raid1.c 86.8 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
	struct bio *bio;
100
	int need_pages;
L
Linus Torvalds 已提交
101 102 103
	int i, j;

	r1_bio = r1bio_pool_alloc(gfp_flags, pi);
J
Jens Axboe 已提交
104
	if (!r1_bio)
L
Linus Torvalds 已提交
105 106 107 108 109 110
		return NULL;

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

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

	r1_bio->master_bio = NULL;

	return r1_bio;

145 146 147 148 149 150 151 152
out_free_pages:
	while (--j >= 0) {
		struct bio_vec *bv;

		bio_for_each_segment_all(bv, r1_bio->bios[j], i)
			__free_page(bv->bv_page);
	}

L
Linus Torvalds 已提交
153
out_free_bio:
154
	while (++j < pi->raid_disks)
L
Linus Torvalds 已提交
155 156 157 158 159 160 161 162
		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;
163
	int i,j;
164
	struct r1bio *r1bio = __r1_bio;
L
Linus Torvalds 已提交
165

166 167 168 169 170
	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)
171
				safe_put_page(r1bio->bios[j]->bi_io_vec[i].bv_page);
172
		}
L
Linus Torvalds 已提交
173 174 175 176 177 178
	for (i=0 ; i < pi->raid_disks; i++)
		bio_put(r1bio->bios[i]);

	r1bio_pool_free(r1bio, data);
}

179
static void put_all_bios(struct r1conf *conf, struct r1bio *r1_bio)
L
Linus Torvalds 已提交
180 181 182
{
	int i;

183
	for (i = 0; i < conf->raid_disks * 2; i++) {
L
Linus Torvalds 已提交
184
		struct bio **bio = r1_bio->bios + i;
185
		if (!BIO_SPECIAL(*bio))
L
Linus Torvalds 已提交
186 187 188 189 190
			bio_put(*bio);
		*bio = NULL;
	}
}

191
static void free_r1bio(struct r1bio *r1_bio)
L
Linus Torvalds 已提交
192
{
193
	struct r1conf *conf = r1_bio->mddev->private;
L
Linus Torvalds 已提交
194 195 196 197 198

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

199
static void put_buf(struct r1bio *r1_bio)
L
Linus Torvalds 已提交
200
{
201
	struct r1conf *conf = r1_bio->mddev->private;
202 203
	int i;

204
	for (i = 0; i < conf->raid_disks * 2; i++) {
205 206 207 208
		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
209 210 211

	mempool_free(r1_bio, conf->r1buf_pool);

212
	lower_barrier(conf);
L
Linus Torvalds 已提交
213 214
}

215
static void reschedule_retry(struct r1bio *r1_bio)
L
Linus Torvalds 已提交
216 217
{
	unsigned long flags;
218
	struct mddev *mddev = r1_bio->mddev;
219
	struct r1conf *conf = mddev->private;
L
Linus Torvalds 已提交
220 221 222

	spin_lock_irqsave(&conf->device_lock, flags);
	list_add(&r1_bio->retry_list, &conf->retry_list);
223
	conf->nr_queued ++;
L
Linus Torvalds 已提交
224 225
	spin_unlock_irqrestore(&conf->device_lock, flags);

226
	wake_up(&conf->wait_barrier);
L
Linus Torvalds 已提交
227 228 229 230 231 232 233 234
	md_wakeup_thread(mddev->thread);
}

/*
 * raid_end_bio_io() is called when we have finished servicing a mirrored
 * operation and are ready to return a success/failure code to the buffer
 * cache layer.
 */
235
static void call_bio_endio(struct r1bio *r1_bio)
236 237 238
{
	struct bio *bio = r1_bio->master_bio;
	int done;
239
	struct r1conf *conf = r1_bio->mddev->private;
240
	sector_t start_next_window = r1_bio->start_next_window;
241
	sector_t bi_sector = bio->bi_iter.bi_sector;
242 243 244 245 246 247 248

	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);
249 250 251 252 253
		/*
		 * make_request() might be waiting for
		 * bi_phys_segments to decrease
		 */
		wake_up(&conf->wait_barrier);
254 255 256 257 258 259 260 261 262 263 264
	} 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.
		 */
265
		allow_barrier(conf, start_next_window, bi_sector);
266 267 268
	}
}

269
static void raid_end_bio_io(struct r1bio *r1_bio)
L
Linus Torvalds 已提交
270 271 272
{
	struct bio *bio = r1_bio->master_bio;

273 274
	/* if nobody has done the final endio yet, do it now */
	if (!test_and_set_bit(R1BIO_Returned, &r1_bio->state)) {
275 276
		pr_debug("raid1: sync end %s on sectors %llu-%llu\n",
			 (bio_data_dir(bio) == WRITE) ? "write" : "read",
277 278
			 (unsigned long long) bio->bi_iter.bi_sector,
			 (unsigned long long) bio_end_sector(bio) - 1);
279

280
		call_bio_endio(r1_bio);
281
	}
L
Linus Torvalds 已提交
282 283 284 285 286 287
	free_r1bio(r1_bio);
}

/*
 * Update disk head position estimator based on IRQ completion info.
 */
288
static inline void update_head_pos(int disk, struct r1bio *r1_bio)
L
Linus Torvalds 已提交
289
{
290
	struct r1conf *conf = r1_bio->mddev->private;
L
Linus Torvalds 已提交
291 292 293 294 295

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

296 297 298
/*
 * Find the disk number which triggered given bio
 */
299
static int find_bio_disk(struct r1bio *r1_bio, struct bio *bio)
300 301
{
	int mirror;
302 303
	struct r1conf *conf = r1_bio->mddev->private;
	int raid_disks = conf->raid_disks;
304

305
	for (mirror = 0; mirror < raid_disks * 2; mirror++)
306 307 308
		if (r1_bio->bios[mirror] == bio)
			break;

309
	BUG_ON(mirror == raid_disks * 2);
310 311 312 313 314
	update_head_pos(mirror, r1_bio);

	return mirror;
}

315
static void raid1_end_read_request(struct bio *bio, int error)
L
Linus Torvalds 已提交
316 317
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
318
	struct r1bio *r1_bio = bio->bi_private;
L
Linus Torvalds 已提交
319
	int mirror;
320
	struct r1conf *conf = r1_bio->mddev->private;
L
Linus Torvalds 已提交
321 322 323 324 325

	mirror = r1_bio->read_disk;
	/*
	 * this branch is our 'one mirror IO has finished' event handler:
	 */
326 327
	update_head_pos(mirror, r1_bio);

328 329 330 331 332 333
	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
334
		 */
335 336 337 338 339 340 341 342
		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
343

344
	if (uptodate) {
L
Linus Torvalds 已提交
345
		raid_end_bio_io(r1_bio);
346 347
		rdev_dec_pending(conf->mirrors[mirror].rdev, conf->mddev);
	} else {
L
Linus Torvalds 已提交
348 349 350 351
		/*
		 * oops, read error:
		 */
		char b[BDEVNAME_SIZE];
352 353 354 355 356 357 358
		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);
359
		set_bit(R1BIO_ReadError, &r1_bio->state);
L
Linus Torvalds 已提交
360
		reschedule_retry(r1_bio);
361
		/* don't drop the reference on read_disk yet */
L
Linus Torvalds 已提交
362 363 364
	}
}

365
static void close_write(struct r1bio *r1_bio)
366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383
{
	/* 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);
}

384
static void r1_bio_write_done(struct r1bio *r1_bio)
385
{
386 387 388 389 390 391 392
	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);
393 394 395 396
		if (test_bit(R1BIO_MadeGood, &r1_bio->state))
			reschedule_retry(r1_bio);
		else
			raid_end_bio_io(r1_bio);
397 398 399
	}
}

400
static void raid1_end_write_request(struct bio *bio, int error)
L
Linus Torvalds 已提交
401 402
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
403
	struct r1bio *r1_bio = bio->bi_private;
404
	int mirror, behind = test_bit(R1BIO_BehindIO, &r1_bio->state);
405
	struct r1conf *conf = r1_bio->mddev->private;
406
	struct bio *to_put = NULL;
L
Linus Torvalds 已提交
407

408
	mirror = find_bio_disk(r1_bio, bio);
L
Linus Torvalds 已提交
409

T
Tejun Heo 已提交
410 411 412 413
	/*
	 * 'one mirror IO has finished' event handler:
	 */
	if (!uptodate) {
414 415
		set_bit(WriteErrorSeen,
			&conf->mirrors[mirror].rdev->flags);
416 417 418 419 420
		if (!test_and_set_bit(WantReplacement,
				      &conf->mirrors[mirror].rdev->flags))
			set_bit(MD_RECOVERY_NEEDED, &
				conf->mddev->recovery);

421
		set_bit(R1BIO_WriteError, &r1_bio->state);
422
	} else {
L
Linus Torvalds 已提交
423
		/*
T
Tejun Heo 已提交
424 425 426 427 428 429 430 431
		 * 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
432
		 */
433 434 435
		sector_t first_bad;
		int bad_sectors;

436 437
		r1_bio->bios[mirror] = NULL;
		to_put = bio;
438 439 440 441 442 443 444 445 446 447 448
		/*
		 * 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
449

450 451 452 453 454 455 456 457 458
		/* 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
459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474
	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;
475 476
				pr_debug("raid1: behind end write sectors"
					 " %llu-%llu\n",
477 478
					 (unsigned long long) mbio->bi_iter.bi_sector,
					 (unsigned long long) bio_end_sector(mbio) - 1);
479
				call_bio_endio(r1_bio);
480 481 482
			}
		}
	}
483 484 485
	if (r1_bio->bios[mirror] == NULL)
		rdev_dec_pending(conf->mirrors[mirror].rdev,
				 conf->mddev);
T
Tejun Heo 已提交
486

L
Linus Torvalds 已提交
487 488 489 490
	/*
	 * Let's see if all mirrored write operations have finished
	 * already.
	 */
491
	r1_bio_write_done(r1_bio);
492

493 494
	if (to_put)
		bio_put(to_put);
L
Linus Torvalds 已提交
495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510
}

/*
 * 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.
 */
511
static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sectors)
L
Linus Torvalds 已提交
512
{
513
	const sector_t this_sector = r1_bio->sector;
514 515
	int sectors;
	int best_good_sectors;
516 517
	int best_disk, best_dist_disk, best_pending_disk;
	int has_nonrot_disk;
518
	int disk;
N
NeilBrown 已提交
519
	sector_t best_dist;
520
	unsigned int min_pending;
521
	struct md_rdev *rdev;
522
	int choose_first;
523
	int choose_next_idle;
L
Linus Torvalds 已提交
524 525 526

	rcu_read_lock();
	/*
527
	 * Check if we can balance. We can balance on the whole
L
Linus Torvalds 已提交
528 529 530 531
	 * device if no resync is going on, or below the resync window.
	 * We take the first readable disk when above the resync window.
	 */
 retry:
532
	sectors = r1_bio->sectors;
N
NeilBrown 已提交
533
	best_disk = -1;
534
	best_dist_disk = -1;
N
NeilBrown 已提交
535
	best_dist = MaxSector;
536 537
	best_pending_disk = -1;
	min_pending = UINT_MAX;
538
	best_good_sectors = 0;
539
	has_nonrot_disk = 0;
540
	choose_next_idle = 0;
541

542 543 544 545 546 547 548
	if ((conf->mddev->recovery_cp < this_sector + sectors) ||
	    (mddev_is_clustered(conf->mddev) &&
	    md_cluster_ops->area_resyncing(conf->mddev, this_sector,
		    this_sector + sectors)))
		choose_first = 1;
	else
		choose_first = 0;
L
Linus Torvalds 已提交
549

550
	for (disk = 0 ; disk < conf->raid_disks * 2 ; disk++) {
N
NeilBrown 已提交
551
		sector_t dist;
552 553
		sector_t first_bad;
		int bad_sectors;
554
		unsigned int pending;
555
		bool nonrot;
556

557 558 559
		rdev = rcu_dereference(conf->mirrors[disk].rdev);
		if (r1_bio->bios[disk] == IO_BLOCKED
		    || rdev == NULL
560
		    || test_bit(Unmerged, &rdev->flags)
N
NeilBrown 已提交
561
		    || test_bit(Faulty, &rdev->flags))
562
			continue;
N
NeilBrown 已提交
563 564
		if (!test_bit(In_sync, &rdev->flags) &&
		    rdev->recovery_offset < this_sector + sectors)
L
Linus Torvalds 已提交
565
			continue;
N
NeilBrown 已提交
566 567 568
		if (test_bit(WriteMostly, &rdev->flags)) {
			/* Don't balance among write-mostly, just
			 * use the first as a last resort */
569
			if (best_dist_disk < 0) {
570 571 572 573 574 575 576 577
				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;
578 579
				best_dist_disk = disk;
				best_pending_disk = disk;
580
			}
N
NeilBrown 已提交
581 582 583 584 585
			continue;
		}
		/* This is a reasonable device to use.  It might
		 * even be best.
		 */
586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614
		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;

615 616
		nonrot = blk_queue_nonrot(bdev_get_queue(rdev->bdev));
		has_nonrot_disk |= nonrot;
617
		pending = atomic_read(&rdev->nr_pending);
N
NeilBrown 已提交
618
		dist = abs(this_sector - conf->mirrors[disk].head_position);
619
		if (choose_first) {
N
NeilBrown 已提交
620
			best_disk = disk;
L
Linus Torvalds 已提交
621 622
			break;
		}
623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660
		/* 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;
661 662 663 664 665 666

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

N
NeilBrown 已提交
667 668
		if (dist < best_dist) {
			best_dist = dist;
669
			best_dist_disk = disk;
L
Linus Torvalds 已提交
670
		}
671
	}
L
Linus Torvalds 已提交
672

673 674 675 676 677 678 679 680 681 682 683 684 685
	/*
	 * 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
686 687
	if (best_disk >= 0) {
		rdev = rcu_dereference(conf->mirrors[best_disk].rdev);
688 689 690
		if (!rdev)
			goto retry;
		atomic_inc(&rdev->nr_pending);
N
NeilBrown 已提交
691
		if (test_bit(Faulty, &rdev->flags)) {
L
Linus Torvalds 已提交
692 693 694
			/* cannot risk returning a device that failed
			 * before we inc'ed nr_pending
			 */
695
			rdev_dec_pending(rdev, conf->mddev);
L
Linus Torvalds 已提交
696 697
			goto retry;
		}
698
		sectors = best_good_sectors;
699 700 701 702

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

703
		conf->mirrors[best_disk].next_seq_sect = this_sector + sectors;
L
Linus Torvalds 已提交
704 705
	}
	rcu_read_unlock();
706
	*max_sectors = sectors;
L
Linus Torvalds 已提交
707

N
NeilBrown 已提交
708
	return best_disk;
L
Linus Torvalds 已提交
709 710
}

711
static int raid1_mergeable_bvec(struct mddev *mddev,
712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742
				struct bvec_merge_data *bvm,
				struct bio_vec *biovec)
{
	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;

}

743
static int raid1_congested(struct mddev *mddev, int bits)
744
{
745
	struct r1conf *conf = mddev->private;
746 747
	int i, ret = 0;

748 749 750 751
	if ((bits & (1 << BDI_async_congested)) &&
	    conf->pending_count >= max_queued_requests)
		return 1;

752
	rcu_read_lock();
753
	for (i = 0; i < conf->raid_disks * 2; i++) {
754
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
755
		if (rdev && !test_bit(Faulty, &rdev->flags)) {
756
			struct request_queue *q = bdev_get_queue(rdev->bdev);
757

758 759
			BUG_ON(!q);

760 761 762
			/* Note the '|| 1' - when read_balance prefers
			 * non-congested targets, it can be removed
			 */
763
			if ((bits & (1<<BDI_async_congested)) || 1)
764 765 766 767 768 769 770 771 772
				ret |= bdi_congested(&q->backing_dev_info, bits);
			else
				ret &= bdi_congested(&q->backing_dev_info, bits);
		}
	}
	rcu_read_unlock();
	return ret;
}

773
static void flush_pending_writes(struct r1conf *conf)
774 775 776 777 778 779 780 781 782
{
	/* 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);
783
		conf->pending_count = 0;
784 785 786 787
		spin_unlock_irq(&conf->device_lock);
		/* flush any pending bitmap writes to
		 * disk before proceeding w/ I/O */
		bitmap_unplug(conf->mddev->bitmap);
788
		wake_up(&conf->wait_barrier);
789 790 791 792

		while (bio) { /* submit pending writes */
			struct bio *next = bio->bi_next;
			bio->bi_next = NULL;
S
Shaohua Li 已提交
793 794 795 796 797 798
			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);
799 800 801 802
			bio = next;
		}
	} else
		spin_unlock_irq(&conf->device_lock);
J
Jens Axboe 已提交
803 804
}

805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824
/* 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
825
 */
826
static void raise_barrier(struct r1conf *conf, sector_t sector_nr)
L
Linus Torvalds 已提交
827 828
{
	spin_lock_irq(&conf->resync_lock);
829 830 831

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

	/* block any new IO from starting */
	conf->barrier++;
836
	conf->next_resync = sector_nr;
837

838 839 840 841 842 843 844
	/* 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.
845
	 * D: while there are any active requests in the current window.
846
	 */
847
	wait_event_lock_irq(conf->wait_barrier,
848
			    !conf->array_frozen &&
849
			    conf->barrier < RESYNC_DEPTH &&
850
			    conf->current_window_requests == 0 &&
851 852
			    (conf->start_next_window >=
			     conf->next_resync + RESYNC_SECTORS),
853
			    conf->resync_lock);
854

855
	conf->nr_pending++;
856 857 858
	spin_unlock_irq(&conf->resync_lock);
}

859
static void lower_barrier(struct r1conf *conf)
860 861
{
	unsigned long flags;
862
	BUG_ON(conf->barrier <= 0);
863 864
	spin_lock_irqsave(&conf->resync_lock, flags);
	conf->barrier--;
865
	conf->nr_pending--;
866 867 868 869
	spin_unlock_irqrestore(&conf->resync_lock, flags);
	wake_up(&conf->wait_barrier);
}

870
static bool need_to_wait_for_sync(struct r1conf *conf, struct bio *bio)
871
{
872 873 874 875 876
	bool wait = false;

	if (conf->array_frozen || !bio)
		wait = true;
	else if (conf->barrier && bio_data_dir(bio) == WRITE) {
877 878 879 880
		if ((conf->mddev->curr_resync_completed
		     >= bio_end_sector(bio)) ||
		    (conf->next_resync + NEXT_NORMALIO_DISTANCE
		     <= bio->bi_iter.bi_sector))
881 882 883 884 885 886 887 888 889 890 891 892
			wait = false;
		else
			wait = true;
	}

	return wait;
}

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

893
	spin_lock_irq(&conf->resync_lock);
894
	if (need_to_wait_for_sync(conf, bio)) {
895
		conf->nr_waiting++;
896 897 898 899
		/* Wait for the barrier to drop.
		 * However if there are already pending
		 * requests (preventing the barrier from
		 * rising completely), and the
900
		 * per-process bio queue isn't empty,
901
		 * then don't wait, as we need to empty
902 903
		 * that queue to allow conf->start_next_window
		 * to increase.
904 905
		 */
		wait_event_lock_irq(conf->wait_barrier,
906 907
				    !conf->array_frozen &&
				    (!conf->barrier ||
908 909 910 911
				     ((conf->start_next_window <
				       conf->next_resync + RESYNC_SECTORS) &&
				      current->bio_list &&
				      !bio_list_empty(current->bio_list))),
912
				    conf->resync_lock);
913
		conf->nr_waiting--;
L
Linus Torvalds 已提交
914
	}
915 916

	if (bio && bio_data_dir(bio) == WRITE) {
917
		if (bio->bi_iter.bi_sector >=
918
		    conf->mddev->curr_resync_completed) {
919 920 921 922 923 924
			if (conf->start_next_window == MaxSector)
				conf->start_next_window =
					conf->next_resync +
					NEXT_NORMALIO_DISTANCE;

			if ((conf->start_next_window + NEXT_NORMALIO_DISTANCE)
925
			    <= bio->bi_iter.bi_sector)
926 927 928 929
				conf->next_window_requests++;
			else
				conf->current_window_requests++;
			sector = conf->start_next_window;
930
		}
931 932
	}

933
	conf->nr_pending++;
L
Linus Torvalds 已提交
934
	spin_unlock_irq(&conf->resync_lock);
935
	return sector;
L
Linus Torvalds 已提交
936 937
}

938 939
static void allow_barrier(struct r1conf *conf, sector_t start_next_window,
			  sector_t bi_sector)
940 941
{
	unsigned long flags;
942

943 944
	spin_lock_irqsave(&conf->resync_lock, flags);
	conf->nr_pending--;
945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965
	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;
		}
	}
966 967 968 969
	spin_unlock_irqrestore(&conf->resync_lock, flags);
	wake_up(&conf->wait_barrier);
}

970
static void freeze_array(struct r1conf *conf, int extra)
971 972 973
{
	/* stop syncio and normal IO and wait for everything to
	 * go quite.
974
	 * We wait until nr_pending match nr_queued+extra
975 976 977 978
	 * 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.
979
	 * Thus the number queued (nr_queued) plus this request (extra)
980 981
	 * must match the number of pending IOs (nr_pending) before
	 * we continue.
982 983
	 */
	spin_lock_irq(&conf->resync_lock);
984
	conf->array_frozen = 1;
985
	wait_event_lock_irq_cmd(conf->wait_barrier,
986
				conf->nr_pending == conf->nr_queued+extra,
987 988
				conf->resync_lock,
				flush_pending_writes(conf));
989 990
	spin_unlock_irq(&conf->resync_lock);
}
991
static void unfreeze_array(struct r1conf *conf)
992 993 994
{
	/* reverse the effect of the freeze */
	spin_lock_irq(&conf->resync_lock);
995
	conf->array_frozen = 0;
996 997 998 999
	wake_up(&conf->wait_barrier);
	spin_unlock_irq(&conf->resync_lock);
}

1000
/* duplicate the data pages for behind I/O
1001
 */
1002
static void alloc_behind_pages(struct bio *bio, struct r1bio *r1_bio)
1003 1004 1005
{
	int i;
	struct bio_vec *bvec;
1006
	struct bio_vec *bvecs = kzalloc(bio->bi_vcnt * sizeof(struct bio_vec),
1007
					GFP_NOIO);
1008
	if (unlikely(!bvecs))
1009
		return;
1010

1011
	bio_for_each_segment_all(bvec, bio, i) {
1012 1013 1014
		bvecs[i] = *bvec;
		bvecs[i].bv_page = alloc_page(GFP_NOIO);
		if (unlikely(!bvecs[i].bv_page))
1015
			goto do_sync_io;
1016 1017 1018
		memcpy(kmap(bvecs[i].bv_page) + bvec->bv_offset,
		       kmap(bvec->bv_page) + bvec->bv_offset, bvec->bv_len);
		kunmap(bvecs[i].bv_page);
1019 1020
		kunmap(bvec->bv_page);
	}
1021
	r1_bio->behind_bvecs = bvecs;
1022 1023 1024
	r1_bio->behind_page_count = bio->bi_vcnt;
	set_bit(R1BIO_BehindIO, &r1_bio->state);
	return;
1025 1026

do_sync_io:
1027
	for (i = 0; i < bio->bi_vcnt; i++)
1028 1029 1030
		if (bvecs[i].bv_page)
			put_page(bvecs[i].bv_page);
	kfree(bvecs);
1031 1032
	pr_debug("%dB behind alloc failed, doing sync I/O\n",
		 bio->bi_iter.bi_size);
1033 1034
}

1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048
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;

1049
	if (from_schedule || current->bio_list) {
1050 1051 1052 1053
		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);
1054
		wake_up(&conf->wait_barrier);
1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067
		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;
1068 1069 1070 1071 1072 1073
		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);
1074 1075 1076 1077 1078
		bio = next;
	}
	kfree(plug);
}

1079
static void make_request(struct mddev *mddev, struct bio * bio)
L
Linus Torvalds 已提交
1080
{
1081
	struct r1conf *conf = mddev->private;
1082
	struct raid1_info *mirror;
1083
	struct r1bio *r1_bio;
L
Linus Torvalds 已提交
1084
	struct bio *read_bio;
1085
	int i, disks;
1086
	struct bitmap *bitmap;
1087
	unsigned long flags;
1088
	const int rw = bio_data_dir(bio);
1089
	const unsigned long do_sync = (bio->bi_rw & REQ_SYNC);
T
Tejun Heo 已提交
1090
	const unsigned long do_flush_fua = (bio->bi_rw & (REQ_FLUSH | REQ_FUA));
S
Shaohua Li 已提交
1091 1092
	const unsigned long do_discard = (bio->bi_rw
					  & (REQ_DISCARD | REQ_SECURE));
1093
	const unsigned long do_same = (bio->bi_rw & REQ_WRITE_SAME);
1094
	struct md_rdev *blocked_rdev;
1095 1096
	struct blk_plug_cb *cb;
	struct raid1_plug_cb *plug = NULL;
1097 1098 1099
	int first_clone;
	int sectors_handled;
	int max_sectors;
1100
	sector_t start_next_window;
1101

L
Linus Torvalds 已提交
1102 1103 1104 1105 1106
	/*
	 * 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.
	 */
1107

1108 1109
	md_write_start(mddev, bio); /* wait on superblock update early */

1110
	if (bio_data_dir(bio) == WRITE &&
1111 1112 1113 1114
	    ((bio_end_sector(bio) > mddev->suspend_lo &&
	    bio->bi_iter.bi_sector < mddev->suspend_hi) ||
	    (mddev_is_clustered(mddev) &&
	     md_cluster_ops->area_resyncing(mddev, bio->bi_iter.bi_sector, bio_end_sector(bio))))) {
1115 1116 1117 1118 1119 1120 1121 1122 1123
		/* 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
1124
			if (bio_end_sector(bio) <= mddev->suspend_lo ||
1125 1126 1127 1128
			    bio->bi_iter.bi_sector >= mddev->suspend_hi ||
			    (mddev_is_clustered(mddev) &&
			     !md_cluster_ops->area_resyncing(mddev,
				     bio->bi_iter.bi_sector, bio_end_sector(bio))))
1129 1130 1131 1132 1133
				break;
			schedule();
		}
		finish_wait(&conf->wait_barrier, &w);
	}
1134

1135
	start_next_window = wait_barrier(conf, bio);
L
Linus Torvalds 已提交
1136

1137 1138
	bitmap = mddev->bitmap;

L
Linus Torvalds 已提交
1139 1140 1141 1142 1143 1144 1145 1146
	/*
	 * 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;
1147
	r1_bio->sectors = bio_sectors(bio);
1148
	r1_bio->state = 0;
L
Linus Torvalds 已提交
1149
	r1_bio->mddev = mddev;
1150
	r1_bio->sector = bio->bi_iter.bi_sector;
L
Linus Torvalds 已提交
1151

1152 1153 1154 1155 1156 1157 1158 1159 1160 1161
	/* 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);

1162
	if (rw == READ) {
L
Linus Torvalds 已提交
1163 1164 1165
		/*
		 * read balancing logic:
		 */
1166 1167 1168 1169
		int rdisk;

read_again:
		rdisk = read_balance(conf, r1_bio, &max_sectors);
L
Linus Torvalds 已提交
1170 1171 1172 1173

		if (rdisk < 0) {
			/* couldn't find anywhere to read from */
			raid_end_bio_io(r1_bio);
1174
			return;
L
Linus Torvalds 已提交
1175 1176 1177
		}
		mirror = conf->mirrors + rdisk;

1178 1179 1180 1181 1182 1183 1184 1185 1186
		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
1187
		r1_bio->read_disk = rdisk;
1188
		r1_bio->start_next_window = 0;
L
Linus Torvalds 已提交
1189

1190
		read_bio = bio_clone_mddev(bio, GFP_NOIO, mddev);
1191
		bio_trim(read_bio, r1_bio->sector - bio->bi_iter.bi_sector,
1192
			 max_sectors);
L
Linus Torvalds 已提交
1193 1194 1195

		r1_bio->bios[rdisk] = read_bio;

1196 1197
		read_bio->bi_iter.bi_sector = r1_bio->sector +
			mirror->rdev->data_offset;
L
Linus Torvalds 已提交
1198 1199
		read_bio->bi_bdev = mirror->rdev->bdev;
		read_bio->bi_end_io = raid1_end_read_request;
1200
		read_bio->bi_rw = READ | do_sync;
L
Linus Torvalds 已提交
1201 1202
		read_bio->bi_private = r1_bio;

1203 1204 1205 1206 1207 1208
		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
1209
					   - bio->bi_iter.bi_sector);
1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226
			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;
1227
			r1_bio->sectors = bio_sectors(bio) - sectors_handled;
1228 1229
			r1_bio->state = 0;
			r1_bio->mddev = mddev;
1230 1231
			r1_bio->sector = bio->bi_iter.bi_sector +
				sectors_handled;
1232 1233 1234
			goto read_again;
		} else
			generic_make_request(read_bio);
1235
		return;
L
Linus Torvalds 已提交
1236 1237 1238 1239 1240
	}

	/*
	 * WRITE:
	 */
1241 1242 1243 1244 1245
	if (conf->pending_count >= max_queued_requests) {
		md_wakeup_thread(mddev->thread);
		wait_event(conf->wait_barrier,
			   conf->pending_count < max_queued_requests);
	}
1246
	/* first select target devices under rcu_lock and
L
Linus Torvalds 已提交
1247 1248
	 * inc refcount on their rdev.  Record them by setting
	 * bios[x] to bio
1249 1250 1251 1252 1253 1254
	 * 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
1255
	 */
N
NeilBrown 已提交
1256

1257
	disks = conf->raid_disks * 2;
1258
 retry_write:
1259
	r1_bio->start_next_window = start_next_window;
1260
	blocked_rdev = NULL;
L
Linus Torvalds 已提交
1261
	rcu_read_lock();
1262
	max_sectors = r1_bio->sectors;
L
Linus Torvalds 已提交
1263
	for (i = 0;  i < disks; i++) {
1264
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
1265 1266 1267 1268 1269
		if (rdev && unlikely(test_bit(Blocked, &rdev->flags))) {
			atomic_inc(&rdev->nr_pending);
			blocked_rdev = rdev;
			break;
		}
1270
		r1_bio->bios[i] = NULL;
1271 1272
		if (!rdev || test_bit(Faulty, &rdev->flags)
		    || test_bit(Unmerged, &rdev->flags)) {
1273 1274
			if (i < conf->raid_disks)
				set_bit(R1BIO_Degraded, &r1_bio->state);
1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301
			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;
1302
				rdev_dec_pending(rdev, mddev);
1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313
				/* 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;
1314
			}
1315 1316 1317 1318 1319 1320 1321
			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
1322 1323 1324
	}
	rcu_read_unlock();

1325 1326 1327
	if (unlikely(blocked_rdev)) {
		/* Wait for this device to become unblocked */
		int j;
1328
		sector_t old = start_next_window;
1329 1330 1331 1332

		for (j = 0; j < i; j++)
			if (r1_bio->bios[j])
				rdev_dec_pending(conf->mirrors[j].rdev, mddev);
1333
		r1_bio->state = 0;
1334
		allow_barrier(conf, start_next_window, bio->bi_iter.bi_sector);
1335
		md_wait_for_blocked_rdev(blocked_rdev, mddev);
1336 1337 1338 1339 1340 1341 1342 1343 1344 1345
		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);
1346 1347 1348
		goto retry_write;
	}

1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359
	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);
1360
	}
1361
	sectors_handled = r1_bio->sector + max_sectors - bio->bi_iter.bi_sector;
1362

1363
	atomic_set(&r1_bio->remaining, 1);
1364
	atomic_set(&r1_bio->behind_remaining, 0);
1365

1366
	first_clone = 1;
L
Linus Torvalds 已提交
1367 1368 1369 1370 1371
	for (i = 0; i < disks; i++) {
		struct bio *mbio;
		if (!r1_bio->bios[i])
			continue;

1372
		mbio = bio_clone_mddev(bio, GFP_NOIO, mddev);
1373
		bio_trim(mbio, r1_bio->sector - bio->bi_iter.bi_sector, max_sectors);
1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391

		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;
		}
1392
		if (r1_bio->behind_bvecs) {
1393 1394 1395
			struct bio_vec *bvec;
			int j;

1396 1397
			/*
			 * We trimmed the bio, so _all is legit
1398
			 */
1399
			bio_for_each_segment_all(bvec, mbio, j)
1400
				bvec->bv_page = r1_bio->behind_bvecs[j].bv_page;
1401 1402 1403 1404
			if (test_bit(WriteMostly, &conf->mirrors[i].rdev->flags))
				atomic_inc(&r1_bio->behind_remaining);
		}

1405 1406
		r1_bio->bios[i] = mbio;

1407
		mbio->bi_iter.bi_sector	= (r1_bio->sector +
1408 1409 1410
				   conf->mirrors[i].rdev->data_offset);
		mbio->bi_bdev = conf->mirrors[i].rdev->bdev;
		mbio->bi_end_io	= raid1_end_write_request;
1411 1412
		mbio->bi_rw =
			WRITE | do_flush_fua | do_sync | do_discard | do_same;
1413 1414
		mbio->bi_private = r1_bio;

L
Linus Torvalds 已提交
1415
		atomic_inc(&r1_bio->remaining);
1416 1417 1418 1419 1420 1421

		cb = blk_check_plugged(raid1_unplug, mddev, sizeof(*plug));
		if (cb)
			plug = container_of(cb, struct raid1_plug_cb, cb);
		else
			plug = NULL;
1422
		spin_lock_irqsave(&conf->device_lock, flags);
1423 1424 1425 1426 1427 1428 1429
		if (plug) {
			bio_list_add(&plug->pending, mbio);
			plug->pending_cnt++;
		} else {
			bio_list_add(&conf->pending_bio_list, mbio);
			conf->pending_count++;
		}
1430
		spin_unlock_irqrestore(&conf->device_lock, flags);
1431
		if (!plug)
N
NeilBrown 已提交
1432
			md_wakeup_thread(mddev->thread);
L
Linus Torvalds 已提交
1433
	}
1434 1435 1436
	/* Mustn't call r1_bio_write_done before this next test,
	 * as it could result in the bio being freed.
	 */
1437
	if (sectors_handled < bio_sectors(bio)) {
1438
		r1_bio_write_done(r1_bio);
1439 1440 1441 1442 1443
		/* 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;
1444
		r1_bio->sectors = bio_sectors(bio) - sectors_handled;
1445 1446
		r1_bio->state = 0;
		r1_bio->mddev = mddev;
1447
		r1_bio->sector = bio->bi_iter.bi_sector + sectors_handled;
1448 1449 1450
		goto retry_write;
	}

1451 1452 1453 1454
	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
1455 1456
}

1457
static void status(struct seq_file *seq, struct mddev *mddev)
L
Linus Torvalds 已提交
1458
{
1459
	struct r1conf *conf = mddev->private;
L
Linus Torvalds 已提交
1460 1461 1462
	int i;

	seq_printf(seq, " [%d/%d] [", conf->raid_disks,
1463
		   conf->raid_disks - mddev->degraded);
1464 1465
	rcu_read_lock();
	for (i = 0; i < conf->raid_disks; i++) {
1466
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
L
Linus Torvalds 已提交
1467
		seq_printf(seq, "%s",
1468 1469 1470
			   rdev && test_bit(In_sync, &rdev->flags) ? "U" : "_");
	}
	rcu_read_unlock();
L
Linus Torvalds 已提交
1471 1472 1473
	seq_printf(seq, "]");
}

1474
static void error(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
1475 1476
{
	char b[BDEVNAME_SIZE];
1477
	struct r1conf *conf = mddev->private;
L
Linus Torvalds 已提交
1478 1479 1480 1481 1482 1483 1484

	/*
	 * 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
	 */
1485
	if (test_bit(In_sync, &rdev->flags)
1486
	    && (conf->raid_disks - mddev->degraded) == 1) {
L
Linus Torvalds 已提交
1487 1488
		/*
		 * Don't fail the drive, act as though we were just a
1489 1490 1491
		 * 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
1492
		 */
1493
		conf->recovery_disabled = mddev->recovery_disabled;
L
Linus Torvalds 已提交
1494
		return;
1495
	}
1496
	set_bit(Blocked, &rdev->flags);
1497 1498 1499
	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
1500
		mddev->degraded++;
1501
		set_bit(Faulty, &rdev->flags);
1502
		spin_unlock_irqrestore(&conf->device_lock, flags);
1503 1504
	} else
		set_bit(Faulty, &rdev->flags);
1505 1506 1507 1508
	/*
	 * if recovery is running, make sure it aborts.
	 */
	set_bit(MD_RECOVERY_INTR, &mddev->recovery);
1509
	set_bit(MD_CHANGE_DEVS, &mddev->flags);
1510 1511 1512
	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
1513 1514
	       mdname(mddev), bdevname(rdev->bdev, b),
	       mdname(mddev), conf->raid_disks - mddev->degraded);
L
Linus Torvalds 已提交
1515 1516
}

1517
static void print_conf(struct r1conf *conf)
L
Linus Torvalds 已提交
1518 1519 1520
{
	int i;

N
NeilBrown 已提交
1521
	printk(KERN_DEBUG "RAID1 conf printout:\n");
L
Linus Torvalds 已提交
1522
	if (!conf) {
N
NeilBrown 已提交
1523
		printk(KERN_DEBUG "(!conf)\n");
L
Linus Torvalds 已提交
1524 1525
		return;
	}
N
NeilBrown 已提交
1526
	printk(KERN_DEBUG " --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
L
Linus Torvalds 已提交
1527 1528
		conf->raid_disks);

1529
	rcu_read_lock();
L
Linus Torvalds 已提交
1530 1531
	for (i = 0; i < conf->raid_disks; i++) {
		char b[BDEVNAME_SIZE];
1532
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
1533
		if (rdev)
N
NeilBrown 已提交
1534
			printk(KERN_DEBUG " disk %d, wo:%d, o:%d, dev:%s\n",
1535 1536 1537
			       i, !test_bit(In_sync, &rdev->flags),
			       !test_bit(Faulty, &rdev->flags),
			       bdevname(rdev->bdev,b));
L
Linus Torvalds 已提交
1538
	}
1539
	rcu_read_unlock();
L
Linus Torvalds 已提交
1540 1541
}

1542
static void close_sync(struct r1conf *conf)
L
Linus Torvalds 已提交
1543
{
1544 1545
	wait_barrier(conf, NULL);
	allow_barrier(conf, 0, 0);
L
Linus Torvalds 已提交
1546 1547 1548

	mempool_destroy(conf->r1buf_pool);
	conf->r1buf_pool = NULL;
1549

1550
	spin_lock_irq(&conf->resync_lock);
1551 1552
	conf->next_resync = 0;
	conf->start_next_window = MaxSector;
1553 1554 1555 1556
	conf->current_window_requests +=
		conf->next_window_requests;
	conf->next_window_requests = 0;
	spin_unlock_irq(&conf->resync_lock);
L
Linus Torvalds 已提交
1557 1558
}

1559
static int raid1_spare_active(struct mddev *mddev)
L
Linus Torvalds 已提交
1560 1561
{
	int i;
1562
	struct r1conf *conf = mddev->private;
1563 1564
	int count = 0;
	unsigned long flags;
L
Linus Torvalds 已提交
1565 1566

	/*
1567
	 * Find all failed disks within the RAID1 configuration
1568 1569
	 * and mark them readable.
	 * Called under mddev lock, so rcu protection not needed.
L
Linus Torvalds 已提交
1570 1571
	 */
	for (i = 0; i < conf->raid_disks; i++) {
1572
		struct md_rdev *rdev = conf->mirrors[i].rdev;
1573 1574
		struct md_rdev *repl = conf->mirrors[conf->raid_disks + i].rdev;
		if (repl
1575
		    && !test_bit(Candidate, &repl->flags)
1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592
		    && 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);
			}
		}
1593
		if (rdev
1594
		    && rdev->recovery_offset == MaxSector
1595
		    && !test_bit(Faulty, &rdev->flags)
1596
		    && !test_and_set_bit(In_sync, &rdev->flags)) {
1597
			count++;
1598
			sysfs_notify_dirent_safe(rdev->sysfs_state);
L
Linus Torvalds 已提交
1599 1600
		}
	}
1601 1602 1603
	spin_lock_irqsave(&conf->device_lock, flags);
	mddev->degraded -= count;
	spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1604 1605

	print_conf(conf);
1606
	return count;
L
Linus Torvalds 已提交
1607 1608
}

1609
static int raid1_add_disk(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
1610
{
1611
	struct r1conf *conf = mddev->private;
1612
	int err = -EEXIST;
1613
	int mirror = 0;
1614
	struct raid1_info *p;
1615
	int first = 0;
1616
	int last = conf->raid_disks - 1;
1617
	struct request_queue *q = bdev_get_queue(rdev->bdev);
L
Linus Torvalds 已提交
1618

1619 1620 1621
	if (mddev->recovery_disabled == conf->recovery_disabled)
		return -EBUSY;

1622 1623 1624
	if (rdev->raid_disk >= 0)
		first = last = rdev->raid_disk;

1625 1626 1627 1628 1629
	if (q->merge_bvec_fn) {
		set_bit(Unmerged, &rdev->flags);
		mddev->merge_check_needed = 1;
	}

1630 1631 1632
	for (mirror = first; mirror <= last; mirror++) {
		p = conf->mirrors+mirror;
		if (!p->rdev) {
L
Linus Torvalds 已提交
1633

1634 1635 1636
			if (mddev->gendisk)
				disk_stack_limits(mddev->gendisk, rdev->bdev,
						  rdev->data_offset << 9);
L
Linus Torvalds 已提交
1637 1638 1639

			p->head_position = 0;
			rdev->raid_disk = mirror;
1640
			err = 0;
1641 1642 1643 1644
			/* 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)
1645
				conf->fullsync = 1;
1646
			rcu_assign_pointer(p->rdev, rdev);
L
Linus Torvalds 已提交
1647 1648
			break;
		}
1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660
		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;
		}
	}
1661 1662 1663 1664 1665 1666 1667 1668 1669
	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();
1670 1671
		freeze_array(conf, 0);
		unfreeze_array(conf);
1672 1673
		clear_bit(Unmerged, &rdev->flags);
	}
1674
	md_integrity_add_rdev(rdev, mddev);
1675
	if (mddev->queue && blk_queue_discard(bdev_get_queue(rdev->bdev)))
S
Shaohua Li 已提交
1676
		queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, mddev->queue);
L
Linus Torvalds 已提交
1677
	print_conf(conf);
1678
	return err;
L
Linus Torvalds 已提交
1679 1680
}

1681
static int raid1_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
L
Linus Torvalds 已提交
1682
{
1683
	struct r1conf *conf = mddev->private;
L
Linus Torvalds 已提交
1684
	int err = 0;
1685
	int number = rdev->raid_disk;
1686
	struct raid1_info *p = conf->mirrors + number;
L
Linus Torvalds 已提交
1687

1688 1689 1690
	if (rdev != p->rdev)
		p = conf->mirrors + conf->raid_disks + number;

L
Linus Torvalds 已提交
1691
	print_conf(conf);
1692
	if (rdev == p->rdev) {
1693
		if (test_bit(In_sync, &rdev->flags) ||
L
Linus Torvalds 已提交
1694 1695 1696 1697
		    atomic_read(&rdev->nr_pending)) {
			err = -EBUSY;
			goto abort;
		}
N
NeilBrown 已提交
1698
		/* Only remove non-faulty devices if recovery
1699 1700 1701
		 * is not possible.
		 */
		if (!test_bit(Faulty, &rdev->flags) &&
1702
		    mddev->recovery_disabled != conf->recovery_disabled &&
1703 1704 1705 1706
		    mddev->degraded < conf->raid_disks) {
			err = -EBUSY;
			goto abort;
		}
L
Linus Torvalds 已提交
1707
		p->rdev = NULL;
1708
		synchronize_rcu();
L
Linus Torvalds 已提交
1709 1710 1711 1712
		if (atomic_read(&rdev->nr_pending)) {
			/* lost the race, try later */
			err = -EBUSY;
			p->rdev = rdev;
1713
			goto abort;
1714 1715 1716 1717 1718 1719 1720
		} 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;
1721
			freeze_array(conf, 0);
1722 1723 1724
			clear_bit(Replacement, &repl->flags);
			p->rdev = repl;
			conf->mirrors[conf->raid_disks + number].rdev = NULL;
1725
			unfreeze_array(conf);
1726 1727
			clear_bit(WantReplacement, &rdev->flags);
		} else
1728
			clear_bit(WantReplacement, &rdev->flags);
1729
		err = md_integrity_register(mddev);
L
Linus Torvalds 已提交
1730 1731 1732 1733 1734 1735 1736
	}
abort:

	print_conf(conf);
	return err;
}

1737
static void end_sync_read(struct bio *bio, int error)
L
Linus Torvalds 已提交
1738
{
1739
	struct r1bio *r1_bio = bio->bi_private;
L
Linus Torvalds 已提交
1740

1741
	update_head_pos(r1_bio->read_disk, r1_bio);
1742

L
Linus Torvalds 已提交
1743 1744 1745 1746 1747
	/*
	 * 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
	 */
1748
	if (test_bit(BIO_UPTODATE, &bio->bi_flags))
L
Linus Torvalds 已提交
1749
		set_bit(R1BIO_Uptodate, &r1_bio->state);
1750 1751 1752

	if (atomic_dec_and_test(&r1_bio->remaining))
		reschedule_retry(r1_bio);
L
Linus Torvalds 已提交
1753 1754
}

1755
static void end_sync_write(struct bio *bio, int error)
L
Linus Torvalds 已提交
1756 1757
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
1758
	struct r1bio *r1_bio = bio->bi_private;
1759
	struct mddev *mddev = r1_bio->mddev;
1760
	struct r1conf *conf = mddev->private;
L
Linus Torvalds 已提交
1761
	int mirror=0;
1762 1763
	sector_t first_bad;
	int bad_sectors;
L
Linus Torvalds 已提交
1764

1765 1766
	mirror = find_bio_disk(r1_bio, bio);

1767
	if (!uptodate) {
N
NeilBrown 已提交
1768
		sector_t sync_blocks = 0;
1769 1770 1771 1772
		sector_t s = r1_bio->sector;
		long sectors_to_go = r1_bio->sectors;
		/* make sure these bits doesn't get cleared. */
		do {
1773
			bitmap_end_sync(mddev->bitmap, s,
1774 1775 1776 1777
					&sync_blocks, 1);
			s += sync_blocks;
			sectors_to_go -= sync_blocks;
		} while (sectors_to_go > 0);
1778 1779
		set_bit(WriteErrorSeen,
			&conf->mirrors[mirror].rdev->flags);
1780 1781 1782 1783
		if (!test_and_set_bit(WantReplacement,
				      &conf->mirrors[mirror].rdev->flags))
			set_bit(MD_RECOVERY_NEEDED, &
				mddev->recovery);
1784
		set_bit(R1BIO_WriteError, &r1_bio->state);
1785 1786 1787
	} else if (is_badblock(conf->mirrors[mirror].rdev,
			       r1_bio->sector,
			       r1_bio->sectors,
1788 1789 1790 1791 1792 1793
			       &first_bad, &bad_sectors) &&
		   !is_badblock(conf->mirrors[r1_bio->read_disk].rdev,
				r1_bio->sector,
				r1_bio->sectors,
				&first_bad, &bad_sectors)
		)
1794
		set_bit(R1BIO_MadeGood, &r1_bio->state);
1795

L
Linus Torvalds 已提交
1796
	if (atomic_dec_and_test(&r1_bio->remaining)) {
1797
		int s = r1_bio->sectors;
1798 1799
		if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
		    test_bit(R1BIO_WriteError, &r1_bio->state))
1800 1801 1802 1803 1804
			reschedule_retry(r1_bio);
		else {
			put_buf(r1_bio);
			md_done_sync(mddev, s, uptodate);
		}
L
Linus Torvalds 已提交
1805 1806 1807
	}
}

1808
static int r1_sync_page_io(struct md_rdev *rdev, sector_t sector,
1809 1810 1811 1812 1813
			    int sectors, struct page *page, int rw)
{
	if (sync_page_io(rdev, sector, sectors << 9, page, rw, false))
		/* success */
		return 1;
1814
	if (rw == WRITE) {
1815
		set_bit(WriteErrorSeen, &rdev->flags);
1816 1817 1818 1819 1820
		if (!test_and_set_bit(WantReplacement,
				      &rdev->flags))
			set_bit(MD_RECOVERY_NEEDED, &
				rdev->mddev->recovery);
	}
1821 1822 1823 1824 1825 1826
	/* 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;
}

1827
static int fix_sync_read_error(struct r1bio *r1_bio)
L
Linus Torvalds 已提交
1828
{
1829 1830 1831 1832 1833 1834 1835
	/* 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.
1836 1837 1838
	 * 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.
1839
	 */
1840
	struct mddev *mddev = r1_bio->mddev;
1841
	struct r1conf *conf = mddev->private;
1842 1843 1844 1845 1846 1847 1848 1849 1850
	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;
1851
		struct md_rdev *rdev;
1852
		int start;
1853 1854 1855 1856 1857 1858 1859 1860 1861 1862

		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;
1863
				if (sync_page_io(rdev, sect, s<<9,
1864 1865 1866 1867 1868 1869 1870
						 bio->bi_io_vec[idx].bv_page,
						 READ, false)) {
					success = 1;
					break;
				}
			}
			d++;
1871
			if (d == conf->raid_disks * 2)
1872 1873 1874
				d = 0;
		} while (!success && d != r1_bio->read_disk);

1875
		if (!success) {
1876
			char b[BDEVNAME_SIZE];
1877 1878 1879 1880 1881 1882
			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.
			 */
1883 1884 1885 1886 1887
			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);
1888
			for (d = 0; d < conf->raid_disks * 2; d++) {
1889 1890 1891 1892 1893 1894 1895
				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) {
1896 1897
				conf->recovery_disabled =
					mddev->recovery_disabled;
1898 1899 1900 1901 1902 1903 1904 1905 1906 1907
				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;
1908
		}
1909 1910 1911 1912 1913

		start = d;
		/* write it back and re-read */
		while (d != r1_bio->read_disk) {
			if (d == 0)
1914
				d = conf->raid_disks * 2;
1915 1916 1917 1918
			d--;
			if (r1_bio->bios[d]->bi_end_io != end_sync_read)
				continue;
			rdev = conf->mirrors[d].rdev;
1919 1920 1921
			if (r1_sync_page_io(rdev, sect, s,
					    bio->bi_io_vec[idx].bv_page,
					    WRITE) == 0) {
1922 1923
				r1_bio->bios[d]->bi_end_io = NULL;
				rdev_dec_pending(rdev, mddev);
1924
			}
1925 1926 1927 1928
		}
		d = start;
		while (d != r1_bio->read_disk) {
			if (d == 0)
1929
				d = conf->raid_disks * 2;
1930 1931 1932 1933
			d--;
			if (r1_bio->bios[d]->bi_end_io != end_sync_read)
				continue;
			rdev = conf->mirrors[d].rdev;
1934 1935 1936
			if (r1_sync_page_io(rdev, sect, s,
					    bio->bi_io_vec[idx].bv_page,
					    READ) != 0)
1937
				atomic_add(s, &rdev->corrected_errors);
1938
		}
1939 1940 1941 1942
		sectors -= s;
		sect += s;
		idx ++;
	}
1943
	set_bit(R1BIO_Uptodate, &r1_bio->state);
1944
	set_bit(BIO_UPTODATE, &bio->bi_flags);
1945 1946 1947
	return 1;
}

1948
static void process_checks(struct r1bio *r1_bio)
1949 1950 1951 1952 1953 1954 1955 1956
{
	/* 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
	 */
1957
	struct mddev *mddev = r1_bio->mddev;
1958
	struct r1conf *conf = mddev->private;
1959 1960
	int primary;
	int i;
1961
	int vcnt;
1962

1963 1964 1965 1966 1967
	/* 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;
1968
		int uptodate;
1969 1970 1971
		struct bio *b = r1_bio->bios[i];
		if (b->bi_end_io != end_sync_read)
			continue;
1972 1973
		/* fixup the bio for reuse, but preserve BIO_UPTODATE */
		uptodate = test_bit(BIO_UPTODATE, &b->bi_flags);
1974
		bio_reset(b);
1975 1976
		if (!uptodate)
			clear_bit(BIO_UPTODATE, &b->bi_flags);
1977
		b->bi_vcnt = vcnt;
1978 1979
		b->bi_iter.bi_size = r1_bio->sectors << 9;
		b->bi_iter.bi_sector = r1_bio->sector +
1980 1981 1982 1983 1984
			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;

1985
		size = b->bi_iter.bi_size;
1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996
		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;
		}
	}
1997
	for (primary = 0; primary < conf->raid_disks * 2; primary++)
1998 1999 2000 2001 2002 2003 2004
		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;
2005
	for (i = 0; i < conf->raid_disks * 2; i++) {
2006 2007 2008
		int j;
		struct bio *pbio = r1_bio->bios[primary];
		struct bio *sbio = r1_bio->bios[i];
2009
		int uptodate = test_bit(BIO_UPTODATE, &sbio->bi_flags);
2010

K
Kent Overstreet 已提交
2011
		if (sbio->bi_end_io != end_sync_read)
2012
			continue;
2013 2014
		/* Now we can 'fixup' the BIO_UPTODATE flag */
		set_bit(BIO_UPTODATE, &sbio->bi_flags);
2015

2016
		if (uptodate) {
2017 2018 2019 2020 2021 2022
			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),
2023
					   sbio->bi_io_vec[j].bv_len))
2024
					break;
2025
			}
2026 2027 2028
		} else
			j = 0;
		if (j >= 0)
2029
			atomic64_add(r1_bio->sectors, &mddev->resync_mismatches);
2030
		if (j < 0 || (test_bit(MD_RECOVERY_CHECK, &mddev->recovery)
2031
			      && uptodate)) {
2032 2033 2034 2035 2036
			/* 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
2037 2038

		bio_copy_data(sbio, pbio);
2039
	}
2040 2041
}

2042
static void sync_request_write(struct mddev *mddev, struct r1bio *r1_bio)
2043
{
2044
	struct r1conf *conf = mddev->private;
2045
	int i;
2046
	int disks = conf->raid_disks * 2;
2047 2048 2049 2050 2051 2052 2053 2054
	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;
2055 2056

	if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
2057 2058
		process_checks(r1_bio);

2059 2060 2061
	/*
	 * schedule writes
	 */
L
Linus Torvalds 已提交
2062 2063 2064
	atomic_set(&r1_bio->remaining, 1);
	for (i = 0; i < disks ; i++) {
		wbio = r1_bio->bios[i];
2065 2066 2067 2068
		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
2069 2070
			continue;

2071 2072
		wbio->bi_rw = WRITE;
		wbio->bi_end_io = end_sync_write;
L
Linus Torvalds 已提交
2073
		atomic_inc(&r1_bio->remaining);
2074
		md_sync_acct(conf->mirrors[i].rdev->bdev, bio_sectors(wbio));
2075

L
Linus Torvalds 已提交
2076 2077 2078 2079
		generic_make_request(wbio);
	}

	if (atomic_dec_and_test(&r1_bio->remaining)) {
2080
		/* if we're here, all write(s) have completed, so clean up */
2081 2082 2083 2084 2085 2086 2087 2088
		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
2089 2090 2091 2092 2093 2094 2095 2096
	}
}

/*
 * This is a kernel thread which:
 *
 *	1.	Retries failed read operations on working mirrors.
 *	2.	Updates the raid superblock when problems encounter.
2097
 *	3.	Performs writes following reads for array synchronising.
L
Linus Torvalds 已提交
2098 2099
 */

2100
static void fix_read_error(struct r1conf *conf, int read_disk,
2101 2102
			   sector_t sect, int sectors)
{
2103
	struct mddev *mddev = conf->mddev;
2104 2105 2106 2107 2108
	while(sectors) {
		int s = sectors;
		int d = read_disk;
		int success = 0;
		int start;
2109
		struct md_rdev *rdev;
2110 2111 2112 2113 2114 2115 2116 2117 2118 2119

		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....
			 */
2120 2121 2122
			sector_t first_bad;
			int bad_sectors;

2123 2124
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
2125 2126 2127
			    (test_bit(In_sync, &rdev->flags) ||
			     (!test_bit(Faulty, &rdev->flags) &&
			      rdev->recovery_offset >= sect + s)) &&
2128 2129
			    is_badblock(rdev, sect, s,
					&first_bad, &bad_sectors) == 0 &&
J
Jonathan Brassow 已提交
2130 2131
			    sync_page_io(rdev, sect, s<<9,
					 conf->tmppage, READ, false))
2132 2133 2134
				success = 1;
			else {
				d++;
2135
				if (d == conf->raid_disks * 2)
2136 2137 2138 2139 2140
					d = 0;
			}
		} while (!success && d != read_disk);

		if (!success) {
2141
			/* Cannot read from anywhere - mark it bad */
2142
			struct md_rdev *rdev = conf->mirrors[read_disk].rdev;
2143 2144
			if (!rdev_set_badblocks(rdev, sect, s, 0))
				md_error(mddev, rdev);
2145 2146 2147 2148 2149 2150
			break;
		}
		/* write it back and re-read */
		start = d;
		while (d != read_disk) {
			if (d==0)
2151
				d = conf->raid_disks * 2;
2152 2153 2154
			d--;
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
2155
			    !test_bit(Faulty, &rdev->flags))
2156 2157
				r1_sync_page_io(rdev, sect, s,
						conf->tmppage, WRITE);
2158 2159 2160 2161 2162
		}
		d = start;
		while (d != read_disk) {
			char b[BDEVNAME_SIZE];
			if (d==0)
2163
				d = conf->raid_disks * 2;
2164 2165 2166
			d--;
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
2167
			    !test_bit(Faulty, &rdev->flags)) {
2168 2169
				if (r1_sync_page_io(rdev, sect, s,
						    conf->tmppage, READ)) {
2170 2171
					atomic_add(s, &rdev->corrected_errors);
					printk(KERN_INFO
N
NeilBrown 已提交
2172
					       "md/raid1:%s: read error corrected "
2173 2174
					       "(%d sectors at %llu on %s)\n",
					       mdname(mddev), s,
2175 2176
					       (unsigned long long)(sect +
					           rdev->data_offset),
2177 2178 2179 2180 2181 2182 2183 2184 2185
					       bdevname(rdev->bdev, b));
				}
			}
		}
		sectors -= s;
		sect += s;
	}
}

2186
static int narrow_write_error(struct r1bio *r1_bio, int i)
2187
{
2188
	struct mddev *mddev = r1_bio->mddev;
2189
	struct r1conf *conf = mddev->private;
2190
	struct md_rdev *rdev = conf->mirrors[i].rdev;
2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211

	/* 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;

2212 2213
	block_sectors = roundup(1 << rdev->badblocks.shift,
				bdev_logical_block_size(rdev->bdev) >> 9);
2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224
	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'*/

2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241
		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);
		}

2242
		wbio->bi_rw = WRITE;
2243 2244
		wbio->bi_iter.bi_sector = r1_bio->sector;
		wbio->bi_iter.bi_size = r1_bio->sectors << 9;
2245

2246
		bio_trim(wbio, sector - r1_bio->sector, sectors);
2247
		wbio->bi_iter.bi_sector += rdev->data_offset;
2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262
		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;
}

2263
static void handle_sync_write_finished(struct r1conf *conf, struct r1bio *r1_bio)
2264 2265 2266
{
	int m;
	int s = r1_bio->sectors;
2267
	for (m = 0; m < conf->raid_disks * 2 ; m++) {
2268
		struct md_rdev *rdev = conf->mirrors[m].rdev;
2269 2270 2271 2272 2273
		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)) {
2274
			rdev_clear_badblocks(rdev, r1_bio->sector, s, 0);
2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285
		}
		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);
}

2286
static void handle_write_finished(struct r1conf *conf, struct r1bio *r1_bio)
2287 2288
{
	int m;
2289
	for (m = 0; m < conf->raid_disks * 2 ; m++)
2290
		if (r1_bio->bios[m] == IO_MADE_GOOD) {
2291
			struct md_rdev *rdev = conf->mirrors[m].rdev;
2292 2293
			rdev_clear_badblocks(rdev,
					     r1_bio->sector,
2294
					     r1_bio->sectors, 0);
2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314
			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);
}

2315
static void handle_read_error(struct r1conf *conf, struct r1bio *r1_bio)
2316 2317 2318
{
	int disk;
	int max_sectors;
2319
	struct mddev *mddev = conf->mddev;
2320 2321
	struct bio *bio;
	char b[BDEVNAME_SIZE];
2322
	struct md_rdev *rdev;
2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333

	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) {
2334
		freeze_array(conf, 1);
2335 2336 2337 2338 2339
		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);
2340
	rdev_dec_pending(conf->mirrors[r1_bio->read_disk].rdev, conf->mddev);
2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360

	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);
2361 2362
		bio_trim(bio, r1_bio->sector - bio->bi_iter.bi_sector,
			 max_sectors);
2363 2364 2365 2366 2367 2368 2369 2370
		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));
2371
		bio->bi_iter.bi_sector = r1_bio->sector + rdev->data_offset;
2372 2373 2374 2375 2376 2377 2378 2379
		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
2380
					       - mbio->bi_iter.bi_sector);
2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393
			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;
2394
			r1_bio->sectors = bio_sectors(mbio) - sectors_handled;
2395 2396 2397
			r1_bio->state = 0;
			set_bit(R1BIO_ReadError, &r1_bio->state);
			r1_bio->mddev = mddev;
2398 2399
			r1_bio->sector = mbio->bi_iter.bi_sector +
				sectors_handled;
2400 2401 2402 2403 2404 2405 2406

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

S
Shaohua Li 已提交
2407
static void raid1d(struct md_thread *thread)
L
Linus Torvalds 已提交
2408
{
S
Shaohua Li 已提交
2409
	struct mddev *mddev = thread->mddev;
2410
	struct r1bio *r1_bio;
L
Linus Torvalds 已提交
2411
	unsigned long flags;
2412
	struct r1conf *conf = mddev->private;
L
Linus Torvalds 已提交
2413
	struct list_head *head = &conf->retry_list;
2414
	struct blk_plug plug;
L
Linus Torvalds 已提交
2415 2416

	md_check_recovery(mddev);
2417 2418

	blk_start_plug(&plug);
L
Linus Torvalds 已提交
2419
	for (;;) {
2420

2421
		flush_pending_writes(conf);
2422

2423 2424 2425
		spin_lock_irqsave(&conf->device_lock, flags);
		if (list_empty(head)) {
			spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
2426
			break;
2427
		}
2428
		r1_bio = list_entry(head->prev, struct r1bio, retry_list);
L
Linus Torvalds 已提交
2429
		list_del(head->prev);
2430
		conf->nr_queued--;
L
Linus Torvalds 已提交
2431 2432 2433
		spin_unlock_irqrestore(&conf->device_lock, flags);

		mddev = r1_bio->mddev;
2434
		conf = mddev->private;
2435
		if (test_bit(R1BIO_IsSync, &r1_bio->state)) {
2436
			if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
2437 2438 2439
			    test_bit(R1BIO_WriteError, &r1_bio->state))
				handle_sync_write_finished(conf, r1_bio);
			else
2440
				sync_request_write(mddev, r1_bio);
2441
		} else if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
2442 2443 2444 2445 2446
			   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
2447 2448 2449 2450
			/* just a partial read to be scheduled from separate
			 * context
			 */
			generic_make_request(r1_bio->bios[r1_bio->read_disk]);
2451

N
NeilBrown 已提交
2452
		cond_resched();
2453 2454
		if (mddev->flags & ~(1<<MD_CHANGE_PENDING))
			md_check_recovery(mddev);
L
Linus Torvalds 已提交
2455
	}
2456
	blk_finish_plug(&plug);
L
Linus Torvalds 已提交
2457 2458
}

2459
static int init_resync(struct r1conf *conf)
L
Linus Torvalds 已提交
2460 2461 2462 2463
{
	int buffs;

	buffs = RESYNC_WINDOW / RESYNC_BLOCK_SIZE;
2464
	BUG_ON(conf->r1buf_pool);
L
Linus Torvalds 已提交
2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482
	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.
 */

2483
static sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipped, int go_faster)
L
Linus Torvalds 已提交
2484
{
2485
	struct r1conf *conf = mddev->private;
2486
	struct r1bio *r1_bio;
L
Linus Torvalds 已提交
2487 2488
	struct bio *bio;
	sector_t max_sector, nr_sectors;
2489
	int disk = -1;
L
Linus Torvalds 已提交
2490
	int i;
2491 2492
	int wonly = -1;
	int write_targets = 0, read_targets = 0;
N
NeilBrown 已提交
2493
	sector_t sync_blocks;
2494
	int still_degraded = 0;
2495 2496
	int good_sectors = RESYNC_SECTORS;
	int min_bad = 0; /* number of sectors that are bad in all devices */
L
Linus Torvalds 已提交
2497 2498 2499

	if (!conf->r1buf_pool)
		if (init_resync(conf))
2500
			return 0;
L
Linus Torvalds 已提交
2501

A
Andre Noll 已提交
2502
	max_sector = mddev->dev_sectors;
L
Linus Torvalds 已提交
2503
	if (sector_nr >= max_sector) {
2504 2505 2506 2507 2508
		/* 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
		 */
2509 2510
		if (mddev->curr_resync < max_sector) /* aborted */
			bitmap_end_sync(mddev->bitmap, mddev->curr_resync,
2511
						&sync_blocks, 1);
2512
		else /* completed sync */
2513
			conf->fullsync = 0;
2514 2515

		bitmap_close_sync(mddev->bitmap);
L
Linus Torvalds 已提交
2516 2517 2518 2519
		close_sync(conf);
		return 0;
	}

2520 2521
	if (mddev->bitmap == NULL &&
	    mddev->recovery_cp == MaxSector &&
2522
	    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
2523 2524 2525 2526
	    conf->fullsync == 0) {
		*skipped = 1;
		return max_sector - sector_nr;
	}
2527 2528 2529
	/* before building a request, check if we can skip these blocks..
	 * This call the bitmap_start_sync doesn't actually record anything
	 */
2530
	if (!bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, 1) &&
2531
	    !conf->fullsync && !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
2532 2533 2534 2535
		/* We can skip this block, and probably several more */
		*skipped = 1;
		return sync_blocks;
	}
L
Linus Torvalds 已提交
2536
	/*
2537 2538 2539
	 * 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
2540
	 */
2541
	if (!go_faster && conf->nr_waiting)
L
Linus Torvalds 已提交
2542
		msleep_interruptible(1000);
2543

N
NeilBrown 已提交
2544
	bitmap_cond_end_sync(mddev->bitmap, sector_nr);
2545
	r1_bio = mempool_alloc(conf->r1buf_pool, GFP_NOIO);
2546

2547
	raise_barrier(conf, sector_nr);
L
Linus Torvalds 已提交
2548

2549
	rcu_read_lock();
L
Linus Torvalds 已提交
2550
	/*
2551 2552 2553 2554 2555 2556
	 * 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
2557 2558 2559 2560
	 */

	r1_bio->mddev = mddev;
	r1_bio->sector = sector_nr;
2561
	r1_bio->state = 0;
L
Linus Torvalds 已提交
2562 2563
	set_bit(R1BIO_IsSync, &r1_bio->state);

2564
	for (i = 0; i < conf->raid_disks * 2; i++) {
2565
		struct md_rdev *rdev;
L
Linus Torvalds 已提交
2566
		bio = r1_bio->bios[i];
K
Kent Overstreet 已提交
2567
		bio_reset(bio);
L
Linus Torvalds 已提交
2568

2569 2570
		rdev = rcu_dereference(conf->mirrors[i].rdev);
		if (rdev == NULL ||
2571
		    test_bit(Faulty, &rdev->flags)) {
2572 2573
			if (i < conf->raid_disks)
				still_degraded = 1;
2574
		} else if (!test_bit(In_sync, &rdev->flags)) {
L
Linus Torvalds 已提交
2575 2576 2577
			bio->bi_rw = WRITE;
			bio->bi_end_io = end_sync_write;
			write_targets ++;
2578 2579
		} else {
			/* may need to read from here */
2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604
			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++;
2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616
			} 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++;
2617 2618
			}
		}
2619 2620
		if (bio->bi_end_io) {
			atomic_inc(&rdev->nr_pending);
2621
			bio->bi_iter.bi_sector = sector_nr + rdev->data_offset;
2622 2623 2624
			bio->bi_bdev = rdev->bdev;
			bio->bi_private = r1_bio;
		}
L
Linus Torvalds 已提交
2625
	}
2626 2627 2628 2629
	rcu_read_unlock();
	if (disk < 0)
		disk = wonly;
	r1_bio->read_disk = disk;
2630

2631 2632 2633 2634 2635
	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;
2636
		for (i = 0 ; i < conf->raid_disks * 2 ; i++)
2637
			if (r1_bio->bios[i]->bi_end_io == end_sync_write) {
2638
				struct md_rdev *rdev = conf->mirrors[i].rdev;
2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665
				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;
	}

2666 2667 2668 2669 2670
	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
2671 2672 2673
		/* There is nowhere to write, so all non-sync
		 * drives must be failed - so we are finished
		 */
2674 2675 2676 2677
		sector_t rv;
		if (min_bad > 0)
			max_sector = sector_nr + min_bad;
		rv = max_sector - sector_nr;
2678
		*skipped = 1;
L
Linus Torvalds 已提交
2679 2680 2681 2682
		put_buf(r1_bio);
		return rv;
	}

2683 2684
	if (max_sector > mddev->resync_max)
		max_sector = mddev->resync_max; /* Don't do IO beyond here */
2685 2686
	if (max_sector > sector_nr + good_sectors)
		max_sector = sector_nr + good_sectors;
L
Linus Torvalds 已提交
2687
	nr_sectors = 0;
2688
	sync_blocks = 0;
L
Linus Torvalds 已提交
2689 2690 2691 2692 2693 2694 2695
	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;
2696 2697
		if (sync_blocks == 0) {
			if (!bitmap_start_sync(mddev->bitmap, sector_nr,
2698 2699 2700
					       &sync_blocks, still_degraded) &&
			    !conf->fullsync &&
			    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
2701
				break;
2702
			BUG_ON(sync_blocks < (PAGE_SIZE>>9));
2703
			if ((len >> 9) > sync_blocks)
2704
				len = sync_blocks<<9;
2705
		}
2706

2707
		for (i = 0 ; i < conf->raid_disks * 2; i++) {
L
Linus Torvalds 已提交
2708 2709
			bio = r1_bio->bios[i];
			if (bio->bi_end_io) {
2710
				page = bio->bi_io_vec[bio->bi_vcnt].bv_page;
L
Linus Torvalds 已提交
2711 2712
				if (bio_add_page(bio, page, len, 0) == 0) {
					/* stop here */
2713
					bio->bi_io_vec[bio->bi_vcnt].bv_page = page;
L
Linus Torvalds 已提交
2714 2715 2716
					while (i > 0) {
						i--;
						bio = r1_bio->bios[i];
2717 2718
						if (bio->bi_end_io==NULL)
							continue;
L
Linus Torvalds 已提交
2719 2720
						/* remove last page from this bio */
						bio->bi_vcnt--;
2721
						bio->bi_iter.bi_size -= len;
2722
						__clear_bit(BIO_SEG_VALID, &bio->bi_flags);
L
Linus Torvalds 已提交
2723 2724 2725 2726 2727 2728 2729
					}
					goto bio_full;
				}
			}
		}
		nr_sectors += len>>9;
		sector_nr += len>>9;
2730
		sync_blocks -= (len>>9);
L
Linus Torvalds 已提交
2731 2732 2733 2734
	} while (r1_bio->bios[disk]->bi_vcnt < RESYNC_PAGES);
 bio_full:
	r1_bio->sectors = nr_sectors;

2735 2736 2737 2738 2739
	/* 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);
2740
		for (i = 0; i < conf->raid_disks * 2 && read_targets; i++) {
2741 2742
			bio = r1_bio->bios[i];
			if (bio->bi_end_io == end_sync_read) {
2743
				read_targets--;
2744
				md_sync_acct(bio->bi_bdev, nr_sectors);
2745 2746 2747 2748 2749 2750
				generic_make_request(bio);
			}
		}
	} else {
		atomic_set(&r1_bio->remaining, 1);
		bio = r1_bio->bios[r1_bio->read_disk];
2751
		md_sync_acct(bio->bi_bdev, nr_sectors);
2752
		generic_make_request(bio);
L
Linus Torvalds 已提交
2753

2754
	}
L
Linus Torvalds 已提交
2755 2756 2757
	return nr_sectors;
}

2758
static sector_t raid1_size(struct mddev *mddev, sector_t sectors, int raid_disks)
2759 2760 2761 2762 2763 2764 2765
{
	if (sectors)
		return sectors;

	return mddev->dev_sectors;
}

2766
static struct r1conf *setup_conf(struct mddev *mddev)
L
Linus Torvalds 已提交
2767
{
2768
	struct r1conf *conf;
2769
	int i;
2770
	struct raid1_info *disk;
2771
	struct md_rdev *rdev;
2772
	int err = -ENOMEM;
L
Linus Torvalds 已提交
2773

2774
	conf = kzalloc(sizeof(struct r1conf), GFP_KERNEL);
L
Linus Torvalds 已提交
2775
	if (!conf)
2776
		goto abort;
L
Linus Torvalds 已提交
2777

2778
	conf->mirrors = kzalloc(sizeof(struct raid1_info)
2779
				* mddev->raid_disks * 2,
L
Linus Torvalds 已提交
2780 2781
				 GFP_KERNEL);
	if (!conf->mirrors)
2782
		goto abort;
L
Linus Torvalds 已提交
2783

2784 2785
	conf->tmppage = alloc_page(GFP_KERNEL);
	if (!conf->tmppage)
2786
		goto abort;
2787

2788
	conf->poolinfo = kzalloc(sizeof(*conf->poolinfo), GFP_KERNEL);
L
Linus Torvalds 已提交
2789
	if (!conf->poolinfo)
2790
		goto abort;
2791
	conf->poolinfo->raid_disks = mddev->raid_disks * 2;
L
Linus Torvalds 已提交
2792 2793 2794 2795
	conf->r1bio_pool = mempool_create(NR_RAID1_BIOS, r1bio_pool_alloc,
					  r1bio_pool_free,
					  conf->poolinfo);
	if (!conf->r1bio_pool)
2796 2797
		goto abort;

2798
	conf->poolinfo->mddev = mddev;
L
Linus Torvalds 已提交
2799

2800
	err = -EINVAL;
2801
	spin_lock_init(&conf->device_lock);
N
NeilBrown 已提交
2802
	rdev_for_each(rdev, mddev) {
2803
		struct request_queue *q;
2804
		int disk_idx = rdev->raid_disk;
L
Linus Torvalds 已提交
2805 2806 2807
		if (disk_idx >= mddev->raid_disks
		    || disk_idx < 0)
			continue;
2808
		if (test_bit(Replacement, &rdev->flags))
2809
			disk = conf->mirrors + mddev->raid_disks + disk_idx;
2810 2811
		else
			disk = conf->mirrors + disk_idx;
L
Linus Torvalds 已提交
2812

2813 2814
		if (disk->rdev)
			goto abort;
L
Linus Torvalds 已提交
2815
		disk->rdev = rdev;
2816 2817 2818
		q = bdev_get_queue(rdev->bdev);
		if (q->merge_bvec_fn)
			mddev->merge_check_needed = 1;
L
Linus Torvalds 已提交
2819 2820

		disk->head_position = 0;
2821
		disk->seq_start = MaxSector;
L
Linus Torvalds 已提交
2822 2823 2824 2825 2826 2827
	}
	conf->raid_disks = mddev->raid_disks;
	conf->mddev = mddev;
	INIT_LIST_HEAD(&conf->retry_list);

	spin_lock_init(&conf->resync_lock);
2828
	init_waitqueue_head(&conf->wait_barrier);
L
Linus Torvalds 已提交
2829

2830
	bio_list_init(&conf->pending_bio_list);
2831
	conf->pending_count = 0;
2832
	conf->recovery_disabled = mddev->recovery_disabled - 1;
2833

2834 2835 2836
	conf->start_next_window = MaxSector;
	conf->current_window_requests = conf->next_window_requests = 0;

2837
	err = -EIO;
2838
	for (i = 0; i < conf->raid_disks * 2; i++) {
L
Linus Torvalds 已提交
2839 2840 2841

		disk = conf->mirrors + i;

2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856
		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;
		}

2857 2858
		if (!disk->rdev ||
		    !test_bit(In_sync, &disk->rdev->flags)) {
L
Linus Torvalds 已提交
2859
			disk->head_position = 0;
2860 2861
			if (disk->rdev &&
			    (disk->rdev->saved_raid_disk < 0))
2862
				conf->fullsync = 1;
2863
		}
L
Linus Torvalds 已提交
2864
	}
2865 2866

	err = -ENOMEM;
2867
	conf->thread = md_register_thread(raid1d, mddev, "raid1");
2868 2869
	if (!conf->thread) {
		printk(KERN_ERR
N
NeilBrown 已提交
2870
		       "md/raid1:%s: couldn't allocate thread\n",
2871 2872
		       mdname(mddev));
		goto abort;
2873
	}
L
Linus Torvalds 已提交
2874

2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888
	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);
}

N
NeilBrown 已提交
2889
static void raid1_free(struct mddev *mddev, void *priv);
2890
static int run(struct mddev *mddev)
2891
{
2892
	struct r1conf *conf;
2893
	int i;
2894
	struct md_rdev *rdev;
2895
	int ret;
S
Shaohua Li 已提交
2896
	bool discard_supported = false;
2897 2898

	if (mddev->level != 1) {
N
NeilBrown 已提交
2899
		printk(KERN_ERR "md/raid1:%s: raid level not set to mirroring (%d)\n",
2900 2901 2902 2903
		       mdname(mddev), mddev->level);
		return -EIO;
	}
	if (mddev->reshape_position != MaxSector) {
N
NeilBrown 已提交
2904
		printk(KERN_ERR "md/raid1:%s: reshape_position set but not supported\n",
2905 2906 2907
		       mdname(mddev));
		return -EIO;
	}
L
Linus Torvalds 已提交
2908
	/*
2909 2910
	 * copy the already verified devices into our private RAID1
	 * bookkeeping area. [whatever we allocate in run(),
N
NeilBrown 已提交
2911
	 * should be freed in raid1_free()]
L
Linus Torvalds 已提交
2912
	 */
2913 2914 2915 2916
	if (mddev->private == NULL)
		conf = setup_conf(mddev);
	else
		conf = mddev->private;
L
Linus Torvalds 已提交
2917

2918 2919
	if (IS_ERR(conf))
		return PTR_ERR(conf);
L
Linus Torvalds 已提交
2920

2921
	if (mddev->queue)
2922 2923
		blk_queue_max_write_same_sectors(mddev->queue, 0);

N
NeilBrown 已提交
2924
	rdev_for_each(rdev, mddev) {
2925 2926
		if (!mddev->gendisk)
			continue;
2927 2928
		disk_stack_limits(mddev->gendisk, rdev->bdev,
				  rdev->data_offset << 9);
S
Shaohua Li 已提交
2929 2930
		if (blk_queue_discard(bdev_get_queue(rdev->bdev)))
			discard_supported = true;
L
Linus Torvalds 已提交
2931
	}
2932

2933 2934 2935 2936 2937 2938 2939 2940 2941 2942
	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;

2943
	if (mddev->recovery_cp != MaxSector)
N
NeilBrown 已提交
2944
		printk(KERN_NOTICE "md/raid1:%s: not clean"
2945 2946
		       " -- starting background reconstruction\n",
		       mdname(mddev));
2947
	printk(KERN_INFO
N
NeilBrown 已提交
2948
		"md/raid1:%s: active with %d out of %d mirrors\n",
2949
		mdname(mddev), mddev->raid_disks - mddev->degraded,
L
Linus Torvalds 已提交
2950
		mddev->raid_disks);
2951

L
Linus Torvalds 已提交
2952 2953 2954
	/*
	 * Ok, everything is just fine now
	 */
2955 2956 2957 2958
	mddev->thread = conf->thread;
	conf->thread = NULL;
	mddev->private = conf;

2959
	md_set_array_sectors(mddev, raid1_size(mddev, 0, 0));
L
Linus Torvalds 已提交
2960

2961
	if (mddev->queue) {
S
Shaohua Li 已提交
2962 2963 2964 2965 2966 2967
		if (discard_supported)
			queue_flag_set_unlocked(QUEUE_FLAG_DISCARD,
						mddev->queue);
		else
			queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD,
						  mddev->queue);
2968
	}
2969 2970

	ret =  md_integrity_register(mddev);
2971 2972
	if (ret) {
		md_unregister_thread(&mddev->thread);
N
NeilBrown 已提交
2973
		raid1_free(mddev, conf);
2974
	}
2975
	return ret;
L
Linus Torvalds 已提交
2976 2977
}

N
NeilBrown 已提交
2978
static void raid1_free(struct mddev *mddev, void *priv)
L
Linus Torvalds 已提交
2979
{
N
NeilBrown 已提交
2980
	struct r1conf *conf = priv;
2981

L
Linus Torvalds 已提交
2982 2983
	if (conf->r1bio_pool)
		mempool_destroy(conf->r1bio_pool);
2984
	kfree(conf->mirrors);
2985
	safe_put_page(conf->tmppage);
2986
	kfree(conf->poolinfo);
L
Linus Torvalds 已提交
2987 2988 2989
	kfree(conf);
}

2990
static int raid1_resize(struct mddev *mddev, sector_t sectors)
L
Linus Torvalds 已提交
2991 2992 2993 2994 2995 2996 2997 2998
{
	/* 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.
	 */
2999 3000 3001
	sector_t newsize = raid1_size(mddev, sectors, 0);
	if (mddev->external_size &&
	    mddev->array_sectors > newsize)
D
Dan Williams 已提交
3002
		return -EINVAL;
3003 3004 3005 3006 3007 3008
	if (mddev->bitmap) {
		int ret = bitmap_resize(mddev->bitmap, newsize, 0, 0);
		if (ret)
			return ret;
	}
	md_set_array_sectors(mddev, newsize);
3009
	set_capacity(mddev->gendisk, mddev->array_sectors);
3010
	revalidate_disk(mddev->gendisk);
D
Dan Williams 已提交
3011
	if (sectors > mddev->dev_sectors &&
3012
	    mddev->recovery_cp > mddev->dev_sectors) {
A
Andre Noll 已提交
3013
		mddev->recovery_cp = mddev->dev_sectors;
L
Linus Torvalds 已提交
3014 3015
		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
	}
D
Dan Williams 已提交
3016
	mddev->dev_sectors = sectors;
3017
	mddev->resync_max_sectors = sectors;
L
Linus Torvalds 已提交
3018 3019 3020
	return 0;
}

3021
static int raid1_reshape(struct mddev *mddev)
L
Linus Torvalds 已提交
3022 3023 3024 3025 3026 3027 3028 3029
{
	/* 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.
3030 3031 3032
	 *
	 * 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
3033 3034 3035
	 */
	mempool_t *newpool, *oldpool;
	struct pool_info *newpoolinfo;
3036
	struct raid1_info *newmirrors;
3037
	struct r1conf *conf = mddev->private;
3038
	int cnt, raid_disks;
3039
	unsigned long flags;
3040
	int d, d2, err;
L
Linus Torvalds 已提交
3041

3042
	/* Cannot change chunk_size, layout, or level */
3043
	if (mddev->chunk_sectors != mddev->new_chunk_sectors ||
3044 3045
	    mddev->layout != mddev->new_layout ||
	    mddev->level != mddev->new_level) {
3046
		mddev->new_chunk_sectors = mddev->chunk_sectors;
3047 3048 3049 3050 3051
		mddev->new_layout = mddev->layout;
		mddev->new_level = mddev->level;
		return -EINVAL;
	}

3052 3053 3054
	err = md_allow_write(mddev);
	if (err)
		return err;
3055

3056 3057
	raid_disks = mddev->raid_disks + mddev->delta_disks;

3058 3059 3060 3061 3062 3063
	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
3064
			return -EBUSY;
3065
	}
L
Linus Torvalds 已提交
3066 3067 3068 3069 3070

	newpoolinfo = kmalloc(sizeof(*newpoolinfo), GFP_KERNEL);
	if (!newpoolinfo)
		return -ENOMEM;
	newpoolinfo->mddev = mddev;
3071
	newpoolinfo->raid_disks = raid_disks * 2;
L
Linus Torvalds 已提交
3072 3073 3074 3075 3076 3077 3078

	newpool = mempool_create(NR_RAID1_BIOS, r1bio_pool_alloc,
				 r1bio_pool_free, newpoolinfo);
	if (!newpool) {
		kfree(newpoolinfo);
		return -ENOMEM;
	}
3079
	newmirrors = kzalloc(sizeof(struct raid1_info) * raid_disks * 2,
3080
			     GFP_KERNEL);
L
Linus Torvalds 已提交
3081 3082 3083 3084 3085 3086
	if (!newmirrors) {
		kfree(newpoolinfo);
		mempool_destroy(newpool);
		return -ENOMEM;
	}

3087
	freeze_array(conf, 0);
L
Linus Torvalds 已提交
3088 3089 3090 3091

	/* ok, everything is stopped */
	oldpool = conf->r1bio_pool;
	conf->r1bio_pool = newpool;
3092

3093
	for (d = d2 = 0; d < conf->raid_disks; d++) {
3094
		struct md_rdev *rdev = conf->mirrors[d].rdev;
3095
		if (rdev && rdev->raid_disk != d2) {
3096
			sysfs_unlink_rdev(mddev, rdev);
3097
			rdev->raid_disk = d2;
3098 3099
			sysfs_unlink_rdev(mddev, rdev);
			if (sysfs_link_rdev(mddev, rdev))
3100
				printk(KERN_WARNING
3101 3102
				       "md/raid1:%s: cannot register rd%d\n",
				       mdname(mddev), rdev->raid_disk);
3103
		}
3104 3105 3106
		if (rdev)
			newmirrors[d2++].rdev = rdev;
	}
L
Linus Torvalds 已提交
3107 3108 3109 3110 3111
	kfree(conf->mirrors);
	conf->mirrors = newmirrors;
	kfree(conf->poolinfo);
	conf->poolinfo = newpoolinfo;

3112
	spin_lock_irqsave(&conf->device_lock, flags);
L
Linus Torvalds 已提交
3113
	mddev->degraded += (raid_disks - conf->raid_disks);
3114
	spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
3115
	conf->raid_disks = mddev->raid_disks = raid_disks;
3116
	mddev->delta_disks = 0;
L
Linus Torvalds 已提交
3117

3118
	unfreeze_array(conf);
L
Linus Torvalds 已提交
3119 3120 3121 3122 3123 3124 3125 3126

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

	mempool_destroy(oldpool);
	return 0;
}

3127
static void raid1_quiesce(struct mddev *mddev, int state)
3128
{
3129
	struct r1conf *conf = mddev->private;
3130 3131

	switch(state) {
3132 3133 3134
	case 2: /* wake for suspend */
		wake_up(&conf->wait_barrier);
		break;
3135
	case 1:
3136
		freeze_array(conf, 0);
3137
		break;
3138
	case 0:
3139
		unfreeze_array(conf);
3140 3141 3142 3143
		break;
	}
}

3144
static void *raid1_takeover(struct mddev *mddev)
3145 3146 3147 3148 3149
{
	/* raid1 can take over:
	 *  raid5 with 2 devices, any layout or chunk size
	 */
	if (mddev->level == 5 && mddev->raid_disks == 2) {
3150
		struct r1conf *conf;
3151 3152 3153 3154 3155
		mddev->new_level = 1;
		mddev->new_layout = 0;
		mddev->new_chunk_sectors = 0;
		conf = setup_conf(mddev);
		if (!IS_ERR(conf))
3156 3157
			/* Array must appear to be quiesced */
			conf->array_frozen = 1;
3158 3159 3160 3161
		return conf;
	}
	return ERR_PTR(-EINVAL);
}
L
Linus Torvalds 已提交
3162

3163
static struct md_personality raid1_personality =
L
Linus Torvalds 已提交
3164 3165
{
	.name		= "raid1",
3166
	.level		= 1,
L
Linus Torvalds 已提交
3167 3168 3169
	.owner		= THIS_MODULE,
	.make_request	= make_request,
	.run		= run,
N
NeilBrown 已提交
3170
	.free		= raid1_free,
L
Linus Torvalds 已提交
3171 3172 3173 3174 3175 3176 3177
	.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,
3178
	.size		= raid1_size,
3179
	.check_reshape	= raid1_reshape,
3180
	.quiesce	= raid1_quiesce,
3181
	.takeover	= raid1_takeover,
3182
	.congested	= raid1_congested,
3183
	.mergeable_bvec	= raid1_mergeable_bvec,
L
Linus Torvalds 已提交
3184 3185 3186 3187
};

static int __init raid_init(void)
{
3188
	return register_md_personality(&raid1_personality);
L
Linus Torvalds 已提交
3189 3190 3191 3192
}

static void raid_exit(void)
{
3193
	unregister_md_personality(&raid1_personality);
L
Linus Torvalds 已提交
3194 3195 3196 3197 3198
}

module_init(raid_init);
module_exit(raid_exit);
MODULE_LICENSE("GPL");
3199
MODULE_DESCRIPTION("RAID1 (mirroring) personality for MD");
L
Linus Torvalds 已提交
3200
MODULE_ALIAS("md-personality-3"); /* RAID1 */
3201
MODULE_ALIAS("md-raid1");
3202
MODULE_ALIAS("md-level-1");
3203 3204

module_param(max_queued_requests, int, S_IRUGO|S_IWUSR);