raid1.c 60.2 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/delay.h>
35 36
#include <linux/blkdev.h>
#include <linux/seq_file.h>
37
#include "md.h"
38
#include "dm-bio-list.h"
39 40
#include "raid1.h"
#include "bitmap.h"
41 42 43 44 45 46 47

#define DEBUG 0
#if DEBUG
#define PRINTK(x...) printk(x)
#else
#define PRINTK(x...)
#endif
L
Linus Torvalds 已提交
48 49 50 51 52 53 54 55 56

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


static void unplug_slaves(mddev_t *mddev);

57 58
static void allow_barrier(conf_t *conf);
static void lower_barrier(conf_t *conf);
L
Linus Torvalds 已提交
59

A
Al Viro 已提交
60
static void * r1bio_pool_alloc(gfp_t gfp_flags, void *data)
L
Linus Torvalds 已提交
61 62 63 64 65 66
{
	struct pool_info *pi = data;
	r1bio_t *r1_bio;
	int size = offsetof(r1bio_t, bios[pi->raid_disks]);

	/* allocate a r1bio with room for raid_disks entries in the bios array */
67 68
	r1_bio = kzalloc(size, gfp_flags);
	if (!r1_bio)
L
Linus Torvalds 已提交
69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84
		unplug_slaves(pi->mddev);

	return r1_bio;
}

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
85
static void * r1buf_pool_alloc(gfp_t gfp_flags, void *data)
L
Linus Torvalds 已提交
86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109
{
	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);
	if (!r1_bio) {
		unplug_slaves(pi->mddev);
		return NULL;
	}

	/*
	 * Allocate bios : 1 for reading, n-1 for writing
	 */
	for (j = pi->raid_disks ; j-- ; ) {
		bio = bio_alloc(gfp_flags, RESYNC_PAGES);
		if (!bio)
			goto out_free_bio;
		r1_bio->bios[j] = bio;
	}
	/*
	 * Allocate RESYNC_PAGES data pages and attach them to
110 111 112
	 * the first bio.
	 * If this is a user-requested check/repair, allocate
	 * RESYNC_PAGES for each bio.
L
Linus Torvalds 已提交
113
	 */
114 115 116 117 118 119 120 121 122 123 124 125
	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;
126
			bio->bi_vcnt = i+1;
127 128 129 130 131 132 133 134
		}
	}
	/* 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
135 136 137 138 139 140 141
	}

	r1_bio->master_bio = NULL;

	return r1_bio;

out_free_pages:
142 143 144
	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);
145
	j = -1;
L
Linus Torvalds 已提交
146 147 148 149 150 151 152 153 154 155
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;
156
	int i,j;
L
Linus Torvalds 已提交
157 158
	r1bio_t *r1bio = __r1_bio;

159 160 161 162 163
	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)
164
				safe_put_page(r1bio->bios[j]->bi_io_vec[i].bv_page);
165
		}
L
Linus Torvalds 已提交
166 167 168 169 170 171 172 173 174 175 176 177
	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;
178
		if (*bio && *bio != IO_BLOCKED)
L
Linus Torvalds 已提交
179 180 181 182 183
			bio_put(*bio);
		*bio = NULL;
	}
}

184
static void free_r1bio(r1bio_t *r1_bio)
L
Linus Torvalds 已提交
185 186 187 188 189 190 191
{
	conf_t *conf = mddev_to_conf(r1_bio->mddev);

	/*
	 * Wake up any possible resync thread that waits for the device
	 * to go idle.
	 */
192
	allow_barrier(conf);
L
Linus Torvalds 已提交
193 194 195 196 197

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

198
static void put_buf(r1bio_t *r1_bio)
L
Linus Torvalds 已提交
199 200
{
	conf_t *conf = mddev_to_conf(r1_bio->mddev);
201 202 203 204 205 206 207
	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
208 209 210

	mempool_free(r1_bio, conf->r1buf_pool);

211
	lower_barrier(conf);
L
Linus Torvalds 已提交
212 213 214 215 216 217 218 219 220 221
}

static void reschedule_retry(r1bio_t *r1_bio)
{
	unsigned long flags;
	mddev_t *mddev = r1_bio->mddev;
	conf_t *conf = mddev_to_conf(mddev);

	spin_lock_irqsave(&conf->device_lock, flags);
	list_add(&r1_bio->retry_list, &conf->retry_list);
222
	conf->nr_queued ++;
L
Linus Torvalds 已提交
223 224
	spin_unlock_irqrestore(&conf->device_lock, flags);

225
	wake_up(&conf->wait_barrier);
L
Linus Torvalds 已提交
226 227 228 229 230 231 232 233 234 235 236 237
	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.
 */
static void raid_end_bio_io(r1bio_t *r1_bio)
{
	struct bio *bio = r1_bio->master_bio;

238 239 240 241 242 243 244 245
	/* 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);

246
		bio_endio(bio,
247 248
			test_bit(R1BIO_Uptodate, &r1_bio->state) ? 0 : -EIO);
	}
L
Linus Torvalds 已提交
249 250 251 252 253 254 255 256 257 258 259 260 261 262
	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)
{
	conf_t *conf = mddev_to_conf(r1_bio->mddev);

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

263
static void raid1_end_read_request(struct bio *bio, int error)
L
Linus Torvalds 已提交
264 265 266 267 268 269 270 271 272 273
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
	r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private);
	int mirror;
	conf_t *conf = mddev_to_conf(r1_bio->mddev);

	mirror = r1_bio->read_disk;
	/*
	 * this branch is our 'one mirror IO has finished' event handler:
	 */
274 275
	update_head_pos(mirror, r1_bio);

276 277 278 279 280 281
	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
282
		 */
283 284 285 286 287 288 289 290
		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
291

292
	if (uptodate)
L
Linus Torvalds 已提交
293
		raid_end_bio_io(r1_bio);
294
	else {
L
Linus Torvalds 已提交
295 296 297 298 299 300 301 302 303 304 305 306 307
		/*
		 * oops, read error:
		 */
		char b[BDEVNAME_SIZE];
		if (printk_ratelimit())
			printk(KERN_ERR "raid1: %s: rescheduling sector %llu\n",
			       bdevname(conf->mirrors[mirror].rdev->bdev,b), (unsigned long long)r1_bio->sector);
		reschedule_retry(r1_bio);
	}

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

308
static void raid1_end_write_request(struct bio *bio, int error)
L
Linus Torvalds 已提交
309 310 311
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
	r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private);
312
	int mirror, behind = test_bit(R1BIO_BehindIO, &r1_bio->state);
L
Linus Torvalds 已提交
313
	conf_t *conf = mddev_to_conf(r1_bio->mddev);
314
	struct bio *to_put = NULL;
L
Linus Torvalds 已提交
315 316 317 318 319 320


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

321
	if (error == -EOPNOTSUPP && test_bit(R1BIO_Barrier, &r1_bio->state)) {
322 323 324
		set_bit(BarriersNotsupp, &conf->mirrors[mirror].rdev->flags);
		set_bit(R1BIO_BarrierRetry, &r1_bio->state);
		r1_bio->mddev->barriers_work = 0;
325
		/* Don't rdev_dec_pending in this branch - keep it for the retry */
326
	} else {
L
Linus Torvalds 已提交
327
		/*
328
		 * this branch is our 'one mirror IO has finished' event handler:
L
Linus Torvalds 已提交
329
		 */
330
		r1_bio->bios[mirror] = NULL;
331
		to_put = bio;
332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367
		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);
		} else
			/*
			 * 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.
			 */
			set_bit(R1BIO_Uptodate, &r1_bio->state);

		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);
368
					bio_endio(mbio, 0);
369
				}
370 371
			}
		}
372
		rdev_dec_pending(conf->mirrors[mirror].rdev, conf->mddev);
373
	}
L
Linus Torvalds 已提交
374 375 376 377 378 379
	/*
	 *
	 * Let's see if all mirrored write operations have finished
	 * already.
	 */
	if (atomic_dec_and_test(&r1_bio->remaining)) {
380
		if (test_bit(R1BIO_BarrierRetry, &r1_bio->state))
381
			reschedule_retry(r1_bio);
382 383 384 385 386 387 388 389 390 391 392 393 394 395 396
		else {
			/* 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 = bio->bi_vcnt;
				while (i--)
					safe_put_page(bio->bi_io_vec[i].bv_page);
			}
			/* 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),
					behind);
			md_write_end(r1_bio->mddev);
			raid_end_bio_io(r1_bio);
397
		}
L
Linus Torvalds 已提交
398
	}
399

400 401
	if (to_put)
		bio_put(to_put);
L
Linus Torvalds 已提交
402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422
}


/*
 * 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.
 */
static int read_balance(conf_t *conf, r1bio_t *r1_bio)
{
	const unsigned long this_sector = r1_bio->sector;
	int new_disk = conf->last_used, disk = new_disk;
423
	int wonly_disk = -1;
L
Linus Torvalds 已提交
424 425
	const int sectors = r1_bio->sectors;
	sector_t new_distance, current_distance;
426
	mdk_rdev_t *rdev;
L
Linus Torvalds 已提交
427 428 429

	rcu_read_lock();
	/*
430
	 * Check if we can balance. We can balance on the whole
L
Linus Torvalds 已提交
431 432 433 434 435 436 437 438 439
	 * device if no resync is going on, or below the resync window.
	 * We take the first readable disk when above the resync window.
	 */
 retry:
	if (conf->mddev->recovery_cp < MaxSector &&
	    (this_sector + sectors >= conf->next_resync)) {
		/* Choose the first operation device, for consistancy */
		new_disk = 0;

440
		for (rdev = rcu_dereference(conf->mirrors[new_disk].rdev);
441
		     r1_bio->bios[new_disk] == IO_BLOCKED ||
442
		     !rdev || !test_bit(In_sync, &rdev->flags)
443
			     || test_bit(WriteMostly, &rdev->flags);
444
		     rdev = rcu_dereference(conf->mirrors[++new_disk].rdev)) {
445

446 447
			if (rdev && test_bit(In_sync, &rdev->flags) &&
				r1_bio->bios[new_disk] != IO_BLOCKED)
448 449 450 451
				wonly_disk = new_disk;

			if (new_disk == conf->raid_disks - 1) {
				new_disk = wonly_disk;
L
Linus Torvalds 已提交
452 453 454 455 456 457 458 459
				break;
			}
		}
		goto rb_out;
	}


	/* make sure the disk is operational */
460
	for (rdev = rcu_dereference(conf->mirrors[new_disk].rdev);
461
	     r1_bio->bios[new_disk] == IO_BLOCKED ||
462
	     !rdev || !test_bit(In_sync, &rdev->flags) ||
463
		     test_bit(WriteMostly, &rdev->flags);
464
	     rdev = rcu_dereference(conf->mirrors[new_disk].rdev)) {
465

466 467
		if (rdev && test_bit(In_sync, &rdev->flags) &&
		    r1_bio->bios[new_disk] != IO_BLOCKED)
468 469
			wonly_disk = new_disk;

L
Linus Torvalds 已提交
470 471 472 473
		if (new_disk <= 0)
			new_disk = conf->raid_disks;
		new_disk--;
		if (new_disk == disk) {
474 475
			new_disk = wonly_disk;
			break;
L
Linus Torvalds 已提交
476 477
		}
	}
478 479 480 481

	if (new_disk < 0)
		goto rb_out;

L
Linus Torvalds 已提交
482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501
	disk = new_disk;
	/* now disk == new_disk == starting point for search */

	/*
	 * Don't change to another disk for sequential reads:
	 */
	if (conf->next_seq_sect == this_sector)
		goto rb_out;
	if (this_sector == conf->mirrors[new_disk].head_position)
		goto rb_out;

	current_distance = abs(this_sector - conf->mirrors[disk].head_position);

	/* Find the disk whose head is closest */

	do {
		if (disk <= 0)
			disk = conf->raid_disks;
		disk--;

502
		rdev = rcu_dereference(conf->mirrors[disk].rdev);
503

504
		if (!rdev || r1_bio->bios[disk] == IO_BLOCKED ||
505
		    !test_bit(In_sync, &rdev->flags) ||
506
		    test_bit(WriteMostly, &rdev->flags))
L
Linus Torvalds 已提交
507 508 509 510 511 512 513 514 515 516 517 518 519
			continue;

		if (!atomic_read(&rdev->nr_pending)) {
			new_disk = disk;
			break;
		}
		new_distance = abs(this_sector - conf->mirrors[disk].head_position);
		if (new_distance < current_distance) {
			current_distance = new_distance;
			new_disk = disk;
		}
	} while (disk != conf->last_used);

520
 rb_out:
L
Linus Torvalds 已提交
521 522 523


	if (new_disk >= 0) {
524
		rdev = rcu_dereference(conf->mirrors[new_disk].rdev);
525 526 527
		if (!rdev)
			goto retry;
		atomic_inc(&rdev->nr_pending);
528
		if (!test_bit(In_sync, &rdev->flags)) {
L
Linus Torvalds 已提交
529 530 531
			/* cannot risk returning a device that failed
			 * before we inc'ed nr_pending
			 */
532
			rdev_dec_pending(rdev, conf->mddev);
L
Linus Torvalds 已提交
533 534
			goto retry;
		}
535 536
		conf->next_seq_sect = this_sector + sectors;
		conf->last_used = new_disk;
L
Linus Torvalds 已提交
537 538 539 540 541 542 543 544 545 546 547 548 549
	}
	rcu_read_unlock();

	return new_disk;
}

static void unplug_slaves(mddev_t *mddev)
{
	conf_t *conf = mddev_to_conf(mddev);
	int i;

	rcu_read_lock();
	for (i=0; i<mddev->raid_disks; i++) {
550
		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
551
		if (rdev && !test_bit(Faulty, &rdev->flags) && atomic_read(&rdev->nr_pending)) {
552
			struct request_queue *r_queue = bdev_get_queue(rdev->bdev);
L
Linus Torvalds 已提交
553 554 555 556

			atomic_inc(&rdev->nr_pending);
			rcu_read_unlock();

557
			blk_unplug(r_queue);
L
Linus Torvalds 已提交
558 559 560 561 562 563 564 565

			rdev_dec_pending(rdev, mddev);
			rcu_read_lock();
		}
	}
	rcu_read_unlock();
}

566
static void raid1_unplug(struct request_queue *q)
L
Linus Torvalds 已提交
567
{
568 569 570 571
	mddev_t *mddev = q->queuedata;

	unplug_slaves(mddev);
	md_wakeup_thread(mddev->thread);
L
Linus Torvalds 已提交
572 573
}

574 575 576 577 578 579 580 581 582 583
static int raid1_congested(void *data, int bits)
{
	mddev_t *mddev = data;
	conf_t *conf = mddev_to_conf(mddev);
	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)) {
584
			struct request_queue *q = bdev_get_queue(rdev->bdev);
585 586 587 588 589 590 591 592 593 594 595 596 597 598 599

			/* Note the '|| 1' - when read_balance prefers
			 * non-congested targets, it can be removed
			 */
			if ((bits & (1<<BDI_write_congested)) || 1)
				ret |= bdi_congested(&q->backing_dev_info, bits);
			else
				ret &= bdi_congested(&q->backing_dev_info, bits);
		}
	}
	rcu_read_unlock();
	return ret;
}


600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630
static int flush_pending_writes(conf_t *conf)
{
	/* Any writes that have been queued but are awaiting
	 * bitmap updates get flushed here.
	 * We return 1 if any requests were actually submitted.
	 */
	int rv = 0;

	spin_lock_irq(&conf->device_lock);

	if (conf->pending_bio_list.head) {
		struct bio *bio;
		bio = bio_list_get(&conf->pending_bio_list);
		blk_remove_plug(conf->mddev->queue);
		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;
		}
		rv = 1;
	} else
		spin_unlock_irq(&conf->device_lock);
	return rv;
}

631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650
/* 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
651 652 653
 */
#define RESYNC_DEPTH 32

654
static void raise_barrier(conf_t *conf)
L
Linus Torvalds 已提交
655 656
{
	spin_lock_irq(&conf->resync_lock);
657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692

	/* Wait until no block IO is waiting */
	wait_event_lock_irq(conf->wait_barrier, !conf->nr_waiting,
			    conf->resync_lock,
			    raid1_unplug(conf->mddev->queue));

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

	/* No wait for all pending IO to complete */
	wait_event_lock_irq(conf->wait_barrier,
			    !conf->nr_pending && conf->barrier < RESYNC_DEPTH,
			    conf->resync_lock,
			    raid1_unplug(conf->mddev->queue));

	spin_unlock_irq(&conf->resync_lock);
}

static void lower_barrier(conf_t *conf)
{
	unsigned long flags;
	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,
				    raid1_unplug(conf->mddev->queue));
		conf->nr_waiting--;
L
Linus Torvalds 已提交
693
	}
694
	conf->nr_pending++;
L
Linus Torvalds 已提交
695 696 697
	spin_unlock_irq(&conf->resync_lock);
}

698 699 700 701 702 703 704 705 706
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);
}

707 708 709 710 711
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
712 713 714 715 716 717 718 719
	 * 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.
720 721 722 723 724
	 */
	spin_lock_irq(&conf->resync_lock);
	conf->barrier++;
	conf->nr_waiting++;
	wait_event_lock_irq(conf->wait_barrier,
725
			    conf->nr_pending == conf->nr_queued+1,
726
			    conf->resync_lock,
727 728
			    ({ flush_pending_writes(conf);
			       raid1_unplug(conf->mddev->queue); }));
729 730 731 732 733 734 735 736 737 738 739 740
	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);
}

741

742 743 744 745 746
/* duplicate the data pages for behind I/O */
static struct page **alloc_behind_pages(struct bio *bio)
{
	int i;
	struct bio_vec *bvec;
747
	struct page **pages = kzalloc(bio->bi_vcnt * sizeof(struct page *),
748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766
					GFP_NOIO);
	if (unlikely(!pages))
		goto do_sync_io;

	bio_for_each_segment(bvec, bio, i) {
		pages[i] = alloc_page(GFP_NOIO);
		if (unlikely(!pages[i]))
			goto do_sync_io;
		memcpy(kmap(pages[i]) + bvec->bv_offset,
			kmap(bvec->bv_page) + bvec->bv_offset, bvec->bv_len);
		kunmap(pages[i]);
		kunmap(bvec->bv_page);
	}

	return pages;

do_sync_io:
	if (pages)
		for (i = 0; i < bio->bi_vcnt && pages[i]; i++)
767
			put_page(pages[i]);
768 769 770 771 772
	kfree(pages);
	PRINTK("%dB behind alloc failed, doing sync I/O\n", bio->bi_size);
	return NULL;
}

773
static int make_request(struct request_queue *q, struct bio * bio)
L
Linus Torvalds 已提交
774 775 776 777 778 779
{
	mddev_t *mddev = q->queuedata;
	conf_t *conf = mddev_to_conf(mddev);
	mirror_info_t *mirror;
	r1bio_t *r1_bio;
	struct bio *read_bio;
780
	int i, targets = 0, disks;
781
	struct bitmap *bitmap;
782 783
	unsigned long flags;
	struct bio_list bl;
784
	struct page **behind_pages = NULL;
785
	const int rw = bio_data_dir(bio);
786
	const int do_sync = bio_sync(bio);
T
Tejun Heo 已提交
787
	int cpu, do_barriers;
788
	mdk_rdev_t *blocked_rdev;
789

L
Linus Torvalds 已提交
790 791 792 793
	/*
	 * 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.
794 795 796
	 * We test barriers_work *after* md_write_start as md_write_start
	 * may cause the first superblock write, and that will check out
	 * if barriers work.
L
Linus Torvalds 已提交
797
	 */
798

799 800
	md_write_start(mddev, bio); /* wait on superblock update early */

801 802 803
	if (unlikely(!mddev->barriers_work && bio_barrier(bio))) {
		if (rw == WRITE)
			md_write_end(mddev);
804
		bio_endio(bio, -EOPNOTSUPP);
805 806 807
		return 0;
	}

808
	wait_barrier(conf);
L
Linus Torvalds 已提交
809

810 811
	bitmap = mddev->bitmap;

T
Tejun Heo 已提交
812 813 814 815 816
	cpu = part_stat_lock();
	part_stat_inc(cpu, &mddev->gendisk->part0, ios[rw]);
	part_stat_add(cpu, &mddev->gendisk->part0, sectors[rw],
		      bio_sectors(bio));
	part_stat_unlock();
L
Linus Torvalds 已提交
817 818 819 820 821 822 823 824 825 826

	/*
	 * 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;
827
	r1_bio->state = 0;
L
Linus Torvalds 已提交
828 829 830
	r1_bio->mddev = mddev;
	r1_bio->sector = bio->bi_sector;

831
	if (rw == READ) {
L
Linus Torvalds 已提交
832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852
		/*
		 * read balancing logic:
		 */
		int rdisk = read_balance(conf, r1_bio);

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

		r1_bio->read_disk = rdisk;

		read_bio = bio_clone(bio, GFP_NOIO);

		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;
853
		read_bio->bi_rw = READ | do_sync;
L
Linus Torvalds 已提交
854 855 856 857 858 859 860 861 862 863 864 865 866 867
		read_bio->bi_private = r1_bio;

		generic_make_request(read_bio);
		return 0;
	}

	/*
	 * WRITE:
	 */
	/* first select target devices under spinlock and
	 * inc refcount on their rdev.  Record them by setting
	 * bios[x] to bio
	 */
	disks = conf->raid_disks;
868 869 870 871 872 873 874
#if 0
	{ static int first=1;
	if (first) printk("First Write sector %llu disks %d\n",
			  (unsigned long long)r1_bio->sector, disks);
	first = 0;
	}
#endif
875 876
 retry_write:
	blocked_rdev = NULL;
L
Linus Torvalds 已提交
877 878
	rcu_read_lock();
	for (i = 0;  i < disks; i++) {
879 880 881 882 883 884 885
		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;
		}
		if (rdev && !test_bit(Faulty, &rdev->flags)) {
L
Linus Torvalds 已提交
886
			atomic_inc(&rdev->nr_pending);
887
			if (test_bit(Faulty, &rdev->flags)) {
888
				rdev_dec_pending(rdev, mddev);
L
Linus Torvalds 已提交
889 890 891
				r1_bio->bios[i] = NULL;
			} else
				r1_bio->bios[i] = bio;
892
			targets++;
L
Linus Torvalds 已提交
893 894 895 896 897
		} else
			r1_bio->bios[i] = NULL;
	}
	rcu_read_unlock();

898 899 900 901 902 903 904 905 906 907 908 909 910 911
	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);

		allow_barrier(conf);
		md_wait_for_blocked_rdev(blocked_rdev, mddev);
		wait_barrier(conf);
		goto retry_write;
	}

912 913
	BUG_ON(targets == 0); /* we never fail the last device */

914 915 916 917 918 919
	if (targets < conf->raid_disks) {
		/* array is degraded, we will not clear the bitmap
		 * on I/O completion (see raid1_end_write_request) */
		set_bit(R1BIO_Degraded, &r1_bio->state);
	}

920 921 922 923 924 925
	/* do behind I/O ? */
	if (bitmap &&
	    atomic_read(&bitmap->behind_writes) < bitmap->max_write_behind &&
	    (behind_pages = alloc_behind_pages(bio)) != NULL)
		set_bit(R1BIO_BehindIO, &r1_bio->state);

926
	atomic_set(&r1_bio->remaining, 0);
927
	atomic_set(&r1_bio->behind_remaining, 0);
928

929
	do_barriers = bio_barrier(bio);
930 931 932
	if (do_barriers)
		set_bit(R1BIO_Barrier, &r1_bio->state);

933
	bio_list_init(&bl);
L
Linus Torvalds 已提交
934 935 936 937 938 939 940 941 942 943 944
	for (i = 0; i < disks; i++) {
		struct bio *mbio;
		if (!r1_bio->bios[i])
			continue;

		mbio = bio_clone(bio, GFP_NOIO);
		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;
945
		mbio->bi_rw = WRITE | do_barriers | do_sync;
L
Linus Torvalds 已提交
946 947
		mbio->bi_private = r1_bio;

948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964
		if (behind_pages) {
			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
			 * know the originial bi_idx, so we just free
			 * them all
			 */
			__bio_for_each_segment(bvec, mbio, j, 0)
				bvec->bv_page = behind_pages[j];
			if (test_bit(WriteMostly, &conf->mirrors[i].rdev->flags))
				atomic_inc(&r1_bio->behind_remaining);
		}

L
Linus Torvalds 已提交
965 966
		atomic_inc(&r1_bio->remaining);

967
		bio_list_add(&bl, mbio);
L
Linus Torvalds 已提交
968
	}
969
	kfree(behind_pages); /* the behind pages are attached to the bios now */
L
Linus Torvalds 已提交
970

971 972
	bitmap_startwrite(bitmap, bio->bi_sector, r1_bio->sectors,
				test_bit(R1BIO_BehindIO, &r1_bio->state));
973 974 975 976 977 978 979
	spin_lock_irqsave(&conf->device_lock, flags);
	bio_list_merge(&conf->pending_bio_list, &bl);
	bio_list_init(&bl);

	blk_plug_device(mddev->queue);
	spin_unlock_irqrestore(&conf->device_lock, flags);

980 981 982
	/* In case raid1d snuck into freeze_array */
	wake_up(&conf->wait_barrier);

983 984
	if (do_sync)
		md_wakeup_thread(mddev->thread);
985 986 987 988 989
#if 0
	while ((bio = bio_list_pop(&bl)) != NULL)
		generic_make_request(bio);
#endif

L
Linus Torvalds 已提交
990 991 992 993 994 995 996 997 998
	return 0;
}

static void status(struct seq_file *seq, mddev_t *mddev)
{
	conf_t *conf = mddev_to_conf(mddev);
	int i;

	seq_printf(seq, " [%d/%d] [", conf->raid_disks,
999
		   conf->raid_disks - mddev->degraded);
1000 1001 1002
	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
1003
		seq_printf(seq, "%s",
1004 1005 1006
			   rdev && test_bit(In_sync, &rdev->flags) ? "U" : "_");
	}
	rcu_read_unlock();
L
Linus Torvalds 已提交
1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021
	seq_printf(seq, "]");
}


static void error(mddev_t *mddev, mdk_rdev_t *rdev)
{
	char b[BDEVNAME_SIZE];
	conf_t *conf = mddev_to_conf(mddev);

	/*
	 * 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
	 */
1022
	if (test_bit(In_sync, &rdev->flags)
1023
	    && (conf->raid_disks - mddev->degraded) == 1) {
L
Linus Torvalds 已提交
1024 1025
		/*
		 * Don't fail the drive, act as though we were just a
1026 1027 1028
		 * 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
1029
		 */
1030
		mddev->recovery_disabled = 1;
L
Linus Torvalds 已提交
1031
		return;
1032
	}
1033 1034 1035
	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
1036
		mddev->degraded++;
1037
		set_bit(Faulty, &rdev->flags);
1038
		spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1039 1040 1041
		/*
		 * if recovery is running, make sure it aborts.
		 */
1042
		set_bit(MD_RECOVERY_INTR, &mddev->recovery);
1043 1044
	} else
		set_bit(Faulty, &rdev->flags);
1045
	set_bit(MD_CHANGE_DEVS, &mddev->flags);
1046 1047
	printk(KERN_ALERT "raid1: Disk failure on %s, disabling device.\n"
		"raid1: Operation continuing on %d devices.\n",
1048
		bdevname(rdev->bdev,b), conf->raid_disks - mddev->degraded);
L
Linus Torvalds 已提交
1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059
}

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

	printk("RAID1 conf printout:\n");
	if (!conf) {
		printk("(!conf)\n");
		return;
	}
1060
	printk(" --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
L
Linus Torvalds 已提交
1061 1062
		conf->raid_disks);

1063
	rcu_read_lock();
L
Linus Torvalds 已提交
1064 1065
	for (i = 0; i < conf->raid_disks; i++) {
		char b[BDEVNAME_SIZE];
1066 1067
		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
		if (rdev)
L
Linus Torvalds 已提交
1068
			printk(" disk %d, wo:%d, o:%d, dev:%s\n",
1069 1070 1071
			       i, !test_bit(In_sync, &rdev->flags),
			       !test_bit(Faulty, &rdev->flags),
			       bdevname(rdev->bdev,b));
L
Linus Torvalds 已提交
1072
	}
1073
	rcu_read_unlock();
L
Linus Torvalds 已提交
1074 1075 1076 1077
}

static void close_sync(conf_t *conf)
{
1078 1079
	wait_barrier(conf);
	allow_barrier(conf);
L
Linus Torvalds 已提交
1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091

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

static int raid1_spare_active(mddev_t *mddev)
{
	int i;
	conf_t *conf = mddev->private;

	/*
	 * Find all failed disks within the RAID1 configuration 
1092 1093
	 * and mark them readable.
	 * Called under mddev lock, so rcu protection not needed.
L
Linus Torvalds 已提交
1094 1095
	 */
	for (i = 0; i < conf->raid_disks; i++) {
1096 1097 1098
		mdk_rdev_t *rdev = conf->mirrors[i].rdev;
		if (rdev
		    && !test_bit(Faulty, &rdev->flags)
1099 1100 1101
		    && !test_and_set_bit(In_sync, &rdev->flags)) {
			unsigned long flags;
			spin_lock_irqsave(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1102
			mddev->degraded--;
1103
			spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114
		}
	}

	print_conf(conf);
	return 0;
}


static int raid1_add_disk(mddev_t *mddev, mdk_rdev_t *rdev)
{
	conf_t *conf = mddev->private;
1115
	int err = -EEXIST;
1116
	int mirror = 0;
L
Linus Torvalds 已提交
1117
	mirror_info_t *p;
1118 1119
	int first = 0;
	int last = mddev->raid_disks - 1;
L
Linus Torvalds 已提交
1120

1121 1122 1123 1124
	if (rdev->raid_disk >= 0)
		first = last = rdev->raid_disk;

	for (mirror = first; mirror <= last; mirror++)
L
Linus Torvalds 已提交
1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138
		if ( !(p=conf->mirrors+mirror)->rdev) {

			blk_queue_stack_limits(mddev->queue,
					       rdev->bdev->bd_disk->queue);
			/* as we don't honour merge_bvec_fn, we must never risk
			 * violating it, so limit ->max_sector to one PAGE, as
			 * a one page request is never in violation.
			 */
			if (rdev->bdev->bd_disk->queue->merge_bvec_fn &&
			    mddev->queue->max_sectors > (PAGE_SIZE>>9))
				blk_queue_max_sectors(mddev->queue, PAGE_SIZE>>9);

			p->head_position = 0;
			rdev->raid_disk = mirror;
1139
			err = 0;
1140 1141 1142 1143
			/* 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)
1144
				conf->fullsync = 1;
1145
			rcu_assign_pointer(p->rdev, rdev);
L
Linus Torvalds 已提交
1146 1147 1148 1149
			break;
		}

	print_conf(conf);
1150
	return err;
L
Linus Torvalds 已提交
1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162
}

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) {
1163
		if (test_bit(In_sync, &rdev->flags) ||
L
Linus Torvalds 已提交
1164 1165 1166 1167
		    atomic_read(&rdev->nr_pending)) {
			err = -EBUSY;
			goto abort;
		}
1168 1169 1170 1171 1172 1173 1174 1175
		/* Only remove non-faulty devices is recovery
		 * is not possible.
		 */
		if (!test_bit(Faulty, &rdev->flags) &&
		    mddev->degraded < conf->raid_disks) {
			err = -EBUSY;
			goto abort;
		}
L
Linus Torvalds 已提交
1176
		p->rdev = NULL;
1177
		synchronize_rcu();
L
Linus Torvalds 已提交
1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190
		if (atomic_read(&rdev->nr_pending)) {
			/* lost the race, try later */
			err = -EBUSY;
			p->rdev = rdev;
		}
	}
abort:

	print_conf(conf);
	return err;
}


1191
static void end_sync_read(struct bio *bio, int error)
L
Linus Torvalds 已提交
1192 1193
{
	r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private);
1194
	int i;
L
Linus Torvalds 已提交
1195

1196 1197 1198 1199 1200
	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
1201 1202 1203 1204 1205
	/*
	 * 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
	 */
1206
	if (test_bit(BIO_UPTODATE, &bio->bi_flags))
L
Linus Torvalds 已提交
1207
		set_bit(R1BIO_Uptodate, &r1_bio->state);
1208 1209 1210

	if (atomic_dec_and_test(&r1_bio->remaining))
		reschedule_retry(r1_bio);
L
Linus Torvalds 已提交
1211 1212
}

1213
static void end_sync_write(struct bio *bio, int error)
L
Linus Torvalds 已提交
1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
	r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private);
	mddev_t *mddev = r1_bio->mddev;
	conf_t *conf = mddev_to_conf(mddev);
	int i;
	int mirror=0;

	for (i = 0; i < conf->raid_disks; i++)
		if (r1_bio->bios[i] == bio) {
			mirror = i;
			break;
		}
1227 1228 1229 1230 1231 1232
	if (!uptodate) {
		int sync_blocks = 0;
		sector_t s = r1_bio->sector;
		long sectors_to_go = r1_bio->sectors;
		/* make sure these bits doesn't get cleared. */
		do {
1233
			bitmap_end_sync(mddev->bitmap, s,
1234 1235 1236 1237
					&sync_blocks, 1);
			s += sync_blocks;
			sectors_to_go -= sync_blocks;
		} while (sectors_to_go > 0);
L
Linus Torvalds 已提交
1238
		md_error(mddev, conf->mirrors[mirror].rdev);
1239
	}
1240

L
Linus Torvalds 已提交
1241 1242 1243
	update_head_pos(mirror, r1_bio);

	if (atomic_dec_and_test(&r1_bio->remaining)) {
1244
		sector_t s = r1_bio->sectors;
L
Linus Torvalds 已提交
1245
		put_buf(r1_bio);
1246
		md_done_sync(mddev, s, uptodate);
L
Linus Torvalds 已提交
1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258
	}
}

static void sync_request_write(mddev_t *mddev, r1bio_t *r1_bio)
{
	conf_t *conf = mddev_to_conf(mddev);
	int i;
	int disks = conf->raid_disks;
	struct bio *bio, *wbio;

	bio = r1_bio->bios[r1_bio->read_disk];

1259

1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281
	if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
		/* 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
		 */
		int primary;
		if (!test_bit(R1BIO_Uptodate, &r1_bio->state)) {
			for (i=0; i<mddev->raid_disks; i++)
				if (r1_bio->bios[i]->bi_end_io == end_sync_read)
					md_error(mddev, conf->mirrors[i].rdev);

			md_done_sync(mddev, r1_bio->sectors, 1);
			put_buf(r1_bio);
			return;
		}
		for (primary=0; primary<mddev->raid_disks; primary++)
			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;
1282
				rdev_dec_pending(conf->mirrors[primary].rdev, mddev);
1283 1284 1285 1286
				break;
			}
		r1_bio->read_disk = primary;
		for (i=0; i<mddev->raid_disks; i++)
1287
			if (r1_bio->bios[i]->bi_end_io == end_sync_read) {
1288 1289 1290 1291
				int j;
				int vcnt = r1_bio->sectors >> (PAGE_SHIFT- 9);
				struct bio *pbio = r1_bio->bios[primary];
				struct bio *sbio = r1_bio->bios[i];
1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304

				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;
					}
				} else
					j = 0;
1305 1306
				if (j >= 0)
					mddev->resync_mismatches += r1_bio->sectors;
1307 1308
				if (j < 0 || (test_bit(MD_RECOVERY_CHECK, &mddev->recovery)
					      && test_bit(BIO_UPTODATE, &sbio->bi_flags))) {
1309
					sbio->bi_end_io = NULL;
1310 1311
					rdev_dec_pending(conf->mirrors[i].rdev, mddev);
				} else {
1312
					/* fixup the bio for reuse */
1313
					int size;
1314 1315 1316 1317 1318 1319 1320 1321 1322 1323
					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;
1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334
					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),
1335 1336
						       page_address(pbio->bi_io_vec[j].bv_page),
						       PAGE_SIZE);
1337
					}
1338

1339 1340 1341
				}
			}
	}
L
Linus Torvalds 已提交
1342
	if (!test_bit(R1BIO_Uptodate, &r1_bio->state)) {
1343 1344 1345
		/* ouch - failed to read all of that.
		 * Try some synchronous reads of other devices to get
		 * good data, much like with normal read errors.  Only
1346
		 * read into the pages we already have so we don't
1347 1348 1349 1350
		 * 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.
L
Linus Torvalds 已提交
1351
		 */
1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365
		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;

			if (s > (PAGE_SIZE>>9))
				s = PAGE_SIZE >> 9;
			do {
				if (r1_bio->bios[d]->bi_end_io == end_sync_read) {
1366 1367 1368 1369
					/* No rcu protection needed here devices
					 * can only be removed when no resync is
					 * active, and resync is currently active
					 */
1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385
					rdev = conf->mirrors[d].rdev;
					if (sync_page_io(rdev->bdev,
							 sect + rdev->data_offset,
							 s<<9,
							 bio->bi_io_vec[idx].bv_page,
							 READ)) {
						success = 1;
						break;
					}
				}
				d++;
				if (d == conf->raid_disks)
					d = 0;
			} while (!success && d != r1_bio->read_disk);

			if (success) {
1386
				int start = d;
1387 1388 1389 1390 1391 1392 1393 1394 1395
				/* write it back and re-read */
				set_bit(R1BIO_Uptodate, &r1_bio->state);
				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;
1396
					atomic_add(s, &rdev->corrected_errors);
1397 1398 1399 1400
					if (sync_page_io(rdev->bdev,
							 sect + rdev->data_offset,
							 s<<9,
							 bio->bi_io_vec[idx].bv_page,
1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412
							 WRITE) == 0)
						md_error(mddev, rdev);
				}
				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;
					if (sync_page_io(rdev->bdev,
1413 1414 1415
							 sect + rdev->data_offset,
							 s<<9,
							 bio->bi_io_vec[idx].bv_page,
1416
							 READ) == 0)
1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434
						md_error(mddev, rdev);
				}
			} else {
				char b[BDEVNAME_SIZE];
				/* Cannot read from anywhere, array is toast */
				md_error(mddev, conf->mirrors[r1_bio->read_disk].rdev);
				printk(KERN_ALERT "raid1: %s: unrecoverable I/O read error"
				       " for block %llu\n",
				       bdevname(bio->bi_bdev,b),
				       (unsigned long long)r1_bio->sector);
				md_done_sync(mddev, r1_bio->sectors, 0);
				put_buf(r1_bio);
				return;
			}
			sectors -= s;
			sect += s;
			idx ++;
		}
L
Linus Torvalds 已提交
1435
	}
1436 1437 1438 1439

	/*
	 * schedule writes
	 */
L
Linus Torvalds 已提交
1440 1441 1442
	atomic_set(&r1_bio->remaining, 1);
	for (i = 0; i < disks ; i++) {
		wbio = r1_bio->bios[i];
1443 1444 1445 1446
		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
1447 1448
			continue;

1449 1450
		wbio->bi_rw = WRITE;
		wbio->bi_end_io = end_sync_write;
L
Linus Torvalds 已提交
1451 1452
		atomic_inc(&r1_bio->remaining);
		md_sync_acct(conf->mirrors[i].rdev->bdev, wbio->bi_size >> 9);
1453

L
Linus Torvalds 已提交
1454 1455 1456 1457
		generic_make_request(wbio);
	}

	if (atomic_dec_and_test(&r1_bio->remaining)) {
1458
		/* if we're here, all write(s) have completed, so clean up */
L
Linus Torvalds 已提交
1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471
		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.
 *	3.	Performs writes following reads for array syncronising.
 */

1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549
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....
			 */
			rdev = conf->mirrors[d].rdev;
			if (rdev &&
			    test_bit(In_sync, &rdev->flags) &&
			    sync_page_io(rdev->bdev,
					 sect + rdev->data_offset,
					 s<<9,
					 conf->tmppage, READ))
				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)) {
				if (sync_page_io(rdev->bdev,
						 sect + rdev->data_offset,
						 s<<9, conf->tmppage, WRITE)
				    == 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)) {
				if (sync_page_io(rdev->bdev,
						 sect + rdev->data_offset,
						 s<<9, conf->tmppage, READ)
				    == 0)
					/* Well, this device is dead */
					md_error(mddev, rdev);
				else {
					atomic_add(s, &rdev->corrected_errors);
					printk(KERN_INFO
					       "raid1:%s: read error corrected "
					       "(%d sectors at %llu on %s)\n",
					       mdname(mddev), s,
1550 1551
					       (unsigned long long)(sect +
					           rdev->data_offset),
1552 1553 1554 1555 1556 1557 1558 1559 1560
					       bdevname(rdev->bdev, b));
				}
			}
		}
		sectors -= s;
		sect += s;
	}
}

L
Linus Torvalds 已提交
1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574
static void raid1d(mddev_t *mddev)
{
	r1bio_t *r1_bio;
	struct bio *bio;
	unsigned long flags;
	conf_t *conf = mddev_to_conf(mddev);
	struct list_head *head = &conf->retry_list;
	int unplug=0;
	mdk_rdev_t *rdev;

	md_check_recovery(mddev);
	
	for (;;) {
		char b[BDEVNAME_SIZE];
1575

1576
		unplug += flush_pending_writes(conf);
1577

1578 1579 1580
		spin_lock_irqsave(&conf->device_lock, flags);
		if (list_empty(head)) {
			spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
1581
			break;
1582
		}
L
Linus Torvalds 已提交
1583 1584
		r1_bio = list_entry(head->prev, r1bio_t, retry_list);
		list_del(head->prev);
1585
		conf->nr_queued--;
L
Linus Torvalds 已提交
1586 1587 1588 1589 1590 1591 1592
		spin_unlock_irqrestore(&conf->device_lock, flags);

		mddev = r1_bio->mddev;
		conf = mddev_to_conf(mddev);
		if (test_bit(R1BIO_IsSync, &r1_bio->state)) {
			sync_request_write(mddev, r1_bio);
			unplug = 1;
1593 1594
		} else if (test_bit(R1BIO_BarrierRetry, &r1_bio->state)) {
			/* some requests in the r1bio were BIO_RW_BARRIER
1595
			 * requests which failed with -EOPNOTSUPP.  Hohumm..
1596 1597 1598
			 * Better resubmit without the barrier.
			 * We know which devices to resubmit for, because
			 * all others have had their bios[] entry cleared.
1599
			 * We already have a nr_pending reference on these rdevs.
1600 1601
			 */
			int i;
1602
			const int do_sync = bio_sync(r1_bio->master_bio);
1603 1604
			clear_bit(R1BIO_BarrierRetry, &r1_bio->state);
			clear_bit(R1BIO_Barrier, &r1_bio->state);
1605 1606 1607
			for (i=0; i < conf->raid_disks; i++)
				if (r1_bio->bios[i])
					atomic_inc(&r1_bio->remaining);
1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622
			for (i=0; i < conf->raid_disks; i++)
				if (r1_bio->bios[i]) {
					struct bio_vec *bvec;
					int j;

					bio = bio_clone(r1_bio->master_bio, GFP_NOIO);
					/* copy pages from the failed bio, as
					 * this might be a write-behind device */
					__bio_for_each_segment(bvec, bio, j, 0)
						bvec->bv_page = bio_iovec_idx(r1_bio->bios[i], j)->bv_page;
					bio_put(r1_bio->bios[i]);
					bio->bi_sector = r1_bio->sector +
						conf->mirrors[i].rdev->data_offset;
					bio->bi_bdev = conf->mirrors[i].rdev->bdev;
					bio->bi_end_io = raid1_end_write_request;
1623
					bio->bi_rw = WRITE | do_sync;
1624 1625 1626 1627
					bio->bi_private = r1_bio;
					r1_bio->bios[i] = bio;
					generic_make_request(bio);
				}
L
Linus Torvalds 已提交
1628 1629
		} else {
			int disk;
1630 1631 1632 1633 1634 1635 1636 1637 1638

			/* 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
			 */
1639 1640 1641 1642 1643 1644
			if (mddev->ro == 0) {
				freeze_array(conf);
				fix_read_error(conf, r1_bio->read_disk,
					       r1_bio->sector,
					       r1_bio->sectors);
				unfreeze_array(conf);
1645 1646
			}

L
Linus Torvalds 已提交
1647
			bio = r1_bio->bios[r1_bio->read_disk];
1648 1649
			if ((disk=read_balance(conf, r1_bio)) == -1 ||
			    disk == r1_bio->read_disk) {
L
Linus Torvalds 已提交
1650 1651 1652 1653 1654 1655
				printk(KERN_ALERT "raid1: %s: unrecoverable I/O"
				       " read error for block %llu\n",
				       bdevname(bio->bi_bdev,b),
				       (unsigned long long)r1_bio->sector);
				raid_end_bio_io(r1_bio);
			} else {
1656
				const int do_sync = bio_sync(r1_bio->master_bio);
1657 1658
				r1_bio->bios[r1_bio->read_disk] =
					mddev->ro ? IO_BLOCKED : NULL;
L
Linus Torvalds 已提交
1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671
				r1_bio->read_disk = disk;
				bio_put(bio);
				bio = bio_clone(r1_bio->master_bio, GFP_NOIO);
				r1_bio->bios[r1_bio->read_disk] = bio;
				rdev = conf->mirrors[disk].rdev;
				if (printk_ratelimit())
					printk(KERN_ERR "raid1: %s: redirecting sector %llu to"
					       " another mirror\n",
					       bdevname(rdev->bdev,b),
					       (unsigned long long)r1_bio->sector);
				bio->bi_sector = r1_bio->sector + rdev->data_offset;
				bio->bi_bdev = rdev->bdev;
				bio->bi_end_io = raid1_end_read_request;
1672
				bio->bi_rw = READ | do_sync;
L
Linus Torvalds 已提交
1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688
				bio->bi_private = r1_bio;
				unplug = 1;
				generic_make_request(bio);
			}
		}
	}
	if (unplug)
		unplug_slaves(mddev);
}


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

	buffs = RESYNC_WINDOW / RESYNC_BLOCK_SIZE;
1689
	BUG_ON(conf->r1buf_pool);
L
Linus Torvalds 已提交
1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707
	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.
 */

1708
static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster)
L
Linus Torvalds 已提交
1709 1710 1711 1712 1713
{
	conf_t *conf = mddev_to_conf(mddev);
	r1bio_t *r1_bio;
	struct bio *bio;
	sector_t max_sector, nr_sectors;
1714
	int disk = -1;
L
Linus Torvalds 已提交
1715
	int i;
1716 1717
	int wonly = -1;
	int write_targets = 0, read_targets = 0;
1718
	int sync_blocks;
1719
	int still_degraded = 0;
L
Linus Torvalds 已提交
1720 1721

	if (!conf->r1buf_pool)
1722 1723 1724 1725
	{
/*
		printk("sync start - bitmap %p\n", mddev->bitmap);
*/
L
Linus Torvalds 已提交
1726
		if (init_resync(conf))
1727
			return 0;
1728
	}
L
Linus Torvalds 已提交
1729

A
Andre Noll 已提交
1730
	max_sector = mddev->dev_sectors;
L
Linus Torvalds 已提交
1731
	if (sector_nr >= max_sector) {
1732 1733 1734 1735 1736
		/* 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
		 */
1737 1738
		if (mddev->curr_resync < max_sector) /* aborted */
			bitmap_end_sync(mddev->bitmap, mddev->curr_resync,
1739
						&sync_blocks, 1);
1740
		else /* completed sync */
1741
			conf->fullsync = 0;
1742 1743

		bitmap_close_sync(mddev->bitmap);
L
Linus Torvalds 已提交
1744 1745 1746 1747
		close_sync(conf);
		return 0;
	}

1748 1749
	if (mddev->bitmap == NULL &&
	    mddev->recovery_cp == MaxSector &&
1750
	    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
1751 1752 1753 1754
	    conf->fullsync == 0) {
		*skipped = 1;
		return max_sector - sector_nr;
	}
1755 1756 1757
	/* before building a request, check if we can skip these blocks..
	 * This call the bitmap_start_sync doesn't actually record anything
	 */
1758
	if (!bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, 1) &&
1759
	    !conf->fullsync && !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
1760 1761 1762 1763
		/* We can skip this block, and probably several more */
		*skipped = 1;
		return sync_blocks;
	}
L
Linus Torvalds 已提交
1764
	/*
1765 1766 1767
	 * 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
1768
	 */
1769
	if (!go_faster && conf->nr_waiting)
L
Linus Torvalds 已提交
1770
		msleep_interruptible(1000);
1771

N
NeilBrown 已提交
1772
	bitmap_cond_end_sync(mddev->bitmap, sector_nr);
1773 1774 1775
	raise_barrier(conf);

	conf->next_resync = sector_nr;
L
Linus Torvalds 已提交
1776

1777 1778
	r1_bio = mempool_alloc(conf->r1buf_pool, GFP_NOIO);
	rcu_read_lock();
L
Linus Torvalds 已提交
1779
	/*
1780 1781 1782 1783 1784 1785
	 * 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
1786 1787 1788 1789
	 */

	r1_bio->mddev = mddev;
	r1_bio->sector = sector_nr;
1790
	r1_bio->state = 0;
L
Linus Torvalds 已提交
1791 1792 1793
	set_bit(R1BIO_IsSync, &r1_bio->state);

	for (i=0; i < conf->raid_disks; i++) {
1794
		mdk_rdev_t *rdev;
L
Linus Torvalds 已提交
1795 1796 1797 1798 1799
		bio = r1_bio->bios[i];

		/* take from bio_init */
		bio->bi_next = NULL;
		bio->bi_flags |= 1 << BIO_UPTODATE;
1800
		bio->bi_rw = READ;
L
Linus Torvalds 已提交
1801 1802 1803 1804 1805 1806 1807
		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;

1808 1809 1810
		rdev = rcu_dereference(conf->mirrors[i].rdev);
		if (rdev == NULL ||
			   test_bit(Faulty, &rdev->flags)) {
1811 1812
			still_degraded = 1;
			continue;
1813
		} else if (!test_bit(In_sync, &rdev->flags)) {
L
Linus Torvalds 已提交
1814 1815 1816
			bio->bi_rw = WRITE;
			bio->bi_end_io = end_sync_write;
			write_targets ++;
1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832
		} else {
			/* may need to read from here */
			bio->bi_rw = READ;
			bio->bi_end_io = end_sync_read;
			if (test_bit(WriteMostly, &rdev->flags)) {
				if (wonly < 0)
					wonly = i;
			} else {
				if (disk < 0)
					disk = i;
			}
			read_targets++;
		}
		atomic_inc(&rdev->nr_pending);
		bio->bi_sector = sector_nr + rdev->data_offset;
		bio->bi_bdev = rdev->bdev;
L
Linus Torvalds 已提交
1833 1834
		bio->bi_private = r1_bio;
	}
1835 1836 1837 1838
	rcu_read_unlock();
	if (disk < 0)
		disk = wonly;
	r1_bio->read_disk = disk;
1839

1840 1841 1842 1843 1844
	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
1845 1846 1847
		/* There is nowhere to write, so all non-sync
		 * drives must be failed - so we are finished
		 */
1848 1849
		sector_t rv = max_sector - sector_nr;
		*skipped = 1;
L
Linus Torvalds 已提交
1850 1851 1852 1853
		put_buf(r1_bio);
		return rv;
	}

1854 1855
	if (max_sector > mddev->resync_max)
		max_sector = mddev->resync_max; /* Don't do IO beyond here */
L
Linus Torvalds 已提交
1856
	nr_sectors = 0;
1857
	sync_blocks = 0;
L
Linus Torvalds 已提交
1858 1859 1860 1861 1862 1863 1864
	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;
1865 1866
		if (sync_blocks == 0) {
			if (!bitmap_start_sync(mddev->bitmap, sector_nr,
1867 1868 1869
					       &sync_blocks, still_degraded) &&
			    !conf->fullsync &&
			    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
1870
				break;
1871
			BUG_ON(sync_blocks < (PAGE_SIZE>>9));
1872 1873
			if (len > (sync_blocks<<9))
				len = sync_blocks<<9;
1874
		}
1875

L
Linus Torvalds 已提交
1876 1877 1878
		for (i=0 ; i < conf->raid_disks; i++) {
			bio = r1_bio->bios[i];
			if (bio->bi_end_io) {
1879
				page = bio->bi_io_vec[bio->bi_vcnt].bv_page;
L
Linus Torvalds 已提交
1880 1881
				if (bio_add_page(bio, page, len, 0) == 0) {
					/* stop here */
1882
					bio->bi_io_vec[bio->bi_vcnt].bv_page = page;
L
Linus Torvalds 已提交
1883 1884 1885
					while (i > 0) {
						i--;
						bio = r1_bio->bios[i];
1886 1887
						if (bio->bi_end_io==NULL)
							continue;
L
Linus Torvalds 已提交
1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898
						/* 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;
1899
		sync_blocks -= (len>>9);
L
Linus Torvalds 已提交
1900 1901 1902 1903
	} while (r1_bio->bios[disk]->bi_vcnt < RESYNC_PAGES);
 bio_full:
	r1_bio->sectors = nr_sectors;

1904 1905 1906 1907 1908 1909 1910 1911
	/* 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) {
1912
				md_sync_acct(bio->bi_bdev, nr_sectors);
1913 1914 1915 1916 1917 1918
				generic_make_request(bio);
			}
		}
	} else {
		atomic_set(&r1_bio->remaining, 1);
		bio = r1_bio->bios[r1_bio->read_disk];
1919
		md_sync_acct(bio->bi_bdev, nr_sectors);
1920
		generic_make_request(bio);
L
Linus Torvalds 已提交
1921

1922
	}
L
Linus Torvalds 已提交
1923 1924 1925
	return nr_sectors;
}

1926 1927 1928 1929 1930 1931 1932 1933
static sector_t raid1_size(mddev_t *mddev, sector_t sectors, int raid_disks)
{
	if (sectors)
		return sectors;

	return mddev->dev_sectors;
}

L
Linus Torvalds 已提交
1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945
static int run(mddev_t *mddev)
{
	conf_t *conf;
	int i, j, disk_idx;
	mirror_info_t *disk;
	mdk_rdev_t *rdev;

	if (mddev->level != 1) {
		printk("raid1: %s: raid level not set to mirroring (%d)\n",
		       mdname(mddev), mddev->level);
		goto out;
	}
1946 1947 1948 1949 1950
	if (mddev->reshape_position != MaxSector) {
		printk("raid1: %s: reshape_position set but not supported\n",
		       mdname(mddev));
		goto out;
	}
L
Linus Torvalds 已提交
1951 1952 1953 1954 1955
	/*
	 * copy the already verified devices into our private RAID1
	 * bookkeeping area. [whatever we allocate in run(),
	 * should be freed in stop()]
	 */
1956
	conf = kzalloc(sizeof(conf_t), GFP_KERNEL);
L
Linus Torvalds 已提交
1957 1958 1959 1960
	mddev->private = conf;
	if (!conf)
		goto out_no_mem;

1961
	conf->mirrors = kzalloc(sizeof(struct mirror_info)*mddev->raid_disks,
L
Linus Torvalds 已提交
1962 1963 1964 1965
				 GFP_KERNEL);
	if (!conf->mirrors)
		goto out_no_mem;

1966 1967 1968 1969
	conf->tmppage = alloc_page(GFP_KERNEL);
	if (!conf->tmppage)
		goto out_no_mem;

L
Linus Torvalds 已提交
1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980
	conf->poolinfo = kmalloc(sizeof(*conf->poolinfo), GFP_KERNEL);
	if (!conf->poolinfo)
		goto out_no_mem;
	conf->poolinfo->mddev = mddev;
	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)
		goto out_no_mem;

1981 1982 1983
	spin_lock_init(&conf->device_lock);
	mddev->queue->queue_lock = &conf->device_lock;

1984
	list_for_each_entry(rdev, &mddev->disks, same_set) {
L
Linus Torvalds 已提交
1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
		disk_idx = rdev->raid_disk;
		if (disk_idx >= mddev->raid_disks
		    || disk_idx < 0)
			continue;
		disk = conf->mirrors + disk_idx;

		disk->rdev = rdev;

		blk_queue_stack_limits(mddev->queue,
				       rdev->bdev->bd_disk->queue);
		/* as we don't honour merge_bvec_fn, we must never risk
		 * violating it, so limit ->max_sector to one PAGE, as
		 * a one page request is never in violation.
		 */
		if (rdev->bdev->bd_disk->queue->merge_bvec_fn &&
		    mddev->queue->max_sectors > (PAGE_SIZE>>9))
			blk_queue_max_sectors(mddev->queue, PAGE_SIZE>>9);

		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);
2010
	init_waitqueue_head(&conf->wait_barrier);
L
Linus Torvalds 已提交
2011

2012 2013 2014
	bio_list_init(&conf->pending_bio_list);
	bio_list_init(&conf->flushing_bio_list);

L
Linus Torvalds 已提交
2015 2016 2017 2018 2019 2020

	mddev->degraded = 0;
	for (i = 0; i < conf->raid_disks; i++) {

		disk = conf->mirrors + i;

2021 2022
		if (!disk->rdev ||
		    !test_bit(In_sync, &disk->rdev->flags)) {
L
Linus Torvalds 已提交
2023 2024
			disk->head_position = 0;
			mddev->degraded++;
2025 2026
			if (disk->rdev)
				conf->fullsync = 1;
L
Linus Torvalds 已提交
2027 2028
		}
	}
2029 2030 2031 2032 2033 2034 2035
	if (mddev->degraded == conf->raid_disks) {
		printk(KERN_ERR "raid1: no operational mirrors for %s\n",
			mdname(mddev));
		goto out_free_conf;
	}
	if (conf->raid_disks - mddev->degraded == 1)
		mddev->recovery_cp = MaxSector;
L
Linus Torvalds 已提交
2036 2037 2038 2039 2040 2041 2042

	/*
	 * find the first working one and use it as a starting point
	 * to read balancing.
	 */
	for (j = 0; j < conf->raid_disks &&
		     (!conf->mirrors[j].rdev ||
2043
		      !test_bit(In_sync, &conf->mirrors[j].rdev->flags)) ; j++)
L
Linus Torvalds 已提交
2044 2045 2046 2047
		/* nothing */;
	conf->last_used = j;


2048 2049 2050 2051 2052 2053
	mddev->thread = md_register_thread(raid1d, mddev, "%s_raid1");
	if (!mddev->thread) {
		printk(KERN_ERR
		       "raid1: couldn't allocate thread for %s\n",
		       mdname(mddev));
		goto out_free_conf;
L
Linus Torvalds 已提交
2054
	}
2055

L
Linus Torvalds 已提交
2056 2057 2058 2059 2060 2061 2062
	printk(KERN_INFO 
		"raid1: raid set %s active with %d out of %d mirrors\n",
		mdname(mddev), mddev->raid_disks - mddev->degraded, 
		mddev->raid_disks);
	/*
	 * Ok, everything is just fine now
	 */
2063
	md_set_array_sectors(mddev, raid1_size(mddev, 0, 0));
L
Linus Torvalds 已提交
2064

2065
	mddev->queue->unplug_fn = raid1_unplug;
2066 2067
	mddev->queue->backing_dev_info.congested_fn = raid1_congested;
	mddev->queue->backing_dev_info.congested_data = mddev;
2068

L
Linus Torvalds 已提交
2069 2070 2071 2072 2073 2074 2075 2076 2077 2078
	return 0;

out_no_mem:
	printk(KERN_ERR "raid1: couldn't allocate memory for %s\n",
	       mdname(mddev));

out_free_conf:
	if (conf) {
		if (conf->r1bio_pool)
			mempool_destroy(conf->r1bio_pool);
2079
		kfree(conf->mirrors);
2080
		safe_put_page(conf->tmppage);
2081
		kfree(conf->poolinfo);
L
Linus Torvalds 已提交
2082 2083 2084 2085 2086 2087 2088 2089 2090 2091
		kfree(conf);
		mddev->private = NULL;
	}
out:
	return -EIO;
}

static int stop(mddev_t *mddev)
{
	conf_t *conf = mddev_to_conf(mddev);
2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102
	struct bitmap *bitmap = mddev->bitmap;
	int behind_wait = 0;

	/* wait for behind writes to complete */
	while (bitmap && atomic_read(&bitmap->behind_writes) > 0) {
		behind_wait++;
		printk(KERN_INFO "raid1: behind writes in progress on device %s, waiting to stop (%d)\n", mdname(mddev), behind_wait);
		set_current_state(TASK_UNINTERRUPTIBLE);
		schedule_timeout(HZ); /* wait a second */
		/* need to kick something here to make sure I/O goes? */
	}
L
Linus Torvalds 已提交
2103

2104 2105 2106
	raise_barrier(conf);
	lower_barrier(conf);

L
Linus Torvalds 已提交
2107 2108 2109 2110 2111
	md_unregister_thread(mddev->thread);
	mddev->thread = NULL;
	blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
	if (conf->r1bio_pool)
		mempool_destroy(conf->r1bio_pool);
2112 2113
	kfree(conf->mirrors);
	kfree(conf->poolinfo);
L
Linus Torvalds 已提交
2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127
	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.
	 */
2128
	md_set_array_sectors(mddev, raid1_size(mddev, sectors, 0));
D
Dan Williams 已提交
2129 2130
	if (mddev->array_sectors > raid1_size(mddev, sectors, 0))
		return -EINVAL;
2131
	set_capacity(mddev->gendisk, mddev->array_sectors);
2132
	mddev->changed = 1;
D
Dan Williams 已提交
2133
	if (sectors > mddev->dev_sectors &&
2134
	    mddev->recovery_cp == MaxSector) {
A
Andre Noll 已提交
2135
		mddev->recovery_cp = mddev->dev_sectors;
L
Linus Torvalds 已提交
2136 2137
		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
	}
D
Dan Williams 已提交
2138
	mddev->dev_sectors = sectors;
2139
	mddev->resync_max_sectors = sectors;
L
Linus Torvalds 已提交
2140 2141 2142
	return 0;
}

2143
static int raid1_reshape(mddev_t *mddev)
L
Linus Torvalds 已提交
2144 2145 2146 2147 2148 2149 2150 2151
{
	/* 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.
2152 2153 2154
	 *
	 * 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
2155 2156 2157 2158 2159
	 */
	mempool_t *newpool, *oldpool;
	struct pool_info *newpoolinfo;
	mirror_info_t *newmirrors;
	conf_t *conf = mddev_to_conf(mddev);
2160
	int cnt, raid_disks;
2161
	unsigned long flags;
2162
	int d, d2, err;
L
Linus Torvalds 已提交
2163

2164 2165 2166 2167 2168 2169 2170 2171 2172 2173
	/* Cannot change chunk_size, layout, or level */
	if (mddev->chunk_size != mddev->new_chunk ||
	    mddev->layout != mddev->new_layout ||
	    mddev->level != mddev->new_level) {
		mddev->new_chunk = mddev->chunk_size;
		mddev->new_layout = mddev->layout;
		mddev->new_level = mddev->level;
		return -EINVAL;
	}

2174 2175 2176
	err = md_allow_write(mddev);
	if (err)
		return err;
2177

2178 2179
	raid_disks = mddev->raid_disks + mddev->delta_disks;

2180 2181 2182 2183 2184 2185
	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
2186
			return -EBUSY;
2187
	}
L
Linus Torvalds 已提交
2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200

	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;
	}
2201
	newmirrors = kzalloc(sizeof(struct mirror_info) * raid_disks, GFP_KERNEL);
L
Linus Torvalds 已提交
2202 2203 2204 2205 2206 2207
	if (!newmirrors) {
		kfree(newpoolinfo);
		mempool_destroy(newpool);
		return -ENOMEM;
	}

2208
	raise_barrier(conf);
L
Linus Torvalds 已提交
2209 2210 2211 2212

	/* ok, everything is stopped */
	oldpool = conf->r1bio_pool;
	conf->r1bio_pool = newpool;
2213

2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228
	for (d = d2 = 0; d < conf->raid_disks; d++) {
		mdk_rdev_t *rdev = conf->mirrors[d].rdev;
		if (rdev && rdev->raid_disk != d2) {
			char nm[20];
			sprintf(nm, "rd%d", rdev->raid_disk);
			sysfs_remove_link(&mddev->kobj, nm);
			rdev->raid_disk = d2;
			sprintf(nm, "rd%d", rdev->raid_disk);
			sysfs_remove_link(&mddev->kobj, nm);
			if (sysfs_create_link(&mddev->kobj,
					      &rdev->kobj, nm))
				printk(KERN_WARNING
				       "md/raid1: cannot register "
				       "%s for %s\n",
				       nm, mdname(mddev));
2229
		}
2230 2231 2232
		if (rdev)
			newmirrors[d2++].rdev = rdev;
	}
L
Linus Torvalds 已提交
2233 2234 2235 2236 2237
	kfree(conf->mirrors);
	conf->mirrors = newmirrors;
	kfree(conf->poolinfo);
	conf->poolinfo = newpoolinfo;

2238
	spin_lock_irqsave(&conf->device_lock, flags);
L
Linus Torvalds 已提交
2239
	mddev->degraded += (raid_disks - conf->raid_disks);
2240
	spin_unlock_irqrestore(&conf->device_lock, flags);
L
Linus Torvalds 已提交
2241
	conf->raid_disks = mddev->raid_disks = raid_disks;
2242
	mddev->delta_disks = 0;
L
Linus Torvalds 已提交
2243

2244
	conf->last_used = 0; /* just make sure it is in-range */
2245
	lower_barrier(conf);
L
Linus Torvalds 已提交
2246 2247 2248 2249 2250 2251 2252 2253

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

	mempool_destroy(oldpool);
	return 0;
}

2254
static void raid1_quiesce(mddev_t *mddev, int state)
2255 2256 2257 2258
{
	conf_t *conf = mddev_to_conf(mddev);

	switch(state) {
2259
	case 1:
2260
		raise_barrier(conf);
2261
		break;
2262
	case 0:
2263
		lower_barrier(conf);
2264 2265 2266 2267
		break;
	}
}

L
Linus Torvalds 已提交
2268

2269
static struct mdk_personality raid1_personality =
L
Linus Torvalds 已提交
2270 2271
{
	.name		= "raid1",
2272
	.level		= 1,
L
Linus Torvalds 已提交
2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283
	.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,
2284
	.size		= raid1_size,
2285
	.check_reshape	= raid1_reshape,
2286
	.quiesce	= raid1_quiesce,
L
Linus Torvalds 已提交
2287 2288 2289 2290
};

static int __init raid_init(void)
{
2291
	return register_md_personality(&raid1_personality);
L
Linus Torvalds 已提交
2292 2293 2294 2295
}

static void raid_exit(void)
{
2296
	unregister_md_personality(&raid1_personality);
L
Linus Torvalds 已提交
2297 2298 2299 2300 2301 2302
}

module_init(raid_init);
module_exit(raid_exit);
MODULE_LICENSE("GPL");
MODULE_ALIAS("md-personality-3"); /* RAID1 */
2303
MODULE_ALIAS("md-raid1");
2304
MODULE_ALIAS("md-level-1");