extent_io.c 92.3 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
#include <linux/bitops.h>
#include <linux/slab.h>
#include <linux/bio.h>
#include <linux/mm.h>
#include <linux/gfp.h>
#include <linux/pagemap.h>
#include <linux/page-flags.h>
#include <linux/module.h>
#include <linux/spinlock.h>
#include <linux/blkdev.h>
#include <linux/swap.h>
#include <linux/writeback.h>
#include <linux/pagevec.h>
#include "extent_io.h"
#include "extent_map.h"
16
#include "compat.h"
17 18
#include "ctree.h"
#include "btrfs_inode.h"
19 20 21 22 23 24 25 26 27 28 29 30

/* temporary define until extent_map moves out of btrfs */
struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
				       unsigned long extra_flags,
				       void (*ctor)(void *, struct kmem_cache *,
						    unsigned long));

static struct kmem_cache *extent_state_cache;
static struct kmem_cache *extent_buffer_cache;

static LIST_HEAD(buffers);
static LIST_HEAD(states);
C
Chris Mason 已提交
31

32
#define LEAK_DEBUG 0
33
#if LEAK_DEBUG
C
Chris Mason 已提交
34
static DEFINE_SPINLOCK(leak_lock);
C
Chris Mason 已提交
35
#endif
36 37 38 39 40 41 42 43 44 45 46 47 48

#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;
49 50 51 52

	/* tells writepage not to lock the state bits for this range
	 * it still does the unlocking
	 */
53 54 55 56
	unsigned int extent_locked:1;

	/* tells the submit_bio code to use a WRITE_SYNC */
	unsigned int sync_io:1;
57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81
};

int __init extent_io_init(void)
{
	extent_state_cache = btrfs_cache_create("extent_state",
					    sizeof(struct extent_state), 0,
					    NULL);
	if (!extent_state_cache)
		return -ENOMEM;

	extent_buffer_cache = btrfs_cache_create("extent_buffers",
					    sizeof(struct extent_buffer), 0,
					    NULL);
	if (!extent_buffer_cache)
		goto free_state_cache;
	return 0;

free_state_cache:
	kmem_cache_destroy(extent_state_cache);
	return -ENOMEM;
}

void extent_io_exit(void)
{
	struct extent_state *state;
82
	struct extent_buffer *eb;
83 84

	while (!list_empty(&states)) {
85
		state = list_entry(states.next, struct extent_state, leak_list);
C
Chris Mason 已提交
86 87 88 89 90
		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));
91
		list_del(&state->leak_list);
92 93 94 95
		kmem_cache_free(extent_state_cache, state);

	}

96 97
	while (!list_empty(&buffers)) {
		eb = list_entry(buffers.next, struct extent_buffer, leak_list);
C
Chris Mason 已提交
98 99 100
		printk(KERN_ERR "btrfs buffer leak start %llu len %lu "
		       "refs %d\n", (unsigned long long)eb->start,
		       eb->len, atomic_read(&eb->refs));
101 102 103
		list_del(&eb->leak_list);
		kmem_cache_free(extent_buffer_cache, eb);
	}
104 105 106 107 108 109 110 111 112 113
	if (extent_state_cache)
		kmem_cache_destroy(extent_state_cache);
	if (extent_buffer_cache)
		kmem_cache_destroy(extent_buffer_cache);
}

void extent_io_tree_init(struct extent_io_tree *tree,
			  struct address_space *mapping, gfp_t mask)
{
	tree->state.rb_node = NULL;
114
	tree->buffer.rb_node = NULL;
115 116
	tree->ops = NULL;
	tree->dirty_bytes = 0;
117
	spin_lock_init(&tree->lock);
118
	spin_lock_init(&tree->buffer_lock);
119 120 121
	tree->mapping = mapping;
}

122
static struct extent_state *alloc_extent_state(gfp_t mask)
123 124
{
	struct extent_state *state;
125
#if LEAK_DEBUG
126
	unsigned long flags;
C
Chris Mason 已提交
127
#endif
128 129

	state = kmem_cache_alloc(extent_state_cache, mask);
130
	if (!state)
131 132 133
		return state;
	state->state = 0;
	state->private = 0;
134
	state->tree = NULL;
135
#if LEAK_DEBUG
136 137 138
	spin_lock_irqsave(&leak_lock, flags);
	list_add(&state->leak_list, &states);
	spin_unlock_irqrestore(&leak_lock, flags);
C
Chris Mason 已提交
139
#endif
140 141 142 143 144
	atomic_set(&state->refs, 1);
	init_waitqueue_head(&state->wq);
	return state;
}

145
static void free_extent_state(struct extent_state *state)
146 147 148 149
{
	if (!state)
		return;
	if (atomic_dec_and_test(&state->refs)) {
150
#if LEAK_DEBUG
151
		unsigned long flags;
C
Chris Mason 已提交
152
#endif
153
		WARN_ON(state->tree);
154
#if LEAK_DEBUG
155 156 157
		spin_lock_irqsave(&leak_lock, flags);
		list_del(&state->leak_list);
		spin_unlock_irqrestore(&leak_lock, flags);
C
Chris Mason 已提交
158
#endif
159 160 161 162 163 164 165
		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
166 167
	struct rb_node **p = &root->rb_node;
	struct rb_node *parent = NULL;
168 169
	struct tree_entry *entry;

C
Chris Mason 已提交
170
	while (*p) {
171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187
		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;
	}

	entry = rb_entry(node, struct tree_entry, rb_node);
	rb_link_node(node, parent, p);
	rb_insert_color(node, root);
	return NULL;
}

188
static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset,
189 190 191
				     struct rb_node **prev_ret,
				     struct rb_node **next_ret)
{
192
	struct rb_root *root = &tree->state;
C
Chris Mason 已提交
193
	struct rb_node *n = root->rb_node;
194 195 196 197 198
	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
199
	while (n) {
200 201 202 203 204 205 206 207
		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
208
		else
209 210 211 212 213
			return n;
	}

	if (prev_ret) {
		orig_prev = prev;
C
Chris Mason 已提交
214
		while (prev && offset > prev_entry->end) {
215 216 217 218 219 220 221 222 223
			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
224
		while (prev && offset < prev_entry->start) {
225 226 227 228 229 230 231 232
			prev = rb_prev(prev);
			prev_entry = rb_entry(prev, struct tree_entry, rb_node);
		}
		*next_ret = prev;
	}
	return NULL;
}

233 234
static inline struct rb_node *tree_search(struct extent_io_tree *tree,
					  u64 offset)
235
{
236
	struct rb_node *prev = NULL;
237
	struct rb_node *ret;
238

239
	ret = __etree_search(tree, offset, &prev, NULL);
C
Chris Mason 已提交
240
	if (!ret)
241 242 243 244
		return prev;
	return ret;
}

245 246 247 248
static struct extent_buffer *buffer_tree_insert(struct extent_io_tree *tree,
					  u64 offset, struct rb_node *node)
{
	struct rb_root *root = &tree->buffer;
C
Chris Mason 已提交
249 250
	struct rb_node **p = &root->rb_node;
	struct rb_node *parent = NULL;
251 252
	struct extent_buffer *eb;

C
Chris Mason 已提交
253
	while (*p) {
254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273
		parent = *p;
		eb = rb_entry(parent, struct extent_buffer, rb_node);

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

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

static struct extent_buffer *buffer_search(struct extent_io_tree *tree,
					   u64 offset)
{
	struct rb_root *root = &tree->buffer;
C
Chris Mason 已提交
274
	struct rb_node *n = root->rb_node;
275 276
	struct extent_buffer *eb;

C
Chris Mason 已提交
277
	while (n) {
278 279 280 281 282 283 284 285 286 287 288
		eb = rb_entry(n, struct extent_buffer, rb_node);
		if (offset < eb->start)
			n = n->rb_left;
		else if (offset > eb->start)
			n = n->rb_right;
		else
			return eb;
	}
	return NULL;
}

289 290 291 292 293 294 295 296 297 298 299 300 301 302 303
/*
 * 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.
 */
static int merge_state(struct extent_io_tree *tree,
		       struct extent_state *state)
{
	struct extent_state *other;
	struct rb_node *other_node;

304
	if (state->state & (EXTENT_IOBITS | EXTENT_BOUNDARY))
305 306 307 308 309 310 311 312
		return 0;

	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) {
			state->start = other->start;
313
			other->tree = NULL;
314 315 316 317 318 319 320 321 322 323
			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) {
			other->start = state->start;
324
			state->tree = NULL;
325 326 327 328 329 330 331
			rb_erase(&state->rb_node, &tree->state);
			free_extent_state(state);
		}
	}
	return 0;
}

332 333 334 335 336 337
static void set_state_cb(struct extent_io_tree *tree,
			 struct extent_state *state,
			 unsigned long bits)
{
	if (tree->ops && tree->ops->set_bit_hook) {
		tree->ops->set_bit_hook(tree->mapping->host, state->start,
C
Chris Mason 已提交
338
					state->end, state->state, bits);
339 340 341 342 343 344 345
	}
}

static void clear_state_cb(struct extent_io_tree *tree,
			   struct extent_state *state,
			   unsigned long bits)
{
L
Liu Hui 已提交
346
	if (tree->ops && tree->ops->clear_bit_hook) {
347
		tree->ops->clear_bit_hook(tree->mapping->host, state->start,
C
Chris Mason 已提交
348
					  state->end, state->state, bits);
349 350 351
	}
}

352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368
/*
 * 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,
			int bits)
{
	struct rb_node *node;

	if (end < start) {
C
Chris Mason 已提交
369 370 371
		printk(KERN_ERR "btrfs end < start %llu %llu\n",
		       (unsigned long long)end,
		       (unsigned long long)start);
372 373 374 375
		WARN_ON(1);
	}
	if (bits & EXTENT_DIRTY)
		tree->dirty_bytes += end - start + 1;
C
Chris Mason 已提交
376
	set_state_cb(tree, state, bits);
377 378 379 380 381 382 383
	state->state |= bits;
	state->start = start;
	state->end = end;
	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
384 385 386 387
		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);
388 389 390
		free_extent_state(state);
		return -EEXIST;
	}
391
	state->tree = tree;
392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423
	merge_state(tree, state);
	return 0;
}

/*
 * 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;
	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;
	}
424
	prealloc->tree = tree;
425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446
	return 0;
}

/*
 * utility function to clear some bits in an extent state struct.
 * it will optionally wake up any one waiting on this state (wake == 1), or
 * forcibly remove the state from the tree (delete == 1).
 *
 * If no bits are set on the state struct after clearing things, the
 * struct is freed and removed from the tree
 */
static int clear_state_bit(struct extent_io_tree *tree,
			    struct extent_state *state, int bits, int wake,
			    int delete)
{
	int ret = state->state & bits;

	if ((bits & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) {
		u64 range = state->end - state->start + 1;
		WARN_ON(range > tree->dirty_bytes);
		tree->dirty_bytes -= range;
	}
447
	clear_state_cb(tree, state, bits);
C
Chris Mason 已提交
448
	state->state &= ~bits;
449 450 451
	if (wake)
		wake_up(&state->wq);
	if (delete || state->state == 0) {
452
		if (state->tree) {
453
			clear_state_cb(tree, state, state->state);
454
			rb_erase(&state->rb_node, &tree->state);
455
			state->tree = NULL;
456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494
			free_extent_state(state);
		} else {
			WARN_ON(1);
		}
	} else {
		merge_state(tree, state);
	}
	return ret;
}

/*
 * 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.
 *
 * This takes the tree lock, and returns < 0 on error, > 0 if any of the
 * bits were already set, or zero if none of the bits were already set.
 */
int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
		     int bits, int wake, int delete, gfp_t mask)
{
	struct extent_state *state;
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
	int err;
	int set = 0;

again:
	if (!prealloc && (mask & __GFP_WAIT)) {
		prealloc = alloc_extent_state(mask);
		if (!prealloc)
			return -ENOMEM;
	}

495
	spin_lock(&tree->lock);
496 497 498 499
	/*
	 * this search will find the extents that end after
	 * our range starts
	 */
500
	node = tree_search(tree, start);
501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524
	if (!node)
		goto out;
	state = rb_entry(node, struct extent_state, rb_node);
	if (state->start > end)
		goto out;
	WARN_ON(state->end < start);

	/*
	 *     | ---- 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) {
525 526
		if (!prealloc)
			prealloc = alloc_extent_state(GFP_ATOMIC);
527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547
		err = split_state(tree, state, prealloc, start);
		BUG_ON(err == -EEXIST);
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
			start = state->end + 1;
			set |= clear_state_bit(tree, state, bits,
					wake, delete);
		} else {
			start = state->start;
		}
		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) {
548 549
		if (!prealloc)
			prealloc = alloc_extent_state(GFP_ATOMIC);
550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565
		err = split_state(tree, state, prealloc, end + 1);
		BUG_ON(err == -EEXIST);

		if (wake)
			wake_up(&state->wq);
		set |= clear_state_bit(tree, prealloc, bits,
				       wake, delete);
		prealloc = NULL;
		goto out;
	}

	start = state->end + 1;
	set |= clear_state_bit(tree, state, bits, wake, delete);
	goto search_again;

out:
566
	spin_unlock(&tree->lock);
567 568 569 570 571 572 573 574
	if (prealloc)
		free_extent_state(prealloc);

	return set;

search_again:
	if (start > end)
		goto out;
575
	spin_unlock(&tree->lock);
576 577 578 579 580 581 582
	if (mask & __GFP_WAIT)
		cond_resched();
	goto again;
}

static int wait_on_state(struct extent_io_tree *tree,
			 struct extent_state *state)
583 584
		__releases(tree->lock)
		__acquires(tree->lock)
585 586 587
{
	DEFINE_WAIT(wait);
	prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE);
588
	spin_unlock(&tree->lock);
589
	schedule();
590
	spin_lock(&tree->lock);
591 592 593 594 595 596 597 598 599 600 601 602 603 604
	finish_wait(&state->wq, &wait);
	return 0;
}

/*
 * 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
 */
int wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, int bits)
{
	struct extent_state *state;
	struct rb_node *node;

605
	spin_lock(&tree->lock);
606 607 608 609 610 611
again:
	while (1) {
		/*
		 * this search will find all the extents that end after
		 * our range starts
		 */
612
		node = tree_search(tree, start);
613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633
		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;

		if (need_resched()) {
634
			spin_unlock(&tree->lock);
635
			cond_resched();
636
			spin_lock(&tree->lock);
637 638 639
		}
	}
out:
640
	spin_unlock(&tree->lock);
641 642 643 644 645 646 647 648 649 650 651
	return 0;
}

static void set_state_bits(struct extent_io_tree *tree,
			   struct extent_state *state,
			   int bits)
{
	if ((bits & EXTENT_DIRTY) && !(state->state & EXTENT_DIRTY)) {
		u64 range = state->end - state->start + 1;
		tree->dirty_bytes += range;
	}
652
	set_state_cb(tree, state, bits);
C
Chris Mason 已提交
653
	state->state |= bits;
654 655 656 657 658 659 660 661 662 663 664 665 666
}

/*
 * 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.
 *
 * If 'exclusive' == 1, 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.
 *
 * [start, end] is inclusive
 * This takes the tree lock.
 */
C
Chris Mason 已提交
667 668 669
static int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
			  int bits, int exclusive, u64 *failed_start,
			  gfp_t mask)
670 671 672 673 674 675 676 677 678 679 680 681 682 683 684
{
	struct extent_state *state;
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
	int err = 0;
	int set;
	u64 last_start;
	u64 last_end;
again:
	if (!prealloc && (mask & __GFP_WAIT)) {
		prealloc = alloc_extent_state(mask);
		if (!prealloc)
			return -ENOMEM;
	}

685
	spin_lock(&tree->lock);
686 687 688 689
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
690
	node = tree_search(tree, start);
691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769
	if (!node) {
		err = insert_state(tree, prealloc, start, end, bits);
		prealloc = NULL;
		BUG_ON(err == -EEXIST);
		goto out;
	}

	state = rb_entry(node, struct extent_state, rb_node);
	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->state & bits;
		if (set && exclusive) {
			*failed_start = state->start;
			err = -EEXIST;
			goto out;
		}
		set_state_bits(tree, state, bits);
		start = state->end + 1;
		merge_state(tree, state);
		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) {
		set = state->state & bits;
		if (exclusive && set) {
			*failed_start = start;
			err = -EEXIST;
			goto out;
		}
		err = split_state(tree, state, prealloc, start);
		BUG_ON(err == -EEXIST);
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
			set_state_bits(tree, state, bits);
			start = state->end + 1;
			merge_state(tree, state);
		} else {
			start = state->start;
		}
		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
770
			this_end = last_start - 1;
771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804
		err = insert_state(tree, prealloc, start, this_end,
				   bits);
		prealloc = NULL;
		BUG_ON(err == -EEXIST);
		if (err)
			goto out;
		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) {
		set = state->state & bits;
		if (exclusive && set) {
			*failed_start = start;
			err = -EEXIST;
			goto out;
		}
		err = split_state(tree, state, prealloc, end + 1);
		BUG_ON(err == -EEXIST);

		set_state_bits(tree, prealloc, bits);
		merge_state(tree, prealloc);
		prealloc = NULL;
		goto out;
	}

	goto search_again;

out:
805
	spin_unlock(&tree->lock);
806 807 808 809 810 811 812 813
	if (prealloc)
		free_extent_state(prealloc);

	return err;

search_again:
	if (start > end)
		goto out;
814
	spin_unlock(&tree->lock);
815 816 817 818 819 820 821 822 823 824 825 826 827
	if (mask & __GFP_WAIT)
		cond_resched();
	goto again;
}

/* wrappers around set/clear extent bit */
int set_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
		     gfp_t mask)
{
	return set_extent_bit(tree, start, end, EXTENT_DIRTY, 0, NULL,
			      mask);
}

828 829 830 831 832 833
int set_extent_ordered(struct extent_io_tree *tree, u64 start, u64 end,
		       gfp_t mask)
{
	return set_extent_bit(tree, start, end, EXTENT_ORDERED, 0, NULL, mask);
}

834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850
int set_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
		    int bits, gfp_t mask)
{
	return set_extent_bit(tree, start, end, bits, 0, NULL,
			      mask);
}

int clear_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
		      int bits, gfp_t mask)
{
	return clear_extent_bit(tree, start, end, bits, 0, 0, mask);
}

int set_extent_delalloc(struct extent_io_tree *tree, u64 start, u64 end,
		     gfp_t mask)
{
	return set_extent_bit(tree, start, end,
851 852
			      EXTENT_DELALLOC | EXTENT_DIRTY,
			      0, NULL, mask);
853 854 855 856 857 858 859 860 861
}

int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
		       gfp_t mask)
{
	return clear_extent_bit(tree, start, end,
				EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0, mask);
}

862 863 864 865 866 867
int clear_extent_ordered(struct extent_io_tree *tree, u64 start, u64 end,
			 gfp_t mask)
{
	return clear_extent_bit(tree, start, end, EXTENT_ORDERED, 1, 0, mask);
}

868 869 870 871 872 873 874
int set_extent_new(struct extent_io_tree *tree, u64 start, u64 end,
		     gfp_t mask)
{
	return set_extent_bit(tree, start, end, EXTENT_NEW, 0, NULL,
			      mask);
}

875
static int clear_extent_new(struct extent_io_tree *tree, u64 start, u64 end,
876 877 878 879 880 881 882 883 884 885 886 887
		       gfp_t mask)
{
	return clear_extent_bit(tree, start, end, EXTENT_NEW, 0, 0, mask);
}

int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end,
			gfp_t mask)
{
	return set_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, NULL,
			      mask);
}

C
Chris Mason 已提交
888 889
static int clear_extent_uptodate(struct extent_io_tree *tree, u64 start,
				 u64 end, gfp_t mask)
890 891 892 893
{
	return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0, mask);
}

894
static int set_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end,
895 896 897 898 899 900
			 gfp_t mask)
{
	return set_extent_bit(tree, start, end, EXTENT_WRITEBACK,
			      0, NULL, mask);
}

C
Chris Mason 已提交
901 902
static int clear_extent_writeback(struct extent_io_tree *tree, u64 start,
				  u64 end, gfp_t mask)
903 904 905 906 907 908 909 910 911
{
	return clear_extent_bit(tree, start, end, EXTENT_WRITEBACK, 1, 0, mask);
}

int wait_on_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end)
{
	return wait_extent_bit(tree, start, end, EXTENT_WRITEBACK);
}

C
Chris Mason 已提交
912 913 914 915
/*
 * either insert or lock state struct between start and end use mask to tell
 * us if waiting is desired.
 */
916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933
int lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask)
{
	int err;
	u64 failed_start;
	while (1) {
		err = set_extent_bit(tree, start, end, EXTENT_LOCKED, 1,
				     &failed_start, mask);
		if (err == -EEXIST && (mask & __GFP_WAIT)) {
			wait_extent_bit(tree, failed_start, end, EXTENT_LOCKED);
			start = failed_start;
		} else {
			break;
		}
		WARN_ON(start > end);
	}
	return err;
}

934 935 936 937 938 939 940 941
int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end,
		    gfp_t mask)
{
	int err;
	u64 failed_start;

	err = set_extent_bit(tree, start, end, EXTENT_LOCKED, 1,
			     &failed_start, mask);
Y
Yan Zheng 已提交
942 943 944 945
	if (err == -EEXIST) {
		if (failed_start > start)
			clear_extent_bit(tree, start, failed_start - 1,
					 EXTENT_LOCKED, 1, 0, mask);
946
		return 0;
Y
Yan Zheng 已提交
947
	}
948 949 950
	return 1;
}

951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979
int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end,
		  gfp_t mask)
{
	return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, mask);
}

/*
 * helper function to set pages and extents in the tree dirty
 */
int set_range_dirty(struct extent_io_tree *tree, 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(tree->mapping, index);
		BUG_ON(!page);
		__set_page_dirty_nobuffers(page);
		page_cache_release(page);
		index++;
	}
	set_extent_dirty(tree, start, end, GFP_NOFS);
	return 0;
}

/*
 * helper function to set both pages and extents in the tree writeback
 */
980
static int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end)
981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996
{
	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);
		BUG_ON(!page);
		set_page_writeback(page);
		page_cache_release(page);
		index++;
	}
	set_extent_writeback(tree, start, end, GFP_NOFS);
	return 0;
}

C
Chris Mason 已提交
997 998 999 1000 1001 1002 1003
/*
 * 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.
 *
 * If nothing was found, 1 is returned, < 0 on error
 */
1004 1005 1006 1007 1008 1009 1010
int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
			  u64 *start_ret, u64 *end_ret, int bits)
{
	struct rb_node *node;
	struct extent_state *state;
	int ret = 1;

1011
	spin_lock(&tree->lock);
1012 1013 1014 1015
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1016
	node = tree_search(tree, start);
C
Chris Mason 已提交
1017
	if (!node)
1018 1019
		goto out;

C
Chris Mason 已提交
1020
	while (1) {
1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032
		state = rb_entry(node, struct extent_state, rb_node);
		if (state->end >= start && (state->state & bits)) {
			*start_ret = state->start;
			*end_ret = state->end;
			ret = 0;
			break;
		}
		node = rb_next(node);
		if (!node)
			break;
	}
out:
1033
	spin_unlock(&tree->lock);
1034 1035 1036
	return ret;
}

C
Chris Mason 已提交
1037 1038 1039 1040
/* 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'
 */
C
Chris Mason 已提交
1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051
struct extent_state *find_first_extent_bit_state(struct extent_io_tree *tree,
						 u64 start, int bits)
{
	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
1052
	if (!node)
C
Chris Mason 已提交
1053 1054
		goto out;

C
Chris Mason 已提交
1055
	while (1) {
C
Chris Mason 已提交
1056
		state = rb_entry(node, struct extent_state, rb_node);
C
Chris Mason 已提交
1057
		if (state->end >= start && (state->state & bits))
C
Chris Mason 已提交
1058
			return state;
C
Chris Mason 已提交
1059

C
Chris Mason 已提交
1060 1061 1062 1063 1064 1065 1066 1067
		node = rb_next(node);
		if (!node)
			break;
	}
out:
	return NULL;
}

C
Chris Mason 已提交
1068 1069 1070 1071 1072 1073
/*
 * 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
1074 1075
static noinline u64 find_delalloc_range(struct extent_io_tree *tree,
					u64 *start, u64 *end, u64 max_bytes)
1076 1077 1078 1079 1080 1081 1082
{
	struct rb_node *node;
	struct extent_state *state;
	u64 cur_start = *start;
	u64 found = 0;
	u64 total_bytes = 0;

1083
	spin_lock(&tree->lock);
C
Chris Mason 已提交
1084

1085 1086 1087 1088
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1089
	node = tree_search(tree, cur_start);
1090
	if (!node) {
1091 1092
		if (!found)
			*end = (u64)-1;
1093 1094 1095
		goto out;
	}

C
Chris Mason 已提交
1096
	while (1) {
1097
		state = rb_entry(node, struct extent_state, rb_node);
1098 1099
		if (found && (state->start != cur_start ||
			      (state->state & EXTENT_BOUNDARY))) {
1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119
			goto out;
		}
		if (!(state->state & EXTENT_DELALLOC)) {
			if (!found)
				*end = state->end;
			goto out;
		}
		if (!found)
			*start = state->start;
		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:
1120
	spin_unlock(&tree->lock);
1121 1122 1123
	return found;
}

C
Chris Mason 已提交
1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137
static noinline int __unlock_for_delalloc(struct inode *inode,
					  struct page *locked_page,
					  u64 start, u64 end)
{
	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)
		return 0;

C
Chris Mason 已提交
1138
	while (nr_pages > 0) {
C
Chris Mason 已提交
1139
		ret = find_get_pages_contig(inode->i_mapping, index,
1140 1141
				     min_t(unsigned long, nr_pages,
				     ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173
		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();
	}
	return 0;
}

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
1174
	while (nrpages > 0) {
C
Chris Mason 已提交
1175
		ret = find_get_pages_contig(inode->i_mapping, index,
1176 1177
				     min_t(unsigned long,
				     nrpages, ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1178 1179 1180 1181 1182 1183 1184 1185 1186 1187
		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
			 */
1188
			if (pages[i] != locked_page) {
C
Chris Mason 已提交
1189
				lock_page(pages[i]);
1190 1191
				if (!PageDirty(pages[i]) ||
				    pages[i]->mapping != inode->i_mapping) {
1192 1193 1194 1195 1196 1197
					ret = -EAGAIN;
					unlock_page(pages[i]);
					page_cache_release(pages[i]);
					goto done;
				}
			}
C
Chris Mason 已提交
1198
			page_cache_release(pages[i]);
1199
			pages_locked++;
C
Chris Mason 已提交
1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239
		}
		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;
	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,
				    max_bytes);
C
Chris Mason 已提交
1240
	if (!found || delalloc_end <= *start) {
C
Chris Mason 已提交
1241 1242 1243 1244 1245
		*start = delalloc_start;
		*end = delalloc_end;
		return found;
	}

C
Chris Mason 已提交
1246 1247 1248 1249 1250
	/*
	 * 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
1251
	if (delalloc_start < *start)
C
Chris Mason 已提交
1252 1253
		delalloc_start = *start;

C
Chris Mason 已提交
1254 1255 1256 1257
	/*
	 * make sure to limit the number of pages we try to lock down
	 * if we're looping.
	 */
C
Chris Mason 已提交
1258
	if (delalloc_end + 1 - delalloc_start > max_bytes && loops)
1259
		delalloc_end = delalloc_start + PAGE_CACHE_SIZE - 1;
C
Chris Mason 已提交
1260

C
Chris Mason 已提交
1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301
	/* 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
		 */
		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;
		}
	}
	BUG_ON(ret);

	/* step three, lock the state bits for the whole range */
	lock_extent(tree, delalloc_start, delalloc_end, GFP_NOFS);

	/* then test to make sure it is all still delalloc */
	ret = test_range_bit(tree, delalloc_start, delalloc_end,
			     EXTENT_DELALLOC, 1);
	if (!ret) {
		unlock_extent(tree, delalloc_start, delalloc_end, GFP_NOFS);
		__unlock_for_delalloc(inode, locked_page,
			      delalloc_start, delalloc_end);
		cond_resched();
		goto again;
	}
	*start = delalloc_start;
	*end = delalloc_end;
out_failed:
	return found;
}

int extent_clear_unlock_delalloc(struct inode *inode,
				struct extent_io_tree *tree,
				u64 start, u64 end, struct page *locked_page,
1302 1303 1304 1305
				int unlock_pages,
				int clear_unlock,
				int clear_delalloc, int clear_dirty,
				int set_writeback,
C
Chris Mason 已提交
1306 1307 1308 1309 1310 1311 1312 1313
				int end_writeback)
{
	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;
1314
	int clear_bits = 0;
C
Chris Mason 已提交
1315

1316 1317
	if (clear_unlock)
		clear_bits |= EXTENT_LOCKED;
C
Chris Mason 已提交
1318 1319 1320
	if (clear_dirty)
		clear_bits |= EXTENT_DIRTY;

1321 1322 1323
	if (clear_delalloc)
		clear_bits |= EXTENT_DELALLOC;

C
Chris Mason 已提交
1324
	clear_extent_bit(tree, start, end, clear_bits, 1, 0, GFP_NOFS);
1325 1326
	if (!(unlock_pages || clear_dirty || set_writeback || end_writeback))
		return 0;
C
Chris Mason 已提交
1327

C
Chris Mason 已提交
1328
	while (nr_pages > 0) {
C
Chris Mason 已提交
1329
		ret = find_get_pages_contig(inode->i_mapping, index,
1330 1331
				     min_t(unsigned long,
				     nr_pages, ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342
		for (i = 0; i < ret; i++) {
			if (pages[i] == locked_page) {
				page_cache_release(pages[i]);
				continue;
			}
			if (clear_dirty)
				clear_page_dirty_for_io(pages[i]);
			if (set_writeback)
				set_page_writeback(pages[i]);
			if (end_writeback)
				end_page_writeback(pages[i]);
1343 1344
			if (unlock_pages)
				unlock_page(pages[i]);
C
Chris Mason 已提交
1345 1346 1347 1348 1349 1350 1351 1352 1353
			page_cache_release(pages[i]);
		}
		nr_pages -= ret;
		index += ret;
		cond_resched();
	}
	return 0;
}

C
Chris Mason 已提交
1354 1355 1356 1357 1358
/*
 * 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.
 */
1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373
u64 count_range_bits(struct extent_io_tree *tree,
		     u64 *start, u64 search_end, u64 max_bytes,
		     unsigned long bits)
{
	struct rb_node *node;
	struct extent_state *state;
	u64 cur_start = *start;
	u64 total_bytes = 0;
	int found = 0;

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

1374
	spin_lock(&tree->lock);
1375 1376 1377 1378 1379 1380 1381 1382
	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.
	 */
1383
	node = tree_search(tree, cur_start);
C
Chris Mason 已提交
1384
	if (!node)
1385 1386
		goto out;

C
Chris Mason 已提交
1387
	while (1) {
1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405
		state = rb_entry(node, struct extent_state, rb_node);
		if (state->start > search_end)
			break;
		if (state->end >= cur_start && (state->state & bits)) {
			total_bytes += min(search_end, state->end) + 1 -
				       max(cur_start, state->start);
			if (total_bytes >= max_bytes)
				break;
			if (!found) {
				*start = state->start;
				found = 1;
			}
		}
		node = rb_next(node);
		if (!node)
			break;
	}
out:
1406
	spin_unlock(&tree->lock);
1407 1408
	return total_bytes;
}
1409 1410

#if 0
1411 1412 1413 1414
/*
 * helper function to lock both pages and extents in the tree.
 * pages must be locked first.
 */
1415
static int lock_range(struct extent_io_tree *tree, u64 start, u64 end)
1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455
{
	unsigned long index = start >> PAGE_CACHE_SHIFT;
	unsigned long end_index = end >> PAGE_CACHE_SHIFT;
	struct page *page;
	int err;

	while (index <= end_index) {
		page = grab_cache_page(tree->mapping, index);
		if (!page) {
			err = -ENOMEM;
			goto failed;
		}
		if (IS_ERR(page)) {
			err = PTR_ERR(page);
			goto failed;
		}
		index++;
	}
	lock_extent(tree, start, end, GFP_NOFS);
	return 0;

failed:
	/*
	 * we failed above in getting the page at 'index', so we undo here
	 * up to but not including the page at 'index'
	 */
	end_index = index;
	index = start >> PAGE_CACHE_SHIFT;
	while (index < end_index) {
		page = find_get_page(tree->mapping, index);
		unlock_page(page);
		page_cache_release(page);
		index++;
	}
	return err;
}

/*
 * helper function to unlock both pages and extents in the tree.
 */
1456
static int unlock_range(struct extent_io_tree *tree, u64 start, u64 end)
1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470
{
	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);
		unlock_page(page);
		page_cache_release(page);
		index++;
	}
	unlock_extent(tree, start, end, GFP_NOFS);
	return 0;
}
1471
#endif
1472

C
Chris Mason 已提交
1473 1474 1475 1476
/*
 * set the private field for a given byte offset in the tree.  If there isn't
 * an extent_state there already, this does nothing.
 */
1477 1478 1479 1480 1481 1482
int set_state_private(struct extent_io_tree *tree, u64 start, u64 private)
{
	struct rb_node *node;
	struct extent_state *state;
	int ret = 0;

1483
	spin_lock(&tree->lock);
1484 1485 1486 1487
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1488
	node = tree_search(tree, start);
1489
	if (!node) {
1490 1491 1492 1493 1494 1495 1496 1497 1498 1499
		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:
1500
	spin_unlock(&tree->lock);
1501 1502 1503 1504 1505 1506 1507 1508 1509
	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;

1510
	spin_lock(&tree->lock);
1511 1512 1513 1514
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1515
	node = tree_search(tree, start);
1516
	if (!node) {
1517 1518 1519 1520 1521 1522 1523 1524 1525 1526
		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:
1527
	spin_unlock(&tree->lock);
1528 1529 1530 1531 1532
	return ret;
}

/*
 * searches a range in the state tree for a given mask.
1533
 * If 'filled' == 1, this returns 1 only if every extent in the tree
1534 1535 1536 1537 1538 1539 1540 1541 1542 1543
 * 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,
		   int bits, int filled)
{
	struct extent_state *state = NULL;
	struct rb_node *node;
	int bitset = 0;

1544
	spin_lock(&tree->lock);
1545
	node = tree_search(tree, start);
1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574
	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;
		}
		start = state->end + 1;
		if (start > end)
			break;
		node = rb_next(node);
		if (!node) {
			if (filled)
				bitset = 0;
			break;
		}
	}
1575
	spin_unlock(&tree->lock);
1576 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 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633
	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
 */
static int check_page_uptodate(struct extent_io_tree *tree,
			       struct page *page)
{
	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
	u64 end = start + PAGE_CACHE_SIZE - 1;
	if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1))
		SetPageUptodate(page);
	return 0;
}

/*
 * helper function to unlock a page if all the extents in the tree
 * for that page are unlocked
 */
static int check_page_locked(struct extent_io_tree *tree,
			     struct page *page)
{
	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
	u64 end = start + PAGE_CACHE_SIZE - 1;
	if (!test_range_bit(tree, start, end, EXTENT_LOCKED, 0))
		unlock_page(page);
	return 0;
}

/*
 * helper function to end page writeback if all the extents
 * in the tree for that page are done with writeback
 */
static int check_page_writeback(struct extent_io_tree *tree,
			     struct page *page)
{
	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
	u64 end = start + PAGE_CACHE_SIZE - 1;
	if (!test_range_bit(tree, start, end, EXTENT_WRITEBACK, 0))
		end_page_writeback(page);
	return 0;
}

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

/*
 * 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)
{
1634
	int uptodate = err == 0;
1635
	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
1636
	struct extent_io_tree *tree;
1637 1638 1639
	u64 start;
	u64 end;
	int whole_page;
1640
	int ret;
1641 1642 1643

	do {
		struct page *page = bvec->bv_page;
1644 1645
		tree = &BTRFS_I(page->mapping->host)->io_tree;

1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656
		start = ((u64)page->index << PAGE_CACHE_SHIFT) +
			 bvec->bv_offset;
		end = start + bvec->bv_len - 1;

		if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE)
			whole_page = 1;
		else
			whole_page = 0;

		if (--bvec >= bio->bi_io_vec)
			prefetchw(&bvec->bv_page->flags);
1657 1658
		if (tree->ops && tree->ops->writepage_end_io_hook) {
			ret = tree->ops->writepage_end_io_hook(page, start,
1659
						       end, NULL, uptodate);
1660 1661 1662 1663 1664 1665 1666
			if (ret)
				uptodate = 0;
		}

		if (!uptodate && tree->ops &&
		    tree->ops->writepage_io_failed_hook) {
			ret = tree->ops->writepage_io_failed_hook(bio, page,
1667
							 start, end, NULL);
1668 1669 1670 1671 1672 1673
			if (ret == 0) {
				uptodate = (err == 0);
				continue;
			}
		}

1674 1675 1676 1677 1678
		if (!uptodate) {
			clear_extent_uptodate(tree, start, end, GFP_ATOMIC);
			ClearPageUptodate(page);
			SetPageError(page);
		}
1679

1680
		clear_extent_writeback(tree, start, end, GFP_ATOMIC);
1681 1682 1683 1684 1685 1686

		if (whole_page)
			end_page_writeback(page);
		else
			check_page_writeback(tree, page);
	} while (bvec >= bio->bi_io_vec);
1687

1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705
	bio_put(bio);
}

/*
 * 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);
	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
1706
	struct extent_io_tree *tree;
1707 1708 1709 1710 1711
	u64 start;
	u64 end;
	int whole_page;
	int ret;

1712 1713 1714
	if (err)
		uptodate = 0;

1715 1716
	do {
		struct page *page = bvec->bv_page;
1717 1718
		tree = &BTRFS_I(page->mapping->host)->io_tree;

1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731
		start = ((u64)page->index << PAGE_CACHE_SHIFT) +
			bvec->bv_offset;
		end = start + bvec->bv_len - 1;

		if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE)
			whole_page = 1;
		else
			whole_page = 0;

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

		if (uptodate && tree->ops && tree->ops->readpage_end_io_hook) {
1732
			ret = tree->ops->readpage_end_io_hook(page, start, end,
1733
							      NULL);
1734 1735 1736
			if (ret)
				uptodate = 0;
		}
1737 1738 1739
		if (!uptodate && tree->ops &&
		    tree->ops->readpage_io_failed_hook) {
			ret = tree->ops->readpage_io_failed_hook(bio, page,
1740
							 start, end, NULL);
1741
			if (ret == 0) {
1742 1743
				uptodate =
					test_bit(BIO_UPTODATE, &bio->bi_flags);
1744 1745
				if (err)
					uptodate = 0;
1746 1747 1748
				continue;
			}
		}
1749

1750
		if (uptodate) {
1751 1752
			set_extent_uptodate(tree, start, end,
					    GFP_ATOMIC);
1753
		}
1754
		unlock_extent(tree, start, end, GFP_ATOMIC);
1755

1756 1757 1758 1759 1760 1761 1762
		if (whole_page) {
			if (uptodate) {
				SetPageUptodate(page);
			} else {
				ClearPageUptodate(page);
				SetPageError(page);
			}
1763
			unlock_page(page);
1764 1765 1766 1767 1768 1769 1770
		} else {
			if (uptodate) {
				check_page_uptodate(tree, page);
			} else {
				ClearPageUptodate(page);
				SetPageError(page);
			}
1771
			check_page_locked(tree, page);
1772
		}
1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786
	} while (bvec >= bio->bi_io_vec);

	bio_put(bio);
}

/*
 * IO done from prepare_write is pretty simple, we just unlock
 * the structs in the extent tree when done, and set the uptodate bits
 * as appropriate.
 */
static void end_bio_extent_preparewrite(struct bio *bio, int err)
{
	const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
1787
	struct extent_io_tree *tree;
1788 1789 1790 1791 1792
	u64 start;
	u64 end;

	do {
		struct page *page = bvec->bv_page;
1793 1794
		tree = &BTRFS_I(page->mapping->host)->io_tree;

1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829
		start = ((u64)page->index << PAGE_CACHE_SHIFT) +
			bvec->bv_offset;
		end = start + bvec->bv_len - 1;

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

		if (uptodate) {
			set_extent_uptodate(tree, start, end, GFP_ATOMIC);
		} else {
			ClearPageUptodate(page);
			SetPageError(page);
		}

		unlock_extent(tree, start, end, GFP_ATOMIC);

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

	bio_put(bio);
}

static struct bio *
extent_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
		 gfp_t gfp_flags)
{
	struct bio *bio;

	bio = bio_alloc(gfp_flags, nr_vecs);

	if (bio == NULL && (current->flags & PF_MEMALLOC)) {
		while (!bio && (nr_vecs /= 2))
			bio = bio_alloc(gfp_flags, nr_vecs);
	}

	if (bio) {
1830
		bio->bi_size = 0;
1831 1832 1833 1834 1835 1836
		bio->bi_bdev = bdev;
		bio->bi_sector = first_sector;
	}
	return bio;
}

C
Chris Mason 已提交
1837 1838
static int submit_one_bio(int rw, struct bio *bio, int mirror_num,
			  unsigned long bio_flags)
1839 1840
{
	int ret = 0;
1841 1842 1843 1844 1845 1846 1847 1848 1849
	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;
	u64 end;

	start = ((u64)page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset;
	end = start + bvec->bv_len - 1;

1850
	bio->bi_private = NULL;
1851 1852 1853

	bio_get(bio);

1854
	if (tree->ops && tree->ops->submit_bio_hook)
1855
		tree->ops->submit_bio_hook(page->mapping->host, rw, bio,
C
Chris Mason 已提交
1856
					   mirror_num, bio_flags);
1857 1858
	else
		submit_bio(rw, bio);
1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870
	if (bio_flagged(bio, BIO_EOPNOTSUPP))
		ret = -EOPNOTSUPP;
	bio_put(bio);
	return ret;
}

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,
1871
			      bio_end_io_t end_io_func,
C
Chris Mason 已提交
1872 1873 1874
			      int mirror_num,
			      unsigned long prev_bio_flags,
			      unsigned long bio_flags)
1875 1876 1877 1878
{
	int ret = 0;
	struct bio *bio;
	int nr;
C
Chris Mason 已提交
1879 1880 1881
	int contig = 0;
	int this_compressed = bio_flags & EXTENT_BIO_COMPRESSED;
	int old_compressed = prev_bio_flags & EXTENT_BIO_COMPRESSED;
1882
	size_t page_size = min_t(size_t, size, PAGE_CACHE_SIZE);
1883 1884 1885

	if (bio_ret && *bio_ret) {
		bio = *bio_ret;
C
Chris Mason 已提交
1886 1887 1888 1889 1890 1891 1892
		if (old_compressed)
			contig = bio->bi_sector == sector;
		else
			contig = bio->bi_sector + (bio->bi_size >> 9) ==
				sector;

		if (prev_bio_flags != bio_flags || !contig ||
1893
		    (tree->ops && tree->ops->merge_bio_hook &&
C
Chris Mason 已提交
1894 1895 1896 1897 1898
		     tree->ops->merge_bio_hook(page, offset, page_size, bio,
					       bio_flags)) ||
		    bio_add_page(bio, page, page_size, offset) < page_size) {
			ret = submit_one_bio(rw, bio, mirror_num,
					     prev_bio_flags);
1899 1900 1901 1902 1903
			bio = NULL;
		} else {
			return 0;
		}
	}
C
Chris Mason 已提交
1904 1905 1906 1907 1908
	if (this_compressed)
		nr = BIO_MAX_PAGES;
	else
		nr = bio_get_nr_vecs(bdev);

1909
	bio = extent_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH);
1910

C
Chris Mason 已提交
1911
	bio_add_page(bio, page, page_size, offset);
1912 1913
	bio->bi_end_io = end_io_func;
	bio->bi_private = tree;
1914

C
Chris Mason 已提交
1915
	if (bio_ret)
1916
		*bio_ret = bio;
C
Chris Mason 已提交
1917
	else
C
Chris Mason 已提交
1918
		ret = submit_one_bio(rw, bio, mirror_num, bio_flags);
1919 1920 1921 1922 1923 1924 1925 1926 1927

	return ret;
}

void set_page_extent_mapped(struct page *page)
{
	if (!PagePrivate(page)) {
		SetPagePrivate(page);
		page_cache_get(page);
1928
		set_page_private(page, EXTENT_PAGE_PRIVATE);
1929 1930 1931
	}
}

1932
static void set_page_extent_head(struct page *page, unsigned long len)
1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944
{
	set_page_private(page, EXTENT_PAGE_PRIVATE_FIRST_PAGE | len << 2);
}

/*
 * 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)
 */
static int __extent_read_full_page(struct extent_io_tree *tree,
				   struct page *page,
				   get_extent_t *get_extent,
C
Chris Mason 已提交
1945 1946
				   struct bio **bio, int mirror_num,
				   unsigned long *bio_flags)
1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963
{
	struct inode *inode = page->mapping->host;
	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
	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;
	size_t page_offset = 0;
	size_t iosize;
C
Chris Mason 已提交
1964
	size_t disk_io_size;
1965
	size_t blocksize = inode->i_sb->s_blocksize;
C
Chris Mason 已提交
1966
	unsigned long this_bio_flag = 0;
1967 1968 1969 1970 1971 1972

	set_page_extent_mapped(page);

	end = page_end;
	lock_extent(tree, start, end, GFP_NOFS);

C
Chris Mason 已提交
1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984
	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;
			userpage = kmap_atomic(page, KM_USER0);
			memset(userpage + zero_offset, 0, iosize);
			flush_dcache_page(page);
			kunmap_atomic(userpage, KM_USER0);
		}
	}
1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
	while (cur <= end) {
		if (cur >= last_byte) {
			char *userpage;
			iosize = PAGE_CACHE_SIZE - page_offset;
			userpage = kmap_atomic(page, KM_USER0);
			memset(userpage + page_offset, 0, iosize);
			flush_dcache_page(page);
			kunmap_atomic(userpage, KM_USER0);
			set_extent_uptodate(tree, cur, cur + iosize - 1,
					    GFP_NOFS);
			unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
			break;
		}
		em = get_extent(inode, page, page_offset, cur,
				end - cur + 1, 0);
		if (IS_ERR(em) || !em) {
			SetPageError(page);
			unlock_extent(tree, cur, end, GFP_NOFS);
			break;
		}
		extent_offset = cur - em->start;
		BUG_ON(extent_map_end(em) <= cur);
		BUG_ON(end < cur);

C
Chris Mason 已提交
2009 2010 2011
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
			this_bio_flag = EXTENT_BIO_COMPRESSED;

2012 2013 2014
		iosize = min(extent_map_end(em) - cur, end - cur + 1);
		cur_end = min(extent_map_end(em) - 1, end);
		iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1);
C
Chris Mason 已提交
2015 2016 2017 2018 2019 2020 2021
		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;
		}
2022 2023
		bdev = em->bdev;
		block_start = em->block_start;
Y
Yan Zheng 已提交
2024 2025
		if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			block_start = EXTENT_MAP_HOLE;
2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045
		free_extent_map(em);
		em = NULL;

		/* we've found a hole, just zero and go on */
		if (block_start == EXTENT_MAP_HOLE) {
			char *userpage;
			userpage = kmap_atomic(page, KM_USER0);
			memset(userpage + page_offset, 0, iosize);
			flush_dcache_page(page);
			kunmap_atomic(userpage, KM_USER0);

			set_extent_uptodate(tree, cur, cur + iosize - 1,
					    GFP_NOFS);
			unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
			cur = cur + iosize;
			page_offset += iosize;
			continue;
		}
		/* the get_extent function already copied into the page */
		if (test_range_bit(tree, cur, cur_end, EXTENT_UPTODATE, 1)) {
2046
			check_page_uptodate(tree, page);
2047 2048 2049 2050 2051
			unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
			cur = cur + iosize;
			page_offset += iosize;
			continue;
		}
2052 2053 2054 2055 2056 2057 2058 2059 2060 2061
		/* we have an inline extent but it didn't get marked up
		 * to date.  Error out
		 */
		if (block_start == EXTENT_MAP_INLINE) {
			SetPageError(page);
			unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
			cur = cur + iosize;
			page_offset += iosize;
			continue;
		}
2062 2063 2064 2065 2066 2067 2068

		ret = 0;
		if (tree->ops && tree->ops->readpage_io_hook) {
			ret = tree->ops->readpage_io_hook(page, cur,
							  cur + iosize - 1);
		}
		if (!ret) {
2069 2070
			unsigned long pnr = (last_byte >> PAGE_CACHE_SHIFT) + 1;
			pnr -= page->index;
2071
			ret = submit_extent_page(READ, tree, page,
C
Chris Mason 已提交
2072
					 sector, disk_io_size, page_offset,
2073
					 bdev, bio, pnr,
C
Chris Mason 已提交
2074 2075 2076
					 end_bio_extent_readpage, mirror_num,
					 *bio_flags,
					 this_bio_flag);
2077
			nr++;
C
Chris Mason 已提交
2078
			*bio_flags = this_bio_flag;
2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096
		}
		if (ret)
			SetPageError(page);
		cur = cur + iosize;
		page_offset += iosize;
	}
	if (!nr) {
		if (!PageError(page))
			SetPageUptodate(page);
		unlock_page(page);
	}
	return 0;
}

int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
			    get_extent_t *get_extent)
{
	struct bio *bio = NULL;
C
Chris Mason 已提交
2097
	unsigned long bio_flags = 0;
2098 2099
	int ret;

C
Chris Mason 已提交
2100 2101
	ret = __extent_read_full_page(tree, page, get_extent, &bio, 0,
				      &bio_flags);
2102
	if (bio)
C
Chris Mason 已提交
2103
		submit_one_bio(READ, bio, 0, bio_flags);
2104 2105 2106
	return ret;
}

2107 2108 2109 2110 2111 2112 2113 2114 2115 2116
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;
}

2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137
/*
 * 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;
	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
	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;
2138
	u64 unlock_start;
2139 2140 2141 2142 2143
	sector_t sector;
	struct extent_map *em;
	struct block_device *bdev;
	int ret;
	int nr = 0;
2144
	size_t pg_offset = 0;
2145 2146 2147 2148 2149
	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
2150 2151
	int page_started;
	int compressed;
2152
	int write_flags;
2153
	unsigned long nr_written = 0;
2154

2155 2156 2157 2158 2159
	if (wbc->sync_mode == WB_SYNC_ALL)
		write_flags = WRITE_SYNC_PLUG;
	else
		write_flags = WRITE;

2160
	WARN_ON(!PageLocked(page));
2161
	pg_offset = i_size & (PAGE_CACHE_SIZE - 1);
2162
	if (page->index > end_index ||
2163
	   (page->index == end_index && !pg_offset)) {
2164
		page->mapping->a_ops->invalidatepage(page, 0);
2165 2166 2167 2168 2169 2170 2171 2172
		unlock_page(page);
		return 0;
	}

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

		userpage = kmap_atomic(page, KM_USER0);
2173 2174
		memset(userpage + pg_offset, 0,
		       PAGE_CACHE_SIZE - pg_offset);
2175
		kunmap_atomic(userpage, KM_USER0);
2176
		flush_dcache_page(page);
2177
	}
2178
	pg_offset = 0;
2179 2180 2181 2182 2183

	set_page_extent_mapped(page);

	delalloc_start = start;
	delalloc_end = 0;
C
Chris Mason 已提交
2184
	page_started = 0;
2185
	if (!epd->extent_locked) {
2186 2187 2188 2189 2190 2191
		/*
		 * 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
2192
		while (delalloc_end < page_end) {
2193
			nr_delalloc = find_lock_delalloc_range(inode, tree,
C
Chris Mason 已提交
2194 2195
						       page,
						       &delalloc_start,
2196 2197
						       &delalloc_end,
						       128 * 1024 * 1024);
2198 2199 2200 2201 2202 2203 2204
			if (nr_delalloc == 0) {
				delalloc_start = delalloc_end + 1;
				continue;
			}
			tree->ops->fill_delalloc(inode, page, delalloc_start,
						 delalloc_end, &page_started,
						 &nr_written);
2205 2206
			delalloc_start = delalloc_end + 1;
		}
C
Chris Mason 已提交
2207

2208 2209 2210 2211 2212
		/* did the fill delalloc function already unlock and start
		 * the IO?
		 */
		if (page_started) {
			ret = 0;
2213 2214 2215 2216 2217 2218 2219
			/*
			 * 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;
2220
		}
C
Chris Mason 已提交
2221
	}
2222
	lock_extent(tree, start, page_end, GFP_NOFS);
2223

2224
	unlock_start = start;
2225

2226
	if (tree->ops && tree->ops->writepage_start_hook) {
C
Chris Mason 已提交
2227 2228
		ret = tree->ops->writepage_start_hook(page, start,
						      page_end);
2229 2230 2231
		if (ret == -EAGAIN) {
			unlock_extent(tree, start, page_end, GFP_NOFS);
			redirty_page_for_writepage(wbc, page);
2232
			update_nr_written(page, wbc, nr_written);
2233
			unlock_page(page);
2234
			ret = 0;
2235
			goto done_unlocked;
2236 2237 2238
		}
	}

2239 2240 2241 2242 2243
	/*
	 * 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);
2244

2245
	end = page_end;
C
Chris Mason 已提交
2246 2247
	if (test_range_bit(tree, start, page_end, EXTENT_DELALLOC, 0))
		printk(KERN_ERR "btrfs delalloc bits after lock_extent\n");
2248 2249 2250

	if (last_byte <= start) {
		clear_extent_dirty(tree, start, page_end, GFP_NOFS);
2251 2252 2253 2254 2255
		unlock_extent(tree, start, page_end, GFP_NOFS);
		if (tree->ops && tree->ops->writepage_end_io_hook)
			tree->ops->writepage_end_io_hook(page, start,
							 page_end, NULL, 1);
		unlock_start = page_end + 1;
2256 2257 2258 2259 2260 2261 2262 2263 2264
		goto done;
	}

	set_extent_uptodate(tree, start, page_end, GFP_NOFS);
	blocksize = inode->i_sb->s_blocksize;

	while (cur <= end) {
		if (cur >= last_byte) {
			clear_extent_dirty(tree, cur, page_end, GFP_NOFS);
2265 2266 2267 2268 2269
			unlock_extent(tree, unlock_start, page_end, GFP_NOFS);
			if (tree->ops && tree->ops->writepage_end_io_hook)
				tree->ops->writepage_end_io_hook(page, cur,
							 page_end, NULL, 1);
			unlock_start = page_end + 1;
2270 2271
			break;
		}
2272
		em = epd->get_extent(inode, page, pg_offset, cur,
2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286
				     end - cur + 1, 1);
		if (IS_ERR(em) || !em) {
			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);
		iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1);
		sector = (em->block_start + extent_offset) >> 9;
		bdev = em->bdev;
		block_start = em->block_start;
C
Chris Mason 已提交
2287
		compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
2288 2289 2290
		free_extent_map(em);
		em = NULL;

C
Chris Mason 已提交
2291 2292 2293 2294 2295
		/*
		 * compressed and inline extents are written through other
		 * paths in the FS
		 */
		if (compressed || block_start == EXTENT_MAP_HOLE ||
2296 2297 2298
		    block_start == EXTENT_MAP_INLINE) {
			clear_extent_dirty(tree, cur,
					   cur + iosize - 1, GFP_NOFS);
2299

C
Chris Mason 已提交
2300
			unlock_extent(tree, unlock_start, cur + iosize - 1,
2301
				      GFP_NOFS);
2302

C
Chris Mason 已提交
2303 2304 2305 2306 2307 2308
			/*
			 * end_io notification does not happen here for
			 * compressed extents
			 */
			if (!compressed && tree->ops &&
			    tree->ops->writepage_end_io_hook)
2309 2310 2311
				tree->ops->writepage_end_io_hook(page, cur,
							 cur + iosize - 1,
							 NULL, 1);
C
Chris Mason 已提交
2312 2313 2314 2315 2316 2317 2318 2319 2320
			else if (compressed) {
				/* we don't want to end_page_writeback on
				 * a compressed extent.  this happens
				 * elsewhere
				 */
				nr++;
			}

			cur += iosize;
2321
			pg_offset += iosize;
2322
			unlock_start = cur;
2323 2324 2325 2326 2327 2328
			continue;
		}
		/* leave this out until we have a page_mkwrite call */
		if (0 && !test_range_bit(tree, cur, cur + iosize - 1,
				   EXTENT_DIRTY, 0)) {
			cur = cur + iosize;
2329
			pg_offset += iosize;
2330 2331
			continue;
		}
C
Chris Mason 已提交
2332

2333 2334 2335 2336 2337 2338 2339
		clear_extent_dirty(tree, cur, cur + iosize - 1, GFP_NOFS);
		if (tree->ops && tree->ops->writepage_io_hook) {
			ret = tree->ops->writepage_io_hook(page, cur,
						cur + iosize - 1);
		} else {
			ret = 0;
		}
2340
		if (ret) {
2341
			SetPageError(page);
2342
		} else {
2343
			unsigned long max_nr = end_index + 1;
2344

2345 2346
			set_range_writeback(tree, cur, cur + iosize - 1);
			if (!PageWriteback(page)) {
C
Chris Mason 已提交
2347 2348 2349
				printk(KERN_ERR "btrfs warning page %lu not "
				       "writeback, cur %llu end %llu\n",
				       page->index, (unsigned long long)cur,
2350 2351 2352
				       (unsigned long long)end);
			}

2353 2354 2355
			ret = submit_extent_page(write_flags, tree, page,
						 sector, iosize, pg_offset,
						 bdev, &epd->bio, max_nr,
C
Chris Mason 已提交
2356 2357
						 end_bio_extent_writepage,
						 0, 0, 0);
2358 2359 2360 2361
			if (ret)
				SetPageError(page);
		}
		cur = cur + iosize;
2362
		pg_offset += iosize;
2363 2364 2365 2366 2367 2368 2369 2370
		nr++;
	}
done:
	if (nr == 0) {
		/* make sure the mapping tag for page dirty gets cleared */
		set_page_writeback(page);
		end_page_writeback(page);
	}
2371 2372
	if (unlock_start <= page_end)
		unlock_extent(tree, unlock_start, page_end, GFP_NOFS);
2373
	unlock_page(page);
2374

2375 2376
done_unlocked:

2377 2378 2379 2380
	return 0;
}

/**
C
Chris Mason 已提交
2381
 * write_cache_pages - walk the list of dirty pages of the given address space and write all of them.
2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394
 * @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.
 */
2395
static int extent_write_cache_pages(struct extent_io_tree *tree,
C
Chris Mason 已提交
2396 2397
			     struct address_space *mapping,
			     struct writeback_control *wbc,
C
Chris Mason 已提交
2398 2399
			     writepage_t writepage, void *data,
			     void (*flush_fn)(void *))
2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424
{
	struct backing_dev_info *bdi = mapping->backing_dev_info;
	int ret = 0;
	int done = 0;
	struct pagevec pvec;
	int nr_pages;
	pgoff_t index;
	pgoff_t end;		/* Inclusive */
	int scanned = 0;
	int range_whole = 0;

	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;
		if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
			range_whole = 1;
		scanned = 1;
	}
retry:
	while (!done && (index <= end) &&
	       (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index,
C
Chris Mason 已提交
2425 2426
			      PAGECACHE_TAG_DIRTY, min(end - index,
				  (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439
		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
			 */
C
Chris Mason 已提交
2440 2441 2442 2443
			if (tree->ops && tree->ops->write_cache_pages_lock_hook)
				tree->ops->write_cache_pages_lock_hook(page);
			else
				lock_page(page);
2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455

			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
2456
			if (wbc->sync_mode != WB_SYNC_NONE) {
2457 2458
				if (PageWriteback(page))
					flush_fn(data);
2459
				wait_on_page_writeback(page);
C
Chris Mason 已提交
2460
			}
2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473

			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;
			}
2474
			if (ret || wbc->nr_to_write <= 0)
2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495
				done = 1;
			if (wbc->nonblocking && bdi_write_congested(bdi)) {
				wbc->encountered_congestion = 1;
				done = 1;
			}
		}
		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;
	}
	return ret;
}

2496
static void flush_epd_write_bio(struct extent_page_data *epd)
C
Chris Mason 已提交
2497 2498
{
	if (epd->bio) {
2499 2500 2501 2502
		if (epd->sync_io)
			submit_one_bio(WRITE_SYNC, epd->bio, 0, 0);
		else
			submit_one_bio(WRITE, epd->bio, 0, 0);
C
Chris Mason 已提交
2503 2504 2505 2506
		epd->bio = NULL;
	}
}

2507 2508 2509 2510 2511 2512
static noinline void flush_write_bio(void *data)
{
	struct extent_page_data *epd = data;
	flush_epd_write_bio(epd);
}

2513 2514 2515 2516 2517 2518 2519 2520 2521 2522
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 address_space *mapping = page->mapping;
	struct extent_page_data epd = {
		.bio = NULL,
		.tree = tree,
		.get_extent = get_extent,
2523
		.extent_locked = 0,
2524
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
2525 2526 2527
	};
	struct writeback_control wbc_writepages = {
		.bdi		= wbc->bdi,
2528
		.sync_mode	= wbc->sync_mode,
2529 2530 2531 2532 2533 2534 2535 2536
		.older_than_this = NULL,
		.nr_to_write	= 64,
		.range_start	= page_offset(page) + PAGE_CACHE_SIZE,
		.range_end	= (loff_t)-1,
	};

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

C
Chris Mason 已提交
2537
	extent_write_cache_pages(tree, mapping, &wbc_writepages,
C
Chris Mason 已提交
2538
				 __extent_writepage, &epd, flush_write_bio);
2539
	flush_epd_write_bio(&epd);
2540 2541 2542
	return ret;
}

2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557
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,
2558
		.sync_io = mode == WB_SYNC_ALL,
2559 2560 2561 2562 2563 2564 2565 2566 2567 2568
	};
	struct writeback_control wbc_writepages = {
		.bdi		= inode->i_mapping->backing_dev_info,
		.sync_mode	= mode,
		.older_than_this = NULL,
		.nr_to_write	= nr_pages * 2,
		.range_start	= start,
		.range_end	= end + 1,
	};

C
Chris Mason 已提交
2569
	while (start <= end) {
2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583
		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;
	}

2584
	flush_epd_write_bio(&epd);
2585 2586
	return ret;
}
2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597

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,
2598
		.extent_locked = 0,
2599
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
2600 2601
	};

C
Chris Mason 已提交
2602
	ret = extent_write_cache_pages(tree, mapping, wbc,
C
Chris Mason 已提交
2603 2604
				       __extent_writepage, &epd,
				       flush_write_bio);
2605
	flush_epd_write_bio(&epd);
2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616
	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;
	struct pagevec pvec;
C
Chris Mason 已提交
2617
	unsigned long bio_flags = 0;
2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634

	pagevec_init(&pvec, 0);
	for (page_idx = 0; page_idx < nr_pages; page_idx++) {
		struct page *page = list_entry(pages->prev, struct page, lru);

		prefetchw(&page->flags);
		list_del(&page->lru);
		/*
		 * what we want to do here is call add_to_page_cache_lru,
		 * but that isn't exported, so we reproduce it here
		 */
		if (!add_to_page_cache(page, mapping,
					page->index, GFP_KERNEL)) {

			/* open coding of lru_cache_add, also not exported */
			page_cache_get(page);
			if (!pagevec_add(&pvec, page))
2635
				__pagevec_lru_add_file(&pvec);
2636
			__extent_read_full_page(tree, page, get_extent,
C
Chris Mason 已提交
2637
						&bio, 0, &bio_flags);
2638 2639 2640 2641
		}
		page_cache_release(page);
	}
	if (pagevec_count(&pvec))
2642
		__pagevec_lru_add_file(&pvec);
2643 2644
	BUG_ON(!list_empty(pages));
	if (bio)
C
Chris Mason 已提交
2645
		submit_one_bio(READ, bio, 0, bio_flags);
2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660
	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)
{
	u64 start = ((u64)page->index << PAGE_CACHE_SHIFT);
	u64 end = start + PAGE_CACHE_SIZE - 1;
	size_t blocksize = page->mapping->host->i_sb->s_blocksize;

C
Chris Mason 已提交
2661
	start += (offset + blocksize - 1) & ~(blocksize - 1);
2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719
	if (start > end)
		return 0;

	lock_extent(tree, start, end, GFP_NOFS);
	wait_on_extent_writeback(tree, start, end);
	clear_extent_bit(tree, start, end,
			 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC,
			 1, 1, GFP_NOFS);
	return 0;
}

/*
 * simple commit_write call, set_range_dirty is used to mark both
 * the pages and the extent records as dirty
 */
int extent_commit_write(struct extent_io_tree *tree,
			struct inode *inode, struct page *page,
			unsigned from, unsigned to)
{
	loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to;

	set_page_extent_mapped(page);
	set_page_dirty(page);

	if (pos > inode->i_size) {
		i_size_write(inode, pos);
		mark_inode_dirty(inode);
	}
	return 0;
}

int extent_prepare_write(struct extent_io_tree *tree,
			 struct inode *inode, struct page *page,
			 unsigned from, unsigned to, get_extent_t *get_extent)
{
	u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
	u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
	u64 block_start;
	u64 orig_block_start;
	u64 block_end;
	u64 cur_end;
	struct extent_map *em;
	unsigned blocksize = 1 << inode->i_blkbits;
	size_t page_offset = 0;
	size_t block_off_start;
	size_t block_off_end;
	int err = 0;
	int iocount = 0;
	int ret = 0;
	int isnew;

	set_page_extent_mapped(page);

	block_start = (page_start + from) & ~((u64)blocksize - 1);
	block_end = (page_start + to - 1) | (blocksize - 1);
	orig_block_start = block_start;

	lock_extent(tree, page_start, page_end, GFP_NOFS);
C
Chris Mason 已提交
2720
	while (block_start <= block_end) {
2721 2722
		em = get_extent(inode, page, page_offset, block_start,
				block_end - block_start + 1, 1);
C
Chris Mason 已提交
2723
		if (IS_ERR(em) || !em)
2724
			goto err;
C
Chris Mason 已提交
2725

2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766
		cur_end = min(block_end, extent_map_end(em) - 1);
		block_off_start = block_start & (PAGE_CACHE_SIZE - 1);
		block_off_end = block_off_start + blocksize;
		isnew = clear_extent_new(tree, block_start, cur_end, GFP_NOFS);

		if (!PageUptodate(page) && isnew &&
		    (block_off_end > to || block_off_start < from)) {
			void *kaddr;

			kaddr = kmap_atomic(page, KM_USER0);
			if (block_off_end > to)
				memset(kaddr + to, 0, block_off_end - to);
			if (block_off_start < from)
				memset(kaddr + block_off_start, 0,
				       from - block_off_start);
			flush_dcache_page(page);
			kunmap_atomic(kaddr, KM_USER0);
		}
		if ((em->block_start != EXTENT_MAP_HOLE &&
		     em->block_start != EXTENT_MAP_INLINE) &&
		    !isnew && !PageUptodate(page) &&
		    (block_off_end > to || block_off_start < from) &&
		    !test_range_bit(tree, block_start, cur_end,
				    EXTENT_UPTODATE, 1)) {
			u64 sector;
			u64 extent_offset = block_start - em->start;
			size_t iosize;
			sector = (em->block_start + extent_offset) >> 9;
			iosize = (cur_end - block_start + blocksize) &
				~((u64)blocksize - 1);
			/*
			 * we've already got the extent locked, but we
			 * need to split the state such that our end_bio
			 * handler can clear the lock.
			 */
			set_extent_bit(tree, block_start,
				       block_start + iosize - 1,
				       EXTENT_LOCKED, 0, NULL, GFP_NOFS);
			ret = submit_extent_page(READ, tree, page,
					 sector, iosize, page_offset, em->bdev,
					 NULL, 1,
C
Chris Mason 已提交
2767 2768
					 end_bio_extent_preparewrite, 0,
					 0, 0);
2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789
			iocount++;
			block_start = block_start + iosize;
		} else {
			set_extent_uptodate(tree, block_start, cur_end,
					    GFP_NOFS);
			unlock_extent(tree, block_start, cur_end, GFP_NOFS);
			block_start = cur_end + 1;
		}
		page_offset = block_start & (PAGE_CACHE_SIZE - 1);
		free_extent_map(em);
	}
	if (iocount) {
		wait_extent_bit(tree, orig_block_start,
				block_end, EXTENT_LOCKED);
	}
	check_page_uptodate(tree, page);
err:
	/* FIXME, zero out newly allocated blocks on error */
	return err;
}

2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802
/*
 * 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.
 */
int try_release_extent_state(struct extent_map_tree *map,
			     struct extent_io_tree *tree, struct page *page,
			     gfp_t mask)
{
	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
	u64 end = start + PAGE_CACHE_SIZE - 1;
	int ret = 1;

2803 2804
	if (test_range_bit(tree, start, end,
			   EXTENT_IOBITS | EXTENT_ORDERED, 0))
2805 2806 2807 2808 2809 2810 2811 2812 2813 2814
		ret = 0;
	else {
		if ((mask & GFP_NOFS) == GFP_NOFS)
			mask = GFP_NOFS;
		clear_extent_bit(tree, start, end, EXTENT_UPTODATE,
				 1, 1, mask);
	}
	return ret;
}

2815 2816 2817 2818 2819 2820
/*
 * 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,
2821 2822
			       struct extent_io_tree *tree, struct page *page,
			       gfp_t mask)
2823 2824 2825 2826
{
	struct extent_map *em;
	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
	u64 end = start + PAGE_CACHE_SIZE - 1;
2827

2828 2829
	if ((mask & __GFP_WAIT) &&
	    page->mapping->host->i_size > 16 * 1024 * 1024) {
2830
		u64 len;
2831
		while (start <= end) {
2832
			len = end - start + 1;
2833
			spin_lock(&map->lock);
2834
			em = lookup_extent_mapping(map, start, len);
2835 2836 2837 2838
			if (!em || IS_ERR(em)) {
				spin_unlock(&map->lock);
				break;
			}
2839 2840
			if (test_bit(EXTENT_FLAG_PINNED, &em->flags) ||
			    em->start != start) {
2841 2842 2843 2844 2845 2846
				spin_unlock(&map->lock);
				free_extent_map(em);
				break;
			}
			if (!test_range_bit(tree, em->start,
					    extent_map_end(em) - 1,
C
Chris Mason 已提交
2847 2848 2849
					    EXTENT_LOCKED | EXTENT_WRITEBACK |
					    EXTENT_ORDERED,
					    0)) {
2850 2851 2852 2853 2854
				remove_extent_mapping(map, em);
				/* once for the rb tree */
				free_extent_map(em);
			}
			start = extent_map_end(em);
2855
			spin_unlock(&map->lock);
2856 2857

			/* once for us */
2858 2859 2860
			free_extent_map(em);
		}
	}
2861
	return try_release_extent_state(map, tree, page, mask);
2862 2863 2864 2865 2866 2867 2868 2869
}

sector_t extent_bmap(struct address_space *mapping, sector_t iblock,
		get_extent_t *get_extent)
{
	struct inode *inode = mapping->host;
	u64 start = iblock << inode->i_blkbits;
	sector_t sector = 0;
Y
Yan Zheng 已提交
2870
	size_t blksize = (1 << inode->i_blkbits);
2871 2872
	struct extent_map *em;

Y
Yan Zheng 已提交
2873 2874 2875 2876 2877
	lock_extent(&BTRFS_I(inode)->io_tree, start, start + blksize - 1,
		    GFP_NOFS);
	em = get_extent(inode, NULL, 0, start, blksize, 0);
	unlock_extent(&BTRFS_I(inode)->io_tree, start, start + blksize - 1,
		      GFP_NOFS);
2878 2879 2880
	if (!em || IS_ERR(em))
		return 0;

Y
Yan Zheng 已提交
2881
	if (em->block_start > EXTENT_MAP_LAST_BYTE)
2882 2883 2884 2885 2886 2887 2888 2889
		goto out;

	sector = (em->block_start + start - em->start) >> inode->i_blkbits;
out:
	free_extent_map(em);
	return sector;
}

Y
Yehuda Sadeh 已提交
2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920 2921 2922 2923 2924 2925 2926
int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
		__u64 start, __u64 len, get_extent_t *get_extent)
{
	int ret;
	u64 off = start;
	u64 max = start + len;
	u32 flags = 0;
	u64 disko = 0;
	struct extent_map *em = NULL;
	int end = 0;
	u64 em_start = 0, em_len = 0;
	unsigned long emflags;
	ret = 0;

	if (len == 0)
		return -EINVAL;

	lock_extent(&BTRFS_I(inode)->io_tree, start, start + len,
		GFP_NOFS);
	em = get_extent(inode, NULL, 0, off, max - off, 0);
	if (!em)
		goto out;
	if (IS_ERR(em)) {
		ret = PTR_ERR(em);
		goto out;
	}
	while (!end) {
		off = em->start + em->len;
		if (off >= max)
			end = 1;

		em_start = em->start;
		em_len = em->len;

		disko = 0;
		flags = 0;

2927
		if (em->block_start == EXTENT_MAP_LAST_BYTE) {
Y
Yehuda Sadeh 已提交
2928 2929
			end = 1;
			flags |= FIEMAP_EXTENT_LAST;
2930
		} else if (em->block_start == EXTENT_MAP_HOLE) {
Y
Yehuda Sadeh 已提交
2931
			flags |= FIEMAP_EXTENT_UNWRITTEN;
2932
		} else if (em->block_start == EXTENT_MAP_INLINE) {
Y
Yehuda Sadeh 已提交
2933 2934
			flags |= (FIEMAP_EXTENT_DATA_INLINE |
				  FIEMAP_EXTENT_NOT_ALIGNED);
2935
		} else if (em->block_start == EXTENT_MAP_DELALLOC) {
Y
Yehuda Sadeh 已提交
2936 2937
			flags |= (FIEMAP_EXTENT_DELALLOC |
				  FIEMAP_EXTENT_UNKNOWN);
2938
		} else {
Y
Yehuda Sadeh 已提交
2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975
			disko = em->block_start;
		}
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
			flags |= FIEMAP_EXTENT_ENCODED;

		emflags = em->flags;
		free_extent_map(em);
		em = NULL;

		if (!end) {
			em = get_extent(inode, NULL, 0, off, max - off, 0);
			if (!em)
				goto out;
			if (IS_ERR(em)) {
				ret = PTR_ERR(em);
				goto out;
			}
			emflags = em->flags;
		}
		if (test_bit(EXTENT_FLAG_VACANCY, &emflags)) {
			flags |= FIEMAP_EXTENT_LAST;
			end = 1;
		}

		ret = fiemap_fill_next_extent(fieinfo, em_start, disko,
					em_len, flags);
		if (ret)
			goto out_free;
	}
out_free:
	free_extent_map(em);
out:
	unlock_extent(&BTRFS_I(inode)->io_tree, start, start + len,
			GFP_NOFS);
	return ret;
}

2976 2977 2978 2979 2980 2981 2982 2983 2984 2985
static inline struct page *extent_buffer_page(struct extent_buffer *eb,
					      unsigned long i)
{
	struct page *p;
	struct address_space *mapping;

	if (i == 0)
		return eb->first_page;
	i += eb->start >> PAGE_CACHE_SHIFT;
	mapping = eb->first_page->mapping;
C
Chris Mason 已提交
2986 2987
	if (!mapping)
		return NULL;
2988 2989 2990 2991 2992 2993 2994

	/*
	 * extent_buffer_page is only called after pinning the page
	 * by increasing the reference count.  So we know the page must
	 * be in the radix tree.
	 */
	rcu_read_lock();
2995
	p = radix_tree_lookup(&mapping->page_tree, i);
2996
	rcu_read_unlock();
2997

2998 2999 3000
	return p;
}

3001
static inline unsigned long num_extent_pages(u64 start, u64 len)
3002
{
3003 3004
	return ((start + len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT) -
		(start >> PAGE_CACHE_SHIFT);
3005 3006
}

3007 3008 3009 3010 3011 3012
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;
3013
#if LEAK_DEBUG
3014
	unsigned long flags;
C
Chris Mason 已提交
3015
#endif
3016 3017 3018 3019

	eb = kmem_cache_zalloc(extent_buffer_cache, mask);
	eb->start = start;
	eb->len = len;
3020 3021 3022
	spin_lock_init(&eb->lock);
	init_waitqueue_head(&eb->lock_wq);

3023
#if LEAK_DEBUG
3024 3025 3026
	spin_lock_irqsave(&leak_lock, flags);
	list_add(&eb->leak_list, &buffers);
	spin_unlock_irqrestore(&leak_lock, flags);
C
Chris Mason 已提交
3027
#endif
3028 3029 3030 3031 3032 3033 3034
	atomic_set(&eb->refs, 1);

	return eb;
}

static void __free_extent_buffer(struct extent_buffer *eb)
{
3035
#if LEAK_DEBUG
3036 3037 3038 3039
	unsigned long flags;
	spin_lock_irqsave(&leak_lock, flags);
	list_del(&eb->leak_list);
	spin_unlock_irqrestore(&leak_lock, flags);
C
Chris Mason 已提交
3040
#endif
3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052
	kmem_cache_free(extent_buffer_cache, eb);
}

struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree,
					  u64 start, unsigned long len,
					  struct page *page0,
					  gfp_t mask)
{
	unsigned long num_pages = num_extent_pages(start, len);
	unsigned long i;
	unsigned long index = start >> PAGE_CACHE_SHIFT;
	struct extent_buffer *eb;
3053
	struct extent_buffer *exists = NULL;
3054 3055 3056 3057
	struct page *p;
	struct address_space *mapping = tree->mapping;
	int uptodate = 1;

3058 3059 3060 3061 3062
	spin_lock(&tree->buffer_lock);
	eb = buffer_search(tree, start);
	if (eb) {
		atomic_inc(&eb->refs);
		spin_unlock(&tree->buffer_lock);
J
Josef Bacik 已提交
3063
		mark_page_accessed(eb->first_page);
3064 3065 3066 3067
		return eb;
	}
	spin_unlock(&tree->buffer_lock);

3068
	eb = __alloc_extent_buffer(tree, start, len, mask);
3069
	if (!eb)
3070 3071 3072 3073 3074 3075 3076 3077 3078 3079
		return NULL;

	if (page0) {
		eb->first_page = page0;
		i = 1;
		index++;
		page_cache_get(page0);
		mark_page_accessed(page0);
		set_page_extent_mapped(page0);
		set_page_extent_head(page0, len);
3080
		uptodate = PageUptodate(page0);
3081 3082 3083 3084 3085 3086 3087
	} else {
		i = 0;
	}
	for (; i < num_pages; i++, index++) {
		p = find_or_create_page(mapping, index, mask | __GFP_HIGHMEM);
		if (!p) {
			WARN_ON(1);
3088
			goto free_eb;
3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102
		}
		set_page_extent_mapped(p);
		mark_page_accessed(p);
		if (i == 0) {
			eb->first_page = p;
			set_page_extent_head(p, len);
		} else {
			set_page_private(p, EXTENT_PAGE_PRIVATE);
		}
		if (!PageUptodate(p))
			uptodate = 0;
		unlock_page(p);
	}
	if (uptodate)
3103
		set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
3104

3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116
	spin_lock(&tree->buffer_lock);
	exists = buffer_tree_insert(tree, start, &eb->rb_node);
	if (exists) {
		/* add one reference for the caller */
		atomic_inc(&exists->refs);
		spin_unlock(&tree->buffer_lock);
		goto free_eb;
	}
	spin_unlock(&tree->buffer_lock);

	/* add one reference for the tree */
	atomic_inc(&eb->refs);
3117 3118
	return eb;

3119
free_eb:
3120
	if (!atomic_dec_and_test(&eb->refs))
3121 3122
		return exists;
	for (index = 1; index < i; index++)
3123
		page_cache_release(extent_buffer_page(eb, index));
3124
	page_cache_release(extent_buffer_page(eb, 0));
3125
	__free_extent_buffer(eb);
3126
	return exists;
3127 3128 3129 3130 3131 3132 3133 3134
}

struct extent_buffer *find_extent_buffer(struct extent_io_tree *tree,
					 u64 start, unsigned long len,
					  gfp_t mask)
{
	struct extent_buffer *eb;

3135 3136 3137 3138 3139
	spin_lock(&tree->buffer_lock);
	eb = buffer_search(tree, start);
	if (eb)
		atomic_inc(&eb->refs);
	spin_unlock(&tree->buffer_lock);
3140

J
Josef Bacik 已提交
3141 3142 3143
	if (eb)
		mark_page_accessed(eb->first_page);

3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154
	return eb;
}

void free_extent_buffer(struct extent_buffer *eb)
{
	if (!eb)
		return;

	if (!atomic_dec_and_test(&eb->refs))
		return;

3155
	WARN_ON(1);
3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168
}

int clear_extent_buffer_dirty(struct extent_io_tree *tree,
			      struct extent_buffer *eb)
{
	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);
3169
		if (!PageDirty(page))
C
Chris Mason 已提交
3170 3171
			continue;

3172
		lock_page(page);
3173 3174 3175 3176 3177 3178
		if (i == 0)
			set_page_extent_head(page, eb->len);
		else
			set_page_private(page, EXTENT_PAGE_PRIVATE);

		clear_page_dirty_for_io(page);
3179
		spin_lock_irq(&page->mapping->tree_lock);
3180 3181 3182 3183 3184
		if (!PageDirty(page)) {
			radix_tree_tag_clear(&page->mapping->page_tree,
						page_index(page),
						PAGECACHE_TAG_DIRTY);
		}
3185
		spin_unlock_irq(&page->mapping->tree_lock);
3186
		unlock_page(page);
3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202
	}
	return 0;
}

int wait_on_extent_buffer_writeback(struct extent_io_tree *tree,
				    struct extent_buffer *eb)
{
	return wait_on_extent_writeback(tree, eb->start,
					eb->start + eb->len - 1);
}

int set_extent_buffer_dirty(struct extent_io_tree *tree,
			     struct extent_buffer *eb)
{
	unsigned long i;
	unsigned long num_pages;
3203
	int was_dirty = 0;
3204

3205
	was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
3206
	num_pages = num_extent_pages(eb->start, eb->len);
3207
	for (i = 0; i < num_pages; i++)
3208
		__set_page_dirty_nobuffers(extent_buffer_page(eb, i));
3209
	return was_dirty;
3210 3211
}

3212 3213 3214 3215 3216 3217 3218 3219
int clear_extent_buffer_uptodate(struct extent_io_tree *tree,
				struct extent_buffer *eb)
{
	unsigned long i;
	struct page *page;
	unsigned long num_pages;

	num_pages = num_extent_pages(eb->start, eb->len);
3220
	clear_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
3221 3222 3223 3224 3225

	clear_extent_uptodate(tree, eb->start, eb->start + eb->len - 1,
			      GFP_NOFS);
	for (i = 0; i < num_pages; i++) {
		page = extent_buffer_page(eb, i);
C
Chris Mason 已提交
3226 3227
		if (page)
			ClearPageUptodate(page);
3228 3229 3230 3231
	}
	return 0;
}

3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255
int set_extent_buffer_uptodate(struct extent_io_tree *tree,
				struct extent_buffer *eb)
{
	unsigned long i;
	struct page *page;
	unsigned long num_pages;

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

	set_extent_uptodate(tree, eb->start, eb->start + eb->len - 1,
			    GFP_NOFS);
	for (i = 0; i < num_pages; i++) {
		page = extent_buffer_page(eb, i);
		if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) ||
		    ((i == num_pages - 1) &&
		     ((eb->start + eb->len) & (PAGE_CACHE_SIZE - 1)))) {
			check_page_uptodate(tree, page);
			continue;
		}
		SetPageUptodate(page);
	}
	return 0;
}

3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267
int extent_range_uptodate(struct extent_io_tree *tree,
			  u64 start, u64 end)
{
	struct page *page;
	int ret;
	int pg_uptodate = 1;
	int uptodate;
	unsigned long index;

	ret = test_range_bit(tree, start, end, EXTENT_UPTODATE, 1);
	if (ret)
		return 1;
C
Chris Mason 已提交
3268
	while (start <= end) {
3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281
		index = start >> PAGE_CACHE_SHIFT;
		page = find_get_page(tree->mapping, index);
		uptodate = PageUptodate(page);
		page_cache_release(page);
		if (!uptodate) {
			pg_uptodate = 0;
			break;
		}
		start += PAGE_CACHE_SIZE;
	}
	return pg_uptodate;
}

3282
int extent_buffer_uptodate(struct extent_io_tree *tree,
3283
			   struct extent_buffer *eb)
3284
{
3285
	int ret = 0;
3286 3287
	unsigned long num_pages;
	unsigned long i;
3288 3289 3290
	struct page *page;
	int pg_uptodate = 1;

3291
	if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags))
C
Chris Mason 已提交
3292
		return 1;
3293

C
Chris Mason 已提交
3294
	ret = test_range_bit(tree, eb->start, eb->start + eb->len - 1,
3295
			   EXTENT_UPTODATE, 1);
C
Chris Mason 已提交
3296 3297
	if (ret)
		return ret;
3298 3299 3300 3301 3302 3303 3304 3305 3306

	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = 0; i < num_pages; i++) {
		page = extent_buffer_page(eb, i);
		if (!PageUptodate(page)) {
			pg_uptodate = 0;
			break;
		}
	}
C
Chris Mason 已提交
3307
	return pg_uptodate;
3308 3309 3310 3311
}

int read_extent_buffer_pages(struct extent_io_tree *tree,
			     struct extent_buffer *eb,
3312
			     u64 start, int wait,
3313
			     get_extent_t *get_extent, int mirror_num)
3314 3315 3316 3317 3318 3319
{
	unsigned long i;
	unsigned long start_i;
	struct page *page;
	int err;
	int ret = 0;
3320 3321 3322
	int locked_pages = 0;
	int all_uptodate = 1;
	int inc_all_pages = 0;
3323
	unsigned long num_pages;
3324
	struct bio *bio = NULL;
C
Chris Mason 已提交
3325
	unsigned long bio_flags = 0;
3326

3327
	if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags))
3328 3329
		return 0;

3330
	if (test_range_bit(tree, eb->start, eb->start + eb->len - 1,
3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346
			   EXTENT_UPTODATE, 1)) {
		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);
		if (!wait) {
3347
			if (!trylock_page(page))
3348
				goto unlock_exit;
3349 3350 3351
		} else {
			lock_page(page);
		}
3352
		locked_pages++;
C
Chris Mason 已提交
3353
		if (!PageUptodate(page))
3354 3355 3356 3357
			all_uptodate = 0;
	}
	if (all_uptodate) {
		if (start_i == 0)
3358
			set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
3359 3360 3361 3362 3363 3364 3365 3366 3367 3368
		goto unlock_exit;
	}

	for (i = start_i; i < num_pages; i++) {
		page = extent_buffer_page(eb, i);
		if (inc_all_pages)
			page_cache_get(page);
		if (!PageUptodate(page)) {
			if (start_i == 0)
				inc_all_pages = 1;
3369
			ClearPageError(page);
3370
			err = __extent_read_full_page(tree, page,
3371
						      get_extent, &bio,
C
Chris Mason 已提交
3372
						      mirror_num, &bio_flags);
C
Chris Mason 已提交
3373
			if (err)
3374 3375 3376 3377 3378 3379
				ret = err;
		} else {
			unlock_page(page);
		}
	}

3380
	if (bio)
C
Chris Mason 已提交
3381
		submit_one_bio(READ, bio, mirror_num, bio_flags);
3382

C
Chris Mason 已提交
3383
	if (ret || !wait)
3384
		return ret;
C
Chris Mason 已提交
3385

3386 3387 3388
	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
3389
		if (!PageUptodate(page))
3390 3391
			ret = -EIO;
	}
C
Chris Mason 已提交
3392

3393
	if (!ret)
3394
		set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
3395
	return ret;
3396 3397 3398

unlock_exit:
	i = start_i;
C
Chris Mason 已提交
3399
	while (locked_pages > 0) {
3400 3401 3402 3403 3404 3405
		page = extent_buffer_page(eb, i);
		i++;
		unlock_page(page);
		locked_pages--;
	}
	return ret;
3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424
}

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
3425
	while (len > 0) {
3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462
		page = extent_buffer_page(eb, i);

		cur = min(len, (PAGE_CACHE_SIZE - offset));
		kaddr = kmap_atomic(page, KM_USER1);
		memcpy(dst, kaddr + offset, cur);
		kunmap_atomic(kaddr, KM_USER1);

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

int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start,
			       unsigned long min_len, char **token, char **map,
			       unsigned long *map_start,
			       unsigned long *map_len, int km)
{
	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
3463

3464
	if (start + min_len > eb->len) {
C
Chris Mason 已提交
3465 3466 3467
		printk(KERN_ERR "btrfs bad mapping eb start %llu len %lu, "
		       "wanted %lu %lu\n", (unsigned long long)eb->start,
		       eb->len, start, min_len);
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 3495 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
		WARN_ON(1);
	}

	p = extent_buffer_page(eb, i);
	kaddr = kmap_atomic(p, km);
	*token = kaddr;
	*map = kaddr + offset;
	*map_len = PAGE_CACHE_SIZE - offset;
	return 0;
}

int map_extent_buffer(struct extent_buffer *eb, unsigned long start,
		      unsigned long min_len,
		      char **token, char **map,
		      unsigned long *map_start,
		      unsigned long *map_len, int km)
{
	int err;
	int save = 0;
	if (eb->map_token) {
		unmap_extent_buffer(eb, eb->map_token, km);
		eb->map_token = NULL;
		save = 1;
	}
	err = map_private_extent_buffer(eb, start, min_len, token, map,
				       map_start, map_len, km);
	if (!err && save) {
		eb->map_token = *token;
		eb->kaddr = *map;
		eb->map_start = *map_start;
		eb->map_len = *map_len;
	}
	return err;
}

void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km)
{
	kunmap_atomic(token, km);
}

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
3526
	while (len > 0) {
3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560
		page = extent_buffer_page(eb, i);

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

		kaddr = kmap_atomic(page, KM_USER0);
		ret = memcmp(ptr, kaddr + offset, cur);
		kunmap_atomic(kaddr, KM_USER0);
		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
3561
	while (len > 0) {
3562 3563 3564 3565 3566 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
		page = extent_buffer_page(eb, i);
		WARN_ON(!PageUptodate(page));

		cur = min(len, PAGE_CACHE_SIZE - offset);
		kaddr = kmap_atomic(page, KM_USER1);
		memcpy(kaddr + offset, src, cur);
		kunmap_atomic(kaddr, KM_USER1);

		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
3592
	while (len > 0) {
3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623
		page = extent_buffer_page(eb, i);
		WARN_ON(!PageUptodate(page));

		cur = min(len, PAGE_CACHE_SIZE - offset);
		kaddr = kmap_atomic(page, KM_USER0);
		memset(kaddr + offset, c, cur);
		kunmap_atomic(kaddr, KM_USER0);

		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
3624
	while (len > 0) {
3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689
		page = extent_buffer_page(dst, i);
		WARN_ON(!PageUptodate(page));

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

		kaddr = kmap_atomic(page, KM_USER0);
		read_extent_buffer(src, kaddr + offset, src_offset, cur);
		kunmap_atomic(kaddr, KM_USER0);

		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)
{
	char *dst_kaddr = kmap_atomic(dst_page, KM_USER0);
	if (dst_page == src_page) {
		memmove(dst_kaddr + dst_off, dst_kaddr + src_off, len);
	} else {
		char *src_kaddr = kmap_atomic(src_page, KM_USER1);
		char *p = dst_kaddr + dst_off + len;
		char *s = src_kaddr + src_off + len;

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

		kunmap_atomic(src_kaddr, KM_USER1);
	}
	kunmap_atomic(dst_kaddr, KM_USER0);
}

static void copy_pages(struct page *dst_page, struct page *src_page,
		       unsigned long dst_off, unsigned long src_off,
		       unsigned long len)
{
	char *dst_kaddr = kmap_atomic(dst_page, KM_USER0);
	char *src_kaddr;

	if (dst_page != src_page)
		src_kaddr = kmap_atomic(src_page, KM_USER1);
	else
		src_kaddr = dst_kaddr;

	memcpy(dst_kaddr + dst_off, src_kaddr + src_off, len);
	kunmap_atomic(dst_kaddr, KM_USER0);
	if (dst_page != src_page)
		kunmap_atomic(src_kaddr, KM_USER1);
}

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
3690 3691
		printk(KERN_ERR "btrfs memmove bogus src_offset %lu move "
		       "len %lu dst len %lu\n", src_offset, len, dst->len);
3692 3693 3694
		BUG_ON(1);
	}
	if (dst_offset + len > dst->len) {
C
Chris Mason 已提交
3695 3696
		printk(KERN_ERR "btrfs memmove bogus dst_offset %lu move "
		       "len %lu dst len %lu\n", dst_offset, len, dst->len);
3697 3698 3699
		BUG_ON(1);
	}

C
Chris Mason 已提交
3700
	while (len > 0) {
3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736
		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
3737 3738
		printk(KERN_ERR "btrfs memmove bogus src_offset %lu move "
		       "len %lu len %lu\n", src_offset, len, dst->len);
3739 3740 3741
		BUG_ON(1);
	}
	if (dst_offset + len > dst->len) {
C
Chris Mason 已提交
3742 3743
		printk(KERN_ERR "btrfs memmove bogus dst_offset %lu move "
		       "len %lu len %lu\n", dst_offset, len, dst->len);
3744 3745 3746 3747 3748 3749
		BUG_ON(1);
	}
	if (dst_offset < src_offset) {
		memcpy_extent_buffer(dst, dst_offset, src_offset, len);
		return;
	}
C
Chris Mason 已提交
3750
	while (len > 0) {
3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770
		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;
	}
}
3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788

int try_release_extent_buffer(struct extent_io_tree *tree, struct page *page)
{
	u64 start = page_offset(page);
	struct extent_buffer *eb;
	int ret = 1;
	unsigned long i;
	unsigned long num_pages;

	spin_lock(&tree->buffer_lock);
	eb = buffer_search(tree, start);
	if (!eb)
		goto out;

	if (atomic_read(&eb->refs) > 1) {
		ret = 0;
		goto out;
	}
3789 3790 3791 3792
	if (test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
		ret = 0;
		goto out;
	}
3793 3794
	/* at this point we can safely release the extent buffer */
	num_pages = num_extent_pages(eb->start, eb->len);
C
Christoph Hellwig 已提交
3795 3796
	for (i = 0; i < num_pages; i++)
		page_cache_release(extent_buffer_page(eb, i));
3797 3798 3799 3800 3801 3802
	rb_erase(&eb->rb_node, &tree->buffer);
	__free_extent_buffer(eb);
out:
	spin_unlock(&tree->buffer_lock);
	return ret;
}