?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
writeback: split writeback_inodes_wb
[linux-2.6/next.git] / fs / btrfs / extent_io.c
blobd74e6af9b53a598a26bf83357f1bed61d2103ef2
1 #include <linux/bitops.h>
2 #include <linux/slab.h>
3 #include <linux/bio.h>
4 #include <linux/mm.h>
5 #include <linux/pagemap.h>
6 #include <linux/page-flags.h>
7 #include <linux/module.h>
8 #include <linux/spinlock.h>
9 #include <linux/blkdev.h>
10 #include <linux/swap.h>
11 #include <linux/writeback.h>
12 #include <linux/pagevec.h>
13 #include "extent_io.h"
14 #include "extent_map.h"
15 #include "compat.h"
16 #include "ctree.h"
17 #include "btrfs_inode.h"
18
19 static struct kmem_cache *extent_state_cache;
20 static struct kmem_cache *extent_buffer_cache;
21
22 static LIST_HEAD(buffers);
23 static LIST_HEAD(states);
24
25 #define LEAK_DEBUG 0
26 #if LEAK_DEBUG
27 static DEFINE_SPINLOCK(leak_lock);
28 #endif
29
30 #define BUFFER_LRU_MAX 64
31
32 struct tree_entry {
33 u64 start;
34 u64 end;
35 struct rb_node rb_node;
36 };
37
38 struct extent_page_data {
39 struct bio *bio;
40 struct extent_io_tree *tree;
41 get_extent_t *get_extent;
42
43 /* tells writepage not to lock the state bits for this range
44 * it still does the unlocking
45 */
46 unsigned int extent_locked:1;
47
48 /* tells the submit_bio code to use a WRITE_SYNC */
49 unsigned int sync_io:1;
50 };
51
52 int __init extent_io_init(void)
53 {
54 extent_state_cache = kmem_cache_create("extent_state",
55 sizeof(struct extent_state), 0,
56 SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
57 if (!extent_state_cache)
58 return -ENOMEM;
59
60 extent_buffer_cache = kmem_cache_create("extent_buffers",
61 sizeof(struct extent_buffer), 0,
62 SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
63 if (!extent_buffer_cache)
64 goto free_state_cache;
65 return 0;
66
67 free_state_cache:
68 kmem_cache_destroy(extent_state_cache);
69 return -ENOMEM;
70 }
71
72 void extent_io_exit(void)
73 {
74 struct extent_state *state;
75 struct extent_buffer *eb;
76
77 while (!list_empty(&states)) {
78 state = list_entry(states.next, struct extent_state, leak_list);
79 printk(KERN_ERR "btrfs state leak: start %llu end %llu "
80 "state %lu in tree %p refs %d\n",
81 (unsigned long long)state->start,
82 (unsigned long long)state->end,
83 state->state, state->tree, atomic_read(&state->refs));
84 list_del(&state->leak_list);
85 kmem_cache_free(extent_state_cache, state);
86
87 }
88
89 while (!list_empty(&buffers)) {
90 eb = list_entry(buffers.next, struct extent_buffer, leak_list);
91 printk(KERN_ERR "btrfs buffer leak start %llu len %lu "
92 "refs %d\n", (unsigned long long)eb->start,
93 eb->len, atomic_read(&eb->refs));
94 list_del(&eb->leak_list);
95 kmem_cache_free(extent_buffer_cache, eb);
96 }
97 if (extent_state_cache)
98 kmem_cache_destroy(extent_state_cache);
99 if (extent_buffer_cache)
100 kmem_cache_destroy(extent_buffer_cache);
101 }
102
103 void extent_io_tree_init(struct extent_io_tree *tree,
104 struct address_space *mapping, gfp_t mask)
105 {
106 tree->state = RB_ROOT;
107 tree->buffer = RB_ROOT;
108 tree->ops = NULL;
109 tree->dirty_bytes = 0;
110 spin_lock_init(&tree->lock);
111 spin_lock_init(&tree->buffer_lock);
112 tree->mapping = mapping;
113 }
114
115 static struct extent_state *alloc_extent_state(gfp_t mask)
116 {
117 struct extent_state *state;
118 #if LEAK_DEBUG
119 unsigned long flags;
120 #endif
121
122 state = kmem_cache_alloc(extent_state_cache, mask);
123 if (!state)
124 return state;
125 state->state = 0;
126 state->private = 0;
127 state->tree = NULL;
128 #if LEAK_DEBUG
129 spin_lock_irqsave(&leak_lock, flags);
130 list_add(&state->leak_list, &states);
131 spin_unlock_irqrestore(&leak_lock, flags);
132 #endif
133 atomic_set(&state->refs, 1);
134 init_waitqueue_head(&state->wq);
135 return state;
136 }
137
138 void free_extent_state(struct extent_state *state)
139 {
140 if (!state)
141 return;
142 if (atomic_dec_and_test(&state->refs)) {
143 #if LEAK_DEBUG
144 unsigned long flags;
145 #endif
146 WARN_ON(state->tree);
147 #if LEAK_DEBUG
148 spin_lock_irqsave(&leak_lock, flags);
149 list_del(&state->leak_list);
150 spin_unlock_irqrestore(&leak_lock, flags);
151 #endif
152 kmem_cache_free(extent_state_cache, state);
153 }
154 }
155
156 static struct rb_node *tree_insert(struct rb_root *root, u64 offset,
157 struct rb_node *node)
158 {
159 struct rb_node **p = &root->rb_node;
160 struct rb_node *parent = NULL;
161 struct tree_entry *entry;
162
163 while (*p) {
164 parent = *p;
165 entry = rb_entry(parent, struct tree_entry, rb_node);
166
167 if (offset < entry->start)
168 p = &(*p)->rb_left;
169 else if (offset > entry->end)
170 p = &(*p)->rb_right;
171 else
172 return parent;
173 }
174
175 entry = rb_entry(node, struct tree_entry, rb_node);
176 rb_link_node(node, parent, p);
177 rb_insert_color(node, root);
178 return NULL;
179 }
180
181 static struct rb_node *__etree_search(struct extent_io_tree *tree, u64 offset,
182 struct rb_node **prev_ret,
183 struct rb_node **next_ret)
184 {
185 struct rb_root *root = &tree->state;
186 struct rb_node *n = root->rb_node;
187 struct rb_node *prev = NULL;
188 struct rb_node *orig_prev = NULL;
189 struct tree_entry *entry;
190 struct tree_entry *prev_entry = NULL;
191
192 while (n) {
193 entry = rb_entry(n, struct tree_entry, rb_node);
194 prev = n;
195 prev_entry = entry;
196
197 if (offset < entry->start)
198 n = n->rb_left;
199 else if (offset > entry->end)
200 n = n->rb_right;
201 else
202 return n;
203 }
204
205 if (prev_ret) {
206 orig_prev = prev;
207 while (prev && offset > prev_entry->end) {
208 prev = rb_next(prev);
209 prev_entry = rb_entry(prev, struct tree_entry, rb_node);
210 }
211 *prev_ret = prev;
212 prev = orig_prev;
213 }
214
215 if (next_ret) {
216 prev_entry = rb_entry(prev, struct tree_entry, rb_node);
217 while (prev && offset < prev_entry->start) {
218 prev = rb_prev(prev);
219 prev_entry = rb_entry(prev, struct tree_entry, rb_node);
220 }
221 *next_ret = prev;
222 }
223 return NULL;
224 }
225
226 static inline struct rb_node *tree_search(struct extent_io_tree *tree,
227 u64 offset)
228 {
229 struct rb_node *prev = NULL;
230 struct rb_node *ret;
231
232 ret = __etree_search(tree, offset, &prev, NULL);
233 if (!ret)
234 return prev;
235 return ret;
236 }
237
238 static struct extent_buffer *buffer_tree_insert(struct extent_io_tree *tree,
239 u64 offset, struct rb_node *node)
240 {
241 struct rb_root *root = &tree->buffer;
242 struct rb_node **p = &root->rb_node;
243 struct rb_node *parent = NULL;
244 struct extent_buffer *eb;
245
246 while (*p) {
247 parent = *p;
248 eb = rb_entry(parent, struct extent_buffer, rb_node);
249
250 if (offset < eb->start)
251 p = &(*p)->rb_left;
252 else if (offset > eb->start)
253 p = &(*p)->rb_right;
254 else
255 return eb;
256 }
257
258 rb_link_node(node, parent, p);
259 rb_insert_color(node, root);
260 return NULL;
261 }
262
263 static struct extent_buffer *buffer_search(struct extent_io_tree *tree,
264 u64 offset)
265 {
266 struct rb_root *root = &tree->buffer;
267 struct rb_node *n = root->rb_node;
268 struct extent_buffer *eb;
269
270 while (n) {
271 eb = rb_entry(n, struct extent_buffer, rb_node);
272 if (offset < eb->start)
273 n = n->rb_left;
274 else if (offset > eb->start)
275 n = n->rb_right;
276 else
277 return eb;
278 }
279 return NULL;
280 }
281
282 static void merge_cb(struct extent_io_tree *tree, struct extent_state *new,
283 struct extent_state *other)
284 {
285 if (tree->ops && tree->ops->merge_extent_hook)
286 tree->ops->merge_extent_hook(tree->mapping->host, new,
287 other);
288 }
289
290 /*
291 * utility function to look for merge candidates inside a given range.
292 * Any extents with matching state are merged together into a single
293 * extent in the tree. Extents with EXTENT_IO in their state field
294 * are not merged because the end_io handlers need to be able to do
295 * operations on them without sleeping (or doing allocations/splits).
296 *
297 * This should be called with the tree lock held.
298 */
299 static int merge_state(struct extent_io_tree *tree,
300 struct extent_state *state)
301 {
302 struct extent_state *other;
303 struct rb_node *other_node;
304
305 if (state->state & (EXTENT_IOBITS | EXTENT_BOUNDARY))
306 return 0;
307
308 other_node = rb_prev(&state->rb_node);
309 if (other_node) {
310 other = rb_entry(other_node, struct extent_state, rb_node);
311 if (other->end == state->start - 1 &&
312 other->state == state->state) {
313 merge_cb(tree, state, other);
314 state->start = other->start;
315 other->tree = NULL;
316 rb_erase(&other->rb_node, &tree->state);
317 free_extent_state(other);
318 }
319 }
320 other_node = rb_next(&state->rb_node);
321 if (other_node) {
322 other = rb_entry(other_node, struct extent_state, rb_node);
323 if (other->start == state->end + 1 &&
324 other->state == state->state) {
325 merge_cb(tree, state, other);
326 other->start = state->start;
327 state->tree = NULL;
328 rb_erase(&state->rb_node, &tree->state);
329 free_extent_state(state);
330 state = NULL;
331 }
332 }
333
334 return 0;
335 }
336
337 static int set_state_cb(struct extent_io_tree *tree,
338 struct extent_state *state, int *bits)
339 {
340 if (tree->ops && tree->ops->set_bit_hook) {
341 return tree->ops->set_bit_hook(tree->mapping->host,
342 state, bits);
343 }
344
345 return 0;
346 }
347
348 static void clear_state_cb(struct extent_io_tree *tree,
349 struct extent_state *state, int *bits)
350 {
351 if (tree->ops && tree->ops->clear_bit_hook)
352 tree->ops->clear_bit_hook(tree->mapping->host, state, bits);
353 }
354
355 /*
356 * insert an extent_state struct into the tree. 'bits' are set on the
357 * struct before it is inserted.
358 *
359 * This may return -EEXIST if the extent is already there, in which case the
360 * state struct is freed.
361 *
362 * The tree lock is not taken internally. This is a utility function and
363 * probably isn't what you want to call (see set/clear_extent_bit).
364 */
365 static int insert_state(struct extent_io_tree *tree,
366 struct extent_state *state, u64 start, u64 end,
367 int *bits)
368 {
369 struct rb_node *node;
370 int bits_to_set = *bits & ~EXTENT_CTLBITS;
371 int ret;
372
373 if (end < start) {
374 printk(KERN_ERR "btrfs end < start %llu %llu\n",
375 (unsigned long long)end,
376 (unsigned long long)start);
377 WARN_ON(1);
378 }
379 state->start = start;
380 state->end = end;
381 ret = set_state_cb(tree, state, bits);
382 if (ret)
383 return ret;
384
385 if (bits_to_set & EXTENT_DIRTY)
386 tree->dirty_bytes += end - start + 1;
387 state->state |= bits_to_set;
388 node = tree_insert(&tree->state, end, &state->rb_node);
389 if (node) {
390 struct extent_state *found;
391 found = rb_entry(node, struct extent_state, rb_node);
392 printk(KERN_ERR "btrfs found node %llu %llu on insert of "
393 "%llu %llu\n", (unsigned long long)found->start,
394 (unsigned long long)found->end,
395 (unsigned long long)start, (unsigned long long)end);
396 free_extent_state(state);
397 return -EEXIST;
398 }
399 state->tree = tree;
400 merge_state(tree, state);
401 return 0;
402 }
403
404 static int split_cb(struct extent_io_tree *tree, struct extent_state *orig,
405 u64 split)
406 {
407 if (tree->ops && tree->ops->split_extent_hook)
408 return tree->ops->split_extent_hook(tree->mapping->host,
409 orig, split);
410 return 0;
411 }
412
413 /*
414 * split a given extent state struct in two, inserting the preallocated
415 * struct 'prealloc' as the newly created second half. 'split' indicates an
416 * offset inside 'orig' where it should be split.
417 *
418 * Before calling,
419 * the tree has 'orig' at [orig->start, orig->end]. After calling, there
420 * are two extent state structs in the tree:
421 * prealloc: [orig->start, split - 1]
422 * orig: [ split, orig->end ]
423 *
424 * The tree locks are not taken by this function. They need to be held
425 * by the caller.
426 */
427 static int split_state(struct extent_io_tree *tree, struct extent_state *orig,
428 struct extent_state *prealloc, u64 split)
429 {
430 struct rb_node *node;
431
432 split_cb(tree, orig, split);
433
434 prealloc->start = orig->start;
435 prealloc->end = split - 1;
436 prealloc->state = orig->state;
437 orig->start = split;
438
439 node = tree_insert(&tree->state, prealloc->end, &prealloc->rb_node);
440 if (node) {
441 free_extent_state(prealloc);
442 return -EEXIST;
443 }
444 prealloc->tree = tree;
445 return 0;
446 }
447
448 /*
449 * utility function to clear some bits in an extent state struct.
450 * it will optionally wake up any one waiting on this state (wake == 1), or
451 * forcibly remove the state from the tree (delete == 1).
452 *
453 * If no bits are set on the state struct after clearing things, the
454 * struct is freed and removed from the tree
455 */
456 static int clear_state_bit(struct extent_io_tree *tree,
457 struct extent_state *state,
458 int *bits, int wake)
459 {
460 int bits_to_clear = *bits & ~EXTENT_CTLBITS;
461 int ret = state->state & bits_to_clear;
462
463 if ((bits_to_clear & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) {
464 u64 range = state->end - state->start + 1;
465 WARN_ON(range > tree->dirty_bytes);
466 tree->dirty_bytes -= range;
467 }
468 clear_state_cb(tree, state, bits);
469 state->state &= ~bits_to_clear;
470 if (wake)
471 wake_up(&state->wq);
472 if (state->state == 0) {
473 if (state->tree) {
474 rb_erase(&state->rb_node, &tree->state);
475 state->tree = NULL;
476 free_extent_state(state);
477 } else {
478 WARN_ON(1);
479 }
480 } else {
481 merge_state(tree, state);
482 }
483 return ret;
484 }
485
486 /*
487 * clear some bits on a range in the tree. This may require splitting
488 * or inserting elements in the tree, so the gfp mask is used to
489 * indicate which allocations or sleeping are allowed.
490 *
491 * pass 'wake' == 1 to kick any sleepers, and 'delete' == 1 to remove
492 * the given range from the tree regardless of state (ie for truncate).
493 *
494 * the range [start, end] is inclusive.
495 *
496 * This takes the tree lock, and returns < 0 on error, > 0 if any of the
497 * bits were already set, or zero if none of the bits were already set.
498 */
499 int clear_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
500 int bits, int wake, int delete,
501 struct extent_state **cached_state,
502 gfp_t mask)
503 {
504 struct extent_state *state;
505 struct extent_state *cached;
506 struct extent_state *prealloc = NULL;
507 struct rb_node *next_node;
508 struct rb_node *node;
509 u64 last_end;
510 int err;
511 int set = 0;
512 int clear = 0;
513
514 if (delete)
515 bits |= ~EXTENT_CTLBITS;
516 bits |= EXTENT_FIRST_DELALLOC;
517
518 if (bits & (EXTENT_IOBITS | EXTENT_BOUNDARY))
519 clear = 1;
520 again:
521 if (!prealloc && (mask & __GFP_WAIT)) {
522 prealloc = alloc_extent_state(mask);
523 if (!prealloc)
524 return -ENOMEM;
525 }
526
527 spin_lock(&tree->lock);
528 if (cached_state) {
529 cached = *cached_state;
530
531 if (clear) {
532 *cached_state = NULL;
533 cached_state = NULL;
534 }
535
536 if (cached && cached->tree && cached->start == start) {
537 if (clear)
538 atomic_dec(&cached->refs);
539 state = cached;
540 goto hit_next;
541 }
542 if (clear)
543 free_extent_state(cached);
544 }
545 /*
546 * this search will find the extents that end after
547 * our range starts
548 */
549 node = tree_search(tree, start);
550 if (!node)
551 goto out;
552 state = rb_entry(node, struct extent_state, rb_node);
553 hit_next:
554 if (state->start > end)
555 goto out;
556 WARN_ON(state->end < start);
557 last_end = state->end;
558
559 /*
560 * | ---- desired range ---- |
561 * | state | or
562 * | ------------- state -------------- |
563 *
564 * We need to split the extent we found, and may flip
565 * bits on second half.
566 *
567 * If the extent we found extends past our range, we
568 * just split and search again. It'll get split again
569 * the next time though.
570 *
571 * If the extent we found is inside our range, we clear
572 * the desired bit on it.
573 */
574
575 if (state->start < start) {
576 if (!prealloc)
577 prealloc = alloc_extent_state(GFP_ATOMIC);
578 err = split_state(tree, state, prealloc, start);
579 BUG_ON(err == -EEXIST);
580 prealloc = NULL;
581 if (err)
582 goto out;
583 if (state->end <= end) {
584 set |= clear_state_bit(tree, state, &bits, wake);
585 if (last_end == (u64)-1)
586 goto out;
587 start = last_end + 1;
588 }
589 goto search_again;
590 }
591 /*
592 * | ---- desired range ---- |
593 * | state |
594 * We need to split the extent, and clear the bit
595 * on the first half
596 */
597 if (state->start <= end && state->end > end) {
598 if (!prealloc)
599 prealloc = alloc_extent_state(GFP_ATOMIC);
600 err = split_state(tree, state, prealloc, end + 1);
601 BUG_ON(err == -EEXIST);
602 if (wake)
603 wake_up(&state->wq);
604
605 set |= clear_state_bit(tree, prealloc, &bits, wake);
606
607 prealloc = NULL;
608 goto out;
609 }
610
611 if (state->end < end && prealloc && !need_resched())
612 next_node = rb_next(&state->rb_node);
613 else
614 next_node = NULL;
615
616 set |= clear_state_bit(tree, state, &bits, wake);
617 if (last_end == (u64)-1)
618 goto out;
619 start = last_end + 1;
620 if (start <= end && next_node) {
621 state = rb_entry(next_node, struct extent_state,
622 rb_node);
623 if (state->start == start)
624 goto hit_next;
625 }
626 goto search_again;
627
628 out:
629 spin_unlock(&tree->lock);
630 if (prealloc)
631 free_extent_state(prealloc);
632
633 return set;
634
635 search_again:
636 if (start > end)
637 goto out;
638 spin_unlock(&tree->lock);
639 if (mask & __GFP_WAIT)
640 cond_resched();
641 goto again;
642 }
643
644 static int wait_on_state(struct extent_io_tree *tree,
645 struct extent_state *state)
646 __releases(tree->lock)
647 __acquires(tree->lock)
648 {
649 DEFINE_WAIT(wait);
650 prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE);
651 spin_unlock(&tree->lock);
652 schedule();
653 spin_lock(&tree->lock);
654 finish_wait(&state->wq, &wait);
655 return 0;
656 }
657
658 /*
659 * waits for one or more bits to clear on a range in the state tree.
660 * The range [start, end] is inclusive.
661 * The tree lock is taken by this function
662 */
663 int wait_extent_bit(struct extent_io_tree *tree, u64 start, u64 end, int bits)
664 {
665 struct extent_state *state;
666 struct rb_node *node;
667
668 spin_lock(&tree->lock);
669 again:
670 while (1) {
671 /*
672 * this search will find all the extents that end after
673 * our range starts
674 */
675 node = tree_search(tree, start);
676 if (!node)
677 break;
678
679 state = rb_entry(node, struct extent_state, rb_node);
680
681 if (state->start > end)
682 goto out;
683
684 if (state->state & bits) {
685 start = state->start;
686 atomic_inc(&state->refs);
687 wait_on_state(tree, state);
688 free_extent_state(state);
689 goto again;
690 }
691 start = state->end + 1;
692
693 if (start > end)
694 break;
695
696 if (need_resched()) {
697 spin_unlock(&tree->lock);
698 cond_resched();
699 spin_lock(&tree->lock);
700 }
701 }
702 out:
703 spin_unlock(&tree->lock);
704 return 0;
705 }
706
707 static int set_state_bits(struct extent_io_tree *tree,
708 struct extent_state *state,
709 int *bits)
710 {
711 int ret;
712 int bits_to_set = *bits & ~EXTENT_CTLBITS;
713
714 ret = set_state_cb(tree, state, bits);
715 if (ret)
716 return ret;
717 if ((bits_to_set & EXTENT_DIRTY) && !(state->state & EXTENT_DIRTY)) {
718 u64 range = state->end - state->start + 1;
719 tree->dirty_bytes += range;
720 }
721 state->state |= bits_to_set;
722
723 return 0;
724 }
725
726 static void cache_state(struct extent_state *state,
727 struct extent_state **cached_ptr)
728 {
729 if (cached_ptr && !(*cached_ptr)) {
730 if (state->state & (EXTENT_IOBITS | EXTENT_BOUNDARY)) {
731 *cached_ptr = state;
732 atomic_inc(&state->refs);
733 }
734 }
735 }
736
737 /*
738 * set some bits on a range in the tree. This may require allocations or
739 * sleeping, so the gfp mask is used to indicate what is allowed.
740 *
741 * If any of the exclusive bits are set, this will fail with -EEXIST if some
742 * part of the range already has the desired bits set. The start of the
743 * existing range is returned in failed_start in this case.
744 *
745 * [start, end] is inclusive This takes the tree lock.
746 */
747
748 int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
749 int bits, int exclusive_bits, u64 *failed_start,
750 struct extent_state **cached_state, gfp_t mask)
751 {
752 struct extent_state *state;
753 struct extent_state *prealloc = NULL;
754 struct rb_node *node;
755 int err = 0;
756 u64 last_start;
757 u64 last_end;
758
759 bits |= EXTENT_FIRST_DELALLOC;
760 again:
761 if (!prealloc && (mask & __GFP_WAIT)) {
762 prealloc = alloc_extent_state(mask);
763 if (!prealloc)
764 return -ENOMEM;
765 }
766
767 spin_lock(&tree->lock);
768 if (cached_state && *cached_state) {
769 state = *cached_state;
770 if (state->start == start && state->tree) {
771 node = &state->rb_node;
772 goto hit_next;
773 }
774 }
775 /*
776 * this search will find all the extents that end after
777 * our range starts.
778 */
779 node = tree_search(tree, start);
780 if (!node) {
781 err = insert_state(tree, prealloc, start, end, &bits);
782 prealloc = NULL;
783 BUG_ON(err == -EEXIST);
784 goto out;
785 }
786 state = rb_entry(node, struct extent_state, rb_node);
787 hit_next:
788 last_start = state->start;
789 last_end = state->end;
790
791 /*
792 * | ---- desired range ---- |
793 * | state |
794 *
795 * Just lock what we found and keep going
796 */
797 if (state->start == start && state->end <= end) {
798 struct rb_node *next_node;
799 if (state->state & exclusive_bits) {
800 *failed_start = state->start;
801 err = -EEXIST;
802 goto out;
803 }
804
805 err = set_state_bits(tree, state, &bits);
806 if (err)
807 goto out;
808
809 cache_state(state, cached_state);
810 merge_state(tree, state);
811 if (last_end == (u64)-1)
812 goto out;
813
814 start = last_end + 1;
815 if (start < end && prealloc && !need_resched()) {
816 next_node = rb_next(node);
817 if (next_node) {
818 state = rb_entry(next_node, struct extent_state,
819 rb_node);
820 if (state->start == start)
821 goto hit_next;
822 }
823 }
824 goto search_again;
825 }
826
827 /*
828 * | ---- desired range ---- |
829 * | state |
830 * or
831 * | ------------- state -------------- |
832 *
833 * We need to split the extent we found, and may flip bits on
834 * second half.
835 *
836 * If the extent we found extends past our
837 * range, we just split and search again. It'll get split
838 * again the next time though.
839 *
840 * If the extent we found is inside our range, we set the
841 * desired bit on it.
842 */
843 if (state->start < start) {
844 if (state->state & exclusive_bits) {
845 *failed_start = start;
846 err = -EEXIST;
847 goto out;
848 }
849 err = split_state(tree, state, prealloc, start);
850 BUG_ON(err == -EEXIST);
851 prealloc = NULL;
852 if (err)
853 goto out;
854 if (state->end <= end) {
855 err = set_state_bits(tree, state, &bits);
856 if (err)
857 goto out;
858 cache_state(state, cached_state);
859 merge_state(tree, state);
860 if (last_end == (u64)-1)
861 goto out;
862 start = last_end + 1;
863 }
864 goto search_again;
865 }
866 /*
867 * | ---- desired range ---- |
868 * | state | or | state |
869 *
870 * There's a hole, we need to insert something in it and
871 * ignore the extent we found.
872 */
873 if (state->start > start) {
874 u64 this_end;
875 if (end < last_start)
876 this_end = end;
877 else
878 this_end = last_start - 1;
879 err = insert_state(tree, prealloc, start, this_end,
880 &bits);
881 BUG_ON(err == -EEXIST);
882 if (err) {
883 prealloc = NULL;
884 goto out;
885 }
886 cache_state(prealloc, cached_state);
887 prealloc = NULL;
888 start = this_end + 1;
889 goto search_again;
890 }
891 /*
892 * | ---- desired range ---- |
893 * | state |
894 * We need to split the extent, and set the bit
895 * on the first half
896 */
897 if (state->start <= end && state->end > end) {
898 if (state->state & exclusive_bits) {
899 *failed_start = start;
900 err = -EEXIST;
901 goto out;
902 }
903 err = split_state(tree, state, prealloc, end + 1);
904 BUG_ON(err == -EEXIST);
905
906 err = set_state_bits(tree, prealloc, &bits);
907 if (err) {
908 prealloc = NULL;
909 goto out;
910 }
911 cache_state(prealloc, cached_state);
912 merge_state(tree, prealloc);
913 prealloc = NULL;
914 goto out;
915 }
916
917 goto search_again;
918
919 out:
920 spin_unlock(&tree->lock);
921 if (prealloc)
922 free_extent_state(prealloc);
923
924 return err;
925
926 search_again:
927 if (start > end)
928 goto out;
929 spin_unlock(&tree->lock);
930 if (mask & __GFP_WAIT)
931 cond_resched();
932 goto again;
933 }
934
935 /* wrappers around set/clear extent bit */
936 int set_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
937 gfp_t mask)
938 {
939 return set_extent_bit(tree, start, end, EXTENT_DIRTY, 0, NULL,
940 NULL, mask);
941 }
942
943 int set_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
944 int bits, gfp_t mask)
945 {
946 return set_extent_bit(tree, start, end, bits, 0, NULL,
947 NULL, mask);
948 }
949
950 int clear_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
951 int bits, gfp_t mask)
952 {
953 return clear_extent_bit(tree, start, end, bits, 0, 0, NULL, mask);
954 }
955
956 int set_extent_delalloc(struct extent_io_tree *tree, u64 start, u64 end,
957 struct extent_state **cached_state, gfp_t mask)
958 {
959 return set_extent_bit(tree, start, end,
960 EXTENT_DELALLOC | EXTENT_DIRTY | EXTENT_UPTODATE,
961 0, NULL, cached_state, mask);
962 }
963
964 int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
965 gfp_t mask)
966 {
967 return clear_extent_bit(tree, start, end,
968 EXTENT_DIRTY | EXTENT_DELALLOC |
969 EXTENT_DO_ACCOUNTING, 0, 0, NULL, mask);
970 }
971
972 int set_extent_new(struct extent_io_tree *tree, u64 start, u64 end,
973 gfp_t mask)
974 {
975 return set_extent_bit(tree, start, end, EXTENT_NEW, 0, NULL,
976 NULL, mask);
977 }
978
979 static int clear_extent_new(struct extent_io_tree *tree, u64 start, u64 end,
980 gfp_t mask)
981 {
982 return clear_extent_bit(tree, start, end, EXTENT_NEW, 0, 0,
983 NULL, mask);
984 }
985
986 int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end,
987 gfp_t mask)
988 {
989 return set_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, NULL,
990 NULL, mask);
991 }
992
993 static int clear_extent_uptodate(struct extent_io_tree *tree, u64 start,
994 u64 end, struct extent_state **cached_state,
995 gfp_t mask)
996 {
997 return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0,
998 cached_state, mask);
999 }
1000
1001 int wait_on_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end)
1002 {
1003 return wait_extent_bit(tree, start, end, EXTENT_WRITEBACK);
1004 }
1005
1006 /*
1007 * either insert or lock state struct between start and end use mask to tell
1008 * us if waiting is desired.
1009 */
1010 int lock_extent_bits(struct extent_io_tree *tree, u64 start, u64 end,
1011 int bits, struct extent_state **cached_state, gfp_t mask)
1012 {
1013 int err;
1014 u64 failed_start;
1015 while (1) {
1016 err = set_extent_bit(tree, start, end, EXTENT_LOCKED | bits,
1017 EXTENT_LOCKED, &failed_start,
1018 cached_state, mask);
1019 if (err == -EEXIST && (mask & __GFP_WAIT)) {
1020 wait_extent_bit(tree, failed_start, end, EXTENT_LOCKED);
1021 start = failed_start;
1022 } else {
1023 break;
1024 }
1025 WARN_ON(start > end);
1026 }
1027 return err;
1028 }
1029
1030 int lock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask)
1031 {
1032 return lock_extent_bits(tree, start, end, 0, NULL, mask);
1033 }
1034
1035 int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end,
1036 gfp_t mask)
1037 {
1038 int err;
1039 u64 failed_start;
1040
1041 err = set_extent_bit(tree, start, end, EXTENT_LOCKED, EXTENT_LOCKED,
1042 &failed_start, NULL, mask);
1043 if (err == -EEXIST) {
1044 if (failed_start > start)
1045 clear_extent_bit(tree, start, failed_start - 1,
1046 EXTENT_LOCKED, 1, 0, NULL, mask);
1047 return 0;
1048 }
1049 return 1;
1050 }
1051
1052 int unlock_extent_cached(struct extent_io_tree *tree, u64 start, u64 end,
1053 struct extent_state **cached, gfp_t mask)
1054 {
1055 return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, cached,
1056 mask);
1057 }
1058
1059 int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end,
1060 gfp_t mask)
1061 {
1062 return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL,
1063 mask);
1064 }
1065
1066 /*
1067 * helper function to set pages and extents in the tree dirty
1068 */
1069 int set_range_dirty(struct extent_io_tree *tree, u64 start, u64 end)
1070 {
1071 unsigned long index = start >> PAGE_CACHE_SHIFT;
1072 unsigned long end_index = end >> PAGE_CACHE_SHIFT;
1073 struct page *page;
1074
1075 while (index <= end_index) {
1076 page = find_get_page(tree->mapping, index);
1077 BUG_ON(!page);
1078 __set_page_dirty_nobuffers(page);
1079 page_cache_release(page);
1080 index++;
1081 }
1082 return 0;
1083 }
1084
1085 /*
1086 * helper function to set both pages and extents in the tree writeback
1087 */
1088 static int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end)
1089 {
1090 unsigned long index = start >> PAGE_CACHE_SHIFT;
1091 unsigned long end_index = end >> PAGE_CACHE_SHIFT;
1092 struct page *page;
1093
1094 while (index <= end_index) {
1095 page = find_get_page(tree->mapping, index);
1096 BUG_ON(!page);
1097 set_page_writeback(page);
1098 page_cache_release(page);
1099 index++;
1100 }
1101 return 0;
1102 }
1103
1104 /*
1105 * find the first offset in the io tree with 'bits' set. zero is
1106 * returned if we find something, and *start_ret and *end_ret are
1107 * set to reflect the state struct that was found.
1108 *
1109 * If nothing was found, 1 is returned, < 0 on error
1110 */
1111 int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
1112 u64 *start_ret, u64 *end_ret, int bits)
1113 {
1114 struct rb_node *node;
1115 struct extent_state *state;
1116 int ret = 1;
1117
1118 spin_lock(&tree->lock);
1119 /*
1120 * this search will find all the extents that end after
1121 * our range starts.
1122 */
1123 node = tree_search(tree, start);
1124 if (!node)
1125 goto out;
1126
1127 while (1) {
1128 state = rb_entry(node, struct extent_state, rb_node);
1129 if (state->end >= start && (state->state & bits)) {
1130 *start_ret = state->start;
1131 *end_ret = state->end;
1132 ret = 0;
1133 break;
1134 }
1135 node = rb_next(node);
1136 if (!node)
1137 break;
1138 }
1139 out:
1140 spin_unlock(&tree->lock);
1141 return ret;
1142 }
1143
1144 /* find the first state struct with 'bits' set after 'start', and
1145 * return it. tree->lock must be held. NULL will returned if
1146 * nothing was found after 'start'
1147 */
1148 struct extent_state *find_first_extent_bit_state(struct extent_io_tree *tree,
1149 u64 start, int bits)
1150 {
1151 struct rb_node *node;
1152 struct extent_state *state;
1153
1154 /*
1155 * this search will find all the extents that end after
1156 * our range starts.
1157 */
1158 node = tree_search(tree, start);
1159 if (!node)
1160 goto out;
1161
1162 while (1) {
1163 state = rb_entry(node, struct extent_state, rb_node);
1164 if (state->end >= start && (state->state & bits))
1165 return state;
1166
1167 node = rb_next(node);
1168 if (!node)
1169 break;
1170 }
1171 out:
1172 return NULL;
1173 }
1174
1175 /*
1176 * find a contiguous range of bytes in the file marked as delalloc, not
1177 * more than 'max_bytes'. start and end are used to return the range,
1178 *
1179 * 1 is returned if we find something, 0 if nothing was in the tree
1180 */
1181 static noinline u64 find_delalloc_range(struct extent_io_tree *tree,
1182 u64 *start, u64 *end, u64 max_bytes,
1183 struct extent_state **cached_state)
1184 {
1185 struct rb_node *node;
1186 struct extent_state *state;
1187 u64 cur_start = *start;
1188 u64 found = 0;
1189 u64 total_bytes = 0;
1190
1191 spin_lock(&tree->lock);
1192
1193 /*
1194 * this search will find all the extents that end after
1195 * our range starts.
1196 */
1197 node = tree_search(tree, cur_start);
1198 if (!node) {
1199 if (!found)
1200 *end = (u64)-1;
1201 goto out;
1202 }
1203
1204 while (1) {
1205 state = rb_entry(node, struct extent_state, rb_node);
1206 if (found && (state->start != cur_start ||
1207 (state->state & EXTENT_BOUNDARY))) {
1208 goto out;
1209 }
1210 if (!(state->state & EXTENT_DELALLOC)) {
1211 if (!found)
1212 *end = state->end;
1213 goto out;
1214 }
1215 if (!found) {
1216 *start = state->start;
1217 *cached_state = state;
1218 atomic_inc(&state->refs);
1219 }
1220 found++;
1221 *end = state->end;
1222 cur_start = state->end + 1;
1223 node = rb_next(node);
1224 if (!node)
1225 break;
1226 total_bytes += state->end - state->start + 1;
1227 if (total_bytes >= max_bytes)
1228 break;
1229 }
1230 out:
1231 spin_unlock(&tree->lock);
1232 return found;
1233 }
1234
1235 static noinline int __unlock_for_delalloc(struct inode *inode,
1236 struct page *locked_page,
1237 u64 start, u64 end)
1238 {
1239 int ret;
1240 struct page *pages[16];
1241 unsigned long index = start >> PAGE_CACHE_SHIFT;
1242 unsigned long end_index = end >> PAGE_CACHE_SHIFT;
1243 unsigned long nr_pages = end_index - index + 1;
1244 int i;
1245
1246 if (index == locked_page->index && end_index == index)
1247 return 0;
1248
1249 while (nr_pages > 0) {
1250 ret = find_get_pages_contig(inode->i_mapping, index,
1251 min_t(unsigned long, nr_pages,
1252 ARRAY_SIZE(pages)), pages);
1253 for (i = 0; i < ret; i++) {
1254 if (pages[i] != locked_page)
1255 unlock_page(pages[i]);
1256 page_cache_release(pages[i]);
1257 }
1258 nr_pages -= ret;
1259 index += ret;
1260 cond_resched();
1261 }
1262 return 0;
1263 }
1264
1265 static noinline int lock_delalloc_pages(struct inode *inode,
1266 struct page *locked_page,
1267 u64 delalloc_start,
1268 u64 delalloc_end)
1269 {
1270 unsigned long index = delalloc_start >> PAGE_CACHE_SHIFT;
1271 unsigned long start_index = index;
1272 unsigned long end_index = delalloc_end >> PAGE_CACHE_SHIFT;
1273 unsigned long pages_locked = 0;
1274 struct page *pages[16];
1275 unsigned long nrpages;
1276 int ret;
1277 int i;
1278
1279 /* the caller is responsible for locking the start index */
1280 if (index == locked_page->index && index == end_index)
1281 return 0;
1282
1283 /* skip the page at the start index */
1284 nrpages = end_index - index + 1;
1285 while (nrpages > 0) {
1286 ret = find_get_pages_contig(inode->i_mapping, index,
1287 min_t(unsigned long,
1288 nrpages, ARRAY_SIZE(pages)), pages);
1289 if (ret == 0) {
1290 ret = -EAGAIN;
1291 goto done;
1292 }
1293 /* now we have an array of pages, lock them all */
1294 for (i = 0; i < ret; i++) {
1295 /*
1296 * the caller is taking responsibility for
1297 * locked_page
1298 */
1299 if (pages[i] != locked_page) {
1300 lock_page(pages[i]);
1301 if (!PageDirty(pages[i]) ||
1302 pages[i]->mapping != inode->i_mapping) {
1303 ret = -EAGAIN;
1304 unlock_page(pages[i]);
1305 page_cache_release(pages[i]);
1306 goto done;
1307 }
1308 }
1309 page_cache_release(pages[i]);
1310 pages_locked++;
1311 }
1312 nrpages -= ret;
1313 index += ret;
1314 cond_resched();
1315 }
1316 ret = 0;
1317 done:
1318 if (ret && pages_locked) {
1319 __unlock_for_delalloc(inode, locked_page,
1320 delalloc_start,
1321 ((u64)(start_index + pages_locked - 1)) <<
1322 PAGE_CACHE_SHIFT);
1323 }
1324 return ret;
1325 }
1326
1327 /*
1328 * find a contiguous range of bytes in the file marked as delalloc, not
1329 * more than 'max_bytes'. start and end are used to return the range,
1330 *
1331 * 1 is returned if we find something, 0 if nothing was in the tree
1332 */
1333 static noinline u64 find_lock_delalloc_range(struct inode *inode,
1334 struct extent_io_tree *tree,
1335 struct page *locked_page,
1336 u64 *start, u64 *end,
1337 u64 max_bytes)
1338 {
1339 u64 delalloc_start;
1340 u64 delalloc_end;
1341 u64 found;
1342 struct extent_state *cached_state = NULL;
1343 int ret;
1344 int loops = 0;
1345
1346 again:
1347 /* step one, find a bunch of delalloc bytes starting at start */
1348 delalloc_start = *start;
1349 delalloc_end = 0;
1350 found = find_delalloc_range(tree, &delalloc_start, &delalloc_end,
1351 max_bytes, &cached_state);
1352 if (!found || delalloc_end <= *start) {
1353 *start = delalloc_start;
1354 *end = delalloc_end;
1355 free_extent_state(cached_state);
1356 return found;
1357 }
1358
1359 /*
1360 * start comes from the offset of locked_page. We have to lock
1361 * pages in order, so we can't process delalloc bytes before
1362 * locked_page
1363 */
1364 if (delalloc_start < *start)
1365 delalloc_start = *start;
1366
1367 /*
1368 * make sure to limit the number of pages we try to lock down
1369 * if we're looping.
1370 */
1371 if (delalloc_end + 1 - delalloc_start > max_bytes && loops)
1372 delalloc_end = delalloc_start + PAGE_CACHE_SIZE - 1;
1373
1374 /* step two, lock all the pages after the page that has start */
1375 ret = lock_delalloc_pages(inode, locked_page,
1376 delalloc_start, delalloc_end);
1377 if (ret == -EAGAIN) {
1378 /* some of the pages are gone, lets avoid looping by
1379 * shortening the size of the delalloc range we're searching
1380 */
1381 free_extent_state(cached_state);
1382 if (!loops) {
1383 unsigned long offset = (*start) & (PAGE_CACHE_SIZE - 1);
1384 max_bytes = PAGE_CACHE_SIZE - offset;
1385 loops = 1;
1386 goto again;
1387 } else {
1388 found = 0;
1389 goto out_failed;
1390 }
1391 }
1392 BUG_ON(ret);
1393
1394 /* step three, lock the state bits for the whole range */
1395 lock_extent_bits(tree, delalloc_start, delalloc_end,
1396 0, &cached_state, GFP_NOFS);
1397
1398 /* then test to make sure it is all still delalloc */
1399 ret = test_range_bit(tree, delalloc_start, delalloc_end,
1400 EXTENT_DELALLOC, 1, cached_state);
1401 if (!ret) {
1402 unlock_extent_cached(tree, delalloc_start, delalloc_end,
1403 &cached_state, GFP_NOFS);
1404 __unlock_for_delalloc(inode, locked_page,
1405 delalloc_start, delalloc_end);
1406 cond_resched();
1407 goto again;
1408 }
1409 free_extent_state(cached_state);
1410 *start = delalloc_start;
1411 *end = delalloc_end;
1412 out_failed:
1413 return found;
1414 }
1415
1416 int extent_clear_unlock_delalloc(struct inode *inode,
1417 struct extent_io_tree *tree,
1418 u64 start, u64 end, struct page *locked_page,
1419 unsigned long op)
1420 {
1421 int ret;
1422 struct page *pages[16];
1423 unsigned long index = start >> PAGE_CACHE_SHIFT;
1424 unsigned long end_index = end >> PAGE_CACHE_SHIFT;
1425 unsigned long nr_pages = end_index - index + 1;
1426 int i;
1427 int clear_bits = 0;
1428
1429 if (op & EXTENT_CLEAR_UNLOCK)
1430 clear_bits |= EXTENT_LOCKED;
1431 if (op & EXTENT_CLEAR_DIRTY)
1432 clear_bits |= EXTENT_DIRTY;
1433
1434 if (op & EXTENT_CLEAR_DELALLOC)
1435 clear_bits |= EXTENT_DELALLOC;
1436
1437 clear_extent_bit(tree, start, end, clear_bits, 1, 0, NULL, GFP_NOFS);
1438 if (!(op & (EXTENT_CLEAR_UNLOCK_PAGE | EXTENT_CLEAR_DIRTY |
1439 EXTENT_SET_WRITEBACK | EXTENT_END_WRITEBACK |
1440 EXTENT_SET_PRIVATE2)))
1441 return 0;
1442
1443 while (nr_pages > 0) {
1444 ret = find_get_pages_contig(inode->i_mapping, index,
1445 min_t(unsigned long,
1446 nr_pages, ARRAY_SIZE(pages)), pages);
1447 for (i = 0; i < ret; i++) {
1448
1449 if (op & EXTENT_SET_PRIVATE2)
1450 SetPagePrivate2(pages[i]);
1451
1452 if (pages[i] == locked_page) {
1453 page_cache_release(pages[i]);
1454 continue;
1455 }
1456 if (op & EXTENT_CLEAR_DIRTY)
1457 clear_page_dirty_for_io(pages[i]);
1458 if (op & EXTENT_SET_WRITEBACK)
1459 set_page_writeback(pages[i]);
1460 if (op & EXTENT_END_WRITEBACK)
1461 end_page_writeback(pages[i]);
1462 if (op & EXTENT_CLEAR_UNLOCK_PAGE)
1463 unlock_page(pages[i]);
1464 page_cache_release(pages[i]);
1465 }
1466 nr_pages -= ret;
1467 index += ret;
1468 cond_resched();
1469 }
1470 return 0;
1471 }
1472
1473 /*
1474 * count the number of bytes in the tree that have a given bit(s)
1475 * set. This can be fairly slow, except for EXTENT_DIRTY which is
1476 * cached. The total number found is returned.
1477 */
1478 u64 count_range_bits(struct extent_io_tree *tree,
1479 u64 *start, u64 search_end, u64 max_bytes,
1480 unsigned long bits)
1481 {
1482 struct rb_node *node;
1483 struct extent_state *state;
1484 u64 cur_start = *start;
1485 u64 total_bytes = 0;
1486 int found = 0;
1487
1488 if (search_end <= cur_start) {
1489 WARN_ON(1);
1490 return 0;
1491 }
1492
1493 spin_lock(&tree->lock);
1494 if (cur_start == 0 && bits == EXTENT_DIRTY) {
1495 total_bytes = tree->dirty_bytes;
1496 goto out;
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 goto out;
1505
1506 while (1) {
1507 state = rb_entry(node, struct extent_state, rb_node);
1508 if (state->start > search_end)
1509 break;
1510 if (state->end >= cur_start && (state->state & bits)) {
1511 total_bytes += min(search_end, state->end) + 1 -
1512 max(cur_start, state->start);
1513 if (total_bytes >= max_bytes)
1514 break;
1515 if (!found) {
1516 *start = state->start;
1517 found = 1;
1518 }
1519 }
1520 node = rb_next(node);
1521 if (!node)
1522 break;
1523 }
1524 out:
1525 spin_unlock(&tree->lock);
1526 return total_bytes;
1527 }
1528
1529 /*
1530 * set the private field for a given byte offset in the tree. If there isn't
1531 * an extent_state there already, this does nothing.
1532 */
1533 int set_state_private(struct extent_io_tree *tree, u64 start, u64 private)
1534 {
1535 struct rb_node *node;
1536 struct extent_state *state;
1537 int ret = 0;
1538
1539 spin_lock(&tree->lock);
1540 /*
1541 * this search will find all the extents that end after
1542 * our range starts.
1543 */
1544 node = tree_search(tree, start);
1545 if (!node) {
1546 ret = -ENOENT;
1547 goto out;
1548 }
1549 state = rb_entry(node, struct extent_state, rb_node);
1550 if (state->start != start) {
1551 ret = -ENOENT;
1552 goto out;
1553 }
1554 state->private = private;
1555 out:
1556 spin_unlock(&tree->lock);
1557 return ret;
1558 }
1559
1560 int get_state_private(struct extent_io_tree *tree, u64 start, u64 *private)
1561 {
1562 struct rb_node *node;
1563 struct extent_state *state;
1564 int ret = 0;
1565
1566 spin_lock(&tree->lock);
1567 /*
1568 * this search will find all the extents that end after
1569 * our range starts.
1570 */
1571 node = tree_search(tree, start);
1572 if (!node) {
1573 ret = -ENOENT;
1574 goto out;
1575 }
1576 state = rb_entry(node, struct extent_state, rb_node);
1577 if (state->start != start) {
1578 ret = -ENOENT;
1579 goto out;
1580 }
1581 *private = state->private;
1582 out:
1583 spin_unlock(&tree->lock);
1584 return ret;
1585 }
1586
1587 /*
1588 * searches a range in the state tree for a given mask.
1589 * If 'filled' == 1, this returns 1 only if every extent in the tree
1590 * has the bits set. Otherwise, 1 is returned if any bit in the
1591 * range is found set.
1592 */
1593 int test_range_bit(struct extent_io_tree *tree, u64 start, u64 end,
1594 int bits, int filled, struct extent_state *cached)
1595 {
1596 struct extent_state *state = NULL;
1597 struct rb_node *node;
1598 int bitset = 0;
1599
1600 spin_lock(&tree->lock);
1601 if (cached && cached->tree && cached->start == start)
1602 node = &cached->rb_node;
1603 else
1604 node = tree_search(tree, start);
1605 while (node && start <= end) {
1606 state = rb_entry(node, struct extent_state, rb_node);
1607
1608 if (filled && state->start > start) {
1609 bitset = 0;
1610 break;
1611 }
1612
1613 if (state->start > end)
1614 break;
1615
1616 if (state->state & bits) {
1617 bitset = 1;
1618 if (!filled)
1619 break;
1620 } else if (filled) {
1621 bitset = 0;
1622 break;
1623 }
1624
1625 if (state->end == (u64)-1)
1626 break;
1627
1628 start = state->end + 1;
1629 if (start > end)
1630 break;
1631 node = rb_next(node);
1632 if (!node) {
1633 if (filled)
1634 bitset = 0;
1635 break;
1636 }
1637 }
1638 spin_unlock(&tree->lock);
1639 return bitset;
1640 }
1641
1642 /*
1643 * helper function to set a given page up to date if all the
1644 * extents in the tree for that page are up to date
1645 */
1646 static int check_page_uptodate(struct extent_io_tree *tree,
1647 struct page *page)
1648 {
1649 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
1650 u64 end = start + PAGE_CACHE_SIZE - 1;
1651 if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1, NULL))
1652 SetPageUptodate(page);
1653 return 0;
1654 }
1655
1656 /*
1657 * helper function to unlock a page if all the extents in the tree
1658 * for that page are unlocked
1659 */
1660 static int check_page_locked(struct extent_io_tree *tree,
1661 struct page *page)
1662 {
1663 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
1664 u64 end = start + PAGE_CACHE_SIZE - 1;
1665 if (!test_range_bit(tree, start, end, EXTENT_LOCKED, 0, NULL))
1666 unlock_page(page);
1667 return 0;
1668 }
1669
1670 /*
1671 * helper function to end page writeback if all the extents
1672 * in the tree for that page are done with writeback
1673 */
1674 static int check_page_writeback(struct extent_io_tree *tree,
1675 struct page *page)
1676 {
1677 end_page_writeback(page);
1678 return 0;
1679 }
1680
1681 /* lots and lots of room for performance fixes in the end_bio funcs */
1682
1683 /*
1684 * after a writepage IO is done, we need to:
1685 * clear the uptodate bits on error
1686 * clear the writeback bits in the extent tree for this IO
1687 * end_page_writeback if the page has no more pending IO
1688 *
1689 * Scheduling is not allowed, so the extent state tree is expected
1690 * to have one and only one object corresponding to this IO.
1691 */
1692 static void end_bio_extent_writepage(struct bio *bio, int err)
1693 {
1694 int uptodate = err == 0;
1695 struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
1696 struct extent_io_tree *tree;
1697 u64 start;
1698 u64 end;
1699 int whole_page;
1700 int ret;
1701
1702 do {
1703 struct page *page = bvec->bv_page;
1704 tree = &BTRFS_I(page->mapping->host)->io_tree;
1705
1706 start = ((u64)page->index << PAGE_CACHE_SHIFT) +
1707 bvec->bv_offset;
1708 end = start + bvec->bv_len - 1;
1709
1710 if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE)
1711 whole_page = 1;
1712 else
1713 whole_page = 0;
1714
1715 if (--bvec >= bio->bi_io_vec)
1716 prefetchw(&bvec->bv_page->flags);
1717 if (tree->ops && tree->ops->writepage_end_io_hook) {
1718 ret = tree->ops->writepage_end_io_hook(page, start,
1719 end, NULL, uptodate);
1720 if (ret)
1721 uptodate = 0;
1722 }
1723
1724 if (!uptodate && tree->ops &&
1725 tree->ops->writepage_io_failed_hook) {
1726 ret = tree->ops->writepage_io_failed_hook(bio, page,
1727 start, end, NULL);
1728 if (ret == 0) {
1729 uptodate = (err == 0);
1730 continue;
1731 }
1732 }
1733
1734 if (!uptodate) {
1735 clear_extent_uptodate(tree, start, end, NULL, GFP_NOFS);
1736 ClearPageUptodate(page);
1737 SetPageError(page);
1738 }
1739
1740 if (whole_page)
1741 end_page_writeback(page);
1742 else
1743 check_page_writeback(tree, page);
1744 } while (bvec >= bio->bi_io_vec);
1745
1746 bio_put(bio);
1747 }
1748
1749 /*
1750 * after a readpage IO is done, we need to:
1751 * clear the uptodate bits on error
1752 * set the uptodate bits if things worked
1753 * set the page up to date if all extents in the tree are uptodate
1754 * clear the lock bit in the extent tree
1755 * unlock the page if there are no other extents locked for it
1756 *
1757 * Scheduling is not allowed, so the extent state tree is expected
1758 * to have one and only one object corresponding to this IO.
1759 */
1760 static void end_bio_extent_readpage(struct bio *bio, int err)
1761 {
1762 int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
1763 struct bio_vec *bvec_end = bio->bi_io_vec + bio->bi_vcnt - 1;
1764 struct bio_vec *bvec = bio->bi_io_vec;
1765 struct extent_io_tree *tree;
1766 u64 start;
1767 u64 end;
1768 int whole_page;
1769 int ret;
1770
1771 if (err)
1772 uptodate = 0;
1773
1774 do {
1775 struct page *page = bvec->bv_page;
1776 tree = &BTRFS_I(page->mapping->host)->io_tree;
1777
1778 start = ((u64)page->index << PAGE_CACHE_SHIFT) +
1779 bvec->bv_offset;
1780 end = start + bvec->bv_len - 1;
1781
1782 if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE)
1783 whole_page = 1;
1784 else
1785 whole_page = 0;
1786
1787 if (++bvec <= bvec_end)
1788 prefetchw(&bvec->bv_page->flags);
1789
1790 if (uptodate && tree->ops && tree->ops->readpage_end_io_hook) {
1791 ret = tree->ops->readpage_end_io_hook(page, start, end,
1792 NULL);
1793 if (ret)
1794 uptodate = 0;
1795 }
1796 if (!uptodate && tree->ops &&
1797 tree->ops->readpage_io_failed_hook) {
1798 ret = tree->ops->readpage_io_failed_hook(bio, page,
1799 start, end, NULL);
1800 if (ret == 0) {
1801 uptodate =
1802 test_bit(BIO_UPTODATE, &bio->bi_flags);
1803 if (err)
1804 uptodate = 0;
1805 continue;
1806 }
1807 }
1808
1809 if (uptodate) {
1810 set_extent_uptodate(tree, start, end,
1811 GFP_ATOMIC);
1812 }
1813 unlock_extent(tree, start, end, GFP_ATOMIC);
1814
1815 if (whole_page) {
1816 if (uptodate) {
1817 SetPageUptodate(page);
1818 } else {
1819 ClearPageUptodate(page);
1820 SetPageError(page);
1821 }
1822 unlock_page(page);
1823 } else {
1824 if (uptodate) {
1825 check_page_uptodate(tree, page);
1826 } else {
1827 ClearPageUptodate(page);
1828 SetPageError(page);
1829 }
1830 check_page_locked(tree, page);
1831 }
1832 } while (bvec <= bvec_end);
1833
1834 bio_put(bio);
1835 }
1836
1837 /*
1838 * IO done from prepare_write is pretty simple, we just unlock
1839 * the structs in the extent tree when done, and set the uptodate bits
1840 * as appropriate.
1841 */
1842 static void end_bio_extent_preparewrite(struct bio *bio, int err)
1843 {
1844 const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
1845 struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
1846 struct extent_io_tree *tree;
1847 u64 start;
1848 u64 end;
1849
1850 do {
1851 struct page *page = bvec->bv_page;
1852 tree = &BTRFS_I(page->mapping->host)->io_tree;
1853
1854 start = ((u64)page->index << PAGE_CACHE_SHIFT) +
1855 bvec->bv_offset;
1856 end = start + bvec->bv_len - 1;
1857
1858 if (--bvec >= bio->bi_io_vec)
1859 prefetchw(&bvec->bv_page->flags);
1860
1861 if (uptodate) {
1862 set_extent_uptodate(tree, start, end, GFP_ATOMIC);
1863 } else {
1864 ClearPageUptodate(page);
1865 SetPageError(page);
1866 }
1867
1868 unlock_extent(tree, start, end, GFP_ATOMIC);
1869
1870 } while (bvec >= bio->bi_io_vec);
1871
1872 bio_put(bio);
1873 }
1874
1875 static struct bio *
1876 extent_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
1877 gfp_t gfp_flags)
1878 {
1879 struct bio *bio;
1880
1881 bio = bio_alloc(gfp_flags, nr_vecs);
1882
1883 if (bio == NULL && (current->flags & PF_MEMALLOC)) {
1884 while (!bio && (nr_vecs /= 2))
1885 bio = bio_alloc(gfp_flags, nr_vecs);
1886 }
1887
1888 if (bio) {
1889 bio->bi_size = 0;
1890 bio->bi_bdev = bdev;
1891 bio->bi_sector = first_sector;
1892 }
1893 return bio;
1894 }
1895
1896 static int submit_one_bio(int rw, struct bio *bio, int mirror_num,
1897 unsigned long bio_flags)
1898 {
1899 int ret = 0;
1900 struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
1901 struct page *page = bvec->bv_page;
1902 struct extent_io_tree *tree = bio->bi_private;
1903 u64 start;
1904 u64 end;
1905
1906 start = ((u64)page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset;
1907 end = start + bvec->bv_len - 1;
1908
1909 bio->bi_private = NULL;
1910
1911 bio_get(bio);
1912
1913 if (tree->ops && tree->ops->submit_bio_hook)
1914 tree->ops->submit_bio_hook(page->mapping->host, rw, bio,
1915 mirror_num, bio_flags, start);
1916 else
1917 submit_bio(rw, bio);
1918 if (bio_flagged(bio, BIO_EOPNOTSUPP))
1919 ret = -EOPNOTSUPP;
1920 bio_put(bio);
1921 return ret;
1922 }
1923
1924 static int submit_extent_page(int rw, struct extent_io_tree *tree,
1925 struct page *page, sector_t sector,
1926 size_t size, unsigned long offset,
1927 struct block_device *bdev,
1928 struct bio **bio_ret,
1929 unsigned long max_pages,
1930 bio_end_io_t end_io_func,
1931 int mirror_num,
1932 unsigned long prev_bio_flags,
1933 unsigned long bio_flags)
1934 {
1935 int ret = 0;
1936 struct bio *bio;
1937 int nr;
1938 int contig = 0;
1939 int this_compressed = bio_flags & EXTENT_BIO_COMPRESSED;
1940 int old_compressed = prev_bio_flags & EXTENT_BIO_COMPRESSED;
1941 size_t page_size = min_t(size_t, size, PAGE_CACHE_SIZE);
1942
1943 if (bio_ret && *bio_ret) {
1944 bio = *bio_ret;
1945 if (old_compressed)
1946 contig = bio->bi_sector == sector;
1947 else
1948 contig = bio->bi_sector + (bio->bi_size >> 9) ==
1949 sector;
1950
1951 if (prev_bio_flags != bio_flags || !contig ||
1952 (tree->ops && tree->ops->merge_bio_hook &&
1953 tree->ops->merge_bio_hook(page, offset, page_size, bio,
1954 bio_flags)) ||
1955 bio_add_page(bio, page, page_size, offset) < page_size) {
1956 ret = submit_one_bio(rw, bio, mirror_num,
1957 prev_bio_flags);
1958 bio = NULL;
1959 } else {
1960 return 0;
1961 }
1962 }
1963 if (this_compressed)
1964 nr = BIO_MAX_PAGES;
1965 else
1966 nr = bio_get_nr_vecs(bdev);
1967
1968 bio = extent_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH);
1969
1970 bio_add_page(bio, page, page_size, offset);
1971 bio->bi_end_io = end_io_func;
1972 bio->bi_private = tree;
1973
1974 if (bio_ret)
1975 *bio_ret = bio;
1976 else
1977 ret = submit_one_bio(rw, bio, mirror_num, bio_flags);
1978
1979 return ret;
1980 }
1981
1982 void set_page_extent_mapped(struct page *page)
1983 {
1984 if (!PagePrivate(page)) {
1985 SetPagePrivate(page);
1986 page_cache_get(page);
1987 set_page_private(page, EXTENT_PAGE_PRIVATE);
1988 }
1989 }
1990
1991 static void set_page_extent_head(struct page *page, unsigned long len)
1992 {
1993 set_page_private(page, EXTENT_PAGE_PRIVATE_FIRST_PAGE | len << 2);
1994 }
1995
1996 /*
1997 * basic readpage implementation. Locked extent state structs are inserted
1998 * into the tree that are removed when the IO is done (by the end_io
1999 * handlers)
2000 */
2001 static int __extent_read_full_page(struct extent_io_tree *tree,
2002 struct page *page,
2003 get_extent_t *get_extent,
2004 struct bio **bio, int mirror_num,
2005 unsigned long *bio_flags)
2006 {
2007 struct inode *inode = page->mapping->host;
2008 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
2009 u64 page_end = start + PAGE_CACHE_SIZE - 1;
2010 u64 end;
2011 u64 cur = start;
2012 u64 extent_offset;
2013 u64 last_byte = i_size_read(inode);
2014 u64 block_start;
2015 u64 cur_end;
2016 sector_t sector;
2017 struct extent_map *em;
2018 struct block_device *bdev;
2019 struct btrfs_ordered_extent *ordered;
2020 int ret;
2021 int nr = 0;
2022 size_t page_offset = 0;
2023 size_t iosize;
2024 size_t disk_io_size;
2025 size_t blocksize = inode->i_sb->s_blocksize;
2026 unsigned long this_bio_flag = 0;
2027
2028 set_page_extent_mapped(page);
2029
2030 end = page_end;
2031 while (1) {
2032 lock_extent(tree, start, end, GFP_NOFS);
2033 ordered = btrfs_lookup_ordered_extent(inode, start);
2034 if (!ordered)
2035 break;
2036 unlock_extent(tree, start, end, GFP_NOFS);
2037 btrfs_start_ordered_extent(inode, ordered, 1);
2038 btrfs_put_ordered_extent(ordered);
2039 }
2040
2041 if (page->index == last_byte >> PAGE_CACHE_SHIFT) {
2042 char *userpage;
2043 size_t zero_offset = last_byte & (PAGE_CACHE_SIZE - 1);
2044
2045 if (zero_offset) {
2046 iosize = PAGE_CACHE_SIZE - zero_offset;
2047 userpage = kmap_atomic(page, KM_USER0);
2048 memset(userpage + zero_offset, 0, iosize);
2049 flush_dcache_page(page);
2050 kunmap_atomic(userpage, KM_USER0);
2051 }
2052 }
2053 while (cur <= end) {
2054 if (cur >= last_byte) {
2055 char *userpage;
2056 iosize = PAGE_CACHE_SIZE - page_offset;
2057 userpage = kmap_atomic(page, KM_USER0);
2058 memset(userpage + page_offset, 0, iosize);
2059 flush_dcache_page(page);
2060 kunmap_atomic(userpage, KM_USER0);
2061 set_extent_uptodate(tree, cur, cur + iosize - 1,
2062 GFP_NOFS);
2063 unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
2064 break;
2065 }
2066 em = get_extent(inode, page, page_offset, cur,
2067 end - cur + 1, 0);
2068 if (IS_ERR(em) || !em) {
2069 SetPageError(page);
2070 unlock_extent(tree, cur, end, GFP_NOFS);
2071 break;
2072 }
2073 extent_offset = cur - em->start;
2074 BUG_ON(extent_map_end(em) <= cur);
2075 BUG_ON(end < cur);
2076
2077 if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
2078 this_bio_flag = EXTENT_BIO_COMPRESSED;
2079
2080 iosize = min(extent_map_end(em) - cur, end - cur + 1);
2081 cur_end = min(extent_map_end(em) - 1, end);
2082 iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1);
2083 if (this_bio_flag & EXTENT_BIO_COMPRESSED) {
2084 disk_io_size = em->block_len;
2085 sector = em->block_start >> 9;
2086 } else {
2087 sector = (em->block_start + extent_offset) >> 9;
2088 disk_io_size = iosize;
2089 }
2090 bdev = em->bdev;
2091 block_start = em->block_start;
2092 if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
2093 block_start = EXTENT_MAP_HOLE;
2094 free_extent_map(em);
2095 em = NULL;
2096
2097 /* we've found a hole, just zero and go on */
2098 if (block_start == EXTENT_MAP_HOLE) {
2099 char *userpage;
2100 userpage = kmap_atomic(page, KM_USER0);
2101 memset(userpage + page_offset, 0, iosize);
2102 flush_dcache_page(page);
2103 kunmap_atomic(userpage, KM_USER0);
2104
2105 set_extent_uptodate(tree, cur, cur + iosize - 1,
2106 GFP_NOFS);
2107 unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
2108 cur = cur + iosize;
2109 page_offset += iosize;
2110 continue;
2111 }
2112 /* the get_extent function already copied into the page */
2113 if (test_range_bit(tree, cur, cur_end,
2114 EXTENT_UPTODATE, 1, NULL)) {
2115 check_page_uptodate(tree, page);
2116 unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
2117 cur = cur + iosize;
2118 page_offset += iosize;
2119 continue;
2120 }
2121 /* we have an inline extent but it didn't get marked up
2122 * to date. Error out
2123 */
2124 if (block_start == EXTENT_MAP_INLINE) {
2125 SetPageError(page);
2126 unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
2127 cur = cur + iosize;
2128 page_offset += iosize;
2129 continue;
2130 }
2131
2132 ret = 0;
2133 if (tree->ops && tree->ops->readpage_io_hook) {
2134 ret = tree->ops->readpage_io_hook(page, cur,
2135 cur + iosize - 1);
2136 }
2137 if (!ret) {
2138 unsigned long pnr = (last_byte >> PAGE_CACHE_SHIFT) + 1;
2139 pnr -= page->index;
2140 ret = submit_extent_page(READ, tree, page,
2141 sector, disk_io_size, page_offset,
2142 bdev, bio, pnr,
2143 end_bio_extent_readpage, mirror_num,
2144 *bio_flags,
2145 this_bio_flag);
2146 nr++;
2147 *bio_flags = this_bio_flag;
2148 }
2149 if (ret)
2150 SetPageError(page);
2151 cur = cur + iosize;
2152 page_offset += iosize;
2153 }
2154 if (!nr) {
2155 if (!PageError(page))
2156 SetPageUptodate(page);
2157 unlock_page(page);
2158 }
2159 return 0;
2160 }
2161
2162 int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
2163 get_extent_t *get_extent)
2164 {
2165 struct bio *bio = NULL;
2166 unsigned long bio_flags = 0;
2167 int ret;
2168
2169 ret = __extent_read_full_page(tree, page, get_extent, &bio, 0,
2170 &bio_flags);
2171 if (bio)
2172 submit_one_bio(READ, bio, 0, bio_flags);
2173 return ret;
2174 }
2175
2176 static noinline void update_nr_written(struct page *page,
2177 struct writeback_control *wbc,
2178 unsigned long nr_written)
2179 {
2180 wbc->nr_to_write -= nr_written;
2181 if (wbc->range_cyclic || (wbc->nr_to_write > 0 &&
2182 wbc->range_start == 0 && wbc->range_end == LLONG_MAX))
2183 page->mapping->writeback_index = page->index + nr_written;
2184 }
2185
2186 /*
2187 * the writepage semantics are similar to regular writepage. extent
2188 * records are inserted to lock ranges in the tree, and as dirty areas
2189 * are found, they are marked writeback. Then the lock bits are removed
2190 * and the end_io handler clears the writeback ranges
2191 */
2192 static int __extent_writepage(struct page *page, struct writeback_control *wbc,
2193 void *data)
2194 {
2195 struct inode *inode = page->mapping->host;
2196 struct extent_page_data *epd = data;
2197 struct extent_io_tree *tree = epd->tree;
2198 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
2199 u64 delalloc_start;
2200 u64 page_end = start + PAGE_CACHE_SIZE - 1;
2201 u64 end;
2202 u64 cur = start;
2203 u64 extent_offset;
2204 u64 last_byte = i_size_read(inode);
2205 u64 block_start;
2206 u64 iosize;
2207 u64 unlock_start;
2208 sector_t sector;
2209 struct extent_state *cached_state = NULL;
2210 struct extent_map *em;
2211 struct block_device *bdev;
2212 int ret;
2213 int nr = 0;
2214 size_t pg_offset = 0;
2215 size_t blocksize;
2216 loff_t i_size = i_size_read(inode);
2217 unsigned long end_index = i_size >> PAGE_CACHE_SHIFT;
2218 u64 nr_delalloc;
2219 u64 delalloc_end;
2220 int page_started;
2221 int compressed;
2222 int write_flags;
2223 unsigned long nr_written = 0;
2224
2225 if (wbc->sync_mode == WB_SYNC_ALL)
2226 write_flags = WRITE_SYNC_PLUG;
2227 else
2228 write_flags = WRITE;
2229
2230 WARN_ON(!PageLocked(page));
2231 pg_offset = i_size & (PAGE_CACHE_SIZE - 1);
2232 if (page->index > end_index ||
2233 (page->index == end_index && !pg_offset)) {
2234 page->mapping->a_ops->invalidatepage(page, 0);
2235 unlock_page(page);
2236 return 0;
2237 }
2238
2239 if (page->index == end_index) {
2240 char *userpage;
2241
2242 userpage = kmap_atomic(page, KM_USER0);
2243 memset(userpage + pg_offset, 0,
2244 PAGE_CACHE_SIZE - pg_offset);
2245 kunmap_atomic(userpage, KM_USER0);
2246 flush_dcache_page(page);
2247 }
2248 pg_offset = 0;
2249
2250 set_page_extent_mapped(page);
2251
2252 delalloc_start = start;
2253 delalloc_end = 0;
2254 page_started = 0;
2255 if (!epd->extent_locked) {
2256 u64 delalloc_to_write = 0;
2257 /*
2258 * make sure the wbc mapping index is at least updated
2259 * to this page.
2260 */
2261 update_nr_written(page, wbc, 0);
2262
2263 while (delalloc_end < page_end) {
2264 nr_delalloc = find_lock_delalloc_range(inode, tree,
2265 page,
2266 &delalloc_start,
2267 &delalloc_end,
2268 128 * 1024 * 1024);
2269 if (nr_delalloc == 0) {
2270 delalloc_start = delalloc_end + 1;
2271 continue;
2272 }
2273 tree->ops->fill_delalloc(inode, page, delalloc_start,
2274 delalloc_end, &page_started,
2275 &nr_written);
2276 /*
2277 * delalloc_end is already one less than the total
2278 * length, so we don't subtract one from
2279 * PAGE_CACHE_SIZE
2280 */
2281 delalloc_to_write += (delalloc_end - delalloc_start +
2282 PAGE_CACHE_SIZE) >>
2283 PAGE_CACHE_SHIFT;
2284 delalloc_start = delalloc_end + 1;
2285 }
2286 if (wbc->nr_to_write < delalloc_to_write) {
2287 int thresh = 8192;
2288
2289 if (delalloc_to_write < thresh * 2)
2290 thresh = delalloc_to_write;
2291 wbc->nr_to_write = min_t(u64, delalloc_to_write,
2292 thresh);
2293 }
2294
2295 /* did the fill delalloc function already unlock and start
2296 * the IO?
2297 */
2298 if (page_started) {
2299 ret = 0;
2300 /*
2301 * we've unlocked the page, so we can't update
2302 * the mapping's writeback index, just update
2303 * nr_to_write.
2304 */
2305 wbc->nr_to_write -= nr_written;
2306 goto done_unlocked;
2307 }
2308 }
2309 if (tree->ops && tree->ops->writepage_start_hook) {
2310 ret = tree->ops->writepage_start_hook(page, start,
2311 page_end);
2312 if (ret == -EAGAIN) {
2313 redirty_page_for_writepage(wbc, page);
2314 update_nr_written(page, wbc, nr_written);
2315 unlock_page(page);
2316 ret = 0;
2317 goto done_unlocked;
2318 }
2319 }
2320
2321 /*
2322 * we don't want to touch the inode after unlocking the page,
2323 * so we update the mapping writeback index now
2324 */
2325 update_nr_written(page, wbc, nr_written + 1);
2326
2327 end = page_end;
2328 if (last_byte <= start) {
2329 if (tree->ops && tree->ops->writepage_end_io_hook)
2330 tree->ops->writepage_end_io_hook(page, start,
2331 page_end, NULL, 1);
2332 unlock_start = page_end + 1;
2333 goto done;
2334 }
2335
2336 blocksize = inode->i_sb->s_blocksize;
2337
2338 while (cur <= end) {
2339 if (cur >= last_byte) {
2340 if (tree->ops && tree->ops->writepage_end_io_hook)
2341 tree->ops->writepage_end_io_hook(page, cur,
2342 page_end, NULL, 1);
2343 unlock_start = page_end + 1;
2344 break;
2345 }
2346 em = epd->get_extent(inode, page, pg_offset, cur,
2347 end - cur + 1, 1);
2348 if (IS_ERR(em) || !em) {
2349 SetPageError(page);
2350 break;
2351 }
2352
2353 extent_offset = cur - em->start;
2354 BUG_ON(extent_map_end(em) <= cur);
2355 BUG_ON(end < cur);
2356 iosize = min(extent_map_end(em) - cur, end - cur + 1);
2357 iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1);
2358 sector = (em->block_start + extent_offset) >> 9;
2359 bdev = em->bdev;
2360 block_start = em->block_start;
2361 compressed = test_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
2362 free_extent_map(em);
2363 em = NULL;
2364
2365 /*
2366 * compressed and inline extents are written through other
2367 * paths in the FS
2368 */
2369 if (compressed || block_start == EXTENT_MAP_HOLE ||
2370 block_start == EXTENT_MAP_INLINE) {
2371 /*
2372 * end_io notification does not happen here for
2373 * compressed extents
2374 */
2375 if (!compressed && tree->ops &&
2376 tree->ops->writepage_end_io_hook)
2377 tree->ops->writepage_end_io_hook(page, cur,
2378 cur + iosize - 1,
2379 NULL, 1);
2380 else if (compressed) {
2381 /* we don't want to end_page_writeback on
2382 * a compressed extent. this happens
2383 * elsewhere
2384 */
2385 nr++;
2386 }
2387
2388 cur += iosize;
2389 pg_offset += iosize;
2390 unlock_start = cur;
2391 continue;
2392 }
2393 /* leave this out until we have a page_mkwrite call */
2394 if (0 && !test_range_bit(tree, cur, cur + iosize - 1,
2395 EXTENT_DIRTY, 0, NULL)) {
2396 cur = cur + iosize;
2397 pg_offset += iosize;
2398 continue;
2399 }
2400
2401 if (tree->ops && tree->ops->writepage_io_hook) {
2402 ret = tree->ops->writepage_io_hook(page, cur,
2403 cur + iosize - 1);
2404 } else {
2405 ret = 0;
2406 }
2407 if (ret) {
2408 SetPageError(page);
2409 } else {
2410 unsigned long max_nr = end_index + 1;
2411
2412 set_range_writeback(tree, cur, cur + iosize - 1);
2413 if (!PageWriteback(page)) {
2414 printk(KERN_ERR "btrfs warning page %lu not "
2415 "writeback, cur %llu end %llu\n",
2416 page->index, (unsigned long long)cur,
2417 (unsigned long long)end);
2418 }
2419
2420 ret = submit_extent_page(write_flags, tree, page,
2421 sector, iosize, pg_offset,
2422 bdev, &epd->bio, max_nr,
2423 end_bio_extent_writepage,
2424 0, 0, 0);
2425 if (ret)
2426 SetPageError(page);
2427 }
2428 cur = cur + iosize;
2429 pg_offset += iosize;
2430 nr++;
2431 }
2432 done:
2433 if (nr == 0) {
2434 /* make sure the mapping tag for page dirty gets cleared */
2435 set_page_writeback(page);
2436 end_page_writeback(page);
2437 }
2438 unlock_page(page);
2439
2440 done_unlocked:
2441
2442 /* drop our reference on any cached states */
2443 free_extent_state(cached_state);
2444 return 0;
2445 }
2446
2447 /**
2448 * write_cache_pages - walk the list of dirty pages of the given address space and write all of them.
2449 * @mapping: address space structure to write
2450 * @wbc: subtract the number of written pages from *@wbc->nr_to_write
2451 * @writepage: function called for each page
2452 * @data: data passed to writepage function
2453 *
2454 * If a page is already under I/O, write_cache_pages() skips it, even
2455 * if it's dirty. This is desirable behaviour for memory-cleaning writeback,
2456 * but it is INCORRECT for data-integrity system calls such as fsync(). fsync()
2457 * and msync() need to guarantee that all the data which was dirty at the time
2458 * the call was made get new I/O started against them. If wbc->sync_mode is
2459 * WB_SYNC_ALL then we were called for data integrity and we must wait for
2460 * existing IO to complete.
2461 */
2462 static int extent_write_cache_pages(struct extent_io_tree *tree,
2463 struct address_space *mapping,
2464 struct writeback_control *wbc,
2465 writepage_t writepage, void *data,
2466 void (*flush_fn)(void *))
2467 {
2468 int ret = 0;
2469 int done = 0;
2470 int nr_to_write_done = 0;
2471 struct pagevec pvec;
2472 int nr_pages;
2473 pgoff_t index;
2474 pgoff_t end; /* Inclusive */
2475 int scanned = 0;
2476 int range_whole = 0;
2477
2478 pagevec_init(&pvec, 0);
2479 if (wbc->range_cyclic) {
2480 index = mapping->writeback_index; /* Start from prev offset */
2481 end = -1;
2482 } else {
2483 index = wbc->range_start >> PAGE_CACHE_SHIFT;
2484 end = wbc->range_end >> PAGE_CACHE_SHIFT;
2485 if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
2486 range_whole = 1;
2487 scanned = 1;
2488 }
2489 retry:
2490 while (!done && !nr_to_write_done && (index <= end) &&
2491 (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index,
2492 PAGECACHE_TAG_DIRTY, min(end - index,
2493 (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
2494 unsigned i;
2495
2496 scanned = 1;
2497 for (i = 0; i < nr_pages; i++) {
2498 struct page *page = pvec.pages[i];
2499
2500 /*
2501 * At this point we hold neither mapping->tree_lock nor
2502 * lock on the page itself: the page may be truncated or
2503 * invalidated (changing page->mapping to NULL), or even
2504 * swizzled back from swapper_space to tmpfs file
2505 * mapping
2506 */
2507 if (tree->ops && tree->ops->write_cache_pages_lock_hook)
2508 tree->ops->write_cache_pages_lock_hook(page);
2509 else
2510 lock_page(page);
2511
2512 if (unlikely(page->mapping != mapping)) {
2513 unlock_page(page);
2514 continue;
2515 }
2516
2517 if (!wbc->range_cyclic && page->index > end) {
2518 done = 1;
2519 unlock_page(page);
2520 continue;
2521 }
2522
2523 if (wbc->sync_mode != WB_SYNC_NONE) {
2524 if (PageWriteback(page))
2525 flush_fn(data);
2526 wait_on_page_writeback(page);
2527 }
2528
2529 if (PageWriteback(page) ||
2530 !clear_page_dirty_for_io(page)) {
2531 unlock_page(page);
2532 continue;
2533 }
2534
2535 ret = (*writepage)(page, wbc, data);
2536
2537 if (unlikely(ret == AOP_WRITEPAGE_ACTIVATE)) {
2538 unlock_page(page);
2539 ret = 0;
2540 }
2541 if (ret)
2542 done = 1;
2543
2544 /*
2545 * the filesystem may choose to bump up nr_to_write.
2546 * We have to make sure to honor the new nr_to_write
2547 * at any time
2548 */
2549 nr_to_write_done = wbc->nr_to_write <= 0;
2550 }
2551 pagevec_release(&pvec);
2552 cond_resched();
2553 }
2554 if (!scanned && !done) {
2555 /*
2556 * We hit the last page and there is more work to be done: wrap
2557 * back to the start of the file
2558 */
2559 scanned = 1;
2560 index = 0;
2561 goto retry;
2562 }
2563 return ret;
2564 }
2565
2566 static void flush_epd_write_bio(struct extent_page_data *epd)
2567 {
2568 if (epd->bio) {
2569 if (epd->sync_io)
2570 submit_one_bio(WRITE_SYNC, epd->bio, 0, 0);
2571 else
2572 submit_one_bio(WRITE, epd->bio, 0, 0);
2573 epd->bio = NULL;
2574 }
2575 }
2576
2577 static noinline void flush_write_bio(void *data)
2578 {
2579 struct extent_page_data *epd = data;
2580 flush_epd_write_bio(epd);
2581 }
2582
2583 int extent_write_full_page(struct extent_io_tree *tree, struct page *page,
2584 get_extent_t *get_extent,
2585 struct writeback_control *wbc)
2586 {
2587 int ret;
2588 struct address_space *mapping = page->mapping;
2589 struct extent_page_data epd = {
2590 .bio = NULL,
2591 .tree = tree,
2592 .get_extent = get_extent,
2593 .extent_locked = 0,
2594 .sync_io = wbc->sync_mode == WB_SYNC_ALL,
2595 };
2596 struct writeback_control wbc_writepages = {
2597 .sync_mode = wbc->sync_mode,
2598 .older_than_this = NULL,
2599 .nr_to_write = 64,
2600 .range_start = page_offset(page) + PAGE_CACHE_SIZE,
2601 .range_end = (loff_t)-1,
2602 };
2603
2604 ret = __extent_writepage(page, wbc, &epd);
2605
2606 extent_write_cache_pages(tree, mapping, &wbc_writepages,
2607 __extent_writepage, &epd, flush_write_bio);
2608 flush_epd_write_bio(&epd);
2609 return ret;
2610 }
2611
2612 int extent_write_locked_range(struct extent_io_tree *tree, struct inode *inode,
2613 u64 start, u64 end, get_extent_t *get_extent,
2614 int mode)
2615 {
2616 int ret = 0;
2617 struct address_space *mapping = inode->i_mapping;
2618 struct page *page;
2619 unsigned long nr_pages = (end - start + PAGE_CACHE_SIZE) >>
2620 PAGE_CACHE_SHIFT;
2621
2622 struct extent_page_data epd = {
2623 .bio = NULL,
2624 .tree = tree,
2625 .get_extent = get_extent,
2626 .extent_locked = 1,
2627 .sync_io = mode == WB_SYNC_ALL,
2628 };
2629 struct writeback_control wbc_writepages = {
2630 .sync_mode = mode,
2631 .older_than_this = NULL,
2632 .nr_to_write = nr_pages * 2,
2633 .range_start = start,
2634 .range_end = end + 1,
2635 };
2636
2637 while (start <= end) {
2638 page = find_get_page(mapping, start >> PAGE_CACHE_SHIFT);
2639 if (clear_page_dirty_for_io(page))
2640 ret = __extent_writepage(page, &wbc_writepages, &epd);
2641 else {
2642 if (tree->ops && tree->ops->writepage_end_io_hook)
2643 tree->ops->writepage_end_io_hook(page, start,
2644 start + PAGE_CACHE_SIZE - 1,
2645 NULL, 1);
2646 unlock_page(page);
2647 }
2648 page_cache_release(page);
2649 start += PAGE_CACHE_SIZE;
2650 }
2651
2652 flush_epd_write_bio(&epd);
2653 return ret;
2654 }
2655
2656 int extent_writepages(struct extent_io_tree *tree,
2657 struct address_space *mapping,
2658 get_extent_t *get_extent,
2659 struct writeback_control *wbc)
2660 {
2661 int ret = 0;
2662 struct extent_page_data epd = {
2663 .bio = NULL,
2664 .tree = tree,
2665 .get_extent = get_extent,
2666 .extent_locked = 0,
2667 .sync_io = wbc->sync_mode == WB_SYNC_ALL,
2668 };
2669
2670 ret = extent_write_cache_pages(tree, mapping, wbc,
2671 __extent_writepage, &epd,
2672 flush_write_bio);
2673 flush_epd_write_bio(&epd);
2674 return ret;
2675 }
2676
2677 int extent_readpages(struct extent_io_tree *tree,
2678 struct address_space *mapping,
2679 struct list_head *pages, unsigned nr_pages,
2680 get_extent_t get_extent)
2681 {
2682 struct bio *bio = NULL;
2683 unsigned page_idx;
2684 unsigned long bio_flags = 0;
2685
2686 for (page_idx = 0; page_idx < nr_pages; page_idx++) {
2687 struct page *page = list_entry(pages->prev, struct page, lru);
2688
2689 prefetchw(&page->flags);
2690 list_del(&page->lru);
2691 if (!add_to_page_cache_lru(page, mapping,
2692 page->index, GFP_KERNEL)) {
2693 __extent_read_full_page(tree, page, get_extent,
2694 &bio, 0, &bio_flags);
2695 }
2696 page_cache_release(page);
2697 }
2698 BUG_ON(!list_empty(pages));
2699 if (bio)
2700 submit_one_bio(READ, bio, 0, bio_flags);
2701 return 0;
2702 }
2703
2704 /*
2705 * basic invalidatepage code, this waits on any locked or writeback
2706 * ranges corresponding to the page, and then deletes any extent state
2707 * records from the tree
2708 */
2709 int extent_invalidatepage(struct extent_io_tree *tree,
2710 struct page *page, unsigned long offset)
2711 {
2712 struct extent_state *cached_state = NULL;
2713 u64 start = ((u64)page->index << PAGE_CACHE_SHIFT);
2714 u64 end = start + PAGE_CACHE_SIZE - 1;
2715 size_t blocksize = page->mapping->host->i_sb->s_blocksize;
2716
2717 start += (offset + blocksize - 1) & ~(blocksize - 1);
2718 if (start > end)
2719 return 0;
2720
2721 lock_extent_bits(tree, start, end, 0, &cached_state, GFP_NOFS);
2722 wait_on_page_writeback(page);
2723 clear_extent_bit(tree, start, end,
2724 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
2725 EXTENT_DO_ACCOUNTING,
2726 1, 1, &cached_state, GFP_NOFS);
2727 return 0;
2728 }
2729
2730 /*
2731 * simple commit_write call, set_range_dirty is used to mark both
2732 * the pages and the extent records as dirty
2733 */
2734 int extent_commit_write(struct extent_io_tree *tree,
2735 struct inode *inode, struct page *page,
2736 unsigned from, unsigned to)
2737 {
2738 loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to;
2739
2740 set_page_extent_mapped(page);
2741 set_page_dirty(page);
2742
2743 if (pos > inode->i_size) {
2744 i_size_write(inode, pos);
2745 mark_inode_dirty(inode);
2746 }
2747 return 0;
2748 }
2749
2750 int extent_prepare_write(struct extent_io_tree *tree,
2751 struct inode *inode, struct page *page,
2752 unsigned from, unsigned to, get_extent_t *get_extent)
2753 {
2754 u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
2755 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
2756 u64 block_start;
2757 u64 orig_block_start;
2758 u64 block_end;
2759 u64 cur_end;
2760 struct extent_map *em;
2761 unsigned blocksize = 1 << inode->i_blkbits;
2762 size_t page_offset = 0;
2763 size_t block_off_start;
2764 size_t block_off_end;
2765 int err = 0;
2766 int iocount = 0;
2767 int ret = 0;
2768 int isnew;
2769
2770 set_page_extent_mapped(page);
2771
2772 block_start = (page_start + from) & ~((u64)blocksize - 1);
2773 block_end = (page_start + to - 1) | (blocksize - 1);
2774 orig_block_start = block_start;
2775
2776 lock_extent(tree, page_start, page_end, GFP_NOFS);
2777 while (block_start <= block_end) {
2778 em = get_extent(inode, page, page_offset, block_start,
2779 block_end - block_start + 1, 1);
2780 if (IS_ERR(em) || !em)
2781 goto err;
2782
2783 cur_end = min(block_end, extent_map_end(em) - 1);
2784 block_off_start = block_start & (PAGE_CACHE_SIZE - 1);
2785 block_off_end = block_off_start + blocksize;
2786 isnew = clear_extent_new(tree, block_start, cur_end, GFP_NOFS);
2787
2788 if (!PageUptodate(page) && isnew &&
2789 (block_off_end > to || block_off_start < from)) {
2790 void *kaddr;
2791
2792 kaddr = kmap_atomic(page, KM_USER0);
2793 if (block_off_end > to)
2794 memset(kaddr + to, 0, block_off_end - to);
2795 if (block_off_start < from)
2796 memset(kaddr + block_off_start, 0,
2797 from - block_off_start);
2798 flush_dcache_page(page);
2799 kunmap_atomic(kaddr, KM_USER0);
2800 }
2801 if ((em->block_start != EXTENT_MAP_HOLE &&
2802 em->block_start != EXTENT_MAP_INLINE) &&
2803 !isnew && !PageUptodate(page) &&
2804 (block_off_end > to || block_off_start < from) &&
2805 !test_range_bit(tree, block_start, cur_end,
2806 EXTENT_UPTODATE, 1, NULL)) {
2807 u64 sector;
2808 u64 extent_offset = block_start - em->start;
2809 size_t iosize;
2810 sector = (em->block_start + extent_offset) >> 9;
2811 iosize = (cur_end - block_start + blocksize) &
2812 ~((u64)blocksize - 1);
2813 /*
2814 * we've already got the extent locked, but we
2815 * need to split the state such that our end_bio
2816 * handler can clear the lock.
2817 */
2818 set_extent_bit(tree, block_start,
2819 block_start + iosize - 1,
2820 EXTENT_LOCKED, 0, NULL, NULL, GFP_NOFS);
2821 ret = submit_extent_page(READ, tree, page,
2822 sector, iosize, page_offset, em->bdev,
2823 NULL, 1,
2824 end_bio_extent_preparewrite, 0,
2825 0, 0);
2826 iocount++;
2827 block_start = block_start + iosize;
2828 } else {
2829 set_extent_uptodate(tree, block_start, cur_end,
2830 GFP_NOFS);
2831 unlock_extent(tree, block_start, cur_end, GFP_NOFS);
2832 block_start = cur_end + 1;
2833 }
2834 page_offset = block_start & (PAGE_CACHE_SIZE - 1);
2835 free_extent_map(em);
2836 }
2837 if (iocount) {
2838 wait_extent_bit(tree, orig_block_start,
2839 block_end, EXTENT_LOCKED);
2840 }
2841 check_page_uptodate(tree, page);
2842 err:
2843 /* FIXME, zero out newly allocated blocks on error */
2844 return err;
2845 }
2846
2847 /*
2848 * a helper for releasepage, this tests for areas of the page that
2849 * are locked or under IO and drops the related state bits if it is safe
2850 * to drop the page.
2851 */
2852 int try_release_extent_state(struct extent_map_tree *map,
2853 struct extent_io_tree *tree, struct page *page,
2854 gfp_t mask)
2855 {
2856 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
2857 u64 end = start + PAGE_CACHE_SIZE - 1;
2858 int ret = 1;
2859
2860 if (test_range_bit(tree, start, end,
2861 EXTENT_IOBITS, 0, NULL))
2862 ret = 0;
2863 else {
2864 if ((mask & GFP_NOFS) == GFP_NOFS)
2865 mask = GFP_NOFS;
2866 /*
2867 * at this point we can safely clear everything except the
2868 * locked bit and the nodatasum bit
2869 */
2870 clear_extent_bit(tree, start, end,
2871 ~(EXTENT_LOCKED | EXTENT_NODATASUM),
2872 0, 0, NULL, mask);
2873 }
2874 return ret;
2875 }
2876
2877 /*
2878 * a helper for releasepage. As long as there are no locked extents
2879 * in the range corresponding to the page, both state records and extent
2880 * map records are removed
2881 */
2882 int try_release_extent_mapping(struct extent_map_tree *map,
2883 struct extent_io_tree *tree, struct page *page,
2884 gfp_t mask)
2885 {
2886 struct extent_map *em;
2887 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
2888 u64 end = start + PAGE_CACHE_SIZE - 1;
2889
2890 if ((mask & __GFP_WAIT) &&
2891 page->mapping->host->i_size > 16 * 1024 * 1024) {
2892 u64 len;
2893 while (start <= end) {
2894 len = end - start + 1;
2895 write_lock(&map->lock);
2896 em = lookup_extent_mapping(map, start, len);
2897 if (!em || IS_ERR(em)) {
2898 write_unlock(&map->lock);
2899 break;
2900 }
2901 if (test_bit(EXTENT_FLAG_PINNED, &em->flags) ||
2902 em->start != start) {
2903 write_unlock(&map->lock);
2904 free_extent_map(em);
2905 break;
2906 }
2907 if (!test_range_bit(tree, em->start,
2908 extent_map_end(em) - 1,
2909 EXTENT_LOCKED | EXTENT_WRITEBACK,
2910 0, NULL)) {
2911 remove_extent_mapping(map, em);
2912 /* once for the rb tree */
2913 free_extent_map(em);
2914 }
2915 start = extent_map_end(em);
2916 write_unlock(&map->lock);
2917
2918 /* once for us */
2919 free_extent_map(em);
2920 }
2921 }
2922 return try_release_extent_state(map, tree, page, mask);
2923 }
2924
2925 sector_t extent_bmap(struct address_space *mapping, sector_t iblock,
2926 get_extent_t *get_extent)
2927 {
2928 struct inode *inode = mapping->host;
2929 struct extent_state *cached_state = NULL;
2930 u64 start = iblock << inode->i_blkbits;
2931 sector_t sector = 0;
2932 size_t blksize = (1 << inode->i_blkbits);
2933 struct extent_map *em;
2934
2935 lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + blksize - 1,
2936 0, &cached_state, GFP_NOFS);
2937 em = get_extent(inode, NULL, 0, start, blksize, 0);
2938 unlock_extent_cached(&BTRFS_I(inode)->io_tree, start,
2939 start + blksize - 1, &cached_state, GFP_NOFS);
2940 if (!em || IS_ERR(em))
2941 return 0;
2942
2943 if (em->block_start > EXTENT_MAP_LAST_BYTE)
2944 goto out;
2945
2946 sector = (em->block_start + start - em->start) >> inode->i_blkbits;
2947 out:
2948 free_extent_map(em);
2949 return sector;
2950 }
2951
2952 int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2953 __u64 start, __u64 len, get_extent_t *get_extent)
2954 {
2955 int ret;
2956 u64 off = start;
2957 u64 max = start + len;
2958 u32 flags = 0;
2959 u64 disko = 0;
2960 struct extent_map *em = NULL;
2961 struct extent_state *cached_state = NULL;
2962 int end = 0;
2963 u64 em_start = 0, em_len = 0;
2964 unsigned long emflags;
2965 ret = 0;
2966
2967 if (len == 0)
2968 return -EINVAL;
2969
2970 lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + len, 0,
2971 &cached_state, GFP_NOFS);
2972 em = get_extent(inode, NULL, 0, off, max - off, 0);
2973 if (!em)
2974 goto out;
2975 if (IS_ERR(em)) {
2976 ret = PTR_ERR(em);
2977 goto out;
2978 }
2979 while (!end) {
2980 off = em->start + em->len;
2981 if (off >= max)
2982 end = 1;
2983
2984 em_start = em->start;
2985 em_len = em->len;
2986
2987 disko = 0;
2988 flags = 0;
2989
2990 if (em->block_start == EXTENT_MAP_LAST_BYTE) {
2991 end = 1;
2992 flags |= FIEMAP_EXTENT_LAST;
2993 } else if (em->block_start == EXTENT_MAP_HOLE) {
2994 flags |= FIEMAP_EXTENT_UNWRITTEN;
2995 } else if (em->block_start == EXTENT_MAP_INLINE) {
2996 flags |= (FIEMAP_EXTENT_DATA_INLINE |
2997 FIEMAP_EXTENT_NOT_ALIGNED);
2998 } else if (em->block_start == EXTENT_MAP_DELALLOC) {
2999 flags |= (FIEMAP_EXTENT_DELALLOC |
3000 FIEMAP_EXTENT_UNKNOWN);
3001 } else {
3002 disko = em->block_start;
3003 }
3004 if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
3005 flags |= FIEMAP_EXTENT_ENCODED;
3006
3007 emflags = em->flags;
3008 free_extent_map(em);
3009 em = NULL;
3010
3011 if (!end) {
3012 em = get_extent(inode, NULL, 0, off, max - off, 0);
3013 if (!em)
3014 goto out;
3015 if (IS_ERR(em)) {
3016 ret = PTR_ERR(em);
3017 goto out;
3018 }
3019 emflags = em->flags;
3020 }
3021 if (test_bit(EXTENT_FLAG_VACANCY, &emflags)) {
3022 flags |= FIEMAP_EXTENT_LAST;
3023 end = 1;
3024 }
3025
3026 ret = fiemap_fill_next_extent(fieinfo, em_start, disko,
3027 em_len, flags);
3028 if (ret)
3029 goto out_free;
3030 }
3031 out_free:
3032 free_extent_map(em);
3033 out:
3034 unlock_extent_cached(&BTRFS_I(inode)->io_tree, start, start + len,
3035 &cached_state, GFP_NOFS);
3036 return ret;
3037 }
3038
3039 static inline struct page *extent_buffer_page(struct extent_buffer *eb,
3040 unsigned long i)
3041 {
3042 struct page *p;
3043 struct address_space *mapping;
3044
3045 if (i == 0)
3046 return eb->first_page;
3047 i += eb->start >> PAGE_CACHE_SHIFT;
3048 mapping = eb->first_page->mapping;
3049 if (!mapping)
3050 return NULL;
3051
3052 /*
3053 * extent_buffer_page is only called after pinning the page
3054 * by increasing the reference count. So we know the page must
3055 * be in the radix tree.
3056 */
3057 rcu_read_lock();
3058 p = radix_tree_lookup(&mapping->page_tree, i);
3059 rcu_read_unlock();
3060
3061 return p;
3062 }
3063
3064 static inline unsigned long num_extent_pages(u64 start, u64 len)
3065 {
3066 return ((start + len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT) -
3067 (start >> PAGE_CACHE_SHIFT);
3068 }
3069
3070 static struct extent_buffer *__alloc_extent_buffer(struct extent_io_tree *tree,
3071 u64 start,
3072 unsigned long len,
3073 gfp_t mask)
3074 {
3075 struct extent_buffer *eb = NULL;
3076 #if LEAK_DEBUG
3077 unsigned long flags;
3078 #endif
3079
3080 eb = kmem_cache_zalloc(extent_buffer_cache, mask);
3081 eb->start = start;
3082 eb->len = len;
3083 spin_lock_init(&eb->lock);
3084 init_waitqueue_head(&eb->lock_wq);
3085
3086 #if LEAK_DEBUG
3087 spin_lock_irqsave(&leak_lock, flags);
3088 list_add(&eb->leak_list, &buffers);
3089 spin_unlock_irqrestore(&leak_lock, flags);
3090 #endif
3091 atomic_set(&eb->refs, 1);
3092
3093 return eb;
3094 }
3095
3096 static void __free_extent_buffer(struct extent_buffer *eb)
3097 {
3098 #if LEAK_DEBUG
3099 unsigned long flags;
3100 spin_lock_irqsave(&leak_lock, flags);
3101 list_del(&eb->leak_list);
3102 spin_unlock_irqrestore(&leak_lock, flags);
3103 #endif
3104 kmem_cache_free(extent_buffer_cache, eb);
3105 }
3106
3107 struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree,
3108 u64 start, unsigned long len,
3109 struct page *page0,
3110 gfp_t mask)
3111 {
3112 unsigned long num_pages = num_extent_pages(start, len);
3113 unsigned long i;
3114 unsigned long index = start >> PAGE_CACHE_SHIFT;
3115 struct extent_buffer *eb;
3116 struct extent_buffer *exists = NULL;
3117 struct page *p;
3118 struct address_space *mapping = tree->mapping;
3119 int uptodate = 1;
3120
3121 spin_lock(&tree->buffer_lock);
3122 eb = buffer_search(tree, start);
3123 if (eb) {
3124 atomic_inc(&eb->refs);
3125 spin_unlock(&tree->buffer_lock);
3126 mark_page_accessed(eb->first_page);
3127 return eb;
3128 }
3129 spin_unlock(&tree->buffer_lock);
3130
3131 eb = __alloc_extent_buffer(tree, start, len, mask);
3132 if (!eb)
3133 return NULL;
3134
3135 if (page0) {
3136 eb->first_page = page0;
3137 i = 1;
3138 index++;
3139 page_cache_get(page0);
3140 mark_page_accessed(page0);
3141 set_page_extent_mapped(page0);
3142 set_page_extent_head(page0, len);
3143 uptodate = PageUptodate(page0);
3144 } else {
3145 i = 0;
3146 }
3147 for (; i < num_pages; i++, index++) {
3148 p = find_or_create_page(mapping, index, mask | __GFP_HIGHMEM);
3149 if (!p) {
3150 WARN_ON(1);
3151 goto free_eb;
3152 }
3153 set_page_extent_mapped(p);
3154 mark_page_accessed(p);
3155 if (i == 0) {
3156 eb->first_page = p;
3157 set_page_extent_head(p, len);
3158 } else {
3159 set_page_private(p, EXTENT_PAGE_PRIVATE);
3160 }
3161 if (!PageUptodate(p))
3162 uptodate = 0;
3163 unlock_page(p);
3164 }
3165 if (uptodate)
3166 set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
3167
3168 spin_lock(&tree->buffer_lock);
3169 exists = buffer_tree_insert(tree, start, &eb->rb_node);
3170 if (exists) {
3171 /* add one reference for the caller */
3172 atomic_inc(&exists->refs);
3173 spin_unlock(&tree->buffer_lock);
3174 goto free_eb;
3175 }
3176 /* add one reference for the tree */
3177 atomic_inc(&eb->refs);
3178 spin_unlock(&tree->buffer_lock);
3179 return eb;
3180
3181 free_eb:
3182 if (!atomic_dec_and_test(&eb->refs))
3183 return exists;
3184 for (index = 1; index < i; index++)
3185 page_cache_release(extent_buffer_page(eb, index));
3186 page_cache_release(extent_buffer_page(eb, 0));
3187 __free_extent_buffer(eb);
3188 return exists;
3189 }
3190
3191 struct extent_buffer *find_extent_buffer(struct extent_io_tree *tree,
3192 u64 start, unsigned long len,
3193 gfp_t mask)
3194 {
3195 struct extent_buffer *eb;
3196
3197 spin_lock(&tree->buffer_lock);
3198 eb = buffer_search(tree, start);
3199 if (eb)
3200 atomic_inc(&eb->refs);
3201 spin_unlock(&tree->buffer_lock);
3202
3203 if (eb)
3204 mark_page_accessed(eb->first_page);
3205
3206 return eb;
3207 }
3208
3209 void free_extent_buffer(struct extent_buffer *eb)
3210 {
3211 if (!eb)
3212 return;
3213
3214 if (!atomic_dec_and_test(&eb->refs))
3215 return;
3216
3217 WARN_ON(1);
3218 }
3219
3220 int clear_extent_buffer_dirty(struct extent_io_tree *tree,
3221 struct extent_buffer *eb)
3222 {
3223 unsigned long i;
3224 unsigned long num_pages;
3225 struct page *page;
3226
3227 num_pages = num_extent_pages(eb->start, eb->len);
3228
3229 for (i = 0; i < num_pages; i++) {
3230 page = extent_buffer_page(eb, i);
3231 if (!PageDirty(page))
3232 continue;
3233
3234 lock_page(page);
3235 if (i == 0)
3236 set_page_extent_head(page, eb->len);
3237 else
3238 set_page_private(page, EXTENT_PAGE_PRIVATE);
3239
3240 clear_page_dirty_for_io(page);
3241 spin_lock_irq(&page->mapping->tree_lock);
3242 if (!PageDirty(page)) {
3243 radix_tree_tag_clear(&page->mapping->page_tree,
3244 page_index(page),
3245 PAGECACHE_TAG_DIRTY);
3246 }
3247 spin_unlock_irq(&page->mapping->tree_lock);
3248 unlock_page(page);
3249 }
3250 return 0;
3251 }
3252
3253 int wait_on_extent_buffer_writeback(struct extent_io_tree *tree,
3254 struct extent_buffer *eb)
3255 {
3256 return wait_on_extent_writeback(tree, eb->start,
3257 eb->start + eb->len - 1);
3258 }
3259
3260 int set_extent_buffer_dirty(struct extent_io_tree *tree,
3261 struct extent_buffer *eb)
3262 {
3263 unsigned long i;
3264 unsigned long num_pages;
3265 int was_dirty = 0;
3266
3267 was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
3268 num_pages = num_extent_pages(eb->start, eb->len);
3269 for (i = 0; i < num_pages; i++)
3270 __set_page_dirty_nobuffers(extent_buffer_page(eb, i));
3271 return was_dirty;
3272 }
3273
3274 int clear_extent_buffer_uptodate(struct extent_io_tree *tree,
3275 struct extent_buffer *eb,
3276 struct extent_state **cached_state)
3277 {
3278 unsigned long i;
3279 struct page *page;
3280 unsigned long num_pages;
3281
3282 num_pages = num_extent_pages(eb->start, eb->len);
3283 clear_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
3284
3285 clear_extent_uptodate(tree, eb->start, eb->start + eb->len - 1,
3286 cached_state, GFP_NOFS);
3287 for (i = 0; i < num_pages; i++) {
3288 page = extent_buffer_page(eb, i);
3289 if (page)
3290 ClearPageUptodate(page);
3291 }
3292 return 0;
3293 }
3294
3295 int set_extent_buffer_uptodate(struct extent_io_tree *tree,
3296 struct extent_buffer *eb)
3297 {
3298 unsigned long i;
3299 struct page *page;
3300 unsigned long num_pages;
3301
3302 num_pages = num_extent_pages(eb->start, eb->len);
3303
3304 set_extent_uptodate(tree, eb->start, eb->start + eb->len - 1,
3305 GFP_NOFS);
3306 for (i = 0; i < num_pages; i++) {
3307 page = extent_buffer_page(eb, i);
3308 if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) ||
3309 ((i == num_pages - 1) &&
3310 ((eb->start + eb->len) & (PAGE_CACHE_SIZE - 1)))) {
3311 check_page_uptodate(tree, page);
3312 continue;
3313 }
3314 SetPageUptodate(page);
3315 }
3316 return 0;
3317 }
3318
3319 int extent_range_uptodate(struct extent_io_tree *tree,
3320 u64 start, u64 end)
3321 {
3322 struct page *page;
3323 int ret;
3324 int pg_uptodate = 1;
3325 int uptodate;
3326 unsigned long index;
3327
3328 ret = test_range_bit(tree, start, end, EXTENT_UPTODATE, 1, NULL);
3329 if (ret)
3330 return 1;
3331 while (start <= end) {
3332 index = start >> PAGE_CACHE_SHIFT;
3333 page = find_get_page(tree->mapping, index);
3334 uptodate = PageUptodate(page);
3335 page_cache_release(page);
3336 if (!uptodate) {
3337 pg_uptodate = 0;
3338 break;
3339 }
3340 start += PAGE_CACHE_SIZE;
3341 }
3342 return pg_uptodate;
3343 }
3344
3345 int extent_buffer_uptodate(struct extent_io_tree *tree,
3346 struct extent_buffer *eb,
3347 struct extent_state *cached_state)
3348 {
3349 int ret = 0;
3350 unsigned long num_pages;
3351 unsigned long i;
3352 struct page *page;
3353 int pg_uptodate = 1;
3354
3355 if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags))
3356 return 1;
3357
3358 ret = test_range_bit(tree, eb->start, eb->start + eb->len - 1,
3359 EXTENT_UPTODATE, 1, cached_state);
3360 if (ret)
3361 return ret;
3362
3363 num_pages = num_extent_pages(eb->start, eb->len);
3364 for (i = 0; i < num_pages; i++) {
3365 page = extent_buffer_page(eb, i);
3366 if (!PageUptodate(page)) {
3367 pg_uptodate = 0;
3368 break;
3369 }
3370 }
3371 return pg_uptodate;
3372 }
3373
3374 int read_extent_buffer_pages(struct extent_io_tree *tree,
3375 struct extent_buffer *eb,
3376 u64 start, int wait,
3377 get_extent_t *get_extent, int mirror_num)
3378 {
3379 unsigned long i;
3380 unsigned long start_i;
3381 struct page *page;
3382 int err;
3383 int ret = 0;
3384 int locked_pages = 0;
3385 int all_uptodate = 1;
3386 int inc_all_pages = 0;
3387 unsigned long num_pages;
3388 struct bio *bio = NULL;
3389 unsigned long bio_flags = 0;
3390
3391 if (test_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags))
3392 return 0;
3393
3394 if (test_range_bit(tree, eb->start, eb->start + eb->len - 1,
3395 EXTENT_UPTODATE, 1, NULL)) {
3396 return 0;
3397 }
3398
3399 if (start) {
3400 WARN_ON(start < eb->start);
3401 start_i = (start >> PAGE_CACHE_SHIFT) -
3402 (eb->start >> PAGE_CACHE_SHIFT);
3403 } else {
3404 start_i = 0;
3405 }
3406
3407 num_pages = num_extent_pages(eb->start, eb->len);
3408 for (i = start_i; i < num_pages; i++) {
3409 page = extent_buffer_page(eb, i);
3410 if (!wait) {
3411 if (!trylock_page(page))
3412 goto unlock_exit;
3413 } else {
3414 lock_page(page);
3415 }
3416 locked_pages++;
3417 if (!PageUptodate(page))
3418 all_uptodate = 0;
3419 }
3420 if (all_uptodate) {
3421 if (start_i == 0)
3422 set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
3423 goto unlock_exit;
3424 }
3425
3426 for (i = start_i; i < num_pages; i++) {
3427 page = extent_buffer_page(eb, i);
3428 if (inc_all_pages)
3429 page_cache_get(page);
3430 if (!PageUptodate(page)) {
3431 if (start_i == 0)
3432 inc_all_pages = 1;
3433 ClearPageError(page);
3434 err = __extent_read_full_page(tree, page,
3435 get_extent, &bio,
3436 mirror_num, &bio_flags);
3437 if (err)
3438 ret = err;
3439 } else {
3440 unlock_page(page);
3441 }
3442 }
3443
3444 if (bio)
3445 submit_one_bio(READ, bio, mirror_num, bio_flags);
3446
3447 if (ret || !wait)
3448 return ret;
3449
3450 for (i = start_i; i < num_pages; i++) {
3451 page = extent_buffer_page(eb, i);
3452 wait_on_page_locked(page);
3453 if (!PageUptodate(page))
3454 ret = -EIO;
3455 }
3456
3457 if (!ret)
3458 set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
3459 return ret;
3460
3461 unlock_exit:
3462 i = start_i;
3463 while (locked_pages > 0) {
3464 page = extent_buffer_page(eb, i);
3465 i++;
3466 unlock_page(page);
3467 locked_pages--;
3468 }
3469 return ret;
3470 }
3471
3472 void read_extent_buffer(struct extent_buffer *eb, void *dstv,
3473 unsigned long start,
3474 unsigned long len)
3475 {
3476 size_t cur;
3477 size_t offset;
3478 struct page *page;
3479 char *kaddr;
3480 char *dst = (char *)dstv;
3481 size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
3482 unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
3483
3484 WARN_ON(start > eb->len);
3485 WARN_ON(start + len > eb->start + eb->len);
3486
3487 offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1);
3488
3489 while (len > 0) {
3490 page = extent_buffer_page(eb, i);
3491
3492 cur = min(len, (PAGE_CACHE_SIZE - offset));
3493 kaddr = kmap_atomic(page, KM_USER1);
3494 memcpy(dst, kaddr + offset, cur);
3495 kunmap_atomic(kaddr, KM_USER1);
3496
3497 dst += cur;
3498 len -= cur;
3499 offset = 0;
3500 i++;
3501 }
3502 }
3503
3504 int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start,
3505 unsigned long min_len, char **token, char **map,
3506 unsigned long *map_start,
3507 unsigned long *map_len, int km)
3508 {
3509 size_t offset = start & (PAGE_CACHE_SIZE - 1);
3510 char *kaddr;
3511 struct page *p;
3512 size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
3513 unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
3514 unsigned long end_i = (start_offset + start + min_len - 1) >>
3515 PAGE_CACHE_SHIFT;
3516
3517 if (i != end_i)
3518 return -EINVAL;
3519
3520 if (i == 0) {
3521 offset = start_offset;
3522 *map_start = 0;
3523 } else {
3524 offset = 0;
3525 *map_start = ((u64)i << PAGE_CACHE_SHIFT) - start_offset;
3526 }
3527
3528 if (start + min_len > eb->len) {
3529 printk(KERN_ERR "btrfs bad mapping eb start %llu len %lu, "
3530 "wanted %lu %lu\n", (unsigned long long)eb->start,
3531 eb->len, start, min_len);
3532 WARN_ON(1);
3533 }
3534
3535 p = extent_buffer_page(eb, i);
3536 kaddr = kmap_atomic(p, km);
3537 *token = kaddr;
3538 *map = kaddr + offset;
3539 *map_len = PAGE_CACHE_SIZE - offset;
3540 return 0;
3541 }
3542
3543 int map_extent_buffer(struct extent_buffer *eb, unsigned long start,
3544 unsigned long min_len,
3545 char **token, char **map,
3546 unsigned long *map_start,
3547 unsigned long *map_len, int km)
3548 {
3549 int err;
3550 int save = 0;
3551 if (eb->map_token) {
3552 unmap_extent_buffer(eb, eb->map_token, km);
3553 eb->map_token = NULL;
3554 save = 1;
3555 }
3556 err = map_private_extent_buffer(eb, start, min_len, token, map,
3557 map_start, map_len, km);
3558 if (!err && save) {
3559 eb->map_token = *token;
3560 eb->kaddr = *map;
3561 eb->map_start = *map_start;
3562 eb->map_len = *map_len;
3563 }
3564 return err;
3565 }
3566
3567 void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km)
3568 {
3569 kunmap_atomic(token, km);
3570 }
3571
3572 int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv,
3573 unsigned long start,
3574 unsigned long len)
3575 {
3576 size_t cur;
3577 size_t offset;
3578 struct page *page;
3579 char *kaddr;
3580 char *ptr = (char *)ptrv;
3581 size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
3582 unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
3583 int ret = 0;
3584
3585 WARN_ON(start > eb->len);
3586 WARN_ON(start + len > eb->start + eb->len);
3587
3588 offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1);
3589
3590 while (len > 0) {
3591 page = extent_buffer_page(eb, i);
3592
3593 cur = min(len, (PAGE_CACHE_SIZE - offset));
3594
3595 kaddr = kmap_atomic(page, KM_USER0);
3596 ret = memcmp(ptr, kaddr + offset, cur);
3597 kunmap_atomic(kaddr, KM_USER0);
3598 if (ret)
3599 break;
3600
3601 ptr += cur;
3602 len -= cur;
3603 offset = 0;
3604 i++;
3605 }
3606 return ret;
3607 }
3608
3609 void write_extent_buffer(struct extent_buffer *eb, const void *srcv,
3610 unsigned long start, unsigned long len)
3611 {
3612 size_t cur;
3613 size_t offset;
3614 struct page *page;
3615 char *kaddr;
3616 char *src = (char *)srcv;
3617 size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
3618 unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
3619
3620 WARN_ON(start > eb->len);
3621 WARN_ON(start + len > eb->start + eb->len);
3622
3623 offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1);
3624
3625 while (len > 0) {
3626 page = extent_buffer_page(eb, i);
3627 WARN_ON(!PageUptodate(page));
3628
3629 cur = min(len, PAGE_CACHE_SIZE - offset);
3630 kaddr = kmap_atomic(page, KM_USER1);
3631 memcpy(kaddr + offset, src, cur);
3632 kunmap_atomic(kaddr, KM_USER1);
3633
3634 src += cur;
3635 len -= cur;
3636 offset = 0;
3637 i++;
3638 }
3639 }
3640
3641 void memset_extent_buffer(struct extent_buffer *eb, char c,
3642 unsigned long start, unsigned long len)
3643 {
3644 size_t cur;
3645 size_t offset;
3646 struct page *page;
3647 char *kaddr;
3648 size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
3649 unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
3650
3651 WARN_ON(start > eb->len);
3652 WARN_ON(start + len > eb->start + eb->len);
3653
3654 offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1);
3655
3656 while (len > 0) {
3657 page = extent_buffer_page(eb, i);
3658 WARN_ON(!PageUptodate(page));
3659
3660 cur = min(len, PAGE_CACHE_SIZE - offset);
3661 kaddr = kmap_atomic(page, KM_USER0);
3662 memset(kaddr + offset, c, cur);
3663 kunmap_atomic(kaddr, KM_USER0);
3664
3665 len -= cur;
3666 offset = 0;
3667 i++;
3668 }
3669 }
3670
3671 void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src,
3672 unsigned long dst_offset, unsigned long src_offset,
3673 unsigned long len)
3674 {
3675 u64 dst_len = dst->len;
3676 size_t cur;
3677 size_t offset;
3678 struct page *page;
3679 char *kaddr;
3680 size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1);
3681 unsigned long i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT;
3682
3683 WARN_ON(src->len != dst_len);
3684
3685 offset = (start_offset + dst_offset) &
3686 ((unsigned long)PAGE_CACHE_SIZE - 1);
3687
3688 while (len > 0) {
3689 page = extent_buffer_page(dst, i);
3690 WARN_ON(!PageUptodate(page));
3691
3692 cur = min(len, (unsigned long)(PAGE_CACHE_SIZE - offset));
3693
3694 kaddr = kmap_atomic(page, KM_USER0);
3695 read_extent_buffer(src, kaddr + offset, src_offset, cur);
3696 kunmap_atomic(kaddr, KM_USER0);
3697
3698 src_offset += cur;
3699 len -= cur;
3700 offset = 0;
3701 i++;
3702 }
3703 }
3704
3705 static void move_pages(struct page *dst_page, struct page *src_page,
3706 unsigned long dst_off, unsigned long src_off,
3707 unsigned long len)
3708 {
3709 char *dst_kaddr = kmap_atomic(dst_page, KM_USER0);
3710 if (dst_page == src_page) {
3711 memmove(dst_kaddr + dst_off, dst_kaddr + src_off, len);
3712 } else {
3713 char *src_kaddr = kmap_atomic(src_page, KM_USER1);
3714 char *p = dst_kaddr + dst_off + len;
3715 char *s = src_kaddr + src_off + len;
3716
3717 while (len--)
3718 *--p = *--s;
3719
3720 kunmap_atomic(src_kaddr, KM_USER1);
3721 }
3722 kunmap_atomic(dst_kaddr, KM_USER0);
3723 }
3724
3725 static void copy_pages(struct page *dst_page, struct page *src_page,
3726 unsigned long dst_off, unsigned long src_off,
3727 unsigned long len)
3728 {
3729 char *dst_kaddr = kmap_atomic(dst_page, KM_USER0);
3730 char *src_kaddr;
3731
3732 if (dst_page != src_page)
3733 src_kaddr = kmap_atomic(src_page, KM_USER1);
3734 else
3735 src_kaddr = dst_kaddr;
3736
3737 memcpy(dst_kaddr + dst_off, src_kaddr + src_off, len);
3738 kunmap_atomic(dst_kaddr, KM_USER0);
3739 if (dst_page != src_page)
3740 kunmap_atomic(src_kaddr, KM_USER1);
3741 }
3742
3743 void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
3744 unsigned long src_offset, unsigned long len)
3745 {
3746 size_t cur;
3747 size_t dst_off_in_page;
3748 size_t src_off_in_page;
3749 size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1);
3750 unsigned long dst_i;
3751 unsigned long src_i;
3752
3753 if (src_offset + len > dst->len) {
3754 printk(KERN_ERR "btrfs memmove bogus src_offset %lu move "
3755 "len %lu dst len %lu\n", src_offset, len, dst->len);
3756 BUG_ON(1);
3757 }
3758 if (dst_offset + len > dst->len) {
3759 printk(KERN_ERR "btrfs memmove bogus dst_offset %lu move "
3760 "len %lu dst len %lu\n", dst_offset, len, dst->len);
3761 BUG_ON(1);
3762 }
3763
3764 while (len > 0) {
3765 dst_off_in_page = (start_offset + dst_offset) &
3766 ((unsigned long)PAGE_CACHE_SIZE - 1);
3767 src_off_in_page = (start_offset + src_offset) &
3768 ((unsigned long)PAGE_CACHE_SIZE - 1);
3769
3770 dst_i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT;
3771 src_i = (start_offset + src_offset) >> PAGE_CACHE_SHIFT;
3772
3773 cur = min(len, (unsigned long)(PAGE_CACHE_SIZE -
3774 src_off_in_page));
3775 cur = min_t(unsigned long, cur,
3776 (unsigned long)(PAGE_CACHE_SIZE - dst_off_in_page));
3777
3778 copy_pages(extent_buffer_page(dst, dst_i),
3779 extent_buffer_page(dst, src_i),
3780 dst_off_in_page, src_off_in_page, cur);
3781
3782 src_offset += cur;
3783 dst_offset += cur;
3784 len -= cur;
3785 }
3786 }
3787
3788 void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
3789 unsigned long src_offset, unsigned long len)
3790 {
3791 size_t cur;
3792 size_t dst_off_in_page;
3793 size_t src_off_in_page;
3794 unsigned long dst_end = dst_offset + len - 1;
3795 unsigned long src_end = src_offset + len - 1;
3796 size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1);
3797 unsigned long dst_i;
3798 unsigned long src_i;
3799
3800 if (src_offset + len > dst->len) {
3801 printk(KERN_ERR "btrfs memmove bogus src_offset %lu move "
3802 "len %lu len %lu\n", src_offset, len, dst->len);
3803 BUG_ON(1);
3804 }
3805 if (dst_offset + len > dst->len) {
3806 printk(KERN_ERR "btrfs memmove bogus dst_offset %lu move "
3807 "len %lu len %lu\n", dst_offset, len, dst->len);
3808 BUG_ON(1);
3809 }
3810 if (dst_offset < src_offset) {
3811 memcpy_extent_buffer(dst, dst_offset, src_offset, len);
3812 return;
3813 }
3814 while (len > 0) {
3815 dst_i = (start_offset + dst_end) >> PAGE_CACHE_SHIFT;
3816 src_i = (start_offset + src_end) >> PAGE_CACHE_SHIFT;
3817
3818 dst_off_in_page = (start_offset + dst_end) &
3819 ((unsigned long)PAGE_CACHE_SIZE - 1);
3820 src_off_in_page = (start_offset + src_end) &
3821 ((unsigned long)PAGE_CACHE_SIZE - 1);
3822
3823 cur = min_t(unsigned long, len, src_off_in_page + 1);
3824 cur = min(cur, dst_off_in_page + 1);
3825 move_pages(extent_buffer_page(dst, dst_i),
3826 extent_buffer_page(dst, src_i),
3827 dst_off_in_page - cur + 1,
3828 src_off_in_page - cur + 1, cur);
3829
3830 dst_end -= cur;
3831 src_end -= cur;
3832 len -= cur;
3833 }
3834 }
3835
3836 int try_release_extent_buffer(struct extent_io_tree *tree, struct page *page)
3837 {
3838 u64 start = page_offset(page);
3839 struct extent_buffer *eb;
3840 int ret = 1;
3841 unsigned long i;
3842 unsigned long num_pages;
3843
3844 spin_lock(&tree->buffer_lock);
3845 eb = buffer_search(tree, start);
3846 if (!eb)
3847 goto out;
3848
3849 if (atomic_read(&eb->refs) > 1) {
3850 ret = 0;
3851 goto out;
3852 }
3853 if (test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
3854 ret = 0;
3855 goto out;
3856 }
3857 /* at this point we can safely release the extent buffer */
3858 num_pages = num_extent_pages(eb->start, eb->len);
3859 for (i = 0; i < num_pages; i++)
3860 page_cache_release(extent_buffer_page(eb, i));
3861 rb_erase(&eb->rb_node, &tree->buffer);
3862 __free_extent_buffer(eb);
3863 out:
3864 spin_unlock(&tree->buffer_lock);
3865 return ret;
3866 }
linux-next