extent_io.c 145.9 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
};

134 135 136 137 138 139 140 141 142 143
static void add_extent_changeset(struct extent_state *state, unsigned bits,
				 struct extent_changeset *changeset,
				 int set)
{
	int ret;

	if (!changeset)
		return;
	if (set && (state->state & bits) == bits)
		return;
144 145
	if (!set && (state->state & bits) == 0)
		return;
146 147 148 149 150 151 152
	changeset->bytes_changed += state->end - state->start + 1;
	ret = ulist_add(changeset->range_changed, state->start, state->end,
			GFP_ATOMIC);
	/* ENOMEM */
	BUG_ON(ret < 0);
}

153
static noinline void flush_write_bio(void *data);
154 155 156
static inline struct btrfs_fs_info *
tree_fs_info(struct extent_io_tree *tree)
{
157 158
	if (!tree->mapping)
		return NULL;
159 160
	return btrfs_sb(tree->mapping->host->i_sb);
}
161

162 163
int __init extent_io_init(void)
{
D
David Sterba 已提交
164
	extent_state_cache = kmem_cache_create("btrfs_extent_state",
165 166
			sizeof(struct extent_state), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
167 168 169
	if (!extent_state_cache)
		return -ENOMEM;

D
David Sterba 已提交
170
	extent_buffer_cache = kmem_cache_create("btrfs_extent_buffer",
171 172
			sizeof(struct extent_buffer), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
173 174
	if (!extent_buffer_cache)
		goto free_state_cache;
175 176 177 178 179

	btrfs_bioset = bioset_create(BIO_POOL_SIZE,
				     offsetof(struct btrfs_io_bio, bio));
	if (!btrfs_bioset)
		goto free_buffer_cache;
180 181 182 183

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

184 185
	return 0;

186 187 188 189
free_bioset:
	bioset_free(btrfs_bioset);
	btrfs_bioset = NULL;

190 191 192 193
free_buffer_cache:
	kmem_cache_destroy(extent_buffer_cache);
	extent_buffer_cache = NULL;

194 195
free_state_cache:
	kmem_cache_destroy(extent_state_cache);
196
	extent_state_cache = NULL;
197 198 199 200 201
	return -ENOMEM;
}

void extent_io_exit(void)
{
202
	btrfs_leak_debug_check();
203 204 205 206 207 208

	/*
	 * Make sure all delayed rcu free are flushed before we
	 * destroy caches.
	 */
	rcu_barrier();
209 210
	kmem_cache_destroy(extent_state_cache);
	kmem_cache_destroy(extent_buffer_cache);
211 212
	if (btrfs_bioset)
		bioset_free(btrfs_bioset);
213 214 215
}

void extent_io_tree_init(struct extent_io_tree *tree,
216
			 struct address_space *mapping)
217
{
218
	tree->state = RB_ROOT;
219 220
	tree->ops = NULL;
	tree->dirty_bytes = 0;
221
	spin_lock_init(&tree->lock);
222 223 224
	tree->mapping = mapping;
}

225
static struct extent_state *alloc_extent_state(gfp_t mask)
226 227 228 229
{
	struct extent_state *state;

	state = kmem_cache_alloc(extent_state_cache, mask);
230
	if (!state)
231 232
		return state;
	state->state = 0;
233
	state->failrec = NULL;
234
	RB_CLEAR_NODE(&state->rb_node);
235
	btrfs_leak_debug_add(&state->leak_list, &states);
236 237
	atomic_set(&state->refs, 1);
	init_waitqueue_head(&state->wq);
238
	trace_alloc_extent_state(state, mask, _RET_IP_);
239 240 241
	return state;
}

242
void free_extent_state(struct extent_state *state)
243 244 245 246
{
	if (!state)
		return;
	if (atomic_dec_and_test(&state->refs)) {
247
		WARN_ON(extent_state_in_tree(state));
248
		btrfs_leak_debug_del(&state->leak_list);
249
		trace_free_extent_state(state, _RET_IP_);
250 251 252 253
		kmem_cache_free(extent_state_cache, state);
	}
}

254 255 256
static struct rb_node *tree_insert(struct rb_root *root,
				   struct rb_node *search_start,
				   u64 offset,
257 258 259
				   struct rb_node *node,
				   struct rb_node ***p_in,
				   struct rb_node **parent_in)
260
{
261
	struct rb_node **p;
C
Chris Mason 已提交
262
	struct rb_node *parent = NULL;
263 264
	struct tree_entry *entry;

265 266 267 268 269 270
	if (p_in && parent_in) {
		p = *p_in;
		parent = *parent_in;
		goto do_insert;
	}

271
	p = search_start ? &search_start : &root->rb_node;
C
Chris Mason 已提交
272
	while (*p) {
273 274 275 276 277 278 279 280 281 282 283
		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;
	}

284
do_insert:
285 286 287 288 289
	rb_link_node(node, parent, p);
	rb_insert_color(node, root);
	return NULL;
}

290
static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset,
291 292 293 294
				      struct rb_node **prev_ret,
				      struct rb_node **next_ret,
				      struct rb_node ***p_ret,
				      struct rb_node **parent_ret)
295
{
296
	struct rb_root *root = &tree->state;
297
	struct rb_node **n = &root->rb_node;
298 299 300 301 302
	struct rb_node *prev = NULL;
	struct rb_node *orig_prev = NULL;
	struct tree_entry *entry;
	struct tree_entry *prev_entry = NULL;

303 304 305
	while (*n) {
		prev = *n;
		entry = rb_entry(prev, struct tree_entry, rb_node);
306 307 308
		prev_entry = entry;

		if (offset < entry->start)
309
			n = &(*n)->rb_left;
310
		else if (offset > entry->end)
311
			n = &(*n)->rb_right;
C
Chris Mason 已提交
312
		else
313
			return *n;
314 315
	}

316 317 318 319 320
	if (p_ret)
		*p_ret = n;
	if (parent_ret)
		*parent_ret = prev;

321 322
	if (prev_ret) {
		orig_prev = prev;
C
Chris Mason 已提交
323
		while (prev && offset > prev_entry->end) {
324 325 326 327 328 329 330 331 332
			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
333
		while (prev && offset < prev_entry->start) {
334 335 336 337 338 339 340 341
			prev = rb_prev(prev);
			prev_entry = rb_entry(prev, struct tree_entry, rb_node);
		}
		*next_ret = prev;
	}
	return NULL;
}

342 343 344 345 346
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)
347
{
348
	struct rb_node *prev = NULL;
349
	struct rb_node *ret;
350

351
	ret = __etree_search(tree, offset, &prev, NULL, p_ret, parent_ret);
C
Chris Mason 已提交
352
	if (!ret)
353 354 355 356
		return prev;
	return ret;
}

357 358 359 360 361 362
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
363 364 365 366 367 368 369 370
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);
}

371 372 373 374 375 376 377 378 379
/*
 * 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.
 */
380 381
static void merge_state(struct extent_io_tree *tree,
		        struct extent_state *state)
382 383 384 385
{
	struct extent_state *other;
	struct rb_node *other_node;

386
	if (state->state & (EXTENT_IOBITS | EXTENT_BOUNDARY))
387
		return;
388 389 390 391 392 393

	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
394
			merge_cb(tree, state, other);
395 396
			state->start = other->start;
			rb_erase(&other->rb_node, &tree->state);
397
			RB_CLEAR_NODE(&other->rb_node);
398 399 400 401 402 403 404 405
			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
406
			merge_cb(tree, state, other);
407 408
			state->end = other->end;
			rb_erase(&other->rb_node, &tree->state);
409
			RB_CLEAR_NODE(&other->rb_node);
410
			free_extent_state(other);
411 412 413 414
		}
	}
}

415
static void set_state_cb(struct extent_io_tree *tree,
416
			 struct extent_state *state, unsigned *bits)
417
{
418 419
	if (tree->ops && tree->ops->set_bit_hook)
		tree->ops->set_bit_hook(tree->mapping->host, state, bits);
420 421 422
}

static void clear_state_cb(struct extent_io_tree *tree,
423
			   struct extent_state *state, unsigned *bits)
424
{
J
Josef Bacik 已提交
425 426
	if (tree->ops && tree->ops->clear_bit_hook)
		tree->ops->clear_bit_hook(tree->mapping->host, state, bits);
427 428
}

429
static void set_state_bits(struct extent_io_tree *tree,
430 431
			   struct extent_state *state, unsigned *bits,
			   struct extent_changeset *changeset);
432

433 434 435 436 437 438 439 440 441 442 443 444
/*
 * 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,
445 446
			struct rb_node ***p,
			struct rb_node **parent,
447
			unsigned *bits, struct extent_changeset *changeset)
448 449 450
{
	struct rb_node *node;

J
Julia Lawall 已提交
451
	if (end < start)
452
		WARN(1, KERN_ERR "BTRFS: end < start %llu %llu\n",
453
		       end, start);
454 455
	state->start = start;
	state->end = end;
J
Josef Bacik 已提交
456

457
	set_state_bits(tree, state, bits, changeset);
458

459
	node = tree_insert(&tree->state, NULL, end, &state->rb_node, p, parent);
460 461 462
	if (node) {
		struct extent_state *found;
		found = rb_entry(node, struct extent_state, rb_node);
463
		printk(KERN_ERR "BTRFS: found node %llu %llu on insert of "
464 465
		       "%llu %llu\n",
		       found->start, found->end, start, end);
466 467 468 469 470 471
		return -EEXIST;
	}
	merge_state(tree, state);
	return 0;
}

472
static void split_cb(struct extent_io_tree *tree, struct extent_state *orig,
J
Josef Bacik 已提交
473 474 475
		     u64 split)
{
	if (tree->ops && tree->ops->split_extent_hook)
476
		tree->ops->split_extent_hook(tree->mapping->host, orig, split);
J
Josef Bacik 已提交
477 478
}

479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496
/*
 * 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
497 498 499

	split_cb(tree, orig, split);

500 501 502 503 504
	prealloc->start = orig->start;
	prealloc->end = split - 1;
	prealloc->state = orig->state;
	orig->start = split;

505 506
	node = tree_insert(&tree->state, &orig->rb_node, prealloc->end,
			   &prealloc->rb_node, NULL, NULL);
507 508 509 510 511 512 513
	if (node) {
		free_extent_state(prealloc);
		return -EEXIST;
	}
	return 0;
}

514 515 516 517 518 519 520 521 522
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;
}

523 524
/*
 * utility function to clear some bits in an extent state struct.
525
 * it will optionally wake up any one waiting on this state (wake == 1).
526 527 528 529
 *
 * If no bits are set on the state struct after clearing things, the
 * struct is freed and removed from the tree
 */
530 531
static struct extent_state *clear_state_bit(struct extent_io_tree *tree,
					    struct extent_state *state,
532 533
					    unsigned *bits, int wake,
					    struct extent_changeset *changeset)
534
{
535
	struct extent_state *next;
536
	unsigned bits_to_clear = *bits & ~EXTENT_CTLBITS;
537

538
	if ((bits_to_clear & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) {
539 540 541 542
		u64 range = state->end - state->start + 1;
		WARN_ON(range > tree->dirty_bytes);
		tree->dirty_bytes -= range;
	}
543
	clear_state_cb(tree, state, bits);
544
	add_extent_changeset(state, bits_to_clear, changeset, 0);
545
	state->state &= ~bits_to_clear;
546 547
	if (wake)
		wake_up(&state->wq);
548
	if (state->state == 0) {
549
		next = next_state(state);
550
		if (extent_state_in_tree(state)) {
551
			rb_erase(&state->rb_node, &tree->state);
552
			RB_CLEAR_NODE(&state->rb_node);
553 554 555 556 557 558
			free_extent_state(state);
		} else {
			WARN_ON(1);
		}
	} else {
		merge_state(tree, state);
559
		next = next_state(state);
560
	}
561
	return next;
562 563
}

564 565 566 567 568 569 570 571 572
static struct extent_state *
alloc_extent_state_atomic(struct extent_state *prealloc)
{
	if (!prealloc)
		prealloc = alloc_extent_state(GFP_ATOMIC);

	return prealloc;
}

573
static void extent_io_tree_panic(struct extent_io_tree *tree, int err)
574 575 576 577 578 579
{
	btrfs_panic(tree_fs_info(tree), err, "Locking error: "
		    "Extent tree was modified by another "
		    "thread while locked.");
}

580 581 582 583 584 585 586 587 588 589
/*
 * 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.
 *
590
 * This takes the tree lock, and returns 0 on success and < 0 on error.
591
 */
592 593 594 595
static int __clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
			      unsigned bits, int wake, int delete,
			      struct extent_state **cached_state,
			      gfp_t mask, struct extent_changeset *changeset)
596 597
{
	struct extent_state *state;
598
	struct extent_state *cached;
599 600
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
601
	u64 last_end;
602
	int err;
603
	int clear = 0;
604

605
	btrfs_debug_check_extent_io_range(tree, start, end);
606

607 608 609
	if (bits & EXTENT_DELALLOC)
		bits |= EXTENT_NORESERVE;

610 611 612 613
	if (delete)
		bits |= ~EXTENT_CTLBITS;
	bits |= EXTENT_FIRST_DELALLOC;

614 615
	if (bits & (EXTENT_IOBITS | EXTENT_BOUNDARY))
		clear = 1;
616
again:
617
	if (!prealloc && gfpflags_allow_blocking(mask)) {
618 619 620 621 622 623 624
		/*
		 * 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.
		 */
625 626 627
		prealloc = alloc_extent_state(mask);
	}

628
	spin_lock(&tree->lock);
629 630
	if (cached_state) {
		cached = *cached_state;
631 632 633 634 635 636

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

637 638
		if (cached && extent_state_in_tree(cached) &&
		    cached->start <= start && cached->end > start) {
639 640
			if (clear)
				atomic_dec(&cached->refs);
641
			state = cached;
642
			goto hit_next;
643
		}
644 645
		if (clear)
			free_extent_state(cached);
646
	}
647 648 649 650
	/*
	 * this search will find the extents that end after
	 * our range starts
	 */
651
	node = tree_search(tree, start);
652 653 654
	if (!node)
		goto out;
	state = rb_entry(node, struct extent_state, rb_node);
655
hit_next:
656 657 658
	if (state->start > end)
		goto out;
	WARN_ON(state->end < start);
659
	last_end = state->end;
660

661
	/* the state doesn't have the wanted bits, go ahead */
662 663
	if (!(state->state & bits)) {
		state = next_state(state);
664
		goto next;
665
	}
666

667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683
	/*
	 *     | ---- 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) {
684 685
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
686
		err = split_state(tree, state, prealloc, start);
687 688 689
		if (err)
			extent_io_tree_panic(tree, err);

690 691 692 693
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
694 695
			state = clear_state_bit(tree, state, &bits, wake,
						changeset);
696
			goto next;
697 698 699 700 701 702 703 704 705 706
		}
		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) {
707 708
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
709
		err = split_state(tree, state, prealloc, end + 1);
710 711 712
		if (err)
			extent_io_tree_panic(tree, err);

713 714
		if (wake)
			wake_up(&state->wq);
715

716
		clear_state_bit(tree, prealloc, &bits, wake, changeset);
J
Josef Bacik 已提交
717

718 719 720
		prealloc = NULL;
		goto out;
	}
721

722
	state = clear_state_bit(tree, state, &bits, wake, changeset);
723
next:
724 725 726
	if (last_end == (u64)-1)
		goto out;
	start = last_end + 1;
727
	if (start <= end && state && !need_resched())
728
		goto hit_next;
729 730 731
	goto search_again;

out:
732
	spin_unlock(&tree->lock);
733 734 735
	if (prealloc)
		free_extent_state(prealloc);

736
	return 0;
737 738 739 740

search_again:
	if (start > end)
		goto out;
741
	spin_unlock(&tree->lock);
742
	if (gfpflags_allow_blocking(mask))
743 744 745 746
		cond_resched();
	goto again;
}

747 748
static void wait_on_state(struct extent_io_tree *tree,
			  struct extent_state *state)
749 750
		__releases(tree->lock)
		__acquires(tree->lock)
751 752 753
{
	DEFINE_WAIT(wait);
	prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE);
754
	spin_unlock(&tree->lock);
755
	schedule();
756
	spin_lock(&tree->lock);
757 758 759 760 761 762 763 764
	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
 */
765 766
static void wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
			    unsigned long bits)
767 768 769 770
{
	struct extent_state *state;
	struct rb_node *node;

771
	btrfs_debug_check_extent_io_range(tree, start, end);
772

773
	spin_lock(&tree->lock);
774 775 776 777 778 779
again:
	while (1) {
		/*
		 * this search will find all the extents that end after
		 * our range starts
		 */
780
		node = tree_search(tree, start);
781
process_node:
782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801
		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;

802 803 804 805
		if (!cond_resched_lock(&tree->lock)) {
			node = rb_next(node);
			goto process_node;
		}
806 807
	}
out:
808
	spin_unlock(&tree->lock);
809 810
}

811
static void set_state_bits(struct extent_io_tree *tree,
812
			   struct extent_state *state,
813
			   unsigned *bits, struct extent_changeset *changeset)
814
{
815
	unsigned bits_to_set = *bits & ~EXTENT_CTLBITS;
J
Josef Bacik 已提交
816

817
	set_state_cb(tree, state, bits);
818
	if ((bits_to_set & EXTENT_DIRTY) && !(state->state & EXTENT_DIRTY)) {
819 820 821
		u64 range = state->end - state->start + 1;
		tree->dirty_bytes += range;
	}
822
	add_extent_changeset(state, bits_to_set, changeset, 1);
823
	state->state |= bits_to_set;
824 825
}

826 827
static void cache_state_if_flags(struct extent_state *state,
				 struct extent_state **cached_ptr,
828
				 unsigned flags)
829 830
{
	if (cached_ptr && !(*cached_ptr)) {
831
		if (!flags || (state->state & flags)) {
832 833 834 835 836 837
			*cached_ptr = state;
			atomic_inc(&state->refs);
		}
	}
}

838 839 840 841 842 843 844
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);
}

845
/*
846 847
 * 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.
848
 *
849 850 851
 * 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.
852
 *
853
 * [start, end] is inclusive This takes the tree lock.
854
 */
855

J
Jeff Mahoney 已提交
856 857
static int __must_check
__set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
858
		 unsigned bits, unsigned exclusive_bits,
859
		 u64 *failed_start, struct extent_state **cached_state,
860
		 gfp_t mask, struct extent_changeset *changeset)
861 862 863 864
{
	struct extent_state *state;
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
865 866
	struct rb_node **p;
	struct rb_node *parent;
867 868 869
	int err = 0;
	u64 last_start;
	u64 last_end;
870

871
	btrfs_debug_check_extent_io_range(tree, start, end);
872

873
	bits |= EXTENT_FIRST_DELALLOC;
874
again:
875
	if (!prealloc && gfpflags_allow_blocking(mask)) {
876
		prealloc = alloc_extent_state(mask);
877
		BUG_ON(!prealloc);
878 879
	}

880
	spin_lock(&tree->lock);
881 882
	if (cached_state && *cached_state) {
		state = *cached_state;
883
		if (state->start <= start && state->end > start &&
884
		    extent_state_in_tree(state)) {
885 886 887 888
			node = &state->rb_node;
			goto hit_next;
		}
	}
889 890 891 892
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
893
	node = tree_search_for_insert(tree, start, &p, &parent);
894
	if (!node) {
895 896
		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
897
		err = insert_state(tree, prealloc, start, end,
898
				   &p, &parent, &bits, changeset);
899 900 901
		if (err)
			extent_io_tree_panic(tree, err);

902
		cache_state(prealloc, cached_state);
903 904 905 906
		prealloc = NULL;
		goto out;
	}
	state = rb_entry(node, struct extent_state, rb_node);
C
Chris Mason 已提交
907
hit_next:
908 909 910 911 912 913 914 915 916 917
	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) {
918
		if (state->state & exclusive_bits) {
919 920 921 922
			*failed_start = state->start;
			err = -EEXIST;
			goto out;
		}
923

924
		set_state_bits(tree, state, &bits, changeset);
925
		cache_state(state, cached_state);
926
		merge_state(tree, state);
927 928 929
		if (last_end == (u64)-1)
			goto out;
		start = last_end + 1;
930 931 932 933
		state = next_state(state);
		if (start < end && state && state->start == start &&
		    !need_resched())
			goto hit_next;
934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953
		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) {
954
		if (state->state & exclusive_bits) {
955 956 957 958
			*failed_start = start;
			err = -EEXIST;
			goto out;
		}
959 960 961

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
962
		err = split_state(tree, state, prealloc, start);
963 964 965
		if (err)
			extent_io_tree_panic(tree, err);

966 967 968 969
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
970
			set_state_bits(tree, state, &bits, changeset);
971
			cache_state(state, cached_state);
972
			merge_state(tree, state);
973 974 975
			if (last_end == (u64)-1)
				goto out;
			start = last_end + 1;
976 977 978 979
			state = next_state(state);
			if (start < end && state && state->start == start &&
			    !need_resched())
				goto hit_next;
980 981 982 983 984 985 986 987 988 989 990 991 992 993 994
		}
		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
995
			this_end = last_start - 1;
996 997 998

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
999 1000 1001 1002 1003

		/*
		 * Avoid to free 'prealloc' if it can be merged with
		 * the later extent.
		 */
1004
		err = insert_state(tree, prealloc, start, this_end,
1005
				   NULL, NULL, &bits, changeset);
1006 1007 1008
		if (err)
			extent_io_tree_panic(tree, err);

J
Josef Bacik 已提交
1009 1010
		cache_state(prealloc, cached_state);
		prealloc = NULL;
1011 1012 1013 1014 1015 1016 1017 1018 1019 1020
		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) {
1021
		if (state->state & exclusive_bits) {
1022 1023 1024 1025
			*failed_start = start;
			err = -EEXIST;
			goto out;
		}
1026 1027 1028

		prealloc = alloc_extent_state_atomic(prealloc);
		BUG_ON(!prealloc);
1029
		err = split_state(tree, state, prealloc, end + 1);
1030 1031
		if (err)
			extent_io_tree_panic(tree, err);
1032

1033
		set_state_bits(tree, prealloc, &bits, changeset);
1034
		cache_state(prealloc, cached_state);
1035 1036 1037 1038 1039 1040 1041 1042
		merge_state(tree, prealloc);
		prealloc = NULL;
		goto out;
	}

	goto search_again;

out:
1043
	spin_unlock(&tree->lock);
1044 1045 1046 1047 1048 1049 1050 1051
	if (prealloc)
		free_extent_state(prealloc);

	return err;

search_again:
	if (start > end)
		goto out;
1052
	spin_unlock(&tree->lock);
1053
	if (gfpflags_allow_blocking(mask))
1054 1055 1056 1057
		cond_resched();
	goto again;
}

1058
int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
1059
		   unsigned bits, u64 * failed_start,
1060
		   struct extent_state **cached_state, gfp_t mask)
J
Jeff Mahoney 已提交
1061 1062
{
	return __set_extent_bit(tree, start, end, bits, 0, failed_start,
1063
				cached_state, mask, NULL);
J
Jeff Mahoney 已提交
1064 1065 1066
}


J
Josef Bacik 已提交
1067
/**
L
Liu Bo 已提交
1068 1069
 * convert_extent_bit - convert all bits in a given range from one bit to
 * 			another
J
Josef Bacik 已提交
1070 1071 1072 1073 1074
 * @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
1075
 * @cached_state:	state that we're going to cache
J
Josef Bacik 已提交
1076 1077 1078 1079 1080 1081 1082 1083 1084
 * @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,
1085
		       unsigned bits, unsigned clear_bits,
1086
		       struct extent_state **cached_state, gfp_t mask)
J
Josef Bacik 已提交
1087 1088 1089 1090
{
	struct extent_state *state;
	struct extent_state *prealloc = NULL;
	struct rb_node *node;
1091 1092
	struct rb_node **p;
	struct rb_node *parent;
J
Josef Bacik 已提交
1093 1094 1095
	int err = 0;
	u64 last_start;
	u64 last_end;
1096
	bool first_iteration = true;
J
Josef Bacik 已提交
1097

1098
	btrfs_debug_check_extent_io_range(tree, start, end);
1099

J
Josef Bacik 已提交
1100
again:
1101
	if (!prealloc && gfpflags_allow_blocking(mask)) {
1102 1103 1104 1105 1106 1107 1108
		/*
		 * 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
1109
		prealloc = alloc_extent_state(mask);
1110
		if (!prealloc && !first_iteration)
J
Josef Bacik 已提交
1111 1112 1113 1114
			return -ENOMEM;
	}

	spin_lock(&tree->lock);
1115 1116 1117
	if (cached_state && *cached_state) {
		state = *cached_state;
		if (state->start <= start && state->end > start &&
1118
		    extent_state_in_tree(state)) {
1119 1120 1121 1122 1123
			node = &state->rb_node;
			goto hit_next;
		}
	}

J
Josef Bacik 已提交
1124 1125 1126 1127
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1128
	node = tree_search_for_insert(tree, start, &p, &parent);
J
Josef Bacik 已提交
1129 1130
	if (!node) {
		prealloc = alloc_extent_state_atomic(prealloc);
1131 1132 1133 1134
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
1135
		err = insert_state(tree, prealloc, start, end,
1136
				   &p, &parent, &bits, NULL);
1137 1138
		if (err)
			extent_io_tree_panic(tree, err);
1139 1140
		cache_state(prealloc, cached_state);
		prealloc = NULL;
J
Josef Bacik 已提交
1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154
		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) {
1155
		set_state_bits(tree, state, &bits, NULL);
1156
		cache_state(state, cached_state);
1157
		state = clear_state_bit(tree, state, &clear_bits, 0, NULL);
J
Josef Bacik 已提交
1158 1159 1160
		if (last_end == (u64)-1)
			goto out;
		start = last_end + 1;
1161 1162 1163
		if (start < end && state && state->start == start &&
		    !need_resched())
			goto hit_next;
J
Josef Bacik 已提交
1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184
		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);
1185 1186 1187 1188
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1189
		err = split_state(tree, state, prealloc, start);
1190 1191
		if (err)
			extent_io_tree_panic(tree, err);
J
Josef Bacik 已提交
1192 1193 1194 1195
		prealloc = NULL;
		if (err)
			goto out;
		if (state->end <= end) {
1196
			set_state_bits(tree, state, &bits, NULL);
1197
			cache_state(state, cached_state);
1198 1199
			state = clear_state_bit(tree, state, &clear_bits, 0,
						NULL);
J
Josef Bacik 已提交
1200 1201 1202
			if (last_end == (u64)-1)
				goto out;
			start = last_end + 1;
1203 1204 1205
			if (start < end && state && state->start == start &&
			    !need_resched())
				goto hit_next;
J
Josef Bacik 已提交
1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223
		}
		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);
1224 1225 1226 1227
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1228 1229 1230 1231 1232 1233

		/*
		 * Avoid to free 'prealloc' if it can be merged with
		 * the later extent.
		 */
		err = insert_state(tree, prealloc, start, this_end,
1234
				   NULL, NULL, &bits, NULL);
1235 1236
		if (err)
			extent_io_tree_panic(tree, err);
1237
		cache_state(prealloc, cached_state);
J
Josef Bacik 已提交
1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249
		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);
1250 1251 1252 1253
		if (!prealloc) {
			err = -ENOMEM;
			goto out;
		}
J
Josef Bacik 已提交
1254 1255

		err = split_state(tree, state, prealloc, end + 1);
1256 1257
		if (err)
			extent_io_tree_panic(tree, err);
J
Josef Bacik 已提交
1258

1259
		set_state_bits(tree, prealloc, &bits, NULL);
1260
		cache_state(prealloc, cached_state);
1261
		clear_state_bit(tree, prealloc, &clear_bits, 0, NULL);
J
Josef Bacik 已提交
1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278
		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);
1279
	if (gfpflags_allow_blocking(mask))
J
Josef Bacik 已提交
1280
		cond_resched();
1281
	first_iteration = false;
J
Josef Bacik 已提交
1282 1283 1284
	goto again;
}

1285
/* wrappers around set/clear extent bit */
1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301
int set_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
			   unsigned bits, gfp_t mask,
			   struct extent_changeset *changeset)
{
	/*
	 * We don't support EXTENT_LOCKED yet, as current changeset will
	 * record any bits changed, so for EXTENT_LOCKED case, it will
	 * either fail with -EEXIST or changeset will record the whole
	 * range.
	 */
	BUG_ON(bits & EXTENT_LOCKED);

	return __set_extent_bit(tree, start, end, bits, 0, NULL, NULL, mask,
				changeset);
}

1302 1303 1304 1305 1306 1307 1308 1309 1310
int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
		     unsigned bits, int wake, int delete,
		     struct extent_state **cached, gfp_t mask)
{
	return __clear_extent_bit(tree, start, end, bits, wake, delete,
				  cached, mask, NULL);
}

int clear_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1311
		unsigned bits, struct extent_changeset *changeset)
1312 1313 1314 1315 1316 1317 1318
{
	/*
	 * Don't support EXTENT_LOCKED case, same reason as
	 * set_record_extent_bits().
	 */
	BUG_ON(bits & EXTENT_LOCKED);

1319
	return __clear_extent_bit(tree, start, end, bits, 0, 0, NULL, GFP_NOFS,
1320 1321 1322
				  changeset);
}

C
Chris Mason 已提交
1323 1324 1325 1326
/*
 * either insert or lock state struct between start and end use mask to tell
 * us if waiting is desired.
 */
1327
int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1328
		     struct extent_state **cached_state)
1329 1330 1331
{
	int err;
	u64 failed_start;
1332

1333
	while (1) {
1334
		err = __set_extent_bit(tree, start, end, EXTENT_LOCKED,
J
Jeff Mahoney 已提交
1335
				       EXTENT_LOCKED, &failed_start,
1336
				       cached_state, GFP_NOFS, NULL);
1337
		if (err == -EEXIST) {
1338 1339
			wait_extent_bit(tree, failed_start, end, EXTENT_LOCKED);
			start = failed_start;
1340
		} else
1341 1342 1343 1344 1345 1346
			break;
		WARN_ON(start > end);
	}
	return err;
}

1347
int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
1348 1349 1350 1351
{
	int err;
	u64 failed_start;

J
Jeff Mahoney 已提交
1352
	err = __set_extent_bit(tree, start, end, EXTENT_LOCKED, EXTENT_LOCKED,
1353
			       &failed_start, NULL, GFP_NOFS, NULL);
Y
Yan Zheng 已提交
1354 1355 1356
	if (err == -EEXIST) {
		if (failed_start > start)
			clear_extent_bit(tree, start, failed_start - 1,
1357
					 EXTENT_LOCKED, 1, 0, NULL, GFP_NOFS);
1358
		return 0;
Y
Yan Zheng 已提交
1359
	}
1360 1361 1362
	return 1;
}

1363
void extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end)
1364
{
1365 1366
	unsigned long index = start >> PAGE_SHIFT;
	unsigned long end_index = end >> PAGE_SHIFT;
1367 1368 1369 1370 1371 1372
	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);
1373
		put_page(page);
1374 1375 1376 1377
		index++;
	}
}

1378
void extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end)
1379
{
1380 1381
	unsigned long index = start >> PAGE_SHIFT;
	unsigned long end_index = end >> PAGE_SHIFT;
1382 1383 1384 1385 1386 1387
	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);
1388
		account_page_redirty(page);
1389
		put_page(page);
1390 1391 1392 1393
		index++;
	}
}

1394 1395 1396
/*
 * helper function to set both pages and extents in the tree writeback
 */
1397
static void set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end)
1398
{
1399 1400
	unsigned long index = start >> PAGE_SHIFT;
	unsigned long end_index = end >> PAGE_SHIFT;
1401 1402 1403 1404
	struct page *page;

	while (index <= end_index) {
		page = find_get_page(tree->mapping, index);
1405
		BUG_ON(!page); /* Pages should be in the extent_io_tree */
1406
		set_page_writeback(page);
1407
		put_page(page);
1408 1409 1410 1411
		index++;
	}
}

C
Chris Mason 已提交
1412 1413 1414 1415
/* find the first state struct with 'bits' set after 'start', and
 * return it.  tree->lock must be held.  NULL will returned if
 * nothing was found after 'start'
 */
1416 1417
static struct extent_state *
find_first_extent_bit_state(struct extent_io_tree *tree,
1418
			    u64 start, unsigned bits)
C
Chris Mason 已提交
1419 1420 1421 1422 1423 1424 1425 1426 1427
{
	struct rb_node *node;
	struct extent_state *state;

	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
	node = tree_search(tree, start);
C
Chris Mason 已提交
1428
	if (!node)
C
Chris Mason 已提交
1429 1430
		goto out;

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

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

1444 1445 1446 1447 1448
/*
 * find the first offset in the io tree with 'bits' set. zero is
 * returned if we find something, and *start_ret and *end_ret are
 * set to reflect the state struct that was found.
 *
1449
 * If nothing was found, 1 is returned. If found something, return 0.
1450 1451
 */
int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
1452
			  u64 *start_ret, u64 *end_ret, unsigned bits,
1453
			  struct extent_state **cached_state)
1454 1455
{
	struct extent_state *state;
1456
	struct rb_node *n;
1457 1458 1459
	int ret = 1;

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

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

C
Chris Mason 已提交
1492 1493 1494 1495 1496 1497
/*
 * find a contiguous range of bytes in the file marked as delalloc, not
 * more than 'max_bytes'.  start and end are used to return the range,
 *
 * 1 is returned if we find something, 0 if nothing was in the tree
 */
C
Chris Mason 已提交
1498
static noinline u64 find_delalloc_range(struct extent_io_tree *tree,
1499 1500
					u64 *start, u64 *end, u64 max_bytes,
					struct extent_state **cached_state)
1501 1502 1503 1504 1505 1506 1507
{
	struct rb_node *node;
	struct extent_state *state;
	u64 cur_start = *start;
	u64 found = 0;
	u64 total_bytes = 0;

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

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

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

1552 1553 1554
static noinline void __unlock_for_delalloc(struct inode *inode,
					   struct page *locked_page,
					   u64 start, u64 end)
C
Chris Mason 已提交
1555 1556 1557
{
	int ret;
	struct page *pages[16];
1558 1559
	unsigned long index = start >> PAGE_SHIFT;
	unsigned long end_index = end >> PAGE_SHIFT;
C
Chris Mason 已提交
1560 1561 1562 1563
	unsigned long nr_pages = end_index - index + 1;
	int i;

	if (index == locked_page->index && end_index == index)
1564
		return;
C
Chris Mason 已提交
1565

C
Chris Mason 已提交
1566
	while (nr_pages > 0) {
C
Chris Mason 已提交
1567
		ret = find_get_pages_contig(inode->i_mapping, index,
1568 1569
				     min_t(unsigned long, nr_pages,
				     ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1570 1571 1572
		for (i = 0; i < ret; i++) {
			if (pages[i] != locked_page)
				unlock_page(pages[i]);
1573
			put_page(pages[i]);
C
Chris Mason 已提交
1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585
		}
		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)
{
1586
	unsigned long index = delalloc_start >> PAGE_SHIFT;
C
Chris Mason 已提交
1587
	unsigned long start_index = index;
1588
	unsigned long end_index = delalloc_end >> PAGE_SHIFT;
C
Chris Mason 已提交
1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600
	unsigned long pages_locked = 0;
	struct page *pages[16];
	unsigned long nrpages;
	int ret;
	int i;

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

	/* skip the page at the start index */
	nrpages = end_index - index + 1;
C
Chris Mason 已提交
1601
	while (nrpages > 0) {
C
Chris Mason 已提交
1602
		ret = find_get_pages_contig(inode->i_mapping, index,
1603 1604
				     min_t(unsigned long,
				     nrpages, ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1605 1606 1607 1608 1609 1610 1611 1612 1613 1614
		if (ret == 0) {
			ret = -EAGAIN;
			goto done;
		}
		/* now we have an array of pages, lock them all */
		for (i = 0; i < ret; i++) {
			/*
			 * the caller is taking responsibility for
			 * locked_page
			 */
1615
			if (pages[i] != locked_page) {
C
Chris Mason 已提交
1616
				lock_page(pages[i]);
1617 1618
				if (!PageDirty(pages[i]) ||
				    pages[i]->mapping != inode->i_mapping) {
1619 1620
					ret = -EAGAIN;
					unlock_page(pages[i]);
1621
					put_page(pages[i]);
1622 1623 1624
					goto done;
				}
			}
1625
			put_page(pages[i]);
1626
			pages_locked++;
C
Chris Mason 已提交
1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637
		}
		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)) <<
1638
			      PAGE_SHIFT);
C
Chris Mason 已提交
1639 1640 1641 1642 1643 1644 1645 1646 1647 1648
	}
	return ret;
}

/*
 * find a contiguous range of bytes in the file marked as delalloc, not
 * more than 'max_bytes'.  start and end are used to return the range,
 *
 * 1 is returned if we find something, 0 if nothing was in the tree
 */
1649 1650 1651 1652
STATIC u64 find_lock_delalloc_range(struct inode *inode,
				    struct extent_io_tree *tree,
				    struct page *locked_page, u64 *start,
				    u64 *end, u64 max_bytes)
C
Chris Mason 已提交
1653 1654 1655 1656
{
	u64 delalloc_start;
	u64 delalloc_end;
	u64 found;
1657
	struct extent_state *cached_state = NULL;
C
Chris Mason 已提交
1658 1659 1660 1661 1662 1663 1664 1665
	int ret;
	int loops = 0;

again:
	/* step one, find a bunch of delalloc bytes starting at start */
	delalloc_start = *start;
	delalloc_end = 0;
	found = find_delalloc_range(tree, &delalloc_start, &delalloc_end,
1666
				    max_bytes, &cached_state);
C
Chris Mason 已提交
1667
	if (!found || delalloc_end <= *start) {
C
Chris Mason 已提交
1668 1669
		*start = delalloc_start;
		*end = delalloc_end;
1670
		free_extent_state(cached_state);
L
Liu Bo 已提交
1671
		return 0;
C
Chris Mason 已提交
1672 1673
	}

C
Chris Mason 已提交
1674 1675 1676 1677 1678
	/*
	 * start comes from the offset of locked_page.  We have to lock
	 * pages in order, so we can't process delalloc bytes before
	 * locked_page
	 */
C
Chris Mason 已提交
1679
	if (delalloc_start < *start)
C
Chris Mason 已提交
1680 1681
		delalloc_start = *start;

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

C
Chris Mason 已提交
1688 1689 1690 1691 1692 1693 1694
	/* step two, lock all the pages after the page that has start */
	ret = lock_delalloc_pages(inode, locked_page,
				  delalloc_start, delalloc_end);
	if (ret == -EAGAIN) {
		/* some of the pages are gone, lets avoid looping by
		 * shortening the size of the delalloc range we're searching
		 */
1695
		free_extent_state(cached_state);
1696
		cached_state = NULL;
C
Chris Mason 已提交
1697
		if (!loops) {
1698
			max_bytes = PAGE_SIZE;
C
Chris Mason 已提交
1699 1700 1701 1702 1703 1704 1705
			loops = 1;
			goto again;
		} else {
			found = 0;
			goto out_failed;
		}
	}
1706
	BUG_ON(ret); /* Only valid values are 0 and -EAGAIN */
C
Chris Mason 已提交
1707 1708

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

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

1729
void extent_clear_unlock_delalloc(struct inode *inode, u64 start, u64 end,
1730
				 struct page *locked_page,
1731
				 unsigned clear_bits,
1732
				 unsigned long page_ops)
C
Chris Mason 已提交
1733
{
1734
	struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
C
Chris Mason 已提交
1735 1736
	int ret;
	struct page *pages[16];
1737 1738
	unsigned long index = start >> PAGE_SHIFT;
	unsigned long end_index = end >> PAGE_SHIFT;
C
Chris Mason 已提交
1739 1740
	unsigned long nr_pages = end_index - index + 1;
	int i;
1741

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

1746 1747 1748
	if ((page_ops & PAGE_SET_ERROR) && nr_pages > 0)
		mapping_set_error(inode->i_mapping, -EIO);

C
Chris Mason 已提交
1749
	while (nr_pages > 0) {
C
Chris Mason 已提交
1750
		ret = find_get_pages_contig(inode->i_mapping, index,
1751 1752
				     min_t(unsigned long,
				     nr_pages, ARRAY_SIZE(pages)), pages);
C
Chris Mason 已提交
1753
		for (i = 0; i < ret; i++) {
1754

1755
			if (page_ops & PAGE_SET_PRIVATE2)
1756 1757
				SetPagePrivate2(pages[i]);

C
Chris Mason 已提交
1758
			if (pages[i] == locked_page) {
1759
				put_page(pages[i]);
C
Chris Mason 已提交
1760 1761
				continue;
			}
1762
			if (page_ops & PAGE_CLEAR_DIRTY)
C
Chris Mason 已提交
1763
				clear_page_dirty_for_io(pages[i]);
1764
			if (page_ops & PAGE_SET_WRITEBACK)
C
Chris Mason 已提交
1765
				set_page_writeback(pages[i]);
1766 1767
			if (page_ops & PAGE_SET_ERROR)
				SetPageError(pages[i]);
1768
			if (page_ops & PAGE_END_WRITEBACK)
C
Chris Mason 已提交
1769
				end_page_writeback(pages[i]);
1770
			if (page_ops & PAGE_UNLOCK)
1771
				unlock_page(pages[i]);
1772
			put_page(pages[i]);
C
Chris Mason 已提交
1773 1774 1775 1776 1777 1778 1779
		}
		nr_pages -= ret;
		index += ret;
		cond_resched();
	}
}

C
Chris Mason 已提交
1780 1781 1782 1783 1784
/*
 * 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.
 */
1785 1786
u64 count_range_bits(struct extent_io_tree *tree,
		     u64 *start, u64 search_end, u64 max_bytes,
1787
		     unsigned bits, int contig)
1788 1789 1790 1791 1792
{
	struct rb_node *node;
	struct extent_state *state;
	u64 cur_start = *start;
	u64 total_bytes = 0;
1793
	u64 last = 0;
1794 1795
	int found = 0;

1796
	if (WARN_ON(search_end <= cur_start))
1797 1798
		return 0;

1799
	spin_lock(&tree->lock);
1800 1801 1802 1803 1804 1805 1806 1807
	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.
	 */
1808
	node = tree_search(tree, cur_start);
C
Chris Mason 已提交
1809
	if (!node)
1810 1811
		goto out;

C
Chris Mason 已提交
1812
	while (1) {
1813 1814 1815
		state = rb_entry(node, struct extent_state, rb_node);
		if (state->start > search_end)
			break;
1816 1817 1818
		if (contig && found && state->start > last + 1)
			break;
		if (state->end >= cur_start && (state->state & bits) == bits) {
1819 1820 1821 1822 1823
			total_bytes += min(search_end, state->end) + 1 -
				       max(cur_start, state->start);
			if (total_bytes >= max_bytes)
				break;
			if (!found) {
1824
				*start = max(cur_start, state->start);
1825 1826
				found = 1;
			}
1827 1828 1829
			last = state->end;
		} else if (contig && found) {
			break;
1830 1831 1832 1833 1834 1835
		}
		node = rb_next(node);
		if (!node)
			break;
	}
out:
1836
	spin_unlock(&tree->lock);
1837 1838
	return total_bytes;
}
1839

C
Chris Mason 已提交
1840 1841 1842 1843
/*
 * set the private field for a given byte offset in the tree.  If there isn't
 * an extent_state there already, this does nothing.
 */
1844
static noinline int set_state_failrec(struct extent_io_tree *tree, u64 start,
1845
		struct io_failure_record *failrec)
1846 1847 1848 1849 1850
{
	struct rb_node *node;
	struct extent_state *state;
	int ret = 0;

1851
	spin_lock(&tree->lock);
1852 1853 1854 1855
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1856
	node = tree_search(tree, start);
1857
	if (!node) {
1858 1859 1860 1861 1862 1863 1864 1865
		ret = -ENOENT;
		goto out;
	}
	state = rb_entry(node, struct extent_state, rb_node);
	if (state->start != start) {
		ret = -ENOENT;
		goto out;
	}
1866
	state->failrec = failrec;
1867
out:
1868
	spin_unlock(&tree->lock);
1869 1870 1871
	return ret;
}

1872
static noinline int get_state_failrec(struct extent_io_tree *tree, u64 start,
1873
		struct io_failure_record **failrec)
1874 1875 1876 1877 1878
{
	struct rb_node *node;
	struct extent_state *state;
	int ret = 0;

1879
	spin_lock(&tree->lock);
1880 1881 1882 1883
	/*
	 * this search will find all the extents that end after
	 * our range starts.
	 */
1884
	node = tree_search(tree, start);
1885
	if (!node) {
1886 1887 1888 1889 1890 1891 1892 1893
		ret = -ENOENT;
		goto out;
	}
	state = rb_entry(node, struct extent_state, rb_node);
	if (state->start != start) {
		ret = -ENOENT;
		goto out;
	}
1894
	*failrec = state->failrec;
1895
out:
1896
	spin_unlock(&tree->lock);
1897 1898 1899 1900 1901
	return ret;
}

/*
 * searches a range in the state tree for a given mask.
1902
 * If 'filled' == 1, this returns 1 only if every extent in the tree
1903 1904 1905 1906
 * 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,
1907
		   unsigned bits, int filled, struct extent_state *cached)
1908 1909 1910 1911 1912
{
	struct extent_state *state = NULL;
	struct rb_node *node;
	int bitset = 0;

1913
	spin_lock(&tree->lock);
1914
	if (cached && extent_state_in_tree(cached) && cached->start <= start &&
1915
	    cached->end > start)
1916 1917 1918
		node = &cached->rb_node;
	else
		node = tree_search(tree, start);
1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937
	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;
		}
1938 1939 1940 1941

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

1942 1943 1944 1945 1946 1947 1948 1949 1950 1951
		start = state->end + 1;
		if (start > end)
			break;
		node = rb_next(node);
		if (!node) {
			if (filled)
				bitset = 0;
			break;
		}
	}
1952
	spin_unlock(&tree->lock);
1953 1954 1955 1956 1957 1958 1959
	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
 */
1960
static void check_page_uptodate(struct extent_io_tree *tree, struct page *page)
1961
{
M
Miao Xie 已提交
1962
	u64 start = page_offset(page);
1963
	u64 end = start + PAGE_SIZE - 1;
1964
	if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1, NULL))
1965 1966 1967
		SetPageUptodate(page);
}

1968
int free_io_failure(struct inode *inode, struct io_failure_record *rec)
1969 1970 1971 1972 1973
{
	int ret;
	int err = 0;
	struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;

1974
	set_state_failrec(failure_tree, rec->start, NULL);
1975 1976
	ret = clear_extent_bits(failure_tree, rec->start,
				rec->start + rec->len - 1,
1977
				EXTENT_LOCKED | EXTENT_DIRTY);
1978 1979 1980
	if (ret)
		err = ret;

D
David Woodhouse 已提交
1981 1982
	ret = clear_extent_bits(&BTRFS_I(inode)->io_tree, rec->start,
				rec->start + rec->len - 1,
1983
				EXTENT_DAMAGED);
D
David Woodhouse 已提交
1984 1985
	if (ret && !err)
		err = ret;
1986 1987 1988 1989 1990 1991 1992 1993 1994 1995

	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.
1996
 * to avoid any synchronization issues, wait for the data after writing, which
1997 1998 1999 2000
 * 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.
 */
2001 2002
int repair_io_failure(struct inode *inode, u64 start, u64 length, u64 logical,
		      struct page *page, unsigned int pg_offset, int mirror_num)
2003
{
2004
	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2005 2006 2007 2008 2009
	struct bio *bio;
	struct btrfs_device *dev;
	u64 map_length = 0;
	u64 sector;
	struct btrfs_bio *bbio = NULL;
D
David Woodhouse 已提交
2010
	struct btrfs_mapping_tree *map_tree = &fs_info->mapping_tree;
2011 2012
	int ret;

2013
	ASSERT(!(fs_info->sb->s_flags & MS_RDONLY));
2014 2015
	BUG_ON(!mirror_num);

D
David Woodhouse 已提交
2016 2017 2018 2019
	/* we can't repair anything in raid56 yet */
	if (btrfs_is_parity_mirror(map_tree, logical, length, mirror_num))
		return 0;

2020
	bio = btrfs_io_bio_alloc(GFP_NOFS, 1);
2021 2022
	if (!bio)
		return -EIO;
2023
	bio->bi_iter.bi_size = 0;
2024 2025
	map_length = length;

2026
	ret = btrfs_map_block(fs_info, WRITE, logical,
2027 2028 2029 2030 2031 2032 2033
			      &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;
2034
	bio->bi_iter.bi_sector = sector;
2035
	dev = bbio->stripes[mirror_num-1].dev;
2036
	btrfs_put_bbio(bbio);
2037 2038 2039 2040 2041
	if (!dev || !dev->bdev || !dev->writeable) {
		bio_put(bio);
		return -EIO;
	}
	bio->bi_bdev = dev->bdev;
2042
	bio_add_page(bio, page, length, pg_offset);
2043

2044
	if (btrfsic_submit_bio_wait(WRITE_SYNC, bio)) {
2045 2046
		/* try to remap that extent elsewhere? */
		bio_put(bio);
2047
		btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_WRITE_ERRS);
2048 2049 2050
		return -EIO;
	}

2051 2052
	btrfs_info_rl_in_rcu(fs_info,
		"read error corrected: ino %llu off %llu (dev %s sector %llu)",
2053 2054
				  btrfs_ino(inode), start,
				  rcu_str_deref(dev->name), sector);
2055 2056 2057 2058
	bio_put(bio);
	return 0;
}

2059 2060 2061 2062 2063
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);
2064
	int ret = 0;
2065

2066 2067 2068
	if (root->fs_info->sb->s_flags & MS_RDONLY)
		return -EROFS;

2069
	for (i = 0; i < num_pages; i++) {
2070
		struct page *p = eb->pages[i];
2071 2072

		ret = repair_io_failure(root->fs_info->btree_inode, start,
2073
					PAGE_SIZE, start, p,
2074
					start - page_offset(p), mirror_num);
2075 2076
		if (ret)
			break;
2077
		start += PAGE_SIZE;
2078 2079 2080 2081 2082
	}

	return ret;
}

2083 2084 2085 2086
/*
 * 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
 */
2087 2088
int clean_io_failure(struct inode *inode, u64 start, struct page *page,
		     unsigned int pg_offset)
2089 2090 2091
{
	u64 private;
	struct io_failure_record *failrec;
2092
	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2093 2094 2095 2096 2097 2098 2099 2100 2101 2102
	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;

2103 2104
	ret = get_state_failrec(&BTRFS_I(inode)->io_failure_tree, start,
			&failrec);
2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115
	if (ret)
		return 0;

	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;
	}
2116 2117
	if (fs_info->sb->s_flags & MS_RDONLY)
		goto out;
2118 2119 2120 2121 2122 2123 2124

	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);

2125 2126
	if (state && state->start <= failrec->start &&
	    state->end >= failrec->start + failrec->len - 1) {
2127 2128
		num_copies = btrfs_num_copies(fs_info, failrec->logical,
					      failrec->len);
2129
		if (num_copies > 1)  {
2130
			repair_io_failure(inode, start, failrec->len,
2131
					  failrec->logical, page,
2132
					  pg_offset, failrec->failed_mirror);
2133 2134 2135 2136
		}
	}

out:
2137
	free_io_failure(inode, failrec);
2138

2139
	return 0;
2140 2141
}

2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166
/*
 * 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);

2167
		failrec = state->failrec;
2168 2169 2170 2171 2172 2173 2174 2175
		free_extent_state(state);
		kfree(failrec);

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

2176
int btrfs_get_io_failure_record(struct inode *inode, u64 start, u64 end,
2177
		struct io_failure_record **failrec_ret)
2178
{
2179
	struct io_failure_record *failrec;
2180 2181 2182 2183 2184 2185 2186
	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;

2187
	ret = get_state_failrec(failure_tree, start, &failrec);
2188 2189 2190 2191
	if (ret) {
		failrec = kzalloc(sizeof(*failrec), GFP_NOFS);
		if (!failrec)
			return -ENOMEM;
2192

2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206
		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;
		}

2207
		if (em->start > start || em->start + em->len <= start) {
2208 2209 2210 2211
			free_extent_map(em);
			em = NULL;
		}
		read_unlock(&em_tree->lock);
2212
		if (!em) {
2213 2214 2215
			kfree(failrec);
			return -EIO;
		}
2216

2217 2218 2219 2220 2221 2222 2223 2224
		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);
		}
2225 2226 2227 2228

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

2229 2230 2231 2232 2233
		failrec->logical = logical;
		free_extent_map(em);

		/* set the bits in the private failure tree */
		ret = set_extent_bits(failure_tree, start, end,
2234
					EXTENT_LOCKED | EXTENT_DIRTY);
2235
		if (ret >= 0)
2236
			ret = set_state_failrec(failure_tree, start, failrec);
2237 2238
		/* set the bits in the inode's tree */
		if (ret >= 0)
2239
			ret = set_extent_bits(tree, start, end, EXTENT_DAMAGED);
2240 2241 2242 2243 2244
		if (ret < 0) {
			kfree(failrec);
			return ret;
		}
	} else {
2245
		pr_debug("Get IO Failure Record: (found) logical=%llu, start=%llu, len=%llu, validation=%d\n",
2246 2247 2248 2249 2250 2251 2252 2253
			 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.
		 */
	}
2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264

	*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;

2265 2266
	num_copies = btrfs_num_copies(BTRFS_I(inode)->root->fs_info,
				      failrec->logical, failrec->len);
2267 2268 2269 2270 2271 2272
	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.
		 */
2273
		pr_debug("Check Repairable: cannot repair, num_copies=%d, next_mirror %d, failed_mirror %d\n",
2274
			 num_copies, failrec->this_mirror, failed_mirror);
2275
		return 0;
2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311
	}

	/*
	 * 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++;
	}

2312
	if (failrec->this_mirror > num_copies) {
2313
		pr_debug("Check Repairable: (fail) num_copies=%d, next_mirror %d, failed_mirror %d\n",
2314
			 num_copies, failrec->this_mirror, failed_mirror);
2315
		return 0;
2316 2317
	}

2318 2319 2320 2321 2322 2323 2324
	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,
2325
				    bio_end_io_t *endio_func, void *data)
2326 2327 2328 2329 2330
{
	struct bio *bio;
	struct btrfs_io_bio *btrfs_failed_bio;
	struct btrfs_io_bio *btrfs_bio;

2331
	bio = btrfs_io_bio_alloc(GFP_NOFS, 1);
2332 2333 2334 2335
	if (!bio)
		return NULL;

	bio->bi_end_io = endio_func;
2336
	bio->bi_iter.bi_sector = failrec->logical >> 9;
2337
	bio->bi_bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
2338
	bio->bi_iter.bi_size = 0;
2339
	bio->bi_private = data;
2340

2341 2342 2343 2344 2345 2346 2347
	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;
2348 2349
		icsum *= csum_size;
		memcpy(btrfs_bio->csum, btrfs_failed_bio->csum + icsum,
2350 2351 2352
		       csum_size);
	}

2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396
	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),
2397 2398
				      (int)phy_offset, failed_bio->bi_end_io,
				      NULL);
2399 2400 2401 2402
	if (!bio) {
		free_io_failure(inode, failrec);
		return -EIO;
	}
2403

2404 2405
	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);
2406

2407 2408 2409
	ret = tree->ops->submit_bio_hook(inode, read_mode, bio,
					 failrec->this_mirror,
					 failrec->bio_flags, 0);
2410
	if (ret) {
2411
		free_io_failure(inode, failrec);
2412 2413 2414
		bio_put(bio);
	}

2415
	return ret;
2416 2417
}

2418 2419
/* lots and lots of room for performance fixes in the end_bio funcs */

2420
void end_extent_writepage(struct page *page, int err, u64 start, u64 end)
2421 2422 2423
{
	int uptodate = (err == 0);
	struct extent_io_tree *tree;
2424
	int ret = 0;
2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437

	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);
2438 2439
		ret = ret < 0 ? ret : -EIO;
		mapping_set_error(page->mapping, ret);
2440 2441 2442
	}
}

2443 2444 2445 2446 2447 2448 2449 2450 2451
/*
 * 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.
 */
2452
static void end_bio_extent_writepage(struct bio *bio)
2453
{
2454
	struct bio_vec *bvec;
2455 2456
	u64 start;
	u64 end;
2457
	int i;
2458

2459
	bio_for_each_segment_all(bvec, bio, i) {
2460
		struct page *page = bvec->bv_page;
2461

2462 2463 2464 2465 2466
		/* 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.  */
2467 2468
		if (bvec->bv_offset || bvec->bv_len != PAGE_SIZE) {
			if (bvec->bv_offset + bvec->bv_len != PAGE_SIZE)
2469 2470 2471 2472 2473 2474 2475 2476 2477
				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);
		}
2478

2479 2480
		start = page_offset(page);
		end = start + bvec->bv_offset + bvec->bv_len - 1;
2481

2482
		end_extent_writepage(page, bio->bi_error, start, end);
2483
		end_page_writeback(page);
2484
	}
2485

2486 2487 2488
	bio_put(bio);
}

2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500
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);
}

2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511
/*
 * 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.
 */
2512
static void end_bio_extent_readpage(struct bio *bio)
2513
{
2514
	struct bio_vec *bvec;
2515
	int uptodate = !bio->bi_error;
2516
	struct btrfs_io_bio *io_bio = btrfs_io_bio(bio);
2517
	struct extent_io_tree *tree;
2518
	u64 offset = 0;
2519 2520
	u64 start;
	u64 end;
2521
	u64 len;
2522 2523
	u64 extent_start = 0;
	u64 extent_len = 0;
2524
	int mirror;
2525
	int ret;
2526
	int i;
2527

2528
	bio_for_each_segment_all(bvec, bio, i) {
2529
		struct page *page = bvec->bv_page;
2530
		struct inode *inode = page->mapping->host;
2531

2532
		pr_debug("end_bio_extent_readpage: bi_sector=%llu, err=%d, "
2533 2534
			 "mirror=%u\n", (u64)bio->bi_iter.bi_sector,
			 bio->bi_error, io_bio->mirror_num);
2535
		tree = &BTRFS_I(inode)->io_tree;
2536

2537 2538 2539 2540 2541
		/* 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.  */
2542 2543
		if (bvec->bv_offset || bvec->bv_len != PAGE_SIZE) {
			if (bvec->bv_offset + bvec->bv_len != PAGE_SIZE)
2544 2545 2546 2547 2548 2549 2550 2551 2552
				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);
		}
2553

2554 2555
		start = page_offset(page);
		end = start + bvec->bv_offset + bvec->bv_len - 1;
2556
		len = bvec->bv_len;
2557

2558
		mirror = io_bio->mirror_num;
2559 2560
		if (likely(uptodate && tree->ops &&
			   tree->ops->readpage_end_io_hook)) {
2561 2562 2563
			ret = tree->ops->readpage_end_io_hook(io_bio, offset,
							      page, start, end,
							      mirror);
2564
			if (ret)
2565
				uptodate = 0;
2566
			else
2567
				clean_io_failure(inode, start, page, 0);
2568
		}
2569

2570 2571 2572 2573
		if (likely(uptodate))
			goto readpage_ok;

		if (tree->ops && tree->ops->readpage_io_failed_hook) {
2574
			ret = tree->ops->readpage_io_failed_hook(page, mirror);
2575
			if (!ret && !bio->bi_error)
2576
				uptodate = 1;
2577
		} else {
2578 2579 2580 2581 2582 2583 2584 2585 2586 2587
			/*
			 * 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.
			 */
2588 2589
			ret = bio_readpage_error(bio, offset, page, start, end,
						 mirror);
2590
			if (ret == 0) {
2591
				uptodate = !bio->bi_error;
2592
				offset += len;
2593 2594 2595
				continue;
			}
		}
2596
readpage_ok:
2597
		if (likely(uptodate)) {
2598
			loff_t i_size = i_size_read(inode);
2599
			pgoff_t end_index = i_size >> PAGE_SHIFT;
2600
			unsigned off;
2601 2602

			/* Zero out the end if this page straddles i_size */
2603
			off = i_size & (PAGE_SIZE-1);
2604
			if (page->index == end_index && off)
2605
				zero_user_segment(page, off, PAGE_SIZE);
2606
			SetPageUptodate(page);
2607
		} else {
2608 2609
			ClearPageUptodate(page);
			SetPageError(page);
2610
		}
2611
		unlock_page(page);
2612
		offset += len;
2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634

		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;
		}
2635
	}
2636

2637 2638 2639
	if (extent_len)
		endio_readpage_release_extent(tree, extent_start, extent_len,
					      uptodate);
2640
	if (io_bio->end_io)
2641
		io_bio->end_io(io_bio, bio->bi_error);
2642 2643 2644
	bio_put(bio);
}

2645 2646 2647 2648
/*
 * 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
 */
2649 2650 2651
struct bio *
btrfs_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
		gfp_t gfp_flags)
2652
{
2653
	struct btrfs_io_bio *btrfs_bio;
2654 2655
	struct bio *bio;

2656
	bio = bio_alloc_bioset(gfp_flags, nr_vecs, btrfs_bioset);
2657 2658

	if (bio == NULL && (current->flags & PF_MEMALLOC)) {
2659 2660 2661 2662
		while (!bio && (nr_vecs /= 2)) {
			bio = bio_alloc_bioset(gfp_flags,
					       nr_vecs, btrfs_bioset);
		}
2663 2664 2665 2666
	}

	if (bio) {
		bio->bi_bdev = bdev;
2667
		bio->bi_iter.bi_sector = first_sector;
2668 2669 2670 2671
		btrfs_bio = btrfs_io_bio(bio);
		btrfs_bio->csum = NULL;
		btrfs_bio->csum_allocated = NULL;
		btrfs_bio->end_io = NULL;
2672 2673 2674 2675
	}
	return bio;
}

2676 2677
struct bio *btrfs_bio_clone(struct bio *bio, gfp_t gfp_mask)
{
2678 2679
	struct btrfs_io_bio *btrfs_bio;
	struct bio *new;
2680

2681 2682 2683 2684 2685 2686
	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;
2687 2688

#ifdef CONFIG_BLK_CGROUP
2689 2690 2691
		/* FIXME, put this into bio_clone_bioset */
		if (bio->bi_css)
			bio_associate_blkcg(new, bio->bi_css);
2692
#endif
2693 2694 2695
	}
	return new;
}
2696 2697 2698 2699

/* this also allocates from the btrfs_bioset */
struct bio *btrfs_io_bio_alloc(gfp_t gfp_mask, unsigned int nr_iovecs)
{
2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710
	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;
2711 2712 2713
}


2714 2715
static int __must_check submit_one_bio(int rw, struct bio *bio,
				       int mirror_num, unsigned long bio_flags)
2716 2717
{
	int ret = 0;
2718 2719 2720 2721 2722
	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
2723
	start = page_offset(page) + bvec->bv_offset;
2724

2725
	bio->bi_private = NULL;
2726 2727 2728

	bio_get(bio);

2729
	if (tree->ops && tree->ops->submit_bio_hook)
2730
		ret = tree->ops->submit_bio_hook(page->mapping->host, rw, bio,
2731
					   mirror_num, bio_flags, start);
2732
	else
2733
		btrfsic_submit_bio(rw, bio);
2734

2735 2736 2737 2738
	bio_put(bio);
	return ret;
}

2739
static int merge_bio(int rw, struct extent_io_tree *tree, struct page *page,
2740 2741 2742 2743 2744
		     unsigned long offset, size_t size, struct bio *bio,
		     unsigned long bio_flags)
{
	int ret = 0;
	if (tree->ops && tree->ops->merge_bio_hook)
2745
		ret = tree->ops->merge_bio_hook(rw, page, offset, size, bio,
2746 2747 2748 2749 2750 2751
						bio_flags);
	BUG_ON(ret < 0);
	return ret;

}

2752
static int submit_extent_page(int rw, struct extent_io_tree *tree,
2753
			      struct writeback_control *wbc,
2754 2755 2756 2757 2758
			      struct page *page, sector_t sector,
			      size_t size, unsigned long offset,
			      struct block_device *bdev,
			      struct bio **bio_ret,
			      unsigned long max_pages,
2759
			      bio_end_io_t end_io_func,
C
Chris Mason 已提交
2760 2761
			      int mirror_num,
			      unsigned long prev_bio_flags,
2762 2763
			      unsigned long bio_flags,
			      bool force_bio_submit)
2764 2765 2766
{
	int ret = 0;
	struct bio *bio;
C
Chris Mason 已提交
2767 2768
	int contig = 0;
	int old_compressed = prev_bio_flags & EXTENT_BIO_COMPRESSED;
2769
	size_t page_size = min_t(size_t, size, PAGE_SIZE);
2770 2771 2772

	if (bio_ret && *bio_ret) {
		bio = *bio_ret;
C
Chris Mason 已提交
2773
		if (old_compressed)
2774
			contig = bio->bi_iter.bi_sector == sector;
C
Chris Mason 已提交
2775
		else
K
Kent Overstreet 已提交
2776
			contig = bio_end_sector(bio) == sector;
C
Chris Mason 已提交
2777 2778

		if (prev_bio_flags != bio_flags || !contig ||
2779
		    force_bio_submit ||
2780
		    merge_bio(rw, tree, page, offset, page_size, bio, bio_flags) ||
C
Chris Mason 已提交
2781 2782 2783
		    bio_add_page(bio, page, page_size, offset) < page_size) {
			ret = submit_one_bio(rw, bio, mirror_num,
					     prev_bio_flags);
2784 2785
			if (ret < 0) {
				*bio_ret = NULL;
2786
				return ret;
2787
			}
2788 2789
			bio = NULL;
		} else {
2790 2791
			if (wbc)
				wbc_account_io(wbc, page, page_size);
2792 2793 2794
			return 0;
		}
	}
C
Chris Mason 已提交
2795

2796 2797
	bio = btrfs_bio_alloc(bdev, sector, BIO_MAX_PAGES,
			GFP_NOFS | __GFP_HIGH);
2798 2799
	if (!bio)
		return -ENOMEM;
2800

C
Chris Mason 已提交
2801
	bio_add_page(bio, page, page_size, offset);
2802 2803
	bio->bi_end_io = end_io_func;
	bio->bi_private = tree;
2804 2805 2806 2807
	if (wbc) {
		wbc_init_bio(wbc, bio);
		wbc_account_io(wbc, page, page_size);
	}
2808

C
Chris Mason 已提交
2809
	if (bio_ret)
2810
		*bio_ret = bio;
C
Chris Mason 已提交
2811
	else
C
Chris Mason 已提交
2812
		ret = submit_one_bio(rw, bio, mirror_num, bio_flags);
2813 2814 2815 2816

	return ret;
}

2817 2818
static void attach_extent_buffer_page(struct extent_buffer *eb,
				      struct page *page)
2819 2820 2821
{
	if (!PagePrivate(page)) {
		SetPagePrivate(page);
2822
		get_page(page);
J
Josef Bacik 已提交
2823 2824 2825
		set_page_private(page, (unsigned long)eb);
	} else {
		WARN_ON(page->private != (unsigned long)eb);
2826 2827 2828
	}
}

J
Josef Bacik 已提交
2829
void set_page_extent_mapped(struct page *page)
2830
{
J
Josef Bacik 已提交
2831 2832
	if (!PagePrivate(page)) {
		SetPagePrivate(page);
2833
		get_page(page);
J
Josef Bacik 已提交
2834 2835
		set_page_private(page, EXTENT_PAGE_PRIVATE);
	}
2836 2837
}

2838 2839 2840 2841 2842 2843 2844 2845 2846
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;
2847
		if (extent_map_in_tree(em) && start >= em->start &&
2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864
		    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;
}
2865 2866 2867 2868
/*
 * 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)
2869
 * XXX JDM: This needs looking at to ensure proper page locking
2870
 */
2871 2872 2873
static int __do_readpage(struct extent_io_tree *tree,
			 struct page *page,
			 get_extent_t *get_extent,
2874
			 struct extent_map **em_cached,
2875
			 struct bio **bio, int mirror_num,
2876 2877
			 unsigned long *bio_flags, int rw,
			 u64 *prev_em_start)
2878 2879
{
	struct inode *inode = page->mapping->host;
M
Miao Xie 已提交
2880
	u64 start = page_offset(page);
2881
	u64 page_end = start + PAGE_SIZE - 1;
2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892
	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;
2893
	size_t pg_offset = 0;
2894
	size_t iosize;
C
Chris Mason 已提交
2895
	size_t disk_io_size;
2896
	size_t blocksize = inode->i_sb->s_blocksize;
2897
	unsigned long this_bio_flag = 0;
2898 2899 2900

	set_page_extent_mapped(page);

2901
	end = page_end;
D
Dan Magenheimer 已提交
2902 2903 2904
	if (!PageUptodate(page)) {
		if (cleancache_get_page(page) == 0) {
			BUG_ON(blocksize != PAGE_SIZE);
2905
			unlock_extent(tree, start, end);
D
Dan Magenheimer 已提交
2906 2907 2908 2909
			goto out;
		}
	}

2910
	if (page->index == last_byte >> PAGE_SHIFT) {
C
Chris Mason 已提交
2911
		char *userpage;
2912
		size_t zero_offset = last_byte & (PAGE_SIZE - 1);
C
Chris Mason 已提交
2913 2914

		if (zero_offset) {
2915
			iosize = PAGE_SIZE - zero_offset;
2916
			userpage = kmap_atomic(page);
C
Chris Mason 已提交
2917 2918
			memset(userpage + zero_offset, 0, iosize);
			flush_dcache_page(page);
2919
			kunmap_atomic(userpage);
C
Chris Mason 已提交
2920 2921
		}
	}
2922
	while (cur <= end) {
2923
		unsigned long pnr = (last_byte >> PAGE_SHIFT) + 1;
2924
		bool force_bio_submit = false;
2925

2926 2927
		if (cur >= last_byte) {
			char *userpage;
2928 2929
			struct extent_state *cached = NULL;

2930
			iosize = PAGE_SIZE - pg_offset;
2931
			userpage = kmap_atomic(page);
2932
			memset(userpage + pg_offset, 0, iosize);
2933
			flush_dcache_page(page);
2934
			kunmap_atomic(userpage);
2935
			set_extent_uptodate(tree, cur, cur + iosize - 1,
2936
					    &cached, GFP_NOFS);
2937 2938 2939
			unlock_extent_cached(tree, cur,
					     cur + iosize - 1,
					     &cached, GFP_NOFS);
2940 2941
			break;
		}
2942 2943
		em = __get_extent_map(inode, page, pg_offset, cur,
				      end - cur + 1, get_extent, em_cached);
2944
		if (IS_ERR_OR_NULL(em)) {
2945
			SetPageError(page);
2946
			unlock_extent(tree, cur, end);
2947 2948 2949 2950 2951 2952
			break;
		}
		extent_offset = cur - em->start;
		BUG_ON(extent_map_end(em) <= cur);
		BUG_ON(end < cur);

2953
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
2954
			this_bio_flag |= EXTENT_BIO_COMPRESSED;
2955 2956 2957
			extent_set_compress_type(&this_bio_flag,
						 em->compress_type);
		}
C
Chris Mason 已提交
2958

2959 2960
		iosize = min(extent_map_end(em) - cur, end - cur + 1);
		cur_end = min(extent_map_end(em) - 1, end);
2961
		iosize = ALIGN(iosize, blocksize);
C
Chris Mason 已提交
2962 2963 2964 2965 2966 2967 2968
		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;
		}
2969 2970
		bdev = em->bdev;
		block_start = em->block_start;
Y
Yan Zheng 已提交
2971 2972
		if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			block_start = EXTENT_MAP_HOLE;
2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015

		/*
		 * 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;

3016 3017 3018 3019 3020 3021
		free_extent_map(em);
		em = NULL;

		/* we've found a hole, just zero and go on */
		if (block_start == EXTENT_MAP_HOLE) {
			char *userpage;
3022 3023
			struct extent_state *cached = NULL;

3024
			userpage = kmap_atomic(page);
3025
			memset(userpage + pg_offset, 0, iosize);
3026
			flush_dcache_page(page);
3027
			kunmap_atomic(userpage);
3028 3029

			set_extent_uptodate(tree, cur, cur + iosize - 1,
3030
					    &cached, GFP_NOFS);
3031 3032 3033
			unlock_extent_cached(tree, cur,
					     cur + iosize - 1,
					     &cached, GFP_NOFS);
3034
			cur = cur + iosize;
3035
			pg_offset += iosize;
3036 3037 3038
			continue;
		}
		/* the get_extent function already copied into the page */
3039 3040
		if (test_range_bit(tree, cur, cur_end,
				   EXTENT_UPTODATE, 1, NULL)) {
3041
			check_page_uptodate(tree, page);
3042
			unlock_extent(tree, cur, cur + iosize - 1);
3043
			cur = cur + iosize;
3044
			pg_offset += iosize;
3045 3046
			continue;
		}
3047 3048 3049 3050 3051
		/* we have an inline extent but it didn't get marked up
		 * to date.  Error out
		 */
		if (block_start == EXTENT_MAP_INLINE) {
			SetPageError(page);
3052
			unlock_extent(tree, cur, cur + iosize - 1);
3053
			cur = cur + iosize;
3054
			pg_offset += iosize;
3055 3056
			continue;
		}
3057

3058
		pnr -= page->index;
3059
		ret = submit_extent_page(rw, tree, NULL, page,
3060
					 sector, disk_io_size, pg_offset,
3061
					 bdev, bio, pnr,
C
Chris Mason 已提交
3062 3063
					 end_bio_extent_readpage, mirror_num,
					 *bio_flags,
3064 3065
					 this_bio_flag,
					 force_bio_submit);
3066 3067 3068 3069
		if (!ret) {
			nr++;
			*bio_flags = this_bio_flag;
		} else {
3070
			SetPageError(page);
3071
			unlock_extent(tree, cur, cur + iosize - 1);
3072
		}
3073
		cur = cur + iosize;
3074
		pg_offset += iosize;
3075
	}
D
Dan Magenheimer 已提交
3076
out:
3077 3078 3079 3080 3081 3082 3083 3084
	if (!nr) {
		if (!PageError(page))
			SetPageUptodate(page);
		unlock_page(page);
	}
	return 0;
}

3085 3086 3087 3088
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,
3089
					     struct extent_map **em_cached,
3090
					     struct bio **bio, int mirror_num,
3091 3092
					     unsigned long *bio_flags, int rw,
					     u64 *prev_em_start)
3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110
{
	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++) {
3111
		__do_readpage(tree, pages[index], get_extent, em_cached, bio,
3112
			      mirror_num, bio_flags, rw, prev_em_start);
3113
		put_page(pages[index]);
3114 3115 3116 3117 3118 3119
	}
}

static void __extent_readpages(struct extent_io_tree *tree,
			       struct page *pages[],
			       int nr_pages, get_extent_t *get_extent,
3120
			       struct extent_map **em_cached,
3121
			       struct bio **bio, int mirror_num,
3122 3123
			       unsigned long *bio_flags, int rw,
			       u64 *prev_em_start)
3124
{
3125
	u64 start = 0;
3126 3127 3128
	u64 end = 0;
	u64 page_start;
	int index;
3129
	int first_index = 0;
3130 3131 3132 3133 3134

	for (index = 0; index < nr_pages; index++) {
		page_start = page_offset(pages[index]);
		if (!end) {
			start = page_start;
3135
			end = start + PAGE_SIZE - 1;
3136 3137
			first_index = index;
		} else if (end + 1 == page_start) {
3138
			end += PAGE_SIZE;
3139 3140 3141
		} else {
			__do_contiguous_readpages(tree, &pages[first_index],
						  index - first_index, start,
3142 3143
						  end, get_extent, em_cached,
						  bio, mirror_num, bio_flags,
3144
						  rw, prev_em_start);
3145
			start = page_start;
3146
			end = start + PAGE_SIZE - 1;
3147 3148 3149 3150 3151 3152 3153
			first_index = index;
		}
	}

	if (end)
		__do_contiguous_readpages(tree, &pages[first_index],
					  index - first_index, start,
3154
					  end, get_extent, em_cached, bio,
3155 3156
					  mirror_num, bio_flags, rw,
					  prev_em_start);
3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167
}

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);
3168
	u64 end = start + PAGE_SIZE - 1;
3169 3170 3171 3172
	int ret;

	while (1) {
		lock_extent(tree, start, end);
3173
		ordered = btrfs_lookup_ordered_range(inode, start,
3174
						PAGE_SIZE);
3175 3176 3177 3178 3179 3180 3181
		if (!ordered)
			break;
		unlock_extent(tree, start, end);
		btrfs_start_ordered_extent(inode, ordered, 1);
		btrfs_put_ordered_extent(ordered);
	}

3182
	ret = __do_readpage(tree, page, get_extent, NULL, bio, mirror_num,
3183
			    bio_flags, rw, NULL);
3184 3185 3186
	return ret;
}

3187
int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
3188
			    get_extent_t *get_extent, int mirror_num)
3189 3190
{
	struct bio *bio = NULL;
C
Chris Mason 已提交
3191
	unsigned long bio_flags = 0;
3192 3193
	int ret;

3194
	ret = __extent_read_full_page(tree, page, get_extent, &bio, mirror_num,
3195
				      &bio_flags, READ);
3196
	if (bio)
3197
		ret = submit_one_bio(READ, bio, mirror_num, bio_flags);
3198 3199 3200
	return ret;
}

3201 3202 3203 3204 3205 3206 3207 3208 3209 3210
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;
}

3211
/*
3212 3213 3214 3215 3216 3217 3218 3219
 * 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)
3220
 */
3221 3222 3223 3224 3225 3226 3227
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;
3228
	u64 page_end = delalloc_start + PAGE_SIZE - 1;
3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242
	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,
3243
					       BTRFS_MAX_EXTENT_SIZE);
3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264
		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;
		}
		/*
3265 3266
		 * delalloc_end is already one less than the total length, so
		 * we don't subtract one from PAGE_SIZE
3267 3268
		 */
		delalloc_to_write += (delalloc_end - delalloc_start +
3269
				      PAGE_SIZE) >> PAGE_SHIFT;
3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314
		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)
3315 3316
{
	struct extent_io_tree *tree = epd->tree;
M
Miao Xie 已提交
3317
	u64 start = page_offset(page);
3318
	u64 page_end = start + PAGE_SIZE - 1;
3319 3320 3321 3322 3323 3324
	u64 end;
	u64 cur = start;
	u64 extent_offset;
	u64 block_start;
	u64 iosize;
	sector_t sector;
3325
	struct extent_state *cached_state = NULL;
3326 3327
	struct extent_map *em;
	struct block_device *bdev;
3328
	size_t pg_offset = 0;
3329
	size_t blocksize;
3330 3331 3332
	int ret = 0;
	int nr = 0;
	bool compressed;
C
Chris Mason 已提交
3333

3334
	if (tree->ops && tree->ops->writepage_start_hook) {
C
Chris Mason 已提交
3335 3336
		ret = tree->ops->writepage_start_hook(page, start,
						      page_end);
3337 3338 3339 3340 3341 3342
		if (ret) {
			/* Fixup worker will requeue */
			if (ret == -EBUSY)
				wbc->pages_skipped++;
			else
				redirty_page_for_writepage(wbc, page);
3343

3344
			update_nr_written(page, wbc, nr_written);
3345
			unlock_page(page);
3346
			ret = 1;
3347
			goto done_unlocked;
3348 3349 3350
		}
	}

3351 3352 3353 3354 3355
	/*
	 * 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);
3356

3357
	end = page_end;
3358
	if (i_size <= start) {
3359 3360 3361
		if (tree->ops && tree->ops->writepage_end_io_hook)
			tree->ops->writepage_end_io_hook(page, start,
							 page_end, NULL, 1);
3362 3363 3364 3365 3366 3367
		goto done;
	}

	blocksize = inode->i_sb->s_blocksize;

	while (cur <= end) {
3368 3369
		u64 em_end;
		if (cur >= i_size) {
3370 3371 3372
			if (tree->ops && tree->ops->writepage_end_io_hook)
				tree->ops->writepage_end_io_hook(page, cur,
							 page_end, NULL, 1);
3373 3374
			break;
		}
3375
		em = epd->get_extent(inode, page, pg_offset, cur,
3376
				     end - cur + 1, 1);
3377
		if (IS_ERR_OR_NULL(em)) {
3378
			SetPageError(page);
3379
			ret = PTR_ERR_OR_ZERO(em);
3380 3381 3382 3383
			break;
		}

		extent_offset = cur - em->start;
3384 3385
		em_end = extent_map_end(em);
		BUG_ON(em_end <= cur);
3386
		BUG_ON(end < cur);
3387
		iosize = min(em_end - cur, end - cur + 1);
3388
		iosize = ALIGN(iosize, blocksize);
3389 3390 3391
		sector = (em->block_start + extent_offset) >> 9;
		bdev = em->bdev;
		block_start = em->block_start;
C
Chris Mason 已提交
3392
		compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
3393 3394 3395
		free_extent_map(em);
		em = NULL;

C
Chris Mason 已提交
3396 3397 3398 3399 3400
		/*
		 * compressed and inline extents are written through other
		 * paths in the FS
		 */
		if (compressed || block_start == EXTENT_MAP_HOLE ||
3401
		    block_start == EXTENT_MAP_INLINE) {
C
Chris Mason 已提交
3402 3403 3404 3405 3406 3407
			/*
			 * end_io notification does not happen here for
			 * compressed extents
			 */
			if (!compressed && tree->ops &&
			    tree->ops->writepage_end_io_hook)
3408 3409 3410
				tree->ops->writepage_end_io_hook(page, cur,
							 cur + iosize - 1,
							 NULL, 1);
C
Chris Mason 已提交
3411 3412 3413 3414 3415 3416 3417 3418 3419
			else if (compressed) {
				/* we don't want to end_page_writeback on
				 * a compressed extent.  this happens
				 * elsewhere
				 */
				nr++;
			}

			cur += iosize;
3420
			pg_offset += iosize;
3421 3422
			continue;
		}
C
Chris Mason 已提交
3423

3424 3425 3426 3427 3428 3429
		if (tree->ops && tree->ops->writepage_io_hook) {
			ret = tree->ops->writepage_io_hook(page, cur,
						cur + iosize - 1);
		} else {
			ret = 0;
		}
3430
		if (ret) {
3431
			SetPageError(page);
3432
		} else {
3433
			unsigned long max_nr = (i_size >> PAGE_SHIFT) + 1;
3434

3435 3436
			set_range_writeback(tree, cur, cur + iosize - 1);
			if (!PageWriteback(page)) {
3437 3438
				btrfs_err(BTRFS_I(inode)->root->fs_info,
					   "page %lu not writeback, cur %llu end %llu",
3439
				       page->index, cur, end);
3440 3441
			}

3442
			ret = submit_extent_page(write_flags, tree, wbc, page,
3443 3444
						 sector, iosize, pg_offset,
						 bdev, &epd->bio, max_nr,
C
Chris Mason 已提交
3445
						 end_bio_extent_writepage,
3446
						 0, 0, 0, false);
3447 3448 3449 3450
			if (ret)
				SetPageError(page);
		}
		cur = cur + iosize;
3451
		pg_offset += iosize;
3452 3453
		nr++;
	}
3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475
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);
3476
	u64 page_end = start + PAGE_SIZE - 1;
3477 3478 3479 3480
	int ret;
	int nr = 0;
	size_t pg_offset = 0;
	loff_t i_size = i_size_read(inode);
3481
	unsigned long end_index = i_size >> PAGE_SHIFT;
3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495
	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);

3496
	pg_offset = i_size & (PAGE_SIZE - 1);
3497 3498
	if (page->index > end_index ||
	   (page->index == end_index && !pg_offset)) {
3499
		page->mapping->a_ops->invalidatepage(page, 0, PAGE_SIZE);
3500 3501 3502 3503 3504 3505 3506 3507 3508
		unlock_page(page);
		return 0;
	}

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

		userpage = kmap_atomic(page);
		memset(userpage + pg_offset, 0,
3509
		       PAGE_SIZE - pg_offset);
3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528
		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;

3529 3530 3531 3532 3533 3534
done:
	if (nr == 0) {
		/* make sure the mapping tag for page dirty gets cleared */
		set_page_writeback(page);
		end_page_writeback(page);
	}
3535 3536 3537 3538
	if (PageError(page)) {
		ret = ret < 0 ? ret : -EIO;
		end_extent_writepage(page, ret, start, page_end);
	}
3539
	unlock_page(page);
3540
	return ret;
3541

3542
done_unlocked:
3543 3544 3545
	return 0;
}

3546
void wait_on_extent_buffer_writeback(struct extent_buffer *eb)
3547
{
3548 3549
	wait_on_bit_io(&eb->bflags, EXTENT_BUFFER_WRITEBACK,
		       TASK_UNINTERRUPTIBLE);
3550 3551
}

3552 3553 3554 3555
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)
3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574
{
	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
3575 3576 3577 3578 3579
		while (1) {
			wait_on_extent_buffer_writeback(eb);
			btrfs_tree_lock(eb);
			if (!test_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags))
				break;
3580 3581 3582 3583
			btrfs_tree_unlock(eb);
		}
	}

3584 3585 3586 3587 3588 3589
	/*
	 * 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);
3590 3591
	if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
		set_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags);
3592
		spin_unlock(&eb->refs_lock);
3593
		btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
3594 3595 3596
		__percpu_counter_add(&fs_info->dirty_metadata_bytes,
				     -eb->len,
				     fs_info->dirty_metadata_batch);
3597
		ret = 1;
3598 3599
	} else {
		spin_unlock(&eb->refs_lock);
3600 3601 3602 3603 3604 3605 3606 3607 3608
	}

	btrfs_tree_unlock(eb);

	if (!ret)
		return ret;

	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = 0; i < num_pages; i++) {
3609
		struct page *p = eb->pages[i];
3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625

		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);
3626
	smp_mb__after_atomic();
3627 3628 3629
	wake_up_bit(&eb->bflags, EXTENT_BUFFER_WRITEBACK);
}

3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691
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 */
	}
}

3692
static void end_bio_extent_buffer_writepage(struct bio *bio)
3693
{
3694
	struct bio_vec *bvec;
3695
	struct extent_buffer *eb;
3696
	int i, done;
3697

3698
	bio_for_each_segment_all(bvec, bio, i) {
3699 3700 3701 3702 3703 3704
		struct page *page = bvec->bv_page;

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

3705 3706
		if (bio->bi_error ||
		    test_bit(EXTENT_BUFFER_WRITE_ERR, &eb->bflags)) {
3707
			ClearPageUptodate(page);
3708
			set_btree_ioerr(page);
3709 3710 3711 3712 3713 3714 3715 3716
		}

		end_page_writeback(page);

		if (!done)
			continue;

		end_extent_buffer_writeback(eb);
3717
	}
3718 3719 3720 3721

	bio_put(bio);
}

3722
static noinline_for_stack int write_one_eb(struct extent_buffer *eb,
3723 3724 3725 3726 3727
			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;
3728
	struct extent_io_tree *tree = &BTRFS_I(fs_info->btree_inode)->io_tree;
3729 3730
	u64 offset = eb->start;
	unsigned long i, num_pages;
3731
	unsigned long bio_flags = 0;
3732
	int rw = (epd->sync_io ? WRITE_SYNC : WRITE) | REQ_META;
3733
	int ret = 0;
3734

3735
	clear_bit(EXTENT_BUFFER_WRITE_ERR, &eb->bflags);
3736 3737
	num_pages = num_extent_pages(eb->start, eb->len);
	atomic_set(&eb->io_pages, num_pages);
3738 3739 3740
	if (btrfs_header_owner(eb) == BTRFS_TREE_LOG_OBJECTID)
		bio_flags = EXTENT_BIO_TREE_LOG;

3741
	for (i = 0; i < num_pages; i++) {
3742
		struct page *p = eb->pages[i];
3743 3744 3745

		clear_page_dirty_for_io(p);
		set_page_writeback(p);
3746
		ret = submit_extent_page(rw, tree, wbc, p, offset >> 9,
3747
					 PAGE_SIZE, 0, bdev, &epd->bio,
3748
					 -1, end_bio_extent_buffer_writepage,
3749
					 0, epd->bio_flags, bio_flags, false);
3750
		epd->bio_flags = bio_flags;
3751
		if (ret) {
3752
			set_btree_ioerr(p);
3753
			end_page_writeback(p);
3754 3755 3756 3757 3758
			if (atomic_sub_and_test(num_pages - i, &eb->io_pages))
				end_extent_buffer_writeback(eb);
			ret = -EIO;
			break;
		}
3759
		offset += PAGE_SIZE;
3760 3761 3762 3763 3764 3765
		update_nr_written(p, wbc, 1);
		unlock_page(p);
	}

	if (unlikely(ret)) {
		for (; i < num_pages; i++) {
3766
			struct page *p = eb->pages[i];
3767
			clear_page_dirty_for_io(p);
3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785
			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,
3786
		.bio_flags = 0,
3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802
	};
	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 {
3803 3804
		index = wbc->range_start >> PAGE_SHIFT;
		end = wbc->range_end >> PAGE_SHIFT;
3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826 3827 3828 3829 3830
		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;
			}

3831 3832 3833 3834 3835 3836
			spin_lock(&mapping->private_lock);
			if (!PagePrivate(page)) {
				spin_unlock(&mapping->private_lock);
				continue;
			}

3837
			eb = (struct extent_buffer *)page->private;
3838 3839 3840 3841 3842 3843

			/*
			 * 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.
			 */
3844
			if (WARN_ON(!eb)) {
3845
				spin_unlock(&mapping->private_lock);
3846 3847 3848
				continue;
			}

3849 3850
			if (eb == prev_eb) {
				spin_unlock(&mapping->private_lock);
3851
				continue;
3852
			}
3853

3854 3855 3856
			ret = atomic_inc_not_zero(&eb->refs);
			spin_unlock(&mapping->private_lock);
			if (!ret)
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 3894 3895 3896
				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;
}

3897
/**
C
Chris Mason 已提交
3898
 * write_cache_pages - walk the list of dirty pages of the given address space and write all of them.
3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910 3911
 * @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.
 */
3912
static int extent_write_cache_pages(struct extent_io_tree *tree,
C
Chris Mason 已提交
3913 3914
			     struct address_space *mapping,
			     struct writeback_control *wbc,
C
Chris Mason 已提交
3915 3916
			     writepage_t writepage, void *data,
			     void (*flush_fn)(void *))
3917
{
3918
	struct inode *inode = mapping->host;
3919 3920
	int ret = 0;
	int done = 0;
3921
	int err = 0;
3922
	int nr_to_write_done = 0;
3923 3924 3925 3926 3927
	struct pagevec pvec;
	int nr_pages;
	pgoff_t index;
	pgoff_t end;		/* Inclusive */
	int scanned = 0;
3928
	int tag;
3929

3930 3931 3932 3933 3934 3935 3936 3937 3938 3939 3940 3941
	/*
	 * 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;

3942 3943 3944 3945 3946
	pagevec_init(&pvec, 0);
	if (wbc->range_cyclic) {
		index = mapping->writeback_index; /* Start from prev offset */
		end = -1;
	} else {
3947 3948
		index = wbc->range_start >> PAGE_SHIFT;
		end = wbc->range_end >> PAGE_SHIFT;
3949 3950
		scanned = 1;
	}
3951 3952 3953 3954
	if (wbc->sync_mode == WB_SYNC_ALL)
		tag = PAGECACHE_TAG_TOWRITE;
	else
		tag = PAGECACHE_TAG_DIRTY;
3955
retry:
3956 3957
	if (wbc->sync_mode == WB_SYNC_ALL)
		tag_pages_for_writeback(mapping, index, end);
3958
	while (!done && !nr_to_write_done && (index <= end) &&
3959 3960
	       (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag,
			min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973
		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
			 */
3974 3975 3976
			if (!trylock_page(page)) {
				flush_fn(data);
				lock_page(page);
3977
			}
3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989

			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
3990
			if (wbc->sync_mode != WB_SYNC_NONE) {
3991 3992
				if (PageWriteback(page))
					flush_fn(data);
3993
				wait_on_page_writeback(page);
C
Chris Mason 已提交
3994
			}
3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007

			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;
			}
4008 4009
			if (!err && ret < 0)
				err = ret;
4010 4011 4012 4013 4014 4015 4016

			/*
			 * 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;
4017 4018 4019 4020
		}
		pagevec_release(&pvec);
		cond_resched();
	}
4021
	if (!scanned && !done && !err) {
4022 4023 4024 4025 4026 4027 4028 4029
		/*
		 * 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;
	}
4030
	btrfs_add_delayed_iput(inode);
4031
	return err;
4032 4033
}

4034
static void flush_epd_write_bio(struct extent_page_data *epd)
C
Chris Mason 已提交
4035 4036
{
	if (epd->bio) {
4037 4038 4039
		int rw = WRITE;
		int ret;

4040
		if (epd->sync_io)
4041 4042
			rw = WRITE_SYNC;

4043
		ret = submit_one_bio(rw, epd->bio, 0, epd->bio_flags);
4044
		BUG_ON(ret < 0); /* -ENOMEM */
C
Chris Mason 已提交
4045 4046 4047 4048
		epd->bio = NULL;
	}
}

4049 4050 4051 4052 4053 4054
static noinline void flush_write_bio(void *data)
{
	struct extent_page_data *epd = data;
	flush_epd_write_bio(epd);
}

4055 4056 4057 4058 4059 4060 4061 4062 4063
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,
4064
		.extent_locked = 0,
4065
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
4066
		.bio_flags = 0,
4067 4068 4069 4070
	};

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

4071
	flush_epd_write_bio(&epd);
4072 4073 4074
	return ret;
}

4075 4076 4077 4078 4079 4080 4081
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;
4082 4083
	unsigned long nr_pages = (end - start + PAGE_SIZE) >>
		PAGE_SHIFT;
4084 4085 4086 4087 4088 4089

	struct extent_page_data epd = {
		.bio = NULL,
		.tree = tree,
		.get_extent = get_extent,
		.extent_locked = 1,
4090
		.sync_io = mode == WB_SYNC_ALL,
4091
		.bio_flags = 0,
4092 4093 4094 4095 4096 4097 4098 4099
	};
	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
4100
	while (start <= end) {
4101
		page = find_get_page(mapping, start >> PAGE_SHIFT);
4102 4103 4104 4105 4106
		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,
4107
						 start + PAGE_SIZE - 1,
4108 4109 4110
						 NULL, 1);
			unlock_page(page);
		}
4111 4112
		put_page(page);
		start += PAGE_SIZE;
4113 4114
	}

4115
	flush_epd_write_bio(&epd);
4116 4117
	return ret;
}
4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128

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,
4129
		.extent_locked = 0,
4130
		.sync_io = wbc->sync_mode == WB_SYNC_ALL,
4131
		.bio_flags = 0,
4132 4133
	};

C
Chris Mason 已提交
4134
	ret = extent_write_cache_pages(tree, mapping, wbc,
C
Chris Mason 已提交
4135 4136
				       __extent_writepage, &epd,
				       flush_write_bio);
4137
	flush_epd_write_bio(&epd);
4138 4139 4140 4141 4142 4143 4144 4145 4146 4147
	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
4148
	unsigned long bio_flags = 0;
L
Liu Bo 已提交
4149 4150
	struct page *pagepool[16];
	struct page *page;
4151
	struct extent_map *em_cached = NULL;
L
Liu Bo 已提交
4152
	int nr = 0;
4153
	u64 prev_em_start = (u64)-1;
4154 4155

	for (page_idx = 0; page_idx < nr_pages; page_idx++) {
L
Liu Bo 已提交
4156
		page = list_entry(pages->prev, struct page, lru);
4157 4158 4159

		prefetchw(&page->flags);
		list_del(&page->lru);
L
Liu Bo 已提交
4160
		if (add_to_page_cache_lru(page, mapping,
4161
					page->index, GFP_NOFS)) {
4162
			put_page(page);
L
Liu Bo 已提交
4163
			continue;
4164
		}
L
Liu Bo 已提交
4165 4166 4167 4168

		pagepool[nr++] = page;
		if (nr < ARRAY_SIZE(pagepool))
			continue;
4169
		__extent_readpages(tree, pagepool, nr, get_extent, &em_cached,
4170
				   &bio, 0, &bio_flags, READ, &prev_em_start);
L
Liu Bo 已提交
4171
		nr = 0;
4172
	}
4173
	if (nr)
4174
		__extent_readpages(tree, pagepool, nr, get_extent, &em_cached,
4175
				   &bio, 0, &bio_flags, READ, &prev_em_start);
L
Liu Bo 已提交
4176

4177 4178 4179
	if (em_cached)
		free_extent_map(em_cached);

4180 4181
	BUG_ON(!list_empty(pages));
	if (bio)
4182
		return submit_one_bio(READ, bio, 0, bio_flags);
4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193
	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)
{
4194
	struct extent_state *cached_state = NULL;
M
Miao Xie 已提交
4195
	u64 start = page_offset(page);
4196
	u64 end = start + PAGE_SIZE - 1;
4197 4198
	size_t blocksize = page->mapping->host->i_sb->s_blocksize;

4199
	start += ALIGN(offset, blocksize);
4200 4201 4202
	if (start > end)
		return 0;

4203
	lock_extent_bits(tree, start, end, &cached_state);
4204
	wait_on_page_writeback(page);
4205
	clear_extent_bit(tree, start, end,
4206 4207
			 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
			 EXTENT_DO_ACCOUNTING,
4208
			 1, 1, &cached_state, GFP_NOFS);
4209 4210 4211
	return 0;
}

4212 4213 4214 4215 4216
/*
 * 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.
 */
4217 4218 4219
static int try_release_extent_state(struct extent_map_tree *map,
				    struct extent_io_tree *tree,
				    struct page *page, gfp_t mask)
4220
{
M
Miao Xie 已提交
4221
	u64 start = page_offset(page);
4222
	u64 end = start + PAGE_SIZE - 1;
4223 4224
	int ret = 1;

4225
	if (test_range_bit(tree, start, end,
4226
			   EXTENT_IOBITS, 0, NULL))
4227 4228 4229 4230
		ret = 0;
	else {
		if ((mask & GFP_NOFS) == GFP_NOFS)
			mask = GFP_NOFS;
4231 4232 4233 4234
		/*
		 * at this point we can safely clear everything except the
		 * locked bit and the nodatasum bit
		 */
4235
		ret = clear_extent_bit(tree, start, end,
4236 4237
				 ~(EXTENT_LOCKED | EXTENT_NODATASUM),
				 0, 0, NULL, mask);
4238 4239 4240 4241 4242 4243 4244 4245

		/* if clear_extent_bit failed for enomem reasons,
		 * we can't allow the release to continue.
		 */
		if (ret < 0)
			ret = 0;
		else
			ret = 1;
4246 4247 4248 4249
	}
	return ret;
}

4250 4251 4252 4253 4254 4255
/*
 * 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,
4256 4257
			       struct extent_io_tree *tree, struct page *page,
			       gfp_t mask)
4258 4259
{
	struct extent_map *em;
M
Miao Xie 已提交
4260
	u64 start = page_offset(page);
4261
	u64 end = start + PAGE_SIZE - 1;
4262

4263
	if (gfpflags_allow_blocking(mask) &&
4264
	    page->mapping->host->i_size > SZ_16M) {
4265
		u64 len;
4266
		while (start <= end) {
4267
			len = end - start + 1;
4268
			write_lock(&map->lock);
4269
			em = lookup_extent_mapping(map, start, len);
4270
			if (!em) {
4271
				write_unlock(&map->lock);
4272 4273
				break;
			}
4274 4275
			if (test_bit(EXTENT_FLAG_PINNED, &em->flags) ||
			    em->start != start) {
4276
				write_unlock(&map->lock);
4277 4278 4279 4280 4281
				free_extent_map(em);
				break;
			}
			if (!test_range_bit(tree, em->start,
					    extent_map_end(em) - 1,
4282
					    EXTENT_LOCKED | EXTENT_WRITEBACK,
4283
					    0, NULL)) {
4284 4285 4286 4287 4288
				remove_extent_mapping(map, em);
				/* once for the rb tree */
				free_extent_map(em);
			}
			start = extent_map_end(em);
4289
			write_unlock(&map->lock);
4290 4291

			/* once for us */
4292 4293 4294
			free_extent_map(em);
		}
	}
4295
	return try_release_extent_state(map, tree, page, mask);
4296 4297
}

4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313
/*
 * 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;

4314
	while (1) {
4315 4316 4317
		len = last - offset;
		if (len == 0)
			break;
4318
		len = ALIGN(len, sectorsize);
4319
		em = get_extent(inode, NULL, 0, offset, len, 0);
4320
		if (IS_ERR_OR_NULL(em))
4321 4322 4323 4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334 4335 4336 4337
			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
4338 4339 4340
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
4341
	int ret = 0;
Y
Yehuda Sadeh 已提交
4342 4343 4344
	u64 off = start;
	u64 max = start + len;
	u32 flags = 0;
J
Josef Bacik 已提交
4345 4346
	u32 found_type;
	u64 last;
4347
	u64 last_for_get_extent = 0;
Y
Yehuda Sadeh 已提交
4348
	u64 disko = 0;
4349
	u64 isize = i_size_read(inode);
J
Josef Bacik 已提交
4350
	struct btrfs_key found_key;
Y
Yehuda Sadeh 已提交
4351
	struct extent_map *em = NULL;
4352
	struct extent_state *cached_state = NULL;
J
Josef Bacik 已提交
4353
	struct btrfs_path *path;
4354
	struct btrfs_root *root = BTRFS_I(inode)->root;
Y
Yehuda Sadeh 已提交
4355
	int end = 0;
4356 4357 4358
	u64 em_start = 0;
	u64 em_len = 0;
	u64 em_end = 0;
Y
Yehuda Sadeh 已提交
4359 4360 4361 4362

	if (len == 0)
		return -EINVAL;

J
Josef Bacik 已提交
4363 4364 4365 4366 4367
	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;
	path->leave_spinning = 1;

4368 4369
	start = round_down(start, BTRFS_I(inode)->root->sectorsize);
	len = round_up(max, BTRFS_I(inode)->root->sectorsize) - start;
4370

4371 4372 4373 4374
	/*
	 * lookup the last file extent.  We're not using i_size here
	 * because there might be preallocation past i_size
	 */
4375 4376
	ret = btrfs_lookup_file_extent(NULL, root, path, btrfs_ino(inode), -1,
				       0);
J
Josef Bacik 已提交
4377 4378 4379 4380 4381 4382 4383
	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]);
4384
	found_type = found_key.type;
J
Josef Bacik 已提交
4385

4386
	/* No extents, but there might be delalloc bits */
L
Li Zefan 已提交
4387
	if (found_key.objectid != btrfs_ino(inode) ||
J
Josef Bacik 已提交
4388
	    found_type != BTRFS_EXTENT_DATA_KEY) {
4389 4390 4391 4392 4393 4394 4395 4396 4397 4398 4399
		/* 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
4400
	}
4401
	btrfs_release_path(path);
J
Josef Bacik 已提交
4402

4403 4404 4405 4406 4407 4408 4409 4410 4411 4412
	/*
	 * 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;
	}

4413
	lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + len - 1,
4414
			 &cached_state);
4415

4416
	em = get_extent_skip_holes(inode, start, last_for_get_extent,
4417
				   get_extent);
Y
Yehuda Sadeh 已提交
4418 4419 4420 4421 4422 4423
	if (!em)
		goto out;
	if (IS_ERR(em)) {
		ret = PTR_ERR(em);
		goto out;
	}
J
Josef Bacik 已提交
4424

Y
Yehuda Sadeh 已提交
4425
	while (!end) {
4426
		u64 offset_in_extent = 0;
4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438

		/* 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
4439

4440 4441
		/*
		 * record the offset from the start of the extent
4442 4443 4444
		 * 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.
4445
		 */
4446 4447
		if (!test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
			offset_in_extent = em_start - em->start;
4448
		em_end = extent_map_end(em);
4449
		em_len = em_end - em_start;
Y
Yehuda Sadeh 已提交
4450 4451 4452
		disko = 0;
		flags = 0;

4453 4454 4455 4456 4457 4458 4459
		/*
		 * bump off for our next call to get_extent
		 */
		off = extent_map_end(em);
		if (off >= max)
			end = 1;

4460
		if (em->block_start == EXTENT_MAP_LAST_BYTE) {
Y
Yehuda Sadeh 已提交
4461 4462
			end = 1;
			flags |= FIEMAP_EXTENT_LAST;
4463
		} else if (em->block_start == EXTENT_MAP_INLINE) {
Y
Yehuda Sadeh 已提交
4464 4465
			flags |= (FIEMAP_EXTENT_DATA_INLINE |
				  FIEMAP_EXTENT_NOT_ALIGNED);
4466
		} else if (em->block_start == EXTENT_MAP_DELALLOC) {
Y
Yehuda Sadeh 已提交
4467 4468
			flags |= (FIEMAP_EXTENT_DELALLOC |
				  FIEMAP_EXTENT_UNKNOWN);
4469 4470 4471
		} else if (fieinfo->fi_extents_max) {
			u64 bytenr = em->block_start -
				(em->start - em->orig_start);
4472

4473
			disko = em->block_start + offset_in_extent;
4474 4475 4476 4477

			/*
			 * As btrfs supports shared space, this information
			 * can be exported to userspace tools via
4478 4479 4480
			 * flag FIEMAP_EXTENT_SHARED.  If fi_extents_max == 0
			 * then we're just getting a count and we can skip the
			 * lookup stuff.
4481
			 */
4482 4483 4484 4485
			ret = btrfs_check_shared(NULL, root->fs_info,
						 root->objectid,
						 btrfs_ino(inode), bytenr);
			if (ret < 0)
4486
				goto out_free;
4487
			if (ret)
4488
				flags |= FIEMAP_EXTENT_SHARED;
4489
			ret = 0;
Y
Yehuda Sadeh 已提交
4490 4491 4492
		}
		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
			flags |= FIEMAP_EXTENT_ENCODED;
4493 4494
		if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			flags |= FIEMAP_EXTENT_UNWRITTEN;
Y
Yehuda Sadeh 已提交
4495 4496 4497

		free_extent_map(em);
		em = NULL;
4498 4499
		if ((em_start >= last) || em_len == (u64)-1 ||
		   (last == (u64)-1 && isize <= em_end)) {
Y
Yehuda Sadeh 已提交
4500 4501 4502 4503
			flags |= FIEMAP_EXTENT_LAST;
			end = 1;
		}

4504 4505 4506 4507 4508 4509 4510 4511
		/* 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
4512 4513 4514
			flags |= FIEMAP_EXTENT_LAST;
			end = 1;
		}
4515 4516
		ret = fiemap_fill_next_extent(fieinfo, em_start, disko,
					      em_len, flags);
4517 4518 4519
		if (ret) {
			if (ret == 1)
				ret = 0;
4520
			goto out_free;
4521
		}
Y
Yehuda Sadeh 已提交
4522 4523 4524 4525
	}
out_free:
	free_extent_map(em);
out:
4526
	btrfs_free_path(path);
L
Liu Bo 已提交
4527
	unlock_extent_cached(&BTRFS_I(inode)->io_tree, start, start + len - 1,
4528
			     &cached_state, GFP_NOFS);
Y
Yehuda Sadeh 已提交
4529 4530 4531
	return ret;
}

4532 4533
static void __free_extent_buffer(struct extent_buffer *eb)
{
4534
	btrfs_leak_debug_del(&eb->leak_list);
4535 4536 4537
	kmem_cache_free(extent_buffer_cache, eb);
}

4538
int extent_buffer_under_io(struct extent_buffer *eb)
4539 4540 4541 4542 4543 4544 4545 4546 4547
{
	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.
 */
4548
static void btrfs_release_extent_buffer_page(struct extent_buffer *eb)
4549 4550 4551 4552 4553 4554 4555
{
	unsigned long index;
	struct page *page;
	int mapped = !test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags);

	BUG_ON(extent_buffer_under_io(eb));

4556 4557
	index = num_extent_pages(eb->start, eb->len);
	if (index == 0)
4558 4559 4560 4561
		return;

	do {
		index--;
4562
		page = eb->pages[index];
4563 4564 4565
		if (!page)
			continue;
		if (mapped)
4566
			spin_lock(&page->mapping->private_lock);
4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578
		/*
		 * 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));
4579
			/*
4580 4581
			 * We need to make sure we haven't be attached
			 * to a new eb.
4582
			 */
4583 4584 4585
			ClearPagePrivate(page);
			set_page_private(page, 0);
			/* One for the page private */
4586
			put_page(page);
4587
		}
4588 4589 4590 4591 4592

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

		/* One for when we alloced the page */
4593
		put_page(page);
4594
	} while (index != 0);
4595 4596 4597 4598 4599 4600 4601
}

/*
 * Helper for releasing the extent buffer.
 */
static inline void btrfs_release_extent_buffer(struct extent_buffer *eb)
{
4602
	btrfs_release_extent_buffer_page(eb);
4603 4604 4605
	__free_extent_buffer(eb);
}

4606 4607
static struct extent_buffer *
__alloc_extent_buffer(struct btrfs_fs_info *fs_info, u64 start,
4608
		      unsigned long len)
4609 4610 4611
{
	struct extent_buffer *eb = NULL;

4612
	eb = kmem_cache_zalloc(extent_buffer_cache, GFP_NOFS|__GFP_NOFAIL);
4613 4614
	eb->start = start;
	eb->len = len;
4615
	eb->fs_info = fs_info;
4616
	eb->bflags = 0;
4617 4618 4619 4620 4621 4622 4623
	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);
4624
	eb->lock_nested = 0;
4625 4626
	init_waitqueue_head(&eb->write_lock_wq);
	init_waitqueue_head(&eb->read_lock_wq);
4627

4628 4629
	btrfs_leak_debug_add(&eb->leak_list, &buffers);

4630
	spin_lock_init(&eb->refs_lock);
4631
	atomic_set(&eb->refs, 1);
4632
	atomic_set(&eb->io_pages, 0);
4633

4634 4635 4636 4637 4638 4639
	/*
	 * 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);
4640 4641 4642 4643

	return eb;
}

4644 4645 4646 4647 4648 4649 4650
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);

4651
	new = __alloc_extent_buffer(src->fs_info, src->start, src->len);
4652 4653 4654 4655
	if (new == NULL)
		return NULL;

	for (i = 0; i < num_pages; i++) {
4656
		p = alloc_page(GFP_NOFS);
4657 4658 4659 4660
		if (!p) {
			btrfs_release_extent_buffer(new);
			return NULL;
		}
4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673
		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;
}

4674 4675
struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
						  u64 start, unsigned long len)
4676 4677
{
	struct extent_buffer *eb;
4678
	unsigned long num_pages;
4679 4680
	unsigned long i;

4681
	num_pages = num_extent_pages(start, len);
4682 4683

	eb = __alloc_extent_buffer(fs_info, start, len);
4684 4685 4686 4687
	if (!eb)
		return NULL;

	for (i = 0; i < num_pages; i++) {
4688
		eb->pages[i] = alloc_page(GFP_NOFS);
4689 4690 4691 4692 4693 4694 4695 4696 4697
		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:
4698 4699
	for (; i > 0; i--)
		__free_page(eb->pages[i - 1]);
4700 4701 4702 4703
	__free_extent_buffer(eb);
	return NULL;
}

4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715 4716 4717 4718 4719 4720 4721
struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
						u64 start)
{
	unsigned long len;

	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;
	}

	return __alloc_dummy_extent_buffer(fs_info, start, len);
}

4722 4723
static void check_buffer_tree_ref(struct extent_buffer *eb)
{
4724
	int refs;
4725 4726 4727 4728 4729 4730 4731 4732 4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743 4744
	/* 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.
	 */
4745 4746 4747 4748
	refs = atomic_read(&eb->refs);
	if (refs >= 2 && test_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
		return;

4749 4750
	spin_lock(&eb->refs_lock);
	if (!test_and_set_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
4751
		atomic_inc(&eb->refs);
4752
	spin_unlock(&eb->refs_lock);
4753 4754
}

4755 4756
static void mark_extent_buffer_accessed(struct extent_buffer *eb,
		struct page *accessed)
4757 4758 4759
{
	unsigned long num_pages, i;

4760 4761
	check_buffer_tree_ref(eb);

4762 4763
	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = 0; i < num_pages; i++) {
4764 4765
		struct page *p = eb->pages[i];

4766 4767
		if (p != accessed)
			mark_page_accessed(p);
4768 4769 4770
	}
}

4771 4772
struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
					 u64 start)
4773 4774 4775 4776
{
	struct extent_buffer *eb;

	rcu_read_lock();
4777
	eb = radix_tree_lookup(&fs_info->buffer_radix,
4778
			       start >> PAGE_SHIFT);
4779 4780
	if (eb && atomic_inc_not_zero(&eb->refs)) {
		rcu_read_unlock();
4781 4782 4783 4784 4785 4786 4787 4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 4798 4799
		/*
		 * 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);
		}
4800
		mark_extent_buffer_accessed(eb, NULL);
4801 4802 4803 4804 4805 4806 4807
		return eb;
	}
	rcu_read_unlock();

	return NULL;
}

4808 4809
#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
4810
					       u64 start)
4811 4812 4813 4814 4815 4816 4817
{
	struct extent_buffer *eb, *exists = NULL;
	int ret;

	eb = find_extent_buffer(fs_info, start);
	if (eb)
		return eb;
4818
	eb = alloc_dummy_extent_buffer(fs_info, start);
4819 4820 4821 4822 4823 4824 4825 4826 4827
	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,
4828
				start >> PAGE_SHIFT, eb);
4829 4830 4831 4832 4833 4834 4835 4836 4837 4838 4839 4840 4841 4842 4843 4844 4845 4846 4847 4848 4849 4850 4851 4852 4853 4854
	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

4855
struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
4856
					  u64 start)
4857
{
4858
	unsigned long len = fs_info->tree_root->nodesize;
4859 4860
	unsigned long num_pages = num_extent_pages(start, len);
	unsigned long i;
4861
	unsigned long index = start >> PAGE_SHIFT;
4862
	struct extent_buffer *eb;
4863
	struct extent_buffer *exists = NULL;
4864
	struct page *p;
4865
	struct address_space *mapping = fs_info->btree_inode->i_mapping;
4866
	int uptodate = 1;
4867
	int ret;
4868

4869
	eb = find_extent_buffer(fs_info, start);
4870
	if (eb)
4871 4872
		return eb;

4873
	eb = __alloc_extent_buffer(fs_info, start, len);
4874
	if (!eb)
4875 4876
		return NULL;

4877
	for (i = 0; i < num_pages; i++, index++) {
4878
		p = find_or_create_page(mapping, index, GFP_NOFS|__GFP_NOFAIL);
4879
		if (!p)
4880
			goto free_eb;
J
Josef Bacik 已提交
4881 4882 4883 4884 4885 4886 4887 4888 4889 4890 4891 4892 4893 4894

		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);
4895
				put_page(p);
4896
				mark_extent_buffer_accessed(exists, p);
J
Josef Bacik 已提交
4897 4898
				goto free_eb;
			}
4899
			exists = NULL;
J
Josef Bacik 已提交
4900

4901
			/*
J
Josef Bacik 已提交
4902 4903 4904 4905
			 * Do this so attach doesn't complain and we need to
			 * drop the ref the old guy had.
			 */
			ClearPagePrivate(p);
4906
			WARN_ON(PageDirty(p));
4907
			put_page(p);
4908
		}
J
Josef Bacik 已提交
4909 4910
		attach_extent_buffer_page(eb, p);
		spin_unlock(&mapping->private_lock);
4911
		WARN_ON(PageDirty(p));
4912
		eb->pages[i] = p;
4913 4914
		if (!PageUptodate(p))
			uptodate = 0;
C
Chris Mason 已提交
4915 4916 4917 4918 4919

		/*
		 * see below about how we avoid a nasty race with release page
		 * and why we unlock later
		 */
4920 4921
	}
	if (uptodate)
4922
		set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
4923
again:
4924 4925 4926 4927
	ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
	if (ret)
		goto free_eb;

4928 4929
	spin_lock(&fs_info->buffer_lock);
	ret = radix_tree_insert(&fs_info->buffer_radix,
4930
				start >> PAGE_SHIFT, eb);
4931
	spin_unlock(&fs_info->buffer_lock);
4932
	radix_tree_preload_end();
4933
	if (ret == -EEXIST) {
4934
		exists = find_extent_buffer(fs_info, start);
4935 4936 4937
		if (exists)
			goto free_eb;
		else
4938
			goto again;
4939 4940
	}
	/* add one reference for the tree */
4941
	check_buffer_tree_ref(eb);
4942
	set_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags);
C
Chris Mason 已提交
4943 4944 4945 4946 4947 4948 4949 4950 4951 4952

	/*
	 * 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
	 */
4953 4954
	SetPageChecked(eb->pages[0]);
	for (i = 1; i < num_pages; i++) {
4955
		p = eb->pages[i];
4956 4957 4958 4959
		ClearPageChecked(p);
		unlock_page(p);
	}
	unlock_page(eb->pages[0]);
4960 4961
	return eb;

4962
free_eb:
4963
	WARN_ON(!atomic_dec_and_test(&eb->refs));
4964 4965 4966 4967
	for (i = 0; i < num_pages; i++) {
		if (eb->pages[i])
			unlock_page(eb->pages[i]);
	}
C
Chris Mason 已提交
4968

4969
	btrfs_release_extent_buffer(eb);
4970
	return exists;
4971 4972
}

4973 4974 4975 4976 4977 4978 4979 4980 4981
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 */
4982
static int release_extent_buffer(struct extent_buffer *eb)
4983 4984 4985
{
	WARN_ON(atomic_read(&eb->refs) == 0);
	if (atomic_dec_and_test(&eb->refs)) {
4986
		if (test_and_clear_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags)) {
4987
			struct btrfs_fs_info *fs_info = eb->fs_info;
4988

4989
			spin_unlock(&eb->refs_lock);
4990

4991 4992
			spin_lock(&fs_info->buffer_lock);
			radix_tree_delete(&fs_info->buffer_radix,
4993
					  eb->start >> PAGE_SHIFT);
4994
			spin_unlock(&fs_info->buffer_lock);
4995 4996
		} else {
			spin_unlock(&eb->refs_lock);
4997
		}
4998 4999

		/* Should be safe to release our pages at this point */
5000
		btrfs_release_extent_buffer_page(eb);
5001 5002 5003 5004 5005 5006
#ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
		if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags))) {
			__free_extent_buffer(eb);
			return 1;
		}
#endif
5007
		call_rcu(&eb->rcu_head, btrfs_release_extent_buffer_rcu);
5008
		return 1;
5009 5010
	}
	spin_unlock(&eb->refs_lock);
5011 5012

	return 0;
5013 5014
}

5015 5016
void free_extent_buffer(struct extent_buffer *eb)
{
5017 5018
	int refs;
	int old;
5019 5020 5021
	if (!eb)
		return;

5022 5023 5024 5025 5026 5027 5028 5029 5030
	while (1) {
		refs = atomic_read(&eb->refs);
		if (refs <= 3)
			break;
		old = atomic_cmpxchg(&eb->refs, refs, refs - 1);
		if (old == refs)
			return;
	}

5031
	spin_lock(&eb->refs_lock);
5032 5033 5034 5035
	if (atomic_read(&eb->refs) == 2 &&
	    test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags))
		atomic_dec(&eb->refs);

5036 5037
	if (atomic_read(&eb->refs) == 2 &&
	    test_bit(EXTENT_BUFFER_STALE, &eb->bflags) &&
5038
	    !extent_buffer_under_io(eb) &&
5039 5040 5041 5042 5043 5044 5045
	    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.
	 */
5046
	release_extent_buffer(eb);
5047 5048 5049 5050 5051
}

void free_extent_buffer_stale(struct extent_buffer *eb)
{
	if (!eb)
5052 5053
		return;

5054 5055 5056
	spin_lock(&eb->refs_lock);
	set_bit(EXTENT_BUFFER_STALE, &eb->bflags);

5057
	if (atomic_read(&eb->refs) == 2 && !extent_buffer_under_io(eb) &&
5058 5059
	    test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
		atomic_dec(&eb->refs);
5060
	release_extent_buffer(eb);
5061 5062
}

5063
void clear_extent_buffer_dirty(struct extent_buffer *eb)
5064 5065 5066 5067 5068 5069 5070 5071
{
	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++) {
5072
		page = eb->pages[i];
5073
		if (!PageDirty(page))
C
Chris Mason 已提交
5074 5075
			continue;

5076
		lock_page(page);
C
Chris Mason 已提交
5077 5078
		WARN_ON(!PagePrivate(page));

5079
		clear_page_dirty_for_io(page);
5080
		spin_lock_irq(&page->mapping->tree_lock);
5081 5082 5083 5084 5085
		if (!PageDirty(page)) {
			radix_tree_tag_clear(&page->mapping->page_tree,
						page_index(page),
						PAGECACHE_TAG_DIRTY);
		}
5086
		spin_unlock_irq(&page->mapping->tree_lock);
5087
		ClearPageError(page);
5088
		unlock_page(page);
5089
	}
5090
	WARN_ON(atomic_read(&eb->refs) == 0);
5091 5092
}

5093
int set_extent_buffer_dirty(struct extent_buffer *eb)
5094 5095 5096
{
	unsigned long i;
	unsigned long num_pages;
5097
	int was_dirty = 0;
5098

5099 5100
	check_buffer_tree_ref(eb);

5101
	was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
5102

5103
	num_pages = num_extent_pages(eb->start, eb->len);
5104
	WARN_ON(atomic_read(&eb->refs) == 0);
5105 5106
	WARN_ON(!test_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags));

5107
	for (i = 0; i < num_pages; i++)
5108
		set_page_dirty(eb->pages[i]);
5109
	return was_dirty;
5110 5111
}

5112
void clear_extent_buffer_uptodate(struct extent_buffer *eb)
5113 5114 5115 5116 5117
{
	unsigned long i;
	struct page *page;
	unsigned long num_pages;

5118
	clear_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5119
	num_pages = num_extent_pages(eb->start, eb->len);
5120
	for (i = 0; i < num_pages; i++) {
5121
		page = eb->pages[i];
C
Chris Mason 已提交
5122 5123
		if (page)
			ClearPageUptodate(page);
5124 5125 5126
	}
}

5127
void set_extent_buffer_uptodate(struct extent_buffer *eb)
5128 5129 5130 5131 5132
{
	unsigned long i;
	struct page *page;
	unsigned long num_pages;

5133
	set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5134 5135
	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = 0; i < num_pages; i++) {
5136
		page = eb->pages[i];
5137 5138 5139 5140
		SetPageUptodate(page);
	}
}

5141
int extent_buffer_uptodate(struct extent_buffer *eb)
5142
{
5143
	return test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5144 5145 5146
}

int read_extent_buffer_pages(struct extent_io_tree *tree,
5147
			     struct extent_buffer *eb, u64 start, int wait,
5148
			     get_extent_t *get_extent, int mirror_num)
5149 5150 5151 5152 5153 5154
{
	unsigned long i;
	unsigned long start_i;
	struct page *page;
	int err;
	int ret = 0;
5155 5156
	int locked_pages = 0;
	int all_uptodate = 1;
5157
	unsigned long num_pages;
5158
	unsigned long num_reads = 0;
5159
	struct bio *bio = NULL;
C
Chris Mason 已提交
5160
	unsigned long bio_flags = 0;
5161

5162
	if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags))
5163 5164 5165 5166
		return 0;

	if (start) {
		WARN_ON(start < eb->start);
5167 5168
		start_i = (start >> PAGE_SHIFT) -
			(eb->start >> PAGE_SHIFT);
5169 5170 5171 5172 5173 5174
	} else {
		start_i = 0;
	}

	num_pages = num_extent_pages(eb->start, eb->len);
	for (i = start_i; i < num_pages; i++) {
5175
		page = eb->pages[i];
5176
		if (wait == WAIT_NONE) {
5177
			if (!trylock_page(page))
5178
				goto unlock_exit;
5179 5180 5181
		} else {
			lock_page(page);
		}
5182
		locked_pages++;
5183 5184
		if (!PageUptodate(page)) {
			num_reads++;
5185
			all_uptodate = 0;
5186
		}
5187 5188 5189
	}
	if (all_uptodate) {
		if (start_i == 0)
5190
			set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5191 5192 5193
		goto unlock_exit;
	}

5194
	clear_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags);
5195
	eb->read_mirror = 0;
5196
	atomic_set(&eb->io_pages, num_reads);
5197
	for (i = start_i; i < num_pages; i++) {
5198
		page = eb->pages[i];
5199
		if (!PageUptodate(page)) {
5200
			ClearPageError(page);
5201
			err = __extent_read_full_page(tree, page,
5202
						      get_extent, &bio,
5203 5204
						      mirror_num, &bio_flags,
						      READ | REQ_META);
C
Chris Mason 已提交
5205
			if (err)
5206 5207 5208 5209 5210 5211
				ret = err;
		} else {
			unlock_page(page);
		}
	}

5212
	if (bio) {
5213 5214
		err = submit_one_bio(READ | REQ_META, bio, mirror_num,
				     bio_flags);
5215 5216
		if (err)
			return err;
5217
	}
5218

5219
	if (ret || wait != WAIT_COMPLETE)
5220
		return ret;
C
Chris Mason 已提交
5221

5222
	for (i = start_i; i < num_pages; i++) {
5223
		page = eb->pages[i];
5224
		wait_on_page_locked(page);
C
Chris Mason 已提交
5225
		if (!PageUptodate(page))
5226 5227
			ret = -EIO;
	}
C
Chris Mason 已提交
5228

5229
	return ret;
5230 5231 5232

unlock_exit:
	i = start_i;
C
Chris Mason 已提交
5233
	while (locked_pages > 0) {
5234
		page = eb->pages[i];
5235 5236 5237 5238 5239
		i++;
		unlock_page(page);
		locked_pages--;
	}
	return ret;
5240 5241 5242 5243 5244 5245 5246 5247 5248 5249 5250
}

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;
5251 5252
	size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_SHIFT;
5253 5254 5255 5256

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

5257
	offset = (start_offset + start) & (PAGE_SIZE - 1);
5258

C
Chris Mason 已提交
5259
	while (len > 0) {
5260
		page = eb->pages[i];
5261

5262
		cur = min(len, (PAGE_SIZE - offset));
5263
		kaddr = page_address(page);
5264 5265 5266 5267 5268 5269 5270 5271 5272
		memcpy(dst, kaddr + offset, cur);

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

5273 5274 5275 5276 5277 5278 5279 5280 5281
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;
5282 5283
	size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_SHIFT;
5284 5285 5286 5287 5288
	int ret = 0;

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

5289
	offset = (start_offset + start) & (PAGE_SIZE - 1);
5290 5291

	while (len > 0) {
5292
		page = eb->pages[i];
5293

5294
		cur = min(len, (PAGE_SIZE - offset));
5295 5296 5297 5298 5299 5300 5301 5302 5303 5304 5305 5306 5307 5308 5309
		kaddr = page_address(page);
		if (copy_to_user(dst, kaddr + offset, cur)) {
			ret = -EFAULT;
			break;
		}

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

	return ret;
}

5310
int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start,
5311
			       unsigned long min_len, char **map,
5312
			       unsigned long *map_start,
5313
			       unsigned long *map_len)
5314
{
5315
	size_t offset = start & (PAGE_SIZE - 1);
5316 5317
	char *kaddr;
	struct page *p;
5318 5319
	size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_SHIFT;
5320
	unsigned long end_i = (start_offset + start + min_len - 1) >>
5321
		PAGE_SHIFT;
5322 5323 5324 5325 5326 5327 5328 5329 5330

	if (i != end_i)
		return -EINVAL;

	if (i == 0) {
		offset = start_offset;
		*map_start = 0;
	} else {
		offset = 0;
5331
		*map_start = ((u64)i << PAGE_SHIFT) - start_offset;
5332
	}
C
Chris Mason 已提交
5333

5334
	if (start + min_len > eb->len) {
J
Julia Lawall 已提交
5335
		WARN(1, KERN_ERR "btrfs bad mapping eb start %llu len %lu, "
5336 5337
		       "wanted %lu %lu\n",
		       eb->start, eb->len, start, min_len);
5338
		return -EINVAL;
5339 5340
	}

5341
	p = eb->pages[i];
5342
	kaddr = page_address(p);
5343
	*map = kaddr + offset;
5344
	*map_len = PAGE_SIZE - offset;
5345 5346 5347 5348 5349 5350 5351 5352 5353 5354 5355 5356
	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;
5357 5358
	size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_SHIFT;
5359 5360 5361 5362 5363
	int ret = 0;

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

5364
	offset = (start_offset + start) & (PAGE_SIZE - 1);
5365

C
Chris Mason 已提交
5366
	while (len > 0) {
5367
		page = eb->pages[i];
5368

5369
		cur = min(len, (PAGE_SIZE - offset));
5370

5371
		kaddr = page_address(page);
5372 5373 5374 5375 5376 5377 5378 5379 5380 5381 5382 5383 5384 5385 5386 5387 5388 5389 5390 5391
		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;
5392 5393
	size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_SHIFT;
5394 5395 5396 5397

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

5398
	offset = (start_offset + start) & (PAGE_SIZE - 1);
5399

C
Chris Mason 已提交
5400
	while (len > 0) {
5401
		page = eb->pages[i];
5402 5403
		WARN_ON(!PageUptodate(page));

5404
		cur = min(len, PAGE_SIZE - offset);
5405
		kaddr = page_address(page);
5406 5407 5408 5409 5410 5411 5412 5413 5414 5415 5416 5417 5418 5419 5420 5421
		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;
5422 5423
	size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
	unsigned long i = (start_offset + start) >> PAGE_SHIFT;
5424 5425 5426 5427

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

5428
	offset = (start_offset + start) & (PAGE_SIZE - 1);
5429

C
Chris Mason 已提交
5430
	while (len > 0) {
5431
		page = eb->pages[i];
5432 5433
		WARN_ON(!PageUptodate(page));

5434
		cur = min(len, PAGE_SIZE - offset);
5435
		kaddr = page_address(page);
5436 5437 5438 5439 5440 5441 5442 5443 5444 5445 5446 5447 5448 5449 5450 5451 5452
		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;
5453 5454
	size_t start_offset = dst->start & ((u64)PAGE_SIZE - 1);
	unsigned long i = (start_offset + dst_offset) >> PAGE_SHIFT;
5455 5456 5457 5458

	WARN_ON(src->len != dst_len);

	offset = (start_offset + dst_offset) &
5459
		(PAGE_SIZE - 1);
5460

C
Chris Mason 已提交
5461
	while (len > 0) {
5462
		page = dst->pages[i];
5463 5464
		WARN_ON(!PageUptodate(page));

5465
		cur = min(len, (unsigned long)(PAGE_SIZE - offset));
5466

5467
		kaddr = page_address(page);
5468 5469 5470 5471 5472 5473 5474 5475 5476
		read_extent_buffer(src, kaddr + offset, src_offset, cur);

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

5477 5478 5479 5480 5481 5482 5483 5484 5485 5486 5487 5488 5489 5490 5491 5492 5493 5494 5495 5496 5497 5498 5499 5500 5501 5502 5503 5504 5505 5506
/*
 * The extent buffer bitmap operations are done with byte granularity because
 * bitmap items are not guaranteed to be aligned to a word and therefore a
 * single word in a bitmap may straddle two pages in the extent buffer.
 */
#define BIT_BYTE(nr) ((nr) / BITS_PER_BYTE)
#define BYTE_MASK ((1 << BITS_PER_BYTE) - 1)
#define BITMAP_FIRST_BYTE_MASK(start) \
	((BYTE_MASK << ((start) & (BITS_PER_BYTE - 1))) & BYTE_MASK)
#define BITMAP_LAST_BYTE_MASK(nbits) \
	(BYTE_MASK >> (-(nbits) & (BITS_PER_BYTE - 1)))

/*
 * eb_bitmap_offset() - calculate the page and offset of the byte containing the
 * given bit number
 * @eb: the extent buffer
 * @start: offset of the bitmap item in the extent buffer
 * @nr: bit number
 * @page_index: return index of the page in the extent buffer that contains the
 * given bit number
 * @page_offset: return offset into the page given by page_index
 *
 * This helper hides the ugliness of finding the byte in an extent buffer which
 * contains a given bit.
 */
static inline void eb_bitmap_offset(struct extent_buffer *eb,
				    unsigned long start, unsigned long nr,
				    unsigned long *page_index,
				    size_t *page_offset)
{
5507
	size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
5508 5509 5510 5511 5512 5513 5514 5515 5516 5517
	size_t byte_offset = BIT_BYTE(nr);
	size_t offset;

	/*
	 * The byte we want is the offset of the extent buffer + the offset of
	 * the bitmap item in the extent buffer + the offset of the byte in the
	 * bitmap item.
	 */
	offset = start_offset + start + byte_offset;

5518 5519
	*page_index = offset >> PAGE_SHIFT;
	*page_offset = offset & (PAGE_SIZE - 1);
5520 5521 5522 5523 5524 5525 5526 5527 5528 5529 5530 5531 5532 5533 5534 5535 5536 5537 5538 5539 5540 5541 5542 5543 5544 5545 5546 5547 5548 5549 5550 5551 5552 5553 5554 5555 5556 5557 5558 5559 5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570
}

/**
 * extent_buffer_test_bit - determine whether a bit in a bitmap item is set
 * @eb: the extent buffer
 * @start: offset of the bitmap item in the extent buffer
 * @nr: bit number to test
 */
int extent_buffer_test_bit(struct extent_buffer *eb, unsigned long start,
			   unsigned long nr)
{
	char *kaddr;
	struct page *page;
	unsigned long i;
	size_t offset;

	eb_bitmap_offset(eb, start, nr, &i, &offset);
	page = eb->pages[i];
	WARN_ON(!PageUptodate(page));
	kaddr = page_address(page);
	return 1U & (kaddr[offset] >> (nr & (BITS_PER_BYTE - 1)));
}

/**
 * extent_buffer_bitmap_set - set an area of a bitmap
 * @eb: the extent buffer
 * @start: offset of the bitmap item in the extent buffer
 * @pos: bit number of the first bit
 * @len: number of bits to set
 */
void extent_buffer_bitmap_set(struct extent_buffer *eb, unsigned long start,
			      unsigned long pos, unsigned long len)
{
	char *kaddr;
	struct page *page;
	unsigned long i;
	size_t offset;
	const unsigned int size = pos + len;
	int bits_to_set = BITS_PER_BYTE - (pos % BITS_PER_BYTE);
	unsigned int mask_to_set = BITMAP_FIRST_BYTE_MASK(pos);

	eb_bitmap_offset(eb, start, pos, &i, &offset);
	page = eb->pages[i];
	WARN_ON(!PageUptodate(page));
	kaddr = page_address(page);

	while (len >= bits_to_set) {
		kaddr[offset] |= mask_to_set;
		len -= bits_to_set;
		bits_to_set = BITS_PER_BYTE;
		mask_to_set = ~0U;
5571
		if (++offset >= PAGE_SIZE && len > 0) {
5572 5573 5574 5575 5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594 5595 5596 5597 5598 5599 5600 5601 5602 5603 5604 5605 5606 5607 5608 5609 5610 5611 5612
			offset = 0;
			page = eb->pages[++i];
			WARN_ON(!PageUptodate(page));
			kaddr = page_address(page);
		}
	}
	if (len) {
		mask_to_set &= BITMAP_LAST_BYTE_MASK(size);
		kaddr[offset] |= mask_to_set;
	}
}


/**
 * extent_buffer_bitmap_clear - clear an area of a bitmap
 * @eb: the extent buffer
 * @start: offset of the bitmap item in the extent buffer
 * @pos: bit number of the first bit
 * @len: number of bits to clear
 */
void extent_buffer_bitmap_clear(struct extent_buffer *eb, unsigned long start,
				unsigned long pos, unsigned long len)
{
	char *kaddr;
	struct page *page;
	unsigned long i;
	size_t offset;
	const unsigned int size = pos + len;
	int bits_to_clear = BITS_PER_BYTE - (pos % BITS_PER_BYTE);
	unsigned int mask_to_clear = BITMAP_FIRST_BYTE_MASK(pos);

	eb_bitmap_offset(eb, start, pos, &i, &offset);
	page = eb->pages[i];
	WARN_ON(!PageUptodate(page));
	kaddr = page_address(page);

	while (len >= bits_to_clear) {
		kaddr[offset] &= ~mask_to_clear;
		len -= bits_to_clear;
		bits_to_clear = BITS_PER_BYTE;
		mask_to_clear = ~0U;
5613
		if (++offset >= PAGE_SIZE && len > 0) {
5614 5615 5616 5617 5618 5619 5620 5621 5622 5623 5624 5625
			offset = 0;
			page = eb->pages[++i];
			WARN_ON(!PageUptodate(page));
			kaddr = page_address(page);
		}
	}
	if (len) {
		mask_to_clear &= BITMAP_LAST_BYTE_MASK(size);
		kaddr[offset] &= ~mask_to_clear;
	}
}

5626 5627 5628 5629 5630 5631
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;
}

5632 5633 5634 5635
static void copy_pages(struct page *dst_page, struct page *src_page,
		       unsigned long dst_off, unsigned long src_off,
		       unsigned long len)
{
5636
	char *dst_kaddr = page_address(dst_page);
5637
	char *src_kaddr;
5638
	int must_memmove = 0;
5639

5640
	if (dst_page != src_page) {
5641
		src_kaddr = page_address(src_page);
5642
	} else {
5643
		src_kaddr = dst_kaddr;
5644 5645
		if (areas_overlap(src_off, dst_off, len))
			must_memmove = 1;
5646
	}
5647

5648 5649 5650 5651
	if (must_memmove)
		memmove(dst_kaddr + dst_off, src_kaddr + src_off, len);
	else
		memcpy(dst_kaddr + dst_off, src_kaddr + src_off, len);
5652 5653 5654 5655 5656 5657 5658 5659
}

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;
5660
	size_t start_offset = dst->start & ((u64)PAGE_SIZE - 1);
5661 5662 5663 5664
	unsigned long dst_i;
	unsigned long src_i;

	if (src_offset + len > dst->len) {
5665 5666 5667
		btrfs_err(dst->fs_info,
			"memmove bogus src_offset %lu move "
		       "len %lu dst len %lu", src_offset, len, dst->len);
5668 5669 5670
		BUG_ON(1);
	}
	if (dst_offset + len > dst->len) {
5671 5672 5673
		btrfs_err(dst->fs_info,
			"memmove bogus dst_offset %lu move "
		       "len %lu dst len %lu", dst_offset, len, dst->len);
5674 5675 5676
		BUG_ON(1);
	}

C
Chris Mason 已提交
5677
	while (len > 0) {
5678
		dst_off_in_page = (start_offset + dst_offset) &
5679
			(PAGE_SIZE - 1);
5680
		src_off_in_page = (start_offset + src_offset) &
5681
			(PAGE_SIZE - 1);
5682

5683 5684
		dst_i = (start_offset + dst_offset) >> PAGE_SHIFT;
		src_i = (start_offset + src_offset) >> PAGE_SHIFT;
5685

5686
		cur = min(len, (unsigned long)(PAGE_SIZE -
5687 5688
					       src_off_in_page));
		cur = min_t(unsigned long, cur,
5689
			(unsigned long)(PAGE_SIZE - dst_off_in_page));
5690

5691
		copy_pages(dst->pages[dst_i], dst->pages[src_i],
5692 5693 5694 5695 5696 5697 5698 5699 5700 5701 5702 5703 5704 5705 5706 5707
			   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;
5708
	size_t start_offset = dst->start & ((u64)PAGE_SIZE - 1);
5709 5710 5711 5712
	unsigned long dst_i;
	unsigned long src_i;

	if (src_offset + len > dst->len) {
5713 5714
		btrfs_err(dst->fs_info, "memmove bogus src_offset %lu move "
		       "len %lu len %lu", src_offset, len, dst->len);
5715 5716 5717
		BUG_ON(1);
	}
	if (dst_offset + len > dst->len) {
5718 5719
		btrfs_err(dst->fs_info, "memmove bogus dst_offset %lu move "
		       "len %lu len %lu", dst_offset, len, dst->len);
5720 5721
		BUG_ON(1);
	}
5722
	if (dst_offset < src_offset) {
5723 5724 5725
		memcpy_extent_buffer(dst, dst_offset, src_offset, len);
		return;
	}
C
Chris Mason 已提交
5726
	while (len > 0) {
5727 5728
		dst_i = (start_offset + dst_end) >> PAGE_SHIFT;
		src_i = (start_offset + src_end) >> PAGE_SHIFT;
5729 5730

		dst_off_in_page = (start_offset + dst_end) &
5731
			(PAGE_SIZE - 1);
5732
		src_off_in_page = (start_offset + src_end) &
5733
			(PAGE_SIZE - 1);
5734 5735 5736

		cur = min_t(unsigned long, len, src_off_in_page + 1);
		cur = min(cur, dst_off_in_page + 1);
5737
		copy_pages(dst->pages[dst_i], dst->pages[src_i],
5738 5739 5740 5741 5742 5743 5744 5745
			   dst_off_in_page - cur + 1,
			   src_off_in_page - cur + 1, cur);

		dst_end -= cur;
		src_end -= cur;
		len -= cur;
	}
}
5746

5747
int try_release_extent_buffer(struct page *page)
5748
{
5749 5750
	struct extent_buffer *eb;

5751 5752 5753 5754 5755 5756 5757
	/*
	 * 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
5758
		return 1;
5759
	}
5760

5761 5762
	eb = (struct extent_buffer *)page->private;
	BUG_ON(!eb);
5763 5764

	/*
5765 5766 5767
	 * 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.
5768
	 */
5769
	spin_lock(&eb->refs_lock);
5770
	if (atomic_read(&eb->refs) != 1 || extent_buffer_under_io(eb)) {
5771 5772 5773
		spin_unlock(&eb->refs_lock);
		spin_unlock(&page->mapping->private_lock);
		return 0;
5774
	}
5775
	spin_unlock(&page->mapping->private_lock);
5776

5777
	/*
5778 5779
	 * 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.
5780
	 */
5781 5782 5783
	if (!test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags)) {
		spin_unlock(&eb->refs_lock);
		return 0;
5784
	}
5785

5786
	return release_extent_buffer(eb);
5787
}