raid1.c 59.7 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11
/*
 * raid1.c : Multiple Devices driver for Linux
 *
 * Copyright (C) 1999, 2000, 2001 Ingo Molnar, Red Hat
 *
 * Copyright (C) 1996, 1997, 1998 Ingo Molnar, Miguel de Icaza, Gadi Oxman
 *
 * RAID-1 management functions.
 *
 * Better read-balancing code written by Mika Kuoppala <miku@iki.fi>, 2000
 *
12
 * Fixes to reconstruction by Jakob Østergaard" <jakob@ostenfeld.dk>
L
Linus Torvalds 已提交
13 14
 * Various fixes by Neil Brown <neilb@cse.unsw.edu.au>
 *
15 16 17 18 19 20 21 22 23
 * Changes by Peter T. Breuer <ptb@it.uc3m.es> 31/1/2003 to support
 * bitmapped intelligence in resync:
 *
 *      - bitmap marked during normal i/o
 *      - bitmap used to skip nondirty blocks during sync
 *
 * Additions to bitmap code, (C) 2003-2004 Paul Clements, SteelEye Technology:
 * - persistent bitmap code
 *
L
Linus Torvalds 已提交
24 25 26 27 28 29 30 31 32 33
 * 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.
 */

34
#include "dm-bio-list.h"
35
#include <linux/delay.h>
L
Linus Torvalds 已提交
36
#include <linux/raid/raid1.h>
37 38 39 40 41 42 43 44
#include <linux/raid/bitmap.h>

#define DEBUG 0
#if DEBUG
#define PRINTK(x...) printk(x)
#else
#define PRINTK(x...)
#endif
L
Linus Torvalds 已提交
45 46 47 48 49 50 51 52 53

/*
 * Number of guaranteed r1bios in case of extreme VM load:
 */
#define	NR_RAID1_BIOS 256


static void unplug_slaves(mddev_t *mddev);

54 55
static void allow_barrier(conf_t *conf);
static void lower_barrier(conf_t *conf);
L
Linus Torvalds 已提交
56

A
Al Viro 已提交
57
static void * r1bio_pool_alloc(gfp_t gfp_flags, void *data)
L
Linus Torvalds 已提交
58 59 60 61 62 63
{
	struct pool_info *pi = data;
	r1bio_t *r1_bio;
	int size = offsetof(r1bio_t, bios[pi->raid_disks]);

	/* allocate a r1bio with room for raid_disks entries in the bios array */
64 65
	r1_bio = kzalloc(size, gfp_flags);
	if (!r1_bio)
L
Linus Torvalds 已提交
66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81
		unplug_slaves(pi->mddev);

	return r1_bio;
}

static void r1bio_pool_free(void *r1_bio, void *data)
{
	kfree(r1_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)

A
Al Viro 已提交
82
static void * r1buf_pool_alloc(gfp_t gfp_flags, void *data)
L
Linus Torvalds 已提交
83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106
{
	struct pool_info *pi = data;
	struct page *page;
	r1bio_t *r1_bio;
	struct bio *bio;
	int i, j;

	r1_bio = r1bio_pool_alloc(gfp_flags, pi);
	if (!r1_bio) {
		unplug_slaves(pi->mddev);
		return NULL;
	}

	/*
	 * Allocate bios : 1 for reading, n-1 for writing
	 */
	for (j = pi->raid_disks ; j-- ; ) {
		bio = bio_alloc(gfp_flags, RESYNC_PAGES);
		if (!bio)
			goto out_free_bio;
		r1_bio->bios[j] = bio;
	}
	/*
	 * Allocate RESYNC_PAGES data pages and attach them to
107 108 109
	 * the first bio.
	 * If this is a user-requested check/repair, allocate
	 * RESYNC_PAGES for each bio.
L
Linus Torvalds 已提交
110
	 */
111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130
	if (test_bit(MD_RECOVERY_REQUESTED, &pi->mddev->recovery))
		j = pi->raid_disks;
	else
		j = 1;
	while(j--) {
		bio = r1_bio->bios[j];
		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;
		}
	}
	/* If not user-requests, copy the page pointers to all bios */
	if (!test_bit(MD_RECOVERY_REQUESTED, &pi->mddev->recovery)) {
		for (i=0; i<RESYNC_PAGES ; i++)
			for (j=1; j<pi->raid_disks; j++)
				r1_bio->bios[j]->bi_io_vec[i].bv_page =
					r1_bio->bios[0]->bi_io_vec[i].bv_page;
L
Linus Torvalds 已提交
131 132 133 134 135 136 137
	}

	r1_bio->master_bio = NULL;

	return r1_bio;

out_free_pages:
138 139
	for (i=0; i < RESYNC_PAGES ; i++)
		for (j=0 ; j < pi->raid_disks; j++)
140
			safe_put_page(r1_bio->bios[j]->bi_io_vec[i].bv_page);
141
	j = -1;
L
Linus Torvalds 已提交
142 143 144 145 146 147 148 149 150 151
out_free_bio:
	while ( ++j < pi->raid_disks )
		bio_put(r1_bio->bios[j]);
	r1bio_pool_free(r1_bio, data);
	return NULL;
}

static void r1buf_pool_free(void *__r1_bio, void *data)
{
	struct pool_info *pi = data;
152
	int i,j;
L
Linus Torvalds 已提交
153 154
	r1bio_t *r1bio = __r1_bio;

155 156 157 158 159
	for (i = 0; i < RESYNC_PAGES; i++)
		for (j = pi->raid_disks; j-- ;) {
			if (j == 0 ||
			    r1bio->bios[j]->bi_io_vec[i].bv_page !=
			    r1bio->bios[0]->bi_io_vec[i].bv_page)
160
				safe_put_page(r1bio->bios[j]->bi_io_vec[i].bv_page);
161
		}
L
Linus Torvalds 已提交
162 163 164 165 166 167 168 169 170 171 172 173
	for (i=0 ; i < pi->raid_disks; i++)
		bio_put(r1bio->bios[i]);

	r1bio_pool_free(r1bio, data);
}

static void put_all_bios(conf_t *conf, r1bio_t *r1_bio)
{
	int i;

	for (i = 0; i < conf->raid_disks; i++) {
		struct bio **bio = r1_bio->bios + i;
174
		if (*bio && *bio != IO_BLOCKED)
L
Linus Torvalds 已提交
175 176 177 178 179
			bio_put(*bio);
		*bio = NULL;
	}
}

180
static void free_r1bio(r1bio_t *r1_bio)
L
Linus Torvalds 已提交
181 182 183 184 185 186 187
{
	conf_t *conf = mddev_to_conf(r1_bio->mddev);

	/*
	 * Wake up any possible resync thread that waits for the device
	 * to go idle.
	 */
188
	allow_barrier(conf);
L
Linus Torvalds 已提交
189 190 191 192 193

	put_all_bios(conf, r1_bio);
	mempool_free(r1_bio, conf->r1bio_pool);
}

194
static void put_buf(r1bio_t *r1_bio)
L
Linus Torvalds 已提交
195 196
{
	conf_t *conf = mddev_to_conf(r1_bio->mddev);
197 198 199 200 201 202 203
	int i;

	for (i=0; i<conf->raid_disks; i++) {
		struct bio *bio = r1_bio->bios[i];
		if (bio->bi_end_io)
			rdev_dec_pending(conf->mirrors[i].rdev, r1_bio->mddev);
	}
L
Linus Torvalds 已提交
204 205 206

	mempool_free(r1_bio, conf->r1buf_pool);

207
	lower_barrier(conf);
L
Linus Torvalds 已提交
208 209 210 211 212 213 214 215 216 217
}

static void reschedule_retry(r1bio_t *r1_bio)
{
	unsigned long flags;
	mddev_t *mddev = r1_bio->mddev;
	conf_t *conf = mddev_to_conf(mddev);

	spin_lock_irqsave(&conf->device_lock, flags);
	list_add(&r1_bio->retry_list, &conf->retry_list);
218
	conf->nr_queued ++;
L
Linus Torvalds 已提交
219 220
	spin_unlock_irqrestore(&conf->device_lock, flags);

221
	wake_up(&conf->wait_barrier);
L
Linus Torvalds 已提交
222 223 224 225 226 227 228 229 230 231 232 233
	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(r1bio_t *r1_bio)
{
	struct bio *bio = r1_bio->master_bio;

234 235 236 237 238 239 240 241
	/* if nobody has done the final endio yet, do it now */
	if (!test_and_set_bit(R1BIO_Returned, &r1_bio->state)) {
		PRINTK(KERN_DEBUG "raid1: sync end %s on sectors %llu-%llu\n",
			(bio_data_dir(bio) == WRITE) ? "write" : "read",
			(unsigned long long) bio->bi_sector,
			(unsigned long long) bio->bi_sector +
				(bio->bi_size >> 9) - 1);

242
		bio_endio(bio,
243 244
			test_bit(R1BIO_Uptodate, &r1_bio->state) ? 0 : -EIO);
	}
L
Linus Torvalds 已提交
245 246 247 248 249 250 251 252 253 254 255 256 257 258
	free_r1bio(r1_bio);
}

/*
 * Update disk head position estimator based on IRQ completion info.
 */
static inline void update_head_pos(int disk, r1bio_t *r1_bio)
{
	conf_t *conf = mddev_to_conf(r1_bio->mddev);

	conf->mirrors[disk].head_position =
		r1_bio->sector + (r1_bio->sectors);
}

259
static void raid1_end_read_request(struct bio *bio, int error)
L
Linus Torvalds 已提交
260 261 262 263 264 265 266 267 268 269
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
	r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private);
	int mirror;
	conf_t *conf = mddev_to_conf(r1_bio->mddev);

	mirror = r1_bio->read_disk;
	/*
	 * this branch is our 'one mirror IO has finished' event handler:
	 */
270 271
	update_head_pos(mirror, r1_bio);

272 273 274 275 276 277
	if (uptodate)
		set_bit(R1BIO_Uptodate, &r1_bio->state);
	else {
		/* If all other devices have failed, we want to return
		 * the error upwards rather than fail the last device.
		 * Here we redefine "uptodate" to mean "Don't want to retry"
L
Linus Torvalds 已提交
278
		 */
279 280 281 282 283 284 285 286
		unsigned long flags;
		spin_lock_irqsave(&conf->device_lock, flags);
		if (r1_bio->mddev->degraded == conf->raid_disks ||
		    (r1_bio->mddev->degraded == conf->raid_disks-1 &&
		     !test_bit(Faulty, &conf->mirrors[mirror].rdev->flags)))
			uptodate = 1;
		spin_unlock_irqrestore(&conf->device_lock, flags);
	}
L
Linus Torvalds 已提交
287

288
	if (uptodate)
L
Linus Torvalds 已提交
289
		raid_end_bio_io(r1_bio);
290
	else {
L
Linus Torvalds 已提交
291 292 293 294 295 296 297 298 299 300 301 302 303
		/*
		 * oops, read error:
		 */
		char b[BDEVNAME_SIZE];
		if (printk_ratelimit())
			printk(KERN_ERR "raid1: %s: rescheduling sector %llu\n",
			       bdevname(conf->mirrors[mirror].rdev->bdev,b), (unsigned long long)r1_bio->sector);
		reschedule_retry(r1_bio);
	}

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

304
static void raid1_end_write_request(struct bio *bio, int error)
L
Linus Torvalds 已提交
305 306 307
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
	r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private);
308
	int mirror, behind = test_bit(R1BIO_BehindIO, &r1_bio->state);
L
Linus Torvalds 已提交
309
	conf_t *conf = mddev_to_conf(r1_bio->mddev);
310
	struct bio *to_put = NULL;
L
Linus Torvalds 已提交
311 312 313 314 315 316


	for (mirror = 0; mirror < conf->raid_disks; mirror++)
		if (r1_bio->bios[mirror] == bio)
			break;

317
	if (error == -EOPNOTSUPP && test_bit(R1BIO_Barrier, &r1_bio->state)) {
318 319 320
		set_bit(BarriersNotsupp, &conf->mirrors[mirror].rdev->flags);
		set_bit(R1BIO_BarrierRetry, &r1_bio->state);
		r1_bio->mddev->barriers_work = 0;
321
		/* Don't rdev_dec_pending in this branch - keep it for the retry */
322
	} else {
L
Linus Torvalds 已提交
323
		/*
324
		 * this branch is our 'one mirror IO has finished' event handler:
L
Linus Torvalds 已提交
325
		 */
326
		r1_bio->bios[mirror] = NULL;
327
		to_put = bio;
328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363
		if (!uptodate) {
			md_error(r1_bio->mddev, conf->mirrors[mirror].rdev);
			/* an I/O failed, we can't clear the bitmap */
			set_bit(R1BIO_Degraded, &r1_bio->state);
		} else
			/*
			 * Set R1BIO_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(R1BIO_Uptodate, &r1_bio->state);

		update_head_pos(mirror, r1_bio);

		if (behind) {
			if (test_bit(WriteMostly, &conf->mirrors[mirror].rdev->flags))
				atomic_dec(&r1_bio->behind_remaining);

			/* In behind mode, we ACK the master bio once the I/O has safely
			 * reached all non-writemostly disks. Setting the Returned bit
			 * ensures that this gets done only once -- we don't ever want to
			 * return -EIO here, instead we'll wait */

			if (atomic_read(&r1_bio->behind_remaining) >= (atomic_read(&r1_bio->remaining)-1) &&
			    test_bit(R1BIO_Uptodate, &r1_bio->state)) {
				/* Maybe we can return now */
				if (!test_and_set_bit(R1BIO_Returned, &r1_bio->state)) {
					struct bio *mbio = r1_bio->master_bio;
					PRINTK(KERN_DEBUG "raid1: behind end write sectors %llu-%llu\n",
					       (unsigned long long) mbio->bi_sector,
					       (unsigned long long) mbio->bi_sector +
					       (mbio->bi_size >> 9) - 1);
364
					bio_endio(mbio, 0);
365
				}
366 367
			}
		}
368
		rdev_dec_pending(conf->mirrors[mirror].rdev, conf->mddev);
369
	}
L
Linus Torvalds 已提交
370 371 372 373 374 375
	/*
	 *
	 * Let's see if all mirrored write operations have finished
	 * already.
	 */
	if (atomic_dec_and_test(&r1_bio->remaining)) {
376
		if (test_bit(R1BIO_BarrierRetry, &r1_bio->state))
377
			reschedule_retry(r1_bio);
378 379 380 381 382 383 384 385 386 387 388 389 390 391 392
		else {
			/* it really is the end of this request */
			if (test_bit(R1BIO_BehindIO, &r1_bio->state)) {
				/* free extra copy of the data pages */
				int i = bio->bi_vcnt;
				while (i--)
					safe_put_page(bio->bi_io_vec[i].bv_page);
			}
			/* clear the bitmap if all writes complete successfully */
			bitmap_endwrite(r1_bio->mddev->bitmap, r1_bio->sector,
					r1_bio->sectors,
					!test_bit(R1BIO_Degraded, &r1_bio->state),
					behind);
			md_write_end(r1_bio->mddev);
			raid_end_bio_io(r1_bio);
393
		}
L
Linus Torvalds 已提交
394
	}
395

396 397
	if (to_put)
		bio_put(to_put);
L
Linus Torvalds 已提交
398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418
}


/*
 * 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.
 */
static int read_balance(conf_t *conf, r1bio_t *r1_bio)
{
	const unsigned long this_sector = r1_bio->sector;
	int new_disk = conf->last_used, disk = new_disk;
419
	int wonly_disk = -1;
L
Linus Torvalds 已提交
420 421
	const int sectors = r1_bio->sectors;
	sector_t new_distance, current_distance;
422
	mdk_rdev_t *rdev;
L
Linus Torvalds 已提交
423 424 425

	rcu_read_lock();
	/*
426
	 * Check if we can balance. We can balance on the whole
L
Linus Torvalds 已提交
427 428 429 430 431 432 433 434 435
	 * device if no resync is going on, or below the resync window.
	 * We take the first readable disk when above the resync window.
	 */
 retry:
	if (conf->mddev->recovery_cp < MaxSector &&
	    (this_sector + sectors >= conf->next_resync)) {
		/* Choose the first operation device, for consistancy */
		new_disk = 0;

436
		for (rdev = rcu_dereference(conf->mirrors[new_disk].rdev);
437
		     r1_bio->bios[new_disk] == IO_BLOCKED ||
438
		     !rdev || !test_bit(In_sync, &rdev->flags)
439
			     || test_bit(WriteMostly, &rdev->flags);
440
		     rdev = rcu_dereference(conf->mirrors[++new_disk].rdev)) {
441

442 443
			if (rdev && test_bit(In_sync, &rdev->flags) &&
				r1_bio->bios[new_disk] != IO_BLOCKED)
444 445 446 447
				wonly_disk = new_disk;

			if (new_disk == conf->raid_disks - 1) {
				new_disk = wonly_disk;
L
Linus Torvalds 已提交
448 449 450 451 452 453 454 455
				break;
			}
		}
		goto rb_out;
	}


	/* make sure the disk is operational */
456
	for (rdev = rcu_dereference(conf->mirrors[new_disk].rdev);
457
	     r1_bio->bios[new_disk] == IO_BLOCKED ||
458
	     !rdev || !test_bit(In_sync, &rdev->flags) ||
459
		     test_bit(WriteMostly, &rdev->flags);
460
	     rdev = rcu_dereference(conf->mirrors[new_disk].rdev)) {
461

462 463
		if (rdev && test_bit(In_sync, &rdev->flags) &&
		    r1_bio->bios[new_disk] != IO_BLOCKED)
464 465
			wonly_disk = new_disk;

L
Linus Torvalds 已提交
466 467 468 469
		if (new_disk <= 0)
			new_disk = conf->raid_disks;
		new_disk--;
		if (new_disk == disk) {
470 471
			new_disk = wonly_disk;
			break;
L
Linus Torvalds 已提交
472 473
		}
	}
474 475 476 477

	if (new_disk < 0)
		goto rb_out;

L
Linus Torvalds 已提交
478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497
	disk = new_disk;
	/* now disk == new_disk == starting point for search */

	/*
	 * Don't change to another disk for sequential reads:
	 */
	if (conf->next_seq_sect == this_sector)
		goto rb_out;
	if (this_sector == conf->mirrors[new_disk].head_position)
		goto rb_out;

	current_distance = abs(this_sector - conf->mirrors[disk].head_position);

	/* Find the disk whose head is closest */

	do {
		if (disk <= 0)
			disk = conf->raid_disks;
		disk--;

498
		rdev = rcu_dereference(conf->mirrors[disk].rdev);
499

500
		if (!rdev || r1_bio->bios[disk] == IO_BLOCKED ||
501
		    !test_bit(In_sync, &rdev->flags) ||
502
		    test_bit(WriteMostly, &rdev->flags))
L
Linus Torvalds 已提交
503 504 505 506 507 508 509 510 511 512 513 514 515
			continue;

		if (!atomic_read(&rdev->nr_pending)) {
			new_disk = disk;
			break;
		}
		new_distance = abs(this_sector - conf->mirrors[disk].head_position);
		if (new_distance < current_distance) {
			current_distance = new_distance;
			new_disk = disk;
		}
	} while (disk != conf->last_used);

516
 rb_out:
L
Linus Torvalds 已提交
517 518 519


	if (new_disk >= 0) {
520
		rdev = rcu_dereference(conf->mirrors[new_disk].rdev);
521 522 523
		if (!rdev)
			goto retry;
		atomic_inc(&rdev->nr_pending);
524
		if (!test_bit(In_sync, &rdev->flags)) {
L
Linus Torvalds 已提交
525 526 527
			/* cannot risk returning a device that failed
			 * before we inc'ed nr_pending
			 */
528
			rdev_dec_pending(rdev, conf->mddev);
L
Linus Torvalds 已提交
529 530
			goto retry;
		}
531 532
		conf->next_seq_sect = this_sector + sectors;
		conf->last_used = new_disk;
L
Linus Torvalds 已提交
533 534 535 536 537 538 539 540 541 542 543 544 545
	}
	rcu_read_unlock();

	return new_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++) {
546
		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
547
		if (rdev && !test_bit(Faulty, &rdev->flags) && atomic_read(&rdev->nr_pending)) {
548
			struct request_queue *r_queue = bdev_get_queue(rdev->bdev);
L
Linus Torvalds 已提交
549 550 551 552

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

553
			blk_unplug(r_queue);
L
Linus Torvalds 已提交
554 555 556 557 558 559 560 561

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

562
static void raid1_unplug(struct request_queue *q)
L
Linus Torvalds 已提交
563
{
564 565 566 567
	mddev_t *mddev = q->queuedata;

	unplug_slaves(mddev);
	md_wakeup_thread(mddev->thread);
L
Linus Torvalds 已提交
568 569
}

570 571 572 573 574 575 576 577 578 579
static int raid1_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; i++) {
		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
		if (rdev && !test_bit(Faulty, &rdev->flags)) {
580
			struct request_queue *q = bdev_get_queue(rdev->bdev);
581 582 583 584 585 586 587 588 589 590 591 592 593 594 595

			/* Note the '|| 1' - when read_balance prefers
			 * non-congested targets, it can be removed
			 */
			if ((bits & (1<<BDI_write_congested)) || 1)
				ret |= bdi_congested(&q->backing_dev_info, bits);
			else
				ret &= bdi_congested(&q->backing_dev_info, bits);
		}
	}
	rcu_read_unlock();
	return ret;
}


596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626
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;
}

627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646
/* 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
647 648 649
 */
#define RESYNC_DEPTH 32

650
static void raise_barrier(conf_t *conf)
L
Linus Torvalds 已提交
651 652
{
	spin_lock_irq(&conf->resync_lock);
653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688

	/* Wait until no block IO is waiting */
	wait_event_lock_irq(conf->wait_barrier, !conf->nr_waiting,
			    conf->resync_lock,
			    raid1_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,
			    raid1_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,
				    raid1_unplug(conf->mddev->queue));
		conf->nr_waiting--;
L
Linus Torvalds 已提交
689
	}
690
	conf->nr_pending++;
L
Linus Torvalds 已提交
691 692 693
	spin_unlock_irq(&conf->resync_lock);
}

694 695 696 697 698 699 700 701 702
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);
}

703 704 705 706 707
static void freeze_array(conf_t *conf)
{
	/* stop syncio and normal IO and wait for everything to
	 * go quite.
	 * We increment barrier and nr_waiting, and then
708 709 710 711 712 713 714 715
	 * 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.
716 717 718 719 720
	 */
	spin_lock_irq(&conf->resync_lock);
	conf->barrier++;
	conf->nr_waiting++;
	wait_event_lock_irq(conf->wait_barrier,
721
			    conf->nr_pending == conf->nr_queued+1,
722
			    conf->resync_lock,
723 724
			    ({ flush_pending_writes(conf);
			       raid1_unplug(conf->mddev->queue); }));
725 726 727 728 729 730 731 732 733 734 735 736
	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);
}

737

738 739 740 741 742
/* duplicate the data pages for behind I/O */
static struct page **alloc_behind_pages(struct bio *bio)
{
	int i;
	struct bio_vec *bvec;
743
	struct page **pages = kzalloc(bio->bi_vcnt * sizeof(struct page *),
744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762
					GFP_NOIO);
	if (unlikely(!pages))
		goto do_sync_io;

	bio_for_each_segment(bvec, bio, i) {
		pages[i] = alloc_page(GFP_NOIO);
		if (unlikely(!pages[i]))
			goto do_sync_io;
		memcpy(kmap(pages[i]) + bvec->bv_offset,
			kmap(bvec->bv_page) + bvec->bv_offset, bvec->bv_len);
		kunmap(pages[i]);
		kunmap(bvec->bv_page);
	}

	return pages;

do_sync_io:
	if (pages)
		for (i = 0; i < bio->bi_vcnt && pages[i]; i++)
763
			put_page(pages[i]);
764 765 766 767 768
	kfree(pages);
	PRINTK("%dB behind alloc failed, doing sync I/O\n", bio->bi_size);
	return NULL;
}

769
static int make_request(struct request_queue *q, struct bio * bio)
L
Linus Torvalds 已提交
770 771 772 773 774 775
{
	mddev_t *mddev = q->queuedata;
	conf_t *conf = mddev_to_conf(mddev);
	mirror_info_t *mirror;
	r1bio_t *r1_bio;
	struct bio *read_bio;
776
	int i, targets = 0, disks;
777
	struct bitmap *bitmap;
778 779
	unsigned long flags;
	struct bio_list bl;
780
	struct page **behind_pages = NULL;
781
	const int rw = bio_data_dir(bio);
782
	const int do_sync = bio_sync(bio);
T
Tejun Heo 已提交
783
	int cpu, do_barriers;
784
	mdk_rdev_t *blocked_rdev;
785

L
Linus Torvalds 已提交
786 787 788 789
	/*
	 * 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.
790 791 792
	 * We test barriers_work *after* md_write_start as md_write_start
	 * may cause the first superblock write, and that will check out
	 * if barriers work.
L
Linus Torvalds 已提交
793
	 */
794

795 796
	md_write_start(mddev, bio); /* wait on superblock update early */

797 798 799
	if (unlikely(!mddev->barriers_work && bio_barrier(bio))) {
		if (rw == WRITE)
			md_write_end(mddev);
800
		bio_endio(bio, -EOPNOTSUPP);
801 802 803
		return 0;
	}

804
	wait_barrier(conf);
L
Linus Torvalds 已提交
805

806 807
	bitmap = mddev->bitmap;

T
Tejun Heo 已提交
808 809 810 811 812
	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();
L
Linus Torvalds 已提交
813 814 815 816 817 818 819 820 821 822

	/*
	 * make_request() can abort the operation when READA is being
	 * used and no empty request is available.
	 *
	 */
	r1_bio = mempool_alloc(conf->r1bio_pool, GFP_NOIO);

	r1_bio->master_bio = bio;
	r1_bio->sectors = bio->bi_size >> 9;
823
	r1_bio->state = 0;
L
Linus Torvalds 已提交
824 825 826
	r1_bio->mddev = mddev;
	r1_bio->sector = bio->bi_sector;

827
	if (rw == READ) {
L
Linus Torvalds 已提交
828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848
		/*
		 * read balancing logic:
		 */
		int rdisk = read_balance(conf, r1_bio);

		if (rdisk < 0) {
			/* couldn't find anywhere to read from */
			raid_end_bio_io(r1_bio);
			return 0;
		}
		mirror = conf->mirrors + rdisk;

		r1_bio->read_disk = rdisk;

		read_bio = bio_clone(bio, GFP_NOIO);

		r1_bio->bios[rdisk] = read_bio;

		read_bio->bi_sector = r1_bio->sector + mirror->rdev->data_offset;
		read_bio->bi_bdev = mirror->rdev->bdev;
		read_bio->bi_end_io = raid1_end_read_request;
849
		read_bio->bi_rw = READ | do_sync;
L
Linus Torvalds 已提交
850 851 852 853 854 855 856 857 858 859 860 861 862 863
		read_bio->bi_private = r1_bio;

		generic_make_request(read_bio);
		return 0;
	}

	/*
	 * WRITE:
	 */
	/* first select target devices under spinlock and
	 * inc refcount on their rdev.  Record them by setting
	 * bios[x] to bio
	 */
	disks = conf->raid_disks;
864 865 866 867 868 869 870
#if 0
	{ static int first=1;
	if (first) printk("First Write sector %llu disks %d\n",
			  (unsigned long long)r1_bio->sector, disks);
	first = 0;
	}
#endif
871 872
 retry_write:
	blocked_rdev = NULL;
L
Linus Torvalds 已提交
873 874
	rcu_read_lock();
	for (i = 0;  i < disks; i++) {
875 876 877 878 879 880 881
		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
		if (rdev && unlikely(test_bit(Blocked, &rdev->flags))) {
			atomic_inc(&rdev->nr_pending);
			blocked_rdev = rdev;
			break;
		}
		if (rdev && !test_bit(Faulty, &rdev->flags)) {
L
Linus Torvalds 已提交
882
			atomic_inc(&rdev->nr_pending);
883
			if (test_bit(Faulty, &rdev->flags)) {
884
				rdev_dec_pending(rdev, mddev);
L
Linus Torvalds 已提交
885 886 887
				r1_bio->bios[i] = NULL;
			} else
				r1_bio->bios[i] = bio;
888
			targets++;
L
Linus Torvalds 已提交
889 890 891 892 893
		} else
			r1_bio->bios[i] = NULL;
	}
	rcu_read_unlock();

894 895 896 897 898 899 900 901 902 903 904 905 906 907
	if (unlikely(blocked_rdev)) {
		/* Wait for this device to become unblocked */
		int j;

		for (j = 0; j < i; j++)
			if (r1_bio->bios[j])
				rdev_dec_pending(conf->mirrors[j].rdev, mddev);

		allow_barrier(conf);
		md_wait_for_blocked_rdev(blocked_rdev, mddev);
		wait_barrier(conf);
		goto retry_write;
	}

908 909
	BUG_ON(targets == 0); /* we never fail the last device */

910 911 912 913 914 915
	if (targets < conf->raid_disks) {
		/* array is degraded, we will not clear the bitmap
		 * on I/O completion (see raid1_end_write_request) */
		set_bit(R1BIO_Degraded, &r1_bio->state);
	}

916 917 918 919 920 921
	/* do behind I/O ? */
	if (bitmap &&
	    atomic_read(&bitmap->behind_writes) < bitmap->max_write_behind &&
	    (behind_pages = alloc_behind_pages(bio)) != NULL)
		set_bit(R1BIO_BehindIO, &r1_bio->state);

922
	atomic_set(&r1_bio->remaining, 0);
923
	atomic_set(&r1_bio->behind_remaining, 0);
924

925
	do_barriers = bio_barrier(bio);
926 927 928
	if (do_barriers)
		set_bit(R1BIO_Barrier, &r1_bio->state);

929
	bio_list_init(&bl);
L
Linus Torvalds 已提交
930 931 932 933 934 935 936 937 938 939 940
	for (i = 0; i < disks; i++) {
		struct bio *mbio;
		if (!r1_bio->bios[i])
			continue;

		mbio = bio_clone(bio, GFP_NOIO);
		r1_bio->bios[i] = mbio;

		mbio->bi_sector	= r1_bio->sector + conf->mirrors[i].rdev->data_offset;
		mbio->bi_bdev = conf->mirrors[i].rdev->bdev;
		mbio->bi_end_io	= raid1_end_write_request;
941
		mbio->bi_rw = WRITE | do_barriers | do_sync;
L
Linus Torvalds 已提交
942 943
		mbio->bi_private = r1_bio;

944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960
		if (behind_pages) {
			struct bio_vec *bvec;
			int j;

			/* Yes, I really want the '__' version so that
			 * we clear any unused pointer in the io_vec, rather
			 * than leave them unchanged.  This is important
			 * because when we come to free the pages, we won't
			 * know the originial bi_idx, so we just free
			 * them all
			 */
			__bio_for_each_segment(bvec, mbio, j, 0)
				bvec->bv_page = behind_pages[j];
			if (test_bit(WriteMostly, &conf->mirrors[i].rdev->flags))
				atomic_inc(&r1_bio->behind_remaining);
		}

L
Linus Torvalds 已提交
961 962
		atomic_inc(&r1_bio->remaining);

963
		bio_list_add(&bl, mbio);
L
Linus Torvalds 已提交
964
	}
965
	kfree(behind_pages); /* the behind pages are attached to the bios now */
L
Linus Torvalds 已提交
966

967 968
	bitmap_startwrite(bitmap, bio->bi_sector, r1_bio->sectors,
				test_bit(R1BIO_BehindIO, &r1_bio->state));
969 970 971 972 973 974 975
	spin_lock_irqsave(&conf->device_lock, flags);
	bio_list_merge(&conf->pending_bio_list, &bl);
	bio_list_init(&bl);

	blk_plug_device(mddev->queue);
	spin_unlock_irqrestore(&conf->device_lock, flags);

976 977 978
	/* In case raid1d snuck into freeze_array */
	wake_up(&conf->wait_barrier);

979 980
	if (do_sync)
		md_wakeup_thread(mddev->thread);
981 982 983 984 985
#if 0
	while ((bio = bio_list_pop(&bl)) != NULL)
		generic_make_request(bio);
#endif

L
Linus Torvalds 已提交
986 987 988 989 990 991 992 993 994
	return 0;
}

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

	seq_printf(seq, " [%d/%d] [", conf->raid_disks,
995
		   conf->raid_disks - mddev->degraded);
996 997 998
	rcu_read_lock();
	for (i = 0; i < conf->raid_disks; i++) {
		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
L
Linus Torvalds 已提交
999
		seq_printf(seq, "%s",
1000 1001 1002
			   rdev && test_bit(In_sync, &rdev->flags) ? "U" : "_");
	}
	rcu_read_unlock();
L
Linus Torvalds 已提交
1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017
	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
	 */
1018
	if (test_bit(In_sync, &rdev->flags)
1019
	    && (conf->raid_disks - mddev->degraded) == 1)
L
Linus Torvalds 已提交
1020 1021 1022 1023 1024
		/*
		 * Don't fail the drive, act as though we were just a
		 * normal single drive
		 */
		return;
1025 1026 1027
	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
1028
		mddev->degraded++;
1029
		set_bit(Faulty, &rdev->flags);
1030
		spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1031 1032 1033
		/*
		 * if recovery is running, make sure it aborts.
		 */
1034
		set_bit(MD_RECOVERY_INTR, &mddev->recovery);
1035 1036
	} else
		set_bit(Faulty, &rdev->flags);
1037
	set_bit(MD_CHANGE_DEVS, &mddev->flags);
1038 1039
	printk(KERN_ALERT "raid1: Disk failure on %s, disabling device.\n"
		"raid1: Operation continuing on %d devices.\n",
1040
		bdevname(rdev->bdev,b), conf->raid_disks - mddev->degraded);
L
Linus Torvalds 已提交
1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051
}

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

	printk("RAID1 conf printout:\n");
	if (!conf) {
		printk("(!conf)\n");
		return;
	}
1052
	printk(" --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
L
Linus Torvalds 已提交
1053 1054
		conf->raid_disks);

1055
	rcu_read_lock();
L
Linus Torvalds 已提交
1056 1057
	for (i = 0; i < conf->raid_disks; i++) {
		char b[BDEVNAME_SIZE];
1058 1059
		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
		if (rdev)
L
Linus Torvalds 已提交
1060
			printk(" disk %d, wo:%d, o:%d, dev:%s\n",
1061 1062 1063
			       i, !test_bit(In_sync, &rdev->flags),
			       !test_bit(Faulty, &rdev->flags),
			       bdevname(rdev->bdev,b));
L
Linus Torvalds 已提交
1064
	}
1065
	rcu_read_unlock();
L
Linus Torvalds 已提交
1066 1067 1068 1069
}

static void close_sync(conf_t *conf)
{
1070 1071
	wait_barrier(conf);
	allow_barrier(conf);
L
Linus Torvalds 已提交
1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083

	mempool_destroy(conf->r1buf_pool);
	conf->r1buf_pool = NULL;
}

static int raid1_spare_active(mddev_t *mddev)
{
	int i;
	conf_t *conf = mddev->private;

	/*
	 * Find all failed disks within the RAID1 configuration 
1084 1085
	 * and mark them readable.
	 * Called under mddev lock, so rcu protection not needed.
L
Linus Torvalds 已提交
1086 1087
	 */
	for (i = 0; i < conf->raid_disks; i++) {
1088 1089 1090
		mdk_rdev_t *rdev = conf->mirrors[i].rdev;
		if (rdev
		    && !test_bit(Faulty, &rdev->flags)
1091 1092 1093
		    && !test_and_set_bit(In_sync, &rdev->flags)) {
			unsigned long flags;
			spin_lock_irqsave(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1094
			mddev->degraded--;
1095
			spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106
		}
	}

	print_conf(conf);
	return 0;
}


static int raid1_add_disk(mddev_t *mddev, mdk_rdev_t *rdev)
{
	conf_t *conf = mddev->private;
1107
	int err = -EEXIST;
1108
	int mirror = 0;
L
Linus Torvalds 已提交
1109
	mirror_info_t *p;
1110 1111
	int first = 0;
	int last = mddev->raid_disks - 1;
L
Linus Torvalds 已提交
1112

1113 1114 1115 1116
	if (rdev->raid_disk >= 0)
		first = last = rdev->raid_disk;

	for (mirror = first; mirror <= last; mirror++)
L
Linus Torvalds 已提交
1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130
		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))
				blk_queue_max_sectors(mddev->queue, PAGE_SIZE>>9);

			p->head_position = 0;
			rdev->raid_disk = mirror;
1131
			err = 0;
1132 1133 1134 1135
			/* As all devices are equivalent, we don't need a full recovery
			 * if this was recently any drive of the array
			 */
			if (rdev->saved_raid_disk < 0)
1136
				conf->fullsync = 1;
1137
			rcu_assign_pointer(p->rdev, rdev);
L
Linus Torvalds 已提交
1138 1139 1140 1141
			break;
		}

	print_conf(conf);
1142
	return err;
L
Linus Torvalds 已提交
1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154
}

static int raid1_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) {
1155
		if (test_bit(In_sync, &rdev->flags) ||
L
Linus Torvalds 已提交
1156 1157 1158 1159
		    atomic_read(&rdev->nr_pending)) {
			err = -EBUSY;
			goto abort;
		}
1160 1161 1162 1163 1164 1165 1166 1167
		/* Only remove non-faulty devices is recovery
		 * is not possible.
		 */
		if (!test_bit(Faulty, &rdev->flags) &&
		    mddev->degraded < conf->raid_disks) {
			err = -EBUSY;
			goto abort;
		}
L
Linus Torvalds 已提交
1168
		p->rdev = NULL;
1169
		synchronize_rcu();
L
Linus Torvalds 已提交
1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182
		if (atomic_read(&rdev->nr_pending)) {
			/* lost the race, try later */
			err = -EBUSY;
			p->rdev = rdev;
		}
	}
abort:

	print_conf(conf);
	return err;
}


1183
static void end_sync_read(struct bio *bio, int error)
L
Linus Torvalds 已提交
1184 1185
{
	r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private);
1186
	int i;
L
Linus Torvalds 已提交
1187

1188 1189 1190 1191 1192
	for (i=r1_bio->mddev->raid_disks; i--; )
		if (r1_bio->bios[i] == bio)
			break;
	BUG_ON(i < 0);
	update_head_pos(i, r1_bio);
L
Linus Torvalds 已提交
1193 1194 1195 1196 1197
	/*
	 * we have read a block, now it needs to be re-written,
	 * or re-read if the read failed.
	 * We don't do much here, just schedule handling by raid1d
	 */
1198
	if (test_bit(BIO_UPTODATE, &bio->bi_flags))
L
Linus Torvalds 已提交
1199
		set_bit(R1BIO_Uptodate, &r1_bio->state);
1200 1201 1202

	if (atomic_dec_and_test(&r1_bio->remaining))
		reschedule_retry(r1_bio);
L
Linus Torvalds 已提交
1203 1204
}

1205
static void end_sync_write(struct bio *bio, int error)
L
Linus Torvalds 已提交
1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
	r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private);
	mddev_t *mddev = r1_bio->mddev;
	conf_t *conf = mddev_to_conf(mddev);
	int i;
	int mirror=0;

	for (i = 0; i < conf->raid_disks; i++)
		if (r1_bio->bios[i] == bio) {
			mirror = i;
			break;
		}
1219 1220 1221 1222 1223 1224
	if (!uptodate) {
		int sync_blocks = 0;
		sector_t s = r1_bio->sector;
		long sectors_to_go = r1_bio->sectors;
		/* make sure these bits doesn't get cleared. */
		do {
1225
			bitmap_end_sync(mddev->bitmap, s,
1226 1227 1228 1229
					&sync_blocks, 1);
			s += sync_blocks;
			sectors_to_go -= sync_blocks;
		} while (sectors_to_go > 0);
L
Linus Torvalds 已提交
1230
		md_error(mddev, conf->mirrors[mirror].rdev);
1231
	}
1232

L
Linus Torvalds 已提交
1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249
	update_head_pos(mirror, r1_bio);

	if (atomic_dec_and_test(&r1_bio->remaining)) {
		md_done_sync(mddev, r1_bio->sectors, uptodate);
		put_buf(r1_bio);
	}
}

static void sync_request_write(mddev_t *mddev, r1bio_t *r1_bio)
{
	conf_t *conf = mddev_to_conf(mddev);
	int i;
	int disks = conf->raid_disks;
	struct bio *bio, *wbio;

	bio = r1_bio->bios[r1_bio->read_disk];

1250

1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272
	if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
		/* We have read all readable devices.  If we haven't
		 * got the block, then there is no hope left.
		 * If we have, then we want to do a comparison
		 * and skip the write if everything is the same.
		 * If any blocks failed to read, then we need to
		 * attempt an over-write
		 */
		int primary;
		if (!test_bit(R1BIO_Uptodate, &r1_bio->state)) {
			for (i=0; i<mddev->raid_disks; i++)
				if (r1_bio->bios[i]->bi_end_io == end_sync_read)
					md_error(mddev, conf->mirrors[i].rdev);

			md_done_sync(mddev, r1_bio->sectors, 1);
			put_buf(r1_bio);
			return;
		}
		for (primary=0; primary<mddev->raid_disks; primary++)
			if (r1_bio->bios[primary]->bi_end_io == end_sync_read &&
			    test_bit(BIO_UPTODATE, &r1_bio->bios[primary]->bi_flags)) {
				r1_bio->bios[primary]->bi_end_io = NULL;
1273
				rdev_dec_pending(conf->mirrors[primary].rdev, mddev);
1274 1275 1276 1277
				break;
			}
		r1_bio->read_disk = primary;
		for (i=0; i<mddev->raid_disks; i++)
1278
			if (r1_bio->bios[i]->bi_end_io == end_sync_read) {
1279 1280 1281 1282
				int j;
				int vcnt = r1_bio->sectors >> (PAGE_SHIFT- 9);
				struct bio *pbio = r1_bio->bios[primary];
				struct bio *sbio = r1_bio->bios[i];
1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295

				if (test_bit(BIO_UPTODATE, &sbio->bi_flags)) {
					for (j = vcnt; j-- ; ) {
						struct page *p, *s;
						p = pbio->bi_io_vec[j].bv_page;
						s = sbio->bi_io_vec[j].bv_page;
						if (memcmp(page_address(p),
							   page_address(s),
							   PAGE_SIZE))
							break;
					}
				} else
					j = 0;
1296 1297
				if (j >= 0)
					mddev->resync_mismatches += r1_bio->sectors;
1298 1299
				if (j < 0 || (test_bit(MD_RECOVERY_CHECK, &mddev->recovery)
					      && test_bit(BIO_UPTODATE, &sbio->bi_flags))) {
1300
					sbio->bi_end_io = NULL;
1301 1302
					rdev_dec_pending(conf->mirrors[i].rdev, mddev);
				} else {
1303
					/* fixup the bio for reuse */
1304
					int size;
1305 1306 1307 1308 1309 1310 1311 1312 1313 1314
					sbio->bi_vcnt = vcnt;
					sbio->bi_size = r1_bio->sectors << 9;
					sbio->bi_idx = 0;
					sbio->bi_phys_segments = 0;
					sbio->bi_flags &= ~(BIO_POOL_MASK - 1);
					sbio->bi_flags |= 1 << BIO_UPTODATE;
					sbio->bi_next = NULL;
					sbio->bi_sector = r1_bio->sector +
						conf->mirrors[i].rdev->data_offset;
					sbio->bi_bdev = conf->mirrors[i].rdev->bdev;
1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325
					size = sbio->bi_size;
					for (j = 0; j < vcnt ; j++) {
						struct bio_vec *bi;
						bi = &sbio->bi_io_vec[j];
						bi->bv_offset = 0;
						if (size > PAGE_SIZE)
							bi->bv_len = PAGE_SIZE;
						else
							bi->bv_len = size;
						size -= PAGE_SIZE;
						memcpy(page_address(bi->bv_page),
1326 1327
						       page_address(pbio->bi_io_vec[j].bv_page),
						       PAGE_SIZE);
1328
					}
1329

1330 1331 1332
				}
			}
	}
L
Linus Torvalds 已提交
1333
	if (!test_bit(R1BIO_Uptodate, &r1_bio->state)) {
1334 1335 1336
		/* ouch - failed to read all of that.
		 * Try some synchronous reads of other devices to get
		 * good data, much like with normal read errors.  Only
1337
		 * read into the pages we already have so we don't
1338 1339 1340 1341
		 * need to re-issue the read request.
		 * We don't need to freeze the array, because being in an
		 * active sync request, there is no normal IO, and
		 * no overlapping syncs.
L
Linus Torvalds 已提交
1342
		 */
1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356
		sector_t sect = r1_bio->sector;
		int sectors = r1_bio->sectors;
		int idx = 0;

		while(sectors) {
			int s = sectors;
			int d = r1_bio->read_disk;
			int success = 0;
			mdk_rdev_t *rdev;

			if (s > (PAGE_SIZE>>9))
				s = PAGE_SIZE >> 9;
			do {
				if (r1_bio->bios[d]->bi_end_io == end_sync_read) {
1357 1358 1359 1360
					/* No rcu protection needed here devices
					 * can only be removed when no resync is
					 * active, and resync is currently active
					 */
1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376
					rdev = conf->mirrors[d].rdev;
					if (sync_page_io(rdev->bdev,
							 sect + rdev->data_offset,
							 s<<9,
							 bio->bi_io_vec[idx].bv_page,
							 READ)) {
						success = 1;
						break;
					}
				}
				d++;
				if (d == conf->raid_disks)
					d = 0;
			} while (!success && d != r1_bio->read_disk);

			if (success) {
1377
				int start = d;
1378 1379 1380 1381 1382 1383 1384 1385 1386
				/* write it back and re-read */
				set_bit(R1BIO_Uptodate, &r1_bio->state);
				while (d != r1_bio->read_disk) {
					if (d == 0)
						d = conf->raid_disks;
					d--;
					if (r1_bio->bios[d]->bi_end_io != end_sync_read)
						continue;
					rdev = conf->mirrors[d].rdev;
1387
					atomic_add(s, &rdev->corrected_errors);
1388 1389 1390 1391
					if (sync_page_io(rdev->bdev,
							 sect + rdev->data_offset,
							 s<<9,
							 bio->bi_io_vec[idx].bv_page,
1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403
							 WRITE) == 0)
						md_error(mddev, rdev);
				}
				d = start;
				while (d != r1_bio->read_disk) {
					if (d == 0)
						d = conf->raid_disks;
					d--;
					if (r1_bio->bios[d]->bi_end_io != end_sync_read)
						continue;
					rdev = conf->mirrors[d].rdev;
					if (sync_page_io(rdev->bdev,
1404 1405 1406
							 sect + rdev->data_offset,
							 s<<9,
							 bio->bi_io_vec[idx].bv_page,
1407
							 READ) == 0)
1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425
						md_error(mddev, rdev);
				}
			} else {
				char b[BDEVNAME_SIZE];
				/* Cannot read from anywhere, array is toast */
				md_error(mddev, conf->mirrors[r1_bio->read_disk].rdev);
				printk(KERN_ALERT "raid1: %s: unrecoverable I/O read error"
				       " for block %llu\n",
				       bdevname(bio->bi_bdev,b),
				       (unsigned long long)r1_bio->sector);
				md_done_sync(mddev, r1_bio->sectors, 0);
				put_buf(r1_bio);
				return;
			}
			sectors -= s;
			sect += s;
			idx ++;
		}
L
Linus Torvalds 已提交
1426
	}
1427 1428 1429 1430

	/*
	 * schedule writes
	 */
L
Linus Torvalds 已提交
1431 1432 1433
	atomic_set(&r1_bio->remaining, 1);
	for (i = 0; i < disks ; i++) {
		wbio = r1_bio->bios[i];
1434 1435 1436 1437
		if (wbio->bi_end_io == NULL ||
		    (wbio->bi_end_io == end_sync_read &&
		     (i == r1_bio->read_disk ||
		      !test_bit(MD_RECOVERY_SYNC, &mddev->recovery))))
L
Linus Torvalds 已提交
1438 1439
			continue;

1440 1441
		wbio->bi_rw = WRITE;
		wbio->bi_end_io = end_sync_write;
L
Linus Torvalds 已提交
1442 1443
		atomic_inc(&r1_bio->remaining);
		md_sync_acct(conf->mirrors[i].rdev->bdev, wbio->bi_size >> 9);
1444

L
Linus Torvalds 已提交
1445 1446 1447 1448
		generic_make_request(wbio);
	}

	if (atomic_dec_and_test(&r1_bio->remaining)) {
1449
		/* if we're here, all write(s) have completed, so clean up */
L
Linus Torvalds 已提交
1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462
		md_done_sync(mddev, r1_bio->sectors, 1);
		put_buf(r1_bio);
	}
}

/*
 * This is a kernel thread which:
 *
 *	1.	Retries failed read operations on working mirrors.
 *	2.	Updates the raid superblock when problems encounter.
 *	3.	Performs writes following reads for array syncronising.
 */

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 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540
static void fix_read_error(conf_t *conf, int read_disk,
			   sector_t sect, int sectors)
{
	mddev_t *mddev = conf->mddev;
	while(sectors) {
		int s = sectors;
		int d = read_disk;
		int success = 0;
		int start;
		mdk_rdev_t *rdev;

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

		do {
			/* Note: no rcu protection needed here
			 * as this is synchronous in the raid1d thread
			 * which is the thread that might remove
			 * a device.  If raid1d ever becomes multi-threaded....
			 */
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
			    test_bit(In_sync, &rdev->flags) &&
			    sync_page_io(rdev->bdev,
					 sect + rdev->data_offset,
					 s<<9,
					 conf->tmppage, READ))
				success = 1;
			else {
				d++;
				if (d == conf->raid_disks)
					d = 0;
			}
		} while (!success && d != read_disk);

		if (!success) {
			/* Cannot read from anywhere -- bye bye array */
			md_error(mddev, conf->mirrors[read_disk].rdev);
			break;
		}
		/* write it back and re-read */
		start = d;
		while (d != read_disk) {
			if (d==0)
				d = conf->raid_disks;
			d--;
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
			    test_bit(In_sync, &rdev->flags)) {
				if (sync_page_io(rdev->bdev,
						 sect + rdev->data_offset,
						 s<<9, conf->tmppage, WRITE)
				    == 0)
					/* Well, this device is dead */
					md_error(mddev, rdev);
			}
		}
		d = start;
		while (d != read_disk) {
			char b[BDEVNAME_SIZE];
			if (d==0)
				d = conf->raid_disks;
			d--;
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
			    test_bit(In_sync, &rdev->flags)) {
				if (sync_page_io(rdev->bdev,
						 sect + rdev->data_offset,
						 s<<9, conf->tmppage, READ)
				    == 0)
					/* Well, this device is dead */
					md_error(mddev, rdev);
				else {
					atomic_add(s, &rdev->corrected_errors);
					printk(KERN_INFO
					       "raid1:%s: read error corrected "
					       "(%d sectors at %llu on %s)\n",
					       mdname(mddev), s,
1541 1542
					       (unsigned long long)(sect +
					           rdev->data_offset),
1543 1544 1545 1546 1547 1548 1549 1550 1551
					       bdevname(rdev->bdev, b));
				}
			}
		}
		sectors -= s;
		sect += s;
	}
}

L
Linus Torvalds 已提交
1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565
static void raid1d(mddev_t *mddev)
{
	r1bio_t *r1_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];
1566

1567
		unplug += flush_pending_writes(conf);
1568

1569 1570 1571
		spin_lock_irqsave(&conf->device_lock, flags);
		if (list_empty(head)) {
			spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1572
			break;
1573
		}
L
Linus Torvalds 已提交
1574 1575
		r1_bio = list_entry(head->prev, r1bio_t, retry_list);
		list_del(head->prev);
1576
		conf->nr_queued--;
L
Linus Torvalds 已提交
1577 1578 1579 1580 1581 1582 1583
		spin_unlock_irqrestore(&conf->device_lock, flags);

		mddev = r1_bio->mddev;
		conf = mddev_to_conf(mddev);
		if (test_bit(R1BIO_IsSync, &r1_bio->state)) {
			sync_request_write(mddev, r1_bio);
			unplug = 1;
1584 1585
		} else if (test_bit(R1BIO_BarrierRetry, &r1_bio->state)) {
			/* some requests in the r1bio were BIO_RW_BARRIER
1586
			 * requests which failed with -EOPNOTSUPP.  Hohumm..
1587 1588 1589
			 * Better resubmit without the barrier.
			 * We know which devices to resubmit for, because
			 * all others have had their bios[] entry cleared.
1590
			 * We already have a nr_pending reference on these rdevs.
1591 1592
			 */
			int i;
1593
			const int do_sync = bio_sync(r1_bio->master_bio);
1594 1595
			clear_bit(R1BIO_BarrierRetry, &r1_bio->state);
			clear_bit(R1BIO_Barrier, &r1_bio->state);
1596 1597 1598
			for (i=0; i < conf->raid_disks; i++)
				if (r1_bio->bios[i])
					atomic_inc(&r1_bio->remaining);
1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613
			for (i=0; i < conf->raid_disks; i++)
				if (r1_bio->bios[i]) {
					struct bio_vec *bvec;
					int j;

					bio = bio_clone(r1_bio->master_bio, GFP_NOIO);
					/* copy pages from the failed bio, as
					 * this might be a write-behind device */
					__bio_for_each_segment(bvec, bio, j, 0)
						bvec->bv_page = bio_iovec_idx(r1_bio->bios[i], j)->bv_page;
					bio_put(r1_bio->bios[i]);
					bio->bi_sector = r1_bio->sector +
						conf->mirrors[i].rdev->data_offset;
					bio->bi_bdev = conf->mirrors[i].rdev->bdev;
					bio->bi_end_io = raid1_end_write_request;
1614
					bio->bi_rw = WRITE | do_sync;
1615 1616 1617 1618
					bio->bi_private = r1_bio;
					r1_bio->bios[i] = bio;
					generic_make_request(bio);
				}
L
Linus Torvalds 已提交
1619 1620
		} else {
			int disk;
1621 1622 1623 1624 1625 1626 1627 1628 1629

			/* 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
			 */
1630 1631 1632 1633 1634 1635
			if (mddev->ro == 0) {
				freeze_array(conf);
				fix_read_error(conf, r1_bio->read_disk,
					       r1_bio->sector,
					       r1_bio->sectors);
				unfreeze_array(conf);
1636 1637
			}

L
Linus Torvalds 已提交
1638 1639 1640 1641 1642 1643 1644 1645
			bio = r1_bio->bios[r1_bio->read_disk];
			if ((disk=read_balance(conf, r1_bio)) == -1) {
				printk(KERN_ALERT "raid1: %s: unrecoverable I/O"
				       " read error for block %llu\n",
				       bdevname(bio->bi_bdev,b),
				       (unsigned long long)r1_bio->sector);
				raid_end_bio_io(r1_bio);
			} else {
1646
				const int do_sync = bio_sync(r1_bio->master_bio);
1647 1648
				r1_bio->bios[r1_bio->read_disk] =
					mddev->ro ? IO_BLOCKED : NULL;
L
Linus Torvalds 已提交
1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661
				r1_bio->read_disk = disk;
				bio_put(bio);
				bio = bio_clone(r1_bio->master_bio, GFP_NOIO);
				r1_bio->bios[r1_bio->read_disk] = bio;
				rdev = conf->mirrors[disk].rdev;
				if (printk_ratelimit())
					printk(KERN_ERR "raid1: %s: redirecting sector %llu to"
					       " another mirror\n",
					       bdevname(rdev->bdev,b),
					       (unsigned long long)r1_bio->sector);
				bio->bi_sector = r1_bio->sector + rdev->data_offset;
				bio->bi_bdev = rdev->bdev;
				bio->bi_end_io = raid1_end_read_request;
1662
				bio->bi_rw = READ | do_sync;
L
Linus Torvalds 已提交
1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678
				bio->bi_private = r1_bio;
				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;
1679
	BUG_ON(conf->r1buf_pool);
L
Linus Torvalds 已提交
1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697
	conf->r1buf_pool = mempool_create(buffs, r1buf_pool_alloc, r1buf_pool_free,
					  conf->poolinfo);
	if (!conf->r1buf_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.
 */

1698
static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster)
L
Linus Torvalds 已提交
1699 1700 1701 1702 1703
{
	conf_t *conf = mddev_to_conf(mddev);
	r1bio_t *r1_bio;
	struct bio *bio;
	sector_t max_sector, nr_sectors;
1704
	int disk = -1;
L
Linus Torvalds 已提交
1705
	int i;
1706 1707
	int wonly = -1;
	int write_targets = 0, read_targets = 0;
1708
	int sync_blocks;
1709
	int still_degraded = 0;
L
Linus Torvalds 已提交
1710 1711

	if (!conf->r1buf_pool)
1712 1713 1714 1715
	{
/*
		printk("sync start - bitmap %p\n", mddev->bitmap);
*/
L
Linus Torvalds 已提交
1716
		if (init_resync(conf))
1717
			return 0;
1718
	}
L
Linus Torvalds 已提交
1719 1720 1721

	max_sector = mddev->size << 1;
	if (sector_nr >= max_sector) {
1722 1723 1724 1725 1726
		/* If we aborted, we need to abort the
		 * sync on the 'current' bitmap chunk (there will
		 * only be one in raid1 resync.
		 * We can find the current addess in mddev->curr_resync
		 */
1727 1728
		if (mddev->curr_resync < max_sector) /* aborted */
			bitmap_end_sync(mddev->bitmap, mddev->curr_resync,
1729
						&sync_blocks, 1);
1730
		else /* completed sync */
1731
			conf->fullsync = 0;
1732 1733

		bitmap_close_sync(mddev->bitmap);
L
Linus Torvalds 已提交
1734 1735 1736 1737
		close_sync(conf);
		return 0;
	}

1738 1739
	if (mddev->bitmap == NULL &&
	    mddev->recovery_cp == MaxSector &&
1740
	    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
1741 1742 1743 1744
	    conf->fullsync == 0) {
		*skipped = 1;
		return max_sector - sector_nr;
	}
1745 1746 1747
	/* before building a request, check if we can skip these blocks..
	 * This call the bitmap_start_sync doesn't actually record anything
	 */
1748
	if (!bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, 1) &&
1749
	    !conf->fullsync && !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
1750 1751 1752 1753
		/* We can skip this block, and probably several more */
		*skipped = 1;
		return sync_blocks;
	}
L
Linus Torvalds 已提交
1754
	/*
1755 1756 1757
	 * If there is non-resync activity waiting for a turn,
	 * and resync is going fast enough,
	 * then let it though before starting on this new sync request.
L
Linus Torvalds 已提交
1758
	 */
1759
	if (!go_faster && conf->nr_waiting)
L
Linus Torvalds 已提交
1760
		msleep_interruptible(1000);
1761

N
NeilBrown 已提交
1762
	bitmap_cond_end_sync(mddev->bitmap, sector_nr);
1763 1764 1765
	raise_barrier(conf);

	conf->next_resync = sector_nr;
L
Linus Torvalds 已提交
1766

1767 1768
	r1_bio = mempool_alloc(conf->r1buf_pool, GFP_NOIO);
	rcu_read_lock();
L
Linus Torvalds 已提交
1769
	/*
1770 1771 1772 1773 1774 1775
	 * If we get a correctably read error during resync or recovery,
	 * we might want to read from a different device.  So we
	 * flag all drives that could conceivably be read from for READ,
	 * and any others (which will be non-In_sync devices) for WRITE.
	 * If a read fails, we try reading from something else for which READ
	 * is OK.
L
Linus Torvalds 已提交
1776 1777 1778 1779
	 */

	r1_bio->mddev = mddev;
	r1_bio->sector = sector_nr;
1780
	r1_bio->state = 0;
L
Linus Torvalds 已提交
1781 1782 1783
	set_bit(R1BIO_IsSync, &r1_bio->state);

	for (i=0; i < conf->raid_disks; i++) {
1784
		mdk_rdev_t *rdev;
L
Linus Torvalds 已提交
1785 1786 1787 1788 1789
		bio = r1_bio->bios[i];

		/* take from bio_init */
		bio->bi_next = NULL;
		bio->bi_flags |= 1 << BIO_UPTODATE;
1790
		bio->bi_rw = READ;
L
Linus Torvalds 已提交
1791 1792 1793 1794 1795 1796 1797
		bio->bi_vcnt = 0;
		bio->bi_idx = 0;
		bio->bi_phys_segments = 0;
		bio->bi_size = 0;
		bio->bi_end_io = NULL;
		bio->bi_private = NULL;

1798 1799 1800
		rdev = rcu_dereference(conf->mirrors[i].rdev);
		if (rdev == NULL ||
			   test_bit(Faulty, &rdev->flags)) {
1801 1802
			still_degraded = 1;
			continue;
1803
		} else if (!test_bit(In_sync, &rdev->flags)) {
L
Linus Torvalds 已提交
1804 1805 1806
			bio->bi_rw = WRITE;
			bio->bi_end_io = end_sync_write;
			write_targets ++;
1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822
		} else {
			/* may need to read from here */
			bio->bi_rw = READ;
			bio->bi_end_io = end_sync_read;
			if (test_bit(WriteMostly, &rdev->flags)) {
				if (wonly < 0)
					wonly = i;
			} else {
				if (disk < 0)
					disk = i;
			}
			read_targets++;
		}
		atomic_inc(&rdev->nr_pending);
		bio->bi_sector = sector_nr + rdev->data_offset;
		bio->bi_bdev = rdev->bdev;
L
Linus Torvalds 已提交
1823 1824
		bio->bi_private = r1_bio;
	}
1825 1826 1827 1828
	rcu_read_unlock();
	if (disk < 0)
		disk = wonly;
	r1_bio->read_disk = disk;
1829

1830 1831 1832 1833 1834
	if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery) && read_targets > 0)
		/* extra read targets are also write targets */
		write_targets += read_targets-1;

	if (write_targets == 0 || read_targets == 0) {
L
Linus Torvalds 已提交
1835 1836 1837
		/* There is nowhere to write, so all non-sync
		 * drives must be failed - so we are finished
		 */
1838 1839
		sector_t rv = max_sector - sector_nr;
		*skipped = 1;
L
Linus Torvalds 已提交
1840 1841 1842 1843
		put_buf(r1_bio);
		return rv;
	}

1844 1845
	if (max_sector > mddev->resync_max)
		max_sector = mddev->resync_max; /* Don't do IO beyond here */
L
Linus Torvalds 已提交
1846
	nr_sectors = 0;
1847
	sync_blocks = 0;
L
Linus Torvalds 已提交
1848 1849 1850 1851 1852 1853 1854
	do {
		struct page *page;
		int len = PAGE_SIZE;
		if (sector_nr + (len>>9) > max_sector)
			len = (max_sector - sector_nr) << 9;
		if (len == 0)
			break;
1855 1856
		if (sync_blocks == 0) {
			if (!bitmap_start_sync(mddev->bitmap, sector_nr,
1857 1858 1859
					       &sync_blocks, still_degraded) &&
			    !conf->fullsync &&
			    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
1860
				break;
1861
			BUG_ON(sync_blocks < (PAGE_SIZE>>9));
1862 1863
			if (len > (sync_blocks<<9))
				len = sync_blocks<<9;
1864
		}
1865

L
Linus Torvalds 已提交
1866 1867 1868
		for (i=0 ; i < conf->raid_disks; i++) {
			bio = r1_bio->bios[i];
			if (bio->bi_end_io) {
1869
				page = bio->bi_io_vec[bio->bi_vcnt].bv_page;
L
Linus Torvalds 已提交
1870 1871
				if (bio_add_page(bio, page, len, 0) == 0) {
					/* stop here */
1872
					bio->bi_io_vec[bio->bi_vcnt].bv_page = page;
L
Linus Torvalds 已提交
1873 1874 1875
					while (i > 0) {
						i--;
						bio = r1_bio->bios[i];
1876 1877
						if (bio->bi_end_io==NULL)
							continue;
L
Linus Torvalds 已提交
1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888
						/* remove last page from this bio */
						bio->bi_vcnt--;
						bio->bi_size -= len;
						bio->bi_flags &= ~(1<< BIO_SEG_VALID);
					}
					goto bio_full;
				}
			}
		}
		nr_sectors += len>>9;
		sector_nr += len>>9;
1889
		sync_blocks -= (len>>9);
L
Linus Torvalds 已提交
1890 1891 1892 1893
	} while (r1_bio->bios[disk]->bi_vcnt < RESYNC_PAGES);
 bio_full:
	r1_bio->sectors = nr_sectors;

1894 1895 1896 1897 1898 1899 1900 1901
	/* For a user-requested sync, we read all readable devices and do a
	 * compare
	 */
	if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
		atomic_set(&r1_bio->remaining, read_targets);
		for (i=0; i<conf->raid_disks; i++) {
			bio = r1_bio->bios[i];
			if (bio->bi_end_io == end_sync_read) {
1902
				md_sync_acct(bio->bi_bdev, nr_sectors);
1903 1904 1905 1906 1907 1908
				generic_make_request(bio);
			}
		}
	} else {
		atomic_set(&r1_bio->remaining, 1);
		bio = r1_bio->bios[r1_bio->read_disk];
1909
		md_sync_acct(bio->bi_bdev, nr_sectors);
1910
		generic_make_request(bio);
L
Linus Torvalds 已提交
1911

1912
	}
L
Linus Torvalds 已提交
1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927
	return nr_sectors;
}

static int run(mddev_t *mddev)
{
	conf_t *conf;
	int i, j, disk_idx;
	mirror_info_t *disk;
	mdk_rdev_t *rdev;

	if (mddev->level != 1) {
		printk("raid1: %s: raid level not set to mirroring (%d)\n",
		       mdname(mddev), mddev->level);
		goto out;
	}
1928 1929 1930 1931 1932
	if (mddev->reshape_position != MaxSector) {
		printk("raid1: %s: reshape_position set but not supported\n",
		       mdname(mddev));
		goto out;
	}
L
Linus Torvalds 已提交
1933 1934 1935 1936 1937
	/*
	 * copy the already verified devices into our private RAID1
	 * bookkeeping area. [whatever we allocate in run(),
	 * should be freed in stop()]
	 */
1938
	conf = kzalloc(sizeof(conf_t), GFP_KERNEL);
L
Linus Torvalds 已提交
1939 1940 1941 1942
	mddev->private = conf;
	if (!conf)
		goto out_no_mem;

1943
	conf->mirrors = kzalloc(sizeof(struct mirror_info)*mddev->raid_disks,
L
Linus Torvalds 已提交
1944 1945 1946 1947
				 GFP_KERNEL);
	if (!conf->mirrors)
		goto out_no_mem;

1948 1949 1950 1951
	conf->tmppage = alloc_page(GFP_KERNEL);
	if (!conf->tmppage)
		goto out_no_mem;

L
Linus Torvalds 已提交
1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962
	conf->poolinfo = kmalloc(sizeof(*conf->poolinfo), GFP_KERNEL);
	if (!conf->poolinfo)
		goto out_no_mem;
	conf->poolinfo->mddev = mddev;
	conf->poolinfo->raid_disks = mddev->raid_disks;
	conf->r1bio_pool = mempool_create(NR_RAID1_BIOS, r1bio_pool_alloc,
					  r1bio_pool_free,
					  conf->poolinfo);
	if (!conf->r1bio_pool)
		goto out_no_mem;

1963 1964 1965
	spin_lock_init(&conf->device_lock);
	mddev->queue->queue_lock = &conf->device_lock;

1966
	list_for_each_entry(rdev, &mddev->disks, same_set) {
L
Linus Torvalds 已提交
1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991
		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))
			blk_queue_max_sectors(mddev->queue, PAGE_SIZE>>9);

		disk->head_position = 0;
	}
	conf->raid_disks = mddev->raid_disks;
	conf->mddev = mddev;
	INIT_LIST_HEAD(&conf->retry_list);

	spin_lock_init(&conf->resync_lock);
1992
	init_waitqueue_head(&conf->wait_barrier);
L
Linus Torvalds 已提交
1993

1994 1995 1996
	bio_list_init(&conf->pending_bio_list);
	bio_list_init(&conf->flushing_bio_list);

L
Linus Torvalds 已提交
1997 1998 1999 2000 2001 2002

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

		disk = conf->mirrors + i;

2003 2004
		if (!disk->rdev ||
		    !test_bit(In_sync, &disk->rdev->flags)) {
L
Linus Torvalds 已提交
2005 2006
			disk->head_position = 0;
			mddev->degraded++;
2007 2008
			if (disk->rdev)
				conf->fullsync = 1;
L
Linus Torvalds 已提交
2009 2010
		}
	}
2011 2012 2013 2014 2015 2016 2017
	if (mddev->degraded == conf->raid_disks) {
		printk(KERN_ERR "raid1: no operational mirrors for %s\n",
			mdname(mddev));
		goto out_free_conf;
	}
	if (conf->raid_disks - mddev->degraded == 1)
		mddev->recovery_cp = MaxSector;
L
Linus Torvalds 已提交
2018 2019 2020 2021 2022 2023 2024

	/*
	 * find the first working one and use it as a starting point
	 * to read balancing.
	 */
	for (j = 0; j < conf->raid_disks &&
		     (!conf->mirrors[j].rdev ||
2025
		      !test_bit(In_sync, &conf->mirrors[j].rdev->flags)) ; j++)
L
Linus Torvalds 已提交
2026 2027 2028 2029
		/* nothing */;
	conf->last_used = j;


2030 2031 2032 2033 2034 2035
	mddev->thread = md_register_thread(raid1d, mddev, "%s_raid1");
	if (!mddev->thread) {
		printk(KERN_ERR
		       "raid1: couldn't allocate thread for %s\n",
		       mdname(mddev));
		goto out_free_conf;
L
Linus Torvalds 已提交
2036
	}
2037

L
Linus Torvalds 已提交
2038 2039 2040 2041 2042 2043 2044
	printk(KERN_INFO 
		"raid1: raid set %s active with %d out of %d mirrors\n",
		mdname(mddev), mddev->raid_disks - mddev->degraded, 
		mddev->raid_disks);
	/*
	 * Ok, everything is just fine now
	 */
2045
	mddev->array_sectors = mddev->size * 2;
L
Linus Torvalds 已提交
2046

2047
	mddev->queue->unplug_fn = raid1_unplug;
2048 2049
	mddev->queue->backing_dev_info.congested_fn = raid1_congested;
	mddev->queue->backing_dev_info.congested_data = mddev;
2050

L
Linus Torvalds 已提交
2051 2052 2053 2054 2055 2056 2057 2058 2059 2060
	return 0;

out_no_mem:
	printk(KERN_ERR "raid1: couldn't allocate memory for %s\n",
	       mdname(mddev));

out_free_conf:
	if (conf) {
		if (conf->r1bio_pool)
			mempool_destroy(conf->r1bio_pool);
2061
		kfree(conf->mirrors);
2062
		safe_put_page(conf->tmppage);
2063
		kfree(conf->poolinfo);
L
Linus Torvalds 已提交
2064 2065 2066 2067 2068 2069 2070 2071 2072 2073
		kfree(conf);
		mddev->private = NULL;
	}
out:
	return -EIO;
}

static int stop(mddev_t *mddev)
{
	conf_t *conf = mddev_to_conf(mddev);
2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084
	struct bitmap *bitmap = mddev->bitmap;
	int behind_wait = 0;

	/* wait for behind writes to complete */
	while (bitmap && atomic_read(&bitmap->behind_writes) > 0) {
		behind_wait++;
		printk(KERN_INFO "raid1: behind writes in progress on device %s, waiting to stop (%d)\n", mdname(mddev), behind_wait);
		set_current_state(TASK_UNINTERRUPTIBLE);
		schedule_timeout(HZ); /* wait a second */
		/* need to kick something here to make sure I/O goes? */
	}
L
Linus Torvalds 已提交
2085 2086 2087 2088 2089 2090

	md_unregister_thread(mddev->thread);
	mddev->thread = NULL;
	blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
	if (conf->r1bio_pool)
		mempool_destroy(conf->r1bio_pool);
2091 2092
	kfree(conf->mirrors);
	kfree(conf->poolinfo);
L
Linus Torvalds 已提交
2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106
	kfree(conf);
	mddev->private = NULL;
	return 0;
}

static int raid1_resize(mddev_t *mddev, sector_t sectors)
{
	/* no resync is happening, and there is enough space
	 * on all devices, so we can resize.
	 * We need to make sure resync covers any new space.
	 * If the array is shrinking we should possibly wait until
	 * any io in the removed space completes, but it hardly seems
	 * worth it.
	 */
2107 2108
	mddev->array_sectors = sectors;
	set_capacity(mddev->gendisk, mddev->array_sectors);
2109
	mddev->changed = 1;
2110 2111
	if (mddev->array_sectors / 2 > mddev->size &&
	    mddev->recovery_cp == MaxSector) {
L
Linus Torvalds 已提交
2112 2113 2114
		mddev->recovery_cp = mddev->size << 1;
		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
	}
2115
	mddev->size = mddev->array_sectors / 2;
2116
	mddev->resync_max_sectors = sectors;
L
Linus Torvalds 已提交
2117 2118 2119
	return 0;
}

2120
static int raid1_reshape(mddev_t *mddev)
L
Linus Torvalds 已提交
2121 2122 2123 2124 2125 2126 2127 2128
{
	/* We need to:
	 * 1/ resize the r1bio_pool
	 * 2/ resize conf->mirrors
	 *
	 * We allocate a new r1bio_pool if we can.
	 * Then raise a device barrier and wait until all IO stops.
	 * Then resize conf->mirrors and swap in the new r1bio pool.
2129 2130 2131
	 *
	 * At the same time, we "pack" the devices so that all the missing
	 * devices have the higher raid_disk numbers.
L
Linus Torvalds 已提交
2132 2133 2134 2135 2136
	 */
	mempool_t *newpool, *oldpool;
	struct pool_info *newpoolinfo;
	mirror_info_t *newmirrors;
	conf_t *conf = mddev_to_conf(mddev);
2137
	int cnt, raid_disks;
2138
	unsigned long flags;
2139
	int d, d2, err;
L
Linus Torvalds 已提交
2140

2141 2142 2143 2144 2145 2146 2147 2148 2149 2150
	/* Cannot change chunk_size, layout, or level */
	if (mddev->chunk_size != mddev->new_chunk ||
	    mddev->layout != mddev->new_layout ||
	    mddev->level != mddev->new_level) {
		mddev->new_chunk = mddev->chunk_size;
		mddev->new_layout = mddev->layout;
		mddev->new_level = mddev->level;
		return -EINVAL;
	}

2151 2152 2153
	err = md_allow_write(mddev);
	if (err)
		return err;
2154

2155 2156
	raid_disks = mddev->raid_disks + mddev->delta_disks;

2157 2158 2159 2160 2161 2162
	if (raid_disks < conf->raid_disks) {
		cnt=0;
		for (d= 0; d < conf->raid_disks; d++)
			if (conf->mirrors[d].rdev)
				cnt++;
		if (cnt > raid_disks)
L
Linus Torvalds 已提交
2163
			return -EBUSY;
2164
	}
L
Linus Torvalds 已提交
2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177

	newpoolinfo = kmalloc(sizeof(*newpoolinfo), GFP_KERNEL);
	if (!newpoolinfo)
		return -ENOMEM;
	newpoolinfo->mddev = mddev;
	newpoolinfo->raid_disks = raid_disks;

	newpool = mempool_create(NR_RAID1_BIOS, r1bio_pool_alloc,
				 r1bio_pool_free, newpoolinfo);
	if (!newpool) {
		kfree(newpoolinfo);
		return -ENOMEM;
	}
2178
	newmirrors = kzalloc(sizeof(struct mirror_info) * raid_disks, GFP_KERNEL);
L
Linus Torvalds 已提交
2179 2180 2181 2182 2183 2184
	if (!newmirrors) {
		kfree(newpoolinfo);
		mempool_destroy(newpool);
		return -ENOMEM;
	}

2185
	raise_barrier(conf);
L
Linus Torvalds 已提交
2186 2187 2188 2189

	/* ok, everything is stopped */
	oldpool = conf->r1bio_pool;
	conf->r1bio_pool = newpool;
2190

2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205
	for (d = d2 = 0; d < conf->raid_disks; d++) {
		mdk_rdev_t *rdev = conf->mirrors[d].rdev;
		if (rdev && rdev->raid_disk != d2) {
			char nm[20];
			sprintf(nm, "rd%d", rdev->raid_disk);
			sysfs_remove_link(&mddev->kobj, nm);
			rdev->raid_disk = d2;
			sprintf(nm, "rd%d", rdev->raid_disk);
			sysfs_remove_link(&mddev->kobj, nm);
			if (sysfs_create_link(&mddev->kobj,
					      &rdev->kobj, nm))
				printk(KERN_WARNING
				       "md/raid1: cannot register "
				       "%s for %s\n",
				       nm, mdname(mddev));
2206
		}
2207 2208 2209
		if (rdev)
			newmirrors[d2++].rdev = rdev;
	}
L
Linus Torvalds 已提交
2210 2211 2212 2213 2214
	kfree(conf->mirrors);
	conf->mirrors = newmirrors;
	kfree(conf->poolinfo);
	conf->poolinfo = newpoolinfo;

2215
	spin_lock_irqsave(&conf->device_lock, flags);
L
Linus Torvalds 已提交
2216
	mddev->degraded += (raid_disks - conf->raid_disks);
2217
	spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
2218
	conf->raid_disks = mddev->raid_disks = raid_disks;
2219
	mddev->delta_disks = 0;
L
Linus Torvalds 已提交
2220

2221
	conf->last_used = 0; /* just make sure it is in-range */
2222
	lower_barrier(conf);
L
Linus Torvalds 已提交
2223 2224 2225 2226 2227 2228 2229 2230

	set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
	md_wakeup_thread(mddev->thread);

	mempool_destroy(oldpool);
	return 0;
}

2231
static void raid1_quiesce(mddev_t *mddev, int state)
2232 2233 2234 2235
{
	conf_t *conf = mddev_to_conf(mddev);

	switch(state) {
2236
	case 1:
2237
		raise_barrier(conf);
2238
		break;
2239
	case 0:
2240
		lower_barrier(conf);
2241 2242 2243 2244
		break;
	}
}

L
Linus Torvalds 已提交
2245

2246
static struct mdk_personality raid1_personality =
L
Linus Torvalds 已提交
2247 2248
{
	.name		= "raid1",
2249
	.level		= 1,
L
Linus Torvalds 已提交
2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260
	.owner		= THIS_MODULE,
	.make_request	= make_request,
	.run		= run,
	.stop		= stop,
	.status		= status,
	.error_handler	= error,
	.hot_add_disk	= raid1_add_disk,
	.hot_remove_disk= raid1_remove_disk,
	.spare_active	= raid1_spare_active,
	.sync_request	= sync_request,
	.resize		= raid1_resize,
2261
	.check_reshape	= raid1_reshape,
2262
	.quiesce	= raid1_quiesce,
L
Linus Torvalds 已提交
2263 2264 2265 2266
};

static int __init raid_init(void)
{
2267
	return register_md_personality(&raid1_personality);
L
Linus Torvalds 已提交
2268 2269 2270 2271
}

static void raid_exit(void)
{
2272
	unregister_md_personality(&raid1_personality);
L
Linus Torvalds 已提交
2273 2274 2275 2276 2277 2278
}

module_init(raid_init);
module_exit(raid_exit);
MODULE_LICENSE("GPL");
MODULE_ALIAS("md-personality-3"); /* RAID1 */
2279
MODULE_ALIAS("md-raid1");
2280
MODULE_ALIAS("md-level-1");