1 // SPDX-License-Identifier: GPL-2.0
2
3 #include <linux/bitops.h>
4 #include <linux/slab.h>
5 #include <linux/bio.h>
6 #include <linux/mm.h>
7 #include <linux/pagemap.h>
8 #include <linux/page-flags.h>
9 #include <linux/spinlock.h>
10 #include <linux/blkdev.h>
11 #include <linux/swap.h>
12 #include <linux/writeback.h>
13 #include <linux/pagevec.h>
14 #include <linux/prefetch.h>
15 #include <linux/cleancache.h>
16 #include "extent_io.h"
17 #include "extent_map.h"
18 #include "ctree.h"
19 #include "btrfs_inode.h"
20 #include "volumes.h"
21 #include "check-integrity.h"
22 #include "locking.h"
23 #include "rcu-string.h"
24 #include "backref.h"
25 #include "disk-io.h"
26
27 static struct kmem_cache *extent_state_cache;
28 static struct kmem_cache *extent_buffer_cache;
29 static struct bio_set *btrfs_bioset;
30
31 static inline bool extent_state_in_tree(const struct extent_state *state)
32 {
33 return !RB_EMPTY_NODE(&state->rb_node);
34 }
35
36 #ifdef CONFIG_BTRFS_DEBUG
37 static LIST_HEAD(buffers);
38 static LIST_HEAD(states);
39
40 static DEFINE_SPINLOCK(leak_lock);
41
42 static inline
43 void btrfs_leak_debug_add(struct list_head *new, struct list_head *head)
44 {
45 unsigned long flags;
46
47 spin_lock_irqsave(&leak_lock, flags);
48 list_add(new, head);
49 spin_unlock_irqrestore(&leak_lock, flags);
50 }
51
52 static inline
53 void btrfs_leak_debug_del(struct list_head *entry)
54 {
55 unsigned long flags;
56
57 spin_lock_irqsave(&leak_lock, flags);
58 list_del(entry);
59 spin_unlock_irqrestore(&leak_lock, flags);
60 }
61
62 static inline
63 void btrfs_leak_debug_check(void)
64 {
65 struct extent_state *state;
66 struct extent_buffer *eb;
67
68 while (!list_empty(&states)) {
69 state = list_entry(states.next, struct extent_state, leak_list);
70 pr_err("BTRFS: state leak: start %llu end %llu state %u in tree %d refs %d\n",
71 state->start, state->end, state->state,
72 extent_state_in_tree(state),
73 refcount_read(&state->refs));
74 list_del(&state->leak_list);
75 kmem_cache_free(extent_state_cache, state);
76 }
77
78 while (!list_empty(&buffers)) {
79 eb = list_entry(buffers.next, struct extent_buffer, leak_list);
80 pr_err("BTRFS: buffer leak start %llu len %lu refs %d bflags %lu\n",
81 eb->start, eb->len, atomic_read(&eb->refs), eb->bflags);
82 list_del(&eb->leak_list);
83 kmem_cache_free(extent_buffer_cache, eb);
84 }
85 }
86
87 #define btrfs_debug_check_extent_io_range(tree, start, end) \
88 __btrfs_debug_check_extent_io_range(__func__, (tree), (start), (end))
89 static inline void __btrfs_debug_check_extent_io_range(const char *caller,
90 struct extent_io_tree *tree, u64 start, u64 end)
91 {
92 if (tree->ops && tree->ops->check_extent_io_range)
93 tree->ops->check_extent_io_range(tree->private_data, caller,
94 start, end);
95 }
96 #else
97 #define btrfs_leak_debug_add(new, head) do {} while (0)
98 #define btrfs_leak_debug_del(entry) do {} while (0)
99 #define btrfs_leak_debug_check() do {} while (0)
100 #define btrfs_debug_check_extent_io_range(c, s, e) do {} while (0)
101 #endif
102
103 #define BUFFER_LRU_MAX 64
104
105 struct tree_entry {
106 u64 start;
107 u64 end;
108 struct rb_node rb_node;
109 };
110
111 struct extent_page_data {
112 struct bio *bio;
113 struct extent_io_tree *tree;
114 /* tells writepage not to lock the state bits for this range
115 * it still does the unlocking
116 */
117 unsigned int extent_locked:1;
118
119 /* tells the submit_bio code to use REQ_SYNC */
120 unsigned int sync_io:1;
121 };
122
123 static int add_extent_changeset(struct extent_state *state, unsigned bits,
124 struct extent_changeset *changeset,
125 int set)
126 {
127 int ret;
128
129 if (!changeset)
130 return 0;
131 if (set && (state->state & bits) == bits)
132 return 0;
133 if (!set && (state->state & bits) == 0)
134 return 0;
135 changeset->bytes_changed += state->end - state->start + 1;
136 ret = ulist_add(&changeset->range_changed, state->start, state->end,
137 GFP_ATOMIC);
138 return ret;
139 }
140
141 static void flush_write_bio(struct extent_page_data *epd);
142
143 static inline struct btrfs_fs_info *
144 tree_fs_info(struct extent_io_tree *tree)
145 {
146 if (tree->ops)
147 return tree->ops->tree_fs_info(tree->private_data);
148 return NULL;
149 }
150
151 int __init extent_io_init(void)
152 {
153 extent_state_cache = kmem_cache_create("btrfs_extent_state",
154 sizeof(struct extent_state), 0,
155 SLAB_MEM_SPREAD, NULL);
156 if (!extent_state_cache)
157 return -ENOMEM;
158
159 extent_buffer_cache = kmem_cache_create("btrfs_extent_buffer",
160 sizeof(struct extent_buffer), 0,
161 SLAB_MEM_SPREAD, NULL);
162 if (!extent_buffer_cache)
163 goto free_state_cache;
164
165 btrfs_bioset = bioset_create(BIO_POOL_SIZE,
166 offsetof(struct btrfs_io_bio, bio),
167 BIOSET_NEED_BVECS);
168 if (!btrfs_bioset)
169 goto free_buffer_cache;
170
171 if (bioset_integrity_create(btrfs_bioset, BIO_POOL_SIZE))
172 goto free_bioset;
173
174 return 0;
175
176 free_bioset:
177 bioset_free(btrfs_bioset);
178 btrfs_bioset = NULL;
179
180 free_buffer_cache:
181 kmem_cache_destroy(extent_buffer_cache);
182 extent_buffer_cache = NULL;
183
184 free_state_cache:
185 kmem_cache_destroy(extent_state_cache);
186 extent_state_cache = NULL;
187 return -ENOMEM;
188 }
189
190 void __cold extent_io_exit(void)
191 {
192 btrfs_leak_debug_check();
193
194 /*
195 * Make sure all delayed rcu free are flushed before we
196 * destroy caches.
197 */
198 rcu_barrier();
199 kmem_cache_destroy(extent_state_cache);
200 kmem_cache_destroy(extent_buffer_cache);
201 if (btrfs_bioset)
202 bioset_free(btrfs_bioset);
203 }
204
205 void extent_io_tree_init(struct extent_io_tree *tree,
206 void *private_data)
207 {
208 tree->state = RB_ROOT;
209 tree->ops = NULL;
210 tree->dirty_bytes = 0;
211 spin_lock_init(&tree->lock);
212 tree->private_data = private_data;
213 }
214
215 static struct extent_state *alloc_extent_state(gfp_t mask)
216 {
217 struct extent_state *state;
218
219 /*
220 * The given mask might be not appropriate for the slab allocator,
221 * drop the unsupported bits
222 */
223 mask &= ~(__GFP_DMA32|__GFP_HIGHMEM);
224 state = kmem_cache_alloc(extent_state_cache, mask);
225 if (!state)
226 return state;
227 state->state = 0;
228 state->failrec = NULL;
229 RB_CLEAR_NODE(&state->rb_node);
230 btrfs_leak_debug_add(&state->leak_list, &states);
231 refcount_set(&state->refs, 1);
232 init_waitqueue_head(&state->wq);
233 trace_alloc_extent_state(state, mask, _RET_IP_);
234 return state;
235 }
236
237 void free_extent_state(struct extent_state *state)
238 {
239 if (!state)
240 return;
241 if (refcount_dec_and_test(&state->refs)) {
242 WARN_ON(extent_state_in_tree(state));
243 btrfs_leak_debug_del(&state->leak_list);
244 trace_free_extent_state(state, _RET_IP_);
245 kmem_cache_free(extent_state_cache, state);
246 }
247 }
248
249 static struct rb_node *tree_insert(struct rb_root *root,
250 struct rb_node *search_start,
251 u64 offset,
252 struct rb_node *node,
253 struct rb_node ***p_in,
254 struct rb_node **parent_in)
255 {
256 struct rb_node **p;
257 struct rb_node *parent = NULL;
258 struct tree_entry *entry;
259
260 if (p_in && parent_in) {
261 p = *p_in;
262 parent = *parent_in;
263 goto do_insert;
264 }
265
266 p = search_start ? &search_start : &root->rb_node;
267 while (*p) {
268 parent = *p;
269 entry = rb_entry(parent, struct tree_entry, rb_node);
270
271 if (offset < entry->start)
272 p = &(*p)->rb_left;
273 else if (offset > entry->end)
274 p = &(*p)->rb_right;
275 else
276 return parent;
277 }
278
279 do_insert:
280 rb_link_node(node, parent, p);
281 rb_insert_color(node, root);
282 return NULL;
283 }
284
285 static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset,
286 struct rb_node **prev_ret,
287 struct rb_node **next_ret,
288 struct rb_node ***p_ret,
289 struct rb_node **parent_ret)
290 {
291 struct rb_root *root = &tree->state;
292 struct rb_node **n = &root->rb_node;
293 struct rb_node *prev = NULL;
294 struct rb_node *orig_prev = NULL;
295 struct tree_entry *entry;
296 struct tree_entry *prev_entry = NULL;
297
298 while (*n) {
299 prev = *n;
300 entry = rb_entry(prev, struct tree_entry, rb_node);
301 prev_entry = entry;
302
303 if (offset < entry->start)
304 n = &(*n)->rb_left;
305 else if (offset > entry->end)
306 n = &(*n)->rb_right;
307 else
308 return *n;
309 }
310
311 if (p_ret)
312 *p_ret = n;
313 if (parent_ret)
314 *parent_ret = prev;
315
316 if (prev_ret) {
317 orig_prev = prev;
318 while (prev && offset > prev_entry->end) {
319 prev = rb_next(prev);
320 prev_entry = rb_entry(prev, struct tree_entry, rb_node);
321 }
322 *prev_ret = prev;
323 prev = orig_prev;
324 }
325
326 if (next_ret) {
327 prev_entry = rb_entry(prev, struct tree_entry, rb_node);
328 while (prev && offset < prev_entry->start) {
329 prev = rb_prev(prev);
330 prev_entry = rb_entry(prev, struct tree_entry, rb_node);
331 }
332 *next_ret = prev;
333 }
334 return NULL;
335 }
336
337 static inline struct rb_node *
338 tree_search_for_insert(struct extent_io_tree *tree,
339 u64 offset,
340 struct rb_node ***p_ret,
341 struct rb_node **parent_ret)
342 {
343 struct rb_node *prev = NULL;
344 struct rb_node *ret;
345
346 ret = __etree_search(tree, offset, &prev, NULL, p_ret, parent_ret);
347 if (!ret)
348 return prev;
349 return ret;
350 }
351
352 static inline struct rb_node *tree_search(struct extent_io_tree *tree,
353 u64 offset)
354 {
355 return tree_search_for_insert(tree, offset, NULL, NULL);
356 }
357
358 static void merge_cb(struct extent_io_tree *tree, struct extent_state *new,
359 struct extent_state *other)
360 {
361 if (tree->ops && tree->ops->merge_extent_hook)
362 tree->ops->merge_extent_hook(tree->private_data, new, other);
363 }
364
365 /*
366 * utility function to look for merge candidates inside a given range.
367 * Any extents with matching state are merged together into a single
368 * extent in the tree. Extents with EXTENT_IO in their state field
369 * are not merged because the end_io handlers need to be able to do
370 * operations on them without sleeping (or doing allocations/splits).
371 *
372 * This should be called with the tree lock held.
373 */
374 static void merge_state(struct extent_io_tree *tree,
375 struct extent_state *state)
376 {
377 struct extent_state *other;
378 struct rb_node *other_node;
379
380 if (state->state & (EXTENT_IOBITS | EXTENT_BOUNDARY))
381 return;
382
383 other_node = rb_prev(&state->rb_node);
384 if (other_node) {
385 other = rb_entry(other_node, struct extent_state, rb_node);
386 if (other->end == state->start - 1 &&
387 other->state == state->state) {
388 merge_cb(tree, state, other);
389 state->start = other->start;
390 rb_erase(&other->rb_node, &tree->state);
391 RB_CLEAR_NODE(&other->rb_node);
392 free_extent_state(other);
393 }
394 }
395 other_node = rb_next(&state->rb_node);
396 if (other_node) {
397 other = rb_entry(other_node, struct extent_state, rb_node);
398 if (other->start == state->end + 1 &&
399 other->state == state->state) {
400 merge_cb(tree, state, other);
401 state->end = other->end;
402 rb_erase(&other->rb_node, &tree->state);
403 RB_CLEAR_NODE(&other->rb_node);
404 free_extent_state(other);
405 }
406 }
407 }
408
409 static void set_state_cb(struct extent_io_tree *tree,
410 struct extent_state *state, unsigned *bits)
411 {
412 if (tree->ops && tree->ops->set_bit_hook)
413 tree->ops->set_bit_hook(tree->private_data, state, bits);
414 }
415
416 static void clear_state_cb(struct extent_io_tree *tree,
417 struct extent_state *state, unsigned *bits)
418 {
419 if (tree->ops && tree->ops->clear_bit_hook)
420 tree->ops->clear_bit_hook(tree->private_data, state, bits);
421 }
422
423 static void set_state_bits(struct extent_io_tree *tree,
424 struct extent_state *state, unsigned *bits,
425 struct extent_changeset *changeset);
426
427 /*
428 * insert an extent_state struct into the tree. 'bits' are set on the
429 * struct before it is inserted.
430 *
431 * This may return -EEXIST if the extent is already there, in which case the
432 * state struct is freed.
433 *
434 * The tree lock is not taken internally. This is a utility function and
435 * probably isn't what you want to call (see set/clear_extent_bit).
436 */
437 static int insert_state(struct extent_io_tree *tree,
438 struct extent_state *state, u64 start, u64 end,
439 struct rb_node ***p,
440 struct rb_node **parent,
441 unsigned *bits, struct extent_changeset *changeset)
442 {
443 struct rb_node *node;
444
445 if (end < start)
446 WARN(1, KERN_ERR "BTRFS: end < start %llu %llu\n",
447 end, start);
448 state->start = start;
449 state->end = end;
450
451 set_state_bits(tree, state, bits, changeset);
452
453 node = tree_insert(&tree->state, NULL, end, &state->rb_node, p, parent);
454 if (node) {
455 struct extent_state *found;
456 found = rb_entry(node, struct extent_state, rb_node);
457 pr_err("BTRFS: found node %llu %llu on insert of %llu %llu\n",
458 found->start, found->end, start, end);
459 return -EEXIST;
460 }
461 merge_state(tree, state);
462 return 0;
463 }
464
465 static void split_cb(struct extent_io_tree *tree, struct extent_state *orig,
466 u64 split)
467 {
468 if (tree->ops && tree->ops->split_extent_hook)
469 tree->ops->split_extent_hook(tree->private_data, orig, split);
470 }
471
472 /*
473 * split a given extent state struct in two, inserting the preallocated
474 * struct 'prealloc' as the newly created second half. 'split' indicates an
475 * offset inside 'orig' where it should be split.
476 *
477 * Before calling,
478 * the tree has 'orig' at [orig->start, orig->end]. After calling, there
479 * are two extent state structs in the tree:
480 * prealloc: [orig->start, split - 1]
481 * orig: [ split, orig->end ]
482 *
483 * The tree locks are not taken by this function. They need to be held
484 * by the caller.
485 */
486 static int split_state(struct extent_io_tree *tree, struct extent_state *orig,
487 struct extent_state *prealloc, u64 split)
488 {
489 struct rb_node *node;
490
491 split_cb(tree, orig, split);
492
493 prealloc->start = orig->start;
494 prealloc->end = split - 1;
495 prealloc->state = orig->state;
496 orig->start = split;
497
498 node = tree_insert(&tree->state, &orig->rb_node, prealloc->end,
499 &prealloc->rb_node, NULL, NULL);
500 if (node) {
501 free_extent_state(prealloc);
502 return -EEXIST;
503 }
504 return 0;
505 }
506
507 static struct extent_state *next_state(struct extent_state *state)
508 {
509 struct rb_node *next = rb_next(&state->rb_node);
510 if (next)
511 return rb_entry(next, struct extent_state, rb_node);
512 else
513 return NULL;
514 }
515
516 /*
517 * utility function to clear some bits in an extent state struct.
518 * it will optionally wake up any one waiting on this state (wake == 1).
519 *
520 * If no bits are set on the state struct after clearing things, the
521 * struct is freed and removed from the tree
522 */
523 static struct extent_state *clear_state_bit(struct extent_io_tree *tree,
524 struct extent_state *state,
525 unsigned *bits, int wake,
526 struct extent_changeset *changeset)
527 {
528 struct extent_state *next;
529 unsigned bits_to_clear = *bits & ~EXTENT_CTLBITS;
530 int ret;
531
532 if ((bits_to_clear & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) {
533 u64 range = state->end - state->start + 1;
534 WARN_ON(range > tree->dirty_bytes);
535 tree->dirty_bytes -= range;
536 }
537 clear_state_cb(tree, state, bits);
538 ret = add_extent_changeset(state, bits_to_clear, changeset, 0);
539 BUG_ON(ret < 0);
540 state->state &= ~bits_to_clear;
541 if (wake)
542 wake_up(&state->wq);
543 if (state->state == 0) {
544 next = next_state(state);
545 if (extent_state_in_tree(state)) {
546 rb_erase(&state->rb_node, &tree->state);
547 RB_CLEAR_NODE(&state->rb_node);
548 free_extent_state(state);
549 } else {
550 WARN_ON(1);
551 }
552 } else {
553 merge_state(tree, state);
554 next = next_state(state);
555 }
556 return next;
557 }
558
559 static struct extent_state *
560 alloc_extent_state_atomic(struct extent_state *prealloc)
561 {
562 if (!prealloc)
563 prealloc = alloc_extent_state(GFP_ATOMIC);
564
565 return prealloc;
566 }
567
568 static void extent_io_tree_panic(struct extent_io_tree *tree, int err)
569 {
570 btrfs_panic(tree_fs_info(tree), err,
571 "Locking error: Extent tree was modified by another thread while locked.");
572 }
573
574 /*
575 * clear some bits on a range in the tree. This may require splitting
576 * or inserting elements in the tree, so the gfp mask is used to
577 * indicate which allocations or sleeping are allowed.
578 *
579 * pass 'wake' == 1 to kick any sleepers, and 'delete' == 1 to remove
580 * the given range from the tree regardless of state (ie for truncate).
581 *
582 * the range [start, end] is inclusive.
583 *
584 * This takes the tree lock, and returns 0 on success and < 0 on error.
585 */
586 int __clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
587 unsigned bits, int wake, int delete,
588 struct extent_state **cached_state,
589 gfp_t mask, struct extent_changeset *changeset)
590 {
591 struct extent_state *state;
592 struct extent_state *cached;
593 struct extent_state *prealloc = NULL;
594 struct rb_node *node;
595 u64 last_end;
596 int err;
597 int clear = 0;
598
599 btrfs_debug_check_extent_io_range(tree, start, end);
600
601 if (bits & EXTENT_DELALLOC)
602 bits |= EXTENT_NORESERVE;
603
604 if (delete)
605 bits |= ~EXTENT_CTLBITS;
606 bits |= EXTENT_FIRST_DELALLOC;
607
608 if (bits & (EXTENT_IOBITS | EXTENT_BOUNDARY))
609 clear = 1;
610 again:
611 if (!prealloc && gfpflags_allow_blocking(mask)) {
612 /*
613 * Don't care for allocation failure here because we might end
614 * up not needing the pre-allocated extent state at all, which
615 * is the case if we only have in the tree extent states that
616 * cover our input range and don't cover too any other range.
617 * If we end up needing a new extent state we allocate it later.
618 */
619 prealloc = alloc_extent_state(mask);
620 }
621
622 spin_lock(&tree->lock);
623 if (cached_state) {
624 cached = *cached_state;
625
626 if (clear) {
627 *cached_state = NULL;
628 cached_state = NULL;
629 }
630
631 if (cached && extent_state_in_tree(cached) &&
632 cached->start <= start && cached->end > start) {
633 if (clear)
634 refcount_dec(&cached->refs);
635 state = cached;
636 goto hit_next;
637 }
638 if (clear)
639 free_extent_state(cached);
640 }
641 /*
642 * this search will find the extents that end after
643 * our range starts
644 */
645 node = tree_search(tree, start);
646 if (!node)
647 goto out;
648 state = rb_entry(node, struct extent_state, rb_node);
649 hit_next:
650 if (state->start > end)
651 goto out;
652 WARN_ON(state->end < start);
653 last_end = state->end;
654
655 /* the state doesn't have the wanted bits, go ahead */
656 if (!(state->state & bits)) {
657 state = next_state(state);
658 goto next;
659 }
660
661 /*
662 * | ---- desired range ---- |
663 * | state | or
664 * | ------------- state -------------- |
665 *
666 * We need to split the extent we found, and may flip
667 * bits on second half.
668 *
669 * If the extent we found extends past our range, we
670 * just split and search again. It'll get split again
671 * the next time though.
672 *
673 * If the extent we found is inside our range, we clear
674 * the desired bit on it.
675 */
676
677 if (state->start < start) {
678 prealloc = alloc_extent_state_atomic(prealloc);
679 BUG_ON(!prealloc);
680 err = split_state(tree, state, prealloc, start);
681 if (err)
682 extent_io_tree_panic(tree, err);
683
684 prealloc = NULL;
685 if (err)
686 goto out;
687 if (state->end <= end) {
688 state = clear_state_bit(tree, state, &bits, wake,
689 changeset);
690 goto next;
691 }
692 goto search_again;
693 }
694 /*
695 * | ---- desired range ---- |
696 * | state |
697 * We need to split the extent, and clear the bit
698 * on the first half
699 */
700 if (state->start <= end && state->end > end) {
701 prealloc = alloc_extent_state_atomic(prealloc);
702 BUG_ON(!prealloc);
703 err = split_state(tree, state, prealloc, end + 1);
704 if (err)
705 extent_io_tree_panic(tree, err);
706
707 if (wake)
708 wake_up(&state->wq);
709
710 clear_state_bit(tree, prealloc, &bits, wake, changeset);
711
712 prealloc = NULL;
713 goto out;
714 }
715
716 state = clear_state_bit(tree, state, &bits, wake, changeset);
717 next:
718 if (last_end == (u64)-1)
719 goto out;
720 start = last_end + 1;
721 if (start <= end && state && !need_resched())
722 goto hit_next;
723
724 search_again:
725 if (start > end)
726 goto out;
727 spin_unlock(&tree->lock);
728 if (gfpflags_allow_blocking(mask))
729 cond_resched();
730 goto again;
731
732 out:
733 spin_unlock(&tree->lock);
734 if (prealloc)
735 free_extent_state(prealloc);
736
737 return 0;
738
739 }
740
741 static void wait_on_state(struct extent_io_tree *tree,
742 struct extent_state *state)
743 __releases(tree->lock)
744 __acquires(tree->lock)
745 {
746 DEFINE_WAIT(wait);
747 prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE);
748 spin_unlock(&tree->lock);
749 schedule();
750 spin_lock(&tree->lock);
751 finish_wait(&state->wq, &wait);
752 }
753
754 /*
755 * waits for one or more bits to clear on a range in the state tree.
756 * The range [start, end] is inclusive.
757 * The tree lock is taken by this function
758 */
759 static void wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
760 unsigned long bits)
761 {
762 struct extent_state *state;
763 struct rb_node *node;
764
765 btrfs_debug_check_extent_io_range(tree, start, end);
766
767 spin_lock(&tree->lock);
768 again:
769 while (1) {
770 /*
771 * this search will find all the extents that end after
772 * our range starts
773 */
774 node = tree_search(tree, start);
775 process_node:
776 if (!node)
777 break;
778
779 state = rb_entry(node, struct extent_state, rb_node);
780
781 if (state->start > end)
782 goto out;
783
784 if (state->state & bits) {
785 start = state->start;
786 refcount_inc(&state->refs);
787 wait_on_state(tree, state);
788 free_extent_state(state);
789 goto again;
790 }
791 start = state->end + 1;
792
793 if (start > end)
794 break;
795
796 if (!cond_resched_lock(&tree->lock)) {
797 node = rb_next(node);
798 goto process_node;
799 }
800 }
801 out:
802 spin_unlock(&tree->lock);
803 }
804
805 static void set_state_bits(struct extent_io_tree *tree,
806 struct extent_state *state,
807 unsigned *bits, struct extent_changeset *changeset)
808 {
809 unsigned bits_to_set = *bits & ~EXTENT_CTLBITS;
810 int ret;
811
812 set_state_cb(tree, state, bits);
813 if ((bits_to_set & EXTENT_DIRTY) && !(state->state & EXTENT_DIRTY)) {
814 u64 range = state->end - state->start + 1;
815 tree->dirty_bytes += range;
816 }
817 ret = add_extent_changeset(state, bits_to_set, changeset, 1);
818 BUG_ON(ret < 0);
819 state->state |= bits_to_set;
820 }
821
822 static void cache_state_if_flags(struct extent_state *state,
823 struct extent_state **cached_ptr,
824 unsigned flags)
825 {
826 if (cached_ptr && !(*cached_ptr)) {
827 if (!flags || (state->state & flags)) {
828 *cached_ptr = state;
829 refcount_inc(&state->refs);
830 }
831 }
832 }
833
834 static void cache_state(struct extent_state *state,
835 struct extent_state **cached_ptr)
836 {
837 return cache_state_if_flags(state, cached_ptr,
838 EXTENT_IOBITS | EXTENT_BOUNDARY);
839 }
840
841 /*
842 * set some bits on a range in the tree. This may require allocations or
843 * sleeping, so the gfp mask is used to indicate what is allowed.
844 *
845 * If any of the exclusive bits are set, this will fail with -EEXIST if some
846 * part of the range already has the desired bits set. The start of the
847 * existing range is returned in failed_start in this case.
848 *
849 * [start, end] is inclusive This takes the tree lock.
850 */
851
852 static int __must_check
853 __set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
854 unsigned bits, unsigned exclusive_bits,
855 u64 *failed_start, struct extent_state **cached_state,
856 gfp_t mask, struct extent_changeset *changeset)
857 {
858 struct extent_state *state;
859 struct extent_state *prealloc = NULL;
860 struct rb_node *node;
861 struct rb_node **p;
862 struct rb_node *parent;
863 int err = 0;
864 u64 last_start;
865 u64 last_end;
866
867 btrfs_debug_check_extent_io_range(tree, start, end);
868
869 bits |= EXTENT_FIRST_DELALLOC;
870 again:
871 if (!prealloc && gfpflags_allow_blocking(mask)) {
872 /*
873 * Don't care for allocation failure here because we might end
874 * up not needing the pre-allocated extent state at all, which
875 * is the case if we only have in the tree extent states that
876 * cover our input range and don't cover too any other range.
877 * If we end up needing a new extent state we allocate it later.
878 */
879 prealloc = alloc_extent_state(mask);
880 }
881
882 spin_lock(&tree->lock);
883 if (cached_state && *cached_state) {
884 state = *cached_state;
885 if (state->start <= start && state->end > start &&
886 extent_state_in_tree(state)) {
887 node = &state->rb_node;
888 goto hit_next;
889 }
890 }
891 /*
892 * this search will find all the extents that end after
893 * our range starts.
894 */
895 node = tree_search_for_insert(tree, start, &p, &parent);
896 if (!node) {
897 prealloc = alloc_extent_state_atomic(prealloc);
898 BUG_ON(!prealloc);
899 err = insert_state(tree, prealloc, start, end,
900 &p, &parent, &bits, changeset);
901 if (err)
902 extent_io_tree_panic(tree, err);
903
904 cache_state(prealloc, cached_state);
905 prealloc = NULL;
906 goto out;
907 }
908 state = rb_entry(node, struct extent_state, rb_node);
909 hit_next:
910 last_start = state->start;
911 last_end = state->end;
912
913 /*
914 * | ---- desired range ---- |
915 * | state |
916 *
917 * Just lock what we found and keep going
918 */
919 if (state->start == start && state->end <= end) {
920 if (state->state & exclusive_bits) {
921 *failed_start = state->start;
922 err = -EEXIST;
923 goto out;
924 }
925
926 set_state_bits(tree, state, &bits, changeset);
927 cache_state(state, cached_state);
928 merge_state(tree, state);
929 if (last_end == (u64)-1)
930 goto out;
931 start = last_end + 1;
932 state = next_state(state);
933 if (start < end && state && state->start == start &&
934 !need_resched())
935 goto hit_next;
936 goto search_again;
937 }
938
939 /*
940 * | ---- desired range ---- |
941 * | state |
942 * or
943 * | ------------- state -------------- |
944 *
945 * We need to split the extent we found, and may flip bits on
946 * second half.
947 *
948 * If the extent we found extends past our
949 * range, we just split and search again. It'll get split
950 * again the next time though.
951 *
952 * If the extent we found is inside our range, we set the
953 * desired bit on it.
954 */
955 if (state->start < start) {
956 if (state->state & exclusive_bits) {
957 *failed_start = start;
958 err = -EEXIST;
959 goto out;
960 }
961
962 prealloc = alloc_extent_state_atomic(prealloc);
963 BUG_ON(!prealloc);
964 err = split_state(tree, state, prealloc, start);
965 if (err)
966 extent_io_tree_panic(tree, err);
967
968 prealloc = NULL;
969 if (err)
970 goto out;
971 if (state->end <= end) {
972 set_state_bits(tree, state, &bits, changeset);
973 cache_state(state, cached_state);
974 merge_state(tree, state);
975 if (last_end == (u64)-1)
976 goto out;
977 start = last_end + 1;
978 state = next_state(state);
979 if (start < end && state && state->start == start &&
980 !need_resched())
981 goto hit_next;
982 }
983 goto search_again;
984 }
985 /*
986 * | ---- desired range ---- |
987 * | state | or | state |
988 *
989 * There's a hole, we need to insert something in it and
990 * ignore the extent we found.
991 */
992 if (state->start > start) {
993 u64 this_end;
994 if (end < last_start)
995 this_end = end;
996 else
997 this_end = last_start - 1;
998
999 prealloc = alloc_extent_state_atomic(prealloc);
1000 BUG_ON(!prealloc);
1001
1002 /*
1003 * Avoid to free 'prealloc' if it can be merged with
1004 * the later extent.
1005 */
1006 err = insert_state(tree, prealloc, start, this_end,
1007 NULL, NULL, &bits, changeset);
1008 if (err)
1009 extent_io_tree_panic(tree, err);
1010
1011 cache_state(prealloc, cached_state);
1012 prealloc = NULL;
1013 start = this_end + 1;
1014 goto search_again;
1015 }
1016 /*
1017 * | ---- desired range ---- |
1018 * | state |
1019 * We need to split the extent, and set the bit
1020 * on the first half
1021 */
1022 if (state->start <= end && state->end > end) {
1023 if (state->state & exclusive_bits) {
1024 *failed_start = start;
1025 err = -EEXIST;
1026 goto out;
1027 }
1028
1029 prealloc = alloc_extent_state_atomic(prealloc);
1030 BUG_ON(!prealloc);
1031 err = split_state(tree, state, prealloc, end + 1);
1032 if (err)
1033 extent_io_tree_panic(tree, err);
1034
1035 set_state_bits(tree, prealloc, &bits, changeset);
1036 cache_state(prealloc, cached_state);
1037 merge_state(tree, prealloc);
1038 prealloc = NULL;
1039 goto out;
1040 }
1041
1042 search_again:
1043 if (start > end)
1044 goto out;
1045 spin_unlock(&tree->lock);
1046 if (gfpflags_allow_blocking(mask))
1047 cond_resched();
1048 goto again;
1049
1050 out:
1051 spin_unlock(&tree->lock);
1052 if (prealloc)
1053 free_extent_state(prealloc);
1054
1055 return err;
1056
1057 }
1058
1059 int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
1060 unsigned bits, u64 * failed_start,
1061 struct extent_state **cached_state, gfp_t mask)
1062 {
1063 return __set_extent_bit(tree, start, end, bits, 0, failed_start,
1064 cached_state, mask, NULL);
1065 }
1066
1067
1068 /**
1069 * convert_extent_bit - convert all bits in a given range from one bit to
1070 * another
1071 * @tree: the io tree to search
1072 * @start: the start offset in bytes
1073 * @end: the end offset in bytes (inclusive)
1074 * @bits: the bits to set in this range
1075 * @clear_bits: the bits to clear in this range
1076 * @cached_state: state that we're going to cache
1077 *
1078 * This will go through and set bits for the given range. If any states exist
1079 * already in this range they are set with the given bit and cleared of the
1080 * clear_bits. This is only meant to be used by things that are mergeable, ie
1081 * converting from say DELALLOC to DIRTY. This is not meant to be used with
1082 * boundary bits like LOCK.
1083 *
1084 * All allocations are done with GFP_NOFS.
1085 */
1086 int convert_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
1087 unsigned bits, unsigned clear_bits,
1088 struct extent_state **cached_state)
1089 {
1090 struct extent_state *state;
1091 struct extent_state *prealloc = NULL;
1092 struct rb_node *node;
1093 struct rb_node **p;
1094 struct rb_node *parent;
1095 int err = 0;
1096 u64 last_start;
1097 u64 last_end;
1098 bool first_iteration = true;
1099
1100 btrfs_debug_check_extent_io_range(tree, start, end);
1101
1102 again:
1103 if (!prealloc) {
1104 /*
1105 * Best effort, don't worry if extent state allocation fails
1106 * here for the first iteration. We might have a cached state
1107 * that matches exactly the target range, in which case no
1108 * extent state allocations are needed. We'll only know this
1109 * after locking the tree.
1110 */
1111 prealloc = alloc_extent_state(GFP_NOFS);
1112 if (!prealloc && !first_iteration)
1113 return -ENOMEM;
1114 }
1115
1116 spin_lock(&tree->lock);
1117 if (cached_state && *cached_state) {
1118 state = *cached_state;
1119 if (state->start <= start && state->end > start &&
1120 extent_state_in_tree(state)) {
1121 node = &state->rb_node;
1122 goto hit_next;
1123 }
1124 }
1125
1126 /*
1127 * this search will find all the extents that end after
1128 * our range starts.
1129 */
1130 node = tree_search_for_insert(tree, start, &p, &parent);
1131 if (!node) {
1132 prealloc = alloc_extent_state_atomic(prealloc);
1133 if (!prealloc) {
1134 err = -ENOMEM;
1135 goto out;
1136 }
1137 err = insert_state(tree, prealloc, start, end,
1138 &p, &parent, &bits, NULL);
1139 if (err)
1140 extent_io_tree_panic(tree, err);
1141 cache_state(prealloc, cached_state);
1142 prealloc = NULL;
1143 goto out;
1144 }
1145 state = rb_entry(node, struct extent_state, rb_node);
1146 hit_next:
1147 last_start = state->start;
1148 last_end = state->end;
1149
1150 /*
1151 * | ---- desired range ---- |
1152 * | state |
1153 *
1154 * Just lock what we found and keep going
1155 */
1156 if (state->start == start && state->end <= end) {
1157 set_state_bits(tree, state, &bits, NULL);
1158 cache_state(state, cached_state);
1159 state = clear_state_bit(tree, state, &clear_bits, 0, NULL);
1160 if (last_end == (u64)-1)
1161 goto out;
1162 start = last_end + 1;
1163 if (start < end && state && state->start == start &&
1164 !need_resched())
1165 goto hit_next;
1166 goto search_again;
1167 }
1168
1169 /*
1170 * | ---- desired range ---- |
1171 * | state |
1172 * or
1173 * | ------------- state -------------- |
1174 *
1175 * We need to split the extent we found, and may flip bits on
1176 * second half.
1177 *
1178 * If the extent we found extends past our
1179 * range, we just split and search again. It'll get split
1180 * again the next time though.
1181 *
1182 * If the extent we found is inside our range, we set the
1183 * desired bit on it.
1184 */
1185 if (state->start < start) {
1186 prealloc = alloc_extent_state_atomic(prealloc);
1187 if (!prealloc) {
1188 err = -ENOMEM;
1189 goto out;
1190 }
1191 err = split_state(tree, state, prealloc, start);
1192 if (err)
1193 extent_io_tree_panic(tree, err);
1194 prealloc = NULL;
1195 if (err)
1196 goto out;
1197 if (state->end <= end) {
1198 set_state_bits(tree, state, &bits, NULL);
1199 cache_state(state, cached_state);
1200 state = clear_state_bit(tree, state, &clear_bits, 0,
1201 NULL);
1202 if (last_end == (u64)-1)
1203 goto out;
1204 start = last_end + 1;
1205 if (start < end && state && state->start == start &&
1206 !need_resched())
1207 goto hit_next;
1208 }
1209 goto search_again;
1210 }
1211 /*
1212 * | ---- desired range ---- |
1213 * | state | or | state |
1214 *
1215 * There's a hole, we need to insert something in it and
1216 * ignore the extent we found.
1217 */
1218 if (state->start > start) {
1219 u64 this_end;
1220 if (end < last_start)
1221 this_end = end;
1222 else
1223 this_end = last_start - 1;
1224
1225 prealloc = alloc_extent_state_atomic(prealloc);
1226 if (!prealloc) {
1227 err = -ENOMEM;
1228 goto out;
1229 }
1230
1231 /*
1232 * Avoid to free 'prealloc' if it can be merged with
1233 * the later extent.
1234 */
1235 err = insert_state(tree, prealloc, start, this_end,
1236 NULL, NULL, &bits, NULL);
1237 if (err)
1238 extent_io_tree_panic(tree, err);
1239 cache_state(prealloc, cached_state);
1240 prealloc = NULL;
1241 start = this_end + 1;
1242 goto search_again;
1243 }
1244 /*
1245 * | ---- desired range ---- |
1246 * | state |
1247 * We need to split the extent, and set the bit
1248 * on the first half
1249 */
1250 if (state->start <= end && state->end > end) {
1251 prealloc = alloc_extent_state_atomic(prealloc);
1252 if (!prealloc) {
1253 err = -ENOMEM;
1254 goto out;
1255 }
1256
1257 err = split_state(tree, state, prealloc, end + 1);
1258 if (err)
1259 extent_io_tree_panic(tree, err);
1260
1261 set_state_bits(tree, prealloc, &bits, NULL);
1262 cache_state(prealloc, cached_state);
1263 clear_state_bit(tree, prealloc, &clear_bits, 0, NULL);
1264 prealloc = NULL;
1265 goto out;
1266 }
1267
1268 search_again:
1269 if (start > end)
1270 goto out;
1271 spin_unlock(&tree->lock);
1272 cond_resched();
1273 first_iteration = false;
1274 goto again;
1275
1276 out:
1277 spin_unlock(&tree->lock);
1278 if (prealloc)
1279 free_extent_state(prealloc);
1280
1281 return err;
1282 }
1283
1284 /* wrappers around set/clear extent bit */
1285 int set_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1286 unsigned bits, struct extent_changeset *changeset)
1287 {
1288 /*
1289 * We don't support EXTENT_LOCKED yet, as current changeset will
1290 * record any bits changed, so for EXTENT_LOCKED case, it will
1291 * either fail with -EEXIST or changeset will record the whole
1292 * range.
1293 */
1294 BUG_ON(bits & EXTENT_LOCKED);
1295
1296 return __set_extent_bit(tree, start, end, bits, 0, NULL, NULL, GFP_NOFS,
1297 changeset);
1298 }
1299
1300 int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
1301 unsigned bits, int wake, int delete,
1302 struct extent_state **cached)
1303 {
1304 return __clear_extent_bit(tree, start, end, bits, wake, delete,
1305 cached, GFP_NOFS, NULL);
1306 }
1307
1308 int clear_record_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1309 unsigned bits, struct extent_changeset *changeset)
1310 {
1311 /*
1312 * Don't support EXTENT_LOCKED case, same reason as
1313 * set_record_extent_bits().
1314 */
1315 BUG_ON(bits & EXTENT_LOCKED);
1316
1317 return __clear_extent_bit(tree, start, end, bits, 0, 0, NULL, GFP_NOFS,
1318 changeset);
1319 }
1320
1321 /*
1322 * either insert or lock state struct between start and end use mask to tell
1323 * us if waiting is desired.
1324 */
1325 int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1326 struct extent_state **cached_state)
1327 {
1328 int err;
1329 u64 failed_start;
1330
1331 while (1) {
1332 err = __set_extent_bit(tree, start, end, EXTENT_LOCKED,
1333 EXTENT_LOCKED, &failed_start,
1334 cached_state, GFP_NOFS, NULL);
1335 if (err == -EEXIST) {
1336 wait_extent_bit(tree, failed_start, end, EXTENT_LOCKED);
1337 start = failed_start;
1338 } else
1339 break;
1340 WARN_ON(start > end);
1341 }
1342 return err;
1343 }
1344
1345 int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end)
1346 {
1347 int err;
1348 u64 failed_start;
1349
1350 err = __set_extent_bit(tree, start, end, EXTENT_LOCKED, EXTENT_LOCKED,
1351 &failed_start, NULL, GFP_NOFS, NULL);
1352 if (err == -EEXIST) {
1353 if (failed_start > start)
1354 clear_extent_bit(tree, start, failed_start - 1,
1355 EXTENT_LOCKED, 1, 0, NULL);
1356 return 0;
1357 }
1358 return 1;
1359 }
1360
1361 void extent_range_clear_dirty_for_io(struct inode *inode, u64 start, u64 end)
1362 {
1363 unsigned long index = start >> PAGE_SHIFT;
1364 unsigned long end_index = end >> PAGE_SHIFT;
1365 struct page *page;
1366
1367 while (index <= end_index) {
1368 page = find_get_page(inode->i_mapping, index);
1369 BUG_ON(!page); /* Pages should be in the extent_io_tree */
1370 clear_page_dirty_for_io(page);
1371 put_page(page);
1372 index++;
1373 }
1374 }
1375
1376 void extent_range_redirty_for_io(struct inode *inode, u64 start, u64 end)
1377 {
1378 unsigned long index = start >> PAGE_SHIFT;
1379 unsigned long end_index = end >> PAGE_SHIFT;
1380 struct page *page;
1381
1382 while (index <= end_index) {
1383 page = find_get_page(inode->i_mapping, index);
1384 BUG_ON(!page); /* Pages should be in the extent_io_tree */
1385 __set_page_dirty_nobuffers(page);
1386 account_page_redirty(page);
1387 put_page(page);
1388 index++;
1389 }
1390 }
1391
1392 /*
1393 * helper function to set both pages and extents in the tree writeback
1394 */
1395 static void set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end)
1396 {
1397 tree->ops->set_range_writeback(tree->private_data, start, end);
1398 }
1399
1400 /* find the first state struct with 'bits' set after 'start', and
1401 * return it. tree->lock must be held. NULL will returned if
1402 * nothing was found after 'start'
1403 */
1404 static struct extent_state *
1405 find_first_extent_bit_state(struct extent_io_tree *tree,
1406 u64 start, unsigned bits)
1407 {
1408 struct rb_node *node;
1409 struct extent_state *state;
1410
1411 /*
1412 * this search will find all the extents that end after
1413 * our range starts.
1414 */
1415 node = tree_search(tree, start);
1416 if (!node)
1417 goto out;
1418
1419 while (1) {
1420 state = rb_entry(node, struct extent_state, rb_node);
1421 if (state->end >= start && (state->state & bits))
1422 return state;
1423
1424 node = rb_next(node);
1425 if (!node)
1426 break;
1427 }
1428 out:
1429 return NULL;
1430 }
1431
1432 /*
1433 * find the first offset in the io tree with 'bits' set. zero is
1434 * returned if we find something, and *start_ret and *end_ret are
1435 * set to reflect the state struct that was found.
1436 *
1437 * If nothing was found, 1 is returned. If found something, return 0.
1438 */
1439 int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
1440 u64 *start_ret, u64 *end_ret, unsigned bits,
1441 struct extent_state **cached_state)
1442 {
1443 struct extent_state *state;
1444 struct rb_node *n;
1445 int ret = 1;
1446
1447 spin_lock(&tree->lock);
1448 if (cached_state && *cached_state) {
1449 state = *cached_state;
1450 if (state->end == start - 1 && extent_state_in_tree(state)) {
1451 n = rb_next(&state->rb_node);
1452 while (n) {
1453 state = rb_entry(n, struct extent_state,
1454 rb_node);
1455 if (state->state & bits)
1456 goto got_it;
1457 n = rb_next(n);
1458 }
1459 free_extent_state(*cached_state);
1460 *cached_state = NULL;
1461 goto out;
1462 }
1463 free_extent_state(*cached_state);
1464 *cached_state = NULL;
1465 }
1466
1467 state = find_first_extent_bit_state(tree, start, bits);
1468 got_it:
1469 if (state) {
1470 cache_state_if_flags(state, cached_state, 0);
1471 *start_ret = state->start;
1472 *end_ret = state->end;
1473 ret = 0;
1474 }
1475 out:
1476 spin_unlock(&tree->lock);
1477 return ret;
1478 }
1479
1480 /*
1481 * find a contiguous range of bytes in the file marked as delalloc, not
1482 * more than 'max_bytes'. start and end are used to return the range,
1483 *
1484 * 1 is returned if we find something, 0 if nothing was in the tree
1485 */
1486 static noinline u64 find_delalloc_range(struct extent_io_tree *tree,
1487 u64 *start, u64 *end, u64 max_bytes,
1488 struct extent_state **cached_state)
1489 {
1490 struct rb_node *node;
1491 struct extent_state *state;
1492 u64 cur_start = *start;
1493 u64 found = 0;
1494 u64 total_bytes = 0;
1495
1496 spin_lock(&tree->lock);
1497
1498 /*
1499 * this search will find all the extents that end after
1500 * our range starts.
1501 */
1502 node = tree_search(tree, cur_start);
1503 if (!node) {
1504 if (!found)
1505 *end = (u64)-1;
1506 goto out;
1507 }
1508
1509 while (1) {
1510 state = rb_entry(node, struct extent_state, rb_node);
1511 if (found && (state->start != cur_start ||
1512 (state->state & EXTENT_BOUNDARY))) {
1513 goto out;
1514 }
1515 if (!(state->state & EXTENT_DELALLOC)) {
1516 if (!found)
1517 *end = state->end;
1518 goto out;
1519 }
1520 if (!found) {
1521 *start = state->start;
1522 *cached_state = state;
1523 refcount_inc(&state->refs);
1524 }
1525 found++;
1526 *end = state->end;
1527 cur_start = state->end + 1;
1528 node = rb_next(node);
1529 total_bytes += state->end - state->start + 1;
1530 if (total_bytes >= max_bytes)
1531 break;
1532 if (!node)
1533 break;
1534 }
1535 out:
1536 spin_unlock(&tree->lock);
1537 return found;
1538 }
1539
1540 static int __process_pages_contig(struct address_space *mapping,
1541 struct page *locked_page,
1542 pgoff_t start_index, pgoff_t end_index,
1543 unsigned long page_ops, pgoff_t *index_ret);
1544
1545 static noinline void __unlock_for_delalloc(struct inode *inode,
1546 struct page *locked_page,
1547 u64 start, u64 end)
1548 {
1549 unsigned long index = start >> PAGE_SHIFT;
1550 unsigned long end_index = end >> PAGE_SHIFT;
1551
1552 ASSERT(locked_page);
1553 if (index == locked_page->index && end_index == index)
1554 return;
1555
1556 __process_pages_contig(inode->i_mapping, locked_page, index, end_index,
1557 PAGE_UNLOCK, NULL);
1558 }
1559
1560 static noinline int lock_delalloc_pages(struct inode *inode,
1561 struct page *locked_page,
1562 u64 delalloc_start,
1563 u64 delalloc_end)
1564 {
1565 unsigned long index = delalloc_start >> PAGE_SHIFT;
1566 unsigned long index_ret = index;
1567 unsigned long end_index = delalloc_end >> PAGE_SHIFT;
1568 int ret;
1569
1570 ASSERT(locked_page);
1571 if (index == locked_page->index && index == end_index)
1572 return 0;
1573
1574 ret = __process_pages_contig(inode->i_mapping, locked_page, index,
1575 end_index, PAGE_LOCK, &index_ret);
1576 if (ret == -EAGAIN)
1577 __unlock_for_delalloc(inode, locked_page, delalloc_start,
1578 (u64)index_ret << PAGE_SHIFT);
1579 return ret;
1580 }
1581
1582 /*
1583 * find a contiguous range of bytes in the file marked as delalloc, not
1584 * more than 'max_bytes'. start and end are used to return the range,
1585 *
1586 * 1 is returned if we find something, 0 if nothing was in the tree
1587 */
1588 STATIC u64 find_lock_delalloc_range(struct inode *inode,
1589 struct extent_io_tree *tree,
1590 struct page *locked_page, u64 *start,
1591 u64 *end, u64 max_bytes)
1592 {
1593 u64 delalloc_start;
1594 u64 delalloc_end;
1595 u64 found;
1596 struct extent_state *cached_state = NULL;
1597 int ret;
1598 int loops = 0;
1599
1600 again:
1601 /* step one, find a bunch of delalloc bytes starting at start */
1602 delalloc_start = *start;
1603 delalloc_end = 0;
1604 found = find_delalloc_range(tree, &delalloc_start, &delalloc_end,
1605 max_bytes, &cached_state);
1606 if (!found || delalloc_end <= *start) {
1607 *start = delalloc_start;
1608 *end = delalloc_end;
1609 free_extent_state(cached_state);
1610 return 0;
1611 }
1612
1613 /*
1614 * start comes from the offset of locked_page. We have to lock
1615 * pages in order, so we can't process delalloc bytes before
1616 * locked_page
1617 */
1618 if (delalloc_start < *start)
1619 delalloc_start = *start;
1620
1621 /*
1622 * make sure to limit the number of pages we try to lock down
1623 */
1624 if (delalloc_end + 1 - delalloc_start > max_bytes)
1625 delalloc_end = delalloc_start + max_bytes - 1;
1626
1627 /* step two, lock all the pages after the page that has start */
1628 ret = lock_delalloc_pages(inode, locked_page,
1629 delalloc_start, delalloc_end);
1630 if (ret == -EAGAIN) {
1631 /* some of the pages are gone, lets avoid looping by
1632 * shortening the size of the delalloc range we're searching
1633 */
1634 free_extent_state(cached_state);
1635 cached_state = NULL;
1636 if (!loops) {
1637 max_bytes = PAGE_SIZE;
1638 loops = 1;
1639 goto again;
1640 } else {
1641 found = 0;
1642 goto out_failed;
1643 }
1644 }
1645 BUG_ON(ret); /* Only valid values are 0 and -EAGAIN */
1646
1647 /* step three, lock the state bits for the whole range */
1648 lock_extent_bits(tree, delalloc_start, delalloc_end, &cached_state);
1649
1650 /* then test to make sure it is all still delalloc */
1651 ret = test_range_bit(tree, delalloc_start, delalloc_end,
1652 EXTENT_DELALLOC, 1, cached_state);
1653 if (!ret) {
1654 unlock_extent_cached(tree, delalloc_start, delalloc_end,
1655 &cached_state);
1656 __unlock_for_delalloc(inode, locked_page,
1657 delalloc_start, delalloc_end);
1658 cond_resched();
1659 goto again;
1660 }
1661 free_extent_state(cached_state);
1662 *start = delalloc_start;
1663 *end = delalloc_end;
1664 out_failed:
1665 return found;
1666 }
1667
1668 static int __process_pages_contig(struct address_space *mapping,
1669 struct page *locked_page,
1670 pgoff_t start_index, pgoff_t end_index,
1671 unsigned long page_ops, pgoff_t *index_ret)
1672 {
1673 unsigned long nr_pages = end_index - start_index + 1;
1674 unsigned long pages_locked = 0;
1675 pgoff_t index = start_index;
1676 struct page *pages[16];
1677 unsigned ret;
1678 int err = 0;
1679 int i;
1680
1681 if (page_ops & PAGE_LOCK) {
1682 ASSERT(page_ops == PAGE_LOCK);
1683 ASSERT(index_ret && *index_ret == start_index);
1684 }
1685
1686 if ((page_ops & PAGE_SET_ERROR) && nr_pages > 0)
1687 mapping_set_error(mapping, -EIO);
1688
1689 while (nr_pages > 0) {
1690 ret = find_get_pages_contig(mapping, index,
1691 min_t(unsigned long,
1692 nr_pages, ARRAY_SIZE(pages)), pages);
1693 if (ret == 0) {
1694 /*
1695 * Only if we're going to lock these pages,
1696 * can we find nothing at @index.
1697 */
1698 ASSERT(page_ops & PAGE_LOCK);
1699 err = -EAGAIN;
1700 goto out;
1701 }
1702
1703 for (i = 0; i < ret; i++) {
1704 if (page_ops & PAGE_SET_PRIVATE2)
1705 SetPagePrivate2(pages[i]);
1706
1707 if (pages[i] == locked_page) {
1708 put_page(pages[i]);
1709 pages_locked++;
1710 continue;
1711 }
1712 if (page_ops & PAGE_CLEAR_DIRTY)
1713 clear_page_dirty_for_io(pages[i]);
1714 if (page_ops & PAGE_SET_WRITEBACK)
1715 set_page_writeback(pages[i]);
1716 if (page_ops & PAGE_SET_ERROR)
1717 SetPageError(pages[i]);
1718 if (page_ops & PAGE_END_WRITEBACK)
1719 end_page_writeback(pages[i]);
1720 if (page_ops & PAGE_UNLOCK)
1721 unlock_page(pages[i]);
1722 if (page_ops & PAGE_LOCK) {
1723 lock_page(pages[i]);
1724 if (!PageDirty(pages[i]) ||
1725 pages[i]->mapping != mapping) {
1726 unlock_page(pages[i]);
1727 put_page(pages[i]);
1728 err = -EAGAIN;
1729 goto out;
1730 }
1731 }
1732 put_page(pages[i]);
1733 pages_locked++;
1734 }
1735 nr_pages -= ret;
1736 index += ret;
1737 cond_resched();
1738 }
1739 out:
1740 if (err && index_ret)
1741 *index_ret = start_index + pages_locked - 1;
1742 return err;
1743 }
1744
1745 void extent_clear_unlock_delalloc(struct inode *inode, u64 start, u64 end,
1746 u64 delalloc_end, struct page *locked_page,
1747 unsigned clear_bits,
1748 unsigned long page_ops)
1749 {
1750 clear_extent_bit(&BTRFS_I(inode)->io_tree, start, end, clear_bits, 1, 0,
1751 NULL);
1752
1753 __process_pages_contig(inode->i_mapping, locked_page,
1754 start >> PAGE_SHIFT, end >> PAGE_SHIFT,
1755 page_ops, NULL);
1756 }
1757
1758 /*
1759 * count the number of bytes in the tree that have a given bit(s)
1760 * set. This can be fairly slow, except for EXTENT_DIRTY which is
1761 * cached. The total number found is returned.
1762 */
1763 u64 count_range_bits(struct extent_io_tree *tree,
1764 u64 *start, u64 search_end, u64 max_bytes,
1765 unsigned bits, int contig)
1766 {
1767 struct rb_node *node;
1768 struct extent_state *state;
1769 u64 cur_start = *start;
1770 u64 total_bytes = 0;
1771 u64 last = 0;
1772 int found = 0;
1773
1774 if (WARN_ON(search_end <= cur_start))
1775 return 0;
1776
1777 spin_lock(&tree->lock);
1778 if (cur_start == 0 && bits == EXTENT_DIRTY) {
1779 total_bytes = tree->dirty_bytes;
1780 goto out;
1781 }
1782 /*
1783 * this search will find all the extents that end after
1784 * our range starts.
1785 */
1786 node = tree_search(tree, cur_start);
1787 if (!node)
1788 goto out;
1789
1790 while (1) {
1791 state = rb_entry(node, struct extent_state, rb_node);
1792 if (state->start > search_end)
1793 break;
1794 if (contig && found && state->start > last + 1)
1795 break;
1796 if (state->end >= cur_start && (state->state & bits) == bits) {
1797 total_bytes += min(search_end, state->end) + 1 -
1798 max(cur_start, state->start);
1799 if (total_bytes >= max_bytes)
1800 break;
1801 if (!found) {
1802 *start = max(cur_start, state->start);
1803 found = 1;
1804 }
1805 last = state->end;
1806 } else if (contig && found) {
1807 break;
1808 }
1809 node = rb_next(node);
1810 if (!node)
1811 break;
1812 }
1813 out:
1814 spin_unlock(&tree->lock);
1815 return total_bytes;
1816 }
1817
1818 /*
1819 * set the private field for a given byte offset in the tree. If there isn't
1820 * an extent_state there already, this does nothing.
1821 */
1822 static noinline int set_state_failrec(struct extent_io_tree *tree, u64 start,
1823 struct io_failure_record *failrec)
1824 {
1825 struct rb_node *node;
1826 struct extent_state *state;
1827 int ret = 0;
1828
1829 spin_lock(&tree->lock);
1830 /*
1831 * this search will find all the extents that end after
1832 * our range starts.
1833 */
1834 node = tree_search(tree, start);
1835 if (!node) {
1836 ret = -ENOENT;
1837 goto out;
1838 }
1839 state = rb_entry(node, struct extent_state, rb_node);
1840 if (state->start != start) {
1841 ret = -ENOENT;
1842 goto out;
1843 }
1844 state->failrec = failrec;
1845 out:
1846 spin_unlock(&tree->lock);
1847 return ret;
1848 }
1849
1850 static noinline int get_state_failrec(struct extent_io_tree *tree, u64 start,
1851 struct io_failure_record **failrec)
1852 {
1853 struct rb_node *node;
1854 struct extent_state *state;
1855 int ret = 0;
1856
1857 spin_lock(&tree->lock);
1858 /*
1859 * this search will find all the extents that end after
1860 * our range starts.
1861 */
1862 node = tree_search(tree, start);
1863 if (!node) {
1864 ret = -ENOENT;
1865 goto out;
1866 }
1867 state = rb_entry(node, struct extent_state, rb_node);
1868 if (state->start != start) {
1869 ret = -ENOENT;
1870 goto out;
1871 }
1872 *failrec = state->failrec;
1873 out:
1874 spin_unlock(&tree->lock);
1875 return ret;
1876 }
1877
1878 /*
1879 * searches a range in the state tree for a given mask.
1880 * If 'filled' == 1, this returns 1 only if every extent in the tree
1881 * has the bits set. Otherwise, 1 is returned if any bit in the
1882 * range is found set.
1883 */
1884 int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end,
1885 unsigned bits, int filled, struct extent_state *cached)
1886 {
1887 struct extent_state *state = NULL;
1888 struct rb_node *node;
1889 int bitset = 0;
1890
1891 spin_lock(&tree->lock);
1892 if (cached && extent_state_in_tree(cached) && cached->start <= start &&
1893 cached->end > start)
1894 node = &cached->rb_node;
1895 else
1896 node = tree_search(tree, start);
1897 while (node && start <= end) {
1898 state = rb_entry(node, struct extent_state, rb_node);
1899
1900 if (filled && state->start > start) {
1901 bitset = 0;
1902 break;
1903 }
1904
1905 if (state->start > end)
1906 break;
1907
1908 if (state->state & bits) {
1909 bitset = 1;
1910 if (!filled)
1911 break;
1912 } else if (filled) {
1913 bitset = 0;
1914 break;
1915 }
1916
1917 if (state->end == (u64)-1)
1918 break;
1919
1920 start = state->end + 1;
1921 if (start > end)
1922 break;
1923 node = rb_next(node);
1924 if (!node) {
1925 if (filled)
1926 bitset = 0;
1927 break;
1928 }
1929 }
1930 spin_unlock(&tree->lock);
1931 return bitset;
1932 }
1933
1934 /*
1935 * helper function to set a given page up to date if all the
1936 * extents in the tree for that page are up to date
1937 */
1938 static void check_page_uptodate(struct extent_io_tree *tree, struct page *page)
1939 {
1940 u64 start = page_offset(page);
1941 u64 end = start + PAGE_SIZE - 1;
1942 if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1, NULL))
1943 SetPageUptodate(page);
1944 }
1945
1946 int free_io_failure(struct extent_io_tree *failure_tree,
1947 struct extent_io_tree *io_tree,
1948 struct io_failure_record *rec)
1949 {
1950 int ret;
1951 int err = 0;
1952
1953 set_state_failrec(failure_tree, rec->start, NULL);
1954 ret = clear_extent_bits(failure_tree, rec->start,
1955 rec->start + rec->len - 1,
1956 EXTENT_LOCKED | EXTENT_DIRTY);
1957 if (ret)
1958 err = ret;
1959
1960 ret = clear_extent_bits(io_tree, rec->start,
1961 rec->start + rec->len - 1,
1962 EXTENT_DAMAGED);
1963 if (ret && !err)
1964 err = ret;
1965
1966 kfree(rec);
1967 return err;
1968 }
1969
1970 /*
1971 * this bypasses the standard btrfs submit functions deliberately, as
1972 * the standard behavior is to write all copies in a raid setup. here we only
1973 * want to write the one bad copy. so we do the mapping for ourselves and issue
1974 * submit_bio directly.
1975 * to avoid any synchronization issues, wait for the data after writing, which
1976 * actually prevents the read that triggered the error from finishing.
1977 * currently, there can be no more than two copies of every data bit. thus,
1978 * exactly one rewrite is required.
1979 */
1980 int repair_io_failure(struct btrfs_fs_info *fs_info, u64 ino, u64 start,
1981 u64 length, u64 logical, struct page *page,
1982 unsigned int pg_offset, int mirror_num)
1983 {
1984 struct bio *bio;
1985 struct btrfs_device *dev;
1986 u64 map_length = 0;
1987 u64 sector;
1988 struct btrfs_bio *bbio = NULL;
1989 int ret;
1990
1991 ASSERT(!(fs_info->sb->s_flags & SB_RDONLY));
1992 BUG_ON(!mirror_num);
1993
1994 bio = btrfs_io_bio_alloc(1);
1995 bio->bi_iter.bi_size = 0;
1996 map_length = length;
1997
1998 /*
1999 * Avoid races with device replace and make sure our bbio has devices
2000 * associated to its stripes that don't go away while we are doing the
2001 * read repair operation.
2002 */
2003 btrfs_bio_counter_inc_blocked(fs_info);
2004 if (btrfs_is_parity_mirror(fs_info, logical, length)) {
2005 /*
2006 * Note that we don't use BTRFS_MAP_WRITE because it's supposed
2007 * to update all raid stripes, but here we just want to correct
2008 * bad stripe, thus BTRFS_MAP_READ is abused to only get the bad
2009 * stripe's dev and sector.
2010 */
2011 ret = btrfs_map_block(fs_info, BTRFS_MAP_READ, logical,
2012 &map_length, &bbio, 0);
2013 if (ret) {
2014 btrfs_bio_counter_dec(fs_info);
2015 bio_put(bio);
2016 return -EIO;
2017 }
2018 ASSERT(bbio->mirror_num == 1);
2019 } else {
2020 ret = btrfs_map_block(fs_info, BTRFS_MAP_WRITE, logical,
2021 &map_length, &bbio, mirror_num);
2022 if (ret) {
2023 btrfs_bio_counter_dec(fs_info);
2024 bio_put(bio);
2025 return -EIO;
2026 }
2027 BUG_ON(mirror_num != bbio->mirror_num);
2028 }
2029
2030 sector = bbio->stripes[bbio->mirror_num - 1].physical >> 9;
2031 bio->bi_iter.bi_sector = sector;
2032 dev = bbio->stripes[bbio->mirror_num - 1].dev;
2033 btrfs_put_bbio(bbio);
2034 if (!dev || !dev->bdev ||
2035 !test_bit(BTRFS_DEV_STATE_WRITEABLE, &dev->dev_state)) {
2036 btrfs_bio_counter_dec(fs_info);
2037 bio_put(bio);
2038 return -EIO;
2039 }
2040 bio_set_dev(bio, dev->bdev);
2041 bio->bi_opf = REQ_OP_WRITE | REQ_SYNC;
2042 bio_add_page(bio, page, length, pg_offset);
2043
2044 if (btrfsic_submit_bio_wait(bio)) {
2045 /* try to remap that extent elsewhere? */
2046 btrfs_bio_counter_dec(fs_info);
2047 bio_put(bio);
2048 btrfs_dev_stat_inc_and_print(dev, BTRFS_DEV_STAT_WRITE_ERRS);
2049 return -EIO;
2050 }
2051
2052 btrfs_info_rl_in_rcu(fs_info,
2053 "read error corrected: ino %llu off %llu (dev %s sector %llu)",
2054 ino, start,
2055 rcu_str_deref(dev->name), sector);
2056 btrfs_bio_counter_dec(fs_info);
2057 bio_put(bio);
2058 return 0;
2059 }
2060
2061 int repair_eb_io_failure(struct btrfs_fs_info *fs_info,
2062 struct extent_buffer *eb, int mirror_num)
2063 {
2064 u64 start = eb->start;
2065 unsigned long i, num_pages = num_extent_pages(eb->start, eb->len);
2066 int ret = 0;
2067
2068 if (sb_rdonly(fs_info->sb))
2069 return -EROFS;
2070
2071 for (i = 0; i < num_pages; i++) {
2072 struct page *p = eb->pages[i];
2073
2074 ret = repair_io_failure(fs_info, 0, start, PAGE_SIZE, start, p,
2075 start - page_offset(p), mirror_num);
2076 if (ret)
2077 break;
2078 start += PAGE_SIZE;
2079 }
2080
2081 return ret;
2082 }
2083
2084 /*
2085 * each time an IO finishes, we do a fast check in the IO failure tree
2086 * to see if we need to process or clean up an io_failure_record
2087 */
2088 int clean_io_failure(struct btrfs_fs_info *fs_info,
2089 struct extent_io_tree *failure_tree,
2090 struct extent_io_tree *io_tree, u64 start,
2091 struct page *page, u64 ino, unsigned int pg_offset)
2092 {
2093 u64 private;
2094 struct io_failure_record *failrec;
2095 struct extent_state *state;
2096 int num_copies;
2097 int ret;
2098
2099 private = 0;
2100 ret = count_range_bits(failure_tree, &private, (u64)-1, 1,
2101 EXTENT_DIRTY, 0);
2102 if (!ret)
2103 return 0;
2104
2105 ret = get_state_failrec(failure_tree, start, &failrec);
2106 if (ret)
2107 return 0;
2108
2109 BUG_ON(!failrec->this_mirror);
2110
2111 if (failrec->in_validation) {
2112 /* there was no real error, just free the record */
2113 btrfs_debug(fs_info,
2114 "clean_io_failure: freeing dummy error at %llu",
2115 failrec->start);
2116 goto out;
2117 }
2118 if (sb_rdonly(fs_info->sb))
2119 goto out;
2120
2121 spin_lock(&io_tree->lock);
2122 state = find_first_extent_bit_state(io_tree,
2123 failrec->start,
2124 EXTENT_LOCKED);
2125 spin_unlock(&io_tree->lock);
2126
2127 if (state && state->start <= failrec->start &&
2128 state->end >= failrec->start + failrec->len - 1) {
2129 num_copies = btrfs_num_copies(fs_info, failrec->logical,
2130 failrec->len);
2131 if (num_copies > 1) {
2132 repair_io_failure(fs_info, ino, start, failrec->len,
2133 failrec->logical, page, pg_offset,
2134 failrec->failed_mirror);
2135 }
2136 }
2137
2138 out:
2139 free_io_failure(failure_tree, io_tree, failrec);
2140
2141 return 0;
2142 }
2143
2144 /*
2145 * Can be called when
2146 * - hold extent lock
2147 * - under ordered extent
2148 * - the inode is freeing
2149 */
2150 void btrfs_free_io_failure_record(struct btrfs_inode *inode, u64 start, u64 end)
2151 {
2152 struct extent_io_tree *failure_tree = &inode->io_failure_tree;
2153 struct io_failure_record *failrec;
2154 struct extent_state *state, *next;
2155
2156 if (RB_EMPTY_ROOT(&failure_tree->state))
2157 return;
2158
2159 spin_lock(&failure_tree->lock);
2160 state = find_first_extent_bit_state(failure_tree, start, EXTENT_DIRTY);
2161 while (state) {
2162 if (state->start > end)
2163 break;
2164
2165 ASSERT(state->end <= end);
2166
2167 next = next_state(state);
2168
2169 failrec = state->failrec;
2170 free_extent_state(state);
2171 kfree(failrec);
2172
2173 state = next;
2174 }
2175 spin_unlock(&failure_tree->lock);
2176 }
2177
2178 int btrfs_get_io_failure_record(struct inode *inode, u64 start, u64 end,
2179 struct io_failure_record **failrec_ret)
2180 {
2181 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2182 struct io_failure_record *failrec;
2183 struct extent_map *em;
2184 struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
2185 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2186 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
2187 int ret;
2188 u64 logical;
2189
2190 ret = get_state_failrec(failure_tree, start, &failrec);
2191 if (ret) {
2192 failrec = kzalloc(sizeof(*failrec), GFP_NOFS);
2193 if (!failrec)
2194 return -ENOMEM;
2195
2196 failrec->start = start;
2197 failrec->len = end - start + 1;
2198 failrec->this_mirror = 0;
2199 failrec->bio_flags = 0;
2200 failrec->in_validation = 0;
2201
2202 read_lock(&em_tree->lock);
2203 em = lookup_extent_mapping(em_tree, start, failrec->len);
2204 if (!em) {
2205 read_unlock(&em_tree->lock);
2206 kfree(failrec);
2207 return -EIO;
2208 }
2209
2210 if (em->start > start || em->start + em->len <= start) {
2211 free_extent_map(em);
2212 em = NULL;
2213 }
2214 read_unlock(&em_tree->lock);
2215 if (!em) {
2216 kfree(failrec);
2217 return -EIO;
2218 }
2219
2220 logical = start - em->start;
2221 logical = em->block_start + logical;
2222 if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
2223 logical = em->block_start;
2224 failrec->bio_flags = EXTENT_BIO_COMPRESSED;
2225 extent_set_compress_type(&failrec->bio_flags,
2226 em->compress_type);
2227 }
2228
2229 btrfs_debug(fs_info,
2230 "Get IO Failure Record: (new) logical=%llu, start=%llu, len=%llu",
2231 logical, start, failrec->len);
2232
2233 failrec->logical = logical;
2234 free_extent_map(em);
2235
2236 /* set the bits in the private failure tree */
2237 ret = set_extent_bits(failure_tree, start, end,
2238 EXTENT_LOCKED | EXTENT_DIRTY);
2239 if (ret >= 0)
2240 ret = set_state_failrec(failure_tree, start, failrec);
2241 /* set the bits in the inode's tree */
2242 if (ret >= 0)
2243 ret = set_extent_bits(tree, start, end, EXTENT_DAMAGED);
2244 if (ret < 0) {
2245 kfree(failrec);
2246 return ret;
2247 }
2248 } else {
2249 btrfs_debug(fs_info,
2250 "Get IO Failure Record: (found) logical=%llu, start=%llu, len=%llu, validation=%d",
2251 failrec->logical, failrec->start, failrec->len,
2252 failrec->in_validation);
2253 /*
2254 * when data can be on disk more than twice, add to failrec here
2255 * (e.g. with a list for failed_mirror) to make
2256 * clean_io_failure() clean all those errors at once.
2257 */
2258 }
2259
2260 *failrec_ret = failrec;
2261
2262 return 0;
2263 }
2264
2265 bool btrfs_check_repairable(struct inode *inode, unsigned failed_bio_pages,
2266 struct io_failure_record *failrec, int failed_mirror)
2267 {
2268 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2269 int num_copies;
2270
2271 num_copies = btrfs_num_copies(fs_info, failrec->logical, failrec->len);
2272 if (num_copies == 1) {
2273 /*
2274 * we only have a single copy of the data, so don't bother with
2275 * all the retry and error correction code that follows. no
2276 * matter what the error is, it is very likely to persist.
2277 */
2278 btrfs_debug(fs_info,
2279 "Check Repairable: cannot repair, num_copies=%d, next_mirror %d, failed_mirror %d",
2280 num_copies, failrec->this_mirror, failed_mirror);
2281 return false;
2282 }
2283
2284 /*
2285 * there are two premises:
2286 * a) deliver good data to the caller
2287 * b) correct the bad sectors on disk
2288 */
2289 if (failed_bio_pages > 1) {
2290 /*
2291 * to fulfill b), we need to know the exact failing sectors, as
2292 * we don't want to rewrite any more than the failed ones. thus,
2293 * we need separate read requests for the failed bio
2294 *
2295 * if the following BUG_ON triggers, our validation request got
2296 * merged. we need separate requests for our algorithm to work.
2297 */
2298 BUG_ON(failrec->in_validation);
2299 failrec->in_validation = 1;
2300 failrec->this_mirror = failed_mirror;
2301 } else {
2302 /*
2303 * we're ready to fulfill a) and b) alongside. get a good copy
2304 * of the failed sector and if we succeed, we have setup
2305 * everything for repair_io_failure to do the rest for us.
2306 */
2307 if (failrec->in_validation) {
2308 BUG_ON(failrec->this_mirror != failed_mirror);
2309 failrec->in_validation = 0;
2310 failrec->this_mirror = 0;
2311 }
2312 failrec->failed_mirror = failed_mirror;
2313 failrec->this_mirror++;
2314 if (failrec->this_mirror == failed_mirror)
2315 failrec->this_mirror++;
2316 }
2317
2318 if (failrec->this_mirror > num_copies) {
2319 btrfs_debug(fs_info,
2320 "Check Repairable: (fail) num_copies=%d, next_mirror %d, failed_mirror %d",
2321 num_copies, failrec->this_mirror, failed_mirror);
2322 return false;
2323 }
2324
2325 return true;
2326 }
2327
2328
2329 struct bio *btrfs_create_repair_bio(struct inode *inode, struct bio *failed_bio,
2330 struct io_failure_record *failrec,
2331 struct page *page, int pg_offset, int icsum,
2332 bio_end_io_t *endio_func, void *data)
2333 {
2334 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2335 struct bio *bio;
2336 struct btrfs_io_bio *btrfs_failed_bio;
2337 struct btrfs_io_bio *btrfs_bio;
2338
2339 bio = btrfs_io_bio_alloc(1);
2340 bio->bi_end_io = endio_func;
2341 bio->bi_iter.bi_sector = failrec->logical >> 9;
2342 bio_set_dev(bio, fs_info->fs_devices->latest_bdev);
2343 bio->bi_iter.bi_size = 0;
2344 bio->bi_private = data;
2345
2346 btrfs_failed_bio = btrfs_io_bio(failed_bio);
2347 if (btrfs_failed_bio->csum) {
2348 u16 csum_size = btrfs_super_csum_size(fs_info->super_copy);
2349
2350 btrfs_bio = btrfs_io_bio(bio);
2351 btrfs_bio->csum = btrfs_bio->csum_inline;
2352 icsum *= csum_size;
2353 memcpy(btrfs_bio->csum, btrfs_failed_bio->csum + icsum,
2354 csum_size);
2355 }
2356
2357 bio_add_page(bio, page, failrec->len, pg_offset);
2358
2359 return bio;
2360 }
2361
2362 /*
2363 * this is a generic handler for readpage errors (default
2364 * readpage_io_failed_hook). if other copies exist, read those and write back
2365 * good data to the failed position. does not investigate in remapping the
2366 * failed extent elsewhere, hoping the device will be smart enough to do this as
2367 * needed
2368 */
2369
2370 static int bio_readpage_error(struct bio *failed_bio, u64 phy_offset,
2371 struct page *page, u64 start, u64 end,
2372 int failed_mirror)
2373 {
2374 struct io_failure_record *failrec;
2375 struct inode *inode = page->mapping->host;
2376 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2377 struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
2378 struct bio *bio;
2379 int read_mode = 0;
2380 blk_status_t status;
2381 int ret;
2382 unsigned failed_bio_pages = bio_pages_all(failed_bio);
2383
2384 BUG_ON(bio_op(failed_bio) == REQ_OP_WRITE);
2385
2386 ret = btrfs_get_io_failure_record(inode, start, end, &failrec);
2387 if (ret)
2388 return ret;
2389
2390 if (!btrfs_check_repairable(inode, failed_bio_pages, failrec,
2391 failed_mirror)) {
2392 free_io_failure(failure_tree, tree, failrec);
2393 return -EIO;
2394 }
2395
2396 if (failed_bio_pages > 1)
2397 read_mode |= REQ_FAILFAST_DEV;
2398
2399 phy_offset >>= inode->i_sb->s_blocksize_bits;
2400 bio = btrfs_create_repair_bio(inode, failed_bio, failrec, page,
2401 start - page_offset(page),
2402 (int)phy_offset, failed_bio->bi_end_io,
2403 NULL);
2404 bio_set_op_attrs(bio, REQ_OP_READ, read_mode);
2405
2406 btrfs_debug(btrfs_sb(inode->i_sb),
2407 "Repair Read Error: submitting new read[%#x] to this_mirror=%d, in_validation=%d",
2408 read_mode, failrec->this_mirror, failrec->in_validation);
2409
2410 status = tree->ops->submit_bio_hook(tree->private_data, bio, failrec->this_mirror,
2411 failrec->bio_flags, 0);
2412 if (status) {
2413 free_io_failure(failure_tree, tree, failrec);
2414 bio_put(bio);
2415 ret = blk_status_to_errno(status);
2416 }
2417
2418 return ret;
2419 }
2420
2421 /* lots and lots of room for performance fixes in the end_bio funcs */
2422
2423 void end_extent_writepage(struct page *page, int err, u64 start, u64 end)
2424 {
2425 int uptodate = (err == 0);
2426 struct extent_io_tree *tree;
2427 int ret = 0;
2428
2429 tree = &BTRFS_I(page->mapping->host)->io_tree;
2430
2431 if (tree->ops && tree->ops->writepage_end_io_hook)
2432 tree->ops->writepage_end_io_hook(page, start, end, NULL,
2433 uptodate);
2434
2435 if (!uptodate) {
2436 ClearPageUptodate(page);
2437 SetPageError(page);
2438 ret = err < 0 ? err : -EIO;
2439 mapping_set_error(page->mapping, ret);
2440 }
2441 }
2442
2443 /*
2444 * after a writepage IO is done, we need to:
2445 * clear the uptodate bits on error
2446 * clear the writeback bits in the extent tree for this IO
2447 * end_page_writeback if the page has no more pending IO
2448 *
2449 * Scheduling is not allowed, so the extent state tree is expected
2450 * to have one and only one object corresponding to this IO.
2451 */
2452 static void end_bio_extent_writepage(struct bio *bio)
2453 {
2454 int error = blk_status_to_errno(bio->bi_status);
2455 struct bio_vec *bvec;
2456 u64 start;
2457 u64 end;
2458 int i;
2459
2460 ASSERT(!bio_flagged(bio, BIO_CLONED));
2461 bio_for_each_segment_all(bvec, bio, i) {
2462 struct page *page = bvec->bv_page;
2463 struct inode *inode = page->mapping->host;
2464 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2465
2466 /* We always issue full-page reads, but if some block
2467 * in a page fails to read, blk_update_request() will
2468 * advance bv_offset and adjust bv_len to compensate.
2469 * Print a warning for nonzero offsets, and an error
2470 * if they don't add up to a full page. */
2471 if (bvec->bv_offset || bvec->bv_len != PAGE_SIZE) {
2472 if (bvec->bv_offset + bvec->bv_len != PAGE_SIZE)
2473 btrfs_err(fs_info,
2474 "partial page write in btrfs with offset %u and length %u",
2475 bvec->bv_offset, bvec->bv_len);
2476 else
2477 btrfs_info(fs_info,
2478 "incomplete page write in btrfs with offset %u and length %u",
2479 bvec->bv_offset, bvec->bv_len);
2480 }
2481
2482 start = page_offset(page);
2483 end = start + bvec->bv_offset + bvec->bv_len - 1;
2484
2485 end_extent_writepage(page, error, start, end);
2486 end_page_writeback(page);
2487 }
2488
2489 bio_put(bio);
2490 }
2491
2492 static void
2493 endio_readpage_release_extent(struct extent_io_tree *tree, u64 start, u64 len,
2494 int uptodate)
2495 {
2496 struct extent_state *cached = NULL;
2497 u64 end = start + len - 1;
2498
2499 if (uptodate && tree->track_uptodate)
2500 set_extent_uptodate(tree, start, end, &cached, GFP_ATOMIC);
2501 unlock_extent_cached_atomic(tree, start, end, &cached);
2502 }
2503
2504 /*
2505 * after a readpage IO is done, we need to:
2506 * clear the uptodate bits on error
2507 * set the uptodate bits if things worked
2508 * set the page up to date if all extents in the tree are uptodate
2509 * clear the lock bit in the extent tree
2510 * unlock the page if there are no other extents locked for it
2511 *
2512 * Scheduling is not allowed, so the extent state tree is expected
2513 * to have one and only one object corresponding to this IO.
2514 */
2515 static void end_bio_extent_readpage(struct bio *bio)
2516 {
2517 struct bio_vec *bvec;
2518 int uptodate = !bio->bi_status;
2519 struct btrfs_io_bio *io_bio = btrfs_io_bio(bio);
2520 struct extent_io_tree *tree, *failure_tree;
2521 u64 offset = 0;
2522 u64 start;
2523 u64 end;
2524 u64 len;
2525 u64 extent_start = 0;
2526 u64 extent_len = 0;
2527 int mirror;
2528 int ret;
2529 int i;
2530
2531 ASSERT(!bio_flagged(bio, BIO_CLONED));
2532 bio_for_each_segment_all(bvec, bio, i) {
2533 struct page *page = bvec->bv_page;
2534 struct inode *inode = page->mapping->host;
2535 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2536
2537 btrfs_debug(fs_info,
2538 "end_bio_extent_readpage: bi_sector=%llu, err=%d, mirror=%u",
2539 (u64)bio->bi_iter.bi_sector, bio->bi_status,
2540 io_bio->mirror_num);
2541 tree = &BTRFS_I(inode)->io_tree;
2542 failure_tree = &BTRFS_I(inode)->io_failure_tree;
2543
2544 /* We always issue full-page reads, but if some block
2545 * in a page fails to read, blk_update_request() will
2546 * advance bv_offset and adjust bv_len to compensate.
2547 * Print a warning for nonzero offsets, and an error
2548 * if they don't add up to a full page. */
2549 if (bvec->bv_offset || bvec->bv_len != PAGE_SIZE) {
2550 if (bvec->bv_offset + bvec->bv_len != PAGE_SIZE)
2551 btrfs_err(fs_info,
2552 "partial page read in btrfs with offset %u and length %u",
2553 bvec->bv_offset, bvec->bv_len);
2554 else
2555 btrfs_info(fs_info,
2556 "incomplete page read in btrfs with offset %u and length %u",
2557 bvec->bv_offset, bvec->bv_len);
2558 }
2559
2560 start = page_offset(page);
2561 end = start + bvec->bv_offset + bvec->bv_len - 1;
2562 len = bvec->bv_len;
2563
2564 mirror = io_bio->mirror_num;
2565 if (likely(uptodate && tree->ops)) {
2566 ret = tree->ops->readpage_end_io_hook(io_bio, offset,
2567 page, start, end,
2568 mirror);
2569 if (ret)
2570 uptodate = 0;
2571 else
2572 clean_io_failure(BTRFS_I(inode)->root->fs_info,
2573 failure_tree, tree, start,
2574 page,
2575 btrfs_ino(BTRFS_I(inode)), 0);
2576 }
2577
2578 if (likely(uptodate))
2579 goto readpage_ok;
2580
2581 if (tree->ops) {
2582 ret = tree->ops->readpage_io_failed_hook(page, mirror);
2583 if (ret == -EAGAIN) {
2584 /*
2585 * Data inode's readpage_io_failed_hook() always
2586 * returns -EAGAIN.
2587 *
2588 * The generic bio_readpage_error handles errors
2589 * the following way: If possible, new read
2590 * requests are created and submitted and will
2591 * end up in end_bio_extent_readpage as well (if
2592 * we're lucky, not in the !uptodate case). In
2593 * that case it returns 0 and we just go on with
2594 * the next page in our bio. If it can't handle
2595 * the error it will return -EIO and we remain
2596 * responsible for that page.
2597 */
2598 ret = bio_readpage_error(bio, offset, page,
2599 start, end, mirror);
2600 if (ret == 0) {
2601 uptodate = !bio->bi_status;
2602 offset += len;
2603 continue;
2604 }
2605 }
2606
2607 /*
2608 * metadata's readpage_io_failed_hook() always returns
2609 * -EIO and fixes nothing. -EIO is also returned if
2610 * data inode error could not be fixed.
2611 */
2612 ASSERT(ret == -EIO);
2613 }
2614 readpage_ok:
2615 if (likely(uptodate)) {
2616 loff_t i_size = i_size_read(inode);
2617 pgoff_t end_index = i_size >> PAGE_SHIFT;
2618 unsigned off;
2619
2620 /* Zero out the end if this page straddles i_size */
2621 off = i_size & (PAGE_SIZE-1);
2622 if (page->index == end_index && off)
2623 zero_user_segment(page, off, PAGE_SIZE);
2624 SetPageUptodate(page);
2625 } else {
2626 ClearPageUptodate(page);
2627 SetPageError(page);
2628 }
2629 unlock_page(page);
2630 offset += len;
2631
2632 if (unlikely(!uptodate)) {
2633 if (extent_len) {
2634 endio_readpage_release_extent(tree,
2635 extent_start,
2636 extent_len, 1);
2637 extent_start = 0;
2638 extent_len = 0;
2639 }
2640 endio_readpage_release_extent(tree, start,
2641 end - start + 1, 0);
2642 } else if (!extent_len) {
2643 extent_start = start;
2644 extent_len = end + 1 - start;
2645 } else if (extent_start + extent_len == start) {
2646 extent_len += end + 1 - start;
2647 } else {
2648 endio_readpage_release_extent(tree, extent_start,
2649 extent_len, uptodate);
2650 extent_start = start;
2651 extent_len = end + 1 - start;
2652 }
2653 }
2654
2655 if (extent_len)
2656 endio_readpage_release_extent(tree, extent_start, extent_len,
2657 uptodate);
2658 if (io_bio->end_io)
2659 io_bio->end_io(io_bio, blk_status_to_errno(bio->bi_status));
2660 bio_put(bio);
2661 }
2662
2663 /*
2664 * Initialize the members up to but not including 'bio'. Use after allocating a
2665 * new bio by bio_alloc_bioset as it does not initialize the bytes outside of
2666 * 'bio' because use of __GFP_ZERO is not supported.
2667 */
2668 static inline void btrfs_io_bio_init(struct btrfs_io_bio *btrfs_bio)
2669 {
2670 memset(btrfs_bio, 0, offsetof(struct btrfs_io_bio, bio));
2671 }
2672
2673 /*
2674 * The following helpers allocate a bio. As it's backed by a bioset, it'll
2675 * never fail. We're returning a bio right now but you can call btrfs_io_bio
2676 * for the appropriate container_of magic
2677 */
2678 struct bio *btrfs_bio_alloc(struct block_device *bdev, u64 first_byte)
2679 {
2680 struct bio *bio;
2681
2682 bio = bio_alloc_bioset(GFP_NOFS, BIO_MAX_PAGES, btrfs_bioset);
2683 bio_set_dev(bio, bdev);
2684 bio->bi_iter.bi_sector = first_byte >> 9;
2685 btrfs_io_bio_init(btrfs_io_bio(bio));
2686 return bio;
2687 }
2688
2689 struct bio *btrfs_bio_clone(struct bio *bio)
2690 {
2691 struct btrfs_io_bio *btrfs_bio;
2692 struct bio *new;
2693
2694 /* Bio allocation backed by a bioset does not fail */
2695 new = bio_clone_fast(bio, GFP_NOFS, btrfs_bioset);
2696 btrfs_bio = btrfs_io_bio(new);
2697 btrfs_io_bio_init(btrfs_bio);
2698 btrfs_bio->iter = bio->bi_iter;
2699 return new;
2700 }
2701
2702 struct bio *btrfs_io_bio_alloc(unsigned int nr_iovecs)
2703 {
2704 struct bio *bio;
2705
2706 /* Bio allocation backed by a bioset does not fail */
2707 bio = bio_alloc_bioset(GFP_NOFS, nr_iovecs, btrfs_bioset);
2708 btrfs_io_bio_init(btrfs_io_bio(bio));
2709 return bio;
2710 }
2711
2712 struct bio *btrfs_bio_clone_partial(struct bio *orig, int offset, int size)
2713 {
2714 struct bio *bio;
2715 struct btrfs_io_bio *btrfs_bio;
2716
2717 /* this will never fail when it's backed by a bioset */
2718 bio = bio_clone_fast(orig, GFP_NOFS, btrfs_bioset);
2719 ASSERT(bio);
2720
2721 btrfs_bio = btrfs_io_bio(bio);
2722 btrfs_io_bio_init(btrfs_bio);
2723
2724 bio_trim(bio, offset >> 9, size >> 9);
2725 btrfs_bio->iter = bio->bi_iter;
2726 return bio;
2727 }
2728
2729 static int __must_check submit_one_bio(struct bio *bio, int mirror_num,
2730 unsigned long bio_flags)
2731 {
2732 blk_status_t ret = 0;
2733 struct bio_vec *bvec = bio_last_bvec_all(bio);
2734 struct page *page = bvec->bv_page;
2735 struct extent_io_tree *tree = bio->bi_private;
2736 u64 start;
2737
2738 start = page_offset(page) + bvec->bv_offset;
2739
2740 bio->bi_private = NULL;
2741
2742 if (tree->ops)
2743 ret = tree->ops->submit_bio_hook(tree->private_data, bio,
2744 mirror_num, bio_flags, start);
2745 else
2746 btrfsic_submit_bio(bio);
2747
2748 return blk_status_to_errno(ret);
2749 }
2750
2751 /*
2752 * @opf: bio REQ_OP_* and REQ_* flags as one value
2753 * @tree: tree so we can call our merge_bio hook
2754 * @wbc: optional writeback control for io accounting
2755 * @page: page to add to the bio
2756 * @pg_offset: offset of the new bio or to check whether we are adding
2757 * a contiguous page to the previous one
2758 * @size: portion of page that we want to write
2759 * @offset: starting offset in the page
2760 * @bdev: attach newly created bios to this bdev
2761 * @bio_ret: must be valid pointer, newly allocated bio will be stored there
2762 * @end_io_func: end_io callback for new bio
2763 * @mirror_num: desired mirror to read/write
2764 * @prev_bio_flags: flags of previous bio to see if we can merge the current one
2765 * @bio_flags: flags of the current bio to see if we can merge them
2766 */
2767 static int submit_extent_page(unsigned int opf, struct extent_io_tree *tree,
2768 struct writeback_control *wbc,
2769 struct page *page, u64 offset,
2770 size_t size, unsigned long pg_offset,
2771 struct block_device *bdev,
2772 struct bio **bio_ret,
2773 bio_end_io_t end_io_func,
2774 int mirror_num,
2775 unsigned long prev_bio_flags,
2776 unsigned long bio_flags,
2777 bool force_bio_submit)
2778 {
2779 int ret = 0;
2780 struct bio *bio;
2781 size_t page_size = min_t(size_t, size, PAGE_SIZE);
2782 sector_t sector = offset >> 9;
2783
2784 ASSERT(bio_ret);
2785
2786 if (*bio_ret) {
2787 bool contig;
2788 bool can_merge = true;
2789
2790 bio = *bio_ret;
2791 if (prev_bio_flags & EXTENT_BIO_COMPRESSED)
2792 contig = bio->bi_iter.bi_sector == sector;
2793 else
2794 contig = bio_end_sector(bio) == sector;
2795
2796 if (tree->ops && tree->ops->merge_bio_hook(page, offset,
2797 page_size, bio, bio_flags))
2798 can_merge = false;
2799
2800 if (prev_bio_flags != bio_flags || !contig || !can_merge ||
2801 force_bio_submit ||
2802 bio_add_page(bio, page, page_size, pg_offset) < page_size) {
2803 ret = submit_one_bio(bio, mirror_num, prev_bio_flags);
2804 if (ret < 0) {
2805 *bio_ret = NULL;
2806 return ret;
2807 }
2808 bio = NULL;
2809 } else {
2810 if (wbc)
2811 wbc_account_io(wbc, page, page_size);
2812 return 0;
2813 }
2814 }
2815
2816 bio = btrfs_bio_alloc(bdev, offset);
2817 bio_add_page(bio, page, page_size, pg_offset);
2818 bio->bi_end_io = end_io_func;
2819 bio->bi_private = tree;
2820 bio->bi_write_hint = page->mapping->host->i_write_hint;
2821 bio->bi_opf = opf;
2822 if (wbc) {
2823 wbc_init_bio(wbc, bio);
2824 wbc_account_io(wbc, page, page_size);
2825 }
2826
2827 *bio_ret = bio;
2828
2829 return ret;
2830 }
2831
2832 static void attach_extent_buffer_page(struct extent_buffer *eb,
2833 struct page *page)
2834 {
2835 if (!PagePrivate(page)) {
2836 SetPagePrivate(page);
2837 get_page(page);
2838 set_page_private(page, (unsigned long)eb);
2839 } else {
2840 WARN_ON(page->private != (unsigned long)eb);
2841 }
2842 }
2843
2844 void set_page_extent_mapped(struct page *page)
2845 {
2846 if (!PagePrivate(page)) {
2847 SetPagePrivate(page);
2848 get_page(page);
2849 set_page_private(page, EXTENT_PAGE_PRIVATE);
2850 }
2851 }
2852
2853 static struct extent_map *
2854 __get_extent_map(struct inode *inode, struct page *page, size_t pg_offset,
2855 u64 start, u64 len, get_extent_t *get_extent,
2856 struct extent_map **em_cached)
2857 {
2858 struct extent_map *em;
2859
2860 if (em_cached && *em_cached) {
2861 em = *em_cached;
2862 if (extent_map_in_tree(em) && start >= em->start &&
2863 start < extent_map_end(em)) {
2864 refcount_inc(&em->refs);
2865 return em;
2866 }
2867
2868 free_extent_map(em);
2869 *em_cached = NULL;
2870 }
2871
2872 em = get_extent(BTRFS_I(inode), page, pg_offset, start, len, 0);
2873 if (em_cached && !IS_ERR_OR_NULL(em)) {
2874 BUG_ON(*em_cached);
2875 refcount_inc(&em->refs);
2876 *em_cached = em;
2877 }
2878 return em;
2879 }
2880 /*
2881 * basic readpage implementation. Locked extent state structs are inserted
2882 * into the tree that are removed when the IO is done (by the end_io
2883 * handlers)
2884 * XXX JDM: This needs looking at to ensure proper page locking
2885 * return 0 on success, otherwise return error
2886 */
2887 static int __do_readpage(struct extent_io_tree *tree,
2888 struct page *page,
2889 get_extent_t *get_extent,
2890 struct extent_map **em_cached,
2891 struct bio **bio, int mirror_num,
2892 unsigned long *bio_flags, unsigned int read_flags,
2893 u64 *prev_em_start)
2894 {
2895 struct inode *inode = page->mapping->host;
2896 u64 start = page_offset(page);
2897 const u64 end = start + PAGE_SIZE - 1;
2898 u64 cur = start;
2899 u64 extent_offset;
2900 u64 last_byte = i_size_read(inode);
2901 u64 block_start;
2902 u64 cur_end;
2903 struct extent_map *em;
2904 struct block_device *bdev;
2905 int ret = 0;
2906 int nr = 0;
2907 size_t pg_offset = 0;
2908 size_t iosize;
2909 size_t disk_io_size;
2910 size_t blocksize = inode->i_sb->s_blocksize;
2911 unsigned long this_bio_flag = 0;
2912
2913 set_page_extent_mapped(page);
2914
2915 if (!PageUptodate(page)) {
2916 if (cleancache_get_page(page) == 0) {
2917 BUG_ON(blocksize != PAGE_SIZE);
2918 unlock_extent(tree, start, end);
2919 goto out;
2920 }
2921 }
2922
2923 if (page->index == last_byte >> PAGE_SHIFT) {
2924 char *userpage;
2925 size_t zero_offset = last_byte & (PAGE_SIZE - 1);
2926
2927 if (zero_offset) {
2928 iosize = PAGE_SIZE - zero_offset;
2929 userpage = kmap_atomic(page);
2930 memset(userpage + zero_offset, 0, iosize);
2931 flush_dcache_page(page);
2932 kunmap_atomic(userpage);
2933 }
2934 }
2935 while (cur <= end) {
2936 bool force_bio_submit = false;
2937 u64 offset;
2938
2939 if (cur >= last_byte) {
2940 char *userpage;
2941 struct extent_state *cached = NULL;
2942
2943 iosize = PAGE_SIZE - pg_offset;
2944 userpage = kmap_atomic(page);
2945 memset(userpage + pg_offset, 0, iosize);
2946 flush_dcache_page(page);
2947 kunmap_atomic(userpage);
2948 set_extent_uptodate(tree, cur, cur + iosize - 1,
2949 &cached, GFP_NOFS);
2950 unlock_extent_cached(tree, cur,
2951 cur + iosize - 1, &cached);
2952 break;
2953 }
2954 em = __get_extent_map(inode, page, pg_offset, cur,
2955 end - cur + 1, get_extent, em_cached);
2956 if (IS_ERR_OR_NULL(em)) {
2957 SetPageError(page);
2958 unlock_extent(tree, cur, end);
2959 break;
2960 }
2961 extent_offset = cur - em->start;
2962 BUG_ON(extent_map_end(em) <= cur);
2963 BUG_ON(end < cur);
2964
2965 if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
2966 this_bio_flag |= EXTENT_BIO_COMPRESSED;
2967 extent_set_compress_type(&this_bio_flag,
2968 em->compress_type);
2969 }
2970
2971 iosize = min(extent_map_end(em) - cur, end - cur + 1);
2972 cur_end = min(extent_map_end(em) - 1, end);
2973 iosize = ALIGN(iosize, blocksize);
2974 if (this_bio_flag & EXTENT_BIO_COMPRESSED) {
2975 disk_io_size = em->block_len;
2976 offset = em->block_start;
2977 } else {
2978 offset = em->block_start + extent_offset;
2979 disk_io_size = iosize;
2980 }
2981 bdev = em->bdev;
2982 block_start = em->block_start;
2983 if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
2984 block_start = EXTENT_MAP_HOLE;
2985
2986 /*
2987 * If we have a file range that points to a compressed extent
2988 * and it's followed by a consecutive file range that points to
2989 * to the same compressed extent (possibly with a different
2990 * offset and/or length, so it either points to the whole extent
2991 * or only part of it), we must make sure we do not submit a
2992 * single bio to populate the pages for the 2 ranges because
2993 * this makes the compressed extent read zero out the pages
2994 * belonging to the 2nd range. Imagine the following scenario:
2995 *
2996 * File layout
2997 * [0 - 8K] [8K - 24K]
2998 * | |
2999 * | |
3000 * points to extent X, points to extent X,
3001 * offset 4K, length of 8K offset 0, length 16K
3002 *
3003 * [extent X, compressed length = 4K uncompressed length = 16K]
3004 *
3005 * If the bio to read the compressed extent covers both ranges,
3006 * it will decompress extent X into the pages belonging to the
3007 * first range and then it will stop, zeroing out the remaining
3008 * pages that belong to the other range that points to extent X.
3009 * So here we make sure we submit 2 bios, one for the first
3010 * range and another one for the third range. Both will target
3011 * the same physical extent from disk, but we can't currently
3012 * make the compressed bio endio callback populate the pages
3013 * for both ranges because each compressed bio is tightly
3014 * coupled with a single extent map, and each range can have
3015 * an extent map with a different offset value relative to the
3016 * uncompressed data of our extent and different lengths. This
3017 * is a corner case so we prioritize correctness over
3018 * non-optimal behavior (submitting 2 bios for the same extent).
3019 */
3020 if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags) &&
3021 prev_em_start && *prev_em_start != (u64)-1 &&
3022 *prev_em_start != em->orig_start)
3023 force_bio_submit = true;
3024
3025 if (prev_em_start)
3026 *prev_em_start = em->orig_start;
3027
3028 free_extent_map(em);
3029 em = NULL;
3030
3031 /* we've found a hole, just zero and go on */
3032 if (block_start == EXTENT_MAP_HOLE) {
3033 char *userpage;
3034 struct extent_state *cached = NULL;
3035
3036 userpage = kmap_atomic(page);
3037 memset(userpage + pg_offset, 0, iosize);
3038 flush_dcache_page(page);
3039 kunmap_atomic(userpage);
3040
3041 set_extent_uptodate(tree, cur, cur + iosize - 1,
3042 &cached, GFP_NOFS);
3043 unlock_extent_cached(tree, cur,
3044 cur + iosize - 1, &cached);
3045 cur = cur + iosize;
3046 pg_offset += iosize;
3047 continue;
3048 }
3049 /* the get_extent function already copied into the page */
3050 if (test_range_bit(tree, cur, cur_end,
3051 EXTENT_UPTODATE, 1, NULL)) {
3052 check_page_uptodate(tree, page);
3053 unlock_extent(tree, cur, cur + iosize - 1);
3054 cur = cur + iosize;
3055 pg_offset += iosize;
3056 continue;
3057 }
3058 /* we have an inline extent but it didn't get marked up
3059 * to date. Error out
3060 */
3061 if (block_start == EXTENT_MAP_INLINE) {
3062 SetPageError(page);
3063 unlock_extent(tree, cur, cur + iosize - 1);
3064 cur = cur + iosize;
3065 pg_offset += iosize;
3066 continue;
3067 }
3068
3069 ret = submit_extent_page(REQ_OP_READ | read_flags, tree, NULL,
3070 page, offset, disk_io_size,
3071 pg_offset, bdev, bio,
3072 end_bio_extent_readpage, mirror_num,
3073 *bio_flags,
3074 this_bio_flag,
3075 force_bio_submit);
3076 if (!ret) {
3077 nr++;
3078 *bio_flags = this_bio_flag;
3079 } else {
3080 SetPageError(page);
3081 unlock_extent(tree, cur, cur + iosize - 1);
3082 goto out;
3083 }
3084 cur = cur + iosize;
3085 pg_offset += iosize;
3086 }
3087 out:
3088 if (!nr) {
3089 if (!PageError(page))
3090 SetPageUptodate(page);
3091 unlock_page(page);
3092 }
3093 return ret;
3094 }
3095
3096 static inline void __do_contiguous_readpages(struct extent_io_tree *tree,
3097 struct page *pages[], int nr_pages,
3098 u64 start, u64 end,
3099 struct extent_map **em_cached,
3100 struct bio **bio,
3101 unsigned long *bio_flags,
3102 u64 *prev_em_start)
3103 {
3104 struct inode *inode;
3105 struct btrfs_ordered_extent *ordered;
3106 int index;
3107
3108 inode = pages[0]->mapping->host;
3109 while (1) {
3110 lock_extent(tree, start, end);
3111 ordered = btrfs_lookup_ordered_range(BTRFS_I(inode), start,
3112 end - start + 1);
3113 if (!ordered)
3114 break;
3115 unlock_extent(tree, start, end);
3116 btrfs_start_ordered_extent(inode, ordered, 1);
3117 btrfs_put_ordered_extent(ordered);
3118 }
3119
3120 for (index = 0; index < nr_pages; index++) {
3121 __do_readpage(tree, pages[index], btrfs_get_extent, em_cached,
3122 bio, 0, bio_flags, 0, prev_em_start);
3123 put_page(pages[index]);
3124 }
3125 }
3126
3127 static void __extent_readpages(struct extent_io_tree *tree,
3128 struct page *pages[],
3129 int nr_pages,
3130 struct extent_map **em_cached,
3131 struct bio **bio, unsigned long *bio_flags,
3132 u64 *prev_em_start)
3133 {
3134 u64 start = 0;
3135 u64 end = 0;
3136 u64 page_start;
3137 int index;
3138 int first_index = 0;
3139
3140 for (index = 0; index < nr_pages; index++) {
3141 page_start = page_offset(pages[index]);
3142 if (!end) {
3143 start = page_start;
3144 end = start + PAGE_SIZE - 1;
3145 first_index = index;
3146 } else if (end + 1 == page_start) {
3147 end += PAGE_SIZE;
3148 } else {
3149 __do_contiguous_readpages(tree, &pages[first_index],
3150 index - first_index, start,
3151 end, em_cached,
3152 bio, bio_flags,
3153 prev_em_start);
3154 start = page_start;
3155 end = start + PAGE_SIZE - 1;
3156 first_index = index;
3157 }
3158 }
3159
3160 if (end)
3161 __do_contiguous_readpages(tree, &pages[first_index],
3162 index - first_index, start,
3163 end, em_cached, bio,
3164 bio_flags, prev_em_start);
3165 }
3166
3167 static int __extent_read_full_page(struct extent_io_tree *tree,
3168 struct page *page,
3169 get_extent_t *get_extent,
3170 struct bio **bio, int mirror_num,
3171 unsigned long *bio_flags,
3172 unsigned int read_flags)
3173 {
3174 struct inode *inode = page->mapping->host;
3175 struct btrfs_ordered_extent *ordered;
3176 u64 start = page_offset(page);
3177 u64 end = start + PAGE_SIZE - 1;
3178 int ret;
3179
3180 while (1) {
3181 lock_extent(tree, start, end);
3182 ordered = btrfs_lookup_ordered_range(BTRFS_I(inode), start,
3183 PAGE_SIZE);
3184 if (!ordered)
3185 break;
3186 unlock_extent(tree, start, end);
3187 btrfs_start_ordered_extent(inode, ordered, 1);
3188 btrfs_put_ordered_extent(ordered);
3189 }
3190
3191 ret = __do_readpage(tree, page, get_extent, NULL, bio, mirror_num,
3192 bio_flags, read_flags, NULL);
3193 return ret;
3194 }
3195
3196 int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
3197 get_extent_t *get_extent, int mirror_num)
3198 {
3199 struct bio *bio = NULL;
3200 unsigned long bio_flags = 0;
3201 int ret;
3202
3203 ret = __extent_read_full_page(tree, page, get_extent, &bio, mirror_num,
3204 &bio_flags, 0);
3205 if (bio)
3206 ret = submit_one_bio(bio, mirror_num, bio_flags);
3207 return ret;
3208 }
3209
3210 static void update_nr_written(struct writeback_control *wbc,
3211 unsigned long nr_written)
3212 {
3213 wbc->nr_to_write -= nr_written;
3214 }
3215
3216 /*
3217 * helper for __extent_writepage, doing all of the delayed allocation setup.
3218 *
3219 * This returns 1 if our fill_delalloc function did all the work required
3220 * to write the page (copy into inline extent). In this case the IO has
3221 * been started and the page is already unlocked.
3222 *
3223 * This returns 0 if all went well (page still locked)
3224 * This returns < 0 if there were errors (page still locked)
3225 */
3226 static noinline_for_stack int writepage_delalloc(struct inode *inode,
3227 struct page *page, struct writeback_control *wbc,
3228 struct extent_page_data *epd,
3229 u64 delalloc_start,
3230 unsigned long *nr_written)
3231 {
3232 struct extent_io_tree *tree = epd->tree;
3233 u64 page_end = delalloc_start + PAGE_SIZE - 1;
3234 u64 nr_delalloc;
3235 u64 delalloc_to_write = 0;
3236 u64 delalloc_end = 0;
3237 int ret;
3238 int page_started = 0;
3239
3240 if (epd->extent_locked || !tree->ops || !tree->ops->fill_delalloc)
3241 return 0;
3242
3243 while (delalloc_end < page_end) {
3244 nr_delalloc = find_lock_delalloc_range(inode, tree,
3245 page,
3246 &delalloc_start,
3247 &delalloc_end,
3248 BTRFS_MAX_EXTENT_SIZE);
3249 if (nr_delalloc == 0) {
3250 delalloc_start = delalloc_end + 1;
3251 continue;
3252 }
3253 ret = tree->ops->fill_delalloc(inode, page,
3254 delalloc_start,
3255 delalloc_end,
3256 &page_started,
3257 nr_written, wbc);
3258 /* File system has been set read-only */
3259 if (ret) {
3260 SetPageError(page);
3261 /* fill_delalloc should be return < 0 for error
3262 * but just in case, we use > 0 here meaning the
3263 * IO is started, so we don't want to return > 0
3264 * unless things are going well.
3265 */
3266 ret = ret < 0 ? ret : -EIO;
3267 goto done;
3268 }
3269 /*
3270 * delalloc_end is already one less than the total length, so
3271 * we don't subtract one from PAGE_SIZE
3272 */
3273 delalloc_to_write += (delalloc_end - delalloc_start +
3274 PAGE_SIZE) >> PAGE_SHIFT;
3275 delalloc_start = delalloc_end + 1;
3276 }
3277 if (wbc->nr_to_write < delalloc_to_write) {
3278 int thresh = 8192;
3279
3280 if (delalloc_to_write < thresh * 2)
3281 thresh = delalloc_to_write;
3282 wbc->nr_to_write = min_t(u64, delalloc_to_write,
3283 thresh);
3284 }
3285
3286 /* did the fill delalloc function already unlock and start
3287 * the IO?
3288 */
3289 if (page_started) {
3290 /*
3291 * we've unlocked the page, so we can't update
3292 * the mapping's writeback index, just update
3293 * nr_to_write.
3294 */
3295 wbc->nr_to_write -= *nr_written;
3296 return 1;
3297 }
3298
3299 ret = 0;
3300
3301 done:
3302 return ret;
3303 }
3304
3305 /*
3306 * helper for __extent_writepage. This calls the writepage start hooks,
3307 * and does the loop to map the page into extents and bios.
3308 *
3309 * We return 1 if the IO is started and the page is unlocked,
3310 * 0 if all went well (page still locked)
3311 * < 0 if there were errors (page still locked)
3312 */
3313 static noinline_for_stack int __extent_writepage_io(struct inode *inode,
3314 struct page *page,
3315 struct writeback_control *wbc,
3316 struct extent_page_data *epd,
3317 loff_t i_size,
3318 unsigned long nr_written,
3319 unsigned int write_flags, int *nr_ret)
3320 {
3321 struct extent_io_tree *tree = epd->tree;
3322 u64 start = page_offset(page);
3323 u64 page_end = start + PAGE_SIZE - 1;
3324 u64 end;
3325 u64 cur = start;
3326 u64 extent_offset;
3327 u64 block_start;
3328 u64 iosize;
3329 struct extent_map *em;
3330 struct block_device *bdev;
3331 size_t pg_offset = 0;
3332 size_t blocksize;
3333 int ret = 0;
3334 int nr = 0;
3335 bool compressed;
3336
3337 if (tree->ops && tree->ops->writepage_start_hook) {
3338 ret = tree->ops->writepage_start_hook(page, start,
3339 page_end);
3340 if (ret) {
3341 /* Fixup worker will requeue */
3342 if (ret == -EBUSY)
3343 wbc->pages_skipped++;
3344 else
3345 redirty_page_for_writepage(wbc, page);
3346
3347 update_nr_written(wbc, nr_written);
3348 unlock_page(page);
3349 return 1;
3350 }
3351 }
3352
3353 /*
3354 * we don't want to touch the inode after unlocking the page,
3355 * so we update the mapping writeback index now
3356 */
3357 update_nr_written(wbc, nr_written + 1);
3358
3359 end = page_end;
3360 if (i_size <= start) {
3361 if (tree->ops && tree->ops->writepage_end_io_hook)
3362 tree->ops->writepage_end_io_hook(page, start,
3363 page_end, NULL, 1);
3364 goto done;
3365 }
3366
3367 blocksize = inode->i_sb->s_blocksize;
3368
3369 while (cur <= end) {
3370 u64 em_end;
3371 u64 offset;
3372
3373 if (cur >= i_size) {
3374 if (tree->ops && tree->ops->writepage_end_io_hook)
3375 tree->ops->writepage_end_io_hook(page, cur,
3376 page_end, NULL, 1);
3377 break;
3378 }
3379 em = btrfs_get_extent(BTRFS_I(inode), page, pg_offset, cur,
3380 end - cur + 1, 1);
3381 if (IS_ERR_OR_NULL(em)) {
3382 SetPageError(page);
3383 ret = PTR_ERR_OR_ZERO(em);
3384 break;
3385 }
3386
3387 extent_offset = cur - em->start;
3388 em_end = extent_map_end(em);
3389 BUG_ON(em_end <= cur);
3390 BUG_ON(end < cur);
3391 iosize = min(em_end - cur, end - cur + 1);
3392 iosize = ALIGN(iosize, blocksize);
3393 offset = em->block_start + extent_offset;
3394 bdev = em->bdev;
3395 block_start = em->block_start;
3396 compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
3397 free_extent_map(em);
3398 em = NULL;
3399
3400 /*
3401 * compressed and inline extents are written through other
3402 * paths in the FS
3403 */
3404 if (compressed || block_start == EXTENT_MAP_HOLE ||
3405 block_start == EXTENT_MAP_INLINE) {
3406 /*
3407 * end_io notification does not happen here for
3408 * compressed extents
3409 */
3410 if (!compressed && tree->ops &&
3411 tree->ops->writepage_end_io_hook)
3412 tree->ops->writepage_end_io_hook(page, cur,
3413 cur + iosize - 1,
3414 NULL, 1);
3415 else if (compressed) {
3416 /* we don't want to end_page_writeback on
3417 * a compressed extent. this happens
3418 * elsewhere
3419 */
3420 nr++;
3421 }
3422
3423 cur += iosize;
3424 pg_offset += iosize;
3425 continue;
3426 }
3427
3428 set_range_writeback(tree, cur, cur + iosize - 1);
3429 if (!PageWriteback(page)) {
3430 btrfs_err(BTRFS_I(inode)->root->fs_info,
3431 "page %lu not writeback, cur %llu end %llu",
3432 page->index, cur, end);
3433 }
3434
3435 ret = submit_extent_page(REQ_OP_WRITE | write_flags, tree, wbc,
3436 page, offset, iosize, pg_offset,
3437 bdev, &epd->bio,
3438 end_bio_extent_writepage,
3439 0, 0, 0, false);
3440 if (ret) {
3441 SetPageError(page);
3442 if (PageWriteback(page))
3443 end_page_writeback(page);
3444 }
3445
3446 cur = cur + iosize;
3447 pg_offset += iosize;
3448 nr++;
3449 }
3450 done:
3451 *nr_ret = nr;
3452 return ret;
3453 }
3454
3455 /*
3456 * the writepage semantics are similar to regular writepage. extent
3457 * records are inserted to lock ranges in the tree, and as dirty areas
3458 * are found, they are marked writeback. Then the lock bits are removed
3459 * and the end_io handler clears the writeback ranges
3460 */
3461 static int __extent_writepage(struct page *page, struct writeback_control *wbc,
3462 struct extent_page_data *epd)
3463 {
3464 struct inode *inode = page->mapping->host;
3465 u64 start = page_offset(page);
3466 u64 page_end = start + PAGE_SIZE - 1;
3467 int ret;
3468 int nr = 0;
3469 size_t pg_offset = 0;
3470 loff_t i_size = i_size_read(inode);
3471 unsigned long end_index = i_size >> PAGE_SHIFT;
3472 unsigned int write_flags = 0;
3473 unsigned long nr_written = 0;
3474
3475 write_flags = wbc_to_write_flags(wbc);
3476
3477 trace___extent_writepage(page, inode, wbc);
3478
3479 WARN_ON(!PageLocked(page));
3480
3481 ClearPageError(page);
3482
3483 pg_offset = i_size & (PAGE_SIZE - 1);
3484 if (page->index > end_index ||
3485 (page->index == end_index && !pg_offset)) {
3486 page->mapping->a_ops->invalidatepage(page, 0, PAGE_SIZE);
3487 unlock_page(page);
3488 return 0;
3489 }
3490
3491 if (page->index == end_index) {
3492 char *userpage;
3493
3494 userpage = kmap_atomic(page);
3495 memset(userpage + pg_offset, 0,
3496 PAGE_SIZE - pg_offset);
3497 kunmap_atomic(userpage);
3498 flush_dcache_page(page);
3499 }
3500
3501 pg_offset = 0;
3502
3503 set_page_extent_mapped(page);
3504
3505 ret = writepage_delalloc(inode, page, wbc, epd, start, &nr_written);
3506 if (ret == 1)
3507 goto done_unlocked;
3508 if (ret)
3509 goto done;
3510
3511 ret = __extent_writepage_io(inode, page, wbc, epd,
3512 i_size, nr_written, write_flags, &nr);
3513 if (ret == 1)
3514 goto done_unlocked;
3515
3516 done:
3517 if (nr == 0) {
3518 /* make sure the mapping tag for page dirty gets cleared */
3519 set_page_writeback(page);
3520 end_page_writeback(page);
3521 }
3522 if (PageError(page)) {
3523 ret = ret < 0 ? ret : -EIO;
3524 end_extent_writepage(page, ret, start, page_end);
3525 }
3526 unlock_page(page);
3527 return ret;
3528
3529 done_unlocked:
3530 return 0;
3531 }
3532
3533 void wait_on_extent_buffer_writeback(struct extent_buffer *eb)
3534 {
3535 wait_on_bit_io(&eb->bflags, EXTENT_BUFFER_WRITEBACK,
3536 TASK_UNINTERRUPTIBLE);
3537 }
3538
3539 static noinline_for_stack int
3540 lock_extent_buffer_for_io(struct extent_buffer *eb,
3541 struct btrfs_fs_info *fs_info,
3542 struct extent_page_data *epd)
3543 {
3544 unsigned long i, num_pages;
3545 int flush = 0;
3546 int ret = 0;
3547
3548 if (!btrfs_try_tree_write_lock(eb)) {
3549 flush = 1;
3550 flush_write_bio(epd);
3551 btrfs_tree_lock(eb);
3552 }
3553
3554 if (test_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags)) {
3555 btrfs_tree_unlock(eb);
3556 if (!epd->sync_io)
3557 return 0;
3558 if (!flush) {
3559 flush_write_bio(epd);
3560 flush = 1;
3561 }
3562 while (1) {
3563 wait_on_extent_buffer_writeback(eb);
3564 btrfs_tree_lock(eb);
3565 if (!test_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags))
3566 break;
3567 btrfs_tree_unlock(eb);
3568 }
3569 }
3570
3571 /*
3572 * We need to do this to prevent races in people who check if the eb is
3573 * under IO since we can end up having no IO bits set for a short period
3574 * of time.
3575 */
3576 spin_lock(&eb->refs_lock);
3577 if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
3578 set_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags);
3579 spin_unlock(&eb->refs_lock);
3580 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
3581 percpu_counter_add_batch(&fs_info->dirty_metadata_bytes,
3582 -eb->len,
3583 fs_info->dirty_metadata_batch);
3584 ret = 1;
3585 } else {
3586 spin_unlock(&eb->refs_lock);
3587 }
3588
3589 btrfs_tree_unlock(eb);
3590
3591 if (!ret)
3592 return ret;
3593
3594 num_pages = num_extent_pages(eb->start, eb->len);
3595 for (i = 0; i < num_pages; i++) {
3596 struct page *p = eb->pages[i];
3597
3598 if (!trylock_page(p)) {
3599 if (!flush) {
3600 flush_write_bio(epd);
3601 flush = 1;
3602 }
3603 lock_page(p);
3604 }
3605 }
3606
3607 return ret;
3608 }
3609
3610 static void end_extent_buffer_writeback(struct extent_buffer *eb)
3611 {
3612 clear_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags);
3613 smp_mb__after_atomic();
3614 wake_up_bit(&eb->bflags, EXTENT_BUFFER_WRITEBACK);
3615 }
3616
3617 static void set_btree_ioerr(struct page *page)
3618 {
3619 struct extent_buffer *eb = (struct extent_buffer *)page->private;
3620
3621 SetPageError(page);
3622 if (test_and_set_bit(EXTENT_BUFFER_WRITE_ERR, &eb->bflags))
3623 return;
3624
3625 /*
3626 * If writeback for a btree extent that doesn't belong to a log tree
3627 * failed, increment the counter transaction->eb_write_errors.
3628 * We do this because while the transaction is running and before it's
3629 * committing (when we call filemap_fdata[write|wait]_range against
3630 * the btree inode), we might have
3631 * btree_inode->i_mapping->a_ops->writepages() called by the VM - if it
3632 * returns an error or an error happens during writeback, when we're
3633 * committing the transaction we wouldn't know about it, since the pages
3634 * can be no longer dirty nor marked anymore for writeback (if a
3635 * subsequent modification to the extent buffer didn't happen before the
3636 * transaction commit), which makes filemap_fdata[write|wait]_range not
3637 * able to find the pages tagged with SetPageError at transaction
3638 * commit time. So if this happens we must abort the transaction,
3639 * otherwise we commit a super block with btree roots that point to
3640 * btree nodes/leafs whose content on disk is invalid - either garbage
3641 * or the content of some node/leaf from a past generation that got
3642 * cowed or deleted and is no longer valid.
3643 *
3644 * Note: setting AS_EIO/AS_ENOSPC in the btree inode's i_mapping would
3645 * not be enough - we need to distinguish between log tree extents vs
3646 * non-log tree extents, and the next filemap_fdatawait_range() call
3647 * will catch and clear such errors in the mapping - and that call might
3648 * be from a log sync and not from a transaction commit. Also, checking
3649 * for the eb flag EXTENT_BUFFER_WRITE_ERR at transaction commit time is
3650 * not done and would not be reliable - the eb might have been released
3651 * from memory and reading it back again means that flag would not be
3652 * set (since it's a runtime flag, not persisted on disk).
3653 *
3654 * Using the flags below in the btree inode also makes us achieve the
3655 * goal of AS_EIO/AS_ENOSPC when writepages() returns success, started
3656 * writeback for all dirty pages and before filemap_fdatawait_range()
3657 * is called, the writeback for all dirty pages had already finished
3658 * with errors - because we were not using AS_EIO/AS_ENOSPC,
3659 * filemap_fdatawait_range() would return success, as it could not know
3660 * that writeback errors happened (the pages were no longer tagged for
3661 * writeback).
3662 */
3663 switch (eb->log_index) {
3664 case -1:
3665 set_bit(BTRFS_FS_BTREE_ERR, &eb->fs_info->flags);
3666 break;
3667 case 0:
3668 set_bit(BTRFS_FS_LOG1_ERR, &eb->fs_info->flags);
3669 break;
3670 case 1:
3671 set_bit(BTRFS_FS_LOG2_ERR, &eb->fs_info->flags);
3672 break;
3673 default:
3674 BUG(); /* unexpected, logic error */
3675 }
3676 }
3677
3678 static void end_bio_extent_buffer_writepage(struct bio *bio)
3679 {
3680 struct bio_vec *bvec;
3681 struct extent_buffer *eb;
3682 int i, done;
3683
3684 ASSERT(!bio_flagged(bio, BIO_CLONED));
3685 bio_for_each_segment_all(bvec, bio, i) {
3686 struct page *page = bvec->bv_page;
3687
3688 eb = (struct extent_buffer *)page->private;
3689 BUG_ON(!eb);
3690 done = atomic_dec_and_test(&eb->io_pages);
3691
3692 if (bio->bi_status ||
3693 test_bit(EXTENT_BUFFER_WRITE_ERR, &eb->bflags)) {
3694 ClearPageUptodate(page);
3695 set_btree_ioerr(page);
3696 }
3697
3698 end_page_writeback(page);
3699
3700 if (!done)
3701 continue;
3702
3703 end_extent_buffer_writeback(eb);
3704 }
3705
3706 bio_put(bio);
3707 }
3708
3709 static noinline_for_stack int write_one_eb(struct extent_buffer *eb,
3710 struct btrfs_fs_info *fs_info,
3711 struct writeback_control *wbc,
3712 struct extent_page_data *epd)
3713 {
3714 struct block_device *bdev = fs_info->fs_devices->latest_bdev;
3715 struct extent_io_tree *tree = &BTRFS_I(fs_info->btree_inode)->io_tree;
3716 u64 offset = eb->start;
3717 u32 nritems;
3718 unsigned long i, num_pages;
3719 unsigned long start, end;
3720 unsigned int write_flags = wbc_to_write_flags(wbc) | REQ_META;
3721 int ret = 0;
3722
3723 clear_bit(EXTENT_BUFFER_WRITE_ERR, &eb->bflags);
3724 num_pages = num_extent_pages(eb->start, eb->len);
3725 atomic_set(&eb->io_pages, num_pages);
3726
3727 /* set btree blocks beyond nritems with 0 to avoid stale content. */
3728 nritems = btrfs_header_nritems(eb);
3729 if (btrfs_header_level(eb) > 0) {
3730 end = btrfs_node_key_ptr_offset(nritems);
3731
3732 memzero_extent_buffer(eb, end, eb->len - end);
3733 } else {
3734 /*
3735 * leaf:
3736 * header 0 1 2 .. N ... data_N .. data_2 data_1 data_0
3737 */
3738 start = btrfs_item_nr_offset(nritems);
3739 end = BTRFS_LEAF_DATA_OFFSET + leaf_data_end(fs_info, eb);
3740 memzero_extent_buffer(eb, start, end - start);
3741 }
3742
3743 for (i = 0; i < num_pages; i++) {
3744 struct page *p = eb->pages[i];
3745
3746 clear_page_dirty_for_io(p);
3747 set_page_writeback(p);
3748 ret = submit_extent_page(REQ_OP_WRITE | write_flags, tree, wbc,
3749 p, offset, PAGE_SIZE, 0, bdev,
3750 &epd->bio,
3751 end_bio_extent_buffer_writepage,
3752 0, 0, 0, false);
3753 if (ret) {
3754 set_btree_ioerr(p);
3755 if (PageWriteback(p))
3756 end_page_writeback(p);
3757 if (atomic_sub_and_test(num_pages - i, &eb->io_pages))
3758 end_extent_buffer_writeback(eb);
3759 ret = -EIO;
3760 break;
3761 }
3762 offset += PAGE_SIZE;
3763 update_nr_written(wbc, 1);
3764 unlock_page(p);
3765 }
3766
3767 if (unlikely(ret)) {
3768 for (; i < num_pages; i++) {
3769 struct page *p = eb->pages[i];
3770 clear_page_dirty_for_io(p);
3771 unlock_page(p);
3772 }
3773 }
3774
3775 return ret;
3776 }
3777
3778 int btree_write_cache_pages(struct address_space *mapping,
3779 struct writeback_control *wbc)
3780 {
3781 struct extent_io_tree *tree = &BTRFS_I(mapping->host)->io_tree;
3782 struct btrfs_fs_info *fs_info = BTRFS_I(mapping->host)->root->fs_info;
3783 struct extent_buffer *eb, *prev_eb = NULL;
3784 struct extent_page_data epd = {
3785 .bio = NULL,
3786 .tree = tree,
3787 .extent_locked = 0,
3788 .sync_io = wbc->sync_mode == WB_SYNC_ALL,
3789 };
3790 int ret = 0;
3791 int done = 0;
3792 int nr_to_write_done = 0;
3793 struct pagevec pvec;
3794 int nr_pages;
3795 pgoff_t index;
3796 pgoff_t end; /* Inclusive */
3797 int scanned = 0;
3798 int tag;
3799
3800 pagevec_init(&pvec);
3801 if (wbc->range_cyclic) {
3802 index = mapping->writeback_index; /* Start from prev offset */
3803 end = -1;
3804 } else {
3805 index = wbc->range_start >> PAGE_SHIFT;
3806 end = wbc->range_end >> PAGE_SHIFT;
3807 scanned = 1;
3808 }
3809 if (wbc->sync_mode == WB_SYNC_ALL)
3810 tag = PAGECACHE_TAG_TOWRITE;
3811 else
3812 tag = PAGECACHE_TAG_DIRTY;
3813 retry:
3814 if (wbc->sync_mode == WB_SYNC_ALL)
3815 tag_pages_for_writeback(mapping, index, end);
3816 while (!done && !nr_to_write_done && (index <= end) &&
3817 (nr_pages = pagevec_lookup_range_tag(&pvec, mapping, &index, end,
3818 tag))) {
3819 unsigned i;
3820
3821 scanned = 1;
3822 for (i = 0; i < nr_pages; i++) {
3823 struct page *page = pvec.pages[i];
3824
3825 if (!PagePrivate(page))
3826 continue;
3827
3828 spin_lock(&mapping->private_lock);
3829 if (!PagePrivate(page)) {
3830 spin_unlock(&mapping->private_lock);
3831 continue;
3832 }
3833
3834 eb = (struct extent_buffer *)page->private;
3835
3836 /*
3837 * Shouldn't happen and normally this would be a BUG_ON
3838 * but no sense in crashing the users box for something
3839 * we can survive anyway.
3840 */
3841 if (WARN_ON(!eb)) {
3842 spin_unlock(&mapping->private_lock);
3843 continue;
3844 }
3845
3846 if (eb == prev_eb) {
3847 spin_unlock(&mapping->private_lock);
3848 continue;
3849 }
3850
3851 ret = atomic_inc_not_zero(&eb->refs);
3852 spin_unlock(&mapping->private_lock);
3853 if (!ret)
3854 continue;
3855
3856 prev_eb = eb;
3857 ret = lock_extent_buffer_for_io(eb, fs_info, &epd);
3858 if (!ret) {
3859 free_extent_buffer(eb);
3860 continue;
3861 }
3862
3863 ret = write_one_eb(eb, fs_info, wbc, &epd);
3864 if (ret) {
3865 done = 1;
3866 free_extent_buffer(eb);
3867 break;
3868 }
3869 free_extent_buffer(eb);
3870
3871 /*
3872 * the filesystem may choose to bump up nr_to_write.
3873 * We have to make sure to honor the new nr_to_write
3874 * at any time
3875 */
3876 nr_to_write_done = wbc->nr_to_write <= 0;
3877 }
3878 pagevec_release(&pvec);
3879 cond_resched();
3880 }
3881 if (!scanned && !done) {
3882 /*
3883 * We hit the last page and there is more work to be done: wrap
3884 * back to the start of the file
3885 */
3886 scanned = 1;
3887 index = 0;
3888 goto retry;
3889 }
3890 flush_write_bio(&epd);
3891 return ret;
3892 }
3893
3894 /**
3895 * write_cache_pages - walk the list of dirty pages of the given address space and write all of them.
3896 * @mapping: address space structure to write
3897 * @wbc: subtract the number of written pages from *@wbc->nr_to_write
3898 * @data: data passed to __extent_writepage function
3899 *
3900 * If a page is already under I/O, write_cache_pages() skips it, even
3901 * if it's dirty. This is desirable behaviour for memory-cleaning writeback,
3902 * but it is INCORRECT for data-integrity system calls such as fsync(). fsync()
3903 * and msync() need to guarantee that all the data which was dirty at the time
3904 * the call was made get new I/O started against them. If wbc->sync_mode is
3905 * WB_SYNC_ALL then we were called for data integrity and we must wait for
3906 * existing IO to complete.
3907 */
3908 static int extent_write_cache_pages(struct address_space *mapping,
3909 struct writeback_control *wbc,
3910 struct extent_page_data *epd)
3911 {
3912 struct inode *inode = mapping->host;
3913 int ret = 0;
3914 int done = 0;
3915 int nr_to_write_done = 0;
3916 struct pagevec pvec;
3917 int nr_pages;
3918 pgoff_t index;
3919 pgoff_t end; /* Inclusive */
3920 pgoff_t done_index;
3921 int range_whole = 0;
3922 int scanned = 0;
3923 int tag;
3924
3925 /*
3926 * We have to hold onto the inode so that ordered extents can do their
3927 * work when the IO finishes. The alternative to this is failing to add
3928 * an ordered extent if the igrab() fails there and that is a huge pain
3929 * to deal with, so instead just hold onto the inode throughout the
3930 * writepages operation. If it fails here we are freeing up the inode
3931 * anyway and we'd rather not waste our time writing out stuff that is
3932 * going to be truncated anyway.
3933 */
3934 if (!igrab(inode))
3935 return 0;
3936
3937 pagevec_init(&pvec);
3938 if (wbc->range_cyclic) {
3939 index = mapping->writeback_index; /* Start from prev offset */
3940 end = -1;
3941 } else {
3942 index = wbc->range_start >> PAGE_SHIFT;
3943 end = wbc->range_end >> PAGE_SHIFT;
3944 if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
3945 range_whole = 1;
3946 scanned = 1;
3947 }
3948 if (wbc->sync_mode == WB_SYNC_ALL)
3949 tag = PAGECACHE_TAG_TOWRITE;
3950 else
3951 tag = PAGECACHE_TAG_DIRTY;
3952 retry:
3953 if (wbc->sync_mode == WB_SYNC_ALL)
3954 tag_pages_for_writeback(mapping, index, end);
3955 done_index = index;
3956 while (!done && !nr_to_write_done && (index <= end) &&
3957 (nr_pages = pagevec_lookup_range_tag(&pvec, mapping,
3958 &index, end, tag))) {
3959 unsigned i;
3960
3961 scanned = 1;
3962 for (i = 0; i < nr_pages; i++) {
3963 struct page *page = pvec.pages[i];
3964
3965 done_index = page->index;
3966 /*
3967 * At this point we hold neither the i_pages lock nor
3968 * the page lock: the page may be truncated or
3969 * invalidated (changing page->mapping to NULL),
3970 * or even swizzled back from swapper_space to
3971 * tmpfs file mapping
3972 */
3973 if (!trylock_page(page)) {
3974 flush_write_bio(epd);
3975 lock_page(page);
3976 }
3977
3978 if (unlikely(page->mapping != mapping)) {
3979 unlock_page(page);
3980 continue;
3981 }
3982
3983 if (wbc->sync_mode != WB_SYNC_NONE) {
3984 if (PageWriteback(page))
3985 flush_write_bio(epd);
3986 wait_on_page_writeback(page);
3987 }
3988
3989 if (PageWriteback(page) ||
3990 !clear_page_dirty_for_io(page)) {
3991 unlock_page(page);
3992 continue;
3993 }
3994
3995 ret = __extent_writepage(page, wbc, epd);
3996
3997 if (unlikely(ret == AOP_WRITEPAGE_ACTIVATE)) {
3998 unlock_page(page);
3999 ret = 0;
4000 }
4001 if (ret < 0) {
4002 /*
4003 * done_index is set past this page,
4004 * so media errors will not choke
4005 * background writeout for the entire
4006 * file. This has consequences for
4007 * range_cyclic semantics (ie. it may
4008 * not be suitable for data integrity
4009 * writeout).
4010 */
4011 done_index = page->index + 1;
4012 done = 1;
4013 break;
4014 }
4015
4016 /*
4017 * the filesystem may choose to bump up nr_to_write.
4018 * We have to make sure to honor the new nr_to_write
4019 * at any time
4020 */
4021 nr_to_write_done = wbc->nr_to_write <= 0;
4022 }
4023 pagevec_release(&pvec);
4024 cond_resched();
4025 }
4026 if (!scanned && !done) {
4027 /*
4028 * We hit the last page and there is more work to be done: wrap
4029 * back to the start of the file
4030 */
4031 scanned = 1;
4032 index = 0;
4033 goto retry;
4034 }
4035
4036 if (wbc->range_cyclic || (wbc->nr_to_write > 0 && range_whole))
4037 mapping->writeback_index = done_index;
4038
4039 btrfs_add_delayed_iput(inode);
4040 return ret;
4041 }
4042
4043 static void flush_write_bio(struct extent_page_data *epd)
4044 {
4045 if (epd->bio) {
4046 int ret;
4047
4048 ret = submit_one_bio(epd->bio, 0, 0);
4049 BUG_ON(ret < 0); /* -ENOMEM */
4050 epd->bio = NULL;
4051 }
4052 }
4053
4054 int extent_write_full_page(struct page *page, struct writeback_control *wbc)
4055 {
4056 int ret;
4057 struct extent_page_data epd = {
4058 .bio = NULL,
4059 .tree = &BTRFS_I(page->mapping->host)->io_tree,
4060 .extent_locked = 0,
4061 .sync_io = wbc->sync_mode == WB_SYNC_ALL,
4062 };
4063
4064 ret = __extent_writepage(page, wbc, &epd);
4065
4066 flush_write_bio(&epd);
4067 return ret;
4068 }
4069
4070 int extent_write_locked_range(struct inode *inode, u64 start, u64 end,
4071 int mode)
4072 {
4073 int ret = 0;
4074 struct address_space *mapping = inode->i_mapping;
4075 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
4076 struct page *page;
4077 unsigned long nr_pages = (end - start + PAGE_SIZE) >>
4078 PAGE_SHIFT;
4079
4080 struct extent_page_data epd = {
4081 .bio = NULL,
4082 .tree = tree,
4083 .extent_locked = 1,
4084 .sync_io = mode == WB_SYNC_ALL,
4085 };
4086 struct writeback_control wbc_writepages = {
4087 .sync_mode = mode,
4088 .nr_to_write = nr_pages * 2,
4089 .range_start = start,
4090 .range_end = end + 1,
4091 };
4092
4093 while (start <= end) {
4094 page = find_get_page(mapping, start >> PAGE_SHIFT);
4095 if (clear_page_dirty_for_io(page))
4096 ret = __extent_writepage(page, &wbc_writepages, &epd);
4097 else {
4098 if (tree->ops && tree->ops->writepage_end_io_hook)
4099 tree->ops->writepage_end_io_hook(page, start,
4100 start + PAGE_SIZE - 1,
4101 NULL, 1);
4102 unlock_page(page);
4103 }
4104 put_page(page);
4105 start += PAGE_SIZE;
4106 }
4107
4108 flush_write_bio(&epd);
4109 return ret;
4110 }
4111
4112 int extent_writepages(struct extent_io_tree *tree,
4113 struct address_space *mapping,
4114 struct writeback_control *wbc)
4115 {
4116 int ret = 0;
4117 struct extent_page_data epd = {
4118 .bio = NULL,
4119 .tree = tree,
4120 .extent_locked = 0,
4121 .sync_io = wbc->sync_mode == WB_SYNC_ALL,
4122 };
4123
4124 ret = extent_write_cache_pages(mapping, wbc, &epd);
4125 flush_write_bio(&epd);
4126 return ret;
4127 }
4128
4129 int extent_readpages(struct extent_io_tree *tree,
4130 struct address_space *mapping,
4131 struct list_head *pages, unsigned nr_pages)
4132 {
4133 struct bio *bio = NULL;
4134 unsigned page_idx;
4135 unsigned long bio_flags = 0;
4136 struct page *pagepool[16];
4137 struct page *page;
4138 struct extent_map *em_cached = NULL;
4139 int nr = 0;
4140 u64 prev_em_start = (u64)-1;
4141
4142 for (page_idx = 0; page_idx < nr_pages; page_idx++) {
4143 page = list_entry(pages->prev, struct page, lru);
4144
4145 prefetchw(&page->flags);
4146 list_del(&page->lru);
4147 if (add_to_page_cache_lru(page, mapping,
4148 page->index,
4149 readahead_gfp_mask(mapping))) {
4150 put_page(page);
4151 continue;
4152 }
4153
4154 pagepool[nr++] = page;
4155 if (nr < ARRAY_SIZE(pagepool))
4156 continue;
4157 __extent_readpages(tree, pagepool, nr, &em_cached, &bio,
4158 &bio_flags, &prev_em_start);
4159 nr = 0;
4160 }
4161 if (nr)
4162 __extent_readpages(tree, pagepool, nr, &em_cached, &bio,
4163 &bio_flags, &prev_em_start);
4164
4165 if (em_cached)
4166 free_extent_map(em_cached);
4167
4168 BUG_ON(!list_empty(pages));
4169 if (bio)
4170 return submit_one_bio(bio, 0, bio_flags);
4171 return 0;
4172 }
4173
4174 /*
4175 * basic invalidatepage code, this waits on any locked or writeback
4176 * ranges corresponding to the page, and then deletes any extent state
4177 * records from the tree
4178 */
4179 int extent_invalidatepage(struct extent_io_tree *tree,
4180 struct page *page, unsigned long offset)
4181 {
4182 struct extent_state *cached_state = NULL;
4183 u64 start = page_offset(page);
4184 u64 end = start + PAGE_SIZE - 1;
4185 size_t blocksize = page->mapping->host->i_sb->s_blocksize;
4186
4187 start += ALIGN(offset, blocksize);
4188 if (start > end)
4189 return 0;
4190
4191 lock_extent_bits(tree, start, end, &cached_state);
4192 wait_on_page_writeback(page);
4193 clear_extent_bit(tree, start, end,
4194 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
4195 EXTENT_DO_ACCOUNTING,
4196 1, 1, &cached_state);
4197 return 0;
4198 }
4199
4200 /*
4201 * a helper for releasepage, this tests for areas of the page that
4202 * are locked or under IO and drops the related state bits if it is safe
4203 * to drop the page.
4204 */
4205 static int try_release_extent_state(struct extent_map_tree *map,
4206 struct extent_io_tree *tree,
4207 struct page *page, gfp_t mask)
4208 {
4209 u64 start = page_offset(page);
4210 u64 end = start + PAGE_SIZE - 1;
4211 int ret = 1;
4212
4213 if (test_range_bit(tree, start, end,
4214 EXTENT_IOBITS, 0, NULL))
4215 ret = 0;
4216 else {
4217 /*
4218 * at this point we can safely clear everything except the
4219 * locked bit and the nodatasum bit
4220 */
4221 ret = __clear_extent_bit(tree, start, end,
4222 ~(EXTENT_LOCKED | EXTENT_NODATASUM),
4223 0, 0, NULL, mask, NULL);
4224
4225 /* if clear_extent_bit failed for enomem reasons,
4226 * we can't allow the release to continue.
4227 */
4228 if (ret < 0)
4229 ret = 0;
4230 else
4231 ret = 1;
4232 }
4233 return ret;
4234 }
4235
4236 /*
4237 * a helper for releasepage. As long as there are no locked extents
4238 * in the range corresponding to the page, both state records and extent
4239 * map records are removed
4240 */
4241 int try_release_extent_mapping(struct extent_map_tree *map,
4242 struct extent_io_tree *tree, struct page *page,
4243 gfp_t mask)
4244 {
4245 struct extent_map *em;
4246 u64 start = page_offset(page);
4247 u64 end = start + PAGE_SIZE - 1;
4248 struct btrfs_inode *btrfs_inode = BTRFS_I(page->mapping->host);
4249
4250 if (gfpflags_allow_blocking(mask) &&
4251 page->mapping->host->i_size > SZ_16M) {
4252 u64 len;
4253 while (start <= end) {
4254 len = end - start + 1;
4255 write_lock(&map->lock);
4256 em = lookup_extent_mapping(map, start, len);
4257 if (!em) {
4258 write_unlock(&map->lock);
4259 break;
4260 }
4261 if (test_bit(EXTENT_FLAG_PINNED, &em->flags) ||
4262 em->start != start) {
4263 write_unlock(&map->lock);
4264 free_extent_map(em);
4265 break;
4266 }
4267 if (!test_range_bit(tree, em->start,
4268 extent_map_end(em) - 1,
4269 EXTENT_LOCKED | EXTENT_WRITEBACK,
4270 0, NULL)) {
4271 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
4272 &btrfs_inode->runtime_flags);
4273 remove_extent_mapping(map, em);
4274 /* once for the rb tree */
4275 free_extent_map(em);
4276 }
4277 start = extent_map_end(em);
4278 write_unlock(&map->lock);
4279
4280 /* once for us */
4281 free_extent_map(em);
4282 }
4283 }
4284 return try_release_extent_state(map, tree, page, mask);
4285 }
4286
4287 /*
4288 * helper function for fiemap, which doesn't want to see any holes.
4289 * This maps until we find something past 'last'
4290 */
4291 static struct extent_map *get_extent_skip_holes(struct inode *inode,
4292 u64 offset, u64 last)
4293 {
4294 u64 sectorsize = btrfs_inode_sectorsize(inode);
4295 struct extent_map *em;
4296 u64 len;
4297
4298 if (offset >= last)
4299 return NULL;
4300
4301 while (1) {
4302 len = last - offset;
4303 if (len == 0)
4304 break;
4305 len = ALIGN(len, sectorsize);
4306 em = btrfs_get_extent_fiemap(BTRFS_I(inode), NULL, 0, offset,
4307 len, 0);
4308 if (IS_ERR_OR_NULL(em))
4309 return em;
4310
4311 /* if this isn't a hole return it */
4312 if (em->block_start != EXTENT_MAP_HOLE)
4313 return em;
4314
4315 /* this is a hole, advance to the next extent */
4316 offset = extent_map_end(em);
4317 free_extent_map(em);
4318 if (offset >= last)
4319 break;
4320 }
4321 return NULL;
4322 }
4323
4324 /*
4325 * To cache previous fiemap extent
4326 *
4327 * Will be used for merging fiemap extent
4328 */
4329 struct fiemap_cache {
4330 u64 offset;
4331 u64 phys;
4332 u64 len;
4333 u32 flags;
4334 bool cached;
4335 };
4336
4337 /*
4338 * Helper to submit fiemap extent.
4339 *
4340 * Will try to merge current fiemap extent specified by @offset, @phys,
4341 * @len and @flags with cached one.
4342 * And only when we fails to merge, cached one will be submitted as
4343 * fiemap extent.
4344 *
4345 * Return value is the same as fiemap_fill_next_extent().
4346 */
4347 static int emit_fiemap_extent(struct fiemap_extent_info *fieinfo,
4348 struct fiemap_cache *cache,
4349 u64 offset, u64 phys, u64 len, u32 flags)
4350 {
4351 int ret = 0;
4352
4353 if (!cache->cached)
4354 goto assign;
4355
4356 /*
4357 * Sanity check, extent_fiemap() should have ensured that new
4358 * fiemap extent won't overlap with cahced one.
4359 * Not recoverable.
4360 *
4361 * NOTE: Physical address can overlap, due to compression
4362 */
4363 if (cache->offset + cache->len > offset) {
4364 WARN_ON(1);
4365 return -EINVAL;
4366 }
4367
4368 /*
4369 * Only merges fiemap extents if
4370 * 1) Their logical addresses are continuous
4371 *
4372 * 2) Their physical addresses are continuous
4373 * So truly compressed (physical size smaller than logical size)
4374 * extents won't get merged with each other
4375 *
4376 * 3) Share same flags except FIEMAP_EXTENT_LAST
4377 * So regular extent won't get merged with prealloc extent
4378 */
4379 if (cache->offset + cache->len == offset &&
4380 cache->phys + cache->len == phys &&
4381 (cache->flags & ~FIEMAP_EXTENT_LAST) ==
4382 (flags & ~FIEMAP_EXTENT_LAST)) {
4383 cache->len += len;
4384 cache->flags |= flags;
4385 goto try_submit_last;
4386 }
4387
4388 /* Not mergeable, need to submit cached one */
4389 ret = fiemap_fill_next_extent(fieinfo, cache->offset, cache->phys,
4390 cache->len, cache->flags);
4391 cache->cached = false;
4392 if (ret)
4393 return ret;
4394 assign:
4395 cache->cached = true;
4396 cache->offset = offset;
4397 cache->phys = phys;
4398 cache->len = len;
4399 cache->flags = flags;
4400 try_submit_last:
4401 if (cache->flags & FIEMAP_EXTENT_LAST) {
4402 ret = fiemap_fill_next_extent(fieinfo, cache->offset,
4403 cache->phys, cache->len, cache->flags);
4404 cache->cached = false;
4405 }
4406 return ret;
4407 }
4408
4409 /*
4410 * Emit last fiemap cache
4411 *
4412 * The last fiemap cache may still be cached in the following case:
4413 * 0 4k 8k
4414 * |<- Fiemap range ->|
4415 * |<------------ First extent ----------->|
4416 *
4417 * In this case, the first extent range will be cached but not emitted.
4418 * So we must emit it before ending extent_fiemap().
4419 */
4420 static int emit_last_fiemap_cache(struct btrfs_fs_info *fs_info,
4421 struct fiemap_extent_info *fieinfo,
4422 struct fiemap_cache *cache)
4423 {
4424 int ret;
4425
4426 if (!cache->cached)
4427 return 0;
4428
4429 ret = fiemap_fill_next_extent(fieinfo, cache->offset, cache->phys,
4430 cache->len, cache->flags);
4431 cache->cached = false;
4432 if (ret > 0)
4433 ret = 0;
4434 return ret;
4435 }
4436
4437 int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
4438 __u64 start, __u64 len)
4439 {
4440 int ret = 0;
4441 u64 off = start;
4442 u64 max = start + len;
4443 u32 flags = 0;
4444 u32 found_type;
4445 u64 last;
4446 u64 last_for_get_extent = 0;
4447 u64 disko = 0;
4448 u64 isize = i_size_read(inode);
4449 struct btrfs_key found_key;
4450 struct extent_map *em = NULL;
4451 struct extent_state *cached_state = NULL;
4452 struct btrfs_path *path;
4453 struct btrfs_root *root = BTRFS_I(inode)->root;
4454 struct fiemap_cache cache = { 0 };
4455 int end = 0;
4456 u64 em_start = 0;
4457 u64 em_len = 0;
4458 u64 em_end = 0;
4459
4460 if (len == 0)
4461 return -EINVAL;
4462
4463 path = btrfs_alloc_path();
4464 if (!path)
4465 return -ENOMEM;
4466 path->leave_spinning = 1;
4467
4468 start = round_down(start, btrfs_inode_sectorsize(inode));
4469 len = round_up(max, btrfs_inode_sectorsize(inode)) - start;
4470
4471 /*
4472 * lookup the last file extent. We're not using i_size here
4473 * because there might be preallocation past i_size
4474 */
4475 ret = btrfs_lookup_file_extent(NULL, root, path,
4476 btrfs_ino(BTRFS_I(inode)), -1, 0);
4477 if (ret < 0) {
4478 btrfs_free_path(path);
4479 return ret;
4480 } else {
4481 WARN_ON(!ret);
4482 if (ret == 1)
4483 ret = 0;
4484 }
4485
4486 path->slots[0]--;
4487 btrfs_item_key_to_cpu(path->nodes[0], &found_key, path->slots[0]);
4488 found_type = found_key.type;
4489
4490 /* No extents, but there might be delalloc bits */
4491 if (found_key.objectid != btrfs_ino(BTRFS_I(inode)) ||
4492 found_type != BTRFS_EXTENT_DATA_KEY) {
4493 /* have to trust i_size as the end */
4494 last = (u64)-1;
4495 last_for_get_extent = isize;
4496 } else {
4497 /*
4498 * remember the start of the last extent. There are a
4499 * bunch of different factors that go into the length of the
4500 * extent, so its much less complex to remember where it started
4501 */
4502 last = found_key.offset;
4503 last_for_get_extent = last + 1;
4504 }
4505 btrfs_release_path(path);
4506
4507 /*
4508 * we might have some extents allocated but more delalloc past those
4509 * extents. so, we trust isize unless the start of the last extent is
4510 * beyond isize
4511 */
4512 if (last < isize) {
4513 last = (u64)-1;
4514 last_for_get_extent = isize;
4515 }
4516
4517 lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + len - 1,
4518 &cached_state);
4519
4520 em = get_extent_skip_holes(inode, start, last_for_get_extent);
4521 if (!em)
4522 goto out;
4523 if (IS_ERR(em)) {
4524 ret = PTR_ERR(em);
4525 goto out;
4526 }
4527
4528 while (!end) {
4529 u64 offset_in_extent = 0;
4530
4531 /* break if the extent we found is outside the range */
4532 if (em->start >= max || extent_map_end(em) < off)
4533 break;
4534
4535 /*
4536 * get_extent may return an extent that starts before our
4537 * requested range. We have to make sure the ranges
4538 * we return to fiemap always move forward and don't
4539 * overlap, so adjust the offsets here
4540 */
4541 em_start = max(em->start, off);
4542
4543 /*
4544 * record the offset from the start of the extent
4545 * for adjusting the disk offset below. Only do this if the
4546 * extent isn't compressed since our in ram offset may be past
4547 * what we have actually allocated on disk.
4548 */
4549 if (!test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
4550 offset_in_extent = em_start - em->start;
4551 em_end = extent_map_end(em);
4552 em_len = em_end - em_start;
4553 disko = 0;
4554 flags = 0;
4555
4556 /*
4557 * bump off for our next call to get_extent
4558 */
4559 off = extent_map_end(em);
4560 if (off >= max)
4561 end = 1;
4562
4563 if (em->block_start == EXTENT_MAP_LAST_BYTE) {
4564 end = 1;
4565 flags |= FIEMAP_EXTENT_LAST;
4566 } else if (em->block_start == EXTENT_MAP_INLINE) {
4567 flags |= (FIEMAP_EXTENT_DATA_INLINE |
4568 FIEMAP_EXTENT_NOT_ALIGNED);
4569 } else if (em->block_start == EXTENT_MAP_DELALLOC) {
4570 flags |= (FIEMAP_EXTENT_DELALLOC |
4571 FIEMAP_EXTENT_UNKNOWN);
4572 } else if (fieinfo->fi_extents_max) {
4573 u64 bytenr = em->block_start -
4574 (em->start - em->orig_start);
4575
4576 disko = em->block_start + offset_in_extent;
4577
4578 /*
4579 * As btrfs supports shared space, this information
4580 * can be exported to userspace tools via
4581 * flag FIEMAP_EXTENT_SHARED. If fi_extents_max == 0
4582 * then we're just getting a count and we can skip the
4583 * lookup stuff.
4584 */
4585 ret = btrfs_check_shared(root,
4586 btrfs_ino(BTRFS_I(inode)),
4587 bytenr);
4588 if (ret < 0)
4589 goto out_free;
4590 if (ret)
4591 flags |= FIEMAP_EXTENT_SHARED;
4592 ret = 0;
4593 }
4594 if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
4595 flags |= FIEMAP_EXTENT_ENCODED;
4596 if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
4597 flags |= FIEMAP_EXTENT_UNWRITTEN;
4598
4599 free_extent_map(em);
4600 em = NULL;
4601 if ((em_start >= last) || em_len == (u64)-1 ||
4602 (last == (u64)-1 && isize <= em_end)) {
4603 flags |= FIEMAP_EXTENT_LAST;
4604 end = 1;
4605 }
4606
4607 /* now scan forward to see if this is really the last extent. */
4608 em = get_extent_skip_holes(inode, off, last_for_get_extent);
4609 if (IS_ERR(em)) {
4610 ret = PTR_ERR(em);
4611 goto out;
4612 }
4613 if (!em) {
4614 flags |= FIEMAP_EXTENT_LAST;
4615 end = 1;
4616 }
4617 ret = emit_fiemap_extent(fieinfo, &cache, em_start, disko,
4618 em_len, flags);
4619 if (ret) {
4620 if (ret == 1)
4621 ret = 0;
4622 goto out_free;
4623 }
4624 }
4625 out_free:
4626 if (!ret)
4627 ret = emit_last_fiemap_cache(root->fs_info, fieinfo, &cache);
4628 free_extent_map(em);
4629 out:
4630 btrfs_free_path(path);
4631 unlock_extent_cached(&BTRFS_I(inode)->io_tree, start, start + len - 1,
4632 &cached_state);
4633 return ret;
4634 }
4635
4636 static void __free_extent_buffer(struct extent_buffer *eb)
4637 {
4638 btrfs_leak_debug_del(&eb->leak_list);
4639 kmem_cache_free(extent_buffer_cache, eb);
4640 }
4641
4642 int extent_buffer_under_io(struct extent_buffer *eb)
4643 {
4644 return (atomic_read(&eb->io_pages) ||
4645 test_bit(EXTENT_BUFFER_WRITEBACK, &eb->bflags) ||
4646 test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
4647 }
4648
4649 /*
4650 * Helper for releasing extent buffer page.
4651 */
4652 static void btrfs_release_extent_buffer_page(struct extent_buffer *eb)
4653 {
4654 unsigned long index;
4655 struct page *page;
4656 int mapped = !test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags);
4657
4658 BUG_ON(extent_buffer_under_io(eb));
4659
4660 index = num_extent_pages(eb->start, eb->len);
4661 if (index == 0)
4662 return;
4663
4664 do {
4665 index--;
4666 page = eb->pages[index];
4667 if (!page)
4668 continue;
4669 if (mapped)
4670 spin_lock(&page->mapping->private_lock);
4671 /*
4672 * We do this since we'll remove the pages after we've
4673 * removed the eb from the radix tree, so we could race
4674 * and have this page now attached to the new eb. So
4675 * only clear page_private if it's still connected to
4676 * this eb.
4677 */
4678 if (PagePrivate(page) &&
4679 page->private == (unsigned long)eb) {
4680 BUG_ON(test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags));
4681 BUG_ON(PageDirty(page));
4682 BUG_ON(PageWriteback(page));
4683 /*
4684 * We need to make sure we haven't be attached
4685 * to a new eb.
4686 */
4687 ClearPagePrivate(page);
4688 set_page_private(page, 0);
4689 /* One for the page private */
4690 put_page(page);
4691 }
4692
4693 if (mapped)
4694 spin_unlock(&page->mapping->private_lock);
4695
4696 /* One for when we allocated the page */
4697 put_page(page);
4698 } while (index != 0);
4699 }
4700
4701 /*
4702 * Helper for releasing the extent buffer.
4703 */
4704 static inline void btrfs_release_extent_buffer(struct extent_buffer *eb)
4705 {
4706 btrfs_release_extent_buffer_page(eb);
4707 __free_extent_buffer(eb);
4708 }
4709
4710 static struct extent_buffer *
4711 __alloc_extent_buffer(struct btrfs_fs_info *fs_info, u64 start,
4712 unsigned long len)
4713 {
4714 struct extent_buffer *eb = NULL;
4715
4716 eb = kmem_cache_zalloc(extent_buffer_cache, GFP_NOFS|__GFP_NOFAIL);
4717 eb->start = start;
4718 eb->len = len;
4719 eb->fs_info = fs_info;
4720 eb->bflags = 0;
4721 rwlock_init(&eb->lock);
4722 atomic_set(&eb->write_locks, 0);
4723 atomic_set(&eb->read_locks, 0);
4724 atomic_set(&eb->blocking_readers, 0);
4725 atomic_set(&eb->blocking_writers, 0);
4726 atomic_set(&eb->spinning_readers, 0);
4727 atomic_set(&eb->spinning_writers, 0);
4728 eb->lock_nested = 0;
4729 init_waitqueue_head(&eb->write_lock_wq);
4730 init_waitqueue_head(&eb->read_lock_wq);
4731
4732 btrfs_leak_debug_add(&eb->leak_list, &buffers);
4733
4734 spin_lock_init(&eb->refs_lock);
4735 atomic_set(&eb->refs, 1);
4736 atomic_set(&eb->io_pages, 0);
4737
4738 /*
4739 * Sanity checks, currently the maximum is 64k covered by 16x 4k pages
4740 */
4741 BUILD_BUG_ON(BTRFS_MAX_METADATA_BLOCKSIZE
4742 > MAX_INLINE_EXTENT_BUFFER_SIZE);
4743 BUG_ON(len > MAX_INLINE_EXTENT_BUFFER_SIZE);
4744
4745 return eb;
4746 }
4747
4748 struct extent_buffer *btrfs_clone_extent_buffer(struct extent_buffer *src)
4749 {
4750 unsigned long i;
4751 struct page *p;
4752 struct extent_buffer *new;
4753 unsigned long num_pages = num_extent_pages(src->start, src->len);
4754
4755 new = __alloc_extent_buffer(src->fs_info, src->start, src->len);
4756 if (new == NULL)
4757 return NULL;
4758
4759 for (i = 0; i < num_pages; i++) {
4760 p = alloc_page(GFP_NOFS);
4761 if (!p) {
4762 btrfs_release_extent_buffer(new);
4763 return NULL;
4764 }
4765 attach_extent_buffer_page(new, p);
4766 WARN_ON(PageDirty(p));
4767 SetPageUptodate(p);
4768 new->pages[i] = p;
4769 copy_page(page_address(p), page_address(src->pages[i]));
4770 }
4771
4772 set_bit(EXTENT_BUFFER_UPTODATE, &new->bflags);
4773 set_bit(EXTENT_BUFFER_DUMMY, &new->bflags);
4774
4775 return new;
4776 }
4777
4778 struct extent_buffer *__alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
4779 u64 start, unsigned long len)
4780 {
4781 struct extent_buffer *eb;
4782 unsigned long num_pages;
4783 unsigned long i;
4784
4785 num_pages = num_extent_pages(start, len);
4786
4787 eb = __alloc_extent_buffer(fs_info, start, len);
4788 if (!eb)
4789 return NULL;
4790
4791 for (i = 0; i < num_pages; i++) {
4792 eb->pages[i] = alloc_page(GFP_NOFS);
4793 if (!eb->pages[i])
4794 goto err;
4795 }
4796 set_extent_buffer_uptodate(eb);
4797 btrfs_set_header_nritems(eb, 0);
4798 set_bit(EXTENT_BUFFER_DUMMY, &eb->bflags);
4799
4800 return eb;
4801 err:
4802 for (; i > 0; i--)
4803 __free_page(eb->pages[i - 1]);
4804 __free_extent_buffer(eb);
4805 return NULL;
4806 }
4807
4808 struct extent_buffer *alloc_dummy_extent_buffer(struct btrfs_fs_info *fs_info,
4809 u64 start)
4810 {
4811 return __alloc_dummy_extent_buffer(fs_info, start, fs_info->nodesize);
4812 }
4813
4814 static void check_buffer_tree_ref(struct extent_buffer *eb)
4815 {
4816 int refs;
4817 /* the ref bit is tricky. We have to make sure it is set
4818 * if we have the buffer dirty. Otherwise the
4819 * code to free a buffer can end up dropping a dirty
4820 * page
4821 *
4822 * Once the ref bit is set, it won't go away while the
4823 * buffer is dirty or in writeback, and it also won't
4824 * go away while we have the reference count on the
4825 * eb bumped.
4826 *
4827 * We can't just set the ref bit without bumping the
4828 * ref on the eb because free_extent_buffer might
4829 * see the ref bit and try to clear it. If this happens
4830 * free_extent_buffer might end up dropping our original
4831 * ref by mistake and freeing the page before we are able
4832 * to add one more ref.
4833 *
4834 * So bump the ref count first, then set the bit. If someone
4835 * beat us to it, drop the ref we added.
4836 */
4837 refs = atomic_read(&eb->refs);
4838 if (refs >= 2 && test_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
4839 return;
4840
4841 spin_lock(&eb->refs_lock);
4842 if (!test_and_set_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
4843 atomic_inc(&eb->refs);
4844 spin_unlock(&eb->refs_lock);
4845 }
4846
4847 static void mark_extent_buffer_accessed(struct extent_buffer *eb,
4848 struct page *accessed)
4849 {
4850 unsigned long num_pages, i;
4851
4852 check_buffer_tree_ref(eb);
4853
4854 num_pages = num_extent_pages(eb->start, eb->len);
4855 for (i = 0; i < num_pages; i++) {
4856 struct page *p = eb->pages[i];
4857
4858 if (p != accessed)
4859 mark_page_accessed(p);
4860 }
4861 }
4862
4863 struct extent_buffer *find_extent_buffer(struct btrfs_fs_info *fs_info,
4864 u64 start)
4865 {
4866 struct extent_buffer *eb;
4867
4868 rcu_read_lock();
4869 eb = radix_tree_lookup(&fs_info->buffer_radix,
4870 start >> PAGE_SHIFT);
4871 if (eb && atomic_inc_not_zero(&eb->refs)) {
4872 rcu_read_unlock();
4873 /*
4874 * Lock our eb's refs_lock to avoid races with
4875 * free_extent_buffer. When we get our eb it might be flagged
4876 * with EXTENT_BUFFER_STALE and another task running
4877 * free_extent_buffer might have seen that flag set,
4878 * eb->refs == 2, that the buffer isn't under IO (dirty and
4879 * writeback flags not set) and it's still in the tree (flag
4880 * EXTENT_BUFFER_TREE_REF set), therefore being in the process
4881 * of decrementing the extent buffer's reference count twice.
4882 * So here we could race and increment the eb's reference count,
4883 * clear its stale flag, mark it as dirty and drop our reference
4884 * before the other task finishes executing free_extent_buffer,
4885 * which would later result in an attempt to free an extent
4886 * buffer that is dirty.
4887 */
4888 if (test_bit(EXTENT_BUFFER_STALE, &eb->bflags)) {
4889 spin_lock(&eb->refs_lock);
4890 spin_unlock(&eb->refs_lock);
4891 }
4892 mark_extent_buffer_accessed(eb, NULL);
4893 return eb;
4894 }
4895 rcu_read_unlock();
4896
4897 return NULL;
4898 }
4899
4900 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
4901 struct extent_buffer *alloc_test_extent_buffer(struct btrfs_fs_info *fs_info,
4902 u64 start)
4903 {
4904 struct extent_buffer *eb, *exists = NULL;
4905 int ret;
4906
4907 eb = find_extent_buffer(fs_info, start);
4908 if (eb)
4909 return eb;
4910 eb = alloc_dummy_extent_buffer(fs_info, start);
4911 if (!eb)
4912 return NULL;
4913 eb->fs_info = fs_info;
4914 again:
4915 ret = radix_tree_preload(GFP_NOFS);
4916 if (ret)
4917 goto free_eb;
4918 spin_lock(&fs_info->buffer_lock);
4919 ret = radix_tree_insert(&fs_info->buffer_radix,
4920 start >> PAGE_SHIFT, eb);
4921 spin_unlock(&fs_info->buffer_lock);
4922 radix_tree_preload_end();
4923 if (ret == -EEXIST) {
4924 exists = find_extent_buffer(fs_info, start);
4925 if (exists)
4926 goto free_eb;
4927 else
4928 goto again;
4929 }
4930 check_buffer_tree_ref(eb);
4931 set_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags);
4932
4933 /*
4934 * We will free dummy extent buffer's if they come into
4935 * free_extent_buffer with a ref count of 2, but if we are using this we
4936 * want the buffers to stay in memory until we're done with them, so
4937 * bump the ref count again.
4938 */
4939 atomic_inc(&eb->refs);
4940 return eb;
4941 free_eb:
4942 btrfs_release_extent_buffer(eb);
4943 return exists;
4944 }
4945 #endif
4946
4947 struct extent_buffer *alloc_extent_buffer(struct btrfs_fs_info *fs_info,
4948 u64 start)
4949 {
4950 unsigned long len = fs_info->nodesize;
4951 unsigned long num_pages = num_extent_pages(start, len);
4952 unsigned long i;
4953 unsigned long index = start >> PAGE_SHIFT;
4954 struct extent_buffer *eb;
4955 struct extent_buffer *exists = NULL;
4956 struct page *p;
4957 struct address_space *mapping = fs_info->btree_inode->i_mapping;
4958 int uptodate = 1;
4959 int ret;
4960
4961 if (!IS_ALIGNED(start, fs_info->sectorsize)) {
4962 btrfs_err(fs_info, "bad tree block start %llu", start);
4963 return ERR_PTR(-EINVAL);
4964 }
4965
4966 eb = find_extent_buffer(fs_info, start);
4967 if (eb)
4968 return eb;
4969
4970 eb = __alloc_extent_buffer(fs_info, start, len);
4971 if (!eb)
4972 return ERR_PTR(-ENOMEM);
4973
4974 for (i = 0; i < num_pages; i++, index++) {
4975 p = find_or_create_page(mapping, index, GFP_NOFS|__GFP_NOFAIL);
4976 if (!p) {
4977 exists = ERR_PTR(-ENOMEM);
4978 goto free_eb;
4979 }
4980
4981 spin_lock(&mapping->private_lock);
4982 if (PagePrivate(p)) {
4983 /*
4984 * We could have already allocated an eb for this page
4985 * and attached one so lets see if we can get a ref on
4986 * the existing eb, and if we can we know it's good and
4987 * we can just return that one, else we know we can just
4988 * overwrite page->private.
4989 */
4990 exists = (struct extent_buffer *)p->private;
4991 if (atomic_inc_not_zero(&exists->refs)) {
4992 spin_unlock(&mapping->private_lock);
4993 unlock_page(p);
4994 put_page(p);
4995 mark_extent_buffer_accessed(exists, p);
4996 goto free_eb;
4997 }
4998 exists = NULL;
4999
5000 /*
5001 * Do this so attach doesn't complain and we need to
5002 * drop the ref the old guy had.
5003 */
5004 ClearPagePrivate(p);
5005 WARN_ON(PageDirty(p));
5006 put_page(p);
5007 }
5008 attach_extent_buffer_page(eb, p);
5009 spin_unlock(&mapping->private_lock);
5010 WARN_ON(PageDirty(p));
5011 eb->pages[i] = p;
5012 if (!PageUptodate(p))
5013 uptodate = 0;
5014
5015 /*
5016 * see below about how we avoid a nasty race with release page
5017 * and why we unlock later
5018 */
5019 }
5020 if (uptodate)
5021 set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5022 again:
5023 ret = radix_tree_preload(GFP_NOFS);
5024 if (ret) {
5025 exists = ERR_PTR(ret);
5026 goto free_eb;
5027 }
5028
5029 spin_lock(&fs_info->buffer_lock);
5030 ret = radix_tree_insert(&fs_info->buffer_radix,
5031 start >> PAGE_SHIFT, eb);
5032 spin_unlock(&fs_info->buffer_lock);
5033 radix_tree_preload_end();
5034 if (ret == -EEXIST) {
5035 exists = find_extent_buffer(fs_info, start);
5036 if (exists)
5037 goto free_eb;
5038 else
5039 goto again;
5040 }
5041 /* add one reference for the tree */
5042 check_buffer_tree_ref(eb);
5043 set_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags);
5044
5045 /*
5046 * there is a race where release page may have
5047 * tried to find this extent buffer in the radix
5048 * but failed. It will tell the VM it is safe to
5049 * reclaim the, and it will clear the page private bit.
5050 * We must make sure to set the page private bit properly
5051 * after the extent buffer is in the radix tree so
5052 * it doesn't get lost
5053 */
5054 SetPageChecked(eb->pages[0]);
5055 for (i = 1; i < num_pages; i++) {
5056 p = eb->pages[i];
5057 ClearPageChecked(p);
5058 unlock_page(p);
5059 }
5060 unlock_page(eb->pages[0]);
5061 return eb;
5062
5063 free_eb:
5064 WARN_ON(!atomic_dec_and_test(&eb->refs));
5065 for (i = 0; i < num_pages; i++) {
5066 if (eb->pages[i])
5067 unlock_page(eb->pages[i]);
5068 }
5069
5070 btrfs_release_extent_buffer(eb);
5071 return exists;
5072 }
5073
5074 static inline void btrfs_release_extent_buffer_rcu(struct rcu_head *head)
5075 {
5076 struct extent_buffer *eb =
5077 container_of(head, struct extent_buffer, rcu_head);
5078
5079 __free_extent_buffer(eb);
5080 }
5081
5082 /* Expects to have eb->eb_lock already held */
5083 static int release_extent_buffer(struct extent_buffer *eb)
5084 {
5085 WARN_ON(atomic_read(&eb->refs) == 0);
5086 if (atomic_dec_and_test(&eb->refs)) {
5087 if (test_and_clear_bit(EXTENT_BUFFER_IN_TREE, &eb->bflags)) {
5088 struct btrfs_fs_info *fs_info = eb->fs_info;
5089
5090 spin_unlock(&eb->refs_lock);
5091
5092 spin_lock(&fs_info->buffer_lock);
5093 radix_tree_delete(&fs_info->buffer_radix,
5094 eb->start >> PAGE_SHIFT);
5095 spin_unlock(&fs_info->buffer_lock);
5096 } else {
5097 spin_unlock(&eb->refs_lock);
5098 }
5099
5100 /* Should be safe to release our pages at this point */
5101 btrfs_release_extent_buffer_page(eb);
5102 #ifdef CONFIG_BTRFS_FS_RUN_SANITY_TESTS
5103 if (unlikely(test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags))) {
5104 __free_extent_buffer(eb);
5105 return 1;
5106 }
5107 #endif
5108 call_rcu(&eb->rcu_head, btrfs_release_extent_buffer_rcu);
5109 return 1;
5110 }
5111 spin_unlock(&eb->refs_lock);
5112
5113 return 0;
5114 }
5115
5116 void free_extent_buffer(struct extent_buffer *eb)
5117 {
5118 int refs;
5119 int old;
5120 if (!eb)
5121 return;
5122
5123 while (1) {
5124 refs = atomic_read(&eb->refs);
5125 if (refs <= 3)
5126 break;
5127 old = atomic_cmpxchg(&eb->refs, refs, refs - 1);
5128 if (old == refs)
5129 return;
5130 }
5131
5132 spin_lock(&eb->refs_lock);
5133 if (atomic_read(&eb->refs) == 2 &&
5134 test_bit(EXTENT_BUFFER_DUMMY, &eb->bflags))
5135 atomic_dec(&eb->refs);
5136
5137 if (atomic_read(&eb->refs) == 2 &&
5138 test_bit(EXTENT_BUFFER_STALE, &eb->bflags) &&
5139 !extent_buffer_under_io(eb) &&
5140 test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
5141 atomic_dec(&eb->refs);
5142
5143 /*
5144 * I know this is terrible, but it's temporary until we stop tracking
5145 * the uptodate bits and such for the extent buffers.
5146 */
5147 release_extent_buffer(eb);
5148 }
5149
5150 void free_extent_buffer_stale(struct extent_buffer *eb)
5151 {
5152 if (!eb)
5153 return;
5154
5155 spin_lock(&eb->refs_lock);
5156 set_bit(EXTENT_BUFFER_STALE, &eb->bflags);
5157
5158 if (atomic_read(&eb->refs) == 2 && !extent_buffer_under_io(eb) &&
5159 test_and_clear_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
5160 atomic_dec(&eb->refs);
5161 release_extent_buffer(eb);
5162 }
5163
5164 void clear_extent_buffer_dirty(struct extent_buffer *eb)
5165 {
5166 unsigned long i;
5167 unsigned long num_pages;
5168 struct page *page;
5169
5170 num_pages = num_extent_pages(eb->start, eb->len);
5171
5172 for (i = 0; i < num_pages; i++) {
5173 page = eb->pages[i];
5174 if (!PageDirty(page))
5175 continue;
5176
5177 lock_page(page);
5178 WARN_ON(!PagePrivate(page));
5179
5180 clear_page_dirty_for_io(page);
5181 xa_lock_irq(&page->mapping->i_pages);
5182 if (!PageDirty(page)) {
5183 radix_tree_tag_clear(&page->mapping->i_pages,
5184 page_index(page),
5185 PAGECACHE_TAG_DIRTY);
5186 }
5187 xa_unlock_irq(&page->mapping->i_pages);
5188 ClearPageError(page);
5189 unlock_page(page);
5190 }
5191 WARN_ON(atomic_read(&eb->refs) == 0);
5192 }
5193
5194 int set_extent_buffer_dirty(struct extent_buffer *eb)
5195 {
5196 unsigned long i;
5197 unsigned long num_pages;
5198 int was_dirty = 0;
5199
5200 check_buffer_tree_ref(eb);
5201
5202 was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
5203
5204 num_pages = num_extent_pages(eb->start, eb->len);
5205 WARN_ON(atomic_read(&eb->refs) == 0);
5206 WARN_ON(!test_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags));
5207
5208 for (i = 0; i < num_pages; i++)
5209 set_page_dirty(eb->pages[i]);
5210 return was_dirty;
5211 }
5212
5213 void clear_extent_buffer_uptodate(struct extent_buffer *eb)
5214 {
5215 unsigned long i;
5216 struct page *page;
5217 unsigned long num_pages;
5218
5219 clear_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5220 num_pages = num_extent_pages(eb->start, eb->len);
5221 for (i = 0; i < num_pages; i++) {
5222 page = eb->pages[i];
5223 if (page)
5224 ClearPageUptodate(page);
5225 }
5226 }
5227
5228 void set_extent_buffer_uptodate(struct extent_buffer *eb)
5229 {
5230 unsigned long i;
5231 struct page *page;
5232 unsigned long num_pages;
5233
5234 set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5235 num_pages = num_extent_pages(eb->start, eb->len);
5236 for (i = 0; i < num_pages; i++) {
5237 page = eb->pages[i];
5238 SetPageUptodate(page);
5239 }
5240 }
5241
5242 int read_extent_buffer_pages(struct extent_io_tree *tree,
5243 struct extent_buffer *eb, int wait, int mirror_num)
5244 {
5245 unsigned long i;
5246 struct page *page;
5247 int err;
5248 int ret = 0;
5249 int locked_pages = 0;
5250 int all_uptodate = 1;
5251 unsigned long num_pages;
5252 unsigned long num_reads = 0;
5253 struct bio *bio = NULL;
5254 unsigned long bio_flags = 0;
5255
5256 if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags))
5257 return 0;
5258
5259 num_pages = num_extent_pages(eb->start, eb->len);
5260 for (i = 0; i < num_pages; i++) {
5261 page = eb->pages[i];
5262 if (wait == WAIT_NONE) {
5263 if (!trylock_page(page))
5264 goto unlock_exit;
5265 } else {
5266 lock_page(page);
5267 }
5268 locked_pages++;
5269 }
5270 /*
5271 * We need to firstly lock all pages to make sure that
5272 * the uptodate bit of our pages won't be affected by
5273 * clear_extent_buffer_uptodate().
5274 */
5275 for (i = 0; i < num_pages; i++) {
5276 page = eb->pages[i];
5277 if (!PageUptodate(page)) {
5278 num_reads++;
5279 all_uptodate = 0;
5280 }
5281 }
5282
5283 if (all_uptodate) {
5284 set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
5285 goto unlock_exit;
5286 }
5287
5288 clear_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags);
5289 eb->read_mirror = 0;
5290 atomic_set(&eb->io_pages, num_reads);
5291 for (i = 0; i < num_pages; i++) {
5292 page = eb->pages[i];
5293
5294 if (!PageUptodate(page)) {
5295 if (ret) {
5296 atomic_dec(&eb->io_pages);
5297 unlock_page(page);
5298 continue;
5299 }
5300
5301 ClearPageError(page);
5302 err = __extent_read_full_page(tree, page,
5303 btree_get_extent, &bio,
5304 mirror_num, &bio_flags,
5305 REQ_META);
5306 if (err) {
5307 ret = err;
5308 /*
5309 * We use &bio in above __extent_read_full_page,
5310 * so we ensure that if it returns error, the
5311 * current page fails to add itself to bio and
5312 * it's been unlocked.
5313 *
5314 * We must dec io_pages by ourselves.
5315 */
5316 atomic_dec(&eb->io_pages);
5317 }
5318 } else {
5319 unlock_page(page);
5320 }
5321 }
5322
5323 if (bio) {
5324 err = submit_one_bio(bio, mirror_num, bio_flags);
5325 if (err)
5326 return err;
5327 }
5328
5329 if (ret || wait != WAIT_COMPLETE)
5330 return ret;
5331
5332 for (i = 0; i < num_pages; i++) {
5333 page = eb->pages[i];
5334 wait_on_page_locked(page);
5335 if (!PageUptodate(page))
5336 ret = -EIO;
5337 }
5338
5339 return ret;
5340
5341 unlock_exit:
5342 while (locked_pages > 0) {
5343 locked_pages--;
5344 page = eb->pages[locked_pages];
5345 unlock_page(page);
5346 }
5347 return ret;
5348 }
5349
5350 void read_extent_buffer(const struct extent_buffer *eb, void *dstv,
5351 unsigned long start, unsigned long len)
5352 {
5353 size_t cur;
5354 size_t offset;
5355 struct page *page;
5356 char *kaddr;
5357 char *dst = (char *)dstv;
5358 size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
5359 unsigned long i = (start_offset + start) >> PAGE_SHIFT;
5360
5361 if (start + len > eb->len) {
5362 WARN(1, KERN_ERR "btrfs bad mapping eb start %llu len %lu, wanted %lu %lu\n",
5363 eb->start, eb->len, start, len);
5364 memset(dst, 0, len);
5365 return;
5366 }
5367
5368 offset = (start_offset + start) & (PAGE_SIZE - 1);
5369
5370 while (len > 0) {
5371 page = eb->pages[i];
5372
5373 cur = min(len, (PAGE_SIZE - offset));
5374 kaddr = page_address(page);
5375 memcpy(dst, kaddr + offset, cur);
5376
5377 dst += cur;
5378 len -= cur;
5379 offset = 0;
5380 i++;
5381 }
5382 }
5383
5384 int read_extent_buffer_to_user(const struct extent_buffer *eb,
5385 void __user *dstv,
5386 unsigned long start, unsigned long len)
5387 {
5388 size_t cur;
5389 size_t offset;
5390 struct page *page;
5391 char *kaddr;
5392 char __user *dst = (char __user *)dstv;
5393 size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
5394 unsigned long i = (start_offset + start) >> PAGE_SHIFT;
5395 int ret = 0;
5396
5397 WARN_ON(start > eb->len);
5398 WARN_ON(start + len > eb->start + eb->len);
5399
5400 offset = (start_offset + start) & (PAGE_SIZE - 1);
5401
5402 while (len > 0) {
5403 page = eb->pages[i];
5404
5405 cur = min(len, (PAGE_SIZE - offset));
5406 kaddr = page_address(page);
5407 if (copy_to_user(dst, kaddr + offset, cur)) {
5408 ret = -EFAULT;
5409 break;
5410 }
5411
5412 dst += cur;
5413 len -= cur;
5414 offset = 0;
5415 i++;
5416 }
5417
5418 return ret;
5419 }
5420
5421 /*
5422 * return 0 if the item is found within a page.
5423 * return 1 if the item spans two pages.
5424 * return -EINVAL otherwise.
5425 */
5426 int map_private_extent_buffer(const struct extent_buffer *eb,
5427 unsigned long start, unsigned long min_len,
5428 char **map, unsigned long *map_start,
5429 unsigned long *map_len)
5430 {
5431 size_t offset = start & (PAGE_SIZE - 1);
5432 char *kaddr;
5433 struct page *p;
5434 size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
5435 unsigned long i = (start_offset + start) >> PAGE_SHIFT;
5436 unsigned long end_i = (start_offset + start + min_len - 1) >>
5437 PAGE_SHIFT;
5438
5439 if (start + min_len > eb->len) {
5440 WARN(1, KERN_ERR "btrfs bad mapping eb start %llu len %lu, wanted %lu %lu\n",
5441 eb->start, eb->len, start, min_len);
5442 return -EINVAL;
5443 }
5444
5445 if (i != end_i)
5446 return 1;
5447
5448 if (i == 0) {
5449 offset = start_offset;
5450 *map_start = 0;
5451 } else {
5452 offset = 0;
5453 *map_start = ((u64)i << PAGE_SHIFT) - start_offset;
5454 }
5455
5456 p = eb->pages[i];
5457 kaddr = page_address(p);
5458 *map = kaddr + offset;
5459 *map_len = PAGE_SIZE - offset;
5460 return 0;
5461 }
5462
5463 int memcmp_extent_buffer(const struct extent_buffer *eb, const void *ptrv,
5464 unsigned long start, unsigned long len)
5465 {
5466 size_t cur;
5467 size_t offset;
5468 struct page *page;
5469 char *kaddr;
5470 char *ptr = (char *)ptrv;
5471 size_t start_offset = eb->start & ((u64)PAGE_SIZE - 1);
5472 unsigned long i = (start_offset + start) >> PAGE_SHIFT;
5473 int ret = 0;
5474
5475 WARN_ON(start > eb->len);
5476 WARN_ON(start + len > eb->start + eb->len);
5477
5478 offset = (start_offset + start) & (PAGE_SIZE - 1);
5479
5480 while (len > 0) {
5481 page = eb->pages[i];
5482
5483 cur = min(len, (PAGE_SIZE - offset));
5484
5485 kaddr = page_address(page);
5486 ret = memcmp(ptr, kaddr + offset, cur);
5487 if (ret)
5488 break;
5489
5490 ptr += cur;
5491 len -= cur;
5492 offset = 0;
5493 i++;
5494 }
5495 return ret;
5496 }
5497
5498 void write_extent_buffer_chunk_tree_uuid(struct extent_buffer *eb,
5499 const void *srcv)
5500 {
5501 char *kaddr;
5502
5503 WARN_ON(!PageUptodate(eb->pages[0]));
5504 kaddr = page_address(eb->pages[0]);
5505 memcpy(kaddr + offsetof(struct btrfs_header, chunk_tree_uuid), srcv,
5506 BTRFS_FSID_SIZE);
5507 }
5508
5509 void write_extent_buffer_fsid(struct extent_buffer *eb, const void *srcv)
5510 {
5511 char *kaddr;
5512
5513 WARN_ON(!PageUptodate(eb->pages[0]));
5514 kaddr = page_address(eb->pages[0]);
5515 memcpy(kaddr + offsetof(struct btrfs_header, fsid), srcv,
5516 BTRFS_FSID_SIZE);
5517 }
5518
5519 void write_extent_buffer(struct extent_buffer *eb, const void *srcv,
5520 unsigned long start, unsigned long len)
5521 {
5522 size_t cur;
5523 size_t offset;
5524 struct page *page;
5525 char *kaddr;