1 #include <linux/bitops.h>
2 #include <linux/slab.h>
6 #include <linux/pagemap.h>
7 #include <linux/page-flags.h>
8 #include <linux/module.h>
9 #include <linux/spinlock.h>
10 #include <linux/blkdev.h>
11 #include <linux/swap.h>
12 #include <linux/version.h>
13 #include "extent_map.h"
15 /* temporary define until extent_map moves out of btrfs */
16 struct kmem_cache *btrfs_cache_create(const char *name, size_t size,
17 unsigned long extra_flags,
18 void (*ctor)(void *, struct kmem_cache *,
21 static struct kmem_cache *extent_map_cache;
22 static struct kmem_cache *extent_state_cache;
23 static struct kmem_cache *extent_buffer_cache;
25 static LIST_HEAD(buffers);
26 static LIST_HEAD(states);
28 static spinlock_t state_lock = SPIN_LOCK_UNLOCKED;
29 #define BUFFER_LRU_MAX 64
35 struct rb_node rb_node;
38 void __init extent_map_init(void)
40 extent_map_cache = btrfs_cache_create("extent_map",
41 sizeof(struct extent_map), 0,
43 extent_state_cache = btrfs_cache_create("extent_state",
44 sizeof(struct extent_state), 0,
46 extent_buffer_cache = btrfs_cache_create("extent_buffers",
47 sizeof(struct extent_buffer), 0,
51 void __exit extent_map_exit(void)
53 struct extent_state *state;
55 while (!list_empty(&states)) {
56 state = list_entry(states.next, struct extent_state, list);
57 printk("state leak: start %Lu end %Lu state %lu in tree %d refs %d\n", state->start, state->end, state->state, state->in_tree, atomic_read(&state->refs));
58 list_del(&state->list);
59 kmem_cache_free(extent_state_cache, state);
64 kmem_cache_destroy(extent_map_cache);
65 if (extent_state_cache)
66 kmem_cache_destroy(extent_state_cache);
67 if (extent_buffer_cache)
68 kmem_cache_destroy(extent_buffer_cache);
71 void extent_map_tree_init(struct extent_map_tree *tree,
72 struct address_space *mapping, gfp_t mask)
74 tree->map.rb_node = NULL;
75 tree->state.rb_node = NULL;
77 rwlock_init(&tree->lock);
78 spin_lock_init(&tree->lru_lock);
79 tree->mapping = mapping;
80 INIT_LIST_HEAD(&tree->buffer_lru);
83 EXPORT_SYMBOL(extent_map_tree_init);
85 void extent_map_tree_empty_lru(struct extent_map_tree *tree)
87 struct extent_buffer *eb;
88 while(!list_empty(&tree->buffer_lru)) {
89 eb = list_entry(tree->buffer_lru.next, struct extent_buffer,
92 free_extent_buffer(eb);
95 EXPORT_SYMBOL(extent_map_tree_empty_lru);
97 struct extent_map *alloc_extent_map(gfp_t mask)
99 struct extent_map *em;
100 em = kmem_cache_alloc(extent_map_cache, mask);
101 if (!em || IS_ERR(em))
104 atomic_set(&em->refs, 1);
107 EXPORT_SYMBOL(alloc_extent_map);
109 void free_extent_map(struct extent_map *em)
113 if (atomic_dec_and_test(&em->refs)) {
114 WARN_ON(em->in_tree);
115 kmem_cache_free(extent_map_cache, em);
118 EXPORT_SYMBOL(free_extent_map);
121 struct extent_state *alloc_extent_state(gfp_t mask)
123 struct extent_state *state;
126 state = kmem_cache_alloc(extent_state_cache, mask);
127 if (!state || IS_ERR(state))
133 spin_lock_irqsave(&state_lock, flags);
134 list_add(&state->list, &states);
135 spin_unlock_irqrestore(&state_lock, flags);
137 atomic_set(&state->refs, 1);
138 init_waitqueue_head(&state->wq);
141 EXPORT_SYMBOL(alloc_extent_state);
143 void free_extent_state(struct extent_state *state)
148 if (atomic_dec_and_test(&state->refs)) {
149 WARN_ON(state->in_tree);
150 spin_lock_irqsave(&state_lock, flags);
151 list_del(&state->list);
152 spin_unlock_irqrestore(&state_lock, flags);
153 kmem_cache_free(extent_state_cache, state);
156 EXPORT_SYMBOL(free_extent_state);
158 static struct rb_node *tree_insert(struct rb_root *root, u64 offset,
159 struct rb_node *node)
161 struct rb_node ** p = &root->rb_node;
162 struct rb_node * parent = NULL;
163 struct tree_entry *entry;
167 entry = rb_entry(parent, struct tree_entry, rb_node);
169 if (offset < entry->start)
171 else if (offset > entry->end)
177 entry = rb_entry(node, struct tree_entry, rb_node);
179 rb_link_node(node, parent, p);
180 rb_insert_color(node, root);
184 static struct rb_node *__tree_search(struct rb_root *root, u64 offset,
185 struct rb_node **prev_ret)
187 struct rb_node * n = root->rb_node;
188 struct rb_node *prev = NULL;
189 struct tree_entry *entry;
190 struct tree_entry *prev_entry = NULL;
193 entry = rb_entry(n, struct tree_entry, rb_node);
197 if (offset < entry->start)
199 else if (offset > entry->end)
206 while(prev && offset > prev_entry->end) {
207 prev = rb_next(prev);
208 prev_entry = rb_entry(prev, struct tree_entry, rb_node);
214 static inline struct rb_node *tree_search(struct rb_root *root, u64 offset)
216 struct rb_node *prev;
218 ret = __tree_search(root, offset, &prev);
224 static int tree_delete(struct rb_root *root, u64 offset)
226 struct rb_node *node;
227 struct tree_entry *entry;
229 node = __tree_search(root, offset, NULL);
232 entry = rb_entry(node, struct tree_entry, rb_node);
234 rb_erase(node, root);
239 * add_extent_mapping tries a simple backward merge with existing
240 * mappings. The extent_map struct passed in will be inserted into
241 * the tree directly (no copies made, just a reference taken).
243 int add_extent_mapping(struct extent_map_tree *tree,
244 struct extent_map *em)
247 struct extent_map *prev = NULL;
250 write_lock_irq(&tree->lock);
251 rb = tree_insert(&tree->map, em->end, &em->rb_node);
253 prev = rb_entry(rb, struct extent_map, rb_node);
254 printk("found extent map %Lu %Lu on insert of %Lu %Lu\n", prev->start, prev->end, em->start, em->end);
258 atomic_inc(&em->refs);
259 if (em->start != 0) {
260 rb = rb_prev(&em->rb_node);
262 prev = rb_entry(rb, struct extent_map, rb_node);
263 if (prev && prev->end + 1 == em->start &&
264 ((em->block_start == EXTENT_MAP_HOLE &&
265 prev->block_start == EXTENT_MAP_HOLE) ||
266 (em->block_start == EXTENT_MAP_INLINE &&
267 prev->block_start == EXTENT_MAP_INLINE) ||
268 (em->block_start == EXTENT_MAP_DELALLOC &&
269 prev->block_start == EXTENT_MAP_DELALLOC) ||
270 (em->block_start < EXTENT_MAP_DELALLOC - 1 &&
271 em->block_start == prev->block_end + 1))) {
272 em->start = prev->start;
273 em->block_start = prev->block_start;
274 rb_erase(&prev->rb_node, &tree->map);
276 free_extent_map(prev);
280 write_unlock_irq(&tree->lock);
283 EXPORT_SYMBOL(add_extent_mapping);
286 * lookup_extent_mapping returns the first extent_map struct in the
287 * tree that intersects the [start, end] (inclusive) range. There may
288 * be additional objects in the tree that intersect, so check the object
289 * returned carefully to make sure you don't need additional lookups.
291 struct extent_map *lookup_extent_mapping(struct extent_map_tree *tree,
294 struct extent_map *em;
295 struct rb_node *rb_node;
297 read_lock_irq(&tree->lock);
298 rb_node = tree_search(&tree->map, start);
303 if (IS_ERR(rb_node)) {
304 em = ERR_PTR(PTR_ERR(rb_node));
307 em = rb_entry(rb_node, struct extent_map, rb_node);
308 if (em->end < start || em->start > end) {
312 atomic_inc(&em->refs);
314 read_unlock_irq(&tree->lock);
317 EXPORT_SYMBOL(lookup_extent_mapping);
320 * removes an extent_map struct from the tree. No reference counts are
321 * dropped, and no checks are done to see if the range is in use
323 int remove_extent_mapping(struct extent_map_tree *tree, struct extent_map *em)
327 write_lock_irq(&tree->lock);
328 ret = tree_delete(&tree->map, em->end);
329 write_unlock_irq(&tree->lock);
332 EXPORT_SYMBOL(remove_extent_mapping);
335 * utility function to look for merge candidates inside a given range.
336 * Any extents with matching state are merged together into a single
337 * extent in the tree. Extents with EXTENT_IO in their state field
338 * are not merged because the end_io handlers need to be able to do
339 * operations on them without sleeping (or doing allocations/splits).
341 * This should be called with the tree lock held.
343 static int merge_state(struct extent_map_tree *tree,
344 struct extent_state *state)
346 struct extent_state *other;
347 struct rb_node *other_node;
349 if (state->state & EXTENT_IOBITS)
352 other_node = rb_prev(&state->rb_node);
354 other = rb_entry(other_node, struct extent_state, rb_node);
355 if (other->end == state->start - 1 &&
356 other->state == state->state) {
357 state->start = other->start;
359 rb_erase(&other->rb_node, &tree->state);
360 free_extent_state(other);
363 other_node = rb_next(&state->rb_node);
365 other = rb_entry(other_node, struct extent_state, rb_node);
366 if (other->start == state->end + 1 &&
367 other->state == state->state) {
368 other->start = state->start;
370 rb_erase(&state->rb_node, &tree->state);
371 free_extent_state(state);
378 * insert an extent_state struct into the tree. 'bits' are set on the
379 * struct before it is inserted.
381 * This may return -EEXIST if the extent is already there, in which case the
382 * state struct is freed.
384 * The tree lock is not taken internally. This is a utility function and
385 * probably isn't what you want to call (see set/clear_extent_bit).
387 static int insert_state(struct extent_map_tree *tree,
388 struct extent_state *state, u64 start, u64 end,
391 struct rb_node *node;
394 printk("end < start %Lu %Lu\n", end, start);
397 state->state |= bits;
398 state->start = start;
400 node = tree_insert(&tree->state, end, &state->rb_node);
402 struct extent_state *found;
403 found = rb_entry(node, struct extent_state, rb_node);
404 printk("found node %Lu %Lu on insert of %Lu %Lu\n", found->start, found->end, start, end);
405 free_extent_state(state);
408 merge_state(tree, state);
413 * split a given extent state struct in two, inserting the preallocated
414 * struct 'prealloc' as the newly created second half. 'split' indicates an
415 * offset inside 'orig' where it should be split.
418 * the tree has 'orig' at [orig->start, orig->end]. After calling, there
419 * are two extent state structs in the tree:
420 * prealloc: [orig->start, split - 1]
421 * orig: [ split, orig->end ]
423 * The tree locks are not taken by this function. They need to be held
426 static int split_state(struct extent_map_tree *tree, struct extent_state *orig,
427 struct extent_state *prealloc, u64 split)
429 struct rb_node *node;
430 prealloc->start = orig->start;
431 prealloc->end = split - 1;
432 prealloc->state = orig->state;
435 node = tree_insert(&tree->state, prealloc->end, &prealloc->rb_node);
437 struct extent_state *found;
438 found = rb_entry(node, struct extent_state, rb_node);
439 printk("found node %Lu %Lu on insert of %Lu %Lu\n", found->start, found->end, prealloc->start, prealloc->end);
440 free_extent_state(prealloc);
447 * utility function to clear some bits in an extent state struct.
448 * it will optionally wake up any one waiting on this state (wake == 1), or
449 * forcibly remove the state from the tree (delete == 1).
451 * If no bits are set on the state struct after clearing things, the
452 * struct is freed and removed from the tree
454 static int clear_state_bit(struct extent_map_tree *tree,
455 struct extent_state *state, int bits, int wake,
458 int ret = state->state & bits;
459 state->state &= ~bits;
462 if (delete || state->state == 0) {
463 if (state->in_tree) {
464 rb_erase(&state->rb_node, &tree->state);
466 free_extent_state(state);
471 merge_state(tree, state);
477 * clear some bits on a range in the tree. This may require splitting
478 * or inserting elements in the tree, so the gfp mask is used to
479 * indicate which allocations or sleeping are allowed.
481 * pass 'wake' == 1 to kick any sleepers, and 'delete' == 1 to remove
482 * the given range from the tree regardless of state (ie for truncate).
484 * the range [start, end] is inclusive.
486 * This takes the tree lock, and returns < 0 on error, > 0 if any of the
487 * bits were already set, or zero if none of the bits were already set.
489 int clear_extent_bit(struct extent_map_tree *tree, u64 start, u64 end,
490 int bits, int wake, int delete, gfp_t mask)
492 struct extent_state *state;
493 struct extent_state *prealloc = NULL;
494 struct rb_node *node;
500 if (!prealloc && (mask & __GFP_WAIT)) {
501 prealloc = alloc_extent_state(mask);
506 write_lock_irqsave(&tree->lock, flags);
508 * this search will find the extents that end after
511 node = tree_search(&tree->state, start);
514 state = rb_entry(node, struct extent_state, rb_node);
515 if (state->start > end)
517 WARN_ON(state->end < start);
520 * | ---- desired range ---- |
522 * | ------------- state -------------- |
524 * We need to split the extent we found, and may flip
525 * bits on second half.
527 * If the extent we found extends past our range, we
528 * just split and search again. It'll get split again
529 * the next time though.
531 * If the extent we found is inside our range, we clear
532 * the desired bit on it.
535 if (state->start < start) {
536 err = split_state(tree, state, prealloc, start);
537 BUG_ON(err == -EEXIST);
541 if (state->end <= end) {
542 start = state->end + 1;
543 set |= clear_state_bit(tree, state, bits,
546 start = state->start;
551 * | ---- desired range ---- |
553 * We need to split the extent, and clear the bit
556 if (state->start <= end && state->end > end) {
557 err = split_state(tree, state, prealloc, end + 1);
558 BUG_ON(err == -EEXIST);
562 set |= clear_state_bit(tree, prealloc, bits,
568 start = state->end + 1;
569 set |= clear_state_bit(tree, state, bits, wake, delete);
573 write_unlock_irqrestore(&tree->lock, flags);
575 free_extent_state(prealloc);
582 write_unlock_irqrestore(&tree->lock, flags);
583 if (mask & __GFP_WAIT)
587 EXPORT_SYMBOL(clear_extent_bit);
589 static int wait_on_state(struct extent_map_tree *tree,
590 struct extent_state *state)
593 prepare_to_wait(&state->wq, &wait, TASK_UNINTERRUPTIBLE);
594 read_unlock_irq(&tree->lock);
596 read_lock_irq(&tree->lock);
597 finish_wait(&state->wq, &wait);
602 * waits for one or more bits to clear on a range in the state tree.
603 * The range [start, end] is inclusive.
604 * The tree lock is taken by this function
606 int wait_extent_bit(struct extent_map_tree *tree, u64 start, u64 end, int bits)
608 struct extent_state *state;
609 struct rb_node *node;
611 read_lock_irq(&tree->lock);
615 * this search will find all the extents that end after
618 node = tree_search(&tree->state, start);
622 state = rb_entry(node, struct extent_state, rb_node);
624 if (state->start > end)
627 if (state->state & bits) {
628 start = state->start;
629 atomic_inc(&state->refs);
630 wait_on_state(tree, state);
631 free_extent_state(state);
634 start = state->end + 1;
639 if (need_resched()) {
640 read_unlock_irq(&tree->lock);
642 read_lock_irq(&tree->lock);
646 read_unlock_irq(&tree->lock);
649 EXPORT_SYMBOL(wait_extent_bit);
652 * set some bits on a range in the tree. This may require allocations
653 * or sleeping, so the gfp mask is used to indicate what is allowed.
655 * If 'exclusive' == 1, this will fail with -EEXIST if some part of the
656 * range already has the desired bits set. The start of the existing
657 * range is returned in failed_start in this case.
659 * [start, end] is inclusive
660 * This takes the tree lock.
662 int set_extent_bit(struct extent_map_tree *tree, u64 start, u64 end, int bits,
663 int exclusive, u64 *failed_start, gfp_t mask)
665 struct extent_state *state;
666 struct extent_state *prealloc = NULL;
667 struct rb_node *node;
674 if (!prealloc && (mask & __GFP_WAIT)) {
675 prealloc = alloc_extent_state(mask);
680 write_lock_irqsave(&tree->lock, flags);
682 * this search will find all the extents that end after
685 node = tree_search(&tree->state, start);
687 err = insert_state(tree, prealloc, start, end, bits);
689 BUG_ON(err == -EEXIST);
693 state = rb_entry(node, struct extent_state, rb_node);
694 last_start = state->start;
695 last_end = state->end;
698 * | ---- desired range ---- |
701 * Just lock what we found and keep going
703 if (state->start == start && state->end <= end) {
704 set = state->state & bits;
705 if (set && exclusive) {
706 *failed_start = state->start;
710 state->state |= bits;
711 start = state->end + 1;
712 merge_state(tree, state);
717 * | ---- desired range ---- |
720 * | ------------- state -------------- |
722 * We need to split the extent we found, and may flip bits on
725 * If the extent we found extends past our
726 * range, we just split and search again. It'll get split
727 * again the next time though.
729 * If the extent we found is inside our range, we set the
732 if (state->start < start) {
733 set = state->state & bits;
734 if (exclusive && set) {
735 *failed_start = start;
739 err = split_state(tree, state, prealloc, start);
740 BUG_ON(err == -EEXIST);
744 if (state->end <= end) {
745 state->state |= bits;
746 start = state->end + 1;
747 merge_state(tree, state);
749 start = state->start;
754 * | ---- desired range ---- |
755 * | state | or | state |
757 * There's a hole, we need to insert something in it and
758 * ignore the extent we found.
760 if (state->start > start) {
762 if (end < last_start)
765 this_end = last_start -1;
766 err = insert_state(tree, prealloc, start, this_end,
769 BUG_ON(err == -EEXIST);
772 start = this_end + 1;
776 * | ---- desired range ---- |
778 * We need to split the extent, and set the bit
781 if (state->start <= end && state->end > end) {
782 set = state->state & bits;
783 if (exclusive && set) {
784 *failed_start = start;
788 err = split_state(tree, state, prealloc, end + 1);
789 BUG_ON(err == -EEXIST);
791 prealloc->state |= bits;
792 merge_state(tree, prealloc);
800 write_unlock_irqrestore(&tree->lock, flags);
802 free_extent_state(prealloc);
809 write_unlock_irqrestore(&tree->lock, flags);
810 if (mask & __GFP_WAIT)
814 EXPORT_SYMBOL(set_extent_bit);
816 /* wrappers around set/clear extent bit */
817 int set_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end,
820 return set_extent_bit(tree, start, end, EXTENT_DIRTY, 0, NULL,
823 EXPORT_SYMBOL(set_extent_dirty);
825 int set_extent_bits(struct extent_map_tree *tree, u64 start, u64 end,
826 int bits, gfp_t mask)
828 return set_extent_bit(tree, start, end, bits, 0, NULL,
831 EXPORT_SYMBOL(set_extent_bits);
833 int clear_extent_bits(struct extent_map_tree *tree, u64 start, u64 end,
834 int bits, gfp_t mask)
836 return clear_extent_bit(tree, start, end, bits, 0, 0, mask);
838 EXPORT_SYMBOL(clear_extent_bits);
840 int set_extent_delalloc(struct extent_map_tree *tree, u64 start, u64 end,
843 return set_extent_bit(tree, start, end,
844 EXTENT_DELALLOC | EXTENT_DIRTY, 0, NULL,
847 EXPORT_SYMBOL(set_extent_delalloc);
849 int clear_extent_dirty(struct extent_map_tree *tree, u64 start, u64 end,
852 return clear_extent_bit(tree, start, end,
853 EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0, mask);
855 EXPORT_SYMBOL(clear_extent_dirty);
857 int set_extent_new(struct extent_map_tree *tree, u64 start, u64 end,
860 return set_extent_bit(tree, start, end, EXTENT_NEW, 0, NULL,
863 EXPORT_SYMBOL(set_extent_new);
865 int clear_extent_new(struct extent_map_tree *tree, u64 start, u64 end,
868 return clear_extent_bit(tree, start, end, EXTENT_NEW, 0, 0, mask);
870 EXPORT_SYMBOL(clear_extent_new);
872 int set_extent_uptodate(struct extent_map_tree *tree, u64 start, u64 end,
875 return set_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, NULL,
878 EXPORT_SYMBOL(set_extent_uptodate);
880 int clear_extent_uptodate(struct extent_map_tree *tree, u64 start, u64 end,
883 return clear_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, 0, mask);
885 EXPORT_SYMBOL(clear_extent_uptodate);
887 int set_extent_writeback(struct extent_map_tree *tree, u64 start, u64 end,
890 return set_extent_bit(tree, start, end, EXTENT_WRITEBACK,
893 EXPORT_SYMBOL(set_extent_writeback);
895 int clear_extent_writeback(struct extent_map_tree *tree, u64 start, u64 end,
898 return clear_extent_bit(tree, start, end, EXTENT_WRITEBACK, 1, 0, mask);
900 EXPORT_SYMBOL(clear_extent_writeback);
902 int wait_on_extent_writeback(struct extent_map_tree *tree, u64 start, u64 end)
904 return wait_extent_bit(tree, start, end, EXTENT_WRITEBACK);
906 EXPORT_SYMBOL(wait_on_extent_writeback);
909 * locks a range in ascending order, waiting for any locked regions
910 * it hits on the way. [start,end] are inclusive, and this will sleep.
912 int lock_extent(struct extent_map_tree *tree, u64 start, u64 end, gfp_t mask)
917 err = set_extent_bit(tree, start, end, EXTENT_LOCKED, 1,
918 &failed_start, mask);
919 if (err == -EEXIST && (mask & __GFP_WAIT)) {
920 wait_extent_bit(tree, failed_start, end, EXTENT_LOCKED);
921 start = failed_start;
925 WARN_ON(start > end);
929 EXPORT_SYMBOL(lock_extent);
931 int unlock_extent(struct extent_map_tree *tree, u64 start, u64 end,
934 return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, mask);
936 EXPORT_SYMBOL(unlock_extent);
939 * helper function to set pages and extents in the tree dirty
941 int set_range_dirty(struct extent_map_tree *tree, u64 start, u64 end)
943 unsigned long index = start >> PAGE_CACHE_SHIFT;
944 unsigned long end_index = end >> PAGE_CACHE_SHIFT;
947 while (index <= end_index) {
948 page = find_get_page(tree->mapping, index);
950 __set_page_dirty_nobuffers(page);
951 page_cache_release(page);
954 set_extent_dirty(tree, start, end, GFP_NOFS);
957 EXPORT_SYMBOL(set_range_dirty);
960 * helper function to set both pages and extents in the tree writeback
962 int set_range_writeback(struct extent_map_tree *tree, u64 start, u64 end)
964 unsigned long index = start >> PAGE_CACHE_SHIFT;
965 unsigned long end_index = end >> PAGE_CACHE_SHIFT;
968 while (index <= end_index) {
969 page = find_get_page(tree->mapping, index);
971 set_page_writeback(page);
972 page_cache_release(page);
975 set_extent_writeback(tree, start, end, GFP_NOFS);
978 EXPORT_SYMBOL(set_range_writeback);
980 int find_first_extent_bit(struct extent_map_tree *tree, u64 start,
981 u64 *start_ret, u64 *end_ret, int bits)
983 struct rb_node *node;
984 struct extent_state *state;
987 read_lock_irq(&tree->lock);
989 * this search will find all the extents that end after
992 node = tree_search(&tree->state, start);
993 if (!node || IS_ERR(node)) {
998 state = rb_entry(node, struct extent_state, rb_node);
999 if (state->end >= start && (state->state & bits)) {
1000 *start_ret = state->start;
1001 *end_ret = state->end;
1005 node = rb_next(node);
1010 read_unlock_irq(&tree->lock);
1013 EXPORT_SYMBOL(find_first_extent_bit);
1015 u64 find_lock_delalloc_range(struct extent_map_tree *tree,
1016 u64 start, u64 lock_start, u64 *end, u64 max_bytes)
1018 struct rb_node *node;
1019 struct extent_state *state;
1020 u64 cur_start = start;
1022 u64 total_bytes = 0;
1024 write_lock_irq(&tree->lock);
1026 * this search will find all the extents that end after
1030 node = tree_search(&tree->state, cur_start);
1031 if (!node || IS_ERR(node)) {
1036 state = rb_entry(node, struct extent_state, rb_node);
1037 if (state->start != cur_start) {
1040 if (!(state->state & EXTENT_DELALLOC)) {
1043 if (state->start >= lock_start) {
1044 if (state->state & EXTENT_LOCKED) {
1046 atomic_inc(&state->refs);
1047 prepare_to_wait(&state->wq, &wait,
1048 TASK_UNINTERRUPTIBLE);
1049 write_unlock_irq(&tree->lock);
1051 write_lock_irq(&tree->lock);
1052 finish_wait(&state->wq, &wait);
1053 free_extent_state(state);
1056 state->state |= EXTENT_LOCKED;
1060 cur_start = state->end + 1;
1061 node = rb_next(node);
1064 total_bytes += state->end - state->start + 1;
1065 if (total_bytes >= max_bytes)
1069 write_unlock_irq(&tree->lock);
1074 * helper function to lock both pages and extents in the tree.
1075 * pages must be locked first.
1077 int lock_range(struct extent_map_tree *tree, u64 start, u64 end)
1079 unsigned long index = start >> PAGE_CACHE_SHIFT;
1080 unsigned long end_index = end >> PAGE_CACHE_SHIFT;
1084 while (index <= end_index) {
1085 page = grab_cache_page(tree->mapping, index);
1091 err = PTR_ERR(page);
1096 lock_extent(tree, start, end, GFP_NOFS);
1101 * we failed above in getting the page at 'index', so we undo here
1102 * up to but not including the page at 'index'
1105 index = start >> PAGE_CACHE_SHIFT;
1106 while (index < end_index) {
1107 page = find_get_page(tree->mapping, index);
1109 page_cache_release(page);
1114 EXPORT_SYMBOL(lock_range);
1117 * helper function to unlock both pages and extents in the tree.
1119 int unlock_range(struct extent_map_tree *tree, u64 start, u64 end)
1121 unsigned long index = start >> PAGE_CACHE_SHIFT;
1122 unsigned long end_index = end >> PAGE_CACHE_SHIFT;
1125 while (index <= end_index) {
1126 page = find_get_page(tree->mapping, index);
1128 page_cache_release(page);
1131 unlock_extent(tree, start, end, GFP_NOFS);
1134 EXPORT_SYMBOL(unlock_range);
1136 int set_state_private(struct extent_map_tree *tree, u64 start, u64 private)
1138 struct rb_node *node;
1139 struct extent_state *state;
1142 write_lock_irq(&tree->lock);
1144 * this search will find all the extents that end after
1147 node = tree_search(&tree->state, start);
1148 if (!node || IS_ERR(node)) {
1152 state = rb_entry(node, struct extent_state, rb_node);
1153 if (state->start != start) {
1157 state->private = private;
1159 write_unlock_irq(&tree->lock);
1163 int get_state_private(struct extent_map_tree *tree, u64 start, u64 *private)
1165 struct rb_node *node;
1166 struct extent_state *state;
1169 read_lock_irq(&tree->lock);
1171 * this search will find all the extents that end after
1174 node = tree_search(&tree->state, start);
1175 if (!node || IS_ERR(node)) {
1179 state = rb_entry(node, struct extent_state, rb_node);
1180 if (state->start != start) {
1184 *private = state->private;
1186 read_unlock_irq(&tree->lock);
1191 * searches a range in the state tree for a given mask.
1192 * If 'filled' == 1, this returns 1 only if ever extent in the tree
1193 * has the bits set. Otherwise, 1 is returned if any bit in the
1194 * range is found set.
1196 int test_range_bit(struct extent_map_tree *tree, u64 start, u64 end,
1197 int bits, int filled)
1199 struct extent_state *state = NULL;
1200 struct rb_node *node;
1203 read_lock_irq(&tree->lock);
1204 node = tree_search(&tree->state, start);
1205 while (node && start <= end) {
1206 state = rb_entry(node, struct extent_state, rb_node);
1207 if (state->start > end)
1210 if (filled && state->start > start) {
1214 if (state->state & bits) {
1218 } else if (filled) {
1222 start = state->end + 1;
1225 node = rb_next(node);
1227 read_unlock_irq(&tree->lock);
1230 EXPORT_SYMBOL(test_range_bit);
1233 * helper function to set a given page up to date if all the
1234 * extents in the tree for that page are up to date
1236 static int check_page_uptodate(struct extent_map_tree *tree,
1239 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
1240 u64 end = start + PAGE_CACHE_SIZE - 1;
1241 if (test_range_bit(tree, start, end, EXTENT_UPTODATE, 1))
1242 SetPageUptodate(page);
1247 * helper function to unlock a page if all the extents in the tree
1248 * for that page are unlocked
1250 static int check_page_locked(struct extent_map_tree *tree,
1253 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
1254 u64 end = start + PAGE_CACHE_SIZE - 1;
1255 if (!test_range_bit(tree, start, end, EXTENT_LOCKED, 0))
1261 * helper function to end page writeback if all the extents
1262 * in the tree for that page are done with writeback
1264 static int check_page_writeback(struct extent_map_tree *tree,
1267 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
1268 u64 end = start + PAGE_CACHE_SIZE - 1;
1269 if (!test_range_bit(tree, start, end, EXTENT_WRITEBACK, 0))
1270 end_page_writeback(page);
1274 /* lots and lots of room for performance fixes in the end_bio funcs */
1277 * after a writepage IO is done, we need to:
1278 * clear the uptodate bits on error
1279 * clear the writeback bits in the extent tree for this IO
1280 * end_page_writeback if the page has no more pending IO
1282 * Scheduling is not allowed, so the extent state tree is expected
1283 * to have one and only one object corresponding to this IO.
1285 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
1286 static void end_bio_extent_writepage(struct bio *bio, int err)
1288 static int end_bio_extent_writepage(struct bio *bio,
1289 unsigned int bytes_done, int err)
1292 const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
1293 struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
1294 struct extent_map_tree *tree = bio->bi_private;
1299 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
1305 struct page *page = bvec->bv_page;
1306 start = ((u64)page->index << PAGE_CACHE_SHIFT) +
1308 end = start + bvec->bv_len - 1;
1310 if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE)
1315 if (--bvec >= bio->bi_io_vec)
1316 prefetchw(&bvec->bv_page->flags);
1319 clear_extent_uptodate(tree, start, end, GFP_ATOMIC);
1320 ClearPageUptodate(page);
1323 clear_extent_writeback(tree, start, end, GFP_ATOMIC);
1326 end_page_writeback(page);
1328 check_page_writeback(tree, page);
1329 if (tree->ops && tree->ops->writepage_end_io_hook)
1330 tree->ops->writepage_end_io_hook(page, start, end);
1331 } while (bvec >= bio->bi_io_vec);
1334 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
1340 * after a readpage IO is done, we need to:
1341 * clear the uptodate bits on error
1342 * set the uptodate bits if things worked
1343 * set the page up to date if all extents in the tree are uptodate
1344 * clear the lock bit in the extent tree
1345 * unlock the page if there are no other extents locked for it
1347 * Scheduling is not allowed, so the extent state tree is expected
1348 * to have one and only one object corresponding to this IO.
1350 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
1351 static void end_bio_extent_readpage(struct bio *bio, int err)
1353 static int end_bio_extent_readpage(struct bio *bio,
1354 unsigned int bytes_done, int err)
1357 int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
1358 struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
1359 struct extent_map_tree *tree = bio->bi_private;
1365 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
1371 struct page *page = bvec->bv_page;
1372 start = ((u64)page->index << PAGE_CACHE_SHIFT) +
1374 end = start + bvec->bv_len - 1;
1376 if (bvec->bv_offset == 0 && bvec->bv_len == PAGE_CACHE_SIZE)
1381 if (--bvec >= bio->bi_io_vec)
1382 prefetchw(&bvec->bv_page->flags);
1384 if (uptodate && tree->ops && tree->ops->readpage_end_io_hook) {
1385 ret = tree->ops->readpage_end_io_hook(page, start, end);
1390 set_extent_uptodate(tree, start, end, GFP_ATOMIC);
1392 SetPageUptodate(page);
1394 check_page_uptodate(tree, page);
1396 ClearPageUptodate(page);
1400 unlock_extent(tree, start, end, GFP_ATOMIC);
1405 check_page_locked(tree, page);
1406 } while (bvec >= bio->bi_io_vec);
1409 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
1415 * IO done from prepare_write is pretty simple, we just unlock
1416 * the structs in the extent tree when done, and set the uptodate bits
1419 #if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
1420 static void end_bio_extent_preparewrite(struct bio *bio, int err)
1422 static int end_bio_extent_preparewrite(struct bio *bio,
1423 unsigned int bytes_done, int err)
1426 const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
1427 struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
1428 struct extent_map_tree *tree = bio->bi_private;
1432 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
1438 struct page *page = bvec->bv_page;
1439 start = ((u64)page->index << PAGE_CACHE_SHIFT) +
1441 end = start + bvec->bv_len - 1;
1443 if (--bvec >= bio->bi_io_vec)
1444 prefetchw(&bvec->bv_page->flags);
1447 set_extent_uptodate(tree, start, end, GFP_ATOMIC);
1449 ClearPageUptodate(page);
1453 unlock_extent(tree, start, end, GFP_ATOMIC);
1455 } while (bvec >= bio->bi_io_vec);
1458 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
1463 static int submit_extent_page(int rw, struct extent_map_tree *tree,
1464 struct page *page, sector_t sector,
1465 size_t size, unsigned long offset,
1466 struct block_device *bdev,
1467 bio_end_io_t end_io_func)
1472 bio = bio_alloc(GFP_NOIO, 1);
1474 bio->bi_sector = sector;
1475 bio->bi_bdev = bdev;
1476 bio->bi_io_vec[0].bv_page = page;
1477 bio->bi_io_vec[0].bv_len = size;
1478 bio->bi_io_vec[0].bv_offset = offset;
1482 bio->bi_size = size;
1484 bio->bi_end_io = end_io_func;
1485 bio->bi_private = tree;
1488 submit_bio(rw, bio);
1490 if (bio_flagged(bio, BIO_EOPNOTSUPP))
1497 void set_page_extent_mapped(struct page *page)
1499 if (!PagePrivate(page)) {
1500 SetPagePrivate(page);
1501 WARN_ON(!page->mapping->a_ops->invalidatepage);
1502 set_page_private(page, EXTENT_PAGE_PRIVATE);
1503 page_cache_get(page);
1508 * basic readpage implementation. Locked extent state structs are inserted
1509 * into the tree that are removed when the IO is done (by the end_io
1512 int extent_read_full_page(struct extent_map_tree *tree, struct page *page,
1513 get_extent_t *get_extent)
1515 struct inode *inode = page->mapping->host;
1516 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
1517 u64 page_end = start + PAGE_CACHE_SIZE - 1;
1521 u64 last_byte = i_size_read(inode);
1525 struct extent_map *em;
1526 struct block_device *bdev;
1529 size_t page_offset = 0;
1531 size_t blocksize = inode->i_sb->s_blocksize;
1533 set_page_extent_mapped(page);
1536 lock_extent(tree, start, end, GFP_NOFS);
1538 while (cur <= end) {
1539 if (cur >= last_byte) {
1540 iosize = PAGE_CACHE_SIZE - page_offset;
1541 zero_user_page(page, page_offset, iosize, KM_USER0);
1542 set_extent_uptodate(tree, cur, cur + iosize - 1,
1544 unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
1547 em = get_extent(inode, page, page_offset, cur, end, 0);
1548 if (IS_ERR(em) || !em) {
1550 unlock_extent(tree, cur, end, GFP_NOFS);
1554 extent_offset = cur - em->start;
1555 BUG_ON(em->end < cur);
1558 iosize = min(em->end - cur, end - cur) + 1;
1559 cur_end = min(em->end, end);
1560 iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1);
1561 sector = (em->block_start + extent_offset) >> 9;
1563 block_start = em->block_start;
1564 free_extent_map(em);
1567 /* we've found a hole, just zero and go on */
1568 if (block_start == EXTENT_MAP_HOLE) {
1569 zero_user_page(page, page_offset, iosize, KM_USER0);
1570 set_extent_uptodate(tree, cur, cur + iosize - 1,
1572 unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
1574 page_offset += iosize;
1577 /* the get_extent function already copied into the page */
1578 if (test_range_bit(tree, cur, cur_end, EXTENT_UPTODATE, 1)) {
1579 unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
1581 page_offset += iosize;
1586 if (tree->ops && tree->ops->readpage_io_hook) {
1587 ret = tree->ops->readpage_io_hook(page, cur,
1591 ret = submit_extent_page(READ, tree, page,
1592 sector, iosize, page_offset,
1593 bdev, end_bio_extent_readpage);
1598 page_offset += iosize;
1602 if (!PageError(page))
1603 SetPageUptodate(page);
1608 EXPORT_SYMBOL(extent_read_full_page);
1611 * the writepage semantics are similar to regular writepage. extent
1612 * records are inserted to lock ranges in the tree, and as dirty areas
1613 * are found, they are marked writeback. Then the lock bits are removed
1614 * and the end_io handler clears the writeback ranges
1616 int extent_write_full_page(struct extent_map_tree *tree, struct page *page,
1617 get_extent_t *get_extent,
1618 struct writeback_control *wbc)
1620 struct inode *inode = page->mapping->host;
1621 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
1622 u64 page_end = start + PAGE_CACHE_SIZE - 1;
1626 u64 last_byte = i_size_read(inode);
1630 struct extent_map *em;
1631 struct block_device *bdev;
1634 size_t page_offset = 0;
1636 loff_t i_size = i_size_read(inode);
1637 unsigned long end_index = i_size >> PAGE_CACHE_SHIFT;
1641 WARN_ON(!PageLocked(page));
1642 if (page->index > end_index) {
1643 clear_extent_dirty(tree, start, page_end, GFP_NOFS);
1648 if (page->index == end_index) {
1649 size_t offset = i_size & (PAGE_CACHE_SIZE - 1);
1650 zero_user_page(page, offset,
1651 PAGE_CACHE_SIZE - offset, KM_USER0);
1654 set_page_extent_mapped(page);
1656 lock_extent(tree, start, page_end, GFP_NOFS);
1657 nr_delalloc = find_lock_delalloc_range(tree, start, page_end + 1,
1661 tree->ops->fill_delalloc(inode, start, delalloc_end);
1662 if (delalloc_end >= page_end + 1) {
1663 clear_extent_bit(tree, page_end + 1, delalloc_end,
1664 EXTENT_LOCKED | EXTENT_DELALLOC,
1667 clear_extent_bit(tree, start, page_end, EXTENT_DELALLOC,
1669 if (test_range_bit(tree, start, page_end, EXTENT_DELALLOC, 0)) {
1670 printk("found delalloc bits after clear extent_bit\n");
1672 } else if (test_range_bit(tree, start, page_end, EXTENT_DELALLOC, 0)) {
1673 printk("found delalloc bits after find_delalloc_range returns 0\n");
1677 if (test_range_bit(tree, start, page_end, EXTENT_DELALLOC, 0)) {
1678 printk("found delalloc bits after lock_extent\n");
1681 if (last_byte <= start) {
1682 clear_extent_dirty(tree, start, page_end, GFP_NOFS);
1686 set_extent_uptodate(tree, start, page_end, GFP_NOFS);
1687 blocksize = inode->i_sb->s_blocksize;
1689 while (cur <= end) {
1690 if (cur >= last_byte) {
1691 clear_extent_dirty(tree, cur, page_end, GFP_NOFS);
1694 em = get_extent(inode, page, page_offset, cur, end, 1);
1695 if (IS_ERR(em) || !em) {
1700 extent_offset = cur - em->start;
1701 BUG_ON(em->end < cur);
1703 iosize = min(em->end - cur, end - cur) + 1;
1704 iosize = (iosize + blocksize - 1) & ~((u64)blocksize - 1);
1705 sector = (em->block_start + extent_offset) >> 9;
1707 block_start = em->block_start;
1708 free_extent_map(em);
1711 if (block_start == EXTENT_MAP_HOLE ||
1712 block_start == EXTENT_MAP_INLINE) {
1713 clear_extent_dirty(tree, cur,
1714 cur + iosize - 1, GFP_NOFS);
1716 page_offset += iosize;
1720 /* leave this out until we have a page_mkwrite call */
1721 if (0 && !test_range_bit(tree, cur, cur + iosize - 1,
1724 page_offset += iosize;
1727 clear_extent_dirty(tree, cur, cur + iosize - 1, GFP_NOFS);
1728 if (tree->ops && tree->ops->writepage_io_hook) {
1729 ret = tree->ops->writepage_io_hook(page, cur,
1737 set_range_writeback(tree, cur, cur + iosize - 1);
1738 ret = submit_extent_page(WRITE, tree, page, sector,
1739 iosize, page_offset, bdev,
1740 end_bio_extent_writepage);
1745 page_offset += iosize;
1749 unlock_extent(tree, start, page_end, GFP_NOFS);
1753 EXPORT_SYMBOL(extent_write_full_page);
1756 * basic invalidatepage code, this waits on any locked or writeback
1757 * ranges corresponding to the page, and then deletes any extent state
1758 * records from the tree
1760 int extent_invalidatepage(struct extent_map_tree *tree,
1761 struct page *page, unsigned long offset)
1763 u64 start = ((u64)page->index << PAGE_CACHE_SHIFT);
1764 u64 end = start + PAGE_CACHE_SIZE - 1;
1765 size_t blocksize = page->mapping->host->i_sb->s_blocksize;
1767 start += (offset + blocksize -1) & ~(blocksize - 1);
1771 lock_extent(tree, start, end, GFP_NOFS);
1772 wait_on_extent_writeback(tree, start, end);
1773 clear_extent_bit(tree, start, end,
1774 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC,
1778 EXPORT_SYMBOL(extent_invalidatepage);
1781 * simple commit_write call, set_range_dirty is used to mark both
1782 * the pages and the extent records as dirty
1784 int extent_commit_write(struct extent_map_tree *tree,
1785 struct inode *inode, struct page *page,
1786 unsigned from, unsigned to)
1788 loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to;
1790 set_page_extent_mapped(page);
1791 set_page_dirty(page);
1793 if (pos > inode->i_size) {
1794 i_size_write(inode, pos);
1795 mark_inode_dirty(inode);
1799 EXPORT_SYMBOL(extent_commit_write);
1801 int extent_prepare_write(struct extent_map_tree *tree,
1802 struct inode *inode, struct page *page,
1803 unsigned from, unsigned to, get_extent_t *get_extent)
1805 u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
1806 u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
1808 u64 orig_block_start;
1811 struct extent_map *em;
1812 unsigned blocksize = 1 << inode->i_blkbits;
1813 size_t page_offset = 0;
1814 size_t block_off_start;
1815 size_t block_off_end;
1821 set_page_extent_mapped(page);
1823 block_start = (page_start + from) & ~((u64)blocksize - 1);
1824 block_end = (page_start + to - 1) | (blocksize - 1);
1825 orig_block_start = block_start;
1827 lock_extent(tree, page_start, page_end, GFP_NOFS);
1828 while(block_start <= block_end) {
1829 em = get_extent(inode, page, page_offset, block_start,
1831 if (IS_ERR(em) || !em) {
1834 cur_end = min(block_end, em->end);
1835 block_off_start = block_start & (PAGE_CACHE_SIZE - 1);
1836 block_off_end = block_off_start + blocksize;
1837 isnew = clear_extent_new(tree, block_start, cur_end, GFP_NOFS);
1839 if (!PageUptodate(page) && isnew &&
1840 (block_off_end > to || block_off_start < from)) {
1843 kaddr = kmap_atomic(page, KM_USER0);
1844 if (block_off_end > to)
1845 memset(kaddr + to, 0, block_off_end - to);
1846 if (block_off_start < from)
1847 memset(kaddr + block_off_start, 0,
1848 from - block_off_start);
1849 flush_dcache_page(page);
1850 kunmap_atomic(kaddr, KM_USER0);
1852 if (!isnew && !PageUptodate(page) &&
1853 (block_off_end > to || block_off_start < from) &&
1854 !test_range_bit(tree, block_start, cur_end,
1855 EXTENT_UPTODATE, 1)) {
1857 u64 extent_offset = block_start - em->start;
1859 sector = (em->block_start + extent_offset) >> 9;
1860 iosize = (cur_end - block_start + blocksize - 1) &
1861 ~((u64)blocksize - 1);
1863 * we've already got the extent locked, but we
1864 * need to split the state such that our end_bio
1865 * handler can clear the lock.
1867 set_extent_bit(tree, block_start,
1868 block_start + iosize - 1,
1869 EXTENT_LOCKED, 0, NULL, GFP_NOFS);
1870 ret = submit_extent_page(READ, tree, page,
1871 sector, iosize, page_offset, em->bdev,
1872 end_bio_extent_preparewrite);
1874 block_start = block_start + iosize;
1876 set_extent_uptodate(tree, block_start, cur_end,
1878 unlock_extent(tree, block_start, cur_end, GFP_NOFS);
1879 block_start = cur_end + 1;
1881 page_offset = block_start & (PAGE_CACHE_SIZE - 1);
1882 free_extent_map(em);
1885 wait_extent_bit(tree, orig_block_start,
1886 block_end, EXTENT_LOCKED);
1888 check_page_uptodate(tree, page);
1890 /* FIXME, zero out newly allocated blocks on error */
1893 EXPORT_SYMBOL(extent_prepare_write);
1896 * a helper for releasepage. As long as there are no locked extents
1897 * in the range corresponding to the page, both state records and extent
1898 * map records are removed
1900 int try_release_extent_mapping(struct extent_map_tree *tree, struct page *page)
1902 struct extent_map *em;
1903 u64 start = (u64)page->index << PAGE_CACHE_SHIFT;
1904 u64 end = start + PAGE_CACHE_SIZE - 1;
1905 u64 orig_start = start;
1908 while (start <= end) {
1909 em = lookup_extent_mapping(tree, start, end);
1910 if (!em || IS_ERR(em))
1912 if (!test_range_bit(tree, em->start, em->end,
1913 EXTENT_LOCKED, 0)) {
1914 remove_extent_mapping(tree, em);
1915 /* once for the rb tree */
1916 free_extent_map(em);
1918 start = em->end + 1;
1920 free_extent_map(em);
1922 if (test_range_bit(tree, orig_start, end, EXTENT_LOCKED, 0))
1925 clear_extent_bit(tree, orig_start, end, EXTENT_UPTODATE,
1929 EXPORT_SYMBOL(try_release_extent_mapping);
1931 sector_t extent_bmap(struct address_space *mapping, sector_t iblock,
1932 get_extent_t *get_extent)
1934 struct inode *inode = mapping->host;
1935 u64 start = iblock << inode->i_blkbits;
1936 u64 end = start + (1 << inode->i_blkbits) - 1;
1937 sector_t sector = 0;
1938 struct extent_map *em;
1940 em = get_extent(inode, NULL, 0, start, end, 0);
1941 if (!em || IS_ERR(em))
1944 if (em->block_start == EXTENT_MAP_INLINE ||
1945 em->block_start == EXTENT_MAP_HOLE)
1948 sector = (em->block_start + start - em->start) >> inode->i_blkbits;
1950 free_extent_map(em);
1954 static int add_lru(struct extent_map_tree *tree, struct extent_buffer *eb)
1956 if (list_empty(&eb->lru)) {
1957 extent_buffer_get(eb);
1958 list_add(&eb->lru, &tree->buffer_lru);
1960 if (tree->lru_size >= BUFFER_LRU_MAX) {
1961 struct extent_buffer *rm;
1962 rm = list_entry(tree->buffer_lru.prev,
1963 struct extent_buffer, lru);
1966 free_extent_buffer(rm);
1969 list_move(&eb->lru, &tree->buffer_lru);
1972 static struct extent_buffer *find_lru(struct extent_map_tree *tree,
1973 u64 start, unsigned long len)
1975 struct list_head *lru = &tree->buffer_lru;
1976 struct list_head *cur = lru->next;
1977 struct extent_buffer *eb;
1979 if (list_empty(lru))
1983 eb = list_entry(cur, struct extent_buffer, lru);
1984 if (eb->start == start && eb->len == len) {
1985 extent_buffer_get(eb);
1989 } while (cur != lru);
1993 static inline unsigned long num_extent_pages(u64 start, u64 len)
1995 return ((start + len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT) -
1996 (start >> PAGE_CACHE_SHIFT);
1999 static inline struct page *extent_buffer_page(struct extent_buffer *eb,
2003 struct address_space *mapping;
2006 return eb->first_page;
2007 i += eb->start >> PAGE_CACHE_SHIFT;
2008 mapping = eb->first_page->mapping;
2009 read_lock_irq(&mapping->tree_lock);
2010 p = radix_tree_lookup(&mapping->page_tree, i);
2011 read_unlock_irq(&mapping->tree_lock);
2015 static struct extent_buffer *__alloc_extent_buffer(struct extent_map_tree *tree,
2020 struct extent_buffer *eb = NULL;
2022 spin_lock(&tree->lru_lock);
2023 eb = find_lru(tree, start, len);
2027 spin_unlock(&tree->lru_lock);
2030 memset(eb, 0, sizeof(*eb));
2032 eb = kmem_cache_zalloc(extent_buffer_cache, mask);
2034 INIT_LIST_HEAD(&eb->lru);
2037 atomic_set(&eb->refs, 1);
2039 spin_lock(&tree->lru_lock);
2042 spin_unlock(&tree->lru_lock);
2046 static void __free_extent_buffer(struct extent_buffer *eb)
2048 kmem_cache_free(extent_buffer_cache, eb);
2051 struct extent_buffer *alloc_extent_buffer(struct extent_map_tree *tree,
2052 u64 start, unsigned long len,
2056 unsigned long num_pages = num_extent_pages(start, len);
2058 unsigned long index = start >> PAGE_CACHE_SHIFT;
2059 struct extent_buffer *eb;
2061 struct address_space *mapping = tree->mapping;
2064 eb = __alloc_extent_buffer(tree, start, len, mask);
2065 if (!eb || IS_ERR(eb))
2068 if (eb->flags & EXTENT_BUFFER_FILLED)
2072 eb->first_page = page0;
2075 page_cache_get(page0);
2076 mark_page_accessed(page0);
2077 set_page_extent_mapped(page0);
2078 set_page_private(page0, EXTENT_PAGE_PRIVATE_FIRST_PAGE |
2083 for (; i < num_pages; i++, index++) {
2084 p = find_or_create_page(mapping, index, mask | __GFP_HIGHMEM);
2087 /* make sure the free only frees the pages we've
2088 * grabbed a reference on
2090 eb->len = i << PAGE_CACHE_SHIFT;
2091 eb->start &= ~((u64)PAGE_CACHE_SIZE - 1);
2094 set_page_extent_mapped(p);
2095 mark_page_accessed(p);
2098 set_page_private(p, EXTENT_PAGE_PRIVATE_FIRST_PAGE |
2101 set_page_private(p, EXTENT_PAGE_PRIVATE);
2103 if (!PageUptodate(p))
2108 eb->flags |= EXTENT_UPTODATE;
2109 eb->flags |= EXTENT_BUFFER_FILLED;
2112 free_extent_buffer(eb);
2115 EXPORT_SYMBOL(alloc_extent_buffer);
2117 struct extent_buffer *find_extent_buffer(struct extent_map_tree *tree,
2118 u64 start, unsigned long len,
2121 unsigned long num_pages = num_extent_pages(start, len);
2122 unsigned long i; unsigned long index = start >> PAGE_CACHE_SHIFT;
2123 struct extent_buffer *eb;
2125 struct address_space *mapping = tree->mapping;
2128 eb = __alloc_extent_buffer(tree, start, len, mask);
2129 if (!eb || IS_ERR(eb))
2132 if (eb->flags & EXTENT_BUFFER_FILLED)
2135 for (i = 0; i < num_pages; i++, index++) {
2136 p = find_lock_page(mapping, index);
2138 /* make sure the free only frees the pages we've
2139 * grabbed a reference on
2141 eb->len = i << PAGE_CACHE_SHIFT;
2142 eb->start &= ~((u64)PAGE_CACHE_SIZE - 1);
2145 set_page_extent_mapped(p);
2146 mark_page_accessed(p);
2150 set_page_private(p, EXTENT_PAGE_PRIVATE_FIRST_PAGE |
2153 set_page_private(p, EXTENT_PAGE_PRIVATE);
2156 if (!PageUptodate(p))
2161 eb->flags |= EXTENT_UPTODATE;
2162 eb->flags |= EXTENT_BUFFER_FILLED;
2165 free_extent_buffer(eb);
2168 EXPORT_SYMBOL(find_extent_buffer);
2170 void free_extent_buffer(struct extent_buffer *eb)
2173 unsigned long num_pages;
2178 if (!atomic_dec_and_test(&eb->refs))
2181 num_pages = num_extent_pages(eb->start, eb->len);
2183 for (i = 0; i < num_pages; i++) {
2184 page_cache_release(extent_buffer_page(eb, i));
2186 __free_extent_buffer(eb);
2188 EXPORT_SYMBOL(free_extent_buffer);
2190 int clear_extent_buffer_dirty(struct extent_map_tree *tree,
2191 struct extent_buffer *eb)
2195 unsigned long num_pages;
2198 u64 start = eb->start;
2199 u64 end = start + eb->len - 1;
2201 set = clear_extent_dirty(tree, start, end, GFP_NOFS);
2202 num_pages = num_extent_pages(eb->start, eb->len);
2204 for (i = 0; i < num_pages; i++) {
2205 page = extent_buffer_page(eb, i);
2208 * if we're on the last page or the first page and the
2209 * block isn't aligned on a page boundary, do extra checks
2210 * to make sure we don't clean page that is partially dirty
2212 if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) ||
2213 ((i == num_pages - 1) &&
2214 ((eb->start + eb->len) & (PAGE_CACHE_SIZE - 1)))) {
2215 start = (u64)page->index << PAGE_CACHE_SHIFT;
2216 end = start + PAGE_CACHE_SIZE - 1;
2217 if (test_range_bit(tree, start, end,
2223 clear_page_dirty_for_io(page);
2228 EXPORT_SYMBOL(clear_extent_buffer_dirty);
2230 int wait_on_extent_buffer_writeback(struct extent_map_tree *tree,
2231 struct extent_buffer *eb)
2233 return wait_on_extent_writeback(tree, eb->start,
2234 eb->start + eb->len - 1);
2236 EXPORT_SYMBOL(wait_on_extent_buffer_writeback);
2238 int set_extent_buffer_dirty(struct extent_map_tree *tree,
2239 struct extent_buffer *eb)
2242 unsigned long num_pages;
2244 num_pages = num_extent_pages(eb->start, eb->len);
2245 for (i = 0; i < num_pages; i++) {
2246 struct page *page = extent_buffer_page(eb, i);
2247 /* writepage may need to do something special for the
2248 * first page, we have to make sure page->private is
2249 * properly set. releasepage may drop page->private
2250 * on us if the page isn't already dirty.
2254 set_page_private(page,
2255 EXTENT_PAGE_PRIVATE_FIRST_PAGE |
2258 __set_page_dirty_nobuffers(extent_buffer_page(eb, i));
2262 return set_extent_dirty(tree, eb->start,
2263 eb->start + eb->len - 1, GFP_NOFS);
2265 EXPORT_SYMBOL(set_extent_buffer_dirty);
2267 int set_extent_buffer_uptodate(struct extent_map_tree *tree,
2268 struct extent_buffer *eb)
2272 unsigned long num_pages;
2274 num_pages = num_extent_pages(eb->start, eb->len);
2276 set_extent_uptodate(tree, eb->start, eb->start + eb->len - 1,
2278 for (i = 0; i < num_pages; i++) {
2279 page = extent_buffer_page(eb, i);
2280 if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) ||
2281 ((i == num_pages - 1) &&
2282 ((eb->start + eb->len) & (PAGE_CACHE_SIZE - 1)))) {
2283 check_page_uptodate(tree, page);
2286 SetPageUptodate(page);
2290 EXPORT_SYMBOL(set_extent_buffer_uptodate);
2292 int extent_buffer_uptodate(struct extent_map_tree *tree,
2293 struct extent_buffer *eb)
2295 if (eb->flags & EXTENT_UPTODATE)
2297 return test_range_bit(tree, eb->start, eb->start + eb->len - 1,
2298 EXTENT_UPTODATE, 1);
2300 EXPORT_SYMBOL(extent_buffer_uptodate);
2302 int read_extent_buffer_pages(struct extent_map_tree *tree,
2303 struct extent_buffer *eb,
2308 unsigned long start_i;
2312 unsigned long num_pages;
2314 if (eb->flags & EXTENT_UPTODATE)
2317 if (0 && test_range_bit(tree, eb->start, eb->start + eb->len - 1,
2318 EXTENT_UPTODATE, 1)) {
2322 WARN_ON(start < eb->start);
2323 start_i = (start >> PAGE_CACHE_SHIFT) -
2324 (eb->start >> PAGE_CACHE_SHIFT);
2329 num_pages = num_extent_pages(eb->start, eb->len);
2330 for (i = start_i; i < num_pages; i++) {
2331 page = extent_buffer_page(eb, i);
2332 if (PageUptodate(page)) {
2336 if (TestSetPageLocked(page)) {
2342 if (!PageUptodate(page)) {
2343 err = page->mapping->a_ops->readpage(NULL, page);
2356 for (i = start_i; i < num_pages; i++) {
2357 page = extent_buffer_page(eb, i);
2358 wait_on_page_locked(page);
2359 if (!PageUptodate(page)) {
2364 eb->flags |= EXTENT_UPTODATE;
2367 EXPORT_SYMBOL(read_extent_buffer_pages);
2369 void read_extent_buffer(struct extent_buffer *eb, void *dstv,
2370 unsigned long start,
2377 char *dst = (char *)dstv;
2378 size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
2379 unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
2380 unsigned long num_pages = num_extent_pages(eb->start, eb->len);
2382 WARN_ON(start > eb->len);
2383 WARN_ON(start + len > eb->start + eb->len);
2385 offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1);
2388 page = extent_buffer_page(eb, i);
2389 if (!PageUptodate(page)) {
2390 printk("page %lu not up to date i %lu, total %lu, len %lu\n", page->index, i, num_pages, eb->len);
2393 WARN_ON(!PageUptodate(page));
2395 cur = min(len, (PAGE_CACHE_SIZE - offset));
2396 kaddr = kmap_atomic(page, KM_USER1);
2397 memcpy(dst, kaddr + offset, cur);
2398 kunmap_atomic(kaddr, KM_USER1);
2406 EXPORT_SYMBOL(read_extent_buffer);
2408 int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start,
2409 unsigned long min_len, char **token, char **map,
2410 unsigned long *map_start,
2411 unsigned long *map_len, int km)
2413 size_t offset = start & (PAGE_CACHE_SIZE - 1);
2416 size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
2417 unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
2418 unsigned long end_i = (start_offset + start + min_len - 1) >>
2425 offset = start_offset;
2429 *map_start = (i << PAGE_CACHE_SHIFT) - start_offset;
2431 if (start + min_len > eb->len) {
2432 printk("bad mapping eb start %Lu len %lu, wanted %lu %lu\n", eb->start, eb->len, start, min_len);
2436 p = extent_buffer_page(eb, i);
2437 WARN_ON(!PageUptodate(p));
2438 kaddr = kmap_atomic(p, km);
2440 *map = kaddr + offset;
2441 *map_len = PAGE_CACHE_SIZE - offset;
2444 EXPORT_SYMBOL(map_private_extent_buffer);
2446 int map_extent_buffer(struct extent_buffer *eb, unsigned long start,
2447 unsigned long min_len,
2448 char **token, char **map,
2449 unsigned long *map_start,
2450 unsigned long *map_len, int km)
2454 if (eb->map_token) {
2455 unmap_extent_buffer(eb, eb->map_token, km);
2456 eb->map_token = NULL;
2459 err = map_private_extent_buffer(eb, start, min_len, token, map,
2460 map_start, map_len, km);
2462 eb->map_token = *token;
2464 eb->map_start = *map_start;
2465 eb->map_len = *map_len;
2469 EXPORT_SYMBOL(map_extent_buffer);
2471 void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km)
2473 kunmap_atomic(token, km);
2475 EXPORT_SYMBOL(unmap_extent_buffer);
2477 int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv,
2478 unsigned long start,
2485 char *ptr = (char *)ptrv;
2486 size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
2487 unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
2490 WARN_ON(start > eb->len);
2491 WARN_ON(start + len > eb->start + eb->len);
2493 offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1);
2496 page = extent_buffer_page(eb, i);
2497 WARN_ON(!PageUptodate(page));
2499 cur = min(len, (PAGE_CACHE_SIZE - offset));
2501 kaddr = kmap_atomic(page, KM_USER0);
2502 ret = memcmp(ptr, kaddr + offset, cur);
2503 kunmap_atomic(kaddr, KM_USER0);
2514 EXPORT_SYMBOL(memcmp_extent_buffer);
2516 void write_extent_buffer(struct extent_buffer *eb, const void *srcv,
2517 unsigned long start, unsigned long len)
2523 char *src = (char *)srcv;
2524 size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
2525 unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
2527 WARN_ON(start > eb->len);
2528 WARN_ON(start + len > eb->start + eb->len);
2530 offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1);
2533 page = extent_buffer_page(eb, i);
2534 WARN_ON(!PageUptodate(page));
2536 cur = min(len, PAGE_CACHE_SIZE - offset);
2537 kaddr = kmap_atomic(page, KM_USER1);
2538 memcpy(kaddr + offset, src, cur);
2539 kunmap_atomic(kaddr, KM_USER1);
2547 EXPORT_SYMBOL(write_extent_buffer);
2549 void memset_extent_buffer(struct extent_buffer *eb, char c,
2550 unsigned long start, unsigned long len)
2556 size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
2557 unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
2559 WARN_ON(start > eb->len);
2560 WARN_ON(start + len > eb->start + eb->len);
2562 offset = (start_offset + start) & ((unsigned long)PAGE_CACHE_SIZE - 1);
2565 page = extent_buffer_page(eb, i);
2566 WARN_ON(!PageUptodate(page));
2568 cur = min(len, PAGE_CACHE_SIZE - offset);
2569 kaddr = kmap_atomic(page, KM_USER0);
2570 memset(kaddr + offset, c, cur);
2571 kunmap_atomic(kaddr, KM_USER0);
2578 EXPORT_SYMBOL(memset_extent_buffer);
2580 void copy_extent_buffer(struct extent_buffer *dst, struct extent_buffer *src,
2581 unsigned long dst_offset, unsigned long src_offset,
2584 u64 dst_len = dst->len;
2589 size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1);
2590 unsigned long i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT;
2592 WARN_ON(src->len != dst_len);
2594 offset = (start_offset + dst_offset) &
2595 ((unsigned long)PAGE_CACHE_SIZE - 1);
2598 page = extent_buffer_page(dst, i);
2599 WARN_ON(!PageUptodate(page));
2601 cur = min(len, (unsigned long)(PAGE_CACHE_SIZE - offset));
2603 kaddr = kmap_atomic(page, KM_USER0);
2604 read_extent_buffer(src, kaddr + offset, src_offset, cur);
2605 kunmap_atomic(kaddr, KM_USER0);
2613 EXPORT_SYMBOL(copy_extent_buffer);
2615 static void move_pages(struct page *dst_page, struct page *src_page,
2616 unsigned long dst_off, unsigned long src_off,
2619 char *dst_kaddr = kmap_atomic(dst_page, KM_USER0);
2620 if (dst_page == src_page) {
2621 memmove(dst_kaddr + dst_off, dst_kaddr + src_off, len);
2623 char *src_kaddr = kmap_atomic(src_page, KM_USER1);
2624 char *p = dst_kaddr + dst_off + len;
2625 char *s = src_kaddr + src_off + len;
2630 kunmap_atomic(src_kaddr, KM_USER1);
2632 kunmap_atomic(dst_kaddr, KM_USER0);
2635 static void copy_pages(struct page *dst_page, struct page *src_page,
2636 unsigned long dst_off, unsigned long src_off,
2639 char *dst_kaddr = kmap_atomic(dst_page, KM_USER0);
2642 if (dst_page != src_page)
2643 src_kaddr = kmap_atomic(src_page, KM_USER1);
2645 src_kaddr = dst_kaddr;
2647 memcpy(dst_kaddr + dst_off, src_kaddr + src_off, len);
2648 kunmap_atomic(dst_kaddr, KM_USER0);
2649 if (dst_page != src_page)
2650 kunmap_atomic(src_kaddr, KM_USER1);
2653 void memcpy_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
2654 unsigned long src_offset, unsigned long len)
2657 size_t dst_off_in_page;
2658 size_t src_off_in_page;
2659 size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1);
2660 unsigned long dst_i;
2661 unsigned long src_i;
2663 if (src_offset + len > dst->len) {
2664 printk("memmove bogus src_offset %lu move len %lu len %lu\n",
2665 src_offset, len, dst->len);
2668 if (dst_offset + len > dst->len) {
2669 printk("memmove bogus dst_offset %lu move len %lu len %lu\n",
2670 dst_offset, len, dst->len);
2675 dst_off_in_page = (start_offset + dst_offset) &
2676 ((unsigned long)PAGE_CACHE_SIZE - 1);
2677 src_off_in_page = (start_offset + src_offset) &
2678 ((unsigned long)PAGE_CACHE_SIZE - 1);
2680 dst_i = (start_offset + dst_offset) >> PAGE_CACHE_SHIFT;
2681 src_i = (start_offset + src_offset) >> PAGE_CACHE_SHIFT;
2683 cur = min(len, (unsigned long)(PAGE_CACHE_SIZE -
2685 cur = min_t(unsigned long, cur,
2686 (unsigned long)(PAGE_CACHE_SIZE - dst_off_in_page));
2688 copy_pages(extent_buffer_page(dst, dst_i),
2689 extent_buffer_page(dst, src_i),
2690 dst_off_in_page, src_off_in_page, cur);
2697 EXPORT_SYMBOL(memcpy_extent_buffer);
2699 void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
2700 unsigned long src_offset, unsigned long len)
2703 size_t dst_off_in_page;
2704 size_t src_off_in_page;
2705 unsigned long dst_end = dst_offset + len - 1;
2706 unsigned long src_end = src_offset + len - 1;
2707 size_t start_offset = dst->start & ((u64)PAGE_CACHE_SIZE - 1);
2708 unsigned long dst_i;
2709 unsigned long src_i;
2711 if (src_offset + len > dst->len) {
2712 printk("memmove bogus src_offset %lu move len %lu len %lu\n",
2713 src_offset, len, dst->len);
2716 if (dst_offset + len > dst->len) {
2717 printk("memmove bogus dst_offset %lu move len %lu len %lu\n",
2718 dst_offset, len, dst->len);
2721 if (dst_offset < src_offset) {
2722 memcpy_extent_buffer(dst, dst_offset, src_offset, len);
2726 dst_i = (start_offset + dst_end) >> PAGE_CACHE_SHIFT;
2727 src_i = (start_offset + src_end) >> PAGE_CACHE_SHIFT;
2729 dst_off_in_page = (start_offset + dst_end) &
2730 ((unsigned long)PAGE_CACHE_SIZE - 1);
2731 src_off_in_page = (start_offset + src_end) &
2732 ((unsigned long)PAGE_CACHE_SIZE - 1);
2734 cur = min_t(unsigned long, len, src_off_in_page + 1);
2735 cur = min(cur, dst_off_in_page + 1);
2736 move_pages(extent_buffer_page(dst, dst_i),
2737 extent_buffer_page(dst, src_i),
2738 dst_off_in_page - cur + 1,
2739 src_off_in_page - cur + 1, cur);
2746 EXPORT_SYMBOL(memmove_extent_buffer);