extent_io.c 135.0 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11
#include <linux/bitops.h>
#include <linux/slab.h>
#include <linux/bio.h>
#include <linux/mm.h>
#include <linux/pagemap.h>
#include <linux/page-flags.h>
#include <linux/spinlock.h>
#include <linux/blkdev.h>
#include <linux/swap.h>
#include <linux/writeback.h>
#include <linux/pagevec.h>
12
#include <linux/prefetch.h>
D
Dan Magenheimer 已提交
13
#include <linux/cleancache.h>
14 15
#include "extent_io.h"
#include "extent_map.h"
16 17
#include "ctree.h"
#include "btrfs_inode.h"
18
#include "volumes.h"
19
#include "check-integrity.h"
20
#include "locking.h"
21
#include "rcu-string.h"
22
#include "backref.h"
23 24 25

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

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

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

static inline
void btrfs_leak_debug_add(struct list_head *new, struct list_head *head)
{
	unsigned long flags;

	spin_lock_irqsave(&leak_lock, flags);
	list_add(new, head);
	spin_unlock_irqrestore(&leak_lock, flags);
}

static inline
void btrfs_leak_debug_del(struct list_head *entry)
{
	unsigned long flags;

	spin_lock_irqsave(&leak_lock, flags);
	list_del(entry);
	spin_unlock_irqrestore(&leak_lock, flags);
}

static inline
void btrfs_leak_debug_check(void)
{
	struct extent_state *state;
	struct extent_buffer *eb;

	while (!list_empty(&states)) {
		state = list_entry(states.next, struct extent_state, leak_list);
62
		printk(KERN_ERR "BTRFS: state leak: start %llu end %llu "
63
		       "state %lu in tree %p refs %d\n",
64 65
		       state->start, state->end, state->state, state->tree,
		       atomic_read(&state->refs));
66 67 68 69 70 71
		list_del(&state->leak_list);
		kmem_cache_free(extent_state_cache, state);
	}

	while (!list_empty(&buffers)) {
		eb = list_entry(buffers.next, struct extent_buffer, leak_list);
72
		printk(KERN_ERR "BTRFS: buffer leak start %llu len %lu "
73 74
		       "refs %d\n",
		       eb->start, eb->len, atomic_read(&eb->refs));
75 76 77 78
		list_del(&eb->leak_list);
		kmem_cache_free(extent_buffer_cache, eb);
	}
}
79

80 81
#define btrfs_debug_check_extent_io_range(tree, start, end)		\
	__btrfs_debug_check_extent_io_range(__func__, (tree), (start), (end))
82
static inline void __btrfs_debug_check_extent_io_range(const char *caller,
83
		struct extent_io_tree *tree, u64 start, u64 end)
84
{
85 86
	struct inode *inode;
	u64 isize;
87

88 89
	if (!tree->mapping)
		return;
90

91 92
	inode = tree->mapping->host;
	isize = i_size_read(inode);
93 94
	if (end >= PAGE_SIZE && (end % 2) == 0 && end != isize - 1) {
		printk_ratelimited(KERN_DEBUG
95
		    "BTRFS: %s: ino %llu isize %llu odd range [%llu,%llu]\n",
96
				caller, btrfs_ino(inode), isize, start, end);
97 98
	}
}
99 100 101 102
#else
#define btrfs_leak_debug_add(new, head)	do {} while (0)
#define btrfs_leak_debug_del(entry)	do {} while (0)
#define btrfs_leak_debug_check()	do {} while (0)
103
#define btrfs_debug_check_extent_io_range(c, s, e)	do {} while (0)
C
Chris Mason 已提交
104
#endif
105 106 107 108 109 110 111 112 113 114 115 116 117

#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;
118
	unsigned long bio_flags;
119 120 121 122

	/* tells writepage not to lock the state bits for this range
	 * it still does the unlocking
	 */
123 124 125 126
	unsigned int extent_locked:1;

	/* tells the submit_bio code to use a WRITE_SYNC */
	unsigned int sync_io:1;
127 128
};

129
static noinline void flush_write_bio(void *data);
130 131 132
static inline struct btrfs_fs_info *
tree_fs_info(struct extent_io_tree *tree)
{
133 134
	if (!tree->mapping)
		return NULL;
135 136
	return btrfs_sb(tree->mapping->host->i_sb);
}
137

138 139
int __init extent_io_init(void)
{
D
David Sterba 已提交
140
	extent_state_cache = kmem_cache_create("btrfs_extent_state",
141 142
			sizeof(struct extent_state), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
143 144 145
	if (!extent_state_cache)
		return -ENOMEM;

D
David Sterba 已提交
146
	extent_buffer_cache = kmem_cache_create("btrfs_extent_buffer",
147 148
			sizeof(struct extent_buffer), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
149 150
	if (!extent_buffer_cache)
		goto free_state_cache;
151 152 153 154 155

	btrfs_bioset = bioset_create(BIO_POOL_SIZE,
				     offsetof(struct btrfs_io_bio, bio));
	if (!btrfs_bioset)
		goto free_buffer_cache;
156 157 158 159

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

160 161
	return 0;

162 163 164 165
free_bioset:
	bioset_free(btrfs_bioset);
	btrfs_bioset = NULL;

166 167 168 169
free_buffer_cache:
	kmem_cache_destroy(extent_buffer_cache);
	extent_buffer_cache = NULL;

170 171
free_state_cache:
	kmem_cache_destroy(extent_state_cache);
172
	extent_state_cache = NULL;
173 174 175 176 177
	return -ENOMEM;
}

void extent_io_exit(void)
{
178
	btrfs_leak_debug_check();
179 180 181 182 183 184

	/*
	 * Make sure all delayed rcu free are flushed before we
	 * destroy caches.
	 */
	rcu_barrier();
185 186 187 188
	if (extent_state_cache)
		kmem_cache_destroy(extent_state_cache);
	if (extent_buffer_cache)
		kmem_cache_destroy(extent_buffer_cache);
189 190
	if (btrfs_bioset)
		bioset_free(btrfs_bioset);
191 192 193
}

void extent_io_tree_init(struct extent_io_tree *tree,
194
			 struct address_space *mapping)
195
{
196
	tree->state = RB_ROOT;
197 198
	tree->ops = NULL;
	tree->dirty_bytes = 0;
199
	spin_lock_init(&tree->lock);
200 201 202
	tree->mapping = mapping;
}

203
static struct extent_state *alloc_extent_state(gfp_t mask)
204 205 206 207
{
	struct extent_state *state;

	state = kmem_cache_alloc(extent_state_cache, mask);
208
	if (!state)
209 210 211
		return state;
	state->state = 0;
	state->private = 0;
212
	state->tree = NULL;
213
	btrfs_leak_debug_add(&state->leak_list, &states);
214 215
	atomic_set(&state->refs, 1);
	init_waitqueue_head(&state->wq);
216
	trace_alloc_extent_state(state, mask, _RET_IP_);
217 218 219
	return state;
}

220
void free_extent_state(struct extent_state *state)
221 222 223 224
{
	if (!state)
		return;
	if (atomic_dec_and_test(&state->refs)) {
225
		WARN_ON(state->tree);
226
		btrfs_leak_debug_del(&state->leak_list);
227
		trace_free_extent_state(state, _RET_IP_);
228 229 230 231
		kmem_cache_free(extent_state_cache, state);
	}
}

232 233 234
static struct rb_node *tree_insert(struct rb_root *root,
				   struct rb_node *search_start,
				   u64 offset,
235 236 237
				   struct rb_node *node,
				   struct rb_node ***p_in,
				   struct rb_node **parent_in)
238
{
239
	struct rb_node **p;
C
Chris Mason 已提交
240
	struct rb_node *parent = NULL;
241 242
	struct tree_entry *entry;

243 244 245 246 247 248
	if (p_in && parent_in) {
		p = *p_in;
		parent = *parent_in;
		goto do_insert;
	}

249
	p = search_start ? &search_start : &root->rb_node;
C
Chris Mason 已提交
250
	while (*p) {
251 252 253 254 255 256 257 258 259 260 261
		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;
	}

262
do_insert:
263 264 265 266 267
	rb_link_node(node, parent, p);
	rb_insert_color(node, root);
	return NULL;
}

268
static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset,
269 270 271 272
				      struct rb_node **prev_ret,
				      struct rb_node **next_ret,
				      struct rb_node ***p_ret,
				      struct rb_node **parent_ret)
273
{
274
	struct rb_root *root = &tree->state;
275
	struct rb_node **n = &root->rb_node;
276 277 278 279 280
	struct rb_node *prev = NULL;
	struct rb_node *orig_prev = NULL;
	struct tree_entry *entry;
	struct tree_entry *prev_entry = NULL;

281 282 283
	while (*n) {
		prev = *n;
		entry = rb_entry(prev, struct tree_entry, rb_node);
284 285 286
		prev_entry = entry;

		if (offset < entry->start)
287
			n = &(*n)->rb_left;
288
		else if (offset > entry->end)
289
			n = &(*n)->rb_right;
C
Chris Mason 已提交
290
		else
291
			return *n;
292 293
	}

294 295 296 297 298
	if (p_ret)
		*p_ret = n;
	if (parent_ret)
		*parent_ret = prev;

299 300
	if (prev_ret) {
		orig_prev = prev;
C
Chris Mason 已提交
301
		while (prev && offset > prev_entry->end) {
302 303 304 305 306 307 308 309 310
			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
311
		while (prev && offset < prev_entry->start) {
312 313 314 315 316 317 318 319
			prev = rb_prev(prev);
			prev_entry = rb_entry(prev, struct tree_entry, rb_node);
		}
		*next_ret = prev;
	}
	return NULL;
}

320 321 322 323 324
static inline struct rb_node *
tree_search_for_insert(struct extent_io_tree *tree,
		       u64 offset,
		       struct rb_node ***p_ret,
		       struct rb_node **parent_ret)
325
{
326
	struct rb_node *prev = NULL;
327
	struct rb_node *ret;
328

329
	ret = __etree_search(tree, offset, &prev, NULL, p_ret, parent_ret);
C
Chris Mason 已提交
330
	if (!ret)
331 332 333 334
		return prev;
	return ret;
}

335 336 337 338 339 340
static inline struct rb_node *tree_search(struct extent_io_tree *tree,
					  u64 offset)
{
	return tree_search_for_insert(tree, offset, NULL, NULL);
}

J
Josef Bacik 已提交
341 342 343 344 345 346 347 348
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);
}

349 350 351 352 353 354 355 356 357
/*
 * 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.
 */
358 359
static void merge_state(struct extent_io_tree *tree,
		        struct extent_state *state)
360 361 362 363
{
	struct extent_state *other;
	struct rb_node *other_node;

364
	if (state->state & (EXTENT_IOBITS | EXTENT_BOUNDARY))
365
		return;
366 367 368 369 370 371

	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
372
			merge_cb(tree, state, other);
373
			state->start = other->start;
374
			other->tree = NULL;
375 376 377 378 379 380 381 382 383
			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
384
			merge_cb(tree, state, other);
385 386 387 388
			state->end = other->end;
			other->tree = NULL;
			rb_erase(&other->rb_node, &tree->state);
			free_extent_state(other);
389 390 391 392
		}
	}
}

393
static void set_state_cb(struct extent_io_tree *tree,
394
			 struct extent_state *state, unsigned long *bits)
395
{
396 397
	if (tree->ops && tree->ops->set_bit_hook)
		tree->ops->set_bit_hook(tree->mapping->host, state, bits);
398 399 400
}

static void clear_state_cb(struct extent_io_tree *tree,
401
			   struct extent_state *state, unsigned long *bits)
402
{
J
Josef Bacik 已提交
403 404
	if (tree->ops && tree->ops->clear_bit_hook)
		tree->ops->clear_bit_hook(tree->mapping->host, state, bits);
405 406
}

407
static void set_state_bits(struct extent_io_tree *tree,
408
			   struct extent_state *state, unsigned long *bits);
409

410 411 412 413 414 415 416 417 418 419 420 421
/*
 * 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,
422 423
			struct rb_node ***p,
			struct rb_node **parent,
424
			unsigned long *bits)
425 426 427
{
	struct rb_node *node;

J
Julia Lawall 已提交
428
	if (end < start)
429
		WARN(1, KERN_ERR "BTRFS: end < start %llu %llu\n",
430
		       end, start);
431 432
	state->start = start;
	state->end = end;
J
Josef Bacik 已提交
433

434 435
	set_state_bits(tree, state, bits);

436
	node = tree_insert(&tree->state, NULL, end, &state->rb_node, p, parent);
437 438 439
	if (node) {
		struct extent_state *found;
		found = rb_entry(node, struct extent_state, rb_node);
440
		printk(KERN_ERR "BTRFS: found node %llu %llu on insert of "
441 442
		       "%llu %llu\n",
		       found->start, found->end, start, end);
443 444
		return -EEXIST;
	}
445
	state->tree = tree;
446 447 448 449
	merge_state(tree, state);
	return 0;
}

450
static void split_cb(struct extent_io_tree *tree, struct extent_state *orig,
J
Josef Bacik 已提交
451 452 453
		     u64 split)
{
	if (tree->ops && tree->ops->split_extent_hook)
454
		tree->ops->split_extent_hook(tree->mapping->host, orig, split);
J
Josef Bacik 已提交
455 456
}

457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474
/*
 * 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
475 476 477

	split_cb(tree, orig, split);

478 479 480 481 482
	prealloc->start = orig->start;
	prealloc->end = split - 1;
	prealloc->state = orig->state;
	orig->start = split;

483 484
	node = tree_insert(&tree->state, &orig->rb_node, prealloc->end,
			   &prealloc->rb_node, NULL, NULL);
485 486 487 488
	if (node) {
		free_extent_state(prealloc);
		return -EEXIST;
	}
489
	prealloc->tree = tree;
490 491 492
	return 0;
}

493 494 495 496 497 498 499 500 501
static struct extent_state *next_state(struct extent_state *state)
{
	struct rb_node *next = rb_next(&state->rb_node);
	if (next)
		return rb_entry(next, struct extent_state, rb_node);
	else
		return NULL;
}

502 503
/*
 * utility function to clear some bits in an extent state struct.
504
 * it will optionally wake up any one waiting on this state (wake == 1).
505 506 507 508
 *
 * If no bits are set on the state struct after clearing things, the
 * struct is freed and removed from the tree
 */
509 510
static struct extent_state *clear_state_bit(struct extent_io_tree *tree,
					    struct extent_state *state,
511
					    unsigned long *bits, int wake)
512
{
513
	struct extent_state *next;
514
	unsigned long bits_to_clear = *bits & ~EXTENT_CTLBITS;
515

516
	if ((bits_to_clear & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) {
517 518 519 520
		u64 range = state->end - state->start + 1;
		WARN_ON(range > tree->dirty_bytes);
		tree->dirty_bytes -= range;
	}
521
	clear_state_cb(tree, state, bits);
522
	state->state &= ~bits_to_clear;
523 524
	if (wake)
		wake_up(&state->wq);
525
	if (state->state == 0) {
526
		next = next_state(state);
527
		if (state->tree) {
528
			rb_erase(&state->rb_node, &tree->state);
529
			state->tree = NULL;
530 531 532 533 534 535
			free_extent_state(state);
		} else {
			WARN_ON(1);
		}
	} else {
		merge_state(tree, state);
536
		next = next_state(state);
537
	}
538
	return next;
539 540
}

541 542 543 544 545 546 547 548 549
static struct extent_state *
alloc_extent_state_atomic(struct extent_state *prealloc)
{
	if (!prealloc)
		prealloc = alloc_extent_state(GFP_ATOMIC);

	return prealloc;
}

550
static void extent_io_tree_panic(struct extent_io_tree *tree, int err)
551 552 553 554 555 556
{
	btrfs_panic(tree_fs_info(tree), err, "Locking error: "
		    "Extent tree was modified by another "
		    "thread while locked.");
}

557 558 559 560 561 562 563 564 565 566
/*
 * 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.
 *
567
 * This takes the tree lock, and returns 0 on success and < 0 on error.
568 569
 */
int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
570
		     unsigned long bits, int wake, int delete,
571 572
		     struct extent_state **cached_state,
		     gfp_t mask)
573 574
{
	struct extent_state *state;
575
	struct extent_state *cached;
576 577
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
578
	u64 last_end;
579
	int err;
580
	int clear = 0;
581

582
	btrfs_debug_check_extent_io_range(tree, start, end);
583

584 585 586
	if (bits & EXTENT_DELALLOC)
		bits |= EXTENT_NORESERVE;

587 588 589 590
	if (delete)
		bits |= ~EXTENT_CTLBITS;
	bits |= EXTENT_FIRST_DELALLOC;

591 592
	if (bits & (EXTENT_IOBITS | EXTENT_BOUNDARY))
		clear = 1;
593 594 595 596 597 598 599
again:
	if (!prealloc && (mask & __GFP_WAIT)) {
		prealloc = alloc_extent_state(mask);
		if (!prealloc)
			return -ENOMEM;
	}

600
	spin_lock(&tree->lock);
601 602
	if (cached_state) {
		cached = *cached_state;
603 604 605 606 607 608

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

609 610
		if (cached && cached->tree && cached->start <= start &&
		    cached->end > start) {
611 612
			if (clear)
				atomic_dec(&cached->refs);
613
			state = cached;
614
			goto hit_next;
615
		}
616 617
		if (clear)
			free_extent_state(cached);
618
	}
619 620 621 622
	/*
	 * this search will find the extents that end after
	 * our range starts
	 */
623
	node = tree_search(tree, start);
624 625 626
	if (!node)
		goto out;
	state = rb_entry(node, struct extent_state, rb_node);
627
hit_next:
628 629 630
	if (state->start > end)
		goto out;
	WARN_ON(state->end < start);
631
	last_end = state->end;
632

633
	/* the state doesn't have the wanted bits, go ahead */
634 635
	if (!(state->state & bits)) {
		state = next_state(state);
636
		goto next;
637
	}
638

639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655
	/*
	 *     | ---- 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) {
656 657
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
658
		err = split_state(tree, state, prealloc, start);
659 660 661
		if (err)
			extent_io_tree_panic(tree, err);

662 663 664 665
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
666 667
			state = clear_state_bit(tree, state, &bits, wake);
			goto next;
668 669 670 671 672 673 674 675 676 677
		}
		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) {
678 679
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
680
		err = split_state(tree, state, prealloc, end + 1);
681 682 683
		if (err)
			extent_io_tree_panic(tree, err);

684 685
		if (wake)
			wake_up(&state->wq);
686

687
		clear_state_bit(tree, prealloc, &bits, wake);
J
Josef Bacik 已提交
688

689 690 691
		prealloc = NULL;
		goto out;
	}
692

693
	state = clear_state_bit(tree, state, &bits, wake);
694
next:
695 696 697
	if (last_end == (u64)-1)
		goto out;
	start = last_end + 1;
698
	if (start <= end && state && !need_resched())
699
		goto hit_next;
700 701 702
	goto search_again;

out:
703
	spin_unlock(&tree->lock);
704 705 706
	if (prealloc)
		free_extent_state(prealloc);

707
	return 0;
708 709 710 711

search_again:
	if (start > end)
		goto out;
712
	spin_unlock(&tree->lock);
713 714 715 716 717
	if (mask & __GFP_WAIT)
		cond_resched();
	goto again;
}

718 719
static void wait_on_state(struct extent_io_tree *tree,
			  struct extent_state *state)
720 721
		__releases(tree->lock)
		__acquires(tree->lock)
722 723 724
{
	DEFINE_WAIT(wait);
	prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE);
725
	spin_unlock(&tree->lock);
726
	schedule();
727
	spin_lock(&tree->lock);
728 729 730 731 732 733 734 735
	finish_wait(&state->wq, &wait);
}

/*
 * waits for one or more bits to clear on a range in the state tree.
 * The range [start, end] is inclusive.
 * The tree lock is taken by this function
 */
736 737
static void wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
			    unsigned long bits)
738 739 740 741
{
	struct extent_state *state;
	struct rb_node *node;

742
	btrfs_debug_check_extent_io_range(tree, start, end);
743

744
	spin_lock(&tree->lock);
745 746 747 748 749 750
again:
	while (1) {
		/*
		 * this search will find all the extents that end after
		 * our range starts
		 */
751
		node = tree_search(tree, start);
752
process_node:
753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772
		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;

773 774 775 776
		if (!cond_resched_lock(&tree->lock)) {
			node = rb_next(node);
			goto process_node;
		}
777 778
	}
out:
779
	spin_unlock(&tree->lock);
780 781
}

782
static void set_state_bits(struct extent_io_tree *tree,
783
			   struct extent_state *state,
784
			   unsigned long *bits)
785
{
786
	unsigned long bits_to_set = *bits & ~EXTENT_CTLBITS;
J
Josef Bacik 已提交
787

788
	set_state_cb(tree, state, bits);
789
	if ((bits_to_set & EXTENT_DIRTY) && !(state->state & EXTENT_DIRTY)) {
790 791 792
		u64 range = state->end - state->start + 1;
		tree->dirty_bytes += range;
	}
793
	state->state |= bits_to_set;
794 795
}

796 797 798 799 800 801 802 803 804 805 806
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);
		}
	}
}

807
/*
808 809
 * 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.
810
 *
811 812 813
 * 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.
814
 *
815
 * [start, end] is inclusive This takes the tree lock.
816
 */
817

J
Jeff Mahoney 已提交
818 819
static int __must_check
__set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
820 821 822
		 unsigned long bits, unsigned long exclusive_bits,
		 u64 *failed_start, struct extent_state **cached_state,
		 gfp_t mask)
823 824 825 826
{
	struct extent_state *state;
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
827 828
	struct rb_node **p;
	struct rb_node *parent;
829 830 831
	int err = 0;
	u64 last_start;
	u64 last_end;
832

833
	btrfs_debug_check_extent_io_range(tree, start, end);
834

835
	bits |= EXTENT_FIRST_DELALLOC;
836 837 838
again:
	if (!prealloc && (mask & __GFP_WAIT)) {
		prealloc = alloc_extent_state(mask);
839
		BUG_ON(!prealloc);
840 841
	}

842
	spin_lock(&tree->lock);
843 844
	if (cached_state && *cached_state) {
		state = *cached_state;
845 846
		if (state->start <= start && state->end > start &&
		    state->tree) {
847 848 849 850
			node = &state->rb_node;
			goto hit_next;
		}
	}
851 852 853 854
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
855
	node = tree_search_for_insert(tree, start, &p, &parent);
856
	if (!node) {
857 858
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
859 860
		err = insert_state(tree, prealloc, start, end,
				   &p, &parent, &bits);
861 862 863
		if (err)
			extent_io_tree_panic(tree, err);

864
		cache_state(prealloc, cached_state);
865 866 867 868
		prealloc = NULL;
		goto out;
	}
	state = rb_entry(node, struct extent_state, rb_node);
C
Chris Mason 已提交
869
hit_next:
870 871 872 873 874 875 876 877 878 879
	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) {
880
		if (state->state & exclusive_bits) {
881 882 883 884
			*failed_start = state->start;
			err = -EEXIST;
			goto out;
		}
885

886
		set_state_bits(tree, state, &bits);
887
		cache_state(state, cached_state);
888
		merge_state(tree, state);
889 890 891
		if (last_end == (u64)-1)
			goto out;
		start = last_end + 1;
892 893 894 895
		state = next_state(state);
		if (start < end && state && state->start == start &&
		    !need_resched())
			goto hit_next;
896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915
		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) {
916
		if (state->state & exclusive_bits) {
917 918 919 920
			*failed_start = start;
			err = -EEXIST;
			goto out;
		}
921 922 923

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
924
		err = split_state(tree, state, prealloc, start);
925 926 927
		if (err)
			extent_io_tree_panic(tree, err);

928 929 930 931
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
932
			set_state_bits(tree, state, &bits);
933
			cache_state(state, cached_state);
934
			merge_state(tree, state);
935 936 937
			if (last_end == (u64)-1)
				goto out;
			start = last_end + 1;
938 939 940 941
			state = next_state(state);
			if (start < end && state && state->start == start &&
			    !need_resched())
				goto hit_next;
942 943 944 945 946 947 948 949 950 951 952 953 954 955 956
		}
		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
957
			this_end = last_start - 1;
958 959 960

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
961 962 963 964 965

		/*
		 * Avoid to free 'prealloc' if it can be merged with
		 * the later extent.
		 */
966
		err = insert_state(tree, prealloc, start, this_end,
967
				   NULL, NULL, &bits);
968 969 970
		if (err)
			extent_io_tree_panic(tree, err);

J
Josef Bacik 已提交
971 972
		cache_state(prealloc, cached_state);
		prealloc = NULL;
973 974 975 976 977 978 979 980 981 982
		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) {
983
		if (state->state & exclusive_bits) {
984 985 986 987
			*failed_start = start;
			err = -EEXIST;
			goto out;
		}
988 989 990

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
991
		err = split_state(tree, state, prealloc, end + 1);
992 993
		if (err)
			extent_io_tree_panic(tree, err);
994

995
		set_state_bits(tree, prealloc, &bits);
996
		cache_state(prealloc, cached_state);
997 998 999 1000 1001 1002 1003 1004
		merge_state(tree, prealloc);
		prealloc = NULL;
		goto out;
	}

	goto search_again;

out:
1005
	spin_unlock(&tree->lock);
1006 1007 1008 1009 1010 1011 1012 1013
	if (prealloc)
		free_extent_state(prealloc);

	return err;

search_again:
	if (start > end)
		goto out;
1014
	spin_unlock(&tree->lock);
1015 1016 1017 1018 1019
	if (mask & __GFP_WAIT)
		cond_resched();
	goto again;
}

1020 1021 1022
int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
		   unsigned long bits, u64 * failed_start,
		   struct extent_state **cached_state, gfp_t mask)
J
Jeff Mahoney 已提交
1023 1024 1025 1026 1027 1028
{
	return __set_extent_bit(tree, start, end, bits, 0, failed_start,
				cached_state, mask);
}


J
Josef Bacik 已提交
1029
/**
L
Liu Bo 已提交
1030 1031
 * convert_extent_bit - convert all bits in a given range from one bit to
 * 			another
J
Josef Bacik 已提交
1032 1033 1034 1035 1036
 * @tree:	the io tree to search
 * @start:	the start offset in bytes
 * @end:	the end offset in bytes (inclusive)
 * @bits:	the bits to set in this range
 * @clear_bits:	the bits to clear in this range
1037
 * @cached_state:	state that we're going to cache
J
Josef Bacik 已提交
1038 1039 1040 1041 1042 1043 1044 1045 1046
 * @mask:	the allocation mask
 *
 * This will go through and set bits for the given range.  If any states exist
 * already in this range they are set with the given bit and cleared of the
 * clear_bits.  This is only meant to be used by things that are mergeable, ie
 * converting from say DELALLOC to DIRTY.  This is not meant to be used with
 * boundary bits like LOCK.
 */
int convert_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
1047
		       unsigned long bits, unsigned long clear_bits,
1048
		       struct extent_state **cached_state, gfp_t mask)
J
Josef Bacik 已提交
1049 1050 1051 1052
{
	struct extent_state *state;
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
1053 1054
	struct rb_node **p;
	struct rb_node *parent;
J
Josef Bacik 已提交
1055 1056 1057 1058
	int err = 0;
	u64 last_start;
	u64 last_end;

1059
	btrfs_debug_check_extent_io_range(tree, start, end);
1060

J
Josef Bacik 已提交
1061 1062 1063 1064 1065 1066 1067 1068
again:
	if (!prealloc && (mask & __GFP_WAIT)) {
		prealloc = alloc_extent_state(mask);
		if (!prealloc)
			return -ENOMEM;
	}

	spin_lock(&tree->lock);
1069 1070 1071 1072 1073 1074 1075 1076 1077
	if (cached_state && *cached_state) {
		state = *cached_state;
		if (state->start <= start && state->end > start &&
		    state->tree) {
			node = &state->rb_node;
			goto hit_next;
		}
	}

J
Josef Bacik 已提交
1078 1079 1080 1081
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1082
	node = tree_search_for_insert(tree, start, &p, &parent);
J
Josef Bacik 已提交
1083 1084
	if (!node) {
		prealloc = alloc_extent_state_atomic(prealloc);
1085 1086 1087 1088
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
1089 1090
		err = insert_state(tree, prealloc, start, end,
				   &p, &parent, &bits);
1091 1092
		if (err)
			extent_io_tree_panic(tree, err);
1093 1094
		cache_state(prealloc, cached_state);
		prealloc = NULL;
J
Josef Bacik 已提交
1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109
		goto out;
	}
	state = rb_entry(node, struct extent_state, rb_node);
hit_next:
	last_start = state->start;
	last_end = state->end;

	/*
	 * | ---- desired range ---- |
	 * | state |
	 *
	 * Just lock what we found and keep going
	 */
	if (state->start == start && state->end <= end) {
		set_state_bits(tree, state, &bits);
1110
		cache_state(state, cached_state);
1111
		state = clear_state_bit(tree, state, &clear_bits, 0);
J
Josef Bacik 已提交
1112 1113 1114
		if (last_end == (u64)-1)
			goto out;
		start = last_end + 1;
1115 1116 1117
		if (start < end && state && state->start == start &&
		    !need_resched())
			goto hit_next;
J
Josef Bacik 已提交
1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138
		goto search_again;
	}

	/*
	 *     | ---- desired range ---- |
	 * | state |
	 *   or
	 * | ------------- state -------------- |
	 *
	 * We need to split the extent we found, and may flip bits on
	 * second half.
	 *
	 * If the extent we found extends past our
	 * range, we just split and search again.  It'll get split
	 * again the next time though.
	 *
	 * If the extent we found is inside our range, we set the
	 * desired bit on it.
	 */
	if (state->start < start) {
		prealloc = alloc_extent_state_atomic(prealloc);
1139 1140 1141 1142
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1143
		err = split_state(tree, state, prealloc, start);
1144 1145
		if (err)
			extent_io_tree_panic(tree, err);
J
Josef Bacik 已提交
1146 1147 1148 1149 1150
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
			set_state_bits(tree, state, &bits);
1151
			cache_state(state, cached_state);
1152
			state = clear_state_bit(tree, state, &clear_bits, 0);
J
Josef Bacik 已提交
1153 1154 1155
			if (last_end == (u64)-1)
				goto out;
			start = last_end + 1;
1156 1157 1158
			if (start < end && state && state->start == start &&
			    !need_resched())
				goto hit_next;
J
Josef Bacik 已提交
1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176
		}
		goto search_again;
	}
	/*
	 * | ---- desired range ---- |
	 *     | state | or               | state |
	 *
	 * There's a hole, we need to insert something in it and
	 * ignore the extent we found.
	 */
	if (state->start > start) {
		u64 this_end;
		if (end < last_start)
			this_end = end;
		else
			this_end = last_start - 1;

		prealloc = alloc_extent_state_atomic(prealloc);
1177 1178 1179 1180
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1181 1182 1183 1184 1185 1186

		/*
		 * Avoid to free 'prealloc' if it can be merged with
		 * the later extent.
		 */
		err = insert_state(tree, prealloc, start, this_end,
1187
				   NULL, NULL, &bits);
1188 1189
		if (err)
			extent_io_tree_panic(tree, err);
1190
		cache_state(prealloc, cached_state);
J
Josef Bacik 已提交
1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202
		prealloc = NULL;
		start = this_end + 1;
		goto search_again;
	}
	/*
	 * | ---- desired range ---- |
	 *                        | state |
	 * We need to split the extent, and set the bit
	 * on the first half
	 */
	if (state->start <= end && state->end > end) {
		prealloc = alloc_extent_state_atomic(prealloc);
1203 1204 1205 1206
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1207 1208

		err = split_state(tree, state, prealloc, end + 1);
1209 1210
		if (err)
			extent_io_tree_panic(tree, err);
J
Josef Bacik 已提交
1211 1212

		set_state_bits(tree, prealloc, &bits);
1213
		cache_state(prealloc, cached_state);
J
Josef Bacik 已提交
1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236
		clear_state_bit(tree, prealloc, &clear_bits, 0);
		prealloc = NULL;
		goto out;
	}

	goto search_again;

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

	return err;

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

1237 1238 1239 1240
/* wrappers around set/clear extent bit */
int set_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
		     gfp_t mask)
{
J
Jeff Mahoney 已提交
1241
	return set_extent_bit(tree, start, end, EXTENT_DIRTY, NULL,
1242
			      NULL, mask);
1243 1244 1245
}

int set_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1246
		    unsigned long bits, gfp_t mask)
1247
{
J
Jeff Mahoney 已提交
1248
	return set_extent_bit(tree, start, end, bits, NULL,
1249
			      NULL, mask);
1250 1251 1252
}

int clear_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1253
		      unsigned long bits, gfp_t mask)
1254
{
1255
	return clear_extent_bit(tree, start, end, bits, 0, 0, NULL, mask);
1256 1257 1258
}

int set_extent_delalloc(struct extent_io_tree *tree, u64 start, u64 end,
1259
			struct extent_state **cached_state, gfp_t mask)
1260 1261
{
	return set_extent_bit(tree, start, end,
1262
			      EXTENT_DELALLOC | EXTENT_UPTODATE,
J
Jeff Mahoney 已提交
1263
			      NULL, cached_state, mask);
1264 1265
}

1266 1267 1268 1269 1270 1271 1272 1273
int set_extent_defrag(struct extent_io_tree *tree, u64 start, u64 end,
		      struct extent_state **cached_state, gfp_t mask)
{
	return set_extent_bit(tree, start, end,
			      EXTENT_DELALLOC | EXTENT_UPTODATE | EXTENT_DEFRAG,
			      NULL, cached_state, mask);
}

1274 1275 1276 1277
int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
		       gfp_t mask)
{
	return clear_extent_bit(tree, start, end,
1278
				EXTENT_DIRTY | EXTENT_DELALLOC |
1279
				EXTENT_DO_ACCOUNTING, 0, 0, NULL, mask);
1280 1281 1282 1283 1284
}

int set_extent_new(struct extent_io_tree *tree, u64 start, u64 end,
		     gfp_t mask)
{
J
Jeff Mahoney 已提交
1285
	return set_extent_bit(tree, start, end, EXTENT_NEW, NULL,
1286
			      NULL, mask);
1287 1288 1289
}

int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end,
1290
			struct extent_state **cached_state, gfp_t mask)
1291
{
L
Liu Bo 已提交
1292
	return set_extent_bit(tree, start, end, EXTENT_UPTODATE, NULL,
J
Jeff Mahoney 已提交
1293
			      cached_state, mask);
1294 1295
}

1296 1297
int clear_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end,
			  struct extent_state **cached_state, gfp_t mask)
1298
{
1299
	return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0,
1300
				cached_state, mask);
1301 1302
}

C
Chris Mason 已提交
1303 1304 1305 1306
/*
 * either insert or lock state struct between start and end use mask to tell
 * us if waiting is desired.
 */
1307
int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1308
		     unsigned long bits, struct extent_state **cached_state)
1309 1310 1311 1312
{
	int err;
	u64 failed_start;
	while (1) {
J
Jeff Mahoney 已提交
1313 1314 1315
		err = __set_extent_bit(tree, start, end, EXTENT_LOCKED | bits,
				       EXTENT_LOCKED, &failed_start,
				       cached_state, GFP_NOFS);
1316
		if (err == -EEXIST) {
1317 1318
			wait_extent_bit(tree, failed_start, end, EXTENT_LOCKED);
			start = failed_start;
1319
		} else
1320 1321 1322 1323 1324 1325
			break;
		WARN_ON(start > end);
	}
	return err;
}

1326
int lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
1327
{
1328
	return lock_extent_bits(tree, start, end, 0, NULL);
1329 1330
}

1331
int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
1332 1333 1334 1335
{
	int err;
	u64 failed_start;

J
Jeff Mahoney 已提交
1336 1337
	err = __set_extent_bit(tree, start, end, EXTENT_LOCKED, EXTENT_LOCKED,
			       &failed_start, NULL, GFP_NOFS);
Y
Yan Zheng 已提交
1338 1339 1340
	if (err == -EEXIST) {
		if (failed_start > start)
			clear_extent_bit(tree, start, failed_start - 1,
1341
					 EXTENT_LOCKED, 1, 0, NULL, GFP_NOFS);
1342
		return 0;
Y
Yan Zheng 已提交
1343
	}
1344 1345 1346
	return 1;
}

1347 1348 1349 1350 1351 1352 1353
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);
}

1354
int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end)
1355
{
1356
	return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL,
1357
				GFP_NOFS);
1358 1359
}

1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392
int extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end)
{
	unsigned long index = start >> PAGE_CACHE_SHIFT;
	unsigned long end_index = end >> PAGE_CACHE_SHIFT;
	struct page *page;

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

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

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

1393 1394 1395
/*
 * helper function to set both pages and extents in the tree writeback
 */
1396
static int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end)
1397 1398 1399 1400 1401 1402 1403
{
	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);
1404
		BUG_ON(!page); /* Pages should be in the extent_io_tree */
1405 1406 1407 1408 1409 1410 1411
		set_page_writeback(page);
		page_cache_release(page);
		index++;
	}
	return 0;
}

C
Chris Mason 已提交
1412 1413 1414 1415
/* 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'
 */
1416 1417
static struct extent_state *
find_first_extent_bit_state(struct extent_io_tree *tree,
1418
			    u64 start, unsigned long bits)
C
Chris Mason 已提交
1419 1420 1421 1422 1423 1424 1425 1426 1427
{
	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
1428
	if (!node)
C
Chris Mason 已提交
1429 1430
		goto out;

C
Chris Mason 已提交
1431
	while (1) {
C
Chris Mason 已提交
1432
		state = rb_entry(node, struct extent_state, rb_node);
C
Chris Mason 已提交
1433
		if (state->end >= start && (state->state & bits))
C
Chris Mason 已提交
1434
			return state;
C
Chris Mason 已提交
1435

C
Chris Mason 已提交
1436 1437 1438 1439 1440 1441 1442 1443
		node = rb_next(node);
		if (!node)
			break;
	}
out:
	return NULL;
}

1444 1445 1446 1447 1448
/*
 * 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.
 *
1449
 * If nothing was found, 1 is returned. If found something, return 0.
1450 1451
 */
int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
1452
			  u64 *start_ret, u64 *end_ret, unsigned long bits,
1453
			  struct extent_state **cached_state)
1454 1455
{
	struct extent_state *state;
1456
	struct rb_node *n;
1457 1458 1459
	int ret = 1;

	spin_lock(&tree->lock);
1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478
	if (cached_state && *cached_state) {
		state = *cached_state;
		if (state->end == start - 1 && state->tree) {
			n = rb_next(&state->rb_node);
			while (n) {
				state = rb_entry(n, struct extent_state,
						 rb_node);
				if (state->state & bits)
					goto got_it;
				n = rb_next(n);
			}
			free_extent_state(*cached_state);
			*cached_state = NULL;
			goto out;
		}
		free_extent_state(*cached_state);
		*cached_state = NULL;
	}

1479
	state = find_first_extent_bit_state(tree, start, bits);
1480
got_it:
1481
	if (state) {
1482
		cache_state(state, cached_state);
1483 1484 1485 1486
		*start_ret = state->start;
		*end_ret = state->end;
		ret = 0;
	}
1487
out:
1488 1489 1490 1491
	spin_unlock(&tree->lock);
	return ret;
}

C
Chris Mason 已提交
1492 1493 1494 1495 1496 1497
/*
 * 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
1498
static noinline u64 find_delalloc_range(struct extent_io_tree *tree,
1499 1500
					u64 *start, u64 *end, u64 max_bytes,
					struct extent_state **cached_state)
1501 1502 1503 1504 1505 1506 1507
{
	struct rb_node *node;
	struct extent_state *state;
	u64 cur_start = *start;
	u64 found = 0;
	u64 total_bytes = 0;

1508
	spin_lock(&tree->lock);
C
Chris Mason 已提交
1509

1510 1511 1512 1513
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1514
	node = tree_search(tree, cur_start);
1515
	if (!node) {
1516 1517
		if (!found)
			*end = (u64)-1;
1518 1519 1520
		goto out;
	}

C
Chris Mason 已提交
1521
	while (1) {
1522
		state = rb_entry(node, struct extent_state, rb_node);
1523 1524
		if (found && (state->start != cur_start ||
			      (state->state & EXTENT_BOUNDARY))) {
1525 1526 1527 1528 1529 1530 1531
			goto out;
		}
		if (!(state->state & EXTENT_DELALLOC)) {
			if (!found)
				*end = state->end;
			goto out;
		}
1532
		if (!found) {
1533
			*start = state->start;
1534 1535 1536
			*cached_state = state;
			atomic_inc(&state->refs);
		}
1537 1538 1539 1540 1541
		found++;
		*end = state->end;
		cur_start = state->end + 1;
		node = rb_next(node);
		total_bytes += state->end - state->start + 1;
1542
		if (total_bytes >= max_bytes)
1543 1544
			break;
		if (!node)
1545 1546 1547
			break;
	}
out:
1548
	spin_unlock(&tree->lock);
1549 1550 1551
	return found;
}

1552 1553 1554
static noinline void __unlock_for_delalloc(struct inode *inode,
					   struct page *locked_page,
					   u64 start, u64 end)
C
Chris Mason 已提交
1555 1556 1557 1558 1559 1560 1561 1562 1563
{
	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)
1564
		return;
C
Chris Mason 已提交
1565

C
Chris Mason 已提交
1566
	while (nr_pages > 0) {
C
Chris Mason 已提交
1567
		ret = find_get_pages_contig(inode->i_mapping, index,
1568 1569
				     min_t(unsigned long, nr_pages,
				     ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600
		for (i = 0; i < ret; i++) {
			if (pages[i] != locked_page)
				unlock_page(pages[i]);
			page_cache_release(pages[i]);
		}
		nr_pages -= ret;
		index += ret;
		cond_resched();
	}
}

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

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

	/* skip the page at the start index */
	nrpages = end_index - index + 1;
C
Chris Mason 已提交
1601
	while (nrpages > 0) {
C
Chris Mason 已提交
1602
		ret = find_get_pages_contig(inode->i_mapping, index,
1603 1604
				     min_t(unsigned long,
				     nrpages, ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1605 1606 1607 1608 1609 1610 1611 1612 1613 1614
		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
			 */
1615
			if (pages[i] != locked_page) {
C
Chris Mason 已提交
1616
				lock_page(pages[i]);
1617 1618
				if (!PageDirty(pages[i]) ||
				    pages[i]->mapping != inode->i_mapping) {
1619 1620 1621 1622 1623 1624
					ret = -EAGAIN;
					unlock_page(pages[i]);
					page_cache_release(pages[i]);
					goto done;
				}
			}
C
Chris Mason 已提交
1625
			page_cache_release(pages[i]);
1626
			pages_locked++;
C
Chris Mason 已提交
1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648
		}
		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
 */
1649 1650 1651 1652
STATIC u64 find_lock_delalloc_range(struct inode *inode,
				    struct extent_io_tree *tree,
				    struct page *locked_page, u64 *start,
				    u64 *end, u64 max_bytes)
C
Chris Mason 已提交
1653 1654 1655 1656
{
	u64 delalloc_start;
	u64 delalloc_end;
	u64 found;
1657
	struct extent_state *cached_state = NULL;
C
Chris Mason 已提交
1658 1659 1660 1661 1662 1663 1664 1665
	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,
1666
				    max_bytes, &cached_state);
C
Chris Mason 已提交
1667
	if (!found || delalloc_end <= *start) {
C
Chris Mason 已提交
1668 1669
		*start = delalloc_start;
		*end = delalloc_end;
1670
		free_extent_state(cached_state);
L
Liu Bo 已提交
1671
		return 0;
C
Chris Mason 已提交
1672 1673
	}

C
Chris Mason 已提交
1674 1675 1676 1677 1678
	/*
	 * 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
1679
	if (delalloc_start < *start)
C
Chris Mason 已提交
1680 1681
		delalloc_start = *start;

C
Chris Mason 已提交
1682 1683 1684
	/*
	 * make sure to limit the number of pages we try to lock down
	 */
1685 1686
	if (delalloc_end + 1 - delalloc_start > max_bytes)
		delalloc_end = delalloc_start + max_bytes - 1;
C
Chris Mason 已提交
1687

C
Chris Mason 已提交
1688 1689 1690 1691 1692 1693 1694
	/* 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
		 */
1695
		free_extent_state(cached_state);
1696
		cached_state = NULL;
C
Chris Mason 已提交
1697
		if (!loops) {
1698
			max_bytes = PAGE_CACHE_SIZE;
C
Chris Mason 已提交
1699 1700 1701 1702 1703 1704 1705
			loops = 1;
			goto again;
		} else {
			found = 0;
			goto out_failed;
		}
	}
1706
	BUG_ON(ret); /* Only valid values are 0 and -EAGAIN */
C
Chris Mason 已提交
1707 1708

	/* step three, lock the state bits for the whole range */
1709
	lock_extent_bits(tree, delalloc_start, delalloc_end, 0, &cached_state);
C
Chris Mason 已提交
1710 1711 1712

	/* then test to make sure it is all still delalloc */
	ret = test_range_bit(tree, delalloc_start, delalloc_end,
1713
			     EXTENT_DELALLOC, 1, cached_state);
C
Chris Mason 已提交
1714
	if (!ret) {
1715 1716
		unlock_extent_cached(tree, delalloc_start, delalloc_end,
				     &cached_state, GFP_NOFS);
C
Chris Mason 已提交
1717 1718 1719 1720 1721
		__unlock_for_delalloc(inode, locked_page,
			      delalloc_start, delalloc_end);
		cond_resched();
		goto again;
	}
1722
	free_extent_state(cached_state);
C
Chris Mason 已提交
1723 1724 1725 1726 1727 1728
	*start = delalloc_start;
	*end = delalloc_end;
out_failed:
	return found;
}

1729 1730 1731 1732
int extent_clear_unlock_delalloc(struct inode *inode, u64 start, u64 end,
				 struct page *locked_page,
				 unsigned long clear_bits,
				 unsigned long page_ops)
C
Chris Mason 已提交
1733
{
1734
	struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
C
Chris Mason 已提交
1735 1736 1737 1738 1739 1740
	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;
1741

1742
	clear_extent_bit(tree, start, end, clear_bits, 1, 0, NULL, GFP_NOFS);
1743
	if (page_ops == 0)
1744
		return 0;
C
Chris Mason 已提交
1745

C
Chris Mason 已提交
1746
	while (nr_pages > 0) {
C
Chris Mason 已提交
1747
		ret = find_get_pages_contig(inode->i_mapping, index,
1748 1749
				     min_t(unsigned long,
				     nr_pages, ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1750
		for (i = 0; i < ret; i++) {
1751

1752
			if (page_ops & PAGE_SET_PRIVATE2)
1753 1754
				SetPagePrivate2(pages[i]);

C
Chris Mason 已提交
1755 1756 1757 1758
			if (pages[i] == locked_page) {
				page_cache_release(pages[i]);
				continue;
			}
1759
			if (page_ops & PAGE_CLEAR_DIRTY)
C
Chris Mason 已提交
1760
				clear_page_dirty_for_io(pages[i]);
1761
			if (page_ops & PAGE_SET_WRITEBACK)
C
Chris Mason 已提交
1762
				set_page_writeback(pages[i]);
1763
			if (page_ops & PAGE_END_WRITEBACK)
C
Chris Mason 已提交
1764
				end_page_writeback(pages[i]);
1765
			if (page_ops & PAGE_UNLOCK)
1766
				unlock_page(pages[i]);
C
Chris Mason 已提交
1767 1768 1769 1770 1771 1772 1773 1774 1775
			page_cache_release(pages[i]);
		}
		nr_pages -= ret;
		index += ret;
		cond_resched();
	}
	return 0;
}

C
Chris Mason 已提交
1776 1777 1778 1779 1780
/*
 * 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.
 */
1781 1782
u64 count_range_bits(struct extent_io_tree *tree,
		     u64 *start, u64 search_end, u64 max_bytes,
1783
		     unsigned long bits, int contig)
1784 1785 1786 1787 1788
{
	struct rb_node *node;
	struct extent_state *state;
	u64 cur_start = *start;
	u64 total_bytes = 0;
1789
	u64 last = 0;
1790 1791
	int found = 0;

1792
	if (WARN_ON(search_end <= cur_start))
1793 1794
		return 0;

1795
	spin_lock(&tree->lock);
1796 1797 1798 1799 1800 1801 1802 1803
	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.
	 */
1804
	node = tree_search(tree, cur_start);
C
Chris Mason 已提交
1805
	if (!node)
1806 1807
		goto out;

C
Chris Mason 已提交
1808
	while (1) {
1809 1810 1811
		state = rb_entry(node, struct extent_state, rb_node);
		if (state->start > search_end)
			break;
1812 1813 1814
		if (contig && found && state->start > last + 1)
			break;
		if (state->end >= cur_start && (state->state & bits) == bits) {
1815 1816 1817 1818 1819
			total_bytes += min(search_end, state->end) + 1 -
				       max(cur_start, state->start);
			if (total_bytes >= max_bytes)
				break;
			if (!found) {
1820
				*start = max(cur_start, state->start);
1821 1822
				found = 1;
			}
1823 1824 1825
			last = state->end;
		} else if (contig && found) {
			break;
1826 1827 1828 1829 1830 1831
		}
		node = rb_next(node);
		if (!node)
			break;
	}
out:
1832
	spin_unlock(&tree->lock);
1833 1834
	return total_bytes;
}
1835

C
Chris Mason 已提交
1836 1837 1838 1839
/*
 * set the private field for a given byte offset in the tree.  If there isn't
 * an extent_state there already, this does nothing.
 */
1840
static int set_state_private(struct extent_io_tree *tree, u64 start, u64 private)
1841 1842 1843 1844 1845
{
	struct rb_node *node;
	struct extent_state *state;
	int ret = 0;

1846
	spin_lock(&tree->lock);
1847 1848 1849 1850
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1851
	node = tree_search(tree, start);
1852
	if (!node) {
1853 1854 1855 1856 1857 1858 1859 1860 1861 1862
		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:
1863
	spin_unlock(&tree->lock);
1864 1865 1866 1867 1868 1869 1870 1871 1872
	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;

1873
	spin_lock(&tree->lock);
1874 1875 1876 1877
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1878
	node = tree_search(tree, start);
1879
	if (!node) {
1880 1881 1882 1883 1884 1885 1886 1887 1888 1889
		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:
1890
	spin_unlock(&tree->lock);
1891 1892 1893 1894 1895
	return ret;
}

/*
 * searches a range in the state tree for a given mask.
1896
 * If 'filled' == 1, this returns 1 only if every extent in the tree
1897 1898 1899 1900
 * 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,
1901
		   unsigned long bits, int filled, struct extent_state *cached)
1902 1903 1904 1905 1906
{
	struct extent_state *state = NULL;
	struct rb_node *node;
	int bitset = 0;

1907
	spin_lock(&tree->lock);
1908 1909
	if (cached && cached->tree && cached->start <= start &&
	    cached->end > start)
1910 1911 1912
		node = &cached->rb_node;
	else
		node = tree_search(tree, start);
1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931
	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;
		}
1932 1933 1934 1935

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

1936 1937 1938 1939 1940 1941 1942 1943 1944 1945
		start = state->end + 1;
		if (start > end)
			break;
		node = rb_next(node);
		if (!node) {
			if (filled)
				bitset = 0;
			break;
		}
	}
1946
	spin_unlock(&tree->lock);
1947 1948 1949 1950 1951 1952 1953
	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
 */
1954
static void check_page_uptodate(struct extent_io_tree *tree, struct page *page)
1955
{
M
Miao Xie 已提交
1956
	u64 start = page_offset(page);
1957
	u64 end = start + PAGE_CACHE_SIZE - 1;
1958
	if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1, NULL))
1959 1960 1961
		SetPageUptodate(page);
}

1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994
/*
 * When IO fails, either with EIO or csum verification fails, we
 * try other mirrors that might have a good copy of the data.  This
 * io_failure_record is used to record state as we go through all the
 * mirrors.  If another mirror has good data, the page is set up to date
 * and things continue.  If a good mirror can't be found, the original
 * bio end_io callback is called to indicate things have failed.
 */
struct io_failure_record {
	struct page *page;
	u64 start;
	u64 len;
	u64 logical;
	unsigned long bio_flags;
	int this_mirror;
	int failed_mirror;
	int in_validation;
};

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

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

D
David Woodhouse 已提交
1995 1996 1997 1998 1999
	ret = clear_extent_bits(&BTRFS_I(inode)->io_tree, rec->start,
				rec->start + rec->len - 1,
				EXTENT_DAMAGED, GFP_NOFS);
	if (ret && !err)
		err = ret;
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

	kfree(rec);
	return err;
}

/*
 * this bypasses the standard btrfs submit functions deliberately, as
 * the standard behavior is to write all copies in a raid setup. here we only
 * want to write the one bad copy. so we do the mapping for ourselves and issue
 * submit_bio directly.
2010
 * to avoid any synchronization issues, wait for the data after writing, which
2011 2012 2013 2014
 * actually prevents the read that triggered the error from finishing.
 * currently, there can be no more than two copies of every data bit. thus,
 * exactly one rewrite is required.
 */
2015
int repair_io_failure(struct btrfs_fs_info *fs_info, u64 start,
2016 2017 2018 2019 2020 2021 2022 2023
			u64 length, u64 logical, struct page *page,
			int mirror_num)
{
	struct bio *bio;
	struct btrfs_device *dev;
	u64 map_length = 0;
	u64 sector;
	struct btrfs_bio *bbio = NULL;
D
David Woodhouse 已提交
2024
	struct btrfs_mapping_tree *map_tree = &fs_info->mapping_tree;
2025 2026
	int ret;

2027
	ASSERT(!(fs_info->sb->s_flags & MS_RDONLY));
2028 2029
	BUG_ON(!mirror_num);

D
David Woodhouse 已提交
2030 2031 2032 2033
	/* we can't repair anything in raid56 yet */
	if (btrfs_is_parity_mirror(map_tree, logical, length, mirror_num))
		return 0;

2034
	bio = btrfs_io_bio_alloc(GFP_NOFS, 1);
2035 2036
	if (!bio)
		return -EIO;
2037
	bio->bi_iter.bi_size = 0;
2038 2039
	map_length = length;

2040
	ret = btrfs_map_block(fs_info, WRITE, logical,
2041 2042 2043 2044 2045 2046 2047
			      &map_length, &bbio, mirror_num);
	if (ret) {
		bio_put(bio);
		return -EIO;
	}
	BUG_ON(mirror_num != bbio->mirror_num);
	sector = bbio->stripes[mirror_num-1].physical >> 9;
2048
	bio->bi_iter.bi_sector = sector;
2049 2050 2051 2052 2053 2054 2055
	dev = bbio->stripes[mirror_num-1].dev;
	kfree(bbio);
	if (!dev || !dev->bdev || !dev->writeable) {
		bio_put(bio);
		return -EIO;
	}
	bio->bi_bdev = dev->bdev;
M
Miao Xie 已提交
2056
	bio_add_page(bio, page, length, start - page_offset(page));
2057

2058
	if (btrfsic_submit_bio_wait(WRITE_SYNC, bio)) {
2059 2060
		/* try to remap that extent elsewhere? */
		bio_put(bio);
2061
		btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_WRITE_ERRS);
2062 2063 2064
		return -EIO;
	}

2065 2066 2067 2068
	printk_ratelimited_in_rcu(KERN_INFO
			"BTRFS: read error corrected: ino %lu off %llu "
		    "(dev %s sector %llu)\n", page->mapping->host->i_ino,
		    start, rcu_str_deref(dev->name), sector);
2069 2070 2071 2072 2073

	bio_put(bio);
	return 0;
}

2074 2075 2076 2077 2078
int repair_eb_io_failure(struct btrfs_root *root, struct extent_buffer *eb,
			 int mirror_num)
{
	u64 start = eb->start;
	unsigned long i, num_pages = num_extent_pages(eb->start, eb->len);
2079
	int ret = 0;
2080

2081 2082 2083
	if (root->fs_info->sb->s_flags & MS_RDONLY)
		return -EROFS;

2084 2085
	for (i = 0; i < num_pages; i++) {
		struct page *p = extent_buffer_page(eb, i);
2086
		ret = repair_io_failure(root->fs_info, start, PAGE_CACHE_SIZE,
2087 2088 2089 2090 2091 2092 2093 2094 2095
					start, p, mirror_num);
		if (ret)
			break;
		start += PAGE_CACHE_SIZE;
	}

	return ret;
}

2096 2097 2098 2099 2100 2101 2102 2103 2104
/*
 * each time an IO finishes, we do a fast check in the IO failure tree
 * to see if we need to process or clean up an io_failure_record
 */
static int clean_io_failure(u64 start, struct page *page)
{
	u64 private;
	u64 private_failure;
	struct io_failure_record *failrec;
2105 2106
	struct inode *inode = page->mapping->host;
	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132
	struct extent_state *state;
	int num_copies;
	int did_repair = 0;
	int ret;

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

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

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

	if (failrec->in_validation) {
		/* there was no real error, just free the record */
		pr_debug("clean_io_failure: freeing dummy error at %llu\n",
			 failrec->start);
		did_repair = 1;
		goto out;
	}
2133 2134
	if (fs_info->sb->s_flags & MS_RDONLY)
		goto out;
2135 2136 2137 2138 2139 2140 2141

	spin_lock(&BTRFS_I(inode)->io_tree.lock);
	state = find_first_extent_bit_state(&BTRFS_I(inode)->io_tree,
					    failrec->start,
					    EXTENT_LOCKED);
	spin_unlock(&BTRFS_I(inode)->io_tree.lock);

2142 2143
	if (state && state->start <= failrec->start &&
	    state->end >= failrec->start + failrec->len - 1) {
2144 2145
		num_copies = btrfs_num_copies(fs_info, failrec->logical,
					      failrec->len);
2146
		if (num_copies > 1)  {
2147
			ret = repair_io_failure(fs_info, start, failrec->len,
2148 2149 2150 2151
						failrec->logical, page,
						failrec->failed_mirror);
			did_repair = !ret;
		}
D
David Woodhouse 已提交
2152
		ret = 0;
2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169
	}

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

	return ret;
}

/*
 * this is a generic handler for readpage errors (default
 * readpage_io_failed_hook). if other copies exist, read those and write back
 * good data to the failed position. does not investigate in remapping the
 * failed extent elsewhere, hoping the device will be smart enough to do this as
 * needed
 */

2170 2171 2172
static int bio_readpage_error(struct bio *failed_bio, u64 phy_offset,
			      struct page *page, u64 start, u64 end,
			      int failed_mirror)
2173 2174 2175 2176 2177 2178 2179 2180 2181
{
	struct io_failure_record *failrec = NULL;
	u64 private;
	struct extent_map *em;
	struct inode *inode = page->mapping->host;
	struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
	struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
	struct bio *bio;
2182 2183
	struct btrfs_io_bio *btrfs_failed_bio;
	struct btrfs_io_bio *btrfs_bio;
2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209
	int num_copies;
	int ret;
	int read_mode;
	u64 logical;

	BUG_ON(failed_bio->bi_rw & REQ_WRITE);

	ret = get_state_private(failure_tree, start, &private);
	if (ret) {
		failrec = kzalloc(sizeof(*failrec), GFP_NOFS);
		if (!failrec)
			return -ENOMEM;
		failrec->start = start;
		failrec->len = end - start + 1;
		failrec->this_mirror = 0;
		failrec->bio_flags = 0;
		failrec->in_validation = 0;

		read_lock(&em_tree->lock);
		em = lookup_extent_mapping(em_tree, start, failrec->len);
		if (!em) {
			read_unlock(&em_tree->lock);
			kfree(failrec);
			return -EIO;
		}

2210
		if (em->start > start || em->start + em->len <= start) {
2211 2212 2213 2214 2215
			free_extent_map(em);
			em = NULL;
		}
		read_unlock(&em_tree->lock);

2216
		if (!em) {
2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258
			kfree(failrec);
			return -EIO;
		}
		logical = start - em->start;
		logical = em->block_start + logical;
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
			logical = em->block_start;
			failrec->bio_flags = EXTENT_BIO_COMPRESSED;
			extent_set_compress_type(&failrec->bio_flags,
						 em->compress_type);
		}
		pr_debug("bio_readpage_error: (new) logical=%llu, start=%llu, "
			 "len=%llu\n", logical, start, failrec->len);
		failrec->logical = logical;
		free_extent_map(em);

		/* set the bits in the private failure tree */
		ret = set_extent_bits(failure_tree, start, end,
					EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
		if (ret >= 0)
			ret = set_state_private(failure_tree, start,
						(u64)(unsigned long)failrec);
		/* set the bits in the inode's tree */
		if (ret >= 0)
			ret = set_extent_bits(tree, start, end, EXTENT_DAMAGED,
						GFP_NOFS);
		if (ret < 0) {
			kfree(failrec);
			return ret;
		}
	} else {
		failrec = (struct io_failure_record *)(unsigned long)private;
		pr_debug("bio_readpage_error: (found) logical=%llu, "
			 "start=%llu, len=%llu, validation=%d\n",
			 failrec->logical, failrec->start, failrec->len,
			 failrec->in_validation);
		/*
		 * when data can be on disk more than twice, add to failrec here
		 * (e.g. with a list for failed_mirror) to make
		 * clean_io_failure() clean all those errors at once.
		 */
	}
2259 2260
	num_copies = btrfs_num_copies(BTRFS_I(inode)->root->fs_info,
				      failrec->logical, failrec->len);
2261 2262 2263 2264 2265 2266
	if (num_copies == 1) {
		/*
		 * we only have a single copy of the data, so don't bother with
		 * all the retry and error correction code that follows. no
		 * matter what the error is, it is very likely to persist.
		 */
2267 2268
		pr_debug("bio_readpage_error: cannot repair, num_copies=%d, next_mirror %d, failed_mirror %d\n",
			 num_copies, failrec->this_mirror, failed_mirror);
2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308
		free_io_failure(inode, failrec, 0);
		return -EIO;
	}

	/*
	 * there are two premises:
	 *	a) deliver good data to the caller
	 *	b) correct the bad sectors on disk
	 */
	if (failed_bio->bi_vcnt > 1) {
		/*
		 * to fulfill b), we need to know the exact failing sectors, as
		 * we don't want to rewrite any more than the failed ones. thus,
		 * we need separate read requests for the failed bio
		 *
		 * if the following BUG_ON triggers, our validation request got
		 * merged. we need separate requests for our algorithm to work.
		 */
		BUG_ON(failrec->in_validation);
		failrec->in_validation = 1;
		failrec->this_mirror = failed_mirror;
		read_mode = READ_SYNC | REQ_FAILFAST_DEV;
	} else {
		/*
		 * we're ready to fulfill a) and b) alongside. get a good copy
		 * of the failed sector and if we succeed, we have setup
		 * everything for repair_io_failure to do the rest for us.
		 */
		if (failrec->in_validation) {
			BUG_ON(failrec->this_mirror != failed_mirror);
			failrec->in_validation = 0;
			failrec->this_mirror = 0;
		}
		failrec->failed_mirror = failed_mirror;
		failrec->this_mirror++;
		if (failrec->this_mirror == failed_mirror)
			failrec->this_mirror++;
		read_mode = READ_SYNC;
	}

2309 2310
	if (failrec->this_mirror > num_copies) {
		pr_debug("bio_readpage_error: (fail) num_copies=%d, next_mirror %d, failed_mirror %d\n",
2311 2312 2313 2314 2315
			 num_copies, failrec->this_mirror, failed_mirror);
		free_io_failure(inode, failrec, 0);
		return -EIO;
	}

2316
	bio = btrfs_io_bio_alloc(GFP_NOFS, 1);
2317 2318 2319 2320
	if (!bio) {
		free_io_failure(inode, failrec, 0);
		return -EIO;
	}
2321
	bio->bi_end_io = failed_bio->bi_end_io;
2322
	bio->bi_iter.bi_sector = failrec->logical >> 9;
2323
	bio->bi_bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
2324
	bio->bi_iter.bi_size = 0;
2325

2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338
	btrfs_failed_bio = btrfs_io_bio(failed_bio);
	if (btrfs_failed_bio->csum) {
		struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
		u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);

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

2339 2340 2341 2342 2343 2344
	bio_add_page(bio, page, failrec->len, start - page_offset(page));

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

2345 2346 2347 2348
	ret = tree->ops->submit_bio_hook(inode, read_mode, bio,
					 failrec->this_mirror,
					 failrec->bio_flags, 0);
	return ret;
2349 2350
}

2351 2352
/* lots and lots of room for performance fixes in the end_bio funcs */

2353 2354 2355 2356
int end_extent_writepage(struct page *page, int err, u64 start, u64 end)
{
	int uptodate = (err == 0);
	struct extent_io_tree *tree;
2357
	int ret = 0;
2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370

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

	if (tree->ops && tree->ops->writepage_end_io_hook) {
		ret = tree->ops->writepage_end_io_hook(page, start,
					       end, NULL, uptodate);
		if (ret)
			uptodate = 0;
	}

	if (!uptodate) {
		ClearPageUptodate(page);
		SetPageError(page);
2371 2372
		ret = ret < 0 ? ret : -EIO;
		mapping_set_error(page->mapping, ret);
2373 2374 2375 2376
	}
	return 0;
}

2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387
/*
 * 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)
{
2388
	struct bio_vec *bvec;
2389 2390
	u64 start;
	u64 end;
2391
	int i;
2392

2393
	bio_for_each_segment_all(bvec, bio, i) {
2394
		struct page *page = bvec->bv_page;
2395

2396 2397 2398 2399 2400
		/* We always issue full-page reads, but if some block
		 * in a page fails to read, blk_update_request() will
		 * advance bv_offset and adjust bv_len to compensate.
		 * Print a warning for nonzero offsets, and an error
		 * if they don't add up to a full page.  */
2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411
		if (bvec->bv_offset || bvec->bv_len != PAGE_CACHE_SIZE) {
			if (bvec->bv_offset + bvec->bv_len != PAGE_CACHE_SIZE)
				btrfs_err(BTRFS_I(page->mapping->host)->root->fs_info,
				   "partial page write in btrfs with offset %u and length %u",
					bvec->bv_offset, bvec->bv_len);
			else
				btrfs_info(BTRFS_I(page->mapping->host)->root->fs_info,
				   "incomplete page write in btrfs with offset %u and "
				   "length %u",
					bvec->bv_offset, bvec->bv_len);
		}
2412

2413 2414
		start = page_offset(page);
		end = start + bvec->bv_offset + bvec->bv_len - 1;
2415

2416 2417
		if (end_extent_writepage(page, err, start, end))
			continue;
2418

2419
		end_page_writeback(page);
2420
	}
2421

2422 2423 2424
	bio_put(bio);
}

2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436
static void
endio_readpage_release_extent(struct extent_io_tree *tree, u64 start, u64 len,
			      int uptodate)
{
	struct extent_state *cached = NULL;
	u64 end = start + len - 1;

	if (uptodate && tree->track_uptodate)
		set_extent_uptodate(tree, start, end, &cached, GFP_ATOMIC);
	unlock_extent_cached(tree, start, end, &cached, GFP_ATOMIC);
}

2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449
/*
 * 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)
{
2450
	struct bio_vec *bvec;
2451
	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
2452
	struct btrfs_io_bio *io_bio = btrfs_io_bio(bio);
2453
	struct extent_io_tree *tree;
2454
	u64 offset = 0;
2455 2456
	u64 start;
	u64 end;
2457
	u64 len;
2458 2459
	u64 extent_start = 0;
	u64 extent_len = 0;
2460
	int mirror;
2461
	int ret;
2462
	int i;
2463

2464 2465 2466
	if (err)
		uptodate = 0;

2467
	bio_for_each_segment_all(bvec, bio, i) {
2468
		struct page *page = bvec->bv_page;
2469
		struct inode *inode = page->mapping->host;
2470

2471
		pr_debug("end_bio_extent_readpage: bi_sector=%llu, err=%d, "
2472
			 "mirror=%lu\n", (u64)bio->bi_iter.bi_sector, err,
2473
			 io_bio->mirror_num);
2474
		tree = &BTRFS_I(inode)->io_tree;
2475

2476 2477 2478 2479 2480
		/* We always issue full-page reads, but if some block
		 * in a page fails to read, blk_update_request() will
		 * advance bv_offset and adjust bv_len to compensate.
		 * Print a warning for nonzero offsets, and an error
		 * if they don't add up to a full page.  */
2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491
		if (bvec->bv_offset || bvec->bv_len != PAGE_CACHE_SIZE) {
			if (bvec->bv_offset + bvec->bv_len != PAGE_CACHE_SIZE)
				btrfs_err(BTRFS_I(page->mapping->host)->root->fs_info,
				   "partial page read in btrfs with offset %u and length %u",
					bvec->bv_offset, bvec->bv_len);
			else
				btrfs_info(BTRFS_I(page->mapping->host)->root->fs_info,
				   "incomplete page read in btrfs with offset %u and "
				   "length %u",
					bvec->bv_offset, bvec->bv_len);
		}
2492

2493 2494
		start = page_offset(page);
		end = start + bvec->bv_offset + bvec->bv_len - 1;
2495
		len = bvec->bv_len;
2496

2497
		mirror = io_bio->mirror_num;
2498 2499
		if (likely(uptodate && tree->ops &&
			   tree->ops->readpage_end_io_hook)) {
2500 2501 2502
			ret = tree->ops->readpage_end_io_hook(io_bio, offset,
							      page, start, end,
							      mirror);
2503
			if (ret)
2504
				uptodate = 0;
2505
			else
2506
				clean_io_failure(start, page);
2507
		}
2508

2509 2510 2511 2512
		if (likely(uptodate))
			goto readpage_ok;

		if (tree->ops && tree->ops->readpage_io_failed_hook) {
2513
			ret = tree->ops->readpage_io_failed_hook(page, mirror);
2514 2515 2516
			if (!ret && !err &&
			    test_bit(BIO_UPTODATE, &bio->bi_flags))
				uptodate = 1;
2517
		} else {
2518 2519 2520 2521 2522 2523 2524 2525 2526 2527
			/*
			 * The generic bio_readpage_error handles errors the
			 * following way: If possible, new read requests are
			 * created and submitted and will end up in
			 * end_bio_extent_readpage as well (if we're lucky, not
			 * in the !uptodate case). In that case it returns 0 and
			 * we just go on with the next page in our bio. If it
			 * can't handle the error it will return -EIO and we
			 * remain responsible for that page.
			 */
2528 2529
			ret = bio_readpage_error(bio, offset, page, start, end,
						 mirror);
2530
			if (ret == 0) {
2531 2532
				uptodate =
					test_bit(BIO_UPTODATE, &bio->bi_flags);
2533 2534
				if (err)
					uptodate = 0;
2535
				offset += len;
2536 2537 2538
				continue;
			}
		}
2539
readpage_ok:
2540
		if (likely(uptodate)) {
2541 2542 2543 2544 2545 2546 2547 2548
			loff_t i_size = i_size_read(inode);
			pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT;
			unsigned offset;

			/* Zero out the end if this page straddles i_size */
			offset = i_size & (PAGE_CACHE_SIZE-1);
			if (page->index == end_index && offset)
				zero_user_segment(page, offset, PAGE_CACHE_SIZE);
2549
			SetPageUptodate(page);
2550
		} else {
2551 2552
			ClearPageUptodate(page);
			SetPageError(page);
2553
		}
2554
		unlock_page(page);
2555
		offset += len;
2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577

		if (unlikely(!uptodate)) {
			if (extent_len) {
				endio_readpage_release_extent(tree,
							      extent_start,
							      extent_len, 1);
				extent_start = 0;
				extent_len = 0;
			}
			endio_readpage_release_extent(tree, start,
						      end - start + 1, 0);
		} else if (!extent_len) {
			extent_start = start;
			extent_len = end + 1 - start;
		} else if (extent_start + extent_len == start) {
			extent_len += end + 1 - start;
		} else {
			endio_readpage_release_extent(tree, extent_start,
						      extent_len, uptodate);
			extent_start = start;
			extent_len = end + 1 - start;
		}
2578
	}
2579

2580 2581 2582
	if (extent_len)
		endio_readpage_release_extent(tree, extent_start, extent_len,
					      uptodate);
2583 2584
	if (io_bio->end_io)
		io_bio->end_io(io_bio, err);
2585 2586 2587
	bio_put(bio);
}

2588 2589 2590 2591
/*
 * this allocates from the btrfs_bioset.  We're returning a bio right now
 * but you can call btrfs_io_bio for the appropriate container_of magic
 */
2592 2593 2594
struct bio *
btrfs_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
		gfp_t gfp_flags)
2595
{
2596
	struct btrfs_io_bio *btrfs_bio;
2597 2598
	struct bio *bio;

2599
	bio = bio_alloc_bioset(gfp_flags, nr_vecs, btrfs_bioset);
2600 2601

	if (bio == NULL && (current->flags & PF_MEMALLOC)) {
2602 2603 2604 2605
		while (!bio && (nr_vecs /= 2)) {
			bio = bio_alloc_bioset(gfp_flags,
					       nr_vecs, btrfs_bioset);
		}
2606 2607 2608 2609
	}

	if (bio) {
		bio->bi_bdev = bdev;
2610
		bio->bi_iter.bi_sector = first_sector;
2611 2612 2613 2614
		btrfs_bio = btrfs_io_bio(bio);
		btrfs_bio->csum = NULL;
		btrfs_bio->csum_allocated = NULL;
		btrfs_bio->end_io = NULL;
2615 2616 2617 2618
	}
	return bio;
}

2619 2620 2621 2622 2623 2624 2625 2626 2627
struct bio *btrfs_bio_clone(struct bio *bio, gfp_t gfp_mask)
{
	return bio_clone_bioset(bio, gfp_mask, btrfs_bioset);
}


/* this also allocates from the btrfs_bioset */
struct bio *btrfs_io_bio_alloc(gfp_t gfp_mask, unsigned int nr_iovecs)
{
2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638
	struct btrfs_io_bio *btrfs_bio;
	struct bio *bio;

	bio = bio_alloc_bioset(gfp_mask, nr_iovecs, btrfs_bioset);
	if (bio) {
		btrfs_bio = btrfs_io_bio(bio);
		btrfs_bio->csum = NULL;
		btrfs_bio->csum_allocated = NULL;
		btrfs_bio->end_io = NULL;
	}
	return bio;
2639 2640 2641
}


2642 2643
static int __must_check submit_one_bio(int rw, struct bio *bio,
				       int mirror_num, unsigned long bio_flags)
2644 2645
{
	int ret = 0;
2646 2647 2648 2649 2650
	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
	struct page *page = bvec->bv_page;
	struct extent_io_tree *tree = bio->bi_private;
	u64 start;

M
Miao Xie 已提交
2651
	start = page_offset(page) + bvec->bv_offset;
2652

2653
	bio->bi_private = NULL;
2654 2655 2656

	bio_get(bio);

2657
	if (tree->ops && tree->ops->submit_bio_hook)
2658
		ret = tree->ops->submit_bio_hook(page->mapping->host, rw, bio,
2659
					   mirror_num, bio_flags, start);
2660
	else
2661
		btrfsic_submit_bio(rw, bio);
2662

2663 2664 2665 2666 2667 2668
	if (bio_flagged(bio, BIO_EOPNOTSUPP))
		ret = -EOPNOTSUPP;
	bio_put(bio);
	return ret;
}

2669
static int merge_bio(int rw, struct extent_io_tree *tree, struct page *page,
2670 2671 2672 2673 2674
		     unsigned long offset, size_t size, struct bio *bio,
		     unsigned long bio_flags)
{
	int ret = 0;
	if (tree->ops && tree->ops->merge_bio_hook)
2675
		ret = tree->ops->merge_bio_hook(rw, page, offset, size, bio,
2676 2677 2678 2679 2680 2681
						bio_flags);
	BUG_ON(ret < 0);
	return ret;

}

2682 2683 2684 2685 2686 2687
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,
2688
			      bio_end_io_t end_io_func,
C
Chris Mason 已提交
2689 2690 2691
			      int mirror_num,
			      unsigned long prev_bio_flags,
			      unsigned long bio_flags)
2692 2693 2694 2695
{
	int ret = 0;
	struct bio *bio;
	int nr;
C
Chris Mason 已提交
2696 2697 2698
	int contig = 0;
	int this_compressed = bio_flags & EXTENT_BIO_COMPRESSED;
	int old_compressed = prev_bio_flags & EXTENT_BIO_COMPRESSED;
2699
	size_t page_size = min_t(size_t, size, PAGE_CACHE_SIZE);
2700 2701 2702

	if (bio_ret && *bio_ret) {
		bio = *bio_ret;
C
Chris Mason 已提交
2703
		if (old_compressed)
2704
			contig = bio->bi_iter.bi_sector == sector;
C
Chris Mason 已提交
2705
		else
K
Kent Overstreet 已提交
2706
			contig = bio_end_sector(bio) == sector;
C
Chris Mason 已提交
2707 2708

		if (prev_bio_flags != bio_flags || !contig ||
2709
		    merge_bio(rw, tree, page, offset, page_size, bio, bio_flags) ||
C
Chris Mason 已提交
2710 2711 2712
		    bio_add_page(bio, page, page_size, offset) < page_size) {
			ret = submit_one_bio(rw, bio, mirror_num,
					     prev_bio_flags);
2713 2714
			if (ret < 0)
				return ret;
2715 2716 2717 2718 2719
			bio = NULL;
		} else {
			return 0;
		}
	}
C
Chris Mason 已提交
2720 2721 2722 2723 2724
	if (this_compressed)
		nr = BIO_MAX_PAGES;
	else
		nr = bio_get_nr_vecs(bdev);

2725
	bio = btrfs_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH);
2726 2727
	if (!bio)
		return -ENOMEM;
2728

C
Chris Mason 已提交
2729
	bio_add_page(bio, page, page_size, offset);
2730 2731
	bio->bi_end_io = end_io_func;
	bio->bi_private = tree;
2732

C
Chris Mason 已提交
2733
	if (bio_ret)
2734
		*bio_ret = bio;
C
Chris Mason 已提交
2735
	else
C
Chris Mason 已提交
2736
		ret = submit_one_bio(rw, bio, mirror_num, bio_flags);
2737 2738 2739 2740

	return ret;
}

2741 2742
static void attach_extent_buffer_page(struct extent_buffer *eb,
				      struct page *page)
2743 2744 2745 2746
{
	if (!PagePrivate(page)) {
		SetPagePrivate(page);
		page_cache_get(page);
J
Josef Bacik 已提交
2747 2748 2749
		set_page_private(page, (unsigned long)eb);
	} else {
		WARN_ON(page->private != (unsigned long)eb);
2750 2751 2752
	}
}

J
Josef Bacik 已提交
2753
void set_page_extent_mapped(struct page *page)
2754
{
J
Josef Bacik 已提交
2755 2756 2757 2758 2759
	if (!PagePrivate(page)) {
		SetPagePrivate(page);
		page_cache_get(page);
		set_page_private(page, EXTENT_PAGE_PRIVATE);
	}
2760 2761
}

2762 2763 2764 2765 2766 2767 2768 2769 2770
static struct extent_map *
__get_extent_map(struct inode *inode, struct page *page, size_t pg_offset,
		 u64 start, u64 len, get_extent_t *get_extent,
		 struct extent_map **em_cached)
{
	struct extent_map *em;

	if (em_cached && *em_cached) {
		em = *em_cached;
2771
		if (extent_map_in_tree(em) && start >= em->start &&
2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788
		    start < extent_map_end(em)) {
			atomic_inc(&em->refs);
			return em;
		}

		free_extent_map(em);
		*em_cached = NULL;
	}

	em = get_extent(inode, page, pg_offset, start, len, 0);
	if (em_cached && !IS_ERR_OR_NULL(em)) {
		BUG_ON(*em_cached);
		atomic_inc(&em->refs);
		*em_cached = em;
	}
	return em;
}
2789 2790 2791 2792
/*
 * 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)
2793
 * XXX JDM: This needs looking at to ensure proper page locking
2794
 */
2795 2796 2797
static int __do_readpage(struct extent_io_tree *tree,
			 struct page *page,
			 get_extent_t *get_extent,
2798
			 struct extent_map **em_cached,
2799 2800
			 struct bio **bio, int mirror_num,
			 unsigned long *bio_flags, int rw)
2801 2802
{
	struct inode *inode = page->mapping->host;
M
Miao Xie 已提交
2803
	u64 start = page_offset(page);
2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815
	u64 page_end = start + PAGE_CACHE_SIZE - 1;
	u64 end;
	u64 cur = start;
	u64 extent_offset;
	u64 last_byte = i_size_read(inode);
	u64 block_start;
	u64 cur_end;
	sector_t sector;
	struct extent_map *em;
	struct block_device *bdev;
	int ret;
	int nr = 0;
2816
	int parent_locked = *bio_flags & EXTENT_BIO_PARENT_LOCKED;
2817
	size_t pg_offset = 0;
2818
	size_t iosize;
C
Chris Mason 已提交
2819
	size_t disk_io_size;
2820
	size_t blocksize = inode->i_sb->s_blocksize;
2821
	unsigned long this_bio_flag = *bio_flags & EXTENT_BIO_PARENT_LOCKED;
2822 2823 2824

	set_page_extent_mapped(page);

2825
	end = page_end;
D
Dan Magenheimer 已提交
2826 2827 2828
	if (!PageUptodate(page)) {
		if (cleancache_get_page(page) == 0) {
			BUG_ON(blocksize != PAGE_SIZE);
2829
			unlock_extent(tree, start, end);
D
Dan Magenheimer 已提交
2830 2831 2832 2833
			goto out;
		}
	}

C
Chris Mason 已提交
2834 2835 2836 2837 2838 2839
	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;
2840
			userpage = kmap_atomic(page);
C
Chris Mason 已提交
2841 2842
			memset(userpage + zero_offset, 0, iosize);
			flush_dcache_page(page);
2843
			kunmap_atomic(userpage);
C
Chris Mason 已提交
2844 2845
		}
	}
2846
	while (cur <= end) {
2847 2848
		unsigned long pnr = (last_byte >> PAGE_CACHE_SHIFT) + 1;

2849 2850
		if (cur >= last_byte) {
			char *userpage;
2851 2852
			struct extent_state *cached = NULL;

2853
			iosize = PAGE_CACHE_SIZE - pg_offset;
2854
			userpage = kmap_atomic(page);
2855
			memset(userpage + pg_offset, 0, iosize);
2856
			flush_dcache_page(page);
2857
			kunmap_atomic(userpage);
2858
			set_extent_uptodate(tree, cur, cur + iosize - 1,
2859
					    &cached, GFP_NOFS);
2860 2861 2862 2863
			if (!parent_locked)
				unlock_extent_cached(tree, cur,
						     cur + iosize - 1,
						     &cached, GFP_NOFS);
2864 2865
			break;
		}
2866 2867
		em = __get_extent_map(inode, page, pg_offset, cur,
				      end - cur + 1, get_extent, em_cached);
2868
		if (IS_ERR_OR_NULL(em)) {
2869
			SetPageError(page);
2870 2871
			if (!parent_locked)
				unlock_extent(tree, cur, end);
2872 2873 2874 2875 2876 2877
			break;
		}
		extent_offset = cur - em->start;
		BUG_ON(extent_map_end(em) <= cur);
		BUG_ON(end < cur);

2878
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
2879
			this_bio_flag |= EXTENT_BIO_COMPRESSED;
2880 2881 2882
			extent_set_compress_type(&this_bio_flag,
						 em->compress_type);
		}
C
Chris Mason 已提交
2883

2884 2885
		iosize = min(extent_map_end(em) - cur, end - cur + 1);
		cur_end = min(extent_map_end(em) - 1, end);
2886
		iosize = ALIGN(iosize, blocksize);
C
Chris Mason 已提交
2887 2888 2889 2890 2891 2892 2893
		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;
		}
2894 2895
		bdev = em->bdev;
		block_start = em->block_start;
Y
Yan Zheng 已提交
2896 2897
		if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			block_start = EXTENT_MAP_HOLE;
2898 2899 2900 2901 2902 2903
		free_extent_map(em);
		em = NULL;

		/* we've found a hole, just zero and go on */
		if (block_start == EXTENT_MAP_HOLE) {
			char *userpage;
2904 2905
			struct extent_state *cached = NULL;

2906
			userpage = kmap_atomic(page);
2907
			memset(userpage + pg_offset, 0, iosize);
2908
			flush_dcache_page(page);
2909
			kunmap_atomic(userpage);
2910 2911

			set_extent_uptodate(tree, cur, cur + iosize - 1,
2912 2913 2914
					    &cached, GFP_NOFS);
			unlock_extent_cached(tree, cur, cur + iosize - 1,
			                     &cached, GFP_NOFS);
2915
			cur = cur + iosize;
2916
			pg_offset += iosize;
2917 2918 2919
			continue;
		}
		/* the get_extent function already copied into the page */
2920 2921
		if (test_range_bit(tree, cur, cur_end,
				   EXTENT_UPTODATE, 1, NULL)) {
2922
			check_page_uptodate(tree, page);
2923 2924
			if (!parent_locked)
				unlock_extent(tree, cur, cur + iosize - 1);
2925
			cur = cur + iosize;
2926
			pg_offset += iosize;
2927 2928
			continue;
		}
2929 2930 2931 2932 2933
		/* we have an inline extent but it didn't get marked up
		 * to date.  Error out
		 */
		if (block_start == EXTENT_MAP_INLINE) {
			SetPageError(page);
2934 2935
			if (!parent_locked)
				unlock_extent(tree, cur, cur + iosize - 1);
2936
			cur = cur + iosize;
2937
			pg_offset += iosize;
2938 2939
			continue;
		}
2940

2941
		pnr -= page->index;
2942
		ret = submit_extent_page(rw, tree, page,
2943
					 sector, disk_io_size, pg_offset,
2944
					 bdev, bio, pnr,
C
Chris Mason 已提交
2945 2946 2947
					 end_bio_extent_readpage, mirror_num,
					 *bio_flags,
					 this_bio_flag);
2948 2949 2950 2951
		if (!ret) {
			nr++;
			*bio_flags = this_bio_flag;
		} else {
2952
			SetPageError(page);
2953 2954
			if (!parent_locked)
				unlock_extent(tree, cur, cur + iosize - 1);
2955
		}
2956
		cur = cur + iosize;
2957
		pg_offset += iosize;
2958
	}
D
Dan Magenheimer 已提交
2959
out:
2960 2961 2962 2963 2964 2965 2966 2967
	if (!nr) {
		if (!PageError(page))
			SetPageUptodate(page);
		unlock_page(page);
	}
	return 0;
}

2968 2969 2970 2971
static inline void __do_contiguous_readpages(struct extent_io_tree *tree,
					     struct page *pages[], int nr_pages,
					     u64 start, u64 end,
					     get_extent_t *get_extent,
2972
					     struct extent_map **em_cached,
2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992
					     struct bio **bio, int mirror_num,
					     unsigned long *bio_flags, int rw)
{
	struct inode *inode;
	struct btrfs_ordered_extent *ordered;
	int index;

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

	for (index = 0; index < nr_pages; index++) {
2993 2994
		__do_readpage(tree, pages[index], get_extent, em_cached, bio,
			      mirror_num, bio_flags, rw);
2995 2996 2997 2998 2999 3000 3001
		page_cache_release(pages[index]);
	}
}

static void __extent_readpages(struct extent_io_tree *tree,
			       struct page *pages[],
			       int nr_pages, get_extent_t *get_extent,
3002
			       struct extent_map **em_cached,
3003 3004 3005
			       struct bio **bio, int mirror_num,
			       unsigned long *bio_flags, int rw)
{
3006
	u64 start = 0;
3007 3008 3009
	u64 end = 0;
	u64 page_start;
	int index;
3010
	int first_index = 0;
3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022

	for (index = 0; index < nr_pages; index++) {
		page_start = page_offset(pages[index]);
		if (!end) {
			start = page_start;
			end = start + PAGE_CACHE_SIZE - 1;
			first_index = index;
		} else if (end + 1 == page_start) {
			end += PAGE_CACHE_SIZE;
		} else {
			__do_contiguous_readpages(tree, &pages[first_index],
						  index - first_index, start,
3023 3024 3025
						  end, get_extent, em_cached,
						  bio, mirror_num, bio_flags,
						  rw);
3026 3027 3028 3029 3030 3031 3032 3033 3034
			start = page_start;
			end = start + PAGE_CACHE_SIZE - 1;
			first_index = index;
		}
	}

	if (end)
		__do_contiguous_readpages(tree, &pages[first_index],
					  index - first_index, start,
3035
					  end, get_extent, em_cached, bio,
3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060
					  mirror_num, bio_flags, rw);
}

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

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

3061 3062
	ret = __do_readpage(tree, page, get_extent, NULL, bio, mirror_num,
			    bio_flags, rw);
3063 3064 3065
	return ret;
}

3066
int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
3067
			    get_extent_t *get_extent, int mirror_num)
3068 3069
{
	struct bio *bio = NULL;
C
Chris Mason 已提交
3070
	unsigned long bio_flags = 0;
3071 3072
	int ret;

3073
	ret = __extent_read_full_page(tree, page, get_extent, &bio, mirror_num,
3074
				      &bio_flags, READ);
3075
	if (bio)
3076
		ret = submit_one_bio(READ, bio, mirror_num, bio_flags);
3077 3078 3079
	return ret;
}

3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093
int extent_read_full_page_nolock(struct extent_io_tree *tree, struct page *page,
				 get_extent_t *get_extent, int mirror_num)
{
	struct bio *bio = NULL;
	unsigned long bio_flags = EXTENT_BIO_PARENT_LOCKED;
	int ret;

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

3094 3095 3096 3097 3098 3099 3100 3101 3102 3103
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;
}

3104
/*
3105 3106 3107 3108 3109 3110 3111 3112
 * helper for __extent_writepage, doing all of the delayed allocation setup.
 *
 * This returns 1 if our fill_delalloc function did all the work required
 * to write the page (copy into inline extent).  In this case the IO has
 * been started and the page is already unlocked.
 *
 * This returns 0 if all went well (page still locked)
 * This returns < 0 if there were errors (page still locked)
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 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209
static noinline_for_stack int writepage_delalloc(struct inode *inode,
			      struct page *page, struct writeback_control *wbc,
			      struct extent_page_data *epd,
			      u64 delalloc_start,
			      unsigned long *nr_written)
{
	struct extent_io_tree *tree = epd->tree;
	u64 page_end = delalloc_start + PAGE_CACHE_SIZE - 1;
	u64 nr_delalloc;
	u64 delalloc_to_write = 0;
	u64 delalloc_end = 0;
	int ret;
	int page_started = 0;

	if (epd->extent_locked || !tree->ops || !tree->ops->fill_delalloc)
		return 0;

	while (delalloc_end < page_end) {
		nr_delalloc = find_lock_delalloc_range(inode, tree,
					       page,
					       &delalloc_start,
					       &delalloc_end,
					       128 * 1024 * 1024);
		if (nr_delalloc == 0) {
			delalloc_start = delalloc_end + 1;
			continue;
		}
		ret = tree->ops->fill_delalloc(inode, page,
					       delalloc_start,
					       delalloc_end,
					       &page_started,
					       nr_written);
		/* File system has been set read-only */
		if (ret) {
			SetPageError(page);
			/* fill_delalloc should be return < 0 for error
			 * but just in case, we use > 0 here meaning the
			 * IO is started, so we don't want to return > 0
			 * unless things are going well.
			 */
			ret = ret < 0 ? ret : -EIO;
			goto done;
		}
		/*
		 * 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;
		delalloc_start = delalloc_end + 1;
	}
	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);
	}

	/* did the fill delalloc function already unlock and start
	 * the IO?
	 */
	if (page_started) {
		/*
		 * 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;
		return 1;
	}

	ret = 0;

done:
	return ret;
}

/*
 * helper for __extent_writepage.  This calls the writepage start hooks,
 * and does the loop to map the page into extents and bios.
 *
 * We return 1 if the IO is started and the page is unlocked,
 * 0 if all went well (page still locked)
 * < 0 if there were errors (page still locked)
 */
static noinline_for_stack int __extent_writepage_io(struct inode *inode,
				 struct page *page,
				 struct writeback_control *wbc,
				 struct extent_page_data *epd,
				 loff_t i_size,
				 unsigned long nr_written,
				 int write_flags, int *nr_ret)
3210 3211
{
	struct extent_io_tree *tree = epd->tree;
M
Miao Xie 已提交
3212
	u64 start = page_offset(page);
3213 3214 3215 3216 3217 3218 3219
	u64 page_end = start + PAGE_CACHE_SIZE - 1;
	u64 end;
	u64 cur = start;
	u64 extent_offset;
	u64 block_start;
	u64 iosize;
	sector_t sector;
3220
	struct extent_state *cached_state = NULL;
3221 3222
	struct extent_map *em;
	struct block_device *bdev;
3223
	size_t pg_offset = 0;
3224
	size_t blocksize;
3225 3226 3227
	int ret = 0;
	int nr = 0;
	bool compressed;
C
Chris Mason 已提交
3228

3229
	if (tree->ops && tree->ops->writepage_start_hook) {
C
Chris Mason 已提交
3230 3231
		ret = tree->ops->writepage_start_hook(page, start,
						      page_end);
3232 3233 3234 3235 3236 3237
		if (ret) {
			/* Fixup worker will requeue */
			if (ret == -EBUSY)
				wbc->pages_skipped++;
			else
				redirty_page_for_writepage(wbc, page);
3238

3239
			update_nr_written(page, wbc, nr_written);
3240
			unlock_page(page);
3241
			ret = 1;
3242
			goto done_unlocked;
3243 3244 3245
		}
	}

3246 3247 3248 3249 3250
	/*
	 * 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);
3251

3252
	end = page_end;
3253
	if (i_size <= start) {
3254 3255 3256
		if (tree->ops && tree->ops->writepage_end_io_hook)
			tree->ops->writepage_end_io_hook(page, start,
							 page_end, NULL, 1);
3257 3258 3259 3260 3261 3262
		goto done;
	}

	blocksize = inode->i_sb->s_blocksize;

	while (cur <= end) {
3263 3264
		u64 em_end;
		if (cur >= i_size) {
3265 3266 3267
			if (tree->ops && tree->ops->writepage_end_io_hook)
				tree->ops->writepage_end_io_hook(page, cur,
							 page_end, NULL, 1);
3268 3269
			break;
		}
3270
		em = epd->get_extent(inode, page, pg_offset, cur,
3271
				     end - cur + 1, 1);
3272
		if (IS_ERR_OR_NULL(em)) {
3273
			SetPageError(page);
3274
			ret = PTR_ERR_OR_ZERO(em);
3275 3276 3277 3278
			break;
		}

		extent_offset = cur - em->start;
3279 3280
		em_end = extent_map_end(em);
		BUG_ON(em_end <= cur);
3281
		BUG_ON(end < cur);
3282
		iosize = min(em_end - cur, end - cur + 1);
3283
		iosize = ALIGN(iosize, blocksize);
3284 3285 3286
		sector = (em->block_start + extent_offset) >> 9;
		bdev = em->bdev;
		block_start = em->block_start;
C
Chris Mason 已提交
3287
		compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
3288 3289 3290
		free_extent_map(em);
		em = NULL;

C
Chris Mason 已提交
3291 3292 3293 3294 3295
		/*
		 * compressed and inline extents are written through other
		 * paths in the FS
		 */
		if (compressed || block_start == EXTENT_MAP_HOLE ||
3296
		    block_start == EXTENT_MAP_INLINE) {
C
Chris Mason 已提交
3297 3298 3299 3300 3301 3302
			/*
			 * end_io notification does not happen here for
			 * compressed extents
			 */
			if (!compressed && tree->ops &&
			    tree->ops->writepage_end_io_hook)
3303 3304 3305
				tree->ops->writepage_end_io_hook(page, cur,
							 cur + iosize - 1,
							 NULL, 1);
C
Chris Mason 已提交
3306 3307 3308 3309 3310 3311 3312 3313 3314
			else if (compressed) {
				/* we don't want to end_page_writeback on
				 * a compressed extent.  this happens
				 * elsewhere
				 */
				nr++;
			}

			cur += iosize;
3315
			pg_offset += iosize;
3316 3317
			continue;
		}
C
Chris Mason 已提交
3318

3319 3320 3321 3322 3323 3324
		if (tree->ops && tree->ops->writepage_io_hook) {
			ret = tree->ops->writepage_io_hook(page, cur,
						cur + iosize - 1);
		} else {
			ret = 0;
		}
3325
		if (ret) {
3326
			SetPageError(page);
3327
		} else {
3328
			unsigned long max_nr = (i_size >> PAGE_CACHE_SHIFT) + 1;
3329

3330 3331
			set_range_writeback(tree, cur, cur + iosize - 1);
			if (!PageWriteback(page)) {
3332 3333
				btrfs_err(BTRFS_I(inode)->root->fs_info,
					   "page %lu not writeback, cur %llu end %llu",
3334
				       page->index, cur, end);
3335 3336
			}

3337 3338 3339
			ret = submit_extent_page(write_flags, tree, page,
						 sector, iosize, pg_offset,
						 bdev, &epd->bio, max_nr,
C
Chris Mason 已提交
3340 3341
						 end_bio_extent_writepage,
						 0, 0, 0);
3342 3343 3344 3345
			if (ret)
				SetPageError(page);
		}
		cur = cur + iosize;
3346
		pg_offset += iosize;
3347 3348
		nr++;
	}
3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423
done:
	*nr_ret = nr;

done_unlocked:

	/* drop our reference on any cached states */
	free_extent_state(cached_state);
	return ret;
}

/*
 * 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;
	u64 start = page_offset(page);
	u64 page_end = start + PAGE_CACHE_SIZE - 1;
	int ret;
	int nr = 0;
	size_t pg_offset = 0;
	loff_t i_size = i_size_read(inode);
	unsigned long end_index = i_size >> PAGE_CACHE_SHIFT;
	int write_flags;
	unsigned long nr_written = 0;

	if (wbc->sync_mode == WB_SYNC_ALL)
		write_flags = WRITE_SYNC;
	else
		write_flags = WRITE;

	trace___extent_writepage(page, inode, wbc);

	WARN_ON(!PageLocked(page));

	ClearPageError(page);

	pg_offset = i_size & (PAGE_CACHE_SIZE - 1);
	if (page->index > end_index ||
	   (page->index == end_index && !pg_offset)) {
		page->mapping->a_ops->invalidatepage(page, 0, PAGE_CACHE_SIZE);
		unlock_page(page);
		return 0;
	}

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

		userpage = kmap_atomic(page);
		memset(userpage + pg_offset, 0,
		       PAGE_CACHE_SIZE - pg_offset);
		kunmap_atomic(userpage);
		flush_dcache_page(page);
	}

	pg_offset = 0;

	set_page_extent_mapped(page);

	ret = writepage_delalloc(inode, page, wbc, epd, start, &nr_written);
	if (ret == 1)
		goto done_unlocked;
	if (ret)
		goto done;

	ret = __extent_writepage_io(inode, page, wbc, epd,
				    i_size, nr_written, write_flags, &nr);
	if (ret == 1)
		goto done_unlocked;

3424 3425 3426 3427 3428 3429
done:
	if (nr == 0) {
		/* make sure the mapping tag for page dirty gets cleared */
		set_page_writeback(page);
		end_page_writeback(page);
	}
3430 3431 3432 3433
	if (PageError(page)) {
		ret = ret < 0 ? ret : -EIO;
		end_extent_writepage(page, ret, start, page_end);
	}
3434
	unlock_page(page);
3435
	return ret;
3436

3437
done_unlocked:
3438 3439 3440
	return 0;
}

3441
void wait_on_extent_buffer_writeback(struct extent_buffer *eb)
3442
{
3443 3444
	wait_on_bit_io(&eb->bflags, EXTENT_BUFFER_WRITEBACK,
		       TASK_UNINTERRUPTIBLE);
3445 3446
}

3447 3448 3449 3450
static noinline_for_stack int
lock_extent_buffer_for_io(struct extent_buffer *eb,
			  struct btrfs_fs_info *fs_info,
			  struct extent_page_data *epd)
3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469
{
	unsigned long i, num_pages;
	int flush = 0;
	int ret = 0;

	if (!btrfs_try_tree_write_lock(eb)) {
		flush = 1;
		flush_write_bio(epd);
		btrfs_tree_lock(eb);
	}

	if (test_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags)) {
		btrfs_tree_unlock(eb);
		if (!epd->sync_io)
			return 0;
		if (!flush) {
			flush_write_bio(epd);
			flush = 1;
		}
C
Chris Mason 已提交
3470 3471 3472 3473 3474
		while (1) {
			wait_on_extent_buffer_writeback(eb);
			btrfs_tree_lock(eb);
			if (!test_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags))
				break;
3475 3476 3477 3478
			btrfs_tree_unlock(eb);
		}
	}

3479 3480 3481 3482 3483 3484
	/*
	 * We need to do this to prevent races in people who check if the eb is
	 * under IO since we can end up having no IO bits set for a short period
	 * of time.
	 */
	spin_lock(&eb->refs_lock);
3485 3486
	if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
		set_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags);
3487
		spin_unlock(&eb->refs_lock);
3488
		btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
3489 3490 3491
		__percpu_counter_add(&fs_info->dirty_metadata_bytes,
				     -eb->len,
				     fs_info->dirty_metadata_batch);
3492
		ret = 1;
3493 3494
	} else {
		spin_unlock(&eb->refs_lock);
3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520
	}

	btrfs_tree_unlock(eb);

	if (!ret)
		return ret;

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

		if (!trylock_page(p)) {
			if (!flush) {
				flush_write_bio(epd);
				flush = 1;
			}
			lock_page(p);
		}
	}

	return ret;
}

static void end_extent_buffer_writeback(struct extent_buffer *eb)
{
	clear_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags);
3521
	smp_mb__after_atomic();
3522 3523 3524 3525 3526
	wake_up_bit(&eb->bflags, EXTENT_BUFFER_WRITEBACK);
}

static void end_bio_extent_buffer_writepage(struct bio *bio, int err)
{
3527
	struct bio_vec *bvec;
3528
	struct extent_buffer *eb;
3529
	int i, done;
3530

3531
	bio_for_each_segment_all(bvec, bio, i) {
3532 3533 3534 3535 3536 3537
		struct page *page = bvec->bv_page;

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

3538
		if (err || test_bit(EXTENT_BUFFER_IOERR, &eb->bflags)) {
3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549
			set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
			ClearPageUptodate(page);
			SetPageError(page);
		}

		end_page_writeback(page);

		if (!done)
			continue;

		end_extent_buffer_writeback(eb);
3550
	}
3551 3552 3553 3554

	bio_put(bio);
}

3555
static noinline_for_stack int write_one_eb(struct extent_buffer *eb,
3556 3557 3558 3559 3560
			struct btrfs_fs_info *fs_info,
			struct writeback_control *wbc,
			struct extent_page_data *epd)
{
	struct block_device *bdev = fs_info->fs_devices->latest_bdev;
3561
	struct extent_io_tree *tree = &BTRFS_I(fs_info->btree_inode)->io_tree;
3562 3563
	u64 offset = eb->start;
	unsigned long i, num_pages;
3564
	unsigned long bio_flags = 0;
3565
	int rw = (epd->sync_io ? WRITE_SYNC : WRITE) | REQ_META;
3566
	int ret = 0;
3567 3568 3569 3570

	clear_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
	num_pages = num_extent_pages(eb->start, eb->len);
	atomic_set(&eb->io_pages, num_pages);
3571 3572 3573
	if (btrfs_header_owner(eb) == BTRFS_TREE_LOG_OBJECTID)
		bio_flags = EXTENT_BIO_TREE_LOG;

3574 3575 3576 3577 3578
	for (i = 0; i < num_pages; i++) {
		struct page *p = extent_buffer_page(eb, i);

		clear_page_dirty_for_io(p);
		set_page_writeback(p);
3579
		ret = submit_extent_page(rw, tree, p, offset >> 9,
3580 3581
					 PAGE_CACHE_SIZE, 0, bdev, &epd->bio,
					 -1, end_bio_extent_buffer_writepage,
3582 3583
					 0, epd->bio_flags, bio_flags);
		epd->bio_flags = bio_flags;
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
		if (ret) {
			set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
			SetPageError(p);
			if (atomic_sub_and_test(num_pages - i, &eb->io_pages))
				end_extent_buffer_writeback(eb);
			ret = -EIO;
			break;
		}
		offset += PAGE_CACHE_SIZE;
		update_nr_written(p, wbc, 1);
		unlock_page(p);
	}

	if (unlikely(ret)) {
		for (; i < num_pages; i++) {
			struct page *p = extent_buffer_page(eb, i);
			unlock_page(p);
		}
	}

	return ret;
}

int btree_write_cache_pages(struct address_space *mapping,
				   struct writeback_control *wbc)
{
	struct extent_io_tree *tree = &BTRFS_I(mapping->host)->io_tree;
	struct btrfs_fs_info *fs_info = BTRFS_I(mapping->host)->root->fs_info;
	struct extent_buffer *eb, *prev_eb = NULL;
	struct extent_page_data epd = {
		.bio = NULL,
		.tree = tree,
		.extent_locked = 0,
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
3618
		.bio_flags = 0,
3619 3620 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 3655 3656 3657 3658 3659 3660 3661 3662
	};
	int ret = 0;
	int done = 0;
	int nr_to_write_done = 0;
	struct pagevec pvec;
	int nr_pages;
	pgoff_t index;
	pgoff_t end;		/* Inclusive */
	int scanned = 0;
	int tag;

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

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

			if (!PagePrivate(page))
				continue;

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

3663 3664 3665 3666 3667 3668
			spin_lock(&mapping->private_lock);
			if (!PagePrivate(page)) {
				spin_unlock(&mapping->private_lock);
				continue;
			}

3669
			eb = (struct extent_buffer *)page->private;
3670 3671 3672 3673 3674 3675

			/*
			 * Shouldn't happen and normally this would be a BUG_ON
			 * but no sense in crashing the users box for something
			 * we can survive anyway.
			 */
3676
			if (WARN_ON(!eb)) {
3677
				spin_unlock(&mapping->private_lock);
3678 3679 3680
				continue;
			}

3681 3682
			if (eb == prev_eb) {
				spin_unlock(&mapping->private_lock);
3683
				continue;
3684
			}
3685

3686 3687 3688
			ret = atomic_inc_not_zero(&eb->refs);
			spin_unlock(&mapping->private_lock);
			if (!ret)
3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728
				continue;

			prev_eb = eb;
			ret = lock_extent_buffer_for_io(eb, fs_info, &epd);
			if (!ret) {
				free_extent_buffer(eb);
				continue;
			}

			ret = write_one_eb(eb, fs_info, wbc, &epd);
			if (ret) {
				done = 1;
				free_extent_buffer(eb);
				break;
			}
			free_extent_buffer(eb);

			/*
			 * the filesystem may choose to bump up nr_to_write.
			 * We have to make sure to honor the new nr_to_write
			 * at any time
			 */
			nr_to_write_done = wbc->nr_to_write <= 0;
		}
		pagevec_release(&pvec);
		cond_resched();
	}
	if (!scanned && !done) {
		/*
		 * We hit the last page and there is more work to be done: wrap
		 * back to the start of the file
		 */
		scanned = 1;
		index = 0;
		goto retry;
	}
	flush_write_bio(&epd);
	return ret;
}

3729
/**
C
Chris Mason 已提交
3730
 * write_cache_pages - walk the list of dirty pages of the given address space and write all of them.
3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743
 * @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.
 */
3744
static int extent_write_cache_pages(struct extent_io_tree *tree,
C
Chris Mason 已提交
3745 3746
			     struct address_space *mapping,
			     struct writeback_control *wbc,
C
Chris Mason 已提交
3747 3748
			     writepage_t writepage, void *data,
			     void (*flush_fn)(void *))
3749
{
3750
	struct inode *inode = mapping->host;
3751 3752
	int ret = 0;
	int done = 0;
3753
	int err = 0;
3754
	int nr_to_write_done = 0;
3755 3756 3757 3758 3759
	struct pagevec pvec;
	int nr_pages;
	pgoff_t index;
	pgoff_t end;		/* Inclusive */
	int scanned = 0;
3760
	int tag;
3761

3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773
	/*
	 * We have to hold onto the inode so that ordered extents can do their
	 * work when the IO finishes.  The alternative to this is failing to add
	 * an ordered extent if the igrab() fails there and that is a huge pain
	 * to deal with, so instead just hold onto the inode throughout the
	 * writepages operation.  If it fails here we are freeing up the inode
	 * anyway and we'd rather not waste our time writing out stuff that is
	 * going to be truncated anyway.
	 */
	if (!igrab(inode))
		return 0;

3774 3775 3776 3777 3778 3779 3780 3781 3782
	pagevec_init(&pvec, 0);
	if (wbc->range_cyclic) {
		index = mapping->writeback_index; /* Start from prev offset */
		end = -1;
	} else {
		index = wbc->range_start >> PAGE_CACHE_SHIFT;
		end = wbc->range_end >> PAGE_CACHE_SHIFT;
		scanned = 1;
	}
3783 3784 3785 3786
	if (wbc->sync_mode == WB_SYNC_ALL)
		tag = PAGECACHE_TAG_TOWRITE;
	else
		tag = PAGECACHE_TAG_DIRTY;
3787
retry:
3788 3789
	if (wbc->sync_mode == WB_SYNC_ALL)
		tag_pages_for_writeback(mapping, index, end);
3790
	while (!done && !nr_to_write_done && (index <= end) &&
3791 3792
	       (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag,
			min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805
		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
			 */
3806 3807 3808
			if (!trylock_page(page)) {
				flush_fn(data);
				lock_page(page);
3809
			}
3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821

			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
3822
			if (wbc->sync_mode != WB_SYNC_NONE) {
3823 3824
				if (PageWriteback(page))
					flush_fn(data);
3825
				wait_on_page_writeback(page);
C
Chris Mason 已提交
3826
			}
3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838 3839

			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;
			}
3840 3841
			if (!err && ret < 0)
				err = ret;
3842 3843 3844 3845 3846 3847 3848

			/*
			 * 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;
3849 3850 3851 3852
		}
		pagevec_release(&pvec);
		cond_resched();
	}
3853
	if (!scanned && !done && !err) {
3854 3855 3856 3857 3858 3859 3860 3861
		/*
		 * 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;
	}
3862
	btrfs_add_delayed_iput(inode);
3863
	return err;
3864 3865
}

3866
static void flush_epd_write_bio(struct extent_page_data *epd)
C
Chris Mason 已提交
3867 3868
{
	if (epd->bio) {
3869 3870 3871
		int rw = WRITE;
		int ret;

3872
		if (epd->sync_io)
3873 3874
			rw = WRITE_SYNC;

3875
		ret = submit_one_bio(rw, epd->bio, 0, epd->bio_flags);
3876
		BUG_ON(ret < 0); /* -ENOMEM */
C
Chris Mason 已提交
3877 3878 3879 3880
		epd->bio = NULL;
	}
}

3881 3882 3883 3884 3885 3886
static noinline void flush_write_bio(void *data)
{
	struct extent_page_data *epd = data;
	flush_epd_write_bio(epd);
}

3887 3888 3889 3890 3891 3892 3893 3894 3895
int extent_write_full_page(struct extent_io_tree *tree, struct page *page,
			  get_extent_t *get_extent,
			  struct writeback_control *wbc)
{
	int ret;
	struct extent_page_data epd = {
		.bio = NULL,
		.tree = tree,
		.get_extent = get_extent,
3896
		.extent_locked = 0,
3897
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
3898
		.bio_flags = 0,
3899 3900 3901 3902
	};

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

3903
	flush_epd_write_bio(&epd);
3904 3905 3906
	return ret;
}

3907 3908 3909 3910 3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921
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,
3922
		.sync_io = mode == WB_SYNC_ALL,
3923
		.bio_flags = 0,
3924 3925 3926 3927 3928 3929 3930 3931
	};
	struct writeback_control wbc_writepages = {
		.sync_mode	= mode,
		.nr_to_write	= nr_pages * 2,
		.range_start	= start,
		.range_end	= end + 1,
	};

C
Chris Mason 已提交
3932
	while (start <= end) {
3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946
		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;
	}

3947
	flush_epd_write_bio(&epd);
3948 3949
	return ret;
}
3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960

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,
3961
		.extent_locked = 0,
3962
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
3963
		.bio_flags = 0,
3964 3965
	};

C
Chris Mason 已提交
3966
	ret = extent_write_cache_pages(tree, mapping, wbc,
C
Chris Mason 已提交
3967 3968
				       __extent_writepage, &epd,
				       flush_write_bio);
3969
	flush_epd_write_bio(&epd);
3970 3971 3972 3973 3974 3975 3976 3977 3978 3979
	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
3980
	unsigned long bio_flags = 0;
L
Liu Bo 已提交
3981 3982
	struct page *pagepool[16];
	struct page *page;
3983
	struct extent_map *em_cached = NULL;
L
Liu Bo 已提交
3984
	int nr = 0;
3985 3986

	for (page_idx = 0; page_idx < nr_pages; page_idx++) {
L
Liu Bo 已提交
3987
		page = list_entry(pages->prev, struct page, lru);
3988 3989 3990

		prefetchw(&page->flags);
		list_del(&page->lru);
L
Liu Bo 已提交
3991
		if (add_to_page_cache_lru(page, mapping,
3992
					page->index, GFP_NOFS)) {
L
Liu Bo 已提交
3993 3994
			page_cache_release(page);
			continue;
3995
		}
L
Liu Bo 已提交
3996 3997 3998 3999

		pagepool[nr++] = page;
		if (nr < ARRAY_SIZE(pagepool))
			continue;
4000
		__extent_readpages(tree, pagepool, nr, get_extent, &em_cached,
4001
				   &bio, 0, &bio_flags, READ);
L
Liu Bo 已提交
4002
		nr = 0;
4003
	}
4004
	if (nr)
4005
		__extent_readpages(tree, pagepool, nr, get_extent, &em_cached,
4006
				   &bio, 0, &bio_flags, READ);
L
Liu Bo 已提交
4007

4008 4009 4010
	if (em_cached)
		free_extent_map(em_cached);

4011 4012
	BUG_ON(!list_empty(pages));
	if (bio)
4013
		return submit_one_bio(READ, bio, 0, bio_flags);
4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024
	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)
{
4025
	struct extent_state *cached_state = NULL;
M
Miao Xie 已提交
4026
	u64 start = page_offset(page);
4027 4028 4029
	u64 end = start + PAGE_CACHE_SIZE - 1;
	size_t blocksize = page->mapping->host->i_sb->s_blocksize;

4030
	start += ALIGN(offset, blocksize);
4031 4032 4033
	if (start > end)
		return 0;

4034
	lock_extent_bits(tree, start, end, 0, &cached_state);
4035
	wait_on_page_writeback(page);
4036
	clear_extent_bit(tree, start, end,
4037 4038
			 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
			 EXTENT_DO_ACCOUNTING,
4039
			 1, 1, &cached_state, GFP_NOFS);
4040 4041 4042
	return 0;
}

4043 4044 4045 4046 4047
/*
 * 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.
 */
4048 4049 4050
static int try_release_extent_state(struct extent_map_tree *map,
				    struct extent_io_tree *tree,
				    struct page *page, gfp_t mask)
4051
{
M
Miao Xie 已提交
4052
	u64 start = page_offset(page);
4053 4054 4055
	u64 end = start + PAGE_CACHE_SIZE - 1;
	int ret = 1;

4056
	if (test_range_bit(tree, start, end,
4057
			   EXTENT_IOBITS, 0, NULL))
4058 4059 4060 4061
		ret = 0;
	else {
		if ((mask & GFP_NOFS) == GFP_NOFS)
			mask = GFP_NOFS;
4062 4063 4064 4065
		/*
		 * at this point we can safely clear everything except the
		 * locked bit and the nodatasum bit
		 */
4066
		ret = clear_extent_bit(tree, start, end,
4067 4068
				 ~(EXTENT_LOCKED | EXTENT_NODATASUM),
				 0, 0, NULL, mask);
4069 4070 4071 4072 4073 4074 4075 4076

		/* if clear_extent_bit failed for enomem reasons,
		 * we can't allow the release to continue.
		 */
		if (ret < 0)
			ret = 0;
		else
			ret = 1;
4077 4078 4079 4080
	}
	return ret;
}

4081 4082 4083 4084 4085 4086
/*
 * 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,
4087 4088
			       struct extent_io_tree *tree, struct page *page,
			       gfp_t mask)
4089 4090
{
	struct extent_map *em;
M
Miao Xie 已提交
4091
	u64 start = page_offset(page);
4092
	u64 end = start + PAGE_CACHE_SIZE - 1;
4093

4094 4095
	if ((mask & __GFP_WAIT) &&
	    page->mapping->host->i_size > 16 * 1024 * 1024) {
4096
		u64 len;
4097
		while (start <= end) {
4098
			len = end - start + 1;
4099
			write_lock(&map->lock);
4100
			em = lookup_extent_mapping(map, start, len);
4101
			if (!em) {
4102
				write_unlock(&map->lock);
4103 4104
				break;
			}
4105 4106
			if (test_bit(EXTENT_FLAG_PINNED, &em->flags) ||
			    em->start != start) {
4107
				write_unlock(&map->lock);
4108 4109 4110 4111 4112
				free_extent_map(em);
				break;
			}
			if (!test_range_bit(tree, em->start,
					    extent_map_end(em) - 1,
4113
					    EXTENT_LOCKED | EXTENT_WRITEBACK,
4114
					    0, NULL)) {
4115 4116 4117 4118 4119
				remove_extent_mapping(map, em);
				/* once for the rb tree */
				free_extent_map(em);
			}
			start = extent_map_end(em);
4120
			write_unlock(&map->lock);
4121 4122

			/* once for us */
4123 4124 4125
			free_extent_map(em);
		}
	}
4126
	return try_release_extent_state(map, tree, page, mask);
4127 4128
}

4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141 4142 4143 4144
/*
 * helper function for fiemap, which doesn't want to see any holes.
 * This maps until we find something past 'last'
 */
static struct extent_map *get_extent_skip_holes(struct inode *inode,
						u64 offset,
						u64 last,
						get_extent_t *get_extent)
{
	u64 sectorsize = BTRFS_I(inode)->root->sectorsize;
	struct extent_map *em;
	u64 len;

	if (offset >= last)
		return NULL;

4145
	while (1) {
4146 4147 4148
		len = last - offset;
		if (len == 0)
			break;
4149
		len = ALIGN(len, sectorsize);
4150
		em = get_extent(inode, NULL, 0, offset, len, 0);
4151
		if (IS_ERR_OR_NULL(em))
4152 4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168
			return em;

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

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

4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181
static noinline int count_ext_ref(u64 inum, u64 offset, u64 root_id, void *ctx)
{
	unsigned long cnt = *((unsigned long *)ctx);

	cnt++;
	*((unsigned long *)ctx) = cnt;

	/* Now we're sure that the extent is shared. */
	if (cnt > 1)
		return 1;
	return 0;
}

Y
Yehuda Sadeh 已提交
4182 4183 4184
int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
		__u64 start, __u64 len, get_extent_t *get_extent)
{
J
Josef Bacik 已提交
4185
	int ret = 0;
Y
Yehuda Sadeh 已提交
4186 4187 4188
	u64 off = start;
	u64 max = start + len;
	u32 flags = 0;
J
Josef Bacik 已提交
4189 4190
	u32 found_type;
	u64 last;
4191
	u64 last_for_get_extent = 0;
Y
Yehuda Sadeh 已提交
4192
	u64 disko = 0;
4193
	u64 isize = i_size_read(inode);
J
Josef Bacik 已提交
4194
	struct btrfs_key found_key;
Y
Yehuda Sadeh 已提交
4195
	struct extent_map *em = NULL;
4196
	struct extent_state *cached_state = NULL;
J
Josef Bacik 已提交
4197
	struct btrfs_path *path;
Y
Yehuda Sadeh 已提交
4198
	int end = 0;
4199 4200 4201
	u64 em_start = 0;
	u64 em_len = 0;
	u64 em_end = 0;
Y
Yehuda Sadeh 已提交
4202 4203 4204 4205

	if (len == 0)
		return -EINVAL;

J
Josef Bacik 已提交
4206 4207 4208 4209 4210
	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;
	path->leave_spinning = 1;

4211 4212
	start = round_down(start, BTRFS_I(inode)->root->sectorsize);
	len = round_up(max, BTRFS_I(inode)->root->sectorsize) - start;
4213

4214 4215 4216 4217
	/*
	 * lookup the last file extent.  We're not using i_size here
	 * because there might be preallocation past i_size
	 */
J
Josef Bacik 已提交
4218
	ret = btrfs_lookup_file_extent(NULL, BTRFS_I(inode)->root,
L
Li Zefan 已提交
4219
				       path, btrfs_ino(inode), -1, 0);
J
Josef Bacik 已提交
4220 4221 4222 4223 4224 4225 4226 4227 4228
	if (ret < 0) {
		btrfs_free_path(path);
		return ret;
	}
	WARN_ON(!ret);
	path->slots[0]--;
	btrfs_item_key_to_cpu(path->nodes[0], &found_key, path->slots[0]);
	found_type = btrfs_key_type(&found_key);

4229
	/* No extents, but there might be delalloc bits */
L
Li Zefan 已提交
4230
	if (found_key.objectid != btrfs_ino(inode) ||
J
Josef Bacik 已提交
4231
	    found_type != BTRFS_EXTENT_DATA_KEY) {
4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242
		/* have to trust i_size as the end */
		last = (u64)-1;
		last_for_get_extent = isize;
	} else {
		/*
		 * remember the start of the last extent.  There are a
		 * bunch of different factors that go into the length of the
		 * extent, so its much less complex to remember where it started
		 */
		last = found_key.offset;
		last_for_get_extent = last + 1;
J
Josef Bacik 已提交
4243
	}
4244
	btrfs_release_path(path);
J
Josef Bacik 已提交
4245

4246 4247 4248 4249 4250 4251 4252 4253 4254 4255
	/*
	 * we might have some extents allocated but more delalloc past those
	 * extents.  so, we trust isize unless the start of the last extent is
	 * beyond isize
	 */
	if (last < isize) {
		last = (u64)-1;
		last_for_get_extent = isize;
	}

L
Liu Bo 已提交
4256
	lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + len - 1, 0,
4257
			 &cached_state);
4258

4259
	em = get_extent_skip_holes(inode, start, last_for_get_extent,
4260
				   get_extent);
Y
Yehuda Sadeh 已提交
4261 4262 4263 4264 4265 4266
	if (!em)
		goto out;
	if (IS_ERR(em)) {
		ret = PTR_ERR(em);
		goto out;
	}
J
Josef Bacik 已提交
4267

Y
Yehuda Sadeh 已提交
4268
	while (!end) {
4269
		u64 offset_in_extent = 0;
4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281

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

		/*
		 * get_extent may return an extent that starts before our
		 * requested range.  We have to make sure the ranges
		 * we return to fiemap always move forward and don't
		 * overlap, so adjust the offsets here
		 */
		em_start = max(em->start, off);
Y
Yehuda Sadeh 已提交
4282

4283 4284
		/*
		 * record the offset from the start of the extent
4285 4286 4287
		 * for adjusting the disk offset below.  Only do this if the
		 * extent isn't compressed since our in ram offset may be past
		 * what we have actually allocated on disk.
4288
		 */
4289 4290
		if (!test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
			offset_in_extent = em_start - em->start;
4291
		em_end = extent_map_end(em);
4292
		em_len = em_end - em_start;
Y
Yehuda Sadeh 已提交
4293 4294 4295
		disko = 0;
		flags = 0;

4296 4297 4298 4299 4300 4301 4302
		/*
		 * bump off for our next call to get_extent
		 */
		off = extent_map_end(em);
		if (off >= max)
			end = 1;

4303
		if (em->block_start == EXTENT_MAP_LAST_BYTE) {
Y
Yehuda Sadeh 已提交
4304 4305
			end = 1;
			flags |= FIEMAP_EXTENT_LAST;
4306
		} else if (em->block_start == EXTENT_MAP_INLINE) {
Y
Yehuda Sadeh 已提交
4307 4308
			flags |= (FIEMAP_EXTENT_DATA_INLINE |
				  FIEMAP_EXTENT_NOT_ALIGNED);
4309
		} else if (em->block_start == EXTENT_MAP_DELALLOC) {
Y
Yehuda Sadeh 已提交
4310 4311
			flags |= (FIEMAP_EXTENT_DELALLOC |
				  FIEMAP_EXTENT_UNKNOWN);
4312
		} else {
4313 4314
			unsigned long ref_cnt = 0;

4315
			disko = em->block_start + offset_in_extent;
4316 4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330

			/*
			 * As btrfs supports shared space, this information
			 * can be exported to userspace tools via
			 * flag FIEMAP_EXTENT_SHARED.
			 */
			ret = iterate_inodes_from_logical(
					em->block_start,
					BTRFS_I(inode)->root->fs_info,
					path, count_ext_ref, &ref_cnt);
			if (ret < 0 && ret != -ENOENT)
				goto out_free;

			if (ref_cnt > 1)
				flags |= FIEMAP_EXTENT_SHARED;
Y
Yehuda Sadeh 已提交
4331 4332 4333 4334 4335 4336
		}
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
			flags |= FIEMAP_EXTENT_ENCODED;

		free_extent_map(em);
		em = NULL;
4337 4338
		if ((em_start >= last) || em_len == (u64)-1 ||
		   (last == (u64)-1 && isize <= em_end)) {
Y
Yehuda Sadeh 已提交
4339 4340 4341 4342
			flags |= FIEMAP_EXTENT_LAST;
			end = 1;
		}

4343 4344 4345 4346 4347 4348 4349 4350
		/* now scan forward to see if this is really the last extent. */
		em = get_extent_skip_holes(inode, off, last_for_get_extent,
					   get_extent);
		if (IS_ERR(em)) {
			ret = PTR_ERR(em);
			goto out;
		}
		if (!em) {
J
Josef Bacik 已提交
4351 4352 4353
			flags |= FIEMAP_EXTENT_LAST;
			end = 1;
		}
4354 4355 4356 4357
		ret = fiemap_fill_next_extent(fieinfo, em_start, disko,
					      em_len, flags);
		if (ret)
			goto out_free;
Y
Yehuda Sadeh 已提交
4358 4359 4360 4361
	}
out_free:
	free_extent_map(em);
out:
4362
	btrfs_free_path(path);
L
Liu Bo 已提交
4363
	unlock_extent_cached(&BTRFS_I(inode)->io_tree, start, start + len - 1,
4364
			     &cached_state, GFP_NOFS);
Y
Yehuda Sadeh 已提交
4365 4366 4367
	return ret;
}

4368 4369
static void __free_extent_buffer(struct extent_buffer *eb)
{
4370
	btrfs_leak_debug_del(&eb->leak_list);
4371 4372 4373
	kmem_cache_free(extent_buffer_cache, eb);
}

4374
int extent_buffer_under_io(struct extent_buffer *eb)
4375 4376 4377 4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404 4405 4406 4407 4408 4409 4410 4411 4412 4413 4414 4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443
{
	return (atomic_read(&eb->io_pages) ||
		test_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags) ||
		test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
}

/*
 * Helper for releasing extent buffer page.
 */
static void btrfs_release_extent_buffer_page(struct extent_buffer *eb,
						unsigned long start_idx)
{
	unsigned long index;
	unsigned long num_pages;
	struct page *page;
	int mapped = !test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags);

	BUG_ON(extent_buffer_under_io(eb));

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

	do {
		index--;
		page = extent_buffer_page(eb, index);
		if (page && mapped) {
			spin_lock(&page->mapping->private_lock);
			/*
			 * We do this since we'll remove the pages after we've
			 * removed the eb from the radix tree, so we could race
			 * and have this page now attached to the new eb.  So
			 * only clear page_private if it's still connected to
			 * this eb.
			 */
			if (PagePrivate(page) &&
			    page->private == (unsigned long)eb) {
				BUG_ON(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
				BUG_ON(PageDirty(page));
				BUG_ON(PageWriteback(page));
				/*
				 * We need to make sure we haven't be attached
				 * to a new eb.
				 */
				ClearPagePrivate(page);
				set_page_private(page, 0);
				/* One for the page private */
				page_cache_release(page);
			}
			spin_unlock(&page->mapping->private_lock);

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

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

4444 4445 4446
static struct extent_buffer *
__alloc_extent_buffer(struct btrfs_fs_info *fs_info, u64 start,
		      unsigned long len, gfp_t mask)
4447 4448 4449 4450
{
	struct extent_buffer *eb = NULL;

	eb = kmem_cache_zalloc(extent_buffer_cache, mask);
T
Tsutomu Itoh 已提交
4451 4452
	if (eb == NULL)
		return NULL;
4453 4454
	eb->start = start;
	eb->len = len;
4455
	eb->fs_info = fs_info;
4456
	eb->bflags = 0;
4457 4458 4459 4460 4461 4462 4463
	rwlock_init(&eb->lock);
	atomic_set(&eb->write_locks, 0);
	atomic_set(&eb->read_locks, 0);
	atomic_set(&eb->blocking_readers, 0);
	atomic_set(&eb->blocking_writers, 0);
	atomic_set(&eb->spinning_readers, 0);
	atomic_set(&eb->spinning_writers, 0);
4464
	eb->lock_nested = 0;
4465 4466
	init_waitqueue_head(&eb->write_lock_wq);
	init_waitqueue_head(&eb->read_lock_wq);
4467

4468 4469
	btrfs_leak_debug_add(&eb->leak_list, &buffers);

4470
	spin_lock_init(&eb->refs_lock);
4471
	atomic_set(&eb->refs, 1);
4472
	atomic_set(&eb->io_pages, 0);
4473

4474 4475 4476 4477 4478 4479
	/*
	 * Sanity checks, currently the maximum is 64k covered by 16x 4k pages
	 */
	BUILD_BUG_ON(BTRFS_MAX_METADATA_BLOCKSIZE
		> MAX_INLINE_EXTENT_BUFFER_SIZE);
	BUG_ON(len > MAX_INLINE_EXTENT_BUFFER_SIZE);
4480 4481 4482 4483

	return eb;
}

4484 4485 4486 4487 4488 4489 4490
struct extent_buffer *btrfs_clone_extent_buffer(struct extent_buffer *src)
{
	unsigned long i;
	struct page *p;
	struct extent_buffer *new;
	unsigned long num_pages = num_extent_pages(src->start, src->len);

4491
	new = __alloc_extent_buffer(NULL, src->start, src->len, GFP_NOFS);
4492 4493 4494 4495
	if (new == NULL)
		return NULL;

	for (i = 0; i < num_pages; i++) {
4496
		p = alloc_page(GFP_NOFS);
4497 4498 4499 4500
		if (!p) {
			btrfs_release_extent_buffer(new);
			return NULL;
		}
4501 4502 4503 4504 4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518 4519
		attach_extent_buffer_page(new, p);
		WARN_ON(PageDirty(p));
		SetPageUptodate(p);
		new->pages[i] = p;
	}

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

	return new;
}

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

4520
	eb = __alloc_extent_buffer(NULL, start, len, GFP_NOFS);
4521 4522 4523 4524
	if (!eb)
		return NULL;

	for (i = 0; i < num_pages; i++) {
4525
		eb->pages[i] = alloc_page(GFP_NOFS);
4526 4527 4528 4529 4530 4531 4532 4533 4534
		if (!eb->pages[i])
			goto err;
	}
	set_extent_buffer_uptodate(eb);
	btrfs_set_header_nritems(eb, 0);
	set_bit(EXTENT_BUFFER_DUMMY, &eb->bflags);

	return eb;
err:
4535 4536
	for (; i > 0; i--)
		__free_page(eb->pages[i - 1]);
4537 4538 4539 4540
	__free_extent_buffer(eb);
	return NULL;
}

4541 4542
static void check_buffer_tree_ref(struct extent_buffer *eb)
{
4543
	int refs;
4544 4545 4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563
	/* the ref bit is tricky.  We have to make sure it is set
	 * if we have the buffer dirty.   Otherwise the
	 * code to free a buffer can end up dropping a dirty
	 * page
	 *
	 * Once the ref bit is set, it won't go away while the
	 * buffer is dirty or in writeback, and it also won't
	 * go away while we have the reference count on the
	 * eb bumped.
	 *
	 * We can't just set the ref bit without bumping the
	 * ref on the eb because free_extent_buffer might
	 * see the ref bit and try to clear it.  If this happens
	 * free_extent_buffer might end up dropping our original
	 * ref by mistake and freeing the page before we are able
	 * to add one more ref.
	 *
	 * So bump the ref count first, then set the bit.  If someone
	 * beat us to it, drop the ref we added.
	 */
4564 4565 4566 4567
	refs = atomic_read(&eb->refs);
	if (refs >= 2 && test_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
		return;

4568 4569
	spin_lock(&eb->refs_lock);
	if (!test_and_set_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
4570
		atomic_inc(&eb->refs);
4571
	spin_unlock(&eb->refs_lock);
4572 4573
}

4574 4575
static void mark_extent_buffer_accessed(struct extent_buffer *eb,
		struct page *accessed)
4576 4577 4578
{
	unsigned long num_pages, i;

4579 4580
	check_buffer_tree_ref(eb);

4581 4582 4583
	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = 0; i < num_pages; i++) {
		struct page *p = extent_buffer_page(eb, i);
4584 4585
		if (p != accessed)
			mark_page_accessed(p);
4586 4587 4588
	}
}

4589 4590
struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
					 u64 start)
4591 4592 4593 4594
{
	struct extent_buffer *eb;

	rcu_read_lock();
4595 4596
	eb = radix_tree_lookup(&fs_info->buffer_radix,
			       start >> PAGE_CACHE_SHIFT);
4597 4598
	if (eb && atomic_inc_not_zero(&eb->refs)) {
		rcu_read_unlock();
4599
		mark_extent_buffer_accessed(eb, NULL);
4600 4601 4602 4603 4604 4605 4606
		return eb;
	}
	rcu_read_unlock();

	return NULL;
}

4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650 4651 4652 4653
#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
					       u64 start, unsigned long len)
{
	struct extent_buffer *eb, *exists = NULL;
	int ret;

	eb = find_extent_buffer(fs_info, start);
	if (eb)
		return eb;
	eb = alloc_dummy_extent_buffer(start, len);
	if (!eb)
		return NULL;
	eb->fs_info = fs_info;
again:
	ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
	if (ret)
		goto free_eb;
	spin_lock(&fs_info->buffer_lock);
	ret = radix_tree_insert(&fs_info->buffer_radix,
				start >> PAGE_CACHE_SHIFT, eb);
	spin_unlock(&fs_info->buffer_lock);
	radix_tree_preload_end();
	if (ret == -EEXIST) {
		exists = find_extent_buffer(fs_info, start);
		if (exists)
			goto free_eb;
		else
			goto again;
	}
	check_buffer_tree_ref(eb);
	set_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags);

	/*
	 * We will free dummy extent buffer's if they come into
	 * free_extent_buffer with a ref count of 2, but if we are using this we
	 * want the buffers to stay in memory until we're done with them, so
	 * bump the ref count again.
	 */
	atomic_inc(&eb->refs);
	return eb;
free_eb:
	btrfs_release_extent_buffer(eb);
	return exists;
}
#endif

4654
struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
4655
					  u64 start, unsigned long len)
4656 4657 4658 4659 4660
{
	unsigned long num_pages = num_extent_pages(start, len);
	unsigned long i;
	unsigned long index = start >> PAGE_CACHE_SHIFT;
	struct extent_buffer *eb;
4661
	struct extent_buffer *exists = NULL;
4662
	struct page *p;
4663
	struct address_space *mapping = fs_info->btree_inode->i_mapping;
4664
	int uptodate = 1;
4665
	int ret;
4666

4667
	eb = find_extent_buffer(fs_info, start);
4668
	if (eb)
4669 4670
		return eb;

4671
	eb = __alloc_extent_buffer(fs_info, start, len, GFP_NOFS);
4672
	if (!eb)
4673 4674
		return NULL;

4675
	for (i = 0; i < num_pages; i++, index++) {
4676
		p = find_or_create_page(mapping, index, GFP_NOFS);
4677
		if (!p)
4678
			goto free_eb;
J
Josef Bacik 已提交
4679 4680 4681 4682 4683 4684 4685 4686 4687 4688 4689 4690 4691 4692

		spin_lock(&mapping->private_lock);
		if (PagePrivate(p)) {
			/*
			 * We could have already allocated an eb for this page
			 * and attached one so lets see if we can get a ref on
			 * the existing eb, and if we can we know it's good and
			 * we can just return that one, else we know we can just
			 * overwrite page->private.
			 */
			exists = (struct extent_buffer *)p->private;
			if (atomic_inc_not_zero(&exists->refs)) {
				spin_unlock(&mapping->private_lock);
				unlock_page(p);
4693
				page_cache_release(p);
4694
				mark_extent_buffer_accessed(exists, p);
J
Josef Bacik 已提交
4695 4696 4697
				goto free_eb;
			}

4698
			/*
J
Josef Bacik 已提交
4699 4700 4701 4702
			 * Do this so attach doesn't complain and we need to
			 * drop the ref the old guy had.
			 */
			ClearPagePrivate(p);
4703
			WARN_ON(PageDirty(p));
J
Josef Bacik 已提交
4704
			page_cache_release(p);
4705
		}
J
Josef Bacik 已提交
4706 4707
		attach_extent_buffer_page(eb, p);
		spin_unlock(&mapping->private_lock);
4708
		WARN_ON(PageDirty(p));
4709
		eb->pages[i] = p;
4710 4711
		if (!PageUptodate(p))
			uptodate = 0;
C
Chris Mason 已提交
4712 4713 4714 4715 4716

		/*
		 * see below about how we avoid a nasty race with release page
		 * and why we unlock later
		 */
4717 4718
	}
	if (uptodate)
4719
		set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
4720
again:
4721 4722 4723 4724
	ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
	if (ret)
		goto free_eb;

4725 4726 4727 4728
	spin_lock(&fs_info->buffer_lock);
	ret = radix_tree_insert(&fs_info->buffer_radix,
				start >> PAGE_CACHE_SHIFT, eb);
	spin_unlock(&fs_info->buffer_lock);
4729
	radix_tree_preload_end();
4730
	if (ret == -EEXIST) {
4731
		exists = find_extent_buffer(fs_info, start);
4732 4733 4734
		if (exists)
			goto free_eb;
		else
4735
			goto again;
4736 4737
	}
	/* add one reference for the tree */
4738
	check_buffer_tree_ref(eb);
4739
	set_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags);
C
Chris Mason 已提交
4740 4741 4742 4743 4744 4745 4746 4747 4748 4749

	/*
	 * there is a race where release page may have
	 * tried to find this extent buffer in the radix
	 * but failed.  It will tell the VM it is safe to
	 * reclaim the, and it will clear the page private bit.
	 * We must make sure to set the page private bit properly
	 * after the extent buffer is in the radix tree so
	 * it doesn't get lost
	 */
4750 4751 4752 4753 4754 4755 4756
	SetPageChecked(eb->pages[0]);
	for (i = 1; i < num_pages; i++) {
		p = extent_buffer_page(eb, i);
		ClearPageChecked(p);
		unlock_page(p);
	}
	unlock_page(eb->pages[0]);
4757 4758
	return eb;

4759
free_eb:
4760 4761 4762 4763
	for (i = 0; i < num_pages; i++) {
		if (eb->pages[i])
			unlock_page(eb->pages[i]);
	}
C
Chris Mason 已提交
4764

4765
	WARN_ON(!atomic_dec_and_test(&eb->refs));
4766
	btrfs_release_extent_buffer(eb);
4767
	return exists;
4768 4769
}

4770 4771 4772 4773 4774 4775 4776 4777 4778
static inline void btrfs_release_extent_buffer_rcu(struct rcu_head *head)
{
	struct extent_buffer *eb =
			container_of(head, struct extent_buffer, rcu_head);

	__free_extent_buffer(eb);
}

/* Expects to have eb->eb_lock already held */
4779
static int release_extent_buffer(struct extent_buffer *eb)
4780 4781 4782
{
	WARN_ON(atomic_read(&eb->refs) == 0);
	if (atomic_dec_and_test(&eb->refs)) {
4783
		if (test_and_clear_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags)) {
4784
			struct btrfs_fs_info *fs_info = eb->fs_info;
4785

4786
			spin_unlock(&eb->refs_lock);
4787

4788 4789
			spin_lock(&fs_info->buffer_lock);
			radix_tree_delete(&fs_info->buffer_radix,
4790
					  eb->start >> PAGE_CACHE_SHIFT);
4791
			spin_unlock(&fs_info->buffer_lock);
4792 4793
		} else {
			spin_unlock(&eb->refs_lock);
4794
		}
4795 4796 4797 4798

		/* Should be safe to release our pages at this point */
		btrfs_release_extent_buffer_page(eb, 0);
		call_rcu(&eb->rcu_head, btrfs_release_extent_buffer_rcu);
4799
		return 1;
4800 4801
	}
	spin_unlock(&eb->refs_lock);
4802 4803

	return 0;
4804 4805
}

4806 4807
void free_extent_buffer(struct extent_buffer *eb)
{
4808 4809
	int refs;
	int old;
4810 4811 4812
	if (!eb)
		return;

4813 4814 4815 4816 4817 4818 4819 4820 4821
	while (1) {
		refs = atomic_read(&eb->refs);
		if (refs <= 3)
			break;
		old = atomic_cmpxchg(&eb->refs, refs, refs - 1);
		if (old == refs)
			return;
	}

4822
	spin_lock(&eb->refs_lock);
4823 4824 4825 4826
	if (atomic_read(&eb->refs) == 2 &&
	    test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags))
		atomic_dec(&eb->refs);

4827 4828
	if (atomic_read(&eb->refs) == 2 &&
	    test_bit(EXTENT_BUFFER_STALE, &eb->bflags) &&
4829
	    !extent_buffer_under_io(eb) &&
4830 4831 4832 4833 4834 4835 4836
	    test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
		atomic_dec(&eb->refs);

	/*
	 * I know this is terrible, but it's temporary until we stop tracking
	 * the uptodate bits and such for the extent buffers.
	 */
4837
	release_extent_buffer(eb);
4838 4839 4840 4841 4842
}

void free_extent_buffer_stale(struct extent_buffer *eb)
{
	if (!eb)
4843 4844
		return;

4845 4846 4847
	spin_lock(&eb->refs_lock);
	set_bit(EXTENT_BUFFER_STALE, &eb->bflags);

4848
	if (atomic_read(&eb->refs) == 2 && !extent_buffer_under_io(eb) &&
4849 4850
	    test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
		atomic_dec(&eb->refs);
4851
	release_extent_buffer(eb);
4852 4853
}

4854
void clear_extent_buffer_dirty(struct extent_buffer *eb)
4855 4856 4857 4858 4859 4860 4861 4862 4863
{
	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);
4864
		if (!PageDirty(page))
C
Chris Mason 已提交
4865 4866
			continue;

4867
		lock_page(page);
C
Chris Mason 已提交
4868 4869
		WARN_ON(!PagePrivate(page));

4870
		clear_page_dirty_for_io(page);
4871
		spin_lock_irq(&page->mapping->tree_lock);
4872 4873 4874 4875 4876
		if (!PageDirty(page)) {
			radix_tree_tag_clear(&page->mapping->page_tree,
						page_index(page),
						PAGECACHE_TAG_DIRTY);
		}
4877
		spin_unlock_irq(&page->mapping->tree_lock);
4878
		ClearPageError(page);
4879
		unlock_page(page);
4880
	}
4881
	WARN_ON(atomic_read(&eb->refs) == 0);
4882 4883
}

4884
int set_extent_buffer_dirty(struct extent_buffer *eb)
4885 4886 4887
{
	unsigned long i;
	unsigned long num_pages;
4888
	int was_dirty = 0;
4889

4890 4891
	check_buffer_tree_ref(eb);

4892
	was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
4893

4894
	num_pages = num_extent_pages(eb->start, eb->len);
4895
	WARN_ON(atomic_read(&eb->refs) == 0);
4896 4897
	WARN_ON(!test_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags));

4898
	for (i = 0; i < num_pages; i++)
4899
		set_page_dirty(extent_buffer_page(eb, i));
4900
	return was_dirty;
4901 4902
}

4903
int clear_extent_buffer_uptodate(struct extent_buffer *eb)
4904 4905 4906 4907 4908
{
	unsigned long i;
	struct page *page;
	unsigned long num_pages;

4909
	clear_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
4910
	num_pages = num_extent_pages(eb->start, eb->len);
4911 4912
	for (i = 0; i < num_pages; i++) {
		page = extent_buffer_page(eb, i);
C
Chris Mason 已提交
4913 4914
		if (page)
			ClearPageUptodate(page);
4915 4916 4917 4918
	}
	return 0;
}

4919
int set_extent_buffer_uptodate(struct extent_buffer *eb)
4920 4921 4922 4923 4924
{
	unsigned long i;
	struct page *page;
	unsigned long num_pages;

4925
	set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
4926 4927 4928 4929 4930 4931 4932 4933
	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = 0; i < num_pages; i++) {
		page = extent_buffer_page(eb, i);
		SetPageUptodate(page);
	}
	return 0;
}

4934
int extent_buffer_uptodate(struct extent_buffer *eb)
4935
{
4936
	return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
4937 4938 4939
}

int read_extent_buffer_pages(struct extent_io_tree *tree,
4940
			     struct extent_buffer *eb, u64 start, int wait,
4941
			     get_extent_t *get_extent, int mirror_num)
4942 4943 4944 4945 4946 4947
{
	unsigned long i;
	unsigned long start_i;
	struct page *page;
	int err;
	int ret = 0;
4948 4949
	int locked_pages = 0;
	int all_uptodate = 1;
4950
	unsigned long num_pages;
4951
	unsigned long num_reads = 0;
4952
	struct bio *bio = NULL;
C
Chris Mason 已提交
4953
	unsigned long bio_flags = 0;
4954

4955
	if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags))
4956 4957 4958 4959 4960 4961 4962 4963 4964 4965 4966 4967 4968
		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);
4969
		if (wait == WAIT_NONE) {
4970
			if (!trylock_page(page))
4971
				goto unlock_exit;
4972 4973 4974
		} else {
			lock_page(page);
		}
4975
		locked_pages++;
4976 4977
		if (!PageUptodate(page)) {
			num_reads++;
4978
			all_uptodate = 0;
4979
		}
4980 4981 4982
	}
	if (all_uptodate) {
		if (start_i == 0)
4983
			set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
4984 4985 4986
		goto unlock_exit;
	}

4987
	clear_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
4988
	eb->read_mirror = 0;
4989
	atomic_set(&eb->io_pages, num_reads);
4990 4991 4992
	for (i = start_i; i < num_pages; i++) {
		page = extent_buffer_page(eb, i);
		if (!PageUptodate(page)) {
4993
			ClearPageError(page);
4994
			err = __extent_read_full_page(tree, page,
4995
						      get_extent, &bio,
4996 4997
						      mirror_num, &bio_flags,
						      READ | REQ_META);
C
Chris Mason 已提交
4998
			if (err)
4999 5000 5001 5002 5003 5004
				ret = err;
		} else {
			unlock_page(page);
		}
	}

5005
	if (bio) {
5006 5007
		err = submit_one_bio(READ | REQ_META, bio, mirror_num,
				     bio_flags);
5008 5009
		if (err)
			return err;
5010
	}
5011

5012
	if (ret || wait != WAIT_COMPLETE)
5013
		return ret;
C
Chris Mason 已提交
5014

5015 5016 5017
	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
5018
		if (!PageUptodate(page))
5019 5020
			ret = -EIO;
	}
C
Chris Mason 已提交
5021

5022
	return ret;
5023 5024 5025

unlock_exit:
	i = start_i;
C
Chris Mason 已提交
5026
	while (locked_pages > 0) {
5027 5028 5029 5030 5031 5032
		page = extent_buffer_page(eb, i);
		i++;
		unlock_page(page);
		locked_pages--;
	}
	return ret;
5033 5034 5035 5036 5037 5038 5039 5040 5041 5042 5043 5044 5045 5046 5047 5048 5049
}

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);

5050
	offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
5051

C
Chris Mason 已提交
5052
	while (len > 0) {
5053 5054 5055
		page = extent_buffer_page(eb, i);

		cur = min(len, (PAGE_CACHE_SIZE - offset));
5056
		kaddr = page_address(page);
5057 5058 5059 5060 5061 5062 5063 5064 5065
		memcpy(dst, kaddr + offset, cur);

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

5066 5067 5068 5069 5070 5071 5072 5073 5074 5075 5076 5077 5078 5079 5080 5081 5082 5083 5084 5085 5086 5087 5088 5089 5090 5091 5092 5093 5094 5095 5096 5097 5098 5099 5100 5101 5102
int read_extent_buffer_to_user(struct extent_buffer *eb, void __user *dstv,
			unsigned long start,
			unsigned long len)
{
	size_t cur;
	size_t offset;
	struct page *page;
	char *kaddr;
	char __user *dst = (char __user *)dstv;
	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) & (PAGE_CACHE_SIZE - 1);

	while (len > 0) {
		page = extent_buffer_page(eb, i);

		cur = min(len, (PAGE_CACHE_SIZE - offset));
		kaddr = page_address(page);
		if (copy_to_user(dst, kaddr + offset, cur)) {
			ret = -EFAULT;
			break;
		}

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

	return ret;
}

5103
int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start,
5104
			       unsigned long min_len, char **map,
5105
			       unsigned long *map_start,
5106
			       unsigned long *map_len)
5107 5108 5109 5110 5111 5112 5113 5114 5115 5116 5117 5118 5119 5120 5121 5122 5123 5124 5125
{
	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
5126

5127
	if (start + min_len > eb->len) {
J
Julia Lawall 已提交
5128
		WARN(1, KERN_ERR "btrfs bad mapping eb start %llu len %lu, "
5129 5130
		       "wanted %lu %lu\n",
		       eb->start, eb->len, start, min_len);
5131
		return -EINVAL;
5132 5133 5134
	}

	p = extent_buffer_page(eb, i);
5135
	kaddr = page_address(p);
5136 5137 5138 5139 5140 5141 5142 5143 5144 5145 5146 5147 5148 5149 5150 5151 5152 5153 5154 5155 5156
	*map = kaddr + offset;
	*map_len = PAGE_CACHE_SIZE - offset;
	return 0;
}

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

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

5157
	offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
5158

C
Chris Mason 已提交
5159
	while (len > 0) {
5160 5161 5162 5163
		page = extent_buffer_page(eb, i);

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

5164
		kaddr = page_address(page);
5165 5166 5167 5168 5169 5170 5171 5172 5173 5174 5175 5176 5177 5178 5179 5180 5181 5182 5183 5184 5185 5186 5187 5188 5189 5190
		ret = memcmp(ptr, kaddr + offset, cur);
		if (ret)
			break;

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

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

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

5191
	offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
5192

C
Chris Mason 已提交
5193
	while (len > 0) {
5194 5195 5196 5197
		page = extent_buffer_page(eb, i);
		WARN_ON(!PageUptodate(page));

		cur = min(len, PAGE_CACHE_SIZE - offset);
5198
		kaddr = page_address(page);
5199 5200 5201 5202 5203 5204 5205 5206 5207 5208 5209 5210 5211 5212 5213 5214 5215 5216 5217 5218 5219 5220
		memcpy(kaddr + offset, src, cur);

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

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

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

5221
	offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
5222

C
Chris Mason 已提交
5223
	while (len > 0) {
5224 5225 5226 5227
		page = extent_buffer_page(eb, i);
		WARN_ON(!PageUptodate(page));

		cur = min(len, PAGE_CACHE_SIZE - offset);
5228
		kaddr = page_address(page);
5229 5230 5231 5232 5233 5234 5235 5236 5237 5238 5239 5240 5241 5242 5243 5244 5245 5246 5247 5248 5249 5250 5251
		memset(kaddr + offset, c, cur);

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

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

	WARN_ON(src->len != dst_len);

	offset = (start_offset + dst_offset) &
5252
		(PAGE_CACHE_SIZE - 1);
5253

C
Chris Mason 已提交
5254
	while (len > 0) {
5255 5256 5257 5258 5259
		page = extent_buffer_page(dst, i);
		WARN_ON(!PageUptodate(page));

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

5260
		kaddr = page_address(page);
5261 5262 5263 5264 5265 5266 5267 5268 5269
		read_extent_buffer(src, kaddr + offset, src_offset, cur);

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

5270 5271 5272 5273 5274 5275
static inline bool areas_overlap(unsigned long src, unsigned long dst, unsigned long len)
{
	unsigned long distance = (src > dst) ? src - dst : dst - src;
	return distance < len;
}

5276 5277 5278 5279
static void copy_pages(struct page *dst_page, struct page *src_page,
		       unsigned long dst_off, unsigned long src_off,
		       unsigned long len)
{
5280
	char *dst_kaddr = page_address(dst_page);
5281
	char *src_kaddr;
5282
	int must_memmove = 0;
5283

5284
	if (dst_page != src_page) {
5285
		src_kaddr = page_address(src_page);
5286
	} else {
5287
		src_kaddr = dst_kaddr;
5288 5289
		if (areas_overlap(src_off, dst_off, len))
			must_memmove = 1;
5290
	}
5291

5292 5293 5294 5295
	if (must_memmove)
		memmove(dst_kaddr + dst_off, src_kaddr + src_off, len);
	else
		memcpy(dst_kaddr + dst_off, src_kaddr + src_off, len);
5296 5297 5298 5299 5300 5301 5302 5303 5304 5305 5306 5307 5308
}

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) {
5309
		printk(KERN_ERR "BTRFS: memmove bogus src_offset %lu move "
C
Chris Mason 已提交
5310
		       "len %lu dst len %lu\n", src_offset, len, dst->len);
5311 5312 5313
		BUG_ON(1);
	}
	if (dst_offset + len > dst->len) {
5314
		printk(KERN_ERR "BTRFS: memmove bogus dst_offset %lu move "
C
Chris Mason 已提交
5315
		       "len %lu dst len %lu\n", dst_offset, len, dst->len);
5316 5317 5318
		BUG_ON(1);
	}

C
Chris Mason 已提交
5319
	while (len > 0) {
5320
		dst_off_in_page = (start_offset + dst_offset) &
5321
			(PAGE_CACHE_SIZE - 1);
5322
		src_off_in_page = (start_offset + src_offset) &
5323
			(PAGE_CACHE_SIZE - 1);
5324 5325 5326 5327 5328 5329 5330 5331 5332 5333 5334 5335 5336 5337 5338 5339 5340 5341 5342 5343 5344 5345 5346 5347 5348 5349 5350 5351 5352 5353 5354 5355

		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) {
5356
		printk(KERN_ERR "BTRFS: memmove bogus src_offset %lu move "
C
Chris Mason 已提交
5357
		       "len %lu len %lu\n", src_offset, len, dst->len);
5358 5359 5360
		BUG_ON(1);
	}
	if (dst_offset + len > dst->len) {
5361
		printk(KERN_ERR "BTRFS: memmove bogus dst_offset %lu move "
C
Chris Mason 已提交
5362
		       "len %lu len %lu\n", dst_offset, len, dst->len);
5363 5364
		BUG_ON(1);
	}
5365
	if (dst_offset < src_offset) {
5366 5367 5368
		memcpy_extent_buffer(dst, dst_offset, src_offset, len);
		return;
	}
C
Chris Mason 已提交
5369
	while (len > 0) {
5370 5371 5372 5373
		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) &
5374
			(PAGE_CACHE_SIZE - 1);
5375
		src_off_in_page = (start_offset + src_end) &
5376
			(PAGE_CACHE_SIZE - 1);
5377 5378 5379

		cur = min_t(unsigned long, len, src_off_in_page + 1);
		cur = min(cur, dst_off_in_page + 1);
Z
Zach Brown 已提交
5380
		copy_pages(extent_buffer_page(dst, dst_i),
5381 5382 5383 5384 5385 5386 5387 5388 5389
			   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;
	}
}
5390

5391
int try_release_extent_buffer(struct page *page)
5392
{
5393 5394
	struct extent_buffer *eb;

5395 5396 5397 5398 5399 5400 5401
	/*
	 * We need to make sure noboody is attaching this page to an eb right
	 * now.
	 */
	spin_lock(&page->mapping->private_lock);
	if (!PagePrivate(page)) {
		spin_unlock(&page->mapping->private_lock);
J
Josef Bacik 已提交
5402
		return 1;
5403
	}
5404

5405 5406
	eb = (struct extent_buffer *)page->private;
	BUG_ON(!eb);
5407 5408

	/*
5409 5410 5411
	 * This is a little awful but should be ok, we need to make sure that
	 * the eb doesn't disappear out from under us while we're looking at
	 * this page.
5412
	 */
5413
	spin_lock(&eb->refs_lock);
5414
	if (atomic_read(&eb->refs) != 1 || extent_buffer_under_io(eb)) {
5415 5416 5417
		spin_unlock(&eb->refs_lock);
		spin_unlock(&page->mapping->private_lock);
		return 0;
5418
	}
5419
	spin_unlock(&page->mapping->private_lock);
5420

5421
	/*
5422 5423
	 * If tree ref isn't set then we know the ref on this eb is a real ref,
	 * so just return, this page will likely be freed soon anyway.
5424
	 */
5425 5426 5427
	if (!test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags)) {
		spin_unlock(&eb->refs_lock);
		return 0;
5428
	}
5429

5430
	return release_extent_buffer(eb);
5431
}