extent_io.c 146.7 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11
#include <linux/bitops.h>
#include <linux/slab.h>
#include <linux/bio.h>
#include <linux/mm.h>
#include <linux/pagemap.h>
#include <linux/page-flags.h>
#include <linux/spinlock.h>
#include <linux/blkdev.h>
#include <linux/swap.h>
#include <linux/writeback.h>
#include <linux/pagevec.h>
12
#include <linux/prefetch.h>
D
Dan Magenheimer 已提交
13
#include <linux/cleancache.h>
14 15
#include "extent_io.h"
#include "extent_map.h"
16 17
#include "ctree.h"
#include "btrfs_inode.h"
18
#include "volumes.h"
19
#include "check-integrity.h"
20
#include "locking.h"
21
#include "rcu-string.h"
22
#include "backref.h"
23 24 25

static struct kmem_cache *extent_state_cache;
static struct kmem_cache *extent_buffer_cache;
26
static struct bio_set *btrfs_bioset;
27

28 29 30 31 32
static inline bool extent_state_in_tree(const struct extent_state *state)
{
	return !RB_EMPTY_NODE(&state->rb_node);
}

33
#ifdef CONFIG_BTRFS_DEBUG
34 35
static LIST_HEAD(buffers);
static LIST_HEAD(states);
C
Chris Mason 已提交
36

C
Chris Mason 已提交
37
static DEFINE_SPINLOCK(leak_lock);
38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66

static inline
void btrfs_leak_debug_add(struct list_head *new, struct list_head *head)
{
	unsigned long flags;

	spin_lock_irqsave(&leak_lock, flags);
	list_add(new, head);
	spin_unlock_irqrestore(&leak_lock, flags);
}

static inline
void btrfs_leak_debug_del(struct list_head *entry)
{
	unsigned long flags;

	spin_lock_irqsave(&leak_lock, flags);
	list_del(entry);
	spin_unlock_irqrestore(&leak_lock, flags);
}

static inline
void btrfs_leak_debug_check(void)
{
	struct extent_state *state;
	struct extent_buffer *eb;

	while (!list_empty(&states)) {
		state = list_entry(states.next, struct extent_state, leak_list);
67
		pr_err("BTRFS: state leak: start %llu end %llu state %u in tree %d refs %d\n",
68 69
		       state->start, state->end, state->state,
		       extent_state_in_tree(state),
70
		       atomic_read(&state->refs));
71 72 73 74 75 76
		list_del(&state->leak_list);
		kmem_cache_free(extent_state_cache, state);
	}

	while (!list_empty(&buffers)) {
		eb = list_entry(buffers.next, struct extent_buffer, leak_list);
77
		printk(KERN_ERR "BTRFS: buffer leak start %llu len %lu "
78 79
		       "refs %d\n",
		       eb->start, eb->len, atomic_read(&eb->refs));
80 81 82 83
		list_del(&eb->leak_list);
		kmem_cache_free(extent_buffer_cache, eb);
	}
}
84

85 86
#define btrfs_debug_check_extent_io_range(tree, start, end)		\
	__btrfs_debug_check_extent_io_range(__func__, (tree), (start), (end))
87
static inline void __btrfs_debug_check_extent_io_range(const char *caller,
88
		struct extent_io_tree *tree, u64 start, u64 end)
89
{
90 91
	struct inode *inode;
	u64 isize;
92

93 94
	if (!tree->mapping)
		return;
95

96 97
	inode = tree->mapping->host;
	isize = i_size_read(inode);
98
	if (end >= PAGE_SIZE && (end % 2) == 0 && end != isize - 1) {
99 100
		btrfs_debug_rl(BTRFS_I(inode)->root->fs_info,
		    "%s: ino %llu isize %llu odd range [%llu,%llu]",
101
				caller, btrfs_ino(inode), isize, start, end);
102 103
	}
}
104 105 106 107
#else
#define btrfs_leak_debug_add(new, head)	do {} while (0)
#define btrfs_leak_debug_del(entry)	do {} while (0)
#define btrfs_leak_debug_check()	do {} while (0)
108
#define btrfs_debug_check_extent_io_range(c, s, e)	do {} while (0)
C
Chris Mason 已提交
109
#endif
110 111 112 113 114 115 116 117 118 119 120 121 122

#define BUFFER_LRU_MAX 64

struct tree_entry {
	u64 start;
	u64 end;
	struct rb_node rb_node;
};

struct extent_page_data {
	struct bio *bio;
	struct extent_io_tree *tree;
	get_extent_t *get_extent;
123
	unsigned long bio_flags;
124 125 126 127

	/* tells writepage not to lock the state bits for this range
	 * it still does the unlocking
	 */
128 129 130 131
	unsigned int extent_locked:1;

	/* tells the submit_bio code to use a WRITE_SYNC */
	unsigned int sync_io:1;
132 133
};

134 135 136 137 138 139 140 141 142 143
static void add_extent_changeset(struct extent_state *state, unsigned bits,
				 struct extent_changeset *changeset,
				 int set)
{
	int ret;

	if (!changeset)
		return;
	if (set && (state->state & bits) == bits)
		return;
144 145
	if (!set && (state->state & bits) == 0)
		return;
146 147 148 149 150 151 152
	changeset->bytes_changed += state->end - state->start + 1;
	ret = ulist_add(changeset->range_changed, state->start, state->end,
			GFP_ATOMIC);
	/* ENOMEM */
	BUG_ON(ret < 0);
}

153
static noinline void flush_write_bio(void *data);
154 155 156
static inline struct btrfs_fs_info *
tree_fs_info(struct extent_io_tree *tree)
{
157 158
	if (!tree->mapping)
		return NULL;
159 160
	return btrfs_sb(tree->mapping->host->i_sb);
}
161

162 163
int __init extent_io_init(void)
{
D
David Sterba 已提交
164
	extent_state_cache = kmem_cache_create("btrfs_extent_state",
165 166
			sizeof(struct extent_state), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
167 168 169
	if (!extent_state_cache)
		return -ENOMEM;

D
David Sterba 已提交
170
	extent_buffer_cache = kmem_cache_create("btrfs_extent_buffer",
171 172
			sizeof(struct extent_buffer), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
173 174
	if (!extent_buffer_cache)
		goto free_state_cache;
175 176 177 178 179

	btrfs_bioset = bioset_create(BIO_POOL_SIZE,
				     offsetof(struct btrfs_io_bio, bio));
	if (!btrfs_bioset)
		goto free_buffer_cache;
180 181 182 183

	if (bioset_integrity_create(btrfs_bioset, BIO_POOL_SIZE))
		goto free_bioset;

184 185
	return 0;

186 187 188 189
free_bioset:
	bioset_free(btrfs_bioset);
	btrfs_bioset = NULL;

190 191 192 193
free_buffer_cache:
	kmem_cache_destroy(extent_buffer_cache);
	extent_buffer_cache = NULL;

194 195
free_state_cache:
	kmem_cache_destroy(extent_state_cache);
196
	extent_state_cache = NULL;
197 198 199 200 201
	return -ENOMEM;
}

void extent_io_exit(void)
{
202
	btrfs_leak_debug_check();
203 204 205 206 207 208

	/*
	 * Make sure all delayed rcu free are flushed before we
	 * destroy caches.
	 */
	rcu_barrier();
209 210
	kmem_cache_destroy(extent_state_cache);
	kmem_cache_destroy(extent_buffer_cache);
211 212
	if (btrfs_bioset)
		bioset_free(btrfs_bioset);
213 214 215
}

void extent_io_tree_init(struct extent_io_tree *tree,
216
			 struct address_space *mapping)
217
{
218
	tree->state = RB_ROOT;
219 220
	tree->ops = NULL;
	tree->dirty_bytes = 0;
221
	spin_lock_init(&tree->lock);
222 223 224
	tree->mapping = mapping;
}

225
static struct extent_state *alloc_extent_state(gfp_t mask)
226 227 228 229
{
	struct extent_state *state;

	state = kmem_cache_alloc(extent_state_cache, mask);
230
	if (!state)
231 232
		return state;
	state->state = 0;
233
	state->failrec = NULL;
234
	RB_CLEAR_NODE(&state->rb_node);
235
	btrfs_leak_debug_add(&state->leak_list, &states);
236 237
	atomic_set(&state->refs, 1);
	init_waitqueue_head(&state->wq);
238
	trace_alloc_extent_state(state, mask, _RET_IP_);
239 240 241
	return state;
}

242
void free_extent_state(struct extent_state *state)
243 244 245 246
{
	if (!state)
		return;
	if (atomic_dec_and_test(&state->refs)) {
247
		WARN_ON(extent_state_in_tree(state));
248
		btrfs_leak_debug_del(&state->leak_list);
249
		trace_free_extent_state(state, _RET_IP_);
250 251 252 253
		kmem_cache_free(extent_state_cache, state);
	}
}

254 255 256
static struct rb_node *tree_insert(struct rb_root *root,
				   struct rb_node *search_start,
				   u64 offset,
257 258 259
				   struct rb_node *node,
				   struct rb_node ***p_in,
				   struct rb_node **parent_in)
260
{
261
	struct rb_node **p;
C
Chris Mason 已提交
262
	struct rb_node *parent = NULL;
263 264
	struct tree_entry *entry;

265 266 267 268 269 270
	if (p_in && parent_in) {
		p = *p_in;
		parent = *parent_in;
		goto do_insert;
	}

271
	p = search_start ? &search_start : &root->rb_node;
C
Chris Mason 已提交
272
	while (*p) {
273 274 275 276 277 278 279 280 281 282 283
		parent = *p;
		entry = rb_entry(parent, struct tree_entry, rb_node);

		if (offset < entry->start)
			p = &(*p)->rb_left;
		else if (offset > entry->end)
			p = &(*p)->rb_right;
		else
			return parent;
	}

284
do_insert:
285 286 287 288 289
	rb_link_node(node, parent, p);
	rb_insert_color(node, root);
	return NULL;
}

290
static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset,
291 292 293 294
				      struct rb_node **prev_ret,
				      struct rb_node **next_ret,
				      struct rb_node ***p_ret,
				      struct rb_node **parent_ret)
295
{
296
	struct rb_root *root = &tree->state;
297
	struct rb_node **n = &root->rb_node;
298 299 300 301 302
	struct rb_node *prev = NULL;
	struct rb_node *orig_prev = NULL;
	struct tree_entry *entry;
	struct tree_entry *prev_entry = NULL;

303 304 305
	while (*n) {
		prev = *n;
		entry = rb_entry(prev, struct tree_entry, rb_node);
306 307 308
		prev_entry = entry;

		if (offset < entry->start)
309
			n = &(*n)->rb_left;
310
		else if (offset > entry->end)
311
			n = &(*n)->rb_right;
C
Chris Mason 已提交
312
		else
313
			return *n;
314 315
	}

316 317 318 319 320
	if (p_ret)
		*p_ret = n;
	if (parent_ret)
		*parent_ret = prev;

321 322
	if (prev_ret) {
		orig_prev = prev;
C
Chris Mason 已提交
323
		while (prev && offset > prev_entry->end) {
324 325 326 327 328 329 330 331 332
			prev = rb_next(prev);
			prev_entry = rb_entry(prev, struct tree_entry, rb_node);
		}
		*prev_ret = prev;
		prev = orig_prev;
	}

	if (next_ret) {
		prev_entry = rb_entry(prev, struct tree_entry, rb_node);
C
Chris Mason 已提交
333
		while (prev && offset < prev_entry->start) {
334 335 336 337 338 339 340 341
			prev = rb_prev(prev);
			prev_entry = rb_entry(prev, struct tree_entry, rb_node);
		}
		*next_ret = prev;
	}
	return NULL;
}

342 343 344 345 346
static inline struct rb_node *
tree_search_for_insert(struct extent_io_tree *tree,
		       u64 offset,
		       struct rb_node ***p_ret,
		       struct rb_node **parent_ret)
347
{
348
	struct rb_node *prev = NULL;
349
	struct rb_node *ret;
350

351
	ret = __etree_search(tree, offset, &prev, NULL, p_ret, parent_ret);
C
Chris Mason 已提交
352
	if (!ret)
353 354 355 356
		return prev;
	return ret;
}

357 358 359 360 361 362
static inline struct rb_node *tree_search(struct extent_io_tree *tree,
					  u64 offset)
{
	return tree_search_for_insert(tree, offset, NULL, NULL);
}

J
Josef Bacik 已提交
363 364 365 366 367 368 369 370
static void merge_cb(struct extent_io_tree *tree, struct extent_state *new,
		     struct extent_state *other)
{
	if (tree->ops && tree->ops->merge_extent_hook)
		tree->ops->merge_extent_hook(tree->mapping->host, new,
					     other);
}

371 372 373 374 375 376 377 378 379
/*
 * utility function to look for merge candidates inside a given range.
 * Any extents with matching state are merged together into a single
 * extent in the tree.  Extents with EXTENT_IO in their state field
 * are not merged because the end_io handlers need to be able to do
 * operations on them without sleeping (or doing allocations/splits).
 *
 * This should be called with the tree lock held.
 */
380 381
static void merge_state(struct extent_io_tree *tree,
		        struct extent_state *state)
382 383 384 385
{
	struct extent_state *other;
	struct rb_node *other_node;

386
	if (state->state & (EXTENT_IOBITS | EXTENT_BOUNDARY))
387
		return;
388 389 390 391 392 393

	other_node = rb_prev(&state->rb_node);
	if (other_node) {
		other = rb_entry(other_node, struct extent_state, rb_node);
		if (other->end == state->start - 1 &&
		    other->state == state->state) {
J
Josef Bacik 已提交
394
			merge_cb(tree, state, other);
395 396
			state->start = other->start;
			rb_erase(&other->rb_node, &tree->state);
397
			RB_CLEAR_NODE(&other->rb_node);
398 399 400 401 402 403 404 405
			free_extent_state(other);
		}
	}
	other_node = rb_next(&state->rb_node);
	if (other_node) {
		other = rb_entry(other_node, struct extent_state, rb_node);
		if (other->start == state->end + 1 &&
		    other->state == state->state) {
J
Josef Bacik 已提交
406
			merge_cb(tree, state, other);
407 408
			state->end = other->end;
			rb_erase(&other->rb_node, &tree->state);
409
			RB_CLEAR_NODE(&other->rb_node);
410
			free_extent_state(other);
411 412 413 414
		}
	}
}

415
static void set_state_cb(struct extent_io_tree *tree,
416
			 struct extent_state *state, unsigned *bits)
417
{
418 419
	if (tree->ops && tree->ops->set_bit_hook)
		tree->ops->set_bit_hook(tree->mapping->host, state, bits);
420 421 422
}

static void clear_state_cb(struct extent_io_tree *tree,
423
			   struct extent_state *state, unsigned *bits)
424
{
J
Josef Bacik 已提交
425 426
	if (tree->ops && tree->ops->clear_bit_hook)
		tree->ops->clear_bit_hook(tree->mapping->host, state, bits);
427 428
}

429
static void set_state_bits(struct extent_io_tree *tree,
430 431
			   struct extent_state *state, unsigned *bits,
			   struct extent_changeset *changeset);
432

433 434 435 436 437 438 439 440 441 442 443 444
/*
 * insert an extent_state struct into the tree.  'bits' are set on the
 * struct before it is inserted.
 *
 * This may return -EEXIST if the extent is already there, in which case the
 * state struct is freed.
 *
 * The tree lock is not taken internally.  This is a utility function and
 * probably isn't what you want to call (see set/clear_extent_bit).
 */
static int insert_state(struct extent_io_tree *tree,
			struct extent_state *state, u64 start, u64 end,
445 446
			struct rb_node ***p,
			struct rb_node **parent,
447
			unsigned *bits, struct extent_changeset *changeset)
448 449 450
{
	struct rb_node *node;

J
Julia Lawall 已提交
451
	if (end < start)
452
		WARN(1, KERN_ERR "BTRFS: end < start %llu %llu\n",
453
		       end, start);
454 455
	state->start = start;
	state->end = end;
J
Josef Bacik 已提交
456

457
	set_state_bits(tree, state, bits, changeset);
458

459
	node = tree_insert(&tree->state, NULL, end, &state->rb_node, p, parent);
460 461 462
	if (node) {
		struct extent_state *found;
		found = rb_entry(node, struct extent_state, rb_node);
463
		printk(KERN_ERR "BTRFS: found node %llu %llu on insert of "
464 465
		       "%llu %llu\n",
		       found->start, found->end, start, end);
466 467 468 469 470 471
		return -EEXIST;
	}
	merge_state(tree, state);
	return 0;
}

472
static void split_cb(struct extent_io_tree *tree, struct extent_state *orig,
J
Josef Bacik 已提交
473 474 475
		     u64 split)
{
	if (tree->ops && tree->ops->split_extent_hook)
476
		tree->ops->split_extent_hook(tree->mapping->host, orig, split);
J
Josef Bacik 已提交
477 478
}

479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496
/*
 * split a given extent state struct in two, inserting the preallocated
 * struct 'prealloc' as the newly created second half.  'split' indicates an
 * offset inside 'orig' where it should be split.
 *
 * Before calling,
 * the tree has 'orig' at [orig->start, orig->end].  After calling, there
 * are two extent state structs in the tree:
 * prealloc: [orig->start, split - 1]
 * orig: [ split, orig->end ]
 *
 * The tree locks are not taken by this function. They need to be held
 * by the caller.
 */
static int split_state(struct extent_io_tree *tree, struct extent_state *orig,
		       struct extent_state *prealloc, u64 split)
{
	struct rb_node *node;
J
Josef Bacik 已提交
497 498 499

	split_cb(tree, orig, split);

500 501 502 503 504
	prealloc->start = orig->start;
	prealloc->end = split - 1;
	prealloc->state = orig->state;
	orig->start = split;

505 506
	node = tree_insert(&tree->state, &orig->rb_node, prealloc->end,
			   &prealloc->rb_node, NULL, NULL);
507 508 509 510 511 512 513
	if (node) {
		free_extent_state(prealloc);
		return -EEXIST;
	}
	return 0;
}

514 515 516 517 518 519 520 521 522
static struct extent_state *next_state(struct extent_state *state)
{
	struct rb_node *next = rb_next(&state->rb_node);
	if (next)
		return rb_entry(next, struct extent_state, rb_node);
	else
		return NULL;
}

523 524
/*
 * utility function to clear some bits in an extent state struct.
525
 * it will optionally wake up any one waiting on this state (wake == 1).
526 527 528 529
 *
 * If no bits are set on the state struct after clearing things, the
 * struct is freed and removed from the tree
 */
530 531
static struct extent_state *clear_state_bit(struct extent_io_tree *tree,
					    struct extent_state *state,
532 533
					    unsigned *bits, int wake,
					    struct extent_changeset *changeset)
534
{
535
	struct extent_state *next;
536
	unsigned bits_to_clear = *bits & ~EXTENT_CTLBITS;
537

538
	if ((bits_to_clear & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) {
539 540 541 542
		u64 range = state->end - state->start + 1;
		WARN_ON(range > tree->dirty_bytes);
		tree->dirty_bytes -= range;
	}
543
	clear_state_cb(tree, state, bits);
544
	add_extent_changeset(state, bits_to_clear, changeset, 0);
545
	state->state &= ~bits_to_clear;
546 547
	if (wake)
		wake_up(&state->wq);
548
	if (state->state == 0) {
549
		next = next_state(state);
550
		if (extent_state_in_tree(state)) {
551
			rb_erase(&state->rb_node, &tree->state);
552
			RB_CLEAR_NODE(&state->rb_node);
553 554 555 556 557 558
			free_extent_state(state);
		} else {
			WARN_ON(1);
		}
	} else {
		merge_state(tree, state);
559
		next = next_state(state);
560
	}
561
	return next;
562 563
}

564 565 566 567 568 569 570 571 572
static struct extent_state *
alloc_extent_state_atomic(struct extent_state *prealloc)
{
	if (!prealloc)
		prealloc = alloc_extent_state(GFP_ATOMIC);

	return prealloc;
}

573
static void extent_io_tree_panic(struct extent_io_tree *tree, int err)
574 575 576 577 578 579
{
	btrfs_panic(tree_fs_info(tree), err, "Locking error: "
		    "Extent tree was modified by another "
		    "thread while locked.");
}

580 581 582 583 584 585 586 587 588 589
/*
 * clear some bits on a range in the tree.  This may require splitting
 * or inserting elements in the tree, so the gfp mask is used to
 * indicate which allocations or sleeping are allowed.
 *
 * pass 'wake' == 1 to kick any sleepers, and 'delete' == 1 to remove
 * the given range from the tree regardless of state (ie for truncate).
 *
 * the range [start, end] is inclusive.
 *
590
 * This takes the tree lock, and returns 0 on success and < 0 on error.
591
 */
592 593 594 595
static int __clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
			      unsigned bits, int wake, int delete,
			      struct extent_state **cached_state,
			      gfp_t mask, struct extent_changeset *changeset)
596 597
{
	struct extent_state *state;
598
	struct extent_state *cached;
599 600
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
601
	u64 last_end;
602
	int err;
603
	int clear = 0;
604

605
	btrfs_debug_check_extent_io_range(tree, start, end);
606

607 608 609
	if (bits & EXTENT_DELALLOC)
		bits |= EXTENT_NORESERVE;

610 611 612 613
	if (delete)
		bits |= ~EXTENT_CTLBITS;
	bits |= EXTENT_FIRST_DELALLOC;

614 615
	if (bits & (EXTENT_IOBITS | EXTENT_BOUNDARY))
		clear = 1;
616
again:
617
	if (!prealloc && gfpflags_allow_blocking(mask)) {
618 619 620 621 622 623 624
		/*
		 * Don't care for allocation failure here because we might end
		 * up not needing the pre-allocated extent state at all, which
		 * is the case if we only have in the tree extent states that
		 * cover our input range and don't cover too any other range.
		 * If we end up needing a new extent state we allocate it later.
		 */
625 626 627
		prealloc = alloc_extent_state(mask);
	}

628
	spin_lock(&tree->lock);
629 630
	if (cached_state) {
		cached = *cached_state;
631 632 633 634 635 636

		if (clear) {
			*cached_state = NULL;
			cached_state = NULL;
		}

637 638
		if (cached && extent_state_in_tree(cached) &&
		    cached->start <= start && cached->end > start) {
639 640
			if (clear)
				atomic_dec(&cached->refs);
641
			state = cached;
642
			goto hit_next;
643
		}
644 645
		if (clear)
			free_extent_state(cached);
646
	}
647 648 649 650
	/*
	 * this search will find the extents that end after
	 * our range starts
	 */
651
	node = tree_search(tree, start);
652 653 654
	if (!node)
		goto out;
	state = rb_entry(node, struct extent_state, rb_node);
655
hit_next:
656 657 658
	if (state->start > end)
		goto out;
	WARN_ON(state->end < start);
659
	last_end = state->end;
660

661
	/* the state doesn't have the wanted bits, go ahead */
662 663
	if (!(state->state & bits)) {
		state = next_state(state);
664
		goto next;
665
	}
666

667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683
	/*
	 *     | ---- desired range ---- |
	 *  | state | or
	 *  | ------------- state -------------- |
	 *
	 * We need to split the extent we found, and may flip
	 * bits on second half.
	 *
	 * If the extent we found extends past our range, we
	 * just split and search again.  It'll get split again
	 * the next time though.
	 *
	 * If the extent we found is inside our range, we clear
	 * the desired bit on it.
	 */

	if (state->start < start) {
684 685
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
686
		err = split_state(tree, state, prealloc, start);
687 688 689
		if (err)
			extent_io_tree_panic(tree, err);

690 691 692 693
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
694 695
			state = clear_state_bit(tree, state, &bits, wake,
						changeset);
696
			goto next;
697 698 699 700 701 702 703 704 705 706
		}
		goto search_again;
	}
	/*
	 * | ---- desired range ---- |
	 *                        | state |
	 * We need to split the extent, and clear the bit
	 * on the first half
	 */
	if (state->start <= end && state->end > end) {
707 708
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
709
		err = split_state(tree, state, prealloc, end + 1);
710 711 712
		if (err)
			extent_io_tree_panic(tree, err);

713 714
		if (wake)
			wake_up(&state->wq);
715

716
		clear_state_bit(tree, prealloc, &bits, wake, changeset);
J
Josef Bacik 已提交
717

718 719 720
		prealloc = NULL;
		goto out;
	}
721

722
	state = clear_state_bit(tree, state, &bits, wake, changeset);
723
next:
724 725 726
	if (last_end == (u64)-1)
		goto out;
	start = last_end + 1;
727
	if (start <= end && state && !need_resched())
728
		goto hit_next;
729 730 731 732

search_again:
	if (start > end)
		goto out;
733
	spin_unlock(&tree->lock);
734
	if (gfpflags_allow_blocking(mask))
735 736
		cond_resched();
	goto again;
737 738 739 740 741 742 743 744

out:
	spin_unlock(&tree->lock);
	if (prealloc)
		free_extent_state(prealloc);

	return 0;

745 746
}

747 748
static void wait_on_state(struct extent_io_tree *tree,
			  struct extent_state *state)
749 750
		__releases(tree->lock)
		__acquires(tree->lock)
751 752 753
{
	DEFINE_WAIT(wait);
	prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE);
754
	spin_unlock(&tree->lock);
755
	schedule();
756
	spin_lock(&tree->lock);
757 758 759 760 761 762 763 764
	finish_wait(&state->wq, &wait);
}

/*
 * waits for one or more bits to clear on a range in the state tree.
 * The range [start, end] is inclusive.
 * The tree lock is taken by this function
 */
765 766
static void wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
			    unsigned long bits)
767 768 769 770
{
	struct extent_state *state;
	struct rb_node *node;

771
	btrfs_debug_check_extent_io_range(tree, start, end);
772

773
	spin_lock(&tree->lock);
774 775 776 777 778 779
again:
	while (1) {
		/*
		 * this search will find all the extents that end after
		 * our range starts
		 */
780
		node = tree_search(tree, start);
781
process_node:
782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801
		if (!node)
			break;

		state = rb_entry(node, struct extent_state, rb_node);

		if (state->start > end)
			goto out;

		if (state->state & bits) {
			start = state->start;
			atomic_inc(&state->refs);
			wait_on_state(tree, state);
			free_extent_state(state);
			goto again;
		}
		start = state->end + 1;

		if (start > end)
			break;

802 803 804 805
		if (!cond_resched_lock(&tree->lock)) {
			node = rb_next(node);
			goto process_node;
		}
806 807
	}
out:
808
	spin_unlock(&tree->lock);
809 810
}

811
static void set_state_bits(struct extent_io_tree *tree,
812
			   struct extent_state *state,
813
			   unsigned *bits, struct extent_changeset *changeset)
814
{
815
	unsigned bits_to_set = *bits & ~EXTENT_CTLBITS;
J
Josef Bacik 已提交
816

817
	set_state_cb(tree, state, bits);
818
	if ((bits_to_set & EXTENT_DIRTY) && !(state->state & EXTENT_DIRTY)) {
819 820 821
		u64 range = state->end - state->start + 1;
		tree->dirty_bytes += range;
	}
822
	add_extent_changeset(state, bits_to_set, changeset, 1);
823
	state->state |= bits_to_set;
824 825
}

826 827
static void cache_state_if_flags(struct extent_state *state,
				 struct extent_state **cached_ptr,
828
				 unsigned flags)
829 830
{
	if (cached_ptr && !(*cached_ptr)) {
831
		if (!flags || (state->state & flags)) {
832 833 834 835 836 837
			*cached_ptr = state;
			atomic_inc(&state->refs);
		}
	}
}

838 839 840 841 842 843 844
static void cache_state(struct extent_state *state,
			struct extent_state **cached_ptr)
{
	return cache_state_if_flags(state, cached_ptr,
				    EXTENT_IOBITS | EXTENT_BOUNDARY);
}

845
/*
846 847
 * set some bits on a range in the tree.  This may require allocations or
 * sleeping, so the gfp mask is used to indicate what is allowed.
848
 *
849 850 851
 * If any of the exclusive bits are set, this will fail with -EEXIST if some
 * part of the range already has the desired bits set.  The start of the
 * existing range is returned in failed_start in this case.
852
 *
853
 * [start, end] is inclusive This takes the tree lock.
854
 */
855

J
Jeff Mahoney 已提交
856 857
static int __must_check
__set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
858
		 unsigned bits, unsigned exclusive_bits,
859
		 u64 *failed_start, struct extent_state **cached_state,
860
		 gfp_t mask, struct extent_changeset *changeset)
861 862 863 864
{
	struct extent_state *state;
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
865 866
	struct rb_node **p;
	struct rb_node *parent;
867 868 869
	int err = 0;
	u64 last_start;
	u64 last_end;
870

871
	btrfs_debug_check_extent_io_range(tree, start, end);
872

873
	bits |= EXTENT_FIRST_DELALLOC;
874
again:
875
	if (!prealloc && gfpflags_allow_blocking(mask)) {
876 877 878 879 880 881 882
		/*
		 * Don't care for allocation failure here because we might end
		 * up not needing the pre-allocated extent state at all, which
		 * is the case if we only have in the tree extent states that
		 * cover our input range and don't cover too any other range.
		 * If we end up needing a new extent state we allocate it later.
		 */
883 884 885
		prealloc = alloc_extent_state(mask);
	}

886
	spin_lock(&tree->lock);
887 888
	if (cached_state && *cached_state) {
		state = *cached_state;
889
		if (state->start <= start && state->end > start &&
890
		    extent_state_in_tree(state)) {
891 892 893 894
			node = &state->rb_node;
			goto hit_next;
		}
	}
895 896 897 898
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
899
	node = tree_search_for_insert(tree, start, &p, &parent);
900
	if (!node) {
901 902
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
903
		err = insert_state(tree, prealloc, start, end,
904
				   &p, &parent, &bits, changeset);
905 906 907
		if (err)
			extent_io_tree_panic(tree, err);

908
		cache_state(prealloc, cached_state);
909 910 911 912
		prealloc = NULL;
		goto out;
	}
	state = rb_entry(node, struct extent_state, rb_node);
C
Chris Mason 已提交
913
hit_next:
914 915 916 917 918 919 920 921 922 923
	last_start = state->start;
	last_end = state->end;

	/*
	 * | ---- desired range ---- |
	 * | state |
	 *
	 * Just lock what we found and keep going
	 */
	if (state->start == start && state->end <= end) {
924
		if (state->state & exclusive_bits) {
925 926 927 928
			*failed_start = state->start;
			err = -EEXIST;
			goto out;
		}
929

930
		set_state_bits(tree, state, &bits, changeset);
931
		cache_state(state, cached_state);
932
		merge_state(tree, state);
933 934 935
		if (last_end == (u64)-1)
			goto out;
		start = last_end + 1;
936 937 938 939
		state = next_state(state);
		if (start < end && state && state->start == start &&
		    !need_resched())
			goto hit_next;
940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959
		goto search_again;
	}

	/*
	 *     | ---- desired range ---- |
	 * | state |
	 *   or
	 * | ------------- state -------------- |
	 *
	 * We need to split the extent we found, and may flip bits on
	 * second half.
	 *
	 * If the extent we found extends past our
	 * range, we just split and search again.  It'll get split
	 * again the next time though.
	 *
	 * If the extent we found is inside our range, we set the
	 * desired bit on it.
	 */
	if (state->start < start) {
960
		if (state->state & exclusive_bits) {
961 962 963 964
			*failed_start = start;
			err = -EEXIST;
			goto out;
		}
965 966 967

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
968
		err = split_state(tree, state, prealloc, start);
969 970 971
		if (err)
			extent_io_tree_panic(tree, err);

972 973 974 975
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
976
			set_state_bits(tree, state, &bits, changeset);
977
			cache_state(state, cached_state);
978
			merge_state(tree, state);
979 980 981
			if (last_end == (u64)-1)
				goto out;
			start = last_end + 1;
982 983 984 985
			state = next_state(state);
			if (start < end && state && state->start == start &&
			    !need_resched())
				goto hit_next;
986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000
		}
		goto search_again;
	}
	/*
	 * | ---- desired range ---- |
	 *     | state | or               | state |
	 *
	 * There's a hole, we need to insert something in it and
	 * ignore the extent we found.
	 */
	if (state->start > start) {
		u64 this_end;
		if (end < last_start)
			this_end = end;
		else
C
Chris Mason 已提交
1001
			this_end = last_start - 1;
1002 1003 1004

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
1005 1006 1007 1008 1009

		/*
		 * Avoid to free 'prealloc' if it can be merged with
		 * the later extent.
		 */
1010
		err = insert_state(tree, prealloc, start, this_end,
1011
				   NULL, NULL, &bits, changeset);
1012 1013 1014
		if (err)
			extent_io_tree_panic(tree, err);

J
Josef Bacik 已提交
1015 1016
		cache_state(prealloc, cached_state);
		prealloc = NULL;
1017 1018 1019 1020 1021 1022 1023 1024 1025 1026
		start = this_end + 1;
		goto search_again;
	}
	/*
	 * | ---- desired range ---- |
	 *                        | state |
	 * We need to split the extent, and set the bit
	 * on the first half
	 */
	if (state->start <= end && state->end > end) {
1027
		if (state->state & exclusive_bits) {
1028 1029 1030 1031
			*failed_start = start;
			err = -EEXIST;
			goto out;
		}
1032 1033 1034

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
1035
		err = split_state(tree, state, prealloc, end + 1);
1036 1037
		if (err)
			extent_io_tree_panic(tree, err);
1038

1039
		set_state_bits(tree, prealloc, &bits, changeset);
1040
		cache_state(prealloc, cached_state);
1041 1042 1043 1044 1045
		merge_state(tree, prealloc);
		prealloc = NULL;
		goto out;
	}

1046 1047 1048 1049 1050 1051 1052
search_again:
	if (start > end)
		goto out;
	spin_unlock(&tree->lock);
	if (gfpflags_allow_blocking(mask))
		cond_resched();
	goto again;
1053 1054

out:
1055
	spin_unlock(&tree->lock);
1056 1057 1058 1059 1060 1061 1062
	if (prealloc)
		free_extent_state(prealloc);

	return err;

}

1063
int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
1064
		   unsigned bits, u64 * failed_start,
1065
		   struct extent_state **cached_state, gfp_t mask)
J
Jeff Mahoney 已提交
1066 1067
{
	return __set_extent_bit(tree, start, end, bits, 0, failed_start,
1068
				cached_state, mask, NULL);
J
Jeff Mahoney 已提交
1069 1070 1071
}


J
Josef Bacik 已提交
1072
/**
L
Liu Bo 已提交
1073 1074
 * convert_extent_bit - convert all bits in a given range from one bit to
 * 			another
J
Josef Bacik 已提交
1075 1076 1077 1078 1079
 * @tree:	the io tree to search
 * @start:	the start offset in bytes
 * @end:	the end offset in bytes (inclusive)
 * @bits:	the bits to set in this range
 * @clear_bits:	the bits to clear in this range
1080
 * @cached_state:	state that we're going to cache
J
Josef Bacik 已提交
1081 1082 1083 1084 1085 1086
 *
 * This will go through and set bits for the given range.  If any states exist
 * already in this range they are set with the given bit and cleared of the
 * clear_bits.  This is only meant to be used by things that are mergeable, ie
 * converting from say DELALLOC to DIRTY.  This is not meant to be used with
 * boundary bits like LOCK.
1087 1088
 *
 * All allocations are done with GFP_NOFS.
J
Josef Bacik 已提交
1089 1090
 */
int convert_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
1091
		       unsigned bits, unsigned clear_bits,
1092
		       struct extent_state **cached_state)
J
Josef Bacik 已提交
1093 1094 1095 1096
{
	struct extent_state *state;
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
1097 1098
	struct rb_node **p;
	struct rb_node *parent;
J
Josef Bacik 已提交
1099 1100 1101
	int err = 0;
	u64 last_start;
	u64 last_end;
1102
	bool first_iteration = true;
J
Josef Bacik 已提交
1103

1104
	btrfs_debug_check_extent_io_range(tree, start, end);
1105

J
Josef Bacik 已提交
1106
again:
1107
	if (!prealloc) {
1108 1109 1110 1111 1112 1113 1114
		/*
		 * Best effort, don't worry if extent state allocation fails
		 * here for the first iteration. We might have a cached state
		 * that matches exactly the target range, in which case no
		 * extent state allocations are needed. We'll only know this
		 * after locking the tree.
		 */
1115
		prealloc = alloc_extent_state(GFP_NOFS);
1116
		if (!prealloc && !first_iteration)
J
Josef Bacik 已提交
1117 1118 1119 1120
			return -ENOMEM;
	}

	spin_lock(&tree->lock);
1121 1122 1123
	if (cached_state && *cached_state) {
		state = *cached_state;
		if (state->start <= start && state->end > start &&
1124
		    extent_state_in_tree(state)) {
1125 1126 1127 1128 1129
			node = &state->rb_node;
			goto hit_next;
		}
	}

J
Josef Bacik 已提交
1130 1131 1132 1133
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1134
	node = tree_search_for_insert(tree, start, &p, &parent);
J
Josef Bacik 已提交
1135 1136
	if (!node) {
		prealloc = alloc_extent_state_atomic(prealloc);
1137 1138 1139 1140
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
1141
		err = insert_state(tree, prealloc, start, end,
1142
				   &p, &parent, &bits, NULL);
1143 1144
		if (err)
			extent_io_tree_panic(tree, err);
1145 1146
		cache_state(prealloc, cached_state);
		prealloc = NULL;
J
Josef Bacik 已提交
1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160
		goto out;
	}
	state = rb_entry(node, struct extent_state, rb_node);
hit_next:
	last_start = state->start;
	last_end = state->end;

	/*
	 * | ---- desired range ---- |
	 * | state |
	 *
	 * Just lock what we found and keep going
	 */
	if (state->start == start && state->end <= end) {
1161
		set_state_bits(tree, state, &bits, NULL);
1162
		cache_state(state, cached_state);
1163
		state = clear_state_bit(tree, state, &clear_bits, 0, NULL);
J
Josef Bacik 已提交
1164 1165 1166
		if (last_end == (u64)-1)
			goto out;
		start = last_end + 1;
1167 1168 1169
		if (start < end && state && state->start == start &&
		    !need_resched())
			goto hit_next;
J
Josef Bacik 已提交
1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190
		goto search_again;
	}

	/*
	 *     | ---- desired range ---- |
	 * | state |
	 *   or
	 * | ------------- state -------------- |
	 *
	 * We need to split the extent we found, and may flip bits on
	 * second half.
	 *
	 * If the extent we found extends past our
	 * range, we just split and search again.  It'll get split
	 * again the next time though.
	 *
	 * If the extent we found is inside our range, we set the
	 * desired bit on it.
	 */
	if (state->start < start) {
		prealloc = alloc_extent_state_atomic(prealloc);
1191 1192 1193 1194
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1195
		err = split_state(tree, state, prealloc, start);
1196 1197
		if (err)
			extent_io_tree_panic(tree, err);
J
Josef Bacik 已提交
1198 1199 1200 1201
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
1202
			set_state_bits(tree, state, &bits, NULL);
1203
			cache_state(state, cached_state);
1204 1205
			state = clear_state_bit(tree, state, &clear_bits, 0,
						NULL);
J
Josef Bacik 已提交
1206 1207 1208
			if (last_end == (u64)-1)
				goto out;
			start = last_end + 1;
1209 1210 1211
			if (start < end && state && state->start == start &&
			    !need_resched())
				goto hit_next;
J
Josef Bacik 已提交
1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229
		}
		goto search_again;
	}
	/*
	 * | ---- desired range ---- |
	 *     | state | or               | state |
	 *
	 * There's a hole, we need to insert something in it and
	 * ignore the extent we found.
	 */
	if (state->start > start) {
		u64 this_end;
		if (end < last_start)
			this_end = end;
		else
			this_end = last_start - 1;

		prealloc = alloc_extent_state_atomic(prealloc);
1230 1231 1232 1233
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1234 1235 1236 1237 1238 1239

		/*
		 * Avoid to free 'prealloc' if it can be merged with
		 * the later extent.
		 */
		err = insert_state(tree, prealloc, start, this_end,
1240
				   NULL, NULL, &bits, NULL);
1241 1242
		if (err)
			extent_io_tree_panic(tree, err);
1243
		cache_state(prealloc, cached_state);
J
Josef Bacik 已提交
1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255
		prealloc = NULL;
		start = this_end + 1;
		goto search_again;
	}
	/*
	 * | ---- desired range ---- |
	 *                        | state |
	 * We need to split the extent, and set the bit
	 * on the first half
	 */
	if (state->start <= end && state->end > end) {
		prealloc = alloc_extent_state_atomic(prealloc);
1256 1257 1258 1259
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1260 1261

		err = split_state(tree, state, prealloc, end + 1);
1262 1263
		if (err)
			extent_io_tree_panic(tree, err);
J
Josef Bacik 已提交
1264

1265
		set_state_bits(tree, prealloc, &bits, NULL);
1266
		cache_state(prealloc, cached_state);
1267
		clear_state_bit(tree, prealloc, &clear_bits, 0, NULL);
J
Josef Bacik 已提交
1268 1269 1270 1271 1272 1273 1274 1275
		prealloc = NULL;
		goto out;
	}

search_again:
	if (start > end)
		goto out;
	spin_unlock(&tree->lock);
1276
	cond_resched();
1277
	first_iteration = false;
J
Josef Bacik 已提交
1278 1279 1280 1281 1282 1283 1284 1285 1286 1287
	goto again;

out:
	spin_unlock(&tree->lock);
	if (prealloc)
		free_extent_state(prealloc);

	return err;
}

1288
/* wrappers around set/clear extent bit */
1289
int set_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1290
			   unsigned bits, struct extent_changeset *changeset)
1291 1292 1293 1294 1295 1296 1297 1298 1299
{
	/*
	 * We don't support EXTENT_LOCKED yet, as current changeset will
	 * record any bits changed, so for EXTENT_LOCKED case, it will
	 * either fail with -EEXIST or changeset will record the whole
	 * range.
	 */
	BUG_ON(bits & EXTENT_LOCKED);

1300
	return __set_extent_bit(tree, start, end, bits, 0, NULL, NULL, GFP_NOFS,
1301 1302 1303
				changeset);
}

1304 1305 1306 1307 1308 1309 1310 1311 1312
int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
		     unsigned bits, int wake, int delete,
		     struct extent_state **cached, gfp_t mask)
{
	return __clear_extent_bit(tree, start, end, bits, wake, delete,
				  cached, mask, NULL);
}

int clear_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1313
		unsigned bits, struct extent_changeset *changeset)
1314 1315 1316 1317 1318 1319 1320
{
	/*
	 * Don't support EXTENT_LOCKED case, same reason as
	 * set_record_extent_bits().
	 */
	BUG_ON(bits & EXTENT_LOCKED);

1321
	return __clear_extent_bit(tree, start, end, bits, 0, 0, NULL, GFP_NOFS,
1322 1323 1324
				  changeset);
}

C
Chris Mason 已提交
1325 1326 1327 1328
/*
 * either insert or lock state struct between start and end use mask to tell
 * us if waiting is desired.
 */
1329
int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1330
		     struct extent_state **cached_state)
1331 1332 1333
{
	int err;
	u64 failed_start;
1334

1335
	while (1) {
1336
		err = __set_extent_bit(tree, start, end, EXTENT_LOCKED,
J
Jeff Mahoney 已提交
1337
				       EXTENT_LOCKED, &failed_start,
1338
				       cached_state, GFP_NOFS, NULL);
1339
		if (err == -EEXIST) {
1340 1341
			wait_extent_bit(tree, failed_start, end, EXTENT_LOCKED);
			start = failed_start;
1342
		} else
1343 1344 1345 1346 1347 1348
			break;
		WARN_ON(start > end);
	}
	return err;
}

1349
int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
1350 1351 1352 1353
{
	int err;
	u64 failed_start;

J
Jeff Mahoney 已提交
1354
	err = __set_extent_bit(tree, start, end, EXTENT_LOCKED, EXTENT_LOCKED,
1355
			       &failed_start, NULL, GFP_NOFS, NULL);
Y
Yan Zheng 已提交
1356 1357 1358
	if (err == -EEXIST) {
		if (failed_start > start)
			clear_extent_bit(tree, start, failed_start - 1,
1359
					 EXTENT_LOCKED, 1, 0, NULL, GFP_NOFS);
1360
		return 0;
Y
Yan Zheng 已提交
1361
	}
1362 1363 1364
	return 1;
}

1365
void extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end)
1366
{
1367 1368
	unsigned long index = start >> PAGE_SHIFT;
	unsigned long end_index = end >> PAGE_SHIFT;
1369 1370 1371 1372 1373 1374
	struct page *page;

	while (index <= end_index) {
		page = find_get_page(inode->i_mapping, index);
		BUG_ON(!page); /* Pages should be in the extent_io_tree */
		clear_page_dirty_for_io(page);
1375
		put_page(page);
1376 1377 1378 1379
		index++;
	}
}

1380
void extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end)
1381
{
1382 1383
	unsigned long index = start >> PAGE_SHIFT;
	unsigned long end_index = end >> PAGE_SHIFT;
1384 1385 1386 1387 1388 1389
	struct page *page;

	while (index <= end_index) {
		page = find_get_page(inode->i_mapping, index);
		BUG_ON(!page); /* Pages should be in the extent_io_tree */
		__set_page_dirty_nobuffers(page);
1390
		account_page_redirty(page);
1391
		put_page(page);
1392 1393 1394 1395
		index++;
	}
}

1396 1397 1398
/*
 * helper function to set both pages and extents in the tree writeback
 */
1399
static void set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end)
1400
{
1401 1402
	unsigned long index = start >> PAGE_SHIFT;
	unsigned long end_index = end >> PAGE_SHIFT;
1403 1404 1405 1406
	struct page *page;

	while (index <= end_index) {
		page = find_get_page(tree->mapping, index);
1407
		BUG_ON(!page); /* Pages should be in the extent_io_tree */
1408
		set_page_writeback(page);
1409
		put_page(page);
1410 1411 1412 1413
		index++;
	}
}

C
Chris Mason 已提交
1414 1415 1416 1417
/* find the first state struct with 'bits' set after 'start', and
 * return it.  tree->lock must be held.  NULL will returned if
 * nothing was found after 'start'
 */
1418 1419
static struct extent_state *
find_first_extent_bit_state(struct extent_io_tree *tree,
1420
			    u64 start, unsigned bits)
C
Chris Mason 已提交
1421 1422 1423 1424 1425 1426 1427 1428 1429
{
	struct rb_node *node;
	struct extent_state *state;

	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
	node = tree_search(tree, start);
C
Chris Mason 已提交
1430
	if (!node)
C
Chris Mason 已提交
1431 1432
		goto out;

C
Chris Mason 已提交
1433
	while (1) {
C
Chris Mason 已提交
1434
		state = rb_entry(node, struct extent_state, rb_node);
C
Chris Mason 已提交
1435
		if (state->end >= start && (state->state & bits))
C
Chris Mason 已提交
1436
			return state;
C
Chris Mason 已提交
1437

C
Chris Mason 已提交
1438 1439 1440 1441 1442 1443 1444 1445
		node = rb_next(node);
		if (!node)
			break;
	}
out:
	return NULL;
}

1446 1447 1448 1449 1450
/*
 * find the first offset in the io tree with 'bits' set. zero is
 * returned if we find something, and *start_ret and *end_ret are
 * set to reflect the state struct that was found.
 *
1451
 * If nothing was found, 1 is returned. If found something, return 0.
1452 1453
 */
int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
1454
			  u64 *start_ret, u64 *end_ret, unsigned bits,
1455
			  struct extent_state **cached_state)
1456 1457
{
	struct extent_state *state;
1458
	struct rb_node *n;
1459 1460 1461
	int ret = 1;

	spin_lock(&tree->lock);
1462 1463
	if (cached_state && *cached_state) {
		state = *cached_state;
1464
		if (state->end == start - 1 && extent_state_in_tree(state)) {
1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480
			n = rb_next(&state->rb_node);
			while (n) {
				state = rb_entry(n, struct extent_state,
						 rb_node);
				if (state->state & bits)
					goto got_it;
				n = rb_next(n);
			}
			free_extent_state(*cached_state);
			*cached_state = NULL;
			goto out;
		}
		free_extent_state(*cached_state);
		*cached_state = NULL;
	}

1481
	state = find_first_extent_bit_state(tree, start, bits);
1482
got_it:
1483
	if (state) {
1484
		cache_state_if_flags(state, cached_state, 0);
1485 1486 1487 1488
		*start_ret = state->start;
		*end_ret = state->end;
		ret = 0;
	}
1489
out:
1490 1491 1492 1493
	spin_unlock(&tree->lock);
	return ret;
}

C
Chris Mason 已提交
1494 1495 1496 1497 1498 1499
/*
 * find a contiguous range of bytes in the file marked as delalloc, not
 * more than 'max_bytes'.  start and end are used to return the range,
 *
 * 1 is returned if we find something, 0 if nothing was in the tree
 */
C
Chris Mason 已提交
1500
static noinline u64 find_delalloc_range(struct extent_io_tree *tree,
1501 1502
					u64 *start, u64 *end, u64 max_bytes,
					struct extent_state **cached_state)
1503 1504 1505 1506 1507 1508 1509
{
	struct rb_node *node;
	struct extent_state *state;
	u64 cur_start = *start;
	u64 found = 0;
	u64 total_bytes = 0;

1510
	spin_lock(&tree->lock);
C
Chris Mason 已提交
1511

1512 1513 1514 1515
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1516
	node = tree_search(tree, cur_start);
1517
	if (!node) {
1518 1519
		if (!found)
			*end = (u64)-1;
1520 1521 1522
		goto out;
	}

C
Chris Mason 已提交
1523
	while (1) {
1524
		state = rb_entry(node, struct extent_state, rb_node);
1525 1526
		if (found && (state->start != cur_start ||
			      (state->state & EXTENT_BOUNDARY))) {
1527 1528 1529 1530 1531 1532 1533
			goto out;
		}
		if (!(state->state & EXTENT_DELALLOC)) {
			if (!found)
				*end = state->end;
			goto out;
		}
1534
		if (!found) {
1535
			*start = state->start;
1536 1537 1538
			*cached_state = state;
			atomic_inc(&state->refs);
		}
1539 1540 1541 1542 1543
		found++;
		*end = state->end;
		cur_start = state->end + 1;
		node = rb_next(node);
		total_bytes += state->end - state->start + 1;
1544
		if (total_bytes >= max_bytes)
1545 1546
			break;
		if (!node)
1547 1548 1549
			break;
	}
out:
1550
	spin_unlock(&tree->lock);
1551 1552 1553
	return found;
}

1554 1555 1556
static noinline void __unlock_for_delalloc(struct inode *inode,
					   struct page *locked_page,
					   u64 start, u64 end)
C
Chris Mason 已提交
1557 1558 1559
{
	int ret;
	struct page *pages[16];
1560 1561
	unsigned long index = start >> PAGE_SHIFT;
	unsigned long end_index = end >> PAGE_SHIFT;
C
Chris Mason 已提交
1562 1563 1564 1565
	unsigned long nr_pages = end_index - index + 1;
	int i;

	if (index == locked_page->index && end_index == index)
1566
		return;
C
Chris Mason 已提交
1567

C
Chris Mason 已提交
1568
	while (nr_pages > 0) {
C
Chris Mason 已提交
1569
		ret = find_get_pages_contig(inode->i_mapping, index,
1570 1571
				     min_t(unsigned long, nr_pages,
				     ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1572 1573 1574
		for (i = 0; i < ret; i++) {
			if (pages[i] != locked_page)
				unlock_page(pages[i]);
1575
			put_page(pages[i]);
C
Chris Mason 已提交
1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587
		}
		nr_pages -= ret;
		index += ret;
		cond_resched();
	}
}

static noinline int lock_delalloc_pages(struct inode *inode,
					struct page *locked_page,
					u64 delalloc_start,
					u64 delalloc_end)
{
1588
	unsigned long index = delalloc_start >> PAGE_SHIFT;
C
Chris Mason 已提交
1589
	unsigned long start_index = index;
1590
	unsigned long end_index = delalloc_end >> PAGE_SHIFT;
C
Chris Mason 已提交
1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602
	unsigned long pages_locked = 0;
	struct page *pages[16];
	unsigned long nrpages;
	int ret;
	int i;

	/* the caller is responsible for locking the start index */
	if (index == locked_page->index && index == end_index)
		return 0;

	/* skip the page at the start index */
	nrpages = end_index - index + 1;
C
Chris Mason 已提交
1603
	while (nrpages > 0) {
C
Chris Mason 已提交
1604
		ret = find_get_pages_contig(inode->i_mapping, index,
1605 1606
				     min_t(unsigned long,
				     nrpages, ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1607 1608 1609 1610 1611 1612 1613 1614 1615 1616
		if (ret == 0) {
			ret = -EAGAIN;
			goto done;
		}
		/* now we have an array of pages, lock them all */
		for (i = 0; i < ret; i++) {
			/*
			 * the caller is taking responsibility for
			 * locked_page
			 */
1617
			if (pages[i] != locked_page) {
C
Chris Mason 已提交
1618
				lock_page(pages[i]);
1619 1620
				if (!PageDirty(pages[i]) ||
				    pages[i]->mapping != inode->i_mapping) {
1621 1622
					ret = -EAGAIN;
					unlock_page(pages[i]);
1623
					put_page(pages[i]);
1624 1625 1626
					goto done;
				}
			}
1627
			put_page(pages[i]);
1628
			pages_locked++;
C
Chris Mason 已提交
1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639
		}
		nrpages -= ret;
		index += ret;
		cond_resched();
	}
	ret = 0;
done:
	if (ret && pages_locked) {
		__unlock_for_delalloc(inode, locked_page,
			      delalloc_start,
			      ((u64)(start_index + pages_locked - 1)) <<
1640
			      PAGE_SHIFT);
C
Chris Mason 已提交
1641 1642 1643 1644 1645 1646 1647 1648 1649 1650
	}
	return ret;
}

/*
 * find a contiguous range of bytes in the file marked as delalloc, not
 * more than 'max_bytes'.  start and end are used to return the range,
 *
 * 1 is returned if we find something, 0 if nothing was in the tree
 */
1651 1652 1653 1654
STATIC u64 find_lock_delalloc_range(struct inode *inode,
				    struct extent_io_tree *tree,
				    struct page *locked_page, u64 *start,
				    u64 *end, u64 max_bytes)
C
Chris Mason 已提交
1655 1656 1657 1658
{
	u64 delalloc_start;
	u64 delalloc_end;
	u64 found;
1659
	struct extent_state *cached_state = NULL;
C
Chris Mason 已提交
1660 1661 1662 1663 1664 1665 1666 1667
	int ret;
	int loops = 0;

again:
	/* step one, find a bunch of delalloc bytes starting at start */
	delalloc_start = *start;
	delalloc_end = 0;
	found = find_delalloc_range(tree, &delalloc_start, &delalloc_end,
1668
				    max_bytes, &cached_state);
C
Chris Mason 已提交
1669
	if (!found || delalloc_end <= *start) {
C
Chris Mason 已提交
1670 1671
		*start = delalloc_start;
		*end = delalloc_end;
1672
		free_extent_state(cached_state);
L
Liu Bo 已提交
1673
		return 0;
C
Chris Mason 已提交
1674 1675
	}

C
Chris Mason 已提交
1676 1677 1678 1679 1680
	/*
	 * start comes from the offset of locked_page.  We have to lock
	 * pages in order, so we can't process delalloc bytes before
	 * locked_page
	 */
C
Chris Mason 已提交
1681
	if (delalloc_start < *start)
C
Chris Mason 已提交
1682 1683
		delalloc_start = *start;

C
Chris Mason 已提交
1684 1685 1686
	/*
	 * make sure to limit the number of pages we try to lock down
	 */
1687 1688
	if (delalloc_end + 1 - delalloc_start > max_bytes)
		delalloc_end = delalloc_start + max_bytes - 1;
C
Chris Mason 已提交
1689

C
Chris Mason 已提交
1690 1691 1692 1693 1694 1695 1696
	/* step two, lock all the pages after the page that has start */
	ret = lock_delalloc_pages(inode, locked_page,
				  delalloc_start, delalloc_end);
	if (ret == -EAGAIN) {
		/* some of the pages are gone, lets avoid looping by
		 * shortening the size of the delalloc range we're searching
		 */
1697
		free_extent_state(cached_state);
1698
		cached_state = NULL;
C
Chris Mason 已提交
1699
		if (!loops) {
1700
			max_bytes = PAGE_SIZE;
C
Chris Mason 已提交
1701 1702 1703 1704 1705 1706 1707
			loops = 1;
			goto again;
		} else {
			found = 0;
			goto out_failed;
		}
	}
1708
	BUG_ON(ret); /* Only valid values are 0 and -EAGAIN */
C
Chris Mason 已提交
1709 1710

	/* step three, lock the state bits for the whole range */
1711
	lock_extent_bits(tree, delalloc_start, delalloc_end, &cached_state);
C
Chris Mason 已提交
1712 1713 1714

	/* then test to make sure it is all still delalloc */
	ret = test_range_bit(tree, delalloc_start, delalloc_end,
1715
			     EXTENT_DELALLOC, 1, cached_state);
C
Chris Mason 已提交
1716
	if (!ret) {
1717 1718
		unlock_extent_cached(tree, delalloc_start, delalloc_end,
				     &cached_state, GFP_NOFS);
C
Chris Mason 已提交
1719 1720 1721 1722 1723
		__unlock_for_delalloc(inode, locked_page,
			      delalloc_start, delalloc_end);
		cond_resched();
		goto again;
	}
1724
	free_extent_state(cached_state);
C
Chris Mason 已提交
1725 1726 1727 1728 1729 1730
	*start = delalloc_start;
	*end = delalloc_end;
out_failed:
	return found;
}

1731
void extent_clear_unlock_delalloc(struct inode *inode, u64 start, u64 end,
1732
				 struct page *locked_page,
1733
				 unsigned clear_bits,
1734
				 unsigned long page_ops)
C
Chris Mason 已提交
1735
{
1736
	struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
C
Chris Mason 已提交
1737 1738
	int ret;
	struct page *pages[16];
1739 1740
	unsigned long index = start >> PAGE_SHIFT;
	unsigned long end_index = end >> PAGE_SHIFT;
C
Chris Mason 已提交
1741 1742
	unsigned long nr_pages = end_index - index + 1;
	int i;
1743

1744
	clear_extent_bit(tree, start, end, clear_bits, 1, 0, NULL, GFP_NOFS);
1745
	if (page_ops == 0)
1746
		return;
C
Chris Mason 已提交
1747

1748 1749 1750
	if ((page_ops & PAGE_SET_ERROR) && nr_pages > 0)
		mapping_set_error(inode->i_mapping, -EIO);

C
Chris Mason 已提交
1751
	while (nr_pages > 0) {
C
Chris Mason 已提交
1752
		ret = find_get_pages_contig(inode->i_mapping, index,
1753 1754
				     min_t(unsigned long,
				     nr_pages, ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1755
		for (i = 0; i < ret; i++) {
1756

1757
			if (page_ops & PAGE_SET_PRIVATE2)
1758 1759
				SetPagePrivate2(pages[i]);

C
Chris Mason 已提交
1760
			if (pages[i] == locked_page) {
1761
				put_page(pages[i]);
C
Chris Mason 已提交
1762 1763
				continue;
			}
1764
			if (page_ops & PAGE_CLEAR_DIRTY)
C
Chris Mason 已提交
1765
				clear_page_dirty_for_io(pages[i]);
1766
			if (page_ops & PAGE_SET_WRITEBACK)
C
Chris Mason 已提交
1767
				set_page_writeback(pages[i]);
1768 1769
			if (page_ops & PAGE_SET_ERROR)
				SetPageError(pages[i]);
1770
			if (page_ops & PAGE_END_WRITEBACK)
C
Chris Mason 已提交
1771
				end_page_writeback(pages[i]);
1772
			if (page_ops & PAGE_UNLOCK)
1773
				unlock_page(pages[i]);
1774
			put_page(pages[i]);
C
Chris Mason 已提交
1775 1776 1777 1778 1779 1780 1781
		}
		nr_pages -= ret;
		index += ret;
		cond_resched();
	}
}

C
Chris Mason 已提交
1782 1783 1784 1785 1786
/*
 * count the number of bytes in the tree that have a given bit(s)
 * set.  This can be fairly slow, except for EXTENT_DIRTY which is
 * cached.  The total number found is returned.
 */
1787 1788
u64 count_range_bits(struct extent_io_tree *tree,
		     u64 *start, u64 search_end, u64 max_bytes,
1789
		     unsigned bits, int contig)
1790 1791 1792 1793 1794
{
	struct rb_node *node;
	struct extent_state *state;
	u64 cur_start = *start;
	u64 total_bytes = 0;
1795
	u64 last = 0;
1796 1797
	int found = 0;

1798
	if (WARN_ON(search_end <= cur_start))
1799 1800
		return 0;

1801
	spin_lock(&tree->lock);
1802 1803 1804 1805 1806 1807 1808 1809
	if (cur_start == 0 && bits == EXTENT_DIRTY) {
		total_bytes = tree->dirty_bytes;
		goto out;
	}
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1810
	node = tree_search(tree, cur_start);
C
Chris Mason 已提交
1811
	if (!node)
1812 1813
		goto out;

C
Chris Mason 已提交
1814
	while (1) {
1815 1816 1817
		state = rb_entry(node, struct extent_state, rb_node);
		if (state->start > search_end)
			break;
1818 1819 1820
		if (contig && found && state->start > last + 1)
			break;
		if (state->end >= cur_start && (state->state & bits) == bits) {
1821 1822 1823 1824 1825
			total_bytes += min(search_end, state->end) + 1 -
				       max(cur_start, state->start);
			if (total_bytes >= max_bytes)
				break;
			if (!found) {
1826
				*start = max(cur_start, state->start);
1827 1828
				found = 1;
			}
1829 1830 1831
			last = state->end;
		} else if (contig && found) {
			break;
1832 1833 1834 1835 1836 1837
		}
		node = rb_next(node);
		if (!node)
			break;
	}
out:
1838
	spin_unlock(&tree->lock);
1839 1840
	return total_bytes;
}
1841

C
Chris Mason 已提交
1842 1843 1844 1845
/*
 * set the private field for a given byte offset in the tree.  If there isn't
 * an extent_state there already, this does nothing.
 */
1846
static noinline int set_state_failrec(struct extent_io_tree *tree, u64 start,
1847
		struct io_failure_record *failrec)
1848 1849 1850 1851 1852
{
	struct rb_node *node;
	struct extent_state *state;
	int ret = 0;

1853
	spin_lock(&tree->lock);
1854 1855 1856 1857
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1858
	node = tree_search(tree, start);
1859
	if (!node) {
1860 1861 1862 1863 1864 1865 1866 1867
		ret = -ENOENT;
		goto out;
	}
	state = rb_entry(node, struct extent_state, rb_node);
	if (state->start != start) {
		ret = -ENOENT;
		goto out;
	}
1868
	state->failrec = failrec;
1869
out:
1870
	spin_unlock(&tree->lock);
1871 1872 1873
	return ret;
}

1874
static noinline int get_state_failrec(struct extent_io_tree *tree, u64 start,
1875
		struct io_failure_record **failrec)
1876 1877 1878 1879 1880
{
	struct rb_node *node;
	struct extent_state *state;
	int ret = 0;

1881
	spin_lock(&tree->lock);
1882 1883 1884 1885
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1886
	node = tree_search(tree, start);
1887
	if (!node) {
1888 1889 1890 1891 1892 1893 1894 1895
		ret = -ENOENT;
		goto out;
	}
	state = rb_entry(node, struct extent_state, rb_node);
	if (state->start != start) {
		ret = -ENOENT;
		goto out;
	}
1896
	*failrec = state->failrec;
1897
out:
1898
	spin_unlock(&tree->lock);
1899 1900 1901 1902 1903
	return ret;
}

/*
 * searches a range in the state tree for a given mask.
1904
 * If 'filled' == 1, this returns 1 only if every extent in the tree
1905 1906 1907 1908
 * has the bits set.  Otherwise, 1 is returned if any bit in the
 * range is found set.
 */
int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end,
1909
		   unsigned bits, int filled, struct extent_state *cached)
1910 1911 1912 1913 1914
{
	struct extent_state *state = NULL;
	struct rb_node *node;
	int bitset = 0;

1915
	spin_lock(&tree->lock);
1916
	if (cached && extent_state_in_tree(cached) && cached->start <= start &&
1917
	    cached->end > start)
1918 1919 1920
		node = &cached->rb_node;
	else
		node = tree_search(tree, start);
1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939
	while (node && start <= end) {
		state = rb_entry(node, struct extent_state, rb_node);

		if (filled && state->start > start) {
			bitset = 0;
			break;
		}

		if (state->start > end)
			break;

		if (state->state & bits) {
			bitset = 1;
			if (!filled)
				break;
		} else if (filled) {
			bitset = 0;
			break;
		}
1940 1941 1942 1943

		if (state->end == (u64)-1)
			break;

1944 1945 1946 1947 1948 1949 1950 1951 1952 1953
		start = state->end + 1;
		if (start > end)
			break;
		node = rb_next(node);
		if (!node) {
			if (filled)
				bitset = 0;
			break;
		}
	}
1954
	spin_unlock(&tree->lock);
1955 1956 1957 1958 1959 1960 1961
	return bitset;
}

/*
 * helper function to set a given page up to date if all the
 * extents in the tree for that page are up to date
 */
1962
static void check_page_uptodate(struct extent_io_tree *tree, struct page *page)
1963
{
M
Miao Xie 已提交
1964
	u64 start = page_offset(page);
1965
	u64 end = start + PAGE_SIZE - 1;
1966
	if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1, NULL))
1967 1968 1969
		SetPageUptodate(page);
}

1970
int free_io_failure(struct inode *inode, struct io_failure_record *rec)
1971 1972 1973 1974 1975
{
	int ret;
	int err = 0;
	struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;

1976
	set_state_failrec(failure_tree, rec->start, NULL);
1977 1978
	ret = clear_extent_bits(failure_tree, rec->start,
				rec->start + rec->len - 1,
1979
				EXTENT_LOCKED | EXTENT_DIRTY);
1980 1981 1982
	if (ret)
		err = ret;

D
David Woodhouse 已提交
1983 1984
	ret = clear_extent_bits(&BTRFS_I(inode)->io_tree, rec->start,
				rec->start + rec->len - 1,
1985
				EXTENT_DAMAGED);
D
David Woodhouse 已提交
1986 1987
	if (ret && !err)
		err = ret;
1988 1989 1990 1991 1992 1993 1994 1995 1996 1997

	kfree(rec);
	return err;
}

/*
 * this bypasses the standard btrfs submit functions deliberately, as
 * the standard behavior is to write all copies in a raid setup. here we only
 * want to write the one bad copy. so we do the mapping for ourselves and issue
 * submit_bio directly.
1998
 * to avoid any synchronization issues, wait for the data after writing, which
1999 2000 2001 2002
 * actually prevents the read that triggered the error from finishing.
 * currently, there can be no more than two copies of every data bit. thus,
 * exactly one rewrite is required.
 */
2003 2004
int repair_io_failure(struct inode *inode, u64 start, u64 length, u64 logical,
		      struct page *page, unsigned int pg_offset, int mirror_num)
2005
{
2006
	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2007 2008 2009 2010 2011
	struct bio *bio;
	struct btrfs_device *dev;
	u64 map_length = 0;
	u64 sector;
	struct btrfs_bio *bbio = NULL;
D
David Woodhouse 已提交
2012
	struct btrfs_mapping_tree *map_tree = &fs_info->mapping_tree;
2013 2014
	int ret;

2015
	ASSERT(!(fs_info->sb->s_flags & MS_RDONLY));
2016 2017
	BUG_ON(!mirror_num);

D
David Woodhouse 已提交
2018 2019 2020 2021
	/* we can't repair anything in raid56 yet */
	if (btrfs_is_parity_mirror(map_tree, logical, length, mirror_num))
		return 0;

2022
	bio = btrfs_io_bio_alloc(GFP_NOFS, 1);
2023 2024
	if (!bio)
		return -EIO;
2025
	bio->bi_iter.bi_size = 0;
2026 2027
	map_length = length;

2028 2029 2030 2031 2032 2033
	/*
	 * Avoid races with device replace and make sure our bbio has devices
	 * associated to its stripes that don't go away while we are doing the
	 * read repair operation.
	 */
	btrfs_bio_counter_inc_blocked(fs_info);
2034
	ret = btrfs_map_block(fs_info, WRITE, logical,
2035 2036
			      &map_length, &bbio, mirror_num);
	if (ret) {
2037
		btrfs_bio_counter_dec(fs_info);
2038 2039 2040 2041 2042
		bio_put(bio);
		return -EIO;
	}
	BUG_ON(mirror_num != bbio->mirror_num);
	sector = bbio->stripes[mirror_num-1].physical >> 9;
2043
	bio->bi_iter.bi_sector = sector;
2044
	dev = bbio->stripes[mirror_num-1].dev;
2045
	btrfs_put_bbio(bbio);
2046
	if (!dev || !dev->bdev || !dev->writeable) {
2047
		btrfs_bio_counter_dec(fs_info);
2048 2049 2050 2051
		bio_put(bio);
		return -EIO;
	}
	bio->bi_bdev = dev->bdev;
2052
	bio->bi_rw = WRITE_SYNC;
2053
	bio_add_page(bio, page, length, pg_offset);
2054

2055
	if (btrfsic_submit_bio_wait(bio)) {
2056
		/* try to remap that extent elsewhere? */
2057
		btrfs_bio_counter_dec(fs_info);
2058
		bio_put(bio);
2059
		btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_WRITE_ERRS);
2060 2061 2062
		return -EIO;
	}

2063 2064
	btrfs_info_rl_in_rcu(fs_info,
		"read error corrected: ino %llu off %llu (dev %s sector %llu)",
2065 2066
				  btrfs_ino(inode), start,
				  rcu_str_deref(dev->name), sector);
2067
	btrfs_bio_counter_dec(fs_info);
2068 2069 2070 2071
	bio_put(bio);
	return 0;
}

2072 2073 2074 2075 2076
int repair_eb_io_failure(struct btrfs_root *root, struct extent_buffer *eb,
			 int mirror_num)
{
	u64 start = eb->start;
	unsigned long i, num_pages = num_extent_pages(eb->start, eb->len);
2077
	int ret = 0;
2078

2079 2080 2081
	if (root->fs_info->sb->s_flags & MS_RDONLY)
		return -EROFS;

2082
	for (i = 0; i < num_pages; i++) {
2083
		struct page *p = eb->pages[i];
2084 2085

		ret = repair_io_failure(root->fs_info->btree_inode, start,
2086
					PAGE_SIZE, start, p,
2087
					start - page_offset(p), mirror_num);
2088 2089
		if (ret)
			break;
2090
		start += PAGE_SIZE;
2091 2092 2093 2094 2095
	}

	return ret;
}

2096 2097 2098 2099
/*
 * each time an IO finishes, we do a fast check in the IO failure tree
 * to see if we need to process or clean up an io_failure_record
 */
2100 2101
int clean_io_failure(struct inode *inode, u64 start, struct page *page,
		     unsigned int pg_offset)
2102 2103 2104
{
	u64 private;
	struct io_failure_record *failrec;
2105
	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2106 2107 2108 2109 2110 2111 2112 2113 2114 2115
	struct extent_state *state;
	int num_copies;
	int ret;

	private = 0;
	ret = count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private,
				(u64)-1, 1, EXTENT_DIRTY, 0);
	if (!ret)
		return 0;

2116 2117
	ret = get_state_failrec(&BTRFS_I(inode)->io_failure_tree, start,
			&failrec);
2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128
	if (ret)
		return 0;

	BUG_ON(!failrec->this_mirror);

	if (failrec->in_validation) {
		/* there was no real error, just free the record */
		pr_debug("clean_io_failure: freeing dummy error at %llu\n",
			 failrec->start);
		goto out;
	}
2129 2130
	if (fs_info->sb->s_flags & MS_RDONLY)
		goto out;
2131 2132 2133 2134 2135 2136 2137

	spin_lock(&BTRFS_I(inode)->io_tree.lock);
	state = find_first_extent_bit_state(&BTRFS_I(inode)->io_tree,
					    failrec->start,
					    EXTENT_LOCKED);
	spin_unlock(&BTRFS_I(inode)->io_tree.lock);

2138 2139
	if (state && state->start <= failrec->start &&
	    state->end >= failrec->start + failrec->len - 1) {
2140 2141
		num_copies = btrfs_num_copies(fs_info, failrec->logical,
					      failrec->len);
2142
		if (num_copies > 1)  {
2143
			repair_io_failure(inode, start, failrec->len,
2144
					  failrec->logical, page,
2145
					  pg_offset, failrec->failed_mirror);
2146 2147 2148 2149
		}
	}

out:
2150
	free_io_failure(inode, failrec);
2151

2152
	return 0;
2153 2154
}

2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179
/*
 * Can be called when
 * - hold extent lock
 * - under ordered extent
 * - the inode is freeing
 */
void btrfs_free_io_failure_record(struct inode *inode, u64 start, u64 end)
{
	struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
	struct io_failure_record *failrec;
	struct extent_state *state, *next;

	if (RB_EMPTY_ROOT(&failure_tree->state))
		return;

	spin_lock(&failure_tree->lock);
	state = find_first_extent_bit_state(failure_tree, start, EXTENT_DIRTY);
	while (state) {
		if (state->start > end)
			break;

		ASSERT(state->end <= end);

		next = next_state(state);

2180
		failrec = state->failrec;
2181 2182 2183 2184 2185 2186 2187 2188
		free_extent_state(state);
		kfree(failrec);

		state = next;
	}
	spin_unlock(&failure_tree->lock);
}

2189
int btrfs_get_io_failure_record(struct inode *inode, u64 start, u64 end,
2190
		struct io_failure_record **failrec_ret)
2191
{
2192
	struct io_failure_record *failrec;
2193 2194 2195 2196 2197 2198 2199
	struct extent_map *em;
	struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
	struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
	int ret;
	u64 logical;

2200
	ret = get_state_failrec(failure_tree, start, &failrec);
2201 2202 2203 2204
	if (ret) {
		failrec = kzalloc(sizeof(*failrec), GFP_NOFS);
		if (!failrec)
			return -ENOMEM;
2205

2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219
		failrec->start = start;
		failrec->len = end - start + 1;
		failrec->this_mirror = 0;
		failrec->bio_flags = 0;
		failrec->in_validation = 0;

		read_lock(&em_tree->lock);
		em = lookup_extent_mapping(em_tree, start, failrec->len);
		if (!em) {
			read_unlock(&em_tree->lock);
			kfree(failrec);
			return -EIO;
		}

2220
		if (em->start > start || em->start + em->len <= start) {
2221 2222 2223 2224
			free_extent_map(em);
			em = NULL;
		}
		read_unlock(&em_tree->lock);
2225
		if (!em) {
2226 2227 2228
			kfree(failrec);
			return -EIO;
		}
2229

2230 2231 2232 2233 2234 2235 2236 2237
		logical = start - em->start;
		logical = em->block_start + logical;
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
			logical = em->block_start;
			failrec->bio_flags = EXTENT_BIO_COMPRESSED;
			extent_set_compress_type(&failrec->bio_flags,
						 em->compress_type);
		}
2238 2239 2240 2241

		pr_debug("Get IO Failure Record: (new) logical=%llu, start=%llu, len=%llu\n",
			 logical, start, failrec->len);

2242 2243 2244 2245 2246
		failrec->logical = logical;
		free_extent_map(em);

		/* set the bits in the private failure tree */
		ret = set_extent_bits(failure_tree, start, end,
2247
					EXTENT_LOCKED | EXTENT_DIRTY);
2248
		if (ret >= 0)
2249
			ret = set_state_failrec(failure_tree, start, failrec);
2250 2251
		/* set the bits in the inode's tree */
		if (ret >= 0)
2252
			ret = set_extent_bits(tree, start, end, EXTENT_DAMAGED);
2253 2254 2255 2256 2257
		if (ret < 0) {
			kfree(failrec);
			return ret;
		}
	} else {
2258
		pr_debug("Get IO Failure Record: (found) logical=%llu, start=%llu, len=%llu, validation=%d\n",
2259 2260 2261 2262 2263 2264 2265 2266
			 failrec->logical, failrec->start, failrec->len,
			 failrec->in_validation);
		/*
		 * when data can be on disk more than twice, add to failrec here
		 * (e.g. with a list for failed_mirror) to make
		 * clean_io_failure() clean all those errors at once.
		 */
	}
2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277

	*failrec_ret = failrec;

	return 0;
}

int btrfs_check_repairable(struct inode *inode, struct bio *failed_bio,
			   struct io_failure_record *failrec, int failed_mirror)
{
	int num_copies;

2278 2279
	num_copies = btrfs_num_copies(BTRFS_I(inode)->root->fs_info,
				      failrec->logical, failrec->len);
2280 2281 2282 2283 2284 2285
	if (num_copies == 1) {
		/*
		 * we only have a single copy of the data, so don't bother with
		 * all the retry and error correction code that follows. no
		 * matter what the error is, it is very likely to persist.
		 */
2286
		pr_debug("Check Repairable: cannot repair, num_copies=%d, next_mirror %d, failed_mirror %d\n",
2287
			 num_copies, failrec->this_mirror, failed_mirror);
2288
		return 0;
2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324
	}

	/*
	 * there are two premises:
	 *	a) deliver good data to the caller
	 *	b) correct the bad sectors on disk
	 */
	if (failed_bio->bi_vcnt > 1) {
		/*
		 * to fulfill b), we need to know the exact failing sectors, as
		 * we don't want to rewrite any more than the failed ones. thus,
		 * we need separate read requests for the failed bio
		 *
		 * if the following BUG_ON triggers, our validation request got
		 * merged. we need separate requests for our algorithm to work.
		 */
		BUG_ON(failrec->in_validation);
		failrec->in_validation = 1;
		failrec->this_mirror = failed_mirror;
	} else {
		/*
		 * we're ready to fulfill a) and b) alongside. get a good copy
		 * of the failed sector and if we succeed, we have setup
		 * everything for repair_io_failure to do the rest for us.
		 */
		if (failrec->in_validation) {
			BUG_ON(failrec->this_mirror != failed_mirror);
			failrec->in_validation = 0;
			failrec->this_mirror = 0;
		}
		failrec->failed_mirror = failed_mirror;
		failrec->this_mirror++;
		if (failrec->this_mirror == failed_mirror)
			failrec->this_mirror++;
	}

2325
	if (failrec->this_mirror > num_copies) {
2326
		pr_debug("Check Repairable: (fail) num_copies=%d, next_mirror %d, failed_mirror %d\n",
2327
			 num_copies, failrec->this_mirror, failed_mirror);
2328
		return 0;
2329 2330
	}

2331 2332 2333 2334 2335 2336 2337
	return 1;
}


struct bio *btrfs_create_repair_bio(struct inode *inode, struct bio *failed_bio,
				    struct io_failure_record *failrec,
				    struct page *page, int pg_offset, int icsum,
2338
				    bio_end_io_t *endio_func, void *data)
2339 2340 2341 2342 2343
{
	struct bio *bio;
	struct btrfs_io_bio *btrfs_failed_bio;
	struct btrfs_io_bio *btrfs_bio;

2344
	bio = btrfs_io_bio_alloc(GFP_NOFS, 1);
2345 2346 2347 2348
	if (!bio)
		return NULL;

	bio->bi_end_io = endio_func;
2349
	bio->bi_iter.bi_sector = failrec->logical >> 9;
2350
	bio->bi_bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
2351
	bio->bi_iter.bi_size = 0;
2352
	bio->bi_private = data;
2353

2354 2355 2356 2357 2358 2359 2360
	btrfs_failed_bio = btrfs_io_bio(failed_bio);
	if (btrfs_failed_bio->csum) {
		struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
		u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);

		btrfs_bio = btrfs_io_bio(bio);
		btrfs_bio->csum = btrfs_bio->csum_inline;
2361 2362
		icsum *= csum_size;
		memcpy(btrfs_bio->csum, btrfs_failed_bio->csum + icsum,
2363 2364 2365
		       csum_size);
	}

2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389
	bio_add_page(bio, page, failrec->len, pg_offset);

	return bio;
}

/*
 * this is a generic handler for readpage errors (default
 * readpage_io_failed_hook). if other copies exist, read those and write back
 * good data to the failed position. does not investigate in remapping the
 * failed extent elsewhere, hoping the device will be smart enough to do this as
 * needed
 */

static int bio_readpage_error(struct bio *failed_bio, u64 phy_offset,
			      struct page *page, u64 start, u64 end,
			      int failed_mirror)
{
	struct io_failure_record *failrec;
	struct inode *inode = page->mapping->host;
	struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
	struct bio *bio;
	int read_mode;
	int ret;

2390
	BUG_ON(bio_op(failed_bio) == REQ_OP_WRITE);
2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409

	ret = btrfs_get_io_failure_record(inode, start, end, &failrec);
	if (ret)
		return ret;

	ret = btrfs_check_repairable(inode, failed_bio, failrec, failed_mirror);
	if (!ret) {
		free_io_failure(inode, failrec);
		return -EIO;
	}

	if (failed_bio->bi_vcnt > 1)
		read_mode = READ_SYNC | REQ_FAILFAST_DEV;
	else
		read_mode = READ_SYNC;

	phy_offset >>= inode->i_sb->s_blocksize_bits;
	bio = btrfs_create_repair_bio(inode, failed_bio, failrec, page,
				      start - page_offset(page),
2410 2411
				      (int)phy_offset, failed_bio->bi_end_io,
				      NULL);
2412 2413 2414 2415
	if (!bio) {
		free_io_failure(inode, failrec);
		return -EIO;
	}
2416
	bio_set_op_attrs(bio, REQ_OP_READ, read_mode);
2417

2418 2419
	pr_debug("Repair Read Error: submitting new read[%#x] to this_mirror=%d, in_validation=%d\n",
		 read_mode, failrec->this_mirror, failrec->in_validation);
2420

2421 2422 2423
	ret = tree->ops->submit_bio_hook(inode, read_mode, bio,
					 failrec->this_mirror,
					 failrec->bio_flags, 0);
2424
	if (ret) {
2425
		free_io_failure(inode, failrec);
2426 2427 2428
		bio_put(bio);
	}

2429
	return ret;
2430 2431
}

2432 2433
/* lots and lots of room for performance fixes in the end_bio funcs */

2434
void end_extent_writepage(struct page *page, int err, u64 start, u64 end)
2435 2436 2437
{
	int uptodate = (err == 0);
	struct extent_io_tree *tree;
2438
	int ret = 0;
2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451

	tree = &BTRFS_I(page->mapping->host)->io_tree;

	if (tree->ops && tree->ops->writepage_end_io_hook) {
		ret = tree->ops->writepage_end_io_hook(page, start,
					       end, NULL, uptodate);
		if (ret)
			uptodate = 0;
	}

	if (!uptodate) {
		ClearPageUptodate(page);
		SetPageError(page);
2452 2453
		ret = ret < 0 ? ret : -EIO;
		mapping_set_error(page->mapping, ret);
2454 2455 2456
	}
}

2457 2458 2459 2460 2461 2462 2463 2464 2465
/*
 * after a writepage IO is done, we need to:
 * clear the uptodate bits on error
 * clear the writeback bits in the extent tree for this IO
 * end_page_writeback if the page has no more pending IO
 *
 * Scheduling is not allowed, so the extent state tree is expected
 * to have one and only one object corresponding to this IO.
 */
2466
static void end_bio_extent_writepage(struct bio *bio)
2467
{
2468
	struct bio_vec *bvec;
2469 2470
	u64 start;
	u64 end;
2471
	int i;
2472

2473
	bio_for_each_segment_all(bvec, bio, i) {
2474
		struct page *page = bvec->bv_page;
2475

2476 2477 2478 2479 2480
		/* We always issue full-page reads, but if some block
		 * in a page fails to read, blk_update_request() will
		 * advance bv_offset and adjust bv_len to compensate.
		 * Print a warning for nonzero offsets, and an error
		 * if they don't add up to a full page.  */
2481 2482
		if (bvec->bv_offset || bvec->bv_len != PAGE_SIZE) {
			if (bvec->bv_offset + bvec->bv_len != PAGE_SIZE)
2483 2484 2485 2486 2487 2488 2489 2490 2491
				btrfs_err(BTRFS_I(page->mapping->host)->root->fs_info,
				   "partial page write in btrfs with offset %u and length %u",
					bvec->bv_offset, bvec->bv_len);
			else
				btrfs_info(BTRFS_I(page->mapping->host)->root->fs_info,
				   "incomplete page write in btrfs with offset %u and "
				   "length %u",
					bvec->bv_offset, bvec->bv_len);
		}
2492

2493 2494
		start = page_offset(page);
		end = start + bvec->bv_offset + bvec->bv_len - 1;
2495

2496
		end_extent_writepage(page, bio->bi_error, start, end);
2497
		end_page_writeback(page);
2498
	}
2499

2500 2501 2502
	bio_put(bio);
}

2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514
static void
endio_readpage_release_extent(struct extent_io_tree *tree, u64 start, u64 len,
			      int uptodate)
{
	struct extent_state *cached = NULL;
	u64 end = start + len - 1;

	if (uptodate && tree->track_uptodate)
		set_extent_uptodate(tree, start, end, &cached, GFP_ATOMIC);
	unlock_extent_cached(tree, start, end, &cached, GFP_ATOMIC);
}

2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525
/*
 * after a readpage IO is done, we need to:
 * clear the uptodate bits on error
 * set the uptodate bits if things worked
 * set the page up to date if all extents in the tree are uptodate
 * clear the lock bit in the extent tree
 * unlock the page if there are no other extents locked for it
 *
 * Scheduling is not allowed, so the extent state tree is expected
 * to have one and only one object corresponding to this IO.
 */
2526
static void end_bio_extent_readpage(struct bio *bio)
2527
{
2528
	struct bio_vec *bvec;
2529
	int uptodate = !bio->bi_error;
2530
	struct btrfs_io_bio *io_bio = btrfs_io_bio(bio);
2531
	struct extent_io_tree *tree;
2532
	u64 offset = 0;
2533 2534
	u64 start;
	u64 end;
2535
	u64 len;
2536 2537
	u64 extent_start = 0;
	u64 extent_len = 0;
2538
	int mirror;
2539
	int ret;
2540
	int i;
2541

2542
	bio_for_each_segment_all(bvec, bio, i) {
2543
		struct page *page = bvec->bv_page;
2544
		struct inode *inode = page->mapping->host;
2545

2546
		pr_debug("end_bio_extent_readpage: bi_sector=%llu, err=%d, "
2547 2548
			 "mirror=%u\n", (u64)bio->bi_iter.bi_sector,
			 bio->bi_error, io_bio->mirror_num);
2549
		tree = &BTRFS_I(inode)->io_tree;
2550

2551 2552 2553 2554 2555
		/* We always issue full-page reads, but if some block
		 * in a page fails to read, blk_update_request() will
		 * advance bv_offset and adjust bv_len to compensate.
		 * Print a warning for nonzero offsets, and an error
		 * if they don't add up to a full page.  */
2556 2557
		if (bvec->bv_offset || bvec->bv_len != PAGE_SIZE) {
			if (bvec->bv_offset + bvec->bv_len != PAGE_SIZE)
2558 2559 2560 2561 2562 2563 2564 2565 2566
				btrfs_err(BTRFS_I(page->mapping->host)->root->fs_info,
				   "partial page read in btrfs with offset %u and length %u",
					bvec->bv_offset, bvec->bv_len);
			else
				btrfs_info(BTRFS_I(page->mapping->host)->root->fs_info,
				   "incomplete page read in btrfs with offset %u and "
				   "length %u",
					bvec->bv_offset, bvec->bv_len);
		}
2567

2568 2569
		start = page_offset(page);
		end = start + bvec->bv_offset + bvec->bv_len - 1;
2570
		len = bvec->bv_len;
2571

2572
		mirror = io_bio->mirror_num;
2573 2574
		if (likely(uptodate && tree->ops &&
			   tree->ops->readpage_end_io_hook)) {
2575 2576 2577
			ret = tree->ops->readpage_end_io_hook(io_bio, offset,
							      page, start, end,
							      mirror);
2578
			if (ret)
2579
				uptodate = 0;
2580
			else
2581
				clean_io_failure(inode, start, page, 0);
2582
		}
2583

2584 2585 2586 2587
		if (likely(uptodate))
			goto readpage_ok;

		if (tree->ops && tree->ops->readpage_io_failed_hook) {
2588
			ret = tree->ops->readpage_io_failed_hook(page, mirror);
2589
			if (!ret && !bio->bi_error)
2590
				uptodate = 1;
2591
		} else {
2592 2593 2594 2595 2596 2597 2598 2599 2600 2601
			/*
			 * The generic bio_readpage_error handles errors the
			 * following way: If possible, new read requests are
			 * created and submitted and will end up in
			 * end_bio_extent_readpage as well (if we're lucky, not
			 * in the !uptodate case). In that case it returns 0 and
			 * we just go on with the next page in our bio. If it
			 * can't handle the error it will return -EIO and we
			 * remain responsible for that page.
			 */
2602 2603
			ret = bio_readpage_error(bio, offset, page, start, end,
						 mirror);
2604
			if (ret == 0) {
2605
				uptodate = !bio->bi_error;
2606
				offset += len;
2607 2608 2609
				continue;
			}
		}
2610
readpage_ok:
2611
		if (likely(uptodate)) {
2612
			loff_t i_size = i_size_read(inode);
2613
			pgoff_t end_index = i_size >> PAGE_SHIFT;
2614
			unsigned off;
2615 2616

			/* Zero out the end if this page straddles i_size */
2617
			off = i_size & (PAGE_SIZE-1);
2618
			if (page->index == end_index && off)
2619
				zero_user_segment(page, off, PAGE_SIZE);
2620
			SetPageUptodate(page);
2621
		} else {
2622 2623
			ClearPageUptodate(page);
			SetPageError(page);
2624
		}
2625
		unlock_page(page);
2626
		offset += len;
2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648

		if (unlikely(!uptodate)) {
			if (extent_len) {
				endio_readpage_release_extent(tree,
							      extent_start,
							      extent_len, 1);
				extent_start = 0;
				extent_len = 0;
			}
			endio_readpage_release_extent(tree, start,
						      end - start + 1, 0);
		} else if (!extent_len) {
			extent_start = start;
			extent_len = end + 1 - start;
		} else if (extent_start + extent_len == start) {
			extent_len += end + 1 - start;
		} else {
			endio_readpage_release_extent(tree, extent_start,
						      extent_len, uptodate);
			extent_start = start;
			extent_len = end + 1 - start;
		}
2649
	}
2650

2651 2652 2653
	if (extent_len)
		endio_readpage_release_extent(tree, extent_start, extent_len,
					      uptodate);
2654
	if (io_bio->end_io)
2655
		io_bio->end_io(io_bio, bio->bi_error);
2656 2657 2658
	bio_put(bio);
}

2659 2660 2661 2662
/*
 * this allocates from the btrfs_bioset.  We're returning a bio right now
 * but you can call btrfs_io_bio for the appropriate container_of magic
 */
2663 2664 2665
struct bio *
btrfs_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
		gfp_t gfp_flags)
2666
{
2667
	struct btrfs_io_bio *btrfs_bio;
2668 2669
	struct bio *bio;

2670
	bio = bio_alloc_bioset(gfp_flags, nr_vecs, btrfs_bioset);
2671 2672

	if (bio == NULL && (current->flags & PF_MEMALLOC)) {
2673 2674 2675 2676
		while (!bio && (nr_vecs /= 2)) {
			bio = bio_alloc_bioset(gfp_flags,
					       nr_vecs, btrfs_bioset);
		}
2677 2678 2679 2680
	}

	if (bio) {
		bio->bi_bdev = bdev;
2681
		bio->bi_iter.bi_sector = first_sector;
2682 2683 2684 2685
		btrfs_bio = btrfs_io_bio(bio);
		btrfs_bio->csum = NULL;
		btrfs_bio->csum_allocated = NULL;
		btrfs_bio->end_io = NULL;
2686 2687 2688 2689
	}
	return bio;
}

2690 2691
struct bio *btrfs_bio_clone(struct bio *bio, gfp_t gfp_mask)
{
2692 2693
	struct btrfs_io_bio *btrfs_bio;
	struct bio *new;
2694

2695 2696 2697 2698 2699 2700
	new = bio_clone_bioset(bio, gfp_mask, btrfs_bioset);
	if (new) {
		btrfs_bio = btrfs_io_bio(new);
		btrfs_bio->csum = NULL;
		btrfs_bio->csum_allocated = NULL;
		btrfs_bio->end_io = NULL;
2701 2702

#ifdef CONFIG_BLK_CGROUP
2703 2704 2705
		/* FIXME, put this into bio_clone_bioset */
		if (bio->bi_css)
			bio_associate_blkcg(new, bio->bi_css);
2706
#endif
2707 2708 2709
	}
	return new;
}
2710 2711 2712 2713

/* this also allocates from the btrfs_bioset */
struct bio *btrfs_io_bio_alloc(gfp_t gfp_mask, unsigned int nr_iovecs)
{
2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724
	struct btrfs_io_bio *btrfs_bio;
	struct bio *bio;

	bio = bio_alloc_bioset(gfp_mask, nr_iovecs, btrfs_bioset);
	if (bio) {
		btrfs_bio = btrfs_io_bio(bio);
		btrfs_bio->csum = NULL;
		btrfs_bio->csum_allocated = NULL;
		btrfs_bio->end_io = NULL;
	}
	return bio;
2725 2726 2727
}


2728 2729
static int __must_check submit_one_bio(struct bio *bio, int mirror_num,
				       unsigned long bio_flags)
2730 2731
{
	int ret = 0;
2732 2733 2734 2735 2736
	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
	struct page *page = bvec->bv_page;
	struct extent_io_tree *tree = bio->bi_private;
	u64 start;

M
Miao Xie 已提交
2737
	start = page_offset(page) + bvec->bv_offset;
2738

2739
	bio->bi_private = NULL;
2740 2741
	bio_get(bio);

2742
	if (tree->ops && tree->ops->submit_bio_hook)
2743 2744 2745
		ret = tree->ops->submit_bio_hook(page->mapping->host,
						 bio->bi_rw, bio, mirror_num,
						 bio_flags, start);
2746
	else
2747
		btrfsic_submit_bio(bio);
2748

2749 2750 2751 2752
	bio_put(bio);
	return ret;
}

2753
static int merge_bio(struct extent_io_tree *tree, struct page *page,
2754 2755 2756 2757 2758
		     unsigned long offset, size_t size, struct bio *bio,
		     unsigned long bio_flags)
{
	int ret = 0;
	if (tree->ops && tree->ops->merge_bio_hook)
2759 2760
		ret = tree->ops->merge_bio_hook(bio_op(bio), page, offset, size,
						bio, bio_flags);
2761 2762 2763 2764 2765
	BUG_ON(ret < 0);
	return ret;

}

2766
static int submit_extent_page(int op, int op_flags, struct extent_io_tree *tree,
2767
			      struct writeback_control *wbc,
2768 2769 2770 2771 2772
			      struct page *page, sector_t sector,
			      size_t size, unsigned long offset,
			      struct block_device *bdev,
			      struct bio **bio_ret,
			      unsigned long max_pages,
2773
			      bio_end_io_t end_io_func,
C
Chris Mason 已提交
2774 2775
			      int mirror_num,
			      unsigned long prev_bio_flags,
2776 2777
			      unsigned long bio_flags,
			      bool force_bio_submit)
2778 2779 2780
{
	int ret = 0;
	struct bio *bio;
C
Chris Mason 已提交
2781 2782
	int contig = 0;
	int old_compressed = prev_bio_flags & EXTENT_BIO_COMPRESSED;
2783
	size_t page_size = min_t(size_t, size, PAGE_SIZE);
2784 2785 2786

	if (bio_ret && *bio_ret) {
		bio = *bio_ret;
C
Chris Mason 已提交
2787
		if (old_compressed)
2788
			contig = bio->bi_iter.bi_sector == sector;
C
Chris Mason 已提交
2789
		else
K
Kent Overstreet 已提交
2790
			contig = bio_end_sector(bio) == sector;
C
Chris Mason 已提交
2791 2792

		if (prev_bio_flags != bio_flags || !contig ||
2793
		    force_bio_submit ||
2794
		    merge_bio(tree, page, offset, page_size, bio, bio_flags) ||
C
Chris Mason 已提交
2795
		    bio_add_page(bio, page, page_size, offset) < page_size) {
2796
			ret = submit_one_bio(bio, mirror_num, prev_bio_flags);
2797 2798
			if (ret < 0) {
				*bio_ret = NULL;
2799
				return ret;
2800
			}
2801 2802
			bio = NULL;
		} else {
2803 2804
			if (wbc)
				wbc_account_io(wbc, page, page_size);
2805 2806 2807
			return 0;
		}
	}
C
Chris Mason 已提交
2808

2809 2810
	bio = btrfs_bio_alloc(bdev, sector, BIO_MAX_PAGES,
			GFP_NOFS | __GFP_HIGH);
2811 2812
	if (!bio)
		return -ENOMEM;
2813

C
Chris Mason 已提交
2814
	bio_add_page(bio, page, page_size, offset);
2815 2816
	bio->bi_end_io = end_io_func;
	bio->bi_private = tree;
2817
	bio_set_op_attrs(bio, op, op_flags);
2818 2819 2820 2821
	if (wbc) {
		wbc_init_bio(wbc, bio);
		wbc_account_io(wbc, page, page_size);
	}
2822

C
Chris Mason 已提交
2823
	if (bio_ret)
2824
		*bio_ret = bio;
C
Chris Mason 已提交
2825
	else
2826
		ret = submit_one_bio(bio, mirror_num, bio_flags);
2827 2828 2829 2830

	return ret;
}

2831 2832
static void attach_extent_buffer_page(struct extent_buffer *eb,
				      struct page *page)
2833 2834 2835
{
	if (!PagePrivate(page)) {
		SetPagePrivate(page);
2836
		get_page(page);
J
Josef Bacik 已提交
2837 2838 2839
		set_page_private(page, (unsigned long)eb);
	} else {
		WARN_ON(page->private != (unsigned long)eb);
2840 2841 2842
	}
}

J
Josef Bacik 已提交
2843
void set_page_extent_mapped(struct page *page)
2844
{
J
Josef Bacik 已提交
2845 2846
	if (!PagePrivate(page)) {
		SetPagePrivate(page);
2847
		get_page(page);
J
Josef Bacik 已提交
2848 2849
		set_page_private(page, EXTENT_PAGE_PRIVATE);
	}
2850 2851
}

2852 2853 2854 2855 2856 2857 2858 2859 2860
static struct extent_map *
__get_extent_map(struct inode *inode, struct page *page, size_t pg_offset,
		 u64 start, u64 len, get_extent_t *get_extent,
		 struct extent_map **em_cached)
{
	struct extent_map *em;

	if (em_cached && *em_cached) {
		em = *em_cached;
2861
		if (extent_map_in_tree(em) && start >= em->start &&
2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878
		    start < extent_map_end(em)) {
			atomic_inc(&em->refs);
			return em;
		}

		free_extent_map(em);
		*em_cached = NULL;
	}

	em = get_extent(inode, page, pg_offset, start, len, 0);
	if (em_cached && !IS_ERR_OR_NULL(em)) {
		BUG_ON(*em_cached);
		atomic_inc(&em->refs);
		*em_cached = em;
	}
	return em;
}
2879 2880 2881 2882
/*
 * basic readpage implementation.  Locked extent state structs are inserted
 * into the tree that are removed when the IO is done (by the end_io
 * handlers)
2883
 * XXX JDM: This needs looking at to ensure proper page locking
2884
 */
2885 2886 2887
static int __do_readpage(struct extent_io_tree *tree,
			 struct page *page,
			 get_extent_t *get_extent,
2888
			 struct extent_map **em_cached,
2889
			 struct bio **bio, int mirror_num,
2890
			 unsigned long *bio_flags, int read_flags,
2891
			 u64 *prev_em_start)
2892 2893
{
	struct inode *inode = page->mapping->host;
M
Miao Xie 已提交
2894
	u64 start = page_offset(page);
2895
	u64 page_end = start + PAGE_SIZE - 1;
2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906
	u64 end;
	u64 cur = start;
	u64 extent_offset;
	u64 last_byte = i_size_read(inode);
	u64 block_start;
	u64 cur_end;
	sector_t sector;
	struct extent_map *em;
	struct block_device *bdev;
	int ret;
	int nr = 0;
2907
	size_t pg_offset = 0;
2908
	size_t iosize;
C
Chris Mason 已提交
2909
	size_t disk_io_size;
2910
	size_t blocksize = inode->i_sb->s_blocksize;
2911
	unsigned long this_bio_flag = 0;
2912 2913 2914

	set_page_extent_mapped(page);

2915
	end = page_end;
D
Dan Magenheimer 已提交
2916 2917 2918
	if (!PageUptodate(page)) {
		if (cleancache_get_page(page) == 0) {
			BUG_ON(blocksize != PAGE_SIZE);
2919
			unlock_extent(tree, start, end);
D
Dan Magenheimer 已提交
2920 2921 2922 2923
			goto out;
		}
	}

2924
	if (page->index == last_byte >> PAGE_SHIFT) {
C
Chris Mason 已提交
2925
		char *userpage;
2926
		size_t zero_offset = last_byte & (PAGE_SIZE - 1);
C
Chris Mason 已提交
2927 2928

		if (zero_offset) {
2929
			iosize = PAGE_SIZE - zero_offset;
2930
			userpage = kmap_atomic(page);
C
Chris Mason 已提交
2931 2932
			memset(userpage + zero_offset, 0, iosize);
			flush_dcache_page(page);
2933
			kunmap_atomic(userpage);
C
Chris Mason 已提交
2934 2935
		}
	}
2936
	while (cur <= end) {
2937
		unsigned long pnr = (last_byte >> PAGE_SHIFT) + 1;
2938
		bool force_bio_submit = false;
2939

2940 2941
		if (cur >= last_byte) {
			char *userpage;
2942 2943
			struct extent_state *cached = NULL;

2944
			iosize = PAGE_SIZE - pg_offset;
2945
			userpage = kmap_atomic(page);
2946
			memset(userpage + pg_offset, 0, iosize);
2947
			flush_dcache_page(page);
2948
			kunmap_atomic(userpage);
2949
			set_extent_uptodate(tree, cur, cur + iosize - 1,
2950
					    &cached, GFP_NOFS);
2951 2952 2953
			unlock_extent_cached(tree, cur,
					     cur + iosize - 1,
					     &cached, GFP_NOFS);
2954 2955
			break;
		}
2956 2957
		em = __get_extent_map(inode, page, pg_offset, cur,
				      end - cur + 1, get_extent, em_cached);
2958
		if (IS_ERR_OR_NULL(em)) {
2959
			SetPageError(page);
2960
			unlock_extent(tree, cur, end);
2961 2962 2963 2964 2965 2966
			break;
		}
		extent_offset = cur - em->start;
		BUG_ON(extent_map_end(em) <= cur);
		BUG_ON(end < cur);

2967
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
2968
			this_bio_flag |= EXTENT_BIO_COMPRESSED;
2969 2970 2971
			extent_set_compress_type(&this_bio_flag,
						 em->compress_type);
		}
C
Chris Mason 已提交
2972

2973 2974
		iosize = min(extent_map_end(em) - cur, end - cur + 1);
		cur_end = min(extent_map_end(em) - 1, end);
2975
		iosize = ALIGN(iosize, blocksize);
C
Chris Mason 已提交
2976 2977 2978 2979 2980 2981 2982
		if (this_bio_flag & EXTENT_BIO_COMPRESSED) {
			disk_io_size = em->block_len;
			sector = em->block_start >> 9;
		} else {
			sector = (em->block_start + extent_offset) >> 9;
			disk_io_size = iosize;
		}
2983 2984
		bdev = em->bdev;
		block_start = em->block_start;
Y
Yan Zheng 已提交
2985 2986
		if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			block_start = EXTENT_MAP_HOLE;
2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029

		/*
		 * If we have a file range that points to a compressed extent
		 * and it's followed by a consecutive file range that points to
		 * to the same compressed extent (possibly with a different
		 * offset and/or length, so it either points to the whole extent
		 * or only part of it), we must make sure we do not submit a
		 * single bio to populate the pages for the 2 ranges because
		 * this makes the compressed extent read zero out the pages
		 * belonging to the 2nd range. Imagine the following scenario:
		 *
		 *  File layout
		 *  [0 - 8K]                     [8K - 24K]
		 *    |                               |
		 *    |                               |
		 * points to extent X,         points to extent X,
		 * offset 4K, length of 8K     offset 0, length 16K
		 *
		 * [extent X, compressed length = 4K uncompressed length = 16K]
		 *
		 * If the bio to read the compressed extent covers both ranges,
		 * it will decompress extent X into the pages belonging to the
		 * first range and then it will stop, zeroing out the remaining
		 * pages that belong to the other range that points to extent X.
		 * So here we make sure we submit 2 bios, one for the first
		 * range and another one for the third range. Both will target
		 * the same physical extent from disk, but we can't currently
		 * make the compressed bio endio callback populate the pages
		 * for both ranges because each compressed bio is tightly
		 * coupled with a single extent map, and each range can have
		 * an extent map with a different offset value relative to the
		 * uncompressed data of our extent and different lengths. This
		 * is a corner case so we prioritize correctness over
		 * non-optimal behavior (submitting 2 bios for the same extent).
		 */
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags) &&
		    prev_em_start && *prev_em_start != (u64)-1 &&
		    *prev_em_start != em->orig_start)
			force_bio_submit = true;

		if (prev_em_start)
			*prev_em_start = em->orig_start;

3030 3031 3032 3033 3034 3035
		free_extent_map(em);
		em = NULL;

		/* we've found a hole, just zero and go on */
		if (block_start == EXTENT_MAP_HOLE) {
			char *userpage;
3036 3037
			struct extent_state *cached = NULL;

3038
			userpage = kmap_atomic(page);
3039
			memset(userpage + pg_offset, 0, iosize);
3040
			flush_dcache_page(page);
3041
			kunmap_atomic(userpage);
3042 3043

			set_extent_uptodate(tree, cur, cur + iosize - 1,
3044
					    &cached, GFP_NOFS);
3045 3046 3047
			unlock_extent_cached(tree, cur,
					     cur + iosize - 1,
					     &cached, GFP_NOFS);
3048
			cur = cur + iosize;
3049
			pg_offset += iosize;
3050 3051 3052
			continue;
		}
		/* the get_extent function already copied into the page */
3053 3054
		if (test_range_bit(tree, cur, cur_end,
				   EXTENT_UPTODATE, 1, NULL)) {
3055
			check_page_uptodate(tree, page);
3056
			unlock_extent(tree, cur, cur + iosize - 1);
3057
			cur = cur + iosize;
3058
			pg_offset += iosize;
3059 3060
			continue;
		}
3061 3062 3063 3064 3065
		/* we have an inline extent but it didn't get marked up
		 * to date.  Error out
		 */
		if (block_start == EXTENT_MAP_INLINE) {
			SetPageError(page);
3066
			unlock_extent(tree, cur, cur + iosize - 1);
3067
			cur = cur + iosize;
3068
			pg_offset += iosize;
3069 3070
			continue;
		}
3071

3072
		pnr -= page->index;
3073 3074
		ret = submit_extent_page(REQ_OP_READ, read_flags, tree, NULL,
					 page, sector, disk_io_size, pg_offset,
3075
					 bdev, bio, pnr,
C
Chris Mason 已提交
3076 3077
					 end_bio_extent_readpage, mirror_num,
					 *bio_flags,
3078 3079
					 this_bio_flag,
					 force_bio_submit);
3080 3081 3082 3083
		if (!ret) {
			nr++;
			*bio_flags = this_bio_flag;
		} else {
3084
			SetPageError(page);
3085
			unlock_extent(tree, cur, cur + iosize - 1);
3086
		}
3087
		cur = cur + iosize;
3088
		pg_offset += iosize;
3089
	}
D
Dan Magenheimer 已提交
3090
out:
3091 3092 3093 3094 3095 3096 3097 3098
	if (!nr) {
		if (!PageError(page))
			SetPageUptodate(page);
		unlock_page(page);
	}
	return 0;
}

3099 3100 3101 3102
static inline void __do_contiguous_readpages(struct extent_io_tree *tree,
					     struct page *pages[], int nr_pages,
					     u64 start, u64 end,
					     get_extent_t *get_extent,
3103
					     struct extent_map **em_cached,
3104
					     struct bio **bio, int mirror_num,
3105
					     unsigned long *bio_flags,
3106
					     u64 *prev_em_start)
3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124
{
	struct inode *inode;
	struct btrfs_ordered_extent *ordered;
	int index;

	inode = pages[0]->mapping->host;
	while (1) {
		lock_extent(tree, start, end);
		ordered = btrfs_lookup_ordered_range(inode, start,
						     end - start + 1);
		if (!ordered)
			break;
		unlock_extent(tree, start, end);
		btrfs_start_ordered_extent(inode, ordered, 1);
		btrfs_put_ordered_extent(ordered);
	}

	for (index = 0; index < nr_pages; index++) {
3125
		__do_readpage(tree, pages[index], get_extent, em_cached, bio,
3126
			      mirror_num, bio_flags, 0, prev_em_start);
3127
		put_page(pages[index]);
3128 3129 3130 3131 3132 3133
	}
}

static void __extent_readpages(struct extent_io_tree *tree,
			       struct page *pages[],
			       int nr_pages, get_extent_t *get_extent,
3134
			       struct extent_map **em_cached,
3135
			       struct bio **bio, int mirror_num,
3136
			       unsigned long *bio_flags,
3137
			       u64 *prev_em_start)
3138
{
3139
	u64 start = 0;
3140 3141 3142
	u64 end = 0;
	u64 page_start;
	int index;
3143
	int first_index = 0;
3144 3145 3146 3147 3148

	for (index = 0; index < nr_pages; index++) {
		page_start = page_offset(pages[index]);
		if (!end) {
			start = page_start;
3149
			end = start + PAGE_SIZE - 1;
3150 3151
			first_index = index;
		} else if (end + 1 == page_start) {
3152
			end += PAGE_SIZE;
3153 3154 3155
		} else {
			__do_contiguous_readpages(tree, &pages[first_index],
						  index - first_index, start,
3156 3157
						  end, get_extent, em_cached,
						  bio, mirror_num, bio_flags,
3158
						  prev_em_start);
3159
			start = page_start;
3160
			end = start + PAGE_SIZE - 1;
3161 3162 3163 3164 3165 3166 3167
			first_index = index;
		}
	}

	if (end)
		__do_contiguous_readpages(tree, &pages[first_index],
					  index - first_index, start,
3168
					  end, get_extent, em_cached, bio,
3169
					  mirror_num, bio_flags,
3170
					  prev_em_start);
3171 3172 3173 3174 3175 3176
}

static int __extent_read_full_page(struct extent_io_tree *tree,
				   struct page *page,
				   get_extent_t *get_extent,
				   struct bio **bio, int mirror_num,
3177
				   unsigned long *bio_flags, int read_flags)
3178 3179 3180 3181
{
	struct inode *inode = page->mapping->host;
	struct btrfs_ordered_extent *ordered;
	u64 start = page_offset(page);
3182
	u64 end = start + PAGE_SIZE - 1;
3183 3184 3185 3186
	int ret;

	while (1) {
		lock_extent(tree, start, end);
3187
		ordered = btrfs_lookup_ordered_range(inode, start,
3188
						PAGE_SIZE);
3189 3190 3191 3192 3193 3194 3195
		if (!ordered)
			break;
		unlock_extent(tree, start, end);
		btrfs_start_ordered_extent(inode, ordered, 1);
		btrfs_put_ordered_extent(ordered);
	}

3196
	ret = __do_readpage(tree, page, get_extent, NULL, bio, mirror_num,
3197
			    bio_flags, read_flags, NULL);
3198 3199 3200
	return ret;
}

3201
int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
3202
			    get_extent_t *get_extent, int mirror_num)
3203 3204
{
	struct bio *bio = NULL;
C
Chris Mason 已提交
3205
	unsigned long bio_flags = 0;
3206 3207
	int ret;

3208
	ret = __extent_read_full_page(tree, page, get_extent, &bio, mirror_num,
3209
				      &bio_flags, 0);
3210
	if (bio)
3211
		ret = submit_one_bio(bio, mirror_num, bio_flags);
3212 3213 3214
	return ret;
}

3215 3216
static void update_nr_written(struct page *page, struct writeback_control *wbc,
			      unsigned long nr_written)
3217 3218 3219 3220
{
	wbc->nr_to_write -= nr_written;
}

3221
/*
3222 3223 3224 3225 3226 3227 3228 3229
 * helper for __extent_writepage, doing all of the delayed allocation setup.
 *
 * This returns 1 if our fill_delalloc function did all the work required
 * to write the page (copy into inline extent).  In this case the IO has
 * been started and the page is already unlocked.
 *
 * This returns 0 if all went well (page still locked)
 * This returns < 0 if there were errors (page still locked)
3230
 */
3231 3232 3233 3234 3235 3236 3237
static noinline_for_stack int writepage_delalloc(struct inode *inode,
			      struct page *page, struct writeback_control *wbc,
			      struct extent_page_data *epd,
			      u64 delalloc_start,
			      unsigned long *nr_written)
{
	struct extent_io_tree *tree = epd->tree;
3238
	u64 page_end = delalloc_start + PAGE_SIZE - 1;
3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252
	u64 nr_delalloc;
	u64 delalloc_to_write = 0;
	u64 delalloc_end = 0;
	int ret;
	int page_started = 0;

	if (epd->extent_locked || !tree->ops || !tree->ops->fill_delalloc)
		return 0;

	while (delalloc_end < page_end) {
		nr_delalloc = find_lock_delalloc_range(inode, tree,
					       page,
					       &delalloc_start,
					       &delalloc_end,
3253
					       BTRFS_MAX_EXTENT_SIZE);
3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274
		if (nr_delalloc == 0) {
			delalloc_start = delalloc_end + 1;
			continue;
		}
		ret = tree->ops->fill_delalloc(inode, page,
					       delalloc_start,
					       delalloc_end,
					       &page_started,
					       nr_written);
		/* File system has been set read-only */
		if (ret) {
			SetPageError(page);
			/* fill_delalloc should be return < 0 for error
			 * but just in case, we use > 0 here meaning the
			 * IO is started, so we don't want to return > 0
			 * unless things are going well.
			 */
			ret = ret < 0 ? ret : -EIO;
			goto done;
		}
		/*
3275 3276
		 * delalloc_end is already one less than the total length, so
		 * we don't subtract one from PAGE_SIZE
3277 3278
		 */
		delalloc_to_write += (delalloc_end - delalloc_start +
3279
				      PAGE_SIZE) >> PAGE_SHIFT;
3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324
		delalloc_start = delalloc_end + 1;
	}
	if (wbc->nr_to_write < delalloc_to_write) {
		int thresh = 8192;

		if (delalloc_to_write < thresh * 2)
			thresh = delalloc_to_write;
		wbc->nr_to_write = min_t(u64, delalloc_to_write,
					 thresh);
	}

	/* did the fill delalloc function already unlock and start
	 * the IO?
	 */
	if (page_started) {
		/*
		 * we've unlocked the page, so we can't update
		 * the mapping's writeback index, just update
		 * nr_to_write.
		 */
		wbc->nr_to_write -= *nr_written;
		return 1;
	}

	ret = 0;

done:
	return ret;
}

/*
 * helper for __extent_writepage.  This calls the writepage start hooks,
 * and does the loop to map the page into extents and bios.
 *
 * We return 1 if the IO is started and the page is unlocked,
 * 0 if all went well (page still locked)
 * < 0 if there were errors (page still locked)
 */
static noinline_for_stack int __extent_writepage_io(struct inode *inode,
				 struct page *page,
				 struct writeback_control *wbc,
				 struct extent_page_data *epd,
				 loff_t i_size,
				 unsigned long nr_written,
				 int write_flags, int *nr_ret)
3325 3326
{
	struct extent_io_tree *tree = epd->tree;
M
Miao Xie 已提交
3327
	u64 start = page_offset(page);
3328
	u64 page_end = start + PAGE_SIZE - 1;
3329 3330 3331 3332 3333 3334
	u64 end;
	u64 cur = start;
	u64 extent_offset;
	u64 block_start;
	u64 iosize;
	sector_t sector;
3335
	struct extent_state *cached_state = NULL;
3336 3337
	struct extent_map *em;
	struct block_device *bdev;
3338
	size_t pg_offset = 0;
3339
	size_t blocksize;
3340 3341 3342
	int ret = 0;
	int nr = 0;
	bool compressed;
C
Chris Mason 已提交
3343

3344
	if (tree->ops && tree->ops->writepage_start_hook) {
C
Chris Mason 已提交
3345 3346
		ret = tree->ops->writepage_start_hook(page, start,
						      page_end);
3347 3348 3349 3350 3351 3352
		if (ret) {
			/* Fixup worker will requeue */
			if (ret == -EBUSY)
				wbc->pages_skipped++;
			else
				redirty_page_for_writepage(wbc, page);
3353

3354
			update_nr_written(page, wbc, nr_written);
3355
			unlock_page(page);
3356
			ret = 1;
3357
			goto done_unlocked;
3358 3359 3360
		}
	}

3361 3362 3363 3364 3365
	/*
	 * we don't want to touch the inode after unlocking the page,
	 * so we update the mapping writeback index now
	 */
	update_nr_written(page, wbc, nr_written + 1);
3366

3367
	end = page_end;
3368
	if (i_size <= start) {
3369 3370 3371
		if (tree->ops && tree->ops->writepage_end_io_hook)
			tree->ops->writepage_end_io_hook(page, start,
							 page_end, NULL, 1);
3372 3373 3374 3375 3376 3377
		goto done;
	}

	blocksize = inode->i_sb->s_blocksize;

	while (cur <= end) {
3378
		u64 em_end;
3379 3380
		unsigned long max_nr;

3381
		if (cur >= i_size) {
3382 3383 3384
			if (tree->ops && tree->ops->writepage_end_io_hook)
				tree->ops->writepage_end_io_hook(page, cur,
							 page_end, NULL, 1);
3385 3386
			break;
		}
3387
		em = epd->get_extent(inode, page, pg_offset, cur,
3388
				     end - cur + 1, 1);
3389
		if (IS_ERR_OR_NULL(em)) {
3390
			SetPageError(page);
3391
			ret = PTR_ERR_OR_ZERO(em);
3392 3393 3394 3395
			break;
		}

		extent_offset = cur - em->start;
3396 3397
		em_end = extent_map_end(em);
		BUG_ON(em_end <= cur);
3398
		BUG_ON(end < cur);
3399
		iosize = min(em_end - cur, end - cur + 1);
3400
		iosize = ALIGN(iosize, blocksize);
3401 3402 3403
		sector = (em->block_start + extent_offset) >> 9;
		bdev = em->bdev;
		block_start = em->block_start;
C
Chris Mason 已提交
3404
		compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
3405 3406 3407
		free_extent_map(em);
		em = NULL;

C
Chris Mason 已提交
3408 3409 3410 3411 3412
		/*
		 * compressed and inline extents are written through other
		 * paths in the FS
		 */
		if (compressed || block_start == EXTENT_MAP_HOLE ||
3413
		    block_start == EXTENT_MAP_INLINE) {
C
Chris Mason 已提交
3414 3415 3416 3417 3418 3419
			/*
			 * end_io notification does not happen here for
			 * compressed extents
			 */
			if (!compressed && tree->ops &&
			    tree->ops->writepage_end_io_hook)
3420 3421 3422
				tree->ops->writepage_end_io_hook(page, cur,
							 cur + iosize - 1,
							 NULL, 1);
C
Chris Mason 已提交
3423 3424 3425 3426 3427 3428 3429 3430 3431
			else if (compressed) {
				/* we don't want to end_page_writeback on
				 * a compressed extent.  this happens
				 * elsewhere
				 */
				nr++;
			}

			cur += iosize;
3432
			pg_offset += iosize;
3433 3434
			continue;
		}
C
Chris Mason 已提交
3435

3436 3437 3438 3439 3440 3441 3442
		max_nr = (i_size >> PAGE_SHIFT) + 1;

		set_range_writeback(tree, cur, cur + iosize - 1);
		if (!PageWriteback(page)) {
			btrfs_err(BTRFS_I(inode)->root->fs_info,
				   "page %lu not writeback, cur %llu end %llu",
			       page->index, cur, end);
3443
		}
3444

3445 3446
		ret = submit_extent_page(REQ_OP_WRITE, write_flags, tree, wbc,
					 page, sector, iosize, pg_offset,
3447 3448 3449 3450 3451
					 bdev, &epd->bio, max_nr,
					 end_bio_extent_writepage,
					 0, 0, 0, false);
		if (ret)
			SetPageError(page);
3452 3453

		cur = cur + iosize;
3454
		pg_offset += iosize;
3455 3456
		nr++;
	}
3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478
done:
	*nr_ret = nr;

done_unlocked:

	/* drop our reference on any cached states */
	free_extent_state(cached_state);
	return ret;
}

/*
 * the writepage semantics are similar to regular writepage.  extent
 * records are inserted to lock ranges in the tree, and as dirty areas
 * are found, they are marked writeback.  Then the lock bits are removed
 * and the end_io handler clears the writeback ranges
 */
static int __extent_writepage(struct page *page, struct writeback_control *wbc,
			      void *data)
{
	struct inode *inode = page->mapping->host;
	struct extent_page_data *epd = data;
	u64 start = page_offset(page);
3479
	u64 page_end = start + PAGE_SIZE - 1;
3480 3481 3482 3483
	int ret;
	int nr = 0;
	size_t pg_offset = 0;
	loff_t i_size = i_size_read(inode);
3484
	unsigned long end_index = i_size >> PAGE_SHIFT;
3485
	int write_flags = 0;
3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496
	unsigned long nr_written = 0;

	if (wbc->sync_mode == WB_SYNC_ALL)
		write_flags = WRITE_SYNC;

	trace___extent_writepage(page, inode, wbc);

	WARN_ON(!PageLocked(page));

	ClearPageError(page);

3497
	pg_offset = i_size & (PAGE_SIZE - 1);
3498 3499
	if (page->index > end_index ||
	   (page->index == end_index && !pg_offset)) {
3500
		page->mapping->a_ops->invalidatepage(page, 0, PAGE_SIZE);
3501 3502 3503 3504 3505 3506 3507 3508 3509
		unlock_page(page);
		return 0;
	}

	if (page->index == end_index) {
		char *userpage;

		userpage = kmap_atomic(page);
		memset(userpage + pg_offset, 0,
3510
		       PAGE_SIZE - pg_offset);
3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529
		kunmap_atomic(userpage);
		flush_dcache_page(page);
	}

	pg_offset = 0;

	set_page_extent_mapped(page);

	ret = writepage_delalloc(inode, page, wbc, epd, start, &nr_written);
	if (ret == 1)
		goto done_unlocked;
	if (ret)
		goto done;

	ret = __extent_writepage_io(inode, page, wbc, epd,
				    i_size, nr_written, write_flags, &nr);
	if (ret == 1)
		goto done_unlocked;

3530 3531 3532 3533 3534 3535
done:
	if (nr == 0) {
		/* make sure the mapping tag for page dirty gets cleared */
		set_page_writeback(page);
		end_page_writeback(page);
	}
3536 3537 3538 3539
	if (PageError(page)) {
		ret = ret < 0 ? ret : -EIO;
		end_extent_writepage(page, ret, start, page_end);
	}
3540
	unlock_page(page);
3541
	return ret;
3542

3543
done_unlocked:
3544 3545 3546
	return 0;
}

3547
void wait_on_extent_buffer_writeback(struct extent_buffer *eb)
3548
{
3549 3550
	wait_on_bit_io(&eb->bflags, EXTENT_BUFFER_WRITEBACK,
		       TASK_UNINTERRUPTIBLE);
3551 3552
}

3553 3554 3555 3556
static noinline_for_stack int
lock_extent_buffer_for_io(struct extent_buffer *eb,
			  struct btrfs_fs_info *fs_info,
			  struct extent_page_data *epd)
3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575
{
	unsigned long i, num_pages;
	int flush = 0;
	int ret = 0;

	if (!btrfs_try_tree_write_lock(eb)) {
		flush = 1;
		flush_write_bio(epd);
		btrfs_tree_lock(eb);
	}

	if (test_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags)) {
		btrfs_tree_unlock(eb);
		if (!epd->sync_io)
			return 0;
		if (!flush) {
			flush_write_bio(epd);
			flush = 1;
		}
C
Chris Mason 已提交
3576 3577 3578 3579 3580
		while (1) {
			wait_on_extent_buffer_writeback(eb);
			btrfs_tree_lock(eb);
			if (!test_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags))
				break;
3581 3582 3583 3584
			btrfs_tree_unlock(eb);
		}
	}

3585 3586 3587 3588 3589 3590
	/*
	 * We need to do this to prevent races in people who check if the eb is
	 * under IO since we can end up having no IO bits set for a short period
	 * of time.
	 */
	spin_lock(&eb->refs_lock);
3591 3592
	if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
		set_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags);
3593
		spin_unlock(&eb->refs_lock);
3594
		btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
3595 3596 3597
		__percpu_counter_add(&fs_info->dirty_metadata_bytes,
				     -eb->len,
				     fs_info->dirty_metadata_batch);
3598
		ret = 1;
3599 3600
	} else {
		spin_unlock(&eb->refs_lock);
3601 3602 3603 3604 3605 3606 3607 3608 3609
	}

	btrfs_tree_unlock(eb);

	if (!ret)
		return ret;

	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = 0; i < num_pages; i++) {
3610
		struct page *p = eb->pages[i];
3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626

		if (!trylock_page(p)) {
			if (!flush) {
				flush_write_bio(epd);
				flush = 1;
			}
			lock_page(p);
		}
	}

	return ret;
}

static void end_extent_buffer_writeback(struct extent_buffer *eb)
{
	clear_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags);
3627
	smp_mb__after_atomic();
3628 3629 3630
	wake_up_bit(&eb->bflags, EXTENT_BUFFER_WRITEBACK);
}

3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692
static void set_btree_ioerr(struct page *page)
{
	struct extent_buffer *eb = (struct extent_buffer *)page->private;
	struct btrfs_inode *btree_ino = BTRFS_I(eb->fs_info->btree_inode);

	SetPageError(page);
	if (test_and_set_bit(EXTENT_BUFFER_WRITE_ERR, &eb->bflags))
		return;

	/*
	 * If writeback for a btree extent that doesn't belong to a log tree
	 * failed, increment the counter transaction->eb_write_errors.
	 * We do this because while the transaction is running and before it's
	 * committing (when we call filemap_fdata[write|wait]_range against
	 * the btree inode), we might have
	 * btree_inode->i_mapping->a_ops->writepages() called by the VM - if it
	 * returns an error or an error happens during writeback, when we're
	 * committing the transaction we wouldn't know about it, since the pages
	 * can be no longer dirty nor marked anymore for writeback (if a
	 * subsequent modification to the extent buffer didn't happen before the
	 * transaction commit), which makes filemap_fdata[write|wait]_range not
	 * able to find the pages tagged with SetPageError at transaction
	 * commit time. So if this happens we must abort the transaction,
	 * otherwise we commit a super block with btree roots that point to
	 * btree nodes/leafs whose content on disk is invalid - either garbage
	 * or the content of some node/leaf from a past generation that got
	 * cowed or deleted and is no longer valid.
	 *
	 * Note: setting AS_EIO/AS_ENOSPC in the btree inode's i_mapping would
	 * not be enough - we need to distinguish between log tree extents vs
	 * non-log tree extents, and the next filemap_fdatawait_range() call
	 * will catch and clear such errors in the mapping - and that call might
	 * be from a log sync and not from a transaction commit. Also, checking
	 * for the eb flag EXTENT_BUFFER_WRITE_ERR at transaction commit time is
	 * not done and would not be reliable - the eb might have been released
	 * from memory and reading it back again means that flag would not be
	 * set (since it's a runtime flag, not persisted on disk).
	 *
	 * Using the flags below in the btree inode also makes us achieve the
	 * goal of AS_EIO/AS_ENOSPC when writepages() returns success, started
	 * writeback for all dirty pages and before filemap_fdatawait_range()
	 * is called, the writeback for all dirty pages had already finished
	 * with errors - because we were not using AS_EIO/AS_ENOSPC,
	 * filemap_fdatawait_range() would return success, as it could not know
	 * that writeback errors happened (the pages were no longer tagged for
	 * writeback).
	 */
	switch (eb->log_index) {
	case -1:
		set_bit(BTRFS_INODE_BTREE_ERR, &btree_ino->runtime_flags);
		break;
	case 0:
		set_bit(BTRFS_INODE_BTREE_LOG1_ERR, &btree_ino->runtime_flags);
		break;
	case 1:
		set_bit(BTRFS_INODE_BTREE_LOG2_ERR, &btree_ino->runtime_flags);
		break;
	default:
		BUG(); /* unexpected, logic error */
	}
}

3693
static void end_bio_extent_buffer_writepage(struct bio *bio)
3694
{
3695
	struct bio_vec *bvec;
3696
	struct extent_buffer *eb;
3697
	int i, done;
3698

3699
	bio_for_each_segment_all(bvec, bio, i) {
3700 3701 3702 3703 3704 3705
		struct page *page = bvec->bv_page;

		eb = (struct extent_buffer *)page->private;
		BUG_ON(!eb);
		done = atomic_dec_and_test(&eb->io_pages);

3706 3707
		if (bio->bi_error ||
		    test_bit(EXTENT_BUFFER_WRITE_ERR, &eb->bflags)) {
3708
			ClearPageUptodate(page);
3709
			set_btree_ioerr(page);
3710 3711 3712 3713 3714 3715 3716 3717
		}

		end_page_writeback(page);

		if (!done)
			continue;

		end_extent_buffer_writeback(eb);
3718
	}
3719 3720 3721 3722

	bio_put(bio);
}

3723
static noinline_for_stack int write_one_eb(struct extent_buffer *eb,
3724 3725 3726 3727 3728
			struct btrfs_fs_info *fs_info,
			struct writeback_control *wbc,
			struct extent_page_data *epd)
{
	struct block_device *bdev = fs_info->fs_devices->latest_bdev;
3729
	struct extent_io_tree *tree = &BTRFS_I(fs_info->btree_inode)->io_tree;
3730 3731
	u64 offset = eb->start;
	unsigned long i, num_pages;
3732
	unsigned long bio_flags = 0;
3733
	int write_flags = (epd->sync_io ? WRITE_SYNC : 0) | REQ_META;
3734
	int ret = 0;
3735

3736
	clear_bit(EXTENT_BUFFER_WRITE_ERR, &eb->bflags);
3737 3738
	num_pages = num_extent_pages(eb->start, eb->len);
	atomic_set(&eb->io_pages, num_pages);
3739 3740 3741
	if (btrfs_header_owner(eb) == BTRFS_TREE_LOG_OBJECTID)
		bio_flags = EXTENT_BIO_TREE_LOG;

3742
	for (i = 0; i < num_pages; i++) {
3743
		struct page *p = eb->pages[i];
3744 3745 3746

		clear_page_dirty_for_io(p);
		set_page_writeback(p);
3747 3748 3749 3750
		ret = submit_extent_page(REQ_OP_WRITE, write_flags, tree, wbc,
					 p, offset >> 9, PAGE_SIZE, 0, bdev,
					 &epd->bio, -1,
					 end_bio_extent_buffer_writepage,
3751
					 0, epd->bio_flags, bio_flags, false);
3752
		epd->bio_flags = bio_flags;
3753
		if (ret) {
3754
			set_btree_ioerr(p);
3755
			end_page_writeback(p);
3756 3757 3758 3759 3760
			if (atomic_sub_and_test(num_pages - i, &eb->io_pages))
				end_extent_buffer_writeback(eb);
			ret = -EIO;
			break;
		}
3761
		offset += PAGE_SIZE;
3762 3763 3764 3765 3766 3767
		update_nr_written(p, wbc, 1);
		unlock_page(p);
	}

	if (unlikely(ret)) {
		for (; i < num_pages; i++) {
3768
			struct page *p = eb->pages[i];
3769
			clear_page_dirty_for_io(p);
3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787
			unlock_page(p);
		}
	}

	return ret;
}

int btree_write_cache_pages(struct address_space *mapping,
				   struct writeback_control *wbc)
{
	struct extent_io_tree *tree = &BTRFS_I(mapping->host)->io_tree;
	struct btrfs_fs_info *fs_info = BTRFS_I(mapping->host)->root->fs_info;
	struct extent_buffer *eb, *prev_eb = NULL;
	struct extent_page_data epd = {
		.bio = NULL,
		.tree = tree,
		.extent_locked = 0,
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
3788
		.bio_flags = 0,
3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804
	};
	int ret = 0;
	int done = 0;
	int nr_to_write_done = 0;
	struct pagevec pvec;
	int nr_pages;
	pgoff_t index;
	pgoff_t end;		/* Inclusive */
	int scanned = 0;
	int tag;

	pagevec_init(&pvec, 0);
	if (wbc->range_cyclic) {
		index = mapping->writeback_index; /* Start from prev offset */
		end = -1;
	} else {
3805 3806
		index = wbc->range_start >> PAGE_SHIFT;
		end = wbc->range_end >> PAGE_SHIFT;
3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826 3827 3828 3829 3830 3831 3832
		scanned = 1;
	}
	if (wbc->sync_mode == WB_SYNC_ALL)
		tag = PAGECACHE_TAG_TOWRITE;
	else
		tag = PAGECACHE_TAG_DIRTY;
retry:
	if (wbc->sync_mode == WB_SYNC_ALL)
		tag_pages_for_writeback(mapping, index, end);
	while (!done && !nr_to_write_done && (index <= end) &&
	       (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag,
			min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
		unsigned i;

		scanned = 1;
		for (i = 0; i < nr_pages; i++) {
			struct page *page = pvec.pages[i];

			if (!PagePrivate(page))
				continue;

			if (!wbc->range_cyclic && page->index > end) {
				done = 1;
				break;
			}

3833 3834 3835 3836 3837 3838
			spin_lock(&mapping->private_lock);
			if (!PagePrivate(page)) {
				spin_unlock(&mapping->private_lock);
				continue;
			}

3839
			eb = (struct extent_buffer *)page->private;
3840 3841 3842 3843 3844 3845

			/*
			 * Shouldn't happen and normally this would be a BUG_ON
			 * but no sense in crashing the users box for something
			 * we can survive anyway.
			 */
3846
			if (WARN_ON(!eb)) {
3847
				spin_unlock(&mapping->private_lock);
3848 3849 3850
				continue;
			}

3851 3852
			if (eb == prev_eb) {
				spin_unlock(&mapping->private_lock);
3853
				continue;
3854
			}
3855

3856 3857 3858
			ret = atomic_inc_not_zero(&eb->refs);
			spin_unlock(&mapping->private_lock);
			if (!ret)
3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898
				continue;

			prev_eb = eb;
			ret = lock_extent_buffer_for_io(eb, fs_info, &epd);
			if (!ret) {
				free_extent_buffer(eb);
				continue;
			}

			ret = write_one_eb(eb, fs_info, wbc, &epd);
			if (ret) {
				done = 1;
				free_extent_buffer(eb);
				break;
			}
			free_extent_buffer(eb);

			/*
			 * the filesystem may choose to bump up nr_to_write.
			 * We have to make sure to honor the new nr_to_write
			 * at any time
			 */
			nr_to_write_done = wbc->nr_to_write <= 0;
		}
		pagevec_release(&pvec);
		cond_resched();
	}
	if (!scanned && !done) {
		/*
		 * We hit the last page and there is more work to be done: wrap
		 * back to the start of the file
		 */
		scanned = 1;
		index = 0;
		goto retry;
	}
	flush_write_bio(&epd);
	return ret;
}

3899
/**
C
Chris Mason 已提交
3900
 * write_cache_pages - walk the list of dirty pages of the given address space and write all of them.
3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911 3912 3913
 * @mapping: address space structure to write
 * @wbc: subtract the number of written pages from *@wbc->nr_to_write
 * @writepage: function called for each page
 * @data: data passed to writepage function
 *
 * If a page is already under I/O, write_cache_pages() skips it, even
 * if it's dirty.  This is desirable behaviour for memory-cleaning writeback,
 * but it is INCORRECT for data-integrity system calls such as fsync().  fsync()
 * and msync() need to guarantee that all the data which was dirty at the time
 * the call was made get new I/O started against them.  If wbc->sync_mode is
 * WB_SYNC_ALL then we were called for data integrity and we must wait for
 * existing IO to complete.
 */
3914
static int extent_write_cache_pages(struct extent_io_tree *tree,
C
Chris Mason 已提交
3915 3916
			     struct address_space *mapping,
			     struct writeback_control *wbc,
C
Chris Mason 已提交
3917 3918
			     writepage_t writepage, void *data,
			     void (*flush_fn)(void *))
3919
{
3920
	struct inode *inode = mapping->host;
3921 3922
	int ret = 0;
	int done = 0;
3923
	int nr_to_write_done = 0;
3924 3925 3926 3927
	struct pagevec pvec;
	int nr_pages;
	pgoff_t index;
	pgoff_t end;		/* Inclusive */
3928 3929
	pgoff_t done_index;
	int range_whole = 0;
3930
	int scanned = 0;
3931
	int tag;
3932

3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944
	/*
	 * We have to hold onto the inode so that ordered extents can do their
	 * work when the IO finishes.  The alternative to this is failing to add
	 * an ordered extent if the igrab() fails there and that is a huge pain
	 * to deal with, so instead just hold onto the inode throughout the
	 * writepages operation.  If it fails here we are freeing up the inode
	 * anyway and we'd rather not waste our time writing out stuff that is
	 * going to be truncated anyway.
	 */
	if (!igrab(inode))
		return 0;

3945 3946 3947 3948 3949
	pagevec_init(&pvec, 0);
	if (wbc->range_cyclic) {
		index = mapping->writeback_index; /* Start from prev offset */
		end = -1;
	} else {
3950 3951
		index = wbc->range_start >> PAGE_SHIFT;
		end = wbc->range_end >> PAGE_SHIFT;
3952 3953
		if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
			range_whole = 1;
3954 3955
		scanned = 1;
	}
3956 3957 3958 3959
	if (wbc->sync_mode == WB_SYNC_ALL)
		tag = PAGECACHE_TAG_TOWRITE;
	else
		tag = PAGECACHE_TAG_DIRTY;
3960
retry:
3961 3962
	if (wbc->sync_mode == WB_SYNC_ALL)
		tag_pages_for_writeback(mapping, index, end);
3963
	done_index = index;
3964
	while (!done && !nr_to_write_done && (index <= end) &&
3965 3966
	       (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag,
			min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
3967 3968 3969 3970 3971 3972
		unsigned i;

		scanned = 1;
		for (i = 0; i < nr_pages; i++) {
			struct page *page = pvec.pages[i];

3973
			done_index = page->index;
3974 3975 3976 3977 3978 3979 3980
			/*
			 * At this point we hold neither mapping->tree_lock nor
			 * lock on the page itself: the page may be truncated or
			 * invalidated (changing page->mapping to NULL), or even
			 * swizzled back from swapper_space to tmpfs file
			 * mapping
			 */
3981 3982 3983
			if (!trylock_page(page)) {
				flush_fn(data);
				lock_page(page);
3984
			}
3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996

			if (unlikely(page->mapping != mapping)) {
				unlock_page(page);
				continue;
			}

			if (!wbc->range_cyclic && page->index > end) {
				done = 1;
				unlock_page(page);
				continue;
			}

C
Chris Mason 已提交
3997
			if (wbc->sync_mode != WB_SYNC_NONE) {
3998 3999
				if (PageWriteback(page))
					flush_fn(data);
4000
				wait_on_page_writeback(page);
C
Chris Mason 已提交
4001
			}
4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014

			if (PageWriteback(page) ||
			    !clear_page_dirty_for_io(page)) {
				unlock_page(page);
				continue;
			}

			ret = (*writepage)(page, wbc, data);

			if (unlikely(ret == AOP_WRITEPAGE_ACTIVATE)) {
				unlock_page(page);
				ret = 0;
			}
4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028
			if (ret < 0) {
				/*
				 * done_index is set past this page,
				 * so media errors will not choke
				 * background writeout for the entire
				 * file. This has consequences for
				 * range_cyclic semantics (ie. it may
				 * not be suitable for data integrity
				 * writeout).
				 */
				done_index = page->index + 1;
				done = 1;
				break;
			}
4029 4030 4031 4032 4033 4034 4035

			/*
			 * the filesystem may choose to bump up nr_to_write.
			 * We have to make sure to honor the new nr_to_write
			 * at any time
			 */
			nr_to_write_done = wbc->nr_to_write <= 0;
4036 4037 4038 4039
		}
		pagevec_release(&pvec);
		cond_resched();
	}
4040
	if (!scanned && !done) {
4041 4042 4043 4044 4045 4046 4047 4048
		/*
		 * We hit the last page and there is more work to be done: wrap
		 * back to the start of the file
		 */
		scanned = 1;
		index = 0;
		goto retry;
	}
4049 4050 4051 4052

	if (wbc->range_cyclic || (wbc->nr_to_write > 0 && range_whole))
		mapping->writeback_index = done_index;

4053
	btrfs_add_delayed_iput(inode);
4054
	return ret;
4055 4056
}

4057
static void flush_epd_write_bio(struct extent_page_data *epd)
C
Chris Mason 已提交
4058 4059
{
	if (epd->bio) {
4060 4061
		int ret;

4062 4063
		bio_set_op_attrs(epd->bio, REQ_OP_WRITE,
				 epd->sync_io ? WRITE_SYNC : 0);
4064

4065
		ret = submit_one_bio(epd->bio, 0, epd->bio_flags);
4066
		BUG_ON(ret < 0); /* -ENOMEM */
C
Chris Mason 已提交
4067 4068 4069 4070
		epd->bio = NULL;
	}
}

4071 4072 4073 4074 4075 4076
static noinline void flush_write_bio(void *data)
{
	struct extent_page_data *epd = data;
	flush_epd_write_bio(epd);
}

4077 4078 4079 4080 4081 4082 4083 4084 4085
int extent_write_full_page(struct extent_io_tree *tree, struct page *page,
			  get_extent_t *get_extent,
			  struct writeback_control *wbc)
{
	int ret;
	struct extent_page_data epd = {
		.bio = NULL,
		.tree = tree,
		.get_extent = get_extent,
4086
		.extent_locked = 0,
4087
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
4088
		.bio_flags = 0,
4089 4090 4091 4092
	};

	ret = __extent_writepage(page, wbc, &epd);

4093
	flush_epd_write_bio(&epd);
4094 4095 4096
	return ret;
}

4097 4098 4099 4100 4101 4102 4103
int extent_write_locked_range(struct extent_io_tree *tree, struct inode *inode,
			      u64 start, u64 end, get_extent_t *get_extent,
			      int mode)
{
	int ret = 0;
	struct address_space *mapping = inode->i_mapping;
	struct page *page;
4104 4105
	unsigned long nr_pages = (end - start + PAGE_SIZE) >>
		PAGE_SHIFT;
4106 4107 4108 4109 4110 4111

	struct extent_page_data epd = {
		.bio = NULL,
		.tree = tree,
		.get_extent = get_extent,
		.extent_locked = 1,
4112
		.sync_io = mode == WB_SYNC_ALL,
4113
		.bio_flags = 0,
4114 4115 4116 4117 4118 4119 4120 4121
	};
	struct writeback_control wbc_writepages = {
		.sync_mode	= mode,
		.nr_to_write	= nr_pages * 2,
		.range_start	= start,
		.range_end	= end + 1,
	};

C
Chris Mason 已提交
4122
	while (start <= end) {
4123
		page = find_get_page(mapping, start >> PAGE_SHIFT);
4124 4125 4126 4127 4128
		if (clear_page_dirty_for_io(page))
			ret = __extent_writepage(page, &wbc_writepages, &epd);
		else {
			if (tree->ops && tree->ops->writepage_end_io_hook)
				tree->ops->writepage_end_io_hook(page, start,
4129
						 start + PAGE_SIZE - 1,
4130 4131 4132
						 NULL, 1);
			unlock_page(page);
		}
4133 4134
		put_page(page);
		start += PAGE_SIZE;
4135 4136
	}

4137
	flush_epd_write_bio(&epd);
4138 4139
	return ret;
}
4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150

int extent_writepages(struct extent_io_tree *tree,
		      struct address_space *mapping,
		      get_extent_t *get_extent,
		      struct writeback_control *wbc)
{
	int ret = 0;
	struct extent_page_data epd = {
		.bio = NULL,
		.tree = tree,
		.get_extent = get_extent,
4151
		.extent_locked = 0,
4152
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
4153
		.bio_flags = 0,
4154 4155
	};

C
Chris Mason 已提交
4156
	ret = extent_write_cache_pages(tree, mapping, wbc,
C
Chris Mason 已提交
4157 4158
				       __extent_writepage, &epd,
				       flush_write_bio);
4159
	flush_epd_write_bio(&epd);
4160 4161 4162 4163 4164 4165 4166 4167 4168 4169
	return ret;
}

int extent_readpages(struct extent_io_tree *tree,
		     struct address_space *mapping,
		     struct list_head *pages, unsigned nr_pages,
		     get_extent_t get_extent)
{
	struct bio *bio = NULL;
	unsigned page_idx;
C
Chris Mason 已提交
4170
	unsigned long bio_flags = 0;
L
Liu Bo 已提交
4171 4172
	struct page *pagepool[16];
	struct page *page;
4173
	struct extent_map *em_cached = NULL;
L
Liu Bo 已提交
4174
	int nr = 0;
4175
	u64 prev_em_start = (u64)-1;
4176 4177

	for (page_idx = 0; page_idx < nr_pages; page_idx++) {
L
Liu Bo 已提交
4178
		page = list_entry(pages->prev, struct page, lru);
4179 4180 4181

		prefetchw(&page->flags);
		list_del(&page->lru);
L
Liu Bo 已提交
4182
		if (add_to_page_cache_lru(page, mapping,
4183
					page->index, GFP_NOFS)) {
4184
			put_page(page);
L
Liu Bo 已提交
4185
			continue;
4186
		}
L
Liu Bo 已提交
4187 4188 4189 4190

		pagepool[nr++] = page;
		if (nr < ARRAY_SIZE(pagepool))
			continue;
4191
		__extent_readpages(tree, pagepool, nr, get_extent, &em_cached,
4192
				   &bio, 0, &bio_flags, &prev_em_start);
L
Liu Bo 已提交
4193
		nr = 0;
4194
	}
4195
	if (nr)
4196
		__extent_readpages(tree, pagepool, nr, get_extent, &em_cached,
4197
				   &bio, 0, &bio_flags, &prev_em_start);
L
Liu Bo 已提交
4198

4199 4200 4201
	if (em_cached)
		free_extent_map(em_cached);

4202 4203
	BUG_ON(!list_empty(pages));
	if (bio)
4204
		return submit_one_bio(bio, 0, bio_flags);
4205 4206 4207 4208 4209 4210 4211 4212 4213 4214 4215
	return 0;
}

/*
 * basic invalidatepage code, this waits on any locked or writeback
 * ranges corresponding to the page, and then deletes any extent state
 * records from the tree
 */
int extent_invalidatepage(struct extent_io_tree *tree,
			  struct page *page, unsigned long offset)
{
4216
	struct extent_state *cached_state = NULL;
M
Miao Xie 已提交
4217
	u64 start = page_offset(page);
4218
	u64 end = start + PAGE_SIZE - 1;
4219 4220
	size_t blocksize = page->mapping->host->i_sb->s_blocksize;

4221
	start += ALIGN(offset, blocksize);
4222 4223 4224
	if (start > end)
		return 0;

4225
	lock_extent_bits(tree, start, end, &cached_state);
4226
	wait_on_page_writeback(page);
4227
	clear_extent_bit(tree, start, end,
4228 4229
			 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
			 EXTENT_DO_ACCOUNTING,
4230
			 1, 1, &cached_state, GFP_NOFS);
4231 4232 4233
	return 0;
}

4234 4235 4236 4237 4238
/*
 * a helper for releasepage, this tests for areas of the page that
 * are locked or under IO and drops the related state bits if it is safe
 * to drop the page.
 */
4239 4240 4241
static int try_release_extent_state(struct extent_map_tree *map,
				    struct extent_io_tree *tree,
				    struct page *page, gfp_t mask)
4242
{
M
Miao Xie 已提交
4243
	u64 start = page_offset(page);
4244
	u64 end = start + PAGE_SIZE - 1;
4245 4246
	int ret = 1;

4247
	if (test_range_bit(tree, start, end,
4248
			   EXTENT_IOBITS, 0, NULL))
4249 4250 4251 4252
		ret = 0;
	else {
		if ((mask & GFP_NOFS) == GFP_NOFS)
			mask = GFP_NOFS;
4253 4254 4255 4256
		/*
		 * at this point we can safely clear everything except the
		 * locked bit and the nodatasum bit
		 */
4257
		ret = clear_extent_bit(tree, start, end,
4258 4259
				 ~(EXTENT_LOCKED | EXTENT_NODATASUM),
				 0, 0, NULL, mask);
4260 4261 4262 4263 4264 4265 4266 4267

		/* if clear_extent_bit failed for enomem reasons,
		 * we can't allow the release to continue.
		 */
		if (ret < 0)
			ret = 0;
		else
			ret = 1;
4268 4269 4270 4271
	}
	return ret;
}

4272 4273 4274 4275 4276 4277
/*
 * a helper for releasepage.  As long as there are no locked extents
 * in the range corresponding to the page, both state records and extent
 * map records are removed
 */
int try_release_extent_mapping(struct extent_map_tree *map,
4278 4279
			       struct extent_io_tree *tree, struct page *page,
			       gfp_t mask)
4280 4281
{
	struct extent_map *em;
M
Miao Xie 已提交
4282
	u64 start = page_offset(page);
4283
	u64 end = start + PAGE_SIZE - 1;
4284

4285
	if (gfpflags_allow_blocking(mask) &&
4286
	    page->mapping->host->i_size > SZ_16M) {
4287
		u64 len;
4288
		while (start <= end) {
4289
			len = end - start + 1;
4290
			write_lock(&map->lock);
4291
			em = lookup_extent_mapping(map, start, len);
4292
			if (!em) {
4293
				write_unlock(&map->lock);
4294 4295
				break;
			}
4296 4297
			if (test_bit(EXTENT_FLAG_PINNED, &em->flags) ||
			    em->start != start) {
4298
				write_unlock(&map->lock);
4299 4300 4301 4302 4303
				free_extent_map(em);
				break;
			}
			if (!test_range_bit(tree, em->start,
					    extent_map_end(em) - 1,
4304
					    EXTENT_LOCKED | EXTENT_WRITEBACK,
4305
					    0, NULL)) {
4306 4307 4308 4309 4310
				remove_extent_mapping(map, em);
				/* once for the rb tree */
				free_extent_map(em);
			}
			start = extent_map_end(em);
4311
			write_unlock(&map->lock);
4312 4313

			/* once for us */
4314 4315 4316
			free_extent_map(em);
		}
	}
4317
	return try_release_extent_state(map, tree, page, mask);
4318 4319
}

4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334 4335
/*
 * helper function for fiemap, which doesn't want to see any holes.
 * This maps until we find something past 'last'
 */
static struct extent_map *get_extent_skip_holes(struct inode *inode,
						u64 offset,
						u64 last,
						get_extent_t *get_extent)
{
	u64 sectorsize = BTRFS_I(inode)->root->sectorsize;
	struct extent_map *em;
	u64 len;

	if (offset >= last)
		return NULL;

4336
	while (1) {
4337 4338 4339
		len = last - offset;
		if (len == 0)
			break;
4340
		len = ALIGN(len, sectorsize);
4341
		em = get_extent(inode, NULL, 0, offset, len, 0);
4342
		if (IS_ERR_OR_NULL(em))
4343 4344 4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355 4356 4357 4358 4359
			return em;

		/* if this isn't a hole return it */
		if (!test_bit(EXTENT_FLAG_VACANCY, &em->flags) &&
		    em->block_start != EXTENT_MAP_HOLE) {
			return em;
		}

		/* this is a hole, advance to the next extent */
		offset = extent_map_end(em);
		free_extent_map(em);
		if (offset >= last)
			break;
	}
	return NULL;
}

Y
Yehuda Sadeh 已提交
4360 4361 4362
int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
		__u64 start, __u64 len, get_extent_t *get_extent)
{
J
Josef Bacik 已提交
4363
	int ret = 0;
Y
Yehuda Sadeh 已提交
4364 4365 4366
	u64 off = start;
	u64 max = start + len;
	u32 flags = 0;
J
Josef Bacik 已提交
4367 4368
	u32 found_type;
	u64 last;
4369
	u64 last_for_get_extent = 0;
Y
Yehuda Sadeh 已提交
4370
	u64 disko = 0;
4371
	u64 isize = i_size_read(inode);
J
Josef Bacik 已提交
4372
	struct btrfs_key found_key;
Y
Yehuda Sadeh 已提交
4373
	struct extent_map *em = NULL;
4374
	struct extent_state *cached_state = NULL;
J
Josef Bacik 已提交
4375
	struct btrfs_path *path;
4376
	struct btrfs_root *root = BTRFS_I(inode)->root;
Y
Yehuda Sadeh 已提交
4377
	int end = 0;
4378 4379 4380
	u64 em_start = 0;
	u64 em_len = 0;
	u64 em_end = 0;
Y
Yehuda Sadeh 已提交
4381 4382 4383 4384

	if (len == 0)
		return -EINVAL;

J
Josef Bacik 已提交
4385 4386 4387 4388 4389
	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;
	path->leave_spinning = 1;

4390 4391
	start = round_down(start, BTRFS_I(inode)->root->sectorsize);
	len = round_up(max, BTRFS_I(inode)->root->sectorsize) - start;
4392

4393 4394 4395 4396
	/*
	 * lookup the last file extent.  We're not using i_size here
	 * because there might be preallocation past i_size
	 */
4397 4398
	ret = btrfs_lookup_file_extent(NULL, root, path, btrfs_ino(inode), -1,
				       0);
J
Josef Bacik 已提交
4399 4400 4401
	if (ret < 0) {
		btrfs_free_path(path);
		return ret;
4402 4403 4404 4405
	} else {
		WARN_ON(!ret);
		if (ret == 1)
			ret = 0;
J
Josef Bacik 已提交
4406
	}
4407

J
Josef Bacik 已提交
4408 4409
	path->slots[0]--;
	btrfs_item_key_to_cpu(path->nodes[0], &found_key, path->slots[0]);
4410
	found_type = found_key.type;
J
Josef Bacik 已提交
4411

4412
	/* No extents, but there might be delalloc bits */
L
Li Zefan 已提交
4413
	if (found_key.objectid != btrfs_ino(inode) ||
J
Josef Bacik 已提交
4414
	    found_type != BTRFS_EXTENT_DATA_KEY) {
4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425
		/* have to trust i_size as the end */
		last = (u64)-1;
		last_for_get_extent = isize;
	} else {
		/*
		 * remember the start of the last extent.  There are a
		 * bunch of different factors that go into the length of the
		 * extent, so its much less complex to remember where it started
		 */
		last = found_key.offset;
		last_for_get_extent = last + 1;
J
Josef Bacik 已提交
4426
	}
4427
	btrfs_release_path(path);
J
Josef Bacik 已提交
4428

4429 4430 4431 4432 4433 4434 4435 4436 4437 4438
	/*
	 * we might have some extents allocated but more delalloc past those
	 * extents.  so, we trust isize unless the start of the last extent is
	 * beyond isize
	 */
	if (last < isize) {
		last = (u64)-1;
		last_for_get_extent = isize;
	}

4439
	lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + len - 1,
4440
			 &cached_state);
4441

4442
	em = get_extent_skip_holes(inode, start, last_for_get_extent,
4443
				   get_extent);
Y
Yehuda Sadeh 已提交
4444 4445 4446 4447 4448 4449
	if (!em)
		goto out;
	if (IS_ERR(em)) {
		ret = PTR_ERR(em);
		goto out;
	}
J
Josef Bacik 已提交
4450

Y
Yehuda Sadeh 已提交
4451
	while (!end) {
4452
		u64 offset_in_extent = 0;
4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464

		/* break if the extent we found is outside the range */
		if (em->start >= max || extent_map_end(em) < off)
			break;

		/*
		 * get_extent may return an extent that starts before our
		 * requested range.  We have to make sure the ranges
		 * we return to fiemap always move forward and don't
		 * overlap, so adjust the offsets here
		 */
		em_start = max(em->start, off);
Y
Yehuda Sadeh 已提交
4465

4466 4467
		/*
		 * record the offset from the start of the extent
4468 4469 4470
		 * for adjusting the disk offset below.  Only do this if the
		 * extent isn't compressed since our in ram offset may be past
		 * what we have actually allocated on disk.
4471
		 */
4472 4473
		if (!test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
			offset_in_extent = em_start - em->start;
4474
		em_end = extent_map_end(em);
4475
		em_len = em_end - em_start;
Y
Yehuda Sadeh 已提交
4476 4477 4478
		disko = 0;
		flags = 0;

4479 4480 4481 4482 4483 4484 4485
		/*
		 * bump off for our next call to get_extent
		 */
		off = extent_map_end(em);
		if (off >= max)
			end = 1;

4486
		if (em->block_start == EXTENT_MAP_LAST_BYTE) {
Y
Yehuda Sadeh 已提交
4487 4488
			end = 1;
			flags |= FIEMAP_EXTENT_LAST;
4489
		} else if (em->block_start == EXTENT_MAP_INLINE) {
Y
Yehuda Sadeh 已提交
4490 4491
			flags |= (FIEMAP_EXTENT_DATA_INLINE |
				  FIEMAP_EXTENT_NOT_ALIGNED);
4492
		} else if (em->block_start == EXTENT_MAP_DELALLOC) {
Y
Yehuda Sadeh 已提交
4493 4494
			flags |= (FIEMAP_EXTENT_DELALLOC |
				  FIEMAP_EXTENT_UNKNOWN);
4495 4496 4497
		} else if (fieinfo->fi_extents_max) {
			u64 bytenr = em->block_start -
				(em->start - em->orig_start);
4498

4499
			disko = em->block_start + offset_in_extent;
4500 4501 4502 4503

			/*
			 * As btrfs supports shared space, this information
			 * can be exported to userspace tools via
4504 4505 4506
			 * flag FIEMAP_EXTENT_SHARED.  If fi_extents_max == 0
			 * then we're just getting a count and we can skip the
			 * lookup stuff.
4507
			 */
4508 4509 4510 4511
			ret = btrfs_check_shared(NULL, root->fs_info,
						 root->objectid,
						 btrfs_ino(inode), bytenr);
			if (ret < 0)
4512
				goto out_free;
4513
			if (ret)
4514
				flags |= FIEMAP_EXTENT_SHARED;
4515
			ret = 0;
Y
Yehuda Sadeh 已提交
4516 4517 4518
		}
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
			flags |= FIEMAP_EXTENT_ENCODED;
4519 4520
		if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			flags |= FIEMAP_EXTENT_UNWRITTEN;
Y
Yehuda Sadeh 已提交
4521 4522 4523

		free_extent_map(em);
		em = NULL;
4524 4525
		if ((em_start >= last) || em_len == (u64)-1 ||
		   (last == (u64)-1 && isize <= em_end)) {
Y
Yehuda Sadeh 已提交
4526 4527 4528 4529
			flags |= FIEMAP_EXTENT_LAST;
			end = 1;
		}

4530 4531 4532 4533 4534 4535 4536 4537
		/* now scan forward to see if this is really the last extent. */
		em = get_extent_skip_holes(inode, off, last_for_get_extent,
					   get_extent);
		if (IS_ERR(em)) {
			ret = PTR_ERR(em);
			goto out;
		}
		if (!em) {
J
Josef Bacik 已提交
4538 4539 4540
			flags |= FIEMAP_EXTENT_LAST;
			end = 1;
		}
4541 4542
		ret = fiemap_fill_next_extent(fieinfo, em_start, disko,
					      em_len, flags);
4543 4544 4545
		if (ret) {
			if (ret == 1)
				ret = 0;
4546
			goto out_free;
4547
		}
Y
Yehuda Sadeh 已提交
4548 4549 4550 4551
	}
out_free:
	free_extent_map(em);
out:
4552
	btrfs_free_path(path);
L
Liu Bo 已提交
4553
	unlock_extent_cached(&BTRFS_I(inode)->io_tree, start, start + len - 1,
4554
			     &cached_state, GFP_NOFS);
Y
Yehuda Sadeh 已提交
4555 4556 4557
	return ret;
}

4558 4559
static void __free_extent_buffer(struct extent_buffer *eb)
{
4560
	btrfs_leak_debug_del(&eb->leak_list);
4561 4562 4563
	kmem_cache_free(extent_buffer_cache, eb);
}

4564
int extent_buffer_under_io(struct extent_buffer *eb)
4565 4566 4567 4568 4569 4570 4571 4572 4573
{
	return (atomic_read(&eb->io_pages) ||
		test_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags) ||
		test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
}

/*
 * Helper for releasing extent buffer page.
 */
4574
static void btrfs_release_extent_buffer_page(struct extent_buffer *eb)
4575 4576 4577 4578 4579 4580 4581
{
	unsigned long index;
	struct page *page;
	int mapped = !test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags);

	BUG_ON(extent_buffer_under_io(eb));

4582 4583
	index = num_extent_pages(eb->start, eb->len);
	if (index == 0)
4584 4585 4586 4587
		return;

	do {
		index--;
4588
		page = eb->pages[index];
4589 4590 4591
		if (!page)
			continue;
		if (mapped)
4592
			spin_lock(&page->mapping->private_lock);
4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604
		/*
		 * We do this since we'll remove the pages after we've
		 * removed the eb from the radix tree, so we could race
		 * and have this page now attached to the new eb.  So
		 * only clear page_private if it's still connected to
		 * this eb.
		 */
		if (PagePrivate(page) &&
		    page->private == (unsigned long)eb) {
			BUG_ON(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
			BUG_ON(PageDirty(page));
			BUG_ON(PageWriteback(page));
4605
			/*
4606 4607
			 * We need to make sure we haven't be attached
			 * to a new eb.
4608
			 */
4609 4610 4611
			ClearPagePrivate(page);
			set_page_private(page, 0);
			/* One for the page private */
4612
			put_page(page);
4613
		}
4614 4615 4616 4617

		if (mapped)
			spin_unlock(&page->mapping->private_lock);

4618
		/* One for when we allocated the page */
4619
		put_page(page);
4620
	} while (index != 0);
4621 4622 4623 4624 4625 4626 4627
}

/*
 * Helper for releasing the extent buffer.
 */
static inline void btrfs_release_extent_buffer(struct extent_buffer *eb)
{
4628
	btrfs_release_extent_buffer_page(eb);
4629 4630 4631
	__free_extent_buffer(eb);
}

4632 4633
static struct extent_buffer *
__alloc_extent_buffer(struct btrfs_fs_info *fs_info, u64 start,
4634
		      unsigned long len)
4635 4636 4637
{
	struct extent_buffer *eb = NULL;

4638
	eb = kmem_cache_zalloc(extent_buffer_cache, GFP_NOFS|__GFP_NOFAIL);
4639 4640
	eb->start = start;
	eb->len = len;
4641
	eb->fs_info = fs_info;
4642
	eb->bflags = 0;
4643 4644 4645 4646 4647 4648 4649
	rwlock_init(&eb->lock);
	atomic_set(&eb->write_locks, 0);
	atomic_set(&eb->read_locks, 0);
	atomic_set(&eb->blocking_readers, 0);
	atomic_set(&eb->blocking_writers, 0);
	atomic_set(&eb->spinning_readers, 0);
	atomic_set(&eb->spinning_writers, 0);
4650
	eb->lock_nested = 0;
4651 4652
	init_waitqueue_head(&eb->write_lock_wq);
	init_waitqueue_head(&eb->read_lock_wq);
4653

4654 4655
	btrfs_leak_debug_add(&eb->leak_list, &buffers);

4656
	spin_lock_init(&eb->refs_lock);
4657
	atomic_set(&eb->refs, 1);
4658
	atomic_set(&eb->io_pages, 0);
4659

4660 4661 4662 4663 4664 4665
	/*
	 * Sanity checks, currently the maximum is 64k covered by 16x 4k pages
	 */
	BUILD_BUG_ON(BTRFS_MAX_METADATA_BLOCKSIZE
		> MAX_INLINE_EXTENT_BUFFER_SIZE);
	BUG_ON(len > MAX_INLINE_EXTENT_BUFFER_SIZE);
4666 4667 4668 4669

	return eb;
}

4670 4671 4672 4673 4674 4675 4676
struct extent_buffer *btrfs_clone_extent_buffer(struct extent_buffer *src)
{
	unsigned long i;
	struct page *p;
	struct extent_buffer *new;
	unsigned long num_pages = num_extent_pages(src->start, src->len);

4677
	new = __alloc_extent_buffer(src->fs_info, src->start, src->len);
4678 4679 4680 4681
	if (new == NULL)
		return NULL;

	for (i = 0; i < num_pages; i++) {
4682
		p = alloc_page(GFP_NOFS);
4683 4684 4685 4686
		if (!p) {
			btrfs_release_extent_buffer(new);
			return NULL;
		}
4687 4688 4689 4690 4691 4692 4693 4694 4695 4696 4697 4698 4699
		attach_extent_buffer_page(new, p);
		WARN_ON(PageDirty(p));
		SetPageUptodate(p);
		new->pages[i] = p;
	}

	copy_extent_buffer(new, src, 0, 0, src->len);
	set_bit(EXTENT_BUFFER_UPTODATE, &new->bflags);
	set_bit(EXTENT_BUFFER_DUMMY, &new->bflags);

	return new;
}

4700 4701
struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
						  u64 start, unsigned long len)
4702 4703
{
	struct extent_buffer *eb;
4704
	unsigned long num_pages;
4705 4706
	unsigned long i;

4707
	num_pages = num_extent_pages(start, len);
4708 4709

	eb = __alloc_extent_buffer(fs_info, start, len);
4710 4711 4712 4713
	if (!eb)
		return NULL;

	for (i = 0; i < num_pages; i++) {
4714
		eb->pages[i] = alloc_page(GFP_NOFS);
4715 4716 4717 4718 4719 4720 4721 4722 4723
		if (!eb->pages[i])
			goto err;
	}
	set_extent_buffer_uptodate(eb);
	btrfs_set_header_nritems(eb, 0);
	set_bit(EXTENT_BUFFER_DUMMY, &eb->bflags);

	return eb;
err:
4724 4725
	for (; i > 0; i--)
		__free_page(eb->pages[i - 1]);
4726 4727 4728 4729
	__free_extent_buffer(eb);
	return NULL;
}

4730 4731 4732 4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743 4744 4745 4746 4747
struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
						u64 start)
{
	unsigned long len;

	if (!fs_info) {
		/*
		 * Called only from tests that don't always have a fs_info
		 * available, but we know that nodesize is 4096
		 */
		len = 4096;
	} else {
		len = fs_info->tree_root->nodesize;
	}

	return __alloc_dummy_extent_buffer(fs_info, start, len);
}

4748 4749
static void check_buffer_tree_ref(struct extent_buffer *eb)
{
4750
	int refs;
4751 4752 4753 4754 4755 4756 4757 4758 4759 4760 4761 4762 4763 4764 4765 4766 4767 4768 4769 4770
	/* the ref bit is tricky.  We have to make sure it is set
	 * if we have the buffer dirty.   Otherwise the
	 * code to free a buffer can end up dropping a dirty
	 * page
	 *
	 * Once the ref bit is set, it won't go away while the
	 * buffer is dirty or in writeback, and it also won't
	 * go away while we have the reference count on the
	 * eb bumped.
	 *
	 * We can't just set the ref bit without bumping the
	 * ref on the eb because free_extent_buffer might
	 * see the ref bit and try to clear it.  If this happens
	 * free_extent_buffer might end up dropping our original
	 * ref by mistake and freeing the page before we are able
	 * to add one more ref.
	 *
	 * So bump the ref count first, then set the bit.  If someone
	 * beat us to it, drop the ref we added.
	 */
4771 4772 4773 4774
	refs = atomic_read(&eb->refs);
	if (refs >= 2 && test_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
		return;

4775 4776
	spin_lock(&eb->refs_lock);
	if (!test_and_set_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
4777
		atomic_inc(&eb->refs);
4778
	spin_unlock(&eb->refs_lock);
4779 4780
}

4781 4782
static void mark_extent_buffer_accessed(struct extent_buffer *eb,
		struct page *accessed)
4783 4784 4785
{
	unsigned long num_pages, i;

4786 4787
	check_buffer_tree_ref(eb);

4788 4789
	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = 0; i < num_pages; i++) {
4790 4791
		struct page *p = eb->pages[i];

4792 4793
		if (p != accessed)
			mark_page_accessed(p);
4794 4795 4796
	}
}

4797 4798
struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
					 u64 start)
4799 4800 4801 4802
{
	struct extent_buffer *eb;

	rcu_read_lock();
4803
	eb = radix_tree_lookup(&fs_info->buffer_radix,
4804
			       start >> PAGE_SHIFT);
4805 4806
	if (eb && atomic_inc_not_zero(&eb->refs)) {
		rcu_read_unlock();
4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 4817 4818 4819 4820 4821 4822 4823 4824 4825
		/*
		 * Lock our eb's refs_lock to avoid races with
		 * free_extent_buffer. When we get our eb it might be flagged
		 * with EXTENT_BUFFER_STALE and another task running
		 * free_extent_buffer might have seen that flag set,
		 * eb->refs == 2, that the buffer isn't under IO (dirty and
		 * writeback flags not set) and it's still in the tree (flag
		 * EXTENT_BUFFER_TREE_REF set), therefore being in the process
		 * of decrementing the extent buffer's reference count twice.
		 * So here we could race and increment the eb's reference count,
		 * clear its stale flag, mark it as dirty and drop our reference
		 * before the other task finishes executing free_extent_buffer,
		 * which would later result in an attempt to free an extent
		 * buffer that is dirty.
		 */
		if (test_bit(EXTENT_BUFFER_STALE, &eb->bflags)) {
			spin_lock(&eb->refs_lock);
			spin_unlock(&eb->refs_lock);
		}
4826
		mark_extent_buffer_accessed(eb, NULL);
4827 4828 4829 4830 4831 4832 4833
		return eb;
	}
	rcu_read_unlock();

	return NULL;
}

4834 4835
#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
4836
					       u64 start)
4837 4838 4839 4840 4841 4842 4843
{
	struct extent_buffer *eb, *exists = NULL;
	int ret;

	eb = find_extent_buffer(fs_info, start);
	if (eb)
		return eb;
4844
	eb = alloc_dummy_extent_buffer(fs_info, start);
4845 4846 4847 4848
	if (!eb)
		return NULL;
	eb->fs_info = fs_info;
again:
4849
	ret = radix_tree_preload(GFP_NOFS);
4850 4851 4852 4853
	if (ret)
		goto free_eb;
	spin_lock(&fs_info->buffer_lock);
	ret = radix_tree_insert(&fs_info->buffer_radix,
4854
				start >> PAGE_SHIFT, eb);
4855 4856 4857 4858 4859 4860 4861 4862 4863 4864 4865 4866 4867 4868 4869 4870 4871 4872 4873 4874 4875 4876 4877 4878 4879 4880
	spin_unlock(&fs_info->buffer_lock);
	radix_tree_preload_end();
	if (ret == -EEXIST) {
		exists = find_extent_buffer(fs_info, start);
		if (exists)
			goto free_eb;
		else
			goto again;
	}
	check_buffer_tree_ref(eb);
	set_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags);

	/*
	 * We will free dummy extent buffer's if they come into
	 * free_extent_buffer with a ref count of 2, but if we are using this we
	 * want the buffers to stay in memory until we're done with them, so
	 * bump the ref count again.
	 */
	atomic_inc(&eb->refs);
	return eb;
free_eb:
	btrfs_release_extent_buffer(eb);
	return exists;
}
#endif

4881
struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
4882
					  u64 start)
4883
{
4884
	unsigned long len = fs_info->tree_root->nodesize;
4885 4886
	unsigned long num_pages = num_extent_pages(start, len);
	unsigned long i;
4887
	unsigned long index = start >> PAGE_SHIFT;
4888
	struct extent_buffer *eb;
4889
	struct extent_buffer *exists = NULL;
4890
	struct page *p;
4891
	struct address_space *mapping = fs_info->btree_inode->i_mapping;
4892
	int uptodate = 1;
4893
	int ret;
4894

4895
	eb = find_extent_buffer(fs_info, start);
4896
	if (eb)
4897 4898
		return eb;

4899
	eb = __alloc_extent_buffer(fs_info, start, len);
4900
	if (!eb)
4901 4902
		return NULL;

4903
	for (i = 0; i < num_pages; i++, index++) {
4904
		p = find_or_create_page(mapping, index, GFP_NOFS|__GFP_NOFAIL);
4905
		if (!p)
4906
			goto free_eb;
J
Josef Bacik 已提交
4907 4908 4909 4910 4911 4912 4913 4914 4915 4916 4917 4918 4919 4920

		spin_lock(&mapping->private_lock);
		if (PagePrivate(p)) {
			/*
			 * We could have already allocated an eb for this page
			 * and attached one so lets see if we can get a ref on
			 * the existing eb, and if we can we know it's good and
			 * we can just return that one, else we know we can just
			 * overwrite page->private.
			 */
			exists = (struct extent_buffer *)p->private;
			if (atomic_inc_not_zero(&exists->refs)) {
				spin_unlock(&mapping->private_lock);
				unlock_page(p);
4921
				put_page(p);
4922
				mark_extent_buffer_accessed(exists, p);
J
Josef Bacik 已提交
4923 4924
				goto free_eb;
			}
4925
			exists = NULL;
J
Josef Bacik 已提交
4926

4927
			/*
J
Josef Bacik 已提交
4928 4929 4930 4931
			 * Do this so attach doesn't complain and we need to
			 * drop the ref the old guy had.
			 */
			ClearPagePrivate(p);
4932
			WARN_ON(PageDirty(p));
4933
			put_page(p);
4934
		}
J
Josef Bacik 已提交
4935 4936
		attach_extent_buffer_page(eb, p);
		spin_unlock(&mapping->private_lock);
4937
		WARN_ON(PageDirty(p));
4938
		eb->pages[i] = p;
4939 4940
		if (!PageUptodate(p))
			uptodate = 0;
C
Chris Mason 已提交
4941 4942 4943 4944 4945

		/*
		 * see below about how we avoid a nasty race with release page
		 * and why we unlock later
		 */
4946 4947
	}
	if (uptodate)
4948
		set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
4949
again:
4950
	ret = radix_tree_preload(GFP_NOFS);
4951 4952 4953
	if (ret)
		goto free_eb;

4954 4955
	spin_lock(&fs_info->buffer_lock);
	ret = radix_tree_insert(&fs_info->buffer_radix,
4956
				start >> PAGE_SHIFT, eb);
4957
	spin_unlock(&fs_info->buffer_lock);
4958
	radix_tree_preload_end();
4959
	if (ret == -EEXIST) {
4960
		exists = find_extent_buffer(fs_info, start);
4961 4962 4963
		if (exists)
			goto free_eb;
		else
4964
			goto again;
4965 4966
	}
	/* add one reference for the tree */
4967
	check_buffer_tree_ref(eb);
4968
	set_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags);
C
Chris Mason 已提交
4969 4970 4971 4972 4973 4974 4975 4976 4977 4978

	/*
	 * there is a race where release page may have
	 * tried to find this extent buffer in the radix
	 * but failed.  It will tell the VM it is safe to
	 * reclaim the, and it will clear the page private bit.
	 * We must make sure to set the page private bit properly
	 * after the extent buffer is in the radix tree so
	 * it doesn't get lost
	 */
4979 4980
	SetPageChecked(eb->pages[0]);
	for (i = 1; i < num_pages; i++) {
4981
		p = eb->pages[i];
4982 4983 4984 4985
		ClearPageChecked(p);
		unlock_page(p);
	}
	unlock_page(eb->pages[0]);
4986 4987
	return eb;

4988
free_eb:
4989
	WARN_ON(!atomic_dec_and_test(&eb->refs));
4990 4991 4992 4993
	for (i = 0; i < num_pages; i++) {
		if (eb->pages[i])
			unlock_page(eb->pages[i]);
	}
C
Chris Mason 已提交
4994

4995
	btrfs_release_extent_buffer(eb);
4996
	return exists;
4997 4998
}

4999 5000 5001 5002 5003 5004 5005 5006 5007
static inline void btrfs_release_extent_buffer_rcu(struct rcu_head *head)
{
	struct extent_buffer *eb =
			container_of(head, struct extent_buffer, rcu_head);

	__free_extent_buffer(eb);
}

/* Expects to have eb->eb_lock already held */
5008
static int release_extent_buffer(struct extent_buffer *eb)
5009 5010 5011
{
	WARN_ON(atomic_read(&eb->refs) == 0);
	if (atomic_dec_and_test(&eb->refs)) {
5012
		if (test_and_clear_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags)) {
5013
			struct btrfs_fs_info *fs_info = eb->fs_info;
5014

5015
			spin_unlock(&eb->refs_lock);
5016

5017 5018
			spin_lock(&fs_info->buffer_lock);
			radix_tree_delete(&fs_info->buffer_radix,
5019
					  eb->start >> PAGE_SHIFT);
5020
			spin_unlock(&fs_info->buffer_lock);
5021 5022
		} else {
			spin_unlock(&eb->refs_lock);
5023
		}
5024 5025

		/* Should be safe to release our pages at this point */
5026
		btrfs_release_extent_buffer_page(eb);
5027 5028 5029 5030 5031 5032
#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
		if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags))) {
			__free_extent_buffer(eb);
			return 1;
		}
#endif
5033
		call_rcu(&eb->rcu_head, btrfs_release_extent_buffer_rcu);
5034
		return 1;
5035 5036
	}
	spin_unlock(&eb->refs_lock);
5037 5038

	return 0;
5039 5040
}

5041 5042
void free_extent_buffer(struct extent_buffer *eb)
{
5043 5044
	int refs;
	int old;
5045 5046 5047
	if (!eb)
		return;

5048 5049 5050 5051 5052 5053 5054 5055 5056
	while (1) {
		refs = atomic_read(&eb->refs);
		if (refs <= 3)
			break;
		old = atomic_cmpxchg(&eb->refs, refs, refs - 1);
		if (old == refs)
			return;
	}

5057
	spin_lock(&eb->refs_lock);
5058 5059 5060 5061
	if (atomic_read(&eb->refs) == 2 &&
	    test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags))
		atomic_dec(&eb->refs);

5062 5063
	if (atomic_read(&eb->refs) == 2 &&
	    test_bit(EXTENT_BUFFER_STALE, &eb->bflags) &&
5064
	    !extent_buffer_under_io(eb) &&
5065 5066 5067 5068 5069 5070 5071
	    test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
		atomic_dec(&eb->refs);

	/*
	 * I know this is terrible, but it's temporary until we stop tracking
	 * the uptodate bits and such for the extent buffers.
	 */
5072
	release_extent_buffer(eb);
5073 5074 5075 5076 5077
}

void free_extent_buffer_stale(struct extent_buffer *eb)
{
	if (!eb)
5078 5079
		return;

5080 5081 5082
	spin_lock(&eb->refs_lock);
	set_bit(EXTENT_BUFFER_STALE, &eb->bflags);

5083
	if (atomic_read(&eb->refs) == 2 && !extent_buffer_under_io(eb) &&
5084 5085
	    test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
		atomic_dec(&eb->refs);
5086
	release_extent_buffer(eb);
5087 5088
}

5089
void clear_extent_buffer_dirty(struct extent_buffer *eb)
5090 5091 5092 5093 5094 5095 5096 5097
{
	unsigned long i;
	unsigned long num_pages;
	struct page *page;

	num_pages = num_extent_pages(eb->start, eb->len);

	for (i = 0; i < num_pages; i++) {
5098
		page = eb->pages[i];
5099
		if (!PageDirty(page))
C
Chris Mason 已提交
5100 5101
			continue;

5102
		lock_page(page);
C
Chris Mason 已提交
5103 5104
		WARN_ON(!PagePrivate(page));

5105
		clear_page_dirty_for_io(page);
5106
		spin_lock_irq(&page->mapping->tree_lock);
5107 5108 5109 5110 5111
		if (!PageDirty(page)) {
			radix_tree_tag_clear(&page->mapping->page_tree,
						page_index(page),
						PAGECACHE_TAG_DIRTY);
		}
5112
		spin_unlock_irq(&page->mapping->tree_lock);
5113
		ClearPageError(page);
5114
		unlock_page(page);
5115
	}
5116
	WARN_ON(atomic_read(&eb->refs) == 0);
5117 5118
}

5119
int set_extent_buffer_dirty(struct extent_buffer *eb)
5120 5121 5122
{
	unsigned long i;
	unsigned long num_pages;
5123
	int was_dirty = 0;
5124

5125 5126
	check_buffer_tree_ref(eb);

5127
	was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
5128

5129
	num_pages = num_extent_pages(eb->start, eb->len);
5130
	WARN_ON(atomic_read(&eb->refs) == 0);
5131 5132
	WARN_ON(!test_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags));

5133
	for (i = 0; i < num_pages; i++)
5134
		set_page_dirty(eb->pages[i]);
5135
	return was_dirty;
5136 5137
}

5138
void clear_extent_buffer_uptodate(struct extent_buffer *eb)
5139 5140 5141 5142 5143
{
	unsigned long i;
	struct page *page;
	unsigned long num_pages;

5144
	clear_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5145
	num_pages = num_extent_pages(eb->start, eb->len);
5146
	for (i = 0; i < num_pages; i++) {
5147
		page = eb->pages[i];
C
Chris Mason 已提交
5148 5149
		if (page)
			ClearPageUptodate(page);
5150 5151 5152
	}
}

5153
void set_extent_buffer_uptodate(struct extent_buffer *eb)
5154 5155 5156 5157 5158
{
	unsigned long i;
	struct page *page;
	unsigned long num_pages;

5159
	set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5160 5161
	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = 0; i < num_pages; i++) {
5162
		page = eb->pages[i];
5163 5164 5165 5166
		SetPageUptodate(page);
	}
}

5167
int extent_buffer_uptodate(struct extent_buffer *eb)
5168
{
5169
	return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5170 5171 5172
}

int read_extent_buffer_pages(struct extent_io_tree *tree,
5173
			     struct extent_buffer *eb, u64 start, int wait,
5174
			     get_extent_t *get_extent, int mirror_num)
5175 5176 5177 5178 5179 5180
{
	unsigned long i;
	unsigned long start_i;
	struct page *page;
	int err;
	int ret = 0;
5181 5182
	int locked_pages = 0;
	int all_uptodate = 1;
5183
	unsigned long num_pages;
5184
	unsigned long num_reads = 0;
5185
	struct bio *bio = NULL;
C
Chris Mason 已提交
5186
	unsigned long bio_flags = 0;
5187

5188
	if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags))
5189 5190 5191 5192
		return 0;

	if (start) {
		WARN_ON(start < eb->start);
5193 5194
		start_i = (start >> PAGE_SHIFT) -
			(eb->start >> PAGE_SHIFT);
5195 5196 5197 5198 5199 5200
	} else {
		start_i = 0;
	}

	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = start_i; i < num_pages; i++) {
5201
		page = eb->pages[i];
5202
		if (wait == WAIT_NONE) {
5203
			if (!trylock_page(page))
5204
				goto unlock_exit;
5205 5206 5207
		} else {
			lock_page(page);
		}
5208
		locked_pages++;
5209 5210
		if (!PageUptodate(page)) {
			num_reads++;
5211
			all_uptodate = 0;
5212
		}
5213 5214 5215
	}
	if (all_uptodate) {
		if (start_i == 0)
5216
			set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5217 5218 5219
		goto unlock_exit;
	}

5220
	clear_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags);
5221
	eb->read_mirror = 0;
5222
	atomic_set(&eb->io_pages, num_reads);
5223
	for (i = start_i; i < num_pages; i++) {
5224
		page = eb->pages[i];
5225
		if (!PageUptodate(page)) {
5226
			ClearPageError(page);
5227
			err = __extent_read_full_page(tree, page,
5228
						      get_extent, &bio,
5229
						      mirror_num, &bio_flags,
5230
						      REQ_META);
C
Chris Mason 已提交
5231
			if (err)
5232 5233 5234 5235 5236 5237
				ret = err;
		} else {
			unlock_page(page);
		}
	}

5238
	if (bio) {
5239
		err = submit_one_bio(bio, mirror_num, bio_flags);
5240 5241
		if (err)
			return err;
5242
	}
5243

5244
	if (ret || wait != WAIT_COMPLETE)
5245
		return ret;
C
Chris Mason 已提交
5246

5247
	for (i = start_i; i < num_pages; i++) {
5248
		page = eb->pages[i];
5249
		wait_on_page_locked(page);
C
Chris Mason 已提交
5250
		if (!PageUptodate(page))
5251 5252
			ret = -EIO;
	}
C
Chris Mason 已提交
5253

5254
	return ret;
5255 5256 5257

unlock_exit:
	i = start_i;
C
Chris Mason 已提交
5258
	while (locked_pages > 0) {
5259
		page = eb->pages[i];
5260 5261 5262 5263 5264
		i++;
		unlock_page(page);
		locked_pages--;
	}
	return ret;
5265 5266 5267 5268 5269 5270 5271 5272 5273 5274 5275
}

void read_extent_buffer(struct extent_buffer *eb, void *dstv,
			unsigned long start,
			unsigned long len)
{
	size_t cur;
	size_t offset;
	struct page *page;
	char *kaddr;
	char *dst = (char *)dstv;
5276 5277
	size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_SHIFT;
5278 5279 5280 5281

	WARN_ON(start > eb->len);
	WARN_ON(start + len > eb->start + eb->len);

5282
	offset = (start_offset + start) & (PAGE_SIZE - 1);
5283

C
Chris Mason 已提交
5284
	while (len > 0) {
5285
		page = eb->pages[i];
5286

5287
		cur = min(len, (PAGE_SIZE - offset));
5288
		kaddr = page_address(page);
5289 5290 5291 5292 5293 5294 5295 5296 5297
		memcpy(dst, kaddr + offset, cur);

		dst += cur;
		len -= cur;
		offset = 0;
		i++;
	}
}

5298 5299 5300 5301 5302 5303 5304 5305 5306
int read_extent_buffer_to_user(struct extent_buffer *eb, void __user *dstv,
			unsigned long start,
			unsigned long len)
{
	size_t cur;
	size_t offset;
	struct page *page;
	char *kaddr;
	char __user *dst = (char __user *)dstv;
5307 5308
	size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_SHIFT;
5309 5310 5311 5312 5313
	int ret = 0;

	WARN_ON(start > eb->len);
	WARN_ON(start + len > eb->start + eb->len);

5314
	offset = (start_offset + start) & (PAGE_SIZE - 1);
5315 5316

	while (len > 0) {
5317
		page = eb->pages[i];
5318

5319
		cur = min(len, (PAGE_SIZE - offset));
5320 5321 5322 5323 5324 5325 5326 5327 5328 5329 5330 5331 5332 5333 5334
		kaddr = page_address(page);
		if (copy_to_user(dst, kaddr + offset, cur)) {
			ret = -EFAULT;
			break;
		}

		dst += cur;
		len -= cur;
		offset = 0;
		i++;
	}

	return ret;
}

5335
int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start,
5336
			       unsigned long min_len, char **map,
5337
			       unsigned long *map_start,
5338
			       unsigned long *map_len)
5339
{
5340
	size_t offset = start & (PAGE_SIZE - 1);
5341 5342
	char *kaddr;
	struct page *p;
5343 5344
	size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_SHIFT;
5345
	unsigned long end_i = (start_offset + start + min_len - 1) >>
5346
		PAGE_SHIFT;
5347 5348 5349 5350 5351 5352 5353 5354 5355

	if (i != end_i)
		return -EINVAL;

	if (i == 0) {
		offset = start_offset;
		*map_start = 0;
	} else {
		offset = 0;
5356
		*map_start = ((u64)i << PAGE_SHIFT) - start_offset;
5357
	}
C
Chris Mason 已提交
5358

5359
	if (start + min_len > eb->len) {
J
Julia Lawall 已提交
5360
		WARN(1, KERN_ERR "btrfs bad mapping eb start %llu len %lu, "
5361 5362
		       "wanted %lu %lu\n",
		       eb->start, eb->len, start, min_len);
5363
		return -EINVAL;
5364 5365
	}

5366
	p = eb->pages[i];
5367
	kaddr = page_address(p);
5368
	*map = kaddr + offset;
5369
	*map_len = PAGE_SIZE - offset;
5370 5371 5372 5373 5374 5375 5376 5377 5378 5379 5380 5381
	return 0;
}

int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv,
			  unsigned long start,
			  unsigned long len)
{
	size_t cur;
	size_t offset;
	struct page *page;
	char *kaddr;
	char *ptr = (char *)ptrv;
5382 5383
	size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_SHIFT;
5384 5385 5386 5387 5388
	int ret = 0;

	WARN_ON(start > eb->len);
	WARN_ON(start + len > eb->start + eb->len);

5389
	offset = (start_offset + start) & (PAGE_SIZE - 1);
5390

C
Chris Mason 已提交
5391
	while (len > 0) {
5392
		page = eb->pages[i];
5393

5394
		cur = min(len, (PAGE_SIZE - offset));
5395

5396
		kaddr = page_address(page);
5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410 5411 5412 5413 5414 5415 5416
		ret = memcmp(ptr, kaddr + offset, cur);
		if (ret)
			break;

		ptr += cur;
		len -= cur;
		offset = 0;
		i++;
	}
	return ret;
}

void write_extent_buffer(struct extent_buffer *eb, const void *srcv,
			 unsigned long start, unsigned long len)
{
	size_t cur;
	size_t offset;
	struct page *page;
	char *kaddr;
	char *src = (char *)srcv;
5417 5418
	size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_SHIFT;
5419 5420 5421 5422

	WARN_ON(start > eb->len);
	WARN_ON(start + len > eb->start + eb->len);

5423
	offset = (start_offset + start) & (PAGE_SIZE - 1);
5424

C
Chris Mason 已提交
5425
	while (len > 0) {
5426
		page = eb->pages[i];
5427 5428
		WARN_ON(!PageUptodate(page));

5429
		cur = min(len, PAGE_SIZE - offset);
5430
		kaddr = page_address(page);
5431 5432 5433 5434 5435 5436 5437 5438 5439 5440 5441 5442 5443 5444 5445 5446
		memcpy(kaddr + offset, src, cur);

		src += cur;
		len -= cur;
		offset = 0;
		i++;
	}
}

void memset_extent_buffer(struct extent_buffer *eb, char c,
			  unsigned long start, unsigned long len)
{
	size_t cur;
	size_t offset;
	struct page *page;
	char *kaddr;
5447 5448
	size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_SHIFT;
5449 5450 5451 5452

	WARN_ON(start > eb->len);
	WARN_ON(start + len > eb->start + eb->len);

5453
	offset = (start_offset + start) & (PAGE_SIZE - 1);
5454

C
Chris Mason 已提交
5455
	while (len > 0) {
5456
		page = eb->pages[i];
5457 5458
		WARN_ON(!PageUptodate(page));

5459
		cur = min(len, PAGE_SIZE - offset);
5460
		kaddr = page_address(page);
5461 5462 5463 5464 5465 5466 5467 5468 5469 5470 5471 5472 5473 5474 5475 5476 5477
		memset(kaddr + offset, c, cur);

		len -= cur;
		offset = 0;
		i++;
	}
}

void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src,
			unsigned long dst_offset, unsigned long src_offset,
			unsigned long len)
{
	u64 dst_len = dst->len;
	size_t cur;
	size_t offset;
	struct page *page;
	char *kaddr;
5478 5479
	size_t start_offset = dst->start & ((u64)PAGE_SIZE - 1);
	unsigned long i = (start_offset + dst_offset) >> PAGE_SHIFT;
5480 5481 5482 5483

	WARN_ON(src->len != dst_len);

	offset = (start_offset + dst_offset) &
5484
		(PAGE_SIZE - 1);
5485

C
Chris Mason 已提交
5486
	while (len > 0) {
5487
		page = dst->pages[i];
5488 5489
		WARN_ON(!PageUptodate(page));

5490
		cur = min(len, (unsigned long)(PAGE_SIZE - offset));
5491

5492
		kaddr = page_address(page);
5493 5494 5495 5496 5497 5498 5499 5500 5501
		read_extent_buffer(src, kaddr + offset, src_offset, cur);

		src_offset += cur;
		len -= cur;
		offset = 0;
		i++;
	}
}

5502 5503 5504 5505 5506 5507 5508 5509 5510 5511 5512 5513 5514 5515 5516 5517 5518 5519 5520 5521 5522 5523 5524 5525 5526 5527 5528 5529 5530 5531
/*
 * The extent buffer bitmap operations are done with byte granularity because
 * bitmap items are not guaranteed to be aligned to a word and therefore a
 * single word in a bitmap may straddle two pages in the extent buffer.
 */
#define BIT_BYTE(nr) ((nr) / BITS_PER_BYTE)
#define BYTE_MASK ((1 << BITS_PER_BYTE) - 1)
#define BITMAP_FIRST_BYTE_MASK(start) \
	((BYTE_MASK << ((start) & (BITS_PER_BYTE - 1))) & BYTE_MASK)
#define BITMAP_LAST_BYTE_MASK(nbits) \
	(BYTE_MASK >> (-(nbits) & (BITS_PER_BYTE - 1)))

/*
 * eb_bitmap_offset() - calculate the page and offset of the byte containing the
 * given bit number
 * @eb: the extent buffer
 * @start: offset of the bitmap item in the extent buffer
 * @nr: bit number
 * @page_index: return index of the page in the extent buffer that contains the
 * given bit number
 * @page_offset: return offset into the page given by page_index
 *
 * This helper hides the ugliness of finding the byte in an extent buffer which
 * contains a given bit.
 */
static inline void eb_bitmap_offset(struct extent_buffer *eb,
				    unsigned long start, unsigned long nr,
				    unsigned long *page_index,
				    size_t *page_offset)
{
5532
	size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
5533 5534 5535 5536 5537 5538 5539 5540 5541 5542
	size_t byte_offset = BIT_BYTE(nr);
	size_t offset;

	/*
	 * The byte we want is the offset of the extent buffer + the offset of
	 * the bitmap item in the extent buffer + the offset of the byte in the
	 * bitmap item.
	 */
	offset = start_offset + start + byte_offset;

5543 5544
	*page_index = offset >> PAGE_SHIFT;
	*page_offset = offset & (PAGE_SIZE - 1);
5545 5546 5547 5548 5549 5550 5551 5552 5553 5554 5555 5556 5557 5558 5559 5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594 5595
}

/**
 * extent_buffer_test_bit - determine whether a bit in a bitmap item is set
 * @eb: the extent buffer
 * @start: offset of the bitmap item in the extent buffer
 * @nr: bit number to test
 */
int extent_buffer_test_bit(struct extent_buffer *eb, unsigned long start,
			   unsigned long nr)
{
	char *kaddr;
	struct page *page;
	unsigned long i;
	size_t offset;

	eb_bitmap_offset(eb, start, nr, &i, &offset);
	page = eb->pages[i];
	WARN_ON(!PageUptodate(page));
	kaddr = page_address(page);
	return 1U & (kaddr[offset] >> (nr & (BITS_PER_BYTE - 1)));
}

/**
 * extent_buffer_bitmap_set - set an area of a bitmap
 * @eb: the extent buffer
 * @start: offset of the bitmap item in the extent buffer
 * @pos: bit number of the first bit
 * @len: number of bits to set
 */
void extent_buffer_bitmap_set(struct extent_buffer *eb, unsigned long start,
			      unsigned long pos, unsigned long len)
{
	char *kaddr;
	struct page *page;
	unsigned long i;
	size_t offset;
	const unsigned int size = pos + len;
	int bits_to_set = BITS_PER_BYTE - (pos % BITS_PER_BYTE);
	unsigned int mask_to_set = BITMAP_FIRST_BYTE_MASK(pos);

	eb_bitmap_offset(eb, start, pos, &i, &offset);
	page = eb->pages[i];
	WARN_ON(!PageUptodate(page));
	kaddr = page_address(page);

	while (len >= bits_to_set) {
		kaddr[offset] |= mask_to_set;
		len -= bits_to_set;
		bits_to_set = BITS_PER_BYTE;
		mask_to_set = ~0U;
5596
		if (++offset >= PAGE_SIZE && len > 0) {
5597 5598 5599 5600 5601 5602 5603 5604 5605 5606 5607 5608 5609 5610 5611 5612 5613 5614 5615 5616 5617 5618 5619 5620 5621 5622 5623 5624 5625 5626 5627 5628 5629 5630 5631 5632 5633 5634 5635 5636 5637
			offset = 0;
			page = eb->pages[++i];
			WARN_ON(!PageUptodate(page));
			kaddr = page_address(page);
		}
	}
	if (len) {
		mask_to_set &= BITMAP_LAST_BYTE_MASK(size);
		kaddr[offset] |= mask_to_set;
	}
}


/**
 * extent_buffer_bitmap_clear - clear an area of a bitmap
 * @eb: the extent buffer
 * @start: offset of the bitmap item in the extent buffer
 * @pos: bit number of the first bit
 * @len: number of bits to clear
 */
void extent_buffer_bitmap_clear(struct extent_buffer *eb, unsigned long start,
				unsigned long pos, unsigned long len)
{
	char *kaddr;
	struct page *page;
	unsigned long i;
	size_t offset;
	const unsigned int size = pos + len;
	int bits_to_clear = BITS_PER_BYTE - (pos % BITS_PER_BYTE);
	unsigned int mask_to_clear = BITMAP_FIRST_BYTE_MASK(pos);

	eb_bitmap_offset(eb, start, pos, &i, &offset);
	page = eb->pages[i];
	WARN_ON(!PageUptodate(page));
	kaddr = page_address(page);

	while (len >= bits_to_clear) {
		kaddr[offset] &= ~mask_to_clear;
		len -= bits_to_clear;
		bits_to_clear = BITS_PER_BYTE;
		mask_to_clear = ~0U;
5638
		if (++offset >= PAGE_SIZE && len > 0) {
5639 5640 5641 5642 5643 5644 5645 5646 5647 5648 5649 5650
			offset = 0;
			page = eb->pages[++i];
			WARN_ON(!PageUptodate(page));
			kaddr = page_address(page);
		}
	}
	if (len) {
		mask_to_clear &= BITMAP_LAST_BYTE_MASK(size);
		kaddr[offset] &= ~mask_to_clear;
	}
}

5651 5652 5653 5654 5655 5656
static inline bool areas_overlap(unsigned long src, unsigned long dst, unsigned long len)
{
	unsigned long distance = (src > dst) ? src - dst : dst - src;
	return distance < len;
}

5657 5658 5659 5660
static void copy_pages(struct page *dst_page, struct page *src_page,
		       unsigned long dst_off, unsigned long src_off,
		       unsigned long len)
{
5661
	char *dst_kaddr = page_address(dst_page);
5662
	char *src_kaddr;
5663
	int must_memmove = 0;
5664

5665
	if (dst_page != src_page) {
5666
		src_kaddr = page_address(src_page);
5667
	} else {
5668
		src_kaddr = dst_kaddr;
5669 5670
		if (areas_overlap(src_off, dst_off, len))
			must_memmove = 1;
5671
	}
5672

5673 5674 5675 5676
	if (must_memmove)
		memmove(dst_kaddr + dst_off, src_kaddr + src_off, len);
	else
		memcpy(dst_kaddr + dst_off, src_kaddr + src_off, len);
5677 5678 5679 5680 5681 5682 5683 5684
}

void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
			   unsigned long src_offset, unsigned long len)
{
	size_t cur;
	size_t dst_off_in_page;
	size_t src_off_in_page;
5685
	size_t start_offset = dst->start & ((u64)PAGE_SIZE - 1);
5686 5687 5688 5689
	unsigned long dst_i;
	unsigned long src_i;

	if (src_offset + len > dst->len) {
5690 5691 5692
		btrfs_err(dst->fs_info,
			"memmove bogus src_offset %lu move "
		       "len %lu dst len %lu", src_offset, len, dst->len);
5693 5694 5695
		BUG_ON(1);
	}
	if (dst_offset + len > dst->len) {
5696 5697 5698
		btrfs_err(dst->fs_info,
			"memmove bogus dst_offset %lu move "
		       "len %lu dst len %lu", dst_offset, len, dst->len);
5699 5700 5701
		BUG_ON(1);
	}

C
Chris Mason 已提交
5702
	while (len > 0) {
5703
		dst_off_in_page = (start_offset + dst_offset) &
5704
			(PAGE_SIZE - 1);
5705
		src_off_in_page = (start_offset + src_offset) &
5706
			(PAGE_SIZE - 1);
5707

5708 5709
		dst_i = (start_offset + dst_offset) >> PAGE_SHIFT;
		src_i = (start_offset + src_offset) >> PAGE_SHIFT;
5710

5711
		cur = min(len, (unsigned long)(PAGE_SIZE -
5712 5713
					       src_off_in_page));
		cur = min_t(unsigned long, cur,
5714
			(unsigned long)(PAGE_SIZE - dst_off_in_page));
5715

5716
		copy_pages(dst->pages[dst_i], dst->pages[src_i],
5717 5718 5719 5720 5721 5722 5723 5724 5725 5726 5727 5728 5729 5730 5731 5732
			   dst_off_in_page, src_off_in_page, cur);

		src_offset += cur;
		dst_offset += cur;
		len -= cur;
	}
}

void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
			   unsigned long src_offset, unsigned long len)
{
	size_t cur;
	size_t dst_off_in_page;
	size_t src_off_in_page;
	unsigned long dst_end = dst_offset + len - 1;
	unsigned long src_end = src_offset + len - 1;
5733
	size_t start_offset = dst->start & ((u64)PAGE_SIZE - 1);
5734 5735 5736 5737
	unsigned long dst_i;
	unsigned long src_i;

	if (src_offset + len > dst->len) {
5738 5739
		btrfs_err(dst->fs_info, "memmove bogus src_offset %lu move "
		       "len %lu len %lu", src_offset, len, dst->len);
5740 5741 5742
		BUG_ON(1);
	}
	if (dst_offset + len > dst->len) {
5743 5744
		btrfs_err(dst->fs_info, "memmove bogus dst_offset %lu move "
		       "len %lu len %lu", dst_offset, len, dst->len);
5745 5746
		BUG_ON(1);
	}
5747
	if (dst_offset < src_offset) {
5748 5749 5750
		memcpy_extent_buffer(dst, dst_offset, src_offset, len);
		return;
	}
C
Chris Mason 已提交
5751
	while (len > 0) {
5752 5753
		dst_i = (start_offset + dst_end) >> PAGE_SHIFT;
		src_i = (start_offset + src_end) >> PAGE_SHIFT;
5754 5755

		dst_off_in_page = (start_offset + dst_end) &
5756
			(PAGE_SIZE - 1);
5757
		src_off_in_page = (start_offset + src_end) &
5758
			(PAGE_SIZE - 1);
5759 5760 5761

		cur = min_t(unsigned long, len, src_off_in_page + 1);
		cur = min(cur, dst_off_in_page + 1);
5762
		copy_pages(dst->pages[dst_i], dst->pages[src_i],
5763 5764 5765 5766 5767 5768 5769 5770
			   dst_off_in_page - cur + 1,
			   src_off_in_page - cur + 1, cur);

		dst_end -= cur;
		src_end -= cur;
		len -= cur;
	}
}
5771

5772
int try_release_extent_buffer(struct page *page)
5773
{
5774 5775
	struct extent_buffer *eb;

5776
	/*
5777
	 * We need to make sure nobody is attaching this page to an eb right
5778 5779 5780 5781 5782
	 * now.
	 */
	spin_lock(&page->mapping->private_lock);
	if (!PagePrivate(page)) {
		spin_unlock(&page->mapping->private_lock);
J
Josef Bacik 已提交
5783
		return 1;
5784
	}
5785

5786 5787
	eb = (struct extent_buffer *)page->private;
	BUG_ON(!eb);
5788 5789

	/*
5790 5791 5792
	 * This is a little awful but should be ok, we need to make sure that
	 * the eb doesn't disappear out from under us while we're looking at
	 * this page.
5793
	 */
5794
	spin_lock(&eb->refs_lock);
5795
	if (atomic_read(&eb->refs) != 1 || extent_buffer_under_io(eb)) {
5796 5797 5798
		spin_unlock(&eb->refs_lock);
		spin_unlock(&page->mapping->private_lock);
		return 0;
5799
	}
5800
	spin_unlock(&page->mapping->private_lock);
5801

5802
	/*
5803 5804
	 * If tree ref isn't set then we know the ref on this eb is a real ref,
	 * so just return, this page will likely be freed soon anyway.
5805
	 */
5806 5807 5808
	if (!test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags)) {
		spin_unlock(&eb->refs_lock);
		return 0;
5809
	}
5810

5811
	return release_extent_buffer(eb);
5812
}