extent_io.c 94.0 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14
#include <linux/bitops.h>
#include <linux/slab.h>
#include <linux/bio.h>
#include <linux/mm.h>
#include <linux/pagemap.h>
#include <linux/page-flags.h>
#include <linux/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"
15
#include "compat.h"
16 17
#include "ctree.h"
#include "btrfs_inode.h"
18 19 20 21 22 23

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
24

25
#define LEAK_DEBUG 0
26
#if LEAK_DEBUG
C
Chris Mason 已提交
27
static DEFINE_SPINLOCK(leak_lock);
C
Chris Mason 已提交
28
#endif
29 30 31 32 33 34 35 36 37 38 39 40 41

#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;
42 43 44 45

	/* tells writepage not to lock the state bits for this range
	 * it still does the unlocking
	 */
46 47 48 49
	unsigned int extent_locked:1;

	/* tells the submit_bio code to use a WRITE_SYNC */
	unsigned int sync_io:1;
50 51 52 53
};

int __init extent_io_init(void)
{
54 55 56
	extent_state_cache = kmem_cache_create("extent_state",
			sizeof(struct extent_state), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
57 58 59
	if (!extent_state_cache)
		return -ENOMEM;

60 61 62
	extent_buffer_cache = kmem_cache_create("extent_buffers",
			sizeof(struct extent_buffer), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
63 64 65 66 67 68 69 70 71 72 73 74
	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;
75
	struct extent_buffer *eb;
76 77

	while (!list_empty(&states)) {
78
		state = list_entry(states.next, struct extent_state, leak_list);
C
Chris Mason 已提交
79 80 81 82 83
		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));
84
		list_del(&state->leak_list);
85 86 87 88
		kmem_cache_free(extent_state_cache, state);

	}

89 90
	while (!list_empty(&buffers)) {
		eb = list_entry(buffers.next, struct extent_buffer, leak_list);
C
Chris Mason 已提交
91 92 93
		printk(KERN_ERR "btrfs buffer leak start %llu len %lu "
		       "refs %d\n", (unsigned long long)eb->start,
		       eb->len, atomic_read(&eb->refs));
94 95 96
		list_del(&eb->leak_list);
		kmem_cache_free(extent_buffer_cache, eb);
	}
97 98 99 100 101 102 103
	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,
104
			 struct address_space *mapping)
105
{
106
	tree->state = RB_ROOT;
107
	INIT_RADIX_TREE(&tree->buffer, GFP_ATOMIC);
108 109
	tree->ops = NULL;
	tree->dirty_bytes = 0;
110
	spin_lock_init(&tree->lock);
111
	spin_lock_init(&tree->buffer_lock);
112 113 114
	tree->mapping = mapping;
}

115
static struct extent_state *alloc_extent_state(gfp_t mask)
116 117
{
	struct extent_state *state;
118
#if LEAK_DEBUG
119
	unsigned long flags;
C
Chris Mason 已提交
120
#endif
121 122

	state = kmem_cache_alloc(extent_state_cache, mask);
123
	if (!state)
124 125 126
		return state;
	state->state = 0;
	state->private = 0;
127
	state->tree = NULL;
128
#if LEAK_DEBUG
129 130 131
	spin_lock_irqsave(&leak_lock, flags);
	list_add(&state->leak_list, &states);
	spin_unlock_irqrestore(&leak_lock, flags);
C
Chris Mason 已提交
132
#endif
133 134 135 136 137
	atomic_set(&state->refs, 1);
	init_waitqueue_head(&state->wq);
	return state;
}

138
void free_extent_state(struct extent_state *state)
139 140 141 142
{
	if (!state)
		return;
	if (atomic_dec_and_test(&state->refs)) {
143
#if LEAK_DEBUG
144
		unsigned long flags;
C
Chris Mason 已提交
145
#endif
146
		WARN_ON(state->tree);
147
#if LEAK_DEBUG
148 149 150
		spin_lock_irqsave(&leak_lock, flags);
		list_del(&state->leak_list);
		spin_unlock_irqrestore(&leak_lock, flags);
C
Chris Mason 已提交
151
#endif
152 153 154 155 156 157 158
		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
159 160
	struct rb_node **p = &root->rb_node;
	struct rb_node *parent = NULL;
161 162
	struct tree_entry *entry;

C
Chris Mason 已提交
163
	while (*p) {
164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180
		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;
}

181
static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset,
182 183 184
				     struct rb_node **prev_ret,
				     struct rb_node **next_ret)
{
185
	struct rb_root *root = &tree->state;
C
Chris Mason 已提交
186
	struct rb_node *n = root->rb_node;
187 188 189 190 191
	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
192
	while (n) {
193 194 195 196 197 198 199 200
		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
201
		else
202 203 204 205 206
			return n;
	}

	if (prev_ret) {
		orig_prev = prev;
C
Chris Mason 已提交
207
		while (prev && offset > prev_entry->end) {
208 209 210 211 212 213 214 215 216
			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
217
		while (prev && offset < prev_entry->start) {
218 219 220 221 222 223 224 225
			prev = rb_prev(prev);
			prev_entry = rb_entry(prev, struct tree_entry, rb_node);
		}
		*next_ret = prev;
	}
	return NULL;
}

226 227
static inline struct rb_node *tree_search(struct extent_io_tree *tree,
					  u64 offset)
228
{
229
	struct rb_node *prev = NULL;
230
	struct rb_node *ret;
231

232
	ret = __etree_search(tree, offset, &prev, NULL);
C
Chris Mason 已提交
233
	if (!ret)
234 235 236 237
		return prev;
	return ret;
}

J
Josef Bacik 已提交
238 239 240 241 242 243 244 245
static void merge_cb(struct extent_io_tree *tree, struct extent_state *new,
		     struct extent_state *other)
{
	if (tree->ops && tree->ops->merge_extent_hook)
		tree->ops->merge_extent_hook(tree->mapping->host, new,
					     other);
}

246 247 248 249 250 251 252 253 254 255 256 257 258 259 260
/*
 * 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;

261
	if (state->state & (EXTENT_IOBITS | EXTENT_BOUNDARY))
262 263 264 265 266 267 268
		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) {
J
Josef Bacik 已提交
269
			merge_cb(tree, state, other);
270
			state->start = other->start;
271
			other->tree = NULL;
272 273 274 275 276 277 278 279 280
			rb_erase(&other->rb_node, &tree->state);
			free_extent_state(other);
		}
	}
	other_node = rb_next(&state->rb_node);
	if (other_node) {
		other = rb_entry(other_node, struct extent_state, rb_node);
		if (other->start == state->end + 1 &&
		    other->state == state->state) {
J
Josef Bacik 已提交
281
			merge_cb(tree, state, other);
282 283 284 285
			state->end = other->end;
			other->tree = NULL;
			rb_erase(&other->rb_node, &tree->state);
			free_extent_state(other);
286 287
		}
	}
J
Josef Bacik 已提交
288

289 290 291
	return 0;
}

J
Josef Bacik 已提交
292
static int set_state_cb(struct extent_io_tree *tree,
293
			 struct extent_state *state, int *bits)
294 295
{
	if (tree->ops && tree->ops->set_bit_hook) {
J
Josef Bacik 已提交
296
		return tree->ops->set_bit_hook(tree->mapping->host,
297
					       state, bits);
298
	}
J
Josef Bacik 已提交
299 300

	return 0;
301 302 303
}

static void clear_state_cb(struct extent_io_tree *tree,
304
			   struct extent_state *state, int *bits)
305
{
J
Josef Bacik 已提交
306 307
	if (tree->ops && tree->ops->clear_bit_hook)
		tree->ops->clear_bit_hook(tree->mapping->host, state, bits);
308 309
}

310 311 312 313 314 315 316 317 318 319 320 321
/*
 * 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,
322
			int *bits)
323 324
{
	struct rb_node *node;
325
	int bits_to_set = *bits & ~EXTENT_CTLBITS;
J
Josef Bacik 已提交
326
	int ret;
327 328

	if (end < start) {
C
Chris Mason 已提交
329 330 331
		printk(KERN_ERR "btrfs end < start %llu %llu\n",
		       (unsigned long long)end,
		       (unsigned long long)start);
332 333 334 335
		WARN_ON(1);
	}
	state->start = start;
	state->end = end;
J
Josef Bacik 已提交
336 337 338 339
	ret = set_state_cb(tree, state, bits);
	if (ret)
		return ret;

340
	if (bits_to_set & EXTENT_DIRTY)
J
Josef Bacik 已提交
341
		tree->dirty_bytes += end - start + 1;
342
	state->state |= bits_to_set;
343 344 345 346
	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
347 348 349 350
		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);
351 352
		return -EEXIST;
	}
353
	state->tree = tree;
354 355 356 357
	merge_state(tree, state);
	return 0;
}

J
Josef Bacik 已提交
358 359 360 361 362 363 364 365 366
static int split_cb(struct extent_io_tree *tree, struct extent_state *orig,
		     u64 split)
{
	if (tree->ops && tree->ops->split_extent_hook)
		return tree->ops->split_extent_hook(tree->mapping->host,
						    orig, split);
	return 0;
}

367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384
/*
 * split a given extent state struct in two, inserting the preallocated
 * struct 'prealloc' as the newly created second half.  'split' indicates an
 * offset inside 'orig' where it should be split.
 *
 * Before calling,
 * the tree has 'orig' at [orig->start, orig->end].  After calling, there
 * are two extent state structs in the tree:
 * prealloc: [orig->start, split - 1]
 * orig: [ split, orig->end ]
 *
 * The tree locks are not taken by this function. They need to be held
 * by the caller.
 */
static int split_state(struct extent_io_tree *tree, struct extent_state *orig,
		       struct extent_state *prealloc, u64 split)
{
	struct rb_node *node;
J
Josef Bacik 已提交
385 386 387

	split_cb(tree, orig, split);

388 389 390 391 392 393 394 395 396 397
	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;
	}
398
	prealloc->tree = tree;
399 400 401 402 403 404 405 406 407 408 409 410
	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,
411 412
			    struct extent_state *state,
			    int *bits, int wake)
413
{
414
	int bits_to_clear = *bits & ~EXTENT_CTLBITS;
415
	int ret = state->state & bits_to_clear;
416

417
	if ((bits_to_clear & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) {
418 419 420 421
		u64 range = state->end - state->start + 1;
		WARN_ON(range > tree->dirty_bytes);
		tree->dirty_bytes -= range;
	}
422
	clear_state_cb(tree, state, bits);
423
	state->state &= ~bits_to_clear;
424 425
	if (wake)
		wake_up(&state->wq);
426
	if (state->state == 0) {
427
		if (state->tree) {
428
			rb_erase(&state->rb_node, &tree->state);
429
			state->tree = NULL;
430 431 432 433 434 435 436 437 438 439
			free_extent_state(state);
		} else {
			WARN_ON(1);
		}
	} else {
		merge_state(tree, state);
	}
	return ret;
}

440 441 442 443 444 445 446 447 448
static struct extent_state *
alloc_extent_state_atomic(struct extent_state *prealloc)
{
	if (!prealloc)
		prealloc = alloc_extent_state(GFP_ATOMIC);

	return prealloc;
}

449 450 451 452 453 454 455 456 457 458 459 460 461 462
/*
 * 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,
463 464 465
		     int bits, int wake, int delete,
		     struct extent_state **cached_state,
		     gfp_t mask)
466 467
{
	struct extent_state *state;
468
	struct extent_state *cached;
469
	struct extent_state *prealloc = NULL;
470
	struct rb_node *next_node;
471
	struct rb_node *node;
472
	u64 last_end;
473 474
	int err;
	int set = 0;
475
	int clear = 0;
476

477 478 479 480
	if (delete)
		bits |= ~EXTENT_CTLBITS;
	bits |= EXTENT_FIRST_DELALLOC;

481 482
	if (bits & (EXTENT_IOBITS | EXTENT_BOUNDARY))
		clear = 1;
483 484 485
again:
	if (!prealloc && (mask & __GFP_WAIT)) {
		prealloc = alloc_extent_state(mask);
486 487
		if (!prealloc)
			return -ENOMEM;
488 489
	}

490
	spin_lock(&tree->lock);
491 492
	if (cached_state) {
		cached = *cached_state;
493 494 495 496 497 498

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

499 500
		if (cached && cached->tree && cached->start <= start &&
		    cached->end > start) {
501 502
			if (clear)
				atomic_dec(&cached->refs);
503
			state = cached;
504
			goto hit_next;
505
		}
506 507
		if (clear)
			free_extent_state(cached);
508
	}
509 510 511 512
	/*
	 * this search will find the extents that end after
	 * our range starts
	 */
513
	node = tree_search(tree, start);
514 515 516
	if (!node)
		goto out;
	state = rb_entry(node, struct extent_state, rb_node);
517
hit_next:
518 519 520
	if (state->start > end)
		goto out;
	WARN_ON(state->end < start);
521
	last_end = state->end;
522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539

	/*
	 *     | ---- 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) {
540 541
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
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) {
548
			set |= clear_state_bit(tree, state, &bits, wake);
549 550 551
			if (last_end == (u64)-1)
				goto out;
			start = last_end + 1;
552 553 554 555 556 557 558 559 560 561
		}
		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) {
562 563
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
564 565 566 567
		err = split_state(tree, state, prealloc, end + 1);
		BUG_ON(err == -EEXIST);
		if (wake)
			wake_up(&state->wq);
568

569
		set |= clear_state_bit(tree, prealloc, &bits, wake);
J
Josef Bacik 已提交
570

571 572 573
		prealloc = NULL;
		goto out;
	}
574

575 576 577 578
	if (state->end < end && prealloc && !need_resched())
		next_node = rb_next(&state->rb_node);
	else
		next_node = NULL;
579

580
	set |= clear_state_bit(tree, state, &bits, wake);
581 582 583
	if (last_end == (u64)-1)
		goto out;
	start = last_end + 1;
584 585 586 587 588 589
	if (start <= end && next_node) {
		state = rb_entry(next_node, struct extent_state,
				 rb_node);
		if (state->start == start)
			goto hit_next;
	}
590 591 592
	goto search_again;

out:
593
	spin_unlock(&tree->lock);
594 595 596 597 598 599 600 601
	if (prealloc)
		free_extent_state(prealloc);

	return set;

search_again:
	if (start > end)
		goto out;
602
	spin_unlock(&tree->lock);
603 604 605 606 607 608 609
	if (mask & __GFP_WAIT)
		cond_resched();
	goto again;
}

static int wait_on_state(struct extent_io_tree *tree,
			 struct extent_state *state)
610 611
		__releases(tree->lock)
		__acquires(tree->lock)
612 613 614
{
	DEFINE_WAIT(wait);
	prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE);
615
	spin_unlock(&tree->lock);
616
	schedule();
617
	spin_lock(&tree->lock);
618 619 620 621 622 623 624 625 626 627 628 629 630 631
	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;

632
	spin_lock(&tree->lock);
633 634 635 636 637 638
again:
	while (1) {
		/*
		 * this search will find all the extents that end after
		 * our range starts
		 */
639
		node = tree_search(tree, start);
640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660
		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()) {
661
			spin_unlock(&tree->lock);
662
			cond_resched();
663
			spin_lock(&tree->lock);
664 665 666
		}
	}
out:
667
	spin_unlock(&tree->lock);
668 669 670
	return 0;
}

J
Josef Bacik 已提交
671
static int set_state_bits(struct extent_io_tree *tree,
672
			   struct extent_state *state,
673
			   int *bits)
674
{
J
Josef Bacik 已提交
675
	int ret;
676
	int bits_to_set = *bits & ~EXTENT_CTLBITS;
J
Josef Bacik 已提交
677 678 679 680

	ret = set_state_cb(tree, state, bits);
	if (ret)
		return ret;
681
	if ((bits_to_set & EXTENT_DIRTY) && !(state->state & EXTENT_DIRTY)) {
682 683 684
		u64 range = state->end - state->start + 1;
		tree->dirty_bytes += range;
	}
685
	state->state |= bits_to_set;
J
Josef Bacik 已提交
686 687

	return 0;
688 689
}

690 691 692 693 694 695 696 697 698 699 700
static void cache_state(struct extent_state *state,
			struct extent_state **cached_ptr)
{
	if (cached_ptr && !(*cached_ptr)) {
		if (state->state & (EXTENT_IOBITS | EXTENT_BOUNDARY)) {
			*cached_ptr = state;
			atomic_inc(&state->refs);
		}
	}
}

701 702 703 704
static void uncache_state(struct extent_state **cached_ptr)
{
	if (cached_ptr && (*cached_ptr)) {
		struct extent_state *state = *cached_ptr;
705 706
		*cached_ptr = NULL;
		free_extent_state(state);
707 708 709
	}
}

710
/*
711 712
 * 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.
713
 *
714 715 716
 * If any of the exclusive bits are set, this will fail with -EEXIST if some
 * part of the range already has the desired bits set.  The start of the
 * existing range is returned in failed_start in this case.
717
 *
718
 * [start, end] is inclusive This takes the tree lock.
719
 */
720

721 722 723
int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
		   int bits, int exclusive_bits, u64 *failed_start,
		   struct extent_state **cached_state, gfp_t mask)
724 725 726 727 728 729 730
{
	struct extent_state *state;
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
	int err = 0;
	u64 last_start;
	u64 last_end;
731

732
	bits |= EXTENT_FIRST_DELALLOC;
733 734 735
again:
	if (!prealloc && (mask & __GFP_WAIT)) {
		prealloc = alloc_extent_state(mask);
736
		BUG_ON(!prealloc);
737 738
	}

739
	spin_lock(&tree->lock);
740 741
	if (cached_state && *cached_state) {
		state = *cached_state;
742 743
		if (state->start <= start && state->end > start &&
		    state->tree) {
744 745 746 747
			node = &state->rb_node;
			goto hit_next;
		}
	}
748 749 750 751
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
752
	node = tree_search(tree, start);
753
	if (!node) {
754 755
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
756
		err = insert_state(tree, prealloc, start, end, &bits);
757 758 759 760 761
		prealloc = NULL;
		BUG_ON(err == -EEXIST);
		goto out;
	}
	state = rb_entry(node, struct extent_state, rb_node);
C
Chris Mason 已提交
762
hit_next:
763 764 765 766 767 768 769 770 771 772
	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) {
C
Chris Mason 已提交
773
		struct rb_node *next_node;
774
		if (state->state & exclusive_bits) {
775 776 777 778
			*failed_start = state->start;
			err = -EEXIST;
			goto out;
		}
779

780
		err = set_state_bits(tree, state, &bits);
J
Josef Bacik 已提交
781 782 783
		if (err)
			goto out;

784
		cache_state(state, cached_state);
785
		merge_state(tree, state);
786 787
		if (last_end == (u64)-1)
			goto out;
C
Chris Mason 已提交
788

789
		start = last_end + 1;
790
		next_node = rb_next(&state->rb_node);
791 792 793 794 795
		if (next_node && start < end && prealloc && !need_resched()) {
			state = rb_entry(next_node, struct extent_state,
					 rb_node);
			if (state->start == start)
				goto hit_next;
C
Chris Mason 已提交
796
		}
797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816
		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) {
817
		if (state->state & exclusive_bits) {
818 819 820 821
			*failed_start = start;
			err = -EEXIST;
			goto out;
		}
822 823 824

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
825 826 827 828 829 830
		err = split_state(tree, state, prealloc, start);
		BUG_ON(err == -EEXIST);
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
831
			err = set_state_bits(tree, state, &bits);
J
Josef Bacik 已提交
832 833
			if (err)
				goto out;
834
			cache_state(state, cached_state);
835
			merge_state(tree, state);
836 837 838
			if (last_end == (u64)-1)
				goto out;
			start = last_end + 1;
839 840 841 842 843 844 845 846 847 848 849 850 851 852 853
		}
		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
854
			this_end = last_start - 1;
855 856 857

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
858 859 860 861 862

		/*
		 * Avoid to free 'prealloc' if it can be merged with
		 * the later extent.
		 */
863
		err = insert_state(tree, prealloc, start, this_end,
864
				   &bits);
865
		BUG_ON(err == -EEXIST);
J
Josef Bacik 已提交
866
		if (err) {
867
			free_extent_state(prealloc);
J
Josef Bacik 已提交
868
			prealloc = NULL;
869
			goto out;
J
Josef Bacik 已提交
870 871 872
		}
		cache_state(prealloc, cached_state);
		prealloc = NULL;
873 874 875 876 877 878 879 880 881 882
		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) {
883
		if (state->state & exclusive_bits) {
884 885 886 887
			*failed_start = start;
			err = -EEXIST;
			goto out;
		}
888 889 890

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
891 892 893
		err = split_state(tree, state, prealloc, end + 1);
		BUG_ON(err == -EEXIST);

894
		err = set_state_bits(tree, prealloc, &bits);
J
Josef Bacik 已提交
895 896 897 898
		if (err) {
			prealloc = NULL;
			goto out;
		}
899
		cache_state(prealloc, cached_state);
900 901 902 903 904 905 906 907
		merge_state(tree, prealloc);
		prealloc = NULL;
		goto out;
	}

	goto search_again;

out:
908
	spin_unlock(&tree->lock);
909 910 911 912 913 914 915 916
	if (prealloc)
		free_extent_state(prealloc);

	return err;

search_again:
	if (start > end)
		goto out;
917
	spin_unlock(&tree->lock);
918 919 920 921 922 923 924 925 926 927
	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,
928
			      NULL, mask);
929 930 931 932 933 934
}

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,
935
			      NULL, mask);
936 937 938 939 940
}

int clear_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
		      int bits, gfp_t mask)
{
941
	return clear_extent_bit(tree, start, end, bits, 0, 0, NULL, mask);
942 943 944
}

int set_extent_delalloc(struct extent_io_tree *tree, u64 start, u64 end,
945
			struct extent_state **cached_state, gfp_t mask)
946 947
{
	return set_extent_bit(tree, start, end,
C
Chris Mason 已提交
948
			      EXTENT_DELALLOC | EXTENT_DIRTY | EXTENT_UPTODATE,
949
			      0, NULL, cached_state, mask);
950 951 952 953 954 955
}

int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
		       gfp_t mask)
{
	return clear_extent_bit(tree, start, end,
956
				EXTENT_DIRTY | EXTENT_DELALLOC |
957
				EXTENT_DO_ACCOUNTING, 0, 0, NULL, mask);
958 959 960 961 962 963
}

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,
964
			      NULL, mask);
965 966 967
}

int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end,
968
			struct extent_state **cached_state, gfp_t mask)
969
{
970 971
	return set_extent_bit(tree, start, end, EXTENT_UPTODATE, 0,
			      NULL, cached_state, mask);
972 973
}

C
Chris Mason 已提交
974
static int clear_extent_uptodate(struct extent_io_tree *tree, u64 start,
975 976
				 u64 end, struct extent_state **cached_state,
				 gfp_t mask)
977
{
978
	return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0,
979
				cached_state, mask);
980 981
}

C
Chris Mason 已提交
982 983 984 985
/*
 * either insert or lock state struct between start and end use mask to tell
 * us if waiting is desired.
 */
986
int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
987
		     int bits, struct extent_state **cached_state, gfp_t mask)
988 989 990 991
{
	int err;
	u64 failed_start;
	while (1) {
992
		err = set_extent_bit(tree, start, end, EXTENT_LOCKED | bits,
993 994
				     EXTENT_LOCKED, &failed_start,
				     cached_state, mask);
995 996 997 998 999 1000 1001 1002 1003 1004 1005
		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;
}

1006 1007
int lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask)
{
1008
	return lock_extent_bits(tree, start, end, 0, NULL, mask);
1009 1010
}

1011 1012 1013 1014 1015 1016
int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end,
		    gfp_t mask)
{
	int err;
	u64 failed_start;

1017 1018
	err = set_extent_bit(tree, start, end, EXTENT_LOCKED, EXTENT_LOCKED,
			     &failed_start, NULL, mask);
Y
Yan Zheng 已提交
1019 1020 1021
	if (err == -EEXIST) {
		if (failed_start > start)
			clear_extent_bit(tree, start, failed_start - 1,
1022
					 EXTENT_LOCKED, 1, 0, NULL, mask);
1023
		return 0;
Y
Yan Zheng 已提交
1024
	}
1025 1026 1027
	return 1;
}

1028 1029 1030 1031 1032 1033 1034
int unlock_extent_cached(struct extent_io_tree *tree, u64 start, u64 end,
			 struct extent_state **cached, gfp_t mask)
{
	return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, cached,
				mask);
}

1035
int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask)
1036
{
1037 1038
	return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL,
				mask);
1039 1040 1041 1042 1043
}

/*
 * helper function to set both pages and extents in the tree writeback
 */
1044
static int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end)
1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059
{
	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++;
	}
	return 0;
}

C
Chris Mason 已提交
1060 1061 1062 1063 1064 1065 1066
/*
 * 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
 */
1067 1068 1069 1070 1071 1072 1073
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;

1074
	spin_lock(&tree->lock);
1075 1076 1077 1078
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1079
	node = tree_search(tree, start);
C
Chris Mason 已提交
1080
	if (!node)
1081 1082
		goto out;

C
Chris Mason 已提交
1083
	while (1) {
1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095
		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:
1096
	spin_unlock(&tree->lock);
1097 1098 1099
	return ret;
}

C
Chris Mason 已提交
1100 1101 1102 1103
/* 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
1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114
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
1115
	if (!node)
C
Chris Mason 已提交
1116 1117
		goto out;

C
Chris Mason 已提交
1118
	while (1) {
C
Chris Mason 已提交
1119
		state = rb_entry(node, struct extent_state, rb_node);
C
Chris Mason 已提交
1120
		if (state->end >= start && (state->state & bits))
C
Chris Mason 已提交
1121
			return state;
C
Chris Mason 已提交
1122

C
Chris Mason 已提交
1123 1124 1125 1126 1127 1128 1129 1130
		node = rb_next(node);
		if (!node)
			break;
	}
out:
	return NULL;
}

C
Chris Mason 已提交
1131 1132 1133 1134 1135 1136
/*
 * 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
1137
static noinline u64 find_delalloc_range(struct extent_io_tree *tree,
1138 1139
					u64 *start, u64 *end, u64 max_bytes,
					struct extent_state **cached_state)
1140 1141 1142 1143 1144 1145 1146
{
	struct rb_node *node;
	struct extent_state *state;
	u64 cur_start = *start;
	u64 found = 0;
	u64 total_bytes = 0;

1147
	spin_lock(&tree->lock);
C
Chris Mason 已提交
1148

1149 1150 1151 1152
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1153
	node = tree_search(tree, cur_start);
1154
	if (!node) {
1155 1156
		if (!found)
			*end = (u64)-1;
1157 1158 1159
		goto out;
	}

C
Chris Mason 已提交
1160
	while (1) {
1161
		state = rb_entry(node, struct extent_state, rb_node);
1162 1163
		if (found && (state->start != cur_start ||
			      (state->state & EXTENT_BOUNDARY))) {
1164 1165 1166 1167 1168 1169 1170
			goto out;
		}
		if (!(state->state & EXTENT_DELALLOC)) {
			if (!found)
				*end = state->end;
			goto out;
		}
1171
		if (!found) {
1172
			*start = state->start;
1173 1174 1175
			*cached_state = state;
			atomic_inc(&state->refs);
		}
1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186
		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:
1187
	spin_unlock(&tree->lock);
1188 1189 1190
	return found;
}

C
Chris Mason 已提交
1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204
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
1205
	while (nr_pages > 0) {
C
Chris Mason 已提交
1206
		ret = find_get_pages_contig(inode->i_mapping, index,
1207 1208
				     min_t(unsigned long, nr_pages,
				     ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
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 1240
		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
1241
	while (nrpages > 0) {
C
Chris Mason 已提交
1242
		ret = find_get_pages_contig(inode->i_mapping, index,
1243 1244
				     min_t(unsigned long,
				     nrpages, ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1245 1246 1247 1248 1249 1250 1251 1252 1253 1254
		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
			 */
1255
			if (pages[i] != locked_page) {
C
Chris Mason 已提交
1256
				lock_page(pages[i]);
1257 1258
				if (!PageDirty(pages[i]) ||
				    pages[i]->mapping != inode->i_mapping) {
1259 1260 1261 1262 1263 1264
					ret = -EAGAIN;
					unlock_page(pages[i]);
					page_cache_release(pages[i]);
					goto done;
				}
			}
C
Chris Mason 已提交
1265
			page_cache_release(pages[i]);
1266
			pages_locked++;
C
Chris Mason 已提交
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
		}
		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;
1298
	struct extent_state *cached_state = NULL;
C
Chris Mason 已提交
1299 1300 1301 1302 1303 1304 1305 1306
	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,
1307
				    max_bytes, &cached_state);
C
Chris Mason 已提交
1308
	if (!found || delalloc_end <= *start) {
C
Chris Mason 已提交
1309 1310
		*start = delalloc_start;
		*end = delalloc_end;
1311
		free_extent_state(cached_state);
C
Chris Mason 已提交
1312 1313 1314
		return found;
	}

C
Chris Mason 已提交
1315 1316 1317 1318 1319
	/*
	 * 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
1320
	if (delalloc_start < *start)
C
Chris Mason 已提交
1321 1322
		delalloc_start = *start;

C
Chris Mason 已提交
1323 1324 1325 1326
	/*
	 * make sure to limit the number of pages we try to lock down
	 * if we're looping.
	 */
C
Chris Mason 已提交
1327
	if (delalloc_end + 1 - delalloc_start > max_bytes && loops)
1328
		delalloc_end = delalloc_start + PAGE_CACHE_SIZE - 1;
C
Chris Mason 已提交
1329

C
Chris Mason 已提交
1330 1331 1332 1333 1334 1335 1336
	/* 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
		 */
1337
		free_extent_state(cached_state);
C
Chris Mason 已提交
1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350
		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 */
1351 1352
	lock_extent_bits(tree, delalloc_start, delalloc_end,
			 0, &cached_state, GFP_NOFS);
C
Chris Mason 已提交
1353 1354 1355

	/* then test to make sure it is all still delalloc */
	ret = test_range_bit(tree, delalloc_start, delalloc_end,
1356
			     EXTENT_DELALLOC, 1, cached_state);
C
Chris Mason 已提交
1357
	if (!ret) {
1358 1359
		unlock_extent_cached(tree, delalloc_start, delalloc_end,
				     &cached_state, GFP_NOFS);
C
Chris Mason 已提交
1360 1361 1362 1363 1364
		__unlock_for_delalloc(inode, locked_page,
			      delalloc_start, delalloc_end);
		cond_resched();
		goto again;
	}
1365
	free_extent_state(cached_state);
C
Chris Mason 已提交
1366 1367 1368 1369 1370 1371 1372 1373 1374
	*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,
1375
				unsigned long op)
C
Chris Mason 已提交
1376 1377 1378 1379 1380 1381 1382
{
	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;
1383
	int clear_bits = 0;
C
Chris Mason 已提交
1384

1385
	if (op & EXTENT_CLEAR_UNLOCK)
1386
		clear_bits |= EXTENT_LOCKED;
1387
	if (op & EXTENT_CLEAR_DIRTY)
C
Chris Mason 已提交
1388 1389
		clear_bits |= EXTENT_DIRTY;

1390
	if (op & EXTENT_CLEAR_DELALLOC)
1391 1392
		clear_bits |= EXTENT_DELALLOC;

1393
	clear_extent_bit(tree, start, end, clear_bits, 1, 0, NULL, GFP_NOFS);
1394 1395 1396
	if (!(op & (EXTENT_CLEAR_UNLOCK_PAGE | EXTENT_CLEAR_DIRTY |
		    EXTENT_SET_WRITEBACK | EXTENT_END_WRITEBACK |
		    EXTENT_SET_PRIVATE2)))
1397
		return 0;
C
Chris Mason 已提交
1398

C
Chris Mason 已提交
1399
	while (nr_pages > 0) {
C
Chris Mason 已提交
1400
		ret = find_get_pages_contig(inode->i_mapping, index,
1401 1402
				     min_t(unsigned long,
				     nr_pages, ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1403
		for (i = 0; i < ret; i++) {
1404

1405
			if (op & EXTENT_SET_PRIVATE2)
1406 1407
				SetPagePrivate2(pages[i]);

C
Chris Mason 已提交
1408 1409 1410 1411
			if (pages[i] == locked_page) {
				page_cache_release(pages[i]);
				continue;
			}
1412
			if (op & EXTENT_CLEAR_DIRTY)
C
Chris Mason 已提交
1413
				clear_page_dirty_for_io(pages[i]);
1414
			if (op & EXTENT_SET_WRITEBACK)
C
Chris Mason 已提交
1415
				set_page_writeback(pages[i]);
1416
			if (op & EXTENT_END_WRITEBACK)
C
Chris Mason 已提交
1417
				end_page_writeback(pages[i]);
1418
			if (op & EXTENT_CLEAR_UNLOCK_PAGE)
1419
				unlock_page(pages[i]);
C
Chris Mason 已提交
1420 1421 1422 1423 1424 1425 1426 1427 1428
			page_cache_release(pages[i]);
		}
		nr_pages -= ret;
		index += ret;
		cond_resched();
	}
	return 0;
}

C
Chris Mason 已提交
1429 1430 1431 1432 1433
/*
 * 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.
 */
1434 1435
u64 count_range_bits(struct extent_io_tree *tree,
		     u64 *start, u64 search_end, u64 max_bytes,
1436
		     unsigned long bits, int contig)
1437 1438 1439 1440 1441
{
	struct rb_node *node;
	struct extent_state *state;
	u64 cur_start = *start;
	u64 total_bytes = 0;
1442
	u64 last = 0;
1443 1444 1445 1446 1447 1448 1449
	int found = 0;

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

1450
	spin_lock(&tree->lock);
1451 1452 1453 1454 1455 1456 1457 1458
	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.
	 */
1459
	node = tree_search(tree, cur_start);
C
Chris Mason 已提交
1460
	if (!node)
1461 1462
		goto out;

C
Chris Mason 已提交
1463
	while (1) {
1464 1465 1466
		state = rb_entry(node, struct extent_state, rb_node);
		if (state->start > search_end)
			break;
1467 1468 1469
		if (contig && found && state->start > last + 1)
			break;
		if (state->end >= cur_start && (state->state & bits) == bits) {
1470 1471 1472 1473 1474
			total_bytes += min(search_end, state->end) + 1 -
				       max(cur_start, state->start);
			if (total_bytes >= max_bytes)
				break;
			if (!found) {
1475
				*start = max(cur_start, state->start);
1476 1477
				found = 1;
			}
1478 1479 1480
			last = state->end;
		} else if (contig && found) {
			break;
1481 1482 1483 1484 1485 1486
		}
		node = rb_next(node);
		if (!node)
			break;
	}
out:
1487
	spin_unlock(&tree->lock);
1488 1489
	return total_bytes;
}
1490

C
Chris Mason 已提交
1491 1492 1493 1494
/*
 * set the private field for a given byte offset in the tree.  If there isn't
 * an extent_state there already, this does nothing.
 */
1495 1496 1497 1498 1499 1500
int set_state_private(struct extent_io_tree *tree, u64 start, u64 private)
{
	struct rb_node *node;
	struct extent_state *state;
	int ret = 0;

1501
	spin_lock(&tree->lock);
1502 1503 1504 1505
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1506
	node = tree_search(tree, start);
1507
	if (!node) {
1508 1509 1510 1511 1512 1513 1514 1515 1516 1517
		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:
1518
	spin_unlock(&tree->lock);
1519 1520 1521 1522 1523 1524 1525 1526 1527
	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;

1528
	spin_lock(&tree->lock);
1529 1530 1531 1532
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1533
	node = tree_search(tree, start);
1534
	if (!node) {
1535 1536 1537 1538 1539 1540 1541 1542 1543 1544
		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:
1545
	spin_unlock(&tree->lock);
1546 1547 1548 1549 1550
	return ret;
}

/*
 * searches a range in the state tree for a given mask.
1551
 * If 'filled' == 1, this returns 1 only if every extent in the tree
1552 1553 1554 1555
 * 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,
1556
		   int bits, int filled, struct extent_state *cached)
1557 1558 1559 1560 1561
{
	struct extent_state *state = NULL;
	struct rb_node *node;
	int bitset = 0;

1562
	spin_lock(&tree->lock);
1563 1564
	if (cached && cached->tree && cached->start <= start &&
	    cached->end > start)
1565 1566 1567
		node = &cached->rb_node;
	else
		node = tree_search(tree, start);
1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586
	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;
		}
1587 1588 1589 1590

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

1591 1592 1593 1594 1595 1596 1597 1598 1599 1600
		start = state->end + 1;
		if (start > end)
			break;
		node = rb_next(node);
		if (!node) {
			if (filled)
				bitset = 0;
			break;
		}
	}
1601
	spin_unlock(&tree->lock);
1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613
	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;
1614
	if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1, NULL))
1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627
		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;
1628
	if (!test_range_bit(tree, start, end, EXTENT_LOCKED, 0, NULL))
1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639
		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)
{
1640
	end_page_writeback(page);
1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656
	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)
{
1657
	int uptodate = err == 0;
1658
	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
1659
	struct extent_io_tree *tree;
1660 1661 1662
	u64 start;
	u64 end;
	int whole_page;
1663
	int ret;
1664 1665 1666

	do {
		struct page *page = bvec->bv_page;
1667 1668
		tree = &BTRFS_I(page->mapping->host)->io_tree;

1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679
		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);
1680 1681
		if (tree->ops && tree->ops->writepage_end_io_hook) {
			ret = tree->ops->writepage_end_io_hook(page, start,
1682
						       end, NULL, uptodate);
1683 1684 1685 1686 1687 1688 1689
			if (ret)
				uptodate = 0;
		}

		if (!uptodate && tree->ops &&
		    tree->ops->writepage_io_failed_hook) {
			ret = tree->ops->writepage_io_failed_hook(bio, page,
1690
							 start, end, NULL);
1691 1692 1693 1694 1695 1696
			if (ret == 0) {
				uptodate = (err == 0);
				continue;
			}
		}

1697
		if (!uptodate) {
1698
			clear_extent_uptodate(tree, start, end, NULL, GFP_NOFS);
1699 1700 1701
			ClearPageUptodate(page);
			SetPageError(page);
		}
1702

1703 1704 1705 1706 1707
		if (whole_page)
			end_page_writeback(page);
		else
			check_page_writeback(tree, page);
	} while (bvec >= bio->bi_io_vec);
1708

1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725
	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);
1726 1727
	struct bio_vec *bvec_end = bio->bi_io_vec + bio->bi_vcnt - 1;
	struct bio_vec *bvec = bio->bi_io_vec;
1728
	struct extent_io_tree *tree;
1729 1730 1731 1732 1733
	u64 start;
	u64 end;
	int whole_page;
	int ret;

1734 1735 1736
	if (err)
		uptodate = 0;

1737 1738
	do {
		struct page *page = bvec->bv_page;
1739 1740 1741
		struct extent_state *cached = NULL;
		struct extent_state *state;

1742 1743
		tree = &BTRFS_I(page->mapping->host)->io_tree;

1744 1745 1746 1747 1748 1749 1750 1751 1752
		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;

1753
		if (++bvec <= bvec_end)
1754 1755
			prefetchw(&bvec->bv_page->flags);

1756
		spin_lock(&tree->lock);
1757
		state = find_first_extent_bit_state(tree, start, EXTENT_LOCKED);
1758
		if (state && state->start == start) {
1759 1760 1761 1762 1763 1764 1765 1766
			/*
			 * take a reference on the state, unlock will drop
			 * the ref
			 */
			cache_state(state, &cached);
		}
		spin_unlock(&tree->lock);

1767
		if (uptodate && tree->ops && tree->ops->readpage_end_io_hook) {
1768
			ret = tree->ops->readpage_end_io_hook(page, start, end,
1769
							      state);
1770 1771 1772
			if (ret)
				uptodate = 0;
		}
1773 1774 1775
		if (!uptodate && tree->ops &&
		    tree->ops->readpage_io_failed_hook) {
			ret = tree->ops->readpage_io_failed_hook(bio, page,
1776
							 start, end, NULL);
1777
			if (ret == 0) {
1778 1779
				uptodate =
					test_bit(BIO_UPTODATE, &bio->bi_flags);
1780 1781
				if (err)
					uptodate = 0;
1782
				uncache_state(&cached);
1783 1784 1785
				continue;
			}
		}
1786

1787
		if (uptodate) {
1788
			set_extent_uptodate(tree, start, end, &cached,
1789
					    GFP_ATOMIC);
1790
		}
1791
		unlock_extent_cached(tree, start, end, &cached, GFP_ATOMIC);
1792

1793 1794 1795 1796 1797 1798 1799
		if (whole_page) {
			if (uptodate) {
				SetPageUptodate(page);
			} else {
				ClearPageUptodate(page);
				SetPageError(page);
			}
1800
			unlock_page(page);
1801 1802 1803 1804 1805 1806 1807
		} else {
			if (uptodate) {
				check_page_uptodate(tree, page);
			} else {
				ClearPageUptodate(page);
				SetPageError(page);
			}
1808
			check_page_locked(tree, page);
1809
		}
1810
	} while (bvec <= bvec_end);
1811 1812 1813 1814

	bio_put(bio);
}

1815 1816 1817
struct bio *
btrfs_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
		gfp_t gfp_flags)
1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828
{
	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) {
1829
		bio->bi_size = 0;
1830 1831 1832 1833 1834 1835
		bio->bi_bdev = bdev;
		bio->bi_sector = first_sector;
	}
	return bio;
}

C
Chris Mason 已提交
1836 1837
static int submit_one_bio(int rw, struct bio *bio, int mirror_num,
			  unsigned long bio_flags)
1838 1839
{
	int ret = 0;
1840 1841 1842 1843 1844 1845 1846
	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;

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

1847
	bio->bi_private = NULL;
1848 1849 1850

	bio_get(bio);

1851
	if (tree->ops && tree->ops->submit_bio_hook)
1852
		ret = tree->ops->submit_bio_hook(page->mapping->host, rw, bio,
1853
					   mirror_num, bio_flags, start);
1854 1855
	else
		submit_bio(rw, bio);
1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867
	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,
1868
			      bio_end_io_t end_io_func,
C
Chris Mason 已提交
1869 1870 1871
			      int mirror_num,
			      unsigned long prev_bio_flags,
			      unsigned long bio_flags)
1872 1873 1874 1875
{
	int ret = 0;
	struct bio *bio;
	int nr;
C
Chris Mason 已提交
1876 1877 1878
	int contig = 0;
	int this_compressed = bio_flags & EXTENT_BIO_COMPRESSED;
	int old_compressed = prev_bio_flags & EXTENT_BIO_COMPRESSED;
1879
	size_t page_size = min_t(size_t, size, PAGE_CACHE_SIZE);
1880 1881 1882

	if (bio_ret && *bio_ret) {
		bio = *bio_ret;
C
Chris Mason 已提交
1883 1884 1885 1886 1887 1888 1889
		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 ||
1890
		    (tree->ops && tree->ops->merge_bio_hook &&
C
Chris Mason 已提交
1891 1892 1893 1894 1895
		     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);
1896 1897 1898 1899 1900
			bio = NULL;
		} else {
			return 0;
		}
	}
C
Chris Mason 已提交
1901 1902 1903 1904 1905
	if (this_compressed)
		nr = BIO_MAX_PAGES;
	else
		nr = bio_get_nr_vecs(bdev);

1906
	bio = btrfs_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH);
1907 1908
	if (!bio)
		return -ENOMEM;
1909

C
Chris Mason 已提交
1910
	bio_add_page(bio, page, page_size, offset);
1911 1912
	bio->bi_end_io = end_io_func;
	bio->bi_private = tree;
1913

C
Chris Mason 已提交
1914
	if (bio_ret)
1915
		*bio_ret = bio;
C
Chris Mason 已提交
1916
	else
C
Chris Mason 已提交
1917
		ret = submit_one_bio(rw, bio, mirror_num, bio_flags);
1918 1919 1920 1921 1922 1923 1924 1925 1926

	return ret;
}

void set_page_extent_mapped(struct page *page)
{
	if (!PagePrivate(page)) {
		SetPagePrivate(page);
		page_cache_get(page);
1927
		set_page_private(page, EXTENT_PAGE_PRIVATE);
1928 1929 1930
	}
}

1931
static void set_page_extent_head(struct page *page, unsigned long len)
1932
{
C
Chris Mason 已提交
1933
	WARN_ON(!PagePrivate(page));
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
{
	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;
1960
	struct btrfs_ordered_extent *ordered;
1961 1962
	int ret;
	int nr = 0;
1963
	size_t pg_offset = 0;
1964
	size_t iosize;
C
Chris Mason 已提交
1965
	size_t disk_io_size;
1966
	size_t blocksize = inode->i_sb->s_blocksize;
C
Chris Mason 已提交
1967
	unsigned long this_bio_flag = 0;
1968 1969 1970 1971

	set_page_extent_mapped(page);

	end = page_end;
1972 1973 1974 1975 1976 1977 1978 1979 1980
	while (1) {
		lock_extent(tree, start, end, GFP_NOFS);
		ordered = btrfs_lookup_ordered_extent(inode, start);
		if (!ordered)
			break;
		unlock_extent(tree, start, end, GFP_NOFS);
		btrfs_start_ordered_extent(inode, ordered, 1);
		btrfs_put_ordered_extent(ordered);
	}
1981

C
Chris Mason 已提交
1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993
	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);
		}
	}
1994 1995 1996
	while (cur <= end) {
		if (cur >= last_byte) {
			char *userpage;
1997 1998
			struct extent_state *cached = NULL;

1999
			iosize = PAGE_CACHE_SIZE - pg_offset;
2000
			userpage = kmap_atomic(page, KM_USER0);
2001
			memset(userpage + pg_offset, 0, iosize);
2002 2003 2004
			flush_dcache_page(page);
			kunmap_atomic(userpage, KM_USER0);
			set_extent_uptodate(tree, cur, cur + iosize - 1,
2005 2006 2007
					    &cached, GFP_NOFS);
			unlock_extent_cached(tree, cur, cur + iosize - 1,
					     &cached, GFP_NOFS);
2008 2009
			break;
		}
2010
		em = get_extent(inode, page, pg_offset, cur,
2011
				end - cur + 1, 0);
2012
		if (IS_ERR_OR_NULL(em)) {
2013 2014 2015 2016 2017 2018 2019 2020
			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);

2021
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
C
Chris Mason 已提交
2022
			this_bio_flag = EXTENT_BIO_COMPRESSED;
2023 2024 2025
			extent_set_compress_type(&this_bio_flag,
						 em->compress_type);
		}
C
Chris Mason 已提交
2026

2027 2028 2029
		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
2030 2031 2032 2033 2034 2035 2036
		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;
		}
2037 2038
		bdev = em->bdev;
		block_start = em->block_start;
Y
Yan Zheng 已提交
2039 2040
		if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			block_start = EXTENT_MAP_HOLE;
2041 2042 2043 2044 2045 2046
		free_extent_map(em);
		em = NULL;

		/* we've found a hole, just zero and go on */
		if (block_start == EXTENT_MAP_HOLE) {
			char *userpage;
2047 2048
			struct extent_state *cached = NULL;

2049
			userpage = kmap_atomic(page, KM_USER0);
2050
			memset(userpage + pg_offset, 0, iosize);
2051 2052 2053 2054
			flush_dcache_page(page);
			kunmap_atomic(userpage, KM_USER0);

			set_extent_uptodate(tree, cur, cur + iosize - 1,
2055 2056 2057
					    &cached, GFP_NOFS);
			unlock_extent_cached(tree, cur, cur + iosize - 1,
			                     &cached, GFP_NOFS);
2058
			cur = cur + iosize;
2059
			pg_offset += iosize;
2060 2061 2062
			continue;
		}
		/* the get_extent function already copied into the page */
2063 2064
		if (test_range_bit(tree, cur, cur_end,
				   EXTENT_UPTODATE, 1, NULL)) {
2065
			check_page_uptodate(tree, page);
2066 2067
			unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
			cur = cur + iosize;
2068
			pg_offset += iosize;
2069 2070
			continue;
		}
2071 2072 2073 2074 2075 2076 2077
		/* 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;
2078
			pg_offset += iosize;
2079 2080
			continue;
		}
2081 2082 2083 2084 2085 2086 2087

		ret = 0;
		if (tree->ops && tree->ops->readpage_io_hook) {
			ret = tree->ops->readpage_io_hook(page, cur,
							  cur + iosize - 1);
		}
		if (!ret) {
2088 2089
			unsigned long pnr = (last_byte >> PAGE_CACHE_SHIFT) + 1;
			pnr -= page->index;
2090
			ret = submit_extent_page(READ, tree, page,
2091
					 sector, disk_io_size, pg_offset,
2092
					 bdev, bio, pnr,
C
Chris Mason 已提交
2093 2094 2095
					 end_bio_extent_readpage, mirror_num,
					 *bio_flags,
					 this_bio_flag);
2096
			nr++;
C
Chris Mason 已提交
2097
			*bio_flags = this_bio_flag;
2098 2099 2100 2101
		}
		if (ret)
			SetPageError(page);
		cur = cur + iosize;
2102
		pg_offset += iosize;
2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115
	}
	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
2116
	unsigned long bio_flags = 0;
2117 2118
	int ret;

C
Chris Mason 已提交
2119 2120
	ret = __extent_read_full_page(tree, page, get_extent, &bio, 0,
				      &bio_flags);
2121
	if (bio)
2122
		ret = submit_one_bio(READ, bio, 0, bio_flags);
2123 2124 2125
	return ret;
}

2126 2127 2128 2129 2130 2131 2132 2133 2134 2135
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;
}

2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157
/*
 * 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;
	sector_t sector;
2158
	struct extent_state *cached_state = NULL;
2159 2160 2161 2162
	struct extent_map *em;
	struct block_device *bdev;
	int ret;
	int nr = 0;
2163
	size_t pg_offset = 0;
2164 2165 2166 2167 2168
	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
2169 2170
	int page_started;
	int compressed;
2171
	int write_flags;
2172
	unsigned long nr_written = 0;
2173

2174
	if (wbc->sync_mode == WB_SYNC_ALL)
J
Jens Axboe 已提交
2175
		write_flags = WRITE_SYNC;
2176 2177 2178
	else
		write_flags = WRITE;

2179 2180
	trace___extent_writepage(page, inode, wbc);

2181
	WARN_ON(!PageLocked(page));
2182
	pg_offset = i_size & (PAGE_CACHE_SIZE - 1);
2183
	if (page->index > end_index ||
2184
	   (page->index == end_index && !pg_offset)) {
2185
		page->mapping->a_ops->invalidatepage(page, 0);
2186 2187 2188 2189 2190 2191 2192 2193
		unlock_page(page);
		return 0;
	}

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

		userpage = kmap_atomic(page, KM_USER0);
2194 2195
		memset(userpage + pg_offset, 0,
		       PAGE_CACHE_SIZE - pg_offset);
2196
		kunmap_atomic(userpage, KM_USER0);
2197
		flush_dcache_page(page);
2198
	}
2199
	pg_offset = 0;
2200 2201 2202 2203 2204

	set_page_extent_mapped(page);

	delalloc_start = start;
	delalloc_end = 0;
C
Chris Mason 已提交
2205
	page_started = 0;
2206
	if (!epd->extent_locked) {
2207
		u64 delalloc_to_write = 0;
2208 2209 2210 2211 2212 2213
		/*
		 * 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
2214
		while (delalloc_end < page_end) {
2215
			nr_delalloc = find_lock_delalloc_range(inode, tree,
C
Chris Mason 已提交
2216 2217
						       page,
						       &delalloc_start,
2218 2219
						       &delalloc_end,
						       128 * 1024 * 1024);
2220 2221 2222 2223 2224 2225 2226
			if (nr_delalloc == 0) {
				delalloc_start = delalloc_end + 1;
				continue;
			}
			tree->ops->fill_delalloc(inode, page, delalloc_start,
						 delalloc_end, &page_started,
						 &nr_written);
2227 2228 2229 2230 2231 2232 2233 2234
			/*
			 * delalloc_end is already one less than the total
			 * length, so we don't subtract one from
			 * PAGE_CACHE_SIZE
			 */
			delalloc_to_write += (delalloc_end - delalloc_start +
					      PAGE_CACHE_SIZE) >>
					      PAGE_CACHE_SHIFT;
2235 2236
			delalloc_start = delalloc_end + 1;
		}
2237 2238 2239 2240 2241 2242 2243 2244
		if (wbc->nr_to_write < delalloc_to_write) {
			int thresh = 8192;

			if (delalloc_to_write < thresh * 2)
				thresh = delalloc_to_write;
			wbc->nr_to_write = min_t(u64, delalloc_to_write,
						 thresh);
		}
C
Chris Mason 已提交
2245

2246 2247 2248 2249 2250
		/* did the fill delalloc function already unlock and start
		 * the IO?
		 */
		if (page_started) {
			ret = 0;
2251 2252 2253 2254 2255 2256 2257
			/*
			 * 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;
2258
		}
C
Chris Mason 已提交
2259
	}
2260
	if (tree->ops && tree->ops->writepage_start_hook) {
C
Chris Mason 已提交
2261 2262
		ret = tree->ops->writepage_start_hook(page, start,
						      page_end);
2263 2264
		if (ret == -EAGAIN) {
			redirty_page_for_writepage(wbc, page);
2265
			update_nr_written(page, wbc, nr_written);
2266
			unlock_page(page);
2267
			ret = 0;
2268
			goto done_unlocked;
2269 2270 2271
		}
	}

2272 2273 2274 2275 2276
	/*
	 * 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);
2277

2278 2279
	end = page_end;
	if (last_byte <= start) {
2280 2281 2282
		if (tree->ops && tree->ops->writepage_end_io_hook)
			tree->ops->writepage_end_io_hook(page, start,
							 page_end, NULL, 1);
2283 2284 2285 2286 2287 2288 2289
		goto done;
	}

	blocksize = inode->i_sb->s_blocksize;

	while (cur <= end) {
		if (cur >= last_byte) {
2290 2291 2292
			if (tree->ops && tree->ops->writepage_end_io_hook)
				tree->ops->writepage_end_io_hook(page, cur,
							 page_end, NULL, 1);
2293 2294
			break;
		}
2295
		em = epd->get_extent(inode, page, pg_offset, cur,
2296
				     end - cur + 1, 1);
2297
		if (IS_ERR_OR_NULL(em)) {
2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309
			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
2310
		compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
2311 2312 2313
		free_extent_map(em);
		em = NULL;

C
Chris Mason 已提交
2314 2315 2316 2317 2318
		/*
		 * compressed and inline extents are written through other
		 * paths in the FS
		 */
		if (compressed || block_start == EXTENT_MAP_HOLE ||
2319
		    block_start == EXTENT_MAP_INLINE) {
C
Chris Mason 已提交
2320 2321 2322 2323 2324 2325
			/*
			 * end_io notification does not happen here for
			 * compressed extents
			 */
			if (!compressed && tree->ops &&
			    tree->ops->writepage_end_io_hook)
2326 2327 2328
				tree->ops->writepage_end_io_hook(page, cur,
							 cur + iosize - 1,
							 NULL, 1);
C
Chris Mason 已提交
2329 2330 2331 2332 2333 2334 2335 2336 2337
			else if (compressed) {
				/* we don't want to end_page_writeback on
				 * a compressed extent.  this happens
				 * elsewhere
				 */
				nr++;
			}

			cur += iosize;
2338
			pg_offset += iosize;
2339 2340 2341 2342
			continue;
		}
		/* leave this out until we have a page_mkwrite call */
		if (0 && !test_range_bit(tree, cur, cur + iosize - 1,
2343
				   EXTENT_DIRTY, 0, NULL)) {
2344
			cur = cur + iosize;
2345
			pg_offset += iosize;
2346 2347
			continue;
		}
C
Chris Mason 已提交
2348

2349 2350 2351 2352 2353 2354
		if (tree->ops && tree->ops->writepage_io_hook) {
			ret = tree->ops->writepage_io_hook(page, cur,
						cur + iosize - 1);
		} else {
			ret = 0;
		}
2355
		if (ret) {
2356
			SetPageError(page);
2357
		} else {
2358
			unsigned long max_nr = end_index + 1;
2359

2360 2361
			set_range_writeback(tree, cur, cur + iosize - 1);
			if (!PageWriteback(page)) {
C
Chris Mason 已提交
2362 2363 2364
				printk(KERN_ERR "btrfs warning page %lu not "
				       "writeback, cur %llu end %llu\n",
				       page->index, (unsigned long long)cur,
2365 2366 2367
				       (unsigned long long)end);
			}

2368 2369 2370
			ret = submit_extent_page(write_flags, tree, page,
						 sector, iosize, pg_offset,
						 bdev, &epd->bio, max_nr,
C
Chris Mason 已提交
2371 2372
						 end_bio_extent_writepage,
						 0, 0, 0);
2373 2374 2375 2376
			if (ret)
				SetPageError(page);
		}
		cur = cur + iosize;
2377
		pg_offset += iosize;
2378 2379 2380 2381 2382 2383 2384 2385 2386
		nr++;
	}
done:
	if (nr == 0) {
		/* make sure the mapping tag for page dirty gets cleared */
		set_page_writeback(page);
		end_page_writeback(page);
	}
	unlock_page(page);
2387

2388 2389
done_unlocked:

2390 2391
	/* drop our reference on any cached states */
	free_extent_state(cached_state);
2392 2393 2394 2395
	return 0;
}

/**
C
Chris Mason 已提交
2396
 * write_cache_pages - walk the list of dirty pages of the given address space and write all of them.
2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409
 * @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.
 */
2410
static int extent_write_cache_pages(struct extent_io_tree *tree,
C
Chris Mason 已提交
2411 2412
			     struct address_space *mapping,
			     struct writeback_control *wbc,
C
Chris Mason 已提交
2413 2414
			     writepage_t writepage, void *data,
			     void (*flush_fn)(void *))
2415 2416 2417
{
	int ret = 0;
	int done = 0;
2418
	int nr_to_write_done = 0;
2419 2420 2421 2422 2423
	struct pagevec pvec;
	int nr_pages;
	pgoff_t index;
	pgoff_t end;		/* Inclusive */
	int scanned = 0;
2424
	int tag;
2425 2426 2427 2428 2429 2430 2431 2432 2433 2434

	pagevec_init(&pvec, 0);
	if (wbc->range_cyclic) {
		index = mapping->writeback_index; /* Start from prev offset */
		end = -1;
	} else {
		index = wbc->range_start >> PAGE_CACHE_SHIFT;
		end = wbc->range_end >> PAGE_CACHE_SHIFT;
		scanned = 1;
	}
2435 2436 2437 2438
	if (wbc->sync_mode == WB_SYNC_ALL)
		tag = PAGECACHE_TAG_TOWRITE;
	else
		tag = PAGECACHE_TAG_DIRTY;
2439
retry:
2440 2441
	if (wbc->sync_mode == WB_SYNC_ALL)
		tag_pages_for_writeback(mapping, index, end);
2442
	while (!done && !nr_to_write_done && (index <= end) &&
2443 2444
	       (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag,
			min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457
		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
2458 2459 2460 2461
			if (tree->ops && tree->ops->write_cache_pages_lock_hook)
				tree->ops->write_cache_pages_lock_hook(page);
			else
				lock_page(page);
2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473

			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
2474
			if (wbc->sync_mode != WB_SYNC_NONE) {
2475 2476
				if (PageWriteback(page))
					flush_fn(data);
2477
				wait_on_page_writeback(page);
C
Chris Mason 已提交
2478
			}
2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491

			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;
			}
2492
			if (ret)
2493
				done = 1;
2494 2495 2496 2497 2498 2499 2500

			/*
			 * the filesystem may choose to bump up nr_to_write.
			 * We have to make sure to honor the new nr_to_write
			 * at any time
			 */
			nr_to_write_done = wbc->nr_to_write <= 0;
2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516
		}
		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;
}

2517
static void flush_epd_write_bio(struct extent_page_data *epd)
C
Chris Mason 已提交
2518 2519
{
	if (epd->bio) {
2520 2521 2522 2523
		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
2524 2525 2526 2527
		epd->bio = NULL;
	}
}

2528 2529 2530 2531 2532 2533
static noinline void flush_write_bio(void *data)
{
	struct extent_page_data *epd = data;
	flush_epd_write_bio(epd);
}

2534 2535 2536 2537 2538 2539 2540 2541 2542 2543
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,
2544
		.extent_locked = 0,
2545
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
2546 2547
	};
	struct writeback_control wbc_writepages = {
2548
		.sync_mode	= wbc->sync_mode,
2549 2550 2551 2552 2553 2554 2555 2556
		.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
2557
	extent_write_cache_pages(tree, mapping, &wbc_writepages,
C
Chris Mason 已提交
2558
				 __extent_writepage, &epd, flush_write_bio);
2559
	flush_epd_write_bio(&epd);
2560 2561 2562
	return ret;
}

2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577
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,
2578
		.sync_io = mode == WB_SYNC_ALL,
2579 2580 2581 2582 2583 2584 2585 2586 2587
	};
	struct writeback_control wbc_writepages = {
		.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
2588
	while (start <= end) {
2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602
		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;
	}

2603
	flush_epd_write_bio(&epd);
2604 2605
	return ret;
}
2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616

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,
2617
		.extent_locked = 0,
2618
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
2619 2620
	};

C
Chris Mason 已提交
2621
	ret = extent_write_cache_pages(tree, mapping, wbc,
C
Chris Mason 已提交
2622 2623
				       __extent_writepage, &epd,
				       flush_write_bio);
2624
	flush_epd_write_bio(&epd);
2625 2626 2627 2628 2629 2630 2631 2632 2633 2634
	return ret;
}

int extent_readpages(struct extent_io_tree *tree,
		     struct address_space *mapping,
		     struct list_head *pages, unsigned nr_pages,
		     get_extent_t get_extent)
{
	struct bio *bio = NULL;
	unsigned page_idx;
C
Chris Mason 已提交
2635
	unsigned long bio_flags = 0;
2636 2637 2638 2639 2640 2641

	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);
2642
		if (!add_to_page_cache_lru(page, mapping,
2643
					page->index, GFP_NOFS)) {
2644
			__extent_read_full_page(tree, page, get_extent,
C
Chris Mason 已提交
2645
						&bio, 0, &bio_flags);
2646 2647 2648 2649 2650
		}
		page_cache_release(page);
	}
	BUG_ON(!list_empty(pages));
	if (bio)
C
Chris Mason 已提交
2651
		submit_one_bio(READ, bio, 0, bio_flags);
2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662
	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)
{
2663
	struct extent_state *cached_state = NULL;
2664 2665 2666 2667
	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
2668
	start += (offset + blocksize - 1) & ~(blocksize - 1);
2669 2670 2671
	if (start > end)
		return 0;

2672
	lock_extent_bits(tree, start, end, 0, &cached_state, GFP_NOFS);
2673
	wait_on_page_writeback(page);
2674
	clear_extent_bit(tree, start, end,
2675 2676
			 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
			 EXTENT_DO_ACCOUNTING,
2677
			 1, 1, &cached_state, GFP_NOFS);
2678 2679 2680
	return 0;
}

2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693
/*
 * 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;

2694
	if (test_range_bit(tree, start, end,
2695
			   EXTENT_IOBITS, 0, NULL))
2696 2697 2698 2699
		ret = 0;
	else {
		if ((mask & GFP_NOFS) == GFP_NOFS)
			mask = GFP_NOFS;
2700 2701 2702 2703
		/*
		 * at this point we can safely clear everything except the
		 * locked bit and the nodatasum bit
		 */
2704
		ret = clear_extent_bit(tree, start, end,
2705 2706
				 ~(EXTENT_LOCKED | EXTENT_NODATASUM),
				 0, 0, NULL, mask);
2707 2708 2709 2710 2711 2712 2713 2714

		/* if clear_extent_bit failed for enomem reasons,
		 * we can't allow the release to continue.
		 */
		if (ret < 0)
			ret = 0;
		else
			ret = 1;
2715 2716 2717 2718
	}
	return ret;
}

2719 2720 2721 2722 2723 2724
/*
 * 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,
2725 2726
			       struct extent_io_tree *tree, struct page *page,
			       gfp_t mask)
2727 2728 2729 2730
{
	struct extent_map *em;
	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
	u64 end = start + PAGE_CACHE_SIZE - 1;
2731

2732 2733
	if ((mask & __GFP_WAIT) &&
	    page->mapping->host->i_size > 16 * 1024 * 1024) {
2734
		u64 len;
2735
		while (start <= end) {
2736
			len = end - start + 1;
2737
			write_lock(&map->lock);
2738
			em = lookup_extent_mapping(map, start, len);
2739
			if (IS_ERR_OR_NULL(em)) {
2740
				write_unlock(&map->lock);
2741 2742
				break;
			}
2743 2744
			if (test_bit(EXTENT_FLAG_PINNED, &em->flags) ||
			    em->start != start) {
2745
				write_unlock(&map->lock);
2746 2747 2748 2749 2750
				free_extent_map(em);
				break;
			}
			if (!test_range_bit(tree, em->start,
					    extent_map_end(em) - 1,
2751
					    EXTENT_LOCKED | EXTENT_WRITEBACK,
2752
					    0, NULL)) {
2753 2754 2755 2756 2757
				remove_extent_mapping(map, em);
				/* once for the rb tree */
				free_extent_map(em);
			}
			start = extent_map_end(em);
2758
			write_unlock(&map->lock);
2759 2760

			/* once for us */
2761 2762 2763
			free_extent_map(em);
		}
	}
2764
	return try_release_extent_state(map, tree, page, mask);
2765 2766
}

2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788
/*
 * helper function for fiemap, which doesn't want to see any holes.
 * This maps until we find something past 'last'
 */
static struct extent_map *get_extent_skip_holes(struct inode *inode,
						u64 offset,
						u64 last,
						get_extent_t *get_extent)
{
	u64 sectorsize = BTRFS_I(inode)->root->sectorsize;
	struct extent_map *em;
	u64 len;

	if (offset >= last)
		return NULL;

	while(1) {
		len = last - offset;
		if (len == 0)
			break;
		len = (len + sectorsize - 1) & ~(sectorsize - 1);
		em = get_extent(inode, NULL, 0, offset, len, 0);
2789
		if (IS_ERR_OR_NULL(em))
2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806
			return em;

		/* if this isn't a hole return it */
		if (!test_bit(EXTENT_FLAG_VACANCY, &em->flags) &&
		    em->block_start != EXTENT_MAP_HOLE) {
			return em;
		}

		/* this is a hole, advance to the next extent */
		offset = extent_map_end(em);
		free_extent_map(em);
		if (offset >= last)
			break;
	}
	return NULL;
}

Y
Yehuda Sadeh 已提交
2807 2808 2809
int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
		__u64 start, __u64 len, get_extent_t *get_extent)
{
J
Josef Bacik 已提交
2810
	int ret = 0;
Y
Yehuda Sadeh 已提交
2811 2812 2813
	u64 off = start;
	u64 max = start + len;
	u32 flags = 0;
J
Josef Bacik 已提交
2814 2815
	u32 found_type;
	u64 last;
2816
	u64 last_for_get_extent = 0;
Y
Yehuda Sadeh 已提交
2817
	u64 disko = 0;
2818
	u64 isize = i_size_read(inode);
J
Josef Bacik 已提交
2819
	struct btrfs_key found_key;
Y
Yehuda Sadeh 已提交
2820
	struct extent_map *em = NULL;
2821
	struct extent_state *cached_state = NULL;
J
Josef Bacik 已提交
2822 2823
	struct btrfs_path *path;
	struct btrfs_file_extent_item *item;
Y
Yehuda Sadeh 已提交
2824
	int end = 0;
2825 2826 2827
	u64 em_start = 0;
	u64 em_len = 0;
	u64 em_end = 0;
Y
Yehuda Sadeh 已提交
2828 2829 2830 2831 2832
	unsigned long emflags;

	if (len == 0)
		return -EINVAL;

J
Josef Bacik 已提交
2833 2834 2835 2836 2837
	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;
	path->leave_spinning = 1;

2838 2839 2840 2841
	/*
	 * lookup the last file extent.  We're not using i_size here
	 * because there might be preallocation past i_size
	 */
J
Josef Bacik 已提交
2842
	ret = btrfs_lookup_file_extent(NULL, BTRFS_I(inode)->root,
L
Li Zefan 已提交
2843
				       path, btrfs_ino(inode), -1, 0);
J
Josef Bacik 已提交
2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854
	if (ret < 0) {
		btrfs_free_path(path);
		return ret;
	}
	WARN_ON(!ret);
	path->slots[0]--;
	item = btrfs_item_ptr(path->nodes[0], path->slots[0],
			      struct btrfs_file_extent_item);
	btrfs_item_key_to_cpu(path->nodes[0], &found_key, path->slots[0]);
	found_type = btrfs_key_type(&found_key);

2855
	/* No extents, but there might be delalloc bits */
L
Li Zefan 已提交
2856
	if (found_key.objectid != btrfs_ino(inode) ||
J
Josef Bacik 已提交
2857
	    found_type != BTRFS_EXTENT_DATA_KEY) {
2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868
		/* have to trust i_size as the end */
		last = (u64)-1;
		last_for_get_extent = isize;
	} else {
		/*
		 * remember the start of the last extent.  There are a
		 * bunch of different factors that go into the length of the
		 * extent, so its much less complex to remember where it started
		 */
		last = found_key.offset;
		last_for_get_extent = last + 1;
J
Josef Bacik 已提交
2869 2870 2871
	}
	btrfs_free_path(path);

2872 2873 2874 2875 2876 2877 2878 2879 2880 2881
	/*
	 * we might have some extents allocated but more delalloc past those
	 * extents.  so, we trust isize unless the start of the last extent is
	 * beyond isize
	 */
	if (last < isize) {
		last = (u64)-1;
		last_for_get_extent = isize;
	}

2882 2883
	lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + len, 0,
			 &cached_state, GFP_NOFS);
2884 2885 2886

	em = get_extent_skip_holes(inode, off, last_for_get_extent,
				   get_extent);
Y
Yehuda Sadeh 已提交
2887 2888 2889 2890 2891 2892
	if (!em)
		goto out;
	if (IS_ERR(em)) {
		ret = PTR_ERR(em);
		goto out;
	}
J
Josef Bacik 已提交
2893

Y
Yehuda Sadeh 已提交
2894
	while (!end) {
2895 2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907
		u64 offset_in_extent;

		/* break if the extent we found is outside the range */
		if (em->start >= max || extent_map_end(em) < off)
			break;

		/*
		 * get_extent may return an extent that starts before our
		 * requested range.  We have to make sure the ranges
		 * we return to fiemap always move forward and don't
		 * overlap, so adjust the offsets here
		 */
		em_start = max(em->start, off);
Y
Yehuda Sadeh 已提交
2908

2909 2910 2911 2912 2913
		/*
		 * record the offset from the start of the extent
		 * for adjusting the disk offset below
		 */
		offset_in_extent = em_start - em->start;
2914
		em_end = extent_map_end(em);
2915
		em_len = em_end - em_start;
2916
		emflags = em->flags;
Y
Yehuda Sadeh 已提交
2917 2918 2919
		disko = 0;
		flags = 0;

2920 2921 2922 2923 2924 2925 2926
		/*
		 * bump off for our next call to get_extent
		 */
		off = extent_map_end(em);
		if (off >= max)
			end = 1;

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_INLINE) {
Y
Yehuda Sadeh 已提交
2931 2932
			flags |= (FIEMAP_EXTENT_DATA_INLINE |
				  FIEMAP_EXTENT_NOT_ALIGNED);
2933
		} else if (em->block_start == EXTENT_MAP_DELALLOC) {
Y
Yehuda Sadeh 已提交
2934 2935
			flags |= (FIEMAP_EXTENT_DELALLOC |
				  FIEMAP_EXTENT_UNKNOWN);
2936
		} else {
2937
			disko = em->block_start + offset_in_extent;
Y
Yehuda Sadeh 已提交
2938 2939 2940 2941 2942 2943
		}
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
			flags |= FIEMAP_EXTENT_ENCODED;

		free_extent_map(em);
		em = NULL;
2944 2945
		if ((em_start >= last) || em_len == (u64)-1 ||
		   (last == (u64)-1 && isize <= em_end)) {
Y
Yehuda Sadeh 已提交
2946 2947 2948 2949
			flags |= FIEMAP_EXTENT_LAST;
			end = 1;
		}

2950 2951 2952 2953 2954 2955 2956 2957
		/* now scan forward to see if this is really the last extent. */
		em = get_extent_skip_holes(inode, off, last_for_get_extent,
					   get_extent);
		if (IS_ERR(em)) {
			ret = PTR_ERR(em);
			goto out;
		}
		if (!em) {
J
Josef Bacik 已提交
2958 2959 2960
			flags |= FIEMAP_EXTENT_LAST;
			end = 1;
		}
2961 2962 2963 2964
		ret = fiemap_fill_next_extent(fieinfo, em_start, disko,
					      em_len, flags);
		if (ret)
			goto out_free;
Y
Yehuda Sadeh 已提交
2965 2966 2967 2968
	}
out_free:
	free_extent_map(em);
out:
2969 2970
	unlock_extent_cached(&BTRFS_I(inode)->io_tree, start, start + len,
			     &cached_state, GFP_NOFS);
Y
Yehuda Sadeh 已提交
2971 2972 2973
	return ret;
}

2974 2975 2976 2977 2978 2979 2980 2981 2982 2983
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
2984 2985
	if (!mapping)
		return NULL;
2986 2987 2988 2989 2990 2991 2992

	/*
	 * 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();
2993
	p = radix_tree_lookup(&mapping->page_tree, i);
2994
	rcu_read_unlock();
2995

2996 2997 2998
	return p;
}

2999
static inline unsigned long num_extent_pages(u64 start, u64 len)
3000
{
3001 3002
	return ((start + len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT) -
		(start >> PAGE_CACHE_SHIFT);
3003 3004
}

3005 3006 3007 3008 3009 3010
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;
3011
#if LEAK_DEBUG
3012
	unsigned long flags;
C
Chris Mason 已提交
3013
#endif
3014 3015

	eb = kmem_cache_zalloc(extent_buffer_cache, mask);
T
Tsutomu Itoh 已提交
3016 3017
	if (eb == NULL)
		return NULL;
3018 3019
	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
	kmem_cache_free(extent_buffer_cache, eb);
}

3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076
/*
 * Helper for releasing extent buffer page.
 */
static void btrfs_release_extent_buffer_page(struct extent_buffer *eb,
						unsigned long start_idx)
{
	unsigned long index;
	struct page *page;

	if (!eb->first_page)
		return;

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

	do {
		index--;
		page = extent_buffer_page(eb, index);
		if (page)
			page_cache_release(page);
	} while (index != start_idx);
}

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

3077 3078
struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree,
					  u64 start, unsigned long len,
3079
					  struct page *page0)
3080 3081 3082 3083 3084
{
	unsigned long num_pages = num_extent_pages(start, len);
	unsigned long i;
	unsigned long index = start >> PAGE_CACHE_SHIFT;
	struct extent_buffer *eb;
3085
	struct extent_buffer *exists = NULL;
3086 3087 3088
	struct page *p;
	struct address_space *mapping = tree->mapping;
	int uptodate = 1;
3089
	int ret;
3090

3091 3092 3093 3094
	rcu_read_lock();
	eb = radix_tree_lookup(&tree->buffer, start >> PAGE_CACHE_SHIFT);
	if (eb && atomic_inc_not_zero(&eb->refs)) {
		rcu_read_unlock();
J
Josef Bacik 已提交
3095
		mark_page_accessed(eb->first_page);
3096 3097
		return eb;
	}
3098
	rcu_read_unlock();
3099

3100
	eb = __alloc_extent_buffer(tree, start, len, GFP_NOFS);
3101
	if (!eb)
3102 3103 3104 3105 3106 3107 3108 3109 3110 3111
		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);
3112
		uptodate = PageUptodate(page0);
3113 3114 3115 3116
	} else {
		i = 0;
	}
	for (; i < num_pages; i++, index++) {
3117
		p = find_or_create_page(mapping, index, GFP_NOFS | __GFP_HIGHMEM);
3118 3119
		if (!p) {
			WARN_ON(1);
3120
			goto free_eb;
3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131
		}
		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;
C
Chris Mason 已提交
3132 3133 3134 3135 3136 3137 3138

		/*
		 * see below about how we avoid a nasty race with release page
		 * and why we unlock later
		 */
		if (i != 0)
			unlock_page(p);
3139 3140
	}
	if (uptodate)
3141
		set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
3142

3143 3144 3145 3146
	ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
	if (ret)
		goto free_eb;

3147
	spin_lock(&tree->buffer_lock);
3148 3149 3150 3151
	ret = radix_tree_insert(&tree->buffer, start >> PAGE_CACHE_SHIFT, eb);
	if (ret == -EEXIST) {
		exists = radix_tree_lookup(&tree->buffer,
						start >> PAGE_CACHE_SHIFT);
3152 3153 3154
		/* add one reference for the caller */
		atomic_inc(&exists->refs);
		spin_unlock(&tree->buffer_lock);
3155
		radix_tree_preload_end();
3156 3157 3158 3159
		goto free_eb;
	}
	/* add one reference for the tree */
	atomic_inc(&eb->refs);
3160
	spin_unlock(&tree->buffer_lock);
3161
	radix_tree_preload_end();
C
Chris Mason 已提交
3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175

	/*
	 * there is a race where release page may have
	 * tried to find this extent buffer in the radix
	 * but failed.  It will tell the VM it is safe to
	 * reclaim the, and it will clear the page private bit.
	 * We must make sure to set the page private bit properly
	 * after the extent buffer is in the radix tree so
	 * it doesn't get lost
	 */
	set_page_extent_mapped(eb->first_page);
	set_page_extent_head(eb->first_page, eb->len);
	if (!page0)
		unlock_page(eb->first_page);
3176 3177
	return eb;

3178
free_eb:
C
Chris Mason 已提交
3179 3180 3181
	if (eb->first_page && !page0)
		unlock_page(eb->first_page);

3182
	if (!atomic_dec_and_test(&eb->refs))
3183
		return exists;
3184
	btrfs_release_extent_buffer(eb);
3185
	return exists;
3186 3187 3188
}

struct extent_buffer *find_extent_buffer(struct extent_io_tree *tree,
3189
					 u64 start, unsigned long len)
3190 3191 3192
{
	struct extent_buffer *eb;

3193 3194 3195 3196
	rcu_read_lock();
	eb = radix_tree_lookup(&tree->buffer, start >> PAGE_CACHE_SHIFT);
	if (eb && atomic_inc_not_zero(&eb->refs)) {
		rcu_read_unlock();
J
Josef Bacik 已提交
3197
		mark_page_accessed(eb->first_page);
3198 3199 3200
		return eb;
	}
	rcu_read_unlock();
J
Josef Bacik 已提交
3201

3202
	return NULL;
3203 3204 3205 3206 3207 3208 3209 3210 3211 3212
}

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

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

3213
	WARN_ON(1);
3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226
}

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);
3227
		if (!PageDirty(page))
C
Chris Mason 已提交
3228 3229
			continue;

3230
		lock_page(page);
C
Chris Mason 已提交
3231 3232 3233
		WARN_ON(!PagePrivate(page));

		set_page_extent_mapped(page);
3234 3235 3236 3237
		if (i == 0)
			set_page_extent_head(page, eb->len);

		clear_page_dirty_for_io(page);
3238
		spin_lock_irq(&page->mapping->tree_lock);
3239 3240 3241 3242 3243
		if (!PageDirty(page)) {
			radix_tree_tag_clear(&page->mapping->page_tree,
						page_index(page),
						PAGECACHE_TAG_DIRTY);
		}
3244
		spin_unlock_irq(&page->mapping->tree_lock);
3245
		unlock_page(page);
3246 3247 3248 3249 3250 3251 3252 3253 3254
	}
	return 0;
}

int set_extent_buffer_dirty(struct extent_io_tree *tree,
			     struct extent_buffer *eb)
{
	unsigned long i;
	unsigned long num_pages;
3255
	int was_dirty = 0;
3256

3257
	was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
3258
	num_pages = num_extent_pages(eb->start, eb->len);
3259
	for (i = 0; i < num_pages; i++)
3260
		__set_page_dirty_nobuffers(extent_buffer_page(eb, i));
3261
	return was_dirty;
3262 3263
}

3264
int clear_extent_buffer_uptodate(struct extent_io_tree *tree,
3265 3266
				struct extent_buffer *eb,
				struct extent_state **cached_state)
3267 3268 3269 3270 3271 3272
{
	unsigned long i;
	struct page *page;
	unsigned long num_pages;

	num_pages = num_extent_pages(eb->start, eb->len);
3273
	clear_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
3274 3275

	clear_extent_uptodate(tree, eb->start, eb->start + eb->len - 1,
3276
			      cached_state, GFP_NOFS);
3277 3278
	for (i = 0; i < num_pages; i++) {
		page = extent_buffer_page(eb, i);
C
Chris Mason 已提交
3279 3280
		if (page)
			ClearPageUptodate(page);
3281 3282 3283 3284
	}
	return 0;
}

3285 3286 3287 3288 3289 3290 3291 3292 3293 3294
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,
3295
			    NULL, GFP_NOFS);
3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308
	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;
}

3309 3310 3311 3312 3313 3314 3315 3316 3317
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;

3318
	ret = test_range_bit(tree, start, end, EXTENT_UPTODATE, 1, NULL);
3319 3320
	if (ret)
		return 1;
C
Chris Mason 已提交
3321
	while (start <= end) {
3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334
		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;
}

3335
int extent_buffer_uptodate(struct extent_io_tree *tree,
3336 3337
			   struct extent_buffer *eb,
			   struct extent_state *cached_state)
3338
{
3339
	int ret = 0;
3340 3341
	unsigned long num_pages;
	unsigned long i;
3342 3343 3344
	struct page *page;
	int pg_uptodate = 1;

3345
	if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags))
C
Chris Mason 已提交
3346
		return 1;
3347

C
Chris Mason 已提交
3348
	ret = test_range_bit(tree, eb->start, eb->start + eb->len - 1,
3349
			   EXTENT_UPTODATE, 1, cached_state);
C
Chris Mason 已提交
3350 3351
	if (ret)
		return ret;
3352 3353 3354 3355 3356 3357 3358 3359 3360

	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
3361
	return pg_uptodate;
3362 3363 3364 3365
}

int read_extent_buffer_pages(struct extent_io_tree *tree,
			     struct extent_buffer *eb,
3366
			     u64 start, int wait,
3367
			     get_extent_t *get_extent, int mirror_num)
3368 3369 3370 3371 3372 3373
{
	unsigned long i;
	unsigned long start_i;
	struct page *page;
	int err;
	int ret = 0;
3374 3375 3376
	int locked_pages = 0;
	int all_uptodate = 1;
	int inc_all_pages = 0;
3377
	unsigned long num_pages;
3378
	struct bio *bio = NULL;
C
Chris Mason 已提交
3379
	unsigned long bio_flags = 0;
3380

3381
	if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags))
3382 3383
		return 0;

3384
	if (test_range_bit(tree, eb->start, eb->start + eb->len - 1,
3385
			   EXTENT_UPTODATE, 1, NULL)) {
3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400
		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) {
3401
			if (!trylock_page(page))
3402
				goto unlock_exit;
3403 3404 3405
		} else {
			lock_page(page);
		}
3406
		locked_pages++;
C
Chris Mason 已提交
3407
		if (!PageUptodate(page))
3408 3409 3410 3411
			all_uptodate = 0;
	}
	if (all_uptodate) {
		if (start_i == 0)
3412
			set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
3413 3414 3415 3416 3417
		goto unlock_exit;
	}

	for (i = start_i; i < num_pages; i++) {
		page = extent_buffer_page(eb, i);
C
Chris Mason 已提交
3418 3419 3420 3421 3422 3423 3424

		WARN_ON(!PagePrivate(page));

		set_page_extent_mapped(page);
		if (i == 0)
			set_page_extent_head(page, eb->len);

3425 3426 3427 3428 3429
		if (inc_all_pages)
			page_cache_get(page);
		if (!PageUptodate(page)) {
			if (start_i == 0)
				inc_all_pages = 1;
3430
			ClearPageError(page);
3431
			err = __extent_read_full_page(tree, page,
3432
						      get_extent, &bio,
C
Chris Mason 已提交
3433
						      mirror_num, &bio_flags);
C
Chris Mason 已提交
3434
			if (err)
3435 3436 3437 3438 3439 3440
				ret = err;
		} else {
			unlock_page(page);
		}
	}

3441
	if (bio)
C
Chris Mason 已提交
3442
		submit_one_bio(READ, bio, mirror_num, bio_flags);
3443

C
Chris Mason 已提交
3444
	if (ret || !wait)
3445
		return ret;
C
Chris Mason 已提交
3446

3447 3448 3449
	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
3450
		if (!PageUptodate(page))
3451 3452
			ret = -EIO;
	}
C
Chris Mason 已提交
3453

3454
	if (!ret)
3455
		set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
3456
	return ret;
3457 3458 3459

unlock_exit:
	i = start_i;
C
Chris Mason 已提交
3460
	while (locked_pages > 0) {
3461 3462 3463 3464 3465 3466
		page = extent_buffer_page(eb, i);
		i++;
		unlock_page(page);
		locked_pages--;
	}
	return ret;
3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485
}

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
3486
	while (len > 0) {
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
		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
3524

3525
	if (start + min_len > eb->len) {
C
Chris Mason 已提交
3526 3527 3528
		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);
3529
		WARN_ON(1);
3530
		return -EINVAL;
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 3561 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
	}

	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
3588
	while (len > 0) {
3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622
		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
3623
	while (len > 0) {
3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653
		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
3654
	while (len > 0) {
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
		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
3686
	while (len > 0) {
3687 3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722
		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);
}

3723 3724 3725 3726 3727 3728
static inline bool areas_overlap(unsigned long src, unsigned long dst, unsigned long len)
{
	unsigned long distance = (src > dst) ? src - dst : dst - src;
	return distance < len;
}

3729 3730 3731 3732 3733 3734 3735
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;

3736
	if (dst_page != src_page) {
3737
		src_kaddr = kmap_atomic(src_page, KM_USER1);
3738
	} else {
3739
		src_kaddr = dst_kaddr;
3740 3741
		BUG_ON(areas_overlap(src_off, dst_off, len));
	}
3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759

	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
3760 3761
		printk(KERN_ERR "btrfs memmove bogus src_offset %lu move "
		       "len %lu dst len %lu\n", src_offset, len, dst->len);
3762 3763 3764
		BUG_ON(1);
	}
	if (dst_offset + len > dst->len) {
C
Chris Mason 已提交
3765 3766
		printk(KERN_ERR "btrfs memmove bogus dst_offset %lu move "
		       "len %lu dst len %lu\n", dst_offset, len, dst->len);
3767 3768 3769
		BUG_ON(1);
	}

C
Chris Mason 已提交
3770
	while (len > 0) {
3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806
		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
3807 3808
		printk(KERN_ERR "btrfs memmove bogus src_offset %lu move "
		       "len %lu len %lu\n", src_offset, len, dst->len);
3809 3810 3811
		BUG_ON(1);
	}
	if (dst_offset + len > dst->len) {
C
Chris Mason 已提交
3812 3813
		printk(KERN_ERR "btrfs memmove bogus dst_offset %lu move "
		       "len %lu len %lu\n", dst_offset, len, dst->len);
3814 3815
		BUG_ON(1);
	}
3816
	if (!areas_overlap(src_offset, dst_offset, len)) {
3817 3818 3819
		memcpy_extent_buffer(dst, dst_offset, src_offset, len);
		return;
	}
C
Chris Mason 已提交
3820
	while (len > 0) {
3821 3822 3823 3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839 3840
		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;
	}
}
3841

3842 3843 3844 3845 3846 3847 3848 3849
static inline void btrfs_release_extent_buffer_rcu(struct rcu_head *head)
{
	struct extent_buffer *eb =
			container_of(head, struct extent_buffer, rcu_head);

	btrfs_release_extent_buffer(eb);
}

3850 3851 3852 3853 3854 3855 3856
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;

	spin_lock(&tree->buffer_lock);
3857
	eb = radix_tree_lookup(&tree->buffer, start >> PAGE_CACHE_SHIFT);
3858 3859 3860 3861
	if (!eb) {
		spin_unlock(&tree->buffer_lock);
		return ret;
	}
3862

3863
	if (test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
3864 3865 3866
		ret = 0;
		goto out;
	}
3867 3868 3869 3870 3871 3872

	/*
	 * set @eb->refs to 0 if it is already 1, and then release the @eb.
	 * Or go back.
	 */
	if (atomic_cmpxchg(&eb->refs, 1, 0) != 1) {
3873 3874 3875
		ret = 0;
		goto out;
	}
3876

3877
	radix_tree_delete(&tree->buffer, start >> PAGE_CACHE_SHIFT);
3878 3879
out:
	spin_unlock(&tree->buffer_lock);
3880 3881 3882 3883

	/* at this point we can safely release the extent buffer */
	if (atomic_read(&eb->refs) == 0)
		call_rcu(&eb->rcu_head, btrfs_release_extent_buffer_rcu);
3884 3885
	return ret;
}