raid1.c 87.3 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 93
#define RESYNC_WINDOW (RESYNC_BLOCK_SIZE * RESYNC_DEPTH)
#define RESYNC_WINDOW_SECTORS (RESYNC_WINDOW >> 9)
#define NEXT_NORMALIO_DISTANCE (3 * RESYNC_WINDOW_SECTORS)
Linus Torvalds's avatar
Linus Torvalds committed
94

95
static void * r1buf_pool_alloc(gfp_t gfp_flags, void *data)
Linus Torvalds's avatar
Linus Torvalds committed
96 97
{
	struct pool_info *pi = data;
98
	struct r1bio *r1_bio;
Linus Torvalds's avatar
Linus Torvalds committed
99
	struct bio *bio;
100
	int need_pages;
Linus Torvalds's avatar
Linus Torvalds committed
101 102 103
	int i, j;

	r1_bio = r1bio_pool_alloc(gfp_flags, pi);
104
	if (!r1_bio)
Linus Torvalds's avatar
Linus Torvalds committed
105 106 107 108 109 110
		return NULL;

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

130
		if (bio_alloc_pages(bio, gfp_flags))
131
			goto out_free_pages;
132 133 134 135 136 137 138
	}
	/* If not user-requests, copy the page pointers to all bios */
	if (!test_bit(MD_RECOVERY_REQUESTED, &pi->mddev->recovery)) {
		for (i=0; i<RESYNC_PAGES ; i++)
			for (j=1; j<pi->raid_disks; j++)
				r1_bio->bios[j]->bi_io_vec[i].bv_page =
					r1_bio->bios[0]->bi_io_vec[i].bv_page;
Linus Torvalds's avatar
Linus Torvalds committed
139 140 141 142 143 144
	}

	r1_bio->master_bio = NULL;

	return r1_bio;

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

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

Linus Torvalds's avatar
Linus Torvalds committed
153
out_free_bio:
154
	while (++j < pi->raid_disks)
Linus Torvalds's avatar
Linus Torvalds committed
155 156 157 158 159 160 161 162
		bio_put(r1_bio->bios[j]);
	r1bio_pool_free(r1_bio, data);
	return NULL;
}

static void r1buf_pool_free(void *__r1_bio, void *data)
{
	struct pool_info *pi = data;
163
	int i,j;
164
	struct r1bio *r1bio = __r1_bio;
Linus Torvalds's avatar
Linus Torvalds committed
165

166 167 168 169 170
	for (i = 0; i < RESYNC_PAGES; i++)
		for (j = pi->raid_disks; j-- ;) {
			if (j == 0 ||
			    r1bio->bios[j]->bi_io_vec[i].bv_page !=
			    r1bio->bios[0]->bi_io_vec[i].bv_page)
171
				safe_put_page(r1bio->bios[j]->bi_io_vec[i].bv_page);
172
		}
Linus Torvalds's avatar
Linus Torvalds committed
173 174 175 176 177 178
	for (i=0 ; i < pi->raid_disks; i++)
		bio_put(r1bio->bios[i]);

	r1bio_pool_free(r1bio, data);
}

179
static void put_all_bios(struct r1conf *conf, struct r1bio *r1_bio)
Linus Torvalds's avatar
Linus Torvalds committed
180 181 182
{
	int i;

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

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

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

199
static void put_buf(struct r1bio *r1_bio)
Linus Torvalds's avatar
Linus Torvalds committed
200
{
201
	struct r1conf *conf = r1_bio->mddev->private;
202 203
	int i;

204
	for (i = 0; i < conf->raid_disks * 2; i++) {
205 206 207 208
		struct bio *bio = r1_bio->bios[i];
		if (bio->bi_end_io)
			rdev_dec_pending(conf->mirrors[i].rdev, r1_bio->mddev);
	}
Linus Torvalds's avatar
Linus Torvalds committed
209 210 211

	mempool_free(r1_bio, conf->r1buf_pool);

212
	lower_barrier(conf);
Linus Torvalds's avatar
Linus Torvalds committed
213 214
}

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

	spin_lock_irqsave(&conf->device_lock, flags);
	list_add(&r1_bio->retry_list, &conf->retry_list);
223
	conf->nr_queued ++;
Linus Torvalds's avatar
Linus Torvalds committed
224 225
	spin_unlock_irqrestore(&conf->device_lock, flags);

226
	wake_up(&conf->wait_barrier);
Linus Torvalds's avatar
Linus Torvalds committed
227 228 229 230 231 232 233 234
	md_wakeup_thread(mddev->thread);
}

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

	if (bio->bi_phys_segments) {
		unsigned long flags;
		spin_lock_irqsave(&conf->device_lock, flags);
		bio->bi_phys_segments--;
		done = (bio->bi_phys_segments == 0);
		spin_unlock_irqrestore(&conf->device_lock, flags);
249 250 251 252 253
		/*
		 * make_request() might be waiting for
		 * bi_phys_segments to decrease
		 */
		wake_up(&conf->wait_barrier);
254 255 256 257 258 259 260 261 262 263 264
	} else
		done = 1;

	if (!test_bit(R1BIO_Uptodate, &r1_bio->state))
		clear_bit(BIO_UPTODATE, &bio->bi_flags);
	if (done) {
		bio_endio(bio, 0);
		/*
		 * Wake up any possible resync thread that waits for the device
		 * to go idle.
		 */
265
		allow_barrier(conf, start_next_window, bi_sector);
266 267 268
	}
}

269
static void raid_end_bio_io(struct r1bio *r1_bio)
Linus Torvalds's avatar
Linus Torvalds committed
270 271 272
{
	struct bio *bio = r1_bio->master_bio;

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

280
		call_bio_endio(r1_bio);
281
	}
Linus Torvalds's avatar
Linus Torvalds committed
282 283 284 285 286 287
	free_r1bio(r1_bio);
}

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

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

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

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

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

	return mirror;
}

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

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

328 329 330 331 332 333
	if (uptodate)
		set_bit(R1BIO_Uptodate, &r1_bio->state);
	else {
		/* If all other devices have failed, we want to return
		 * the error upwards rather than fail the last device.
		 * Here we redefine "uptodate" to mean "Don't want to retry"
Linus Torvalds's avatar
Linus Torvalds committed
334
		 */
335 336 337 338
		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 &&
339
		     test_bit(In_sync, &conf->mirrors[mirror].rdev->flags)))
340 341 342
			uptodate = 1;
		spin_unlock_irqrestore(&conf->device_lock, flags);
	}
Linus Torvalds's avatar
Linus Torvalds committed
343

344
	if (uptodate) {
Linus Torvalds's avatar
Linus Torvalds committed
345
		raid_end_bio_io(r1_bio);
346 347
		rdev_dec_pending(conf->mirrors[mirror].rdev, conf->mddev);
	} else {
Linus Torvalds's avatar
Linus Torvalds committed
348 349 350 351
		/*
		 * oops, read error:
		 */
		char b[BDEVNAME_SIZE];
352 353 354 355 356 357 358
		printk_ratelimited(
			KERN_ERR "md/raid1:%s: %s: "
			"rescheduling sector %llu\n",
			mdname(conf->mddev),
			bdevname(conf->mirrors[mirror].rdev->bdev,
				 b),
			(unsigned long long)r1_bio->sector);
359
		set_bit(R1BIO_ReadError, &r1_bio->state);
Linus Torvalds's avatar
Linus Torvalds committed
360
		reschedule_retry(r1_bio);
361
		/* don't drop the reference on read_disk yet */
Linus Torvalds's avatar
Linus Torvalds committed
362 363 364
	}
}

365
static void close_write(struct r1bio *r1_bio)
366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383
{
	/* it really is the end of this request */
	if (test_bit(R1BIO_BehindIO, &r1_bio->state)) {
		/* free extra copy of the data pages */
		int i = r1_bio->behind_page_count;
		while (i--)
			safe_put_page(r1_bio->behind_bvecs[i].bv_page);
		kfree(r1_bio->behind_bvecs);
		r1_bio->behind_bvecs = NULL;
	}
	/* clear the bitmap if all writes complete successfully */
	bitmap_endwrite(r1_bio->mddev->bitmap, r1_bio->sector,
			r1_bio->sectors,
			!test_bit(R1BIO_Degraded, &r1_bio->state),
			test_bit(R1BIO_BehindIO, &r1_bio->state));
	md_write_end(r1_bio->mddev);
}

384
static void r1_bio_write_done(struct r1bio *r1_bio)
385
{
386 387 388 389 390 391 392
	if (!atomic_dec_and_test(&r1_bio->remaining))
		return;

	if (test_bit(R1BIO_WriteError, &r1_bio->state))
		reschedule_retry(r1_bio);
	else {
		close_write(r1_bio);
393 394 395 396
		if (test_bit(R1BIO_MadeGood, &r1_bio->state))
			reschedule_retry(r1_bio);
		else
			raid_end_bio_io(r1_bio);
397 398 399
	}
}

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

408
	mirror = find_bio_disk(r1_bio, bio);
Linus Torvalds's avatar
Linus Torvalds committed
409

410 411 412 413
	/*
	 * 'one mirror IO has finished' event handler:
	 */
	if (!uptodate) {
414 415
		set_bit(WriteErrorSeen,
			&conf->mirrors[mirror].rdev->flags);
416 417 418 419 420
		if (!test_and_set_bit(WantReplacement,
				      &conf->mirrors[mirror].rdev->flags))
			set_bit(MD_RECOVERY_NEEDED, &
				conf->mddev->recovery);

421
		set_bit(R1BIO_WriteError, &r1_bio->state);
422
	} else {
Linus Torvalds's avatar
Linus Torvalds committed
423
		/*
424 425 426 427 428 429 430 431
		 * Set R1BIO_Uptodate in our master bio, so that we
		 * will return a good error code for to the higher
		 * levels even if IO on some other mirrored buffer
		 * fails.
		 *
		 * The 'master' represents the composite IO operation
		 * to user-side. So if something waits for IO, then it
		 * will wait for the 'master' bio.
Linus Torvalds's avatar
Linus Torvalds committed
432
		 */
433 434 435
		sector_t first_bad;
		int bad_sectors;

436 437
		r1_bio->bios[mirror] = NULL;
		to_put = bio;
438 439 440 441 442 443 444 445 446 447 448
		/*
		 * Do not set R1BIO_Uptodate if the current device is
		 * rebuilding or Faulty. This is because we cannot use
		 * such device for properly reading the data back (we could
		 * potentially use it, if the current write would have felt
		 * before rdev->recovery_offset, but for simplicity we don't
		 * check this here.
		 */
		if (test_bit(In_sync, &conf->mirrors[mirror].rdev->flags) &&
		    !test_bit(Faulty, &conf->mirrors[mirror].rdev->flags))
			set_bit(R1BIO_Uptodate, &r1_bio->state);
449

450 451 452 453 454 455 456 457 458
		/* Maybe we can clear some bad blocks. */
		if (is_badblock(conf->mirrors[mirror].rdev,
				r1_bio->sector, r1_bio->sectors,
				&first_bad, &bad_sectors)) {
			r1_bio->bios[mirror] = IO_MADE_GOOD;
			set_bit(R1BIO_MadeGood, &r1_bio->state);
		}
	}

459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474
	if (behind) {
		if (test_bit(WriteMostly, &conf->mirrors[mirror].rdev->flags))
			atomic_dec(&r1_bio->behind_remaining);

		/*
		 * In behind mode, we ACK the master bio once the I/O
		 * has safely reached all non-writemostly
		 * disks. Setting the Returned bit ensures that this
		 * gets done only once -- we don't ever want to return
		 * -EIO here, instead we'll wait
		 */
		if (atomic_read(&r1_bio->behind_remaining) >= (atomic_read(&r1_bio->remaining)-1) &&
		    test_bit(R1BIO_Uptodate, &r1_bio->state)) {
			/* Maybe we can return now */
			if (!test_and_set_bit(R1BIO_Returned, &r1_bio->state)) {
				struct bio *mbio = r1_bio->master_bio;
475 476
				pr_debug("raid1: behind end write sectors"
					 " %llu-%llu\n",
477 478
					 (unsigned long long) mbio->bi_iter.bi_sector,
					 (unsigned long long) bio_end_sector(mbio) - 1);
479
				call_bio_endio(r1_bio);
480 481 482
			}
		}
	}
483 484 485
	if (r1_bio->bios[mirror] == NULL)
		rdev_dec_pending(conf->mirrors[mirror].rdev,
				 conf->mddev);
486

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

493 494
	if (to_put)
		bio_put(to_put);
Linus Torvalds's avatar
Linus Torvalds committed
495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510
}

/*
 * This routine returns the disk from which the requested read should
 * be done. There is a per-array 'next expected sequential IO' sector
 * number - if this matches on the next IO then we use the last disk.
 * There is also a per-disk 'last know head position' sector that is
 * maintained from IRQ contexts, both the normal and the resync IO
 * completion handlers update this position correctly. If there is no
 * perfect sequential match then we pick the disk whose head is closest.
 *
 * If there are 2 mirrors in the same 2 devices, performance degrades
 * because position is mirror, not device based.
 *
 * The rdev for the device selected will have nr_pending incremented.
 */
511
static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sectors)
Linus Torvalds's avatar
Linus Torvalds committed
512
{
513
	const sector_t this_sector = r1_bio->sector;
514 515
	int sectors;
	int best_good_sectors;
516 517
	int best_disk, best_dist_disk, best_pending_disk;
	int has_nonrot_disk;
518
	int disk;
NeilBrown's avatar
NeilBrown committed
519
	sector_t best_dist;
520
	unsigned int min_pending;
521
	struct md_rdev *rdev;
522
	int choose_first;
523
	int choose_next_idle;
Linus Torvalds's avatar
Linus Torvalds committed
524 525 526

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

542
	choose_first = (conf->mddev->recovery_cp < this_sector + sectors);
Linus Torvalds's avatar
Linus Torvalds committed
543

544
	for (disk = 0 ; disk < conf->raid_disks * 2 ; disk++) {
NeilBrown's avatar
NeilBrown committed
545
		sector_t dist;
546 547
		sector_t first_bad;
		int bad_sectors;
548
		unsigned int pending;
549
		bool nonrot;
550

551 552 553
		rdev = rcu_dereference(conf->mirrors[disk].rdev);
		if (r1_bio->bios[disk] == IO_BLOCKED
		    || rdev == NULL
554
		    || test_bit(Unmerged, &rdev->flags)
NeilBrown's avatar
NeilBrown committed
555
		    || test_bit(Faulty, &rdev->flags))
556
			continue;
NeilBrown's avatar
NeilBrown committed
557 558
		if (!test_bit(In_sync, &rdev->flags) &&
		    rdev->recovery_offset < this_sector + sectors)
Linus Torvalds's avatar
Linus Torvalds committed
559
			continue;
NeilBrown's avatar
NeilBrown committed
560 561 562
		if (test_bit(WriteMostly, &rdev->flags)) {
			/* Don't balance among write-mostly, just
			 * use the first as a last resort */
563
			if (best_dist_disk < 0) {
564 565
				if (is_badblock(rdev, this_sector, sectors,
						&first_bad, &bad_sectors)) {
566
					if (first_bad <= this_sector)
567 568 569 570 571
						/* Cannot use this */
						continue;
					best_good_sectors = first_bad - this_sector;
				} else
					best_good_sectors = sectors;
572 573
				best_dist_disk = disk;
				best_pending_disk = disk;
574
			}
NeilBrown's avatar
NeilBrown committed
575 576 577 578 579
			continue;
		}
		/* This is a reasonable device to use.  It might
		 * even be best.
		 */
580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608
		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;

609 610
		nonrot = blk_queue_nonrot(bdev_get_queue(rdev->bdev));
		has_nonrot_disk |= nonrot;
611
		pending = atomic_read(&rdev->nr_pending);
NeilBrown's avatar
NeilBrown committed
612
		dist = abs(this_sector - conf->mirrors[disk].head_position);
613
		if (choose_first) {
NeilBrown's avatar
NeilBrown committed
614
			best_disk = disk;
Linus Torvalds's avatar
Linus Torvalds committed
615 616
			break;
		}
617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654
		/* 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;
655 656 657 658 659 660

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

NeilBrown's avatar
NeilBrown committed
661 662
		if (dist < best_dist) {
			best_dist = dist;
663
			best_dist_disk = disk;
Linus Torvalds's avatar
Linus Torvalds committed
664
		}
665
	}
Linus Torvalds's avatar
Linus Torvalds committed
666

667 668 669 670 671 672 673 674 675 676 677 678 679
	/*
	 * 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
680 681
	if (best_disk >= 0) {
		rdev = rcu_dereference(conf->mirrors[best_disk].rdev);
682 683 684
		if (!rdev)
			goto retry;
		atomic_inc(&rdev->nr_pending);
NeilBrown's avatar
NeilBrown committed
685
		if (test_bit(Faulty, &rdev->flags)) {
Linus Torvalds's avatar
Linus Torvalds committed
686 687 688
			/* cannot risk returning a device that failed
			 * before we inc'ed nr_pending
			 */
689
			rdev_dec_pending(rdev, conf->mddev);
Linus Torvalds's avatar
Linus Torvalds committed
690 691
			goto retry;
		}
692
		sectors = best_good_sectors;
693 694 695 696

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

697
		conf->mirrors[best_disk].next_seq_sect = this_sector + sectors;
Linus Torvalds's avatar
Linus Torvalds committed
698 699
	}
	rcu_read_unlock();
700
	*max_sectors = sectors;
Linus Torvalds's avatar
Linus Torvalds committed
701

NeilBrown's avatar
NeilBrown committed
702
	return best_disk;
Linus Torvalds's avatar
Linus Torvalds committed
703 704
}

705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737
static int raid1_mergeable_bvec(struct request_queue *q,
				struct bvec_merge_data *bvm,
				struct bio_vec *biovec)
{
	struct mddev *mddev = q->queuedata;
	struct r1conf *conf = mddev->private;
	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
	int max = biovec->bv_len;

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

}

738
int md_raid1_congested(struct mddev *mddev, int bits)
739
{
740
	struct r1conf *conf = mddev->private;
741 742
	int i, ret = 0;

743 744 745 746
	if ((bits & (1 << BDI_async_congested)) &&
	    conf->pending_count >= max_queued_requests)
		return 1;

747
	rcu_read_lock();
748
	for (i = 0; i < conf->raid_disks * 2; i++) {
749
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
750
		if (rdev && !test_bit(Faulty, &rdev->flags)) {
751
			struct request_queue *q = bdev_get_queue(rdev->bdev);
752

753 754
			BUG_ON(!q);

755 756 757
			/* Note the '|| 1' - when read_balance prefers
			 * non-congested targets, it can be removed
			 */
758
			if ((bits & (1<<BDI_async_congested)) || 1)
759 760 761 762 763 764 765 766
				ret |= bdi_congested(&q->backing_dev_info, bits);
			else
				ret &= bdi_congested(&q->backing_dev_info, bits);
		}
	}
	rcu_read_unlock();
	return ret;
}
767
EXPORT_SYMBOL_GPL(md_raid1_congested);
768

769 770
static int raid1_congested(void *data, int bits)
{
771
	struct mddev *mddev = data;
772 773 774 775

	return mddev_congested(mddev, bits) ||
		md_raid1_congested(mddev, bits);
}
776

777
static void flush_pending_writes(struct r1conf *conf)
778 779 780 781 782 783 784 785 786
{
	/* 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);
787
		conf->pending_count = 0;
788 789 790 791
		spin_unlock_irq(&conf->device_lock);
		/* flush any pending bitmap writes to
		 * disk before proceeding w/ I/O */
		bitmap_unplug(conf->mddev->bitmap);
792
		wake_up(&conf->wait_barrier);
793 794 795 796

		while (bio) { /* submit pending writes */
			struct bio *next = bio->bi_next;
			bio->bi_next = NULL;
Shaohua Li's avatar
Shaohua Li committed
797 798 799 800 801 802
			if (unlikely((bio->bi_rw & REQ_DISCARD) &&
			    !blk_queue_discard(bdev_get_queue(bio->bi_bdev))))
				/* Just ignore it */
				bio_endio(bio, 0);
			else
				generic_make_request(bio);
803 804 805 806
			bio = next;
		}
	} else
		spin_unlock_irq(&conf->device_lock);
807 808
}

809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828
/* 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
829
 */
830
static void raise_barrier(struct r1conf *conf, sector_t sector_nr)
Linus Torvalds's avatar
Linus Torvalds committed
831 832
{
	spin_lock_irq(&conf->resync_lock);
833 834 835

	/* Wait until no block IO is waiting */
	wait_event_lock_irq(conf->wait_barrier, !conf->nr_waiting,
836
			    conf->resync_lock);
837 838 839

	/* block any new IO from starting */
	conf->barrier++;
840
	conf->next_resync = sector_nr;
841

842 843 844 845 846 847 848
	/* 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.
849
	 * D: while there are any active requests in the current window.
850
	 */
851
	wait_event_lock_irq(conf->wait_barrier,
852
			    !conf->array_frozen &&
853
			    conf->barrier < RESYNC_DEPTH &&
854
			    conf->current_window_requests == 0 &&
855 856
			    (conf->start_next_window >=
			     conf->next_resync + RESYNC_SECTORS),
857
			    conf->resync_lock);
858

859
	conf->nr_pending++;
860 861 862
	spin_unlock_irq(&conf->resync_lock);
}

863
static void lower_barrier(struct r1conf *conf)
864 865
{
	unsigned long flags;
866
	BUG_ON(conf->barrier <= 0);
867 868
	spin_lock_irqsave(&conf->resync_lock, flags);
	conf->barrier--;
869
	conf->nr_pending--;
870 871 872 873
	spin_unlock_irqrestore(&conf->resync_lock, flags);
	wake_up(&conf->wait_barrier);
}

874
static bool need_to_wait_for_sync(struct r1conf *conf, struct bio *bio)
875
{
876 877 878 879 880
	bool wait = false;

	if (conf->array_frozen || !bio)
		wait = true;
	else if (conf->barrier && bio_data_dir(bio) == WRITE) {
881 882 883 884
		if ((conf->mddev->curr_resync_completed
		     >= bio_end_sector(bio)) ||
		    (conf->next_resync + NEXT_NORMALIO_DISTANCE
		     <= bio->bi_iter.bi_sector))
885 886 887 888 889 890 891 892 893 894 895 896
			wait = false;
		else
			wait = true;
	}

	return wait;
}

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

897
	spin_lock_irq(&conf->resync_lock);
898
	if (need_to_wait_for_sync(conf, bio)) {
899
		conf->nr_waiting++;
900 901 902 903
		/* Wait for the barrier to drop.
		 * However if there are already pending
		 * requests (preventing the barrier from
		 * rising completely), and the
904
		 * per-process bio queue isn't empty,
905
		 * then don't wait, as we need to empty
906 907
		 * that queue to allow conf->start_next_window
		 * to increase.
908 909
		 */
		wait_event_lock_irq(conf->wait_barrier,
910 911
				    !conf->array_frozen &&
				    (!conf->barrier ||
912 913 914 915
				     ((conf->start_next_window <
				       conf->next_resync + RESYNC_SECTORS) &&
				      current->bio_list &&
				      !bio_list_empty(current->bio_list))),
916
				    conf->resync_lock);
917
		conf->nr_waiting--;
Linus Torvalds's avatar
Linus Torvalds committed
918
	}
919 920

	if (bio && bio_data_dir(bio) == WRITE) {
921
		if (bio->bi_iter.bi_sector >=
922
		    conf->mddev->curr_resync_completed) {
923 924 925 926 927 928
			if (conf->start_next_window == MaxSector)
				conf->start_next_window =
					conf->next_resync +
					NEXT_NORMALIO_DISTANCE;

			if ((conf->start_next_window + NEXT_NORMALIO_DISTANCE)
929
			    <= bio->bi_iter.bi_sector)
930 931 932 933
				conf->next_window_requests++;
			else
				conf->current_window_requests++;
			sector = conf->start_next_window;
934
		}
935 936
	}

937
	conf->nr_pending++;
Linus Torvalds's avatar
Linus Torvalds committed
938
	spin_unlock_irq(&conf->resync_lock);
939
	return sector;
Linus Torvalds's avatar
Linus Torvalds committed
940 941
}

942 943
static void allow_barrier(struct r1conf *conf, sector_t start_next_window,
			  sector_t bi_sector)
944 945
{
	unsigned long flags;
946

947 948
	spin_lock_irqsave(&conf->resync_lock, flags);
	conf->nr_pending--;
949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969
	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;
		}
	}
970 971 972 973
	spin_unlock_irqrestore(&conf->resync_lock, flags);
	wake_up(&conf->wait_barrier);
}

974
static void freeze_array(struct r1conf *conf, int extra)
975 976 977
{
	/* stop syncio and normal IO and wait for everything to
	 * go quite.
978
	 * We wait until nr_pending match nr_queued+extra
979 980 981 982
	 * 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.
983
	 * Thus the number queued (nr_queued) plus this request (extra)
984 985
	 * must match the number of pending IOs (nr_pending) before
	 * we continue.
986 987
	 */
	spin_lock_irq(&conf->resync_lock);
988
	conf->array_frozen = 1;
989
	wait_event_lock_irq_cmd(conf->wait_barrier,
990
				conf->nr_pending == conf->nr_queued+extra,
991 992
				conf->resync_lock,
				flush_pending_writes(conf));
993 994
	spin_unlock_irq(&conf->resync_lock);
}
995
static void unfreeze_array(struct r1conf *conf)
996 997 998
{
	/* reverse the effect of the freeze */
	spin_lock_irq(&conf->resync_lock);
999
	conf->array_frozen = 0;
1000 1001 1002 1003
	wake_up(&conf->wait_barrier);
	spin_unlock_irq(&conf->resync_lock);
}

1004
/* duplicate the data pages for behind I/O
1005
 */
1006
static void alloc_behind_pages(struct bio *bio, struct r1bio *r1_bio)
1007 1008 1009
{
	int i;
	struct bio_vec *bvec;
1010
	struct bio_vec *bvecs = kzalloc(bio->bi_vcnt * sizeof(struct bio_vec),
1011
					GFP_NOIO);
1012
	if (unlikely(!bvecs))
1013
		return;
1014

1015
	bio_for_each_segment_all(bvec, bio, i) {
1016 1017 1018
		bvecs[i] = *bvec;
		bvecs[i].bv_page = alloc_page(GFP_NOIO);
		if (unlikely(!bvecs[i].bv_page))
1019
			goto do_sync_io;
1020 1021 1022
		memcpy(kmap(bvecs[i].bv_page) + bvec->bv_offset,
		       kmap(bvec->bv_page) + bvec->bv_offset, bvec->bv_len);
		kunmap(bvecs[i].bv_page);
1023 1024
		kunmap(bvec->bv_page);
	}
1025
	r1_bio->behind_bvecs = bvecs;
1026 1027 1028
	r1_bio->behind_page_count = bio->bi_vcnt;
	set_bit(R1BIO_BehindIO, &r1_bio->state);
	return;
1029 1030

do_sync_io:
1031
	for (i = 0; i < bio->bi_vcnt; i++)
1032 1033 1034
		if (bvecs[i].bv_page)
			put_page(bvecs[i].bv_page);
	kfree(bvecs);
1035 1036
	pr_debug("%dB behind alloc failed, doing sync I/O\n",
		 bio->bi_iter.bi_size);
1037 1038
}

1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052
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;

1053
	if (from_schedule || current->bio_list) {
1054 1055 1056 1057
		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);
1058
		wake_up(&conf->wait_barrier);
1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071
		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;
1072 1073 1074 1075 1076 1077
		if (unlikely((bio->bi_rw & REQ_DISCARD) &&
		    !blk_queue_discard(bdev_get_queue(bio->bi_bdev))))
			/* Just ignore it */
			bio_endio(bio, 0);
		else
			generic_make_request(bio);
1078 1079 1080 1081 1082
		bio = next;
	}
	kfree(plug);
}

1083
static void make_request(struct mddev *mddev, struct bio * bio)
Linus Torvalds's avatar
Linus Torvalds committed
1084
{
1085
	struct r1conf *conf = mddev->private;
1086
	struct raid1_info *mirror;
1087
	struct r1bio *r1_bio;
Linus Torvalds's avatar
Linus Torvalds committed
1088
	struct bio *read_bio;
1089
	int i, disks;
1090
	struct bitmap *bitmap;
1091
	unsigned long flags;
1092
	const int rw = bio_data_dir(bio);
1093
	const unsigned long do_sync = (bio->bi_rw & REQ_SYNC);
1094
	const unsigned long do_flush_fua = (bio->bi_rw & (REQ_FLUSH | REQ_FUA));
Shaohua Li's avatar
Shaohua Li committed
1095 1096
	const unsigned long do_discard = (bio->bi_rw
					  & (REQ_DISCARD | REQ_SECURE));
1097
	const unsigned long do_same = (bio->bi_rw & REQ_WRITE_SAME);
1098
	struct md_rdev *blocked_rdev;
1099 1100
	struct blk_plug_cb *cb;
	struct raid1_plug_cb *plug = NULL;
1101 1102 1103
	int first_clone;
	int sectors_handled;
	int max_sectors;
1104
	sector_t start_next_window;
1105

Linus Torvalds's avatar
Linus Torvalds committed
1106 1107 1108 1109 1110
	/*
	 * 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.
	 */
1111

1112 1113
	md_write_start(mddev, bio); /* wait on superblock update early */

1114
	if (bio_data_dir(bio) == WRITE &&
1115
	    bio_end_sector(bio) > mddev->suspend_lo &&
1116
	    bio->bi_iter.bi_sector < mddev->suspend_hi) {
1117 1118 1119 1120 1121 1122
		/* As the suspend_* range is controlled by
		 * userspace, we want an interruptible
		 * wait.
		 */
		DEFINE_WAIT(w);
		for (;;) {
1123
			sigset_t full, old;
1124 1125
			prepare_to_wait(&conf->wait_barrier,
					&w, TASK_INTERRUPTIBLE);
1126
			if (bio_end_sector(bio) <= mddev->suspend_lo ||
1127
			    bio->bi_iter.bi_sector >= mddev->suspend_hi)
1128
				break;
1129 1130
			sigfillset(&full);
			sigprocmask(SIG_BLOCK, &full, &old);
1131
			schedule();
1132
			sigprocmask(SIG_SETMASK, &old, NULL);
1133 1134 1135
		}
		finish_wait(&conf->wait_barrier, &w);
	}
1136

1137
	start_next_window = wait_barrier(conf, bio);
Linus Torvalds's avatar
Linus Torvalds committed
1138

1139 1140
	bitmap = mddev->bitmap;

Linus Torvalds's avatar
Linus Torvalds committed
1141 1142 1143 1144 1145 1146 1147 1148
	/*
	 * 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;
1149
	r1_bio->sectors = bio_sectors(bio);
1150
	r1_bio->state = 0;
Linus Torvalds's avatar
Linus Torvalds committed
1151
	r1_bio->mddev = mddev;
1152
	r1_bio->sector = bio->bi_iter.bi_sector;
Linus Torvalds's avatar
Linus Torvalds committed
1153

1154 1155 1156 1157 1158 1159 1160 1161 1162 1163
	/* We might need to issue multiple reads to different
	 * devices if there are bad blocks around, so we keep
	 * track of the number of reads in bio->bi_phys_segments.
	 * If this is 0, there is only one r1_bio and no locking
	 * will be needed when requests complete.  If it is
	 * non-zero, then it is the number of not-completed requests.
	 */
	bio->bi_phys_segments = 0;
	clear_bit(BIO_SEG_VALID, &bio->bi_flags);

1164
	if (rw == READ) {
Linus Torvalds's avatar
Linus Torvalds committed
1165 1166 1167
		/*
		 * read balancing logic:
		 */
1168 1169 1170 1171
		int rdisk;

read_again:
		rdisk = read_balance(conf, r1_bio, &max_sectors);
Linus Torvalds's avatar
Linus Torvalds committed
1172 1173 1174 1175

		if (rdisk < 0) {
			/* couldn't find anywhere to read from */
			raid_end_bio_io(r1_bio);
1176
			return;
Linus Torvalds's avatar
Linus Torvalds committed
1177 1178 1179
		}
		mirror = conf->mirrors + rdisk;

1180 1181 1182 1183 1184 1185 1186 1187 1188
		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
1189
		r1_bio->read_disk = rdisk;
1190
		r1_bio->start_next_window = 0;
Linus Torvalds's avatar
Linus Torvalds committed
1191

1192
		read_bio = bio_clone_mddev(bio, GFP_NOIO, mddev);
1193
		bio_trim(read_bio, r1_bio->sector - bio->bi_iter.bi_sector,
1194
			 max_sectors);
Linus Torvalds's avatar
Linus Torvalds committed
1195 1196 1197

		r1_bio->bios[rdisk] = read_bio;

1198 1199
		read_bio->bi_iter.bi_sector = r1_bio->sector +
			mirror->rdev->data_offset;
Linus Torvalds's avatar
Linus Torvalds committed
1200 1201
		read_bio->bi_bdev = mirror->rdev->bdev;
		read_bio->bi_end_io = raid1_end_read_request;
1202
		read_bio->bi_rw = READ | do_sync;
Linus Torvalds's avatar
Linus Torvalds committed
1203 1204
		read_bio->bi_private = r1_bio;

1205 1206 1207 1208 1209 1210
		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
1211
					   - bio->bi_iter.bi_sector);
1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228
			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;
1229
			r1_bio->sectors = bio_sectors(bio) - sectors_handled;
1230 1231
			r1_bio->state = 0;
			r1_bio->mddev = mddev;
1232 1233
			r1_bio->sector = bio->bi_iter.bi_sector +
				sectors_handled;
1234 1235 1236
			goto read_again;
		} else
			generic_make_request(read_bio);
1237
		return;
Linus Torvalds's avatar
Linus Torvalds committed
1238 1239 1240 1241 1242
	}

	/*
	 * WRITE:
	 */
1243 1244 1245 1246 1247
	if (conf->pending_count >= max_queued_requests) {
		md_wakeup_thread(mddev->thread);
		wait_event(conf->wait_barrier,
			   conf->pending_count < max_queued_requests);
	}
1248
	/* first select target devices under rcu_lock and
Linus Torvalds's avatar
Linus Torvalds committed
1249 1250
	 * inc refcount on their rdev.  Record them by setting
	 * bios[x] to bio
1251 1252 1253 1254 1255 1256
	 * 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
1257
	 */
1258

1259
	disks = conf->raid_disks * 2;
1260
 retry_write:
1261
	r1_bio->start_next_window = start_next_window;
1262
	blocked_rdev = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
1263
	rcu_read_lock();
1264
	max_sectors = r1_bio->sectors;
Linus Torvalds's avatar
Linus Torvalds committed
1265
	for (i = 0;  i < disks; i++) {
1266
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
1267 1268 1269 1270 1271
		if (rdev && unlikely(test_bit(Blocked, &rdev->flags))) {
			atomic_inc(&rdev->nr_pending);
			blocked_rdev = rdev;
			break;
		}
1272
		r1_bio->bios[i] = NULL;
1273 1274
		if (!rdev || test_bit(Faulty, &rdev->flags)
		    || test_bit(Unmerged, &rdev->flags)) {
1275 1276
			if (i < conf->raid_disks)
				set_bit(R1BIO_Degraded, &r1_bio->state);
1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303
			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;
1304
				rdev_dec_pending(rdev, mddev);
1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315
				/* 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;
1316
			}
1317 1318 1319 1320 1321 1322 1323
			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
1324 1325 1326
	}
	rcu_read_unlock();

1327 1328 1329
	if (unlikely(blocked_rdev)) {
		/* Wait for this device to become unblocked */
		int j;
1330
		sector_t old = start_next_window;
1331 1332 1333 1334

		for (j = 0; j < i; j++)
			if (r1_bio->bios[j])
				rdev_dec_pending(conf->mirrors[j].rdev, mddev);
1335
		r1_bio->state = 0;
1336
		allow_barrier(conf, start_next_window, bio->bi_iter.bi_sector);
1337
		md_wait_for_blocked_rdev(blocked_rdev, mddev);
1338 1339 1340 1341 1342 1343 1344 1345 1346 1347
		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);
1348 1349 1350
		goto retry_write;
	}

1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361
	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);
1362
	}
1363
	sectors_handled = r1_bio->sector + max_sectors - bio->bi_iter.bi_sector;
1364

1365
	atomic_set(&r1_bio->remaining, 1);
1366
	atomic_set(&r1_bio->behind_remaining, 0);
1367

1368
	first_clone = 1;
Linus Torvalds's avatar
Linus Torvalds committed
1369 1370 1371 1372 1373
	for (i = 0; i < disks; i++) {
		struct bio *mbio;
		if (!r1_bio->bios[i])
			continue;

1374
		mbio = bio_clone_mddev(bio, GFP_NOIO, mddev);
1375
		bio_trim(mbio, r1_bio->sector - bio->bi_iter.bi_sector, max_sectors);
1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393

		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;
		}
1394
		if (r1_bio->behind_bvecs) {
1395 1396 1397
			struct bio_vec *bvec;
			int j;

1398 1399
			/*
			 * We trimmed the bio, so _all is legit
1400
			 */
1401
			bio_for_each_segment_all(bvec, mbio, j)
1402
				bvec->bv_page = r1_bio->behind_bvecs[j].bv_page;
1403 1404 1405 1406
			if (test_bit(WriteMostly, &conf->mirrors[i].rdev->flags))
				atomic_inc(&r1_bio->behind_remaining);
		}

1407 1408
		r1_bio->bios[i] = mbio;

1409
		mbio->bi_iter.bi_sector	= (r1_bio->sector +
1410 1411 1412
				   conf->mirrors[i].rdev->data_offset);
		mbio->bi_bdev = conf->mirrors[i].rdev->bdev;
		mbio->bi_end_io	= raid1_end_write_request;
1413 1414
		mbio->bi_rw =
			WRITE | do_flush_fua | do_sync | do_discard | do_same;
1415 1416
		mbio->bi_private = r1_bio;

Linus Torvalds's avatar
Linus Torvalds committed
1417
		atomic_inc(&r1_bio->remaining);
1418 1419 1420 1421 1422 1423

		cb = blk_check_plugged(raid1_unplug, mddev, sizeof(*plug));
		if (cb)
			plug = container_of(cb, struct raid1_plug_cb, cb);
		else
			plug = NULL;
1424
		spin_lock_irqsave(&conf->device_lock, flags);
1425 1426 1427 1428 1429 1430 1431
		if (plug) {
			bio_list_add(&plug->pending, mbio);
			plug->pending_cnt++;
		} else {
			bio_list_add(&conf->pending_bio_list, mbio);
			conf->pending_count++;
		}
1432
		spin_unlock_irqrestore(&conf->device_lock, flags);
1433
		if (!plug)
NeilBrown's avatar
NeilBrown committed
1434
			md_wakeup_thread(mddev->thread);
Linus Torvalds's avatar
Linus Torvalds committed
1435
	}
1436 1437 1438
	/* Mustn't call r1_bio_write_done before this next test,
	 * as it could result in the bio being freed.
	 */
1439
	if (sectors_handled < bio_sectors(bio)) {
1440
		r1_bio_write_done(r1_bio);
1441 1442 1443 1444 1445
		/* 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;
1446
		r1_bio->sectors = bio_sectors(bio) - sectors_handled;
1447 1448
		r1_bio->state = 0;
		r1_bio->mddev = mddev;
1449
		r1_bio->sector = bio->bi_iter.bi_sector + sectors_handled;
1450 1451 1452
		goto retry_write;
	}

1453 1454 1455 1456
	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
1457 1458
}

1459
static void status(struct seq_file *seq, struct mddev *mddev)
Linus Torvalds's avatar
Linus Torvalds committed
1460
{
1461
	struct r1conf *conf = mddev->private;
Linus Torvalds's avatar
Linus Torvalds committed
1462 1463 1464
	int i;

	seq_printf(seq, " [%d/%d] [", conf->raid_disks,
1465
		   conf->raid_disks - mddev->degraded);
1466 1467
	rcu_read_lock();
	for (i = 0; i < conf->raid_disks; i++) {
1468
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
Linus Torvalds's avatar
Linus Torvalds committed
1469
		seq_printf(seq, "%s",
1470 1471 1472
			   rdev && test_bit(In_sync, &rdev->flags) ? "U" : "_");
	}
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
1473 1474 1475
	seq_printf(seq, "]");
}

1476
static void error(struct mddev *mddev, struct md_rdev *rdev)
Linus Torvalds's avatar
Linus Torvalds committed
1477 1478
{
	char b[BDEVNAME_SIZE];
1479
	struct r1conf *conf = mddev->private;
1480
	unsigned long flags;
Linus Torvalds's avatar
Linus Torvalds committed
1481 1482 1483 1484 1485 1486 1487

	/*
	 * 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
	 */
1488
	if (test_bit(In_sync, &rdev->flags)
1489
	    && (conf->raid_disks - mddev->degraded) == 1) {
Linus Torvalds's avatar
Linus Torvalds committed
1490 1491
		/*
		 * Don't fail the drive, act as though we were just a
1492 1493 1494
		 * 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
1495
		 */
1496
		conf->recovery_disabled = mddev->recovery_disabled;
Linus Torvalds's avatar
Linus Torvalds committed
1497
		return;
1498
	}
1499
	set_bit(Blocked, &rdev->flags);
1500
	spin_lock_irqsave(&conf->device_lock, flags);
1501
	if (test_and_clear_bit(In_sync, &rdev->flags)) {
Linus Torvalds's avatar
Linus Torvalds committed
1502
		mddev->degraded++;
1503 1504 1505
		set_bit(Faulty, &rdev->flags);
	} else
		set_bit(Faulty, &rdev->flags);
1506
	spin_unlock_irqrestore(&conf->device_lock, flags);
1507 1508 1509 1510
	/*
	 * if recovery is running, make sure it aborts.
	 */
	set_bit(MD_RECOVERY_INTR, &mddev->recovery);
1511
	set_bit(MD_CHANGE_DEVS, &mddev->flags);
1512 1513 1514
	printk(KERN_ALERT
	       "md/raid1:%s: Disk failure on %s, disabling device.\n"
	       "md/raid1:%s: Operation continuing on %d devices.\n",
1515 1516
	       mdname(mddev), bdevname(rdev->bdev, b),
	       mdname(mddev), conf->raid_disks - mddev->degraded);
Linus Torvalds's avatar
Linus Torvalds committed
1517 1518
}

1519
static void print_conf(struct r1conf *conf)
Linus Torvalds's avatar
Linus Torvalds committed
1520 1521 1522
{
	int i;

1523
	printk(KERN_DEBUG "RAID1 conf printout:\n");
Linus Torvalds's avatar
Linus Torvalds committed
1524
	if (!conf) {
1525
		printk(KERN_DEBUG "(!conf)\n");
Linus Torvalds's avatar
Linus Torvalds committed
1526 1527
		return;
	}
1528
	printk(KERN_DEBUG " --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
Linus Torvalds's avatar
Linus Torvalds committed
1529 1530
		conf->raid_disks);

1531
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
1532 1533
	for (i = 0; i < conf->raid_disks; i++) {
		char b[BDEVNAME_SIZE];
1534
		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
1535
		if (rdev)
1536
			printk(KERN_DEBUG " disk %d, wo:%d, o:%d, dev:%s\n",
1537 1538 1539
			       i, !test_bit(In_sync, &rdev->flags),
			       !test_bit(Faulty, &rdev->flags),
			       bdevname(rdev->bdev,b));
Linus Torvalds's avatar
Linus Torvalds committed
1540
	}
1541
	rcu_read_unlock();
Linus Torvalds's avatar
Linus Torvalds committed
1542 1543
}

1544
static void close_sync(struct r1conf *conf)
Linus Torvalds's avatar
Linus Torvalds committed
1545
{
1546 1547
	wait_barrier(conf, NULL);
	allow_barrier(conf, 0, 0);
Linus Torvalds's avatar
Linus Torvalds committed
1548 1549 1550

	mempool_destroy(conf->r1buf_pool);
	conf->r1buf_pool = NULL;
1551

1552
	spin_lock_irq(&conf->resync_lock);
1553 1554
	conf->next_resync = 0;
	conf->start_next_window = MaxSector;
1555 1556 1557 1558
	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
1559 1560
}

1561
static int raid1_spare_active(struct mddev *mddev)
Linus Torvalds's avatar
Linus Torvalds committed
1562 1563
{
	int i;
1564
	struct r1conf *conf = mddev->private;
1565 1566
	int count = 0;
	unsigned long flags;
Linus Torvalds's avatar
Linus Torvalds committed
1567 1568

	/*
1569
	 * Find all failed disks within the RAID1 configuration
1570 1571
	 * and mark them readable.
	 * Called under mddev lock, so rcu protection not needed.
1572 1573
	 * 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
1574
	 */
1575
	spin_lock_irqsave(&conf->device_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
1576
	for (i = 0; i < conf->raid_disks; i++) {
1577
		struct md_rdev *rdev = conf->mirrors[i].rdev;
1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596
		struct md_rdev *repl = conf->mirrors[conf->raid_disks + i].rdev;
		if (repl
		    && repl->recovery_offset == MaxSector
		    && !test_bit(Faulty, &repl->flags)
		    && !test_and_set_bit(In_sync, &repl->flags)) {
			/* replacement has just become active */
			if (!rdev ||
			    !test_and_clear_bit(In_sync, &rdev->flags))
				count++;
			if (rdev) {
				/* Replaced device not technically
				 * faulty, but we need to be sure
				 * it gets removed and never re-added
				 */
				set_bit(Faulty, &rdev->flags);
				sysfs_notify_dirent_safe(
					rdev->sysfs_state);
			}
		}
1597
		if (rdev
1598
		    && rdev->recovery_offset == MaxSector
1599
		    && !test_bit(Faulty, &rdev->flags)
1600
		    && !test_and_set_bit(In_sync, &rdev->flags)) {
1601
			count++;
1602
			sysfs_notify_dirent_safe(rdev->sysfs_state);
Linus Torvalds's avatar
Linus Torvalds committed
1603 1604
		}
	}
1605 1606
	mddev->degraded -= count;
	spin_unlock_irqrestore(&conf->device_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
1607 1608

	print_conf(conf);
1609
	return count;
Linus Torvalds's avatar
Linus Torvalds committed
1610 1611
}

1612
static int raid1_add_disk(struct mddev *mddev, struct md_rdev *rdev)
Linus Torvalds's avatar
Linus Torvalds committed
1613
{
1614
	struct r1conf *conf = mddev->private;
1615
	int err = -EEXIST;
1616
	int mirror = 0;
1617
	struct raid1_info *p;
1618
	int first = 0;
1619
	int last = conf->raid_disks - 1;
1620
	struct request_queue *q = bdev_get_queue(rdev->bdev);
Linus Torvalds's avatar
Linus Torvalds committed
1621

1622 1623 1624
	if (mddev->recovery_disabled == conf->recovery_disabled)
		return -EBUSY;

1625 1626 1627
	if (rdev->raid_disk >= 0)
		first = last = rdev->raid_disk;

1628 1629 1630 1631 1632
	if (q->merge_bvec_fn) {
		set_bit(Unmerged, &rdev->flags);
		mddev->merge_check_needed = 1;
	}

1633 1634 1635
	for (mirror = first; mirror <= last; mirror++) {
		p = conf->mirrors+mirror;
		if (!p->rdev) {
Linus Torvalds's avatar
Linus Torvalds committed
1636

1637 1638 1639
			if (mddev->gendisk)
				disk_stack_limits(mddev->gendisk, rdev->bdev,
						  rdev->data_offset << 9);
Linus Torvalds's avatar
Linus Torvalds committed
1640 1641 1642

			p->head_position = 0;
			rdev->raid_disk = mirror;
1643
			err = 0;
1644 1645 1646 1647
			/* 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)
1648
				conf->fullsync = 1;
1649
			rcu_assign_pointer(p->rdev, rdev);
Linus Torvalds's avatar
Linus Torvalds committed
1650 1651
			break;
		}
1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663
		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;
		}
	}
1664 1665 1666 1667 1668 1669 1670 1671 1672
	if (err == 0 && test_bit(Unmerged, &rdev->flags)) {
		/* Some requests might not have seen this new
		 * merge_bvec_fn.  We must wait for them to complete
		 * before merging the device fully.
		 * First we make sure any code which has tested
		 * our function has submitted the request, then
		 * we wait for all outstanding requests to complete.
		 */
		synchronize_sched();
1673 1674
		freeze_array(conf, 0);
		unfreeze_array(conf);
1675 1676
		clear_bit(Unmerged, &rdev->flags);
	}
1677
	md_integrity_add_rdev(rdev, mddev);
1678
	if (mddev->queue && blk_queue_discard(bdev_get_queue(rdev->bdev)))
Shaohua Li's avatar
Shaohua Li committed
1679
		queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, mddev->queue);
Linus Torvalds's avatar
Linus Torvalds committed
1680
	print_conf(conf);
1681
	return err;
Linus Torvalds's avatar
Linus Torvalds committed
1682 1683
}

1684
static int raid1_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
Linus Torvalds's avatar
Linus Torvalds committed
1685
{
1686
	struct r1conf *conf = mddev->private;
Linus Torvalds's avatar
Linus Torvalds committed
1687
	int err = 0;
1688
	int number = rdev->raid_disk;
1689
	struct raid1_info *p = conf->mirrors + number;
Linus Torvalds's avatar
Linus Torvalds committed
1690

1691 1692 1693
	if (rdev != p->rdev)
		p = conf->mirrors + conf->raid_disks + number;

Linus Torvalds's avatar
Linus Torvalds committed
1694
	print_conf(conf);
1695
	if (rdev == p->rdev) {
1696
		if (test_bit(In_sync, &rdev->flags) ||
Linus Torvalds's avatar
Linus Torvalds committed
1697 1698 1699 1700
		    atomic_read(&rdev->nr_pending)) {
			err = -EBUSY;
			goto abort;
		}
1701
		/* Only remove non-faulty devices if recovery
1702 1703 1704
		 * is not possible.
		 */
		if (!test_bit(Faulty, &rdev->flags) &&
1705
		    mddev->recovery_disabled != conf->recovery_disabled &&
1706 1707 1708 1709
		    mddev->degraded < conf->raid_disks) {
			err = -EBUSY;
			goto abort;
		}
Linus Torvalds's avatar
Linus Torvalds committed
1710
		p->rdev = NULL;
1711
		synchronize_rcu();
Linus Torvalds's avatar
Linus Torvalds committed
1712 1713 1714 1715
		if (atomic_read(&rdev->nr_pending)) {
			/* lost the race, try later */
			err = -EBUSY;
			p->rdev = rdev;
1716
			goto abort;
1717 1718 1719 1720 1721 1722 1723
		} 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;
1724
			freeze_array(conf, 0);
1725 1726 1727
			clear_bit(Replacement, &repl->flags);
			p->rdev = repl;
			conf->mirrors[conf->raid_disks + number].rdev = NULL;
1728
			unfreeze_array(conf);
1729 1730
			clear_bit(WantReplacement, &rdev->flags);
		} else
1731
			clear_bit(WantReplacement, &rdev->flags);
1732
		err = md_integrity_register(mddev);
Linus Torvalds's avatar
Linus Torvalds committed
1733 1734 1735 1736 1737 1738 1739
	}
abort:

	print_conf(conf);
	return err;
}

1740
static void end_sync_read(struct bio *bio, int error)
Linus Torvalds's avatar
Linus Torvalds committed
1741
{
1742
	struct r1bio *r1_bio = bio->bi_private;
Linus Torvalds's avatar
Linus Torvalds committed
1743

1744
	update_head_pos(r1_bio->read_disk, r1_bio);
1745

Linus Torvalds's avatar
Linus Torvalds committed
1746 1747 1748 1749 1750
	/*
	 * 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
	 */
1751
	if (test_bit(BIO_UPTODATE, &bio->bi_flags))
Linus Torvalds's avatar
Linus Torvalds committed
1752
		set_bit(R1BIO_Uptodate, &r1_bio->state);
1753 1754 1755

	if (atomic_dec_and_test(&r1_bio->remaining))
		reschedule_retry(r1_bio);
Linus Torvalds's avatar
Linus Torvalds committed
1756 1757
}

1758
static void end_sync_write(struct bio *bio, int error)
Linus Torvalds's avatar
Linus Torvalds committed
1759 1760
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
1761
	struct r1bio *r1_bio = bio->bi_private;
1762
	struct mddev *mddev = r1_bio->mddev;
1763
	struct r1conf *conf = mddev->private;
Linus Torvalds's avatar
Linus Torvalds committed
1764
	int mirror=0;
1765 1766
	sector_t first_bad;
	int bad_sectors;
Linus Torvalds's avatar
Linus Torvalds committed
1767

1768 1769
	mirror = find_bio_disk(r1_bio, bio);

1770
	if (!uptodate) {
1771
		sector_t sync_blocks = 0;
1772 1773 1774 1775
		sector_t s = r1_bio->sector;
		long sectors_to_go = r1_bio->sectors;
		/* make sure these bits doesn't get cleared. */
		do {
1776
			bitmap_end_sync(mddev->bitmap, s,
1777 1778 1779 1780
					&sync_blocks, 1);
			s += sync_blocks;
			sectors_to_go -= sync_blocks;
		} while (sectors_to_go > 0);
1781 1782
		set_bit(WriteErrorSeen,
			&conf->mirrors[mirror].rdev->flags);
1783 1784 1785 1786
		if (!test_and_set_bit(WantReplacement,
				      &conf->mirrors[mirror].rdev->flags))
			set_bit(MD_RECOVERY_NEEDED, &
				mddev->recovery);
1787
		set_bit(R1BIO_WriteError, &r1_bio->state);
1788 1789 1790
	} else if (is_badblock(conf->mirrors[mirror].rdev,
			       r1_bio->sector,
			       r1_bio->sectors,
1791 1792 1793 1794 1795 1796
			       &first_bad, &bad_sectors) &&
		   !is_badblock(conf->mirrors[r1_bio->read_disk].rdev,
				r1_bio->sector,
				r1_bio->sectors,
				&first_bad, &bad_sectors)
		)
1797
		set_bit(R1BIO_MadeGood, &r1_bio->state);
1798

Linus Torvalds's avatar
Linus Torvalds committed
1799
	if (atomic_dec_and_test(&r1_bio->remaining)) {
1800
		int s = r1_bio->sectors;
1801 1802
		if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
		    test_bit(R1BIO_WriteError, &r1_bio->state))
1803 1804 1805 1806 1807
			reschedule_retry(r1_bio);
		else {
			put_buf(r1_bio);
			md_done_sync(mddev, s, uptodate);
		}
Linus Torvalds's avatar
Linus Torvalds committed
1808 1809 1810
	}
}

1811
static int r1_sync_page_io(struct md_rdev *rdev, sector_t sector,
1812 1813 1814 1815 1816
			    int sectors, struct page *page, int rw)
{
	if (sync_page_io(rdev, sector, sectors << 9, page, rw, false))
		/* success */
		return 1;
1817
	if (rw == WRITE) {
1818
		set_bit(WriteErrorSeen, &rdev->flags);
1819 1820 1821 1822 1823
		if (!test_and_set_bit(WantReplacement,
				      &rdev->flags))
			set_bit(MD_RECOVERY_NEEDED, &
				rdev->mddev->recovery);
	}
1824 1825 1826 1827 1828 1829
	/* 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;
}

1830
static int fix_sync_read_error(struct r1bio *r1_bio)
Linus Torvalds's avatar
Linus Torvalds committed
1831
{
1832 1833 1834 1835 1836 1837 1838
	/* 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.
1839 1840 1841
	 * 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.
1842
	 */
1843
	struct mddev *mddev = r1_bio->mddev;
1844
	struct r1conf *conf = mddev->private;
1845 1846 1847 1848 1849 1850 1851 1852 1853
	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;
1854
		struct md_rdev *rdev;
1855
		int start;
1856 1857 1858 1859 1860 1861 1862 1863 1864 1865

		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;
1866
				if (sync_page_io(rdev, sect, s<<9,
1867 1868 1869 1870 1871 1872 1873
						 bio->bi_io_vec[idx].bv_page,
						 READ, false)) {
					success = 1;
					break;
				}
			}
			d++;
1874
			if (d == conf->raid_disks * 2)
1875 1876 1877
				d = 0;
		} while (!success && d != r1_bio->read_disk);

1878
		if (!success) {
1879
			char b[BDEVNAME_SIZE];
1880 1881 1882 1883 1884 1885
			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.
			 */
1886 1887 1888 1889 1890
			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);
1891
			for (d = 0; d < conf->raid_disks * 2; d++) {
1892 1893 1894 1895 1896 1897 1898
				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) {
1899 1900
				conf->recovery_disabled =
					mddev->recovery_disabled;
1901 1902 1903 1904 1905 1906 1907 1908 1909 1910
				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;
1911
		}
1912 1913 1914 1915 1916

		start = d;
		/* write it back and re-read */
		while (d != r1_bio->read_disk) {
			if (d == 0)
1917
				d = conf->raid_disks * 2;
1918 1919 1920 1921
			d--;
			if (r1_bio->bios[d]->bi_end_io != end_sync_read)
				continue;
			rdev = conf->mirrors[d].rdev;
1922 1923 1924
			if (r1_sync_page_io(rdev, sect, s,
					    bio->bi_io_vec[idx].bv_page,
					    WRITE) == 0) {
1925 1926
				r1_bio->bios[d]->bi_end_io = NULL;
				rdev_dec_pending(rdev, mddev);
1927
			}
1928 1929 1930 1931
		}
		d = start;
		while (d != r1_bio->read_disk) {
			if (d == 0)
1932
				d = conf->raid_disks * 2;
1933 1934 1935 1936
			d--;
			if (r1_bio->bios[d]->bi_end_io != end_sync_read)
				continue;
			rdev = conf->mirrors[d].rdev;
1937 1938 1939
			if (r1_sync_page_io(rdev, sect, s,
					    bio->bi_io_vec[idx].bv_page,
					    READ) != 0)
1940
				atomic_add(s, &rdev->corrected_errors);
1941
		}
1942 1943 1944 1945
		sectors -= s;
		sect += s;
		idx ++;
	}
1946
	set_bit(R1BIO_Uptodate, &r1_bio->state);
1947
	set_bit(BIO_UPTODATE, &bio->bi_flags);
1948 1949 1950
	return 1;
}

1951
static void process_checks(struct r1bio *r1_bio)
1952 1953 1954 1955 1956 1957 1958 1959
{
	/* 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
	 */
1960
	struct mddev *mddev = r1_bio->mddev;
1961
	struct r1conf *conf = mddev->private;
1962 1963
	int primary;
	int i;
1964
	int vcnt;
1965

1966 1967 1968 1969 1970
	/* 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;
1971
		int uptodate;
1972 1973 1974
		struct bio *b = r1_bio->bios[i];
		if (b->bi_end_io != end_sync_read)
			continue;
1975 1976
		/* fixup the bio for reuse, but preserve BIO_UPTODATE */
		uptodate = test_bit(BIO_UPTODATE, &b->bi_flags);
1977
		bio_reset(b);
1978 1979
		if (!uptodate)
			clear_bit(BIO_UPTODATE, &b->bi_flags);
1980
		b->bi_vcnt = vcnt;
1981 1982
		b->bi_iter.bi_size = r1_bio->sectors << 9;
		b->bi_iter.bi_sector = r1_bio->sector +
1983 1984 1985 1986 1987
			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;

1988
		size = b->bi_iter.bi_size;
1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999
		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;
		}
	}
2000
	for (primary = 0; primary < conf->raid_disks * 2; primary++)
2001 2002 2003 2004 2005 2006 2007
		if (r1_bio->bios[primary]->bi_end_io == end_sync_read &&
		    test_bit(BIO_UPTODATE, &r1_bio->bios[primary]->bi_flags)) {
			r1_bio->bios[primary]->bi_end_io = NULL;
			rdev_dec_pending(conf->mirrors[primary].rdev, mddev);
			break;
		}
	r1_bio->read_disk = primary;
2008
	for (i = 0; i < conf->raid_disks * 2; i++) {
2009 2010 2011
		int j;
		struct bio *pbio = r1_bio->bios[primary];
		struct bio *sbio = r1_bio->bios[i];
2012
		int uptodate = test_bit(BIO_UPTODATE, &sbio->bi_flags);
2013

Kent Overstreet's avatar
Kent Overstreet committed
2014
		if (sbio->bi_end_io != end_sync_read)
2015
			continue;
2016 2017
		/* Now we can 'fixup' the BIO_UPTODATE flag */
		set_bit(BIO_UPTODATE, &sbio->bi_flags);
2018

2019
		if (uptodate) {
2020 2021 2022 2023 2024 2025
			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),
2026
					   sbio->bi_io_vec[j].bv_len))
2027
					break;
2028
			}
2029 2030 2031
		} else
			j = 0;
		if (j >= 0)
2032
			atomic64_add(r1_bio->sectors, &mddev->resync_mismatches);
2033
		if (j < 0 || (test_bit(MD_RECOVERY_CHECK, &mddev->recovery)
2034
			      && uptodate)) {
2035 2036 2037 2038 2039
			/* 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
2040 2041

		bio_copy_data(sbio, pbio);
2042
	}
2043 2044
}

2045
static void sync_request_write(struct mddev *mddev, struct r1bio *r1_bio)
2046
{
2047
	struct r1conf *conf = mddev->private;
2048
	int i;
2049
	int disks = conf->raid_disks * 2;
2050 2051 2052 2053 2054 2055 2056 2057
	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;
2058 2059

	if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
2060 2061
		process_checks(r1_bio);

2062 2063 2064
	/*
	 * schedule writes
	 */
Linus Torvalds's avatar
Linus Torvalds committed
2065 2066 2067
	atomic_set(&r1_bio->remaining, 1);
	for (i = 0; i < disks ; i++) {
		wbio = r1_bio->bios[i];
2068 2069 2070 2071
		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
2072
			continue;
2073 2074
		if (test_bit(Faulty, &conf->mirrors[i].rdev->flags))
			continue;
Linus Torvalds's avatar
Linus Torvalds committed
2075

2076 2077
		wbio->bi_rw = WRITE;
		wbio->bi_end_io = end_sync_write;
Linus Torvalds's avatar
Linus Torvalds committed
2078
		atomic_inc(&r1_bio->remaining);
2079
		md_sync_acct(conf->mirrors[i].rdev->bdev, bio_sectors(wbio));
2080

Linus Torvalds's avatar
Linus Torvalds committed
2081 2082 2083 2084
		generic_make_request(wbio);
	}

	if (atomic_dec_and_test(&r1_bio->remaining)) {
2085
		/* if we're here, all write(s) have completed, so clean up */
2086 2087 2088 2089 2090 2091 2092 2093
		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
2094 2095 2096 2097 2098 2099 2100 2101
	}
}

/*
 * This is a kernel thread which:
 *
 *	1.	Retries failed read operations on working mirrors.
 *	2.	Updates the raid superblock when problems encounter.
2102
 *	3.	Performs writes following reads for array synchronising.
Linus Torvalds's avatar
Linus Torvalds committed
2103 2104
 */

2105
static void fix_read_error(struct r1conf *conf, int read_disk,
2106 2107
			   sector_t sect, int sectors)
{
2108
	struct mddev *mddev = conf->mddev;
2109 2110 2111 2112 2113
	while(sectors) {
		int s = sectors;
		int d = read_disk;
		int success = 0;
		int start;
2114
		struct md_rdev *rdev;
2115 2116 2117 2118 2119 2120 2121 2122 2123 2124

		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....
			 */
2125 2126 2127
			sector_t first_bad;
			int bad_sectors;

2128 2129
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
2130 2131 2132
			    (test_bit(In_sync, &rdev->flags) ||
			     (!test_bit(Faulty, &rdev->flags) &&
			      rdev->recovery_offset >= sect + s)) &&
2133 2134
			    is_badblock(rdev, sect, s,
					&first_bad, &bad_sectors) == 0 &&
2135 2136
			    sync_page_io(rdev, sect, s<<9,
					 conf->tmppage, READ, false))
2137 2138 2139
				success = 1;
			else {
				d++;
2140
				if (d == conf->raid_disks * 2)
2141 2142 2143 2144 2145
					d = 0;
			}
		} while (!success && d != read_disk);

		if (!success) {
2146
			/* Cannot read from anywhere - mark it bad */
2147
			struct md_rdev *rdev = conf->mirrors[read_disk].rdev;
2148 2149
			if (!rdev_set_badblocks(rdev, sect, s, 0))
				md_error(mddev, rdev);
2150 2151 2152 2153 2154 2155
			break;
		}
		/* write it back and re-read */
		start = d;
		while (d != read_disk) {
			if (d==0)
2156
				d = conf->raid_disks * 2;
2157 2158 2159
			d--;
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
2160
			    !test_bit(Faulty, &rdev->flags))
2161 2162
				r1_sync_page_io(rdev, sect, s,
						conf->tmppage, WRITE);
2163 2164 2165 2166 2167
		}
		d = start;
		while (d != read_disk) {
			char b[BDEVNAME_SIZE];
			if (d==0)
2168
				d = conf->raid_disks * 2;
2169 2170 2171
			d--;
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
2172
			    !test_bit(Faulty, &rdev->flags)) {
2173 2174
				if (r1_sync_page_io(rdev, sect, s,
						    conf->tmppage, READ)) {
2175 2176
					atomic_add(s, &rdev->corrected_errors);
					printk(KERN_INFO
2177
					       "md/raid1:%s: read error corrected "
2178 2179
					       "(%d sectors at %llu on %s)\n",
					       mdname(mddev), s,
2180 2181
					       (unsigned long long)(sect +
					           rdev->data_offset),
2182 2183 2184 2185 2186 2187 2188 2189 2190
					       bdevname(rdev->bdev, b));
				}
			}
		}
		sectors -= s;
		sect += s;
	}
}

2191
static int narrow_write_error(struct r1bio *r1_bio, int i)
2192
{
2193
	struct mddev *mddev = r1_bio->mddev;
2194
	struct r1conf *conf = mddev->private;
2195
	struct md_rdev *rdev = conf->mirrors[i].rdev;
2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228

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

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

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

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

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

2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245
		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);
		}

2246
		wbio->bi_rw = WRITE;
2247 2248
		wbio->bi_iter.bi_sector = r1_bio->sector;
		wbio->bi_iter.bi_size = r1_bio->sectors << 9;
2249

2250
		bio_trim(wbio, sector - r1_bio->sector, sectors);
2251
		wbio->bi_iter.bi_sector += rdev->data_offset;
2252
		wbio->bi_bdev = rdev->bdev;
2253
		if (submit_bio_wait(WRITE, wbio) < 0)
2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266
			/* failure! */
			ok = rdev_set_badblocks(rdev, sector,
						sectors, 0)
				&& ok;

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

2267
static void handle_sync_write_finished(struct r1conf *conf, struct r1bio *r1_bio)
2268 2269 2270
{
	int m;
	int s = r1_bio->sectors;
2271
	for (m = 0; m < conf->raid_disks * 2 ; m++) {
2272
		struct md_rdev *rdev = conf->mirrors[m].rdev;
2273 2274 2275 2276 2277
		struct bio *bio = r1_bio->bios[m];
		if (bio->bi_end_io == NULL)
			continue;
		if (test_bit(BIO_UPTODATE, &bio->bi_flags) &&
		    test_bit(R1BIO_MadeGood, &r1_bio->state)) {
2278
			rdev_clear_badblocks(rdev, r1_bio->sector, s, 0);
2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289
		}
		if (!test_bit(BIO_UPTODATE, &bio->bi_flags) &&
		    test_bit(R1BIO_WriteError, &r1_bio->state)) {
			if (!rdev_set_badblocks(rdev, r1_bio->sector, s, 0))
				md_error(conf->mddev, rdev);
		}
	}
	put_buf(r1_bio);
	md_done_sync(conf->mddev, s, 1);
}

2290
static void handle_write_finished(struct r1conf *conf, struct r1bio *r1_bio)
2291 2292
{
	int m;
2293
	for (m = 0; m < conf->raid_disks * 2 ; m++)
2294
		if (r1_bio->bios[m] == IO_MADE_GOOD) {
2295
			struct md_rdev *rdev = conf->mirrors[m].rdev;
2296 2297
			rdev_clear_badblocks(rdev,
					     r1_bio->sector,
2298
					     r1_bio->sectors, 0);
2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318
			rdev_dec_pending(rdev, conf->mddev);
		} else if (r1_bio->bios[m] != NULL) {
			/* This drive got a write error.  We need to
			 * narrow down and record precise write
			 * errors.
			 */
			if (!narrow_write_error(r1_bio, m)) {
				md_error(conf->mddev,
					 conf->mirrors[m].rdev);
				/* an I/O failed, we can't clear the bitmap */
				set_bit(R1BIO_Degraded, &r1_bio->state);
			}
			rdev_dec_pending(conf->mirrors[m].rdev,
					 conf->mddev);
		}
	if (test_bit(R1BIO_WriteError, &r1_bio->state))
		close_write(r1_bio);
	raid_end_bio_io(r1_bio);
}

2319
static void handle_read_error(struct r1conf *conf, struct r1bio *r1_bio)
2320 2321 2322
{
	int disk;
	int max_sectors;
2323
	struct mddev *mddev = conf->mddev;
2324 2325
	struct bio *bio;
	char b[BDEVNAME_SIZE];
2326
	struct md_rdev *rdev;
2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337

	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) {
2338
		freeze_array(conf, 1);
2339 2340 2341 2342 2343
		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);
2344
	rdev_dec_pending(conf->mirrors[r1_bio->read_disk].rdev, conf->mddev);
2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364

	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);
2365 2366
		bio_trim(bio, r1_bio->sector - bio->bi_iter.bi_sector,
			 max_sectors);
2367 2368 2369 2370 2371 2372 2373 2374
		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));
2375
		bio->bi_iter.bi_sector = r1_bio->sector + rdev->data_offset;
2376 2377 2378 2379 2380 2381 2382 2383
		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
2384
					       - mbio->bi_iter.bi_sector);
2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397
			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;
2398
			r1_bio->sectors = bio_sectors(mbio) - sectors_handled;
2399 2400 2401
			r1_bio->state = 0;
			set_bit(R1BIO_ReadError, &r1_bio->state);
			r1_bio->mddev = mddev;
2402 2403
			r1_bio->sector = mbio->bi_iter.bi_sector +
				sectors_handled;
2404 2405 2406 2407 2408 2409 2410

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

2411
static void raid1d(struct md_thread *thread)
Linus Torvalds's avatar
Linus Torvalds committed
2412
{
2413
	struct mddev *mddev = thread->mddev;
2414
	struct r1bio *r1_bio;
Linus Torvalds's avatar
Linus Torvalds committed
2415
	unsigned long flags;
2416
	struct r1conf *conf = mddev->private;
Linus Torvalds's avatar
Linus Torvalds committed
2417
	struct list_head *head = &conf->retry_list;
2418
	struct blk_plug plug;
Linus Torvalds's avatar
Linus Torvalds committed
2419 2420

	md_check_recovery(mddev);
2421 2422

	blk_start_plug(&plug);
Linus Torvalds's avatar
Linus Torvalds committed
2423
	for (;;) {
2424

2425
		flush_pending_writes(conf);
2426

2427 2428 2429
		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
2430
			break;
2431
		}
2432
		r1_bio = list_entry(head->prev, struct r1bio, retry_list);
Linus Torvalds's avatar
Linus Torvalds committed
2433
		list_del(head->prev);
2434
		conf->nr_queued--;
Linus Torvalds's avatar
Linus Torvalds committed
2435 2436 2437
		spin_unlock_irqrestore(&conf->device_lock, flags);

		mddev = r1_bio->mddev;
2438
		conf = mddev->private;
2439
		if (test_bit(R1BIO_IsSync, &r1_bio->state)) {
2440
			if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
2441 2442 2443
			    test_bit(R1BIO_WriteError, &r1_bio->state))
				handle_sync_write_finished(conf, r1_bio);
			else
2444
				sync_request_write(mddev, r1_bio);
2445
		} else if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
2446 2447 2448 2449 2450
			   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
2451 2452 2453 2454
			/* just a partial read to be scheduled from separate
			 * context
			 */
			generic_make_request(r1_bio->bios[r1_bio->read_disk]);
2455

2456
		cond_resched();
2457 2458
		if (mddev->flags & ~(1<<MD_CHANGE_PENDING))
			md_check_recovery(mddev);
Linus Torvalds's avatar
Linus Torvalds committed
2459
	}
2460
	blk_finish_plug(&plug);
Linus Torvalds's avatar
Linus Torvalds committed
2461 2462
}

2463
static int init_resync(struct r1conf *conf)
Linus Torvalds's avatar
Linus Torvalds committed
2464 2465 2466 2467
{
	int buffs;

	buffs = RESYNC_WINDOW / RESYNC_BLOCK_SIZE;
2468
	BUG_ON(conf->r1buf_pool);
Linus Torvalds's avatar
Linus Torvalds committed
2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486
	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.
 */

2487
static sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipped, int go_faster)
Linus Torvalds's avatar
Linus Torvalds committed
2488
{
2489
	struct r1conf *conf = mddev->private;
2490
	struct r1bio *r1_bio;
Linus Torvalds's avatar
Linus Torvalds committed
2491 2492
	struct bio *bio;
	sector_t max_sector, nr_sectors;
2493
	int disk = -1;
Linus Torvalds's avatar
Linus Torvalds committed
2494
	int i;
2495 2496
	int wonly = -1;
	int write_targets = 0, read_targets = 0;
2497
	sector_t sync_blocks;
2498
	int still_degraded = 0;
2499 2500
	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
2501 2502 2503

	if (!conf->r1buf_pool)
		if (init_resync(conf))
2504
			return 0;
Linus Torvalds's avatar
Linus Torvalds committed
2505

2506
	max_sector = mddev->dev_sectors;
Linus Torvalds's avatar
Linus Torvalds committed
2507
	if (sector_nr >= max_sector) {
2508 2509 2510 2511 2512
		/* 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
		 */
2513 2514
		if (mddev->curr_resync < max_sector) /* aborted */
			bitmap_end_sync(mddev->bitmap, mddev->curr_resync,
2515
						&sync_blocks, 1);
2516
		else /* completed sync */
2517
			conf->fullsync = 0;
2518 2519

		bitmap_close_sync(mddev->bitmap);
Linus Torvalds's avatar
Linus Torvalds committed
2520 2521 2522 2523
		close_sync(conf);
		return 0;
	}

2524 2525
	if (mddev->bitmap == NULL &&
	    mddev->recovery_cp == MaxSector &&
2526
	    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
2527 2528 2529 2530
	    conf->fullsync == 0) {
		*skipped = 1;
		return max_sector - sector_nr;
	}
2531 2532 2533
	/* before building a request, check if we can skip these blocks..
	 * This call the bitmap_start_sync doesn't actually record anything
	 */
2534
	if (!bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, 1) &&
2535
	    !conf->fullsync && !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
2536 2537 2538 2539
		/* We can skip this block, and probably several more */
		*skipped = 1;
		return sync_blocks;
	}
Linus Torvalds's avatar
Linus Torvalds committed
2540
	/*
2541 2542 2543
	 * If there is non-resync activity waiting for a turn,
	 * and resync is going fast enough,
	 * then let it though before starting on this new sync request.
Linus Torvalds's avatar
Linus Torvalds committed
2544
	 */
2545
	if (!go_faster && conf->nr_waiting)
Linus Torvalds's avatar
Linus Torvalds committed
2546
		msleep_interruptible(1000);
2547

2548
	bitmap_cond_end_sync(mddev->bitmap, sector_nr);
2549
	r1_bio = mempool_alloc(conf->r1buf_pool, GFP_NOIO);
2550

2551
	raise_barrier(conf, sector_nr);
Linus Torvalds's avatar
Linus Torvalds committed
2552

2553
	rcu_read_lock();
Linus Torvalds's avatar
Linus Torvalds committed
2554
	/*
2555 2556 2557 2558 2559 2560
	 * 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
2561 2562 2563 2564
	 */

	r1_bio->mddev = mddev;
	r1_bio->sector = sector_nr;
2565
	r1_bio->state = 0;
Linus Torvalds's avatar
Linus Torvalds committed
2566 2567
	set_bit(R1BIO_IsSync, &r1_bio->state);

2568
	for (i = 0; i < conf->raid_disks * 2; i++) {
2569
		struct md_rdev *rdev;
Linus Torvalds's avatar
Linus Torvalds committed
2570
		bio = r1_bio->bios[i];
Kent Overstreet's avatar
Kent Overstreet committed
2571
		bio_reset(bio);
Linus Torvalds's avatar
Linus Torvalds committed
2572

2573 2574
		rdev = rcu_dereference(conf->mirrors[i].rdev);
		if (rdev == NULL ||
2575
		    test_bit(Faulty, &rdev->flags)) {
2576 2577
			if (i < conf->raid_disks)
				still_degraded = 1;
2578
		} else if (!test_bit(In_sync, &rdev->flags)) {
Linus Torvalds's avatar
Linus Torvalds committed
2579 2580 2581
			bio->bi_rw = WRITE;
			bio->bi_end_io = end_sync_write;
			write_targets ++;
2582 2583
		} else {
			/* may need to read from here */
2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608
			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++;
2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620
			} 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++;
2621 2622
			}
		}
2623 2624
		if (bio->bi_end_io) {
			atomic_inc(&rdev->nr_pending);
2625
			bio->bi_iter.bi_sector = sector_nr + rdev->data_offset;
2626 2627 2628
			bio->bi_bdev = rdev->bdev;
			bio->bi_private = r1_bio;
		}
Linus Torvalds's avatar
Linus Torvalds committed
2629
	}
2630 2631 2632 2633
	rcu_read_unlock();
	if (disk < 0)
		disk = wonly;
	r1_bio->read_disk = disk;
2634

2635 2636 2637 2638 2639
	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;
2640
		for (i = 0 ; i < conf->raid_disks * 2 ; i++)
2641
			if (r1_bio->bios[i]->bi_end_io == end_sync_write) {
2642
				struct md_rdev *rdev = conf->mirrors[i].rdev;
2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669
				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;
	}

2670 2671 2672 2673 2674
	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
2675 2676 2677
		/* There is nowhere to write, so all non-sync
		 * drives must be failed - so we are finished
		 */
2678 2679 2680 2681
		sector_t rv;
		if (min_bad > 0)
			max_sector = sector_nr + min_bad;
		rv = max_sector - sector_nr;
2682
		*skipped = 1;
Linus Torvalds's avatar
Linus Torvalds committed
2683 2684 2685 2686
		put_buf(r1_bio);
		return rv;
	}

2687 2688
	if (max_sector > mddev->resync_max)
		max_sector = mddev->resync_max; /* Don't do IO beyond here */
2689 2690
	if (max_sector > sector_nr + good_sectors)
		max_sector = sector_nr + good_sectors;
Linus Torvalds's avatar
Linus Torvalds committed
2691
	nr_sectors = 0;
2692
	sync_blocks = 0;
Linus Torvalds's avatar
Linus Torvalds committed
2693 2694 2695 2696 2697 2698 2699
	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;
2700 2701
		if (sync_blocks == 0) {
			if (!bitmap_start_sync(mddev->bitmap, sector_nr,
2702 2703 2704
					       &sync_blocks, still_degraded) &&
			    !conf->fullsync &&
			    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
2705
				break;
2706
			BUG_ON(sync_blocks < (PAGE_SIZE>>9));
2707
			if ((len >> 9) > sync_blocks)
2708
				len = sync_blocks<<9;
2709
		}
2710

2711
		for (i = 0 ; i < conf->raid_disks * 2; i++) {
Linus Torvalds's avatar
Linus Torvalds committed
2712 2713
			bio = r1_bio->bios[i];
			if (bio->bi_end_io) {
2714
				page = bio->bi_io_vec[bio->bi_vcnt].bv_page;
Linus Torvalds's avatar
Linus Torvalds committed
2715 2716
				if (bio_add_page(bio, page, len, 0) == 0) {
					/* stop here */
2717
					bio->bi_io_vec[bio->bi_vcnt].bv_page = page;
Linus Torvalds's avatar
Linus Torvalds committed
2718 2719 2720
					while (i > 0) {
						i--;
						bio = r1_bio->bios[i];
2721 2722
						if (bio->bi_end_io==NULL)
							continue;
Linus Torvalds's avatar
Linus Torvalds committed
2723 2724
						/* remove last page from this bio */
						bio->bi_vcnt--;
2725
						bio->bi_iter.bi_size -= len;
2726
						__clear_bit(BIO_SEG_VALID, &bio->bi_flags);
Linus Torvalds's avatar
Linus Torvalds committed
2727 2728 2729 2730 2731 2732 2733
					}
					goto bio_full;
				}
			}
		}
		nr_sectors += len>>9;
		sector_nr += len>>9;
2734
		sync_blocks -= (len>>9);
Linus Torvalds's avatar
Linus Torvalds committed
2735 2736 2737 2738
	} while (r1_bio->bios[disk]->bi_vcnt < RESYNC_PAGES);
 bio_full:
	r1_bio->sectors = nr_sectors;

2739 2740 2741 2742 2743
	/* 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);
2744
		for (i = 0; i < conf->raid_disks * 2 && read_targets; i++) {
2745 2746
			bio = r1_bio->bios[i];
			if (bio->bi_end_io == end_sync_read) {
2747
				read_targets--;
2748
				md_sync_acct(bio->bi_bdev, nr_sectors);
2749 2750 2751 2752 2753 2754
				generic_make_request(bio);
			}
		}
	} else {
		atomic_set(&r1_bio->remaining, 1);
		bio = r1_bio->bios[r1_bio->read_disk];
2755
		md_sync_acct(bio->bi_bdev, nr_sectors);
2756
		generic_make_request(bio);
Linus Torvalds's avatar
Linus Torvalds committed
2757

2758
	}
Linus Torvalds's avatar
Linus Torvalds committed
2759 2760 2761
	return nr_sectors;
}

2762
static sector_t raid1_size(struct mddev *mddev, sector_t sectors, int raid_disks)
2763 2764 2765 2766 2767 2768 2769
{
	if (sectors)
		return sectors;

	return mddev->dev_sectors;
}

2770
static struct r1conf *setup_conf(struct mddev *mddev)
Linus Torvalds's avatar
Linus Torvalds committed
2771
{
2772
	struct r1conf *conf;
2773
	int i;
2774
	struct raid1_info *disk;
2775
	struct md_rdev *rdev;
2776
	int err = -ENOMEM;
Linus Torvalds's avatar
Linus Torvalds committed
2777

2778
	conf = kzalloc(sizeof(struct r1conf), GFP_KERNEL);
Linus Torvalds's avatar
Linus Torvalds committed
2779
	if (!conf)
2780
		goto abort;
Linus Torvalds's avatar
Linus Torvalds committed
2781

2782
	conf->mirrors = kzalloc(sizeof(struct raid1_info)
2783
				* mddev->raid_disks * 2,
Linus Torvalds's avatar
Linus Torvalds committed
2784 2785
				 GFP_KERNEL);
	if (!conf->mirrors)
2786
		goto abort;
Linus Torvalds's avatar
Linus Torvalds committed
2787

2788 2789
	conf->tmppage = alloc_page(GFP_KERNEL);
	if (!conf->tmppage)
2790
		goto abort;
2791

2792
	conf->poolinfo = kzalloc(sizeof(*conf->poolinfo), GFP_KERNEL);
Linus Torvalds's avatar
Linus Torvalds committed
2793
	if (!conf->poolinfo)
2794
		goto abort;
2795
	conf->poolinfo->raid_disks = mddev->raid_disks * 2;
Linus Torvalds's avatar
Linus Torvalds committed
2796 2797 2798 2799
	conf->r1bio_pool = mempool_create(NR_RAID1_BIOS, r1bio_pool_alloc,
					  r1bio_pool_free,
					  conf->poolinfo);
	if (!conf->r1bio_pool)
2800 2801
		goto abort;

2802
	conf->poolinfo->mddev = mddev;
Linus Torvalds's avatar
Linus Torvalds committed
2803

2804
	err = -EINVAL;
2805
	spin_lock_init(&conf->device_lock);
NeilBrown's avatar
NeilBrown committed
2806
	rdev_for_each(rdev, mddev) {
2807
		struct request_queue *q;
2808
		int disk_idx = rdev->raid_disk;
Linus Torvalds's avatar
Linus Torvalds committed
2809 2810 2811
		if (disk_idx >= mddev->raid_disks
		    || disk_idx < 0)
			continue;
2812
		if (test_bit(Replacement, &rdev->flags))
2813
			disk = conf->mirrors + mddev->raid_disks + disk_idx;
2814 2815
		else
			disk = conf->mirrors + disk_idx;
Linus Torvalds's avatar
Linus Torvalds committed
2816

2817 2818
		if (disk->rdev)
			goto abort;
Linus Torvalds's avatar
Linus Torvalds committed
2819
		disk->rdev = rdev;
2820 2821 2822
		q = bdev_get_queue(rdev->bdev);
		if (q->merge_bvec_fn)
			mddev->merge_check_needed = 1;
Linus Torvalds's avatar
Linus Torvalds committed
2823 2824

		disk->head_position = 0;
2825
		disk->seq_start = MaxSector;
Linus Torvalds's avatar
Linus Torvalds committed
2826 2827 2828 2829 2830 2831
	}
	conf->raid_disks = mddev->raid_disks;
	conf->mddev = mddev;
	INIT_LIST_HEAD(&conf->retry_list);

	spin_lock_init(&conf->resync_lock);
2832
	init_waitqueue_head(&conf->wait_barrier);
Linus Torvalds's avatar
Linus Torvalds committed
2833

2834
	bio_list_init(&conf->pending_bio_list);
2835
	conf->pending_count = 0;
2836
	conf->recovery_disabled = mddev->recovery_disabled - 1;
2837

2838 2839 2840
	conf->start_next_window = MaxSector;
	conf->current_window_requests = conf->next_window_requests = 0;

2841
	err = -EIO;
2842
	for (i = 0; i < conf->raid_disks * 2; i++) {
Linus Torvalds's avatar
Linus Torvalds committed
2843 2844 2845

		disk = conf->mirrors + i;

2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860
		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;
		}

2861 2862
		if (!disk->rdev ||
		    !test_bit(In_sync, &disk->rdev->flags)) {
Linus Torvalds's avatar
Linus Torvalds committed
2863
			disk->head_position = 0;
2864 2865
			if (disk->rdev &&
			    (disk->rdev->saved_raid_disk < 0))
2866
				conf->fullsync = 1;
2867
		}
Linus Torvalds's avatar
Linus Torvalds committed
2868
	}
2869 2870

	err = -ENOMEM;
2871
	conf->thread = md_register_thread(raid1d, mddev, "raid1");
2872 2873
	if (!conf->thread) {
		printk(KERN_ERR
2874
		       "md/raid1:%s: couldn't allocate thread\n",
2875 2876
		       mdname(mddev));
		goto abort;
2877
	}
Linus Torvalds's avatar
Linus Torvalds committed
2878

2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892
	return conf;

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

2893
static int stop(struct mddev *mddev);
2894
static int run(struct mddev *mddev)
2895
{
2896
	struct r1conf *conf;
2897
	int i;
2898
	struct md_rdev *rdev;
2899
	int ret;
Shaohua Li's avatar
Shaohua Li committed
2900
	bool discard_supported = false;
2901 2902

	if (mddev->level != 1) {
2903
		printk(KERN_ERR "md/raid1:%s: raid level not set to mirroring (%d)\n",
2904 2905 2906 2907
		       mdname(mddev), mddev->level);
		return -EIO;
	}
	if (mddev->reshape_position != MaxSector) {
2908
		printk(KERN_ERR "md/raid1:%s: reshape_position set but not supported\n",
2909 2910 2911
		       mdname(mddev));
		return -EIO;
	}
Linus Torvalds's avatar
Linus Torvalds committed
2912
	/*
2913 2914 2915
	 * copy the already verified devices into our private RAID1
	 * bookkeeping area. [whatever we allocate in run(),
	 * should be freed in stop()]
Linus Torvalds's avatar
Linus Torvalds committed
2916
	 */
2917 2918 2919 2920
	if (mddev->private == NULL)
		conf = setup_conf(mddev);
	else
		conf = mddev->private;
Linus Torvalds's avatar
Linus Torvalds committed
2921

2922 2923
	if (IS_ERR(conf))
		return PTR_ERR(conf);
Linus Torvalds's avatar
Linus Torvalds committed
2924

2925
	if (mddev->queue)
2926 2927
		blk_queue_max_write_same_sectors(mddev->queue, 0);

NeilBrown's avatar
NeilBrown committed
2928
	rdev_for_each(rdev, mddev) {
2929 2930
		if (!mddev->gendisk)
			continue;
2931 2932
		disk_stack_limits(mddev->gendisk, rdev->bdev,
				  rdev->data_offset << 9);
Shaohua Li's avatar
Shaohua Li committed
2933 2934
		if (blk_queue_discard(bdev_get_queue(rdev->bdev)))
			discard_supported = true;
Linus Torvalds's avatar
Linus Torvalds committed
2935
	}
2936

2937 2938 2939 2940 2941 2942 2943 2944 2945 2946
	mddev->degraded = 0;
	for (i=0; i < conf->raid_disks; i++)
		if (conf->mirrors[i].rdev == NULL ||
		    !test_bit(In_sync, &conf->mirrors[i].rdev->flags) ||
		    test_bit(Faulty, &conf->mirrors[i].rdev->flags))
			mddev->degraded++;

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

2947
	if (mddev->recovery_cp != MaxSector)
2948
		printk(KERN_NOTICE "md/raid1:%s: not clean"
2949 2950
		       " -- starting background reconstruction\n",
		       mdname(mddev));
2951
	printk(KERN_INFO
2952
		"md/raid1:%s: active with %d out of %d mirrors\n",
2953
		mdname(mddev), mddev->raid_disks - mddev->degraded,
Linus Torvalds's avatar
Linus Torvalds committed
2954
		mddev->raid_disks);
2955

Linus Torvalds's avatar
Linus Torvalds committed
2956 2957 2958
	/*
	 * Ok, everything is just fine now
	 */
2959 2960 2961 2962
	mddev->thread = conf->thread;
	conf->thread = NULL;
	mddev->private = conf;

2963
	md_set_array_sectors(mddev, raid1_size(mddev, 0, 0));
Linus Torvalds's avatar
Linus Torvalds committed
2964

2965 2966 2967
	if (mddev->queue) {
		mddev->queue->backing_dev_info.congested_fn = raid1_congested;
		mddev->queue->backing_dev_info.congested_data = mddev;
2968
		blk_queue_merge_bvec(mddev->queue, raid1_mergeable_bvec);
Shaohua Li's avatar
Shaohua Li committed
2969 2970 2971 2972 2973 2974 2975

		if (discard_supported)
			queue_flag_set_unlocked(QUEUE_FLAG_DISCARD,
						mddev->queue);
		else
			queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD,
						  mddev->queue);
2976
	}
2977 2978 2979 2980 2981

	ret =  md_integrity_register(mddev);
	if (ret)
		stop(mddev);
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
2982 2983
}

2984
static int stop(struct mddev *mddev)
Linus Torvalds's avatar
Linus Torvalds committed
2985
{
2986
	struct r1conf *conf = mddev->private;
2987 2988 2989
	struct bitmap *bitmap = mddev->bitmap;

	/* wait for behind writes to complete */
2990
	if (bitmap && atomic_read(&bitmap->behind_writes) > 0) {
2991 2992
		printk(KERN_INFO "md/raid1:%s: behind writes in progress - waiting to stop.\n",
		       mdname(mddev));
2993
		/* need to kick something here to make sure I/O goes? */
2994 2995
		wait_event(bitmap->behind_wait,
			   atomic_read(&bitmap->behind_writes) == 0);
2996
	}
Linus Torvalds's avatar
Linus Torvalds committed
2997

2998 2999
	freeze_array(conf, 0);
	unfreeze_array(conf);
3000

3001
	md_unregister_thread(&mddev->thread);
Linus Torvalds's avatar
Linus Torvalds committed
3002 3003
	if (conf->r1bio_pool)
		mempool_destroy(conf->r1bio_pool);
3004
	kfree(conf->mirrors);
3005
	safe_put_page(conf->tmppage);
3006
	kfree(conf->poolinfo);
Linus Torvalds's avatar
Linus Torvalds committed
3007 3008 3009 3010 3011
	kfree(conf);
	mddev->private = NULL;
	return 0;
}

3012
static int raid1_resize(struct mddev *mddev, sector_t sectors)
Linus Torvalds's avatar
Linus Torvalds committed
3013 3014 3015 3016 3017 3018 3019 3020
{
	/* 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.
	 */
3021 3022 3023
	sector_t newsize = raid1_size(mddev, sectors, 0);
	if (mddev->external_size &&
	    mddev->array_sectors > newsize)
3024
		return -EINVAL;
3025 3026 3027 3028 3029 3030
	if (mddev->bitmap) {
		int ret = bitmap_resize(mddev->bitmap, newsize, 0, 0);
		if (ret)
			return ret;
	}
	md_set_array_sectors(mddev, newsize);
3031
	set_capacity(mddev->gendisk, mddev->array_sectors);
3032
	revalidate_disk(mddev->gendisk);
3033
	if (sectors > mddev->dev_sectors &&
3034
	    mddev->recovery_cp > mddev->dev_sectors) {
3035
		mddev->recovery_cp = mddev->dev_sectors;
Linus Torvalds's avatar
Linus Torvalds committed
3036 3037
		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
	}
3038
	mddev->dev_sectors = sectors;
3039
	mddev->resync_max_sectors = sectors;
Linus Torvalds's avatar
Linus Torvalds committed
3040 3041 3042
	return 0;
}

3043
static int raid1_reshape(struct mddev *mddev)
Linus Torvalds's avatar
Linus Torvalds committed
3044 3045 3046 3047 3048 3049 3050 3051
{
	/* 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.
3052 3053 3054
	 *
	 * 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
3055 3056 3057
	 */
	mempool_t *newpool, *oldpool;
	struct pool_info *newpoolinfo;
3058
	struct raid1_info *newmirrors;
3059
	struct r1conf *conf = mddev->private;
3060
	int cnt, raid_disks;
3061
	unsigned long flags;
3062
	int d, d2, err;
Linus Torvalds's avatar
Linus Torvalds committed
3063

3064
	/* Cannot change chunk_size, layout, or level */
3065
	if (mddev->chunk_sectors != mddev->new_chunk_sectors ||
3066 3067
	    mddev->layout != mddev->new_layout ||
	    mddev->level != mddev->new_level) {
3068
		mddev->new_chunk_sectors = mddev->chunk_sectors;
3069 3070 3071 3072 3073
		mddev->new_layout = mddev->layout;
		mddev->new_level = mddev->level;
		return -EINVAL;
	}

3074 3075 3076
	err = md_allow_write(mddev);
	if (err)
		return err;
3077

3078 3079
	raid_disks = mddev->raid_disks + mddev->delta_disks;

3080 3081 3082 3083 3084 3085
	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
3086
			return -EBUSY;
3087
	}
Linus Torvalds's avatar
Linus Torvalds committed
3088 3089 3090 3091 3092

	newpoolinfo = kmalloc(sizeof(*newpoolinfo), GFP_KERNEL);
	if (!newpoolinfo)
		return -ENOMEM;
	newpoolinfo->mddev = mddev;
3093
	newpoolinfo->raid_disks = raid_disks * 2;
Linus Torvalds's avatar
Linus Torvalds committed
3094 3095 3096 3097 3098 3099 3100

	newpool = mempool_create(NR_RAID1_BIOS, r1bio_pool_alloc,
				 r1bio_pool_free, newpoolinfo);
	if (!newpool) {
		kfree(newpoolinfo);
		return -ENOMEM;
	}
3101
	newmirrors = kzalloc(sizeof(struct raid1_info) * raid_disks * 2,
3102
			     GFP_KERNEL);
Linus Torvalds's avatar
Linus Torvalds committed
3103 3104 3105 3106 3107 3108
	if (!newmirrors) {
		kfree(newpoolinfo);
		mempool_destroy(newpool);
		return -ENOMEM;
	}

3109
	freeze_array(conf, 0);
Linus Torvalds's avatar
Linus Torvalds committed
3110 3111 3112 3113

	/* ok, everything is stopped */
	oldpool = conf->r1bio_pool;
	conf->r1bio_pool = newpool;
3114

3115
	for (d = d2 = 0; d < conf->raid_disks; d++) {
3116
		struct md_rdev *rdev = conf->mirrors[d].rdev;
3117
		if (rdev && rdev->raid_disk != d2) {
3118
			sysfs_unlink_rdev(mddev, rdev);
3119
			rdev->raid_disk = d2;
3120 3121
			sysfs_unlink_rdev(mddev, rdev);
			if (sysfs_link_rdev(mddev, rdev))
3122
				printk(KERN_WARNING
3123 3124
				       "md/raid1:%s: cannot register rd%d\n",
				       mdname(mddev), rdev->raid_disk);
3125
		}
3126 3127 3128
		if (rdev)
			newmirrors[d2++].rdev = rdev;
	}
Linus Torvalds's avatar
Linus Torvalds committed
3129 3130 3131 3132 3133
	kfree(conf->mirrors);
	conf->mirrors = newmirrors;
	kfree(conf->poolinfo);
	conf->poolinfo = newpoolinfo;

3134
	spin_lock_irqsave(&conf->device_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
3135
	mddev->degraded += (raid_disks - conf->raid_disks);
3136
	spin_unlock_irqrestore(&conf->device_lock, flags);
Linus Torvalds's avatar
Linus Torvalds committed
3137
	conf->raid_disks = mddev->raid_disks = raid_disks;
3138
	mddev->delta_disks = 0;
Linus Torvalds's avatar
Linus Torvalds committed
3139

3140
	unfreeze_array(conf);
Linus Torvalds's avatar
Linus Torvalds committed
3141 3142 3143 3144 3145 3146 3147 3148

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

	mempool_destroy(oldpool);
	return 0;
}

3149
static void raid1_quiesce(struct mddev *mddev, int state)
3150
{
3151
	struct r1conf *conf = mddev->private;
3152 3153

	switch(state) {
3154 3155 3156
	case 2: /* wake for suspend */
		wake_up(&conf->wait_barrier);
		break;
3157
	case 1:
3158
		freeze_array(conf, 0);
3159
		break;
3160
	case 0:
3161
		unfreeze_array(conf);
3162 3163 3164 3165
		break;
	}
}

3166
static void *raid1_takeover(struct mddev *mddev)
3167 3168 3169 3170 3171
{
	/* raid1 can take over:
	 *  raid5 with 2 devices, any layout or chunk size
	 */
	if (mddev->level == 5 && mddev->raid_disks == 2) {
3172
		struct r1conf *conf;
3173 3174 3175 3176 3177
		mddev->new_level = 1;
		mddev->new_layout = 0;
		mddev->new_chunk_sectors = 0;
		conf = setup_conf(mddev);
		if (!IS_ERR(conf))
3178 3179
			/* Array must appear to be quiesced */
			conf->array_frozen = 1;
3180 3181 3182 3183
		return conf;
	}
	return ERR_PTR(-EINVAL);
}
Linus Torvalds's avatar
Linus Torvalds committed
3184

3185
static struct md_personality raid1_personality =
Linus Torvalds's avatar
Linus Torvalds committed
3186 3187
{
	.name		= "raid1",
3188
	.level		= 1,
Linus Torvalds's avatar
Linus Torvalds committed
3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199
	.owner		= THIS_MODULE,
	.make_request	= make_request,
	.run		= run,
	.stop		= stop,
	.status		= status,
	.error_handler	= error,
	.hot_add_disk	= raid1_add_disk,
	.hot_remove_disk= raid1_remove_disk,
	.spare_active	= raid1_spare_active,
	.sync_request	= sync_request,
	.resize		= raid1_resize,
3200
	.size		= raid1_size,
3201
	.check_reshape	= raid1_reshape,
3202
	.quiesce	= raid1_quiesce,
3203
	.takeover	= raid1_takeover,
Linus Torvalds's avatar
Linus Torvalds committed
3204 3205 3206 3207
};

static int __init raid_init(void)
{
3208
	return register_md_personality(&raid1_personality);
Linus Torvalds's avatar
Linus Torvalds committed
3209 3210 3211 3212
}

static void raid_exit(void)
{
3213
	unregister_md_personality(&raid1_personality);
Linus Torvalds's avatar
Linus Torvalds committed
3214 3215 3216 3217 3218
}

module_init(raid_init);
module_exit(raid_exit);
MODULE_LICENSE("GPL");
3219
MODULE_DESCRIPTION("RAID1 (mirroring) personality for MD");
Linus Torvalds's avatar
Linus Torvalds committed
3220
MODULE_ALIAS("md-personality-3"); /* RAID1 */
3221
MODULE_ALIAS("md-raid1");
3222
MODULE_ALIAS("md-level-1");
3223 3224

module_param(max_queued_requests, int, S_IRUGO|S_IWUSR);