scrub.c 114.2 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 44 45 46 47 48 49
/*
 * the following three values only influence the performance.
 * The last one configures the number of parallel and outstanding I/O
 * operations. The first two values configure an upper limit for the number
 * of (dynamically allocated) pages that are added to a bio.
 */
#define SCRUB_PAGES_PER_RD_BIO	32	/* 128k per bio */
#define SCRUB_PAGES_PER_WR_BIO	32	/* 128k per bio */
#define SCRUB_BIOS_PER_SCTX	64	/* 8MB per device in flight */
50 51 52 53 54 55

/*
 * the following value times PAGE_SIZE needs to be large enough to match the
 * largest node/leaf/sector size that shall be supported.
 * Values larger than BTRFS_STRIPE_LEN are not supported.
 */
56
#define SCRUB_MAX_PAGES_PER_BLOCK	16	/* 64k per node/leaf/sector */
A
Arne Jansen 已提交
57

58
struct scrub_recover {
59
	refcount_t		refs;
60
	struct btrfs_io_context	*bioc;
61 62 63
	u64			map_length;
};

A
Arne Jansen 已提交
64
struct scrub_page {
65 66
	struct scrub_block	*sblock;
	struct page		*page;
67
	struct btrfs_device	*dev;
68
	struct list_head	list;
A
Arne Jansen 已提交
69 70
	u64			flags;  /* extent flags */
	u64			generation;
71 72
	u64			logical;
	u64			physical;
73
	u64			physical_for_dev_replace;
74
	atomic_t		refs;
75
	u8			mirror_num;
76 77
	unsigned int		have_csum:1;
	unsigned int		io_error:1;
A
Arne Jansen 已提交
78
	u8			csum[BTRFS_CSUM_SIZE];
79 80

	struct scrub_recover	*recover;
A
Arne Jansen 已提交
81 82 83 84
};

struct scrub_bio {
	int			index;
85
	struct scrub_ctx	*sctx;
86
	struct btrfs_device	*dev;
A
Arne Jansen 已提交
87
	struct bio		*bio;
88
	blk_status_t		status;
A
Arne Jansen 已提交
89 90
	u64			logical;
	u64			physical;
91 92 93 94 95
#if SCRUB_PAGES_PER_WR_BIO >= SCRUB_PAGES_PER_RD_BIO
	struct scrub_page	*pagev[SCRUB_PAGES_PER_WR_BIO];
#else
	struct scrub_page	*pagev[SCRUB_PAGES_PER_RD_BIO];
#endif
96
	int			page_count;
A
Arne Jansen 已提交
97 98 99 100
	int			next_free;
	struct btrfs_work	work;
};

101
struct scrub_block {
102
	struct scrub_page	*pagev[SCRUB_MAX_PAGES_PER_BLOCK];
103 104
	int			page_count;
	atomic_t		outstanding_pages;
105
	refcount_t		refs; /* free mem on transition to zero */
106
	struct scrub_ctx	*sctx;
107
	struct scrub_parity	*sparity;
108 109 110 111
	struct {
		unsigned int	header_error:1;
		unsigned int	checksum_error:1;
		unsigned int	no_io_error_seen:1;
112
		unsigned int	generation_error:1; /* also sets header_error */
113 114 115 116

		/* The following is for the data used to check parity */
		/* It is for the data with checksum */
		unsigned int	data_corrected:1;
117
	};
118
	struct btrfs_work	work;
119 120
};

121 122 123 124 125 126 127 128 129 130 131 132
/* 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;

133
	u32			stripe_len;
134

135
	refcount_t		refs;
136 137 138 139 140 141 142 143 144 145 146 147 148 149 150

	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;

151
	unsigned long		bitmap[];
152 153
};

154
struct scrub_ctx {
155
	struct scrub_bio	*bios[SCRUB_BIOS_PER_SCTX];
156
	struct btrfs_fs_info	*fs_info;
A
Arne Jansen 已提交
157 158
	int			first_free;
	int			curr;
159 160
	atomic_t		bios_in_flight;
	atomic_t		workers_pending;
A
Arne Jansen 已提交
161 162 163 164
	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
165
	int			readonly;
166
	int			pages_per_rd_bio;
167

168 169 170 171
	/* State of IO submission throttling affecting the associated device */
	ktime_t			throttle_deadline;
	u64			throttle_sent;

172
	int			is_dev_replace;
173
	u64			write_pointer;
174 175 176 177 178

	struct scrub_bio        *wr_curr_bio;
	struct mutex            wr_lock;
	int                     pages_per_wr_bio; /* <= SCRUB_PAGES_PER_WR_BIO */
	struct btrfs_device     *wr_tgtdev;
179
	bool                    flush_all_writes;
180

A
Arne Jansen 已提交
181 182 183 184 185
	/*
	 * statistics
	 */
	struct btrfs_scrub_progress stat;
	spinlock_t		stat_lock;
186 187 188 189 190 191 192 193

	/*
	 * 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.
	 */
194
	refcount_t              refs;
A
Arne Jansen 已提交
195 196
};

197 198 199 200
struct scrub_warning {
	struct btrfs_path	*path;
	u64			extent_item_size;
	const char		*errstr;
D
David Sterba 已提交
201
	u64			physical;
202 203 204 205
	u64			logical;
	struct btrfs_device	*dev;
};

206 207 208 209 210 211 212
struct full_stripe_lock {
	struct rb_node node;
	u64 logical;
	u64 refs;
	struct mutex mutex;
};

213
static int scrub_setup_recheck_block(struct scrub_block *original_sblock,
214
				     struct scrub_block *sblocks_for_recheck);
215
static void scrub_recheck_block(struct btrfs_fs_info *fs_info,
216 217
				struct scrub_block *sblock,
				int retry_failed_mirror);
218
static void scrub_recheck_block_checksum(struct scrub_block *sblock);
219
static int scrub_repair_block_from_good_copy(struct scrub_block *sblock_bad,
220
					     struct scrub_block *sblock_good);
221 222 223
static int scrub_repair_page_from_good_copy(struct scrub_block *sblock_bad,
					    struct scrub_block *sblock_good,
					    int page_num, int force_write);
224 225 226
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);
227 228 229 230
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);
231 232
static void scrub_page_get(struct scrub_page *spage);
static void scrub_page_put(struct scrub_page *spage);
233 234
static void scrub_parity_get(struct scrub_parity *sparity);
static void scrub_parity_put(struct scrub_parity *sparity);
235
static int scrub_pages(struct scrub_ctx *sctx, u64 logical, u32 len,
236
		       u64 physical, struct btrfs_device *dev, u64 flags,
237
		       u64 gen, int mirror_num, u8 *csum,
238
		       u64 physical_for_dev_replace);
239
static void scrub_bio_end_io(struct bio *bio);
240 241
static void scrub_bio_end_io_worker(struct btrfs_work *work);
static void scrub_block_complete(struct scrub_block *sblock);
242
static void scrub_remap_extent(struct btrfs_fs_info *fs_info,
243
			       u64 extent_logical, u32 extent_len,
244 245 246 247 248 249
			       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);
250
static void scrub_wr_bio_end_io(struct bio *bio);
251
static void scrub_wr_bio_end_io_worker(struct btrfs_work *work);
252
static void scrub_put_ctx(struct scrub_ctx *sctx);
S
Stefan Behrens 已提交
253

254
static inline int scrub_is_page_on_raid56(struct scrub_page *spage)
255
{
256
	return spage->recover &&
257
	       (spage->recover->bioc->map_type & BTRFS_BLOCK_GROUP_RAID56_MASK);
258
}
S
Stefan Behrens 已提交
259

260 261
static void scrub_pending_bio_inc(struct scrub_ctx *sctx)
{
262
	refcount_inc(&sctx->refs);
263 264 265 266 267 268 269
	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);
270
	scrub_put_ctx(sctx);
271 272
}

273
static void __scrub_blocked_if_needed(struct btrfs_fs_info *fs_info)
274 275 276 277 278 279 280 281 282
{
	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);
	}
}

283
static void scrub_pause_on(struct btrfs_fs_info *fs_info)
284 285 286
{
	atomic_inc(&fs_info->scrubs_paused);
	wake_up(&fs_info->scrub_pause_wait);
287
}
288

289 290
static void scrub_pause_off(struct btrfs_fs_info *fs_info)
{
291 292 293 294 295 296 297 298
	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);
}

299 300 301 302 303 304
static void scrub_blocked_if_needed(struct btrfs_fs_info *fs_info)
{
	scrub_pause_on(fs_info);
	scrub_pause_off(fs_info);
}

305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323
/*
 * 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;

324
	lockdep_assert_held(&locks_root->lock);
325 326 327 328 329 330 331 332 333 334 335 336 337 338 339

	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;
		}
	}

340 341 342
	/*
	 * Insert new lock.
	 */
343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367
	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;

368
	lockdep_assert_held(&locks_root->lock);
369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387

	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.
 */
388
static u64 get_full_stripe_logical(struct btrfs_block_group *cache, u64 bytenr)
389 390 391 392 393 394 395 396 397 398 399 400 401
{
	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.
	 */
402 403
	ret = div64_u64(bytenr - cache->start, cache->full_stripe_len) *
			cache->full_stripe_len + cache->start;
404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420
	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)
{
421
	struct btrfs_block_group *bg_cache;
422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467
	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)
{
468
	struct btrfs_block_group *bg_cache;
469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521
	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;
}

522
static void scrub_free_csums(struct scrub_ctx *sctx)
A
Arne Jansen 已提交
523
{
524
	while (!list_empty(&sctx->csum_list)) {
A
Arne Jansen 已提交
525
		struct btrfs_ordered_sum *sum;
526
		sum = list_first_entry(&sctx->csum_list,
A
Arne Jansen 已提交
527 528 529 530 531 532
				       struct btrfs_ordered_sum, list);
		list_del(&sum->list);
		kfree(sum);
	}
}

533
static noinline_for_stack void scrub_free_ctx(struct scrub_ctx *sctx)
A
Arne Jansen 已提交
534 535 536
{
	int i;

537
	if (!sctx)
A
Arne Jansen 已提交
538 539
		return;

540
	/* this can happen when scrub is cancelled */
541 542
	if (sctx->curr != -1) {
		struct scrub_bio *sbio = sctx->bios[sctx->curr];
543 544

		for (i = 0; i < sbio->page_count; i++) {
545
			WARN_ON(!sbio->pagev[i]->page);
546 547 548 549 550
			scrub_block_put(sbio->pagev[i]->sblock);
		}
		bio_put(sbio->bio);
	}

551
	for (i = 0; i < SCRUB_BIOS_PER_SCTX; ++i) {
552
		struct scrub_bio *sbio = sctx->bios[i];
A
Arne Jansen 已提交
553 554 555 556 557 558

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

559
	kfree(sctx->wr_curr_bio);
560 561
	scrub_free_csums(sctx);
	kfree(sctx);
A
Arne Jansen 已提交
562 563
}

564 565
static void scrub_put_ctx(struct scrub_ctx *sctx)
{
566
	if (refcount_dec_and_test(&sctx->refs))
567 568 569
		scrub_free_ctx(sctx);
}

570 571
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
572
{
573
	struct scrub_ctx *sctx;
A
Arne Jansen 已提交
574 575
	int		i;

576
	sctx = kzalloc(sizeof(*sctx), GFP_KERNEL);
577
	if (!sctx)
A
Arne Jansen 已提交
578
		goto nomem;
579
	refcount_set(&sctx->refs, 1);
580
	sctx->is_dev_replace = is_dev_replace;
581
	sctx->pages_per_rd_bio = SCRUB_PAGES_PER_RD_BIO;
582
	sctx->curr = -1;
583
	sctx->fs_info = fs_info;
584
	INIT_LIST_HEAD(&sctx->csum_list);
585
	for (i = 0; i < SCRUB_BIOS_PER_SCTX; ++i) {
A
Arne Jansen 已提交
586 587
		struct scrub_bio *sbio;

588
		sbio = kzalloc(sizeof(*sbio), GFP_KERNEL);
A
Arne Jansen 已提交
589 590
		if (!sbio)
			goto nomem;
591
		sctx->bios[i] = sbio;
A
Arne Jansen 已提交
592 593

		sbio->index = i;
594
		sbio->sctx = sctx;
595
		sbio->page_count = 0;
596 597
		btrfs_init_work(&sbio->work, scrub_bio_end_io_worker, NULL,
				NULL);
A
Arne Jansen 已提交
598

599
		if (i != SCRUB_BIOS_PER_SCTX - 1)
600
			sctx->bios[i]->next_free = i + 1;
601
		else
602 603 604
			sctx->bios[i]->next_free = -1;
	}
	sctx->first_free = 0;
605 606
	atomic_set(&sctx->bios_in_flight, 0);
	atomic_set(&sctx->workers_pending, 0);
607 608 609 610 611
	atomic_set(&sctx->cancel_req, 0);

	spin_lock_init(&sctx->list_lock);
	spin_lock_init(&sctx->stat_lock);
	init_waitqueue_head(&sctx->list_wait);
612
	sctx->throttle_deadline = 0;
613

614 615 616
	WARN_ON(sctx->wr_curr_bio != NULL);
	mutex_init(&sctx->wr_lock);
	sctx->wr_curr_bio = NULL;
617
	if (is_dev_replace) {
618
		WARN_ON(!fs_info->dev_replace.tgtdev);
619
		sctx->pages_per_wr_bio = SCRUB_PAGES_PER_WR_BIO;
620
		sctx->wr_tgtdev = fs_info->dev_replace.tgtdev;
621
		sctx->flush_all_writes = false;
622
	}
623

624
	return sctx;
A
Arne Jansen 已提交
625 626

nomem:
627
	scrub_free_ctx(sctx);
A
Arne Jansen 已提交
628 629 630
	return ERR_PTR(-ENOMEM);
}

631 632
static int scrub_print_warning_inode(u64 inum, u64 offset, u64 root,
				     void *warn_ctx)
633 634 635 636
{
	u32 nlink;
	int ret;
	int i;
637
	unsigned nofs_flag;
638 639
	struct extent_buffer *eb;
	struct btrfs_inode_item *inode_item;
640
	struct scrub_warning *swarn = warn_ctx;
641
	struct btrfs_fs_info *fs_info = swarn->dev->fs_info;
642 643
	struct inode_fs_paths *ipath = NULL;
	struct btrfs_root *local_root;
644
	struct btrfs_key key;
645

D
David Sterba 已提交
646
	local_root = btrfs_get_fs_root(fs_info, root, true);
647 648 649 650 651
	if (IS_ERR(local_root)) {
		ret = PTR_ERR(local_root);
		goto err;
	}

652 653 654
	/*
	 * this makes the path point to (inum INODE_ITEM ioff)
	 */
655 656 657 658 659
	key.objectid = inum;
	key.type = BTRFS_INODE_ITEM_KEY;
	key.offset = 0;

	ret = btrfs_search_slot(NULL, local_root, &key, swarn->path, 0, 0);
660
	if (ret) {
661
		btrfs_put_root(local_root);
662 663 664 665 666 667 668 669 670 671
		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);

672 673 674 675 676 677
	/*
	 * 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();
678
	ipath = init_ipath(4096, local_root, swarn->path);
679
	memalloc_nofs_restore(nofs_flag);
680
	if (IS_ERR(ipath)) {
681
		btrfs_put_root(local_root);
682 683 684 685
		ret = PTR_ERR(ipath);
		ipath = NULL;
		goto err;
	}
686 687 688 689 690 691 692 693 694 695
	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
696
		btrfs_warn_in_rcu(fs_info,
697
"%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
698 699
				  swarn->errstr, swarn->logical,
				  rcu_str_deref(swarn->dev->name),
D
David Sterba 已提交
700
				  swarn->physical,
J
Jeff Mahoney 已提交
701
				  root, inum, offset,
702
				  fs_info->sectorsize, nlink,
J
Jeff Mahoney 已提交
703
				  (char *)(unsigned long)ipath->fspath->val[i]);
704

705
	btrfs_put_root(local_root);
706 707 708 709
	free_ipath(ipath);
	return 0;

err:
J
Jeff Mahoney 已提交
710
	btrfs_warn_in_rcu(fs_info,
D
David Sterba 已提交
711
			  "%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
712 713
			  swarn->errstr, swarn->logical,
			  rcu_str_deref(swarn->dev->name),
D
David Sterba 已提交
714
			  swarn->physical,
J
Jeff Mahoney 已提交
715
			  root, inum, offset, ret);
716 717 718 719 720

	free_ipath(ipath);
	return 0;
}

721
static void scrub_print_warning(const char *errstr, struct scrub_block *sblock)
722
{
723 724
	struct btrfs_device *dev;
	struct btrfs_fs_info *fs_info;
725 726 727 728 729
	struct btrfs_path *path;
	struct btrfs_key found_key;
	struct extent_buffer *eb;
	struct btrfs_extent_item *ei;
	struct scrub_warning swarn;
730 731 732
	unsigned long ptr = 0;
	u64 extent_item_pos;
	u64 flags = 0;
733
	u64 ref_root;
734
	u32 item_size;
735
	u8 ref_level = 0;
736
	int ret;
737

738
	WARN_ON(sblock->page_count < 1);
739
	dev = sblock->pagev[0]->dev;
740
	fs_info = sblock->sctx->fs_info;
741

742
	path = btrfs_alloc_path();
743 744
	if (!path)
		return;
745

D
David Sterba 已提交
746
	swarn.physical = sblock->pagev[0]->physical;
747
	swarn.logical = sblock->pagev[0]->logical;
748
	swarn.errstr = errstr;
749
	swarn.dev = NULL;
750

751 752
	ret = extent_from_logical(fs_info, swarn.logical, path, &found_key,
				  &flags);
753 754 755
	if (ret < 0)
		goto out;

J
Jan Schmidt 已提交
756
	extent_item_pos = swarn.logical - found_key.objectid;
757 758 759 760
	swarn.extent_item_size = found_key.offset;

	eb = path->nodes[0];
	ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
761
	item_size = btrfs_item_size(eb, path->slots[0]);
762

763
	if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
764
		do {
765 766 767
			ret = tree_backref_for_extent(&ptr, eb, &found_key, ei,
						      item_size, &ref_root,
						      &ref_level);
768
			btrfs_warn_in_rcu(fs_info,
D
David Sterba 已提交
769
"%s at logical %llu on dev %s, physical %llu: metadata %s (level %d) in tree %llu",
J
Jeff Mahoney 已提交
770
				errstr, swarn.logical,
771
				rcu_str_deref(dev->name),
D
David Sterba 已提交
772
				swarn.physical,
773 774 775 776
				ref_level ? "node" : "leaf",
				ret < 0 ? -1 : ref_level,
				ret < 0 ? -1 : ref_root);
		} while (ret != 1);
777
		btrfs_release_path(path);
778
	} else {
779
		btrfs_release_path(path);
780
		swarn.path = path;
781
		swarn.dev = dev;
782 783
		iterate_extent_inodes(fs_info, found_key.objectid,
					extent_item_pos, 1,
784
					scrub_print_warning_inode, &swarn, false);
785 786 787 788 789 790
	}

out:
	btrfs_free_path(path);
}

791 792
static inline void scrub_get_recover(struct scrub_recover *recover)
{
793
	refcount_inc(&recover->refs);
794 795
}

796 797
static inline void scrub_put_recover(struct btrfs_fs_info *fs_info,
				     struct scrub_recover *recover)
798
{
799
	if (refcount_dec_and_test(&recover->refs)) {
800
		btrfs_bio_counter_dec(fs_info);
801
		btrfs_put_bioc(recover->bioc);
802 803 804 805
		kfree(recover);
	}
}

A
Arne Jansen 已提交
806
/*
807 808 809 810 811 812
 * 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
813
 */
814
static int scrub_handle_errored_block(struct scrub_block *sblock_to_check)
A
Arne Jansen 已提交
815
{
816
	struct scrub_ctx *sctx = sblock_to_check->sctx;
817
	struct btrfs_device *dev;
818 819 820 821 822 823 824 825 826 827 828
	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;
829
	bool full_stripe_locked;
830
	unsigned int nofs_flag;
831
	static DEFINE_RATELIMIT_STATE(rs, DEFAULT_RATELIMIT_INTERVAL,
832 833 834
				      DEFAULT_RATELIMIT_BURST);

	BUG_ON(sblock_to_check->page_count < 1);
835
	fs_info = sctx->fs_info;
836 837 838 839 840 841 842 843 844 845 846
	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;
	}
847 848 849 850
	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 &
851
			BTRFS_EXTENT_FLAG_DATA);
852 853
	have_csum = sblock_to_check->pagev[0]->have_csum;
	dev = sblock_to_check->pagev[0]->dev;
854

855 856 857
	if (btrfs_is_zoned(fs_info) && !sctx->is_dev_replace)
		return btrfs_repair_one_zone(fs_info, logical);

858 859 860 861 862 863 864 865 866 867
	/*
	 * 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();
868 869 870 871 872 873 874 875 876
	/*
	 * 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) {
877
		memalloc_nofs_restore(nofs_flag);
878 879 880 881 882 883 884 885 886
		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;
	}

887 888 889 890
	/*
	 * 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,
891
	 * sector by sector this time in order to know which sectors
892 893 894 895
	 * 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
896 897 898 899 900 901 902 903 904 905
	 * 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
906 907 908
	 * 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.
909
	 * Only if this is not possible, the sectors are picked from
910 911 912 913 914 915
	 * 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.
	 */

916
	sblocks_for_recheck = kcalloc(BTRFS_MAX_MIRRORS,
917
				      sizeof(*sblocks_for_recheck), GFP_KERNEL);
918
	if (!sblocks_for_recheck) {
919 920 921 922 923
		spin_lock(&sctx->stat_lock);
		sctx->stat.malloc_errors++;
		sctx->stat.read_errors++;
		sctx->stat.uncorrectable_errors++;
		spin_unlock(&sctx->stat_lock);
924
		btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_READ_ERRS);
925
		goto out;
A
Arne Jansen 已提交
926 927
	}

928
	/* setup the context, map the logical blocks and alloc the pages */
929
	ret = scrub_setup_recheck_block(sblock_to_check, sblocks_for_recheck);
930
	if (ret) {
931 932 933 934
		spin_lock(&sctx->stat_lock);
		sctx->stat.read_errors++;
		sctx->stat.uncorrectable_errors++;
		spin_unlock(&sctx->stat_lock);
935
		btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_READ_ERRS);
936 937 938 939
		goto out;
	}
	BUG_ON(failed_mirror_index >= BTRFS_MAX_MIRRORS);
	sblock_bad = sblocks_for_recheck + failed_mirror_index;
940

941
	/* build and submit the bios for the failed mirror, check checksums */
942
	scrub_recheck_block(fs_info, sblock_bad, 1);
A
Arne Jansen 已提交
943

944 945 946 947 948 949 950 951 952 953
	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)
		 */
954 955
		spin_lock(&sctx->stat_lock);
		sctx->stat.unverified_errors++;
956
		sblock_to_check->data_corrected = 1;
957
		spin_unlock(&sctx->stat_lock);
A
Arne Jansen 已提交
958

959 960
		if (sctx->is_dev_replace)
			scrub_write_block_to_dev_replace(sblock_bad);
961
		goto out;
A
Arne Jansen 已提交
962 963
	}

964
	if (!sblock_bad->no_io_error_seen) {
965 966 967
		spin_lock(&sctx->stat_lock);
		sctx->stat.read_errors++;
		spin_unlock(&sctx->stat_lock);
968
		if (__ratelimit(&rs))
969
			scrub_print_warning("i/o error", sblock_to_check);
970
		btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_READ_ERRS);
971
	} else if (sblock_bad->checksum_error) {
972 973 974
		spin_lock(&sctx->stat_lock);
		sctx->stat.csum_errors++;
		spin_unlock(&sctx->stat_lock);
975
		if (__ratelimit(&rs))
976
			scrub_print_warning("checksum error", sblock_to_check);
977
		btrfs_dev_stat_inc_and_print(dev,
978
					     BTRFS_DEV_STAT_CORRUPTION_ERRS);
979
	} else if (sblock_bad->header_error) {
980 981 982
		spin_lock(&sctx->stat_lock);
		sctx->stat.verify_errors++;
		spin_unlock(&sctx->stat_lock);
983
		if (__ratelimit(&rs))
984 985
			scrub_print_warning("checksum/header error",
					    sblock_to_check);
986
		if (sblock_bad->generation_error)
987
			btrfs_dev_stat_inc_and_print(dev,
988 989
				BTRFS_DEV_STAT_GENERATION_ERRS);
		else
990
			btrfs_dev_stat_inc_and_print(dev,
991
				BTRFS_DEV_STAT_CORRUPTION_ERRS);
992
	}
A
Arne Jansen 已提交
993

994 995 996 997
	if (sctx->readonly) {
		ASSERT(!sctx->is_dev_replace);
		goto out;
	}
A
Arne Jansen 已提交
998

999 1000
	/*
	 * now build and submit the bios for the other mirrors, check
1001 1002
	 * checksums.
	 * First try to pick the mirror which is completely without I/O
1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013
	 * 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).
	 */
1014
	for (mirror_index = 0; ;mirror_index++) {
1015
		struct scrub_block *sblock_other;
1016

1017 1018
		if (mirror_index == failed_mirror_index)
			continue;
1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029

		/* 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;
1030
			int max_allowed = r->bioc->num_stripes - r->bioc->num_tgtdevs;
1031 1032 1033 1034 1035 1036 1037 1038 1039 1040

			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;
		}
1041 1042

		/* build and submit the bios, check checksums */
1043
		scrub_recheck_block(fs_info, sblock_other, 0);
1044 1045

		if (!sblock_other->header_error &&
1046 1047
		    !sblock_other->checksum_error &&
		    sblock_other->no_io_error_seen) {
1048 1049
			if (sctx->is_dev_replace) {
				scrub_write_block_to_dev_replace(sblock_other);
1050
				goto corrected_error;
1051 1052
			} else {
				ret = scrub_repair_block_from_good_copy(
1053 1054 1055
						sblock_bad, sblock_other);
				if (!ret)
					goto corrected_error;
1056
			}
1057 1058
		}
	}
A
Arne Jansen 已提交
1059

1060 1061
	if (sblock_bad->no_io_error_seen && !sctx->is_dev_replace)
		goto did_not_correct_error;
1062 1063 1064

	/*
	 * In case of I/O errors in the area that is supposed to be
1065 1066
	 * repaired, continue by picking good copies of those sectors.
	 * Select the good sectors from mirrors to rewrite bad sectors from
1067 1068 1069 1070 1071
	 * 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
1072
	 * all possible combinations of sectors from the different mirrors
1073
	 * until the checksum verification succeeds. For example, when
1074
	 * the 2nd sector of mirror #1 faces I/O errors, and the 2nd sector
1075
	 * of mirror #2 is readable but the final checksum test fails,
1076
	 * then the 2nd sector of mirror #3 could be tried, whether now
1077
	 * the final checksum succeeds. But this would be a rare
1078 1079 1080 1081
	 * 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
1082
	 * disks) instead of on sectorsize. Then maybe 512 byte of one
1083
	 * mirror could be repaired by taking 512 byte of a different
1084
	 * mirror, even if other 512 byte sectors in the same sectorsize
1085
	 * area are unreadable.
A
Arne Jansen 已提交
1086
	 */
1087
	success = 1;
1088 1089
	for (page_num = 0; page_num < sblock_bad->page_count;
	     page_num++) {
1090
		struct scrub_page *spage_bad = sblock_bad->pagev[page_num];
1091
		struct scrub_block *sblock_other = NULL;
1092

1093
		/* skip no-io-error page in scrub */
1094
		if (!spage_bad->io_error && !sctx->is_dev_replace)
A
Arne Jansen 已提交
1095
			continue;
1096

1097 1098 1099 1100 1101 1102 1103 1104 1105
		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;
1106
		} else if (spage_bad->io_error) {
1107
			/* try to find no-io-error page in mirrors */
1108 1109 1110 1111 1112 1113 1114 1115 1116
			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;
1117 1118
				}
			}
1119 1120
			if (!sblock_other)
				success = 0;
I
Ilya Dryomov 已提交
1121
		}
A
Arne Jansen 已提交
1122

1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135
		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) {
1136
				atomic64_inc(
1137
					&fs_info->dev_replace.num_write_errors);
1138 1139 1140 1141 1142 1143 1144
				success = 0;
			}
		} else if (sblock_other) {
			ret = scrub_repair_page_from_good_copy(sblock_bad,
							       sblock_other,
							       page_num, 0);
			if (0 == ret)
1145
				spage_bad->io_error = 0;
1146 1147
			else
				success = 0;
1148
		}
A
Arne Jansen 已提交
1149 1150
	}

1151
	if (success && !sctx->is_dev_replace) {
1152 1153 1154 1155 1156 1157 1158 1159 1160 1161
		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.
			 */
1162
			scrub_recheck_block(fs_info, sblock_bad, 1);
1163
			if (!sblock_bad->header_error &&
1164 1165 1166 1167 1168 1169 1170
			    !sblock_bad->checksum_error &&
			    sblock_bad->no_io_error_seen)
				goto corrected_error;
			else
				goto did_not_correct_error;
		} else {
corrected_error:
1171 1172
			spin_lock(&sctx->stat_lock);
			sctx->stat.corrected_errors++;
1173
			sblock_to_check->data_corrected = 1;
1174
			spin_unlock(&sctx->stat_lock);
1175 1176
			btrfs_err_rl_in_rcu(fs_info,
				"fixed up error at logical %llu on dev %s",
1177
				logical, rcu_str_deref(dev->name));
A
Arne Jansen 已提交
1178
		}
1179 1180
	} else {
did_not_correct_error:
1181 1182 1183
		spin_lock(&sctx->stat_lock);
		sctx->stat.uncorrectable_errors++;
		spin_unlock(&sctx->stat_lock);
1184 1185
		btrfs_err_rl_in_rcu(fs_info,
			"unable to fixup (regular) error at logical %llu on dev %s",
1186
			logical, rcu_str_deref(dev->name));
I
Ilya Dryomov 已提交
1187
	}
A
Arne Jansen 已提交
1188

1189 1190 1191 1192 1193 1194
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;
1195
			struct scrub_recover *recover;
1196 1197
			int page_index;

1198 1199 1200
			for (page_index = 0; page_index < sblock->page_count;
			     page_index++) {
				sblock->pagev[page_index]->sblock = NULL;
1201 1202
				recover = sblock->pagev[page_index]->recover;
				if (recover) {
1203
					scrub_put_recover(fs_info, recover);
1204 1205 1206
					sblock->pagev[page_index]->recover =
									NULL;
				}
1207 1208
				scrub_page_put(sblock->pagev[page_index]);
			}
1209 1210 1211
		}
		kfree(sblocks_for_recheck);
	}
A
Arne Jansen 已提交
1212

1213
	ret = unlock_full_stripe(fs_info, logical, full_stripe_locked);
1214
	memalloc_nofs_restore(nofs_flag);
1215 1216
	if (ret < 0)
		return ret;
1217 1218
	return 0;
}
A
Arne Jansen 已提交
1219

1220
static inline int scrub_nr_raid_mirrors(struct btrfs_io_context *bioc)
1221
{
1222
	if (bioc->map_type & BTRFS_BLOCK_GROUP_RAID5)
Z
Zhao Lei 已提交
1223
		return 2;
1224
	else if (bioc->map_type & BTRFS_BLOCK_GROUP_RAID6)
Z
Zhao Lei 已提交
1225 1226
		return 3;
	else
1227
		return (int)bioc->num_stripes;
1228 1229
}

Z
Zhao Lei 已提交
1230 1231
static inline void scrub_stripe_index_and_offset(u64 logical, u64 map_type,
						 u64 *raid_map,
1232 1233 1234 1235 1236 1237 1238
						 u64 mapped_length,
						 int nstripes, int mirror,
						 int *stripe_index,
						 u64 *stripe_offset)
{
	int i;

1239
	if (map_type & BTRFS_BLOCK_GROUP_RAID56_MASK) {
1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259
		/* 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;
	}
}

1260
static int scrub_setup_recheck_block(struct scrub_block *original_sblock,
1261 1262
				     struct scrub_block *sblocks_for_recheck)
{
1263
	struct scrub_ctx *sctx = original_sblock->sctx;
1264
	struct btrfs_fs_info *fs_info = sctx->fs_info;
1265
	u64 length = original_sblock->page_count * fs_info->sectorsize;
1266
	u64 logical = original_sblock->pagev[0]->logical;
1267 1268 1269
	u64 generation = original_sblock->pagev[0]->generation;
	u64 flags = original_sblock->pagev[0]->flags;
	u64 have_csum = original_sblock->pagev[0]->have_csum;
1270
	struct scrub_recover *recover;
1271
	struct btrfs_io_context *bioc;
1272 1273 1274 1275
	u64 sublen;
	u64 mapped_length;
	u64 stripe_offset;
	int stripe_index;
1276
	int page_index = 0;
1277
	int mirror_index;
1278
	int nmirrors;
1279 1280 1281
	int ret;

	/*
1282
	 * note: the two members refs and outstanding_pages
1283 1284 1285 1286 1287
	 * are not used (and not set) in the blocks that are used for
	 * the recheck procedure
	 */

	while (length > 0) {
1288
		sublen = min_t(u64, length, fs_info->sectorsize);
1289
		mapped_length = sublen;
1290
		bioc = NULL;
A
Arne Jansen 已提交
1291

1292
		/*
1293 1294
		 * With a length of sectorsize, each returned stripe represents
		 * one mirror
1295
		 */
1296
		btrfs_bio_counter_inc_blocked(fs_info);
1297
		ret = btrfs_map_sblock(fs_info, BTRFS_MAP_GET_READ_MIRRORS,
1298 1299 1300
				       logical, &mapped_length, &bioc);
		if (ret || !bioc || mapped_length < sublen) {
			btrfs_put_bioc(bioc);
1301
			btrfs_bio_counter_dec(fs_info);
1302 1303
			return -EIO;
		}
A
Arne Jansen 已提交
1304

1305 1306
		recover = kzalloc(sizeof(struct scrub_recover), GFP_NOFS);
		if (!recover) {
1307
			btrfs_put_bioc(bioc);
1308
			btrfs_bio_counter_dec(fs_info);
1309 1310 1311
			return -ENOMEM;
		}

1312
		refcount_set(&recover->refs, 1);
1313
		recover->bioc = bioc;
1314 1315
		recover->map_length = mapped_length;

1316
		BUG_ON(page_index >= SCRUB_MAX_PAGES_PER_BLOCK);
1317

1318
		nmirrors = min(scrub_nr_raid_mirrors(bioc), BTRFS_MAX_MIRRORS);
Z
Zhao Lei 已提交
1319

1320
		for (mirror_index = 0; mirror_index < nmirrors;
1321 1322
		     mirror_index++) {
			struct scrub_block *sblock;
1323
			struct scrub_page *spage;
1324 1325

			sblock = sblocks_for_recheck + mirror_index;
1326
			sblock->sctx = sctx;
1327

1328 1329
			spage = kzalloc(sizeof(*spage), GFP_NOFS);
			if (!spage) {
1330
leave_nomem:
1331 1332 1333
				spin_lock(&sctx->stat_lock);
				sctx->stat.malloc_errors++;
				spin_unlock(&sctx->stat_lock);
1334
				scrub_put_recover(fs_info, recover);
1335 1336
				return -ENOMEM;
			}
1337 1338 1339 1340 1341 1342 1343
			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;
1344
			if (have_csum)
1345
				memcpy(spage->csum,
1346
				       original_sblock->pagev[0]->csum,
1347
				       sctx->fs_info->csum_size);
1348

Z
Zhao Lei 已提交
1349
			scrub_stripe_index_and_offset(logical,
1350 1351
						      bioc->map_type,
						      bioc->raid_map,
1352
						      mapped_length,
1353 1354
						      bioc->num_stripes -
						      bioc->num_tgtdevs,
1355 1356 1357
						      mirror_index,
						      &stripe_index,
						      &stripe_offset);
1358
			spage->physical = bioc->stripes[stripe_index].physical +
1359
					 stripe_offset;
1360
			spage->dev = bioc->stripes[stripe_index].dev;
1361

1362
			BUG_ON(page_index >= original_sblock->page_count);
1363
			spage->physical_for_dev_replace =
1364 1365
				original_sblock->pagev[page_index]->
				physical_for_dev_replace;
1366
			/* for missing devices, dev->bdev is NULL */
1367
			spage->mirror_num = mirror_index + 1;
1368
			sblock->page_count++;
1369 1370
			spage->page = alloc_page(GFP_NOFS);
			if (!spage->page)
1371
				goto leave_nomem;
1372 1373

			scrub_get_recover(recover);
1374
			spage->recover = recover;
1375
		}
1376
		scrub_put_recover(fs_info, recover);
1377 1378 1379 1380 1381 1382
		length -= sublen;
		logical += sublen;
		page_index++;
	}

	return 0;
I
Ilya Dryomov 已提交
1383 1384
}

1385
static void scrub_bio_wait_endio(struct bio *bio)
1386
{
1387
	complete(bio->bi_private);
1388 1389 1390 1391
}

static int scrub_submit_raid56_bio_wait(struct btrfs_fs_info *fs_info,
					struct bio *bio,
1392
					struct scrub_page *spage)
1393
{
1394
	DECLARE_COMPLETION_ONSTACK(done);
1395
	int ret;
1396
	int mirror_num;
1397

1398
	bio->bi_iter.bi_sector = spage->logical >> 9;
1399 1400 1401
	bio->bi_private = &done;
	bio->bi_end_io = scrub_bio_wait_endio;

1402
	mirror_num = spage->sblock->pagev[0]->mirror_num;
1403
	ret = raid56_parity_recover(bio, spage->recover->bioc,
1404
				    spage->recover->map_length,
1405
				    mirror_num, 0);
1406 1407 1408
	if (ret)
		return ret;

1409 1410
	wait_for_completion_io(&done);
	return blk_status_to_errno(bio->bi_status);
1411 1412
}

L
Liu Bo 已提交
1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424
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;

1425
	bio = btrfs_bio_alloc(BIO_MAX_VECS);
L
Liu Bo 已提交
1426 1427 1428
	bio_set_dev(bio, first_page->dev->bdev);

	for (page_num = 0; page_num < sblock->page_count; page_num++) {
1429
		struct scrub_page *spage = sblock->pagev[page_num];
L
Liu Bo 已提交
1430

1431 1432
		WARN_ON(!spage->page);
		bio_add_page(bio, spage->page, PAGE_SIZE, 0);
L
Liu Bo 已提交
1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451
	}

	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;
}

1452 1453 1454 1455 1456 1457 1458
/*
 * 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.
 */
1459
static void scrub_recheck_block(struct btrfs_fs_info *fs_info,
1460 1461
				struct scrub_block *sblock,
				int retry_failed_mirror)
I
Ilya Dryomov 已提交
1462
{
1463
	int page_num;
I
Ilya Dryomov 已提交
1464

1465
	sblock->no_io_error_seen = 1;
I
Ilya Dryomov 已提交
1466

L
Liu Bo 已提交
1467 1468 1469 1470
	/* 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);

1471 1472
	for (page_num = 0; page_num < sblock->page_count; page_num++) {
		struct bio *bio;
1473
		struct scrub_page *spage = sblock->pagev[page_num];
1474

1475 1476
		if (spage->dev->bdev == NULL) {
			spage->io_error = 1;
1477 1478 1479 1480
			sblock->no_io_error_seen = 0;
			continue;
		}

1481
		WARN_ON(!spage->page);
1482
		bio = btrfs_bio_alloc(1);
1483
		bio_set_dev(bio, spage->dev->bdev);
1484

1485
		bio_add_page(bio, spage->page, fs_info->sectorsize, 0);
1486
		bio->bi_iter.bi_sector = spage->physical >> 9;
L
Liu Bo 已提交
1487
		bio->bi_opf = REQ_OP_READ;
1488

L
Liu Bo 已提交
1489
		if (btrfsic_submit_bio_wait(bio)) {
1490
			spage->io_error = 1;
L
Liu Bo 已提交
1491
			sblock->no_io_error_seen = 0;
1492
		}
1493

1494 1495
		bio_put(bio);
	}
I
Ilya Dryomov 已提交
1496

1497
	if (sblock->no_io_error_seen)
1498
		scrub_recheck_block_checksum(sblock);
A
Arne Jansen 已提交
1499 1500
}

M
Miao Xie 已提交
1501 1502 1503 1504 1505 1506
static inline int scrub_check_fsid(u8 fsid[],
				   struct scrub_page *spage)
{
	struct btrfs_fs_devices *fs_devices = spage->dev->fs_devices;
	int ret;

1507
	ret = memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE);
M
Miao Xie 已提交
1508 1509 1510
	return !ret;
}

1511
static void scrub_recheck_block_checksum(struct scrub_block *sblock)
A
Arne Jansen 已提交
1512
{
1513 1514 1515
	sblock->header_error = 0;
	sblock->checksum_error = 0;
	sblock->generation_error = 0;
1516

1517 1518 1519 1520
	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
1521 1522
}

1523
static int scrub_repair_block_from_good_copy(struct scrub_block *sblock_bad,
1524
					     struct scrub_block *sblock_good)
1525 1526 1527
{
	int page_num;
	int ret = 0;
I
Ilya Dryomov 已提交
1528

1529 1530
	for (page_num = 0; page_num < sblock_bad->page_count; page_num++) {
		int ret_sub;
I
Ilya Dryomov 已提交
1531

1532 1533
		ret_sub = scrub_repair_page_from_good_copy(sblock_bad,
							   sblock_good,
1534
							   page_num, 1);
1535 1536
		if (ret_sub)
			ret = ret_sub;
A
Arne Jansen 已提交
1537
	}
1538 1539 1540 1541 1542 1543 1544 1545

	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)
{
1546 1547
	struct scrub_page *spage_bad = sblock_bad->pagev[page_num];
	struct scrub_page *spage_good = sblock_good->pagev[page_num];
1548
	struct btrfs_fs_info *fs_info = sblock_bad->sctx->fs_info;
1549
	const u32 sectorsize = fs_info->sectorsize;
1550

1551 1552
	BUG_ON(spage_bad->page == NULL);
	BUG_ON(spage_good->page == NULL);
1553
	if (force_write || sblock_bad->header_error ||
1554
	    sblock_bad->checksum_error || spage_bad->io_error) {
1555 1556 1557
		struct bio *bio;
		int ret;

1558
		if (!spage_bad->dev->bdev) {
1559
			btrfs_warn_rl(fs_info,
J
Jeff Mahoney 已提交
1560
				"scrub_repair_page_from_good_copy(bdev == NULL) is unexpected");
1561 1562 1563
			return -EIO;
		}

1564
		bio = btrfs_bio_alloc(1);
1565 1566
		bio_set_dev(bio, spage_bad->dev->bdev);
		bio->bi_iter.bi_sector = spage_bad->physical >> 9;
D
David Sterba 已提交
1567
		bio->bi_opf = REQ_OP_WRITE;
1568

1569 1570
		ret = bio_add_page(bio, spage_good->page, sectorsize, 0);
		if (ret != sectorsize) {
1571 1572
			bio_put(bio);
			return -EIO;
1573
		}
1574

1575
		if (btrfsic_submit_bio_wait(bio)) {
1576
			btrfs_dev_stat_inc_and_print(spage_bad->dev,
1577
				BTRFS_DEV_STAT_WRITE_ERRS);
1578
			atomic64_inc(&fs_info->dev_replace.num_write_errors);
1579 1580 1581
			bio_put(bio);
			return -EIO;
		}
1582
		bio_put(bio);
A
Arne Jansen 已提交
1583 1584
	}

1585 1586 1587
	return 0;
}

1588 1589
static void scrub_write_block_to_dev_replace(struct scrub_block *sblock)
{
1590
	struct btrfs_fs_info *fs_info = sblock->sctx->fs_info;
1591 1592
	int page_num;

1593 1594 1595 1596 1597 1598 1599
	/*
	 * 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;

1600 1601 1602 1603 1604
	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)
1605
			atomic64_inc(&fs_info->dev_replace.num_write_errors);
1606 1607 1608 1609 1610 1611 1612 1613 1614
	}
}

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);
1615 1616
	if (spage->io_error)
		clear_page(page_address(spage->page));
1617 1618 1619 1620

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

1621 1622 1623 1624 1625 1626 1627 1628
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;

1629 1630 1631
	if (!btrfs_dev_is_sequential(sctx->wr_tgtdev, physical))
		return 0;

1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642
	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;
}

1643 1644 1645 1646 1647
static int scrub_add_page_to_wr_bio(struct scrub_ctx *sctx,
				    struct scrub_page *spage)
{
	struct scrub_bio *sbio;
	int ret;
1648
	const u32 sectorsize = sctx->fs_info->sectorsize;
1649

1650
	mutex_lock(&sctx->wr_lock);
1651
again:
1652 1653
	if (!sctx->wr_curr_bio) {
		sctx->wr_curr_bio = kzalloc(sizeof(*sctx->wr_curr_bio),
1654
					      GFP_KERNEL);
1655 1656
		if (!sctx->wr_curr_bio) {
			mutex_unlock(&sctx->wr_lock);
1657 1658
			return -ENOMEM;
		}
1659 1660
		sctx->wr_curr_bio->sctx = sctx;
		sctx->wr_curr_bio->page_count = 0;
1661
	}
1662
	sbio = sctx->wr_curr_bio;
1663 1664 1665
	if (sbio->page_count == 0) {
		struct bio *bio;

1666 1667 1668 1669 1670 1671 1672
		ret = fill_writer_pointer_gap(sctx,
					      spage->physical_for_dev_replace);
		if (ret) {
			mutex_unlock(&sctx->wr_lock);
			return ret;
		}

1673 1674
		sbio->physical = spage->physical_for_dev_replace;
		sbio->logical = spage->logical;
1675
		sbio->dev = sctx->wr_tgtdev;
1676 1677
		bio = sbio->bio;
		if (!bio) {
1678
			bio = btrfs_bio_alloc(sctx->pages_per_wr_bio);
1679 1680 1681 1682 1683
			sbio->bio = bio;
		}

		bio->bi_private = sbio;
		bio->bi_end_io = scrub_wr_bio_end_io;
1684
		bio_set_dev(bio, sbio->dev->bdev);
1685
		bio->bi_iter.bi_sector = sbio->physical >> 9;
D
David Sterba 已提交
1686
		bio->bi_opf = REQ_OP_WRITE;
1687
		sbio->status = 0;
1688
	} else if (sbio->physical + sbio->page_count * sectorsize !=
1689
		   spage->physical_for_dev_replace ||
1690
		   sbio->logical + sbio->page_count * sectorsize !=
1691 1692 1693 1694 1695
		   spage->logical) {
		scrub_wr_submit(sctx);
		goto again;
	}

1696 1697
	ret = bio_add_page(sbio->bio, spage->page, sectorsize, 0);
	if (ret != sectorsize) {
1698 1699 1700
		if (sbio->page_count < 1) {
			bio_put(sbio->bio);
			sbio->bio = NULL;
1701
			mutex_unlock(&sctx->wr_lock);
1702 1703 1704 1705 1706 1707 1708 1709 1710
			return -EIO;
		}
		scrub_wr_submit(sctx);
		goto again;
	}

	sbio->pagev[sbio->page_count] = spage;
	scrub_page_get(spage);
	sbio->page_count++;
1711
	if (sbio->page_count == sctx->pages_per_wr_bio)
1712
		scrub_wr_submit(sctx);
1713
	mutex_unlock(&sctx->wr_lock);
1714 1715 1716 1717 1718 1719 1720 1721

	return 0;
}

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

1722
	if (!sctx->wr_curr_bio)
1723 1724
		return;

1725 1726
	sbio = sctx->wr_curr_bio;
	sctx->wr_curr_bio = NULL;
1727
	WARN_ON(!sbio->bio->bi_bdev);
1728 1729 1730 1731 1732
	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 */
1733
	btrfsic_submit_bio(sbio->bio);
1734 1735

	if (btrfs_is_zoned(sctx->fs_info))
1736 1737
		sctx->write_pointer = sbio->physical + sbio->page_count *
			sctx->fs_info->sectorsize;
1738 1739
}

1740
static void scrub_wr_bio_end_io(struct bio *bio)
1741 1742
{
	struct scrub_bio *sbio = bio->bi_private;
1743
	struct btrfs_fs_info *fs_info = sbio->dev->fs_info;
1744

1745
	sbio->status = bio->bi_status;
1746 1747
	sbio->bio = bio;

1748
	btrfs_init_work(&sbio->work, scrub_wr_bio_end_io_worker, NULL, NULL);
1749
	btrfs_queue_work(fs_info->scrub_wr_completion_workers, &sbio->work);
1750 1751 1752 1753 1754 1755 1756 1757 1758
}

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;

	WARN_ON(sbio->page_count > SCRUB_PAGES_PER_WR_BIO);
1759
	if (sbio->status) {
1760
		struct btrfs_dev_replace *dev_replace =
1761
			&sbio->sctx->fs_info->dev_replace;
1762 1763 1764 1765 1766

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

			spage->io_error = 1;
1767
			atomic64_inc(&dev_replace->num_write_errors);
1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779
		}
	}

	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)
1780 1781 1782 1783
{
	u64 flags;
	int ret;

1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795
	/*
	 * 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;

1796 1797
	WARN_ON(sblock->page_count < 1);
	flags = sblock->pagev[0]->flags;
1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808
	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);
1809 1810

	return ret;
A
Arne Jansen 已提交
1811 1812
}

1813
static int scrub_checksum_data(struct scrub_block *sblock)
A
Arne Jansen 已提交
1814
{
1815
	struct scrub_ctx *sctx = sblock->sctx;
1816 1817
	struct btrfs_fs_info *fs_info = sctx->fs_info;
	SHASH_DESC_ON_STACK(shash, fs_info->csum_shash);
A
Arne Jansen 已提交
1818
	u8 csum[BTRFS_CSUM_SIZE];
1819
	struct scrub_page *spage;
1820
	char *kaddr;
A
Arne Jansen 已提交
1821

1822
	BUG_ON(sblock->page_count < 1);
1823 1824
	spage = sblock->pagev[0];
	if (!spage->have_csum)
A
Arne Jansen 已提交
1825 1826
		return 0;

1827
	kaddr = page_address(spage->page);
1828

1829 1830
	shash->tfm = fs_info->csum_shash;
	crypto_shash_init(shash);
1831

1832 1833 1834 1835 1836
	/*
	 * 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
1837

1838 1839
	if (memcmp(csum, spage->csum, fs_info->csum_size))
		sblock->checksum_error = 1;
1840
	return sblock->checksum_error;
A
Arne Jansen 已提交
1841 1842
}

1843
static int scrub_checksum_tree_block(struct scrub_block *sblock)
A
Arne Jansen 已提交
1844
{
1845
	struct scrub_ctx *sctx = sblock->sctx;
A
Arne Jansen 已提交
1846
	struct btrfs_header *h;
1847
	struct btrfs_fs_info *fs_info = sctx->fs_info;
1848
	SHASH_DESC_ON_STACK(shash, fs_info->csum_shash);
1849 1850
	u8 calculated_csum[BTRFS_CSUM_SIZE];
	u8 on_disk_csum[BTRFS_CSUM_SIZE];
1851 1852 1853 1854 1855 1856 1857
	/*
	 * 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;
1858
	int i;
1859
	struct scrub_page *spage;
1860
	char *kaddr;
1861

1862
	BUG_ON(sblock->page_count < 1);
1863 1864 1865 1866

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

1867 1868
	spage = sblock->pagev[0];
	kaddr = page_address(spage->page);
1869
	h = (struct btrfs_header *)kaddr;
1870
	memcpy(on_disk_csum, h->csum, sctx->fs_info->csum_size);
A
Arne Jansen 已提交
1871 1872 1873 1874 1875 1876

	/*
	 * we don't use the getter functions here, as we
	 * a) don't have an extent buffer and
	 * b) the page is already kmapped
	 */
1877
	if (spage->logical != btrfs_stack_header_bytenr(h))
1878
		sblock->header_error = 1;
A
Arne Jansen 已提交
1879

1880
	if (spage->generation != btrfs_stack_header_generation(h)) {
1881 1882 1883
		sblock->header_error = 1;
		sblock->generation_error = 1;
	}
A
Arne Jansen 已提交
1884

1885
	if (!scrub_check_fsid(h->fsid, spage))
1886
		sblock->header_error = 1;
A
Arne Jansen 已提交
1887 1888 1889

	if (memcmp(h->chunk_tree_uuid, fs_info->chunk_tree_uuid,
		   BTRFS_UUID_SIZE))
1890
		sblock->header_error = 1;
A
Arne Jansen 已提交
1891

1892 1893 1894
	shash->tfm = fs_info->csum_shash;
	crypto_shash_init(shash);
	crypto_shash_update(shash, kaddr + BTRFS_CSUM_SIZE,
1895
			    sectorsize - BTRFS_CSUM_SIZE);
1896

1897
	for (i = 1; i < num_sectors; i++) {
1898
		kaddr = page_address(sblock->pagev[i]->page);
1899
		crypto_shash_update(shash, kaddr, sectorsize);
1900 1901
	}

1902
	crypto_shash_final(shash, calculated_csum);
1903
	if (memcmp(calculated_csum, on_disk_csum, sctx->fs_info->csum_size))
1904
		sblock->checksum_error = 1;
A
Arne Jansen 已提交
1905

1906
	return sblock->header_error || sblock->checksum_error;
A
Arne Jansen 已提交
1907 1908
}

1909
static int scrub_checksum_super(struct scrub_block *sblock)
A
Arne Jansen 已提交
1910 1911
{
	struct btrfs_super_block *s;
1912
	struct scrub_ctx *sctx = sblock->sctx;
1913 1914
	struct btrfs_fs_info *fs_info = sctx->fs_info;
	SHASH_DESC_ON_STACK(shash, fs_info->csum_shash);
1915
	u8 calculated_csum[BTRFS_CSUM_SIZE];
1916
	struct scrub_page *spage;
1917
	char *kaddr;
1918 1919
	int fail_gen = 0;
	int fail_cor = 0;
1920

1921
	BUG_ON(sblock->page_count < 1);
1922 1923
	spage = sblock->pagev[0];
	kaddr = page_address(spage->page);
1924
	s = (struct btrfs_super_block *)kaddr;
A
Arne Jansen 已提交
1925

1926
	if (spage->logical != btrfs_super_bytenr(s))
1927
		++fail_cor;
A
Arne Jansen 已提交
1928

1929
	if (spage->generation != btrfs_super_generation(s))
1930
		++fail_gen;
A
Arne Jansen 已提交
1931

1932
	if (!scrub_check_fsid(s->fsid, spage))
1933
		++fail_cor;
A
Arne Jansen 已提交
1934

1935 1936 1937 1938
	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);
1939

1940
	if (memcmp(calculated_csum, s->csum, sctx->fs_info->csum_size))
1941
		++fail_cor;
A
Arne Jansen 已提交
1942

1943
	if (fail_cor + fail_gen) {
A
Arne Jansen 已提交
1944 1945 1946 1947 1948
		/*
		 * if we find an error in a super block, we just report it.
		 * They will get written with the next transaction commit
		 * anyway
		 */
1949 1950 1951
		spin_lock(&sctx->stat_lock);
		++sctx->stat.super_errors;
		spin_unlock(&sctx->stat_lock);
1952
		if (fail_cor)
1953
			btrfs_dev_stat_inc_and_print(spage->dev,
1954 1955
				BTRFS_DEV_STAT_CORRUPTION_ERRS);
		else
1956
			btrfs_dev_stat_inc_and_print(spage->dev,
1957
				BTRFS_DEV_STAT_GENERATION_ERRS);
A
Arne Jansen 已提交
1958 1959
	}

1960
	return fail_cor + fail_gen;
A
Arne Jansen 已提交
1961 1962
}

1963 1964
static void scrub_block_get(struct scrub_block *sblock)
{
1965
	refcount_inc(&sblock->refs);
1966 1967 1968 1969
}

static void scrub_block_put(struct scrub_block *sblock)
{
1970
	if (refcount_dec_and_test(&sblock->refs)) {
1971 1972
		int i;

1973 1974 1975
		if (sblock->sparity)
			scrub_parity_put(sblock->sparity);

1976
		for (i = 0; i < sblock->page_count; i++)
1977
			scrub_page_put(sblock->pagev[i]);
1978 1979 1980 1981
		kfree(sblock);
	}
}

1982 1983
static void scrub_page_get(struct scrub_page *spage)
{
1984
	atomic_inc(&spage->refs);
1985 1986 1987 1988
}

static void scrub_page_put(struct scrub_page *spage)
{
1989
	if (atomic_dec_and_test(&spage->refs)) {
1990 1991 1992 1993 1994 1995
		if (spage->page)
			__free_page(spage->page);
		kfree(spage);
	}
}

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 2048 2049 2050 2051 2052 2053 2054
/*
 * 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;
}

2055
static void scrub_submit(struct scrub_ctx *sctx)
A
Arne Jansen 已提交
2056 2057 2058
{
	struct scrub_bio *sbio;

2059
	if (sctx->curr == -1)
S
Stefan Behrens 已提交
2060
		return;
A
Arne Jansen 已提交
2061

2062 2063
	scrub_throttle(sctx);

2064 2065
	sbio = sctx->bios[sctx->curr];
	sctx->curr = -1;
2066
	scrub_pending_bio_inc(sctx);
2067
	btrfsic_submit_bio(sbio->bio);
A
Arne Jansen 已提交
2068 2069
}

2070 2071
static int scrub_add_page_to_rd_bio(struct scrub_ctx *sctx,
				    struct scrub_page *spage)
A
Arne Jansen 已提交
2072
{
2073
	struct scrub_block *sblock = spage->sblock;
A
Arne Jansen 已提交
2074
	struct scrub_bio *sbio;
2075
	const u32 sectorsize = sctx->fs_info->sectorsize;
2076
	int ret;
A
Arne Jansen 已提交
2077 2078 2079 2080 2081

again:
	/*
	 * grab a fresh bio or wait for one to become available
	 */
2082 2083 2084 2085 2086 2087 2088 2089
	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
2090
		} else {
2091 2092
			spin_unlock(&sctx->list_lock);
			wait_event(sctx->list_wait, sctx->first_free != -1);
A
Arne Jansen 已提交
2093 2094
		}
	}
2095
	sbio = sctx->bios[sctx->curr];
2096
	if (sbio->page_count == 0) {
2097 2098
		struct bio *bio;

2099 2100
		sbio->physical = spage->physical;
		sbio->logical = spage->logical;
2101
		sbio->dev = spage->dev;
2102 2103
		bio = sbio->bio;
		if (!bio) {
2104
			bio = btrfs_bio_alloc(sctx->pages_per_rd_bio);
2105 2106
			sbio->bio = bio;
		}
2107 2108 2109

		bio->bi_private = sbio;
		bio->bi_end_io = scrub_bio_end_io;
2110
		bio_set_dev(bio, sbio->dev->bdev);
2111
		bio->bi_iter.bi_sector = sbio->physical >> 9;
D
David Sterba 已提交
2112
		bio->bi_opf = REQ_OP_READ;
2113
		sbio->status = 0;
2114
	} else if (sbio->physical + sbio->page_count * sectorsize !=
2115
		   spage->physical ||
2116
		   sbio->logical + sbio->page_count * sectorsize !=
2117 2118
		   spage->logical ||
		   sbio->dev != spage->dev) {
2119
		scrub_submit(sctx);
A
Arne Jansen 已提交
2120 2121
		goto again;
	}
2122

2123
	sbio->pagev[sbio->page_count] = spage;
2124 2125
	ret = bio_add_page(sbio->bio, spage->page, sectorsize, 0);
	if (ret != sectorsize) {
2126 2127 2128 2129 2130
		if (sbio->page_count < 1) {
			bio_put(sbio->bio);
			sbio->bio = NULL;
			return -EIO;
		}
2131
		scrub_submit(sctx);
2132 2133 2134
		goto again;
	}

2135
	scrub_block_get(sblock); /* one for the page added to the bio */
2136 2137
	atomic_inc(&sblock->outstanding_pages);
	sbio->page_count++;
2138
	if (sbio->page_count == sctx->pages_per_rd_bio)
2139
		scrub_submit(sctx);
2140 2141 2142 2143

	return 0;
}

2144
static void scrub_missing_raid56_end_io(struct bio *bio)
2145 2146
{
	struct scrub_block *sblock = bio->bi_private;
2147
	struct btrfs_fs_info *fs_info = sblock->sctx->fs_info;
2148

2149
	if (bio->bi_status)
2150 2151
		sblock->no_io_error_seen = 0;

2152 2153
	bio_put(bio);

2154 2155 2156 2157 2158 2159 2160
	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;
2161
	struct btrfs_fs_info *fs_info = sctx->fs_info;
2162 2163 2164 2165 2166 2167
	u64 logical;
	struct btrfs_device *dev;

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

2168
	if (sblock->no_io_error_seen)
2169
		scrub_recheck_block_checksum(sblock);
2170 2171 2172 2173 2174

	if (!sblock->no_io_error_seen) {
		spin_lock(&sctx->stat_lock);
		sctx->stat.read_errors++;
		spin_unlock(&sctx->stat_lock);
2175
		btrfs_err_rl_in_rcu(fs_info,
2176
			"IO error rebuilding logical %llu for dev %s",
2177 2178 2179 2180 2181
			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);
2182
		btrfs_err_rl_in_rcu(fs_info,
2183
			"failed to rebuild valid logical %llu for dev %s",
2184 2185 2186 2187 2188
			logical, rcu_str_deref(dev->name));
	} else {
		scrub_write_block_to_dev_replace(sblock);
	}

2189
	if (sctx->is_dev_replace && sctx->flush_all_writes) {
2190
		mutex_lock(&sctx->wr_lock);
2191
		scrub_wr_submit(sctx);
2192
		mutex_unlock(&sctx->wr_lock);
2193 2194
	}

2195
	scrub_block_put(sblock);
2196 2197 2198 2199 2200 2201
	scrub_pending_bio_dec(sctx);
}

static void scrub_missing_raid56_pages(struct scrub_block *sblock)
{
	struct scrub_ctx *sctx = sblock->sctx;
2202
	struct btrfs_fs_info *fs_info = sctx->fs_info;
2203 2204
	u64 length = sblock->page_count * PAGE_SIZE;
	u64 logical = sblock->pagev[0]->logical;
2205
	struct btrfs_io_context *bioc = NULL;
2206 2207 2208 2209 2210
	struct bio *bio;
	struct btrfs_raid_bio *rbio;
	int ret;
	int i;

2211
	btrfs_bio_counter_inc_blocked(fs_info);
2212
	ret = btrfs_map_sblock(fs_info, BTRFS_MAP_GET_READ_MIRRORS, logical,
2213 2214 2215
			       &length, &bioc);
	if (ret || !bioc || !bioc->raid_map)
		goto bioc_out;
2216 2217

	if (WARN_ON(!sctx->is_dev_replace ||
2218
		    !(bioc->map_type & BTRFS_BLOCK_GROUP_RAID56_MASK))) {
2219 2220 2221 2222 2223 2224
		/*
		 * 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().
		 */
2225
		goto bioc_out;
2226 2227
	}

2228
	bio = btrfs_bio_alloc(BIO_MAX_VECS);
2229 2230 2231 2232
	bio->bi_iter.bi_sector = logical >> 9;
	bio->bi_private = sblock;
	bio->bi_end_io = scrub_missing_raid56_end_io;

2233
	rbio = raid56_alloc_missing_rbio(bio, bioc, length);
2234 2235 2236 2237 2238 2239 2240 2241 2242
	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);
	}

2243
	btrfs_init_work(&sblock->work, scrub_missing_raid56_worker, NULL, NULL);
2244 2245 2246 2247 2248 2249 2250
	scrub_block_get(sblock);
	scrub_pending_bio_inc(sctx);
	raid56_submit_missing_rbio(rbio);
	return;

rbio_out:
	bio_put(bio);
2251
bioc_out:
2252
	btrfs_bio_counter_dec(fs_info);
2253
	btrfs_put_bioc(bioc);
2254 2255 2256 2257 2258
	spin_lock(&sctx->stat_lock);
	sctx->stat.malloc_errors++;
	spin_unlock(&sctx->stat_lock);
}

2259
static int scrub_pages(struct scrub_ctx *sctx, u64 logical, u32 len,
2260
		       u64 physical, struct btrfs_device *dev, u64 flags,
2261
		       u64 gen, int mirror_num, u8 *csum,
2262
		       u64 physical_for_dev_replace)
2263 2264
{
	struct scrub_block *sblock;
2265
	const u32 sectorsize = sctx->fs_info->sectorsize;
2266 2267
	int index;

2268
	sblock = kzalloc(sizeof(*sblock), GFP_KERNEL);
2269
	if (!sblock) {
2270 2271 2272
		spin_lock(&sctx->stat_lock);
		sctx->stat.malloc_errors++;
		spin_unlock(&sctx->stat_lock);
2273
		return -ENOMEM;
A
Arne Jansen 已提交
2274
	}
2275

2276 2277
	/* one ref inside this function, plus one for each page added to
	 * a bio later on */
2278
	refcount_set(&sblock->refs, 1);
2279
	sblock->sctx = sctx;
2280 2281 2282
	sblock->no_io_error_seen = 1;

	for (index = 0; len > 0; index++) {
2283
		struct scrub_page *spage;
2284 2285 2286 2287 2288 2289
		/*
		 * 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);
2290

2291
		spage = kzalloc(sizeof(*spage), GFP_KERNEL);
2292 2293
		if (!spage) {
leave_nomem:
2294 2295 2296
			spin_lock(&sctx->stat_lock);
			sctx->stat.malloc_errors++;
			spin_unlock(&sctx->stat_lock);
2297
			scrub_block_put(sblock);
2298 2299
			return -ENOMEM;
		}
2300 2301 2302
		BUG_ON(index >= SCRUB_MAX_PAGES_PER_BLOCK);
		scrub_page_get(spage);
		sblock->pagev[index] = spage;
2303
		spage->sblock = sblock;
2304
		spage->dev = dev;
2305 2306 2307 2308
		spage->flags = flags;
		spage->generation = gen;
		spage->logical = logical;
		spage->physical = physical;
2309
		spage->physical_for_dev_replace = physical_for_dev_replace;
2310 2311 2312
		spage->mirror_num = mirror_num;
		if (csum) {
			spage->have_csum = 1;
2313
			memcpy(spage->csum, csum, sctx->fs_info->csum_size);
2314 2315 2316 2317
		} else {
			spage->have_csum = 0;
		}
		sblock->page_count++;
2318
		spage->page = alloc_page(GFP_KERNEL);
2319 2320
		if (!spage->page)
			goto leave_nomem;
2321 2322 2323
		len -= l;
		logical += l;
		physical += l;
2324
		physical_for_dev_replace += l;
2325 2326
	}

2327
	WARN_ON(sblock->page_count == 0);
2328
	if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) {
2329 2330 2331 2332 2333 2334 2335 2336 2337
		/*
		 * 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;
2338

2339 2340 2341 2342 2343
			ret = scrub_add_page_to_rd_bio(sctx, spage);
			if (ret) {
				scrub_block_put(sblock);
				return ret;
			}
2344
		}
A
Arne Jansen 已提交
2345

2346
		if (flags & BTRFS_EXTENT_FLAG_SUPER)
2347 2348
			scrub_submit(sctx);
	}
A
Arne Jansen 已提交
2349

2350 2351
	/* last one frees, either here or in bio completion for last page */
	scrub_block_put(sblock);
A
Arne Jansen 已提交
2352 2353 2354
	return 0;
}

2355
static void scrub_bio_end_io(struct bio *bio)
2356 2357
{
	struct scrub_bio *sbio = bio->bi_private;
2358
	struct btrfs_fs_info *fs_info = sbio->dev->fs_info;
2359

2360
	sbio->status = bio->bi_status;
2361 2362
	sbio->bio = bio;

2363
	btrfs_queue_work(fs_info->scrub_workers, &sbio->work);
2364 2365 2366 2367 2368
}

static void scrub_bio_end_io_worker(struct btrfs_work *work)
{
	struct scrub_bio *sbio = container_of(work, struct scrub_bio, work);
2369
	struct scrub_ctx *sctx = sbio->sctx;
2370 2371
	int i;

2372
	BUG_ON(sbio->page_count > SCRUB_PAGES_PER_RD_BIO);
2373
	if (sbio->status) {
2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393
		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;
2394 2395 2396 2397
	spin_lock(&sctx->list_lock);
	sbio->next_free = sctx->first_free;
	sctx->first_free = sbio->index;
	spin_unlock(&sctx->list_lock);
2398

2399
	if (sctx->is_dev_replace && sctx->flush_all_writes) {
2400
		mutex_lock(&sctx->wr_lock);
2401
		scrub_wr_submit(sctx);
2402
		mutex_unlock(&sctx->wr_lock);
2403 2404
	}

2405
	scrub_pending_bio_dec(sctx);
2406 2407
}

2408 2409
static inline void __scrub_mark_bitmap(struct scrub_parity *sparity,
				       unsigned long *bitmap,
2410
				       u64 start, u32 len)
2411
{
2412
	u64 offset;
2413
	u32 nsectors;
2414
	u32 sectorsize_bits = sparity->sctx->fs_info->sectorsize_bits;
2415 2416 2417 2418 2419 2420 2421

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

	start -= sparity->logic_start;
2422
	start = div64_u64_rem(start, sparity->stripe_len, &offset);
2423
	offset = offset >> sectorsize_bits;
2424
	nsectors = len >> sectorsize_bits;
2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435

	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,
2436
						   u64 start, u32 len)
2437 2438 2439 2440 2441
{
	__scrub_mark_bitmap(sparity, sparity->ebitmap, start, len);
}

static inline void scrub_parity_mark_sectors_data(struct scrub_parity *sparity,
2442
						  u64 start, u32 len)
2443 2444 2445 2446
{
	__scrub_mark_bitmap(sparity, sparity->dbitmap, start, len);
}

2447 2448
static void scrub_block_complete(struct scrub_block *sblock)
{
2449 2450
	int corrupted = 0;

2451
	if (!sblock->no_io_error_seen) {
2452
		corrupted = 1;
2453
		scrub_handle_errored_block(sblock);
2454 2455 2456 2457 2458 2459
	} else {
		/*
		 * if has checksum error, write via repair mechanism in
		 * dev replace case, otherwise write here in dev replace
		 * case.
		 */
2460 2461
		corrupted = scrub_checksum(sblock);
		if (!corrupted && sblock->sctx->is_dev_replace)
2462 2463
			scrub_write_block_to_dev_replace(sblock);
	}
2464 2465 2466 2467

	if (sblock->sparity && corrupted && !sblock->data_corrected) {
		u64 start = sblock->pagev[0]->logical;
		u64 end = sblock->pagev[sblock->page_count - 1]->logical +
2468
			  sblock->sctx->fs_info->sectorsize;
2469

2470
		ASSERT(end - start <= U32_MAX);
2471 2472 2473
		scrub_parity_mark_sectors_error(sblock->sparity,
						start, end - start);
	}
2474 2475
}

2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487
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
2488
 * storing bytenr ordered csum ranges.  We're responsible to cleanup any range
2489 2490 2491 2492 2493
 * 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.
 */
2494
static int scrub_find_csum(struct scrub_ctx *sctx, u64 logical, u8 *csum)
A
Arne Jansen 已提交
2495
{
2496
	bool found = false;
A
Arne Jansen 已提交
2497

2498
	while (!list_empty(&sctx->csum_list)) {
2499 2500 2501 2502
		struct btrfs_ordered_sum *sum = NULL;
		unsigned long index;
		unsigned long num_sectors;

2503
		sum = list_first_entry(&sctx->csum_list,
A
Arne Jansen 已提交
2504
				       struct btrfs_ordered_sum, list);
2505
		/* The current csum range is beyond our range, no csum found */
A
Arne Jansen 已提交
2506 2507 2508
		if (sum->bytenr > logical)
			break;

2509 2510 2511 2512 2513 2514 2515 2516 2517 2518
		/*
		 * 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
2519

2520 2521 2522 2523
		/* 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;
2524

2525 2526 2527 2528 2529 2530 2531
		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
2532
	}
2533 2534
	if (!found)
		return 0;
2535
	return 1;
A
Arne Jansen 已提交
2536 2537 2538
}

/* scrub extent tries to collect up to 64 kB for each bio */
L
Liu Bo 已提交
2539
static int scrub_extent(struct scrub_ctx *sctx, struct map_lookup *map,
2540
			u64 logical, u32 len,
2541
			u64 physical, struct btrfs_device *dev, u64 flags,
2542
			u64 gen, int mirror_num, u64 physical_for_dev_replace)
A
Arne Jansen 已提交
2543 2544 2545
{
	int ret;
	u8 csum[BTRFS_CSUM_SIZE];
2546 2547 2548
	u32 blocksize;

	if (flags & BTRFS_EXTENT_FLAG_DATA) {
L
Liu Bo 已提交
2549 2550 2551 2552
		if (map->type & BTRFS_BLOCK_GROUP_RAID56_MASK)
			blocksize = map->stripe_len;
		else
			blocksize = sctx->fs_info->sectorsize;
2553 2554 2555 2556
		spin_lock(&sctx->stat_lock);
		sctx->stat.data_extents_scrubbed++;
		sctx->stat.data_bytes_scrubbed += len;
		spin_unlock(&sctx->stat_lock);
2557
	} else if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
L
Liu Bo 已提交
2558 2559 2560 2561
		if (map->type & BTRFS_BLOCK_GROUP_RAID56_MASK)
			blocksize = map->stripe_len;
		else
			blocksize = sctx->fs_info->nodesize;
2562 2563 2564 2565
		spin_lock(&sctx->stat_lock);
		sctx->stat.tree_extents_scrubbed++;
		sctx->stat.tree_bytes_scrubbed += len;
		spin_unlock(&sctx->stat_lock);
2566
	} else {
2567
		blocksize = sctx->fs_info->sectorsize;
2568
		WARN_ON(1);
2569
	}
A
Arne Jansen 已提交
2570 2571

	while (len) {
2572
		u32 l = min(len, blocksize);
A
Arne Jansen 已提交
2573 2574 2575 2576
		int have_csum = 0;

		if (flags & BTRFS_EXTENT_FLAG_DATA) {
			/* push csums to sbio */
2577
			have_csum = scrub_find_csum(sctx, logical, csum);
A
Arne Jansen 已提交
2578
			if (have_csum == 0)
2579
				++sctx->stat.no_csum;
A
Arne Jansen 已提交
2580
		}
2581
		ret = scrub_pages(sctx, logical, l, physical, dev, flags, gen,
2582
				  mirror_num, have_csum ? csum : NULL,
2583
				  physical_for_dev_replace);
A
Arne Jansen 已提交
2584 2585 2586 2587 2588
		if (ret)
			return ret;
		len -= l;
		logical += l;
		physical += l;
2589
		physical_for_dev_replace += l;
A
Arne Jansen 已提交
2590 2591 2592 2593
	}
	return 0;
}

2594
static int scrub_pages_for_parity(struct scrub_parity *sparity,
2595
				  u64 logical, u32 len,
2596 2597 2598 2599 2600
				  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;
2601
	const u32 sectorsize = sctx->fs_info->sectorsize;
2602 2603
	int index;

2604 2605
	ASSERT(IS_ALIGNED(len, sectorsize));

2606
	sblock = kzalloc(sizeof(*sblock), GFP_KERNEL);
2607 2608 2609 2610 2611 2612 2613 2614 2615
	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 */
2616
	refcount_set(&sblock->refs, 1);
2617 2618 2619 2620 2621 2622 2623 2624
	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;

2625
		spage = kzalloc(sizeof(*spage), GFP_KERNEL);
2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649
		if (!spage) {
leave_nomem:
			spin_lock(&sctx->stat_lock);
			sctx->stat.malloc_errors++;
			spin_unlock(&sctx->stat_lock);
			scrub_block_put(sblock);
			return -ENOMEM;
		}
		BUG_ON(index >= SCRUB_MAX_PAGES_PER_BLOCK);
		/* 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;
2650
			memcpy(spage->csum, csum, sctx->fs_info->csum_size);
2651 2652 2653 2654
		} else {
			spage->have_csum = 0;
		}
		sblock->page_count++;
2655
		spage->page = alloc_page(GFP_KERNEL);
2656 2657
		if (!spage->page)
			goto leave_nomem;
2658 2659 2660 2661 2662 2663


		/* Iterate over the stripe range in sectorsize steps */
		len -= sectorsize;
		logical += sectorsize;
		physical += sectorsize;
2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683
	}

	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,
2684
				   u64 logical, u32 len,
2685 2686 2687 2688 2689 2690 2691 2692
				   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;

2693
	if (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state)) {
2694 2695 2696 2697
		scrub_parity_mark_sectors_error(sparity, logical, len);
		return 0;
	}

2698
	if (flags & BTRFS_EXTENT_FLAG_DATA) {
L
Liu Bo 已提交
2699
		blocksize = sparity->stripe_len;
2700
	} else if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
L
Liu Bo 已提交
2701
		blocksize = sparity->stripe_len;
2702
	} else {
2703
		blocksize = sctx->fs_info->sectorsize;
2704 2705 2706 2707
		WARN_ON(1);
	}

	while (len) {
2708
		u32 l = min(len, blocksize);
2709 2710 2711 2712
		int have_csum = 0;

		if (flags & BTRFS_EXTENT_FLAG_DATA) {
			/* push csums to sbio */
2713
			have_csum = scrub_find_csum(sctx, logical, csum);
2714 2715 2716 2717 2718 2719 2720 2721
			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;
2722
skip:
2723 2724 2725 2726 2727 2728 2729
		len -= l;
		logical += l;
		physical += l;
	}
	return 0;
}

2730 2731 2732 2733 2734 2735 2736 2737
/*
 * 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,
2738 2739
				   struct map_lookup *map, u64 *offset,
				   u64 *stripe_start)
2740 2741 2742 2743 2744
{
	int i;
	int j = 0;
	u64 stripe_nr;
	u64 last_offset;
2745 2746
	u32 stripe_index;
	u32 rot;
2747
	const int data_stripes = nr_data_stripes(map);
2748

2749
	last_offset = (physical - map->stripes[num].physical) * data_stripes;
2750 2751 2752
	if (stripe_start)
		*stripe_start = last_offset;

2753
	*offset = last_offset;
2754
	for (i = 0; i < data_stripes; i++) {
2755 2756
		*offset = last_offset + i * map->stripe_len;

2757
		stripe_nr = div64_u64(*offset, map->stripe_len);
2758
		stripe_nr = div_u64(stripe_nr, data_stripes);
2759 2760

		/* Work out the disk rotation on this stripe-set */
2761
		stripe_nr = div_u64_rem(stripe_nr, map->num_stripes, &rot);
2762 2763
		/* calculate which stripe this data locates */
		rot += i;
2764
		stripe_index = rot % map->num_stripes;
2765 2766 2767 2768 2769 2770 2771 2772 2773
		if (stripe_index == num)
			return 0;
		if (stripe_index < num)
			j++;
	}
	*offset = last_offset + j * map->stripe_len;
	return 1;
}

2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795
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);
}

2796 2797 2798 2799 2800 2801 2802 2803 2804 2805
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);
}

2806
static void scrub_parity_bio_endio(struct bio *bio)
2807 2808
{
	struct scrub_parity *sparity = (struct scrub_parity *)bio->bi_private;
2809
	struct btrfs_fs_info *fs_info = sparity->sctx->fs_info;
2810

2811
	if (bio->bi_status)
2812 2813 2814 2815
		bitmap_or(sparity->ebitmap, sparity->ebitmap, sparity->dbitmap,
			  sparity->nsectors);

	bio_put(bio);
2816

2817 2818
	btrfs_init_work(&sparity->work, scrub_parity_bio_endio_worker, NULL,
			NULL);
2819
	btrfs_queue_work(fs_info->scrub_parity_workers, &sparity->work);
2820 2821 2822 2823 2824
}

static void scrub_parity_check_and_repair(struct scrub_parity *sparity)
{
	struct scrub_ctx *sctx = sparity->sctx;
2825
	struct btrfs_fs_info *fs_info = sctx->fs_info;
2826 2827
	struct bio *bio;
	struct btrfs_raid_bio *rbio;
2828
	struct btrfs_io_context *bioc = NULL;
2829 2830 2831 2832 2833 2834 2835
	u64 length;
	int ret;

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

2836
	length = sparity->logic_end - sparity->logic_start;
2837 2838

	btrfs_bio_counter_inc_blocked(fs_info);
2839
	ret = btrfs_map_sblock(fs_info, BTRFS_MAP_WRITE, sparity->logic_start,
2840 2841 2842
			       &length, &bioc);
	if (ret || !bioc || !bioc->raid_map)
		goto bioc_out;
2843

2844
	bio = btrfs_bio_alloc(BIO_MAX_VECS);
2845 2846 2847 2848
	bio->bi_iter.bi_sector = sparity->logic_start >> 9;
	bio->bi_private = sparity;
	bio->bi_end_io = scrub_parity_bio_endio;

2849 2850
	rbio = raid56_parity_alloc_scrub_rbio(bio, bioc, length,
					      sparity->scrub_dev,
2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861
					      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);
2862
bioc_out:
2863
	btrfs_bio_counter_dec(fs_info);
2864
	btrfs_put_bioc(bioc);
2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875
	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)
{
2876
	return DIV_ROUND_UP(nsectors, BITS_PER_LONG) * sizeof(long);
2877 2878 2879 2880
}

static void scrub_parity_get(struct scrub_parity *sparity)
{
2881
	refcount_inc(&sparity->refs);
2882 2883 2884 2885
}

static void scrub_parity_put(struct scrub_parity *sparity)
{
2886
	if (!refcount_dec_and_test(&sparity->refs))
2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898
		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)
{
2899
	struct btrfs_fs_info *fs_info = sctx->fs_info;
2900
	struct btrfs_root *root = btrfs_extent_root(fs_info, logic_start);
2901 2902
	struct btrfs_root *csum_root = fs_info->csum_root;
	struct btrfs_extent_item *extent;
2903
	struct btrfs_io_context *bioc = NULL;
2904 2905 2906 2907 2908 2909 2910 2911
	u64 flags;
	int ret;
	int slot;
	struct extent_buffer *l;
	struct btrfs_key key;
	u64 generation;
	u64 extent_logical;
	u64 extent_physical;
2912 2913
	/* Check the comment in scrub_stripe() for why u32 is enough here */
	u32 extent_len;
2914
	u64 mapped_length;
2915 2916 2917 2918 2919 2920 2921
	struct btrfs_device *extent_dev;
	struct scrub_parity *sparity;
	int nsectors;
	int bitmap_len;
	int extent_mirror_num;
	int stop_loop = 0;

2922
	ASSERT(map->stripe_len <= U32_MAX);
2923
	nsectors = map->stripe_len >> fs_info->sectorsize_bits;
2924 2925 2926 2927 2928 2929 2930 2931 2932 2933
	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;
	}

2934
	ASSERT(map->stripe_len <= U32_MAX);
2935 2936 2937 2938 2939 2940
	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;
2941
	refcount_set(&sparity->refs, 1);
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 2983 2984 2985 2986 2987 2988 2989
	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);

2990 2991 2992 2993
			if (key.type != BTRFS_EXTENT_ITEM_KEY &&
			    key.type != BTRFS_METADATA_ITEM_KEY)
				goto next;

2994
			if (key.type == BTRFS_METADATA_ITEM_KEY)
2995
				bytes = fs_info->nodesize;
2996 2997 2998 2999 3000 3001
			else
				bytes = key.offset;

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

3002
			if (key.objectid >= logic_end) {
3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014
				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);

3015 3016 3017 3018
			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
3019 3020
				btrfs_err(fs_info,
					  "scrub: tree block %llu spanning stripes, ignored. logical=%llu",
3021
					  key.objectid, logic_start);
3022 3023 3024
				spin_lock(&sctx->stat_lock);
				sctx->stat.uncorrectable_errors++;
				spin_unlock(&sctx->stat_lock);
3025 3026 3027 3028
				goto next;
			}
again:
			extent_logical = key.objectid;
3029
			ASSERT(bytes <= U32_MAX);
3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044
			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);

3045
			mapped_length = extent_len;
3046
			bioc = NULL;
3047
			ret = btrfs_map_block(fs_info, BTRFS_MAP_READ,
3048
					extent_logical, &mapped_length, &bioc,
3049
					0);
3050
			if (!ret) {
3051
				if (!bioc || mapped_length < extent_len)
3052 3053 3054
					ret = -EIO;
			}
			if (ret) {
3055
				btrfs_put_bioc(bioc);
3056 3057
				goto out;
			}
3058 3059 3060 3061
			extent_physical = bioc->stripes[0].physical;
			extent_mirror_num = bioc->mirror_num;
			extent_dev = bioc->stripes[0].dev;
			btrfs_put_bioc(bioc);
3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075

			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);
3076 3077 3078

			scrub_free_csums(sctx);

3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107
			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:
3108 3109
	if (ret < 0) {
		ASSERT(logic_end - logic_start <= U32_MAX);
3110
		scrub_parity_mark_sectors_error(sparity, logic_start,
3111
						logic_end - logic_start);
3112
	}
3113 3114
	scrub_parity_put(sparity);
	scrub_submit(sctx);
3115
	mutex_lock(&sctx->wr_lock);
3116
	scrub_wr_submit(sctx);
3117
	mutex_unlock(&sctx->wr_lock);
3118 3119 3120 3121 3122

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

3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136
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);
}

3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162
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;
}

3163
static noinline_for_stack int scrub_stripe(struct scrub_ctx *sctx,
3164 3165
					   struct map_lookup *map,
					   struct btrfs_device *scrub_dev,
3166 3167
					   int num, u64 base, u64 length,
					   struct btrfs_block_group *cache)
A
Arne Jansen 已提交
3168
{
3169
	struct btrfs_path *path, *ppath;
3170
	struct btrfs_fs_info *fs_info = sctx->fs_info;
3171
	struct btrfs_root *root;
A
Arne Jansen 已提交
3172 3173
	struct btrfs_root *csum_root = fs_info->csum_root;
	struct btrfs_extent_item *extent;
3174
	struct blk_plug plug;
A
Arne Jansen 已提交
3175 3176 3177 3178 3179 3180 3181
	u64 flags;
	int ret;
	int slot;
	u64 nstripes;
	struct extent_buffer *l;
	u64 physical;
	u64 logical;
L
Liu Bo 已提交
3182
	u64 logic_end;
3183
	u64 physical_end;
A
Arne Jansen 已提交
3184
	u64 generation;
3185
	int mirror_num;
A
Arne Jansen 已提交
3186 3187
	struct reada_control *reada1;
	struct reada_control *reada2;
3188
	struct btrfs_key key;
A
Arne Jansen 已提交
3189
	struct btrfs_key key_end;
A
Arne Jansen 已提交
3190 3191
	u64 increment = map->stripe_len;
	u64 offset;
3192 3193
	u64 extent_logical;
	u64 extent_physical;
3194 3195 3196 3197 3198
	/*
	 * Unlike chunk length, extent length should never go beyond
	 * BTRFS_MAX_EXTENT_SIZE, thus u32 is enough here.
	 */
	u32 extent_len;
3199 3200
	u64 stripe_logical;
	u64 stripe_end;
3201 3202
	struct btrfs_device *extent_dev;
	int extent_mirror_num;
3203
	int stop_loop = 0;
D
David Woodhouse 已提交
3204

3205
	physical = map->stripes[num].physical;
A
Arne Jansen 已提交
3206
	offset = 0;
3207
	nstripes = div64_u64(length, map->stripe_len);
3208 3209
	mirror_num = 1;
	increment = map->stripe_len;
A
Arne Jansen 已提交
3210 3211 3212 3213 3214 3215 3216
	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;
3217
		mirror_num = num % map->sub_stripes + 1;
3218
	} else if (map->type & BTRFS_BLOCK_GROUP_RAID1_MASK) {
3219
		mirror_num = num % map->num_stripes + 1;
A
Arne Jansen 已提交
3220
	} else if (map->type & BTRFS_BLOCK_GROUP_DUP) {
3221
		mirror_num = num % map->num_stripes + 1;
3222
	} else if (map->type & BTRFS_BLOCK_GROUP_RAID56_MASK) {
3223
		get_raid56_logic_offset(physical, num, map, &offset, NULL);
3224
		increment = map->stripe_len * nr_data_stripes(map);
A
Arne Jansen 已提交
3225 3226 3227 3228 3229 3230
	}

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

3231 3232
	ppath = btrfs_alloc_path();
	if (!ppath) {
3233
		btrfs_free_path(path);
3234 3235 3236
		return -ENOMEM;
	}

3237 3238 3239 3240 3241
	/*
	 * 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
3242 3243 3244
	path->search_commit_root = 1;
	path->skip_locking = 1;

3245 3246
	ppath->search_commit_root = 1;
	ppath->skip_locking = 1;
A
Arne Jansen 已提交
3247
	/*
A
Arne Jansen 已提交
3248 3249 3250
	 * 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
3251 3252
	 */
	logical = base + offset;
3253
	physical_end = physical + nstripes * map->stripe_len;
3254
	if (map->type & BTRFS_BLOCK_GROUP_RAID56_MASK) {
3255
		get_raid56_logic_offset(physical_end, num,
3256
					map, &logic_end, NULL);
3257 3258 3259 3260
		logic_end += base;
	} else {
		logic_end = logical + increment * nstripes;
	}
3261
	wait_event(sctx->list_wait,
3262
		   atomic_read(&sctx->bios_in_flight) == 0);
3263
	scrub_blocked_if_needed(fs_info);
A
Arne Jansen 已提交
3264

3265 3266
	root = btrfs_extent_root(fs_info, logical);

A
Arne Jansen 已提交
3267
	/* FIXME it might be better to start readahead at commit root */
3268 3269 3270
	key.objectid = logical;
	key.type = BTRFS_EXTENT_ITEM_KEY;
	key.offset = (u64)0;
3271
	key_end.objectid = logic_end;
3272 3273
	key_end.type = BTRFS_METADATA_ITEM_KEY;
	key_end.offset = (u64)-1;
3274
	reada1 = btrfs_reada_add(root, &key, &key_end);
A
Arne Jansen 已提交
3275

3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286
	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
3287 3288 3289

	if (!IS_ERR(reada1))
		btrfs_reada_wait(reada1);
3290
	if (!IS_ERR_OR_NULL(reada2))
A
Arne Jansen 已提交
3291 3292
		btrfs_reada_wait(reada2);

A
Arne Jansen 已提交
3293 3294 3295 3296 3297

	/*
	 * collect all data csums for the stripe to avoid seeking during
	 * the scrub. This might currently (crc32) end up to be about 1MB
	 */
3298
	blk_start_plug(&plug);
A
Arne Jansen 已提交
3299

3300 3301 3302 3303 3304 3305 3306 3307
	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
3308 3309 3310 3311
	/*
	 * now find all extents for each stripe and scrub them
	 */
	ret = 0;
3312
	while (physical < physical_end) {
A
Arne Jansen 已提交
3313 3314 3315 3316
		/*
		 * canceled?
		 */
		if (atomic_read(&fs_info->scrub_cancel_req) ||
3317
		    atomic_read(&sctx->cancel_req)) {
A
Arne Jansen 已提交
3318 3319 3320 3321 3322 3323 3324 3325
			ret = -ECANCELED;
			goto out;
		}
		/*
		 * check to see if we have to pause
		 */
		if (atomic_read(&fs_info->scrub_pause_req)) {
			/* push queued extents */
3326
			sctx->flush_all_writes = true;
3327
			scrub_submit(sctx);
3328
			mutex_lock(&sctx->wr_lock);
3329
			scrub_wr_submit(sctx);
3330
			mutex_unlock(&sctx->wr_lock);
3331
			wait_event(sctx->list_wait,
3332
				   atomic_read(&sctx->bios_in_flight) == 0);
3333
			sctx->flush_all_writes = false;
3334
			scrub_blocked_if_needed(fs_info);
A
Arne Jansen 已提交
3335 3336
		}

3337 3338 3339 3340 3341 3342
		if (map->type & BTRFS_BLOCK_GROUP_RAID56_MASK) {
			ret = get_raid56_logic_offset(physical, num, map,
						      &logical,
						      &stripe_logical);
			logical += base;
			if (ret) {
3343
				/* it is parity strip */
3344
				stripe_logical += base;
3345
				stripe_end = stripe_logical + increment;
3346 3347 3348 3349 3350 3351 3352 3353 3354
				ret = scrub_raid56_parity(sctx, map, scrub_dev,
							  ppath, stripe_logical,
							  stripe_end);
				if (ret)
					goto out;
				goto skip;
			}
		}

3355 3356 3357 3358
		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
3359
		key.objectid = logical;
L
Liu Bo 已提交
3360
		key.offset = (u64)-1;
A
Arne Jansen 已提交
3361 3362 3363 3364

		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
		if (ret < 0)
			goto out;
3365

3366
		if (ret > 0) {
3367
			ret = btrfs_previous_extent_item(root, path, 0);
A
Arne Jansen 已提交
3368 3369
			if (ret < 0)
				goto out;
3370 3371 3372 3373 3374 3375 3376 3377 3378
			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
3379 3380
		}

L
Liu Bo 已提交
3381
		stop_loop = 0;
A
Arne Jansen 已提交
3382
		while (1) {
3383 3384
			u64 bytes;

A
Arne Jansen 已提交
3385 3386 3387 3388 3389 3390 3391 3392 3393
			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
3394
				stop_loop = 1;
A
Arne Jansen 已提交
3395 3396 3397 3398
				break;
			}
			btrfs_item_key_to_cpu(l, &key, slot);

3399 3400 3401 3402
			if (key.type != BTRFS_EXTENT_ITEM_KEY &&
			    key.type != BTRFS_METADATA_ITEM_KEY)
				goto next;

3403
			if (key.type == BTRFS_METADATA_ITEM_KEY)
3404
				bytes = fs_info->nodesize;
3405 3406 3407 3408
			else
				bytes = key.offset;

			if (key.objectid + bytes <= logical)
A
Arne Jansen 已提交
3409 3410
				goto next;

L
Liu Bo 已提交
3411 3412 3413 3414 3415 3416
			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
3417

3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431
			/*
			 * 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
3432 3433 3434 3435 3436
			extent = btrfs_item_ptr(l, slot,
						struct btrfs_extent_item);
			flags = btrfs_extent_flags(l, extent);
			generation = btrfs_extent_generation(l, extent);

3437 3438 3439 3440
			if ((flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) &&
			    (key.objectid < logical ||
			     key.objectid + bytes >
			     logical + map->stripe_len)) {
3441
				btrfs_err(fs_info,
J
Jeff Mahoney 已提交
3442
					   "scrub: tree block %llu spanning stripes, ignored. logical=%llu",
3443
				       key.objectid, logical);
3444 3445 3446
				spin_lock(&sctx->stat_lock);
				sctx->stat.uncorrectable_errors++;
				spin_unlock(&sctx->stat_lock);
A
Arne Jansen 已提交
3447 3448 3449
				goto next;
			}

L
Liu Bo 已提交
3450 3451
again:
			extent_logical = key.objectid;
3452
			ASSERT(bytes <= U32_MAX);
L
Liu Bo 已提交
3453 3454
			extent_len = bytes;

A
Arne Jansen 已提交
3455 3456 3457
			/*
			 * trim extent to this stripe
			 */
L
Liu Bo 已提交
3458 3459 3460
			if (extent_logical < logical) {
				extent_len -= logical - extent_logical;
				extent_logical = logical;
A
Arne Jansen 已提交
3461
			}
L
Liu Bo 已提交
3462
			if (extent_logical + extent_len >
A
Arne Jansen 已提交
3463
			    logical + map->stripe_len) {
L
Liu Bo 已提交
3464 3465
				extent_len = logical + map->stripe_len -
					     extent_logical;
A
Arne Jansen 已提交
3466 3467
			}

L
Liu Bo 已提交
3468
			extent_physical = extent_logical - logical + physical;
3469 3470
			extent_dev = scrub_dev;
			extent_mirror_num = mirror_num;
3471
			if (sctx->is_dev_replace)
3472 3473 3474 3475
				scrub_remap_extent(fs_info, extent_logical,
						   extent_len, &extent_physical,
						   &extent_dev,
						   &extent_mirror_num);
L
Liu Bo 已提交
3476

3477 3478 3479 3480 3481 3482 3483 3484
			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
3485

L
Liu Bo 已提交
3486
			ret = scrub_extent(sctx, map, extent_logical, extent_len,
3487 3488
					   extent_physical, extent_dev, flags,
					   generation, extent_mirror_num,
3489
					   extent_logical - logical + physical);
3490 3491 3492

			scrub_free_csums(sctx);

A
Arne Jansen 已提交
3493 3494 3495
			if (ret)
				goto out;

3496 3497 3498
			if (sctx->is_dev_replace)
				sync_replace_for_zoned(sctx);

L
Liu Bo 已提交
3499 3500
			if (extent_logical + extent_len <
			    key.objectid + bytes) {
3501
				if (map->type & BTRFS_BLOCK_GROUP_RAID56_MASK) {
3502 3503 3504 3505
					/*
					 * loop until we find next data stripe
					 * or we have finished all stripes.
					 */
3506 3507 3508 3509 3510 3511 3512 3513 3514 3515
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 +
3516
								increment;
3517 3518 3519 3520 3521 3522 3523 3524
						ret = scrub_raid56_parity(sctx,
							map, scrub_dev, ppath,
							stripe_logical,
							stripe_end);
						if (ret)
							goto out;
						goto loop;
					}
3525 3526 3527 3528
				} else {
					physical += map->stripe_len;
					logical += increment;
				}
L
Liu Bo 已提交
3529 3530 3531 3532 3533
				if (logical < key.objectid + bytes) {
					cond_resched();
					goto again;
				}

3534
				if (physical >= physical_end) {
L
Liu Bo 已提交
3535 3536 3537 3538
					stop_loop = 1;
					break;
				}
			}
A
Arne Jansen 已提交
3539 3540 3541
next:
			path->slots[0]++;
		}
C
Chris Mason 已提交
3542
		btrfs_release_path(path);
3543
skip:
A
Arne Jansen 已提交
3544 3545
		logical += increment;
		physical += map->stripe_len;
3546
		spin_lock(&sctx->stat_lock);
L
Liu Bo 已提交
3547 3548 3549 3550 3551
		if (stop_loop)
			sctx->stat.last_physical = map->stripes[num].physical +
						   length;
		else
			sctx->stat.last_physical = physical;
3552
		spin_unlock(&sctx->stat_lock);
L
Liu Bo 已提交
3553 3554
		if (stop_loop)
			break;
A
Arne Jansen 已提交
3555
	}
3556
out:
A
Arne Jansen 已提交
3557
	/* push queued extents */
3558
	scrub_submit(sctx);
3559
	mutex_lock(&sctx->wr_lock);
3560
	scrub_wr_submit(sctx);
3561
	mutex_unlock(&sctx->wr_lock);
A
Arne Jansen 已提交
3562

3563
	blk_finish_plug(&plug);
A
Arne Jansen 已提交
3564
	btrfs_free_path(path);
3565
	btrfs_free_path(ppath);
3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576

	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
3577 3578 3579
	return ret < 0 ? ret : 0;
}

3580
static noinline_for_stack int scrub_chunk(struct scrub_ctx *sctx,
3581 3582
					  struct btrfs_device *scrub_dev,
					  u64 chunk_offset, u64 length,
3583
					  u64 dev_offset,
3584
					  struct btrfs_block_group *cache)
A
Arne Jansen 已提交
3585
{
3586
	struct btrfs_fs_info *fs_info = sctx->fs_info;
3587
	struct extent_map_tree *map_tree = &fs_info->mapping_tree;
A
Arne Jansen 已提交
3588 3589 3590
	struct map_lookup *map;
	struct extent_map *em;
	int i;
3591
	int ret = 0;
A
Arne Jansen 已提交
3592

3593 3594 3595
	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
3596

3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608
	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
3609

3610
	map = em->map_lookup;
A
Arne Jansen 已提交
3611 3612 3613 3614 3615 3616 3617
	if (em->start != chunk_offset)
		goto out;

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

	for (i = 0; i < map->num_stripes; ++i) {
3618
		if (map->stripes[i].dev->bdev == scrub_dev->bdev &&
3619
		    map->stripes[i].physical == dev_offset) {
3620
			ret = scrub_stripe(sctx, map, scrub_dev, i,
3621
					   chunk_offset, length, cache);
A
Arne Jansen 已提交
3622 3623 3624 3625 3626 3627 3628 3629 3630 3631
			if (ret)
				goto out;
		}
	}
out:
	free_extent_map(em);

	return ret;
}

3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650
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
3651
static noinline_for_stack
3652
int scrub_enumerate_chunks(struct scrub_ctx *sctx,
3653
			   struct btrfs_device *scrub_dev, u64 start, u64 end)
A
Arne Jansen 已提交
3654 3655 3656
{
	struct btrfs_dev_extent *dev_extent = NULL;
	struct btrfs_path *path;
3657 3658
	struct btrfs_fs_info *fs_info = sctx->fs_info;
	struct btrfs_root *root = fs_info->dev_root;
A
Arne Jansen 已提交
3659 3660
	u64 length;
	u64 chunk_offset;
3661
	int ret = 0;
3662
	int ro_set;
A
Arne Jansen 已提交
3663 3664 3665 3666
	int slot;
	struct extent_buffer *l;
	struct btrfs_key key;
	struct btrfs_key found_key;
3667
	struct btrfs_block_group *cache;
3668
	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
A
Arne Jansen 已提交
3669 3670 3671 3672 3673

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

3674
	path->reada = READA_FORWARD;
A
Arne Jansen 已提交
3675 3676 3677
	path->search_commit_root = 1;
	path->skip_locking = 1;

3678
	key.objectid = scrub_dev->devid;
A
Arne Jansen 已提交
3679 3680 3681 3682 3683 3684
	key.offset = 0ull;
	key.type = BTRFS_DEV_EXTENT_KEY;

	while (1) {
		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
		if (ret < 0)
3685 3686 3687 3688 3689
			break;
		if (ret > 0) {
			if (path->slots[0] >=
			    btrfs_header_nritems(path->nodes[0])) {
				ret = btrfs_next_leaf(root, path);
3690 3691 3692 3693
				if (ret < 0)
					break;
				if (ret > 0) {
					ret = 0;
3694
					break;
3695 3696 3697
				}
			} else {
				ret = 0;
3698 3699
			}
		}
A
Arne Jansen 已提交
3700 3701 3702 3703 3704 3705

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

		btrfs_item_key_to_cpu(l, &found_key, slot);

3706
		if (found_key.objectid != scrub_dev->devid)
A
Arne Jansen 已提交
3707 3708
			break;

3709
		if (found_key.type != BTRFS_DEV_EXTENT_KEY)
A
Arne Jansen 已提交
3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720
			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);

3721 3722
		if (found_key.offset + length <= start)
			goto skip;
A
Arne Jansen 已提交
3723 3724 3725 3726 3727 3728 3729 3730

		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);
3731 3732 3733 3734 3735 3736

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

3737 3738 3739 3740
		if (sctx->is_dev_replace && btrfs_is_zoned(fs_info)) {
			spin_lock(&cache->lock);
			if (!cache->to_copy) {
				spin_unlock(&cache->lock);
3741 3742
				btrfs_put_block_group(cache);
				goto skip;
3743 3744 3745 3746
			}
			spin_unlock(&cache->lock);
		}

3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760
		/*
		 * 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;
		}
3761
		btrfs_freeze_block_group(cache);
3762 3763
		spin_unlock(&cache->lock);

3764 3765 3766 3767 3768 3769 3770 3771 3772
		/*
		 * 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);
3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790

		/*
		 * 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
3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802
		 *
		 * 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.
3803
		 */
3804
		ret = btrfs_inc_block_group_ro(cache, sctx->is_dev_replace);
3805 3806 3807 3808 3809 3810 3811 3812 3813 3814
		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;
			}
		}

3815 3816
		if (ret == 0) {
			ro_set = 1;
3817
		} else if (ret == -ENOSPC && !sctx->is_dev_replace) {
3818 3819 3820
			/*
			 * btrfs_inc_block_group_ro return -ENOSPC when it
			 * failed in creating new chunk for metadata.
3821
			 * It is not a problem for scrub, because
3822 3823 3824 3825
			 * metadata are always cowed, and our scrub paused
			 * commit_transactions.
			 */
			ro_set = 0;
3826 3827 3828 3829 3830 3831 3832
		} 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;
3833
		} else {
J
Jeff Mahoney 已提交
3834
			btrfs_warn(fs_info,
3835
				   "failed setting block group ro: %d", ret);
3836
			btrfs_unfreeze_block_group(cache);
3837
			btrfs_put_block_group(cache);
3838
			scrub_pause_off(fs_info);
3839 3840 3841
			break;
		}

3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853
		/*
		 * 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);
3854
		down_write(&dev_replace->rwsem);
3855 3856 3857
		dev_replace->cursor_right = found_key.offset + length;
		dev_replace->cursor_left = found_key.offset;
		dev_replace->item_needs_writeback = 1;
3858 3859
		up_write(&dev_replace->rwsem);

3860
		ret = scrub_chunk(sctx, scrub_dev, chunk_offset, length,
3861
				  found_key.offset, cache);
3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872

		/*
		 * 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.
		 */
3873
		sctx->flush_all_writes = true;
3874
		scrub_submit(sctx);
3875
		mutex_lock(&sctx->wr_lock);
3876
		scrub_wr_submit(sctx);
3877
		mutex_unlock(&sctx->wr_lock);
3878 3879 3880

		wait_event(sctx->list_wait,
			   atomic_read(&sctx->bios_in_flight) == 0);
3881 3882

		scrub_pause_on(fs_info);
3883 3884 3885 3886 3887 3888

		/*
		 * must be called before we decrease @scrub_paused.
		 * make sure we don't block transaction commit while
		 * we are waiting pending workers finished.
		 */
3889 3890
		wait_event(sctx->list_wait,
			   atomic_read(&sctx->workers_pending) == 0);
3891
		sctx->flush_all_writes = false;
3892

3893
		scrub_pause_off(fs_info);
3894

3895 3896 3897 3898 3899
		if (sctx->is_dev_replace &&
		    !btrfs_finish_block_group_to_copy(dev_replace->srcdev,
						      cache, found_key.offset))
			ro_set = 0;

3900
		down_write(&dev_replace->rwsem);
3901 3902
		dev_replace->cursor_left = dev_replace->cursor_right;
		dev_replace->item_needs_writeback = 1;
3903
		up_write(&dev_replace->rwsem);
3904

3905
		if (ro_set)
3906
			btrfs_dec_block_group_ro(cache);
3907

3908 3909 3910 3911 3912 3913 3914 3915 3916
		/*
		 * 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 &&
3917
		    cache->used == 0) {
3918
			spin_unlock(&cache->lock);
3919 3920 3921 3922 3923
			if (btrfs_test_opt(fs_info, DISCARD_ASYNC))
				btrfs_discard_queue_work(&fs_info->discard_ctl,
							 cache);
			else
				btrfs_mark_bg_unused(cache);
3924 3925 3926
		} else {
			spin_unlock(&cache->lock);
		}
3927
skip_unfreeze:
3928
		btrfs_unfreeze_block_group(cache);
A
Arne Jansen 已提交
3929 3930 3931
		btrfs_put_block_group(cache);
		if (ret)
			break;
3932
		if (sctx->is_dev_replace &&
3933
		    atomic64_read(&dev_replace->num_write_errors) > 0) {
3934 3935 3936 3937 3938 3939 3940
			ret = -EIO;
			break;
		}
		if (sctx->stat.malloc_errors > 0) {
			ret = -ENOMEM;
			break;
		}
3941
skip:
A
Arne Jansen 已提交
3942
		key.offset = found_key.offset + length;
C
Chris Mason 已提交
3943
		btrfs_release_path(path);
A
Arne Jansen 已提交
3944 3945 3946
	}

	btrfs_free_path(path);
3947

3948
	return ret;
A
Arne Jansen 已提交
3949 3950
}

3951 3952
static noinline_for_stack int scrub_supers(struct scrub_ctx *sctx,
					   struct btrfs_device *scrub_dev)
A
Arne Jansen 已提交
3953 3954 3955 3956 3957
{
	int	i;
	u64	bytenr;
	u64	gen;
	int	ret;
3958
	struct btrfs_fs_info *fs_info = sctx->fs_info;
A
Arne Jansen 已提交
3959

J
Josef Bacik 已提交
3960
	if (BTRFS_FS_ERROR(fs_info))
3961
		return -EROFS;
3962

3963
	/* Seed devices of a new filesystem has their own generation. */
3964
	if (scrub_dev->fs_devices != fs_info->fs_devices)
3965 3966
		gen = scrub_dev->generation;
	else
3967
		gen = fs_info->last_trans_committed;
A
Arne Jansen 已提交
3968 3969 3970

	for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
		bytenr = btrfs_sb_offset(i);
3971 3972
		if (bytenr + BTRFS_SUPER_INFO_SIZE >
		    scrub_dev->commit_total_bytes)
A
Arne Jansen 已提交
3973
			break;
3974 3975
		if (!btrfs_check_super_location(scrub_dev, bytenr))
			continue;
A
Arne Jansen 已提交
3976

3977
		ret = scrub_pages(sctx, bytenr, BTRFS_SUPER_INFO_SIZE, bytenr,
3978
				  scrub_dev, BTRFS_EXTENT_FLAG_SUPER, gen, i,
3979
				  NULL, bytenr);
A
Arne Jansen 已提交
3980 3981 3982
		if (ret)
			return ret;
	}
3983
	wait_event(sctx->list_wait, atomic_read(&sctx->bios_in_flight) == 0);
A
Arne Jansen 已提交
3984 3985 3986 3987

	return 0;
}

3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010
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
4011 4012 4013
/*
 * get a reference count on fs_info->scrub_workers. start worker if necessary
 */
4014 4015
static noinline_for_stack int scrub_workers_get(struct btrfs_fs_info *fs_info,
						int is_dev_replace)
A
Arne Jansen 已提交
4016
{
4017 4018 4019
	struct btrfs_workqueue *scrub_workers = NULL;
	struct btrfs_workqueue *scrub_wr_comp = NULL;
	struct btrfs_workqueue *scrub_parity = NULL;
4020
	unsigned int flags = WQ_FREEZABLE | WQ_UNBOUND;
4021
	int max_active = fs_info->thread_pool_size;
4022
	int ret = -ENOMEM;
A
Arne Jansen 已提交
4023

4024 4025
	if (refcount_inc_not_zero(&fs_info->scrub_workers_refcnt))
		return 0;
4026

4027 4028 4029 4030
	scrub_workers = btrfs_alloc_workqueue(fs_info, "scrub", flags,
					      is_dev_replace ? 1 : max_active, 4);
	if (!scrub_workers)
		goto fail_scrub_workers;
4031

4032
	scrub_wr_comp = btrfs_alloc_workqueue(fs_info, "scrubwrc", flags,
4033
					      max_active, 2);
4034 4035
	if (!scrub_wr_comp)
		goto fail_scrub_wr_completion_workers;
4036

4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049
	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;
4050
		refcount_set(&fs_info->scrub_workers_refcnt, 1);
4051 4052
		mutex_unlock(&fs_info->scrub_lock);
		return 0;
A
Arne Jansen 已提交
4053
	}
4054 4055 4056
	/* Other thread raced in and created the workers for us */
	refcount_inc(&fs_info->scrub_workers_refcnt);
	mutex_unlock(&fs_info->scrub_lock);
4057

4058 4059
	ret = 0;
	btrfs_destroy_workqueue(scrub_parity);
4060
fail_scrub_parity_workers:
4061
	btrfs_destroy_workqueue(scrub_wr_comp);
4062
fail_scrub_wr_completion_workers:
4063
	btrfs_destroy_workqueue(scrub_workers);
4064
fail_scrub_workers:
4065
	return ret;
A
Arne Jansen 已提交
4066 4067
}

4068 4069
int btrfs_scrub_dev(struct btrfs_fs_info *fs_info, u64 devid, u64 start,
		    u64 end, struct btrfs_scrub_progress *progress,
4070
		    int readonly, int is_dev_replace)
A
Arne Jansen 已提交
4071
{
4072
	struct btrfs_dev_lookup_args args = { .devid = devid };
4073
	struct scrub_ctx *sctx;
A
Arne Jansen 已提交
4074 4075
	int ret;
	struct btrfs_device *dev;
4076
	unsigned int nofs_flag;
A
Arne Jansen 已提交
4077

4078
	if (btrfs_fs_closing(fs_info))
4079
		return -EAGAIN;
A
Arne Jansen 已提交
4080

4081
	if (fs_info->nodesize > BTRFS_STRIPE_LEN) {
4082 4083 4084 4085 4086
		/*
		 * 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.
		 */
4087 4088
		btrfs_err(fs_info,
			   "scrub: size assumption nodesize <= BTRFS_STRIPE_LEN (%d <= %d) fails",
4089 4090
		       fs_info->nodesize,
		       BTRFS_STRIPE_LEN);
4091 4092 4093
		return -EINVAL;
	}

4094
	if (fs_info->nodesize >
4095
	    PAGE_SIZE * SCRUB_MAX_PAGES_PER_BLOCK ||
4096
	    fs_info->sectorsize > PAGE_SIZE * SCRUB_MAX_PAGES_PER_BLOCK) {
4097 4098 4099 4100
		/*
		 * would exhaust the array bounds of pagev member in
		 * struct scrub_block
		 */
J
Jeff Mahoney 已提交
4101 4102
		btrfs_err(fs_info,
			  "scrub: size assumption nodesize and sectorsize <= SCRUB_MAX_PAGES_PER_BLOCK (%d <= %d && %d <= %d) fails",
4103
		       fs_info->nodesize,
4104
		       SCRUB_MAX_PAGES_PER_BLOCK,
4105
		       fs_info->sectorsize,
4106 4107 4108 4109
		       SCRUB_MAX_PAGES_PER_BLOCK);
		return -EINVAL;
	}

4110 4111 4112 4113
	/* 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
4114

4115 4116 4117 4118
	ret = scrub_workers_get(fs_info, is_dev_replace);
	if (ret)
		goto out_free_ctx;

4119
	mutex_lock(&fs_info->fs_devices->device_list_mutex);
4120
	dev = btrfs_find_device(fs_info->fs_devices, &args);
4121 4122
	if (!dev || (test_bit(BTRFS_DEV_STATE_MISSING, &dev->dev_state) &&
		     !is_dev_replace)) {
4123
		mutex_unlock(&fs_info->fs_devices->device_list_mutex);
4124
		ret = -ENODEV;
4125
		goto out;
A
Arne Jansen 已提交
4126 4127
	}

4128 4129
	if (!is_dev_replace && !readonly &&
	    !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) {
4130
		mutex_unlock(&fs_info->fs_devices->device_list_mutex);
4131 4132 4133
		btrfs_err_in_rcu(fs_info,
			"scrub on devid %llu: filesystem on %s is not writable",
				 devid, rcu_str_deref(dev->name));
4134
		ret = -EROFS;
4135
		goto out;
4136 4137
	}

4138
	mutex_lock(&fs_info->scrub_lock);
4139
	if (!test_bit(BTRFS_DEV_STATE_IN_FS_METADATA, &dev->dev_state) ||
4140
	    test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &dev->dev_state)) {
A
Arne Jansen 已提交
4141
		mutex_unlock(&fs_info->scrub_lock);
4142
		mutex_unlock(&fs_info->fs_devices->device_list_mutex);
4143
		ret = -EIO;
4144
		goto out;
A
Arne Jansen 已提交
4145 4146
	}

4147
	down_read(&fs_info->dev_replace.rwsem);
4148
	if (dev->scrub_ctx ||
4149 4150
	    (!is_dev_replace &&
	     btrfs_dev_replace_is_ongoing(&fs_info->dev_replace))) {
4151
		up_read(&fs_info->dev_replace.rwsem);
A
Arne Jansen 已提交
4152
		mutex_unlock(&fs_info->scrub_lock);
4153
		mutex_unlock(&fs_info->fs_devices->device_list_mutex);
4154
		ret = -EINPROGRESS;
4155
		goto out;
A
Arne Jansen 已提交
4156
	}
4157
	up_read(&fs_info->dev_replace.rwsem);
4158

4159
	sctx->readonly = readonly;
4160
	dev->scrub_ctx = sctx;
4161
	mutex_unlock(&fs_info->fs_devices->device_list_mutex);
A
Arne Jansen 已提交
4162

4163 4164 4165 4166
	/*
	 * checking @scrub_pause_req here, we can avoid
	 * race between committing transaction and scrubbing.
	 */
4167
	__scrub_blocked_if_needed(fs_info);
A
Arne Jansen 已提交
4168 4169 4170
	atomic_inc(&fs_info->scrubs_running);
	mutex_unlock(&fs_info->scrub_lock);

4171 4172 4173 4174 4175 4176 4177 4178 4179 4180
	/*
	 * 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();
4181
	if (!is_dev_replace) {
4182
		btrfs_info(fs_info, "scrub: started on devid %llu", devid);
4183 4184 4185 4186
		/*
		 * by holding device list mutex, we can
		 * kick off writing super in log tree sync.
		 */
4187
		mutex_lock(&fs_info->fs_devices->device_list_mutex);
4188
		ret = scrub_supers(sctx, dev);
4189
		mutex_unlock(&fs_info->fs_devices->device_list_mutex);
4190
	}
A
Arne Jansen 已提交
4191 4192

	if (!ret)
4193
		ret = scrub_enumerate_chunks(sctx, dev, start, end);
4194
	memalloc_nofs_restore(nofs_flag);
A
Arne Jansen 已提交
4195

4196
	wait_event(sctx->list_wait, atomic_read(&sctx->bios_in_flight) == 0);
A
Arne Jansen 已提交
4197 4198 4199
	atomic_dec(&fs_info->scrubs_running);
	wake_up(&fs_info->scrub_pause_wait);

4200
	wait_event(sctx->list_wait, atomic_read(&sctx->workers_pending) == 0);
4201

A
Arne Jansen 已提交
4202
	if (progress)
4203
		memcpy(progress, &sctx->stat, sizeof(*progress));
A
Arne Jansen 已提交
4204

4205 4206 4207 4208
	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
4209
	mutex_lock(&fs_info->scrub_lock);
4210
	dev->scrub_ctx = NULL;
A
Arne Jansen 已提交
4211 4212
	mutex_unlock(&fs_info->scrub_lock);

4213
	scrub_workers_put(fs_info);
4214
	scrub_put_ctx(sctx);
A
Arne Jansen 已提交
4215

4216
	return ret;
4217 4218
out:
	scrub_workers_put(fs_info);
4219 4220 4221
out_free_ctx:
	scrub_free_ctx(sctx);

A
Arne Jansen 已提交
4222 4223 4224
	return ret;
}

4225
void btrfs_scrub_pause(struct btrfs_fs_info *fs_info)
A
Arne Jansen 已提交
4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238 4239
{
	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);
}

4240
void btrfs_scrub_continue(struct btrfs_fs_info *fs_info)
A
Arne Jansen 已提交
4241 4242 4243 4244 4245
{
	atomic_dec(&fs_info->scrub_pause_req);
	wake_up(&fs_info->scrub_pause_wait);
}

4246
int btrfs_scrub_cancel(struct btrfs_fs_info *fs_info)
A
Arne Jansen 已提交
4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266
{
	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;
}

4267
int btrfs_scrub_cancel_dev(struct btrfs_device *dev)
4268
{
4269
	struct btrfs_fs_info *fs_info = dev->fs_info;
4270
	struct scrub_ctx *sctx;
A
Arne Jansen 已提交
4271 4272

	mutex_lock(&fs_info->scrub_lock);
4273
	sctx = dev->scrub_ctx;
4274
	if (!sctx) {
A
Arne Jansen 已提交
4275 4276 4277
		mutex_unlock(&fs_info->scrub_lock);
		return -ENOTCONN;
	}
4278
	atomic_inc(&sctx->cancel_req);
4279
	while (dev->scrub_ctx) {
A
Arne Jansen 已提交
4280 4281
		mutex_unlock(&fs_info->scrub_lock);
		wait_event(fs_info->scrub_pause_wait,
4282
			   dev->scrub_ctx == NULL);
A
Arne Jansen 已提交
4283 4284 4285 4286 4287 4288
		mutex_lock(&fs_info->scrub_lock);
	}
	mutex_unlock(&fs_info->scrub_lock);

	return 0;
}
S
Stefan Behrens 已提交
4289

4290
int btrfs_scrub_progress(struct btrfs_fs_info *fs_info, u64 devid,
A
Arne Jansen 已提交
4291 4292
			 struct btrfs_scrub_progress *progress)
{
4293
	struct btrfs_dev_lookup_args args = { .devid = devid };
A
Arne Jansen 已提交
4294
	struct btrfs_device *dev;
4295
	struct scrub_ctx *sctx = NULL;
A
Arne Jansen 已提交
4296

4297
	mutex_lock(&fs_info->fs_devices->device_list_mutex);
4298
	dev = btrfs_find_device(fs_info->fs_devices, &args);
A
Arne Jansen 已提交
4299
	if (dev)
4300
		sctx = dev->scrub_ctx;
4301 4302
	if (sctx)
		memcpy(progress, &sctx->stat, sizeof(*progress));
4303
	mutex_unlock(&fs_info->fs_devices->device_list_mutex);
A
Arne Jansen 已提交
4304

4305
	return dev ? (sctx ? 0 : -ENOTCONN) : -ENODEV;
A
Arne Jansen 已提交
4306
}
4307 4308

static void scrub_remap_extent(struct btrfs_fs_info *fs_info,
4309
			       u64 extent_logical, u32 extent_len,
4310 4311 4312 4313 4314
			       u64 *extent_physical,
			       struct btrfs_device **extent_dev,
			       int *extent_mirror_num)
{
	u64 mapped_length;
4315
	struct btrfs_io_context *bioc = NULL;
4316 4317 4318
	int ret;

	mapped_length = extent_len;
4319
	ret = btrfs_map_block(fs_info, BTRFS_MAP_READ, extent_logical,
4320 4321 4322 4323
			      &mapped_length, &bioc, 0);
	if (ret || !bioc || mapped_length < extent_len ||
	    !bioc->stripes[0].dev->bdev) {
		btrfs_put_bioc(bioc);
4324 4325 4326
		return;
	}

4327 4328 4329 4330
	*extent_physical = bioc->stripes[0].physical;
	*extent_mirror_num = bioc->mirror_num;
	*extent_dev = bioc->stripes[0].dev;
	btrfs_put_bioc(bioc);
4331
}