raid1.c 72.4 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11
/*
 * raid1.c : Multiple Devices driver for Linux
 *
 * Copyright (C) 1999, 2000, 2001 Ingo Molnar, Red Hat
 *
 * Copyright (C) 1996, 1997, 1998 Ingo Molnar, Miguel de Icaza, Gadi Oxman
 *
 * RAID-1 management functions.
 *
 * Better read-balancing code written by Mika Kuoppala <miku@iki.fi>, 2000
 *
12
 * Fixes to reconstruction by Jakob Østergaard" <jakob@ostenfeld.dk>
L
Linus Torvalds 已提交
13 14
 * Various fixes by Neil Brown <neilb@cse.unsw.edu.au>
 *
15 16 17 18 19 20 21 22 23
 * Changes by Peter T. Breuer <ptb@it.uc3m.es> 31/1/2003 to support
 * bitmapped intelligence in resync:
 *
 *      - bitmap marked during normal i/o
 *      - bitmap used to skip nondirty blocks during sync
 *
 * Additions to bitmap code, (C) 2003-2004 Paul Clements, SteelEye Technology:
 * - persistent bitmap code
 *
L
Linus Torvalds 已提交
24 25 26 27 28 29 30 31 32 33
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2, or (at your option)
 * any later version.
 *
 * You should have received a copy of the GNU General Public License
 * (for example /usr/src/linux/COPYING); if not, write to the Free
 * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 */

34
#include <linux/slab.h>
35
#include <linux/delay.h>
36 37
#include <linux/blkdev.h>
#include <linux/seq_file.h>
38
#include <linux/ratelimit.h>
39
#include "md.h"
40 41
#include "raid1.h"
#include "bitmap.h"
42 43

#define DEBUG 0
44
#define PRINTK(x...) do { if (DEBUG) printk(x); } while (0)
L
Linus Torvalds 已提交
45 46 47 48 49 50 51

/*
 * Number of guaranteed r1bios in case of extreme VM load:
 */
#define	NR_RAID1_BIOS 256


52 53
static void allow_barrier(conf_t *conf);
static void lower_barrier(conf_t *conf);
L
Linus Torvalds 已提交
54

A
Al Viro 已提交
55
static void * r1bio_pool_alloc(gfp_t gfp_flags, void *data)
L
Linus Torvalds 已提交
56 57 58 59 60
{
	struct pool_info *pi = data;
	int size = offsetof(r1bio_t, bios[pi->raid_disks]);

	/* allocate a r1bio with room for raid_disks entries in the bios array */
J
Jens Axboe 已提交
61
	return kzalloc(size, gfp_flags);
L
Linus Torvalds 已提交
62 63 64 65 66 67 68 69 70 71 72 73 74
}

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

#define RESYNC_BLOCK_SIZE (64*1024)
//#define RESYNC_BLOCK_SIZE PAGE_SIZE
#define RESYNC_SECTORS (RESYNC_BLOCK_SIZE >> 9)
#define RESYNC_PAGES ((RESYNC_BLOCK_SIZE + PAGE_SIZE-1) / PAGE_SIZE)
#define RESYNC_WINDOW (2048*1024)

A
Al Viro 已提交
75
static void * r1buf_pool_alloc(gfp_t gfp_flags, void *data)
L
Linus Torvalds 已提交
76 77 78 79 80 81 82 83
{
	struct pool_info *pi = data;
	struct page *page;
	r1bio_t *r1_bio;
	struct bio *bio;
	int i, j;

	r1_bio = r1bio_pool_alloc(gfp_flags, pi);
J
Jens Axboe 已提交
84
	if (!r1_bio)
L
Linus Torvalds 已提交
85 86 87 88 89 90
		return NULL;

	/*
	 * Allocate bios : 1 for reading, n-1 for writing
	 */
	for (j = pi->raid_disks ; j-- ; ) {
91
		bio = bio_kmalloc(gfp_flags, RESYNC_PAGES);
L
Linus Torvalds 已提交
92 93 94 95 96 97
		if (!bio)
			goto out_free_bio;
		r1_bio->bios[j] = bio;
	}
	/*
	 * Allocate RESYNC_PAGES data pages and attach them to
98 99 100
	 * the first bio.
	 * If this is a user-requested check/repair, allocate
	 * RESYNC_PAGES for each bio.
L
Linus Torvalds 已提交
101
	 */
102 103 104 105 106 107 108 109 110 111 112 113
	if (test_bit(MD_RECOVERY_REQUESTED, &pi->mddev->recovery))
		j = pi->raid_disks;
	else
		j = 1;
	while(j--) {
		bio = r1_bio->bios[j];
		for (i = 0; i < RESYNC_PAGES; i++) {
			page = alloc_page(gfp_flags);
			if (unlikely(!page))
				goto out_free_pages;

			bio->bi_io_vec[i].bv_page = page;
114
			bio->bi_vcnt = i+1;
115 116 117 118 119 120 121 122
		}
	}
	/* If not user-requests, copy the page pointers to all bios */
	if (!test_bit(MD_RECOVERY_REQUESTED, &pi->mddev->recovery)) {
		for (i=0; i<RESYNC_PAGES ; i++)
			for (j=1; j<pi->raid_disks; j++)
				r1_bio->bios[j]->bi_io_vec[i].bv_page =
					r1_bio->bios[0]->bi_io_vec[i].bv_page;
L
Linus Torvalds 已提交
123 124 125 126 127 128 129
	}

	r1_bio->master_bio = NULL;

	return r1_bio;

out_free_pages:
130 131 132
	for (j=0 ; j < pi->raid_disks; j++)
		for (i=0; i < r1_bio->bios[j]->bi_vcnt ; i++)
			put_page(r1_bio->bios[j]->bi_io_vec[i].bv_page);
133
	j = -1;
L
Linus Torvalds 已提交
134 135 136 137 138 139 140 141 142 143
out_free_bio:
	while ( ++j < pi->raid_disks )
		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;
144
	int i,j;
L
Linus Torvalds 已提交
145 146
	r1bio_t *r1bio = __r1_bio;

147 148 149 150 151
	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)
152
				safe_put_page(r1bio->bios[j]->bi_io_vec[i].bv_page);
153
		}
L
Linus Torvalds 已提交
154 155 156 157 158 159 160 161 162 163 164 165
	for (i=0 ; i < pi->raid_disks; i++)
		bio_put(r1bio->bios[i]);

	r1bio_pool_free(r1bio, data);
}

static void put_all_bios(conf_t *conf, r1bio_t *r1_bio)
{
	int i;

	for (i = 0; i < conf->raid_disks; i++) {
		struct bio **bio = r1_bio->bios + i;
166
		if (!BIO_SPECIAL(*bio))
L
Linus Torvalds 已提交
167 168 169 170 171
			bio_put(*bio);
		*bio = NULL;
	}
}

172
static void free_r1bio(r1bio_t *r1_bio)
L
Linus Torvalds 已提交
173
{
174
	conf_t *conf = r1_bio->mddev->private;
L
Linus Torvalds 已提交
175 176 177 178 179

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

180
static void put_buf(r1bio_t *r1_bio)
L
Linus Torvalds 已提交
181
{
182
	conf_t *conf = r1_bio->mddev->private;
183 184 185 186 187 188 189
	int i;

	for (i=0; i<conf->raid_disks; i++) {
		struct bio *bio = r1_bio->bios[i];
		if (bio->bi_end_io)
			rdev_dec_pending(conf->mirrors[i].rdev, r1_bio->mddev);
	}
L
Linus Torvalds 已提交
190 191 192

	mempool_free(r1_bio, conf->r1buf_pool);

193
	lower_barrier(conf);
L
Linus Torvalds 已提交
194 195 196 197 198 199
}

static void reschedule_retry(r1bio_t *r1_bio)
{
	unsigned long flags;
	mddev_t *mddev = r1_bio->mddev;
200
	conf_t *conf = mddev->private;
L
Linus Torvalds 已提交
201 202 203

	spin_lock_irqsave(&conf->device_lock, flags);
	list_add(&r1_bio->retry_list, &conf->retry_list);
204
	conf->nr_queued ++;
L
Linus Torvalds 已提交
205 206
	spin_unlock_irqrestore(&conf->device_lock, flags);

207
	wake_up(&conf->wait_barrier);
L
Linus Torvalds 已提交
208 209 210 211 212 213 214 215
	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.
 */
216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242
static void call_bio_endio(r1bio_t *r1_bio)
{
	struct bio *bio = r1_bio->master_bio;
	int done;
	conf_t *conf = r1_bio->mddev->private;

	if (bio->bi_phys_segments) {
		unsigned long flags;
		spin_lock_irqsave(&conf->device_lock, flags);
		bio->bi_phys_segments--;
		done = (bio->bi_phys_segments == 0);
		spin_unlock_irqrestore(&conf->device_lock, flags);
	} else
		done = 1;

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

L
Linus Torvalds 已提交
243 244 245 246
static void raid_end_bio_io(r1bio_t *r1_bio)
{
	struct bio *bio = r1_bio->master_bio;

247 248 249 250 251 252 253 254
	/* if nobody has done the final endio yet, do it now */
	if (!test_and_set_bit(R1BIO_Returned, &r1_bio->state)) {
		PRINTK(KERN_DEBUG "raid1: sync end %s on sectors %llu-%llu\n",
			(bio_data_dir(bio) == WRITE) ? "write" : "read",
			(unsigned long long) bio->bi_sector,
			(unsigned long long) bio->bi_sector +
				(bio->bi_size >> 9) - 1);

255
		call_bio_endio(r1_bio);
256
	}
L
Linus Torvalds 已提交
257 258 259 260 261 262 263 264
	free_r1bio(r1_bio);
}

/*
 * Update disk head position estimator based on IRQ completion info.
 */
static inline void update_head_pos(int disk, r1bio_t *r1_bio)
{
265
	conf_t *conf = r1_bio->mddev->private;
L
Linus Torvalds 已提交
266 267 268 269 270

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

271
static void raid1_end_read_request(struct bio *bio, int error)
L
Linus Torvalds 已提交
272 273
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
274
	r1bio_t *r1_bio = bio->bi_private;
L
Linus Torvalds 已提交
275
	int mirror;
276
	conf_t *conf = r1_bio->mddev->private;
L
Linus Torvalds 已提交
277 278 279 280 281

	mirror = r1_bio->read_disk;
	/*
	 * this branch is our 'one mirror IO has finished' event handler:
	 */
282 283
	update_head_pos(mirror, r1_bio);

284 285 286 287 288 289
	if (uptodate)
		set_bit(R1BIO_Uptodate, &r1_bio->state);
	else {
		/* If all other devices have failed, we want to return
		 * the error upwards rather than fail the last device.
		 * Here we redefine "uptodate" to mean "Don't want to retry"
L
Linus Torvalds 已提交
290
		 */
291 292 293 294 295 296 297 298
		unsigned long flags;
		spin_lock_irqsave(&conf->device_lock, flags);
		if (r1_bio->mddev->degraded == conf->raid_disks ||
		    (r1_bio->mddev->degraded == conf->raid_disks-1 &&
		     !test_bit(Faulty, &conf->mirrors[mirror].rdev->flags)))
			uptodate = 1;
		spin_unlock_irqrestore(&conf->device_lock, flags);
	}
L
Linus Torvalds 已提交
299

300
	if (uptodate)
L
Linus Torvalds 已提交
301
		raid_end_bio_io(r1_bio);
302
	else {
L
Linus Torvalds 已提交
303 304 305 306
		/*
		 * oops, read error:
		 */
		char b[BDEVNAME_SIZE];
307 308 309 310 311 312 313
		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);
314
		set_bit(R1BIO_ReadError, &r1_bio->state);
L
Linus Torvalds 已提交
315 316 317 318 319 320
		reschedule_retry(r1_bio);
	}

	rdev_dec_pending(conf->mirrors[mirror].rdev, conf->mddev);
}

321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339
static void close_write(r1bio_t *r1_bio)
{
	/* 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);
}

340
static void r1_bio_write_done(r1bio_t *r1_bio)
341
{
342 343 344 345 346 347 348
	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);
349 350 351 352
		if (test_bit(R1BIO_MadeGood, &r1_bio->state))
			reschedule_retry(r1_bio);
		else
			raid_end_bio_io(r1_bio);
353 354 355
	}
}

356
static void raid1_end_write_request(struct bio *bio, int error)
L
Linus Torvalds 已提交
357 358
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
359
	r1bio_t *r1_bio = bio->bi_private;
360
	int mirror, behind = test_bit(R1BIO_BehindIO, &r1_bio->state);
361
	conf_t *conf = r1_bio->mddev->private;
362
	struct bio *to_put = NULL;
L
Linus Torvalds 已提交
363 364 365 366 367 368


	for (mirror = 0; mirror < conf->raid_disks; mirror++)
		if (r1_bio->bios[mirror] == bio)
			break;

T
Tejun Heo 已提交
369 370 371 372
	/*
	 * 'one mirror IO has finished' event handler:
	 */
	if (!uptodate) {
373 374 375
		set_bit(WriteErrorSeen,
			&conf->mirrors[mirror].rdev->flags);
		set_bit(R1BIO_WriteError, &r1_bio->state);
376
	} else {
L
Linus Torvalds 已提交
377
		/*
T
Tejun Heo 已提交
378 379 380 381 382 383 384 385
		 * Set R1BIO_Uptodate in our master bio, so that we
		 * will return a good error code for to the higher
		 * levels even if IO on some other mirrored buffer
		 * fails.
		 *
		 * The 'master' represents the composite IO operation
		 * to user-side. So if something waits for IO, then it
		 * will wait for the 'master' bio.
L
Linus Torvalds 已提交
386
		 */
387 388 389
		sector_t first_bad;
		int bad_sectors;

390 391
		r1_bio->bios[mirror] = NULL;
		to_put = bio;
T
Tejun Heo 已提交
392 393
		set_bit(R1BIO_Uptodate, &r1_bio->state);

394 395 396 397 398 399 400 401 402
		/* Maybe we can clear some bad blocks. */
		if (is_badblock(conf->mirrors[mirror].rdev,
				r1_bio->sector, r1_bio->sectors,
				&first_bad, &bad_sectors)) {
			r1_bio->bios[mirror] = IO_MADE_GOOD;
			set_bit(R1BIO_MadeGood, &r1_bio->state);
		}
	}

T
Tejun Heo 已提交
403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424
	update_head_pos(mirror, r1_bio);

	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;
				PRINTK(KERN_DEBUG "raid1: behind end write sectors %llu-%llu\n",
				       (unsigned long long) mbio->bi_sector,
				       (unsigned long long) mbio->bi_sector +
				       (mbio->bi_size >> 9) - 1);
425
				call_bio_endio(r1_bio);
426 427 428
			}
		}
	}
429 430 431
	if (r1_bio->bios[mirror] == NULL)
		rdev_dec_pending(conf->mirrors[mirror].rdev,
				 conf->mddev);
T
Tejun Heo 已提交
432

L
Linus Torvalds 已提交
433 434 435 436
	/*
	 * Let's see if all mirrored write operations have finished
	 * already.
	 */
437
	r1_bio_write_done(r1_bio);
438

439 440
	if (to_put)
		bio_put(to_put);
L
Linus Torvalds 已提交
441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457
}


/*
 * 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.
 */
458
static int read_balance(conf_t *conf, r1bio_t *r1_bio, int *max_sectors)
L
Linus Torvalds 已提交
459
{
460
	const sector_t this_sector = r1_bio->sector;
461 462
	int sectors;
	int best_good_sectors;
463
	int start_disk;
N
NeilBrown 已提交
464
	int best_disk;
465
	int i;
N
NeilBrown 已提交
466
	sector_t best_dist;
467
	mdk_rdev_t *rdev;
468
	int choose_first;
L
Linus Torvalds 已提交
469 470 471

	rcu_read_lock();
	/*
472
	 * Check if we can balance. We can balance on the whole
L
Linus Torvalds 已提交
473 474 475 476
	 * device if no resync is going on, or below the resync window.
	 * We take the first readable disk when above the resync window.
	 */
 retry:
477
	sectors = r1_bio->sectors;
N
NeilBrown 已提交
478 479
	best_disk = -1;
	best_dist = MaxSector;
480 481
	best_good_sectors = 0;

L
Linus Torvalds 已提交
482 483
	if (conf->mddev->recovery_cp < MaxSector &&
	    (this_sector + sectors >= conf->next_resync)) {
484 485 486 487 488
		choose_first = 1;
		start_disk = 0;
	} else {
		choose_first = 0;
		start_disk = conf->last_used;
L
Linus Torvalds 已提交
489 490
	}

491
	for (i = 0 ; i < conf->raid_disks ; i++) {
N
NeilBrown 已提交
492
		sector_t dist;
493 494 495
		sector_t first_bad;
		int bad_sectors;

496 497 498 499 500 501 502
		int disk = start_disk + i;
		if (disk >= conf->raid_disks)
			disk -= conf->raid_disks;

		rdev = rcu_dereference(conf->mirrors[disk].rdev);
		if (r1_bio->bios[disk] == IO_BLOCKED
		    || rdev == NULL
N
NeilBrown 已提交
503
		    || test_bit(Faulty, &rdev->flags))
504
			continue;
N
NeilBrown 已提交
505 506
		if (!test_bit(In_sync, &rdev->flags) &&
		    rdev->recovery_offset < this_sector + sectors)
L
Linus Torvalds 已提交
507
			continue;
N
NeilBrown 已提交
508 509 510 511 512 513 514 515 516 517
		if (test_bit(WriteMostly, &rdev->flags)) {
			/* Don't balance among write-mostly, just
			 * use the first as a last resort */
			if (best_disk < 0)
				best_disk = disk;
			continue;
		}
		/* This is a reasonable device to use.  It might
		 * even be best.
		 */
518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546
		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;

N
NeilBrown 已提交
547 548 549 550 551 552 553 554
		dist = abs(this_sector - conf->mirrors[disk].head_position);
		if (choose_first
		    /* Don't change to another disk for sequential reads */
		    || conf->next_seq_sect == this_sector
		    || dist == 0
		    /* If device is idle, use it */
		    || atomic_read(&rdev->nr_pending) == 0) {
			best_disk = disk;
L
Linus Torvalds 已提交
555 556
			break;
		}
N
NeilBrown 已提交
557 558 559
		if (dist < best_dist) {
			best_dist = dist;
			best_disk = disk;
L
Linus Torvalds 已提交
560
		}
561
	}
L
Linus Torvalds 已提交
562

N
NeilBrown 已提交
563 564
	if (best_disk >= 0) {
		rdev = rcu_dereference(conf->mirrors[best_disk].rdev);
565 566 567
		if (!rdev)
			goto retry;
		atomic_inc(&rdev->nr_pending);
N
NeilBrown 已提交
568
		if (test_bit(Faulty, &rdev->flags)) {
L
Linus Torvalds 已提交
569 570 571
			/* cannot risk returning a device that failed
			 * before we inc'ed nr_pending
			 */
572
			rdev_dec_pending(rdev, conf->mddev);
L
Linus Torvalds 已提交
573 574
			goto retry;
		}
575
		sectors = best_good_sectors;
576
		conf->next_seq_sect = this_sector + sectors;
N
NeilBrown 已提交
577
		conf->last_used = best_disk;
L
Linus Torvalds 已提交
578 579
	}
	rcu_read_unlock();
580
	*max_sectors = sectors;
L
Linus Torvalds 已提交
581

N
NeilBrown 已提交
582
	return best_disk;
L
Linus Torvalds 已提交
583 584
}

585
int md_raid1_congested(mddev_t *mddev, int bits)
586
{
587
	conf_t *conf = mddev->private;
588 589 590 591 592 593
	int i, ret = 0;

	rcu_read_lock();
	for (i = 0; i < mddev->raid_disks; i++) {
		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
		if (rdev && !test_bit(Faulty, &rdev->flags)) {
594
			struct request_queue *q = bdev_get_queue(rdev->bdev);
595

596 597
			BUG_ON(!q);

598 599 600
			/* Note the '|| 1' - when read_balance prefers
			 * non-congested targets, it can be removed
			 */
601
			if ((bits & (1<<BDI_async_congested)) || 1)
602 603 604 605 606 607 608 609
				ret |= bdi_congested(&q->backing_dev_info, bits);
			else
				ret &= bdi_congested(&q->backing_dev_info, bits);
		}
	}
	rcu_read_unlock();
	return ret;
}
610
EXPORT_SYMBOL_GPL(md_raid1_congested);
611

612 613 614 615 616 617 618
static int raid1_congested(void *data, int bits)
{
	mddev_t *mddev = data;

	return mddev_congested(mddev, bits) ||
		md_raid1_congested(mddev, bits);
}
619

J
Jens Axboe 已提交
620
static void flush_pending_writes(conf_t *conf)
621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642
{
	/* 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);
		spin_unlock_irq(&conf->device_lock);
		/* flush any pending bitmap writes to
		 * disk before proceeding w/ I/O */
		bitmap_unplug(conf->mddev->bitmap);

		while (bio) { /* submit pending writes */
			struct bio *next = bio->bi_next;
			bio->bi_next = NULL;
			generic_make_request(bio);
			bio = next;
		}
	} else
		spin_unlock_irq(&conf->device_lock);
J
Jens Axboe 已提交
643 644
}

645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664
/* Barriers....
 * Sometimes we need to suspend IO while we do something else,
 * either some resync/recovery, or reconfigure the array.
 * To do this we raise a 'barrier'.
 * The 'barrier' is a counter that can be raised multiple times
 * to count how many activities are happening which preclude
 * normal IO.
 * We can only raise the barrier if there is no pending IO.
 * i.e. if nr_pending == 0.
 * We choose only to raise the barrier if no-one is waiting for the
 * barrier to go down.  This means that as soon as an IO request
 * is ready, no other operations which require a barrier will start
 * until the IO request has had a chance.
 *
 * So: regular IO calls 'wait_barrier'.  When that returns there
 *    is no backgroup IO happening,  It must arrange to call
 *    allow_barrier when it has finished its IO.
 * backgroup IO calls must call raise_barrier.  Once that returns
 *    there is no normal IO happeing.  It must arrange to call
 *    lower_barrier when the particular background IO completes.
L
Linus Torvalds 已提交
665 666 667
 */
#define RESYNC_DEPTH 32

668
static void raise_barrier(conf_t *conf)
L
Linus Torvalds 已提交
669 670
{
	spin_lock_irq(&conf->resync_lock);
671 672 673

	/* Wait until no block IO is waiting */
	wait_event_lock_irq(conf->wait_barrier, !conf->nr_waiting,
N
NeilBrown 已提交
674
			    conf->resync_lock, );
675 676 677 678

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

N
NeilBrown 已提交
679
	/* Now wait for all pending IO to complete */
680 681
	wait_event_lock_irq(conf->wait_barrier,
			    !conf->nr_pending && conf->barrier < RESYNC_DEPTH,
N
NeilBrown 已提交
682
			    conf->resync_lock, );
683 684 685 686 687 688 689

	spin_unlock_irq(&conf->resync_lock);
}

static void lower_barrier(conf_t *conf)
{
	unsigned long flags;
690
	BUG_ON(conf->barrier <= 0);
691 692 693 694 695 696 697 698 699 700 701 702 703
	spin_lock_irqsave(&conf->resync_lock, flags);
	conf->barrier--;
	spin_unlock_irqrestore(&conf->resync_lock, flags);
	wake_up(&conf->wait_barrier);
}

static void wait_barrier(conf_t *conf)
{
	spin_lock_irq(&conf->resync_lock);
	if (conf->barrier) {
		conf->nr_waiting++;
		wait_event_lock_irq(conf->wait_barrier, !conf->barrier,
				    conf->resync_lock,
N
NeilBrown 已提交
704
				    );
705
		conf->nr_waiting--;
L
Linus Torvalds 已提交
706
	}
707
	conf->nr_pending++;
L
Linus Torvalds 已提交
708 709 710
	spin_unlock_irq(&conf->resync_lock);
}

711 712 713 714 715 716 717 718 719
static void allow_barrier(conf_t *conf)
{
	unsigned long flags;
	spin_lock_irqsave(&conf->resync_lock, flags);
	conf->nr_pending--;
	spin_unlock_irqrestore(&conf->resync_lock, flags);
	wake_up(&conf->wait_barrier);
}

720 721 722 723 724
static void freeze_array(conf_t *conf)
{
	/* stop syncio and normal IO and wait for everything to
	 * go quite.
	 * We increment barrier and nr_waiting, and then
725 726 727 728 729 730 731 732
	 * wait until nr_pending match nr_queued+1
	 * This is called in the context of one normal IO request
	 * that has failed. Thus any sync request that might be pending
	 * will be blocked by nr_pending, and we need to wait for
	 * pending IO requests to complete or be queued for re-try.
	 * Thus the number queued (nr_queued) plus this request (1)
	 * must match the number of pending IOs (nr_pending) before
	 * we continue.
733 734 735 736 737
	 */
	spin_lock_irq(&conf->resync_lock);
	conf->barrier++;
	conf->nr_waiting++;
	wait_event_lock_irq(conf->wait_barrier,
738
			    conf->nr_pending == conf->nr_queued+1,
739
			    conf->resync_lock,
N
NeilBrown 已提交
740
			    flush_pending_writes(conf));
741 742 743 744 745 746 747 748 749 750 751 752
	spin_unlock_irq(&conf->resync_lock);
}
static void unfreeze_array(conf_t *conf)
{
	/* reverse the effect of the freeze */
	spin_lock_irq(&conf->resync_lock);
	conf->barrier--;
	conf->nr_waiting--;
	wake_up(&conf->wait_barrier);
	spin_unlock_irq(&conf->resync_lock);
}

753

754 755
/* duplicate the data pages for behind I/O 
 */
756
static void alloc_behind_pages(struct bio *bio, r1bio_t *r1_bio)
757 758 759
{
	int i;
	struct bio_vec *bvec;
760
	struct bio_vec *bvecs = kzalloc(bio->bi_vcnt * sizeof(struct bio_vec),
761
					GFP_NOIO);
762
	if (unlikely(!bvecs))
763
		return;
764 765

	bio_for_each_segment(bvec, bio, i) {
766 767 768
		bvecs[i] = *bvec;
		bvecs[i].bv_page = alloc_page(GFP_NOIO);
		if (unlikely(!bvecs[i].bv_page))
769
			goto do_sync_io;
770 771 772
		memcpy(kmap(bvecs[i].bv_page) + bvec->bv_offset,
		       kmap(bvec->bv_page) + bvec->bv_offset, bvec->bv_len);
		kunmap(bvecs[i].bv_page);
773 774
		kunmap(bvec->bv_page);
	}
775
	r1_bio->behind_bvecs = bvecs;
776 777 778
	r1_bio->behind_page_count = bio->bi_vcnt;
	set_bit(R1BIO_BehindIO, &r1_bio->state);
	return;
779 780

do_sync_io:
781
	for (i = 0; i < bio->bi_vcnt; i++)
782 783 784
		if (bvecs[i].bv_page)
			put_page(bvecs[i].bv_page);
	kfree(bvecs);
785 786 787
	PRINTK("%dB behind alloc failed, doing sync I/O\n", bio->bi_size);
}

788
static int make_request(mddev_t *mddev, struct bio * bio)
L
Linus Torvalds 已提交
789
{
790
	conf_t *conf = mddev->private;
L
Linus Torvalds 已提交
791 792 793
	mirror_info_t *mirror;
	r1bio_t *r1_bio;
	struct bio *read_bio;
794
	int i, disks;
795
	struct bitmap *bitmap;
796
	unsigned long flags;
797
	const int rw = bio_data_dir(bio);
798
	const unsigned long do_sync = (bio->bi_rw & REQ_SYNC);
T
Tejun Heo 已提交
799
	const unsigned long do_flush_fua = (bio->bi_rw & (REQ_FLUSH | REQ_FUA));
800
	mdk_rdev_t *blocked_rdev;
N
NeilBrown 已提交
801
	int plugged;
802 803 804
	int first_clone;
	int sectors_handled;
	int max_sectors;
805

L
Linus Torvalds 已提交
806 807 808 809 810
	/*
	 * 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.
	 */
811

812 813
	md_write_start(mddev, bio); /* wait on superblock update early */

814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832
	if (bio_data_dir(bio) == WRITE &&
	    bio->bi_sector + bio->bi_size/512 > mddev->suspend_lo &&
	    bio->bi_sector < mddev->suspend_hi) {
		/* As the suspend_* range is controlled by
		 * userspace, we want an interruptible
		 * wait.
		 */
		DEFINE_WAIT(w);
		for (;;) {
			flush_signals(current);
			prepare_to_wait(&conf->wait_barrier,
					&w, TASK_INTERRUPTIBLE);
			if (bio->bi_sector + bio->bi_size/512 <= mddev->suspend_lo ||
			    bio->bi_sector >= mddev->suspend_hi)
				break;
			schedule();
		}
		finish_wait(&conf->wait_barrier, &w);
	}
833

834
	wait_barrier(conf);
L
Linus Torvalds 已提交
835

836 837
	bitmap = mddev->bitmap;

L
Linus Torvalds 已提交
838 839 840 841 842 843 844 845 846
	/*
	 * 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;
	r1_bio->sectors = bio->bi_size >> 9;
847
	r1_bio->state = 0;
L
Linus Torvalds 已提交
848 849 850
	r1_bio->mddev = mddev;
	r1_bio->sector = bio->bi_sector;

851 852 853 854 855 856 857 858 859 860
	/* 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);

861
	if (rw == READ) {
L
Linus Torvalds 已提交
862 863 864
		/*
		 * read balancing logic:
		 */
865 866 867 868
		int rdisk;

read_again:
		rdisk = read_balance(conf, r1_bio, &max_sectors);
L
Linus Torvalds 已提交
869 870 871 872 873 874 875 876

		if (rdisk < 0) {
			/* couldn't find anywhere to read from */
			raid_end_bio_io(r1_bio);
			return 0;
		}
		mirror = conf->mirrors + rdisk;

877 878 879 880 881 882 883 884 885
		if (test_bit(WriteMostly, &mirror->rdev->flags) &&
		    bitmap) {
			/* Reading from a write-mostly device must
			 * take care not to over-take any writes
			 * that are 'behind'
			 */
			wait_event(bitmap->behind_wait,
				   atomic_read(&bitmap->behind_writes) == 0);
		}
L
Linus Torvalds 已提交
886 887
		r1_bio->read_disk = rdisk;

888
		read_bio = bio_clone_mddev(bio, GFP_NOIO, mddev);
889 890
		md_trim_bio(read_bio, r1_bio->sector - bio->bi_sector,
			    max_sectors);
L
Linus Torvalds 已提交
891 892 893 894 895 896

		r1_bio->bios[rdisk] = read_bio;

		read_bio->bi_sector = r1_bio->sector + mirror->rdev->data_offset;
		read_bio->bi_bdev = mirror->rdev->bdev;
		read_bio->bi_end_io = raid1_end_read_request;
897
		read_bio->bi_rw = READ | do_sync;
L
Linus Torvalds 已提交
898 899
		read_bio->bi_private = r1_bio;

900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930
		if (max_sectors < r1_bio->sectors) {
			/* could not read all from this device, so we will
			 * need another r1_bio.
			 */

			sectors_handled = (r1_bio->sector + max_sectors
					   - bio->bi_sector);
			r1_bio->sectors = max_sectors;
			spin_lock_irq(&conf->device_lock);
			if (bio->bi_phys_segments == 0)
				bio->bi_phys_segments = 2;
			else
				bio->bi_phys_segments++;
			spin_unlock_irq(&conf->device_lock);
			/* Cannot call generic_make_request directly
			 * as that will be queued in __make_request
			 * and subsequent mempool_alloc might block waiting
			 * for it.  So hand bio over to raid1d.
			 */
			reschedule_retry(r1_bio);

			r1_bio = mempool_alloc(conf->r1bio_pool, GFP_NOIO);

			r1_bio->master_bio = bio;
			r1_bio->sectors = (bio->bi_size >> 9) - sectors_handled;
			r1_bio->state = 0;
			r1_bio->mddev = mddev;
			r1_bio->sector = bio->bi_sector + sectors_handled;
			goto read_again;
		} else
			generic_make_request(read_bio);
L
Linus Torvalds 已提交
931 932 933 934 935 936
		return 0;
	}

	/*
	 * WRITE:
	 */
937
	/* first select target devices under rcu_lock and
L
Linus Torvalds 已提交
938 939
	 * inc refcount on their rdev.  Record them by setting
	 * bios[x] to bio
940 941 942 943 944 945
	 * If there are known/acknowledged bad blocks on any device on
	 * which we have seen a write error, we want to avoid writing those
	 * blocks.
	 * This potentially requires several writes to write around
	 * the bad blocks.  Each set of writes gets it's own r1bio
	 * with a set of bios attached.
L
Linus Torvalds 已提交
946
	 */
N
NeilBrown 已提交
947 948
	plugged = mddev_check_plugged(mddev);

L
Linus Torvalds 已提交
949
	disks = conf->raid_disks;
950 951
 retry_write:
	blocked_rdev = NULL;
L
Linus Torvalds 已提交
952
	rcu_read_lock();
953
	max_sectors = r1_bio->sectors;
L
Linus Torvalds 已提交
954
	for (i = 0;  i < disks; i++) {
955 956 957 958 959 960
		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
		if (rdev && unlikely(test_bit(Blocked, &rdev->flags))) {
			atomic_inc(&rdev->nr_pending);
			blocked_rdev = rdev;
			break;
		}
961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990
		r1_bio->bios[i] = NULL;
		if (!rdev || test_bit(Faulty, &rdev->flags)) {
			set_bit(R1BIO_Degraded, &r1_bio->state);
			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;
991
				rdev_dec_pending(rdev, mddev);
992 993 994 995 996 997 998 999 1000 1001 1002
				/* 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;
1003
			}
1004 1005 1006 1007 1008 1009 1010
			if (is_bad) {
				int good_sectors = first_bad - r1_bio->sector;
				if (good_sectors < max_sectors)
					max_sectors = good_sectors;
			}
		}
		r1_bio->bios[i] = bio;
L
Linus Torvalds 已提交
1011 1012 1013
	}
	rcu_read_unlock();

1014 1015 1016 1017 1018 1019 1020
	if (unlikely(blocked_rdev)) {
		/* Wait for this device to become unblocked */
		int j;

		for (j = 0; j < i; j++)
			if (r1_bio->bios[j])
				rdev_dec_pending(conf->mirrors[j].rdev, mddev);
1021
		r1_bio->state = 0;
1022 1023 1024 1025 1026 1027
		allow_barrier(conf);
		md_wait_for_blocked_rdev(blocked_rdev, mddev);
		wait_barrier(conf);
		goto retry_write;
	}

1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038
	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);
1039
	}
1040
	sectors_handled = r1_bio->sector + max_sectors - bio->bi_sector;
1041

1042
	atomic_set(&r1_bio->remaining, 1);
1043
	atomic_set(&r1_bio->behind_remaining, 0);
1044

1045
	first_clone = 1;
L
Linus Torvalds 已提交
1046 1047 1048 1049 1050
	for (i = 0; i < disks; i++) {
		struct bio *mbio;
		if (!r1_bio->bios[i])
			continue;

1051
		mbio = bio_clone_mddev(bio, GFP_NOIO, mddev);
1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070
		md_trim_bio(mbio, r1_bio->sector - bio->bi_sector, max_sectors);

		if (first_clone) {
			/* do behind I/O ?
			 * Not if there are too many, or cannot
			 * allocate memory, or a reader on WriteMostly
			 * is waiting for behind writes to flush */
			if (bitmap &&
			    (atomic_read(&bitmap->behind_writes)
			     < mddev->bitmap_info.max_write_behind) &&
			    !waitqueue_active(&bitmap->behind_wait))
				alloc_behind_pages(mbio, r1_bio);

			bitmap_startwrite(bitmap, r1_bio->sector,
					  r1_bio->sectors,
					  test_bit(R1BIO_BehindIO,
						   &r1_bio->state));
			first_clone = 0;
		}
1071
		if (r1_bio->behind_bvecs) {
1072 1073 1074 1075 1076 1077 1078
			struct bio_vec *bvec;
			int j;

			/* Yes, I really want the '__' version so that
			 * we clear any unused pointer in the io_vec, rather
			 * than leave them unchanged.  This is important
			 * because when we come to free the pages, we won't
N
NeilBrown 已提交
1079
			 * know the original bi_idx, so we just free
1080 1081 1082
			 * them all
			 */
			__bio_for_each_segment(bvec, mbio, j, 0)
1083
				bvec->bv_page = r1_bio->behind_bvecs[j].bv_page;
1084 1085 1086 1087
			if (test_bit(WriteMostly, &conf->mirrors[i].rdev->flags))
				atomic_inc(&r1_bio->behind_remaining);
		}

1088 1089 1090 1091 1092 1093 1094 1095 1096
		r1_bio->bios[i] = mbio;

		mbio->bi_sector	= (r1_bio->sector +
				   conf->mirrors[i].rdev->data_offset);
		mbio->bi_bdev = conf->mirrors[i].rdev->bdev;
		mbio->bi_end_io	= raid1_end_write_request;
		mbio->bi_rw = WRITE | do_flush_fua | do_sync;
		mbio->bi_private = r1_bio;

L
Linus Torvalds 已提交
1097
		atomic_inc(&r1_bio->remaining);
1098 1099 1100
		spin_lock_irqsave(&conf->device_lock, flags);
		bio_list_add(&conf->pending_bio_list, mbio);
		spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1101
	}
1102
	r1_bio_write_done(r1_bio);
L
Linus Torvalds 已提交
1103

1104
	/* In case raid1d snuck in to freeze_array */
1105 1106
	wake_up(&conf->wait_barrier);

1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119
	if (sectors_handled < (bio->bi_size >> 9)) {
		/* 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;
		r1_bio->sectors = (bio->bi_size >> 9) - sectors_handled;
		r1_bio->state = 0;
		r1_bio->mddev = mddev;
		r1_bio->sector = bio->bi_sector + sectors_handled;
		goto retry_write;
	}

N
NeilBrown 已提交
1120
	if (do_sync || !bitmap || !plugged)
1121
		md_wakeup_thread(mddev->thread);
1122

L
Linus Torvalds 已提交
1123 1124 1125 1126 1127
	return 0;
}

static void status(struct seq_file *seq, mddev_t *mddev)
{
1128
	conf_t *conf = mddev->private;
L
Linus Torvalds 已提交
1129 1130 1131
	int i;

	seq_printf(seq, " [%d/%d] [", conf->raid_disks,
1132
		   conf->raid_disks - mddev->degraded);
1133 1134 1135
	rcu_read_lock();
	for (i = 0; i < conf->raid_disks; i++) {
		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
L
Linus Torvalds 已提交
1136
		seq_printf(seq, "%s",
1137 1138 1139
			   rdev && test_bit(In_sync, &rdev->flags) ? "U" : "_");
	}
	rcu_read_unlock();
L
Linus Torvalds 已提交
1140 1141 1142 1143 1144 1145 1146
	seq_printf(seq, "]");
}


static void error(mddev_t *mddev, mdk_rdev_t *rdev)
{
	char b[BDEVNAME_SIZE];
1147
	conf_t *conf = mddev->private;
L
Linus Torvalds 已提交
1148 1149 1150 1151 1152 1153 1154

	/*
	 * 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
	 */
1155
	if (test_bit(In_sync, &rdev->flags)
1156
	    && (conf->raid_disks - mddev->degraded) == 1) {
L
Linus Torvalds 已提交
1157 1158
		/*
		 * Don't fail the drive, act as though we were just a
1159 1160 1161
		 * normal single drive.
		 * However don't try a recovery from this drive as
		 * it is very likely to fail.
L
Linus Torvalds 已提交
1162
		 */
1163
		conf->recovery_disabled = mddev->recovery_disabled;
L
Linus Torvalds 已提交
1164
		return;
1165
	}
1166
	set_bit(Blocked, &rdev->flags);
1167 1168 1169
	if (test_and_clear_bit(In_sync, &rdev->flags)) {
		unsigned long flags;
		spin_lock_irqsave(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1170
		mddev->degraded++;
1171
		set_bit(Faulty, &rdev->flags);
1172
		spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1173 1174 1175
		/*
		 * if recovery is running, make sure it aborts.
		 */
1176
		set_bit(MD_RECOVERY_INTR, &mddev->recovery);
1177 1178
	} else
		set_bit(Faulty, &rdev->flags);
1179
	set_bit(MD_CHANGE_DEVS, &mddev->flags);
1180 1181 1182
	printk(KERN_ALERT
	       "md/raid1:%s: Disk failure on %s, disabling device.\n"
	       "md/raid1:%s: Operation continuing on %d devices.\n",
N
NeilBrown 已提交
1183 1184
	       mdname(mddev), bdevname(rdev->bdev, b),
	       mdname(mddev), conf->raid_disks - mddev->degraded);
L
Linus Torvalds 已提交
1185 1186 1187 1188 1189 1190
}

static void print_conf(conf_t *conf)
{
	int i;

N
NeilBrown 已提交
1191
	printk(KERN_DEBUG "RAID1 conf printout:\n");
L
Linus Torvalds 已提交
1192
	if (!conf) {
N
NeilBrown 已提交
1193
		printk(KERN_DEBUG "(!conf)\n");
L
Linus Torvalds 已提交
1194 1195
		return;
	}
N
NeilBrown 已提交
1196
	printk(KERN_DEBUG " --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
L
Linus Torvalds 已提交
1197 1198
		conf->raid_disks);

1199
	rcu_read_lock();
L
Linus Torvalds 已提交
1200 1201
	for (i = 0; i < conf->raid_disks; i++) {
		char b[BDEVNAME_SIZE];
1202 1203
		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
		if (rdev)
N
NeilBrown 已提交
1204
			printk(KERN_DEBUG " disk %d, wo:%d, o:%d, dev:%s\n",
1205 1206 1207
			       i, !test_bit(In_sync, &rdev->flags),
			       !test_bit(Faulty, &rdev->flags),
			       bdevname(rdev->bdev,b));
L
Linus Torvalds 已提交
1208
	}
1209
	rcu_read_unlock();
L
Linus Torvalds 已提交
1210 1211 1212 1213
}

static void close_sync(conf_t *conf)
{
1214 1215
	wait_barrier(conf);
	allow_barrier(conf);
L
Linus Torvalds 已提交
1216 1217 1218 1219 1220 1221 1222 1223 1224

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

static int raid1_spare_active(mddev_t *mddev)
{
	int i;
	conf_t *conf = mddev->private;
1225 1226
	int count = 0;
	unsigned long flags;
L
Linus Torvalds 已提交
1227 1228 1229

	/*
	 * Find all failed disks within the RAID1 configuration 
1230 1231
	 * and mark them readable.
	 * Called under mddev lock, so rcu protection not needed.
L
Linus Torvalds 已提交
1232 1233
	 */
	for (i = 0; i < conf->raid_disks; i++) {
1234 1235 1236
		mdk_rdev_t *rdev = conf->mirrors[i].rdev;
		if (rdev
		    && !test_bit(Faulty, &rdev->flags)
1237
		    && !test_and_set_bit(In_sync, &rdev->flags)) {
1238
			count++;
1239
			sysfs_notify_dirent_safe(rdev->sysfs_state);
L
Linus Torvalds 已提交
1240 1241
		}
	}
1242 1243 1244
	spin_lock_irqsave(&conf->device_lock, flags);
	mddev->degraded -= count;
	spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1245 1246

	print_conf(conf);
1247
	return count;
L
Linus Torvalds 已提交
1248 1249 1250 1251 1252 1253
}


static int raid1_add_disk(mddev_t *mddev, mdk_rdev_t *rdev)
{
	conf_t *conf = mddev->private;
1254
	int err = -EEXIST;
1255
	int mirror = 0;
L
Linus Torvalds 已提交
1256
	mirror_info_t *p;
1257 1258
	int first = 0;
	int last = mddev->raid_disks - 1;
L
Linus Torvalds 已提交
1259

1260 1261 1262
	if (mddev->recovery_disabled == conf->recovery_disabled)
		return -EBUSY;

1263 1264 1265 1266
	if (rdev->raid_disk >= 0)
		first = last = rdev->raid_disk;

	for (mirror = first; mirror <= last; mirror++)
L
Linus Torvalds 已提交
1267 1268
		if ( !(p=conf->mirrors+mirror)->rdev) {

1269 1270
			disk_stack_limits(mddev->gendisk, rdev->bdev,
					  rdev->data_offset << 9);
1271 1272 1273 1274 1275
			/* as we don't honour merge_bvec_fn, we must
			 * never risk violating it, so limit
			 * ->max_segments to one lying with a single
			 * page, as a one page request is never in
			 * violation.
L
Linus Torvalds 已提交
1276
			 */
1277 1278 1279 1280 1281
			if (rdev->bdev->bd_disk->queue->merge_bvec_fn) {
				blk_queue_max_segments(mddev->queue, 1);
				blk_queue_segment_boundary(mddev->queue,
							   PAGE_CACHE_SIZE - 1);
			}
L
Linus Torvalds 已提交
1282 1283 1284

			p->head_position = 0;
			rdev->raid_disk = mirror;
1285
			err = 0;
1286 1287 1288 1289
			/* 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)
1290
				conf->fullsync = 1;
1291
			rcu_assign_pointer(p->rdev, rdev);
L
Linus Torvalds 已提交
1292 1293
			break;
		}
1294
	md_integrity_add_rdev(rdev, mddev);
L
Linus Torvalds 已提交
1295
	print_conf(conf);
1296
	return err;
L
Linus Torvalds 已提交
1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308
}

static int raid1_remove_disk(mddev_t *mddev, int number)
{
	conf_t *conf = mddev->private;
	int err = 0;
	mdk_rdev_t *rdev;
	mirror_info_t *p = conf->mirrors+ number;

	print_conf(conf);
	rdev = p->rdev;
	if (rdev) {
1309
		if (test_bit(In_sync, &rdev->flags) ||
L
Linus Torvalds 已提交
1310 1311 1312 1313
		    atomic_read(&rdev->nr_pending)) {
			err = -EBUSY;
			goto abort;
		}
N
NeilBrown 已提交
1314
		/* Only remove non-faulty devices if recovery
1315 1316 1317
		 * is not possible.
		 */
		if (!test_bit(Faulty, &rdev->flags) &&
1318
		    mddev->recovery_disabled != conf->recovery_disabled &&
1319 1320 1321 1322
		    mddev->degraded < conf->raid_disks) {
			err = -EBUSY;
			goto abort;
		}
L
Linus Torvalds 已提交
1323
		p->rdev = NULL;
1324
		synchronize_rcu();
L
Linus Torvalds 已提交
1325 1326 1327 1328
		if (atomic_read(&rdev->nr_pending)) {
			/* lost the race, try later */
			err = -EBUSY;
			p->rdev = rdev;
1329
			goto abort;
L
Linus Torvalds 已提交
1330
		}
1331
		err = md_integrity_register(mddev);
L
Linus Torvalds 已提交
1332 1333 1334 1335 1336 1337 1338 1339
	}
abort:

	print_conf(conf);
	return err;
}


1340
static void end_sync_read(struct bio *bio, int error)
L
Linus Torvalds 已提交
1341
{
1342
	r1bio_t *r1_bio = bio->bi_private;
1343
	int i;
L
Linus Torvalds 已提交
1344

1345 1346 1347 1348 1349
	for (i=r1_bio->mddev->raid_disks; i--; )
		if (r1_bio->bios[i] == bio)
			break;
	BUG_ON(i < 0);
	update_head_pos(i, r1_bio);
L
Linus Torvalds 已提交
1350 1351 1352 1353 1354
	/*
	 * 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
	 */
1355
	if (test_bit(BIO_UPTODATE, &bio->bi_flags))
L
Linus Torvalds 已提交
1356
		set_bit(R1BIO_Uptodate, &r1_bio->state);
1357 1358 1359

	if (atomic_dec_and_test(&r1_bio->remaining))
		reschedule_retry(r1_bio);
L
Linus Torvalds 已提交
1360 1361
}

1362
static void end_sync_write(struct bio *bio, int error)
L
Linus Torvalds 已提交
1363 1364
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
1365
	r1bio_t *r1_bio = bio->bi_private;
L
Linus Torvalds 已提交
1366
	mddev_t *mddev = r1_bio->mddev;
1367
	conf_t *conf = mddev->private;
L
Linus Torvalds 已提交
1368 1369
	int i;
	int mirror=0;
1370 1371
	sector_t first_bad;
	int bad_sectors;
L
Linus Torvalds 已提交
1372 1373 1374 1375 1376 1377

	for (i = 0; i < conf->raid_disks; i++)
		if (r1_bio->bios[i] == bio) {
			mirror = i;
			break;
		}
1378
	if (!uptodate) {
N
NeilBrown 已提交
1379
		sector_t sync_blocks = 0;
1380 1381 1382 1383
		sector_t s = r1_bio->sector;
		long sectors_to_go = r1_bio->sectors;
		/* make sure these bits doesn't get cleared. */
		do {
1384
			bitmap_end_sync(mddev->bitmap, s,
1385 1386 1387 1388
					&sync_blocks, 1);
			s += sync_blocks;
			sectors_to_go -= sync_blocks;
		} while (sectors_to_go > 0);
1389 1390 1391
		set_bit(WriteErrorSeen,
			&conf->mirrors[mirror].rdev->flags);
		set_bit(R1BIO_WriteError, &r1_bio->state);
1392 1393 1394
	} else if (is_badblock(conf->mirrors[mirror].rdev,
			       r1_bio->sector,
			       r1_bio->sectors,
1395 1396 1397 1398 1399 1400
			       &first_bad, &bad_sectors) &&
		   !is_badblock(conf->mirrors[r1_bio->read_disk].rdev,
				r1_bio->sector,
				r1_bio->sectors,
				&first_bad, &bad_sectors)
		)
1401
		set_bit(R1BIO_MadeGood, &r1_bio->state);
1402

L
Linus Torvalds 已提交
1403 1404 1405
	update_head_pos(mirror, r1_bio);

	if (atomic_dec_and_test(&r1_bio->remaining)) {
1406
		int s = r1_bio->sectors;
1407 1408
		if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
		    test_bit(R1BIO_WriteError, &r1_bio->state))
1409 1410 1411 1412 1413
			reschedule_retry(r1_bio);
		else {
			put_buf(r1_bio);
			md_done_sync(mddev, s, uptodate);
		}
L
Linus Torvalds 已提交
1414 1415 1416
	}
}

1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430
static int r1_sync_page_io(mdk_rdev_t *rdev, sector_t sector,
			    int sectors, struct page *page, int rw)
{
	if (sync_page_io(rdev, sector, sectors << 9, page, rw, false))
		/* success */
		return 1;
	if (rw == WRITE)
		set_bit(WriteErrorSeen, &rdev->flags);
	/* 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;
}

1431
static int fix_sync_read_error(r1bio_t *r1_bio)
L
Linus Torvalds 已提交
1432
{
1433 1434 1435 1436 1437 1438 1439
	/* 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.
1440 1441 1442
	 * 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.
1443 1444
	 */
	mddev_t *mddev = r1_bio->mddev;
1445
	conf_t *conf = mddev->private;
1446 1447 1448 1449 1450 1451 1452 1453 1454 1455
	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;
		mdk_rdev_t *rdev;
1456
		int start;
1457 1458 1459 1460 1461 1462 1463 1464 1465 1466

		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;
1467
				if (sync_page_io(rdev, sect, s<<9,
1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478
						 bio->bi_io_vec[idx].bv_page,
						 READ, false)) {
					success = 1;
					break;
				}
			}
			d++;
			if (d == conf->raid_disks)
				d = 0;
		} while (!success && d != r1_bio->read_disk);

1479
		if (!success) {
1480
			char b[BDEVNAME_SIZE];
1481 1482 1483 1484 1485 1486
			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.
			 */
1487 1488 1489 1490 1491
			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);
1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510
			for (d = 0; d < conf->raid_disks; d++) {
				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) {
				mddev->recovery_disabled = 1;
				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;
1511
		}
1512 1513 1514 1515 1516 1517 1518 1519 1520 1521

		start = d;
		/* write it back and re-read */
		while (d != r1_bio->read_disk) {
			if (d == 0)
				d = conf->raid_disks;
			d--;
			if (r1_bio->bios[d]->bi_end_io != end_sync_read)
				continue;
			rdev = conf->mirrors[d].rdev;
1522 1523 1524
			if (r1_sync_page_io(rdev, sect, s,
					    bio->bi_io_vec[idx].bv_page,
					    WRITE) == 0) {
1525 1526
				r1_bio->bios[d]->bi_end_io = NULL;
				rdev_dec_pending(rdev, mddev);
1527
			}
1528 1529 1530 1531 1532 1533 1534 1535 1536
		}
		d = start;
		while (d != r1_bio->read_disk) {
			if (d == 0)
				d = conf->raid_disks;
			d--;
			if (r1_bio->bios[d]->bi_end_io != end_sync_read)
				continue;
			rdev = conf->mirrors[d].rdev;
1537 1538 1539
			if (r1_sync_page_io(rdev, sect, s,
					    bio->bi_io_vec[idx].bv_page,
					    READ) != 0)
1540
				atomic_add(s, &rdev->corrected_errors);
1541
		}
1542 1543 1544 1545
		sectors -= s;
		sect += s;
		idx ++;
	}
1546
	set_bit(R1BIO_Uptodate, &r1_bio->state);
1547
	set_bit(BIO_UPTODATE, &bio->bi_flags);
1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564
	return 1;
}

static int process_checks(r1bio_t *r1_bio)
{
	/* 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
	 */
	mddev_t *mddev = r1_bio->mddev;
	conf_t *conf = mddev->private;
	int primary;
	int i;

1565
	for (primary = 0; primary < conf->raid_disks; primary++)
1566 1567 1568 1569 1570 1571 1572
		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;
1573 1574 1575 1576 1577 1578
	for (i = 0; i < conf->raid_disks; i++) {
		int j;
		int vcnt = r1_bio->sectors >> (PAGE_SHIFT- 9);
		struct bio *pbio = r1_bio->bios[primary];
		struct bio *sbio = r1_bio->bios[i];
		int size;
1579

1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591
		if (r1_bio->bios[i]->bi_end_io != end_sync_read)
			continue;

		if (test_bit(BIO_UPTODATE, &sbio->bi_flags)) {
			for (j = vcnt; j-- ; ) {
				struct page *p, *s;
				p = pbio->bi_io_vec[j].bv_page;
				s = sbio->bi_io_vec[j].bv_page;
				if (memcmp(page_address(p),
					   page_address(s),
					   PAGE_SIZE))
					break;
1592
			}
1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627
		} else
			j = 0;
		if (j >= 0)
			mddev->resync_mismatches += r1_bio->sectors;
		if (j < 0 || (test_bit(MD_RECOVERY_CHECK, &mddev->recovery)
			      && test_bit(BIO_UPTODATE, &sbio->bi_flags))) {
			/* No need to write to this device. */
			sbio->bi_end_io = NULL;
			rdev_dec_pending(conf->mirrors[i].rdev, mddev);
			continue;
		}
		/* fixup the bio for reuse */
		sbio->bi_vcnt = vcnt;
		sbio->bi_size = r1_bio->sectors << 9;
		sbio->bi_idx = 0;
		sbio->bi_phys_segments = 0;
		sbio->bi_flags &= ~(BIO_POOL_MASK - 1);
		sbio->bi_flags |= 1 << BIO_UPTODATE;
		sbio->bi_next = NULL;
		sbio->bi_sector = r1_bio->sector +
			conf->mirrors[i].rdev->data_offset;
		sbio->bi_bdev = conf->mirrors[i].rdev->bdev;
		size = sbio->bi_size;
		for (j = 0; j < vcnt ; j++) {
			struct bio_vec *bi;
			bi = &sbio->bi_io_vec[j];
			bi->bv_offset = 0;
			if (size > PAGE_SIZE)
				bi->bv_len = PAGE_SIZE;
			else
				bi->bv_len = size;
			size -= PAGE_SIZE;
			memcpy(page_address(bi->bv_page),
			       page_address(pbio->bi_io_vec[j].bv_page),
			       PAGE_SIZE);
1628
		}
1629
	}
1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645
	return 0;
}

static void sync_request_write(mddev_t *mddev, r1bio_t *r1_bio)
{
	conf_t *conf = mddev->private;
	int i;
	int disks = conf->raid_disks;
	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;
1646 1647 1648 1649

	if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
		if (process_checks(r1_bio) < 0)
			return;
1650 1651 1652
	/*
	 * schedule writes
	 */
L
Linus Torvalds 已提交
1653 1654 1655
	atomic_set(&r1_bio->remaining, 1);
	for (i = 0; i < disks ; i++) {
		wbio = r1_bio->bios[i];
1656 1657 1658 1659
		if (wbio->bi_end_io == NULL ||
		    (wbio->bi_end_io == end_sync_read &&
		     (i == r1_bio->read_disk ||
		      !test_bit(MD_RECOVERY_SYNC, &mddev->recovery))))
L
Linus Torvalds 已提交
1660 1661
			continue;

1662 1663
		wbio->bi_rw = WRITE;
		wbio->bi_end_io = end_sync_write;
L
Linus Torvalds 已提交
1664 1665
		atomic_inc(&r1_bio->remaining);
		md_sync_acct(conf->mirrors[i].rdev->bdev, wbio->bi_size >> 9);
1666

L
Linus Torvalds 已提交
1667 1668 1669 1670
		generic_make_request(wbio);
	}

	if (atomic_dec_and_test(&r1_bio->remaining)) {
1671
		/* if we're here, all write(s) have completed, so clean up */
L
Linus Torvalds 已提交
1672 1673 1674 1675 1676 1677 1678 1679 1680 1681
		md_done_sync(mddev, r1_bio->sectors, 1);
		put_buf(r1_bio);
	}
}

/*
 * This is a kernel thread which:
 *
 *	1.	Retries failed read operations on working mirrors.
 *	2.	Updates the raid superblock when problems encounter.
1682
 *	3.	Performs writes following reads for array synchronising.
L
Linus Torvalds 已提交
1683 1684
 */

1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704
static void fix_read_error(conf_t *conf, int read_disk,
			   sector_t sect, int sectors)
{
	mddev_t *mddev = conf->mddev;
	while(sectors) {
		int s = sectors;
		int d = read_disk;
		int success = 0;
		int start;
		mdk_rdev_t *rdev;

		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....
			 */
1705 1706 1707
			sector_t first_bad;
			int bad_sectors;

1708 1709 1710
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
			    test_bit(In_sync, &rdev->flags) &&
1711 1712
			    is_badblock(rdev, sect, s,
					&first_bad, &bad_sectors) == 0 &&
J
Jonathan Brassow 已提交
1713 1714
			    sync_page_io(rdev, sect, s<<9,
					 conf->tmppage, READ, false))
1715 1716 1717 1718 1719 1720 1721 1722 1723
				success = 1;
			else {
				d++;
				if (d == conf->raid_disks)
					d = 0;
			}
		} while (!success && d != read_disk);

		if (!success) {
1724 1725 1726 1727
			/* Cannot read from anywhere - mark it bad */
			mdk_rdev_t *rdev = conf->mirrors[read_disk].rdev;
			if (!rdev_set_badblocks(rdev, sect, s, 0))
				md_error(mddev, rdev);
1728 1729 1730 1731 1732 1733 1734 1735 1736 1737
			break;
		}
		/* write it back and re-read */
		start = d;
		while (d != read_disk) {
			if (d==0)
				d = conf->raid_disks;
			d--;
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
1738 1739 1740
			    test_bit(In_sync, &rdev->flags))
				r1_sync_page_io(rdev, sect, s,
						conf->tmppage, WRITE);
1741 1742 1743 1744 1745 1746 1747 1748 1749 1750
		}
		d = start;
		while (d != read_disk) {
			char b[BDEVNAME_SIZE];
			if (d==0)
				d = conf->raid_disks;
			d--;
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
			    test_bit(In_sync, &rdev->flags)) {
1751 1752
				if (r1_sync_page_io(rdev, sect, s,
						    conf->tmppage, READ)) {
1753 1754
					atomic_add(s, &rdev->corrected_errors);
					printk(KERN_INFO
N
NeilBrown 已提交
1755
					       "md/raid1:%s: read error corrected "
1756 1757
					       "(%d sectors at %llu on %s)\n",
					       mdname(mddev), s,
1758 1759
					       (unsigned long long)(sect +
					           rdev->data_offset),
1760 1761 1762 1763 1764 1765 1766 1767 1768
					       bdevname(rdev->bdev, b));
				}
			}
		}
		sectors -= s;
		sect += s;
	}
}

1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863
static void bi_complete(struct bio *bio, int error)
{
	complete((struct completion *)bio->bi_private);
}

static int submit_bio_wait(int rw, struct bio *bio)
{
	struct completion event;
	rw |= REQ_SYNC;

	init_completion(&event);
	bio->bi_private = &event;
	bio->bi_end_io = bi_complete;
	submit_bio(rw, bio);
	wait_for_completion(&event);

	return test_bit(BIO_UPTODATE, &bio->bi_flags);
}

static int narrow_write_error(r1bio_t *r1_bio, int i)
{
	mddev_t *mddev = r1_bio->mddev;
	conf_t *conf = mddev->private;
	mdk_rdev_t *rdev = conf->mirrors[i].rdev;
	int vcnt, idx;
	struct bio_vec *vec;

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

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

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

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

	if (test_bit(R1BIO_BehindIO, &r1_bio->state)) {
		vcnt = r1_bio->behind_page_count;
		vec = r1_bio->behind_bvecs;
		idx = 0;
		while (vec[idx].bv_page == NULL)
			idx++;
	} else {
		vcnt = r1_bio->master_bio->bi_vcnt;
		vec = r1_bio->master_bio->bi_io_vec;
		idx = r1_bio->master_bio->bi_idx;
	}
	while (sect_to_write) {
		struct bio *wbio;
		if (sectors > sect_to_write)
			sectors = sect_to_write;
		/* Write at 'sector' for 'sectors'*/

		wbio = bio_alloc_mddev(GFP_NOIO, vcnt, mddev);
		memcpy(wbio->bi_io_vec, vec, vcnt * sizeof(struct bio_vec));
		wbio->bi_sector = r1_bio->sector;
		wbio->bi_rw = WRITE;
		wbio->bi_vcnt = vcnt;
		wbio->bi_size = r1_bio->sectors << 9;
		wbio->bi_idx = idx;

		md_trim_bio(wbio, sector - r1_bio->sector, sectors);
		wbio->bi_sector += rdev->data_offset;
		wbio->bi_bdev = rdev->bdev;
		if (submit_bio_wait(WRITE, wbio) == 0)
			/* failure! */
			ok = rdev_set_badblocks(rdev, sector,
						sectors, 0)
				&& ok;

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

1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
static void handle_sync_write_finished(conf_t *conf, r1bio_t *r1_bio)
{
	int m;
	int s = r1_bio->sectors;
	for (m = 0; m < conf->raid_disks ; m++) {
		mdk_rdev_t *rdev = conf->mirrors[m].rdev;
		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)) {
			rdev_clear_badblocks(rdev, r1_bio->sector, s);
		}
		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);
}

static void handle_write_finished(conf_t *conf, r1bio_t *r1_bio)
{
	int m;
	for (m = 0; m < conf->raid_disks ; m++)
		if (r1_bio->bios[m] == IO_MADE_GOOD) {
			mdk_rdev_t *rdev = conf->mirrors[m].rdev;
			rdev_clear_badblocks(rdev,
					     r1_bio->sector,
					     r1_bio->sectors);
			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);
}

static void handle_read_error(conf_t *conf, r1bio_t *r1_bio)
{
	int disk;
	int max_sectors;
	mddev_t *mddev = conf->mddev;
	struct bio *bio;
	char b[BDEVNAME_SIZE];
	mdk_rdev_t *rdev;

	clear_bit(R1BIO_ReadError, &r1_bio->state);
	/* we got a read error. Maybe the drive is bad.  Maybe just
	 * the block and we can fix it.
	 * We freeze all other IO, and try reading the block from
	 * other devices.  When we find one, we re-write
	 * and check it that fixes the read error.
	 * This is all done synchronously while the array is
	 * frozen
	 */
	if (mddev->ro == 0) {
		freeze_array(conf);
		fix_read_error(conf, r1_bio->read_disk,
			       r1_bio->sector, r1_bio->sectors);
		unfreeze_array(conf);
	} else
		md_error(mddev, conf->mirrors[r1_bio->read_disk].rdev);

	bio = r1_bio->bios[r1_bio->read_disk];
	bdevname(bio->bi_bdev, b);
read_more:
	disk = read_balance(conf, r1_bio, &max_sectors);
	if (disk == -1) {
		printk(KERN_ALERT "md/raid1:%s: %s: unrecoverable I/O"
		       " read error for block %llu\n",
		       mdname(mddev), b, (unsigned long long)r1_bio->sector);
		raid_end_bio_io(r1_bio);
	} else {
		const unsigned long do_sync
			= r1_bio->master_bio->bi_rw & REQ_SYNC;
		if (bio) {
			r1_bio->bios[r1_bio->read_disk] =
				mddev->ro ? IO_BLOCKED : NULL;
			bio_put(bio);
		}
		r1_bio->read_disk = disk;
		bio = bio_clone_mddev(r1_bio->master_bio, GFP_NOIO, mddev);
		md_trim_bio(bio, r1_bio->sector - bio->bi_sector, max_sectors);
		r1_bio->bios[r1_bio->read_disk] = bio;
		rdev = conf->mirrors[disk].rdev;
		printk_ratelimited(KERN_ERR
				   "md/raid1:%s: redirecting sector %llu"
				   " to other mirror: %s\n",
				   mdname(mddev),
				   (unsigned long long)r1_bio->sector,
				   bdevname(rdev->bdev, b));
		bio->bi_sector = r1_bio->sector + rdev->data_offset;
		bio->bi_bdev = rdev->bdev;
		bio->bi_end_io = raid1_end_read_request;
		bio->bi_rw = READ | do_sync;
		bio->bi_private = r1_bio;
		if (max_sectors < r1_bio->sectors) {
			/* Drat - have to split this up more */
			struct bio *mbio = r1_bio->master_bio;
			int sectors_handled = (r1_bio->sector + max_sectors
					       - mbio->bi_sector);
			r1_bio->sectors = max_sectors;
			spin_lock_irq(&conf->device_lock);
			if (mbio->bi_phys_segments == 0)
				mbio->bi_phys_segments = 2;
			else
				mbio->bi_phys_segments++;
			spin_unlock_irq(&conf->device_lock);
			generic_make_request(bio);
			bio = NULL;

			r1_bio = mempool_alloc(conf->r1bio_pool, GFP_NOIO);

			r1_bio->master_bio = mbio;
			r1_bio->sectors = (mbio->bi_size >> 9)
					  - sectors_handled;
			r1_bio->state = 0;
			set_bit(R1BIO_ReadError, &r1_bio->state);
			r1_bio->mddev = mddev;
			r1_bio->sector = mbio->bi_sector + sectors_handled;

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

L
Linus Torvalds 已提交
2006 2007 2008 2009
static void raid1d(mddev_t *mddev)
{
	r1bio_t *r1_bio;
	unsigned long flags;
2010
	conf_t *conf = mddev->private;
L
Linus Torvalds 已提交
2011
	struct list_head *head = &conf->retry_list;
2012
	struct blk_plug plug;
L
Linus Torvalds 已提交
2013 2014

	md_check_recovery(mddev);
2015 2016

	blk_start_plug(&plug);
L
Linus Torvalds 已提交
2017
	for (;;) {
2018

N
NeilBrown 已提交
2019 2020
		if (atomic_read(&mddev->plug_cnt) == 0)
			flush_pending_writes(conf);
2021

2022 2023 2024
		spin_lock_irqsave(&conf->device_lock, flags);
		if (list_empty(head)) {
			spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
2025
			break;
2026
		}
L
Linus Torvalds 已提交
2027 2028
		r1_bio = list_entry(head->prev, r1bio_t, retry_list);
		list_del(head->prev);
2029
		conf->nr_queued--;
L
Linus Torvalds 已提交
2030 2031 2032
		spin_unlock_irqrestore(&conf->device_lock, flags);

		mddev = r1_bio->mddev;
2033
		conf = mddev->private;
2034
		if (test_bit(R1BIO_IsSync, &r1_bio->state)) {
2035
			if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
2036 2037 2038
			    test_bit(R1BIO_WriteError, &r1_bio->state))
				handle_sync_write_finished(conf, r1_bio);
			else
2039
				sync_request_write(mddev, r1_bio);
2040
		} else if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
2041 2042 2043 2044 2045
			   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
2046 2047 2048 2049
			/* just a partial read to be scheduled from separate
			 * context
			 */
			generic_make_request(r1_bio->bios[r1_bio->read_disk]);
2050

N
NeilBrown 已提交
2051
		cond_resched();
2052 2053
		if (mddev->flags & ~(1<<MD_CHANGE_PENDING))
			md_check_recovery(mddev);
L
Linus Torvalds 已提交
2054
	}
2055
	blk_finish_plug(&plug);
L
Linus Torvalds 已提交
2056 2057 2058 2059 2060 2061 2062 2063
}


static int init_resync(conf_t *conf)
{
	int buffs;

	buffs = RESYNC_WINDOW / RESYNC_BLOCK_SIZE;
2064
	BUG_ON(conf->r1buf_pool);
L
Linus Torvalds 已提交
2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082
	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.
 */

2083
static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster)
L
Linus Torvalds 已提交
2084
{
2085
	conf_t *conf = mddev->private;
L
Linus Torvalds 已提交
2086 2087 2088
	r1bio_t *r1_bio;
	struct bio *bio;
	sector_t max_sector, nr_sectors;
2089
	int disk = -1;
L
Linus Torvalds 已提交
2090
	int i;
2091 2092
	int wonly = -1;
	int write_targets = 0, read_targets = 0;
N
NeilBrown 已提交
2093
	sector_t sync_blocks;
2094
	int still_degraded = 0;
2095 2096
	int good_sectors = RESYNC_SECTORS;
	int min_bad = 0; /* number of sectors that are bad in all devices */
L
Linus Torvalds 已提交
2097 2098 2099

	if (!conf->r1buf_pool)
		if (init_resync(conf))
2100
			return 0;
L
Linus Torvalds 已提交
2101

A
Andre Noll 已提交
2102
	max_sector = mddev->dev_sectors;
L
Linus Torvalds 已提交
2103
	if (sector_nr >= max_sector) {
2104 2105 2106 2107 2108
		/* 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
		 */
2109 2110
		if (mddev->curr_resync < max_sector) /* aborted */
			bitmap_end_sync(mddev->bitmap, mddev->curr_resync,
2111
						&sync_blocks, 1);
2112
		else /* completed sync */
2113
			conf->fullsync = 0;
2114 2115

		bitmap_close_sync(mddev->bitmap);
L
Linus Torvalds 已提交
2116 2117 2118 2119
		close_sync(conf);
		return 0;
	}

2120 2121
	if (mddev->bitmap == NULL &&
	    mddev->recovery_cp == MaxSector &&
2122
	    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
2123 2124 2125 2126
	    conf->fullsync == 0) {
		*skipped = 1;
		return max_sector - sector_nr;
	}
2127 2128 2129
	/* before building a request, check if we can skip these blocks..
	 * This call the bitmap_start_sync doesn't actually record anything
	 */
2130
	if (!bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, 1) &&
2131
	    !conf->fullsync && !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
2132 2133 2134 2135
		/* We can skip this block, and probably several more */
		*skipped = 1;
		return sync_blocks;
	}
L
Linus Torvalds 已提交
2136
	/*
2137 2138 2139
	 * If there is non-resync activity waiting for a turn,
	 * and resync is going fast enough,
	 * then let it though before starting on this new sync request.
L
Linus Torvalds 已提交
2140
	 */
2141
	if (!go_faster && conf->nr_waiting)
L
Linus Torvalds 已提交
2142
		msleep_interruptible(1000);
2143

N
NeilBrown 已提交
2144
	bitmap_cond_end_sync(mddev->bitmap, sector_nr);
2145
	r1_bio = mempool_alloc(conf->r1buf_pool, GFP_NOIO);
2146 2147 2148
	raise_barrier(conf);

	conf->next_resync = sector_nr;
L
Linus Torvalds 已提交
2149

2150
	rcu_read_lock();
L
Linus Torvalds 已提交
2151
	/*
2152 2153 2154 2155 2156 2157
	 * If we get a correctably read error during resync or recovery,
	 * we might want to read from a different device.  So we
	 * flag all drives that could conceivably be read from for READ,
	 * and any others (which will be non-In_sync devices) for WRITE.
	 * If a read fails, we try reading from something else for which READ
	 * is OK.
L
Linus Torvalds 已提交
2158 2159 2160 2161
	 */

	r1_bio->mddev = mddev;
	r1_bio->sector = sector_nr;
2162
	r1_bio->state = 0;
L
Linus Torvalds 已提交
2163 2164 2165
	set_bit(R1BIO_IsSync, &r1_bio->state);

	for (i=0; i < conf->raid_disks; i++) {
2166
		mdk_rdev_t *rdev;
L
Linus Torvalds 已提交
2167 2168 2169 2170
		bio = r1_bio->bios[i];

		/* take from bio_init */
		bio->bi_next = NULL;
2171
		bio->bi_flags &= ~(BIO_POOL_MASK-1);
L
Linus Torvalds 已提交
2172
		bio->bi_flags |= 1 << BIO_UPTODATE;
2173
		bio->bi_comp_cpu = -1;
2174
		bio->bi_rw = READ;
L
Linus Torvalds 已提交
2175 2176 2177 2178 2179 2180 2181
		bio->bi_vcnt = 0;
		bio->bi_idx = 0;
		bio->bi_phys_segments = 0;
		bio->bi_size = 0;
		bio->bi_end_io = NULL;
		bio->bi_private = NULL;

2182 2183
		rdev = rcu_dereference(conf->mirrors[i].rdev);
		if (rdev == NULL ||
2184
		    test_bit(Faulty, &rdev->flags)) {
2185
			still_degraded = 1;
2186
		} else if (!test_bit(In_sync, &rdev->flags)) {
L
Linus Torvalds 已提交
2187 2188 2189
			bio->bi_rw = WRITE;
			bio->bi_end_io = end_sync_write;
			write_targets ++;
2190 2191
		} else {
			/* may need to read from here */
2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216
			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++;
2217 2218
			}
		}
2219 2220 2221 2222 2223 2224
		if (bio->bi_end_io) {
			atomic_inc(&rdev->nr_pending);
			bio->bi_sector = sector_nr + rdev->data_offset;
			bio->bi_bdev = rdev->bdev;
			bio->bi_private = r1_bio;
		}
L
Linus Torvalds 已提交
2225
	}
2226 2227 2228 2229
	rcu_read_unlock();
	if (disk < 0)
		disk = wonly;
	r1_bio->read_disk = disk;
2230

2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266
	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;
		for (i = 0 ; i < conf->raid_disks ; i++)
			if (r1_bio->bios[i]->bi_end_io == end_sync_write) {
				mdk_rdev_t *rdev =
					rcu_dereference(conf->mirrors[i].rdev);
				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;
	}

2267 2268 2269 2270 2271
	if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery) && read_targets > 0)
		/* extra read targets are also write targets */
		write_targets += read_targets-1;

	if (write_targets == 0 || read_targets == 0) {
L
Linus Torvalds 已提交
2272 2273 2274
		/* There is nowhere to write, so all non-sync
		 * drives must be failed - so we are finished
		 */
2275 2276
		sector_t rv = max_sector - sector_nr;
		*skipped = 1;
L
Linus Torvalds 已提交
2277 2278 2279 2280
		put_buf(r1_bio);
		return rv;
	}

2281 2282
	if (max_sector > mddev->resync_max)
		max_sector = mddev->resync_max; /* Don't do IO beyond here */
2283 2284
	if (max_sector > sector_nr + good_sectors)
		max_sector = sector_nr + good_sectors;
L
Linus Torvalds 已提交
2285
	nr_sectors = 0;
2286
	sync_blocks = 0;
L
Linus Torvalds 已提交
2287 2288 2289 2290 2291 2292 2293
	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;
2294 2295
		if (sync_blocks == 0) {
			if (!bitmap_start_sync(mddev->bitmap, sector_nr,
2296 2297 2298
					       &sync_blocks, still_degraded) &&
			    !conf->fullsync &&
			    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
2299
				break;
2300
			BUG_ON(sync_blocks < (PAGE_SIZE>>9));
2301
			if ((len >> 9) > sync_blocks)
2302
				len = sync_blocks<<9;
2303
		}
2304

L
Linus Torvalds 已提交
2305 2306 2307
		for (i=0 ; i < conf->raid_disks; i++) {
			bio = r1_bio->bios[i];
			if (bio->bi_end_io) {
2308
				page = bio->bi_io_vec[bio->bi_vcnt].bv_page;
L
Linus Torvalds 已提交
2309 2310
				if (bio_add_page(bio, page, len, 0) == 0) {
					/* stop here */
2311
					bio->bi_io_vec[bio->bi_vcnt].bv_page = page;
L
Linus Torvalds 已提交
2312 2313 2314
					while (i > 0) {
						i--;
						bio = r1_bio->bios[i];
2315 2316
						if (bio->bi_end_io==NULL)
							continue;
L
Linus Torvalds 已提交
2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327
						/* remove last page from this bio */
						bio->bi_vcnt--;
						bio->bi_size -= len;
						bio->bi_flags &= ~(1<< BIO_SEG_VALID);
					}
					goto bio_full;
				}
			}
		}
		nr_sectors += len>>9;
		sector_nr += len>>9;
2328
		sync_blocks -= (len>>9);
L
Linus Torvalds 已提交
2329 2330 2331 2332
	} while (r1_bio->bios[disk]->bi_vcnt < RESYNC_PAGES);
 bio_full:
	r1_bio->sectors = nr_sectors;

2333 2334 2335 2336 2337 2338 2339 2340
	/* 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);
		for (i=0; i<conf->raid_disks; i++) {
			bio = r1_bio->bios[i];
			if (bio->bi_end_io == end_sync_read) {
2341
				md_sync_acct(bio->bi_bdev, nr_sectors);
2342 2343 2344 2345 2346 2347
				generic_make_request(bio);
			}
		}
	} else {
		atomic_set(&r1_bio->remaining, 1);
		bio = r1_bio->bios[r1_bio->read_disk];
2348
		md_sync_acct(bio->bi_bdev, nr_sectors);
2349
		generic_make_request(bio);
L
Linus Torvalds 已提交
2350

2351
	}
L
Linus Torvalds 已提交
2352 2353 2354
	return nr_sectors;
}

2355 2356 2357 2358 2359 2360 2361 2362
static sector_t raid1_size(mddev_t *mddev, sector_t sectors, int raid_disks)
{
	if (sectors)
		return sectors;

	return mddev->dev_sectors;
}

2363
static conf_t *setup_conf(mddev_t *mddev)
L
Linus Torvalds 已提交
2364 2365
{
	conf_t *conf;
2366
	int i;
L
Linus Torvalds 已提交
2367 2368
	mirror_info_t *disk;
	mdk_rdev_t *rdev;
2369
	int err = -ENOMEM;
L
Linus Torvalds 已提交
2370

2371
	conf = kzalloc(sizeof(conf_t), GFP_KERNEL);
L
Linus Torvalds 已提交
2372
	if (!conf)
2373
		goto abort;
L
Linus Torvalds 已提交
2374

2375
	conf->mirrors = kzalloc(sizeof(struct mirror_info)*mddev->raid_disks,
L
Linus Torvalds 已提交
2376 2377
				 GFP_KERNEL);
	if (!conf->mirrors)
2378
		goto abort;
L
Linus Torvalds 已提交
2379

2380 2381
	conf->tmppage = alloc_page(GFP_KERNEL);
	if (!conf->tmppage)
2382
		goto abort;
2383

2384
	conf->poolinfo = kzalloc(sizeof(*conf->poolinfo), GFP_KERNEL);
L
Linus Torvalds 已提交
2385
	if (!conf->poolinfo)
2386
		goto abort;
L
Linus Torvalds 已提交
2387 2388 2389 2390 2391
	conf->poolinfo->raid_disks = mddev->raid_disks;
	conf->r1bio_pool = mempool_create(NR_RAID1_BIOS, r1bio_pool_alloc,
					  r1bio_pool_free,
					  conf->poolinfo);
	if (!conf->r1bio_pool)
2392 2393
		goto abort;

2394
	conf->poolinfo->mddev = mddev;
L
Linus Torvalds 已提交
2395

2396
	spin_lock_init(&conf->device_lock);
2397
	list_for_each_entry(rdev, &mddev->disks, same_set) {
2398
		int disk_idx = rdev->raid_disk;
L
Linus Torvalds 已提交
2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412
		if (disk_idx >= mddev->raid_disks
		    || disk_idx < 0)
			continue;
		disk = conf->mirrors + disk_idx;

		disk->rdev = rdev;

		disk->head_position = 0;
	}
	conf->raid_disks = mddev->raid_disks;
	conf->mddev = mddev;
	INIT_LIST_HEAD(&conf->retry_list);

	spin_lock_init(&conf->resync_lock);
2413
	init_waitqueue_head(&conf->wait_barrier);
L
Linus Torvalds 已提交
2414

2415 2416
	bio_list_init(&conf->pending_bio_list);

2417
	conf->last_used = -1;
L
Linus Torvalds 已提交
2418 2419 2420 2421
	for (i = 0; i < conf->raid_disks; i++) {

		disk = conf->mirrors + i;

2422 2423
		if (!disk->rdev ||
		    !test_bit(In_sync, &disk->rdev->flags)) {
L
Linus Torvalds 已提交
2424
			disk->head_position = 0;
2425 2426
			if (disk->rdev)
				conf->fullsync = 1;
2427 2428 2429 2430 2431 2432
		} else if (conf->last_used < 0)
			/*
			 * The first working device is used as a
			 * starting point to read balancing.
			 */
			conf->last_used = i;
L
Linus Torvalds 已提交
2433
	}
2434 2435 2436

	err = -EIO;
	if (conf->last_used < 0) {
N
NeilBrown 已提交
2437
		printk(KERN_ERR "md/raid1:%s: no operational mirrors\n",
2438 2439 2440 2441 2442 2443 2444
		       mdname(mddev));
		goto abort;
	}
	err = -ENOMEM;
	conf->thread = md_register_thread(raid1d, mddev, NULL);
	if (!conf->thread) {
		printk(KERN_ERR
N
NeilBrown 已提交
2445
		       "md/raid1:%s: couldn't allocate thread\n",
2446 2447
		       mdname(mddev));
		goto abort;
2448
	}
L
Linus Torvalds 已提交
2449

2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470
	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);
}

static int run(mddev_t *mddev)
{
	conf_t *conf;
	int i;
	mdk_rdev_t *rdev;

	if (mddev->level != 1) {
N
NeilBrown 已提交
2471
		printk(KERN_ERR "md/raid1:%s: raid level not set to mirroring (%d)\n",
2472 2473 2474 2475
		       mdname(mddev), mddev->level);
		return -EIO;
	}
	if (mddev->reshape_position != MaxSector) {
N
NeilBrown 已提交
2476
		printk(KERN_ERR "md/raid1:%s: reshape_position set but not supported\n",
2477 2478 2479
		       mdname(mddev));
		return -EIO;
	}
L
Linus Torvalds 已提交
2480
	/*
2481 2482 2483
	 * copy the already verified devices into our private RAID1
	 * bookkeeping area. [whatever we allocate in run(),
	 * should be freed in stop()]
L
Linus Torvalds 已提交
2484
	 */
2485 2486 2487 2488
	if (mddev->private == NULL)
		conf = setup_conf(mddev);
	else
		conf = mddev->private;
L
Linus Torvalds 已提交
2489

2490 2491
	if (IS_ERR(conf))
		return PTR_ERR(conf);
L
Linus Torvalds 已提交
2492

2493
	list_for_each_entry(rdev, &mddev->disks, same_set) {
2494 2495
		if (!mddev->gendisk)
			continue;
2496 2497 2498
		disk_stack_limits(mddev->gendisk, rdev->bdev,
				  rdev->data_offset << 9);
		/* as we don't honour merge_bvec_fn, we must never risk
2499 2500
		 * violating it, so limit ->max_segments to 1 lying within
		 * a single page, as a one page request is never in violation.
2501
		 */
2502 2503 2504 2505 2506
		if (rdev->bdev->bd_disk->queue->merge_bvec_fn) {
			blk_queue_max_segments(mddev->queue, 1);
			blk_queue_segment_boundary(mddev->queue,
						   PAGE_CACHE_SIZE - 1);
		}
L
Linus Torvalds 已提交
2507
	}
2508

2509 2510 2511 2512 2513 2514 2515 2516 2517 2518
	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;

2519
	if (mddev->recovery_cp != MaxSector)
N
NeilBrown 已提交
2520
		printk(KERN_NOTICE "md/raid1:%s: not clean"
2521 2522
		       " -- starting background reconstruction\n",
		       mdname(mddev));
L
Linus Torvalds 已提交
2523
	printk(KERN_INFO 
N
NeilBrown 已提交
2524
		"md/raid1:%s: active with %d out of %d mirrors\n",
L
Linus Torvalds 已提交
2525 2526
		mdname(mddev), mddev->raid_disks - mddev->degraded, 
		mddev->raid_disks);
2527

L
Linus Torvalds 已提交
2528 2529 2530
	/*
	 * Ok, everything is just fine now
	 */
2531 2532 2533 2534
	mddev->thread = conf->thread;
	conf->thread = NULL;
	mddev->private = conf;

2535
	md_set_array_sectors(mddev, raid1_size(mddev, 0, 0));
L
Linus Torvalds 已提交
2536

2537 2538 2539 2540
	if (mddev->queue) {
		mddev->queue->backing_dev_info.congested_fn = raid1_congested;
		mddev->queue->backing_dev_info.congested_data = mddev;
	}
2541
	return md_integrity_register(mddev);
L
Linus Torvalds 已提交
2542 2543 2544 2545
}

static int stop(mddev_t *mddev)
{
2546
	conf_t *conf = mddev->private;
2547 2548 2549
	struct bitmap *bitmap = mddev->bitmap;

	/* wait for behind writes to complete */
2550
	if (bitmap && atomic_read(&bitmap->behind_writes) > 0) {
N
NeilBrown 已提交
2551 2552
		printk(KERN_INFO "md/raid1:%s: behind writes in progress - waiting to stop.\n",
		       mdname(mddev));
2553
		/* need to kick something here to make sure I/O goes? */
2554 2555
		wait_event(bitmap->behind_wait,
			   atomic_read(&bitmap->behind_writes) == 0);
2556
	}
L
Linus Torvalds 已提交
2557

2558 2559 2560
	raise_barrier(conf);
	lower_barrier(conf);

L
Linus Torvalds 已提交
2561 2562 2563 2564
	md_unregister_thread(mddev->thread);
	mddev->thread = NULL;
	if (conf->r1bio_pool)
		mempool_destroy(conf->r1bio_pool);
2565 2566
	kfree(conf->mirrors);
	kfree(conf->poolinfo);
L
Linus Torvalds 已提交
2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580
	kfree(conf);
	mddev->private = NULL;
	return 0;
}

static int raid1_resize(mddev_t *mddev, sector_t sectors)
{
	/* 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.
	 */
2581
	md_set_array_sectors(mddev, raid1_size(mddev, sectors, 0));
D
Dan Williams 已提交
2582 2583
	if (mddev->array_sectors > raid1_size(mddev, sectors, 0))
		return -EINVAL;
2584
	set_capacity(mddev->gendisk, mddev->array_sectors);
2585
	revalidate_disk(mddev->gendisk);
D
Dan Williams 已提交
2586
	if (sectors > mddev->dev_sectors &&
2587
	    mddev->recovery_cp > mddev->dev_sectors) {
A
Andre Noll 已提交
2588
		mddev->recovery_cp = mddev->dev_sectors;
L
Linus Torvalds 已提交
2589 2590
		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
	}
D
Dan Williams 已提交
2591
	mddev->dev_sectors = sectors;
2592
	mddev->resync_max_sectors = sectors;
L
Linus Torvalds 已提交
2593 2594 2595
	return 0;
}

2596
static int raid1_reshape(mddev_t *mddev)
L
Linus Torvalds 已提交
2597 2598 2599 2600 2601 2602 2603 2604
{
	/* 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.
2605 2606 2607
	 *
	 * At the same time, we "pack" the devices so that all the missing
	 * devices have the higher raid_disk numbers.
L
Linus Torvalds 已提交
2608 2609 2610 2611
	 */
	mempool_t *newpool, *oldpool;
	struct pool_info *newpoolinfo;
	mirror_info_t *newmirrors;
2612
	conf_t *conf = mddev->private;
2613
	int cnt, raid_disks;
2614
	unsigned long flags;
2615
	int d, d2, err;
L
Linus Torvalds 已提交
2616

2617
	/* Cannot change chunk_size, layout, or level */
2618
	if (mddev->chunk_sectors != mddev->new_chunk_sectors ||
2619 2620
	    mddev->layout != mddev->new_layout ||
	    mddev->level != mddev->new_level) {
2621
		mddev->new_chunk_sectors = mddev->chunk_sectors;
2622 2623 2624 2625 2626
		mddev->new_layout = mddev->layout;
		mddev->new_level = mddev->level;
		return -EINVAL;
	}

2627 2628 2629
	err = md_allow_write(mddev);
	if (err)
		return err;
2630

2631 2632
	raid_disks = mddev->raid_disks + mddev->delta_disks;

2633 2634 2635 2636 2637 2638
	if (raid_disks < conf->raid_disks) {
		cnt=0;
		for (d= 0; d < conf->raid_disks; d++)
			if (conf->mirrors[d].rdev)
				cnt++;
		if (cnt > raid_disks)
L
Linus Torvalds 已提交
2639
			return -EBUSY;
2640
	}
L
Linus Torvalds 已提交
2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653

	newpoolinfo = kmalloc(sizeof(*newpoolinfo), GFP_KERNEL);
	if (!newpoolinfo)
		return -ENOMEM;
	newpoolinfo->mddev = mddev;
	newpoolinfo->raid_disks = raid_disks;

	newpool = mempool_create(NR_RAID1_BIOS, r1bio_pool_alloc,
				 r1bio_pool_free, newpoolinfo);
	if (!newpool) {
		kfree(newpoolinfo);
		return -ENOMEM;
	}
2654
	newmirrors = kzalloc(sizeof(struct mirror_info) * raid_disks, GFP_KERNEL);
L
Linus Torvalds 已提交
2655 2656 2657 2658 2659 2660
	if (!newmirrors) {
		kfree(newpoolinfo);
		mempool_destroy(newpool);
		return -ENOMEM;
	}

2661
	raise_barrier(conf);
L
Linus Torvalds 已提交
2662 2663 2664 2665

	/* ok, everything is stopped */
	oldpool = conf->r1bio_pool;
	conf->r1bio_pool = newpool;
2666

2667 2668 2669
	for (d = d2 = 0; d < conf->raid_disks; d++) {
		mdk_rdev_t *rdev = conf->mirrors[d].rdev;
		if (rdev && rdev->raid_disk != d2) {
2670
			sysfs_unlink_rdev(mddev, rdev);
2671
			rdev->raid_disk = d2;
2672 2673
			sysfs_unlink_rdev(mddev, rdev);
			if (sysfs_link_rdev(mddev, rdev))
2674
				printk(KERN_WARNING
2675 2676
				       "md/raid1:%s: cannot register rd%d\n",
				       mdname(mddev), rdev->raid_disk);
2677
		}
2678 2679 2680
		if (rdev)
			newmirrors[d2++].rdev = rdev;
	}
L
Linus Torvalds 已提交
2681 2682 2683 2684 2685
	kfree(conf->mirrors);
	conf->mirrors = newmirrors;
	kfree(conf->poolinfo);
	conf->poolinfo = newpoolinfo;

2686
	spin_lock_irqsave(&conf->device_lock, flags);
L
Linus Torvalds 已提交
2687
	mddev->degraded += (raid_disks - conf->raid_disks);
2688
	spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
2689
	conf->raid_disks = mddev->raid_disks = raid_disks;
2690
	mddev->delta_disks = 0;
L
Linus Torvalds 已提交
2691

2692
	conf->last_used = 0; /* just make sure it is in-range */
2693
	lower_barrier(conf);
L
Linus Torvalds 已提交
2694 2695 2696 2697 2698 2699 2700 2701

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

	mempool_destroy(oldpool);
	return 0;
}

2702
static void raid1_quiesce(mddev_t *mddev, int state)
2703
{
2704
	conf_t *conf = mddev->private;
2705 2706

	switch(state) {
2707 2708 2709
	case 2: /* wake for suspend */
		wake_up(&conf->wait_barrier);
		break;
2710
	case 1:
2711
		raise_barrier(conf);
2712
		break;
2713
	case 0:
2714
		lower_barrier(conf);
2715 2716 2717 2718
		break;
	}
}

2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735
static void *raid1_takeover(mddev_t *mddev)
{
	/* raid1 can take over:
	 *  raid5 with 2 devices, any layout or chunk size
	 */
	if (mddev->level == 5 && mddev->raid_disks == 2) {
		conf_t *conf;
		mddev->new_level = 1;
		mddev->new_layout = 0;
		mddev->new_chunk_sectors = 0;
		conf = setup_conf(mddev);
		if (!IS_ERR(conf))
			conf->barrier = 1;
		return conf;
	}
	return ERR_PTR(-EINVAL);
}
L
Linus Torvalds 已提交
2736

2737
static struct mdk_personality raid1_personality =
L
Linus Torvalds 已提交
2738 2739
{
	.name		= "raid1",
2740
	.level		= 1,
L
Linus Torvalds 已提交
2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751
	.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,
2752
	.size		= raid1_size,
2753
	.check_reshape	= raid1_reshape,
2754
	.quiesce	= raid1_quiesce,
2755
	.takeover	= raid1_takeover,
L
Linus Torvalds 已提交
2756 2757 2758 2759
};

static int __init raid_init(void)
{
2760
	return register_md_personality(&raid1_personality);
L
Linus Torvalds 已提交
2761 2762 2763 2764
}

static void raid_exit(void)
{
2765
	unregister_md_personality(&raid1_personality);
L
Linus Torvalds 已提交
2766 2767 2768 2769 2770
}

module_init(raid_init);
module_exit(raid_exit);
MODULE_LICENSE("GPL");
2771
MODULE_DESCRIPTION("RAID1 (mirroring) personality for MD");
L
Linus Torvalds 已提交
2772
MODULE_ALIAS("md-personality-3"); /* RAID1 */
2773
MODULE_ALIAS("md-raid1");
2774
MODULE_ALIAS("md-level-1");