extent_io.c 143.6 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 29 30 31 32
static inline bool extent_state_in_tree(const struct extent_state *state)
{
	return !RB_EMPTY_NODE(&state->rb_node);
}

33
#ifdef CONFIG_BTRFS_DEBUG
34 35
static LIST_HEAD(buffers);
static LIST_HEAD(states);
C
Chris Mason 已提交
36

C
Chris Mason 已提交
37
static DEFINE_SPINLOCK(leak_lock);
38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66

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);
67
		pr_err("BTRFS: state leak: start %llu end %llu state %u in tree %d refs %d\n",
68 69
		       state->start, state->end, state->state,
		       extent_state_in_tree(state),
70
		       atomic_read(&state->refs));
71 72 73 74 75 76
		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);
77
		printk(KERN_ERR "BTRFS: buffer leak start %llu len %lu "
78 79
		       "refs %d\n",
		       eb->start, eb->len, atomic_read(&eb->refs));
80 81 82 83
		list_del(&eb->leak_list);
		kmem_cache_free(extent_buffer_cache, eb);
	}
}
84

85 86
#define btrfs_debug_check_extent_io_range(tree, start, end)		\
	__btrfs_debug_check_extent_io_range(__func__, (tree), (start), (end))
87
static inline void __btrfs_debug_check_extent_io_range(const char *caller,
88
		struct extent_io_tree *tree, u64 start, u64 end)
89
{
90 91
	struct inode *inode;
	u64 isize;
92

93 94
	if (!tree->mapping)
		return;
95

96 97
	inode = tree->mapping->host;
	isize = i_size_read(inode);
98
	if (end >= PAGE_SIZE && (end % 2) == 0 && end != isize - 1) {
99 100
		btrfs_debug_rl(BTRFS_I(inode)->root->fs_info,
		    "%s: ino %llu isize %llu odd range [%llu,%llu]",
101
				caller, btrfs_ino(inode), isize, start, end);
102 103
	}
}
104 105 106 107
#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)
108
#define btrfs_debug_check_extent_io_range(c, s, e)	do {} while (0)
C
Chris Mason 已提交
109
#endif
110 111 112 113 114 115 116 117 118 119 120 121 122

#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;
123
	unsigned long bio_flags;
124 125 126 127

	/* tells writepage not to lock the state bits for this range
	 * it still does the unlocking
	 */
128 129 130 131
	unsigned int extent_locked:1;

	/* tells the submit_bio code to use a WRITE_SYNC */
	unsigned int sync_io:1;
132 133
};

134
static noinline void flush_write_bio(void *data);
135 136 137
static inline struct btrfs_fs_info *
tree_fs_info(struct extent_io_tree *tree)
{
138 139
	if (!tree->mapping)
		return NULL;
140 141
	return btrfs_sb(tree->mapping->host->i_sb);
}
142

143 144
int __init extent_io_init(void)
{
D
David Sterba 已提交
145
	extent_state_cache = kmem_cache_create("btrfs_extent_state",
146 147
			sizeof(struct extent_state), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
148 149 150
	if (!extent_state_cache)
		return -ENOMEM;

D
David Sterba 已提交
151
	extent_buffer_cache = kmem_cache_create("btrfs_extent_buffer",
152 153
			sizeof(struct extent_buffer), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
154 155
	if (!extent_buffer_cache)
		goto free_state_cache;
156 157 158 159 160

	btrfs_bioset = bioset_create(BIO_POOL_SIZE,
				     offsetof(struct btrfs_io_bio, bio));
	if (!btrfs_bioset)
		goto free_buffer_cache;
161 162 163 164

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

165 166
	return 0;

167 168 169 170
free_bioset:
	bioset_free(btrfs_bioset);
	btrfs_bioset = NULL;

171 172 173 174
free_buffer_cache:
	kmem_cache_destroy(extent_buffer_cache);
	extent_buffer_cache = NULL;

175 176
free_state_cache:
	kmem_cache_destroy(extent_state_cache);
177
	extent_state_cache = NULL;
178 179 180 181 182
	return -ENOMEM;
}

void extent_io_exit(void)
{
183
	btrfs_leak_debug_check();
184 185 186 187 188 189

	/*
	 * Make sure all delayed rcu free are flushed before we
	 * destroy caches.
	 */
	rcu_barrier();
190 191 192 193
	if (extent_state_cache)
		kmem_cache_destroy(extent_state_cache);
	if (extent_buffer_cache)
		kmem_cache_destroy(extent_buffer_cache);
194 195
	if (btrfs_bioset)
		bioset_free(btrfs_bioset);
196 197 198
}

void extent_io_tree_init(struct extent_io_tree *tree,
199
			 struct address_space *mapping)
200
{
201
	tree->state = RB_ROOT;
202 203
	tree->ops = NULL;
	tree->dirty_bytes = 0;
204
	spin_lock_init(&tree->lock);
205 206 207
	tree->mapping = mapping;
}

208
static struct extent_state *alloc_extent_state(gfp_t mask)
209 210 211 212
{
	struct extent_state *state;

	state = kmem_cache_alloc(extent_state_cache, mask);
213
	if (!state)
214 215 216
		return state;
	state->state = 0;
	state->private = 0;
217
	RB_CLEAR_NODE(&state->rb_node);
218
	btrfs_leak_debug_add(&state->leak_list, &states);
219 220
	atomic_set(&state->refs, 1);
	init_waitqueue_head(&state->wq);
221
	trace_alloc_extent_state(state, mask, _RET_IP_);
222 223 224
	return state;
}

225
void free_extent_state(struct extent_state *state)
226 227 228 229
{
	if (!state)
		return;
	if (atomic_dec_and_test(&state->refs)) {
230
		WARN_ON(extent_state_in_tree(state));
231
		btrfs_leak_debug_del(&state->leak_list);
232
		trace_free_extent_state(state, _RET_IP_);
233 234 235 236
		kmem_cache_free(extent_state_cache, state);
	}
}

237 238 239
static struct rb_node *tree_insert(struct rb_root *root,
				   struct rb_node *search_start,
				   u64 offset,
240 241 242
				   struct rb_node *node,
				   struct rb_node ***p_in,
				   struct rb_node **parent_in)
243
{
244
	struct rb_node **p;
C
Chris Mason 已提交
245
	struct rb_node *parent = NULL;
246 247
	struct tree_entry *entry;

248 249 250 251 252 253
	if (p_in && parent_in) {
		p = *p_in;
		parent = *parent_in;
		goto do_insert;
	}

254
	p = search_start ? &search_start : &root->rb_node;
C
Chris Mason 已提交
255
	while (*p) {
256 257 258 259 260 261 262 263 264 265 266
		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;
	}

267
do_insert:
268 269 270 271 272
	rb_link_node(node, parent, p);
	rb_insert_color(node, root);
	return NULL;
}

273
static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset,
274 275 276 277
				      struct rb_node **prev_ret,
				      struct rb_node **next_ret,
				      struct rb_node ***p_ret,
				      struct rb_node **parent_ret)
278
{
279
	struct rb_root *root = &tree->state;
280
	struct rb_node **n = &root->rb_node;
281 282 283 284 285
	struct rb_node *prev = NULL;
	struct rb_node *orig_prev = NULL;
	struct tree_entry *entry;
	struct tree_entry *prev_entry = NULL;

286 287 288
	while (*n) {
		prev = *n;
		entry = rb_entry(prev, struct tree_entry, rb_node);
289 290 291
		prev_entry = entry;

		if (offset < entry->start)
292
			n = &(*n)->rb_left;
293
		else if (offset > entry->end)
294
			n = &(*n)->rb_right;
C
Chris Mason 已提交
295
		else
296
			return *n;
297 298
	}

299 300 301 302 303
	if (p_ret)
		*p_ret = n;
	if (parent_ret)
		*parent_ret = prev;

304 305
	if (prev_ret) {
		orig_prev = prev;
C
Chris Mason 已提交
306
		while (prev && offset > prev_entry->end) {
307 308 309 310 311 312 313 314 315
			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
316
		while (prev && offset < prev_entry->start) {
317 318 319 320 321 322 323 324
			prev = rb_prev(prev);
			prev_entry = rb_entry(prev, struct tree_entry, rb_node);
		}
		*next_ret = prev;
	}
	return NULL;
}

325 326 327 328 329
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)
330
{
331
	struct rb_node *prev = NULL;
332
	struct rb_node *ret;
333

334
	ret = __etree_search(tree, offset, &prev, NULL, p_ret, parent_ret);
C
Chris Mason 已提交
335
	if (!ret)
336 337 338 339
		return prev;
	return ret;
}

340 341 342 343 344 345
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
346 347 348 349 350 351 352 353
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);
}

354 355 356 357 358 359 360 361 362
/*
 * 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.
 */
363 364
static void merge_state(struct extent_io_tree *tree,
		        struct extent_state *state)
365 366 367 368
{
	struct extent_state *other;
	struct rb_node *other_node;

369
	if (state->state & (EXTENT_IOBITS | EXTENT_BOUNDARY))
370
		return;
371 372 373 374 375 376

	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
377
			merge_cb(tree, state, other);
378 379
			state->start = other->start;
			rb_erase(&other->rb_node, &tree->state);
380
			RB_CLEAR_NODE(&other->rb_node);
381 382 383 384 385 386 387 388
			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
389
			merge_cb(tree, state, other);
390 391
			state->end = other->end;
			rb_erase(&other->rb_node, &tree->state);
392
			RB_CLEAR_NODE(&other->rb_node);
393
			free_extent_state(other);
394 395 396 397
		}
	}
}

398
static void set_state_cb(struct extent_io_tree *tree,
399
			 struct extent_state *state, unsigned *bits)
400
{
401 402
	if (tree->ops && tree->ops->set_bit_hook)
		tree->ops->set_bit_hook(tree->mapping->host, state, bits);
403 404 405
}

static void clear_state_cb(struct extent_io_tree *tree,
406
			   struct extent_state *state, unsigned *bits)
407
{
J
Josef Bacik 已提交
408 409
	if (tree->ops && tree->ops->clear_bit_hook)
		tree->ops->clear_bit_hook(tree->mapping->host, state, bits);
410 411
}

412
static void set_state_bits(struct extent_io_tree *tree,
413
			   struct extent_state *state, unsigned *bits);
414

415 416 417 418 419 420 421 422 423 424 425 426
/*
 * 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,
427 428
			struct rb_node ***p,
			struct rb_node **parent,
429
			unsigned *bits)
430 431 432
{
	struct rb_node *node;

J
Julia Lawall 已提交
433
	if (end < start)
434
		WARN(1, KERN_ERR "BTRFS: end < start %llu %llu\n",
435
		       end, start);
436 437
	state->start = start;
	state->end = end;
J
Josef Bacik 已提交
438

439 440
	set_state_bits(tree, state, bits);

441
	node = tree_insert(&tree->state, NULL, end, &state->rb_node, p, parent);
442 443 444
	if (node) {
		struct extent_state *found;
		found = rb_entry(node, struct extent_state, rb_node);
445
		printk(KERN_ERR "BTRFS: found node %llu %llu on insert of "
446 447
		       "%llu %llu\n",
		       found->start, found->end, start, end);
448 449 450 451 452 453
		return -EEXIST;
	}
	merge_state(tree, state);
	return 0;
}

454
static void split_cb(struct extent_io_tree *tree, struct extent_state *orig,
J
Josef Bacik 已提交
455 456 457
		     u64 split)
{
	if (tree->ops && tree->ops->split_extent_hook)
458
		tree->ops->split_extent_hook(tree->mapping->host, orig, split);
J
Josef Bacik 已提交
459 460
}

461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478
/*
 * 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
479 480 481

	split_cb(tree, orig, split);

482 483 484 485 486
	prealloc->start = orig->start;
	prealloc->end = split - 1;
	prealloc->state = orig->state;
	orig->start = split;

487 488
	node = tree_insert(&tree->state, &orig->rb_node, prealloc->end,
			   &prealloc->rb_node, NULL, NULL);
489 490 491 492 493 494 495
	if (node) {
		free_extent_state(prealloc);
		return -EEXIST;
	}
	return 0;
}

496 497 498 499 500 501 502 503 504
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;
}

505 506
/*
 * utility function to clear some bits in an extent state struct.
507
 * it will optionally wake up any one waiting on this state (wake == 1).
508 509 510 511
 *
 * If no bits are set on the state struct after clearing things, the
 * struct is freed and removed from the tree
 */
512 513
static struct extent_state *clear_state_bit(struct extent_io_tree *tree,
					    struct extent_state *state,
514
					    unsigned *bits, int wake)
515
{
516
	struct extent_state *next;
517
	unsigned bits_to_clear = *bits & ~EXTENT_CTLBITS;
518

519
	if ((bits_to_clear & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) {
520 521 522 523
		u64 range = state->end - state->start + 1;
		WARN_ON(range > tree->dirty_bytes);
		tree->dirty_bytes -= range;
	}
524
	clear_state_cb(tree, state, bits);
525
	state->state &= ~bits_to_clear;
526 527
	if (wake)
		wake_up(&state->wq);
528
	if (state->state == 0) {
529
		next = next_state(state);
530
		if (extent_state_in_tree(state)) {
531
			rb_erase(&state->rb_node, &tree->state);
532
			RB_CLEAR_NODE(&state->rb_node);
533 534 535 536 537 538
			free_extent_state(state);
		} else {
			WARN_ON(1);
		}
	} else {
		merge_state(tree, state);
539
		next = next_state(state);
540
	}
541
	return next;
542 543
}

544 545 546 547 548 549 550 551 552
static struct extent_state *
alloc_extent_state_atomic(struct extent_state *prealloc)
{
	if (!prealloc)
		prealloc = alloc_extent_state(GFP_ATOMIC);

	return prealloc;
}

553
static void extent_io_tree_panic(struct extent_io_tree *tree, int err)
554 555 556 557 558 559
{
	btrfs_panic(tree_fs_info(tree), err, "Locking error: "
		    "Extent tree was modified by another "
		    "thread while locked.");
}

560 561 562 563 564 565 566 567 568 569
/*
 * 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.
 *
570
 * This takes the tree lock, and returns 0 on success and < 0 on error.
571 572
 */
int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
573
		     unsigned bits, int wake, int delete,
574 575
		     struct extent_state **cached_state,
		     gfp_t mask)
576 577
{
	struct extent_state *state;
578
	struct extent_state *cached;
579 580
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
581
	u64 last_end;
582
	int err;
583
	int clear = 0;
584

585
	btrfs_debug_check_extent_io_range(tree, start, end);
586

587 588 589
	if (bits & EXTENT_DELALLOC)
		bits |= EXTENT_NORESERVE;

590 591 592 593
	if (delete)
		bits |= ~EXTENT_CTLBITS;
	bits |= EXTENT_FIRST_DELALLOC;

594 595
	if (bits & (EXTENT_IOBITS | EXTENT_BOUNDARY))
		clear = 1;
596 597
again:
	if (!prealloc && (mask & __GFP_WAIT)) {
598 599 600 601 602 603 604
		/*
		 * Don't care for allocation failure here because we might end
		 * up not needing the pre-allocated extent state at all, which
		 * is the case if we only have in the tree extent states that
		 * cover our input range and don't cover too any other range.
		 * If we end up needing a new extent state we allocate it later.
		 */
605 606 607
		prealloc = alloc_extent_state(mask);
	}

608
	spin_lock(&tree->lock);
609 610
	if (cached_state) {
		cached = *cached_state;
611 612 613 614 615 616

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

617 618
		if (cached && extent_state_in_tree(cached) &&
		    cached->start <= start && cached->end > start) {
619 620
			if (clear)
				atomic_dec(&cached->refs);
621
			state = cached;
622
			goto hit_next;
623
		}
624 625
		if (clear)
			free_extent_state(cached);
626
	}
627 628 629 630
	/*
	 * this search will find the extents that end after
	 * our range starts
	 */
631
	node = tree_search(tree, start);
632 633 634
	if (!node)
		goto out;
	state = rb_entry(node, struct extent_state, rb_node);
635
hit_next:
636 637 638
	if (state->start > end)
		goto out;
	WARN_ON(state->end < start);
639
	last_end = state->end;
640

641
	/* the state doesn't have the wanted bits, go ahead */
642 643
	if (!(state->state & bits)) {
		state = next_state(state);
644
		goto next;
645
	}
646

647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663
	/*
	 *     | ---- 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) {
664 665
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
666
		err = split_state(tree, state, prealloc, start);
667 668 669
		if (err)
			extent_io_tree_panic(tree, err);

670 671 672 673
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
674 675
			state = clear_state_bit(tree, state, &bits, wake);
			goto next;
676 677 678 679 680 681 682 683 684 685
		}
		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) {
686 687
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
688
		err = split_state(tree, state, prealloc, end + 1);
689 690 691
		if (err)
			extent_io_tree_panic(tree, err);

692 693
		if (wake)
			wake_up(&state->wq);
694

695
		clear_state_bit(tree, prealloc, &bits, wake);
J
Josef Bacik 已提交
696

697 698 699
		prealloc = NULL;
		goto out;
	}
700

701
	state = clear_state_bit(tree, state, &bits, wake);
702
next:
703 704 705
	if (last_end == (u64)-1)
		goto out;
	start = last_end + 1;
706
	if (start <= end && state && !need_resched())
707
		goto hit_next;
708 709 710
	goto search_again;

out:
711
	spin_unlock(&tree->lock);
712 713 714
	if (prealloc)
		free_extent_state(prealloc);

715
	return 0;
716 717 718 719

search_again:
	if (start > end)
		goto out;
720
	spin_unlock(&tree->lock);
721 722 723 724 725
	if (mask & __GFP_WAIT)
		cond_resched();
	goto again;
}

726 727
static void wait_on_state(struct extent_io_tree *tree,
			  struct extent_state *state)
728 729
		__releases(tree->lock)
		__acquires(tree->lock)
730 731 732
{
	DEFINE_WAIT(wait);
	prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE);
733
	spin_unlock(&tree->lock);
734
	schedule();
735
	spin_lock(&tree->lock);
736 737 738 739 740 741 742 743
	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
 */
744 745
static void wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
			    unsigned long bits)
746 747 748 749
{
	struct extent_state *state;
	struct rb_node *node;

750
	btrfs_debug_check_extent_io_range(tree, start, end);
751

752
	spin_lock(&tree->lock);
753 754 755 756 757 758
again:
	while (1) {
		/*
		 * this search will find all the extents that end after
		 * our range starts
		 */
759
		node = tree_search(tree, start);
760
process_node:
761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780
		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;

781 782 783 784
		if (!cond_resched_lock(&tree->lock)) {
			node = rb_next(node);
			goto process_node;
		}
785 786
	}
out:
787
	spin_unlock(&tree->lock);
788 789
}

790
static void set_state_bits(struct extent_io_tree *tree,
791
			   struct extent_state *state,
792
			   unsigned *bits)
793
{
794
	unsigned bits_to_set = *bits & ~EXTENT_CTLBITS;
J
Josef Bacik 已提交
795

796
	set_state_cb(tree, state, bits);
797
	if ((bits_to_set & EXTENT_DIRTY) && !(state->state & EXTENT_DIRTY)) {
798 799 800
		u64 range = state->end - state->start + 1;
		tree->dirty_bytes += range;
	}
801
	state->state |= bits_to_set;
802 803
}

804 805
static void cache_state_if_flags(struct extent_state *state,
				 struct extent_state **cached_ptr,
806
				 unsigned flags)
807 808
{
	if (cached_ptr && !(*cached_ptr)) {
809
		if (!flags || (state->state & flags)) {
810 811 812 813 814 815
			*cached_ptr = state;
			atomic_inc(&state->refs);
		}
	}
}

816 817 818 819 820 821 822
static void cache_state(struct extent_state *state,
			struct extent_state **cached_ptr)
{
	return cache_state_if_flags(state, cached_ptr,
				    EXTENT_IOBITS | EXTENT_BOUNDARY);
}

823
/*
824 825
 * 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.
826
 *
827 828 829
 * 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.
830
 *
831
 * [start, end] is inclusive This takes the tree lock.
832
 */
833

J
Jeff Mahoney 已提交
834 835
static int __must_check
__set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
836
		 unsigned bits, unsigned exclusive_bits,
837 838
		 u64 *failed_start, struct extent_state **cached_state,
		 gfp_t mask)
839 840 841 842
{
	struct extent_state *state;
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
843 844
	struct rb_node **p;
	struct rb_node *parent;
845 846 847
	int err = 0;
	u64 last_start;
	u64 last_end;
848

849
	btrfs_debug_check_extent_io_range(tree, start, end);
850

851
	bits |= EXTENT_FIRST_DELALLOC;
852 853 854
again:
	if (!prealloc && (mask & __GFP_WAIT)) {
		prealloc = alloc_extent_state(mask);
855
		BUG_ON(!prealloc);
856 857
	}

858
	spin_lock(&tree->lock);
859 860
	if (cached_state && *cached_state) {
		state = *cached_state;
861
		if (state->start <= start && state->end > start &&
862
		    extent_state_in_tree(state)) {
863 864 865 866
			node = &state->rb_node;
			goto hit_next;
		}
	}
867 868 869 870
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
871
	node = tree_search_for_insert(tree, start, &p, &parent);
872
	if (!node) {
873 874
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
875 876
		err = insert_state(tree, prealloc, start, end,
				   &p, &parent, &bits);
877 878 879
		if (err)
			extent_io_tree_panic(tree, err);

880
		cache_state(prealloc, cached_state);
881 882 883 884
		prealloc = NULL;
		goto out;
	}
	state = rb_entry(node, struct extent_state, rb_node);
C
Chris Mason 已提交
885
hit_next:
886 887 888 889 890 891 892 893 894 895
	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) {
896
		if (state->state & exclusive_bits) {
897 898 899 900
			*failed_start = state->start;
			err = -EEXIST;
			goto out;
		}
901

902
		set_state_bits(tree, state, &bits);
903
		cache_state(state, cached_state);
904
		merge_state(tree, state);
905 906 907
		if (last_end == (u64)-1)
			goto out;
		start = last_end + 1;
908 909 910 911
		state = next_state(state);
		if (start < end && state && state->start == start &&
		    !need_resched())
			goto hit_next;
912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931
		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) {
932
		if (state->state & exclusive_bits) {
933 934 935 936
			*failed_start = start;
			err = -EEXIST;
			goto out;
		}
937 938 939

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
940
		err = split_state(tree, state, prealloc, start);
941 942 943
		if (err)
			extent_io_tree_panic(tree, err);

944 945 946 947
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
948
			set_state_bits(tree, state, &bits);
949
			cache_state(state, cached_state);
950
			merge_state(tree, state);
951 952 953
			if (last_end == (u64)-1)
				goto out;
			start = last_end + 1;
954 955 956 957
			state = next_state(state);
			if (start < end && state && state->start == start &&
			    !need_resched())
				goto hit_next;
958 959 960 961 962 963 964 965 966 967 968 969 970 971 972
		}
		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
973
			this_end = last_start - 1;
974 975 976

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
977 978 979 980 981

		/*
		 * Avoid to free 'prealloc' if it can be merged with
		 * the later extent.
		 */
982
		err = insert_state(tree, prealloc, start, this_end,
983
				   NULL, NULL, &bits);
984 985 986
		if (err)
			extent_io_tree_panic(tree, err);

J
Josef Bacik 已提交
987 988
		cache_state(prealloc, cached_state);
		prealloc = NULL;
989 990 991 992 993 994 995 996 997 998
		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) {
999
		if (state->state & exclusive_bits) {
1000 1001 1002 1003
			*failed_start = start;
			err = -EEXIST;
			goto out;
		}
1004 1005 1006

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
1007
		err = split_state(tree, state, prealloc, end + 1);
1008 1009
		if (err)
			extent_io_tree_panic(tree, err);
1010

1011
		set_state_bits(tree, prealloc, &bits);
1012
		cache_state(prealloc, cached_state);
1013 1014 1015 1016 1017 1018 1019 1020
		merge_state(tree, prealloc);
		prealloc = NULL;
		goto out;
	}

	goto search_again;

out:
1021
	spin_unlock(&tree->lock);
1022 1023 1024 1025 1026 1027 1028 1029
	if (prealloc)
		free_extent_state(prealloc);

	return err;

search_again:
	if (start > end)
		goto out;
1030
	spin_unlock(&tree->lock);
1031 1032 1033 1034 1035
	if (mask & __GFP_WAIT)
		cond_resched();
	goto again;
}

1036
int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
1037
		   unsigned bits, u64 * failed_start,
1038
		   struct extent_state **cached_state, gfp_t mask)
J
Jeff Mahoney 已提交
1039 1040 1041 1042 1043 1044
{
	return __set_extent_bit(tree, start, end, bits, 0, failed_start,
				cached_state, mask);
}


J
Josef Bacik 已提交
1045
/**
L
Liu Bo 已提交
1046 1047
 * convert_extent_bit - convert all bits in a given range from one bit to
 * 			another
J
Josef Bacik 已提交
1048 1049 1050 1051 1052
 * @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
1053
 * @cached_state:	state that we're going to cache
J
Josef Bacik 已提交
1054 1055 1056 1057 1058 1059 1060 1061 1062
 * @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,
1063
		       unsigned bits, unsigned clear_bits,
1064
		       struct extent_state **cached_state, gfp_t mask)
J
Josef Bacik 已提交
1065 1066 1067 1068
{
	struct extent_state *state;
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
1069 1070
	struct rb_node **p;
	struct rb_node *parent;
J
Josef Bacik 已提交
1071 1072 1073
	int err = 0;
	u64 last_start;
	u64 last_end;
1074
	bool first_iteration = true;
J
Josef Bacik 已提交
1075

1076
	btrfs_debug_check_extent_io_range(tree, start, end);
1077

J
Josef Bacik 已提交
1078 1079
again:
	if (!prealloc && (mask & __GFP_WAIT)) {
1080 1081 1082 1083 1084 1085 1086
		/*
		 * Best effort, don't worry if extent state allocation fails
		 * here for the first iteration. We might have a cached state
		 * that matches exactly the target range, in which case no
		 * extent state allocations are needed. We'll only know this
		 * after locking the tree.
		 */
J
Josef Bacik 已提交
1087
		prealloc = alloc_extent_state(mask);
1088
		if (!prealloc && !first_iteration)
J
Josef Bacik 已提交
1089 1090 1091 1092
			return -ENOMEM;
	}

	spin_lock(&tree->lock);
1093 1094 1095
	if (cached_state && *cached_state) {
		state = *cached_state;
		if (state->start <= start && state->end > start &&
1096
		    extent_state_in_tree(state)) {
1097 1098 1099 1100 1101
			node = &state->rb_node;
			goto hit_next;
		}
	}

J
Josef Bacik 已提交
1102 1103 1104 1105
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1106
	node = tree_search_for_insert(tree, start, &p, &parent);
J
Josef Bacik 已提交
1107 1108
	if (!node) {
		prealloc = alloc_extent_state_atomic(prealloc);
1109 1110 1111 1112
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
1113 1114
		err = insert_state(tree, prealloc, start, end,
				   &p, &parent, &bits);
1115 1116
		if (err)
			extent_io_tree_panic(tree, err);
1117 1118
		cache_state(prealloc, cached_state);
		prealloc = NULL;
J
Josef Bacik 已提交
1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133
		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);
1134
		cache_state(state, cached_state);
1135
		state = clear_state_bit(tree, state, &clear_bits, 0);
J
Josef Bacik 已提交
1136 1137 1138
		if (last_end == (u64)-1)
			goto out;
		start = last_end + 1;
1139 1140 1141
		if (start < end && state && state->start == start &&
		    !need_resched())
			goto hit_next;
J
Josef Bacik 已提交
1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162
		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);
1163 1164 1165 1166
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1167
		err = split_state(tree, state, prealloc, start);
1168 1169
		if (err)
			extent_io_tree_panic(tree, err);
J
Josef Bacik 已提交
1170 1171 1172 1173 1174
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
			set_state_bits(tree, state, &bits);
1175
			cache_state(state, cached_state);
1176
			state = clear_state_bit(tree, state, &clear_bits, 0);
J
Josef Bacik 已提交
1177 1178 1179
			if (last_end == (u64)-1)
				goto out;
			start = last_end + 1;
1180 1181 1182
			if (start < end && state && state->start == start &&
			    !need_resched())
				goto hit_next;
J
Josef Bacik 已提交
1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200
		}
		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);
1201 1202 1203 1204
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1205 1206 1207 1208 1209 1210

		/*
		 * Avoid to free 'prealloc' if it can be merged with
		 * the later extent.
		 */
		err = insert_state(tree, prealloc, start, this_end,
1211
				   NULL, NULL, &bits);
1212 1213
		if (err)
			extent_io_tree_panic(tree, err);
1214
		cache_state(prealloc, cached_state);
J
Josef Bacik 已提交
1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226
		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);
1227 1228 1229 1230
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1231 1232

		err = split_state(tree, state, prealloc, end + 1);
1233 1234
		if (err)
			extent_io_tree_panic(tree, err);
J
Josef Bacik 已提交
1235 1236

		set_state_bits(tree, prealloc, &bits);
1237
		cache_state(prealloc, cached_state);
J
Josef Bacik 已提交
1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257
		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();
1258
	first_iteration = false;
J
Josef Bacik 已提交
1259 1260 1261
	goto again;
}

1262 1263 1264 1265
/* 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
1266
	return set_extent_bit(tree, start, end, EXTENT_DIRTY, NULL,
1267
			      NULL, mask);
1268 1269 1270
}

int set_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1271
		    unsigned bits, gfp_t mask)
1272
{
J
Jeff Mahoney 已提交
1273
	return set_extent_bit(tree, start, end, bits, NULL,
1274
			      NULL, mask);
1275 1276 1277
}

int clear_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1278
		      unsigned bits, gfp_t mask)
1279
{
1280 1281 1282 1283 1284 1285
	int wake = 0;

	if (bits & EXTENT_LOCKED)
		wake = 1;

	return clear_extent_bit(tree, start, end, bits, wake, 0, NULL, mask);
1286 1287 1288
}

int set_extent_delalloc(struct extent_io_tree *tree, u64 start, u64 end,
1289
			struct extent_state **cached_state, gfp_t mask)
1290 1291
{
	return set_extent_bit(tree, start, end,
1292
			      EXTENT_DELALLOC | EXTENT_UPTODATE,
J
Jeff Mahoney 已提交
1293
			      NULL, cached_state, mask);
1294 1295
}

1296 1297 1298 1299 1300 1301 1302 1303
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);
}

1304 1305 1306 1307
int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
		       gfp_t mask)
{
	return clear_extent_bit(tree, start, end,
1308
				EXTENT_DIRTY | EXTENT_DELALLOC |
1309
				EXTENT_DO_ACCOUNTING, 0, 0, NULL, mask);
1310 1311 1312 1313 1314
}

int set_extent_new(struct extent_io_tree *tree, u64 start, u64 end,
		     gfp_t mask)
{
J
Jeff Mahoney 已提交
1315
	return set_extent_bit(tree, start, end, EXTENT_NEW, NULL,
1316
			      NULL, mask);
1317 1318 1319
}

int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end,
1320
			struct extent_state **cached_state, gfp_t mask)
1321
{
L
Liu Bo 已提交
1322
	return set_extent_bit(tree, start, end, EXTENT_UPTODATE, NULL,
J
Jeff Mahoney 已提交
1323
			      cached_state, mask);
1324 1325
}

1326 1327
int clear_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end,
			  struct extent_state **cached_state, gfp_t mask)
1328
{
1329
	return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0,
1330
				cached_state, mask);
1331 1332
}

C
Chris Mason 已提交
1333 1334 1335 1336
/*
 * either insert or lock state struct between start and end use mask to tell
 * us if waiting is desired.
 */
1337
int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1338
		     unsigned bits, struct extent_state **cached_state)
1339 1340 1341
{
	int err;
	u64 failed_start;
1342

1343
	while (1) {
J
Jeff Mahoney 已提交
1344 1345 1346
		err = __set_extent_bit(tree, start, end, EXTENT_LOCKED | bits,
				       EXTENT_LOCKED, &failed_start,
				       cached_state, GFP_NOFS);
1347
		if (err == -EEXIST) {
1348 1349
			wait_extent_bit(tree, failed_start, end, EXTENT_LOCKED);
			start = failed_start;
1350
		} else
1351 1352 1353 1354 1355 1356
			break;
		WARN_ON(start > end);
	}
	return err;
}

1357
int lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
1358
{
1359
	return lock_extent_bits(tree, start, end, 0, NULL);
1360 1361
}

1362
int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
1363 1364 1365 1366
{
	int err;
	u64 failed_start;

J
Jeff Mahoney 已提交
1367 1368
	err = __set_extent_bit(tree, start, end, EXTENT_LOCKED, EXTENT_LOCKED,
			       &failed_start, NULL, GFP_NOFS);
Y
Yan Zheng 已提交
1369 1370 1371
	if (err == -EEXIST) {
		if (failed_start > start)
			clear_extent_bit(tree, start, failed_start - 1,
1372
					 EXTENT_LOCKED, 1, 0, NULL, GFP_NOFS);
1373
		return 0;
Y
Yan Zheng 已提交
1374
	}
1375 1376 1377
	return 1;
}

1378 1379 1380 1381 1382 1383 1384
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);
}

1385
int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end)
1386
{
1387
	return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL,
1388
				GFP_NOFS);
1389 1390
}

1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416
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 */
		__set_page_dirty_nobuffers(page);
1417
		account_page_redirty(page);
1418 1419 1420 1421 1422 1423
		page_cache_release(page);
		index++;
	}
	return 0;
}

1424 1425 1426
/*
 * helper function to set both pages and extents in the tree writeback
 */
1427
static int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end)
1428 1429 1430 1431 1432 1433 1434
{
	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);
1435
		BUG_ON(!page); /* Pages should be in the extent_io_tree */
1436 1437 1438 1439 1440 1441 1442
		set_page_writeback(page);
		page_cache_release(page);
		index++;
	}
	return 0;
}

C
Chris Mason 已提交
1443 1444 1445 1446
/* 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'
 */
1447 1448
static struct extent_state *
find_first_extent_bit_state(struct extent_io_tree *tree,
1449
			    u64 start, unsigned bits)
C
Chris Mason 已提交
1450 1451 1452 1453 1454 1455 1456 1457 1458
{
	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
1459
	if (!node)
C
Chris Mason 已提交
1460 1461
		goto out;

C
Chris Mason 已提交
1462
	while (1) {
C
Chris Mason 已提交
1463
		state = rb_entry(node, struct extent_state, rb_node);
C
Chris Mason 已提交
1464
		if (state->end >= start && (state->state & bits))
C
Chris Mason 已提交
1465
			return state;
C
Chris Mason 已提交
1466

C
Chris Mason 已提交
1467 1468 1469 1470 1471 1472 1473 1474
		node = rb_next(node);
		if (!node)
			break;
	}
out:
	return NULL;
}

1475 1476 1477 1478 1479
/*
 * 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.
 *
1480
 * If nothing was found, 1 is returned. If found something, return 0.
1481 1482
 */
int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
1483
			  u64 *start_ret, u64 *end_ret, unsigned bits,
1484
			  struct extent_state **cached_state)
1485 1486
{
	struct extent_state *state;
1487
	struct rb_node *n;
1488 1489 1490
	int ret = 1;

	spin_lock(&tree->lock);
1491 1492
	if (cached_state && *cached_state) {
		state = *cached_state;
1493
		if (state->end == start - 1 && extent_state_in_tree(state)) {
1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509
			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;
	}

1510
	state = find_first_extent_bit_state(tree, start, bits);
1511
got_it:
1512
	if (state) {
1513
		cache_state_if_flags(state, cached_state, 0);
1514 1515 1516 1517
		*start_ret = state->start;
		*end_ret = state->end;
		ret = 0;
	}
1518
out:
1519 1520 1521 1522
	spin_unlock(&tree->lock);
	return ret;
}

C
Chris Mason 已提交
1523 1524 1525 1526 1527 1528
/*
 * 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
1529
static noinline u64 find_delalloc_range(struct extent_io_tree *tree,
1530 1531
					u64 *start, u64 *end, u64 max_bytes,
					struct extent_state **cached_state)
1532 1533 1534 1535 1536 1537 1538
{
	struct rb_node *node;
	struct extent_state *state;
	u64 cur_start = *start;
	u64 found = 0;
	u64 total_bytes = 0;

1539
	spin_lock(&tree->lock);
C
Chris Mason 已提交
1540

1541 1542 1543 1544
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1545
	node = tree_search(tree, cur_start);
1546
	if (!node) {
1547 1548
		if (!found)
			*end = (u64)-1;
1549 1550 1551
		goto out;
	}

C
Chris Mason 已提交
1552
	while (1) {
1553
		state = rb_entry(node, struct extent_state, rb_node);
1554 1555
		if (found && (state->start != cur_start ||
			      (state->state & EXTENT_BOUNDARY))) {
1556 1557 1558 1559 1560 1561 1562
			goto out;
		}
		if (!(state->state & EXTENT_DELALLOC)) {
			if (!found)
				*end = state->end;
			goto out;
		}
1563
		if (!found) {
1564
			*start = state->start;
1565 1566 1567
			*cached_state = state;
			atomic_inc(&state->refs);
		}
1568 1569 1570 1571 1572
		found++;
		*end = state->end;
		cur_start = state->end + 1;
		node = rb_next(node);
		total_bytes += state->end - state->start + 1;
1573
		if (total_bytes >= max_bytes)
1574 1575
			break;
		if (!node)
1576 1577 1578
			break;
	}
out:
1579
	spin_unlock(&tree->lock);
1580 1581 1582
	return found;
}

1583 1584 1585
static noinline void __unlock_for_delalloc(struct inode *inode,
					   struct page *locked_page,
					   u64 start, u64 end)
C
Chris Mason 已提交
1586 1587 1588 1589 1590 1591 1592 1593 1594
{
	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)
1595
		return;
C
Chris Mason 已提交
1596

C
Chris Mason 已提交
1597
	while (nr_pages > 0) {
C
Chris Mason 已提交
1598
		ret = find_get_pages_contig(inode->i_mapping, index,
1599 1600
				     min_t(unsigned long, nr_pages,
				     ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631
		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
1632
	while (nrpages > 0) {
C
Chris Mason 已提交
1633
		ret = find_get_pages_contig(inode->i_mapping, index,
1634 1635
				     min_t(unsigned long,
				     nrpages, ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1636 1637 1638 1639 1640 1641 1642 1643 1644 1645
		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
			 */
1646
			if (pages[i] != locked_page) {
C
Chris Mason 已提交
1647
				lock_page(pages[i]);
1648 1649
				if (!PageDirty(pages[i]) ||
				    pages[i]->mapping != inode->i_mapping) {
1650 1651 1652 1653 1654 1655
					ret = -EAGAIN;
					unlock_page(pages[i]);
					page_cache_release(pages[i]);
					goto done;
				}
			}
C
Chris Mason 已提交
1656
			page_cache_release(pages[i]);
1657
			pages_locked++;
C
Chris Mason 已提交
1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679
		}
		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
 */
1680 1681 1682 1683
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
1684 1685 1686 1687
{
	u64 delalloc_start;
	u64 delalloc_end;
	u64 found;
1688
	struct extent_state *cached_state = NULL;
C
Chris Mason 已提交
1689 1690 1691 1692 1693 1694 1695 1696
	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,
1697
				    max_bytes, &cached_state);
C
Chris Mason 已提交
1698
	if (!found || delalloc_end <= *start) {
C
Chris Mason 已提交
1699 1700
		*start = delalloc_start;
		*end = delalloc_end;
1701
		free_extent_state(cached_state);
L
Liu Bo 已提交
1702
		return 0;
C
Chris Mason 已提交
1703 1704
	}

C
Chris Mason 已提交
1705 1706 1707 1708 1709
	/*
	 * 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
1710
	if (delalloc_start < *start)
C
Chris Mason 已提交
1711 1712
		delalloc_start = *start;

C
Chris Mason 已提交
1713 1714 1715
	/*
	 * make sure to limit the number of pages we try to lock down
	 */
1716 1717
	if (delalloc_end + 1 - delalloc_start > max_bytes)
		delalloc_end = delalloc_start + max_bytes - 1;
C
Chris Mason 已提交
1718

C
Chris Mason 已提交
1719 1720 1721 1722 1723 1724 1725
	/* 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
		 */
1726
		free_extent_state(cached_state);
1727
		cached_state = NULL;
C
Chris Mason 已提交
1728
		if (!loops) {
1729
			max_bytes = PAGE_CACHE_SIZE;
C
Chris Mason 已提交
1730 1731 1732 1733 1734 1735 1736
			loops = 1;
			goto again;
		} else {
			found = 0;
			goto out_failed;
		}
	}
1737
	BUG_ON(ret); /* Only valid values are 0 and -EAGAIN */
C
Chris Mason 已提交
1738 1739

	/* step three, lock the state bits for the whole range */
1740
	lock_extent_bits(tree, delalloc_start, delalloc_end, 0, &cached_state);
C
Chris Mason 已提交
1741 1742 1743

	/* then test to make sure it is all still delalloc */
	ret = test_range_bit(tree, delalloc_start, delalloc_end,
1744
			     EXTENT_DELALLOC, 1, cached_state);
C
Chris Mason 已提交
1745
	if (!ret) {
1746 1747
		unlock_extent_cached(tree, delalloc_start, delalloc_end,
				     &cached_state, GFP_NOFS);
C
Chris Mason 已提交
1748 1749 1750 1751 1752
		__unlock_for_delalloc(inode, locked_page,
			      delalloc_start, delalloc_end);
		cond_resched();
		goto again;
	}
1753
	free_extent_state(cached_state);
C
Chris Mason 已提交
1754 1755 1756 1757 1758 1759
	*start = delalloc_start;
	*end = delalloc_end;
out_failed:
	return found;
}

1760 1761
int extent_clear_unlock_delalloc(struct inode *inode, u64 start, u64 end,
				 struct page *locked_page,
1762
				 unsigned clear_bits,
1763
				 unsigned long page_ops)
C
Chris Mason 已提交
1764
{
1765
	struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
C
Chris Mason 已提交
1766 1767 1768 1769 1770 1771
	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;
1772

1773
	clear_extent_bit(tree, start, end, clear_bits, 1, 0, NULL, GFP_NOFS);
1774
	if (page_ops == 0)
1775
		return 0;
C
Chris Mason 已提交
1776

1777 1778 1779
	if ((page_ops & PAGE_SET_ERROR) && nr_pages > 0)
		mapping_set_error(inode->i_mapping, -EIO);

C
Chris Mason 已提交
1780
	while (nr_pages > 0) {
C
Chris Mason 已提交
1781
		ret = find_get_pages_contig(inode->i_mapping, index,
1782 1783
				     min_t(unsigned long,
				     nr_pages, ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1784
		for (i = 0; i < ret; i++) {
1785

1786
			if (page_ops & PAGE_SET_PRIVATE2)
1787 1788
				SetPagePrivate2(pages[i]);

C
Chris Mason 已提交
1789 1790 1791 1792
			if (pages[i] == locked_page) {
				page_cache_release(pages[i]);
				continue;
			}
1793
			if (page_ops & PAGE_CLEAR_DIRTY)
C
Chris Mason 已提交
1794
				clear_page_dirty_for_io(pages[i]);
1795
			if (page_ops & PAGE_SET_WRITEBACK)
C
Chris Mason 已提交
1796
				set_page_writeback(pages[i]);
1797 1798
			if (page_ops & PAGE_SET_ERROR)
				SetPageError(pages[i]);
1799
			if (page_ops & PAGE_END_WRITEBACK)
C
Chris Mason 已提交
1800
				end_page_writeback(pages[i]);
1801
			if (page_ops & PAGE_UNLOCK)
1802
				unlock_page(pages[i]);
C
Chris Mason 已提交
1803 1804 1805 1806 1807 1808 1809 1810 1811
			page_cache_release(pages[i]);
		}
		nr_pages -= ret;
		index += ret;
		cond_resched();
	}
	return 0;
}

C
Chris Mason 已提交
1812 1813 1814 1815 1816
/*
 * 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.
 */
1817 1818
u64 count_range_bits(struct extent_io_tree *tree,
		     u64 *start, u64 search_end, u64 max_bytes,
1819
		     unsigned bits, int contig)
1820 1821 1822 1823 1824
{
	struct rb_node *node;
	struct extent_state *state;
	u64 cur_start = *start;
	u64 total_bytes = 0;
1825
	u64 last = 0;
1826 1827
	int found = 0;

1828
	if (WARN_ON(search_end <= cur_start))
1829 1830
		return 0;

1831
	spin_lock(&tree->lock);
1832 1833 1834 1835 1836 1837 1838 1839
	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.
	 */
1840
	node = tree_search(tree, cur_start);
C
Chris Mason 已提交
1841
	if (!node)
1842 1843
		goto out;

C
Chris Mason 已提交
1844
	while (1) {
1845 1846 1847
		state = rb_entry(node, struct extent_state, rb_node);
		if (state->start > search_end)
			break;
1848 1849 1850
		if (contig && found && state->start > last + 1)
			break;
		if (state->end >= cur_start && (state->state & bits) == bits) {
1851 1852 1853 1854 1855
			total_bytes += min(search_end, state->end) + 1 -
				       max(cur_start, state->start);
			if (total_bytes >= max_bytes)
				break;
			if (!found) {
1856
				*start = max(cur_start, state->start);
1857 1858
				found = 1;
			}
1859 1860 1861
			last = state->end;
		} else if (contig && found) {
			break;
1862 1863 1864 1865 1866 1867
		}
		node = rb_next(node);
		if (!node)
			break;
	}
out:
1868
	spin_unlock(&tree->lock);
1869 1870
	return total_bytes;
}
1871

C
Chris Mason 已提交
1872 1873 1874 1875
/*
 * set the private field for a given byte offset in the tree.  If there isn't
 * an extent_state there already, this does nothing.
 */
1876
static int set_state_private(struct extent_io_tree *tree, u64 start, u64 private)
1877 1878 1879 1880 1881
{
	struct rb_node *node;
	struct extent_state *state;
	int ret = 0;

1882
	spin_lock(&tree->lock);
1883 1884 1885 1886
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1887
	node = tree_search(tree, start);
1888
	if (!node) {
1889 1890 1891 1892 1893 1894 1895 1896 1897 1898
		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:
1899
	spin_unlock(&tree->lock);
1900 1901 1902 1903 1904 1905 1906 1907 1908
	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;

1909
	spin_lock(&tree->lock);
1910 1911 1912 1913
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1914
	node = tree_search(tree, start);
1915
	if (!node) {
1916 1917 1918 1919 1920 1921 1922 1923 1924 1925
		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:
1926
	spin_unlock(&tree->lock);
1927 1928 1929 1930 1931
	return ret;
}

/*
 * searches a range in the state tree for a given mask.
1932
 * If 'filled' == 1, this returns 1 only if every extent in the tree
1933 1934 1935 1936
 * 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,
1937
		   unsigned bits, int filled, struct extent_state *cached)
1938 1939 1940 1941 1942
{
	struct extent_state *state = NULL;
	struct rb_node *node;
	int bitset = 0;

1943
	spin_lock(&tree->lock);
1944
	if (cached && extent_state_in_tree(cached) && cached->start <= start &&
1945
	    cached->end > start)
1946 1947 1948
		node = &cached->rb_node;
	else
		node = tree_search(tree, start);
1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967
	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;
		}
1968 1969 1970 1971

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

1972 1973 1974 1975 1976 1977 1978 1979 1980 1981
		start = state->end + 1;
		if (start > end)
			break;
		node = rb_next(node);
		if (!node) {
			if (filled)
				bitset = 0;
			break;
		}
	}
1982
	spin_unlock(&tree->lock);
1983 1984 1985 1986 1987 1988 1989
	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
 */
1990
static void check_page_uptodate(struct extent_io_tree *tree, struct page *page)
1991
{
M
Miao Xie 已提交
1992
	u64 start = page_offset(page);
1993
	u64 end = start + PAGE_CACHE_SIZE - 1;
1994
	if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1, NULL))
1995 1996 1997
		SetPageUptodate(page);
}

1998
int free_io_failure(struct inode *inode, struct io_failure_record *rec)
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
{
	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
2011 2012 2013 2014 2015
	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;
2016 2017 2018 2019 2020 2021 2022 2023 2024 2025

	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.
2026
 * to avoid any synchronization issues, wait for the data after writing, which
2027 2028 2029 2030
 * 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.
 */
2031 2032
int repair_io_failure(struct inode *inode, u64 start, u64 length, u64 logical,
		      struct page *page, unsigned int pg_offset, int mirror_num)
2033
{
2034
	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2035 2036 2037 2038 2039
	struct bio *bio;
	struct btrfs_device *dev;
	u64 map_length = 0;
	u64 sector;
	struct btrfs_bio *bbio = NULL;
D
David Woodhouse 已提交
2040
	struct btrfs_mapping_tree *map_tree = &fs_info->mapping_tree;
2041 2042
	int ret;

2043
	ASSERT(!(fs_info->sb->s_flags & MS_RDONLY));
2044 2045
	BUG_ON(!mirror_num);

D
David Woodhouse 已提交
2046 2047 2048 2049
	/* we can't repair anything in raid56 yet */
	if (btrfs_is_parity_mirror(map_tree, logical, length, mirror_num))
		return 0;

2050
	bio = btrfs_io_bio_alloc(GFP_NOFS, 1);
2051 2052
	if (!bio)
		return -EIO;
2053
	bio->bi_iter.bi_size = 0;
2054 2055
	map_length = length;

2056
	ret = btrfs_map_block(fs_info, WRITE, logical,
2057 2058 2059 2060 2061 2062 2063
			      &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;
2064
	bio->bi_iter.bi_sector = sector;
2065
	dev = bbio->stripes[mirror_num-1].dev;
2066
	btrfs_put_bbio(bbio);
2067 2068 2069 2070 2071
	if (!dev || !dev->bdev || !dev->writeable) {
		bio_put(bio);
		return -EIO;
	}
	bio->bi_bdev = dev->bdev;
2072
	bio_add_page(bio, page, length, pg_offset);
2073

2074
	if (btrfsic_submit_bio_wait(WRITE_SYNC, bio)) {
2075 2076
		/* try to remap that extent elsewhere? */
		bio_put(bio);
2077
		btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_WRITE_ERRS);
2078 2079 2080
		return -EIO;
	}

2081 2082
	btrfs_info_rl_in_rcu(fs_info,
		"read error corrected: ino %llu off %llu (dev %s sector %llu)",
2083 2084
				  btrfs_ino(inode), start,
				  rcu_str_deref(dev->name), sector);
2085 2086 2087 2088
	bio_put(bio);
	return 0;
}

2089 2090 2091 2092 2093
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);
2094
	int ret = 0;
2095

2096 2097 2098
	if (root->fs_info->sb->s_flags & MS_RDONLY)
		return -EROFS;

2099
	for (i = 0; i < num_pages; i++) {
2100
		struct page *p = eb->pages[i];
2101 2102 2103 2104

		ret = repair_io_failure(root->fs_info->btree_inode, start,
					PAGE_CACHE_SIZE, start, p,
					start - page_offset(p), mirror_num);
2105 2106 2107 2108 2109 2110 2111 2112
		if (ret)
			break;
		start += PAGE_CACHE_SIZE;
	}

	return ret;
}

2113 2114 2115 2116
/*
 * 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
 */
2117 2118
int clean_io_failure(struct inode *inode, u64 start, struct page *page,
		     unsigned int pg_offset)
2119 2120 2121 2122
{
	u64 private;
	u64 private_failure;
	struct io_failure_record *failrec;
2123
	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147
	struct extent_state *state;
	int num_copies;
	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);
		goto out;
	}
2148 2149
	if (fs_info->sb->s_flags & MS_RDONLY)
		goto out;
2150 2151 2152 2153 2154 2155 2156

	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);

2157 2158
	if (state && state->start <= failrec->start &&
	    state->end >= failrec->start + failrec->len - 1) {
2159 2160
		num_copies = btrfs_num_copies(fs_info, failrec->logical,
					      failrec->len);
2161
		if (num_copies > 1)  {
2162
			repair_io_failure(inode, start, failrec->len,
2163
					  failrec->logical, page,
2164
					  pg_offset, failrec->failed_mirror);
2165 2166 2167 2168
		}
	}

out:
2169
	free_io_failure(inode, failrec);
2170

2171
	return 0;
2172 2173
}

2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198
/*
 * Can be called when
 * - hold extent lock
 * - under ordered extent
 * - the inode is freeing
 */
void btrfs_free_io_failure_record(struct inode *inode, u64 start, u64 end)
{
	struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
	struct io_failure_record *failrec;
	struct extent_state *state, *next;

	if (RB_EMPTY_ROOT(&failure_tree->state))
		return;

	spin_lock(&failure_tree->lock);
	state = find_first_extent_bit_state(failure_tree, start, EXTENT_DIRTY);
	while (state) {
		if (state->start > end)
			break;

		ASSERT(state->end <= end);

		next = next_state(state);

2199
		failrec = (struct io_failure_record *)(unsigned long)state->private;
2200 2201 2202 2203 2204 2205 2206 2207
		free_extent_state(state);
		kfree(failrec);

		state = next;
	}
	spin_unlock(&failure_tree->lock);
}

2208 2209
int btrfs_get_io_failure_record(struct inode *inode, u64 start, u64 end,
				struct io_failure_record **failrec_ret)
2210
{
2211
	struct io_failure_record *failrec;
2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224
	u64 private;
	struct extent_map *em;
	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;
	int ret;
	u64 logical;

	ret = get_state_private(failure_tree, start, &private);
	if (ret) {
		failrec = kzalloc(sizeof(*failrec), GFP_NOFS);
		if (!failrec)
			return -ENOMEM;
2225

2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239
		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;
		}

2240
		if (em->start > start || em->start + em->len <= start) {
2241 2242 2243 2244
			free_extent_map(em);
			em = NULL;
		}
		read_unlock(&em_tree->lock);
2245
		if (!em) {
2246 2247 2248
			kfree(failrec);
			return -EIO;
		}
2249

2250 2251 2252 2253 2254 2255 2256 2257
		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);
		}
2258 2259 2260 2261

		pr_debug("Get IO Failure Record: (new) logical=%llu, start=%llu, len=%llu\n",
			 logical, start, failrec->len);

2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280
		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;
2281
		pr_debug("Get IO Failure Record: (found) logical=%llu, start=%llu, len=%llu, validation=%d\n",
2282 2283 2284 2285 2286 2287 2288 2289
			 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.
		 */
	}
2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300

	*failrec_ret = failrec;

	return 0;
}

int btrfs_check_repairable(struct inode *inode, struct bio *failed_bio,
			   struct io_failure_record *failrec, int failed_mirror)
{
	int num_copies;

2301 2302
	num_copies = btrfs_num_copies(BTRFS_I(inode)->root->fs_info,
				      failrec->logical, failrec->len);
2303 2304 2305 2306 2307 2308
	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.
		 */
2309
		pr_debug("Check Repairable: cannot repair, num_copies=%d, next_mirror %d, failed_mirror %d\n",
2310
			 num_copies, failrec->this_mirror, failed_mirror);
2311
		return 0;
2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347
	}

	/*
	 * 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;
	} 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++;
	}

2348
	if (failrec->this_mirror > num_copies) {
2349
		pr_debug("Check Repairable: (fail) num_copies=%d, next_mirror %d, failed_mirror %d\n",
2350
			 num_copies, failrec->this_mirror, failed_mirror);
2351
		return 0;
2352 2353
	}

2354 2355 2356 2357 2358 2359 2360
	return 1;
}


struct bio *btrfs_create_repair_bio(struct inode *inode, struct bio *failed_bio,
				    struct io_failure_record *failrec,
				    struct page *page, int pg_offset, int icsum,
2361
				    bio_end_io_t *endio_func, void *data)
2362 2363 2364 2365 2366
{
	struct bio *bio;
	struct btrfs_io_bio *btrfs_failed_bio;
	struct btrfs_io_bio *btrfs_bio;

2367
	bio = btrfs_io_bio_alloc(GFP_NOFS, 1);
2368 2369 2370 2371
	if (!bio)
		return NULL;

	bio->bi_end_io = endio_func;
2372
	bio->bi_iter.bi_sector = failrec->logical >> 9;
2373
	bio->bi_bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
2374
	bio->bi_iter.bi_size = 0;
2375
	bio->bi_private = data;
2376

2377 2378 2379 2380 2381 2382 2383
	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;
2384 2385
		icsum *= csum_size;
		memcpy(btrfs_bio->csum, btrfs_failed_bio->csum + icsum,
2386 2387 2388
		       csum_size);
	}

2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432
	bio_add_page(bio, page, failrec->len, pg_offset);

	return bio;
}

/*
 * 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
 */

static int bio_readpage_error(struct bio *failed_bio, u64 phy_offset,
			      struct page *page, u64 start, u64 end,
			      int failed_mirror)
{
	struct io_failure_record *failrec;
	struct inode *inode = page->mapping->host;
	struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
	struct bio *bio;
	int read_mode;
	int ret;

	BUG_ON(failed_bio->bi_rw & REQ_WRITE);

	ret = btrfs_get_io_failure_record(inode, start, end, &failrec);
	if (ret)
		return ret;

	ret = btrfs_check_repairable(inode, failed_bio, failrec, failed_mirror);
	if (!ret) {
		free_io_failure(inode, failrec);
		return -EIO;
	}

	if (failed_bio->bi_vcnt > 1)
		read_mode = READ_SYNC | REQ_FAILFAST_DEV;
	else
		read_mode = READ_SYNC;

	phy_offset >>= inode->i_sb->s_blocksize_bits;
	bio = btrfs_create_repair_bio(inode, failed_bio, failrec, page,
				      start - page_offset(page),
2433 2434
				      (int)phy_offset, failed_bio->bi_end_io,
				      NULL);
2435 2436 2437 2438
	if (!bio) {
		free_io_failure(inode, failrec);
		return -EIO;
	}
2439

2440 2441
	pr_debug("Repair Read Error: submitting new read[%#x] to this_mirror=%d, in_validation=%d\n",
		 read_mode, failrec->this_mirror, failrec->in_validation);
2442

2443 2444 2445
	ret = tree->ops->submit_bio_hook(inode, read_mode, bio,
					 failrec->this_mirror,
					 failrec->bio_flags, 0);
2446
	if (ret) {
2447
		free_io_failure(inode, failrec);
2448 2449 2450
		bio_put(bio);
	}

2451
	return ret;
2452 2453
}

2454 2455
/* lots and lots of room for performance fixes in the end_bio funcs */

2456 2457 2458 2459
int end_extent_writepage(struct page *page, int err, u64 start, u64 end)
{
	int uptodate = (err == 0);
	struct extent_io_tree *tree;
2460
	int ret = 0;
2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473

	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);
2474 2475
		ret = ret < 0 ? ret : -EIO;
		mapping_set_error(page->mapping, ret);
2476 2477 2478 2479
	}
	return 0;
}

2480 2481 2482 2483 2484 2485 2486 2487 2488
/*
 * 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.
 */
2489
static void end_bio_extent_writepage(struct bio *bio)
2490
{
2491
	struct bio_vec *bvec;
2492 2493
	u64 start;
	u64 end;
2494
	int i;
2495

2496
	bio_for_each_segment_all(bvec, bio, i) {
2497
		struct page *page = bvec->bv_page;
2498

2499 2500 2501 2502 2503
		/* 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.  */
2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514
		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);
		}
2515

2516 2517
		start = page_offset(page);
		end = start + bvec->bv_offset + bvec->bv_len - 1;
2518

2519
		if (end_extent_writepage(page, bio->bi_error, start, end))
2520
			continue;
2521

2522
		end_page_writeback(page);
2523
	}
2524

2525 2526 2527
	bio_put(bio);
}

2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539
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);
}

2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550
/*
 * 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.
 */
2551
static void end_bio_extent_readpage(struct bio *bio)
2552
{
2553
	struct bio_vec *bvec;
2554
	int uptodate = !bio->bi_error;
2555
	struct btrfs_io_bio *io_bio = btrfs_io_bio(bio);
2556
	struct extent_io_tree *tree;
2557
	u64 offset = 0;
2558 2559
	u64 start;
	u64 end;
2560
	u64 len;
2561 2562
	u64 extent_start = 0;
	u64 extent_len = 0;
2563
	int mirror;
2564
	int ret;
2565
	int i;
2566

2567
	bio_for_each_segment_all(bvec, bio, i) {
2568
		struct page *page = bvec->bv_page;
2569
		struct inode *inode = page->mapping->host;
2570

2571
		pr_debug("end_bio_extent_readpage: bi_sector=%llu, err=%d, "
2572 2573
			 "mirror=%u\n", (u64)bio->bi_iter.bi_sector,
			 bio->bi_error, io_bio->mirror_num);
2574
		tree = &BTRFS_I(inode)->io_tree;
2575

2576 2577 2578 2579 2580
		/* 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.  */
2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591
		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);
		}
2592

2593 2594
		start = page_offset(page);
		end = start + bvec->bv_offset + bvec->bv_len - 1;
2595
		len = bvec->bv_len;
2596

2597
		mirror = io_bio->mirror_num;
2598 2599
		if (likely(uptodate && tree->ops &&
			   tree->ops->readpage_end_io_hook)) {
2600 2601 2602
			ret = tree->ops->readpage_end_io_hook(io_bio, offset,
							      page, start, end,
							      mirror);
2603
			if (ret)
2604
				uptodate = 0;
2605
			else
2606
				clean_io_failure(inode, start, page, 0);
2607
		}
2608

2609 2610 2611 2612
		if (likely(uptodate))
			goto readpage_ok;

		if (tree->ops && tree->ops->readpage_io_failed_hook) {
2613
			ret = tree->ops->readpage_io_failed_hook(page, mirror);
2614
			if (!ret && !bio->bi_error)
2615
				uptodate = 1;
2616
		} else {
2617 2618 2619 2620 2621 2622 2623 2624 2625 2626
			/*
			 * 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.
			 */
2627 2628
			ret = bio_readpage_error(bio, offset, page, start, end,
						 mirror);
2629
			if (ret == 0) {
2630
				uptodate = !bio->bi_error;
2631
				offset += len;
2632 2633 2634
				continue;
			}
		}
2635
readpage_ok:
2636
		if (likely(uptodate)) {
2637 2638
			loff_t i_size = i_size_read(inode);
			pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT;
2639
			unsigned off;
2640 2641

			/* Zero out the end if this page straddles i_size */
2642 2643 2644
			off = i_size & (PAGE_CACHE_SIZE-1);
			if (page->index == end_index && off)
				zero_user_segment(page, off, PAGE_CACHE_SIZE);
2645
			SetPageUptodate(page);
2646
		} else {
2647 2648
			ClearPageUptodate(page);
			SetPageError(page);
2649
		}
2650
		unlock_page(page);
2651
		offset += len;
2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673

		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;
		}
2674
	}
2675

2676 2677 2678
	if (extent_len)
		endio_readpage_release_extent(tree, extent_start, extent_len,
					      uptodate);
2679
	if (io_bio->end_io)
2680
		io_bio->end_io(io_bio, bio->bi_error);
2681 2682 2683
	bio_put(bio);
}

2684 2685 2686 2687
/*
 * 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
 */
2688 2689 2690
struct bio *
btrfs_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
		gfp_t gfp_flags)
2691
{
2692
	struct btrfs_io_bio *btrfs_bio;
2693 2694
	struct bio *bio;

2695
	bio = bio_alloc_bioset(gfp_flags, nr_vecs, btrfs_bioset);
2696 2697

	if (bio == NULL && (current->flags & PF_MEMALLOC)) {
2698 2699 2700 2701
		while (!bio && (nr_vecs /= 2)) {
			bio = bio_alloc_bioset(gfp_flags,
					       nr_vecs, btrfs_bioset);
		}
2702 2703 2704 2705
	}

	if (bio) {
		bio->bi_bdev = bdev;
2706
		bio->bi_iter.bi_sector = first_sector;
2707 2708 2709 2710
		btrfs_bio = btrfs_io_bio(bio);
		btrfs_bio->csum = NULL;
		btrfs_bio->csum_allocated = NULL;
		btrfs_bio->end_io = NULL;
2711 2712 2713 2714
	}
	return bio;
}

2715 2716
struct bio *btrfs_bio_clone(struct bio *bio, gfp_t gfp_mask)
{
2717 2718
	struct btrfs_io_bio *btrfs_bio;
	struct bio *new;
2719

2720 2721 2722 2723 2724 2725
	new = bio_clone_bioset(bio, gfp_mask, btrfs_bioset);
	if (new) {
		btrfs_bio = btrfs_io_bio(new);
		btrfs_bio->csum = NULL;
		btrfs_bio->csum_allocated = NULL;
		btrfs_bio->end_io = NULL;
2726 2727

#ifdef CONFIG_BLK_CGROUP
2728 2729 2730
		/* FIXME, put this into bio_clone_bioset */
		if (bio->bi_css)
			bio_associate_blkcg(new, bio->bi_css);
2731
#endif
2732 2733 2734
	}
	return new;
}
2735 2736 2737 2738

/* this also allocates from the btrfs_bioset */
struct bio *btrfs_io_bio_alloc(gfp_t gfp_mask, unsigned int nr_iovecs)
{
2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749
	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;
2750 2751 2752
}


2753 2754
static int __must_check submit_one_bio(int rw, struct bio *bio,
				       int mirror_num, unsigned long bio_flags)
2755 2756
{
	int ret = 0;
2757 2758 2759 2760 2761
	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
2762
	start = page_offset(page) + bvec->bv_offset;
2763

2764
	bio->bi_private = NULL;
2765 2766 2767

	bio_get(bio);

2768
	if (tree->ops && tree->ops->submit_bio_hook)
2769
		ret = tree->ops->submit_bio_hook(page->mapping->host, rw, bio,
2770
					   mirror_num, bio_flags, start);
2771
	else
2772
		btrfsic_submit_bio(rw, bio);
2773

2774 2775 2776 2777
	bio_put(bio);
	return ret;
}

2778
static int merge_bio(int rw, struct extent_io_tree *tree, struct page *page,
2779 2780 2781 2782 2783
		     unsigned long offset, size_t size, struct bio *bio,
		     unsigned long bio_flags)
{
	int ret = 0;
	if (tree->ops && tree->ops->merge_bio_hook)
2784
		ret = tree->ops->merge_bio_hook(rw, page, offset, size, bio,
2785 2786 2787 2788 2789 2790
						bio_flags);
	BUG_ON(ret < 0);
	return ret;

}

2791
static int submit_extent_page(int rw, struct extent_io_tree *tree,
2792
			      struct writeback_control *wbc,
2793 2794 2795 2796 2797
			      struct page *page, sector_t sector,
			      size_t size, unsigned long offset,
			      struct block_device *bdev,
			      struct bio **bio_ret,
			      unsigned long max_pages,
2798
			      bio_end_io_t end_io_func,
C
Chris Mason 已提交
2799 2800
			      int mirror_num,
			      unsigned long prev_bio_flags,
2801 2802
			      unsigned long bio_flags,
			      bool force_bio_submit)
2803 2804 2805
{
	int ret = 0;
	struct bio *bio;
C
Chris Mason 已提交
2806 2807
	int contig = 0;
	int old_compressed = prev_bio_flags & EXTENT_BIO_COMPRESSED;
2808
	size_t page_size = min_t(size_t, size, PAGE_CACHE_SIZE);
2809 2810 2811

	if (bio_ret && *bio_ret) {
		bio = *bio_ret;
C
Chris Mason 已提交
2812
		if (old_compressed)
2813
			contig = bio->bi_iter.bi_sector == sector;
C
Chris Mason 已提交
2814
		else
K
Kent Overstreet 已提交
2815
			contig = bio_end_sector(bio) == sector;
C
Chris Mason 已提交
2816 2817

		if (prev_bio_flags != bio_flags || !contig ||
2818
		    force_bio_submit ||
2819
		    merge_bio(rw, tree, page, offset, page_size, bio, bio_flags) ||
C
Chris Mason 已提交
2820 2821 2822
		    bio_add_page(bio, page, page_size, offset) < page_size) {
			ret = submit_one_bio(rw, bio, mirror_num,
					     prev_bio_flags);
2823 2824
			if (ret < 0) {
				*bio_ret = NULL;
2825
				return ret;
2826
			}
2827 2828
			bio = NULL;
		} else {
2829 2830
			if (wbc)
				wbc_account_io(wbc, page, page_size);
2831 2832 2833
			return 0;
		}
	}
C
Chris Mason 已提交
2834

2835 2836
	bio = btrfs_bio_alloc(bdev, sector, BIO_MAX_PAGES,
			GFP_NOFS | __GFP_HIGH);
2837 2838
	if (!bio)
		return -ENOMEM;
2839

C
Chris Mason 已提交
2840
	bio_add_page(bio, page, page_size, offset);
2841 2842
	bio->bi_end_io = end_io_func;
	bio->bi_private = tree;
2843 2844 2845 2846
	if (wbc) {
		wbc_init_bio(wbc, bio);
		wbc_account_io(wbc, page, page_size);
	}
2847

C
Chris Mason 已提交
2848
	if (bio_ret)
2849
		*bio_ret = bio;
C
Chris Mason 已提交
2850
	else
C
Chris Mason 已提交
2851
		ret = submit_one_bio(rw, bio, mirror_num, bio_flags);
2852 2853 2854 2855

	return ret;
}

2856 2857
static void attach_extent_buffer_page(struct extent_buffer *eb,
				      struct page *page)
2858 2859 2860 2861
{
	if (!PagePrivate(page)) {
		SetPagePrivate(page);
		page_cache_get(page);
J
Josef Bacik 已提交
2862 2863 2864
		set_page_private(page, (unsigned long)eb);
	} else {
		WARN_ON(page->private != (unsigned long)eb);
2865 2866 2867
	}
}

J
Josef Bacik 已提交
2868
void set_page_extent_mapped(struct page *page)
2869
{
J
Josef Bacik 已提交
2870 2871 2872 2873 2874
	if (!PagePrivate(page)) {
		SetPagePrivate(page);
		page_cache_get(page);
		set_page_private(page, EXTENT_PAGE_PRIVATE);
	}
2875 2876
}

2877 2878 2879 2880 2881 2882 2883 2884 2885
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;
2886
		if (extent_map_in_tree(em) && start >= em->start &&
2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903
		    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;
}
2904 2905 2906 2907
/*
 * 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)
2908
 * XXX JDM: This needs looking at to ensure proper page locking
2909
 */
2910 2911 2912
static int __do_readpage(struct extent_io_tree *tree,
			 struct page *page,
			 get_extent_t *get_extent,
2913
			 struct extent_map **em_cached,
2914
			 struct bio **bio, int mirror_num,
2915 2916
			 unsigned long *bio_flags, int rw,
			 u64 *prev_em_start)
2917 2918
{
	struct inode *inode = page->mapping->host;
M
Miao Xie 已提交
2919
	u64 start = page_offset(page);
2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931
	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;
2932
	int parent_locked = *bio_flags & EXTENT_BIO_PARENT_LOCKED;
2933
	size_t pg_offset = 0;
2934
	size_t iosize;
C
Chris Mason 已提交
2935
	size_t disk_io_size;
2936
	size_t blocksize = inode->i_sb->s_blocksize;
2937
	unsigned long this_bio_flag = *bio_flags & EXTENT_BIO_PARENT_LOCKED;
2938 2939 2940

	set_page_extent_mapped(page);

2941
	end = page_end;
D
Dan Magenheimer 已提交
2942 2943 2944
	if (!PageUptodate(page)) {
		if (cleancache_get_page(page) == 0) {
			BUG_ON(blocksize != PAGE_SIZE);
2945
			unlock_extent(tree, start, end);
D
Dan Magenheimer 已提交
2946 2947 2948 2949
			goto out;
		}
	}

C
Chris Mason 已提交
2950 2951 2952 2953 2954 2955
	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;
2956
			userpage = kmap_atomic(page);
C
Chris Mason 已提交
2957 2958
			memset(userpage + zero_offset, 0, iosize);
			flush_dcache_page(page);
2959
			kunmap_atomic(userpage);
C
Chris Mason 已提交
2960 2961
		}
	}
2962
	while (cur <= end) {
2963
		unsigned long pnr = (last_byte >> PAGE_CACHE_SHIFT) + 1;
2964
		bool force_bio_submit = false;
2965

2966 2967
		if (cur >= last_byte) {
			char *userpage;
2968 2969
			struct extent_state *cached = NULL;

2970
			iosize = PAGE_CACHE_SIZE - pg_offset;
2971
			userpage = kmap_atomic(page);
2972
			memset(userpage + pg_offset, 0, iosize);
2973
			flush_dcache_page(page);
2974
			kunmap_atomic(userpage);
2975
			set_extent_uptodate(tree, cur, cur + iosize - 1,
2976
					    &cached, GFP_NOFS);
2977 2978 2979 2980
			if (!parent_locked)
				unlock_extent_cached(tree, cur,
						     cur + iosize - 1,
						     &cached, GFP_NOFS);
2981 2982
			break;
		}
2983 2984
		em = __get_extent_map(inode, page, pg_offset, cur,
				      end - cur + 1, get_extent, em_cached);
2985
		if (IS_ERR_OR_NULL(em)) {
2986
			SetPageError(page);
2987 2988
			if (!parent_locked)
				unlock_extent(tree, cur, end);
2989 2990 2991 2992 2993 2994
			break;
		}
		extent_offset = cur - em->start;
		BUG_ON(extent_map_end(em) <= cur);
		BUG_ON(end < cur);

2995
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
2996
			this_bio_flag |= EXTENT_BIO_COMPRESSED;
2997 2998 2999
			extent_set_compress_type(&this_bio_flag,
						 em->compress_type);
		}
C
Chris Mason 已提交
3000

3001 3002
		iosize = min(extent_map_end(em) - cur, end - cur + 1);
		cur_end = min(extent_map_end(em) - 1, end);
3003
		iosize = ALIGN(iosize, blocksize);
C
Chris Mason 已提交
3004 3005 3006 3007 3008 3009 3010
		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;
		}
3011 3012
		bdev = em->bdev;
		block_start = em->block_start;
Y
Yan Zheng 已提交
3013 3014
		if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			block_start = EXTENT_MAP_HOLE;
3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057

		/*
		 * If we have a file range that points to a compressed extent
		 * and it's followed by a consecutive file range that points to
		 * to the same compressed extent (possibly with a different
		 * offset and/or length, so it either points to the whole extent
		 * or only part of it), we must make sure we do not submit a
		 * single bio to populate the pages for the 2 ranges because
		 * this makes the compressed extent read zero out the pages
		 * belonging to the 2nd range. Imagine the following scenario:
		 *
		 *  File layout
		 *  [0 - 8K]                     [8K - 24K]
		 *    |                               |
		 *    |                               |
		 * points to extent X,         points to extent X,
		 * offset 4K, length of 8K     offset 0, length 16K
		 *
		 * [extent X, compressed length = 4K uncompressed length = 16K]
		 *
		 * If the bio to read the compressed extent covers both ranges,
		 * it will decompress extent X into the pages belonging to the
		 * first range and then it will stop, zeroing out the remaining
		 * pages that belong to the other range that points to extent X.
		 * So here we make sure we submit 2 bios, one for the first
		 * range and another one for the third range. Both will target
		 * the same physical extent from disk, but we can't currently
		 * make the compressed bio endio callback populate the pages
		 * for both ranges because each compressed bio is tightly
		 * coupled with a single extent map, and each range can have
		 * an extent map with a different offset value relative to the
		 * uncompressed data of our extent and different lengths. This
		 * is a corner case so we prioritize correctness over
		 * non-optimal behavior (submitting 2 bios for the same extent).
		 */
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags) &&
		    prev_em_start && *prev_em_start != (u64)-1 &&
		    *prev_em_start != em->orig_start)
			force_bio_submit = true;

		if (prev_em_start)
			*prev_em_start = em->orig_start;

3058 3059 3060 3061 3062 3063
		free_extent_map(em);
		em = NULL;

		/* we've found a hole, just zero and go on */
		if (block_start == EXTENT_MAP_HOLE) {
			char *userpage;
3064 3065
			struct extent_state *cached = NULL;

3066
			userpage = kmap_atomic(page);
3067
			memset(userpage + pg_offset, 0, iosize);
3068
			flush_dcache_page(page);
3069
			kunmap_atomic(userpage);
3070 3071

			set_extent_uptodate(tree, cur, cur + iosize - 1,
3072
					    &cached, GFP_NOFS);
3073 3074 3075 3076 3077 3078
			if (parent_locked)
				free_extent_state(cached);
			else
				unlock_extent_cached(tree, cur,
						     cur + iosize - 1,
						     &cached, GFP_NOFS);
3079
			cur = cur + iosize;
3080
			pg_offset += iosize;
3081 3082 3083
			continue;
		}
		/* the get_extent function already copied into the page */
3084 3085
		if (test_range_bit(tree, cur, cur_end,
				   EXTENT_UPTODATE, 1, NULL)) {
3086
			check_page_uptodate(tree, page);
3087 3088
			if (!parent_locked)
				unlock_extent(tree, cur, cur + iosize - 1);
3089
			cur = cur + iosize;
3090
			pg_offset += iosize;
3091 3092
			continue;
		}
3093 3094 3095 3096 3097
		/* we have an inline extent but it didn't get marked up
		 * to date.  Error out
		 */
		if (block_start == EXTENT_MAP_INLINE) {
			SetPageError(page);
3098 3099
			if (!parent_locked)
				unlock_extent(tree, cur, cur + iosize - 1);
3100
			cur = cur + iosize;
3101
			pg_offset += iosize;
3102 3103
			continue;
		}
3104

3105
		pnr -= page->index;
3106
		ret = submit_extent_page(rw, tree, NULL, page,
3107
					 sector, disk_io_size, pg_offset,
3108
					 bdev, bio, pnr,
C
Chris Mason 已提交
3109 3110
					 end_bio_extent_readpage, mirror_num,
					 *bio_flags,
3111 3112
					 this_bio_flag,
					 force_bio_submit);
3113 3114 3115 3116
		if (!ret) {
			nr++;
			*bio_flags = this_bio_flag;
		} else {
3117
			SetPageError(page);
3118 3119
			if (!parent_locked)
				unlock_extent(tree, cur, cur + iosize - 1);
3120
		}
3121
		cur = cur + iosize;
3122
		pg_offset += iosize;
3123
	}
D
Dan Magenheimer 已提交
3124
out:
3125 3126 3127 3128 3129 3130 3131 3132
	if (!nr) {
		if (!PageError(page))
			SetPageUptodate(page);
		unlock_page(page);
	}
	return 0;
}

3133 3134 3135 3136
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,
3137
					     struct extent_map **em_cached,
3138
					     struct bio **bio, int mirror_num,
3139 3140
					     unsigned long *bio_flags, int rw,
					     u64 *prev_em_start)
3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158
{
	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++) {
3159
		__do_readpage(tree, pages[index], get_extent, em_cached, bio,
3160
			      mirror_num, bio_flags, rw, prev_em_start);
3161 3162 3163 3164 3165 3166 3167
		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,
3168
			       struct extent_map **em_cached,
3169
			       struct bio **bio, int mirror_num,
3170 3171
			       unsigned long *bio_flags, int rw,
			       u64 *prev_em_start)
3172
{
3173
	u64 start = 0;
3174 3175 3176
	u64 end = 0;
	u64 page_start;
	int index;
3177
	int first_index = 0;
3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189

	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,
3190 3191
						  end, get_extent, em_cached,
						  bio, mirror_num, bio_flags,
3192
						  rw, prev_em_start);
3193 3194 3195 3196 3197 3198 3199 3200 3201
			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,
3202
					  end, get_extent, em_cached, bio,
3203 3204
					  mirror_num, bio_flags, rw,
					  prev_em_start);
3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228
}

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);
	}

3229
	ret = __do_readpage(tree, page, get_extent, NULL, bio, mirror_num,
3230
			    bio_flags, rw, NULL);
3231 3232 3233
	return ret;
}

3234
int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
3235
			    get_extent_t *get_extent, int mirror_num)
3236 3237
{
	struct bio *bio = NULL;
C
Chris Mason 已提交
3238
	unsigned long bio_flags = 0;
3239 3240
	int ret;

3241
	ret = __extent_read_full_page(tree, page, get_extent, &bio, mirror_num,
3242
				      &bio_flags, READ);
3243
	if (bio)
3244
		ret = submit_one_bio(READ, bio, mirror_num, bio_flags);
3245 3246 3247
	return ret;
}

3248 3249 3250 3251 3252 3253 3254 3255
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,
3256
			    &bio_flags, READ, NULL);
3257 3258 3259 3260 3261
	if (bio)
		ret = submit_one_bio(READ, bio, mirror_num, bio_flags);
	return ret;
}

3262 3263 3264 3265 3266 3267 3268 3269 3270 3271
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;
}

3272
/*
3273 3274 3275 3276 3277 3278 3279 3280
 * 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)
3281
 */
3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303
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,
3304
					       BTRFS_MAX_EXTENT_SIZE);
3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 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
		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)
3378 3379
{
	struct extent_io_tree *tree = epd->tree;
M
Miao Xie 已提交
3380
	u64 start = page_offset(page);
3381 3382 3383 3384 3385 3386 3387
	u64 page_end = start + PAGE_CACHE_SIZE - 1;
	u64 end;
	u64 cur = start;
	u64 extent_offset;
	u64 block_start;
	u64 iosize;
	sector_t sector;
3388
	struct extent_state *cached_state = NULL;
3389 3390
	struct extent_map *em;
	struct block_device *bdev;
3391
	size_t pg_offset = 0;
3392
	size_t blocksize;
3393 3394 3395
	int ret = 0;
	int nr = 0;
	bool compressed;
C
Chris Mason 已提交
3396

3397
	if (tree->ops && tree->ops->writepage_start_hook) {
C
Chris Mason 已提交
3398 3399
		ret = tree->ops->writepage_start_hook(page, start,
						      page_end);
3400 3401 3402 3403 3404 3405
		if (ret) {
			/* Fixup worker will requeue */
			if (ret == -EBUSY)
				wbc->pages_skipped++;
			else
				redirty_page_for_writepage(wbc, page);
3406

3407
			update_nr_written(page, wbc, nr_written);
3408
			unlock_page(page);
3409
			ret = 1;
3410
			goto done_unlocked;
3411 3412 3413
		}
	}

3414 3415 3416 3417 3418
	/*
	 * 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);
3419

3420
	end = page_end;
3421
	if (i_size <= start) {
3422 3423 3424
		if (tree->ops && tree->ops->writepage_end_io_hook)
			tree->ops->writepage_end_io_hook(page, start,
							 page_end, NULL, 1);
3425 3426 3427 3428 3429 3430
		goto done;
	}

	blocksize = inode->i_sb->s_blocksize;

	while (cur <= end) {
3431 3432
		u64 em_end;
		if (cur >= i_size) {
3433 3434 3435
			if (tree->ops && tree->ops->writepage_end_io_hook)
				tree->ops->writepage_end_io_hook(page, cur,
							 page_end, NULL, 1);
3436 3437
			break;
		}
3438
		em = epd->get_extent(inode, page, pg_offset, cur,
3439
				     end - cur + 1, 1);
3440
		if (IS_ERR_OR_NULL(em)) {
3441
			SetPageError(page);
3442
			ret = PTR_ERR_OR_ZERO(em);
3443 3444 3445 3446
			break;
		}

		extent_offset = cur - em->start;
3447 3448
		em_end = extent_map_end(em);
		BUG_ON(em_end <= cur);
3449
		BUG_ON(end < cur);
3450
		iosize = min(em_end - cur, end - cur + 1);
3451
		iosize = ALIGN(iosize, blocksize);
3452 3453 3454
		sector = (em->block_start + extent_offset) >> 9;
		bdev = em->bdev;
		block_start = em->block_start;
C
Chris Mason 已提交
3455
		compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
3456 3457 3458
		free_extent_map(em);
		em = NULL;

C
Chris Mason 已提交
3459 3460 3461 3462 3463
		/*
		 * compressed and inline extents are written through other
		 * paths in the FS
		 */
		if (compressed || block_start == EXTENT_MAP_HOLE ||
3464
		    block_start == EXTENT_MAP_INLINE) {
C
Chris Mason 已提交
3465 3466 3467 3468 3469 3470
			/*
			 * end_io notification does not happen here for
			 * compressed extents
			 */
			if (!compressed && tree->ops &&
			    tree->ops->writepage_end_io_hook)
3471 3472 3473
				tree->ops->writepage_end_io_hook(page, cur,
							 cur + iosize - 1,
							 NULL, 1);
C
Chris Mason 已提交
3474 3475 3476 3477 3478 3479 3480 3481 3482
			else if (compressed) {
				/* we don't want to end_page_writeback on
				 * a compressed extent.  this happens
				 * elsewhere
				 */
				nr++;
			}

			cur += iosize;
3483
			pg_offset += iosize;
3484 3485
			continue;
		}
C
Chris Mason 已提交
3486

3487 3488 3489 3490 3491 3492
		if (tree->ops && tree->ops->writepage_io_hook) {
			ret = tree->ops->writepage_io_hook(page, cur,
						cur + iosize - 1);
		} else {
			ret = 0;
		}
3493
		if (ret) {
3494
			SetPageError(page);
3495
		} else {
3496
			unsigned long max_nr = (i_size >> PAGE_CACHE_SHIFT) + 1;
3497

3498 3499
			set_range_writeback(tree, cur, cur + iosize - 1);
			if (!PageWriteback(page)) {
3500 3501
				btrfs_err(BTRFS_I(inode)->root->fs_info,
					   "page %lu not writeback, cur %llu end %llu",
3502
				       page->index, cur, end);
3503 3504
			}

3505
			ret = submit_extent_page(write_flags, tree, wbc, page,
3506 3507
						 sector, iosize, pg_offset,
						 bdev, &epd->bio, max_nr,
C
Chris Mason 已提交
3508
						 end_bio_extent_writepage,
3509
						 0, 0, 0, false);
3510 3511 3512 3513
			if (ret)
				SetPageError(page);
		}
		cur = cur + iosize;
3514
		pg_offset += iosize;
3515 3516
		nr++;
	}
3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591
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;

3592 3593 3594 3595 3596 3597
done:
	if (nr == 0) {
		/* make sure the mapping tag for page dirty gets cleared */
		set_page_writeback(page);
		end_page_writeback(page);
	}
3598 3599 3600 3601
	if (PageError(page)) {
		ret = ret < 0 ? ret : -EIO;
		end_extent_writepage(page, ret, start, page_end);
	}
3602
	unlock_page(page);
3603
	return ret;
3604

3605
done_unlocked:
3606 3607 3608
	return 0;
}

3609
void wait_on_extent_buffer_writeback(struct extent_buffer *eb)
3610
{
3611 3612
	wait_on_bit_io(&eb->bflags, EXTENT_BUFFER_WRITEBACK,
		       TASK_UNINTERRUPTIBLE);
3613 3614
}

3615 3616 3617 3618
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)
3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637
{
	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
3638 3639 3640 3641 3642
		while (1) {
			wait_on_extent_buffer_writeback(eb);
			btrfs_tree_lock(eb);
			if (!test_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags))
				break;
3643 3644 3645 3646
			btrfs_tree_unlock(eb);
		}
	}

3647 3648 3649 3650 3651 3652
	/*
	 * 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);
3653 3654
	if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
		set_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags);
3655
		spin_unlock(&eb->refs_lock);
3656
		btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
3657 3658 3659
		__percpu_counter_add(&fs_info->dirty_metadata_bytes,
				     -eb->len,
				     fs_info->dirty_metadata_batch);
3660
		ret = 1;
3661 3662
	} else {
		spin_unlock(&eb->refs_lock);
3663 3664 3665 3666 3667 3668 3669 3670 3671
	}

	btrfs_tree_unlock(eb);

	if (!ret)
		return ret;

	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = 0; i < num_pages; i++) {
3672
		struct page *p = eb->pages[i];
3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688

		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);
3689
	smp_mb__after_atomic();
3690 3691 3692
	wake_up_bit(&eb->bflags, EXTENT_BUFFER_WRITEBACK);
}

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 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754
static void set_btree_ioerr(struct page *page)
{
	struct extent_buffer *eb = (struct extent_buffer *)page->private;
	struct btrfs_inode *btree_ino = BTRFS_I(eb->fs_info->btree_inode);

	SetPageError(page);
	if (test_and_set_bit(EXTENT_BUFFER_WRITE_ERR, &eb->bflags))
		return;

	/*
	 * If writeback for a btree extent that doesn't belong to a log tree
	 * failed, increment the counter transaction->eb_write_errors.
	 * We do this because while the transaction is running and before it's
	 * committing (when we call filemap_fdata[write|wait]_range against
	 * the btree inode), we might have
	 * btree_inode->i_mapping->a_ops->writepages() called by the VM - if it
	 * returns an error or an error happens during writeback, when we're
	 * committing the transaction we wouldn't know about it, since the pages
	 * can be no longer dirty nor marked anymore for writeback (if a
	 * subsequent modification to the extent buffer didn't happen before the
	 * transaction commit), which makes filemap_fdata[write|wait]_range not
	 * able to find the pages tagged with SetPageError at transaction
	 * commit time. So if this happens we must abort the transaction,
	 * otherwise we commit a super block with btree roots that point to
	 * btree nodes/leafs whose content on disk is invalid - either garbage
	 * or the content of some node/leaf from a past generation that got
	 * cowed or deleted and is no longer valid.
	 *
	 * Note: setting AS_EIO/AS_ENOSPC in the btree inode's i_mapping would
	 * not be enough - we need to distinguish between log tree extents vs
	 * non-log tree extents, and the next filemap_fdatawait_range() call
	 * will catch and clear such errors in the mapping - and that call might
	 * be from a log sync and not from a transaction commit. Also, checking
	 * for the eb flag EXTENT_BUFFER_WRITE_ERR at transaction commit time is
	 * not done and would not be reliable - the eb might have been released
	 * from memory and reading it back again means that flag would not be
	 * set (since it's a runtime flag, not persisted on disk).
	 *
	 * Using the flags below in the btree inode also makes us achieve the
	 * goal of AS_EIO/AS_ENOSPC when writepages() returns success, started
	 * writeback for all dirty pages and before filemap_fdatawait_range()
	 * is called, the writeback for all dirty pages had already finished
	 * with errors - because we were not using AS_EIO/AS_ENOSPC,
	 * filemap_fdatawait_range() would return success, as it could not know
	 * that writeback errors happened (the pages were no longer tagged for
	 * writeback).
	 */
	switch (eb->log_index) {
	case -1:
		set_bit(BTRFS_INODE_BTREE_ERR, &btree_ino->runtime_flags);
		break;
	case 0:
		set_bit(BTRFS_INODE_BTREE_LOG1_ERR, &btree_ino->runtime_flags);
		break;
	case 1:
		set_bit(BTRFS_INODE_BTREE_LOG2_ERR, &btree_ino->runtime_flags);
		break;
	default:
		BUG(); /* unexpected, logic error */
	}
}

3755
static void end_bio_extent_buffer_writepage(struct bio *bio)
3756
{
3757
	struct bio_vec *bvec;
3758
	struct extent_buffer *eb;
3759
	int i, done;
3760

3761
	bio_for_each_segment_all(bvec, bio, i) {
3762 3763 3764 3765 3766 3767
		struct page *page = bvec->bv_page;

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

3768 3769
		if (bio->bi_error ||
		    test_bit(EXTENT_BUFFER_WRITE_ERR, &eb->bflags)) {
3770
			ClearPageUptodate(page);
3771
			set_btree_ioerr(page);
3772 3773 3774 3775 3776 3777 3778 3779
		}

		end_page_writeback(page);

		if (!done)
			continue;

		end_extent_buffer_writeback(eb);
3780
	}
3781 3782 3783 3784

	bio_put(bio);
}

3785
static noinline_for_stack int write_one_eb(struct extent_buffer *eb,
3786 3787 3788 3789 3790
			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;
3791
	struct extent_io_tree *tree = &BTRFS_I(fs_info->btree_inode)->io_tree;
3792 3793
	u64 offset = eb->start;
	unsigned long i, num_pages;
3794
	unsigned long bio_flags = 0;
3795
	int rw = (epd->sync_io ? WRITE_SYNC : WRITE) | REQ_META;
3796
	int ret = 0;
3797

3798
	clear_bit(EXTENT_BUFFER_WRITE_ERR, &eb->bflags);
3799 3800
	num_pages = num_extent_pages(eb->start, eb->len);
	atomic_set(&eb->io_pages, num_pages);
3801 3802 3803
	if (btrfs_header_owner(eb) == BTRFS_TREE_LOG_OBJECTID)
		bio_flags = EXTENT_BIO_TREE_LOG;

3804
	for (i = 0; i < num_pages; i++) {
3805
		struct page *p = eb->pages[i];
3806 3807 3808

		clear_page_dirty_for_io(p);
		set_page_writeback(p);
3809
		ret = submit_extent_page(rw, tree, wbc, p, offset >> 9,
3810 3811
					 PAGE_CACHE_SIZE, 0, bdev, &epd->bio,
					 -1, end_bio_extent_buffer_writepage,
3812
					 0, epd->bio_flags, bio_flags, false);
3813
		epd->bio_flags = bio_flags;
3814
		if (ret) {
3815
			set_btree_ioerr(p);
3816
			end_page_writeback(p);
3817 3818 3819 3820 3821 3822 3823 3824 3825 3826 3827 3828
			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++) {
3829
			struct page *p = eb->pages[i];
3830
			clear_page_dirty_for_io(p);
3831 3832 3833 3834 3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848
			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,
3849
		.bio_flags = 0,
3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893
	};
	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;
			}

3894 3895 3896 3897 3898 3899
			spin_lock(&mapping->private_lock);
			if (!PagePrivate(page)) {
				spin_unlock(&mapping->private_lock);
				continue;
			}

3900
			eb = (struct extent_buffer *)page->private;
3901 3902 3903 3904 3905 3906

			/*
			 * 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.
			 */
3907
			if (WARN_ON(!eb)) {
3908
				spin_unlock(&mapping->private_lock);
3909 3910 3911
				continue;
			}

3912 3913
			if (eb == prev_eb) {
				spin_unlock(&mapping->private_lock);
3914
				continue;
3915
			}
3916

3917 3918 3919
			ret = atomic_inc_not_zero(&eb->refs);
			spin_unlock(&mapping->private_lock);
			if (!ret)
3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959
				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;
}

3960
/**
C
Chris Mason 已提交
3961
 * write_cache_pages - walk the list of dirty pages of the given address space and write all of them.
3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974
 * @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.
 */
3975
static int extent_write_cache_pages(struct extent_io_tree *tree,
C
Chris Mason 已提交
3976 3977
			     struct address_space *mapping,
			     struct writeback_control *wbc,
C
Chris Mason 已提交
3978 3979
			     writepage_t writepage, void *data,
			     void (*flush_fn)(void *))
3980
{
3981
	struct inode *inode = mapping->host;
3982 3983
	int ret = 0;
	int done = 0;
3984
	int err = 0;
3985
	int nr_to_write_done = 0;
3986 3987 3988 3989 3990
	struct pagevec pvec;
	int nr_pages;
	pgoff_t index;
	pgoff_t end;		/* Inclusive */
	int scanned = 0;
3991
	int tag;
3992

3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004
	/*
	 * 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;

4005 4006 4007 4008 4009 4010 4011 4012 4013
	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;
	}
4014 4015 4016 4017
	if (wbc->sync_mode == WB_SYNC_ALL)
		tag = PAGECACHE_TAG_TOWRITE;
	else
		tag = PAGECACHE_TAG_DIRTY;
4018
retry:
4019 4020
	if (wbc->sync_mode == WB_SYNC_ALL)
		tag_pages_for_writeback(mapping, index, end);
4021
	while (!done && !nr_to_write_done && (index <= end) &&
4022 4023
	       (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag,
			min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036
		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
			 */
4037 4038 4039
			if (!trylock_page(page)) {
				flush_fn(data);
				lock_page(page);
4040
			}
4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052

			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
4053
			if (wbc->sync_mode != WB_SYNC_NONE) {
4054 4055
				if (PageWriteback(page))
					flush_fn(data);
4056
				wait_on_page_writeback(page);
C
Chris Mason 已提交
4057
			}
4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069 4070

			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;
			}
4071 4072
			if (!err && ret < 0)
				err = ret;
4073 4074 4075 4076 4077 4078 4079

			/*
			 * 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;
4080 4081 4082 4083
		}
		pagevec_release(&pvec);
		cond_resched();
	}
4084
	if (!scanned && !done && !err) {
4085 4086 4087 4088 4089 4090 4091 4092
		/*
		 * 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;
	}
4093
	btrfs_add_delayed_iput(inode);
4094
	return err;
4095 4096
}

4097
static void flush_epd_write_bio(struct extent_page_data *epd)
C
Chris Mason 已提交
4098 4099
{
	if (epd->bio) {
4100 4101 4102
		int rw = WRITE;
		int ret;

4103
		if (epd->sync_io)
4104 4105
			rw = WRITE_SYNC;

4106
		ret = submit_one_bio(rw, epd->bio, 0, epd->bio_flags);
4107
		BUG_ON(ret < 0); /* -ENOMEM */
C
Chris Mason 已提交
4108 4109 4110 4111
		epd->bio = NULL;
	}
}

4112 4113 4114 4115 4116 4117
static noinline void flush_write_bio(void *data)
{
	struct extent_page_data *epd = data;
	flush_epd_write_bio(epd);
}

4118 4119 4120 4121 4122 4123 4124 4125 4126
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,
4127
		.extent_locked = 0,
4128
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
4129
		.bio_flags = 0,
4130 4131 4132 4133
	};

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

4134
	flush_epd_write_bio(&epd);
4135 4136 4137
	return ret;
}

4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152
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,
4153
		.sync_io = mode == WB_SYNC_ALL,
4154
		.bio_flags = 0,
4155 4156 4157 4158 4159 4160 4161 4162
	};
	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
4163
	while (start <= end) {
4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175 4176 4177
		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;
	}

4178
	flush_epd_write_bio(&epd);
4179 4180
	return ret;
}
4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191

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,
4192
		.extent_locked = 0,
4193
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
4194
		.bio_flags = 0,
4195 4196
	};

C
Chris Mason 已提交
4197
	ret = extent_write_cache_pages(tree, mapping, wbc,
C
Chris Mason 已提交
4198 4199
				       __extent_writepage, &epd,
				       flush_write_bio);
4200
	flush_epd_write_bio(&epd);
4201 4202 4203 4204 4205 4206 4207 4208 4209 4210
	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
4211
	unsigned long bio_flags = 0;
L
Liu Bo 已提交
4212 4213
	struct page *pagepool[16];
	struct page *page;
4214
	struct extent_map *em_cached = NULL;
L
Liu Bo 已提交
4215
	int nr = 0;
4216
	u64 prev_em_start = (u64)-1;
4217 4218

	for (page_idx = 0; page_idx < nr_pages; page_idx++) {
L
Liu Bo 已提交
4219
		page = list_entry(pages->prev, struct page, lru);
4220 4221 4222

		prefetchw(&page->flags);
		list_del(&page->lru);
L
Liu Bo 已提交
4223
		if (add_to_page_cache_lru(page, mapping,
4224
					page->index, GFP_NOFS)) {
L
Liu Bo 已提交
4225 4226
			page_cache_release(page);
			continue;
4227
		}
L
Liu Bo 已提交
4228 4229 4230 4231

		pagepool[nr++] = page;
		if (nr < ARRAY_SIZE(pagepool))
			continue;
4232
		__extent_readpages(tree, pagepool, nr, get_extent, &em_cached,
4233
				   &bio, 0, &bio_flags, READ, &prev_em_start);
L
Liu Bo 已提交
4234
		nr = 0;
4235
	}
4236
	if (nr)
4237
		__extent_readpages(tree, pagepool, nr, get_extent, &em_cached,
4238
				   &bio, 0, &bio_flags, READ, &prev_em_start);
L
Liu Bo 已提交
4239

4240 4241 4242
	if (em_cached)
		free_extent_map(em_cached);

4243 4244
	BUG_ON(!list_empty(pages));
	if (bio)
4245
		return submit_one_bio(READ, bio, 0, bio_flags);
4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256
	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)
{
4257
	struct extent_state *cached_state = NULL;
M
Miao Xie 已提交
4258
	u64 start = page_offset(page);
4259 4260 4261
	u64 end = start + PAGE_CACHE_SIZE - 1;
	size_t blocksize = page->mapping->host->i_sb->s_blocksize;

4262
	start += ALIGN(offset, blocksize);
4263 4264 4265
	if (start > end)
		return 0;

4266
	lock_extent_bits(tree, start, end, 0, &cached_state);
4267
	wait_on_page_writeback(page);
4268
	clear_extent_bit(tree, start, end,
4269 4270
			 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
			 EXTENT_DO_ACCOUNTING,
4271
			 1, 1, &cached_state, GFP_NOFS);
4272 4273 4274
	return 0;
}

4275 4276 4277 4278 4279
/*
 * 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.
 */
4280 4281 4282
static int try_release_extent_state(struct extent_map_tree *map,
				    struct extent_io_tree *tree,
				    struct page *page, gfp_t mask)
4283
{
M
Miao Xie 已提交
4284
	u64 start = page_offset(page);
4285 4286 4287
	u64 end = start + PAGE_CACHE_SIZE - 1;
	int ret = 1;

4288
	if (test_range_bit(tree, start, end,
4289
			   EXTENT_IOBITS, 0, NULL))
4290 4291 4292 4293
		ret = 0;
	else {
		if ((mask & GFP_NOFS) == GFP_NOFS)
			mask = GFP_NOFS;
4294 4295 4296 4297
		/*
		 * at this point we can safely clear everything except the
		 * locked bit and the nodatasum bit
		 */
4298
		ret = clear_extent_bit(tree, start, end,
4299 4300
				 ~(EXTENT_LOCKED | EXTENT_NODATASUM),
				 0, 0, NULL, mask);
4301 4302 4303 4304 4305 4306 4307 4308

		/* if clear_extent_bit failed for enomem reasons,
		 * we can't allow the release to continue.
		 */
		if (ret < 0)
			ret = 0;
		else
			ret = 1;
4309 4310 4311 4312
	}
	return ret;
}

4313 4314 4315 4316 4317 4318
/*
 * 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,
4319 4320
			       struct extent_io_tree *tree, struct page *page,
			       gfp_t mask)
4321 4322
{
	struct extent_map *em;
M
Miao Xie 已提交
4323
	u64 start = page_offset(page);
4324
	u64 end = start + PAGE_CACHE_SIZE - 1;
4325

4326 4327
	if ((mask & __GFP_WAIT) &&
	    page->mapping->host->i_size > 16 * 1024 * 1024) {
4328
		u64 len;
4329
		while (start <= end) {
4330
			len = end - start + 1;
4331
			write_lock(&map->lock);
4332
			em = lookup_extent_mapping(map, start, len);
4333
			if (!em) {
4334
				write_unlock(&map->lock);
4335 4336
				break;
			}
4337 4338
			if (test_bit(EXTENT_FLAG_PINNED, &em->flags) ||
			    em->start != start) {
4339
				write_unlock(&map->lock);
4340 4341 4342 4343 4344
				free_extent_map(em);
				break;
			}
			if (!test_range_bit(tree, em->start,
					    extent_map_end(em) - 1,
4345
					    EXTENT_LOCKED | EXTENT_WRITEBACK,
4346
					    0, NULL)) {
4347 4348 4349 4350 4351
				remove_extent_mapping(map, em);
				/* once for the rb tree */
				free_extent_map(em);
			}
			start = extent_map_end(em);
4352
			write_unlock(&map->lock);
4353 4354

			/* once for us */
4355 4356 4357
			free_extent_map(em);
		}
	}
4358
	return try_release_extent_state(map, tree, page, mask);
4359 4360
}

4361 4362 4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 4374 4375 4376
/*
 * 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;

4377
	while (1) {
4378 4379 4380
		len = last - offset;
		if (len == 0)
			break;
4381
		len = ALIGN(len, sectorsize);
4382
		em = get_extent(inode, NULL, 0, offset, len, 0);
4383
		if (IS_ERR_OR_NULL(em))
4384 4385 4386 4387 4388 4389 4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400
			return em;

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

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

Y
Yehuda Sadeh 已提交
4401 4402 4403
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
4404
	int ret = 0;
Y
Yehuda Sadeh 已提交
4405 4406 4407
	u64 off = start;
	u64 max = start + len;
	u32 flags = 0;
J
Josef Bacik 已提交
4408 4409
	u32 found_type;
	u64 last;
4410
	u64 last_for_get_extent = 0;
Y
Yehuda Sadeh 已提交
4411
	u64 disko = 0;
4412
	u64 isize = i_size_read(inode);
J
Josef Bacik 已提交
4413
	struct btrfs_key found_key;
Y
Yehuda Sadeh 已提交
4414
	struct extent_map *em = NULL;
4415
	struct extent_state *cached_state = NULL;
J
Josef Bacik 已提交
4416
	struct btrfs_path *path;
4417
	struct btrfs_root *root = BTRFS_I(inode)->root;
Y
Yehuda Sadeh 已提交
4418
	int end = 0;
4419 4420 4421
	u64 em_start = 0;
	u64 em_len = 0;
	u64 em_end = 0;
Y
Yehuda Sadeh 已提交
4422 4423 4424 4425

	if (len == 0)
		return -EINVAL;

J
Josef Bacik 已提交
4426 4427 4428 4429 4430
	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;
	path->leave_spinning = 1;

4431 4432
	start = round_down(start, BTRFS_I(inode)->root->sectorsize);
	len = round_up(max, BTRFS_I(inode)->root->sectorsize) - start;
4433

4434 4435 4436 4437
	/*
	 * lookup the last file extent.  We're not using i_size here
	 * because there might be preallocation past i_size
	 */
4438 4439
	ret = btrfs_lookup_file_extent(NULL, root, path, btrfs_ino(inode), -1,
				       0);
J
Josef Bacik 已提交
4440 4441 4442 4443 4444 4445 4446
	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]);
4447
	found_type = found_key.type;
J
Josef Bacik 已提交
4448

4449
	/* No extents, but there might be delalloc bits */
L
Li Zefan 已提交
4450
	if (found_key.objectid != btrfs_ino(inode) ||
J
Josef Bacik 已提交
4451
	    found_type != BTRFS_EXTENT_DATA_KEY) {
4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462
		/* 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
4463
	}
4464
	btrfs_release_path(path);
J
Josef Bacik 已提交
4465

4466 4467 4468 4469 4470 4471 4472 4473 4474 4475
	/*
	 * 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
4476
	lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + len - 1, 0,
4477
			 &cached_state);
4478

4479
	em = get_extent_skip_holes(inode, start, last_for_get_extent,
4480
				   get_extent);
Y
Yehuda Sadeh 已提交
4481 4482 4483 4484 4485 4486
	if (!em)
		goto out;
	if (IS_ERR(em)) {
		ret = PTR_ERR(em);
		goto out;
	}
J
Josef Bacik 已提交
4487

Y
Yehuda Sadeh 已提交
4488
	while (!end) {
4489
		u64 offset_in_extent = 0;
4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501

		/* 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
4502

4503 4504
		/*
		 * record the offset from the start of the extent
4505 4506 4507
		 * 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.
4508
		 */
4509 4510
		if (!test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
			offset_in_extent = em_start - em->start;
4511
		em_end = extent_map_end(em);
4512
		em_len = em_end - em_start;
Y
Yehuda Sadeh 已提交
4513 4514 4515
		disko = 0;
		flags = 0;

4516 4517 4518 4519 4520 4521 4522
		/*
		 * bump off for our next call to get_extent
		 */
		off = extent_map_end(em);
		if (off >= max)
			end = 1;

4523
		if (em->block_start == EXTENT_MAP_LAST_BYTE) {
Y
Yehuda Sadeh 已提交
4524 4525
			end = 1;
			flags |= FIEMAP_EXTENT_LAST;
4526
		} else if (em->block_start == EXTENT_MAP_INLINE) {
Y
Yehuda Sadeh 已提交
4527 4528
			flags |= (FIEMAP_EXTENT_DATA_INLINE |
				  FIEMAP_EXTENT_NOT_ALIGNED);
4529
		} else if (em->block_start == EXTENT_MAP_DELALLOC) {
Y
Yehuda Sadeh 已提交
4530 4531
			flags |= (FIEMAP_EXTENT_DELALLOC |
				  FIEMAP_EXTENT_UNKNOWN);
4532 4533 4534
		} else if (fieinfo->fi_extents_max) {
			u64 bytenr = em->block_start -
				(em->start - em->orig_start);
4535

4536
			disko = em->block_start + offset_in_extent;
4537 4538 4539 4540

			/*
			 * As btrfs supports shared space, this information
			 * can be exported to userspace tools via
4541 4542 4543
			 * flag FIEMAP_EXTENT_SHARED.  If fi_extents_max == 0
			 * then we're just getting a count and we can skip the
			 * lookup stuff.
4544
			 */
4545 4546 4547 4548
			ret = btrfs_check_shared(NULL, root->fs_info,
						 root->objectid,
						 btrfs_ino(inode), bytenr);
			if (ret < 0)
4549
				goto out_free;
4550
			if (ret)
4551
				flags |= FIEMAP_EXTENT_SHARED;
4552
			ret = 0;
Y
Yehuda Sadeh 已提交
4553 4554 4555
		}
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
			flags |= FIEMAP_EXTENT_ENCODED;
4556 4557
		if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			flags |= FIEMAP_EXTENT_UNWRITTEN;
Y
Yehuda Sadeh 已提交
4558 4559 4560

		free_extent_map(em);
		em = NULL;
4561 4562
		if ((em_start >= last) || em_len == (u64)-1 ||
		   (last == (u64)-1 && isize <= em_end)) {
Y
Yehuda Sadeh 已提交
4563 4564 4565 4566
			flags |= FIEMAP_EXTENT_LAST;
			end = 1;
		}

4567 4568 4569 4570 4571 4572 4573 4574
		/* 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
4575 4576 4577
			flags |= FIEMAP_EXTENT_LAST;
			end = 1;
		}
4578 4579
		ret = fiemap_fill_next_extent(fieinfo, em_start, disko,
					      em_len, flags);
4580 4581 4582
		if (ret) {
			if (ret == 1)
				ret = 0;
4583
			goto out_free;
4584
		}
Y
Yehuda Sadeh 已提交
4585 4586 4587 4588
	}
out_free:
	free_extent_map(em);
out:
4589
	btrfs_free_path(path);
L
Liu Bo 已提交
4590
	unlock_extent_cached(&BTRFS_I(inode)->io_tree, start, start + len - 1,
4591
			     &cached_state, GFP_NOFS);
Y
Yehuda Sadeh 已提交
4592 4593 4594
	return ret;
}

4595 4596
static void __free_extent_buffer(struct extent_buffer *eb)
{
4597
	btrfs_leak_debug_del(&eb->leak_list);
4598 4599 4600
	kmem_cache_free(extent_buffer_cache, eb);
}

4601
int extent_buffer_under_io(struct extent_buffer *eb)
4602 4603 4604 4605 4606 4607 4608 4609 4610
{
	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.
 */
4611
static void btrfs_release_extent_buffer_page(struct extent_buffer *eb)
4612 4613 4614 4615 4616 4617 4618
{
	unsigned long index;
	struct page *page;
	int mapped = !test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags);

	BUG_ON(extent_buffer_under_io(eb));

4619 4620
	index = num_extent_pages(eb->start, eb->len);
	if (index == 0)
4621 4622 4623 4624
		return;

	do {
		index--;
4625
		page = eb->pages[index];
4626 4627 4628
		if (!page)
			continue;
		if (mapped)
4629
			spin_lock(&page->mapping->private_lock);
4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641
		/*
		 * 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));
4642
			/*
4643 4644
			 * We need to make sure we haven't be attached
			 * to a new eb.
4645
			 */
4646 4647 4648
			ClearPagePrivate(page);
			set_page_private(page, 0);
			/* One for the page private */
4649 4650
			page_cache_release(page);
		}
4651 4652 4653 4654 4655 4656

		if (mapped)
			spin_unlock(&page->mapping->private_lock);

		/* One for when we alloced the page */
		page_cache_release(page);
4657
	} while (index != 0);
4658 4659 4660 4661 4662 4663 4664
}

/*
 * Helper for releasing the extent buffer.
 */
static inline void btrfs_release_extent_buffer(struct extent_buffer *eb)
{
4665
	btrfs_release_extent_buffer_page(eb);
4666 4667 4668
	__free_extent_buffer(eb);
}

4669 4670
static struct extent_buffer *
__alloc_extent_buffer(struct btrfs_fs_info *fs_info, u64 start,
4671
		      unsigned long len)
4672 4673 4674
{
	struct extent_buffer *eb = NULL;

4675
	eb = kmem_cache_zalloc(extent_buffer_cache, GFP_NOFS|__GFP_NOFAIL);
4676 4677
	eb->start = start;
	eb->len = len;
4678
	eb->fs_info = fs_info;
4679
	eb->bflags = 0;
4680 4681 4682 4683 4684 4685 4686
	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);
4687
	eb->lock_nested = 0;
4688 4689
	init_waitqueue_head(&eb->write_lock_wq);
	init_waitqueue_head(&eb->read_lock_wq);
4690

4691 4692
	btrfs_leak_debug_add(&eb->leak_list, &buffers);

4693
	spin_lock_init(&eb->refs_lock);
4694
	atomic_set(&eb->refs, 1);
4695
	atomic_set(&eb->io_pages, 0);
4696

4697 4698 4699 4700 4701 4702
	/*
	 * 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);
4703 4704 4705 4706

	return eb;
}

4707 4708 4709 4710 4711 4712 4713
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);

4714
	new = __alloc_extent_buffer(src->fs_info, src->start, src->len);
4715 4716 4717 4718
	if (new == NULL)
		return NULL;

	for (i = 0; i < num_pages; i++) {
4719
		p = alloc_page(GFP_NOFS);
4720 4721 4722 4723
		if (!p) {
			btrfs_release_extent_buffer(new);
			return NULL;
		}
4724 4725 4726 4727 4728 4729 4730 4731 4732 4733 4734 4735 4736
		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;
}

4737 4738
struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
						u64 start)
4739 4740
{
	struct extent_buffer *eb;
4741 4742
	unsigned long len;
	unsigned long num_pages;
4743 4744
	unsigned long i;

4745 4746 4747 4748 4749 4750 4751 4752 4753 4754 4755 4756
	if (!fs_info) {
		/*
		 * Called only from tests that don't always have a fs_info
		 * available, but we know that nodesize is 4096
		 */
		len = 4096;
	} else {
		len = fs_info->tree_root->nodesize;
	}
	num_pages = num_extent_pages(0, len);

	eb = __alloc_extent_buffer(fs_info, start, len);
4757 4758 4759 4760
	if (!eb)
		return NULL;

	for (i = 0; i < num_pages; i++) {
4761
		eb->pages[i] = alloc_page(GFP_NOFS);
4762 4763 4764 4765 4766 4767 4768 4769 4770
		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:
4771 4772
	for (; i > 0; i--)
		__free_page(eb->pages[i - 1]);
4773 4774 4775 4776
	__free_extent_buffer(eb);
	return NULL;
}

4777 4778
static void check_buffer_tree_ref(struct extent_buffer *eb)
{
4779
	int refs;
4780 4781 4782 4783 4784 4785 4786 4787 4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 4798 4799
	/* 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.
	 */
4800 4801 4802 4803
	refs = atomic_read(&eb->refs);
	if (refs >= 2 && test_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
		return;

4804 4805
	spin_lock(&eb->refs_lock);
	if (!test_and_set_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
4806
		atomic_inc(&eb->refs);
4807
	spin_unlock(&eb->refs_lock);
4808 4809
}

4810 4811
static void mark_extent_buffer_accessed(struct extent_buffer *eb,
		struct page *accessed)
4812 4813 4814
{
	unsigned long num_pages, i;

4815 4816
	check_buffer_tree_ref(eb);

4817 4818
	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = 0; i < num_pages; i++) {
4819 4820
		struct page *p = eb->pages[i];

4821 4822
		if (p != accessed)
			mark_page_accessed(p);
4823 4824 4825
	}
}

4826 4827
struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
					 u64 start)
4828 4829 4830 4831
{
	struct extent_buffer *eb;

	rcu_read_lock();
4832 4833
	eb = radix_tree_lookup(&fs_info->buffer_radix,
			       start >> PAGE_CACHE_SHIFT);
4834 4835
	if (eb && atomic_inc_not_zero(&eb->refs)) {
		rcu_read_unlock();
4836 4837 4838 4839 4840 4841 4842 4843 4844 4845 4846 4847 4848 4849 4850 4851 4852 4853 4854
		/*
		 * Lock our eb's refs_lock to avoid races with
		 * free_extent_buffer. When we get our eb it might be flagged
		 * with EXTENT_BUFFER_STALE and another task running
		 * free_extent_buffer might have seen that flag set,
		 * eb->refs == 2, that the buffer isn't under IO (dirty and
		 * writeback flags not set) and it's still in the tree (flag
		 * EXTENT_BUFFER_TREE_REF set), therefore being in the process
		 * of decrementing the extent buffer's reference count twice.
		 * So here we could race and increment the eb's reference count,
		 * clear its stale flag, mark it as dirty and drop our reference
		 * before the other task finishes executing free_extent_buffer,
		 * which would later result in an attempt to free an extent
		 * buffer that is dirty.
		 */
		if (test_bit(EXTENT_BUFFER_STALE, &eb->bflags)) {
			spin_lock(&eb->refs_lock);
			spin_unlock(&eb->refs_lock);
		}
4855
		mark_extent_buffer_accessed(eb, NULL);
4856 4857 4858 4859 4860 4861 4862
		return eb;
	}
	rcu_read_unlock();

	return NULL;
}

4863 4864
#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
4865
					       u64 start)
4866 4867 4868 4869 4870 4871 4872
{
	struct extent_buffer *eb, *exists = NULL;
	int ret;

	eb = find_extent_buffer(fs_info, start);
	if (eb)
		return eb;
4873
	eb = alloc_dummy_extent_buffer(fs_info, start);
4874 4875 4876 4877 4878 4879 4880 4881 4882 4883 4884 4885 4886 4887 4888 4889 4890 4891 4892 4893 4894 4895 4896 4897 4898 4899 4900 4901 4902 4903 4904 4905 4906 4907 4908 4909
	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

4910
struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
4911
					  u64 start)
4912
{
4913
	unsigned long len = fs_info->tree_root->nodesize;
4914 4915 4916 4917
	unsigned long num_pages = num_extent_pages(start, len);
	unsigned long i;
	unsigned long index = start >> PAGE_CACHE_SHIFT;
	struct extent_buffer *eb;
4918
	struct extent_buffer *exists = NULL;
4919
	struct page *p;
4920
	struct address_space *mapping = fs_info->btree_inode->i_mapping;
4921
	int uptodate = 1;
4922
	int ret;
4923

4924
	eb = find_extent_buffer(fs_info, start);
4925
	if (eb)
4926 4927
		return eb;

4928
	eb = __alloc_extent_buffer(fs_info, start, len);
4929
	if (!eb)
4930 4931
		return NULL;

4932
	for (i = 0; i < num_pages; i++, index++) {
4933
		p = find_or_create_page(mapping, index, GFP_NOFS|__GFP_NOFAIL);
4934
		if (!p)
4935
			goto free_eb;
J
Josef Bacik 已提交
4936 4937 4938 4939 4940 4941 4942 4943 4944 4945 4946 4947 4948 4949

		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);
4950
				page_cache_release(p);
4951
				mark_extent_buffer_accessed(exists, p);
J
Josef Bacik 已提交
4952 4953
				goto free_eb;
			}
4954
			exists = NULL;
J
Josef Bacik 已提交
4955

4956
			/*
J
Josef Bacik 已提交
4957 4958 4959 4960
			 * Do this so attach doesn't complain and we need to
			 * drop the ref the old guy had.
			 */
			ClearPagePrivate(p);
4961
			WARN_ON(PageDirty(p));
J
Josef Bacik 已提交
4962
			page_cache_release(p);
4963
		}
J
Josef Bacik 已提交
4964 4965
		attach_extent_buffer_page(eb, p);
		spin_unlock(&mapping->private_lock);
4966
		WARN_ON(PageDirty(p));
4967
		eb->pages[i] = p;
4968 4969
		if (!PageUptodate(p))
			uptodate = 0;
C
Chris Mason 已提交
4970 4971 4972 4973 4974

		/*
		 * see below about how we avoid a nasty race with release page
		 * and why we unlock later
		 */
4975 4976
	}
	if (uptodate)
4977
		set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
4978
again:
4979 4980 4981 4982
	ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
	if (ret)
		goto free_eb;

4983 4984 4985 4986
	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);
4987
	radix_tree_preload_end();
4988
	if (ret == -EEXIST) {
4989
		exists = find_extent_buffer(fs_info, start);
4990 4991 4992
		if (exists)
			goto free_eb;
		else
4993
			goto again;
4994 4995
	}
	/* add one reference for the tree */
4996
	check_buffer_tree_ref(eb);
4997
	set_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags);
C
Chris Mason 已提交
4998 4999 5000 5001 5002 5003 5004 5005 5006 5007

	/*
	 * 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
	 */
5008 5009
	SetPageChecked(eb->pages[0]);
	for (i = 1; i < num_pages; i++) {
5010
		p = eb->pages[i];
5011 5012 5013 5014
		ClearPageChecked(p);
		unlock_page(p);
	}
	unlock_page(eb->pages[0]);
5015 5016
	return eb;

5017
free_eb:
5018
	WARN_ON(!atomic_dec_and_test(&eb->refs));
5019 5020 5021 5022
	for (i = 0; i < num_pages; i++) {
		if (eb->pages[i])
			unlock_page(eb->pages[i]);
	}
C
Chris Mason 已提交
5023

5024
	btrfs_release_extent_buffer(eb);
5025
	return exists;
5026 5027
}

5028 5029 5030 5031 5032 5033 5034 5035 5036
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 */
5037
static int release_extent_buffer(struct extent_buffer *eb)
5038 5039 5040
{
	WARN_ON(atomic_read(&eb->refs) == 0);
	if (atomic_dec_and_test(&eb->refs)) {
5041
		if (test_and_clear_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags)) {
5042
			struct btrfs_fs_info *fs_info = eb->fs_info;
5043

5044
			spin_unlock(&eb->refs_lock);
5045

5046 5047
			spin_lock(&fs_info->buffer_lock);
			radix_tree_delete(&fs_info->buffer_radix,
5048
					  eb->start >> PAGE_CACHE_SHIFT);
5049
			spin_unlock(&fs_info->buffer_lock);
5050 5051
		} else {
			spin_unlock(&eb->refs_lock);
5052
		}
5053 5054

		/* Should be safe to release our pages at this point */
5055
		btrfs_release_extent_buffer_page(eb);
5056 5057 5058 5059 5060 5061
#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
		if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags))) {
			__free_extent_buffer(eb);
			return 1;
		}
#endif
5062
		call_rcu(&eb->rcu_head, btrfs_release_extent_buffer_rcu);
5063
		return 1;
5064 5065
	}
	spin_unlock(&eb->refs_lock);
5066 5067

	return 0;
5068 5069
}

5070 5071
void free_extent_buffer(struct extent_buffer *eb)
{
5072 5073
	int refs;
	int old;
5074 5075 5076
	if (!eb)
		return;

5077 5078 5079 5080 5081 5082 5083 5084 5085
	while (1) {
		refs = atomic_read(&eb->refs);
		if (refs <= 3)
			break;
		old = atomic_cmpxchg(&eb->refs, refs, refs - 1);
		if (old == refs)
			return;
	}

5086
	spin_lock(&eb->refs_lock);
5087 5088 5089 5090
	if (atomic_read(&eb->refs) == 2 &&
	    test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags))
		atomic_dec(&eb->refs);

5091 5092
	if (atomic_read(&eb->refs) == 2 &&
	    test_bit(EXTENT_BUFFER_STALE, &eb->bflags) &&
5093
	    !extent_buffer_under_io(eb) &&
5094 5095 5096 5097 5098 5099 5100
	    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.
	 */
5101
	release_extent_buffer(eb);
5102 5103 5104 5105 5106
}

void free_extent_buffer_stale(struct extent_buffer *eb)
{
	if (!eb)
5107 5108
		return;

5109 5110 5111
	spin_lock(&eb->refs_lock);
	set_bit(EXTENT_BUFFER_STALE, &eb->bflags);

5112
	if (atomic_read(&eb->refs) == 2 && !extent_buffer_under_io(eb) &&
5113 5114
	    test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
		atomic_dec(&eb->refs);
5115
	release_extent_buffer(eb);
5116 5117
}

5118
void clear_extent_buffer_dirty(struct extent_buffer *eb)
5119 5120 5121 5122 5123 5124 5125 5126
{
	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++) {
5127
		page = eb->pages[i];
5128
		if (!PageDirty(page))
C
Chris Mason 已提交
5129 5130
			continue;

5131
		lock_page(page);
C
Chris Mason 已提交
5132 5133
		WARN_ON(!PagePrivate(page));

5134
		clear_page_dirty_for_io(page);
5135
		spin_lock_irq(&page->mapping->tree_lock);
5136 5137 5138 5139 5140
		if (!PageDirty(page)) {
			radix_tree_tag_clear(&page->mapping->page_tree,
						page_index(page),
						PAGECACHE_TAG_DIRTY);
		}
5141
		spin_unlock_irq(&page->mapping->tree_lock);
5142
		ClearPageError(page);
5143
		unlock_page(page);
5144
	}
5145
	WARN_ON(atomic_read(&eb->refs) == 0);
5146 5147
}

5148
int set_extent_buffer_dirty(struct extent_buffer *eb)
5149 5150 5151
{
	unsigned long i;
	unsigned long num_pages;
5152
	int was_dirty = 0;
5153

5154 5155
	check_buffer_tree_ref(eb);

5156
	was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
5157

5158
	num_pages = num_extent_pages(eb->start, eb->len);
5159
	WARN_ON(atomic_read(&eb->refs) == 0);
5160 5161
	WARN_ON(!test_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags));

5162
	for (i = 0; i < num_pages; i++)
5163
		set_page_dirty(eb->pages[i]);
5164
	return was_dirty;
5165 5166
}

5167
int clear_extent_buffer_uptodate(struct extent_buffer *eb)
5168 5169 5170 5171 5172
{
	unsigned long i;
	struct page *page;
	unsigned long num_pages;

5173
	clear_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5174
	num_pages = num_extent_pages(eb->start, eb->len);
5175
	for (i = 0; i < num_pages; i++) {
5176
		page = eb->pages[i];
C
Chris Mason 已提交
5177 5178
		if (page)
			ClearPageUptodate(page);
5179 5180 5181 5182
	}
	return 0;
}

5183
int set_extent_buffer_uptodate(struct extent_buffer *eb)
5184 5185 5186 5187 5188
{
	unsigned long i;
	struct page *page;
	unsigned long num_pages;

5189
	set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5190 5191
	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = 0; i < num_pages; i++) {
5192
		page = eb->pages[i];
5193 5194 5195 5196 5197
		SetPageUptodate(page);
	}
	return 0;
}

5198
int extent_buffer_uptodate(struct extent_buffer *eb)
5199
{
5200
	return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5201 5202 5203
}

int read_extent_buffer_pages(struct extent_io_tree *tree,
5204
			     struct extent_buffer *eb, u64 start, int wait,
5205
			     get_extent_t *get_extent, int mirror_num)
5206 5207 5208 5209 5210 5211
{
	unsigned long i;
	unsigned long start_i;
	struct page *page;
	int err;
	int ret = 0;
5212 5213
	int locked_pages = 0;
	int all_uptodate = 1;
5214
	unsigned long num_pages;
5215
	unsigned long num_reads = 0;
5216
	struct bio *bio = NULL;
C
Chris Mason 已提交
5217
	unsigned long bio_flags = 0;
5218

5219
	if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags))
5220 5221 5222 5223 5224 5225 5226 5227 5228 5229 5230 5231
		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++) {
5232
		page = eb->pages[i];
5233
		if (wait == WAIT_NONE) {
5234
			if (!trylock_page(page))
5235
				goto unlock_exit;
5236 5237 5238
		} else {
			lock_page(page);
		}
5239
		locked_pages++;
5240 5241
		if (!PageUptodate(page)) {
			num_reads++;
5242
			all_uptodate = 0;
5243
		}
5244 5245 5246
	}
	if (all_uptodate) {
		if (start_i == 0)
5247
			set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5248 5249 5250
		goto unlock_exit;
	}

5251
	clear_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags);
5252
	eb->read_mirror = 0;
5253
	atomic_set(&eb->io_pages, num_reads);
5254
	for (i = start_i; i < num_pages; i++) {
5255
		page = eb->pages[i];
5256
		if (!PageUptodate(page)) {
5257
			ClearPageError(page);
5258
			err = __extent_read_full_page(tree, page,
5259
						      get_extent, &bio,
5260 5261
						      mirror_num, &bio_flags,
						      READ | REQ_META);
C
Chris Mason 已提交
5262
			if (err)
5263 5264 5265 5266 5267 5268
				ret = err;
		} else {
			unlock_page(page);
		}
	}

5269
	if (bio) {
5270 5271
		err = submit_one_bio(READ | REQ_META, bio, mirror_num,
				     bio_flags);
5272 5273
		if (err)
			return err;
5274
	}
5275

5276
	if (ret || wait != WAIT_COMPLETE)
5277
		return ret;
C
Chris Mason 已提交
5278

5279
	for (i = start_i; i < num_pages; i++) {
5280
		page = eb->pages[i];
5281
		wait_on_page_locked(page);
C
Chris Mason 已提交
5282
		if (!PageUptodate(page))
5283 5284
			ret = -EIO;
	}
C
Chris Mason 已提交
5285

5286
	return ret;
5287 5288 5289

unlock_exit:
	i = start_i;
C
Chris Mason 已提交
5290
	while (locked_pages > 0) {
5291
		page = eb->pages[i];
5292 5293 5294 5295 5296
		i++;
		unlock_page(page);
		locked_pages--;
	}
	return ret;
5297 5298 5299 5300 5301 5302 5303 5304 5305 5306 5307 5308 5309 5310 5311 5312 5313
}

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);

5314
	offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
5315

C
Chris Mason 已提交
5316
	while (len > 0) {
5317
		page = eb->pages[i];
5318 5319

		cur = min(len, (PAGE_CACHE_SIZE - offset));
5320
		kaddr = page_address(page);
5321 5322 5323 5324 5325 5326 5327 5328 5329
		memcpy(dst, kaddr + offset, cur);

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

5330 5331 5332 5333 5334 5335 5336 5337 5338 5339 5340 5341 5342 5343 5344 5345 5346 5347 5348
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) {
5349
		page = eb->pages[i];
5350 5351 5352 5353 5354 5355 5356 5357 5358 5359 5360 5361 5362 5363 5364 5365 5366

		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;
}

5367
int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start,
5368
			       unsigned long min_len, char **map,
5369
			       unsigned long *map_start,
5370
			       unsigned long *map_len)
5371 5372 5373 5374 5375 5376 5377 5378 5379 5380 5381 5382 5383 5384 5385 5386 5387 5388 5389
{
	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
5390

5391
	if (start + min_len > eb->len) {
J
Julia Lawall 已提交
5392
		WARN(1, KERN_ERR "btrfs bad mapping eb start %llu len %lu, "
5393 5394
		       "wanted %lu %lu\n",
		       eb->start, eb->len, start, min_len);
5395
		return -EINVAL;
5396 5397
	}

5398
	p = eb->pages[i];
5399
	kaddr = page_address(p);
5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410 5411 5412 5413 5414 5415 5416 5417 5418 5419 5420
	*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);

5421
	offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
5422

C
Chris Mason 已提交
5423
	while (len > 0) {
5424
		page = eb->pages[i];
5425 5426 5427

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

5428
		kaddr = page_address(page);
5429 5430 5431 5432 5433 5434 5435 5436 5437 5438 5439 5440 5441 5442 5443 5444 5445 5446 5447 5448 5449 5450 5451 5452 5453 5454
		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);

5455
	offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
5456

C
Chris Mason 已提交
5457
	while (len > 0) {
5458
		page = eb->pages[i];
5459 5460 5461
		WARN_ON(!PageUptodate(page));

		cur = min(len, PAGE_CACHE_SIZE - offset);
5462
		kaddr = page_address(page);
5463 5464 5465 5466 5467 5468 5469 5470 5471 5472 5473 5474 5475 5476 5477 5478 5479 5480 5481 5482 5483 5484
		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);

5485
	offset = (start_offset + start) & (PAGE_CACHE_SIZE - 1);
5486

C
Chris Mason 已提交
5487
	while (len > 0) {
5488
		page = eb->pages[i];
5489 5490 5491
		WARN_ON(!PageUptodate(page));

		cur = min(len, PAGE_CACHE_SIZE - offset);
5492
		kaddr = page_address(page);
5493 5494 5495 5496 5497 5498 5499 5500 5501 5502 5503 5504 5505 5506 5507 5508 5509 5510 5511 5512 5513 5514 5515
		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) &
5516
		(PAGE_CACHE_SIZE - 1);
5517

C
Chris Mason 已提交
5518
	while (len > 0) {
5519
		page = dst->pages[i];
5520 5521 5522 5523
		WARN_ON(!PageUptodate(page));

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

5524
		kaddr = page_address(page);
5525 5526 5527 5528 5529 5530 5531 5532 5533
		read_extent_buffer(src, kaddr + offset, src_offset, cur);

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

5534 5535 5536 5537 5538 5539
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;
}

5540 5541 5542 5543
static void copy_pages(struct page *dst_page, struct page *src_page,
		       unsigned long dst_off, unsigned long src_off,
		       unsigned long len)
{
5544
	char *dst_kaddr = page_address(dst_page);
5545
	char *src_kaddr;
5546
	int must_memmove = 0;
5547

5548
	if (dst_page != src_page) {
5549
		src_kaddr = page_address(src_page);
5550
	} else {
5551
		src_kaddr = dst_kaddr;
5552 5553
		if (areas_overlap(src_off, dst_off, len))
			must_memmove = 1;
5554
	}
5555

5556 5557 5558 5559
	if (must_memmove)
		memmove(dst_kaddr + dst_off, src_kaddr + src_off, len);
	else
		memcpy(dst_kaddr + dst_off, src_kaddr + src_off, len);
5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570 5571 5572
}

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) {
5573 5574 5575
		btrfs_err(dst->fs_info,
			"memmove bogus src_offset %lu move "
		       "len %lu dst len %lu", src_offset, len, dst->len);
5576 5577 5578
		BUG_ON(1);
	}
	if (dst_offset + len > dst->len) {
5579 5580 5581
		btrfs_err(dst->fs_info,
			"memmove bogus dst_offset %lu move "
		       "len %lu dst len %lu", dst_offset, len, dst->len);
5582 5583 5584
		BUG_ON(1);
	}

C
Chris Mason 已提交
5585
	while (len > 0) {
5586
		dst_off_in_page = (start_offset + dst_offset) &
5587
			(PAGE_CACHE_SIZE - 1);
5588
		src_off_in_page = (start_offset + src_offset) &
5589
			(PAGE_CACHE_SIZE - 1);
5590 5591 5592 5593 5594 5595 5596 5597 5598

		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));

5599
		copy_pages(dst->pages[dst_i], dst->pages[src_i],
5600 5601 5602 5603 5604 5605 5606 5607 5608 5609 5610 5611 5612 5613 5614 5615 5616 5617 5618 5619 5620
			   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) {
5621 5622
		btrfs_err(dst->fs_info, "memmove bogus src_offset %lu move "
		       "len %lu len %lu", src_offset, len, dst->len);
5623 5624 5625
		BUG_ON(1);
	}
	if (dst_offset + len > dst->len) {
5626 5627
		btrfs_err(dst->fs_info, "memmove bogus dst_offset %lu move "
		       "len %lu len %lu", dst_offset, len, dst->len);
5628 5629
		BUG_ON(1);
	}
5630
	if (dst_offset < src_offset) {
5631 5632 5633
		memcpy_extent_buffer(dst, dst_offset, src_offset, len);
		return;
	}
C
Chris Mason 已提交
5634
	while (len > 0) {
5635 5636 5637 5638
		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) &
5639
			(PAGE_CACHE_SIZE - 1);
5640
		src_off_in_page = (start_offset + src_end) &
5641
			(PAGE_CACHE_SIZE - 1);
5642 5643 5644

		cur = min_t(unsigned long, len, src_off_in_page + 1);
		cur = min(cur, dst_off_in_page + 1);
5645
		copy_pages(dst->pages[dst_i], dst->pages[src_i],
5646 5647 5648 5649 5650 5651 5652 5653
			   dst_off_in_page - cur + 1,
			   src_off_in_page - cur + 1, cur);

		dst_end -= cur;
		src_end -= cur;
		len -= cur;
	}
}
5654

5655
int try_release_extent_buffer(struct page *page)
5656
{
5657 5658
	struct extent_buffer *eb;

5659 5660 5661 5662 5663 5664 5665
	/*
	 * 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
5666
		return 1;
5667
	}
5668

5669 5670
	eb = (struct extent_buffer *)page->private;
	BUG_ON(!eb);
5671 5672

	/*
5673 5674 5675
	 * 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.
5676
	 */
5677
	spin_lock(&eb->refs_lock);
5678
	if (atomic_read(&eb->refs) != 1 || extent_buffer_under_io(eb)) {
5679 5680 5681
		spin_unlock(&eb->refs_lock);
		spin_unlock(&page->mapping->private_lock);
		return 0;
5682
	}
5683
	spin_unlock(&page->mapping->private_lock);
5684

5685
	/*
5686 5687
	 * 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.
5688
	 */
5689 5690 5691
	if (!test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags)) {
		spin_unlock(&eb->refs_lock);
		return 0;
5692
	}
5693

5694
	return release_extent_buffer(eb);
5695
}