raid10.c 60.7 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
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"
Linus Torvalds's avatar
Linus Torvalds committed
22
#include <linux/raid/raid10.h>
23
#include <linux/raid/bitmap.h>
Linus Torvalds's avatar
Linus Torvalds committed
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 )
Linus Torvalds's avatar
Linus Torvalds committed
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
Linus Torvalds's avatar
Linus Torvalds committed
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.
Linus Torvalds's avatar
Linus Torvalds committed
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);

60
static void * r10bio_pool_alloc(gfp_t gfp_flags, void *data)
Linus Torvalds's avatar
Linus Torvalds committed
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)
Linus Torvalds's avatar
Linus Torvalds committed
69 70 71 72 73 74 75 76 77 78
		unplug_slaves(conf->mddev);

	return r10_bio;
}

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

79
/* Maximum size of each resync request */
Linus Torvalds's avatar
Linus Torvalds committed
80 81
#define RESYNC_BLOCK_SIZE (64*1024)
#define RESYNC_PAGES ((RESYNC_BLOCK_SIZE + PAGE_SIZE-1) / PAGE_SIZE)
82 83 84 85
/* amount of memory to reserve for resync requests */
#define RESYNC_WINDOW (1024*1024)
/* maximum number of concurrent requests, memory permitting */
#define RESYNC_DEPTH (32*1024*1024/RESYNC_BLOCK_SIZE)
Linus Torvalds's avatar
Linus Torvalds committed
86 87 88 89 90 91 92 93

/*
 * 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)
 *
 */
94
static void * r10buf_pool_alloc(gfp_t gfp_flags, void *data)
Linus Torvalds's avatar
Linus Torvalds committed
95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141
{
	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--)
142
		safe_put_page(bio->bi_io_vec[i-1].bv_page);
Linus Torvalds's avatar
Linus Torvalds committed
143 144
	while (j--)
		for (i = 0; i < RESYNC_PAGES ; i++)
145
			safe_put_page(r10_bio->devs[j].bio->bi_io_vec[i].bv_page);
Linus Torvalds's avatar
Linus Torvalds committed
146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164
	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++) {
165
				safe_put_page(bio->bi_io_vec[i].bv_page);
Linus Torvalds's avatar
Linus Torvalds committed
166 167 168 169 170 171 172 173 174 175 176 177 178 179
				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;
180
		if (*bio && *bio != IO_BLOCKED)
Linus Torvalds's avatar
Linus Torvalds committed
181 182 183 184 185
			bio_put(*bio);
		*bio = NULL;
	}
}

186
static void free_r10bio(r10bio_t *r10_bio)
Linus Torvalds's avatar
Linus Torvalds committed
187 188 189 190 191 192 193
{
	conf_t *conf = mddev_to_conf(r10_bio->mddev);

	/*
	 * Wake up any possible resync thread that waits for the device
	 * to go idle.
	 */
194
	allow_barrier(conf);
Linus Torvalds's avatar
Linus Torvalds committed
195 196 197 198 199

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

200
static void put_buf(r10bio_t *r10_bio)
Linus Torvalds's avatar
Linus Torvalds committed
201 202 203 204 205
{
	conf_t *conf = mddev_to_conf(r10_bio->mddev);

	mempool_free(r10_bio, conf->r10buf_pool);

206
	lower_barrier(conf);
Linus Torvalds's avatar
Linus Torvalds committed
207 208 209 210 211 212 213 214 215 216
}

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);
217
	conf->nr_queued ++;
Linus Torvalds's avatar
Linus Torvalds committed
218 219
	spin_unlock_irqrestore(&conf->device_lock, flags);

220 221 222
	/* wake up frozen array... */
	wake_up(&conf->wait_barrier);

Linus Torvalds's avatar
Linus Torvalds committed
223 224 225 226 227 228 229 230 231 232 233 234
	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;

235
	bio_endio(bio,
Linus Torvalds's avatar
Linus Torvalds committed
236 237 238 239 240 241 242 243 244 245 246 247 248 249 250
		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);
}

251
static void raid10_end_read_request(struct bio *bio, int error)
Linus Torvalds's avatar
Linus Torvalds committed
252 253 254 255 256 257 258 259 260 261 262 263
{
	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:
	 */
264 265 266
	update_head_pos(slot, r10_bio);

	if (uptodate) {
Linus Torvalds's avatar
Linus Torvalds committed
267 268 269 270 271 272 273 274 275 276 277
		/*
		 * 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);
278
	} else {
Linus Torvalds's avatar
Linus Torvalds committed
279 280 281 282 283 284 285 286 287 288 289 290 291
		/*
		 * 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);
}

292
static void raid10_end_write_request(struct bio *bio, int error)
Linus Torvalds's avatar
Linus Torvalds committed
293 294 295 296 297 298 299 300 301 302 303 304 305 306
{
	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:
	 */
307
	if (!uptodate) {
Linus Torvalds's avatar
Linus Torvalds committed
308
		md_error(r10_bio->mddev, conf->mirrors[dev].rdev);
309 310 311
		/* an I/O failed, we can't clear the bitmap */
		set_bit(R10BIO_Degraded, &r10_bio->state);
	} else
Linus Torvalds's avatar
Linus Torvalds committed
312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330
		/*
		 * 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)) {
331 332 333 334 335
		/* 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);
Linus Torvalds's avatar
Linus Torvalds committed
336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362
		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
363
 * on each device that it is on.
Linus Torvalds's avatar
Linus Torvalds committed
364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385
 *
 * 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);
386 387
	if (conf->far_offset)
		stripe *= conf->far_copies;
Linus Torvalds's avatar
Linus Torvalds committed
388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421

	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;
422 423 424 425 426 427 428 429
	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 {
430
		while (sector >= conf->stride) {
431 432 433 434 435 436 437 438
			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;
	}
Linus Torvalds's avatar
Linus Torvalds committed
439 440 441 442 443 444 445 446
	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
447
 *	@bvm: properties of new bio
Linus Torvalds's avatar
Linus Torvalds committed
448 449 450 451 452 453
 *	@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.
 */
454 455 456
static int raid10_mergeable_bvec(struct request_queue *q,
				 struct bvec_merge_data *bvm,
				 struct bio_vec *biovec)
Linus Torvalds's avatar
Linus Torvalds committed
457 458
{
	mddev_t *mddev = q->queuedata;
459
	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
Linus Torvalds's avatar
Linus Torvalds committed
460 461
	int max;
	unsigned int chunk_sectors = mddev->chunk_size >> 9;
462
	unsigned int bio_sectors = bvm->bi_size >> 9;
Linus Torvalds's avatar
Linus Torvalds committed
463 464 465

	max =  (chunk_sectors - ((sector & (chunk_sectors - 1)) + bio_sectors)) << 9;
	if (max < 0) max = 0; /* bio_add cannot handle a negative return */
466 467
	if (max <= biovec->bv_len && bio_sectors == 0)
		return biovec->bv_len;
Linus Torvalds's avatar
Linus Torvalds committed
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
	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;
497
	mdk_rdev_t *rdev;
Linus Torvalds's avatar
Linus Torvalds committed
498 499 500 501 502

	raid10_find_phys(conf, r10_bio);
	rcu_read_lock();
	/*
	 * Check if we can balance. We can balance on the whole
503 504 505
	 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
506 507 508 509 510 511 512
	 */
	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;

513
		while ((rdev = rcu_dereference(conf->mirrors[disk].rdev)) == NULL ||
514
		       r10_bio->devs[slot].bio == IO_BLOCKED ||
515
		       !test_bit(In_sync, &rdev->flags)) {
Linus Torvalds's avatar
Linus Torvalds committed
516 517 518 519 520 521 522 523 524 525 526 527 528 529 530
			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;
531
	while ((rdev=rcu_dereference(conf->mirrors[disk].rdev)) == NULL ||
532
	       r10_bio->devs[slot].bio == IO_BLOCKED ||
533
	       !test_bit(In_sync, &rdev->flags)) {
Linus Torvalds's avatar
Linus Torvalds committed
534 535 536 537 538 539 540 541 542
		slot ++;
		if (slot == conf->copies) {
			disk = -1;
			goto rb_out;
		}
		disk = r10_bio->devs[slot].devnum;
	}


543 544
	current_distance = abs(r10_bio->devs[slot].addr -
			       conf->mirrors[disk].head_position);
Linus Torvalds's avatar
Linus Torvalds committed
545

546 547
	/* Find the disk whose head is closest,
	 * or - for far > 1 - find the closest to partition beginning */
Linus Torvalds's avatar
Linus Torvalds committed
548 549 550 551 552

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


553
		if ((rdev=rcu_dereference(conf->mirrors[ndisk].rdev)) == NULL ||
554
		    r10_bio->devs[nslot].bio == IO_BLOCKED ||
555
		    !test_bit(In_sync, &rdev->flags))
Linus Torvalds's avatar
Linus Torvalds committed
556 557
			continue;

558 559 560 561 562
		/* 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)) {
Linus Torvalds's avatar
Linus Torvalds committed
563 564 565 566
			disk = ndisk;
			slot = nslot;
			break;
		}
567 568 569 570 571 572 573

		/* 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);
Linus Torvalds's avatar
Linus Torvalds committed
574 575 576 577 578 579 580 581 582 583 584
		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;*/

585
	if (disk >= 0 && (rdev=rcu_dereference(conf->mirrors[disk].rdev))!= NULL)
Linus Torvalds's avatar
Linus Torvalds committed
586
		atomic_inc(&conf->mirrors[disk].rdev->nr_pending);
587 588
	else
		disk = -1;
Linus Torvalds's avatar
Linus Torvalds committed
589 590 591 592 593 594 595 596 597 598 599 600
	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++) {
601
		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
602
		if (rdev && !test_bit(Faulty, &rdev->flags) && atomic_read(&rdev->nr_pending)) {
603
			struct request_queue *r_queue = bdev_get_queue(rdev->bdev);
Linus Torvalds's avatar
Linus Torvalds committed
604 605 606 607

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

608
			blk_unplug(r_queue);
Linus Torvalds's avatar
Linus Torvalds committed
609 610 611 612 613 614 615 616

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

617
static void raid10_unplug(struct request_queue *q)
Linus Torvalds's avatar
Linus Torvalds committed
618
{
619 620
	mddev_t *mddev = q->queuedata;

Linus Torvalds's avatar
Linus Torvalds committed
621
	unplug_slaves(q->queuedata);
622
	md_wakeup_thread(mddev->thread);
Linus Torvalds's avatar
Linus Torvalds committed
623 624
}

625 626 627 628 629 630 631 632 633 634
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)) {
635
			struct request_queue *q = bdev_get_queue(rdev->bdev);
636 637 638 639 640 641 642 643

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

644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673
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;
}
674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693
/* 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.
Linus Torvalds's avatar
Linus Torvalds committed
694 695
 */

696
static void raise_barrier(conf_t *conf, int force)
Linus Torvalds's avatar
Linus Torvalds committed
697
{
698
	BUG_ON(force && !conf->barrier);
Linus Torvalds's avatar
Linus Torvalds committed
699
	spin_lock_irq(&conf->resync_lock);
700

701 702
	/* Wait until no block IO is waiting (unless 'force') */
	wait_event_lock_irq(conf->wait_barrier, force || !conf->nr_waiting,
703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735
			    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--;
Linus Torvalds's avatar
Linus Torvalds committed
736
	}
737
	conf->nr_pending++;
Linus Torvalds's avatar
Linus Torvalds committed
738 739 740
	spin_unlock_irq(&conf->resync_lock);
}

741 742 743 744 745 746 747 748 749
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);
}

750 751 752
static void freeze_array(conf_t *conf)
{
	/* stop syncio and normal IO and wait for everything to
NeilBrown's avatar
NeilBrown committed
753
	 * go quiet.
754
	 * We increment barrier and nr_waiting, and then
755 756 757 758 759 760 761 762
	 * 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.
763 764 765 766 767
	 */
	spin_lock_irq(&conf->resync_lock);
	conf->barrier++;
	conf->nr_waiting++;
	wait_event_lock_irq(conf->wait_barrier,
768
			    conf->nr_pending == conf->nr_queued+1,
769
			    conf->resync_lock,
770 771
			    ({ flush_pending_writes(conf);
			       raid10_unplug(conf->mddev->queue); }));
772 773 774 775 776 777 778 779 780 781 782 783 784
	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);
}

785
static int make_request(struct request_queue *q, struct bio * bio)
Linus Torvalds's avatar
Linus Torvalds committed
786 787 788 789 790 791
{
	mddev_t *mddev = q->queuedata;
	conf_t *conf = mddev_to_conf(mddev);
	mirror_info_t *mirror;
	r10bio_t *r10_bio;
	struct bio *read_bio;
Tejun Heo's avatar
Tejun Heo committed
792
	int cpu;
Linus Torvalds's avatar
Linus Torvalds committed
793 794
	int i;
	int chunk_sects = conf->chunk_mask + 1;
795
	const int rw = bio_data_dir(bio);
796
	const int do_sync = bio_sync(bio);
797 798
	struct bio_list bl;
	unsigned long flags;
799
	mdk_rdev_t *blocked_rdev;
Linus Torvalds's avatar
Linus Torvalds committed
800

801
	if (unlikely(bio_barrier(bio))) {
802
		bio_endio(bio, -EOPNOTSUPP);
803 804 805
		return 0;
	}

Linus Torvalds's avatar
Linus Torvalds committed
806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833
	/* 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);

834
		bio_io_error(bio);
Linus Torvalds's avatar
Linus Torvalds committed
835 836 837
		return 0;
	}

838
	md_write_start(mddev, bio);
839

Linus Torvalds's avatar
Linus Torvalds committed
840 841 842 843 844
	/*
	 * 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.
	 */
845
	wait_barrier(conf);
Linus Torvalds's avatar
Linus Torvalds committed
846

847 848 849 850 851
	cpu = part_stat_lock();
	part_stat_inc(cpu, &mddev->gendisk->part0, ios[rw]);
	part_stat_add(cpu, &mddev->gendisk->part0, sectors[rw],
		      bio_sectors(bio));
	part_stat_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
852 853 854 855 856 857 858 859

	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;
860
	r10_bio->state = 0;
Linus Torvalds's avatar
Linus Torvalds committed
861

862
	if (rw == READ) {
Linus Torvalds's avatar
Linus Torvalds committed
863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881
		/*
		 * 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;
882
		read_bio->bi_rw = READ | do_sync;
Linus Torvalds's avatar
Linus Torvalds committed
883 884 885 886 887 888 889 890 891
		read_bio->bi_private = r10_bio;

		generic_make_request(read_bio);
		return 0;
	}

	/*
	 * WRITE:
	 */
892
	/* first select target devices under rcu_lock and
Linus Torvalds's avatar
Linus Torvalds committed
893 894 895 896
	 * inc refcount on their rdev.  Record them by setting
	 * bios[x] to bio
	 */
	raid10_find_phys(conf, r10_bio);
897
 retry_write:
898
	blocked_rdev = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
899 900 901
	rcu_read_lock();
	for (i = 0;  i < conf->copies; i++) {
		int d = r10_bio->devs[i].devnum;
902
		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[d].rdev);
903 904 905 906 907 908
		if (rdev && unlikely(test_bit(Blocked, &rdev->flags))) {
			atomic_inc(&rdev->nr_pending);
			blocked_rdev = rdev;
			break;
		}
		if (rdev && !test_bit(Faulty, &rdev->flags)) {
909
			atomic_inc(&rdev->nr_pending);
Linus Torvalds's avatar
Linus Torvalds committed
910
			r10_bio->devs[i].bio = bio;
911
		} else {
Linus Torvalds's avatar
Linus Torvalds committed
912
			r10_bio->devs[i].bio = NULL;
913 914
			set_bit(R10BIO_Degraded, &r10_bio->state);
		}
Linus Torvalds's avatar
Linus Torvalds committed
915 916 917
	}
	rcu_read_unlock();

918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933
	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;
	}

934
	atomic_set(&r10_bio->remaining, 0);
935

936
	bio_list_init(&bl);
Linus Torvalds's avatar
Linus Torvalds committed
937 938 939 940 941 942 943 944 945 946 947 948 949
	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;
950
		mbio->bi_rw = WRITE | do_sync;
Linus Torvalds's avatar
Linus Torvalds committed
951 952 953
		mbio->bi_private = r10_bio;

		atomic_inc(&r10_bio->remaining);
954
		bio_list_add(&bl, mbio);
Linus Torvalds's avatar
Linus Torvalds committed
955 956
	}

957 958 959 960 961 962 963
	if (unlikely(!atomic_read(&r10_bio->remaining))) {
		/* the array is dead */
		md_write_end(mddev);
		raid_end_bio_io(r10_bio);
		return 0;
	}

964 965 966 967 968
	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);
Linus Torvalds's avatar
Linus Torvalds committed
969

970 971 972
	/* In case raid10d snuck in to freeze_array */
	wake_up(&conf->wait_barrier);

973 974 975
	if (do_sync)
		md_wakeup_thread(mddev->thread);

Linus Torvalds's avatar
Linus Torvalds committed
976 977 978 979 980 981 982 983 984 985 986 987
	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);
988 989 990 991 992 993
	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);
	}
Linus Torvalds's avatar
Linus Torvalds committed
994
	seq_printf(seq, " [%d/%d] [", conf->raid_disks,
995
					conf->raid_disks - mddev->degraded);
Linus Torvalds's avatar
Linus Torvalds committed
996 997 998
	for (i = 0; i < conf->raid_disks; i++)
		seq_printf(seq, "%s",
			      conf->mirrors[i].rdev &&
999
			      test_bit(In_sync, &conf->mirrors[i].rdev->flags) ? "U" : "_");
Linus Torvalds's avatar
Linus Torvalds committed
1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013
	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
	 */
1014
	if (test_bit(In_sync, &rdev->flags)
1015
	    && conf->raid_disks-mddev->degraded == 1)
Linus Torvalds's avatar
Linus Torvalds committed
1016 1017 1018 1019 1020 1021 1022 1023
		/*
		 * 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;
1024 1025 1026
	if (test_and_clear_bit(In_sync, &rdev->flags)) {
		unsigned long flags;
		spin_lock_irqsave(&conf->device_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
1027
		mddev->degraded++;
1028
		spin_unlock_irqrestore(&conf->device_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
1029 1030 1031
		/*
		 * if recovery is running, make sure it aborts.
		 */
1032
		set_bit(MD_RECOVERY_INTR, &mddev->recovery);
Linus Torvalds's avatar
Linus Torvalds committed
1033
	}
1034
	set_bit(Faulty, &rdev->flags);
1035
	set_bit(MD_CHANGE_DEVS, &mddev->flags);
1036 1037
	printk(KERN_ALERT "raid10: Disk failure on %s, disabling device.\n"
		"raid10: Operation continuing on %d devices.\n",
1038
		bdevname(rdev->bdev,b), conf->raid_disks - mddev->degraded);
Linus Torvalds's avatar
Linus Torvalds committed
1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050
}

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

	printk("RAID10 conf printout:\n");
	if (!conf) {
		printk("(!conf)\n");
		return;
	}
1051
	printk(" --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
Linus Torvalds's avatar
Linus Torvalds committed
1052 1053 1054 1055 1056 1057 1058
		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",
1059 1060
				i, !test_bit(In_sync, &tmp->rdev->flags),
			        !test_bit(Faulty, &tmp->rdev->flags),
Linus Torvalds's avatar
Linus Torvalds committed
1061 1062 1063 1064 1065 1066
				bdevname(tmp->rdev->bdev,b));
	}
}

static void close_sync(conf_t *conf)
{
1067 1068
	wait_barrier(conf);
	allow_barrier(conf);
Linus Torvalds's avatar
Linus Torvalds committed
1069 1070 1071 1072 1073

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

1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094
/* 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;
}

Linus Torvalds's avatar
Linus Torvalds committed
1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107
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
1108
		    && !test_bit(Faulty, &tmp->rdev->flags)
1109 1110 1111
		    && !test_and_set_bit(In_sync, &tmp->rdev->flags)) {
			unsigned long flags;
			spin_lock_irqsave(&conf->device_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
1112
			mddev->degraded--;
1113
			spin_unlock_irqrestore(&conf->device_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
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;
1125
	int err = -EEXIST;
Linus Torvalds's avatar
Linus Torvalds committed
1126 1127
	int mirror;
	mirror_info_t *p;
1128 1129
	int first = 0;
	int last = mddev->raid_disks - 1;
Linus Torvalds's avatar
Linus Torvalds committed
1130 1131 1132 1133 1134

	if (mddev->recovery_cp < MaxSector)
		/* only hot-add to in-sync arrays, as recovery is
		 * very different from resync
		 */
1135
		return -EBUSY;
1136
	if (!enough(conf))
1137
		return -EINVAL;
Linus Torvalds's avatar
Linus Torvalds committed
1138

1139 1140
	if (rdev->raid_disk)
		first = last = rdev->raid_disk;
Linus Torvalds's avatar
Linus Torvalds committed
1141

1142
	if (rdev->saved_raid_disk >= 0 &&
1143
	    rdev->saved_raid_disk >= first &&
1144 1145 1146
	    conf->mirrors[rdev->saved_raid_disk].rdev == NULL)
		mirror = rdev->saved_raid_disk;
	else
1147 1148
		mirror = first;
	for ( ; mirror <= last ; mirror++)
Linus Torvalds's avatar
Linus Torvalds committed
1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162
		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;
1163
			err = 0;
1164 1165
			if (rdev->saved_raid_disk != mirror)
				conf->fullsync = 1;
1166
			rcu_assign_pointer(p->rdev, rdev);
Linus Torvalds's avatar
Linus Torvalds committed
1167 1168 1169 1170
			break;
		}

	print_conf(conf);
1171
	return err;
Linus Torvalds's avatar
Linus Torvalds committed
1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183
}

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) {
1184
		if (test_bit(In_sync, &rdev->flags) ||
Linus Torvalds's avatar
Linus Torvalds committed
1185 1186 1187 1188
		    atomic_read(&rdev->nr_pending)) {
			err = -EBUSY;
			goto abort;
		}
1189 1190 1191 1192 1193 1194 1195 1196
		/* Only remove faulty devices in recovery
		 * is not possible.
		 */
		if (!test_bit(Faulty, &rdev->flags) &&
		    enough(conf)) {
			err = -EBUSY;
			goto abort;
		}
Linus Torvalds's avatar
Linus Torvalds committed
1197
		p->rdev = NULL;
1198
		synchronize_rcu();
Linus Torvalds's avatar
Linus Torvalds committed
1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211
		if (atomic_read(&rdev->nr_pending)) {
			/* lost the race, try later */
			err = -EBUSY;
			p->rdev = rdev;
		}
	}
abort:

	print_conf(conf);
	return err;
}


1212
static void end_sync_read(struct bio *bio, int error)
Linus Torvalds's avatar
Linus Torvalds committed
1213 1214 1215 1216 1217 1218 1219 1220
{
	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;
1221
	BUG_ON(i == conf->copies);
Linus Torvalds's avatar
Linus Torvalds committed
1222 1223
	update_head_pos(i, r10_bio);
	d = r10_bio->devs[i].devnum;
1224 1225 1226

	if (test_bit(BIO_UPTODATE, &bio->bi_flags))
		set_bit(R10BIO_Uptodate, &r10_bio->state);
1227 1228 1229 1230 1231 1232 1233
	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);
	}
Linus Torvalds's avatar
Linus Torvalds committed
1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247

	/* 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);
}

1248
static void end_sync_write(struct bio *bio, int error)
Linus Torvalds's avatar
Linus Torvalds committed
1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262
{
	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);
1263

Linus Torvalds's avatar
Linus Torvalds committed
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 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316
	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 */
1317 1318 1319
	for (i=0 ; i < conf->copies ; i++) {
		int  j, d;
		int vcnt = r10_bio->sectors >> (PAGE_SHIFT-9);
Linus Torvalds's avatar
Linus Torvalds committed
1320 1321

		tbio = r10_bio->devs[i].bio;
1322 1323 1324 1325

		if (tbio->bi_end_io != end_sync_read)
			continue;
		if (i == first)
Linus Torvalds's avatar
Linus Torvalds committed
1326
			continue;
1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340
		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;
		}
1341 1342 1343
		if (test_bit(MD_RECOVERY_CHECK, &mddev->recovery))
			/* Don't fix anything. */
			continue;
Linus Torvalds's avatar
Linus Torvalds committed
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 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417
		/* 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_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);
1418 1419 1420
	if (test_bit(R10BIO_Uptodate, &r10_bio->state))
		generic_make_request(wbio);
	else
1421
		bio_endio(wbio, -EIO);
Linus Torvalds's avatar
Linus Torvalds committed
1422 1423 1424 1425 1426 1427 1428 1429
}


/*
 * This is a kernel thread which:
 *
 *	1.	Retries failed read operations on working mirrors.
 *	2.	Updates the raid superblock when problems encounter.
1430
 *	3.	Performs writes following reads for array synchronising.
Linus Torvalds's avatar
Linus Torvalds committed
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 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527
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,
1528 1529
					       (unsigned long long)(sect+
					            rdev->data_offset),