk++)
L
Linus Torvalds 已提交
4808
		if ((p=conf->disks + disk)->rdev == NULL) {
4809
			clear_bit(In_sync, &rdev->flags);
L
Linus Torvalds 已提交
4810
			rdev->raid_disk = disk;
4811
			err = 0;
4812 4813
			if (rdev->saved_raid_disk != disk)
				conf->fullsync = 1;
4814
			rcu_assign_pointer(p->rdev, rdev);
L
Linus Torvalds 已提交
4815 4816 4817
			break;
		}
	print_raid5_conf(conf);
4818
	return err;
L
Linus Torvalds 已提交
4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830
}

static int raid5_resize(mddev_t *mddev, sector_t sectors)
{
	/* 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.
	 */
	sectors &= ~((sector_t)mddev->chunk_size/512 - 1);
4831 4832
	md_set_array_sectors(mddev, raid5_size(mddev, sectors,
					       mddev->raid_disks));
D
Dan Williams 已提交
4833 4834 4835
	if (mddev->array_sectors >
	    raid5_size(mddev, sectors, mddev->raid_disks))
		return -EINVAL;
4836
	set_capacity(mddev->gendisk, mddev->array_sectors);
4837
	mddev->changed = 1;
A
Andre Noll 已提交
4838 4839
	if (sectors > mddev->dev_sectors && mddev->recovery_cp == MaxSector) {
		mddev->recovery_cp = mddev->dev_sectors;
L
Linus Torvalds 已提交
4840 4841
		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
	}
A
Andre Noll 已提交
4842
	mddev->dev_sectors = sectors;
4843
	mddev->resync_max_sectors = sectors;
L
Linus Torvalds 已提交
4844 4845 4846
	return 0;
}

4847
static int raid5_check_reshape(mddev_t *mddev)
4848
{
4849
	raid5_conf_t *conf = mddev->private;
4850

4851 4852 4853 4854
	if (mddev->delta_disks == 0 &&
	    mddev->new_layout == mddev->layout &&
	    mddev->new_chunk == mddev->chunk_size)
		return -EINVAL; /* nothing to do */
4855 4856 4857
	if (mddev->bitmap)
		/* Cannot grow a bitmap yet */
		return -EBUSY;
4858 4859 4860 4861 4862 4863 4864 4865 4866 4867 4868 4869 4870 4871
	if (mddev->degraded > conf->max_degraded)
		return -EINVAL;
	if (mddev->delta_disks < 0) {
		/* We might be able to shrink, but the devices must
		 * be made bigger first.
		 * For raid6, 4 is the minimum size.
		 * Otherwise 2 is the minimum
		 */
		int min = 2;
		if (mddev->level == 6)
			min = 4;
		if (mddev->raid_disks + mddev->delta_disks < min)
			return -EINVAL;
	}
4872 4873 4874 4875 4876 4877 4878 4879 4880

	/* 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.
	 */
4881 4882
	if ((mddev->chunk_size / STRIPE_SIZE) * 4 > conf->max_nr_stripes ||
	    (mddev->new_chunk / STRIPE_SIZE) * 4 > conf->max_nr_stripes) {
4883
		printk(KERN_WARNING "raid5: reshape: not enough stripes.  Needed %lu\n",
4884 4885
		       (max(mddev->chunk_size, mddev->new_chunk)
			/ STRIPE_SIZE)*4);
4886 4887 4888
		return -ENOSPC;
	}

4889
	return resize_stripes(conf, conf->raid_disks + mddev->delta_disks);
4890 4891 4892 4893
}

static int raid5_start_reshape(mddev_t *mddev)
{
4894
	raid5_conf_t *conf = mddev->private;
4895 4896 4897
	mdk_rdev_t *rdev;
	int spares = 0;
	int added_devices = 0;
4898
	unsigned long flags;
4899

4900
	if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
4901 4902
		return -EBUSY;

4903
	list_for_each_entry(rdev, &mddev->disks, same_set)
4904 4905 4906
		if (rdev->raid_disk < 0 &&
		    !test_bit(Faulty, &rdev->flags))
			spares++;
4907

4908
	if (spares - mddev->degraded < mddev->delta_disks - conf->max_degraded)
4909 4910 4911 4912 4913
		/* Not enough devices even to make a degraded array
		 * of that size
		 */
		return -EINVAL;

4914 4915 4916 4917 4918 4919 4920 4921 4922 4923 4924
	/* 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) {
		printk(KERN_ERR "md: %s: array size must be reduced "
		       "before number of disks\n", mdname(mddev));
		return -EINVAL;
	}

4925
	atomic_set(&conf->reshape_stripes, 0);
4926 4927
	spin_lock_irq(&conf->device_lock);
	conf->previous_raid_disks = conf->raid_disks;
4928
	conf->raid_disks += mddev->delta_disks;
4929 4930 4931 4932
	conf->prev_chunk = conf->chunk_size;
	conf->chunk_size = mddev->new_chunk;
	conf->prev_algo = conf->algorithm;
	conf->algorithm = mddev->new_layout;
4933 4934 4935 4936 4937
	if (mddev->delta_disks < 0)
		conf->reshape_progress = raid5_size(mddev, 0, 0);
	else
		conf->reshape_progress = 0;
	conf->reshape_safe = conf->reshape_progress;
4938
	conf->generation++;
4939 4940 4941 4942 4943
	spin_unlock_irq(&conf->device_lock);

	/* Add some new drives, as many as will fit.
	 * We know there are enough to make the newly sized array work.
	 */
4944
	list_for_each_entry(rdev, &mddev->disks, same_set)
4945 4946
		if (rdev->raid_disk < 0 &&
		    !test_bit(Faulty, &rdev->flags)) {
4947
			if (raid5_add_disk(mddev, rdev) == 0) {
4948 4949 4950
				char nm[20];
				set_bit(In_sync, &rdev->flags);
				added_devices++;
4951
				rdev->recovery_offset = 0;
4952
				sprintf(nm, "rd%d", rdev->raid_disk);
4953 4954 4955 4956 4957 4958
				if (sysfs_create_link(&mddev->kobj,
						      &rdev->kobj, nm))
					printk(KERN_WARNING
					       "raid5: failed to create "
					       " link %s for %s\n",
					       nm, mdname(mddev));
4959 4960 4961 4962
			} else
				break;
		}

4963 4964 4965 4966 4967 4968
	if (mddev->delta_disks > 0) {
		spin_lock_irqsave(&conf->device_lock, flags);
		mddev->degraded = (conf->raid_disks - conf->previous_raid_disks)
			- added_devices;
		spin_unlock_irqrestore(&conf->device_lock, flags);
	}
4969
	mddev->raid_disks = conf->raid_disks;
4970
	mddev->reshape_position = 0;
4971
	set_bit(MD_CHANGE_DEVS, &mddev->flags);
4972

4973 4974 4975 4976 4977 4978 4979 4980 4981 4982
	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,
						"%s_reshape");
	if (!mddev->sync_thread) {
		mddev->recovery = 0;
		spin_lock_irq(&conf->device_lock);
		mddev->raid_disks = conf->raid_disks = conf->previous_raid_disks;
4983
		conf->reshape_progress = MaxSector;
4984 4985 4986
		spin_unlock_irq(&conf->device_lock);
		return -EAGAIN;
	}
4987
	conf->reshape_checkpoint = jiffies;
4988 4989 4990 4991 4992
	md_wakeup_thread(mddev->sync_thread);
	md_new_event(mddev);
	return 0;
}

4993 4994 4995
/* This is called from the reshape thread and should make any
 * changes needed in 'conf'
 */
4996 4997 4998
static void end_reshape(raid5_conf_t *conf)
{

4999 5000 5001
	if (!test_bit(MD_RECOVERY_INTR, &conf->mddev->recovery)) {

		spin_lock_irq(&conf->device_lock);
5002
		conf->previous_raid_disks = conf->raid_disks;
5003
		conf->reshape_progress = MaxSector;
5004
		spin_unlock_irq(&conf->device_lock);
5005
		wake_up(&conf->wait_for_overlap);
5006 5007 5008 5009 5010

		/* read-ahead size must cover two whole stripes, which is
		 * 2 * (datadisks) * chunksize where 'n' is the number of raid devices
		 */
		{
5011 5012 5013
			int data_disks = conf->raid_disks - conf->max_degraded;
			int stripe = data_disks * (conf->chunk_size
						   / PAGE_SIZE);
5014 5015 5016
			if (conf->mddev->queue->backing_dev_info.ra_pages < 2 * stripe)
				conf->mddev->queue->backing_dev_info.ra_pages = 2 * stripe;
		}
5017 5018 5019
	}
}

5020 5021 5022
/* This is called from the raid5d thread with mddev_lock held.
 * It makes config changes to the device.
 */
5023 5024 5025
static void raid5_finish_reshape(mddev_t *mddev)
{
	struct block_device *bdev;
5026
	raid5_conf_t *conf = mddev->private;
5027 5028 5029

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

5030 5031 5032 5033 5034 5035 5036 5037 5038 5039 5040 5041 5042 5043 5044 5045 5046 5047 5048 5049 5050 5051 5052 5053 5054
		if (mddev->delta_disks > 0) {
			md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
			set_capacity(mddev->gendisk, mddev->array_sectors);
			mddev->changed = 1;

			bdev = bdget_disk(mddev->gendisk, 0);
			if (bdev) {
				mutex_lock(&bdev->bd_inode->i_mutex);
				i_size_write(bdev->bd_inode,
					     (loff_t)mddev->array_sectors << 9);
				mutex_unlock(&bdev->bd_inode->i_mutex);
				bdput(bdev);
			}
		} else {
			int d;
			mddev->degraded = conf->raid_disks;
			for (d = 0; d < conf->raid_disks ; d++)
				if (conf->disks[d].rdev &&
				    test_bit(In_sync,
					     &conf->disks[d].rdev->flags))
					mddev->degraded--;
			for (d = conf->raid_disks ;
			     d < conf->raid_disks - mddev->delta_disks;
			     d++)
				raid5_remove_disk(mddev, d);
5055
		}
5056 5057
		mddev->layout = conf->algorithm;
		mddev->chunk_size = conf->chunk_size;
5058 5059
		mddev->reshape_position = MaxSector;
		mddev->delta_disks = 0;
5060 5061 5062
	}
}

5063 5064
static void raid5_quiesce(mddev_t *mddev, int state)
{
5065
	raid5_conf_t *conf = mddev->private;
5066 5067

	switch(state) {
5068 5069 5070 5071
	case 2: /* resume for a suspend */
		wake_up(&conf->wait_for_overlap);
		break;

5072 5073 5074 5075
	case 1: /* stop all writes */
		spin_lock_irq(&conf->device_lock);
		conf->quiesce = 1;
		wait_event_lock_irq(conf->wait_for_stripe,
5076 5077
				    atomic_read(&conf->active_stripes) == 0 &&
				    atomic_read(&conf->active_aligned_reads) == 0,
5078 5079 5080 5081 5082 5083 5084 5085
				    conf->device_lock, /* nothing */);
		spin_unlock_irq(&conf->device_lock);
		break;

	case 0: /* re-enable writes */
		spin_lock_irq(&conf->device_lock);
		conf->quiesce = 0;
		wake_up(&conf->wait_for_stripe);
5086
		wake_up(&conf->wait_for_overlap);
5087 5088 5089 5090
		spin_unlock_irq(&conf->device_lock);
		break;
	}
}
5091

5092 5093 5094 5095 5096 5097 5098 5099 5100 5101 5102 5103 5104 5105 5106 5107 5108 5109 5110 5111 5112 5113 5114 5115 5116 5117 5118 5119

static void *raid5_takeover_raid1(mddev_t *mddev)
{
	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;
	mddev->new_chunk = chunksect << 9;

	return setup_conf(mddev);
}

5120 5121 5122 5123 5124 5125 5126 5127 5128 5129 5130 5131 5132 5133 5134 5135 5136 5137 5138 5139 5140 5141 5142 5143 5144 5145 5146 5147 5148 5149 5150 5151 5152
static void *raid5_takeover_raid6(mddev_t *mddev)
{
	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);
}

5153

5154 5155
static int raid5_reconfig(mddev_t *mddev, int new_layout, int new_chunk)
{
5156 5157 5158 5159
	/* 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.
5160
	 */
5161
	raid5_conf_t *conf = mddev->private;
5162 5163 5164 5165 5166 5167 5168 5169 5170 5171 5172 5173 5174 5175 5176 5177

	if (new_layout >= 0 && !algorithm_valid_raid5(new_layout))
		return -EINVAL;
	if (new_chunk > 0) {
		if (new_chunk & (new_chunk-1))
			/* not a power of 2 */
			return -EINVAL;
		if (new_chunk < PAGE_SIZE)
			return -EINVAL;
		if (mddev->array_sectors & ((new_chunk>>9)-1))
			/* not factor of array size */
			return -EINVAL;
	}

	/* They look valid */

5178
	if (mddev->raid_disks == 2) {
5179

5180 5181 5182 5183 5184 5185 5186 5187 5188 5189 5190 5191 5192 5193 5194
		if (new_layout >= 0) {
			conf->algorithm = new_layout;
			mddev->layout = mddev->new_layout = new_layout;
		}
		if (new_chunk > 0) {
			conf->chunk_size = new_chunk;
			mddev->chunk_size = mddev->new_chunk = new_chunk;
		}
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
		md_wakeup_thread(mddev->thread);
	} else {
		if (new_layout >= 0)
			mddev->new_layout = new_layout;
		if (new_chunk > 0)
			mddev->new_chunk = new_chunk;
5195
	}
5196 5197 5198 5199 5200 5201 5202
	return 0;
}

static int raid6_reconfig(mddev_t *mddev, int new_layout, int new_chunk)
{
	if (new_layout >= 0 && !algorithm_valid_raid6(new_layout))
		return -EINVAL;
5203
	if (new_chunk > 0) {
5204 5205 5206 5207 5208 5209 5210 5211
		if (new_chunk & (new_chunk-1))
			/* not a power of 2 */
			return -EINVAL;
		if (new_chunk < PAGE_SIZE)
			return -EINVAL;
		if (mddev->array_sectors & ((new_chunk>>9)-1))
			/* not factor of array size */
			return -EINVAL;
5212
	}
5213 5214 5215 5216 5217 5218 5219 5220

	/* They look valid */

	if (new_layout >= 0)
		mddev->new_layout = new_layout;
	if (new_chunk > 0)
		mddev->new_chunk = new_chunk;

5221 5222 5223
	return 0;
}

5224 5225 5226 5227 5228 5229 5230 5231 5232 5233 5234 5235 5236
static void *raid5_takeover(mddev_t *mddev)
{
	/* raid5 can take over:
	 *  raid0 - if all devices are the same - make it a raid4 layout
	 *  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
	 *
	 * For now, just do raid1
	 */

	if (mddev->level == 1)
		return raid5_takeover_raid1(mddev);
5237 5238 5239 5240 5241
	if (mddev->level == 4) {
		mddev->new_layout = ALGORITHM_PARITY_N;
		mddev->new_level = 5;
		return setup_conf(mddev);
	}
5242 5243
	if (mddev->level == 6)
		return raid5_takeover_raid6(mddev);
5244 5245 5246 5247 5248

	return ERR_PTR(-EINVAL);
}


5249 5250 5251 5252 5253 5254 5255 5256 5257 5258 5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269 5270 5271 5272 5273 5274 5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291 5292 5293 5294 5295 5296 5297
static struct mdk_personality raid5_personality;

static void *raid6_takeover(mddev_t *mddev)
{
	/* 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);
}


5298 5299 5300 5301 5302 5303 5304 5305 5306 5307 5308 5309 5310 5311 5312
static struct mdk_personality raid6_personality =
{
	.name		= "raid6",
	.level		= 6,
	.owner		= THIS_MODULE,
	.make_request	= make_request,
	.run		= run,
	.stop		= stop,
	.status		= status,
	.error_handler	= error,
	.hot_add_disk	= raid5_add_disk,
	.hot_remove_disk= raid5_remove_disk,
	.spare_active	= raid5_spare_active,
	.sync_request	= sync_request,
	.resize		= raid5_resize,
5313
	.size		= raid5_size,
5314 5315
	.check_reshape	= raid5_check_reshape,
	.start_reshape  = raid5_start_reshape,
5316
	.finish_reshape = raid5_finish_reshape,
5317
	.quiesce	= raid5_quiesce,
5318
	.takeover	= raid6_takeover,
5319
	.reconfig	= raid6_reconfig,
5320
};
5321
static struct mdk_personality raid5_personality =
L
Linus Torvalds 已提交
5322 5323
{
	.name		= "raid5",
5324
	.level		= 5,
L
Linus Torvalds 已提交
5325 5326 5327 5328 5329 5330 5331 5332 5333 5334 5335
	.owner		= THIS_MODULE,
	.make_request	= make_request,
	.run		= run,
	.stop		= stop,
	.status		= status,
	.error_handler	= error,
	.hot_add_disk	= raid5_add_disk,
	.hot_remove_disk= raid5_remove_disk,
	.spare_active	= raid5_spare_active,
	.sync_request	= sync_request,
	.resize		= raid5_resize,
5336
	.size		= raid5_size,
5337 5338
	.check_reshape	= raid5_check_reshape,
	.start_reshape  = raid5_start_reshape,
5339
	.finish_reshape = raid5_finish_reshape,
5340
	.quiesce	= raid5_quiesce,
5341
	.takeover	= raid5_takeover,
5342
	.reconfig	= raid5_reconfig,
L
Linus Torvalds 已提交
5343 5344
};

5345
static struct mdk_personality raid4_personality =
L
Linus Torvalds 已提交
5346
{
5347 5348 5349 5350 5351 5352 5353 5354 5355 5356 5357 5358 5359
	.name		= "raid4",
	.level		= 4,
	.owner		= THIS_MODULE,
	.make_request	= make_request,
	.run		= run,
	.stop		= stop,
	.status		= status,
	.error_handler	= error,
	.hot_add_disk	= raid5_add_disk,
	.hot_remove_disk= raid5_remove_disk,
	.spare_active	= raid5_spare_active,
	.sync_request	= sync_request,
	.resize		= raid5_resize,
5360
	.size		= raid5_size,
5361 5362
	.check_reshape	= raid5_check_reshape,
	.start_reshape  = raid5_start_reshape,
5363
	.finish_reshape = raid5_finish_reshape,
5364 5365 5366 5367 5368
	.quiesce	= raid5_quiesce,
};

static int __init raid5_init(void)
{
5369
	register_md_personality(&raid6_personality);
5370 5371 5372
	register_md_personality(&raid5_personality);
	register_md_personality(&raid4_personality);
	return 0;
L
Linus Torvalds 已提交
5373 5374
}

5375
static void raid5_exit(void)
L
Linus Torvalds 已提交
5376
{
5377
	unregister_md_personality(&raid6_personality);
5378 5379
	unregister_md_personality(&raid5_personality);
	unregister_md_personality(&raid4_personality);
L
Linus Torvalds 已提交
5380 5381 5382 5383 5384 5385
}

module_init(raid5_init);
module_exit(raid5_exit);
MODULE_LICENSE("GPL");
MODULE_ALIAS("md-personality-4"); /* RAID5 */
5386 5387
MODULE_ALIAS("md-raid5");
MODULE_ALIAS("md-raid4");
5388 5389
MODULE_ALIAS("md-level-5");
MODULE_ALIAS("md-level-4");
5390 5391 5392 5393 5394 5395 5396
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");