raid1.c 86.3 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, sector_t start_next_window,
			  sector_t bi_sector);
71
static void lower_barrier(struct r1conf *conf);
L
Linus Torvalds 已提交
72

A
Al Viro 已提交
73
static void * r1bio_pool_alloc(gfp_t gfp_flags, void *data)
L
Linus Torvalds 已提交
74 75
{
	struct pool_info *pi = data;
76
	int size = offsetof(struct r1bio, bios[pi->raid_disks]);
L
Linus Torvalds 已提交
77 78

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

static void r1bio_pool_free(void *r1_bio, void *data)
{
	kfree(r1_bio);
}

#define RESYNC_BLOCK_SIZE (64*1024)
88
#define RESYNC_DEPTH 32
L
Linus Torvalds 已提交
89 90
#define RESYNC_SECTORS (RESYNC_BLOCK_SIZE >> 9)
#define RESYNC_PAGES ((RESYNC_BLOCK_SIZE + PAGE_SIZE-1) / PAGE_SIZE)
91 92 93
#define RESYNC_WINDOW (RESYNC_BLOCK_SIZE * RESYNC_DEPTH)
#define RESYNC_WINDOW_SECTORS (RESYNC_WINDOW >> 9)
#define NEXT_NORMALIO_DISTANCE (3 * RESYNC_WINDOW_SECTORS)
L
Linus Torvalds 已提交
94

A
Al Viro 已提交
95
static void * r1buf_pool_alloc(gfp_t gfp_flags, void *data)
L
Linus Torvalds 已提交
96 97
{
	struct pool_info *pi = data;
98
	struct r1bio *r1_bio;
L
Linus Torvalds 已提交
99 100 101 102
	struct bio *bio;
	int i, j;

	r1_bio = r1bio_pool_alloc(gfp_flags, pi);
J
Jens Axboe 已提交
103
	if (!r1_bio)
L
Linus Torvalds 已提交
104 105 106 107 108 109
		return NULL;

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

K
Kent Overstreet 已提交
129 130
		if (bio_alloc_pages(bio, gfp_flags))
			goto out_free_bio;
131 132 133 134 135 136 137
	}
	/* 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
138 139 140 141 142 143 144
	}

	r1_bio->master_bio = NULL;

	return r1_bio;

out_free_bio:
145
	while (++j < pi->raid_disks)
L
Linus Torvalds 已提交
146 147 148 149 150 151 152 153
		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;
154
	int i,j;
155
	struct r1bio *r1bio = __r1_bio;
L
Linus Torvalds 已提交
156

157 158 159 160 161
	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)
162
				safe_put_page(r1bio->bios[j]->bi_io_vec[i].bv_page);
163
		}
L
Linus Torvalds 已提交
164 165 166 167 168 169
	for (i=0 ; i < pi->raid_disks; i++)
		bio_put(r1bio->bios[i]);

	r1bio_pool_free(r1bio, data);
}

170
static void put_all_bios(struct r1conf *conf, struct r1bio *r1_bio)
L
Linus Torvalds 已提交
171 172 173
{
	int i;

174
	for (i = 0; i < conf->raid_disks * 2; i++) {
L
Linus Torvalds 已提交
175
		struct bio **bio = r1_bio->bios + i;
176
		if (!BIO_SPECIAL(*bio))
L
Linus Torvalds 已提交
177 178 179 180 181
			bio_put(*bio);
		*bio = NULL;
	}
}

182
static void free_r1bio(struct r1bio *r1_bio)
L
Linus Torvalds 已提交
183
{
184
	struct r1conf *conf = r1_bio->mddev->private;
L
Linus Torvalds 已提交
185 186 187 188 189

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

190
static void put_buf(struct r1bio *r1_bio)
L
Linus Torvalds 已提交
191
{
192
	struct r1conf *conf = r1_bio->mddev->private;
193 194
	int i;

195
	for (i = 0; i < conf->raid_disks * 2; i++) {
196 197 198 199
		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
200 201 202

	mempool_free(r1_bio, conf->r1buf_pool);

203
	lower_barrier(conf);
L
Linus Torvalds 已提交
204 205
}

206
static void reschedule_retry(struct r1bio *r1_bio)
L
Linus Torvalds 已提交
207 208
{
	unsigned long flags;
209
	struct mddev *mddev = r1_bio->mddev;
210
	struct r1conf *conf = mddev->private;
L
Linus Torvalds 已提交
211 212 213

	spin_lock_irqsave(&conf->device_lock, flags);
	list_add(&r1_bio->retry_list, &conf->retry_list);
214
	conf->nr_queued ++;
L
Linus Torvalds 已提交
215 216
	spin_unlock_irqrestore(&conf->device_lock, flags);

217
	wake_up(&conf->wait_barrier);
L
Linus Torvalds 已提交
218 219 220 221 222 223 224 225
	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.
 */
226
static void call_bio_endio(struct r1bio *r1_bio)
227 228 229
{
	struct bio *bio = r1_bio->master_bio;
	int done;
230
	struct r1conf *conf = r1_bio->mddev->private;
231
	sector_t start_next_window = r1_bio->start_next_window;
232
	sector_t bi_sector = bio->bi_iter.bi_sector;
233 234 235 236 237 238 239

	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);
240 241 242 243 244
		/*
		 * make_request() might be waiting for
		 * bi_phys_segments to decrease
		 */
		wake_up(&conf->wait_barrier);
245 246 247 248 249 250 251 252 253 254 255
	} 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.
		 */
256
		allow_barrier(conf, start_next_window, bi_sector);
257 258 259
	}
}

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
		pr_debug("raid1: sync end %s on sectors %llu-%llu\n",
			 (bio_data_dir(bio) == WRITE) ? "write" : "read",
268 269
			 (unsigned long long) bio->bi_iter.bi_sector,
			 (unsigned long long) bio_end_sector(bio) - 1);
270

271
		call_bio_endio(r1_bio);
272
	}
L
Linus Torvalds 已提交
273 274 275 276 277 278
	free_r1bio(r1_bio);
}

/*
 * Update disk head position estimator based on IRQ completion info.
 */
279
static inline void update_head_pos(int disk, struct r1bio *r1_bio)
L
Linus Torvalds 已提交
280
{
281
	struct r1conf *conf = r1_bio->mddev->private;
L
Linus Torvalds 已提交
282 283 284 285 286

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

287 288 289
/*
 * Find the disk number which triggered given bio
 */
290
static int find_bio_disk(struct r1bio *r1_bio, struct bio *bio)
291 292
{
	int mirror;
293 294
	struct r1conf *conf = r1_bio->mddev->private;
	int raid_disks = conf->raid_disks;
295

296
	for (mirror = 0; mirror < raid_disks * 2; mirror++)
297 298 299
		if (r1_bio->bios[mirror] == bio)
			break;

300
	BUG_ON(mirror == raid_disks * 2);
301 302 303 304 305
	update_head_pos(mirror, r1_bio);

	return mirror;
}

306
static void raid1_end_read_request(struct bio *bio, int error)
L
Linus Torvalds 已提交
307 308
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
309
	struct r1bio *r1_bio = bio->bi_private;
L
Linus Torvalds 已提交
310
	int mirror;
311
	struct r1conf *conf = r1_bio->mddev->private;
L
Linus Torvalds 已提交
312 313 314 315 316

	mirror = r1_bio->read_disk;
	/*
	 * this branch is our 'one mirror IO has finished' event handler:
	 */
317 318
	update_head_pos(mirror, r1_bio);

319 320 321 322 323 324
	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
325
		 */
326 327 328 329 330 331 332 333
		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
334

335
	if (uptodate) {
L
Linus Torvalds 已提交
336
		raid_end_bio_io(r1_bio);
337 338
		rdev_dec_pending(conf->mirrors[mirror].rdev, conf->mddev);
	} else {
L
Linus Torvalds 已提交
339 340 341 342
		/*
		 * oops, read error:
		 */
		char b[BDEVNAME_SIZE];
343 344 345 346 347 348 349
		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);
350
		set_bit(R1BIO_ReadError, &r1_bio->state);
L
Linus Torvalds 已提交
351
		reschedule_retry(r1_bio);
352
		/* don't drop the reference on read_disk yet */
L
Linus Torvalds 已提交
353 354 355
	}
}

356
static void close_write(struct r1bio *r1_bio)
357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374
{
	/* 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);
}

375
static void r1_bio_write_done(struct r1bio *r1_bio)
376
{
377 378 379 380 381 382 383
	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);
384 385 386 387
		if (test_bit(R1BIO_MadeGood, &r1_bio->state))
			reschedule_retry(r1_bio);
		else
			raid_end_bio_io(r1_bio);
388 389 390
	}
}

391
static void raid1_end_write_request(struct bio *bio, int error)
L
Linus Torvalds 已提交
392 393
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
394
	struct r1bio *r1_bio = bio->bi_private;
395
	int mirror, behind = test_bit(R1BIO_BehindIO, &r1_bio->state);
396
	struct r1conf *conf = r1_bio->mddev->private;
397
	struct bio *to_put = NULL;
L
Linus Torvalds 已提交
398

399
	mirror = find_bio_disk(r1_bio, bio);
L
Linus Torvalds 已提交
400

T
Tejun Heo 已提交
401 402 403 404
	/*
	 * 'one mirror IO has finished' event handler:
	 */
	if (!uptodate) {
405 406
		set_bit(WriteErrorSeen,
			&conf->mirrors[mirror].rdev->flags);
407 408 409 410 411
		if (!test_and_set_bit(WantReplacement,
				      &conf->mirrors[mirror].rdev->flags))
			set_bit(MD_RECOVERY_NEEDED, &
				conf->mddev->recovery);

412
		set_bit(R1BIO_WriteError, &r1_bio->state);
413
	} else {
L
Linus Torvalds 已提交
414
		/*
T
Tejun Heo 已提交
415 416 417 418 419 420 421 422
		 * 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
423
		 */
424 425 426
		sector_t first_bad;
		int bad_sectors;

427 428
		r1_bio->bios[mirror] = NULL;
		to_put = bio;
429 430 431 432 433 434 435 436 437 438 439
		/*
		 * Do not set R1BIO_Uptodate if the current device is
		 * rebuilding or Faulty. This is because we cannot use
		 * such device for properly reading the data back (we could
		 * potentially use it, if the current write would have felt
		 * before rdev->recovery_offset, but for simplicity we don't
		 * check this here.
		 */
		if (test_bit(In_sync, &conf->mirrors[mirror].rdev->flags) &&
		    !test_bit(Faulty, &conf->mirrors[mirror].rdev->flags))
			set_bit(R1BIO_Uptodate, &r1_bio->state);
T
Tejun Heo 已提交
440

441 442 443 444 445 446 447 448 449
		/* 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
450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465
	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;
466 467
				pr_debug("raid1: behind end write sectors"
					 " %llu-%llu\n",
468 469
					 (unsigned long long) mbio->bi_iter.bi_sector,
					 (unsigned long long) bio_end_sector(mbio) - 1);
470
				call_bio_endio(r1_bio);
471 472 473
			}
		}
	}
474 475 476
	if (r1_bio->bios[mirror] == NULL)
		rdev_dec_pending(conf->mirrors[mirror].rdev,
				 conf->mddev);
T
Tejun Heo 已提交
477

L
Linus Torvalds 已提交
478 479 480 481
	/*
	 * Let's see if all mirrored write operations have finished
	 * already.
	 */
482
	r1_bio_write_done(r1_bio);
483

484 485
	if (to_put)
		bio_put(to_put);
L
Linus Torvalds 已提交
486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502
}


/*
 * 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.
 */
503
static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sectors)
L
Linus Torvalds 已提交
504
{
505
	const sector_t this_sector = r1_bio->sector;
506 507
	int sectors;
	int best_good_sectors;
508 509
	int best_disk, best_dist_disk, best_pending_disk;
	int has_nonrot_disk;
510
	int disk;
N
NeilBrown 已提交
511
	sector_t best_dist;
512
	unsigned int min_pending;
513
	struct md_rdev *rdev;
514
	int choose_first;
515
	int choose_next_idle;
L
Linus Torvalds 已提交
516 517 518

	rcu_read_lock();
	/*
519
	 * Check if we can balance. We can balance on the whole
L
Linus Torvalds 已提交
520 521 522 523
	 * device if no resync is going on, or below the resync window.
	 * We take the first readable disk when above the resync window.
	 */
 retry:
524
	sectors = r1_bio->sectors;
N
NeilBrown 已提交
525
	best_disk = -1;
526
	best_dist_disk = -1;
N
NeilBrown 已提交
527
	best_dist = MaxSector;
528 529
	best_pending_disk = -1;
	min_pending = UINT_MAX;
530
	best_good_sectors = 0;
531
	has_nonrot_disk = 0;
532
	choose_next_idle = 0;
533

L
Linus Torvalds 已提交
534
	if (conf->mddev->recovery_cp < MaxSector &&
535
	    (this_sector + sectors >= conf->next_resync))
536
		choose_first = 1;
537
	else
538
		choose_first = 0;
L
Linus Torvalds 已提交
539

540
	for (disk = 0 ; disk < conf->raid_disks * 2 ; disk++) {
N
NeilBrown 已提交
541
		sector_t dist;
542 543
		sector_t first_bad;
		int bad_sectors;
544
		unsigned int pending;
545
		bool nonrot;
546

547 548 549
		rdev = rcu_dereference(conf->mirrors[disk].rdev);
		if (r1_bio->bios[disk] == IO_BLOCKED
		    || rdev == NULL
550
		    || test_bit(Unmerged, &rdev->flags)
N
NeilBrown 已提交
551
		    || test_bit(Faulty, &rdev->flags))
552
			continue;
N
NeilBrown 已提交
553 554
		if (!test_bit(In_sync, &rdev->flags) &&
		    rdev->recovery_offset < this_sector + sectors)
L
Linus Torvalds 已提交
555
			continue;
N
NeilBrown 已提交
556 557 558
		if (test_bit(WriteMostly, &rdev->flags)) {
			/* Don't balance among write-mostly, just
			 * use the first as a last resort */
559 560 561 562 563 564 565 566 567
			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
568
				best_disk = disk;
569
			}
N
NeilBrown 已提交
570 571 572 573 574
			continue;
		}
		/* This is a reasonable device to use.  It might
		 * even be best.
		 */
575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603
		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;

604 605
		nonrot = blk_queue_nonrot(bdev_get_queue(rdev->bdev));
		has_nonrot_disk |= nonrot;
606
		pending = atomic_read(&rdev->nr_pending);
N
NeilBrown 已提交
607
		dist = abs(this_sector - conf->mirrors[disk].head_position);
608
		if (choose_first) {
N
NeilBrown 已提交
609
			best_disk = disk;
L
Linus Torvalds 已提交
610 611
			break;
		}
612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649
		/* 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;
650 651 652 653 654 655

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

N
NeilBrown 已提交
656 657
		if (dist < best_dist) {
			best_dist = dist;
658
			best_dist_disk = disk;
L
Linus Torvalds 已提交
659
		}
660
	}
L
Linus Torvalds 已提交
661

662 663 664 665 666 667 668 669 670 671 672 673 674
	/*
	 * 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
675 676
	if (best_disk >= 0) {
		rdev = rcu_dereference(conf->mirrors[best_disk].rdev);
677 678 679
		if (!rdev)
			goto retry;
		atomic_inc(&rdev->nr_pending);
N
NeilBrown 已提交
680
		if (test_bit(Faulty, &rdev->flags)) {
L
Linus Torvalds 已提交
681 682 683
			/* cannot risk returning a device that failed
			 * before we inc'ed nr_pending
			 */
684
			rdev_dec_pending(rdev, conf->mddev);
L
Linus Torvalds 已提交
685 686
			goto retry;
		}
687
		sectors = best_good_sectors;
688 689 690 691

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

692
		conf->mirrors[best_disk].next_seq_sect = this_sector + sectors;
L
Linus Torvalds 已提交
693 694
	}
	rcu_read_unlock();
695
	*max_sectors = sectors;
L
Linus Torvalds 已提交
696

N
NeilBrown 已提交
697
	return best_disk;
L
Linus Torvalds 已提交
698 699
}

700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732
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;

}

733
int md_raid1_congested(struct mddev *mddev, int bits)
734
{
735
	struct r1conf *conf = mddev->private;
736 737
	int i, ret = 0;

738 739 740 741
	if ((bits & (1 << BDI_async_congested)) &&
	    conf->pending_count >= max_queued_requests)
		return 1;

742
	rcu_read_lock();
743
	for (i = 0; i < conf->raid_disks * 2; i++) {
744
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
745
		if (rdev && !test_bit(Faulty, &rdev->flags)) {
746
			struct request_queue *q = bdev_get_queue(rdev->bdev);
747

748 749
			BUG_ON(!q);

750 751 752
			/* Note the '|| 1' - when read_balance prefers
			 * non-congested targets, it can be removed
			 */
753
			if ((bits & (1<<BDI_async_congested)) || 1)
754 755 756 757 758 759 760 761
				ret |= bdi_congested(&q->backing_dev_info, bits);
			else
				ret &= bdi_congested(&q->backing_dev_info, bits);
		}
	}
	rcu_read_unlock();
	return ret;
}
762
EXPORT_SYMBOL_GPL(md_raid1_congested);
763

764 765
static int raid1_congested(void *data, int bits)
{
766
	struct mddev *mddev = data;
767 768 769 770

	return mddev_congested(mddev, bits) ||
		md_raid1_congested(mddev, bits);
}
771

772
static void flush_pending_writes(struct r1conf *conf)
773 774 775 776 777 778 779 780 781
{
	/* 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);
782
		conf->pending_count = 0;
783 784 785 786
		spin_unlock_irq(&conf->device_lock);
		/* flush any pending bitmap writes to
		 * disk before proceeding w/ I/O */
		bitmap_unplug(conf->mddev->bitmap);
787
		wake_up(&conf->wait_barrier);
788 789 790 791

		while (bio) { /* submit pending writes */
			struct bio *next = bio->bi_next;
			bio->bi_next = NULL;
S
Shaohua Li 已提交
792 793 794 795 796 797
			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);
798 799 800 801
			bio = next;
		}
	} else
		spin_unlock_irq(&conf->device_lock);
J
Jens Axboe 已提交
802 803
}

804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823
/* 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
824
 */
825
static void raise_barrier(struct r1conf *conf)
L
Linus Torvalds 已提交
826 827
{
	spin_lock_irq(&conf->resync_lock);
828 829 830

	/* Wait until no block IO is waiting */
	wait_event_lock_irq(conf->wait_barrier, !conf->nr_waiting,
831
			    conf->resync_lock);
832 833 834 835

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

836 837 838 839 840 841 842 843
	/* For these conditions we must wait:
	 * A: while the array is in frozen state
	 * B: while barrier >= RESYNC_DEPTH, meaning resync reach
	 *    the max count which allowed.
	 * C: next_resync + RESYNC_SECTORS > start_next_window, meaning
	 *    next resync will reach to the window which normal bios are
	 *    handling.
	 */
844
	wait_event_lock_irq(conf->wait_barrier,
845
			    !conf->array_frozen &&
846 847 848
			    conf->barrier < RESYNC_DEPTH &&
			    (conf->start_next_window >=
			     conf->next_resync + RESYNC_SECTORS),
849
			    conf->resync_lock);
850 851 852 853

	spin_unlock_irq(&conf->resync_lock);
}

854
static void lower_barrier(struct r1conf *conf)
855 856
{
	unsigned long flags;
857
	BUG_ON(conf->barrier <= 0);
858 859 860 861 862 863
	spin_lock_irqsave(&conf->resync_lock, flags);
	conf->barrier--;
	spin_unlock_irqrestore(&conf->resync_lock, flags);
	wake_up(&conf->wait_barrier);
}

864
static bool need_to_wait_for_sync(struct r1conf *conf, struct bio *bio)
865
{
866 867 868 869 870 871 872 873 874 875
	bool wait = false;

	if (conf->array_frozen || !bio)
		wait = true;
	else if (conf->barrier && bio_data_dir(bio) == WRITE) {
		if (conf->next_resync < RESYNC_WINDOW_SECTORS)
			wait = true;
		else if ((conf->next_resync - RESYNC_WINDOW_SECTORS
				>= bio_end_sector(bio)) ||
			 (conf->next_resync + NEXT_NORMALIO_DISTANCE
876
				<= bio->bi_iter.bi_sector))
877 878 879 880 881 882 883 884 885 886 887 888
			wait = false;
		else
			wait = true;
	}

	return wait;
}

static sector_t wait_barrier(struct r1conf *conf, struct bio *bio)
{
	sector_t sector = 0;

889
	spin_lock_irq(&conf->resync_lock);
890
	if (need_to_wait_for_sync(conf, bio)) {
891
		conf->nr_waiting++;
892 893 894 895 896 897 898 899 900 901
		/* 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,
902 903
				    !conf->array_frozen &&
				    (!conf->barrier ||
904 905
				    ((conf->start_next_window <
				      conf->next_resync + RESYNC_SECTORS) &&
906
				     current->bio_list &&
907
				     !bio_list_empty(current->bio_list))),
908
				    conf->resync_lock);
909
		conf->nr_waiting--;
L
Linus Torvalds 已提交
910
	}
911 912 913

	if (bio && bio_data_dir(bio) == WRITE) {
		if (conf->next_resync + NEXT_NORMALIO_DISTANCE
914
		    <= bio->bi_iter.bi_sector) {
915 916 917 918 919 920
			if (conf->start_next_window == MaxSector)
				conf->start_next_window =
					conf->next_resync +
					NEXT_NORMALIO_DISTANCE;

			if ((conf->start_next_window + NEXT_NORMALIO_DISTANCE)
921
			    <= bio->bi_iter.bi_sector)
922 923 924 925
				conf->next_window_requests++;
			else
				conf->current_window_requests++;
			sector = conf->start_next_window;
926
		}
927 928
	}

929
	conf->nr_pending++;
L
Linus Torvalds 已提交
930
	spin_unlock_irq(&conf->resync_lock);
931
	return sector;
L
Linus Torvalds 已提交
932 933
}

934 935
static void allow_barrier(struct r1conf *conf, sector_t start_next_window,
			  sector_t bi_sector)
936 937
{
	unsigned long flags;
938

939 940
	spin_lock_irqsave(&conf->resync_lock, flags);
	conf->nr_pending--;
941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961
	if (start_next_window) {
		if (start_next_window == conf->start_next_window) {
			if (conf->start_next_window + NEXT_NORMALIO_DISTANCE
			    <= bi_sector)
				conf->next_window_requests--;
			else
				conf->current_window_requests--;
		} else
			conf->current_window_requests--;

		if (!conf->current_window_requests) {
			if (conf->next_window_requests) {
				conf->current_window_requests =
					conf->next_window_requests;
				conf->next_window_requests = 0;
				conf->start_next_window +=
					NEXT_NORMALIO_DISTANCE;
			} else
				conf->start_next_window = MaxSector;
		}
	}
962 963 964 965
	spin_unlock_irqrestore(&conf->resync_lock, flags);
	wake_up(&conf->wait_barrier);
}

966
static void freeze_array(struct r1conf *conf, int extra)
967 968 969
{
	/* stop syncio and normal IO and wait for everything to
	 * go quite.
970
	 * We wait until nr_pending match nr_queued+extra
971 972 973 974
	 * 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.
975
	 * Thus the number queued (nr_queued) plus this request (extra)
976 977
	 * must match the number of pending IOs (nr_pending) before
	 * we continue.
978 979
	 */
	spin_lock_irq(&conf->resync_lock);
980
	conf->array_frozen = 1;
981
	wait_event_lock_irq_cmd(conf->wait_barrier,
982
				conf->nr_pending == conf->nr_queued+extra,
983 984
				conf->resync_lock,
				flush_pending_writes(conf));
985 986
	spin_unlock_irq(&conf->resync_lock);
}
987
static void unfreeze_array(struct r1conf *conf)
988 989 990
{
	/* reverse the effect of the freeze */
	spin_lock_irq(&conf->resync_lock);
991
	conf->array_frozen = 0;
992 993 994 995
	wake_up(&conf->wait_barrier);
	spin_unlock_irq(&conf->resync_lock);
}

996

997 998
/* duplicate the data pages for behind I/O 
 */
999
static void alloc_behind_pages(struct bio *bio, struct r1bio *r1_bio)
1000 1001 1002
{
	int i;
	struct bio_vec *bvec;
1003
	struct bio_vec *bvecs = kzalloc(bio->bi_vcnt * sizeof(struct bio_vec),
1004
					GFP_NOIO);
1005
	if (unlikely(!bvecs))
1006
		return;
1007

1008
	bio_for_each_segment_all(bvec, bio, i) {
1009 1010 1011
		bvecs[i] = *bvec;
		bvecs[i].bv_page = alloc_page(GFP_NOIO);
		if (unlikely(!bvecs[i].bv_page))
1012
			goto do_sync_io;
1013 1014 1015
		memcpy(kmap(bvecs[i].bv_page) + bvec->bv_offset,
		       kmap(bvec->bv_page) + bvec->bv_offset, bvec->bv_len);
		kunmap(bvecs[i].bv_page);
1016 1017
		kunmap(bvec->bv_page);
	}
1018
	r1_bio->behind_bvecs = bvecs;
1019 1020 1021
	r1_bio->behind_page_count = bio->bi_vcnt;
	set_bit(R1BIO_BehindIO, &r1_bio->state);
	return;
1022 1023

do_sync_io:
1024
	for (i = 0; i < bio->bi_vcnt; i++)
1025 1026 1027
		if (bvecs[i].bv_page)
			put_page(bvecs[i].bv_page);
	kfree(bvecs);
1028 1029
	pr_debug("%dB behind alloc failed, doing sync I/O\n",
		 bio->bi_iter.bi_size);
1030 1031
}

1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045
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;

1046
	if (from_schedule || current->bio_list) {
1047 1048 1049 1050
		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);
1051
		wake_up(&conf->wait_barrier);
1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064
		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;
1065 1066 1067 1068 1069 1070
		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);
1071 1072 1073 1074 1075
		bio = next;
	}
	kfree(plug);
}

1076
static void make_request(struct mddev *mddev, struct bio * bio)
L
Linus Torvalds 已提交
1077
{
1078
	struct r1conf *conf = mddev->private;
1079
	struct raid1_info *mirror;
1080
	struct r1bio *r1_bio;
L
Linus Torvalds 已提交
1081
	struct bio *read_bio;
1082
	int i, disks;
1083
	struct bitmap *bitmap;
1084
	unsigned long flags;
1085
	const int rw = bio_data_dir(bio);
1086
	const unsigned long do_sync = (bio->bi_rw & REQ_SYNC);
T
Tejun Heo 已提交
1087
	const unsigned long do_flush_fua = (bio->bi_rw & (REQ_FLUSH | REQ_FUA));
S
Shaohua Li 已提交
1088 1089
	const unsigned long do_discard = (bio->bi_rw
					  & (REQ_DISCARD | REQ_SECURE));
1090
	const unsigned long do_same = (bio->bi_rw & REQ_WRITE_SAME);
1091
	struct md_rdev *blocked_rdev;
1092 1093
	struct blk_plug_cb *cb;
	struct raid1_plug_cb *plug = NULL;
1094 1095 1096
	int first_clone;
	int sectors_handled;
	int max_sectors;
1097
	sector_t start_next_window;
1098

L
Linus Torvalds 已提交
1099 1100 1101 1102 1103
	/*
	 * 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.
	 */
1104

1105 1106
	md_write_start(mddev, bio); /* wait on superblock update early */

1107
	if (bio_data_dir(bio) == WRITE &&
K
Kent Overstreet 已提交
1108
	    bio_end_sector(bio) > mddev->suspend_lo &&
1109
	    bio->bi_iter.bi_sector < mddev->suspend_hi) {
1110 1111 1112 1113 1114 1115 1116 1117 1118
		/* 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
1119
			if (bio_end_sector(bio) <= mddev->suspend_lo ||
1120
			    bio->bi_iter.bi_sector >= mddev->suspend_hi)
1121 1122 1123 1124 1125
				break;
			schedule();
		}
		finish_wait(&conf->wait_barrier, &w);
	}
1126

1127
	start_next_window = wait_barrier(conf, bio);
L
Linus Torvalds 已提交
1128

1129 1130
	bitmap = mddev->bitmap;

L
Linus Torvalds 已提交
1131 1132 1133 1134 1135 1136 1137 1138
	/*
	 * 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;
1139
	r1_bio->sectors = bio_sectors(bio);
1140
	r1_bio->state = 0;
L
Linus Torvalds 已提交
1141
	r1_bio->mddev = mddev;
1142
	r1_bio->sector = bio->bi_iter.bi_sector;
L
Linus Torvalds 已提交
1143

1144 1145 1146 1147 1148 1149 1150 1151 1152 1153
	/* 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);

1154
	if (rw == READ) {
L
Linus Torvalds 已提交
1155 1156 1157
		/*
		 * read balancing logic:
		 */
1158 1159 1160 1161
		int rdisk;

read_again:
		rdisk = read_balance(conf, r1_bio, &max_sectors);
L
Linus Torvalds 已提交
1162 1163 1164 1165

		if (rdisk < 0) {
			/* couldn't find anywhere to read from */
			raid_end_bio_io(r1_bio);
1166
			return;
L
Linus Torvalds 已提交
1167 1168 1169
		}
		mirror = conf->mirrors + rdisk;

1170 1171 1172 1173 1174 1175 1176 1177 1178
		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
1179 1180
		r1_bio->read_disk = rdisk;

1181
		read_bio = bio_clone_mddev(bio, GFP_NOIO, mddev);
1182
		bio_trim(read_bio, r1_bio->sector - bio->bi_iter.bi_sector,
1183
			 max_sectors);
L
Linus Torvalds 已提交
1184 1185 1186

		r1_bio->bios[rdisk] = read_bio;

1187 1188
		read_bio->bi_iter.bi_sector = r1_bio->sector +
			mirror->rdev->data_offset;
L
Linus Torvalds 已提交
1189 1190
		read_bio->bi_bdev = mirror->rdev->bdev;
		read_bio->bi_end_io = raid1_end_read_request;
1191
		read_bio->bi_rw = READ | do_sync;
L
Linus Torvalds 已提交
1192 1193
		read_bio->bi_private = r1_bio;

1194 1195 1196 1197 1198 1199
		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
1200
					   - bio->bi_iter.bi_sector);
1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217
			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;
1218
			r1_bio->sectors = bio_sectors(bio) - sectors_handled;
1219 1220
			r1_bio->state = 0;
			r1_bio->mddev = mddev;
1221 1222
			r1_bio->sector = bio->bi_iter.bi_sector +
				sectors_handled;
1223 1224 1225
			goto read_again;
		} else
			generic_make_request(read_bio);
1226
		return;
L
Linus Torvalds 已提交
1227 1228 1229 1230 1231
	}

	/*
	 * WRITE:
	 */
1232 1233 1234 1235 1236
	if (conf->pending_count >= max_queued_requests) {
		md_wakeup_thread(mddev->thread);
		wait_event(conf->wait_barrier,
			   conf->pending_count < max_queued_requests);
	}
1237
	/* first select target devices under rcu_lock and
L
Linus Torvalds 已提交
1238 1239
	 * inc refcount on their rdev.  Record them by setting
	 * bios[x] to bio
1240 1241 1242 1243 1244 1245
	 * 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
1246
	 */
N
NeilBrown 已提交
1247

1248
	disks = conf->raid_disks * 2;
1249
 retry_write:
1250
	r1_bio->start_next_window = start_next_window;
1251
	blocked_rdev = NULL;
L
Linus Torvalds 已提交
1252
	rcu_read_lock();
1253
	max_sectors = r1_bio->sectors;
L
Linus Torvalds 已提交
1254
	for (i = 0;  i < disks; i++) {
1255
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
1256 1257 1258 1259 1260
		if (rdev && unlikely(test_bit(Blocked, &rdev->flags))) {
			atomic_inc(&rdev->nr_pending);
			blocked_rdev = rdev;
			break;
		}
1261
		r1_bio->bios[i] = NULL;
1262 1263
		if (!rdev || test_bit(Faulty, &rdev->flags)
		    || test_bit(Unmerged, &rdev->flags)) {
1264 1265
			if (i < conf->raid_disks)
				set_bit(R1BIO_Degraded, &r1_bio->state);
1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292
			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;
1293
				rdev_dec_pending(rdev, mddev);
1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304
				/* 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;
1305
			}
1306 1307 1308 1309 1310 1311 1312
			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
1313 1314 1315
	}
	rcu_read_unlock();

1316 1317 1318
	if (unlikely(blocked_rdev)) {
		/* Wait for this device to become unblocked */
		int j;
1319
		sector_t old = start_next_window;
1320 1321 1322 1323

		for (j = 0; j < i; j++)
			if (r1_bio->bios[j])
				rdev_dec_pending(conf->mirrors[j].rdev, mddev);
1324
		r1_bio->state = 0;
1325
		allow_barrier(conf, start_next_window, bio->bi_iter.bi_sector);
1326
		md_wait_for_blocked_rdev(blocked_rdev, mddev);
1327 1328 1329 1330 1331 1332 1333 1334 1335 1336
		start_next_window = wait_barrier(conf, bio);
		/*
		 * We must make sure the multi r1bios of bio have
		 * the same value of bi_phys_segments
		 */
		if (bio->bi_phys_segments && old &&
		    old != start_next_window)
			/* Wait for the former r1bio(s) to complete */
			wait_event(conf->wait_barrier,
				   bio->bi_phys_segments == 1);
1337 1338 1339
		goto retry_write;
	}

1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350
	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);
1351
	}
1352
	sectors_handled = r1_bio->sector + max_sectors - bio->bi_iter.bi_sector;
1353

1354
	atomic_set(&r1_bio->remaining, 1);
1355
	atomic_set(&r1_bio->behind_remaining, 0);
1356

1357
	first_clone = 1;
L
Linus Torvalds 已提交
1358 1359 1360 1361 1362
	for (i = 0; i < disks; i++) {
		struct bio *mbio;
		if (!r1_bio->bios[i])
			continue;

1363
		mbio = bio_clone_mddev(bio, GFP_NOIO, mddev);
1364
		bio_trim(mbio, r1_bio->sector - bio->bi_iter.bi_sector, max_sectors);
1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382

		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;
		}
1383
		if (r1_bio->behind_bvecs) {
1384 1385 1386
			struct bio_vec *bvec;
			int j;

1387 1388
			/*
			 * We trimmed the bio, so _all is legit
1389
			 */
1390
			bio_for_each_segment_all(bvec, mbio, j)
1391
				bvec->bv_page = r1_bio->behind_bvecs[j].bv_page;
1392 1393 1394 1395
			if (test_bit(WriteMostly, &conf->mirrors[i].rdev->flags))
				atomic_inc(&r1_bio->behind_remaining);
		}

1396 1397
		r1_bio->bios[i] = mbio;

1398
		mbio->bi_iter.bi_sector	= (r1_bio->sector +
1399 1400 1401
				   conf->mirrors[i].rdev->data_offset);
		mbio->bi_bdev = conf->mirrors[i].rdev->bdev;
		mbio->bi_end_io	= raid1_end_write_request;
1402 1403
		mbio->bi_rw =
			WRITE | do_flush_fua | do_sync | do_discard | do_same;
1404 1405
		mbio->bi_private = r1_bio;

L
Linus Torvalds 已提交
1406
		atomic_inc(&r1_bio->remaining);
1407 1408 1409 1410 1411 1412

		cb = blk_check_plugged(raid1_unplug, mddev, sizeof(*plug));
		if (cb)
			plug = container_of(cb, struct raid1_plug_cb, cb);
		else
			plug = NULL;
1413
		spin_lock_irqsave(&conf->device_lock, flags);
1414 1415 1416 1417 1418 1419 1420
		if (plug) {
			bio_list_add(&plug->pending, mbio);
			plug->pending_cnt++;
		} else {
			bio_list_add(&conf->pending_bio_list, mbio);
			conf->pending_count++;
		}
1421
		spin_unlock_irqrestore(&conf->device_lock, flags);
1422
		if (!plug)
N
NeilBrown 已提交
1423
			md_wakeup_thread(mddev->thread);
L
Linus Torvalds 已提交
1424
	}
1425 1426 1427
	/* Mustn't call r1_bio_write_done before this next test,
	 * as it could result in the bio being freed.
	 */
1428
	if (sectors_handled < bio_sectors(bio)) {
1429
		r1_bio_write_done(r1_bio);
1430 1431 1432 1433 1434
		/* 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;
1435
		r1_bio->sectors = bio_sectors(bio) - sectors_handled;
1436 1437
		r1_bio->state = 0;
		r1_bio->mddev = mddev;
1438
		r1_bio->sector = bio->bi_iter.bi_sector + sectors_handled;
1439 1440 1441
		goto retry_write;
	}

1442 1443 1444 1445
	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
1446 1447
}

1448
static void status(struct seq_file *seq, struct mddev *mddev)
L
Linus Torvalds 已提交
1449
{
1450
	struct r1conf *conf = mddev->private;
L
Linus Torvalds 已提交
1451 1452 1453
	int i;

	seq_printf(seq, " [%d/%d] [", conf->raid_disks,
1454
		   conf->raid_disks - mddev->degraded);
1455 1456
	rcu_read_lock();
	for (i = 0; i < conf->raid_disks; i++) {
1457
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
L
Linus Torvalds 已提交
1458
		seq_printf(seq, "%s",
1459 1460 1461
			   rdev && test_bit(In_sync, &rdev->flags) ? "U" : "_");
	}
	rcu_read_unlock();
L
Linus Torvalds 已提交
1462 1463 1464 1465
	seq_printf(seq, "]");
}


1466
static void error(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
1467 1468
{
	char b[BDEVNAME_SIZE];
1469
	struct r1conf *conf = mddev->private;
L
Linus Torvalds 已提交
1470 1471 1472 1473 1474 1475 1476

	/*
	 * 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
	 */
1477
	if (test_bit(In_sync, &rdev->flags)
1478
	    && (conf->raid_disks - mddev->degraded) == 1) {
L
Linus Torvalds 已提交
1479 1480
		/*
		 * Don't fail the drive, act as though we were just a
1481 1482 1483
		 * 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
1484
		 */
1485
		conf->recovery_disabled = mddev->recovery_disabled;
L
Linus Torvalds 已提交
1486
		return;
1487
	}
1488
	set_bit(Blocked, &rdev->flags);
1489 1490 1491
	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
1492
		mddev->degraded++;
1493
		set_bit(Faulty, &rdev->flags);
1494
		spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1495 1496 1497
		/*
		 * if recovery is running, make sure it aborts.
		 */
1498
		set_bit(MD_RECOVERY_INTR, &mddev->recovery);
1499 1500
	} else
		set_bit(Faulty, &rdev->flags);
1501
	set_bit(MD_CHANGE_DEVS, &mddev->flags);
1502 1503 1504
	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
1505 1506
	       mdname(mddev), bdevname(rdev->bdev, b),
	       mdname(mddev), conf->raid_disks - mddev->degraded);
L
Linus Torvalds 已提交
1507 1508
}

1509
static void print_conf(struct r1conf *conf)
L
Linus Torvalds 已提交
1510 1511 1512
{
	int i;

N
NeilBrown 已提交
1513
	printk(KERN_DEBUG "RAID1 conf printout:\n");
L
Linus Torvalds 已提交
1514
	if (!conf) {
N
NeilBrown 已提交
1515
		printk(KERN_DEBUG "(!conf)\n");
L
Linus Torvalds 已提交
1516 1517
		return;
	}
N
NeilBrown 已提交
1518
	printk(KERN_DEBUG " --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
L
Linus Torvalds 已提交
1519 1520
		conf->raid_disks);

1521
	rcu_read_lock();
L
Linus Torvalds 已提交
1522 1523
	for (i = 0; i < conf->raid_disks; i++) {
		char b[BDEVNAME_SIZE];
1524
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
1525
		if (rdev)
N
NeilBrown 已提交
1526
			printk(KERN_DEBUG " disk %d, wo:%d, o:%d, dev:%s\n",
1527 1528 1529
			       i, !test_bit(In_sync, &rdev->flags),
			       !test_bit(Faulty, &rdev->flags),
			       bdevname(rdev->bdev,b));
L
Linus Torvalds 已提交
1530
	}
1531
	rcu_read_unlock();
L
Linus Torvalds 已提交
1532 1533
}

1534
static void close_sync(struct r1conf *conf)
L
Linus Torvalds 已提交
1535
{
1536 1537
	wait_barrier(conf, NULL);
	allow_barrier(conf, 0, 0);
L
Linus Torvalds 已提交
1538 1539 1540

	mempool_destroy(conf->r1buf_pool);
	conf->r1buf_pool = NULL;
1541 1542 1543

	conf->next_resync = 0;
	conf->start_next_window = MaxSector;
L
Linus Torvalds 已提交
1544 1545
}

1546
static int raid1_spare_active(struct mddev *mddev)
L
Linus Torvalds 已提交
1547 1548
{
	int i;
1549
	struct r1conf *conf = mddev->private;
1550 1551
	int count = 0;
	unsigned long flags;
L
Linus Torvalds 已提交
1552 1553 1554

	/*
	 * Find all failed disks within the RAID1 configuration 
1555 1556
	 * and mark them readable.
	 * Called under mddev lock, so rcu protection not needed.
L
Linus Torvalds 已提交
1557 1558
	 */
	for (i = 0; i < conf->raid_disks; i++) {
1559
		struct md_rdev *rdev = conf->mirrors[i].rdev;
1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578
		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);
			}
		}
1579
		if (rdev
1580
		    && rdev->recovery_offset == MaxSector
1581
		    && !test_bit(Faulty, &rdev->flags)
1582
		    && !test_and_set_bit(In_sync, &rdev->flags)) {
1583
			count++;
1584
			sysfs_notify_dirent_safe(rdev->sysfs_state);
L
Linus Torvalds 已提交
1585 1586
		}
	}
1587 1588 1589
	spin_lock_irqsave(&conf->device_lock, flags);
	mddev->degraded -= count;
	spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1590 1591

	print_conf(conf);
1592
	return count;
L
Linus Torvalds 已提交
1593 1594 1595
}


1596
static int raid1_add_disk(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
1597
{
1598
	struct r1conf *conf = mddev->private;
1599
	int err = -EEXIST;
1600
	int mirror = 0;
1601
	struct raid1_info *p;
1602
	int first = 0;
1603
	int last = conf->raid_disks - 1;
1604
	struct request_queue *q = bdev_get_queue(rdev->bdev);
L
Linus Torvalds 已提交
1605

1606 1607 1608
	if (mddev->recovery_disabled == conf->recovery_disabled)
		return -EBUSY;

1609 1610 1611
	if (rdev->raid_disk >= 0)
		first = last = rdev->raid_disk;

1612 1613 1614 1615 1616
	if (q->merge_bvec_fn) {
		set_bit(Unmerged, &rdev->flags);
		mddev->merge_check_needed = 1;
	}

1617 1618 1619
	for (mirror = first; mirror <= last; mirror++) {
		p = conf->mirrors+mirror;
		if (!p->rdev) {
L
Linus Torvalds 已提交
1620

1621 1622 1623
			if (mddev->gendisk)
				disk_stack_limits(mddev->gendisk, rdev->bdev,
						  rdev->data_offset << 9);
L
Linus Torvalds 已提交
1624 1625 1626

			p->head_position = 0;
			rdev->raid_disk = mirror;
1627
			err = 0;
1628 1629 1630 1631
			/* 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)
1632
				conf->fullsync = 1;
1633
			rcu_assign_pointer(p->rdev, rdev);
L
Linus Torvalds 已提交
1634 1635
			break;
		}
1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647
		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;
		}
	}
1648 1649 1650 1651 1652 1653 1654 1655 1656
	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();
1657 1658
		freeze_array(conf, 0);
		unfreeze_array(conf);
1659 1660
		clear_bit(Unmerged, &rdev->flags);
	}
1661
	md_integrity_add_rdev(rdev, mddev);
1662
	if (mddev->queue && blk_queue_discard(bdev_get_queue(rdev->bdev)))
S
Shaohua Li 已提交
1663
		queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, mddev->queue);
L
Linus Torvalds 已提交
1664
	print_conf(conf);
1665
	return err;
L
Linus Torvalds 已提交
1666 1667
}

1668
static int raid1_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
1669
{
1670
	struct r1conf *conf = mddev->private;
L
Linus Torvalds 已提交
1671
	int err = 0;
1672
	int number = rdev->raid_disk;
1673
	struct raid1_info *p = conf->mirrors + number;
L
Linus Torvalds 已提交
1674

1675 1676 1677
	if (rdev != p->rdev)
		p = conf->mirrors + conf->raid_disks + number;

L
Linus Torvalds 已提交
1678
	print_conf(conf);
1679
	if (rdev == p->rdev) {
1680
		if (test_bit(In_sync, &rdev->flags) ||
L
Linus Torvalds 已提交
1681 1682 1683 1684
		    atomic_read(&rdev->nr_pending)) {
			err = -EBUSY;
			goto abort;
		}
N
NeilBrown 已提交
1685
		/* Only remove non-faulty devices if recovery
1686 1687 1688
		 * is not possible.
		 */
		if (!test_bit(Faulty, &rdev->flags) &&
1689
		    mddev->recovery_disabled != conf->recovery_disabled &&
1690 1691 1692 1693
		    mddev->degraded < conf->raid_disks) {
			err = -EBUSY;
			goto abort;
		}
L
Linus Torvalds 已提交
1694
		p->rdev = NULL;
1695
		synchronize_rcu();
L
Linus Torvalds 已提交
1696 1697 1698 1699
		if (atomic_read(&rdev->nr_pending)) {
			/* lost the race, try later */
			err = -EBUSY;
			p->rdev = rdev;
1700
			goto abort;
1701 1702 1703 1704 1705 1706 1707
		} 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;
1708
			freeze_array(conf, 0);
1709 1710 1711
			clear_bit(Replacement, &repl->flags);
			p->rdev = repl;
			conf->mirrors[conf->raid_disks + number].rdev = NULL;
1712
			unfreeze_array(conf);
1713 1714
			clear_bit(WantReplacement, &rdev->flags);
		} else
1715
			clear_bit(WantReplacement, &rdev->flags);
1716
		err = md_integrity_register(mddev);
L
Linus Torvalds 已提交
1717 1718 1719 1720 1721 1722 1723 1724
	}
abort:

	print_conf(conf);
	return err;
}


1725
static void end_sync_read(struct bio *bio, int error)
L
Linus Torvalds 已提交
1726
{
1727
	struct r1bio *r1_bio = bio->bi_private;
L
Linus Torvalds 已提交
1728

1729
	update_head_pos(r1_bio->read_disk, r1_bio);
1730

L
Linus Torvalds 已提交
1731 1732 1733 1734 1735
	/*
	 * 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
	 */
1736
	if (test_bit(BIO_UPTODATE, &bio->bi_flags))
L
Linus Torvalds 已提交
1737
		set_bit(R1BIO_Uptodate, &r1_bio->state);
1738 1739 1740

	if (atomic_dec_and_test(&r1_bio->remaining))
		reschedule_retry(r1_bio);
L
Linus Torvalds 已提交
1741 1742
}

1743
static void end_sync_write(struct bio *bio, int error)
L
Linus Torvalds 已提交
1744 1745
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
1746
	struct r1bio *r1_bio = bio->bi_private;
1747
	struct mddev *mddev = r1_bio->mddev;
1748
	struct r1conf *conf = mddev->private;
L
Linus Torvalds 已提交
1749
	int mirror=0;
1750 1751
	sector_t first_bad;
	int bad_sectors;
L
Linus Torvalds 已提交
1752

1753 1754
	mirror = find_bio_disk(r1_bio, bio);

1755
	if (!uptodate) {
N
NeilBrown 已提交
1756
		sector_t sync_blocks = 0;
1757 1758 1759 1760
		sector_t s = r1_bio->sector;
		long sectors_to_go = r1_bio->sectors;
		/* make sure these bits doesn't get cleared. */
		do {
1761
			bitmap_end_sync(mddev->bitmap, s,
1762 1763 1764 1765
					&sync_blocks, 1);
			s += sync_blocks;
			sectors_to_go -= sync_blocks;
		} while (sectors_to_go > 0);
1766 1767
		set_bit(WriteErrorSeen,
			&conf->mirrors[mirror].rdev->flags);
1768 1769 1770 1771
		if (!test_and_set_bit(WantReplacement,
				      &conf->mirrors[mirror].rdev->flags))
			set_bit(MD_RECOVERY_NEEDED, &
				mddev->recovery);
1772
		set_bit(R1BIO_WriteError, &r1_bio->state);
1773 1774 1775
	} else if (is_badblock(conf->mirrors[mirror].rdev,
			       r1_bio->sector,
			       r1_bio->sectors,
1776 1777 1778 1779 1780 1781
			       &first_bad, &bad_sectors) &&
		   !is_badblock(conf->mirrors[r1_bio->read_disk].rdev,
				r1_bio->sector,
				r1_bio->sectors,
				&first_bad, &bad_sectors)
		)
1782
		set_bit(R1BIO_MadeGood, &r1_bio->state);
1783

L
Linus Torvalds 已提交
1784
	if (atomic_dec_and_test(&r1_bio->remaining)) {
1785
		int s = r1_bio->sectors;
1786 1787
		if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
		    test_bit(R1BIO_WriteError, &r1_bio->state))
1788 1789 1790 1791 1792
			reschedule_retry(r1_bio);
		else {
			put_buf(r1_bio);
			md_done_sync(mddev, s, uptodate);
		}
L
Linus Torvalds 已提交
1793 1794 1795
	}
}

1796
static int r1_sync_page_io(struct md_rdev *rdev, sector_t sector,
1797 1798 1799 1800 1801
			    int sectors, struct page *page, int rw)
{
	if (sync_page_io(rdev, sector, sectors << 9, page, rw, false))
		/* success */
		return 1;
1802
	if (rw == WRITE) {
1803
		set_bit(WriteErrorSeen, &rdev->flags);
1804 1805 1806 1807 1808
		if (!test_and_set_bit(WantReplacement,
				      &rdev->flags))
			set_bit(MD_RECOVERY_NEEDED, &
				rdev->mddev->recovery);
	}
1809 1810 1811 1812 1813 1814
	/* 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;
}

1815
static int fix_sync_read_error(struct r1bio *r1_bio)
L
Linus Torvalds 已提交
1816
{
1817 1818 1819 1820 1821 1822 1823
	/* 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.
1824 1825 1826
	 * 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.
1827
	 */
1828
	struct mddev *mddev = r1_bio->mddev;
1829
	struct r1conf *conf = mddev->private;
1830 1831 1832 1833 1834 1835 1836 1837 1838
	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;
1839
		struct md_rdev *rdev;
1840
		int start;
1841 1842 1843 1844 1845 1846 1847 1848 1849 1850

		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;
1851
				if (sync_page_io(rdev, sect, s<<9,
1852 1853 1854 1855 1856 1857 1858
						 bio->bi_io_vec[idx].bv_page,
						 READ, false)) {
					success = 1;
					break;
				}
			}
			d++;
1859
			if (d == conf->raid_disks * 2)
1860 1861 1862
				d = 0;
		} while (!success && d != r1_bio->read_disk);

1863
		if (!success) {
1864
			char b[BDEVNAME_SIZE];
1865 1866 1867 1868 1869 1870
			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.
			 */
1871 1872 1873 1874 1875
			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);
1876
			for (d = 0; d < conf->raid_disks * 2; d++) {
1877 1878 1879 1880 1881 1882 1883
				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) {
1884 1885
				conf->recovery_disabled =
					mddev->recovery_disabled;
1886 1887 1888 1889 1890 1891 1892 1893 1894 1895
				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;
1896
		}
1897 1898 1899 1900 1901

		start = d;
		/* write it back and re-read */
		while (d != r1_bio->read_disk) {
			if (d == 0)
1902
				d = conf->raid_disks * 2;
1903 1904 1905 1906
			d--;
			if (r1_bio->bios[d]->bi_end_io != end_sync_read)
				continue;
			rdev = conf->mirrors[d].rdev;
1907 1908 1909
			if (r1_sync_page_io(rdev, sect, s,
					    bio->bi_io_vec[idx].bv_page,
					    WRITE) == 0) {
1910 1911
				r1_bio->bios[d]->bi_end_io = NULL;
				rdev_dec_pending(rdev, mddev);
1912
			}
1913 1914 1915 1916
		}
		d = start;
		while (d != r1_bio->read_disk) {
			if (d == 0)
1917
				d = conf->raid_disks * 2;
1918 1919 1920 1921
			d--;
			if (r1_bio->bios[d]->bi_end_io != end_sync_read)
				continue;
			rdev = conf->mirrors[d].rdev;
1922 1923 1924
			if (r1_sync_page_io(rdev, sect, s,
					    bio->bi_io_vec[idx].bv_page,
					    READ) != 0)
1925
				atomic_add(s, &rdev->corrected_errors);
1926
		}
1927 1928 1929 1930
		sectors -= s;
		sect += s;
		idx ++;
	}
1931
	set_bit(R1BIO_Uptodate, &r1_bio->state);
1932
	set_bit(BIO_UPTODATE, &bio->bi_flags);
1933 1934 1935
	return 1;
}

1936
static int process_checks(struct r1bio *r1_bio)
1937 1938 1939 1940 1941 1942 1943 1944
{
	/* 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
	 */
1945
	struct mddev *mddev = r1_bio->mddev;
1946
	struct r1conf *conf = mddev->private;
1947 1948
	int primary;
	int i;
1949
	int vcnt;
1950

1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961
	/* Fix variable parts of all bios */
	vcnt = (r1_bio->sectors + PAGE_SIZE / 512 - 1) >> (PAGE_SHIFT - 9);
	for (i = 0; i < conf->raid_disks * 2; i++) {
		int j;
		int size;
		struct bio *b = r1_bio->bios[i];
		if (b->bi_end_io != end_sync_read)
			continue;
		/* fixup the bio for reuse */
		bio_reset(b);
		b->bi_vcnt = vcnt;
1962 1963
		b->bi_iter.bi_size = r1_bio->sectors << 9;
		b->bi_iter.bi_sector = r1_bio->sector +
1964 1965 1966 1967 1968
			conf->mirrors[i].rdev->data_offset;
		b->bi_bdev = conf->mirrors[i].rdev->bdev;
		b->bi_end_io = end_sync_read;
		b->bi_private = r1_bio;

1969
		size = b->bi_iter.bi_size;
1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980
		for (j = 0; j < vcnt ; j++) {
			struct bio_vec *bi;
			bi = &b->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;
		}
	}
1981
	for (primary = 0; primary < conf->raid_disks * 2; primary++)
1982 1983 1984 1985 1986 1987 1988
		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;
1989
	for (i = 0; i < conf->raid_disks * 2; i++) {
1990 1991 1992
		int j;
		struct bio *pbio = r1_bio->bios[primary];
		struct bio *sbio = r1_bio->bios[i];
1993

K
Kent Overstreet 已提交
1994
		if (sbio->bi_end_io != end_sync_read)
1995 1996 1997 1998 1999 2000 2001 2002 2003
			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),
2004
					   sbio->bi_io_vec[j].bv_len))
2005
					break;
2006
			}
2007 2008 2009
		} else
			j = 0;
		if (j >= 0)
2010
			atomic64_add(r1_bio->sectors, &mddev->resync_mismatches);
2011 2012 2013 2014 2015 2016 2017
		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;
		}
K
Kent Overstreet 已提交
2018 2019

		bio_copy_data(sbio, pbio);
2020
	}
2021 2022 2023
	return 0;
}

2024
static void sync_request_write(struct mddev *mddev, struct r1bio *r1_bio)
2025
{
2026
	struct r1conf *conf = mddev->private;
2027
	int i;
2028
	int disks = conf->raid_disks * 2;
2029 2030 2031 2032 2033 2034 2035 2036
	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;
2037 2038 2039 2040

	if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
		if (process_checks(r1_bio) < 0)
			return;
2041 2042 2043
	/*
	 * schedule writes
	 */
L
Linus Torvalds 已提交
2044 2045 2046
	atomic_set(&r1_bio->remaining, 1);
	for (i = 0; i < disks ; i++) {
		wbio = r1_bio->bios[i];
2047 2048 2049 2050
		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
2051 2052
			continue;

2053 2054
		wbio->bi_rw = WRITE;
		wbio->bi_end_io = end_sync_write;
L
Linus Torvalds 已提交
2055
		atomic_inc(&r1_bio->remaining);
2056
		md_sync_acct(conf->mirrors[i].rdev->bdev, bio_sectors(wbio));
2057

L
Linus Torvalds 已提交
2058 2059 2060 2061
		generic_make_request(wbio);
	}

	if (atomic_dec_and_test(&r1_bio->remaining)) {
2062
		/* if we're here, all write(s) have completed, so clean up */
2063 2064 2065 2066 2067 2068 2069 2070
		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
2071 2072 2073 2074 2075 2076 2077 2078
	}
}

/*
 * This is a kernel thread which:
 *
 *	1.	Retries failed read operations on working mirrors.
 *	2.	Updates the raid superblock when problems encounter.
2079
 *	3.	Performs writes following reads for array synchronising.
L
Linus Torvalds 已提交
2080 2081
 */

2082
static void fix_read_error(struct r1conf *conf, int read_disk,
2083 2084
			   sector_t sect, int sectors)
{
2085
	struct mddev *mddev = conf->mddev;
2086 2087 2088 2089 2090
	while(sectors) {
		int s = sectors;
		int d = read_disk;
		int success = 0;
		int start;
2091
		struct md_rdev *rdev;
2092 2093 2094 2095 2096 2097 2098 2099 2100 2101

		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....
			 */
2102 2103 2104
			sector_t first_bad;
			int bad_sectors;

2105 2106
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
2107 2108 2109
			    (test_bit(In_sync, &rdev->flags) ||
			     (!test_bit(Faulty, &rdev->flags) &&
			      rdev->recovery_offset >= sect + s)) &&
2110 2111
			    is_badblock(rdev, sect, s,
					&first_bad, &bad_sectors) == 0 &&
J
Jonathan Brassow 已提交
2112 2113
			    sync_page_io(rdev, sect, s<<9,
					 conf->tmppage, READ, false))
2114 2115 2116
				success = 1;
			else {
				d++;
2117
				if (d == conf->raid_disks * 2)
2118 2119 2120 2121 2122
					d = 0;
			}
		} while (!success && d != read_disk);

		if (!success) {
2123
			/* Cannot read from anywhere - mark it bad */
2124
			struct md_rdev *rdev = conf->mirrors[read_disk].rdev;
2125 2126
			if (!rdev_set_badblocks(rdev, sect, s, 0))
				md_error(mddev, rdev);
2127 2128 2129 2130 2131 2132
			break;
		}
		/* write it back and re-read */
		start = d;
		while (d != read_disk) {
			if (d==0)
2133
				d = conf->raid_disks * 2;
2134 2135 2136
			d--;
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
2137 2138 2139
			    test_bit(In_sync, &rdev->flags))
				r1_sync_page_io(rdev, sect, s,
						conf->tmppage, WRITE);
2140 2141 2142 2143 2144
		}
		d = start;
		while (d != read_disk) {
			char b[BDEVNAME_SIZE];
			if (d==0)
2145
				d = conf->raid_disks * 2;
2146 2147 2148 2149
			d--;
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
			    test_bit(In_sync, &rdev->flags)) {
2150 2151
				if (r1_sync_page_io(rdev, sect, s,
						    conf->tmppage, READ)) {
2152 2153
					atomic_add(s, &rdev->corrected_errors);
					printk(KERN_INFO
N
NeilBrown 已提交
2154
					       "md/raid1:%s: read error corrected "
2155 2156
					       "(%d sectors at %llu on %s)\n",
					       mdname(mddev), s,
2157 2158
					       (unsigned long long)(sect +
					           rdev->data_offset),
2159 2160 2161 2162 2163 2164 2165 2166 2167
					       bdevname(rdev->bdev, b));
				}
			}
		}
		sectors -= s;
		sect += s;
	}
}

2168
static int narrow_write_error(struct r1bio *r1_bio, int i)
2169
{
2170
	struct mddev *mddev = r1_bio->mddev;
2171
	struct r1conf *conf = mddev->private;
2172
	struct md_rdev *rdev = conf->mirrors[i].rdev;
2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205

	/* 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;

	while (sect_to_write) {
		struct bio *wbio;
		if (sectors > sect_to_write)
			sectors = sect_to_write;
		/* Write at 'sector' for 'sectors'*/

2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222
		if (test_bit(R1BIO_BehindIO, &r1_bio->state)) {
			unsigned vcnt = r1_bio->behind_page_count;
			struct bio_vec *vec = r1_bio->behind_bvecs;

			while (!vec->bv_page) {
				vec++;
				vcnt--;
			}

			wbio = bio_alloc_mddev(GFP_NOIO, vcnt, mddev);
			memcpy(wbio->bi_io_vec, vec, vcnt * sizeof(struct bio_vec));

			wbio->bi_vcnt = vcnt;
		} else {
			wbio = bio_clone_mddev(r1_bio->master_bio, GFP_NOIO, mddev);
		}

2223
		wbio->bi_rw = WRITE;
2224 2225
		wbio->bi_iter.bi_sector = r1_bio->sector;
		wbio->bi_iter.bi_size = r1_bio->sectors << 9;
2226

2227
		bio_trim(wbio, sector - r1_bio->sector, sectors);
2228
		wbio->bi_iter.bi_sector += rdev->data_offset;
2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243
		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;
}

2244
static void handle_sync_write_finished(struct r1conf *conf, struct r1bio *r1_bio)
2245 2246 2247
{
	int m;
	int s = r1_bio->sectors;
2248
	for (m = 0; m < conf->raid_disks * 2 ; m++) {
2249
		struct md_rdev *rdev = conf->mirrors[m].rdev;
2250 2251 2252 2253 2254
		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)) {
2255
			rdev_clear_badblocks(rdev, r1_bio->sector, s, 0);
2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266
		}
		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);
}

2267
static void handle_write_finished(struct r1conf *conf, struct r1bio *r1_bio)
2268 2269
{
	int m;
2270
	for (m = 0; m < conf->raid_disks * 2 ; m++)
2271
		if (r1_bio->bios[m] == IO_MADE_GOOD) {
2272
			struct md_rdev *rdev = conf->mirrors[m].rdev;
2273 2274
			rdev_clear_badblocks(rdev,
					     r1_bio->sector,
2275
					     r1_bio->sectors, 0);
2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295
			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);
}

2296
static void handle_read_error(struct r1conf *conf, struct r1bio *r1_bio)
2297 2298 2299
{
	int disk;
	int max_sectors;
2300
	struct mddev *mddev = conf->mddev;
2301 2302
	struct bio *bio;
	char b[BDEVNAME_SIZE];
2303
	struct md_rdev *rdev;
2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314

	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) {
2315
		freeze_array(conf, 1);
2316 2317 2318 2319 2320
		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);
2321
	rdev_dec_pending(conf->mirrors[r1_bio->read_disk].rdev, conf->mddev);
2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341

	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);
2342 2343
		bio_trim(bio, r1_bio->sector - bio->bi_iter.bi_sector,
			 max_sectors);
2344 2345 2346 2347 2348 2349 2350 2351
		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));
2352
		bio->bi_iter.bi_sector = r1_bio->sector + rdev->data_offset;
2353 2354 2355 2356 2357 2358 2359 2360
		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
2361
					       - mbio->bi_iter.bi_sector);
2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374
			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;
2375
			r1_bio->sectors = bio_sectors(mbio) - sectors_handled;
2376 2377 2378
			r1_bio->state = 0;
			set_bit(R1BIO_ReadError, &r1_bio->state);
			r1_bio->mddev = mddev;
2379 2380
			r1_bio->sector = mbio->bi_iter.bi_sector +
				sectors_handled;
2381 2382 2383 2384 2385 2386 2387

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

S
Shaohua Li 已提交
2388
static void raid1d(struct md_thread *thread)
L
Linus Torvalds 已提交
2389
{
S
Shaohua Li 已提交
2390
	struct mddev *mddev = thread->mddev;
2391
	struct r1bio *r1_bio;
L
Linus Torvalds 已提交
2392
	unsigned long flags;
2393
	struct r1conf *conf = mddev->private;
L
Linus Torvalds 已提交
2394
	struct list_head *head = &conf->retry_list;
2395
	struct blk_plug plug;
L
Linus Torvalds 已提交
2396 2397

	md_check_recovery(mddev);
2398 2399

	blk_start_plug(&plug);
L
Linus Torvalds 已提交
2400
	for (;;) {
2401

2402
		flush_pending_writes(conf);
2403

2404 2405 2406
		spin_lock_irqsave(&conf->device_lock, flags);
		if (list_empty(head)) {
			spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
2407
			break;
2408
		}
2409
		r1_bio = list_entry(head->prev, struct r1bio, retry_list);
L
Linus Torvalds 已提交
2410
		list_del(head->prev);
2411
		conf->nr_queued--;
L
Linus Torvalds 已提交
2412 2413 2414
		spin_unlock_irqrestore(&conf->device_lock, flags);

		mddev = r1_bio->mddev;
2415
		conf = mddev->private;
2416
		if (test_bit(R1BIO_IsSync, &r1_bio->state)) {
2417
			if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
2418 2419 2420
			    test_bit(R1BIO_WriteError, &r1_bio->state))
				handle_sync_write_finished(conf, r1_bio);
			else
2421
				sync_request_write(mddev, r1_bio);
2422
		} else if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
2423 2424 2425 2426 2427
			   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
2428 2429 2430 2431
			/* just a partial read to be scheduled from separate
			 * context
			 */
			generic_make_request(r1_bio->bios[r1_bio->read_disk]);
2432

N
NeilBrown 已提交
2433
		cond_resched();
2434 2435
		if (mddev->flags & ~(1<<MD_CHANGE_PENDING))
			md_check_recovery(mddev);
L
Linus Torvalds 已提交
2436
	}
2437
	blk_finish_plug(&plug);
L
Linus Torvalds 已提交
2438 2439 2440
}


2441
static int init_resync(struct r1conf *conf)
L
Linus Torvalds 已提交
2442 2443 2444 2445
{
	int buffs;

	buffs = RESYNC_WINDOW / RESYNC_BLOCK_SIZE;
2446
	BUG_ON(conf->r1buf_pool);
L
Linus Torvalds 已提交
2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464
	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.
 */

2465
static sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipped, int go_faster)
L
Linus Torvalds 已提交
2466
{
2467
	struct r1conf *conf = mddev->private;
2468
	struct r1bio *r1_bio;
L
Linus Torvalds 已提交
2469 2470
	struct bio *bio;
	sector_t max_sector, nr_sectors;
2471
	int disk = -1;
L
Linus Torvalds 已提交
2472
	int i;
2473 2474
	int wonly = -1;
	int write_targets = 0, read_targets = 0;
N
NeilBrown 已提交
2475
	sector_t sync_blocks;
2476
	int still_degraded = 0;
2477 2478
	int good_sectors = RESYNC_SECTORS;
	int min_bad = 0; /* number of sectors that are bad in all devices */
L
Linus Torvalds 已提交
2479 2480 2481

	if (!conf->r1buf_pool)
		if (init_resync(conf))
2482
			return 0;
L
Linus Torvalds 已提交
2483

A
Andre Noll 已提交
2484
	max_sector = mddev->dev_sectors;
L
Linus Torvalds 已提交
2485
	if (sector_nr >= max_sector) {
2486 2487 2488 2489 2490
		/* 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
		 */
2491 2492
		if (mddev->curr_resync < max_sector) /* aborted */
			bitmap_end_sync(mddev->bitmap, mddev->curr_resync,
2493
						&sync_blocks, 1);
2494
		else /* completed sync */
2495
			conf->fullsync = 0;
2496 2497

		bitmap_close_sync(mddev->bitmap);
L
Linus Torvalds 已提交
2498 2499 2500 2501
		close_sync(conf);
		return 0;
	}

2502 2503
	if (mddev->bitmap == NULL &&
	    mddev->recovery_cp == MaxSector &&
2504
	    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
2505 2506 2507 2508
	    conf->fullsync == 0) {
		*skipped = 1;
		return max_sector - sector_nr;
	}
2509 2510 2511
	/* before building a request, check if we can skip these blocks..
	 * This call the bitmap_start_sync doesn't actually record anything
	 */
2512
	if (!bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, 1) &&
2513
	    !conf->fullsync && !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
2514 2515 2516 2517
		/* We can skip this block, and probably several more */
		*skipped = 1;
		return sync_blocks;
	}
L
Linus Torvalds 已提交
2518
	/*
2519 2520 2521
	 * 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
2522
	 */
2523
	if (!go_faster && conf->nr_waiting)
L
Linus Torvalds 已提交
2524
		msleep_interruptible(1000);
2525

N
NeilBrown 已提交
2526
	bitmap_cond_end_sync(mddev->bitmap, sector_nr);
2527
	r1_bio = mempool_alloc(conf->r1buf_pool, GFP_NOIO);
2528 2529 2530
	raise_barrier(conf);

	conf->next_resync = sector_nr;
L
Linus Torvalds 已提交
2531

2532
	rcu_read_lock();
L
Linus Torvalds 已提交
2533
	/*
2534 2535 2536 2537 2538 2539
	 * 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
2540 2541 2542 2543
	 */

	r1_bio->mddev = mddev;
	r1_bio->sector = sector_nr;
2544
	r1_bio->state = 0;
L
Linus Torvalds 已提交
2545 2546
	set_bit(R1BIO_IsSync, &r1_bio->state);

2547
	for (i = 0; i < conf->raid_disks * 2; i++) {
2548
		struct md_rdev *rdev;
L
Linus Torvalds 已提交
2549
		bio = r1_bio->bios[i];
K
Kent Overstreet 已提交
2550
		bio_reset(bio);
L
Linus Torvalds 已提交
2551

2552 2553
		rdev = rcu_dereference(conf->mirrors[i].rdev);
		if (rdev == NULL ||
2554
		    test_bit(Faulty, &rdev->flags)) {
2555 2556
			if (i < conf->raid_disks)
				still_degraded = 1;
2557
		} else if (!test_bit(In_sync, &rdev->flags)) {
L
Linus Torvalds 已提交
2558 2559 2560
			bio->bi_rw = WRITE;
			bio->bi_end_io = end_sync_write;
			write_targets ++;
2561 2562
		} else {
			/* may need to read from here */
2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587
			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++;
2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599
			} 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++;
2600 2601
			}
		}
2602 2603
		if (bio->bi_end_io) {
			atomic_inc(&rdev->nr_pending);
2604
			bio->bi_iter.bi_sector = sector_nr + rdev->data_offset;
2605 2606 2607
			bio->bi_bdev = rdev->bdev;
			bio->bi_private = r1_bio;
		}
L
Linus Torvalds 已提交
2608
	}
2609 2610 2611 2612
	rcu_read_unlock();
	if (disk < 0)
		disk = wonly;
	r1_bio->read_disk = disk;
2613

2614 2615 2616 2617 2618
	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;
2619
		for (i = 0 ; i < conf->raid_disks * 2 ; i++)
2620
			if (r1_bio->bios[i]->bi_end_io == end_sync_write) {
2621
				struct md_rdev *rdev = conf->mirrors[i].rdev;
2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648
				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;
	}

2649 2650 2651 2652 2653
	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
2654 2655 2656
		/* There is nowhere to write, so all non-sync
		 * drives must be failed - so we are finished
		 */
2657 2658 2659 2660
		sector_t rv;
		if (min_bad > 0)
			max_sector = sector_nr + min_bad;
		rv = max_sector - sector_nr;
2661
		*skipped = 1;
L
Linus Torvalds 已提交
2662 2663 2664 2665
		put_buf(r1_bio);
		return rv;
	}

2666 2667
	if (max_sector > mddev->resync_max)
		max_sector = mddev->resync_max; /* Don't do IO beyond here */
2668 2669
	if (max_sector > sector_nr + good_sectors)
		max_sector = sector_nr + good_sectors;
L
Linus Torvalds 已提交
2670
	nr_sectors = 0;
2671
	sync_blocks = 0;
L
Linus Torvalds 已提交
2672 2673 2674 2675 2676 2677 2678
	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;
2679 2680
		if (sync_blocks == 0) {
			if (!bitmap_start_sync(mddev->bitmap, sector_nr,
2681 2682 2683
					       &sync_blocks, still_degraded) &&
			    !conf->fullsync &&
			    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
2684
				break;
2685
			BUG_ON(sync_blocks < (PAGE_SIZE>>9));
2686
			if ((len >> 9) > sync_blocks)
2687
				len = sync_blocks<<9;
2688
		}
2689

2690
		for (i = 0 ; i < conf->raid_disks * 2; i++) {
L
Linus Torvalds 已提交
2691 2692
			bio = r1_bio->bios[i];
			if (bio->bi_end_io) {
2693
				page = bio->bi_io_vec[bio->bi_vcnt].bv_page;
L
Linus Torvalds 已提交
2694 2695
				if (bio_add_page(bio, page, len, 0) == 0) {
					/* stop here */
2696
					bio->bi_io_vec[bio->bi_vcnt].bv_page = page;
L
Linus Torvalds 已提交
2697 2698 2699
					while (i > 0) {
						i--;
						bio = r1_bio->bios[i];
2700 2701
						if (bio->bi_end_io==NULL)
							continue;
L
Linus Torvalds 已提交
2702 2703
						/* remove last page from this bio */
						bio->bi_vcnt--;
2704
						bio->bi_iter.bi_size -= len;
L
Linus Torvalds 已提交
2705 2706 2707 2708 2709 2710 2711 2712
						bio->bi_flags &= ~(1<< BIO_SEG_VALID);
					}
					goto bio_full;
				}
			}
		}
		nr_sectors += len>>9;
		sector_nr += len>>9;
2713
		sync_blocks -= (len>>9);
L
Linus Torvalds 已提交
2714 2715 2716 2717
	} while (r1_bio->bios[disk]->bi_vcnt < RESYNC_PAGES);
 bio_full:
	r1_bio->sectors = nr_sectors;

2718 2719 2720 2721 2722
	/* 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);
2723
		for (i = 0; i < conf->raid_disks * 2 && read_targets; i++) {
2724 2725
			bio = r1_bio->bios[i];
			if (bio->bi_end_io == end_sync_read) {
2726
				read_targets--;
2727
				md_sync_acct(bio->bi_bdev, nr_sectors);
2728 2729 2730 2731 2732 2733
				generic_make_request(bio);
			}
		}
	} else {
		atomic_set(&r1_bio->remaining, 1);
		bio = r1_bio->bios[r1_bio->read_disk];
2734
		md_sync_acct(bio->bi_bdev, nr_sectors);
2735
		generic_make_request(bio);
L
Linus Torvalds 已提交
2736

2737
	}
L
Linus Torvalds 已提交
2738 2739 2740
	return nr_sectors;
}

2741
static sector_t raid1_size(struct mddev *mddev, sector_t sectors, int raid_disks)
2742 2743 2744 2745 2746 2747 2748
{
	if (sectors)
		return sectors;

	return mddev->dev_sectors;
}

2749
static struct r1conf *setup_conf(struct mddev *mddev)
L
Linus Torvalds 已提交
2750
{
2751
	struct r1conf *conf;
2752
	int i;
2753
	struct raid1_info *disk;
2754
	struct md_rdev *rdev;
2755
	int err = -ENOMEM;
L
Linus Torvalds 已提交
2756

2757
	conf = kzalloc(sizeof(struct r1conf), GFP_KERNEL);
L
Linus Torvalds 已提交
2758
	if (!conf)
2759
		goto abort;
L
Linus Torvalds 已提交
2760

2761
	conf->mirrors = kzalloc(sizeof(struct raid1_info)
2762
				* mddev->raid_disks * 2,
L
Linus Torvalds 已提交
2763 2764
				 GFP_KERNEL);
	if (!conf->mirrors)
2765
		goto abort;
L
Linus Torvalds 已提交
2766

2767 2768
	conf->tmppage = alloc_page(GFP_KERNEL);
	if (!conf->tmppage)
2769
		goto abort;
2770

2771
	conf->poolinfo = kzalloc(sizeof(*conf->poolinfo), GFP_KERNEL);
L
Linus Torvalds 已提交
2772
	if (!conf->poolinfo)
2773
		goto abort;
2774
	conf->poolinfo->raid_disks = mddev->raid_disks * 2;
L
Linus Torvalds 已提交
2775 2776 2777 2778
	conf->r1bio_pool = mempool_create(NR_RAID1_BIOS, r1bio_pool_alloc,
					  r1bio_pool_free,
					  conf->poolinfo);
	if (!conf->r1bio_pool)
2779 2780
		goto abort;

2781
	conf->poolinfo->mddev = mddev;
L
Linus Torvalds 已提交
2782

2783
	err = -EINVAL;
2784
	spin_lock_init(&conf->device_lock);
N
NeilBrown 已提交
2785
	rdev_for_each(rdev, mddev) {
2786
		struct request_queue *q;
2787
		int disk_idx = rdev->raid_disk;
L
Linus Torvalds 已提交
2788 2789 2790
		if (disk_idx >= mddev->raid_disks
		    || disk_idx < 0)
			continue;
2791
		if (test_bit(Replacement, &rdev->flags))
2792
			disk = conf->mirrors + mddev->raid_disks + disk_idx;
2793 2794
		else
			disk = conf->mirrors + disk_idx;
L
Linus Torvalds 已提交
2795

2796 2797
		if (disk->rdev)
			goto abort;
L
Linus Torvalds 已提交
2798
		disk->rdev = rdev;
2799 2800 2801
		q = bdev_get_queue(rdev->bdev);
		if (q->merge_bvec_fn)
			mddev->merge_check_needed = 1;
L
Linus Torvalds 已提交
2802 2803

		disk->head_position = 0;
2804
		disk->seq_start = MaxSector;
L
Linus Torvalds 已提交
2805 2806 2807 2808 2809 2810
	}
	conf->raid_disks = mddev->raid_disks;
	conf->mddev = mddev;
	INIT_LIST_HEAD(&conf->retry_list);

	spin_lock_init(&conf->resync_lock);
2811
	init_waitqueue_head(&conf->wait_barrier);
L
Linus Torvalds 已提交
2812

2813
	bio_list_init(&conf->pending_bio_list);
2814
	conf->pending_count = 0;
2815
	conf->recovery_disabled = mddev->recovery_disabled - 1;
2816

2817 2818 2819
	conf->start_next_window = MaxSector;
	conf->current_window_requests = conf->next_window_requests = 0;

2820
	err = -EIO;
2821
	for (i = 0; i < conf->raid_disks * 2; i++) {
L
Linus Torvalds 已提交
2822 2823 2824

		disk = conf->mirrors + i;

2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839
		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;
		}

2840 2841
		if (!disk->rdev ||
		    !test_bit(In_sync, &disk->rdev->flags)) {
L
Linus Torvalds 已提交
2842
			disk->head_position = 0;
2843 2844
			if (disk->rdev &&
			    (disk->rdev->saved_raid_disk < 0))
2845
				conf->fullsync = 1;
2846
		}
L
Linus Torvalds 已提交
2847
	}
2848 2849

	err = -ENOMEM;
2850
	conf->thread = md_register_thread(raid1d, mddev, "raid1");
2851 2852
	if (!conf->thread) {
		printk(KERN_ERR
N
NeilBrown 已提交
2853
		       "md/raid1:%s: couldn't allocate thread\n",
2854 2855
		       mdname(mddev));
		goto abort;
2856
	}
L
Linus Torvalds 已提交
2857

2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871
	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);
}

2872
static int stop(struct mddev *mddev);
2873
static int run(struct mddev *mddev)
2874
{
2875
	struct r1conf *conf;
2876
	int i;
2877
	struct md_rdev *rdev;
2878
	int ret;
S
Shaohua Li 已提交
2879
	bool discard_supported = false;
2880 2881

	if (mddev->level != 1) {
N
NeilBrown 已提交
2882
		printk(KERN_ERR "md/raid1:%s: raid level not set to mirroring (%d)\n",
2883 2884 2885 2886
		       mdname(mddev), mddev->level);
		return -EIO;
	}
	if (mddev->reshape_position != MaxSector) {
N
NeilBrown 已提交
2887
		printk(KERN_ERR "md/raid1:%s: reshape_position set but not supported\n",
2888 2889 2890
		       mdname(mddev));
		return -EIO;
	}
L
Linus Torvalds 已提交
2891
	/*
2892 2893 2894
	 * 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
2895
	 */
2896 2897 2898 2899
	if (mddev->private == NULL)
		conf = setup_conf(mddev);
	else
		conf = mddev->private;
L
Linus Torvalds 已提交
2900

2901 2902
	if (IS_ERR(conf))
		return PTR_ERR(conf);
L
Linus Torvalds 已提交
2903

2904
	if (mddev->queue)
2905 2906
		blk_queue_max_write_same_sectors(mddev->queue, 0);

N
NeilBrown 已提交
2907
	rdev_for_each(rdev, mddev) {
2908 2909
		if (!mddev->gendisk)
			continue;
2910 2911
		disk_stack_limits(mddev->gendisk, rdev->bdev,
				  rdev->data_offset << 9);
S
Shaohua Li 已提交
2912 2913
		if (blk_queue_discard(bdev_get_queue(rdev->bdev)))
			discard_supported = true;
L
Linus Torvalds 已提交
2914
	}
2915

2916 2917 2918 2919 2920 2921 2922 2923 2924 2925
	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;

2926
	if (mddev->recovery_cp != MaxSector)
N
NeilBrown 已提交
2927
		printk(KERN_NOTICE "md/raid1:%s: not clean"
2928 2929
		       " -- starting background reconstruction\n",
		       mdname(mddev));
L
Linus Torvalds 已提交
2930
	printk(KERN_INFO 
N
NeilBrown 已提交
2931
		"md/raid1:%s: active with %d out of %d mirrors\n",
L
Linus Torvalds 已提交
2932 2933
		mdname(mddev), mddev->raid_disks - mddev->degraded, 
		mddev->raid_disks);
2934

L
Linus Torvalds 已提交
2935 2936 2937
	/*
	 * Ok, everything is just fine now
	 */
2938 2939 2940 2941
	mddev->thread = conf->thread;
	conf->thread = NULL;
	mddev->private = conf;

2942
	md_set_array_sectors(mddev, raid1_size(mddev, 0, 0));
L
Linus Torvalds 已提交
2943

2944 2945 2946
	if (mddev->queue) {
		mddev->queue->backing_dev_info.congested_fn = raid1_congested;
		mddev->queue->backing_dev_info.congested_data = mddev;
2947
		blk_queue_merge_bvec(mddev->queue, raid1_mergeable_bvec);
S
Shaohua Li 已提交
2948 2949 2950 2951 2952 2953 2954

		if (discard_supported)
			queue_flag_set_unlocked(QUEUE_FLAG_DISCARD,
						mddev->queue);
		else
			queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD,
						  mddev->queue);
2955
	}
2956 2957 2958 2959 2960

	ret =  md_integrity_register(mddev);
	if (ret)
		stop(mddev);
	return ret;
L
Linus Torvalds 已提交
2961 2962
}

2963
static int stop(struct mddev *mddev)
L
Linus Torvalds 已提交
2964
{
2965
	struct r1conf *conf = mddev->private;
2966 2967 2968
	struct bitmap *bitmap = mddev->bitmap;

	/* wait for behind writes to complete */
2969
	if (bitmap && atomic_read(&bitmap->behind_writes) > 0) {
N
NeilBrown 已提交
2970 2971
		printk(KERN_INFO "md/raid1:%s: behind writes in progress - waiting to stop.\n",
		       mdname(mddev));
2972
		/* need to kick something here to make sure I/O goes? */
2973 2974
		wait_event(bitmap->behind_wait,
			   atomic_read(&bitmap->behind_writes) == 0);
2975
	}
L
Linus Torvalds 已提交
2976

2977 2978
	freeze_array(conf, 0);
	unfreeze_array(conf);
2979

2980
	md_unregister_thread(&mddev->thread);
L
Linus Torvalds 已提交
2981 2982
	if (conf->r1bio_pool)
		mempool_destroy(conf->r1bio_pool);
2983
	kfree(conf->mirrors);
2984
	safe_put_page(conf->tmppage);
2985
	kfree(conf->poolinfo);
L
Linus Torvalds 已提交
2986 2987 2988 2989 2990
	kfree(conf);
	mddev->private = NULL;
	return 0;
}

2991
static int raid1_resize(struct mddev *mddev, sector_t sectors)
L
Linus Torvalds 已提交
2992 2993 2994 2995 2996 2997 2998 2999
{
	/* 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.
	 */
3000 3001 3002
	sector_t newsize = raid1_size(mddev, sectors, 0);
	if (mddev->external_size &&
	    mddev->array_sectors > newsize)
D
Dan Williams 已提交
3003
		return -EINVAL;
3004 3005 3006 3007 3008 3009
	if (mddev->bitmap) {
		int ret = bitmap_resize(mddev->bitmap, newsize, 0, 0);
		if (ret)
			return ret;
	}
	md_set_array_sectors(mddev, newsize);
3010
	set_capacity(mddev->gendisk, mddev->array_sectors);
3011
	revalidate_disk(mddev->gendisk);
D
Dan Williams 已提交
3012
	if (sectors > mddev->dev_sectors &&
3013
	    mddev->recovery_cp > mddev->dev_sectors) {
A
Andre Noll 已提交
3014
		mddev->recovery_cp = mddev->dev_sectors;
L
Linus Torvalds 已提交
3015 3016
		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
	}
D
Dan Williams 已提交
3017
	mddev->dev_sectors = sectors;
3018
	mddev->resync_max_sectors = sectors;
L
Linus Torvalds 已提交
3019 3020 3021
	return 0;
}

3022
static int raid1_reshape(struct mddev *mddev)
L
Linus Torvalds 已提交
3023 3024 3025 3026 3027 3028 3029 3030
{
	/* 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.
3031 3032 3033
	 *
	 * 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
3034 3035 3036
	 */
	mempool_t *newpool, *oldpool;
	struct pool_info *newpoolinfo;
3037
	struct raid1_info *newmirrors;
3038
	struct r1conf *conf = mddev->private;
3039
	int cnt, raid_disks;
3040
	unsigned long flags;
3041
	int d, d2, err;
L
Linus Torvalds 已提交
3042

3043
	/* Cannot change chunk_size, layout, or level */
3044
	if (mddev->chunk_sectors != mddev->new_chunk_sectors ||
3045 3046
	    mddev->layout != mddev->new_layout ||
	    mddev->level != mddev->new_level) {
3047
		mddev->new_chunk_sectors = mddev->chunk_sectors;
3048 3049 3050 3051 3052
		mddev->new_layout = mddev->layout;
		mddev->new_level = mddev->level;
		return -EINVAL;
	}

3053 3054 3055
	err = md_allow_write(mddev);
	if (err)
		return err;
3056

3057 3058
	raid_disks = mddev->raid_disks + mddev->delta_disks;

3059 3060 3061 3062 3063 3064
	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
3065
			return -EBUSY;
3066
	}
L
Linus Torvalds 已提交
3067 3068 3069 3070 3071

	newpoolinfo = kmalloc(sizeof(*newpoolinfo), GFP_KERNEL);
	if (!newpoolinfo)
		return -ENOMEM;
	newpoolinfo->mddev = mddev;
3072
	newpoolinfo->raid_disks = raid_disks * 2;
L
Linus Torvalds 已提交
3073 3074 3075 3076 3077 3078 3079

	newpool = mempool_create(NR_RAID1_BIOS, r1bio_pool_alloc,
				 r1bio_pool_free, newpoolinfo);
	if (!newpool) {
		kfree(newpoolinfo);
		return -ENOMEM;
	}
3080
	newmirrors = kzalloc(sizeof(struct raid1_info) * raid_disks * 2,
3081
			     GFP_KERNEL);
L
Linus Torvalds 已提交
3082 3083 3084 3085 3086 3087
	if (!newmirrors) {
		kfree(newpoolinfo);
		mempool_destroy(newpool);
		return -ENOMEM;
	}

3088
	freeze_array(conf, 0);
L
Linus Torvalds 已提交
3089 3090 3091 3092

	/* ok, everything is stopped */
	oldpool = conf->r1bio_pool;
	conf->r1bio_pool = newpool;
3093

3094
	for (d = d2 = 0; d < conf->raid_disks; d++) {
3095
		struct md_rdev *rdev = conf->mirrors[d].rdev;
3096
		if (rdev && rdev->raid_disk != d2) {
3097
			sysfs_unlink_rdev(mddev, rdev);
3098
			rdev->raid_disk = d2;
3099 3100
			sysfs_unlink_rdev(mddev, rdev);
			if (sysfs_link_rdev(mddev, rdev))
3101
				printk(KERN_WARNING
3102 3103
				       "md/raid1:%s: cannot register rd%d\n",
				       mdname(mddev), rdev->raid_disk);
3104
		}
3105 3106 3107
		if (rdev)
			newmirrors[d2++].rdev = rdev;
	}
L
Linus Torvalds 已提交
3108 3109 3110 3111 3112
	kfree(conf->mirrors);
	conf->mirrors = newmirrors;
	kfree(conf->poolinfo);
	conf->poolinfo = newpoolinfo;

3113
	spin_lock_irqsave(&conf->device_lock, flags);
L
Linus Torvalds 已提交
3114
	mddev->degraded += (raid_disks - conf->raid_disks);
3115
	spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
3116
	conf->raid_disks = mddev->raid_disks = raid_disks;
3117
	mddev->delta_disks = 0;
L
Linus Torvalds 已提交
3118

3119
	unfreeze_array(conf);
L
Linus Torvalds 已提交
3120 3121 3122 3123 3124 3125 3126 3127

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

	mempool_destroy(oldpool);
	return 0;
}

3128
static void raid1_quiesce(struct mddev *mddev, int state)
3129
{
3130
	struct r1conf *conf = mddev->private;
3131 3132

	switch(state) {
3133 3134 3135
	case 2: /* wake for suspend */
		wake_up(&conf->wait_barrier);
		break;
3136
	case 1:
3137
		freeze_array(conf, 0);
3138
		break;
3139
	case 0:
3140
		unfreeze_array(conf);
3141 3142 3143 3144
		break;
	}
}

3145
static void *raid1_takeover(struct mddev *mddev)
3146 3147 3148 3149 3150
{
	/* raid1 can take over:
	 *  raid5 with 2 devices, any layout or chunk size
	 */
	if (mddev->level == 5 && mddev->raid_disks == 2) {
3151
		struct r1conf *conf;
3152 3153 3154 3155 3156
		mddev->new_level = 1;
		mddev->new_layout = 0;
		mddev->new_chunk_sectors = 0;
		conf = setup_conf(mddev);
		if (!IS_ERR(conf))
3157 3158
			/* Array must appear to be quiesced */
			conf->array_frozen = 1;
3159 3160 3161 3162
		return conf;
	}
	return ERR_PTR(-EINVAL);
}
L
Linus Torvalds 已提交
3163

3164
static struct md_personality raid1_personality =
L
Linus Torvalds 已提交
3165 3166
{
	.name		= "raid1",
3167
	.level		= 1,
L
Linus Torvalds 已提交
3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178
	.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,
3179
	.size		= raid1_size,
3180
	.check_reshape	= raid1_reshape,
3181
	.quiesce	= raid1_quiesce,
3182
	.takeover	= raid1_takeover,
L
Linus Torvalds 已提交
3183 3184 3185 3186
};

static int __init raid_init(void)
{
3187
	return register_md_personality(&raid1_personality);
L
Linus Torvalds 已提交
3188 3189 3190 3191
}

static void raid_exit(void)
{
3192
	unregister_md_personality(&raid1_personality);
L
Linus Torvalds 已提交
3193 3194 3195 3196 3197
}

module_init(raid_init);
module_exit(raid_exit);
MODULE_LICENSE("GPL");
3198
MODULE_DESCRIPTION("RAID1 (mirroring) personality for MD");
L
Linus Torvalds 已提交
3199
MODULE_ALIAS("md-personality-3"); /* RAID1 */
3200
MODULE_ALIAS("md-raid1");
3201
MODULE_ALIAS("md-level-1");
3202 3203

module_param(max_queued_requests, int, S_IRUGO|S_IWUSR);