raid1.c 87.5 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
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>
Linus Torvalds's avatar
Linus Torvalds committed
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
 *
Linus Torvalds's avatar
Linus Torvalds committed
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

Linus Torvalds's avatar
Linus Torvalds committed
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;
Linus Torvalds's avatar
Linus Torvalds committed
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);
Linus Torvalds's avatar
Linus Torvalds committed
72

73
static void * r1bio_pool_alloc(gfp_t gfp_flags, void *data)
Linus Torvalds's avatar
Linus Torvalds committed
74 75
{
	struct pool_info *pi = data;
76
	int size = offsetof(struct r1bio, bios[pi->raid_disks]);
Linus Torvalds's avatar
Linus Torvalds committed
77 78

	/* allocate a r1bio with room for raid_disks entries in the bios array */
79
	return kzalloc(size, gfp_flags);
Linus Torvalds's avatar
Linus Torvalds committed
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
Linus Torvalds's avatar
Linus Torvalds committed
89 90
#define RESYNC_SECTORS (RESYNC_BLOCK_SIZE >> 9)
#define RESYNC_PAGES ((RESYNC_BLOCK_SIZE + PAGE_SIZE-1) / PAGE_SIZE)
91 92
#define RESYNC_WINDOW (RESYNC_BLOCK_SIZE * RESYNC_DEPTH)
#define RESYNC_WINDOW_SECTORS (RESYNC_WINDOW >> 9)
93 94
#define CLUSTER_RESYNC_WINDOW (16 * RESYNC_WINDOW)
#define CLUSTER_RESYNC_WINDOW_SECTORS (CLUSTER_RESYNC_WINDOW >> 9)
95
#define NEXT_NORMALIO_DISTANCE (3 * RESYNC_WINDOW_SECTORS)
Linus Torvalds's avatar
Linus Torvalds committed
96

97
static void * r1buf_pool_alloc(gfp_t gfp_flags, void *data)
Linus Torvalds's avatar
Linus Torvalds committed
98 99
{
	struct pool_info *pi = data;
100
	struct r1bio *r1_bio;
Linus Torvalds's avatar
Linus Torvalds committed
101
	struct bio *bio;
102
	int need_pages;
Linus Torvalds's avatar
Linus Torvalds committed
103 104 105
	int i, j;

	r1_bio = r1bio_pool_alloc(gfp_flags, pi);
106
	if (!r1_bio)
Linus Torvalds's avatar
Linus Torvalds committed
107 108 109 110 111 112
		return NULL;

	/*
	 * Allocate bios : 1 for reading, n-1 for writing
	 */
	for (j = pi->raid_disks ; j-- ; ) {
113
		bio = bio_kmalloc(gfp_flags, RESYNC_PAGES);
Linus Torvalds's avatar
Linus Torvalds committed
114 115 116 117 118 119
		if (!bio)
			goto out_free_bio;
		r1_bio->bios[j] = bio;
	}
	/*
	 * Allocate RESYNC_PAGES data pages and attach them to
120 121 122
	 * the first bio.
	 * If this is a user-requested check/repair, allocate
	 * RESYNC_PAGES for each bio.
Linus Torvalds's avatar
Linus Torvalds committed
123
	 */
124
	if (test_bit(MD_RECOVERY_REQUESTED, &pi->mddev->recovery))
125
		need_pages = pi->raid_disks;
126
	else
127 128
		need_pages = 1;
	for (j = 0; j < need_pages; j++) {
129
		bio = r1_bio->bios[j];
130
		bio->bi_vcnt = RESYNC_PAGES;
131

132
		if (bio_alloc_pages(bio, gfp_flags))
133
			goto out_free_pages;
134 135 136 137 138 139 140
	}
	/* 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;
Linus Torvalds's avatar
Linus Torvalds committed
141 142 143 144 145 146
	}

	r1_bio->master_bio = NULL;

	return r1_bio;

147 148 149 150 151 152 153 154
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);
	}

Linus Torvalds's avatar
Linus Torvalds committed
155
out_free_bio:
156
	while (++j < pi->raid_disks)
Linus Torvalds's avatar
Linus Torvalds committed
157 158 159 160 161 162 163 164
		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;
165
	int i,j;
166
	struct r1bio *r1bio = __r1_bio;
Linus Torvalds's avatar
Linus Torvalds committed
167

168 169 170 171 172
	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)
173
				safe_put_page(r1bio->bios[j]->bi_io_vec[i].bv_page);
174
		}
Linus Torvalds's avatar
Linus Torvalds committed
175 176 177 178 179 180
	for (i=0 ; i < pi->raid_disks; i++)
		bio_put(r1bio->bios[i]);

	r1bio_pool_free(r1bio, data);
}

181
static void put_all_bios(struct r1conf *conf, struct r1bio *r1_bio)
Linus Torvalds's avatar
Linus Torvalds committed
182 183 184
{
	int i;

185
	for (i = 0; i < conf->raid_disks * 2; i++) {
Linus Torvalds's avatar
Linus Torvalds committed
186
		struct bio **bio = r1_bio->bios + i;
187
		if (!BIO_SPECIAL(*bio))
Linus Torvalds's avatar
Linus Torvalds committed
188 189 190 191 192
			bio_put(*bio);
		*bio = NULL;
	}
}

193
static void free_r1bio(struct r1bio *r1_bio)
Linus Torvalds's avatar
Linus Torvalds committed
194
{
195
	struct r1conf *conf = r1_bio->mddev->private;
Linus Torvalds's avatar
Linus Torvalds committed
196 197 198 199 200

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

201
static void put_buf(struct r1bio *r1_bio)
Linus Torvalds's avatar
Linus Torvalds committed
202
{
203
	struct r1conf *conf = r1_bio->mddev->private;
204 205
	int i;

206
	for (i = 0; i < conf->raid_disks * 2; i++) {
207 208 209 210
		struct bio *bio = r1_bio->bios[i];
		if (bio->bi_end_io)
			rdev_dec_pending(conf->mirrors[i].rdev, r1_bio->mddev);
	}
Linus Torvalds's avatar
Linus Torvalds committed
211 212 213

	mempool_free(r1_bio, conf->r1buf_pool);

214
	lower_barrier(conf);
Linus Torvalds's avatar
Linus Torvalds committed
215 216
}

217
static void reschedule_retry(struct r1bio *r1_bio)
Linus Torvalds's avatar
Linus Torvalds committed
218 219
{
	unsigned long flags;
220
	struct mddev *mddev = r1_bio->mddev;
221
	struct r1conf *conf = mddev->private;
Linus Torvalds's avatar
Linus Torvalds committed
222 223 224

	spin_lock_irqsave(&conf->device_lock, flags);
	list_add(&r1_bio->retry_list, &conf->retry_list);
225
	conf->nr_queued ++;
Linus Torvalds's avatar
Linus Torvalds committed
226 227
	spin_unlock_irqrestore(&conf->device_lock, flags);

228
	wake_up(&conf->wait_barrier);
Linus Torvalds's avatar
Linus Torvalds committed
229 230 231 232 233 234 235 236
	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.
 */
237
static void call_bio_endio(struct r1bio *r1_bio)
238 239 240
{
	struct bio *bio = r1_bio->master_bio;
	int done;
241
	struct r1conf *conf = r1_bio->mddev->private;
242
	sector_t start_next_window = r1_bio->start_next_window;
243
	sector_t bi_sector = bio->bi_iter.bi_sector;
244 245 246 247 248 249 250

	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);
251 252 253 254 255
		/*
		 * make_request() might be waiting for
		 * bi_phys_segments to decrease
		 */
		wake_up(&conf->wait_barrier);
256 257 258 259
	} else
		done = 1;

	if (!test_bit(R1BIO_Uptodate, &r1_bio->state))
260 261
		bio->bi_error = -EIO;

262
	if (done) {
263
		bio_endio(bio);
264 265 266 267
		/*
		 * Wake up any possible resync thread that waits for the device
		 * to go idle.
		 */
268
		allow_barrier(conf, start_next_window, bi_sector);
269 270 271
	}
}

272
static void raid_end_bio_io(struct r1bio *r1_bio)
Linus Torvalds's avatar
Linus Torvalds committed
273 274 275
{
	struct bio *bio = r1_bio->master_bio;

276 277
	/* if nobody has done the final endio yet, do it now */
	if (!test_and_set_bit(R1BIO_Returned, &r1_bio->state)) {
278 279
		pr_debug("raid1: sync end %s on sectors %llu-%llu\n",
			 (bio_data_dir(bio) == WRITE) ? "write" : "read",
280 281
			 (unsigned long long) bio->bi_iter.bi_sector,
			 (unsigned long long) bio_end_sector(bio) - 1);
282

283
		call_bio_endio(r1_bio);
284
	}
Linus Torvalds's avatar
Linus Torvalds committed
285 286 287 288 289 290
	free_r1bio(r1_bio);
}

/*
 * Update disk head position estimator based on IRQ completion info.
 */
291
static inline void update_head_pos(int disk, struct r1bio *r1_bio)
Linus Torvalds's avatar
Linus Torvalds committed
292
{
293
	struct r1conf *conf = r1_bio->mddev->private;
Linus Torvalds's avatar
Linus Torvalds committed
294 295 296 297 298

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

299 300 301
/*
 * Find the disk number which triggered given bio
 */
302
static int find_bio_disk(struct r1bio *r1_bio, struct bio *bio)
303 304
{
	int mirror;
305 306
	struct r1conf *conf = r1_bio->mddev->private;
	int raid_disks = conf->raid_disks;
307

308
	for (mirror = 0; mirror < raid_disks * 2; mirror++)
309 310 311
		if (r1_bio->bios[mirror] == bio)
			break;

312
	BUG_ON(mirror == raid_disks * 2);
313 314 315 316 317
	update_head_pos(mirror, r1_bio);

	return mirror;
}

318
static void raid1_end_read_request(struct bio *bio)
Linus Torvalds's avatar
Linus Torvalds committed
319
{
320
	int uptodate = !bio->bi_error;
321
	struct r1bio *r1_bio = bio->bi_private;
Linus Torvalds's avatar
Linus Torvalds committed
322
	int mirror;
323
	struct r1conf *conf = r1_bio->mddev->private;
Linus Torvalds's avatar
Linus Torvalds committed
324 325 326 327 328

	mirror = r1_bio->read_disk;
	/*
	 * this branch is our 'one mirror IO has finished' event handler:
	 */
329 330
	update_head_pos(mirror, r1_bio);

331 332 333 334 335 336
	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"
Linus Torvalds's avatar
Linus Torvalds committed
337
		 */
338 339 340 341
		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 &&
342
		     test_bit(In_sync, &conf->mirrors[mirror].rdev->flags)))
343 344 345
			uptodate = 1;
		spin_unlock_irqrestore(&conf->device_lock, flags);
	}
Linus Torvalds's avatar
Linus Torvalds committed
346

347
	if (uptodate) {
Linus Torvalds's avatar
Linus Torvalds committed
348
		raid_end_bio_io(r1_bio);
349 350
		rdev_dec_pending(conf->mirrors[mirror].rdev, conf->mddev);
	} else {
Linus Torvalds's avatar
Linus Torvalds committed
351 352 353 354
		/*
		 * oops, read error:
		 */
		char b[BDEVNAME_SIZE];
355 356 357 358 359 360 361
		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);
362
		set_bit(R1BIO_ReadError, &r1_bio->state);
Linus Torvalds's avatar
Linus Torvalds committed
363
		reschedule_retry(r1_bio);
364
		/* don't drop the reference on read_disk yet */
Linus Torvalds's avatar
Linus Torvalds committed
365 366 367
	}
}

368
static void close_write(struct r1bio *r1_bio)
369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386
{
	/* 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);
}

387
static void r1_bio_write_done(struct r1bio *r1_bio)
388
{
389 390 391 392 393 394 395
	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);
396 397 398 399
		if (test_bit(R1BIO_MadeGood, &r1_bio->state))
			reschedule_retry(r1_bio);
		else
			raid_end_bio_io(r1_bio);
400 401 402
	}
}

403
static void raid1_end_write_request(struct bio *bio)
Linus Torvalds's avatar
Linus Torvalds committed
404
{
405
	struct r1bio *r1_bio = bio->bi_private;
406
	int mirror, behind = test_bit(R1BIO_BehindIO, &r1_bio->state);
407
	struct r1conf *conf = r1_bio->mddev->private;
408
	struct bio *to_put = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
409

410
	mirror = find_bio_disk(r1_bio, bio);
Linus Torvalds's avatar
Linus Torvalds committed
411

412 413 414
	/*
	 * 'one mirror IO has finished' event handler:
	 */
415
	if (bio->bi_error) {
416 417
		set_bit(WriteErrorSeen,
			&conf->mirrors[mirror].rdev->flags);
418 419 420 421 422
		if (!test_and_set_bit(WantReplacement,
				      &conf->mirrors[mirror].rdev->flags))
			set_bit(MD_RECOVERY_NEEDED, &
				conf->mddev->recovery);

423
		set_bit(R1BIO_WriteError, &r1_bio->state);
424
	} else {
Linus Torvalds's avatar
Linus Torvalds committed
425
		/*
426 427 428 429 430 431 432 433
		 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
434
		 */
435 436 437
		sector_t first_bad;
		int bad_sectors;

438 439
		r1_bio->bios[mirror] = NULL;
		to_put = bio;
440 441 442 443 444 445 446 447 448 449 450
		/*
		 * 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);
451

452 453 454 455 456 457 458 459 460
		/* 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);
		}
	}

461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476
	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;
477 478
				pr_debug("raid1: behind end write sectors"
					 " %llu-%llu\n",
479 480
					 (unsigned long long) mbio->bi_iter.bi_sector,
					 (unsigned long long) bio_end_sector(mbio) - 1);
481
				call_bio_endio(r1_bio);
482 483 484
			}
		}
	}
485 486 487
	if (r1_bio->bios[mirror] == NULL)
		rdev_dec_pending(conf->mirrors[mirror].rdev,
				 conf->mddev);
488

Linus Torvalds's avatar
Linus Torvalds committed
489 490 491 492
	/*
	 * Let's see if all mirrored write operations have finished
	 * already.
	 */
493
	r1_bio_write_done(r1_bio);
494

495 496
	if (to_put)
		bio_put(to_put);
Linus Torvalds's avatar
Linus Torvalds committed
497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512
}

/*
 * 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.
 */
513
static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sectors)
Linus Torvalds's avatar
Linus Torvalds committed
514
{
515
	const sector_t this_sector = r1_bio->sector;
516 517
	int sectors;
	int best_good_sectors;
518 519
	int best_disk, best_dist_disk, best_pending_disk;
	int has_nonrot_disk;
520
	int disk;
NeilBrown's avatar
NeilBrown committed
521
	sector_t best_dist;
522
	unsigned int min_pending;
523
	struct md_rdev *rdev;
524
	int choose_first;
525
	int choose_next_idle;
Linus Torvalds's avatar
Linus Torvalds committed
526 527 528

	rcu_read_lock();
	/*
529
	 * Check if we can balance. We can balance on the whole
Linus Torvalds's avatar
Linus Torvalds committed
530 531 532 533
	 * device if no resync is going on, or below the resync window.
	 * We take the first readable disk when above the resync window.
	 */
 retry:
534
	sectors = r1_bio->sectors;
NeilBrown's avatar
NeilBrown committed
535
	best_disk = -1;
536
	best_dist_disk = -1;
NeilBrown's avatar
NeilBrown committed
537
	best_dist = MaxSector;
538 539
	best_pending_disk = -1;
	min_pending = UINT_MAX;
540
	best_good_sectors = 0;
541
	has_nonrot_disk = 0;
542
	choose_next_idle = 0;
543

544 545
	if ((conf->mddev->recovery_cp < this_sector + sectors) ||
	    (mddev_is_clustered(conf->mddev) &&
546
	    md_cluster_ops->area_resyncing(conf->mddev, READ, this_sector,
547 548 549 550
		    this_sector + sectors)))
		choose_first = 1;
	else
		choose_first = 0;
Linus Torvalds's avatar
Linus Torvalds committed
551

552
	for (disk = 0 ; disk < conf->raid_disks * 2 ; disk++) {
NeilBrown's avatar
NeilBrown committed
553
		sector_t dist;
554 555
		sector_t first_bad;
		int bad_sectors;
556
		unsigned int pending;
557
		bool nonrot;
558

559 560 561
		rdev = rcu_dereference(conf->mirrors[disk].rdev);
		if (r1_bio->bios[disk] == IO_BLOCKED
		    || rdev == NULL
NeilBrown's avatar
NeilBrown committed
562
		    || test_bit(Faulty, &rdev->flags))
563
			continue;
NeilBrown's avatar
NeilBrown committed
564 565
		if (!test_bit(In_sync, &rdev->flags) &&
		    rdev->recovery_offset < this_sector + sectors)
Linus Torvalds's avatar
Linus Torvalds committed
566
			continue;
NeilBrown's avatar
NeilBrown committed
567 568 569
		if (test_bit(WriteMostly, &rdev->flags)) {
			/* Don't balance among write-mostly, just
			 * use the first as a last resort */
570
			if (best_dist_disk < 0) {
571 572
				if (is_badblock(rdev, this_sector, sectors,
						&first_bad, &bad_sectors)) {
573
					if (first_bad <= this_sector)
574 575 576 577 578
						/* Cannot use this */
						continue;
					best_good_sectors = first_bad - this_sector;
				} else
					best_good_sectors = sectors;
579 580
				best_dist_disk = disk;
				best_pending_disk = disk;
581
			}
NeilBrown's avatar
NeilBrown committed
582 583 584 585 586
			continue;
		}
		/* This is a reasonable device to use.  It might
		 * even be best.
		 */
587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615
		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;

616 617
		nonrot = blk_queue_nonrot(bdev_get_queue(rdev->bdev));
		has_nonrot_disk |= nonrot;
618
		pending = atomic_read(&rdev->nr_pending);
NeilBrown's avatar
NeilBrown committed
619
		dist = abs(this_sector - conf->mirrors[disk].head_position);
620
		if (choose_first) {
NeilBrown's avatar
NeilBrown committed
621
			best_disk = disk;
Linus Torvalds's avatar
Linus Torvalds committed
622 623
			break;
		}
624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661
		/* 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;
662 663 664 665 666 667

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

NeilBrown's avatar
NeilBrown committed
668 669
		if (dist < best_dist) {
			best_dist = dist;
670
			best_dist_disk = disk;
Linus Torvalds's avatar
Linus Torvalds committed
671
		}
672
	}
Linus Torvalds's avatar
Linus Torvalds committed
673

674 675 676 677 678 679 680 681 682 683 684 685 686
	/*
	 * 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;
	}

NeilBrown's avatar
NeilBrown committed
687 688
	if (best_disk >= 0) {
		rdev = rcu_dereference(conf->mirrors[best_disk].rdev);
689 690 691
		if (!rdev)
			goto retry;
		atomic_inc(&rdev->nr_pending);
NeilBrown's avatar
NeilBrown committed
692
		if (test_bit(Faulty, &rdev->flags)) {
Linus Torvalds's avatar
Linus Torvalds committed
693 694 695
			/* cannot risk returning a device that failed
			 * before we inc'ed nr_pending
			 */
696
			rdev_dec_pending(rdev, conf->mddev);
Linus Torvalds's avatar
Linus Torvalds committed
697 698
			goto retry;
		}
699
		sectors = best_good_sectors;
700 701 702 703

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

704
		conf->mirrors[best_disk].next_seq_sect = this_sector + sectors;
Linus Torvalds's avatar
Linus Torvalds committed
705 706
	}
	rcu_read_unlock();
707
	*max_sectors = sectors;
Linus Torvalds's avatar
Linus Torvalds committed
708

NeilBrown's avatar
NeilBrown committed
709
	return best_disk;
Linus Torvalds's avatar
Linus Torvalds committed
710 711
}

712
static int raid1_congested(struct mddev *mddev, int bits)
713
{
714
	struct r1conf *conf = mddev->private;
715 716
	int i, ret = 0;

717
	if ((bits & (1 << WB_async_congested)) &&
718 719 720
	    conf->pending_count >= max_queued_requests)
		return 1;

721
	rcu_read_lock();
722
	for (i = 0; i < conf->raid_disks * 2; i++) {
723
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
724
		if (rdev && !test_bit(Faulty, &rdev->flags)) {
725
			struct request_queue *q = bdev_get_queue(rdev->bdev);
726

727 728
			BUG_ON(!q);

729 730 731
			/* Note the '|| 1' - when read_balance prefers
			 * non-congested targets, it can be removed
			 */
732
			if ((bits & (1 << WB_async_congested)) || 1)
733
				ret |= bdi_congested(q->backing_dev_info, bits);
734
			else
735
				ret &= bdi_congested(q->backing_dev_info, bits);
736 737 738 739 740 741
		}
	}
	rcu_read_unlock();
	return ret;
}

742
static void flush_pending_writes(struct r1conf *conf)
743 744 745 746 747 748 749 750 751
{
	/* 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);
752
		conf->pending_count = 0;
753 754 755 756
		spin_unlock_irq(&conf->device_lock);
		/* flush any pending bitmap writes to
		 * disk before proceeding w/ I/O */
		bitmap_unplug(conf->mddev->bitmap);
757
		wake_up(&conf->wait_barrier);
758 759 760 761

		while (bio) { /* submit pending writes */
			struct bio *next = bio->bi_next;
			bio->bi_next = NULL;
Shaohua Li's avatar
Shaohua Li committed
762 763 764
			if (unlikely((bio->bi_rw & REQ_DISCARD) &&
			    !blk_queue_discard(bdev_get_queue(bio->bi_bdev))))
				/* Just ignore it */
765
				bio_endio(bio);
Shaohua Li's avatar
Shaohua Li committed
766 767
			else
				generic_make_request(bio);
768 769 770 771
			bio = next;
		}
	} else
		spin_unlock_irq(&conf->device_lock);
772 773
}

774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793
/* 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.
Linus Torvalds's avatar
Linus Torvalds committed
794
 */
795
static void raise_barrier(struct r1conf *conf, sector_t sector_nr)
Linus Torvalds's avatar
Linus Torvalds committed
796 797
{
	spin_lock_irq(&conf->resync_lock);
798 799 800

	/* Wait until no block IO is waiting */
	wait_event_lock_irq(conf->wait_barrier, !conf->nr_waiting,
801
			    conf->resync_lock);
802 803 804

	/* block any new IO from starting */
	conf->barrier++;
805
	conf->next_resync = sector_nr;
806

807 808 809 810 811 812 813
	/* 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.
814
	 * D: while there are any active requests in the current window.
815
	 */
816
	wait_event_lock_irq(conf->wait_barrier,
817
			    !conf->array_frozen &&
818
			    conf->barrier < RESYNC_DEPTH &&
819
			    conf->current_window_requests == 0 &&
820 821
			    (conf->start_next_window >=
			     conf->next_resync + RESYNC_SECTORS),
822
			    conf->resync_lock);
823

824
	conf->nr_pending++;
825 826 827
	spin_unlock_irq(&conf->resync_lock);
}

828
static void lower_barrier(struct r1conf *conf)
829 830
{
	unsigned long flags;
831
	BUG_ON(conf->barrier <= 0);
832 833
	spin_lock_irqsave(&conf->resync_lock, flags);
	conf->barrier--;
834
	conf->nr_pending--;
835 836 837 838
	spin_unlock_irqrestore(&conf->resync_lock, flags);
	wake_up(&conf->wait_barrier);
}

839
static bool need_to_wait_for_sync(struct r1conf *conf, struct bio *bio)
840
{
841 842 843 844 845
	bool wait = false;

	if (conf->array_frozen || !bio)
		wait = true;
	else if (conf->barrier && bio_data_dir(bio) == WRITE) {
846 847 848 849
		if ((conf->mddev->curr_resync_completed
		     >= bio_end_sector(bio)) ||
		    (conf->next_resync + NEXT_NORMALIO_DISTANCE
		     <= bio->bi_iter.bi_sector))
850 851 852 853 854 855 856 857 858 859 860 861
			wait = false;
		else
			wait = true;
	}

	return wait;
}

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

862
	spin_lock_irq(&conf->resync_lock);
863
	if (need_to_wait_for_sync(conf, bio)) {
864
		conf->nr_waiting++;
865 866 867 868
		/* Wait for the barrier to drop.
		 * However if there are already pending
		 * requests (preventing the barrier from
		 * rising completely), and the
869
		 * per-process bio queue isn't empty,
870
		 * then don't wait, as we need to empty
871 872
		 * that queue to allow conf->start_next_window
		 * to increase.
873 874
		 */
		wait_event_lock_irq(conf->wait_barrier,
875 876
				    !conf->array_frozen &&
				    (!conf->barrier ||
877 878 879
				     ((conf->start_next_window <
				       conf->next_resync + RESYNC_SECTORS) &&
				      current->bio_list &&
880 881
				     (!bio_list_empty(&current->bio_list[0]) ||
				      !bio_list_empty(&current->bio_list[1])))),
882
				    conf->resync_lock);
883
		conf->nr_waiting--;
Linus Torvalds's avatar
Linus Torvalds committed
884
	}
885 886

	if (bio && bio_data_dir(bio) == WRITE) {
887
		if (bio->bi_iter.bi_sector >= conf->next_resync) {
888 889 890 891 892 893
			if (conf->start_next_window == MaxSector)
				conf->start_next_window =
					conf->next_resync +
					NEXT_NORMALIO_DISTANCE;

			if ((conf->start_next_window + NEXT_NORMALIO_DISTANCE)
894
			    <= bio->bi_iter.bi_sector)
895 896 897 898
				conf->next_window_requests++;
			else
				conf->current_window_requests++;
			sector = conf->start_next_window;
899
		}
900 901
	}

902
	conf->nr_pending++;
Linus Torvalds's avatar
Linus Torvalds committed
903
	spin_unlock_irq(&conf->resync_lock);
904
	return sector;
Linus Torvalds's avatar
Linus Torvalds committed
905 906
}

907 908
static void allow_barrier(struct r1conf *conf, sector_t start_next_window,
			  sector_t bi_sector)
909 910
{
	unsigned long flags;
911

912 913
	spin_lock_irqsave(&conf->resync_lock, flags);
	conf->nr_pending--;
914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934
	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;
		}
	}
935 936 937 938
	spin_unlock_irqrestore(&conf->resync_lock, flags);
	wake_up(&conf->wait_barrier);
}

939
static void freeze_array(struct r1conf *conf, int extra)
940 941 942
{
	/* stop syncio and normal IO and wait for everything to
	 * go quite.
943
	 * We wait until nr_pending match nr_queued+extra
944 945 946 947
	 * 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.
948
	 * Thus the number queued (nr_queued) plus this request (extra)
949 950
	 * must match the number of pending IOs (nr_pending) before
	 * we continue.
951 952
	 */
	spin_lock_irq(&conf->resync_lock);
953
	conf->array_frozen = 1;
954
	wait_event_lock_irq_cmd(conf->wait_barrier,
955
				conf->nr_pending == conf->nr_queued+extra,
956 957
				conf->resync_lock,
				flush_pending_writes(conf));
958 959
	spin_unlock_irq(&conf->resync_lock);
}
960
static void unfreeze_array(struct r1conf *conf)
961 962 963
{
	/* reverse the effect of the freeze */
	spin_lock_irq(&conf->resync_lock);
964
	conf->array_frozen = 0;
965 966 967 968
	wake_up(&conf->wait_barrier);
	spin_unlock_irq(&conf->resync_lock);
}

969
/* duplicate the data pages for behind I/O
970
 */
971
static void alloc_behind_pages(struct bio *bio, struct r1bio *r1_bio)
972 973 974
{
	int i;
	struct bio_vec *bvec;
975
	struct bio_vec *bvecs = kzalloc(bio->bi_vcnt * sizeof(struct bio_vec),
976
					GFP_NOIO);
977
	if (unlikely(!bvecs))
978
		return;
979

980
	bio_for_each_segment_all(bvec, bio, i) {
981 982 983
		bvecs[i] = *bvec;
		bvecs[i].bv_page = alloc_page(GFP_NOIO);
		if (unlikely(!bvecs[i].bv_page))
984
			goto do_sync_io;
985 986 987
		memcpy(kmap(bvecs[i].bv_page) + bvec->bv_offset,
		       kmap(bvec->bv_page) + bvec->bv_offset, bvec->bv_len);
		kunmap(bvecs[i].bv_page);
988 989
		kunmap(bvec->bv_page);
	}
990
	r1_bio->behind_bvecs = bvecs;
991 992 993
	r1_bio->behind_page_count = bio->bi_vcnt;
	set_bit(R1BIO_BehindIO, &r1_bio->state);
	return;
994 995

do_sync_io:
996
	for (i = 0; i < bio->bi_vcnt; i++)
997 998 999
		if (bvecs[i].bv_page)
			put_page(bvecs[i].bv_page);
	kfree(bvecs);
1000 1001
	pr_debug("%dB behind alloc failed, doing sync I/O\n",
		 bio->bi_iter.bi_size);
1002 1003
}

1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017
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;

1018
	if (from_schedule || current->bio_list) {
1019 1020 1021 1022
		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);
1023
		wake_up(&conf->wait_barrier);
1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036
		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;
1037 1038 1039
		if (unlikely((bio->bi_rw & REQ_DISCARD) &&
		    !blk_queue_discard(bdev_get_queue(bio->bi_bdev))))
			/* Just ignore it */
1040
			bio_endio(bio);
1041 1042
		else
			generic_make_request(bio);
1043 1044 1045 1046 1047
		bio = next;
	}
	kfree(plug);
}

1048
static void make_request(struct mddev *mddev, struct bio * bio)
Linus Torvalds's avatar
Linus Torvalds committed
1049
{
1050
	struct r1conf *conf = mddev->private;
1051
	struct raid1_info *mirror;
1052
	struct r1bio *r1_bio;
Linus Torvalds's avatar
Linus Torvalds committed
1053
	struct bio *read_bio;
1054
	int i, disks;
1055
	struct bitmap *bitmap;
1056
	unsigned long flags;
1057
	const int rw = bio_data_dir(bio);
1058
	const unsigned long do_sync = (bio->bi_rw & REQ_SYNC);
1059
	const unsigned long do_flush_fua = (bio->bi_rw & (REQ_FLUSH | REQ_FUA));
Shaohua Li's avatar
Shaohua Li committed
1060 1061
	const unsigned long do_discard = (bio->bi_rw
					  & (REQ_DISCARD | REQ_SECURE));
1062
	const unsigned long do_same = (bio->bi_rw & REQ_WRITE_SAME);
1063
	struct md_rdev *blocked_rdev;
1064 1065
	struct blk_plug_cb *cb;
	struct raid1_plug_cb *plug = NULL;
1066 1067 1068
	int first_clone;
	int sectors_handled;
	int max_sectors;
1069
	sector_t start_next_window;
1070

Linus Torvalds's avatar
Linus Torvalds committed
1071 1072 1073 1074 1075
	/*
	 * 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.
	 */
1076

1077 1078
	md_write_start(mddev, bio); /* wait on superblock update early */

1079
	if (bio_data_dir(bio) == WRITE &&
1080 1081 1082
	    ((bio_end_sector(bio) > mddev->suspend_lo &&
	    bio->bi_iter.bi_sector < mddev->suspend_hi) ||
	    (mddev_is_clustered(mddev) &&
1083 1084
	     md_cluster_ops->area_resyncing(mddev, WRITE,
		     bio->bi_iter.bi_sector, bio_end_sector(bio))))) {
1085 1086 1087 1088 1089 1090
		/* As the suspend_* range is controlled by
		 * userspace, we want an interruptible
		 * wait.
		 */
		DEFINE_WAIT(w);
		for (;;) {
1091
			sigset_t full, old;
1092 1093
			prepare_to_wait(&conf->wait_barrier,
					&w, TASK_INTERRUPTIBLE);
1094
			if (bio_end_sector(bio) <= mddev->suspend_lo ||
1095 1096
			    bio->bi_iter.bi_sector >= mddev->suspend_hi ||
			    (mddev_is_clustered(mddev) &&
1097
			     !md_cluster_ops->area_resyncing(mddev, WRITE,
1098
				     bio->bi_iter.bi_sector, bio_end_sector(bio))))
1099
				break;
1100 1101
			sigfillset(&full);
			sigprocmask(SIG_BLOCK, &full, &old);
1102
			schedule();
1103
			sigprocmask(SIG_SETMASK, &old, NULL);
1104 1105 1106
		}
		finish_wait(&conf->wait_barrier, &w);
	}
1107

1108
	start_next_window = wait_barrier(conf, bio);
Linus Torvalds's avatar
Linus Torvalds committed
1109

1110 1111
	bitmap = mddev->bitmap;

Linus Torvalds's avatar
Linus Torvalds committed
1112 1113 1114 1115 1116 1117 1118 1119
	/*
	 * 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;
1120
	r1_bio->sectors = bio_sectors(bio);
1121
	r1_bio->state = 0;
Linus Torvalds's avatar
Linus Torvalds committed
1122
	r1_bio->mddev = mddev;
1123
	r1_bio->sector = bio->bi_iter.bi_sector;
Linus Torvalds's avatar
Linus Torvalds committed
1124

1125 1126 1127 1128 1129 1130 1131 1132
	/* 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;
1133
	bio_clear_flag(bio, BIO_SEG_VALID);
1134

1135
	if (rw == READ) {
Linus Torvalds's avatar
Linus Torvalds committed
1136 1137 1138
		/*
		 * read balancing logic:
		 */
1139 1140 1141 1142
		int rdisk;

read_again:
		rdisk = read_balance(conf, r1_bio, &max_sectors);
Linus Torvalds's avatar
Linus Torvalds committed
1143 1144 1145 1146

		if (rdisk < 0) {
			/* couldn't find anywhere to read from */
			raid_end_bio_io(r1_bio);
1147
			return;
Linus Torvalds's avatar
Linus Torvalds committed
1148 1149 1150
		}
		mirror = conf->mirrors + rdisk;

1151 1152 1153 1154 1155 1156 1157 1158 1159
		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);
		}
Linus Torvalds's avatar
Linus Torvalds committed
1160
		r1_bio->read_disk = rdisk;
1161
		r1_bio->start_next_window = 0;
Linus Torvalds's avatar
Linus Torvalds committed
1162

1163
		read_bio = bio_clone_mddev(bio, GFP_NOIO, mddev);
1164
		bio_trim(read_bio, r1_bio->sector - bio->bi_iter.bi_sector,
1165
			 max_sectors);
Linus Torvalds's avatar
Linus Torvalds committed
1166 1167 1168

		r1_bio->bios[rdisk] = read_bio;

1169 1170
		read_bio->bi_iter.bi_sector = r1_bio->sector +
			mirror->rdev->data_offset;
Linus Torvalds's avatar
Linus Torvalds committed
1171 1172
		read_bio->bi_bdev = mirror->rdev->bdev;
		read_bio->bi_end_io = raid1_end_read_request;
1173
		read_bio->bi_rw = READ | do_sync;
Linus Torvalds's avatar
Linus Torvalds committed
1174 1175
		read_bio->bi_private = r1_bio;

1176 1177 1178 1179 1180 1181
		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
1182
					   - bio->bi_iter.bi_sector);
1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199
			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;
1200
			r1_bio->sectors = bio_sectors(bio) - sectors_handled;
1201 1202
			r1_bio->state = 0;
			r1_bio->mddev = mddev;
1203 1204
			r1_bio->sector = bio->bi_iter.bi_sector +
				sectors_handled;
1205 1206 1207
			goto read_again;
		} else
			generic_make_request(read_bio);
1208
		return;
Linus Torvalds's avatar
Linus Torvalds committed
1209 1210 1211 1212 1213
	}

	/*
	 * WRITE:
	 */
1214 1215 1216 1217 1218
	if (conf->pending_count >= max_queued_requests) {
		md_wakeup_thread(mddev->thread);
		wait_event(conf->wait_barrier,
			   conf->pending_count < max_queued_requests);
	}
1219
	/* first select target devices under rcu_lock and
Linus Torvalds's avatar
Linus Torvalds committed
1220 1221
	 * inc refcount on their rdev.  Record them by setting
	 * bios[x] to bio
1222 1223 1224 1225 1226 1227
	 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
1228
	 */
1229

1230
	disks = conf->raid_disks * 2;
1231
 retry_write:
1232
	r1_bio->start_next_window = start_next_window;
1233
	blocked_rdev = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
1234
	rcu_read_lock();
1235
	max_sectors = r1_bio->sectors;
Linus Torvalds's avatar
Linus Torvalds committed
1236
	for (i = 0;  i < disks; i++) {
1237
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
1238 1239 1240 1241 1242
		if (rdev && unlikely(test_bit(Blocked, &rdev->flags))) {
			atomic_inc(&rdev->nr_pending);
			blocked_rdev = rdev;
			break;
		}
1243
		r1_bio->bios[i] = NULL;
1244
		if (!rdev || test_bit(Faulty, &rdev->flags)) {
1245 1246
			if (i < conf->raid_disks)
				set_bit(R1BIO_Degraded, &r1_bio->state);
1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273
			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;
1274
				rdev_dec_pending(rdev, mddev);
1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285
				/* 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;
1286
			}
1287 1288 1289 1290 1291 1292 1293
			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;
Linus Torvalds's avatar
Linus Torvalds committed
1294 1295 1296
	}
	rcu_read_unlock();

1297 1298 1299
	if (unlikely(blocked_rdev)) {
		/* Wait for this device to become unblocked */
		int j;
1300
		sector_t old = start_next_window;
1301 1302 1303 1304

		for (j = 0; j < i; j++)
			if (r1_bio->bios[j])
				rdev_dec_pending(conf->mirrors[j].rdev, mddev);
1305
		r1_bio->state = 0;
1306
		allow_barrier(conf, start_next_window, bio->bi_iter.bi_sector);
1307
		md_wait_for_blocked_rdev(blocked_rdev, mddev);
1308 1309 1310 1311 1312 1313 1314 1315 1316 1317
		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);
1318 1319 1320
		goto retry_write;
	}

1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331
	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);
1332
	}
1333
	sectors_handled = r1_bio->sector + max_sectors - bio->bi_iter.bi_sector;
1334

1335
	atomic_set(&r1_bio->remaining, 1);
1336
	atomic_set(&r1_bio->behind_remaining, 0);
1337

1338
	first_clone = 1;
Linus Torvalds's avatar
Linus Torvalds committed
1339 1340 1341 1342 1343
	for (i = 0; i < disks; i++) {
		struct bio *mbio;
		if (!r1_bio->bios[i])
			continue;

1344
		mbio = bio_clone_mddev(bio, GFP_NOIO, mddev);
1345
		bio_trim(mbio, r1_bio->sector - bio->bi_iter.bi_sector, max_sectors);
1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363

		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;
		}
1364
		if (r1_bio->behind_bvecs) {
1365 1366 1367
			struct bio_vec *bvec;
			int j;

1368 1369
			/*
			 * We trimmed the bio, so _all is legit
1370
			 */
1371
			bio_for_each_segment_all(bvec, mbio, j)
1372
				bvec->bv_page = r1_bio->behind_bvecs[j].bv_page;
1373 1374 1375 1376
			if (test_bit(WriteMostly, &conf->mirrors[i].rdev->flags))
				atomic_inc(&r1_bio->behind_remaining);
		}

1377 1378
		r1_bio->bios[i] = mbio;

1379
		mbio->bi_iter.bi_sector	= (r1_bio->sector +
1380 1381 1382
				   conf->mirrors[i].rdev->data_offset);
		mbio->bi_bdev = conf->mirrors[i].rdev->bdev;
		mbio->bi_end_io	= raid1_end_write_request;
1383 1384
		mbio->bi_rw =
			WRITE | do_flush_fua | do_sync | do_discard | do_same;
1385 1386
		mbio->bi_private = r1_bio;

Linus Torvalds's avatar
Linus Torvalds committed
1387
		atomic_inc(&r1_bio->remaining);
1388 1389 1390 1391 1392 1393

		cb = blk_check_plugged(raid1_unplug, mddev, sizeof(*plug));
		if (cb)
			plug = container_of(cb, struct raid1_plug_cb, cb);
		else
			plug = NULL;
1394
		spin_lock_irqsave(&conf->device_lock, flags);
1395 1396 1397 1398 1399 1400 1401
		if (plug) {
			bio_list_add(&plug->pending, mbio);
			plug->pending_cnt++;
		} else {
			bio_list_add(&conf->pending_bio_list, mbio);
			conf->pending_count++;
		}
1402
		spin_unlock_irqrestore(&conf->device_lock, flags);
1403
		if (!plug)
NeilBrown's avatar
NeilBrown committed
1404
			md_wakeup_thread(mddev->thread);
Linus Torvalds's avatar
Linus Torvalds committed
1405
	}
1406 1407 1408
	/* Mustn't call r1_bio_write_done before this next test,
	 * as it could result in the bio being freed.
	 */
1409
	if (sectors_handled < bio_sectors(bio)) {
1410
		r1_bio_write_done(r1_bio);
1411 1412 1413 1414 1415
		/* 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;
1416
		r1_bio->sectors = bio_sectors(bio) - sectors_handled;
1417 1418
		r1_bio->state = 0;
		r1_bio->mddev = mddev;
1419
		r1_bio->sector = bio->bi_iter.bi_sector + sectors_handled;
1420 1421 1422
		goto retry_write;
	}

1423 1424 1425 1426
	r1_bio_write_done(r1_bio);

	/* In case raid1d snuck in to freeze_array */
	wake_up(&conf->wait_barrier);
Linus Torvalds's avatar
Linus Torvalds committed
1427 1428
}

1429
static void status(struct seq_file *seq, struct mddev *mddev)
Linus Torvalds's avatar
Linus Torvalds committed
1430
{
1431
	struct r1conf *conf = mddev->private;
Linus Torvalds's avatar
Linus Torvalds committed
1432 1433 1434
	int i;

	seq_printf(seq, " [%d/%d] [", conf->raid_disks,
1435
		   conf->raid_disks - mddev->degraded);
1436 1437
	rcu_read_lock();
	for (i = 0; i < conf->raid_disks; i++) {
1438
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
Linus Torvalds's avatar
Linus Torvalds committed
1439
		seq_printf(seq, "%s",
1440 1441 1442
			   rdev && test_bit(In_sync, &rdev->flags) ? "U" : "_");
	}
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
1443 1444 1445
	seq_printf(seq, "]");
}

1446
static void error(struct mddev *mddev, struct md_rdev *rdev)
Linus Torvalds's avatar
Linus Torvalds committed
1447 1448
{
	char b[BDEVNAME_SIZE];
1449
	struct r1conf *conf = mddev->private;
1450
	unsigned long flags;
Linus Torvalds's avatar
Linus Torvalds committed
1451 1452 1453 1454 1455 1456 1457

	/*
	 * 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
	 */
1458
	if (test_bit(In_sync, &rdev->flags)
1459
	    && (conf->raid_disks - mddev->degraded) == 1) {
Linus Torvalds's avatar
Linus Torvalds committed
1460 1461
		/*
		 * Don't fail the drive, act as though we were just a
1462 1463 1464
		 * normal single drive.
		 * However don't try a recovery from this drive as
		 * it is very likely to fail.
Linus Torvalds's avatar
Linus Torvalds committed
1465
		 */
1466
		conf->recovery_disabled = mddev->recovery_disabled;
Linus Torvalds's avatar
Linus Torvalds committed
1467
		return;
1468
	}
1469
	set_bit(Blocked, &rdev->flags);
1470
	spin_lock_irqsave(&conf->device_lock, flags);
1471
	if (test_and_clear_bit(In_sync, &rdev->flags)) {
Linus Torvalds's avatar
Linus Torvalds committed
1472
		mddev->degraded++;
1473 1474 1475
		set_bit(Faulty, &rdev->flags);
	} else
		set_bit(Faulty, &rdev->flags);
1476
	spin_unlock_irqrestore(&conf->device_lock, flags);
1477 1478 1479 1480
	/*
	 * if recovery is running, make sure it aborts.
	 */
	set_bit(MD_RECOVERY_INTR, &mddev->recovery);
1481
	set_bit(MD_CHANGE_DEVS, &mddev->flags);
1482
	set_bit(MD_CHANGE_PENDING, &mddev->flags);
1483 1484 1485
	printk(KERN_ALERT
	       "md/raid1:%s: Disk failure on %s, disabling device.\n"
	       "md/raid1:%s: Operation continuing on %d devices.\n",
1486 1487
	       mdname(mddev), bdevname(rdev->bdev, b),
	       mdname(mddev), conf->raid_disks - mddev->degraded);
Linus Torvalds's avatar
Linus Torvalds committed
1488 1489
}

1490
static void print_conf(struct r1conf *conf)
Linus Torvalds's avatar
Linus Torvalds committed
1491 1492 1493
{
	int i;

1494
	printk(KERN_DEBUG "RAID1 conf printout:\n");
Linus Torvalds's avatar
Linus Torvalds committed
1495
	if (!conf) {
1496
		printk(KERN_DEBUG "(!conf)\n");
Linus Torvalds's avatar
Linus Torvalds committed
1497 1498
		return;
	}
1499
	printk(KERN_DEBUG " --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
Linus Torvalds's avatar
Linus Torvalds committed
1500 1501
		conf->raid_disks);

1502
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
1503 1504
	for (i = 0; i < conf->raid_disks; i++) {
		char b[BDEVNAME_SIZE];
1505
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
1506
		if (rdev)
1507
			printk(KERN_DEBUG " disk %d, wo:%d, o:%d, dev:%s\n",
1508 1509 1510
			       i, !test_bit(In_sync, &rdev->flags),
			       !test_bit(Faulty, &rdev->flags),
			       bdevname(rdev->bdev,b));
Linus Torvalds's avatar
Linus Torvalds committed
1511
	}
1512
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
1513 1514
}

1515
static void close_sync(struct r1conf *conf)
Linus Torvalds's avatar
Linus Torvalds committed
1516
{
1517 1518
	wait_barrier(conf, NULL);
	allow_barrier(conf, 0, 0);
Linus Torvalds's avatar
Linus Torvalds committed
1519 1520 1521

	mempool_destroy(conf->r1buf_pool);
	conf->r1buf_pool = NULL;
1522

1523
	spin_lock_irq(&conf->resync_lock);
1524
	conf->next_resync = MaxSector - 2 * NEXT_NORMALIO_DISTANCE;
1525
	conf->start_next_window = MaxSector;
1526 1527 1528 1529
	conf->current_window_requests +=
		conf->next_window_requests;
	conf->next_window_requests = 0;
	spin_unlock_irq(&conf->resync_lock);
Linus Torvalds's avatar
Linus Torvalds committed
1530 1531
}

1532
static int raid1_spare_active(struct mddev *mddev)
Linus Torvalds's avatar
Linus Torvalds committed
1533 1534
{
	int i;
1535
	struct r1conf *conf = mddev->private;
1536 1537
	int count = 0;
	unsigned long flags;
Linus Torvalds's avatar
Linus Torvalds committed
1538 1539

	/*
1540
	 * Find all failed disks within the RAID1 configuration
1541 1542
	 * and mark them readable.
	 * Called under mddev lock, so rcu protection not needed.
1543 1544
	 * device_lock used to avoid races with raid1_end_read_request
	 * which expects 'In_sync' flags and ->degraded to be consistent.
Linus Torvalds's avatar
Linus Torvalds committed
1545
	 */
1546
	spin_lock_irqsave(&conf->device_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
1547
	for (i = 0; i < conf->raid_disks; i++) {
1548
		struct md_rdev *rdev = conf->mirrors[i].rdev;
1549 1550
		struct md_rdev *repl = conf->mirrors[conf->raid_disks + i].rdev;
		if (repl
1551
		    && !test_bit(Candidate, &repl->flags)
1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568
		    && 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);
			}
		}
1569
		if (rdev
1570
		    && rdev->recovery_offset == MaxSector
1571
		    && !test_bit(Faulty, &rdev->flags)
1572
		    && !test_and_set_bit(In_sync, &rdev->flags)) {
1573
			count++;
1574
			sysfs_notify_dirent_safe(rdev->sysfs_state);
Linus Torvalds's avatar
Linus Torvalds committed
1575 1576
		}
	}
1577 1578
	mddev->degraded -= count;
	spin_unlock_irqrestore(&conf->device_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
1579 1580

	print_conf(conf);
1581
	return count;
Linus Torvalds's avatar
Linus Torvalds committed
1582 1583
}

1584
static int raid1_add_disk(struct mddev *mddev, struct md_rdev *rdev)
Linus Torvalds's avatar
Linus Torvalds committed
1585
{
1586
	struct r1conf *conf = mddev->private;
1587
	int err = -EEXIST;
1588
	int mirror = 0;
1589
	struct raid1_info *p;
1590
	int first = 0;
1591
	int last = conf->raid_disks - 1;
Linus Torvalds's avatar
Linus Torvalds committed
1592

1593 1594 1595
	if (mddev->recovery_disabled == conf->recovery_disabled)
		return -EBUSY;

1596 1597 1598
	if (md_integrity_add_rdev(rdev, mddev))
		return -ENXIO;

1599 1600 1601
	if (rdev->raid_disk >= 0)
		first = last = rdev->raid_disk;

1602 1603 1604 1605 1606 1607
	/*
	 * find the disk ... but prefer rdev->saved_raid_disk
	 * if possible.
	 */
	if (rdev->saved_raid_disk >= 0 &&
	    rdev->saved_raid_disk >= first &&
1608
	    rdev->saved_raid_disk < conf->raid_disks &&
1609 1610 1611
	    conf->mirrors[rdev->saved_raid_disk].rdev == NULL)
		first = last = rdev->saved_raid_disk;

1612 1613 1614
	for (mirror = first; mirror <= last; mirror++) {
		p = conf->mirrors+mirror;
		if (!p->rdev) {
Linus Torvalds's avatar
Linus Torvalds committed
1615

1616 1617 1618
			if (mddev->gendisk)
				disk_stack_limits(mddev->gendisk, rdev->bdev,
						  rdev->data_offset << 9);
Linus Torvalds's avatar
Linus Torvalds committed
1619 1620 1621

			p->head_position = 0;
			rdev->raid_disk = mirror;
1622
			err = 0;
1623 1624 1625 1626
			/* 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)
1627
				conf->fullsync = 1;
1628
			rcu_assign_pointer(p->rdev, rdev);
Linus Torvalds's avatar
Linus Torvalds committed
1629 1630
			break;
		}
1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642
		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;
		}
	}
1643
	if (mddev->queue && blk_queue_discard(bdev_get_queue(rdev->bdev)))
Shaohua Li's avatar
Shaohua Li committed
1644
		queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, mddev->queue);
Linus Torvalds's avatar
Linus Torvalds committed
1645
	print_conf(conf);
1646
	return err;
Linus Torvalds's avatar
Linus Torvalds committed
1647 1648
}

1649
static int raid1_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
Linus Torvalds's avatar
Linus Torvalds committed
1650
{
1651
	struct r1conf *conf = mddev->private;
Linus Torvalds's avatar
Linus Torvalds committed
1652
	int err = 0;
1653
	int number = rdev->raid_disk;
1654
	struct raid1_info *p = conf->mirrors + number;
Linus Torvalds's avatar
Linus Torvalds committed
1655

1656 1657 1658
	if (rdev != p->rdev)
		p = conf->mirrors + conf->raid_disks + number;

Linus Torvalds's avatar
Linus Torvalds committed
1659
	print_conf(conf);
1660
	if (rdev == p->rdev) {
1661
		if (test_bit(In_sync, &rdev->flags) ||
Linus Torvalds's avatar
Linus Torvalds committed
1662 1663 1664 1665
		    atomic_read(&rdev->nr_pending)) {
			err = -EBUSY;
			goto abort;
		}
1666
		/* Only remove non-faulty devices if recovery
1667 1668 1669
		 * is not possible.
		 */
		if (!test_bit(Faulty, &rdev->flags) &&
1670
		    mddev->recovery_disabled != conf->recovery_disabled &&
1671 1672 1673 1674
		    mddev->degraded < conf->raid_disks) {
			err = -EBUSY;
			goto abort;
		}
Linus Torvalds's avatar
Linus Torvalds committed
1675
		p->rdev = NULL;
1676
		synchronize_rcu();
Linus Torvalds's avatar
Linus Torvalds committed
1677 1678 1679 1680
		if (atomic_read(&rdev->nr_pending)) {
			/* lost the race, try later */
			err = -EBUSY;
			p->rdev = rdev;
1681
			goto abort;
1682 1683 1684 1685 1686 1687 1688
		} 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;
1689
			freeze_array(conf, 0);
1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700
			if (atomic_read(&repl->nr_pending)) {
				/* It means that some queued IO of retry_list
				 * hold repl. Thus, we cannot set replacement
				 * as NULL, avoiding rdev NULL pointer
				 * dereference in sync_request_write and
				 * handle_write_finished.
				 */
				err = -EBUSY;
				unfreeze_array(conf);
				goto abort;
			}
1701 1702 1703
			clear_bit(Replacement, &repl->flags);
			p->rdev = repl;
			conf->mirrors[conf->raid_disks + number].rdev = NULL;
1704
			unfreeze_array(conf);
1705 1706
			clear_bit(WantReplacement, &rdev->flags);
		} else
1707
			clear_bit(WantReplacement, &rdev->flags);
1708
		err = md_integrity_register(mddev);
Linus Torvalds's avatar
Linus Torvalds committed
1709 1710 1711 1712 1713 1714 1715
	}
abort:

	print_conf(conf);
	return err;
}

1716
static void end_sync_read(struct bio *bio)
Linus Torvalds's avatar
Linus Torvalds committed
1717
{
1718
	struct r1bio *r1_bio = bio->bi_private;
Linus Torvalds's avatar
Linus Torvalds committed
1719

1720
	update_head_pos(r1_bio->read_disk, r1_bio);
1721

Linus Torvalds's avatar
Linus Torvalds committed
1722 1723 1724 1725 1726
	/*
	 * 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
	 */
1727
	if (!bio->bi_error)
Linus Torvalds's avatar
Linus Torvalds committed
1728
		set_bit(R1BIO_Uptodate, &r1_bio->state);
1729 1730 1731

	if (atomic_dec_and_test(&r1_bio->remaining))
		reschedule_retry(r1_bio);
Linus Torvalds's avatar
Linus Torvalds committed
1732 1733
}

1734
static void end_sync_write(struct bio *bio)
Linus Torvalds's avatar
Linus Torvalds committed
1735
{
1736
	int uptodate = !bio->bi_error;
1737
	struct r1bio *r1_bio = bio->bi_private;
1738
	struct mddev *mddev = r1_bio->mddev;
1739
	struct r1conf *conf = mddev->private;
Linus Torvalds's avatar
Linus Torvalds committed
1740
	int mirror=0;
1741 1742
	sector_t first_bad;
	int bad_sectors;
Linus Torvalds's avatar
Linus Torvalds committed
1743

1744 1745
	mirror = find_bio_disk(r1_bio, bio);

1746
	if (!uptodate) {
1747
		sector_t sync_blocks = 0;
1748 1749 1750 1751
		sector_t s = r1_bio->sector;
		long sectors_to_go = r1_bio->sectors;
		/* make sure these bits doesn't get cleared. */
		do {
1752
			bitmap_end_sync(mddev->bitmap, s,
1753 1754 1755 1756
					&sync_blocks, 1);
			s += sync_blocks;
			sectors_to_go -= sync_blocks;
		} while (sectors_to_go > 0);
1757 1758
		set_bit(WriteErrorSeen,
			&conf->mirrors[mirror].rdev->flags);
1759 1760 1761 1762
		if (!test_and_set_bit(WantReplacement,
				      &conf->mirrors[mirror].rdev->flags))
			set_bit(MD_RECOVERY_NEEDED, &
				mddev->recovery);
1763
		set_bit(R1BIO_WriteError, &r1_bio->state);
1764 1765 1766
	} else if (is_badblock(conf->mirrors[mirror].rdev,
			       r1_bio->sector,
			       r1_bio->sectors,
1767 1768 1769 1770 1771 1772
			       &first_bad, &bad_sectors) &&
		   !is_badblock(conf->mirrors[r1_bio->read_disk].rdev,
				r1_bio->sector,
				r1_bio->sectors,
				&first_bad, &bad_sectors)
		)
1773
		set_bit(R1BIO_MadeGood, &r1_bio->state);
1774

Linus Torvalds's avatar
Linus Torvalds committed
1775
	if (atomic_dec_and_test(&r1_bio->remaining)) {
1776
		int s = r1_bio->sectors;
1777 1778
		if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
		    test_bit(R1BIO_WriteError, &r1_bio->state))
1779 1780 1781 1782 1783
			reschedule_retry(r1_bio);
		else {
			put_buf(r1_bio);
			md_done_sync(mddev, s, uptodate);
		}
Linus Torvalds's avatar
Linus Torvalds committed
1784 1785 1786
	}
}

1787
static int r1_sync_page_io(struct md_rdev *rdev, sector_t sector,
1788 1789 1790 1791 1792
			    int sectors, struct page *page, int rw)
{
	if (sync_page_io(rdev, sector, sectors << 9, page, rw, false))
		/* success */
		return 1;
1793
	if (rw == WRITE) {
1794
		set_bit(WriteErrorSeen, &rdev->flags);
1795 1796 1797 1798 1799
		if (!test_and_set_bit(WantReplacement,
				      &rdev->flags))
			set_bit(MD_RECOVERY_NEEDED, &
				rdev->mddev->recovery);
	}
1800 1801 1802 1803 1804 1805
	/* 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;
}

1806
static int fix_sync_read_error(struct r1bio *r1_bio)
Linus Torvalds's avatar
Linus Torvalds committed
1807
{
1808 1809 1810 1811 1812 1813 1814
	/* 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.
1815 1816 1817
	 * 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.
1818
	 */
1819
	struct mddev *mddev = r1_bio->mddev;
1820
	struct r1conf *conf = mddev->private;
1821 1822 1823 1824 1825 1826 1827 1828 1829
	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;
1830
		struct md_rdev *rdev;
1831
		int start;
1832 1833 1834 1835 1836 1837 1838 1839 1840 1841

		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;
1842
				if (sync_page_io(rdev, sect, s<<9,
1843 1844 1845 1846 1847 1848 1849
						 bio->bi_io_vec[idx].bv_page,
						 READ, false)) {
					success = 1;
					break;
				}
			}
			d++;
1850
			if (d == conf->raid_disks * 2)
1851 1852 1853
				d = 0;
		} while (!success && d != r1_bio->read_disk);

1854
		if (!success) {
1855
			char b[BDEVNAME_SIZE];
1856 1857 1858 1859 1860 1861
			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.
			 */
1862 1863 1864 1865 1866
			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);
1867
			for (d = 0; d < conf->raid_disks * 2; d++) {
1868 1869 1870 1871 1872 1873 1874
				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) {
1875 1876
				conf->recovery_disabled =
					mddev->recovery_disabled;
1877 1878 1879 1880 1881 1882 1883 1884 1885 1886
				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;
1887
		}
1888 1889 1890 1891 1892

		start = d;
		/* write it back and re-read */
		while (d != r1_bio->read_disk) {
			if (d == 0)
1893
				d = conf->raid_disks * 2;
1894 1895 1896 1897
			d--;
			if (r1_bio->bios[d]->bi_end_io != end_sync_read)
				continue;
			rdev = conf->mirrors[d].rdev;
1898 1899 1900
			if (r1_sync_page_io(rdev, sect, s,
					    bio->bi_io_vec[idx].bv_page,
					    WRITE) == 0) {
1901 1902
				r1_bio->bios[d]->bi_end_io = NULL;
				rdev_dec_pending(rdev, mddev);
1903
			}
1904 1905 1906 1907
		}
		d = start;
		while (d != r1_bio->read_disk) {
			if (d == 0)
1908
				d = conf->raid_disks * 2;
1909 1910 1911 1912
			d--;
			if (r1_bio->bios[d]->bi_end_io != end_sync_read)
				continue;
			rdev = conf->mirrors[d].rdev;
1913 1914 1915
			if (r1_sync_page_io(rdev, sect, s,
					    bio->bi_io_vec[idx].bv_page,
					    READ) != 0)
1916
				atomic_add(s, &rdev->corrected_errors);
1917
		}
1918 1919 1920 1921
		sectors -= s;
		sect += s;
		idx ++;
	}
1922
	set_bit(R1BIO_Uptodate, &r1_bio->state);
1923
	bio->bi_error = 0;
1924 1925 1926
	return 1;
}

1927
static void process_checks(struct r1bio *r1_bio)
1928 1929 1930 1931 1932 1933 1934 1935
{
	/* 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
	 */
1936
	struct mddev *mddev = r1_bio->mddev;
1937
	struct r1conf *conf = mddev->private;
1938 1939
	int primary;
	int i;
1940
	int vcnt;
1941

1942 1943 1944 1945 1946
	/* 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;
1947
		int error;
1948 1949 1950
		struct bio *b = r1_bio->bios[i];
		if (b->bi_end_io != end_sync_read)
			continue;
1951 1952
		/* fixup the bio for reuse, but preserve errno */
		error = b->bi_error;
1953
		bio_reset(b);
1954
		b->bi_error = error;
1955
		b->bi_vcnt = vcnt;
1956 1957
		b->bi_iter.bi_size = r1_bio->sectors << 9;
		b->bi_iter.bi_sector = r1_bio->sector +
1958 1959 1960 1961 1962
			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;

1963
		size = b->bi_iter.bi_size;
1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974
		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;
		}
	}
1975
	for (primary = 0; primary < conf->raid_disks * 2; primary++)
1976
		if (r1_bio->bios[primary]->bi_end_io == end_sync_read &&
1977
		    !r1_bio->bios[primary]->bi_error) {
1978 1979 1980 1981 1982
			r1_bio->bios[primary]->bi_end_io = NULL;
			rdev_dec_pending(conf->mirrors[primary].rdev, mddev);
			break;
		}
	r1_bio->read_disk = primary;
1983
	for (i = 0; i < conf->raid_disks * 2; i++) {
1984 1985 1986
		int j;
		struct bio *pbio = r1_bio->bios[primary];
		struct bio *sbio = r1_bio->bios[i];
1987
		int error = sbio->bi_error;
1988

Kent Overstreet's avatar
Kent Overstreet committed
1989
		if (sbio->bi_end_io != end_sync_read)
1990
			continue;
1991 1992
		/* Now we can 'fixup' the error value */
		sbio->bi_error = 0;
1993

1994
		if (!error) {
1995 1996 1997 1998 1999 2000
			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),
2001
					   sbio->bi_io_vec[j].bv_len))
2002
					break;
2003
			}
2004 2005 2006
		} else
			j = 0;
		if (j >= 0)
2007
			atomic64_add(r1_bio->sectors, &mddev->resync_mismatches);
2008
		if (j < 0 || (test_bit(MD_RECOVERY_CHECK, &mddev->recovery)
2009
			      && !error)) {
2010 2011 2012 2013 2014
			/* No need to write to this device. */
			sbio->bi_end_io = NULL;
			rdev_dec_pending(conf->mirrors[i].rdev, mddev);
			continue;
		}
Kent Overstreet's avatar
Kent Overstreet committed
2015 2016

		bio_copy_data(sbio, pbio);
2017
	}
2018 2019
}

2020
static void sync_request_write(struct mddev *mddev, struct r1bio *r1_bio)
2021
{
2022
	struct r1conf *conf = mddev->private;
2023
	int i;
2024
	int disks = conf->raid_disks * 2;
2025 2026 2027 2028 2029 2030 2031 2032
	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;
2033 2034

	if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
2035 2036
		process_checks(r1_bio);

2037 2038 2039
	/*
	 * schedule writes
	 */
Linus Torvalds's avatar
Linus Torvalds committed
2040 2041 2042
	atomic_set(&r1_bio->remaining, 1);
	for (i = 0; i < disks ; i++) {
		wbio = r1_bio->bios[i];
2043 2044 2045 2046
		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))))
Linus Torvalds's avatar
Linus Torvalds committed
2047 2048
			continue;

2049 2050
		wbio->bi_rw = WRITE;
		wbio->bi_end_io = end_sync_write;
Linus Torvalds's avatar
Linus Torvalds committed
2051
		atomic_inc(&r1_bio->remaining);
2052
		md_sync_acct(conf->mirrors[i].rdev->bdev, bio_sectors(wbio));
2053

Linus Torvalds's avatar
Linus Torvalds committed
2054 2055 2056 2057
		generic_make_request(wbio);
	}

	if (atomic_dec_and_test(&r1_bio->remaining)) {
2058
		/* if we're here, all write(s) have completed, so clean up */
2059 2060 2061 2062 2063 2064 2065 2066
		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);
		}
Linus Torvalds's avatar
Linus Torvalds committed
2067 2068 2069 2070 2071 2072 2073 2074
	}
}

/*
 * This is a kernel thread which:
 *
 *	1.	Retries failed read operations on working mirrors.
 *	2.	Updates the raid superblock when problems encounter.
2075
 *	3.	Performs writes following reads for array synchronising.
Linus Torvalds's avatar
Linus Torvalds committed
2076 2077
 */

2078
static void fix_read_error(struct r1conf *conf, int read_disk,
2079 2080
			   sector_t sect, int sectors)
{
2081
	struct mddev *mddev = conf->mddev;
2082 2083 2084 2085 2086
	while(sectors) {
		int s = sectors;
		int d = read_disk;
		int success = 0;
		int start;
2087
		struct md_rdev *rdev;
2088 2089 2090 2091 2092 2093 2094 2095 2096 2097

		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....
			 */
2098 2099 2100
			sector_t first_bad;
			int bad_sectors;

2101 2102
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
2103 2104 2105
			    (test_bit(In_sync, &rdev->flags) ||
			     (!test_bit(Faulty, &rdev->flags) &&
			      rdev->recovery_offset >= sect + s)) &&
2106 2107
			    is_badblock(rdev, sect, s,
					&first_bad, &bad_sectors) == 0 &&
2108 2109
			    sync_page_io(rdev, sect, s<<9,
					 conf->tmppage, READ, false))
2110 2111 2112
				success = 1;
			else {
				d++;
2113
				if (d == conf->raid_disks * 2)
2114 2115 2116 2117 2118
					d = 0;
			}
		} while (!success && d != read_disk);

		if (!success) {
2119
			/* Cannot read from anywhere - mark it bad */
2120
			struct md_rdev *rdev = conf->mirrors[read_disk].rdev;
2121 2122
			if (!rdev_set_badblocks(rdev, sect, s, 0))
				md_error(mddev, rdev);
2123 2124 2125 2126 2127 2128
			break;
		}
		/* write it back and re-read */
		start = d;
		while (d != read_disk) {
			if (d==0)
2129
				d = conf->raid_disks * 2;
2130 2131 2132
			d--;
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
2133
			    !test_bit(Faulty, &rdev->flags))
2134 2135
				r1_sync_page_io(rdev, sect, s,
						conf->tmppage, WRITE);
2136 2137 2138 2139 2140
		}
		d = start;
		while (d != read_disk) {
			char b[BDEVNAME_SIZE];
			if (d==0)
2141
				d = conf->raid_disks * 2;
2142 2143 2144
			d--;
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
2145
			    !test_bit(Faulty, &rdev->flags)) {
2146 2147
				if (r1_sync_page_io(rdev, sect, s,
						    conf->tmppage, READ)) {
2148 2149
					atomic_add(s, &rdev->corrected_errors);
					printk(KERN_INFO
2150
					       "md/raid1:%s: read error corrected "
2151 2152
					       "(%d sectors at %llu on %s)\n",
					       mdname(mddev), s,
2153 2154
					       (unsigned long long)(sect +
					           rdev->data_offset),
2155 2156 2157 2158 2159 2160 2161 2162 2163
					       bdevname(rdev->bdev, b));
				}
			}
		}
		sectors -= s;
		sect += s;
	}
}

2164
static int narrow_write_error(struct r1bio *r1_bio, int i)
2165
{
2166
	struct mddev *mddev = r1_bio->mddev;
2167
	struct r1conf *conf = mddev->private;
2168
	struct md_rdev *rdev = conf->mirrors[i].rdev;
2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189

	/* 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;

2190 2191
	block_sectors = roundup(1 << rdev->badblocks.shift,
				bdev_logical_block_size(rdev->bdev) >> 9);
2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202
	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'*/

2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219
		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);
		}

2220
		wbio->bi_rw = WRITE;
2221 2222
		wbio->bi_iter.bi_sector = r1_bio->sector;
		wbio->bi_iter.bi_size = r1_bio->sectors << 9;
2223

2224
		bio_trim(wbio, sector - r1_bio->sector, sectors);
2225
		wbio->bi_iter.bi_sector += rdev->data_offset;
2226
		wbio->bi_bdev = rdev->bdev;
2227
		if (submit_bio_wait(WRITE, wbio) < 0)
2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240
			/* failure! */
			ok = rdev_set_badblocks(rdev, sector,
						sectors, 0)
				&& ok;

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

2241
static void handle_sync_write_finished(struct r1conf *conf, struct r1bio *r1_bio)
2242 2243 2244
{
	int m;
	int s = r1_bio->sectors;
2245
	for (m = 0; m < conf->raid_disks * 2 ; m++) {
2246
		struct md_rdev *rdev = conf->mirrors[m].rdev;
2247 2248 2249
		struct bio *bio = r1_bio->bios[m];
		if (bio->bi_end_io == NULL)
			continue;
2250
		if (!bio->bi_error &&
2251
		    test_bit(R1BIO_MadeGood, &r1_bio->state)) {
2252
			rdev_clear_badblocks(rdev, r1_bio->sector, s, 0);
2253
		}
2254
		if (bio->bi_error &&
2255 2256 2257 2258 2259 2260 2261 2262 2263
		    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);
}

2264
static void handle_write_finished(struct r1conf *conf, struct r1bio *r1_bio)
2265 2266
{
	int m;
2267
	bool fail = false;
2268
	for (m = 0; m < conf->raid_disks * 2 ; m++)
2269
		if (r1_bio->bios[m] == IO_MADE_GOOD) {
2270
			struct md_rdev *rdev = conf->mirrors[m].rdev;
2271 2272
			rdev_clear_badblocks(rdev,
					     r1_bio->sector,
2273
					     r1_bio->sectors, 0);
2274 2275 2276 2277 2278 2279
			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.
			 */
2280
			fail = true;
2281 2282 2283 2284 2285 2286 2287 2288 2289
			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);
		}
2290 2291 2292
	if (fail) {
		spin_lock_irq(&conf->device_lock);
		list_add(&r1_bio->retry_list, &conf->bio_end_io_list);
2293
		conf->nr_queued++;
2294 2295
		spin_unlock_irq(&conf->device_lock);
		md_wakeup_thread(conf->mddev->thread);
2296 2297 2298
	} else {
		if (test_bit(R1BIO_WriteError, &r1_bio->state))
			close_write(r1_bio);
2299
		raid_end_bio_io(r1_bio);
2300
	}
2301 2302
}

2303
static void handle_read_error(struct r1conf *conf, struct r1bio *r1_bio)
2304 2305 2306
{
	int disk;
	int max_sectors;
2307
	struct mddev *mddev = conf->mddev;
2308 2309
	struct bio *bio;
	char b[BDEVNAME_SIZE];
2310
	struct md_rdev *rdev;
2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321

	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) {
2322
		freeze_array(conf, 1);
2323 2324 2325 2326 2327
		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);
2328
	rdev_dec_pending(conf->mirrors[r1_bio->read_disk].rdev, conf->mddev);
2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348

	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);
2349 2350
		bio_trim(bio, r1_bio->sector - bio->bi_iter.bi_sector,
			 max_sectors);
2351 2352 2353 2354 2355 2356 2357 2358
		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));
2359
		bio->bi_iter.bi_sector = r1_bio->sector + rdev->data_offset;
2360 2361 2362 2363 2364 2365 2366 2367
		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
2368
					       - mbio->bi_iter.bi_sector);
2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381
			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;
2382
			r1_bio->sectors = bio_sectors(mbio) - sectors_handled;
2383 2384 2385
			r1_bio->state = 0;
			set_bit(R1BIO_ReadError, &r1_bio->state);
			r1_bio->mddev = mddev;
2386 2387
			r1_bio->sector = mbio->bi_iter.bi_sector +
				sectors_handled;
2388 2389 2390 2391 2392 2393 2394

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

2395
static void raid1d(struct md_thread *thread)
Linus Torvalds's avatar
Linus Torvalds committed
2396
{
2397
	struct mddev *mddev = thread->mddev;
2398
	struct r1bio *r1_bio;
Linus Torvalds's avatar
Linus Torvalds committed
2399
	unsigned long flags;
2400
	struct r1conf *conf = mddev->private;
Linus Torvalds's avatar
Linus Torvalds committed
2401
	struct list_head *head = &conf->retry_list;
2402
	struct blk_plug plug;
Linus Torvalds's avatar
Linus Torvalds committed
2403 2404

	md_check_recovery(mddev);
2405

2406 2407 2408 2409 2410
	if (!list_empty_careful(&conf->bio_end_io_list) &&
	    !test_bit(MD_CHANGE_PENDING, &mddev->flags)) {
		LIST_HEAD(tmp);
		spin_lock_irqsave(&conf->device_lock, flags);
		if (!test_bit(MD_CHANGE_PENDING, &mddev->flags)) {
2411 2412 2413 2414
			while (!list_empty(&conf->bio_end_io_list)) {
				list_move(conf->bio_end_io_list.prev, &tmp);
				conf->nr_queued--;
			}
2415 2416 2417
		}
		spin_unlock_irqrestore(&conf->device_lock, flags);
		while (!list_empty(&tmp)) {
2418 2419
			r1_bio = list_first_entry(&tmp, struct r1bio,
						  retry_list);
2420
			list_del(&r1_bio->retry_list);
2421 2422 2423 2424
			if (mddev->degraded)
				set_bit(R1BIO_Degraded, &r1_bio->state);
			if (test_bit(R1BIO_WriteError, &r1_bio->state))
				close_write(r1_bio);
2425 2426 2427 2428
			raid_end_bio_io(r1_bio);
		}
	}

2429
	blk_start_plug(&plug);
Linus Torvalds's avatar
Linus Torvalds committed
2430
	for (;;) {
2431

2432
		flush_pending_writes(conf);
2433

2434 2435 2436
		spin_lock_irqsave(&conf->device_lock, flags);
		if (list_empty(head)) {
			spin_unlock_irqrestore(&conf->device_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
2437
			break;
2438
		}
2439
		r1_bio = list_entry(head->prev, struct r1bio, retry_list);
Linus Torvalds's avatar
Linus Torvalds committed
2440
		list_del(head->prev);
2441
		conf->nr_queued--;
Linus Torvalds's avatar
Linus Torvalds committed
2442 2443 2444
		spin_unlock_irqrestore(&conf->device_lock, flags);

		mddev = r1_bio->mddev;
2445
		conf = mddev->private;
2446
		if (test_bit(R1BIO_IsSync, &r1_bio->state)) {
2447
			if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
2448 2449 2450
			    test_bit(R1BIO_WriteError, &r1_bio->state))
				handle_sync_write_finished(conf, r1_bio);
			else
2451
				sync_request_write(mddev, r1_bio);
2452
		} else if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
2453 2454 2455 2456 2457
			   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
2458 2459 2460 2461
			/* just a partial read to be scheduled from separate
			 * context
			 */
			generic_make_request(r1_bio->bios[r1_bio->read_disk]);
2462

2463
		cond_resched();
2464 2465
		if (mddev->flags & ~(1<<MD_CHANGE_PENDING))
			md_check_recovery(mddev);
Linus Torvalds's avatar
Linus Torvalds committed
2466
	}
2467
	blk_finish_plug(&plug);
Linus Torvalds's avatar
Linus Torvalds committed
2468 2469
}

2470
static int init_resync(struct r1conf *conf)
Linus Torvalds's avatar
Linus Torvalds committed
2471 2472 2473 2474
{
	int buffs;

	buffs = RESYNC_WINDOW / RESYNC_BLOCK_SIZE;
2475
	BUG_ON(conf->r1buf_pool);
Linus Torvalds's avatar
Linus Torvalds committed
2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493
	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.
 */

2494
static sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipped)
Linus Torvalds's avatar
Linus Torvalds committed
2495
{
2496
	struct r1conf *conf = mddev->private;
2497
	struct r1bio *r1_bio;
Linus Torvalds's avatar
Linus Torvalds committed
2498 2499
	struct bio *bio;
	sector_t max_sector, nr_sectors;
2500
	int disk = -1;
Linus Torvalds's avatar
Linus Torvalds committed
2501
	int i;
2502 2503
	int wonly = -1;
	int write_targets = 0, read_targets = 0;
2504
	sector_t sync_blocks;
2505
	int still_degraded = 0;
2506 2507
	int good_sectors = RESYNC_SECTORS;
	int min_bad = 0; /* number of sectors that are bad in all devices */
Linus Torvalds's avatar
Linus Torvalds committed
2508 2509 2510

	if (!conf->r1buf_pool)
		if (init_resync(conf))
2511
			return 0;
Linus Torvalds's avatar
Linus Torvalds committed
2512

2513
	max_sector = mddev->dev_sectors;
Linus Torvalds's avatar
Linus Torvalds committed
2514
	if (sector_nr >= max_sector) {
2515 2516 2517 2518 2519
		/* 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
		 */
2520 2521
		if (mddev->curr_resync < max_sector) /* aborted */
			bitmap_end_sync(mddev->bitmap, mddev->curr_resync,
2522
						&sync_blocks, 1);
2523
		else /* completed sync */
2524
			conf->fullsync = 0;
2525 2526

		bitmap_close_sync(mddev->bitmap);
Linus Torvalds's avatar
Linus Torvalds committed
2527
		close_sync(conf);
2528 2529 2530 2531 2532

		if (mddev_is_clustered(mddev)) {
			conf->cluster_sync_low = 0;
			conf->cluster_sync_high = 0;
		}
Linus Torvalds's avatar
Linus Torvalds committed
2533 2534 2535
		return 0;
	}

2536 2537
	if (mddev->bitmap == NULL &&
	    mddev->recovery_cp == MaxSector &&
2538
	    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
2539 2540 2541 2542
	    conf->fullsync == 0) {
		*skipped = 1;
		return max_sector - sector_nr;
	}
2543 2544 2545
	/* before building a request, check if we can skip these blocks..
	 * This call the bitmap_start_sync doesn't actually record anything
	 */
2546
	if (!bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, 1) &&
2547
	    !conf->fullsync && !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
2548 2549 2550 2551
		/* We can skip this block, and probably several more */
		*skipped = 1;
		return sync_blocks;
	}
2552

2553 2554 2555 2556 2557 2558
	/* we are incrementing sector_nr below. To be safe, we check against
	 * sector_nr + two times RESYNC_SECTORS
	 */

	bitmap_cond_end_sync(mddev->bitmap, sector_nr,
		mddev_is_clustered(mddev) && (sector_nr + 2 * RESYNC_SECTORS > conf->cluster_sync_high));
2559
	r1_bio = mempool_alloc(conf->r1buf_pool, GFP_NOIO);
2560

2561
	raise_barrier(conf, sector_nr);
Linus Torvalds's avatar
Linus Torvalds committed
2562

2563
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
2564
	/*
2565 2566 2567 2568 2569 2570
	 * 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.
Linus Torvalds's avatar
Linus Torvalds committed
2571 2572 2573 2574
	 */

	r1_bio->mddev = mddev;
	r1_bio->sector = sector_nr;
2575
	r1_bio->state = 0;
Linus Torvalds's avatar
Linus Torvalds committed
2576 2577
	set_bit(R1BIO_IsSync, &r1_bio->state);

2578
	for (i = 0; i < conf->raid_disks * 2; i++) {
2579
		struct md_rdev *rdev;
Linus Torvalds's avatar
Linus Torvalds committed
2580
		bio = r1_bio->bios[i];
Kent Overstreet's avatar
Kent Overstreet committed
2581
		bio_reset(bio);
Linus Torvalds's avatar
Linus Torvalds committed
2582

2583 2584
		rdev = rcu_dereference(conf->mirrors[i].rdev);
		if (rdev == NULL ||
2585
		    test_bit(Faulty, &rdev->flags)) {
2586 2587
			if (i < conf->raid_disks)
				still_degraded = 1;
2588
		} else if (!test_bit(In_sync, &rdev->flags)) {
Linus Torvalds's avatar
Linus Torvalds committed
2589 2590 2591
			bio->bi_rw = WRITE;
			bio->bi_end_io = end_sync_write;
			write_targets ++;
2592 2593
		} else {
			/* may need to read from here */
2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618
			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++;
2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630
			} 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++;
2631 2632
			}
		}
2633
		if (rdev && bio->bi_end_io) {
2634
			atomic_inc(&rdev->nr_pending);
2635
			bio->bi_iter.bi_sector = sector_nr + rdev->data_offset;
2636 2637 2638
			bio->bi_bdev = rdev->bdev;
			bio->bi_private = r1_bio;
		}
Linus Torvalds's avatar
Linus Torvalds committed
2639
	}
2640 2641 2642 2643
	rcu_read_unlock();
	if (disk < 0)
		disk = wonly;
	r1_bio->read_disk = disk;
2644

2645 2646 2647 2648 2649
	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;
2650
		for (i = 0 ; i < conf->raid_disks * 2 ; i++)
2651
			if (r1_bio->bios[i]->bi_end_io == end_sync_write) {
2652
				struct md_rdev *rdev = conf->mirrors[i].rdev;
2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679
				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;
	}

2680 2681 2682 2683 2684
	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) {
Linus Torvalds's avatar
Linus Torvalds committed
2685 2686 2687
		/* There is nowhere to write, so all non-sync
		 * drives must be failed - so we are finished
		 */
2688 2689 2690 2691
		sector_t rv;
		if (min_bad > 0)
			max_sector = sector_nr + min_bad;
		rv = max_sector - sector_nr;
2692
		*skipped = 1;
Linus Torvalds's avatar
Linus Torvalds committed
2693 2694 2695 2696
		put_buf(r1_bio);
		return rv;
	}

2697 2698
	if (max_sector > mddev->resync_max)
		max_sector = mddev->resync_max; /* Don't do IO beyond here */
2699 2700
	if (max_sector > sector_nr + good_sectors)
		max_sector = sector_nr + good_sectors;
Linus Torvalds's avatar
Linus Torvalds committed
2701
	nr_sectors = 0;
2702
	sync_blocks = 0;
Linus Torvalds's avatar
Linus Torvalds committed
2703 2704 2705 2706 2707 2708 2709
	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;
2710 2711
		if (sync_blocks == 0) {
			if (!bitmap_start_sync(mddev->bitmap, sector_nr,
2712 2713 2714
					       &sync_blocks, still_degraded) &&
			    !conf->fullsync &&
			    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
2715
				break;
2716
			BUG_ON(sync_blocks < (PAGE_SIZE>>9));
2717
			if ((len >> 9) > sync_blocks)
2718
				len = sync_blocks<<9;
2719
		}
2720

2721
		for (i = 0 ; i < conf->raid_disks * 2; i++) {
Linus Torvalds's avatar
Linus Torvalds committed
2722 2723
			bio = r1_bio->bios[i];
			if (bio->bi_end_io) {
2724
				page = bio->bi_io_vec[bio->bi_vcnt].bv_page;
Linus Torvalds's avatar
Linus Torvalds committed
2725 2726
				if (bio_add_page(bio, page, len, 0) == 0) {
					/* stop here */
2727
					bio->bi_io_vec[bio->bi_vcnt].bv_page = page;
Linus Torvalds's avatar
Linus Torvalds committed
2728 2729 2730
					while (i > 0) {
						i--;
						bio = r1_bio->bios[i];
2731 2732
						if (bio->bi_end_io==NULL)
							continue;
Linus Torvalds's avatar
Linus Torvalds committed
2733 2734
						/* remove last page from this bio */
						bio->bi_vcnt--;
2735
						bio->bi_iter.bi_size -= len;
2736
						bio_clear_flag(bio, BIO_SEG_VALID);
Linus Torvalds's avatar
Linus Torvalds committed
2737 2738 2739 2740 2741 2742 2743
					}
					goto bio_full;
				}
			}
		}
		nr_sectors += len>>9;
		sector_nr += len>>9;
2744
		sync_blocks -= (len>>9);
Linus Torvalds's avatar
Linus Torvalds committed
2745 2746 2747 2748
	} while (r1_bio->bios[disk]->bi_vcnt < RESYNC_PAGES);
 bio_full:
	r1_bio->sectors = nr_sectors;

2749 2750 2751 2752 2753 2754 2755 2756 2757 2758
	if (mddev_is_clustered(mddev) &&
			conf->cluster_sync_high < sector_nr + nr_sectors) {
		conf->cluster_sync_low = mddev->curr_resync_completed;
		conf->cluster_sync_high = conf->cluster_sync_low + CLUSTER_RESYNC_WINDOW_SECTORS;
		/* Send resync message */
		md_cluster_ops->resync_info_update(mddev,
				conf->cluster_sync_low,
				conf->cluster_sync_high);
	}

2759 2760 2761 2762 2763
	/* 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);
2764
		for (i = 0; i < conf->raid_disks * 2 && read_targets; i++) {
2765 2766
			bio = r1_bio->bios[i];
			if (bio->bi_end_io == end_sync_read) {
2767
				read_targets--;
2768
				md_sync_acct(bio->bi_bdev, nr_sectors);
2769 2770 2771 2772 2773 2774
				generic_make_request(bio);
			}
		}
	} else {
		atomic_set(&r1_bio->remaining, 1);
		bio = r1_bio->bios[r1_bio->read_disk];
2775
		md_sync_acct(bio->bi_bdev, nr_sectors);
2776
		generic_make_request(bio);
Linus Torvalds's avatar
Linus Torvalds committed
2777

2778
	}
Linus Torvalds's avatar
Linus Torvalds committed
2779 2780 2781
	return nr_sectors;
}

2782
static sector_t raid1_size(struct mddev *mddev, sector_t sectors, int raid_disks)
2783 2784 2785 2786 2787 2788 2789
{
	if (sectors)
		return sectors;

	return mddev->dev_sectors;
}

2790
static struct r1conf *setup_conf(struct mddev *mddev)
Linus Torvalds's avatar
Linus Torvalds committed
2791
{
2792
	struct r1conf *conf;
2793
	int i;
2794
	struct raid1_info *disk;
2795
	struct md_rdev *rdev;
2796
	int err = -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
2797

2798
	conf = kzalloc(sizeof(struct r1conf), GFP_KERNEL);
Linus Torvalds's avatar
Linus Torvalds committed
2799
	if (!conf)
2800
		goto abort;
Linus Torvalds's avatar
Linus Torvalds committed
2801

2802
	conf->mirrors = kzalloc(sizeof(struct raid1_info)
2803
				* mddev->raid_disks * 2,
Linus Torvalds's avatar
Linus Torvalds committed
2804 2805
				 GFP_KERNEL);
	if (!conf->mirrors)
2806
		goto abort;
Linus Torvalds's avatar
Linus Torvalds committed
2807

2808 2809
	conf->tmppage = alloc_page(GFP_KERNEL);
	if (!conf->tmppage)
2810
		goto abort;
2811

2812
	conf->poolinfo = kzalloc(sizeof(*conf->poolinfo), GFP_KERNEL);
Linus Torvalds's avatar
Linus Torvalds committed
2813
	if (!conf->poolinfo)
2814
		goto abort;
2815
	conf->poolinfo->raid_disks = mddev->raid_disks * 2;
Linus Torvalds's avatar
Linus Torvalds committed
2816 2817 2818 2819
	conf->r1bio_pool = mempool_create(NR_RAID1_BIOS, r1bio_pool_alloc,
					  r1bio_pool_free,
					  conf->poolinfo);
	if (!conf->r1bio_pool)
2820 2821
		goto abort;

2822
	conf->poolinfo->mddev = mddev;
Linus Torvalds's avatar
Linus Torvalds committed
2823

2824
	err = -EINVAL;
2825
	spin_lock_init(&conf->device_lock);
NeilBrown's avatar
NeilBrown committed
2826
	rdev_for_each(rdev, mddev) {
2827
		struct request_queue *q;
2828
		int disk_idx = rdev->raid_disk;
Linus Torvalds's avatar
Linus Torvalds committed
2829 2830 2831
		if (disk_idx >= mddev->raid_disks
		    || disk_idx < 0)
			continue;
2832
		if (test_bit(Replacement, &rdev->flags))
2833
			disk = conf->mirrors + mddev->raid_disks + disk_idx;
2834 2835
		else
			disk = conf->mirrors + disk_idx;
Linus Torvalds's avatar
Linus Torvalds committed
2836

2837 2838
		if (disk->rdev)
			goto abort;
Linus Torvalds's avatar
Linus Torvalds committed
2839
		disk->rdev = rdev;
2840
		q = bdev_get_queue(rdev->bdev);
Linus Torvalds's avatar
Linus Torvalds committed
2841 2842

		disk->head_position = 0;
2843
		disk->seq_start = MaxSector;
Linus Torvalds's avatar
Linus Torvalds committed
2844 2845 2846 2847
	}
	conf->raid_disks = mddev->raid_disks;
	conf->mddev = mddev;
	INIT_LIST_HEAD(&conf->retry_list);
2848
	INIT_LIST_HEAD(&conf->bio_end_io_list);
Linus Torvalds's avatar
Linus Torvalds committed
2849 2850

	spin_lock_init(&conf->resync_lock);
2851
	init_waitqueue_head(&conf->wait_barrier);
Linus Torvalds's avatar
Linus Torvalds committed
2852

2853
	bio_list_init(&conf->pending_bio_list);
2854
	conf->pending_count = 0;
2855
	conf->recovery_disabled = mddev->recovery_disabled - 1;
2856

2857 2858 2859
	conf->start_next_window = MaxSector;
	conf->current_window_requests = conf->next_window_requests = 0;

2860
	err = -EIO;
2861
	for (i = 0; i < conf->raid_disks * 2; i++) {
Linus Torvalds's avatar
Linus Torvalds committed
2862 2863 2864

		disk = conf->mirrors + i;

2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879
		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;
		}

2880 2881
		if (!disk->rdev ||
		    !test_bit(In_sync, &disk->rdev->flags)) {
Linus Torvalds's avatar
Linus Torvalds committed
2882
			disk->head_position = 0;
2883 2884
			if (disk->rdev &&
			    (disk->rdev->saved_raid_disk < 0))
2885
				conf->fullsync = 1;
2886
		}
Linus Torvalds's avatar
Linus Torvalds committed
2887
	}
2888 2889

	err = -ENOMEM;
2890
	conf->thread = md_register_thread(raid1d, mddev, "raid1");
2891 2892
	if (!conf->thread) {
		printk(KERN_ERR
2893
		       "md/raid1:%s: couldn't allocate thread\n",
2894 2895
		       mdname(mddev));
		goto abort;
2896
	}
Linus Torvalds's avatar
Linus Torvalds committed
2897

2898 2899 2900 2901
	return conf;

 abort:
	if (conf) {
2902
		mempool_destroy(conf->r1bio_pool);
2903 2904 2905 2906 2907 2908 2909 2910
		kfree(conf->mirrors);
		safe_put_page(conf->tmppage);
		kfree(conf->poolinfo);
		kfree(conf);
	}
	return ERR_PTR(err);
}

NeilBrown's avatar
NeilBrown committed
2911
static void raid1_free(struct mddev *mddev, void *priv);
2912
static int run(struct mddev *mddev)
2913
{
2914
	struct r1conf *conf;
2915
	int i;
2916
	struct md_rdev *rdev;
2917
	int ret;
Shaohua Li's avatar
Shaohua Li committed
2918
	bool discard_supported = false;
2919 2920

	if (mddev->level != 1) {
2921
		printk(KERN_ERR "md/raid1:%s: raid level not set to mirroring (%d)\n",
2922 2923 2924 2925
		       mdname(mddev), mddev->level);
		return -EIO;
	}
	if (mddev->reshape_position != MaxSector) {
2926
		printk(KERN_ERR "md/raid1:%s: reshape_position set but not supported\n",
2927 2928 2929
		       mdname(mddev));
		return -EIO;
	}
Linus Torvalds's avatar
Linus Torvalds committed
2930
	/*
2931 2932
	 * copy the already verified devices into our private RAID1
	 * bookkeeping area. [whatever we allocate in run(),
NeilBrown's avatar
NeilBrown committed
2933
	 * should be freed in raid1_free()]
Linus Torvalds's avatar
Linus Torvalds committed
2934
	 */
2935 2936 2937 2938
	if (mddev->private == NULL)
		conf = setup_conf(mddev);
	else
		conf = mddev->private;
Linus Torvalds's avatar
Linus Torvalds committed
2939

2940 2941
	if (IS_ERR(conf))
		return PTR_ERR(conf);
Linus Torvalds's avatar
Linus Torvalds committed
2942

2943
	if (mddev->queue)
2944 2945
		blk_queue_max_write_same_sectors(mddev->queue, 0);

NeilBrown's avatar
NeilBrown committed
2946
	rdev_for_each(rdev, mddev) {
2947 2948
		if (!mddev->gendisk)
			continue;
2949 2950
		disk_stack_limits(mddev->gendisk, rdev->bdev,
				  rdev->data_offset << 9);
Shaohua Li's avatar
Shaohua Li committed
2951 2952
		if (blk_queue_discard(bdev_get_queue(rdev->bdev)))
			discard_supported = true;
Linus Torvalds's avatar
Linus Torvalds committed
2953
	}
2954

2955 2956 2957 2958 2959 2960
	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++;
2961 2962 2963 2964 2965 2966 2967
	/*
	 * RAID1 needs at least one disk in active
	 */
	if (conf->raid_disks - mddev->degraded < 1) {
		ret = -EINVAL;
		goto abort;
	}
2968 2969 2970 2971

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

2972
	if (mddev->recovery_cp != MaxSector)
2973
		printk(KERN_NOTICE "md/raid1:%s: not clean"
2974 2975
		       " -- starting background reconstruction\n",
		       mdname(mddev));
2976
	printk(KERN_INFO
2977
		"md/raid1:%s: active with %d out of %d mirrors\n",
2978
		mdname(mddev), mddev->raid_disks - mddev->degraded,
Linus Torvalds's avatar
Linus Torvalds committed
2979
		mddev->raid_disks);
2980

Linus Torvalds's avatar
Linus Torvalds committed
2981 2982 2983
	/*
	 * Ok, everything is just fine now
	 */
2984 2985 2986 2987
	mddev->thread = conf->thread;
	conf->thread = NULL;
	mddev->private = conf;

2988
	md_set_array_sectors(mddev, raid1_size(mddev, 0, 0));
Linus Torvalds's avatar
Linus Torvalds committed
2989

2990
	if (mddev->queue) {
Shaohua Li's avatar
Shaohua Li committed
2991 2992 2993 2994 2995 2996
		if (discard_supported)
			queue_flag_set_unlocked(QUEUE_FLAG_DISCARD,
						mddev->queue);
		else
			queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD,
						  mddev->queue);
2997
	}
2998 2999

	ret =  md_integrity_register(mddev);
3000 3001
	if (ret) {
		md_unregister_thread(&mddev->thread);
3002
		goto abort;
3003
	}
3004 3005 3006 3007
	return 0;

abort:
	raid1_free(mddev, conf);
3008
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
3009 3010
}

NeilBrown's avatar
NeilBrown committed
3011
static void raid1_free(struct mddev *mddev, void *priv)
Linus Torvalds's avatar
Linus Torvalds committed
3012
{
NeilBrown's avatar
NeilBrown committed
3013
	struct r1conf *conf = priv;
3014

3015
	mempool_destroy(conf->r1bio_pool);
3016
	kfree(conf->mirrors);
3017
	safe_put_page(conf->tmppage);
3018
	kfree(conf->poolinfo);
Linus Torvalds's avatar
Linus Torvalds committed
3019 3020 3021
	kfree(conf);
}

3022
static int raid1_resize(struct mddev *mddev, sector_t sectors)
Linus Torvalds's avatar
Linus Torvalds committed
3023 3024 3025 3026 3027 3028 3029 3030
{
	/* 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.
	 */
3031 3032 3033
	sector_t newsize = raid1_size(mddev, sectors, 0);
	if (mddev->external_size &&
	    mddev->array_sectors > newsize)
3034
		return -EINVAL;
3035 3036 3037 3038 3039 3040
	if (mddev->bitmap) {
		int ret = bitmap_resize(mddev->bitmap, newsize, 0, 0);
		if (ret)
			return ret;
	}
	md_set_array_sectors(mddev, newsize);
3041
	set_capacity(mddev->gendisk, mddev->array_sectors);
3042
	revalidate_disk(mddev->gendisk);
3043
	if (sectors > mddev->dev_sectors &&
3044
	    mddev->recovery_cp > mddev->dev_sectors) {
3045
		mddev->recovery_cp = mddev->dev_sectors;
Linus Torvalds's avatar
Linus Torvalds committed
3046 3047
		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
	}
3048
	mddev->dev_sectors = sectors;
3049
	mddev->resync_max_sectors = sectors;
Linus Torvalds's avatar
Linus Torvalds committed
3050 3051 3052
	return 0;
}

3053
static int raid1_reshape(struct mddev *mddev)
Linus Torvalds's avatar
Linus Torvalds committed
3054 3055 3056 3057 3058 3059 3060 3061
{
	/* 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.
3062 3063 3064
	 *
	 * At the same time, we "pack" the devices so that all the missing
	 * devices have the higher raid_disk numbers.
Linus Torvalds's avatar
Linus Torvalds committed
3065 3066 3067
	 */
	mempool_t *newpool, *oldpool;
	struct pool_info *newpoolinfo;
3068
	struct raid1_info *newmirrors;
3069
	struct r1conf *conf = mddev->private;
3070
	int cnt, raid_disks;
3071
	unsigned long flags;
3072
	int d, d2, err;
Linus Torvalds's avatar
Linus Torvalds committed
3073

3074
	/* Cannot change chunk_size, layout, or level */
3075
	if (mddev->chunk_sectors != mddev->new_chunk_sectors ||
3076 3077
	    mddev->layout != mddev->new_layout ||
	    mddev->level != mddev->new_level) {
3078
		mddev->new_chunk_sectors = mddev->chunk_sectors;
3079 3080 3081 3082 3083
		mddev->new_layout = mddev->layout;
		mddev->new_level = mddev->level;
		return -EINVAL;
	}

3084 3085 3086 3087 3088
	if (!mddev_is_clustered(mddev)) {
		err = md_allow_write(mddev);
		if (err)
			return err;
	}
3089

3090 3091
	raid_disks = mddev->raid_disks + mddev->delta_disks;

3092 3093 3094 3095 3096 3097
	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)
Linus Torvalds's avatar
Linus Torvalds committed
3098
			return -EBUSY;
3099
	}
Linus Torvalds's avatar
Linus Torvalds committed
3100 3101 3102 3103 3104

	newpoolinfo = kmalloc(sizeof(*newpoolinfo), GFP_KERNEL);
	if (!newpoolinfo)
		return -ENOMEM;
	newpoolinfo->mddev = mddev;
3105
	newpoolinfo->raid_disks = raid_disks * 2;
Linus Torvalds's avatar
Linus Torvalds committed
3106 3107 3108 3109 3110 3111 3112

	newpool = mempool_create(NR_RAID1_BIOS, r1bio_pool_alloc,
				 r1bio_pool_free, newpoolinfo);
	if (!newpool) {
		kfree(newpoolinfo);
		return -ENOMEM;
	}
3113
	newmirrors = kzalloc(sizeof(struct raid1_info) * raid_disks * 2,
3114
			     GFP_KERNEL);
Linus Torvalds's avatar
Linus Torvalds committed
3115 3116 3117 3118 3119 3120
	if (!newmirrors) {
		kfree(newpoolinfo);
		mempool_destroy(newpool);
		return -ENOMEM;
	}

3121
	freeze_array(conf, 0);
Linus Torvalds's avatar
Linus Torvalds committed
3122 3123 3124 3125

	/* ok, everything is stopped */
	oldpool = conf->r1bio_pool;
	conf->r1bio_pool = newpool;
3126

3127
	for (d = d2 = 0; d < conf->raid_disks; d++) {
3128
		struct md_rdev *rdev = conf->mirrors[d].rdev;
3129
		if (rdev && rdev->raid_disk != d2) {
3130
			sysfs_unlink_rdev(mddev, rdev);
3131
			rdev->raid_disk = d2;
3132 3133
			sysfs_unlink_rdev(mddev, rdev);
			if (sysfs_link_rdev(mddev, rdev))
3134
				printk(KERN_WARNING
3135 3136
				       "md/raid1:%s: cannot register rd%d\n",
				       mdname(mddev), rdev->raid_disk);
3137
		}
3138 3139 3140
		if (rdev)
			newmirrors[d2++].rdev = rdev;
	}
Linus Torvalds's avatar
Linus Torvalds committed
3141 3142 3143 3144 3145
	kfree(conf->mirrors);
	conf->mirrors = newmirrors;
	kfree(conf->poolinfo);
	conf->poolinfo = newpoolinfo;

3146
	spin_lock_irqsave(&conf->device_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
3147
	mddev->degraded += (raid_disks - conf->raid_disks);
3148
	spin_unlock_irqrestore(&conf->device_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
3149
	conf->raid_disks = mddev->raid_disks = raid_disks;
3150
	mddev->delta_disks = 0;
Linus Torvalds's avatar
Linus Torvalds committed
3151

3152
	unfreeze_array(conf);
Linus Torvalds's avatar
Linus Torvalds committed
3153

3154
	set_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
Linus Torvalds's avatar
Linus Torvalds committed
3155 3156 3157 3158 3159 3160 3161
	set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
	md_wakeup_thread(mddev->thread);

	mempool_destroy(oldpool);
	return 0;
}

3162
static void raid1_quiesce(struct mddev *mddev, int state)
3163
{
3164
	struct r1conf *conf = mddev->private;
3165 3166

	switch(state) {
3167 3168 3169
	case 2: /* wake for suspend */
		wake_up(&conf->wait_barrier);
		break;
3170
	case 1:
3171
		freeze_array(conf, 0);
3172
		break;
3173
	case 0:
3174
		unfreeze_array(conf);
3175 3176 3177 3178
		break;
	}
}

3179
static void *raid1_takeover(struct mddev *mddev)
3180 3181 3182 3183 3184
{
	/* raid1 can take over:
	 *  raid5 with 2 devices, any layout or chunk size
	 */
	if (mddev->level == 5 && mddev->raid_disks == 2) {
3185
		struct r1conf *conf;
3186 3187 3188 3189 3190
		mddev->new_level = 1;
		mddev->new_layout = 0;
		mddev->new_chunk_sectors = 0;
		conf = setup_conf(mddev);
		if (!IS_ERR(conf))
3191 3192
			/* Array must appear to be quiesced */
			conf->array_frozen = 1;
3193 3194 3195 3196
		return conf;
	}
	return ERR_PTR(-EINVAL);
}
Linus Torvalds's avatar
Linus Torvalds committed
3197

3198
static struct md_personality raid1_personality =
Linus Torvalds's avatar
Linus Torvalds committed
3199 3200
{
	.name		= "raid1",
3201
	.level		= 1,
Linus Torvalds's avatar
Linus Torvalds committed
3202 3203 3204
	.owner		= THIS_MODULE,
	.make_request	= make_request,
	.run		= run,
NeilBrown's avatar
NeilBrown committed
3205
	.free		= raid1_free,
Linus Torvalds's avatar
Linus Torvalds committed
3206 3207 3208 3209 3210 3211 3212
	.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,
3213
	.size		= raid1_size,
3214
	.check_reshape	= raid1_reshape,
3215
	.quiesce	= raid1_quiesce,
3216
	.takeover	= raid1_takeover,
3217
	.congested	= raid1_congested,
Linus Torvalds's avatar
Linus Torvalds committed
3218 3219 3220 3221
};

static int __init raid_init(void)
{
3222
	return register_md_personality(&raid1_personality);
Linus Torvalds's avatar
Linus Torvalds committed
3223 3224 3225 3226
}

static void raid_exit(void)
{
3227
	unregister_md_personality(&raid1_personality);
Linus Torvalds's avatar
Linus Torvalds committed
3228 3229 3230 3231 3232
}

module_init(raid_init);
module_exit(raid_exit);
MODULE_LICENSE("GPL");
3233
MODULE_DESCRIPTION("RAID1 (mirroring) personality for MD");
Linus Torvalds's avatar
Linus Torvalds committed
3234
MODULE_ALIAS("md-personality-3"); /* RAID1 */
3235
MODULE_ALIAS("md-raid1");
3236
MODULE_ALIAS("md-level-1");
3237 3238

module_param(max_queued_requests, int, S_IRUGO|S_IWUSR);