raid1.c 72.9 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 <linux/slab.h>
35
#include <linux/delay.h>
36 37
#include <linux/blkdev.h>
#include <linux/seq_file.h>
38
#include <linux/ratelimit.h>
39
#include "md.h"
40 41
#include "raid1.h"
#include "bitmap.h"
42

L
Linus Torvalds 已提交
43 44 45 46 47 48
/*
 * Number of guaranteed r1bios in case of extreme VM load:
 */
#define	NR_RAID1_BIOS 256


49 50
static void allow_barrier(conf_t *conf);
static void lower_barrier(conf_t *conf);
L
Linus Torvalds 已提交
51

A
Al Viro 已提交
52
static void * r1bio_pool_alloc(gfp_t gfp_flags, void *data)
L
Linus Torvalds 已提交
53 54
{
	struct pool_info *pi = data;
55
	int size = offsetof(struct r1bio, bios[pi->raid_disks]);
L
Linus Torvalds 已提交
56 57

	/* allocate a r1bio with room for raid_disks entries in the bios array */
J
Jens Axboe 已提交
58
	return kzalloc(size, gfp_flags);
L
Linus Torvalds 已提交
59 60 61 62 63 64 65 66 67 68 69 70 71
}

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
72
static void * r1buf_pool_alloc(gfp_t gfp_flags, void *data)
L
Linus Torvalds 已提交
73 74 75
{
	struct pool_info *pi = data;
	struct page *page;
76
	struct r1bio *r1_bio;
L
Linus Torvalds 已提交
77 78 79 80
	struct bio *bio;
	int i, j;

	r1_bio = r1bio_pool_alloc(gfp_flags, pi);
J
Jens Axboe 已提交
81
	if (!r1_bio)
L
Linus Torvalds 已提交
82 83 84 85 86 87
		return NULL;

	/*
	 * Allocate bios : 1 for reading, n-1 for writing
	 */
	for (j = pi->raid_disks ; j-- ; ) {
88
		bio = bio_kmalloc(gfp_flags, RESYNC_PAGES);
L
Linus Torvalds 已提交
89 90 91 92 93 94
		if (!bio)
			goto out_free_bio;
		r1_bio->bios[j] = bio;
	}
	/*
	 * Allocate RESYNC_PAGES data pages and attach them to
95 96 97
	 * the first bio.
	 * If this is a user-requested check/repair, allocate
	 * RESYNC_PAGES for each bio.
L
Linus Torvalds 已提交
98
	 */
99 100 101 102 103 104 105 106 107 108 109 110
	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;
111
			bio->bi_vcnt = i+1;
112 113 114 115 116 117 118 119
		}
	}
	/* 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
120 121 122 123 124 125 126
	}

	r1_bio->master_bio = NULL;

	return r1_bio;

out_free_pages:
127 128 129
	for (j=0 ; j < pi->raid_disks; j++)
		for (i=0; i < r1_bio->bios[j]->bi_vcnt ; i++)
			put_page(r1_bio->bios[j]->bi_io_vec[i].bv_page);
130
	j = -1;
L
Linus Torvalds 已提交
131 132 133 134 135 136 137 138 139 140
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;
141
	int i,j;
142
	struct r1bio *r1bio = __r1_bio;
L
Linus Torvalds 已提交
143

144 145 146 147 148
	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)
149
				safe_put_page(r1bio->bios[j]->bi_io_vec[i].bv_page);
150
		}
L
Linus Torvalds 已提交
151 152 153 154 155 156
	for (i=0 ; i < pi->raid_disks; i++)
		bio_put(r1bio->bios[i]);

	r1bio_pool_free(r1bio, data);
}

157
static void put_all_bios(conf_t *conf, struct r1bio *r1_bio)
L
Linus Torvalds 已提交
158 159 160 161 162
{
	int i;

	for (i = 0; i < conf->raid_disks; i++) {
		struct bio **bio = r1_bio->bios + i;
163
		if (!BIO_SPECIAL(*bio))
L
Linus Torvalds 已提交
164 165 166 167 168
			bio_put(*bio);
		*bio = NULL;
	}
}

169
static void free_r1bio(struct r1bio *r1_bio)
L
Linus Torvalds 已提交
170
{
171
	conf_t *conf = r1_bio->mddev->private;
L
Linus Torvalds 已提交
172 173 174 175 176

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

177
static void put_buf(struct r1bio *r1_bio)
L
Linus Torvalds 已提交
178
{
179
	conf_t *conf = r1_bio->mddev->private;
180 181 182 183 184 185 186
	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
187 188 189

	mempool_free(r1_bio, conf->r1buf_pool);

190
	lower_barrier(conf);
L
Linus Torvalds 已提交
191 192
}

193
static void reschedule_retry(struct r1bio *r1_bio)
L
Linus Torvalds 已提交
194 195
{
	unsigned long flags;
196
	struct mddev *mddev = r1_bio->mddev;
197
	conf_t *conf = mddev->private;
L
Linus Torvalds 已提交
198 199 200

	spin_lock_irqsave(&conf->device_lock, flags);
	list_add(&r1_bio->retry_list, &conf->retry_list);
201
	conf->nr_queued ++;
L
Linus Torvalds 已提交
202 203
	spin_unlock_irqrestore(&conf->device_lock, flags);

204
	wake_up(&conf->wait_barrier);
L
Linus Torvalds 已提交
205 206 207 208 209 210 211 212
	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.
 */
213
static void call_bio_endio(struct r1bio *r1_bio)
214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239
{
	struct bio *bio = r1_bio->master_bio;
	int done;
	conf_t *conf = r1_bio->mddev->private;

	if (bio->bi_phys_segments) {
		unsigned long flags;
		spin_lock_irqsave(&conf->device_lock, flags);
		bio->bi_phys_segments--;
		done = (bio->bi_phys_segments == 0);
		spin_unlock_irqrestore(&conf->device_lock, flags);
	} else
		done = 1;

	if (!test_bit(R1BIO_Uptodate, &r1_bio->state))
		clear_bit(BIO_UPTODATE, &bio->bi_flags);
	if (done) {
		bio_endio(bio, 0);
		/*
		 * Wake up any possible resync thread that waits for the device
		 * to go idle.
		 */
		allow_barrier(conf);
	}
}

240
static void raid_end_bio_io(struct r1bio *r1_bio)
L
Linus Torvalds 已提交
241 242 243
{
	struct bio *bio = r1_bio->master_bio;

244 245
	/* if nobody has done the final endio yet, do it now */
	if (!test_and_set_bit(R1BIO_Returned, &r1_bio->state)) {
246 247 248 249 250
		pr_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);
251

252
		call_bio_endio(r1_bio);
253
	}
L
Linus Torvalds 已提交
254 255 256 257 258 259
	free_r1bio(r1_bio);
}

/*
 * Update disk head position estimator based on IRQ completion info.
 */
260
static inline void update_head_pos(int disk, struct r1bio *r1_bio)
L
Linus Torvalds 已提交
261
{
262
	conf_t *conf = r1_bio->mddev->private;
L
Linus Torvalds 已提交
263 264 265 266 267

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

268 269 270
/*
 * Find the disk number which triggered given bio
 */
271
static int find_bio_disk(struct r1bio *r1_bio, struct bio *bio)
272 273 274 275 276 277 278 279 280 281 282 283 284 285
{
	int mirror;
	int raid_disks = r1_bio->mddev->raid_disks;

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

	BUG_ON(mirror == raid_disks);
	update_head_pos(mirror, r1_bio);

	return mirror;
}

286
static void raid1_end_read_request(struct bio *bio, int error)
L
Linus Torvalds 已提交
287 288
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
289
	struct r1bio *r1_bio = bio->bi_private;
L
Linus Torvalds 已提交
290
	int mirror;
291
	conf_t *conf = r1_bio->mddev->private;
L
Linus Torvalds 已提交
292 293 294 295 296

	mirror = r1_bio->read_disk;
	/*
	 * this branch is our 'one mirror IO has finished' event handler:
	 */
297 298
	update_head_pos(mirror, r1_bio);

299 300 301 302 303 304
	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
305
		 */
306 307 308 309 310 311 312 313
		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
314

315
	if (uptodate)
L
Linus Torvalds 已提交
316
		raid_end_bio_io(r1_bio);
317
	else {
L
Linus Torvalds 已提交
318 319 320 321
		/*
		 * oops, read error:
		 */
		char b[BDEVNAME_SIZE];
322 323 324 325 326 327 328
		printk_ratelimited(
			KERN_ERR "md/raid1:%s: %s: "
			"rescheduling sector %llu\n",
			mdname(conf->mddev),
			bdevname(conf->mirrors[mirror].rdev->bdev,
				 b),
			(unsigned long long)r1_bio->sector);
329
		set_bit(R1BIO_ReadError, &r1_bio->state);
L
Linus Torvalds 已提交
330 331 332 333 334 335
		reschedule_retry(r1_bio);
	}

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

336
static void close_write(struct r1bio *r1_bio)
337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354
{
	/* 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 = r1_bio->behind_page_count;
		while (i--)
			safe_put_page(r1_bio->behind_bvecs[i].bv_page);
		kfree(r1_bio->behind_bvecs);
		r1_bio->behind_bvecs = NULL;
	}
	/* 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),
			test_bit(R1BIO_BehindIO, &r1_bio->state));
	md_write_end(r1_bio->mddev);
}

355
static void r1_bio_write_done(struct r1bio *r1_bio)
356
{
357 358 359 360 361 362 363
	if (!atomic_dec_and_test(&r1_bio->remaining))
		return;

	if (test_bit(R1BIO_WriteError, &r1_bio->state))
		reschedule_retry(r1_bio);
	else {
		close_write(r1_bio);
364 365 366 367
		if (test_bit(R1BIO_MadeGood, &r1_bio->state))
			reschedule_retry(r1_bio);
		else
			raid_end_bio_io(r1_bio);
368 369 370
	}
}

371
static void raid1_end_write_request(struct bio *bio, int error)
L
Linus Torvalds 已提交
372 373
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
374
	struct r1bio *r1_bio = bio->bi_private;
375
	int mirror, behind = test_bit(R1BIO_BehindIO, &r1_bio->state);
376
	conf_t *conf = r1_bio->mddev->private;
377
	struct bio *to_put = NULL;
L
Linus Torvalds 已提交
378

379
	mirror = find_bio_disk(r1_bio, bio);
L
Linus Torvalds 已提交
380

T
Tejun Heo 已提交
381 382 383 384
	/*
	 * 'one mirror IO has finished' event handler:
	 */
	if (!uptodate) {
385 386 387
		set_bit(WriteErrorSeen,
			&conf->mirrors[mirror].rdev->flags);
		set_bit(R1BIO_WriteError, &r1_bio->state);
388
	} else {
L
Linus Torvalds 已提交
389
		/*
T
Tejun Heo 已提交
390 391 392 393 394 395 396 397
		 * 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.
L
Linus Torvalds 已提交
398
		 */
399 400 401
		sector_t first_bad;
		int bad_sectors;

402 403
		r1_bio->bios[mirror] = NULL;
		to_put = bio;
T
Tejun Heo 已提交
404 405
		set_bit(R1BIO_Uptodate, &r1_bio->state);

406 407 408 409 410 411 412 413 414
		/* Maybe we can clear some bad blocks. */
		if (is_badblock(conf->mirrors[mirror].rdev,
				r1_bio->sector, r1_bio->sectors,
				&first_bad, &bad_sectors)) {
			r1_bio->bios[mirror] = IO_MADE_GOOD;
			set_bit(R1BIO_MadeGood, &r1_bio->state);
		}
	}

T
Tejun Heo 已提交
415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430
	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;
431 432 433 434 435
				pr_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);
436
				call_bio_endio(r1_bio);
437 438 439
			}
		}
	}
440 441 442
	if (r1_bio->bios[mirror] == NULL)
		rdev_dec_pending(conf->mirrors[mirror].rdev,
				 conf->mddev);
T
Tejun Heo 已提交
443

L
Linus Torvalds 已提交
444 445 446 447
	/*
	 * Let's see if all mirrored write operations have finished
	 * already.
	 */
448
	r1_bio_write_done(r1_bio);
449

450 451
	if (to_put)
		bio_put(to_put);
L
Linus Torvalds 已提交
452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468
}


/*
 * 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.
 */
469
static int read_balance(conf_t *conf, struct r1bio *r1_bio, int *max_sectors)
L
Linus Torvalds 已提交
470
{
471
	const sector_t this_sector = r1_bio->sector;
472 473
	int sectors;
	int best_good_sectors;
474
	int start_disk;
N
NeilBrown 已提交
475
	int best_disk;
476
	int i;
N
NeilBrown 已提交
477
	sector_t best_dist;
478
	struct md_rdev *rdev;
479
	int choose_first;
L
Linus Torvalds 已提交
480 481 482

	rcu_read_lock();
	/*
483
	 * Check if we can balance. We can balance on the whole
L
Linus Torvalds 已提交
484 485 486 487
	 * device if no resync is going on, or below the resync window.
	 * We take the first readable disk when above the resync window.
	 */
 retry:
488
	sectors = r1_bio->sectors;
N
NeilBrown 已提交
489 490
	best_disk = -1;
	best_dist = MaxSector;
491 492
	best_good_sectors = 0;

L
Linus Torvalds 已提交
493 494
	if (conf->mddev->recovery_cp < MaxSector &&
	    (this_sector + sectors >= conf->next_resync)) {
495 496 497 498 499
		choose_first = 1;
		start_disk = 0;
	} else {
		choose_first = 0;
		start_disk = conf->last_used;
L
Linus Torvalds 已提交
500 501
	}

502
	for (i = 0 ; i < conf->raid_disks ; i++) {
N
NeilBrown 已提交
503
		sector_t dist;
504 505 506
		sector_t first_bad;
		int bad_sectors;

507 508 509 510 511 512 513
		int disk = start_disk + i;
		if (disk >= conf->raid_disks)
			disk -= conf->raid_disks;

		rdev = rcu_dereference(conf->mirrors[disk].rdev);
		if (r1_bio->bios[disk] == IO_BLOCKED
		    || rdev == NULL
N
NeilBrown 已提交
514
		    || test_bit(Faulty, &rdev->flags))
515
			continue;
N
NeilBrown 已提交
516 517
		if (!test_bit(In_sync, &rdev->flags) &&
		    rdev->recovery_offset < this_sector + sectors)
L
Linus Torvalds 已提交
518
			continue;
N
NeilBrown 已提交
519 520 521 522 523 524 525 526 527 528
		if (test_bit(WriteMostly, &rdev->flags)) {
			/* Don't balance among write-mostly, just
			 * use the first as a last resort */
			if (best_disk < 0)
				best_disk = disk;
			continue;
		}
		/* This is a reasonable device to use.  It might
		 * even be best.
		 */
529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557
		if (is_badblock(rdev, this_sector, sectors,
				&first_bad, &bad_sectors)) {
			if (best_dist < MaxSector)
				/* already have a better device */
				continue;
			if (first_bad <= this_sector) {
				/* cannot read here. If this is the 'primary'
				 * device, then we must not read beyond
				 * bad_sectors from another device..
				 */
				bad_sectors -= (this_sector - first_bad);
				if (choose_first && sectors > bad_sectors)
					sectors = bad_sectors;
				if (best_good_sectors > sectors)
					best_good_sectors = sectors;

			} else {
				sector_t good_sectors = first_bad - this_sector;
				if (good_sectors > best_good_sectors) {
					best_good_sectors = good_sectors;
					best_disk = disk;
				}
				if (choose_first)
					break;
			}
			continue;
		} else
			best_good_sectors = sectors;

N
NeilBrown 已提交
558 559 560 561 562 563 564 565
		dist = abs(this_sector - conf->mirrors[disk].head_position);
		if (choose_first
		    /* Don't change to another disk for sequential reads */
		    || conf->next_seq_sect == this_sector
		    || dist == 0
		    /* If device is idle, use it */
		    || atomic_read(&rdev->nr_pending) == 0) {
			best_disk = disk;
L
Linus Torvalds 已提交
566 567
			break;
		}
N
NeilBrown 已提交
568 569 570
		if (dist < best_dist) {
			best_dist = dist;
			best_disk = disk;
L
Linus Torvalds 已提交
571
		}
572
	}
L
Linus Torvalds 已提交
573

N
NeilBrown 已提交
574 575
	if (best_disk >= 0) {
		rdev = rcu_dereference(conf->mirrors[best_disk].rdev);
576 577 578
		if (!rdev)
			goto retry;
		atomic_inc(&rdev->nr_pending);
N
NeilBrown 已提交
579
		if (test_bit(Faulty, &rdev->flags)) {
L
Linus Torvalds 已提交
580 581 582
			/* cannot risk returning a device that failed
			 * before we inc'ed nr_pending
			 */
583
			rdev_dec_pending(rdev, conf->mddev);
L
Linus Torvalds 已提交
584 585
			goto retry;
		}
586
		sectors = best_good_sectors;
587
		conf->next_seq_sect = this_sector + sectors;
N
NeilBrown 已提交
588
		conf->last_used = best_disk;
L
Linus Torvalds 已提交
589 590
	}
	rcu_read_unlock();
591
	*max_sectors = sectors;
L
Linus Torvalds 已提交
592

N
NeilBrown 已提交
593
	return best_disk;
L
Linus Torvalds 已提交
594 595
}

596
int md_raid1_congested(struct mddev *mddev, int bits)
597
{
598
	conf_t *conf = mddev->private;
599 600 601 602
	int i, ret = 0;

	rcu_read_lock();
	for (i = 0; i < mddev->raid_disks; i++) {
603
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
604
		if (rdev && !test_bit(Faulty, &rdev->flags)) {
605
			struct request_queue *q = bdev_get_queue(rdev->bdev);
606

607 608
			BUG_ON(!q);

609 610 611
			/* Note the '|| 1' - when read_balance prefers
			 * non-congested targets, it can be removed
			 */
612
			if ((bits & (1<<BDI_async_congested)) || 1)
613 614 615 616 617 618 619 620
				ret |= bdi_congested(&q->backing_dev_info, bits);
			else
				ret &= bdi_congested(&q->backing_dev_info, bits);
		}
	}
	rcu_read_unlock();
	return ret;
}
621
EXPORT_SYMBOL_GPL(md_raid1_congested);
622

623 624
static int raid1_congested(void *data, int bits)
{
625
	struct mddev *mddev = data;
626 627 628 629

	return mddev_congested(mddev, bits) ||
		md_raid1_congested(mddev, bits);
}
630

J
Jens Axboe 已提交
631
static void flush_pending_writes(conf_t *conf)
632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653
{
	/* Any writes that have been queued but are awaiting
	 * bitmap updates get flushed here.
	 */
	spin_lock_irq(&conf->device_lock);

	if (conf->pending_bio_list.head) {
		struct bio *bio;
		bio = bio_list_get(&conf->pending_bio_list);
		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;
		}
	} else
		spin_unlock_irq(&conf->device_lock);
J
Jens Axboe 已提交
654 655
}

656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675
/* 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
676 677 678
 */
#define RESYNC_DEPTH 32

679
static void raise_barrier(conf_t *conf)
L
Linus Torvalds 已提交
680 681
{
	spin_lock_irq(&conf->resync_lock);
682 683 684

	/* Wait until no block IO is waiting */
	wait_event_lock_irq(conf->wait_barrier, !conf->nr_waiting,
N
NeilBrown 已提交
685
			    conf->resync_lock, );
686 687 688 689

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

N
NeilBrown 已提交
690
	/* Now wait for all pending IO to complete */
691 692
	wait_event_lock_irq(conf->wait_barrier,
			    !conf->nr_pending && conf->barrier < RESYNC_DEPTH,
N
NeilBrown 已提交
693
			    conf->resync_lock, );
694 695 696 697 698 699 700

	spin_unlock_irq(&conf->resync_lock);
}

static void lower_barrier(conf_t *conf)
{
	unsigned long flags;
701
	BUG_ON(conf->barrier <= 0);
702 703 704 705 706 707 708 709 710 711 712 713 714
	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,
N
NeilBrown 已提交
715
				    );
716
		conf->nr_waiting--;
L
Linus Torvalds 已提交
717
	}
718
	conf->nr_pending++;
L
Linus Torvalds 已提交
719 720 721
	spin_unlock_irq(&conf->resync_lock);
}

722 723 724 725 726 727 728 729 730
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);
}

731 732 733 734 735
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
736 737 738 739 740 741 742 743
	 * 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.
744 745 746 747 748
	 */
	spin_lock_irq(&conf->resync_lock);
	conf->barrier++;
	conf->nr_waiting++;
	wait_event_lock_irq(conf->wait_barrier,
749
			    conf->nr_pending == conf->nr_queued+1,
750
			    conf->resync_lock,
N
NeilBrown 已提交
751
			    flush_pending_writes(conf));
752 753 754 755 756 757 758 759 760 761 762 763
	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);
}

764

765 766
/* duplicate the data pages for behind I/O 
 */
767
static void alloc_behind_pages(struct bio *bio, struct r1bio *r1_bio)
768 769 770
{
	int i;
	struct bio_vec *bvec;
771
	struct bio_vec *bvecs = kzalloc(bio->bi_vcnt * sizeof(struct bio_vec),
772
					GFP_NOIO);
773
	if (unlikely(!bvecs))
774
		return;
775 776

	bio_for_each_segment(bvec, bio, i) {
777 778 779
		bvecs[i] = *bvec;
		bvecs[i].bv_page = alloc_page(GFP_NOIO);
		if (unlikely(!bvecs[i].bv_page))
780
			goto do_sync_io;
781 782 783
		memcpy(kmap(bvecs[i].bv_page) + bvec->bv_offset,
		       kmap(bvec->bv_page) + bvec->bv_offset, bvec->bv_len);
		kunmap(bvecs[i].bv_page);
784 785
		kunmap(bvec->bv_page);
	}
786
	r1_bio->behind_bvecs = bvecs;
787 788 789
	r1_bio->behind_page_count = bio->bi_vcnt;
	set_bit(R1BIO_BehindIO, &r1_bio->state);
	return;
790 791

do_sync_io:
792
	for (i = 0; i < bio->bi_vcnt; i++)
793 794 795
		if (bvecs[i].bv_page)
			put_page(bvecs[i].bv_page);
	kfree(bvecs);
796
	pr_debug("%dB behind alloc failed, doing sync I/O\n", bio->bi_size);
797 798
}

799
static int make_request(struct mddev *mddev, struct bio * bio)
L
Linus Torvalds 已提交
800
{
801
	conf_t *conf = mddev->private;
802
	struct mirror_info *mirror;
803
	struct r1bio *r1_bio;
L
Linus Torvalds 已提交
804
	struct bio *read_bio;
805
	int i, disks;
806
	struct bitmap *bitmap;
807
	unsigned long flags;
808
	const int rw = bio_data_dir(bio);
809
	const unsigned long do_sync = (bio->bi_rw & REQ_SYNC);
T
Tejun Heo 已提交
810
	const unsigned long do_flush_fua = (bio->bi_rw & (REQ_FLUSH | REQ_FUA));
811
	struct md_rdev *blocked_rdev;
N
NeilBrown 已提交
812
	int plugged;
813 814 815
	int first_clone;
	int sectors_handled;
	int max_sectors;
816

L
Linus Torvalds 已提交
817 818 819 820 821
	/*
	 * 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.
	 */
822

823 824
	md_write_start(mddev, bio); /* wait on superblock update early */

825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843
	if (bio_data_dir(bio) == WRITE &&
	    bio->bi_sector + bio->bi_size/512 > mddev->suspend_lo &&
	    bio->bi_sector < mddev->suspend_hi) {
		/* As the suspend_* range is controlled by
		 * userspace, we want an interruptible
		 * wait.
		 */
		DEFINE_WAIT(w);
		for (;;) {
			flush_signals(current);
			prepare_to_wait(&conf->wait_barrier,
					&w, TASK_INTERRUPTIBLE);
			if (bio->bi_sector + bio->bi_size/512 <= mddev->suspend_lo ||
			    bio->bi_sector >= mddev->suspend_hi)
				break;
			schedule();
		}
		finish_wait(&conf->wait_barrier, &w);
	}
844

845
	wait_barrier(conf);
L
Linus Torvalds 已提交
846

847 848
	bitmap = mddev->bitmap;

L
Linus Torvalds 已提交
849 850 851 852 853 854 855 856 857
	/*
	 * 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;
858
	r1_bio->state = 0;
L
Linus Torvalds 已提交
859 860 861
	r1_bio->mddev = mddev;
	r1_bio->sector = bio->bi_sector;

862 863 864 865 866 867 868 869 870 871
	/* We might need to issue multiple reads to different
	 * devices if there are bad blocks around, so we keep
	 * track of the number of reads in bio->bi_phys_segments.
	 * If this is 0, there is only one r1_bio and no locking
	 * will be needed when requests complete.  If it is
	 * non-zero, then it is the number of not-completed requests.
	 */
	bio->bi_phys_segments = 0;
	clear_bit(BIO_SEG_VALID, &bio->bi_flags);

872
	if (rw == READ) {
L
Linus Torvalds 已提交
873 874 875
		/*
		 * read balancing logic:
		 */
876 877 878 879
		int rdisk;

read_again:
		rdisk = read_balance(conf, r1_bio, &max_sectors);
L
Linus Torvalds 已提交
880 881 882 883 884 885 886 887

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

888 889 890 891 892 893 894 895 896
		if (test_bit(WriteMostly, &mirror->rdev->flags) &&
		    bitmap) {
			/* Reading from a write-mostly device must
			 * take care not to over-take any writes
			 * that are 'behind'
			 */
			wait_event(bitmap->behind_wait,
				   atomic_read(&bitmap->behind_writes) == 0);
		}
L
Linus Torvalds 已提交
897 898
		r1_bio->read_disk = rdisk;

899
		read_bio = bio_clone_mddev(bio, GFP_NOIO, mddev);
900 901
		md_trim_bio(read_bio, r1_bio->sector - bio->bi_sector,
			    max_sectors);
L
Linus Torvalds 已提交
902 903 904 905 906 907

		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;
908
		read_bio->bi_rw = READ | do_sync;
L
Linus Torvalds 已提交
909 910
		read_bio->bi_private = r1_bio;

911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941
		if (max_sectors < r1_bio->sectors) {
			/* could not read all from this device, so we will
			 * need another r1_bio.
			 */

			sectors_handled = (r1_bio->sector + max_sectors
					   - bio->bi_sector);
			r1_bio->sectors = max_sectors;
			spin_lock_irq(&conf->device_lock);
			if (bio->bi_phys_segments == 0)
				bio->bi_phys_segments = 2;
			else
				bio->bi_phys_segments++;
			spin_unlock_irq(&conf->device_lock);
			/* Cannot call generic_make_request directly
			 * as that will be queued in __make_request
			 * and subsequent mempool_alloc might block waiting
			 * for it.  So hand bio over to raid1d.
			 */
			reschedule_retry(r1_bio);

			r1_bio = mempool_alloc(conf->r1bio_pool, GFP_NOIO);

			r1_bio->master_bio = bio;
			r1_bio->sectors = (bio->bi_size >> 9) - sectors_handled;
			r1_bio->state = 0;
			r1_bio->mddev = mddev;
			r1_bio->sector = bio->bi_sector + sectors_handled;
			goto read_again;
		} else
			generic_make_request(read_bio);
L
Linus Torvalds 已提交
942 943 944 945 946 947
		return 0;
	}

	/*
	 * WRITE:
	 */
948
	/* first select target devices under rcu_lock and
L
Linus Torvalds 已提交
949 950
	 * inc refcount on their rdev.  Record them by setting
	 * bios[x] to bio
951 952 953 954 955 956
	 * If there are known/acknowledged bad blocks on any device on
	 * which we have seen a write error, we want to avoid writing those
	 * blocks.
	 * This potentially requires several writes to write around
	 * the bad blocks.  Each set of writes gets it's own r1bio
	 * with a set of bios attached.
L
Linus Torvalds 已提交
957
	 */
N
NeilBrown 已提交
958 959
	plugged = mddev_check_plugged(mddev);

L
Linus Torvalds 已提交
960
	disks = conf->raid_disks;
961 962
 retry_write:
	blocked_rdev = NULL;
L
Linus Torvalds 已提交
963
	rcu_read_lock();
964
	max_sectors = r1_bio->sectors;
L
Linus Torvalds 已提交
965
	for (i = 0;  i < disks; i++) {
966
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
967 968 969 970 971
		if (rdev && unlikely(test_bit(Blocked, &rdev->flags))) {
			atomic_inc(&rdev->nr_pending);
			blocked_rdev = rdev;
			break;
		}
972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001
		r1_bio->bios[i] = NULL;
		if (!rdev || test_bit(Faulty, &rdev->flags)) {
			set_bit(R1BIO_Degraded, &r1_bio->state);
			continue;
		}

		atomic_inc(&rdev->nr_pending);
		if (test_bit(WriteErrorSeen, &rdev->flags)) {
			sector_t first_bad;
			int bad_sectors;
			int is_bad;

			is_bad = is_badblock(rdev, r1_bio->sector,
					     max_sectors,
					     &first_bad, &bad_sectors);
			if (is_bad < 0) {
				/* mustn't write here until the bad block is
				 * acknowledged*/
				set_bit(BlockedBadBlocks, &rdev->flags);
				blocked_rdev = rdev;
				break;
			}
			if (is_bad && first_bad <= r1_bio->sector) {
				/* Cannot write here at all */
				bad_sectors -= (r1_bio->sector - first_bad);
				if (bad_sectors < max_sectors)
					/* mustn't write more than bad_sectors
					 * to other devices yet
					 */
					max_sectors = bad_sectors;
1002
				rdev_dec_pending(rdev, mddev);
1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013
				/* We don't set R1BIO_Degraded as that
				 * only applies if the disk is
				 * missing, so it might be re-added,
				 * and we want to know to recover this
				 * chunk.
				 * In this case the device is here,
				 * and the fact that this chunk is not
				 * in-sync is recorded in the bad
				 * block log
				 */
				continue;
1014
			}
1015 1016 1017 1018 1019 1020 1021
			if (is_bad) {
				int good_sectors = first_bad - r1_bio->sector;
				if (good_sectors < max_sectors)
					max_sectors = good_sectors;
			}
		}
		r1_bio->bios[i] = bio;
L
Linus Torvalds 已提交
1022 1023 1024
	}
	rcu_read_unlock();

1025 1026 1027 1028 1029 1030 1031
	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);
1032
		r1_bio->state = 0;
1033 1034 1035 1036 1037 1038
		allow_barrier(conf);
		md_wait_for_blocked_rdev(blocked_rdev, mddev);
		wait_barrier(conf);
		goto retry_write;
	}

1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049
	if (max_sectors < r1_bio->sectors) {
		/* We are splitting this write into multiple parts, so
		 * we need to prepare for allocating another r1_bio.
		 */
		r1_bio->sectors = max_sectors;
		spin_lock_irq(&conf->device_lock);
		if (bio->bi_phys_segments == 0)
			bio->bi_phys_segments = 2;
		else
			bio->bi_phys_segments++;
		spin_unlock_irq(&conf->device_lock);
1050
	}
1051
	sectors_handled = r1_bio->sector + max_sectors - bio->bi_sector;
1052

1053
	atomic_set(&r1_bio->remaining, 1);
1054
	atomic_set(&r1_bio->behind_remaining, 0);
1055

1056
	first_clone = 1;
L
Linus Torvalds 已提交
1057 1058 1059 1060 1061
	for (i = 0; i < disks; i++) {
		struct bio *mbio;
		if (!r1_bio->bios[i])
			continue;

1062
		mbio = bio_clone_mddev(bio, GFP_NOIO, mddev);
1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081
		md_trim_bio(mbio, r1_bio->sector - bio->bi_sector, max_sectors);

		if (first_clone) {
			/* do behind I/O ?
			 * Not if there are too many, or cannot
			 * allocate memory, or a reader on WriteMostly
			 * is waiting for behind writes to flush */
			if (bitmap &&
			    (atomic_read(&bitmap->behind_writes)
			     < mddev->bitmap_info.max_write_behind) &&
			    !waitqueue_active(&bitmap->behind_wait))
				alloc_behind_pages(mbio, r1_bio);

			bitmap_startwrite(bitmap, r1_bio->sector,
					  r1_bio->sectors,
					  test_bit(R1BIO_BehindIO,
						   &r1_bio->state));
			first_clone = 0;
		}
1082
		if (r1_bio->behind_bvecs) {
1083 1084 1085 1086 1087 1088 1089
			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
N
NeilBrown 已提交
1090
			 * know the original bi_idx, so we just free
1091 1092 1093
			 * them all
			 */
			__bio_for_each_segment(bvec, mbio, j, 0)
1094
				bvec->bv_page = r1_bio->behind_bvecs[j].bv_page;
1095 1096 1097 1098
			if (test_bit(WriteMostly, &conf->mirrors[i].rdev->flags))
				atomic_inc(&r1_bio->behind_remaining);
		}

1099 1100 1101 1102 1103 1104 1105 1106 1107
		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;
		mbio->bi_rw = WRITE | do_flush_fua | do_sync;
		mbio->bi_private = r1_bio;

L
Linus Torvalds 已提交
1108
		atomic_inc(&r1_bio->remaining);
1109 1110 1111
		spin_lock_irqsave(&conf->device_lock, flags);
		bio_list_add(&conf->pending_bio_list, mbio);
		spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1112
	}
1113 1114 1115
	/* Mustn't call r1_bio_write_done before this next test,
	 * as it could result in the bio being freed.
	 */
1116
	if (sectors_handled < (bio->bi_size >> 9)) {
1117
		r1_bio_write_done(r1_bio);
1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129
		/* We need another r1_bio.  It has already been counted
		 * in bio->bi_phys_segments
		 */
		r1_bio = mempool_alloc(conf->r1bio_pool, GFP_NOIO);
		r1_bio->master_bio = bio;
		r1_bio->sectors = (bio->bi_size >> 9) - sectors_handled;
		r1_bio->state = 0;
		r1_bio->mddev = mddev;
		r1_bio->sector = bio->bi_sector + sectors_handled;
		goto retry_write;
	}

1130 1131 1132 1133 1134
	r1_bio_write_done(r1_bio);

	/* In case raid1d snuck in to freeze_array */
	wake_up(&conf->wait_barrier);

N
NeilBrown 已提交
1135
	if (do_sync || !bitmap || !plugged)
1136
		md_wakeup_thread(mddev->thread);
1137

L
Linus Torvalds 已提交
1138 1139 1140
	return 0;
}

1141
static void status(struct seq_file *seq, struct mddev *mddev)
L
Linus Torvalds 已提交
1142
{
1143
	conf_t *conf = mddev->private;
L
Linus Torvalds 已提交
1144 1145 1146
	int i;

	seq_printf(seq, " [%d/%d] [", conf->raid_disks,
1147
		   conf->raid_disks - mddev->degraded);
1148 1149
	rcu_read_lock();
	for (i = 0; i < conf->raid_disks; i++) {
1150
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
L
Linus Torvalds 已提交
1151
		seq_printf(seq, "%s",
1152 1153 1154
			   rdev && test_bit(In_sync, &rdev->flags) ? "U" : "_");
	}
	rcu_read_unlock();
L
Linus Torvalds 已提交
1155 1156 1157 1158
	seq_printf(seq, "]");
}


1159
static void error(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
1160 1161
{
	char b[BDEVNAME_SIZE];
1162
	conf_t *conf = mddev->private;
L
Linus Torvalds 已提交
1163 1164 1165 1166 1167 1168 1169

	/*
	 * 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
	 */
1170
	if (test_bit(In_sync, &rdev->flags)
1171
	    && (conf->raid_disks - mddev->degraded) == 1) {
L
Linus Torvalds 已提交
1172 1173
		/*
		 * Don't fail the drive, act as though we were just a
1174 1175 1176
		 * normal single drive.
		 * However don't try a recovery from this drive as
		 * it is very likely to fail.
L
Linus Torvalds 已提交
1177
		 */
1178
		conf->recovery_disabled = mddev->recovery_disabled;
L
Linus Torvalds 已提交
1179
		return;
1180
	}
1181
	set_bit(Blocked, &rdev->flags);
1182 1183 1184
	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
1185
		mddev->degraded++;
1186
		set_bit(Faulty, &rdev->flags);
1187
		spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1188 1189 1190
		/*
		 * if recovery is running, make sure it aborts.
		 */
1191
		set_bit(MD_RECOVERY_INTR, &mddev->recovery);
1192 1193
	} else
		set_bit(Faulty, &rdev->flags);
1194
	set_bit(MD_CHANGE_DEVS, &mddev->flags);
1195 1196 1197
	printk(KERN_ALERT
	       "md/raid1:%s: Disk failure on %s, disabling device.\n"
	       "md/raid1:%s: Operation continuing on %d devices.\n",
N
NeilBrown 已提交
1198 1199
	       mdname(mddev), bdevname(rdev->bdev, b),
	       mdname(mddev), conf->raid_disks - mddev->degraded);
L
Linus Torvalds 已提交
1200 1201 1202 1203 1204 1205
}

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

N
NeilBrown 已提交
1206
	printk(KERN_DEBUG "RAID1 conf printout:\n");
L
Linus Torvalds 已提交
1207
	if (!conf) {
N
NeilBrown 已提交
1208
		printk(KERN_DEBUG "(!conf)\n");
L
Linus Torvalds 已提交
1209 1210
		return;
	}
N
NeilBrown 已提交
1211
	printk(KERN_DEBUG " --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
L
Linus Torvalds 已提交
1212 1213
		conf->raid_disks);

1214
	rcu_read_lock();
L
Linus Torvalds 已提交
1215 1216
	for (i = 0; i < conf->raid_disks; i++) {
		char b[BDEVNAME_SIZE];
1217
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
1218
		if (rdev)
N
NeilBrown 已提交
1219
			printk(KERN_DEBUG " disk %d, wo:%d, o:%d, dev:%s\n",
1220 1221 1222
			       i, !test_bit(In_sync, &rdev->flags),
			       !test_bit(Faulty, &rdev->flags),
			       bdevname(rdev->bdev,b));
L
Linus Torvalds 已提交
1223
	}
1224
	rcu_read_unlock();
L
Linus Torvalds 已提交
1225 1226 1227 1228
}

static void close_sync(conf_t *conf)
{
1229 1230
	wait_barrier(conf);
	allow_barrier(conf);
L
Linus Torvalds 已提交
1231 1232 1233 1234 1235

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

1236
static int raid1_spare_active(struct mddev *mddev)
L
Linus Torvalds 已提交
1237 1238 1239
{
	int i;
	conf_t *conf = mddev->private;
1240 1241
	int count = 0;
	unsigned long flags;
L
Linus Torvalds 已提交
1242 1243 1244

	/*
	 * Find all failed disks within the RAID1 configuration 
1245 1246
	 * and mark them readable.
	 * Called under mddev lock, so rcu protection not needed.
L
Linus Torvalds 已提交
1247 1248
	 */
	for (i = 0; i < conf->raid_disks; i++) {
1249
		struct md_rdev *rdev = conf->mirrors[i].rdev;
1250 1251
		if (rdev
		    && !test_bit(Faulty, &rdev->flags)
1252
		    && !test_and_set_bit(In_sync, &rdev->flags)) {
1253
			count++;
1254
			sysfs_notify_dirent_safe(rdev->sysfs_state);
L
Linus Torvalds 已提交
1255 1256
		}
	}
1257 1258 1259
	spin_lock_irqsave(&conf->device_lock, flags);
	mddev->degraded -= count;
	spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1260 1261

	print_conf(conf);
1262
	return count;
L
Linus Torvalds 已提交
1263 1264 1265
}


1266
static int raid1_add_disk(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
1267 1268
{
	conf_t *conf = mddev->private;
1269
	int err = -EEXIST;
1270
	int mirror = 0;
1271
	struct mirror_info *p;
1272 1273
	int first = 0;
	int last = mddev->raid_disks - 1;
L
Linus Torvalds 已提交
1274

1275 1276 1277
	if (mddev->recovery_disabled == conf->recovery_disabled)
		return -EBUSY;

1278 1279 1280 1281
	if (rdev->raid_disk >= 0)
		first = last = rdev->raid_disk;

	for (mirror = first; mirror <= last; mirror++)
L
Linus Torvalds 已提交
1282 1283
		if ( !(p=conf->mirrors+mirror)->rdev) {

1284 1285
			disk_stack_limits(mddev->gendisk, rdev->bdev,
					  rdev->data_offset << 9);
1286 1287 1288 1289 1290
			/* as we don't honour merge_bvec_fn, we must
			 * never risk violating it, so limit
			 * ->max_segments to one lying with a single
			 * page, as a one page request is never in
			 * violation.
L
Linus Torvalds 已提交
1291
			 */
1292 1293 1294 1295 1296
			if (rdev->bdev->bd_disk->queue->merge_bvec_fn) {
				blk_queue_max_segments(mddev->queue, 1);
				blk_queue_segment_boundary(mddev->queue,
							   PAGE_CACHE_SIZE - 1);
			}
L
Linus Torvalds 已提交
1297 1298 1299

			p->head_position = 0;
			rdev->raid_disk = mirror;
1300
			err = 0;
1301 1302 1303 1304
			/* 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)
1305
				conf->fullsync = 1;
1306
			rcu_assign_pointer(p->rdev, rdev);
L
Linus Torvalds 已提交
1307 1308
			break;
		}
1309
	md_integrity_add_rdev(rdev, mddev);
L
Linus Torvalds 已提交
1310
	print_conf(conf);
1311
	return err;
L
Linus Torvalds 已提交
1312 1313
}

1314
static int raid1_remove_disk(struct mddev *mddev, int number)
L
Linus Torvalds 已提交
1315 1316 1317
{
	conf_t *conf = mddev->private;
	int err = 0;
1318
	struct md_rdev *rdev;
1319
	struct mirror_info *p = conf->mirrors+ number;
L
Linus Torvalds 已提交
1320 1321 1322 1323

	print_conf(conf);
	rdev = p->rdev;
	if (rdev) {
1324
		if (test_bit(In_sync, &rdev->flags) ||
L
Linus Torvalds 已提交
1325 1326 1327 1328
		    atomic_read(&rdev->nr_pending)) {
			err = -EBUSY;
			goto abort;
		}
N
NeilBrown 已提交
1329
		/* Only remove non-faulty devices if recovery
1330 1331 1332
		 * is not possible.
		 */
		if (!test_bit(Faulty, &rdev->flags) &&
1333
		    mddev->recovery_disabled != conf->recovery_disabled &&
1334 1335 1336 1337
		    mddev->degraded < conf->raid_disks) {
			err = -EBUSY;
			goto abort;
		}
L
Linus Torvalds 已提交
1338
		p->rdev = NULL;
1339
		synchronize_rcu();
L
Linus Torvalds 已提交
1340 1341 1342 1343
		if (atomic_read(&rdev->nr_pending)) {
			/* lost the race, try later */
			err = -EBUSY;
			p->rdev = rdev;
1344
			goto abort;
L
Linus Torvalds 已提交
1345
		}
1346
		err = md_integrity_register(mddev);
L
Linus Torvalds 已提交
1347 1348 1349 1350 1351 1352 1353 1354
	}
abort:

	print_conf(conf);
	return err;
}


1355
static void end_sync_read(struct bio *bio, int error)
L
Linus Torvalds 已提交
1356
{
1357
	struct r1bio *r1_bio = bio->bi_private;
L
Linus Torvalds 已提交
1358

1359
	update_head_pos(r1_bio->read_disk, r1_bio);
1360

L
Linus Torvalds 已提交
1361 1362 1363 1364 1365
	/*
	 * 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
	 */
1366
	if (test_bit(BIO_UPTODATE, &bio->bi_flags))
L
Linus Torvalds 已提交
1367
		set_bit(R1BIO_Uptodate, &r1_bio->state);
1368 1369 1370

	if (atomic_dec_and_test(&r1_bio->remaining))
		reschedule_retry(r1_bio);
L
Linus Torvalds 已提交
1371 1372
}

1373
static void end_sync_write(struct bio *bio, int error)
L
Linus Torvalds 已提交
1374 1375
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
1376
	struct r1bio *r1_bio = bio->bi_private;
1377
	struct mddev *mddev = r1_bio->mddev;
1378
	conf_t *conf = mddev->private;
L
Linus Torvalds 已提交
1379
	int mirror=0;
1380 1381
	sector_t first_bad;
	int bad_sectors;
L
Linus Torvalds 已提交
1382

1383 1384
	mirror = find_bio_disk(r1_bio, bio);

1385
	if (!uptodate) {
N
NeilBrown 已提交
1386
		sector_t sync_blocks = 0;
1387 1388 1389 1390
		sector_t s = r1_bio->sector;
		long sectors_to_go = r1_bio->sectors;
		/* make sure these bits doesn't get cleared. */
		do {
1391
			bitmap_end_sync(mddev->bitmap, s,
1392 1393 1394 1395
					&sync_blocks, 1);
			s += sync_blocks;
			sectors_to_go -= sync_blocks;
		} while (sectors_to_go > 0);
1396 1397 1398
		set_bit(WriteErrorSeen,
			&conf->mirrors[mirror].rdev->flags);
		set_bit(R1BIO_WriteError, &r1_bio->state);
1399 1400 1401
	} else if (is_badblock(conf->mirrors[mirror].rdev,
			       r1_bio->sector,
			       r1_bio->sectors,
1402 1403 1404 1405 1406 1407
			       &first_bad, &bad_sectors) &&
		   !is_badblock(conf->mirrors[r1_bio->read_disk].rdev,
				r1_bio->sector,
				r1_bio->sectors,
				&first_bad, &bad_sectors)
		)
1408
		set_bit(R1BIO_MadeGood, &r1_bio->state);
1409

L
Linus Torvalds 已提交
1410
	if (atomic_dec_and_test(&r1_bio->remaining)) {
1411
		int s = r1_bio->sectors;
1412 1413
		if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
		    test_bit(R1BIO_WriteError, &r1_bio->state))
1414 1415 1416 1417 1418
			reschedule_retry(r1_bio);
		else {
			put_buf(r1_bio);
			md_done_sync(mddev, s, uptodate);
		}
L
Linus Torvalds 已提交
1419 1420 1421
	}
}

1422
static int r1_sync_page_io(struct md_rdev *rdev, sector_t sector,
1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435
			    int sectors, struct page *page, int rw)
{
	if (sync_page_io(rdev, sector, sectors << 9, page, rw, false))
		/* success */
		return 1;
	if (rw == WRITE)
		set_bit(WriteErrorSeen, &rdev->flags);
	/* need to record an error - either for the block or the device */
	if (!rdev_set_badblocks(rdev, sector, sectors, 0))
		md_error(rdev->mddev, rdev);
	return 0;
}

1436
static int fix_sync_read_error(struct r1bio *r1_bio)
L
Linus Torvalds 已提交
1437
{
1438 1439 1440 1441 1442 1443 1444
	/* Try some synchronous reads of other devices to get
	 * good data, much like with normal read errors.  Only
	 * read into the pages we already have so we don't
	 * 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.
1445 1446 1447
	 * We don't need to check is_badblock() again as we
	 * made sure that anything with a bad block in range
	 * will have bi_end_io clear.
1448
	 */
1449
	struct mddev *mddev = r1_bio->mddev;
1450
	conf_t *conf = mddev->private;
1451 1452 1453 1454 1455 1456 1457 1458 1459
	struct bio *bio = r1_bio->bios[r1_bio->read_disk];
	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;
1460
		struct md_rdev *rdev;
1461
		int start;
1462 1463 1464 1465 1466 1467 1468 1469 1470 1471

		if (s > (PAGE_SIZE>>9))
			s = PAGE_SIZE >> 9;
		do {
			if (r1_bio->bios[d]->bi_end_io == end_sync_read) {
				/* No rcu protection needed here devices
				 * can only be removed when no resync is
				 * active, and resync is currently active
				 */
				rdev = conf->mirrors[d].rdev;
1472
				if (sync_page_io(rdev, sect, s<<9,
1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483
						 bio->bi_io_vec[idx].bv_page,
						 READ, false)) {
					success = 1;
					break;
				}
			}
			d++;
			if (d == conf->raid_disks)
				d = 0;
		} while (!success && d != r1_bio->read_disk);

1484
		if (!success) {
1485
			char b[BDEVNAME_SIZE];
1486 1487 1488 1489 1490 1491
			int abort = 0;
			/* Cannot read from anywhere, this block is lost.
			 * Record a bad block on each device.  If that doesn't
			 * work just disable and interrupt the recovery.
			 * Don't fail devices as that won't really help.
			 */
1492 1493 1494 1495 1496
			printk(KERN_ALERT "md/raid1:%s: %s: unrecoverable I/O read error"
			       " for block %llu\n",
			       mdname(mddev),
			       bdevname(bio->bi_bdev, b),
			       (unsigned long long)r1_bio->sector);
1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515
			for (d = 0; d < conf->raid_disks; d++) {
				rdev = conf->mirrors[d].rdev;
				if (!rdev || test_bit(Faulty, &rdev->flags))
					continue;
				if (!rdev_set_badblocks(rdev, sect, s, 0))
					abort = 1;
			}
			if (abort) {
				mddev->recovery_disabled = 1;
				set_bit(MD_RECOVERY_INTR, &mddev->recovery);
				md_done_sync(mddev, r1_bio->sectors, 0);
				put_buf(r1_bio);
				return 0;
			}
			/* Try next page */
			sectors -= s;
			sect += s;
			idx++;
			continue;
1516
		}
1517 1518 1519 1520 1521 1522 1523 1524 1525 1526

		start = d;
		/* write it back and re-read */
		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;
1527 1528 1529
			if (r1_sync_page_io(rdev, sect, s,
					    bio->bi_io_vec[idx].bv_page,
					    WRITE) == 0) {
1530 1531
				r1_bio->bios[d]->bi_end_io = NULL;
				rdev_dec_pending(rdev, mddev);
1532
			}
1533 1534 1535 1536 1537 1538 1539 1540 1541
		}
		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;
1542 1543 1544
			if (r1_sync_page_io(rdev, sect, s,
					    bio->bi_io_vec[idx].bv_page,
					    READ) != 0)
1545
				atomic_add(s, &rdev->corrected_errors);
1546
		}
1547 1548 1549 1550
		sectors -= s;
		sect += s;
		idx ++;
	}
1551
	set_bit(R1BIO_Uptodate, &r1_bio->state);
1552
	set_bit(BIO_UPTODATE, &bio->bi_flags);
1553 1554 1555
	return 1;
}

1556
static int process_checks(struct r1bio *r1_bio)
1557 1558 1559 1560 1561 1562 1563 1564
{
	/* 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
	 */
1565
	struct mddev *mddev = r1_bio->mddev;
1566 1567 1568 1569
	conf_t *conf = mddev->private;
	int primary;
	int i;

1570
	for (primary = 0; primary < conf->raid_disks; primary++)
1571 1572 1573 1574 1575 1576 1577
		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;
			rdev_dec_pending(conf->mirrors[primary].rdev, mddev);
			break;
		}
	r1_bio->read_disk = primary;
1578 1579 1580 1581 1582 1583
	for (i = 0; i < conf->raid_disks; i++) {
		int j;
		int vcnt = r1_bio->sectors >> (PAGE_SHIFT- 9);
		struct bio *pbio = r1_bio->bios[primary];
		struct bio *sbio = r1_bio->bios[i];
		int size;
1584

1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596
		if (r1_bio->bios[i]->bi_end_io != end_sync_read)
			continue;

		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;
1597
			}
1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632
		} else
			j = 0;
		if (j >= 0)
			mddev->resync_mismatches += r1_bio->sectors;
		if (j < 0 || (test_bit(MD_RECOVERY_CHECK, &mddev->recovery)
			      && test_bit(BIO_UPTODATE, &sbio->bi_flags))) {
			/* No need to write to this device. */
			sbio->bi_end_io = NULL;
			rdev_dec_pending(conf->mirrors[i].rdev, mddev);
			continue;
		}
		/* fixup the bio for reuse */
		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;
		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),
			       page_address(pbio->bi_io_vec[j].bv_page),
			       PAGE_SIZE);
1633
		}
1634
	}
1635 1636 1637
	return 0;
}

1638
static void sync_request_write(struct mddev *mddev, struct r1bio *r1_bio)
1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650
{
	conf_t *conf = mddev->private;
	int i;
	int disks = conf->raid_disks;
	struct bio *bio, *wbio;

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

	if (!test_bit(R1BIO_Uptodate, &r1_bio->state))
		/* ouch - failed to read all of that. */
		if (!fix_sync_read_error(r1_bio))
			return;
1651 1652 1653 1654

	if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
		if (process_checks(r1_bio) < 0)
			return;
1655 1656 1657
	/*
	 * schedule writes
	 */
L
Linus Torvalds 已提交
1658 1659 1660
	atomic_set(&r1_bio->remaining, 1);
	for (i = 0; i < disks ; i++) {
		wbio = r1_bio->bios[i];
1661 1662 1663 1664
		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
1665 1666
			continue;

1667 1668
		wbio->bi_rw = WRITE;
		wbio->bi_end_io = end_sync_write;
L
Linus Torvalds 已提交
1669 1670
		atomic_inc(&r1_bio->remaining);
		md_sync_acct(conf->mirrors[i].rdev->bdev, wbio->bi_size >> 9);
1671

L
Linus Torvalds 已提交
1672 1673 1674 1675
		generic_make_request(wbio);
	}

	if (atomic_dec_and_test(&r1_bio->remaining)) {
1676
		/* if we're here, all write(s) have completed, so clean up */
L
Linus Torvalds 已提交
1677 1678 1679 1680 1681 1682 1683 1684 1685 1686
		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.
1687
 *	3.	Performs writes following reads for array synchronising.
L
Linus Torvalds 已提交
1688 1689
 */

1690 1691 1692
static void fix_read_error(conf_t *conf, int read_disk,
			   sector_t sect, int sectors)
{
1693
	struct mddev *mddev = conf->mddev;
1694 1695 1696 1697 1698
	while(sectors) {
		int s = sectors;
		int d = read_disk;
		int success = 0;
		int start;
1699
		struct md_rdev *rdev;
1700 1701 1702 1703 1704 1705 1706 1707 1708 1709

		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....
			 */
1710 1711 1712
			sector_t first_bad;
			int bad_sectors;

1713 1714 1715
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
			    test_bit(In_sync, &rdev->flags) &&
1716 1717
			    is_badblock(rdev, sect, s,
					&first_bad, &bad_sectors) == 0 &&
J
Jonathan Brassow 已提交
1718 1719
			    sync_page_io(rdev, sect, s<<9,
					 conf->tmppage, READ, false))
1720 1721 1722 1723 1724 1725 1726 1727 1728
				success = 1;
			else {
				d++;
				if (d == conf->raid_disks)
					d = 0;
			}
		} while (!success && d != read_disk);

		if (!success) {
1729
			/* Cannot read from anywhere - mark it bad */
1730
			struct md_rdev *rdev = conf->mirrors[read_disk].rdev;
1731 1732
			if (!rdev_set_badblocks(rdev, sect, s, 0))
				md_error(mddev, rdev);
1733 1734 1735 1736 1737 1738 1739 1740 1741 1742
			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 &&
1743 1744 1745
			    test_bit(In_sync, &rdev->flags))
				r1_sync_page_io(rdev, sect, s,
						conf->tmppage, WRITE);
1746 1747 1748 1749 1750 1751 1752 1753 1754 1755
		}
		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)) {
1756 1757
				if (r1_sync_page_io(rdev, sect, s,
						    conf->tmppage, READ)) {
1758 1759
					atomic_add(s, &rdev->corrected_errors);
					printk(KERN_INFO
N
NeilBrown 已提交
1760
					       "md/raid1:%s: read error corrected "
1761 1762
					       "(%d sectors at %llu on %s)\n",
					       mdname(mddev), s,
1763 1764
					       (unsigned long long)(sect +
					           rdev->data_offset),
1765 1766 1767 1768 1769 1770 1771 1772 1773
					       bdevname(rdev->bdev, b));
				}
			}
		}
		sectors -= s;
		sect += s;
	}
}

1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792
static void bi_complete(struct bio *bio, int error)
{
	complete((struct completion *)bio->bi_private);
}

static int submit_bio_wait(int rw, struct bio *bio)
{
	struct completion event;
	rw |= REQ_SYNC;

	init_completion(&event);
	bio->bi_private = &event;
	bio->bi_end_io = bi_complete;
	submit_bio(rw, bio);
	wait_for_completion(&event);

	return test_bit(BIO_UPTODATE, &bio->bi_flags);
}

1793
static int narrow_write_error(struct r1bio *r1_bio, int i)
1794
{
1795
	struct mddev *mddev = r1_bio->mddev;
1796
	conf_t *conf = mddev->private;
1797
	struct md_rdev *rdev = conf->mirrors[i].rdev;
1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868
	int vcnt, idx;
	struct bio_vec *vec;

	/* bio has the data to be written to device 'i' where
	 * we just recently had a write error.
	 * We repeatedly clone the bio and trim down to one block,
	 * then try the write.  Where the write fails we record
	 * a bad block.
	 * It is conceivable that the bio doesn't exactly align with
	 * blocks.  We must handle this somehow.
	 *
	 * We currently own a reference on the rdev.
	 */

	int block_sectors;
	sector_t sector;
	int sectors;
	int sect_to_write = r1_bio->sectors;
	int ok = 1;

	if (rdev->badblocks.shift < 0)
		return 0;

	block_sectors = 1 << rdev->badblocks.shift;
	sector = r1_bio->sector;
	sectors = ((sector + block_sectors)
		   & ~(sector_t)(block_sectors - 1))
		- sector;

	if (test_bit(R1BIO_BehindIO, &r1_bio->state)) {
		vcnt = r1_bio->behind_page_count;
		vec = r1_bio->behind_bvecs;
		idx = 0;
		while (vec[idx].bv_page == NULL)
			idx++;
	} else {
		vcnt = r1_bio->master_bio->bi_vcnt;
		vec = r1_bio->master_bio->bi_io_vec;
		idx = r1_bio->master_bio->bi_idx;
	}
	while (sect_to_write) {
		struct bio *wbio;
		if (sectors > sect_to_write)
			sectors = sect_to_write;
		/* Write at 'sector' for 'sectors'*/

		wbio = bio_alloc_mddev(GFP_NOIO, vcnt, mddev);
		memcpy(wbio->bi_io_vec, vec, vcnt * sizeof(struct bio_vec));
		wbio->bi_sector = r1_bio->sector;
		wbio->bi_rw = WRITE;
		wbio->bi_vcnt = vcnt;
		wbio->bi_size = r1_bio->sectors << 9;
		wbio->bi_idx = idx;

		md_trim_bio(wbio, sector - r1_bio->sector, sectors);
		wbio->bi_sector += rdev->data_offset;
		wbio->bi_bdev = rdev->bdev;
		if (submit_bio_wait(WRITE, wbio) == 0)
			/* failure! */
			ok = rdev_set_badblocks(rdev, sector,
						sectors, 0)
				&& ok;

		bio_put(wbio);
		sect_to_write -= sectors;
		sector += sectors;
		sectors = block_sectors;
	}
	return ok;
}

1869
static void handle_sync_write_finished(conf_t *conf, struct r1bio *r1_bio)
1870 1871 1872 1873
{
	int m;
	int s = r1_bio->sectors;
	for (m = 0; m < conf->raid_disks ; m++) {
1874
		struct md_rdev *rdev = conf->mirrors[m].rdev;
1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891
		struct bio *bio = r1_bio->bios[m];
		if (bio->bi_end_io == NULL)
			continue;
		if (test_bit(BIO_UPTODATE, &bio->bi_flags) &&
		    test_bit(R1BIO_MadeGood, &r1_bio->state)) {
			rdev_clear_badblocks(rdev, r1_bio->sector, s);
		}
		if (!test_bit(BIO_UPTODATE, &bio->bi_flags) &&
		    test_bit(R1BIO_WriteError, &r1_bio->state)) {
			if (!rdev_set_badblocks(rdev, r1_bio->sector, s, 0))
				md_error(conf->mddev, rdev);
		}
	}
	put_buf(r1_bio);
	md_done_sync(conf->mddev, s, 1);
}

1892
static void handle_write_finished(conf_t *conf, struct r1bio *r1_bio)
1893 1894 1895 1896
{
	int m;
	for (m = 0; m < conf->raid_disks ; m++)
		if (r1_bio->bios[m] == IO_MADE_GOOD) {
1897
			struct md_rdev *rdev = conf->mirrors[m].rdev;
1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920
			rdev_clear_badblocks(rdev,
					     r1_bio->sector,
					     r1_bio->sectors);
			rdev_dec_pending(rdev, conf->mddev);
		} else if (r1_bio->bios[m] != NULL) {
			/* This drive got a write error.  We need to
			 * narrow down and record precise write
			 * errors.
			 */
			if (!narrow_write_error(r1_bio, m)) {
				md_error(conf->mddev,
					 conf->mirrors[m].rdev);
				/* an I/O failed, we can't clear the bitmap */
				set_bit(R1BIO_Degraded, &r1_bio->state);
			}
			rdev_dec_pending(conf->mirrors[m].rdev,
					 conf->mddev);
		}
	if (test_bit(R1BIO_WriteError, &r1_bio->state))
		close_write(r1_bio);
	raid_end_bio_io(r1_bio);
}

1921
static void handle_read_error(conf_t *conf, struct r1bio *r1_bio)
1922 1923 1924
{
	int disk;
	int max_sectors;
1925
	struct mddev *mddev = conf->mddev;
1926 1927
	struct bio *bio;
	char b[BDEVNAME_SIZE];
1928
	struct md_rdev *rdev;
1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

	clear_bit(R1BIO_ReadError, &r1_bio->state);
	/* 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
	 */
	if (mddev->ro == 0) {
		freeze_array(conf);
		fix_read_error(conf, r1_bio->read_disk,
			       r1_bio->sector, r1_bio->sectors);
		unfreeze_array(conf);
	} else
		md_error(mddev, conf->mirrors[r1_bio->read_disk].rdev);

	bio = r1_bio->bios[r1_bio->read_disk];
	bdevname(bio->bi_bdev, b);
read_more:
	disk = read_balance(conf, r1_bio, &max_sectors);
	if (disk == -1) {
		printk(KERN_ALERT "md/raid1:%s: %s: unrecoverable I/O"
		       " read error for block %llu\n",
		       mdname(mddev), b, (unsigned long long)r1_bio->sector);
		raid_end_bio_io(r1_bio);
	} else {
		const unsigned long do_sync
			= r1_bio->master_bio->bi_rw & REQ_SYNC;
		if (bio) {
			r1_bio->bios[r1_bio->read_disk] =
				mddev->ro ? IO_BLOCKED : NULL;
			bio_put(bio);
		}
		r1_bio->read_disk = disk;
		bio = bio_clone_mddev(r1_bio->master_bio, GFP_NOIO, mddev);
		md_trim_bio(bio, r1_bio->sector - bio->bi_sector, max_sectors);
		r1_bio->bios[r1_bio->read_disk] = bio;
		rdev = conf->mirrors[disk].rdev;
		printk_ratelimited(KERN_ERR
				   "md/raid1:%s: redirecting sector %llu"
				   " to other mirror: %s\n",
				   mdname(mddev),
				   (unsigned long long)r1_bio->sector,
				   bdevname(rdev->bdev, b));
		bio->bi_sector = r1_bio->sector + rdev->data_offset;
		bio->bi_bdev = rdev->bdev;
		bio->bi_end_io = raid1_end_read_request;
		bio->bi_rw = READ | do_sync;
		bio->bi_private = r1_bio;
		if (max_sectors < r1_bio->sectors) {
			/* Drat - have to split this up more */
			struct bio *mbio = r1_bio->master_bio;
			int sectors_handled = (r1_bio->sector + max_sectors
					       - mbio->bi_sector);
			r1_bio->sectors = max_sectors;
			spin_lock_irq(&conf->device_lock);
			if (mbio->bi_phys_segments == 0)
				mbio->bi_phys_segments = 2;
			else
				mbio->bi_phys_segments++;
			spin_unlock_irq(&conf->device_lock);
			generic_make_request(bio);
			bio = NULL;

			r1_bio = mempool_alloc(conf->r1bio_pool, GFP_NOIO);

			r1_bio->master_bio = mbio;
			r1_bio->sectors = (mbio->bi_size >> 9)
					  - sectors_handled;
			r1_bio->state = 0;
			set_bit(R1BIO_ReadError, &r1_bio->state);
			r1_bio->mddev = mddev;
			r1_bio->sector = mbio->bi_sector + sectors_handled;

			goto read_more;
		} else
			generic_make_request(bio);
	}
}

2011
static void raid1d(struct mddev *mddev)
L
Linus Torvalds 已提交
2012
{
2013
	struct r1bio *r1_bio;
L
Linus Torvalds 已提交
2014
	unsigned long flags;
2015
	conf_t *conf = mddev->private;
L
Linus Torvalds 已提交
2016
	struct list_head *head = &conf->retry_list;
2017
	struct blk_plug plug;
L
Linus Torvalds 已提交
2018 2019

	md_check_recovery(mddev);
2020 2021

	blk_start_plug(&plug);
L
Linus Torvalds 已提交
2022
	for (;;) {
2023

N
NeilBrown 已提交
2024 2025
		if (atomic_read(&mddev->plug_cnt) == 0)
			flush_pending_writes(conf);
2026

2027 2028 2029
		spin_lock_irqsave(&conf->device_lock, flags);
		if (list_empty(head)) {
			spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
2030
			break;
2031
		}
2032
		r1_bio = list_entry(head->prev, struct r1bio, retry_list);
L
Linus Torvalds 已提交
2033
		list_del(head->prev);
2034
		conf->nr_queued--;
L
Linus Torvalds 已提交
2035 2036 2037
		spin_unlock_irqrestore(&conf->device_lock, flags);

		mddev = r1_bio->mddev;
2038
		conf = mddev->private;
2039
		if (test_bit(R1BIO_IsSync, &r1_bio->state)) {
2040
			if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
2041 2042 2043
			    test_bit(R1BIO_WriteError, &r1_bio->state))
				handle_sync_write_finished(conf, r1_bio);
			else
2044
				sync_request_write(mddev, r1_bio);
2045
		} else if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
2046 2047 2048 2049 2050
			   test_bit(R1BIO_WriteError, &r1_bio->state))
			handle_write_finished(conf, r1_bio);
		else if (test_bit(R1BIO_ReadError, &r1_bio->state))
			handle_read_error(conf, r1_bio);
		else
2051 2052 2053 2054
			/* just a partial read to be scheduled from separate
			 * context
			 */
			generic_make_request(r1_bio->bios[r1_bio->read_disk]);
2055

N
NeilBrown 已提交
2056
		cond_resched();
2057 2058
		if (mddev->flags & ~(1<<MD_CHANGE_PENDING))
			md_check_recovery(mddev);
L
Linus Torvalds 已提交
2059
	}
2060
	blk_finish_plug(&plug);
L
Linus Torvalds 已提交
2061 2062 2063 2064 2065 2066 2067 2068
}


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

	buffs = RESYNC_WINDOW / RESYNC_BLOCK_SIZE;
2069
	BUG_ON(conf->r1buf_pool);
L
Linus Torvalds 已提交
2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087
	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.
 */

2088
static sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipped, int go_faster)
L
Linus Torvalds 已提交
2089
{
2090
	conf_t *conf = mddev->private;
2091
	struct r1bio *r1_bio;
L
Linus Torvalds 已提交
2092 2093
	struct bio *bio;
	sector_t max_sector, nr_sectors;
2094
	int disk = -1;
L
Linus Torvalds 已提交
2095
	int i;
2096 2097
	int wonly = -1;
	int write_targets = 0, read_targets = 0;
N
NeilBrown 已提交
2098
	sector_t sync_blocks;
2099
	int still_degraded = 0;
2100 2101
	int good_sectors = RESYNC_SECTORS;
	int min_bad = 0; /* number of sectors that are bad in all devices */
L
Linus Torvalds 已提交
2102 2103 2104

	if (!conf->r1buf_pool)
		if (init_resync(conf))
2105
			return 0;
L
Linus Torvalds 已提交
2106

A
Andre Noll 已提交
2107
	max_sector = mddev->dev_sectors;
L
Linus Torvalds 已提交
2108
	if (sector_nr >= max_sector) {
2109 2110 2111 2112 2113
		/* 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
		 */
2114 2115
		if (mddev->curr_resync < max_sector) /* aborted */
			bitmap_end_sync(mddev->bitmap, mddev->curr_resync,
2116
						&sync_blocks, 1);
2117
		else /* completed sync */
2118
			conf->fullsync = 0;
2119 2120

		bitmap_close_sync(mddev->bitmap);
L
Linus Torvalds 已提交
2121 2122 2123 2124
		close_sync(conf);
		return 0;
	}

2125 2126
	if (mddev->bitmap == NULL &&
	    mddev->recovery_cp == MaxSector &&
2127
	    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
2128 2129 2130 2131
	    conf->fullsync == 0) {
		*skipped = 1;
		return max_sector - sector_nr;
	}
2132 2133 2134
	/* before building a request, check if we can skip these blocks..
	 * This call the bitmap_start_sync doesn't actually record anything
	 */
2135
	if (!bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, 1) &&
2136
	    !conf->fullsync && !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
2137 2138 2139 2140
		/* We can skip this block, and probably several more */
		*skipped = 1;
		return sync_blocks;
	}
L
Linus Torvalds 已提交
2141
	/*
2142 2143 2144
	 * 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
2145
	 */
2146
	if (!go_faster && conf->nr_waiting)
L
Linus Torvalds 已提交
2147
		msleep_interruptible(1000);
2148

N
NeilBrown 已提交
2149
	bitmap_cond_end_sync(mddev->bitmap, sector_nr);
2150
	r1_bio = mempool_alloc(conf->r1buf_pool, GFP_NOIO);
2151 2152 2153
	raise_barrier(conf);

	conf->next_resync = sector_nr;
L
Linus Torvalds 已提交
2154

2155
	rcu_read_lock();
L
Linus Torvalds 已提交
2156
	/*
2157 2158 2159 2160 2161 2162
	 * 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
2163 2164 2165 2166
	 */

	r1_bio->mddev = mddev;
	r1_bio->sector = sector_nr;
2167
	r1_bio->state = 0;
L
Linus Torvalds 已提交
2168 2169 2170
	set_bit(R1BIO_IsSync, &r1_bio->state);

	for (i=0; i < conf->raid_disks; i++) {
2171
		struct md_rdev *rdev;
L
Linus Torvalds 已提交
2172 2173 2174 2175
		bio = r1_bio->bios[i];

		/* take from bio_init */
		bio->bi_next = NULL;
2176
		bio->bi_flags &= ~(BIO_POOL_MASK-1);
L
Linus Torvalds 已提交
2177
		bio->bi_flags |= 1 << BIO_UPTODATE;
2178
		bio->bi_comp_cpu = -1;
2179
		bio->bi_rw = READ;
L
Linus Torvalds 已提交
2180 2181 2182 2183 2184 2185 2186
		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;

2187 2188
		rdev = rcu_dereference(conf->mirrors[i].rdev);
		if (rdev == NULL ||
2189
		    test_bit(Faulty, &rdev->flags)) {
2190
			still_degraded = 1;
2191
		} else if (!test_bit(In_sync, &rdev->flags)) {
L
Linus Torvalds 已提交
2192 2193 2194
			bio->bi_rw = WRITE;
			bio->bi_end_io = end_sync_write;
			write_targets ++;
2195 2196
		} else {
			/* may need to read from here */
2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221
			sector_t first_bad = MaxSector;
			int bad_sectors;

			if (is_badblock(rdev, sector_nr, good_sectors,
					&first_bad, &bad_sectors)) {
				if (first_bad > sector_nr)
					good_sectors = first_bad - sector_nr;
				else {
					bad_sectors -= (sector_nr - first_bad);
					if (min_bad == 0 ||
					    min_bad > bad_sectors)
						min_bad = bad_sectors;
				}
			}
			if (sector_nr < first_bad) {
				if (test_bit(WriteMostly, &rdev->flags)) {
					if (wonly < 0)
						wonly = i;
				} else {
					if (disk < 0)
						disk = i;
				}
				bio->bi_rw = READ;
				bio->bi_end_io = end_sync_read;
				read_targets++;
2222 2223
			}
		}
2224 2225 2226 2227 2228 2229
		if (bio->bi_end_io) {
			atomic_inc(&rdev->nr_pending);
			bio->bi_sector = sector_nr + rdev->data_offset;
			bio->bi_bdev = rdev->bdev;
			bio->bi_private = r1_bio;
		}
L
Linus Torvalds 已提交
2230
	}
2231 2232 2233 2234
	rcu_read_unlock();
	if (disk < 0)
		disk = wonly;
	r1_bio->read_disk = disk;
2235

2236 2237 2238 2239 2240 2241 2242
	if (read_targets == 0 && min_bad > 0) {
		/* These sectors are bad on all InSync devices, so we
		 * need to mark them bad on all write targets
		 */
		int ok = 1;
		for (i = 0 ; i < conf->raid_disks ; i++)
			if (r1_bio->bios[i]->bi_end_io == end_sync_write) {
2243
				struct md_rdev *rdev =
2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271
					rcu_dereference(conf->mirrors[i].rdev);
				ok = rdev_set_badblocks(rdev, sector_nr,
							min_bad, 0
					) && ok;
			}
		set_bit(MD_CHANGE_DEVS, &mddev->flags);
		*skipped = 1;
		put_buf(r1_bio);

		if (!ok) {
			/* Cannot record the badblocks, so need to
			 * abort the resync.
			 * If there are multiple read targets, could just
			 * fail the really bad ones ???
			 */
			conf->recovery_disabled = mddev->recovery_disabled;
			set_bit(MD_RECOVERY_INTR, &mddev->recovery);
			return 0;
		} else
			return min_bad;

	}
	if (min_bad > 0 && min_bad < good_sectors) {
		/* only resync enough to reach the next bad->good
		 * transition */
		good_sectors = min_bad;
	}

2272 2273 2274 2275 2276
	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
2277 2278 2279
		/* There is nowhere to write, so all non-sync
		 * drives must be failed - so we are finished
		 */
2280 2281
		sector_t rv = max_sector - sector_nr;
		*skipped = 1;
L
Linus Torvalds 已提交
2282 2283 2284 2285
		put_buf(r1_bio);
		return rv;
	}

2286 2287
	if (max_sector > mddev->resync_max)
		max_sector = mddev->resync_max; /* Don't do IO beyond here */
2288 2289
	if (max_sector > sector_nr + good_sectors)
		max_sector = sector_nr + good_sectors;
L
Linus Torvalds 已提交
2290
	nr_sectors = 0;
2291
	sync_blocks = 0;
L
Linus Torvalds 已提交
2292 2293 2294 2295 2296 2297 2298
	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;
2299 2300
		if (sync_blocks == 0) {
			if (!bitmap_start_sync(mddev->bitmap, sector_nr,
2301 2302 2303
					       &sync_blocks, still_degraded) &&
			    !conf->fullsync &&
			    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
2304
				break;
2305
			BUG_ON(sync_blocks < (PAGE_SIZE>>9));
2306
			if ((len >> 9) > sync_blocks)
2307
				len = sync_blocks<<9;
2308
		}
2309

L
Linus Torvalds 已提交
2310 2311 2312
		for (i=0 ; i < conf->raid_disks; i++) {
			bio = r1_bio->bios[i];
			if (bio->bi_end_io) {
2313
				page = bio->bi_io_vec[bio->bi_vcnt].bv_page;
L
Linus Torvalds 已提交
2314 2315
				if (bio_add_page(bio, page, len, 0) == 0) {
					/* stop here */
2316
					bio->bi_io_vec[bio->bi_vcnt].bv_page = page;
L
Linus Torvalds 已提交
2317 2318 2319
					while (i > 0) {
						i--;
						bio = r1_bio->bios[i];
2320 2321
						if (bio->bi_end_io==NULL)
							continue;
L
Linus Torvalds 已提交
2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332
						/* 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;
2333
		sync_blocks -= (len>>9);
L
Linus Torvalds 已提交
2334 2335 2336 2337
	} while (r1_bio->bios[disk]->bi_vcnt < RESYNC_PAGES);
 bio_full:
	r1_bio->sectors = nr_sectors;

2338 2339 2340 2341 2342 2343 2344 2345
	/* 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) {
2346
				md_sync_acct(bio->bi_bdev, nr_sectors);
2347 2348 2349 2350 2351 2352
				generic_make_request(bio);
			}
		}
	} else {
		atomic_set(&r1_bio->remaining, 1);
		bio = r1_bio->bios[r1_bio->read_disk];
2353
		md_sync_acct(bio->bi_bdev, nr_sectors);
2354
		generic_make_request(bio);
L
Linus Torvalds 已提交
2355

2356
	}
L
Linus Torvalds 已提交
2357 2358 2359
	return nr_sectors;
}

2360
static sector_t raid1_size(struct mddev *mddev, sector_t sectors, int raid_disks)
2361 2362 2363 2364 2365 2366 2367
{
	if (sectors)
		return sectors;

	return mddev->dev_sectors;
}

2368
static conf_t *setup_conf(struct mddev *mddev)
L
Linus Torvalds 已提交
2369 2370
{
	conf_t *conf;
2371
	int i;
2372
	struct mirror_info *disk;
2373
	struct md_rdev *rdev;
2374
	int err = -ENOMEM;
L
Linus Torvalds 已提交
2375

2376
	conf = kzalloc(sizeof(conf_t), GFP_KERNEL);
L
Linus Torvalds 已提交
2377
	if (!conf)
2378
		goto abort;
L
Linus Torvalds 已提交
2379

2380
	conf->mirrors = kzalloc(sizeof(struct mirror_info)*mddev->raid_disks,
L
Linus Torvalds 已提交
2381 2382
				 GFP_KERNEL);
	if (!conf->mirrors)
2383
		goto abort;
L
Linus Torvalds 已提交
2384

2385 2386
	conf->tmppage = alloc_page(GFP_KERNEL);
	if (!conf->tmppage)
2387
		goto abort;
2388

2389
	conf->poolinfo = kzalloc(sizeof(*conf->poolinfo), GFP_KERNEL);
L
Linus Torvalds 已提交
2390
	if (!conf->poolinfo)
2391
		goto abort;
L
Linus Torvalds 已提交
2392 2393 2394 2395 2396
	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)
2397 2398
		goto abort;

2399
	conf->poolinfo->mddev = mddev;
L
Linus Torvalds 已提交
2400

2401
	spin_lock_init(&conf->device_lock);
2402
	list_for_each_entry(rdev, &mddev->disks, same_set) {
2403
		int disk_idx = rdev->raid_disk;
L
Linus Torvalds 已提交
2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417
		if (disk_idx >= mddev->raid_disks
		    || disk_idx < 0)
			continue;
		disk = conf->mirrors + disk_idx;

		disk->rdev = rdev;

		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);
2418
	init_waitqueue_head(&conf->wait_barrier);
L
Linus Torvalds 已提交
2419

2420 2421
	bio_list_init(&conf->pending_bio_list);

2422
	conf->last_used = -1;
L
Linus Torvalds 已提交
2423 2424 2425 2426
	for (i = 0; i < conf->raid_disks; i++) {

		disk = conf->mirrors + i;

2427 2428
		if (!disk->rdev ||
		    !test_bit(In_sync, &disk->rdev->flags)) {
L
Linus Torvalds 已提交
2429
			disk->head_position = 0;
2430 2431
			if (disk->rdev)
				conf->fullsync = 1;
2432 2433 2434 2435 2436 2437
		} else if (conf->last_used < 0)
			/*
			 * The first working device is used as a
			 * starting point to read balancing.
			 */
			conf->last_used = i;
L
Linus Torvalds 已提交
2438
	}
2439 2440 2441

	err = -EIO;
	if (conf->last_used < 0) {
N
NeilBrown 已提交
2442
		printk(KERN_ERR "md/raid1:%s: no operational mirrors\n",
2443 2444 2445 2446 2447 2448 2449
		       mdname(mddev));
		goto abort;
	}
	err = -ENOMEM;
	conf->thread = md_register_thread(raid1d, mddev, NULL);
	if (!conf->thread) {
		printk(KERN_ERR
N
NeilBrown 已提交
2450
		       "md/raid1:%s: couldn't allocate thread\n",
2451 2452
		       mdname(mddev));
		goto abort;
2453
	}
L
Linus Torvalds 已提交
2454

2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468
	return conf;

 abort:
	if (conf) {
		if (conf->r1bio_pool)
			mempool_destroy(conf->r1bio_pool);
		kfree(conf->mirrors);
		safe_put_page(conf->tmppage);
		kfree(conf->poolinfo);
		kfree(conf);
	}
	return ERR_PTR(err);
}

2469
static int run(struct mddev *mddev)
2470 2471 2472
{
	conf_t *conf;
	int i;
2473
	struct md_rdev *rdev;
2474 2475

	if (mddev->level != 1) {
N
NeilBrown 已提交
2476
		printk(KERN_ERR "md/raid1:%s: raid level not set to mirroring (%d)\n",
2477 2478 2479 2480
		       mdname(mddev), mddev->level);
		return -EIO;
	}
	if (mddev->reshape_position != MaxSector) {
N
NeilBrown 已提交
2481
		printk(KERN_ERR "md/raid1:%s: reshape_position set but not supported\n",
2482 2483 2484
		       mdname(mddev));
		return -EIO;
	}
L
Linus Torvalds 已提交
2485
	/*
2486 2487 2488
	 * copy the already verified devices into our private RAID1
	 * bookkeeping area. [whatever we allocate in run(),
	 * should be freed in stop()]
L
Linus Torvalds 已提交
2489
	 */
2490 2491 2492 2493
	if (mddev->private == NULL)
		conf = setup_conf(mddev);
	else
		conf = mddev->private;
L
Linus Torvalds 已提交
2494

2495 2496
	if (IS_ERR(conf))
		return PTR_ERR(conf);
L
Linus Torvalds 已提交
2497

2498
	list_for_each_entry(rdev, &mddev->disks, same_set) {
2499 2500
		if (!mddev->gendisk)
			continue;
2501 2502 2503
		disk_stack_limits(mddev->gendisk, rdev->bdev,
				  rdev->data_offset << 9);
		/* as we don't honour merge_bvec_fn, we must never risk
2504 2505
		 * violating it, so limit ->max_segments to 1 lying within
		 * a single page, as a one page request is never in violation.
2506
		 */
2507 2508 2509 2510 2511
		if (rdev->bdev->bd_disk->queue->merge_bvec_fn) {
			blk_queue_max_segments(mddev->queue, 1);
			blk_queue_segment_boundary(mddev->queue,
						   PAGE_CACHE_SIZE - 1);
		}
L
Linus Torvalds 已提交
2512
	}
2513

2514 2515 2516 2517 2518 2519 2520 2521 2522 2523
	mddev->degraded = 0;
	for (i=0; i < conf->raid_disks; i++)
		if (conf->mirrors[i].rdev == NULL ||
		    !test_bit(In_sync, &conf->mirrors[i].rdev->flags) ||
		    test_bit(Faulty, &conf->mirrors[i].rdev->flags))
			mddev->degraded++;

	if (conf->raid_disks - mddev->degraded == 1)
		mddev->recovery_cp = MaxSector;

2524
	if (mddev->recovery_cp != MaxSector)
N
NeilBrown 已提交
2525
		printk(KERN_NOTICE "md/raid1:%s: not clean"
2526 2527
		       " -- starting background reconstruction\n",
		       mdname(mddev));
L
Linus Torvalds 已提交
2528
	printk(KERN_INFO 
N
NeilBrown 已提交
2529
		"md/raid1:%s: active with %d out of %d mirrors\n",
L
Linus Torvalds 已提交
2530 2531
		mdname(mddev), mddev->raid_disks - mddev->degraded, 
		mddev->raid_disks);
2532

L
Linus Torvalds 已提交
2533 2534 2535
	/*
	 * Ok, everything is just fine now
	 */
2536 2537 2538 2539
	mddev->thread = conf->thread;
	conf->thread = NULL;
	mddev->private = conf;

2540
	md_set_array_sectors(mddev, raid1_size(mddev, 0, 0));
L
Linus Torvalds 已提交
2541

2542 2543 2544 2545
	if (mddev->queue) {
		mddev->queue->backing_dev_info.congested_fn = raid1_congested;
		mddev->queue->backing_dev_info.congested_data = mddev;
	}
2546
	return md_integrity_register(mddev);
L
Linus Torvalds 已提交
2547 2548
}

2549
static int stop(struct mddev *mddev)
L
Linus Torvalds 已提交
2550
{
2551
	conf_t *conf = mddev->private;
2552 2553 2554
	struct bitmap *bitmap = mddev->bitmap;

	/* wait for behind writes to complete */
2555
	if (bitmap && atomic_read(&bitmap->behind_writes) > 0) {
N
NeilBrown 已提交
2556 2557
		printk(KERN_INFO "md/raid1:%s: behind writes in progress - waiting to stop.\n",
		       mdname(mddev));
2558
		/* need to kick something here to make sure I/O goes? */
2559 2560
		wait_event(bitmap->behind_wait,
			   atomic_read(&bitmap->behind_writes) == 0);
2561
	}
L
Linus Torvalds 已提交
2562

2563 2564 2565
	raise_barrier(conf);
	lower_barrier(conf);

2566
	md_unregister_thread(&mddev->thread);
L
Linus Torvalds 已提交
2567 2568
	if (conf->r1bio_pool)
		mempool_destroy(conf->r1bio_pool);
2569 2570
	kfree(conf->mirrors);
	kfree(conf->poolinfo);
L
Linus Torvalds 已提交
2571 2572 2573 2574 2575
	kfree(conf);
	mddev->private = NULL;
	return 0;
}

2576
static int raid1_resize(struct mddev *mddev, sector_t sectors)
L
Linus Torvalds 已提交
2577 2578 2579 2580 2581 2582 2583 2584
{
	/* 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.
	 */
2585
	md_set_array_sectors(mddev, raid1_size(mddev, sectors, 0));
D
Dan Williams 已提交
2586 2587
	if (mddev->array_sectors > raid1_size(mddev, sectors, 0))
		return -EINVAL;
2588
	set_capacity(mddev->gendisk, mddev->array_sectors);
2589
	revalidate_disk(mddev->gendisk);
D
Dan Williams 已提交
2590
	if (sectors > mddev->dev_sectors &&
2591
	    mddev->recovery_cp > mddev->dev_sectors) {
A
Andre Noll 已提交
2592
		mddev->recovery_cp = mddev->dev_sectors;
L
Linus Torvalds 已提交
2593 2594
		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
	}
D
Dan Williams 已提交
2595
	mddev->dev_sectors = sectors;
2596
	mddev->resync_max_sectors = sectors;
L
Linus Torvalds 已提交
2597 2598 2599
	return 0;
}

2600
static int raid1_reshape(struct mddev *mddev)
L
Linus Torvalds 已提交
2601 2602 2603 2604 2605 2606 2607 2608
{
	/* 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.
2609 2610 2611
	 *
	 * 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
2612 2613 2614
	 */
	mempool_t *newpool, *oldpool;
	struct pool_info *newpoolinfo;
2615
	struct mirror_info *newmirrors;
2616
	conf_t *conf = mddev->private;
2617
	int cnt, raid_disks;
2618
	unsigned long flags;
2619
	int d, d2, err;
L
Linus Torvalds 已提交
2620

2621
	/* Cannot change chunk_size, layout, or level */
2622
	if (mddev->chunk_sectors != mddev->new_chunk_sectors ||
2623 2624
	    mddev->layout != mddev->new_layout ||
	    mddev->level != mddev->new_level) {
2625
		mddev->new_chunk_sectors = mddev->chunk_sectors;
2626 2627 2628 2629 2630
		mddev->new_layout = mddev->layout;
		mddev->new_level = mddev->level;
		return -EINVAL;
	}

2631 2632 2633
	err = md_allow_write(mddev);
	if (err)
		return err;
2634

2635 2636
	raid_disks = mddev->raid_disks + mddev->delta_disks;

2637 2638 2639 2640 2641 2642
	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
2643
			return -EBUSY;
2644
	}
L
Linus Torvalds 已提交
2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657

	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;
	}
2658
	newmirrors = kzalloc(sizeof(struct mirror_info) * raid_disks, GFP_KERNEL);
L
Linus Torvalds 已提交
2659 2660 2661 2662 2663 2664
	if (!newmirrors) {
		kfree(newpoolinfo);
		mempool_destroy(newpool);
		return -ENOMEM;
	}

2665
	raise_barrier(conf);
L
Linus Torvalds 已提交
2666 2667 2668 2669

	/* ok, everything is stopped */
	oldpool = conf->r1bio_pool;
	conf->r1bio_pool = newpool;
2670

2671
	for (d = d2 = 0; d < conf->raid_disks; d++) {
2672
		struct md_rdev *rdev = conf->mirrors[d].rdev;
2673
		if (rdev && rdev->raid_disk != d2) {
2674
			sysfs_unlink_rdev(mddev, rdev);
2675
			rdev->raid_disk = d2;
2676 2677
			sysfs_unlink_rdev(mddev, rdev);
			if (sysfs_link_rdev(mddev, rdev))
2678
				printk(KERN_WARNING
2679 2680
				       "md/raid1:%s: cannot register rd%d\n",
				       mdname(mddev), rdev->raid_disk);
2681
		}
2682 2683 2684
		if (rdev)
			newmirrors[d2++].rdev = rdev;
	}
L
Linus Torvalds 已提交
2685 2686 2687 2688 2689
	kfree(conf->mirrors);
	conf->mirrors = newmirrors;
	kfree(conf->poolinfo);
	conf->poolinfo = newpoolinfo;

2690
	spin_lock_irqsave(&conf->device_lock, flags);
L
Linus Torvalds 已提交
2691
	mddev->degraded += (raid_disks - conf->raid_disks);
2692
	spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
2693
	conf->raid_disks = mddev->raid_disks = raid_disks;
2694
	mddev->delta_disks = 0;
L
Linus Torvalds 已提交
2695

2696
	conf->last_used = 0; /* just make sure it is in-range */
2697
	lower_barrier(conf);
L
Linus Torvalds 已提交
2698 2699 2700 2701 2702 2703 2704 2705

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

	mempool_destroy(oldpool);
	return 0;
}

2706
static void raid1_quiesce(struct mddev *mddev, int state)
2707
{
2708
	conf_t *conf = mddev->private;
2709 2710

	switch(state) {
2711 2712 2713
	case 2: /* wake for suspend */
		wake_up(&conf->wait_barrier);
		break;
2714
	case 1:
2715
		raise_barrier(conf);
2716
		break;
2717
	case 0:
2718
		lower_barrier(conf);
2719 2720 2721 2722
		break;
	}
}

2723
static void *raid1_takeover(struct mddev *mddev)
2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739
{
	/* raid1 can take over:
	 *  raid5 with 2 devices, any layout or chunk size
	 */
	if (mddev->level == 5 && mddev->raid_disks == 2) {
		conf_t *conf;
		mddev->new_level = 1;
		mddev->new_layout = 0;
		mddev->new_chunk_sectors = 0;
		conf = setup_conf(mddev);
		if (!IS_ERR(conf))
			conf->barrier = 1;
		return conf;
	}
	return ERR_PTR(-EINVAL);
}
L
Linus Torvalds 已提交
2740

2741
static struct mdk_personality raid1_personality =
L
Linus Torvalds 已提交
2742 2743
{
	.name		= "raid1",
2744
	.level		= 1,
L
Linus Torvalds 已提交
2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755
	.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,
2756
	.size		= raid1_size,
2757
	.check_reshape	= raid1_reshape,
2758
	.quiesce	= raid1_quiesce,
2759
	.takeover	= raid1_takeover,
L
Linus Torvalds 已提交
2760 2761 2762 2763
};

static int __init raid_init(void)
{
2764
	return register_md_personality(&raid1_personality);
L
Linus Torvalds 已提交
2765 2766 2767 2768
}

static void raid_exit(void)
{
2769
	unregister_md_personality(&raid1_personality);
L
Linus Torvalds 已提交
2770 2771 2772 2773 2774
}

module_init(raid_init);
module_exit(raid_exit);
MODULE_LICENSE("GPL");
2775
MODULE_DESCRIPTION("RAID1 (mirroring) personality for MD");
L
Linus Torvalds 已提交
2776
MODULE_ALIAS("md-personality-3"); /* RAID1 */
2777
MODULE_ALIAS("md-raid1");
2778
MODULE_ALIAS("md-level-1");