raid10.c 60.1 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
/*
 * raid10.c : Multiple Devices driver for Linux
 *
 * Copyright (C) 2000-2004 Neil Brown
 *
 * RAID-10 support for md.
 *
 * Base on code in raid1.c.  See raid1.c for futher copyright information.
 *
 *
 * 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
#include "dm-bio-list.h"
L
Linus Torvalds 已提交
22
#include <linux/raid/raid10.h>
23
#include <linux/raid/bitmap.h>
L
Linus Torvalds 已提交
24 25 26 27 28 29 30 31

/*
 * RAID10 provides a combination of RAID0 and RAID1 functionality.
 * The layout of data is defined by
 *    chunk_size
 *    raid_disks
 *    near_copies (stored in low byte of layout)
 *    far_copies (stored in second byte of layout)
32
 *    far_offset (stored in bit 16 of layout )
L
Linus Torvalds 已提交
33 34 35 36 37 38 39
 *
 * The data to be stored is divided into chunks using chunksize.
 * Each device is divided into far_copies sections.
 * In each section, chunks are laid out in a style similar to raid0, but
 * near_copies copies of each chunk is stored (each on a different drive).
 * The starting device for each section is offset near_copies from the starting
 * device of the previous section.
40
 * Thus they are (near_copies*far_copies) of each chunk, and each is on a different
L
Linus Torvalds 已提交
41 42 43
 * drive.
 * near_copies and far_copies must be at least one, and their product is at most
 * raid_disks.
44 45 46 47
 *
 * If far_offset is true, then the far_copies are handled a bit differently.
 * The copies are still in different stripes, but instead of be very far apart
 * on disk, there are adjacent stripes.
L
Linus Torvalds 已提交
48 49 50 51 52 53 54 55 56
 */

/*
 * Number of guaranteed r10bios in case of extreme VM load:
 */
#define	NR_RAID10_BIOS 256

static void unplug_slaves(mddev_t *mddev);

57 58 59
static void allow_barrier(conf_t *conf);
static void lower_barrier(conf_t *conf);

A
Al Viro 已提交
60
static void * r10bio_pool_alloc(gfp_t gfp_flags, void *data)
L
Linus Torvalds 已提交
61 62 63 64 65 66
{
	conf_t *conf = data;
	r10bio_t *r10_bio;
	int size = offsetof(struct r10bio_s, devs[conf->copies]);

	/* allocate a r10bio with room for raid_disks entries in the bios array */
67 68
	r10_bio = kzalloc(size, gfp_flags);
	if (!r10_bio)
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 90 91
		unplug_slaves(conf->mddev);

	return r10_bio;
}

static void r10bio_pool_free(void *r10_bio, void *data)
{
	kfree(r10_bio);
}

#define RESYNC_BLOCK_SIZE (64*1024)
//#define RESYNC_BLOCK_SIZE PAGE_SIZE
#define RESYNC_SECTORS (RESYNC_BLOCK_SIZE >> 9)
#define RESYNC_PAGES ((RESYNC_BLOCK_SIZE + PAGE_SIZE-1) / PAGE_SIZE)
#define RESYNC_WINDOW (2048*1024)

/*
 * When performing a resync, we need to read and compare, so
 * we need as many pages are there are copies.
 * When performing a recovery, we need 2 bios, one for read,
 * one for write (we recover only one drive per r10buf)
 *
 */
A
Al Viro 已提交
92
static void * r10buf_pool_alloc(gfp_t gfp_flags, void *data)
L
Linus Torvalds 已提交
93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139
{
	conf_t *conf = data;
	struct page *page;
	r10bio_t *r10_bio;
	struct bio *bio;
	int i, j;
	int nalloc;

	r10_bio = r10bio_pool_alloc(gfp_flags, conf);
	if (!r10_bio) {
		unplug_slaves(conf->mddev);
		return NULL;
	}

	if (test_bit(MD_RECOVERY_SYNC, &conf->mddev->recovery))
		nalloc = conf->copies; /* resync */
	else
		nalloc = 2; /* recovery */

	/*
	 * Allocate bios.
	 */
	for (j = nalloc ; j-- ; ) {
		bio = bio_alloc(gfp_flags, RESYNC_PAGES);
		if (!bio)
			goto out_free_bio;
		r10_bio->devs[j].bio = bio;
	}
	/*
	 * Allocate RESYNC_PAGES data pages and attach them
	 * where needed.
	 */
	for (j = 0 ; j < nalloc; j++) {
		bio = r10_bio->devs[j].bio;
		for (i = 0; i < RESYNC_PAGES; i++) {
			page = alloc_page(gfp_flags);
			if (unlikely(!page))
				goto out_free_pages;

			bio->bi_io_vec[i].bv_page = page;
		}
	}

	return r10_bio;

out_free_pages:
	for ( ; i > 0 ; i--)
140
		safe_put_page(bio->bi_io_vec[i-1].bv_page);
L
Linus Torvalds 已提交
141 142
	while (j--)
		for (i = 0; i < RESYNC_PAGES ; i++)
143
			safe_put_page(r10_bio->devs[j].bio->bi_io_vec[i].bv_page);
L
Linus Torvalds 已提交
144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162
	j = -1;
out_free_bio:
	while ( ++j < nalloc )
		bio_put(r10_bio->devs[j].bio);
	r10bio_pool_free(r10_bio, conf);
	return NULL;
}

static void r10buf_pool_free(void *__r10_bio, void *data)
{
	int i;
	conf_t *conf = data;
	r10bio_t *r10bio = __r10_bio;
	int j;

	for (j=0; j < conf->copies; j++) {
		struct bio *bio = r10bio->devs[j].bio;
		if (bio) {
			for (i = 0; i < RESYNC_PAGES; i++) {
163
				safe_put_page(bio->bi_io_vec[i].bv_page);
L
Linus Torvalds 已提交
164 165 166 167 168 169 170 171 172 173 174 175 176 177
				bio->bi_io_vec[i].bv_page = NULL;
			}
			bio_put(bio);
		}
	}
	r10bio_pool_free(r10bio, conf);
}

static void put_all_bios(conf_t *conf, r10bio_t *r10_bio)
{
	int i;

	for (i = 0; i < conf->copies; i++) {
		struct bio **bio = & r10_bio->devs[i].bio;
178
		if (*bio && *bio != IO_BLOCKED)
L
Linus Torvalds 已提交
179 180 181 182 183
			bio_put(*bio);
		*bio = NULL;
	}
}

184
static void free_r10bio(r10bio_t *r10_bio)
L
Linus Torvalds 已提交
185 186 187 188 189 190 191
{
	conf_t *conf = mddev_to_conf(r10_bio->mddev);

	/*
	 * Wake up any possible resync thread that waits for the device
	 * to go idle.
	 */
192
	allow_barrier(conf);
L
Linus Torvalds 已提交
193 194 195 196 197

	put_all_bios(conf, r10_bio);
	mempool_free(r10_bio, conf->r10bio_pool);
}

198
static void put_buf(r10bio_t *r10_bio)
L
Linus Torvalds 已提交
199 200 201 202 203
{
	conf_t *conf = mddev_to_conf(r10_bio->mddev);

	mempool_free(r10_bio, conf->r10buf_pool);

204
	lower_barrier(conf);
L
Linus Torvalds 已提交
205 206 207 208 209 210 211 212 213 214
}

static void reschedule_retry(r10bio_t *r10_bio)
{
	unsigned long flags;
	mddev_t *mddev = r10_bio->mddev;
	conf_t *conf = mddev_to_conf(mddev);

	spin_lock_irqsave(&conf->device_lock, flags);
	list_add(&r10_bio->retry_list, &conf->retry_list);
215
	conf->nr_queued ++;
L
Linus Torvalds 已提交
216 217 218 219 220 221 222 223 224 225 226 227 228 229
	spin_unlock_irqrestore(&conf->device_lock, flags);

	md_wakeup_thread(mddev->thread);
}

/*
 * raid_end_bio_io() is called when we have finished servicing a mirrored
 * operation and are ready to return a success/failure code to the buffer
 * cache layer.
 */
static void raid_end_bio_io(r10bio_t *r10_bio)
{
	struct bio *bio = r10_bio->master_bio;

230
	bio_endio(bio,
L
Linus Torvalds 已提交
231 232 233 234 235 236 237 238 239 240 241 242 243 244 245
		test_bit(R10BIO_Uptodate, &r10_bio->state) ? 0 : -EIO);
	free_r10bio(r10_bio);
}

/*
 * Update disk head position estimator based on IRQ completion info.
 */
static inline void update_head_pos(int slot, r10bio_t *r10_bio)
{
	conf_t *conf = mddev_to_conf(r10_bio->mddev);

	conf->mirrors[r10_bio->devs[slot].devnum].head_position =
		r10_bio->devs[slot].addr + (r10_bio->sectors);
}

246
static void raid10_end_read_request(struct bio *bio, int error)
L
Linus Torvalds 已提交
247 248 249 250 251 252 253 254 255 256 257 258
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
	r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private);
	int slot, dev;
	conf_t *conf = mddev_to_conf(r10_bio->mddev);


	slot = r10_bio->read_slot;
	dev = r10_bio->devs[slot].devnum;
	/*
	 * this branch is our 'one mirror IO has finished' event handler:
	 */
259 260 261
	update_head_pos(slot, r10_bio);

	if (uptodate) {
L
Linus Torvalds 已提交
262 263 264 265 266 267 268 269 270 271 272
		/*
		 * Set R10BIO_Uptodate in our master bio, so that
		 * we will return a good error code to the higher
		 * levels even if IO on some other mirrored buffer fails.
		 *
		 * The 'master' represents the composite IO operation to
		 * user-side. So if something waits for IO, then it will
		 * wait for the 'master' bio.
		 */
		set_bit(R10BIO_Uptodate, &r10_bio->state);
		raid_end_bio_io(r10_bio);
273
	} else {
L
Linus Torvalds 已提交
274 275 276 277 278 279 280 281 282 283 284 285 286
		/*
		 * oops, read error:
		 */
		char b[BDEVNAME_SIZE];
		if (printk_ratelimit())
			printk(KERN_ERR "raid10: %s: rescheduling sector %llu\n",
			       bdevname(conf->mirrors[dev].rdev->bdev,b), (unsigned long long)r10_bio->sector);
		reschedule_retry(r10_bio);
	}

	rdev_dec_pending(conf->mirrors[dev].rdev, conf->mddev);
}

287
static void raid10_end_write_request(struct bio *bio, int error)
L
Linus Torvalds 已提交
288 289 290 291 292 293 294 295 296 297 298 299 300 301
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
	r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private);
	int slot, dev;
	conf_t *conf = mddev_to_conf(r10_bio->mddev);

	for (slot = 0; slot < conf->copies; slot++)
		if (r10_bio->devs[slot].bio == bio)
			break;
	dev = r10_bio->devs[slot].devnum;

	/*
	 * this branch is our 'one mirror IO has finished' event handler:
	 */
302
	if (!uptodate) {
L
Linus Torvalds 已提交
303
		md_error(r10_bio->mddev, conf->mirrors[dev].rdev);
304 305 306
		/* an I/O failed, we can't clear the bitmap */
		set_bit(R10BIO_Degraded, &r10_bio->state);
	} else
L
Linus Torvalds 已提交
307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325
		/*
		 * Set R10BIO_Uptodate in our master bio, so that
		 * we will return a good error code for to the higher
		 * levels even if IO on some other mirrored buffer fails.
		 *
		 * The 'master' represents the composite IO operation to
		 * user-side. So if something waits for IO, then it will
		 * wait for the 'master' bio.
		 */
		set_bit(R10BIO_Uptodate, &r10_bio->state);

	update_head_pos(slot, r10_bio);

	/*
	 *
	 * Let's see if all mirrored write operations have finished
	 * already.
	 */
	if (atomic_dec_and_test(&r10_bio->remaining)) {
326 327 328 329 330
		/* clear the bitmap if all writes complete successfully */
		bitmap_endwrite(r10_bio->mddev->bitmap, r10_bio->sector,
				r10_bio->sectors,
				!test_bit(R10BIO_Degraded, &r10_bio->state),
				0);
L
Linus Torvalds 已提交
331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357
		md_write_end(r10_bio->mddev);
		raid_end_bio_io(r10_bio);
	}

	rdev_dec_pending(conf->mirrors[dev].rdev, conf->mddev);
}


/*
 * RAID10 layout manager
 * Aswell as the chunksize and raid_disks count, there are two
 * parameters: near_copies and far_copies.
 * near_copies * far_copies must be <= raid_disks.
 * Normally one of these will be 1.
 * If both are 1, we get raid0.
 * If near_copies == raid_disks, we get raid1.
 *
 * Chunks are layed out in raid0 style with near_copies copies of the
 * first chunk, followed by near_copies copies of the next chunk and
 * so on.
 * If far_copies > 1, then after 1/far_copies of the array has been assigned
 * as described above, we start again with a device offset of near_copies.
 * So we effectively have another copy of the whole array further down all
 * the drives, but with blocks on different drives.
 * With this layout, and block is never stored twice on the one device.
 *
 * raid10_find_phys finds the sector offset of a given virtual sector
358
 * on each device that it is on.
L
Linus Torvalds 已提交
359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380
 *
 * raid10_find_virt does the reverse mapping, from a device and a
 * sector offset to a virtual address
 */

static void raid10_find_phys(conf_t *conf, r10bio_t *r10bio)
{
	int n,f;
	sector_t sector;
	sector_t chunk;
	sector_t stripe;
	int dev;

	int slot = 0;

	/* now calculate first sector/dev */
	chunk = r10bio->sector >> conf->chunk_shift;
	sector = r10bio->sector & conf->chunk_mask;

	chunk *= conf->near_copies;
	stripe = chunk;
	dev = sector_div(stripe, conf->raid_disks);
381 382
	if (conf->far_offset)
		stripe *= conf->far_copies;
L
Linus Torvalds 已提交
383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416

	sector += stripe << conf->chunk_shift;

	/* and calculate all the others */
	for (n=0; n < conf->near_copies; n++) {
		int d = dev;
		sector_t s = sector;
		r10bio->devs[slot].addr = sector;
		r10bio->devs[slot].devnum = d;
		slot++;

		for (f = 1; f < conf->far_copies; f++) {
			d += conf->near_copies;
			if (d >= conf->raid_disks)
				d -= conf->raid_disks;
			s += conf->stride;
			r10bio->devs[slot].devnum = d;
			r10bio->devs[slot].addr = s;
			slot++;
		}
		dev++;
		if (dev >= conf->raid_disks) {
			dev = 0;
			sector += (conf->chunk_mask + 1);
		}
	}
	BUG_ON(slot != conf->copies);
}

static sector_t raid10_find_virt(conf_t *conf, sector_t sector, int dev)
{
	sector_t offset, chunk, vchunk;

	offset = sector & conf->chunk_mask;
417 418 419 420 421 422 423 424
	if (conf->far_offset) {
		int fc;
		chunk = sector >> conf->chunk_shift;
		fc = sector_div(chunk, conf->far_copies);
		dev -= fc * conf->near_copies;
		if (dev < 0)
			dev += conf->raid_disks;
	} else {
425
		while (sector >= conf->stride) {
426 427 428 429 430 431 432 433
			sector -= conf->stride;
			if (dev < conf->near_copies)
				dev += conf->raid_disks - conf->near_copies;
			else
				dev -= conf->near_copies;
		}
		chunk = sector >> conf->chunk_shift;
	}
L
Linus Torvalds 已提交
434 435 436 437 438 439 440 441 442 443 444 445 446 447 448
	vchunk = chunk * conf->raid_disks + dev;
	sector_div(vchunk, conf->near_copies);
	return (vchunk << conf->chunk_shift) + offset;
}

/**
 *	raid10_mergeable_bvec -- tell bio layer if a two requests can be merged
 *	@q: request queue
 *	@bio: the buffer head that's been built up so far
 *	@biovec: the request that could be merged to it.
 *
 *	Return amount of bytes we can accept at this offset
 *      If near_copies == raid_disk, there are no striping issues,
 *      but in that case, the function isn't called at all.
 */
449
static int raid10_mergeable_bvec(struct request_queue *q, struct bio *bio,
L
Linus Torvalds 已提交
450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490
				struct bio_vec *bio_vec)
{
	mddev_t *mddev = q->queuedata;
	sector_t sector = bio->bi_sector + get_start_sect(bio->bi_bdev);
	int max;
	unsigned int chunk_sectors = mddev->chunk_size >> 9;
	unsigned int bio_sectors = bio->bi_size >> 9;

	max =  (chunk_sectors - ((sector & (chunk_sectors - 1)) + bio_sectors)) << 9;
	if (max < 0) max = 0; /* bio_add cannot handle a negative return */
	if (max <= bio_vec->bv_len && bio_sectors == 0)
		return bio_vec->bv_len;
	else
		return max;
}

/*
 * This routine returns the disk from which the requested read should
 * be done. There is a per-array 'next expected sequential IO' sector
 * number - if this matches on the next IO then we use the last disk.
 * There is also a per-disk 'last know head position' sector that is
 * maintained from IRQ contexts, both the normal and the resync IO
 * completion handlers update this position correctly. If there is no
 * perfect sequential match then we pick the disk whose head is closest.
 *
 * If there are 2 mirrors in the same 2 devices, performance degrades
 * because position is mirror, not device based.
 *
 * The rdev for the device selected will have nr_pending incremented.
 */

/*
 * FIXME: possibly should rethink readbalancing and do it differently
 * depending on near_copies / far_copies geometry.
 */
static int read_balance(conf_t *conf, r10bio_t *r10_bio)
{
	const unsigned long this_sector = r10_bio->sector;
	int disk, slot, nslot;
	const int sectors = r10_bio->sectors;
	sector_t new_distance, current_distance;
491
	mdk_rdev_t *rdev;
L
Linus Torvalds 已提交
492 493 494 495 496

	raid10_find_phys(conf, r10_bio);
	rcu_read_lock();
	/*
	 * Check if we can balance. We can balance on the whole
497 498 499
	 * device if no resync is going on (recovery is ok), or below
	 * the resync window. We take the first readable disk when
	 * above the resync window.
L
Linus Torvalds 已提交
500 501 502 503 504 505 506
	 */
	if (conf->mddev->recovery_cp < MaxSector
	    && (this_sector + sectors >= conf->next_resync)) {
		/* make sure that disk is operational */
		slot = 0;
		disk = r10_bio->devs[slot].devnum;

507
		while ((rdev = rcu_dereference(conf->mirrors[disk].rdev)) == NULL ||
508
		       r10_bio->devs[slot].bio == IO_BLOCKED ||
509
		       !test_bit(In_sync, &rdev->flags)) {
L
Linus Torvalds 已提交
510 511 512 513 514 515 516 517 518 519 520 521 522 523 524
			slot++;
			if (slot == conf->copies) {
				slot = 0;
				disk = -1;
				break;
			}
			disk = r10_bio->devs[slot].devnum;
		}
		goto rb_out;
	}


	/* make sure the disk is operational */
	slot = 0;
	disk = r10_bio->devs[slot].devnum;
525
	while ((rdev=rcu_dereference(conf->mirrors[disk].rdev)) == NULL ||
526
	       r10_bio->devs[slot].bio == IO_BLOCKED ||
527
	       !test_bit(In_sync, &rdev->flags)) {
L
Linus Torvalds 已提交
528 529 530 531 532 533 534 535 536
		slot ++;
		if (slot == conf->copies) {
			disk = -1;
			goto rb_out;
		}
		disk = r10_bio->devs[slot].devnum;
	}


537 538
	current_distance = abs(r10_bio->devs[slot].addr -
			       conf->mirrors[disk].head_position);
L
Linus Torvalds 已提交
539

540 541
	/* Find the disk whose head is closest,
	 * or - for far > 1 - find the closest to partition beginning */
L
Linus Torvalds 已提交
542 543 544 545 546

	for (nslot = slot; nslot < conf->copies; nslot++) {
		int ndisk = r10_bio->devs[nslot].devnum;


547
		if ((rdev=rcu_dereference(conf->mirrors[ndisk].rdev)) == NULL ||
548
		    r10_bio->devs[nslot].bio == IO_BLOCKED ||
549
		    !test_bit(In_sync, &rdev->flags))
L
Linus Torvalds 已提交
550 551
			continue;

552 553 554 555 556
		/* This optimisation is debatable, and completely destroys
		 * sequential read speed for 'far copies' arrays.  So only
		 * keep it for 'near' arrays, and review those later.
		 */
		if (conf->near_copies > 1 && !atomic_read(&rdev->nr_pending)) {
L
Linus Torvalds 已提交
557 558 559 560
			disk = ndisk;
			slot = nslot;
			break;
		}
561 562 563 564 565 566 567

		/* for far > 1 always use the lowest address */
		if (conf->far_copies > 1)
			new_distance = r10_bio->devs[nslot].addr;
		else
			new_distance = abs(r10_bio->devs[nslot].addr -
					   conf->mirrors[ndisk].head_position);
L
Linus Torvalds 已提交
568 569 570 571 572 573 574 575 576 577 578
		if (new_distance < current_distance) {
			current_distance = new_distance;
			disk = ndisk;
			slot = nslot;
		}
	}

rb_out:
	r10_bio->read_slot = slot;
/*	conf->next_seq_sect = this_sector + sectors;*/

579
	if (disk >= 0 && (rdev=rcu_dereference(conf->mirrors[disk].rdev))!= NULL)
L
Linus Torvalds 已提交
580
		atomic_inc(&conf->mirrors[disk].rdev->nr_pending);
N
NeilBrown 已提交
581 582
	else
		disk = -1;
L
Linus Torvalds 已提交
583 584 585 586 587 588 589 590 591 592 593 594
	rcu_read_unlock();

	return disk;
}

static void unplug_slaves(mddev_t *mddev)
{
	conf_t *conf = mddev_to_conf(mddev);
	int i;

	rcu_read_lock();
	for (i=0; i<mddev->raid_disks; i++) {
595
		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
596
		if (rdev && !test_bit(Faulty, &rdev->flags) && atomic_read(&rdev->nr_pending)) {
597
			struct request_queue *r_queue = bdev_get_queue(rdev->bdev);
L
Linus Torvalds 已提交
598 599 600 601

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

602
			blk_unplug(r_queue);
L
Linus Torvalds 已提交
603 604 605 606 607 608 609 610

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

611
static void raid10_unplug(struct request_queue *q)
L
Linus Torvalds 已提交
612
{
613 614
	mddev_t *mddev = q->queuedata;

L
Linus Torvalds 已提交
615
	unplug_slaves(q->queuedata);
616
	md_wakeup_thread(mddev->thread);
L
Linus Torvalds 已提交
617 618
}

619 620 621 622 623 624 625 626 627 628
static int raid10_congested(void *data, int bits)
{
	mddev_t *mddev = data;
	conf_t *conf = mddev_to_conf(mddev);
	int i, ret = 0;

	rcu_read_lock();
	for (i = 0; i < mddev->raid_disks && ret == 0; i++) {
		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
		if (rdev && !test_bit(Faulty, &rdev->flags)) {
629
			struct request_queue *q = bdev_get_queue(rdev->bdev);
630 631 632 633 634 635 636 637

			ret |= bdi_congested(&q->backing_dev_info, bits);
		}
	}
	rcu_read_unlock();
	return ret;
}

638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667
static int flush_pending_writes(conf_t *conf)
{
	/* Any writes that have been queued but are awaiting
	 * bitmap updates get flushed here.
	 * We return 1 if any requests were actually submitted.
	 */
	int rv = 0;

	spin_lock_irq(&conf->device_lock);

	if (conf->pending_bio_list.head) {
		struct bio *bio;
		bio = bio_list_get(&conf->pending_bio_list);
		blk_remove_plug(conf->mddev->queue);
		spin_unlock_irq(&conf->device_lock);
		/* flush any pending bitmap writes to disk
		 * before proceeding w/ I/O */
		bitmap_unplug(conf->mddev->bitmap);

		while (bio) { /* submit pending writes */
			struct bio *next = bio->bi_next;
			bio->bi_next = NULL;
			generic_make_request(bio);
			bio = next;
		}
		rv = 1;
	} else
		spin_unlock_irq(&conf->device_lock);
	return rv;
}
668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687
/* Barriers....
 * Sometimes we need to suspend IO while we do something else,
 * either some resync/recovery, or reconfigure the array.
 * To do this we raise a 'barrier'.
 * The 'barrier' is a counter that can be raised multiple times
 * to count how many activities are happening which preclude
 * normal IO.
 * We can only raise the barrier if there is no pending IO.
 * i.e. if nr_pending == 0.
 * We choose only to raise the barrier if no-one is waiting for the
 * barrier to go down.  This means that as soon as an IO request
 * is ready, no other operations which require a barrier will start
 * until the IO request has had a chance.
 *
 * So: regular IO calls 'wait_barrier'.  When that returns there
 *    is no backgroup IO happening,  It must arrange to call
 *    allow_barrier when it has finished its IO.
 * backgroup IO calls must call raise_barrier.  Once that returns
 *    there is no normal IO happeing.  It must arrange to call
 *    lower_barrier when the particular background IO completes.
L
Linus Torvalds 已提交
688 689 690
 */
#define RESYNC_DEPTH 32

691
static void raise_barrier(conf_t *conf, int force)
L
Linus Torvalds 已提交
692
{
693
	BUG_ON(force && !conf->barrier);
L
Linus Torvalds 已提交
694
	spin_lock_irq(&conf->resync_lock);
695

696 697
	/* Wait until no block IO is waiting (unless 'force') */
	wait_event_lock_irq(conf->wait_barrier, force || !conf->nr_waiting,
698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730
			    conf->resync_lock,
			    raid10_unplug(conf->mddev->queue));

	/* block any new IO from starting */
	conf->barrier++;

	/* No wait for all pending IO to complete */
	wait_event_lock_irq(conf->wait_barrier,
			    !conf->nr_pending && conf->barrier < RESYNC_DEPTH,
			    conf->resync_lock,
			    raid10_unplug(conf->mddev->queue));

	spin_unlock_irq(&conf->resync_lock);
}

static void lower_barrier(conf_t *conf)
{
	unsigned long flags;
	spin_lock_irqsave(&conf->resync_lock, flags);
	conf->barrier--;
	spin_unlock_irqrestore(&conf->resync_lock, flags);
	wake_up(&conf->wait_barrier);
}

static void wait_barrier(conf_t *conf)
{
	spin_lock_irq(&conf->resync_lock);
	if (conf->barrier) {
		conf->nr_waiting++;
		wait_event_lock_irq(conf->wait_barrier, !conf->barrier,
				    conf->resync_lock,
				    raid10_unplug(conf->mddev->queue));
		conf->nr_waiting--;
L
Linus Torvalds 已提交
731
	}
732
	conf->nr_pending++;
L
Linus Torvalds 已提交
733 734 735
	spin_unlock_irq(&conf->resync_lock);
}

736 737 738 739 740 741 742 743 744
static void allow_barrier(conf_t *conf)
{
	unsigned long flags;
	spin_lock_irqsave(&conf->resync_lock, flags);
	conf->nr_pending--;
	spin_unlock_irqrestore(&conf->resync_lock, flags);
	wake_up(&conf->wait_barrier);
}

745 746 747
static void freeze_array(conf_t *conf)
{
	/* stop syncio and normal IO and wait for everything to
N
NeilBrown 已提交
748
	 * go quiet.
749
	 * We increment barrier and nr_waiting, and then
750 751 752 753 754 755 756 757
	 * wait until nr_pending match nr_queued+1
	 * This is called in the context of one normal IO request
	 * that has failed. Thus any sync request that might be pending
	 * will be blocked by nr_pending, and we need to wait for
	 * pending IO requests to complete or be queued for re-try.
	 * Thus the number queued (nr_queued) plus this request (1)
	 * must match the number of pending IOs (nr_pending) before
	 * we continue.
758 759 760 761 762
	 */
	spin_lock_irq(&conf->resync_lock);
	conf->barrier++;
	conf->nr_waiting++;
	wait_event_lock_irq(conf->wait_barrier,
763
			    conf->nr_pending == conf->nr_queued+1,
764
			    conf->resync_lock,
765 766
			    ({ flush_pending_writes(conf);
			       raid10_unplug(conf->mddev->queue); }));
767 768 769 770 771 772 773 774 775 776 777 778 779
	spin_unlock_irq(&conf->resync_lock);
}

static void unfreeze_array(conf_t *conf)
{
	/* reverse the effect of the freeze */
	spin_lock_irq(&conf->resync_lock);
	conf->barrier--;
	conf->nr_waiting--;
	wake_up(&conf->wait_barrier);
	spin_unlock_irq(&conf->resync_lock);
}

780
static int make_request(struct request_queue *q, struct bio * bio)
L
Linus Torvalds 已提交
781 782 783 784 785 786 787 788
{
	mddev_t *mddev = q->queuedata;
	conf_t *conf = mddev_to_conf(mddev);
	mirror_info_t *mirror;
	r10bio_t *r10_bio;
	struct bio *read_bio;
	int i;
	int chunk_sects = conf->chunk_mask + 1;
789
	const int rw = bio_data_dir(bio);
790
	const int do_sync = bio_sync(bio);
791 792
	struct bio_list bl;
	unsigned long flags;
793
	mdk_rdev_t *blocked_rdev;
L
Linus Torvalds 已提交
794

795
	if (unlikely(bio_barrier(bio))) {
796
		bio_endio(bio, -EOPNOTSUPP);
797 798 799
		return 0;
	}

L
Linus Torvalds 已提交
800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827
	/* If this request crosses a chunk boundary, we need to
	 * split it.  This will only happen for 1 PAGE (or less) requests.
	 */
	if (unlikely( (bio->bi_sector & conf->chunk_mask) + (bio->bi_size >> 9)
		      > chunk_sects &&
		    conf->near_copies < conf->raid_disks)) {
		struct bio_pair *bp;
		/* Sanity check -- queue functions should prevent this happening */
		if (bio->bi_vcnt != 1 ||
		    bio->bi_idx != 0)
			goto bad_map;
		/* This is a one page bio that upper layers
		 * refuse to split for us, so we need to split it.
		 */
		bp = bio_split(bio, bio_split_pool,
			       chunk_sects - (bio->bi_sector & (chunk_sects - 1)) );
		if (make_request(q, &bp->bio1))
			generic_make_request(&bp->bio1);
		if (make_request(q, &bp->bio2))
			generic_make_request(&bp->bio2);

		bio_pair_release(bp);
		return 0;
	bad_map:
		printk("raid10_make_request bug: can't convert block across chunks"
		       " or bigger than %dk %llu %d\n", chunk_sects/2,
		       (unsigned long long)bio->bi_sector, bio->bi_size >> 10);

828
		bio_io_error(bio);
L
Linus Torvalds 已提交
829 830 831
		return 0;
	}

832
	md_write_start(mddev, bio);
833

L
Linus Torvalds 已提交
834 835 836 837 838
	/*
	 * Register the new request and wait if the reconstruction
	 * thread has put up a bar for new requests.
	 * Continue immediately if no resync is active currently.
	 */
839
	wait_barrier(conf);
L
Linus Torvalds 已提交
840

841 842
	disk_stat_inc(mddev->gendisk, ios[rw]);
	disk_stat_add(mddev->gendisk, sectors[rw], bio_sectors(bio));
L
Linus Torvalds 已提交
843 844 845 846 847 848 849 850

	r10_bio = mempool_alloc(conf->r10bio_pool, GFP_NOIO);

	r10_bio->master_bio = bio;
	r10_bio->sectors = bio->bi_size >> 9;

	r10_bio->mddev = mddev;
	r10_bio->sector = bio->bi_sector;
851
	r10_bio->state = 0;
L
Linus Torvalds 已提交
852

853
	if (rw == READ) {
L
Linus Torvalds 已提交
854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872
		/*
		 * read balancing logic:
		 */
		int disk = read_balance(conf, r10_bio);
		int slot = r10_bio->read_slot;
		if (disk < 0) {
			raid_end_bio_io(r10_bio);
			return 0;
		}
		mirror = conf->mirrors + disk;

		read_bio = bio_clone(bio, GFP_NOIO);

		r10_bio->devs[slot].bio = read_bio;

		read_bio->bi_sector = r10_bio->devs[slot].addr +
			mirror->rdev->data_offset;
		read_bio->bi_bdev = mirror->rdev->bdev;
		read_bio->bi_end_io = raid10_end_read_request;
873
		read_bio->bi_rw = READ | do_sync;
L
Linus Torvalds 已提交
874 875 876 877 878 879 880 881 882
		read_bio->bi_private = r10_bio;

		generic_make_request(read_bio);
		return 0;
	}

	/*
	 * WRITE:
	 */
883
	/* first select target devices under rcu_lock and
L
Linus Torvalds 已提交
884 885 886 887
	 * inc refcount on their rdev.  Record them by setting
	 * bios[x] to bio
	 */
	raid10_find_phys(conf, r10_bio);
888 889
 retry_write:
	blocked_rdev = 0;
L
Linus Torvalds 已提交
890 891 892
	rcu_read_lock();
	for (i = 0;  i < conf->copies; i++) {
		int d = r10_bio->devs[i].devnum;
893
		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[d].rdev);
894 895 896 897 898 899
		if (rdev && unlikely(test_bit(Blocked, &rdev->flags))) {
			atomic_inc(&rdev->nr_pending);
			blocked_rdev = rdev;
			break;
		}
		if (rdev && !test_bit(Faulty, &rdev->flags)) {
900
			atomic_inc(&rdev->nr_pending);
L
Linus Torvalds 已提交
901
			r10_bio->devs[i].bio = bio;
902
		} else {
L
Linus Torvalds 已提交
903
			r10_bio->devs[i].bio = NULL;
904 905
			set_bit(R10BIO_Degraded, &r10_bio->state);
		}
L
Linus Torvalds 已提交
906 907 908
	}
	rcu_read_unlock();

909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924
	if (unlikely(blocked_rdev)) {
		/* Have to wait for this device to get unblocked, then retry */
		int j;
		int d;

		for (j = 0; j < i; j++)
			if (r10_bio->devs[j].bio) {
				d = r10_bio->devs[j].devnum;
				rdev_dec_pending(conf->mirrors[d].rdev, mddev);
			}
		allow_barrier(conf);
		md_wait_for_blocked_rdev(blocked_rdev, mddev);
		wait_barrier(conf);
		goto retry_write;
	}

925
	atomic_set(&r10_bio->remaining, 0);
926

927
	bio_list_init(&bl);
L
Linus Torvalds 已提交
928 929 930 931 932 933 934 935 936 937 938 939 940
	for (i = 0; i < conf->copies; i++) {
		struct bio *mbio;
		int d = r10_bio->devs[i].devnum;
		if (!r10_bio->devs[i].bio)
			continue;

		mbio = bio_clone(bio, GFP_NOIO);
		r10_bio->devs[i].bio = mbio;

		mbio->bi_sector	= r10_bio->devs[i].addr+
			conf->mirrors[d].rdev->data_offset;
		mbio->bi_bdev = conf->mirrors[d].rdev->bdev;
		mbio->bi_end_io	= raid10_end_write_request;
941
		mbio->bi_rw = WRITE | do_sync;
L
Linus Torvalds 已提交
942 943 944
		mbio->bi_private = r10_bio;

		atomic_inc(&r10_bio->remaining);
945
		bio_list_add(&bl, mbio);
L
Linus Torvalds 已提交
946 947
	}

948 949 950 951 952 953 954
	if (unlikely(!atomic_read(&r10_bio->remaining))) {
		/* the array is dead */
		md_write_end(mddev);
		raid_end_bio_io(r10_bio);
		return 0;
	}

955 956 957 958 959
	bitmap_startwrite(mddev->bitmap, bio->bi_sector, r10_bio->sectors, 0);
	spin_lock_irqsave(&conf->device_lock, flags);
	bio_list_merge(&conf->pending_bio_list, &bl);
	blk_plug_device(mddev->queue);
	spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
960

961 962 963
	/* In case raid10d snuck in to freeze_array */
	wake_up(&conf->wait_barrier);

964 965 966
	if (do_sync)
		md_wakeup_thread(mddev->thread);

L
Linus Torvalds 已提交
967 968 969 970 971 972 973 974 975 976 977 978
	return 0;
}

static void status(struct seq_file *seq, mddev_t *mddev)
{
	conf_t *conf = mddev_to_conf(mddev);
	int i;

	if (conf->near_copies < conf->raid_disks)
		seq_printf(seq, " %dK chunks", mddev->chunk_size/1024);
	if (conf->near_copies > 1)
		seq_printf(seq, " %d near-copies", conf->near_copies);
979 980 981 982 983 984
	if (conf->far_copies > 1) {
		if (conf->far_offset)
			seq_printf(seq, " %d offset-copies", conf->far_copies);
		else
			seq_printf(seq, " %d far-copies", conf->far_copies);
	}
L
Linus Torvalds 已提交
985
	seq_printf(seq, " [%d/%d] [", conf->raid_disks,
986
					conf->raid_disks - mddev->degraded);
L
Linus Torvalds 已提交
987 988 989
	for (i = 0; i < conf->raid_disks; i++)
		seq_printf(seq, "%s",
			      conf->mirrors[i].rdev &&
990
			      test_bit(In_sync, &conf->mirrors[i].rdev->flags) ? "U" : "_");
L
Linus Torvalds 已提交
991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004
	seq_printf(seq, "]");
}

static void error(mddev_t *mddev, mdk_rdev_t *rdev)
{
	char b[BDEVNAME_SIZE];
	conf_t *conf = mddev_to_conf(mddev);

	/*
	 * If it is not operational, then we have already marked it as dead
	 * else if it is the last working disks, ignore the error, let the
	 * next level up know.
	 * else mark the drive as failed
	 */
1005
	if (test_bit(In_sync, &rdev->flags)
1006
	    && conf->raid_disks-mddev->degraded == 1)
L
Linus Torvalds 已提交
1007 1008 1009 1010 1011 1012 1013 1014
		/*
		 * Don't fail the drive, just return an IO error.
		 * The test should really be more sophisticated than
		 * "working_disks == 1", but it isn't critical, and
		 * can wait until we do more sophisticated "is the drive
		 * really dead" tests...
		 */
		return;
1015 1016 1017
	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
1018
		mddev->degraded++;
1019
		spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1020 1021 1022 1023 1024
		/*
		 * if recovery is running, make sure it aborts.
		 */
		set_bit(MD_RECOVERY_ERR, &mddev->recovery);
	}
1025
	set_bit(Faulty, &rdev->flags);
1026
	set_bit(MD_CHANGE_DEVS, &mddev->flags);
1027 1028
	printk(KERN_ALERT "raid10: Disk failure on %s, disabling device.\n"
		"raid10: Operation continuing on %d devices.\n",
1029
		bdevname(rdev->bdev,b), conf->raid_disks - mddev->degraded);
L
Linus Torvalds 已提交
1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041
}

static void print_conf(conf_t *conf)
{
	int i;
	mirror_info_t *tmp;

	printk("RAID10 conf printout:\n");
	if (!conf) {
		printk("(!conf)\n");
		return;
	}
1042
	printk(" --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
L
Linus Torvalds 已提交
1043 1044 1045 1046 1047 1048 1049
		conf->raid_disks);

	for (i = 0; i < conf->raid_disks; i++) {
		char b[BDEVNAME_SIZE];
		tmp = conf->mirrors + i;
		if (tmp->rdev)
			printk(" disk %d, wo:%d, o:%d, dev:%s\n",
1050 1051
				i, !test_bit(In_sync, &tmp->rdev->flags),
			        !test_bit(Faulty, &tmp->rdev->flags),
L
Linus Torvalds 已提交
1052 1053 1054 1055 1056 1057
				bdevname(tmp->rdev->bdev,b));
	}
}

static void close_sync(conf_t *conf)
{
1058 1059
	wait_barrier(conf);
	allow_barrier(conf);
L
Linus Torvalds 已提交
1060 1061 1062 1063 1064

	mempool_destroy(conf->r10buf_pool);
	conf->r10buf_pool = NULL;
}

1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085
/* check if there are enough drives for
 * every block to appear on atleast one
 */
static int enough(conf_t *conf)
{
	int first = 0;

	do {
		int n = conf->copies;
		int cnt = 0;
		while (n--) {
			if (conf->mirrors[first].rdev)
				cnt++;
			first = (first+1) % conf->raid_disks;
		}
		if (cnt == 0)
			return 0;
	} while (first != 0);
	return 1;
}

L
Linus Torvalds 已提交
1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098
static int raid10_spare_active(mddev_t *mddev)
{
	int i;
	conf_t *conf = mddev->private;
	mirror_info_t *tmp;

	/*
	 * Find all non-in_sync disks within the RAID10 configuration
	 * and mark them in_sync
	 */
	for (i = 0; i < conf->raid_disks; i++) {
		tmp = conf->mirrors + i;
		if (tmp->rdev
1099
		    && !test_bit(Faulty, &tmp->rdev->flags)
1100 1101 1102
		    && !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
1103
			mddev->degraded--;
1104
			spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124
		}
	}

	print_conf(conf);
	return 0;
}


static int raid10_add_disk(mddev_t *mddev, mdk_rdev_t *rdev)
{
	conf_t *conf = mddev->private;
	int found = 0;
	int mirror;
	mirror_info_t *p;

	if (mddev->recovery_cp < MaxSector)
		/* only hot-add to in-sync arrays, as recovery is
		 * very different from resync
		 */
		return 0;
1125 1126
	if (!enough(conf))
		return 0;
L
Linus Torvalds 已提交
1127

1128 1129 1130 1131 1132 1133
	if (rdev->saved_raid_disk >= 0 &&
	    conf->mirrors[rdev->saved_raid_disk].rdev == NULL)
		mirror = rdev->saved_raid_disk;
	else
		mirror = 0;
	for ( ; mirror < mddev->raid_disks; mirror++)
L
Linus Torvalds 已提交
1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148
		if ( !(p=conf->mirrors+mirror)->rdev) {

			blk_queue_stack_limits(mddev->queue,
					       rdev->bdev->bd_disk->queue);
			/* as we don't honour merge_bvec_fn, we must never risk
			 * violating it, so limit ->max_sector to one PAGE, as
			 * a one page request is never in violation.
			 */
			if (rdev->bdev->bd_disk->queue->merge_bvec_fn &&
			    mddev->queue->max_sectors > (PAGE_SIZE>>9))
				mddev->queue->max_sectors = (PAGE_SIZE>>9);

			p->head_position = 0;
			rdev->raid_disk = mirror;
			found = 1;
1149 1150
			if (rdev->saved_raid_disk != mirror)
				conf->fullsync = 1;
1151
			rcu_assign_pointer(p->rdev, rdev);
L
Linus Torvalds 已提交
1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168
			break;
		}

	print_conf(conf);
	return found;
}

static int raid10_remove_disk(mddev_t *mddev, int number)
{
	conf_t *conf = mddev->private;
	int err = 0;
	mdk_rdev_t *rdev;
	mirror_info_t *p = conf->mirrors+ number;

	print_conf(conf);
	rdev = p->rdev;
	if (rdev) {
1169
		if (test_bit(In_sync, &rdev->flags) ||
L
Linus Torvalds 已提交
1170 1171 1172 1173 1174
		    atomic_read(&rdev->nr_pending)) {
			err = -EBUSY;
			goto abort;
		}
		p->rdev = NULL;
1175
		synchronize_rcu();
L
Linus Torvalds 已提交
1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188
		if (atomic_read(&rdev->nr_pending)) {
			/* lost the race, try later */
			err = -EBUSY;
			p->rdev = rdev;
		}
	}
abort:

	print_conf(conf);
	return err;
}


1189
static void end_sync_read(struct bio *bio, int error)
L
Linus Torvalds 已提交
1190 1191 1192 1193 1194 1195 1196 1197
{
	r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private);
	conf_t *conf = mddev_to_conf(r10_bio->mddev);
	int i,d;

	for (i=0; i<conf->copies; i++)
		if (r10_bio->devs[i].bio == bio)
			break;
1198
	BUG_ON(i == conf->copies);
L
Linus Torvalds 已提交
1199 1200
	update_head_pos(i, r10_bio);
	d = r10_bio->devs[i].devnum;
1201 1202 1203

	if (test_bit(BIO_UPTODATE, &bio->bi_flags))
		set_bit(R10BIO_Uptodate, &r10_bio->state);
1204 1205 1206 1207 1208 1209 1210
	else {
		atomic_add(r10_bio->sectors,
			   &conf->mirrors[d].rdev->corrected_errors);
		if (!test_bit(MD_RECOVERY_SYNC, &conf->mddev->recovery))
			md_error(r10_bio->mddev,
				 conf->mirrors[d].rdev);
	}
L
Linus Torvalds 已提交
1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224

	/* for reconstruct, we always reschedule after a read.
	 * for resync, only after all reads
	 */
	if (test_bit(R10BIO_IsRecover, &r10_bio->state) ||
	    atomic_dec_and_test(&r10_bio->remaining)) {
		/* we have read all the blocks,
		 * do the comparison in process context in raid10d
		 */
		reschedule_retry(r10_bio);
	}
	rdev_dec_pending(conf->mirrors[d].rdev, conf->mddev);
}

1225
static void end_sync_write(struct bio *bio, int error)
L
Linus Torvalds 已提交
1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
	r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private);
	mddev_t *mddev = r10_bio->mddev;
	conf_t *conf = mddev_to_conf(mddev);
	int i,d;

	for (i = 0; i < conf->copies; i++)
		if (r10_bio->devs[i].bio == bio)
			break;
	d = r10_bio->devs[i].devnum;

	if (!uptodate)
		md_error(mddev, conf->mirrors[d].rdev);
	update_head_pos(i, r10_bio);

	while (atomic_dec_and_test(&r10_bio->remaining)) {
		if (r10_bio->master_bio == NULL) {
			/* the primary of several recovery bios */
			md_done_sync(mddev, r10_bio->sectors, 1);
			put_buf(r10_bio);
			break;
		} else {
			r10bio_t *r10_bio2 = (r10bio_t *)r10_bio->master_bio;
			put_buf(r10_bio);
			r10_bio = r10_bio2;
		}
	}
	rdev_dec_pending(conf->mirrors[d].rdev, mddev);
}

/*
 * Note: sync and recover and handled very differently for raid10
 * This code is for resync.
 * For resync, we read through virtual addresses and read all blocks.
 * If there is any error, we schedule a write.  The lowest numbered
 * drive is authoritative.
 * However requests come for physical address, so we need to map.
 * For every physical address there are raid_disks/copies virtual addresses,
 * which is always are least one, but is not necessarly an integer.
 * This means that a physical address can span multiple chunks, so we may
 * have to submit multiple io requests for a single sync request.
 */
/*
 * We check if all blocks are in-sync and only write to blocks that
 * aren't in sync
 */
static void sync_request_write(mddev_t *mddev, r10bio_t *r10_bio)
{
	conf_t *conf = mddev_to_conf(mddev);
	int i, first;
	struct bio *tbio, *fbio;

	atomic_set(&r10_bio->remaining, 1);

	/* find the first device with a block */
	for (i=0; i<conf->copies; i++)
		if (test_bit(BIO_UPTODATE, &r10_bio->devs[i].bio->bi_flags))
			break;

	if (i == conf->copies)
		goto done;

	first = i;
	fbio = r10_bio->devs[i].bio;

	/* now find blocks with errors */
1293 1294 1295
	for (i=0 ; i < conf->copies ; i++) {
		int  j, d;
		int vcnt = r10_bio->sectors >> (PAGE_SHIFT-9);
L
Linus Torvalds 已提交
1296 1297

		tbio = r10_bio->devs[i].bio;
1298 1299 1300 1301

		if (tbio->bi_end_io != end_sync_read)
			continue;
		if (i == first)
L
Linus Torvalds 已提交
1302
			continue;
1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316
		if (test_bit(BIO_UPTODATE, &r10_bio->devs[i].bio->bi_flags)) {
			/* We know that the bi_io_vec layout is the same for
			 * both 'first' and 'i', so we just compare them.
			 * All vec entries are PAGE_SIZE;
			 */
			for (j = 0; j < vcnt; j++)
				if (memcmp(page_address(fbio->bi_io_vec[j].bv_page),
					   page_address(tbio->bi_io_vec[j].bv_page),
					   PAGE_SIZE))
					break;
			if (j == vcnt)
				continue;
			mddev->resync_mismatches += r10_bio->sectors;
		}
1317 1318 1319
		if (test_bit(MD_RECOVERY_CHECK, &mddev->recovery))
			/* Don't fix anything. */
			continue;
L
Linus Torvalds 已提交
1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396
		/* Ok, we need to write this bio
		 * First we need to fixup bv_offset, bv_len and
		 * bi_vecs, as the read request might have corrupted these
		 */
		tbio->bi_vcnt = vcnt;
		tbio->bi_size = r10_bio->sectors << 9;
		tbio->bi_idx = 0;
		tbio->bi_phys_segments = 0;
		tbio->bi_hw_segments = 0;
		tbio->bi_hw_front_size = 0;
		tbio->bi_hw_back_size = 0;
		tbio->bi_flags &= ~(BIO_POOL_MASK - 1);
		tbio->bi_flags |= 1 << BIO_UPTODATE;
		tbio->bi_next = NULL;
		tbio->bi_rw = WRITE;
		tbio->bi_private = r10_bio;
		tbio->bi_sector = r10_bio->devs[i].addr;

		for (j=0; j < vcnt ; j++) {
			tbio->bi_io_vec[j].bv_offset = 0;
			tbio->bi_io_vec[j].bv_len = PAGE_SIZE;

			memcpy(page_address(tbio->bi_io_vec[j].bv_page),
			       page_address(fbio->bi_io_vec[j].bv_page),
			       PAGE_SIZE);
		}
		tbio->bi_end_io = end_sync_write;

		d = r10_bio->devs[i].devnum;
		atomic_inc(&conf->mirrors[d].rdev->nr_pending);
		atomic_inc(&r10_bio->remaining);
		md_sync_acct(conf->mirrors[d].rdev->bdev, tbio->bi_size >> 9);

		tbio->bi_sector += conf->mirrors[d].rdev->data_offset;
		tbio->bi_bdev = conf->mirrors[d].rdev->bdev;
		generic_make_request(tbio);
	}

done:
	if (atomic_dec_and_test(&r10_bio->remaining)) {
		md_done_sync(mddev, r10_bio->sectors, 1);
		put_buf(r10_bio);
	}
}

/*
 * Now for the recovery code.
 * Recovery happens across physical sectors.
 * We recover all non-is_sync drives by finding the virtual address of
 * each, and then choose a working drive that also has that virt address.
 * There is a separate r10_bio for each non-in_sync drive.
 * Only the first two slots are in use. The first for reading,
 * The second for writing.
 *
 */

static void recovery_request_write(mddev_t *mddev, r10bio_t *r10_bio)
{
	conf_t *conf = mddev_to_conf(mddev);
	int i, d;
	struct bio *bio, *wbio;


	/* move the pages across to the second bio
	 * and submit the write request
	 */
	bio = r10_bio->devs[0].bio;
	wbio = r10_bio->devs[1].bio;
	for (i=0; i < wbio->bi_vcnt; i++) {
		struct page *p = bio->bi_io_vec[i].bv_page;
		bio->bi_io_vec[i].bv_page = wbio->bi_io_vec[i].bv_page;
		wbio->bi_io_vec[i].bv_page = p;
	}
	d = r10_bio->devs[1].devnum;

	atomic_inc(&conf->mirrors[d].rdev->nr_pending);
	md_sync_acct(conf->mirrors[d].rdev->bdev, wbio->bi_size >> 9);
1397 1398 1399
	if (test_bit(R10BIO_Uptodate, &r10_bio->state))
		generic_make_request(wbio);
	else
1400
		bio_endio(wbio, -EIO);
L
Linus Torvalds 已提交
1401 1402 1403 1404 1405 1406 1407 1408
}


/*
 * This is a kernel thread which:
 *
 *	1.	Retries failed read operations on working mirrors.
 *	2.	Updates the raid superblock when problems encounter.
1409
 *	3.	Performs writes following reads for array synchronising.
L
Linus Torvalds 已提交
1410 1411
 */

1412 1413 1414 1415 1416 1417 1418 1419 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 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506
static void fix_read_error(conf_t *conf, mddev_t *mddev, r10bio_t *r10_bio)
{
	int sect = 0; /* Offset from r10_bio->sector */
	int sectors = r10_bio->sectors;
	mdk_rdev_t*rdev;
	while(sectors) {
		int s = sectors;
		int sl = r10_bio->read_slot;
		int success = 0;
		int start;

		if (s > (PAGE_SIZE>>9))
			s = PAGE_SIZE >> 9;

		rcu_read_lock();
		do {
			int d = r10_bio->devs[sl].devnum;
			rdev = rcu_dereference(conf->mirrors[d].rdev);
			if (rdev &&
			    test_bit(In_sync, &rdev->flags)) {
				atomic_inc(&rdev->nr_pending);
				rcu_read_unlock();
				success = sync_page_io(rdev->bdev,
						       r10_bio->devs[sl].addr +
						       sect + rdev->data_offset,
						       s<<9,
						       conf->tmppage, READ);
				rdev_dec_pending(rdev, mddev);
				rcu_read_lock();
				if (success)
					break;
			}
			sl++;
			if (sl == conf->copies)
				sl = 0;
		} while (!success && sl != r10_bio->read_slot);
		rcu_read_unlock();

		if (!success) {
			/* Cannot read from anywhere -- bye bye array */
			int dn = r10_bio->devs[r10_bio->read_slot].devnum;
			md_error(mddev, conf->mirrors[dn].rdev);
			break;
		}

		start = sl;
		/* write it back and re-read */
		rcu_read_lock();
		while (sl != r10_bio->read_slot) {
			int d;
			if (sl==0)
				sl = conf->copies;
			sl--;
			d = r10_bio->devs[sl].devnum;
			rdev = rcu_dereference(conf->mirrors[d].rdev);
			if (rdev &&
			    test_bit(In_sync, &rdev->flags)) {
				atomic_inc(&rdev->nr_pending);
				rcu_read_unlock();
				atomic_add(s, &rdev->corrected_errors);
				if (sync_page_io(rdev->bdev,
						 r10_bio->devs[sl].addr +
						 sect + rdev->data_offset,
						 s<<9, conf->tmppage, WRITE)
				    == 0)
					/* Well, this device is dead */
					md_error(mddev, rdev);
				rdev_dec_pending(rdev, mddev);
				rcu_read_lock();
			}
		}
		sl = start;
		while (sl != r10_bio->read_slot) {
			int d;
			if (sl==0)
				sl = conf->copies;
			sl--;
			d = r10_bio->devs[sl].devnum;
			rdev = rcu_dereference(conf->mirrors[d].rdev);
			if (rdev &&
			    test_bit(In_sync, &rdev->flags)) {
				char b[BDEVNAME_SIZE];
				atomic_inc(&rdev->nr_pending);
				rcu_read_unlock();
				if (sync_page_io(rdev->bdev,
						 r10_bio->devs[sl].addr +
						 sect + rdev->data_offset,
						 s<<9, conf->tmppage, READ) == 0)
					/* Well, this device is dead */
					md_error(mddev, rdev);
				else
					printk(KERN_INFO
					       "raid10:%s: read error corrected"
					       " (%d sectors at %llu on %s)\n",
					       mdname(mddev), s,
1507 1508
					       (unsigned long long)(sect+
					            rdev->data_offset),
1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521
					       bdevname(rdev->bdev, b));

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

		sectors -= s;
		sect += s;
	}
}

L
Linus Torvalds 已提交
1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535
static void raid10d(mddev_t *mddev)
{
	r10bio_t *r10_bio;
	struct bio *bio;
	unsigned long flags;
	conf_t *conf = mddev_to_conf(mddev);
	struct list_head *head = &conf->retry_list;
	int unplug=0;
	mdk_rdev_t *rdev;

	md_check_recovery(mddev);

	for (;;) {
		char b[BDEVNAME_SIZE];
1536

1537
		unplug += flush_pending_writes(conf);
1538

1539 1540 1541
		spin_lock_irqsave(&conf->device_lock, flags);
		if (list_empty(head)) {
			spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1542
			break;
1543
		}
L
Linus Torvalds 已提交
1544 1545
		r10_bio = list_entry(head->prev, r10bio_t, retry_list);
		list_del(head->prev);
1546
		conf->nr_queued--;
L
Linus Torvalds 已提交
1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558
		spin_unlock_irqrestore(&conf->device_lock, flags);

		mddev = r10_bio->mddev;
		conf = mddev_to_conf(mddev);
		if (test_bit(R10BIO_IsSync, &r10_bio->state)) {
			sync_request_write(mddev, r10_bio);
			unplug = 1;
		} else 	if (test_bit(R10BIO_IsRecover, &r10_bio->state)) {
			recovery_request_write(mddev, r10_bio);
			unplug = 1;
		} else {
			int mirror;
1559 1560 1561 1562 1563 1564 1565 1566
			/* we got a read error. Maybe the drive is bad.  Maybe just
			 * the block and we can fix it.
			 * We freeze all other IO, and try reading the block from
			 * other devices.  When we find one, we re-write
			 * and check it that fixes the read error.
			 * This is all done synchronously while the array is
			 * frozen.
			 */
1567 1568 1569 1570
			if (mddev->ro == 0) {
				freeze_array(conf);
				fix_read_error(conf, mddev, r10_bio);
				unfreeze_array(conf);
1571 1572
			}

L
Linus Torvalds 已提交
1573
			bio = r10_bio->devs[r10_bio->read_slot].bio;
1574 1575
			r10_bio->devs[r10_bio->read_slot].bio =
				mddev->ro ? IO_BLOCKED : NULL;
L
Linus Torvalds 已提交
1576 1577 1578 1579 1580 1581 1582
			mirror = read_balance(conf, r10_bio);
			if (mirror == -1) {
				printk(KERN_ALERT "raid10: %s: unrecoverable I/O"
				       " read error for block %llu\n",
				       bdevname(bio->bi_bdev,b),
				       (unsigned long long)r10_bio->sector);
				raid_end_bio_io(r10_bio);
1583
				bio_put(bio);
L
Linus Torvalds 已提交
1584
			} else {
1585
				const int do_sync = bio_sync(r10_bio->master_bio);
1586
				bio_put(bio);
L
Linus Torvalds 已提交
1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597
				rdev = conf->mirrors[mirror].rdev;
				if (printk_ratelimit())
					printk(KERN_ERR "raid10: %s: redirecting sector %llu to"
					       " another mirror\n",
					       bdevname(rdev->bdev,b),
					       (unsigned long long)r10_bio->sector);
				bio = bio_clone(r10_bio->master_bio, GFP_NOIO);
				r10_bio->devs[r10_bio->read_slot].bio = bio;
				bio->bi_sector = r10_bio->devs[r10_bio->read_slot].addr
					+ rdev->data_offset;
				bio->bi_bdev = rdev->bdev;
1598
				bio->bi_rw = READ | do_sync;
L
Linus Torvalds 已提交
1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615
				bio->bi_private = r10_bio;
				bio->bi_end_io = raid10_end_read_request;
				unplug = 1;
				generic_make_request(bio);
			}
		}
	}
	if (unplug)
		unplug_slaves(mddev);
}


static int init_resync(conf_t *conf)
{
	int buffs;

	buffs = RESYNC_WINDOW / RESYNC_BLOCK_SIZE;
1616
	BUG_ON(conf->r10buf_pool);
L
Linus Torvalds 已提交
1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655
	conf->r10buf_pool = mempool_create(buffs, r10buf_pool_alloc, r10buf_pool_free, conf);
	if (!conf->r10buf_pool)
		return -ENOMEM;
	conf->next_resync = 0;
	return 0;
}

/*
 * perform a "sync" on one "block"
 *
 * We need to make sure that no normal I/O request - particularly write
 * requests - conflict with active sync requests.
 *
 * This is achieved by tracking pending requests and a 'barrier' concept
 * that can be installed to exclude normal IO requests.
 *
 * Resync and recovery are handled very differently.
 * We differentiate by looking at MD_RECOVERY_SYNC in mddev->recovery.
 *
 * For resync, we iterate over virtual addresses, read all copies,
 * and update if there are differences.  If only one copy is live,
 * skip it.
 * For recovery, we iterate over physical addresses, read a good
 * value for each non-in_sync drive, and over-write.
 *
 * So, for recovery we may have several outstanding complex requests for a
 * given address, one for each out-of-sync device.  We model this by allocating
 * a number of r10_bio structures, one for each out-of-sync device.
 * As we setup these structures, we collect all bio's together into a list
 * which we then process collectively to add pages, and then process again
 * to pass to generic_make_request.
 *
 * The r10_bio structures are linked using a borrowed master_bio pointer.
 * This link is counted in ->remaining.  When the r10_bio that points to NULL
 * has its remaining count decremented to 0, the whole complex operation
 * is complete.
 *
 */

1656
static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster)
L
Linus Torvalds 已提交
1657 1658 1659 1660 1661 1662 1663
{
	conf_t *conf = mddev_to_conf(mddev);
	r10bio_t *r10_bio;
	struct bio *biolist = NULL, *bio;
	sector_t max_sector, nr_sectors;
	int disk;
	int i;
1664 1665
	int max_sync;
	int sync_blocks;
L
Linus Torvalds 已提交
1666 1667 1668 1669 1670 1671

	sector_t sectors_skipped = 0;
	int chunks_skipped = 0;

	if (!conf->r10buf_pool)
		if (init_resync(conf))
1672
			return 0;
L
Linus Torvalds 已提交
1673 1674 1675 1676 1677 1678

 skipped:
	max_sector = mddev->size << 1;
	if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery))
		max_sector = mddev->resync_max_sectors;
	if (sector_nr >= max_sector) {
1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701
		/* If we aborted, we need to abort the
		 * sync on the 'current' bitmap chucks (there can
		 * be several when recovering multiple devices).
		 * as we may have started syncing it but not finished.
		 * We can find the current address in
		 * mddev->curr_resync, but for recovery,
		 * we need to convert that to several
		 * virtual addresses.
		 */
		if (mddev->curr_resync < max_sector) { /* aborted */
			if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery))
				bitmap_end_sync(mddev->bitmap, mddev->curr_resync,
						&sync_blocks, 1);
			else for (i=0; i<conf->raid_disks; i++) {
				sector_t sect =
					raid10_find_virt(conf, mddev->curr_resync, i);
				bitmap_end_sync(mddev->bitmap, sect,
						&sync_blocks, 1);
			}
		} else /* completed sync */
			conf->fullsync = 0;

		bitmap_close_sync(mddev->bitmap);
L
Linus Torvalds 已提交
1702
		close_sync(conf);
1703
		*skipped = 1;
L
Linus Torvalds 已提交
1704 1705 1706 1707 1708 1709
		return sectors_skipped;
	}
	if (chunks_skipped >= conf->raid_disks) {
		/* if there has been nothing to do on any drive,
		 * then there is nothing to do at all..
		 */
1710 1711
		*skipped = 1;
		return (max_sector - sector_nr) + sectors_skipped;
L
Linus Torvalds 已提交
1712 1713
	}

1714 1715 1716
	if (max_sector > mddev->resync_max)
		max_sector = mddev->resync_max; /* Don't do IO beyond here */

L
Linus Torvalds 已提交
1717 1718 1719 1720 1721 1722 1723 1724 1725 1726
	/* make sure whole request will fit in a chunk - if chunks
	 * are meaningful
	 */
	if (conf->near_copies < conf->raid_disks &&
	    max_sector > (sector_nr | conf->chunk_mask))
		max_sector = (sector_nr | conf->chunk_mask) + 1;
	/*
	 * If there is non-resync activity waiting for us then
	 * put in a delay to throttle resync.
	 */
1727
	if (!go_faster && conf->nr_waiting)
L
Linus Torvalds 已提交
1728 1729
		msleep_interruptible(1000);

N
NeilBrown 已提交
1730 1731
	bitmap_cond_end_sync(mddev->bitmap, sector_nr);

L
Linus Torvalds 已提交
1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746
	/* Again, very different code for resync and recovery.
	 * Both must result in an r10bio with a list of bios that
	 * have bi_end_io, bi_sector, bi_bdev set,
	 * and bi_private set to the r10bio.
	 * For recovery, we may actually create several r10bios
	 * with 2 bios in each, that correspond to the bios in the main one.
	 * In this case, the subordinate r10bios link back through a
	 * borrowed master_bio pointer, and the counter in the master
	 * includes a ref from each subordinate.
	 */
	/* First, we decide what to do and set ->bi_end_io
	 * To end_sync_read if we want to read, and
	 * end_sync_write if we will want to write.
	 */

1747
	max_sync = RESYNC_PAGES << (PAGE_SHIFT-9);
L
Linus Torvalds 已提交
1748 1749 1750 1751 1752 1753 1754
	if (!test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
		/* recovery... the complicated one */
		int i, j, k;
		r10_bio = NULL;

		for (i=0 ; i<conf->raid_disks; i++)
			if (conf->mirrors[i].rdev &&
1755
			    !test_bit(In_sync, &conf->mirrors[i].rdev->flags)) {
1756
				int still_degraded = 0;
L
Linus Torvalds 已提交
1757 1758
				/* want to reconstruct this device */
				r10bio_t *rb2 = r10_bio;
1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775
				sector_t sect = raid10_find_virt(conf, sector_nr, i);
				int must_sync;
				/* Unless we are doing a full sync, we only need
				 * to recover the block if it is set in the bitmap
				 */
				must_sync = bitmap_start_sync(mddev->bitmap, sect,
							      &sync_blocks, 1);
				if (sync_blocks < max_sync)
					max_sync = sync_blocks;
				if (!must_sync &&
				    !conf->fullsync) {
					/* yep, skip the sync_blocks here, but don't assume
					 * that there will never be anything to do here
					 */
					chunks_skipped = -1;
					continue;
				}
L
Linus Torvalds 已提交
1776 1777

				r10_bio = mempool_alloc(conf->r10buf_pool, GFP_NOIO);
1778
				raise_barrier(conf, rb2 != NULL);
L
Linus Torvalds 已提交
1779 1780 1781 1782 1783 1784 1785
				atomic_set(&r10_bio->remaining, 0);

				r10_bio->master_bio = (struct bio*)rb2;
				if (rb2)
					atomic_inc(&rb2->remaining);
				r10_bio->mddev = mddev;
				set_bit(R10BIO_IsRecover, &r10_bio->state);
1786 1787
				r10_bio->sector = sect;

L
Linus Torvalds 已提交
1788
				raid10_find_phys(conf, r10_bio);
1789 1790 1791 1792 1793 1794
				/* Need to check if this section will still be
				 * degraded
				 */
				for (j=0; j<conf->copies;j++) {
					int d = r10_bio->devs[j].devnum;
					if (conf->mirrors[d].rdev == NULL ||
1795
					    test_bit(Faulty, &conf->mirrors[d].rdev->flags)) {
1796
						still_degraded = 1;
1797 1798
						break;
					}
1799 1800 1801 1802
				}
				must_sync = bitmap_start_sync(mddev->bitmap, sect,
							      &sync_blocks, still_degraded);

L
Linus Torvalds 已提交
1803 1804 1805
				for (j=0; j<conf->copies;j++) {
					int d = r10_bio->devs[j].devnum;
					if (conf->mirrors[d].rdev &&
1806
					    test_bit(In_sync, &conf->mirrors[d].rdev->flags)) {
L
Linus Torvalds 已提交
1807 1808 1809 1810 1811 1812
						/* This is where we read from */
						bio = r10_bio->devs[0].bio;
						bio->bi_next = biolist;
						biolist = bio;
						bio->bi_private = r10_bio;
						bio->bi_end_io = end_sync_read;
1813
						bio->bi_rw = READ;
L
Linus Torvalds 已提交
1814 1815 1816 1817 1818 1819 1820 1821 1822 1823
						bio->bi_sector = r10_bio->devs[j].addr +
							conf->mirrors[d].rdev->data_offset;
						bio->bi_bdev = conf->mirrors[d].rdev->bdev;
						atomic_inc(&conf->mirrors[d].rdev->nr_pending);
						atomic_inc(&r10_bio->remaining);
						/* and we write to 'i' */

						for (k=0; k<conf->copies; k++)
							if (r10_bio->devs[k].devnum == i)
								break;
1824
						BUG_ON(k == conf->copies);
L
Linus Torvalds 已提交
1825 1826 1827 1828 1829
						bio = r10_bio->devs[1].bio;
						bio->bi_next = biolist;
						biolist = bio;
						bio->bi_private = r10_bio;
						bio->bi_end_io = end_sync_write;
1830
						bio->bi_rw = WRITE;
L
Linus Torvalds 已提交
1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841
						bio->bi_sector = r10_bio->devs[k].addr +
							conf->mirrors[i].rdev->data_offset;
						bio->bi_bdev = conf->mirrors[i].rdev->bdev;

						r10_bio->devs[0].devnum = d;
						r10_bio->devs[1].devnum = i;

						break;
					}
				}
				if (j == conf->copies) {
1842 1843
					/* Cannot recover, so abort the recovery */
					put_buf(r10_bio);
1844 1845
					if (rb2)
						atomic_dec(&rb2->remaining);
1846 1847 1848 1849 1850
					r10_bio = rb2;
					if (!test_and_set_bit(MD_RECOVERY_ERR, &mddev->recovery))
						printk(KERN_INFO "raid10: %s: insufficient working devices for recovery.\n",
						       mdname(mddev));
					break;
L
Linus Torvalds 已提交
1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864
				}
			}
		if (biolist == NULL) {
			while (r10_bio) {
				r10bio_t *rb2 = r10_bio;
				r10_bio = (r10bio_t*) rb2->master_bio;
				rb2->master_bio = NULL;
				put_buf(rb2);
			}
			goto giveup;
		}
	} else {
		/* resync. Schedule a read for every block at this virt offset */
		int count = 0;
1865 1866 1867 1868 1869 1870 1871 1872 1873 1874

		if (!bitmap_start_sync(mddev->bitmap, sector_nr,
				       &sync_blocks, mddev->degraded) &&
		    !conf->fullsync && !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
			/* We can skip this block */
			*skipped = 1;
			return sync_blocks + sectors_skipped;
		}
		if (sync_blocks < max_sync)
			max_sync = sync_blocks;
L
Linus Torvalds 已提交
1875 1876 1877 1878
		r10_bio = mempool_alloc(conf->r10buf_pool, GFP_NOIO);

		r10_bio->mddev = mddev;
		atomic_set(&r10_bio->remaining, 0);
1879 1880
		raise_barrier(conf, 0);
		conf->next_resync = sector_nr;
L
Linus Torvalds 已提交
1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891

		r10_bio->master_bio = NULL;
		r10_bio->sector = sector_nr;
		set_bit(R10BIO_IsSync, &r10_bio->state);
		raid10_find_phys(conf, r10_bio);
		r10_bio->sectors = (sector_nr | conf->chunk_mask) - sector_nr +1;

		for (i=0; i<conf->copies; i++) {
			int d = r10_bio->devs[i].devnum;
			bio = r10_bio->devs[i].bio;
			bio->bi_end_io = NULL;
N
NeilBrown 已提交
1892
			clear_bit(BIO_UPTODATE, &bio->bi_flags);
L
Linus Torvalds 已提交
1893
			if (conf->mirrors[d].rdev == NULL ||
1894
			    test_bit(Faulty, &conf->mirrors[d].rdev->flags))
L
Linus Torvalds 已提交
1895 1896 1897 1898 1899 1900 1901
				continue;
			atomic_inc(&conf->mirrors[d].rdev->nr_pending);
			atomic_inc(&r10_bio->remaining);
			bio->bi_next = biolist;
			biolist = bio;
			bio->bi_private = r10_bio;
			bio->bi_end_io = end_sync_read;
1902
			bio->bi_rw = READ;
L
Linus Torvalds 已提交
1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933
			bio->bi_sector = r10_bio->devs[i].addr +
				conf->mirrors[d].rdev->data_offset;
			bio->bi_bdev = conf->mirrors[d].rdev->bdev;
			count++;
		}

		if (count < 2) {
			for (i=0; i<conf->copies; i++) {
				int d = r10_bio->devs[i].devnum;
				if (r10_bio->devs[i].bio->bi_end_io)
					rdev_dec_pending(conf->mirrors[d].rdev, mddev);
			}
			put_buf(r10_bio);
			biolist = NULL;
			goto giveup;
		}
	}

	for (bio = biolist; bio ; bio=bio->bi_next) {

		bio->bi_flags &= ~(BIO_POOL_MASK - 1);
		if (bio->bi_end_io)
			bio->bi_flags |= 1 << BIO_UPTODATE;
		bio->bi_vcnt = 0;
		bio->bi_idx = 0;
		bio->bi_phys_segments = 0;
		bio->bi_hw_segments = 0;
		bio->bi_size = 0;
	}

	nr_sectors = 0;
1934 1935
	if (sector_nr + max_sync < max_sector)
		max_sector = sector_nr + max_sync;
L
Linus Torvalds 已提交
1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979
	do {
		struct page *page;
		int len = PAGE_SIZE;
		disk = 0;
		if (sector_nr + (len>>9) > max_sector)
			len = (max_sector - sector_nr) << 9;
		if (len == 0)
			break;
		for (bio= biolist ; bio ; bio=bio->bi_next) {
			page = bio->bi_io_vec[bio->bi_vcnt].bv_page;
			if (bio_add_page(bio, page, len, 0) == 0) {
				/* stop here */
				struct bio *bio2;
				bio->bi_io_vec[bio->bi_vcnt].bv_page = page;
				for (bio2 = biolist; bio2 && bio2 != bio; bio2 = bio2->bi_next) {
					/* remove last page from this bio */
					bio2->bi_vcnt--;
					bio2->bi_size -= len;
					bio2->bi_flags &= ~(1<< BIO_SEG_VALID);
				}
				goto bio_full;
			}
			disk = i;
		}
		nr_sectors += len>>9;
		sector_nr += len>>9;
	} while (biolist->bi_vcnt < RESYNC_PAGES);
 bio_full:
	r10_bio->sectors = nr_sectors;

	while (biolist) {
		bio = biolist;
		biolist = biolist->bi_next;

		bio->bi_next = NULL;
		r10_bio = bio->bi_private;
		r10_bio->sectors = nr_sectors;

		if (bio->bi_end_io == end_sync_read) {
			md_sync_acct(bio->bi_bdev, nr_sectors);
			generic_make_request(bio);
		}
	}

1980 1981 1982 1983 1984 1985
	if (sectors_skipped)
		/* pretend they weren't skipped, it makes
		 * no important difference in this case
		 */
		md_done_sync(mddev, sectors_skipped, 1);

L
Linus Torvalds 已提交
1986 1987 1988 1989 1990 1991
	return sectors_skipped + nr_sectors;
 giveup:
	/* There is nowhere to write, so all non-sync
	 * drives must be failed, so try the next chunk...
	 */
	{
1992
	sector_t sec = max_sector - sector_nr;
L
Linus Torvalds 已提交
1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
	sectors_skipped += sec;
	chunks_skipped ++;
	sector_nr = max_sector;
	goto skipped;
	}
}

static int run(mddev_t *mddev)
{
	conf_t *conf;
	int i, disk_idx;
	mirror_info_t *disk;
	mdk_rdev_t *rdev;
	struct list_head *tmp;
2007
	int nc, fc, fo;
L
Linus Torvalds 已提交
2008 2009
	sector_t stride, size;

2010 2011 2012
	if (mddev->chunk_size == 0) {
		printk(KERN_ERR "md/raid10: non-zero chunk size required.\n");
		return -EINVAL;
L
Linus Torvalds 已提交
2013
	}
2014

L
Linus Torvalds 已提交
2015 2016
	nc = mddev->layout & 255;
	fc = (mddev->layout >> 8) & 255;
2017
	fo = mddev->layout & (1<<16);
L
Linus Torvalds 已提交
2018
	if ((nc*fc) <2 || (nc*fc) > mddev->raid_disks ||
2019
	    (mddev->layout >> 17)) {
L
Linus Torvalds 已提交
2020 2021 2022 2023 2024 2025 2026 2027 2028
		printk(KERN_ERR "raid10: %s: unsupported raid10 layout: 0x%8x\n",
		       mdname(mddev), mddev->layout);
		goto out;
	}
	/*
	 * copy the already verified devices into our private RAID10
	 * bookkeeping area. [whatever we allocate in run(),
	 * should be freed in stop()]
	 */
2029
	conf = kzalloc(sizeof(conf_t), GFP_KERNEL);
L
Linus Torvalds 已提交
2030 2031 2032 2033 2034 2035
	mddev->private = conf;
	if (!conf) {
		printk(KERN_ERR "raid10: couldn't allocate memory for %s\n",
			mdname(mddev));
		goto out;
	}
2036
	conf->mirrors = kzalloc(sizeof(struct mirror_info)*mddev->raid_disks,
L
Linus Torvalds 已提交
2037 2038 2039 2040 2041 2042
				 GFP_KERNEL);
	if (!conf->mirrors) {
		printk(KERN_ERR "raid10: couldn't allocate memory for %s\n",
		       mdname(mddev));
		goto out_free_conf;
	}
2043 2044 2045 2046

	conf->tmppage = alloc_page(GFP_KERNEL);
	if (!conf->tmppage)
		goto out_free_conf;
L
Linus Torvalds 已提交
2047

2048 2049
	conf->mddev = mddev;
	conf->raid_disks = mddev->raid_disks;
L
Linus Torvalds 已提交
2050 2051 2052
	conf->near_copies = nc;
	conf->far_copies = fc;
	conf->copies = nc*fc;
2053
	conf->far_offset = fo;
L
Linus Torvalds 已提交
2054 2055
	conf->chunk_mask = (sector_t)(mddev->chunk_size>>9)-1;
	conf->chunk_shift = ffz(~mddev->chunk_size) - 9;
2056 2057 2058 2059 2060 2061 2062
	size = mddev->size >> (conf->chunk_shift-1);
	sector_div(size, fc);
	size = size * conf->raid_disks;
	sector_div(size, nc);
	/* 'size' is now the number of chunks in the array */
	/* calculate "used chunks per device" in 'stride' */
	stride = size * conf->copies;
N
NeilBrown 已提交
2063 2064 2065 2066 2067

	/* We need to round up when dividing by raid_disks to
	 * get the stride size.
	 */
	stride += conf->raid_disks - 1;
2068 2069 2070
	sector_div(stride, conf->raid_disks);
	mddev->size = stride  << (conf->chunk_shift-1);

2071
	if (fo)
2072 2073
		stride = 1;
	else
2074
		sector_div(stride, fc);
2075 2076
	conf->stride = stride << conf->chunk_shift;

L
Linus Torvalds 已提交
2077 2078 2079 2080 2081 2082 2083 2084
	conf->r10bio_pool = mempool_create(NR_RAID10_BIOS, r10bio_pool_alloc,
						r10bio_pool_free, conf);
	if (!conf->r10bio_pool) {
		printk(KERN_ERR "raid10: couldn't allocate memory for %s\n",
			mdname(mddev));
		goto out_free_conf;
	}

2085
	rdev_for_each(rdev, tmp, mddev) {
L
Linus Torvalds 已提交
2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109
		disk_idx = rdev->raid_disk;
		if (disk_idx >= mddev->raid_disks
		    || disk_idx < 0)
			continue;
		disk = conf->mirrors + disk_idx;

		disk->rdev = rdev;

		blk_queue_stack_limits(mddev->queue,
				       rdev->bdev->bd_disk->queue);
		/* as we don't honour merge_bvec_fn, we must never risk
		 * violating it, so limit ->max_sector to one PAGE, as
		 * a one page request is never in violation.
		 */
		if (rdev->bdev->bd_disk->queue->merge_bvec_fn &&
		    mddev->queue->max_sectors > (PAGE_SIZE>>9))
			mddev->queue->max_sectors = (PAGE_SIZE>>9);

		disk->head_position = 0;
	}
	spin_lock_init(&conf->device_lock);
	INIT_LIST_HEAD(&conf->retry_list);

	spin_lock_init(&conf->resync_lock);
2110
	init_waitqueue_head(&conf->wait_barrier);
L
Linus Torvalds 已提交
2111

2112 2113 2114 2115
	/* need to check that every block has at least one working mirror */
	if (!enough(conf)) {
		printk(KERN_ERR "raid10: not enough operational mirrors for %s\n",
		       mdname(mddev));
L
Linus Torvalds 已提交
2116 2117 2118 2119 2120 2121 2122 2123
		goto out_free_conf;
	}

	mddev->degraded = 0;
	for (i = 0; i < conf->raid_disks; i++) {

		disk = conf->mirrors + i;

2124
		if (!disk->rdev ||
2125
		    !test_bit(In_sync, &disk->rdev->flags)) {
L
Linus Torvalds 已提交
2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146
			disk->head_position = 0;
			mddev->degraded++;
		}
	}


	mddev->thread = md_register_thread(raid10d, mddev, "%s_raid10");
	if (!mddev->thread) {
		printk(KERN_ERR
		       "raid10: couldn't allocate thread for %s\n",
		       mdname(mddev));
		goto out_free_conf;
	}

	printk(KERN_INFO
		"raid10: raid set %s active with %d out of %d devices\n",
		mdname(mddev), mddev->raid_disks - mddev->degraded,
		mddev->raid_disks);
	/*
	 * Ok, everything is just fine now
	 */
2147 2148
	mddev->array_size = size << (conf->chunk_shift-1);
	mddev->resync_max_sectors = size << conf->chunk_shift;
L
Linus Torvalds 已提交
2149

2150
	mddev->queue->unplug_fn = raid10_unplug;
2151 2152
	mddev->queue->backing_dev_info.congested_fn = raid10_congested;
	mddev->queue->backing_dev_info.congested_data = mddev;
2153

L
Linus Torvalds 已提交
2154 2155 2156 2157 2158
	/* Calculate max read-ahead size.
	 * We need to readahead at least twice a whole stripe....
	 * maybe...
	 */
	{
2159
		int stripe = conf->raid_disks * (mddev->chunk_size / PAGE_SIZE);
L
Linus Torvalds 已提交
2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171
		stripe /= conf->near_copies;
		if (mddev->queue->backing_dev_info.ra_pages < 2* stripe)
			mddev->queue->backing_dev_info.ra_pages = 2* stripe;
	}

	if (conf->near_copies < mddev->raid_disks)
		blk_queue_merge_bvec(mddev->queue, raid10_mergeable_bvec);
	return 0;

out_free_conf:
	if (conf->r10bio_pool)
		mempool_destroy(conf->r10bio_pool);
2172
	safe_put_page(conf->tmppage);
2173
	kfree(conf->mirrors);
L
Linus Torvalds 已提交
2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188
	kfree(conf);
	mddev->private = NULL;
out:
	return -EIO;
}

static int stop(mddev_t *mddev)
{
	conf_t *conf = mddev_to_conf(mddev);

	md_unregister_thread(mddev->thread);
	mddev->thread = NULL;
	blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
	if (conf->r10bio_pool)
		mempool_destroy(conf->r10bio_pool);
2189
	kfree(conf->mirrors);
L
Linus Torvalds 已提交
2190 2191 2192 2193 2194
	kfree(conf);
	mddev->private = NULL;
	return 0;
}

2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214
static void raid10_quiesce(mddev_t *mddev, int state)
{
	conf_t *conf = mddev_to_conf(mddev);

	switch(state) {
	case 1:
		raise_barrier(conf, 0);
		break;
	case 0:
		lower_barrier(conf);
		break;
	}
	if (mddev->thread) {
		if (mddev->bitmap)
			mddev->thread->timeout = mddev->bitmap->daemon_sleep * HZ;
		else
			mddev->thread->timeout = MAX_SCHEDULE_TIMEOUT;
		md_wakeup_thread(mddev->thread);
	}
}
L
Linus Torvalds 已提交
2215

2216
static struct mdk_personality raid10_personality =
L
Linus Torvalds 已提交
2217 2218
{
	.name		= "raid10",
2219
	.level		= 10,
L
Linus Torvalds 已提交
2220 2221 2222 2223 2224 2225 2226 2227 2228 2229
	.owner		= THIS_MODULE,
	.make_request	= make_request,
	.run		= run,
	.stop		= stop,
	.status		= status,
	.error_handler	= error,
	.hot_add_disk	= raid10_add_disk,
	.hot_remove_disk= raid10_remove_disk,
	.spare_active	= raid10_spare_active,
	.sync_request	= sync_request,
2230
	.quiesce	= raid10_quiesce,
L
Linus Torvalds 已提交
2231 2232 2233 2234
};

static int __init raid_init(void)
{
2235
	return register_md_personality(&raid10_personality);
L
Linus Torvalds 已提交
2236 2237 2238 2239
}

static void raid_exit(void)
{
2240
	unregister_md_personality(&raid10_personality);
L
Linus Torvalds 已提交
2241 2242 2243 2244 2245 2246
}

module_init(raid_init);
module_exit(raid_exit);
MODULE_LICENSE("GPL");
MODULE_ALIAS("md-personality-9"); /* RAID10 */
2247
MODULE_ALIAS("md-raid10");
2248
MODULE_ALIAS("md-level-10");