scrub.c 113.9 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0
A
Arne Jansen 已提交
2
/*
3
 * Copyright (C) 2011, 2012 STRATO.  All rights reserved.
A
Arne Jansen 已提交
4 5 6
 */

#include <linux/blkdev.h>
7
#include <linux/ratelimit.h>
8
#include <linux/sched/mm.h>
9
#include <crypto/hash.h>
A
Arne Jansen 已提交
10
#include "ctree.h"
11
#include "discard.h"
A
Arne Jansen 已提交
12 13 14
#include "volumes.h"
#include "disk-io.h"
#include "ordered-data.h"
15
#include "transaction.h"
16
#include "backref.h"
17
#include "extent_io.h"
18
#include "dev-replace.h"
19
#include "check-integrity.h"
20
#include "rcu-string.h"
D
David Woodhouse 已提交
21
#include "raid56.h"
22
#include "block-group.h"
23
#include "zoned.h"
A
Arne Jansen 已提交
24 25 26 27 28 29 30 31 32 33 34 35 36 37

/*
 * 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
 *  - track and record media errors, throw out bad devices
 *  - add a mode to also read unallocated space
 */

38
struct scrub_block;
39
struct scrub_ctx;
A
Arne Jansen 已提交
40

41
/*
42 43
 * The following three values only influence the performance.
 *
44
 * The last one configures the number of parallel and outstanding I/O
45
 * operations. The first one configures an upper limit for the number
46 47
 * of (dynamically allocated) pages that are added to a bio.
 */
48 49
#define SCRUB_PAGES_PER_BIO	32	/* 128KiB per bio for x86 */
#define SCRUB_BIOS_PER_SCTX	64	/* 8MiB per device in flight for x86 */
50 51

/*
52
 * The following value times PAGE_SIZE needs to be large enough to match the
53 54
 * largest node/leaf/sector size that shall be supported.
 */
55
#define SCRUB_MAX_PAGES_PER_BLOCK	(BTRFS_MAX_METADATA_BLOCKSIZE / SZ_4K)
A
Arne Jansen 已提交
56

57
struct scrub_recover {
58
	refcount_t		refs;
59
	struct btrfs_io_context	*bioc;
60 61 62
	u64			map_length;
};

A
Arne Jansen 已提交
63
struct scrub_page {
64 65
	struct scrub_block	*sblock;
	struct page		*page;
66
	struct btrfs_device	*dev;
67
	struct list_head	list;
A
Arne Jansen 已提交
68 69
	u64			flags;  /* extent flags */
	u64			generation;
70 71
	u64			logical;
	u64			physical;
72
	u64			physical_for_dev_replace;
73
	atomic_t		refs;
74
	u8			mirror_num;
75 76
	unsigned int		have_csum:1;
	unsigned int		io_error:1;
A
Arne Jansen 已提交
77
	u8			csum[BTRFS_CSUM_SIZE];
78 79

	struct scrub_recover	*recover;
A
Arne Jansen 已提交
80 81 82 83
};

struct scrub_bio {
	int			index;
84
	struct scrub_ctx	*sctx;
85
	struct btrfs_device	*dev;
A
Arne Jansen 已提交
86
	struct bio		*bio;
87
	blk_status_t		status;
A
Arne Jansen 已提交
88 89
	u64			logical;
	u64			physical;
90
	struct scrub_page	*pagev[SCRUB_PAGES_PER_BIO];
91
	int			page_count;
A
Arne Jansen 已提交
92 93 94 95
	int			next_free;
	struct btrfs_work	work;
};

96
struct scrub_block {
97
	struct scrub_page	*pagev[SCRUB_MAX_PAGES_PER_BLOCK];
98 99
	int			page_count;
	atomic_t		outstanding_pages;
100
	refcount_t		refs; /* free mem on transition to zero */
101
	struct scrub_ctx	*sctx;
102
	struct scrub_parity	*sparity;
103 104 105 106
	struct {
		unsigned int	header_error:1;
		unsigned int	checksum_error:1;
		unsigned int	no_io_error_seen:1;
107
		unsigned int	generation_error:1; /* also sets header_error */
108 109 110 111

		/* The following is for the data used to check parity */
		/* It is for the data with checksum */
		unsigned int	data_corrected:1;
112
	};
113
	struct btrfs_work	work;
114 115
};

116 117 118 119 120 121 122 123 124 125 126 127
/* Used for the chunks with parity stripe such RAID5/6 */
struct scrub_parity {
	struct scrub_ctx	*sctx;

	struct btrfs_device	*scrub_dev;

	u64			logic_start;

	u64			logic_end;

	int			nsectors;

128
	u32			stripe_len;
129

130
	refcount_t		refs;
131 132 133 134 135 136 137 138 139 140 141 142 143 144 145

	struct list_head	spages;

	/* Work of parity check and repair */
	struct btrfs_work	work;

	/* Mark the parity blocks which have data */
	unsigned long		*dbitmap;

	/*
	 * Mark the parity blocks which have data, but errors happen when
	 * read data or check data
	 */
	unsigned long		*ebitmap;

146
	unsigned long		bitmap[];
147 148
};

149
struct scrub_ctx {
150
	struct scrub_bio	*bios[SCRUB_BIOS_PER_SCTX];
151
	struct btrfs_fs_info	*fs_info;
A
Arne Jansen 已提交
152 153
	int			first_free;
	int			curr;
154 155
	atomic_t		bios_in_flight;
	atomic_t		workers_pending;
A
Arne Jansen 已提交
156 157 158 159
	spinlock_t		list_lock;
	wait_queue_head_t	list_wait;
	struct list_head	csum_list;
	atomic_t		cancel_req;
A
Arne Jansen 已提交
160
	int			readonly;
161
	int			pages_per_bio;
162

163 164 165 166
	/* State of IO submission throttling affecting the associated device */
	ktime_t			throttle_deadline;
	u64			throttle_sent;

167
	int			is_dev_replace;
168
	u64			write_pointer;
169 170 171 172

	struct scrub_bio        *wr_curr_bio;
	struct mutex            wr_lock;
	struct btrfs_device     *wr_tgtdev;
173
	bool                    flush_all_writes;
174

A
Arne Jansen 已提交
175 176 177 178 179
	/*
	 * statistics
	 */
	struct btrfs_scrub_progress stat;
	spinlock_t		stat_lock;
180 181 182 183 184 185 186 187

	/*
	 * Use a ref counter to avoid use-after-free issues. Scrub workers
	 * decrement bios_in_flight and workers_pending and then do a wakeup
	 * on the list_wait wait queue. We must ensure the main scrub task
	 * doesn't free the scrub context before or while the workers are
	 * doing the wakeup() call.
	 */
188
	refcount_t              refs;
A
Arne Jansen 已提交
189 190
};

191 192 193 194
struct scrub_warning {
	struct btrfs_path	*path;
	u64			extent_item_size;
	const char		*errstr;
D
David Sterba 已提交
195
	u64			physical;
196 197 198 199
	u64			logical;
	struct btrfs_device	*dev;
};

200 201 202 203 204 205 206
struct full_stripe_lock {
	struct rb_node node;
	u64 logical;
	u64 refs;
	struct mutex mutex;
};

207
static int scrub_setup_recheck_block(struct scrub_block *original_sblock,
208
				     struct scrub_block *sblocks_for_recheck);
209
static void scrub_recheck_block(struct btrfs_fs_info *fs_info,
210 211
				struct scrub_block *sblock,
				int retry_failed_mirror);
212
static void scrub_recheck_block_checksum(struct scrub_block *sblock);
213
static int scrub_repair_block_from_good_copy(struct scrub_block *sblock_bad,
214
					     struct scrub_block *sblock_good);
215 216 217
static int scrub_repair_page_from_good_copy(struct scrub_block *sblock_bad,
					    struct scrub_block *sblock_good,
					    int page_num, int force_write);
218 219 220
static void scrub_write_block_to_dev_replace(struct scrub_block *sblock);
static int scrub_write_page_to_dev_replace(struct scrub_block *sblock,
					   int page_num);
221 222 223 224
static int scrub_checksum_data(struct scrub_block *sblock);
static int scrub_checksum_tree_block(struct scrub_block *sblock);
static int scrub_checksum_super(struct scrub_block *sblock);
static void scrub_block_put(struct scrub_block *sblock);
225 226
static void scrub_page_get(struct scrub_page *spage);
static void scrub_page_put(struct scrub_page *spage);
227 228
static void scrub_parity_get(struct scrub_parity *sparity);
static void scrub_parity_put(struct scrub_parity *sparity);
229
static int scrub_pages(struct scrub_ctx *sctx, u64 logical, u32 len,
230
		       u64 physical, struct btrfs_device *dev, u64 flags,
231
		       u64 gen, int mirror_num, u8 *csum,
232
		       u64 physical_for_dev_replace);
233
static void scrub_bio_end_io(struct bio *bio);
234 235
static void scrub_bio_end_io_worker(struct btrfs_work *work);
static void scrub_block_complete(struct scrub_block *sblock);
236
static void scrub_remap_extent(struct btrfs_fs_info *fs_info,
237
			       u64 extent_logical, u32 extent_len,
238 239 240 241 242 243
			       u64 *extent_physical,
			       struct btrfs_device **extent_dev,
			       int *extent_mirror_num);
static int scrub_add_page_to_wr_bio(struct scrub_ctx *sctx,
				    struct scrub_page *spage);
static void scrub_wr_submit(struct scrub_ctx *sctx);
244
static void scrub_wr_bio_end_io(struct bio *bio);
245
static void scrub_wr_bio_end_io_worker(struct btrfs_work *work);
246
static void scrub_put_ctx(struct scrub_ctx *sctx);
S
Stefan Behrens 已提交
247

248
static inline int scrub_is_page_on_raid56(struct scrub_page *spage)
249
{
250
	return spage->recover &&
251
	       (spage->recover->bioc->map_type & BTRFS_BLOCK_GROUP_RAID56_MASK);
252
}
S
Stefan Behrens 已提交
253

254 255
static void scrub_pending_bio_inc(struct scrub_ctx *sctx)
{
256
	refcount_inc(&sctx->refs);
257 258 259 260 261 262 263
	atomic_inc(&sctx->bios_in_flight);
}

static void scrub_pending_bio_dec(struct scrub_ctx *sctx)
{
	atomic_dec(&sctx->bios_in_flight);
	wake_up(&sctx->list_wait);
264
	scrub_put_ctx(sctx);
265 266
}

267
static void __scrub_blocked_if_needed(struct btrfs_fs_info *fs_info)
268 269 270 271 272 273 274 275 276
{
	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);
	}
}

277
static void scrub_pause_on(struct btrfs_fs_info *fs_info)
278 279 280
{
	atomic_inc(&fs_info->scrubs_paused);
	wake_up(&fs_info->scrub_pause_wait);
281
}
282

283 284
static void scrub_pause_off(struct btrfs_fs_info *fs_info)
{
285 286 287 288 289 290 291 292
	mutex_lock(&fs_info->scrub_lock);
	__scrub_blocked_if_needed(fs_info);
	atomic_dec(&fs_info->scrubs_paused);
	mutex_unlock(&fs_info->scrub_lock);

	wake_up(&fs_info->scrub_pause_wait);
}

293 294 295 296 297 298
static void scrub_blocked_if_needed(struct btrfs_fs_info *fs_info)
{
	scrub_pause_on(fs_info);
	scrub_pause_off(fs_info);
}

299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317
/*
 * Insert new full stripe lock into full stripe locks tree
 *
 * Return pointer to existing or newly inserted full_stripe_lock structure if
 * everything works well.
 * Return ERR_PTR(-ENOMEM) if we failed to allocate memory
 *
 * NOTE: caller must hold full_stripe_locks_root->lock before calling this
 * function
 */
static struct full_stripe_lock *insert_full_stripe_lock(
		struct btrfs_full_stripe_locks_tree *locks_root,
		u64 fstripe_logical)
{
	struct rb_node **p;
	struct rb_node *parent = NULL;
	struct full_stripe_lock *entry;
	struct full_stripe_lock *ret;

318
	lockdep_assert_held(&locks_root->lock);
319 320 321 322 323 324 325 326 327 328 329 330 331 332 333

	p = &locks_root->root.rb_node;
	while (*p) {
		parent = *p;
		entry = rb_entry(parent, struct full_stripe_lock, node);
		if (fstripe_logical < entry->logical) {
			p = &(*p)->rb_left;
		} else if (fstripe_logical > entry->logical) {
			p = &(*p)->rb_right;
		} else {
			entry->refs++;
			return entry;
		}
	}

334 335 336
	/*
	 * Insert new lock.
	 */
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
	ret = kmalloc(sizeof(*ret), GFP_KERNEL);
	if (!ret)
		return ERR_PTR(-ENOMEM);
	ret->logical = fstripe_logical;
	ret->refs = 1;
	mutex_init(&ret->mutex);

	rb_link_node(&ret->node, parent, p);
	rb_insert_color(&ret->node, &locks_root->root);
	return ret;
}

/*
 * Search for a full stripe lock of a block group
 *
 * Return pointer to existing full stripe lock if found
 * Return NULL if not found
 */
static struct full_stripe_lock *search_full_stripe_lock(
		struct btrfs_full_stripe_locks_tree *locks_root,
		u64 fstripe_logical)
{
	struct rb_node *node;
	struct full_stripe_lock *entry;

362
	lockdep_assert_held(&locks_root->lock);
363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381

	node = locks_root->root.rb_node;
	while (node) {
		entry = rb_entry(node, struct full_stripe_lock, node);
		if (fstripe_logical < entry->logical)
			node = node->rb_left;
		else if (fstripe_logical > entry->logical)
			node = node->rb_right;
		else
			return entry;
	}
	return NULL;
}

/*
 * Helper to get full stripe logical from a normal bytenr.
 *
 * Caller must ensure @cache is a RAID56 block group.
 */
382
static u64 get_full_stripe_logical(struct btrfs_block_group *cache, u64 bytenr)
383 384 385 386 387 388 389 390 391 392 393 394 395
{
	u64 ret;

	/*
	 * Due to chunk item size limit, full stripe length should not be
	 * larger than U32_MAX. Just a sanity check here.
	 */
	WARN_ON_ONCE(cache->full_stripe_len >= U32_MAX);

	/*
	 * round_down() can only handle power of 2, while RAID56 full
	 * stripe length can be 64KiB * n, so we need to manually round down.
	 */
396 397
	ret = div64_u64(bytenr - cache->start, cache->full_stripe_len) *
			cache->full_stripe_len + cache->start;
398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414
	return ret;
}

/*
 * Lock a full stripe to avoid concurrency of recovery and read
 *
 * It's only used for profiles with parities (RAID5/6), for other profiles it
 * does nothing.
 *
 * Return 0 if we locked full stripe covering @bytenr, with a mutex held.
 * So caller must call unlock_full_stripe() at the same context.
 *
 * Return <0 if encounters error.
 */
static int lock_full_stripe(struct btrfs_fs_info *fs_info, u64 bytenr,
			    bool *locked_ret)
{
415
	struct btrfs_block_group *bg_cache;
416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461
	struct btrfs_full_stripe_locks_tree *locks_root;
	struct full_stripe_lock *existing;
	u64 fstripe_start;
	int ret = 0;

	*locked_ret = false;
	bg_cache = btrfs_lookup_block_group(fs_info, bytenr);
	if (!bg_cache) {
		ASSERT(0);
		return -ENOENT;
	}

	/* Profiles not based on parity don't need full stripe lock */
	if (!(bg_cache->flags & BTRFS_BLOCK_GROUP_RAID56_MASK))
		goto out;
	locks_root = &bg_cache->full_stripe_locks_root;

	fstripe_start = get_full_stripe_logical(bg_cache, bytenr);

	/* Now insert the full stripe lock */
	mutex_lock(&locks_root->lock);
	existing = insert_full_stripe_lock(locks_root, fstripe_start);
	mutex_unlock(&locks_root->lock);
	if (IS_ERR(existing)) {
		ret = PTR_ERR(existing);
		goto out;
	}
	mutex_lock(&existing->mutex);
	*locked_ret = true;
out:
	btrfs_put_block_group(bg_cache);
	return ret;
}

/*
 * Unlock a full stripe.
 *
 * NOTE: Caller must ensure it's the same context calling corresponding
 * lock_full_stripe().
 *
 * Return 0 if we unlock full stripe without problem.
 * Return <0 for error
 */
static int unlock_full_stripe(struct btrfs_fs_info *fs_info, u64 bytenr,
			      bool locked)
{
462
	struct btrfs_block_group *bg_cache;
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
	struct btrfs_full_stripe_locks_tree *locks_root;
	struct full_stripe_lock *fstripe_lock;
	u64 fstripe_start;
	bool freeit = false;
	int ret = 0;

	/* If we didn't acquire full stripe lock, no need to continue */
	if (!locked)
		return 0;

	bg_cache = btrfs_lookup_block_group(fs_info, bytenr);
	if (!bg_cache) {
		ASSERT(0);
		return -ENOENT;
	}
	if (!(bg_cache->flags & BTRFS_BLOCK_GROUP_RAID56_MASK))
		goto out;

	locks_root = &bg_cache->full_stripe_locks_root;
	fstripe_start = get_full_stripe_logical(bg_cache, bytenr);

	mutex_lock(&locks_root->lock);
	fstripe_lock = search_full_stripe_lock(locks_root, fstripe_start);
	/* Unpaired unlock_full_stripe() detected */
	if (!fstripe_lock) {
		WARN_ON(1);
		ret = -ENOENT;
		mutex_unlock(&locks_root->lock);
		goto out;
	}

	if (fstripe_lock->refs == 0) {
		WARN_ON(1);
		btrfs_warn(fs_info, "full stripe lock at %llu refcount underflow",
			fstripe_lock->logical);
	} else {
		fstripe_lock->refs--;
	}

	if (fstripe_lock->refs == 0) {
		rb_erase(&fstripe_lock->node, &locks_root->root);
		freeit = true;
	}
	mutex_unlock(&locks_root->lock);

	mutex_unlock(&fstripe_lock->mutex);
	if (freeit)
		kfree(fstripe_lock);
out:
	btrfs_put_block_group(bg_cache);
	return ret;
}

516
static void scrub_free_csums(struct scrub_ctx *sctx)
A
Arne Jansen 已提交
517
{
518
	while (!list_empty(&sctx->csum_list)) {
A
Arne Jansen 已提交
519
		struct btrfs_ordered_sum *sum;
520
		sum = list_first_entry(&sctx->csum_list,
A
Arne Jansen 已提交
521 522 523 524 525 526
				       struct btrfs_ordered_sum, list);
		list_del(&sum->list);
		kfree(sum);
	}
}

527
static noinline_for_stack void scrub_free_ctx(struct scrub_ctx *sctx)
A
Arne Jansen 已提交
528 529 530
{
	int i;

531
	if (!sctx)
A
Arne Jansen 已提交
532 533
		return;

534
	/* this can happen when scrub is cancelled */
535 536
	if (sctx->curr != -1) {
		struct scrub_bio *sbio = sctx->bios[sctx->curr];
537 538

		for (i = 0; i < sbio->page_count; i++) {
539
			WARN_ON(!sbio->pagev[i]->page);
540 541 542 543 544
			scrub_block_put(sbio->pagev[i]->sblock);
		}
		bio_put(sbio->bio);
	}

545
	for (i = 0; i < SCRUB_BIOS_PER_SCTX; ++i) {
546
		struct scrub_bio *sbio = sctx->bios[i];
A
Arne Jansen 已提交
547 548 549 550 551 552

		if (!sbio)
			break;
		kfree(sbio);
	}

553
	kfree(sctx->wr_curr_bio);
554 555
	scrub_free_csums(sctx);
	kfree(sctx);
A
Arne Jansen 已提交
556 557
}

558 559
static void scrub_put_ctx(struct scrub_ctx *sctx)
{
560
	if (refcount_dec_and_test(&sctx->refs))
561 562 563
		scrub_free_ctx(sctx);
}

564 565
static noinline_for_stack struct scrub_ctx *scrub_setup_ctx(
		struct btrfs_fs_info *fs_info, int is_dev_replace)
A
Arne Jansen 已提交
566
{
567
	struct scrub_ctx *sctx;
A
Arne Jansen 已提交
568 569
	int		i;

570
	sctx = kzalloc(sizeof(*sctx), GFP_KERNEL);
571
	if (!sctx)
A
Arne Jansen 已提交
572
		goto nomem;
573
	refcount_set(&sctx->refs, 1);
574
	sctx->is_dev_replace = is_dev_replace;
575
	sctx->pages_per_bio = SCRUB_PAGES_PER_BIO;
576
	sctx->curr = -1;
577
	sctx->fs_info = fs_info;
578
	INIT_LIST_HEAD(&sctx->csum_list);
579
	for (i = 0; i < SCRUB_BIOS_PER_SCTX; ++i) {
A
Arne Jansen 已提交
580 581
		struct scrub_bio *sbio;

582
		sbio = kzalloc(sizeof(*sbio), GFP_KERNEL);
A
Arne Jansen 已提交
583 584
		if (!sbio)
			goto nomem;
585
		sctx->bios[i] = sbio;
A
Arne Jansen 已提交
586 587

		sbio->index = i;
588
		sbio->sctx = sctx;
589
		sbio->page_count = 0;
590 591
		btrfs_init_work(&sbio->work, scrub_bio_end_io_worker, NULL,
				NULL);
A
Arne Jansen 已提交
592

593
		if (i != SCRUB_BIOS_PER_SCTX - 1)
594
			sctx->bios[i]->next_free = i + 1;
595
		else
596 597 598
			sctx->bios[i]->next_free = -1;
	}
	sctx->first_free = 0;
599 600
	atomic_set(&sctx->bios_in_flight, 0);
	atomic_set(&sctx->workers_pending, 0);
601 602 603 604 605
	atomic_set(&sctx->cancel_req, 0);

	spin_lock_init(&sctx->list_lock);
	spin_lock_init(&sctx->stat_lock);
	init_waitqueue_head(&sctx->list_wait);
606
	sctx->throttle_deadline = 0;
607

608 609 610
	WARN_ON(sctx->wr_curr_bio != NULL);
	mutex_init(&sctx->wr_lock);
	sctx->wr_curr_bio = NULL;
611
	if (is_dev_replace) {
612 613
		WARN_ON(!fs_info->dev_replace.tgtdev);
		sctx->wr_tgtdev = fs_info->dev_replace.tgtdev;
614
		sctx->flush_all_writes = false;
615
	}
616

617
	return sctx;
A
Arne Jansen 已提交
618 619

nomem:
620
	scrub_free_ctx(sctx);
A
Arne Jansen 已提交
621 622 623
	return ERR_PTR(-ENOMEM);
}

624 625
static int scrub_print_warning_inode(u64 inum, u64 offset, u64 root,
				     void *warn_ctx)
626 627 628 629
{
	u32 nlink;
	int ret;
	int i;
630
	unsigned nofs_flag;
631 632
	struct extent_buffer *eb;
	struct btrfs_inode_item *inode_item;
633
	struct scrub_warning *swarn = warn_ctx;
634
	struct btrfs_fs_info *fs_info = swarn->dev->fs_info;
635 636
	struct inode_fs_paths *ipath = NULL;
	struct btrfs_root *local_root;
637
	struct btrfs_key key;
638

D
David Sterba 已提交
639
	local_root = btrfs_get_fs_root(fs_info, root, true);
640 641 642 643 644
	if (IS_ERR(local_root)) {
		ret = PTR_ERR(local_root);
		goto err;
	}

645 646 647
	/*
	 * this makes the path point to (inum INODE_ITEM ioff)
	 */
648 649 650 651 652
	key.objectid = inum;
	key.type = BTRFS_INODE_ITEM_KEY;
	key.offset = 0;

	ret = btrfs_search_slot(NULL, local_root, &key, swarn->path, 0, 0);
653
	if (ret) {
654
		btrfs_put_root(local_root);
655 656 657 658 659 660 661 662 663 664
		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);
	nlink = btrfs_inode_nlink(eb, inode_item);
	btrfs_release_path(swarn->path);

665 666 667 668 669 670
	/*
	 * init_path might indirectly call vmalloc, or use GFP_KERNEL. Scrub
	 * uses GFP_NOFS in this context, so we keep it consistent but it does
	 * not seem to be strictly necessary.
	 */
	nofs_flag = memalloc_nofs_save();
671
	ipath = init_ipath(4096, local_root, swarn->path);
672
	memalloc_nofs_restore(nofs_flag);
673
	if (IS_ERR(ipath)) {
674
		btrfs_put_root(local_root);
675 676 677 678
		ret = PTR_ERR(ipath);
		ipath = NULL;
		goto err;
	}
679 680 681 682 683 684 685 686 687 688
	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)
J
Jeff Mahoney 已提交
689
		btrfs_warn_in_rcu(fs_info,
690
"%s at logical %llu on dev %s, physical %llu, root %llu, inode %llu, offset %llu, length %u, links %u (path: %s)",
J
Jeff Mahoney 已提交
691 692
				  swarn->errstr, swarn->logical,
				  rcu_str_deref(swarn->dev->name),
D
David Sterba 已提交
693
				  swarn->physical,
J
Jeff Mahoney 已提交
694
				  root, inum, offset,
695
				  fs_info->sectorsize, nlink,
J
Jeff Mahoney 已提交
696
				  (char *)(unsigned long)ipath->fspath->val[i]);
697

698
	btrfs_put_root(local_root);
699 700 701 702
	free_ipath(ipath);
	return 0;

err:
J
Jeff Mahoney 已提交
703
	btrfs_warn_in_rcu(fs_info,
D
David Sterba 已提交
704
			  "%s at logical %llu on dev %s, physical %llu, root %llu, inode %llu, offset %llu: path resolving failed with ret=%d",
J
Jeff Mahoney 已提交
705 706
			  swarn->errstr, swarn->logical,
			  rcu_str_deref(swarn->dev->name),
D
David Sterba 已提交
707
			  swarn->physical,
J
Jeff Mahoney 已提交
708
			  root, inum, offset, ret);
709 710 711 712 713

	free_ipath(ipath);
	return 0;
}

714
static void scrub_print_warning(const char *errstr, struct scrub_block *sblock)
715
{
716 717
	struct btrfs_device *dev;
	struct btrfs_fs_info *fs_info;
718 719 720 721 722
	struct btrfs_path *path;
	struct btrfs_key found_key;
	struct extent_buffer *eb;
	struct btrfs_extent_item *ei;
	struct scrub_warning swarn;
723 724 725
	unsigned long ptr = 0;
	u64 extent_item_pos;
	u64 flags = 0;
726
	u64 ref_root;
727
	u32 item_size;
728
	u8 ref_level = 0;
729
	int ret;
730

731
	WARN_ON(sblock->page_count < 1);
732
	dev = sblock->pagev[0]->dev;
733
	fs_info = sblock->sctx->fs_info;
734

735
	path = btrfs_alloc_path();
736 737
	if (!path)
		return;
738

D
David Sterba 已提交
739
	swarn.physical = sblock->pagev[0]->physical;
740
	swarn.logical = sblock->pagev[0]->logical;
741
	swarn.errstr = errstr;
742
	swarn.dev = NULL;
743

744 745
	ret = extent_from_logical(fs_info, swarn.logical, path, &found_key,
				  &flags);
746 747 748
	if (ret < 0)
		goto out;

J
Jan Schmidt 已提交
749
	extent_item_pos = swarn.logical - found_key.objectid;
750 751 752 753
	swarn.extent_item_size = found_key.offset;

	eb = path->nodes[0];
	ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
754
	item_size = btrfs_item_size(eb, path->slots[0]);
755

756
	if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
757
		do {
758 759 760
			ret = tree_backref_for_extent(&ptr, eb, &found_key, ei,
						      item_size, &ref_root,
						      &ref_level);
761
			btrfs_warn_in_rcu(fs_info,
D
David Sterba 已提交
762
"%s at logical %llu on dev %s, physical %llu: metadata %s (level %d) in tree %llu",
J
Jeff Mahoney 已提交
763
				errstr, swarn.logical,
764
				rcu_str_deref(dev->name),
D
David Sterba 已提交
765
				swarn.physical,
766 767 768 769
				ref_level ? "node" : "leaf",
				ret < 0 ? -1 : ref_level,
				ret < 0 ? -1 : ref_root);
		} while (ret != 1);
770
		btrfs_release_path(path);
771
	} else {
772
		btrfs_release_path(path);
773
		swarn.path = path;
774
		swarn.dev = dev;
775 776
		iterate_extent_inodes(fs_info, found_key.objectid,
					extent_item_pos, 1,
777
					scrub_print_warning_inode, &swarn, false);
778 779 780 781 782 783
	}

out:
	btrfs_free_path(path);
}

784 785
static inline void scrub_get_recover(struct scrub_recover *recover)
{
786
	refcount_inc(&recover->refs);
787 788
}

789 790
static inline void scrub_put_recover(struct btrfs_fs_info *fs_info,
				     struct scrub_recover *recover)
791
{
792
	if (refcount_dec_and_test(&recover->refs)) {
793
		btrfs_bio_counter_dec(fs_info);
794
		btrfs_put_bioc(recover->bioc);
795 796 797 798
		kfree(recover);
	}
}

A
Arne Jansen 已提交
799
/*
800 801 802 803 804 805
 * scrub_handle_errored_block gets called when either verification of the
 * pages failed or the bio failed to read, e.g. with EIO. In the latter
 * case, this function handles all pages in the bio, even though only one
 * may be bad.
 * The goal of this function is to repair the errored block by using the
 * contents of one of the mirrors.
A
Arne Jansen 已提交
806
 */
807
static int scrub_handle_errored_block(struct scrub_block *sblock_to_check)
A
Arne Jansen 已提交
808
{
809
	struct scrub_ctx *sctx = sblock_to_check->sctx;
810
	struct btrfs_device *dev;
811 812 813 814 815 816 817 818 819 820 821
	struct btrfs_fs_info *fs_info;
	u64 logical;
	unsigned int failed_mirror_index;
	unsigned int is_metadata;
	unsigned int have_csum;
	struct scrub_block *sblocks_for_recheck; /* holds one for each mirror */
	struct scrub_block *sblock_bad;
	int ret;
	int mirror_index;
	int page_num;
	int success;
822
	bool full_stripe_locked;
823
	unsigned int nofs_flag;
824
	static DEFINE_RATELIMIT_STATE(rs, DEFAULT_RATELIMIT_INTERVAL,
825 826 827
				      DEFAULT_RATELIMIT_BURST);

	BUG_ON(sblock_to_check->page_count < 1);
828
	fs_info = sctx->fs_info;
829 830 831 832 833 834 835 836 837 838 839
	if (sblock_to_check->pagev[0]->flags & BTRFS_EXTENT_FLAG_SUPER) {
		/*
		 * 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(&sctx->stat_lock);
		++sctx->stat.super_errors;
		spin_unlock(&sctx->stat_lock);
		return 0;
	}
840 841 842 843
	logical = sblock_to_check->pagev[0]->logical;
	BUG_ON(sblock_to_check->pagev[0]->mirror_num < 1);
	failed_mirror_index = sblock_to_check->pagev[0]->mirror_num - 1;
	is_metadata = !(sblock_to_check->pagev[0]->flags &
844
			BTRFS_EXTENT_FLAG_DATA);
845 846
	have_csum = sblock_to_check->pagev[0]->have_csum;
	dev = sblock_to_check->pagev[0]->dev;
847

848 849 850
	if (btrfs_is_zoned(fs_info) && !sctx->is_dev_replace)
		return btrfs_repair_one_zone(fs_info, logical);

851 852 853 854 855 856 857 858 859 860
	/*
	 * We must use GFP_NOFS because the scrub task might be waiting for a
	 * worker task executing this function and in turn a transaction commit
	 * might be waiting the scrub task to pause (which needs to wait for all
	 * the worker tasks to complete before pausing).
	 * We do allocations in the workers through insert_full_stripe_lock()
	 * and scrub_add_page_to_wr_bio(), which happens down the call chain of
	 * this function.
	 */
	nofs_flag = memalloc_nofs_save();
861 862 863 864 865 866 867 868 869
	/*
	 * For RAID5/6, race can happen for a different device scrub thread.
	 * For data corruption, Parity and Data threads will both try
	 * to recovery the data.
	 * Race can lead to doubly added csum error, or even unrecoverable
	 * error.
	 */
	ret = lock_full_stripe(fs_info, logical, &full_stripe_locked);
	if (ret < 0) {
870
		memalloc_nofs_restore(nofs_flag);
871 872 873 874 875 876 877 878 879
		spin_lock(&sctx->stat_lock);
		if (ret == -ENOMEM)
			sctx->stat.malloc_errors++;
		sctx->stat.read_errors++;
		sctx->stat.uncorrectable_errors++;
		spin_unlock(&sctx->stat_lock);
		return ret;
	}

880 881 882 883
	/*
	 * read all mirrors one after the other. This includes to
	 * re-read the extent or metadata block that failed (that was
	 * the cause that this fixup code is called) another time,
884
	 * sector by sector this time in order to know which sectors
885 886 887 888
	 * caused I/O errors and which ones are good (for all mirrors).
	 * It is the goal to handle the situation when more than one
	 * mirror contains I/O errors, but the errors do not
	 * overlap, i.e. the data can be repaired by selecting the
889 890 891 892 893 894 895 896 897 898
	 * sectors from those mirrors without I/O error on the
	 * particular sectors. One example (with blocks >= 2 * sectorsize)
	 * would be that mirror #1 has an I/O error on the first sector,
	 * the second sector is good, and mirror #2 has an I/O error on
	 * the second sector, but the first sector is good.
	 * Then the first sector of the first mirror can be repaired by
	 * taking the first sector of the second mirror, and the
	 * second sector of the second mirror can be repaired by
	 * copying the contents of the 2nd sector of the 1st mirror.
	 * One more note: if the sectors of one mirror contain I/O
899 900 901
	 * errors, the checksum cannot be verified. In order to get
	 * the best data for repairing, the first attempt is to find
	 * a mirror without I/O errors and with a validated checksum.
902
	 * Only if this is not possible, the sectors are picked from
903 904 905 906 907 908
	 * mirrors with I/O errors without considering the checksum.
	 * If the latter is the case, at the end, the checksum of the
	 * repaired area is verified in order to correctly maintain
	 * the statistics.
	 */

909
	sblocks_for_recheck = kcalloc(BTRFS_MAX_MIRRORS,
910
				      sizeof(*sblocks_for_recheck), GFP_KERNEL);
911
	if (!sblocks_for_recheck) {
912 913 914 915 916
		spin_lock(&sctx->stat_lock);
		sctx->stat.malloc_errors++;
		sctx->stat.read_errors++;
		sctx->stat.uncorrectable_errors++;
		spin_unlock(&sctx->stat_lock);
917
		btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_READ_ERRS);
918
		goto out;
A
Arne Jansen 已提交
919 920
	}

921
	/* setup the context, map the logical blocks and alloc the pages */
922
	ret = scrub_setup_recheck_block(sblock_to_check, sblocks_for_recheck);
923
	if (ret) {
924 925 926 927
		spin_lock(&sctx->stat_lock);
		sctx->stat.read_errors++;
		sctx->stat.uncorrectable_errors++;
		spin_unlock(&sctx->stat_lock);
928
		btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_READ_ERRS);
929 930 931 932
		goto out;
	}
	BUG_ON(failed_mirror_index >= BTRFS_MAX_MIRRORS);
	sblock_bad = sblocks_for_recheck + failed_mirror_index;
933

934
	/* build and submit the bios for the failed mirror, check checksums */
935
	scrub_recheck_block(fs_info, sblock_bad, 1);
A
Arne Jansen 已提交
936

937 938 939 940 941 942 943 944 945 946
	if (!sblock_bad->header_error && !sblock_bad->checksum_error &&
	    sblock_bad->no_io_error_seen) {
		/*
		 * the error disappeared after reading page by page, or
		 * the area was part of a huge bio and other parts of the
		 * bio caused I/O errors, or the block layer merged several
		 * read requests into one and the error is caused by a
		 * different bio (usually one of the two latter cases is
		 * the cause)
		 */
947 948
		spin_lock(&sctx->stat_lock);
		sctx->stat.unverified_errors++;
949
		sblock_to_check->data_corrected = 1;
950
		spin_unlock(&sctx->stat_lock);
A
Arne Jansen 已提交
951

952 953
		if (sctx->is_dev_replace)
			scrub_write_block_to_dev_replace(sblock_bad);
954
		goto out;
A
Arne Jansen 已提交
955 956
	}

957
	if (!sblock_bad->no_io_error_seen) {
958 959 960
		spin_lock(&sctx->stat_lock);
		sctx->stat.read_errors++;
		spin_unlock(&sctx->stat_lock);
961
		if (__ratelimit(&rs))
962
			scrub_print_warning("i/o error", sblock_to_check);
963
		btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_READ_ERRS);
964
	} else if (sblock_bad->checksum_error) {
965 966 967
		spin_lock(&sctx->stat_lock);
		sctx->stat.csum_errors++;
		spin_unlock(&sctx->stat_lock);
968
		if (__ratelimit(&rs))
969
			scrub_print_warning("checksum error", sblock_to_check);
970
		btrfs_dev_stat_inc_and_print(dev,
971
					     BTRFS_DEV_STAT_CORRUPTION_ERRS);
972
	} else if (sblock_bad->header_error) {
973 974 975
		spin_lock(&sctx->stat_lock);
		sctx->stat.verify_errors++;
		spin_unlock(&sctx->stat_lock);
976
		if (__ratelimit(&rs))
977 978
			scrub_print_warning("checksum/header error",
					    sblock_to_check);
979
		if (sblock_bad->generation_error)
980
			btrfs_dev_stat_inc_and_print(dev,
981 982
				BTRFS_DEV_STAT_GENERATION_ERRS);
		else
983
			btrfs_dev_stat_inc_and_print(dev,
984
				BTRFS_DEV_STAT_CORRUPTION_ERRS);
985
	}
A
Arne Jansen 已提交
986

987 988 989 990
	if (sctx->readonly) {
		ASSERT(!sctx->is_dev_replace);
		goto out;
	}
A
Arne Jansen 已提交
991

992 993
	/*
	 * now build and submit the bios for the other mirrors, check
994 995
	 * checksums.
	 * First try to pick the mirror which is completely without I/O
996 997 998 999 1000 1001 1002 1003 1004 1005 1006
	 * errors and also does not have a checksum error.
	 * If one is found, and if a checksum is present, the full block
	 * that is known to contain an error is rewritten. Afterwards
	 * the block is known to be corrected.
	 * If a mirror is found which is completely correct, and no
	 * checksum is present, only those pages are rewritten that had
	 * an I/O error in the block to be repaired, since it cannot be
	 * determined, which copy of the other pages is better (and it
	 * could happen otherwise that a correct page would be
	 * overwritten by a bad one).
	 */
1007
	for (mirror_index = 0; ;mirror_index++) {
1008
		struct scrub_block *sblock_other;
1009

1010 1011
		if (mirror_index == failed_mirror_index)
			continue;
1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022

		/* raid56's mirror can be more than BTRFS_MAX_MIRRORS */
		if (!scrub_is_page_on_raid56(sblock_bad->pagev[0])) {
			if (mirror_index >= BTRFS_MAX_MIRRORS)
				break;
			if (!sblocks_for_recheck[mirror_index].page_count)
				break;

			sblock_other = sblocks_for_recheck + mirror_index;
		} else {
			struct scrub_recover *r = sblock_bad->pagev[0]->recover;
1023
			int max_allowed = r->bioc->num_stripes - r->bioc->num_tgtdevs;
1024 1025 1026 1027 1028 1029 1030 1031 1032 1033

			if (mirror_index >= max_allowed)
				break;
			if (!sblocks_for_recheck[1].page_count)
				break;

			ASSERT(failed_mirror_index == 0);
			sblock_other = sblocks_for_recheck + 1;
			sblock_other->pagev[0]->mirror_num = 1 + mirror_index;
		}
1034 1035

		/* build and submit the bios, check checksums */
1036
		scrub_recheck_block(fs_info, sblock_other, 0);
1037 1038

		if (!sblock_other->header_error &&
1039 1040
		    !sblock_other->checksum_error &&
		    sblock_other->no_io_error_seen) {
1041 1042
			if (sctx->is_dev_replace) {
				scrub_write_block_to_dev_replace(sblock_other);
1043
				goto corrected_error;
1044 1045
			} else {
				ret = scrub_repair_block_from_good_copy(
1046 1047 1048
						sblock_bad, sblock_other);
				if (!ret)
					goto corrected_error;
1049
			}
1050 1051
		}
	}
A
Arne Jansen 已提交
1052

1053 1054
	if (sblock_bad->no_io_error_seen && !sctx->is_dev_replace)
		goto did_not_correct_error;
1055 1056 1057

	/*
	 * In case of I/O errors in the area that is supposed to be
1058 1059
	 * repaired, continue by picking good copies of those sectors.
	 * Select the good sectors from mirrors to rewrite bad sectors from
1060 1061 1062 1063 1064
	 * the area to fix. Afterwards verify the checksum of the block
	 * that is supposed to be repaired. This verification step is
	 * only done for the purpose of statistic counting and for the
	 * final scrub report, whether errors remain.
	 * A perfect algorithm could make use of the checksum and try
1065
	 * all possible combinations of sectors from the different mirrors
1066
	 * until the checksum verification succeeds. For example, when
1067
	 * the 2nd sector of mirror #1 faces I/O errors, and the 2nd sector
1068
	 * of mirror #2 is readable but the final checksum test fails,
1069
	 * then the 2nd sector of mirror #3 could be tried, whether now
1070
	 * the final checksum succeeds. But this would be a rare
1071 1072 1073 1074
	 * exception and is therefore not implemented. At least it is
	 * avoided that the good copy is overwritten.
	 * A more useful improvement would be to pick the sectors
	 * without I/O error based on sector sizes (512 bytes on legacy
1075
	 * disks) instead of on sectorsize. Then maybe 512 byte of one
1076
	 * mirror could be repaired by taking 512 byte of a different
1077
	 * mirror, even if other 512 byte sectors in the same sectorsize
1078
	 * area are unreadable.
A
Arne Jansen 已提交
1079
	 */
1080
	success = 1;
1081 1082
	for (page_num = 0; page_num < sblock_bad->page_count;
	     page_num++) {
1083
		struct scrub_page *spage_bad = sblock_bad->pagev[page_num];
1084
		struct scrub_block *sblock_other = NULL;
1085

1086
		/* skip no-io-error page in scrub */
1087
		if (!spage_bad->io_error && !sctx->is_dev_replace)
A
Arne Jansen 已提交
1088
			continue;
1089

1090 1091 1092 1093 1094 1095 1096 1097 1098
		if (scrub_is_page_on_raid56(sblock_bad->pagev[0])) {
			/*
			 * In case of dev replace, if raid56 rebuild process
			 * didn't work out correct data, then copy the content
			 * in sblock_bad to make sure target device is identical
			 * to source device, instead of writing garbage data in
			 * sblock_for_recheck array to target device.
			 */
			sblock_other = NULL;
1099
		} else if (spage_bad->io_error) {
1100
			/* try to find no-io-error page in mirrors */
1101 1102 1103 1104 1105 1106 1107 1108 1109
			for (mirror_index = 0;
			     mirror_index < BTRFS_MAX_MIRRORS &&
			     sblocks_for_recheck[mirror_index].page_count > 0;
			     mirror_index++) {
				if (!sblocks_for_recheck[mirror_index].
				    pagev[page_num]->io_error) {
					sblock_other = sblocks_for_recheck +
						       mirror_index;
					break;
1110 1111
				}
			}
1112 1113
			if (!sblock_other)
				success = 0;
I
Ilya Dryomov 已提交
1114
		}
A
Arne Jansen 已提交
1115

1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128
		if (sctx->is_dev_replace) {
			/*
			 * did not find a mirror to fetch the page
			 * from. scrub_write_page_to_dev_replace()
			 * handles this case (page->io_error), by
			 * filling the block with zeros before
			 * submitting the write request
			 */
			if (!sblock_other)
				sblock_other = sblock_bad;

			if (scrub_write_page_to_dev_replace(sblock_other,
							    page_num) != 0) {
1129
				atomic64_inc(
1130
					&fs_info->dev_replace.num_write_errors);
1131 1132 1133 1134 1135 1136 1137
				success = 0;
			}
		} else if (sblock_other) {
			ret = scrub_repair_page_from_good_copy(sblock_bad,
							       sblock_other,
							       page_num, 0);
			if (0 == ret)
1138
				spage_bad->io_error = 0;
1139 1140
			else
				success = 0;
1141
		}
A
Arne Jansen 已提交
1142 1143
	}

1144
	if (success && !sctx->is_dev_replace) {
1145 1146 1147 1148 1149 1150 1151 1152 1153 1154
		if (is_metadata || have_csum) {
			/*
			 * need to verify the checksum now that all
			 * sectors on disk are repaired (the write
			 * request for data to be repaired is on its way).
			 * Just be lazy and use scrub_recheck_block()
			 * which re-reads the data before the checksum
			 * is verified, but most likely the data comes out
			 * of the page cache.
			 */
1155
			scrub_recheck_block(fs_info, sblock_bad, 1);
1156
			if (!sblock_bad->header_error &&
1157 1158 1159 1160 1161 1162 1163
			    !sblock_bad->checksum_error &&
			    sblock_bad->no_io_error_seen)
				goto corrected_error;
			else
				goto did_not_correct_error;
		} else {
corrected_error:
1164 1165
			spin_lock(&sctx->stat_lock);
			sctx->stat.corrected_errors++;
1166
			sblock_to_check->data_corrected = 1;
1167
			spin_unlock(&sctx->stat_lock);
1168 1169
			btrfs_err_rl_in_rcu(fs_info,
				"fixed up error at logical %llu on dev %s",
1170
				logical, rcu_str_deref(dev->name));
A
Arne Jansen 已提交
1171
		}
1172 1173
	} else {
did_not_correct_error:
1174 1175 1176
		spin_lock(&sctx->stat_lock);
		sctx->stat.uncorrectable_errors++;
		spin_unlock(&sctx->stat_lock);
1177 1178
		btrfs_err_rl_in_rcu(fs_info,
			"unable to fixup (regular) error at logical %llu on dev %s",
1179
			logical, rcu_str_deref(dev->name));
I
Ilya Dryomov 已提交
1180
	}
A
Arne Jansen 已提交
1181

1182 1183 1184 1185 1186 1187
out:
	if (sblocks_for_recheck) {
		for (mirror_index = 0; mirror_index < BTRFS_MAX_MIRRORS;
		     mirror_index++) {
			struct scrub_block *sblock = sblocks_for_recheck +
						     mirror_index;
1188
			struct scrub_recover *recover;
1189 1190
			int page_index;

1191 1192 1193
			for (page_index = 0; page_index < sblock->page_count;
			     page_index++) {
				sblock->pagev[page_index]->sblock = NULL;
1194 1195
				recover = sblock->pagev[page_index]->recover;
				if (recover) {
1196
					scrub_put_recover(fs_info, recover);
1197 1198 1199
					sblock->pagev[page_index]->recover =
									NULL;
				}
1200 1201
				scrub_page_put(sblock->pagev[page_index]);
			}
1202 1203 1204
		}
		kfree(sblocks_for_recheck);
	}
A
Arne Jansen 已提交
1205

1206
	ret = unlock_full_stripe(fs_info, logical, full_stripe_locked);
1207
	memalloc_nofs_restore(nofs_flag);
1208 1209
	if (ret < 0)
		return ret;
1210 1211
	return 0;
}
A
Arne Jansen 已提交
1212

1213
static inline int scrub_nr_raid_mirrors(struct btrfs_io_context *bioc)
1214
{
1215
	if (bioc->map_type & BTRFS_BLOCK_GROUP_RAID5)
Z
Zhao Lei 已提交
1216
		return 2;
1217
	else if (bioc->map_type & BTRFS_BLOCK_GROUP_RAID6)
Z
Zhao Lei 已提交
1218 1219
		return 3;
	else
1220
		return (int)bioc->num_stripes;
1221 1222
}

Z
Zhao Lei 已提交
1223 1224
static inline void scrub_stripe_index_and_offset(u64 logical, u64 map_type,
						 u64 *raid_map,
1225 1226 1227 1228 1229 1230 1231
						 u64 mapped_length,
						 int nstripes, int mirror,
						 int *stripe_index,
						 u64 *stripe_offset)
{
	int i;

1232
	if (map_type & BTRFS_BLOCK_GROUP_RAID56_MASK) {
1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252
		/* RAID5/6 */
		for (i = 0; i < nstripes; i++) {
			if (raid_map[i] == RAID6_Q_STRIPE ||
			    raid_map[i] == RAID5_P_STRIPE)
				continue;

			if (logical >= raid_map[i] &&
			    logical < raid_map[i] + mapped_length)
				break;
		}

		*stripe_index = i;
		*stripe_offset = logical - raid_map[i];
	} else {
		/* The other RAID type */
		*stripe_index = mirror;
		*stripe_offset = 0;
	}
}

1253
static int scrub_setup_recheck_block(struct scrub_block *original_sblock,
1254 1255
				     struct scrub_block *sblocks_for_recheck)
{
1256
	struct scrub_ctx *sctx = original_sblock->sctx;
1257
	struct btrfs_fs_info *fs_info = sctx->fs_info;
1258
	u64 length = original_sblock->page_count * fs_info->sectorsize;
1259
	u64 logical = original_sblock->pagev[0]->logical;
1260 1261 1262
	u64 generation = original_sblock->pagev[0]->generation;
	u64 flags = original_sblock->pagev[0]->flags;
	u64 have_csum = original_sblock->pagev[0]->have_csum;
1263
	struct scrub_recover *recover;
1264
	struct btrfs_io_context *bioc;
1265 1266 1267 1268
	u64 sublen;
	u64 mapped_length;
	u64 stripe_offset;
	int stripe_index;
1269
	int page_index = 0;
1270
	int mirror_index;
1271
	int nmirrors;
1272 1273 1274
	int ret;

	/*
1275
	 * note: the two members refs and outstanding_pages
1276 1277 1278 1279 1280
	 * are not used (and not set) in the blocks that are used for
	 * the recheck procedure
	 */

	while (length > 0) {
1281
		sublen = min_t(u64, length, fs_info->sectorsize);
1282
		mapped_length = sublen;
1283
		bioc = NULL;
A
Arne Jansen 已提交
1284

1285
		/*
1286 1287
		 * With a length of sectorsize, each returned stripe represents
		 * one mirror
1288
		 */
1289
		btrfs_bio_counter_inc_blocked(fs_info);
1290
		ret = btrfs_map_sblock(fs_info, BTRFS_MAP_GET_READ_MIRRORS,
1291 1292 1293
				       logical, &mapped_length, &bioc);
		if (ret || !bioc || mapped_length < sublen) {
			btrfs_put_bioc(bioc);
1294
			btrfs_bio_counter_dec(fs_info);
1295 1296
			return -EIO;
		}
A
Arne Jansen 已提交
1297

1298 1299
		recover = kzalloc(sizeof(struct scrub_recover), GFP_NOFS);
		if (!recover) {
1300
			btrfs_put_bioc(bioc);
1301
			btrfs_bio_counter_dec(fs_info);
1302 1303 1304
			return -ENOMEM;
		}

1305
		refcount_set(&recover->refs, 1);
1306
		recover->bioc = bioc;
1307 1308
		recover->map_length = mapped_length;

1309
		ASSERT(page_index < SCRUB_MAX_PAGES_PER_BLOCK);
1310

1311
		nmirrors = min(scrub_nr_raid_mirrors(bioc), BTRFS_MAX_MIRRORS);
Z
Zhao Lei 已提交
1312

1313
		for (mirror_index = 0; mirror_index < nmirrors;
1314 1315
		     mirror_index++) {
			struct scrub_block *sblock;
1316
			struct scrub_page *spage;
1317 1318

			sblock = sblocks_for_recheck + mirror_index;
1319
			sblock->sctx = sctx;
1320

1321 1322
			spage = kzalloc(sizeof(*spage), GFP_NOFS);
			if (!spage) {
1323
leave_nomem:
1324 1325 1326
				spin_lock(&sctx->stat_lock);
				sctx->stat.malloc_errors++;
				spin_unlock(&sctx->stat_lock);
1327
				scrub_put_recover(fs_info, recover);
1328 1329
				return -ENOMEM;
			}
1330 1331 1332 1333 1334 1335 1336
			scrub_page_get(spage);
			sblock->pagev[page_index] = spage;
			spage->sblock = sblock;
			spage->flags = flags;
			spage->generation = generation;
			spage->logical = logical;
			spage->have_csum = have_csum;
1337
			if (have_csum)
1338
				memcpy(spage->csum,
1339
				       original_sblock->pagev[0]->csum,
1340
				       sctx->fs_info->csum_size);
1341

Z
Zhao Lei 已提交
1342
			scrub_stripe_index_and_offset(logical,
1343 1344
						      bioc->map_type,
						      bioc->raid_map,
1345
						      mapped_length,
1346 1347
						      bioc->num_stripes -
						      bioc->num_tgtdevs,
1348 1349 1350
						      mirror_index,
						      &stripe_index,
						      &stripe_offset);
1351
			spage->physical = bioc->stripes[stripe_index].physical +
1352
					 stripe_offset;
1353
			spage->dev = bioc->stripes[stripe_index].dev;
1354

1355
			BUG_ON(page_index >= original_sblock->page_count);
1356
			spage->physical_for_dev_replace =
1357 1358
				original_sblock->pagev[page_index]->
				physical_for_dev_replace;
1359
			/* for missing devices, dev->bdev is NULL */
1360
			spage->mirror_num = mirror_index + 1;
1361
			sblock->page_count++;
1362 1363
			spage->page = alloc_page(GFP_NOFS);
			if (!spage->page)
1364
				goto leave_nomem;
1365 1366

			scrub_get_recover(recover);
1367
			spage->recover = recover;
1368
		}
1369
		scrub_put_recover(fs_info, recover);
1370 1371 1372 1373 1374 1375
		length -= sublen;
		logical += sublen;
		page_index++;
	}

	return 0;
I
Ilya Dryomov 已提交
1376 1377
}

1378
static void scrub_bio_wait_endio(struct bio *bio)
1379
{
1380
	complete(bio->bi_private);
1381 1382 1383 1384
}

static int scrub_submit_raid56_bio_wait(struct btrfs_fs_info *fs_info,
					struct bio *bio,
1385
					struct scrub_page *spage)
1386
{
1387
	DECLARE_COMPLETION_ONSTACK(done);
1388
	int ret;
1389
	int mirror_num;
1390

1391
	bio->bi_iter.bi_sector = spage->logical >> 9;
1392 1393 1394
	bio->bi_private = &done;
	bio->bi_end_io = scrub_bio_wait_endio;

1395
	mirror_num = spage->sblock->pagev[0]->mirror_num;
1396
	ret = raid56_parity_recover(bio, spage->recover->bioc,
1397
				    spage->recover->map_length,
1398
				    mirror_num, 0);
1399 1400 1401
	if (ret)
		return ret;

1402 1403
	wait_for_completion_io(&done);
	return blk_status_to_errno(bio->bi_status);
1404 1405
}

L
Liu Bo 已提交
1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417
static void scrub_recheck_block_on_raid56(struct btrfs_fs_info *fs_info,
					  struct scrub_block *sblock)
{
	struct scrub_page *first_page = sblock->pagev[0];
	struct bio *bio;
	int page_num;

	/* All pages in sblock belong to the same stripe on the same device. */
	ASSERT(first_page->dev);
	if (!first_page->dev->bdev)
		goto out;

1418
	bio = btrfs_bio_alloc(BIO_MAX_VECS);
L
Liu Bo 已提交
1419 1420 1421
	bio_set_dev(bio, first_page->dev->bdev);

	for (page_num = 0; page_num < sblock->page_count; page_num++) {
1422
		struct scrub_page *spage = sblock->pagev[page_num];
L
Liu Bo 已提交
1423

1424 1425
		WARN_ON(!spage->page);
		bio_add_page(bio, spage->page, PAGE_SIZE, 0);
L
Liu Bo 已提交
1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444
	}

	if (scrub_submit_raid56_bio_wait(fs_info, bio, first_page)) {
		bio_put(bio);
		goto out;
	}

	bio_put(bio);

	scrub_recheck_block_checksum(sblock);

	return;
out:
	for (page_num = 0; page_num < sblock->page_count; page_num++)
		sblock->pagev[page_num]->io_error = 1;

	sblock->no_io_error_seen = 0;
}

1445 1446 1447 1448 1449 1450 1451
/*
 * this function will check the on disk data for checksum errors, header
 * errors and read I/O errors. If any I/O errors happen, the exact pages
 * which are errored are marked as being bad. The goal is to enable scrub
 * to take those pages that are not errored from all the mirrors so that
 * the pages that are errored in the just handled mirror can be repaired.
 */
1452
static void scrub_recheck_block(struct btrfs_fs_info *fs_info,
1453 1454
				struct scrub_block *sblock,
				int retry_failed_mirror)
I
Ilya Dryomov 已提交
1455
{
1456
	int page_num;
I
Ilya Dryomov 已提交
1457

1458
	sblock->no_io_error_seen = 1;
I
Ilya Dryomov 已提交
1459

L
Liu Bo 已提交
1460 1461 1462 1463
	/* short cut for raid56 */
	if (!retry_failed_mirror && scrub_is_page_on_raid56(sblock->pagev[0]))
		return scrub_recheck_block_on_raid56(fs_info, sblock);

1464 1465
	for (page_num = 0; page_num < sblock->page_count; page_num++) {
		struct bio *bio;
1466
		struct scrub_page *spage = sblock->pagev[page_num];
1467

1468 1469
		if (spage->dev->bdev == NULL) {
			spage->io_error = 1;
1470 1471 1472 1473
			sblock->no_io_error_seen = 0;
			continue;
		}

1474
		WARN_ON(!spage->page);
1475
		bio = btrfs_bio_alloc(1);
1476
		bio_set_dev(bio, spage->dev->bdev);
1477

1478
		bio_add_page(bio, spage->page, fs_info->sectorsize, 0);
1479
		bio->bi_iter.bi_sector = spage->physical >> 9;
L
Liu Bo 已提交
1480
		bio->bi_opf = REQ_OP_READ;
1481

L
Liu Bo 已提交
1482
		if (btrfsic_submit_bio_wait(bio)) {
1483
			spage->io_error = 1;
L
Liu Bo 已提交
1484
			sblock->no_io_error_seen = 0;
1485
		}
1486

1487 1488
		bio_put(bio);
	}
I
Ilya Dryomov 已提交
1489

1490
	if (sblock->no_io_error_seen)
1491
		scrub_recheck_block_checksum(sblock);
A
Arne Jansen 已提交
1492 1493
}

M
Miao Xie 已提交
1494 1495 1496 1497 1498 1499
static inline int scrub_check_fsid(u8 fsid[],
				   struct scrub_page *spage)
{
	struct btrfs_fs_devices *fs_devices = spage->dev->fs_devices;
	int ret;

1500
	ret = memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE);
M
Miao Xie 已提交
1501 1502 1503
	return !ret;
}

1504
static void scrub_recheck_block_checksum(struct scrub_block *sblock)
A
Arne Jansen 已提交
1505
{
1506 1507 1508
	sblock->header_error = 0;
	sblock->checksum_error = 0;
	sblock->generation_error = 0;
1509

1510 1511 1512 1513
	if (sblock->pagev[0]->flags & BTRFS_EXTENT_FLAG_DATA)
		scrub_checksum_data(sblock);
	else
		scrub_checksum_tree_block(sblock);
A
Arne Jansen 已提交
1514 1515
}

1516
static int scrub_repair_block_from_good_copy(struct scrub_block *sblock_bad,
1517
					     struct scrub_block *sblock_good)
1518 1519 1520
{
	int page_num;
	int ret = 0;
I
Ilya Dryomov 已提交
1521

1522 1523
	for (page_num = 0; page_num < sblock_bad->page_count; page_num++) {
		int ret_sub;
I
Ilya Dryomov 已提交
1524

1525 1526
		ret_sub = scrub_repair_page_from_good_copy(sblock_bad,
							   sblock_good,
1527
							   page_num, 1);
1528 1529
		if (ret_sub)
			ret = ret_sub;
A
Arne Jansen 已提交
1530
	}
1531 1532 1533 1534 1535 1536 1537 1538

	return ret;
}

static int scrub_repair_page_from_good_copy(struct scrub_block *sblock_bad,
					    struct scrub_block *sblock_good,
					    int page_num, int force_write)
{
1539 1540
	struct scrub_page *spage_bad = sblock_bad->pagev[page_num];
	struct scrub_page *spage_good = sblock_good->pagev[page_num];
1541
	struct btrfs_fs_info *fs_info = sblock_bad->sctx->fs_info;
1542
	const u32 sectorsize = fs_info->sectorsize;
1543

1544 1545
	BUG_ON(spage_bad->page == NULL);
	BUG_ON(spage_good->page == NULL);
1546
	if (force_write || sblock_bad->header_error ||
1547
	    sblock_bad->checksum_error || spage_bad->io_error) {
1548 1549 1550
		struct bio *bio;
		int ret;

1551
		if (!spage_bad->dev->bdev) {
1552
			btrfs_warn_rl(fs_info,
J
Jeff Mahoney 已提交
1553
				"scrub_repair_page_from_good_copy(bdev == NULL) is unexpected");
1554 1555 1556
			return -EIO;
		}

1557
		bio = btrfs_bio_alloc(1);
1558 1559
		bio_set_dev(bio, spage_bad->dev->bdev);
		bio->bi_iter.bi_sector = spage_bad->physical >> 9;
D
David Sterba 已提交
1560
		bio->bi_opf = REQ_OP_WRITE;
1561

1562 1563
		ret = bio_add_page(bio, spage_good->page, sectorsize, 0);
		if (ret != sectorsize) {
1564 1565
			bio_put(bio);
			return -EIO;
1566
		}
1567

1568
		if (btrfsic_submit_bio_wait(bio)) {
1569
			btrfs_dev_stat_inc_and_print(spage_bad->dev,
1570
				BTRFS_DEV_STAT_WRITE_ERRS);
1571
			atomic64_inc(&fs_info->dev_replace.num_write_errors);
1572 1573 1574
			bio_put(bio);
			return -EIO;
		}
1575
		bio_put(bio);
A
Arne Jansen 已提交
1576 1577
	}

1578 1579 1580
	return 0;
}

1581 1582
static void scrub_write_block_to_dev_replace(struct scrub_block *sblock)
{
1583
	struct btrfs_fs_info *fs_info = sblock->sctx->fs_info;
1584 1585
	int page_num;

1586 1587 1588 1589 1590 1591 1592
	/*
	 * This block is used for the check of the parity on the source device,
	 * so the data needn't be written into the destination device.
	 */
	if (sblock->sparity)
		return;

1593 1594 1595 1596 1597
	for (page_num = 0; page_num < sblock->page_count; page_num++) {
		int ret;

		ret = scrub_write_page_to_dev_replace(sblock, page_num);
		if (ret)
1598
			atomic64_inc(&fs_info->dev_replace.num_write_errors);
1599 1600 1601 1602 1603 1604 1605 1606 1607
	}
}

static int scrub_write_page_to_dev_replace(struct scrub_block *sblock,
					   int page_num)
{
	struct scrub_page *spage = sblock->pagev[page_num];

	BUG_ON(spage->page == NULL);
1608 1609
	if (spage->io_error)
		clear_page(page_address(spage->page));
1610 1611 1612 1613

	return scrub_add_page_to_wr_bio(sblock->sctx, spage);
}

1614 1615 1616 1617 1618 1619 1620 1621
static int fill_writer_pointer_gap(struct scrub_ctx *sctx, u64 physical)
{
	int ret = 0;
	u64 length;

	if (!btrfs_is_zoned(sctx->fs_info))
		return 0;

1622 1623 1624
	if (!btrfs_dev_is_sequential(sctx->wr_tgtdev, physical))
		return 0;

1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635
	if (sctx->write_pointer < physical) {
		length = physical - sctx->write_pointer;

		ret = btrfs_zoned_issue_zeroout(sctx->wr_tgtdev,
						sctx->write_pointer, length);
		if (!ret)
			sctx->write_pointer = physical;
	}
	return ret;
}

1636 1637 1638 1639 1640
static int scrub_add_page_to_wr_bio(struct scrub_ctx *sctx,
				    struct scrub_page *spage)
{
	struct scrub_bio *sbio;
	int ret;
1641
	const u32 sectorsize = sctx->fs_info->sectorsize;
1642

1643
	mutex_lock(&sctx->wr_lock);
1644
again:
1645 1646
	if (!sctx->wr_curr_bio) {
		sctx->wr_curr_bio = kzalloc(sizeof(*sctx->wr_curr_bio),
1647
					      GFP_KERNEL);
1648 1649
		if (!sctx->wr_curr_bio) {
			mutex_unlock(&sctx->wr_lock);
1650 1651
			return -ENOMEM;
		}
1652 1653
		sctx->wr_curr_bio->sctx = sctx;
		sctx->wr_curr_bio->page_count = 0;
1654
	}
1655
	sbio = sctx->wr_curr_bio;
1656 1657 1658
	if (sbio->page_count == 0) {
		struct bio *bio;

1659 1660 1661 1662 1663 1664 1665
		ret = fill_writer_pointer_gap(sctx,
					      spage->physical_for_dev_replace);
		if (ret) {
			mutex_unlock(&sctx->wr_lock);
			return ret;
		}

1666 1667
		sbio->physical = spage->physical_for_dev_replace;
		sbio->logical = spage->logical;
1668
		sbio->dev = sctx->wr_tgtdev;
1669 1670
		bio = sbio->bio;
		if (!bio) {
1671
			bio = btrfs_bio_alloc(sctx->pages_per_bio);
1672 1673 1674 1675 1676
			sbio->bio = bio;
		}

		bio->bi_private = sbio;
		bio->bi_end_io = scrub_wr_bio_end_io;
1677
		bio_set_dev(bio, sbio->dev->bdev);
1678
		bio->bi_iter.bi_sector = sbio->physical >> 9;
D
David Sterba 已提交
1679
		bio->bi_opf = REQ_OP_WRITE;
1680
		sbio->status = 0;
1681
	} else if (sbio->physical + sbio->page_count * sectorsize !=
1682
		   spage->physical_for_dev_replace ||
1683
		   sbio->logical + sbio->page_count * sectorsize !=
1684 1685 1686 1687 1688
		   spage->logical) {
		scrub_wr_submit(sctx);
		goto again;
	}

1689 1690
	ret = bio_add_page(sbio->bio, spage->page, sectorsize, 0);
	if (ret != sectorsize) {
1691 1692 1693
		if (sbio->page_count < 1) {
			bio_put(sbio->bio);
			sbio->bio = NULL;
1694
			mutex_unlock(&sctx->wr_lock);
1695 1696 1697 1698 1699 1700 1701 1702 1703
			return -EIO;
		}
		scrub_wr_submit(sctx);
		goto again;
	}

	sbio->pagev[sbio->page_count] = spage;
	scrub_page_get(spage);
	sbio->page_count++;
1704
	if (sbio->page_count == sctx->pages_per_bio)
1705
		scrub_wr_submit(sctx);
1706
	mutex_unlock(&sctx->wr_lock);
1707 1708 1709 1710 1711 1712 1713 1714

	return 0;
}

static void scrub_wr_submit(struct scrub_ctx *sctx)
{
	struct scrub_bio *sbio;

1715
	if (!sctx->wr_curr_bio)
1716 1717
		return;

1718 1719
	sbio = sctx->wr_curr_bio;
	sctx->wr_curr_bio = NULL;
1720
	WARN_ON(!sbio->bio->bi_bdev);
1721 1722 1723 1724 1725
	scrub_pending_bio_inc(sctx);
	/* process all writes in a single worker thread. Then the block layer
	 * orders the requests before sending them to the driver which
	 * doubled the write performance on spinning disks when measured
	 * with Linux 3.5 */
1726
	btrfsic_submit_bio(sbio->bio);
1727 1728

	if (btrfs_is_zoned(sctx->fs_info))
1729 1730
		sctx->write_pointer = sbio->physical + sbio->page_count *
			sctx->fs_info->sectorsize;
1731 1732
}

1733
static void scrub_wr_bio_end_io(struct bio *bio)
1734 1735
{
	struct scrub_bio *sbio = bio->bi_private;
1736
	struct btrfs_fs_info *fs_info = sbio->dev->fs_info;
1737

1738
	sbio->status = bio->bi_status;
1739 1740
	sbio->bio = bio;

1741
	btrfs_init_work(&sbio->work, scrub_wr_bio_end_io_worker, NULL, NULL);
1742
	btrfs_queue_work(fs_info->scrub_wr_completion_workers, &sbio->work);
1743 1744 1745 1746 1747 1748 1749 1750
}

static void scrub_wr_bio_end_io_worker(struct btrfs_work *work)
{
	struct scrub_bio *sbio = container_of(work, struct scrub_bio, work);
	struct scrub_ctx *sctx = sbio->sctx;
	int i;

1751
	ASSERT(sbio->page_count <= SCRUB_PAGES_PER_BIO);
1752
	if (sbio->status) {
1753
		struct btrfs_dev_replace *dev_replace =
1754
			&sbio->sctx->fs_info->dev_replace;
1755 1756 1757 1758 1759

		for (i = 0; i < sbio->page_count; i++) {
			struct scrub_page *spage = sbio->pagev[i];

			spage->io_error = 1;
1760
			atomic64_inc(&dev_replace->num_write_errors);
1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772
		}
	}

	for (i = 0; i < sbio->page_count; i++)
		scrub_page_put(sbio->pagev[i]);

	bio_put(sbio->bio);
	kfree(sbio);
	scrub_pending_bio_dec(sctx);
}

static int scrub_checksum(struct scrub_block *sblock)
1773 1774 1775 1776
{
	u64 flags;
	int ret;

1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788
	/*
	 * No need to initialize these stats currently,
	 * because this function only use return value
	 * instead of these stats value.
	 *
	 * Todo:
	 * always use stats
	 */
	sblock->header_error = 0;
	sblock->generation_error = 0;
	sblock->checksum_error = 0;

1789 1790
	WARN_ON(sblock->page_count < 1);
	flags = sblock->pagev[0]->flags;
1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801
	ret = 0;
	if (flags & BTRFS_EXTENT_FLAG_DATA)
		ret = scrub_checksum_data(sblock);
	else if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
		ret = scrub_checksum_tree_block(sblock);
	else if (flags & BTRFS_EXTENT_FLAG_SUPER)
		(void)scrub_checksum_super(sblock);
	else
		WARN_ON(1);
	if (ret)
		scrub_handle_errored_block(sblock);
1802 1803

	return ret;
A
Arne Jansen 已提交
1804 1805
}

1806
static int scrub_checksum_data(struct scrub_block *sblock)
A
Arne Jansen 已提交
1807
{
1808
	struct scrub_ctx *sctx = sblock->sctx;
1809 1810
	struct btrfs_fs_info *fs_info = sctx->fs_info;
	SHASH_DESC_ON_STACK(shash, fs_info->csum_shash);
A
Arne Jansen 已提交
1811
	u8 csum[BTRFS_CSUM_SIZE];
1812
	struct scrub_page *spage;
1813
	char *kaddr;
A
Arne Jansen 已提交
1814

1815
	BUG_ON(sblock->page_count < 1);
1816 1817
	spage = sblock->pagev[0];
	if (!spage->have_csum)
A
Arne Jansen 已提交
1818 1819
		return 0;

1820
	kaddr = page_address(spage->page);
1821

1822 1823
	shash->tfm = fs_info->csum_shash;
	crypto_shash_init(shash);
1824

1825 1826 1827 1828 1829
	/*
	 * In scrub_pages() and scrub_pages_for_parity() we ensure each spage
	 * only contains one sector of data.
	 */
	crypto_shash_digest(shash, kaddr, fs_info->sectorsize, csum);
A
Arne Jansen 已提交
1830

1831 1832
	if (memcmp(csum, spage->csum, fs_info->csum_size))
		sblock->checksum_error = 1;
1833
	return sblock->checksum_error;
A
Arne Jansen 已提交
1834 1835
}

1836
static int scrub_checksum_tree_block(struct scrub_block *sblock)
A
Arne Jansen 已提交
1837
{
1838
	struct scrub_ctx *sctx = sblock->sctx;
A
Arne Jansen 已提交
1839
	struct btrfs_header *h;
1840
	struct btrfs_fs_info *fs_info = sctx->fs_info;
1841
	SHASH_DESC_ON_STACK(shash, fs_info->csum_shash);
1842 1843
	u8 calculated_csum[BTRFS_CSUM_SIZE];
	u8 on_disk_csum[BTRFS_CSUM_SIZE];
1844 1845 1846 1847 1848 1849 1850
	/*
	 * This is done in sectorsize steps even for metadata as there's a
	 * constraint for nodesize to be aligned to sectorsize. This will need
	 * to change so we don't misuse data and metadata units like that.
	 */
	const u32 sectorsize = sctx->fs_info->sectorsize;
	const int num_sectors = fs_info->nodesize >> fs_info->sectorsize_bits;
1851
	int i;
1852
	struct scrub_page *spage;
1853
	char *kaddr;
1854

1855
	BUG_ON(sblock->page_count < 1);
1856 1857 1858 1859

	/* Each member in pagev is just one block, not a full page */
	ASSERT(sblock->page_count == num_sectors);

1860 1861
	spage = sblock->pagev[0];
	kaddr = page_address(spage->page);
1862
	h = (struct btrfs_header *)kaddr;
1863
	memcpy(on_disk_csum, h->csum, sctx->fs_info->csum_size);
A
Arne Jansen 已提交
1864 1865 1866 1867 1868 1869

	/*
	 * we don't use the getter functions here, as we
	 * a) don't have an extent buffer and
	 * b) the page is already kmapped
	 */
1870
	if (spage->logical != btrfs_stack_header_bytenr(h))
1871
		sblock->header_error = 1;
A
Arne Jansen 已提交
1872

1873
	if (spage->generation != btrfs_stack_header_generation(h)) {
1874 1875 1876
		sblock->header_error = 1;
		sblock->generation_error = 1;
	}
A
Arne Jansen 已提交
1877

1878
	if (!scrub_check_fsid(h->fsid, spage))
1879
		sblock->header_error = 1;
A
Arne Jansen 已提交
1880 1881 1882

	if (memcmp(h->chunk_tree_uuid, fs_info->chunk_tree_uuid,
		   BTRFS_UUID_SIZE))
1883
		sblock->header_error = 1;
A
Arne Jansen 已提交
1884

1885 1886 1887
	shash->tfm = fs_info->csum_shash;
	crypto_shash_init(shash);
	crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE,
1888
			    sectorsize - BTRFS_CSUM_SIZE);
1889

1890
	for (i = 1; i < num_sectors; i++) {
1891
		kaddr = page_address(sblock->pagev[i]->page);
1892
		crypto_shash_update(shash, kaddr, sectorsize);
1893 1894
	}

1895
	crypto_shash_final(shash, calculated_csum);
1896
	if (memcmp(calculated_csum, on_disk_csum, sctx->fs_info->csum_size))
1897
		sblock->checksum_error = 1;
A
Arne Jansen 已提交
1898

1899
	return sblock->header_error || sblock->checksum_error;
A
Arne Jansen 已提交
1900 1901
}

1902
static int scrub_checksum_super(struct scrub_block *sblock)
A
Arne Jansen 已提交
1903 1904
{
	struct btrfs_super_block *s;
1905
	struct scrub_ctx *sctx = sblock->sctx;
1906 1907
	struct btrfs_fs_info *fs_info = sctx->fs_info;
	SHASH_DESC_ON_STACK(shash, fs_info->csum_shash);
1908
	u8 calculated_csum[BTRFS_CSUM_SIZE];
1909
	struct scrub_page *spage;
1910
	char *kaddr;
1911 1912
	int fail_gen = 0;
	int fail_cor = 0;
1913

1914
	BUG_ON(sblock->page_count < 1);
1915 1916
	spage = sblock->pagev[0];
	kaddr = page_address(spage->page);
1917
	s = (struct btrfs_super_block *)kaddr;
A
Arne Jansen 已提交
1918

1919
	if (spage->logical != btrfs_super_bytenr(s))
1920
		++fail_cor;
A
Arne Jansen 已提交
1921

1922
	if (spage->generation != btrfs_super_generation(s))
1923
		++fail_gen;
A
Arne Jansen 已提交
1924

1925
	if (!scrub_check_fsid(s->fsid, spage))
1926
		++fail_cor;
A
Arne Jansen 已提交
1927

1928 1929 1930 1931
	shash->tfm = fs_info->csum_shash;
	crypto_shash_init(shash);
	crypto_shash_digest(shash, kaddr + BTRFS_CSUM_SIZE,
			BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE, calculated_csum);
1932

1933
	if (memcmp(calculated_csum, s->csum, sctx->fs_info->csum_size))
1934
		++fail_cor;
A
Arne Jansen 已提交
1935

1936
	if (fail_cor + fail_gen) {
A
Arne Jansen 已提交
1937 1938 1939 1940 1941
		/*
		 * if we find an error in a super block, we just report it.
		 * They will get written with the next transaction commit
		 * anyway
		 */
1942 1943 1944
		spin_lock(&sctx->stat_lock);
		++sctx->stat.super_errors;
		spin_unlock(&sctx->stat_lock);
1945
		if (fail_cor)
1946
			btrfs_dev_stat_inc_and_print(spage->dev,
1947 1948
				BTRFS_DEV_STAT_CORRUPTION_ERRS);
		else
1949
			btrfs_dev_stat_inc_and_print(spage->dev,
1950
				BTRFS_DEV_STAT_GENERATION_ERRS);
A
Arne Jansen 已提交
1951 1952
	}

1953
	return fail_cor + fail_gen;
A
Arne Jansen 已提交
1954 1955
}

1956 1957
static void scrub_block_get(struct scrub_block *sblock)
{
1958
	refcount_inc(&sblock->refs);
1959 1960 1961 1962
}

static void scrub_block_put(struct scrub_block *sblock)
{
1963
	if (refcount_dec_and_test(&sblock->refs)) {
1964 1965
		int i;

1966 1967 1968
		if (sblock->sparity)
			scrub_parity_put(sblock->sparity);

1969
		for (i = 0; i < sblock->page_count; i++)
1970
			scrub_page_put(sblock->pagev[i]);
1971 1972 1973 1974
		kfree(sblock);
	}
}

1975 1976
static void scrub_page_get(struct scrub_page *spage)
{
1977
	atomic_inc(&spage->refs);
1978 1979 1980 1981
}

static void scrub_page_put(struct scrub_page *spage)
{
1982
	if (atomic_dec_and_test(&spage->refs)) {
1983 1984 1985 1986 1987 1988
		if (spage->page)
			__free_page(spage->page);
		kfree(spage);
	}
}

1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047
/*
 * Throttling of IO submission, bandwidth-limit based, the timeslice is 1
 * second.  Limit can be set via /sys/fs/UUID/devinfo/devid/scrub_speed_max.
 */
static void scrub_throttle(struct scrub_ctx *sctx)
{
	const int time_slice = 1000;
	struct scrub_bio *sbio;
	struct btrfs_device *device;
	s64 delta;
	ktime_t now;
	u32 div;
	u64 bwlimit;

	sbio = sctx->bios[sctx->curr];
	device = sbio->dev;
	bwlimit = READ_ONCE(device->scrub_speed_max);
	if (bwlimit == 0)
		return;

	/*
	 * Slice is divided into intervals when the IO is submitted, adjust by
	 * bwlimit and maximum of 64 intervals.
	 */
	div = max_t(u32, 1, (u32)(bwlimit / (16 * 1024 * 1024)));
	div = min_t(u32, 64, div);

	/* Start new epoch, set deadline */
	now = ktime_get();
	if (sctx->throttle_deadline == 0) {
		sctx->throttle_deadline = ktime_add_ms(now, time_slice / div);
		sctx->throttle_sent = 0;
	}

	/* Still in the time to send? */
	if (ktime_before(now, sctx->throttle_deadline)) {
		/* If current bio is within the limit, send it */
		sctx->throttle_sent += sbio->bio->bi_iter.bi_size;
		if (sctx->throttle_sent <= div_u64(bwlimit, div))
			return;

		/* We're over the limit, sleep until the rest of the slice */
		delta = ktime_ms_delta(sctx->throttle_deadline, now);
	} else {
		/* New request after deadline, start new epoch */
		delta = 0;
	}

	if (delta) {
		long timeout;

		timeout = div_u64(delta * HZ, 1000);
		schedule_timeout_interruptible(timeout);
	}

	/* Next call will start the deadline period */
	sctx->throttle_deadline = 0;
}

2048
static void scrub_submit(struct scrub_ctx *sctx)
A
Arne Jansen 已提交
2049 2050 2051
{
	struct scrub_bio *sbio;

2052
	if (sctx->curr == -1)
S
Stefan Behrens 已提交
2053
		return;
A
Arne Jansen 已提交
2054

2055 2056
	scrub_throttle(sctx);

2057 2058
	sbio = sctx->bios[sctx->curr];
	sctx->curr = -1;
2059
	scrub_pending_bio_inc(sctx);
2060
	btrfsic_submit_bio(sbio->bio);
A
Arne Jansen 已提交
2061 2062
}

2063 2064
static int scrub_add_page_to_rd_bio(struct scrub_ctx *sctx,
				    struct scrub_page *spage)
A
Arne Jansen 已提交
2065
{
2066
	struct scrub_block *sblock = spage->sblock;
A
Arne Jansen 已提交
2067
	struct scrub_bio *sbio;
2068
	const u32 sectorsize = sctx->fs_info->sectorsize;
2069
	int ret;
A
Arne Jansen 已提交
2070 2071 2072 2073 2074

again:
	/*
	 * grab a fresh bio or wait for one to become available
	 */
2075 2076 2077 2078 2079 2080 2081 2082
	while (sctx->curr == -1) {
		spin_lock(&sctx->list_lock);
		sctx->curr = sctx->first_free;
		if (sctx->curr != -1) {
			sctx->first_free = sctx->bios[sctx->curr]->next_free;
			sctx->bios[sctx->curr]->next_free = -1;
			sctx->bios[sctx->curr]->page_count = 0;
			spin_unlock(&sctx->list_lock);
A
Arne Jansen 已提交
2083
		} else {
2084 2085
			spin_unlock(&sctx->list_lock);
			wait_event(sctx->list_wait, sctx->first_free != -1);
A
Arne Jansen 已提交
2086 2087
		}
	}
2088
	sbio = sctx->bios[sctx->curr];
2089
	if (sbio->page_count == 0) {
2090 2091
		struct bio *bio;

2092 2093
		sbio->physical = spage->physical;
		sbio->logical = spage->logical;
2094
		sbio->dev = spage->dev;
2095 2096
		bio = sbio->bio;
		if (!bio) {
2097
			bio = btrfs_bio_alloc(sctx->pages_per_bio);
2098 2099
			sbio->bio = bio;
		}
2100 2101 2102

		bio->bi_private = sbio;
		bio->bi_end_io = scrub_bio_end_io;
2103
		bio_set_dev(bio, sbio->dev->bdev);
2104
		bio->bi_iter.bi_sector = sbio->physical >> 9;
D
David Sterba 已提交
2105
		bio->bi_opf = REQ_OP_READ;
2106
		sbio->status = 0;
2107
	} else if (sbio->physical + sbio->page_count * sectorsize !=
2108
		   spage->physical ||
2109
		   sbio->logical + sbio->page_count * sectorsize !=
2110 2111
		   spage->logical ||
		   sbio->dev != spage->dev) {
2112
		scrub_submit(sctx);
A
Arne Jansen 已提交
2113 2114
		goto again;
	}
2115

2116
	sbio->pagev[sbio->page_count] = spage;
2117 2118
	ret = bio_add_page(sbio->bio, spage->page, sectorsize, 0);
	if (ret != sectorsize) {
2119 2120 2121 2122 2123
		if (sbio->page_count < 1) {
			bio_put(sbio->bio);
			sbio->bio = NULL;
			return -EIO;
		}
2124
		scrub_submit(sctx);
2125 2126 2127
		goto again;
	}

2128
	scrub_block_get(sblock); /* one for the page added to the bio */
2129 2130
	atomic_inc(&sblock->outstanding_pages);
	sbio->page_count++;
2131
	if (sbio->page_count == sctx->pages_per_bio)
2132
		scrub_submit(sctx);
2133 2134 2135 2136

	return 0;
}

2137
static void scrub_missing_raid56_end_io(struct bio *bio)
2138 2139
{
	struct scrub_block *sblock = bio->bi_private;
2140
	struct btrfs_fs_info *fs_info = sblock->sctx->fs_info;
2141

2142
	if (bio->bi_status)
2143 2144
		sblock->no_io_error_seen = 0;

2145 2146
	bio_put(bio);

2147 2148 2149 2150 2151 2152 2153
	btrfs_queue_work(fs_info->scrub_workers, &sblock->work);
}

static void scrub_missing_raid56_worker(struct btrfs_work *work)
{
	struct scrub_block *sblock = container_of(work, struct scrub_block, work);
	struct scrub_ctx *sctx = sblock->sctx;
2154
	struct btrfs_fs_info *fs_info = sctx->fs_info;
2155 2156 2157 2158 2159 2160
	u64 logical;
	struct btrfs_device *dev;

	logical = sblock->pagev[0]->logical;
	dev = sblock->pagev[0]->dev;

2161
	if (sblock->no_io_error_seen)
2162
		scrub_recheck_block_checksum(sblock);
2163 2164 2165 2166 2167

	if (!sblock->no_io_error_seen) {
		spin_lock(&sctx->stat_lock);
		sctx->stat.read_errors++;
		spin_unlock(&sctx->stat_lock);
2168
		btrfs_err_rl_in_rcu(fs_info,
2169
			"IO error rebuilding logical %llu for dev %s",
2170 2171 2172 2173 2174
			logical, rcu_str_deref(dev->name));
	} else if (sblock->header_error || sblock->checksum_error) {
		spin_lock(&sctx->stat_lock);
		sctx->stat.uncorrectable_errors++;
		spin_unlock(&sctx->stat_lock);
2175
		btrfs_err_rl_in_rcu(fs_info,
2176
			"failed to rebuild valid logical %llu for dev %s",
2177 2178 2179 2180 2181
			logical, rcu_str_deref(dev->name));
	} else {
		scrub_write_block_to_dev_replace(sblock);
	}

2182
	if (sctx->is_dev_replace && sctx->flush_all_writes) {
2183
		mutex_lock(&sctx->wr_lock);
2184
		scrub_wr_submit(sctx);
2185
		mutex_unlock(&sctx->wr_lock);
2186 2187
	}

2188
	scrub_block_put(sblock);
2189 2190 2191 2192 2193 2194
	scrub_pending_bio_dec(sctx);
}

static void scrub_missing_raid56_pages(struct scrub_block *sblock)
{
	struct scrub_ctx *sctx = sblock->sctx;
2195
	struct btrfs_fs_info *fs_info = sctx->fs_info;
2196 2197
	u64 length = sblock->page_count * PAGE_SIZE;
	u64 logical = sblock->pagev[0]->logical;
2198
	struct btrfs_io_context *bioc = NULL;
2199 2200 2201 2202 2203
	struct bio *bio;
	struct btrfs_raid_bio *rbio;
	int ret;
	int i;

2204
	btrfs_bio_counter_inc_blocked(fs_info);
2205
	ret = btrfs_map_sblock(fs_info, BTRFS_MAP_GET_READ_MIRRORS, logical,
2206 2207 2208
			       &length, &bioc);
	if (ret || !bioc || !bioc->raid_map)
		goto bioc_out;
2209 2210

	if (WARN_ON(!sctx->is_dev_replace ||
2211
		    !(bioc->map_type & BTRFS_BLOCK_GROUP_RAID56_MASK))) {
2212 2213 2214 2215 2216 2217
		/*
		 * We shouldn't be scrubbing a missing device. Even for dev
		 * replace, we should only get here for RAID 5/6. We either
		 * managed to mount something with no mirrors remaining or
		 * there's a bug in scrub_remap_extent()/btrfs_map_block().
		 */
2218
		goto bioc_out;
2219 2220
	}

2221
	bio = btrfs_bio_alloc(BIO_MAX_VECS);
2222 2223 2224 2225
	bio->bi_iter.bi_sector = logical >> 9;
	bio->bi_private = sblock;
	bio->bi_end_io = scrub_missing_raid56_end_io;

2226
	rbio = raid56_alloc_missing_rbio(bio, bioc, length);
2227 2228 2229 2230 2231 2232 2233 2234 2235
	if (!rbio)
		goto rbio_out;

	for (i = 0; i < sblock->page_count; i++) {
		struct scrub_page *spage = sblock->pagev[i];

		raid56_add_scrub_pages(rbio, spage->page, spage->logical);
	}

2236
	btrfs_init_work(&sblock->work, scrub_missing_raid56_worker, NULL, NULL);
2237 2238 2239 2240 2241 2242 2243
	scrub_block_get(sblock);
	scrub_pending_bio_inc(sctx);
	raid56_submit_missing_rbio(rbio);
	return;

rbio_out:
	bio_put(bio);
2244
bioc_out:
2245
	btrfs_bio_counter_dec(fs_info);
2246
	btrfs_put_bioc(bioc);
2247 2248 2249 2250 2251
	spin_lock(&sctx->stat_lock);
	sctx->stat.malloc_errors++;
	spin_unlock(&sctx->stat_lock);
}

2252
static int scrub_pages(struct scrub_ctx *sctx, u64 logical, u32 len,
2253
		       u64 physical, struct btrfs_device *dev, u64 flags,
2254
		       u64 gen, int mirror_num, u8 *csum,
2255
		       u64 physical_for_dev_replace)
2256 2257
{
	struct scrub_block *sblock;
2258
	const u32 sectorsize = sctx->fs_info->sectorsize;
2259 2260
	int index;

2261
	sblock = kzalloc(sizeof(*sblock), GFP_KERNEL);
2262
	if (!sblock) {
2263 2264 2265
		spin_lock(&sctx->stat_lock);
		sctx->stat.malloc_errors++;
		spin_unlock(&sctx->stat_lock);
2266
		return -ENOMEM;
A
Arne Jansen 已提交
2267
	}
2268

2269 2270
	/* one ref inside this function, plus one for each page added to
	 * a bio later on */
2271
	refcount_set(&sblock->refs, 1);
2272
	sblock->sctx = sctx;
2273 2274 2275
	sblock->no_io_error_seen = 1;

	for (index = 0; len > 0; index++) {
2276
		struct scrub_page *spage;
2277 2278 2279 2280 2281 2282
		/*
		 * Here we will allocate one page for one sector to scrub.
		 * This is fine if PAGE_SIZE == sectorsize, but will cost
		 * more memory for PAGE_SIZE > sectorsize case.
		 */
		u32 l = min(sectorsize, len);
2283

2284
		spage = kzalloc(sizeof(*spage), GFP_KERNEL);
2285 2286
		if (!spage) {
leave_nomem:
2287 2288 2289
			spin_lock(&sctx->stat_lock);
			sctx->stat.malloc_errors++;
			spin_unlock(&sctx->stat_lock);
2290
			scrub_block_put(sblock);
2291 2292
			return -ENOMEM;
		}
2293
		ASSERT(index < SCRUB_MAX_PAGES_PER_BLOCK);
2294 2295
		scrub_page_get(spage);
		sblock->pagev[index] = spage;
2296
		spage->sblock = sblock;
2297
		spage->dev = dev;
2298 2299 2300 2301
		spage->flags = flags;
		spage->generation = gen;
		spage->logical = logical;
		spage->physical = physical;
2302
		spage->physical_for_dev_replace = physical_for_dev_replace;
2303 2304 2305
		spage->mirror_num = mirror_num;
		if (csum) {
			spage->have_csum = 1;
2306
			memcpy(spage->csum, csum, sctx->fs_info->csum_size);
2307 2308 2309 2310
		} else {
			spage->have_csum = 0;
		}
		sblock->page_count++;
2311
		spage->page = alloc_page(GFP_KERNEL);
2312 2313
		if (!spage->page)
			goto leave_nomem;
2314 2315 2316
		len -= l;
		logical += l;
		physical += l;
2317
		physical_for_dev_replace += l;
2318 2319
	}

2320
	WARN_ON(sblock->page_count == 0);
2321
	if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) {
2322 2323 2324 2325 2326 2327 2328 2329 2330
		/*
		 * This case should only be hit for RAID 5/6 device replace. See
		 * the comment in scrub_missing_raid56_pages() for details.
		 */
		scrub_missing_raid56_pages(sblock);
	} else {
		for (index = 0; index < sblock->page_count; index++) {
			struct scrub_page *spage = sblock->pagev[index];
			int ret;
2331

2332 2333 2334 2335 2336
			ret = scrub_add_page_to_rd_bio(sctx, spage);
			if (ret) {
				scrub_block_put(sblock);
				return ret;
			}
2337
		}
A
Arne Jansen 已提交
2338

2339
		if (flags & BTRFS_EXTENT_FLAG_SUPER)
2340 2341
			scrub_submit(sctx);
	}
A
Arne Jansen 已提交
2342

2343 2344
	/* last one frees, either here or in bio completion for last page */
	scrub_block_put(sblock);
A
Arne Jansen 已提交
2345 2346 2347
	return 0;
}

2348
static void scrub_bio_end_io(struct bio *bio)
2349 2350
{
	struct scrub_bio *sbio = bio->bi_private;
2351
	struct btrfs_fs_info *fs_info = sbio->dev->fs_info;
2352

2353
	sbio->status = bio->bi_status;
2354 2355
	sbio->bio = bio;

2356
	btrfs_queue_work(fs_info->scrub_workers, &sbio->work);
2357 2358 2359 2360 2361
}

static void scrub_bio_end_io_worker(struct btrfs_work *work)
{
	struct scrub_bio *sbio = container_of(work, struct scrub_bio, work);
2362
	struct scrub_ctx *sctx = sbio->sctx;
2363 2364
	int i;

2365
	ASSERT(sbio->page_count <= SCRUB_PAGES_PER_BIO);
2366
	if (sbio->status) {
2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386
		for (i = 0; i < sbio->page_count; i++) {
			struct scrub_page *spage = sbio->pagev[i];

			spage->io_error = 1;
			spage->sblock->no_io_error_seen = 0;
		}
	}

	/* now complete the scrub_block items that have all pages completed */
	for (i = 0; i < sbio->page_count; i++) {
		struct scrub_page *spage = sbio->pagev[i];
		struct scrub_block *sblock = spage->sblock;

		if (atomic_dec_and_test(&sblock->outstanding_pages))
			scrub_block_complete(sblock);
		scrub_block_put(sblock);
	}

	bio_put(sbio->bio);
	sbio->bio = NULL;
2387 2388 2389 2390
	spin_lock(&sctx->list_lock);
	sbio->next_free = sctx->first_free;
	sctx->first_free = sbio->index;
	spin_unlock(&sctx->list_lock);
2391

2392
	if (sctx->is_dev_replace && sctx->flush_all_writes) {
2393
		mutex_lock(&sctx->wr_lock);
2394
		scrub_wr_submit(sctx);
2395
		mutex_unlock(&sctx->wr_lock);
2396 2397
	}

2398
	scrub_pending_bio_dec(sctx);
2399 2400
}

2401 2402
static inline void __scrub_mark_bitmap(struct scrub_parity *sparity,
				       unsigned long *bitmap,
2403
				       u64 start, u32 len)
2404
{
2405
	u64 offset;
2406
	u32 nsectors;
2407
	u32 sectorsize_bits = sparity->sctx->fs_info->sectorsize_bits;
2408 2409 2410 2411 2412 2413 2414

	if (len >= sparity->stripe_len) {
		bitmap_set(bitmap, 0, sparity->nsectors);
		return;
	}

	start -= sparity->logic_start;
2415
	start = div64_u64_rem(start, sparity->stripe_len, &offset);
2416
	offset = offset >> sectorsize_bits;
2417
	nsectors = len >> sectorsize_bits;
2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428

	if (offset + nsectors <= sparity->nsectors) {
		bitmap_set(bitmap, offset, nsectors);
		return;
	}

	bitmap_set(bitmap, offset, sparity->nsectors - offset);
	bitmap_set(bitmap, 0, nsectors - (sparity->nsectors - offset));
}

static inline void scrub_parity_mark_sectors_error(struct scrub_parity *sparity,
2429
						   u64 start, u32 len)
2430 2431 2432 2433 2434
{
	__scrub_mark_bitmap(sparity, sparity->ebitmap, start, len);
}

static inline void scrub_parity_mark_sectors_data(struct scrub_parity *sparity,
2435
						  u64 start, u32 len)
2436 2437 2438 2439
{
	__scrub_mark_bitmap(sparity, sparity->dbitmap, start, len);
}

2440 2441
static void scrub_block_complete(struct scrub_block *sblock)
{
2442 2443
	int corrupted = 0;

2444
	if (!sblock->no_io_error_seen) {
2445
		corrupted = 1;
2446
		scrub_handle_errored_block(sblock);
2447 2448 2449 2450 2451 2452
	} else {
		/*
		 * if has checksum error, write via repair mechanism in
		 * dev replace case, otherwise write here in dev replace
		 * case.
		 */
2453 2454
		corrupted = scrub_checksum(sblock);
		if (!corrupted && sblock->sctx->is_dev_replace)
2455 2456
			scrub_write_block_to_dev_replace(sblock);
	}
2457 2458 2459 2460

	if (sblock->sparity && corrupted && !sblock->data_corrected) {
		u64 start = sblock->pagev[0]->logical;
		u64 end = sblock->pagev[sblock->page_count - 1]->logical +
2461
			  sblock->sctx->fs_info->sectorsize;
2462

2463
		ASSERT(end - start <= U32_MAX);
2464 2465 2466
		scrub_parity_mark_sectors_error(sblock->sparity,
						start, end - start);
	}
2467 2468
}

2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480
static void drop_csum_range(struct scrub_ctx *sctx, struct btrfs_ordered_sum *sum)
{
	sctx->stat.csum_discards += sum->len >> sctx->fs_info->sectorsize_bits;
	list_del(&sum->list);
	kfree(sum);
}

/*
 * Find the desired csum for range [logical, logical + sectorsize), and store
 * the csum into @csum.
 *
 * The search source is sctx->csum_list, which is a pre-populated list
D
David Sterba 已提交
2481
 * storing bytenr ordered csum ranges.  We're responsible to cleanup any range
2482 2483 2484 2485 2486
 * that is before @logical.
 *
 * Return 0 if there is no csum for the range.
 * Return 1 if there is csum for the range and copied to @csum.
 */
2487
static int scrub_find_csum(struct scrub_ctx *sctx, u64 logical, u8 *csum)
A
Arne Jansen 已提交
2488
{
2489
	bool found = false;
A
Arne Jansen 已提交
2490

2491
	while (!list_empty(&sctx->csum_list)) {
2492 2493 2494 2495
		struct btrfs_ordered_sum *sum = NULL;
		unsigned long index;
		unsigned long num_sectors;

2496
		sum = list_first_entry(&sctx->csum_list,
A
Arne Jansen 已提交
2497
				       struct btrfs_ordered_sum, list);
2498
		/* The current csum range is beyond our range, no csum found */
A
Arne Jansen 已提交
2499 2500 2501
		if (sum->bytenr > logical)
			break;

2502 2503 2504 2505 2506 2507 2508 2509 2510 2511
		/*
		 * The current sum is before our bytenr, since scrub is always
		 * done in bytenr order, the csum will never be used anymore,
		 * clean it up so that later calls won't bother with the range,
		 * and continue search the next range.
		 */
		if (sum->bytenr + sum->len <= logical) {
			drop_csum_range(sctx, sum);
			continue;
		}
A
Arne Jansen 已提交
2512

2513 2514 2515 2516
		/* Now the csum range covers our bytenr, copy the csum */
		found = true;
		index = (logical - sum->bytenr) >> sctx->fs_info->sectorsize_bits;
		num_sectors = sum->len >> sctx->fs_info->sectorsize_bits;
2517

2518 2519 2520 2521 2522 2523 2524
		memcpy(csum, sum->sums + index * sctx->fs_info->csum_size,
		       sctx->fs_info->csum_size);

		/* Cleanup the range if we're at the end of the csum range */
		if (index == num_sectors - 1)
			drop_csum_range(sctx, sum);
		break;
A
Arne Jansen 已提交
2525
	}
2526 2527
	if (!found)
		return 0;
2528
	return 1;
A
Arne Jansen 已提交
2529 2530 2531
}

/* scrub extent tries to collect up to 64 kB for each bio */
L
Liu Bo 已提交
2532
static int scrub_extent(struct scrub_ctx *sctx, struct map_lookup *map,
2533
			u64 logical, u32 len,
2534
			u64 physical, struct btrfs_device *dev, u64 flags,
2535
			u64 gen, int mirror_num, u64 physical_for_dev_replace)
A
Arne Jansen 已提交
2536 2537 2538
{
	int ret;
	u8 csum[BTRFS_CSUM_SIZE];
2539 2540 2541
	u32 blocksize;

	if (flags & BTRFS_EXTENT_FLAG_DATA) {
L
Liu Bo 已提交
2542 2543 2544 2545
		if (map->type & BTRFS_BLOCK_GROUP_RAID56_MASK)
			blocksize = map->stripe_len;
		else
			blocksize = sctx->fs_info->sectorsize;
2546 2547 2548 2549
		spin_lock(&sctx->stat_lock);
		sctx->stat.data_extents_scrubbed++;
		sctx->stat.data_bytes_scrubbed += len;
		spin_unlock(&sctx->stat_lock);
2550
	} else if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
L
Liu Bo 已提交
2551 2552 2553 2554
		if (map->type & BTRFS_BLOCK_GROUP_RAID56_MASK)
			blocksize = map->stripe_len;
		else
			blocksize = sctx->fs_info->nodesize;
2555 2556 2557 2558
		spin_lock(&sctx->stat_lock);
		sctx->stat.tree_extents_scrubbed++;
		sctx->stat.tree_bytes_scrubbed += len;
		spin_unlock(&sctx->stat_lock);
2559
	} else {
2560
		blocksize = sctx->fs_info->sectorsize;
2561
		WARN_ON(1);
2562
	}
A
Arne Jansen 已提交
2563 2564

	while (len) {
2565
		u32 l = min(len, blocksize);
A
Arne Jansen 已提交
2566 2567 2568 2569
		int have_csum = 0;

		if (flags & BTRFS_EXTENT_FLAG_DATA) {
			/* push csums to sbio */
2570
			have_csum = scrub_find_csum(sctx, logical, csum);
A
Arne Jansen 已提交
2571
			if (have_csum == 0)
2572
				++sctx->stat.no_csum;
A
Arne Jansen 已提交
2573
		}
2574
		ret = scrub_pages(sctx, logical, l, physical, dev, flags, gen,
2575
				  mirror_num, have_csum ? csum : NULL,
2576
				  physical_for_dev_replace);
A
Arne Jansen 已提交
2577 2578 2579 2580 2581
		if (ret)
			return ret;
		len -= l;
		logical += l;
		physical += l;
2582
		physical_for_dev_replace += l;
A
Arne Jansen 已提交
2583 2584 2585 2586
	}
	return 0;
}

2587
static int scrub_pages_for_parity(struct scrub_parity *sparity,
2588
				  u64 logical, u32 len,
2589 2590 2591 2592 2593
				  u64 physical, struct btrfs_device *dev,
				  u64 flags, u64 gen, int mirror_num, u8 *csum)
{
	struct scrub_ctx *sctx = sparity->sctx;
	struct scrub_block *sblock;
2594
	const u32 sectorsize = sctx->fs_info->sectorsize;
2595 2596
	int index;

2597 2598
	ASSERT(IS_ALIGNED(len, sectorsize));

2599
	sblock = kzalloc(sizeof(*sblock), GFP_KERNEL);
2600 2601 2602 2603 2604 2605 2606 2607 2608
	if (!sblock) {
		spin_lock(&sctx->stat_lock);
		sctx->stat.malloc_errors++;
		spin_unlock(&sctx->stat_lock);
		return -ENOMEM;
	}

	/* one ref inside this function, plus one for each page added to
	 * a bio later on */
2609
	refcount_set(&sblock->refs, 1);
2610 2611 2612 2613 2614 2615 2616 2617
	sblock->sctx = sctx;
	sblock->no_io_error_seen = 1;
	sblock->sparity = sparity;
	scrub_parity_get(sparity);

	for (index = 0; len > 0; index++) {
		struct scrub_page *spage;

2618
		spage = kzalloc(sizeof(*spage), GFP_KERNEL);
2619 2620 2621 2622 2623 2624 2625 2626
		if (!spage) {
leave_nomem:
			spin_lock(&sctx->stat_lock);
			sctx->stat.malloc_errors++;
			spin_unlock(&sctx->stat_lock);
			scrub_block_put(sblock);
			return -ENOMEM;
		}
2627
		ASSERT(index < SCRUB_MAX_PAGES_PER_BLOCK);
2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642
		/* For scrub block */
		scrub_page_get(spage);
		sblock->pagev[index] = spage;
		/* For scrub parity */
		scrub_page_get(spage);
		list_add_tail(&spage->list, &sparity->spages);
		spage->sblock = sblock;
		spage->dev = dev;
		spage->flags = flags;
		spage->generation = gen;
		spage->logical = logical;
		spage->physical = physical;
		spage->mirror_num = mirror_num;
		if (csum) {
			spage->have_csum = 1;
2643
			memcpy(spage->csum, csum, sctx->fs_info->csum_size);
2644 2645 2646 2647
		} else {
			spage->have_csum = 0;
		}
		sblock->page_count++;
2648
		spage->page = alloc_page(GFP_KERNEL);
2649 2650
		if (!spage->page)
			goto leave_nomem;
2651 2652 2653 2654 2655 2656


		/* Iterate over the stripe range in sectorsize steps */
		len -= sectorsize;
		logical += sectorsize;
		physical += sectorsize;
2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676
	}

	WARN_ON(sblock->page_count == 0);
	for (index = 0; index < sblock->page_count; index++) {
		struct scrub_page *spage = sblock->pagev[index];
		int ret;

		ret = scrub_add_page_to_rd_bio(sctx, spage);
		if (ret) {
			scrub_block_put(sblock);
			return ret;
		}
	}

	/* last one frees, either here or in bio completion for last page */
	scrub_block_put(sblock);
	return 0;
}

static int scrub_extent_for_parity(struct scrub_parity *sparity,
2677
				   u64 logical, u32 len,
2678 2679 2680 2681 2682 2683 2684 2685
				   u64 physical, struct btrfs_device *dev,
				   u64 flags, u64 gen, int mirror_num)
{
	struct scrub_ctx *sctx = sparity->sctx;
	int ret;
	u8 csum[BTRFS_CSUM_SIZE];
	u32 blocksize;

2686
	if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) {
2687 2688 2689 2690
		scrub_parity_mark_sectors_error(sparity, logical, len);
		return 0;
	}

2691
	if (flags & BTRFS_EXTENT_FLAG_DATA) {
L
Liu Bo 已提交
2692
		blocksize = sparity->stripe_len;
2693
	} else if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
L
Liu Bo 已提交
2694
		blocksize = sparity->stripe_len;
2695
	} else {
2696
		blocksize = sctx->fs_info->sectorsize;
2697 2698 2699 2700
		WARN_ON(1);
	}

	while (len) {
2701
		u32 l = min(len, blocksize);
2702 2703 2704 2705
		int have_csum = 0;

		if (flags & BTRFS_EXTENT_FLAG_DATA) {
			/* push csums to sbio */
2706
			have_csum = scrub_find_csum(sctx, logical, csum);
2707 2708 2709 2710 2711 2712 2713 2714
			if (have_csum == 0)
				goto skip;
		}
		ret = scrub_pages_for_parity(sparity, logical, l, physical, dev,
					     flags, gen, mirror_num,
					     have_csum ? csum : NULL);
		if (ret)
			return ret;
2715
skip:
2716 2717 2718 2719 2720 2721 2722
		len -= l;
		logical += l;
		physical += l;
	}
	return 0;
}

2723 2724 2725 2726 2727 2728 2729 2730
/*
 * Given a physical address, this will calculate it's
 * logical offset. if this is a parity stripe, it will return
 * the most left data stripe's logical offset.
 *
 * return 0 if it is a data stripe, 1 means parity stripe.
 */
static int get_raid56_logic_offset(u64 physical, int num,
2731 2732
				   struct map_lookup *map, u64 *offset,
				   u64 *stripe_start)
2733 2734 2735 2736 2737
{
	int i;
	int j = 0;
	u64 stripe_nr;
	u64 last_offset;
2738 2739
	u32 stripe_index;
	u32 rot;
2740
	const int data_stripes = nr_data_stripes(map);
2741

2742
	last_offset = (physical - map->stripes[num].physical) * data_stripes;
2743 2744 2745
	if (stripe_start)
		*stripe_start = last_offset;

2746
	*offset = last_offset;
2747
	for (i = 0; i < data_stripes; i++) {
2748 2749
		*offset = last_offset + i * map->stripe_len;

2750
		stripe_nr = div64_u64(*offset, map->stripe_len);
2751
		stripe_nr = div_u64(stripe_nr, data_stripes);
2752 2753

		/* Work out the disk rotation on this stripe-set */
2754
		stripe_nr = div_u64_rem(stripe_nr, map->num_stripes, &rot);
2755 2756
		/* calculate which stripe this data locates */
		rot += i;
2757
		stripe_index = rot % map->num_stripes;
2758 2759 2760 2761 2762 2763 2764 2765 2766
		if (stripe_index == num)
			return 0;
		if (stripe_index < num)
			j++;
	}
	*offset = last_offset + j * map->stripe_len;
	return 1;
}

2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788
static void scrub_free_parity(struct scrub_parity *sparity)
{
	struct scrub_ctx *sctx = sparity->sctx;
	struct scrub_page *curr, *next;
	int nbits;

	nbits = bitmap_weight(sparity->ebitmap, sparity->nsectors);
	if (nbits) {
		spin_lock(&sctx->stat_lock);
		sctx->stat.read_errors += nbits;
		sctx->stat.uncorrectable_errors += nbits;
		spin_unlock(&sctx->stat_lock);
	}

	list_for_each_entry_safe(curr, next, &sparity->spages, list) {
		list_del_init(&curr->list);
		scrub_page_put(curr);
	}

	kfree(sparity);
}

2789 2790 2791 2792 2793 2794 2795 2796 2797 2798
static void scrub_parity_bio_endio_worker(struct btrfs_work *work)
{
	struct scrub_parity *sparity = container_of(work, struct scrub_parity,
						    work);
	struct scrub_ctx *sctx = sparity->sctx;

	scrub_free_parity(sparity);
	scrub_pending_bio_dec(sctx);
}

2799
static void scrub_parity_bio_endio(struct bio *bio)
2800 2801
{
	struct scrub_parity *sparity = (struct scrub_parity *)bio->bi_private;
2802
	struct btrfs_fs_info *fs_info = sparity->sctx->fs_info;
2803

2804
	if (bio->bi_status)
2805 2806 2807 2808
		bitmap_or(sparity->ebitmap, sparity->ebitmap, sparity->dbitmap,
			  sparity->nsectors);

	bio_put(bio);
2809

2810 2811
	btrfs_init_work(&sparity->work, scrub_parity_bio_endio_worker, NULL,
			NULL);
2812
	btrfs_queue_work(fs_info->scrub_parity_workers, &sparity->work);
2813 2814 2815 2816 2817
}

static void scrub_parity_check_and_repair(struct scrub_parity *sparity)
{
	struct scrub_ctx *sctx = sparity->sctx;
2818
	struct btrfs_fs_info *fs_info = sctx->fs_info;
2819 2820
	struct bio *bio;
	struct btrfs_raid_bio *rbio;
2821
	struct btrfs_io_context *bioc = NULL;
2822 2823 2824 2825 2826 2827 2828
	u64 length;
	int ret;

	if (!bitmap_andnot(sparity->dbitmap, sparity->dbitmap, sparity->ebitmap,
			   sparity->nsectors))
		goto out;

2829
	length = sparity->logic_end - sparity->logic_start;
2830 2831

	btrfs_bio_counter_inc_blocked(fs_info);
2832
	ret = btrfs_map_sblock(fs_info, BTRFS_MAP_WRITE, sparity->logic_start,
2833 2834 2835
			       &length, &bioc);
	if (ret || !bioc || !bioc->raid_map)
		goto bioc_out;
2836

2837
	bio = btrfs_bio_alloc(BIO_MAX_VECS);
2838 2839 2840 2841
	bio->bi_iter.bi_sector = sparity->logic_start >> 9;
	bio->bi_private = sparity;
	bio->bi_end_io = scrub_parity_bio_endio;

2842 2843
	rbio = raid56_parity_alloc_scrub_rbio(bio, bioc, length,
					      sparity->scrub_dev,
2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854
					      sparity->dbitmap,
					      sparity->nsectors);
	if (!rbio)
		goto rbio_out;

	scrub_pending_bio_inc(sctx);
	raid56_parity_submit_scrub_rbio(rbio);
	return;

rbio_out:
	bio_put(bio);
2855
bioc_out:
2856
	btrfs_bio_counter_dec(fs_info);
2857
	btrfs_put_bioc(bioc);
2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868
	bitmap_or(sparity->ebitmap, sparity->ebitmap, sparity->dbitmap,
		  sparity->nsectors);
	spin_lock(&sctx->stat_lock);
	sctx->stat.malloc_errors++;
	spin_unlock(&sctx->stat_lock);
out:
	scrub_free_parity(sparity);
}

static inline int scrub_calc_parity_bitmap_len(int nsectors)
{
2869
	return DIV_ROUND_UP(nsectors, BITS_PER_LONG) * sizeof(long);
2870 2871 2872 2873
}

static void scrub_parity_get(struct scrub_parity *sparity)
{
2874
	refcount_inc(&sparity->refs);
2875 2876 2877 2878
}

static void scrub_parity_put(struct scrub_parity *sparity)
{
2879
	if (!refcount_dec_and_test(&sparity->refs))
2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891
		return;

	scrub_parity_check_and_repair(sparity);
}

static noinline_for_stack int scrub_raid56_parity(struct scrub_ctx *sctx,
						  struct map_lookup *map,
						  struct btrfs_device *sdev,
						  struct btrfs_path *path,
						  u64 logic_start,
						  u64 logic_end)
{
2892
	struct btrfs_fs_info *fs_info = sctx->fs_info;
2893
	struct btrfs_root *root = btrfs_extent_root(fs_info, logic_start);
2894
	struct btrfs_root *csum_root;
2895
	struct btrfs_extent_item *extent;
2896
	struct btrfs_io_context *bioc = NULL;
2897 2898 2899 2900 2901 2902 2903 2904
	u64 flags;
	int ret;
	int slot;
	struct extent_buffer *l;
	struct btrfs_key key;
	u64 generation;
	u64 extent_logical;
	u64 extent_physical;
2905 2906
	/* Check the comment in scrub_stripe() for why u32 is enough here */
	u32 extent_len;
2907
	u64 mapped_length;
2908 2909 2910 2911 2912 2913 2914
	struct btrfs_device *extent_dev;
	struct scrub_parity *sparity;
	int nsectors;
	int bitmap_len;
	int extent_mirror_num;
	int stop_loop = 0;

2915
	ASSERT(map->stripe_len <= U32_MAX);
2916
	nsectors = map->stripe_len >> fs_info->sectorsize_bits;
2917 2918 2919 2920 2921 2922 2923 2924 2925 2926
	bitmap_len = scrub_calc_parity_bitmap_len(nsectors);
	sparity = kzalloc(sizeof(struct scrub_parity) + 2 * bitmap_len,
			  GFP_NOFS);
	if (!sparity) {
		spin_lock(&sctx->stat_lock);
		sctx->stat.malloc_errors++;
		spin_unlock(&sctx->stat_lock);
		return -ENOMEM;
	}

2927
	ASSERT(map->stripe_len <= U32_MAX);
2928 2929 2930 2931 2932 2933
	sparity->stripe_len = map->stripe_len;
	sparity->nsectors = nsectors;
	sparity->sctx = sctx;
	sparity->scrub_dev = sdev;
	sparity->logic_start = logic_start;
	sparity->logic_end = logic_end;
2934
	refcount_set(&sparity->refs, 1);
2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982
	INIT_LIST_HEAD(&sparity->spages);
	sparity->dbitmap = sparity->bitmap;
	sparity->ebitmap = (void *)sparity->bitmap + bitmap_len;

	ret = 0;
	while (logic_start < logic_end) {
		if (btrfs_fs_incompat(fs_info, SKINNY_METADATA))
			key.type = BTRFS_METADATA_ITEM_KEY;
		else
			key.type = BTRFS_EXTENT_ITEM_KEY;
		key.objectid = logic_start;
		key.offset = (u64)-1;

		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
		if (ret < 0)
			goto out;

		if (ret > 0) {
			ret = btrfs_previous_extent_item(root, path, 0);
			if (ret < 0)
				goto out;
			if (ret > 0) {
				btrfs_release_path(path);
				ret = btrfs_search_slot(NULL, root, &key,
							path, 0, 0);
				if (ret < 0)
					goto out;
			}
		}

		stop_loop = 0;
		while (1) {
			u64 bytes;

			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;

				stop_loop = 1;
				break;
			}
			btrfs_item_key_to_cpu(l, &key, slot);

2983 2984 2985 2986
			if (key.type != BTRFS_EXTENT_ITEM_KEY &&
			    key.type != BTRFS_METADATA_ITEM_KEY)
				goto next;

2987
			if (key.type == BTRFS_METADATA_ITEM_KEY)
2988
				bytes = fs_info->nodesize;
2989 2990 2991 2992 2993 2994
			else
				bytes = key.offset;

			if (key.objectid + bytes <= logic_start)
				goto next;

2995
			if (key.objectid >= logic_end) {
2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007
				stop_loop = 1;
				break;
			}

			while (key.objectid >= logic_start + map->stripe_len)
				logic_start += map->stripe_len;

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

3008 3009 3010 3011
			if ((flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) &&
			    (key.objectid < logic_start ||
			     key.objectid + bytes >
			     logic_start + map->stripe_len)) {
J
Jeff Mahoney 已提交
3012 3013
				btrfs_err(fs_info,
					  "scrub: tree block %llu spanning stripes, ignored. logical=%llu",
3014
					  key.objectid, logic_start);
3015 3016 3017
				spin_lock(&sctx->stat_lock);
				sctx->stat.uncorrectable_errors++;
				spin_unlock(&sctx->stat_lock);
3018 3019 3020 3021
				goto next;
			}
again:
			extent_logical = key.objectid;
3022
			ASSERT(bytes <= U32_MAX);
3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037
			extent_len = bytes;

			if (extent_logical < logic_start) {
				extent_len -= logic_start - extent_logical;
				extent_logical = logic_start;
			}

			if (extent_logical + extent_len >
			    logic_start + map->stripe_len)
				extent_len = logic_start + map->stripe_len -
					     extent_logical;

			scrub_parity_mark_sectors_data(sparity, extent_logical,
						       extent_len);

3038
			mapped_length = extent_len;
3039
			bioc = NULL;
3040
			ret = btrfs_map_block(fs_info, BTRFS_MAP_READ,
3041
					extent_logical, &mapped_length, &bioc,
3042
					0);
3043
			if (!ret) {
3044
				if (!bioc || mapped_length < extent_len)
3045 3046 3047
					ret = -EIO;
			}
			if (ret) {
3048
				btrfs_put_bioc(bioc);
3049 3050
				goto out;
			}
3051 3052 3053 3054
			extent_physical = bioc->stripes[0].physical;
			extent_mirror_num = bioc->mirror_num;
			extent_dev = bioc->stripes[0].dev;
			btrfs_put_bioc(bioc);
3055

3056
			csum_root = btrfs_csum_root(fs_info, extent_logical);
3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069
			ret = btrfs_lookup_csums_range(csum_root,
						extent_logical,
						extent_logical + extent_len - 1,
						&sctx->csum_list, 1);
			if (ret)
				goto out;

			ret = scrub_extent_for_parity(sparity, extent_logical,
						      extent_len,
						      extent_physical,
						      extent_dev, flags,
						      generation,
						      extent_mirror_num);
3070 3071 3072

			scrub_free_csums(sctx);

3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101
			if (ret)
				goto out;

			if (extent_logical + extent_len <
			    key.objectid + bytes) {
				logic_start += map->stripe_len;

				if (logic_start >= logic_end) {
					stop_loop = 1;
					break;
				}

				if (logic_start < key.objectid + bytes) {
					cond_resched();
					goto again;
				}
			}
next:
			path->slots[0]++;
		}

		btrfs_release_path(path);

		if (stop_loop)
			break;

		logic_start += map->stripe_len;
	}
out:
3102 3103
	if (ret < 0) {
		ASSERT(logic_end - logic_start <= U32_MAX);
3104
		scrub_parity_mark_sectors_error(sparity, logic_start,
3105
						logic_end - logic_start);
3106
	}
3107 3108
	scrub_parity_put(sparity);
	scrub_submit(sctx);
3109
	mutex_lock(&sctx->wr_lock);
3110
	scrub_wr_submit(sctx);
3111
	mutex_unlock(&sctx->wr_lock);
3112 3113 3114 3115 3116

	btrfs_release_path(path);
	return ret < 0 ? ret : 0;
}

3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130
static void sync_replace_for_zoned(struct scrub_ctx *sctx)
{
	if (!btrfs_is_zoned(sctx->fs_info))
		return;

	sctx->flush_all_writes = true;
	scrub_submit(sctx);
	mutex_lock(&sctx->wr_lock);
	scrub_wr_submit(sctx);
	mutex_unlock(&sctx->wr_lock);

	wait_event(sctx->list_wait, atomic_read(&sctx->bios_in_flight) == 0);
}

3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156
static int sync_write_pointer_for_zoned(struct scrub_ctx *sctx, u64 logical,
					u64 physical, u64 physical_end)
{
	struct btrfs_fs_info *fs_info = sctx->fs_info;
	int ret = 0;

	if (!btrfs_is_zoned(fs_info))
		return 0;

	wait_event(sctx->list_wait, atomic_read(&sctx->bios_in_flight) == 0);

	mutex_lock(&sctx->wr_lock);
	if (sctx->write_pointer < physical_end) {
		ret = btrfs_sync_zone_write_pointer(sctx->wr_tgtdev, logical,
						    physical,
						    sctx->write_pointer);
		if (ret)
			btrfs_err(fs_info,
				  "zoned: failed to recover write pointer");
	}
	mutex_unlock(&sctx->wr_lock);
	btrfs_dev_clear_zone_empty(sctx->wr_tgtdev, physical);

	return ret;
}

3157
static noinline_for_stack int scrub_stripe(struct scrub_ctx *sctx,
3158 3159
					   struct map_lookup *map,
					   struct btrfs_device *scrub_dev,
3160 3161
					   int num, u64 base, u64 length,
					   struct btrfs_block_group *cache)
A
Arne Jansen 已提交
3162
{
3163
	struct btrfs_path *path, *ppath;
3164
	struct btrfs_fs_info *fs_info = sctx->fs_info;
3165
	struct btrfs_root *root;
3166
	struct btrfs_root *csum_root;
A
Arne Jansen 已提交
3167
	struct btrfs_extent_item *extent;
3168
	struct blk_plug plug;
A
Arne Jansen 已提交
3169 3170 3171 3172 3173 3174 3175
	u64 flags;
	int ret;
	int slot;
	u64 nstripes;
	struct extent_buffer *l;
	u64 physical;
	u64 logical;
L
Liu Bo 已提交
3176
	u64 logic_end;
3177
	u64 physical_end;
A
Arne Jansen 已提交
3178
	u64 generation;
3179
	int mirror_num;
A
Arne Jansen 已提交
3180 3181
	struct reada_control *reada1;
	struct reada_control *reada2;
3182
	struct btrfs_key key;
A
Arne Jansen 已提交
3183
	struct btrfs_key key_end;
A
Arne Jansen 已提交
3184 3185
	u64 increment = map->stripe_len;
	u64 offset;
3186 3187
	u64 extent_logical;
	u64 extent_physical;
3188 3189 3190 3191 3192
	/*
	 * Unlike chunk length, extent length should never go beyond
	 * BTRFS_MAX_EXTENT_SIZE, thus u32 is enough here.
	 */
	u32 extent_len;
3193 3194
	u64 stripe_logical;
	u64 stripe_end;
3195 3196
	struct btrfs_device *extent_dev;
	int extent_mirror_num;
3197
	int stop_loop = 0;
D
David Woodhouse 已提交
3198

3199
	physical = map->stripes[num].physical;
A
Arne Jansen 已提交
3200
	offset = 0;
3201
	nstripes = div64_u64(length, map->stripe_len);
3202 3203
	mirror_num = 1;
	increment = map->stripe_len;
A
Arne Jansen 已提交
3204 3205 3206 3207 3208 3209 3210
	if (map->type & BTRFS_BLOCK_GROUP_RAID0) {
		offset = map->stripe_len * num;
		increment = map->stripe_len * map->num_stripes;
	} 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;
3211
		mirror_num = num % map->sub_stripes + 1;
3212
	} else if (map->type & BTRFS_BLOCK_GROUP_RAID1_MASK) {
3213
		mirror_num = num % map->num_stripes + 1;
A
Arne Jansen 已提交
3214
	} else if (map->type & BTRFS_BLOCK_GROUP_DUP) {
3215
		mirror_num = num % map->num_stripes + 1;
3216
	} else if (map->type & BTRFS_BLOCK_GROUP_RAID56_MASK) {
3217
		get_raid56_logic_offset(physical, num, map, &offset, NULL);
3218
		increment = map->stripe_len * nr_data_stripes(map);
A
Arne Jansen 已提交
3219 3220 3221 3222 3223 3224
	}

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

3225 3226
	ppath = btrfs_alloc_path();
	if (!ppath) {
3227
		btrfs_free_path(path);
3228 3229 3230
		return -ENOMEM;
	}

3231 3232 3233 3234 3235
	/*
	 * work on commit root. The related disk blocks are static as
	 * long as COW is applied. This means, it is save to rewrite
	 * them to repair disk errors without any race conditions
	 */
A
Arne Jansen 已提交
3236 3237 3238
	path->search_commit_root = 1;
	path->skip_locking = 1;

3239 3240
	ppath->search_commit_root = 1;
	ppath->skip_locking = 1;
A
Arne Jansen 已提交
3241
	/*
A
Arne Jansen 已提交
3242 3243 3244
	 * 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
3245 3246
	 */
	logical = base + offset;
3247
	physical_end = physical + nstripes * map->stripe_len;
3248
	if (map->type & BTRFS_BLOCK_GROUP_RAID56_MASK) {
3249
		get_raid56_logic_offset(physical_end, num,
3250
					map, &logic_end, NULL);
3251 3252 3253 3254
		logic_end += base;
	} else {
		logic_end = logical + increment * nstripes;
	}
3255
	wait_event(sctx->list_wait,
3256
		   atomic_read(&sctx->bios_in_flight) == 0);
3257
	scrub_blocked_if_needed(fs_info);
A
Arne Jansen 已提交
3258

3259 3260
	root = btrfs_extent_root(fs_info, logical);

A
Arne Jansen 已提交
3261
	/* FIXME it might be better to start readahead at commit root */
3262 3263 3264
	key.objectid = logical;
	key.type = BTRFS_EXTENT_ITEM_KEY;
	key.offset = (u64)0;
3265
	key_end.objectid = logic_end;
3266 3267
	key_end.type = BTRFS_METADATA_ITEM_KEY;
	key_end.offset = (u64)-1;
3268
	reada1 = btrfs_reada_add(root, &key, &key_end);
A
Arne Jansen 已提交
3269

3270 3271
	csum_root = btrfs_csum_root(fs_info, logical);

3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282
	if (cache->flags & BTRFS_BLOCK_GROUP_DATA) {
		key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
		key.type = BTRFS_EXTENT_CSUM_KEY;
		key.offset = logical;
		key_end.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
		key_end.type = BTRFS_EXTENT_CSUM_KEY;
		key_end.offset = logic_end;
		reada2 = btrfs_reada_add(csum_root, &key, &key_end);
	} else {
		reada2 = NULL;
	}
A
Arne Jansen 已提交
3283 3284 3285

	if (!IS_ERR(reada1))
		btrfs_reada_wait(reada1);
3286
	if (!IS_ERR_OR_NULL(reada2))
A
Arne Jansen 已提交
3287 3288
		btrfs_reada_wait(reada2);

A
Arne Jansen 已提交
3289 3290 3291 3292 3293

	/*
	 * collect all data csums for the stripe to avoid seeking during
	 * the scrub. This might currently (crc32) end up to be about 1MB
	 */
3294
	blk_start_plug(&plug);
A
Arne Jansen 已提交
3295

3296 3297 3298 3299 3300 3301 3302 3303
	if (sctx->is_dev_replace &&
	    btrfs_dev_is_sequential(sctx->wr_tgtdev, physical)) {
		mutex_lock(&sctx->wr_lock);
		sctx->write_pointer = physical;
		mutex_unlock(&sctx->wr_lock);
		sctx->flush_all_writes = true;
	}

A
Arne Jansen 已提交
3304 3305 3306 3307
	/*
	 * now find all extents for each stripe and scrub them
	 */
	ret = 0;
3308
	while (physical < physical_end) {
A
Arne Jansen 已提交
3309 3310 3311 3312
		/*
		 * canceled?
		 */
		if (atomic_read(&fs_info->scrub_cancel_req) ||
3313
		    atomic_read(&sctx->cancel_req)) {
A
Arne Jansen 已提交
3314 3315 3316 3317 3318 3319 3320 3321
			ret = -ECANCELED;
			goto out;
		}
		/*
		 * check to see if we have to pause
		 */
		if (atomic_read(&fs_info->scrub_pause_req)) {
			/* push queued extents */
3322
			sctx->flush_all_writes = true;
3323
			scrub_submit(sctx);
3324
			mutex_lock(&sctx->wr_lock);
3325
			scrub_wr_submit(sctx);
3326
			mutex_unlock(&sctx->wr_lock);
3327
			wait_event(sctx->list_wait,
3328
				   atomic_read(&sctx->bios_in_flight) == 0);
3329
			sctx->flush_all_writes = false;
3330
			scrub_blocked_if_needed(fs_info);
A
Arne Jansen 已提交
3331 3332
		}

3333 3334 3335 3336 3337 3338
		if (map->type & BTRFS_BLOCK_GROUP_RAID56_MASK) {
			ret = get_raid56_logic_offset(physical, num, map,
						      &logical,
						      &stripe_logical);
			logical += base;
			if (ret) {
3339
				/* it is parity strip */
3340
				stripe_logical += base;
3341
				stripe_end = stripe_logical + increment;
3342 3343 3344 3345 3346 3347 3348 3349 3350
				ret = scrub_raid56_parity(sctx, map, scrub_dev,
							  ppath, stripe_logical,
							  stripe_end);
				if (ret)
					goto out;
				goto skip;
			}
		}

3351 3352 3353 3354
		if (btrfs_fs_incompat(fs_info, SKINNY_METADATA))
			key.type = BTRFS_METADATA_ITEM_KEY;
		else
			key.type = BTRFS_EXTENT_ITEM_KEY;
A
Arne Jansen 已提交
3355
		key.objectid = logical;
L
Liu Bo 已提交
3356
		key.offset = (u64)-1;
A
Arne Jansen 已提交
3357 3358 3359 3360

		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
		if (ret < 0)
			goto out;
3361

3362
		if (ret > 0) {
3363
			ret = btrfs_previous_extent_item(root, path, 0);
A
Arne Jansen 已提交
3364 3365
			if (ret < 0)
				goto out;
3366 3367 3368 3369 3370 3371 3372 3373 3374
			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
3375 3376
		}

L
Liu Bo 已提交
3377
		stop_loop = 0;
A
Arne Jansen 已提交
3378
		while (1) {
3379 3380
			u64 bytes;

A
Arne Jansen 已提交
3381 3382 3383 3384 3385 3386 3387 3388 3389
			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;

L
Liu Bo 已提交
3390
				stop_loop = 1;
A
Arne Jansen 已提交
3391 3392 3393 3394
				break;
			}
			btrfs_item_key_to_cpu(l, &key, slot);

3395 3396 3397 3398
			if (key.type != BTRFS_EXTENT_ITEM_KEY &&
			    key.type != BTRFS_METADATA_ITEM_KEY)
				goto next;

3399
			if (key.type == BTRFS_METADATA_ITEM_KEY)
3400
				bytes = fs_info->nodesize;
3401 3402 3403 3404
			else
				bytes = key.offset;

			if (key.objectid + bytes <= logical)
A
Arne Jansen 已提交
3405 3406
				goto next;

L
Liu Bo 已提交
3407 3408 3409 3410 3411 3412
			if (key.objectid >= logical + map->stripe_len) {
				/* out of this device extent */
				if (key.objectid >= logic_end)
					stop_loop = 1;
				break;
			}
A
Arne Jansen 已提交
3413

3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427
			/*
			 * If our block group was removed in the meanwhile, just
			 * stop scrubbing since there is no point in continuing.
			 * Continuing would prevent reusing its device extents
			 * for new block groups for a long time.
			 */
			spin_lock(&cache->lock);
			if (cache->removed) {
				spin_unlock(&cache->lock);
				ret = 0;
				goto out;
			}
			spin_unlock(&cache->lock);

A
Arne Jansen 已提交
3428 3429 3430 3431 3432
			extent = btrfs_item_ptr(l, slot,
						struct btrfs_extent_item);
			flags = btrfs_extent_flags(l, extent);
			generation = btrfs_extent_generation(l, extent);

3433 3434 3435 3436
			if ((flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) &&
			    (key.objectid < logical ||
			     key.objectid + bytes >
			     logical + map->stripe_len)) {
3437
				btrfs_err(fs_info,
J
Jeff Mahoney 已提交
3438
					   "scrub: tree block %llu spanning stripes, ignored. logical=%llu",
3439
				       key.objectid, logical);
3440 3441 3442
				spin_lock(&sctx->stat_lock);
				sctx->stat.uncorrectable_errors++;
				spin_unlock(&sctx->stat_lock);
A
Arne Jansen 已提交
3443 3444 3445
				goto next;
			}

L
Liu Bo 已提交
3446 3447
again:
			extent_logical = key.objectid;
3448
			ASSERT(bytes <= U32_MAX);
L
Liu Bo 已提交
3449 3450
			extent_len = bytes;

A
Arne Jansen 已提交
3451 3452 3453
			/*
			 * trim extent to this stripe
			 */
L
Liu Bo 已提交
3454 3455 3456
			if (extent_logical < logical) {
				extent_len -= logical - extent_logical;
				extent_logical = logical;
A
Arne Jansen 已提交
3457
			}
L
Liu Bo 已提交
3458
			if (extent_logical + extent_len >
A
Arne Jansen 已提交
3459
			    logical + map->stripe_len) {
L
Liu Bo 已提交
3460 3461
				extent_len = logical + map->stripe_len -
					     extent_logical;
A
Arne Jansen 已提交
3462 3463
			}

L
Liu Bo 已提交
3464
			extent_physical = extent_logical - logical + physical;
3465 3466
			extent_dev = scrub_dev;
			extent_mirror_num = mirror_num;
3467
			if (sctx->is_dev_replace)
3468 3469 3470 3471
				scrub_remap_extent(fs_info, extent_logical,
						   extent_len, &extent_physical,
						   &extent_dev,
						   &extent_mirror_num);
L
Liu Bo 已提交
3472

3473 3474 3475 3476 3477 3478 3479 3480
			if (flags & BTRFS_EXTENT_FLAG_DATA) {
				ret = btrfs_lookup_csums_range(csum_root,
						extent_logical,
						extent_logical + extent_len - 1,
						&sctx->csum_list, 1);
				if (ret)
					goto out;
			}
L
Liu Bo 已提交
3481

L
Liu Bo 已提交
3482
			ret = scrub_extent(sctx, map, extent_logical, extent_len,
3483 3484
					   extent_physical, extent_dev, flags,
					   generation, extent_mirror_num,
3485
					   extent_logical - logical + physical);
3486 3487 3488

			scrub_free_csums(sctx);

A
Arne Jansen 已提交
3489 3490 3491
			if (ret)
				goto out;

3492 3493 3494
			if (sctx->is_dev_replace)
				sync_replace_for_zoned(sctx);

L
Liu Bo 已提交
3495 3496
			if (extent_logical + extent_len <
			    key.objectid + bytes) {
3497
				if (map->type & BTRFS_BLOCK_GROUP_RAID56_MASK) {
3498 3499 3500 3501
					/*
					 * loop until we find next data stripe
					 * or we have finished all stripes.
					 */
3502 3503 3504 3505 3506 3507 3508 3509 3510 3511
loop:
					physical += map->stripe_len;
					ret = get_raid56_logic_offset(physical,
							num, map, &logical,
							&stripe_logical);
					logical += base;

					if (ret && physical < physical_end) {
						stripe_logical += base;
						stripe_end = stripe_logical +
3512
								increment;
3513 3514 3515 3516 3517 3518 3519 3520
						ret = scrub_raid56_parity(sctx,
							map, scrub_dev, ppath,
							stripe_logical,
							stripe_end);
						if (ret)
							goto out;
						goto loop;
					}
3521 3522 3523 3524
				} else {
					physical += map->stripe_len;
					logical += increment;
				}
L
Liu Bo 已提交
3525 3526 3527 3528 3529
				if (logical < key.objectid + bytes) {
					cond_resched();
					goto again;
				}

3530
				if (physical >= physical_end) {
L
Liu Bo 已提交
3531 3532 3533 3534
					stop_loop = 1;
					break;
				}
			}
A
Arne Jansen 已提交
3535 3536 3537
next:
			path->slots[0]++;
		}
C
Chris Mason 已提交
3538
		btrfs_release_path(path);
3539
skip:
A
Arne Jansen 已提交
3540 3541
		logical += increment;
		physical += map->stripe_len;
3542
		spin_lock(&sctx->stat_lock);
L
Liu Bo 已提交
3543 3544 3545 3546 3547
		if (stop_loop)
			sctx->stat.last_physical = map->stripes[num].physical +
						   length;
		else
			sctx->stat.last_physical = physical;
3548
		spin_unlock(&sctx->stat_lock);
L
Liu Bo 已提交
3549 3550
		if (stop_loop)
			break;
A
Arne Jansen 已提交
3551
	}
3552
out:
A
Arne Jansen 已提交
3553
	/* push queued extents */
3554
	scrub_submit(sctx);
3555
	mutex_lock(&sctx->wr_lock);
3556
	scrub_wr_submit(sctx);
3557
	mutex_unlock(&sctx->wr_lock);
A
Arne Jansen 已提交
3558

3559
	blk_finish_plug(&plug);
A
Arne Jansen 已提交
3560
	btrfs_free_path(path);
3561
	btrfs_free_path(ppath);
3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572

	if (sctx->is_dev_replace && ret >= 0) {
		int ret2;

		ret2 = sync_write_pointer_for_zoned(sctx, base + offset,
						    map->stripes[num].physical,
						    physical_end);
		if (ret2)
			ret = ret2;
	}

A
Arne Jansen 已提交
3573 3574 3575
	return ret < 0 ? ret : 0;
}

3576
static noinline_for_stack int scrub_chunk(struct scrub_ctx *sctx,
3577 3578
					  struct btrfs_device *scrub_dev,
					  u64 chunk_offset, u64 length,
3579
					  u64 dev_offset,
3580
					  struct btrfs_block_group *cache)
A
Arne Jansen 已提交
3581
{
3582
	struct btrfs_fs_info *fs_info = sctx->fs_info;
3583
	struct extent_map_tree *map_tree = &fs_info->mapping_tree;
A
Arne Jansen 已提交
3584 3585 3586
	struct map_lookup *map;
	struct extent_map *em;
	int i;
3587
	int ret = 0;
A
Arne Jansen 已提交
3588

3589 3590 3591
	read_lock(&map_tree->lock);
	em = lookup_extent_mapping(map_tree, chunk_offset, 1);
	read_unlock(&map_tree->lock);
A
Arne Jansen 已提交
3592

3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604
	if (!em) {
		/*
		 * Might have been an unused block group deleted by the cleaner
		 * kthread or relocation.
		 */
		spin_lock(&cache->lock);
		if (!cache->removed)
			ret = -EINVAL;
		spin_unlock(&cache->lock);

		return ret;
	}
A
Arne Jansen 已提交
3605

3606
	map = em->map_lookup;
A
Arne Jansen 已提交
3607 3608 3609 3610 3611 3612 3613
	if (em->start != chunk_offset)
		goto out;

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

	for (i = 0; i < map->num_stripes; ++i) {
3614
		if (map->stripes[i].dev->bdev == scrub_dev->bdev &&
3615
		    map->stripes[i].physical == dev_offset) {
3616
			ret = scrub_stripe(sctx, map, scrub_dev, i,
3617
					   chunk_offset, length, cache);
A
Arne Jansen 已提交
3618 3619 3620 3621 3622 3623 3624 3625 3626 3627
			if (ret)
				goto out;
		}
	}
out:
	free_extent_map(em);

	return ret;
}

3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646
static int finish_extent_writes_for_zoned(struct btrfs_root *root,
					  struct btrfs_block_group *cache)
{
	struct btrfs_fs_info *fs_info = cache->fs_info;
	struct btrfs_trans_handle *trans;

	if (!btrfs_is_zoned(fs_info))
		return 0;

	btrfs_wait_block_group_reservations(cache);
	btrfs_wait_nocow_writers(cache);
	btrfs_wait_ordered_roots(fs_info, U64_MAX, cache->start, cache->length);

	trans = btrfs_join_transaction(root);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
	return btrfs_commit_transaction(trans);
}

A
Arne Jansen 已提交
3647
static noinline_for_stack
3648
int scrub_enumerate_chunks(struct scrub_ctx *sctx,
3649
			   struct btrfs_device *scrub_dev, u64 start, u64 end)
A
Arne Jansen 已提交
3650 3651 3652
{
	struct btrfs_dev_extent *dev_extent = NULL;
	struct btrfs_path *path;
3653 3654
	struct btrfs_fs_info *fs_info = sctx->fs_info;
	struct btrfs_root *root = fs_info->dev_root;
A
Arne Jansen 已提交
3655 3656
	u64 length;
	u64 chunk_offset;
3657
	int ret = 0;
3658
	int ro_set;
A
Arne Jansen 已提交
3659 3660 3661 3662
	int slot;
	struct extent_buffer *l;
	struct btrfs_key key;
	struct btrfs_key found_key;
3663
	struct btrfs_block_group *cache;
3664
	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
A
Arne Jansen 已提交
3665 3666 3667 3668 3669

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

3670
	path->reada = READA_FORWARD;
A
Arne Jansen 已提交
3671 3672 3673
	path->search_commit_root = 1;
	path->skip_locking = 1;

3674
	key.objectid = scrub_dev->devid;
A
Arne Jansen 已提交
3675 3676 3677 3678 3679 3680
	key.offset = 0ull;
	key.type = BTRFS_DEV_EXTENT_KEY;

	while (1) {
		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
		if (ret < 0)
3681 3682 3683 3684 3685
			break;
		if (ret > 0) {
			if (path->slots[0] >=
			    btrfs_header_nritems(path->nodes[0])) {
				ret = btrfs_next_leaf(root, path);
3686 3687 3688 3689
				if (ret < 0)
					break;
				if (ret > 0) {
					ret = 0;
3690
					break;
3691 3692 3693
				}
			} else {
				ret = 0;
3694 3695
			}
		}
A
Arne Jansen 已提交
3696 3697 3698 3699 3700 3701

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

		btrfs_item_key_to_cpu(l, &found_key, slot);

3702
		if (found_key.objectid != scrub_dev->devid)
A
Arne Jansen 已提交
3703 3704
			break;

3705
		if (found_key.type != BTRFS_DEV_EXTENT_KEY)
A
Arne Jansen 已提交
3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716
			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);

3717 3718
		if (found_key.offset + length <= start)
			goto skip;
A
Arne Jansen 已提交
3719 3720 3721 3722 3723 3724 3725 3726

		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);
3727 3728 3729 3730 3731 3732

		/* some chunks are removed but not committed to disk yet,
		 * continue scrubbing */
		if (!cache)
			goto skip;

3733 3734 3735 3736
		if (sctx->is_dev_replace && btrfs_is_zoned(fs_info)) {
			spin_lock(&cache->lock);
			if (!cache->to_copy) {
				spin_unlock(&cache->lock);
3737 3738
				btrfs_put_block_group(cache);
				goto skip;
3739 3740 3741 3742
			}
			spin_unlock(&cache->lock);
		}

3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756
		/*
		 * Make sure that while we are scrubbing the corresponding block
		 * group doesn't get its logical address and its device extents
		 * reused for another block group, which can possibly be of a
		 * different type and different profile. We do this to prevent
		 * false error detections and crashes due to bogus attempts to
		 * repair extents.
		 */
		spin_lock(&cache->lock);
		if (cache->removed) {
			spin_unlock(&cache->lock);
			btrfs_put_block_group(cache);
			goto skip;
		}
3757
		btrfs_freeze_block_group(cache);
3758 3759
		spin_unlock(&cache->lock);

3760 3761 3762 3763 3764 3765 3766 3767 3768
		/*
		 * we need call btrfs_inc_block_group_ro() with scrubs_paused,
		 * to avoid deadlock caused by:
		 * btrfs_inc_block_group_ro()
		 * -> btrfs_wait_for_commit()
		 * -> btrfs_commit_transaction()
		 * -> btrfs_scrub_pause()
		 */
		scrub_pause_on(fs_info);
3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786

		/*
		 * Don't do chunk preallocation for scrub.
		 *
		 * This is especially important for SYSTEM bgs, or we can hit
		 * -EFBIG from btrfs_finish_chunk_alloc() like:
		 * 1. The only SYSTEM bg is marked RO.
		 *    Since SYSTEM bg is small, that's pretty common.
		 * 2. New SYSTEM bg will be allocated
		 *    Due to regular version will allocate new chunk.
		 * 3. New SYSTEM bg is empty and will get cleaned up
		 *    Before cleanup really happens, it's marked RO again.
		 * 4. Empty SYSTEM bg get scrubbed
		 *    We go back to 2.
		 *
		 * This can easily boost the amount of SYSTEM chunks if cleaner
		 * thread can't be triggered fast enough, and use up all space
		 * of btrfs_super_block::sys_chunk_array
3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798
		 *
		 * While for dev replace, we need to try our best to mark block
		 * group RO, to prevent race between:
		 * - Write duplication
		 *   Contains latest data
		 * - Scrub copy
		 *   Contains data from commit tree
		 *
		 * If target block group is not marked RO, nocow writes can
		 * be overwritten by scrub copy, causing data corruption.
		 * So for dev-replace, it's not allowed to continue if a block
		 * group is not RO.
3799
		 */
3800
		ret = btrfs_inc_block_group_ro(cache, sctx->is_dev_replace);
3801 3802 3803 3804 3805 3806 3807 3808 3809 3810
		if (!ret && sctx->is_dev_replace) {
			ret = finish_extent_writes_for_zoned(root, cache);
			if (ret) {
				btrfs_dec_block_group_ro(cache);
				scrub_pause_off(fs_info);
				btrfs_put_block_group(cache);
				break;
			}
		}

3811 3812
		if (ret == 0) {
			ro_set = 1;
3813
		} else if (ret == -ENOSPC && !sctx->is_dev_replace) {
3814 3815 3816
			/*
			 * btrfs_inc_block_group_ro return -ENOSPC when it
			 * failed in creating new chunk for metadata.
3817
			 * It is not a problem for scrub, because
3818 3819 3820 3821
			 * metadata are always cowed, and our scrub paused
			 * commit_transactions.
			 */
			ro_set = 0;
3822 3823 3824 3825 3826 3827 3828
		} else if (ret == -ETXTBSY) {
			btrfs_warn(fs_info,
		   "skipping scrub of block group %llu due to active swapfile",
				   cache->start);
			scrub_pause_off(fs_info);
			ret = 0;
			goto skip_unfreeze;
3829
		} else {
J
Jeff Mahoney 已提交
3830
			btrfs_warn(fs_info,
3831
				   "failed setting block group ro: %d", ret);
3832
			btrfs_unfreeze_block_group(cache);
3833
			btrfs_put_block_group(cache);
3834
			scrub_pause_off(fs_info);
3835 3836 3837
			break;
		}

3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849
		/*
		 * Now the target block is marked RO, wait for nocow writes to
		 * finish before dev-replace.
		 * COW is fine, as COW never overwrites extents in commit tree.
		 */
		if (sctx->is_dev_replace) {
			btrfs_wait_nocow_writers(cache);
			btrfs_wait_ordered_roots(fs_info, U64_MAX, cache->start,
					cache->length);
		}

		scrub_pause_off(fs_info);
3850
		down_write(&dev_replace->rwsem);
3851 3852 3853
		dev_replace->cursor_right = found_key.offset + length;
		dev_replace->cursor_left = found_key.offset;
		dev_replace->item_needs_writeback = 1;
3854 3855
		up_write(&dev_replace->rwsem);

3856
		ret = scrub_chunk(sctx, scrub_dev, chunk_offset, length,
3857
				  found_key.offset, cache);
3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868

		/*
		 * flush, submit all pending read and write bios, afterwards
		 * wait for them.
		 * Note that in the dev replace case, a read request causes
		 * write requests that are submitted in the read completion
		 * worker. Therefore in the current situation, it is required
		 * that all write requests are flushed, so that all read and
		 * write requests are really completed when bios_in_flight
		 * changes to 0.
		 */
3869
		sctx->flush_all_writes = true;
3870
		scrub_submit(sctx);
3871
		mutex_lock(&sctx->wr_lock);
3872
		scrub_wr_submit(sctx);
3873
		mutex_unlock(&sctx->wr_lock);
3874 3875 3876

		wait_event(sctx->list_wait,
			   atomic_read(&sctx->bios_in_flight) == 0);
3877 3878

		scrub_pause_on(fs_info);
3879 3880 3881 3882 3883 3884

		/*
		 * must be called before we decrease @scrub_paused.
		 * make sure we don't block transaction commit while
		 * we are waiting pending workers finished.
		 */
3885 3886
		wait_event(sctx->list_wait,
			   atomic_read(&sctx->workers_pending) == 0);
3887
		sctx->flush_all_writes = false;
3888

3889
		scrub_pause_off(fs_info);
3890

3891 3892 3893 3894 3895
		if (sctx->is_dev_replace &&
		    !btrfs_finish_block_group_to_copy(dev_replace->srcdev,
						      cache, found_key.offset))
			ro_set = 0;

3896
		down_write(&dev_replace->rwsem);
3897 3898
		dev_replace->cursor_left = dev_replace->cursor_right;
		dev_replace->item_needs_writeback = 1;
3899
		up_write(&dev_replace->rwsem);
3900

3901
		if (ro_set)
3902
			btrfs_dec_block_group_ro(cache);
3903

3904 3905 3906 3907 3908 3909 3910 3911 3912
		/*
		 * We might have prevented the cleaner kthread from deleting
		 * this block group if it was already unused because we raced
		 * and set it to RO mode first. So add it back to the unused
		 * list, otherwise it might not ever be deleted unless a manual
		 * balance is triggered or it becomes used and unused again.
		 */
		spin_lock(&cache->lock);
		if (!cache->removed && !cache->ro && cache->reserved == 0 &&
3913
		    cache->used == 0) {
3914
			spin_unlock(&cache->lock);
3915 3916 3917 3918 3919
			if (btrfs_test_opt(fs_info, DISCARD_ASYNC))
				btrfs_discard_queue_work(&fs_info->discard_ctl,
							 cache);
			else
				btrfs_mark_bg_unused(cache);
3920 3921 3922
		} else {
			spin_unlock(&cache->lock);
		}
3923
skip_unfreeze:
3924
		btrfs_unfreeze_block_group(cache);
A
Arne Jansen 已提交
3925 3926 3927
		btrfs_put_block_group(cache);
		if (ret)
			break;
3928
		if (sctx->is_dev_replace &&
3929
		    atomic64_read(&dev_replace->num_write_errors) > 0) {
3930 3931 3932 3933 3934 3935 3936
			ret = -EIO;
			break;
		}
		if (sctx->stat.malloc_errors > 0) {
			ret = -ENOMEM;
			break;
		}
3937
skip:
A
Arne Jansen 已提交
3938
		key.offset = found_key.offset + length;
C
Chris Mason 已提交
3939
		btrfs_release_path(path);
A
Arne Jansen 已提交
3940 3941 3942
	}

	btrfs_free_path(path);
3943

3944
	return ret;
A
Arne Jansen 已提交
3945 3946
}

3947 3948
static noinline_for_stack int scrub_supers(struct scrub_ctx *sctx,
					   struct btrfs_device *scrub_dev)
A
Arne Jansen 已提交
3949 3950 3951 3952 3953
{
	int	i;
	u64	bytenr;
	u64	gen;
	int	ret;
3954
	struct btrfs_fs_info *fs_info = sctx->fs_info;
A
Arne Jansen 已提交
3955

J
Josef Bacik 已提交
3956
	if (BTRFS_FS_ERROR(fs_info))
3957
		return -EROFS;
3958

3959
	/* Seed devices of a new filesystem has their own generation. */
3960
	if (scrub_dev->fs_devices != fs_info->fs_devices)
3961 3962
		gen = scrub_dev->generation;
	else
3963
		gen = fs_info->last_trans_committed;
A
Arne Jansen 已提交
3964 3965 3966

	for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
		bytenr = btrfs_sb_offset(i);
3967 3968
		if (bytenr + BTRFS_SUPER_INFO_SIZE >
		    scrub_dev->commit_total_bytes)
A
Arne Jansen 已提交
3969
			break;
3970 3971
		if (!btrfs_check_super_location(scrub_dev, bytenr))
			continue;
A
Arne Jansen 已提交
3972

3973
		ret = scrub_pages(sctx, bytenr, BTRFS_SUPER_INFO_SIZE, bytenr,
3974
				  scrub_dev, BTRFS_EXTENT_FLAG_SUPER, gen, i,
3975
				  NULL, bytenr);
A
Arne Jansen 已提交
3976 3977 3978
		if (ret)
			return ret;
	}
3979
	wait_event(sctx->list_wait, atomic_read(&sctx->bios_in_flight) == 0);
A
Arne Jansen 已提交
3980 3981 3982 3983

	return 0;
}

3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006
static void scrub_workers_put(struct btrfs_fs_info *fs_info)
{
	if (refcount_dec_and_mutex_lock(&fs_info->scrub_workers_refcnt,
					&fs_info->scrub_lock)) {
		struct btrfs_workqueue *scrub_workers = NULL;
		struct btrfs_workqueue *scrub_wr_comp = NULL;
		struct btrfs_workqueue *scrub_parity = NULL;

		scrub_workers = fs_info->scrub_workers;
		scrub_wr_comp = fs_info->scrub_wr_completion_workers;
		scrub_parity = fs_info->scrub_parity_workers;

		fs_info->scrub_workers = NULL;
		fs_info->scrub_wr_completion_workers = NULL;
		fs_info->scrub_parity_workers = NULL;
		mutex_unlock(&fs_info->scrub_lock);

		btrfs_destroy_workqueue(scrub_workers);
		btrfs_destroy_workqueue(scrub_wr_comp);
		btrfs_destroy_workqueue(scrub_parity);
	}
}

A
Arne Jansen 已提交
4007 4008 4009
/*
 * get a reference count on fs_info->scrub_workers. start worker if necessary
 */
4010 4011
static noinline_for_stack int scrub_workers_get(struct btrfs_fs_info *fs_info,
						int is_dev_replace)
A
Arne Jansen 已提交
4012
{
4013 4014 4015
	struct btrfs_workqueue *scrub_workers = NULL;
	struct btrfs_workqueue *scrub_wr_comp = NULL;
	struct btrfs_workqueue *scrub_parity = NULL;
4016
	unsigned int flags = WQ_FREEZABLE | WQ_UNBOUND;
4017
	int max_active = fs_info->thread_pool_size;
4018
	int ret = -ENOMEM;
A
Arne Jansen 已提交
4019

4020 4021
	if (refcount_inc_not_zero(&fs_info->scrub_workers_refcnt))
		return 0;
4022

4023 4024 4025 4026
	scrub_workers = btrfs_alloc_workqueue(fs_info, "scrub", flags,
					      is_dev_replace ? 1 : max_active, 4);
	if (!scrub_workers)
		goto fail_scrub_workers;
4027

4028
	scrub_wr_comp = btrfs_alloc_workqueue(fs_info, "scrubwrc", flags,
4029
					      max_active, 2);
4030 4031
	if (!scrub_wr_comp)
		goto fail_scrub_wr_completion_workers;
4032

4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045
	scrub_parity = btrfs_alloc_workqueue(fs_info, "scrubparity", flags,
					     max_active, 2);
	if (!scrub_parity)
		goto fail_scrub_parity_workers;

	mutex_lock(&fs_info->scrub_lock);
	if (refcount_read(&fs_info->scrub_workers_refcnt) == 0) {
		ASSERT(fs_info->scrub_workers == NULL &&
		       fs_info->scrub_wr_completion_workers == NULL &&
		       fs_info->scrub_parity_workers == NULL);
		fs_info->scrub_workers = scrub_workers;
		fs_info->scrub_wr_completion_workers = scrub_wr_comp;
		fs_info->scrub_parity_workers = scrub_parity;
4046
		refcount_set(&fs_info->scrub_workers_refcnt, 1);
4047 4048
		mutex_unlock(&fs_info->scrub_lock);
		return 0;
A
Arne Jansen 已提交
4049
	}
4050 4051 4052
	/* Other thread raced in and created the workers for us */
	refcount_inc(&fs_info->scrub_workers_refcnt);
	mutex_unlock(&fs_info->scrub_lock);
4053

4054 4055
	ret = 0;
	btrfs_destroy_workqueue(scrub_parity);
4056
fail_scrub_parity_workers:
4057
	btrfs_destroy_workqueue(scrub_wr_comp);
4058
fail_scrub_wr_completion_workers:
4059
	btrfs_destroy_workqueue(scrub_workers);
4060
fail_scrub_workers:
4061
	return ret;
A
Arne Jansen 已提交
4062 4063
}

4064 4065
int btrfs_scrub_dev(struct btrfs_fs_info *fs_info, u64 devid, u64 start,
		    u64 end, struct btrfs_scrub_progress *progress,
4066
		    int readonly, int is_dev_replace)
A
Arne Jansen 已提交
4067
{
4068
	struct btrfs_dev_lookup_args args = { .devid = devid };
4069
	struct scrub_ctx *sctx;
A
Arne Jansen 已提交
4070 4071
	int ret;
	struct btrfs_device *dev;
4072
	unsigned int nofs_flag;
A
Arne Jansen 已提交
4073

4074
	if (btrfs_fs_closing(fs_info))
4075
		return -EAGAIN;
A
Arne Jansen 已提交
4076

4077
	if (fs_info->nodesize > BTRFS_STRIPE_LEN) {
4078 4079 4080 4081 4082
		/*
		 * in this case scrub is unable to calculate the checksum
		 * the way scrub is implemented. Do not handle this
		 * situation at all because it won't ever happen.
		 */
4083 4084
		btrfs_err(fs_info,
			   "scrub: size assumption nodesize <= BTRFS_STRIPE_LEN (%d <= %d) fails",
4085 4086
		       fs_info->nodesize,
		       BTRFS_STRIPE_LEN);
4087 4088 4089
		return -EINVAL;
	}

4090
	if (fs_info->nodesize >
4091
	    PAGE_SIZE * SCRUB_MAX_PAGES_PER_BLOCK ||
4092
	    fs_info->sectorsize > PAGE_SIZE * SCRUB_MAX_PAGES_PER_BLOCK) {
4093 4094 4095 4096
		/*
		 * would exhaust the array bounds of pagev member in
		 * struct scrub_block
		 */
J
Jeff Mahoney 已提交
4097 4098
		btrfs_err(fs_info,
			  "scrub: size assumption nodesize and sectorsize <= SCRUB_MAX_PAGES_PER_BLOCK (%d <= %d && %d <= %d) fails",
4099
		       fs_info->nodesize,
4100
		       SCRUB_MAX_PAGES_PER_BLOCK,
4101
		       fs_info->sectorsize,
4102 4103 4104 4105
		       SCRUB_MAX_PAGES_PER_BLOCK);
		return -EINVAL;
	}

4106 4107 4108 4109
	/* Allocate outside of device_list_mutex */
	sctx = scrub_setup_ctx(fs_info, is_dev_replace);
	if (IS_ERR(sctx))
		return PTR_ERR(sctx);
A
Arne Jansen 已提交
4110

4111 4112 4113 4114
	ret = scrub_workers_get(fs_info, is_dev_replace);
	if (ret)
		goto out_free_ctx;

4115
	mutex_lock(&fs_info->fs_devices->device_list_mutex);
4116
	dev = btrfs_find_device(fs_info->fs_devices, &args);
4117 4118
	if (!dev || (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state) &&
		     !is_dev_replace)) {
4119
		mutex_unlock(&fs_info->fs_devices->device_list_mutex);
4120
		ret = -ENODEV;
4121
		goto out;
A
Arne Jansen 已提交
4122 4123
	}

4124 4125
	if (!is_dev_replace && !readonly &&
	    !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) {
4126
		mutex_unlock(&fs_info->fs_devices->device_list_mutex);
4127 4128 4129
		btrfs_err_in_rcu(fs_info,
			"scrub on devid %llu: filesystem on %s is not writable",
				 devid, rcu_str_deref(dev->name));
4130
		ret = -EROFS;
4131
		goto out;
4132 4133
	}

4134
	mutex_lock(&fs_info->scrub_lock);
4135
	if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) ||
4136
	    test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &dev->dev_state)) {
A
Arne Jansen 已提交
4137
		mutex_unlock(&fs_info->scrub_lock);
4138
		mutex_unlock(&fs_info->fs_devices->device_list_mutex);
4139
		ret = -EIO;
4140
		goto out;
A
Arne Jansen 已提交
4141 4142
	}

4143
	down_read(&fs_info->dev_replace.rwsem);
4144
	if (dev->scrub_ctx ||
4145 4146
	    (!is_dev_replace &&
	     btrfs_dev_replace_is_ongoing(&fs_info->dev_replace))) {
4147
		up_read(&fs_info->dev_replace.rwsem);
A
Arne Jansen 已提交
4148
		mutex_unlock(&fs_info->scrub_lock);
4149
		mutex_unlock(&fs_info->fs_devices->device_list_mutex);
4150
		ret = -EINPROGRESS;
4151
		goto out;
A
Arne Jansen 已提交
4152
	}
4153
	up_read(&fs_info->dev_replace.rwsem);
4154

4155
	sctx->readonly = readonly;
4156
	dev->scrub_ctx = sctx;
4157
	mutex_unlock(&fs_info->fs_devices->device_list_mutex);
A
Arne Jansen 已提交
4158

4159 4160 4161 4162
	/*
	 * checking @scrub_pause_req here, we can avoid
	 * race between committing transaction and scrubbing.
	 */
4163
	__scrub_blocked_if_needed(fs_info);
A
Arne Jansen 已提交
4164 4165 4166
	atomic_inc(&fs_info->scrubs_running);
	mutex_unlock(&fs_info->scrub_lock);

4167 4168 4169 4170 4171 4172 4173 4174 4175 4176
	/*
	 * In order to avoid deadlock with reclaim when there is a transaction
	 * trying to pause scrub, make sure we use GFP_NOFS for all the
	 * allocations done at btrfs_scrub_pages() and scrub_pages_for_parity()
	 * invoked by our callees. The pausing request is done when the
	 * transaction commit starts, and it blocks the transaction until scrub
	 * is paused (done at specific points at scrub_stripe() or right above
	 * before incrementing fs_info->scrubs_running).
	 */
	nofs_flag = memalloc_nofs_save();
4177
	if (!is_dev_replace) {
4178
		btrfs_info(fs_info, "scrub: started on devid %llu", devid);
4179 4180 4181 4182
		/*
		 * by holding device list mutex, we can
		 * kick off writing super in log tree sync.
		 */
4183
		mutex_lock(&fs_info->fs_devices->device_list_mutex);
4184
		ret = scrub_supers(sctx, dev);
4185
		mutex_unlock(&fs_info->fs_devices->device_list_mutex);
4186
	}
A
Arne Jansen 已提交
4187 4188

	if (!ret)
4189
		ret = scrub_enumerate_chunks(sctx, dev, start, end);
4190
	memalloc_nofs_restore(nofs_flag);
A
Arne Jansen 已提交
4191

4192
	wait_event(sctx->list_wait, atomic_read(&sctx->bios_in_flight) == 0);
A
Arne Jansen 已提交
4193 4194 4195
	atomic_dec(&fs_info->scrubs_running);
	wake_up(&fs_info->scrub_pause_wait);

4196
	wait_event(sctx->list_wait, atomic_read(&sctx->workers_pending) == 0);
4197

A
Arne Jansen 已提交
4198
	if (progress)
4199
		memcpy(progress, &sctx->stat, sizeof(*progress));
A
Arne Jansen 已提交
4200

4201 4202 4203 4204
	if (!is_dev_replace)
		btrfs_info(fs_info, "scrub: %s on devid %llu with status: %d",
			ret ? "not finished" : "finished", devid, ret);

A
Arne Jansen 已提交
4205
	mutex_lock(&fs_info->scrub_lock);
4206
	dev->scrub_ctx = NULL;
A
Arne Jansen 已提交
4207 4208
	mutex_unlock(&fs_info->scrub_lock);

4209
	scrub_workers_put(fs_info);
4210
	scrub_put_ctx(sctx);
A
Arne Jansen 已提交
4211

4212
	return ret;
4213 4214
out:
	scrub_workers_put(fs_info);
4215 4216 4217
out_free_ctx:
	scrub_free_ctx(sctx);

A
Arne Jansen 已提交
4218 4219 4220
	return ret;
}

4221
void btrfs_scrub_pause(struct btrfs_fs_info *fs_info)
A
Arne Jansen 已提交
4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235
{
	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);
}

4236
void btrfs_scrub_continue(struct btrfs_fs_info *fs_info)
A
Arne Jansen 已提交
4237 4238 4239 4240 4241
{
	atomic_dec(&fs_info->scrub_pause_req);
	wake_up(&fs_info->scrub_pause_wait);
}

4242
int btrfs_scrub_cancel(struct btrfs_fs_info *fs_info)
A
Arne Jansen 已提交
4243 4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261 4262
{
	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;
}

4263
int btrfs_scrub_cancel_dev(struct btrfs_device *dev)
4264
{
4265
	struct btrfs_fs_info *fs_info = dev->fs_info;
4266
	struct scrub_ctx *sctx;
A
Arne Jansen 已提交
4267 4268

	mutex_lock(&fs_info->scrub_lock);
4269
	sctx = dev->scrub_ctx;
4270
	if (!sctx) {
A
Arne Jansen 已提交
4271 4272 4273
		mutex_unlock(&fs_info->scrub_lock);
		return -ENOTCONN;
	}
4274
	atomic_inc(&sctx->cancel_req);
4275
	while (dev->scrub_ctx) {
A
Arne Jansen 已提交
4276 4277
		mutex_unlock(&fs_info->scrub_lock);
		wait_event(fs_info->scrub_pause_wait,
4278
			   dev->scrub_ctx == NULL);
A
Arne Jansen 已提交
4279 4280 4281 4282 4283 4284
		mutex_lock(&fs_info->scrub_lock);
	}
	mutex_unlock(&fs_info->scrub_lock);

	return 0;
}
S
Stefan Behrens 已提交
4285

4286
int btrfs_scrub_progress(struct btrfs_fs_info *fs_info, u64 devid,
A
Arne Jansen 已提交
4287 4288
			 struct btrfs_scrub_progress *progress)
{
4289
	struct btrfs_dev_lookup_args args = { .devid = devid };
A
Arne Jansen 已提交
4290
	struct btrfs_device *dev;
4291
	struct scrub_ctx *sctx = NULL;
A
Arne Jansen 已提交
4292

4293
	mutex_lock(&fs_info->fs_devices->device_list_mutex);
4294
	dev = btrfs_find_device(fs_info->fs_devices, &args);
A
Arne Jansen 已提交
4295
	if (dev)
4296
		sctx = dev->scrub_ctx;
4297 4298
	if (sctx)
		memcpy(progress, &sctx->stat, sizeof(*progress));
4299
	mutex_unlock(&fs_info->fs_devices->device_list_mutex);
A
Arne Jansen 已提交
4300

4301
	return dev ? (sctx ? 0 : -ENOTCONN) : -ENODEV;
A
Arne Jansen 已提交
4302
}
4303 4304

static void scrub_remap_extent(struct btrfs_fs_info *fs_info,
4305
			       u64 extent_logical, u32 extent_len,
4306 4307 4308 4309 4310
			       u64 *extent_physical,
			       struct btrfs_device **extent_dev,
			       int *extent_mirror_num)
{
	u64 mapped_length;
4311
	struct btrfs_io_context *bioc = NULL;
4312 4313 4314
	int ret;

	mapped_length = extent_len;
4315
	ret = btrfs_map_block(fs_info, BTRFS_MAP_READ, extent_logical,
4316 4317 4318 4319
			      &mapped_length, &bioc, 0);
	if (ret || !bioc || mapped_length < extent_len ||
	    !bioc->stripes[0].dev->bdev) {
		btrfs_put_bioc(bioc);
4320 4321 4322
		return;
	}

4323 4324 4325 4326
	*extent_physical = bioc->stripes[0].physical;
	*extent_mirror_num = bioc->mirror_num;
	*extent_dev = bioc->stripes[0].dev;
	btrfs_put_bioc(bioc);
4327
}