extent_io.c 94.2 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 99 100 101 102 103 104 105
	if (extent_state_cache)
		kmem_cache_destroy(extent_state_cache);
	if (extent_buffer_cache)
		kmem_cache_destroy(extent_buffer_cache);
}

void extent_io_tree_init(struct extent_io_tree *tree,
			  struct address_space *mapping, gfp_t mask)
{
106 107
	tree->state = RB_ROOT;
	tree->buffer = RB_ROOT;
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;
}

238 239 240 241
static struct extent_buffer *buffer_tree_insert(struct extent_io_tree *tree,
					  u64 offset, struct rb_node *node)
{
	struct rb_root *root = &tree->buffer;
C
Chris Mason 已提交
242 243
	struct rb_node **p = &root->rb_node;
	struct rb_node *parent = NULL;
244 245
	struct extent_buffer *eb;

C
Chris Mason 已提交
246
	while (*p) {
247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266
		parent = *p;
		eb = rb_entry(parent, struct extent_buffer, rb_node);

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

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

static struct extent_buffer *buffer_search(struct extent_io_tree *tree,
					   u64 offset)
{
	struct rb_root *root = &tree->buffer;
C
Chris Mason 已提交
267
	struct rb_node *n = root->rb_node;
268 269
	struct extent_buffer *eb;

C
Chris Mason 已提交
270
	while (n) {
271 272 273 274 275 276 277 278 279 280 281
		eb = rb_entry(n, struct extent_buffer, rb_node);
		if (offset < eb->start)
			n = n->rb_left;
		else if (offset > eb->start)
			n = n->rb_right;
		else
			return eb;
	}
	return NULL;
}

J
Josef Bacik 已提交
282 283 284 285 286 287 288 289
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);
}

290 291 292 293 294 295 296 297 298 299 300 301 302 303 304
/*
 * 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;

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

	other_node = rb_prev(&state->rb_node);
	if (other_node) {
		other = rb_entry(other_node, struct extent_state, rb_node);
		if (other->end == state->start - 1 &&
		    other->state == state->state) {
J
Josef Bacik 已提交
313
			merge_cb(tree, state, other);
314
			state->start = other->start;
315
			other->tree = NULL;
316 317 318 319 320 321 322 323 324
			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
325
			merge_cb(tree, state, other);
326
			other->start = state->start;
327
			state->tree = NULL;
328 329
			rb_erase(&state->rb_node, &tree->state);
			free_extent_state(state);
J
Josef Bacik 已提交
330
			state = NULL;
331 332
		}
	}
J
Josef Bacik 已提交
333

334 335 336
	return 0;
}

J
Josef Bacik 已提交
337
static int set_state_cb(struct extent_io_tree *tree,
338
			 struct extent_state *state, int *bits)
339 340
{
	if (tree->ops && tree->ops->set_bit_hook) {
J
Josef Bacik 已提交
341
		return tree->ops->set_bit_hook(tree->mapping->host,
342
					       state, bits);
343
	}
J
Josef Bacik 已提交
344 345

	return 0;
346 347 348
}

static void clear_state_cb(struct extent_io_tree *tree,
349
			   struct extent_state *state, int *bits)
350
{
J
Josef Bacik 已提交
351 352
	if (tree->ops && tree->ops->clear_bit_hook)
		tree->ops->clear_bit_hook(tree->mapping->host, state, bits);
353 354
}

355 356 357 358 359 360 361 362 363 364 365 366
/*
 * 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,
367
			int *bits)
368 369
{
	struct rb_node *node;
370
	int bits_to_set = *bits & ~EXTENT_CTLBITS;
J
Josef Bacik 已提交
371
	int ret;
372 373

	if (end < start) {
C
Chris Mason 已提交
374 375 376
		printk(KERN_ERR "btrfs end < start %llu %llu\n",
		       (unsigned long long)end,
		       (unsigned long long)start);
377 378 379 380
		WARN_ON(1);
	}
	state->start = start;
	state->end = end;
J
Josef Bacik 已提交
381 382 383 384
	ret = set_state_cb(tree, state, bits);
	if (ret)
		return ret;

385
	if (bits_to_set & EXTENT_DIRTY)
J
Josef Bacik 已提交
386
		tree->dirty_bytes += end - start + 1;
387
	state->state |= bits_to_set;
388 389 390 391
	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
392 393 394 395
		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);
396 397 398
		free_extent_state(state);
		return -EEXIST;
	}
399
	state->tree = tree;
400 401 402 403
	merge_state(tree, state);
	return 0;
}

J
Josef Bacik 已提交
404 405 406 407 408 409 410 411 412
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;
}

413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430
/*
 * 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
431 432 433

	split_cb(tree, orig, split);

434 435 436 437 438 439 440 441 442 443
	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;
	}
444
	prealloc->tree = tree;
445 446 447 448 449 450 451 452 453 454 455 456
	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,
457 458
			    struct extent_state *state,
			    int *bits, int wake)
459
{
460
	int bits_to_clear = *bits & ~EXTENT_CTLBITS;
461
	int ret = state->state & bits_to_clear;
462

463
	if ((bits_to_clear & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) {
464 465 466 467
		u64 range = state->end - state->start + 1;
		WARN_ON(range > tree->dirty_bytes);
		tree->dirty_bytes -= range;
	}
468
	clear_state_cb(tree, state, bits);
469
	state->state &= ~bits_to_clear;
470 471
	if (wake)
		wake_up(&state->wq);
472
	if (state->state == 0) {
473
		if (state->tree) {
474
			rb_erase(&state->rb_node, &tree->state);
475
			state->tree = NULL;
476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499
			free_extent_state(state);
		} else {
			WARN_ON(1);
		}
	} else {
		merge_state(tree, state);
	}
	return ret;
}

/*
 * clear some bits on a range in the tree.  This may require splitting
 * or inserting elements in the tree, so the gfp mask is used to
 * indicate which allocations or sleeping are allowed.
 *
 * pass 'wake' == 1 to kick any sleepers, and 'delete' == 1 to remove
 * the given range from the tree regardless of state (ie for truncate).
 *
 * the range [start, end] is inclusive.
 *
 * This takes the tree lock, and returns < 0 on error, > 0 if any of the
 * bits were already set, or zero if none of the bits were already set.
 */
int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
500 501 502
		     int bits, int wake, int delete,
		     struct extent_state **cached_state,
		     gfp_t mask)
503 504
{
	struct extent_state *state;
505
	struct extent_state *cached;
506
	struct extent_state *prealloc = NULL;
507
	struct rb_node *next_node;
508
	struct rb_node *node;
509
	u64 last_end;
510 511
	int err;
	int set = 0;
512
	int clear = 0;
513

514 515 516 517
	if (delete)
		bits |= ~EXTENT_CTLBITS;
	bits |= EXTENT_FIRST_DELALLOC;

518 519
	if (bits & (EXTENT_IOBITS | EXTENT_BOUNDARY))
		clear = 1;
520 521 522 523 524 525 526
again:
	if (!prealloc && (mask & __GFP_WAIT)) {
		prealloc = alloc_extent_state(mask);
		if (!prealloc)
			return -ENOMEM;
	}

527
	spin_lock(&tree->lock);
528 529
	if (cached_state) {
		cached = *cached_state;
530 531 532 533 534 535

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

536
		if (cached && cached->tree && cached->start == start) {
537 538
			if (clear)
				atomic_dec(&cached->refs);
539
			state = cached;
540
			goto hit_next;
541
		}
542 543
		if (clear)
			free_extent_state(cached);
544
	}
545 546 547 548
	/*
	 * this search will find the extents that end after
	 * our range starts
	 */
549
	node = tree_search(tree, start);
550 551 552
	if (!node)
		goto out;
	state = rb_entry(node, struct extent_state, rb_node);
553
hit_next:
554 555 556
	if (state->start > end)
		goto out;
	WARN_ON(state->end < start);
557
	last_end = state->end;
558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575

	/*
	 *     | ---- 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) {
576 577
		if (!prealloc)
			prealloc = alloc_extent_state(GFP_ATOMIC);
578 579 580 581 582 583
		err = split_state(tree, state, prealloc, start);
		BUG_ON(err == -EEXIST);
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
584
			set |= clear_state_bit(tree, state, &bits, wake);
585 586 587
			if (last_end == (u64)-1)
				goto out;
			start = last_end + 1;
588 589 590 591 592 593 594 595 596 597
		}
		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) {
598 599
		if (!prealloc)
			prealloc = alloc_extent_state(GFP_ATOMIC);
600 601 602 603
		err = split_state(tree, state, prealloc, end + 1);
		BUG_ON(err == -EEXIST);
		if (wake)
			wake_up(&state->wq);
604

605
		set |= clear_state_bit(tree, prealloc, &bits, wake);
J
Josef Bacik 已提交
606

607 608 609
		prealloc = NULL;
		goto out;
	}
610

611 612 613 614
	if (state->end < end && prealloc && !need_resched())
		next_node = rb_next(&state->rb_node);
	else
		next_node = NULL;
615

616
	set |= clear_state_bit(tree, state, &bits, wake);
617 618 619
	if (last_end == (u64)-1)
		goto out;
	start = last_end + 1;
620 621 622 623 624 625
	if (start <= end && next_node) {
		state = rb_entry(next_node, struct extent_state,
				 rb_node);
		if (state->start == start)
			goto hit_next;
	}
626 627 628
	goto search_again;

out:
629
	spin_unlock(&tree->lock);
630 631 632 633 634 635 636 637
	if (prealloc)
		free_extent_state(prealloc);

	return set;

search_again:
	if (start > end)
		goto out;
638
	spin_unlock(&tree->lock);
639 640 641 642 643 644 645
	if (mask & __GFP_WAIT)
		cond_resched();
	goto again;
}

static int wait_on_state(struct extent_io_tree *tree,
			 struct extent_state *state)
646 647
		__releases(tree->lock)
		__acquires(tree->lock)
648 649 650
{
	DEFINE_WAIT(wait);
	prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE);
651
	spin_unlock(&tree->lock);
652
	schedule();
653
	spin_lock(&tree->lock);
654 655 656 657 658 659 660 661 662 663 664 665 666 667
	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;

668
	spin_lock(&tree->lock);
669 670 671 672 673 674
again:
	while (1) {
		/*
		 * this search will find all the extents that end after
		 * our range starts
		 */
675
		node = tree_search(tree, start);
676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696
		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()) {
697
			spin_unlock(&tree->lock);
698
			cond_resched();
699
			spin_lock(&tree->lock);
700 701 702
		}
	}
out:
703
	spin_unlock(&tree->lock);
704 705 706
	return 0;
}

J
Josef Bacik 已提交
707
static int set_state_bits(struct extent_io_tree *tree,
708
			   struct extent_state *state,
709
			   int *bits)
710
{
J
Josef Bacik 已提交
711
	int ret;
712
	int bits_to_set = *bits & ~EXTENT_CTLBITS;
J
Josef Bacik 已提交
713 714 715 716

	ret = set_state_cb(tree, state, bits);
	if (ret)
		return ret;
717
	if ((bits_to_set & EXTENT_DIRTY) && !(state->state & EXTENT_DIRTY)) {
718 719 720
		u64 range = state->end - state->start + 1;
		tree->dirty_bytes += range;
	}
721
	state->state |= bits_to_set;
J
Josef Bacik 已提交
722 723

	return 0;
724 725
}

726 727 728 729 730 731 732 733 734 735 736
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);
		}
	}
}

737
/*
738 739
 * 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.
740
 *
741 742 743
 * 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.
744
 *
745
 * [start, end] is inclusive This takes the tree lock.
746
 */
747

748 749 750
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)
751 752 753 754 755 756 757
{
	struct extent_state *state;
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
	int err = 0;
	u64 last_start;
	u64 last_end;
758

759
	bits |= EXTENT_FIRST_DELALLOC;
760 761 762 763 764 765 766
again:
	if (!prealloc && (mask & __GFP_WAIT)) {
		prealloc = alloc_extent_state(mask);
		if (!prealloc)
			return -ENOMEM;
	}

767
	spin_lock(&tree->lock);
768 769 770 771 772 773 774
	if (cached_state && *cached_state) {
		state = *cached_state;
		if (state->start == start && state->tree) {
			node = &state->rb_node;
			goto hit_next;
		}
	}
775 776 777 778
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
779
	node = tree_search(tree, start);
780
	if (!node) {
781
		err = insert_state(tree, prealloc, start, end, &bits);
782 783 784 785 786
		prealloc = NULL;
		BUG_ON(err == -EEXIST);
		goto out;
	}
	state = rb_entry(node, struct extent_state, rb_node);
C
Chris Mason 已提交
787
hit_next:
788 789 790 791 792 793 794 795 796 797
	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
798
		struct rb_node *next_node;
799
		if (state->state & exclusive_bits) {
800 801 802 803
			*failed_start = state->start;
			err = -EEXIST;
			goto out;
		}
804

805
		err = set_state_bits(tree, state, &bits);
J
Josef Bacik 已提交
806 807 808
		if (err)
			goto out;

809
		cache_state(state, cached_state);
810
		merge_state(tree, state);
811 812
		if (last_end == (u64)-1)
			goto out;
C
Chris Mason 已提交
813

814
		start = last_end + 1;
C
Chris Mason 已提交
815 816 817 818 819 820 821 822 823
		if (start < end && prealloc && !need_resched()) {
			next_node = rb_next(node);
			if (next_node) {
				state = rb_entry(next_node, struct extent_state,
						 rb_node);
				if (state->start == start)
					goto hit_next;
			}
		}
824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843
		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) {
844
		if (state->state & exclusive_bits) {
845 846 847 848 849 850 851 852 853 854
			*failed_start = start;
			err = -EEXIST;
			goto out;
		}
		err = split_state(tree, state, prealloc, start);
		BUG_ON(err == -EEXIST);
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
855
			err = set_state_bits(tree, state, &bits);
J
Josef Bacik 已提交
856 857
			if (err)
				goto out;
858
			cache_state(state, cached_state);
859
			merge_state(tree, state);
860 861 862
			if (last_end == (u64)-1)
				goto out;
			start = last_end + 1;
863 864 865 866 867 868 869 870 871 872 873 874 875 876 877
		}
		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
878
			this_end = last_start - 1;
879
		err = insert_state(tree, prealloc, start, this_end,
880
				   &bits);
881
		BUG_ON(err == -EEXIST);
J
Josef Bacik 已提交
882 883
		if (err) {
			prealloc = NULL;
884
			goto out;
J
Josef Bacik 已提交
885 886 887
		}
		cache_state(prealloc, cached_state);
		prealloc = NULL;
888 889 890 891 892 893 894 895 896 897
		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) {
898
		if (state->state & exclusive_bits) {
899 900 901 902 903 904 905
			*failed_start = start;
			err = -EEXIST;
			goto out;
		}
		err = split_state(tree, state, prealloc, end + 1);
		BUG_ON(err == -EEXIST);

906
		err = set_state_bits(tree, prealloc, &bits);
J
Josef Bacik 已提交
907 908 909 910
		if (err) {
			prealloc = NULL;
			goto out;
		}
911
		cache_state(prealloc, cached_state);
912 913 914 915 916 917 918 919
		merge_state(tree, prealloc);
		prealloc = NULL;
		goto out;
	}

	goto search_again;

out:
920
	spin_unlock(&tree->lock);
921 922 923 924 925 926 927 928
	if (prealloc)
		free_extent_state(prealloc);

	return err;

search_again:
	if (start > end)
		goto out;
929
	spin_unlock(&tree->lock);
930 931 932 933 934 935 936 937 938 939
	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,
940
			      NULL, mask);
941 942 943 944 945 946
}

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,
947
			      NULL, mask);
948 949 950 951 952
}

int clear_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
		      int bits, gfp_t mask)
{
953
	return clear_extent_bit(tree, start, end, bits, 0, 0, NULL, mask);
954 955 956
}

int set_extent_delalloc(struct extent_io_tree *tree, u64 start, u64 end,
957
			struct extent_state **cached_state, gfp_t mask)
958 959
{
	return set_extent_bit(tree, start, end,
C
Chris Mason 已提交
960
			      EXTENT_DELALLOC | EXTENT_DIRTY | EXTENT_UPTODATE,
961
			      0, NULL, cached_state, mask);
962 963 964 965 966 967
}

int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
		       gfp_t mask)
{
	return clear_extent_bit(tree, start, end,
968
				EXTENT_DIRTY | EXTENT_DELALLOC |
969
				EXTENT_DO_ACCOUNTING, 0, 0, NULL, mask);
970 971 972 973 974 975
}

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,
976
			      NULL, mask);
977 978
}

979
static int clear_extent_new(struct extent_io_tree *tree, u64 start, u64 end,
980 981
		       gfp_t mask)
{
982 983
	return clear_extent_bit(tree, start, end, EXTENT_NEW, 0, 0,
				NULL, mask);
984 985 986 987 988 989
}

int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end,
			gfp_t mask)
{
	return set_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, NULL,
990
			      NULL, mask);
991 992
}

C
Chris Mason 已提交
993
static int clear_extent_uptodate(struct extent_io_tree *tree, u64 start,
994 995
				 u64 end, struct extent_state **cached_state,
				 gfp_t mask)
996
{
997
	return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0,
998
				cached_state, mask);
999 1000 1001 1002 1003 1004 1005
}

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

C
Chris Mason 已提交
1006 1007 1008 1009
/*
 * either insert or lock state struct between start and end use mask to tell
 * us if waiting is desired.
 */
1010
int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1011
		     int bits, struct extent_state **cached_state, gfp_t mask)
1012 1013 1014 1015
{
	int err;
	u64 failed_start;
	while (1) {
1016
		err = set_extent_bit(tree, start, end, EXTENT_LOCKED | bits,
1017 1018
				     EXTENT_LOCKED, &failed_start,
				     cached_state, mask);
1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029
		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;
}

1030 1031
int lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask)
{
1032
	return lock_extent_bits(tree, start, end, 0, NULL, mask);
1033 1034
}

1035 1036 1037 1038 1039 1040
int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end,
		    gfp_t mask)
{
	int err;
	u64 failed_start;

1041 1042
	err = set_extent_bit(tree, start, end, EXTENT_LOCKED, EXTENT_LOCKED,
			     &failed_start, NULL, mask);
Y
Yan Zheng 已提交
1043 1044 1045
	if (err == -EEXIST) {
		if (failed_start > start)
			clear_extent_bit(tree, start, failed_start - 1,
1046
					 EXTENT_LOCKED, 1, 0, NULL, mask);
1047
		return 0;
Y
Yan Zheng 已提交
1048
	}
1049 1050 1051
	return 1;
}

1052 1053 1054 1055 1056 1057 1058
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);
}

1059 1060 1061
int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end,
		  gfp_t mask)
{
1062 1063
	return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL,
				mask);
1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087
}

/*
 * helper function to set pages and extents in the tree dirty
 */
int set_range_dirty(struct extent_io_tree *tree, u64 start, u64 end)
{
	unsigned long index = start >> PAGE_CACHE_SHIFT;
	unsigned long end_index = end >> PAGE_CACHE_SHIFT;
	struct page *page;

	while (index <= end_index) {
		page = find_get_page(tree->mapping, index);
		BUG_ON(!page);
		__set_page_dirty_nobuffers(page);
		page_cache_release(page);
		index++;
	}
	return 0;
}

/*
 * helper function to set both pages and extents in the tree writeback
 */
1088
static int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end)
1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103
{
	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
1104 1105 1106 1107 1108 1109 1110
/*
 * 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
 */
1111 1112 1113 1114 1115 1116 1117
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;

1118
	spin_lock(&tree->lock);
1119 1120 1121 1122
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1123
	node = tree_search(tree, start);
C
Chris Mason 已提交
1124
	if (!node)
1125 1126
		goto out;

C
Chris Mason 已提交
1127
	while (1) {
1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139
		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:
1140
	spin_unlock(&tree->lock);
1141 1142 1143
	return ret;
}

C
Chris Mason 已提交
1144 1145 1146 1147
/* 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
1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158
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
1159
	if (!node)
C
Chris Mason 已提交
1160 1161
		goto out;

C
Chris Mason 已提交
1162
	while (1) {
C
Chris Mason 已提交
1163
		state = rb_entry(node, struct extent_state, rb_node);
C
Chris Mason 已提交
1164
		if (state->end >= start && (state->state & bits))
C
Chris Mason 已提交
1165
			return state;
C
Chris Mason 已提交
1166

C
Chris Mason 已提交
1167 1168 1169 1170 1171 1172 1173 1174
		node = rb_next(node);
		if (!node)
			break;
	}
out:
	return NULL;
}

C
Chris Mason 已提交
1175 1176 1177 1178 1179 1180
/*
 * 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
1181
static noinline u64 find_delalloc_range(struct extent_io_tree *tree,
1182 1183
					u64 *start, u64 *end, u64 max_bytes,
					struct extent_state **cached_state)
1184 1185 1186 1187 1188 1189 1190
{
	struct rb_node *node;
	struct extent_state *state;
	u64 cur_start = *start;
	u64 found = 0;
	u64 total_bytes = 0;

1191
	spin_lock(&tree->lock);
C
Chris Mason 已提交
1192

1193 1194 1195 1196
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1197
	node = tree_search(tree, cur_start);
1198
	if (!node) {
1199 1200
		if (!found)
			*end = (u64)-1;
1201 1202 1203
		goto out;
	}

C
Chris Mason 已提交
1204
	while (1) {
1205
		state = rb_entry(node, struct extent_state, rb_node);
1206 1207
		if (found && (state->start != cur_start ||
			      (state->state & EXTENT_BOUNDARY))) {
1208 1209 1210 1211 1212 1213 1214
			goto out;
		}
		if (!(state->state & EXTENT_DELALLOC)) {
			if (!found)
				*end = state->end;
			goto out;
		}
1215
		if (!found) {
1216
			*start = state->start;
1217 1218 1219
			*cached_state = state;
			atomic_inc(&state->refs);
		}
1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230
		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:
1231
	spin_unlock(&tree->lock);
1232 1233 1234
	return found;
}

C
Chris Mason 已提交
1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248
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
1249
	while (nr_pages > 0) {
C
Chris Mason 已提交
1250
		ret = find_get_pages_contig(inode->i_mapping, index,
1251 1252
				     min_t(unsigned long, nr_pages,
				     ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284
		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
1285
	while (nrpages > 0) {
C
Chris Mason 已提交
1286
		ret = find_get_pages_contig(inode->i_mapping, index,
1287 1288
				     min_t(unsigned long,
				     nrpages, ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1289 1290 1291 1292 1293 1294 1295 1296 1297 1298
		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
			 */
1299
			if (pages[i] != locked_page) {
C
Chris Mason 已提交
1300
				lock_page(pages[i]);
1301 1302
				if (!PageDirty(pages[i]) ||
				    pages[i]->mapping != inode->i_mapping) {
1303 1304 1305 1306 1307 1308
					ret = -EAGAIN;
					unlock_page(pages[i]);
					page_cache_release(pages[i]);
					goto done;
				}
			}
C
Chris Mason 已提交
1309
			page_cache_release(pages[i]);
1310
			pages_locked++;
C
Chris Mason 已提交
1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341
		}
		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;
1342
	struct extent_state *cached_state = NULL;
C
Chris Mason 已提交
1343 1344 1345 1346 1347 1348 1349 1350
	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,
1351
				    max_bytes, &cached_state);
C
Chris Mason 已提交
1352
	if (!found || delalloc_end <= *start) {
C
Chris Mason 已提交
1353 1354
		*start = delalloc_start;
		*end = delalloc_end;
1355
		free_extent_state(cached_state);
C
Chris Mason 已提交
1356 1357 1358
		return found;
	}

C
Chris Mason 已提交
1359 1360 1361 1362 1363
	/*
	 * 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
1364
	if (delalloc_start < *start)
C
Chris Mason 已提交
1365 1366
		delalloc_start = *start;

C
Chris Mason 已提交
1367 1368 1369 1370
	/*
	 * make sure to limit the number of pages we try to lock down
	 * if we're looping.
	 */
C
Chris Mason 已提交
1371
	if (delalloc_end + 1 - delalloc_start > max_bytes && loops)
1372
		delalloc_end = delalloc_start + PAGE_CACHE_SIZE - 1;
C
Chris Mason 已提交
1373

C
Chris Mason 已提交
1374 1375 1376 1377 1378 1379 1380
	/* 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
		 */
1381
		free_extent_state(cached_state);
C
Chris Mason 已提交
1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394
		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 */
1395 1396
	lock_extent_bits(tree, delalloc_start, delalloc_end,
			 0, &cached_state, GFP_NOFS);
C
Chris Mason 已提交
1397 1398 1399

	/* then test to make sure it is all still delalloc */
	ret = test_range_bit(tree, delalloc_start, delalloc_end,
1400
			     EXTENT_DELALLOC, 1, cached_state);
C
Chris Mason 已提交
1401
	if (!ret) {
1402 1403
		unlock_extent_cached(tree, delalloc_start, delalloc_end,
				     &cached_state, GFP_NOFS);
C
Chris Mason 已提交
1404 1405 1406 1407 1408
		__unlock_for_delalloc(inode, locked_page,
			      delalloc_start, delalloc_end);
		cond_resched();
		goto again;
	}
1409
	free_extent_state(cached_state);
C
Chris Mason 已提交
1410 1411 1412 1413 1414 1415 1416 1417 1418
	*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,
1419
				unsigned long op)
C
Chris Mason 已提交
1420 1421 1422 1423 1424 1425 1426
{
	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;
1427
	int clear_bits = 0;
C
Chris Mason 已提交
1428

1429
	if (op & EXTENT_CLEAR_UNLOCK)
1430
		clear_bits |= EXTENT_LOCKED;
1431
	if (op & EXTENT_CLEAR_DIRTY)
C
Chris Mason 已提交
1432 1433
		clear_bits |= EXTENT_DIRTY;

1434
	if (op & EXTENT_CLEAR_DELALLOC)
1435 1436
		clear_bits |= EXTENT_DELALLOC;

1437
	clear_extent_bit(tree, start, end, clear_bits, 1, 0, NULL, GFP_NOFS);
1438 1439 1440
	if (!(op & (EXTENT_CLEAR_UNLOCK_PAGE | EXTENT_CLEAR_DIRTY |
		    EXTENT_SET_WRITEBACK | EXTENT_END_WRITEBACK |
		    EXTENT_SET_PRIVATE2)))
1441
		return 0;
C
Chris Mason 已提交
1442

C
Chris Mason 已提交
1443
	while (nr_pages > 0) {
C
Chris Mason 已提交
1444
		ret = find_get_pages_contig(inode->i_mapping, index,
1445 1446
				     min_t(unsigned long,
				     nr_pages, ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1447
		for (i = 0; i < ret; i++) {
1448

1449
			if (op & EXTENT_SET_PRIVATE2)
1450 1451
				SetPagePrivate2(pages[i]);

C
Chris Mason 已提交
1452 1453 1454 1455
			if (pages[i] == locked_page) {
				page_cache_release(pages[i]);
				continue;
			}
1456
			if (op & EXTENT_CLEAR_DIRTY)
C
Chris Mason 已提交
1457
				clear_page_dirty_for_io(pages[i]);
1458
			if (op & EXTENT_SET_WRITEBACK)
C
Chris Mason 已提交
1459
				set_page_writeback(pages[i]);
1460
			if (op & EXTENT_END_WRITEBACK)
C
Chris Mason 已提交
1461
				end_page_writeback(pages[i]);
1462
			if (op & EXTENT_CLEAR_UNLOCK_PAGE)
1463
				unlock_page(pages[i]);
C
Chris Mason 已提交
1464 1465 1466 1467 1468 1469 1470 1471 1472
			page_cache_release(pages[i]);
		}
		nr_pages -= ret;
		index += ret;
		cond_resched();
	}
	return 0;
}

C
Chris Mason 已提交
1473 1474 1475 1476 1477
/*
 * 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.
 */
1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492
u64 count_range_bits(struct extent_io_tree *tree,
		     u64 *start, u64 search_end, u64 max_bytes,
		     unsigned long bits)
{
	struct rb_node *node;
	struct extent_state *state;
	u64 cur_start = *start;
	u64 total_bytes = 0;
	int found = 0;

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

1493
	spin_lock(&tree->lock);
1494 1495 1496 1497 1498 1499 1500 1501
	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.
	 */
1502
	node = tree_search(tree, cur_start);
C
Chris Mason 已提交
1503
	if (!node)
1504 1505
		goto out;

C
Chris Mason 已提交
1506
	while (1) {
1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524
		state = rb_entry(node, struct extent_state, rb_node);
		if (state->start > search_end)
			break;
		if (state->end >= cur_start && (state->state & bits)) {
			total_bytes += min(search_end, state->end) + 1 -
				       max(cur_start, state->start);
			if (total_bytes >= max_bytes)
				break;
			if (!found) {
				*start = state->start;
				found = 1;
			}
		}
		node = rb_next(node);
		if (!node)
			break;
	}
out:
1525
	spin_unlock(&tree->lock);
1526 1527
	return total_bytes;
}
1528

C
Chris Mason 已提交
1529 1530 1531 1532
/*
 * set the private field for a given byte offset in the tree.  If there isn't
 * an extent_state there already, this does nothing.
 */
1533 1534 1535 1536 1537 1538
int set_state_private(struct extent_io_tree *tree, u64 start, u64 private)
{
	struct rb_node *node;
	struct extent_state *state;
	int ret = 0;

1539
	spin_lock(&tree->lock);
1540 1541 1542 1543
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1544
	node = tree_search(tree, start);
1545
	if (!node) {
1546 1547 1548 1549 1550 1551 1552 1553 1554 1555
		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:
1556
	spin_unlock(&tree->lock);
1557 1558 1559 1560 1561 1562 1563 1564 1565
	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;

1566
	spin_lock(&tree->lock);
1567 1568 1569 1570
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1571
	node = tree_search(tree, start);
1572
	if (!node) {
1573 1574 1575 1576 1577 1578 1579 1580 1581 1582
		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:
1583
	spin_unlock(&tree->lock);
1584 1585 1586 1587 1588
	return ret;
}

/*
 * searches a range in the state tree for a given mask.
1589
 * If 'filled' == 1, this returns 1 only if every extent in the tree
1590 1591 1592 1593
 * 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,
1594
		   int bits, int filled, struct extent_state *cached)
1595 1596 1597 1598 1599
{
	struct extent_state *state = NULL;
	struct rb_node *node;
	int bitset = 0;

1600
	spin_lock(&tree->lock);
1601 1602 1603 1604
	if (cached && cached->tree && cached->start == start)
		node = &cached->rb_node;
	else
		node = tree_search(tree, start);
1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623
	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;
		}
1624 1625 1626 1627

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

1628 1629 1630 1631 1632 1633 1634 1635 1636 1637
		start = state->end + 1;
		if (start > end)
			break;
		node = rb_next(node);
		if (!node) {
			if (filled)
				bitset = 0;
			break;
		}
	}
1638
	spin_unlock(&tree->lock);
1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650
	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;
1651
	if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1, NULL))
1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664
		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;
1665
	if (!test_range_bit(tree, start, end, EXTENT_LOCKED, 0, NULL))
1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676
		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)
{
1677
	end_page_writeback(page);
1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693
	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)
{
1694
	int uptodate = err == 0;
1695
	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
1696
	struct extent_io_tree *tree;
1697 1698 1699
	u64 start;
	u64 end;
	int whole_page;
1700
	int ret;
1701 1702 1703

	do {
		struct page *page = bvec->bv_page;
1704 1705
		tree = &BTRFS_I(page->mapping->host)->io_tree;

1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716
		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);
1717 1718
		if (tree->ops && tree->ops->writepage_end_io_hook) {
			ret = tree->ops->writepage_end_io_hook(page, start,
1719
						       end, NULL, uptodate);
1720 1721 1722 1723 1724 1725 1726
			if (ret)
				uptodate = 0;
		}

		if (!uptodate && tree->ops &&
		    tree->ops->writepage_io_failed_hook) {
			ret = tree->ops->writepage_io_failed_hook(bio, page,
1727
							 start, end, NULL);
1728 1729 1730 1731 1732 1733
			if (ret == 0) {
				uptodate = (err == 0);
				continue;
			}
		}

1734
		if (!uptodate) {
1735
			clear_extent_uptodate(tree, start, end, NULL, GFP_NOFS);
1736 1737 1738
			ClearPageUptodate(page);
			SetPageError(page);
		}
1739

1740 1741 1742 1743 1744
		if (whole_page)
			end_page_writeback(page);
		else
			check_page_writeback(tree, page);
	} while (bvec >= bio->bi_io_vec);
1745

1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762
	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);
1763 1764
	struct bio_vec *bvec_end = bio->bi_io_vec + bio->bi_vcnt - 1;
	struct bio_vec *bvec = bio->bi_io_vec;
1765
	struct extent_io_tree *tree;
1766 1767 1768 1769 1770
	u64 start;
	u64 end;
	int whole_page;
	int ret;

1771 1772 1773
	if (err)
		uptodate = 0;

1774 1775
	do {
		struct page *page = bvec->bv_page;
1776 1777
		tree = &BTRFS_I(page->mapping->host)->io_tree;

1778 1779 1780 1781 1782 1783 1784 1785 1786
		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;

1787
		if (++bvec <= bvec_end)
1788 1789 1790
			prefetchw(&bvec->bv_page->flags);

		if (uptodate && tree->ops && tree->ops->readpage_end_io_hook) {
1791
			ret = tree->ops->readpage_end_io_hook(page, start, end,
1792
							      NULL);
1793 1794 1795
			if (ret)
				uptodate = 0;
		}
1796 1797 1798
		if (!uptodate && tree->ops &&
		    tree->ops->readpage_io_failed_hook) {
			ret = tree->ops->readpage_io_failed_hook(bio, page,
1799
							 start, end, NULL);
1800
			if (ret == 0) {
1801 1802
				uptodate =
					test_bit(BIO_UPTODATE, &bio->bi_flags);
1803 1804
				if (err)
					uptodate = 0;
1805 1806 1807
				continue;
			}
		}
1808

1809
		if (uptodate) {
1810 1811
			set_extent_uptodate(tree, start, end,
					    GFP_ATOMIC);
1812
		}
1813
		unlock_extent(tree, start, end, GFP_ATOMIC);
1814

1815 1816 1817 1818 1819 1820 1821
		if (whole_page) {
			if (uptodate) {
				SetPageUptodate(page);
			} else {
				ClearPageUptodate(page);
				SetPageError(page);
			}
1822
			unlock_page(page);
1823 1824 1825 1826 1827 1828 1829
		} else {
			if (uptodate) {
				check_page_uptodate(tree, page);
			} else {
				ClearPageUptodate(page);
				SetPageError(page);
			}
1830
			check_page_locked(tree, page);
1831
		}
1832
	} while (bvec <= bvec_end);
1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845

	bio_put(bio);
}

/*
 * IO done from prepare_write is pretty simple, we just unlock
 * the structs in the extent tree when done, and set the uptodate bits
 * as appropriate.
 */
static void end_bio_extent_preparewrite(struct bio *bio, int err)
{
	const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
1846
	struct extent_io_tree *tree;
1847 1848 1849 1850 1851
	u64 start;
	u64 end;

	do {
		struct page *page = bvec->bv_page;
1852 1853
		tree = &BTRFS_I(page->mapping->host)->io_tree;

1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888
		start = ((u64)page->index << PAGE_CACHE_SHIFT) +
			bvec->bv_offset;
		end = start + bvec->bv_len - 1;

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

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

		unlock_extent(tree, start, end, GFP_ATOMIC);

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

	bio_put(bio);
}

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

	bio = bio_alloc(gfp_flags, nr_vecs);

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

	if (bio) {
1889
		bio->bi_size = 0;
1890 1891 1892 1893 1894 1895
		bio->bi_bdev = bdev;
		bio->bi_sector = first_sector;
	}
	return bio;
}

C
Chris Mason 已提交
1896 1897
static int submit_one_bio(int rw, struct bio *bio, int mirror_num,
			  unsigned long bio_flags)
1898 1899
{
	int ret = 0;
1900 1901 1902 1903 1904 1905 1906 1907 1908
	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
	struct page *page = bvec->bv_page;
	struct extent_io_tree *tree = bio->bi_private;
	u64 start;
	u64 end;

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

1909
	bio->bi_private = NULL;
1910 1911 1912

	bio_get(bio);

1913
	if (tree->ops && tree->ops->submit_bio_hook)
1914
		tree->ops->submit_bio_hook(page->mapping->host, rw, bio,
1915
					   mirror_num, bio_flags, start);
1916 1917
	else
		submit_bio(rw, bio);
1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929
	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,
1930
			      bio_end_io_t end_io_func,
C
Chris Mason 已提交
1931 1932 1933
			      int mirror_num,
			      unsigned long prev_bio_flags,
			      unsigned long bio_flags)
1934 1935 1936 1937
{
	int ret = 0;
	struct bio *bio;
	int nr;
C
Chris Mason 已提交
1938 1939 1940
	int contig = 0;
	int this_compressed = bio_flags & EXTENT_BIO_COMPRESSED;
	int old_compressed = prev_bio_flags & EXTENT_BIO_COMPRESSED;
1941
	size_t page_size = min_t(size_t, size, PAGE_CACHE_SIZE);
1942 1943 1944

	if (bio_ret && *bio_ret) {
		bio = *bio_ret;
C
Chris Mason 已提交
1945 1946 1947 1948 1949 1950 1951
		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 ||
1952
		    (tree->ops && tree->ops->merge_bio_hook &&
C
Chris Mason 已提交
1953 1954 1955 1956 1957
		     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);
1958 1959 1960 1961 1962
			bio = NULL;
		} else {
			return 0;
		}
	}
C
Chris Mason 已提交
1963 1964 1965 1966 1967
	if (this_compressed)
		nr = BIO_MAX_PAGES;
	else
		nr = bio_get_nr_vecs(bdev);

1968
	bio = extent_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH);
1969

C
Chris Mason 已提交
1970
	bio_add_page(bio, page, page_size, offset);
1971 1972
	bio->bi_end_io = end_io_func;
	bio->bi_private = tree;
1973

C
Chris Mason 已提交
1974
	if (bio_ret)
1975
		*bio_ret = bio;
C
Chris Mason 已提交
1976
	else
C
Chris Mason 已提交
1977
		ret = submit_one_bio(rw, bio, mirror_num, bio_flags);
1978 1979 1980 1981 1982 1983 1984 1985 1986

	return ret;
}

void set_page_extent_mapped(struct page *page)
{
	if (!PagePrivate(page)) {
		SetPagePrivate(page);
		page_cache_get(page);
1987
		set_page_private(page, EXTENT_PAGE_PRIVATE);
1988 1989 1990
	}
}

1991
static void set_page_extent_head(struct page *page, unsigned long len)
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
{
	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
2004 2005
				   struct bio **bio, int mirror_num,
				   unsigned long *bio_flags)
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
{
	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;
2019
	struct btrfs_ordered_extent *ordered;
2020 2021 2022 2023
	int ret;
	int nr = 0;
	size_t page_offset = 0;
	size_t iosize;
C
Chris Mason 已提交
2024
	size_t disk_io_size;
2025
	size_t blocksize = inode->i_sb->s_blocksize;
C
Chris Mason 已提交
2026
	unsigned long this_bio_flag = 0;
2027 2028 2029 2030

	set_page_extent_mapped(page);

	end = page_end;
2031 2032 2033 2034 2035 2036 2037 2038 2039
	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);
	}
2040

C
Chris Mason 已提交
2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052
	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);
		}
	}
2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076
	while (cur <= end) {
		if (cur >= last_byte) {
			char *userpage;
			iosize = PAGE_CACHE_SIZE - page_offset;
			userpage = kmap_atomic(page, KM_USER0);
			memset(userpage + page_offset, 0, iosize);
			flush_dcache_page(page);
			kunmap_atomic(userpage, KM_USER0);
			set_extent_uptodate(tree, cur, cur + iosize - 1,
					    GFP_NOFS);
			unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
			break;
		}
		em = get_extent(inode, page, page_offset, cur,
				end - cur + 1, 0);
		if (IS_ERR(em) || !em) {
			SetPageError(page);
			unlock_extent(tree, cur, end, GFP_NOFS);
			break;
		}
		extent_offset = cur - em->start;
		BUG_ON(extent_map_end(em) <= cur);
		BUG_ON(end < cur);

C
Chris Mason 已提交
2077 2078 2079
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
			this_bio_flag = EXTENT_BIO_COMPRESSED;

2080 2081 2082
		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
2083 2084 2085 2086 2087 2088 2089
		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;
		}
2090 2091
		bdev = em->bdev;
		block_start = em->block_start;
Y
Yan Zheng 已提交
2092 2093
		if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			block_start = EXTENT_MAP_HOLE;
2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112
		free_extent_map(em);
		em = NULL;

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

			set_extent_uptodate(tree, cur, cur + iosize - 1,
					    GFP_NOFS);
			unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
			cur = cur + iosize;
			page_offset += iosize;
			continue;
		}
		/* the get_extent function already copied into the page */
2113 2114
		if (test_range_bit(tree, cur, cur_end,
				   EXTENT_UPTODATE, 1, NULL)) {
2115
			check_page_uptodate(tree, page);
2116 2117 2118 2119 2120
			unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
			cur = cur + iosize;
			page_offset += iosize;
			continue;
		}
2121 2122 2123 2124 2125 2126 2127 2128 2129 2130
		/* we have an inline extent but it didn't get marked up
		 * to date.  Error out
		 */
		if (block_start == EXTENT_MAP_INLINE) {
			SetPageError(page);
			unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
			cur = cur + iosize;
			page_offset += iosize;
			continue;
		}
2131 2132 2133 2134 2135 2136 2137

		ret = 0;
		if (tree->ops && tree->ops->readpage_io_hook) {
			ret = tree->ops->readpage_io_hook(page, cur,
							  cur + iosize - 1);
		}
		if (!ret) {
2138 2139
			unsigned long pnr = (last_byte >> PAGE_CACHE_SHIFT) + 1;
			pnr -= page->index;
2140
			ret = submit_extent_page(READ, tree, page,
C
Chris Mason 已提交
2141
					 sector, disk_io_size, page_offset,
2142
					 bdev, bio, pnr,
C
Chris Mason 已提交
2143 2144 2145
					 end_bio_extent_readpage, mirror_num,
					 *bio_flags,
					 this_bio_flag);
2146
			nr++;
C
Chris Mason 已提交
2147
			*bio_flags = this_bio_flag;
2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165
		}
		if (ret)
			SetPageError(page);
		cur = cur + iosize;
		page_offset += iosize;
	}
	if (!nr) {
		if (!PageError(page))
			SetPageUptodate(page);
		unlock_page(page);
	}
	return 0;
}

int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
			    get_extent_t *get_extent)
{
	struct bio *bio = NULL;
C
Chris Mason 已提交
2166
	unsigned long bio_flags = 0;
2167 2168
	int ret;

C
Chris Mason 已提交
2169 2170
	ret = __extent_read_full_page(tree, page, get_extent, &bio, 0,
				      &bio_flags);
2171
	if (bio)
C
Chris Mason 已提交
2172
		submit_one_bio(READ, bio, 0, bio_flags);
2173 2174 2175
	return ret;
}

2176 2177 2178 2179 2180 2181 2182 2183 2184 2185
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;
}

2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206
/*
 * 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;
2207
	u64 unlock_start;
2208
	sector_t sector;
2209
	struct extent_state *cached_state = NULL;
2210 2211 2212 2213
	struct extent_map *em;
	struct block_device *bdev;
	int ret;
	int nr = 0;
2214
	size_t pg_offset = 0;
2215 2216 2217 2218 2219
	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
2220 2221
	int page_started;
	int compressed;
2222
	int write_flags;
2223
	unsigned long nr_written = 0;
2224

2225 2226 2227 2228 2229
	if (wbc->sync_mode == WB_SYNC_ALL)
		write_flags = WRITE_SYNC_PLUG;
	else
		write_flags = WRITE;

2230
	WARN_ON(!PageLocked(page));
2231
	pg_offset = i_size & (PAGE_CACHE_SIZE - 1);
2232
	if (page->index > end_index ||
2233
	   (page->index == end_index && !pg_offset)) {
2234
		page->mapping->a_ops->invalidatepage(page, 0);
2235 2236 2237 2238 2239 2240 2241 2242
		unlock_page(page);
		return 0;
	}

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

		userpage = kmap_atomic(page, KM_USER0);
2243 2244
		memset(userpage + pg_offset, 0,
		       PAGE_CACHE_SIZE - pg_offset);
2245
		kunmap_atomic(userpage, KM_USER0);
2246
		flush_dcache_page(page);
2247
	}
2248
	pg_offset = 0;
2249 2250 2251 2252 2253

	set_page_extent_mapped(page);

	delalloc_start = start;
	delalloc_end = 0;
C
Chris Mason 已提交
2254
	page_started = 0;
2255
	if (!epd->extent_locked) {
2256
		u64 delalloc_to_write = 0;
2257 2258 2259 2260 2261 2262
		/*
		 * 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
2263
		while (delalloc_end < page_end) {
2264
			nr_delalloc = find_lock_delalloc_range(inode, tree,
C
Chris Mason 已提交
2265 2266
						       page,
						       &delalloc_start,
2267 2268
						       &delalloc_end,
						       128 * 1024 * 1024);
2269 2270 2271 2272 2273 2274 2275
			if (nr_delalloc == 0) {
				delalloc_start = delalloc_end + 1;
				continue;
			}
			tree->ops->fill_delalloc(inode, page, delalloc_start,
						 delalloc_end, &page_started,
						 &nr_written);
2276 2277 2278 2279 2280 2281 2282 2283
			/*
			 * 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;
2284 2285
			delalloc_start = delalloc_end + 1;
		}
2286 2287 2288 2289 2290 2291 2292 2293
		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
2294

2295 2296 2297 2298 2299
		/* did the fill delalloc function already unlock and start
		 * the IO?
		 */
		if (page_started) {
			ret = 0;
2300 2301 2302 2303 2304 2305 2306
			/*
			 * 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;
2307
		}
C
Chris Mason 已提交
2308
	}
2309
	if (tree->ops && tree->ops->writepage_start_hook) {
C
Chris Mason 已提交
2310 2311
		ret = tree->ops->writepage_start_hook(page, start,
						      page_end);
2312 2313
		if (ret == -EAGAIN) {
			redirty_page_for_writepage(wbc, page);
2314
			update_nr_written(page, wbc, nr_written);
2315
			unlock_page(page);
2316
			ret = 0;
2317
			goto done_unlocked;
2318 2319 2320
		}
	}

2321 2322 2323 2324 2325
	/*
	 * 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);
2326

2327 2328
	end = page_end;
	if (last_byte <= start) {
2329 2330 2331 2332
		if (tree->ops && tree->ops->writepage_end_io_hook)
			tree->ops->writepage_end_io_hook(page, start,
							 page_end, NULL, 1);
		unlock_start = page_end + 1;
2333 2334 2335 2336 2337 2338 2339
		goto done;
	}

	blocksize = inode->i_sb->s_blocksize;

	while (cur <= end) {
		if (cur >= last_byte) {
2340 2341 2342 2343
			if (tree->ops && tree->ops->writepage_end_io_hook)
				tree->ops->writepage_end_io_hook(page, cur,
							 page_end, NULL, 1);
			unlock_start = page_end + 1;
2344 2345
			break;
		}
2346
		em = epd->get_extent(inode, page, pg_offset, cur,
2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360
				     end - cur + 1, 1);
		if (IS_ERR(em) || !em) {
			SetPageError(page);
			break;
		}

		extent_offset = cur - em->start;
		BUG_ON(extent_map_end(em) <= cur);
		BUG_ON(end < cur);
		iosize = min(extent_map_end(em) - cur, end - cur + 1);
		iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1);
		sector = (em->block_start + extent_offset) >> 9;
		bdev = em->bdev;
		block_start = em->block_start;
C
Chris Mason 已提交
2361
		compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
2362 2363 2364
		free_extent_map(em);
		em = NULL;

C
Chris Mason 已提交
2365 2366 2367 2368 2369
		/*
		 * compressed and inline extents are written through other
		 * paths in the FS
		 */
		if (compressed || block_start == EXTENT_MAP_HOLE ||
2370
		    block_start == EXTENT_MAP_INLINE) {
C
Chris Mason 已提交
2371 2372 2373 2374 2375 2376
			/*
			 * end_io notification does not happen here for
			 * compressed extents
			 */
			if (!compressed && tree->ops &&
			    tree->ops->writepage_end_io_hook)
2377 2378 2379
				tree->ops->writepage_end_io_hook(page, cur,
							 cur + iosize - 1,
							 NULL, 1);
C
Chris Mason 已提交
2380 2381 2382 2383 2384 2385 2386 2387 2388
			else if (compressed) {
				/* we don't want to end_page_writeback on
				 * a compressed extent.  this happens
				 * elsewhere
				 */
				nr++;
			}

			cur += iosize;
2389
			pg_offset += iosize;
2390
			unlock_start = cur;
2391 2392 2393 2394
			continue;
		}
		/* leave this out until we have a page_mkwrite call */
		if (0 && !test_range_bit(tree, cur, cur + iosize - 1,
2395
				   EXTENT_DIRTY, 0, NULL)) {
2396
			cur = cur + iosize;
2397
			pg_offset += iosize;
2398 2399
			continue;
		}
C
Chris Mason 已提交
2400

2401 2402 2403 2404 2405 2406
		if (tree->ops && tree->ops->writepage_io_hook) {
			ret = tree->ops->writepage_io_hook(page, cur,
						cur + iosize - 1);
		} else {
			ret = 0;
		}
2407
		if (ret) {
2408
			SetPageError(page);
2409
		} else {
2410
			unsigned long max_nr = end_index + 1;
2411

2412 2413
			set_range_writeback(tree, cur, cur + iosize - 1);
			if (!PageWriteback(page)) {
C
Chris Mason 已提交
2414 2415 2416
				printk(KERN_ERR "btrfs warning page %lu not "
				       "writeback, cur %llu end %llu\n",
				       page->index, (unsigned long long)cur,
2417 2418 2419
				       (unsigned long long)end);
			}

2420 2421 2422
			ret = submit_extent_page(write_flags, tree, page,
						 sector, iosize, pg_offset,
						 bdev, &epd->bio, max_nr,
C
Chris Mason 已提交
2423 2424
						 end_bio_extent_writepage,
						 0, 0, 0);
2425 2426 2427 2428
			if (ret)
				SetPageError(page);
		}
		cur = cur + iosize;
2429
		pg_offset += iosize;
2430 2431 2432 2433 2434 2435 2436 2437 2438
		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);
2439

2440 2441
done_unlocked:

2442 2443
	/* drop our reference on any cached states */
	free_extent_state(cached_state);
2444 2445 2446 2447
	return 0;
}

/**
C
Chris Mason 已提交
2448
 * write_cache_pages - walk the list of dirty pages of the given address space and write all of them.
2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461
 * @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.
 */
2462
static int extent_write_cache_pages(struct extent_io_tree *tree,
C
Chris Mason 已提交
2463 2464
			     struct address_space *mapping,
			     struct writeback_control *wbc,
C
Chris Mason 已提交
2465 2466
			     writepage_t writepage, void *data,
			     void (*flush_fn)(void *))
2467 2468 2469
{
	int ret = 0;
	int done = 0;
2470
	int nr_to_write_done = 0;
2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489
	struct pagevec pvec;
	int nr_pages;
	pgoff_t index;
	pgoff_t end;		/* Inclusive */
	int scanned = 0;
	int range_whole = 0;

	pagevec_init(&pvec, 0);
	if (wbc->range_cyclic) {
		index = mapping->writeback_index; /* Start from prev offset */
		end = -1;
	} else {
		index = wbc->range_start >> PAGE_CACHE_SHIFT;
		end = wbc->range_end >> PAGE_CACHE_SHIFT;
		if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
			range_whole = 1;
		scanned = 1;
	}
retry:
2490
	while (!done && !nr_to_write_done && (index <= end) &&
2491
	       (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index,
C
Chris Mason 已提交
2492 2493
			      PAGECACHE_TAG_DIRTY, min(end - index,
				  (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506
		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
2507 2508 2509 2510
			if (tree->ops && tree->ops->write_cache_pages_lock_hook)
				tree->ops->write_cache_pages_lock_hook(page);
			else
				lock_page(page);
2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522

			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
2523
			if (wbc->sync_mode != WB_SYNC_NONE) {
2524 2525
				if (PageWriteback(page))
					flush_fn(data);
2526
				wait_on_page_writeback(page);
C
Chris Mason 已提交
2527
			}
2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540

			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;
			}
2541
			if (ret)
2542
				done = 1;
2543 2544 2545 2546 2547 2548 2549

			/*
			 * 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;
2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565
		}
		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;
}

2566
static void flush_epd_write_bio(struct extent_page_data *epd)
C
Chris Mason 已提交
2567 2568
{
	if (epd->bio) {
2569 2570 2571 2572
		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
2573 2574 2575 2576
		epd->bio = NULL;
	}
}

2577 2578 2579 2580 2581 2582
static noinline void flush_write_bio(void *data)
{
	struct extent_page_data *epd = data;
	flush_epd_write_bio(epd);
}

2583 2584 2585 2586 2587 2588 2589 2590 2591 2592
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,
2593
		.extent_locked = 0,
2594
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
2595 2596
	};
	struct writeback_control wbc_writepages = {
2597
		.sync_mode	= wbc->sync_mode,
2598 2599 2600 2601 2602 2603 2604 2605
		.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
2606
	extent_write_cache_pages(tree, mapping, &wbc_writepages,
C
Chris Mason 已提交
2607
				 __extent_writepage, &epd, flush_write_bio);
2608
	flush_epd_write_bio(&epd);
2609 2610 2611
	return ret;
}

2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626
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,
2627
		.sync_io = mode == WB_SYNC_ALL,
2628 2629 2630 2631 2632 2633 2634 2635 2636
	};
	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
2637
	while (start <= end) {
2638 2639 2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651
		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;
	}

2652
	flush_epd_write_bio(&epd);
2653 2654
	return ret;
}
2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665

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,
2666
		.extent_locked = 0,
2667
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
2668 2669
	};

C
Chris Mason 已提交
2670
	ret = extent_write_cache_pages(tree, mapping, wbc,
C
Chris Mason 已提交
2671 2672
				       __extent_writepage, &epd,
				       flush_write_bio);
2673
	flush_epd_write_bio(&epd);
2674 2675 2676 2677 2678 2679 2680 2681 2682 2683
	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
2684
	unsigned long bio_flags = 0;
2685 2686 2687 2688 2689 2690

	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);
2691
		if (!add_to_page_cache_lru(page, mapping,
2692
					page->index, GFP_KERNEL)) {
2693
			__extent_read_full_page(tree, page, get_extent,
C
Chris Mason 已提交
2694
						&bio, 0, &bio_flags);
2695 2696 2697 2698 2699
		}
		page_cache_release(page);
	}
	BUG_ON(!list_empty(pages));
	if (bio)
C
Chris Mason 已提交
2700
		submit_one_bio(READ, bio, 0, bio_flags);
2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711
	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)
{
2712
	struct extent_state *cached_state = NULL;
2713 2714 2715 2716
	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
2717
	start += (offset + blocksize - 1) & ~(blocksize - 1);
2718 2719 2720
	if (start > end)
		return 0;

2721
	lock_extent_bits(tree, start, end, 0, &cached_state, GFP_NOFS);
2722
	wait_on_page_writeback(page);
2723
	clear_extent_bit(tree, start, end,
2724 2725
			 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
			 EXTENT_DO_ACCOUNTING,
2726
			 1, 1, &cached_state, GFP_NOFS);
2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776
	return 0;
}

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

	set_page_extent_mapped(page);
	set_page_dirty(page);

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

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

	set_page_extent_mapped(page);

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

	lock_extent(tree, page_start, page_end, GFP_NOFS);
C
Chris Mason 已提交
2777
	while (block_start <= block_end) {
2778 2779
		em = get_extent(inode, page, page_offset, block_start,
				block_end - block_start + 1, 1);
C
Chris Mason 已提交
2780
		if (IS_ERR(em) || !em)
2781
			goto err;
C
Chris Mason 已提交
2782

2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805
		cur_end = min(block_end, extent_map_end(em) - 1);
		block_off_start = block_start & (PAGE_CACHE_SIZE - 1);
		block_off_end = block_off_start + blocksize;
		isnew = clear_extent_new(tree, block_start, cur_end, GFP_NOFS);

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

			kaddr = kmap_atomic(page, KM_USER0);
			if (block_off_end > to)
				memset(kaddr + to, 0, block_off_end - to);
			if (block_off_start < from)
				memset(kaddr + block_off_start, 0,
				       from - block_off_start);
			flush_dcache_page(page);
			kunmap_atomic(kaddr, KM_USER0);
		}
		if ((em->block_start != EXTENT_MAP_HOLE &&
		     em->block_start != EXTENT_MAP_INLINE) &&
		    !isnew && !PageUptodate(page) &&
		    (block_off_end > to || block_off_start < from) &&
		    !test_range_bit(tree, block_start, cur_end,
2806
				    EXTENT_UPTODATE, 1, NULL)) {
2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819
			u64 sector;
			u64 extent_offset = block_start - em->start;
			size_t iosize;
			sector = (em->block_start + extent_offset) >> 9;
			iosize = (cur_end - block_start + blocksize) &
				~((u64)blocksize - 1);
			/*
			 * we've already got the extent locked, but we
			 * need to split the state such that our end_bio
			 * handler can clear the lock.
			 */
			set_extent_bit(tree, block_start,
				       block_start + iosize - 1,
2820
				       EXTENT_LOCKED, 0, NULL, NULL, GFP_NOFS);
2821 2822 2823
			ret = submit_extent_page(READ, tree, page,
					 sector, iosize, page_offset, em->bdev,
					 NULL, 1,
C
Chris Mason 已提交
2824 2825
					 end_bio_extent_preparewrite, 0,
					 0, 0);
2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846
			iocount++;
			block_start = block_start + iosize;
		} else {
			set_extent_uptodate(tree, block_start, cur_end,
					    GFP_NOFS);
			unlock_extent(tree, block_start, cur_end, GFP_NOFS);
			block_start = cur_end + 1;
		}
		page_offset = block_start & (PAGE_CACHE_SIZE - 1);
		free_extent_map(em);
	}
	if (iocount) {
		wait_extent_bit(tree, orig_block_start,
				block_end, EXTENT_LOCKED);
	}
	check_page_uptodate(tree, page);
err:
	/* FIXME, zero out newly allocated blocks on error */
	return err;
}

2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859
/*
 * 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;

2860
	if (test_range_bit(tree, start, end,
2861
			   EXTENT_IOBITS, 0, NULL))
2862 2863 2864 2865
		ret = 0;
	else {
		if ((mask & GFP_NOFS) == GFP_NOFS)
			mask = GFP_NOFS;
2866 2867 2868 2869 2870 2871 2872
		/*
		 * at this point we can safely clear everything except the
		 * locked bit and the nodatasum bit
		 */
		clear_extent_bit(tree, start, end,
				 ~(EXTENT_LOCKED | EXTENT_NODATASUM),
				 0, 0, NULL, mask);
2873 2874 2875 2876
	}
	return ret;
}

2877 2878 2879 2880 2881 2882
/*
 * 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,
2883 2884
			       struct extent_io_tree *tree, struct page *page,
			       gfp_t mask)
2885 2886 2887 2888
{
	struct extent_map *em;
	u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
	u64 end = start + PAGE_CACHE_SIZE - 1;
2889

2890 2891
	if ((mask & __GFP_WAIT) &&
	    page->mapping->host->i_size > 16 * 1024 * 1024) {
2892
		u64 len;
2893
		while (start <= end) {
2894
			len = end - start + 1;
2895
			write_lock(&map->lock);
2896
			em = lookup_extent_mapping(map, start, len);
2897
			if (!em || IS_ERR(em)) {
2898
				write_unlock(&map->lock);
2899 2900
				break;
			}
2901 2902
			if (test_bit(EXTENT_FLAG_PINNED, &em->flags) ||
			    em->start != start) {
2903
				write_unlock(&map->lock);
2904 2905 2906 2907 2908
				free_extent_map(em);
				break;
			}
			if (!test_range_bit(tree, em->start,
					    extent_map_end(em) - 1,
2909
					    EXTENT_LOCKED | EXTENT_WRITEBACK,
2910
					    0, NULL)) {
2911 2912 2913 2914 2915
				remove_extent_mapping(map, em);
				/* once for the rb tree */
				free_extent_map(em);
			}
			start = extent_map_end(em);
2916
			write_unlock(&map->lock);
2917 2918

			/* once for us */
2919 2920 2921
			free_extent_map(em);
		}
	}
2922
	return try_release_extent_state(map, tree, page, mask);
2923 2924 2925 2926 2927 2928
}

sector_t extent_bmap(struct address_space *mapping, sector_t iblock,
		get_extent_t *get_extent)
{
	struct inode *inode = mapping->host;
2929
	struct extent_state *cached_state = NULL;
2930 2931
	u64 start = iblock << inode->i_blkbits;
	sector_t sector = 0;
Y
Yan Zheng 已提交
2932
	size_t blksize = (1 << inode->i_blkbits);
2933 2934
	struct extent_map *em;

2935 2936
	lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + blksize - 1,
			 0, &cached_state, GFP_NOFS);
Y
Yan Zheng 已提交
2937
	em = get_extent(inode, NULL, 0, start, blksize, 0);
2938 2939
	unlock_extent_cached(&BTRFS_I(inode)->io_tree, start,
			     start + blksize - 1, &cached_state, GFP_NOFS);
2940 2941 2942
	if (!em || IS_ERR(em))
		return 0;

Y
Yan Zheng 已提交
2943
	if (em->block_start > EXTENT_MAP_LAST_BYTE)
2944 2945 2946 2947 2948 2949 2950 2951
		goto out;

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

Y
Yehuda Sadeh 已提交
2952 2953 2954 2955 2956 2957 2958 2959 2960
int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
		__u64 start, __u64 len, get_extent_t *get_extent)
{
	int ret;
	u64 off = start;
	u64 max = start + len;
	u32 flags = 0;
	u64 disko = 0;
	struct extent_map *em = NULL;
2961
	struct extent_state *cached_state = NULL;
Y
Yehuda Sadeh 已提交
2962 2963 2964 2965 2966 2967 2968 2969
	int end = 0;
	u64 em_start = 0, em_len = 0;
	unsigned long emflags;
	ret = 0;

	if (len == 0)
		return -EINVAL;

2970 2971
	lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + len, 0,
			 &cached_state, GFP_NOFS);
Y
Yehuda Sadeh 已提交
2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989
	em = get_extent(inode, NULL, 0, off, max - off, 0);
	if (!em)
		goto out;
	if (IS_ERR(em)) {
		ret = PTR_ERR(em);
		goto out;
	}
	while (!end) {
		off = em->start + em->len;
		if (off >= max)
			end = 1;

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

		disko = 0;
		flags = 0;

2990
		if (em->block_start == EXTENT_MAP_LAST_BYTE) {
Y
Yehuda Sadeh 已提交
2991 2992
			end = 1;
			flags |= FIEMAP_EXTENT_LAST;
2993
		} else if (em->block_start == EXTENT_MAP_HOLE) {
Y
Yehuda Sadeh 已提交
2994
			flags |= FIEMAP_EXTENT_UNWRITTEN;
2995
		} else if (em->block_start == EXTENT_MAP_INLINE) {
Y
Yehuda Sadeh 已提交
2996 2997
			flags |= (FIEMAP_EXTENT_DATA_INLINE |
				  FIEMAP_EXTENT_NOT_ALIGNED);
2998
		} else if (em->block_start == EXTENT_MAP_DELALLOC) {
Y
Yehuda Sadeh 已提交
2999 3000
			flags |= (FIEMAP_EXTENT_DELALLOC |
				  FIEMAP_EXTENT_UNKNOWN);
3001
		} else {
Y
Yehuda Sadeh 已提交
3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033
			disko = em->block_start;
		}
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
			flags |= FIEMAP_EXTENT_ENCODED;

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

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

		ret = fiemap_fill_next_extent(fieinfo, em_start, disko,
					em_len, flags);
		if (ret)
			goto out_free;
	}
out_free:
	free_extent_map(em);
out:
3034 3035
	unlock_extent_cached(&BTRFS_I(inode)->io_tree, start, start + len,
			     &cached_state, GFP_NOFS);
Y
Yehuda Sadeh 已提交
3036 3037 3038
	return ret;
}

3039 3040 3041 3042 3043 3044 3045 3046 3047 3048
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
3049 3050
	if (!mapping)
		return NULL;
3051 3052 3053 3054 3055 3056 3057

	/*
	 * 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();
3058
	p = radix_tree_lookup(&mapping->page_tree, i);
3059
	rcu_read_unlock();
3060

3061 3062 3063
	return p;
}

3064
static inline unsigned long num_extent_pages(u64 start, u64 len)
3065
{
3066 3067
	return ((start + len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT) -
		(start >> PAGE_CACHE_SHIFT);
3068 3069
}

3070 3071 3072 3073 3074 3075
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;
3076
#if LEAK_DEBUG
3077
	unsigned long flags;
C
Chris Mason 已提交
3078
#endif
3079 3080 3081 3082

	eb = kmem_cache_zalloc(extent_buffer_cache, mask);
	eb->start = start;
	eb->len = len;
3083 3084 3085
	spin_lock_init(&eb->lock);
	init_waitqueue_head(&eb->lock_wq);

3086
#if LEAK_DEBUG
3087 3088 3089
	spin_lock_irqsave(&leak_lock, flags);
	list_add(&eb->leak_list, &buffers);
	spin_unlock_irqrestore(&leak_lock, flags);
C
Chris Mason 已提交
3090
#endif
3091 3092 3093 3094 3095 3096 3097
	atomic_set(&eb->refs, 1);

	return eb;
}

static void __free_extent_buffer(struct extent_buffer *eb)
{
3098
#if LEAK_DEBUG
3099 3100 3101 3102
	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
3103
#endif
3104 3105 3106
	kmem_cache_free(extent_buffer_cache, eb);
}

3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139
/*
 * 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);
}

3140 3141 3142 3143 3144 3145 3146 3147 3148
struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree,
					  u64 start, unsigned long len,
					  struct page *page0,
					  gfp_t mask)
{
	unsigned long num_pages = num_extent_pages(start, len);
	unsigned long i;
	unsigned long index = start >> PAGE_CACHE_SHIFT;
	struct extent_buffer *eb;
3149
	struct extent_buffer *exists = NULL;
3150 3151 3152 3153
	struct page *p;
	struct address_space *mapping = tree->mapping;
	int uptodate = 1;

3154 3155 3156 3157 3158
	spin_lock(&tree->buffer_lock);
	eb = buffer_search(tree, start);
	if (eb) {
		atomic_inc(&eb->refs);
		spin_unlock(&tree->buffer_lock);
J
Josef Bacik 已提交
3159
		mark_page_accessed(eb->first_page);
3160 3161 3162 3163
		return eb;
	}
	spin_unlock(&tree->buffer_lock);

3164
	eb = __alloc_extent_buffer(tree, start, len, mask);
3165
	if (!eb)
3166 3167 3168 3169 3170 3171 3172 3173 3174 3175
		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);
3176
		uptodate = PageUptodate(page0);
3177 3178 3179 3180 3181 3182 3183
	} else {
		i = 0;
	}
	for (; i < num_pages; i++, index++) {
		p = find_or_create_page(mapping, index, mask | __GFP_HIGHMEM);
		if (!p) {
			WARN_ON(1);
3184
			goto free_eb;
3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198
		}
		set_page_extent_mapped(p);
		mark_page_accessed(p);
		if (i == 0) {
			eb->first_page = p;
			set_page_extent_head(p, len);
		} else {
			set_page_private(p, EXTENT_PAGE_PRIVATE);
		}
		if (!PageUptodate(p))
			uptodate = 0;
		unlock_page(p);
	}
	if (uptodate)
3199
		set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
3200

3201 3202 3203 3204 3205 3206 3207 3208 3209 3210
	spin_lock(&tree->buffer_lock);
	exists = buffer_tree_insert(tree, start, &eb->rb_node);
	if (exists) {
		/* add one reference for the caller */
		atomic_inc(&exists->refs);
		spin_unlock(&tree->buffer_lock);
		goto free_eb;
	}
	/* add one reference for the tree */
	atomic_inc(&eb->refs);
3211
	spin_unlock(&tree->buffer_lock);
3212 3213
	return eb;

3214
free_eb:
3215
	if (!atomic_dec_and_test(&eb->refs))
3216
		return exists;
3217
	btrfs_release_extent_buffer(eb);
3218
	return exists;
3219 3220 3221 3222 3223 3224 3225 3226
}

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

3227 3228 3229 3230 3231
	spin_lock(&tree->buffer_lock);
	eb = buffer_search(tree, start);
	if (eb)
		atomic_inc(&eb->refs);
	spin_unlock(&tree->buffer_lock);
3232

J
Josef Bacik 已提交
3233 3234 3235
	if (eb)
		mark_page_accessed(eb->first_page);

3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246
	return eb;
}

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

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

3247
	WARN_ON(1);
3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260
}

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);
3261
		if (!PageDirty(page))
C
Chris Mason 已提交
3262 3263
			continue;

3264
		lock_page(page);
3265 3266 3267 3268 3269 3270
		if (i == 0)
			set_page_extent_head(page, eb->len);
		else
			set_page_private(page, EXTENT_PAGE_PRIVATE);

		clear_page_dirty_for_io(page);
3271
		spin_lock_irq(&page->mapping->tree_lock);
3272 3273 3274 3275 3276
		if (!PageDirty(page)) {
			radix_tree_tag_clear(&page->mapping->page_tree,
						page_index(page),
						PAGECACHE_TAG_DIRTY);
		}
3277
		spin_unlock_irq(&page->mapping->tree_lock);
3278
		unlock_page(page);
3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294
	}
	return 0;
}

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

int set_extent_buffer_dirty(struct extent_io_tree *tree,
			     struct extent_buffer *eb)
{
	unsigned long i;
	unsigned long num_pages;
3295
	int was_dirty = 0;
3296

3297
	was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
3298
	num_pages = num_extent_pages(eb->start, eb->len);
3299
	for (i = 0; i < num_pages; i++)
3300
		__set_page_dirty_nobuffers(extent_buffer_page(eb, i));
3301
	return was_dirty;
3302 3303
}

3304
int clear_extent_buffer_uptodate(struct extent_io_tree *tree,
3305 3306
				struct extent_buffer *eb,
				struct extent_state **cached_state)
3307 3308 3309 3310 3311 3312
{
	unsigned long i;
	struct page *page;
	unsigned long num_pages;

	num_pages = num_extent_pages(eb->start, eb->len);
3313
	clear_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
3314 3315

	clear_extent_uptodate(tree, eb->start, eb->start + eb->len - 1,
3316
			      cached_state, GFP_NOFS);
3317 3318
	for (i = 0; i < num_pages; i++) {
		page = extent_buffer_page(eb, i);
C
Chris Mason 已提交
3319 3320
		if (page)
			ClearPageUptodate(page);
3321 3322 3323 3324
	}
	return 0;
}

3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348
int set_extent_buffer_uptodate(struct extent_io_tree *tree,
				struct extent_buffer *eb)
{
	unsigned long i;
	struct page *page;
	unsigned long num_pages;

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

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

3349 3350 3351 3352 3353 3354 3355 3356 3357
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;

3358
	ret = test_range_bit(tree, start, end, EXTENT_UPTODATE, 1, NULL);
3359 3360
	if (ret)
		return 1;
C
Chris Mason 已提交
3361
	while (start <= end) {
3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374
		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;
}

3375
int extent_buffer_uptodate(struct extent_io_tree *tree,
3376 3377
			   struct extent_buffer *eb,
			   struct extent_state *cached_state)
3378
{
3379
	int ret = 0;
3380 3381
	unsigned long num_pages;
	unsigned long i;
3382 3383 3384
	struct page *page;
	int pg_uptodate = 1;

3385
	if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags))
C
Chris Mason 已提交
3386
		return 1;
3387

C
Chris Mason 已提交
3388
	ret = test_range_bit(tree, eb->start, eb->start + eb->len - 1,
3389
			   EXTENT_UPTODATE, 1, cached_state);
C
Chris Mason 已提交
3390 3391
	if (ret)
		return ret;
3392 3393 3394 3395 3396 3397 3398 3399 3400

	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
3401
	return pg_uptodate;
3402 3403 3404 3405
}

int read_extent_buffer_pages(struct extent_io_tree *tree,
			     struct extent_buffer *eb,
3406
			     u64 start, int wait,
3407
			     get_extent_t *get_extent, int mirror_num)
3408 3409 3410 3411 3412 3413
{
	unsigned long i;
	unsigned long start_i;
	struct page *page;
	int err;
	int ret = 0;
3414 3415 3416
	int locked_pages = 0;
	int all_uptodate = 1;
	int inc_all_pages = 0;
3417
	unsigned long num_pages;
3418
	struct bio *bio = NULL;
C
Chris Mason 已提交
3419
	unsigned long bio_flags = 0;
3420

3421
	if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags))
3422 3423
		return 0;

3424
	if (test_range_bit(tree, eb->start, eb->start + eb->len - 1,
3425
			   EXTENT_UPTODATE, 1, NULL)) {
3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440
		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) {
3441
			if (!trylock_page(page))
3442
				goto unlock_exit;
3443 3444 3445
		} else {
			lock_page(page);
		}
3446
		locked_pages++;
C
Chris Mason 已提交
3447
		if (!PageUptodate(page))
3448 3449 3450 3451
			all_uptodate = 0;
	}
	if (all_uptodate) {
		if (start_i == 0)
3452
			set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
3453 3454 3455 3456 3457 3458 3459 3460 3461 3462
		goto unlock_exit;
	}

	for (i = start_i; i < num_pages; i++) {
		page = extent_buffer_page(eb, i);
		if (inc_all_pages)
			page_cache_get(page);
		if (!PageUptodate(page)) {
			if (start_i == 0)
				inc_all_pages = 1;
3463
			ClearPageError(page);
3464
			err = __extent_read_full_page(tree, page,
3465
						      get_extent, &bio,
C
Chris Mason 已提交
3466
						      mirror_num, &bio_flags);
C
Chris Mason 已提交
3467
			if (err)
3468 3469 3470 3471 3472 3473
				ret = err;
		} else {
			unlock_page(page);
		}
	}

3474
	if (bio)
C
Chris Mason 已提交
3475
		submit_one_bio(READ, bio, mirror_num, bio_flags);
3476

C
Chris Mason 已提交
3477
	if (ret || !wait)
3478
		return ret;
C
Chris Mason 已提交
3479

3480 3481 3482
	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
3483
		if (!PageUptodate(page))
3484 3485
			ret = -EIO;
	}
C
Chris Mason 已提交
3486

3487
	if (!ret)
3488
		set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
3489
	return ret;
3490 3491 3492

unlock_exit:
	i = start_i;
C
Chris Mason 已提交
3493
	while (locked_pages > 0) {
3494 3495 3496 3497 3498 3499
		page = extent_buffer_page(eb, i);
		i++;
		unlock_page(page);
		locked_pages--;
	}
	return ret;
3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518
}

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
3519
	while (len > 0) {
3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556
		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
3557

3558
	if (start + min_len > eb->len) {
C
Chris Mason 已提交
3559 3560 3561
		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);
3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 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
		WARN_ON(1);
	}

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

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

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

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

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

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

C
Chris Mason 已提交
3620
	while (len > 0) {
3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654
		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
3655
	while (len > 0) {
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_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
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
		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
3718
	while (len > 0) {
3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783
		page = extent_buffer_page(dst, i);
		WARN_ON(!PageUptodate(page));

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

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

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

static void move_pages(struct page *dst_page, struct page *src_page,
		       unsigned long dst_off, unsigned long src_off,
		       unsigned long len)
{
	char *dst_kaddr = kmap_atomic(dst_page, KM_USER0);
	if (dst_page == src_page) {
		memmove(dst_kaddr + dst_off, dst_kaddr + src_off, len);
	} else {
		char *src_kaddr = kmap_atomic(src_page, KM_USER1);
		char *p = dst_kaddr + dst_off + len;
		char *s = src_kaddr + src_off + len;

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

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

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

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

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

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

	if (src_offset + len > dst->len) {
C
Chris Mason 已提交
3784 3785
		printk(KERN_ERR "btrfs memmove bogus src_offset %lu move "
		       "len %lu dst len %lu\n", src_offset, len, dst->len);
3786 3787 3788
		BUG_ON(1);
	}
	if (dst_offset + len > dst->len) {
C
Chris Mason 已提交
3789 3790
		printk(KERN_ERR "btrfs memmove bogus dst_offset %lu move "
		       "len %lu dst len %lu\n", dst_offset, len, dst->len);
3791 3792 3793
		BUG_ON(1);
	}

C
Chris Mason 已提交
3794
	while (len > 0) {
3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826 3827 3828 3829 3830
		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
3831 3832
		printk(KERN_ERR "btrfs memmove bogus src_offset %lu move "
		       "len %lu len %lu\n", src_offset, len, dst->len);
3833 3834 3835
		BUG_ON(1);
	}
	if (dst_offset + len > dst->len) {
C
Chris Mason 已提交
3836 3837
		printk(KERN_ERR "btrfs memmove bogus dst_offset %lu move "
		       "len %lu len %lu\n", dst_offset, len, dst->len);
3838 3839 3840 3841 3842 3843
		BUG_ON(1);
	}
	if (dst_offset < src_offset) {
		memcpy_extent_buffer(dst, dst_offset, src_offset, len);
		return;
	}
C
Chris Mason 已提交
3844
	while (len > 0) {
3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862 3863 3864
		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;
	}
}
3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880

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);
	eb = buffer_search(tree, start);
	if (!eb)
		goto out;

	if (atomic_read(&eb->refs) > 1) {
		ret = 0;
		goto out;
	}
3881 3882 3883 3884
	if (test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
		ret = 0;
		goto out;
	}
3885

3886
	rb_erase(&eb->rb_node, &tree->buffer);
3887 3888
	/* at this point we can safely release the extent buffer */
	btrfs_release_extent_buffer(eb);
3889 3890 3891 3892
out:
	spin_unlock(&tree->buffer_lock);
	return ret;
}