extent_io.c 130.5 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
#include "compat.h"
17 18
#include "ctree.h"
#include "btrfs_inode.h"
19
#include "volumes.h"
20
#include "check-integrity.h"
21
#include "locking.h"
22
#include "rcu-string.h"
23 24 25

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

28
#ifdef CONFIG_BTRFS_DEBUG
29 30
static LIST_HEAD(buffers);
static LIST_HEAD(states);
C
Chris Mason 已提交
31

C
Chris Mason 已提交
32
static DEFINE_SPINLOCK(leak_lock);
33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63

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);
		printk(KERN_ERR "btrfs state leak: start %llu end %llu "
		       "state %lu in tree %p refs %d\n",
64 65
		       state->start, state->end, state->state, state->tree,
		       atomic_read(&state->refs));
66 67 68 69 70 71 72
		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);
		printk(KERN_ERR "btrfs buffer leak start %llu len %lu "
73 74
		       "refs %d\n",
		       eb->start, eb->len, atomic_read(&eb->refs));
75 76 77 78
		list_del(&eb->leak_list);
		kmem_cache_free(extent_buffer_cache, eb);
	}
}
79 80 81 82 83 84 85 86 87 88 89

#define btrfs_debug_check_extent_io_range(inode, start, end)		\
	__btrfs_debug_check_extent_io_range(__func__, (inode), (start), (end))
static inline void __btrfs_debug_check_extent_io_range(const char *caller,
		struct inode *inode, u64 start, u64 end)
{
	u64 isize = i_size_read(inode);

	if (end >= PAGE_SIZE && (end % 2) == 0 && end != isize - 1) {
		printk_ratelimited(KERN_DEBUG
		    "btrfs: %s: ino %llu isize %llu odd range [%llu,%llu]\n",
90
				caller, btrfs_ino(inode), isize, start, end);
91 92
	}
}
93 94 95 96
#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)
97
#define btrfs_debug_check_extent_io_range(c, s, e)	do {} while (0)
C
Chris Mason 已提交
98
#endif
99 100 101 102 103 104 105 106 107 108 109 110 111

#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;
112
	unsigned long bio_flags;
113 114 115 116

	/* tells writepage not to lock the state bits for this range
	 * it still does the unlocking
	 */
117 118 119 120
	unsigned int extent_locked:1;

	/* tells the submit_bio code to use a WRITE_SYNC */
	unsigned int sync_io:1;
121 122
};

123
static noinline void flush_write_bio(void *data);
124 125 126 127 128
static inline struct btrfs_fs_info *
tree_fs_info(struct extent_io_tree *tree)
{
	return btrfs_sb(tree->mapping->host->i_sb);
}
129

130 131
int __init extent_io_init(void)
{
D
David Sterba 已提交
132
	extent_state_cache = kmem_cache_create("btrfs_extent_state",
133 134
			sizeof(struct extent_state), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
135 136 137
	if (!extent_state_cache)
		return -ENOMEM;

D
David Sterba 已提交
138
	extent_buffer_cache = kmem_cache_create("btrfs_extent_buffer",
139 140
			sizeof(struct extent_buffer), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
141 142
	if (!extent_buffer_cache)
		goto free_state_cache;
143 144 145 146 147

	btrfs_bioset = bioset_create(BIO_POOL_SIZE,
				     offsetof(struct btrfs_io_bio, bio));
	if (!btrfs_bioset)
		goto free_buffer_cache;
148 149 150 151

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

152 153
	return 0;

154 155 156 157
free_bioset:
	bioset_free(btrfs_bioset);
	btrfs_bioset = NULL;

158 159 160 161
free_buffer_cache:
	kmem_cache_destroy(extent_buffer_cache);
	extent_buffer_cache = NULL;

162 163
free_state_cache:
	kmem_cache_destroy(extent_state_cache);
164
	extent_state_cache = NULL;
165 166 167 168 169
	return -ENOMEM;
}

void extent_io_exit(void)
{
170
	btrfs_leak_debug_check();
171 172 173 174 175 176

	/*
	 * Make sure all delayed rcu free are flushed before we
	 * destroy caches.
	 */
	rcu_barrier();
177 178 179 180
	if (extent_state_cache)
		kmem_cache_destroy(extent_state_cache);
	if (extent_buffer_cache)
		kmem_cache_destroy(extent_buffer_cache);
181 182
	if (btrfs_bioset)
		bioset_free(btrfs_bioset);
183 184 185
}

void extent_io_tree_init(struct extent_io_tree *tree,
186
			 struct address_space *mapping)
187
{
188
	tree->state = RB_ROOT;
189
	INIT_RADIX_TREE(&tree->buffer, GFP_ATOMIC);
190 191
	tree->ops = NULL;
	tree->dirty_bytes = 0;
192
	spin_lock_init(&tree->lock);
193
	spin_lock_init(&tree->buffer_lock);
194 195 196
	tree->mapping = mapping;
}

197
static struct extent_state *alloc_extent_state(gfp_t mask)
198 199 200 201
{
	struct extent_state *state;

	state = kmem_cache_alloc(extent_state_cache, mask);
202
	if (!state)
203 204 205
		return state;
	state->state = 0;
	state->private = 0;
206
	state->tree = NULL;
207
	btrfs_leak_debug_add(&state->leak_list, &states);
208 209
	atomic_set(&state->refs, 1);
	init_waitqueue_head(&state->wq);
210
	trace_alloc_extent_state(state, mask, _RET_IP_);
211 212 213
	return state;
}

214
void free_extent_state(struct extent_state *state)
215 216 217 218
{
	if (!state)
		return;
	if (atomic_dec_and_test(&state->refs)) {
219
		WARN_ON(state->tree);
220
		btrfs_leak_debug_del(&state->leak_list);
221
		trace_free_extent_state(state, _RET_IP_);
222 223 224 225 226 227 228
		kmem_cache_free(extent_state_cache, state);
	}
}

static struct rb_node *tree_insert(struct rb_root *root, u64 offset,
				   struct rb_node *node)
{
C
Chris Mason 已提交
229 230
	struct rb_node **p = &root->rb_node;
	struct rb_node *parent = NULL;
231 232
	struct tree_entry *entry;

C
Chris Mason 已提交
233
	while (*p) {
234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249
		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;
	}

	rb_link_node(node, parent, p);
	rb_insert_color(node, root);
	return NULL;
}

250
static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset,
251 252 253
				     struct rb_node **prev_ret,
				     struct rb_node **next_ret)
{
254
	struct rb_root *root = &tree->state;
C
Chris Mason 已提交
255
	struct rb_node *n = root->rb_node;
256 257 258 259 260
	struct rb_node *prev = NULL;
	struct rb_node *orig_prev = NULL;
	struct tree_entry *entry;
	struct tree_entry *prev_entry = NULL;

C
Chris Mason 已提交
261
	while (n) {
262 263 264 265 266 267 268 269
		entry = rb_entry(n, struct tree_entry, rb_node);
		prev = n;
		prev_entry = entry;

		if (offset < entry->start)
			n = n->rb_left;
		else if (offset > entry->end)
			n = n->rb_right;
C
Chris Mason 已提交
270
		else
271 272 273 274 275
			return n;
	}

	if (prev_ret) {
		orig_prev = prev;
C
Chris Mason 已提交
276
		while (prev && offset > prev_entry->end) {
277 278 279 280 281 282 283 284 285
			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
286
		while (prev && offset < prev_entry->start) {
287 288 289 290 291 292 293 294
			prev = rb_prev(prev);
			prev_entry = rb_entry(prev, struct tree_entry, rb_node);
		}
		*next_ret = prev;
	}
	return NULL;
}

295 296
static inline struct rb_node *tree_search(struct extent_io_tree *tree,
					  u64 offset)
297
{
298
	struct rb_node *prev = NULL;
299
	struct rb_node *ret;
300

301
	ret = __etree_search(tree, offset, &prev, NULL);
C
Chris Mason 已提交
302
	if (!ret)
303 304 305 306
		return prev;
	return ret;
}

J
Josef Bacik 已提交
307 308 309 310 311 312 313 314
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);
}

315 316 317 318 319 320 321 322 323
/*
 * 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.
 */
324 325
static void merge_state(struct extent_io_tree *tree,
		        struct extent_state *state)
326 327 328 329
{
	struct extent_state *other;
	struct rb_node *other_node;

330
	if (state->state & (EXTENT_IOBITS | EXTENT_BOUNDARY))
331
		return;
332 333 334 335 336 337

	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
338
			merge_cb(tree, state, other);
339
			state->start = other->start;
340
			other->tree = NULL;
341 342 343 344 345 346 347 348 349
			rb_erase(&other->rb_node, &tree->state);
			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
350
			merge_cb(tree, state, other);
351 352 353 354
			state->end = other->end;
			other->tree = NULL;
			rb_erase(&other->rb_node, &tree->state);
			free_extent_state(other);
355 356 357 358
		}
	}
}

359
static void set_state_cb(struct extent_io_tree *tree,
360
			 struct extent_state *state, unsigned long *bits)
361
{
362 363
	if (tree->ops && tree->ops->set_bit_hook)
		tree->ops->set_bit_hook(tree->mapping->host, state, bits);
364 365 366
}

static void clear_state_cb(struct extent_io_tree *tree,
367
			   struct extent_state *state, unsigned long *bits)
368
{
J
Josef Bacik 已提交
369 370
	if (tree->ops && tree->ops->clear_bit_hook)
		tree->ops->clear_bit_hook(tree->mapping->host, state, bits);
371 372
}

373
static void set_state_bits(struct extent_io_tree *tree,
374
			   struct extent_state *state, unsigned long *bits);
375

376 377 378 379 380 381 382 383 384 385 386 387
/*
 * 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,
388
			unsigned long *bits)
389 390 391
{
	struct rb_node *node;

J
Julia Lawall 已提交
392 393
	if (end < start)
		WARN(1, KERN_ERR "btrfs end < start %llu %llu\n",
394
		       end, start);
395 396
	state->start = start;
	state->end = end;
J
Josef Bacik 已提交
397

398 399
	set_state_bits(tree, state, bits);

400 401 402 403
	node = tree_insert(&tree->state, end, &state->rb_node);
	if (node) {
		struct extent_state *found;
		found = rb_entry(node, struct extent_state, rb_node);
C
Chris Mason 已提交
404
		printk(KERN_ERR "btrfs found node %llu %llu on insert of "
405 406
		       "%llu %llu\n",
		       found->start, found->end, start, end);
407 408
		return -EEXIST;
	}
409
	state->tree = tree;
410 411 412 413
	merge_state(tree, state);
	return 0;
}

414
static void split_cb(struct extent_io_tree *tree, struct extent_state *orig,
J
Josef Bacik 已提交
415 416 417
		     u64 split)
{
	if (tree->ops && tree->ops->split_extent_hook)
418
		tree->ops->split_extent_hook(tree->mapping->host, orig, split);
J
Josef Bacik 已提交
419 420
}

421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438
/*
 * 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
439 440 441

	split_cb(tree, orig, split);

442 443 444 445 446 447 448 449 450 451
	prealloc->start = orig->start;
	prealloc->end = split - 1;
	prealloc->state = orig->state;
	orig->start = split;

	node = tree_insert(&tree->state, prealloc->end, &prealloc->rb_node);
	if (node) {
		free_extent_state(prealloc);
		return -EEXIST;
	}
452
	prealloc->tree = tree;
453 454 455
	return 0;
}

456 457 458 459 460 461 462 463 464
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;
}

465 466
/*
 * utility function to clear some bits in an extent state struct.
467
 * it will optionally wake up any one waiting on this state (wake == 1).
468 469 470 471
 *
 * If no bits are set on the state struct after clearing things, the
 * struct is freed and removed from the tree
 */
472 473
static struct extent_state *clear_state_bit(struct extent_io_tree *tree,
					    struct extent_state *state,
474
					    unsigned long *bits, int wake)
475
{
476
	struct extent_state *next;
477
	unsigned long bits_to_clear = *bits & ~EXTENT_CTLBITS;
478

479
	if ((bits_to_clear & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) {
480 481 482 483
		u64 range = state->end - state->start + 1;
		WARN_ON(range > tree->dirty_bytes);
		tree->dirty_bytes -= range;
	}
484
	clear_state_cb(tree, state, bits);
485
	state->state &= ~bits_to_clear;
486 487
	if (wake)
		wake_up(&state->wq);
488
	if (state->state == 0) {
489
		next = next_state(state);
490
		if (state->tree) {
491
			rb_erase(&state->rb_node, &tree->state);
492
			state->tree = NULL;
493 494 495 496 497 498
			free_extent_state(state);
		} else {
			WARN_ON(1);
		}
	} else {
		merge_state(tree, state);
499
		next = next_state(state);
500
	}
501
	return next;
502 503
}

504 505 506 507 508 509 510 511 512
static struct extent_state *
alloc_extent_state_atomic(struct extent_state *prealloc)
{
	if (!prealloc)
		prealloc = alloc_extent_state(GFP_ATOMIC);

	return prealloc;
}

513
static void extent_io_tree_panic(struct extent_io_tree *tree, int err)
514 515 516 517 518 519
{
	btrfs_panic(tree_fs_info(tree), err, "Locking error: "
		    "Extent tree was modified by another "
		    "thread while locked.");
}

520 521 522 523 524 525 526 527 528 529
/*
 * 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.
 *
530
 * This takes the tree lock, and returns 0 on success and < 0 on error.
531 532
 */
int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
533
		     unsigned long bits, int wake, int delete,
534 535
		     struct extent_state **cached_state,
		     gfp_t mask)
536 537
{
	struct extent_state *state;
538
	struct extent_state *cached;
539 540
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
541
	u64 last_end;
542
	int err;
543
	int clear = 0;
544

545 546
	btrfs_debug_check_extent_io_range(tree->mapping->host, start, end);

547 548 549
	if (bits & EXTENT_DELALLOC)
		bits |= EXTENT_NORESERVE;

550 551 552 553
	if (delete)
		bits |= ~EXTENT_CTLBITS;
	bits |= EXTENT_FIRST_DELALLOC;

554 555
	if (bits & (EXTENT_IOBITS | EXTENT_BOUNDARY))
		clear = 1;
556 557 558 559 560 561 562
again:
	if (!prealloc && (mask & __GFP_WAIT)) {
		prealloc = alloc_extent_state(mask);
		if (!prealloc)
			return -ENOMEM;
	}

563
	spin_lock(&tree->lock);
564 565
	if (cached_state) {
		cached = *cached_state;
566 567 568 569 570 571

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

572 573
		if (cached && cached->tree && cached->start <= start &&
		    cached->end > start) {
574 575
			if (clear)
				atomic_dec(&cached->refs);
576
			state = cached;
577
			goto hit_next;
578
		}
579 580
		if (clear)
			free_extent_state(cached);
581
	}
582 583 584 585
	/*
	 * this search will find the extents that end after
	 * our range starts
	 */
586
	node = tree_search(tree, start);
587 588 589
	if (!node)
		goto out;
	state = rb_entry(node, struct extent_state, rb_node);
590
hit_next:
591 592 593
	if (state->start > end)
		goto out;
	WARN_ON(state->end < start);
594
	last_end = state->end;
595

596
	/* the state doesn't have the wanted bits, go ahead */
597 598
	if (!(state->state & bits)) {
		state = next_state(state);
599
		goto next;
600
	}
601

602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618
	/*
	 *     | ---- 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) {
619 620
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
621
		err = split_state(tree, state, prealloc, start);
622 623 624
		if (err)
			extent_io_tree_panic(tree, err);

625 626 627 628
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
629 630
			state = clear_state_bit(tree, state, &bits, wake);
			goto next;
631 632 633 634 635 636 637 638 639 640
		}
		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) {
641 642
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
643
		err = split_state(tree, state, prealloc, end + 1);
644 645 646
		if (err)
			extent_io_tree_panic(tree, err);

647 648
		if (wake)
			wake_up(&state->wq);
649

650
		clear_state_bit(tree, prealloc, &bits, wake);
J
Josef Bacik 已提交
651

652 653 654
		prealloc = NULL;
		goto out;
	}
655

656
	state = clear_state_bit(tree, state, &bits, wake);
657
next:
658 659 660
	if (last_end == (u64)-1)
		goto out;
	start = last_end + 1;
661
	if (start <= end && state && !need_resched())
662
		goto hit_next;
663 664 665
	goto search_again;

out:
666
	spin_unlock(&tree->lock);
667 668 669
	if (prealloc)
		free_extent_state(prealloc);

670
	return 0;
671 672 673 674

search_again:
	if (start > end)
		goto out;
675
	spin_unlock(&tree->lock);
676 677 678 679 680
	if (mask & __GFP_WAIT)
		cond_resched();
	goto again;
}

681 682
static void wait_on_state(struct extent_io_tree *tree,
			  struct extent_state *state)
683 684
		__releases(tree->lock)
		__acquires(tree->lock)
685 686 687
{
	DEFINE_WAIT(wait);
	prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE);
688
	spin_unlock(&tree->lock);
689
	schedule();
690
	spin_lock(&tree->lock);
691 692 693 694 695 696 697 698
	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
 */
699 700
static void wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
			    unsigned long bits)
701 702 703 704
{
	struct extent_state *state;
	struct rb_node *node;

705 706
	btrfs_debug_check_extent_io_range(tree->mapping->host, start, end);

707
	spin_lock(&tree->lock);
708 709 710 711 712 713
again:
	while (1) {
		/*
		 * this search will find all the extents that end after
		 * our range starts
		 */
714
		node = tree_search(tree, start);
715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734
		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;

735
		cond_resched_lock(&tree->lock);
736 737
	}
out:
738
	spin_unlock(&tree->lock);
739 740
}

741
static void set_state_bits(struct extent_io_tree *tree,
742
			   struct extent_state *state,
743
			   unsigned long *bits)
744
{
745
	unsigned long bits_to_set = *bits & ~EXTENT_CTLBITS;
J
Josef Bacik 已提交
746

747
	set_state_cb(tree, state, bits);
748
	if ((bits_to_set & EXTENT_DIRTY) && !(state->state & EXTENT_DIRTY)) {
749 750 751
		u64 range = state->end - state->start + 1;
		tree->dirty_bytes += range;
	}
752
	state->state |= bits_to_set;
753 754
}

755 756 757 758 759 760 761 762 763 764 765
static void cache_state(struct extent_state *state,
			struct extent_state **cached_ptr)
{
	if (cached_ptr && !(*cached_ptr)) {
		if (state->state & (EXTENT_IOBITS | EXTENT_BOUNDARY)) {
			*cached_ptr = state;
			atomic_inc(&state->refs);
		}
	}
}

766
/*
767 768
 * 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.
769
 *
770 771 772
 * 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.
773
 *
774
 * [start, end] is inclusive This takes the tree lock.
775
 */
776

J
Jeff Mahoney 已提交
777 778
static int __must_check
__set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
779 780 781
		 unsigned long bits, unsigned long exclusive_bits,
		 u64 *failed_start, struct extent_state **cached_state,
		 gfp_t mask)
782 783 784 785 786 787 788
{
	struct extent_state *state;
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
	int err = 0;
	u64 last_start;
	u64 last_end;
789

790 791
	btrfs_debug_check_extent_io_range(tree->mapping->host, start, end);

792
	bits |= EXTENT_FIRST_DELALLOC;
793 794 795
again:
	if (!prealloc && (mask & __GFP_WAIT)) {
		prealloc = alloc_extent_state(mask);
796
		BUG_ON(!prealloc);
797 798
	}

799
	spin_lock(&tree->lock);
800 801
	if (cached_state && *cached_state) {
		state = *cached_state;
802 803
		if (state->start <= start && state->end > start &&
		    state->tree) {
804 805 806 807
			node = &state->rb_node;
			goto hit_next;
		}
	}
808 809 810 811
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
812
	node = tree_search(tree, start);
813
	if (!node) {
814 815
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
816
		err = insert_state(tree, prealloc, start, end, &bits);
817 818 819
		if (err)
			extent_io_tree_panic(tree, err);

820 821 822 823
		prealloc = NULL;
		goto out;
	}
	state = rb_entry(node, struct extent_state, rb_node);
C
Chris Mason 已提交
824
hit_next:
825 826 827 828 829 830 831 832 833 834
	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) {
835
		if (state->state & exclusive_bits) {
836 837 838 839
			*failed_start = state->start;
			err = -EEXIST;
			goto out;
		}
840

841
		set_state_bits(tree, state, &bits);
842
		cache_state(state, cached_state);
843
		merge_state(tree, state);
844 845 846
		if (last_end == (u64)-1)
			goto out;
		start = last_end + 1;
847 848 849 850
		state = next_state(state);
		if (start < end && state && state->start == start &&
		    !need_resched())
			goto hit_next;
851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870
		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) {
871
		if (state->state & exclusive_bits) {
872 873 874 875
			*failed_start = start;
			err = -EEXIST;
			goto out;
		}
876 877 878

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
879
		err = split_state(tree, state, prealloc, start);
880 881 882
		if (err)
			extent_io_tree_panic(tree, err);

883 884 885 886
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
887
			set_state_bits(tree, state, &bits);
888
			cache_state(state, cached_state);
889
			merge_state(tree, state);
890 891 892
			if (last_end == (u64)-1)
				goto out;
			start = last_end + 1;
893 894 895 896
			state = next_state(state);
			if (start < end && state && state->start == start &&
			    !need_resched())
				goto hit_next;
897 898 899 900 901 902 903 904 905 906 907 908 909 910 911
		}
		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
912
			this_end = last_start - 1;
913 914 915

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
916 917 918 919 920

		/*
		 * Avoid to free 'prealloc' if it can be merged with
		 * the later extent.
		 */
921
		err = insert_state(tree, prealloc, start, this_end,
922
				   &bits);
923 924 925
		if (err)
			extent_io_tree_panic(tree, err);

J
Josef Bacik 已提交
926 927
		cache_state(prealloc, cached_state);
		prealloc = NULL;
928 929 930 931 932 933 934 935 936 937
		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) {
938
		if (state->state & exclusive_bits) {
939 940 941 942
			*failed_start = start;
			err = -EEXIST;
			goto out;
		}
943 944 945

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
946
		err = split_state(tree, state, prealloc, end + 1);
947 948
		if (err)
			extent_io_tree_panic(tree, err);
949

950
		set_state_bits(tree, prealloc, &bits);
951
		cache_state(prealloc, cached_state);
952 953 954 955 956 957 958 959
		merge_state(tree, prealloc);
		prealloc = NULL;
		goto out;
	}

	goto search_again;

out:
960
	spin_unlock(&tree->lock);
961 962 963 964 965 966 967 968
	if (prealloc)
		free_extent_state(prealloc);

	return err;

search_again:
	if (start > end)
		goto out;
969
	spin_unlock(&tree->lock);
970 971 972 973 974
	if (mask & __GFP_WAIT)
		cond_resched();
	goto again;
}

975 976 977
int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
		   unsigned long bits, u64 * failed_start,
		   struct extent_state **cached_state, gfp_t mask)
J
Jeff Mahoney 已提交
978 979 980 981 982 983
{
	return __set_extent_bit(tree, start, end, bits, 0, failed_start,
				cached_state, mask);
}


J
Josef Bacik 已提交
984
/**
L
Liu Bo 已提交
985 986
 * convert_extent_bit - convert all bits in a given range from one bit to
 * 			another
J
Josef Bacik 已提交
987 988 989 990 991
 * @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
992
 * @cached_state:	state that we're going to cache
J
Josef Bacik 已提交
993 994 995 996 997 998 999 1000 1001
 * @mask:	the allocation mask
 *
 * This will go through and set bits for the given range.  If any states exist
 * already in this range they are set with the given bit and cleared of the
 * clear_bits.  This is only meant to be used by things that are mergeable, ie
 * converting from say DELALLOC to DIRTY.  This is not meant to be used with
 * boundary bits like LOCK.
 */
int convert_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
1002
		       unsigned long bits, unsigned long clear_bits,
1003
		       struct extent_state **cached_state, gfp_t mask)
J
Josef Bacik 已提交
1004 1005 1006 1007 1008 1009 1010 1011
{
	struct extent_state *state;
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
	int err = 0;
	u64 last_start;
	u64 last_end;

1012 1013
	btrfs_debug_check_extent_io_range(tree->mapping->host, start, end);

J
Josef Bacik 已提交
1014 1015 1016 1017 1018 1019 1020 1021
again:
	if (!prealloc && (mask & __GFP_WAIT)) {
		prealloc = alloc_extent_state(mask);
		if (!prealloc)
			return -ENOMEM;
	}

	spin_lock(&tree->lock);
1022 1023 1024 1025 1026 1027 1028 1029 1030
	if (cached_state && *cached_state) {
		state = *cached_state;
		if (state->start <= start && state->end > start &&
		    state->tree) {
			node = &state->rb_node;
			goto hit_next;
		}
	}

J
Josef Bacik 已提交
1031 1032 1033 1034 1035 1036 1037
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
	node = tree_search(tree, start);
	if (!node) {
		prealloc = alloc_extent_state_atomic(prealloc);
1038 1039 1040 1041
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1042 1043
		err = insert_state(tree, prealloc, start, end, &bits);
		prealloc = NULL;
1044 1045
		if (err)
			extent_io_tree_panic(tree, err);
J
Josef Bacik 已提交
1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060
		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) {
		set_state_bits(tree, state, &bits);
1061
		cache_state(state, cached_state);
1062
		state = clear_state_bit(tree, state, &clear_bits, 0);
J
Josef Bacik 已提交
1063 1064 1065
		if (last_end == (u64)-1)
			goto out;
		start = last_end + 1;
1066 1067 1068
		if (start < end && state && state->start == start &&
		    !need_resched())
			goto hit_next;
J
Josef Bacik 已提交
1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089
		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);
1090 1091 1092 1093
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1094
		err = split_state(tree, state, prealloc, start);
1095 1096
		if (err)
			extent_io_tree_panic(tree, err);
J
Josef Bacik 已提交
1097 1098 1099 1100 1101
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
			set_state_bits(tree, state, &bits);
1102
			cache_state(state, cached_state);
1103
			state = clear_state_bit(tree, state, &clear_bits, 0);
J
Josef Bacik 已提交
1104 1105 1106
			if (last_end == (u64)-1)
				goto out;
			start = last_end + 1;
1107 1108 1109
			if (start < end && state && state->start == start &&
			    !need_resched())
				goto hit_next;
J
Josef Bacik 已提交
1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127
		}
		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);
1128 1129 1130 1131
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1132 1133 1134 1135 1136 1137 1138

		/*
		 * Avoid to free 'prealloc' if it can be merged with
		 * the later extent.
		 */
		err = insert_state(tree, prealloc, start, this_end,
				   &bits);
1139 1140
		if (err)
			extent_io_tree_panic(tree, err);
1141
		cache_state(prealloc, cached_state);
J
Josef Bacik 已提交
1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153
		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);
1154 1155 1156 1157
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1158 1159

		err = split_state(tree, state, prealloc, end + 1);
1160 1161
		if (err)
			extent_io_tree_panic(tree, err);
J
Josef Bacik 已提交
1162 1163

		set_state_bits(tree, prealloc, &bits);
1164
		cache_state(prealloc, cached_state);
J
Josef Bacik 已提交
1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187
		clear_state_bit(tree, prealloc, &clear_bits, 0);
		prealloc = NULL;
		goto out;
	}

	goto search_again;

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

	return err;

search_again:
	if (start > end)
		goto out;
	spin_unlock(&tree->lock);
	if (mask & __GFP_WAIT)
		cond_resched();
	goto again;
}

1188 1189 1190 1191
/* wrappers around set/clear extent bit */
int set_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
		     gfp_t mask)
{
J
Jeff Mahoney 已提交
1192
	return set_extent_bit(tree, start, end, EXTENT_DIRTY, NULL,
1193
			      NULL, mask);
1194 1195 1196
}

int set_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1197
		    unsigned long bits, gfp_t mask)
1198
{
J
Jeff Mahoney 已提交
1199
	return set_extent_bit(tree, start, end, bits, NULL,
1200
			      NULL, mask);
1201 1202 1203
}

int clear_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1204
		      unsigned long bits, gfp_t mask)
1205
{
1206
	return clear_extent_bit(tree, start, end, bits, 0, 0, NULL, mask);
1207 1208 1209
}

int set_extent_delalloc(struct extent_io_tree *tree, u64 start, u64 end,
1210
			struct extent_state **cached_state, gfp_t mask)
1211 1212
{
	return set_extent_bit(tree, start, end,
1213
			      EXTENT_DELALLOC | EXTENT_UPTODATE,
J
Jeff Mahoney 已提交
1214
			      NULL, cached_state, mask);
1215 1216
}

1217 1218 1219 1220 1221 1222 1223 1224
int set_extent_defrag(struct extent_io_tree *tree, u64 start, u64 end,
		      struct extent_state **cached_state, gfp_t mask)
{
	return set_extent_bit(tree, start, end,
			      EXTENT_DELALLOC | EXTENT_UPTODATE | EXTENT_DEFRAG,
			      NULL, cached_state, mask);
}

1225 1226 1227 1228
int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
		       gfp_t mask)
{
	return clear_extent_bit(tree, start, end,
1229
				EXTENT_DIRTY | EXTENT_DELALLOC |
1230
				EXTENT_DO_ACCOUNTING, 0, 0, NULL, mask);
1231 1232 1233 1234 1235
}

int set_extent_new(struct extent_io_tree *tree, u64 start, u64 end,
		     gfp_t mask)
{
J
Jeff Mahoney 已提交
1236
	return set_extent_bit(tree, start, end, EXTENT_NEW, NULL,
1237
			      NULL, mask);
1238 1239 1240
}

int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end,
1241
			struct extent_state **cached_state, gfp_t mask)
1242
{
L
Liu Bo 已提交
1243
	return set_extent_bit(tree, start, end, EXTENT_UPTODATE, NULL,
J
Jeff Mahoney 已提交
1244
			      cached_state, mask);
1245 1246
}

1247 1248
int clear_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end,
			  struct extent_state **cached_state, gfp_t mask)
1249
{
1250
	return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0,
1251
				cached_state, mask);
1252 1253
}

C
Chris Mason 已提交
1254 1255 1256 1257
/*
 * either insert or lock state struct between start and end use mask to tell
 * us if waiting is desired.
 */
1258
int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1259
		     unsigned long bits, struct extent_state **cached_state)
1260 1261 1262 1263
{
	int err;
	u64 failed_start;
	while (1) {
J
Jeff Mahoney 已提交
1264 1265 1266
		err = __set_extent_bit(tree, start, end, EXTENT_LOCKED | bits,
				       EXTENT_LOCKED, &failed_start,
				       cached_state, GFP_NOFS);
1267
		if (err == -EEXIST) {
1268 1269
			wait_extent_bit(tree, failed_start, end, EXTENT_LOCKED);
			start = failed_start;
1270
		} else
1271 1272 1273 1274 1275 1276
			break;
		WARN_ON(start > end);
	}
	return err;
}

1277
int lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
1278
{
1279
	return lock_extent_bits(tree, start, end, 0, NULL);
1280 1281
}

1282
int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
1283 1284 1285 1286
{
	int err;
	u64 failed_start;

J
Jeff Mahoney 已提交
1287 1288
	err = __set_extent_bit(tree, start, end, EXTENT_LOCKED, EXTENT_LOCKED,
			       &failed_start, NULL, GFP_NOFS);
Y
Yan Zheng 已提交
1289 1290 1291
	if (err == -EEXIST) {
		if (failed_start > start)
			clear_extent_bit(tree, start, failed_start - 1,
1292
					 EXTENT_LOCKED, 1, 0, NULL, GFP_NOFS);
1293
		return 0;
Y
Yan Zheng 已提交
1294
	}
1295 1296 1297
	return 1;
}

1298 1299 1300 1301 1302 1303 1304
int unlock_extent_cached(struct extent_io_tree *tree, u64 start, u64 end,
			 struct extent_state **cached, gfp_t mask)
{
	return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, cached,
				mask);
}

1305
int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end)
1306
{
1307
	return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL,
1308
				GFP_NOFS);
1309 1310
}

1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343
int extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end)
{
	unsigned long index = start >> PAGE_CACHE_SHIFT;
	unsigned long end_index = end >> PAGE_CACHE_SHIFT;
	struct page *page;

	while (index <= end_index) {
		page = find_get_page(inode->i_mapping, index);
		BUG_ON(!page); /* Pages should be in the extent_io_tree */
		clear_page_dirty_for_io(page);
		page_cache_release(page);
		index++;
	}
	return 0;
}

int extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end)
{
	unsigned long index = start >> PAGE_CACHE_SHIFT;
	unsigned long end_index = end >> PAGE_CACHE_SHIFT;
	struct page *page;

	while (index <= end_index) {
		page = find_get_page(inode->i_mapping, index);
		BUG_ON(!page); /* Pages should be in the extent_io_tree */
		account_page_redirty(page);
		__set_page_dirty_nobuffers(page);
		page_cache_release(page);
		index++;
	}
	return 0;
}

1344 1345 1346
/*
 * helper function to set both pages and extents in the tree writeback
 */
1347
static int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end)
1348 1349 1350 1351 1352 1353 1354
{
	unsigned long index = start >> PAGE_CACHE_SHIFT;
	unsigned long end_index = end >> PAGE_CACHE_SHIFT;
	struct page *page;

	while (index <= end_index) {
		page = find_get_page(tree->mapping, index);
1355
		BUG_ON(!page); /* Pages should be in the extent_io_tree */
1356 1357 1358 1359 1360 1361 1362
		set_page_writeback(page);
		page_cache_release(page);
		index++;
	}
	return 0;
}

C
Chris Mason 已提交
1363 1364 1365 1366
/* 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'
 */
1367 1368
static struct extent_state *
find_first_extent_bit_state(struct extent_io_tree *tree,
1369
			    u64 start, unsigned long bits)
C
Chris Mason 已提交
1370 1371 1372 1373 1374 1375 1376 1377 1378
{
	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
1379
	if (!node)
C
Chris Mason 已提交
1380 1381
		goto out;

C
Chris Mason 已提交
1382
	while (1) {
C
Chris Mason 已提交
1383
		state = rb_entry(node, struct extent_state, rb_node);
C
Chris Mason 已提交
1384
		if (state->end >= start && (state->state & bits))
C
Chris Mason 已提交
1385
			return state;
C
Chris Mason 已提交
1386

C
Chris Mason 已提交
1387 1388 1389 1390 1391 1392 1393 1394
		node = rb_next(node);
		if (!node)
			break;
	}
out:
	return NULL;
}

1395 1396 1397 1398 1399
/*
 * 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.
 *
1400
 * If nothing was found, 1 is returned. If found something, return 0.
1401 1402
 */
int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
1403
			  u64 *start_ret, u64 *end_ret, unsigned long bits,
1404
			  struct extent_state **cached_state)
1405 1406
{
	struct extent_state *state;
1407
	struct rb_node *n;
1408 1409 1410
	int ret = 1;

	spin_lock(&tree->lock);
1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429
	if (cached_state && *cached_state) {
		state = *cached_state;
		if (state->end == start - 1 && state->tree) {
			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;
	}

1430
	state = find_first_extent_bit_state(tree, start, bits);
1431
got_it:
1432
	if (state) {
1433
		cache_state(state, cached_state);
1434 1435 1436 1437
		*start_ret = state->start;
		*end_ret = state->end;
		ret = 0;
	}
1438
out:
1439 1440 1441 1442
	spin_unlock(&tree->lock);
	return ret;
}

C
Chris Mason 已提交
1443 1444 1445 1446 1447 1448
/*
 * 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
1449
static noinline u64 find_delalloc_range(struct extent_io_tree *tree,
1450 1451
					u64 *start, u64 *end, u64 max_bytes,
					struct extent_state **cached_state)
1452 1453 1454 1455 1456 1457 1458
{
	struct rb_node *node;
	struct extent_state *state;
	u64 cur_start = *start;
	u64 found = 0;
	u64 total_bytes = 0;

1459
	spin_lock(&tree->lock);
C
Chris Mason 已提交
1460

1461 1462 1463 1464
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1465
	node = tree_search(tree, cur_start);
1466
	if (!node) {
1467 1468
		if (!found)
			*end = (u64)-1;
1469 1470 1471
		goto out;
	}

C
Chris Mason 已提交
1472
	while (1) {
1473
		state = rb_entry(node, struct extent_state, rb_node);
1474 1475
		if (found && (state->start != cur_start ||
			      (state->state & EXTENT_BOUNDARY))) {
1476 1477 1478 1479 1480 1481 1482
			goto out;
		}
		if (!(state->state & EXTENT_DELALLOC)) {
			if (!found)
				*end = state->end;
			goto out;
		}
1483
		if (!found) {
1484
			*start = state->start;
1485 1486 1487
			*cached_state = state;
			atomic_inc(&state->refs);
		}
1488 1489 1490 1491 1492
		found++;
		*end = state->end;
		cur_start = state->end + 1;
		node = rb_next(node);
		total_bytes += state->end - state->start + 1;
1493
		if (total_bytes >= max_bytes)
1494 1495
			break;
		if (!node)
1496 1497 1498
			break;
	}
out:
1499
	spin_unlock(&tree->lock);
1500 1501 1502
	return found;
}

1503 1504 1505
static noinline void __unlock_for_delalloc(struct inode *inode,
					   struct page *locked_page,
					   u64 start, u64 end)
C
Chris Mason 已提交
1506 1507 1508 1509 1510 1511 1512 1513 1514
{
	int ret;
	struct page *pages[16];
	unsigned long index = start >> PAGE_CACHE_SHIFT;
	unsigned long end_index = end >> PAGE_CACHE_SHIFT;
	unsigned long nr_pages = end_index - index + 1;
	int i;

	if (index == locked_page->index && end_index == index)
1515
		return;
C
Chris Mason 已提交
1516

C
Chris Mason 已提交
1517
	while (nr_pages > 0) {
C
Chris Mason 已提交
1518
		ret = find_get_pages_contig(inode->i_mapping, index,
1519 1520
				     min_t(unsigned long, nr_pages,
				     ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551
		for (i = 0; i < ret; i++) {
			if (pages[i] != locked_page)
				unlock_page(pages[i]);
			page_cache_release(pages[i]);
		}
		nr_pages -= ret;
		index += ret;
		cond_resched();
	}
}

static noinline int lock_delalloc_pages(struct inode *inode,
					struct page *locked_page,
					u64 delalloc_start,
					u64 delalloc_end)
{
	unsigned long index = delalloc_start >> PAGE_CACHE_SHIFT;
	unsigned long start_index = index;
	unsigned long end_index = delalloc_end >> PAGE_CACHE_SHIFT;
	unsigned long pages_locked = 0;
	struct page *pages[16];
	unsigned long nrpages;
	int ret;
	int i;

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

	/* skip the page at the start index */
	nrpages = end_index - index + 1;
C
Chris Mason 已提交
1552
	while (nrpages > 0) {
C
Chris Mason 已提交
1553
		ret = find_get_pages_contig(inode->i_mapping, index,
1554 1555
				     min_t(unsigned long,
				     nrpages, ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1556 1557 1558 1559 1560 1561 1562 1563 1564 1565
		if (ret == 0) {
			ret = -EAGAIN;
			goto done;
		}
		/* now we have an array of pages, lock them all */
		for (i = 0; i < ret; i++) {
			/*
			 * the caller is taking responsibility for
			 * locked_page
			 */
1566
			if (pages[i] != locked_page) {
C
Chris Mason 已提交
1567
				lock_page(pages[i]);
1568 1569
				if (!PageDirty(pages[i]) ||
				    pages[i]->mapping != inode->i_mapping) {
1570 1571 1572 1573 1574 1575
					ret = -EAGAIN;
					unlock_page(pages[i]);
					page_cache_release(pages[i]);
					goto done;
				}
			}
C
Chris Mason 已提交
1576
			page_cache_release(pages[i]);
1577
			pages_locked++;
C
Chris Mason 已提交
1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608
		}
		nrpages -= ret;
		index += ret;
		cond_resched();
	}
	ret = 0;
done:
	if (ret && pages_locked) {
		__unlock_for_delalloc(inode, locked_page,
			      delalloc_start,
			      ((u64)(start_index + pages_locked - 1)) <<
			      PAGE_CACHE_SHIFT);
	}
	return ret;
}

/*
 * find a contiguous range of bytes in the file marked as delalloc, not
 * more than 'max_bytes'.  start and end are used to return the range,
 *
 * 1 is returned if we find something, 0 if nothing was in the tree
 */
static noinline u64 find_lock_delalloc_range(struct inode *inode,
					     struct extent_io_tree *tree,
					     struct page *locked_page,
					     u64 *start, u64 *end,
					     u64 max_bytes)
{
	u64 delalloc_start;
	u64 delalloc_end;
	u64 found;
1609
	struct extent_state *cached_state = NULL;
C
Chris Mason 已提交
1610 1611 1612 1613 1614 1615 1616 1617
	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,
1618
				    max_bytes, &cached_state);
C
Chris Mason 已提交
1619
	if (!found || delalloc_end <= *start) {
C
Chris Mason 已提交
1620 1621
		*start = delalloc_start;
		*end = delalloc_end;
1622
		free_extent_state(cached_state);
L
Liu Bo 已提交
1623
		return 0;
C
Chris Mason 已提交
1624 1625
	}

C
Chris Mason 已提交
1626 1627 1628 1629 1630
	/*
	 * 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
1631
	if (delalloc_start < *start)
C
Chris Mason 已提交
1632 1633
		delalloc_start = *start;

C
Chris Mason 已提交
1634 1635 1636
	/*
	 * make sure to limit the number of pages we try to lock down
	 */
1637 1638
	if (delalloc_end + 1 - delalloc_start > max_bytes)
		delalloc_end = delalloc_start + max_bytes - 1;
C
Chris Mason 已提交
1639

C
Chris Mason 已提交
1640 1641 1642 1643 1644 1645 1646
	/* 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
		 */
1647
		free_extent_state(cached_state);
C
Chris Mason 已提交
1648
		if (!loops) {
1649
			max_bytes = PAGE_CACHE_SIZE;
C
Chris Mason 已提交
1650 1651 1652 1653 1654 1655 1656
			loops = 1;
			goto again;
		} else {
			found = 0;
			goto out_failed;
		}
	}
1657
	BUG_ON(ret); /* Only valid values are 0 and -EAGAIN */
C
Chris Mason 已提交
1658 1659

	/* step three, lock the state bits for the whole range */
1660
	lock_extent_bits(tree, delalloc_start, delalloc_end, 0, &cached_state);
C
Chris Mason 已提交
1661 1662 1663

	/* then test to make sure it is all still delalloc */
	ret = test_range_bit(tree, delalloc_start, delalloc_end,
1664
			     EXTENT_DELALLOC, 1, cached_state);
C
Chris Mason 已提交
1665
	if (!ret) {
1666 1667
		unlock_extent_cached(tree, delalloc_start, delalloc_end,
				     &cached_state, GFP_NOFS);
C
Chris Mason 已提交
1668 1669 1670 1671 1672
		__unlock_for_delalloc(inode, locked_page,
			      delalloc_start, delalloc_end);
		cond_resched();
		goto again;
	}
1673
	free_extent_state(cached_state);
C
Chris Mason 已提交
1674 1675 1676 1677 1678 1679
	*start = delalloc_start;
	*end = delalloc_end;
out_failed:
	return found;
}

1680 1681 1682 1683
int extent_clear_unlock_delalloc(struct inode *inode, u64 start, u64 end,
				 struct page *locked_page,
				 unsigned long clear_bits,
				 unsigned long page_ops)
C
Chris Mason 已提交
1684
{
1685
	struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
C
Chris Mason 已提交
1686 1687 1688 1689 1690 1691
	int ret;
	struct page *pages[16];
	unsigned long index = start >> PAGE_CACHE_SHIFT;
	unsigned long end_index = end >> PAGE_CACHE_SHIFT;
	unsigned long nr_pages = end_index - index + 1;
	int i;
1692

1693
	clear_extent_bit(tree, start, end, clear_bits, 1, 0, NULL, GFP_NOFS);
1694
	if (page_ops == 0)
1695
		return 0;
C
Chris Mason 已提交
1696

C
Chris Mason 已提交
1697
	while (nr_pages > 0) {
C
Chris Mason 已提交
1698
		ret = find_get_pages_contig(inode->i_mapping, index,
1699 1700
				     min_t(unsigned long,
				     nr_pages, ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1701
		for (i = 0; i < ret; i++) {
1702

1703
			if (page_ops & PAGE_SET_PRIVATE2)
1704 1705
				SetPagePrivate2(pages[i]);

C
Chris Mason 已提交
1706 1707 1708 1709
			if (pages[i] == locked_page) {
				page_cache_release(pages[i]);
				continue;
			}
1710
			if (page_ops & PAGE_CLEAR_DIRTY)
C
Chris Mason 已提交
1711
				clear_page_dirty_for_io(pages[i]);
1712
			if (page_ops & PAGE_SET_WRITEBACK)
C
Chris Mason 已提交
1713
				set_page_writeback(pages[i]);
1714
			if (page_ops & PAGE_END_WRITEBACK)
C
Chris Mason 已提交
1715
				end_page_writeback(pages[i]);
1716
			if (page_ops & PAGE_UNLOCK)
1717
				unlock_page(pages[i]);
C
Chris Mason 已提交
1718 1719 1720 1721 1722 1723 1724 1725 1726
			page_cache_release(pages[i]);
		}
		nr_pages -= ret;
		index += ret;
		cond_resched();
	}
	return 0;
}

C
Chris Mason 已提交
1727 1728 1729 1730 1731
/*
 * 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.
 */
1732 1733
u64 count_range_bits(struct extent_io_tree *tree,
		     u64 *start, u64 search_end, u64 max_bytes,
1734
		     unsigned long bits, int contig)
1735 1736 1737 1738 1739
{
	struct rb_node *node;
	struct extent_state *state;
	u64 cur_start = *start;
	u64 total_bytes = 0;
1740
	u64 last = 0;
1741 1742 1743 1744 1745 1746 1747
	int found = 0;

	if (search_end <= cur_start) {
		WARN_ON(1);
		return 0;
	}

1748
	spin_lock(&tree->lock);
1749 1750 1751 1752 1753 1754 1755 1756
	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.
	 */
1757
	node = tree_search(tree, cur_start);
C
Chris Mason 已提交
1758
	if (!node)
1759 1760
		goto out;

C
Chris Mason 已提交
1761
	while (1) {
1762 1763 1764
		state = rb_entry(node, struct extent_state, rb_node);
		if (state->start > search_end)
			break;
1765 1766 1767
		if (contig && found && state->start > last + 1)
			break;
		if (state->end >= cur_start && (state->state & bits) == bits) {
1768 1769 1770 1771 1772
			total_bytes += min(search_end, state->end) + 1 -
				       max(cur_start, state->start);
			if (total_bytes >= max_bytes)
				break;
			if (!found) {
1773
				*start = max(cur_start, state->start);
1774 1775
				found = 1;
			}
1776 1777 1778
			last = state->end;
		} else if (contig && found) {
			break;
1779 1780 1781 1782 1783 1784
		}
		node = rb_next(node);
		if (!node)
			break;
	}
out:
1785
	spin_unlock(&tree->lock);
1786 1787
	return total_bytes;
}
1788

C
Chris Mason 已提交
1789 1790 1791 1792
/*
 * set the private field for a given byte offset in the tree.  If there isn't
 * an extent_state there already, this does nothing.
 */
1793
static int set_state_private(struct extent_io_tree *tree, u64 start, u64 private)
1794 1795 1796 1797 1798
{
	struct rb_node *node;
	struct extent_state *state;
	int ret = 0;

1799
	spin_lock(&tree->lock);
1800 1801 1802 1803
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1804
	node = tree_search(tree, start);
1805
	if (!node) {
1806 1807 1808 1809 1810 1811 1812 1813 1814 1815
		ret = -ENOENT;
		goto out;
	}
	state = rb_entry(node, struct extent_state, rb_node);
	if (state->start != start) {
		ret = -ENOENT;
		goto out;
	}
	state->private = private;
out:
1816
	spin_unlock(&tree->lock);
1817 1818 1819 1820 1821 1822 1823 1824 1825
	return ret;
}

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

1826
	spin_lock(&tree->lock);
1827 1828 1829 1830
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1831
	node = tree_search(tree, start);
1832
	if (!node) {
1833 1834 1835 1836 1837 1838 1839 1840 1841 1842
		ret = -ENOENT;
		goto out;
	}
	state = rb_entry(node, struct extent_state, rb_node);
	if (state->start != start) {
		ret = -ENOENT;
		goto out;
	}
	*private = state->private;
out:
1843
	spin_unlock(&tree->lock);
1844 1845 1846 1847 1848
	return ret;
}

/*
 * searches a range in the state tree for a given mask.
1849
 * If 'filled' == 1, this returns 1 only if every extent in the tree
1850 1851 1852 1853
 * 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,
1854
		   unsigned long bits, int filled, struct extent_state *cached)
1855 1856 1857 1858 1859
{
	struct extent_state *state = NULL;
	struct rb_node *node;
	int bitset = 0;

1860
	spin_lock(&tree->lock);
1861 1862
	if (cached && cached->tree && cached->start <= start &&
	    cached->end > start)
1863 1864 1865
		node = &cached->rb_node;
	else
		node = tree_search(tree, start);
1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884
	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;
		}
1885 1886 1887 1888

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

1889 1890 1891 1892 1893 1894 1895 1896 1897 1898
		start = state->end + 1;
		if (start > end)
			break;
		node = rb_next(node);
		if (!node) {
			if (filled)
				bitset = 0;
			break;
		}
	}
1899
	spin_unlock(&tree->lock);
1900 1901 1902 1903 1904 1905 1906
	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
 */
1907
static void check_page_uptodate(struct extent_io_tree *tree, struct page *page)
1908
{
M
Miao Xie 已提交
1909
	u64 start = page_offset(page);
1910
	u64 end = start + PAGE_CACHE_SIZE - 1;
1911
	if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1, NULL))
1912 1913 1914
		SetPageUptodate(page);
}

1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947
/*
 * When IO fails, either with EIO or csum verification fails, we
 * try other mirrors that might have a good copy of the data.  This
 * io_failure_record is used to record state as we go through all the
 * mirrors.  If another mirror has good data, the page is set up to date
 * and things continue.  If a good mirror can't be found, the original
 * bio end_io callback is called to indicate things have failed.
 */
struct io_failure_record {
	struct page *page;
	u64 start;
	u64 len;
	u64 logical;
	unsigned long bio_flags;
	int this_mirror;
	int failed_mirror;
	int in_validation;
};

static int free_io_failure(struct inode *inode, struct io_failure_record *rec,
				int did_repair)
{
	int ret;
	int err = 0;
	struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;

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

D
David Woodhouse 已提交
1948 1949 1950 1951 1952
	ret = clear_extent_bits(&BTRFS_I(inode)->io_tree, rec->start,
				rec->start + rec->len - 1,
				EXTENT_DAMAGED, GFP_NOFS);
	if (ret && !err)
		err = ret;
1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967

	kfree(rec);
	return err;
}

static void repair_io_failure_callback(struct bio *bio, int err)
{
	complete(bio->bi_private);
}

/*
 * 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.
1968
 * to avoid any synchronization issues, wait for the data after writing, which
1969 1970 1971 1972
 * 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.
 */
1973
int repair_io_failure(struct btrfs_fs_info *fs_info, u64 start,
1974 1975 1976 1977 1978 1979 1980 1981 1982
			u64 length, u64 logical, struct page *page,
			int mirror_num)
{
	struct bio *bio;
	struct btrfs_device *dev;
	DECLARE_COMPLETION_ONSTACK(compl);
	u64 map_length = 0;
	u64 sector;
	struct btrfs_bio *bbio = NULL;
D
David Woodhouse 已提交
1983
	struct btrfs_mapping_tree *map_tree = &fs_info->mapping_tree;
1984 1985 1986 1987
	int ret;

	BUG_ON(!mirror_num);

D
David Woodhouse 已提交
1988 1989 1990 1991
	/* we can't repair anything in raid56 yet */
	if (btrfs_is_parity_mirror(map_tree, logical, length, mirror_num))
		return 0;

1992
	bio = btrfs_io_bio_alloc(GFP_NOFS, 1);
1993 1994 1995 1996 1997 1998 1999
	if (!bio)
		return -EIO;
	bio->bi_private = &compl;
	bio->bi_end_io = repair_io_failure_callback;
	bio->bi_size = 0;
	map_length = length;

2000
	ret = btrfs_map_block(fs_info, WRITE, logical,
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
			      &map_length, &bbio, mirror_num);
	if (ret) {
		bio_put(bio);
		return -EIO;
	}
	BUG_ON(mirror_num != bbio->mirror_num);
	sector = bbio->stripes[mirror_num-1].physical >> 9;
	bio->bi_sector = sector;
	dev = bbio->stripes[mirror_num-1].dev;
	kfree(bbio);
	if (!dev || !dev->bdev || !dev->writeable) {
		bio_put(bio);
		return -EIO;
	}
	bio->bi_bdev = dev->bdev;
M
Miao Xie 已提交
2016
	bio_add_page(bio, page, length, start - page_offset(page));
2017
	btrfsic_submit_bio(WRITE_SYNC, bio);
2018 2019 2020 2021 2022
	wait_for_completion(&compl);

	if (!test_bit(BIO_UPTODATE, &bio->bi_flags)) {
		/* try to remap that extent elsewhere? */
		bio_put(bio);
2023
		btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_WRITE_ERRS);
2024 2025 2026
		return -EIO;
	}

2027
	printk_ratelimited_in_rcu(KERN_INFO "btrfs read error corrected: ino %lu off %llu "
2028 2029
		      "(dev %s sector %llu)\n", page->mapping->host->i_ino,
		      start, rcu_str_deref(dev->name), sector);
2030 2031 2032 2033 2034

	bio_put(bio);
	return 0;
}

2035 2036 2037 2038 2039
int repair_eb_io_failure(struct btrfs_root *root, struct extent_buffer *eb,
			 int mirror_num)
{
	u64 start = eb->start;
	unsigned long i, num_pages = num_extent_pages(eb->start, eb->len);
2040
	int ret = 0;
2041 2042 2043

	for (i = 0; i < num_pages; i++) {
		struct page *p = extent_buffer_page(eb, i);
2044
		ret = repair_io_failure(root->fs_info, start, PAGE_CACHE_SIZE,
2045 2046 2047 2048 2049 2050 2051 2052 2053
					start, p, mirror_num);
		if (ret)
			break;
		start += PAGE_CACHE_SIZE;
	}

	return ret;
}

2054 2055 2056 2057 2058 2059 2060 2061 2062
/*
 * 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
 */
static int clean_io_failure(u64 start, struct page *page)
{
	u64 private;
	u64 private_failure;
	struct io_failure_record *failrec;
2063
	struct btrfs_fs_info *fs_info;
2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097
	struct extent_state *state;
	int num_copies;
	int did_repair = 0;
	int ret;
	struct inode *inode = page->mapping->host;

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

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

	failrec = (struct io_failure_record *)(unsigned long) private_failure;
	BUG_ON(!failrec->this_mirror);

	if (failrec->in_validation) {
		/* there was no real error, just free the record */
		pr_debug("clean_io_failure: freeing dummy error at %llu\n",
			 failrec->start);
		did_repair = 1;
		goto out;
	}

	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);

2098 2099
	if (state && state->start <= failrec->start &&
	    state->end >= failrec->start + failrec->len - 1) {
2100 2101 2102
		fs_info = BTRFS_I(inode)->root->fs_info;
		num_copies = btrfs_num_copies(fs_info, failrec->logical,
					      failrec->len);
2103
		if (num_copies > 1)  {
2104
			ret = repair_io_failure(fs_info, start, failrec->len,
2105 2106 2107 2108
						failrec->logical, page,
						failrec->failed_mirror);
			did_repair = !ret;
		}
D
David Woodhouse 已提交
2109
		ret = 0;
2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126
	}

out:
	if (!ret)
		ret = free_io_failure(inode, failrec, did_repair);

	return ret;
}

/*
 * 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
 */

2127 2128 2129
static int bio_readpage_error(struct bio *failed_bio, u64 phy_offset,
			      struct page *page, u64 start, u64 end,
			      int failed_mirror)
2130 2131 2132 2133 2134 2135 2136 2137 2138
{
	struct io_failure_record *failrec = NULL;
	u64 private;
	struct extent_map *em;
	struct inode *inode = page->mapping->host;
	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;
	struct bio *bio;
2139 2140
	struct btrfs_io_bio *btrfs_failed_bio;
	struct btrfs_io_bio *btrfs_bio;
2141 2142 2143 2144 2145 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
	int num_copies;
	int ret;
	int read_mode;
	u64 logical;

	BUG_ON(failed_bio->bi_rw & REQ_WRITE);

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

		if (em->start > start || em->start + em->len < start) {
			free_extent_map(em);
			em = NULL;
		}
		read_unlock(&em_tree->lock);

2173
		if (!em) {
2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215
			kfree(failrec);
			return -EIO;
		}
		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);
		}
		pr_debug("bio_readpage_error: (new) logical=%llu, start=%llu, "
			 "len=%llu\n", logical, start, failrec->len);
		failrec->logical = logical;
		free_extent_map(em);

		/* set the bits in the private failure tree */
		ret = set_extent_bits(failure_tree, start, end,
					EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
		if (ret >= 0)
			ret = set_state_private(failure_tree, start,
						(u64)(unsigned long)failrec);
		/* set the bits in the inode's tree */
		if (ret >= 0)
			ret = set_extent_bits(tree, start, end, EXTENT_DAMAGED,
						GFP_NOFS);
		if (ret < 0) {
			kfree(failrec);
			return ret;
		}
	} else {
		failrec = (struct io_failure_record *)(unsigned long)private;
		pr_debug("bio_readpage_error: (found) logical=%llu, "
			 "start=%llu, len=%llu, validation=%d\n",
			 failrec->logical, failrec->start, failrec->len,
			 failrec->in_validation);
		/*
		 * when data can be on disk more than twice, add to failrec here
		 * (e.g. with a list for failed_mirror) to make
		 * clean_io_failure() clean all those errors at once.
		 */
	}
2216 2217
	num_copies = btrfs_num_copies(BTRFS_I(inode)->root->fs_info,
				      failrec->logical, failrec->len);
2218 2219 2220 2221 2222 2223
	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.
		 */
2224 2225
		pr_debug("bio_readpage_error: cannot repair, num_copies=%d, next_mirror %d, failed_mirror %d\n",
			 num_copies, failrec->this_mirror, failed_mirror);
2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265
		free_io_failure(inode, failrec, 0);
		return -EIO;
	}

	/*
	 * 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;
		read_mode = READ_SYNC | REQ_FAILFAST_DEV;
	} 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++;
		read_mode = READ_SYNC;
	}

2266 2267
	if (failrec->this_mirror > num_copies) {
		pr_debug("bio_readpage_error: (fail) num_copies=%d, next_mirror %d, failed_mirror %d\n",
2268 2269 2270 2271 2272
			 num_copies, failrec->this_mirror, failed_mirror);
		free_io_failure(inode, failrec, 0);
		return -EIO;
	}

2273
	bio = btrfs_io_bio_alloc(GFP_NOFS, 1);
2274 2275 2276 2277
	if (!bio) {
		free_io_failure(inode, failrec, 0);
		return -EIO;
	}
2278 2279 2280 2281 2282
	bio->bi_end_io = failed_bio->bi_end_io;
	bio->bi_sector = failrec->logical >> 9;
	bio->bi_bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
	bio->bi_size = 0;

2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295
	btrfs_failed_bio = btrfs_io_bio(failed_bio);
	if (btrfs_failed_bio->csum) {
		struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
		u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);

		btrfs_bio = btrfs_io_bio(bio);
		btrfs_bio->csum = btrfs_bio->csum_inline;
		phy_offset >>= inode->i_sb->s_blocksize_bits;
		phy_offset *= csum_size;
		memcpy(btrfs_bio->csum, btrfs_failed_bio->csum + phy_offset,
		       csum_size);
	}

2296 2297 2298 2299 2300 2301
	bio_add_page(bio, page, failrec->len, start - page_offset(page));

	pr_debug("bio_readpage_error: submitting new read[%#x] to "
		 "this_mirror=%d, num_copies=%d, in_validation=%d\n", read_mode,
		 failrec->this_mirror, num_copies, failrec->in_validation);

2302 2303 2304 2305
	ret = tree->ops->submit_bio_hook(inode, read_mode, bio,
					 failrec->this_mirror,
					 failrec->bio_flags, 0);
	return ret;
2306 2307
}

2308 2309
/* lots and lots of room for performance fixes in the end_bio funcs */

2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331
int end_extent_writepage(struct page *page, int err, u64 start, u64 end)
{
	int uptodate = (err == 0);
	struct extent_io_tree *tree;
	int ret;

	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);
	}
	return 0;
}

2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343
/*
 * 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.
 */
static void end_bio_extent_writepage(struct bio *bio, int err)
{
	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
2344
	struct extent_io_tree *tree;
2345 2346 2347 2348 2349
	u64 start;
	u64 end;

	do {
		struct page *page = bvec->bv_page;
2350 2351
		tree = &BTRFS_I(page->mapping->host)->io_tree;

2352 2353 2354 2355 2356 2357 2358 2359 2360 2361
		/* 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.  */
		if (bvec->bv_offset || bvec->bv_len != PAGE_CACHE_SIZE)
			printk("%s page write in btrfs with offset %u and length %u\n",
			       bvec->bv_offset + bvec->bv_len != PAGE_CACHE_SIZE
			       ? KERN_ERR "partial" : KERN_INFO "incomplete",
			       bvec->bv_offset, bvec->bv_len);
2362

2363 2364
		start = page_offset(page);
		end = start + bvec->bv_offset + bvec->bv_len - 1;
2365 2366 2367

		if (--bvec >= bio->bi_io_vec)
			prefetchw(&bvec->bv_page->flags);
2368

2369 2370
		if (end_extent_writepage(page, err, start, end))
			continue;
2371

2372
		end_page_writeback(page);
2373
	} while (bvec >= bio->bi_io_vec);
2374

2375 2376 2377
	bio_put(bio);
}

2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389
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);
}

2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403
/*
 * 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.
 */
static void end_bio_extent_readpage(struct bio *bio, int err)
{
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
2404 2405
	struct bio_vec *bvec_end = bio->bi_io_vec + bio->bi_vcnt - 1;
	struct bio_vec *bvec = bio->bi_io_vec;
2406
	struct btrfs_io_bio *io_bio = btrfs_io_bio(bio);
2407
	struct extent_io_tree *tree;
2408
	u64 offset = 0;
2409 2410
	u64 start;
	u64 end;
2411
	u64 len;
2412 2413
	u64 extent_start = 0;
	u64 extent_len = 0;
2414
	int mirror;
2415 2416
	int ret;

2417 2418 2419
	if (err)
		uptodate = 0;

2420 2421
	do {
		struct page *page = bvec->bv_page;
2422
		struct inode *inode = page->mapping->host;
2423

2424
		pr_debug("end_bio_extent_readpage: bi_sector=%llu, err=%d, "
2425 2426
			 "mirror=%lu\n", (u64)bio->bi_sector, err,
			 io_bio->mirror_num);
2427
		tree = &BTRFS_I(inode)->io_tree;
2428

2429 2430 2431 2432 2433 2434 2435 2436 2437 2438
		/* 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.  */
		if (bvec->bv_offset || bvec->bv_len != PAGE_CACHE_SIZE)
			printk("%s page read in btrfs with offset %u and length %u\n",
			       bvec->bv_offset + bvec->bv_len != PAGE_CACHE_SIZE
			       ? KERN_ERR "partial" : KERN_INFO "incomplete",
			       bvec->bv_offset, bvec->bv_len);
2439

2440 2441
		start = page_offset(page);
		end = start + bvec->bv_offset + bvec->bv_len - 1;
2442
		len = bvec->bv_len;
2443

2444
		if (++bvec <= bvec_end)
2445 2446
			prefetchw(&bvec->bv_page->flags);

2447
		mirror = io_bio->mirror_num;
2448 2449
		if (likely(uptodate && tree->ops &&
			   tree->ops->readpage_end_io_hook)) {
2450 2451 2452
			ret = tree->ops->readpage_end_io_hook(io_bio, offset,
							      page, start, end,
							      mirror);
2453
			if (ret)
2454
				uptodate = 0;
2455
			else
2456
				clean_io_failure(start, page);
2457
		}
2458

2459 2460 2461 2462
		if (likely(uptodate))
			goto readpage_ok;

		if (tree->ops && tree->ops->readpage_io_failed_hook) {
2463
			ret = tree->ops->readpage_io_failed_hook(page, mirror);
2464 2465 2466
			if (!ret && !err &&
			    test_bit(BIO_UPTODATE, &bio->bi_flags))
				uptodate = 1;
2467
		} else {
2468 2469 2470 2471 2472 2473 2474 2475 2476 2477
			/*
			 * 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.
			 */
2478 2479
			ret = bio_readpage_error(bio, offset, page, start, end,
						 mirror);
2480
			if (ret == 0) {
2481 2482
				uptodate =
					test_bit(BIO_UPTODATE, &bio->bi_flags);
2483 2484
				if (err)
					uptodate = 0;
2485 2486 2487
				continue;
			}
		}
2488
readpage_ok:
2489
		if (likely(uptodate)) {
2490 2491 2492 2493 2494 2495 2496 2497
			loff_t i_size = i_size_read(inode);
			pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT;
			unsigned offset;

			/* Zero out the end if this page straddles i_size */
			offset = i_size & (PAGE_CACHE_SIZE-1);
			if (page->index == end_index && offset)
				zero_user_segment(page, offset, PAGE_CACHE_SIZE);
2498
			SetPageUptodate(page);
2499
		} else {
2500 2501
			ClearPageUptodate(page);
			SetPageError(page);
2502
		}
2503
		unlock_page(page);
2504
		offset += len;
2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526

		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;
		}
2527
	} while (bvec <= bvec_end);
2528

2529 2530 2531
	if (extent_len)
		endio_readpage_release_extent(tree, extent_start, extent_len,
					      uptodate);
2532 2533
	if (io_bio->end_io)
		io_bio->end_io(io_bio, err);
2534 2535 2536
	bio_put(bio);
}

2537 2538 2539 2540
/*
 * 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
 */
2541 2542 2543
struct bio *
btrfs_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
		gfp_t gfp_flags)
2544
{
2545
	struct btrfs_io_bio *btrfs_bio;
2546 2547
	struct bio *bio;

2548
	bio = bio_alloc_bioset(gfp_flags, nr_vecs, btrfs_bioset);
2549 2550

	if (bio == NULL && (current->flags & PF_MEMALLOC)) {
2551 2552 2553 2554
		while (!bio && (nr_vecs /= 2)) {
			bio = bio_alloc_bioset(gfp_flags,
					       nr_vecs, btrfs_bioset);
		}
2555 2556 2557
	}

	if (bio) {
2558
		bio->bi_size = 0;
2559 2560
		bio->bi_bdev = bdev;
		bio->bi_sector = first_sector;
2561 2562 2563 2564
		btrfs_bio = btrfs_io_bio(bio);
		btrfs_bio->csum = NULL;
		btrfs_bio->csum_allocated = NULL;
		btrfs_bio->end_io = NULL;
2565 2566 2567 2568
	}
	return bio;
}

2569 2570 2571 2572 2573 2574 2575 2576 2577
struct bio *btrfs_bio_clone(struct bio *bio, gfp_t gfp_mask)
{
	return bio_clone_bioset(bio, gfp_mask, btrfs_bioset);
}


/* this also allocates from the btrfs_bioset */
struct bio *btrfs_io_bio_alloc(gfp_t gfp_mask, unsigned int nr_iovecs)
{
2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588
	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;
2589 2590 2591
}


2592 2593
static int __must_check submit_one_bio(int rw, struct bio *bio,
				       int mirror_num, unsigned long bio_flags)
2594 2595
{
	int ret = 0;
2596 2597 2598 2599 2600
	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
2601
	start = page_offset(page) + bvec->bv_offset;
2602

2603
	bio->bi_private = NULL;
2604 2605 2606

	bio_get(bio);

2607
	if (tree->ops && tree->ops->submit_bio_hook)
2608
		ret = tree->ops->submit_bio_hook(page->mapping->host, rw, bio,
2609
					   mirror_num, bio_flags, start);
2610
	else
2611
		btrfsic_submit_bio(rw, bio);
2612

2613 2614 2615 2616 2617 2618
	if (bio_flagged(bio, BIO_EOPNOTSUPP))
		ret = -EOPNOTSUPP;
	bio_put(bio);
	return ret;
}

2619
static int merge_bio(int rw, struct extent_io_tree *tree, struct page *page,
2620 2621 2622 2623 2624
		     unsigned long offset, size_t size, struct bio *bio,
		     unsigned long bio_flags)
{
	int ret = 0;
	if (tree->ops && tree->ops->merge_bio_hook)
2625
		ret = tree->ops->merge_bio_hook(rw, page, offset, size, bio,
2626 2627 2628 2629 2630 2631
						bio_flags);
	BUG_ON(ret < 0);
	return ret;

}

2632 2633 2634 2635 2636 2637
static int submit_extent_page(int rw, struct extent_io_tree *tree,
			      struct page *page, sector_t sector,
			      size_t size, unsigned long offset,
			      struct block_device *bdev,
			      struct bio **bio_ret,
			      unsigned long max_pages,
2638
			      bio_end_io_t end_io_func,
C
Chris Mason 已提交
2639 2640 2641
			      int mirror_num,
			      unsigned long prev_bio_flags,
			      unsigned long bio_flags)
2642 2643 2644 2645
{
	int ret = 0;
	struct bio *bio;
	int nr;
C
Chris Mason 已提交
2646 2647 2648
	int contig = 0;
	int this_compressed = bio_flags & EXTENT_BIO_COMPRESSED;
	int old_compressed = prev_bio_flags & EXTENT_BIO_COMPRESSED;
2649
	size_t page_size = min_t(size_t, size, PAGE_CACHE_SIZE);
2650 2651 2652

	if (bio_ret && *bio_ret) {
		bio = *bio_ret;
C
Chris Mason 已提交
2653 2654 2655
		if (old_compressed)
			contig = bio->bi_sector == sector;
		else
K
Kent Overstreet 已提交
2656
			contig = bio_end_sector(bio) == sector;
C
Chris Mason 已提交
2657 2658

		if (prev_bio_flags != bio_flags || !contig ||
2659
		    merge_bio(rw, tree, page, offset, page_size, bio, bio_flags) ||
C
Chris Mason 已提交
2660 2661 2662
		    bio_add_page(bio, page, page_size, offset) < page_size) {
			ret = submit_one_bio(rw, bio, mirror_num,
					     prev_bio_flags);
2663 2664
			if (ret < 0)
				return ret;
2665 2666 2667 2668 2669
			bio = NULL;
		} else {
			return 0;
		}
	}
C
Chris Mason 已提交
2670 2671 2672 2673 2674
	if (this_compressed)
		nr = BIO_MAX_PAGES;
	else
		nr = bio_get_nr_vecs(bdev);

2675
	bio = btrfs_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH);
2676 2677
	if (!bio)
		return -ENOMEM;
2678

C
Chris Mason 已提交
2679
	bio_add_page(bio, page, page_size, offset);
2680 2681
	bio->bi_end_io = end_io_func;
	bio->bi_private = tree;
2682

C
Chris Mason 已提交
2683
	if (bio_ret)
2684
		*bio_ret = bio;
C
Chris Mason 已提交
2685
	else
C
Chris Mason 已提交
2686
		ret = submit_one_bio(rw, bio, mirror_num, bio_flags);
2687 2688 2689 2690

	return ret;
}

2691 2692
static void attach_extent_buffer_page(struct extent_buffer *eb,
				      struct page *page)
2693 2694 2695 2696
{
	if (!PagePrivate(page)) {
		SetPagePrivate(page);
		page_cache_get(page);
J
Josef Bacik 已提交
2697 2698 2699
		set_page_private(page, (unsigned long)eb);
	} else {
		WARN_ON(page->private != (unsigned long)eb);
2700 2701 2702
	}
}

J
Josef Bacik 已提交
2703
void set_page_extent_mapped(struct page *page)
2704
{
J
Josef Bacik 已提交
2705 2706 2707 2708 2709
	if (!PagePrivate(page)) {
		SetPagePrivate(page);
		page_cache_get(page);
		set_page_private(page, EXTENT_PAGE_PRIVATE);
	}
2710 2711
}

2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738
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;
		if (em->in_tree && start >= em->start &&
		    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;
}
2739 2740 2741 2742
/*
 * 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)
2743
 * XXX JDM: This needs looking at to ensure proper page locking
2744
 */
2745 2746 2747
static int __do_readpage(struct extent_io_tree *tree,
			 struct page *page,
			 get_extent_t *get_extent,
2748
			 struct extent_map **em_cached,
2749 2750
			 struct bio **bio, int mirror_num,
			 unsigned long *bio_flags, int rw)
2751 2752
{
	struct inode *inode = page->mapping->host;
M
Miao Xie 已提交
2753
	u64 start = page_offset(page);
2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765
	u64 page_end = start + PAGE_CACHE_SIZE - 1;
	u64 end;
	u64 cur = start;
	u64 extent_offset;
	u64 last_byte = i_size_read(inode);
	u64 block_start;
	u64 cur_end;
	sector_t sector;
	struct extent_map *em;
	struct block_device *bdev;
	int ret;
	int nr = 0;
2766
	int parent_locked = *bio_flags & EXTENT_BIO_PARENT_LOCKED;
2767
	size_t pg_offset = 0;
2768
	size_t iosize;
C
Chris Mason 已提交
2769
	size_t disk_io_size;
2770
	size_t blocksize = inode->i_sb->s_blocksize;
2771
	unsigned long this_bio_flag = *bio_flags & EXTENT_BIO_PARENT_LOCKED;
2772 2773 2774

	set_page_extent_mapped(page);

2775
	end = page_end;
D
Dan Magenheimer 已提交
2776 2777 2778
	if (!PageUptodate(page)) {
		if (cleancache_get_page(page) == 0) {
			BUG_ON(blocksize != PAGE_SIZE);
2779
			unlock_extent(tree, start, end);
D
Dan Magenheimer 已提交
2780 2781 2782 2783
			goto out;
		}
	}

C
Chris Mason 已提交
2784 2785 2786 2787 2788 2789
	if (page->index == last_byte >> PAGE_CACHE_SHIFT) {
		char *userpage;
		size_t zero_offset = last_byte & (PAGE_CACHE_SIZE - 1);

		if (zero_offset) {
			iosize = PAGE_CACHE_SIZE - zero_offset;
2790
			userpage = kmap_atomic(page);
C
Chris Mason 已提交
2791 2792
			memset(userpage + zero_offset, 0, iosize);
			flush_dcache_page(page);
2793
			kunmap_atomic(userpage);
C
Chris Mason 已提交
2794 2795
		}
	}
2796
	while (cur <= end) {
2797 2798
		unsigned long pnr = (last_byte >> PAGE_CACHE_SHIFT) + 1;

2799 2800
		if (cur >= last_byte) {
			char *userpage;
2801 2802
			struct extent_state *cached = NULL;

2803
			iosize = PAGE_CACHE_SIZE - pg_offset;
2804
			userpage = kmap_atomic(page);
2805
			memset(userpage + pg_offset, 0, iosize);
2806
			flush_dcache_page(page);
2807
			kunmap_atomic(userpage);
2808
			set_extent_uptodate(tree, cur, cur + iosize - 1,
2809
					    &cached, GFP_NOFS);
2810 2811 2812 2813
			if (!parent_locked)
				unlock_extent_cached(tree, cur,
						     cur + iosize - 1,
						     &cached, GFP_NOFS);
2814 2815
			break;
		}
2816 2817
		em = __get_extent_map(inode, page, pg_offset, cur,
				      end - cur + 1, get_extent, em_cached);
2818
		if (IS_ERR_OR_NULL(em)) {
2819
			SetPageError(page);
2820 2821
			if (!parent_locked)
				unlock_extent(tree, cur, end);
2822 2823 2824 2825 2826 2827
			break;
		}
		extent_offset = cur - em->start;
		BUG_ON(extent_map_end(em) <= cur);
		BUG_ON(end < cur);

2828
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
2829
			this_bio_flag |= EXTENT_BIO_COMPRESSED;
2830 2831 2832
			extent_set_compress_type(&this_bio_flag,
						 em->compress_type);
		}
C
Chris Mason 已提交
2833

2834 2835
		iosize = min(extent_map_end(em) - cur, end - cur + 1);
		cur_end = min(extent_map_end(em) - 1, end);
2836
		iosize = ALIGN(iosize, blocksize);
C
Chris Mason 已提交
2837 2838 2839 2840 2841 2842 2843
		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;
		}
2844 2845
		bdev = em->bdev;
		block_start = em->block_start;
Y
Yan Zheng 已提交
2846 2847
		if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			block_start = EXTENT_MAP_HOLE;
2848 2849 2850 2851 2852 2853
		free_extent_map(em);
		em = NULL;

		/* we've found a hole, just zero and go on */
		if (block_start == EXTENT_MAP_HOLE) {
			char *userpage;
2854 2855
			struct extent_state *cached = NULL;

2856
			userpage = kmap_atomic(page);
2857
			memset(userpage + pg_offset, 0, iosize);
2858
			flush_dcache_page(page);
2859
			kunmap_atomic(userpage);
2860 2861

			set_extent_uptodate(tree, cur, cur + iosize - 1,
2862 2863 2864
					    &cached, GFP_NOFS);
			unlock_extent_cached(tree, cur, cur + iosize - 1,
			                     &cached, GFP_NOFS);
2865
			cur = cur + iosize;
2866
			pg_offset += iosize;
2867 2868 2869
			continue;
		}
		/* the get_extent function already copied into the page */
2870 2871
		if (test_range_bit(tree, cur, cur_end,
				   EXTENT_UPTODATE, 1, NULL)) {
2872
			check_page_uptodate(tree, page);
2873 2874
			if (!parent_locked)
				unlock_extent(tree, cur, cur + iosize - 1);
2875
			cur = cur + iosize;
2876
			pg_offset += iosize;
2877 2878
			continue;
		}
2879 2880 2881 2882 2883
		/* we have an inline extent but it didn't get marked up
		 * to date.  Error out
		 */
		if (block_start == EXTENT_MAP_INLINE) {
			SetPageError(page);
2884 2885
			if (!parent_locked)
				unlock_extent(tree, cur, cur + iosize - 1);
2886
			cur = cur + iosize;
2887
			pg_offset += iosize;
2888 2889
			continue;
		}
2890

2891
		pnr -= page->index;
2892
		ret = submit_extent_page(rw, tree, page,
2893
					 sector, disk_io_size, pg_offset,
2894
					 bdev, bio, pnr,
C
Chris Mason 已提交
2895 2896 2897
					 end_bio_extent_readpage, mirror_num,
					 *bio_flags,
					 this_bio_flag);
2898 2899 2900 2901
		if (!ret) {
			nr++;
			*bio_flags = this_bio_flag;
		} else {
2902
			SetPageError(page);
2903 2904
			if (!parent_locked)
				unlock_extent(tree, cur, cur + iosize - 1);
2905
		}
2906
		cur = cur + iosize;
2907
		pg_offset += iosize;
2908
	}
D
Dan Magenheimer 已提交
2909
out:
2910 2911 2912 2913 2914 2915 2916 2917
	if (!nr) {
		if (!PageError(page))
			SetPageUptodate(page);
		unlock_page(page);
	}
	return 0;
}

2918 2919 2920 2921
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,
2922
					     struct extent_map **em_cached,
2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942
					     struct bio **bio, int mirror_num,
					     unsigned long *bio_flags, int rw)
{
	struct inode *inode;
	struct btrfs_ordered_extent *ordered;
	int index;

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

	for (index = 0; index < nr_pages; index++) {
2943 2944
		__do_readpage(tree, pages[index], get_extent, em_cached, bio,
			      mirror_num, bio_flags, rw);
2945 2946 2947 2948 2949 2950 2951
		page_cache_release(pages[index]);
	}
}

static void __extent_readpages(struct extent_io_tree *tree,
			       struct page *pages[],
			       int nr_pages, get_extent_t *get_extent,
2952
			       struct extent_map **em_cached,
2953 2954 2955
			       struct bio **bio, int mirror_num,
			       unsigned long *bio_flags, int rw)
{
2956
	u64 start = 0;
2957 2958 2959
	u64 end = 0;
	u64 page_start;
	int index;
2960
	int first_index = 0;
2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972

	for (index = 0; index < nr_pages; index++) {
		page_start = page_offset(pages[index]);
		if (!end) {
			start = page_start;
			end = start + PAGE_CACHE_SIZE - 1;
			first_index = index;
		} else if (end + 1 == page_start) {
			end += PAGE_CACHE_SIZE;
		} else {
			__do_contiguous_readpages(tree, &pages[first_index],
						  index - first_index, start,
2973 2974 2975
						  end, get_extent, em_cached,
						  bio, mirror_num, bio_flags,
						  rw);
2976 2977 2978 2979 2980 2981 2982 2983 2984
			start = page_start;
			end = start + PAGE_CACHE_SIZE - 1;
			first_index = index;
		}
	}

	if (end)
		__do_contiguous_readpages(tree, &pages[first_index],
					  index - first_index, start,
2985
					  end, get_extent, em_cached, bio,
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
					  mirror_num, bio_flags, rw);
}

static int __extent_read_full_page(struct extent_io_tree *tree,
				   struct page *page,
				   get_extent_t *get_extent,
				   struct bio **bio, int mirror_num,
				   unsigned long *bio_flags, int rw)
{
	struct inode *inode = page->mapping->host;
	struct btrfs_ordered_extent *ordered;
	u64 start = page_offset(page);
	u64 end = start + PAGE_CACHE_SIZE - 1;
	int ret;

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

3011 3012
	ret = __do_readpage(tree, page, get_extent, NULL, bio, mirror_num,
			    bio_flags, rw);
3013 3014 3015
	return ret;
}

3016
int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
3017
			    get_extent_t *get_extent, int mirror_num)
3018 3019
{
	struct bio *bio = NULL;
C
Chris Mason 已提交
3020
	unsigned long bio_flags = 0;
3021 3022
	int ret;

3023
	ret = __extent_read_full_page(tree, page, get_extent, &bio, mirror_num,
3024
				      &bio_flags, READ);
3025
	if (bio)
3026
		ret = submit_one_bio(READ, bio, mirror_num, bio_flags);
3027 3028 3029
	return ret;
}

3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043
int extent_read_full_page_nolock(struct extent_io_tree *tree, struct page *page,
				 get_extent_t *get_extent, int mirror_num)
{
	struct bio *bio = NULL;
	unsigned long bio_flags = EXTENT_BIO_PARENT_LOCKED;
	int ret;

	ret = __do_readpage(tree, page, get_extent, NULL, &bio, mirror_num,
				      &bio_flags, READ);
	if (bio)
		ret = submit_one_bio(READ, bio, mirror_num, bio_flags);
	return ret;
}

3044 3045 3046 3047 3048 3049 3050 3051 3052 3053
static noinline void update_nr_written(struct page *page,
				      struct writeback_control *wbc,
				      unsigned long nr_written)
{
	wbc->nr_to_write -= nr_written;
	if (wbc->range_cyclic || (wbc->nr_to_write > 0 &&
	    wbc->range_start == 0 && wbc->range_end == LLONG_MAX))
		page->mapping->writeback_index = page->index + nr_written;
}

3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065
/*
 * 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;
	struct extent_io_tree *tree = epd->tree;
M
Miao Xie 已提交
3066
	u64 start = page_offset(page);
3067 3068 3069 3070 3071 3072 3073 3074 3075
	u64 delalloc_start;
	u64 page_end = start + PAGE_CACHE_SIZE - 1;
	u64 end;
	u64 cur = start;
	u64 extent_offset;
	u64 last_byte = i_size_read(inode);
	u64 block_start;
	u64 iosize;
	sector_t sector;
3076
	struct extent_state *cached_state = NULL;
3077 3078 3079 3080
	struct extent_map *em;
	struct block_device *bdev;
	int ret;
	int nr = 0;
3081
	size_t pg_offset = 0;
3082 3083 3084 3085 3086
	size_t blocksize;
	loff_t i_size = i_size_read(inode);
	unsigned long end_index = i_size >> PAGE_CACHE_SHIFT;
	u64 nr_delalloc;
	u64 delalloc_end;
C
Chris Mason 已提交
3087 3088
	int page_started;
	int compressed;
3089
	int write_flags;
3090
	unsigned long nr_written = 0;
3091
	bool fill_delalloc = true;
3092

3093
	if (wbc->sync_mode == WB_SYNC_ALL)
J
Jens Axboe 已提交
3094
		write_flags = WRITE_SYNC;
3095 3096 3097
	else
		write_flags = WRITE;

3098 3099
	trace___extent_writepage(page, inode, wbc);

3100
	WARN_ON(!PageLocked(page));
3101 3102 3103

	ClearPageError(page);

3104
	pg_offset = i_size & (PAGE_CACHE_SIZE - 1);
3105
	if (page->index > end_index ||
3106
	   (page->index == end_index && !pg_offset)) {
3107
		page->mapping->a_ops->invalidatepage(page, 0, PAGE_CACHE_SIZE);
3108 3109 3110 3111 3112 3113 3114
		unlock_page(page);
		return 0;
	}

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

3115
		userpage = kmap_atomic(page);
3116 3117
		memset(userpage + pg_offset, 0,
		       PAGE_CACHE_SIZE - pg_offset);
3118
		kunmap_atomic(userpage);
3119
		flush_dcache_page(page);
3120
	}
3121
	pg_offset = 0;
3122 3123 3124

	set_page_extent_mapped(page);

3125 3126 3127
	if (!tree->ops || !tree->ops->fill_delalloc)
		fill_delalloc = false;

3128 3129
	delalloc_start = start;
	delalloc_end = 0;
C
Chris Mason 已提交
3130
	page_started = 0;
3131
	if (!epd->extent_locked && fill_delalloc) {
3132
		u64 delalloc_to_write = 0;
3133 3134 3135 3136 3137 3138
		/*
		 * make sure the wbc mapping index is at least updated
		 * to this page.
		 */
		update_nr_written(page, wbc, 0);

C
Chris Mason 已提交
3139
		while (delalloc_end < page_end) {
3140
			nr_delalloc = find_lock_delalloc_range(inode, tree,
C
Chris Mason 已提交
3141 3142
						       page,
						       &delalloc_start,
3143 3144
						       &delalloc_end,
						       128 * 1024 * 1024);
3145 3146 3147 3148
			if (nr_delalloc == 0) {
				delalloc_start = delalloc_end + 1;
				continue;
			}
3149 3150 3151 3152 3153
			ret = tree->ops->fill_delalloc(inode, page,
						       delalloc_start,
						       delalloc_end,
						       &page_started,
						       &nr_written);
3154 3155 3156 3157 3158
			/* File system has been set read-only */
			if (ret) {
				SetPageError(page);
				goto done;
			}
3159 3160 3161 3162 3163 3164 3165 3166
			/*
			 * delalloc_end is already one less than the total
			 * length, so we don't subtract one from
			 * PAGE_CACHE_SIZE
			 */
			delalloc_to_write += (delalloc_end - delalloc_start +
					      PAGE_CACHE_SIZE) >>
					      PAGE_CACHE_SHIFT;
3167 3168
			delalloc_start = delalloc_end + 1;
		}
3169 3170 3171 3172 3173 3174 3175 3176
		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);
		}
C
Chris Mason 已提交
3177

3178 3179 3180 3181 3182
		/* did the fill delalloc function already unlock and start
		 * the IO?
		 */
		if (page_started) {
			ret = 0;
3183 3184 3185 3186 3187 3188 3189
			/*
			 * 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;
			goto done_unlocked;
3190
		}
C
Chris Mason 已提交
3191
	}
3192
	if (tree->ops && tree->ops->writepage_start_hook) {
C
Chris Mason 已提交
3193 3194
		ret = tree->ops->writepage_start_hook(page, start,
						      page_end);
3195 3196 3197 3198 3199 3200
		if (ret) {
			/* Fixup worker will requeue */
			if (ret == -EBUSY)
				wbc->pages_skipped++;
			else
				redirty_page_for_writepage(wbc, page);
3201
			update_nr_written(page, wbc, nr_written);
3202
			unlock_page(page);
3203
			ret = 0;
3204
			goto done_unlocked;
3205 3206 3207
		}
	}

3208 3209 3210 3211 3212
	/*
	 * we don't want to touch the inode after unlocking the page,
	 * so we update the mapping writeback index now
	 */
	update_nr_written(page, wbc, nr_written + 1);
3213

3214 3215
	end = page_end;
	if (last_byte <= start) {
3216 3217 3218
		if (tree->ops && tree->ops->writepage_end_io_hook)
			tree->ops->writepage_end_io_hook(page, start,
							 page_end, NULL, 1);
3219 3220 3221 3222 3223 3224 3225
		goto done;
	}

	blocksize = inode->i_sb->s_blocksize;

	while (cur <= end) {
		if (cur >= last_byte) {
3226 3227 3228
			if (tree->ops && tree->ops->writepage_end_io_hook)
				tree->ops->writepage_end_io_hook(page, cur,
							 page_end, NULL, 1);
3229 3230
			break;
		}
3231
		em = epd->get_extent(inode, page, pg_offset, cur,
3232
				     end - cur + 1, 1);
3233
		if (IS_ERR_OR_NULL(em)) {
3234 3235 3236 3237 3238 3239 3240 3241
			SetPageError(page);
			break;
		}

		extent_offset = cur - em->start;
		BUG_ON(extent_map_end(em) <= cur);
		BUG_ON(end < cur);
		iosize = min(extent_map_end(em) - cur, end - cur + 1);
3242
		iosize = ALIGN(iosize, blocksize);
3243 3244 3245
		sector = (em->block_start + extent_offset) >> 9;
		bdev = em->bdev;
		block_start = em->block_start;
C
Chris Mason 已提交
3246
		compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
3247 3248 3249
		free_extent_map(em);
		em = NULL;

C
Chris Mason 已提交
3250 3251 3252 3253 3254
		/*
		 * compressed and inline extents are written through other
		 * paths in the FS
		 */
		if (compressed || block_start == EXTENT_MAP_HOLE ||
3255
		    block_start == EXTENT_MAP_INLINE) {
C
Chris Mason 已提交
3256 3257 3258 3259 3260 3261
			/*
			 * end_io notification does not happen here for
			 * compressed extents
			 */
			if (!compressed && tree->ops &&
			    tree->ops->writepage_end_io_hook)
3262 3263 3264
				tree->ops->writepage_end_io_hook(page, cur,
							 cur + iosize - 1,
							 NULL, 1);
C
Chris Mason 已提交
3265 3266 3267 3268 3269 3270 3271 3272 3273
			else if (compressed) {
				/* we don't want to end_page_writeback on
				 * a compressed extent.  this happens
				 * elsewhere
				 */
				nr++;
			}

			cur += iosize;
3274
			pg_offset += iosize;
3275 3276 3277 3278
			continue;
		}
		/* leave this out until we have a page_mkwrite call */
		if (0 && !test_range_bit(tree, cur, cur + iosize - 1,
3279
				   EXTENT_DIRTY, 0, NULL)) {
3280
			cur = cur + iosize;
3281
			pg_offset += iosize;
3282 3283
			continue;
		}
C
Chris Mason 已提交
3284

3285 3286 3287 3288 3289 3290
		if (tree->ops && tree->ops->writepage_io_hook) {
			ret = tree->ops->writepage_io_hook(page, cur,
						cur + iosize - 1);
		} else {
			ret = 0;
		}
3291
		if (ret) {
3292
			SetPageError(page);
3293
		} else {
3294
			unsigned long max_nr = end_index + 1;
3295

3296 3297
			set_range_writeback(tree, cur, cur + iosize - 1);
			if (!PageWriteback(page)) {
C
Chris Mason 已提交
3298 3299
				printk(KERN_ERR "btrfs warning page %lu not "
				       "writeback, cur %llu end %llu\n",
3300
				       page->index, cur, end);
3301 3302
			}

3303 3304 3305
			ret = submit_extent_page(write_flags, tree, page,
						 sector, iosize, pg_offset,
						 bdev, &epd->bio, max_nr,
C
Chris Mason 已提交
3306 3307
						 end_bio_extent_writepage,
						 0, 0, 0);
3308 3309 3310 3311
			if (ret)
				SetPageError(page);
		}
		cur = cur + iosize;
3312
		pg_offset += iosize;
3313 3314 3315 3316 3317 3318 3319 3320 3321
		nr++;
	}
done:
	if (nr == 0) {
		/* make sure the mapping tag for page dirty gets cleared */
		set_page_writeback(page);
		end_page_writeback(page);
	}
	unlock_page(page);
3322

3323 3324
done_unlocked:

3325 3326
	/* drop our reference on any cached states */
	free_extent_state(cached_state);
3327 3328 3329
	return 0;
}

3330 3331 3332 3333 3334 3335
static int eb_wait(void *word)
{
	io_schedule();
	return 0;
}

3336
void wait_on_extent_buffer_writeback(struct extent_buffer *eb)
3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363
{
	wait_on_bit(&eb->bflags, EXTENT_BUFFER_WRITEBACK, eb_wait,
		    TASK_UNINTERRUPTIBLE);
}

static int lock_extent_buffer_for_io(struct extent_buffer *eb,
				     struct btrfs_fs_info *fs_info,
				     struct extent_page_data *epd)
{
	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
3364 3365 3366 3367 3368
		while (1) {
			wait_on_extent_buffer_writeback(eb);
			btrfs_tree_lock(eb);
			if (!test_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags))
				break;
3369 3370 3371 3372
			btrfs_tree_unlock(eb);
		}
	}

3373 3374 3375 3376 3377 3378
	/*
	 * 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);
3379 3380
	if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
		set_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags);
3381
		spin_unlock(&eb->refs_lock);
3382
		btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
3383 3384 3385
		__percpu_counter_add(&fs_info->dirty_metadata_bytes,
				     -eb->len,
				     fs_info->dirty_metadata_batch);
3386
		ret = 1;
3387 3388
	} else {
		spin_unlock(&eb->refs_lock);
3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459
	}

	btrfs_tree_unlock(eb);

	if (!ret)
		return ret;

	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = 0; i < num_pages; i++) {
		struct page *p = extent_buffer_page(eb, i);

		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);
	smp_mb__after_clear_bit();
	wake_up_bit(&eb->bflags, EXTENT_BUFFER_WRITEBACK);
}

static void end_bio_extent_buffer_writepage(struct bio *bio, int err)
{
	int uptodate = err == 0;
	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
	struct extent_buffer *eb;
	int done;

	do {
		struct page *page = bvec->bv_page;

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

		if (!uptodate || test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) {
			set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
			ClearPageUptodate(page);
			SetPageError(page);
		}

		end_page_writeback(page);

		if (!done)
			continue;

		end_extent_buffer_writeback(eb);
	} while (bvec >= bio->bi_io_vec);

	bio_put(bio);

}

static int write_one_eb(struct extent_buffer *eb,
			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;
	u64 offset = eb->start;
	unsigned long i, num_pages;
3460
	unsigned long bio_flags = 0;
3461
	int rw = (epd->sync_io ? WRITE_SYNC : WRITE) | REQ_META;
3462
	int ret = 0;
3463 3464 3465 3466

	clear_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
	num_pages = num_extent_pages(eb->start, eb->len);
	atomic_set(&eb->io_pages, num_pages);
3467 3468 3469
	if (btrfs_header_owner(eb) == BTRFS_TREE_LOG_OBJECTID)
		bio_flags = EXTENT_BIO_TREE_LOG;

3470 3471 3472 3473 3474 3475 3476 3477
	for (i = 0; i < num_pages; i++) {
		struct page *p = extent_buffer_page(eb, i);

		clear_page_dirty_for_io(p);
		set_page_writeback(p);
		ret = submit_extent_page(rw, eb->tree, p, offset >> 9,
					 PAGE_CACHE_SIZE, 0, bdev, &epd->bio,
					 -1, end_bio_extent_buffer_writepage,
3478 3479
					 0, epd->bio_flags, bio_flags);
		epd->bio_flags = bio_flags;
3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513
		if (ret) {
			set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
			SetPageError(p);
			if (atomic_sub_and_test(num_pages - i, &eb->io_pages))
				end_extent_buffer_writeback(eb);
			ret = -EIO;
			break;
		}
		offset += PAGE_CACHE_SIZE;
		update_nr_written(p, wbc, 1);
		unlock_page(p);
	}

	if (unlikely(ret)) {
		for (; i < num_pages; i++) {
			struct page *p = extent_buffer_page(eb, i);
			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,
3514
		.bio_flags = 0,
3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558
	};
	int ret = 0;
	int done = 0;
	int nr_to_write_done = 0;
	struct pagevec pvec;
	int nr_pages;
	pgoff_t index;
	pgoff_t end;		/* Inclusive */
	int scanned = 0;
	int tag;

	pagevec_init(&pvec, 0);
	if (wbc->range_cyclic) {
		index = mapping->writeback_index; /* Start from prev offset */
		end = -1;
	} else {
		index = wbc->range_start >> PAGE_CACHE_SHIFT;
		end = wbc->range_end >> PAGE_CACHE_SHIFT;
		scanned = 1;
	}
	if (wbc->sync_mode == WB_SYNC_ALL)
		tag = PAGECACHE_TAG_TOWRITE;
	else
		tag = PAGECACHE_TAG_DIRTY;
retry:
	if (wbc->sync_mode == WB_SYNC_ALL)
		tag_pages_for_writeback(mapping, index, end);
	while (!done && !nr_to_write_done && (index <= end) &&
	       (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag,
			min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
		unsigned i;

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

			if (!PagePrivate(page))
				continue;

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

3559 3560 3561 3562 3563 3564
			spin_lock(&mapping->private_lock);
			if (!PagePrivate(page)) {
				spin_unlock(&mapping->private_lock);
				continue;
			}

3565
			eb = (struct extent_buffer *)page->private;
3566 3567 3568 3569 3570 3571

			/*
			 * 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.
			 */
3572
			if (!eb) {
3573
				spin_unlock(&mapping->private_lock);
3574 3575 3576 3577
				WARN_ON(1);
				continue;
			}

3578 3579
			if (eb == prev_eb) {
				spin_unlock(&mapping->private_lock);
3580
				continue;
3581
			}
3582

3583 3584 3585
			ret = atomic_inc_not_zero(&eb->refs);
			spin_unlock(&mapping->private_lock);
			if (!ret)
3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625
				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;
}

3626
/**
C
Chris Mason 已提交
3627
 * write_cache_pages - walk the list of dirty pages of the given address space and write all of them.
3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640
 * @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.
 */
3641
static int extent_write_cache_pages(struct extent_io_tree *tree,
C
Chris Mason 已提交
3642 3643
			     struct address_space *mapping,
			     struct writeback_control *wbc,
C
Chris Mason 已提交
3644 3645
			     writepage_t writepage, void *data,
			     void (*flush_fn)(void *))
3646
{
3647
	struct inode *inode = mapping->host;
3648 3649
	int ret = 0;
	int done = 0;
3650
	int nr_to_write_done = 0;
3651 3652 3653 3654 3655
	struct pagevec pvec;
	int nr_pages;
	pgoff_t index;
	pgoff_t end;		/* Inclusive */
	int scanned = 0;
3656
	int tag;
3657

3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669
	/*
	 * 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;

3670 3671 3672 3673 3674 3675 3676 3677 3678
	pagevec_init(&pvec, 0);
	if (wbc->range_cyclic) {
		index = mapping->writeback_index; /* Start from prev offset */
		end = -1;
	} else {
		index = wbc->range_start >> PAGE_CACHE_SHIFT;
		end = wbc->range_end >> PAGE_CACHE_SHIFT;
		scanned = 1;
	}
3679 3680 3681 3682
	if (wbc->sync_mode == WB_SYNC_ALL)
		tag = PAGECACHE_TAG_TOWRITE;
	else
		tag = PAGECACHE_TAG_DIRTY;
3683
retry:
3684 3685
	if (wbc->sync_mode == WB_SYNC_ALL)
		tag_pages_for_writeback(mapping, index, end);
3686
	while (!done && !nr_to_write_done && (index <= end) &&
3687 3688
	       (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag,
			min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701
		unsigned i;

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

			/*
			 * At this point we hold neither mapping->tree_lock nor
			 * lock on the page itself: the page may be truncated or
			 * invalidated (changing page->mapping to NULL), or even
			 * swizzled back from swapper_space to tmpfs file
			 * mapping
			 */
3702 3703 3704
			if (!trylock_page(page)) {
				flush_fn(data);
				lock_page(page);
3705
			}
3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717

			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
3718
			if (wbc->sync_mode != WB_SYNC_NONE) {
3719 3720
				if (PageWriteback(page))
					flush_fn(data);
3721
				wait_on_page_writeback(page);
C
Chris Mason 已提交
3722
			}
3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735

			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;
			}
3736
			if (ret)
3737
				done = 1;
3738 3739 3740 3741 3742 3743 3744

			/*
			 * 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;
3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757
		}
		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;
	}
3758
	btrfs_add_delayed_iput(inode);
3759 3760 3761
	return ret;
}

3762
static void flush_epd_write_bio(struct extent_page_data *epd)
C
Chris Mason 已提交
3763 3764
{
	if (epd->bio) {
3765 3766 3767
		int rw = WRITE;
		int ret;

3768
		if (epd->sync_io)
3769 3770
			rw = WRITE_SYNC;

3771
		ret = submit_one_bio(rw, epd->bio, 0, epd->bio_flags);
3772
		BUG_ON(ret < 0); /* -ENOMEM */
C
Chris Mason 已提交
3773 3774 3775 3776
		epd->bio = NULL;
	}
}

3777 3778 3779 3780 3781 3782
static noinline void flush_write_bio(void *data)
{
	struct extent_page_data *epd = data;
	flush_epd_write_bio(epd);
}

3783 3784 3785 3786 3787 3788 3789 3790 3791
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,
3792
		.extent_locked = 0,
3793
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
3794
		.bio_flags = 0,
3795 3796 3797 3798
	};

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

3799
	flush_epd_write_bio(&epd);
3800 3801 3802
	return ret;
}

3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817
int extent_write_locked_range(struct extent_io_tree *tree, struct inode *inode,
			      u64 start, u64 end, get_extent_t *get_extent,
			      int mode)
{
	int ret = 0;
	struct address_space *mapping = inode->i_mapping;
	struct page *page;
	unsigned long nr_pages = (end - start + PAGE_CACHE_SIZE) >>
		PAGE_CACHE_SHIFT;

	struct extent_page_data epd = {
		.bio = NULL,
		.tree = tree,
		.get_extent = get_extent,
		.extent_locked = 1,
3818
		.sync_io = mode == WB_SYNC_ALL,
3819
		.bio_flags = 0,
3820 3821 3822 3823 3824 3825 3826 3827
	};
	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
3828
	while (start <= end) {
3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 3842
		page = find_get_page(mapping, start >> PAGE_CACHE_SHIFT);
		if (clear_page_dirty_for_io(page))
			ret = __extent_writepage(page, &wbc_writepages, &epd);
		else {
			if (tree->ops && tree->ops->writepage_end_io_hook)
				tree->ops->writepage_end_io_hook(page, start,
						 start + PAGE_CACHE_SIZE - 1,
						 NULL, 1);
			unlock_page(page);
		}
		page_cache_release(page);
		start += PAGE_CACHE_SIZE;
	}

3843
	flush_epd_write_bio(&epd);
3844 3845
	return ret;
}
3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856

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,
3857
		.extent_locked = 0,
3858
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
3859
		.bio_flags = 0,
3860 3861
	};

C
Chris Mason 已提交
3862
	ret = extent_write_cache_pages(tree, mapping, wbc,
C
Chris Mason 已提交
3863 3864
				       __extent_writepage, &epd,
				       flush_write_bio);
3865
	flush_epd_write_bio(&epd);
3866 3867 3868 3869 3870 3871 3872 3873 3874 3875
	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
3876
	unsigned long bio_flags = 0;
L
Liu Bo 已提交
3877 3878
	struct page *pagepool[16];
	struct page *page;
3879
	struct extent_map *em_cached = NULL;
L
Liu Bo 已提交
3880
	int nr = 0;
3881 3882

	for (page_idx = 0; page_idx < nr_pages; page_idx++) {
L
Liu Bo 已提交
3883
		page = list_entry(pages->prev, struct page, lru);
3884 3885 3886

		prefetchw(&page->flags);
		list_del(&page->lru);
L
Liu Bo 已提交
3887
		if (add_to_page_cache_lru(page, mapping,
3888
					page->index, GFP_NOFS)) {
L
Liu Bo 已提交
3889 3890
			page_cache_release(page);
			continue;
3891
		}
L
Liu Bo 已提交
3892 3893 3894 3895

		pagepool[nr++] = page;
		if (nr < ARRAY_SIZE(pagepool))
			continue;
3896
		__extent_readpages(tree, pagepool, nr, get_extent, &em_cached,
3897
				   &bio, 0, &bio_flags, READ);
L
Liu Bo 已提交
3898
		nr = 0;
3899
	}
3900
	if (nr)
3901
		__extent_readpages(tree, pagepool, nr, get_extent, &em_cached,
3902
				   &bio, 0, &bio_flags, READ);
L
Liu Bo 已提交
3903

3904 3905 3906
	if (em_cached)
		free_extent_map(em_cached);

3907 3908
	BUG_ON(!list_empty(pages));
	if (bio)
3909
		return submit_one_bio(READ, bio, 0, bio_flags);
3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920
	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)
{
3921
	struct extent_state *cached_state = NULL;
M
Miao Xie 已提交
3922
	u64 start = page_offset(page);
3923 3924 3925
	u64 end = start + PAGE_CACHE_SIZE - 1;
	size_t blocksize = page->mapping->host->i_sb->s_blocksize;

3926
	start += ALIGN(offset, blocksize);
3927 3928 3929
	if (start > end)
		return 0;

3930
	lock_extent_bits(tree, start, end, 0, &cached_state);
3931
	wait_on_page_writeback(page);
3932
	clear_extent_bit(tree, start, end,
3933 3934
			 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
			 EXTENT_DO_ACCOUNTING,
3935
			 1, 1, &cached_state, GFP_NOFS);
3936 3937 3938
	return 0;
}

3939 3940 3941 3942 3943
/*
 * 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.
 */
3944 3945 3946
static int try_release_extent_state(struct extent_map_tree *map,
				    struct extent_io_tree *tree,
				    struct page *page, gfp_t mask)
3947
{
M
Miao Xie 已提交
3948
	u64 start = page_offset(page);
3949 3950 3951
	u64 end = start + PAGE_CACHE_SIZE - 1;
	int ret = 1;

3952
	if (test_range_bit(tree, start, end,
3953
			   EXTENT_IOBITS, 0, NULL))
3954 3955 3956 3957
		ret = 0;
	else {
		if ((mask & GFP_NOFS) == GFP_NOFS)
			mask = GFP_NOFS;
3958 3959 3960 3961
		/*
		 * at this point we can safely clear everything except the
		 * locked bit and the nodatasum bit
		 */
3962
		ret = clear_extent_bit(tree, start, end,
3963 3964
				 ~(EXTENT_LOCKED | EXTENT_NODATASUM),
				 0, 0, NULL, mask);
3965 3966 3967 3968 3969 3970 3971 3972

		/* if clear_extent_bit failed for enomem reasons,
		 * we can't allow the release to continue.
		 */
		if (ret < 0)
			ret = 0;
		else
			ret = 1;
3973 3974 3975 3976
	}
	return ret;
}

3977 3978 3979 3980 3981 3982
/*
 * 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,
3983 3984
			       struct extent_io_tree *tree, struct page *page,
			       gfp_t mask)
3985 3986
{
	struct extent_map *em;
M
Miao Xie 已提交
3987
	u64 start = page_offset(page);
3988
	u64 end = start + PAGE_CACHE_SIZE - 1;
3989

3990 3991
	if ((mask & __GFP_WAIT) &&
	    page->mapping->host->i_size > 16 * 1024 * 1024) {
3992
		u64 len;
3993
		while (start <= end) {
3994
			len = end - start + 1;
3995
			write_lock(&map->lock);
3996
			em = lookup_extent_mapping(map, start, len);
3997
			if (!em) {
3998
				write_unlock(&map->lock);
3999 4000
				break;
			}
4001 4002
			if (test_bit(EXTENT_FLAG_PINNED, &em->flags) ||
			    em->start != start) {
4003
				write_unlock(&map->lock);
4004 4005 4006 4007 4008
				free_extent_map(em);
				break;
			}
			if (!test_range_bit(tree, em->start,
					    extent_map_end(em) - 1,
4009
					    EXTENT_LOCKED | EXTENT_WRITEBACK,
4010
					    0, NULL)) {
4011 4012 4013 4014 4015
				remove_extent_mapping(map, em);
				/* once for the rb tree */
				free_extent_map(em);
			}
			start = extent_map_end(em);
4016
			write_unlock(&map->lock);
4017 4018

			/* once for us */
4019 4020 4021
			free_extent_map(em);
		}
	}
4022
	return try_release_extent_state(map, tree, page, mask);
4023 4024
}

4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044
/*
 * helper function for fiemap, which doesn't want to see any holes.
 * This maps until we find something past 'last'
 */
static struct extent_map *get_extent_skip_holes(struct inode *inode,
						u64 offset,
						u64 last,
						get_extent_t *get_extent)
{
	u64 sectorsize = BTRFS_I(inode)->root->sectorsize;
	struct extent_map *em;
	u64 len;

	if (offset >= last)
		return NULL;

	while(1) {
		len = last - offset;
		if (len == 0)
			break;
4045
		len = ALIGN(len, sectorsize);
4046
		em = get_extent(inode, NULL, 0, offset, len, 0);
4047
		if (IS_ERR_OR_NULL(em))
4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064
			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
4065 4066 4067
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
4068
	int ret = 0;
Y
Yehuda Sadeh 已提交
4069 4070 4071
	u64 off = start;
	u64 max = start + len;
	u32 flags = 0;
J
Josef Bacik 已提交
4072 4073
	u32 found_type;
	u64 last;
4074
	u64 last_for_get_extent = 0;
Y
Yehuda Sadeh 已提交
4075
	u64 disko = 0;
4076
	u64 isize = i_size_read(inode);
J
Josef Bacik 已提交
4077
	struct btrfs_key found_key;
Y
Yehuda Sadeh 已提交
4078
	struct extent_map *em = NULL;
4079
	struct extent_state *cached_state = NULL;
J
Josef Bacik 已提交
4080 4081
	struct btrfs_path *path;
	struct btrfs_file_extent_item *item;
Y
Yehuda Sadeh 已提交
4082
	int end = 0;
4083 4084 4085
	u64 em_start = 0;
	u64 em_len = 0;
	u64 em_end = 0;
Y
Yehuda Sadeh 已提交
4086 4087 4088 4089 4090
	unsigned long emflags;

	if (len == 0)
		return -EINVAL;

J
Josef Bacik 已提交
4091 4092 4093 4094 4095
	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;
	path->leave_spinning = 1;

4096 4097 4098
	start = ALIGN(start, BTRFS_I(inode)->root->sectorsize);
	len = ALIGN(len, BTRFS_I(inode)->root->sectorsize);

4099 4100 4101 4102
	/*
	 * lookup the last file extent.  We're not using i_size here
	 * because there might be preallocation past i_size
	 */
J
Josef Bacik 已提交
4103
	ret = btrfs_lookup_file_extent(NULL, BTRFS_I(inode)->root,
L
Li Zefan 已提交
4104
				       path, btrfs_ino(inode), -1, 0);
J
Josef Bacik 已提交
4105 4106 4107 4108 4109 4110 4111 4112 4113 4114 4115
	if (ret < 0) {
		btrfs_free_path(path);
		return ret;
	}
	WARN_ON(!ret);
	path->slots[0]--;
	item = btrfs_item_ptr(path->nodes[0], path->slots[0],
			      struct btrfs_file_extent_item);
	btrfs_item_key_to_cpu(path->nodes[0], &found_key, path->slots[0]);
	found_type = btrfs_key_type(&found_key);

4116
	/* No extents, but there might be delalloc bits */
L
Li Zefan 已提交
4117
	if (found_key.objectid != btrfs_ino(inode) ||
J
Josef Bacik 已提交
4118
	    found_type != BTRFS_EXTENT_DATA_KEY) {
4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129
		/* 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
4130 4131 4132
	}
	btrfs_free_path(path);

4133 4134 4135 4136 4137 4138 4139 4140 4141 4142
	/*
	 * we might have some extents allocated but more delalloc past those
	 * extents.  so, we trust isize unless the start of the last extent is
	 * beyond isize
	 */
	if (last < isize) {
		last = (u64)-1;
		last_for_get_extent = isize;
	}

L
Liu Bo 已提交
4143
	lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + len - 1, 0,
4144
			 &cached_state);
4145

4146
	em = get_extent_skip_holes(inode, start, last_for_get_extent,
4147
				   get_extent);
Y
Yehuda Sadeh 已提交
4148 4149 4150 4151 4152 4153
	if (!em)
		goto out;
	if (IS_ERR(em)) {
		ret = PTR_ERR(em);
		goto out;
	}
J
Josef Bacik 已提交
4154

Y
Yehuda Sadeh 已提交
4155
	while (!end) {
4156
		u64 offset_in_extent = 0;
4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168

		/* 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
4169

4170 4171
		/*
		 * record the offset from the start of the extent
4172 4173 4174
		 * 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.
4175
		 */
4176 4177
		if (!test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
			offset_in_extent = em_start - em->start;
4178
		em_end = extent_map_end(em);
4179
		em_len = em_end - em_start;
4180
		emflags = em->flags;
Y
Yehuda Sadeh 已提交
4181 4182 4183
		disko = 0;
		flags = 0;

4184 4185 4186 4187 4188 4189 4190
		/*
		 * bump off for our next call to get_extent
		 */
		off = extent_map_end(em);
		if (off >= max)
			end = 1;

4191
		if (em->block_start == EXTENT_MAP_LAST_BYTE) {
Y
Yehuda Sadeh 已提交
4192 4193
			end = 1;
			flags |= FIEMAP_EXTENT_LAST;
4194
		} else if (em->block_start == EXTENT_MAP_INLINE) {
Y
Yehuda Sadeh 已提交
4195 4196
			flags |= (FIEMAP_EXTENT_DATA_INLINE |
				  FIEMAP_EXTENT_NOT_ALIGNED);
4197
		} else if (em->block_start == EXTENT_MAP_DELALLOC) {
Y
Yehuda Sadeh 已提交
4198 4199
			flags |= (FIEMAP_EXTENT_DELALLOC |
				  FIEMAP_EXTENT_UNKNOWN);
4200
		} else {
4201
			disko = em->block_start + offset_in_extent;
Y
Yehuda Sadeh 已提交
4202 4203 4204 4205 4206 4207
		}
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
			flags |= FIEMAP_EXTENT_ENCODED;

		free_extent_map(em);
		em = NULL;
4208 4209
		if ((em_start >= last) || em_len == (u64)-1 ||
		   (last == (u64)-1 && isize <= em_end)) {
Y
Yehuda Sadeh 已提交
4210 4211 4212 4213
			flags |= FIEMAP_EXTENT_LAST;
			end = 1;
		}

4214 4215 4216 4217 4218 4219 4220 4221
		/* 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
4222 4223 4224
			flags |= FIEMAP_EXTENT_LAST;
			end = 1;
		}
4225 4226 4227 4228
		ret = fiemap_fill_next_extent(fieinfo, em_start, disko,
					      em_len, flags);
		if (ret)
			goto out_free;
Y
Yehuda Sadeh 已提交
4229 4230 4231 4232
	}
out_free:
	free_extent_map(em);
out:
L
Liu Bo 已提交
4233
	unlock_extent_cached(&BTRFS_I(inode)->io_tree, start, start + len - 1,
4234
			     &cached_state, GFP_NOFS);
Y
Yehuda Sadeh 已提交
4235 4236 4237
	return ret;
}

4238 4239
static void __free_extent_buffer(struct extent_buffer *eb)
{
4240
	btrfs_leak_debug_del(&eb->leak_list);
4241 4242 4243
	kmem_cache_free(extent_buffer_cache, eb);
}

4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313
static int extent_buffer_under_io(struct extent_buffer *eb)
{
	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.
 */
static void btrfs_release_extent_buffer_page(struct extent_buffer *eb,
						unsigned long start_idx)
{
	unsigned long index;
	unsigned long num_pages;
	struct page *page;
	int mapped = !test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags);

	BUG_ON(extent_buffer_under_io(eb));

	num_pages = num_extent_pages(eb->start, eb->len);
	index = start_idx + num_pages;
	if (start_idx >= index)
		return;

	do {
		index--;
		page = extent_buffer_page(eb, index);
		if (page && mapped) {
			spin_lock(&page->mapping->private_lock);
			/*
			 * 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));
				/*
				 * We need to make sure we haven't be attached
				 * to a new eb.
				 */
				ClearPagePrivate(page);
				set_page_private(page, 0);
				/* One for the page private */
				page_cache_release(page);
			}
			spin_unlock(&page->mapping->private_lock);

		}
		if (page) {
			/* One for when we alloced the page */
			page_cache_release(page);
		}
	} while (index != start_idx);
}

/*
 * Helper for releasing the extent buffer.
 */
static inline void btrfs_release_extent_buffer(struct extent_buffer *eb)
{
	btrfs_release_extent_buffer_page(eb, 0);
	__free_extent_buffer(eb);
}

4314 4315 4316 4317 4318 4319 4320 4321
static struct extent_buffer *__alloc_extent_buffer(struct extent_io_tree *tree,
						   u64 start,
						   unsigned long len,
						   gfp_t mask)
{
	struct extent_buffer *eb = NULL;

	eb = kmem_cache_zalloc(extent_buffer_cache, mask);
T
Tsutomu Itoh 已提交
4322 4323
	if (eb == NULL)
		return NULL;
4324 4325
	eb->start = start;
	eb->len = len;
J
Josef Bacik 已提交
4326
	eb->tree = tree;
4327
	eb->bflags = 0;
4328 4329 4330 4331 4332 4333 4334
	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);
4335
	eb->lock_nested = 0;
4336 4337
	init_waitqueue_head(&eb->write_lock_wq);
	init_waitqueue_head(&eb->read_lock_wq);
4338

4339 4340
	btrfs_leak_debug_add(&eb->leak_list, &buffers);

4341
	spin_lock_init(&eb->refs_lock);
4342
	atomic_set(&eb->refs, 1);
4343
	atomic_set(&eb->io_pages, 0);
4344

4345 4346 4347 4348 4349 4350
	/*
	 * 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);
4351 4352 4353 4354

	return eb;
}

4355 4356 4357 4358 4359 4360 4361
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);

4362
	new = __alloc_extent_buffer(NULL, src->start, src->len, GFP_NOFS);
4363 4364 4365 4366
	if (new == NULL)
		return NULL;

	for (i = 0; i < num_pages; i++) {
4367
		p = alloc_page(GFP_NOFS);
4368 4369 4370 4371
		if (!p) {
			btrfs_release_extent_buffer(new);
			return NULL;
		}
4372 4373 4374 4375 4376 4377 4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390
		attach_extent_buffer_page(new, p);
		WARN_ON(PageDirty(p));
		SetPageUptodate(p);
		new->pages[i] = p;
	}

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

	return new;
}

struct extent_buffer *alloc_dummy_extent_buffer(u64 start, unsigned long len)
{
	struct extent_buffer *eb;
	unsigned long num_pages = num_extent_pages(0, len);
	unsigned long i;

4391
	eb = __alloc_extent_buffer(NULL, start, len, GFP_NOFS);
4392 4393 4394 4395
	if (!eb)
		return NULL;

	for (i = 0; i < num_pages; i++) {
4396
		eb->pages[i] = alloc_page(GFP_NOFS);
4397 4398 4399 4400 4401 4402 4403 4404 4405
		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:
4406 4407
	for (; i > 0; i--)
		__free_page(eb->pages[i - 1]);
4408 4409 4410 4411
	__free_extent_buffer(eb);
	return NULL;
}

4412 4413
static void check_buffer_tree_ref(struct extent_buffer *eb)
{
4414
	int refs;
4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434
	/* 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.
	 */
4435 4436 4437 4438
	refs = atomic_read(&eb->refs);
	if (refs >= 2 && test_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
		return;

4439 4440
	spin_lock(&eb->refs_lock);
	if (!test_and_set_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
4441
		atomic_inc(&eb->refs);
4442
	spin_unlock(&eb->refs_lock);
4443 4444
}

4445 4446 4447 4448
static void mark_extent_buffer_accessed(struct extent_buffer *eb)
{
	unsigned long num_pages, i;

4449 4450
	check_buffer_tree_ref(eb);

4451 4452 4453 4454 4455 4456 4457
	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = 0; i < num_pages; i++) {
		struct page *p = extent_buffer_page(eb, i);
		mark_page_accessed(p);
	}
}

4458
struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree,
4459
					  u64 start, unsigned long len)
4460 4461 4462 4463 4464
{
	unsigned long num_pages = num_extent_pages(start, len);
	unsigned long i;
	unsigned long index = start >> PAGE_CACHE_SHIFT;
	struct extent_buffer *eb;
4465
	struct extent_buffer *exists = NULL;
4466 4467 4468
	struct page *p;
	struct address_space *mapping = tree->mapping;
	int uptodate = 1;
4469
	int ret;
4470

4471 4472 4473 4474
	rcu_read_lock();
	eb = radix_tree_lookup(&tree->buffer, start >> PAGE_CACHE_SHIFT);
	if (eb && atomic_inc_not_zero(&eb->refs)) {
		rcu_read_unlock();
4475
		mark_extent_buffer_accessed(eb);
4476 4477
		return eb;
	}
4478
	rcu_read_unlock();
4479

4480
	eb = __alloc_extent_buffer(tree, start, len, GFP_NOFS);
4481
	if (!eb)
4482 4483
		return NULL;

4484
	for (i = 0; i < num_pages; i++, index++) {
4485
		p = find_or_create_page(mapping, index, GFP_NOFS);
4486
		if (!p)
4487
			goto free_eb;
J
Josef Bacik 已提交
4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501

		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);
4502
				page_cache_release(p);
4503
				mark_extent_buffer_accessed(exists);
J
Josef Bacik 已提交
4504 4505 4506
				goto free_eb;
			}

4507
			/*
J
Josef Bacik 已提交
4508 4509 4510 4511
			 * Do this so attach doesn't complain and we need to
			 * drop the ref the old guy had.
			 */
			ClearPagePrivate(p);
4512
			WARN_ON(PageDirty(p));
J
Josef Bacik 已提交
4513
			page_cache_release(p);
4514
		}
J
Josef Bacik 已提交
4515 4516
		attach_extent_buffer_page(eb, p);
		spin_unlock(&mapping->private_lock);
4517
		WARN_ON(PageDirty(p));
4518
		mark_page_accessed(p);
4519
		eb->pages[i] = p;
4520 4521
		if (!PageUptodate(p))
			uptodate = 0;
C
Chris Mason 已提交
4522 4523 4524 4525 4526

		/*
		 * see below about how we avoid a nasty race with release page
		 * and why we unlock later
		 */
4527 4528
	}
	if (uptodate)
4529
		set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
4530
again:
4531 4532 4533 4534
	ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
	if (ret)
		goto free_eb;

4535
	spin_lock(&tree->buffer_lock);
4536 4537 4538 4539
	ret = radix_tree_insert(&tree->buffer, start >> PAGE_CACHE_SHIFT, eb);
	if (ret == -EEXIST) {
		exists = radix_tree_lookup(&tree->buffer,
						start >> PAGE_CACHE_SHIFT);
4540 4541 4542 4543 4544 4545
		if (!atomic_inc_not_zero(&exists->refs)) {
			spin_unlock(&tree->buffer_lock);
			radix_tree_preload_end();
			exists = NULL;
			goto again;
		}
4546
		spin_unlock(&tree->buffer_lock);
4547
		radix_tree_preload_end();
4548
		mark_extent_buffer_accessed(exists);
4549 4550 4551
		goto free_eb;
	}
	/* add one reference for the tree */
4552
	check_buffer_tree_ref(eb);
4553
	spin_unlock(&tree->buffer_lock);
4554
	radix_tree_preload_end();
C
Chris Mason 已提交
4555 4556 4557 4558 4559 4560 4561 4562 4563 4564

	/*
	 * 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
	 */
4565 4566 4567 4568 4569 4570 4571
	SetPageChecked(eb->pages[0]);
	for (i = 1; i < num_pages; i++) {
		p = extent_buffer_page(eb, i);
		ClearPageChecked(p);
		unlock_page(p);
	}
	unlock_page(eb->pages[0]);
4572 4573
	return eb;

4574
free_eb:
4575 4576 4577 4578
	for (i = 0; i < num_pages; i++) {
		if (eb->pages[i])
			unlock_page(eb->pages[i]);
	}
C
Chris Mason 已提交
4579

4580
	WARN_ON(!atomic_dec_and_test(&eb->refs));
4581
	btrfs_release_extent_buffer(eb);
4582
	return exists;
4583 4584 4585
}

struct extent_buffer *find_extent_buffer(struct extent_io_tree *tree,
4586
					 u64 start, unsigned long len)
4587 4588 4589
{
	struct extent_buffer *eb;

4590 4591 4592 4593
	rcu_read_lock();
	eb = radix_tree_lookup(&tree->buffer, start >> PAGE_CACHE_SHIFT);
	if (eb && atomic_inc_not_zero(&eb->refs)) {
		rcu_read_unlock();
4594
		mark_extent_buffer_accessed(eb);
4595 4596 4597
		return eb;
	}
	rcu_read_unlock();
J
Josef Bacik 已提交
4598

4599
	return NULL;
4600 4601
}

4602 4603 4604 4605 4606 4607 4608 4609 4610
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 */
4611
static int release_extent_buffer(struct extent_buffer *eb)
4612 4613 4614
{
	WARN_ON(atomic_read(&eb->refs) == 0);
	if (atomic_dec_and_test(&eb->refs)) {
4615 4616 4617 4618
		if (test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags)) {
			spin_unlock(&eb->refs_lock);
		} else {
			struct extent_io_tree *tree = eb->tree;
4619

4620
			spin_unlock(&eb->refs_lock);
4621

4622 4623 4624 4625 4626
			spin_lock(&tree->buffer_lock);
			radix_tree_delete(&tree->buffer,
					  eb->start >> PAGE_CACHE_SHIFT);
			spin_unlock(&tree->buffer_lock);
		}
4627 4628 4629 4630

		/* Should be safe to release our pages at this point */
		btrfs_release_extent_buffer_page(eb, 0);
		call_rcu(&eb->rcu_head, btrfs_release_extent_buffer_rcu);
4631
		return 1;
4632 4633
	}
	spin_unlock(&eb->refs_lock);
4634 4635

	return 0;
4636 4637
}

4638 4639
void free_extent_buffer(struct extent_buffer *eb)
{
4640 4641
	int refs;
	int old;
4642 4643 4644
	if (!eb)
		return;

4645 4646 4647 4648 4649 4650 4651 4652 4653
	while (1) {
		refs = atomic_read(&eb->refs);
		if (refs <= 3)
			break;
		old = atomic_cmpxchg(&eb->refs, refs, refs - 1);
		if (old == refs)
			return;
	}

4654
	spin_lock(&eb->refs_lock);
4655 4656 4657 4658
	if (atomic_read(&eb->refs) == 2 &&
	    test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags))
		atomic_dec(&eb->refs);

4659 4660
	if (atomic_read(&eb->refs) == 2 &&
	    test_bit(EXTENT_BUFFER_STALE, &eb->bflags) &&
4661
	    !extent_buffer_under_io(eb) &&
4662 4663 4664 4665 4666 4667 4668
	    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.
	 */
4669
	release_extent_buffer(eb);
4670 4671 4672 4673 4674
}

void free_extent_buffer_stale(struct extent_buffer *eb)
{
	if (!eb)
4675 4676
		return;

4677 4678 4679
	spin_lock(&eb->refs_lock);
	set_bit(EXTENT_BUFFER_STALE, &eb->bflags);

4680
	if (atomic_read(&eb->refs) == 2 && !extent_buffer_under_io(eb) &&
4681 4682
	    test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
		atomic_dec(&eb->refs);
4683
	release_extent_buffer(eb);
4684 4685
}

4686
void clear_extent_buffer_dirty(struct extent_buffer *eb)
4687 4688 4689 4690 4691 4692 4693 4694 4695
{
	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++) {
		page = extent_buffer_page(eb, i);
4696
		if (!PageDirty(page))
C
Chris Mason 已提交
4697 4698
			continue;

4699
		lock_page(page);
C
Chris Mason 已提交
4700 4701
		WARN_ON(!PagePrivate(page));

4702
		clear_page_dirty_for_io(page);
4703
		spin_lock_irq(&page->mapping->tree_lock);
4704 4705 4706 4707 4708
		if (!PageDirty(page)) {
			radix_tree_tag_clear(&page->mapping->page_tree,
						page_index(page),
						PAGECACHE_TAG_DIRTY);
		}
4709
		spin_unlock_irq(&page->mapping->tree_lock);
4710
		ClearPageError(page);
4711
		unlock_page(page);
4712
	}
4713
	WARN_ON(atomic_read(&eb->refs) == 0);
4714 4715
}

4716
int set_extent_buffer_dirty(struct extent_buffer *eb)
4717 4718 4719
{
	unsigned long i;
	unsigned long num_pages;
4720
	int was_dirty = 0;
4721

4722 4723
	check_buffer_tree_ref(eb);

4724
	was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
4725

4726
	num_pages = num_extent_pages(eb->start, eb->len);
4727
	WARN_ON(atomic_read(&eb->refs) == 0);
4728 4729
	WARN_ON(!test_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags));

4730
	for (i = 0; i < num_pages; i++)
4731
		set_page_dirty(extent_buffer_page(eb, i));
4732
	return was_dirty;
4733 4734
}

4735
int clear_extent_buffer_uptodate(struct extent_buffer *eb)
4736 4737 4738 4739 4740
{
	unsigned long i;
	struct page *page;
	unsigned long num_pages;

4741
	clear_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
4742
	num_pages = num_extent_pages(eb->start, eb->len);
4743 4744
	for (i = 0; i < num_pages; i++) {
		page = extent_buffer_page(eb, i);
C
Chris Mason 已提交
4745 4746
		if (page)
			ClearPageUptodate(page);
4747 4748 4749 4750
	}
	return 0;
}

4751
int set_extent_buffer_uptodate(struct extent_buffer *eb)
4752 4753 4754 4755 4756
{
	unsigned long i;
	struct page *page;
	unsigned long num_pages;

4757
	set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
4758 4759 4760 4761 4762 4763 4764 4765
	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = 0; i < num_pages; i++) {
		page = extent_buffer_page(eb, i);
		SetPageUptodate(page);
	}
	return 0;
}

4766
int extent_buffer_uptodate(struct extent_buffer *eb)
4767
{
4768
	return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
4769 4770 4771
}

int read_extent_buffer_pages(struct extent_io_tree *tree,
4772
			     struct extent_buffer *eb, u64 start, int wait,
4773
			     get_extent_t *get_extent, int mirror_num)
4774 4775 4776 4777 4778 4779
{
	unsigned long i;
	unsigned long start_i;
	struct page *page;
	int err;
	int ret = 0;
4780 4781
	int locked_pages = 0;
	int all_uptodate = 1;
4782
	unsigned long num_pages;
4783
	unsigned long num_reads = 0;
4784
	struct bio *bio = NULL;
C
Chris Mason 已提交
4785
	unsigned long bio_flags = 0;
4786

4787
	if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags))
4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 4798 4799 4800
		return 0;

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

	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = start_i; i < num_pages; i++) {
		page = extent_buffer_page(eb, i);
4801
		if (wait == WAIT_NONE) {
4802
			if (!trylock_page(page))
4803
				goto unlock_exit;
4804 4805 4806
		} else {
			lock_page(page);
		}
4807
		locked_pages++;
4808 4809
		if (!PageUptodate(page)) {
			num_reads++;
4810
			all_uptodate = 0;
4811
		}
4812 4813 4814
	}
	if (all_uptodate) {
		if (start_i == 0)
4815
			set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
4816 4817 4818
		goto unlock_exit;
	}

4819
	clear_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
4820
	eb->read_mirror = 0;
4821
	atomic_set(&eb->io_pages, num_reads);
4822 4823 4824
	for (i = start_i; i < num_pages; i++) {
		page = extent_buffer_page(eb, i);
		if (!PageUptodate(page)) {
4825
			ClearPageError(page);
4826
			err = __extent_read_full_page(tree, page,
4827
						      get_extent, &bio,
4828 4829
						      mirror_num, &bio_flags,
						      READ | REQ_META);
C
Chris Mason 已提交
4830
			if (err)
4831 4832 4833 4834 4835 4836
				ret = err;
		} else {
			unlock_page(page);
		}
	}

4837
	if (bio) {
4838 4839
		err = submit_one_bio(READ | REQ_META, bio, mirror_num,
				     bio_flags);
4840 4841
		if (err)
			return err;
4842
	}
4843

4844
	if (ret || wait != WAIT_COMPLETE)
4845
		return ret;
C
Chris Mason 已提交
4846

4847 4848 4849
	for (i = start_i; i < num_pages; i++) {
		page = extent_buffer_page(eb, i);
		wait_on_page_locked(page);
C
Chris Mason 已提交
4850
		if (!PageUptodate(page))
4851 4852
			ret = -EIO;
	}
C
Chris Mason 已提交
4853

4854
	return ret;
4855 4856 4857

unlock_exit:
	i = start_i;
C
Chris Mason 已提交
4858
	while (locked_pages > 0) {
4859 4860 4861 4862 4863 4864
		page = extent_buffer_page(eb, i);
		i++;
		unlock_page(page);
		locked_pages--;
	}
	return ret;
4865 4866 4867 4868 4869 4870 4871 4872 4873 4874 4875 4876 4877 4878 4879 4880 4881
}

void read_extent_buffer(struct extent_buffer *eb, void *dstv,
			unsigned long start,
			unsigned long len)
{
	size_t cur;
	size_t offset;
	struct page *page;
	char *kaddr;
	char *dst = (char *)dstv;
	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;

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

4882
	offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
4883

C
Chris Mason 已提交
4884
	while (len > 0) {
4885 4886 4887
		page = extent_buffer_page(eb, i);

		cur = min(len, (PAGE_CACHE_SIZE - offset));
4888
		kaddr = page_address(page);
4889 4890 4891 4892 4893 4894 4895 4896 4897 4898
		memcpy(dst, kaddr + offset, cur);

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

int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start,
4899
			       unsigned long min_len, char **map,
4900
			       unsigned long *map_start,
4901
			       unsigned long *map_len)
4902 4903 4904 4905 4906 4907 4908 4909 4910 4911 4912 4913 4914 4915 4916 4917 4918 4919 4920
{
	size_t offset = start & (PAGE_CACHE_SIZE - 1);
	char *kaddr;
	struct page *p;
	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
	unsigned long end_i = (start_offset + start + min_len - 1) >>
		PAGE_CACHE_SHIFT;

	if (i != end_i)
		return -EINVAL;

	if (i == 0) {
		offset = start_offset;
		*map_start = 0;
	} else {
		offset = 0;
		*map_start = ((u64)i << PAGE_CACHE_SHIFT) - start_offset;
	}
C
Chris Mason 已提交
4921

4922
	if (start + min_len > eb->len) {
J
Julia Lawall 已提交
4923
		WARN(1, KERN_ERR "btrfs bad mapping eb start %llu len %lu, "
4924 4925
		       "wanted %lu %lu\n",
		       eb->start, eb->len, start, min_len);
4926
		return -EINVAL;
4927 4928 4929
	}

	p = extent_buffer_page(eb, i);
4930
	kaddr = page_address(p);
4931 4932 4933 4934 4935 4936 4937 4938 4939 4940 4941 4942 4943 4944 4945 4946 4947 4948 4949 4950 4951
	*map = kaddr + offset;
	*map_len = PAGE_CACHE_SIZE - offset;
	return 0;
}

int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv,
			  unsigned long start,
			  unsigned long len)
{
	size_t cur;
	size_t offset;
	struct page *page;
	char *kaddr;
	char *ptr = (char *)ptrv;
	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
	int ret = 0;

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

4952
	offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
4953

C
Chris Mason 已提交
4954
	while (len > 0) {
4955 4956 4957 4958
		page = extent_buffer_page(eb, i);

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

4959
		kaddr = page_address(page);
4960 4961 4962 4963 4964 4965 4966 4967 4968 4969 4970 4971 4972 4973 4974 4975 4976 4977 4978 4979 4980 4981 4982 4983 4984 4985
		ret = memcmp(ptr, kaddr + offset, cur);
		if (ret)
			break;

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

void write_extent_buffer(struct extent_buffer *eb, const void *srcv,
			 unsigned long start, unsigned long len)
{
	size_t cur;
	size_t offset;
	struct page *page;
	char *kaddr;
	char *src = (char *)srcv;
	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;

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

4986
	offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
4987

C
Chris Mason 已提交
4988
	while (len > 0) {
4989 4990 4991 4992
		page = extent_buffer_page(eb, i);
		WARN_ON(!PageUptodate(page));

		cur = min(len, PAGE_CACHE_SIZE - offset);
4993
		kaddr = page_address(page);
4994 4995 4996 4997 4998 4999 5000 5001 5002 5003 5004 5005 5006 5007 5008 5009 5010 5011 5012 5013 5014 5015
		memcpy(kaddr + offset, src, cur);

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

void memset_extent_buffer(struct extent_buffer *eb, char c,
			  unsigned long start, unsigned long len)
{
	size_t cur;
	size_t offset;
	struct page *page;
	char *kaddr;
	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;

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

5016
	offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
5017

C
Chris Mason 已提交
5018
	while (len > 0) {
5019 5020 5021 5022
		page = extent_buffer_page(eb, i);
		WARN_ON(!PageUptodate(page));

		cur = min(len, PAGE_CACHE_SIZE - offset);
5023
		kaddr = page_address(page);
5024 5025 5026 5027 5028 5029 5030 5031 5032 5033 5034 5035 5036 5037 5038 5039 5040 5041 5042 5043 5044 5045 5046
		memset(kaddr + offset, c, cur);

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

void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src,
			unsigned long dst_offset, unsigned long src_offset,
			unsigned long len)
{
	u64 dst_len = dst->len;
	size_t cur;
	size_t offset;
	struct page *page;
	char *kaddr;
	size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1);
	unsigned long i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT;

	WARN_ON(src->len != dst_len);

	offset = (start_offset + dst_offset) &
5047
		(PAGE_CACHE_SIZE - 1);
5048

C
Chris Mason 已提交
5049
	while (len > 0) {
5050 5051 5052 5053 5054
		page = extent_buffer_page(dst, i);
		WARN_ON(!PageUptodate(page));

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

5055
		kaddr = page_address(page);
5056 5057 5058 5059 5060 5061 5062 5063 5064 5065 5066 5067 5068
		read_extent_buffer(src, kaddr + offset, src_offset, cur);

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

static void move_pages(struct page *dst_page, struct page *src_page,
		       unsigned long dst_off, unsigned long src_off,
		       unsigned long len)
{
5069
	char *dst_kaddr = page_address(dst_page);
5070 5071 5072
	if (dst_page == src_page) {
		memmove(dst_kaddr + dst_off, dst_kaddr + src_off, len);
	} else {
5073
		char *src_kaddr = page_address(src_page);
5074 5075 5076 5077 5078 5079 5080 5081
		char *p = dst_kaddr + dst_off + len;
		char *s = src_kaddr + src_off + len;

		while (len--)
			*--p = *--s;
	}
}

5082 5083 5084 5085 5086 5087
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;
}

5088 5089 5090 5091
static void copy_pages(struct page *dst_page, struct page *src_page,
		       unsigned long dst_off, unsigned long src_off,
		       unsigned long len)
{
5092
	char *dst_kaddr = page_address(dst_page);
5093
	char *src_kaddr;
5094
	int must_memmove = 0;
5095

5096
	if (dst_page != src_page) {
5097
		src_kaddr = page_address(src_page);
5098
	} else {
5099
		src_kaddr = dst_kaddr;
5100 5101
		if (areas_overlap(src_off, dst_off, len))
			must_memmove = 1;
5102
	}
5103

5104 5105 5106 5107
	if (must_memmove)
		memmove(dst_kaddr + dst_off, src_kaddr + src_off, len);
	else
		memcpy(dst_kaddr + dst_off, src_kaddr + src_off, len);
5108 5109 5110 5111 5112 5113 5114 5115 5116 5117 5118 5119 5120
}

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

	if (src_offset + len > dst->len) {
C
Chris Mason 已提交
5121 5122
		printk(KERN_ERR "btrfs memmove bogus src_offset %lu move "
		       "len %lu dst len %lu\n", src_offset, len, dst->len);
5123 5124 5125
		BUG_ON(1);
	}
	if (dst_offset + len > dst->len) {
C
Chris Mason 已提交
5126 5127
		printk(KERN_ERR "btrfs memmove bogus dst_offset %lu move "
		       "len %lu dst len %lu\n", dst_offset, len, dst->len);
5128 5129 5130
		BUG_ON(1);
	}

C
Chris Mason 已提交
5131
	while (len > 0) {
5132
		dst_off_in_page = (start_offset + dst_offset) &
5133
			(PAGE_CACHE_SIZE - 1);
5134
		src_off_in_page = (start_offset + src_offset) &
5135
			(PAGE_CACHE_SIZE - 1);
5136 5137 5138 5139 5140 5141 5142 5143 5144 5145 5146 5147 5148 5149 5150 5151 5152 5153 5154 5155 5156 5157 5158 5159 5160 5161 5162 5163 5164 5165 5166 5167

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

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

		copy_pages(extent_buffer_page(dst, dst_i),
			   extent_buffer_page(dst, src_i),
			   dst_off_in_page, src_off_in_page, cur);

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

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

	if (src_offset + len > dst->len) {
C
Chris Mason 已提交
5168 5169
		printk(KERN_ERR "btrfs memmove bogus src_offset %lu move "
		       "len %lu len %lu\n", src_offset, len, dst->len);
5170 5171 5172
		BUG_ON(1);
	}
	if (dst_offset + len > dst->len) {
C
Chris Mason 已提交
5173 5174
		printk(KERN_ERR "btrfs memmove bogus dst_offset %lu move "
		       "len %lu len %lu\n", dst_offset, len, dst->len);
5175 5176
		BUG_ON(1);
	}
5177
	if (dst_offset < src_offset) {
5178 5179 5180
		memcpy_extent_buffer(dst, dst_offset, src_offset, len);
		return;
	}
C
Chris Mason 已提交
5181
	while (len > 0) {
5182 5183 5184 5185
		dst_i = (start_offset + dst_end) >> PAGE_CACHE_SHIFT;
		src_i = (start_offset + src_end) >> PAGE_CACHE_SHIFT;

		dst_off_in_page = (start_offset + dst_end) &
5186
			(PAGE_CACHE_SIZE - 1);
5187
		src_off_in_page = (start_offset + src_end) &
5188
			(PAGE_CACHE_SIZE - 1);
5189 5190 5191 5192 5193 5194 5195 5196 5197 5198 5199 5200 5201

		cur = min_t(unsigned long, len, src_off_in_page + 1);
		cur = min(cur, dst_off_in_page + 1);
		move_pages(extent_buffer_page(dst, dst_i),
			   extent_buffer_page(dst, src_i),
			   dst_off_in_page - cur + 1,
			   src_off_in_page - cur + 1, cur);

		dst_end -= cur;
		src_end -= cur;
		len -= cur;
	}
}
5202

5203
int try_release_extent_buffer(struct page *page)
5204
{
5205 5206
	struct extent_buffer *eb;

5207 5208 5209 5210 5211 5212 5213
	/*
	 * We need to make sure noboody is attaching this page to an eb right
	 * now.
	 */
	spin_lock(&page->mapping->private_lock);
	if (!PagePrivate(page)) {
		spin_unlock(&page->mapping->private_lock);
J
Josef Bacik 已提交
5214
		return 1;
5215
	}
5216

5217 5218
	eb = (struct extent_buffer *)page->private;
	BUG_ON(!eb);
5219 5220

	/*
5221 5222 5223
	 * 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.
5224
	 */
5225
	spin_lock(&eb->refs_lock);
5226
	if (atomic_read(&eb->refs) != 1 || extent_buffer_under_io(eb)) {
5227 5228 5229
		spin_unlock(&eb->refs_lock);
		spin_unlock(&page->mapping->private_lock);
		return 0;
5230
	}
5231
	spin_unlock(&page->mapping->private_lock);
5232

5233
	/*
5234 5235
	 * 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.
5236
	 */
5237 5238 5239
	if (!test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags)) {
		spin_unlock(&eb->refs_lock);
		return 0;
5240
	}
5241

5242
	return release_extent_buffer(eb);
5243
}