extent_io.c 130.4 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 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79

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",
		       (unsigned long long)state->start,
		       (unsigned long long)state->end,
		       state->state, state->tree, atomic_read(&state->refs));
		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 "
		       "refs %d\n", (unsigned long long)eb->start,
		       eb->len, atomic_read(&eb->refs));
		list_del(&eb->leak_list);
		kmem_cache_free(extent_buffer_cache, eb);
	}
}
80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97

#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",
				caller,
				(unsigned long long)btrfs_ino(inode),
				(unsigned long long)isize,
				(unsigned long long)start,
				(unsigned long long)end);
	}
}
98 99 100 101
#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)
102
#define btrfs_debug_check_extent_io_range(c, s, e)	do {} while (0)
C
Chris Mason 已提交
103
#endif
104 105 106 107 108 109 110 111 112 113 114 115 116

#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;
117
	unsigned long bio_flags;
118 119 120 121

	/* tells writepage not to lock the state bits for this range
	 * it still does the unlocking
	 */
122 123 124 125
	unsigned int extent_locked:1;

	/* tells the submit_bio code to use a WRITE_SYNC */
	unsigned int sync_io:1;
126 127
};

128
static noinline void flush_write_bio(void *data);
129 130 131 132 133
static inline struct btrfs_fs_info *
tree_fs_info(struct extent_io_tree *tree)
{
	return btrfs_sb(tree->mapping->host->i_sb);
}
134

135 136
int __init extent_io_init(void)
{
D
David Sterba 已提交
137
	extent_state_cache = kmem_cache_create("btrfs_extent_state",
138 139
			sizeof(struct extent_state), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
140 141 142
	if (!extent_state_cache)
		return -ENOMEM;

D
David Sterba 已提交
143
	extent_buffer_cache = kmem_cache_create("btrfs_extent_buffer",
144 145
			sizeof(struct extent_buffer), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
146 147
	if (!extent_buffer_cache)
		goto free_state_cache;
148 149 150 151 152

	btrfs_bioset = bioset_create(BIO_POOL_SIZE,
				     offsetof(struct btrfs_io_bio, bio));
	if (!btrfs_bioset)
		goto free_buffer_cache;
153 154
	return 0;

155 156 157 158
free_buffer_cache:
	kmem_cache_destroy(extent_buffer_cache);
	extent_buffer_cache = NULL;

159 160
free_state_cache:
	kmem_cache_destroy(extent_state_cache);
161
	extent_state_cache = NULL;
162 163 164 165 166
	return -ENOMEM;
}

void extent_io_exit(void)
{
167
	btrfs_leak_debug_check();
168 169 170 171 172 173

	/*
	 * Make sure all delayed rcu free are flushed before we
	 * destroy caches.
	 */
	rcu_barrier();
174 175 176 177
	if (extent_state_cache)
		kmem_cache_destroy(extent_state_cache);
	if (extent_buffer_cache)
		kmem_cache_destroy(extent_buffer_cache);
178 179
	if (btrfs_bioset)
		bioset_free(btrfs_bioset);
180 181 182
}

void extent_io_tree_init(struct extent_io_tree *tree,
183
			 struct address_space *mapping)
184
{
185
	tree->state = RB_ROOT;
186
	INIT_RADIX_TREE(&tree->buffer, GFP_ATOMIC);
187 188
	tree->ops = NULL;
	tree->dirty_bytes = 0;
189
	spin_lock_init(&tree->lock);
190
	spin_lock_init(&tree->buffer_lock);
191 192 193
	tree->mapping = mapping;
}

194
static struct extent_state *alloc_extent_state(gfp_t mask)
195 196 197 198
{
	struct extent_state *state;

	state = kmem_cache_alloc(extent_state_cache, mask);
199
	if (!state)
200 201 202
		return state;
	state->state = 0;
	state->private = 0;
203
	state->tree = NULL;
204
	btrfs_leak_debug_add(&state->leak_list, &states);
205 206
	atomic_set(&state->refs, 1);
	init_waitqueue_head(&state->wq);
207
	trace_alloc_extent_state(state, mask, _RET_IP_);
208 209 210
	return state;
}

211
void free_extent_state(struct extent_state *state)
212 213 214 215
{
	if (!state)
		return;
	if (atomic_dec_and_test(&state->refs)) {
216
		WARN_ON(state->tree);
217
		btrfs_leak_debug_del(&state->leak_list);
218
		trace_free_extent_state(state, _RET_IP_);
219 220 221 222 223 224 225
		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
226 227
	struct rb_node **p = &root->rb_node;
	struct rb_node *parent = NULL;
228 229
	struct tree_entry *entry;

C
Chris Mason 已提交
230
	while (*p) {
231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246
		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;
}

247
static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset,
248 249 250
				     struct rb_node **prev_ret,
				     struct rb_node **next_ret)
{
251
	struct rb_root *root = &tree->state;
C
Chris Mason 已提交
252
	struct rb_node *n = root->rb_node;
253 254 255 256 257
	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
258
	while (n) {
259 260 261 262 263 264 265 266
		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
267
		else
268 269 270 271 272
			return n;
	}

	if (prev_ret) {
		orig_prev = prev;
C
Chris Mason 已提交
273
		while (prev && offset > prev_entry->end) {
274 275 276 277 278 279 280 281 282
			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
283
		while (prev && offset < prev_entry->start) {
284 285 286 287 288 289 290 291
			prev = rb_prev(prev);
			prev_entry = rb_entry(prev, struct tree_entry, rb_node);
		}
		*next_ret = prev;
	}
	return NULL;
}

292 293
static inline struct rb_node *tree_search(struct extent_io_tree *tree,
					  u64 offset)
294
{
295
	struct rb_node *prev = NULL;
296
	struct rb_node *ret;
297

298
	ret = __etree_search(tree, offset, &prev, NULL);
C
Chris Mason 已提交
299
	if (!ret)
300 301 302 303
		return prev;
	return ret;
}

J
Josef Bacik 已提交
304 305 306 307 308 309 310 311
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);
}

312 313 314 315 316 317 318 319 320
/*
 * 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.
 */
321 322
static void merge_state(struct extent_io_tree *tree,
		        struct extent_state *state)
323 324 325 326
{
	struct extent_state *other;
	struct rb_node *other_node;

327
	if (state->state & (EXTENT_IOBITS | EXTENT_BOUNDARY))
328
		return;
329 330 331 332 333 334

	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
335
			merge_cb(tree, state, other);
336
			state->start = other->start;
337
			other->tree = NULL;
338 339 340 341 342 343 344 345 346
			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
347
			merge_cb(tree, state, other);
348 349 350 351
			state->end = other->end;
			other->tree = NULL;
			rb_erase(&other->rb_node, &tree->state);
			free_extent_state(other);
352 353 354 355
		}
	}
}

356
static void set_state_cb(struct extent_io_tree *tree,
357
			 struct extent_state *state, unsigned long *bits)
358
{
359 360
	if (tree->ops && tree->ops->set_bit_hook)
		tree->ops->set_bit_hook(tree->mapping->host, state, bits);
361 362 363
}

static void clear_state_cb(struct extent_io_tree *tree,
364
			   struct extent_state *state, unsigned long *bits)
365
{
J
Josef Bacik 已提交
366 367
	if (tree->ops && tree->ops->clear_bit_hook)
		tree->ops->clear_bit_hook(tree->mapping->host, state, bits);
368 369
}

370
static void set_state_bits(struct extent_io_tree *tree,
371
			   struct extent_state *state, unsigned long *bits);
372

373 374 375 376 377 378 379 380 381 382 383 384
/*
 * 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,
385
			unsigned long *bits)
386 387 388
{
	struct rb_node *node;

J
Julia Lawall 已提交
389 390
	if (end < start)
		WARN(1, KERN_ERR "btrfs end < start %llu %llu\n",
C
Chris Mason 已提交
391 392
		       (unsigned long long)end,
		       (unsigned long long)start);
393 394
	state->start = start;
	state->end = end;
J
Josef Bacik 已提交
395

396 397
	set_state_bits(tree, state, bits);

398 399 400 401
	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
402 403 404 405
		printk(KERN_ERR "btrfs found node %llu %llu on insert of "
		       "%llu %llu\n", (unsigned long long)found->start,
		       (unsigned long long)found->end,
		       (unsigned long long)start, (unsigned long long)end);
406 407
		return -EEXIST;
	}
408
	state->tree = tree;
409 410 411 412
	merge_state(tree, state);
	return 0;
}

413
static void split_cb(struct extent_io_tree *tree, struct extent_state *orig,
J
Josef Bacik 已提交
414 415 416
		     u64 split)
{
	if (tree->ops && tree->ops->split_extent_hook)
417
		tree->ops->split_extent_hook(tree->mapping->host, orig, split);
J
Josef Bacik 已提交
418 419
}

420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437
/*
 * 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
438 439 440

	split_cb(tree, orig, split);

441 442 443 444 445 446 447 448 449 450
	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;
	}
451
	prealloc->tree = tree;
452 453 454
	return 0;
}

455 456 457 458 459 460 461 462 463
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;
}

464 465
/*
 * utility function to clear some bits in an extent state struct.
466
 * it will optionally wake up any one waiting on this state (wake == 1).
467 468 469 470
 *
 * If no bits are set on the state struct after clearing things, the
 * struct is freed and removed from the tree
 */
471 472
static struct extent_state *clear_state_bit(struct extent_io_tree *tree,
					    struct extent_state *state,
473
					    unsigned long *bits, int wake)
474
{
475
	struct extent_state *next;
476
	unsigned long bits_to_clear = *bits & ~EXTENT_CTLBITS;
477

478
	if ((bits_to_clear & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) {
479 480 481 482
		u64 range = state->end - state->start + 1;
		WARN_ON(range > tree->dirty_bytes);
		tree->dirty_bytes -= range;
	}
483
	clear_state_cb(tree, state, bits);
484
	state->state &= ~bits_to_clear;
485 486
	if (wake)
		wake_up(&state->wq);
487
	if (state->state == 0) {
488
		next = next_state(state);
489
		if (state->tree) {
490
			rb_erase(&state->rb_node, &tree->state);
491
			state->tree = NULL;
492 493 494 495 496 497
			free_extent_state(state);
		} else {
			WARN_ON(1);
		}
	} else {
		merge_state(tree, state);
498
		next = next_state(state);
499
	}
500
	return next;
501 502
}

503 504 505 506 507 508 509 510 511
static struct extent_state *
alloc_extent_state_atomic(struct extent_state *prealloc)
{
	if (!prealloc)
		prealloc = alloc_extent_state(GFP_ATOMIC);

	return prealloc;
}

512
static void extent_io_tree_panic(struct extent_io_tree *tree, int err)
513 514 515 516 517 518
{
	btrfs_panic(tree_fs_info(tree), err, "Locking error: "
		    "Extent tree was modified by another "
		    "thread while locked.");
}

519 520 521 522 523 524 525 526 527 528
/*
 * 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.
 *
529
 * This takes the tree lock, and returns 0 on success and < 0 on error.
530 531
 */
int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
532
		     unsigned long bits, int wake, int delete,
533 534
		     struct extent_state **cached_state,
		     gfp_t mask)
535 536
{
	struct extent_state *state;
537
	struct extent_state *cached;
538 539
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
540
	u64 last_end;
541
	int err;
542
	int clear = 0;
543

544 545
	btrfs_debug_check_extent_io_range(tree->mapping->host, start, end);

546 547 548
	if (bits & EXTENT_DELALLOC)
		bits |= EXTENT_NORESERVE;

549 550 551 552
	if (delete)
		bits |= ~EXTENT_CTLBITS;
	bits |= EXTENT_FIRST_DELALLOC;

553 554
	if (bits & (EXTENT_IOBITS | EXTENT_BOUNDARY))
		clear = 1;
555 556 557 558 559 560 561
again:
	if (!prealloc && (mask & __GFP_WAIT)) {
		prealloc = alloc_extent_state(mask);
		if (!prealloc)
			return -ENOMEM;
	}

562
	spin_lock(&tree->lock);
563 564
	if (cached_state) {
		cached = *cached_state;
565 566 567 568 569 570

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

571 572
		if (cached && cached->tree && cached->start <= start &&
		    cached->end > start) {
573 574
			if (clear)
				atomic_dec(&cached->refs);
575
			state = cached;
576
			goto hit_next;
577
		}
578 579
		if (clear)
			free_extent_state(cached);
580
	}
581 582 583 584
	/*
	 * this search will find the extents that end after
	 * our range starts
	 */
585
	node = tree_search(tree, start);
586 587 588
	if (!node)
		goto out;
	state = rb_entry(node, struct extent_state, rb_node);
589
hit_next:
590 591 592
	if (state->start > end)
		goto out;
	WARN_ON(state->end < start);
593
	last_end = state->end;
594

595
	/* the state doesn't have the wanted bits, go ahead */
596 597
	if (!(state->state & bits)) {
		state = next_state(state);
598
		goto next;
599
	}
600

601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617
	/*
	 *     | ---- 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) {
618 619
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
620
		err = split_state(tree, state, prealloc, start);
621 622 623
		if (err)
			extent_io_tree_panic(tree, err);

624 625 626 627
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
628 629
			state = clear_state_bit(tree, state, &bits, wake);
			goto next;
630 631 632 633 634 635 636 637 638 639
		}
		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) {
640 641
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
642
		err = split_state(tree, state, prealloc, end + 1);
643 644 645
		if (err)
			extent_io_tree_panic(tree, err);

646 647
		if (wake)
			wake_up(&state->wq);
648

649
		clear_state_bit(tree, prealloc, &bits, wake);
J
Josef Bacik 已提交
650

651 652 653
		prealloc = NULL;
		goto out;
	}
654

655
	state = clear_state_bit(tree, state, &bits, wake);
656
next:
657 658 659
	if (last_end == (u64)-1)
		goto out;
	start = last_end + 1;
660
	if (start <= end && state && !need_resched())
661
		goto hit_next;
662 663 664
	goto search_again;

out:
665
	spin_unlock(&tree->lock);
666 667 668
	if (prealloc)
		free_extent_state(prealloc);

669
	return 0;
670 671 672 673

search_again:
	if (start > end)
		goto out;
674
	spin_unlock(&tree->lock);
675 676 677 678 679
	if (mask & __GFP_WAIT)
		cond_resched();
	goto again;
}

680 681
static void wait_on_state(struct extent_io_tree *tree,
			  struct extent_state *state)
682 683
		__releases(tree->lock)
		__acquires(tree->lock)
684 685 686
{
	DEFINE_WAIT(wait);
	prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE);
687
	spin_unlock(&tree->lock);
688
	schedule();
689
	spin_lock(&tree->lock);
690 691 692 693 694 695 696 697
	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
 */
698 699
static void wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
			    unsigned long bits)
700 701 702 703
{
	struct extent_state *state;
	struct rb_node *node;

704 705
	btrfs_debug_check_extent_io_range(tree->mapping->host, start, end);

706
	spin_lock(&tree->lock);
707 708 709 710 711 712
again:
	while (1) {
		/*
		 * this search will find all the extents that end after
		 * our range starts
		 */
713
		node = tree_search(tree, start);
714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733
		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;

734
		cond_resched_lock(&tree->lock);
735 736
	}
out:
737
	spin_unlock(&tree->lock);
738 739
}

740
static void set_state_bits(struct extent_io_tree *tree,
741
			   struct extent_state *state,
742
			   unsigned long *bits)
743
{
744
	unsigned long bits_to_set = *bits & ~EXTENT_CTLBITS;
J
Josef Bacik 已提交
745

746
	set_state_cb(tree, state, bits);
747
	if ((bits_to_set & EXTENT_DIRTY) && !(state->state & EXTENT_DIRTY)) {
748 749 750
		u64 range = state->end - state->start + 1;
		tree->dirty_bytes += range;
	}
751
	state->state |= bits_to_set;
752 753
}

754 755 756 757 758 759 760 761 762 763 764
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);
		}
	}
}

765
/*
766 767
 * 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.
768
 *
769 770 771
 * 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.
772
 *
773
 * [start, end] is inclusive This takes the tree lock.
774
 */
775

J
Jeff Mahoney 已提交
776 777
static int __must_check
__set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
778 779 780
		 unsigned long bits, unsigned long exclusive_bits,
		 u64 *failed_start, struct extent_state **cached_state,
		 gfp_t mask)
781 782 783 784 785 786 787
{
	struct extent_state *state;
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
	int err = 0;
	u64 last_start;
	u64 last_end;
788

789 790
	btrfs_debug_check_extent_io_range(tree->mapping->host, start, end);

791
	bits |= EXTENT_FIRST_DELALLOC;
792 793 794
again:
	if (!prealloc && (mask & __GFP_WAIT)) {
		prealloc = alloc_extent_state(mask);
795
		BUG_ON(!prealloc);
796 797
	}

798
	spin_lock(&tree->lock);
799 800
	if (cached_state && *cached_state) {
		state = *cached_state;
801 802
		if (state->start <= start && state->end > start &&
		    state->tree) {
803 804 805 806
			node = &state->rb_node;
			goto hit_next;
		}
	}
807 808 809 810
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
811
	node = tree_search(tree, start);
812
	if (!node) {
813 814
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
815
		err = insert_state(tree, prealloc, start, end, &bits);
816 817 818
		if (err)
			extent_io_tree_panic(tree, err);

819 820 821 822
		prealloc = NULL;
		goto out;
	}
	state = rb_entry(node, struct extent_state, rb_node);
C
Chris Mason 已提交
823
hit_next:
824 825 826 827 828 829 830 831 832 833
	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) {
834
		if (state->state & exclusive_bits) {
835 836 837 838
			*failed_start = state->start;
			err = -EEXIST;
			goto out;
		}
839

840
		set_state_bits(tree, state, &bits);
841
		cache_state(state, cached_state);
842
		merge_state(tree, state);
843 844 845
		if (last_end == (u64)-1)
			goto out;
		start = last_end + 1;
846 847 848 849
		state = next_state(state);
		if (start < end && state && state->start == start &&
		    !need_resched())
			goto hit_next;
850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869
		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) {
870
		if (state->state & exclusive_bits) {
871 872 873 874
			*failed_start = start;
			err = -EEXIST;
			goto out;
		}
875 876 877

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
878
		err = split_state(tree, state, prealloc, start);
879 880 881
		if (err)
			extent_io_tree_panic(tree, err);

882 883 884 885
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
886
			set_state_bits(tree, state, &bits);
887
			cache_state(state, cached_state);
888
			merge_state(tree, state);
889 890 891
			if (last_end == (u64)-1)
				goto out;
			start = last_end + 1;
892 893 894 895
			state = next_state(state);
			if (start < end && state && state->start == start &&
			    !need_resched())
				goto hit_next;
896 897 898 899 900 901 902 903 904 905 906 907 908 909 910
		}
		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
911
			this_end = last_start - 1;
912 913 914

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
915 916 917 918 919

		/*
		 * Avoid to free 'prealloc' if it can be merged with
		 * the later extent.
		 */
920
		err = insert_state(tree, prealloc, start, this_end,
921
				   &bits);
922 923 924
		if (err)
			extent_io_tree_panic(tree, err);

J
Josef Bacik 已提交
925 926
		cache_state(prealloc, cached_state);
		prealloc = NULL;
927 928 929 930 931 932 933 934 935 936
		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) {
937
		if (state->state & exclusive_bits) {
938 939 940 941
			*failed_start = start;
			err = -EEXIST;
			goto out;
		}
942 943 944

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
945
		err = split_state(tree, state, prealloc, end + 1);
946 947
		if (err)
			extent_io_tree_panic(tree, err);
948

949
		set_state_bits(tree, prealloc, &bits);
950
		cache_state(prealloc, cached_state);
951 952 953 954 955 956 957 958
		merge_state(tree, prealloc);
		prealloc = NULL;
		goto out;
	}

	goto search_again;

out:
959
	spin_unlock(&tree->lock);
960 961 962 963 964 965 966 967
	if (prealloc)
		free_extent_state(prealloc);

	return err;

search_again:
	if (start > end)
		goto out;
968
	spin_unlock(&tree->lock);
969 970 971 972 973
	if (mask & __GFP_WAIT)
		cond_resched();
	goto again;
}

974 975 976
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
977 978 979 980 981 982
{
	return __set_extent_bit(tree, start, end, bits, 0, failed_start,
				cached_state, mask);
}


J
Josef Bacik 已提交
983
/**
L
Liu Bo 已提交
984 985
 * convert_extent_bit - convert all bits in a given range from one bit to
 * 			another
J
Josef Bacik 已提交
986 987 988 989 990
 * @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
991
 * @cached_state:	state that we're going to cache
J
Josef Bacik 已提交
992 993 994 995 996 997 998 999 1000
 * @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,
1001
		       unsigned long bits, unsigned long clear_bits,
1002
		       struct extent_state **cached_state, gfp_t mask)
J
Josef Bacik 已提交
1003 1004 1005 1006 1007 1008 1009 1010
{
	struct extent_state *state;
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
	int err = 0;
	u64 last_start;
	u64 last_end;

1011 1012
	btrfs_debug_check_extent_io_range(tree->mapping->host, start, end);

J
Josef Bacik 已提交
1013 1014 1015 1016 1017 1018 1019 1020
again:
	if (!prealloc && (mask & __GFP_WAIT)) {
		prealloc = alloc_extent_state(mask);
		if (!prealloc)
			return -ENOMEM;
	}

	spin_lock(&tree->lock);
1021 1022 1023 1024 1025 1026 1027 1028 1029
	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
1030 1031 1032 1033 1034 1035 1036
	/*
	 * 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);
1037 1038 1039 1040
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1041 1042
		err = insert_state(tree, prealloc, start, end, &bits);
		prealloc = NULL;
1043 1044
		if (err)
			extent_io_tree_panic(tree, err);
J
Josef Bacik 已提交
1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059
		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);
1060
		cache_state(state, cached_state);
1061
		state = clear_state_bit(tree, state, &clear_bits, 0);
J
Josef Bacik 已提交
1062 1063 1064
		if (last_end == (u64)-1)
			goto out;
		start = last_end + 1;
1065 1066 1067
		if (start < end && state && state->start == start &&
		    !need_resched())
			goto hit_next;
J
Josef Bacik 已提交
1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088
		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);
1089 1090 1091 1092
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1093
		err = split_state(tree, state, prealloc, start);
1094 1095
		if (err)
			extent_io_tree_panic(tree, err);
J
Josef Bacik 已提交
1096 1097 1098 1099 1100
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
			set_state_bits(tree, state, &bits);
1101
			cache_state(state, cached_state);
1102
			state = clear_state_bit(tree, state, &clear_bits, 0);
J
Josef Bacik 已提交
1103 1104 1105
			if (last_end == (u64)-1)
				goto out;
			start = last_end + 1;
1106 1107 1108
			if (start < end && state && state->start == start &&
			    !need_resched())
				goto hit_next;
J
Josef Bacik 已提交
1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126
		}
		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);
1127 1128 1129 1130
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1131 1132 1133 1134 1135 1136 1137

		/*
		 * Avoid to free 'prealloc' if it can be merged with
		 * the later extent.
		 */
		err = insert_state(tree, prealloc, start, this_end,
				   &bits);
1138 1139
		if (err)
			extent_io_tree_panic(tree, err);
1140
		cache_state(prealloc, cached_state);
J
Josef Bacik 已提交
1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152
		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);
1153 1154 1155 1156
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1157 1158

		err = split_state(tree, state, prealloc, end + 1);
1159 1160
		if (err)
			extent_io_tree_panic(tree, err);
J
Josef Bacik 已提交
1161 1162

		set_state_bits(tree, prealloc, &bits);
1163
		cache_state(prealloc, cached_state);
J
Josef Bacik 已提交
1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186
		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;
}

1187 1188 1189 1190
/* 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
1191
	return set_extent_bit(tree, start, end, EXTENT_DIRTY, NULL,
1192
			      NULL, mask);
1193 1194 1195
}

int set_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1196
		    unsigned long bits, gfp_t mask)
1197
{
J
Jeff Mahoney 已提交
1198
	return set_extent_bit(tree, start, end, bits, NULL,
1199
			      NULL, mask);
1200 1201 1202
}

int clear_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1203
		      unsigned long bits, gfp_t mask)
1204
{
1205
	return clear_extent_bit(tree, start, end, bits, 0, 0, NULL, mask);
1206 1207 1208
}

int set_extent_delalloc(struct extent_io_tree *tree, u64 start, u64 end,
1209
			struct extent_state **cached_state, gfp_t mask)
1210 1211
{
	return set_extent_bit(tree, start, end,
1212
			      EXTENT_DELALLOC | EXTENT_UPTODATE,
J
Jeff Mahoney 已提交
1213
			      NULL, cached_state, mask);
1214 1215
}

1216 1217 1218 1219 1220 1221 1222 1223
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);
}

1224 1225 1226 1227
int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
		       gfp_t mask)
{
	return clear_extent_bit(tree, start, end,
1228
				EXTENT_DIRTY | EXTENT_DELALLOC |
1229
				EXTENT_DO_ACCOUNTING, 0, 0, NULL, mask);
1230 1231 1232 1233 1234
}

int set_extent_new(struct extent_io_tree *tree, u64 start, u64 end,
		     gfp_t mask)
{
J
Jeff Mahoney 已提交
1235
	return set_extent_bit(tree, start, end, EXTENT_NEW, NULL,
1236
			      NULL, mask);
1237 1238 1239
}

int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end,
1240
			struct extent_state **cached_state, gfp_t mask)
1241
{
L
Liu Bo 已提交
1242
	return set_extent_bit(tree, start, end, EXTENT_UPTODATE, NULL,
J
Jeff Mahoney 已提交
1243
			      cached_state, mask);
1244 1245
}

1246 1247
int clear_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end,
			  struct extent_state **cached_state, gfp_t mask)
1248
{
1249
	return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0,
1250
				cached_state, mask);
1251 1252
}

C
Chris Mason 已提交
1253 1254 1255 1256
/*
 * either insert or lock state struct between start and end use mask to tell
 * us if waiting is desired.
 */
1257
int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1258
		     unsigned long bits, struct extent_state **cached_state)
1259 1260 1261 1262
{
	int err;
	u64 failed_start;
	while (1) {
J
Jeff Mahoney 已提交
1263 1264 1265
		err = __set_extent_bit(tree, start, end, EXTENT_LOCKED | bits,
				       EXTENT_LOCKED, &failed_start,
				       cached_state, GFP_NOFS);
1266
		if (err == -EEXIST) {
1267 1268
			wait_extent_bit(tree, failed_start, end, EXTENT_LOCKED);
			start = failed_start;
1269
		} else
1270 1271 1272 1273 1274 1275
			break;
		WARN_ON(start > end);
	}
	return err;
}

1276
int lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
1277
{
1278
	return lock_extent_bits(tree, start, end, 0, NULL);
1279 1280
}

1281
int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
1282 1283 1284 1285
{
	int err;
	u64 failed_start;

J
Jeff Mahoney 已提交
1286 1287
	err = __set_extent_bit(tree, start, end, EXTENT_LOCKED, EXTENT_LOCKED,
			       &failed_start, NULL, GFP_NOFS);
Y
Yan Zheng 已提交
1288 1289 1290
	if (err == -EEXIST) {
		if (failed_start > start)
			clear_extent_bit(tree, start, failed_start - 1,
1291
					 EXTENT_LOCKED, 1, 0, NULL, GFP_NOFS);
1292
		return 0;
Y
Yan Zheng 已提交
1293
	}
1294 1295 1296
	return 1;
}

1297 1298 1299 1300 1301 1302 1303
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);
}

1304
int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end)
1305
{
1306
	return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL,
1307
				GFP_NOFS);
1308 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
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;
}

1343 1344 1345
/*
 * helper function to set both pages and extents in the tree writeback
 */
1346
static int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end)
1347 1348 1349 1350 1351 1352 1353
{
	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);
1354
		BUG_ON(!page); /* Pages should be in the extent_io_tree */
1355 1356 1357 1358 1359 1360 1361
		set_page_writeback(page);
		page_cache_release(page);
		index++;
	}
	return 0;
}

C
Chris Mason 已提交
1362 1363 1364 1365
/* 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'
 */
1366 1367
static struct extent_state *
find_first_extent_bit_state(struct extent_io_tree *tree,
1368
			    u64 start, unsigned long bits)
C
Chris Mason 已提交
1369 1370 1371 1372 1373 1374 1375 1376 1377
{
	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
1378
	if (!node)
C
Chris Mason 已提交
1379 1380
		goto out;

C
Chris Mason 已提交
1381
	while (1) {
C
Chris Mason 已提交
1382
		state = rb_entry(node, struct extent_state, rb_node);
C
Chris Mason 已提交
1383
		if (state->end >= start && (state->state & bits))
C
Chris Mason 已提交
1384
			return state;
C
Chris Mason 已提交
1385

C
Chris Mason 已提交
1386 1387 1388 1389 1390 1391 1392 1393
		node = rb_next(node);
		if (!node)
			break;
	}
out:
	return NULL;
}

1394 1395 1396 1397 1398
/*
 * 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.
 *
1399
 * If nothing was found, 1 is returned. If found something, return 0.
1400 1401
 */
int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
1402
			  u64 *start_ret, u64 *end_ret, unsigned long bits,
1403
			  struct extent_state **cached_state)
1404 1405
{
	struct extent_state *state;
1406
	struct rb_node *n;
1407 1408 1409
	int ret = 1;

	spin_lock(&tree->lock);
1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428
	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;
	}

1429
	state = find_first_extent_bit_state(tree, start, bits);
1430
got_it:
1431
	if (state) {
1432
		cache_state(state, cached_state);
1433 1434 1435 1436
		*start_ret = state->start;
		*end_ret = state->end;
		ret = 0;
	}
1437
out:
1438 1439 1440 1441
	spin_unlock(&tree->lock);
	return ret;
}

C
Chris Mason 已提交
1442 1443 1444 1445 1446 1447
/*
 * 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
1448
static noinline u64 find_delalloc_range(struct extent_io_tree *tree,
1449 1450
					u64 *start, u64 *end, u64 max_bytes,
					struct extent_state **cached_state)
1451 1452 1453 1454 1455 1456 1457
{
	struct rb_node *node;
	struct extent_state *state;
	u64 cur_start = *start;
	u64 found = 0;
	u64 total_bytes = 0;

1458
	spin_lock(&tree->lock);
C
Chris Mason 已提交
1459

1460 1461 1462 1463
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1464
	node = tree_search(tree, cur_start);
1465
	if (!node) {
1466 1467
		if (!found)
			*end = (u64)-1;
1468 1469 1470
		goto out;
	}

C
Chris Mason 已提交
1471
	while (1) {
1472
		state = rb_entry(node, struct extent_state, rb_node);
1473 1474
		if (found && (state->start != cur_start ||
			      (state->state & EXTENT_BOUNDARY))) {
1475 1476 1477 1478 1479 1480 1481
			goto out;
		}
		if (!(state->state & EXTENT_DELALLOC)) {
			if (!found)
				*end = state->end;
			goto out;
		}
1482
		if (!found) {
1483
			*start = state->start;
1484 1485 1486
			*cached_state = state;
			atomic_inc(&state->refs);
		}
1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497
		found++;
		*end = state->end;
		cur_start = state->end + 1;
		node = rb_next(node);
		if (!node)
			break;
		total_bytes += state->end - state->start + 1;
		if (total_bytes >= max_bytes)
			break;
	}
out:
1498
	spin_unlock(&tree->lock);
1499 1500 1501
	return found;
}

1502 1503 1504
static noinline void __unlock_for_delalloc(struct inode *inode,
					   struct page *locked_page,
					   u64 start, u64 end)
C
Chris Mason 已提交
1505 1506 1507 1508 1509 1510 1511 1512 1513
{
	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)
1514
		return;
C
Chris Mason 已提交
1515

C
Chris Mason 已提交
1516
	while (nr_pages > 0) {
C
Chris Mason 已提交
1517
		ret = find_get_pages_contig(inode->i_mapping, index,
1518 1519
				     min_t(unsigned long, nr_pages,
				     ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1520 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
		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
1551
	while (nrpages > 0) {
C
Chris Mason 已提交
1552
		ret = find_get_pages_contig(inode->i_mapping, index,
1553 1554
				     min_t(unsigned long,
				     nrpages, ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1555 1556 1557 1558 1559 1560 1561 1562 1563 1564
		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
			 */
1565
			if (pages[i] != locked_page) {
C
Chris Mason 已提交
1566
				lock_page(pages[i]);
1567 1568
				if (!PageDirty(pages[i]) ||
				    pages[i]->mapping != inode->i_mapping) {
1569 1570 1571 1572 1573 1574
					ret = -EAGAIN;
					unlock_page(pages[i]);
					page_cache_release(pages[i]);
					goto done;
				}
			}
C
Chris Mason 已提交
1575
			page_cache_release(pages[i]);
1576
			pages_locked++;
C
Chris Mason 已提交
1577 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
		}
		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;
1608
	struct extent_state *cached_state = NULL;
C
Chris Mason 已提交
1609 1610 1611 1612 1613 1614 1615 1616
	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,
1617
				    max_bytes, &cached_state);
C
Chris Mason 已提交
1618
	if (!found || delalloc_end <= *start) {
C
Chris Mason 已提交
1619 1620
		*start = delalloc_start;
		*end = delalloc_end;
1621
		free_extent_state(cached_state);
C
Chris Mason 已提交
1622 1623 1624
		return found;
	}

C
Chris Mason 已提交
1625 1626 1627 1628 1629
	/*
	 * 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
1630
	if (delalloc_start < *start)
C
Chris Mason 已提交
1631 1632
		delalloc_start = *start;

C
Chris Mason 已提交
1633 1634 1635 1636
	/*
	 * make sure to limit the number of pages we try to lock down
	 * if we're looping.
	 */
C
Chris Mason 已提交
1637
	if (delalloc_end + 1 - delalloc_start > max_bytes && loops)
1638
		delalloc_end = delalloc_start + PAGE_CACHE_SIZE - 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 1649 1650 1651 1652 1653 1654 1655 1656 1657
		if (!loops) {
			unsigned long offset = (*start) & (PAGE_CACHE_SIZE - 1);
			max_bytes = PAGE_CACHE_SIZE - offset;
			loops = 1;
			goto again;
		} else {
			found = 0;
			goto out_failed;
		}
	}
1658
	BUG_ON(ret); /* Only valid values are 0 and -EAGAIN */
C
Chris Mason 已提交
1659 1660

	/* step three, lock the state bits for the whole range */
1661
	lock_extent_bits(tree, delalloc_start, delalloc_end, 0, &cached_state);
C
Chris Mason 已提交
1662 1663 1664

	/* then test to make sure it is all still delalloc */
	ret = test_range_bit(tree, delalloc_start, delalloc_end,
1665
			     EXTENT_DELALLOC, 1, cached_state);
C
Chris Mason 已提交
1666
	if (!ret) {
1667 1668
		unlock_extent_cached(tree, delalloc_start, delalloc_end,
				     &cached_state, GFP_NOFS);
C
Chris Mason 已提交
1669 1670 1671 1672 1673
		__unlock_for_delalloc(inode, locked_page,
			      delalloc_start, delalloc_end);
		cond_resched();
		goto again;
	}
1674
	free_extent_state(cached_state);
C
Chris Mason 已提交
1675 1676 1677 1678 1679 1680
	*start = delalloc_start;
	*end = delalloc_end;
out_failed:
	return found;
}

1681 1682 1683 1684
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
1685
{
1686
	struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
C
Chris Mason 已提交
1687 1688 1689 1690 1691 1692
	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;
1693

1694
	clear_extent_bit(tree, start, end, clear_bits, 1, 0, NULL, GFP_NOFS);
1695
	if (page_ops == 0)
1696
		return 0;
C
Chris Mason 已提交
1697

C
Chris Mason 已提交
1698
	while (nr_pages > 0) {
C
Chris Mason 已提交
1699
		ret = find_get_pages_contig(inode->i_mapping, index,
1700 1701
				     min_t(unsigned long,
				     nr_pages, ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1702
		for (i = 0; i < ret; i++) {
1703

1704
			if (page_ops & PAGE_SET_PRIVATE2)
1705 1706
				SetPagePrivate2(pages[i]);

C
Chris Mason 已提交
1707 1708 1709 1710
			if (pages[i] == locked_page) {
				page_cache_release(pages[i]);
				continue;
			}
1711
			if (page_ops & PAGE_CLEAR_DIRTY)
C
Chris Mason 已提交
1712
				clear_page_dirty_for_io(pages[i]);
1713
			if (page_ops & PAGE_SET_WRITEBACK)
C
Chris Mason 已提交
1714
				set_page_writeback(pages[i]);
1715
			if (page_ops & PAGE_END_WRITEBACK)
C
Chris Mason 已提交
1716
				end_page_writeback(pages[i]);
1717
			if (page_ops & PAGE_UNLOCK)
1718
				unlock_page(pages[i]);
C
Chris Mason 已提交
1719 1720 1721 1722 1723 1724 1725 1726 1727
			page_cache_release(pages[i]);
		}
		nr_pages -= ret;
		index += ret;
		cond_resched();
	}
	return 0;
}

C
Chris Mason 已提交
1728 1729 1730 1731 1732
/*
 * 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.
 */
1733 1734
u64 count_range_bits(struct extent_io_tree *tree,
		     u64 *start, u64 search_end, u64 max_bytes,
1735
		     unsigned long bits, int contig)
1736 1737 1738 1739 1740
{
	struct rb_node *node;
	struct extent_state *state;
	u64 cur_start = *start;
	u64 total_bytes = 0;
1741
	u64 last = 0;
1742 1743 1744 1745 1746 1747 1748
	int found = 0;

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

1749
	spin_lock(&tree->lock);
1750 1751 1752 1753 1754 1755 1756 1757
	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.
	 */
1758
	node = tree_search(tree, cur_start);
C
Chris Mason 已提交
1759
	if (!node)
1760 1761
		goto out;

C
Chris Mason 已提交
1762
	while (1) {
1763 1764 1765
		state = rb_entry(node, struct extent_state, rb_node);
		if (state->start > search_end)
			break;
1766 1767 1768
		if (contig && found && state->start > last + 1)
			break;
		if (state->end >= cur_start && (state->state & bits) == bits) {
1769 1770 1771 1772 1773
			total_bytes += min(search_end, state->end) + 1 -
				       max(cur_start, state->start);
			if (total_bytes >= max_bytes)
				break;
			if (!found) {
1774
				*start = max(cur_start, state->start);
1775 1776
				found = 1;
			}
1777 1778 1779
			last = state->end;
		} else if (contig && found) {
			break;
1780 1781 1782 1783 1784 1785
		}
		node = rb_next(node);
		if (!node)
			break;
	}
out:
1786
	spin_unlock(&tree->lock);
1787 1788
	return total_bytes;
}
1789

C
Chris Mason 已提交
1790 1791 1792 1793
/*
 * set the private field for a given byte offset in the tree.  If there isn't
 * an extent_state there already, this does nothing.
 */
1794 1795 1796 1797 1798 1799
int set_state_private(struct extent_io_tree *tree, u64 start, u64 private)
{
	struct rb_node *node;
	struct extent_state *state;
	int ret = 0;

1800
	spin_lock(&tree->lock);
1801 1802 1803 1804
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1805
	node = tree_search(tree, start);
1806
	if (!node) {
1807 1808 1809 1810 1811 1812 1813 1814 1815 1816
		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:
1817
	spin_unlock(&tree->lock);
1818 1819 1820 1821 1822 1823 1824 1825 1826
	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;

1827
	spin_lock(&tree->lock);
1828 1829 1830 1831
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1832
	node = tree_search(tree, start);
1833
	if (!node) {
1834 1835 1836 1837 1838 1839 1840 1841 1842 1843
		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:
1844
	spin_unlock(&tree->lock);
1845 1846 1847 1848 1849
	return ret;
}

/*
 * searches a range in the state tree for a given mask.
1850
 * If 'filled' == 1, this returns 1 only if every extent in the tree
1851 1852 1853 1854
 * 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,
1855
		   unsigned long bits, int filled, struct extent_state *cached)
1856 1857 1858 1859 1860
{
	struct extent_state *state = NULL;
	struct rb_node *node;
	int bitset = 0;

1861
	spin_lock(&tree->lock);
1862 1863
	if (cached && cached->tree && cached->start <= start &&
	    cached->end > start)
1864 1865 1866
		node = &cached->rb_node;
	else
		node = tree_search(tree, start);
1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885
	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;
		}
1886 1887 1888 1889

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

1890 1891 1892 1893 1894 1895 1896 1897 1898 1899
		start = state->end + 1;
		if (start > end)
			break;
		node = rb_next(node);
		if (!node) {
			if (filled)
				bitset = 0;
			break;
		}
	}
1900
	spin_unlock(&tree->lock);
1901 1902 1903 1904 1905 1906 1907
	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
 */
1908
static void check_page_uptodate(struct extent_io_tree *tree, struct page *page)
1909
{
M
Miao Xie 已提交
1910
	u64 start = page_offset(page);
1911
	u64 end = start + PAGE_CACHE_SIZE - 1;
1912
	if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1, NULL))
1913 1914 1915
		SetPageUptodate(page);
}

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 1948
/*
 * 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
1949 1950 1951 1952 1953
	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;
1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968

	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.
1969
 * to avoid any synchronization issues, wait for the data after writing, which
1970 1971 1972 1973
 * 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.
 */
1974
int repair_io_failure(struct btrfs_fs_info *fs_info, u64 start,
1975 1976 1977 1978 1979 1980 1981 1982 1983
			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
1984
	struct btrfs_mapping_tree *map_tree = &fs_info->mapping_tree;
1985 1986 1987 1988
	int ret;

	BUG_ON(!mirror_num);

D
David Woodhouse 已提交
1989 1990 1991 1992
	/* we can't repair anything in raid56 yet */
	if (btrfs_is_parity_mirror(map_tree, logical, length, mirror_num))
		return 0;

1993
	bio = btrfs_io_bio_alloc(GFP_NOFS, 1);
1994 1995 1996 1997 1998 1999 2000
	if (!bio)
		return -EIO;
	bio->bi_private = &compl;
	bio->bi_end_io = repair_io_failure_callback;
	bio->bi_size = 0;
	map_length = length;

2001
	ret = btrfs_map_block(fs_info, WRITE, logical,
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
			      &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
2017
	bio_add_page(bio, page, length, start - page_offset(page));
2018
	btrfsic_submit_bio(WRITE_SYNC, bio);
2019 2020 2021 2022 2023
	wait_for_completion(&compl);

	if (!test_bit(BIO_UPTODATE, &bio->bi_flags)) {
		/* try to remap that extent elsewhere? */
		bio_put(bio);
2024
		btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_WRITE_ERRS);
2025 2026 2027
		return -EIO;
	}

2028
	printk_ratelimited_in_rcu(KERN_INFO "btrfs read error corrected: ino %lu off %llu "
2029 2030
		      "(dev %s sector %llu)\n", page->mapping->host->i_ino,
		      start, rcu_str_deref(dev->name), sector);
2031 2032 2033 2034 2035

	bio_put(bio);
	return 0;
}

2036 2037 2038 2039 2040
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);
2041
	int ret = 0;
2042 2043 2044

	for (i = 0; i < num_pages; i++) {
		struct page *p = extent_buffer_page(eb, i);
2045
		ret = repair_io_failure(root->fs_info, start, PAGE_CACHE_SIZE,
2046 2047 2048 2049 2050 2051 2052 2053 2054
					start, p, mirror_num);
		if (ret)
			break;
		start += PAGE_CACHE_SIZE;
	}

	return ret;
}

2055 2056 2057 2058 2059 2060 2061 2062 2063
/*
 * 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;
2064
	struct btrfs_fs_info *fs_info;
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 2098
	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);

2099 2100
	if (state && state->start <= failrec->start &&
	    state->end >= failrec->start + failrec->len - 1) {
2101 2102 2103
		fs_info = BTRFS_I(inode)->root->fs_info;
		num_copies = btrfs_num_copies(fs_info, failrec->logical,
					      failrec->len);
2104
		if (num_copies > 1)  {
2105
			ret = repair_io_failure(fs_info, start, failrec->len,
2106 2107 2108 2109
						failrec->logical, page,
						failrec->failed_mirror);
			did_repair = !ret;
		}
D
David Woodhouse 已提交
2110
		ret = 0;
2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127
	}

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
 */

2128 2129 2130
static int bio_readpage_error(struct bio *failed_bio, u64 phy_offset,
			      struct page *page, u64 start, u64 end,
			      int failed_mirror)
2131 2132 2133 2134 2135 2136 2137 2138 2139
{
	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;
2140 2141
	struct btrfs_io_bio *btrfs_failed_bio;
	struct btrfs_io_bio *btrfs_bio;
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 2173
	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);

2174
		if (!em) {
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 2216
			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.
		 */
	}
2217 2218
	num_copies = btrfs_num_copies(BTRFS_I(inode)->root->fs_info,
				      failrec->logical, failrec->len);
2219 2220 2221 2222 2223 2224
	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.
		 */
2225 2226
		pr_debug("bio_readpage_error: cannot repair, num_copies=%d, next_mirror %d, failed_mirror %d\n",
			 num_copies, failrec->this_mirror, failed_mirror);
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 2266
		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;
	}

2267 2268
	if (failrec->this_mirror > num_copies) {
		pr_debug("bio_readpage_error: (fail) num_copies=%d, next_mirror %d, failed_mirror %d\n",
2269 2270 2271 2272 2273
			 num_copies, failrec->this_mirror, failed_mirror);
		free_io_failure(inode, failrec, 0);
		return -EIO;
	}

2274
	bio = btrfs_io_bio_alloc(GFP_NOFS, 1);
2275 2276 2277 2278
	if (!bio) {
		free_io_failure(inode, failrec, 0);
		return -EIO;
	}
2279 2280 2281 2282 2283
	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;

2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296
	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);
	}

2297 2298 2299 2300 2301 2302
	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);

2303 2304 2305 2306
	ret = tree->ops->submit_bio_hook(inode, read_mode, bio,
					 failrec->this_mirror,
					 failrec->bio_flags, 0);
	return ret;
2307 2308
}

2309 2310
/* lots and lots of room for performance fixes in the end_bio funcs */

2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332
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;
}

2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344
/*
 * 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;
2345
	struct extent_io_tree *tree;
2346 2347 2348 2349 2350
	u64 start;
	u64 end;

	do {
		struct page *page = bvec->bv_page;
2351 2352
		tree = &BTRFS_I(page->mapping->host)->io_tree;

2353 2354 2355 2356 2357 2358 2359 2360 2361 2362
		/* 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);
2363

2364 2365
		start = page_offset(page);
		end = start + bvec->bv_offset + bvec->bv_len - 1;
2366 2367 2368

		if (--bvec >= bio->bi_io_vec)
			prefetchw(&bvec->bv_page->flags);
2369

2370 2371
		if (end_extent_writepage(page, err, start, end))
			continue;
2372

2373
		end_page_writeback(page);
2374
	} while (bvec >= bio->bi_io_vec);
2375

2376 2377 2378
	bio_put(bio);
}

2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390
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);
}

2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404
/*
 * 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);
2405 2406
	struct bio_vec *bvec_end = bio->bi_io_vec + bio->bi_vcnt - 1;
	struct bio_vec *bvec = bio->bi_io_vec;
2407
	struct btrfs_io_bio *io_bio = btrfs_io_bio(bio);
2408
	struct extent_io_tree *tree;
2409
	u64 offset = 0;
2410 2411
	u64 start;
	u64 end;
2412
	u64 len;
2413 2414
	u64 extent_start = 0;
	u64 extent_len = 0;
2415
	int mirror;
2416 2417
	int ret;

2418 2419 2420
	if (err)
		uptodate = 0;

2421 2422
	do {
		struct page *page = bvec->bv_page;
2423
		struct inode *inode = page->mapping->host;
2424

2425
		pr_debug("end_bio_extent_readpage: bi_sector=%llu, err=%d, "
2426 2427
			 "mirror=%lu\n", (u64)bio->bi_sector, err,
			 io_bio->mirror_num);
2428
		tree = &BTRFS_I(inode)->io_tree;
2429

2430 2431 2432 2433 2434 2435 2436 2437 2438 2439
		/* 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);
2440

2441 2442
		start = page_offset(page);
		end = start + bvec->bv_offset + bvec->bv_len - 1;
2443
		len = bvec->bv_len;
2444

2445
		if (++bvec <= bvec_end)
2446 2447
			prefetchw(&bvec->bv_page->flags);

2448
		mirror = io_bio->mirror_num;
2449 2450
		if (likely(uptodate && tree->ops &&
			   tree->ops->readpage_end_io_hook)) {
2451 2452 2453
			ret = tree->ops->readpage_end_io_hook(io_bio, offset,
							      page, start, end,
							      mirror);
2454
			if (ret)
2455
				uptodate = 0;
2456
			else
2457
				clean_io_failure(start, page);
2458
		}
2459

2460 2461 2462 2463
		if (likely(uptodate))
			goto readpage_ok;

		if (tree->ops && tree->ops->readpage_io_failed_hook) {
2464
			ret = tree->ops->readpage_io_failed_hook(page, mirror);
2465 2466 2467
			if (!ret && !err &&
			    test_bit(BIO_UPTODATE, &bio->bi_flags))
				uptodate = 1;
2468
		} else {
2469 2470 2471 2472 2473 2474 2475 2476 2477 2478
			/*
			 * 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.
			 */
2479 2480
			ret = bio_readpage_error(bio, offset, page, start, end,
						 mirror);
2481
			if (ret == 0) {
2482 2483
				uptodate =
					test_bit(BIO_UPTODATE, &bio->bi_flags);
2484 2485
				if (err)
					uptodate = 0;
2486 2487 2488
				continue;
			}
		}
2489
readpage_ok:
2490
		if (likely(uptodate)) {
2491 2492 2493 2494 2495 2496 2497 2498
			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);
2499
			SetPageUptodate(page);
2500
		} else {
2501 2502
			ClearPageUptodate(page);
			SetPageError(page);
2503
		}
2504
		unlock_page(page);
2505
		offset += len;
2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527

		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;
		}
2528
	} while (bvec <= bvec_end);
2529

2530 2531 2532
	if (extent_len)
		endio_readpage_release_extent(tree, extent_start, extent_len,
					      uptodate);
2533 2534
	if (io_bio->end_io)
		io_bio->end_io(io_bio, err);
2535 2536 2537
	bio_put(bio);
}

2538 2539 2540 2541
/*
 * 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
 */
2542 2543 2544
struct bio *
btrfs_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
		gfp_t gfp_flags)
2545
{
2546
	struct btrfs_io_bio *btrfs_bio;
2547 2548
	struct bio *bio;

2549
	bio = bio_alloc_bioset(gfp_flags, nr_vecs, btrfs_bioset);
2550 2551

	if (bio == NULL && (current->flags & PF_MEMALLOC)) {
2552 2553 2554 2555
		while (!bio && (nr_vecs /= 2)) {
			bio = bio_alloc_bioset(gfp_flags,
					       nr_vecs, btrfs_bioset);
		}
2556 2557 2558
	}

	if (bio) {
2559
		bio->bi_size = 0;
2560 2561
		bio->bi_bdev = bdev;
		bio->bi_sector = first_sector;
2562 2563 2564 2565
		btrfs_bio = btrfs_io_bio(bio);
		btrfs_bio->csum = NULL;
		btrfs_bio->csum_allocated = NULL;
		btrfs_bio->end_io = NULL;
2566 2567 2568 2569
	}
	return bio;
}

2570 2571 2572 2573 2574 2575 2576 2577 2578
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)
{
2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589
	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;
2590 2591 2592
}


2593 2594
static int __must_check submit_one_bio(int rw, struct bio *bio,
				       int mirror_num, unsigned long bio_flags)
2595 2596
{
	int ret = 0;
2597 2598 2599 2600 2601
	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
2602
	start = page_offset(page) + bvec->bv_offset;
2603

2604
	bio->bi_private = NULL;
2605 2606 2607

	bio_get(bio);

2608
	if (tree->ops && tree->ops->submit_bio_hook)
2609
		ret = tree->ops->submit_bio_hook(page->mapping->host, rw, bio,
2610
					   mirror_num, bio_flags, start);
2611
	else
2612
		btrfsic_submit_bio(rw, bio);
2613

2614 2615 2616 2617 2618 2619
	if (bio_flagged(bio, BIO_EOPNOTSUPP))
		ret = -EOPNOTSUPP;
	bio_put(bio);
	return ret;
}

2620
static int merge_bio(int rw, struct extent_io_tree *tree, struct page *page,
2621 2622 2623 2624 2625
		     unsigned long offset, size_t size, struct bio *bio,
		     unsigned long bio_flags)
{
	int ret = 0;
	if (tree->ops && tree->ops->merge_bio_hook)
2626
		ret = tree->ops->merge_bio_hook(rw, page, offset, size, bio,
2627 2628 2629 2630 2631 2632
						bio_flags);
	BUG_ON(ret < 0);
	return ret;

}

2633 2634 2635 2636 2637 2638
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,
2639
			      bio_end_io_t end_io_func,
C
Chris Mason 已提交
2640 2641 2642
			      int mirror_num,
			      unsigned long prev_bio_flags,
			      unsigned long bio_flags)
2643 2644 2645 2646
{
	int ret = 0;
	struct bio *bio;
	int nr;
C
Chris Mason 已提交
2647 2648 2649
	int contig = 0;
	int this_compressed = bio_flags & EXTENT_BIO_COMPRESSED;
	int old_compressed = prev_bio_flags & EXTENT_BIO_COMPRESSED;
2650
	size_t page_size = min_t(size_t, size, PAGE_CACHE_SIZE);
2651 2652 2653

	if (bio_ret && *bio_ret) {
		bio = *bio_ret;
C
Chris Mason 已提交
2654 2655 2656
		if (old_compressed)
			contig = bio->bi_sector == sector;
		else
K
Kent Overstreet 已提交
2657
			contig = bio_end_sector(bio) == sector;
C
Chris Mason 已提交
2658 2659

		if (prev_bio_flags != bio_flags || !contig ||
2660
		    merge_bio(rw, tree, page, offset, page_size, bio, bio_flags) ||
C
Chris Mason 已提交
2661 2662 2663
		    bio_add_page(bio, page, page_size, offset) < page_size) {
			ret = submit_one_bio(rw, bio, mirror_num,
					     prev_bio_flags);
2664 2665
			if (ret < 0)
				return ret;
2666 2667 2668 2669 2670
			bio = NULL;
		} else {
			return 0;
		}
	}
C
Chris Mason 已提交
2671 2672 2673 2674 2675
	if (this_compressed)
		nr = BIO_MAX_PAGES;
	else
		nr = bio_get_nr_vecs(bdev);

2676
	bio = btrfs_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH);
2677 2678
	if (!bio)
		return -ENOMEM;
2679

C
Chris Mason 已提交
2680
	bio_add_page(bio, page, page_size, offset);
2681 2682
	bio->bi_end_io = end_io_func;
	bio->bi_private = tree;
2683

C
Chris Mason 已提交
2684
	if (bio_ret)
2685
		*bio_ret = bio;
C
Chris Mason 已提交
2686
	else
C
Chris Mason 已提交
2687
		ret = submit_one_bio(rw, bio, mirror_num, bio_flags);
2688 2689 2690 2691

	return ret;
}

2692 2693
static void attach_extent_buffer_page(struct extent_buffer *eb,
				      struct page *page)
2694 2695 2696 2697
{
	if (!PagePrivate(page)) {
		SetPagePrivate(page);
		page_cache_get(page);
J
Josef Bacik 已提交
2698 2699 2700
		set_page_private(page, (unsigned long)eb);
	} else {
		WARN_ON(page->private != (unsigned long)eb);
2701 2702 2703
	}
}

J
Josef Bacik 已提交
2704
void set_page_extent_mapped(struct page *page)
2705
{
J
Josef Bacik 已提交
2706 2707 2708 2709 2710
	if (!PagePrivate(page)) {
		SetPagePrivate(page);
		page_cache_get(page);
		set_page_private(page, EXTENT_PAGE_PRIVATE);
	}
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 2739
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;
}
2740 2741 2742 2743
/*
 * 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)
2744
 * XXX JDM: This needs looking at to ensure proper page locking
2745
 */
2746 2747 2748
static int __do_readpage(struct extent_io_tree *tree,
			 struct page *page,
			 get_extent_t *get_extent,
2749
			 struct extent_map **em_cached,
2750 2751
			 struct bio **bio, int mirror_num,
			 unsigned long *bio_flags, int rw)
2752 2753
{
	struct inode *inode = page->mapping->host;
M
Miao Xie 已提交
2754
	u64 start = page_offset(page);
2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766
	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;
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;
C
Chris Mason 已提交
2771
	unsigned long this_bio_flag = 0;
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 2810 2811
					    &cached, GFP_NOFS);
			unlock_extent_cached(tree, cur, cur + iosize - 1,
					     &cached, GFP_NOFS);
2812 2813
			break;
		}
2814 2815
		em = __get_extent_map(inode, page, pg_offset, cur,
				      end - cur + 1, get_extent, em_cached);
2816
		if (IS_ERR_OR_NULL(em)) {
2817
			SetPageError(page);
2818
			unlock_extent(tree, cur, end);
2819 2820 2821 2822 2823 2824
			break;
		}
		extent_offset = cur - em->start;
		BUG_ON(extent_map_end(em) <= cur);
		BUG_ON(end < cur);

2825
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
C
Chris Mason 已提交
2826
			this_bio_flag = EXTENT_BIO_COMPRESSED;
2827 2828 2829
			extent_set_compress_type(&this_bio_flag,
						 em->compress_type);
		}
C
Chris Mason 已提交
2830

2831 2832
		iosize = min(extent_map_end(em) - cur, end - cur + 1);
		cur_end = min(extent_map_end(em) - 1, end);
2833
		iosize = ALIGN(iosize, blocksize);
C
Chris Mason 已提交
2834 2835 2836 2837 2838 2839 2840
		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;
		}
2841 2842
		bdev = em->bdev;
		block_start = em->block_start;
Y
Yan Zheng 已提交
2843 2844
		if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			block_start = EXTENT_MAP_HOLE;
2845 2846 2847 2848 2849 2850
		free_extent_map(em);
		em = NULL;

		/* we've found a hole, just zero and go on */
		if (block_start == EXTENT_MAP_HOLE) {
			char *userpage;
2851 2852
			struct extent_state *cached = NULL;

2853
			userpage = kmap_atomic(page);
2854
			memset(userpage + pg_offset, 0, iosize);
2855
			flush_dcache_page(page);
2856
			kunmap_atomic(userpage);
2857 2858

			set_extent_uptodate(tree, cur, cur + iosize - 1,
2859 2860 2861
					    &cached, GFP_NOFS);
			unlock_extent_cached(tree, cur, cur + iosize - 1,
			                     &cached, GFP_NOFS);
2862
			cur = cur + iosize;
2863
			pg_offset += iosize;
2864 2865 2866
			continue;
		}
		/* the get_extent function already copied into the page */
2867 2868
		if (test_range_bit(tree, cur, cur_end,
				   EXTENT_UPTODATE, 1, NULL)) {
2869
			check_page_uptodate(tree, page);
2870
			unlock_extent(tree, cur, cur + iosize - 1);
2871
			cur = cur + iosize;
2872
			pg_offset += iosize;
2873 2874
			continue;
		}
2875 2876 2877 2878 2879
		/* we have an inline extent but it didn't get marked up
		 * to date.  Error out
		 */
		if (block_start == EXTENT_MAP_INLINE) {
			SetPageError(page);
2880
			unlock_extent(tree, cur, cur + iosize - 1);
2881
			cur = cur + iosize;
2882
			pg_offset += iosize;
2883 2884
			continue;
		}
2885

2886
		pnr -= page->index;
2887
		ret = submit_extent_page(rw, tree, page,
2888
					 sector, disk_io_size, pg_offset,
2889
					 bdev, bio, pnr,
C
Chris Mason 已提交
2890 2891 2892
					 end_bio_extent_readpage, mirror_num,
					 *bio_flags,
					 this_bio_flag);
2893 2894 2895 2896
		if (!ret) {
			nr++;
			*bio_flags = this_bio_flag;
		} else {
2897
			SetPageError(page);
2898 2899
			unlock_extent(tree, cur, cur + iosize - 1);
		}
2900
		cur = cur + iosize;
2901
		pg_offset += iosize;
2902
	}
D
Dan Magenheimer 已提交
2903
out:
2904 2905 2906 2907 2908 2909 2910 2911
	if (!nr) {
		if (!PageError(page))
			SetPageUptodate(page);
		unlock_page(page);
	}
	return 0;
}

2912 2913 2914 2915
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,
2916
					     struct extent_map **em_cached,
2917 2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935 2936
					     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++) {
2937 2938
		__do_readpage(tree, pages[index], get_extent, em_cached, bio,
			      mirror_num, bio_flags, rw);
2939 2940 2941 2942 2943 2944 2945
		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,
2946
			       struct extent_map **em_cached,
2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966
			       struct bio **bio, int mirror_num,
			       unsigned long *bio_flags, int rw)
{
	u64 start;
	u64 end = 0;
	u64 page_start;
	int index;
	int first_index;

	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,
2967 2968 2969
						  end, get_extent, em_cached,
						  bio, mirror_num, bio_flags,
						  rw);
2970 2971 2972 2973 2974 2975 2976 2977 2978
			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,
2979
					  end, get_extent, em_cached, bio,
2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004
					  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);
	}

3005 3006
	ret = __do_readpage(tree, page, get_extent, NULL, bio, mirror_num,
			    bio_flags, rw);
3007 3008 3009
	return ret;
}

3010
int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
3011
			    get_extent_t *get_extent, int mirror_num)
3012 3013
{
	struct bio *bio = NULL;
C
Chris Mason 已提交
3014
	unsigned long bio_flags = 0;
3015 3016
	int ret;

3017
	ret = __extent_read_full_page(tree, page, get_extent, &bio, mirror_num,
3018
				      &bio_flags, READ);
3019
	if (bio)
3020
		ret = submit_one_bio(READ, bio, mirror_num, bio_flags);
3021 3022 3023
	return ret;
}

3024 3025 3026 3027 3028 3029 3030 3031 3032 3033
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;
}

3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045
/*
 * 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
3046
	u64 start = page_offset(page);
3047 3048 3049 3050 3051 3052 3053 3054 3055
	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;
3056
	struct extent_state *cached_state = NULL;
3057 3058 3059 3060
	struct extent_map *em;
	struct block_device *bdev;
	int ret;
	int nr = 0;
3061
	size_t pg_offset = 0;
3062 3063 3064 3065 3066
	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
3067 3068
	int page_started;
	int compressed;
3069
	int write_flags;
3070
	unsigned long nr_written = 0;
3071
	bool fill_delalloc = true;
3072

3073
	if (wbc->sync_mode == WB_SYNC_ALL)
J
Jens Axboe 已提交
3074
		write_flags = WRITE_SYNC;
3075 3076 3077
	else
		write_flags = WRITE;

3078 3079
	trace___extent_writepage(page, inode, wbc);

3080
	WARN_ON(!PageLocked(page));
3081 3082 3083

	ClearPageError(page);

3084
	pg_offset = i_size & (PAGE_CACHE_SIZE - 1);
3085
	if (page->index > end_index ||
3086
	   (page->index == end_index && !pg_offset)) {
3087
		page->mapping->a_ops->invalidatepage(page, 0, PAGE_CACHE_SIZE);
3088 3089 3090 3091 3092 3093 3094
		unlock_page(page);
		return 0;
	}

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

3095
		userpage = kmap_atomic(page);
3096 3097
		memset(userpage + pg_offset, 0,
		       PAGE_CACHE_SIZE - pg_offset);
3098
		kunmap_atomic(userpage);
3099
		flush_dcache_page(page);
3100
	}
3101
	pg_offset = 0;
3102 3103 3104

	set_page_extent_mapped(page);

3105 3106 3107
	if (!tree->ops || !tree->ops->fill_delalloc)
		fill_delalloc = false;

3108 3109
	delalloc_start = start;
	delalloc_end = 0;
C
Chris Mason 已提交
3110
	page_started = 0;
3111
	if (!epd->extent_locked && fill_delalloc) {
3112
		u64 delalloc_to_write = 0;
3113 3114 3115 3116 3117 3118
		/*
		 * 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
3119
		while (delalloc_end < page_end) {
3120
			nr_delalloc = find_lock_delalloc_range(inode, tree,
C
Chris Mason 已提交
3121 3122
						       page,
						       &delalloc_start,
3123 3124
						       &delalloc_end,
						       128 * 1024 * 1024);
3125 3126 3127 3128
			if (nr_delalloc == 0) {
				delalloc_start = delalloc_end + 1;
				continue;
			}
3129 3130 3131 3132 3133
			ret = tree->ops->fill_delalloc(inode, page,
						       delalloc_start,
						       delalloc_end,
						       &page_started,
						       &nr_written);
3134 3135 3136 3137 3138
			/* File system has been set read-only */
			if (ret) {
				SetPageError(page);
				goto done;
			}
3139 3140 3141 3142 3143 3144 3145 3146
			/*
			 * 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;
3147 3148
			delalloc_start = delalloc_end + 1;
		}
3149 3150 3151 3152 3153 3154 3155 3156
		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
3157

3158 3159 3160 3161 3162
		/* did the fill delalloc function already unlock and start
		 * the IO?
		 */
		if (page_started) {
			ret = 0;
3163 3164 3165 3166 3167 3168 3169
			/*
			 * 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;
3170
		}
C
Chris Mason 已提交
3171
	}
3172
	if (tree->ops && tree->ops->writepage_start_hook) {
C
Chris Mason 已提交
3173 3174
		ret = tree->ops->writepage_start_hook(page, start,
						      page_end);
3175 3176 3177 3178 3179 3180
		if (ret) {
			/* Fixup worker will requeue */
			if (ret == -EBUSY)
				wbc->pages_skipped++;
			else
				redirty_page_for_writepage(wbc, page);
3181
			update_nr_written(page, wbc, nr_written);
3182
			unlock_page(page);
3183
			ret = 0;
3184
			goto done_unlocked;
3185 3186 3187
		}
	}

3188 3189 3190 3191 3192
	/*
	 * 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);
3193

3194 3195
	end = page_end;
	if (last_byte <= start) {
3196 3197 3198
		if (tree->ops && tree->ops->writepage_end_io_hook)
			tree->ops->writepage_end_io_hook(page, start,
							 page_end, NULL, 1);
3199 3200 3201 3202 3203 3204 3205
		goto done;
	}

	blocksize = inode->i_sb->s_blocksize;

	while (cur <= end) {
		if (cur >= last_byte) {
3206 3207 3208
			if (tree->ops && tree->ops->writepage_end_io_hook)
				tree->ops->writepage_end_io_hook(page, cur,
							 page_end, NULL, 1);
3209 3210
			break;
		}
3211
		em = epd->get_extent(inode, page, pg_offset, cur,
3212
				     end - cur + 1, 1);
3213
		if (IS_ERR_OR_NULL(em)) {
3214 3215 3216 3217 3218 3219 3220 3221
			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);
3222
		iosize = ALIGN(iosize, blocksize);
3223 3224 3225
		sector = (em->block_start + extent_offset) >> 9;
		bdev = em->bdev;
		block_start = em->block_start;
C
Chris Mason 已提交
3226
		compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
3227 3228 3229
		free_extent_map(em);
		em = NULL;

C
Chris Mason 已提交
3230 3231 3232 3233 3234
		/*
		 * compressed and inline extents are written through other
		 * paths in the FS
		 */
		if (compressed || block_start == EXTENT_MAP_HOLE ||
3235
		    block_start == EXTENT_MAP_INLINE) {
C
Chris Mason 已提交
3236 3237 3238 3239 3240 3241
			/*
			 * end_io notification does not happen here for
			 * compressed extents
			 */
			if (!compressed && tree->ops &&
			    tree->ops->writepage_end_io_hook)
3242 3243 3244
				tree->ops->writepage_end_io_hook(page, cur,
							 cur + iosize - 1,
							 NULL, 1);
C
Chris Mason 已提交
3245 3246 3247 3248 3249 3250 3251 3252 3253
			else if (compressed) {
				/* we don't want to end_page_writeback on
				 * a compressed extent.  this happens
				 * elsewhere
				 */
				nr++;
			}

			cur += iosize;
3254
			pg_offset += iosize;
3255 3256 3257 3258
			continue;
		}
		/* leave this out until we have a page_mkwrite call */
		if (0 && !test_range_bit(tree, cur, cur + iosize - 1,
3259
				   EXTENT_DIRTY, 0, NULL)) {
3260
			cur = cur + iosize;
3261
			pg_offset += iosize;
3262 3263
			continue;
		}
C
Chris Mason 已提交
3264

3265 3266 3267 3268 3269 3270
		if (tree->ops && tree->ops->writepage_io_hook) {
			ret = tree->ops->writepage_io_hook(page, cur,
						cur + iosize - 1);
		} else {
			ret = 0;
		}
3271
		if (ret) {
3272
			SetPageError(page);
3273
		} else {
3274
			unsigned long max_nr = end_index + 1;
3275

3276 3277
			set_range_writeback(tree, cur, cur + iosize - 1);
			if (!PageWriteback(page)) {
C
Chris Mason 已提交
3278 3279 3280
				printk(KERN_ERR "btrfs warning page %lu not "
				       "writeback, cur %llu end %llu\n",
				       page->index, (unsigned long long)cur,
3281 3282 3283
				       (unsigned long long)end);
			}

3284 3285 3286
			ret = submit_extent_page(write_flags, tree, page,
						 sector, iosize, pg_offset,
						 bdev, &epd->bio, max_nr,
C
Chris Mason 已提交
3287 3288
						 end_bio_extent_writepage,
						 0, 0, 0);
3289 3290 3291 3292
			if (ret)
				SetPageError(page);
		}
		cur = cur + iosize;
3293
		pg_offset += iosize;
3294 3295 3296 3297 3298 3299 3300 3301 3302
		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);
3303

3304 3305
done_unlocked:

3306 3307
	/* drop our reference on any cached states */
	free_extent_state(cached_state);
3308 3309 3310
	return 0;
}

3311 3312 3313 3314 3315 3316
static int eb_wait(void *word)
{
	io_schedule();
	return 0;
}

3317
void wait_on_extent_buffer_writeback(struct extent_buffer *eb)
3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344
{
	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
3345 3346 3347 3348 3349
		while (1) {
			wait_on_extent_buffer_writeback(eb);
			btrfs_tree_lock(eb);
			if (!test_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags))
				break;
3350 3351 3352 3353
			btrfs_tree_unlock(eb);
		}
	}

3354 3355 3356 3357 3358 3359
	/*
	 * 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);
3360 3361
	if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
		set_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags);
3362
		spin_unlock(&eb->refs_lock);
3363
		btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
3364 3365 3366
		__percpu_counter_add(&fs_info->dirty_metadata_bytes,
				     -eb->len,
				     fs_info->dirty_metadata_batch);
3367
		ret = 1;
3368 3369
	} else {
		spin_unlock(&eb->refs_lock);
3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440
	}

	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;
3441
	unsigned long bio_flags = 0;
3442
	int rw = (epd->sync_io ? WRITE_SYNC : WRITE) | REQ_META;
3443
	int ret = 0;
3444 3445 3446 3447

	clear_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
	num_pages = num_extent_pages(eb->start, eb->len);
	atomic_set(&eb->io_pages, num_pages);
3448 3449 3450
	if (btrfs_header_owner(eb) == BTRFS_TREE_LOG_OBJECTID)
		bio_flags = EXTENT_BIO_TREE_LOG;

3451 3452 3453 3454 3455 3456 3457 3458
	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,
3459 3460
					 0, epd->bio_flags, bio_flags);
		epd->bio_flags = bio_flags;
3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494
		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,
3495
		.bio_flags = 0,
3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 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
	};
	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;
			}

3540 3541 3542 3543 3544 3545
			spin_lock(&mapping->private_lock);
			if (!PagePrivate(page)) {
				spin_unlock(&mapping->private_lock);
				continue;
			}

3546
			eb = (struct extent_buffer *)page->private;
3547 3548 3549 3550 3551 3552

			/*
			 * 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.
			 */
3553
			if (!eb) {
3554
				spin_unlock(&mapping->private_lock);
3555 3556 3557 3558
				WARN_ON(1);
				continue;
			}

3559 3560
			if (eb == prev_eb) {
				spin_unlock(&mapping->private_lock);
3561
				continue;
3562
			}
3563

3564 3565 3566
			ret = atomic_inc_not_zero(&eb->refs);
			spin_unlock(&mapping->private_lock);
			if (!ret)
3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606
				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;
}

3607
/**
C
Chris Mason 已提交
3608
 * write_cache_pages - walk the list of dirty pages of the given address space and write all of them.
3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621
 * @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.
 */
3622
static int extent_write_cache_pages(struct extent_io_tree *tree,
C
Chris Mason 已提交
3623 3624
			     struct address_space *mapping,
			     struct writeback_control *wbc,
C
Chris Mason 已提交
3625 3626
			     writepage_t writepage, void *data,
			     void (*flush_fn)(void *))
3627
{
3628
	struct inode *inode = mapping->host;
3629 3630
	int ret = 0;
	int done = 0;
3631
	int nr_to_write_done = 0;
3632 3633 3634 3635 3636
	struct pagevec pvec;
	int nr_pages;
	pgoff_t index;
	pgoff_t end;		/* Inclusive */
	int scanned = 0;
3637
	int tag;
3638

3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650
	/*
	 * 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;

3651 3652 3653 3654 3655 3656 3657 3658 3659
	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;
	}
3660 3661 3662 3663
	if (wbc->sync_mode == WB_SYNC_ALL)
		tag = PAGECACHE_TAG_TOWRITE;
	else
		tag = PAGECACHE_TAG_DIRTY;
3664
retry:
3665 3666
	if (wbc->sync_mode == WB_SYNC_ALL)
		tag_pages_for_writeback(mapping, index, end);
3667
	while (!done && !nr_to_write_done && (index <= end) &&
3668 3669
	       (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag,
			min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682
		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
			 */
3683 3684 3685
			if (!trylock_page(page)) {
				flush_fn(data);
				lock_page(page);
3686
			}
3687 3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698

			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
3699
			if (wbc->sync_mode != WB_SYNC_NONE) {
3700 3701
				if (PageWriteback(page))
					flush_fn(data);
3702
				wait_on_page_writeback(page);
C
Chris Mason 已提交
3703
			}
3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716

			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;
			}
3717
			if (ret)
3718
				done = 1;
3719 3720 3721 3722 3723 3724 3725

			/*
			 * 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;
3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738
		}
		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;
	}
3739
	btrfs_add_delayed_iput(inode);
3740 3741 3742
	return ret;
}

3743
static void flush_epd_write_bio(struct extent_page_data *epd)
C
Chris Mason 已提交
3744 3745
{
	if (epd->bio) {
3746 3747 3748
		int rw = WRITE;
		int ret;

3749
		if (epd->sync_io)
3750 3751
			rw = WRITE_SYNC;

3752
		ret = submit_one_bio(rw, epd->bio, 0, epd->bio_flags);
3753
		BUG_ON(ret < 0); /* -ENOMEM */
C
Chris Mason 已提交
3754 3755 3756 3757
		epd->bio = NULL;
	}
}

3758 3759 3760 3761 3762 3763
static noinline void flush_write_bio(void *data)
{
	struct extent_page_data *epd = data;
	flush_epd_write_bio(epd);
}

3764 3765 3766 3767 3768 3769 3770 3771 3772
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,
3773
		.extent_locked = 0,
3774
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
3775
		.bio_flags = 0,
3776 3777 3778 3779
	};

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

3780
	flush_epd_write_bio(&epd);
3781 3782 3783
	return ret;
}

3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798
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,
3799
		.sync_io = mode == WB_SYNC_ALL,
3800
		.bio_flags = 0,
3801 3802 3803 3804 3805 3806 3807 3808
	};
	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
3809
	while (start <= end) {
3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823
		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;
	}

3824
	flush_epd_write_bio(&epd);
3825 3826
	return ret;
}
3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837

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,
3838
		.extent_locked = 0,
3839
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
3840
		.bio_flags = 0,
3841 3842
	};

C
Chris Mason 已提交
3843
	ret = extent_write_cache_pages(tree, mapping, wbc,
C
Chris Mason 已提交
3844 3845
				       __extent_writepage, &epd,
				       flush_write_bio);
3846
	flush_epd_write_bio(&epd);
3847 3848 3849 3850 3851 3852 3853 3854 3855 3856
	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
3857
	unsigned long bio_flags = 0;
L
Liu Bo 已提交
3858 3859
	struct page *pagepool[16];
	struct page *page;
3860
	struct extent_map *em_cached = NULL;
L
Liu Bo 已提交
3861
	int nr = 0;
3862 3863

	for (page_idx = 0; page_idx < nr_pages; page_idx++) {
L
Liu Bo 已提交
3864
		page = list_entry(pages->prev, struct page, lru);
3865 3866 3867

		prefetchw(&page->flags);
		list_del(&page->lru);
L
Liu Bo 已提交
3868
		if (add_to_page_cache_lru(page, mapping,
3869
					page->index, GFP_NOFS)) {
L
Liu Bo 已提交
3870 3871
			page_cache_release(page);
			continue;
3872
		}
L
Liu Bo 已提交
3873 3874 3875 3876

		pagepool[nr++] = page;
		if (nr < ARRAY_SIZE(pagepool))
			continue;
3877
		__extent_readpages(tree, pagepool, nr, get_extent, &em_cached,
3878
				   &bio, 0, &bio_flags, READ);
L
Liu Bo 已提交
3879
		nr = 0;
3880
	}
3881
	if (nr)
3882
		__extent_readpages(tree, pagepool, nr, get_extent, &em_cached,
3883
				   &bio, 0, &bio_flags, READ);
L
Liu Bo 已提交
3884

3885 3886 3887
	if (em_cached)
		free_extent_map(em_cached);

3888 3889
	BUG_ON(!list_empty(pages));
	if (bio)
3890
		return submit_one_bio(READ, bio, 0, bio_flags);
3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901
	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)
{
3902
	struct extent_state *cached_state = NULL;
M
Miao Xie 已提交
3903
	u64 start = page_offset(page);
3904 3905 3906
	u64 end = start + PAGE_CACHE_SIZE - 1;
	size_t blocksize = page->mapping->host->i_sb->s_blocksize;

3907
	start += ALIGN(offset, blocksize);
3908 3909 3910
	if (start > end)
		return 0;

3911
	lock_extent_bits(tree, start, end, 0, &cached_state);
3912
	wait_on_page_writeback(page);
3913
	clear_extent_bit(tree, start, end,
3914 3915
			 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
			 EXTENT_DO_ACCOUNTING,
3916
			 1, 1, &cached_state, GFP_NOFS);
3917 3918 3919
	return 0;
}

3920 3921 3922 3923 3924
/*
 * 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.
 */
3925 3926 3927
static int try_release_extent_state(struct extent_map_tree *map,
				    struct extent_io_tree *tree,
				    struct page *page, gfp_t mask)
3928
{
M
Miao Xie 已提交
3929
	u64 start = page_offset(page);
3930 3931 3932
	u64 end = start + PAGE_CACHE_SIZE - 1;
	int ret = 1;

3933
	if (test_range_bit(tree, start, end,
3934
			   EXTENT_IOBITS, 0, NULL))
3935 3936 3937 3938
		ret = 0;
	else {
		if ((mask & GFP_NOFS) == GFP_NOFS)
			mask = GFP_NOFS;
3939 3940 3941 3942
		/*
		 * at this point we can safely clear everything except the
		 * locked bit and the nodatasum bit
		 */
3943
		ret = clear_extent_bit(tree, start, end,
3944 3945
				 ~(EXTENT_LOCKED | EXTENT_NODATASUM),
				 0, 0, NULL, mask);
3946 3947 3948 3949 3950 3951 3952 3953

		/* if clear_extent_bit failed for enomem reasons,
		 * we can't allow the release to continue.
		 */
		if (ret < 0)
			ret = 0;
		else
			ret = 1;
3954 3955 3956 3957
	}
	return ret;
}

3958 3959 3960 3961 3962 3963
/*
 * 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,
3964 3965
			       struct extent_io_tree *tree, struct page *page,
			       gfp_t mask)
3966 3967
{
	struct extent_map *em;
M
Miao Xie 已提交
3968
	u64 start = page_offset(page);
3969
	u64 end = start + PAGE_CACHE_SIZE - 1;
3970

3971 3972
	if ((mask & __GFP_WAIT) &&
	    page->mapping->host->i_size > 16 * 1024 * 1024) {
3973
		u64 len;
3974
		while (start <= end) {
3975
			len = end - start + 1;
3976
			write_lock(&map->lock);
3977
			em = lookup_extent_mapping(map, start, len);
3978
			if (!em) {
3979
				write_unlock(&map->lock);
3980 3981
				break;
			}
3982 3983
			if (test_bit(EXTENT_FLAG_PINNED, &em->flags) ||
			    em->start != start) {
3984
				write_unlock(&map->lock);
3985 3986 3987 3988 3989
				free_extent_map(em);
				break;
			}
			if (!test_range_bit(tree, em->start,
					    extent_map_end(em) - 1,
3990
					    EXTENT_LOCKED | EXTENT_WRITEBACK,
3991
					    0, NULL)) {
3992 3993 3994 3995 3996
				remove_extent_mapping(map, em);
				/* once for the rb tree */
				free_extent_map(em);
			}
			start = extent_map_end(em);
3997
			write_unlock(&map->lock);
3998 3999

			/* once for us */
4000 4001 4002
			free_extent_map(em);
		}
	}
4003
	return try_release_extent_state(map, tree, page, mask);
4004 4005
}

4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025
/*
 * 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;
4026
		len = ALIGN(len, sectorsize);
4027
		em = get_extent(inode, NULL, 0, offset, len, 0);
4028
		if (IS_ERR_OR_NULL(em))
4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045
			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
4046 4047 4048
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
4049
	int ret = 0;
Y
Yehuda Sadeh 已提交
4050 4051 4052
	u64 off = start;
	u64 max = start + len;
	u32 flags = 0;
J
Josef Bacik 已提交
4053 4054
	u32 found_type;
	u64 last;
4055
	u64 last_for_get_extent = 0;
Y
Yehuda Sadeh 已提交
4056
	u64 disko = 0;
4057
	u64 isize = i_size_read(inode);
J
Josef Bacik 已提交
4058
	struct btrfs_key found_key;
Y
Yehuda Sadeh 已提交
4059
	struct extent_map *em = NULL;
4060
	struct extent_state *cached_state = NULL;
J
Josef Bacik 已提交
4061 4062
	struct btrfs_path *path;
	struct btrfs_file_extent_item *item;
Y
Yehuda Sadeh 已提交
4063
	int end = 0;
4064 4065 4066
	u64 em_start = 0;
	u64 em_len = 0;
	u64 em_end = 0;
Y
Yehuda Sadeh 已提交
4067 4068 4069 4070 4071
	unsigned long emflags;

	if (len == 0)
		return -EINVAL;

J
Josef Bacik 已提交
4072 4073 4074 4075 4076
	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;
	path->leave_spinning = 1;

4077 4078 4079
	start = ALIGN(start, BTRFS_I(inode)->root->sectorsize);
	len = ALIGN(len, BTRFS_I(inode)->root->sectorsize);

4080 4081 4082 4083
	/*
	 * 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
4084
	ret = btrfs_lookup_file_extent(NULL, BTRFS_I(inode)->root,
L
Li Zefan 已提交
4085
				       path, btrfs_ino(inode), -1, 0);
J
Josef Bacik 已提交
4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096
	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);

4097
	/* No extents, but there might be delalloc bits */
L
Li Zefan 已提交
4098
	if (found_key.objectid != btrfs_ino(inode) ||
J
Josef Bacik 已提交
4099
	    found_type != BTRFS_EXTENT_DATA_KEY) {
4100 4101 4102 4103 4104 4105 4106 4107 4108 4109 4110
		/* 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
4111 4112 4113
	}
	btrfs_free_path(path);

4114 4115 4116 4117 4118 4119 4120 4121 4122 4123
	/*
	 * 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
4124
	lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + len - 1, 0,
4125
			 &cached_state);
4126

4127
	em = get_extent_skip_holes(inode, start, last_for_get_extent,
4128
				   get_extent);
Y
Yehuda Sadeh 已提交
4129 4130 4131 4132 4133 4134
	if (!em)
		goto out;
	if (IS_ERR(em)) {
		ret = PTR_ERR(em);
		goto out;
	}
J
Josef Bacik 已提交
4135

Y
Yehuda Sadeh 已提交
4136
	while (!end) {
4137
		u64 offset_in_extent = 0;
4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149

		/* 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
4150

4151 4152
		/*
		 * record the offset from the start of the extent
4153 4154 4155
		 * 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.
4156
		 */
4157 4158
		if (!test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
			offset_in_extent = em_start - em->start;
4159
		em_end = extent_map_end(em);
4160
		em_len = em_end - em_start;
4161
		emflags = em->flags;
Y
Yehuda Sadeh 已提交
4162 4163 4164
		disko = 0;
		flags = 0;

4165 4166 4167 4168 4169 4170 4171
		/*
		 * bump off for our next call to get_extent
		 */
		off = extent_map_end(em);
		if (off >= max)
			end = 1;

4172
		if (em->block_start == EXTENT_MAP_LAST_BYTE) {
Y
Yehuda Sadeh 已提交
4173 4174
			end = 1;
			flags |= FIEMAP_EXTENT_LAST;
4175
		} else if (em->block_start == EXTENT_MAP_INLINE) {
Y
Yehuda Sadeh 已提交
4176 4177
			flags |= (FIEMAP_EXTENT_DATA_INLINE |
				  FIEMAP_EXTENT_NOT_ALIGNED);
4178
		} else if (em->block_start == EXTENT_MAP_DELALLOC) {
Y
Yehuda Sadeh 已提交
4179 4180
			flags |= (FIEMAP_EXTENT_DELALLOC |
				  FIEMAP_EXTENT_UNKNOWN);
4181
		} else {
4182
			disko = em->block_start + offset_in_extent;
Y
Yehuda Sadeh 已提交
4183 4184 4185 4186 4187 4188
		}
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
			flags |= FIEMAP_EXTENT_ENCODED;

		free_extent_map(em);
		em = NULL;
4189 4190
		if ((em_start >= last) || em_len == (u64)-1 ||
		   (last == (u64)-1 && isize <= em_end)) {
Y
Yehuda Sadeh 已提交
4191 4192 4193 4194
			flags |= FIEMAP_EXTENT_LAST;
			end = 1;
		}

4195 4196 4197 4198 4199 4200 4201 4202
		/* 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
4203 4204 4205
			flags |= FIEMAP_EXTENT_LAST;
			end = 1;
		}
4206 4207 4208 4209
		ret = fiemap_fill_next_extent(fieinfo, em_start, disko,
					      em_len, flags);
		if (ret)
			goto out_free;
Y
Yehuda Sadeh 已提交
4210 4211 4212 4213
	}
out_free:
	free_extent_map(em);
out:
L
Liu Bo 已提交
4214
	unlock_extent_cached(&BTRFS_I(inode)->io_tree, start, start + len - 1,
4215
			     &cached_state, GFP_NOFS);
Y
Yehuda Sadeh 已提交
4216 4217 4218
	return ret;
}

4219 4220
static void __free_extent_buffer(struct extent_buffer *eb)
{
4221
	btrfs_leak_debug_del(&eb->leak_list);
4222 4223 4224
	kmem_cache_free(extent_buffer_cache, eb);
}

4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242 4243 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
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);
}

4295 4296 4297 4298 4299 4300 4301 4302
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
4303 4304
	if (eb == NULL)
		return NULL;
4305 4306
	eb->start = start;
	eb->len = len;
J
Josef Bacik 已提交
4307
	eb->tree = tree;
4308
	eb->bflags = 0;
4309 4310 4311 4312 4313 4314 4315
	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);
4316
	eb->lock_nested = 0;
4317 4318
	init_waitqueue_head(&eb->write_lock_wq);
	init_waitqueue_head(&eb->read_lock_wq);
4319

4320 4321
	btrfs_leak_debug_add(&eb->leak_list, &buffers);

4322
	spin_lock_init(&eb->refs_lock);
4323
	atomic_set(&eb->refs, 1);
4324
	atomic_set(&eb->io_pages, 0);
4325

4326 4327 4328 4329 4330 4331
	/*
	 * 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);
4332 4333 4334 4335

	return eb;
}

4336 4337 4338 4339 4340 4341 4342
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);

4343
	new = __alloc_extent_buffer(NULL, src->start, src->len, GFP_NOFS);
4344 4345 4346 4347
	if (new == NULL)
		return NULL;

	for (i = 0; i < num_pages; i++) {
4348
		p = alloc_page(GFP_NOFS);
4349 4350 4351 4352
		if (!p) {
			btrfs_release_extent_buffer(new);
			return NULL;
		}
4353 4354 4355 4356 4357 4358 4359 4360 4361 4362 4363 4364 4365 4366 4367 4368 4369 4370 4371
		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;

4372
	eb = __alloc_extent_buffer(NULL, start, len, GFP_NOFS);
4373 4374 4375 4376
	if (!eb)
		return NULL;

	for (i = 0; i < num_pages; i++) {
4377
		eb->pages[i] = alloc_page(GFP_NOFS);
4378 4379 4380 4381 4382 4383 4384 4385 4386
		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:
4387 4388
	for (; i > 0; i--)
		__free_page(eb->pages[i - 1]);
4389 4390 4391 4392
	__free_extent_buffer(eb);
	return NULL;
}

4393 4394
static void check_buffer_tree_ref(struct extent_buffer *eb)
{
4395
	int refs;
4396 4397 4398 4399 4400 4401 4402 4403 4404 4405 4406 4407 4408 4409 4410 4411 4412 4413 4414 4415
	/* 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.
	 */
4416 4417 4418 4419
	refs = atomic_read(&eb->refs);
	if (refs >= 2 && test_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
		return;

4420 4421
	spin_lock(&eb->refs_lock);
	if (!test_and_set_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
4422
		atomic_inc(&eb->refs);
4423
	spin_unlock(&eb->refs_lock);
4424 4425
}

4426 4427 4428 4429
static void mark_extent_buffer_accessed(struct extent_buffer *eb)
{
	unsigned long num_pages, i;

4430 4431
	check_buffer_tree_ref(eb);

4432 4433 4434 4435 4436 4437 4438
	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);
	}
}

4439
struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree,
4440
					  u64 start, unsigned long len)
4441 4442 4443 4444 4445
{
	unsigned long num_pages = num_extent_pages(start, len);
	unsigned long i;
	unsigned long index = start >> PAGE_CACHE_SHIFT;
	struct extent_buffer *eb;
4446
	struct extent_buffer *exists = NULL;
4447 4448 4449
	struct page *p;
	struct address_space *mapping = tree->mapping;
	int uptodate = 1;
4450
	int ret;
4451

4452 4453 4454 4455
	rcu_read_lock();
	eb = radix_tree_lookup(&tree->buffer, start >> PAGE_CACHE_SHIFT);
	if (eb && atomic_inc_not_zero(&eb->refs)) {
		rcu_read_unlock();
4456
		mark_extent_buffer_accessed(eb);
4457 4458
		return eb;
	}
4459
	rcu_read_unlock();
4460

4461
	eb = __alloc_extent_buffer(tree, start, len, GFP_NOFS);
4462
	if (!eb)
4463 4464
		return NULL;

4465
	for (i = 0; i < num_pages; i++, index++) {
4466
		p = find_or_create_page(mapping, index, GFP_NOFS);
4467
		if (!p)
4468
			goto free_eb;
J
Josef Bacik 已提交
4469 4470 4471 4472 4473 4474 4475 4476 4477 4478 4479 4480 4481 4482

		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);
4483
				page_cache_release(p);
4484
				mark_extent_buffer_accessed(exists);
J
Josef Bacik 已提交
4485 4486 4487
				goto free_eb;
			}

4488
			/*
J
Josef Bacik 已提交
4489 4490 4491 4492
			 * Do this so attach doesn't complain and we need to
			 * drop the ref the old guy had.
			 */
			ClearPagePrivate(p);
4493
			WARN_ON(PageDirty(p));
J
Josef Bacik 已提交
4494
			page_cache_release(p);
4495
		}
J
Josef Bacik 已提交
4496 4497
		attach_extent_buffer_page(eb, p);
		spin_unlock(&mapping->private_lock);
4498
		WARN_ON(PageDirty(p));
4499
		mark_page_accessed(p);
4500
		eb->pages[i] = p;
4501 4502
		if (!PageUptodate(p))
			uptodate = 0;
C
Chris Mason 已提交
4503 4504 4505 4506 4507

		/*
		 * see below about how we avoid a nasty race with release page
		 * and why we unlock later
		 */
4508 4509
	}
	if (uptodate)
4510
		set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
4511
again:
4512 4513 4514 4515
	ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
	if (ret)
		goto free_eb;

4516
	spin_lock(&tree->buffer_lock);
4517 4518 4519 4520
	ret = radix_tree_insert(&tree->buffer, start >> PAGE_CACHE_SHIFT, eb);
	if (ret == -EEXIST) {
		exists = radix_tree_lookup(&tree->buffer,
						start >> PAGE_CACHE_SHIFT);
4521 4522 4523 4524 4525 4526
		if (!atomic_inc_not_zero(&exists->refs)) {
			spin_unlock(&tree->buffer_lock);
			radix_tree_preload_end();
			exists = NULL;
			goto again;
		}
4527
		spin_unlock(&tree->buffer_lock);
4528
		radix_tree_preload_end();
4529
		mark_extent_buffer_accessed(exists);
4530 4531 4532
		goto free_eb;
	}
	/* add one reference for the tree */
4533
	check_buffer_tree_ref(eb);
4534
	spin_unlock(&tree->buffer_lock);
4535
	radix_tree_preload_end();
C
Chris Mason 已提交
4536 4537 4538 4539 4540 4541 4542 4543 4544 4545

	/*
	 * 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
	 */
4546 4547 4548 4549 4550 4551 4552
	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]);
4553 4554
	return eb;

4555
free_eb:
4556 4557 4558 4559
	for (i = 0; i < num_pages; i++) {
		if (eb->pages[i])
			unlock_page(eb->pages[i]);
	}
C
Chris Mason 已提交
4560

4561
	WARN_ON(!atomic_dec_and_test(&eb->refs));
4562
	btrfs_release_extent_buffer(eb);
4563
	return exists;
4564 4565 4566
}

struct extent_buffer *find_extent_buffer(struct extent_io_tree *tree,
4567
					 u64 start, unsigned long len)
4568 4569 4570
{
	struct extent_buffer *eb;

4571 4572 4573 4574
	rcu_read_lock();
	eb = radix_tree_lookup(&tree->buffer, start >> PAGE_CACHE_SHIFT);
	if (eb && atomic_inc_not_zero(&eb->refs)) {
		rcu_read_unlock();
4575
		mark_extent_buffer_accessed(eb);
4576 4577 4578
		return eb;
	}
	rcu_read_unlock();
J
Josef Bacik 已提交
4579

4580
	return NULL;
4581 4582
}

4583 4584 4585 4586 4587 4588 4589 4590 4591
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 */
4592
static int release_extent_buffer(struct extent_buffer *eb)
4593 4594 4595
{
	WARN_ON(atomic_read(&eb->refs) == 0);
	if (atomic_dec_and_test(&eb->refs)) {
4596 4597 4598 4599
		if (test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags)) {
			spin_unlock(&eb->refs_lock);
		} else {
			struct extent_io_tree *tree = eb->tree;
4600

4601
			spin_unlock(&eb->refs_lock);
4602

4603 4604 4605 4606 4607
			spin_lock(&tree->buffer_lock);
			radix_tree_delete(&tree->buffer,
					  eb->start >> PAGE_CACHE_SHIFT);
			spin_unlock(&tree->buffer_lock);
		}
4608 4609 4610 4611

		/* 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);
4612
		return 1;
4613 4614
	}
	spin_unlock(&eb->refs_lock);
4615 4616

	return 0;
4617 4618
}

4619 4620
void free_extent_buffer(struct extent_buffer *eb)
{
4621 4622
	int refs;
	int old;
4623 4624 4625
	if (!eb)
		return;

4626 4627 4628 4629 4630 4631 4632 4633 4634
	while (1) {
		refs = atomic_read(&eb->refs);
		if (refs <= 3)
			break;
		old = atomic_cmpxchg(&eb->refs, refs, refs - 1);
		if (old == refs)
			return;
	}

4635
	spin_lock(&eb->refs_lock);
4636 4637 4638 4639
	if (atomic_read(&eb->refs) == 2 &&
	    test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags))
		atomic_dec(&eb->refs);

4640 4641
	if (atomic_read(&eb->refs) == 2 &&
	    test_bit(EXTENT_BUFFER_STALE, &eb->bflags) &&
4642
	    !extent_buffer_under_io(eb) &&
4643 4644 4645 4646 4647 4648 4649
	    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.
	 */
4650
	release_extent_buffer(eb);
4651 4652 4653 4654 4655
}

void free_extent_buffer_stale(struct extent_buffer *eb)
{
	if (!eb)
4656 4657
		return;

4658 4659 4660
	spin_lock(&eb->refs_lock);
	set_bit(EXTENT_BUFFER_STALE, &eb->bflags);

4661
	if (atomic_read(&eb->refs) == 2 && !extent_buffer_under_io(eb) &&
4662 4663
	    test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
		atomic_dec(&eb->refs);
4664
	release_extent_buffer(eb);
4665 4666
}

4667
void clear_extent_buffer_dirty(struct extent_buffer *eb)
4668 4669 4670 4671 4672 4673 4674 4675 4676
{
	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);
4677
		if (!PageDirty(page))
C
Chris Mason 已提交
4678 4679
			continue;

4680
		lock_page(page);
C
Chris Mason 已提交
4681 4682
		WARN_ON(!PagePrivate(page));

4683
		clear_page_dirty_for_io(page);
4684
		spin_lock_irq(&page->mapping->tree_lock);
4685 4686 4687 4688 4689
		if (!PageDirty(page)) {
			radix_tree_tag_clear(&page->mapping->page_tree,
						page_index(page),
						PAGECACHE_TAG_DIRTY);
		}
4690
		spin_unlock_irq(&page->mapping->tree_lock);
4691
		ClearPageError(page);
4692
		unlock_page(page);
4693
	}
4694
	WARN_ON(atomic_read(&eb->refs) == 0);
4695 4696
}

4697
int set_extent_buffer_dirty(struct extent_buffer *eb)
4698 4699 4700
{
	unsigned long i;
	unsigned long num_pages;
4701
	int was_dirty = 0;
4702

4703 4704
	check_buffer_tree_ref(eb);

4705
	was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
4706

4707
	num_pages = num_extent_pages(eb->start, eb->len);
4708
	WARN_ON(atomic_read(&eb->refs) == 0);
4709 4710
	WARN_ON(!test_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags));

4711
	for (i = 0; i < num_pages; i++)
4712
		set_page_dirty(extent_buffer_page(eb, i));
4713
	return was_dirty;
4714 4715
}

4716
int clear_extent_buffer_uptodate(struct extent_buffer *eb)
4717 4718 4719 4720 4721
{
	unsigned long i;
	struct page *page;
	unsigned long num_pages;

4722
	clear_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
4723
	num_pages = num_extent_pages(eb->start, eb->len);
4724 4725
	for (i = 0; i < num_pages; i++) {
		page = extent_buffer_page(eb, i);
C
Chris Mason 已提交
4726 4727
		if (page)
			ClearPageUptodate(page);
4728 4729 4730 4731
	}
	return 0;
}

4732
int set_extent_buffer_uptodate(struct extent_buffer *eb)
4733 4734 4735 4736 4737
{
	unsigned long i;
	struct page *page;
	unsigned long num_pages;

4738
	set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
4739 4740 4741 4742 4743 4744 4745 4746
	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;
}

4747
int extent_buffer_uptodate(struct extent_buffer *eb)
4748
{
4749
	return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
4750 4751 4752
}

int read_extent_buffer_pages(struct extent_io_tree *tree,
4753
			     struct extent_buffer *eb, u64 start, int wait,
4754
			     get_extent_t *get_extent, int mirror_num)
4755 4756 4757 4758 4759 4760
{
	unsigned long i;
	unsigned long start_i;
	struct page *page;
	int err;
	int ret = 0;
4761 4762
	int locked_pages = 0;
	int all_uptodate = 1;
4763
	unsigned long num_pages;
4764
	unsigned long num_reads = 0;
4765
	struct bio *bio = NULL;
C
Chris Mason 已提交
4766
	unsigned long bio_flags = 0;
4767

4768
	if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags))
4769 4770 4771 4772 4773 4774 4775 4776 4777 4778 4779 4780 4781
		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);
4782
		if (wait == WAIT_NONE) {
4783
			if (!trylock_page(page))
4784
				goto unlock_exit;
4785 4786 4787
		} else {
			lock_page(page);
		}
4788
		locked_pages++;
4789 4790
		if (!PageUptodate(page)) {
			num_reads++;
4791
			all_uptodate = 0;
4792
		}
4793 4794 4795
	}
	if (all_uptodate) {
		if (start_i == 0)
4796
			set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
4797 4798 4799
		goto unlock_exit;
	}

4800
	clear_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
4801
	eb->read_mirror = 0;
4802
	atomic_set(&eb->io_pages, num_reads);
4803 4804 4805
	for (i = start_i; i < num_pages; i++) {
		page = extent_buffer_page(eb, i);
		if (!PageUptodate(page)) {
4806
			ClearPageError(page);
4807
			err = __extent_read_full_page(tree, page,
4808
						      get_extent, &bio,
4809 4810
						      mirror_num, &bio_flags,
						      READ | REQ_META);
C
Chris Mason 已提交
4811
			if (err)
4812 4813 4814 4815 4816 4817
				ret = err;
		} else {
			unlock_page(page);
		}
	}

4818
	if (bio) {
4819 4820
		err = submit_one_bio(READ | REQ_META, bio, mirror_num,
				     bio_flags);
4821 4822
		if (err)
			return err;
4823
	}
4824

4825
	if (ret || wait != WAIT_COMPLETE)
4826
		return ret;
C
Chris Mason 已提交
4827

4828 4829 4830
	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
4831
		if (!PageUptodate(page))
4832 4833
			ret = -EIO;
	}
C
Chris Mason 已提交
4834

4835
	return ret;
4836 4837 4838

unlock_exit:
	i = start_i;
C
Chris Mason 已提交
4839
	while (locked_pages > 0) {
4840 4841 4842 4843 4844 4845
		page = extent_buffer_page(eb, i);
		i++;
		unlock_page(page);
		locked_pages--;
	}
	return ret;
4846 4847 4848 4849 4850 4851 4852 4853 4854 4855 4856 4857 4858 4859 4860 4861 4862 4863 4864
}

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);

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

C
Chris Mason 已提交
4865
	while (len > 0) {
4866 4867 4868
		page = extent_buffer_page(eb, i);

		cur = min(len, (PAGE_CACHE_SIZE - offset));
4869
		kaddr = page_address(page);
4870 4871 4872 4873 4874 4875 4876 4877 4878 4879
		memcpy(dst, kaddr + offset, cur);

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

int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start,
4880
			       unsigned long min_len, char **map,
4881
			       unsigned long *map_start,
4882
			       unsigned long *map_len)
4883 4884 4885 4886 4887 4888 4889 4890 4891 4892 4893 4894 4895 4896 4897 4898 4899 4900 4901
{
	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
4902

4903
	if (start + min_len > eb->len) {
J
Julia Lawall 已提交
4904
		WARN(1, KERN_ERR "btrfs bad mapping eb start %llu len %lu, "
C
Chris Mason 已提交
4905 4906
		       "wanted %lu %lu\n", (unsigned long long)eb->start,
		       eb->len, start, min_len);
4907
		return -EINVAL;
4908 4909 4910
	}

	p = extent_buffer_page(eb, i);
4911
	kaddr = page_address(p);
4912 4913 4914 4915 4916 4917 4918 4919 4920 4921 4922 4923 4924 4925 4926 4927 4928 4929 4930 4931 4932 4933 4934
	*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);

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

C
Chris Mason 已提交
4935
	while (len > 0) {
4936 4937 4938 4939
		page = extent_buffer_page(eb, i);

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

4940
		kaddr = page_address(page);
4941 4942 4943 4944 4945 4946 4947 4948 4949 4950 4951 4952 4953 4954 4955 4956 4957 4958 4959 4960 4961 4962 4963 4964 4965 4966 4967 4968
		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);

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

C
Chris Mason 已提交
4969
	while (len > 0) {
4970 4971 4972 4973
		page = extent_buffer_page(eb, i);
		WARN_ON(!PageUptodate(page));

		cur = min(len, PAGE_CACHE_SIZE - offset);
4974
		kaddr = page_address(page);
4975 4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986 4987 4988 4989 4990 4991 4992 4993 4994 4995 4996 4997 4998
		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);

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

C
Chris Mason 已提交
4999
	while (len > 0) {
5000 5001 5002 5003
		page = extent_buffer_page(eb, i);
		WARN_ON(!PageUptodate(page));

		cur = min(len, PAGE_CACHE_SIZE - offset);
5004
		kaddr = page_address(page);
5005 5006 5007 5008 5009 5010 5011 5012 5013 5014 5015 5016 5017 5018 5019 5020 5021 5022 5023 5024 5025 5026 5027 5028 5029
		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) &
		((unsigned long)PAGE_CACHE_SIZE - 1);

C
Chris Mason 已提交
5030
	while (len > 0) {
5031 5032 5033 5034 5035
		page = extent_buffer_page(dst, i);
		WARN_ON(!PageUptodate(page));

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

5036
		kaddr = page_address(page);
5037 5038 5039 5040 5041 5042 5043 5044 5045 5046 5047 5048 5049
		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)
{
5050
	char *dst_kaddr = page_address(dst_page);
5051 5052 5053
	if (dst_page == src_page) {
		memmove(dst_kaddr + dst_off, dst_kaddr + src_off, len);
	} else {
5054
		char *src_kaddr = page_address(src_page);
5055 5056 5057 5058 5059 5060 5061 5062
		char *p = dst_kaddr + dst_off + len;
		char *s = src_kaddr + src_off + len;

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

5063 5064 5065 5066 5067 5068
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;
}

5069 5070 5071 5072
static void copy_pages(struct page *dst_page, struct page *src_page,
		       unsigned long dst_off, unsigned long src_off,
		       unsigned long len)
{
5073
	char *dst_kaddr = page_address(dst_page);
5074
	char *src_kaddr;
5075
	int must_memmove = 0;
5076

5077
	if (dst_page != src_page) {
5078
		src_kaddr = page_address(src_page);
5079
	} else {
5080
		src_kaddr = dst_kaddr;
5081 5082
		if (areas_overlap(src_off, dst_off, len))
			must_memmove = 1;
5083
	}
5084

5085 5086 5087 5088
	if (must_memmove)
		memmove(dst_kaddr + dst_off, src_kaddr + src_off, len);
	else
		memcpy(dst_kaddr + dst_off, src_kaddr + src_off, len);
5089 5090 5091 5092 5093 5094 5095 5096 5097 5098 5099 5100 5101
}

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
5102 5103
		printk(KERN_ERR "btrfs memmove bogus src_offset %lu move "
		       "len %lu dst len %lu\n", src_offset, len, dst->len);
5104 5105 5106
		BUG_ON(1);
	}
	if (dst_offset + len > dst->len) {
C
Chris Mason 已提交
5107 5108
		printk(KERN_ERR "btrfs memmove bogus dst_offset %lu move "
		       "len %lu dst len %lu\n", dst_offset, len, dst->len);
5109 5110 5111
		BUG_ON(1);
	}

C
Chris Mason 已提交
5112
	while (len > 0) {
5113 5114 5115 5116 5117 5118 5119 5120 5121 5122 5123 5124 5125 5126 5127 5128 5129 5130 5131 5132 5133 5134 5135 5136 5137 5138 5139 5140 5141 5142 5143 5144 5145 5146 5147 5148
		dst_off_in_page = (start_offset + dst_offset) &
			((unsigned long)PAGE_CACHE_SIZE - 1);
		src_off_in_page = (start_offset + src_offset) &
			((unsigned long)PAGE_CACHE_SIZE - 1);

		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
5149 5150
		printk(KERN_ERR "btrfs memmove bogus src_offset %lu move "
		       "len %lu len %lu\n", src_offset, len, dst->len);
5151 5152 5153
		BUG_ON(1);
	}
	if (dst_offset + len > dst->len) {
C
Chris Mason 已提交
5154 5155
		printk(KERN_ERR "btrfs memmove bogus dst_offset %lu move "
		       "len %lu len %lu\n", dst_offset, len, dst->len);
5156 5157
		BUG_ON(1);
	}
5158
	if (dst_offset < src_offset) {
5159 5160 5161
		memcpy_extent_buffer(dst, dst_offset, src_offset, len);
		return;
	}
C
Chris Mason 已提交
5162
	while (len > 0) {
5163 5164 5165 5166 5167 5168 5169 5170 5171 5172 5173 5174 5175 5176 5177 5178 5179 5180 5181 5182
		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) &
			((unsigned long)PAGE_CACHE_SIZE - 1);
		src_off_in_page = (start_offset + src_end) &
			((unsigned long)PAGE_CACHE_SIZE - 1);

		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;
	}
}
5183

5184
int try_release_extent_buffer(struct page *page)
5185
{
5186 5187
	struct extent_buffer *eb;

5188 5189 5190 5191 5192 5193 5194
	/*
	 * 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
5195
		return 1;
5196
	}
5197

5198 5199
	eb = (struct extent_buffer *)page->private;
	BUG_ON(!eb);
5200 5201

	/*
5202 5203 5204
	 * 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.
5205
	 */
5206
	spin_lock(&eb->refs_lock);
5207
	if (atomic_read(&eb->refs) != 1 || extent_buffer_under_io(eb)) {
5208 5209 5210
		spin_unlock(&eb->refs_lock);
		spin_unlock(&page->mapping->private_lock);
		return 0;
5211
	}
5212
	spin_unlock(&page->mapping->private_lock);
5213

5214
	/*
5215 5216
	 * 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.
5217
	 */
5218 5219 5220
	if (!test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags)) {
		spin_unlock(&eb->refs_lock);
		return 0;
5221
	}
5222

5223
	return release_extent_buffer(eb);
5224
}