raid1.c 83.1 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
#include <linux/blkdev.h>
37
#include <linux/module.h>
38
#include <linux/seq_file.h>
39
#include <linux/ratelimit.h>
40
#include "md.h"
41 42
#include "raid1.h"
#include "bitmap.h"
43

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

49 50 51 52 53 54 55 56 57 58 59 60 61 62
/* when we get a read error on a read-only array, we redirect to another
 * device without failing the first device, or trying to over-write to
 * correct the read error.  To keep track of bad blocks on a per-bio
 * level, we store IO_BLOCKED in the appropriate 'bios' pointer
 */
#define IO_BLOCKED ((struct bio *)1)
/* When we successfully write to a known bad-block, we need to remove the
 * bad-block marking which must be done from process context.  So we record
 * the success by setting devs[n].bio to IO_MADE_GOOD
 */
#define IO_MADE_GOOD ((struct bio *)2)

#define BIO_SPECIAL(bio) ((unsigned long)bio <= 2)

63 64 65 66 67
/* When there are this many requests queue to be written by
 * the raid1 thread, we become 'congested' to provide back-pressure
 * for writeback.
 */
static int max_queued_requests = 1024;
L
Linus Torvalds 已提交
68

69 70
static void allow_barrier(struct r1conf *conf);
static void lower_barrier(struct r1conf *conf);
L
Linus Torvalds 已提交
71

A
Al Viro 已提交
72
static void * r1bio_pool_alloc(gfp_t gfp_flags, void *data)
L
Linus Torvalds 已提交
73 74
{
	struct pool_info *pi = data;
75
	int size = offsetof(struct r1bio, bios[pi->raid_disks]);
L
Linus Torvalds 已提交
76 77

	/* allocate a r1bio with room for raid_disks entries in the bios array */
J
Jens Axboe 已提交
78
	return kzalloc(size, gfp_flags);
L
Linus Torvalds 已提交
79 80 81 82 83 84 85 86 87 88 89 90 91
}

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
92
static void * r1buf_pool_alloc(gfp_t gfp_flags, void *data)
L
Linus Torvalds 已提交
93 94 95
{
	struct pool_info *pi = data;
	struct page *page;
96
	struct r1bio *r1_bio;
L
Linus Torvalds 已提交
97 98 99 100
	struct bio *bio;
	int i, j;

	r1_bio = r1bio_pool_alloc(gfp_flags, pi);
J
Jens Axboe 已提交
101
	if (!r1_bio)
L
Linus Torvalds 已提交
102 103 104 105 106 107
		return NULL;

	/*
	 * Allocate bios : 1 for reading, n-1 for writing
	 */
	for (j = pi->raid_disks ; j-- ; ) {
108
		bio = bio_kmalloc(gfp_flags, RESYNC_PAGES);
L
Linus Torvalds 已提交
109 110 111 112 113 114
		if (!bio)
			goto out_free_bio;
		r1_bio->bios[j] = bio;
	}
	/*
	 * Allocate RESYNC_PAGES data pages and attach them to
115 116 117
	 * the first bio.
	 * If this is a user-requested check/repair, allocate
	 * RESYNC_PAGES for each bio.
L
Linus Torvalds 已提交
118
	 */
119 120 121 122 123 124 125 126 127 128 129 130
	if (test_bit(MD_RECOVERY_REQUESTED, &pi->mddev->recovery))
		j = pi->raid_disks;
	else
		j = 1;
	while(j--) {
		bio = r1_bio->bios[j];
		for (i = 0; i < RESYNC_PAGES; i++) {
			page = alloc_page(gfp_flags);
			if (unlikely(!page))
				goto out_free_pages;

			bio->bi_io_vec[i].bv_page = page;
131
			bio->bi_vcnt = i+1;
132 133 134 135 136 137 138 139
		}
	}
	/* 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
140 141 142 143 144 145 146
	}

	r1_bio->master_bio = NULL;

	return r1_bio;

out_free_pages:
147 148 149
	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);
150
	j = -1;
L
Linus Torvalds 已提交
151
out_free_bio:
152
	while (++j < pi->raid_disks)
L
Linus Torvalds 已提交
153 154 155 156 157 158 159 160
		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;
161
	int i,j;
162
	struct r1bio *r1bio = __r1_bio;
L
Linus Torvalds 已提交
163

164 165 166 167 168
	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)
169
				safe_put_page(r1bio->bios[j]->bi_io_vec[i].bv_page);
170
		}
L
Linus Torvalds 已提交
171 172 173 174 175 176
	for (i=0 ; i < pi->raid_disks; i++)
		bio_put(r1bio->bios[i]);

	r1bio_pool_free(r1bio, data);
}

177
static void put_all_bios(struct r1conf *conf, struct r1bio *r1_bio)
L
Linus Torvalds 已提交
178 179 180
{
	int i;

181
	for (i = 0; i < conf->raid_disks * 2; i++) {
L
Linus Torvalds 已提交
182
		struct bio **bio = r1_bio->bios + i;
183
		if (!BIO_SPECIAL(*bio))
L
Linus Torvalds 已提交
184 185 186 187 188
			bio_put(*bio);
		*bio = NULL;
	}
}

189
static void free_r1bio(struct r1bio *r1_bio)
L
Linus Torvalds 已提交
190
{
191
	struct r1conf *conf = r1_bio->mddev->private;
L
Linus Torvalds 已提交
192 193 194 195 196

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

197
static void put_buf(struct r1bio *r1_bio)
L
Linus Torvalds 已提交
198
{
199
	struct r1conf *conf = r1_bio->mddev->private;
200 201
	int i;

202
	for (i = 0; i < conf->raid_disks * 2; i++) {
203 204 205 206
		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
207 208 209

	mempool_free(r1_bio, conf->r1buf_pool);

210
	lower_barrier(conf);
L
Linus Torvalds 已提交
211 212
}

213
static void reschedule_retry(struct r1bio *r1_bio)
L
Linus Torvalds 已提交
214 215
{
	unsigned long flags;
216
	struct mddev *mddev = r1_bio->mddev;
217
	struct r1conf *conf = mddev->private;
L
Linus Torvalds 已提交
218 219 220

	spin_lock_irqsave(&conf->device_lock, flags);
	list_add(&r1_bio->retry_list, &conf->retry_list);
221
	conf->nr_queued ++;
L
Linus Torvalds 已提交
222 223
	spin_unlock_irqrestore(&conf->device_lock, flags);

224
	wake_up(&conf->wait_barrier);
L
Linus Torvalds 已提交
225 226 227 228 229 230 231 232
	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.
 */
233
static void call_bio_endio(struct r1bio *r1_bio)
234 235 236
{
	struct bio *bio = r1_bio->master_bio;
	int done;
237
	struct r1conf *conf = r1_bio->mddev->private;
238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259

	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);
	}
}

260
static void raid_end_bio_io(struct r1bio *r1_bio)
L
Linus Torvalds 已提交
261 262 263
{
	struct bio *bio = r1_bio->master_bio;

264 265
	/* if nobody has done the final endio yet, do it now */
	if (!test_and_set_bit(R1BIO_Returned, &r1_bio->state)) {
266 267 268 269
		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 +
270
			 bio_sectors(bio) - 1);
271

272
		call_bio_endio(r1_bio);
273
	}
L
Linus Torvalds 已提交
274 275 276 277 278 279
	free_r1bio(r1_bio);
}

/*
 * Update disk head position estimator based on IRQ completion info.
 */
280
static inline void update_head_pos(int disk, struct r1bio *r1_bio)
L
Linus Torvalds 已提交
281
{
282
	struct r1conf *conf = r1_bio->mddev->private;
L
Linus Torvalds 已提交
283 284 285 286 287

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

288 289 290
/*
 * Find the disk number which triggered given bio
 */
291
static int find_bio_disk(struct r1bio *r1_bio, struct bio *bio)
292 293
{
	int mirror;
294 295
	struct r1conf *conf = r1_bio->mddev->private;
	int raid_disks = conf->raid_disks;
296

297
	for (mirror = 0; mirror < raid_disks * 2; mirror++)
298 299 300
		if (r1_bio->bios[mirror] == bio)
			break;

301
	BUG_ON(mirror == raid_disks * 2);
302 303 304 305 306
	update_head_pos(mirror, r1_bio);

	return mirror;
}

307
static void raid1_end_read_request(struct bio *bio, int error)
L
Linus Torvalds 已提交
308 309
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
310
	struct r1bio *r1_bio = bio->bi_private;
L
Linus Torvalds 已提交
311
	int mirror;
312
	struct r1conf *conf = r1_bio->mddev->private;
L
Linus Torvalds 已提交
313 314 315 316 317

	mirror = r1_bio->read_disk;
	/*
	 * this branch is our 'one mirror IO has finished' event handler:
	 */
318 319
	update_head_pos(mirror, r1_bio);

320 321 322 323 324 325
	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
326
		 */
327 328 329 330 331 332 333 334
		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
335

336
	if (uptodate) {
L
Linus Torvalds 已提交
337
		raid_end_bio_io(r1_bio);
338 339
		rdev_dec_pending(conf->mirrors[mirror].rdev, conf->mddev);
	} else {
L
Linus Torvalds 已提交
340 341 342 343
		/*
		 * oops, read error:
		 */
		char b[BDEVNAME_SIZE];
344 345 346 347 348 349 350
		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);
351
		set_bit(R1BIO_ReadError, &r1_bio->state);
L
Linus Torvalds 已提交
352
		reschedule_retry(r1_bio);
353
		/* don't drop the reference on read_disk yet */
L
Linus Torvalds 已提交
354 355 356
	}
}

357
static void close_write(struct r1bio *r1_bio)
358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375
{
	/* 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);
}

376
static void r1_bio_write_done(struct r1bio *r1_bio)
377
{
378 379 380 381 382 383 384
	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);
385 386 387 388
		if (test_bit(R1BIO_MadeGood, &r1_bio->state))
			reschedule_retry(r1_bio);
		else
			raid_end_bio_io(r1_bio);
389 390 391
	}
}

392
static void raid1_end_write_request(struct bio *bio, int error)
L
Linus Torvalds 已提交
393 394
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
395
	struct r1bio *r1_bio = bio->bi_private;
396
	int mirror, behind = test_bit(R1BIO_BehindIO, &r1_bio->state);
397
	struct r1conf *conf = r1_bio->mddev->private;
398
	struct bio *to_put = NULL;
L
Linus Torvalds 已提交
399

400
	mirror = find_bio_disk(r1_bio, bio);
L
Linus Torvalds 已提交
401

T
Tejun Heo 已提交
402 403 404 405
	/*
	 * 'one mirror IO has finished' event handler:
	 */
	if (!uptodate) {
406 407
		set_bit(WriteErrorSeen,
			&conf->mirrors[mirror].rdev->flags);
408 409 410 411 412
		if (!test_and_set_bit(WantReplacement,
				      &conf->mirrors[mirror].rdev->flags))
			set_bit(MD_RECOVERY_NEEDED, &
				conf->mddev->recovery);

413
		set_bit(R1BIO_WriteError, &r1_bio->state);
414
	} else {
L
Linus Torvalds 已提交
415
		/*
T
Tejun Heo 已提交
416 417 418 419 420 421 422 423
		 * 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
424
		 */
425 426 427
		sector_t first_bad;
		int bad_sectors;

428 429
		r1_bio->bios[mirror] = NULL;
		to_put = bio;
T
Tejun Heo 已提交
430 431
		set_bit(R1BIO_Uptodate, &r1_bio->state);

432 433 434 435 436 437 438 439 440
		/* 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
441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456
	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;
457 458 459 460
				pr_debug("raid1: behind end write sectors"
					 " %llu-%llu\n",
					 (unsigned long long) mbio->bi_sector,
					 (unsigned long long) mbio->bi_sector +
461
					 bio_sectors(mbio) - 1);
462
				call_bio_endio(r1_bio);
463 464 465
			}
		}
	}
466 467 468
	if (r1_bio->bios[mirror] == NULL)
		rdev_dec_pending(conf->mirrors[mirror].rdev,
				 conf->mddev);
T
Tejun Heo 已提交
469

L
Linus Torvalds 已提交
470 471 472 473
	/*
	 * Let's see if all mirrored write operations have finished
	 * already.
	 */
474
	r1_bio_write_done(r1_bio);
475

476 477
	if (to_put)
		bio_put(to_put);
L
Linus Torvalds 已提交
478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494
}


/*
 * 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.
 */
495
static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sectors)
L
Linus Torvalds 已提交
496
{
497
	const sector_t this_sector = r1_bio->sector;
498 499
	int sectors;
	int best_good_sectors;
500 501
	int best_disk, best_dist_disk, best_pending_disk;
	int has_nonrot_disk;
502
	int disk;
N
NeilBrown 已提交
503
	sector_t best_dist;
504
	unsigned int min_pending;
505
	struct md_rdev *rdev;
506
	int choose_first;
507
	int choose_next_idle;
L
Linus Torvalds 已提交
508 509 510

	rcu_read_lock();
	/*
511
	 * Check if we can balance. We can balance on the whole
L
Linus Torvalds 已提交
512 513 514 515
	 * device if no resync is going on, or below the resync window.
	 * We take the first readable disk when above the resync window.
	 */
 retry:
516
	sectors = r1_bio->sectors;
N
NeilBrown 已提交
517
	best_disk = -1;
518
	best_dist_disk = -1;
N
NeilBrown 已提交
519
	best_dist = MaxSector;
520 521
	best_pending_disk = -1;
	min_pending = UINT_MAX;
522
	best_good_sectors = 0;
523
	has_nonrot_disk = 0;
524
	choose_next_idle = 0;
525

L
Linus Torvalds 已提交
526
	if (conf->mddev->recovery_cp < MaxSector &&
527
	    (this_sector + sectors >= conf->next_resync))
528
		choose_first = 1;
529
	else
530
		choose_first = 0;
L
Linus Torvalds 已提交
531

532
	for (disk = 0 ; disk < conf->raid_disks * 2 ; disk++) {
N
NeilBrown 已提交
533
		sector_t dist;
534 535
		sector_t first_bad;
		int bad_sectors;
536
		unsigned int pending;
537
		bool nonrot;
538

539 540 541
		rdev = rcu_dereference(conf->mirrors[disk].rdev);
		if (r1_bio->bios[disk] == IO_BLOCKED
		    || rdev == NULL
542
		    || test_bit(Unmerged, &rdev->flags)
N
NeilBrown 已提交
543
		    || test_bit(Faulty, &rdev->flags))
544
			continue;
N
NeilBrown 已提交
545 546
		if (!test_bit(In_sync, &rdev->flags) &&
		    rdev->recovery_offset < this_sector + sectors)
L
Linus Torvalds 已提交
547
			continue;
N
NeilBrown 已提交
548 549 550
		if (test_bit(WriteMostly, &rdev->flags)) {
			/* Don't balance among write-mostly, just
			 * use the first as a last resort */
551 552 553 554 555 556 557 558 559
			if (best_disk < 0) {
				if (is_badblock(rdev, this_sector, sectors,
						&first_bad, &bad_sectors)) {
					if (first_bad < this_sector)
						/* Cannot use this */
						continue;
					best_good_sectors = first_bad - this_sector;
				} else
					best_good_sectors = sectors;
N
NeilBrown 已提交
560
				best_disk = disk;
561
			}
N
NeilBrown 已提交
562 563 564 565 566
			continue;
		}
		/* This is a reasonable device to use.  It might
		 * even be best.
		 */
567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595
		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;

596 597
		nonrot = blk_queue_nonrot(bdev_get_queue(rdev->bdev));
		has_nonrot_disk |= nonrot;
598
		pending = atomic_read(&rdev->nr_pending);
N
NeilBrown 已提交
599
		dist = abs(this_sector - conf->mirrors[disk].head_position);
600
		if (choose_first) {
N
NeilBrown 已提交
601
			best_disk = disk;
L
Linus Torvalds 已提交
602 603
			break;
		}
604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641
		/* Don't change to another disk for sequential reads */
		if (conf->mirrors[disk].next_seq_sect == this_sector
		    || dist == 0) {
			int opt_iosize = bdev_io_opt(rdev->bdev) >> 9;
			struct raid1_info *mirror = &conf->mirrors[disk];

			best_disk = disk;
			/*
			 * If buffered sequential IO size exceeds optimal
			 * iosize, check if there is idle disk. If yes, choose
			 * the idle disk. read_balance could already choose an
			 * idle disk before noticing it's a sequential IO in
			 * this disk. This doesn't matter because this disk
			 * will idle, next time it will be utilized after the
			 * first disk has IO size exceeds optimal iosize. In
			 * this way, iosize of the first disk will be optimal
			 * iosize at least. iosize of the second disk might be
			 * small, but not a big deal since when the second disk
			 * starts IO, the first disk is likely still busy.
			 */
			if (nonrot && opt_iosize > 0 &&
			    mirror->seq_start != MaxSector &&
			    mirror->next_seq_sect > opt_iosize &&
			    mirror->next_seq_sect - opt_iosize >=
			    mirror->seq_start) {
				choose_next_idle = 1;
				continue;
			}
			break;
		}
		/* If device is idle, use it */
		if (pending == 0) {
			best_disk = disk;
			break;
		}

		if (choose_next_idle)
			continue;
642 643 644 645 646 647

		if (min_pending > pending) {
			min_pending = pending;
			best_pending_disk = disk;
		}

N
NeilBrown 已提交
648 649
		if (dist < best_dist) {
			best_dist = dist;
650
			best_dist_disk = disk;
L
Linus Torvalds 已提交
651
		}
652
	}
L
Linus Torvalds 已提交
653

654 655 656 657 658 659 660 661 662 663 664 665 666
	/*
	 * If all disks are rotational, choose the closest disk. If any disk is
	 * non-rotational, choose the disk with less pending request even the
	 * disk is rotational, which might/might not be optimal for raids with
	 * mixed ratation/non-rotational disks depending on workload.
	 */
	if (best_disk == -1) {
		if (has_nonrot_disk)
			best_disk = best_pending_disk;
		else
			best_disk = best_dist_disk;
	}

N
NeilBrown 已提交
667 668
	if (best_disk >= 0) {
		rdev = rcu_dereference(conf->mirrors[best_disk].rdev);
669 670 671
		if (!rdev)
			goto retry;
		atomic_inc(&rdev->nr_pending);
N
NeilBrown 已提交
672
		if (test_bit(Faulty, &rdev->flags)) {
L
Linus Torvalds 已提交
673 674 675
			/* cannot risk returning a device that failed
			 * before we inc'ed nr_pending
			 */
676
			rdev_dec_pending(rdev, conf->mddev);
L
Linus Torvalds 已提交
677 678
			goto retry;
		}
679
		sectors = best_good_sectors;
680 681 682 683

		if (conf->mirrors[best_disk].next_seq_sect != this_sector)
			conf->mirrors[best_disk].seq_start = this_sector;

684
		conf->mirrors[best_disk].next_seq_sect = this_sector + sectors;
L
Linus Torvalds 已提交
685 686
	}
	rcu_read_unlock();
687
	*max_sectors = sectors;
L
Linus Torvalds 已提交
688

N
NeilBrown 已提交
689
	return best_disk;
L
Linus Torvalds 已提交
690 691
}

692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724
static int raid1_mergeable_bvec(struct request_queue *q,
				struct bvec_merge_data *bvm,
				struct bio_vec *biovec)
{
	struct mddev *mddev = q->queuedata;
	struct r1conf *conf = mddev->private;
	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
	int max = biovec->bv_len;

	if (mddev->merge_check_needed) {
		int disk;
		rcu_read_lock();
		for (disk = 0; disk < conf->raid_disks * 2; disk++) {
			struct md_rdev *rdev = rcu_dereference(
				conf->mirrors[disk].rdev);
			if (rdev && !test_bit(Faulty, &rdev->flags)) {
				struct request_queue *q =
					bdev_get_queue(rdev->bdev);
				if (q->merge_bvec_fn) {
					bvm->bi_sector = sector +
						rdev->data_offset;
					bvm->bi_bdev = rdev->bdev;
					max = min(max, q->merge_bvec_fn(
							  q, bvm, biovec));
				}
			}
		}
		rcu_read_unlock();
	}
	return max;

}

725
int md_raid1_congested(struct mddev *mddev, int bits)
726
{
727
	struct r1conf *conf = mddev->private;
728 729
	int i, ret = 0;

730 731 732 733
	if ((bits & (1 << BDI_async_congested)) &&
	    conf->pending_count >= max_queued_requests)
		return 1;

734
	rcu_read_lock();
735
	for (i = 0; i < conf->raid_disks * 2; i++) {
736
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
737
		if (rdev && !test_bit(Faulty, &rdev->flags)) {
738
			struct request_queue *q = bdev_get_queue(rdev->bdev);
739

740 741
			BUG_ON(!q);

742 743 744
			/* Note the '|| 1' - when read_balance prefers
			 * non-congested targets, it can be removed
			 */
745
			if ((bits & (1<<BDI_async_congested)) || 1)
746 747 748 749 750 751 752 753
				ret |= bdi_congested(&q->backing_dev_info, bits);
			else
				ret &= bdi_congested(&q->backing_dev_info, bits);
		}
	}
	rcu_read_unlock();
	return ret;
}
754
EXPORT_SYMBOL_GPL(md_raid1_congested);
755

756 757
static int raid1_congested(void *data, int bits)
{
758
	struct mddev *mddev = data;
759 760 761 762

	return mddev_congested(mddev, bits) ||
		md_raid1_congested(mddev, bits);
}
763

764
static void flush_pending_writes(struct r1conf *conf)
765 766 767 768 769 770 771 772 773
{
	/* 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);
774
		conf->pending_count = 0;
775 776 777 778
		spin_unlock_irq(&conf->device_lock);
		/* flush any pending bitmap writes to
		 * disk before proceeding w/ I/O */
		bitmap_unplug(conf->mddev->bitmap);
779
		wake_up(&conf->wait_barrier);
780 781 782 783

		while (bio) { /* submit pending writes */
			struct bio *next = bio->bi_next;
			bio->bi_next = NULL;
S
Shaohua Li 已提交
784 785 786 787 788 789
			if (unlikely((bio->bi_rw & REQ_DISCARD) &&
			    !blk_queue_discard(bdev_get_queue(bio->bi_bdev))))
				/* Just ignore it */
				bio_endio(bio, 0);
			else
				generic_make_request(bio);
790 791 792 793
			bio = next;
		}
	} else
		spin_unlock_irq(&conf->device_lock);
J
Jens Axboe 已提交
794 795
}

796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815
/* 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
816 817 818
 */
#define RESYNC_DEPTH 32

819
static void raise_barrier(struct r1conf *conf)
L
Linus Torvalds 已提交
820 821
{
	spin_lock_irq(&conf->resync_lock);
822 823 824

	/* Wait until no block IO is waiting */
	wait_event_lock_irq(conf->wait_barrier, !conf->nr_waiting,
825
			    conf->resync_lock);
826 827 828 829

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

N
NeilBrown 已提交
830
	/* Now wait for all pending IO to complete */
831 832
	wait_event_lock_irq(conf->wait_barrier,
			    !conf->nr_pending && conf->barrier < RESYNC_DEPTH,
833
			    conf->resync_lock);
834 835 836 837

	spin_unlock_irq(&conf->resync_lock);
}

838
static void lower_barrier(struct r1conf *conf)
839 840
{
	unsigned long flags;
841
	BUG_ON(conf->barrier <= 0);
842 843 844 845 846 847
	spin_lock_irqsave(&conf->resync_lock, flags);
	conf->barrier--;
	spin_unlock_irqrestore(&conf->resync_lock, flags);
	wake_up(&conf->wait_barrier);
}

848
static void wait_barrier(struct r1conf *conf)
849 850 851 852
{
	spin_lock_irq(&conf->resync_lock);
	if (conf->barrier) {
		conf->nr_waiting++;
853 854 855 856 857 858 859 860 861 862 863 864 865 866
		/* Wait for the barrier to drop.
		 * However if there are already pending
		 * requests (preventing the barrier from
		 * rising completely), and the
		 * pre-process bio queue isn't empty,
		 * then don't wait, as we need to empty
		 * that queue to get the nr_pending
		 * count down.
		 */
		wait_event_lock_irq(conf->wait_barrier,
				    !conf->barrier ||
				    (conf->nr_pending &&
				     current->bio_list &&
				     !bio_list_empty(current->bio_list)),
867
				    conf->resync_lock);
868
		conf->nr_waiting--;
L
Linus Torvalds 已提交
869
	}
870
	conf->nr_pending++;
L
Linus Torvalds 已提交
871 872 873
	spin_unlock_irq(&conf->resync_lock);
}

874
static void allow_barrier(struct r1conf *conf)
875 876 877 878 879 880 881 882
{
	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);
}

883
static void freeze_array(struct r1conf *conf)
884 885 886 887
{
	/* stop syncio and normal IO and wait for everything to
	 * go quite.
	 * We increment barrier and nr_waiting, and then
888 889 890 891 892 893 894 895
	 * 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.
896 897 898 899
	 */
	spin_lock_irq(&conf->resync_lock);
	conf->barrier++;
	conf->nr_waiting++;
900 901 902 903
	wait_event_lock_irq_cmd(conf->wait_barrier,
				conf->nr_pending == conf->nr_queued+1,
				conf->resync_lock,
				flush_pending_writes(conf));
904 905
	spin_unlock_irq(&conf->resync_lock);
}
906
static void unfreeze_array(struct r1conf *conf)
907 908 909 910 911 912 913 914 915
{
	/* 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);
}

916

917 918
/* duplicate the data pages for behind I/O 
 */
919
static void alloc_behind_pages(struct bio *bio, struct r1bio *r1_bio)
920 921 922
{
	int i;
	struct bio_vec *bvec;
923
	struct bio_vec *bvecs = kzalloc(bio->bi_vcnt * sizeof(struct bio_vec),
924
					GFP_NOIO);
925
	if (unlikely(!bvecs))
926
		return;
927 928

	bio_for_each_segment(bvec, bio, i) {
929 930 931
		bvecs[i] = *bvec;
		bvecs[i].bv_page = alloc_page(GFP_NOIO);
		if (unlikely(!bvecs[i].bv_page))
932
			goto do_sync_io;
933 934 935
		memcpy(kmap(bvecs[i].bv_page) + bvec->bv_offset,
		       kmap(bvec->bv_page) + bvec->bv_offset, bvec->bv_len);
		kunmap(bvecs[i].bv_page);
936 937
		kunmap(bvec->bv_page);
	}
938
	r1_bio->behind_bvecs = bvecs;
939 940 941
	r1_bio->behind_page_count = bio->bi_vcnt;
	set_bit(R1BIO_BehindIO, &r1_bio->state);
	return;
942 943

do_sync_io:
944
	for (i = 0; i < bio->bi_vcnt; i++)
945 946 947
		if (bvecs[i].bv_page)
			put_page(bvecs[i].bv_page);
	kfree(bvecs);
948
	pr_debug("%dB behind alloc failed, doing sync I/O\n", bio->bi_size);
949 950
}

951 952 953 954 955 956 957 958 959 960 961 962 963 964
struct raid1_plug_cb {
	struct blk_plug_cb	cb;
	struct bio_list		pending;
	int			pending_cnt;
};

static void raid1_unplug(struct blk_plug_cb *cb, bool from_schedule)
{
	struct raid1_plug_cb *plug = container_of(cb, struct raid1_plug_cb,
						  cb);
	struct mddev *mddev = plug->cb.data;
	struct r1conf *conf = mddev->private;
	struct bio *bio;

965
	if (from_schedule || current->bio_list) {
966 967 968 969
		spin_lock_irq(&conf->device_lock);
		bio_list_merge(&conf->pending_bio_list, &plug->pending);
		conf->pending_count += plug->pending_cnt;
		spin_unlock_irq(&conf->device_lock);
970
		wake_up(&conf->wait_barrier);
971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989
		md_wakeup_thread(mddev->thread);
		kfree(plug);
		return;
	}

	/* we aren't scheduling, so we can do the write-out directly. */
	bio = bio_list_get(&plug->pending);
	bitmap_unplug(mddev->bitmap);
	wake_up(&conf->wait_barrier);

	while (bio) { /* submit pending writes */
		struct bio *next = bio->bi_next;
		bio->bi_next = NULL;
		generic_make_request(bio);
		bio = next;
	}
	kfree(plug);
}

990
static void make_request(struct mddev *mddev, struct bio * bio)
L
Linus Torvalds 已提交
991
{
992
	struct r1conf *conf = mddev->private;
993
	struct raid1_info *mirror;
994
	struct r1bio *r1_bio;
L
Linus Torvalds 已提交
995
	struct bio *read_bio;
996
	int i, disks;
997
	struct bitmap *bitmap;
998
	unsigned long flags;
999
	const int rw = bio_data_dir(bio);
1000
	const unsigned long do_sync = (bio->bi_rw & REQ_SYNC);
T
Tejun Heo 已提交
1001
	const unsigned long do_flush_fua = (bio->bi_rw & (REQ_FLUSH | REQ_FUA));
S
Shaohua Li 已提交
1002 1003
	const unsigned long do_discard = (bio->bi_rw
					  & (REQ_DISCARD | REQ_SECURE));
1004
	const unsigned long do_same = (bio->bi_rw & REQ_WRITE_SAME);
1005
	struct md_rdev *blocked_rdev;
1006 1007
	struct blk_plug_cb *cb;
	struct raid1_plug_cb *plug = NULL;
1008 1009 1010
	int first_clone;
	int sectors_handled;
	int max_sectors;
1011

L
Linus Torvalds 已提交
1012 1013 1014 1015 1016
	/*
	 * 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.
	 */
1017

1018 1019
	md_write_start(mddev, bio); /* wait on superblock update early */

1020
	if (bio_data_dir(bio) == WRITE &&
K
Kent Overstreet 已提交
1021
	    bio_end_sector(bio) > mddev->suspend_lo &&
1022 1023 1024 1025 1026 1027 1028 1029 1030 1031
	    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);
K
Kent Overstreet 已提交
1032
			if (bio_end_sector(bio) <= mddev->suspend_lo ||
1033 1034 1035 1036 1037 1038
			    bio->bi_sector >= mddev->suspend_hi)
				break;
			schedule();
		}
		finish_wait(&conf->wait_barrier, &w);
	}
1039

1040
	wait_barrier(conf);
L
Linus Torvalds 已提交
1041

1042 1043
	bitmap = mddev->bitmap;

L
Linus Torvalds 已提交
1044 1045 1046 1047 1048 1049 1050 1051
	/*
	 * 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;
1052
	r1_bio->sectors = bio_sectors(bio);
1053
	r1_bio->state = 0;
L
Linus Torvalds 已提交
1054 1055 1056
	r1_bio->mddev = mddev;
	r1_bio->sector = bio->bi_sector;

1057 1058 1059 1060 1061 1062 1063 1064 1065 1066
	/* 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);

1067
	if (rw == READ) {
L
Linus Torvalds 已提交
1068 1069 1070
		/*
		 * read balancing logic:
		 */
1071 1072 1073 1074
		int rdisk;

read_again:
		rdisk = read_balance(conf, r1_bio, &max_sectors);
L
Linus Torvalds 已提交
1075 1076 1077 1078

		if (rdisk < 0) {
			/* couldn't find anywhere to read from */
			raid_end_bio_io(r1_bio);
1079
			return;
L
Linus Torvalds 已提交
1080 1081 1082
		}
		mirror = conf->mirrors + rdisk;

1083 1084 1085 1086 1087 1088 1089 1090 1091
		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
1092 1093
		r1_bio->read_disk = rdisk;

1094
		read_bio = bio_clone_mddev(bio, GFP_NOIO, mddev);
1095 1096
		md_trim_bio(read_bio, r1_bio->sector - bio->bi_sector,
			    max_sectors);
L
Linus Torvalds 已提交
1097 1098 1099 1100 1101 1102

		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;
1103
		read_bio->bi_rw = READ | do_sync;
L
Linus Torvalds 已提交
1104 1105
		read_bio->bi_private = r1_bio;

1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129
		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;
1130
			r1_bio->sectors = bio_sectors(bio) - sectors_handled;
1131 1132 1133 1134 1135 1136
			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);
1137
		return;
L
Linus Torvalds 已提交
1138 1139 1140 1141 1142
	}

	/*
	 * WRITE:
	 */
1143 1144 1145 1146 1147
	if (conf->pending_count >= max_queued_requests) {
		md_wakeup_thread(mddev->thread);
		wait_event(conf->wait_barrier,
			   conf->pending_count < max_queued_requests);
	}
1148
	/* first select target devices under rcu_lock and
L
Linus Torvalds 已提交
1149 1150
	 * inc refcount on their rdev.  Record them by setting
	 * bios[x] to bio
1151 1152 1153 1154 1155 1156
	 * 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
1157
	 */
N
NeilBrown 已提交
1158

1159
	disks = conf->raid_disks * 2;
1160 1161
 retry_write:
	blocked_rdev = NULL;
L
Linus Torvalds 已提交
1162
	rcu_read_lock();
1163
	max_sectors = r1_bio->sectors;
L
Linus Torvalds 已提交
1164
	for (i = 0;  i < disks; i++) {
1165
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
1166 1167 1168 1169 1170
		if (rdev && unlikely(test_bit(Blocked, &rdev->flags))) {
			atomic_inc(&rdev->nr_pending);
			blocked_rdev = rdev;
			break;
		}
1171
		r1_bio->bios[i] = NULL;
1172 1173
		if (!rdev || test_bit(Faulty, &rdev->flags)
		    || test_bit(Unmerged, &rdev->flags)) {
1174 1175
			if (i < conf->raid_disks)
				set_bit(R1BIO_Degraded, &r1_bio->state);
1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202
			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;
1203
				rdev_dec_pending(rdev, mddev);
1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214
				/* 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;
1215
			}
1216 1217 1218 1219 1220 1221 1222
			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
1223 1224 1225
	}
	rcu_read_unlock();

1226 1227 1228 1229 1230 1231 1232
	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);
1233
		r1_bio->state = 0;
1234 1235 1236 1237 1238 1239
		allow_barrier(conf);
		md_wait_for_blocked_rdev(blocked_rdev, mddev);
		wait_barrier(conf);
		goto retry_write;
	}

1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250
	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);
1251
	}
1252
	sectors_handled = r1_bio->sector + max_sectors - bio->bi_sector;
1253

1254
	atomic_set(&r1_bio->remaining, 1);
1255
	atomic_set(&r1_bio->behind_remaining, 0);
1256

1257
	first_clone = 1;
L
Linus Torvalds 已提交
1258 1259 1260 1261 1262
	for (i = 0; i < disks; i++) {
		struct bio *mbio;
		if (!r1_bio->bios[i])
			continue;

1263
		mbio = bio_clone_mddev(bio, GFP_NOIO, mddev);
1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282
		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;
		}
1283
		if (r1_bio->behind_bvecs) {
1284 1285 1286 1287 1288 1289 1290
			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
1291
			 * know the original bi_idx, so we just free
1292 1293 1294
			 * them all
			 */
			__bio_for_each_segment(bvec, mbio, j, 0)
1295
				bvec->bv_page = r1_bio->behind_bvecs[j].bv_page;
1296 1297 1298 1299
			if (test_bit(WriteMostly, &conf->mirrors[i].rdev->flags))
				atomic_inc(&r1_bio->behind_remaining);
		}

1300 1301 1302 1303 1304 1305
		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;
1306 1307
		mbio->bi_rw =
			WRITE | do_flush_fua | do_sync | do_discard | do_same;
1308 1309
		mbio->bi_private = r1_bio;

L
Linus Torvalds 已提交
1310
		atomic_inc(&r1_bio->remaining);
1311 1312 1313 1314 1315 1316

		cb = blk_check_plugged(raid1_unplug, mddev, sizeof(*plug));
		if (cb)
			plug = container_of(cb, struct raid1_plug_cb, cb);
		else
			plug = NULL;
1317
		spin_lock_irqsave(&conf->device_lock, flags);
1318 1319 1320 1321 1322 1323 1324
		if (plug) {
			bio_list_add(&plug->pending, mbio);
			plug->pending_cnt++;
		} else {
			bio_list_add(&conf->pending_bio_list, mbio);
			conf->pending_count++;
		}
1325
		spin_unlock_irqrestore(&conf->device_lock, flags);
1326
		if (!plug)
N
NeilBrown 已提交
1327
			md_wakeup_thread(mddev->thread);
L
Linus Torvalds 已提交
1328
	}
1329 1330 1331
	/* Mustn't call r1_bio_write_done before this next test,
	 * as it could result in the bio being freed.
	 */
1332
	if (sectors_handled < bio_sectors(bio)) {
1333
		r1_bio_write_done(r1_bio);
1334 1335 1336 1337 1338
		/* 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;
1339
		r1_bio->sectors = bio_sectors(bio) - sectors_handled;
1340 1341 1342 1343 1344 1345
		r1_bio->state = 0;
		r1_bio->mddev = mddev;
		r1_bio->sector = bio->bi_sector + sectors_handled;
		goto retry_write;
	}

1346 1347 1348 1349
	r1_bio_write_done(r1_bio);

	/* In case raid1d snuck in to freeze_array */
	wake_up(&conf->wait_barrier);
L
Linus Torvalds 已提交
1350 1351
}

1352
static void status(struct seq_file *seq, struct mddev *mddev)
L
Linus Torvalds 已提交
1353
{
1354
	struct r1conf *conf = mddev->private;
L
Linus Torvalds 已提交
1355 1356 1357
	int i;

	seq_printf(seq, " [%d/%d] [", conf->raid_disks,
1358
		   conf->raid_disks - mddev->degraded);
1359 1360
	rcu_read_lock();
	for (i = 0; i < conf->raid_disks; i++) {
1361
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
L
Linus Torvalds 已提交
1362
		seq_printf(seq, "%s",
1363 1364 1365
			   rdev && test_bit(In_sync, &rdev->flags) ? "U" : "_");
	}
	rcu_read_unlock();
L
Linus Torvalds 已提交
1366 1367 1368 1369
	seq_printf(seq, "]");
}


1370
static void error(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
1371 1372
{
	char b[BDEVNAME_SIZE];
1373
	struct r1conf *conf = mddev->private;
L
Linus Torvalds 已提交
1374 1375 1376 1377 1378 1379 1380

	/*
	 * 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
	 */
1381
	if (test_bit(In_sync, &rdev->flags)
1382
	    && (conf->raid_disks - mddev->degraded) == 1) {
L
Linus Torvalds 已提交
1383 1384
		/*
		 * Don't fail the drive, act as though we were just a
1385 1386 1387
		 * 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
1388
		 */
1389
		conf->recovery_disabled = mddev->recovery_disabled;
L
Linus Torvalds 已提交
1390
		return;
1391
	}
1392
	set_bit(Blocked, &rdev->flags);
1393 1394 1395
	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
1396
		mddev->degraded++;
1397
		set_bit(Faulty, &rdev->flags);
1398
		spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1399 1400 1401
		/*
		 * if recovery is running, make sure it aborts.
		 */
1402
		set_bit(MD_RECOVERY_INTR, &mddev->recovery);
1403 1404
	} else
		set_bit(Faulty, &rdev->flags);
1405
	set_bit(MD_CHANGE_DEVS, &mddev->flags);
1406 1407 1408
	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
1409 1410
	       mdname(mddev), bdevname(rdev->bdev, b),
	       mdname(mddev), conf->raid_disks - mddev->degraded);
L
Linus Torvalds 已提交
1411 1412
}

1413
static void print_conf(struct r1conf *conf)
L
Linus Torvalds 已提交
1414 1415 1416
{
	int i;

N
NeilBrown 已提交
1417
	printk(KERN_DEBUG "RAID1 conf printout:\n");
L
Linus Torvalds 已提交
1418
	if (!conf) {
N
NeilBrown 已提交
1419
		printk(KERN_DEBUG "(!conf)\n");
L
Linus Torvalds 已提交
1420 1421
		return;
	}
N
NeilBrown 已提交
1422
	printk(KERN_DEBUG " --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
L
Linus Torvalds 已提交
1423 1424
		conf->raid_disks);

1425
	rcu_read_lock();
L
Linus Torvalds 已提交
1426 1427
	for (i = 0; i < conf->raid_disks; i++) {
		char b[BDEVNAME_SIZE];
1428
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
1429
		if (rdev)
N
NeilBrown 已提交
1430
			printk(KERN_DEBUG " disk %d, wo:%d, o:%d, dev:%s\n",
1431 1432 1433
			       i, !test_bit(In_sync, &rdev->flags),
			       !test_bit(Faulty, &rdev->flags),
			       bdevname(rdev->bdev,b));
L
Linus Torvalds 已提交
1434
	}
1435
	rcu_read_unlock();
L
Linus Torvalds 已提交
1436 1437
}

1438
static void close_sync(struct r1conf *conf)
L
Linus Torvalds 已提交
1439
{
1440 1441
	wait_barrier(conf);
	allow_barrier(conf);
L
Linus Torvalds 已提交
1442 1443 1444 1445 1446

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

1447
static int raid1_spare_active(struct mddev *mddev)
L
Linus Torvalds 已提交
1448 1449
{
	int i;
1450
	struct r1conf *conf = mddev->private;
1451 1452
	int count = 0;
	unsigned long flags;
L
Linus Torvalds 已提交
1453 1454 1455

	/*
	 * Find all failed disks within the RAID1 configuration 
1456 1457
	 * and mark them readable.
	 * Called under mddev lock, so rcu protection not needed.
L
Linus Torvalds 已提交
1458 1459
	 */
	for (i = 0; i < conf->raid_disks; i++) {
1460
		struct md_rdev *rdev = conf->mirrors[i].rdev;
1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479
		struct md_rdev *repl = conf->mirrors[conf->raid_disks + i].rdev;
		if (repl
		    && repl->recovery_offset == MaxSector
		    && !test_bit(Faulty, &repl->flags)
		    && !test_and_set_bit(In_sync, &repl->flags)) {
			/* replacement has just become active */
			if (!rdev ||
			    !test_and_clear_bit(In_sync, &rdev->flags))
				count++;
			if (rdev) {
				/* Replaced device not technically
				 * faulty, but we need to be sure
				 * it gets removed and never re-added
				 */
				set_bit(Faulty, &rdev->flags);
				sysfs_notify_dirent_safe(
					rdev->sysfs_state);
			}
		}
1480 1481
		if (rdev
		    && !test_bit(Faulty, &rdev->flags)
1482
		    && !test_and_set_bit(In_sync, &rdev->flags)) {
1483
			count++;
1484
			sysfs_notify_dirent_safe(rdev->sysfs_state);
L
Linus Torvalds 已提交
1485 1486
		}
	}
1487 1488 1489
	spin_lock_irqsave(&conf->device_lock, flags);
	mddev->degraded -= count;
	spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1490 1491

	print_conf(conf);
1492
	return count;
L
Linus Torvalds 已提交
1493 1494 1495
}


1496
static int raid1_add_disk(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
1497
{
1498
	struct r1conf *conf = mddev->private;
1499
	int err = -EEXIST;
1500
	int mirror = 0;
1501
	struct raid1_info *p;
1502
	int first = 0;
1503
	int last = conf->raid_disks - 1;
1504
	struct request_queue *q = bdev_get_queue(rdev->bdev);
L
Linus Torvalds 已提交
1505

1506 1507 1508
	if (mddev->recovery_disabled == conf->recovery_disabled)
		return -EBUSY;

1509 1510 1511
	if (rdev->raid_disk >= 0)
		first = last = rdev->raid_disk;

1512 1513 1514 1515 1516
	if (q->merge_bvec_fn) {
		set_bit(Unmerged, &rdev->flags);
		mddev->merge_check_needed = 1;
	}

1517 1518 1519
	for (mirror = first; mirror <= last; mirror++) {
		p = conf->mirrors+mirror;
		if (!p->rdev) {
L
Linus Torvalds 已提交
1520

1521 1522
			disk_stack_limits(mddev->gendisk, rdev->bdev,
					  rdev->data_offset << 9);
L
Linus Torvalds 已提交
1523 1524 1525

			p->head_position = 0;
			rdev->raid_disk = mirror;
1526
			err = 0;
1527 1528 1529 1530
			/* 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)
1531
				conf->fullsync = 1;
1532
			rcu_assign_pointer(p->rdev, rdev);
L
Linus Torvalds 已提交
1533 1534
			break;
		}
1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546
		if (test_bit(WantReplacement, &p->rdev->flags) &&
		    p[conf->raid_disks].rdev == NULL) {
			/* Add this device as a replacement */
			clear_bit(In_sync, &rdev->flags);
			set_bit(Replacement, &rdev->flags);
			rdev->raid_disk = mirror;
			err = 0;
			conf->fullsync = 1;
			rcu_assign_pointer(p[conf->raid_disks].rdev, rdev);
			break;
		}
	}
1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559
	if (err == 0 && test_bit(Unmerged, &rdev->flags)) {
		/* Some requests might not have seen this new
		 * merge_bvec_fn.  We must wait for them to complete
		 * before merging the device fully.
		 * First we make sure any code which has tested
		 * our function has submitted the request, then
		 * we wait for all outstanding requests to complete.
		 */
		synchronize_sched();
		raise_barrier(conf);
		lower_barrier(conf);
		clear_bit(Unmerged, &rdev->flags);
	}
1560
	md_integrity_add_rdev(rdev, mddev);
S
Shaohua Li 已提交
1561 1562
	if (blk_queue_discard(bdev_get_queue(rdev->bdev)))
		queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, mddev->queue);
L
Linus Torvalds 已提交
1563
	print_conf(conf);
1564
	return err;
L
Linus Torvalds 已提交
1565 1566
}

1567
static int raid1_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
1568
{
1569
	struct r1conf *conf = mddev->private;
L
Linus Torvalds 已提交
1570
	int err = 0;
1571
	int number = rdev->raid_disk;
1572
	struct raid1_info *p = conf->mirrors + number;
L
Linus Torvalds 已提交
1573

1574 1575 1576
	if (rdev != p->rdev)
		p = conf->mirrors + conf->raid_disks + number;

L
Linus Torvalds 已提交
1577
	print_conf(conf);
1578
	if (rdev == p->rdev) {
1579
		if (test_bit(In_sync, &rdev->flags) ||
L
Linus Torvalds 已提交
1580 1581 1582 1583
		    atomic_read(&rdev->nr_pending)) {
			err = -EBUSY;
			goto abort;
		}
N
NeilBrown 已提交
1584
		/* Only remove non-faulty devices if recovery
1585 1586 1587
		 * is not possible.
		 */
		if (!test_bit(Faulty, &rdev->flags) &&
1588
		    mddev->recovery_disabled != conf->recovery_disabled &&
1589 1590 1591 1592
		    mddev->degraded < conf->raid_disks) {
			err = -EBUSY;
			goto abort;
		}
L
Linus Torvalds 已提交
1593
		p->rdev = NULL;
1594
		synchronize_rcu();
L
Linus Torvalds 已提交
1595 1596 1597 1598
		if (atomic_read(&rdev->nr_pending)) {
			/* lost the race, try later */
			err = -EBUSY;
			p->rdev = rdev;
1599
			goto abort;
1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613
		} else if (conf->mirrors[conf->raid_disks + number].rdev) {
			/* We just removed a device that is being replaced.
			 * Move down the replacement.  We drain all IO before
			 * doing this to avoid confusion.
			 */
			struct md_rdev *repl =
				conf->mirrors[conf->raid_disks + number].rdev;
			raise_barrier(conf);
			clear_bit(Replacement, &repl->flags);
			p->rdev = repl;
			conf->mirrors[conf->raid_disks + number].rdev = NULL;
			lower_barrier(conf);
			clear_bit(WantReplacement, &rdev->flags);
		} else
1614
			clear_bit(WantReplacement, &rdev->flags);
1615
		err = md_integrity_register(mddev);
L
Linus Torvalds 已提交
1616 1617 1618 1619 1620 1621 1622 1623
	}
abort:

	print_conf(conf);
	return err;
}


1624
static void end_sync_read(struct bio *bio, int error)
L
Linus Torvalds 已提交
1625
{
1626
	struct r1bio *r1_bio = bio->bi_private;
L
Linus Torvalds 已提交
1627

1628
	update_head_pos(r1_bio->read_disk, r1_bio);
1629

L
Linus Torvalds 已提交
1630 1631 1632 1633 1634
	/*
	 * 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
	 */
1635
	if (test_bit(BIO_UPTODATE, &bio->bi_flags))
L
Linus Torvalds 已提交
1636
		set_bit(R1BIO_Uptodate, &r1_bio->state);
1637 1638 1639

	if (atomic_dec_and_test(&r1_bio->remaining))
		reschedule_retry(r1_bio);
L
Linus Torvalds 已提交
1640 1641
}

1642
static void end_sync_write(struct bio *bio, int error)
L
Linus Torvalds 已提交
1643 1644
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
1645
	struct r1bio *r1_bio = bio->bi_private;
1646
	struct mddev *mddev = r1_bio->mddev;
1647
	struct r1conf *conf = mddev->private;
L
Linus Torvalds 已提交
1648
	int mirror=0;
1649 1650
	sector_t first_bad;
	int bad_sectors;
L
Linus Torvalds 已提交
1651

1652 1653
	mirror = find_bio_disk(r1_bio, bio);

1654
	if (!uptodate) {
N
NeilBrown 已提交
1655
		sector_t sync_blocks = 0;
1656 1657 1658 1659
		sector_t s = r1_bio->sector;
		long sectors_to_go = r1_bio->sectors;
		/* make sure these bits doesn't get cleared. */
		do {
1660
			bitmap_end_sync(mddev->bitmap, s,
1661 1662 1663 1664
					&sync_blocks, 1);
			s += sync_blocks;
			sectors_to_go -= sync_blocks;
		} while (sectors_to_go > 0);
1665 1666
		set_bit(WriteErrorSeen,
			&conf->mirrors[mirror].rdev->flags);
1667 1668 1669 1670
		if (!test_and_set_bit(WantReplacement,
				      &conf->mirrors[mirror].rdev->flags))
			set_bit(MD_RECOVERY_NEEDED, &
				mddev->recovery);
1671
		set_bit(R1BIO_WriteError, &r1_bio->state);
1672 1673 1674
	} else if (is_badblock(conf->mirrors[mirror].rdev,
			       r1_bio->sector,
			       r1_bio->sectors,
1675 1676 1677 1678 1679 1680
			       &first_bad, &bad_sectors) &&
		   !is_badblock(conf->mirrors[r1_bio->read_disk].rdev,
				r1_bio->sector,
				r1_bio->sectors,
				&first_bad, &bad_sectors)
		)
1681
		set_bit(R1BIO_MadeGood, &r1_bio->state);
1682

L
Linus Torvalds 已提交
1683
	if (atomic_dec_and_test(&r1_bio->remaining)) {
1684
		int s = r1_bio->sectors;
1685 1686
		if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
		    test_bit(R1BIO_WriteError, &r1_bio->state))
1687 1688 1689 1690 1691
			reschedule_retry(r1_bio);
		else {
			put_buf(r1_bio);
			md_done_sync(mddev, s, uptodate);
		}
L
Linus Torvalds 已提交
1692 1693 1694
	}
}

1695
static int r1_sync_page_io(struct md_rdev *rdev, sector_t sector,
1696 1697 1698 1699 1700
			    int sectors, struct page *page, int rw)
{
	if (sync_page_io(rdev, sector, sectors << 9, page, rw, false))
		/* success */
		return 1;
1701
	if (rw == WRITE) {
1702
		set_bit(WriteErrorSeen, &rdev->flags);
1703 1704 1705 1706 1707
		if (!test_and_set_bit(WantReplacement,
				      &rdev->flags))
			set_bit(MD_RECOVERY_NEEDED, &
				rdev->mddev->recovery);
	}
1708 1709 1710 1711 1712 1713
	/* 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;
}

1714
static int fix_sync_read_error(struct r1bio *r1_bio)
L
Linus Torvalds 已提交
1715
{
1716 1717 1718 1719 1720 1721 1722
	/* 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.
1723 1724 1725
	 * 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.
1726
	 */
1727
	struct mddev *mddev = r1_bio->mddev;
1728
	struct r1conf *conf = mddev->private;
1729 1730 1731 1732 1733 1734 1735 1736 1737
	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;
1738
		struct md_rdev *rdev;
1739
		int start;
1740 1741 1742 1743 1744 1745 1746 1747 1748 1749

		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;
1750
				if (sync_page_io(rdev, sect, s<<9,
1751 1752 1753 1754 1755 1756 1757
						 bio->bi_io_vec[idx].bv_page,
						 READ, false)) {
					success = 1;
					break;
				}
			}
			d++;
1758
			if (d == conf->raid_disks * 2)
1759 1760 1761
				d = 0;
		} while (!success && d != r1_bio->read_disk);

1762
		if (!success) {
1763
			char b[BDEVNAME_SIZE];
1764 1765 1766 1767 1768 1769
			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.
			 */
1770 1771 1772 1773 1774
			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);
1775
			for (d = 0; d < conf->raid_disks * 2; d++) {
1776 1777 1778 1779 1780 1781 1782
				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) {
1783 1784
				conf->recovery_disabled =
					mddev->recovery_disabled;
1785 1786 1787 1788 1789 1790 1791 1792 1793 1794
				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;
1795
		}
1796 1797 1798 1799 1800

		start = d;
		/* write it back and re-read */
		while (d != r1_bio->read_disk) {
			if (d == 0)
1801
				d = conf->raid_disks * 2;
1802 1803 1804 1805
			d--;
			if (r1_bio->bios[d]->bi_end_io != end_sync_read)
				continue;
			rdev = conf->mirrors[d].rdev;
1806 1807 1808
			if (r1_sync_page_io(rdev, sect, s,
					    bio->bi_io_vec[idx].bv_page,
					    WRITE) == 0) {
1809 1810
				r1_bio->bios[d]->bi_end_io = NULL;
				rdev_dec_pending(rdev, mddev);
1811
			}
1812 1813 1814 1815
		}
		d = start;
		while (d != r1_bio->read_disk) {
			if (d == 0)
1816
				d = conf->raid_disks * 2;
1817 1818 1819 1820
			d--;
			if (r1_bio->bios[d]->bi_end_io != end_sync_read)
				continue;
			rdev = conf->mirrors[d].rdev;
1821 1822 1823
			if (r1_sync_page_io(rdev, sect, s,
					    bio->bi_io_vec[idx].bv_page,
					    READ) != 0)
1824
				atomic_add(s, &rdev->corrected_errors);
1825
		}
1826 1827 1828 1829
		sectors -= s;
		sect += s;
		idx ++;
	}
1830
	set_bit(R1BIO_Uptodate, &r1_bio->state);
1831
	set_bit(BIO_UPTODATE, &bio->bi_flags);
1832 1833 1834
	return 1;
}

1835
static int process_checks(struct r1bio *r1_bio)
1836 1837 1838 1839 1840 1841 1842 1843
{
	/* 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
	 */
1844
	struct mddev *mddev = r1_bio->mddev;
1845
	struct r1conf *conf = mddev->private;
1846 1847
	int primary;
	int i;
1848
	int vcnt;
1849

1850
	for (primary = 0; primary < conf->raid_disks * 2; primary++)
1851 1852 1853 1854 1855 1856 1857
		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;
1858
	vcnt = (r1_bio->sectors + PAGE_SIZE / 512 - 1) >> (PAGE_SHIFT - 9);
1859
	for (i = 0; i < conf->raid_disks * 2; i++) {
1860 1861 1862 1863
		int j;
		struct bio *pbio = r1_bio->bios[primary];
		struct bio *sbio = r1_bio->bios[i];
		int size;
1864

1865 1866 1867 1868 1869 1870 1871 1872 1873 1874
		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),
1875
					   sbio->bi_io_vec[j].bv_len))
1876
					break;
1877
			}
1878 1879 1880
		} else
			j = 0;
		if (j >= 0)
1881
			atomic64_add(r1_bio->sectors, &mddev->resync_mismatches);
1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912
		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);
1913
		}
1914
	}
1915 1916 1917
	return 0;
}

1918
static void sync_request_write(struct mddev *mddev, struct r1bio *r1_bio)
1919
{
1920
	struct r1conf *conf = mddev->private;
1921
	int i;
1922
	int disks = conf->raid_disks * 2;
1923 1924 1925 1926 1927 1928 1929 1930
	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;
1931 1932 1933 1934

	if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
		if (process_checks(r1_bio) < 0)
			return;
1935 1936 1937
	/*
	 * schedule writes
	 */
L
Linus Torvalds 已提交
1938 1939 1940
	atomic_set(&r1_bio->remaining, 1);
	for (i = 0; i < disks ; i++) {
		wbio = r1_bio->bios[i];
1941 1942 1943 1944
		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
1945 1946
			continue;

1947 1948
		wbio->bi_rw = WRITE;
		wbio->bi_end_io = end_sync_write;
L
Linus Torvalds 已提交
1949
		atomic_inc(&r1_bio->remaining);
1950
		md_sync_acct(conf->mirrors[i].rdev->bdev, bio_sectors(wbio));
1951

L
Linus Torvalds 已提交
1952 1953 1954 1955
		generic_make_request(wbio);
	}

	if (atomic_dec_and_test(&r1_bio->remaining)) {
1956
		/* if we're here, all write(s) have completed, so clean up */
1957 1958 1959 1960 1961 1962 1963 1964
		int s = r1_bio->sectors;
		if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
		    test_bit(R1BIO_WriteError, &r1_bio->state))
			reschedule_retry(r1_bio);
		else {
			put_buf(r1_bio);
			md_done_sync(mddev, s, 1);
		}
L
Linus Torvalds 已提交
1965 1966 1967 1968 1969 1970 1971 1972
	}
}

/*
 * This is a kernel thread which:
 *
 *	1.	Retries failed read operations on working mirrors.
 *	2.	Updates the raid superblock when problems encounter.
1973
 *	3.	Performs writes following reads for array synchronising.
L
Linus Torvalds 已提交
1974 1975
 */

1976
static void fix_read_error(struct r1conf *conf, int read_disk,
1977 1978
			   sector_t sect, int sectors)
{
1979
	struct mddev *mddev = conf->mddev;
1980 1981 1982 1983 1984
	while(sectors) {
		int s = sectors;
		int d = read_disk;
		int success = 0;
		int start;
1985
		struct md_rdev *rdev;
1986 1987 1988 1989 1990 1991 1992 1993 1994 1995

		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....
			 */
1996 1997 1998
			sector_t first_bad;
			int bad_sectors;

1999 2000
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
2001 2002 2003
			    (test_bit(In_sync, &rdev->flags) ||
			     (!test_bit(Faulty, &rdev->flags) &&
			      rdev->recovery_offset >= sect + s)) &&
2004 2005
			    is_badblock(rdev, sect, s,
					&first_bad, &bad_sectors) == 0 &&
J
Jonathan Brassow 已提交
2006 2007
			    sync_page_io(rdev, sect, s<<9,
					 conf->tmppage, READ, false))
2008 2009 2010
				success = 1;
			else {
				d++;
2011
				if (d == conf->raid_disks * 2)
2012 2013 2014 2015 2016
					d = 0;
			}
		} while (!success && d != read_disk);

		if (!success) {
2017
			/* Cannot read from anywhere - mark it bad */
2018
			struct md_rdev *rdev = conf->mirrors[read_disk].rdev;
2019 2020
			if (!rdev_set_badblocks(rdev, sect, s, 0))
				md_error(mddev, rdev);
2021 2022 2023 2024 2025 2026
			break;
		}
		/* write it back and re-read */
		start = d;
		while (d != read_disk) {
			if (d==0)
2027
				d = conf->raid_disks * 2;
2028 2029 2030
			d--;
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
2031 2032 2033
			    test_bit(In_sync, &rdev->flags))
				r1_sync_page_io(rdev, sect, s,
						conf->tmppage, WRITE);
2034 2035 2036 2037 2038
		}
		d = start;
		while (d != read_disk) {
			char b[BDEVNAME_SIZE];
			if (d==0)
2039
				d = conf->raid_disks * 2;
2040 2041 2042 2043
			d--;
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
			    test_bit(In_sync, &rdev->flags)) {
2044 2045
				if (r1_sync_page_io(rdev, sect, s,
						    conf->tmppage, READ)) {
2046 2047
					atomic_add(s, &rdev->corrected_errors);
					printk(KERN_INFO
N
NeilBrown 已提交
2048
					       "md/raid1:%s: read error corrected "
2049 2050
					       "(%d sectors at %llu on %s)\n",
					       mdname(mddev), s,
2051 2052
					       (unsigned long long)(sect +
					           rdev->data_offset),
2053 2054 2055 2056 2057 2058 2059 2060 2061
					       bdevname(rdev->bdev, b));
				}
			}
		}
		sectors -= s;
		sect += s;
	}
}

2062
static int narrow_write_error(struct r1bio *r1_bio, int i)
2063
{
2064
	struct mddev *mddev = r1_bio->mddev;
2065
	struct r1conf *conf = mddev->private;
2066
	struct md_rdev *rdev = conf->mirrors[i].rdev;
2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137
	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;
}

2138
static void handle_sync_write_finished(struct r1conf *conf, struct r1bio *r1_bio)
2139 2140 2141
{
	int m;
	int s = r1_bio->sectors;
2142
	for (m = 0; m < conf->raid_disks * 2 ; m++) {
2143
		struct md_rdev *rdev = conf->mirrors[m].rdev;
2144 2145 2146 2147 2148
		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)) {
2149
			rdev_clear_badblocks(rdev, r1_bio->sector, s, 0);
2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160
		}
		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);
}

2161
static void handle_write_finished(struct r1conf *conf, struct r1bio *r1_bio)
2162 2163
{
	int m;
2164
	for (m = 0; m < conf->raid_disks * 2 ; m++)
2165
		if (r1_bio->bios[m] == IO_MADE_GOOD) {
2166
			struct md_rdev *rdev = conf->mirrors[m].rdev;
2167 2168
			rdev_clear_badblocks(rdev,
					     r1_bio->sector,
2169
					     r1_bio->sectors, 0);
2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189
			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);
}

2190
static void handle_read_error(struct r1conf *conf, struct r1bio *r1_bio)
2191 2192 2193
{
	int disk;
	int max_sectors;
2194
	struct mddev *mddev = conf->mddev;
2195 2196
	struct bio *bio;
	char b[BDEVNAME_SIZE];
2197
	struct md_rdev *rdev;
2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214

	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);
2215
	rdev_dec_pending(conf->mirrors[r1_bio->read_disk].rdev, conf->mddev);
2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267

	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;
2268
			r1_bio->sectors = bio_sectors(mbio) - sectors_handled;
2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279
			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);
	}
}

S
Shaohua Li 已提交
2280
static void raid1d(struct md_thread *thread)
L
Linus Torvalds 已提交
2281
{
S
Shaohua Li 已提交
2282
	struct mddev *mddev = thread->mddev;
2283
	struct r1bio *r1_bio;
L
Linus Torvalds 已提交
2284
	unsigned long flags;
2285
	struct r1conf *conf = mddev->private;
L
Linus Torvalds 已提交
2286
	struct list_head *head = &conf->retry_list;
2287
	struct blk_plug plug;
L
Linus Torvalds 已提交
2288 2289

	md_check_recovery(mddev);
2290 2291

	blk_start_plug(&plug);
L
Linus Torvalds 已提交
2292
	for (;;) {
2293

2294
		flush_pending_writes(conf);
2295

2296 2297 2298
		spin_lock_irqsave(&conf->device_lock, flags);
		if (list_empty(head)) {
			spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
2299
			break;
2300
		}
2301
		r1_bio = list_entry(head->prev, struct r1bio, retry_list);
L
Linus Torvalds 已提交
2302
		list_del(head->prev);
2303
		conf->nr_queued--;
L
Linus Torvalds 已提交
2304 2305 2306
		spin_unlock_irqrestore(&conf->device_lock, flags);

		mddev = r1_bio->mddev;
2307
		conf = mddev->private;
2308
		if (test_bit(R1BIO_IsSync, &r1_bio->state)) {
2309
			if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
2310 2311 2312
			    test_bit(R1BIO_WriteError, &r1_bio->state))
				handle_sync_write_finished(conf, r1_bio);
			else
2313
				sync_request_write(mddev, r1_bio);
2314
		} else if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
2315 2316 2317 2318 2319
			   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
2320 2321 2322 2323
			/* just a partial read to be scheduled from separate
			 * context
			 */
			generic_make_request(r1_bio->bios[r1_bio->read_disk]);
2324

N
NeilBrown 已提交
2325
		cond_resched();
2326 2327
		if (mddev->flags & ~(1<<MD_CHANGE_PENDING))
			md_check_recovery(mddev);
L
Linus Torvalds 已提交
2328
	}
2329
	blk_finish_plug(&plug);
L
Linus Torvalds 已提交
2330 2331 2332
}


2333
static int init_resync(struct r1conf *conf)
L
Linus Torvalds 已提交
2334 2335 2336 2337
{
	int buffs;

	buffs = RESYNC_WINDOW / RESYNC_BLOCK_SIZE;
2338
	BUG_ON(conf->r1buf_pool);
L
Linus Torvalds 已提交
2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356
	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.
 */

2357
static sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipped, int go_faster)
L
Linus Torvalds 已提交
2358
{
2359
	struct r1conf *conf = mddev->private;
2360
	struct r1bio *r1_bio;
L
Linus Torvalds 已提交
2361 2362
	struct bio *bio;
	sector_t max_sector, nr_sectors;
2363
	int disk = -1;
L
Linus Torvalds 已提交
2364
	int i;
2365 2366
	int wonly = -1;
	int write_targets = 0, read_targets = 0;
N
NeilBrown 已提交
2367
	sector_t sync_blocks;
2368
	int still_degraded = 0;
2369 2370
	int good_sectors = RESYNC_SECTORS;
	int min_bad = 0; /* number of sectors that are bad in all devices */
L
Linus Torvalds 已提交
2371 2372 2373

	if (!conf->r1buf_pool)
		if (init_resync(conf))
2374
			return 0;
L
Linus Torvalds 已提交
2375

A
Andre Noll 已提交
2376
	max_sector = mddev->dev_sectors;
L
Linus Torvalds 已提交
2377
	if (sector_nr >= max_sector) {
2378 2379 2380 2381 2382
		/* 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
		 */
2383 2384
		if (mddev->curr_resync < max_sector) /* aborted */
			bitmap_end_sync(mddev->bitmap, mddev->curr_resync,
2385
						&sync_blocks, 1);
2386
		else /* completed sync */
2387
			conf->fullsync = 0;
2388 2389

		bitmap_close_sync(mddev->bitmap);
L
Linus Torvalds 已提交
2390 2391 2392 2393
		close_sync(conf);
		return 0;
	}

2394 2395
	if (mddev->bitmap == NULL &&
	    mddev->recovery_cp == MaxSector &&
2396
	    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
2397 2398 2399 2400
	    conf->fullsync == 0) {
		*skipped = 1;
		return max_sector - sector_nr;
	}
2401 2402 2403
	/* before building a request, check if we can skip these blocks..
	 * This call the bitmap_start_sync doesn't actually record anything
	 */
2404
	if (!bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, 1) &&
2405
	    !conf->fullsync && !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
2406 2407 2408 2409
		/* We can skip this block, and probably several more */
		*skipped = 1;
		return sync_blocks;
	}
L
Linus Torvalds 已提交
2410
	/*
2411 2412 2413
	 * 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
2414
	 */
2415
	if (!go_faster && conf->nr_waiting)
L
Linus Torvalds 已提交
2416
		msleep_interruptible(1000);
2417

N
NeilBrown 已提交
2418
	bitmap_cond_end_sync(mddev->bitmap, sector_nr);
2419
	r1_bio = mempool_alloc(conf->r1buf_pool, GFP_NOIO);
2420 2421 2422
	raise_barrier(conf);

	conf->next_resync = sector_nr;
L
Linus Torvalds 已提交
2423

2424
	rcu_read_lock();
L
Linus Torvalds 已提交
2425
	/*
2426 2427 2428 2429 2430 2431
	 * 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
2432 2433 2434 2435
	 */

	r1_bio->mddev = mddev;
	r1_bio->sector = sector_nr;
2436
	r1_bio->state = 0;
L
Linus Torvalds 已提交
2437 2438
	set_bit(R1BIO_IsSync, &r1_bio->state);

2439
	for (i = 0; i < conf->raid_disks * 2; i++) {
2440
		struct md_rdev *rdev;
L
Linus Torvalds 已提交
2441 2442 2443 2444
		bio = r1_bio->bios[i];

		/* take from bio_init */
		bio->bi_next = NULL;
2445
		bio->bi_flags &= ~(BIO_POOL_MASK-1);
L
Linus Torvalds 已提交
2446
		bio->bi_flags |= 1 << BIO_UPTODATE;
2447
		bio->bi_rw = READ;
L
Linus Torvalds 已提交
2448 2449 2450 2451 2452 2453 2454
		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;

2455 2456
		rdev = rcu_dereference(conf->mirrors[i].rdev);
		if (rdev == NULL ||
2457
		    test_bit(Faulty, &rdev->flags)) {
2458 2459
			if (i < conf->raid_disks)
				still_degraded = 1;
2460
		} else if (!test_bit(In_sync, &rdev->flags)) {
L
Linus Torvalds 已提交
2461 2462 2463
			bio->bi_rw = WRITE;
			bio->bi_end_io = end_sync_write;
			write_targets ++;
2464 2465
		} else {
			/* may need to read from here */
2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490
			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++;
2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502
			} else if (!test_bit(WriteErrorSeen, &rdev->flags) &&
				test_bit(MD_RECOVERY_SYNC, &mddev->recovery) &&
				!test_bit(MD_RECOVERY_CHECK, &mddev->recovery)) {
				/*
				 * The device is suitable for reading (InSync),
				 * but has bad block(s) here. Let's try to correct them,
				 * if we are doing resync or repair. Otherwise, leave
				 * this device alone for this sync request.
				 */
				bio->bi_rw = WRITE;
				bio->bi_end_io = end_sync_write;
				write_targets++;
2503 2504
			}
		}
2505 2506 2507 2508 2509 2510
		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
2511
	}
2512 2513 2514 2515
	rcu_read_unlock();
	if (disk < 0)
		disk = wonly;
	r1_bio->read_disk = disk;
2516

2517 2518 2519 2520 2521
	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;
2522
		for (i = 0 ; i < conf->raid_disks * 2 ; i++)
2523
			if (r1_bio->bios[i]->bi_end_io == end_sync_write) {
2524
				struct md_rdev *rdev = conf->mirrors[i].rdev;
2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551
				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;
	}

2552 2553 2554 2555 2556
	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
2557 2558 2559
		/* There is nowhere to write, so all non-sync
		 * drives must be failed - so we are finished
		 */
2560 2561 2562 2563
		sector_t rv;
		if (min_bad > 0)
			max_sector = sector_nr + min_bad;
		rv = max_sector - sector_nr;
2564
		*skipped = 1;
L
Linus Torvalds 已提交
2565 2566 2567 2568
		put_buf(r1_bio);
		return rv;
	}

2569 2570
	if (max_sector > mddev->resync_max)
		max_sector = mddev->resync_max; /* Don't do IO beyond here */
2571 2572
	if (max_sector > sector_nr + good_sectors)
		max_sector = sector_nr + good_sectors;
L
Linus Torvalds 已提交
2573
	nr_sectors = 0;
2574
	sync_blocks = 0;
L
Linus Torvalds 已提交
2575 2576 2577 2578 2579 2580 2581
	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;
2582 2583
		if (sync_blocks == 0) {
			if (!bitmap_start_sync(mddev->bitmap, sector_nr,
2584 2585 2586
					       &sync_blocks, still_degraded) &&
			    !conf->fullsync &&
			    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
2587
				break;
2588
			BUG_ON(sync_blocks < (PAGE_SIZE>>9));
2589
			if ((len >> 9) > sync_blocks)
2590
				len = sync_blocks<<9;
2591
		}
2592

2593
		for (i = 0 ; i < conf->raid_disks * 2; i++) {
L
Linus Torvalds 已提交
2594 2595
			bio = r1_bio->bios[i];
			if (bio->bi_end_io) {
2596
				page = bio->bi_io_vec[bio->bi_vcnt].bv_page;
L
Linus Torvalds 已提交
2597 2598
				if (bio_add_page(bio, page, len, 0) == 0) {
					/* stop here */
2599
					bio->bi_io_vec[bio->bi_vcnt].bv_page = page;
L
Linus Torvalds 已提交
2600 2601 2602
					while (i > 0) {
						i--;
						bio = r1_bio->bios[i];
2603 2604
						if (bio->bi_end_io==NULL)
							continue;
L
Linus Torvalds 已提交
2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615
						/* 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;
2616
		sync_blocks -= (len>>9);
L
Linus Torvalds 已提交
2617 2618 2619 2620
	} while (r1_bio->bios[disk]->bi_vcnt < RESYNC_PAGES);
 bio_full:
	r1_bio->sectors = nr_sectors;

2621 2622 2623 2624 2625
	/* 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);
2626
		for (i = 0; i < conf->raid_disks * 2 && read_targets; i++) {
2627 2628
			bio = r1_bio->bios[i];
			if (bio->bi_end_io == end_sync_read) {
2629
				read_targets--;
2630
				md_sync_acct(bio->bi_bdev, nr_sectors);
2631 2632 2633 2634 2635 2636
				generic_make_request(bio);
			}
		}
	} else {
		atomic_set(&r1_bio->remaining, 1);
		bio = r1_bio->bios[r1_bio->read_disk];
2637
		md_sync_acct(bio->bi_bdev, nr_sectors);
2638
		generic_make_request(bio);
L
Linus Torvalds 已提交
2639

2640
	}
L
Linus Torvalds 已提交
2641 2642 2643
	return nr_sectors;
}

2644
static sector_t raid1_size(struct mddev *mddev, sector_t sectors, int raid_disks)
2645 2646 2647 2648 2649 2650 2651
{
	if (sectors)
		return sectors;

	return mddev->dev_sectors;
}

2652
static struct r1conf *setup_conf(struct mddev *mddev)
L
Linus Torvalds 已提交
2653
{
2654
	struct r1conf *conf;
2655
	int i;
2656
	struct raid1_info *disk;
2657
	struct md_rdev *rdev;
2658
	int err = -ENOMEM;
L
Linus Torvalds 已提交
2659

2660
	conf = kzalloc(sizeof(struct r1conf), GFP_KERNEL);
L
Linus Torvalds 已提交
2661
	if (!conf)
2662
		goto abort;
L
Linus Torvalds 已提交
2663

2664
	conf->mirrors = kzalloc(sizeof(struct raid1_info)
2665
				* mddev->raid_disks * 2,
L
Linus Torvalds 已提交
2666 2667
				 GFP_KERNEL);
	if (!conf->mirrors)
2668
		goto abort;
L
Linus Torvalds 已提交
2669

2670 2671
	conf->tmppage = alloc_page(GFP_KERNEL);
	if (!conf->tmppage)
2672
		goto abort;
2673

2674
	conf->poolinfo = kzalloc(sizeof(*conf->poolinfo), GFP_KERNEL);
L
Linus Torvalds 已提交
2675
	if (!conf->poolinfo)
2676
		goto abort;
2677
	conf->poolinfo->raid_disks = mddev->raid_disks * 2;
L
Linus Torvalds 已提交
2678 2679 2680 2681
	conf->r1bio_pool = mempool_create(NR_RAID1_BIOS, r1bio_pool_alloc,
					  r1bio_pool_free,
					  conf->poolinfo);
	if (!conf->r1bio_pool)
2682 2683
		goto abort;

2684
	conf->poolinfo->mddev = mddev;
L
Linus Torvalds 已提交
2685

2686
	err = -EINVAL;
2687
	spin_lock_init(&conf->device_lock);
N
NeilBrown 已提交
2688
	rdev_for_each(rdev, mddev) {
2689
		struct request_queue *q;
2690
		int disk_idx = rdev->raid_disk;
L
Linus Torvalds 已提交
2691 2692 2693
		if (disk_idx >= mddev->raid_disks
		    || disk_idx < 0)
			continue;
2694
		if (test_bit(Replacement, &rdev->flags))
2695
			disk = conf->mirrors + mddev->raid_disks + disk_idx;
2696 2697
		else
			disk = conf->mirrors + disk_idx;
L
Linus Torvalds 已提交
2698

2699 2700
		if (disk->rdev)
			goto abort;
L
Linus Torvalds 已提交
2701
		disk->rdev = rdev;
2702 2703 2704
		q = bdev_get_queue(rdev->bdev);
		if (q->merge_bvec_fn)
			mddev->merge_check_needed = 1;
L
Linus Torvalds 已提交
2705 2706

		disk->head_position = 0;
2707
		disk->seq_start = MaxSector;
L
Linus Torvalds 已提交
2708 2709 2710 2711 2712 2713
	}
	conf->raid_disks = mddev->raid_disks;
	conf->mddev = mddev;
	INIT_LIST_HEAD(&conf->retry_list);

	spin_lock_init(&conf->resync_lock);
2714
	init_waitqueue_head(&conf->wait_barrier);
L
Linus Torvalds 已提交
2715

2716
	bio_list_init(&conf->pending_bio_list);
2717
	conf->pending_count = 0;
2718
	conf->recovery_disabled = mddev->recovery_disabled - 1;
2719

2720
	err = -EIO;
2721
	for (i = 0; i < conf->raid_disks * 2; i++) {
L
Linus Torvalds 已提交
2722 2723 2724

		disk = conf->mirrors + i;

2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739
		if (i < conf->raid_disks &&
		    disk[conf->raid_disks].rdev) {
			/* This slot has a replacement. */
			if (!disk->rdev) {
				/* No original, just make the replacement
				 * a recovering spare
				 */
				disk->rdev =
					disk[conf->raid_disks].rdev;
				disk[conf->raid_disks].rdev = NULL;
			} else if (!test_bit(In_sync, &disk->rdev->flags))
				/* Original is not in_sync - bad */
				goto abort;
		}

2740 2741
		if (!disk->rdev ||
		    !test_bit(In_sync, &disk->rdev->flags)) {
L
Linus Torvalds 已提交
2742
			disk->head_position = 0;
2743 2744
			if (disk->rdev &&
			    (disk->rdev->saved_raid_disk < 0))
2745
				conf->fullsync = 1;
2746
		}
L
Linus Torvalds 已提交
2747
	}
2748 2749

	err = -ENOMEM;
2750
	conf->thread = md_register_thread(raid1d, mddev, "raid1");
2751 2752
	if (!conf->thread) {
		printk(KERN_ERR
N
NeilBrown 已提交
2753
		       "md/raid1:%s: couldn't allocate thread\n",
2754 2755
		       mdname(mddev));
		goto abort;
2756
	}
L
Linus Torvalds 已提交
2757

2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771
	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);
}

2772
static int stop(struct mddev *mddev);
2773
static int run(struct mddev *mddev)
2774
{
2775
	struct r1conf *conf;
2776
	int i;
2777
	struct md_rdev *rdev;
2778
	int ret;
S
Shaohua Li 已提交
2779
	bool discard_supported = false;
2780 2781

	if (mddev->level != 1) {
N
NeilBrown 已提交
2782
		printk(KERN_ERR "md/raid1:%s: raid level not set to mirroring (%d)\n",
2783 2784 2785 2786
		       mdname(mddev), mddev->level);
		return -EIO;
	}
	if (mddev->reshape_position != MaxSector) {
N
NeilBrown 已提交
2787
		printk(KERN_ERR "md/raid1:%s: reshape_position set but not supported\n",
2788 2789 2790
		       mdname(mddev));
		return -EIO;
	}
L
Linus Torvalds 已提交
2791
	/*
2792 2793 2794
	 * 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
2795
	 */
2796 2797 2798 2799
	if (mddev->private == NULL)
		conf = setup_conf(mddev);
	else
		conf = mddev->private;
L
Linus Torvalds 已提交
2800

2801 2802
	if (IS_ERR(conf))
		return PTR_ERR(conf);
L
Linus Torvalds 已提交
2803

2804 2805 2806
	if (mddev->queue)
		blk_queue_max_write_same_sectors(mddev->queue,
						 mddev->chunk_sectors);
N
NeilBrown 已提交
2807
	rdev_for_each(rdev, mddev) {
2808 2809
		if (!mddev->gendisk)
			continue;
2810 2811
		disk_stack_limits(mddev->gendisk, rdev->bdev,
				  rdev->data_offset << 9);
S
Shaohua Li 已提交
2812 2813
		if (blk_queue_discard(bdev_get_queue(rdev->bdev)))
			discard_supported = true;
L
Linus Torvalds 已提交
2814
	}
2815

2816 2817 2818 2819 2820 2821 2822 2823 2824 2825
	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;

2826
	if (mddev->recovery_cp != MaxSector)
N
NeilBrown 已提交
2827
		printk(KERN_NOTICE "md/raid1:%s: not clean"
2828 2829
		       " -- starting background reconstruction\n",
		       mdname(mddev));
L
Linus Torvalds 已提交
2830
	printk(KERN_INFO 
N
NeilBrown 已提交
2831
		"md/raid1:%s: active with %d out of %d mirrors\n",
L
Linus Torvalds 已提交
2832 2833
		mdname(mddev), mddev->raid_disks - mddev->degraded, 
		mddev->raid_disks);
2834

L
Linus Torvalds 已提交
2835 2836 2837
	/*
	 * Ok, everything is just fine now
	 */
2838 2839 2840 2841
	mddev->thread = conf->thread;
	conf->thread = NULL;
	mddev->private = conf;

2842
	md_set_array_sectors(mddev, raid1_size(mddev, 0, 0));
L
Linus Torvalds 已提交
2843

2844 2845 2846
	if (mddev->queue) {
		mddev->queue->backing_dev_info.congested_fn = raid1_congested;
		mddev->queue->backing_dev_info.congested_data = mddev;
2847
		blk_queue_merge_bvec(mddev->queue, raid1_mergeable_bvec);
S
Shaohua Li 已提交
2848 2849 2850 2851 2852 2853 2854

		if (discard_supported)
			queue_flag_set_unlocked(QUEUE_FLAG_DISCARD,
						mddev->queue);
		else
			queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD,
						  mddev->queue);
2855
	}
2856 2857 2858 2859 2860

	ret =  md_integrity_register(mddev);
	if (ret)
		stop(mddev);
	return ret;
L
Linus Torvalds 已提交
2861 2862
}

2863
static int stop(struct mddev *mddev)
L
Linus Torvalds 已提交
2864
{
2865
	struct r1conf *conf = mddev->private;
2866 2867 2868
	struct bitmap *bitmap = mddev->bitmap;

	/* wait for behind writes to complete */
2869
	if (bitmap && atomic_read(&bitmap->behind_writes) > 0) {
N
NeilBrown 已提交
2870 2871
		printk(KERN_INFO "md/raid1:%s: behind writes in progress - waiting to stop.\n",
		       mdname(mddev));
2872
		/* need to kick something here to make sure I/O goes? */
2873 2874
		wait_event(bitmap->behind_wait,
			   atomic_read(&bitmap->behind_writes) == 0);
2875
	}
L
Linus Torvalds 已提交
2876

2877 2878 2879
	raise_barrier(conf);
	lower_barrier(conf);

2880
	md_unregister_thread(&mddev->thread);
L
Linus Torvalds 已提交
2881 2882
	if (conf->r1bio_pool)
		mempool_destroy(conf->r1bio_pool);
2883 2884
	kfree(conf->mirrors);
	kfree(conf->poolinfo);
L
Linus Torvalds 已提交
2885 2886 2887 2888 2889
	kfree(conf);
	mddev->private = NULL;
	return 0;
}

2890
static int raid1_resize(struct mddev *mddev, sector_t sectors)
L
Linus Torvalds 已提交
2891 2892 2893 2894 2895 2896 2897 2898
{
	/* 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.
	 */
2899 2900 2901
	sector_t newsize = raid1_size(mddev, sectors, 0);
	if (mddev->external_size &&
	    mddev->array_sectors > newsize)
D
Dan Williams 已提交
2902
		return -EINVAL;
2903 2904 2905 2906 2907 2908
	if (mddev->bitmap) {
		int ret = bitmap_resize(mddev->bitmap, newsize, 0, 0);
		if (ret)
			return ret;
	}
	md_set_array_sectors(mddev, newsize);
2909
	set_capacity(mddev->gendisk, mddev->array_sectors);
2910
	revalidate_disk(mddev->gendisk);
D
Dan Williams 已提交
2911
	if (sectors > mddev->dev_sectors &&
2912
	    mddev->recovery_cp > mddev->dev_sectors) {
A
Andre Noll 已提交
2913
		mddev->recovery_cp = mddev->dev_sectors;
L
Linus Torvalds 已提交
2914 2915
		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
	}
D
Dan Williams 已提交
2916
	mddev->dev_sectors = sectors;
2917
	mddev->resync_max_sectors = sectors;
L
Linus Torvalds 已提交
2918 2919 2920
	return 0;
}

2921
static int raid1_reshape(struct mddev *mddev)
L
Linus Torvalds 已提交
2922 2923 2924 2925 2926 2927 2928 2929
{
	/* 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.
2930 2931 2932
	 *
	 * 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
2933 2934 2935
	 */
	mempool_t *newpool, *oldpool;
	struct pool_info *newpoolinfo;
2936
	struct raid1_info *newmirrors;
2937
	struct r1conf *conf = mddev->private;
2938
	int cnt, raid_disks;
2939
	unsigned long flags;
2940
	int d, d2, err;
L
Linus Torvalds 已提交
2941

2942
	/* Cannot change chunk_size, layout, or level */
2943
	if (mddev->chunk_sectors != mddev->new_chunk_sectors ||
2944 2945
	    mddev->layout != mddev->new_layout ||
	    mddev->level != mddev->new_level) {
2946
		mddev->new_chunk_sectors = mddev->chunk_sectors;
2947 2948 2949 2950 2951
		mddev->new_layout = mddev->layout;
		mddev->new_level = mddev->level;
		return -EINVAL;
	}

2952 2953 2954
	err = md_allow_write(mddev);
	if (err)
		return err;
2955

2956 2957
	raid_disks = mddev->raid_disks + mddev->delta_disks;

2958 2959 2960 2961 2962 2963
	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
2964
			return -EBUSY;
2965
	}
L
Linus Torvalds 已提交
2966 2967 2968 2969 2970

	newpoolinfo = kmalloc(sizeof(*newpoolinfo), GFP_KERNEL);
	if (!newpoolinfo)
		return -ENOMEM;
	newpoolinfo->mddev = mddev;
2971
	newpoolinfo->raid_disks = raid_disks * 2;
L
Linus Torvalds 已提交
2972 2973 2974 2975 2976 2977 2978

	newpool = mempool_create(NR_RAID1_BIOS, r1bio_pool_alloc,
				 r1bio_pool_free, newpoolinfo);
	if (!newpool) {
		kfree(newpoolinfo);
		return -ENOMEM;
	}
2979
	newmirrors = kzalloc(sizeof(struct raid1_info) * raid_disks * 2,
2980
			     GFP_KERNEL);
L
Linus Torvalds 已提交
2981 2982 2983 2984 2985 2986
	if (!newmirrors) {
		kfree(newpoolinfo);
		mempool_destroy(newpool);
		return -ENOMEM;
	}

2987
	raise_barrier(conf);
L
Linus Torvalds 已提交
2988 2989 2990 2991

	/* ok, everything is stopped */
	oldpool = conf->r1bio_pool;
	conf->r1bio_pool = newpool;
2992

2993
	for (d = d2 = 0; d < conf->raid_disks; d++) {
2994
		struct md_rdev *rdev = conf->mirrors[d].rdev;
2995
		if (rdev && rdev->raid_disk != d2) {
2996
			sysfs_unlink_rdev(mddev, rdev);
2997
			rdev->raid_disk = d2;
2998 2999
			sysfs_unlink_rdev(mddev, rdev);
			if (sysfs_link_rdev(mddev, rdev))
3000
				printk(KERN_WARNING
3001 3002
				       "md/raid1:%s: cannot register rd%d\n",
				       mdname(mddev), rdev->raid_disk);
3003
		}
3004 3005 3006
		if (rdev)
			newmirrors[d2++].rdev = rdev;
	}
L
Linus Torvalds 已提交
3007 3008 3009 3010 3011
	kfree(conf->mirrors);
	conf->mirrors = newmirrors;
	kfree(conf->poolinfo);
	conf->poolinfo = newpoolinfo;

3012
	spin_lock_irqsave(&conf->device_lock, flags);
L
Linus Torvalds 已提交
3013
	mddev->degraded += (raid_disks - conf->raid_disks);
3014
	spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
3015
	conf->raid_disks = mddev->raid_disks = raid_disks;
3016
	mddev->delta_disks = 0;
L
Linus Torvalds 已提交
3017

3018
	lower_barrier(conf);
L
Linus Torvalds 已提交
3019 3020 3021 3022 3023 3024 3025 3026

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

	mempool_destroy(oldpool);
	return 0;
}

3027
static void raid1_quiesce(struct mddev *mddev, int state)
3028
{
3029
	struct r1conf *conf = mddev->private;
3030 3031

	switch(state) {
3032 3033 3034
	case 2: /* wake for suspend */
		wake_up(&conf->wait_barrier);
		break;
3035
	case 1:
3036
		raise_barrier(conf);
3037
		break;
3038
	case 0:
3039
		lower_barrier(conf);
3040 3041 3042 3043
		break;
	}
}

3044
static void *raid1_takeover(struct mddev *mddev)
3045 3046 3047 3048 3049
{
	/* raid1 can take over:
	 *  raid5 with 2 devices, any layout or chunk size
	 */
	if (mddev->level == 5 && mddev->raid_disks == 2) {
3050
		struct r1conf *conf;
3051 3052 3053 3054 3055 3056 3057 3058 3059 3060
		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
3061

3062
static struct md_personality raid1_personality =
L
Linus Torvalds 已提交
3063 3064
{
	.name		= "raid1",
3065
	.level		= 1,
L
Linus Torvalds 已提交
3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076
	.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,
3077
	.size		= raid1_size,
3078
	.check_reshape	= raid1_reshape,
3079
	.quiesce	= raid1_quiesce,
3080
	.takeover	= raid1_takeover,
L
Linus Torvalds 已提交
3081 3082 3083 3084
};

static int __init raid_init(void)
{
3085
	return register_md_personality(&raid1_personality);
L
Linus Torvalds 已提交
3086 3087 3088 3089
}

static void raid_exit(void)
{
3090
	unregister_md_personality(&raid1_personality);
L
Linus Torvalds 已提交
3091 3092 3093 3094 3095
}

module_init(raid_init);
module_exit(raid_exit);
MODULE_LICENSE("GPL");
3096
MODULE_DESCRIPTION("RAID1 (mirroring) personality for MD");
L
Linus Torvalds 已提交
3097
MODULE_ALIAS("md-personality-3"); /* RAID1 */
3098
MODULE_ALIAS("md-raid1");
3099
MODULE_ALIAS("md-level-1");
3100 3101

module_param(max_queued_requests, int, S_IRUGO|S_IWUSR);