extent_io.c 145.7 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();
209 210 211 212
	if (extent_state_cache)
		kmem_cache_destroy(extent_state_cache);
	if (extent_buffer_cache)
		kmem_cache_destroy(extent_buffer_cache);
213 214
	if (btrfs_bioset)
		bioset_free(btrfs_bioset);
215 216 217
}

void extent_io_tree_init(struct extent_io_tree *tree,
218
			 struct address_space *mapping)
219
{
220
	tree->state = RB_ROOT;
221 222
	tree->ops = NULL;
	tree->dirty_bytes = 0;
223
	spin_lock_init(&tree->lock);
224 225 226
	tree->mapping = mapping;
}

227
static struct extent_state *alloc_extent_state(gfp_t mask)
228 229 230 231
{
	struct extent_state *state;

	state = kmem_cache_alloc(extent_state_cache, mask);
232
	if (!state)
233 234 235
		return state;
	state->state = 0;
	state->private = 0;
236
	RB_CLEAR_NODE(&state->rb_node);
237
	btrfs_leak_debug_add(&state->leak_list, &states);
238 239
	atomic_set(&state->refs, 1);
	init_waitqueue_head(&state->wq);
240
	trace_alloc_extent_state(state, mask, _RET_IP_);
241 242 243
	return state;
}

244
void free_extent_state(struct extent_state *state)
245 246 247 248
{
	if (!state)
		return;
	if (atomic_dec_and_test(&state->refs)) {
249
		WARN_ON(extent_state_in_tree(state));
250
		btrfs_leak_debug_del(&state->leak_list);
251
		trace_free_extent_state(state, _RET_IP_);
252 253 254 255
		kmem_cache_free(extent_state_cache, state);
	}
}

256 257 258
static struct rb_node *tree_insert(struct rb_root *root,
				   struct rb_node *search_start,
				   u64 offset,
259 260 261
				   struct rb_node *node,
				   struct rb_node ***p_in,
				   struct rb_node **parent_in)
262
{
263
	struct rb_node **p;
C
Chris Mason 已提交
264
	struct rb_node *parent = NULL;
265 266
	struct tree_entry *entry;

267 268 269 270 271 272
	if (p_in && parent_in) {
		p = *p_in;
		parent = *parent_in;
		goto do_insert;
	}

273
	p = search_start ? &search_start : &root->rb_node;
C
Chris Mason 已提交
274
	while (*p) {
275 276 277 278 279 280 281 282 283 284 285
		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;
	}

286
do_insert:
287 288 289 290 291
	rb_link_node(node, parent, p);
	rb_insert_color(node, root);
	return NULL;
}

292
static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset,
293 294 295 296
				      struct rb_node **prev_ret,
				      struct rb_node **next_ret,
				      struct rb_node ***p_ret,
				      struct rb_node **parent_ret)
297
{
298
	struct rb_root *root = &tree->state;
299
	struct rb_node **n = &root->rb_node;
300 301 302 303 304
	struct rb_node *prev = NULL;
	struct rb_node *orig_prev = NULL;
	struct tree_entry *entry;
	struct tree_entry *prev_entry = NULL;

305 306 307
	while (*n) {
		prev = *n;
		entry = rb_entry(prev, struct tree_entry, rb_node);
308 309 310
		prev_entry = entry;

		if (offset < entry->start)
311
			n = &(*n)->rb_left;
312
		else if (offset > entry->end)
313
			n = &(*n)->rb_right;
C
Chris Mason 已提交
314
		else
315
			return *n;
316 317
	}

318 319 320 321 322
	if (p_ret)
		*p_ret = n;
	if (parent_ret)
		*parent_ret = prev;

323 324
	if (prev_ret) {
		orig_prev = prev;
C
Chris Mason 已提交
325
		while (prev && offset > prev_entry->end) {
326 327 328 329 330 331 332 333 334
			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
335
		while (prev && offset < prev_entry->start) {
336 337 338 339 340 341 342 343
			prev = rb_prev(prev);
			prev_entry = rb_entry(prev, struct tree_entry, rb_node);
		}
		*next_ret = prev;
	}
	return NULL;
}

344 345 346 347 348
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)
349
{
350
	struct rb_node *prev = NULL;
351
	struct rb_node *ret;
352

353
	ret = __etree_search(tree, offset, &prev, NULL, p_ret, parent_ret);
C
Chris Mason 已提交
354
	if (!ret)
355 356 357 358
		return prev;
	return ret;
}

359 360 361 362 363 364
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
365 366 367 368 369 370 371 372
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);
}

373 374 375 376 377 378 379 380 381
/*
 * 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.
 */
382 383
static void merge_state(struct extent_io_tree *tree,
		        struct extent_state *state)
384 385 386 387
{
	struct extent_state *other;
	struct rb_node *other_node;

388
	if (state->state & (EXTENT_IOBITS | EXTENT_BOUNDARY))
389
		return;
390 391 392 393 394 395

	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
396
			merge_cb(tree, state, other);
397 398
			state->start = other->start;
			rb_erase(&other->rb_node, &tree->state);
399
			RB_CLEAR_NODE(&other->rb_node);
400 401 402 403 404 405 406 407
			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
408
			merge_cb(tree, state, other);
409 410
			state->end = other->end;
			rb_erase(&other->rb_node, &tree->state);
411
			RB_CLEAR_NODE(&other->rb_node);
412
			free_extent_state(other);
413 414 415 416
		}
	}
}

417
static void set_state_cb(struct extent_io_tree *tree,
418
			 struct extent_state *state, unsigned *bits)
419
{
420 421
	if (tree->ops && tree->ops->set_bit_hook)
		tree->ops->set_bit_hook(tree->mapping->host, state, bits);
422 423 424
}

static void clear_state_cb(struct extent_io_tree *tree,
425
			   struct extent_state *state, unsigned *bits)
426
{
J
Josef Bacik 已提交
427 428
	if (tree->ops && tree->ops->clear_bit_hook)
		tree->ops->clear_bit_hook(tree->mapping->host, state, bits);
429 430
}

431
static void set_state_bits(struct extent_io_tree *tree,
432 433
			   struct extent_state *state, unsigned *bits,
			   struct extent_changeset *changeset);
434

435 436 437 438 439 440 441 442 443 444 445 446
/*
 * 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,
447 448
			struct rb_node ***p,
			struct rb_node **parent,
449
			unsigned *bits, struct extent_changeset *changeset)
450 451 452
{
	struct rb_node *node;

J
Julia Lawall 已提交
453
	if (end < start)
454
		WARN(1, KERN_ERR "BTRFS: end < start %llu %llu\n",
455
		       end, start);
456 457
	state->start = start;
	state->end = end;
J
Josef Bacik 已提交
458

459
	set_state_bits(tree, state, bits, changeset);
460

461
	node = tree_insert(&tree->state, NULL, end, &state->rb_node, p, parent);
462 463 464
	if (node) {
		struct extent_state *found;
		found = rb_entry(node, struct extent_state, rb_node);
465
		printk(KERN_ERR "BTRFS: found node %llu %llu on insert of "
466 467
		       "%llu %llu\n",
		       found->start, found->end, start, end);
468 469 470 471 472 473
		return -EEXIST;
	}
	merge_state(tree, state);
	return 0;
}

474
static void split_cb(struct extent_io_tree *tree, struct extent_state *orig,
J
Josef Bacik 已提交
475 476 477
		     u64 split)
{
	if (tree->ops && tree->ops->split_extent_hook)
478
		tree->ops->split_extent_hook(tree->mapping->host, orig, split);
J
Josef Bacik 已提交
479 480
}

481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498
/*
 * 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
499 500 501

	split_cb(tree, orig, split);

502 503 504 505 506
	prealloc->start = orig->start;
	prealloc->end = split - 1;
	prealloc->state = orig->state;
	orig->start = split;

507 508
	node = tree_insert(&tree->state, &orig->rb_node, prealloc->end,
			   &prealloc->rb_node, NULL, NULL);
509 510 511 512 513 514 515
	if (node) {
		free_extent_state(prealloc);
		return -EEXIST;
	}
	return 0;
}

516 517 518 519 520 521 522 523 524
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;
}

525 526
/*
 * utility function to clear some bits in an extent state struct.
527
 * it will optionally wake up any one waiting on this state (wake == 1).
528 529 530 531
 *
 * If no bits are set on the state struct after clearing things, the
 * struct is freed and removed from the tree
 */
532 533
static struct extent_state *clear_state_bit(struct extent_io_tree *tree,
					    struct extent_state *state,
534 535
					    unsigned *bits, int wake,
					    struct extent_changeset *changeset)
536
{
537
	struct extent_state *next;
538
	unsigned bits_to_clear = *bits & ~EXTENT_CTLBITS;
539

540
	if ((bits_to_clear & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) {
541 542 543 544
		u64 range = state->end - state->start + 1;
		WARN_ON(range > tree->dirty_bytes);
		tree->dirty_bytes -= range;
	}
545
	clear_state_cb(tree, state, bits);
546
	add_extent_changeset(state, bits_to_clear, changeset, 0);
547
	state->state &= ~bits_to_clear;
548 549
	if (wake)
		wake_up(&state->wq);
550
	if (state->state == 0) {
551
		next = next_state(state);
552
		if (extent_state_in_tree(state)) {
553
			rb_erase(&state->rb_node, &tree->state);
554
			RB_CLEAR_NODE(&state->rb_node);
555 556 557 558 559 560
			free_extent_state(state);
		} else {
			WARN_ON(1);
		}
	} else {
		merge_state(tree, state);
561
		next = next_state(state);
562
	}
563
	return next;
564 565
}

566 567 568 569 570 571 572 573 574
static struct extent_state *
alloc_extent_state_atomic(struct extent_state *prealloc)
{
	if (!prealloc)
		prealloc = alloc_extent_state(GFP_ATOMIC);

	return prealloc;
}

575
static void extent_io_tree_panic(struct extent_io_tree *tree, int err)
576 577 578 579 580 581
{
	btrfs_panic(tree_fs_info(tree), err, "Locking error: "
		    "Extent tree was modified by another "
		    "thread while locked.");
}

582 583 584 585 586 587 588 589 590 591
/*
 * 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.
 *
592
 * This takes the tree lock, and returns 0 on success and < 0 on error.
593
 */
594 595 596 597
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)
598 599
{
	struct extent_state *state;
600
	struct extent_state *cached;
601 602
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
603
	u64 last_end;
604
	int err;
605
	int clear = 0;
606

607
	btrfs_debug_check_extent_io_range(tree, start, end);
608

609 610 611
	if (bits & EXTENT_DELALLOC)
		bits |= EXTENT_NORESERVE;

612 613 614 615
	if (delete)
		bits |= ~EXTENT_CTLBITS;
	bits |= EXTENT_FIRST_DELALLOC;

616 617
	if (bits & (EXTENT_IOBITS | EXTENT_BOUNDARY))
		clear = 1;
618
again:
619
	if (!prealloc && gfpflags_allow_blocking(mask)) {
620 621 622 623 624 625 626
		/*
		 * 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.
		 */
627 628 629
		prealloc = alloc_extent_state(mask);
	}

630
	spin_lock(&tree->lock);
631 632
	if (cached_state) {
		cached = *cached_state;
633 634 635 636 637 638

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

639 640
		if (cached && extent_state_in_tree(cached) &&
		    cached->start <= start && cached->end > start) {
641 642
			if (clear)
				atomic_dec(&cached->refs);
643
			state = cached;
644
			goto hit_next;
645
		}
646 647
		if (clear)
			free_extent_state(cached);
648
	}
649 650 651 652
	/*
	 * this search will find the extents that end after
	 * our range starts
	 */
653
	node = tree_search(tree, start);
654 655 656
	if (!node)
		goto out;
	state = rb_entry(node, struct extent_state, rb_node);
657
hit_next:
658 659 660
	if (state->start > end)
		goto out;
	WARN_ON(state->end < start);
661
	last_end = state->end;
662

663
	/* the state doesn't have the wanted bits, go ahead */
664 665
	if (!(state->state & bits)) {
		state = next_state(state);
666
		goto next;
667
	}
668

669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685
	/*
	 *     | ---- 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) {
686 687
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
688
		err = split_state(tree, state, prealloc, start);
689 690 691
		if (err)
			extent_io_tree_panic(tree, err);

692 693 694 695
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
696 697
			state = clear_state_bit(tree, state, &bits, wake,
						changeset);
698
			goto next;
699 700 701 702 703 704 705 706 707 708
		}
		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) {
709 710
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
711
		err = split_state(tree, state, prealloc, end + 1);
712 713 714
		if (err)
			extent_io_tree_panic(tree, err);

715 716
		if (wake)
			wake_up(&state->wq);
717

718
		clear_state_bit(tree, prealloc, &bits, wake, changeset);
J
Josef Bacik 已提交
719

720 721 722
		prealloc = NULL;
		goto out;
	}
723

724
	state = clear_state_bit(tree, state, &bits, wake, changeset);
725
next:
726 727 728
	if (last_end == (u64)-1)
		goto out;
	start = last_end + 1;
729
	if (start <= end && state && !need_resched())
730
		goto hit_next;
731 732 733
	goto search_again;

out:
734
	spin_unlock(&tree->lock);
735 736 737
	if (prealloc)
		free_extent_state(prealloc);

738
	return 0;
739 740 741 742

search_again:
	if (start > end)
		goto out;
743
	spin_unlock(&tree->lock);
744
	if (gfpflags_allow_blocking(mask))
745 746 747 748
		cond_resched();
	goto again;
}

749 750
static void wait_on_state(struct extent_io_tree *tree,
			  struct extent_state *state)
751 752
		__releases(tree->lock)
		__acquires(tree->lock)
753 754 755
{
	DEFINE_WAIT(wait);
	prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE);
756
	spin_unlock(&tree->lock);
757
	schedule();
758
	spin_lock(&tree->lock);
759 760 761 762 763 764 765 766
	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
 */
767 768
static void wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
			    unsigned long bits)
769 770 771 772
{
	struct extent_state *state;
	struct rb_node *node;

773
	btrfs_debug_check_extent_io_range(tree, start, end);
774

775
	spin_lock(&tree->lock);
776 777 778 779 780 781
again:
	while (1) {
		/*
		 * this search will find all the extents that end after
		 * our range starts
		 */
782
		node = tree_search(tree, start);
783
process_node:
784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803
		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;

804 805 806 807
		if (!cond_resched_lock(&tree->lock)) {
			node = rb_next(node);
			goto process_node;
		}
808 809
	}
out:
810
	spin_unlock(&tree->lock);
811 812
}

813
static void set_state_bits(struct extent_io_tree *tree,
814
			   struct extent_state *state,
815
			   unsigned *bits, struct extent_changeset *changeset)
816
{
817
	unsigned bits_to_set = *bits & ~EXTENT_CTLBITS;
J
Josef Bacik 已提交
818

819
	set_state_cb(tree, state, bits);
820
	if ((bits_to_set & EXTENT_DIRTY) && !(state->state & EXTENT_DIRTY)) {
821 822 823
		u64 range = state->end - state->start + 1;
		tree->dirty_bytes += range;
	}
824
	add_extent_changeset(state, bits_to_set, changeset, 1);
825
	state->state |= bits_to_set;
826 827
}

828 829
static void cache_state_if_flags(struct extent_state *state,
				 struct extent_state **cached_ptr,
830
				 unsigned flags)
831 832
{
	if (cached_ptr && !(*cached_ptr)) {
833
		if (!flags || (state->state & flags)) {
834 835 836 837 838 839
			*cached_ptr = state;
			atomic_inc(&state->refs);
		}
	}
}

840 841 842 843 844 845 846
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);
}

847
/*
848 849
 * 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.
850
 *
851 852 853
 * 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.
854
 *
855
 * [start, end] is inclusive This takes the tree lock.
856
 */
857

J
Jeff Mahoney 已提交
858 859
static int __must_check
__set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
860
		 unsigned bits, unsigned exclusive_bits,
861
		 u64 *failed_start, struct extent_state **cached_state,
862
		 gfp_t mask, struct extent_changeset *changeset)
863 864 865 866
{
	struct extent_state *state;
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
867 868
	struct rb_node **p;
	struct rb_node *parent;
869 870 871
	int err = 0;
	u64 last_start;
	u64 last_end;
872

873
	btrfs_debug_check_extent_io_range(tree, start, end);
874

875
	bits |= EXTENT_FIRST_DELALLOC;
876
again:
877
	if (!prealloc && gfpflags_allow_blocking(mask)) {
878
		prealloc = alloc_extent_state(mask);
879
		BUG_ON(!prealloc);
880 881
	}

882
	spin_lock(&tree->lock);
883 884
	if (cached_state && *cached_state) {
		state = *cached_state;
885
		if (state->start <= start && state->end > start &&
886
		    extent_state_in_tree(state)) {
887 888 889 890
			node = &state->rb_node;
			goto hit_next;
		}
	}
891 892 893 894
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
895
	node = tree_search_for_insert(tree, start, &p, &parent);
896
	if (!node) {
897 898
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
899
		err = insert_state(tree, prealloc, start, end,
900
				   &p, &parent, &bits, changeset);
901 902 903
		if (err)
			extent_io_tree_panic(tree, err);

904
		cache_state(prealloc, cached_state);
905 906 907 908
		prealloc = NULL;
		goto out;
	}
	state = rb_entry(node, struct extent_state, rb_node);
C
Chris Mason 已提交
909
hit_next:
910 911 912 913 914 915 916 917 918 919
	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) {
920
		if (state->state & exclusive_bits) {
921 922 923 924
			*failed_start = state->start;
			err = -EEXIST;
			goto out;
		}
925

926
		set_state_bits(tree, state, &bits, changeset);
927
		cache_state(state, cached_state);
928
		merge_state(tree, state);
929 930 931
		if (last_end == (u64)-1)
			goto out;
		start = last_end + 1;
932 933 934 935
		state = next_state(state);
		if (start < end && state && state->start == start &&
		    !need_resched())
			goto hit_next;
936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955
		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) {
956
		if (state->state & exclusive_bits) {
957 958 959 960
			*failed_start = start;
			err = -EEXIST;
			goto out;
		}
961 962 963

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
964
		err = split_state(tree, state, prealloc, start);
965 966 967
		if (err)
			extent_io_tree_panic(tree, err);

968 969 970 971
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
972
			set_state_bits(tree, state, &bits, changeset);
973
			cache_state(state, cached_state);
974
			merge_state(tree, state);
975 976 977
			if (last_end == (u64)-1)
				goto out;
			start = last_end + 1;
978 979 980 981
			state = next_state(state);
			if (start < end && state && state->start == start &&
			    !need_resched())
				goto hit_next;
982 983 984 985 986 987 988 989 990 991 992 993 994 995 996
		}
		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
997
			this_end = last_start - 1;
998 999 1000

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
1001 1002 1003 1004 1005

		/*
		 * Avoid to free 'prealloc' if it can be merged with
		 * the later extent.
		 */
1006
		err = insert_state(tree, prealloc, start, this_end,
1007
				   NULL, NULL, &bits, changeset);
1008 1009 1010
		if (err)
			extent_io_tree_panic(tree, err);

J
Josef Bacik 已提交
1011 1012
		cache_state(prealloc, cached_state);
		prealloc = NULL;
1013 1014 1015 1016 1017 1018 1019 1020 1021 1022
		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) {
1023
		if (state->state & exclusive_bits) {
1024 1025 1026 1027
			*failed_start = start;
			err = -EEXIST;
			goto out;
		}
1028 1029 1030

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
1031
		err = split_state(tree, state, prealloc, end + 1);
1032 1033
		if (err)
			extent_io_tree_panic(tree, err);
1034

1035
		set_state_bits(tree, prealloc, &bits, changeset);
1036
		cache_state(prealloc, cached_state);
1037 1038 1039 1040 1041 1042 1043 1044
		merge_state(tree, prealloc);
		prealloc = NULL;
		goto out;
	}

	goto search_again;

out:
1045
	spin_unlock(&tree->lock);
1046 1047 1048 1049 1050 1051 1052 1053
	if (prealloc)
		free_extent_state(prealloc);

	return err;

search_again:
	if (start > end)
		goto out;
1054
	spin_unlock(&tree->lock);
1055
	if (gfpflags_allow_blocking(mask))
1056 1057 1058 1059
		cond_resched();
	goto again;
}

1060
int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
1061
		   unsigned bits, u64 * failed_start,
1062
		   struct extent_state **cached_state, gfp_t mask)
J
Jeff Mahoney 已提交
1063 1064
{
	return __set_extent_bit(tree, start, end, bits, 0, failed_start,
1065
				cached_state, mask, NULL);
J
Jeff Mahoney 已提交
1066 1067 1068
}


J
Josef Bacik 已提交
1069
/**
L
Liu Bo 已提交
1070 1071
 * convert_extent_bit - convert all bits in a given range from one bit to
 * 			another
J
Josef Bacik 已提交
1072 1073 1074 1075 1076
 * @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
1077
 * @cached_state:	state that we're going to cache
J
Josef Bacik 已提交
1078 1079 1080 1081 1082 1083 1084 1085 1086
 * @mask:	the allocation mask
 *
 * 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.
 */
int convert_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
1087
		       unsigned bits, unsigned clear_bits,
1088
		       struct extent_state **cached_state, gfp_t mask)
J
Josef Bacik 已提交
1089 1090 1091 1092
{
	struct extent_state *state;
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
1093 1094
	struct rb_node **p;
	struct rb_node *parent;
J
Josef Bacik 已提交
1095 1096 1097
	int err = 0;
	u64 last_start;
	u64 last_end;
1098
	bool first_iteration = true;
J
Josef Bacik 已提交
1099

1100
	btrfs_debug_check_extent_io_range(tree, start, end);
1101

J
Josef Bacik 已提交
1102
again:
1103
	if (!prealloc && gfpflags_allow_blocking(mask)) {
1104 1105 1106 1107 1108 1109 1110
		/*
		 * 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.
		 */
J
Josef Bacik 已提交
1111
		prealloc = alloc_extent_state(mask);
1112
		if (!prealloc && !first_iteration)
J
Josef Bacik 已提交
1113 1114 1115 1116
			return -ENOMEM;
	}

	spin_lock(&tree->lock);
1117 1118 1119
	if (cached_state && *cached_state) {
		state = *cached_state;
		if (state->start <= start && state->end > start &&
1120
		    extent_state_in_tree(state)) {
1121 1122 1123 1124 1125
			node = &state->rb_node;
			goto hit_next;
		}
	}

J
Josef Bacik 已提交
1126 1127 1128 1129
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1130
	node = tree_search_for_insert(tree, start, &p, &parent);
J
Josef Bacik 已提交
1131 1132
	if (!node) {
		prealloc = alloc_extent_state_atomic(prealloc);
1133 1134 1135 1136
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
1137
		err = insert_state(tree, prealloc, start, end,
1138
				   &p, &parent, &bits, NULL);
1139 1140
		if (err)
			extent_io_tree_panic(tree, err);
1141 1142
		cache_state(prealloc, cached_state);
		prealloc = NULL;
J
Josef Bacik 已提交
1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156
		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) {
1157
		set_state_bits(tree, state, &bits, NULL);
1158
		cache_state(state, cached_state);
1159
		state = clear_state_bit(tree, state, &clear_bits, 0, NULL);
J
Josef Bacik 已提交
1160 1161 1162
		if (last_end == (u64)-1)
			goto out;
		start = last_end + 1;
1163 1164 1165
		if (start < end && state && state->start == start &&
		    !need_resched())
			goto hit_next;
J
Josef Bacik 已提交
1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186
		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);
1187 1188 1189 1190
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1191
		err = split_state(tree, state, prealloc, start);
1192 1193
		if (err)
			extent_io_tree_panic(tree, err);
J
Josef Bacik 已提交
1194 1195 1196 1197
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
1198
			set_state_bits(tree, state, &bits, NULL);
1199
			cache_state(state, cached_state);
1200 1201
			state = clear_state_bit(tree, state, &clear_bits, 0,
						NULL);
J
Josef Bacik 已提交
1202 1203 1204
			if (last_end == (u64)-1)
				goto out;
			start = last_end + 1;
1205 1206 1207
			if (start < end && state && state->start == start &&
			    !need_resched())
				goto hit_next;
J
Josef Bacik 已提交
1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225
		}
		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);
1226 1227 1228 1229
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1230 1231 1232 1233 1234 1235

		/*
		 * Avoid to free 'prealloc' if it can be merged with
		 * the later extent.
		 */
		err = insert_state(tree, prealloc, start, this_end,
1236
				   NULL, NULL, &bits, NULL);
1237 1238
		if (err)
			extent_io_tree_panic(tree, err);
1239
		cache_state(prealloc, cached_state);
J
Josef Bacik 已提交
1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251
		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);
1252 1253 1254 1255
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1256 1257

		err = split_state(tree, state, prealloc, end + 1);
1258 1259
		if (err)
			extent_io_tree_panic(tree, err);
J
Josef Bacik 已提交
1260

1261
		set_state_bits(tree, prealloc, &bits, NULL);
1262
		cache_state(prealloc, cached_state);
1263
		clear_state_bit(tree, prealloc, &clear_bits, 0, NULL);
J
Josef Bacik 已提交
1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280
		prealloc = NULL;
		goto out;
	}

	goto search_again;

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

	return err;

search_again:
	if (start > end)
		goto out;
	spin_unlock(&tree->lock);
1281
	if (gfpflags_allow_blocking(mask))
J
Josef Bacik 已提交
1282
		cond_resched();
1283
	first_iteration = false;
J
Josef Bacik 已提交
1284 1285 1286
	goto again;
}

1287 1288 1289 1290
/* wrappers around set/clear extent bit */
int set_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
		     gfp_t mask)
{
J
Jeff Mahoney 已提交
1291
	return set_extent_bit(tree, start, end, EXTENT_DIRTY, NULL,
1292
			      NULL, mask);
1293 1294 1295
}

int set_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1296
		    unsigned bits, gfp_t mask)
1297
{
J
Jeff Mahoney 已提交
1298
	return set_extent_bit(tree, start, end, bits, NULL,
1299
			      NULL, mask);
1300 1301
}

1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317
int set_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
			   unsigned bits, gfp_t mask,
			   struct extent_changeset *changeset)
{
	/*
	 * 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);

	return __set_extent_bit(tree, start, end, bits, 0, NULL, NULL, mask,
				changeset);
}

1318 1319 1320 1321 1322 1323 1324 1325
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);
}

1326
int clear_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1327
		      unsigned bits, gfp_t mask)
1328
{
1329 1330 1331 1332 1333 1334
	int wake = 0;

	if (bits & EXTENT_LOCKED)
		wake = 1;

	return clear_extent_bit(tree, start, end, bits, wake, 0, NULL, mask);
1335 1336
}

1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350
int clear_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
			     unsigned bits, gfp_t mask,
			     struct extent_changeset *changeset)
{
	/*
	 * Don't support EXTENT_LOCKED case, same reason as
	 * set_record_extent_bits().
	 */
	BUG_ON(bits & EXTENT_LOCKED);

	return __clear_extent_bit(tree, start, end, bits, 0, 0, NULL, mask,
				  changeset);
}

1351
int set_extent_delalloc(struct extent_io_tree *tree, u64 start, u64 end,
1352
			struct extent_state **cached_state, gfp_t mask)
1353 1354
{
	return set_extent_bit(tree, start, end,
1355
			      EXTENT_DELALLOC | EXTENT_UPTODATE,
J
Jeff Mahoney 已提交
1356
			      NULL, cached_state, mask);
1357 1358
}

1359 1360 1361 1362 1363 1364 1365 1366
int set_extent_defrag(struct extent_io_tree *tree, u64 start, u64 end,
		      struct extent_state **cached_state, gfp_t mask)
{
	return set_extent_bit(tree, start, end,
			      EXTENT_DELALLOC | EXTENT_UPTODATE | EXTENT_DEFRAG,
			      NULL, cached_state, mask);
}

1367 1368 1369 1370
int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
		       gfp_t mask)
{
	return clear_extent_bit(tree, start, end,
1371
				EXTENT_DIRTY | EXTENT_DELALLOC |
1372
				EXTENT_DO_ACCOUNTING, 0, 0, NULL, mask);
1373 1374 1375 1376 1377
}

int set_extent_new(struct extent_io_tree *tree, u64 start, u64 end,
		     gfp_t mask)
{
J
Jeff Mahoney 已提交
1378
	return set_extent_bit(tree, start, end, EXTENT_NEW, NULL,
1379
			      NULL, mask);
1380 1381 1382
}

int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end,
1383
			struct extent_state **cached_state, gfp_t mask)
1384
{
L
Liu Bo 已提交
1385
	return set_extent_bit(tree, start, end, EXTENT_UPTODATE, NULL,
J
Jeff Mahoney 已提交
1386
			      cached_state, mask);
1387 1388
}

1389 1390
int clear_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end,
			  struct extent_state **cached_state, gfp_t mask)
1391
{
1392
	return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0,
1393
				cached_state, mask);
1394 1395
}

C
Chris Mason 已提交
1396 1397 1398 1399
/*
 * either insert or lock state struct between start and end use mask to tell
 * us if waiting is desired.
 */
1400
int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1401
		     unsigned bits, struct extent_state **cached_state)
1402 1403 1404
{
	int err;
	u64 failed_start;
1405

1406
	while (1) {
J
Jeff Mahoney 已提交
1407 1408
		err = __set_extent_bit(tree, start, end, EXTENT_LOCKED | bits,
				       EXTENT_LOCKED, &failed_start,
1409
				       cached_state, GFP_NOFS, NULL);
1410
		if (err == -EEXIST) {
1411 1412
			wait_extent_bit(tree, failed_start, end, EXTENT_LOCKED);
			start = failed_start;
1413
		} else
1414 1415 1416 1417 1418 1419
			break;
		WARN_ON(start > end);
	}
	return err;
}

1420
int lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
1421
{
1422
	return lock_extent_bits(tree, start, end, 0, NULL);
1423 1424
}

1425
int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
1426 1427 1428 1429
{
	int err;
	u64 failed_start;

J
Jeff Mahoney 已提交
1430
	err = __set_extent_bit(tree, start, end, EXTENT_LOCKED, EXTENT_LOCKED,
1431
			       &failed_start, NULL, GFP_NOFS, NULL);
Y
Yan Zheng 已提交
1432 1433 1434
	if (err == -EEXIST) {
		if (failed_start > start)
			clear_extent_bit(tree, start, failed_start - 1,
1435
					 EXTENT_LOCKED, 1, 0, NULL, GFP_NOFS);
1436
		return 0;
Y
Yan Zheng 已提交
1437
	}
1438 1439 1440
	return 1;
}

1441 1442 1443 1444 1445 1446 1447
int unlock_extent_cached(struct extent_io_tree *tree, u64 start, u64 end,
			 struct extent_state **cached, gfp_t mask)
{
	return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, cached,
				mask);
}

1448
int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end)
1449
{
1450
	return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL,
1451
				GFP_NOFS);
1452 1453
}

1454
void extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end)
1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468
{
	unsigned long index = start >> PAGE_CACHE_SHIFT;
	unsigned long end_index = end >> PAGE_CACHE_SHIFT;
	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);
		page_cache_release(page);
		index++;
	}
}

1469
void extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end)
1470 1471 1472 1473 1474 1475 1476 1477 1478
{
	unsigned long index = start >> PAGE_CACHE_SHIFT;
	unsigned long end_index = end >> PAGE_CACHE_SHIFT;
	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);
1479
		account_page_redirty(page);
1480 1481 1482 1483 1484
		page_cache_release(page);
		index++;
	}
}

1485 1486 1487
/*
 * helper function to set both pages and extents in the tree writeback
 */
1488
static int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end)
1489 1490 1491 1492 1493 1494 1495
{
	unsigned long index = start >> PAGE_CACHE_SHIFT;
	unsigned long end_index = end >> PAGE_CACHE_SHIFT;
	struct page *page;

	while (index <= end_index) {
		page = find_get_page(tree->mapping, index);
1496
		BUG_ON(!page); /* Pages should be in the extent_io_tree */
1497 1498 1499 1500 1501 1502 1503
		set_page_writeback(page);
		page_cache_release(page);
		index++;
	}
	return 0;
}

C
Chris Mason 已提交
1504 1505 1506 1507
/* 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'
 */
1508 1509
static struct extent_state *
find_first_extent_bit_state(struct extent_io_tree *tree,
1510
			    u64 start, unsigned bits)
C
Chris Mason 已提交
1511 1512 1513 1514 1515 1516 1517 1518 1519
{
	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
1520
	if (!node)
C
Chris Mason 已提交
1521 1522
		goto out;

C
Chris Mason 已提交
1523
	while (1) {
C
Chris Mason 已提交
1524
		state = rb_entry(node, struct extent_state, rb_node);
C
Chris Mason 已提交
1525
		if (state->end >= start && (state->state & bits))
C
Chris Mason 已提交
1526
			return state;
C
Chris Mason 已提交
1527

C
Chris Mason 已提交
1528 1529 1530 1531 1532 1533 1534 1535
		node = rb_next(node);
		if (!node)
			break;
	}
out:
	return NULL;
}

1536 1537 1538 1539 1540
/*
 * 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.
 *
1541
 * If nothing was found, 1 is returned. If found something, return 0.
1542 1543
 */
int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
1544
			  u64 *start_ret, u64 *end_ret, unsigned bits,
1545
			  struct extent_state **cached_state)
1546 1547
{
	struct extent_state *state;
1548
	struct rb_node *n;
1549 1550 1551
	int ret = 1;

	spin_lock(&tree->lock);
1552 1553
	if (cached_state && *cached_state) {
		state = *cached_state;
1554
		if (state->end == start - 1 && extent_state_in_tree(state)) {
1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570
			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;
	}

1571
	state = find_first_extent_bit_state(tree, start, bits);
1572
got_it:
1573
	if (state) {
1574
		cache_state_if_flags(state, cached_state, 0);
1575 1576 1577 1578
		*start_ret = state->start;
		*end_ret = state->end;
		ret = 0;
	}
1579
out:
1580 1581 1582 1583
	spin_unlock(&tree->lock);
	return ret;
}

C
Chris Mason 已提交
1584 1585 1586 1587 1588 1589
/*
 * 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
1590
static noinline u64 find_delalloc_range(struct extent_io_tree *tree,
1591 1592
					u64 *start, u64 *end, u64 max_bytes,
					struct extent_state **cached_state)
1593 1594 1595 1596 1597 1598 1599
{
	struct rb_node *node;
	struct extent_state *state;
	u64 cur_start = *start;
	u64 found = 0;
	u64 total_bytes = 0;

1600
	spin_lock(&tree->lock);
C
Chris Mason 已提交
1601

1602 1603 1604 1605
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1606
	node = tree_search(tree, cur_start);
1607
	if (!node) {
1608 1609
		if (!found)
			*end = (u64)-1;
1610 1611 1612
		goto out;
	}

C
Chris Mason 已提交
1613
	while (1) {
1614
		state = rb_entry(node, struct extent_state, rb_node);
1615 1616
		if (found && (state->start != cur_start ||
			      (state->state & EXTENT_BOUNDARY))) {
1617 1618 1619 1620 1621 1622 1623
			goto out;
		}
		if (!(state->state & EXTENT_DELALLOC)) {
			if (!found)
				*end = state->end;
			goto out;
		}
1624
		if (!found) {
1625
			*start = state->start;
1626 1627 1628
			*cached_state = state;
			atomic_inc(&state->refs);
		}
1629 1630 1631 1632 1633
		found++;
		*end = state->end;
		cur_start = state->end + 1;
		node = rb_next(node);
		total_bytes += state->end - state->start + 1;
1634
		if (total_bytes >= max_bytes)
1635 1636
			break;
		if (!node)
1637 1638 1639
			break;
	}
out:
1640
	spin_unlock(&tree->lock);
1641 1642 1643
	return found;
}

1644 1645 1646
static noinline void __unlock_for_delalloc(struct inode *inode,
					   struct page *locked_page,
					   u64 start, u64 end)
C
Chris Mason 已提交
1647 1648 1649 1650 1651 1652 1653 1654 1655
{
	int ret;
	struct page *pages[16];
	unsigned long index = start >> PAGE_CACHE_SHIFT;
	unsigned long end_index = end >> PAGE_CACHE_SHIFT;
	unsigned long nr_pages = end_index - index + 1;
	int i;

	if (index == locked_page->index && end_index == index)
1656
		return;
C
Chris Mason 已提交
1657

C
Chris Mason 已提交
1658
	while (nr_pages > 0) {
C
Chris Mason 已提交
1659
		ret = find_get_pages_contig(inode->i_mapping, index,
1660 1661
				     min_t(unsigned long, nr_pages,
				     ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692
		for (i = 0; i < ret; i++) {
			if (pages[i] != locked_page)
				unlock_page(pages[i]);
			page_cache_release(pages[i]);
		}
		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)
{
	unsigned long index = delalloc_start >> PAGE_CACHE_SHIFT;
	unsigned long start_index = index;
	unsigned long end_index = delalloc_end >> PAGE_CACHE_SHIFT;
	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
1693
	while (nrpages > 0) {
C
Chris Mason 已提交
1694
		ret = find_get_pages_contig(inode->i_mapping, index,
1695 1696
				     min_t(unsigned long,
				     nrpages, ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1697 1698 1699 1700 1701 1702 1703 1704 1705 1706
		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
			 */
1707
			if (pages[i] != locked_page) {
C
Chris Mason 已提交
1708
				lock_page(pages[i]);
1709 1710
				if (!PageDirty(pages[i]) ||
				    pages[i]->mapping != inode->i_mapping) {
1711 1712 1713 1714 1715 1716
					ret = -EAGAIN;
					unlock_page(pages[i]);
					page_cache_release(pages[i]);
					goto done;
				}
			}
C
Chris Mason 已提交
1717
			page_cache_release(pages[i]);
1718
			pages_locked++;
C
Chris Mason 已提交
1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740
		}
		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)) <<
			      PAGE_CACHE_SHIFT);
	}
	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
 */
1741 1742 1743 1744
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
1745 1746 1747 1748
{
	u64 delalloc_start;
	u64 delalloc_end;
	u64 found;
1749
	struct extent_state *cached_state = NULL;
C
Chris Mason 已提交
1750 1751 1752 1753 1754 1755 1756 1757
	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,
1758
				    max_bytes, &cached_state);
C
Chris Mason 已提交
1759
	if (!found || delalloc_end <= *start) {
C
Chris Mason 已提交
1760 1761
		*start = delalloc_start;
		*end = delalloc_end;
1762
		free_extent_state(cached_state);
L
Liu Bo 已提交
1763
		return 0;
C
Chris Mason 已提交
1764 1765
	}

C
Chris Mason 已提交
1766 1767 1768 1769 1770
	/*
	 * 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
1771
	if (delalloc_start < *start)
C
Chris Mason 已提交
1772 1773
		delalloc_start = *start;

C
Chris Mason 已提交
1774 1775 1776
	/*
	 * make sure to limit the number of pages we try to lock down
	 */
1777 1778
	if (delalloc_end + 1 - delalloc_start > max_bytes)
		delalloc_end = delalloc_start + max_bytes - 1;
C
Chris Mason 已提交
1779

C
Chris Mason 已提交
1780 1781 1782 1783 1784 1785 1786
	/* 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
		 */
1787
		free_extent_state(cached_state);
1788
		cached_state = NULL;
C
Chris Mason 已提交
1789
		if (!loops) {
1790
			max_bytes = PAGE_CACHE_SIZE;
C
Chris Mason 已提交
1791 1792 1793 1794 1795 1796 1797
			loops = 1;
			goto again;
		} else {
			found = 0;
			goto out_failed;
		}
	}
1798
	BUG_ON(ret); /* Only valid values are 0 and -EAGAIN */
C
Chris Mason 已提交
1799 1800

	/* step three, lock the state bits for the whole range */
1801
	lock_extent_bits(tree, delalloc_start, delalloc_end, 0, &cached_state);
C
Chris Mason 已提交
1802 1803 1804

	/* then test to make sure it is all still delalloc */
	ret = test_range_bit(tree, delalloc_start, delalloc_end,
1805
			     EXTENT_DELALLOC, 1, cached_state);
C
Chris Mason 已提交
1806
	if (!ret) {
1807 1808
		unlock_extent_cached(tree, delalloc_start, delalloc_end,
				     &cached_state, GFP_NOFS);
C
Chris Mason 已提交
1809 1810 1811 1812 1813
		__unlock_for_delalloc(inode, locked_page,
			      delalloc_start, delalloc_end);
		cond_resched();
		goto again;
	}
1814
	free_extent_state(cached_state);
C
Chris Mason 已提交
1815 1816 1817 1818 1819 1820
	*start = delalloc_start;
	*end = delalloc_end;
out_failed:
	return found;
}

1821
void extent_clear_unlock_delalloc(struct inode *inode, u64 start, u64 end,
1822
				 struct page *locked_page,
1823
				 unsigned clear_bits,
1824
				 unsigned long page_ops)
C
Chris Mason 已提交
1825
{
1826
	struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
C
Chris Mason 已提交
1827 1828 1829 1830 1831 1832
	int ret;
	struct page *pages[16];
	unsigned long index = start >> PAGE_CACHE_SHIFT;
	unsigned long end_index = end >> PAGE_CACHE_SHIFT;
	unsigned long nr_pages = end_index - index + 1;
	int i;
1833

1834
	clear_extent_bit(tree, start, end, clear_bits, 1, 0, NULL, GFP_NOFS);
1835
	if (page_ops == 0)
1836
		return;
C
Chris Mason 已提交
1837

1838 1839 1840
	if ((page_ops & PAGE_SET_ERROR) && nr_pages > 0)
		mapping_set_error(inode->i_mapping, -EIO);

C
Chris Mason 已提交
1841
	while (nr_pages > 0) {
C
Chris Mason 已提交
1842
		ret = find_get_pages_contig(inode->i_mapping, index,
1843 1844
				     min_t(unsigned long,
				     nr_pages, ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1845
		for (i = 0; i < ret; i++) {
1846

1847
			if (page_ops & PAGE_SET_PRIVATE2)
1848 1849
				SetPagePrivate2(pages[i]);

C
Chris Mason 已提交
1850 1851 1852 1853
			if (pages[i] == locked_page) {
				page_cache_release(pages[i]);
				continue;
			}
1854
			if (page_ops & PAGE_CLEAR_DIRTY)
C
Chris Mason 已提交
1855
				clear_page_dirty_for_io(pages[i]);
1856
			if (page_ops & PAGE_SET_WRITEBACK)
C
Chris Mason 已提交
1857
				set_page_writeback(pages[i]);
1858 1859
			if (page_ops & PAGE_SET_ERROR)
				SetPageError(pages[i]);
1860
			if (page_ops & PAGE_END_WRITEBACK)
C
Chris Mason 已提交
1861
				end_page_writeback(pages[i]);
1862
			if (page_ops & PAGE_UNLOCK)
1863
				unlock_page(pages[i]);
C
Chris Mason 已提交
1864 1865 1866 1867 1868 1869 1870 1871
			page_cache_release(pages[i]);
		}
		nr_pages -= ret;
		index += ret;
		cond_resched();
	}
}

C
Chris Mason 已提交
1872 1873 1874 1875 1876
/*
 * 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.
 */
1877 1878
u64 count_range_bits(struct extent_io_tree *tree,
		     u64 *start, u64 search_end, u64 max_bytes,
1879
		     unsigned bits, int contig)
1880 1881 1882 1883 1884
{
	struct rb_node *node;
	struct extent_state *state;
	u64 cur_start = *start;
	u64 total_bytes = 0;
1885
	u64 last = 0;
1886 1887
	int found = 0;

1888
	if (WARN_ON(search_end <= cur_start))
1889 1890
		return 0;

1891
	spin_lock(&tree->lock);
1892 1893 1894 1895 1896 1897 1898 1899
	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.
	 */
1900
	node = tree_search(tree, cur_start);
C
Chris Mason 已提交
1901
	if (!node)
1902 1903
		goto out;

C
Chris Mason 已提交
1904
	while (1) {
1905 1906 1907
		state = rb_entry(node, struct extent_state, rb_node);
		if (state->start > search_end)
			break;
1908 1909 1910
		if (contig && found && state->start > last + 1)
			break;
		if (state->end >= cur_start && (state->state & bits) == bits) {
1911 1912 1913 1914 1915
			total_bytes += min(search_end, state->end) + 1 -
				       max(cur_start, state->start);
			if (total_bytes >= max_bytes)
				break;
			if (!found) {
1916
				*start = max(cur_start, state->start);
1917 1918
				found = 1;
			}
1919 1920 1921
			last = state->end;
		} else if (contig && found) {
			break;
1922 1923 1924 1925 1926 1927
		}
		node = rb_next(node);
		if (!node)
			break;
	}
out:
1928
	spin_unlock(&tree->lock);
1929 1930
	return total_bytes;
}
1931

C
Chris Mason 已提交
1932 1933 1934 1935
/*
 * set the private field for a given byte offset in the tree.  If there isn't
 * an extent_state there already, this does nothing.
 */
1936
static int set_state_private(struct extent_io_tree *tree, u64 start, u64 private)
1937 1938 1939 1940 1941
{
	struct rb_node *node;
	struct extent_state *state;
	int ret = 0;

1942
	spin_lock(&tree->lock);
1943 1944 1945 1946
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1947
	node = tree_search(tree, start);
1948
	if (!node) {
1949 1950 1951 1952 1953 1954 1955 1956 1957 1958
		ret = -ENOENT;
		goto out;
	}
	state = rb_entry(node, struct extent_state, rb_node);
	if (state->start != start) {
		ret = -ENOENT;
		goto out;
	}
	state->private = private;
out:
1959
	spin_unlock(&tree->lock);
1960 1961 1962 1963 1964 1965 1966 1967 1968
	return ret;
}

int get_state_private(struct extent_io_tree *tree, u64 start, u64 *private)
{
	struct rb_node *node;
	struct extent_state *state;
	int ret = 0;

1969
	spin_lock(&tree->lock);
1970 1971 1972 1973
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1974
	node = tree_search(tree, start);
1975
	if (!node) {
1976 1977 1978 1979 1980 1981 1982 1983 1984 1985
		ret = -ENOENT;
		goto out;
	}
	state = rb_entry(node, struct extent_state, rb_node);
	if (state->start != start) {
		ret = -ENOENT;
		goto out;
	}
	*private = state->private;
out:
1986
	spin_unlock(&tree->lock);
1987 1988 1989 1990 1991
	return ret;
}

/*
 * searches a range in the state tree for a given mask.
1992
 * If 'filled' == 1, this returns 1 only if every extent in the tree
1993 1994 1995 1996
 * 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,
1997
		   unsigned bits, int filled, struct extent_state *cached)
1998 1999 2000 2001 2002
{
	struct extent_state *state = NULL;
	struct rb_node *node;
	int bitset = 0;

2003
	spin_lock(&tree->lock);
2004
	if (cached && extent_state_in_tree(cached) && cached->start <= start &&
2005
	    cached->end > start)
2006 2007 2008
		node = &cached->rb_node;
	else
		node = tree_search(tree, start);
2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027
	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;
		}
2028 2029 2030 2031

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

2032 2033 2034 2035 2036 2037 2038 2039 2040 2041
		start = state->end + 1;
		if (start > end)
			break;
		node = rb_next(node);
		if (!node) {
			if (filled)
				bitset = 0;
			break;
		}
	}
2042
	spin_unlock(&tree->lock);
2043 2044 2045 2046 2047 2048 2049
	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
 */
2050
static void check_page_uptodate(struct extent_io_tree *tree, struct page *page)
2051
{
M
Miao Xie 已提交
2052
	u64 start = page_offset(page);
2053
	u64 end = start + PAGE_CACHE_SIZE - 1;
2054
	if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1, NULL))
2055 2056 2057
		SetPageUptodate(page);
}

2058
int free_io_failure(struct inode *inode, struct io_failure_record *rec)
2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070
{
	int ret;
	int err = 0;
	struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;

	set_state_private(failure_tree, rec->start, 0);
	ret = clear_extent_bits(failure_tree, rec->start,
				rec->start + rec->len - 1,
				EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
	if (ret)
		err = ret;

D
David Woodhouse 已提交
2071 2072 2073 2074 2075
	ret = clear_extent_bits(&BTRFS_I(inode)->io_tree, rec->start,
				rec->start + rec->len - 1,
				EXTENT_DAMAGED, GFP_NOFS);
	if (ret && !err)
		err = ret;
2076 2077 2078 2079 2080 2081 2082 2083 2084 2085

	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.
2086
 * to avoid any synchronization issues, wait for the data after writing, which
2087 2088 2089 2090
 * 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.
 */
2091 2092
int repair_io_failure(struct inode *inode, u64 start, u64 length, u64 logical,
		      struct page *page, unsigned int pg_offset, int mirror_num)
2093
{
2094
	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2095 2096 2097 2098 2099
	struct bio *bio;
	struct btrfs_device *dev;
	u64 map_length = 0;
	u64 sector;
	struct btrfs_bio *bbio = NULL;
D
David Woodhouse 已提交
2100
	struct btrfs_mapping_tree *map_tree = &fs_info->mapping_tree;
2101 2102
	int ret;

2103
	ASSERT(!(fs_info->sb->s_flags & MS_RDONLY));
2104 2105
	BUG_ON(!mirror_num);

D
David Woodhouse 已提交
2106 2107 2108 2109
	/* we can't repair anything in raid56 yet */
	if (btrfs_is_parity_mirror(map_tree, logical, length, mirror_num))
		return 0;

2110
	bio = btrfs_io_bio_alloc(GFP_NOFS, 1);
2111 2112
	if (!bio)
		return -EIO;
2113
	bio->bi_iter.bi_size = 0;
2114 2115
	map_length = length;

2116
	ret = btrfs_map_block(fs_info, WRITE, logical,
2117 2118 2119 2120 2121 2122 2123
			      &map_length, &bbio, mirror_num);
	if (ret) {
		bio_put(bio);
		return -EIO;
	}
	BUG_ON(mirror_num != bbio->mirror_num);
	sector = bbio->stripes[mirror_num-1].physical >> 9;
2124
	bio->bi_iter.bi_sector = sector;
2125
	dev = bbio->stripes[mirror_num-1].dev;
2126
	btrfs_put_bbio(bbio);
2127 2128 2129 2130 2131
	if (!dev || !dev->bdev || !dev->writeable) {
		bio_put(bio);
		return -EIO;
	}
	bio->bi_bdev = dev->bdev;
2132
	bio_add_page(bio, page, length, pg_offset);
2133

2134
	if (btrfsic_submit_bio_wait(WRITE_SYNC, bio)) {
2135 2136
		/* try to remap that extent elsewhere? */
		bio_put(bio);
2137
		btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_WRITE_ERRS);
2138 2139 2140
		return -EIO;
	}

2141 2142
	btrfs_info_rl_in_rcu(fs_info,
		"read error corrected: ino %llu off %llu (dev %s sector %llu)",
2143 2144
				  btrfs_ino(inode), start,
				  rcu_str_deref(dev->name), sector);
2145 2146 2147 2148
	bio_put(bio);
	return 0;
}

2149 2150 2151 2152 2153
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);
2154
	int ret = 0;
2155

2156 2157 2158
	if (root->fs_info->sb->s_flags & MS_RDONLY)
		return -EROFS;

2159
	for (i = 0; i < num_pages; i++) {
2160
		struct page *p = eb->pages[i];
2161 2162 2163 2164

		ret = repair_io_failure(root->fs_info->btree_inode, start,
					PAGE_CACHE_SIZE, start, p,
					start - page_offset(p), mirror_num);
2165 2166 2167 2168 2169 2170 2171 2172
		if (ret)
			break;
		start += PAGE_CACHE_SIZE;
	}

	return ret;
}

2173 2174 2175 2176
/*
 * 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
 */
2177 2178
int clean_io_failure(struct inode *inode, u64 start, struct page *page,
		     unsigned int pg_offset)
2179 2180 2181 2182
{
	u64 private;
	u64 private_failure;
	struct io_failure_record *failrec;
2183
	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207
	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;

	ret = get_state_private(&BTRFS_I(inode)->io_failure_tree, start,
				&private_failure);
	if (ret)
		return 0;

	failrec = (struct io_failure_record *)(unsigned long) private_failure;
	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;
	}
2208 2209
	if (fs_info->sb->s_flags & MS_RDONLY)
		goto out;
2210 2211 2212 2213 2214 2215 2216

	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);

2217 2218
	if (state && state->start <= failrec->start &&
	    state->end >= failrec->start + failrec->len - 1) {
2219 2220
		num_copies = btrfs_num_copies(fs_info, failrec->logical,
					      failrec->len);
2221
		if (num_copies > 1)  {
2222
			repair_io_failure(inode, start, failrec->len,
2223
					  failrec->logical, page,
2224
					  pg_offset, failrec->failed_mirror);
2225 2226 2227 2228
		}
	}

out:
2229
	free_io_failure(inode, failrec);
2230

2231
	return 0;
2232 2233
}

2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258
/*
 * 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);

2259
		failrec = (struct io_failure_record *)(unsigned long)state->private;
2260 2261 2262 2263 2264 2265 2266 2267
		free_extent_state(state);
		kfree(failrec);

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

2268 2269
int btrfs_get_io_failure_record(struct inode *inode, u64 start, u64 end,
				struct io_failure_record **failrec_ret)
2270
{
2271
	struct io_failure_record *failrec;
2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284
	u64 private;
	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;

	ret = get_state_private(failure_tree, start, &private);
	if (ret) {
		failrec = kzalloc(sizeof(*failrec), GFP_NOFS);
		if (!failrec)
			return -ENOMEM;
2285

2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299
		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;
		}

2300
		if (em->start > start || em->start + em->len <= start) {
2301 2302 2303 2304
			free_extent_map(em);
			em = NULL;
		}
		read_unlock(&em_tree->lock);
2305
		if (!em) {
2306 2307 2308
			kfree(failrec);
			return -EIO;
		}
2309

2310 2311 2312 2313 2314 2315 2316 2317
		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);
		}
2318 2319 2320 2321

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

2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340
		failrec->logical = logical;
		free_extent_map(em);

		/* set the bits in the private failure tree */
		ret = set_extent_bits(failure_tree, start, end,
					EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
		if (ret >= 0)
			ret = set_state_private(failure_tree, start,
						(u64)(unsigned long)failrec);
		/* set the bits in the inode's tree */
		if (ret >= 0)
			ret = set_extent_bits(tree, start, end, EXTENT_DAMAGED,
						GFP_NOFS);
		if (ret < 0) {
			kfree(failrec);
			return ret;
		}
	} else {
		failrec = (struct io_failure_record *)(unsigned long)private;
2341
		pr_debug("Get IO Failure Record: (found) logical=%llu, start=%llu, len=%llu, validation=%d\n",
2342 2343 2344 2345 2346 2347 2348 2349
			 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.
		 */
	}
2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360

	*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;

2361 2362
	num_copies = btrfs_num_copies(BTRFS_I(inode)->root->fs_info,
				      failrec->logical, failrec->len);
2363 2364 2365 2366 2367 2368
	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.
		 */
2369
		pr_debug("Check Repairable: cannot repair, num_copies=%d, next_mirror %d, failed_mirror %d\n",
2370
			 num_copies, failrec->this_mirror, failed_mirror);
2371
		return 0;
2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407
	}

	/*
	 * 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++;
	}

2408
	if (failrec->this_mirror > num_copies) {
2409
		pr_debug("Check Repairable: (fail) num_copies=%d, next_mirror %d, failed_mirror %d\n",
2410
			 num_copies, failrec->this_mirror, failed_mirror);
2411
		return 0;
2412 2413
	}

2414 2415 2416 2417 2418 2419 2420
	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,
2421
				    bio_end_io_t *endio_func, void *data)
2422 2423 2424 2425 2426
{
	struct bio *bio;
	struct btrfs_io_bio *btrfs_failed_bio;
	struct btrfs_io_bio *btrfs_bio;

2427
	bio = btrfs_io_bio_alloc(GFP_NOFS, 1);
2428 2429 2430 2431
	if (!bio)
		return NULL;

	bio->bi_end_io = endio_func;
2432
	bio->bi_iter.bi_sector = failrec->logical >> 9;
2433
	bio->bi_bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
2434
	bio->bi_iter.bi_size = 0;
2435
	bio->bi_private = data;
2436

2437 2438 2439 2440 2441 2442 2443
	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;
2444 2445
		icsum *= csum_size;
		memcpy(btrfs_bio->csum, btrfs_failed_bio->csum + icsum,
2446 2447 2448
		       csum_size);
	}

2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492
	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;

	BUG_ON(failed_bio->bi_rw & REQ_WRITE);

	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),
2493 2494
				      (int)phy_offset, failed_bio->bi_end_io,
				      NULL);
2495 2496 2497 2498
	if (!bio) {
		free_io_failure(inode, failrec);
		return -EIO;
	}
2499

2500 2501
	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);
2502

2503 2504 2505
	ret = tree->ops->submit_bio_hook(inode, read_mode, bio,
					 failrec->this_mirror,
					 failrec->bio_flags, 0);
2506
	if (ret) {
2507
		free_io_failure(inode, failrec);
2508 2509 2510
		bio_put(bio);
	}

2511
	return ret;
2512 2513
}

2514 2515
/* lots and lots of room for performance fixes in the end_bio funcs */

2516
void end_extent_writepage(struct page *page, int err, u64 start, u64 end)
2517 2518 2519
{
	int uptodate = (err == 0);
	struct extent_io_tree *tree;
2520
	int ret = 0;
2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533

	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);
2534 2535
		ret = ret < 0 ? ret : -EIO;
		mapping_set_error(page->mapping, ret);
2536 2537 2538
	}
}

2539 2540 2541 2542 2543 2544 2545 2546 2547
/*
 * 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.
 */
2548
static void end_bio_extent_writepage(struct bio *bio)
2549
{
2550
	struct bio_vec *bvec;
2551 2552
	u64 start;
	u64 end;
2553
	int i;
2554

2555
	bio_for_each_segment_all(bvec, bio, i) {
2556
		struct page *page = bvec->bv_page;
2557

2558 2559 2560 2561 2562
		/* 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.  */
2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573
		if (bvec->bv_offset || bvec->bv_len != PAGE_CACHE_SIZE) {
			if (bvec->bv_offset + bvec->bv_len != PAGE_CACHE_SIZE)
				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);
		}
2574

2575 2576
		start = page_offset(page);
		end = start + bvec->bv_offset + bvec->bv_len - 1;
2577

2578
		end_extent_writepage(page, bio->bi_error, start, end);
2579
		end_page_writeback(page);
2580
	}
2581

2582 2583 2584
	bio_put(bio);
}

2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596
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);
}

2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607
/*
 * 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.
 */
2608
static void end_bio_extent_readpage(struct bio *bio)
2609
{
2610
	struct bio_vec *bvec;
2611
	int uptodate = !bio->bi_error;
2612
	struct btrfs_io_bio *io_bio = btrfs_io_bio(bio);
2613
	struct extent_io_tree *tree;
2614
	u64 offset = 0;
2615 2616
	u64 start;
	u64 end;
2617
	u64 len;
2618 2619
	u64 extent_start = 0;
	u64 extent_len = 0;
2620
	int mirror;
2621
	int ret;
2622
	int i;
2623

2624
	bio_for_each_segment_all(bvec, bio, i) {
2625
		struct page *page = bvec->bv_page;
2626
		struct inode *inode = page->mapping->host;
2627

2628
		pr_debug("end_bio_extent_readpage: bi_sector=%llu, err=%d, "
2629 2630
			 "mirror=%u\n", (u64)bio->bi_iter.bi_sector,
			 bio->bi_error, io_bio->mirror_num);
2631
		tree = &BTRFS_I(inode)->io_tree;
2632

2633 2634 2635 2636 2637
		/* 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.  */
2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648
		if (bvec->bv_offset || bvec->bv_len != PAGE_CACHE_SIZE) {
			if (bvec->bv_offset + bvec->bv_len != PAGE_CACHE_SIZE)
				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);
		}
2649

2650 2651
		start = page_offset(page);
		end = start + bvec->bv_offset + bvec->bv_len - 1;
2652
		len = bvec->bv_len;
2653

2654
		mirror = io_bio->mirror_num;
2655 2656
		if (likely(uptodate && tree->ops &&
			   tree->ops->readpage_end_io_hook)) {
2657 2658 2659
			ret = tree->ops->readpage_end_io_hook(io_bio, offset,
							      page, start, end,
							      mirror);
2660
			if (ret)
2661
				uptodate = 0;
2662
			else
2663
				clean_io_failure(inode, start, page, 0);
2664
		}
2665

2666 2667 2668 2669
		if (likely(uptodate))
			goto readpage_ok;

		if (tree->ops && tree->ops->readpage_io_failed_hook) {
2670
			ret = tree->ops->readpage_io_failed_hook(page, mirror);
2671
			if (!ret && !bio->bi_error)
2672
				uptodate = 1;
2673
		} else {
2674 2675 2676 2677 2678 2679 2680 2681 2682 2683
			/*
			 * 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.
			 */
2684 2685
			ret = bio_readpage_error(bio, offset, page, start, end,
						 mirror);
2686
			if (ret == 0) {
2687
				uptodate = !bio->bi_error;
2688
				offset += len;
2689 2690 2691
				continue;
			}
		}
2692
readpage_ok:
2693
		if (likely(uptodate)) {
2694 2695
			loff_t i_size = i_size_read(inode);
			pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT;
2696
			unsigned off;
2697 2698

			/* Zero out the end if this page straddles i_size */
2699 2700 2701
			off = i_size & (PAGE_CACHE_SIZE-1);
			if (page->index == end_index && off)
				zero_user_segment(page, off, PAGE_CACHE_SIZE);
2702
			SetPageUptodate(page);
2703
		} else {
2704 2705
			ClearPageUptodate(page);
			SetPageError(page);
2706
		}
2707
		unlock_page(page);
2708
		offset += len;
2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730

		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;
		}
2731
	}
2732

2733 2734 2735
	if (extent_len)
		endio_readpage_release_extent(tree, extent_start, extent_len,
					      uptodate);
2736
	if (io_bio->end_io)
2737
		io_bio->end_io(io_bio, bio->bi_error);
2738 2739 2740
	bio_put(bio);
}

2741 2742 2743 2744
/*
 * 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
 */
2745 2746 2747
struct bio *
btrfs_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
		gfp_t gfp_flags)
2748
{
2749
	struct btrfs_io_bio *btrfs_bio;
2750 2751
	struct bio *bio;

2752
	bio = bio_alloc_bioset(gfp_flags, nr_vecs, btrfs_bioset);
2753 2754

	if (bio == NULL && (current->flags & PF_MEMALLOC)) {
2755 2756 2757 2758
		while (!bio && (nr_vecs /= 2)) {
			bio = bio_alloc_bioset(gfp_flags,
					       nr_vecs, btrfs_bioset);
		}
2759 2760 2761 2762
	}

	if (bio) {
		bio->bi_bdev = bdev;
2763
		bio->bi_iter.bi_sector = first_sector;
2764 2765 2766 2767
		btrfs_bio = btrfs_io_bio(bio);
		btrfs_bio->csum = NULL;
		btrfs_bio->csum_allocated = NULL;
		btrfs_bio->end_io = NULL;
2768 2769 2770 2771
	}
	return bio;
}

2772 2773
struct bio *btrfs_bio_clone(struct bio *bio, gfp_t gfp_mask)
{
2774 2775
	struct btrfs_io_bio *btrfs_bio;
	struct bio *new;
2776

2777 2778 2779 2780 2781 2782
	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;
2783 2784

#ifdef CONFIG_BLK_CGROUP
2785 2786 2787
		/* FIXME, put this into bio_clone_bioset */
		if (bio->bi_css)
			bio_associate_blkcg(new, bio->bi_css);
2788
#endif
2789 2790 2791
	}
	return new;
}
2792 2793 2794 2795

/* this also allocates from the btrfs_bioset */
struct bio *btrfs_io_bio_alloc(gfp_t gfp_mask, unsigned int nr_iovecs)
{
2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806
	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;
2807 2808 2809
}


2810 2811
static int __must_check submit_one_bio(int rw, struct bio *bio,
				       int mirror_num, unsigned long bio_flags)
2812 2813
{
	int ret = 0;
2814 2815 2816 2817 2818
	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
2819
	start = page_offset(page) + bvec->bv_offset;
2820

2821
	bio->bi_private = NULL;
2822 2823 2824

	bio_get(bio);

2825
	if (tree->ops && tree->ops->submit_bio_hook)
2826
		ret = tree->ops->submit_bio_hook(page->mapping->host, rw, bio,
2827
					   mirror_num, bio_flags, start);
2828
	else
2829
		btrfsic_submit_bio(rw, bio);
2830

2831 2832 2833 2834
	bio_put(bio);
	return ret;
}

2835
static int merge_bio(int rw, struct extent_io_tree *tree, struct page *page,
2836 2837 2838 2839 2840
		     unsigned long offset, size_t size, struct bio *bio,
		     unsigned long bio_flags)
{
	int ret = 0;
	if (tree->ops && tree->ops->merge_bio_hook)
2841
		ret = tree->ops->merge_bio_hook(rw, page, offset, size, bio,
2842 2843 2844 2845 2846 2847
						bio_flags);
	BUG_ON(ret < 0);
	return ret;

}

2848
static int submit_extent_page(int rw, struct extent_io_tree *tree,
2849
			      struct writeback_control *wbc,
2850 2851 2852 2853 2854
			      struct page *page, sector_t sector,
			      size_t size, unsigned long offset,
			      struct block_device *bdev,
			      struct bio **bio_ret,
			      unsigned long max_pages,
2855
			      bio_end_io_t end_io_func,
C
Chris Mason 已提交
2856 2857
			      int mirror_num,
			      unsigned long prev_bio_flags,
2858 2859
			      unsigned long bio_flags,
			      bool force_bio_submit)
2860 2861 2862
{
	int ret = 0;
	struct bio *bio;
C
Chris Mason 已提交
2863 2864
	int contig = 0;
	int old_compressed = prev_bio_flags & EXTENT_BIO_COMPRESSED;
2865
	size_t page_size = min_t(size_t, size, PAGE_CACHE_SIZE);
2866 2867 2868

	if (bio_ret && *bio_ret) {
		bio = *bio_ret;
C
Chris Mason 已提交
2869
		if (old_compressed)
2870
			contig = bio->bi_iter.bi_sector == sector;
C
Chris Mason 已提交
2871
		else
K
Kent Overstreet 已提交
2872
			contig = bio_end_sector(bio) == sector;
C
Chris Mason 已提交
2873 2874

		if (prev_bio_flags != bio_flags || !contig ||
2875
		    force_bio_submit ||
2876
		    merge_bio(rw, tree, page, offset, page_size, bio, bio_flags) ||
C
Chris Mason 已提交
2877 2878 2879
		    bio_add_page(bio, page, page_size, offset) < page_size) {
			ret = submit_one_bio(rw, bio, mirror_num,
					     prev_bio_flags);
2880 2881
			if (ret < 0) {
				*bio_ret = NULL;
2882
				return ret;
2883
			}
2884 2885
			bio = NULL;
		} else {
2886 2887
			if (wbc)
				wbc_account_io(wbc, page, page_size);
2888 2889 2890
			return 0;
		}
	}
C
Chris Mason 已提交
2891

2892 2893
	bio = btrfs_bio_alloc(bdev, sector, BIO_MAX_PAGES,
			GFP_NOFS | __GFP_HIGH);
2894 2895
	if (!bio)
		return -ENOMEM;
2896

C
Chris Mason 已提交
2897
	bio_add_page(bio, page, page_size, offset);
2898 2899
	bio->bi_end_io = end_io_func;
	bio->bi_private = tree;
2900 2901 2902 2903
	if (wbc) {
		wbc_init_bio(wbc, bio);
		wbc_account_io(wbc, page, page_size);
	}
2904

C
Chris Mason 已提交
2905
	if (bio_ret)
2906
		*bio_ret = bio;
C
Chris Mason 已提交
2907
	else
C
Chris Mason 已提交
2908
		ret = submit_one_bio(rw, bio, mirror_num, bio_flags);
2909 2910 2911 2912

	return ret;
}

2913 2914
static void attach_extent_buffer_page(struct extent_buffer *eb,
				      struct page *page)
2915 2916 2917 2918
{
	if (!PagePrivate(page)) {
		SetPagePrivate(page);
		page_cache_get(page);
J
Josef Bacik 已提交
2919 2920 2921
		set_page_private(page, (unsigned long)eb);
	} else {
		WARN_ON(page->private != (unsigned long)eb);
2922 2923 2924
	}
}

J
Josef Bacik 已提交
2925
void set_page_extent_mapped(struct page *page)
2926
{
J
Josef Bacik 已提交
2927 2928 2929 2930 2931
	if (!PagePrivate(page)) {
		SetPagePrivate(page);
		page_cache_get(page);
		set_page_private(page, EXTENT_PAGE_PRIVATE);
	}
2932 2933
}

2934 2935 2936 2937 2938 2939 2940 2941 2942
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;
2943
		if (extent_map_in_tree(em) && start >= em->start &&
2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960
		    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;
}
2961 2962 2963 2964
/*
 * 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)
2965
 * XXX JDM: This needs looking at to ensure proper page locking
2966
 */
2967 2968 2969
static int __do_readpage(struct extent_io_tree *tree,
			 struct page *page,
			 get_extent_t *get_extent,
2970
			 struct extent_map **em_cached,
2971
			 struct bio **bio, int mirror_num,
2972 2973
			 unsigned long *bio_flags, int rw,
			 u64 *prev_em_start)
2974 2975
{
	struct inode *inode = page->mapping->host;
M
Miao Xie 已提交
2976
	u64 start = page_offset(page);
2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988
	u64 page_end = start + PAGE_CACHE_SIZE - 1;
	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;
2989
	int parent_locked = *bio_flags & EXTENT_BIO_PARENT_LOCKED;
2990
	size_t pg_offset = 0;
2991
	size_t iosize;
C
Chris Mason 已提交
2992
	size_t disk_io_size;
2993
	size_t blocksize = inode->i_sb->s_blocksize;
2994
	unsigned long this_bio_flag = *bio_flags & EXTENT_BIO_PARENT_LOCKED;
2995 2996 2997

	set_page_extent_mapped(page);

2998
	end = page_end;
D
Dan Magenheimer 已提交
2999 3000 3001
	if (!PageUptodate(page)) {
		if (cleancache_get_page(page) == 0) {
			BUG_ON(blocksize != PAGE_SIZE);
3002
			unlock_extent(tree, start, end);
D
Dan Magenheimer 已提交
3003 3004 3005 3006
			goto out;
		}
	}

C
Chris Mason 已提交
3007 3008 3009 3010 3011 3012
	if (page->index == last_byte >> PAGE_CACHE_SHIFT) {
		char *userpage;
		size_t zero_offset = last_byte & (PAGE_CACHE_SIZE - 1);

		if (zero_offset) {
			iosize = PAGE_CACHE_SIZE - zero_offset;
3013
			userpage = kmap_atomic(page);
C
Chris Mason 已提交
3014 3015
			memset(userpage + zero_offset, 0, iosize);
			flush_dcache_page(page);
3016
			kunmap_atomic(userpage);
C
Chris Mason 已提交
3017 3018
		}
	}
3019
	while (cur <= end) {
3020
		unsigned long pnr = (last_byte >> PAGE_CACHE_SHIFT) + 1;
3021
		bool force_bio_submit = false;
3022

3023 3024
		if (cur >= last_byte) {
			char *userpage;
3025 3026
			struct extent_state *cached = NULL;

3027
			iosize = PAGE_CACHE_SIZE - pg_offset;
3028
			userpage = kmap_atomic(page);
3029
			memset(userpage + pg_offset, 0, iosize);
3030
			flush_dcache_page(page);
3031
			kunmap_atomic(userpage);
3032
			set_extent_uptodate(tree, cur, cur + iosize - 1,
3033
					    &cached, GFP_NOFS);
3034 3035 3036 3037
			if (!parent_locked)
				unlock_extent_cached(tree, cur,
						     cur + iosize - 1,
						     &cached, GFP_NOFS);
3038 3039
			break;
		}
3040 3041
		em = __get_extent_map(inode, page, pg_offset, cur,
				      end - cur + 1, get_extent, em_cached);
3042
		if (IS_ERR_OR_NULL(em)) {
3043
			SetPageError(page);
3044 3045
			if (!parent_locked)
				unlock_extent(tree, cur, end);
3046 3047 3048 3049 3050 3051
			break;
		}
		extent_offset = cur - em->start;
		BUG_ON(extent_map_end(em) <= cur);
		BUG_ON(end < cur);

3052
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
3053
			this_bio_flag |= EXTENT_BIO_COMPRESSED;
3054 3055 3056
			extent_set_compress_type(&this_bio_flag,
						 em->compress_type);
		}
C
Chris Mason 已提交
3057

3058 3059
		iosize = min(extent_map_end(em) - cur, end - cur + 1);
		cur_end = min(extent_map_end(em) - 1, end);
3060
		iosize = ALIGN(iosize, blocksize);
C
Chris Mason 已提交
3061 3062 3063 3064 3065 3066 3067
		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;
		}
3068 3069
		bdev = em->bdev;
		block_start = em->block_start;
Y
Yan Zheng 已提交
3070 3071
		if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			block_start = EXTENT_MAP_HOLE;
3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114

		/*
		 * 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;

3115 3116 3117 3118 3119 3120
		free_extent_map(em);
		em = NULL;

		/* we've found a hole, just zero and go on */
		if (block_start == EXTENT_MAP_HOLE) {
			char *userpage;
3121 3122
			struct extent_state *cached = NULL;

3123
			userpage = kmap_atomic(page);
3124
			memset(userpage + pg_offset, 0, iosize);
3125
			flush_dcache_page(page);
3126
			kunmap_atomic(userpage);
3127 3128

			set_extent_uptodate(tree, cur, cur + iosize - 1,
3129
					    &cached, GFP_NOFS);
3130 3131 3132 3133 3134 3135
			if (parent_locked)
				free_extent_state(cached);
			else
				unlock_extent_cached(tree, cur,
						     cur + iosize - 1,
						     &cached, GFP_NOFS);
3136
			cur = cur + iosize;
3137
			pg_offset += iosize;
3138 3139 3140
			continue;
		}
		/* the get_extent function already copied into the page */
3141 3142
		if (test_range_bit(tree, cur, cur_end,
				   EXTENT_UPTODATE, 1, NULL)) {
3143
			check_page_uptodate(tree, page);
3144 3145
			if (!parent_locked)
				unlock_extent(tree, cur, cur + iosize - 1);
3146
			cur = cur + iosize;
3147
			pg_offset += iosize;
3148 3149
			continue;
		}
3150 3151 3152 3153 3154
		/* we have an inline extent but it didn't get marked up
		 * to date.  Error out
		 */
		if (block_start == EXTENT_MAP_INLINE) {
			SetPageError(page);
3155 3156
			if (!parent_locked)
				unlock_extent(tree, cur, cur + iosize - 1);
3157
			cur = cur + iosize;
3158
			pg_offset += iosize;
3159 3160
			continue;
		}
3161

3162
		pnr -= page->index;
3163
		ret = submit_extent_page(rw, tree, NULL, page,
3164
					 sector, disk_io_size, pg_offset,
3165
					 bdev, bio, pnr,
C
Chris Mason 已提交
3166 3167
					 end_bio_extent_readpage, mirror_num,
					 *bio_flags,
3168 3169
					 this_bio_flag,
					 force_bio_submit);
3170 3171 3172 3173
		if (!ret) {
			nr++;
			*bio_flags = this_bio_flag;
		} else {
3174
			SetPageError(page);
3175 3176
			if (!parent_locked)
				unlock_extent(tree, cur, cur + iosize - 1);
3177
		}
3178
		cur = cur + iosize;
3179
		pg_offset += iosize;
3180
	}
D
Dan Magenheimer 已提交
3181
out:
3182 3183 3184 3185 3186 3187 3188 3189
	if (!nr) {
		if (!PageError(page))
			SetPageUptodate(page);
		unlock_page(page);
	}
	return 0;
}

3190 3191 3192 3193
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,
3194
					     struct extent_map **em_cached,
3195
					     struct bio **bio, int mirror_num,
3196 3197
					     unsigned long *bio_flags, int rw,
					     u64 *prev_em_start)
3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215
{
	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++) {
3216
		__do_readpage(tree, pages[index], get_extent, em_cached, bio,
3217
			      mirror_num, bio_flags, rw, prev_em_start);
3218 3219 3220 3221 3222 3223 3224
		page_cache_release(pages[index]);
	}
}

static void __extent_readpages(struct extent_io_tree *tree,
			       struct page *pages[],
			       int nr_pages, get_extent_t *get_extent,
3225
			       struct extent_map **em_cached,
3226
			       struct bio **bio, int mirror_num,
3227 3228
			       unsigned long *bio_flags, int rw,
			       u64 *prev_em_start)
3229
{
3230
	u64 start = 0;
3231 3232 3233
	u64 end = 0;
	u64 page_start;
	int index;
3234
	int first_index = 0;
3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246

	for (index = 0; index < nr_pages; index++) {
		page_start = page_offset(pages[index]);
		if (!end) {
			start = page_start;
			end = start + PAGE_CACHE_SIZE - 1;
			first_index = index;
		} else if (end + 1 == page_start) {
			end += PAGE_CACHE_SIZE;
		} else {
			__do_contiguous_readpages(tree, &pages[first_index],
						  index - first_index, start,
3247 3248
						  end, get_extent, em_cached,
						  bio, mirror_num, bio_flags,
3249
						  rw, prev_em_start);
3250 3251 3252 3253 3254 3255 3256 3257 3258
			start = page_start;
			end = start + PAGE_CACHE_SIZE - 1;
			first_index = index;
		}
	}

	if (end)
		__do_contiguous_readpages(tree, &pages[first_index],
					  index - first_index, start,
3259
					  end, get_extent, em_cached, bio,
3260 3261
					  mirror_num, bio_flags, rw,
					  prev_em_start);
3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285
}

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,
				   unsigned long *bio_flags, int rw)
{
	struct inode *inode = page->mapping->host;
	struct btrfs_ordered_extent *ordered;
	u64 start = page_offset(page);
	u64 end = start + PAGE_CACHE_SIZE - 1;
	int ret;

	while (1) {
		lock_extent(tree, start, end);
		ordered = btrfs_lookup_ordered_extent(inode, start);
		if (!ordered)
			break;
		unlock_extent(tree, start, end);
		btrfs_start_ordered_extent(inode, ordered, 1);
		btrfs_put_ordered_extent(ordered);
	}

3286
	ret = __do_readpage(tree, page, get_extent, NULL, bio, mirror_num,
3287
			    bio_flags, rw, NULL);
3288 3289 3290
	return ret;
}

3291
int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
3292
			    get_extent_t *get_extent, int mirror_num)
3293 3294
{
	struct bio *bio = NULL;
C
Chris Mason 已提交
3295
	unsigned long bio_flags = 0;
3296 3297
	int ret;

3298
	ret = __extent_read_full_page(tree, page, get_extent, &bio, mirror_num,
3299
				      &bio_flags, READ);
3300
	if (bio)
3301
		ret = submit_one_bio(READ, bio, mirror_num, bio_flags);
3302 3303 3304
	return ret;
}

3305 3306 3307 3308 3309 3310 3311 3312
int extent_read_full_page_nolock(struct extent_io_tree *tree, struct page *page,
				 get_extent_t *get_extent, int mirror_num)
{
	struct bio *bio = NULL;
	unsigned long bio_flags = EXTENT_BIO_PARENT_LOCKED;
	int ret;

	ret = __do_readpage(tree, page, get_extent, NULL, &bio, mirror_num,
3313
			    &bio_flags, READ, NULL);
3314 3315 3316 3317 3318
	if (bio)
		ret = submit_one_bio(READ, bio, mirror_num, bio_flags);
	return ret;
}

3319 3320 3321 3322 3323 3324 3325 3326 3327 3328
static noinline void update_nr_written(struct page *page,
				      struct writeback_control *wbc,
				      unsigned long nr_written)
{
	wbc->nr_to_write -= nr_written;
	if (wbc->range_cyclic || (wbc->nr_to_write > 0 &&
	    wbc->range_start == 0 && wbc->range_end == LLONG_MAX))
		page->mapping->writeback_index = page->index + nr_written;
}

3329
/*
3330 3331 3332 3333 3334 3335 3336 3337
 * 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)
3338
 */
3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360
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;
	u64 page_end = delalloc_start + PAGE_CACHE_SIZE - 1;
	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,
3361
					       BTRFS_MAX_EXTENT_SIZE);
3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434
		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;
		}
		/*
		 * delalloc_end is already one less than the total
		 * length, so we don't subtract one from
		 * PAGE_CACHE_SIZE
		 */
		delalloc_to_write += (delalloc_end - delalloc_start +
				      PAGE_CACHE_SIZE) >>
				      PAGE_CACHE_SHIFT;
		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)
3435 3436
{
	struct extent_io_tree *tree = epd->tree;
M
Miao Xie 已提交
3437
	u64 start = page_offset(page);
3438 3439 3440 3441 3442 3443 3444
	u64 page_end = start + PAGE_CACHE_SIZE - 1;
	u64 end;
	u64 cur = start;
	u64 extent_offset;
	u64 block_start;
	u64 iosize;
	sector_t sector;
3445
	struct extent_state *cached_state = NULL;
3446 3447
	struct extent_map *em;
	struct block_device *bdev;
3448
	size_t pg_offset = 0;
3449
	size_t blocksize;
3450 3451 3452
	int ret = 0;
	int nr = 0;
	bool compressed;
C
Chris Mason 已提交
3453

3454
	if (tree->ops && tree->ops->writepage_start_hook) {
C
Chris Mason 已提交
3455 3456
		ret = tree->ops->writepage_start_hook(page, start,
						      page_end);
3457 3458 3459 3460 3461 3462
		if (ret) {
			/* Fixup worker will requeue */
			if (ret == -EBUSY)
				wbc->pages_skipped++;
			else
				redirty_page_for_writepage(wbc, page);
3463

3464
			update_nr_written(page, wbc, nr_written);
3465
			unlock_page(page);
3466
			ret = 1;
3467
			goto done_unlocked;
3468 3469 3470
		}
	}

3471 3472 3473 3474 3475
	/*
	 * 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);
3476

3477
	end = page_end;
3478
	if (i_size <= start) {
3479 3480 3481
		if (tree->ops && tree->ops->writepage_end_io_hook)
			tree->ops->writepage_end_io_hook(page, start,
							 page_end, NULL, 1);
3482 3483 3484 3485 3486 3487
		goto done;
	}

	blocksize = inode->i_sb->s_blocksize;

	while (cur <= end) {
3488 3489
		u64 em_end;
		if (cur >= i_size) {
3490 3491 3492
			if (tree->ops && tree->ops->writepage_end_io_hook)
				tree->ops->writepage_end_io_hook(page, cur,
							 page_end, NULL, 1);
3493 3494
			break;
		}
3495
		em = epd->get_extent(inode, page, pg_offset, cur,
3496
				     end - cur + 1, 1);
3497
		if (IS_ERR_OR_NULL(em)) {
3498
			SetPageError(page);
3499
			ret = PTR_ERR_OR_ZERO(em);
3500 3501 3502 3503
			break;
		}

		extent_offset = cur - em->start;
3504 3505
		em_end = extent_map_end(em);
		BUG_ON(em_end <= cur);
3506
		BUG_ON(end < cur);
3507
		iosize = min(em_end - cur, end - cur + 1);
3508
		iosize = ALIGN(iosize, blocksize);
3509 3510 3511
		sector = (em->block_start + extent_offset) >> 9;
		bdev = em->bdev;
		block_start = em->block_start;
C
Chris Mason 已提交
3512
		compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
3513 3514 3515
		free_extent_map(em);
		em = NULL;

C
Chris Mason 已提交
3516 3517 3518 3519 3520
		/*
		 * compressed and inline extents are written through other
		 * paths in the FS
		 */
		if (compressed || block_start == EXTENT_MAP_HOLE ||
3521
		    block_start == EXTENT_MAP_INLINE) {
C
Chris Mason 已提交
3522 3523 3524 3525 3526 3527
			/*
			 * end_io notification does not happen here for
			 * compressed extents
			 */
			if (!compressed && tree->ops &&
			    tree->ops->writepage_end_io_hook)
3528 3529 3530
				tree->ops->writepage_end_io_hook(page, cur,
							 cur + iosize - 1,
							 NULL, 1);
C
Chris Mason 已提交
3531 3532 3533 3534 3535 3536 3537 3538 3539
			else if (compressed) {
				/* we don't want to end_page_writeback on
				 * a compressed extent.  this happens
				 * elsewhere
				 */
				nr++;
			}

			cur += iosize;
3540
			pg_offset += iosize;
3541 3542
			continue;
		}
C
Chris Mason 已提交
3543

3544 3545 3546 3547 3548 3549
		if (tree->ops && tree->ops->writepage_io_hook) {
			ret = tree->ops->writepage_io_hook(page, cur,
						cur + iosize - 1);
		} else {
			ret = 0;
		}
3550
		if (ret) {
3551
			SetPageError(page);
3552
		} else {
3553
			unsigned long max_nr = (i_size >> PAGE_CACHE_SHIFT) + 1;
3554

3555 3556
			set_range_writeback(tree, cur, cur + iosize - 1);
			if (!PageWriteback(page)) {
3557 3558
				btrfs_err(BTRFS_I(inode)->root->fs_info,
					   "page %lu not writeback, cur %llu end %llu",
3559
				       page->index, cur, end);
3560 3561
			}

3562
			ret = submit_extent_page(write_flags, tree, wbc, page,
3563 3564
						 sector, iosize, pg_offset,
						 bdev, &epd->bio, max_nr,
C
Chris Mason 已提交
3565
						 end_bio_extent_writepage,
3566
						 0, 0, 0, false);
3567 3568 3569 3570
			if (ret)
				SetPageError(page);
		}
		cur = cur + iosize;
3571
		pg_offset += iosize;
3572 3573
		nr++;
	}
3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648
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);
	u64 page_end = start + PAGE_CACHE_SIZE - 1;
	int ret;
	int nr = 0;
	size_t pg_offset = 0;
	loff_t i_size = i_size_read(inode);
	unsigned long end_index = i_size >> PAGE_CACHE_SHIFT;
	int write_flags;
	unsigned long nr_written = 0;

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

	trace___extent_writepage(page, inode, wbc);

	WARN_ON(!PageLocked(page));

	ClearPageError(page);

	pg_offset = i_size & (PAGE_CACHE_SIZE - 1);
	if (page->index > end_index ||
	   (page->index == end_index && !pg_offset)) {
		page->mapping->a_ops->invalidatepage(page, 0, PAGE_CACHE_SIZE);
		unlock_page(page);
		return 0;
	}

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

		userpage = kmap_atomic(page);
		memset(userpage + pg_offset, 0,
		       PAGE_CACHE_SIZE - pg_offset);
		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;

3649 3650 3651 3652 3653 3654
done:
	if (nr == 0) {
		/* make sure the mapping tag for page dirty gets cleared */
		set_page_writeback(page);
		end_page_writeback(page);
	}
3655 3656 3657 3658
	if (PageError(page)) {
		ret = ret < 0 ? ret : -EIO;
		end_extent_writepage(page, ret, start, page_end);
	}
3659
	unlock_page(page);
3660
	return ret;
3661

3662
done_unlocked:
3663 3664 3665
	return 0;
}

3666
void wait_on_extent_buffer_writeback(struct extent_buffer *eb)
3667
{
3668 3669
	wait_on_bit_io(&eb->bflags, EXTENT_BUFFER_WRITEBACK,
		       TASK_UNINTERRUPTIBLE);
3670 3671
}

3672 3673 3674 3675
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)
3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694
{
	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
3695 3696 3697 3698 3699
		while (1) {
			wait_on_extent_buffer_writeback(eb);
			btrfs_tree_lock(eb);
			if (!test_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags))
				break;
3700 3701 3702 3703
			btrfs_tree_unlock(eb);
		}
	}

3704 3705 3706 3707 3708 3709
	/*
	 * 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);
3710 3711
	if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
		set_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags);
3712
		spin_unlock(&eb->refs_lock);
3713
		btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
3714 3715 3716
		__percpu_counter_add(&fs_info->dirty_metadata_bytes,
				     -eb->len,
				     fs_info->dirty_metadata_batch);
3717
		ret = 1;
3718 3719
	} else {
		spin_unlock(&eb->refs_lock);
3720 3721 3722 3723 3724 3725 3726 3727 3728
	}

	btrfs_tree_unlock(eb);

	if (!ret)
		return ret;

	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = 0; i < num_pages; i++) {
3729
		struct page *p = eb->pages[i];
3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745

		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);
3746
	smp_mb__after_atomic();
3747 3748 3749
	wake_up_bit(&eb->bflags, EXTENT_BUFFER_WRITEBACK);
}

3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811
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 */
	}
}

3812
static void end_bio_extent_buffer_writepage(struct bio *bio)
3813
{
3814
	struct bio_vec *bvec;
3815
	struct extent_buffer *eb;
3816
	int i, done;
3817

3818
	bio_for_each_segment_all(bvec, bio, i) {
3819 3820 3821 3822 3823 3824
		struct page *page = bvec->bv_page;

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

3825 3826
		if (bio->bi_error ||
		    test_bit(EXTENT_BUFFER_WRITE_ERR, &eb->bflags)) {
3827
			ClearPageUptodate(page);
3828
			set_btree_ioerr(page);
3829 3830 3831 3832 3833 3834 3835 3836
		}

		end_page_writeback(page);

		if (!done)
			continue;

		end_extent_buffer_writeback(eb);
3837
	}
3838 3839 3840 3841

	bio_put(bio);
}

3842
static noinline_for_stack int write_one_eb(struct extent_buffer *eb,
3843 3844 3845 3846 3847
			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;
3848
	struct extent_io_tree *tree = &BTRFS_I(fs_info->btree_inode)->io_tree;
3849 3850
	u64 offset = eb->start;
	unsigned long i, num_pages;
3851
	unsigned long bio_flags = 0;
3852
	int rw = (epd->sync_io ? WRITE_SYNC : WRITE) | REQ_META;
3853
	int ret = 0;
3854

3855
	clear_bit(EXTENT_BUFFER_WRITE_ERR, &eb->bflags);
3856 3857
	num_pages = num_extent_pages(eb->start, eb->len);
	atomic_set(&eb->io_pages, num_pages);
3858 3859 3860
	if (btrfs_header_owner(eb) == BTRFS_TREE_LOG_OBJECTID)
		bio_flags = EXTENT_BIO_TREE_LOG;

3861
	for (i = 0; i < num_pages; i++) {
3862
		struct page *p = eb->pages[i];
3863 3864 3865

		clear_page_dirty_for_io(p);
		set_page_writeback(p);
3866
		ret = submit_extent_page(rw, tree, wbc, p, offset >> 9,
3867 3868
					 PAGE_CACHE_SIZE, 0, bdev, &epd->bio,
					 -1, end_bio_extent_buffer_writepage,
3869
					 0, epd->bio_flags, bio_flags, false);
3870
		epd->bio_flags = bio_flags;
3871
		if (ret) {
3872
			set_btree_ioerr(p);
3873
			end_page_writeback(p);
3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885
			if (atomic_sub_and_test(num_pages - i, &eb->io_pages))
				end_extent_buffer_writeback(eb);
			ret = -EIO;
			break;
		}
		offset += PAGE_CACHE_SIZE;
		update_nr_written(p, wbc, 1);
		unlock_page(p);
	}

	if (unlikely(ret)) {
		for (; i < num_pages; i++) {
3886
			struct page *p = eb->pages[i];
3887
			clear_page_dirty_for_io(p);
3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905
			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,
3906
		.bio_flags = 0,
3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950
	};
	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 {
		index = wbc->range_start >> PAGE_CACHE_SHIFT;
		end = wbc->range_end >> PAGE_CACHE_SHIFT;
		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;
			}

3951 3952 3953 3954 3955 3956
			spin_lock(&mapping->private_lock);
			if (!PagePrivate(page)) {
				spin_unlock(&mapping->private_lock);
				continue;
			}

3957
			eb = (struct extent_buffer *)page->private;
3958 3959 3960 3961 3962 3963

			/*
			 * 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.
			 */
3964
			if (WARN_ON(!eb)) {
3965
				spin_unlock(&mapping->private_lock);
3966 3967 3968
				continue;
			}

3969 3970
			if (eb == prev_eb) {
				spin_unlock(&mapping->private_lock);
3971
				continue;
3972
			}
3973

3974 3975 3976
			ret = atomic_inc_not_zero(&eb->refs);
			spin_unlock(&mapping->private_lock);
			if (!ret)
3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016
				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;
}

4017
/**
C
Chris Mason 已提交
4018
 * write_cache_pages - walk the list of dirty pages of the given address space and write all of them.
4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031
 * @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.
 */
4032
static int extent_write_cache_pages(struct extent_io_tree *tree,
C
Chris Mason 已提交
4033 4034
			     struct address_space *mapping,
			     struct writeback_control *wbc,
C
Chris Mason 已提交
4035 4036
			     writepage_t writepage, void *data,
			     void (*flush_fn)(void *))
4037
{
4038
	struct inode *inode = mapping->host;
4039 4040
	int ret = 0;
	int done = 0;
4041
	int err = 0;
4042
	int nr_to_write_done = 0;
4043 4044 4045 4046 4047
	struct pagevec pvec;
	int nr_pages;
	pgoff_t index;
	pgoff_t end;		/* Inclusive */
	int scanned = 0;
4048
	int tag;
4049

4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061
	/*
	 * 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;

4062 4063 4064 4065 4066 4067 4068 4069 4070
	pagevec_init(&pvec, 0);
	if (wbc->range_cyclic) {
		index = mapping->writeback_index; /* Start from prev offset */
		end = -1;
	} else {
		index = wbc->range_start >> PAGE_CACHE_SHIFT;
		end = wbc->range_end >> PAGE_CACHE_SHIFT;
		scanned = 1;
	}
4071 4072 4073 4074
	if (wbc->sync_mode == WB_SYNC_ALL)
		tag = PAGECACHE_TAG_TOWRITE;
	else
		tag = PAGECACHE_TAG_DIRTY;
4075
retry:
4076 4077
	if (wbc->sync_mode == WB_SYNC_ALL)
		tag_pages_for_writeback(mapping, index, end);
4078
	while (!done && !nr_to_write_done && (index <= end) &&
4079 4080
	       (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag,
			min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093
		unsigned i;

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

			/*
			 * 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
			 */
4094 4095 4096
			if (!trylock_page(page)) {
				flush_fn(data);
				lock_page(page);
4097
			}
4098 4099 4100 4101 4102 4103 4104 4105 4106 4107 4108 4109

			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
4110
			if (wbc->sync_mode != WB_SYNC_NONE) {
4111 4112
				if (PageWriteback(page))
					flush_fn(data);
4113
				wait_on_page_writeback(page);
C
Chris Mason 已提交
4114
			}
4115 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127

			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;
			}
4128 4129
			if (!err && ret < 0)
				err = ret;
4130 4131 4132 4133 4134 4135 4136

			/*
			 * 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;
4137 4138 4139 4140
		}
		pagevec_release(&pvec);
		cond_resched();
	}
4141
	if (!scanned && !done && !err) {
4142 4143 4144 4145 4146 4147 4148 4149
		/*
		 * 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;
	}
4150
	btrfs_add_delayed_iput(inode);
4151
	return err;
4152 4153
}

4154
static void flush_epd_write_bio(struct extent_page_data *epd)
C
Chris Mason 已提交
4155 4156
{
	if (epd->bio) {
4157 4158 4159
		int rw = WRITE;
		int ret;

4160
		if (epd->sync_io)
4161 4162
			rw = WRITE_SYNC;

4163
		ret = submit_one_bio(rw, epd->bio, 0, epd->bio_flags);
4164
		BUG_ON(ret < 0); /* -ENOMEM */
C
Chris Mason 已提交
4165 4166 4167 4168
		epd->bio = NULL;
	}
}

4169 4170 4171 4172 4173 4174
static noinline void flush_write_bio(void *data)
{
	struct extent_page_data *epd = data;
	flush_epd_write_bio(epd);
}

4175 4176 4177 4178 4179 4180 4181 4182 4183
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,
4184
		.extent_locked = 0,
4185
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
4186
		.bio_flags = 0,
4187 4188 4189 4190
	};

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

4191
	flush_epd_write_bio(&epd);
4192 4193 4194
	return ret;
}

4195 4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209
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;
	unsigned long nr_pages = (end - start + PAGE_CACHE_SIZE) >>
		PAGE_CACHE_SHIFT;

	struct extent_page_data epd = {
		.bio = NULL,
		.tree = tree,
		.get_extent = get_extent,
		.extent_locked = 1,
4210
		.sync_io = mode == WB_SYNC_ALL,
4211
		.bio_flags = 0,
4212 4213 4214 4215 4216 4217 4218 4219
	};
	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
4220
	while (start <= end) {
4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234
		page = find_get_page(mapping, start >> PAGE_CACHE_SHIFT);
		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,
						 start + PAGE_CACHE_SIZE - 1,
						 NULL, 1);
			unlock_page(page);
		}
		page_cache_release(page);
		start += PAGE_CACHE_SIZE;
	}

4235
	flush_epd_write_bio(&epd);
4236 4237
	return ret;
}
4238 4239 4240 4241 4242 4243 4244 4245 4246 4247 4248

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,
4249
		.extent_locked = 0,
4250
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
4251
		.bio_flags = 0,
4252 4253
	};

C
Chris Mason 已提交
4254
	ret = extent_write_cache_pages(tree, mapping, wbc,
C
Chris Mason 已提交
4255 4256
				       __extent_writepage, &epd,
				       flush_write_bio);
4257
	flush_epd_write_bio(&epd);
4258 4259 4260 4261 4262 4263 4264 4265 4266 4267
	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
4268
	unsigned long bio_flags = 0;
L
Liu Bo 已提交
4269 4270
	struct page *pagepool[16];
	struct page *page;
4271
	struct extent_map *em_cached = NULL;
L
Liu Bo 已提交
4272
	int nr = 0;
4273
	u64 prev_em_start = (u64)-1;
4274 4275

	for (page_idx = 0; page_idx < nr_pages; page_idx++) {
L
Liu Bo 已提交
4276
		page = list_entry(pages->prev, struct page, lru);
4277 4278 4279

		prefetchw(&page->flags);
		list_del(&page->lru);
L
Liu Bo 已提交
4280
		if (add_to_page_cache_lru(page, mapping,
4281
					page->index, GFP_NOFS)) {
L
Liu Bo 已提交
4282 4283
			page_cache_release(page);
			continue;
4284
		}
L
Liu Bo 已提交
4285 4286 4287 4288

		pagepool[nr++] = page;
		if (nr < ARRAY_SIZE(pagepool))
			continue;
4289
		__extent_readpages(tree, pagepool, nr, get_extent, &em_cached,
4290
				   &bio, 0, &bio_flags, READ, &prev_em_start);
L
Liu Bo 已提交
4291
		nr = 0;
4292
	}
4293
	if (nr)
4294
		__extent_readpages(tree, pagepool, nr, get_extent, &em_cached,
4295
				   &bio, 0, &bio_flags, READ, &prev_em_start);
L
Liu Bo 已提交
4296

4297 4298 4299
	if (em_cached)
		free_extent_map(em_cached);

4300 4301
	BUG_ON(!list_empty(pages));
	if (bio)
4302
		return submit_one_bio(READ, bio, 0, bio_flags);
4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313
	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)
{
4314
	struct extent_state *cached_state = NULL;
M
Miao Xie 已提交
4315
	u64 start = page_offset(page);
4316 4317 4318
	u64 end = start + PAGE_CACHE_SIZE - 1;
	size_t blocksize = page->mapping->host->i_sb->s_blocksize;

4319
	start += ALIGN(offset, blocksize);
4320 4321 4322
	if (start > end)
		return 0;

4323
	lock_extent_bits(tree, start, end, 0, &cached_state);
4324
	wait_on_page_writeback(page);
4325
	clear_extent_bit(tree, start, end,
4326 4327
			 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
			 EXTENT_DO_ACCOUNTING,
4328
			 1, 1, &cached_state, GFP_NOFS);
4329 4330 4331
	return 0;
}

4332 4333 4334 4335 4336
/*
 * 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.
 */
4337 4338 4339
static int try_release_extent_state(struct extent_map_tree *map,
				    struct extent_io_tree *tree,
				    struct page *page, gfp_t mask)
4340
{
M
Miao Xie 已提交
4341
	u64 start = page_offset(page);
4342 4343 4344
	u64 end = start + PAGE_CACHE_SIZE - 1;
	int ret = 1;

4345
	if (test_range_bit(tree, start, end,
4346
			   EXTENT_IOBITS, 0, NULL))
4347 4348 4349 4350
		ret = 0;
	else {
		if ((mask & GFP_NOFS) == GFP_NOFS)
			mask = GFP_NOFS;
4351 4352 4353 4354
		/*
		 * at this point we can safely clear everything except the
		 * locked bit and the nodatasum bit
		 */
4355
		ret = clear_extent_bit(tree, start, end,
4356 4357
				 ~(EXTENT_LOCKED | EXTENT_NODATASUM),
				 0, 0, NULL, mask);
4358 4359 4360 4361 4362 4363 4364 4365

		/* if clear_extent_bit failed for enomem reasons,
		 * we can't allow the release to continue.
		 */
		if (ret < 0)
			ret = 0;
		else
			ret = 1;
4366 4367 4368 4369
	}
	return ret;
}

4370 4371 4372 4373 4374 4375
/*
 * 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,
4376 4377
			       struct extent_io_tree *tree, struct page *page,
			       gfp_t mask)
4378 4379
{
	struct extent_map *em;
M
Miao Xie 已提交
4380
	u64 start = page_offset(page);
4381
	u64 end = start + PAGE_CACHE_SIZE - 1;
4382

4383
	if (gfpflags_allow_blocking(mask) &&
4384
	    page->mapping->host->i_size > 16 * 1024 * 1024) {
4385
		u64 len;
4386
		while (start <= end) {
4387
			len = end - start + 1;
4388
			write_lock(&map->lock);
4389
			em = lookup_extent_mapping(map, start, len);
4390
			if (!em) {
4391
				write_unlock(&map->lock);
4392 4393
				break;
			}
4394 4395
			if (test_bit(EXTENT_FLAG_PINNED, &em->flags) ||
			    em->start != start) {
4396
				write_unlock(&map->lock);
4397 4398 4399 4400 4401
				free_extent_map(em);
				break;
			}
			if (!test_range_bit(tree, em->start,
					    extent_map_end(em) - 1,
4402
					    EXTENT_LOCKED | EXTENT_WRITEBACK,
4403
					    0, NULL)) {
4404 4405 4406 4407 4408
				remove_extent_mapping(map, em);
				/* once for the rb tree */
				free_extent_map(em);
			}
			start = extent_map_end(em);
4409
			write_unlock(&map->lock);
4410 4411

			/* once for us */
4412 4413 4414
			free_extent_map(em);
		}
	}
4415
	return try_release_extent_state(map, tree, page, mask);
4416 4417
}

4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433
/*
 * 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;

4434
	while (1) {
4435 4436 4437
		len = last - offset;
		if (len == 0)
			break;
4438
		len = ALIGN(len, sectorsize);
4439
		em = get_extent(inode, NULL, 0, offset, len, 0);
4440
		if (IS_ERR_OR_NULL(em))
4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457
			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
4458 4459 4460
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
4461
	int ret = 0;
Y
Yehuda Sadeh 已提交
4462 4463 4464
	u64 off = start;
	u64 max = start + len;
	u32 flags = 0;
J
Josef Bacik 已提交
4465 4466
	u32 found_type;
	u64 last;
4467
	u64 last_for_get_extent = 0;
Y
Yehuda Sadeh 已提交
4468
	u64 disko = 0;
4469
	u64 isize = i_size_read(inode);
J
Josef Bacik 已提交
4470
	struct btrfs_key found_key;
Y
Yehuda Sadeh 已提交
4471
	struct extent_map *em = NULL;
4472
	struct extent_state *cached_state = NULL;
J
Josef Bacik 已提交
4473
	struct btrfs_path *path;
4474
	struct btrfs_root *root = BTRFS_I(inode)->root;
Y
Yehuda Sadeh 已提交
4475
	int end = 0;
4476 4477 4478
	u64 em_start = 0;
	u64 em_len = 0;
	u64 em_end = 0;
Y
Yehuda Sadeh 已提交
4479 4480 4481 4482

	if (len == 0)
		return -EINVAL;

J
Josef Bacik 已提交
4483 4484 4485 4486 4487
	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;
	path->leave_spinning = 1;

4488 4489
	start = round_down(start, BTRFS_I(inode)->root->sectorsize);
	len = round_up(max, BTRFS_I(inode)->root->sectorsize) - start;
4490

4491 4492 4493 4494
	/*
	 * lookup the last file extent.  We're not using i_size here
	 * because there might be preallocation past i_size
	 */
4495 4496
	ret = btrfs_lookup_file_extent(NULL, root, path, btrfs_ino(inode), -1,
				       0);
J
Josef Bacik 已提交
4497 4498 4499 4500 4501 4502 4503
	if (ret < 0) {
		btrfs_free_path(path);
		return ret;
	}
	WARN_ON(!ret);
	path->slots[0]--;
	btrfs_item_key_to_cpu(path->nodes[0], &found_key, path->slots[0]);
4504
	found_type = found_key.type;
J
Josef Bacik 已提交
4505

4506
	/* No extents, but there might be delalloc bits */
L
Li Zefan 已提交
4507
	if (found_key.objectid != btrfs_ino(inode) ||
J
Josef Bacik 已提交
4508
	    found_type != BTRFS_EXTENT_DATA_KEY) {
4509 4510 4511 4512 4513 4514 4515 4516 4517 4518 4519
		/* 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
4520
	}
4521
	btrfs_release_path(path);
J
Josef Bacik 已提交
4522

4523 4524 4525 4526 4527 4528 4529 4530 4531 4532
	/*
	 * 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;
	}

L
Liu Bo 已提交
4533
	lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + len - 1, 0,
4534
			 &cached_state);
4535

4536
	em = get_extent_skip_holes(inode, start, last_for_get_extent,
4537
				   get_extent);
Y
Yehuda Sadeh 已提交
4538 4539 4540 4541 4542 4543
	if (!em)
		goto out;
	if (IS_ERR(em)) {
		ret = PTR_ERR(em);
		goto out;
	}
J
Josef Bacik 已提交
4544

Y
Yehuda Sadeh 已提交
4545
	while (!end) {
4546
		u64 offset_in_extent = 0;
4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558

		/* 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
4559

4560 4561
		/*
		 * record the offset from the start of the extent
4562 4563 4564
		 * 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.
4565
		 */
4566 4567
		if (!test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
			offset_in_extent = em_start - em->start;
4568
		em_end = extent_map_end(em);
4569
		em_len = em_end - em_start;
Y
Yehuda Sadeh 已提交
4570 4571 4572
		disko = 0;
		flags = 0;

4573 4574 4575 4576 4577 4578 4579
		/*
		 * bump off for our next call to get_extent
		 */
		off = extent_map_end(em);
		if (off >= max)
			end = 1;

4580
		if (em->block_start == EXTENT_MAP_LAST_BYTE) {
Y
Yehuda Sadeh 已提交
4581 4582
			end = 1;
			flags |= FIEMAP_EXTENT_LAST;
4583
		} else if (em->block_start == EXTENT_MAP_INLINE) {
Y
Yehuda Sadeh 已提交
4584 4585
			flags |= (FIEMAP_EXTENT_DATA_INLINE |
				  FIEMAP_EXTENT_NOT_ALIGNED);
4586
		} else if (em->block_start == EXTENT_MAP_DELALLOC) {
Y
Yehuda Sadeh 已提交
4587 4588
			flags |= (FIEMAP_EXTENT_DELALLOC |
				  FIEMAP_EXTENT_UNKNOWN);
4589 4590 4591
		} else if (fieinfo->fi_extents_max) {
			u64 bytenr = em->block_start -
				(em->start - em->orig_start);
4592

4593
			disko = em->block_start + offset_in_extent;
4594 4595 4596 4597

			/*
			 * As btrfs supports shared space, this information
			 * can be exported to userspace tools via
4598 4599 4600
			 * flag FIEMAP_EXTENT_SHARED.  If fi_extents_max == 0
			 * then we're just getting a count and we can skip the
			 * lookup stuff.
4601
			 */
4602 4603 4604 4605
			ret = btrfs_check_shared(NULL, root->fs_info,
						 root->objectid,
						 btrfs_ino(inode), bytenr);
			if (ret < 0)
4606
				goto out_free;
4607
			if (ret)
4608
				flags |= FIEMAP_EXTENT_SHARED;
4609
			ret = 0;
Y
Yehuda Sadeh 已提交
4610 4611 4612
		}
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
			flags |= FIEMAP_EXTENT_ENCODED;
4613 4614
		if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			flags |= FIEMAP_EXTENT_UNWRITTEN;
Y
Yehuda Sadeh 已提交
4615 4616 4617

		free_extent_map(em);
		em = NULL;
4618 4619
		if ((em_start >= last) || em_len == (u64)-1 ||
		   (last == (u64)-1 && isize <= em_end)) {
Y
Yehuda Sadeh 已提交
4620 4621 4622 4623
			flags |= FIEMAP_EXTENT_LAST;
			end = 1;
		}

4624 4625 4626 4627 4628 4629 4630 4631
		/* 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
4632 4633 4634
			flags |= FIEMAP_EXTENT_LAST;
			end = 1;
		}
4635 4636
		ret = fiemap_fill_next_extent(fieinfo, em_start, disko,
					      em_len, flags);
4637 4638 4639
		if (ret) {
			if (ret == 1)
				ret = 0;
4640
			goto out_free;
4641
		}
Y
Yehuda Sadeh 已提交
4642 4643 4644 4645
	}
out_free:
	free_extent_map(em);
out:
4646
	btrfs_free_path(path);
L
Liu Bo 已提交
4647
	unlock_extent_cached(&BTRFS_I(inode)->io_tree, start, start + len - 1,
4648
			     &cached_state, GFP_NOFS);
Y
Yehuda Sadeh 已提交
4649 4650 4651
	return ret;
}

4652 4653
static void __free_extent_buffer(struct extent_buffer *eb)
{
4654
	btrfs_leak_debug_del(&eb->leak_list);
4655 4656 4657
	kmem_cache_free(extent_buffer_cache, eb);
}

4658
int extent_buffer_under_io(struct extent_buffer *eb)
4659 4660 4661 4662 4663 4664 4665 4666 4667
{
	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.
 */
4668
static void btrfs_release_extent_buffer_page(struct extent_buffer *eb)
4669 4670 4671 4672 4673 4674 4675
{
	unsigned long index;
	struct page *page;
	int mapped = !test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags);

	BUG_ON(extent_buffer_under_io(eb));

4676 4677
	index = num_extent_pages(eb->start, eb->len);
	if (index == 0)
4678 4679 4680 4681
		return;

	do {
		index--;
4682
		page = eb->pages[index];
4683 4684 4685
		if (!page)
			continue;
		if (mapped)
4686
			spin_lock(&page->mapping->private_lock);
4687 4688 4689 4690 4691 4692 4693 4694 4695 4696 4697 4698
		/*
		 * 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));
4699
			/*
4700 4701
			 * We need to make sure we haven't be attached
			 * to a new eb.
4702
			 */
4703 4704 4705
			ClearPagePrivate(page);
			set_page_private(page, 0);
			/* One for the page private */
4706 4707
			page_cache_release(page);
		}
4708 4709 4710 4711 4712 4713

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

		/* One for when we alloced the page */
		page_cache_release(page);
4714
	} while (index != 0);
4715 4716 4717 4718 4719 4720 4721
}

/*
 * Helper for releasing the extent buffer.
 */
static inline void btrfs_release_extent_buffer(struct extent_buffer *eb)
{
4722
	btrfs_release_extent_buffer_page(eb);
4723 4724 4725
	__free_extent_buffer(eb);
}

4726 4727
static struct extent_buffer *
__alloc_extent_buffer(struct btrfs_fs_info *fs_info, u64 start,
4728
		      unsigned long len)
4729 4730 4731
{
	struct extent_buffer *eb = NULL;

4732
	eb = kmem_cache_zalloc(extent_buffer_cache, GFP_NOFS|__GFP_NOFAIL);
4733 4734
	eb->start = start;
	eb->len = len;
4735
	eb->fs_info = fs_info;
4736
	eb->bflags = 0;
4737 4738 4739 4740 4741 4742 4743
	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);
4744
	eb->lock_nested = 0;
4745 4746
	init_waitqueue_head(&eb->write_lock_wq);
	init_waitqueue_head(&eb->read_lock_wq);
4747

4748 4749
	btrfs_leak_debug_add(&eb->leak_list, &buffers);

4750
	spin_lock_init(&eb->refs_lock);
4751
	atomic_set(&eb->refs, 1);
4752
	atomic_set(&eb->io_pages, 0);
4753

4754 4755 4756 4757 4758 4759
	/*
	 * 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);
4760 4761 4762 4763

	return eb;
}

4764 4765 4766 4767 4768 4769 4770
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);

4771
	new = __alloc_extent_buffer(src->fs_info, src->start, src->len);
4772 4773 4774 4775
	if (new == NULL)
		return NULL;

	for (i = 0; i < num_pages; i++) {
4776
		p = alloc_page(GFP_NOFS);
4777 4778 4779 4780
		if (!p) {
			btrfs_release_extent_buffer(new);
			return NULL;
		}
4781 4782 4783 4784 4785 4786 4787 4788 4789 4790 4791 4792 4793
		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;
}

4794 4795
struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
						u64 start)
4796 4797
{
	struct extent_buffer *eb;
4798 4799
	unsigned long len;
	unsigned long num_pages;
4800 4801
	unsigned long i;

4802 4803 4804 4805 4806 4807 4808 4809 4810 4811 4812 4813
	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;
	}
	num_pages = num_extent_pages(0, len);

	eb = __alloc_extent_buffer(fs_info, start, len);
4814 4815 4816 4817
	if (!eb)
		return NULL;

	for (i = 0; i < num_pages; i++) {
4818
		eb->pages[i] = alloc_page(GFP_NOFS);
4819 4820 4821 4822 4823 4824 4825 4826 4827
		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:
4828 4829
	for (; i > 0; i--)
		__free_page(eb->pages[i - 1]);
4830 4831 4832 4833
	__free_extent_buffer(eb);
	return NULL;
}

4834 4835
static void check_buffer_tree_ref(struct extent_buffer *eb)
{
4836
	int refs;
4837 4838 4839 4840 4841 4842 4843 4844 4845 4846 4847 4848 4849 4850 4851 4852 4853 4854 4855 4856
	/* 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.
	 */
4857 4858 4859 4860
	refs = atomic_read(&eb->refs);
	if (refs >= 2 && test_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
		return;

4861 4862
	spin_lock(&eb->refs_lock);
	if (!test_and_set_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
4863
		atomic_inc(&eb->refs);
4864
	spin_unlock(&eb->refs_lock);
4865 4866
}

4867 4868
static void mark_extent_buffer_accessed(struct extent_buffer *eb,
		struct page *accessed)
4869 4870 4871
{
	unsigned long num_pages, i;

4872 4873
	check_buffer_tree_ref(eb);

4874 4875
	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = 0; i < num_pages; i++) {
4876 4877
		struct page *p = eb->pages[i];

4878 4879
		if (p != accessed)
			mark_page_accessed(p);
4880 4881 4882
	}
}

4883 4884
struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
					 u64 start)
4885 4886 4887 4888
{
	struct extent_buffer *eb;

	rcu_read_lock();
4889 4890
	eb = radix_tree_lookup(&fs_info->buffer_radix,
			       start >> PAGE_CACHE_SHIFT);
4891 4892
	if (eb && atomic_inc_not_zero(&eb->refs)) {
		rcu_read_unlock();
4893 4894 4895 4896 4897 4898 4899 4900 4901 4902 4903 4904 4905 4906 4907 4908 4909 4910 4911
		/*
		 * 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);
		}
4912
		mark_extent_buffer_accessed(eb, NULL);
4913 4914 4915 4916 4917 4918 4919
		return eb;
	}
	rcu_read_unlock();

	return NULL;
}

4920 4921
#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
4922
					       u64 start)
4923 4924 4925 4926 4927 4928 4929
{
	struct extent_buffer *eb, *exists = NULL;
	int ret;

	eb = find_extent_buffer(fs_info, start);
	if (eb)
		return eb;
4930
	eb = alloc_dummy_extent_buffer(fs_info, start);
4931 4932 4933 4934 4935 4936 4937 4938 4939 4940 4941 4942 4943 4944 4945 4946 4947 4948 4949 4950 4951 4952 4953 4954 4955 4956 4957 4958 4959 4960 4961 4962 4963 4964 4965 4966
	if (!eb)
		return NULL;
	eb->fs_info = fs_info;
again:
	ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
	if (ret)
		goto free_eb;
	spin_lock(&fs_info->buffer_lock);
	ret = radix_tree_insert(&fs_info->buffer_radix,
				start >> PAGE_CACHE_SHIFT, eb);
	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

4967
struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
4968
					  u64 start)
4969
{
4970
	unsigned long len = fs_info->tree_root->nodesize;
4971 4972 4973 4974
	unsigned long num_pages = num_extent_pages(start, len);
	unsigned long i;
	unsigned long index = start >> PAGE_CACHE_SHIFT;
	struct extent_buffer *eb;
4975
	struct extent_buffer *exists = NULL;
4976
	struct page *p;
4977
	struct address_space *mapping = fs_info->btree_inode->i_mapping;
4978
	int uptodate = 1;
4979
	int ret;
4980

4981
	eb = find_extent_buffer(fs_info, start);
4982
	if (eb)
4983 4984
		return eb;

4985
	eb = __alloc_extent_buffer(fs_info, start, len);
4986
	if (!eb)
4987 4988
		return NULL;

4989
	for (i = 0; i < num_pages; i++, index++) {
4990
		p = find_or_create_page(mapping, index, GFP_NOFS|__GFP_NOFAIL);
4991
		if (!p)
4992
			goto free_eb;
J
Josef Bacik 已提交
4993 4994 4995 4996 4997 4998 4999 5000 5001 5002 5003 5004 5005 5006

		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);
5007
				page_cache_release(p);
5008
				mark_extent_buffer_accessed(exists, p);
J
Josef Bacik 已提交
5009 5010
				goto free_eb;
			}
5011
			exists = NULL;
J
Josef Bacik 已提交
5012

5013
			/*
J
Josef Bacik 已提交
5014 5015 5016 5017
			 * Do this so attach doesn't complain and we need to
			 * drop the ref the old guy had.
			 */
			ClearPagePrivate(p);
5018
			WARN_ON(PageDirty(p));
J
Josef Bacik 已提交
5019
			page_cache_release(p);
5020
		}
J
Josef Bacik 已提交
5021 5022
		attach_extent_buffer_page(eb, p);
		spin_unlock(&mapping->private_lock);
5023
		WARN_ON(PageDirty(p));
5024
		eb->pages[i] = p;
5025 5026
		if (!PageUptodate(p))
			uptodate = 0;
C
Chris Mason 已提交
5027 5028 5029 5030 5031

		/*
		 * see below about how we avoid a nasty race with release page
		 * and why we unlock later
		 */
5032 5033
	}
	if (uptodate)
5034
		set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5035
again:
5036 5037 5038 5039
	ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
	if (ret)
		goto free_eb;

5040 5041 5042 5043
	spin_lock(&fs_info->buffer_lock);
	ret = radix_tree_insert(&fs_info->buffer_radix,
				start >> PAGE_CACHE_SHIFT, eb);
	spin_unlock(&fs_info->buffer_lock);
5044
	radix_tree_preload_end();
5045
	if (ret == -EEXIST) {
5046
		exists = find_extent_buffer(fs_info, start);
5047 5048 5049
		if (exists)
			goto free_eb;
		else
5050
			goto again;
5051 5052
	}
	/* add one reference for the tree */
5053
	check_buffer_tree_ref(eb);
5054
	set_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags);
C
Chris Mason 已提交
5055 5056 5057 5058 5059 5060 5061 5062 5063 5064

	/*
	 * 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
	 */
5065 5066
	SetPageChecked(eb->pages[0]);
	for (i = 1; i < num_pages; i++) {
5067
		p = eb->pages[i];
5068 5069 5070 5071
		ClearPageChecked(p);
		unlock_page(p);
	}
	unlock_page(eb->pages[0]);
5072 5073
	return eb;

5074
free_eb:
5075
	WARN_ON(!atomic_dec_and_test(&eb->refs));
5076 5077 5078 5079
	for (i = 0; i < num_pages; i++) {
		if (eb->pages[i])
			unlock_page(eb->pages[i]);
	}
C
Chris Mason 已提交
5080

5081
	btrfs_release_extent_buffer(eb);
5082
	return exists;
5083 5084
}

5085 5086 5087 5088 5089 5090 5091 5092 5093
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 */
5094
static int release_extent_buffer(struct extent_buffer *eb)
5095 5096 5097
{
	WARN_ON(atomic_read(&eb->refs) == 0);
	if (atomic_dec_and_test(&eb->refs)) {
5098
		if (test_and_clear_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags)) {
5099
			struct btrfs_fs_info *fs_info = eb->fs_info;
5100

5101
			spin_unlock(&eb->refs_lock);
5102

5103 5104
			spin_lock(&fs_info->buffer_lock);
			radix_tree_delete(&fs_info->buffer_radix,
5105
					  eb->start >> PAGE_CACHE_SHIFT);
5106
			spin_unlock(&fs_info->buffer_lock);
5107 5108
		} else {
			spin_unlock(&eb->refs_lock);
5109
		}
5110 5111

		/* Should be safe to release our pages at this point */
5112
		btrfs_release_extent_buffer_page(eb);
5113 5114 5115 5116 5117 5118
#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
		if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags))) {
			__free_extent_buffer(eb);
			return 1;
		}
#endif
5119
		call_rcu(&eb->rcu_head, btrfs_release_extent_buffer_rcu);
5120
		return 1;
5121 5122
	}
	spin_unlock(&eb->refs_lock);
5123 5124

	return 0;
5125 5126
}

5127 5128
void free_extent_buffer(struct extent_buffer *eb)
{
5129 5130
	int refs;
	int old;
5131 5132 5133
	if (!eb)
		return;

5134 5135 5136 5137 5138 5139 5140 5141 5142
	while (1) {
		refs = atomic_read(&eb->refs);
		if (refs <= 3)
			break;
		old = atomic_cmpxchg(&eb->refs, refs, refs - 1);
		if (old == refs)
			return;
	}

5143
	spin_lock(&eb->refs_lock);
5144 5145 5146 5147
	if (atomic_read(&eb->refs) == 2 &&
	    test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags))
		atomic_dec(&eb->refs);

5148 5149
	if (atomic_read(&eb->refs) == 2 &&
	    test_bit(EXTENT_BUFFER_STALE, &eb->bflags) &&
5150
	    !extent_buffer_under_io(eb) &&
5151 5152 5153 5154 5155 5156 5157
	    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.
	 */
5158
	release_extent_buffer(eb);
5159 5160 5161 5162 5163
}

void free_extent_buffer_stale(struct extent_buffer *eb)
{
	if (!eb)
5164 5165
		return;

5166 5167 5168
	spin_lock(&eb->refs_lock);
	set_bit(EXTENT_BUFFER_STALE, &eb->bflags);

5169
	if (atomic_read(&eb->refs) == 2 && !extent_buffer_under_io(eb) &&
5170 5171
	    test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
		atomic_dec(&eb->refs);
5172
	release_extent_buffer(eb);
5173 5174
}

5175
void clear_extent_buffer_dirty(struct extent_buffer *eb)
5176 5177 5178 5179 5180 5181 5182 5183
{
	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++) {
5184
		page = eb->pages[i];
5185
		if (!PageDirty(page))
C
Chris Mason 已提交
5186 5187
			continue;

5188
		lock_page(page);
C
Chris Mason 已提交
5189 5190
		WARN_ON(!PagePrivate(page));

5191
		clear_page_dirty_for_io(page);
5192
		spin_lock_irq(&page->mapping->tree_lock);
5193 5194 5195 5196 5197
		if (!PageDirty(page)) {
			radix_tree_tag_clear(&page->mapping->page_tree,
						page_index(page),
						PAGECACHE_TAG_DIRTY);
		}
5198
		spin_unlock_irq(&page->mapping->tree_lock);
5199
		ClearPageError(page);
5200
		unlock_page(page);
5201
	}
5202
	WARN_ON(atomic_read(&eb->refs) == 0);
5203 5204
}

5205
int set_extent_buffer_dirty(struct extent_buffer *eb)
5206 5207 5208
{
	unsigned long i;
	unsigned long num_pages;
5209
	int was_dirty = 0;
5210

5211 5212
	check_buffer_tree_ref(eb);

5213
	was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
5214

5215
	num_pages = num_extent_pages(eb->start, eb->len);
5216
	WARN_ON(atomic_read(&eb->refs) == 0);
5217 5218
	WARN_ON(!test_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags));

5219
	for (i = 0; i < num_pages; i++)
5220
		set_page_dirty(eb->pages[i]);
5221
	return was_dirty;
5222 5223
}

5224
void clear_extent_buffer_uptodate(struct extent_buffer *eb)
5225 5226 5227 5228 5229
{
	unsigned long i;
	struct page *page;
	unsigned long num_pages;

5230
	clear_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5231
	num_pages = num_extent_pages(eb->start, eb->len);
5232
	for (i = 0; i < num_pages; i++) {
5233
		page = eb->pages[i];
C
Chris Mason 已提交
5234 5235
		if (page)
			ClearPageUptodate(page);
5236 5237 5238
	}
}

5239
void set_extent_buffer_uptodate(struct extent_buffer *eb)
5240 5241 5242 5243 5244
{
	unsigned long i;
	struct page *page;
	unsigned long num_pages;

5245
	set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5246 5247
	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = 0; i < num_pages; i++) {
5248
		page = eb->pages[i];
5249 5250 5251 5252
		SetPageUptodate(page);
	}
}

5253
int extent_buffer_uptodate(struct extent_buffer *eb)
5254
{
5255
	return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5256 5257 5258
}

int read_extent_buffer_pages(struct extent_io_tree *tree,
5259
			     struct extent_buffer *eb, u64 start, int wait,
5260
			     get_extent_t *get_extent, int mirror_num)
5261 5262 5263 5264 5265 5266
{
	unsigned long i;
	unsigned long start_i;
	struct page *page;
	int err;
	int ret = 0;
5267 5268
	int locked_pages = 0;
	int all_uptodate = 1;
5269
	unsigned long num_pages;
5270
	unsigned long num_reads = 0;
5271
	struct bio *bio = NULL;
C
Chris Mason 已提交
5272
	unsigned long bio_flags = 0;
5273

5274
	if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags))
5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286
		return 0;

	if (start) {
		WARN_ON(start < eb->start);
		start_i = (start >> PAGE_CACHE_SHIFT) -
			(eb->start >> PAGE_CACHE_SHIFT);
	} else {
		start_i = 0;
	}

	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = start_i; i < num_pages; i++) {
5287
		page = eb->pages[i];
5288
		if (wait == WAIT_NONE) {
5289
			if (!trylock_page(page))
5290
				goto unlock_exit;
5291 5292 5293
		} else {
			lock_page(page);
		}
5294
		locked_pages++;
5295 5296
		if (!PageUptodate(page)) {
			num_reads++;
5297
			all_uptodate = 0;
5298
		}
5299 5300 5301
	}
	if (all_uptodate) {
		if (start_i == 0)
5302
			set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5303 5304 5305
		goto unlock_exit;
	}

5306
	clear_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags);
5307
	eb->read_mirror = 0;
5308
	atomic_set(&eb->io_pages, num_reads);
5309
	for (i = start_i; i < num_pages; i++) {
5310
		page = eb->pages[i];
5311
		if (!PageUptodate(page)) {
5312
			ClearPageError(page);
5313
			err = __extent_read_full_page(tree, page,
5314
						      get_extent, &bio,
5315 5316
						      mirror_num, &bio_flags,
						      READ | REQ_META);
C
Chris Mason 已提交
5317
			if (err)
5318 5319 5320 5321 5322 5323
				ret = err;
		} else {
			unlock_page(page);
		}
	}

5324
	if (bio) {
5325 5326
		err = submit_one_bio(READ | REQ_META, bio, mirror_num,
				     bio_flags);
5327 5328
		if (err)
			return err;
5329
	}
5330

5331
	if (ret || wait != WAIT_COMPLETE)
5332
		return ret;
C
Chris Mason 已提交
5333

5334
	for (i = start_i; i < num_pages; i++) {
5335
		page = eb->pages[i];
5336
		wait_on_page_locked(page);
C
Chris Mason 已提交
5337
		if (!PageUptodate(page))
5338 5339
			ret = -EIO;
	}
C
Chris Mason 已提交
5340

5341
	return ret;
5342 5343 5344

unlock_exit:
	i = start_i;
C
Chris Mason 已提交
5345
	while (locked_pages > 0) {
5346
		page = eb->pages[i];
5347 5348 5349 5350 5351
		i++;
		unlock_page(page);
		locked_pages--;
	}
	return ret;
5352 5353 5354 5355 5356 5357 5358 5359 5360 5361 5362 5363 5364 5365 5366 5367 5368
}

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;
	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;

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

5369
	offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
5370

C
Chris Mason 已提交
5371
	while (len > 0) {
5372
		page = eb->pages[i];
5373 5374

		cur = min(len, (PAGE_CACHE_SIZE - offset));
5375
		kaddr = page_address(page);
5376 5377 5378 5379 5380 5381 5382 5383 5384
		memcpy(dst, kaddr + offset, cur);

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

5385 5386 5387 5388 5389 5390 5391 5392 5393 5394 5395 5396 5397 5398 5399 5400 5401 5402 5403
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;
	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
	int ret = 0;

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

	offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);

	while (len > 0) {
5404
		page = eb->pages[i];
5405 5406 5407 5408 5409 5410 5411 5412 5413 5414 5415 5416 5417 5418 5419 5420 5421

		cur = min(len, (PAGE_CACHE_SIZE - offset));
		kaddr = page_address(page);
		if (copy_to_user(dst, kaddr + offset, cur)) {
			ret = -EFAULT;
			break;
		}

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

	return ret;
}

5422
int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start,
5423
			       unsigned long min_len, char **map,
5424
			       unsigned long *map_start,
5425
			       unsigned long *map_len)
5426 5427 5428 5429 5430 5431 5432 5433 5434 5435 5436 5437 5438 5439 5440 5441 5442 5443 5444
{
	size_t offset = start & (PAGE_CACHE_SIZE - 1);
	char *kaddr;
	struct page *p;
	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
	unsigned long end_i = (start_offset + start + min_len - 1) >>
		PAGE_CACHE_SHIFT;

	if (i != end_i)
		return -EINVAL;

	if (i == 0) {
		offset = start_offset;
		*map_start = 0;
	} else {
		offset = 0;
		*map_start = ((u64)i << PAGE_CACHE_SHIFT) - start_offset;
	}
C
Chris Mason 已提交
5445

5446
	if (start + min_len > eb->len) {
J
Julia Lawall 已提交
5447
		WARN(1, KERN_ERR "btrfs bad mapping eb start %llu len %lu, "
5448 5449
		       "wanted %lu %lu\n",
		       eb->start, eb->len, start, min_len);
5450
		return -EINVAL;
5451 5452
	}

5453
	p = eb->pages[i];
5454
	kaddr = page_address(p);
5455 5456 5457 5458 5459 5460 5461 5462 5463 5464 5465 5466 5467 5468 5469 5470 5471 5472 5473 5474 5475
	*map = kaddr + offset;
	*map_len = PAGE_CACHE_SIZE - offset;
	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;
	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
	int ret = 0;

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

5476
	offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
5477

C
Chris Mason 已提交
5478
	while (len > 0) {
5479
		page = eb->pages[i];
5480 5481 5482

		cur = min(len, (PAGE_CACHE_SIZE - offset));

5483
		kaddr = page_address(page);
5484 5485 5486 5487 5488 5489 5490 5491 5492 5493 5494 5495 5496 5497 5498 5499 5500 5501 5502 5503 5504 5505 5506 5507 5508 5509
		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;
	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;

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

5510
	offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
5511

C
Chris Mason 已提交
5512
	while (len > 0) {
5513
		page = eb->pages[i];
5514 5515 5516
		WARN_ON(!PageUptodate(page));

		cur = min(len, PAGE_CACHE_SIZE - offset);
5517
		kaddr = page_address(page);
5518 5519 5520 5521 5522 5523 5524 5525 5526 5527 5528 5529 5530 5531 5532 5533 5534 5535 5536 5537 5538 5539
		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;
	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;

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

5540
	offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
5541

C
Chris Mason 已提交
5542
	while (len > 0) {
5543
		page = eb->pages[i];
5544 5545 5546
		WARN_ON(!PageUptodate(page));

		cur = min(len, PAGE_CACHE_SIZE - offset);
5547
		kaddr = page_address(page);
5548 5549 5550 5551 5552 5553 5554 5555 5556 5557 5558 5559 5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570
		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;
	size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1);
	unsigned long i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT;

	WARN_ON(src->len != dst_len);

	offset = (start_offset + dst_offset) &
5571
		(PAGE_CACHE_SIZE - 1);
5572

C
Chris Mason 已提交
5573
	while (len > 0) {
5574
		page = dst->pages[i];
5575 5576 5577 5578
		WARN_ON(!PageUptodate(page));

		cur = min(len, (unsigned long)(PAGE_CACHE_SIZE - offset));

5579
		kaddr = page_address(page);
5580 5581 5582 5583 5584 5585 5586 5587 5588
		read_extent_buffer(src, kaddr + offset, src_offset, cur);

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

5589 5590 5591 5592 5593 5594
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;
}

5595 5596 5597 5598
static void copy_pages(struct page *dst_page, struct page *src_page,
		       unsigned long dst_off, unsigned long src_off,
		       unsigned long len)
{
5599
	char *dst_kaddr = page_address(dst_page);
5600
	char *src_kaddr;
5601
	int must_memmove = 0;
5602

5603
	if (dst_page != src_page) {
5604
		src_kaddr = page_address(src_page);
5605
	} else {
5606
		src_kaddr = dst_kaddr;
5607 5608
		if (areas_overlap(src_off, dst_off, len))
			must_memmove = 1;
5609
	}
5610

5611 5612 5613 5614
	if (must_memmove)
		memmove(dst_kaddr + dst_off, src_kaddr + src_off, len);
	else
		memcpy(dst_kaddr + dst_off, src_kaddr + src_off, len);
5615 5616 5617 5618 5619 5620 5621 5622 5623 5624 5625 5626 5627
}

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;
	size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1);
	unsigned long dst_i;
	unsigned long src_i;

	if (src_offset + len > dst->len) {
5628 5629 5630
		btrfs_err(dst->fs_info,
			"memmove bogus src_offset %lu move "
		       "len %lu dst len %lu", src_offset, len, dst->len);
5631 5632 5633
		BUG_ON(1);
	}
	if (dst_offset + len > dst->len) {
5634 5635 5636
		btrfs_err(dst->fs_info,
			"memmove bogus dst_offset %lu move "
		       "len %lu dst len %lu", dst_offset, len, dst->len);
5637 5638 5639
		BUG_ON(1);
	}

C
Chris Mason 已提交
5640
	while (len > 0) {
5641
		dst_off_in_page = (start_offset + dst_offset) &
5642
			(PAGE_CACHE_SIZE - 1);
5643
		src_off_in_page = (start_offset + src_offset) &
5644
			(PAGE_CACHE_SIZE - 1);
5645 5646 5647 5648 5649 5650 5651 5652 5653

		dst_i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT;
		src_i = (start_offset + src_offset) >> PAGE_CACHE_SHIFT;

		cur = min(len, (unsigned long)(PAGE_CACHE_SIZE -
					       src_off_in_page));
		cur = min_t(unsigned long, cur,
			(unsigned long)(PAGE_CACHE_SIZE - dst_off_in_page));

5654
		copy_pages(dst->pages[dst_i], dst->pages[src_i],
5655 5656 5657 5658 5659 5660 5661 5662 5663 5664 5665 5666 5667 5668 5669 5670 5671 5672 5673 5674 5675
			   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;
	size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1);
	unsigned long dst_i;
	unsigned long src_i;

	if (src_offset + len > dst->len) {
5676 5677
		btrfs_err(dst->fs_info, "memmove bogus src_offset %lu move "
		       "len %lu len %lu", src_offset, len, dst->len);
5678 5679 5680
		BUG_ON(1);
	}
	if (dst_offset + len > dst->len) {
5681 5682
		btrfs_err(dst->fs_info, "memmove bogus dst_offset %lu move "
		       "len %lu len %lu", dst_offset, len, dst->len);
5683 5684
		BUG_ON(1);
	}
5685
	if (dst_offset < src_offset) {
5686 5687 5688
		memcpy_extent_buffer(dst, dst_offset, src_offset, len);
		return;
	}
C
Chris Mason 已提交
5689
	while (len > 0) {
5690 5691 5692 5693
		dst_i = (start_offset + dst_end) >> PAGE_CACHE_SHIFT;
		src_i = (start_offset + src_end) >> PAGE_CACHE_SHIFT;

		dst_off_in_page = (start_offset + dst_end) &
5694
			(PAGE_CACHE_SIZE - 1);
5695
		src_off_in_page = (start_offset + src_end) &
5696
			(PAGE_CACHE_SIZE - 1);
5697 5698 5699

		cur = min_t(unsigned long, len, src_off_in_page + 1);
		cur = min(cur, dst_off_in_page + 1);
5700
		copy_pages(dst->pages[dst_i], dst->pages[src_i],
5701 5702 5703 5704 5705 5706 5707 5708
			   dst_off_in_page - cur + 1,
			   src_off_in_page - cur + 1, cur);

		dst_end -= cur;
		src_end -= cur;
		len -= cur;
	}
}
5709

5710
int try_release_extent_buffer(struct page *page)
5711
{
5712 5713
	struct extent_buffer *eb;

5714 5715 5716 5717 5718 5719 5720
	/*
	 * We need to make sure noboody is attaching this page to an eb right
	 * now.
	 */
	spin_lock(&page->mapping->private_lock);
	if (!PagePrivate(page)) {
		spin_unlock(&page->mapping->private_lock);
J
Josef Bacik 已提交
5721
		return 1;
5722
	}
5723

5724 5725
	eb = (struct extent_buffer *)page->private;
	BUG_ON(!eb);
5726 5727

	/*
5728 5729 5730
	 * 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.
5731
	 */
5732
	spin_lock(&eb->refs_lock);
5733
	if (atomic_read(&eb->refs) != 1 || extent_buffer_under_io(eb)) {
5734 5735 5736
		spin_unlock(&eb->refs_lock);
		spin_unlock(&page->mapping->private_lock);
		return 0;
5737
	}
5738
	spin_unlock(&page->mapping->private_lock);
5739

5740
	/*
5741 5742
	 * 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.
5743
	 */
5744 5745 5746
	if (!test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags)) {
		spin_unlock(&eb->refs_lock);
		return 0;
5747
	}
5748

5749
	return release_extent_buffer(eb);
5750
}