raid1.c 68.1 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
static void r1_bio_write_done(r1bio_t *r1_bio)
322 323 324 325 326 327
{
	if (atomic_dec_and_test(&r1_bio->remaining))
	{
		/* it really is the end of this request */
		if (test_bit(R1BIO_BehindIO, &r1_bio->state)) {
			/* free extra copy of the data pages */
328
			int i = r1_bio->behind_page_count;
329
			while (i--)
330 331 332
				safe_put_page(r1_bio->behind_bvecs[i].bv_page);
			kfree(r1_bio->behind_bvecs);
			r1_bio->behind_bvecs = NULL;
333 334 335 336 337
		}
		/* 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),
338
				test_bit(R1BIO_BehindIO, &r1_bio->state));
339
		md_write_end(r1_bio->mddev);
340 341 342 343
		if (test_bit(R1BIO_MadeGood, &r1_bio->state))
			reschedule_retry(r1_bio);
		else
			raid_end_bio_io(r1_bio);
344 345 346
	}
}

347
static void raid1_end_write_request(struct bio *bio, int error)
L
Linus Torvalds 已提交
348 349
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
350
	r1bio_t *r1_bio = bio->bi_private;
351
	int mirror, behind = test_bit(R1BIO_BehindIO, &r1_bio->state);
352
	conf_t *conf = r1_bio->mddev->private;
353
	struct bio *to_put = NULL;
L
Linus Torvalds 已提交
354 355 356 357 358 359


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

T
Tejun Heo 已提交
360 361 362 363 364 365 366 367 368
	/*
	 * 'one mirror IO has finished' event handler:
	 */
	r1_bio->bios[mirror] = NULL;
	to_put = bio;
	if (!uptodate) {
		md_error(r1_bio->mddev, conf->mirrors[mirror].rdev);
		/* an I/O failed, we can't clear the bitmap */
		set_bit(R1BIO_Degraded, &r1_bio->state);
369
	} else {
L
Linus Torvalds 已提交
370
		/*
T
Tejun Heo 已提交
371 372 373 374 375 376 377 378
		 * 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
379
		 */
380 381 382
		sector_t first_bad;
		int bad_sectors;

T
Tejun Heo 已提交
383 384
		set_bit(R1BIO_Uptodate, &r1_bio->state);

385 386 387 388 389 390 391 392 393
		/* 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
394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415
	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);
416
				call_bio_endio(r1_bio);
417 418 419
			}
		}
	}
420 421 422
	if (r1_bio->bios[mirror] == NULL)
		rdev_dec_pending(conf->mirrors[mirror].rdev,
				 conf->mddev);
T
Tejun Heo 已提交
423

L
Linus Torvalds 已提交
424 425 426 427
	/*
	 * Let's see if all mirrored write operations have finished
	 * already.
	 */
428
	r1_bio_write_done(r1_bio);
429

430 431
	if (to_put)
		bio_put(to_put);
L
Linus Torvalds 已提交
432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448
}


/*
 * 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.
 */
449
static int read_balance(conf_t *conf, r1bio_t *r1_bio, int *max_sectors)
L
Linus Torvalds 已提交
450
{
451
	const sector_t this_sector = r1_bio->sector;
452 453
	int sectors;
	int best_good_sectors;
454
	int start_disk;
N
NeilBrown 已提交
455
	int best_disk;
456
	int i;
N
NeilBrown 已提交
457
	sector_t best_dist;
458
	mdk_rdev_t *rdev;
459
	int choose_first;
L
Linus Torvalds 已提交
460 461 462

	rcu_read_lock();
	/*
463
	 * Check if we can balance. We can balance on the whole
L
Linus Torvalds 已提交
464 465 466 467
	 * device if no resync is going on, or below the resync window.
	 * We take the first readable disk when above the resync window.
	 */
 retry:
468
	sectors = r1_bio->sectors;
N
NeilBrown 已提交
469 470
	best_disk = -1;
	best_dist = MaxSector;
471 472
	best_good_sectors = 0;

L
Linus Torvalds 已提交
473 474
	if (conf->mddev->recovery_cp < MaxSector &&
	    (this_sector + sectors >= conf->next_resync)) {
475 476 477 478 479
		choose_first = 1;
		start_disk = 0;
	} else {
		choose_first = 0;
		start_disk = conf->last_used;
L
Linus Torvalds 已提交
480 481
	}

482
	for (i = 0 ; i < conf->raid_disks ; i++) {
N
NeilBrown 已提交
483
		sector_t dist;
484 485 486
		sector_t first_bad;
		int bad_sectors;

487 488 489 490 491 492 493
		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
494
		    || test_bit(Faulty, &rdev->flags))
495
			continue;
N
NeilBrown 已提交
496 497
		if (!test_bit(In_sync, &rdev->flags) &&
		    rdev->recovery_offset < this_sector + sectors)
L
Linus Torvalds 已提交
498
			continue;
N
NeilBrown 已提交
499 500 501 502 503 504 505 506 507 508
		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.
		 */
509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537
		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
538 539 540 541 542 543 544 545
		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
546 547
			break;
		}
N
NeilBrown 已提交
548 549 550
		if (dist < best_dist) {
			best_dist = dist;
			best_disk = disk;
L
Linus Torvalds 已提交
551
		}
552
	}
L
Linus Torvalds 已提交
553

N
NeilBrown 已提交
554 555
	if (best_disk >= 0) {
		rdev = rcu_dereference(conf->mirrors[best_disk].rdev);
556 557 558
		if (!rdev)
			goto retry;
		atomic_inc(&rdev->nr_pending);
N
NeilBrown 已提交
559
		if (test_bit(Faulty, &rdev->flags)) {
L
Linus Torvalds 已提交
560 561 562
			/* cannot risk returning a device that failed
			 * before we inc'ed nr_pending
			 */
563
			rdev_dec_pending(rdev, conf->mddev);
L
Linus Torvalds 已提交
564 565
			goto retry;
		}
566
		sectors = best_good_sectors;
567
		conf->next_seq_sect = this_sector + sectors;
N
NeilBrown 已提交
568
		conf->last_used = best_disk;
L
Linus Torvalds 已提交
569 570
	}
	rcu_read_unlock();
571
	*max_sectors = sectors;
L
Linus Torvalds 已提交
572

N
NeilBrown 已提交
573
	return best_disk;
L
Linus Torvalds 已提交
574 575
}

576
int md_raid1_congested(mddev_t *mddev, int bits)
577
{
578
	conf_t *conf = mddev->private;
579 580 581 582 583 584
	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)) {
585
			struct request_queue *q = bdev_get_queue(rdev->bdev);
586

587 588
			BUG_ON(!q);

589 590 591
			/* Note the '|| 1' - when read_balance prefers
			 * non-congested targets, it can be removed
			 */
592
			if ((bits & (1<<BDI_async_congested)) || 1)
593 594 595 596 597 598 599 600
				ret |= bdi_congested(&q->backing_dev_info, bits);
			else
				ret &= bdi_congested(&q->backing_dev_info, bits);
		}
	}
	rcu_read_unlock();
	return ret;
}
601
EXPORT_SYMBOL_GPL(md_raid1_congested);
602

603 604 605 606 607 608 609
static int raid1_congested(void *data, int bits)
{
	mddev_t *mddev = data;

	return mddev_congested(mddev, bits) ||
		md_raid1_congested(mddev, bits);
}
610

J
Jens Axboe 已提交
611
static void flush_pending_writes(conf_t *conf)
612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633
{
	/* 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
634 635
}

636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655
/* 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
656 657 658
 */
#define RESYNC_DEPTH 32

659
static void raise_barrier(conf_t *conf)
L
Linus Torvalds 已提交
660 661
{
	spin_lock_irq(&conf->resync_lock);
662 663 664

	/* Wait until no block IO is waiting */
	wait_event_lock_irq(conf->wait_barrier, !conf->nr_waiting,
N
NeilBrown 已提交
665
			    conf->resync_lock, );
666 667 668 669

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

N
NeilBrown 已提交
670
	/* Now wait for all pending IO to complete */
671 672
	wait_event_lock_irq(conf->wait_barrier,
			    !conf->nr_pending && conf->barrier < RESYNC_DEPTH,
N
NeilBrown 已提交
673
			    conf->resync_lock, );
674 675 676 677 678 679 680

	spin_unlock_irq(&conf->resync_lock);
}

static void lower_barrier(conf_t *conf)
{
	unsigned long flags;
681
	BUG_ON(conf->barrier <= 0);
682 683 684 685 686 687 688 689 690 691 692 693 694
	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
695
				    );
696
		conf->nr_waiting--;
L
Linus Torvalds 已提交
697
	}
698
	conf->nr_pending++;
L
Linus Torvalds 已提交
699 700 701
	spin_unlock_irq(&conf->resync_lock);
}

702 703 704 705 706 707 708 709 710
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);
}

711 712 713 714 715
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
716 717 718 719 720 721 722 723
	 * 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.
724 725 726 727 728
	 */
	spin_lock_irq(&conf->resync_lock);
	conf->barrier++;
	conf->nr_waiting++;
	wait_event_lock_irq(conf->wait_barrier,
729
			    conf->nr_pending == conf->nr_queued+1,
730
			    conf->resync_lock,
N
NeilBrown 已提交
731
			    flush_pending_writes(conf));
732 733 734 735 736 737 738 739 740 741 742 743
	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);
}

744

745 746
/* duplicate the data pages for behind I/O 
 */
747
static void alloc_behind_pages(struct bio *bio, r1bio_t *r1_bio)
748 749 750
{
	int i;
	struct bio_vec *bvec;
751
	struct bio_vec *bvecs = kzalloc(bio->bi_vcnt * sizeof(struct bio_vec),
752
					GFP_NOIO);
753
	if (unlikely(!bvecs))
754
		return;
755 756

	bio_for_each_segment(bvec, bio, i) {
757 758 759
		bvecs[i] = *bvec;
		bvecs[i].bv_page = alloc_page(GFP_NOIO);
		if (unlikely(!bvecs[i].bv_page))
760
			goto do_sync_io;
761 762 763
		memcpy(kmap(bvecs[i].bv_page) + bvec->bv_offset,
		       kmap(bvec->bv_page) + bvec->bv_offset, bvec->bv_len);
		kunmap(bvecs[i].bv_page);
764 765
		kunmap(bvec->bv_page);
	}
766
	r1_bio->behind_bvecs = bvecs;
767 768 769
	r1_bio->behind_page_count = bio->bi_vcnt;
	set_bit(R1BIO_BehindIO, &r1_bio->state);
	return;
770 771

do_sync_io:
772
	for (i = 0; i < bio->bi_vcnt; i++)
773 774 775
		if (bvecs[i].bv_page)
			put_page(bvecs[i].bv_page);
	kfree(bvecs);
776 777 778
	PRINTK("%dB behind alloc failed, doing sync I/O\n", bio->bi_size);
}

779
static int make_request(mddev_t *mddev, struct bio * bio)
L
Linus Torvalds 已提交
780
{
781
	conf_t *conf = mddev->private;
L
Linus Torvalds 已提交
782 783 784
	mirror_info_t *mirror;
	r1bio_t *r1_bio;
	struct bio *read_bio;
785
	int i, disks;
786
	struct bitmap *bitmap;
787
	unsigned long flags;
788
	const int rw = bio_data_dir(bio);
789
	const unsigned long do_sync = (bio->bi_rw & REQ_SYNC);
T
Tejun Heo 已提交
790
	const unsigned long do_flush_fua = (bio->bi_rw & (REQ_FLUSH | REQ_FUA));
791
	mdk_rdev_t *blocked_rdev;
N
NeilBrown 已提交
792
	int plugged;
793 794 795
	int first_clone;
	int sectors_handled;
	int max_sectors;
796

L
Linus Torvalds 已提交
797 798 799 800 801
	/*
	 * 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.
	 */
802

803 804
	md_write_start(mddev, bio); /* wait on superblock update early */

805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823
	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);
	}
824

825
	wait_barrier(conf);
L
Linus Torvalds 已提交
826

827 828
	bitmap = mddev->bitmap;

L
Linus Torvalds 已提交
829 830 831 832 833 834 835 836 837
	/*
	 * 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;
838
	r1_bio->state = 0;
L
Linus Torvalds 已提交
839 840 841
	r1_bio->mddev = mddev;
	r1_bio->sector = bio->bi_sector;

842 843 844 845 846 847 848 849 850 851
	/* 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);

852
	if (rw == READ) {
L
Linus Torvalds 已提交
853 854 855
		/*
		 * read balancing logic:
		 */
856 857 858 859
		int rdisk;

read_again:
		rdisk = read_balance(conf, r1_bio, &max_sectors);
L
Linus Torvalds 已提交
860 861 862 863 864 865 866 867

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

868 869 870 871 872 873 874 875 876
		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
877 878
		r1_bio->read_disk = rdisk;

879
		read_bio = bio_clone_mddev(bio, GFP_NOIO, mddev);
880 881
		md_trim_bio(read_bio, r1_bio->sector - bio->bi_sector,
			    max_sectors);
L
Linus Torvalds 已提交
882 883 884 885 886 887

		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;
888
		read_bio->bi_rw = READ | do_sync;
L
Linus Torvalds 已提交
889 890
		read_bio->bi_private = r1_bio;

891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921
		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
922 923 924 925 926 927
		return 0;
	}

	/*
	 * WRITE:
	 */
928
	/* first select target devices under rcu_lock and
L
Linus Torvalds 已提交
929 930
	 * inc refcount on their rdev.  Record them by setting
	 * bios[x] to bio
931 932 933 934 935 936
	 * 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
937
	 */
N
NeilBrown 已提交
938 939
	plugged = mddev_check_plugged(mddev);

L
Linus Torvalds 已提交
940
	disks = conf->raid_disks;
941 942
 retry_write:
	blocked_rdev = NULL;
L
Linus Torvalds 已提交
943
	rcu_read_lock();
944
	max_sectors = r1_bio->sectors;
L
Linus Torvalds 已提交
945
	for (i = 0;  i < disks; i++) {
946 947 948 949 950 951
		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;
		}
952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981
		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;
982
				rdev_dec_pending(rdev, mddev);
983 984 985 986 987 988 989 990 991 992 993
				/* 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;
994
			}
995 996 997 998 999 1000 1001
			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
1002 1003 1004
	}
	rcu_read_unlock();

1005 1006 1007 1008 1009 1010 1011
	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);
1012
		r1_bio->state = 0;
1013 1014 1015 1016 1017 1018
		allow_barrier(conf);
		md_wait_for_blocked_rdev(blocked_rdev, mddev);
		wait_barrier(conf);
		goto retry_write;
	}

1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029
	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);
1030
	}
1031
	sectors_handled = r1_bio->sector + max_sectors - bio->bi_sector;
1032

1033
	atomic_set(&r1_bio->remaining, 1);
1034
	atomic_set(&r1_bio->behind_remaining, 0);
1035

1036
	first_clone = 1;
L
Linus Torvalds 已提交
1037 1038 1039 1040 1041
	for (i = 0; i < disks; i++) {
		struct bio *mbio;
		if (!r1_bio->bios[i])
			continue;

1042
		mbio = bio_clone_mddev(bio, GFP_NOIO, mddev);
1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061
		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;
		}
1062
		if (r1_bio->behind_bvecs) {
1063 1064 1065 1066 1067 1068 1069
			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
1070
			 * know the original bi_idx, so we just free
1071 1072 1073
			 * them all
			 */
			__bio_for_each_segment(bvec, mbio, j, 0)
1074
				bvec->bv_page = r1_bio->behind_bvecs[j].bv_page;
1075 1076 1077 1078
			if (test_bit(WriteMostly, &conf->mirrors[i].rdev->flags))
				atomic_inc(&r1_bio->behind_remaining);
		}

1079 1080 1081 1082 1083 1084 1085 1086 1087
		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
1088
		atomic_inc(&r1_bio->remaining);
1089 1090 1091
		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
1092
	}
1093
	r1_bio_write_done(r1_bio);
L
Linus Torvalds 已提交
1094

1095
	/* In case raid1d snuck in to freeze_array */
1096 1097
	wake_up(&conf->wait_barrier);

1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110
	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
1111
	if (do_sync || !bitmap || !plugged)
1112
		md_wakeup_thread(mddev->thread);
1113

L
Linus Torvalds 已提交
1114 1115 1116 1117 1118
	return 0;
}

static void status(struct seq_file *seq, mddev_t *mddev)
{
1119
	conf_t *conf = mddev->private;
L
Linus Torvalds 已提交
1120 1121 1122
	int i;

	seq_printf(seq, " [%d/%d] [", conf->raid_disks,
1123
		   conf->raid_disks - mddev->degraded);
1124 1125 1126
	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
1127
		seq_printf(seq, "%s",
1128 1129 1130
			   rdev && test_bit(In_sync, &rdev->flags) ? "U" : "_");
	}
	rcu_read_unlock();
L
Linus Torvalds 已提交
1131 1132 1133 1134 1135 1136 1137
	seq_printf(seq, "]");
}


static void error(mddev_t *mddev, mdk_rdev_t *rdev)
{
	char b[BDEVNAME_SIZE];
1138
	conf_t *conf = mddev->private;
L
Linus Torvalds 已提交
1139 1140 1141 1142 1143 1144 1145

	/*
	 * 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
	 */
1146
	if (test_bit(In_sync, &rdev->flags)
1147
	    && (conf->raid_disks - mddev->degraded) == 1) {
L
Linus Torvalds 已提交
1148 1149
		/*
		 * Don't fail the drive, act as though we were just a
1150 1151 1152
		 * 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
1153
		 */
1154
		conf->recovery_disabled = mddev->recovery_disabled;
L
Linus Torvalds 已提交
1155
		return;
1156
	}
1157
	set_bit(Blocked, &rdev->flags);
1158 1159 1160
	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
1161
		mddev->degraded++;
1162
		set_bit(Faulty, &rdev->flags);
1163
		spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1164 1165 1166
		/*
		 * if recovery is running, make sure it aborts.
		 */
1167
		set_bit(MD_RECOVERY_INTR, &mddev->recovery);
1168 1169
	} else
		set_bit(Faulty, &rdev->flags);
1170
	set_bit(MD_CHANGE_DEVS, &mddev->flags);
1171 1172 1173
	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
1174 1175
	       mdname(mddev), bdevname(rdev->bdev, b),
	       mdname(mddev), conf->raid_disks - mddev->degraded);
L
Linus Torvalds 已提交
1176 1177 1178 1179 1180 1181
}

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

N
NeilBrown 已提交
1182
	printk(KERN_DEBUG "RAID1 conf printout:\n");
L
Linus Torvalds 已提交
1183
	if (!conf) {
N
NeilBrown 已提交
1184
		printk(KERN_DEBUG "(!conf)\n");
L
Linus Torvalds 已提交
1185 1186
		return;
	}
N
NeilBrown 已提交
1187
	printk(KERN_DEBUG " --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
L
Linus Torvalds 已提交
1188 1189
		conf->raid_disks);

1190
	rcu_read_lock();
L
Linus Torvalds 已提交
1191 1192
	for (i = 0; i < conf->raid_disks; i++) {
		char b[BDEVNAME_SIZE];
1193 1194
		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
		if (rdev)
N
NeilBrown 已提交
1195
			printk(KERN_DEBUG " disk %d, wo:%d, o:%d, dev:%s\n",
1196 1197 1198
			       i, !test_bit(In_sync, &rdev->flags),
			       !test_bit(Faulty, &rdev->flags),
			       bdevname(rdev->bdev,b));
L
Linus Torvalds 已提交
1199
	}
1200
	rcu_read_unlock();
L
Linus Torvalds 已提交
1201 1202 1203 1204
}

static void close_sync(conf_t *conf)
{
1205 1206
	wait_barrier(conf);
	allow_barrier(conf);
L
Linus Torvalds 已提交
1207 1208 1209 1210 1211 1212 1213 1214 1215

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

static int raid1_spare_active(mddev_t *mddev)
{
	int i;
	conf_t *conf = mddev->private;
1216 1217
	int count = 0;
	unsigned long flags;
L
Linus Torvalds 已提交
1218 1219 1220

	/*
	 * Find all failed disks within the RAID1 configuration 
1221 1222
	 * and mark them readable.
	 * Called under mddev lock, so rcu protection not needed.
L
Linus Torvalds 已提交
1223 1224
	 */
	for (i = 0; i < conf->raid_disks; i++) {
1225 1226 1227
		mdk_rdev_t *rdev = conf->mirrors[i].rdev;
		if (rdev
		    && !test_bit(Faulty, &rdev->flags)
1228
		    && !test_and_set_bit(In_sync, &rdev->flags)) {
1229
			count++;
1230
			sysfs_notify_dirent_safe(rdev->sysfs_state);
L
Linus Torvalds 已提交
1231 1232
		}
	}
1233 1234 1235
	spin_lock_irqsave(&conf->device_lock, flags);
	mddev->degraded -= count;
	spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1236 1237

	print_conf(conf);
1238
	return count;
L
Linus Torvalds 已提交
1239 1240 1241 1242 1243 1244
}


static int raid1_add_disk(mddev_t *mddev, mdk_rdev_t *rdev)
{
	conf_t *conf = mddev->private;
1245
	int err = -EEXIST;
1246
	int mirror = 0;
L
Linus Torvalds 已提交
1247
	mirror_info_t *p;
1248 1249
	int first = 0;
	int last = mddev->raid_disks - 1;
L
Linus Torvalds 已提交
1250

1251 1252 1253
	if (mddev->recovery_disabled == conf->recovery_disabled)
		return -EBUSY;

1254 1255 1256 1257
	if (rdev->raid_disk >= 0)
		first = last = rdev->raid_disk;

	for (mirror = first; mirror <= last; mirror++)
L
Linus Torvalds 已提交
1258 1259
		if ( !(p=conf->mirrors+mirror)->rdev) {

1260 1261
			disk_stack_limits(mddev->gendisk, rdev->bdev,
					  rdev->data_offset << 9);
1262 1263 1264 1265 1266
			/* 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
1267
			 */
1268 1269 1270 1271 1272
			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
1273 1274 1275

			p->head_position = 0;
			rdev->raid_disk = mirror;
1276
			err = 0;
1277 1278 1279 1280
			/* 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)
1281
				conf->fullsync = 1;
1282
			rcu_assign_pointer(p->rdev, rdev);
L
Linus Torvalds 已提交
1283 1284
			break;
		}
1285
	md_integrity_add_rdev(rdev, mddev);
L
Linus Torvalds 已提交
1286
	print_conf(conf);
1287
	return err;
L
Linus Torvalds 已提交
1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299
}

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) {
1300
		if (test_bit(In_sync, &rdev->flags) ||
L
Linus Torvalds 已提交
1301 1302 1303 1304
		    atomic_read(&rdev->nr_pending)) {
			err = -EBUSY;
			goto abort;
		}
N
NeilBrown 已提交
1305
		/* Only remove non-faulty devices if recovery
1306 1307 1308
		 * is not possible.
		 */
		if (!test_bit(Faulty, &rdev->flags) &&
1309
		    mddev->recovery_disabled != conf->recovery_disabled &&
1310 1311 1312 1313
		    mddev->degraded < conf->raid_disks) {
			err = -EBUSY;
			goto abort;
		}
L
Linus Torvalds 已提交
1314
		p->rdev = NULL;
1315
		synchronize_rcu();
L
Linus Torvalds 已提交
1316 1317 1318 1319
		if (atomic_read(&rdev->nr_pending)) {
			/* lost the race, try later */
			err = -EBUSY;
			p->rdev = rdev;
1320
			goto abort;
L
Linus Torvalds 已提交
1321
		}
1322
		err = md_integrity_register(mddev);
L
Linus Torvalds 已提交
1323 1324 1325 1326 1327 1328 1329 1330
	}
abort:

	print_conf(conf);
	return err;
}


1331
static void end_sync_read(struct bio *bio, int error)
L
Linus Torvalds 已提交
1332
{
1333
	r1bio_t *r1_bio = bio->bi_private;
1334
	int i;
L
Linus Torvalds 已提交
1335

1336 1337 1338 1339 1340
	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
1341 1342 1343 1344 1345
	/*
	 * 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
	 */
1346
	if (test_bit(BIO_UPTODATE, &bio->bi_flags))
L
Linus Torvalds 已提交
1347
		set_bit(R1BIO_Uptodate, &r1_bio->state);
1348 1349 1350

	if (atomic_dec_and_test(&r1_bio->remaining))
		reschedule_retry(r1_bio);
L
Linus Torvalds 已提交
1351 1352
}

1353
static void end_sync_write(struct bio *bio, int error)
L
Linus Torvalds 已提交
1354 1355
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
1356
	r1bio_t *r1_bio = bio->bi_private;
L
Linus Torvalds 已提交
1357
	mddev_t *mddev = r1_bio->mddev;
1358
	conf_t *conf = mddev->private;
L
Linus Torvalds 已提交
1359 1360
	int i;
	int mirror=0;
1361 1362
	sector_t first_bad;
	int bad_sectors;
L
Linus Torvalds 已提交
1363 1364 1365 1366 1367 1368

	for (i = 0; i < conf->raid_disks; i++)
		if (r1_bio->bios[i] == bio) {
			mirror = i;
			break;
		}
1369
	if (!uptodate) {
N
NeilBrown 已提交
1370
		sector_t sync_blocks = 0;
1371 1372 1373 1374
		sector_t s = r1_bio->sector;
		long sectors_to_go = r1_bio->sectors;
		/* make sure these bits doesn't get cleared. */
		do {
1375
			bitmap_end_sync(mddev->bitmap, s,
1376 1377 1378 1379
					&sync_blocks, 1);
			s += sync_blocks;
			sectors_to_go -= sync_blocks;
		} while (sectors_to_go > 0);
L
Linus Torvalds 已提交
1380
		md_error(mddev, conf->mirrors[mirror].rdev);
1381 1382 1383 1384 1385
	} else if (is_badblock(conf->mirrors[mirror].rdev,
			       r1_bio->sector,
			       r1_bio->sectors,
			       &first_bad, &bad_sectors))
		set_bit(R1BIO_MadeGood, &r1_bio->state);
1386

L
Linus Torvalds 已提交
1387 1388 1389
	update_head_pos(mirror, r1_bio);

	if (atomic_dec_and_test(&r1_bio->remaining)) {
1390 1391 1392 1393 1394 1395 1396
		int s = r1_bio->sectors;
		if (test_bit(R1BIO_MadeGood, &r1_bio->state))
			reschedule_retry(r1_bio);
		else {
			put_buf(r1_bio);
			md_done_sync(mddev, s, uptodate);
		}
L
Linus Torvalds 已提交
1397 1398 1399
	}
}

1400
static int fix_sync_read_error(r1bio_t *r1_bio)
L
Linus Torvalds 已提交
1401
{
1402 1403 1404 1405 1406 1407 1408
	/* 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.
1409 1410 1411
	 * 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.
1412 1413
	 */
	mddev_t *mddev = r1_bio->mddev;
1414
	conf_t *conf = mddev->private;
1415 1416 1417 1418 1419 1420 1421 1422 1423 1424
	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;
1425
		int start;
1426 1427 1428 1429 1430 1431 1432 1433 1434 1435

		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;
1436
				if (sync_page_io(rdev, sect, s<<9,
1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447
						 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);

1448
		if (!success) {
1449 1450 1451 1452 1453 1454 1455 1456 1457
			char b[BDEVNAME_SIZE];
			/* Cannot read from anywhere, array is toast */
			md_error(mddev, conf->mirrors[r1_bio->read_disk].rdev);
			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);
			md_done_sync(mddev, r1_bio->sectors, 0);
1458
			put_buf(r1_bio);
1459
			return 0;
1460
		}
1461 1462 1463 1464 1465 1466 1467 1468 1469 1470

		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;
1471
			if (sync_page_io(rdev, sect, s<<9,
1472 1473 1474 1475 1476
					 bio->bi_io_vec[idx].bv_page,
					 WRITE, false) == 0) {
				r1_bio->bios[d]->bi_end_io = NULL;
				rdev_dec_pending(rdev, mddev);
				md_error(mddev, rdev);
1477
			}
1478 1479 1480 1481 1482 1483 1484 1485 1486
		}
		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;
1487
			if (sync_page_io(rdev, sect, s<<9,
1488 1489 1490
					 bio->bi_io_vec[idx].bv_page,
					 READ, false) == 0)
				md_error(mddev, rdev);
1491 1492
			else
				atomic_add(s, &rdev->corrected_errors);
1493
		}
1494 1495 1496 1497
		sectors -= s;
		sect += s;
		idx ++;
	}
1498
	set_bit(R1BIO_Uptodate, &r1_bio->state);
1499
	set_bit(BIO_UPTODATE, &bio->bi_flags);
1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516
	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;

1517
	for (primary = 0; primary < conf->raid_disks; primary++)
1518 1519 1520 1521 1522 1523 1524
		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;
1525 1526 1527 1528 1529 1530
	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;
1531

1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543
		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;
1544
			}
1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579
		} 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);
1580
		}
1581
	}
1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597
	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;
1598 1599 1600 1601

	if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
		if (process_checks(r1_bio) < 0)
			return;
1602 1603 1604
	/*
	 * schedule writes
	 */
L
Linus Torvalds 已提交
1605 1606 1607
	atomic_set(&r1_bio->remaining, 1);
	for (i = 0; i < disks ; i++) {
		wbio = r1_bio->bios[i];
1608 1609 1610 1611
		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
1612 1613
			continue;

1614 1615
		wbio->bi_rw = WRITE;
		wbio->bi_end_io = end_sync_write;
L
Linus Torvalds 已提交
1616 1617
		atomic_inc(&r1_bio->remaining);
		md_sync_acct(conf->mirrors[i].rdev->bdev, wbio->bi_size >> 9);
1618

L
Linus Torvalds 已提交
1619 1620 1621 1622
		generic_make_request(wbio);
	}

	if (atomic_dec_and_test(&r1_bio->remaining)) {
1623
		/* if we're here, all write(s) have completed, so clean up */
L
Linus Torvalds 已提交
1624 1625 1626 1627 1628 1629 1630 1631 1632 1633
		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.
1634
 *	3.	Performs writes following reads for array synchronising.
L
Linus Torvalds 已提交
1635 1636
 */

1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656
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....
			 */
1657 1658 1659
			sector_t first_bad;
			int bad_sectors;

1660 1661 1662
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
			    test_bit(In_sync, &rdev->flags) &&
1663 1664
			    is_badblock(rdev, sect, s,
					&first_bad, &bad_sectors) == 0 &&
J
Jonathan Brassow 已提交
1665 1666
			    sync_page_io(rdev, sect, s<<9,
					 conf->tmppage, READ, false))
1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688
				success = 1;
			else {
				d++;
				if (d == conf->raid_disks)
					d = 0;
			}
		} while (!success && d != read_disk);

		if (!success) {
			/* Cannot read from anywhere -- bye bye array */
			md_error(mddev, conf->mirrors[read_disk].rdev);
			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 &&
			    test_bit(In_sync, &rdev->flags)) {
J
Jonathan Brassow 已提交
1689 1690
				if (sync_page_io(rdev, sect, s<<9,
						 conf->tmppage, WRITE, false)
1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704
				    == 0)
					/* Well, this device is dead */
					md_error(mddev, rdev);
			}
		}
		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)) {
J
Jonathan Brassow 已提交
1705 1706
				if (sync_page_io(rdev, sect, s<<9,
						 conf->tmppage, READ, false)
1707 1708 1709 1710 1711 1712
				    == 0)
					/* Well, this device is dead */
					md_error(mddev, rdev);
				else {
					atomic_add(s, &rdev->corrected_errors);
					printk(KERN_INFO
N
NeilBrown 已提交
1713
					       "md/raid1:%s: read error corrected "
1714 1715
					       "(%d sectors at %llu on %s)\n",
					       mdname(mddev), s,
1716 1717
					       (unsigned long long)(sect +
					           rdev->data_offset),
1718 1719 1720 1721 1722 1723 1724 1725 1726
					       bdevname(rdev->bdev, b));
				}
			}
		}
		sectors -= s;
		sect += s;
	}
}

L
Linus Torvalds 已提交
1727 1728 1729 1730 1731
static void raid1d(mddev_t *mddev)
{
	r1bio_t *r1_bio;
	struct bio *bio;
	unsigned long flags;
1732
	conf_t *conf = mddev->private;
L
Linus Torvalds 已提交
1733 1734
	struct list_head *head = &conf->retry_list;
	mdk_rdev_t *rdev;
1735
	struct blk_plug plug;
L
Linus Torvalds 已提交
1736 1737

	md_check_recovery(mddev);
1738 1739

	blk_start_plug(&plug);
L
Linus Torvalds 已提交
1740 1741
	for (;;) {
		char b[BDEVNAME_SIZE];
1742

N
NeilBrown 已提交
1743 1744
		if (atomic_read(&mddev->plug_cnt) == 0)
			flush_pending_writes(conf);
1745

1746 1747 1748
		spin_lock_irqsave(&conf->device_lock, flags);
		if (list_empty(head)) {
			spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1749
			break;
1750
		}
L
Linus Torvalds 已提交
1751 1752
		r1_bio = list_entry(head->prev, r1bio_t, retry_list);
		list_del(head->prev);
1753
		conf->nr_queued--;
L
Linus Torvalds 已提交
1754 1755 1756
		spin_unlock_irqrestore(&conf->device_lock, flags);

		mddev = r1_bio->mddev;
1757
		conf = mddev->private;
1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790
		if (test_bit(R1BIO_IsSync, &r1_bio->state)) {
			if (test_bit(R1BIO_MadeGood, &r1_bio->state)) {
				int m;
				int s = r1_bio->sectors;
				for (m = 0; m < conf->raid_disks ; m++) {
					struct bio *bio = r1_bio->bios[m];
					if (bio->bi_end_io != NULL &&
					    test_bit(BIO_UPTODATE,
						     &bio->bi_flags)) {
						rdev = conf->mirrors[m].rdev;
						rdev_clear_badblocks(
							rdev,
							r1_bio->sector,
							r1_bio->sectors);
					}
				}
				put_buf(r1_bio);
				md_done_sync(mddev, s, 1);
			} else
				sync_request_write(mddev, r1_bio);
		} else if (test_bit(R1BIO_MadeGood, &r1_bio->state)) {
			int m;
			for (m = 0; m < conf->raid_disks ; m++)
				if (r1_bio->bios[m] == IO_MADE_GOOD) {
					rdev = conf->mirrors[m].rdev;
					rdev_clear_badblocks(
						rdev,
						r1_bio->sector,
						r1_bio->sectors);
					rdev_dec_pending(rdev, mddev);
				}
			raid_end_bio_io(r1_bio);
		} else if (test_bit(R1BIO_ReadError, &r1_bio->state)) {
L
Linus Torvalds 已提交
1791
			int disk;
1792
			int max_sectors;
1793

1794
			clear_bit(R1BIO_ReadError, &r1_bio->state);
1795 1796 1797 1798 1799 1800 1801 1802
			/* 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
			 */
1803 1804 1805 1806 1807 1808
			if (mddev->ro == 0) {
				freeze_array(conf);
				fix_read_error(conf, r1_bio->read_disk,
					       r1_bio->sector,
					       r1_bio->sectors);
				unfreeze_array(conf);
1809 1810 1811
			} else
				md_error(mddev,
					 conf->mirrors[r1_bio->read_disk].rdev);
1812

L
Linus Torvalds 已提交
1813
			bio = r1_bio->bios[r1_bio->read_disk];
1814 1815 1816 1817
			bdevname(bio->bi_bdev, b);
read_more:
			disk = read_balance(conf, r1_bio, &max_sectors);
			if (disk == -1) {
N
NeilBrown 已提交
1818
				printk(KERN_ALERT "md/raid1:%s: %s: unrecoverable I/O"
L
Linus Torvalds 已提交
1819
				       " read error for block %llu\n",
1820
				       mdname(mddev), b,
L
Linus Torvalds 已提交
1821 1822 1823
				       (unsigned long long)r1_bio->sector);
				raid_end_bio_io(r1_bio);
			} else {
1824
				const unsigned long do_sync = r1_bio->master_bio->bi_rw & REQ_SYNC;
1825 1826 1827 1828 1829
				if (bio) {
					r1_bio->bios[r1_bio->read_disk] =
						mddev->ro ? IO_BLOCKED : NULL;
					bio_put(bio);
				}
L
Linus Torvalds 已提交
1830
				r1_bio->read_disk = disk;
1831 1832
				bio = bio_clone_mddev(r1_bio->master_bio,
						      GFP_NOIO, mddev);
1833 1834 1835
				md_trim_bio(bio,
					    r1_bio->sector - bio->bi_sector,
					    max_sectors);
L
Linus Torvalds 已提交
1836 1837
				r1_bio->bios[r1_bio->read_disk] = bio;
				rdev = conf->mirrors[disk].rdev;
1838 1839 1840 1841 1842 1843 1844
				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));
L
Linus Torvalds 已提交
1845 1846 1847
				bio->bi_sector = r1_bio->sector + rdev->data_offset;
				bio->bi_bdev = rdev->bdev;
				bio->bi_end_io = raid1_end_read_request;
1848
				bio->bi_rw = READ | do_sync;
L
Linus Torvalds 已提交
1849
				bio->bi_private = r1_bio;
1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881
				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
1882
			}
1883 1884 1885 1886 1887
		} else {
			/* just a partial read to be scheduled from separate
			 * context
			 */
			generic_make_request(r1_bio->bios[r1_bio->read_disk]);
L
Linus Torvalds 已提交
1888
		}
N
NeilBrown 已提交
1889
		cond_resched();
1890 1891
		if (mddev->flags & ~(1<<MD_CHANGE_PENDING))
			md_check_recovery(mddev);
L
Linus Torvalds 已提交
1892
	}
1893
	blk_finish_plug(&plug);
L
Linus Torvalds 已提交
1894 1895 1896 1897 1898 1899 1900 1901
}


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

	buffs = RESYNC_WINDOW / RESYNC_BLOCK_SIZE;
1902
	BUG_ON(conf->r1buf_pool);
L
Linus Torvalds 已提交
1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920
	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.
 */

1921
static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster)
L
Linus Torvalds 已提交
1922
{
1923
	conf_t *conf = mddev->private;
L
Linus Torvalds 已提交
1924 1925 1926
	r1bio_t *r1_bio;
	struct bio *bio;
	sector_t max_sector, nr_sectors;
1927
	int disk = -1;
L
Linus Torvalds 已提交
1928
	int i;
1929 1930
	int wonly = -1;
	int write_targets = 0, read_targets = 0;
N
NeilBrown 已提交
1931
	sector_t sync_blocks;
1932
	int still_degraded = 0;
1933 1934
	int good_sectors = RESYNC_SECTORS;
	int min_bad = 0; /* number of sectors that are bad in all devices */
L
Linus Torvalds 已提交
1935 1936 1937

	if (!conf->r1buf_pool)
		if (init_resync(conf))
1938
			return 0;
L
Linus Torvalds 已提交
1939

A
Andre Noll 已提交
1940
	max_sector = mddev->dev_sectors;
L
Linus Torvalds 已提交
1941
	if (sector_nr >= max_sector) {
1942 1943 1944 1945 1946
		/* 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
		 */
1947 1948
		if (mddev->curr_resync < max_sector) /* aborted */
			bitmap_end_sync(mddev->bitmap, mddev->curr_resync,
1949
						&sync_blocks, 1);
1950
		else /* completed sync */
1951
			conf->fullsync = 0;
1952 1953

		bitmap_close_sync(mddev->bitmap);
L
Linus Torvalds 已提交
1954 1955 1956 1957
		close_sync(conf);
		return 0;
	}

1958 1959
	if (mddev->bitmap == NULL &&
	    mddev->recovery_cp == MaxSector &&
1960
	    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
1961 1962 1963 1964
	    conf->fullsync == 0) {
		*skipped = 1;
		return max_sector - sector_nr;
	}
1965 1966 1967
	/* before building a request, check if we can skip these blocks..
	 * This call the bitmap_start_sync doesn't actually record anything
	 */
1968
	if (!bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, 1) &&
1969
	    !conf->fullsync && !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
1970 1971 1972 1973
		/* We can skip this block, and probably several more */
		*skipped = 1;
		return sync_blocks;
	}
L
Linus Torvalds 已提交
1974
	/*
1975 1976 1977
	 * 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
1978
	 */
1979
	if (!go_faster && conf->nr_waiting)
L
Linus Torvalds 已提交
1980
		msleep_interruptible(1000);
1981

N
NeilBrown 已提交
1982
	bitmap_cond_end_sync(mddev->bitmap, sector_nr);
1983
	r1_bio = mempool_alloc(conf->r1buf_pool, GFP_NOIO);
1984 1985 1986
	raise_barrier(conf);

	conf->next_resync = sector_nr;
L
Linus Torvalds 已提交
1987

1988
	rcu_read_lock();
L
Linus Torvalds 已提交
1989
	/*
1990 1991 1992 1993 1994 1995
	 * 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
1996 1997 1998 1999
	 */

	r1_bio->mddev = mddev;
	r1_bio->sector = sector_nr;
2000
	r1_bio->state = 0;
L
Linus Torvalds 已提交
2001 2002 2003
	set_bit(R1BIO_IsSync, &r1_bio->state);

	for (i=0; i < conf->raid_disks; i++) {
2004
		mdk_rdev_t *rdev;
L
Linus Torvalds 已提交
2005 2006 2007 2008
		bio = r1_bio->bios[i];

		/* take from bio_init */
		bio->bi_next = NULL;
2009
		bio->bi_flags &= ~(BIO_POOL_MASK-1);
L
Linus Torvalds 已提交
2010
		bio->bi_flags |= 1 << BIO_UPTODATE;
2011
		bio->bi_comp_cpu = -1;
2012
		bio->bi_rw = READ;
L
Linus Torvalds 已提交
2013 2014 2015 2016 2017 2018 2019
		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;

2020 2021
		rdev = rcu_dereference(conf->mirrors[i].rdev);
		if (rdev == NULL ||
2022
		    test_bit(Faulty, &rdev->flags)) {
2023
			still_degraded = 1;
2024
		} else if (!test_bit(In_sync, &rdev->flags)) {
L
Linus Torvalds 已提交
2025 2026 2027
			bio->bi_rw = WRITE;
			bio->bi_end_io = end_sync_write;
			write_targets ++;
2028 2029
		} else {
			/* may need to read from here */
2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054
			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++;
2055 2056
			}
		}
2057 2058 2059 2060 2061 2062
		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
2063
	}
2064 2065 2066 2067
	rcu_read_unlock();
	if (disk < 0)
		disk = wonly;
	r1_bio->read_disk = disk;
2068

2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104
	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;
	}

2105 2106 2107 2108 2109
	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
2110 2111 2112
		/* There is nowhere to write, so all non-sync
		 * drives must be failed - so we are finished
		 */
2113 2114
		sector_t rv = max_sector - sector_nr;
		*skipped = 1;
L
Linus Torvalds 已提交
2115 2116 2117 2118
		put_buf(r1_bio);
		return rv;
	}

2119 2120
	if (max_sector > mddev->resync_max)
		max_sector = mddev->resync_max; /* Don't do IO beyond here */
2121 2122
	if (max_sector > sector_nr + good_sectors)
		max_sector = sector_nr + good_sectors;
L
Linus Torvalds 已提交
2123
	nr_sectors = 0;
2124
	sync_blocks = 0;
L
Linus Torvalds 已提交
2125 2126 2127 2128 2129 2130 2131
	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;
2132 2133
		if (sync_blocks == 0) {
			if (!bitmap_start_sync(mddev->bitmap, sector_nr,
2134 2135 2136
					       &sync_blocks, still_degraded) &&
			    !conf->fullsync &&
			    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
2137
				break;
2138
			BUG_ON(sync_blocks < (PAGE_SIZE>>9));
2139
			if ((len >> 9) > sync_blocks)
2140
				len = sync_blocks<<9;
2141
		}
2142

L
Linus Torvalds 已提交
2143 2144 2145
		for (i=0 ; i < conf->raid_disks; i++) {
			bio = r1_bio->bios[i];
			if (bio->bi_end_io) {
2146
				page = bio->bi_io_vec[bio->bi_vcnt].bv_page;
L
Linus Torvalds 已提交
2147 2148
				if (bio_add_page(bio, page, len, 0) == 0) {
					/* stop here */
2149
					bio->bi_io_vec[bio->bi_vcnt].bv_page = page;
L
Linus Torvalds 已提交
2150 2151 2152
					while (i > 0) {
						i--;
						bio = r1_bio->bios[i];
2153 2154
						if (bio->bi_end_io==NULL)
							continue;
L
Linus Torvalds 已提交
2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165
						/* 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;
2166
		sync_blocks -= (len>>9);
L
Linus Torvalds 已提交
2167 2168 2169 2170
	} while (r1_bio->bios[disk]->bi_vcnt < RESYNC_PAGES);
 bio_full:
	r1_bio->sectors = nr_sectors;

2171 2172 2173 2174 2175 2176 2177 2178
	/* 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) {
2179
				md_sync_acct(bio->bi_bdev, nr_sectors);
2180 2181 2182 2183 2184 2185
				generic_make_request(bio);
			}
		}
	} else {
		atomic_set(&r1_bio->remaining, 1);
		bio = r1_bio->bios[r1_bio->read_disk];
2186
		md_sync_acct(bio->bi_bdev, nr_sectors);
2187
		generic_make_request(bio);
L
Linus Torvalds 已提交
2188

2189
	}
L
Linus Torvalds 已提交
2190 2191 2192
	return nr_sectors;
}

2193 2194 2195 2196 2197 2198 2199 2200
static sector_t raid1_size(mddev_t *mddev, sector_t sectors, int raid_disks)
{
	if (sectors)
		return sectors;

	return mddev->dev_sectors;
}

2201
static conf_t *setup_conf(mddev_t *mddev)
L
Linus Torvalds 已提交
2202 2203
{
	conf_t *conf;
2204
	int i;
L
Linus Torvalds 已提交
2205 2206
	mirror_info_t *disk;
	mdk_rdev_t *rdev;
2207
	int err = -ENOMEM;
L
Linus Torvalds 已提交
2208

2209
	conf = kzalloc(sizeof(conf_t), GFP_KERNEL);
L
Linus Torvalds 已提交
2210
	if (!conf)
2211
		goto abort;
L
Linus Torvalds 已提交
2212

2213
	conf->mirrors = kzalloc(sizeof(struct mirror_info)*mddev->raid_disks,
L
Linus Torvalds 已提交
2214 2215
				 GFP_KERNEL);
	if (!conf->mirrors)
2216
		goto abort;
L
Linus Torvalds 已提交
2217

2218 2219
	conf->tmppage = alloc_page(GFP_KERNEL);
	if (!conf->tmppage)
2220
		goto abort;
2221

2222
	conf->poolinfo = kzalloc(sizeof(*conf->poolinfo), GFP_KERNEL);
L
Linus Torvalds 已提交
2223
	if (!conf->poolinfo)
2224
		goto abort;
L
Linus Torvalds 已提交
2225 2226 2227 2228 2229
	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)
2230 2231
		goto abort;

2232
	conf->poolinfo->mddev = mddev;
L
Linus Torvalds 已提交
2233

2234
	spin_lock_init(&conf->device_lock);
2235
	list_for_each_entry(rdev, &mddev->disks, same_set) {
2236
		int disk_idx = rdev->raid_disk;
L
Linus Torvalds 已提交
2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250
		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);
2251
	init_waitqueue_head(&conf->wait_barrier);
L
Linus Torvalds 已提交
2252

2253 2254
	bio_list_init(&conf->pending_bio_list);

2255
	conf->last_used = -1;
L
Linus Torvalds 已提交
2256 2257 2258 2259
	for (i = 0; i < conf->raid_disks; i++) {

		disk = conf->mirrors + i;

2260 2261
		if (!disk->rdev ||
		    !test_bit(In_sync, &disk->rdev->flags)) {
L
Linus Torvalds 已提交
2262
			disk->head_position = 0;
2263 2264
			if (disk->rdev)
				conf->fullsync = 1;
2265 2266 2267 2268 2269 2270
		} 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
2271
	}
2272 2273 2274

	err = -EIO;
	if (conf->last_used < 0) {
N
NeilBrown 已提交
2275
		printk(KERN_ERR "md/raid1:%s: no operational mirrors\n",
2276 2277 2278 2279 2280 2281 2282
		       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
2283
		       "md/raid1:%s: couldn't allocate thread\n",
2284 2285
		       mdname(mddev));
		goto abort;
2286
	}
L
Linus Torvalds 已提交
2287

2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308
	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
2309
		printk(KERN_ERR "md/raid1:%s: raid level not set to mirroring (%d)\n",
2310 2311 2312 2313
		       mdname(mddev), mddev->level);
		return -EIO;
	}
	if (mddev->reshape_position != MaxSector) {
N
NeilBrown 已提交
2314
		printk(KERN_ERR "md/raid1:%s: reshape_position set but not supported\n",
2315 2316 2317
		       mdname(mddev));
		return -EIO;
	}
L
Linus Torvalds 已提交
2318
	/*
2319 2320 2321
	 * 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
2322
	 */
2323 2324 2325 2326
	if (mddev->private == NULL)
		conf = setup_conf(mddev);
	else
		conf = mddev->private;
L
Linus Torvalds 已提交
2327

2328 2329
	if (IS_ERR(conf))
		return PTR_ERR(conf);
L
Linus Torvalds 已提交
2330

2331
	list_for_each_entry(rdev, &mddev->disks, same_set) {
2332 2333
		if (!mddev->gendisk)
			continue;
2334 2335 2336
		disk_stack_limits(mddev->gendisk, rdev->bdev,
				  rdev->data_offset << 9);
		/* as we don't honour merge_bvec_fn, we must never risk
2337 2338
		 * violating it, so limit ->max_segments to 1 lying within
		 * a single page, as a one page request is never in violation.
2339
		 */
2340 2341 2342 2343 2344
		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
2345
	}
2346

2347 2348 2349 2350 2351 2352 2353 2354 2355 2356
	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;

2357
	if (mddev->recovery_cp != MaxSector)
N
NeilBrown 已提交
2358
		printk(KERN_NOTICE "md/raid1:%s: not clean"
2359 2360
		       " -- starting background reconstruction\n",
		       mdname(mddev));
L
Linus Torvalds 已提交
2361
	printk(KERN_INFO 
N
NeilBrown 已提交
2362
		"md/raid1:%s: active with %d out of %d mirrors\n",
L
Linus Torvalds 已提交
2363 2364
		mdname(mddev), mddev->raid_disks - mddev->degraded, 
		mddev->raid_disks);
2365

L
Linus Torvalds 已提交
2366 2367 2368
	/*
	 * Ok, everything is just fine now
	 */
2369 2370 2371 2372
	mddev->thread = conf->thread;
	conf->thread = NULL;
	mddev->private = conf;

2373
	md_set_array_sectors(mddev, raid1_size(mddev, 0, 0));
L
Linus Torvalds 已提交
2374

2375 2376 2377 2378
	if (mddev->queue) {
		mddev->queue->backing_dev_info.congested_fn = raid1_congested;
		mddev->queue->backing_dev_info.congested_data = mddev;
	}
2379
	return md_integrity_register(mddev);
L
Linus Torvalds 已提交
2380 2381 2382 2383
}

static int stop(mddev_t *mddev)
{
2384
	conf_t *conf = mddev->private;
2385 2386 2387
	struct bitmap *bitmap = mddev->bitmap;

	/* wait for behind writes to complete */
2388
	if (bitmap && atomic_read(&bitmap->behind_writes) > 0) {
N
NeilBrown 已提交
2389 2390
		printk(KERN_INFO "md/raid1:%s: behind writes in progress - waiting to stop.\n",
		       mdname(mddev));
2391
		/* need to kick something here to make sure I/O goes? */
2392 2393
		wait_event(bitmap->behind_wait,
			   atomic_read(&bitmap->behind_writes) == 0);
2394
	}
L
Linus Torvalds 已提交
2395

2396 2397 2398
	raise_barrier(conf);
	lower_barrier(conf);

L
Linus Torvalds 已提交
2399 2400 2401 2402
	md_unregister_thread(mddev->thread);
	mddev->thread = NULL;
	if (conf->r1bio_pool)
		mempool_destroy(conf->r1bio_pool);
2403 2404
	kfree(conf->mirrors);
	kfree(conf->poolinfo);
L
Linus Torvalds 已提交
2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418
	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.
	 */
2419
	md_set_array_sectors(mddev, raid1_size(mddev, sectors, 0));
D
Dan Williams 已提交
2420 2421
	if (mddev->array_sectors > raid1_size(mddev, sectors, 0))
		return -EINVAL;
2422
	set_capacity(mddev->gendisk, mddev->array_sectors);
2423
	revalidate_disk(mddev->gendisk);
D
Dan Williams 已提交
2424
	if (sectors > mddev->dev_sectors &&
2425
	    mddev->recovery_cp > mddev->dev_sectors) {
A
Andre Noll 已提交
2426
		mddev->recovery_cp = mddev->dev_sectors;
L
Linus Torvalds 已提交
2427 2428
		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
	}
D
Dan Williams 已提交
2429
	mddev->dev_sectors = sectors;
2430
	mddev->resync_max_sectors = sectors;
L
Linus Torvalds 已提交
2431 2432 2433
	return 0;
}

2434
static int raid1_reshape(mddev_t *mddev)
L
Linus Torvalds 已提交
2435 2436 2437 2438 2439 2440 2441 2442
{
	/* 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.
2443 2444 2445
	 *
	 * 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
2446 2447 2448 2449
	 */
	mempool_t *newpool, *oldpool;
	struct pool_info *newpoolinfo;
	mirror_info_t *newmirrors;
2450
	conf_t *conf = mddev->private;
2451
	int cnt, raid_disks;
2452
	unsigned long flags;
2453
	int d, d2, err;
L
Linus Torvalds 已提交
2454

2455
	/* Cannot change chunk_size, layout, or level */
2456
	if (mddev->chunk_sectors != mddev->new_chunk_sectors ||
2457 2458
	    mddev->layout != mddev->new_layout ||
	    mddev->level != mddev->new_level) {
2459
		mddev->new_chunk_sectors = mddev->chunk_sectors;
2460 2461 2462 2463 2464
		mddev->new_layout = mddev->layout;
		mddev->new_level = mddev->level;
		return -EINVAL;
	}

2465 2466 2467
	err = md_allow_write(mddev);
	if (err)
		return err;
2468

2469 2470
	raid_disks = mddev->raid_disks + mddev->delta_disks;

2471 2472 2473 2474 2475 2476
	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
2477
			return -EBUSY;
2478
	}
L
Linus Torvalds 已提交
2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491

	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;
	}
2492
	newmirrors = kzalloc(sizeof(struct mirror_info) * raid_disks, GFP_KERNEL);
L
Linus Torvalds 已提交
2493 2494 2495 2496 2497 2498
	if (!newmirrors) {
		kfree(newpoolinfo);
		mempool_destroy(newpool);
		return -ENOMEM;
	}

2499
	raise_barrier(conf);
L
Linus Torvalds 已提交
2500 2501 2502 2503

	/* ok, everything is stopped */
	oldpool = conf->r1bio_pool;
	conf->r1bio_pool = newpool;
2504

2505 2506 2507
	for (d = d2 = 0; d < conf->raid_disks; d++) {
		mdk_rdev_t *rdev = conf->mirrors[d].rdev;
		if (rdev && rdev->raid_disk != d2) {
2508
			sysfs_unlink_rdev(mddev, rdev);
2509
			rdev->raid_disk = d2;
2510 2511
			sysfs_unlink_rdev(mddev, rdev);
			if (sysfs_link_rdev(mddev, rdev))
2512
				printk(KERN_WARNING
2513 2514
				       "md/raid1:%s: cannot register rd%d\n",
				       mdname(mddev), rdev->raid_disk);
2515
		}
2516 2517 2518
		if (rdev)
			newmirrors[d2++].rdev = rdev;
	}
L
Linus Torvalds 已提交
2519 2520 2521 2522 2523
	kfree(conf->mirrors);
	conf->mirrors = newmirrors;
	kfree(conf->poolinfo);
	conf->poolinfo = newpoolinfo;

2524
	spin_lock_irqsave(&conf->device_lock, flags);
L
Linus Torvalds 已提交
2525
	mddev->degraded += (raid_disks - conf->raid_disks);
2526
	spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
2527
	conf->raid_disks = mddev->raid_disks = raid_disks;
2528
	mddev->delta_disks = 0;
L
Linus Torvalds 已提交
2529

2530
	conf->last_used = 0; /* just make sure it is in-range */
2531
	lower_barrier(conf);
L
Linus Torvalds 已提交
2532 2533 2534 2535 2536 2537 2538 2539

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

	mempool_destroy(oldpool);
	return 0;
}

2540
static void raid1_quiesce(mddev_t *mddev, int state)
2541
{
2542
	conf_t *conf = mddev->private;
2543 2544

	switch(state) {
2545 2546 2547
	case 2: /* wake for suspend */
		wake_up(&conf->wait_barrier);
		break;
2548
	case 1:
2549
		raise_barrier(conf);
2550
		break;
2551
	case 0:
2552
		lower_barrier(conf);
2553 2554 2555 2556
		break;
	}
}

2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573
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
2574

2575
static struct mdk_personality raid1_personality =
L
Linus Torvalds 已提交
2576 2577
{
	.name		= "raid1",
2578
	.level		= 1,
L
Linus Torvalds 已提交
2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589
	.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,
2590
	.size		= raid1_size,
2591
	.check_reshape	= raid1_reshape,
2592
	.quiesce	= raid1_quiesce,
2593
	.takeover	= raid1_takeover,
L
Linus Torvalds 已提交
2594 2595 2596 2597
};

static int __init raid_init(void)
{
2598
	return register_md_personality(&raid1_personality);
L
Linus Torvalds 已提交
2599 2600 2601 2602
}

static void raid_exit(void)
{
2603
	unregister_md_personality(&raid1_personality);
L
Linus Torvalds 已提交
2604 2605 2606 2607 2608
}

module_init(raid_init);
module_exit(raid_exit);
MODULE_LICENSE("GPL");
2609
MODULE_DESCRIPTION("RAID1 (mirroring) personality for MD");
L
Linus Torvalds 已提交
2610
MODULE_ALIAS("md-personality-3"); /* RAID1 */
2611
MODULE_ALIAS("md-raid1");
2612
MODULE_ALIAS("md-level-1");