extent_io.c 148.8 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
#include "transaction.h"
24 25 26

static struct kmem_cache *extent_state_cache;
static struct kmem_cache *extent_buffer_cache;
27
static struct bio_set *btrfs_bioset;
28

29 30 31 32 33
static inline bool extent_state_in_tree(const struct extent_state *state)
{
	return !RB_EMPTY_NODE(&state->rb_node);
}

34
#ifdef CONFIG_BTRFS_DEBUG
35 36
static LIST_HEAD(buffers);
static LIST_HEAD(states);
C
Chris Mason 已提交
37

C
Chris Mason 已提交
38
static DEFINE_SPINLOCK(leak_lock);
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 67

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);
68
		pr_err("BTRFS: state leak: start %llu end %llu state %u in tree %d refs %d\n",
69 70
		       state->start, state->end, state->state,
		       extent_state_in_tree(state),
71
		       atomic_read(&state->refs));
72 73 74 75 76 77
		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);
78
		pr_err("BTRFS: buffer leak start %llu len %lu refs %d\n",
79
		       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(BTRFS_I(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
	unsigned int extent_locked:1;

130
	/* tells the submit_bio code to use REQ_SYNC */
131
	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
	changeset->bytes_changed += state->end - state->start + 1;
147
	ret = ulist_add(&changeset->range_changed, state->start, state->end,
148 149 150 151 152
			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
			sizeof(struct extent_state), 0,
166
			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
			sizeof(struct extent_buffer), 0,
172
			SLAB_MEM_SPREAD, NULL);
173 174
	if (!extent_buffer_cache)
		goto free_state_cache;
175 176 177 178 179

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

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

184 185
	return 0;

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

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

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

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

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

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

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

229 230 231 232 233
	/*
	 * The given mask might be not appropriate for the slab allocator,
	 * drop the unsupported bits
	 */
	mask &= ~(__GFP_DMA32|__GFP_HIGHMEM);
234
	state = kmem_cache_alloc(extent_state_cache, mask);
235
	if (!state)
236 237
		return state;
	state->state = 0;
238
	state->failrec = NULL;
239
	RB_CLEAR_NODE(&state->rb_node);
240
	btrfs_leak_debug_add(&state->leak_list, &states);
241 242
	atomic_set(&state->refs, 1);
	init_waitqueue_head(&state->wq);
243
	trace_alloc_extent_state(state, mask, _RET_IP_);
244 245 246
	return state;
}

247
void free_extent_state(struct extent_state *state)
248 249 250 251
{
	if (!state)
		return;
	if (atomic_dec_and_test(&state->refs)) {
252
		WARN_ON(extent_state_in_tree(state));
253
		btrfs_leak_debug_del(&state->leak_list);
254
		trace_free_extent_state(state, _RET_IP_);
255 256 257 258
		kmem_cache_free(extent_state_cache, state);
	}
}

259 260 261
static struct rb_node *tree_insert(struct rb_root *root,
				   struct rb_node *search_start,
				   u64 offset,
262 263 264
				   struct rb_node *node,
				   struct rb_node ***p_in,
				   struct rb_node **parent_in)
265
{
266
	struct rb_node **p;
C
Chris Mason 已提交
267
	struct rb_node *parent = NULL;
268 269
	struct tree_entry *entry;

270 271 272 273 274 275
	if (p_in && parent_in) {
		p = *p_in;
		parent = *parent_in;
		goto do_insert;
	}

276
	p = search_start ? &search_start : &root->rb_node;
C
Chris Mason 已提交
277
	while (*p) {
278 279 280 281 282 283 284 285 286 287 288
		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;
	}

289
do_insert:
290 291 292 293 294
	rb_link_node(node, parent, p);
	rb_insert_color(node, root);
	return NULL;
}

295
static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset,
296 297 298 299
				      struct rb_node **prev_ret,
				      struct rb_node **next_ret,
				      struct rb_node ***p_ret,
				      struct rb_node **parent_ret)
300
{
301
	struct rb_root *root = &tree->state;
302
	struct rb_node **n = &root->rb_node;
303 304 305 306 307
	struct rb_node *prev = NULL;
	struct rb_node *orig_prev = NULL;
	struct tree_entry *entry;
	struct tree_entry *prev_entry = NULL;

308 309 310
	while (*n) {
		prev = *n;
		entry = rb_entry(prev, struct tree_entry, rb_node);
311 312 313
		prev_entry = entry;

		if (offset < entry->start)
314
			n = &(*n)->rb_left;
315
		else if (offset > entry->end)
316
			n = &(*n)->rb_right;
C
Chris Mason 已提交
317
		else
318
			return *n;
319 320
	}

321 322 323 324 325
	if (p_ret)
		*p_ret = n;
	if (parent_ret)
		*parent_ret = prev;

326 327
	if (prev_ret) {
		orig_prev = prev;
C
Chris Mason 已提交
328
		while (prev && offset > prev_entry->end) {
329 330 331 332 333 334 335 336 337
			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
338
		while (prev && offset < prev_entry->start) {
339 340 341 342 343 344 345 346
			prev = rb_prev(prev);
			prev_entry = rb_entry(prev, struct tree_entry, rb_node);
		}
		*next_ret = prev;
	}
	return NULL;
}

347 348 349 350 351
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)
352
{
353
	struct rb_node *prev = NULL;
354
	struct rb_node *ret;
355

356
	ret = __etree_search(tree, offset, &prev, NULL, p_ret, parent_ret);
C
Chris Mason 已提交
357
	if (!ret)
358 359 360 361
		return prev;
	return ret;
}

362 363 364 365 366 367
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
368 369 370 371 372 373 374 375
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);
}

376 377 378 379 380 381 382 383 384
/*
 * 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.
 */
385 386
static void merge_state(struct extent_io_tree *tree,
		        struct extent_state *state)
387 388 389 390
{
	struct extent_state *other;
	struct rb_node *other_node;

391
	if (state->state & (EXTENT_IOBITS | EXTENT_BOUNDARY))
392
		return;
393 394 395 396 397 398

	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
399
			merge_cb(tree, state, other);
400 401
			state->start = other->start;
			rb_erase(&other->rb_node, &tree->state);
402
			RB_CLEAR_NODE(&other->rb_node);
403 404 405 406 407 408 409 410
			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
411
			merge_cb(tree, state, other);
412 413
			state->end = other->end;
			rb_erase(&other->rb_node, &tree->state);
414
			RB_CLEAR_NODE(&other->rb_node);
415
			free_extent_state(other);
416 417 418 419
		}
	}
}

420
static void set_state_cb(struct extent_io_tree *tree,
421
			 struct extent_state *state, unsigned *bits)
422
{
423 424
	if (tree->ops && tree->ops->set_bit_hook)
		tree->ops->set_bit_hook(tree->mapping->host, state, bits);
425 426 427
}

static void clear_state_cb(struct extent_io_tree *tree,
428
			   struct extent_state *state, unsigned *bits)
429
{
J
Josef Bacik 已提交
430 431
	if (tree->ops && tree->ops->clear_bit_hook)
		tree->ops->clear_bit_hook(tree->mapping->host, state, bits);
432 433
}

434
static void set_state_bits(struct extent_io_tree *tree,
435 436
			   struct extent_state *state, unsigned *bits,
			   struct extent_changeset *changeset);
437

438 439 440 441 442 443 444 445 446 447 448 449
/*
 * 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,
450 451
			struct rb_node ***p,
			struct rb_node **parent,
452
			unsigned *bits, struct extent_changeset *changeset)
453 454 455
{
	struct rb_node *node;

J
Julia Lawall 已提交
456
	if (end < start)
457
		WARN(1, KERN_ERR "BTRFS: end < start %llu %llu\n",
458
		       end, start);
459 460
	state->start = start;
	state->end = end;
J
Josef Bacik 已提交
461

462
	set_state_bits(tree, state, bits, changeset);
463

464
	node = tree_insert(&tree->state, NULL, end, &state->rb_node, p, parent);
465 466 467
	if (node) {
		struct extent_state *found;
		found = rb_entry(node, struct extent_state, rb_node);
468
		pr_err("BTRFS: found node %llu %llu on insert of %llu %llu\n",
469
		       found->start, found->end, start, end);
470 471 472 473 474 475
		return -EEXIST;
	}
	merge_state(tree, state);
	return 0;
}

476
static void split_cb(struct extent_io_tree *tree, struct extent_state *orig,
J
Josef Bacik 已提交
477 478 479
		     u64 split)
{
	if (tree->ops && tree->ops->split_extent_hook)
480
		tree->ops->split_extent_hook(tree->mapping->host, orig, split);
J
Josef Bacik 已提交
481 482
}

483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500
/*
 * 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
501 502 503

	split_cb(tree, orig, split);

504 505 506 507 508
	prealloc->start = orig->start;
	prealloc->end = split - 1;
	prealloc->state = orig->state;
	orig->start = split;

509 510
	node = tree_insert(&tree->state, &orig->rb_node, prealloc->end,
			   &prealloc->rb_node, NULL, NULL);
511 512 513 514 515 516 517
	if (node) {
		free_extent_state(prealloc);
		return -EEXIST;
	}
	return 0;
}

518 519 520 521 522 523 524 525 526
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;
}

527 528
/*
 * utility function to clear some bits in an extent state struct.
529
 * it will optionally wake up any one waiting on this state (wake == 1).
530 531 532 533
 *
 * If no bits are set on the state struct after clearing things, the
 * struct is freed and removed from the tree
 */
534 535
static struct extent_state *clear_state_bit(struct extent_io_tree *tree,
					    struct extent_state *state,
536 537
					    unsigned *bits, int wake,
					    struct extent_changeset *changeset)
538
{
539
	struct extent_state *next;
540
	unsigned bits_to_clear = *bits & ~EXTENT_CTLBITS;
541

542
	if ((bits_to_clear & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) {
543 544 545 546
		u64 range = state->end - state->start + 1;
		WARN_ON(range > tree->dirty_bytes);
		tree->dirty_bytes -= range;
	}
547
	clear_state_cb(tree, state, bits);
548
	add_extent_changeset(state, bits_to_clear, changeset, 0);
549
	state->state &= ~bits_to_clear;
550 551
	if (wake)
		wake_up(&state->wq);
552
	if (state->state == 0) {
553
		next = next_state(state);
554
		if (extent_state_in_tree(state)) {
555
			rb_erase(&state->rb_node, &tree->state);
556
			RB_CLEAR_NODE(&state->rb_node);
557 558 559 560 561 562
			free_extent_state(state);
		} else {
			WARN_ON(1);
		}
	} else {
		merge_state(tree, state);
563
		next = next_state(state);
564
	}
565
	return next;
566 567
}

568 569 570 571 572 573 574 575 576
static struct extent_state *
alloc_extent_state_atomic(struct extent_state *prealloc)
{
	if (!prealloc)
		prealloc = alloc_extent_state(GFP_ATOMIC);

	return prealloc;
}

577
static void extent_io_tree_panic(struct extent_io_tree *tree, int err)
578
{
J
Jeff Mahoney 已提交
579 580
	btrfs_panic(tree_fs_info(tree), err,
		    "Locking error: Extent tree was modified by another thread while locked.");
581 582
}

583 584 585 586 587 588 589 590 591 592
/*
 * 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.
 *
593
 * This takes the tree lock, and returns 0 on success and < 0 on error.
594
 */
595 596 597 598
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)
599 600
{
	struct extent_state *state;
601
	struct extent_state *cached;
602 603
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
604
	u64 last_end;
605
	int err;
606
	int clear = 0;
607

608
	btrfs_debug_check_extent_io_range(tree, start, end);
609

610 611 612
	if (bits & EXTENT_DELALLOC)
		bits |= EXTENT_NORESERVE;

613 614 615 616
	if (delete)
		bits |= ~EXTENT_CTLBITS;
	bits |= EXTENT_FIRST_DELALLOC;

617 618
	if (bits & (EXTENT_IOBITS | EXTENT_BOUNDARY))
		clear = 1;
619
again:
620
	if (!prealloc && gfpflags_allow_blocking(mask)) {
621 622 623 624 625 626 627
		/*
		 * 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.
		 */
628 629 630
		prealloc = alloc_extent_state(mask);
	}

631
	spin_lock(&tree->lock);
632 633
	if (cached_state) {
		cached = *cached_state;
634 635 636 637 638 639

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

640 641
		if (cached && extent_state_in_tree(cached) &&
		    cached->start <= start && cached->end > start) {
642 643
			if (clear)
				atomic_dec(&cached->refs);
644
			state = cached;
645
			goto hit_next;
646
		}
647 648
		if (clear)
			free_extent_state(cached);
649
	}
650 651 652 653
	/*
	 * this search will find the extents that end after
	 * our range starts
	 */
654
	node = tree_search(tree, start);
655 656 657
	if (!node)
		goto out;
	state = rb_entry(node, struct extent_state, rb_node);
658
hit_next:
659 660 661
	if (state->start > end)
		goto out;
	WARN_ON(state->end < start);
662
	last_end = state->end;
663

664
	/* the state doesn't have the wanted bits, go ahead */
665 666
	if (!(state->state & bits)) {
		state = next_state(state);
667
		goto next;
668
	}
669

670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686
	/*
	 *     | ---- 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) {
687 688
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
689
		err = split_state(tree, state, prealloc, start);
690 691 692
		if (err)
			extent_io_tree_panic(tree, err);

693 694 695 696
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
697 698
			state = clear_state_bit(tree, state, &bits, wake,
						changeset);
699
			goto next;
700 701 702 703 704 705 706 707 708 709
		}
		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) {
710 711
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
712
		err = split_state(tree, state, prealloc, end + 1);
713 714 715
		if (err)
			extent_io_tree_panic(tree, err);

716 717
		if (wake)
			wake_up(&state->wq);
718

719
		clear_state_bit(tree, prealloc, &bits, wake, changeset);
J
Josef Bacik 已提交
720

721 722 723
		prealloc = NULL;
		goto out;
	}
724

725
	state = clear_state_bit(tree, state, &bits, wake, changeset);
726
next:
727 728 729
	if (last_end == (u64)-1)
		goto out;
	start = last_end + 1;
730
	if (start <= end && state && !need_resched())
731
		goto hit_next;
732 733 734 735

search_again:
	if (start > end)
		goto out;
736
	spin_unlock(&tree->lock);
737
	if (gfpflags_allow_blocking(mask))
738 739
		cond_resched();
	goto again;
740 741 742 743 744 745 746 747

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

	return 0;

748 749
}

750 751
static void wait_on_state(struct extent_io_tree *tree,
			  struct extent_state *state)
752 753
		__releases(tree->lock)
		__acquires(tree->lock)
754 755 756
{
	DEFINE_WAIT(wait);
	prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE);
757
	spin_unlock(&tree->lock);
758
	schedule();
759
	spin_lock(&tree->lock);
760 761 762 763 764 765 766 767
	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
 */
768 769
static void wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
			    unsigned long bits)
770 771 772 773
{
	struct extent_state *state;
	struct rb_node *node;

774
	btrfs_debug_check_extent_io_range(tree, start, end);
775

776
	spin_lock(&tree->lock);
777 778 779 780 781 782
again:
	while (1) {
		/*
		 * this search will find all the extents that end after
		 * our range starts
		 */
783
		node = tree_search(tree, start);
784
process_node:
785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804
		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;

805 806 807 808
		if (!cond_resched_lock(&tree->lock)) {
			node = rb_next(node);
			goto process_node;
		}
809 810
	}
out:
811
	spin_unlock(&tree->lock);
812 813
}

814
static void set_state_bits(struct extent_io_tree *tree,
815
			   struct extent_state *state,
816
			   unsigned *bits, struct extent_changeset *changeset)
817
{
818
	unsigned bits_to_set = *bits & ~EXTENT_CTLBITS;
J
Josef Bacik 已提交
819

820
	set_state_cb(tree, state, bits);
821
	if ((bits_to_set & EXTENT_DIRTY) && !(state->state & EXTENT_DIRTY)) {
822 823 824
		u64 range = state->end - state->start + 1;
		tree->dirty_bytes += range;
	}
825
	add_extent_changeset(state, bits_to_set, changeset, 1);
826
	state->state |= bits_to_set;
827 828
}

829 830
static void cache_state_if_flags(struct extent_state *state,
				 struct extent_state **cached_ptr,
831
				 unsigned flags)
832 833
{
	if (cached_ptr && !(*cached_ptr)) {
834
		if (!flags || (state->state & flags)) {
835 836 837 838 839 840
			*cached_ptr = state;
			atomic_inc(&state->refs);
		}
	}
}

841 842 843 844 845 846 847
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);
}

848
/*
849 850
 * 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.
851
 *
852 853 854
 * 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.
855
 *
856
 * [start, end] is inclusive This takes the tree lock.
857
 */
858

J
Jeff Mahoney 已提交
859 860
static int __must_check
__set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
861
		 unsigned bits, unsigned exclusive_bits,
862
		 u64 *failed_start, struct extent_state **cached_state,
863
		 gfp_t mask, struct extent_changeset *changeset)
864 865 866 867
{
	struct extent_state *state;
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
868 869
	struct rb_node **p;
	struct rb_node *parent;
870 871 872
	int err = 0;
	u64 last_start;
	u64 last_end;
873

874
	btrfs_debug_check_extent_io_range(tree, start, end);
875

876
	bits |= EXTENT_FIRST_DELALLOC;
877
again:
878
	if (!prealloc && gfpflags_allow_blocking(mask)) {
879 880 881 882 883 884 885
		/*
		 * 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.
		 */
886 887 888
		prealloc = alloc_extent_state(mask);
	}

889
	spin_lock(&tree->lock);
890 891
	if (cached_state && *cached_state) {
		state = *cached_state;
892
		if (state->start <= start && state->end > start &&
893
		    extent_state_in_tree(state)) {
894 895 896 897
			node = &state->rb_node;
			goto hit_next;
		}
	}
898 899 900 901
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
902
	node = tree_search_for_insert(tree, start, &p, &parent);
903
	if (!node) {
904 905
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
906
		err = insert_state(tree, prealloc, start, end,
907
				   &p, &parent, &bits, changeset);
908 909 910
		if (err)
			extent_io_tree_panic(tree, err);

911
		cache_state(prealloc, cached_state);
912 913 914 915
		prealloc = NULL;
		goto out;
	}
	state = rb_entry(node, struct extent_state, rb_node);
C
Chris Mason 已提交
916
hit_next:
917 918 919 920 921 922 923 924 925 926
	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) {
927
		if (state->state & exclusive_bits) {
928 929 930 931
			*failed_start = state->start;
			err = -EEXIST;
			goto out;
		}
932

933
		set_state_bits(tree, state, &bits, changeset);
934
		cache_state(state, cached_state);
935
		merge_state(tree, state);
936 937 938
		if (last_end == (u64)-1)
			goto out;
		start = last_end + 1;
939 940 941 942
		state = next_state(state);
		if (start < end && state && state->start == start &&
		    !need_resched())
			goto hit_next;
943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962
		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) {
963
		if (state->state & exclusive_bits) {
964 965 966 967
			*failed_start = start;
			err = -EEXIST;
			goto out;
		}
968 969 970

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
971
		err = split_state(tree, state, prealloc, start);
972 973 974
		if (err)
			extent_io_tree_panic(tree, err);

975 976 977 978
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
979
			set_state_bits(tree, state, &bits, changeset);
980
			cache_state(state, cached_state);
981
			merge_state(tree, state);
982 983 984
			if (last_end == (u64)-1)
				goto out;
			start = last_end + 1;
985 986 987 988
			state = next_state(state);
			if (start < end && state && state->start == start &&
			    !need_resched())
				goto hit_next;
989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003
		}
		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
1004
			this_end = last_start - 1;
1005 1006 1007

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
1008 1009 1010 1011 1012

		/*
		 * Avoid to free 'prealloc' if it can be merged with
		 * the later extent.
		 */
1013
		err = insert_state(tree, prealloc, start, this_end,
1014
				   NULL, NULL, &bits, changeset);
1015 1016 1017
		if (err)
			extent_io_tree_panic(tree, err);

J
Josef Bacik 已提交
1018 1019
		cache_state(prealloc, cached_state);
		prealloc = NULL;
1020 1021 1022 1023 1024 1025 1026 1027 1028 1029
		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) {
1030
		if (state->state & exclusive_bits) {
1031 1032 1033 1034
			*failed_start = start;
			err = -EEXIST;
			goto out;
		}
1035 1036 1037

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
1038
		err = split_state(tree, state, prealloc, end + 1);
1039 1040
		if (err)
			extent_io_tree_panic(tree, err);
1041

1042
		set_state_bits(tree, prealloc, &bits, changeset);
1043
		cache_state(prealloc, cached_state);
1044 1045 1046 1047 1048
		merge_state(tree, prealloc);
		prealloc = NULL;
		goto out;
	}

1049 1050 1051 1052 1053 1054 1055
search_again:
	if (start > end)
		goto out;
	spin_unlock(&tree->lock);
	if (gfpflags_allow_blocking(mask))
		cond_resched();
	goto again;
1056 1057

out:
1058
	spin_unlock(&tree->lock);
1059 1060 1061 1062 1063 1064 1065
	if (prealloc)
		free_extent_state(prealloc);

	return err;

}

1066
int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
1067
		   unsigned bits, u64 * failed_start,
1068
		   struct extent_state **cached_state, gfp_t mask)
J
Jeff Mahoney 已提交
1069 1070
{
	return __set_extent_bit(tree, start, end, bits, 0, failed_start,
1071
				cached_state, mask, NULL);
J
Jeff Mahoney 已提交
1072 1073 1074
}


J
Josef Bacik 已提交
1075
/**
L
Liu Bo 已提交
1076 1077
 * convert_extent_bit - convert all bits in a given range from one bit to
 * 			another
J
Josef Bacik 已提交
1078 1079 1080 1081 1082
 * @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
1083
 * @cached_state:	state that we're going to cache
J
Josef Bacik 已提交
1084 1085 1086 1087 1088 1089
 *
 * 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.
1090 1091
 *
 * All allocations are done with GFP_NOFS.
J
Josef Bacik 已提交
1092 1093
 */
int convert_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
1094
		       unsigned bits, unsigned clear_bits,
1095
		       struct extent_state **cached_state)
J
Josef Bacik 已提交
1096 1097 1098 1099
{
	struct extent_state *state;
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
1100 1101
	struct rb_node **p;
	struct rb_node *parent;
J
Josef Bacik 已提交
1102 1103 1104
	int err = 0;
	u64 last_start;
	u64 last_end;
1105
	bool first_iteration = true;
J
Josef Bacik 已提交
1106

1107
	btrfs_debug_check_extent_io_range(tree, start, end);
1108

J
Josef Bacik 已提交
1109
again:
1110
	if (!prealloc) {
1111 1112 1113 1114 1115 1116 1117
		/*
		 * 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.
		 */
1118
		prealloc = alloc_extent_state(GFP_NOFS);
1119
		if (!prealloc && !first_iteration)
J
Josef Bacik 已提交
1120 1121 1122 1123
			return -ENOMEM;
	}

	spin_lock(&tree->lock);
1124 1125 1126
	if (cached_state && *cached_state) {
		state = *cached_state;
		if (state->start <= start && state->end > start &&
1127
		    extent_state_in_tree(state)) {
1128 1129 1130 1131 1132
			node = &state->rb_node;
			goto hit_next;
		}
	}

J
Josef Bacik 已提交
1133 1134 1135 1136
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1137
	node = tree_search_for_insert(tree, start, &p, &parent);
J
Josef Bacik 已提交
1138 1139
	if (!node) {
		prealloc = alloc_extent_state_atomic(prealloc);
1140 1141 1142 1143
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
1144
		err = insert_state(tree, prealloc, start, end,
1145
				   &p, &parent, &bits, NULL);
1146 1147
		if (err)
			extent_io_tree_panic(tree, err);
1148 1149
		cache_state(prealloc, cached_state);
		prealloc = NULL;
J
Josef Bacik 已提交
1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163
		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) {
1164
		set_state_bits(tree, state, &bits, NULL);
1165
		cache_state(state, cached_state);
1166
		state = clear_state_bit(tree, state, &clear_bits, 0, NULL);
J
Josef Bacik 已提交
1167 1168 1169
		if (last_end == (u64)-1)
			goto out;
		start = last_end + 1;
1170 1171 1172
		if (start < end && state && state->start == start &&
		    !need_resched())
			goto hit_next;
J
Josef Bacik 已提交
1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193
		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);
1194 1195 1196 1197
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1198
		err = split_state(tree, state, prealloc, start);
1199 1200
		if (err)
			extent_io_tree_panic(tree, err);
J
Josef Bacik 已提交
1201 1202 1203 1204
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
1205
			set_state_bits(tree, state, &bits, NULL);
1206
			cache_state(state, cached_state);
1207 1208
			state = clear_state_bit(tree, state, &clear_bits, 0,
						NULL);
J
Josef Bacik 已提交
1209 1210 1211
			if (last_end == (u64)-1)
				goto out;
			start = last_end + 1;
1212 1213 1214
			if (start < end && state && state->start == start &&
			    !need_resched())
				goto hit_next;
J
Josef Bacik 已提交
1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232
		}
		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);
1233 1234 1235 1236
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1237 1238 1239 1240 1241 1242

		/*
		 * Avoid to free 'prealloc' if it can be merged with
		 * the later extent.
		 */
		err = insert_state(tree, prealloc, start, this_end,
1243
				   NULL, NULL, &bits, NULL);
1244 1245
		if (err)
			extent_io_tree_panic(tree, err);
1246
		cache_state(prealloc, cached_state);
J
Josef Bacik 已提交
1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258
		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);
1259 1260 1261 1262
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1263 1264

		err = split_state(tree, state, prealloc, end + 1);
1265 1266
		if (err)
			extent_io_tree_panic(tree, err);
J
Josef Bacik 已提交
1267

1268
		set_state_bits(tree, prealloc, &bits, NULL);
1269
		cache_state(prealloc, cached_state);
1270
		clear_state_bit(tree, prealloc, &clear_bits, 0, NULL);
J
Josef Bacik 已提交
1271 1272 1273 1274 1275 1276 1277 1278
		prealloc = NULL;
		goto out;
	}

search_again:
	if (start > end)
		goto out;
	spin_unlock(&tree->lock);
1279
	cond_resched();
1280
	first_iteration = false;
J
Josef Bacik 已提交
1281 1282 1283 1284 1285 1286 1287 1288 1289 1290
	goto again;

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

	return err;
}

1291
/* wrappers around set/clear extent bit */
1292
int set_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1293
			   unsigned bits, struct extent_changeset *changeset)
1294 1295 1296 1297 1298 1299 1300 1301 1302
{
	/*
	 * 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);

1303
	return __set_extent_bit(tree, start, end, bits, 0, NULL, NULL, GFP_NOFS,
1304 1305 1306
				changeset);
}

1307 1308 1309 1310 1311 1312 1313 1314 1315
int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
		     unsigned bits, int wake, int delete,
		     struct extent_state **cached, gfp_t mask)
{
	return __clear_extent_bit(tree, start, end, bits, wake, delete,
				  cached, mask, NULL);
}

int clear_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1316
		unsigned bits, struct extent_changeset *changeset)
1317 1318 1319 1320 1321 1322 1323
{
	/*
	 * Don't support EXTENT_LOCKED case, same reason as
	 * set_record_extent_bits().
	 */
	BUG_ON(bits & EXTENT_LOCKED);

1324
	return __clear_extent_bit(tree, start, end, bits, 0, 0, NULL, GFP_NOFS,
1325 1326 1327
				  changeset);
}

C
Chris Mason 已提交
1328 1329 1330 1331
/*
 * either insert or lock state struct between start and end use mask to tell
 * us if waiting is desired.
 */
1332
int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1333
		     struct extent_state **cached_state)
1334 1335 1336
{
	int err;
	u64 failed_start;
1337

1338
	while (1) {
1339
		err = __set_extent_bit(tree, start, end, EXTENT_LOCKED,
J
Jeff Mahoney 已提交
1340
				       EXTENT_LOCKED, &failed_start,
1341
				       cached_state, GFP_NOFS, NULL);
1342
		if (err == -EEXIST) {
1343 1344
			wait_extent_bit(tree, failed_start, end, EXTENT_LOCKED);
			start = failed_start;
1345
		} else
1346 1347 1348 1349 1350 1351
			break;
		WARN_ON(start > end);
	}
	return err;
}

1352
int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
1353 1354 1355 1356
{
	int err;
	u64 failed_start;

J
Jeff Mahoney 已提交
1357
	err = __set_extent_bit(tree, start, end, EXTENT_LOCKED, EXTENT_LOCKED,
1358
			       &failed_start, NULL, GFP_NOFS, NULL);
Y
Yan Zheng 已提交
1359 1360 1361
	if (err == -EEXIST) {
		if (failed_start > start)
			clear_extent_bit(tree, start, failed_start - 1,
1362
					 EXTENT_LOCKED, 1, 0, NULL, GFP_NOFS);
1363
		return 0;
Y
Yan Zheng 已提交
1364
	}
1365 1366 1367
	return 1;
}

1368
void extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end)
1369
{
1370 1371
	unsigned long index = start >> PAGE_SHIFT;
	unsigned long end_index = end >> PAGE_SHIFT;
1372 1373 1374 1375 1376 1377
	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);
1378
		put_page(page);
1379 1380 1381 1382
		index++;
	}
}

1383
void extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end)
1384
{
1385 1386
	unsigned long index = start >> PAGE_SHIFT;
	unsigned long end_index = end >> PAGE_SHIFT;
1387 1388 1389 1390 1391 1392
	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);
1393
		account_page_redirty(page);
1394
		put_page(page);
1395 1396 1397 1398
		index++;
	}
}

1399 1400 1401
/*
 * helper function to set both pages and extents in the tree writeback
 */
1402
static void set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end)
1403
{
1404 1405
	unsigned long index = start >> PAGE_SHIFT;
	unsigned long end_index = end >> PAGE_SHIFT;
1406 1407 1408 1409
	struct page *page;

	while (index <= end_index) {
		page = find_get_page(tree->mapping, index);
1410
		BUG_ON(!page); /* Pages should be in the extent_io_tree */
1411
		set_page_writeback(page);
1412
		put_page(page);
1413 1414 1415 1416
		index++;
	}
}

C
Chris Mason 已提交
1417 1418 1419 1420
/* 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'
 */
1421 1422
static struct extent_state *
find_first_extent_bit_state(struct extent_io_tree *tree,
1423
			    u64 start, unsigned bits)
C
Chris Mason 已提交
1424 1425 1426 1427 1428 1429 1430 1431 1432
{
	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
1433
	if (!node)
C
Chris Mason 已提交
1434 1435
		goto out;

C
Chris Mason 已提交
1436
	while (1) {
C
Chris Mason 已提交
1437
		state = rb_entry(node, struct extent_state, rb_node);
C
Chris Mason 已提交
1438
		if (state->end >= start && (state->state & bits))
C
Chris Mason 已提交
1439
			return state;
C
Chris Mason 已提交
1440

C
Chris Mason 已提交
1441 1442 1443 1444 1445 1446 1447 1448
		node = rb_next(node);
		if (!node)
			break;
	}
out:
	return NULL;
}

1449 1450 1451 1452 1453
/*
 * 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.
 *
1454
 * If nothing was found, 1 is returned. If found something, return 0.
1455 1456
 */
int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
1457
			  u64 *start_ret, u64 *end_ret, unsigned bits,
1458
			  struct extent_state **cached_state)
1459 1460
{
	struct extent_state *state;
1461
	struct rb_node *n;
1462 1463 1464
	int ret = 1;

	spin_lock(&tree->lock);
1465 1466
	if (cached_state && *cached_state) {
		state = *cached_state;
1467
		if (state->end == start - 1 && extent_state_in_tree(state)) {
1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483
			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;
	}

1484
	state = find_first_extent_bit_state(tree, start, bits);
1485
got_it:
1486
	if (state) {
1487
		cache_state_if_flags(state, cached_state, 0);
1488 1489 1490 1491
		*start_ret = state->start;
		*end_ret = state->end;
		ret = 0;
	}
1492
out:
1493 1494 1495 1496
	spin_unlock(&tree->lock);
	return ret;
}

C
Chris Mason 已提交
1497 1498 1499 1500 1501 1502
/*
 * 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
1503
static noinline u64 find_delalloc_range(struct extent_io_tree *tree,
1504 1505
					u64 *start, u64 *end, u64 max_bytes,
					struct extent_state **cached_state)
1506 1507 1508 1509 1510 1511 1512
{
	struct rb_node *node;
	struct extent_state *state;
	u64 cur_start = *start;
	u64 found = 0;
	u64 total_bytes = 0;

1513
	spin_lock(&tree->lock);
C
Chris Mason 已提交
1514

1515 1516 1517 1518
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1519
	node = tree_search(tree, cur_start);
1520
	if (!node) {
1521 1522
		if (!found)
			*end = (u64)-1;
1523 1524 1525
		goto out;
	}

C
Chris Mason 已提交
1526
	while (1) {
1527
		state = rb_entry(node, struct extent_state, rb_node);
1528 1529
		if (found && (state->start != cur_start ||
			      (state->state & EXTENT_BOUNDARY))) {
1530 1531 1532 1533 1534 1535 1536
			goto out;
		}
		if (!(state->state & EXTENT_DELALLOC)) {
			if (!found)
				*end = state->end;
			goto out;
		}
1537
		if (!found) {
1538
			*start = state->start;
1539 1540 1541
			*cached_state = state;
			atomic_inc(&state->refs);
		}
1542 1543 1544 1545 1546
		found++;
		*end = state->end;
		cur_start = state->end + 1;
		node = rb_next(node);
		total_bytes += state->end - state->start + 1;
1547
		if (total_bytes >= max_bytes)
1548 1549
			break;
		if (!node)
1550 1551 1552
			break;
	}
out:
1553
	spin_unlock(&tree->lock);
1554 1555 1556
	return found;
}

1557 1558 1559 1560 1561
static int __process_pages_contig(struct address_space *mapping,
				  struct page *locked_page,
				  pgoff_t start_index, pgoff_t end_index,
				  unsigned long page_ops, pgoff_t *index_ret);

1562 1563 1564
static noinline void __unlock_for_delalloc(struct inode *inode,
					   struct page *locked_page,
					   u64 start, u64 end)
C
Chris Mason 已提交
1565
{
1566 1567
	unsigned long index = start >> PAGE_SHIFT;
	unsigned long end_index = end >> PAGE_SHIFT;
C
Chris Mason 已提交
1568

1569
	ASSERT(locked_page);
C
Chris Mason 已提交
1570
	if (index == locked_page->index && end_index == index)
1571
		return;
C
Chris Mason 已提交
1572

1573 1574
	__process_pages_contig(inode->i_mapping, locked_page, index, end_index,
			       PAGE_UNLOCK, NULL);
C
Chris Mason 已提交
1575 1576 1577 1578 1579 1580 1581
}

static noinline int lock_delalloc_pages(struct inode *inode,
					struct page *locked_page,
					u64 delalloc_start,
					u64 delalloc_end)
{
1582
	unsigned long index = delalloc_start >> PAGE_SHIFT;
1583
	unsigned long index_ret = index;
1584
	unsigned long end_index = delalloc_end >> PAGE_SHIFT;
C
Chris Mason 已提交
1585 1586
	int ret;

1587
	ASSERT(locked_page);
C
Chris Mason 已提交
1588 1589 1590
	if (index == locked_page->index && index == end_index)
		return 0;

1591 1592 1593 1594 1595
	ret = __process_pages_contig(inode->i_mapping, locked_page, index,
				     end_index, PAGE_LOCK, &index_ret);
	if (ret == -EAGAIN)
		__unlock_for_delalloc(inode, locked_page, delalloc_start,
				      (u64)index_ret << PAGE_SHIFT);
C
Chris Mason 已提交
1596 1597 1598 1599 1600 1601 1602 1603 1604
	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
 */
1605 1606 1607 1608
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
1609 1610 1611 1612
{
	u64 delalloc_start;
	u64 delalloc_end;
	u64 found;
1613
	struct extent_state *cached_state = NULL;
C
Chris Mason 已提交
1614 1615 1616 1617 1618 1619 1620 1621
	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,
1622
				    max_bytes, &cached_state);
C
Chris Mason 已提交
1623
	if (!found || delalloc_end <= *start) {
C
Chris Mason 已提交
1624 1625
		*start = delalloc_start;
		*end = delalloc_end;
1626
		free_extent_state(cached_state);
L
Liu Bo 已提交
1627
		return 0;
C
Chris Mason 已提交
1628 1629
	}

C
Chris Mason 已提交
1630 1631 1632 1633 1634
	/*
	 * 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
1635
	if (delalloc_start < *start)
C
Chris Mason 已提交
1636 1637
		delalloc_start = *start;

C
Chris Mason 已提交
1638 1639 1640
	/*
	 * make sure to limit the number of pages we try to lock down
	 */
1641 1642
	if (delalloc_end + 1 - delalloc_start > max_bytes)
		delalloc_end = delalloc_start + max_bytes - 1;
C
Chris Mason 已提交
1643

C
Chris Mason 已提交
1644 1645 1646 1647 1648 1649 1650
	/* 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
		 */
1651
		free_extent_state(cached_state);
1652
		cached_state = NULL;
C
Chris Mason 已提交
1653
		if (!loops) {
1654
			max_bytes = PAGE_SIZE;
C
Chris Mason 已提交
1655 1656 1657 1658 1659 1660 1661
			loops = 1;
			goto again;
		} else {
			found = 0;
			goto out_failed;
		}
	}
1662
	BUG_ON(ret); /* Only valid values are 0 and -EAGAIN */
C
Chris Mason 已提交
1663 1664

	/* step three, lock the state bits for the whole range */
1665
	lock_extent_bits(tree, delalloc_start, delalloc_end, &cached_state);
C
Chris Mason 已提交
1666 1667 1668

	/* then test to make sure it is all still delalloc */
	ret = test_range_bit(tree, delalloc_start, delalloc_end,
1669
			     EXTENT_DELALLOC, 1, cached_state);
C
Chris Mason 已提交
1670
	if (!ret) {
1671 1672
		unlock_extent_cached(tree, delalloc_start, delalloc_end,
				     &cached_state, GFP_NOFS);
C
Chris Mason 已提交
1673 1674 1675 1676 1677
		__unlock_for_delalloc(inode, locked_page,
			      delalloc_start, delalloc_end);
		cond_resched();
		goto again;
	}
1678
	free_extent_state(cached_state);
C
Chris Mason 已提交
1679 1680 1681 1682 1683 1684
	*start = delalloc_start;
	*end = delalloc_end;
out_failed:
	return found;
}

1685 1686 1687 1688
static int __process_pages_contig(struct address_space *mapping,
				  struct page *locked_page,
				  pgoff_t start_index, pgoff_t end_index,
				  unsigned long page_ops, pgoff_t *index_ret)
C
Chris Mason 已提交
1689
{
1690
	unsigned long nr_pages = end_index - start_index + 1;
1691
	unsigned long pages_locked = 0;
1692
	pgoff_t index = start_index;
C
Chris Mason 已提交
1693
	struct page *pages[16];
1694
	unsigned ret;
1695
	int err = 0;
C
Chris Mason 已提交
1696
	int i;
1697

1698 1699 1700 1701 1702
	if (page_ops & PAGE_LOCK) {
		ASSERT(page_ops == PAGE_LOCK);
		ASSERT(index_ret && *index_ret == start_index);
	}

1703
	if ((page_ops & PAGE_SET_ERROR) && nr_pages > 0)
1704
		mapping_set_error(mapping, -EIO);
1705

C
Chris Mason 已提交
1706
	while (nr_pages > 0) {
1707
		ret = find_get_pages_contig(mapping, index,
1708 1709
				     min_t(unsigned long,
				     nr_pages, ARRAY_SIZE(pages)), pages);
1710 1711 1712 1713 1714 1715 1716 1717
		if (ret == 0) {
			/*
			 * Only if we're going to lock these pages,
			 * can we find nothing at @index.
			 */
			ASSERT(page_ops & PAGE_LOCK);
			return ret;
		}
1718

1719
		for (i = 0; i < ret; i++) {
1720
			if (page_ops & PAGE_SET_PRIVATE2)
1721 1722
				SetPagePrivate2(pages[i]);

C
Chris Mason 已提交
1723
			if (pages[i] == locked_page) {
1724
				put_page(pages[i]);
1725
				pages_locked++;
C
Chris Mason 已提交
1726 1727
				continue;
			}
1728
			if (page_ops & PAGE_CLEAR_DIRTY)
C
Chris Mason 已提交
1729
				clear_page_dirty_for_io(pages[i]);
1730
			if (page_ops & PAGE_SET_WRITEBACK)
C
Chris Mason 已提交
1731
				set_page_writeback(pages[i]);
1732 1733
			if (page_ops & PAGE_SET_ERROR)
				SetPageError(pages[i]);
1734
			if (page_ops & PAGE_END_WRITEBACK)
C
Chris Mason 已提交
1735
				end_page_writeback(pages[i]);
1736
			if (page_ops & PAGE_UNLOCK)
1737
				unlock_page(pages[i]);
1738 1739 1740 1741 1742 1743 1744 1745 1746 1747
			if (page_ops & PAGE_LOCK) {
				lock_page(pages[i]);
				if (!PageDirty(pages[i]) ||
				    pages[i]->mapping != mapping) {
					unlock_page(pages[i]);
					put_page(pages[i]);
					err = -EAGAIN;
					goto out;
				}
			}
1748
			put_page(pages[i]);
1749
			pages_locked++;
C
Chris Mason 已提交
1750 1751 1752 1753 1754
		}
		nr_pages -= ret;
		index += ret;
		cond_resched();
	}
1755 1756 1757 1758
out:
	if (err && index_ret)
		*index_ret = start_index + pages_locked - 1;
	return err;
C
Chris Mason 已提交
1759 1760
}

1761 1762 1763 1764 1765 1766 1767 1768 1769 1770
void extent_clear_unlock_delalloc(struct inode *inode, u64 start, u64 end,
				 u64 delalloc_end, struct page *locked_page,
				 unsigned clear_bits,
				 unsigned long page_ops)
{
	clear_extent_bit(&BTRFS_I(inode)->io_tree, start, end, clear_bits, 1, 0,
			 NULL, GFP_NOFS);

	__process_pages_contig(inode->i_mapping, locked_page,
			       start >> PAGE_SHIFT, end >> PAGE_SHIFT,
1771
			       page_ops, NULL);
1772 1773
}

C
Chris Mason 已提交
1774 1775 1776 1777 1778
/*
 * 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.
 */
1779 1780
u64 count_range_bits(struct extent_io_tree *tree,
		     u64 *start, u64 search_end, u64 max_bytes,
1781
		     unsigned bits, int contig)
1782 1783 1784 1785 1786
{
	struct rb_node *node;
	struct extent_state *state;
	u64 cur_start = *start;
	u64 total_bytes = 0;
1787
	u64 last = 0;
1788 1789
	int found = 0;

1790
	if (WARN_ON(search_end <= cur_start))
1791 1792
		return 0;

1793
	spin_lock(&tree->lock);
1794 1795 1796 1797 1798 1799 1800 1801
	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.
	 */
1802
	node = tree_search(tree, cur_start);
C
Chris Mason 已提交
1803
	if (!node)
1804 1805
		goto out;

C
Chris Mason 已提交
1806
	while (1) {
1807 1808 1809
		state = rb_entry(node, struct extent_state, rb_node);
		if (state->start > search_end)
			break;
1810 1811 1812
		if (contig && found && state->start > last + 1)
			break;
		if (state->end >= cur_start && (state->state & bits) == bits) {
1813 1814 1815 1816 1817
			total_bytes += min(search_end, state->end) + 1 -
				       max(cur_start, state->start);
			if (total_bytes >= max_bytes)
				break;
			if (!found) {
1818
				*start = max(cur_start, state->start);
1819 1820
				found = 1;
			}
1821 1822 1823
			last = state->end;
		} else if (contig && found) {
			break;
1824 1825 1826 1827 1828 1829
		}
		node = rb_next(node);
		if (!node)
			break;
	}
out:
1830
	spin_unlock(&tree->lock);
1831 1832
	return total_bytes;
}
1833

C
Chris Mason 已提交
1834 1835 1836 1837
/*
 * set the private field for a given byte offset in the tree.  If there isn't
 * an extent_state there already, this does nothing.
 */
1838
static noinline int set_state_failrec(struct extent_io_tree *tree, u64 start,
1839
		struct io_failure_record *failrec)
1840 1841 1842 1843 1844
{
	struct rb_node *node;
	struct extent_state *state;
	int ret = 0;

1845
	spin_lock(&tree->lock);
1846 1847 1848 1849
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1850
	node = tree_search(tree, start);
1851
	if (!node) {
1852 1853 1854 1855 1856 1857 1858 1859
		ret = -ENOENT;
		goto out;
	}
	state = rb_entry(node, struct extent_state, rb_node);
	if (state->start != start) {
		ret = -ENOENT;
		goto out;
	}
1860
	state->failrec = failrec;
1861
out:
1862
	spin_unlock(&tree->lock);
1863 1864 1865
	return ret;
}

1866
static noinline int get_state_failrec(struct extent_io_tree *tree, u64 start,
1867
		struct io_failure_record **failrec)
1868 1869 1870 1871 1872
{
	struct rb_node *node;
	struct extent_state *state;
	int ret = 0;

1873
	spin_lock(&tree->lock);
1874 1875 1876 1877
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1878
	node = tree_search(tree, start);
1879
	if (!node) {
1880 1881 1882 1883 1884 1885 1886 1887
		ret = -ENOENT;
		goto out;
	}
	state = rb_entry(node, struct extent_state, rb_node);
	if (state->start != start) {
		ret = -ENOENT;
		goto out;
	}
1888
	*failrec = state->failrec;
1889
out:
1890
	spin_unlock(&tree->lock);
1891 1892 1893 1894 1895
	return ret;
}

/*
 * searches a range in the state tree for a given mask.
1896
 * If 'filled' == 1, this returns 1 only if every extent in the tree
1897 1898 1899 1900
 * 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,
1901
		   unsigned bits, int filled, struct extent_state *cached)
1902 1903 1904 1905 1906
{
	struct extent_state *state = NULL;
	struct rb_node *node;
	int bitset = 0;

1907
	spin_lock(&tree->lock);
1908
	if (cached && extent_state_in_tree(cached) && cached->start <= start &&
1909
	    cached->end > start)
1910 1911 1912
		node = &cached->rb_node;
	else
		node = tree_search(tree, start);
1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931
	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;
		}
1932 1933 1934 1935

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

1936 1937 1938 1939 1940 1941 1942 1943 1944 1945
		start = state->end + 1;
		if (start > end)
			break;
		node = rb_next(node);
		if (!node) {
			if (filled)
				bitset = 0;
			break;
		}
	}
1946
	spin_unlock(&tree->lock);
1947 1948 1949 1950 1951 1952 1953
	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
 */
1954
static void check_page_uptodate(struct extent_io_tree *tree, struct page *page)
1955
{
M
Miao Xie 已提交
1956
	u64 start = page_offset(page);
1957
	u64 end = start + PAGE_SIZE - 1;
1958
	if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1, NULL))
1959 1960 1961
		SetPageUptodate(page);
}

1962
int free_io_failure(struct btrfs_inode *inode, struct io_failure_record *rec)
1963 1964 1965
{
	int ret;
	int err = 0;
1966
	struct extent_io_tree *failure_tree = &inode->io_failure_tree;
1967

1968
	set_state_failrec(failure_tree, rec->start, NULL);
1969 1970
	ret = clear_extent_bits(failure_tree, rec->start,
				rec->start + rec->len - 1,
1971
				EXTENT_LOCKED | EXTENT_DIRTY);
1972 1973 1974
	if (ret)
		err = ret;

1975
	ret = clear_extent_bits(&inode->io_tree, rec->start,
D
David Woodhouse 已提交
1976
				rec->start + rec->len - 1,
1977
				EXTENT_DAMAGED);
D
David Woodhouse 已提交
1978 1979
	if (ret && !err)
		err = ret;
1980 1981 1982 1983 1984 1985 1986 1987 1988 1989

	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.
1990
 * to avoid any synchronization issues, wait for the data after writing, which
1991 1992 1993 1994
 * 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.
 */
1995 1996
int repair_io_failure(struct inode *inode, u64 start, u64 length, u64 logical,
		      struct page *page, unsigned int pg_offset, int mirror_num)
1997
{
1998
	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
1999 2000 2001 2002 2003
	struct bio *bio;
	struct btrfs_device *dev;
	u64 map_length = 0;
	u64 sector;
	struct btrfs_bio *bbio = NULL;
D
David Woodhouse 已提交
2004
	struct btrfs_mapping_tree *map_tree = &fs_info->mapping_tree;
2005 2006
	int ret;

2007
	ASSERT(!(fs_info->sb->s_flags & MS_RDONLY));
2008 2009
	BUG_ON(!mirror_num);

D
David Woodhouse 已提交
2010 2011 2012 2013
	/* we can't repair anything in raid56 yet */
	if (btrfs_is_parity_mirror(map_tree, logical, length, mirror_num))
		return 0;

2014
	bio = btrfs_io_bio_alloc(GFP_NOFS, 1);
2015 2016
	if (!bio)
		return -EIO;
2017
	bio->bi_iter.bi_size = 0;
2018 2019
	map_length = length;

2020 2021 2022 2023 2024 2025
	/*
	 * Avoid races with device replace and make sure our bbio has devices
	 * associated to its stripes that don't go away while we are doing the
	 * read repair operation.
	 */
	btrfs_bio_counter_inc_blocked(fs_info);
2026
	ret = btrfs_map_block(fs_info, BTRFS_MAP_WRITE, logical,
2027 2028
			      &map_length, &bbio, mirror_num);
	if (ret) {
2029
		btrfs_bio_counter_dec(fs_info);
2030 2031 2032 2033 2034
		bio_put(bio);
		return -EIO;
	}
	BUG_ON(mirror_num != bbio->mirror_num);
	sector = bbio->stripes[mirror_num-1].physical >> 9;
2035
	bio->bi_iter.bi_sector = sector;
2036
	dev = bbio->stripes[mirror_num-1].dev;
2037
	btrfs_put_bbio(bbio);
2038
	if (!dev || !dev->bdev || !dev->writeable) {
2039
		btrfs_bio_counter_dec(fs_info);
2040 2041 2042 2043
		bio_put(bio);
		return -EIO;
	}
	bio->bi_bdev = dev->bdev;
2044
	bio->bi_opf = REQ_OP_WRITE | REQ_SYNC;
2045
	bio_add_page(bio, page, length, pg_offset);
2046

2047
	if (btrfsic_submit_bio_wait(bio)) {
2048
		/* try to remap that extent elsewhere? */
2049
		btrfs_bio_counter_dec(fs_info);
2050
		bio_put(bio);
2051
		btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_WRITE_ERRS);
2052 2053 2054
		return -EIO;
	}

2055 2056
	btrfs_info_rl_in_rcu(fs_info,
		"read error corrected: ino %llu off %llu (dev %s sector %llu)",
2057
				  btrfs_ino(BTRFS_I(inode)), start,
2058
				  rcu_str_deref(dev->name), sector);
2059
	btrfs_bio_counter_dec(fs_info);
2060 2061 2062 2063
	bio_put(bio);
	return 0;
}

2064 2065
int repair_eb_io_failure(struct btrfs_fs_info *fs_info,
			 struct extent_buffer *eb, int mirror_num)
2066 2067 2068
{
	u64 start = eb->start;
	unsigned long i, num_pages = num_extent_pages(eb->start, eb->len);
2069
	int ret = 0;
2070

2071
	if (fs_info->sb->s_flags & MS_RDONLY)
2072 2073
		return -EROFS;

2074
	for (i = 0; i < num_pages; i++) {
2075
		struct page *p = eb->pages[i];
2076

2077
		ret = repair_io_failure(fs_info->btree_inode, start,
2078
					PAGE_SIZE, start, p,
2079
					start - page_offset(p), mirror_num);
2080 2081
		if (ret)
			break;
2082
		start += PAGE_SIZE;
2083 2084 2085 2086 2087
	}

	return ret;
}

2088 2089 2090 2091
/*
 * 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
 */
2092 2093
int clean_io_failure(struct inode *inode, u64 start, struct page *page,
		     unsigned int pg_offset)
2094 2095 2096
{
	u64 private;
	struct io_failure_record *failrec;
2097
	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2098 2099 2100 2101 2102 2103 2104 2105 2106 2107
	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;

2108 2109
	ret = get_state_failrec(&BTRFS_I(inode)->io_failure_tree, start,
			&failrec);
2110 2111 2112 2113 2114 2115 2116
	if (ret)
		return 0;

	BUG_ON(!failrec->this_mirror);

	if (failrec->in_validation) {
		/* there was no real error, just free the record */
2117 2118 2119
		btrfs_debug(fs_info,
			"clean_io_failure: freeing dummy error at %llu",
			failrec->start);
2120 2121
		goto out;
	}
2122 2123
	if (fs_info->sb->s_flags & MS_RDONLY)
		goto out;
2124 2125 2126 2127 2128 2129 2130

	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);

2131 2132
	if (state && state->start <= failrec->start &&
	    state->end >= failrec->start + failrec->len - 1) {
2133 2134
		num_copies = btrfs_num_copies(fs_info, failrec->logical,
					      failrec->len);
2135
		if (num_copies > 1)  {
2136
			repair_io_failure(inode, start, failrec->len,
2137
					  failrec->logical, page,
2138
					  pg_offset, failrec->failed_mirror);
2139 2140 2141 2142
		}
	}

out:
2143
	free_io_failure(BTRFS_I(inode), failrec);
2144

2145
	return 0;
2146 2147
}

2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172
/*
 * 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);

2173
		failrec = state->failrec;
2174 2175 2176 2177 2178 2179 2180 2181
		free_extent_state(state);
		kfree(failrec);

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

2182
int btrfs_get_io_failure_record(struct inode *inode, u64 start, u64 end,
2183
		struct io_failure_record **failrec_ret)
2184
{
2185
	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2186
	struct io_failure_record *failrec;
2187 2188 2189 2190 2191 2192 2193
	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;

2194
	ret = get_state_failrec(failure_tree, start, &failrec);
2195 2196 2197 2198
	if (ret) {
		failrec = kzalloc(sizeof(*failrec), GFP_NOFS);
		if (!failrec)
			return -ENOMEM;
2199

2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213
		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;
		}

2214
		if (em->start > start || em->start + em->len <= start) {
2215 2216 2217 2218
			free_extent_map(em);
			em = NULL;
		}
		read_unlock(&em_tree->lock);
2219
		if (!em) {
2220 2221 2222
			kfree(failrec);
			return -EIO;
		}
2223

2224 2225 2226 2227 2228 2229 2230 2231
		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);
		}
2232

2233 2234 2235
		btrfs_debug(fs_info,
			"Get IO Failure Record: (new) logical=%llu, start=%llu, len=%llu",
			logical, start, failrec->len);
2236

2237 2238 2239 2240 2241
		failrec->logical = logical;
		free_extent_map(em);

		/* set the bits in the private failure tree */
		ret = set_extent_bits(failure_tree, start, end,
2242
					EXTENT_LOCKED | EXTENT_DIRTY);
2243
		if (ret >= 0)
2244
			ret = set_state_failrec(failure_tree, start, failrec);
2245 2246
		/* set the bits in the inode's tree */
		if (ret >= 0)
2247
			ret = set_extent_bits(tree, start, end, EXTENT_DAMAGED);
2248 2249 2250 2251 2252
		if (ret < 0) {
			kfree(failrec);
			return ret;
		}
	} else {
2253 2254 2255 2256
		btrfs_debug(fs_info,
			"Get IO Failure Record: (found) logical=%llu, start=%llu, len=%llu, validation=%d",
			failrec->logical, failrec->start, failrec->len,
			failrec->in_validation);
2257 2258 2259 2260 2261 2262
		/*
		 * 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.
		 */
	}
2263 2264 2265 2266 2267 2268 2269 2270 2271

	*failrec_ret = failrec;

	return 0;
}

int btrfs_check_repairable(struct inode *inode, struct bio *failed_bio,
			   struct io_failure_record *failrec, int failed_mirror)
{
2272
	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2273 2274
	int num_copies;

2275
	num_copies = btrfs_num_copies(fs_info, failrec->logical, failrec->len);
2276 2277 2278 2279 2280 2281
	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.
		 */
2282 2283 2284
		btrfs_debug(fs_info,
			"Check Repairable: cannot repair, num_copies=%d, next_mirror %d, failed_mirror %d",
			num_copies, failrec->this_mirror, failed_mirror);
2285
		return 0;
2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321
	}

	/*
	 * 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++;
	}

2322
	if (failrec->this_mirror > num_copies) {
2323 2324 2325
		btrfs_debug(fs_info,
			"Check Repairable: (fail) num_copies=%d, next_mirror %d, failed_mirror %d",
			num_copies, failrec->this_mirror, failed_mirror);
2326
		return 0;
2327 2328
	}

2329 2330 2331 2332 2333 2334 2335
	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,
2336
				    bio_end_io_t *endio_func, void *data)
2337
{
2338
	struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2339 2340 2341 2342
	struct bio *bio;
	struct btrfs_io_bio *btrfs_failed_bio;
	struct btrfs_io_bio *btrfs_bio;

2343
	bio = btrfs_io_bio_alloc(GFP_NOFS, 1);
2344 2345 2346 2347
	if (!bio)
		return NULL;

	bio->bi_end_io = endio_func;
2348
	bio->bi_iter.bi_sector = failrec->logical >> 9;
2349
	bio->bi_bdev = fs_info->fs_devices->latest_bdev;
2350
	bio->bi_iter.bi_size = 0;
2351
	bio->bi_private = data;
2352

2353 2354 2355 2356 2357 2358
	btrfs_failed_bio = btrfs_io_bio(failed_bio);
	if (btrfs_failed_bio->csum) {
		u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);

		btrfs_bio = btrfs_io_bio(bio);
		btrfs_bio->csum = btrfs_bio->csum_inline;
2359 2360
		icsum *= csum_size;
		memcpy(btrfs_bio->csum, btrfs_failed_bio->csum + icsum,
2361 2362 2363
		       csum_size);
	}

2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384
	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;
2385
	int read_mode = 0;
2386 2387
	int ret;

2388
	BUG_ON(bio_op(failed_bio) == REQ_OP_WRITE);
2389 2390 2391 2392 2393 2394 2395

	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) {
2396
		free_io_failure(BTRFS_I(inode), failrec);
2397 2398 2399 2400
		return -EIO;
	}

	if (failed_bio->bi_vcnt > 1)
2401
		read_mode |= REQ_FAILFAST_DEV;
2402 2403 2404 2405

	phy_offset >>= inode->i_sb->s_blocksize_bits;
	bio = btrfs_create_repair_bio(inode, failed_bio, failrec, page,
				      start - page_offset(page),
2406 2407
				      (int)phy_offset, failed_bio->bi_end_io,
				      NULL);
2408
	if (!bio) {
2409
		free_io_failure(BTRFS_I(inode), failrec);
2410 2411
		return -EIO;
	}
2412
	bio_set_op_attrs(bio, REQ_OP_READ, read_mode);
2413

2414 2415 2416
	btrfs_debug(btrfs_sb(inode->i_sb),
		"Repair Read Error: submitting new read[%#x] to this_mirror=%d, in_validation=%d",
		read_mode, failrec->this_mirror, failrec->in_validation);
2417

2418
	ret = tree->ops->submit_bio_hook(inode, bio, failrec->this_mirror,
2419
					 failrec->bio_flags, 0);
2420
	if (ret) {
2421
		free_io_failure(BTRFS_I(inode), failrec);
2422 2423 2424
		bio_put(bio);
	}

2425
	return ret;
2426 2427
}

2428 2429
/* lots and lots of room for performance fixes in the end_bio funcs */

2430
void end_extent_writepage(struct page *page, int err, u64 start, u64 end)
2431 2432 2433
{
	int uptodate = (err == 0);
	struct extent_io_tree *tree;
2434
	int ret = 0;
2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447

	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);
2448 2449
		ret = ret < 0 ? ret : -EIO;
		mapping_set_error(page->mapping, ret);
2450 2451 2452
	}
}

2453 2454 2455 2456 2457 2458 2459 2460 2461
/*
 * 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.
 */
2462
static void end_bio_extent_writepage(struct bio *bio)
2463
{
2464
	struct bio_vec *bvec;
2465 2466
	u64 start;
	u64 end;
2467
	int i;
2468

2469
	bio_for_each_segment_all(bvec, bio, i) {
2470
		struct page *page = bvec->bv_page;
2471 2472
		struct inode *inode = page->mapping->host;
		struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2473

2474 2475 2476 2477 2478
		/* 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.  */
2479 2480
		if (bvec->bv_offset || bvec->bv_len != PAGE_SIZE) {
			if (bvec->bv_offset + bvec->bv_len != PAGE_SIZE)
2481
				btrfs_err(fs_info,
2482 2483 2484
				   "partial page write in btrfs with offset %u and length %u",
					bvec->bv_offset, bvec->bv_len);
			else
2485
				btrfs_info(fs_info,
J
Jeff Mahoney 已提交
2486
				   "incomplete page write in btrfs with offset %u and length %u",
2487 2488
					bvec->bv_offset, bvec->bv_len);
		}
2489

2490 2491
		start = page_offset(page);
		end = start + bvec->bv_offset + bvec->bv_len - 1;
2492

2493
		end_extent_writepage(page, bio->bi_error, start, end);
2494
		end_page_writeback(page);
2495
	}
2496

2497 2498 2499
	bio_put(bio);
}

2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511
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);
}

2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522
/*
 * 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.
 */
2523
static void end_bio_extent_readpage(struct bio *bio)
2524
{
2525
	struct bio_vec *bvec;
2526
	int uptodate = !bio->bi_error;
2527
	struct btrfs_io_bio *io_bio = btrfs_io_bio(bio);
2528
	struct extent_io_tree *tree;
2529
	u64 offset = 0;
2530 2531
	u64 start;
	u64 end;
2532
	u64 len;
2533 2534
	u64 extent_start = 0;
	u64 extent_len = 0;
2535
	int mirror;
2536
	int ret;
2537
	int i;
2538

2539
	bio_for_each_segment_all(bvec, bio, i) {
2540
		struct page *page = bvec->bv_page;
2541
		struct inode *inode = page->mapping->host;
2542
		struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2543

2544 2545 2546 2547
		btrfs_debug(fs_info,
			"end_bio_extent_readpage: bi_sector=%llu, err=%d, mirror=%u",
			(u64)bio->bi_iter.bi_sector, bio->bi_error,
			io_bio->mirror_num);
2548
		tree = &BTRFS_I(inode)->io_tree;
2549

2550 2551 2552 2553 2554
		/* 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.  */
2555 2556
		if (bvec->bv_offset || bvec->bv_len != PAGE_SIZE) {
			if (bvec->bv_offset + bvec->bv_len != PAGE_SIZE)
2557 2558
				btrfs_err(fs_info,
					"partial page read in btrfs with offset %u and length %u",
2559 2560
					bvec->bv_offset, bvec->bv_len);
			else
2561 2562
				btrfs_info(fs_info,
					"incomplete page read in btrfs with offset %u and length %u",
2563 2564
					bvec->bv_offset, bvec->bv_len);
		}
2565

2566 2567
		start = page_offset(page);
		end = start + bvec->bv_offset + bvec->bv_len - 1;
2568
		len = bvec->bv_len;
2569

2570
		mirror = io_bio->mirror_num;
2571 2572
		if (likely(uptodate && tree->ops &&
			   tree->ops->readpage_end_io_hook)) {
2573 2574 2575
			ret = tree->ops->readpage_end_io_hook(io_bio, offset,
							      page, start, end,
							      mirror);
2576
			if (ret)
2577
				uptodate = 0;
2578
			else
2579
				clean_io_failure(inode, start, page, 0);
2580
		}
2581

2582 2583 2584 2585
		if (likely(uptodate))
			goto readpage_ok;

		if (tree->ops && tree->ops->readpage_io_failed_hook) {
2586
			ret = tree->ops->readpage_io_failed_hook(page, mirror);
2587
			if (!ret && !bio->bi_error)
2588
				uptodate = 1;
2589
		} else {
2590 2591 2592 2593 2594 2595 2596 2597 2598 2599
			/*
			 * 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.
			 */
2600 2601
			ret = bio_readpage_error(bio, offset, page, start, end,
						 mirror);
2602
			if (ret == 0) {
2603
				uptodate = !bio->bi_error;
2604
				offset += len;
2605 2606 2607
				continue;
			}
		}
2608
readpage_ok:
2609
		if (likely(uptodate)) {
2610
			loff_t i_size = i_size_read(inode);
2611
			pgoff_t end_index = i_size >> PAGE_SHIFT;
2612
			unsigned off;
2613 2614

			/* Zero out the end if this page straddles i_size */
2615
			off = i_size & (PAGE_SIZE-1);
2616
			if (page->index == end_index && off)
2617
				zero_user_segment(page, off, PAGE_SIZE);
2618
			SetPageUptodate(page);
2619
		} else {
2620 2621
			ClearPageUptodate(page);
			SetPageError(page);
2622
		}
2623
		unlock_page(page);
2624
		offset += len;
2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645 2646

		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;
		}
2647
	}
2648

2649 2650 2651
	if (extent_len)
		endio_readpage_release_extent(tree, extent_start, extent_len,
					      uptodate);
2652
	if (io_bio->end_io)
2653
		io_bio->end_io(io_bio, bio->bi_error);
2654 2655 2656
	bio_put(bio);
}

2657 2658 2659 2660
/*
 * 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
 */
2661 2662 2663
struct bio *
btrfs_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
		gfp_t gfp_flags)
2664
{
2665
	struct btrfs_io_bio *btrfs_bio;
2666 2667
	struct bio *bio;

2668
	bio = bio_alloc_bioset(gfp_flags, nr_vecs, btrfs_bioset);
2669 2670

	if (bio == NULL && (current->flags & PF_MEMALLOC)) {
2671 2672 2673 2674
		while (!bio && (nr_vecs /= 2)) {
			bio = bio_alloc_bioset(gfp_flags,
					       nr_vecs, btrfs_bioset);
		}
2675 2676 2677 2678
	}

	if (bio) {
		bio->bi_bdev = bdev;
2679
		bio->bi_iter.bi_sector = first_sector;
2680 2681 2682 2683
		btrfs_bio = btrfs_io_bio(bio);
		btrfs_bio->csum = NULL;
		btrfs_bio->csum_allocated = NULL;
		btrfs_bio->end_io = NULL;
2684 2685 2686 2687
	}
	return bio;
}

2688 2689
struct bio *btrfs_bio_clone(struct bio *bio, gfp_t gfp_mask)
{
2690 2691
	struct btrfs_io_bio *btrfs_bio;
	struct bio *new;
2692

2693 2694 2695 2696 2697 2698 2699 2700 2701
	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;
	}
	return new;
}
2702 2703 2704 2705

/* this also allocates from the btrfs_bioset */
struct bio *btrfs_io_bio_alloc(gfp_t gfp_mask, unsigned int nr_iovecs)
{
2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716
	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;
2717 2718 2719
}


2720 2721
static int __must_check submit_one_bio(struct bio *bio, int mirror_num,
				       unsigned long bio_flags)
2722 2723
{
	int ret = 0;
2724 2725 2726 2727 2728
	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
2729
	start = page_offset(page) + bvec->bv_offset;
2730

2731
	bio->bi_private = NULL;
2732 2733
	bio_get(bio);

2734
	if (tree->ops && tree->ops->submit_bio_hook)
2735
		ret = tree->ops->submit_bio_hook(page->mapping->host, bio,
2736
					   mirror_num, bio_flags, start);
2737
	else
2738
		btrfsic_submit_bio(bio);
2739

2740 2741 2742 2743
	bio_put(bio);
	return ret;
}

2744
static int merge_bio(struct extent_io_tree *tree, struct page *page,
2745 2746 2747 2748 2749
		     unsigned long offset, size_t size, struct bio *bio,
		     unsigned long bio_flags)
{
	int ret = 0;
	if (tree->ops && tree->ops->merge_bio_hook)
2750
		ret = tree->ops->merge_bio_hook(page, offset, size, bio,
2751 2752 2753 2754 2755
						bio_flags);
	return ret;

}

2756
static int submit_extent_page(int op, int op_flags, struct extent_io_tree *tree,
2757
			      struct writeback_control *wbc,
2758 2759 2760 2761
			      struct page *page, sector_t sector,
			      size_t size, unsigned long offset,
			      struct block_device *bdev,
			      struct bio **bio_ret,
2762
			      bio_end_io_t end_io_func,
C
Chris Mason 已提交
2763 2764
			      int mirror_num,
			      unsigned long prev_bio_flags,
2765 2766
			      unsigned long bio_flags,
			      bool force_bio_submit)
2767 2768 2769
{
	int ret = 0;
	struct bio *bio;
C
Chris Mason 已提交
2770 2771
	int contig = 0;
	int old_compressed = prev_bio_flags & EXTENT_BIO_COMPRESSED;
2772
	size_t page_size = min_t(size_t, size, PAGE_SIZE);
2773 2774 2775

	if (bio_ret && *bio_ret) {
		bio = *bio_ret;
C
Chris Mason 已提交
2776
		if (old_compressed)
2777
			contig = bio->bi_iter.bi_sector == sector;
C
Chris Mason 已提交
2778
		else
K
Kent Overstreet 已提交
2779
			contig = bio_end_sector(bio) == sector;
C
Chris Mason 已提交
2780 2781

		if (prev_bio_flags != bio_flags || !contig ||
2782
		    force_bio_submit ||
2783
		    merge_bio(tree, page, offset, page_size, bio, bio_flags) ||
C
Chris Mason 已提交
2784
		    bio_add_page(bio, page, page_size, offset) < page_size) {
2785
			ret = submit_one_bio(bio, mirror_num, prev_bio_flags);
2786 2787
			if (ret < 0) {
				*bio_ret = NULL;
2788
				return ret;
2789
			}
2790 2791
			bio = NULL;
		} else {
2792 2793
			if (wbc)
				wbc_account_io(wbc, page, page_size);
2794 2795 2796
			return 0;
		}
	}
C
Chris Mason 已提交
2797

2798 2799
	bio = btrfs_bio_alloc(bdev, sector, BIO_MAX_PAGES,
			GFP_NOFS | __GFP_HIGH);
2800 2801
	if (!bio)
		return -ENOMEM;
2802

C
Chris Mason 已提交
2803
	bio_add_page(bio, page, page_size, offset);
2804 2805
	bio->bi_end_io = end_io_func;
	bio->bi_private = tree;
2806
	bio_set_op_attrs(bio, op, op_flags);
2807 2808 2809 2810
	if (wbc) {
		wbc_init_bio(wbc, bio);
		wbc_account_io(wbc, page, page_size);
	}
2811

C
Chris Mason 已提交
2812
	if (bio_ret)
2813
		*bio_ret = bio;
C
Chris Mason 已提交
2814
	else
2815
		ret = submit_one_bio(bio, mirror_num, bio_flags);
2816 2817 2818 2819

	return ret;
}

2820 2821
static void attach_extent_buffer_page(struct extent_buffer *eb,
				      struct page *page)
2822 2823 2824
{
	if (!PagePrivate(page)) {
		SetPagePrivate(page);
2825
		get_page(page);
J
Josef Bacik 已提交
2826 2827 2828
		set_page_private(page, (unsigned long)eb);
	} else {
		WARN_ON(page->private != (unsigned long)eb);
2829 2830 2831
	}
}

J
Josef Bacik 已提交
2832
void set_page_extent_mapped(struct page *page)
2833
{
J
Josef Bacik 已提交
2834 2835
	if (!PagePrivate(page)) {
		SetPagePrivate(page);
2836
		get_page(page);
J
Josef Bacik 已提交
2837 2838
		set_page_private(page, EXTENT_PAGE_PRIVATE);
	}
2839 2840
}

2841 2842 2843 2844 2845 2846 2847 2848 2849
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;
2850
		if (extent_map_in_tree(em) && start >= em->start &&
2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867
		    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;
}
2868 2869 2870 2871
/*
 * 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)
2872
 * XXX JDM: This needs looking at to ensure proper page locking
2873
 * return 0 on success, otherwise return error
2874
 */
2875 2876 2877
static int __do_readpage(struct extent_io_tree *tree,
			 struct page *page,
			 get_extent_t *get_extent,
2878
			 struct extent_map **em_cached,
2879
			 struct bio **bio, int mirror_num,
2880
			 unsigned long *bio_flags, int read_flags,
2881
			 u64 *prev_em_start)
2882 2883
{
	struct inode *inode = page->mapping->host;
M
Miao Xie 已提交
2884
	u64 start = page_offset(page);
2885
	u64 page_end = start + PAGE_SIZE - 1;
2886 2887 2888 2889 2890 2891 2892 2893 2894
	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;
2895
	int ret = 0;
2896
	int nr = 0;
2897
	size_t pg_offset = 0;
2898
	size_t iosize;
C
Chris Mason 已提交
2899
	size_t disk_io_size;
2900
	size_t blocksize = inode->i_sb->s_blocksize;
2901
	unsigned long this_bio_flag = 0;
2902 2903 2904

	set_page_extent_mapped(page);

2905
	end = page_end;
D
Dan Magenheimer 已提交
2906 2907 2908
	if (!PageUptodate(page)) {
		if (cleancache_get_page(page) == 0) {
			BUG_ON(blocksize != PAGE_SIZE);
2909
			unlock_extent(tree, start, end);
D
Dan Magenheimer 已提交
2910 2911 2912 2913
			goto out;
		}
	}

2914
	if (page->index == last_byte >> PAGE_SHIFT) {
C
Chris Mason 已提交
2915
		char *userpage;
2916
		size_t zero_offset = last_byte & (PAGE_SIZE - 1);
C
Chris Mason 已提交
2917 2918

		if (zero_offset) {
2919
			iosize = PAGE_SIZE - zero_offset;
2920
			userpage = kmap_atomic(page);
C
Chris Mason 已提交
2921 2922
			memset(userpage + zero_offset, 0, iosize);
			flush_dcache_page(page);
2923
			kunmap_atomic(userpage);
C
Chris Mason 已提交
2924 2925
		}
	}
2926
	while (cur <= end) {
2927
		bool force_bio_submit = false;
2928

2929 2930
		if (cur >= last_byte) {
			char *userpage;
2931 2932
			struct extent_state *cached = NULL;

2933
			iosize = PAGE_SIZE - pg_offset;
2934
			userpage = kmap_atomic(page);
2935
			memset(userpage + pg_offset, 0, iosize);
2936
			flush_dcache_page(page);
2937
			kunmap_atomic(userpage);
2938
			set_extent_uptodate(tree, cur, cur + iosize - 1,
2939
					    &cached, GFP_NOFS);
2940 2941 2942
			unlock_extent_cached(tree, cur,
					     cur + iosize - 1,
					     &cached, GFP_NOFS);
2943 2944
			break;
		}
2945 2946
		em = __get_extent_map(inode, page, pg_offset, cur,
				      end - cur + 1, get_extent, em_cached);
2947
		if (IS_ERR_OR_NULL(em)) {
2948
			SetPageError(page);
2949
			unlock_extent(tree, cur, end);
2950 2951 2952 2953 2954 2955
			break;
		}
		extent_offset = cur - em->start;
		BUG_ON(extent_map_end(em) <= cur);
		BUG_ON(end < cur);

2956
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
2957
			this_bio_flag |= EXTENT_BIO_COMPRESSED;
2958 2959 2960
			extent_set_compress_type(&this_bio_flag,
						 em->compress_type);
		}
C
Chris Mason 已提交
2961

2962 2963
		iosize = min(extent_map_end(em) - cur, end - cur + 1);
		cur_end = min(extent_map_end(em) - 1, end);
2964
		iosize = ALIGN(iosize, blocksize);
C
Chris Mason 已提交
2965 2966 2967 2968 2969 2970 2971
		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;
		}
2972 2973
		bdev = em->bdev;
		block_start = em->block_start;
Y
Yan Zheng 已提交
2974 2975
		if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			block_start = EXTENT_MAP_HOLE;
2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018

		/*
		 * 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;

3019 3020 3021 3022 3023 3024
		free_extent_map(em);
		em = NULL;

		/* we've found a hole, just zero and go on */
		if (block_start == EXTENT_MAP_HOLE) {
			char *userpage;
3025 3026
			struct extent_state *cached = NULL;

3027
			userpage = kmap_atomic(page);
3028
			memset(userpage + pg_offset, 0, iosize);
3029
			flush_dcache_page(page);
3030
			kunmap_atomic(userpage);
3031 3032

			set_extent_uptodate(tree, cur, cur + iosize - 1,
3033
					    &cached, GFP_NOFS);
3034 3035 3036
			unlock_extent_cached(tree, cur,
					     cur + iosize - 1,
					     &cached, GFP_NOFS);
3037
			cur = cur + iosize;
3038
			pg_offset += iosize;
3039 3040 3041
			continue;
		}
		/* the get_extent function already copied into the page */
3042 3043
		if (test_range_bit(tree, cur, cur_end,
				   EXTENT_UPTODATE, 1, NULL)) {
3044
			check_page_uptodate(tree, page);
3045
			unlock_extent(tree, cur, cur + iosize - 1);
3046
			cur = cur + iosize;
3047
			pg_offset += iosize;
3048 3049
			continue;
		}
3050 3051 3052 3053 3054
		/* we have an inline extent but it didn't get marked up
		 * to date.  Error out
		 */
		if (block_start == EXTENT_MAP_INLINE) {
			SetPageError(page);
3055
			unlock_extent(tree, cur, cur + iosize - 1);
3056
			cur = cur + iosize;
3057
			pg_offset += iosize;
3058 3059
			continue;
		}
3060

3061 3062
		ret = submit_extent_page(REQ_OP_READ, read_flags, tree, NULL,
					 page, sector, disk_io_size, pg_offset,
3063
					 bdev, bio,
C
Chris Mason 已提交
3064 3065
					 end_bio_extent_readpage, mirror_num,
					 *bio_flags,
3066 3067
					 this_bio_flag,
					 force_bio_submit);
3068 3069 3070 3071
		if (!ret) {
			nr++;
			*bio_flags = this_bio_flag;
		} else {
3072
			SetPageError(page);
3073
			unlock_extent(tree, cur, cur + iosize - 1);
3074
			goto out;
3075
		}
3076
		cur = cur + iosize;
3077
		pg_offset += iosize;
3078
	}
D
Dan Magenheimer 已提交
3079
out:
3080 3081 3082 3083 3084
	if (!nr) {
		if (!PageError(page))
			SetPageUptodate(page);
		unlock_page(page);
	}
3085
	return ret;
3086 3087
}

3088 3089 3090 3091
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,
3092
					     struct extent_map **em_cached,
3093
					     struct bio **bio, int mirror_num,
3094
					     unsigned long *bio_flags,
3095
					     u64 *prev_em_start)
3096 3097 3098 3099 3100 3101 3102 3103
{
	struct inode *inode;
	struct btrfs_ordered_extent *ordered;
	int index;

	inode = pages[0]->mapping->host;
	while (1) {
		lock_extent(tree, start, end);
3104
		ordered = btrfs_lookup_ordered_range(BTRFS_I(inode), start,
3105 3106 3107 3108 3109 3110 3111 3112 3113
						     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++) {
3114
		__do_readpage(tree, pages[index], get_extent, em_cached, bio,
3115
			      mirror_num, bio_flags, 0, prev_em_start);
3116
		put_page(pages[index]);
3117 3118 3119 3120 3121 3122
	}
}

static void __extent_readpages(struct extent_io_tree *tree,
			       struct page *pages[],
			       int nr_pages, get_extent_t *get_extent,
3123
			       struct extent_map **em_cached,
3124
			       struct bio **bio, int mirror_num,
3125
			       unsigned long *bio_flags,
3126
			       u64 *prev_em_start)
3127
{
3128
	u64 start = 0;
3129 3130 3131
	u64 end = 0;
	u64 page_start;
	int index;
3132
	int first_index = 0;
3133 3134 3135 3136 3137

	for (index = 0; index < nr_pages; index++) {
		page_start = page_offset(pages[index]);
		if (!end) {
			start = page_start;
3138
			end = start + PAGE_SIZE - 1;
3139 3140
			first_index = index;
		} else if (end + 1 == page_start) {
3141
			end += PAGE_SIZE;
3142 3143 3144
		} else {
			__do_contiguous_readpages(tree, &pages[first_index],
						  index - first_index, start,
3145 3146
						  end, get_extent, em_cached,
						  bio, mirror_num, bio_flags,
3147
						  prev_em_start);
3148
			start = page_start;
3149
			end = start + PAGE_SIZE - 1;
3150 3151 3152 3153 3154 3155 3156
			first_index = index;
		}
	}

	if (end)
		__do_contiguous_readpages(tree, &pages[first_index],
					  index - first_index, start,
3157
					  end, get_extent, em_cached, bio,
3158
					  mirror_num, bio_flags,
3159
					  prev_em_start);
3160 3161 3162 3163 3164 3165
}

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,
3166
				   unsigned long *bio_flags, int read_flags)
3167 3168 3169 3170
{
	struct inode *inode = page->mapping->host;
	struct btrfs_ordered_extent *ordered;
	u64 start = page_offset(page);
3171
	u64 end = start + PAGE_SIZE - 1;
3172 3173 3174 3175
	int ret;

	while (1) {
		lock_extent(tree, start, end);
3176
		ordered = btrfs_lookup_ordered_range(BTRFS_I(inode), start,
3177
						PAGE_SIZE);
3178 3179 3180 3181 3182 3183 3184
		if (!ordered)
			break;
		unlock_extent(tree, start, end);
		btrfs_start_ordered_extent(inode, ordered, 1);
		btrfs_put_ordered_extent(ordered);
	}

3185
	ret = __do_readpage(tree, page, get_extent, NULL, bio, mirror_num,
3186
			    bio_flags, read_flags, NULL);
3187 3188 3189
	return ret;
}

3190
int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
3191
			    get_extent_t *get_extent, int mirror_num)
3192 3193
{
	struct bio *bio = NULL;
C
Chris Mason 已提交
3194
	unsigned long bio_flags = 0;
3195 3196
	int ret;

3197
	ret = __extent_read_full_page(tree, page, get_extent, &bio, mirror_num,
3198
				      &bio_flags, 0);
3199
	if (bio)
3200
		ret = submit_one_bio(bio, mirror_num, bio_flags);
3201 3202 3203
	return ret;
}

3204
static void update_nr_written(struct writeback_control *wbc,
3205
			      unsigned long nr_written)
3206 3207 3208 3209
{
	wbc->nr_to_write -= nr_written;
}

3210
/*
3211 3212 3213 3214 3215 3216 3217 3218
 * 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)
3219
 */
3220 3221 3222 3223 3224 3225 3226
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;
3227
	u64 page_end = delalloc_start + PAGE_SIZE - 1;
3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241
	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,
3242
					       BTRFS_MAX_EXTENT_SIZE);
3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263
		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;
		}
		/*
3264 3265
		 * delalloc_end is already one less than the total length, so
		 * we don't subtract one from PAGE_SIZE
3266 3267
		 */
		delalloc_to_write += (delalloc_end - delalloc_start +
3268
				      PAGE_SIZE) >> PAGE_SHIFT;
3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313
		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)
3314 3315
{
	struct extent_io_tree *tree = epd->tree;
M
Miao Xie 已提交
3316
	u64 start = page_offset(page);
3317
	u64 page_end = start + PAGE_SIZE - 1;
3318 3319 3320 3321 3322 3323 3324 3325
	u64 end;
	u64 cur = start;
	u64 extent_offset;
	u64 block_start;
	u64 iosize;
	sector_t sector;
	struct extent_map *em;
	struct block_device *bdev;
3326
	size_t pg_offset = 0;
3327
	size_t blocksize;
3328 3329 3330
	int ret = 0;
	int nr = 0;
	bool compressed;
C
Chris Mason 已提交
3331

3332
	if (tree->ops && tree->ops->writepage_start_hook) {
C
Chris Mason 已提交
3333 3334
		ret = tree->ops->writepage_start_hook(page, start,
						      page_end);
3335 3336 3337 3338 3339 3340
		if (ret) {
			/* Fixup worker will requeue */
			if (ret == -EBUSY)
				wbc->pages_skipped++;
			else
				redirty_page_for_writepage(wbc, page);
3341

3342
			update_nr_written(wbc, nr_written);
3343
			unlock_page(page);
3344
			return 1;
3345 3346 3347
		}
	}

3348 3349 3350 3351
	/*
	 * we don't want to touch the inode after unlocking the page,
	 * so we update the mapping writeback index now
	 */
3352
	update_nr_written(wbc, nr_written + 1);
3353

3354
	end = page_end;
3355
	if (i_size <= start) {
3356 3357 3358
		if (tree->ops && tree->ops->writepage_end_io_hook)
			tree->ops->writepage_end_io_hook(page, start,
							 page_end, NULL, 1);
3359 3360 3361 3362 3363 3364
		goto done;
	}

	blocksize = inode->i_sb->s_blocksize;

	while (cur <= end) {
3365
		u64 em_end;
3366

3367
		if (cur >= i_size) {
3368 3369 3370
			if (tree->ops && tree->ops->writepage_end_io_hook)
				tree->ops->writepage_end_io_hook(page, cur,
							 page_end, NULL, 1);
3371 3372
			break;
		}
3373
		em = epd->get_extent(inode, page, pg_offset, cur,
3374
				     end - cur + 1, 1);
3375
		if (IS_ERR_OR_NULL(em)) {
3376
			SetPageError(page);
3377
			ret = PTR_ERR_OR_ZERO(em);
3378 3379 3380 3381
			break;
		}

		extent_offset = cur - em->start;
3382 3383
		em_end = extent_map_end(em);
		BUG_ON(em_end <= cur);
3384
		BUG_ON(end < cur);
3385
		iosize = min(em_end - cur, end - cur + 1);
3386
		iosize = ALIGN(iosize, blocksize);
3387 3388 3389
		sector = (em->block_start + extent_offset) >> 9;
		bdev = em->bdev;
		block_start = em->block_start;
C
Chris Mason 已提交
3390
		compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
3391 3392 3393
		free_extent_map(em);
		em = NULL;

C
Chris Mason 已提交
3394 3395 3396 3397 3398
		/*
		 * compressed and inline extents are written through other
		 * paths in the FS
		 */
		if (compressed || block_start == EXTENT_MAP_HOLE ||
3399
		    block_start == EXTENT_MAP_INLINE) {
C
Chris Mason 已提交
3400 3401 3402 3403 3404 3405
			/*
			 * end_io notification does not happen here for
			 * compressed extents
			 */
			if (!compressed && tree->ops &&
			    tree->ops->writepage_end_io_hook)
3406 3407 3408
				tree->ops->writepage_end_io_hook(page, cur,
							 cur + iosize - 1,
							 NULL, 1);
C
Chris Mason 已提交
3409 3410 3411 3412 3413 3414 3415 3416 3417
			else if (compressed) {
				/* we don't want to end_page_writeback on
				 * a compressed extent.  this happens
				 * elsewhere
				 */
				nr++;
			}

			cur += iosize;
3418
			pg_offset += iosize;
3419 3420
			continue;
		}
C
Chris Mason 已提交
3421

3422 3423 3424 3425 3426
		set_range_writeback(tree, cur, cur + iosize - 1);
		if (!PageWriteback(page)) {
			btrfs_err(BTRFS_I(inode)->root->fs_info,
				   "page %lu not writeback, cur %llu end %llu",
			       page->index, cur, end);
3427
		}
3428

3429 3430
		ret = submit_extent_page(REQ_OP_WRITE, write_flags, tree, wbc,
					 page, sector, iosize, pg_offset,
3431
					 bdev, &epd->bio,
3432 3433
					 end_bio_extent_writepage,
					 0, 0, 0, false);
3434
		if (ret) {
3435
			SetPageError(page);
3436 3437 3438
			if (PageWriteback(page))
				end_page_writeback(page);
		}
3439 3440

		cur = cur + iosize;
3441
		pg_offset += iosize;
3442 3443
		nr++;
	}
3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460
done:
	*nr_ret = nr;
	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);
3461
	u64 page_end = start + PAGE_SIZE - 1;
3462 3463 3464 3465
	int ret;
	int nr = 0;
	size_t pg_offset = 0;
	loff_t i_size = i_size_read(inode);
3466
	unsigned long end_index = i_size >> PAGE_SHIFT;
3467
	int write_flags = 0;
3468 3469 3470
	unsigned long nr_written = 0;

	if (wbc->sync_mode == WB_SYNC_ALL)
3471
		write_flags = REQ_SYNC;
3472 3473 3474 3475 3476 3477 3478

	trace___extent_writepage(page, inode, wbc);

	WARN_ON(!PageLocked(page));

	ClearPageError(page);

3479
	pg_offset = i_size & (PAGE_SIZE - 1);
3480 3481
	if (page->index > end_index ||
	   (page->index == end_index && !pg_offset)) {
3482
		page->mapping->a_ops->invalidatepage(page, 0, PAGE_SIZE);
3483 3484 3485 3486 3487 3488 3489 3490 3491
		unlock_page(page);
		return 0;
	}

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

		userpage = kmap_atomic(page);
		memset(userpage + pg_offset, 0,
3492
		       PAGE_SIZE - pg_offset);
3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511
		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;

3512 3513 3514 3515 3516 3517
done:
	if (nr == 0) {
		/* make sure the mapping tag for page dirty gets cleared */
		set_page_writeback(page);
		end_page_writeback(page);
	}
3518 3519 3520 3521
	if (PageError(page)) {
		ret = ret < 0 ? ret : -EIO;
		end_extent_writepage(page, ret, start, page_end);
	}
3522
	unlock_page(page);
3523
	return ret;
3524

3525
done_unlocked:
3526 3527 3528
	return 0;
}

3529
void wait_on_extent_buffer_writeback(struct extent_buffer *eb)
3530
{
3531 3532
	wait_on_bit_io(&eb->bflags, EXTENT_BUFFER_WRITEBACK,
		       TASK_UNINTERRUPTIBLE);
3533 3534
}

3535 3536 3537 3538
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)
3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557
{
	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
3558 3559 3560 3561 3562
		while (1) {
			wait_on_extent_buffer_writeback(eb);
			btrfs_tree_lock(eb);
			if (!test_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags))
				break;
3563 3564 3565 3566
			btrfs_tree_unlock(eb);
		}
	}

3567 3568 3569 3570 3571 3572
	/*
	 * 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);
3573 3574
	if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
		set_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags);
3575
		spin_unlock(&eb->refs_lock);
3576
		btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
3577 3578 3579
		__percpu_counter_add(&fs_info->dirty_metadata_bytes,
				     -eb->len,
				     fs_info->dirty_metadata_batch);
3580
		ret = 1;
3581 3582
	} else {
		spin_unlock(&eb->refs_lock);
3583 3584 3585 3586 3587 3588 3589 3590 3591
	}

	btrfs_tree_unlock(eb);

	if (!ret)
		return ret;

	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = 0; i < num_pages; i++) {
3592
		struct page *p = eb->pages[i];
3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608

		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);
3609
	smp_mb__after_atomic();
3610 3611 3612
	wake_up_bit(&eb->bflags, EXTENT_BUFFER_WRITEBACK);
}

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 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660
static void set_btree_ioerr(struct page *page)
{
	struct extent_buffer *eb = (struct extent_buffer *)page->private;

	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:
3661
		set_bit(BTRFS_FS_BTREE_ERR, &eb->fs_info->flags);
3662 3663
		break;
	case 0:
3664
		set_bit(BTRFS_FS_LOG1_ERR, &eb->fs_info->flags);
3665 3666
		break;
	case 1:
3667
		set_bit(BTRFS_FS_LOG2_ERR, &eb->fs_info->flags);
3668 3669 3670 3671 3672 3673
		break;
	default:
		BUG(); /* unexpected, logic error */
	}
}

3674
static void end_bio_extent_buffer_writepage(struct bio *bio)
3675
{
3676
	struct bio_vec *bvec;
3677
	struct extent_buffer *eb;
3678
	int i, done;
3679

3680
	bio_for_each_segment_all(bvec, bio, i) {
3681 3682 3683 3684 3685 3686
		struct page *page = bvec->bv_page;

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

3687 3688
		if (bio->bi_error ||
		    test_bit(EXTENT_BUFFER_WRITE_ERR, &eb->bflags)) {
3689
			ClearPageUptodate(page);
3690
			set_btree_ioerr(page);
3691 3692 3693 3694 3695 3696 3697 3698
		}

		end_page_writeback(page);

		if (!done)
			continue;

		end_extent_buffer_writeback(eb);
3699
	}
3700 3701 3702 3703

	bio_put(bio);
}

3704
static noinline_for_stack int write_one_eb(struct extent_buffer *eb,
3705 3706 3707 3708 3709
			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;
3710
	struct extent_io_tree *tree = &BTRFS_I(fs_info->btree_inode)->io_tree;
3711
	u64 offset = eb->start;
3712
	u32 nritems;
3713
	unsigned long i, num_pages;
3714
	unsigned long bio_flags = 0;
3715
	unsigned long start, end;
3716
	int write_flags = (epd->sync_io ? REQ_SYNC : 0) | REQ_META;
3717
	int ret = 0;
3718

3719
	clear_bit(EXTENT_BUFFER_WRITE_ERR, &eb->bflags);
3720 3721
	num_pages = num_extent_pages(eb->start, eb->len);
	atomic_set(&eb->io_pages, num_pages);
3722 3723 3724
	if (btrfs_header_owner(eb) == BTRFS_TREE_LOG_OBJECTID)
		bio_flags = EXTENT_BIO_TREE_LOG;

3725 3726
	/* set btree blocks beyond nritems with 0 to avoid stale content. */
	nritems = btrfs_header_nritems(eb);
3727 3728 3729
	if (btrfs_header_level(eb) > 0) {
		end = btrfs_node_key_ptr_offset(nritems);

3730
		memzero_extent_buffer(eb, end, eb->len - end);
3731 3732 3733 3734 3735 3736
	} else {
		/*
		 * leaf:
		 * header 0 1 2 .. N ... data_N .. data_2 data_1 data_0
		 */
		start = btrfs_item_nr_offset(nritems);
3737
		end = btrfs_leaf_data(eb) + leaf_data_end(fs_info, eb);
3738
		memzero_extent_buffer(eb, start, end - start);
3739 3740
	}

3741
	for (i = 0; i < num_pages; i++) {
3742
		struct page *p = eb->pages[i];
3743 3744 3745

		clear_page_dirty_for_io(p);
		set_page_writeback(p);
3746 3747
		ret = submit_extent_page(REQ_OP_WRITE, write_flags, tree, wbc,
					 p, offset >> 9, PAGE_SIZE, 0, bdev,
3748
					 &epd->bio,
3749
					 end_bio_extent_buffer_writepage,
3750
					 0, epd->bio_flags, bio_flags, false);
3751
		epd->bio_flags = bio_flags;
3752
		if (ret) {
3753
			set_btree_ioerr(p);
3754 3755
			if (PageWriteback(p))
				end_page_writeback(p);
3756 3757 3758 3759 3760
			if (atomic_sub_and_test(num_pages - i, &eb->io_pages))
				end_extent_buffer_writeback(eb);
			ret = -EIO;
			break;
		}
3761
		offset += PAGE_SIZE;
3762
		update_nr_written(wbc, 1);
3763 3764 3765 3766 3767
		unlock_page(p);
	}

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

	return ret;
}

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

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

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

			if (!PagePrivate(page))
				continue;

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

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

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

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

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

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

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

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

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

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

3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943
	/*
	 * 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;

3944 3945 3946 3947 3948
	pagevec_init(&pvec, 0);
	if (wbc->range_cyclic) {
		index = mapping->writeback_index; /* Start from prev offset */
		end = -1;
	} else {
3949 3950
		index = wbc->range_start >> PAGE_SHIFT;
		end = wbc->range_end >> PAGE_SHIFT;
3951 3952
		if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
			range_whole = 1;
3953 3954
		scanned = 1;
	}
3955 3956 3957 3958
	if (wbc->sync_mode == WB_SYNC_ALL)
		tag = PAGECACHE_TAG_TOWRITE;
	else
		tag = PAGECACHE_TAG_DIRTY;
3959
retry:
3960 3961
	if (wbc->sync_mode == WB_SYNC_ALL)
		tag_pages_for_writeback(mapping, index, end);
3962
	done_index = index;
3963
	while (!done && !nr_to_write_done && (index <= end) &&
3964 3965
	       (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag,
			min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
3966 3967 3968 3969 3970 3971
		unsigned i;

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

3972
			done_index = page->index;
3973 3974 3975 3976 3977 3978 3979
			/*
			 * 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
			 */
3980 3981 3982
			if (!trylock_page(page)) {
				flush_fn(data);
				lock_page(page);
3983
			}
3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995

			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
3996
			if (wbc->sync_mode != WB_SYNC_NONE) {
3997 3998
				if (PageWriteback(page))
					flush_fn(data);
3999
				wait_on_page_writeback(page);
C
Chris Mason 已提交
4000
			}
4001 4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012 4013

			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;
			}
4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027
			if (ret < 0) {
				/*
				 * done_index is set past this page,
				 * so media errors will not choke
				 * background writeout for the entire
				 * file. This has consequences for
				 * range_cyclic semantics (ie. it may
				 * not be suitable for data integrity
				 * writeout).
				 */
				done_index = page->index + 1;
				done = 1;
				break;
			}
4028 4029 4030 4031 4032 4033 4034

			/*
			 * 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;
4035 4036 4037 4038
		}
		pagevec_release(&pvec);
		cond_resched();
	}
4039
	if (!scanned && !done) {
4040 4041 4042 4043 4044 4045 4046 4047
		/*
		 * 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;
	}
4048 4049 4050 4051

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

4052
	btrfs_add_delayed_iput(inode);
4053
	return ret;
4054 4055
}

4056
static void flush_epd_write_bio(struct extent_page_data *epd)
C
Chris Mason 已提交
4057 4058
{
	if (epd->bio) {
4059 4060
		int ret;

4061
		bio_set_op_attrs(epd->bio, REQ_OP_WRITE,
4062
				 epd->sync_io ? REQ_SYNC : 0);
4063

4064
		ret = submit_one_bio(epd->bio, 0, epd->bio_flags);
4065
		BUG_ON(ret < 0); /* -ENOMEM */
C
Chris Mason 已提交
4066 4067 4068 4069
		epd->bio = NULL;
	}
}

4070 4071 4072 4073 4074 4075
static noinline void flush_write_bio(void *data)
{
	struct extent_page_data *epd = data;
	flush_epd_write_bio(epd);
}

4076 4077 4078 4079 4080 4081 4082 4083 4084
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,
4085
		.extent_locked = 0,
4086
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
4087
		.bio_flags = 0,
4088 4089 4090 4091
	};

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

4092
	flush_epd_write_bio(&epd);
4093 4094 4095
	return ret;
}

4096 4097 4098 4099 4100 4101 4102
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;
4103 4104
	unsigned long nr_pages = (end - start + PAGE_SIZE) >>
		PAGE_SHIFT;
4105 4106 4107 4108 4109 4110

	struct extent_page_data epd = {
		.bio = NULL,
		.tree = tree,
		.get_extent = get_extent,
		.extent_locked = 1,
4111
		.sync_io = mode == WB_SYNC_ALL,
4112
		.bio_flags = 0,
4113 4114 4115 4116 4117 4118 4119 4120
	};
	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
4121
	while (start <= end) {
4122
		page = find_get_page(mapping, start >> PAGE_SHIFT);
4123 4124 4125 4126 4127
		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,
4128
						 start + PAGE_SIZE - 1,
4129 4130 4131
						 NULL, 1);
			unlock_page(page);
		}
4132 4133
		put_page(page);
		start += PAGE_SIZE;
4134 4135
	}

4136
	flush_epd_write_bio(&epd);
4137 4138
	return ret;
}
4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149

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,
4150
		.extent_locked = 0,
4151
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
4152
		.bio_flags = 0,
4153 4154
	};

4155
	ret = extent_write_cache_pages(mapping, wbc, __extent_writepage, &epd,
C
Chris Mason 已提交
4156
				       flush_write_bio);
4157
	flush_epd_write_bio(&epd);
4158 4159 4160 4161 4162 4163 4164 4165 4166 4167
	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
4168
	unsigned long bio_flags = 0;
L
Liu Bo 已提交
4169 4170
	struct page *pagepool[16];
	struct page *page;
4171
	struct extent_map *em_cached = NULL;
L
Liu Bo 已提交
4172
	int nr = 0;
4173
	u64 prev_em_start = (u64)-1;
4174 4175

	for (page_idx = 0; page_idx < nr_pages; page_idx++) {
L
Liu Bo 已提交
4176
		page = list_entry(pages->prev, struct page, lru);
4177 4178 4179

		prefetchw(&page->flags);
		list_del(&page->lru);
L
Liu Bo 已提交
4180
		if (add_to_page_cache_lru(page, mapping,
4181 4182
					page->index,
					readahead_gfp_mask(mapping))) {
4183
			put_page(page);
L
Liu Bo 已提交
4184
			continue;
4185
		}
L
Liu Bo 已提交
4186 4187 4188 4189

		pagepool[nr++] = page;
		if (nr < ARRAY_SIZE(pagepool))
			continue;
4190
		__extent_readpages(tree, pagepool, nr, get_extent, &em_cached,
4191
				   &bio, 0, &bio_flags, &prev_em_start);
L
Liu Bo 已提交
4192
		nr = 0;
4193
	}
4194
	if (nr)
4195
		__extent_readpages(tree, pagepool, nr, get_extent, &em_cached,
4196
				   &bio, 0, &bio_flags, &prev_em_start);
L
Liu Bo 已提交
4197

4198 4199 4200
	if (em_cached)
		free_extent_map(em_cached);

4201 4202
	BUG_ON(!list_empty(pages));
	if (bio)
4203
		return submit_one_bio(bio, 0, bio_flags);
4204 4205 4206 4207 4208 4209 4210 4211 4212 4213 4214
	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)
{
4215
	struct extent_state *cached_state = NULL;
M
Miao Xie 已提交
4216
	u64 start = page_offset(page);
4217
	u64 end = start + PAGE_SIZE - 1;
4218 4219
	size_t blocksize = page->mapping->host->i_sb->s_blocksize;

4220
	start += ALIGN(offset, blocksize);
4221 4222 4223
	if (start > end)
		return 0;

4224
	lock_extent_bits(tree, start, end, &cached_state);
4225
	wait_on_page_writeback(page);
4226
	clear_extent_bit(tree, start, end,
4227 4228
			 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
			 EXTENT_DO_ACCOUNTING,
4229
			 1, 1, &cached_state, GFP_NOFS);
4230 4231 4232
	return 0;
}

4233 4234 4235 4236 4237
/*
 * 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.
 */
4238 4239 4240
static int try_release_extent_state(struct extent_map_tree *map,
				    struct extent_io_tree *tree,
				    struct page *page, gfp_t mask)
4241
{
M
Miao Xie 已提交
4242
	u64 start = page_offset(page);
4243
	u64 end = start + PAGE_SIZE - 1;
4244 4245
	int ret = 1;

4246
	if (test_range_bit(tree, start, end,
4247
			   EXTENT_IOBITS, 0, NULL))
4248 4249
		ret = 0;
	else {
4250 4251 4252 4253
		/*
		 * at this point we can safely clear everything except the
		 * locked bit and the nodatasum bit
		 */
4254
		ret = clear_extent_bit(tree, start, end,
4255 4256
				 ~(EXTENT_LOCKED | EXTENT_NODATASUM),
				 0, 0, NULL, mask);
4257 4258 4259 4260 4261 4262 4263 4264

		/* if clear_extent_bit failed for enomem reasons,
		 * we can't allow the release to continue.
		 */
		if (ret < 0)
			ret = 0;
		else
			ret = 1;
4265 4266 4267 4268
	}
	return ret;
}

4269 4270 4271 4272 4273 4274
/*
 * 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,
4275 4276
			       struct extent_io_tree *tree, struct page *page,
			       gfp_t mask)
4277 4278
{
	struct extent_map *em;
M
Miao Xie 已提交
4279
	u64 start = page_offset(page);
4280
	u64 end = start + PAGE_SIZE - 1;
4281

4282
	if (gfpflags_allow_blocking(mask) &&
4283
	    page->mapping->host->i_size > SZ_16M) {
4284
		u64 len;
4285
		while (start <= end) {
4286
			len = end - start + 1;
4287
			write_lock(&map->lock);
4288
			em = lookup_extent_mapping(map, start, len);
4289
			if (!em) {
4290
				write_unlock(&map->lock);
4291 4292
				break;
			}
4293 4294
			if (test_bit(EXTENT_FLAG_PINNED, &em->flags) ||
			    em->start != start) {
4295
				write_unlock(&map->lock);
4296 4297 4298 4299 4300
				free_extent_map(em);
				break;
			}
			if (!test_range_bit(tree, em->start,
					    extent_map_end(em) - 1,
4301
					    EXTENT_LOCKED | EXTENT_WRITEBACK,
4302
					    0, NULL)) {
4303 4304 4305 4306 4307
				remove_extent_mapping(map, em);
				/* once for the rb tree */
				free_extent_map(em);
			}
			start = extent_map_end(em);
4308
			write_unlock(&map->lock);
4309 4310

			/* once for us */
4311 4312 4313
			free_extent_map(em);
		}
	}
4314
	return try_release_extent_state(map, tree, page, mask);
4315 4316
}

4317 4318 4319 4320 4321 4322 4323 4324 4325
/*
 * 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)
{
4326
	u64 sectorsize = btrfs_inode_sectorsize(inode);
4327 4328 4329 4330 4331 4332
	struct extent_map *em;
	u64 len;

	if (offset >= last)
		return NULL;

4333
	while (1) {
4334 4335 4336
		len = last - offset;
		if (len == 0)
			break;
4337
		len = ALIGN(len, sectorsize);
4338
		em = get_extent(inode, NULL, 0, offset, len, 0);
4339
		if (IS_ERR_OR_NULL(em))
4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355 4356
			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
4357 4358 4359
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
4360
	int ret = 0;
Y
Yehuda Sadeh 已提交
4361 4362 4363
	u64 off = start;
	u64 max = start + len;
	u32 flags = 0;
J
Josef Bacik 已提交
4364 4365
	u32 found_type;
	u64 last;
4366
	u64 last_for_get_extent = 0;
Y
Yehuda Sadeh 已提交
4367
	u64 disko = 0;
4368
	u64 isize = i_size_read(inode);
J
Josef Bacik 已提交
4369
	struct btrfs_key found_key;
Y
Yehuda Sadeh 已提交
4370
	struct extent_map *em = NULL;
4371
	struct extent_state *cached_state = NULL;
J
Josef Bacik 已提交
4372
	struct btrfs_path *path;
4373
	struct btrfs_root *root = BTRFS_I(inode)->root;
Y
Yehuda Sadeh 已提交
4374
	int end = 0;
4375 4376 4377
	u64 em_start = 0;
	u64 em_len = 0;
	u64 em_end = 0;
Y
Yehuda Sadeh 已提交
4378 4379 4380 4381

	if (len == 0)
		return -EINVAL;

J
Josef Bacik 已提交
4382 4383 4384 4385 4386
	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;
	path->leave_spinning = 1;

4387 4388
	start = round_down(start, btrfs_inode_sectorsize(inode));
	len = round_up(max, btrfs_inode_sectorsize(inode)) - start;
4389

4390 4391 4392 4393
	/*
	 * lookup the last file extent.  We're not using i_size here
	 * because there might be preallocation past i_size
	 */
4394 4395
	ret = btrfs_lookup_file_extent(NULL, root, path,
			btrfs_ino(BTRFS_I(inode)), -1, 0);
J
Josef Bacik 已提交
4396 4397 4398
	if (ret < 0) {
		btrfs_free_path(path);
		return ret;
4399 4400 4401 4402
	} else {
		WARN_ON(!ret);
		if (ret == 1)
			ret = 0;
J
Josef Bacik 已提交
4403
	}
4404

J
Josef Bacik 已提交
4405 4406
	path->slots[0]--;
	btrfs_item_key_to_cpu(path->nodes[0], &found_key, path->slots[0]);
4407
	found_type = found_key.type;
J
Josef Bacik 已提交
4408

4409
	/* No extents, but there might be delalloc bits */
4410
	if (found_key.objectid != btrfs_ino(BTRFS_I(inode)) ||
J
Josef Bacik 已提交
4411
	    found_type != BTRFS_EXTENT_DATA_KEY) {
4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422
		/* 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
4423
	}
4424
	btrfs_release_path(path);
J
Josef Bacik 已提交
4425

4426 4427 4428 4429 4430 4431 4432 4433 4434 4435
	/*
	 * 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;
	}

4436
	lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + len - 1,
4437
			 &cached_state);
4438

4439
	em = get_extent_skip_holes(inode, start, last_for_get_extent,
4440
				   get_extent);
Y
Yehuda Sadeh 已提交
4441 4442 4443 4444 4445 4446
	if (!em)
		goto out;
	if (IS_ERR(em)) {
		ret = PTR_ERR(em);
		goto out;
	}
J
Josef Bacik 已提交
4447

Y
Yehuda Sadeh 已提交
4448
	while (!end) {
4449
		u64 offset_in_extent = 0;
4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461

		/* 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
4462

4463 4464
		/*
		 * record the offset from the start of the extent
4465 4466 4467
		 * 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.
4468
		 */
4469 4470
		if (!test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
			offset_in_extent = em_start - em->start;
4471
		em_end = extent_map_end(em);
4472
		em_len = em_end - em_start;
Y
Yehuda Sadeh 已提交
4473 4474 4475
		disko = 0;
		flags = 0;

4476 4477 4478 4479 4480 4481 4482
		/*
		 * bump off for our next call to get_extent
		 */
		off = extent_map_end(em);
		if (off >= max)
			end = 1;

4483
		if (em->block_start == EXTENT_MAP_LAST_BYTE) {
Y
Yehuda Sadeh 已提交
4484 4485
			end = 1;
			flags |= FIEMAP_EXTENT_LAST;
4486
		} else if (em->block_start == EXTENT_MAP_INLINE) {
Y
Yehuda Sadeh 已提交
4487 4488
			flags |= (FIEMAP_EXTENT_DATA_INLINE |
				  FIEMAP_EXTENT_NOT_ALIGNED);
4489
		} else if (em->block_start == EXTENT_MAP_DELALLOC) {
Y
Yehuda Sadeh 已提交
4490 4491
			flags |= (FIEMAP_EXTENT_DELALLOC |
				  FIEMAP_EXTENT_UNKNOWN);
4492
		} else if (fieinfo->fi_extents_max) {
4493 4494
			struct btrfs_trans_handle *trans;

4495 4496
			u64 bytenr = em->block_start -
				(em->start - em->orig_start);
4497

4498
			disko = em->block_start + offset_in_extent;
4499

4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 4510
			/*
			 * We need a trans handle to get delayed refs
			 */
			trans = btrfs_join_transaction(root);
			/*
			 * It's OK if we can't start a trans we can still check
			 * from commit_root
			 */
			if (IS_ERR(trans))
				trans = NULL;

4511 4512 4513
			/*
			 * As btrfs supports shared space, this information
			 * can be exported to userspace tools via
4514 4515 4516
			 * flag FIEMAP_EXTENT_SHARED.  If fi_extents_max == 0
			 * then we're just getting a count and we can skip the
			 * lookup stuff.
4517
			 */
4518
			ret = btrfs_check_shared(trans, root->fs_info,
4519 4520
					root->objectid,
					btrfs_ino(BTRFS_I(inode)), bytenr);
4521
			if (trans)
4522
				btrfs_end_transaction(trans);
4523
			if (ret < 0)
4524
				goto out_free;
4525
			if (ret)
4526
				flags |= FIEMAP_EXTENT_SHARED;
4527
			ret = 0;
Y
Yehuda Sadeh 已提交
4528 4529 4530
		}
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
			flags |= FIEMAP_EXTENT_ENCODED;
4531 4532
		if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			flags |= FIEMAP_EXTENT_UNWRITTEN;
Y
Yehuda Sadeh 已提交
4533 4534 4535

		free_extent_map(em);
		em = NULL;
4536 4537
		if ((em_start >= last) || em_len == (u64)-1 ||
		   (last == (u64)-1 && isize <= em_end)) {
Y
Yehuda Sadeh 已提交
4538 4539 4540 4541
			flags |= FIEMAP_EXTENT_LAST;
			end = 1;
		}

4542 4543 4544 4545 4546 4547 4548 4549
		/* 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
4550 4551 4552
			flags |= FIEMAP_EXTENT_LAST;
			end = 1;
		}
4553 4554
		ret = fiemap_fill_next_extent(fieinfo, em_start, disko,
					      em_len, flags);
4555 4556 4557
		if (ret) {
			if (ret == 1)
				ret = 0;
4558
			goto out_free;
4559
		}
Y
Yehuda Sadeh 已提交
4560 4561 4562 4563
	}
out_free:
	free_extent_map(em);
out:
4564
	btrfs_free_path(path);
L
Liu Bo 已提交
4565
	unlock_extent_cached(&BTRFS_I(inode)->io_tree, start, start + len - 1,
4566
			     &cached_state, GFP_NOFS);
Y
Yehuda Sadeh 已提交
4567 4568 4569
	return ret;
}

4570 4571
static void __free_extent_buffer(struct extent_buffer *eb)
{
4572
	btrfs_leak_debug_del(&eb->leak_list);
4573 4574 4575
	kmem_cache_free(extent_buffer_cache, eb);
}

4576
int extent_buffer_under_io(struct extent_buffer *eb)
4577 4578 4579 4580 4581 4582 4583 4584 4585
{
	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.
 */
4586
static void btrfs_release_extent_buffer_page(struct extent_buffer *eb)
4587 4588 4589 4590 4591 4592 4593
{
	unsigned long index;
	struct page *page;
	int mapped = !test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags);

	BUG_ON(extent_buffer_under_io(eb));

4594 4595
	index = num_extent_pages(eb->start, eb->len);
	if (index == 0)
4596 4597 4598 4599
		return;

	do {
		index--;
4600
		page = eb->pages[index];
4601 4602 4603
		if (!page)
			continue;
		if (mapped)
4604
			spin_lock(&page->mapping->private_lock);
4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616
		/*
		 * 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));
4617
			/*
4618 4619
			 * We need to make sure we haven't be attached
			 * to a new eb.
4620
			 */
4621 4622 4623
			ClearPagePrivate(page);
			set_page_private(page, 0);
			/* One for the page private */
4624
			put_page(page);
4625
		}
4626 4627 4628 4629

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

4630
		/* One for when we allocated the page */
4631
		put_page(page);
4632
	} while (index != 0);
4633 4634 4635 4636 4637 4638 4639
}

/*
 * Helper for releasing the extent buffer.
 */
static inline void btrfs_release_extent_buffer(struct extent_buffer *eb)
{
4640
	btrfs_release_extent_buffer_page(eb);
4641 4642 4643
	__free_extent_buffer(eb);
}

4644 4645
static struct extent_buffer *
__alloc_extent_buffer(struct btrfs_fs_info *fs_info, u64 start,
4646
		      unsigned long len)
4647 4648 4649
{
	struct extent_buffer *eb = NULL;

4650
	eb = kmem_cache_zalloc(extent_buffer_cache, GFP_NOFS|__GFP_NOFAIL);
4651 4652
	eb->start = start;
	eb->len = len;
4653
	eb->fs_info = fs_info;
4654
	eb->bflags = 0;
4655 4656 4657 4658 4659 4660 4661
	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);
4662
	eb->lock_nested = 0;
4663 4664
	init_waitqueue_head(&eb->write_lock_wq);
	init_waitqueue_head(&eb->read_lock_wq);
4665

4666 4667
	btrfs_leak_debug_add(&eb->leak_list, &buffers);

4668
	spin_lock_init(&eb->refs_lock);
4669
	atomic_set(&eb->refs, 1);
4670
	atomic_set(&eb->io_pages, 0);
4671

4672 4673 4674 4675 4676 4677
	/*
	 * 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);
4678 4679 4680 4681

	return eb;
}

4682 4683 4684 4685 4686 4687 4688
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);

4689
	new = __alloc_extent_buffer(src->fs_info, src->start, src->len);
4690 4691 4692 4693
	if (new == NULL)
		return NULL;

	for (i = 0; i < num_pages; i++) {
4694
		p = alloc_page(GFP_NOFS);
4695 4696 4697 4698
		if (!p) {
			btrfs_release_extent_buffer(new);
			return NULL;
		}
4699 4700 4701 4702
		attach_extent_buffer_page(new, p);
		WARN_ON(PageDirty(p));
		SetPageUptodate(p);
		new->pages[i] = p;
4703
		copy_page(page_address(p), page_address(src->pages[i]));
4704 4705 4706 4707 4708 4709 4710 4711
	}

	set_bit(EXTENT_BUFFER_UPTODATE, &new->bflags);
	set_bit(EXTENT_BUFFER_DUMMY, &new->bflags);

	return new;
}

4712 4713
struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
						  u64 start, unsigned long len)
4714 4715
{
	struct extent_buffer *eb;
4716
	unsigned long num_pages;
4717 4718
	unsigned long i;

4719
	num_pages = num_extent_pages(start, len);
4720 4721

	eb = __alloc_extent_buffer(fs_info, start, len);
4722 4723 4724 4725
	if (!eb)
		return NULL;

	for (i = 0; i < num_pages; i++) {
4726
		eb->pages[i] = alloc_page(GFP_NOFS);
4727 4728 4729 4730 4731 4732 4733 4734 4735
		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:
4736 4737
	for (; i > 0; i--)
		__free_page(eb->pages[i - 1]);
4738 4739 4740 4741
	__free_extent_buffer(eb);
	return NULL;
}

4742
struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
4743
						u64 start)
4744
{
4745
	return __alloc_dummy_extent_buffer(fs_info, start, fs_info->nodesize);
4746 4747
}

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

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

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

4786 4787
	check_buffer_tree_ref(eb);

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

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

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

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

	return NULL;
}

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

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

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

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

4895
	if (!IS_ALIGNED(start, fs_info->sectorsize)) {
4896 4897 4898 4899
		btrfs_err(fs_info, "bad tree block start %llu", start);
		return ERR_PTR(-EINVAL);
	}

4900
	eb = find_extent_buffer(fs_info, start);
4901
	if (eb)
4902 4903
		return eb;

4904
	eb = __alloc_extent_buffer(fs_info, start, len);
4905
	if (!eb)
4906
		return ERR_PTR(-ENOMEM);
4907

4908
	for (i = 0; i < num_pages; i++, index++) {
4909
		p = find_or_create_page(mapping, index, GFP_NOFS|__GFP_NOFAIL);
4910 4911
		if (!p) {
			exists = ERR_PTR(-ENOMEM);
4912
			goto free_eb;
4913
		}
J
Josef Bacik 已提交
4914 4915 4916 4917 4918 4919 4920 4921 4922 4923 4924 4925 4926 4927

		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);
4928
				put_page(p);
4929
				mark_extent_buffer_accessed(exists, p);
J
Josef Bacik 已提交
4930 4931
				goto free_eb;
			}
4932
			exists = NULL;
J
Josef Bacik 已提交
4933

4934
			/*
J
Josef Bacik 已提交
4935 4936 4937 4938
			 * Do this so attach doesn't complain and we need to
			 * drop the ref the old guy had.
			 */
			ClearPagePrivate(p);
4939
			WARN_ON(PageDirty(p));
4940
			put_page(p);
4941
		}
J
Josef Bacik 已提交
4942 4943
		attach_extent_buffer_page(eb, p);
		spin_unlock(&mapping->private_lock);
4944
		WARN_ON(PageDirty(p));
4945
		eb->pages[i] = p;
4946 4947
		if (!PageUptodate(p))
			uptodate = 0;
C
Chris Mason 已提交
4948 4949 4950 4951 4952

		/*
		 * see below about how we avoid a nasty race with release page
		 * and why we unlock later
		 */
4953 4954
	}
	if (uptodate)
4955
		set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
4956
again:
4957
	ret = radix_tree_preload(GFP_NOFS);
4958 4959
	if (ret) {
		exists = ERR_PTR(ret);
4960
		goto free_eb;
4961
	}
4962

4963 4964
	spin_lock(&fs_info->buffer_lock);
	ret = radix_tree_insert(&fs_info->buffer_radix,
4965
				start >> PAGE_SHIFT, eb);
4966
	spin_unlock(&fs_info->buffer_lock);
4967
	radix_tree_preload_end();
4968
	if (ret == -EEXIST) {
4969
		exists = find_extent_buffer(fs_info, start);
4970 4971 4972
		if (exists)
			goto free_eb;
		else
4973
			goto again;
4974 4975
	}
	/* add one reference for the tree */
4976
	check_buffer_tree_ref(eb);
4977
	set_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags);
C
Chris Mason 已提交
4978 4979 4980 4981 4982 4983 4984 4985 4986 4987

	/*
	 * 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
	 */
4988 4989
	SetPageChecked(eb->pages[0]);
	for (i = 1; i < num_pages; i++) {
4990
		p = eb->pages[i];
4991 4992 4993 4994
		ClearPageChecked(p);
		unlock_page(p);
	}
	unlock_page(eb->pages[0]);
4995 4996
	return eb;

4997
free_eb:
4998
	WARN_ON(!atomic_dec_and_test(&eb->refs));
4999 5000 5001 5002
	for (i = 0; i < num_pages; i++) {
		if (eb->pages[i])
			unlock_page(eb->pages[i]);
	}
C
Chris Mason 已提交
5003

5004
	btrfs_release_extent_buffer(eb);
5005
	return exists;
5006 5007
}

5008 5009 5010 5011 5012 5013 5014 5015 5016
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 */
5017
static int release_extent_buffer(struct extent_buffer *eb)
5018 5019 5020
{
	WARN_ON(atomic_read(&eb->refs) == 0);
	if (atomic_dec_and_test(&eb->refs)) {
5021
		if (test_and_clear_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags)) {
5022
			struct btrfs_fs_info *fs_info = eb->fs_info;
5023

5024
			spin_unlock(&eb->refs_lock);
5025

5026 5027
			spin_lock(&fs_info->buffer_lock);
			radix_tree_delete(&fs_info->buffer_radix,
5028
					  eb->start >> PAGE_SHIFT);
5029
			spin_unlock(&fs_info->buffer_lock);
5030 5031
		} else {
			spin_unlock(&eb->refs_lock);
5032
		}
5033 5034

		/* Should be safe to release our pages at this point */
5035
		btrfs_release_extent_buffer_page(eb);
5036 5037 5038 5039 5040 5041
#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
		if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags))) {
			__free_extent_buffer(eb);
			return 1;
		}
#endif
5042
		call_rcu(&eb->rcu_head, btrfs_release_extent_buffer_rcu);
5043
		return 1;
5044 5045
	}
	spin_unlock(&eb->refs_lock);
5046 5047

	return 0;
5048 5049
}

5050 5051
void free_extent_buffer(struct extent_buffer *eb)
{
5052 5053
	int refs;
	int old;
5054 5055 5056
	if (!eb)
		return;

5057 5058 5059 5060 5061 5062 5063 5064 5065
	while (1) {
		refs = atomic_read(&eb->refs);
		if (refs <= 3)
			break;
		old = atomic_cmpxchg(&eb->refs, refs, refs - 1);
		if (old == refs)
			return;
	}

5066
	spin_lock(&eb->refs_lock);
5067 5068 5069 5070
	if (atomic_read(&eb->refs) == 2 &&
	    test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags))
		atomic_dec(&eb->refs);

5071 5072
	if (atomic_read(&eb->refs) == 2 &&
	    test_bit(EXTENT_BUFFER_STALE, &eb->bflags) &&
5073
	    !extent_buffer_under_io(eb) &&
5074 5075 5076 5077 5078 5079 5080
	    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.
	 */
5081
	release_extent_buffer(eb);
5082 5083 5084 5085 5086
}

void free_extent_buffer_stale(struct extent_buffer *eb)
{
	if (!eb)
5087 5088
		return;

5089 5090 5091
	spin_lock(&eb->refs_lock);
	set_bit(EXTENT_BUFFER_STALE, &eb->bflags);

5092
	if (atomic_read(&eb->refs) == 2 && !extent_buffer_under_io(eb) &&
5093 5094
	    test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
		atomic_dec(&eb->refs);
5095
	release_extent_buffer(eb);
5096 5097
}

5098
void clear_extent_buffer_dirty(struct extent_buffer *eb)
5099 5100 5101 5102 5103 5104 5105 5106
{
	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++) {
5107
		page = eb->pages[i];
5108
		if (!PageDirty(page))
C
Chris Mason 已提交
5109 5110
			continue;

5111
		lock_page(page);
C
Chris Mason 已提交
5112 5113
		WARN_ON(!PagePrivate(page));

5114
		clear_page_dirty_for_io(page);
5115
		spin_lock_irq(&page->mapping->tree_lock);
5116 5117 5118 5119 5120
		if (!PageDirty(page)) {
			radix_tree_tag_clear(&page->mapping->page_tree,
						page_index(page),
						PAGECACHE_TAG_DIRTY);
		}
5121
		spin_unlock_irq(&page->mapping->tree_lock);
5122
		ClearPageError(page);
5123
		unlock_page(page);
5124
	}
5125
	WARN_ON(atomic_read(&eb->refs) == 0);
5126 5127
}

5128
int set_extent_buffer_dirty(struct extent_buffer *eb)
5129 5130 5131
{
	unsigned long i;
	unsigned long num_pages;
5132
	int was_dirty = 0;
5133

5134 5135
	check_buffer_tree_ref(eb);

5136
	was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
5137

5138
	num_pages = num_extent_pages(eb->start, eb->len);
5139
	WARN_ON(atomic_read(&eb->refs) == 0);
5140 5141
	WARN_ON(!test_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags));

5142
	for (i = 0; i < num_pages; i++)
5143
		set_page_dirty(eb->pages[i]);
5144
	return was_dirty;
5145 5146
}

5147
void clear_extent_buffer_uptodate(struct extent_buffer *eb)
5148 5149 5150 5151 5152
{
	unsigned long i;
	struct page *page;
	unsigned long num_pages;

5153
	clear_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5154
	num_pages = num_extent_pages(eb->start, eb->len);
5155
	for (i = 0; i < num_pages; i++) {
5156
		page = eb->pages[i];
C
Chris Mason 已提交
5157 5158
		if (page)
			ClearPageUptodate(page);
5159 5160 5161
	}
}

5162
void set_extent_buffer_uptodate(struct extent_buffer *eb)
5163 5164 5165 5166 5167
{
	unsigned long i;
	struct page *page;
	unsigned long num_pages;

5168
	set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5169 5170
	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = 0; i < num_pages; i++) {
5171
		page = eb->pages[i];
5172 5173 5174 5175
		SetPageUptodate(page);
	}
}

5176
int extent_buffer_uptodate(struct extent_buffer *eb)
5177
{
5178
	return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5179 5180 5181
}

int read_extent_buffer_pages(struct extent_io_tree *tree,
5182
			     struct extent_buffer *eb, int wait,
5183
			     get_extent_t *get_extent, int mirror_num)
5184 5185 5186 5187 5188
{
	unsigned long i;
	struct page *page;
	int err;
	int ret = 0;
5189 5190
	int locked_pages = 0;
	int all_uptodate = 1;
5191
	unsigned long num_pages;
5192
	unsigned long num_reads = 0;
5193
	struct bio *bio = NULL;
C
Chris Mason 已提交
5194
	unsigned long bio_flags = 0;
5195

5196
	if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags))
5197 5198 5199
		return 0;

	num_pages = num_extent_pages(eb->start, eb->len);
5200
	for (i = 0; i < num_pages; i++) {
5201
		page = eb->pages[i];
5202
		if (wait == WAIT_NONE) {
5203
			if (!trylock_page(page))
5204
				goto unlock_exit;
5205 5206 5207
		} else {
			lock_page(page);
		}
5208
		locked_pages++;
5209 5210 5211 5212 5213 5214
	}
	/*
	 * We need to firstly lock all pages to make sure that
	 * the uptodate bit of our pages won't be affected by
	 * clear_extent_buffer_uptodate().
	 */
5215
	for (i = 0; i < num_pages; i++) {
5216
		page = eb->pages[i];
5217 5218
		if (!PageUptodate(page)) {
			num_reads++;
5219
			all_uptodate = 0;
5220
		}
5221
	}
5222

5223
	if (all_uptodate) {
5224
		set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5225 5226 5227
		goto unlock_exit;
	}

5228
	clear_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags);
5229
	eb->read_mirror = 0;
5230
	atomic_set(&eb->io_pages, num_reads);
5231
	for (i = 0; i < num_pages; i++) {
5232
		page = eb->pages[i];
5233

5234
		if (!PageUptodate(page)) {
5235 5236 5237 5238 5239 5240
			if (ret) {
				atomic_dec(&eb->io_pages);
				unlock_page(page);
				continue;
			}

5241
			ClearPageError(page);
5242
			err = __extent_read_full_page(tree, page,
5243
						      get_extent, &bio,
5244
						      mirror_num, &bio_flags,
5245
						      REQ_META);
5246
			if (err) {
5247
				ret = err;
5248 5249 5250 5251 5252 5253 5254 5255 5256 5257
				/*
				 * We use &bio in above __extent_read_full_page,
				 * so we ensure that if it returns error, the
				 * current page fails to add itself to bio and
				 * it's been unlocked.
				 *
				 * We must dec io_pages by ourselves.
				 */
				atomic_dec(&eb->io_pages);
			}
5258 5259 5260 5261 5262
		} else {
			unlock_page(page);
		}
	}

5263
	if (bio) {
5264
		err = submit_one_bio(bio, mirror_num, bio_flags);
5265 5266
		if (err)
			return err;
5267
	}
5268

5269
	if (ret || wait != WAIT_COMPLETE)
5270
		return ret;
C
Chris Mason 已提交
5271

5272
	for (i = 0; i < num_pages; i++) {
5273
		page = eb->pages[i];
5274
		wait_on_page_locked(page);
C
Chris Mason 已提交
5275
		if (!PageUptodate(page))
5276 5277
			ret = -EIO;
	}
C
Chris Mason 已提交
5278

5279
	return ret;
5280 5281

unlock_exit:
C
Chris Mason 已提交
5282
	while (locked_pages > 0) {
5283
		locked_pages--;
5284 5285
		page = eb->pages[locked_pages];
		unlock_page(page);
5286 5287
	}
	return ret;
5288 5289 5290 5291 5292 5293 5294 5295 5296 5297 5298
}

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;
5299 5300
	size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_SHIFT;
5301 5302 5303 5304

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

5305
	offset = (start_offset + start) & (PAGE_SIZE - 1);
5306

C
Chris Mason 已提交
5307
	while (len > 0) {
5308
		page = eb->pages[i];
5309

5310
		cur = min(len, (PAGE_SIZE - offset));
5311
		kaddr = page_address(page);
5312 5313 5314 5315 5316 5317 5318 5319 5320
		memcpy(dst, kaddr + offset, cur);

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

5321 5322 5323 5324 5325 5326 5327 5328 5329
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;
5330 5331
	size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_SHIFT;
5332 5333 5334 5335 5336
	int ret = 0;

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

5337
	offset = (start_offset + start) & (PAGE_SIZE - 1);
5338 5339

	while (len > 0) {
5340
		page = eb->pages[i];
5341

5342
		cur = min(len, (PAGE_SIZE - offset));
5343 5344 5345 5346 5347 5348 5349 5350 5351 5352 5353 5354 5355 5356 5357
		kaddr = page_address(page);
		if (copy_to_user(dst, kaddr + offset, cur)) {
			ret = -EFAULT;
			break;
		}

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

	return ret;
}

5358 5359 5360 5361 5362
/*
 * return 0 if the item is found within a page.
 * return 1 if the item spans two pages.
 * return -EINVAL otherwise.
 */
5363
int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start,
5364
			       unsigned long min_len, char **map,
5365
			       unsigned long *map_start,
5366
			       unsigned long *map_len)
5367
{
5368
	size_t offset = start & (PAGE_SIZE - 1);
5369 5370
	char *kaddr;
	struct page *p;
5371 5372
	size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_SHIFT;
5373
	unsigned long end_i = (start_offset + start + min_len - 1) >>
5374
		PAGE_SHIFT;
5375 5376

	if (i != end_i)
5377
		return 1;
5378 5379 5380 5381 5382 5383

	if (i == 0) {
		offset = start_offset;
		*map_start = 0;
	} else {
		offset = 0;
5384
		*map_start = ((u64)i << PAGE_SHIFT) - start_offset;
5385
	}
C
Chris Mason 已提交
5386

5387
	if (start + min_len > eb->len) {
J
Jeff Mahoney 已提交
5388
		WARN(1, KERN_ERR "btrfs bad mapping eb start %llu len %lu, wanted %lu %lu\n",
5389
		       eb->start, eb->len, start, min_len);
5390
		return -EINVAL;
5391 5392
	}

5393
	p = eb->pages[i];
5394
	kaddr = page_address(p);
5395
	*map = kaddr + offset;
5396
	*map_len = PAGE_SIZE - offset;
5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408
	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;
5409 5410
	size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_SHIFT;
5411 5412 5413 5414 5415
	int ret = 0;

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

5416
	offset = (start_offset + start) & (PAGE_SIZE - 1);
5417

C
Chris Mason 已提交
5418
	while (len > 0) {
5419
		page = eb->pages[i];
5420

5421
		cur = min(len, (PAGE_SIZE - offset));
5422

5423
		kaddr = page_address(page);
5424 5425 5426 5427 5428 5429 5430 5431 5432 5433 5434 5435
		ret = memcmp(ptr, kaddr + offset, cur);
		if (ret)
			break;

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

5436 5437 5438 5439 5440 5441 5442 5443 5444 5445 5446 5447 5448 5449 5450 5451 5452 5453 5454 5455 5456
void write_extent_buffer_chunk_tree_uuid(struct extent_buffer *eb,
		const void *srcv)
{
	char *kaddr;

	WARN_ON(!PageUptodate(eb->pages[0]));
	kaddr = page_address(eb->pages[0]);
	memcpy(kaddr + offsetof(struct btrfs_header, chunk_tree_uuid), srcv,
			BTRFS_FSID_SIZE);
}

void write_extent_buffer_fsid(struct extent_buffer *eb, const void *srcv)
{
	char *kaddr;

	WARN_ON(!PageUptodate(eb->pages[0]));
	kaddr = page_address(eb->pages[0]);
	memcpy(kaddr + offsetof(struct btrfs_header, fsid), srcv,
			BTRFS_FSID_SIZE);
}

5457 5458 5459 5460 5461 5462 5463 5464
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;
5465 5466
	size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_SHIFT;
5467 5468 5469 5470

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

5471
	offset = (start_offset + start) & (PAGE_SIZE - 1);
5472

C
Chris Mason 已提交
5473
	while (len > 0) {
5474
		page = eb->pages[i];
5475 5476
		WARN_ON(!PageUptodate(page));

5477
		cur = min(len, PAGE_SIZE - offset);
5478
		kaddr = page_address(page);
5479 5480 5481 5482 5483 5484 5485 5486 5487
		memcpy(kaddr + offset, src, cur);

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

5488 5489
void memzero_extent_buffer(struct extent_buffer *eb, unsigned long start,
		unsigned long len)
5490 5491 5492 5493 5494
{
	size_t cur;
	size_t offset;
	struct page *page;
	char *kaddr;
5495 5496
	size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_SHIFT;
5497 5498 5499 5500

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

5501
	offset = (start_offset + start) & (PAGE_SIZE - 1);
5502

C
Chris Mason 已提交
5503
	while (len > 0) {
5504
		page = eb->pages[i];
5505 5506
		WARN_ON(!PageUptodate(page));

5507
		cur = min(len, PAGE_SIZE - offset);
5508
		kaddr = page_address(page);
5509
		memset(kaddr + offset, 0, cur);
5510 5511 5512 5513 5514 5515 5516

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

5517 5518 5519 5520 5521 5522 5523 5524 5525 5526 5527 5528 5529 5530
void copy_extent_buffer_full(struct extent_buffer *dst,
			     struct extent_buffer *src)
{
	int i;
	unsigned num_pages;

	ASSERT(dst->len == src->len);

	num_pages = num_extent_pages(dst->start, dst->len);
	for (i = 0; i < num_pages; i++)
		copy_page(page_address(dst->pages[i]),
				page_address(src->pages[i]));
}

5531 5532 5533 5534 5535 5536 5537 5538 5539
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;
5540 5541
	size_t start_offset = dst->start & ((u64)PAGE_SIZE - 1);
	unsigned long i = (start_offset + dst_offset) >> PAGE_SHIFT;
5542 5543 5544 5545

	WARN_ON(src->len != dst_len);

	offset = (start_offset + dst_offset) &
5546
		(PAGE_SIZE - 1);
5547

C
Chris Mason 已提交
5548
	while (len > 0) {
5549
		page = dst->pages[i];
5550 5551
		WARN_ON(!PageUptodate(page));

5552
		cur = min(len, (unsigned long)(PAGE_SIZE - offset));
5553

5554
		kaddr = page_address(page);
5555 5556 5557 5558 5559 5560 5561 5562 5563
		read_extent_buffer(src, kaddr + offset, src_offset, cur);

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

5564 5565 5566 5567 5568 5569 5570 5571 5572 5573 5574
void le_bitmap_set(u8 *map, unsigned int start, int len)
{
	u8 *p = map + BIT_BYTE(start);
	const unsigned int size = start + len;
	int bits_to_set = BITS_PER_BYTE - (start % BITS_PER_BYTE);
	u8 mask_to_set = BITMAP_FIRST_BYTE_MASK(start);

	while (len - bits_to_set >= 0) {
		*p |= mask_to_set;
		len -= bits_to_set;
		bits_to_set = BITS_PER_BYTE;
D
Dan Carpenter 已提交
5575
		mask_to_set = ~0;
5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594
		p++;
	}
	if (len) {
		mask_to_set &= BITMAP_LAST_BYTE_MASK(size);
		*p |= mask_to_set;
	}
}

void le_bitmap_clear(u8 *map, unsigned int start, int len)
{
	u8 *p = map + BIT_BYTE(start);
	const unsigned int size = start + len;
	int bits_to_clear = BITS_PER_BYTE - (start % BITS_PER_BYTE);
	u8 mask_to_clear = BITMAP_FIRST_BYTE_MASK(start);

	while (len - bits_to_clear >= 0) {
		*p &= ~mask_to_clear;
		len -= bits_to_clear;
		bits_to_clear = BITS_PER_BYTE;
D
Dan Carpenter 已提交
5595
		mask_to_clear = ~0;
5596 5597 5598 5599 5600 5601 5602
		p++;
	}
	if (len) {
		mask_to_clear &= BITMAP_LAST_BYTE_MASK(size);
		*p &= ~mask_to_clear;
	}
}
5603 5604 5605 5606 5607 5608 5609 5610 5611 5612 5613 5614 5615 5616 5617 5618 5619 5620 5621

/*
 * eb_bitmap_offset() - calculate the page and offset of the byte containing the
 * given bit number
 * @eb: the extent buffer
 * @start: offset of the bitmap item in the extent buffer
 * @nr: bit number
 * @page_index: return index of the page in the extent buffer that contains the
 * given bit number
 * @page_offset: return offset into the page given by page_index
 *
 * This helper hides the ugliness of finding the byte in an extent buffer which
 * contains a given bit.
 */
static inline void eb_bitmap_offset(struct extent_buffer *eb,
				    unsigned long start, unsigned long nr,
				    unsigned long *page_index,
				    size_t *page_offset)
{
5622
	size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
5623 5624 5625 5626 5627 5628 5629 5630 5631 5632
	size_t byte_offset = BIT_BYTE(nr);
	size_t offset;

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

5633 5634
	*page_index = offset >> PAGE_SHIFT;
	*page_offset = offset & (PAGE_SIZE - 1);
5635 5636 5637 5638 5639 5640 5641 5642 5643 5644 5645
}

/**
 * extent_buffer_test_bit - determine whether a bit in a bitmap item is set
 * @eb: the extent buffer
 * @start: offset of the bitmap item in the extent buffer
 * @nr: bit number to test
 */
int extent_buffer_test_bit(struct extent_buffer *eb, unsigned long start,
			   unsigned long nr)
{
5646
	u8 *kaddr;
5647 5648 5649 5650 5651 5652 5653 5654 5655 5656 5657 5658 5659 5660 5661 5662 5663 5664 5665 5666 5667
	struct page *page;
	unsigned long i;
	size_t offset;

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

/**
 * extent_buffer_bitmap_set - set an area of a bitmap
 * @eb: the extent buffer
 * @start: offset of the bitmap item in the extent buffer
 * @pos: bit number of the first bit
 * @len: number of bits to set
 */
void extent_buffer_bitmap_set(struct extent_buffer *eb, unsigned long start,
			      unsigned long pos, unsigned long len)
{
5668
	u8 *kaddr;
5669 5670 5671 5672 5673
	struct page *page;
	unsigned long i;
	size_t offset;
	const unsigned int size = pos + len;
	int bits_to_set = BITS_PER_BYTE - (pos % BITS_PER_BYTE);
5674
	u8 mask_to_set = BITMAP_FIRST_BYTE_MASK(pos);
5675 5676 5677 5678 5679 5680 5681 5682 5683 5684

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

	while (len >= bits_to_set) {
		kaddr[offset] |= mask_to_set;
		len -= bits_to_set;
		bits_to_set = BITS_PER_BYTE;
D
Dan Carpenter 已提交
5685
		mask_to_set = ~0;
5686
		if (++offset >= PAGE_SIZE && len > 0) {
5687 5688 5689 5690 5691 5692 5693 5694 5695 5696 5697 5698 5699 5700 5701 5702 5703 5704 5705 5706 5707 5708 5709
			offset = 0;
			page = eb->pages[++i];
			WARN_ON(!PageUptodate(page));
			kaddr = page_address(page);
		}
	}
	if (len) {
		mask_to_set &= BITMAP_LAST_BYTE_MASK(size);
		kaddr[offset] |= mask_to_set;
	}
}


/**
 * extent_buffer_bitmap_clear - clear an area of a bitmap
 * @eb: the extent buffer
 * @start: offset of the bitmap item in the extent buffer
 * @pos: bit number of the first bit
 * @len: number of bits to clear
 */
void extent_buffer_bitmap_clear(struct extent_buffer *eb, unsigned long start,
				unsigned long pos, unsigned long len)
{
5710
	u8 *kaddr;
5711 5712 5713 5714 5715
	struct page *page;
	unsigned long i;
	size_t offset;
	const unsigned int size = pos + len;
	int bits_to_clear = BITS_PER_BYTE - (pos % BITS_PER_BYTE);
5716
	u8 mask_to_clear = BITMAP_FIRST_BYTE_MASK(pos);
5717 5718 5719 5720 5721 5722 5723 5724 5725 5726

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

	while (len >= bits_to_clear) {
		kaddr[offset] &= ~mask_to_clear;
		len -= bits_to_clear;
		bits_to_clear = BITS_PER_BYTE;
D
Dan Carpenter 已提交
5727
		mask_to_clear = ~0;
5728
		if (++offset >= PAGE_SIZE && len > 0) {
5729 5730 5731 5732 5733 5734 5735 5736 5737 5738 5739 5740
			offset = 0;
			page = eb->pages[++i];
			WARN_ON(!PageUptodate(page));
			kaddr = page_address(page);
		}
	}
	if (len) {
		mask_to_clear &= BITMAP_LAST_BYTE_MASK(size);
		kaddr[offset] &= ~mask_to_clear;
	}
}

5741 5742 5743 5744 5745 5746
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;
}

5747 5748 5749 5750
static void copy_pages(struct page *dst_page, struct page *src_page,
		       unsigned long dst_off, unsigned long src_off,
		       unsigned long len)
{
5751
	char *dst_kaddr = page_address(dst_page);
5752
	char *src_kaddr;
5753
	int must_memmove = 0;
5754

5755
	if (dst_page != src_page) {
5756
		src_kaddr = page_address(src_page);
5757
	} else {
5758
		src_kaddr = dst_kaddr;
5759 5760
		if (areas_overlap(src_off, dst_off, len))
			must_memmove = 1;
5761
	}
5762

5763 5764 5765 5766
	if (must_memmove)
		memmove(dst_kaddr + dst_off, src_kaddr + src_off, len);
	else
		memcpy(dst_kaddr + dst_off, src_kaddr + src_off, len);
5767 5768 5769 5770 5771
}

void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
			   unsigned long src_offset, unsigned long len)
{
5772
	struct btrfs_fs_info *fs_info = dst->fs_info;
5773 5774 5775
	size_t cur;
	size_t dst_off_in_page;
	size_t src_off_in_page;
5776
	size_t start_offset = dst->start & ((u64)PAGE_SIZE - 1);
5777 5778 5779 5780
	unsigned long dst_i;
	unsigned long src_i;

	if (src_offset + len > dst->len) {
5781
		btrfs_err(fs_info,
J
Jeff Mahoney 已提交
5782 5783
			"memmove bogus src_offset %lu move len %lu dst len %lu",
			 src_offset, len, dst->len);
5784 5785 5786
		BUG_ON(1);
	}
	if (dst_offset + len > dst->len) {
5787
		btrfs_err(fs_info,
J
Jeff Mahoney 已提交
5788 5789
			"memmove bogus dst_offset %lu move len %lu dst len %lu",
			 dst_offset, len, dst->len);
5790 5791 5792
		BUG_ON(1);
	}

C
Chris Mason 已提交
5793
	while (len > 0) {
5794
		dst_off_in_page = (start_offset + dst_offset) &
5795
			(PAGE_SIZE - 1);
5796
		src_off_in_page = (start_offset + src_offset) &
5797
			(PAGE_SIZE - 1);
5798

5799 5800
		dst_i = (start_offset + dst_offset) >> PAGE_SHIFT;
		src_i = (start_offset + src_offset) >> PAGE_SHIFT;
5801

5802
		cur = min(len, (unsigned long)(PAGE_SIZE -
5803 5804
					       src_off_in_page));
		cur = min_t(unsigned long, cur,
5805
			(unsigned long)(PAGE_SIZE - dst_off_in_page));
5806

5807
		copy_pages(dst->pages[dst_i], dst->pages[src_i],
5808 5809 5810 5811 5812 5813 5814 5815 5816 5817 5818
			   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)
{
5819
	struct btrfs_fs_info *fs_info = dst->fs_info;
5820 5821 5822 5823 5824
	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;
5825
	size_t start_offset = dst->start & ((u64)PAGE_SIZE - 1);
5826 5827 5828 5829
	unsigned long dst_i;
	unsigned long src_i;

	if (src_offset + len > dst->len) {
5830
		btrfs_err(fs_info,
J
Jeff Mahoney 已提交
5831 5832
			  "memmove bogus src_offset %lu move len %lu len %lu",
			  src_offset, len, dst->len);
5833 5834 5835
		BUG_ON(1);
	}
	if (dst_offset + len > dst->len) {
5836
		btrfs_err(fs_info,
J
Jeff Mahoney 已提交
5837 5838
			  "memmove bogus dst_offset %lu move len %lu len %lu",
			  dst_offset, len, dst->len);
5839 5840
		BUG_ON(1);
	}
5841
	if (dst_offset < src_offset) {
5842 5843 5844
		memcpy_extent_buffer(dst, dst_offset, src_offset, len);
		return;
	}
C
Chris Mason 已提交
5845
	while (len > 0) {
5846 5847
		dst_i = (start_offset + dst_end) >> PAGE_SHIFT;
		src_i = (start_offset + src_end) >> PAGE_SHIFT;
5848 5849

		dst_off_in_page = (start_offset + dst_end) &
5850
			(PAGE_SIZE - 1);
5851
		src_off_in_page = (start_offset + src_end) &
5852
			(PAGE_SIZE - 1);
5853 5854 5855

		cur = min_t(unsigned long, len, src_off_in_page + 1);
		cur = min(cur, dst_off_in_page + 1);
5856
		copy_pages(dst->pages[dst_i], dst->pages[src_i],
5857 5858 5859 5860 5861 5862 5863 5864
			   dst_off_in_page - cur + 1,
			   src_off_in_page - cur + 1, cur);

		dst_end -= cur;
		src_end -= cur;
		len -= cur;
	}
}
5865

5866
int try_release_extent_buffer(struct page *page)
5867
{
5868 5869
	struct extent_buffer *eb;

5870
	/*
5871
	 * We need to make sure nobody is attaching this page to an eb right
5872 5873 5874 5875 5876
	 * now.
	 */
	spin_lock(&page->mapping->private_lock);
	if (!PagePrivate(page)) {
		spin_unlock(&page->mapping->private_lock);
J
Josef Bacik 已提交
5877
		return 1;
5878
	}
5879

5880 5881
	eb = (struct extent_buffer *)page->private;
	BUG_ON(!eb);
5882 5883

	/*
5884 5885 5886
	 * 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.
5887
	 */
5888
	spin_lock(&eb->refs_lock);
5889
	if (atomic_read(&eb->refs) != 1 || extent_buffer_under_io(eb)) {
5890 5891 5892
		spin_unlock(&eb->refs_lock);
		spin_unlock(&page->mapping->private_lock);
		return 0;
5893
	}
5894
	spin_unlock(&page->mapping->private_lock);
5895

5896
	/*
5897 5898
	 * 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.
5899
	 */
5900 5901 5902
	if (!test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags)) {
		spin_unlock(&eb->refs_lock);
		return 0;
5903
	}
5904

5905
	return release_extent_buffer(eb);
5906
}