2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/sched.h>
22 #include "print-tree.h"
23 #include "transaction.h"
25 #define BLOCK_GROUP_DATA EXTENT_WRITEBACK
26 #define BLOCK_GROUP_METADATA EXTENT_UPTODATE
27 #define BLOCK_GROUP_DIRTY EXTENT_DIRTY
29 static int finish_current_insert(struct btrfs_trans_handle *trans, struct
30 btrfs_root *extent_root);
31 static int del_pending_extents(struct btrfs_trans_handle *trans, struct
32 btrfs_root *extent_root);
34 static int cache_block_group(struct btrfs_root *root,
35 struct btrfs_block_group_cache *block_group)
37 struct btrfs_path *path;
40 struct extent_buffer *leaf;
41 struct extent_map_tree *free_space_cache;
48 root = root->fs_info->extent_root;
49 free_space_cache = &root->fs_info->free_space_cache;
51 if (block_group->cached)
54 path = btrfs_alloc_path();
59 first_free = block_group->key.objectid;
60 key.objectid = block_group->key.objectid;
63 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
64 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
69 if (ret && path->slots[0] > 0)
73 leaf = path->nodes[0];
74 slot = path->slots[0];
75 if (slot >= btrfs_header_nritems(leaf)) {
76 ret = btrfs_next_leaf(root, path);
86 btrfs_item_key_to_cpu(leaf, &key, slot);
87 if (key.objectid < block_group->key.objectid) {
88 if (key.objectid + key.offset > first_free)
89 first_free = key.objectid + key.offset;
93 if (key.objectid >= block_group->key.objectid +
94 block_group->key.offset) {
98 if (btrfs_key_type(&key) == BTRFS_EXTENT_ITEM_KEY) {
103 if (key.objectid > last) {
104 hole_size = key.objectid - last;
105 set_extent_dirty(free_space_cache, last,
106 last + hole_size - 1,
109 last = key.objectid + key.offset;
117 if (block_group->key.objectid +
118 block_group->key.offset > last) {
119 hole_size = block_group->key.objectid +
120 block_group->key.offset - last;
121 set_extent_dirty(free_space_cache, last,
122 last + hole_size - 1, GFP_NOFS);
124 block_group->cached = 1;
126 btrfs_free_path(path);
130 struct btrfs_block_group_cache *btrfs_lookup_block_group(struct
134 struct extent_map_tree *block_group_cache;
135 struct btrfs_block_group_cache *block_group = NULL;
141 block_group_cache = &info->block_group_cache;
142 ret = find_first_extent_bit(block_group_cache,
143 blocknr, &start, &end,
144 BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA);
148 ret = get_state_private(block_group_cache, start, &ptr);
152 block_group = (struct btrfs_block_group_cache *)ptr;
155 if (block_group->key.objectid <= blocknr && blocknr <=
156 block_group->key.objectid + block_group->key.offset)
162 static u64 find_search_start(struct btrfs_root *root,
163 struct btrfs_block_group_cache **cache_ret,
164 u64 search_start, int num, int data)
167 struct btrfs_block_group_cache *cache = *cache_ret;
168 u64 last = max(search_start, cache->key.objectid);
173 ret = cache_block_group(root, cache);
177 ret = find_first_extent_bit(&root->fs_info->free_space_cache,
178 last, &start, &end, EXTENT_DIRTY);
182 start = max(last, start);
184 if (end + 1 - start < num)
186 if (start + num >= cache->key.objectid + cache->key.offset)
194 cache = btrfs_lookup_block_group(root->fs_info,
195 last + cache->key.offset - 1);
199 cache = btrfs_find_block_group(root, cache,
200 last + cache->key.offset - 1, data, 0);
202 last = min(cache->key.objectid, last);
206 static u64 div_factor(u64 num, int factor)
213 struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root,
214 struct btrfs_block_group_cache
215 *hint, u64 search_start,
218 struct btrfs_block_group_cache *cache;
219 struct extent_map_tree *block_group_cache;
220 struct btrfs_block_group_cache *found_group = NULL;
221 struct btrfs_fs_info *info = root->fs_info;
235 block_group_cache = &info->block_group_cache;
241 bit = BLOCK_GROUP_DATA;
243 bit = BLOCK_GROUP_METADATA;
246 struct btrfs_block_group_cache *shint;
247 shint = btrfs_lookup_block_group(info, search_start);
248 if (shint && shint->data == data) {
249 used = btrfs_block_group_used(&shint->item);
250 if (used < div_factor(shint->key.offset, factor)) {
255 if (hint && hint->data == data) {
256 used = btrfs_block_group_used(&hint->item);
257 if (used < div_factor(hint->key.offset, factor)) {
260 last = hint->key.offset * 3;
261 if (hint->key.objectid >= last)
262 last = max(search_start + hint->key.offset - 1,
263 hint->key.objectid - last);
265 last = hint->key.objectid + hint->key.offset;
269 hint_last = max(hint->key.objectid, search_start);
271 hint_last = search_start;
277 ret = find_first_extent_bit(block_group_cache, last,
282 ret = get_state_private(block_group_cache, start, &ptr);
286 cache = (struct btrfs_block_group_cache *)ptr;
287 last = cache->key.objectid + cache->key.offset;
288 used = btrfs_block_group_used(&cache->item);
291 free_check = cache->key.offset;
293 free_check = div_factor(cache->key.offset, factor);
295 if (used < free_check) {
308 bit = BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA;
316 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
317 struct btrfs_root *root,
318 u64 blocknr, u64 num_blocks)
320 struct btrfs_path *path;
322 struct btrfs_key key;
323 struct extent_buffer *l;
324 struct btrfs_extent_item *item;
327 path = btrfs_alloc_path();
331 key.objectid = blocknr;
332 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
333 key.offset = num_blocks;
334 ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
343 item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
344 refs = btrfs_extent_refs(l, item);
345 btrfs_set_extent_refs(l, item, refs + 1);
346 btrfs_mark_buffer_dirty(path->nodes[0]);
348 btrfs_release_path(root->fs_info->extent_root, path);
349 btrfs_free_path(path);
350 finish_current_insert(trans, root->fs_info->extent_root);
351 del_pending_extents(trans, root->fs_info->extent_root);
355 int btrfs_extent_post_op(struct btrfs_trans_handle *trans,
356 struct btrfs_root *root)
358 finish_current_insert(trans, root->fs_info->extent_root);
359 del_pending_extents(trans, root->fs_info->extent_root);
363 static int lookup_extent_ref(struct btrfs_trans_handle *trans,
364 struct btrfs_root *root, u64 blocknr,
365 u64 num_blocks, u32 *refs)
367 struct btrfs_path *path;
369 struct btrfs_key key;
370 struct extent_buffer *l;
371 struct btrfs_extent_item *item;
373 path = btrfs_alloc_path();
374 key.objectid = blocknr;
375 key.offset = num_blocks;
376 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
377 ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
382 btrfs_print_leaf(root, path->nodes[0]);
383 printk("failed to find block number %Lu\n", blocknr);
387 item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
388 *refs = btrfs_extent_refs(l, item);
390 btrfs_free_path(path);
394 int btrfs_inc_root_ref(struct btrfs_trans_handle *trans,
395 struct btrfs_root *root)
397 return btrfs_inc_extent_ref(trans, root,
398 extent_buffer_blocknr(root->node), 1);
401 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
402 struct extent_buffer *buf)
406 struct btrfs_key key;
407 struct btrfs_file_extent_item *fi;
417 leaf = btrfs_is_leaf(buf);
418 nritems = btrfs_header_nritems(buf);
419 for (i = 0; i < nritems; i++) {
422 btrfs_item_key_to_cpu(buf, &key, i);
423 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
425 fi = btrfs_item_ptr(buf, i,
426 struct btrfs_file_extent_item);
427 if (btrfs_file_extent_type(buf, fi) ==
428 BTRFS_FILE_EXTENT_INLINE)
430 disk_blocknr = btrfs_file_extent_disk_blocknr(buf, fi);
431 if (disk_blocknr == 0)
433 ret = btrfs_inc_extent_ref(trans, root, disk_blocknr,
434 btrfs_file_extent_disk_num_blocks(buf, fi));
440 blocknr = btrfs_node_blockptr(buf, i);
441 ret = btrfs_inc_extent_ref(trans, root, blocknr, 1);
451 for (i =0; i < faili; i++) {
454 btrfs_item_key_to_cpu(buf, &key, i);
455 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
457 fi = btrfs_item_ptr(buf, i,
458 struct btrfs_file_extent_item);
459 if (btrfs_file_extent_type(buf, fi) ==
460 BTRFS_FILE_EXTENT_INLINE)
462 disk_blocknr = btrfs_file_extent_disk_blocknr(buf, fi);
463 if (disk_blocknr == 0)
465 err = btrfs_free_extent(trans, root, disk_blocknr,
466 btrfs_file_extent_disk_num_blocks(buf,
470 blocknr = btrfs_node_blockptr(buf, i);
471 err = btrfs_free_extent(trans, root, blocknr, 1, 0);
478 static int write_one_cache_group(struct btrfs_trans_handle *trans,
479 struct btrfs_root *root,
480 struct btrfs_path *path,
481 struct btrfs_block_group_cache *cache)
485 struct btrfs_root *extent_root = root->fs_info->extent_root;
487 struct extent_buffer *leaf;
489 ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1);
494 leaf = path->nodes[0];
495 bi = btrfs_item_ptr_offset(leaf, path->slots[0]);
496 write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item));
497 btrfs_mark_buffer_dirty(leaf);
498 btrfs_release_path(extent_root, path);
500 finish_current_insert(trans, extent_root);
501 pending_ret = del_pending_extents(trans, extent_root);
510 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
511 struct btrfs_root *root)
513 struct extent_map_tree *block_group_cache;
514 struct btrfs_block_group_cache *cache;
518 struct btrfs_path *path;
524 block_group_cache = &root->fs_info->block_group_cache;
525 path = btrfs_alloc_path();
530 ret = find_first_extent_bit(block_group_cache, last,
531 &start, &end, BLOCK_GROUP_DIRTY);
536 ret = get_state_private(block_group_cache, start, &ptr);
540 cache = (struct btrfs_block_group_cache *)ptr;
541 err = write_one_cache_group(trans, root,
544 * if we fail to write the cache group, we want
545 * to keep it marked dirty in hopes that a later
552 clear_extent_bits(block_group_cache, start, end,
553 BLOCK_GROUP_DIRTY, GFP_NOFS);
555 btrfs_free_path(path);
559 static int update_block_group(struct btrfs_trans_handle *trans,
560 struct btrfs_root *root,
561 u64 blocknr, u64 num, int alloc, int mark_free,
564 struct btrfs_block_group_cache *cache;
565 struct btrfs_fs_info *info = root->fs_info;
573 cache = btrfs_lookup_block_group(info, blocknr);
577 block_in_group = blocknr - cache->key.objectid;
578 WARN_ON(block_in_group > cache->key.offset);
579 start = cache->key.objectid;
580 end = start + cache->key.offset - 1;
581 set_extent_bits(&info->block_group_cache, start, end,
582 BLOCK_GROUP_DIRTY, GFP_NOFS);
584 old_val = btrfs_block_group_used(&cache->item);
585 num = min(total, cache->key.offset - block_in_group);
587 if (cache->data != data &&
588 old_val < (cache->key.offset >> 1)) {
594 bit_to_clear = BLOCK_GROUP_DATA;
595 bit_to_set = BLOCK_GROUP_METADATA;
597 BTRFS_BLOCK_GROUP_DATA;
599 bit_to_clear = BLOCK_GROUP_METADATA;
600 bit_to_set = BLOCK_GROUP_DATA;
602 ~BTRFS_BLOCK_GROUP_DATA;
604 clear_extent_bits(&info->block_group_cache,
605 start, end, bit_to_clear,
607 set_extent_bits(&info->block_group_cache,
608 start, end, bit_to_set,
615 set_extent_dirty(&info->free_space_cache,
616 blocknr, blocknr + num - 1,
620 btrfs_set_block_group_used(&cache->item, old_val);
627 int btrfs_copy_pinned(struct btrfs_root *root, struct extent_map_tree *copy)
632 struct extent_map_tree *pinned_extents = &root->fs_info->pinned_extents;
636 ret = find_first_extent_bit(pinned_extents, last,
637 &start, &end, EXTENT_DIRTY);
640 set_extent_dirty(copy, start, end, GFP_NOFS);
646 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
647 struct btrfs_root *root,
648 struct extent_map_tree *unpin)
653 struct extent_map_tree *pinned_extents = &root->fs_info->pinned_extents;
654 struct extent_map_tree *free_space_cache;
656 free_space_cache = &root->fs_info->free_space_cache;
659 ret = find_first_extent_bit(unpin, 0, &start, &end,
664 clear_extent_dirty(pinned_extents, start, end,
666 clear_extent_dirty(unpin, start, end, GFP_NOFS);
667 set_extent_dirty(free_space_cache, start, end, GFP_NOFS);
672 static int finish_current_insert(struct btrfs_trans_handle *trans, struct
673 btrfs_root *extent_root)
675 struct btrfs_key ins;
676 struct btrfs_extent_item extent_item;
681 struct btrfs_fs_info *info = extent_root->fs_info;
683 btrfs_set_stack_extent_refs(&extent_item, 1);
684 btrfs_set_key_type(&ins, BTRFS_EXTENT_ITEM_KEY);
685 btrfs_set_stack_extent_owner(&extent_item,
686 extent_root->root_key.objectid);
689 ret = find_first_extent_bit(&info->extent_ins, 0, &start,
690 &end, EXTENT_LOCKED);
694 ins.objectid = start;
695 ins.offset = end + 1 - start;
696 err = btrfs_insert_item(trans, extent_root, &ins,
697 &extent_item, sizeof(extent_item));
698 clear_extent_bits(&info->extent_ins, start, end, EXTENT_LOCKED,
704 static int pin_down_block(struct btrfs_root *root, u64 blocknr, int pending)
707 struct extent_buffer *buf;
710 buf = btrfs_find_tree_block(root, blocknr);
712 if (btrfs_buffer_uptodate(buf)) {
714 root->fs_info->running_transaction->transid;
715 if (btrfs_header_generation(buf) == transid) {
716 free_extent_buffer(buf);
720 free_extent_buffer(buf);
722 set_extent_dirty(&root->fs_info->pinned_extents,
723 blocknr, blocknr, GFP_NOFS);
725 set_extent_bits(&root->fs_info->pending_del,
726 blocknr, blocknr, EXTENT_LOCKED, GFP_NOFS);
733 * remove an extent from the root, returns 0 on success
735 static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
736 *root, u64 blocknr, u64 num_blocks, int pin,
739 struct btrfs_path *path;
740 struct btrfs_key key;
741 struct btrfs_fs_info *info = root->fs_info;
742 struct btrfs_root *extent_root = info->extent_root;
743 struct extent_buffer *leaf;
745 struct btrfs_extent_item *ei;
748 key.objectid = blocknr;
749 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
750 key.offset = num_blocks;
752 path = btrfs_alloc_path();
756 ret = btrfs_search_slot(trans, extent_root, &key, path, -1, 1);
761 leaf = path->nodes[0];
762 ei = btrfs_item_ptr(leaf, path->slots[0],
763 struct btrfs_extent_item);
764 refs = btrfs_extent_refs(leaf, ei);
767 btrfs_set_extent_refs(leaf, ei, refs);
768 btrfs_mark_buffer_dirty(leaf);
771 u64 super_blocks_used, root_blocks_used;
774 ret = pin_down_block(root, blocknr, 0);
778 /* block accounting for super block */
779 super_blocks_used = btrfs_super_blocks_used(&info->super_copy);
780 btrfs_set_super_blocks_used(&info->super_copy,
781 super_blocks_used - num_blocks);
783 /* block accounting for root item */
784 root_blocks_used = btrfs_root_used(&root->root_item);
785 btrfs_set_root_used(&root->root_item,
786 root_blocks_used - num_blocks);
788 ret = btrfs_del_item(trans, extent_root, path);
792 ret = update_block_group(trans, root, blocknr, num_blocks, 0,
796 btrfs_free_path(path);
797 finish_current_insert(trans, extent_root);
802 * find all the blocks marked as pending in the radix tree and remove
803 * them from the extent map
805 static int del_pending_extents(struct btrfs_trans_handle *trans, struct
806 btrfs_root *extent_root)
812 struct extent_map_tree *pending_del;
813 struct extent_map_tree *pinned_extents;
815 pending_del = &extent_root->fs_info->pending_del;
816 pinned_extents = &extent_root->fs_info->pinned_extents;
819 ret = find_first_extent_bit(pending_del, 0, &start, &end,
824 set_extent_dirty(pinned_extents, start, end, GFP_NOFS);
825 clear_extent_bits(pending_del, start, end, EXTENT_LOCKED,
827 ret = __free_extent(trans, extent_root,
828 start, end + 1 - start, 0, 0);
836 * remove an extent from the root, returns 0 on success
838 int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
839 *root, u64 blocknr, u64 num_blocks, int pin)
841 struct btrfs_root *extent_root = root->fs_info->extent_root;
845 if (root == extent_root) {
846 pin_down_block(root, blocknr, 1);
849 ret = __free_extent(trans, root, blocknr, num_blocks, pin, pin == 0);
850 pending_ret = del_pending_extents(trans, root->fs_info->extent_root);
851 return ret ? ret : pending_ret;
855 * walks the btree of allocated extents and find a hole of a given size.
856 * The key ins is changed to record the hole:
857 * ins->objectid == block start
858 * ins->flags = BTRFS_EXTENT_ITEM_KEY
859 * ins->offset == number of blocks
860 * Any available blocks before search_start are skipped.
862 static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
863 *orig_root, u64 num_blocks, u64 empty_size,
864 u64 search_start, u64 search_end, u64 hint_block,
865 struct btrfs_key *ins, u64 exclude_start,
866 u64 exclude_nr, int data)
868 struct btrfs_path *path;
869 struct btrfs_key key;
874 u64 orig_search_start = search_start;
876 struct extent_buffer *l;
877 struct btrfs_root * root = orig_root->fs_info->extent_root;
878 struct btrfs_fs_info *info = root->fs_info;
879 int total_needed = num_blocks;
881 struct btrfs_block_group_cache *block_group;
884 u64 cached_search_start = 0;
886 WARN_ON(num_blocks < 1);
887 btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
889 level = btrfs_header_level(root->node);
891 if (search_end == (u64)-1)
892 search_end = btrfs_super_total_blocks(&info->super_copy);
894 block_group = btrfs_lookup_block_group(info, hint_block);
895 block_group = btrfs_find_block_group(root, block_group,
896 hint_block, data, 1);
898 block_group = btrfs_find_block_group(root,
899 trans->block_group, 0,
903 total_needed += empty_size;
904 path = btrfs_alloc_path();
907 search_start = find_search_start(root, &block_group,
908 search_start, total_needed, data);
909 cached_search_start = search_start;
911 btrfs_init_path(path);
912 ins->objectid = search_start;
917 ret = btrfs_search_slot(trans, root, ins, path, 0, 0);
921 if (path->slots[0] > 0) {
926 btrfs_item_key_to_cpu(l, &key, path->slots[0]);
929 * a rare case, go back one key if we hit a block group item
930 * instead of an extent item
932 if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY &&
933 key.objectid + key.offset >= search_start) {
934 ins->objectid = key.objectid;
935 ins->offset = key.offset - 1;
936 btrfs_release_path(root, path);
937 ret = btrfs_search_slot(trans, root, ins, path, 0, 0);
941 if (path->slots[0] > 0) {
948 slot = path->slots[0];
949 if (slot >= btrfs_header_nritems(l)) {
950 ret = btrfs_next_leaf(root, path);
956 ins->objectid = search_start;
957 ins->offset = search_end - search_start;
961 ins->objectid = last_block > search_start ?
962 last_block : search_start;
963 ins->offset = search_end - ins->objectid;
966 btrfs_item_key_to_cpu(l, &key, slot);
968 if (key.objectid >= search_start && key.objectid > last_block &&
970 if (last_block < search_start)
971 last_block = search_start;
972 hole_size = key.objectid - last_block;
973 if (hole_size >= num_blocks) {
974 ins->objectid = last_block;
975 ins->offset = hole_size;
979 if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY) {
981 last_block = key.objectid;
989 last_block = key.objectid + key.offset;
991 if (!full_scan && last_block >= block_group->key.objectid +
992 block_group->key.offset) {
993 btrfs_release_path(root, path);
994 search_start = block_group->key.objectid +
995 block_group->key.offset * 2;
1003 /* we have to make sure we didn't find an extent that has already
1004 * been allocated by the map tree or the original allocation
1006 btrfs_release_path(root, path);
1007 BUG_ON(ins->objectid < search_start);
1009 if (ins->objectid + num_blocks >= search_end)
1012 if (test_range_bit(&info->extent_ins, ins->objectid,
1013 ins->objectid + num_blocks -1, EXTENT_LOCKED, 0)) {
1014 search_start = ins->objectid + num_blocks;
1017 if (test_range_bit(&info->pinned_extents, ins->objectid,
1018 ins->objectid + num_blocks -1, EXTENT_DIRTY, 0)) {
1019 search_start = ins->objectid + num_blocks;
1022 if (exclude_nr > 0 && (ins->objectid + num_blocks > exclude_start &&
1023 ins->objectid < exclude_start + exclude_nr)) {
1024 search_start = exclude_start + exclude_nr;
1028 block_group = btrfs_lookup_block_group(info, ins->objectid);
1030 trans->block_group = block_group;
1032 ins->offset = num_blocks;
1033 btrfs_free_path(path);
1037 if (search_start + num_blocks >= search_end) {
1039 search_start = orig_search_start;
1046 total_needed -= empty_size;
1051 block_group = btrfs_lookup_block_group(info, search_start);
1054 block_group = btrfs_find_block_group(root, block_group,
1055 search_start, data, 0);
1059 btrfs_release_path(root, path);
1060 btrfs_free_path(path);
1064 * finds a free extent and does all the dirty work required for allocation
1065 * returns the key for the extent through ins, and a tree buffer for
1066 * the first block of the extent through buf.
1068 * returns 0 if everything worked, non-zero otherwise.
1070 int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
1071 struct btrfs_root *root, u64 owner,
1072 u64 num_blocks, u64 empty_size, u64 hint_block,
1073 u64 search_end, struct btrfs_key *ins, int data)
1077 u64 super_blocks_used, root_blocks_used;
1078 u64 search_start = 0;
1079 struct btrfs_fs_info *info = root->fs_info;
1080 struct btrfs_root *extent_root = info->extent_root;
1081 struct btrfs_extent_item extent_item;
1083 btrfs_set_stack_extent_refs(&extent_item, 1);
1084 btrfs_set_stack_extent_owner(&extent_item, owner);
1086 WARN_ON(num_blocks < 1);
1087 ret = find_free_extent(trans, root, num_blocks, empty_size,
1088 search_start, search_end, hint_block, ins,
1089 trans->alloc_exclude_start,
1090 trans->alloc_exclude_nr, data);
1095 /* block accounting for super block */
1096 super_blocks_used = btrfs_super_blocks_used(&info->super_copy);
1097 btrfs_set_super_blocks_used(&info->super_copy, super_blocks_used +
1100 /* block accounting for root item */
1101 root_blocks_used = btrfs_root_used(&root->root_item);
1102 btrfs_set_root_used(&root->root_item, root_blocks_used +
1105 clear_extent_dirty(&root->fs_info->free_space_cache,
1106 ins->objectid, ins->objectid + ins->offset - 1,
1109 if (root == extent_root) {
1110 BUG_ON(num_blocks != 1);
1111 set_extent_bits(&root->fs_info->extent_ins, ins->objectid,
1112 ins->objectid + ins->offset - 1,
1113 EXTENT_LOCKED, GFP_NOFS);
1117 WARN_ON(trans->alloc_exclude_nr);
1118 trans->alloc_exclude_start = ins->objectid;
1119 trans->alloc_exclude_nr = ins->offset;
1120 ret = btrfs_insert_item(trans, extent_root, ins, &extent_item,
1121 sizeof(extent_item));
1123 trans->alloc_exclude_start = 0;
1124 trans->alloc_exclude_nr = 0;
1127 finish_current_insert(trans, extent_root);
1128 pending_ret = del_pending_extents(trans, extent_root);
1138 ret = update_block_group(trans, root, ins->objectid, ins->offset, 1, 0,
1145 * helper function to allocate a block for a given tree
1146 * returns the tree buffer or NULL.
1148 struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
1149 struct btrfs_root *root, u64 hint,
1152 struct btrfs_key ins;
1154 struct extent_buffer *buf;
1156 ret = btrfs_alloc_extent(trans, root, root->root_key.objectid,
1157 1, empty_size, hint, (u64)-1, &ins, 0);
1160 return ERR_PTR(ret);
1162 buf = btrfs_find_create_tree_block(root, ins.objectid);
1164 btrfs_free_extent(trans, root, ins.objectid, 1, 0);
1165 return ERR_PTR(-ENOMEM);
1167 btrfs_set_buffer_uptodate(buf);
1168 buf->alloc_addr = (unsigned long)__builtin_return_address(0);
1169 set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
1170 buf->start + buf->len - 1, GFP_NOFS);
1172 set_buffer_checked(buf);
1173 set_buffer_defrag(buf);
1175 /* FIXME!!!!!!!!!!!!!!!!
1176 set_radix_bit(&trans->transaction->dirty_pages, buf->pages[0]->index);
1178 trans->blocks_used++;
1182 static int drop_leaf_ref(struct btrfs_trans_handle *trans,
1183 struct btrfs_root *root, struct extent_buffer *leaf)
1185 struct btrfs_key key;
1186 struct btrfs_file_extent_item *fi;
1191 BUG_ON(!btrfs_is_leaf(leaf));
1192 nritems = btrfs_header_nritems(leaf);
1193 for (i = 0; i < nritems; i++) {
1196 btrfs_item_key_to_cpu(leaf, &key, i);
1197 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
1199 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
1200 if (btrfs_file_extent_type(leaf, fi) ==
1201 BTRFS_FILE_EXTENT_INLINE)
1204 * FIXME make sure to insert a trans record that
1205 * repeats the snapshot del on crash
1207 disk_blocknr = btrfs_file_extent_disk_blocknr(leaf, fi);
1208 if (disk_blocknr == 0)
1210 ret = btrfs_free_extent(trans, root, disk_blocknr,
1211 btrfs_file_extent_disk_num_blocks(leaf, fi), 0);
1217 static void reada_walk_down(struct btrfs_root *root,
1218 struct extent_buffer *node)
1226 nritems = btrfs_header_nritems(node);
1227 for (i = 0; i < nritems; i++) {
1228 blocknr = btrfs_node_blockptr(node, i);
1229 ret = lookup_extent_ref(NULL, root, blocknr, 1, &refs);
1233 mutex_unlock(&root->fs_info->fs_mutex);
1234 ret = readahead_tree_block(root, blocknr);
1236 mutex_lock(&root->fs_info->fs_mutex);
1243 * helper function for drop_snapshot, this walks down the tree dropping ref
1244 * counts as it goes.
1246 static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
1247 *root, struct btrfs_path *path, int *level)
1249 struct extent_buffer *next;
1250 struct extent_buffer *cur;
1255 WARN_ON(*level < 0);
1256 WARN_ON(*level >= BTRFS_MAX_LEVEL);
1257 ret = lookup_extent_ref(trans, root,
1258 extent_buffer_blocknr(path->nodes[*level]),
1265 * walk down to the last node level and free all the leaves
1267 while(*level >= 0) {
1268 WARN_ON(*level < 0);
1269 WARN_ON(*level >= BTRFS_MAX_LEVEL);
1270 cur = path->nodes[*level];
1272 if (*level > 0 && path->slots[*level] == 0)
1273 reada_walk_down(root, cur);
1275 if (btrfs_header_level(cur) != *level)
1278 if (path->slots[*level] >=
1279 btrfs_header_nritems(cur))
1282 ret = drop_leaf_ref(trans, root, cur);
1286 blocknr = btrfs_node_blockptr(cur, path->slots[*level]);
1287 ret = lookup_extent_ref(trans, root, blocknr, 1, &refs);
1290 path->slots[*level]++;
1291 ret = btrfs_free_extent(trans, root, blocknr, 1, 1);
1295 next = btrfs_find_tree_block(root, blocknr);
1296 if (!next || !btrfs_buffer_uptodate(next)) {
1297 free_extent_buffer(next);
1298 mutex_unlock(&root->fs_info->fs_mutex);
1299 next = read_tree_block(root, blocknr);
1300 mutex_lock(&root->fs_info->fs_mutex);
1302 /* we dropped the lock, check one more time */
1303 ret = lookup_extent_ref(trans, root, blocknr, 1, &refs);
1306 path->slots[*level]++;
1307 free_extent_buffer(next);
1308 ret = btrfs_free_extent(trans, root,
1314 WARN_ON(*level <= 0);
1315 if (path->nodes[*level-1])
1316 free_extent_buffer(path->nodes[*level-1]);
1317 path->nodes[*level-1] = next;
1318 *level = btrfs_header_level(next);
1319 path->slots[*level] = 0;
1322 WARN_ON(*level < 0);
1323 WARN_ON(*level >= BTRFS_MAX_LEVEL);
1324 ret = btrfs_free_extent(trans, root,
1325 extent_buffer_blocknr(path->nodes[*level]), 1, 1);
1326 free_extent_buffer(path->nodes[*level]);
1327 path->nodes[*level] = NULL;
1334 * helper for dropping snapshots. This walks back up the tree in the path
1335 * to find the first node higher up where we haven't yet gone through
1338 static int walk_up_tree(struct btrfs_trans_handle *trans, struct btrfs_root
1339 *root, struct btrfs_path *path, int *level)
1344 struct btrfs_root_item *root_item = &root->root_item;
1346 for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) {
1347 slot = path->slots[i];
1348 if (slot < btrfs_header_nritems(path->nodes[i]) - 1) {
1349 struct extent_buffer *node;
1350 struct btrfs_disk_key disk_key;
1351 node = path->nodes[i];
1354 WARN_ON(*level == 0);
1355 btrfs_node_key(node, &disk_key, path->slots[i]);
1356 memcpy(&root_item->drop_progress,
1357 &disk_key, sizeof(disk_key));
1358 root_item->drop_level = i;
1361 ret = btrfs_free_extent(trans, root,
1362 extent_buffer_blocknr(path->nodes[*level]),
1365 free_extent_buffer(path->nodes[*level]);
1366 path->nodes[*level] = NULL;
1374 * drop the reference count on the tree rooted at 'snap'. This traverses
1375 * the tree freeing any blocks that have a ref count of zero after being
1378 int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
1384 struct btrfs_path *path;
1387 struct btrfs_root_item *root_item = &root->root_item;
1389 path = btrfs_alloc_path();
1392 level = btrfs_header_level(root->node);
1394 if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
1395 path->nodes[level] = root->node;
1396 extent_buffer_get(root->node);
1397 path->slots[level] = 0;
1399 struct btrfs_key key;
1400 struct btrfs_disk_key found_key;
1401 struct extent_buffer *node;
1403 btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
1404 level = root_item->drop_level;
1405 path->lowest_level = level;
1406 wret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1411 node = path->nodes[level];
1412 btrfs_node_key(node, &found_key, path->slots[level]);
1413 WARN_ON(memcmp(&found_key, &root_item->drop_progress,
1414 sizeof(found_key)));
1417 wret = walk_down_tree(trans, root, path, &level);
1423 wret = walk_up_tree(trans, root, path, &level);
1431 for (i = 0; i <= orig_level; i++) {
1432 if (path->nodes[i]) {
1433 free_extent_buffer(path->nodes[i]);
1438 btrfs_free_path(path);
1442 int btrfs_free_block_groups(struct btrfs_fs_info *info)
1449 ret = find_first_extent_bit(&info->block_group_cache, 0,
1450 &start, &end, (unsigned int)-1);
1453 clear_extent_bits(&info->block_group_cache, start,
1454 end, (unsigned int)-1, GFP_NOFS);
1457 ret = find_first_extent_bit(&info->free_space_cache, 0,
1458 &start, &end, EXTENT_DIRTY);
1461 clear_extent_dirty(&info->free_space_cache, start,
1467 int btrfs_read_block_groups(struct btrfs_root *root)
1469 struct btrfs_path *path;
1473 struct btrfs_block_group_cache *cache;
1474 struct btrfs_fs_info *info = root->fs_info;
1475 struct extent_map_tree *block_group_cache;
1476 struct btrfs_key key;
1477 struct btrfs_key found_key;
1478 struct extent_buffer *leaf;
1479 u64 group_size_blocks;
1481 block_group_cache = &info->block_group_cache;
1483 group_size_blocks = BTRFS_BLOCK_GROUP_SIZE >>
1484 info->sb->s_blocksize_bits;
1486 root = info->extent_root;
1488 key.offset = group_size_blocks;
1489 btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
1491 path = btrfs_alloc_path();
1496 ret = btrfs_search_slot(NULL, info->extent_root,
1502 leaf = path->nodes[0];
1503 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1504 cache = kmalloc(sizeof(*cache), GFP_NOFS);
1510 read_extent_buffer(leaf, &cache->item,
1511 btrfs_item_ptr_offset(leaf, path->slots[0]),
1512 sizeof(cache->item));
1513 memcpy(&cache->key, &found_key, sizeof(found_key));
1516 key.objectid = found_key.objectid + found_key.offset;
1517 btrfs_release_path(root, path);
1519 if (cache->item.flags & BTRFS_BLOCK_GROUP_DATA) {
1520 bit = BLOCK_GROUP_DATA;
1523 bit = BLOCK_GROUP_METADATA;
1527 /* use EXTENT_LOCKED to prevent merging */
1528 set_extent_bits(block_group_cache, found_key.objectid,
1529 found_key.objectid + found_key.offset - 1,
1530 bit | EXTENT_LOCKED, GFP_NOFS);
1531 set_state_private(block_group_cache, found_key.objectid,
1535 btrfs_super_total_blocks(&info->super_copy))
1539 btrfs_free_path(path);