scrub.c 44.0 KB
Newer Older
A
Arne Jansen 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
/*
 * Copyright (C) 2011 STRATO.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License v2 as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 */

#include <linux/blkdev.h>
20
#include <linux/ratelimit.h>
A
Arne Jansen 已提交
21 22 23 24
#include "ctree.h"
#include "volumes.h"
#include "disk-io.h"
#include "ordered-data.h"
25
#include "transaction.h"
26
#include "backref.h"
27
#include "extent_io.h"
A
Arne Jansen 已提交
28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54

/*
 * This is only the first step towards a full-features scrub. It reads all
 * extent and super block and verifies the checksums. In case a bad checksum
 * is found or the extent cannot be read, good data will be written back if
 * any can be found.
 *
 * Future enhancements:
 *  - In case an unrepairable extent is encountered, track which files are
 *    affected and report them
 *  - In case of a read error on files with nodatasum, map the file and read
 *    the extent to trigger a writeback of the good copy
 *  - track and record media errors, throw out bad devices
 *  - add a mode to also read unallocated space
 */

struct scrub_bio;
struct scrub_page;
struct scrub_dev;
static void scrub_bio_end_io(struct bio *bio, int err);
static void scrub_checksum(struct btrfs_work *work);
static int scrub_checksum_data(struct scrub_dev *sdev,
			       struct scrub_page *spag, void *buffer);
static int scrub_checksum_tree_block(struct scrub_dev *sdev,
				     struct scrub_page *spag, u64 logical,
				     void *buffer);
static int scrub_checksum_super(struct scrub_bio *sbio, void *buffer);
I
Ilya Dryomov 已提交
55 56 57 58 59
static int scrub_fixup_check(struct scrub_bio *sbio, int ix);
static void scrub_fixup_end_io(struct bio *bio, int err);
static int scrub_fixup_io(int rw, struct block_device *bdev, sector_t sector,
			  struct page *page);
static void scrub_fixup(struct scrub_bio *sbio, int ix);
A
Arne Jansen 已提交
60 61 62 63 64 65 66

#define SCRUB_PAGES_PER_BIO	16	/* 64k per bio */
#define SCRUB_BIOS_PER_DEV	16	/* 1 MB per device in flight */

struct scrub_page {
	u64			flags;  /* extent flags */
	u64			generation;
67
	int			mirror_num;
A
Arne Jansen 已提交
68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90
	int			have_csum;
	u8			csum[BTRFS_CSUM_SIZE];
};

struct scrub_bio {
	int			index;
	struct scrub_dev	*sdev;
	struct bio		*bio;
	int			err;
	u64			logical;
	u64			physical;
	struct scrub_page	spag[SCRUB_PAGES_PER_BIO];
	u64			count;
	int			next_free;
	struct btrfs_work	work;
};

struct scrub_dev {
	struct scrub_bio	*bios[SCRUB_BIOS_PER_DEV];
	struct btrfs_device	*dev;
	int			first_free;
	int			curr;
	atomic_t		in_flight;
91
	atomic_t		fixup_cnt;
A
Arne Jansen 已提交
92 93 94 95 96
	spinlock_t		list_lock;
	wait_queue_head_t	list_wait;
	u16			csum_size;
	struct list_head	csum_list;
	atomic_t		cancel_req;
A
Arne Jansen 已提交
97
	int			readonly;
A
Arne Jansen 已提交
98 99 100 101 102 103 104
	/*
	 * statistics
	 */
	struct btrfs_scrub_progress stat;
	spinlock_t		stat_lock;
};

105 106 107 108 109 110 111 112
struct scrub_fixup_nodatasum {
	struct scrub_dev	*sdev;
	u64			logical;
	struct btrfs_root	*root;
	struct btrfs_work	work;
	int			mirror_num;
};

113 114 115 116 117 118 119 120 121 122 123 124 125
struct scrub_warning {
	struct btrfs_path	*path;
	u64			extent_item_size;
	char			*scratch_buf;
	char			*msg_buf;
	const char		*errstr;
	sector_t		sector;
	u64			logical;
	struct btrfs_device	*dev;
	int			msg_bufsize;
	int			scratch_bufsize;
};

A
Arne Jansen 已提交
126 127 128 129 130 131 132 133 134 135 136
static void scrub_free_csums(struct scrub_dev *sdev)
{
	while (!list_empty(&sdev->csum_list)) {
		struct btrfs_ordered_sum *sum;
		sum = list_first_entry(&sdev->csum_list,
				       struct btrfs_ordered_sum, list);
		list_del(&sum->list);
		kfree(sum);
	}
}

137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153
static void scrub_free_bio(struct bio *bio)
{
	int i;
	struct page *last_page = NULL;

	if (!bio)
		return;

	for (i = 0; i < bio->bi_vcnt; ++i) {
		if (bio->bi_io_vec[i].bv_page == last_page)
			continue;
		last_page = bio->bi_io_vec[i].bv_page;
		__free_page(last_page);
	}
	bio_put(bio);
}

A
Arne Jansen 已提交
154 155 156 157 158 159 160 161 162 163 164 165 166
static noinline_for_stack void scrub_free_dev(struct scrub_dev *sdev)
{
	int i;

	if (!sdev)
		return;

	for (i = 0; i < SCRUB_BIOS_PER_DEV; ++i) {
		struct scrub_bio *sbio = sdev->bios[i];

		if (!sbio)
			break;

167
		scrub_free_bio(sbio->bio);
A
Arne Jansen 已提交
168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200
		kfree(sbio);
	}

	scrub_free_csums(sdev);
	kfree(sdev);
}

static noinline_for_stack
struct scrub_dev *scrub_setup_dev(struct btrfs_device *dev)
{
	struct scrub_dev *sdev;
	int		i;
	struct btrfs_fs_info *fs_info = dev->dev_root->fs_info;

	sdev = kzalloc(sizeof(*sdev), GFP_NOFS);
	if (!sdev)
		goto nomem;
	sdev->dev = dev;
	for (i = 0; i < SCRUB_BIOS_PER_DEV; ++i) {
		struct scrub_bio *sbio;

		sbio = kzalloc(sizeof(*sbio), GFP_NOFS);
		if (!sbio)
			goto nomem;
		sdev->bios[i] = sbio;

		sbio->index = i;
		sbio->sdev = sdev;
		sbio->count = 0;
		sbio->work.func = scrub_checksum;

		if (i != SCRUB_BIOS_PER_DEV-1)
			sdev->bios[i]->next_free = i + 1;
201
		else
A
Arne Jansen 已提交
202 203 204 205 206
			sdev->bios[i]->next_free = -1;
	}
	sdev->first_free = 0;
	sdev->curr = -1;
	atomic_set(&sdev->in_flight, 0);
207
	atomic_set(&sdev->fixup_cnt, 0);
A
Arne Jansen 已提交
208
	atomic_set(&sdev->cancel_req, 0);
209
	sdev->csum_size = btrfs_super_csum_size(fs_info->super_copy);
A
Arne Jansen 已提交
210 211 212 213 214 215 216 217 218 219 220 221
	INIT_LIST_HEAD(&sdev->csum_list);

	spin_lock_init(&sdev->list_lock);
	spin_lock_init(&sdev->stat_lock);
	init_waitqueue_head(&sdev->list_wait);
	return sdev;

nomem:
	scrub_free_dev(sdev);
	return ERR_PTR(-ENOMEM);
}

222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274
static int scrub_print_warning_inode(u64 inum, u64 offset, u64 root, void *ctx)
{
	u64 isize;
	u32 nlink;
	int ret;
	int i;
	struct extent_buffer *eb;
	struct btrfs_inode_item *inode_item;
	struct scrub_warning *swarn = ctx;
	struct btrfs_fs_info *fs_info = swarn->dev->dev_root->fs_info;
	struct inode_fs_paths *ipath = NULL;
	struct btrfs_root *local_root;
	struct btrfs_key root_key;

	root_key.objectid = root;
	root_key.type = BTRFS_ROOT_ITEM_KEY;
	root_key.offset = (u64)-1;
	local_root = btrfs_read_fs_root_no_name(fs_info, &root_key);
	if (IS_ERR(local_root)) {
		ret = PTR_ERR(local_root);
		goto err;
	}

	ret = inode_item_info(inum, 0, local_root, swarn->path);
	if (ret) {
		btrfs_release_path(swarn->path);
		goto err;
	}

	eb = swarn->path->nodes[0];
	inode_item = btrfs_item_ptr(eb, swarn->path->slots[0],
					struct btrfs_inode_item);
	isize = btrfs_inode_size(eb, inode_item);
	nlink = btrfs_inode_nlink(eb, inode_item);
	btrfs_release_path(swarn->path);

	ipath = init_ipath(4096, local_root, swarn->path);
	ret = paths_from_inode(inum, ipath);

	if (ret < 0)
		goto err;

	/*
	 * we deliberately ignore the bit ipath might have been too small to
	 * hold all of the paths here
	 */
	for (i = 0; i < ipath->fspath->elem_cnt; ++i)
		printk(KERN_WARNING "btrfs: %s at logical %llu on dev "
			"%s, sector %llu, root %llu, inode %llu, offset %llu, "
			"length %llu, links %u (path: %s)\n", swarn->errstr,
			swarn->logical, swarn->dev->name,
			(unsigned long long)swarn->sector, root, inum, offset,
			min(isize - offset, (u64)PAGE_SIZE), nlink,
275
			(char *)ipath->fspath->val[i]);
276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358

	free_ipath(ipath);
	return 0;

err:
	printk(KERN_WARNING "btrfs: %s at logical %llu on dev "
		"%s, sector %llu, root %llu, inode %llu, offset %llu: path "
		"resolving failed with ret=%d\n", swarn->errstr,
		swarn->logical, swarn->dev->name,
		(unsigned long long)swarn->sector, root, inum, offset, ret);

	free_ipath(ipath);
	return 0;
}

static void scrub_print_warning(const char *errstr, struct scrub_bio *sbio,
				int ix)
{
	struct btrfs_device *dev = sbio->sdev->dev;
	struct btrfs_fs_info *fs_info = dev->dev_root->fs_info;
	struct btrfs_path *path;
	struct btrfs_key found_key;
	struct extent_buffer *eb;
	struct btrfs_extent_item *ei;
	struct scrub_warning swarn;
	u32 item_size;
	int ret;
	u64 ref_root;
	u8 ref_level;
	unsigned long ptr = 0;
	const int bufsize = 4096;
	u64 extent_offset;

	path = btrfs_alloc_path();

	swarn.scratch_buf = kmalloc(bufsize, GFP_NOFS);
	swarn.msg_buf = kmalloc(bufsize, GFP_NOFS);
	swarn.sector = (sbio->physical + ix * PAGE_SIZE) >> 9;
	swarn.logical = sbio->logical + ix * PAGE_SIZE;
	swarn.errstr = errstr;
	swarn.dev = dev;
	swarn.msg_bufsize = bufsize;
	swarn.scratch_bufsize = bufsize;

	if (!path || !swarn.scratch_buf || !swarn.msg_buf)
		goto out;

	ret = extent_from_logical(fs_info, swarn.logical, path, &found_key);
	if (ret < 0)
		goto out;

	extent_offset = swarn.logical - found_key.objectid;
	swarn.extent_item_size = found_key.offset;

	eb = path->nodes[0];
	ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
	item_size = btrfs_item_size_nr(eb, path->slots[0]);

	if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
		do {
			ret = tree_backref_for_extent(&ptr, eb, ei, item_size,
							&ref_root, &ref_level);
			printk(KERN_WARNING "%s at logical %llu on dev %s, "
				"sector %llu: metadata %s (level %d) in tree "
				"%llu\n", errstr, swarn.logical, dev->name,
				(unsigned long long)swarn.sector,
				ref_level ? "node" : "leaf",
				ret < 0 ? -1 : ref_level,
				ret < 0 ? -1 : ref_root);
		} while (ret != 1);
	} else {
		swarn.path = path;
		iterate_extent_inodes(fs_info, path, found_key.objectid,
					extent_offset,
					scrub_print_warning_inode, &swarn);
	}

out:
	btrfs_free_path(path);
	kfree(swarn.scratch_buf);
	kfree(swarn.msg_buf);
}

359 360
static int scrub_fixup_readpage(u64 inum, u64 offset, u64 root, void *ctx)
{
361
	struct page *page = NULL;
362 363 364
	unsigned long index;
	struct scrub_fixup_nodatasum *fixup = ctx;
	int ret;
365
	int corrected = 0;
366
	struct btrfs_key key;
367
	struct inode *inode = NULL;
368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387
	u64 end = offset + PAGE_SIZE - 1;
	struct btrfs_root *local_root;

	key.objectid = root;
	key.type = BTRFS_ROOT_ITEM_KEY;
	key.offset = (u64)-1;
	local_root = btrfs_read_fs_root_no_name(fixup->root->fs_info, &key);
	if (IS_ERR(local_root))
		return PTR_ERR(local_root);

	key.type = BTRFS_INODE_ITEM_KEY;
	key.objectid = inum;
	key.offset = 0;
	inode = btrfs_iget(fixup->root->fs_info->sb, &key, local_root, NULL);
	if (IS_ERR(inode))
		return PTR_ERR(inode);

	index = offset >> PAGE_CACHE_SHIFT;

	page = find_or_create_page(inode->i_mapping, index, GFP_NOFS);
388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453
	if (!page) {
		ret = -ENOMEM;
		goto out;
	}

	if (PageUptodate(page)) {
		struct btrfs_mapping_tree *map_tree;
		if (PageDirty(page)) {
			/*
			 * we need to write the data to the defect sector. the
			 * data that was in that sector is not in memory,
			 * because the page was modified. we must not write the
			 * modified page to that sector.
			 *
			 * TODO: what could be done here: wait for the delalloc
			 *       runner to write out that page (might involve
			 *       COW) and see whether the sector is still
			 *       referenced afterwards.
			 *
			 * For the meantime, we'll treat this error
			 * incorrectable, although there is a chance that a
			 * later scrub will find the bad sector again and that
			 * there's no dirty page in memory, then.
			 */
			ret = -EIO;
			goto out;
		}
		map_tree = &BTRFS_I(inode)->root->fs_info->mapping_tree;
		ret = repair_io_failure(map_tree, offset, PAGE_SIZE,
					fixup->logical, page,
					fixup->mirror_num);
		unlock_page(page);
		corrected = !ret;
	} else {
		/*
		 * we need to get good data first. the general readpage path
		 * will call repair_io_failure for us, we just have to make
		 * sure we read the bad mirror.
		 */
		ret = set_extent_bits(&BTRFS_I(inode)->io_tree, offset, end,
					EXTENT_DAMAGED, GFP_NOFS);
		if (ret) {
			/* set_extent_bits should give proper error */
			WARN_ON(ret > 0);
			if (ret > 0)
				ret = -EFAULT;
			goto out;
		}

		ret = extent_read_full_page(&BTRFS_I(inode)->io_tree, page,
						btrfs_get_extent,
						fixup->mirror_num);
		wait_on_page_locked(page);

		corrected = !test_range_bit(&BTRFS_I(inode)->io_tree, offset,
						end, EXTENT_DAMAGED, 0, NULL);
		if (!corrected)
			clear_extent_bits(&BTRFS_I(inode)->io_tree, offset, end,
						EXTENT_DAMAGED, GFP_NOFS);
	}

out:
	if (page)
		put_page(page);
	if (inode)
		iput(inode);
454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544

	if (ret < 0)
		return ret;

	if (ret == 0 && corrected) {
		/*
		 * we only need to call readpage for one of the inodes belonging
		 * to this extent. so make iterate_extent_inodes stop
		 */
		return 1;
	}

	return -EIO;
}

static void scrub_fixup_nodatasum(struct btrfs_work *work)
{
	int ret;
	struct scrub_fixup_nodatasum *fixup;
	struct scrub_dev *sdev;
	struct btrfs_trans_handle *trans = NULL;
	struct btrfs_fs_info *fs_info;
	struct btrfs_path *path;
	int uncorrectable = 0;

	fixup = container_of(work, struct scrub_fixup_nodatasum, work);
	sdev = fixup->sdev;
	fs_info = fixup->root->fs_info;

	path = btrfs_alloc_path();
	if (!path) {
		spin_lock(&sdev->stat_lock);
		++sdev->stat.malloc_errors;
		spin_unlock(&sdev->stat_lock);
		uncorrectable = 1;
		goto out;
	}

	trans = btrfs_join_transaction(fixup->root);
	if (IS_ERR(trans)) {
		uncorrectable = 1;
		goto out;
	}

	/*
	 * the idea is to trigger a regular read through the standard path. we
	 * read a page from the (failed) logical address by specifying the
	 * corresponding copynum of the failed sector. thus, that readpage is
	 * expected to fail.
	 * that is the point where on-the-fly error correction will kick in
	 * (once it's finished) and rewrite the failed sector if a good copy
	 * can be found.
	 */
	ret = iterate_inodes_from_logical(fixup->logical, fixup->root->fs_info,
						path, scrub_fixup_readpage,
						fixup);
	if (ret < 0) {
		uncorrectable = 1;
		goto out;
	}
	WARN_ON(ret != 1);

	spin_lock(&sdev->stat_lock);
	++sdev->stat.corrected_errors;
	spin_unlock(&sdev->stat_lock);

out:
	if (trans && !IS_ERR(trans))
		btrfs_end_transaction(trans, fixup->root);
	if (uncorrectable) {
		spin_lock(&sdev->stat_lock);
		++sdev->stat.uncorrectable_errors;
		spin_unlock(&sdev->stat_lock);
		printk_ratelimited(KERN_ERR "btrfs: unable to fixup "
					"(nodatasum) error at logical %llu\n",
					fixup->logical);
	}

	btrfs_free_path(path);
	kfree(fixup);

	/* see caller why we're pretending to be paused in the scrub counters */
	mutex_lock(&fs_info->scrub_lock);
	atomic_dec(&fs_info->scrubs_running);
	atomic_dec(&fs_info->scrubs_paused);
	mutex_unlock(&fs_info->scrub_lock);
	atomic_dec(&sdev->fixup_cnt);
	wake_up(&fs_info->scrub_pause_wait);
	wake_up(&sdev->list_wait);
}

A
Arne Jansen 已提交
545 546 547 548 549 550
/*
 * scrub_recheck_error gets called when either verification of the page
 * failed or the bio failed to read, e.g. with EIO. In the latter case,
 * recheck_error gets called for every page in the bio, even though only
 * one may be bad
 */
551
static int scrub_recheck_error(struct scrub_bio *sbio, int ix)
A
Arne Jansen 已提交
552
{
553 554
	struct scrub_dev *sdev = sbio->sdev;
	u64 sector = (sbio->physical + ix * PAGE_SIZE) >> 9;
555 556
	static DEFINE_RATELIMIT_STATE(_rs, DEFAULT_RATELIMIT_INTERVAL,
					DEFAULT_RATELIMIT_BURST);
557

I
Ilya Dryomov 已提交
558
	if (sbio->err) {
559
		if (scrub_fixup_io(READ, sbio->sdev->dev->bdev, sector,
I
Ilya Dryomov 已提交
560 561
				   sbio->bio->bi_io_vec[ix].bv_page) == 0) {
			if (scrub_fixup_check(sbio, ix) == 0)
562
				return 0;
I
Ilya Dryomov 已提交
563
		}
564 565 566 567 568
		if (__ratelimit(&_rs))
			scrub_print_warning("i/o error", sbio, ix);
	} else {
		if (__ratelimit(&_rs))
			scrub_print_warning("checksum error", sbio, ix);
A
Arne Jansen 已提交
569 570
	}

571 572 573 574
	spin_lock(&sdev->stat_lock);
	++sdev->stat.read_errors;
	spin_unlock(&sdev->stat_lock);

I
Ilya Dryomov 已提交
575
	scrub_fixup(sbio, ix);
576
	return 1;
A
Arne Jansen 已提交
577 578
}

I
Ilya Dryomov 已提交
579
static int scrub_fixup_check(struct scrub_bio *sbio, int ix)
A
Arne Jansen 已提交
580 581 582 583
{
	int ret = 1;
	struct page *page;
	void *buffer;
I
Ilya Dryomov 已提交
584
	u64 flags = sbio->spag[ix].flags;
A
Arne Jansen 已提交
585

I
Ilya Dryomov 已提交
586
	page = sbio->bio->bi_io_vec[ix].bv_page;
A
Arne Jansen 已提交
587 588
	buffer = kmap_atomic(page, KM_USER0);
	if (flags & BTRFS_EXTENT_FLAG_DATA) {
I
Ilya Dryomov 已提交
589 590
		ret = scrub_checksum_data(sbio->sdev,
					  sbio->spag + ix, buffer);
A
Arne Jansen 已提交
591
	} else if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
I
Ilya Dryomov 已提交
592 593 594
		ret = scrub_checksum_tree_block(sbio->sdev,
						sbio->spag + ix,
						sbio->logical + ix * PAGE_SIZE,
A
Arne Jansen 已提交
595 596 597 598 599 600 601 602 603 604 605 606 607 608
						buffer);
	} else {
		WARN_ON(1);
	}
	kunmap_atomic(buffer, KM_USER0);

	return ret;
}

static void scrub_fixup_end_io(struct bio *bio, int err)
{
	complete((struct completion *)bio->bi_private);
}

I
Ilya Dryomov 已提交
609
static void scrub_fixup(struct scrub_bio *sbio, int ix)
A
Arne Jansen 已提交
610
{
I
Ilya Dryomov 已提交
611
	struct scrub_dev *sdev = sbio->sdev;
A
Arne Jansen 已提交
612 613
	struct btrfs_fs_info *fs_info = sdev->dev->dev_root->fs_info;
	struct btrfs_mapping_tree *map_tree = &fs_info->mapping_tree;
614
	struct btrfs_bio *bbio = NULL;
615
	struct scrub_fixup_nodatasum *fixup;
I
Ilya Dryomov 已提交
616
	u64 logical = sbio->logical + ix * PAGE_SIZE;
A
Arne Jansen 已提交
617 618 619 620 621
	u64 length;
	int i;
	int ret;
	DECLARE_COMPLETION_ONSTACK(complete);

I
Ilya Dryomov 已提交
622 623
	if ((sbio->spag[ix].flags & BTRFS_EXTENT_FLAG_DATA) &&
	    (sbio->spag[ix].have_csum == 0)) {
624 625 626 627 628 629 630
		fixup = kzalloc(sizeof(*fixup), GFP_NOFS);
		if (!fixup)
			goto uncorrectable;
		fixup->sdev = sdev;
		fixup->logical = logical;
		fixup->root = fs_info->extent_root;
		fixup->mirror_num = sbio->spag[ix].mirror_num;
A
Arne Jansen 已提交
631
		/*
632 633 634 635 636 637 638
		 * increment scrubs_running to prevent cancel requests from
		 * completing as long as a fixup worker is running. we must also
		 * increment scrubs_paused to prevent deadlocking on pause
		 * requests used for transactions commits (as the worker uses a
		 * transaction context). it is safe to regard the fixup worker
		 * as paused for all matters practical. effectively, we only
		 * avoid cancellation requests from completing.
A
Arne Jansen 已提交
639
		 */
640 641 642 643 644 645 646 647
		mutex_lock(&fs_info->scrub_lock);
		atomic_inc(&fs_info->scrubs_running);
		atomic_inc(&fs_info->scrubs_paused);
		mutex_unlock(&fs_info->scrub_lock);
		atomic_inc(&sdev->fixup_cnt);
		fixup->work.func = scrub_fixup_nodatasum;
		btrfs_queue_worker(&fs_info->scrub_workers, &fixup->work);
		return;
A
Arne Jansen 已提交
648 649 650
	}

	length = PAGE_SIZE;
I
Ilya Dryomov 已提交
651
	ret = btrfs_map_block(map_tree, REQ_WRITE, logical, &length,
652 653
			      &bbio, 0);
	if (ret || !bbio || length < PAGE_SIZE) {
A
Arne Jansen 已提交
654 655
		printk(KERN_ERR
		       "scrub_fixup: btrfs_map_block failed us for %llu\n",
I
Ilya Dryomov 已提交
656
		       (unsigned long long)logical);
A
Arne Jansen 已提交
657
		WARN_ON(1);
658
		kfree(bbio);
A
Arne Jansen 已提交
659 660 661
		return;
	}

662
	if (bbio->num_stripes == 1)
A
Arne Jansen 已提交
663 664 665 666 667 668
		/* there aren't any replicas */
		goto uncorrectable;

	/*
	 * first find a good copy
	 */
669
	for (i = 0; i < bbio->num_stripes; ++i) {
670
		if (i + 1 == sbio->spag[ix].mirror_num)
A
Arne Jansen 已提交
671 672
			continue;

673 674
		if (scrub_fixup_io(READ, bbio->stripes[i].dev->bdev,
				   bbio->stripes[i].physical >> 9,
I
Ilya Dryomov 已提交
675
				   sbio->bio->bi_io_vec[ix].bv_page)) {
A
Arne Jansen 已提交
676 677
			/* I/O-error, this is not a good copy */
			continue;
I
Ilya Dryomov 已提交
678
		}
A
Arne Jansen 已提交
679

I
Ilya Dryomov 已提交
680
		if (scrub_fixup_check(sbio, ix) == 0)
A
Arne Jansen 已提交
681 682
			break;
	}
683
	if (i == bbio->num_stripes)
A
Arne Jansen 已提交
684 685
		goto uncorrectable;

A
Arne Jansen 已提交
686 687 688 689 690 691 692 693 694 695
	if (!sdev->readonly) {
		/*
		 * bi_io_vec[ix].bv_page now contains good data, write it back
		 */
		if (scrub_fixup_io(WRITE, sdev->dev->bdev,
				   (sbio->physical + ix * PAGE_SIZE) >> 9,
				   sbio->bio->bi_io_vec[ix].bv_page)) {
			/* I/O-error, writeback failed, give up */
			goto uncorrectable;
		}
I
Ilya Dryomov 已提交
696
	}
A
Arne Jansen 已提交
697

698
	kfree(bbio);
A
Arne Jansen 已提交
699 700 701 702
	spin_lock(&sdev->stat_lock);
	++sdev->stat.corrected_errors;
	spin_unlock(&sdev->stat_lock);

703 704
	printk_ratelimited(KERN_ERR "btrfs: fixed up error at logical %llu\n",
			       (unsigned long long)logical);
A
Arne Jansen 已提交
705 706 707
	return;

uncorrectable:
708
	kfree(bbio);
A
Arne Jansen 已提交
709 710 711 712
	spin_lock(&sdev->stat_lock);
	++sdev->stat.uncorrectable_errors;
	spin_unlock(&sdev->stat_lock);

713 714
	printk_ratelimited(KERN_ERR "btrfs: unable to fixup (regular) error at "
				"logical %llu\n", (unsigned long long)logical);
I
Ilya Dryomov 已提交
715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731
}

static int scrub_fixup_io(int rw, struct block_device *bdev, sector_t sector,
			 struct page *page)
{
	struct bio *bio = NULL;
	int ret;
	DECLARE_COMPLETION_ONSTACK(complete);

	bio = bio_alloc(GFP_NOFS, 1);
	bio->bi_bdev = bdev;
	bio->bi_sector = sector;
	bio_add_page(bio, page, PAGE_SIZE, 0);
	bio->bi_end_io = scrub_fixup_end_io;
	bio->bi_private = &complete;
	submit_bio(rw, bio);

732
	/* this will also unplug the queue */
I
Ilya Dryomov 已提交
733 734 735 736 737
	wait_for_completion(&complete);

	ret = !test_bit(BIO_UPTODATE, &bio->bi_flags);
	bio_put(bio);
	return ret;
A
Arne Jansen 已提交
738 739 740 741 742 743 744 745 746
}

static void scrub_bio_end_io(struct bio *bio, int err)
{
	struct scrub_bio *sbio = bio->bi_private;
	struct scrub_dev *sdev = sbio->sdev;
	struct btrfs_fs_info *fs_info = sdev->dev->dev_root->fs_info;

	sbio->err = err;
747
	sbio->bio = bio;
A
Arne Jansen 已提交
748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763

	btrfs_queue_worker(&fs_info->scrub_workers, &sbio->work);
}

static void scrub_checksum(struct btrfs_work *work)
{
	struct scrub_bio *sbio = container_of(work, struct scrub_bio, work);
	struct scrub_dev *sdev = sbio->sdev;
	struct page *page;
	void *buffer;
	int i;
	u64 flags;
	u64 logical;
	int ret;

	if (sbio->err) {
764
		ret = 0;
A
Arne Jansen 已提交
765
		for (i = 0; i < sbio->count; ++i)
766 767 768 769 770 771
			ret |= scrub_recheck_error(sbio, i);
		if (!ret) {
			spin_lock(&sdev->stat_lock);
			++sdev->stat.unverified_errors;
			spin_unlock(&sdev->stat_lock);
		}
I
Ilya Dryomov 已提交
772 773 774 775 776 777 778 779 780 781 782 783

		sbio->bio->bi_flags &= ~(BIO_POOL_MASK - 1);
		sbio->bio->bi_flags |= 1 << BIO_UPTODATE;
		sbio->bio->bi_phys_segments = 0;
		sbio->bio->bi_idx = 0;

		for (i = 0; i < sbio->count; i++) {
			struct bio_vec *bi;
			bi = &sbio->bio->bi_io_vec[i];
			bi->bv_offset = 0;
			bi->bv_len = PAGE_SIZE;
		}
A
Arne Jansen 已提交
784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803
		goto out;
	}
	for (i = 0; i < sbio->count; ++i) {
		page = sbio->bio->bi_io_vec[i].bv_page;
		buffer = kmap_atomic(page, KM_USER0);
		flags = sbio->spag[i].flags;
		logical = sbio->logical + i * PAGE_SIZE;
		ret = 0;
		if (flags & BTRFS_EXTENT_FLAG_DATA) {
			ret = scrub_checksum_data(sdev, sbio->spag + i, buffer);
		} else if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
			ret = scrub_checksum_tree_block(sdev, sbio->spag + i,
							logical, buffer);
		} else if (flags & BTRFS_EXTENT_FLAG_SUPER) {
			BUG_ON(i);
			(void)scrub_checksum_super(sbio, buffer);
		} else {
			WARN_ON(1);
		}
		kunmap_atomic(buffer, KM_USER0);
804 805 806 807 808 809 810 811
		if (ret) {
			ret = scrub_recheck_error(sbio, i);
			if (!ret) {
				spin_lock(&sdev->stat_lock);
				++sdev->stat.unverified_errors;
				spin_unlock(&sdev->stat_lock);
			}
		}
A
Arne Jansen 已提交
812 813 814
	}

out:
815 816
	scrub_free_bio(sbio->bio);
	sbio->bio = NULL;
A
Arne Jansen 已提交
817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955
	spin_lock(&sdev->list_lock);
	sbio->next_free = sdev->first_free;
	sdev->first_free = sbio->index;
	spin_unlock(&sdev->list_lock);
	atomic_dec(&sdev->in_flight);
	wake_up(&sdev->list_wait);
}

static int scrub_checksum_data(struct scrub_dev *sdev,
			       struct scrub_page *spag, void *buffer)
{
	u8 csum[BTRFS_CSUM_SIZE];
	u32 crc = ~(u32)0;
	int fail = 0;
	struct btrfs_root *root = sdev->dev->dev_root;

	if (!spag->have_csum)
		return 0;

	crc = btrfs_csum_data(root, buffer, crc, PAGE_SIZE);
	btrfs_csum_final(crc, csum);
	if (memcmp(csum, spag->csum, sdev->csum_size))
		fail = 1;

	spin_lock(&sdev->stat_lock);
	++sdev->stat.data_extents_scrubbed;
	sdev->stat.data_bytes_scrubbed += PAGE_SIZE;
	if (fail)
		++sdev->stat.csum_errors;
	spin_unlock(&sdev->stat_lock);

	return fail;
}

static int scrub_checksum_tree_block(struct scrub_dev *sdev,
				     struct scrub_page *spag, u64 logical,
				     void *buffer)
{
	struct btrfs_header *h;
	struct btrfs_root *root = sdev->dev->dev_root;
	struct btrfs_fs_info *fs_info = root->fs_info;
	u8 csum[BTRFS_CSUM_SIZE];
	u32 crc = ~(u32)0;
	int fail = 0;
	int crc_fail = 0;

	/*
	 * we don't use the getter functions here, as we
	 * a) don't have an extent buffer and
	 * b) the page is already kmapped
	 */
	h = (struct btrfs_header *)buffer;

	if (logical != le64_to_cpu(h->bytenr))
		++fail;

	if (spag->generation != le64_to_cpu(h->generation))
		++fail;

	if (memcmp(h->fsid, fs_info->fsid, BTRFS_UUID_SIZE))
		++fail;

	if (memcmp(h->chunk_tree_uuid, fs_info->chunk_tree_uuid,
		   BTRFS_UUID_SIZE))
		++fail;

	crc = btrfs_csum_data(root, buffer + BTRFS_CSUM_SIZE, crc,
			      PAGE_SIZE - BTRFS_CSUM_SIZE);
	btrfs_csum_final(crc, csum);
	if (memcmp(csum, h->csum, sdev->csum_size))
		++crc_fail;

	spin_lock(&sdev->stat_lock);
	++sdev->stat.tree_extents_scrubbed;
	sdev->stat.tree_bytes_scrubbed += PAGE_SIZE;
	if (crc_fail)
		++sdev->stat.csum_errors;
	if (fail)
		++sdev->stat.verify_errors;
	spin_unlock(&sdev->stat_lock);

	return fail || crc_fail;
}

static int scrub_checksum_super(struct scrub_bio *sbio, void *buffer)
{
	struct btrfs_super_block *s;
	u64 logical;
	struct scrub_dev *sdev = sbio->sdev;
	struct btrfs_root *root = sdev->dev->dev_root;
	struct btrfs_fs_info *fs_info = root->fs_info;
	u8 csum[BTRFS_CSUM_SIZE];
	u32 crc = ~(u32)0;
	int fail = 0;

	s = (struct btrfs_super_block *)buffer;
	logical = sbio->logical;

	if (logical != le64_to_cpu(s->bytenr))
		++fail;

	if (sbio->spag[0].generation != le64_to_cpu(s->generation))
		++fail;

	if (memcmp(s->fsid, fs_info->fsid, BTRFS_UUID_SIZE))
		++fail;

	crc = btrfs_csum_data(root, buffer + BTRFS_CSUM_SIZE, crc,
			      PAGE_SIZE - BTRFS_CSUM_SIZE);
	btrfs_csum_final(crc, csum);
	if (memcmp(csum, s->csum, sbio->sdev->csum_size))
		++fail;

	if (fail) {
		/*
		 * if we find an error in a super block, we just report it.
		 * They will get written with the next transaction commit
		 * anyway
		 */
		spin_lock(&sdev->stat_lock);
		++sdev->stat.super_errors;
		spin_unlock(&sdev->stat_lock);
	}

	return fail;
}

static int scrub_submit(struct scrub_dev *sdev)
{
	struct scrub_bio *sbio;

	if (sdev->curr == -1)
		return 0;

	sbio = sdev->bios[sdev->curr];
	sbio->err = 0;
	sdev->curr = -1;
	atomic_inc(&sdev->in_flight);

956
	submit_bio(READ, sbio->bio);
A
Arne Jansen 已提交
957 958 959 960 961

	return 0;
}

static int scrub_page(struct scrub_dev *sdev, u64 logical, u64 len,
962
		      u64 physical, u64 flags, u64 gen, int mirror_num,
A
Arne Jansen 已提交
963 964 965
		      u8 *csum, int force)
{
	struct scrub_bio *sbio;
966 967
	struct page *page;
	int ret;
A
Arne Jansen 已提交
968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987

again:
	/*
	 * grab a fresh bio or wait for one to become available
	 */
	while (sdev->curr == -1) {
		spin_lock(&sdev->list_lock);
		sdev->curr = sdev->first_free;
		if (sdev->curr != -1) {
			sdev->first_free = sdev->bios[sdev->curr]->next_free;
			sdev->bios[sdev->curr]->next_free = -1;
			sdev->bios[sdev->curr]->count = 0;
			spin_unlock(&sdev->list_lock);
		} else {
			spin_unlock(&sdev->list_lock);
			wait_event(sdev->list_wait, sdev->first_free != -1);
		}
	}
	sbio = sdev->bios[sdev->curr];
	if (sbio->count == 0) {
988 989
		struct bio *bio;

A
Arne Jansen 已提交
990 991
		sbio->physical = physical;
		sbio->logical = logical;
992 993 994 995 996 997 998 999 1000 1001
		bio = bio_alloc(GFP_NOFS, SCRUB_PAGES_PER_BIO);
		if (!bio)
			return -ENOMEM;

		bio->bi_private = sbio;
		bio->bi_end_io = scrub_bio_end_io;
		bio->bi_bdev = sdev->dev->bdev;
		bio->bi_sector = sbio->physical >> 9;
		sbio->err = 0;
		sbio->bio = bio;
1002 1003
	} else if (sbio->physical + sbio->count * PAGE_SIZE != physical ||
		   sbio->logical + sbio->count * PAGE_SIZE != logical) {
1004 1005 1006
		ret = scrub_submit(sdev);
		if (ret)
			return ret;
A
Arne Jansen 已提交
1007 1008 1009 1010 1011 1012
		goto again;
	}
	sbio->spag[sbio->count].flags = flags;
	sbio->spag[sbio->count].generation = gen;
	sbio->spag[sbio->count].have_csum = 0;
	sbio->spag[sbio->count].mirror_num = mirror_num;
1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026

	page = alloc_page(GFP_NOFS);
	if (!page)
		return -ENOMEM;

	ret = bio_add_page(sbio->bio, page, PAGE_SIZE, 0);
	if (!ret) {
		__free_page(page);
		ret = scrub_submit(sdev);
		if (ret)
			return ret;
		goto again;
	}

A
Arne Jansen 已提交
1027 1028 1029 1030 1031
	if (csum) {
		sbio->spag[sbio->count].have_csum = 1;
		memcpy(sbio->spag[sbio->count].csum, csum, sdev->csum_size);
	}
	++sbio->count;
1032 1033 1034 1035 1036 1037 1038
	if (sbio->count == SCRUB_PAGES_PER_BIO || force) {
		int ret;

		ret = scrub_submit(sdev);
		if (ret)
			return ret;
	}
A
Arne Jansen 已提交
1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084

	return 0;
}

static int scrub_find_csum(struct scrub_dev *sdev, u64 logical, u64 len,
			   u8 *csum)
{
	struct btrfs_ordered_sum *sum = NULL;
	int ret = 0;
	unsigned long i;
	unsigned long num_sectors;
	u32 sectorsize = sdev->dev->dev_root->sectorsize;

	while (!list_empty(&sdev->csum_list)) {
		sum = list_first_entry(&sdev->csum_list,
				       struct btrfs_ordered_sum, list);
		if (sum->bytenr > logical)
			return 0;
		if (sum->bytenr + sum->len > logical)
			break;

		++sdev->stat.csum_discards;
		list_del(&sum->list);
		kfree(sum);
		sum = NULL;
	}
	if (!sum)
		return 0;

	num_sectors = sum->len / sectorsize;
	for (i = 0; i < num_sectors; ++i) {
		if (sum->sums[i].bytenr == logical) {
			memcpy(csum, &sum->sums[i].sum, sdev->csum_size);
			ret = 1;
			break;
		}
	}
	if (ret && i == num_sectors - 1) {
		list_del(&sum->list);
		kfree(sum);
	}
	return ret;
}

/* scrub extent tries to collect up to 64 kB for each bio */
static int scrub_extent(struct scrub_dev *sdev, u64 logical, u64 len,
1085
			u64 physical, u64 flags, u64 gen, int mirror_num)
A
Arne Jansen 已提交
1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118
{
	int ret;
	u8 csum[BTRFS_CSUM_SIZE];

	while (len) {
		u64 l = min_t(u64, len, PAGE_SIZE);
		int have_csum = 0;

		if (flags & BTRFS_EXTENT_FLAG_DATA) {
			/* push csums to sbio */
			have_csum = scrub_find_csum(sdev, logical, l, csum);
			if (have_csum == 0)
				++sdev->stat.no_csum;
		}
		ret = scrub_page(sdev, logical, l, physical, flags, gen,
				 mirror_num, have_csum ? csum : NULL, 0);
		if (ret)
			return ret;
		len -= l;
		logical += l;
		physical += l;
	}
	return 0;
}

static noinline_for_stack int scrub_stripe(struct scrub_dev *sdev,
	struct map_lookup *map, int num, u64 base, u64 length)
{
	struct btrfs_path *path;
	struct btrfs_fs_info *fs_info = sdev->dev->dev_root->fs_info;
	struct btrfs_root *root = fs_info->extent_root;
	struct btrfs_root *csum_root = fs_info->csum_root;
	struct btrfs_extent_item *extent;
1119
	struct blk_plug plug;
A
Arne Jansen 已提交
1120 1121 1122 1123 1124 1125 1126 1127 1128 1129
	u64 flags;
	int ret;
	int slot;
	int i;
	u64 nstripes;
	struct extent_buffer *l;
	struct btrfs_key key;
	u64 physical;
	u64 logical;
	u64 generation;
1130
	int mirror_num;
A
Arne Jansen 已提交
1131 1132 1133 1134
	struct reada_control *reada1;
	struct reada_control *reada2;
	struct btrfs_key key_start;
	struct btrfs_key key_end;
A
Arne Jansen 已提交
1135 1136 1137 1138 1139 1140 1141 1142 1143 1144

	u64 increment = map->stripe_len;
	u64 offset;

	nstripes = length;
	offset = 0;
	do_div(nstripes, map->stripe_len);
	if (map->type & BTRFS_BLOCK_GROUP_RAID0) {
		offset = map->stripe_len * num;
		increment = map->stripe_len * map->num_stripes;
1145
		mirror_num = 1;
A
Arne Jansen 已提交
1146 1147 1148 1149
	} else if (map->type & BTRFS_BLOCK_GROUP_RAID10) {
		int factor = map->num_stripes / map->sub_stripes;
		offset = map->stripe_len * (num / map->sub_stripes);
		increment = map->stripe_len * factor;
1150
		mirror_num = num % map->sub_stripes + 1;
A
Arne Jansen 已提交
1151 1152
	} else if (map->type & BTRFS_BLOCK_GROUP_RAID1) {
		increment = map->stripe_len;
1153
		mirror_num = num % map->num_stripes + 1;
A
Arne Jansen 已提交
1154 1155
	} else if (map->type & BTRFS_BLOCK_GROUP_DUP) {
		increment = map->stripe_len;
1156
		mirror_num = num % map->num_stripes + 1;
A
Arne Jansen 已提交
1157 1158
	} else {
		increment = map->stripe_len;
1159
		mirror_num = 1;
A
Arne Jansen 已提交
1160 1161 1162 1163 1164 1165 1166 1167 1168 1169
	}

	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;

	path->search_commit_root = 1;
	path->skip_locking = 1;

	/*
A
Arne Jansen 已提交
1170 1171 1172
	 * trigger the readahead for extent tree csum tree and wait for
	 * completion. During readahead, the scrub is officially paused
	 * to not hold off transaction commits
A
Arne Jansen 已提交
1173 1174 1175
	 */
	logical = base + offset;

A
Arne Jansen 已提交
1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208
	wait_event(sdev->list_wait,
		   atomic_read(&sdev->in_flight) == 0);
	atomic_inc(&fs_info->scrubs_paused);
	wake_up(&fs_info->scrub_pause_wait);

	/* FIXME it might be better to start readahead at commit root */
	key_start.objectid = logical;
	key_start.type = BTRFS_EXTENT_ITEM_KEY;
	key_start.offset = (u64)0;
	key_end.objectid = base + offset + nstripes * increment;
	key_end.type = BTRFS_EXTENT_ITEM_KEY;
	key_end.offset = (u64)0;
	reada1 = btrfs_reada_add(root, &key_start, &key_end);

	key_start.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
	key_start.type = BTRFS_EXTENT_CSUM_KEY;
	key_start.offset = logical;
	key_end.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
	key_end.type = BTRFS_EXTENT_CSUM_KEY;
	key_end.offset = base + offset + nstripes * increment;
	reada2 = btrfs_reada_add(csum_root, &key_start, &key_end);

	if (!IS_ERR(reada1))
		btrfs_reada_wait(reada1);
	if (!IS_ERR(reada2))
		btrfs_reada_wait(reada2);

	mutex_lock(&fs_info->scrub_lock);
	while (atomic_read(&fs_info->scrub_pause_req)) {
		mutex_unlock(&fs_info->scrub_lock);
		wait_event(fs_info->scrub_pause_wait,
		   atomic_read(&fs_info->scrub_pause_req) == 0);
		mutex_lock(&fs_info->scrub_lock);
A
Arne Jansen 已提交
1209
	}
A
Arne Jansen 已提交
1210 1211 1212
	atomic_dec(&fs_info->scrubs_paused);
	mutex_unlock(&fs_info->scrub_lock);
	wake_up(&fs_info->scrub_pause_wait);
A
Arne Jansen 已提交
1213 1214 1215 1216 1217

	/*
	 * collect all data csums for the stripe to avoid seeking during
	 * the scrub. This might currently (crc32) end up to be about 1MB
	 */
1218
	blk_start_plug(&plug);
A
Arne Jansen 已提交
1219 1220 1221 1222

	/*
	 * now find all extents for each stripe and scrub them
	 */
A
Arne Jansen 已提交
1223 1224
	logical = base + offset;
	physical = map->stripes[num].physical;
A
Arne Jansen 已提交
1225
	ret = 0;
A
Arne Jansen 已提交
1226
	for (i = 0; i < nstripes; ++i) {
A
Arne Jansen 已提交
1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256
		/*
		 * canceled?
		 */
		if (atomic_read(&fs_info->scrub_cancel_req) ||
		    atomic_read(&sdev->cancel_req)) {
			ret = -ECANCELED;
			goto out;
		}
		/*
		 * check to see if we have to pause
		 */
		if (atomic_read(&fs_info->scrub_pause_req)) {
			/* push queued extents */
			scrub_submit(sdev);
			wait_event(sdev->list_wait,
				   atomic_read(&sdev->in_flight) == 0);
			atomic_inc(&fs_info->scrubs_paused);
			wake_up(&fs_info->scrub_pause_wait);
			mutex_lock(&fs_info->scrub_lock);
			while (atomic_read(&fs_info->scrub_pause_req)) {
				mutex_unlock(&fs_info->scrub_lock);
				wait_event(fs_info->scrub_pause_wait,
				   atomic_read(&fs_info->scrub_pause_req) == 0);
				mutex_lock(&fs_info->scrub_lock);
			}
			atomic_dec(&fs_info->scrubs_paused);
			mutex_unlock(&fs_info->scrub_lock);
			wake_up(&fs_info->scrub_pause_wait);
		}

A
Arne Jansen 已提交
1257 1258 1259 1260 1261 1262
		ret = btrfs_lookup_csums_range(csum_root, logical,
					       logical + map->stripe_len - 1,
					       &sdev->csum_list, 1);
		if (ret)
			goto out;

A
Arne Jansen 已提交
1263 1264 1265 1266 1267 1268 1269
		key.objectid = logical;
		key.type = BTRFS_EXTENT_ITEM_KEY;
		key.offset = (u64)0;

		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
		if (ret < 0)
			goto out;
1270
		if (ret > 0) {
A
Arne Jansen 已提交
1271 1272 1273 1274
			ret = btrfs_previous_item(root, path, 0,
						  BTRFS_EXTENT_ITEM_KEY);
			if (ret < 0)
				goto out;
1275 1276 1277 1278 1279 1280 1281 1282 1283
			if (ret > 0) {
				/* there's no smaller item, so stick with the
				 * larger one */
				btrfs_release_path(path);
				ret = btrfs_search_slot(NULL, root, &key,
							path, 0, 0);
				if (ret < 0)
					goto out;
			}
A
Arne Jansen 已提交
1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345
		}

		while (1) {
			l = path->nodes[0];
			slot = path->slots[0];
			if (slot >= btrfs_header_nritems(l)) {
				ret = btrfs_next_leaf(root, path);
				if (ret == 0)
					continue;
				if (ret < 0)
					goto out;

				break;
			}
			btrfs_item_key_to_cpu(l, &key, slot);

			if (key.objectid + key.offset <= logical)
				goto next;

			if (key.objectid >= logical + map->stripe_len)
				break;

			if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY)
				goto next;

			extent = btrfs_item_ptr(l, slot,
						struct btrfs_extent_item);
			flags = btrfs_extent_flags(l, extent);
			generation = btrfs_extent_generation(l, extent);

			if (key.objectid < logical &&
			    (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)) {
				printk(KERN_ERR
				       "btrfs scrub: tree block %llu spanning "
				       "stripes, ignored. logical=%llu\n",
				       (unsigned long long)key.objectid,
				       (unsigned long long)logical);
				goto next;
			}

			/*
			 * trim extent to this stripe
			 */
			if (key.objectid < logical) {
				key.offset -= logical - key.objectid;
				key.objectid = logical;
			}
			if (key.objectid + key.offset >
			    logical + map->stripe_len) {
				key.offset = logical + map->stripe_len -
					     key.objectid;
			}

			ret = scrub_extent(sdev, key.objectid, key.offset,
					   key.objectid - logical + physical,
					   flags, generation, mirror_num);
			if (ret)
				goto out;

next:
			path->slots[0]++;
		}
C
Chris Mason 已提交
1346
		btrfs_release_path(path);
A
Arne Jansen 已提交
1347 1348 1349 1350 1351 1352 1353 1354 1355 1356
		logical += increment;
		physical += map->stripe_len;
		spin_lock(&sdev->stat_lock);
		sdev->stat.last_physical = physical;
		spin_unlock(&sdev->stat_lock);
	}
	/* push queued extents */
	scrub_submit(sdev);

out:
1357
	blk_finish_plug(&plug);
A
Arne Jansen 已提交
1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432
	btrfs_free_path(path);
	return ret < 0 ? ret : 0;
}

static noinline_for_stack int scrub_chunk(struct scrub_dev *sdev,
	u64 chunk_tree, u64 chunk_objectid, u64 chunk_offset, u64 length)
{
	struct btrfs_mapping_tree *map_tree =
		&sdev->dev->dev_root->fs_info->mapping_tree;
	struct map_lookup *map;
	struct extent_map *em;
	int i;
	int ret = -EINVAL;

	read_lock(&map_tree->map_tree.lock);
	em = lookup_extent_mapping(&map_tree->map_tree, chunk_offset, 1);
	read_unlock(&map_tree->map_tree.lock);

	if (!em)
		return -EINVAL;

	map = (struct map_lookup *)em->bdev;
	if (em->start != chunk_offset)
		goto out;

	if (em->len < length)
		goto out;

	for (i = 0; i < map->num_stripes; ++i) {
		if (map->stripes[i].dev == sdev->dev) {
			ret = scrub_stripe(sdev, map, i, chunk_offset, length);
			if (ret)
				goto out;
		}
	}
out:
	free_extent_map(em);

	return ret;
}

static noinline_for_stack
int scrub_enumerate_chunks(struct scrub_dev *sdev, u64 start, u64 end)
{
	struct btrfs_dev_extent *dev_extent = NULL;
	struct btrfs_path *path;
	struct btrfs_root *root = sdev->dev->dev_root;
	struct btrfs_fs_info *fs_info = root->fs_info;
	u64 length;
	u64 chunk_tree;
	u64 chunk_objectid;
	u64 chunk_offset;
	int ret;
	int slot;
	struct extent_buffer *l;
	struct btrfs_key key;
	struct btrfs_key found_key;
	struct btrfs_block_group_cache *cache;

	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;

	path->reada = 2;
	path->search_commit_root = 1;
	path->skip_locking = 1;

	key.objectid = sdev->dev->devid;
	key.offset = 0ull;
	key.type = BTRFS_DEV_EXTENT_KEY;


	while (1) {
		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
		if (ret < 0)
1433 1434 1435 1436 1437 1438 1439 1440 1441
			break;
		if (ret > 0) {
			if (path->slots[0] >=
			    btrfs_header_nritems(path->nodes[0])) {
				ret = btrfs_next_leaf(root, path);
				if (ret)
					break;
			}
		}
A
Arne Jansen 已提交
1442 1443 1444 1445 1446 1447 1448 1449 1450

		l = path->nodes[0];
		slot = path->slots[0];

		btrfs_item_key_to_cpu(l, &found_key, slot);

		if (found_key.objectid != sdev->dev->devid)
			break;

1451
		if (btrfs_key_type(&found_key) != BTRFS_DEV_EXTENT_KEY)
A
Arne Jansen 已提交
1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464
			break;

		if (found_key.offset >= end)
			break;

		if (found_key.offset < key.offset)
			break;

		dev_extent = btrfs_item_ptr(l, slot, struct btrfs_dev_extent);
		length = btrfs_dev_extent_length(l, dev_extent);

		if (found_key.offset + length <= start) {
			key.offset = found_key.offset + length;
C
Chris Mason 已提交
1465
			btrfs_release_path(path);
A
Arne Jansen 已提交
1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479
			continue;
		}

		chunk_tree = btrfs_dev_extent_chunk_tree(l, dev_extent);
		chunk_objectid = btrfs_dev_extent_chunk_objectid(l, dev_extent);
		chunk_offset = btrfs_dev_extent_chunk_offset(l, dev_extent);

		/*
		 * get a reference on the corresponding block group to prevent
		 * the chunk from going away while we scrub it
		 */
		cache = btrfs_lookup_block_group(fs_info, chunk_offset);
		if (!cache) {
			ret = -ENOENT;
1480
			break;
A
Arne Jansen 已提交
1481 1482 1483 1484 1485 1486 1487 1488
		}
		ret = scrub_chunk(sdev, chunk_tree, chunk_objectid,
				  chunk_offset, length);
		btrfs_put_block_group(cache);
		if (ret)
			break;

		key.offset = found_key.offset + length;
C
Chris Mason 已提交
1489
		btrfs_release_path(path);
A
Arne Jansen 已提交
1490 1491 1492
	}

	btrfs_free_path(path);
1493 1494 1495 1496 1497 1498

	/*
	 * ret can still be 1 from search_slot or next_leaf,
	 * that's not an error
	 */
	return ret < 0 ? ret : 0;
A
Arne Jansen 已提交
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
}

static noinline_for_stack int scrub_supers(struct scrub_dev *sdev)
{
	int	i;
	u64	bytenr;
	u64	gen;
	int	ret;
	struct btrfs_device *device = sdev->dev;
	struct btrfs_root *root = device->dev_root;

	gen = root->fs_info->last_trans_committed;

	for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
		bytenr = btrfs_sb_offset(i);
		if (bytenr + BTRFS_SUPER_INFO_SIZE >= device->total_bytes)
			break;

		ret = scrub_page(sdev, bytenr, PAGE_SIZE, bytenr,
				 BTRFS_EXTENT_FLAG_SUPER, gen, i, NULL, 1);
		if (ret)
			return ret;
	}
	wait_event(sdev->list_wait, atomic_read(&sdev->in_flight) == 0);

	return 0;
}

/*
 * get a reference count on fs_info->scrub_workers. start worker if necessary
 */
static noinline_for_stack int scrub_workers_get(struct btrfs_root *root)
{
	struct btrfs_fs_info *fs_info = root->fs_info;

	mutex_lock(&fs_info->scrub_lock);
A
Arne Jansen 已提交
1535 1536 1537 1538
	if (fs_info->scrub_workers_refcnt == 0) {
		btrfs_init_workers(&fs_info->scrub_workers, "scrub",
			   fs_info->thread_pool_size, &fs_info->generic_worker);
		fs_info->scrub_workers.idle_thresh = 4;
A
Arne Jansen 已提交
1539
		btrfs_start_workers(&fs_info->scrub_workers, 1);
A
Arne Jansen 已提交
1540
	}
A
Arne Jansen 已提交
1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559
	++fs_info->scrub_workers_refcnt;
	mutex_unlock(&fs_info->scrub_lock);

	return 0;
}

static noinline_for_stack void scrub_workers_put(struct btrfs_root *root)
{
	struct btrfs_fs_info *fs_info = root->fs_info;

	mutex_lock(&fs_info->scrub_lock);
	if (--fs_info->scrub_workers_refcnt == 0)
		btrfs_stop_workers(&fs_info->scrub_workers);
	WARN_ON(fs_info->scrub_workers_refcnt < 0);
	mutex_unlock(&fs_info->scrub_lock);
}


int btrfs_scrub_dev(struct btrfs_root *root, u64 devid, u64 start, u64 end,
A
Arne Jansen 已提交
1560
		    struct btrfs_scrub_progress *progress, int readonly)
A
Arne Jansen 已提交
1561 1562 1563 1564 1565 1566
{
	struct scrub_dev *sdev;
	struct btrfs_fs_info *fs_info = root->fs_info;
	int ret;
	struct btrfs_device *dev;

1567
	if (btrfs_fs_closing(root->fs_info))
A
Arne Jansen 已提交
1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612
		return -EINVAL;

	/*
	 * check some assumptions
	 */
	if (root->sectorsize != PAGE_SIZE ||
	    root->sectorsize != root->leafsize ||
	    root->sectorsize != root->nodesize) {
		printk(KERN_ERR "btrfs_scrub: size assumptions fail\n");
		return -EINVAL;
	}

	ret = scrub_workers_get(root);
	if (ret)
		return ret;

	mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
	dev = btrfs_find_device(root, devid, NULL, NULL);
	if (!dev || dev->missing) {
		mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
		scrub_workers_put(root);
		return -ENODEV;
	}
	mutex_lock(&fs_info->scrub_lock);

	if (!dev->in_fs_metadata) {
		mutex_unlock(&fs_info->scrub_lock);
		mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
		scrub_workers_put(root);
		return -ENODEV;
	}

	if (dev->scrub_device) {
		mutex_unlock(&fs_info->scrub_lock);
		mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
		scrub_workers_put(root);
		return -EINPROGRESS;
	}
	sdev = scrub_setup_dev(dev);
	if (IS_ERR(sdev)) {
		mutex_unlock(&fs_info->scrub_lock);
		mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
		scrub_workers_put(root);
		return PTR_ERR(sdev);
	}
A
Arne Jansen 已提交
1613
	sdev->readonly = readonly;
A
Arne Jansen 已提交
1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630
	dev->scrub_device = sdev;

	atomic_inc(&fs_info->scrubs_running);
	mutex_unlock(&fs_info->scrub_lock);
	mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);

	down_read(&fs_info->scrub_super_lock);
	ret = scrub_supers(sdev);
	up_read(&fs_info->scrub_super_lock);

	if (!ret)
		ret = scrub_enumerate_chunks(sdev, start, end);

	wait_event(sdev->list_wait, atomic_read(&sdev->in_flight) == 0);
	atomic_dec(&fs_info->scrubs_running);
	wake_up(&fs_info->scrub_pause_wait);

1631 1632
	wait_event(sdev->list_wait, atomic_read(&sdev->fixup_cnt) == 0);

A
Arne Jansen 已提交
1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768
	if (progress)
		memcpy(progress, &sdev->stat, sizeof(*progress));

	mutex_lock(&fs_info->scrub_lock);
	dev->scrub_device = NULL;
	mutex_unlock(&fs_info->scrub_lock);

	scrub_free_dev(sdev);
	scrub_workers_put(root);

	return ret;
}

int btrfs_scrub_pause(struct btrfs_root *root)
{
	struct btrfs_fs_info *fs_info = root->fs_info;

	mutex_lock(&fs_info->scrub_lock);
	atomic_inc(&fs_info->scrub_pause_req);
	while (atomic_read(&fs_info->scrubs_paused) !=
	       atomic_read(&fs_info->scrubs_running)) {
		mutex_unlock(&fs_info->scrub_lock);
		wait_event(fs_info->scrub_pause_wait,
			   atomic_read(&fs_info->scrubs_paused) ==
			   atomic_read(&fs_info->scrubs_running));
		mutex_lock(&fs_info->scrub_lock);
	}
	mutex_unlock(&fs_info->scrub_lock);

	return 0;
}

int btrfs_scrub_continue(struct btrfs_root *root)
{
	struct btrfs_fs_info *fs_info = root->fs_info;

	atomic_dec(&fs_info->scrub_pause_req);
	wake_up(&fs_info->scrub_pause_wait);
	return 0;
}

int btrfs_scrub_pause_super(struct btrfs_root *root)
{
	down_write(&root->fs_info->scrub_super_lock);
	return 0;
}

int btrfs_scrub_continue_super(struct btrfs_root *root)
{
	up_write(&root->fs_info->scrub_super_lock);
	return 0;
}

int btrfs_scrub_cancel(struct btrfs_root *root)
{
	struct btrfs_fs_info *fs_info = root->fs_info;

	mutex_lock(&fs_info->scrub_lock);
	if (!atomic_read(&fs_info->scrubs_running)) {
		mutex_unlock(&fs_info->scrub_lock);
		return -ENOTCONN;
	}

	atomic_inc(&fs_info->scrub_cancel_req);
	while (atomic_read(&fs_info->scrubs_running)) {
		mutex_unlock(&fs_info->scrub_lock);
		wait_event(fs_info->scrub_pause_wait,
			   atomic_read(&fs_info->scrubs_running) == 0);
		mutex_lock(&fs_info->scrub_lock);
	}
	atomic_dec(&fs_info->scrub_cancel_req);
	mutex_unlock(&fs_info->scrub_lock);

	return 0;
}

int btrfs_scrub_cancel_dev(struct btrfs_root *root, struct btrfs_device *dev)
{
	struct btrfs_fs_info *fs_info = root->fs_info;
	struct scrub_dev *sdev;

	mutex_lock(&fs_info->scrub_lock);
	sdev = dev->scrub_device;
	if (!sdev) {
		mutex_unlock(&fs_info->scrub_lock);
		return -ENOTCONN;
	}
	atomic_inc(&sdev->cancel_req);
	while (dev->scrub_device) {
		mutex_unlock(&fs_info->scrub_lock);
		wait_event(fs_info->scrub_pause_wait,
			   dev->scrub_device == NULL);
		mutex_lock(&fs_info->scrub_lock);
	}
	mutex_unlock(&fs_info->scrub_lock);

	return 0;
}
int btrfs_scrub_cancel_devid(struct btrfs_root *root, u64 devid)
{
	struct btrfs_fs_info *fs_info = root->fs_info;
	struct btrfs_device *dev;
	int ret;

	/*
	 * we have to hold the device_list_mutex here so the device
	 * does not go away in cancel_dev. FIXME: find a better solution
	 */
	mutex_lock(&fs_info->fs_devices->device_list_mutex);
	dev = btrfs_find_device(root, devid, NULL, NULL);
	if (!dev) {
		mutex_unlock(&fs_info->fs_devices->device_list_mutex);
		return -ENODEV;
	}
	ret = btrfs_scrub_cancel_dev(root, dev);
	mutex_unlock(&fs_info->fs_devices->device_list_mutex);

	return ret;
}

int btrfs_scrub_progress(struct btrfs_root *root, u64 devid,
			 struct btrfs_scrub_progress *progress)
{
	struct btrfs_device *dev;
	struct scrub_dev *sdev = NULL;

	mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
	dev = btrfs_find_device(root, devid, NULL, NULL);
	if (dev)
		sdev = dev->scrub_device;
	if (sdev)
		memcpy(progress, &sdev->stat, sizeof(*progress));
	mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);

	return dev ? (sdev ? 0 : -ENOTCONN) : -ENODEV;
}