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.
18 #include <linux/sched.h>
19 #include <linux/pagemap.h>
20 #include <linux/writeback.h>
21 #include <linux/blkdev.h>
22 #include <linux/sort.h>
23 #include <linux/rcupdate.h>
28 #include "print-tree.h"
29 #include "transaction.h"
32 #include "free-space-cache.h"
34 static int update_reserved_extents(struct btrfs_root *root,
35 u64 bytenr, u64 num, int reserve);
36 static int update_block_group(struct btrfs_trans_handle *trans,
37 struct btrfs_root *root,
38 u64 bytenr, u64 num_bytes, int alloc,
40 static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
41 struct btrfs_root *root,
42 u64 bytenr, u64 num_bytes, u64 parent,
43 u64 root_objectid, u64 owner_objectid,
44 u64 owner_offset, int refs_to_drop,
45 struct btrfs_delayed_extent_op *extra_op);
46 static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
47 struct extent_buffer *leaf,
48 struct btrfs_extent_item *ei);
49 static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
50 struct btrfs_root *root,
51 u64 parent, u64 root_objectid,
52 u64 flags, u64 owner, u64 offset,
53 struct btrfs_key *ins, int ref_mod);
54 static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
55 struct btrfs_root *root,
56 u64 parent, u64 root_objectid,
57 u64 flags, struct btrfs_disk_key *key,
58 int level, struct btrfs_key *ins);
60 static int do_chunk_alloc(struct btrfs_trans_handle *trans,
61 struct btrfs_root *extent_root, u64 alloc_bytes,
62 u64 flags, int force);
64 static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits)
66 return (cache->flags & bits) == bits;
70 * this adds the block group to the fs_info rb tree for the block group
73 static int btrfs_add_block_group_cache(struct btrfs_fs_info *info,
74 struct btrfs_block_group_cache *block_group)
77 struct rb_node *parent = NULL;
78 struct btrfs_block_group_cache *cache;
80 spin_lock(&info->block_group_cache_lock);
81 p = &info->block_group_cache_tree.rb_node;
85 cache = rb_entry(parent, struct btrfs_block_group_cache,
87 if (block_group->key.objectid < cache->key.objectid) {
89 } else if (block_group->key.objectid > cache->key.objectid) {
92 spin_unlock(&info->block_group_cache_lock);
97 rb_link_node(&block_group->cache_node, parent, p);
98 rb_insert_color(&block_group->cache_node,
99 &info->block_group_cache_tree);
100 spin_unlock(&info->block_group_cache_lock);
106 * This will return the block group at or after bytenr if contains is 0, else
107 * it will return the block group that contains the bytenr
109 static struct btrfs_block_group_cache *
110 block_group_cache_tree_search(struct btrfs_fs_info *info, u64 bytenr,
113 struct btrfs_block_group_cache *cache, *ret = NULL;
117 spin_lock(&info->block_group_cache_lock);
118 n = info->block_group_cache_tree.rb_node;
121 cache = rb_entry(n, struct btrfs_block_group_cache,
123 end = cache->key.objectid + cache->key.offset - 1;
124 start = cache->key.objectid;
126 if (bytenr < start) {
127 if (!contains && (!ret || start < ret->key.objectid))
130 } else if (bytenr > start) {
131 if (contains && bytenr <= end) {
142 atomic_inc(&ret->count);
143 spin_unlock(&info->block_group_cache_lock);
149 * this is only called by cache_block_group, since we could have freed extents
150 * we need to check the pinned_extents for any extents that can't be used yet
151 * since their free space will be released as soon as the transaction commits.
153 static int add_new_free_space(struct btrfs_block_group_cache *block_group,
154 struct btrfs_fs_info *info, u64 start, u64 end)
156 u64 extent_start, extent_end, size;
159 while (start < end) {
160 ret = find_first_extent_bit(&info->pinned_extents, start,
161 &extent_start, &extent_end,
166 if (extent_start == start) {
167 start = extent_end + 1;
168 } else if (extent_start > start && extent_start < end) {
169 size = extent_start - start;
170 ret = btrfs_add_free_space(block_group, start,
173 start = extent_end + 1;
181 ret = btrfs_add_free_space(block_group, start, size);
188 static int remove_sb_from_cache(struct btrfs_root *root,
189 struct btrfs_block_group_cache *cache)
196 for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
197 bytenr = btrfs_sb_offset(i);
198 ret = btrfs_rmap_block(&root->fs_info->mapping_tree,
199 cache->key.objectid, bytenr, 0,
200 &logical, &nr, &stripe_len);
203 btrfs_remove_free_space(cache, logical[nr],
211 static int cache_block_group(struct btrfs_root *root,
212 struct btrfs_block_group_cache *block_group)
214 struct btrfs_path *path;
216 struct btrfs_key key;
217 struct extent_buffer *leaf;
224 root = root->fs_info->extent_root;
226 if (block_group->cached)
229 path = btrfs_alloc_path();
235 * we get into deadlocks with paths held by callers of this function.
236 * since the alloc_mutex is protecting things right now, just
237 * skip the locking here
239 path->skip_locking = 1;
240 last = max_t(u64, block_group->key.objectid, BTRFS_SUPER_INFO_OFFSET);
243 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
244 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
249 leaf = path->nodes[0];
250 slot = path->slots[0];
251 if (slot >= btrfs_header_nritems(leaf)) {
252 ret = btrfs_next_leaf(root, path);
260 btrfs_item_key_to_cpu(leaf, &key, slot);
261 if (key.objectid < block_group->key.objectid)
264 if (key.objectid >= block_group->key.objectid +
265 block_group->key.offset)
268 if (btrfs_key_type(&key) == BTRFS_EXTENT_ITEM_KEY) {
269 add_new_free_space(block_group, root->fs_info, last,
272 last = key.objectid + key.offset;
278 add_new_free_space(block_group, root->fs_info, last,
279 block_group->key.objectid +
280 block_group->key.offset);
282 block_group->cached = 1;
283 remove_sb_from_cache(root, block_group);
286 btrfs_free_path(path);
291 * return the block group that starts at or after bytenr
293 static struct btrfs_block_group_cache *
294 btrfs_lookup_first_block_group(struct btrfs_fs_info *info, u64 bytenr)
296 struct btrfs_block_group_cache *cache;
298 cache = block_group_cache_tree_search(info, bytenr, 0);
304 * return the block group that contains the given bytenr
306 struct btrfs_block_group_cache *btrfs_lookup_block_group(
307 struct btrfs_fs_info *info,
310 struct btrfs_block_group_cache *cache;
312 cache = block_group_cache_tree_search(info, bytenr, 1);
317 void btrfs_put_block_group(struct btrfs_block_group_cache *cache)
319 if (atomic_dec_and_test(&cache->count))
323 static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info,
326 struct list_head *head = &info->space_info;
327 struct btrfs_space_info *found;
330 list_for_each_entry_rcu(found, head, list) {
331 if (found->flags == flags) {
341 * after adding space to the filesystem, we need to clear the full flags
342 * on all the space infos.
344 void btrfs_clear_space_info_full(struct btrfs_fs_info *info)
346 struct list_head *head = &info->space_info;
347 struct btrfs_space_info *found;
350 list_for_each_entry_rcu(found, head, list)
355 static u64 div_factor(u64 num, int factor)
364 u64 btrfs_find_block_group(struct btrfs_root *root,
365 u64 search_start, u64 search_hint, int owner)
367 struct btrfs_block_group_cache *cache;
369 u64 last = max(search_hint, search_start);
376 cache = btrfs_lookup_first_block_group(root->fs_info, last);
380 spin_lock(&cache->lock);
381 last = cache->key.objectid + cache->key.offset;
382 used = btrfs_block_group_used(&cache->item);
384 if ((full_search || !cache->ro) &&
385 block_group_bits(cache, BTRFS_BLOCK_GROUP_METADATA)) {
386 if (used + cache->pinned + cache->reserved <
387 div_factor(cache->key.offset, factor)) {
388 group_start = cache->key.objectid;
389 spin_unlock(&cache->lock);
390 btrfs_put_block_group(cache);
394 spin_unlock(&cache->lock);
395 btrfs_put_block_group(cache);
403 if (!full_search && factor < 10) {
413 /* simple helper to search for an existing extent at a given offset */
414 int btrfs_lookup_extent(struct btrfs_root *root, u64 start, u64 len)
417 struct btrfs_key key;
418 struct btrfs_path *path;
420 path = btrfs_alloc_path();
422 key.objectid = start;
424 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
425 ret = btrfs_search_slot(NULL, root->fs_info->extent_root, &key, path,
427 btrfs_free_path(path);
432 * Back reference rules. Back refs have three main goals:
434 * 1) differentiate between all holders of references to an extent so that
435 * when a reference is dropped we can make sure it was a valid reference
436 * before freeing the extent.
438 * 2) Provide enough information to quickly find the holders of an extent
439 * if we notice a given block is corrupted or bad.
441 * 3) Make it easy to migrate blocks for FS shrinking or storage pool
442 * maintenance. This is actually the same as #2, but with a slightly
443 * different use case.
445 * There are two kinds of back refs. The implicit back refs is optimized
446 * for pointers in non-shared tree blocks. For a given pointer in a block,
447 * back refs of this kind provide information about the block's owner tree
448 * and the pointer's key. These information allow us to find the block by
449 * b-tree searching. The full back refs is for pointers in tree blocks not
450 * referenced by their owner trees. The location of tree block is recorded
451 * in the back refs. Actually the full back refs is generic, and can be
452 * used in all cases the implicit back refs is used. The major shortcoming
453 * of the full back refs is its overhead. Every time a tree block gets
454 * COWed, we have to update back refs entry for all pointers in it.
456 * For a newly allocated tree block, we use implicit back refs for
457 * pointers in it. This means most tree related operations only involve
458 * implicit back refs. For a tree block created in old transaction, the
459 * only way to drop a reference to it is COW it. So we can detect the
460 * event that tree block loses its owner tree's reference and do the
461 * back refs conversion.
463 * When a tree block is COW'd through a tree, there are four cases:
465 * The reference count of the block is one and the tree is the block's
466 * owner tree. Nothing to do in this case.
468 * The reference count of the block is one and the tree is not the
469 * block's owner tree. In this case, full back refs is used for pointers
470 * in the block. Remove these full back refs, add implicit back refs for
471 * every pointers in the new block.
473 * The reference count of the block is greater than one and the tree is
474 * the block's owner tree. In this case, implicit back refs is used for
475 * pointers in the block. Add full back refs for every pointers in the
476 * block, increase lower level extents' reference counts. The original
477 * implicit back refs are entailed to the new block.
479 * The reference count of the block is greater than one and the tree is
480 * not the block's owner tree. Add implicit back refs for every pointer in
481 * the new block, increase lower level extents' reference count.
483 * Back Reference Key composing:
485 * The key objectid corresponds to the first byte in the extent,
486 * The key type is used to differentiate between types of back refs.
487 * There are different meanings of the key offset for different types
490 * File extents can be referenced by:
492 * - multiple snapshots, subvolumes, or different generations in one subvol
493 * - different files inside a single subvolume
494 * - different offsets inside a file (bookend extents in file.c)
496 * The extent ref structure for the implicit back refs has fields for:
498 * - Objectid of the subvolume root
499 * - objectid of the file holding the reference
500 * - original offset in the file
501 * - how many bookend extents
503 * The key offset for the implicit back refs is hash of the first
506 * The extent ref structure for the full back refs has field for:
508 * - number of pointers in the tree leaf
510 * The key offset for the implicit back refs is the first byte of
513 * When a file extent is allocated, The implicit back refs is used.
514 * the fields are filled in:
516 * (root_key.objectid, inode objectid, offset in file, 1)
518 * When a file extent is removed file truncation, we find the
519 * corresponding implicit back refs and check the following fields:
521 * (btrfs_header_owner(leaf), inode objectid, offset in file)
523 * Btree extents can be referenced by:
525 * - Different subvolumes
527 * Both the implicit back refs and the full back refs for tree blocks
528 * only consist of key. The key offset for the implicit back refs is
529 * objectid of block's owner tree. The key offset for the full back refs
530 * is the first byte of parent block.
532 * When implicit back refs is used, information about the lowest key and
533 * level of the tree block are required. These information are stored in
534 * tree block info structure.
537 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
538 static int convert_extent_item_v0(struct btrfs_trans_handle *trans,
539 struct btrfs_root *root,
540 struct btrfs_path *path,
541 u64 owner, u32 extra_size)
543 struct btrfs_extent_item *item;
544 struct btrfs_extent_item_v0 *ei0;
545 struct btrfs_extent_ref_v0 *ref0;
546 struct btrfs_tree_block_info *bi;
547 struct extent_buffer *leaf;
548 struct btrfs_key key;
549 struct btrfs_key found_key;
550 u32 new_size = sizeof(*item);
554 leaf = path->nodes[0];
555 BUG_ON(btrfs_item_size_nr(leaf, path->slots[0]) != sizeof(*ei0));
557 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
558 ei0 = btrfs_item_ptr(leaf, path->slots[0],
559 struct btrfs_extent_item_v0);
560 refs = btrfs_extent_refs_v0(leaf, ei0);
562 if (owner == (u64)-1) {
564 if (path->slots[0] >= btrfs_header_nritems(leaf)) {
565 ret = btrfs_next_leaf(root, path);
569 leaf = path->nodes[0];
571 btrfs_item_key_to_cpu(leaf, &found_key,
573 BUG_ON(key.objectid != found_key.objectid);
574 if (found_key.type != BTRFS_EXTENT_REF_V0_KEY) {
578 ref0 = btrfs_item_ptr(leaf, path->slots[0],
579 struct btrfs_extent_ref_v0);
580 owner = btrfs_ref_objectid_v0(leaf, ref0);
584 btrfs_release_path(root, path);
586 if (owner < BTRFS_FIRST_FREE_OBJECTID)
587 new_size += sizeof(*bi);
589 new_size -= sizeof(*ei0);
590 ret = btrfs_search_slot(trans, root, &key, path,
591 new_size + extra_size, 1);
596 ret = btrfs_extend_item(trans, root, path, new_size);
599 leaf = path->nodes[0];
600 item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
601 btrfs_set_extent_refs(leaf, item, refs);
602 /* FIXME: get real generation */
603 btrfs_set_extent_generation(leaf, item, 0);
604 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
605 btrfs_set_extent_flags(leaf, item,
606 BTRFS_EXTENT_FLAG_TREE_BLOCK |
607 BTRFS_BLOCK_FLAG_FULL_BACKREF);
608 bi = (struct btrfs_tree_block_info *)(item + 1);
609 /* FIXME: get first key of the block */
610 memset_extent_buffer(leaf, 0, (unsigned long)bi, sizeof(*bi));
611 btrfs_set_tree_block_level(leaf, bi, (int)owner);
613 btrfs_set_extent_flags(leaf, item, BTRFS_EXTENT_FLAG_DATA);
615 btrfs_mark_buffer_dirty(leaf);
620 static u64 hash_extent_data_ref(u64 root_objectid, u64 owner, u64 offset)
622 u32 high_crc = ~(u32)0;
623 u32 low_crc = ~(u32)0;
626 lenum = cpu_to_le64(root_objectid);
627 high_crc = crc32c(high_crc, &lenum, sizeof(lenum));
628 lenum = cpu_to_le64(owner);
629 low_crc = crc32c(low_crc, &lenum, sizeof(lenum));
630 lenum = cpu_to_le64(offset);
631 low_crc = crc32c(low_crc, &lenum, sizeof(lenum));
633 return ((u64)high_crc << 31) ^ (u64)low_crc;
636 static u64 hash_extent_data_ref_item(struct extent_buffer *leaf,
637 struct btrfs_extent_data_ref *ref)
639 return hash_extent_data_ref(btrfs_extent_data_ref_root(leaf, ref),
640 btrfs_extent_data_ref_objectid(leaf, ref),
641 btrfs_extent_data_ref_offset(leaf, ref));
644 static int match_extent_data_ref(struct extent_buffer *leaf,
645 struct btrfs_extent_data_ref *ref,
646 u64 root_objectid, u64 owner, u64 offset)
648 if (btrfs_extent_data_ref_root(leaf, ref) != root_objectid ||
649 btrfs_extent_data_ref_objectid(leaf, ref) != owner ||
650 btrfs_extent_data_ref_offset(leaf, ref) != offset)
655 static noinline int lookup_extent_data_ref(struct btrfs_trans_handle *trans,
656 struct btrfs_root *root,
657 struct btrfs_path *path,
658 u64 bytenr, u64 parent,
660 u64 owner, u64 offset)
662 struct btrfs_key key;
663 struct btrfs_extent_data_ref *ref;
664 struct extent_buffer *leaf;
670 key.objectid = bytenr;
672 key.type = BTRFS_SHARED_DATA_REF_KEY;
675 key.type = BTRFS_EXTENT_DATA_REF_KEY;
676 key.offset = hash_extent_data_ref(root_objectid,
681 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
690 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
691 key.type = BTRFS_EXTENT_REF_V0_KEY;
692 btrfs_release_path(root, path);
693 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
704 leaf = path->nodes[0];
705 nritems = btrfs_header_nritems(leaf);
707 if (path->slots[0] >= nritems) {
708 ret = btrfs_next_leaf(root, path);
714 leaf = path->nodes[0];
715 nritems = btrfs_header_nritems(leaf);
719 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
720 if (key.objectid != bytenr ||
721 key.type != BTRFS_EXTENT_DATA_REF_KEY)
724 ref = btrfs_item_ptr(leaf, path->slots[0],
725 struct btrfs_extent_data_ref);
727 if (match_extent_data_ref(leaf, ref, root_objectid,
730 btrfs_release_path(root, path);
742 static noinline int insert_extent_data_ref(struct btrfs_trans_handle *trans,
743 struct btrfs_root *root,
744 struct btrfs_path *path,
745 u64 bytenr, u64 parent,
746 u64 root_objectid, u64 owner,
747 u64 offset, int refs_to_add)
749 struct btrfs_key key;
750 struct extent_buffer *leaf;
755 key.objectid = bytenr;
757 key.type = BTRFS_SHARED_DATA_REF_KEY;
759 size = sizeof(struct btrfs_shared_data_ref);
761 key.type = BTRFS_EXTENT_DATA_REF_KEY;
762 key.offset = hash_extent_data_ref(root_objectid,
764 size = sizeof(struct btrfs_extent_data_ref);
767 ret = btrfs_insert_empty_item(trans, root, path, &key, size);
768 if (ret && ret != -EEXIST)
771 leaf = path->nodes[0];
773 struct btrfs_shared_data_ref *ref;
774 ref = btrfs_item_ptr(leaf, path->slots[0],
775 struct btrfs_shared_data_ref);
777 btrfs_set_shared_data_ref_count(leaf, ref, refs_to_add);
779 num_refs = btrfs_shared_data_ref_count(leaf, ref);
780 num_refs += refs_to_add;
781 btrfs_set_shared_data_ref_count(leaf, ref, num_refs);
784 struct btrfs_extent_data_ref *ref;
785 while (ret == -EEXIST) {
786 ref = btrfs_item_ptr(leaf, path->slots[0],
787 struct btrfs_extent_data_ref);
788 if (match_extent_data_ref(leaf, ref, root_objectid,
791 btrfs_release_path(root, path);
793 ret = btrfs_insert_empty_item(trans, root, path, &key,
795 if (ret && ret != -EEXIST)
798 leaf = path->nodes[0];
800 ref = btrfs_item_ptr(leaf, path->slots[0],
801 struct btrfs_extent_data_ref);
803 btrfs_set_extent_data_ref_root(leaf, ref,
805 btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
806 btrfs_set_extent_data_ref_offset(leaf, ref, offset);
807 btrfs_set_extent_data_ref_count(leaf, ref, refs_to_add);
809 num_refs = btrfs_extent_data_ref_count(leaf, ref);
810 num_refs += refs_to_add;
811 btrfs_set_extent_data_ref_count(leaf, ref, num_refs);
814 btrfs_mark_buffer_dirty(leaf);
817 btrfs_release_path(root, path);
821 static noinline int remove_extent_data_ref(struct btrfs_trans_handle *trans,
822 struct btrfs_root *root,
823 struct btrfs_path *path,
826 struct btrfs_key key;
827 struct btrfs_extent_data_ref *ref1 = NULL;
828 struct btrfs_shared_data_ref *ref2 = NULL;
829 struct extent_buffer *leaf;
833 leaf = path->nodes[0];
834 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
836 if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
837 ref1 = btrfs_item_ptr(leaf, path->slots[0],
838 struct btrfs_extent_data_ref);
839 num_refs = btrfs_extent_data_ref_count(leaf, ref1);
840 } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
841 ref2 = btrfs_item_ptr(leaf, path->slots[0],
842 struct btrfs_shared_data_ref);
843 num_refs = btrfs_shared_data_ref_count(leaf, ref2);
844 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
845 } else if (key.type == BTRFS_EXTENT_REF_V0_KEY) {
846 struct btrfs_extent_ref_v0 *ref0;
847 ref0 = btrfs_item_ptr(leaf, path->slots[0],
848 struct btrfs_extent_ref_v0);
849 num_refs = btrfs_ref_count_v0(leaf, ref0);
855 BUG_ON(num_refs < refs_to_drop);
856 num_refs -= refs_to_drop;
859 ret = btrfs_del_item(trans, root, path);
861 if (key.type == BTRFS_EXTENT_DATA_REF_KEY)
862 btrfs_set_extent_data_ref_count(leaf, ref1, num_refs);
863 else if (key.type == BTRFS_SHARED_DATA_REF_KEY)
864 btrfs_set_shared_data_ref_count(leaf, ref2, num_refs);
865 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
867 struct btrfs_extent_ref_v0 *ref0;
868 ref0 = btrfs_item_ptr(leaf, path->slots[0],
869 struct btrfs_extent_ref_v0);
870 btrfs_set_ref_count_v0(leaf, ref0, num_refs);
873 btrfs_mark_buffer_dirty(leaf);
878 static noinline u32 extent_data_ref_count(struct btrfs_root *root,
879 struct btrfs_path *path,
880 struct btrfs_extent_inline_ref *iref)
882 struct btrfs_key key;
883 struct extent_buffer *leaf;
884 struct btrfs_extent_data_ref *ref1;
885 struct btrfs_shared_data_ref *ref2;
888 leaf = path->nodes[0];
889 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
891 if (btrfs_extent_inline_ref_type(leaf, iref) ==
892 BTRFS_EXTENT_DATA_REF_KEY) {
893 ref1 = (struct btrfs_extent_data_ref *)(&iref->offset);
894 num_refs = btrfs_extent_data_ref_count(leaf, ref1);
896 ref2 = (struct btrfs_shared_data_ref *)(iref + 1);
897 num_refs = btrfs_shared_data_ref_count(leaf, ref2);
899 } else if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
900 ref1 = btrfs_item_ptr(leaf, path->slots[0],
901 struct btrfs_extent_data_ref);
902 num_refs = btrfs_extent_data_ref_count(leaf, ref1);
903 } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
904 ref2 = btrfs_item_ptr(leaf, path->slots[0],
905 struct btrfs_shared_data_ref);
906 num_refs = btrfs_shared_data_ref_count(leaf, ref2);
907 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
908 } else if (key.type == BTRFS_EXTENT_REF_V0_KEY) {
909 struct btrfs_extent_ref_v0 *ref0;
910 ref0 = btrfs_item_ptr(leaf, path->slots[0],
911 struct btrfs_extent_ref_v0);
912 num_refs = btrfs_ref_count_v0(leaf, ref0);
920 static noinline int lookup_tree_block_ref(struct btrfs_trans_handle *trans,
921 struct btrfs_root *root,
922 struct btrfs_path *path,
923 u64 bytenr, u64 parent,
926 struct btrfs_key key;
929 key.objectid = bytenr;
931 key.type = BTRFS_SHARED_BLOCK_REF_KEY;
934 key.type = BTRFS_TREE_BLOCK_REF_KEY;
935 key.offset = root_objectid;
938 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
941 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
942 if (ret == -ENOENT && parent) {
943 btrfs_release_path(root, path);
944 key.type = BTRFS_EXTENT_REF_V0_KEY;
945 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
953 static noinline int insert_tree_block_ref(struct btrfs_trans_handle *trans,
954 struct btrfs_root *root,
955 struct btrfs_path *path,
956 u64 bytenr, u64 parent,
959 struct btrfs_key key;
962 key.objectid = bytenr;
964 key.type = BTRFS_SHARED_BLOCK_REF_KEY;
967 key.type = BTRFS_TREE_BLOCK_REF_KEY;
968 key.offset = root_objectid;
971 ret = btrfs_insert_empty_item(trans, root, path, &key, 0);
972 btrfs_release_path(root, path);
976 static inline int extent_ref_type(u64 parent, u64 owner)
979 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
981 type = BTRFS_SHARED_BLOCK_REF_KEY;
983 type = BTRFS_TREE_BLOCK_REF_KEY;
986 type = BTRFS_SHARED_DATA_REF_KEY;
988 type = BTRFS_EXTENT_DATA_REF_KEY;
993 static int find_next_key(struct btrfs_path *path, int level,
994 struct btrfs_key *key)
997 for (; level < BTRFS_MAX_LEVEL; level++) {
998 if (!path->nodes[level])
1000 if (path->slots[level] + 1 >=
1001 btrfs_header_nritems(path->nodes[level]))
1004 btrfs_item_key_to_cpu(path->nodes[level], key,
1005 path->slots[level] + 1);
1007 btrfs_node_key_to_cpu(path->nodes[level], key,
1008 path->slots[level] + 1);
1015 * look for inline back ref. if back ref is found, *ref_ret is set
1016 * to the address of inline back ref, and 0 is returned.
1018 * if back ref isn't found, *ref_ret is set to the address where it
1019 * should be inserted, and -ENOENT is returned.
1021 * if insert is true and there are too many inline back refs, the path
1022 * points to the extent item, and -EAGAIN is returned.
1024 * NOTE: inline back refs are ordered in the same way that back ref
1025 * items in the tree are ordered.
1027 static noinline_for_stack
1028 int lookup_inline_extent_backref(struct btrfs_trans_handle *trans,
1029 struct btrfs_root *root,
1030 struct btrfs_path *path,
1031 struct btrfs_extent_inline_ref **ref_ret,
1032 u64 bytenr, u64 num_bytes,
1033 u64 parent, u64 root_objectid,
1034 u64 owner, u64 offset, int insert)
1036 struct btrfs_key key;
1037 struct extent_buffer *leaf;
1038 struct btrfs_extent_item *ei;
1039 struct btrfs_extent_inline_ref *iref;
1050 key.objectid = bytenr;
1051 key.type = BTRFS_EXTENT_ITEM_KEY;
1052 key.offset = num_bytes;
1054 want = extent_ref_type(parent, owner);
1056 extra_size = btrfs_extent_inline_ref_size(want);
1057 path->keep_locks = 1;
1060 ret = btrfs_search_slot(trans, root, &key, path, extra_size, 1);
1067 leaf = path->nodes[0];
1068 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1069 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1070 if (item_size < sizeof(*ei)) {
1075 ret = convert_extent_item_v0(trans, root, path, owner,
1081 leaf = path->nodes[0];
1082 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1085 BUG_ON(item_size < sizeof(*ei));
1087 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1088 flags = btrfs_extent_flags(leaf, ei);
1090 ptr = (unsigned long)(ei + 1);
1091 end = (unsigned long)ei + item_size;
1093 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
1094 ptr += sizeof(struct btrfs_tree_block_info);
1097 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
1106 iref = (struct btrfs_extent_inline_ref *)ptr;
1107 type = btrfs_extent_inline_ref_type(leaf, iref);
1111 ptr += btrfs_extent_inline_ref_size(type);
1115 if (type == BTRFS_EXTENT_DATA_REF_KEY) {
1116 struct btrfs_extent_data_ref *dref;
1117 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1118 if (match_extent_data_ref(leaf, dref, root_objectid,
1123 if (hash_extent_data_ref_item(leaf, dref) <
1124 hash_extent_data_ref(root_objectid, owner, offset))
1128 ref_offset = btrfs_extent_inline_ref_offset(leaf, iref);
1130 if (parent == ref_offset) {
1134 if (ref_offset < parent)
1137 if (root_objectid == ref_offset) {
1141 if (ref_offset < root_objectid)
1145 ptr += btrfs_extent_inline_ref_size(type);
1147 if (err == -ENOENT && insert) {
1148 if (item_size + extra_size >=
1149 BTRFS_MAX_EXTENT_ITEM_SIZE(root)) {
1154 * To add new inline back ref, we have to make sure
1155 * there is no corresponding back ref item.
1156 * For simplicity, we just do not add new inline back
1157 * ref if there is any kind of item for this block
1159 if (find_next_key(path, 0, &key) == 0 &&
1160 key.objectid == bytenr &&
1161 key.type < BTRFS_BLOCK_GROUP_ITEM_KEY) {
1166 *ref_ret = (struct btrfs_extent_inline_ref *)ptr;
1169 path->keep_locks = 0;
1170 btrfs_unlock_up_safe(path, 1);
1176 * helper to add new inline back ref
1178 static noinline_for_stack
1179 int setup_inline_extent_backref(struct btrfs_trans_handle *trans,
1180 struct btrfs_root *root,
1181 struct btrfs_path *path,
1182 struct btrfs_extent_inline_ref *iref,
1183 u64 parent, u64 root_objectid,
1184 u64 owner, u64 offset, int refs_to_add,
1185 struct btrfs_delayed_extent_op *extent_op)
1187 struct extent_buffer *leaf;
1188 struct btrfs_extent_item *ei;
1191 unsigned long item_offset;
1197 leaf = path->nodes[0];
1198 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1199 item_offset = (unsigned long)iref - (unsigned long)ei;
1201 type = extent_ref_type(parent, owner);
1202 size = btrfs_extent_inline_ref_size(type);
1204 ret = btrfs_extend_item(trans, root, path, size);
1207 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1208 refs = btrfs_extent_refs(leaf, ei);
1209 refs += refs_to_add;
1210 btrfs_set_extent_refs(leaf, ei, refs);
1212 __run_delayed_extent_op(extent_op, leaf, ei);
1214 ptr = (unsigned long)ei + item_offset;
1215 end = (unsigned long)ei + btrfs_item_size_nr(leaf, path->slots[0]);
1216 if (ptr < end - size)
1217 memmove_extent_buffer(leaf, ptr + size, ptr,
1220 iref = (struct btrfs_extent_inline_ref *)ptr;
1221 btrfs_set_extent_inline_ref_type(leaf, iref, type);
1222 if (type == BTRFS_EXTENT_DATA_REF_KEY) {
1223 struct btrfs_extent_data_ref *dref;
1224 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1225 btrfs_set_extent_data_ref_root(leaf, dref, root_objectid);
1226 btrfs_set_extent_data_ref_objectid(leaf, dref, owner);
1227 btrfs_set_extent_data_ref_offset(leaf, dref, offset);
1228 btrfs_set_extent_data_ref_count(leaf, dref, refs_to_add);
1229 } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
1230 struct btrfs_shared_data_ref *sref;
1231 sref = (struct btrfs_shared_data_ref *)(iref + 1);
1232 btrfs_set_shared_data_ref_count(leaf, sref, refs_to_add);
1233 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
1234 } else if (type == BTRFS_SHARED_BLOCK_REF_KEY) {
1235 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
1237 btrfs_set_extent_inline_ref_offset(leaf, iref, root_objectid);
1239 btrfs_mark_buffer_dirty(leaf);
1243 static int lookup_extent_backref(struct btrfs_trans_handle *trans,
1244 struct btrfs_root *root,
1245 struct btrfs_path *path,
1246 struct btrfs_extent_inline_ref **ref_ret,
1247 u64 bytenr, u64 num_bytes, u64 parent,
1248 u64 root_objectid, u64 owner, u64 offset)
1252 ret = lookup_inline_extent_backref(trans, root, path, ref_ret,
1253 bytenr, num_bytes, parent,
1254 root_objectid, owner, offset, 0);
1258 btrfs_release_path(root, path);
1261 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1262 ret = lookup_tree_block_ref(trans, root, path, bytenr, parent,
1265 ret = lookup_extent_data_ref(trans, root, path, bytenr, parent,
1266 root_objectid, owner, offset);
1272 * helper to update/remove inline back ref
1274 static noinline_for_stack
1275 int update_inline_extent_backref(struct btrfs_trans_handle *trans,
1276 struct btrfs_root *root,
1277 struct btrfs_path *path,
1278 struct btrfs_extent_inline_ref *iref,
1280 struct btrfs_delayed_extent_op *extent_op)
1282 struct extent_buffer *leaf;
1283 struct btrfs_extent_item *ei;
1284 struct btrfs_extent_data_ref *dref = NULL;
1285 struct btrfs_shared_data_ref *sref = NULL;
1294 leaf = path->nodes[0];
1295 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1296 refs = btrfs_extent_refs(leaf, ei);
1297 WARN_ON(refs_to_mod < 0 && refs + refs_to_mod <= 0);
1298 refs += refs_to_mod;
1299 btrfs_set_extent_refs(leaf, ei, refs);
1301 __run_delayed_extent_op(extent_op, leaf, ei);
1303 type = btrfs_extent_inline_ref_type(leaf, iref);
1305 if (type == BTRFS_EXTENT_DATA_REF_KEY) {
1306 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1307 refs = btrfs_extent_data_ref_count(leaf, dref);
1308 } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
1309 sref = (struct btrfs_shared_data_ref *)(iref + 1);
1310 refs = btrfs_shared_data_ref_count(leaf, sref);
1313 BUG_ON(refs_to_mod != -1);
1316 BUG_ON(refs_to_mod < 0 && refs < -refs_to_mod);
1317 refs += refs_to_mod;
1320 if (type == BTRFS_EXTENT_DATA_REF_KEY)
1321 btrfs_set_extent_data_ref_count(leaf, dref, refs);
1323 btrfs_set_shared_data_ref_count(leaf, sref, refs);
1325 size = btrfs_extent_inline_ref_size(type);
1326 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1327 ptr = (unsigned long)iref;
1328 end = (unsigned long)ei + item_size;
1329 if (ptr + size < end)
1330 memmove_extent_buffer(leaf, ptr, ptr + size,
1333 ret = btrfs_truncate_item(trans, root, path, item_size, 1);
1336 btrfs_mark_buffer_dirty(leaf);
1340 static noinline_for_stack
1341 int insert_inline_extent_backref(struct btrfs_trans_handle *trans,
1342 struct btrfs_root *root,
1343 struct btrfs_path *path,
1344 u64 bytenr, u64 num_bytes, u64 parent,
1345 u64 root_objectid, u64 owner,
1346 u64 offset, int refs_to_add,
1347 struct btrfs_delayed_extent_op *extent_op)
1349 struct btrfs_extent_inline_ref *iref;
1352 ret = lookup_inline_extent_backref(trans, root, path, &iref,
1353 bytenr, num_bytes, parent,
1354 root_objectid, owner, offset, 1);
1356 BUG_ON(owner < BTRFS_FIRST_FREE_OBJECTID);
1357 ret = update_inline_extent_backref(trans, root, path, iref,
1358 refs_to_add, extent_op);
1359 } else if (ret == -ENOENT) {
1360 ret = setup_inline_extent_backref(trans, root, path, iref,
1361 parent, root_objectid,
1362 owner, offset, refs_to_add,
1368 static int insert_extent_backref(struct btrfs_trans_handle *trans,
1369 struct btrfs_root *root,
1370 struct btrfs_path *path,
1371 u64 bytenr, u64 parent, u64 root_objectid,
1372 u64 owner, u64 offset, int refs_to_add)
1375 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1376 BUG_ON(refs_to_add != 1);
1377 ret = insert_tree_block_ref(trans, root, path, bytenr,
1378 parent, root_objectid);
1380 ret = insert_extent_data_ref(trans, root, path, bytenr,
1381 parent, root_objectid,
1382 owner, offset, refs_to_add);
1387 static int remove_extent_backref(struct btrfs_trans_handle *trans,
1388 struct btrfs_root *root,
1389 struct btrfs_path *path,
1390 struct btrfs_extent_inline_ref *iref,
1391 int refs_to_drop, int is_data)
1395 BUG_ON(!is_data && refs_to_drop != 1);
1397 ret = update_inline_extent_backref(trans, root, path, iref,
1398 -refs_to_drop, NULL);
1399 } else if (is_data) {
1400 ret = remove_extent_data_ref(trans, root, path, refs_to_drop);
1402 ret = btrfs_del_item(trans, root, path);
1407 #ifdef BIO_RW_DISCARD
1408 static void btrfs_issue_discard(struct block_device *bdev,
1411 blkdev_issue_discard(bdev, start >> 9, len >> 9, GFP_KERNEL);
1415 static int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr,
1418 #ifdef BIO_RW_DISCARD
1420 u64 map_length = num_bytes;
1421 struct btrfs_multi_bio *multi = NULL;
1423 /* Tell the block device(s) that the sectors can be discarded */
1424 ret = btrfs_map_block(&root->fs_info->mapping_tree, READ,
1425 bytenr, &map_length, &multi, 0);
1427 struct btrfs_bio_stripe *stripe = multi->stripes;
1430 if (map_length > num_bytes)
1431 map_length = num_bytes;
1433 for (i = 0; i < multi->num_stripes; i++, stripe++) {
1434 btrfs_issue_discard(stripe->dev->bdev,
1447 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
1448 struct btrfs_root *root,
1449 u64 bytenr, u64 num_bytes, u64 parent,
1450 u64 root_objectid, u64 owner, u64 offset)
1453 BUG_ON(owner < BTRFS_FIRST_FREE_OBJECTID &&
1454 root_objectid == BTRFS_TREE_LOG_OBJECTID);
1456 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1457 ret = btrfs_add_delayed_tree_ref(trans, bytenr, num_bytes,
1458 parent, root_objectid, (int)owner,
1459 BTRFS_ADD_DELAYED_REF, NULL);
1461 ret = btrfs_add_delayed_data_ref(trans, bytenr, num_bytes,
1462 parent, root_objectid, owner, offset,
1463 BTRFS_ADD_DELAYED_REF, NULL);
1468 static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
1469 struct btrfs_root *root,
1470 u64 bytenr, u64 num_bytes,
1471 u64 parent, u64 root_objectid,
1472 u64 owner, u64 offset, int refs_to_add,
1473 struct btrfs_delayed_extent_op *extent_op)
1475 struct btrfs_path *path;
1476 struct extent_buffer *leaf;
1477 struct btrfs_extent_item *item;
1482 path = btrfs_alloc_path();
1487 path->leave_spinning = 1;
1488 /* this will setup the path even if it fails to insert the back ref */
1489 ret = insert_inline_extent_backref(trans, root->fs_info->extent_root,
1490 path, bytenr, num_bytes, parent,
1491 root_objectid, owner, offset,
1492 refs_to_add, extent_op);
1496 if (ret != -EAGAIN) {
1501 leaf = path->nodes[0];
1502 item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1503 refs = btrfs_extent_refs(leaf, item);
1504 btrfs_set_extent_refs(leaf, item, refs + refs_to_add);
1506 __run_delayed_extent_op(extent_op, leaf, item);
1508 btrfs_mark_buffer_dirty(leaf);
1509 btrfs_release_path(root->fs_info->extent_root, path);
1512 path->leave_spinning = 1;
1514 /* now insert the actual backref */
1515 ret = insert_extent_backref(trans, root->fs_info->extent_root,
1516 path, bytenr, parent, root_objectid,
1517 owner, offset, refs_to_add);
1520 btrfs_free_path(path);
1524 static int run_delayed_data_ref(struct btrfs_trans_handle *trans,
1525 struct btrfs_root *root,
1526 struct btrfs_delayed_ref_node *node,
1527 struct btrfs_delayed_extent_op *extent_op,
1528 int insert_reserved)
1531 struct btrfs_delayed_data_ref *ref;
1532 struct btrfs_key ins;
1537 ins.objectid = node->bytenr;
1538 ins.offset = node->num_bytes;
1539 ins.type = BTRFS_EXTENT_ITEM_KEY;
1541 ref = btrfs_delayed_node_to_data_ref(node);
1542 if (node->type == BTRFS_SHARED_DATA_REF_KEY)
1543 parent = ref->parent;
1545 ref_root = ref->root;
1547 if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
1549 BUG_ON(extent_op->update_key);
1550 flags |= extent_op->flags_to_set;
1552 ret = alloc_reserved_file_extent(trans, root,
1553 parent, ref_root, flags,
1554 ref->objectid, ref->offset,
1555 &ins, node->ref_mod);
1556 update_reserved_extents(root, ins.objectid, ins.offset, 0);
1557 } else if (node->action == BTRFS_ADD_DELAYED_REF) {
1558 ret = __btrfs_inc_extent_ref(trans, root, node->bytenr,
1559 node->num_bytes, parent,
1560 ref_root, ref->objectid,
1561 ref->offset, node->ref_mod,
1563 } else if (node->action == BTRFS_DROP_DELAYED_REF) {
1564 ret = __btrfs_free_extent(trans, root, node->bytenr,
1565 node->num_bytes, parent,
1566 ref_root, ref->objectid,
1567 ref->offset, node->ref_mod,
1575 static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
1576 struct extent_buffer *leaf,
1577 struct btrfs_extent_item *ei)
1579 u64 flags = btrfs_extent_flags(leaf, ei);
1580 if (extent_op->update_flags) {
1581 flags |= extent_op->flags_to_set;
1582 btrfs_set_extent_flags(leaf, ei, flags);
1585 if (extent_op->update_key) {
1586 struct btrfs_tree_block_info *bi;
1587 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK));
1588 bi = (struct btrfs_tree_block_info *)(ei + 1);
1589 btrfs_set_tree_block_key(leaf, bi, &extent_op->key);
1593 static int run_delayed_extent_op(struct btrfs_trans_handle *trans,
1594 struct btrfs_root *root,
1595 struct btrfs_delayed_ref_node *node,
1596 struct btrfs_delayed_extent_op *extent_op)
1598 struct btrfs_key key;
1599 struct btrfs_path *path;
1600 struct btrfs_extent_item *ei;
1601 struct extent_buffer *leaf;
1606 path = btrfs_alloc_path();
1610 key.objectid = node->bytenr;
1611 key.type = BTRFS_EXTENT_ITEM_KEY;
1612 key.offset = node->num_bytes;
1615 path->leave_spinning = 1;
1616 ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key,
1627 leaf = path->nodes[0];
1628 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1629 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1630 if (item_size < sizeof(*ei)) {
1631 ret = convert_extent_item_v0(trans, root->fs_info->extent_root,
1637 leaf = path->nodes[0];
1638 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1641 BUG_ON(item_size < sizeof(*ei));
1642 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1643 __run_delayed_extent_op(extent_op, leaf, ei);
1645 btrfs_mark_buffer_dirty(leaf);
1647 btrfs_free_path(path);
1651 static int run_delayed_tree_ref(struct btrfs_trans_handle *trans,
1652 struct btrfs_root *root,
1653 struct btrfs_delayed_ref_node *node,
1654 struct btrfs_delayed_extent_op *extent_op,
1655 int insert_reserved)
1658 struct btrfs_delayed_tree_ref *ref;
1659 struct btrfs_key ins;
1663 ins.objectid = node->bytenr;
1664 ins.offset = node->num_bytes;
1665 ins.type = BTRFS_EXTENT_ITEM_KEY;
1667 ref = btrfs_delayed_node_to_tree_ref(node);
1668 if (node->type == BTRFS_SHARED_BLOCK_REF_KEY)
1669 parent = ref->parent;
1671 ref_root = ref->root;
1673 BUG_ON(node->ref_mod != 1);
1674 if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
1675 BUG_ON(!extent_op || !extent_op->update_flags ||
1676 !extent_op->update_key);
1677 ret = alloc_reserved_tree_block(trans, root,
1679 extent_op->flags_to_set,
1682 update_reserved_extents(root, ins.objectid, ins.offset, 0);
1683 } else if (node->action == BTRFS_ADD_DELAYED_REF) {
1684 ret = __btrfs_inc_extent_ref(trans, root, node->bytenr,
1685 node->num_bytes, parent, ref_root,
1686 ref->level, 0, 1, extent_op);
1687 } else if (node->action == BTRFS_DROP_DELAYED_REF) {
1688 ret = __btrfs_free_extent(trans, root, node->bytenr,
1689 node->num_bytes, parent, ref_root,
1690 ref->level, 0, 1, extent_op);
1698 /* helper function to actually process a single delayed ref entry */
1699 static int run_one_delayed_ref(struct btrfs_trans_handle *trans,
1700 struct btrfs_root *root,
1701 struct btrfs_delayed_ref_node *node,
1702 struct btrfs_delayed_extent_op *extent_op,
1703 int insert_reserved)
1706 if (btrfs_delayed_ref_is_head(node)) {
1707 struct btrfs_delayed_ref_head *head;
1709 * we've hit the end of the chain and we were supposed
1710 * to insert this extent into the tree. But, it got
1711 * deleted before we ever needed to insert it, so all
1712 * we have to do is clean up the accounting
1715 head = btrfs_delayed_node_to_head(node);
1716 if (insert_reserved) {
1717 if (head->is_data) {
1718 ret = btrfs_del_csums(trans, root,
1723 btrfs_update_pinned_extents(root, node->bytenr,
1724 node->num_bytes, 1);
1725 update_reserved_extents(root, node->bytenr,
1726 node->num_bytes, 0);
1728 mutex_unlock(&head->mutex);
1732 if (node->type == BTRFS_TREE_BLOCK_REF_KEY ||
1733 node->type == BTRFS_SHARED_BLOCK_REF_KEY)
1734 ret = run_delayed_tree_ref(trans, root, node, extent_op,
1736 else if (node->type == BTRFS_EXTENT_DATA_REF_KEY ||
1737 node->type == BTRFS_SHARED_DATA_REF_KEY)
1738 ret = run_delayed_data_ref(trans, root, node, extent_op,
1745 static noinline struct btrfs_delayed_ref_node *
1746 select_delayed_ref(struct btrfs_delayed_ref_head *head)
1748 struct rb_node *node;
1749 struct btrfs_delayed_ref_node *ref;
1750 int action = BTRFS_ADD_DELAYED_REF;
1753 * select delayed ref of type BTRFS_ADD_DELAYED_REF first.
1754 * this prevents ref count from going down to zero when
1755 * there still are pending delayed ref.
1757 node = rb_prev(&head->node.rb_node);
1761 ref = rb_entry(node, struct btrfs_delayed_ref_node,
1763 if (ref->bytenr != head->node.bytenr)
1765 if (ref->action == action)
1767 node = rb_prev(node);
1769 if (action == BTRFS_ADD_DELAYED_REF) {
1770 action = BTRFS_DROP_DELAYED_REF;
1776 static noinline int run_clustered_refs(struct btrfs_trans_handle *trans,
1777 struct btrfs_root *root,
1778 struct list_head *cluster)
1780 struct btrfs_delayed_ref_root *delayed_refs;
1781 struct btrfs_delayed_ref_node *ref;
1782 struct btrfs_delayed_ref_head *locked_ref = NULL;
1783 struct btrfs_delayed_extent_op *extent_op;
1786 int must_insert_reserved = 0;
1788 delayed_refs = &trans->transaction->delayed_refs;
1791 /* pick a new head ref from the cluster list */
1792 if (list_empty(cluster))
1795 locked_ref = list_entry(cluster->next,
1796 struct btrfs_delayed_ref_head, cluster);
1798 /* grab the lock that says we are going to process
1799 * all the refs for this head */
1800 ret = btrfs_delayed_ref_lock(trans, locked_ref);
1803 * we may have dropped the spin lock to get the head
1804 * mutex lock, and that might have given someone else
1805 * time to free the head. If that's true, it has been
1806 * removed from our list and we can move on.
1808 if (ret == -EAGAIN) {
1816 * record the must insert reserved flag before we
1817 * drop the spin lock.
1819 must_insert_reserved = locked_ref->must_insert_reserved;
1820 locked_ref->must_insert_reserved = 0;
1822 extent_op = locked_ref->extent_op;
1823 locked_ref->extent_op = NULL;
1826 * locked_ref is the head node, so we have to go one
1827 * node back for any delayed ref updates
1829 ref = select_delayed_ref(locked_ref);
1831 /* All delayed refs have been processed, Go ahead
1832 * and send the head node to run_one_delayed_ref,
1833 * so that any accounting fixes can happen
1835 ref = &locked_ref->node;
1837 if (extent_op && must_insert_reserved) {
1843 spin_unlock(&delayed_refs->lock);
1845 ret = run_delayed_extent_op(trans, root,
1851 spin_lock(&delayed_refs->lock);
1855 list_del_init(&locked_ref->cluster);
1860 rb_erase(&ref->rb_node, &delayed_refs->root);
1861 delayed_refs->num_entries--;
1863 spin_unlock(&delayed_refs->lock);
1865 ret = run_one_delayed_ref(trans, root, ref, extent_op,
1866 must_insert_reserved);
1869 btrfs_put_delayed_ref(ref);
1874 spin_lock(&delayed_refs->lock);
1880 * this starts processing the delayed reference count updates and
1881 * extent insertions we have queued up so far. count can be
1882 * 0, which means to process everything in the tree at the start
1883 * of the run (but not newly added entries), or it can be some target
1884 * number you'd like to process.
1886 int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
1887 struct btrfs_root *root, unsigned long count)
1889 struct rb_node *node;
1890 struct btrfs_delayed_ref_root *delayed_refs;
1891 struct btrfs_delayed_ref_node *ref;
1892 struct list_head cluster;
1894 int run_all = count == (unsigned long)-1;
1897 if (root == root->fs_info->extent_root)
1898 root = root->fs_info->tree_root;
1900 delayed_refs = &trans->transaction->delayed_refs;
1901 INIT_LIST_HEAD(&cluster);
1903 spin_lock(&delayed_refs->lock);
1905 count = delayed_refs->num_entries * 2;
1909 if (!(run_all || run_most) &&
1910 delayed_refs->num_heads_ready < 64)
1914 * go find something we can process in the rbtree. We start at
1915 * the beginning of the tree, and then build a cluster
1916 * of refs to process starting at the first one we are able to
1919 ret = btrfs_find_ref_cluster(trans, &cluster,
1920 delayed_refs->run_delayed_start);
1924 ret = run_clustered_refs(trans, root, &cluster);
1927 count -= min_t(unsigned long, ret, count);
1934 node = rb_first(&delayed_refs->root);
1937 count = (unsigned long)-1;
1940 ref = rb_entry(node, struct btrfs_delayed_ref_node,
1942 if (btrfs_delayed_ref_is_head(ref)) {
1943 struct btrfs_delayed_ref_head *head;
1945 head = btrfs_delayed_node_to_head(ref);
1946 atomic_inc(&ref->refs);
1948 spin_unlock(&delayed_refs->lock);
1949 mutex_lock(&head->mutex);
1950 mutex_unlock(&head->mutex);
1952 btrfs_put_delayed_ref(ref);
1956 node = rb_next(node);
1958 spin_unlock(&delayed_refs->lock);
1959 schedule_timeout(1);
1963 spin_unlock(&delayed_refs->lock);
1967 int btrfs_set_disk_extent_flags(struct btrfs_trans_handle *trans,
1968 struct btrfs_root *root,
1969 u64 bytenr, u64 num_bytes, u64 flags,
1972 struct btrfs_delayed_extent_op *extent_op;
1975 extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
1979 extent_op->flags_to_set = flags;
1980 extent_op->update_flags = 1;
1981 extent_op->update_key = 0;
1982 extent_op->is_data = is_data ? 1 : 0;
1984 ret = btrfs_add_delayed_extent_op(trans, bytenr, num_bytes, extent_op);
1990 static noinline int check_delayed_ref(struct btrfs_trans_handle *trans,
1991 struct btrfs_root *root,
1992 struct btrfs_path *path,
1993 u64 objectid, u64 offset, u64 bytenr)
1995 struct btrfs_delayed_ref_head *head;
1996 struct btrfs_delayed_ref_node *ref;
1997 struct btrfs_delayed_data_ref *data_ref;
1998 struct btrfs_delayed_ref_root *delayed_refs;
1999 struct rb_node *node;
2003 delayed_refs = &trans->transaction->delayed_refs;
2004 spin_lock(&delayed_refs->lock);
2005 head = btrfs_find_delayed_ref_head(trans, bytenr);
2009 if (!mutex_trylock(&head->mutex)) {
2010 atomic_inc(&head->node.refs);
2011 spin_unlock(&delayed_refs->lock);
2013 btrfs_release_path(root->fs_info->extent_root, path);
2015 mutex_lock(&head->mutex);
2016 mutex_unlock(&head->mutex);
2017 btrfs_put_delayed_ref(&head->node);
2021 node = rb_prev(&head->node.rb_node);
2025 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
2027 if (ref->bytenr != bytenr)
2031 if (ref->type != BTRFS_EXTENT_DATA_REF_KEY)
2034 data_ref = btrfs_delayed_node_to_data_ref(ref);
2036 node = rb_prev(node);
2038 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
2039 if (ref->bytenr == bytenr)
2043 if (data_ref->root != root->root_key.objectid ||
2044 data_ref->objectid != objectid || data_ref->offset != offset)
2049 mutex_unlock(&head->mutex);
2051 spin_unlock(&delayed_refs->lock);
2055 static noinline int check_committed_ref(struct btrfs_trans_handle *trans,
2056 struct btrfs_root *root,
2057 struct btrfs_path *path,
2058 u64 objectid, u64 offset, u64 bytenr)
2060 struct btrfs_root *extent_root = root->fs_info->extent_root;
2061 struct extent_buffer *leaf;
2062 struct btrfs_extent_data_ref *ref;
2063 struct btrfs_extent_inline_ref *iref;
2064 struct btrfs_extent_item *ei;
2065 struct btrfs_key key;
2069 key.objectid = bytenr;
2070 key.offset = (u64)-1;
2071 key.type = BTRFS_EXTENT_ITEM_KEY;
2073 ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
2079 if (path->slots[0] == 0)
2083 leaf = path->nodes[0];
2084 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
2086 if (key.objectid != bytenr || key.type != BTRFS_EXTENT_ITEM_KEY)
2090 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
2091 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
2092 if (item_size < sizeof(*ei)) {
2093 WARN_ON(item_size != sizeof(struct btrfs_extent_item_v0));
2097 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
2099 if (item_size != sizeof(*ei) +
2100 btrfs_extent_inline_ref_size(BTRFS_EXTENT_DATA_REF_KEY))
2103 if (btrfs_extent_generation(leaf, ei) <=
2104 btrfs_root_last_snapshot(&root->root_item))
2107 iref = (struct btrfs_extent_inline_ref *)(ei + 1);
2108 if (btrfs_extent_inline_ref_type(leaf, iref) !=
2109 BTRFS_EXTENT_DATA_REF_KEY)
2112 ref = (struct btrfs_extent_data_ref *)(&iref->offset);
2113 if (btrfs_extent_refs(leaf, ei) !=
2114 btrfs_extent_data_ref_count(leaf, ref) ||
2115 btrfs_extent_data_ref_root(leaf, ref) !=
2116 root->root_key.objectid ||
2117 btrfs_extent_data_ref_objectid(leaf, ref) != objectid ||
2118 btrfs_extent_data_ref_offset(leaf, ref) != offset)
2126 int btrfs_cross_ref_exist(struct btrfs_trans_handle *trans,
2127 struct btrfs_root *root,
2128 u64 objectid, u64 offset, u64 bytenr)
2130 struct btrfs_path *path;
2134 path = btrfs_alloc_path();
2139 ret = check_committed_ref(trans, root, path, objectid,
2141 if (ret && ret != -ENOENT)
2144 ret2 = check_delayed_ref(trans, root, path, objectid,
2146 } while (ret2 == -EAGAIN);
2148 if (ret2 && ret2 != -ENOENT) {
2153 if (ret != -ENOENT || ret2 != -ENOENT)
2156 btrfs_free_path(path);
2161 int btrfs_cache_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2162 struct extent_buffer *buf, u32 nr_extents)
2164 struct btrfs_key key;
2165 struct btrfs_file_extent_item *fi;
2173 if (!root->ref_cows)
2176 if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
2178 root_gen = root->root_key.offset;
2181 root_gen = trans->transid - 1;
2184 level = btrfs_header_level(buf);
2185 nritems = btrfs_header_nritems(buf);
2188 struct btrfs_leaf_ref *ref;
2189 struct btrfs_extent_info *info;
2191 ref = btrfs_alloc_leaf_ref(root, nr_extents);
2197 ref->root_gen = root_gen;
2198 ref->bytenr = buf->start;
2199 ref->owner = btrfs_header_owner(buf);
2200 ref->generation = btrfs_header_generation(buf);
2201 ref->nritems = nr_extents;
2202 info = ref->extents;
2204 for (i = 0; nr_extents > 0 && i < nritems; i++) {
2206 btrfs_item_key_to_cpu(buf, &key, i);
2207 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
2209 fi = btrfs_item_ptr(buf, i,
2210 struct btrfs_file_extent_item);
2211 if (btrfs_file_extent_type(buf, fi) ==
2212 BTRFS_FILE_EXTENT_INLINE)
2214 disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
2215 if (disk_bytenr == 0)
2218 info->bytenr = disk_bytenr;
2220 btrfs_file_extent_disk_num_bytes(buf, fi);
2221 info->objectid = key.objectid;
2222 info->offset = key.offset;
2226 ret = btrfs_add_leaf_ref(root, ref, shared);
2227 if (ret == -EEXIST && shared) {
2228 struct btrfs_leaf_ref *old;
2229 old = btrfs_lookup_leaf_ref(root, ref->bytenr);
2231 btrfs_remove_leaf_ref(root, old);
2232 btrfs_free_leaf_ref(root, old);
2233 ret = btrfs_add_leaf_ref(root, ref, shared);
2236 btrfs_free_leaf_ref(root, ref);
2242 /* when a block goes through cow, we update the reference counts of
2243 * everything that block points to. The internal pointers of the block
2244 * can be in just about any order, and it is likely to have clusters of
2245 * things that are close together and clusters of things that are not.
2247 * To help reduce the seeks that come with updating all of these reference
2248 * counts, sort them by byte number before actual updates are done.
2250 * struct refsort is used to match byte number to slot in the btree block.
2251 * we sort based on the byte number and then use the slot to actually
2254 * struct refsort is smaller than strcut btrfs_item and smaller than
2255 * struct btrfs_key_ptr. Since we're currently limited to the page size
2256 * for a btree block, there's no way for a kmalloc of refsorts for a
2257 * single node to be bigger than a page.
2265 * for passing into sort()
2267 static int refsort_cmp(const void *a_void, const void *b_void)
2269 const struct refsort *a = a_void;
2270 const struct refsort *b = b_void;
2272 if (a->bytenr < b->bytenr)
2274 if (a->bytenr > b->bytenr)
2280 static int __btrfs_mod_ref(struct btrfs_trans_handle *trans,
2281 struct btrfs_root *root,
2282 struct extent_buffer *buf,
2283 int full_backref, int inc)
2290 struct btrfs_key key;
2291 struct btrfs_file_extent_item *fi;
2295 int (*process_func)(struct btrfs_trans_handle *, struct btrfs_root *,
2296 u64, u64, u64, u64, u64, u64);
2298 ref_root = btrfs_header_owner(buf);
2299 nritems = btrfs_header_nritems(buf);
2300 level = btrfs_header_level(buf);
2302 if (!root->ref_cows && level == 0)
2306 process_func = btrfs_inc_extent_ref;
2308 process_func = btrfs_free_extent;
2311 parent = buf->start;
2315 for (i = 0; i < nritems; i++) {
2317 btrfs_item_key_to_cpu(buf, &key, i);
2318 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
2320 fi = btrfs_item_ptr(buf, i,
2321 struct btrfs_file_extent_item);
2322 if (btrfs_file_extent_type(buf, fi) ==
2323 BTRFS_FILE_EXTENT_INLINE)
2325 bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
2329 num_bytes = btrfs_file_extent_disk_num_bytes(buf, fi);
2330 key.offset -= btrfs_file_extent_offset(buf, fi);
2331 ret = process_func(trans, root, bytenr, num_bytes,
2332 parent, ref_root, key.objectid,
2337 bytenr = btrfs_node_blockptr(buf, i);
2338 num_bytes = btrfs_level_size(root, level - 1);
2339 ret = process_func(trans, root, bytenr, num_bytes,
2340 parent, ref_root, level - 1, 0);
2351 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2352 struct extent_buffer *buf, int full_backref)
2354 return __btrfs_mod_ref(trans, root, buf, full_backref, 1);
2357 int btrfs_dec_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2358 struct extent_buffer *buf, int full_backref)
2360 return __btrfs_mod_ref(trans, root, buf, full_backref, 0);
2363 static int write_one_cache_group(struct btrfs_trans_handle *trans,
2364 struct btrfs_root *root,
2365 struct btrfs_path *path,
2366 struct btrfs_block_group_cache *cache)
2369 struct btrfs_root *extent_root = root->fs_info->extent_root;
2371 struct extent_buffer *leaf;
2373 ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1);
2378 leaf = path->nodes[0];
2379 bi = btrfs_item_ptr_offset(leaf, path->slots[0]);
2380 write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item));
2381 btrfs_mark_buffer_dirty(leaf);
2382 btrfs_release_path(extent_root, path);
2390 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
2391 struct btrfs_root *root)
2393 struct btrfs_block_group_cache *cache, *entry;
2397 struct btrfs_path *path;
2400 path = btrfs_alloc_path();
2406 spin_lock(&root->fs_info->block_group_cache_lock);
2407 for (n = rb_first(&root->fs_info->block_group_cache_tree);
2408 n; n = rb_next(n)) {
2409 entry = rb_entry(n, struct btrfs_block_group_cache,
2416 spin_unlock(&root->fs_info->block_group_cache_lock);
2422 last += cache->key.offset;
2424 err = write_one_cache_group(trans, root,
2427 * if we fail to write the cache group, we want
2428 * to keep it marked dirty in hopes that a later
2436 btrfs_free_path(path);
2440 int btrfs_extent_readonly(struct btrfs_root *root, u64 bytenr)
2442 struct btrfs_block_group_cache *block_group;
2445 block_group = btrfs_lookup_block_group(root->fs_info, bytenr);
2446 if (!block_group || block_group->ro)
2449 btrfs_put_block_group(block_group);
2453 static int update_space_info(struct btrfs_fs_info *info, u64 flags,
2454 u64 total_bytes, u64 bytes_used,
2455 struct btrfs_space_info **space_info)
2457 struct btrfs_space_info *found;
2459 found = __find_space_info(info, flags);
2461 spin_lock(&found->lock);
2462 found->total_bytes += total_bytes;
2463 found->bytes_used += bytes_used;
2465 spin_unlock(&found->lock);
2466 *space_info = found;
2469 found = kzalloc(sizeof(*found), GFP_NOFS);
2473 INIT_LIST_HEAD(&found->block_groups);
2474 init_rwsem(&found->groups_sem);
2475 spin_lock_init(&found->lock);
2476 found->flags = flags;
2477 found->total_bytes = total_bytes;
2478 found->bytes_used = bytes_used;
2479 found->bytes_pinned = 0;
2480 found->bytes_reserved = 0;
2481 found->bytes_readonly = 0;
2482 found->bytes_delalloc = 0;
2484 found->force_alloc = 0;
2485 *space_info = found;
2486 list_add_rcu(&found->list, &info->space_info);
2490 static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags)
2492 u64 extra_flags = flags & (BTRFS_BLOCK_GROUP_RAID0 |
2493 BTRFS_BLOCK_GROUP_RAID1 |
2494 BTRFS_BLOCK_GROUP_RAID10 |
2495 BTRFS_BLOCK_GROUP_DUP);
2497 if (flags & BTRFS_BLOCK_GROUP_DATA)
2498 fs_info->avail_data_alloc_bits |= extra_flags;
2499 if (flags & BTRFS_BLOCK_GROUP_METADATA)
2500 fs_info->avail_metadata_alloc_bits |= extra_flags;
2501 if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
2502 fs_info->avail_system_alloc_bits |= extra_flags;
2506 static void set_block_group_readonly(struct btrfs_block_group_cache *cache)
2508 spin_lock(&cache->space_info->lock);
2509 spin_lock(&cache->lock);
2511 cache->space_info->bytes_readonly += cache->key.offset -
2512 btrfs_block_group_used(&cache->item);
2515 spin_unlock(&cache->lock);
2516 spin_unlock(&cache->space_info->lock);
2519 u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags)
2521 u64 num_devices = root->fs_info->fs_devices->rw_devices;
2523 if (num_devices == 1)
2524 flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0);
2525 if (num_devices < 4)
2526 flags &= ~BTRFS_BLOCK_GROUP_RAID10;
2528 if ((flags & BTRFS_BLOCK_GROUP_DUP) &&
2529 (flags & (BTRFS_BLOCK_GROUP_RAID1 |
2530 BTRFS_BLOCK_GROUP_RAID10))) {
2531 flags &= ~BTRFS_BLOCK_GROUP_DUP;
2534 if ((flags & BTRFS_BLOCK_GROUP_RAID1) &&
2535 (flags & BTRFS_BLOCK_GROUP_RAID10)) {
2536 flags &= ~BTRFS_BLOCK_GROUP_RAID1;
2539 if ((flags & BTRFS_BLOCK_GROUP_RAID0) &&
2540 ((flags & BTRFS_BLOCK_GROUP_RAID1) |
2541 (flags & BTRFS_BLOCK_GROUP_RAID10) |
2542 (flags & BTRFS_BLOCK_GROUP_DUP)))
2543 flags &= ~BTRFS_BLOCK_GROUP_RAID0;
2547 static u64 btrfs_get_alloc_profile(struct btrfs_root *root, u64 data)
2549 struct btrfs_fs_info *info = root->fs_info;
2553 alloc_profile = info->avail_data_alloc_bits &
2554 info->data_alloc_profile;
2555 data = BTRFS_BLOCK_GROUP_DATA | alloc_profile;
2556 } else if (root == root->fs_info->chunk_root) {
2557 alloc_profile = info->avail_system_alloc_bits &
2558 info->system_alloc_profile;
2559 data = BTRFS_BLOCK_GROUP_SYSTEM | alloc_profile;
2561 alloc_profile = info->avail_metadata_alloc_bits &
2562 info->metadata_alloc_profile;
2563 data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile;
2566 return btrfs_reduce_alloc_profile(root, data);
2569 void btrfs_set_inode_space_info(struct btrfs_root *root, struct inode *inode)
2573 alloc_target = btrfs_get_alloc_profile(root, 1);
2574 BTRFS_I(inode)->space_info = __find_space_info(root->fs_info,
2579 * for now this just makes sure we have at least 5% of our metadata space free
2582 int btrfs_check_metadata_free_space(struct btrfs_root *root)
2584 struct btrfs_fs_info *info = root->fs_info;
2585 struct btrfs_space_info *meta_sinfo;
2586 u64 alloc_target, thresh;
2587 int committed = 0, ret;
2589 /* get the space info for where the metadata will live */
2590 alloc_target = btrfs_get_alloc_profile(root, 0);
2591 meta_sinfo = __find_space_info(info, alloc_target);
2594 spin_lock(&meta_sinfo->lock);
2595 if (!meta_sinfo->full)
2596 thresh = meta_sinfo->total_bytes * 80;
2598 thresh = meta_sinfo->total_bytes * 95;
2600 do_div(thresh, 100);
2602 if (meta_sinfo->bytes_used + meta_sinfo->bytes_reserved +
2603 meta_sinfo->bytes_pinned + meta_sinfo->bytes_readonly > thresh) {
2604 struct btrfs_trans_handle *trans;
2605 if (!meta_sinfo->full) {
2606 meta_sinfo->force_alloc = 1;
2607 spin_unlock(&meta_sinfo->lock);
2609 trans = btrfs_start_transaction(root, 1);
2613 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
2614 2 * 1024 * 1024, alloc_target, 0);
2615 btrfs_end_transaction(trans, root);
2618 spin_unlock(&meta_sinfo->lock);
2622 trans = btrfs_join_transaction(root, 1);
2625 ret = btrfs_commit_transaction(trans, root);
2632 spin_unlock(&meta_sinfo->lock);
2638 * This will check the space that the inode allocates from to make sure we have
2639 * enough space for bytes.
2641 int btrfs_check_data_free_space(struct btrfs_root *root, struct inode *inode,
2644 struct btrfs_space_info *data_sinfo;
2645 int ret = 0, committed = 0;
2647 /* make sure bytes are sectorsize aligned */
2648 bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
2650 data_sinfo = BTRFS_I(inode)->space_info;
2652 /* make sure we have enough space to handle the data first */
2653 spin_lock(&data_sinfo->lock);
2654 if (data_sinfo->total_bytes - data_sinfo->bytes_used -
2655 data_sinfo->bytes_delalloc - data_sinfo->bytes_reserved -
2656 data_sinfo->bytes_pinned - data_sinfo->bytes_readonly -
2657 data_sinfo->bytes_may_use < bytes) {
2658 struct btrfs_trans_handle *trans;
2661 * if we don't have enough free bytes in this space then we need
2662 * to alloc a new chunk.
2664 if (!data_sinfo->full) {
2667 data_sinfo->force_alloc = 1;
2668 spin_unlock(&data_sinfo->lock);
2670 alloc_target = btrfs_get_alloc_profile(root, 1);
2671 trans = btrfs_start_transaction(root, 1);
2675 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
2676 bytes + 2 * 1024 * 1024,
2678 btrfs_end_transaction(trans, root);
2683 spin_unlock(&data_sinfo->lock);
2685 /* commit the current transaction and try again */
2688 trans = btrfs_join_transaction(root, 1);
2691 ret = btrfs_commit_transaction(trans, root);
2697 printk(KERN_ERR "no space left, need %llu, %llu delalloc bytes"
2698 ", %llu bytes_used, %llu bytes_reserved, "
2699 "%llu bytes_pinned, %llu bytes_readonly, %llu may use "
2700 "%llu total\n", (unsigned long long)bytes,
2701 (unsigned long long)data_sinfo->bytes_delalloc,
2702 (unsigned long long)data_sinfo->bytes_used,
2703 (unsigned long long)data_sinfo->bytes_reserved,
2704 (unsigned long long)data_sinfo->bytes_pinned,
2705 (unsigned long long)data_sinfo->bytes_readonly,
2706 (unsigned long long)data_sinfo->bytes_may_use,
2707 (unsigned long long)data_sinfo->total_bytes);
2710 data_sinfo->bytes_may_use += bytes;
2711 BTRFS_I(inode)->reserved_bytes += bytes;
2712 spin_unlock(&data_sinfo->lock);
2714 return btrfs_check_metadata_free_space(root);
2718 * if there was an error for whatever reason after calling
2719 * btrfs_check_data_free_space, call this so we can cleanup the counters.
2721 void btrfs_free_reserved_data_space(struct btrfs_root *root,
2722 struct inode *inode, u64 bytes)
2724 struct btrfs_space_info *data_sinfo;
2726 /* make sure bytes are sectorsize aligned */
2727 bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
2729 data_sinfo = BTRFS_I(inode)->space_info;
2730 spin_lock(&data_sinfo->lock);
2731 data_sinfo->bytes_may_use -= bytes;
2732 BTRFS_I(inode)->reserved_bytes -= bytes;
2733 spin_unlock(&data_sinfo->lock);
2736 /* called when we are adding a delalloc extent to the inode's io_tree */
2737 void btrfs_delalloc_reserve_space(struct btrfs_root *root, struct inode *inode,
2740 struct btrfs_space_info *data_sinfo;
2742 /* get the space info for where this inode will be storing its data */
2743 data_sinfo = BTRFS_I(inode)->space_info;
2745 /* make sure we have enough space to handle the data first */
2746 spin_lock(&data_sinfo->lock);
2747 data_sinfo->bytes_delalloc += bytes;
2750 * we are adding a delalloc extent without calling
2751 * btrfs_check_data_free_space first. This happens on a weird
2752 * writepage condition, but shouldn't hurt our accounting
2754 if (unlikely(bytes > BTRFS_I(inode)->reserved_bytes)) {
2755 data_sinfo->bytes_may_use -= BTRFS_I(inode)->reserved_bytes;
2756 BTRFS_I(inode)->reserved_bytes = 0;
2758 data_sinfo->bytes_may_use -= bytes;
2759 BTRFS_I(inode)->reserved_bytes -= bytes;
2762 spin_unlock(&data_sinfo->lock);
2765 /* called when we are clearing an delalloc extent from the inode's io_tree */
2766 void btrfs_delalloc_free_space(struct btrfs_root *root, struct inode *inode,
2769 struct btrfs_space_info *info;
2771 info = BTRFS_I(inode)->space_info;
2773 spin_lock(&info->lock);
2774 info->bytes_delalloc -= bytes;
2775 spin_unlock(&info->lock);
2778 static void force_metadata_allocation(struct btrfs_fs_info *info)
2780 struct list_head *head = &info->space_info;
2781 struct btrfs_space_info *found;
2784 list_for_each_entry_rcu(found, head, list) {
2785 if (found->flags & BTRFS_BLOCK_GROUP_METADATA)
2786 found->force_alloc = 1;
2791 static int do_chunk_alloc(struct btrfs_trans_handle *trans,
2792 struct btrfs_root *extent_root, u64 alloc_bytes,
2793 u64 flags, int force)
2795 struct btrfs_space_info *space_info;
2796 struct btrfs_fs_info *fs_info = extent_root->fs_info;
2800 mutex_lock(&fs_info->chunk_mutex);
2802 flags = btrfs_reduce_alloc_profile(extent_root, flags);
2804 space_info = __find_space_info(extent_root->fs_info, flags);
2806 ret = update_space_info(extent_root->fs_info, flags,
2810 BUG_ON(!space_info);
2812 spin_lock(&space_info->lock);
2813 if (space_info->force_alloc) {
2815 space_info->force_alloc = 0;
2817 if (space_info->full) {
2818 spin_unlock(&space_info->lock);
2822 thresh = space_info->total_bytes - space_info->bytes_readonly;
2823 thresh = div_factor(thresh, 6);
2825 (space_info->bytes_used + space_info->bytes_pinned +
2826 space_info->bytes_reserved + alloc_bytes) < thresh) {
2827 spin_unlock(&space_info->lock);
2830 spin_unlock(&space_info->lock);
2833 * if we're doing a data chunk, go ahead and make sure that
2834 * we keep a reasonable number of metadata chunks allocated in the
2837 if (flags & BTRFS_BLOCK_GROUP_DATA) {
2838 fs_info->data_chunk_allocations++;
2839 if (!(fs_info->data_chunk_allocations %
2840 fs_info->metadata_ratio))
2841 force_metadata_allocation(fs_info);
2844 ret = btrfs_alloc_chunk(trans, extent_root, flags);
2846 space_info->full = 1;
2848 mutex_unlock(&extent_root->fs_info->chunk_mutex);
2852 static int update_block_group(struct btrfs_trans_handle *trans,
2853 struct btrfs_root *root,
2854 u64 bytenr, u64 num_bytes, int alloc,
2857 struct btrfs_block_group_cache *cache;
2858 struct btrfs_fs_info *info = root->fs_info;
2859 u64 total = num_bytes;
2863 /* block accounting for super block */
2864 spin_lock(&info->delalloc_lock);
2865 old_val = btrfs_super_bytes_used(&info->super_copy);
2867 old_val += num_bytes;
2869 old_val -= num_bytes;
2870 btrfs_set_super_bytes_used(&info->super_copy, old_val);
2872 /* block accounting for root item */
2873 old_val = btrfs_root_used(&root->root_item);
2875 old_val += num_bytes;
2877 old_val -= num_bytes;
2878 btrfs_set_root_used(&root->root_item, old_val);
2879 spin_unlock(&info->delalloc_lock);
2882 cache = btrfs_lookup_block_group(info, bytenr);
2885 byte_in_group = bytenr - cache->key.objectid;
2886 WARN_ON(byte_in_group > cache->key.offset);
2888 spin_lock(&cache->space_info->lock);
2889 spin_lock(&cache->lock);
2891 old_val = btrfs_block_group_used(&cache->item);
2892 num_bytes = min(total, cache->key.offset - byte_in_group);
2894 old_val += num_bytes;
2895 cache->space_info->bytes_used += num_bytes;
2897 cache->space_info->bytes_readonly -= num_bytes;
2898 btrfs_set_block_group_used(&cache->item, old_val);
2899 spin_unlock(&cache->lock);
2900 spin_unlock(&cache->space_info->lock);
2902 old_val -= num_bytes;
2903 cache->space_info->bytes_used -= num_bytes;
2905 cache->space_info->bytes_readonly += num_bytes;
2906 btrfs_set_block_group_used(&cache->item, old_val);
2907 spin_unlock(&cache->lock);
2908 spin_unlock(&cache->space_info->lock);
2912 ret = btrfs_discard_extent(root, bytenr,
2916 ret = btrfs_add_free_space(cache, bytenr,
2921 btrfs_put_block_group(cache);
2923 bytenr += num_bytes;
2928 static u64 first_logical_byte(struct btrfs_root *root, u64 search_start)
2930 struct btrfs_block_group_cache *cache;
2933 cache = btrfs_lookup_first_block_group(root->fs_info, search_start);
2937 bytenr = cache->key.objectid;
2938 btrfs_put_block_group(cache);
2943 int btrfs_update_pinned_extents(struct btrfs_root *root,
2944 u64 bytenr, u64 num, int pin)
2947 struct btrfs_block_group_cache *cache;
2948 struct btrfs_fs_info *fs_info = root->fs_info;
2951 set_extent_dirty(&fs_info->pinned_extents,
2952 bytenr, bytenr + num - 1, GFP_NOFS);
2954 clear_extent_dirty(&fs_info->pinned_extents,
2955 bytenr, bytenr + num - 1, GFP_NOFS);
2959 cache = btrfs_lookup_block_group(fs_info, bytenr);
2961 len = min(num, cache->key.offset -
2962 (bytenr - cache->key.objectid));
2964 spin_lock(&cache->space_info->lock);
2965 spin_lock(&cache->lock);
2966 cache->pinned += len;
2967 cache->space_info->bytes_pinned += len;
2968 spin_unlock(&cache->lock);
2969 spin_unlock(&cache->space_info->lock);
2970 fs_info->total_pinned += len;
2972 spin_lock(&cache->space_info->lock);
2973 spin_lock(&cache->lock);
2974 cache->pinned -= len;
2975 cache->space_info->bytes_pinned -= len;
2976 spin_unlock(&cache->lock);
2977 spin_unlock(&cache->space_info->lock);
2978 fs_info->total_pinned -= len;
2980 btrfs_add_free_space(cache, bytenr, len);
2982 btrfs_put_block_group(cache);
2989 static int update_reserved_extents(struct btrfs_root *root,
2990 u64 bytenr, u64 num, int reserve)
2993 struct btrfs_block_group_cache *cache;
2994 struct btrfs_fs_info *fs_info = root->fs_info;
2997 cache = btrfs_lookup_block_group(fs_info, bytenr);
2999 len = min(num, cache->key.offset -
3000 (bytenr - cache->key.objectid));
3002 spin_lock(&cache->space_info->lock);
3003 spin_lock(&cache->lock);
3005 cache->reserved += len;
3006 cache->space_info->bytes_reserved += len;
3008 cache->reserved -= len;
3009 cache->space_info->bytes_reserved -= len;
3011 spin_unlock(&cache->lock);
3012 spin_unlock(&cache->space_info->lock);
3013 btrfs_put_block_group(cache);
3020 int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy)
3025 struct extent_io_tree *pinned_extents = &root->fs_info->pinned_extents;
3029 ret = find_first_extent_bit(pinned_extents, last,
3030 &start, &end, EXTENT_DIRTY);
3033 set_extent_dirty(copy, start, end, GFP_NOFS);
3039 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
3040 struct btrfs_root *root,
3041 struct extent_io_tree *unpin)
3048 ret = find_first_extent_bit(unpin, 0, &start, &end,
3053 ret = btrfs_discard_extent(root, start, end + 1 - start);
3055 /* unlocks the pinned mutex */
3056 btrfs_update_pinned_extents(root, start, end + 1 - start, 0);
3057 clear_extent_dirty(unpin, start, end, GFP_NOFS);
3064 static int pin_down_bytes(struct btrfs_trans_handle *trans,
3065 struct btrfs_root *root,
3066 struct btrfs_path *path,
3067 u64 bytenr, u64 num_bytes, int is_data,
3068 struct extent_buffer **must_clean)
3071 struct extent_buffer *buf;
3076 buf = btrfs_find_tree_block(root, bytenr, num_bytes);
3080 /* we can reuse a block if it hasn't been written
3081 * and it is from this transaction. We can't
3082 * reuse anything from the tree log root because
3083 * it has tiny sub-transactions.
3085 if (btrfs_buffer_uptodate(buf, 0) &&
3086 btrfs_try_tree_lock(buf)) {
3087 u64 header_owner = btrfs_header_owner(buf);
3088 u64 header_transid = btrfs_header_generation(buf);
3089 if (header_owner != BTRFS_TREE_LOG_OBJECTID &&
3090 header_transid == trans->transid &&
3091 !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
3095 btrfs_tree_unlock(buf);
3097 free_extent_buffer(buf);
3099 btrfs_set_path_blocking(path);
3100 /* unlocks the pinned mutex */
3101 btrfs_update_pinned_extents(root, bytenr, num_bytes, 1);
3108 static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
3109 struct btrfs_root *root,
3110 u64 bytenr, u64 num_bytes, u64 parent,
3111 u64 root_objectid, u64 owner_objectid,
3112 u64 owner_offset, int refs_to_drop,
3113 struct btrfs_delayed_extent_op *extent_op)
3115 struct btrfs_key key;
3116 struct btrfs_path *path;
3117 struct btrfs_fs_info *info = root->fs_info;
3118 struct btrfs_root *extent_root = info->extent_root;
3119 struct extent_buffer *leaf;
3120 struct btrfs_extent_item *ei;
3121 struct btrfs_extent_inline_ref *iref;
3124 int extent_slot = 0;
3125 int found_extent = 0;
3130 path = btrfs_alloc_path();
3135 path->leave_spinning = 1;
3137 is_data = owner_objectid >= BTRFS_FIRST_FREE_OBJECTID;
3138 BUG_ON(!is_data && refs_to_drop != 1);
3140 ret = lookup_extent_backref(trans, extent_root, path, &iref,
3141 bytenr, num_bytes, parent,
3142 root_objectid, owner_objectid,
3145 extent_slot = path->slots[0];
3146 while (extent_slot >= 0) {
3147 btrfs_item_key_to_cpu(path->nodes[0], &key,
3149 if (key.objectid != bytenr)
3151 if (key.type == BTRFS_EXTENT_ITEM_KEY &&
3152 key.offset == num_bytes) {
3156 if (path->slots[0] - extent_slot > 5)
3160 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
3161 item_size = btrfs_item_size_nr(path->nodes[0], extent_slot);
3162 if (found_extent && item_size < sizeof(*ei))
3165 if (!found_extent) {
3167 ret = remove_extent_backref(trans, extent_root, path,
3171 btrfs_release_path(extent_root, path);
3172 path->leave_spinning = 1;
3174 key.objectid = bytenr;
3175 key.type = BTRFS_EXTENT_ITEM_KEY;
3176 key.offset = num_bytes;
3178 ret = btrfs_search_slot(trans, extent_root,
3181 printk(KERN_ERR "umm, got %d back from search"
3182 ", was looking for %llu\n", ret,
3183 (unsigned long long)bytenr);
3184 btrfs_print_leaf(extent_root, path->nodes[0]);
3187 extent_slot = path->slots[0];
3190 btrfs_print_leaf(extent_root, path->nodes[0]);
3192 printk(KERN_ERR "btrfs unable to find ref byte nr %llu "
3193 "parent %llu root %llu owner %llu offset %llu\n",
3194 (unsigned long long)bytenr,
3195 (unsigned long long)parent,
3196 (unsigned long long)root_objectid,
3197 (unsigned long long)owner_objectid,
3198 (unsigned long long)owner_offset);
3201 leaf = path->nodes[0];
3202 item_size = btrfs_item_size_nr(leaf, extent_slot);
3203 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
3204 if (item_size < sizeof(*ei)) {
3205 BUG_ON(found_extent || extent_slot != path->slots[0]);
3206 ret = convert_extent_item_v0(trans, extent_root, path,
3210 btrfs_release_path(extent_root, path);
3211 path->leave_spinning = 1;
3213 key.objectid = bytenr;
3214 key.type = BTRFS_EXTENT_ITEM_KEY;
3215 key.offset = num_bytes;
3217 ret = btrfs_search_slot(trans, extent_root, &key, path,
3220 printk(KERN_ERR "umm, got %d back from search"
3221 ", was looking for %llu\n", ret,
3222 (unsigned long long)bytenr);
3223 btrfs_print_leaf(extent_root, path->nodes[0]);
3226 extent_slot = path->slots[0];
3227 leaf = path->nodes[0];
3228 item_size = btrfs_item_size_nr(leaf, extent_slot);
3231 BUG_ON(item_size < sizeof(*ei));
3232 ei = btrfs_item_ptr(leaf, extent_slot,
3233 struct btrfs_extent_item);
3234 if (owner_objectid < BTRFS_FIRST_FREE_OBJECTID) {
3235 struct btrfs_tree_block_info *bi;
3236 BUG_ON(item_size < sizeof(*ei) + sizeof(*bi));
3237 bi = (struct btrfs_tree_block_info *)(ei + 1);
3238 WARN_ON(owner_objectid != btrfs_tree_block_level(leaf, bi));
3241 refs = btrfs_extent_refs(leaf, ei);
3242 BUG_ON(refs < refs_to_drop);
3243 refs -= refs_to_drop;
3247 __run_delayed_extent_op(extent_op, leaf, ei);
3249 * In the case of inline back ref, reference count will
3250 * be updated by remove_extent_backref
3253 BUG_ON(!found_extent);
3255 btrfs_set_extent_refs(leaf, ei, refs);
3256 btrfs_mark_buffer_dirty(leaf);
3259 ret = remove_extent_backref(trans, extent_root, path,
3266 struct extent_buffer *must_clean = NULL;
3269 BUG_ON(is_data && refs_to_drop !=
3270 extent_data_ref_count(root, path, iref));
3272 BUG_ON(path->slots[0] != extent_slot);
3274 BUG_ON(path->slots[0] != extent_slot + 1);
3275 path->slots[0] = extent_slot;
3280 ret = pin_down_bytes(trans, root, path, bytenr,
3281 num_bytes, is_data, &must_clean);
3286 * it is going to be very rare for someone to be waiting
3287 * on the block we're freeing. del_items might need to
3288 * schedule, so rather than get fancy, just force it
3292 btrfs_set_lock_blocking(must_clean);
3294 ret = btrfs_del_items(trans, extent_root, path, path->slots[0],
3297 btrfs_release_path(extent_root, path);
3300 clean_tree_block(NULL, root, must_clean);
3301 btrfs_tree_unlock(must_clean);
3302 free_extent_buffer(must_clean);
3306 ret = btrfs_del_csums(trans, root, bytenr, num_bytes);
3309 invalidate_mapping_pages(info->btree_inode->i_mapping,
3310 bytenr >> PAGE_CACHE_SHIFT,
3311 (bytenr + num_bytes - 1) >> PAGE_CACHE_SHIFT);
3314 ret = update_block_group(trans, root, bytenr, num_bytes, 0,
3318 btrfs_free_path(path);
3323 * when we free an extent, it is possible (and likely) that we free the last
3324 * delayed ref for that extent as well. This searches the delayed ref tree for
3325 * a given extent, and if there are no other delayed refs to be processed, it
3326 * removes it from the tree.
3328 static noinline int check_ref_cleanup(struct btrfs_trans_handle *trans,
3329 struct btrfs_root *root, u64 bytenr)
3331 struct btrfs_delayed_ref_head *head;
3332 struct btrfs_delayed_ref_root *delayed_refs;
3333 struct btrfs_delayed_ref_node *ref;
3334 struct rb_node *node;
3337 delayed_refs = &trans->transaction->delayed_refs;
3338 spin_lock(&delayed_refs->lock);
3339 head = btrfs_find_delayed_ref_head(trans, bytenr);
3343 node = rb_prev(&head->node.rb_node);
3347 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
3349 /* there are still entries for this ref, we can't drop it */
3350 if (ref->bytenr == bytenr)
3353 if (head->extent_op) {
3354 if (!head->must_insert_reserved)
3356 kfree(head->extent_op);
3357 head->extent_op = NULL;
3361 * waiting for the lock here would deadlock. If someone else has it
3362 * locked they are already in the process of dropping it anyway
3364 if (!mutex_trylock(&head->mutex))
3368 * at this point we have a head with no other entries. Go
3369 * ahead and process it.
3371 head->node.in_tree = 0;
3372 rb_erase(&head->node.rb_node, &delayed_refs->root);
3374 delayed_refs->num_entries--;
3377 * we don't take a ref on the node because we're removing it from the
3378 * tree, so we just steal the ref the tree was holding.
3380 delayed_refs->num_heads--;
3381 if (list_empty(&head->cluster))
3382 delayed_refs->num_heads_ready--;
3384 list_del_init(&head->cluster);
3385 spin_unlock(&delayed_refs->lock);
3387 ret = run_one_delayed_ref(trans, root->fs_info->tree_root,
3388 &head->node, head->extent_op,
3389 head->must_insert_reserved);
3391 btrfs_put_delayed_ref(&head->node);
3394 spin_unlock(&delayed_refs->lock);
3398 int btrfs_free_extent(struct btrfs_trans_handle *trans,
3399 struct btrfs_root *root,
3400 u64 bytenr, u64 num_bytes, u64 parent,
3401 u64 root_objectid, u64 owner, u64 offset)
3406 * tree log blocks never actually go into the extent allocation
3407 * tree, just update pinning info and exit early.
3409 if (root_objectid == BTRFS_TREE_LOG_OBJECTID) {
3410 WARN_ON(owner >= BTRFS_FIRST_FREE_OBJECTID);
3411 /* unlocks the pinned mutex */
3412 btrfs_update_pinned_extents(root, bytenr, num_bytes, 1);
3413 update_reserved_extents(root, bytenr, num_bytes, 0);
3415 } else if (owner < BTRFS_FIRST_FREE_OBJECTID) {
3416 ret = btrfs_add_delayed_tree_ref(trans, bytenr, num_bytes,
3417 parent, root_objectid, (int)owner,
3418 BTRFS_DROP_DELAYED_REF, NULL);
3420 ret = check_ref_cleanup(trans, root, bytenr);
3423 ret = btrfs_add_delayed_data_ref(trans, bytenr, num_bytes,
3424 parent, root_objectid, owner,
3425 offset, BTRFS_DROP_DELAYED_REF, NULL);
3431 static u64 stripe_align(struct btrfs_root *root, u64 val)
3433 u64 mask = ((u64)root->stripesize - 1);
3434 u64 ret = (val + mask) & ~mask;
3439 * walks the btree of allocated extents and find a hole of a given size.
3440 * The key ins is changed to record the hole:
3441 * ins->objectid == block start
3442 * ins->flags = BTRFS_EXTENT_ITEM_KEY
3443 * ins->offset == number of blocks
3444 * Any available blocks before search_start are skipped.
3446 static noinline int find_free_extent(struct btrfs_trans_handle *trans,
3447 struct btrfs_root *orig_root,
3448 u64 num_bytes, u64 empty_size,
3449 u64 search_start, u64 search_end,
3450 u64 hint_byte, struct btrfs_key *ins,
3451 u64 exclude_start, u64 exclude_nr,
3455 struct btrfs_root *root = orig_root->fs_info->extent_root;
3456 struct btrfs_free_cluster *last_ptr = NULL;
3457 struct btrfs_block_group_cache *block_group = NULL;
3458 int empty_cluster = 2 * 1024 * 1024;
3459 int allowed_chunk_alloc = 0;
3460 struct btrfs_space_info *space_info;
3461 int last_ptr_loop = 0;
3464 WARN_ON(num_bytes < root->sectorsize);
3465 btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
3469 space_info = __find_space_info(root->fs_info, data);
3471 if (orig_root->ref_cows || empty_size)
3472 allowed_chunk_alloc = 1;
3474 if (data & BTRFS_BLOCK_GROUP_METADATA) {
3475 last_ptr = &root->fs_info->meta_alloc_cluster;
3476 if (!btrfs_test_opt(root, SSD))
3477 empty_cluster = 64 * 1024;
3480 if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD)) {
3481 last_ptr = &root->fs_info->data_alloc_cluster;
3485 spin_lock(&last_ptr->lock);
3486 if (last_ptr->block_group)
3487 hint_byte = last_ptr->window_start;
3488 spin_unlock(&last_ptr->lock);
3491 search_start = max(search_start, first_logical_byte(root, 0));
3492 search_start = max(search_start, hint_byte);
3499 if (search_start == hint_byte) {
3500 block_group = btrfs_lookup_block_group(root->fs_info,
3502 if (block_group && block_group_bits(block_group, data)) {
3503 down_read(&space_info->groups_sem);
3504 if (list_empty(&block_group->list) ||
3507 * someone is removing this block group,
3508 * we can't jump into the have_block_group
3509 * target because our list pointers are not
3512 btrfs_put_block_group(block_group);
3513 up_read(&space_info->groups_sem);
3515 goto have_block_group;
3516 } else if (block_group) {
3517 btrfs_put_block_group(block_group);
3522 down_read(&space_info->groups_sem);
3523 list_for_each_entry(block_group, &space_info->block_groups, list) {
3526 atomic_inc(&block_group->count);
3527 search_start = block_group->key.objectid;
3530 if (unlikely(!block_group->cached)) {
3531 mutex_lock(&block_group->cache_mutex);
3532 ret = cache_block_group(root, block_group);
3533 mutex_unlock(&block_group->cache_mutex);
3535 btrfs_put_block_group(block_group);
3540 if (unlikely(block_group->ro))
3545 * the refill lock keeps out other
3546 * people trying to start a new cluster
3548 spin_lock(&last_ptr->refill_lock);
3549 if (last_ptr->block_group &&
3550 (last_ptr->block_group->ro ||
3551 !block_group_bits(last_ptr->block_group, data))) {
3553 goto refill_cluster;
3556 offset = btrfs_alloc_from_cluster(block_group, last_ptr,
3557 num_bytes, search_start);
3559 /* we have a block, we're done */
3560 spin_unlock(&last_ptr->refill_lock);
3564 spin_lock(&last_ptr->lock);
3566 * whoops, this cluster doesn't actually point to
3567 * this block group. Get a ref on the block
3568 * group is does point to and try again
3570 if (!last_ptr_loop && last_ptr->block_group &&
3571 last_ptr->block_group != block_group) {
3573 btrfs_put_block_group(block_group);
3574 block_group = last_ptr->block_group;
3575 atomic_inc(&block_group->count);
3576 spin_unlock(&last_ptr->lock);
3577 spin_unlock(&last_ptr->refill_lock);
3580 search_start = block_group->key.objectid;
3582 * we know this block group is properly
3583 * in the list because
3584 * btrfs_remove_block_group, drops the
3585 * cluster before it removes the block
3586 * group from the list
3588 goto have_block_group;
3590 spin_unlock(&last_ptr->lock);
3593 * this cluster didn't work out, free it and
3596 btrfs_return_cluster_to_free_space(NULL, last_ptr);
3600 /* allocate a cluster in this block group */
3601 ret = btrfs_find_space_cluster(trans, root,
3602 block_group, last_ptr,
3604 empty_cluster + empty_size);
3607 * now pull our allocation out of this
3610 offset = btrfs_alloc_from_cluster(block_group,
3611 last_ptr, num_bytes,
3614 /* we found one, proceed */
3615 spin_unlock(&last_ptr->refill_lock);
3620 * at this point we either didn't find a cluster
3621 * or we weren't able to allocate a block from our
3622 * cluster. Free the cluster we've been trying
3623 * to use, and go to the next block group
3626 btrfs_return_cluster_to_free_space(NULL,
3628 spin_unlock(&last_ptr->refill_lock);
3631 spin_unlock(&last_ptr->refill_lock);
3634 offset = btrfs_find_space_for_alloc(block_group, search_start,
3635 num_bytes, empty_size);
3639 search_start = stripe_align(root, offset);
3641 /* move on to the next group */
3642 if (search_start + num_bytes >= search_end) {
3643 btrfs_add_free_space(block_group, offset, num_bytes);
3647 /* move on to the next group */
3648 if (search_start + num_bytes >
3649 block_group->key.objectid + block_group->key.offset) {
3650 btrfs_add_free_space(block_group, offset, num_bytes);
3654 if (exclude_nr > 0 &&
3655 (search_start + num_bytes > exclude_start &&
3656 search_start < exclude_start + exclude_nr)) {
3657 search_start = exclude_start + exclude_nr;
3659 btrfs_add_free_space(block_group, offset, num_bytes);
3661 * if search_start is still in this block group
3662 * then we just re-search this block group
3664 if (search_start >= block_group->key.objectid &&
3665 search_start < (block_group->key.objectid +
3666 block_group->key.offset))
3667 goto have_block_group;
3671 ins->objectid = search_start;
3672 ins->offset = num_bytes;
3674 if (offset < search_start)
3675 btrfs_add_free_space(block_group, offset,
3676 search_start - offset);
3677 BUG_ON(offset > search_start);
3679 /* we are all good, lets return */
3682 btrfs_put_block_group(block_group);
3684 up_read(&space_info->groups_sem);
3686 /* loop == 0, try to find a clustered alloc in every block group
3687 * loop == 1, try again after forcing a chunk allocation
3688 * loop == 2, set empty_size and empty_cluster to 0 and try again
3690 if (!ins->objectid && loop < 3 &&
3691 (empty_size || empty_cluster || allowed_chunk_alloc)) {
3697 if (allowed_chunk_alloc) {
3698 ret = do_chunk_alloc(trans, root, num_bytes +
3699 2 * 1024 * 1024, data, 1);
3700 allowed_chunk_alloc = 0;
3702 space_info->force_alloc = 1;
3710 } else if (!ins->objectid) {
3714 /* we found what we needed */
3715 if (ins->objectid) {
3716 if (!(data & BTRFS_BLOCK_GROUP_DATA))
3717 trans->block_group = block_group->key.objectid;
3719 btrfs_put_block_group(block_group);
3726 static void dump_space_info(struct btrfs_space_info *info, u64 bytes)
3728 struct btrfs_block_group_cache *cache;
3730 printk(KERN_INFO "space_info has %llu free, is %sfull\n",
3731 (unsigned long long)(info->total_bytes - info->bytes_used -
3732 info->bytes_pinned - info->bytes_reserved),
3733 (info->full) ? "" : "not ");
3734 printk(KERN_INFO "space_info total=%llu, pinned=%llu, delalloc=%llu,"
3735 " may_use=%llu, used=%llu\n",
3736 (unsigned long long)info->total_bytes,
3737 (unsigned long long)info->bytes_pinned,
3738 (unsigned long long)info->bytes_delalloc,
3739 (unsigned long long)info->bytes_may_use,
3740 (unsigned long long)info->bytes_used);
3742 down_read(&info->groups_sem);
3743 list_for_each_entry(cache, &info->block_groups, list) {
3744 spin_lock(&cache->lock);
3745 printk(KERN_INFO "block group %llu has %llu bytes, %llu used "
3746 "%llu pinned %llu reserved\n",
3747 (unsigned long long)cache->key.objectid,
3748 (unsigned long long)cache->key.offset,
3749 (unsigned long long)btrfs_block_group_used(&cache->item),
3750 (unsigned long long)cache->pinned,
3751 (unsigned long long)cache->reserved);
3752 btrfs_dump_free_space(cache, bytes);
3753 spin_unlock(&cache->lock);
3755 up_read(&info->groups_sem);
3758 static int __btrfs_reserve_extent(struct btrfs_trans_handle *trans,
3759 struct btrfs_root *root,
3760 u64 num_bytes, u64 min_alloc_size,
3761 u64 empty_size, u64 hint_byte,
3762 u64 search_end, struct btrfs_key *ins,
3766 u64 search_start = 0;
3767 struct btrfs_fs_info *info = root->fs_info;
3769 data = btrfs_get_alloc_profile(root, data);
3772 * the only place that sets empty_size is btrfs_realloc_node, which
3773 * is not called recursively on allocations
3775 if (empty_size || root->ref_cows) {
3776 if (!(data & BTRFS_BLOCK_GROUP_METADATA)) {
3777 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
3779 BTRFS_BLOCK_GROUP_METADATA |
3780 (info->metadata_alloc_profile &
3781 info->avail_metadata_alloc_bits), 0);
3783 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
3784 num_bytes + 2 * 1024 * 1024, data, 0);
3787 WARN_ON(num_bytes < root->sectorsize);
3788 ret = find_free_extent(trans, root, num_bytes, empty_size,
3789 search_start, search_end, hint_byte, ins,
3790 trans->alloc_exclude_start,
3791 trans->alloc_exclude_nr, data);
3793 if (ret == -ENOSPC && num_bytes > min_alloc_size) {
3794 num_bytes = num_bytes >> 1;
3795 num_bytes = num_bytes & ~(root->sectorsize - 1);
3796 num_bytes = max(num_bytes, min_alloc_size);
3797 do_chunk_alloc(trans, root->fs_info->extent_root,
3798 num_bytes, data, 1);
3802 struct btrfs_space_info *sinfo;
3804 sinfo = __find_space_info(root->fs_info, data);
3805 printk(KERN_ERR "btrfs allocation failed flags %llu, "
3806 "wanted %llu\n", (unsigned long long)data,
3807 (unsigned long long)num_bytes);
3808 dump_space_info(sinfo, num_bytes);
3815 int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len)
3817 struct btrfs_block_group_cache *cache;
3820 cache = btrfs_lookup_block_group(root->fs_info, start);
3822 printk(KERN_ERR "Unable to find block group for %llu\n",
3823 (unsigned long long)start);
3827 ret = btrfs_discard_extent(root, start, len);
3829 btrfs_add_free_space(cache, start, len);
3830 btrfs_put_block_group(cache);
3831 update_reserved_extents(root, start, len, 0);
3836 int btrfs_reserve_extent(struct btrfs_trans_handle *trans,
3837 struct btrfs_root *root,
3838 u64 num_bytes, u64 min_alloc_size,
3839 u64 empty_size, u64 hint_byte,
3840 u64 search_end, struct btrfs_key *ins,
3844 ret = __btrfs_reserve_extent(trans, root, num_bytes, min_alloc_size,
3845 empty_size, hint_byte, search_end, ins,
3847 update_reserved_extents(root, ins->objectid, ins->offset, 1);
3851 static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
3852 struct btrfs_root *root,
3853 u64 parent, u64 root_objectid,
3854 u64 flags, u64 owner, u64 offset,
3855 struct btrfs_key *ins, int ref_mod)
3858 struct btrfs_fs_info *fs_info = root->fs_info;
3859 struct btrfs_extent_item *extent_item;
3860 struct btrfs_extent_inline_ref *iref;
3861 struct btrfs_path *path;
3862 struct extent_buffer *leaf;
3867 type = BTRFS_SHARED_DATA_REF_KEY;
3869 type = BTRFS_EXTENT_DATA_REF_KEY;
3871 size = sizeof(*extent_item) + btrfs_extent_inline_ref_size(type);
3873 path = btrfs_alloc_path();
3876 path->leave_spinning = 1;
3877 ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
3881 leaf = path->nodes[0];
3882 extent_item = btrfs_item_ptr(leaf, path->slots[0],
3883 struct btrfs_extent_item);
3884 btrfs_set_extent_refs(leaf, extent_item, ref_mod);
3885 btrfs_set_extent_generation(leaf, extent_item, trans->transid);
3886 btrfs_set_extent_flags(leaf, extent_item,
3887 flags | BTRFS_EXTENT_FLAG_DATA);
3889 iref = (struct btrfs_extent_inline_ref *)(extent_item + 1);
3890 btrfs_set_extent_inline_ref_type(leaf, iref, type);
3892 struct btrfs_shared_data_ref *ref;
3893 ref = (struct btrfs_shared_data_ref *)(iref + 1);
3894 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
3895 btrfs_set_shared_data_ref_count(leaf, ref, ref_mod);
3897 struct btrfs_extent_data_ref *ref;
3898 ref = (struct btrfs_extent_data_ref *)(&iref->offset);
3899 btrfs_set_extent_data_ref_root(leaf, ref, root_objectid);
3900 btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
3901 btrfs_set_extent_data_ref_offset(leaf, ref, offset);
3902 btrfs_set_extent_data_ref_count(leaf, ref, ref_mod);
3905 btrfs_mark_buffer_dirty(path->nodes[0]);
3906 btrfs_free_path(path);
3908 ret = update_block_group(trans, root, ins->objectid, ins->offset,
3911 printk(KERN_ERR "btrfs update block group failed for %llu "
3912 "%llu\n", (unsigned long long)ins->objectid,
3913 (unsigned long long)ins->offset);
3919 static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
3920 struct btrfs_root *root,
3921 u64 parent, u64 root_objectid,
3922 u64 flags, struct btrfs_disk_key *key,
3923 int level, struct btrfs_key *ins)
3926 struct btrfs_fs_info *fs_info = root->fs_info;
3927 struct btrfs_extent_item *extent_item;
3928 struct btrfs_tree_block_info *block_info;
3929 struct btrfs_extent_inline_ref *iref;
3930 struct btrfs_path *path;
3931 struct extent_buffer *leaf;
3932 u32 size = sizeof(*extent_item) + sizeof(*block_info) + sizeof(*iref);
3934 path = btrfs_alloc_path();
3937 path->leave_spinning = 1;
3938 ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
3942 leaf = path->nodes[0];
3943 extent_item = btrfs_item_ptr(leaf, path->slots[0],
3944 struct btrfs_extent_item);
3945 btrfs_set_extent_refs(leaf, extent_item, 1);
3946 btrfs_set_extent_generation(leaf, extent_item, trans->transid);
3947 btrfs_set_extent_flags(leaf, extent_item,
3948 flags | BTRFS_EXTENT_FLAG_TREE_BLOCK);
3949 block_info = (struct btrfs_tree_block_info *)(extent_item + 1);
3951 btrfs_set_tree_block_key(leaf, block_info, key);
3952 btrfs_set_tree_block_level(leaf, block_info, level);
3954 iref = (struct btrfs_extent_inline_ref *)(block_info + 1);
3956 BUG_ON(!(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF));
3957 btrfs_set_extent_inline_ref_type(leaf, iref,
3958 BTRFS_SHARED_BLOCK_REF_KEY);
3959 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
3961 btrfs_set_extent_inline_ref_type(leaf, iref,
3962 BTRFS_TREE_BLOCK_REF_KEY);
3963 btrfs_set_extent_inline_ref_offset(leaf, iref, root_objectid);
3966 btrfs_mark_buffer_dirty(leaf);
3967 btrfs_free_path(path);
3969 ret = update_block_group(trans, root, ins->objectid, ins->offset,
3972 printk(KERN_ERR "btrfs update block group failed for %llu "
3973 "%llu\n", (unsigned long long)ins->objectid,
3974 (unsigned long long)ins->offset);
3980 int btrfs_alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
3981 struct btrfs_root *root,
3982 u64 root_objectid, u64 owner,
3983 u64 offset, struct btrfs_key *ins)
3987 BUG_ON(root_objectid == BTRFS_TREE_LOG_OBJECTID);
3989 ret = btrfs_add_delayed_data_ref(trans, ins->objectid, ins->offset,
3990 0, root_objectid, owner, offset,
3991 BTRFS_ADD_DELAYED_EXTENT, NULL);
3996 * this is used by the tree logging recovery code. It records that
3997 * an extent has been allocated and makes sure to clear the free
3998 * space cache bits as well
4000 int btrfs_alloc_logged_file_extent(struct btrfs_trans_handle *trans,
4001 struct btrfs_root *root,
4002 u64 root_objectid, u64 owner, u64 offset,
4003 struct btrfs_key *ins)
4006 struct btrfs_block_group_cache *block_group;
4008 block_group = btrfs_lookup_block_group(root->fs_info, ins->objectid);
4009 mutex_lock(&block_group->cache_mutex);
4010 cache_block_group(root, block_group);
4011 mutex_unlock(&block_group->cache_mutex);
4013 ret = btrfs_remove_free_space(block_group, ins->objectid,
4016 btrfs_put_block_group(block_group);
4017 ret = alloc_reserved_file_extent(trans, root, 0, root_objectid,
4018 0, owner, offset, ins, 1);
4023 * finds a free extent and does all the dirty work required for allocation
4024 * returns the key for the extent through ins, and a tree buffer for
4025 * the first block of the extent through buf.
4027 * returns 0 if everything worked, non-zero otherwise.
4029 static int alloc_tree_block(struct btrfs_trans_handle *trans,
4030 struct btrfs_root *root,
4031 u64 num_bytes, u64 parent, u64 root_objectid,
4032 struct btrfs_disk_key *key, int level,
4033 u64 empty_size, u64 hint_byte, u64 search_end,
4034 struct btrfs_key *ins)
4039 ret = __btrfs_reserve_extent(trans, root, num_bytes, num_bytes,
4040 empty_size, hint_byte, search_end,
4044 if (root_objectid == BTRFS_TREE_RELOC_OBJECTID) {
4046 parent = ins->objectid;
4047 flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF;
4051 update_reserved_extents(root, ins->objectid, ins->offset, 1);
4052 if (root_objectid != BTRFS_TREE_LOG_OBJECTID) {
4053 struct btrfs_delayed_extent_op *extent_op;
4054 extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
4057 memcpy(&extent_op->key, key, sizeof(extent_op->key));
4059 memset(&extent_op->key, 0, sizeof(extent_op->key));
4060 extent_op->flags_to_set = flags;
4061 extent_op->update_key = 1;
4062 extent_op->update_flags = 1;
4063 extent_op->is_data = 0;
4065 ret = btrfs_add_delayed_tree_ref(trans, ins->objectid,
4066 ins->offset, parent, root_objectid,
4067 level, BTRFS_ADD_DELAYED_EXTENT,
4074 struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans,
4075 struct btrfs_root *root,
4076 u64 bytenr, u32 blocksize,
4079 struct extent_buffer *buf;
4081 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
4083 return ERR_PTR(-ENOMEM);
4084 btrfs_set_header_generation(buf, trans->transid);
4085 btrfs_set_buffer_lockdep_class(buf, level);
4086 btrfs_tree_lock(buf);
4087 clean_tree_block(trans, root, buf);
4089 btrfs_set_lock_blocking(buf);
4090 btrfs_set_buffer_uptodate(buf);
4092 if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) {
4093 set_extent_dirty(&root->dirty_log_pages, buf->start,
4094 buf->start + buf->len - 1, GFP_NOFS);
4096 set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
4097 buf->start + buf->len - 1, GFP_NOFS);
4099 trans->blocks_used++;
4100 /* this returns a buffer locked for blocking */
4105 * helper function to allocate a block for a given tree
4106 * returns the tree buffer or NULL.
4108 struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
4109 struct btrfs_root *root, u32 blocksize,
4110 u64 parent, u64 root_objectid,
4111 struct btrfs_disk_key *key, int level,
4112 u64 hint, u64 empty_size)
4114 struct btrfs_key ins;
4116 struct extent_buffer *buf;
4118 ret = alloc_tree_block(trans, root, blocksize, parent, root_objectid,
4119 key, level, empty_size, hint, (u64)-1, &ins);
4122 return ERR_PTR(ret);
4125 buf = btrfs_init_new_buffer(trans, root, ins.objectid,
4131 int btrfs_drop_leaf_ref(struct btrfs_trans_handle *trans,
4132 struct btrfs_root *root, struct extent_buffer *leaf)
4136 struct btrfs_key key;
4137 struct btrfs_file_extent_item *fi;
4142 BUG_ON(!btrfs_is_leaf(leaf));
4143 nritems = btrfs_header_nritems(leaf);
4145 for (i = 0; i < nritems; i++) {
4147 btrfs_item_key_to_cpu(leaf, &key, i);
4149 /* only extents have references, skip everything else */
4150 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
4153 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
4155 /* inline extents live in the btree, they don't have refs */
4156 if (btrfs_file_extent_type(leaf, fi) ==
4157 BTRFS_FILE_EXTENT_INLINE)
4160 disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
4162 /* holes don't have refs */
4163 if (disk_bytenr == 0)
4166 num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
4167 ret = btrfs_free_extent(trans, root, disk_bytenr, num_bytes,
4168 leaf->start, 0, key.objectid, 0);
4174 static noinline int cache_drop_leaf_ref(struct btrfs_trans_handle *trans,
4175 struct btrfs_root *root,
4176 struct btrfs_leaf_ref *ref)
4180 struct btrfs_extent_info *info;
4181 struct refsort *sorted;
4183 if (ref->nritems == 0)
4186 sorted = kmalloc(sizeof(*sorted) * ref->nritems, GFP_NOFS);
4187 for (i = 0; i < ref->nritems; i++) {
4188 sorted[i].bytenr = ref->extents[i].bytenr;
4191 sort(sorted, ref->nritems, sizeof(struct refsort), refsort_cmp, NULL);
4194 * the items in the ref were sorted when the ref was inserted
4195 * into the ref cache, so this is already in order
4197 for (i = 0; i < ref->nritems; i++) {
4198 info = ref->extents + sorted[i].slot;
4199 ret = btrfs_free_extent(trans, root, info->bytenr,
4200 info->num_bytes, ref->bytenr,
4201 ref->owner, ref->generation,
4204 atomic_inc(&root->fs_info->throttle_gen);
4205 wake_up(&root->fs_info->transaction_throttle);
4217 static int drop_snap_lookup_refcount(struct btrfs_trans_handle *trans,
4218 struct btrfs_root *root, u64 start,
4223 ret = btrfs_lookup_extent_refs(trans, root, start, len, refs);
4226 #if 0 /* some debugging code in case we see problems here */
4227 /* if the refs count is one, it won't get increased again. But
4228 * if the ref count is > 1, someone may be decreasing it at
4229 * the same time we are.
4232 struct extent_buffer *eb = NULL;
4233 eb = btrfs_find_create_tree_block(root, start, len);
4235 btrfs_tree_lock(eb);
4237 mutex_lock(&root->fs_info->alloc_mutex);
4238 ret = lookup_extent_ref(NULL, root, start, len, refs);
4240 mutex_unlock(&root->fs_info->alloc_mutex);
4243 btrfs_tree_unlock(eb);
4244 free_extent_buffer(eb);
4247 printk(KERN_ERR "btrfs block %llu went down to one "
4248 "during drop_snap\n", (unsigned long long)start);
4260 * this is used while deleting old snapshots, and it drops the refs
4261 * on a whole subtree starting from a level 1 node.
4263 * The idea is to sort all the leaf pointers, and then drop the
4264 * ref on all the leaves in order. Most of the time the leaves
4265 * will have ref cache entries, so no leaf IOs will be required to
4266 * find the extents they have references on.
4268 * For each leaf, any references it has are also dropped in order
4270 * This ends up dropping the references in something close to optimal
4271 * order for reading and modifying the extent allocation tree.
4273 static noinline int drop_level_one_refs(struct btrfs_trans_handle *trans,
4274 struct btrfs_root *root,
4275 struct btrfs_path *path)
4280 struct extent_buffer *eb = path->nodes[1];
4281 struct extent_buffer *leaf;
4282 struct btrfs_leaf_ref *ref;
4283 struct refsort *sorted = NULL;
4284 int nritems = btrfs_header_nritems(eb);
4288 int slot = path->slots[1];
4289 u32 blocksize = btrfs_level_size(root, 0);
4295 root_owner = btrfs_header_owner(eb);
4296 root_gen = btrfs_header_generation(eb);
4297 sorted = kmalloc(sizeof(*sorted) * nritems, GFP_NOFS);
4300 * step one, sort all the leaf pointers so we don't scribble
4301 * randomly into the extent allocation tree
4303 for (i = slot; i < nritems; i++) {
4304 sorted[refi].bytenr = btrfs_node_blockptr(eb, i);
4305 sorted[refi].slot = i;
4310 * nritems won't be zero, but if we're picking up drop_snapshot
4311 * after a crash, slot might be > 0, so double check things
4317 sort(sorted, refi, sizeof(struct refsort), refsort_cmp, NULL);
4320 * the first loop frees everything the leaves point to
4322 for (i = 0; i < refi; i++) {
4325 bytenr = sorted[i].bytenr;
4328 * check the reference count on this leaf. If it is > 1
4329 * we just decrement it below and don't update any
4330 * of the refs the leaf points to.
4332 ret = drop_snap_lookup_refcount(trans, root, bytenr,
4338 ptr_gen = btrfs_node_ptr_generation(eb, sorted[i].slot);
4341 * the leaf only had one reference, which means the
4342 * only thing pointing to this leaf is the snapshot
4343 * we're deleting. It isn't possible for the reference
4344 * count to increase again later
4346 * The reference cache is checked for the leaf,
4347 * and if found we'll be able to drop any refs held by
4348 * the leaf without needing to read it in.
4350 ref = btrfs_lookup_leaf_ref(root, bytenr);
4351 if (ref && ref->generation != ptr_gen) {
4352 btrfs_free_leaf_ref(root, ref);
4356 ret = cache_drop_leaf_ref(trans, root, ref);
4358 btrfs_remove_leaf_ref(root, ref);
4359 btrfs_free_leaf_ref(root, ref);
4362 * the leaf wasn't in the reference cache, so
4363 * we have to read it.
4365 leaf = read_tree_block(root, bytenr, blocksize,
4367 ret = btrfs_drop_leaf_ref(trans, root, leaf);
4369 free_extent_buffer(leaf);
4371 atomic_inc(&root->fs_info->throttle_gen);
4372 wake_up(&root->fs_info->transaction_throttle);
4377 * run through the loop again to free the refs on the leaves.
4378 * This is faster than doing it in the loop above because
4379 * the leaves are likely to be clustered together. We end up
4380 * working in nice chunks on the extent allocation tree.
4382 for (i = 0; i < refi; i++) {
4383 bytenr = sorted[i].bytenr;
4384 ret = btrfs_free_extent(trans, root, bytenr,
4385 blocksize, eb->start,
4386 root_owner, root_gen, 0, 1);
4389 atomic_inc(&root->fs_info->throttle_gen);
4390 wake_up(&root->fs_info->transaction_throttle);
4397 * update the path to show we've processed the entire level 1
4398 * node. This will get saved into the root's drop_snapshot_progress
4399 * field so these drops are not repeated again if this transaction
4402 path->slots[1] = nritems;
4407 * helper function for drop_snapshot, this walks down the tree dropping ref
4408 * counts as it goes.
4410 static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
4411 struct btrfs_root *root,
4412 struct btrfs_path *path, int *level)
4418 struct extent_buffer *next;
4419 struct extent_buffer *cur;
4420 struct extent_buffer *parent;
4425 WARN_ON(*level < 0);
4426 WARN_ON(*level >= BTRFS_MAX_LEVEL);
4427 ret = drop_snap_lookup_refcount(trans, root, path->nodes[*level]->start,
4428 path->nodes[*level]->len, &refs);
4434 * walk down to the last node level and free all the leaves
4436 while (*level >= 0) {
4437 WARN_ON(*level < 0);
4438 WARN_ON(*level >= BTRFS_MAX_LEVEL);
4439 cur = path->nodes[*level];
4441 if (btrfs_header_level(cur) != *level)
4444 if (path->slots[*level] >=
4445 btrfs_header_nritems(cur))
4448 /* the new code goes down to level 1 and does all the
4449 * leaves pointed to that node in bulk. So, this check
4450 * for level 0 will always be false.
4452 * But, the disk format allows the drop_snapshot_progress
4453 * field in the root to leave things in a state where
4454 * a leaf will need cleaning up here. If someone crashes
4455 * with the old code and then boots with the new code,
4456 * we might find a leaf here.
4459 ret = btrfs_drop_leaf_ref(trans, root, cur);
4465 * once we get to level one, process the whole node
4466 * at once, including everything below it.
4469 ret = drop_level_one_refs(trans, root, path);
4474 bytenr = btrfs_node_blockptr(cur, path->slots[*level]);
4475 ptr_gen = btrfs_node_ptr_generation(cur, path->slots[*level]);
4476 blocksize = btrfs_level_size(root, *level - 1);
4478 ret = drop_snap_lookup_refcount(trans, root, bytenr,
4483 * if there is more than one reference, we don't need
4484 * to read that node to drop any references it has. We
4485 * just drop the ref we hold on that node and move on to the
4486 * next slot in this level.
4489 parent = path->nodes[*level];
4490 root_owner = btrfs_header_owner(parent);
4491 root_gen = btrfs_header_generation(parent);
4492 path->slots[*level]++;
4494 ret = btrfs_free_extent(trans, root, bytenr,
4495 blocksize, parent->start,
4496 root_owner, root_gen,
4500 atomic_inc(&root->fs_info->throttle_gen);
4501 wake_up(&root->fs_info->transaction_throttle);
4508 * we need to keep freeing things in the next level down.
4509 * read the block and loop around to process it
4511 next = read_tree_block(root, bytenr, blocksize, ptr_gen);
4512 WARN_ON(*level <= 0);
4513 if (path->nodes[*level-1])
4514 free_extent_buffer(path->nodes[*level-1]);
4515 path->nodes[*level-1] = next;
4516 *level = btrfs_header_level(next);
4517 path->slots[*level] = 0;
4521 WARN_ON(*level < 0);
4522 WARN_ON(*level >= BTRFS_MAX_LEVEL);
4524 if (path->nodes[*level] == root->node) {
4525 parent = path->nodes[*level];
4526 bytenr = path->nodes[*level]->start;
4528 parent = path->nodes[*level + 1];
4529 bytenr = btrfs_node_blockptr(parent, path->slots[*level + 1]);
4532 blocksize = btrfs_level_size(root, *level);
4533 root_owner = btrfs_header_owner(parent);
4534 root_gen = btrfs_header_generation(parent);
4537 * cleanup and free the reference on the last node
4540 ret = btrfs_free_extent(trans, root, bytenr, blocksize,
4541 parent->start, root_owner, root_gen,
4543 free_extent_buffer(path->nodes[*level]);
4544 path->nodes[*level] = NULL;
4554 struct walk_control {
4555 u64 refs[BTRFS_MAX_LEVEL];
4556 u64 flags[BTRFS_MAX_LEVEL];
4557 struct btrfs_key update_progress;
4565 #define DROP_REFERENCE 1
4566 #define UPDATE_BACKREF 2
4569 * hepler to process tree block while walking down the tree.
4571 * when wc->stage == DROP_REFERENCE, this function checks
4572 * reference count of the block. if the block is shared and
4573 * we need update back refs for the subtree rooted at the
4574 * block, this function changes wc->stage to UPDATE_BACKREF
4576 * when wc->stage == UPDATE_BACKREF, this function updates
4577 * back refs for pointers in the block.
4579 * NOTE: return value 1 means we should stop walking down.
4581 static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
4582 struct btrfs_root *root,
4583 struct btrfs_path *path,
4584 struct walk_control *wc)
4586 int level = wc->level;
4587 struct extent_buffer *eb = path->nodes[level];
4588 struct btrfs_key key;
4589 u64 flag = BTRFS_BLOCK_FLAG_FULL_BACKREF;
4592 if (wc->stage == UPDATE_BACKREF &&
4593 btrfs_header_owner(eb) != root->root_key.objectid)
4597 * when reference count of tree block is 1, it won't increase
4598 * again. once full backref flag is set, we never clear it.
4600 if ((wc->stage == DROP_REFERENCE && wc->refs[level] != 1) ||
4601 (wc->stage == UPDATE_BACKREF && !(wc->flags[level] & flag))) {
4602 BUG_ON(!path->locks[level]);
4603 ret = btrfs_lookup_extent_info(trans, root,
4608 BUG_ON(wc->refs[level] == 0);
4611 if (wc->stage == DROP_REFERENCE &&
4612 wc->update_ref && wc->refs[level] > 1) {
4613 BUG_ON(eb == root->node);
4614 BUG_ON(path->slots[level] > 0);
4616 btrfs_item_key_to_cpu(eb, &key, path->slots[level]);
4618 btrfs_node_key_to_cpu(eb, &key, path->slots[level]);
4619 if (btrfs_header_owner(eb) == root->root_key.objectid &&
4620 btrfs_comp_cpu_keys(&key, &wc->update_progress) >= 0) {
4621 wc->stage = UPDATE_BACKREF;
4622 wc->shared_level = level;
4626 if (wc->stage == DROP_REFERENCE) {
4627 if (wc->refs[level] > 1)
4630 if (path->locks[level] && !wc->keep_locks) {
4631 btrfs_tree_unlock(eb);
4632 path->locks[level] = 0;
4637 /* wc->stage == UPDATE_BACKREF */
4638 if (!(wc->flags[level] & flag)) {
4639 BUG_ON(!path->locks[level]);
4640 ret = btrfs_inc_ref(trans, root, eb, 1);
4642 ret = btrfs_dec_ref(trans, root, eb, 0);
4644 ret = btrfs_set_disk_extent_flags(trans, root, eb->start,
4647 wc->flags[level] |= flag;
4651 * the block is shared by multiple trees, so it's not good to
4652 * keep the tree lock
4654 if (path->locks[level] && level > 0) {
4655 btrfs_tree_unlock(eb);
4656 path->locks[level] = 0;
4662 * hepler to process tree block while walking up the tree.
4664 * when wc->stage == DROP_REFERENCE, this function drops
4665 * reference count on the block.
4667 * when wc->stage == UPDATE_BACKREF, this function changes
4668 * wc->stage back to DROP_REFERENCE if we changed wc->stage
4669 * to UPDATE_BACKREF previously while processing the block.
4671 * NOTE: return value 1 means we should stop walking up.
4673 static noinline int walk_up_proc(struct btrfs_trans_handle *trans,
4674 struct btrfs_root *root,
4675 struct btrfs_path *path,
4676 struct walk_control *wc)
4679 int level = wc->level;
4680 struct extent_buffer *eb = path->nodes[level];
4683 if (wc->stage == UPDATE_BACKREF) {
4684 BUG_ON(wc->shared_level < level);
4685 if (level < wc->shared_level)
4688 BUG_ON(wc->refs[level] <= 1);
4689 ret = find_next_key(path, level + 1, &wc->update_progress);
4693 wc->stage = DROP_REFERENCE;
4694 wc->shared_level = -1;
4695 path->slots[level] = 0;
4698 * check reference count again if the block isn't locked.
4699 * we should start walking down the tree again if reference
4702 if (!path->locks[level]) {
4704 btrfs_tree_lock(eb);
4705 btrfs_set_lock_blocking(eb);
4706 path->locks[level] = 1;
4708 ret = btrfs_lookup_extent_info(trans, root,
4713 BUG_ON(wc->refs[level] == 0);
4714 if (wc->refs[level] == 1) {
4715 btrfs_tree_unlock(eb);
4716 path->locks[level] = 0;
4724 /* wc->stage == DROP_REFERENCE */
4725 BUG_ON(wc->refs[level] > 1 && !path->locks[level]);
4727 if (wc->refs[level] == 1) {
4729 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
4730 ret = btrfs_dec_ref(trans, root, eb, 1);
4732 ret = btrfs_dec_ref(trans, root, eb, 0);
4735 /* make block locked assertion in clean_tree_block happy */
4736 if (!path->locks[level] &&
4737 btrfs_header_generation(eb) == trans->transid) {
4738 btrfs_tree_lock(eb);
4739 btrfs_set_lock_blocking(eb);
4740 path->locks[level] = 1;
4742 clean_tree_block(trans, root, eb);
4745 if (eb == root->node) {
4746 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
4749 BUG_ON(root->root_key.objectid !=
4750 btrfs_header_owner(eb));
4752 if (wc->flags[level + 1] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
4753 parent = path->nodes[level + 1]->start;
4755 BUG_ON(root->root_key.objectid !=
4756 btrfs_header_owner(path->nodes[level + 1]));
4759 ret = btrfs_free_extent(trans, root, eb->start, eb->len, parent,
4760 root->root_key.objectid, level, 0);
4763 wc->refs[level] = 0;
4764 wc->flags[level] = 0;
4768 static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
4769 struct btrfs_root *root,
4770 struct btrfs_path *path,
4771 struct walk_control *wc)
4773 struct extent_buffer *next;
4774 struct extent_buffer *cur;
4778 int level = wc->level;
4781 while (level >= 0) {
4782 cur = path->nodes[level];
4783 BUG_ON(path->slots[level] >= btrfs_header_nritems(cur));
4785 ret = walk_down_proc(trans, root, path, wc);
4792 bytenr = btrfs_node_blockptr(cur, path->slots[level]);
4793 blocksize = btrfs_level_size(root, level - 1);
4794 ptr_gen = btrfs_node_ptr_generation(cur, path->slots[level]);
4796 next = read_tree_block(root, bytenr, blocksize, ptr_gen);
4797 btrfs_tree_lock(next);
4798 btrfs_set_lock_blocking(next);
4801 BUG_ON(level != btrfs_header_level(next));
4802 path->nodes[level] = next;
4803 path->slots[level] = 0;
4804 path->locks[level] = 1;
4810 static noinline int walk_up_tree(struct btrfs_trans_handle *trans,
4811 struct btrfs_root *root,
4812 struct btrfs_path *path,
4813 struct walk_control *wc, int max_level)
4815 int level = wc->level;
4818 path->slots[level] = btrfs_header_nritems(path->nodes[level]);
4819 while (level < max_level && path->nodes[level]) {
4821 if (path->slots[level] + 1 <
4822 btrfs_header_nritems(path->nodes[level])) {
4823 path->slots[level]++;
4826 ret = walk_up_proc(trans, root, path, wc);
4830 if (path->locks[level]) {
4831 btrfs_tree_unlock(path->nodes[level]);
4832 path->locks[level] = 0;
4834 free_extent_buffer(path->nodes[level]);
4835 path->nodes[level] = NULL;
4843 * drop a subvolume tree.
4845 * this function traverses the tree freeing any blocks that only
4846 * referenced by the tree.
4848 * when a shared tree block is found. this function decreases its
4849 * reference count by one. if update_ref is true, this function
4850 * also make sure backrefs for the shared block and all lower level
4851 * blocks are properly updated.
4853 int btrfs_drop_snapshot(struct btrfs_root *root, int update_ref)
4855 struct btrfs_path *path;
4856 struct btrfs_trans_handle *trans;
4857 struct btrfs_root *tree_root = root->fs_info->tree_root;
4858 struct btrfs_root_item *root_item = &root->root_item;
4859 struct walk_control *wc;
4860 struct btrfs_key key;
4865 path = btrfs_alloc_path();
4868 wc = kzalloc(sizeof(*wc), GFP_NOFS);
4871 trans = btrfs_start_transaction(tree_root, 1);
4873 if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
4874 level = btrfs_header_level(root->node);
4875 path->nodes[level] = btrfs_lock_root_node(root);
4876 btrfs_set_lock_blocking(path->nodes[level]);
4877 path->slots[level] = 0;
4878 path->locks[level] = 1;
4879 memset(&wc->update_progress, 0,
4880 sizeof(wc->update_progress));
4882 btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
4883 memcpy(&wc->update_progress, &key,
4884 sizeof(wc->update_progress));
4886 level = root_item->drop_level;
4888 path->lowest_level = level;
4889 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
4890 path->lowest_level = 0;
4895 btrfs_node_key_to_cpu(path->nodes[level], &key,
4896 path->slots[level]);
4897 WARN_ON(memcmp(&key, &wc->update_progress, sizeof(key)));
4900 * unlock our path, this is safe because only this
4901 * function is allowed to delete this snapshot
4903 btrfs_unlock_up_safe(path, 0);
4905 level = btrfs_header_level(root->node);
4907 btrfs_tree_lock(path->nodes[level]);
4908 btrfs_set_lock_blocking(path->nodes[level]);
4910 ret = btrfs_lookup_extent_info(trans, root,
4911 path->nodes[level]->start,
4912 path->nodes[level]->len,
4916 BUG_ON(wc->refs[level] == 0);
4918 if (level == root_item->drop_level)
4921 btrfs_tree_unlock(path->nodes[level]);
4922 WARN_ON(wc->refs[level] != 1);
4928 wc->shared_level = -1;
4929 wc->stage = DROP_REFERENCE;
4930 wc->update_ref = update_ref;
4934 ret = walk_down_tree(trans, root, path, wc);
4940 ret = walk_up_tree(trans, root, path, wc, BTRFS_MAX_LEVEL);
4947 BUG_ON(wc->stage != DROP_REFERENCE);
4951 if (wc->stage == DROP_REFERENCE) {
4953 btrfs_node_key(path->nodes[level],
4954 &root_item->drop_progress,
4955 path->slots[level]);
4956 root_item->drop_level = level;
4959 BUG_ON(wc->level == 0);
4960 if (trans->transaction->in_commit ||
4961 trans->transaction->delayed_refs.flushing) {
4962 ret = btrfs_update_root(trans, tree_root,
4967 btrfs_end_transaction(trans, tree_root);
4968 trans = btrfs_start_transaction(tree_root, 1);
4970 unsigned long update;
4971 update = trans->delayed_ref_updates;
4972 trans->delayed_ref_updates = 0;
4974 btrfs_run_delayed_refs(trans, tree_root,
4978 btrfs_release_path(root, path);
4981 ret = btrfs_del_root(trans, tree_root, &root->root_key);
4984 free_extent_buffer(root->node);
4985 free_extent_buffer(root->commit_root);
4988 btrfs_end_transaction(trans, tree_root);
4990 btrfs_free_path(path);
4995 * drop subtree rooted at tree block 'node'.
4997 * NOTE: this function will unlock and release tree block 'node'
4999 int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
5000 struct btrfs_root *root,
5001 struct extent_buffer *node,
5002 struct extent_buffer *parent)
5004 struct btrfs_path *path;
5005 struct walk_control *wc;
5011 BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
5013 path = btrfs_alloc_path();
5016 wc = kzalloc(sizeof(*wc), GFP_NOFS);
5019 btrfs_assert_tree_locked(parent);
5020 parent_level = btrfs_header_level(parent);
5021 extent_buffer_get(parent);
5022 path->nodes[parent_level] = parent;
5023 path->slots[parent_level] = btrfs_header_nritems(parent);
5025 btrfs_assert_tree_locked(node);
5026 level = btrfs_header_level(node);
5027 path->nodes[level] = node;
5028 path->slots[level] = 0;
5029 path->locks[level] = 1;
5031 wc->refs[parent_level] = 1;
5032 wc->flags[parent_level] = BTRFS_BLOCK_FLAG_FULL_BACKREF;
5034 wc->shared_level = -1;
5035 wc->stage = DROP_REFERENCE;
5040 wret = walk_down_tree(trans, root, path, wc);
5046 wret = walk_up_tree(trans, root, path, wc, parent_level);
5054 btrfs_free_path(path);
5059 static unsigned long calc_ra(unsigned long start, unsigned long last,
5062 return min(last, start + nr - 1);
5065 static noinline int relocate_inode_pages(struct inode *inode, u64 start,
5070 unsigned long first_index;
5071 unsigned long last_index;
5074 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
5075 struct file_ra_state *ra;
5076 struct btrfs_ordered_extent *ordered;
5077 unsigned int total_read = 0;
5078 unsigned int total_dirty = 0;
5081 ra = kzalloc(sizeof(*ra), GFP_NOFS);
5083 mutex_lock(&inode->i_mutex);
5084 first_index = start >> PAGE_CACHE_SHIFT;
5085 last_index = (start + len - 1) >> PAGE_CACHE_SHIFT;
5087 /* make sure the dirty trick played by the caller work */
5088 ret = invalidate_inode_pages2_range(inode->i_mapping,
5089 first_index, last_index);
5093 file_ra_state_init(ra, inode->i_mapping);
5095 for (i = first_index ; i <= last_index; i++) {
5096 if (total_read % ra->ra_pages == 0) {
5097 btrfs_force_ra(inode->i_mapping, ra, NULL, i,
5098 calc_ra(i, last_index, ra->ra_pages));
5102 if (((u64)i << PAGE_CACHE_SHIFT) > i_size_read(inode))
5104 page = grab_cache_page(inode->i_mapping, i);
5109 if (!PageUptodate(page)) {
5110 btrfs_readpage(NULL, page);
5112 if (!PageUptodate(page)) {
5114 page_cache_release(page);
5119 wait_on_page_writeback(page);
5121 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
5122 page_end = page_start + PAGE_CACHE_SIZE - 1;
5123 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
5125 ordered = btrfs_lookup_ordered_extent(inode, page_start);
5127 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
5129 page_cache_release(page);
5130 btrfs_start_ordered_extent(inode, ordered, 1);
5131 btrfs_put_ordered_extent(ordered);
5134 set_page_extent_mapped(page);
5136 if (i == first_index)
5137 set_extent_bits(io_tree, page_start, page_end,
5138 EXTENT_BOUNDARY, GFP_NOFS);
5139 btrfs_set_extent_delalloc(inode, page_start, page_end);
5141 set_page_dirty(page);
5144 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
5146 page_cache_release(page);
5151 mutex_unlock(&inode->i_mutex);
5152 balance_dirty_pages_ratelimited_nr(inode->i_mapping, total_dirty);
5156 static noinline int relocate_data_extent(struct inode *reloc_inode,
5157 struct btrfs_key *extent_key,
5160 struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
5161 struct extent_map_tree *em_tree = &BTRFS_I(reloc_inode)->extent_tree;
5162 struct extent_map *em;
5163 u64 start = extent_key->objectid - offset;
5164 u64 end = start + extent_key->offset - 1;
5166 em = alloc_extent_map(GFP_NOFS);
5167 BUG_ON(!em || IS_ERR(em));
5170 em->len = extent_key->offset;
5171 em->block_len = extent_key->offset;
5172 em->block_start = extent_key->objectid;
5173 em->bdev = root->fs_info->fs_devices->latest_bdev;
5174 set_bit(EXTENT_FLAG_PINNED, &em->flags);
5176 /* setup extent map to cheat btrfs_readpage */
5177 lock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS);
5180 spin_lock(&em_tree->lock);
5181 ret = add_extent_mapping(em_tree, em);
5182 spin_unlock(&em_tree->lock);
5183 if (ret != -EEXIST) {
5184 free_extent_map(em);
5187 btrfs_drop_extent_cache(reloc_inode, start, end, 0);
5189 unlock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS);
5191 return relocate_inode_pages(reloc_inode, start, extent_key->offset);
5194 struct btrfs_ref_path {
5196 u64 nodes[BTRFS_MAX_LEVEL];
5198 u64 root_generation;
5205 struct btrfs_key node_keys[BTRFS_MAX_LEVEL];
5206 u64 new_nodes[BTRFS_MAX_LEVEL];
5209 struct disk_extent {
5220 static int is_cowonly_root(u64 root_objectid)
5222 if (root_objectid == BTRFS_ROOT_TREE_OBJECTID ||
5223 root_objectid == BTRFS_EXTENT_TREE_OBJECTID ||
5224 root_objectid == BTRFS_CHUNK_TREE_OBJECTID ||
5225 root_objectid == BTRFS_DEV_TREE_OBJECTID ||
5226 root_objectid == BTRFS_TREE_LOG_OBJECTID ||
5227 root_objectid == BTRFS_CSUM_TREE_OBJECTID)
5232 static noinline int __next_ref_path(struct btrfs_trans_handle *trans,
5233 struct btrfs_root *extent_root,
5234 struct btrfs_ref_path *ref_path,
5237 struct extent_buffer *leaf;
5238 struct btrfs_path *path;
5239 struct btrfs_extent_ref *ref;
5240 struct btrfs_key key;
5241 struct btrfs_key found_key;
5247 path = btrfs_alloc_path();
5252 ref_path->lowest_level = -1;
5253 ref_path->current_level = -1;
5254 ref_path->shared_level = -1;
5258 level = ref_path->current_level - 1;
5259 while (level >= -1) {
5261 if (level < ref_path->lowest_level)
5265 bytenr = ref_path->nodes[level];
5267 bytenr = ref_path->extent_start;
5268 BUG_ON(bytenr == 0);
5270 parent = ref_path->nodes[level + 1];
5271 ref_path->nodes[level + 1] = 0;
5272 ref_path->current_level = level;
5273 BUG_ON(parent == 0);
5275 key.objectid = bytenr;
5276 key.offset = parent + 1;
5277 key.type = BTRFS_EXTENT_REF_KEY;
5279 ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 0);
5284 leaf = path->nodes[0];
5285 nritems = btrfs_header_nritems(leaf);
5286 if (path->slots[0] >= nritems) {
5287 ret = btrfs_next_leaf(extent_root, path);
5292 leaf = path->nodes[0];
5295 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
5296 if (found_key.objectid == bytenr &&
5297 found_key.type == BTRFS_EXTENT_REF_KEY) {
5298 if (level < ref_path->shared_level)
5299 ref_path->shared_level = level;
5304 btrfs_release_path(extent_root, path);
5307 /* reached lowest level */
5311 level = ref_path->current_level;
5312 while (level < BTRFS_MAX_LEVEL - 1) {
5316 bytenr = ref_path->nodes[level];
5318 bytenr = ref_path->extent_start;
5320 BUG_ON(bytenr == 0);
5322 key.objectid = bytenr;
5324 key.type = BTRFS_EXTENT_REF_KEY;
5326 ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 0);
5330 leaf = path->nodes[0];
5331 nritems = btrfs_header_nritems(leaf);
5332 if (path->slots[0] >= nritems) {
5333 ret = btrfs_next_leaf(extent_root, path);
5337 /* the extent was freed by someone */
5338 if (ref_path->lowest_level == level)
5340 btrfs_release_path(extent_root, path);
5343 leaf = path->nodes[0];
5346 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
5347 if (found_key.objectid != bytenr ||
5348 found_key.type != BTRFS_EXTENT_REF_KEY) {
5349 /* the extent was freed by someone */
5350 if (ref_path->lowest_level == level) {
5354 btrfs_release_path(extent_root, path);
5358 ref = btrfs_item_ptr(leaf, path->slots[0],
5359 struct btrfs_extent_ref);
5360 ref_objectid = btrfs_ref_objectid(leaf, ref);
5361 if (ref_objectid < BTRFS_FIRST_FREE_OBJECTID) {
5363 level = (int)ref_objectid;
5364 BUG_ON(level >= BTRFS_MAX_LEVEL);
5365 ref_path->lowest_level = level;
5366 ref_path->current_level = level;
5367 ref_path->nodes[level] = bytenr;
5369 WARN_ON(ref_objectid != level);
5372 WARN_ON(level != -1);
5376 if (ref_path->lowest_level == level) {
5377 ref_path->owner_objectid = ref_objectid;
5378 ref_path->num_refs = btrfs_ref_num_refs(leaf, ref);
5382 * the block is tree root or the block isn't in reference
5385 if (found_key.objectid == found_key.offset ||
5386 is_cowonly_root(btrfs_ref_root(leaf, ref))) {
5387 ref_path->root_objectid = btrfs_ref_root(leaf, ref);
5388 ref_path->root_generation =
5389 btrfs_ref_generation(leaf, ref);
5391 /* special reference from the tree log */
5392 ref_path->nodes[0] = found_key.offset;
5393 ref_path->current_level = 0;
5400 BUG_ON(ref_path->nodes[level] != 0);
5401 ref_path->nodes[level] = found_key.offset;
5402 ref_path->current_level = level;
5405 * the reference was created in the running transaction,
5406 * no need to continue walking up.
5408 if (btrfs_ref_generation(leaf, ref) == trans->transid) {
5409 ref_path->root_objectid = btrfs_ref_root(leaf, ref);
5410 ref_path->root_generation =
5411 btrfs_ref_generation(leaf, ref);
5416 btrfs_release_path(extent_root, path);
5419 /* reached max tree level, but no tree root found. */
5422 btrfs_free_path(path);
5426 static int btrfs_first_ref_path(struct btrfs_trans_handle *trans,
5427 struct btrfs_root *extent_root,
5428 struct btrfs_ref_path *ref_path,
5431 memset(ref_path, 0, sizeof(*ref_path));
5432 ref_path->extent_start = extent_start;
5434 return __next_ref_path(trans, extent_root, ref_path, 1);
5437 static int btrfs_next_ref_path(struct btrfs_trans_handle *trans,
5438 struct btrfs_root *extent_root,
5439 struct btrfs_ref_path *ref_path)
5441 return __next_ref_path(trans, extent_root, ref_path, 0);
5444 static noinline int get_new_locations(struct inode *reloc_inode,
5445 struct btrfs_key *extent_key,
5446 u64 offset, int no_fragment,
5447 struct disk_extent **extents,
5450 struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
5451 struct btrfs_path *path;
5452 struct btrfs_file_extent_item *fi;
5453 struct extent_buffer *leaf;
5454 struct disk_extent *exts = *extents;
5455 struct btrfs_key found_key;
5460 int max = *nr_extents;
5463 WARN_ON(!no_fragment && *extents);
5466 exts = kmalloc(sizeof(*exts) * max, GFP_NOFS);
5471 path = btrfs_alloc_path();
5474 cur_pos = extent_key->objectid - offset;
5475 last_byte = extent_key->objectid + extent_key->offset;
5476 ret = btrfs_lookup_file_extent(NULL, root, path, reloc_inode->i_ino,
5486 leaf = path->nodes[0];
5487 nritems = btrfs_header_nritems(leaf);
5488 if (path->slots[0] >= nritems) {
5489 ret = btrfs_next_leaf(root, path);
5494 leaf = path->nodes[0];
5497 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
5498 if (found_key.offset != cur_pos ||
5499 found_key.type != BTRFS_EXTENT_DATA_KEY ||
5500 found_key.objectid != reloc_inode->i_ino)
5503 fi = btrfs_item_ptr(leaf, path->slots[0],
5504 struct btrfs_file_extent_item);
5505 if (btrfs_file_extent_type(leaf, fi) !=
5506 BTRFS_FILE_EXTENT_REG ||
5507 btrfs_file_extent_disk_bytenr(leaf, fi) == 0)
5511 struct disk_extent *old = exts;
5513 exts = kzalloc(sizeof(*exts) * max, GFP_NOFS);
5514 memcpy(exts, old, sizeof(*exts) * nr);
5515 if (old != *extents)
5519 exts[nr].disk_bytenr =
5520 btrfs_file_extent_disk_bytenr(leaf, fi);
5521 exts[nr].disk_num_bytes =
5522 btrfs_file_extent_disk_num_bytes(leaf, fi);
5523 exts[nr].offset = btrfs_file_extent_offset(leaf, fi);
5524 exts[nr].num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
5525 exts[nr].ram_bytes = btrfs_file_extent_ram_bytes(leaf, fi);
5526 exts[nr].compression = btrfs_file_extent_compression(leaf, fi);
5527 exts[nr].encryption = btrfs_file_extent_encryption(leaf, fi);
5528 exts[nr].other_encoding = btrfs_file_extent_other_encoding(leaf,
5530 BUG_ON(exts[nr].offset > 0);
5531 BUG_ON(exts[nr].compression || exts[nr].encryption);
5532 BUG_ON(exts[nr].num_bytes != exts[nr].disk_num_bytes);
5534 cur_pos += exts[nr].num_bytes;
5537 if (cur_pos + offset >= last_byte)
5547 BUG_ON(cur_pos + offset > last_byte);
5548 if (cur_pos + offset < last_byte) {
5554 btrfs_free_path(path);
5556 if (exts != *extents)
5565 static noinline int replace_one_extent(struct btrfs_trans_handle *trans,
5566 struct btrfs_root *root,
5567 struct btrfs_path *path,
5568 struct btrfs_key *extent_key,
5569 struct btrfs_key *leaf_key,
5570 struct btrfs_ref_path *ref_path,
5571 struct disk_extent *new_extents,
5574 struct extent_buffer *leaf;
5575 struct btrfs_file_extent_item *fi;
5576 struct inode *inode = NULL;
5577 struct btrfs_key key;
5582 u64 search_end = (u64)-1;
5585 int extent_locked = 0;
5589 memcpy(&key, leaf_key, sizeof(key));
5590 if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS) {
5591 if (key.objectid < ref_path->owner_objectid ||
5592 (key.objectid == ref_path->owner_objectid &&
5593 key.type < BTRFS_EXTENT_DATA_KEY)) {
5594 key.objectid = ref_path->owner_objectid;
5595 key.type = BTRFS_EXTENT_DATA_KEY;
5601 ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
5605 leaf = path->nodes[0];
5606 nritems = btrfs_header_nritems(leaf);
5608 if (extent_locked && ret > 0) {
5610 * the file extent item was modified by someone
5611 * before the extent got locked.
5613 unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
5614 lock_end, GFP_NOFS);
5618 if (path->slots[0] >= nritems) {
5619 if (++nr_scaned > 2)
5622 BUG_ON(extent_locked);
5623 ret = btrfs_next_leaf(root, path);
5628 leaf = path->nodes[0];
5629 nritems = btrfs_header_nritems(leaf);
5632 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
5634 if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS) {
5635 if ((key.objectid > ref_path->owner_objectid) ||
5636 (key.objectid == ref_path->owner_objectid &&
5637 key.type > BTRFS_EXTENT_DATA_KEY) ||
5638 key.offset >= search_end)
5642 if (inode && key.objectid != inode->i_ino) {
5643 BUG_ON(extent_locked);
5644 btrfs_release_path(root, path);
5645 mutex_unlock(&inode->i_mutex);
5651 if (key.type != BTRFS_EXTENT_DATA_KEY) {
5656 fi = btrfs_item_ptr(leaf, path->slots[0],
5657 struct btrfs_file_extent_item);
5658 extent_type = btrfs_file_extent_type(leaf, fi);
5659 if ((extent_type != BTRFS_FILE_EXTENT_REG &&
5660 extent_type != BTRFS_FILE_EXTENT_PREALLOC) ||
5661 (btrfs_file_extent_disk_bytenr(leaf, fi) !=
5662 extent_key->objectid)) {
5668 num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
5669 ext_offset = btrfs_file_extent_offset(leaf, fi);
5671 if (search_end == (u64)-1) {
5672 search_end = key.offset - ext_offset +
5673 btrfs_file_extent_ram_bytes(leaf, fi);
5676 if (!extent_locked) {
5677 lock_start = key.offset;
5678 lock_end = lock_start + num_bytes - 1;
5680 if (lock_start > key.offset ||
5681 lock_end + 1 < key.offset + num_bytes) {
5682 unlock_extent(&BTRFS_I(inode)->io_tree,
5683 lock_start, lock_end, GFP_NOFS);
5689 btrfs_release_path(root, path);
5691 inode = btrfs_iget_locked(root->fs_info->sb,
5692 key.objectid, root);
5693 if (inode->i_state & I_NEW) {
5694 BTRFS_I(inode)->root = root;
5695 BTRFS_I(inode)->location.objectid =
5697 BTRFS_I(inode)->location.type =
5698 BTRFS_INODE_ITEM_KEY;
5699 BTRFS_I(inode)->location.offset = 0;
5700 btrfs_read_locked_inode(inode);
5701 unlock_new_inode(inode);
5704 * some code call btrfs_commit_transaction while
5705 * holding the i_mutex, so we can't use mutex_lock
5708 if (is_bad_inode(inode) ||
5709 !mutex_trylock(&inode->i_mutex)) {
5712 key.offset = (u64)-1;
5717 if (!extent_locked) {
5718 struct btrfs_ordered_extent *ordered;
5720 btrfs_release_path(root, path);
5722 lock_extent(&BTRFS_I(inode)->io_tree, lock_start,
5723 lock_end, GFP_NOFS);
5724 ordered = btrfs_lookup_first_ordered_extent(inode,
5727 ordered->file_offset <= lock_end &&
5728 ordered->file_offset + ordered->len > lock_start) {
5729 unlock_extent(&BTRFS_I(inode)->io_tree,
5730 lock_start, lock_end, GFP_NOFS);
5731 btrfs_start_ordered_extent(inode, ordered, 1);
5732 btrfs_put_ordered_extent(ordered);
5733 key.offset += num_bytes;
5737 btrfs_put_ordered_extent(ordered);
5743 if (nr_extents == 1) {
5744 /* update extent pointer in place */
5745 btrfs_set_file_extent_disk_bytenr(leaf, fi,
5746 new_extents[0].disk_bytenr);
5747 btrfs_set_file_extent_disk_num_bytes(leaf, fi,
5748 new_extents[0].disk_num_bytes);
5749 btrfs_mark_buffer_dirty(leaf);
5751 btrfs_drop_extent_cache(inode, key.offset,
5752 key.offset + num_bytes - 1, 0);
5754 ret = btrfs_inc_extent_ref(trans, root,
5755 new_extents[0].disk_bytenr,
5756 new_extents[0].disk_num_bytes,
5758 root->root_key.objectid,
5763 ret = btrfs_free_extent(trans, root,
5764 extent_key->objectid,
5767 btrfs_header_owner(leaf),
5768 btrfs_header_generation(leaf),
5772 btrfs_release_path(root, path);
5773 key.offset += num_bytes;
5781 * drop old extent pointer at first, then insert the
5782 * new pointers one bye one
5784 btrfs_release_path(root, path);
5785 ret = btrfs_drop_extents(trans, root, inode, key.offset,
5786 key.offset + num_bytes,
5787 key.offset, &alloc_hint);
5790 for (i = 0; i < nr_extents; i++) {
5791 if (ext_offset >= new_extents[i].num_bytes) {
5792 ext_offset -= new_extents[i].num_bytes;
5795 extent_len = min(new_extents[i].num_bytes -
5796 ext_offset, num_bytes);
5798 ret = btrfs_insert_empty_item(trans, root,
5803 leaf = path->nodes[0];
5804 fi = btrfs_item_ptr(leaf, path->slots[0],
5805 struct btrfs_file_extent_item);
5806 btrfs_set_file_extent_generation(leaf, fi,
5808 btrfs_set_file_extent_type(leaf, fi,
5809 BTRFS_FILE_EXTENT_REG);
5810 btrfs_set_file_extent_disk_bytenr(leaf, fi,
5811 new_extents[i].disk_bytenr);
5812 btrfs_set_file_extent_disk_num_bytes(leaf, fi,
5813 new_extents[i].disk_num_bytes);
5814 btrfs_set_file_extent_ram_bytes(leaf, fi,
5815 new_extents[i].ram_bytes);
5817 btrfs_set_file_extent_compression(leaf, fi,
5818 new_extents[i].compression);
5819 btrfs_set_file_extent_encryption(leaf, fi,
5820 new_extents[i].encryption);
5821 btrfs_set_file_extent_other_encoding(leaf, fi,
5822 new_extents[i].other_encoding);
5824 btrfs_set_file_extent_num_bytes(leaf, fi,
5826 ext_offset += new_extents[i].offset;
5827 btrfs_set_file_extent_offset(leaf, fi,
5829 btrfs_mark_buffer_dirty(leaf);
5831 btrfs_drop_extent_cache(inode, key.offset,
5832 key.offset + extent_len - 1, 0);
5834 ret = btrfs_inc_extent_ref(trans, root,
5835 new_extents[i].disk_bytenr,
5836 new_extents[i].disk_num_bytes,
5838 root->root_key.objectid,
5839 trans->transid, key.objectid);
5841 btrfs_release_path(root, path);
5843 inode_add_bytes(inode, extent_len);
5846 num_bytes -= extent_len;
5847 key.offset += extent_len;
5852 BUG_ON(i >= nr_extents);
5856 if (extent_locked) {
5857 unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
5858 lock_end, GFP_NOFS);
5862 if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS &&
5863 key.offset >= search_end)
5870 btrfs_release_path(root, path);
5872 mutex_unlock(&inode->i_mutex);
5873 if (extent_locked) {
5874 unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
5875 lock_end, GFP_NOFS);
5882 int btrfs_reloc_tree_cache_ref(struct btrfs_trans_handle *trans,
5883 struct btrfs_root *root,
5884 struct extent_buffer *buf, u64 orig_start)
5889 BUG_ON(btrfs_header_generation(buf) != trans->transid);
5890 BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
5892 level = btrfs_header_level(buf);
5894 struct btrfs_leaf_ref *ref;
5895 struct btrfs_leaf_ref *orig_ref;
5897 orig_ref = btrfs_lookup_leaf_ref(root, orig_start);
5901 ref = btrfs_alloc_leaf_ref(root, orig_ref->nritems);
5903 btrfs_free_leaf_ref(root, orig_ref);
5907 ref->nritems = orig_ref->nritems;
5908 memcpy(ref->extents, orig_ref->extents,
5909 sizeof(ref->extents[0]) * ref->nritems);
5911 btrfs_free_leaf_ref(root, orig_ref);
5913 ref->root_gen = trans->transid;
5914 ref->bytenr = buf->start;
5915 ref->owner = btrfs_header_owner(buf);
5916 ref->generation = btrfs_header_generation(buf);
5918 ret = btrfs_add_leaf_ref(root, ref, 0);
5920 btrfs_free_leaf_ref(root, ref);
5925 static noinline int invalidate_extent_cache(struct btrfs_root *root,
5926 struct extent_buffer *leaf,
5927 struct btrfs_block_group_cache *group,
5928 struct btrfs_root *target_root)
5930 struct btrfs_key key;
5931 struct inode *inode = NULL;
5932 struct btrfs_file_extent_item *fi;
5934 u64 skip_objectid = 0;
5938 nritems = btrfs_header_nritems(leaf);
5939 for (i = 0; i < nritems; i++) {
5940 btrfs_item_key_to_cpu(leaf, &key, i);
5941 if (key.objectid == skip_objectid ||
5942 key.type != BTRFS_EXTENT_DATA_KEY)
5944 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
5945 if (btrfs_file_extent_type(leaf, fi) ==
5946 BTRFS_FILE_EXTENT_INLINE)
5948 if (btrfs_file_extent_disk_bytenr(leaf, fi) == 0)
5950 if (!inode || inode->i_ino != key.objectid) {
5952 inode = btrfs_ilookup(target_root->fs_info->sb,
5953 key.objectid, target_root, 1);
5956 skip_objectid = key.objectid;
5959 num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
5961 lock_extent(&BTRFS_I(inode)->io_tree, key.offset,
5962 key.offset + num_bytes - 1, GFP_NOFS);
5963 btrfs_drop_extent_cache(inode, key.offset,
5964 key.offset + num_bytes - 1, 1);
5965 unlock_extent(&BTRFS_I(inode)->io_tree, key.offset,
5966 key.offset + num_bytes - 1, GFP_NOFS);
5973 static noinline int replace_extents_in_leaf(struct btrfs_trans_handle *trans,
5974 struct btrfs_root *root,
5975 struct extent_buffer *leaf,
5976 struct btrfs_block_group_cache *group,
5977 struct inode *reloc_inode)
5979 struct btrfs_key key;
5980 struct btrfs_key extent_key;
5981 struct btrfs_file_extent_item *fi;
5982 struct btrfs_leaf_ref *ref;
5983 struct disk_extent *new_extent;
5992 new_extent = kmalloc(sizeof(*new_extent), GFP_NOFS);
5993 BUG_ON(!new_extent);
5995 ref = btrfs_lookup_leaf_ref(root, leaf->start);
5999 nritems = btrfs_header_nritems(leaf);
6000 for (i = 0; i < nritems; i++) {
6001 btrfs_item_key_to_cpu(leaf, &key, i);
6002 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
6004 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
6005 if (btrfs_file_extent_type(leaf, fi) ==
6006 BTRFS_FILE_EXTENT_INLINE)
6008 bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
6009 num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
6014 if (bytenr >= group->key.objectid + group->key.offset ||
6015 bytenr + num_bytes <= group->key.objectid)
6018 extent_key.objectid = bytenr;
6019 extent_key.offset = num_bytes;
6020 extent_key.type = BTRFS_EXTENT_ITEM_KEY;
6022 ret = get_new_locations(reloc_inode, &extent_key,
6023 group->key.objectid, 1,
6024 &new_extent, &nr_extent);
6029 BUG_ON(ref->extents[ext_index].bytenr != bytenr);
6030 BUG_ON(ref->extents[ext_index].num_bytes != num_bytes);
6031 ref->extents[ext_index].bytenr = new_extent->disk_bytenr;
6032 ref->extents[ext_index].num_bytes = new_extent->disk_num_bytes;
6034 btrfs_set_file_extent_disk_bytenr(leaf, fi,
6035 new_extent->disk_bytenr);
6036 btrfs_set_file_extent_disk_num_bytes(leaf, fi,
6037 new_extent->disk_num_bytes);
6038 btrfs_mark_buffer_dirty(leaf);
6040 ret = btrfs_inc_extent_ref(trans, root,
6041 new_extent->disk_bytenr,
6042 new_extent->disk_num_bytes,
6044 root->root_key.objectid,
6045 trans->transid, key.objectid);
6048 ret = btrfs_free_extent(trans, root,
6049 bytenr, num_bytes, leaf->start,
6050 btrfs_header_owner(leaf),
6051 btrfs_header_generation(leaf),
6057 BUG_ON(ext_index + 1 != ref->nritems);
6058 btrfs_free_leaf_ref(root, ref);
6062 int btrfs_free_reloc_root(struct btrfs_trans_handle *trans,
6063 struct btrfs_root *root)
6065 struct btrfs_root *reloc_root;
6068 if (root->reloc_root) {
6069 reloc_root = root->reloc_root;
6070 root->reloc_root = NULL;
6071 list_add(&reloc_root->dead_list,
6072 &root->fs_info->dead_reloc_roots);
6074 btrfs_set_root_bytenr(&reloc_root->root_item,
6075 reloc_root->node->start);
6076 btrfs_set_root_level(&root->root_item,
6077 btrfs_header_level(reloc_root->node));
6078 memset(&reloc_root->root_item.drop_progress, 0,
6079 sizeof(struct btrfs_disk_key));
6080 reloc_root->root_item.drop_level = 0;
6082 ret = btrfs_update_root(trans, root->fs_info->tree_root,
6083 &reloc_root->root_key,
6084 &reloc_root->root_item);
6090 int btrfs_drop_dead_reloc_roots(struct btrfs_root *root)
6092 struct btrfs_trans_handle *trans;
6093 struct btrfs_root *reloc_root;
6094 struct btrfs_root *prev_root = NULL;
6095 struct list_head dead_roots;
6099 INIT_LIST_HEAD(&dead_roots);
6100 list_splice_init(&root->fs_info->dead_reloc_roots, &dead_roots);
6102 while (!list_empty(&dead_roots)) {
6103 reloc_root = list_entry(dead_roots.prev,
6104 struct btrfs_root, dead_list);
6105 list_del_init(&reloc_root->dead_list);
6107 BUG_ON(reloc_root->commit_root != NULL);
6109 trans = btrfs_join_transaction(root, 1);
6112 mutex_lock(&root->fs_info->drop_mutex);
6113 ret = btrfs_drop_snapshot(trans, reloc_root);
6116 mutex_unlock(&root->fs_info->drop_mutex);
6118 nr = trans->blocks_used;
6119 ret = btrfs_end_transaction(trans, root);
6121 btrfs_btree_balance_dirty(root, nr);
6124 free_extent_buffer(reloc_root->node);
6126 ret = btrfs_del_root(trans, root->fs_info->tree_root,
6127 &reloc_root->root_key);
6129 mutex_unlock(&root->fs_info->drop_mutex);
6131 nr = trans->blocks_used;
6132 ret = btrfs_end_transaction(trans, root);
6134 btrfs_btree_balance_dirty(root, nr);
6137 prev_root = reloc_root;
6140 btrfs_remove_leaf_refs(prev_root, (u64)-1, 0);
6146 int btrfs_add_dead_reloc_root(struct btrfs_root *root)
6148 list_add(&root->dead_list, &root->fs_info->dead_reloc_roots);
6152 int btrfs_cleanup_reloc_trees(struct btrfs_root *root)
6154 struct btrfs_root *reloc_root;
6155 struct btrfs_trans_handle *trans;
6156 struct btrfs_key location;
6160 mutex_lock(&root->fs_info->tree_reloc_mutex);
6161 ret = btrfs_find_dead_roots(root, BTRFS_TREE_RELOC_OBJECTID, NULL);
6163 found = !list_empty(&root->fs_info->dead_reloc_roots);
6164 mutex_unlock(&root->fs_info->tree_reloc_mutex);
6167 trans = btrfs_start_transaction(root, 1);
6169 ret = btrfs_commit_transaction(trans, root);
6173 location.objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
6174 location.offset = (u64)-1;
6175 location.type = BTRFS_ROOT_ITEM_KEY;
6177 reloc_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
6178 BUG_ON(!reloc_root);
6179 btrfs_orphan_cleanup(reloc_root);
6183 static noinline int init_reloc_tree(struct btrfs_trans_handle *trans,
6184 struct btrfs_root *root)
6186 struct btrfs_root *reloc_root;
6187 struct extent_buffer *eb;
6188 struct btrfs_root_item *root_item;
6189 struct btrfs_key root_key;
6192 BUG_ON(!root->ref_cows);
6193 if (root->reloc_root)
6196 root_item = kmalloc(sizeof(*root_item), GFP_NOFS);
6199 ret = btrfs_copy_root(trans, root, root->commit_root,
6200 &eb, BTRFS_TREE_RELOC_OBJECTID);
6203 root_key.objectid = BTRFS_TREE_RELOC_OBJECTID;
6204 root_key.offset = root->root_key.objectid;
6205 root_key.type = BTRFS_ROOT_ITEM_KEY;
6207 memcpy(root_item, &root->root_item, sizeof(root_item));
6208 btrfs_set_root_refs(root_item, 0);
6209 btrfs_set_root_bytenr(root_item, eb->start);
6210 btrfs_set_root_level(root_item, btrfs_header_level(eb));
6211 btrfs_set_root_generation(root_item, trans->transid);
6213 btrfs_tree_unlock(eb);
6214 free_extent_buffer(eb);
6216 ret = btrfs_insert_root(trans, root->fs_info->tree_root,
6217 &root_key, root_item);
6221 reloc_root = btrfs_read_fs_root_no_radix(root->fs_info->tree_root,
6223 BUG_ON(!reloc_root);
6224 reloc_root->last_trans = trans->transid;
6225 reloc_root->commit_root = NULL;
6226 reloc_root->ref_tree = &root->fs_info->reloc_ref_tree;
6228 root->reloc_root = reloc_root;
6233 * Core function of space balance.
6235 * The idea is using reloc trees to relocate tree blocks in reference
6236 * counted roots. There is one reloc tree for each subvol, and all
6237 * reloc trees share same root key objectid. Reloc trees are snapshots
6238 * of the latest committed roots of subvols (root->commit_root).
6240 * To relocate a tree block referenced by a subvol, there are two steps.
6241 * COW the block through subvol's reloc tree, then update block pointer
6242 * in the subvol to point to the new block. Since all reloc trees share
6243 * same root key objectid, doing special handing for tree blocks owned
6244 * by them is easy. Once a tree block has been COWed in one reloc tree,
6245 * we can use the resulting new block directly when the same block is
6246 * required to COW again through other reloc trees. By this way, relocated
6247 * tree blocks are shared between reloc trees, so they are also shared
6250 static noinline int relocate_one_path(struct btrfs_trans_handle *trans,
6251 struct btrfs_root *root,
6252 struct btrfs_path *path,
6253 struct btrfs_key *first_key,
6254 struct btrfs_ref_path *ref_path,
6255 struct btrfs_block_group_cache *group,
6256 struct inode *reloc_inode)
6258 struct btrfs_root *reloc_root;
6259 struct extent_buffer *eb = NULL;
6260 struct btrfs_key *keys;
6264 int lowest_level = 0;
6267 if (ref_path->owner_objectid < BTRFS_FIRST_FREE_OBJECTID)
6268 lowest_level = ref_path->owner_objectid;
6270 if (!root->ref_cows) {
6271 path->lowest_level = lowest_level;
6272 ret = btrfs_search_slot(trans, root, first_key, path, 0, 1);
6274 path->lowest_level = 0;
6275 btrfs_release_path(root, path);
6279 mutex_lock(&root->fs_info->tree_reloc_mutex);
6280 ret = init_reloc_tree(trans, root);
6282 reloc_root = root->reloc_root;
6284 shared_level = ref_path->shared_level;
6285 ref_path->shared_level = BTRFS_MAX_LEVEL - 1;
6287 keys = ref_path->node_keys;
6288 nodes = ref_path->new_nodes;
6289 memset(&keys[shared_level + 1], 0,
6290 sizeof(*keys) * (BTRFS_MAX_LEVEL - shared_level - 1));
6291 memset(&nodes[shared_level + 1], 0,
6292 sizeof(*nodes) * (BTRFS_MAX_LEVEL - shared_level - 1));
6294 if (nodes[lowest_level] == 0) {
6295 path->lowest_level = lowest_level;
6296 ret = btrfs_search_slot(trans, reloc_root, first_key, path,
6299 for (level = lowest_level; level < BTRFS_MAX_LEVEL; level++) {
6300 eb = path->nodes[level];
6301 if (!eb || eb == reloc_root->node)
6303 nodes[level] = eb->start;
6305 btrfs_item_key_to_cpu(eb, &keys[level], 0);
6307 btrfs_node_key_to_cpu(eb, &keys[level], 0);
6310 ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
6311 eb = path->nodes[0];
6312 ret = replace_extents_in_leaf(trans, reloc_root, eb,
6313 group, reloc_inode);
6316 btrfs_release_path(reloc_root, path);
6318 ret = btrfs_merge_path(trans, reloc_root, keys, nodes,
6324 * replace tree blocks in the fs tree with tree blocks in
6327 ret = btrfs_merge_path(trans, root, keys, nodes, lowest_level);
6330 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
6331 ret = btrfs_search_slot(trans, reloc_root, first_key, path,
6334 extent_buffer_get(path->nodes[0]);
6335 eb = path->nodes[0];
6336 btrfs_release_path(reloc_root, path);
6337 ret = invalidate_extent_cache(reloc_root, eb, group, root);
6339 free_extent_buffer(eb);
6342 mutex_unlock(&root->fs_info->tree_reloc_mutex);
6343 path->lowest_level = 0;
6347 static noinline int relocate_tree_block(struct btrfs_trans_handle *trans,
6348 struct btrfs_root *root,
6349 struct btrfs_path *path,
6350 struct btrfs_key *first_key,
6351 struct btrfs_ref_path *ref_path)
6355 ret = relocate_one_path(trans, root, path, first_key,
6356 ref_path, NULL, NULL);
6362 static noinline int del_extent_zero(struct btrfs_trans_handle *trans,
6363 struct btrfs_root *extent_root,
6364 struct btrfs_path *path,
6365 struct btrfs_key *extent_key)
6369 ret = btrfs_search_slot(trans, extent_root, extent_key, path, -1, 1);
6372 ret = btrfs_del_item(trans, extent_root, path);
6374 btrfs_release_path(extent_root, path);
6378 static noinline struct btrfs_root *read_ref_root(struct btrfs_fs_info *fs_info,
6379 struct btrfs_ref_path *ref_path)
6381 struct btrfs_key root_key;
6383 root_key.objectid = ref_path->root_objectid;
6384 root_key.type = BTRFS_ROOT_ITEM_KEY;
6385 if (is_cowonly_root(ref_path->root_objectid))
6386 root_key.offset = 0;
6388 root_key.offset = (u64)-1;
6390 return btrfs_read_fs_root_no_name(fs_info, &root_key);
6393 static noinline int relocate_one_extent(struct btrfs_root *extent_root,
6394 struct btrfs_path *path,
6395 struct btrfs_key *extent_key,
6396 struct btrfs_block_group_cache *group,
6397 struct inode *reloc_inode, int pass)
6399 struct btrfs_trans_handle *trans;
6400 struct btrfs_root *found_root;
6401 struct btrfs_ref_path *ref_path = NULL;
6402 struct disk_extent *new_extents = NULL;
6407 struct btrfs_key first_key;
6411 trans = btrfs_start_transaction(extent_root, 1);
6414 if (extent_key->objectid == 0) {
6415 ret = del_extent_zero(trans, extent_root, path, extent_key);
6419 ref_path = kmalloc(sizeof(*ref_path), GFP_NOFS);
6425 for (loops = 0; ; loops++) {
6427 ret = btrfs_first_ref_path(trans, extent_root, ref_path,
6428 extent_key->objectid);
6430 ret = btrfs_next_ref_path(trans, extent_root, ref_path);
6437 if (ref_path->root_objectid == BTRFS_TREE_LOG_OBJECTID ||
6438 ref_path->root_objectid == BTRFS_TREE_RELOC_OBJECTID)
6441 found_root = read_ref_root(extent_root->fs_info, ref_path);
6442 BUG_ON(!found_root);
6444 * for reference counted tree, only process reference paths
6445 * rooted at the latest committed root.
6447 if (found_root->ref_cows &&
6448 ref_path->root_generation != found_root->root_key.offset)
6451 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
6454 * copy data extents to new locations
6456 u64 group_start = group->key.objectid;
6457 ret = relocate_data_extent(reloc_inode,
6466 level = ref_path->owner_objectid;
6469 if (prev_block != ref_path->nodes[level]) {
6470 struct extent_buffer *eb;
6471 u64 block_start = ref_path->nodes[level];
6472 u64 block_size = btrfs_level_size(found_root, level);
6474 eb = read_tree_block(found_root, block_start,
6476 btrfs_tree_lock(eb);
6477 BUG_ON(level != btrfs_header_level(eb));
6480 btrfs_item_key_to_cpu(eb, &first_key, 0);
6482 btrfs_node_key_to_cpu(eb, &first_key, 0);
6484 btrfs_tree_unlock(eb);
6485 free_extent_buffer(eb);
6486 prev_block = block_start;
6489 mutex_lock(&extent_root->fs_info->trans_mutex);
6490 btrfs_record_root_in_trans(found_root);
6491 mutex_unlock(&extent_root->fs_info->trans_mutex);
6492 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
6494 * try to update data extent references while
6495 * keeping metadata shared between snapshots.
6498 ret = relocate_one_path(trans, found_root,
6499 path, &first_key, ref_path,
6500 group, reloc_inode);
6506 * use fallback method to process the remaining
6510 u64 group_start = group->key.objectid;
6511 new_extents = kmalloc(sizeof(*new_extents),
6514 ret = get_new_locations(reloc_inode,
6522 ret = replace_one_extent(trans, found_root,
6524 &first_key, ref_path,
6525 new_extents, nr_extents);
6527 ret = relocate_tree_block(trans, found_root, path,
6528 &first_key, ref_path);
6535 btrfs_end_transaction(trans, extent_root);
6542 static u64 update_block_group_flags(struct btrfs_root *root, u64 flags)
6545 u64 stripped = BTRFS_BLOCK_GROUP_RAID0 |
6546 BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10;
6548 num_devices = root->fs_info->fs_devices->rw_devices;
6549 if (num_devices == 1) {
6550 stripped |= BTRFS_BLOCK_GROUP_DUP;
6551 stripped = flags & ~stripped;
6553 /* turn raid0 into single device chunks */
6554 if (flags & BTRFS_BLOCK_GROUP_RAID0)
6557 /* turn mirroring into duplication */
6558 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
6559 BTRFS_BLOCK_GROUP_RAID10))
6560 return stripped | BTRFS_BLOCK_GROUP_DUP;
6563 /* they already had raid on here, just return */
6564 if (flags & stripped)
6567 stripped |= BTRFS_BLOCK_GROUP_DUP;
6568 stripped = flags & ~stripped;
6570 /* switch duplicated blocks with raid1 */
6571 if (flags & BTRFS_BLOCK_GROUP_DUP)
6572 return stripped | BTRFS_BLOCK_GROUP_RAID1;
6574 /* turn single device chunks into raid0 */
6575 return stripped | BTRFS_BLOCK_GROUP_RAID0;
6580 static int __alloc_chunk_for_shrink(struct btrfs_root *root,
6581 struct btrfs_block_group_cache *shrink_block_group,
6584 struct btrfs_trans_handle *trans;
6585 u64 new_alloc_flags;
6588 spin_lock(&shrink_block_group->lock);
6589 if (btrfs_block_group_used(&shrink_block_group->item) +
6590 shrink_block_group->reserved > 0) {
6591 spin_unlock(&shrink_block_group->lock);
6593 trans = btrfs_start_transaction(root, 1);
6594 spin_lock(&shrink_block_group->lock);
6596 new_alloc_flags = update_block_group_flags(root,
6597 shrink_block_group->flags);
6598 if (new_alloc_flags != shrink_block_group->flags) {
6600 btrfs_block_group_used(&shrink_block_group->item);
6602 calc = shrink_block_group->key.offset;
6604 spin_unlock(&shrink_block_group->lock);
6606 do_chunk_alloc(trans, root->fs_info->extent_root,
6607 calc + 2 * 1024 * 1024, new_alloc_flags, force);
6609 btrfs_end_transaction(trans, root);
6611 spin_unlock(&shrink_block_group->lock);
6616 int btrfs_prepare_block_group_relocation(struct btrfs_root *root,
6617 struct btrfs_block_group_cache *group)
6620 __alloc_chunk_for_shrink(root, group, 1);
6621 set_block_group_readonly(group);
6626 static int __insert_orphan_inode(struct btrfs_trans_handle *trans,
6627 struct btrfs_root *root,
6628 u64 objectid, u64 size)
6630 struct btrfs_path *path;
6631 struct btrfs_inode_item *item;
6632 struct extent_buffer *leaf;
6635 path = btrfs_alloc_path();
6639 path->leave_spinning = 1;
6640 ret = btrfs_insert_empty_inode(trans, root, path, objectid);
6644 leaf = path->nodes[0];
6645 item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_inode_item);
6646 memset_extent_buffer(leaf, 0, (unsigned long)item, sizeof(*item));
6647 btrfs_set_inode_generation(leaf, item, 1);
6648 btrfs_set_inode_size(leaf, item, size);
6649 btrfs_set_inode_mode(leaf, item, S_IFREG | 0600);
6650 btrfs_set_inode_flags(leaf, item, BTRFS_INODE_NOCOMPRESS);
6651 btrfs_mark_buffer_dirty(leaf);
6652 btrfs_release_path(root, path);
6654 btrfs_free_path(path);
6658 static noinline struct inode *create_reloc_inode(struct btrfs_fs_info *fs_info,
6659 struct btrfs_block_group_cache *group)
6661 struct inode *inode = NULL;
6662 struct btrfs_trans_handle *trans;
6663 struct btrfs_root *root;
6664 struct btrfs_key root_key;
6665 u64 objectid = BTRFS_FIRST_FREE_OBJECTID;
6668 root_key.objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
6669 root_key.type = BTRFS_ROOT_ITEM_KEY;
6670 root_key.offset = (u64)-1;
6671 root = btrfs_read_fs_root_no_name(fs_info, &root_key);
6673 return ERR_CAST(root);
6675 trans = btrfs_start_transaction(root, 1);
6678 err = btrfs_find_free_objectid(trans, root, objectid, &objectid);
6682 err = __insert_orphan_inode(trans, root, objectid, group->key.offset);
6685 err = btrfs_insert_file_extent(trans, root, objectid, 0, 0, 0,
6686 group->key.offset, 0, group->key.offset,
6690 inode = btrfs_iget_locked(root->fs_info->sb, objectid, root);
6691 if (inode->i_state & I_NEW) {
6692 BTRFS_I(inode)->root = root;
6693 BTRFS_I(inode)->location.objectid = objectid;
6694 BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY;
6695 BTRFS_I(inode)->location.offset = 0;
6696 btrfs_read_locked_inode(inode);
6697 unlock_new_inode(inode);
6698 BUG_ON(is_bad_inode(inode));
6702 BTRFS_I(inode)->index_cnt = group->key.objectid;
6704 err = btrfs_orphan_add(trans, inode);
6706 btrfs_end_transaction(trans, root);
6710 inode = ERR_PTR(err);
6715 int btrfs_reloc_clone_csums(struct inode *inode, u64 file_pos, u64 len)
6718 struct btrfs_ordered_sum *sums;
6719 struct btrfs_sector_sum *sector_sum;
6720 struct btrfs_ordered_extent *ordered;
6721 struct btrfs_root *root = BTRFS_I(inode)->root;
6722 struct list_head list;
6727 INIT_LIST_HEAD(&list);
6729 ordered = btrfs_lookup_ordered_extent(inode, file_pos);
6730 BUG_ON(ordered->file_offset != file_pos || ordered->len != len);
6732 disk_bytenr = file_pos + BTRFS_I(inode)->index_cnt;
6733 ret = btrfs_lookup_csums_range(root->fs_info->csum_root, disk_bytenr,
6734 disk_bytenr + len - 1, &list);
6736 while (!list_empty(&list)) {
6737 sums = list_entry(list.next, struct btrfs_ordered_sum, list);
6738 list_del_init(&sums->list);
6740 sector_sum = sums->sums;
6741 sums->bytenr = ordered->start;
6744 while (offset < sums->len) {
6745 sector_sum->bytenr += ordered->start - disk_bytenr;
6747 offset += root->sectorsize;
6750 btrfs_add_ordered_sum(inode, ordered, sums);
6752 btrfs_put_ordered_extent(ordered);
6756 int btrfs_relocate_block_group(struct btrfs_root *root, u64 group_start)
6758 struct btrfs_trans_handle *trans;
6759 struct btrfs_path *path;
6760 struct btrfs_fs_info *info = root->fs_info;
6761 struct extent_buffer *leaf;
6762 struct inode *reloc_inode;
6763 struct btrfs_block_group_cache *block_group;
6764 struct btrfs_key key;
6773 root = root->fs_info->extent_root;
6775 block_group = btrfs_lookup_block_group(info, group_start);
6776 BUG_ON(!block_group);
6778 printk(KERN_INFO "btrfs relocating block group %llu flags %llu\n",
6779 (unsigned long long)block_group->key.objectid,
6780 (unsigned long long)block_group->flags);
6782 path = btrfs_alloc_path();
6785 reloc_inode = create_reloc_inode(info, block_group);
6786 BUG_ON(IS_ERR(reloc_inode));
6788 __alloc_chunk_for_shrink(root, block_group, 1);
6789 set_block_group_readonly(block_group);
6791 btrfs_start_delalloc_inodes(info->tree_root);
6792 btrfs_wait_ordered_extents(info->tree_root, 0);
6797 key.objectid = block_group->key.objectid;
6800 cur_byte = key.objectid;
6802 trans = btrfs_start_transaction(info->tree_root, 1);
6803 btrfs_commit_transaction(trans, info->tree_root);
6805 mutex_lock(&root->fs_info->cleaner_mutex);
6806 btrfs_clean_old_snapshots(info->tree_root);
6807 btrfs_remove_leaf_refs(info->tree_root, (u64)-1, 1);
6808 mutex_unlock(&root->fs_info->cleaner_mutex);
6810 trans = btrfs_start_transaction(info->tree_root, 1);
6811 btrfs_commit_transaction(trans, info->tree_root);
6814 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
6818 leaf = path->nodes[0];
6819 nritems = btrfs_header_nritems(leaf);
6820 if (path->slots[0] >= nritems) {
6821 ret = btrfs_next_leaf(root, path);
6828 leaf = path->nodes[0];
6829 nritems = btrfs_header_nritems(leaf);
6832 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
6834 if (key.objectid >= block_group->key.objectid +
6835 block_group->key.offset)
6838 if (progress && need_resched()) {
6839 btrfs_release_path(root, path);
6846 if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY ||
6847 key.objectid + key.offset <= cur_byte) {
6853 cur_byte = key.objectid + key.offset;
6854 btrfs_release_path(root, path);
6856 __alloc_chunk_for_shrink(root, block_group, 0);
6857 ret = relocate_one_extent(root, path, &key, block_group,
6863 key.objectid = cur_byte;
6868 btrfs_release_path(root, path);
6871 btrfs_wait_ordered_range(reloc_inode, 0, (u64)-1);
6872 invalidate_mapping_pages(reloc_inode->i_mapping, 0, -1);
6875 if (total_found > 0) {
6876 printk(KERN_INFO "btrfs found %llu extents in pass %d\n",
6877 (unsigned long long)total_found, pass);
6879 if (total_found == skipped && pass > 2) {
6881 reloc_inode = create_reloc_inode(info, block_group);
6887 /* delete reloc_inode */
6890 /* unpin extents in this range */
6891 trans = btrfs_start_transaction(info->tree_root, 1);
6892 btrfs_commit_transaction(trans, info->tree_root);
6894 spin_lock(&block_group->lock);
6895 WARN_ON(block_group->pinned > 0);
6896 WARN_ON(block_group->reserved > 0);
6897 WARN_ON(btrfs_block_group_used(&block_group->item) > 0);
6898 spin_unlock(&block_group->lock);
6899 btrfs_put_block_group(block_group);
6902 btrfs_free_path(path);
6907 static int find_first_block_group(struct btrfs_root *root,
6908 struct btrfs_path *path, struct btrfs_key *key)
6911 struct btrfs_key found_key;
6912 struct extent_buffer *leaf;
6915 ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
6920 slot = path->slots[0];
6921 leaf = path->nodes[0];
6922 if (slot >= btrfs_header_nritems(leaf)) {
6923 ret = btrfs_next_leaf(root, path);
6930 btrfs_item_key_to_cpu(leaf, &found_key, slot);
6932 if (found_key.objectid >= key->objectid &&
6933 found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) {
6944 int btrfs_free_block_groups(struct btrfs_fs_info *info)
6946 struct btrfs_block_group_cache *block_group;
6947 struct btrfs_space_info *space_info;
6950 spin_lock(&info->block_group_cache_lock);
6951 while ((n = rb_last(&info->block_group_cache_tree)) != NULL) {
6952 block_group = rb_entry(n, struct btrfs_block_group_cache,
6954 rb_erase(&block_group->cache_node,
6955 &info->block_group_cache_tree);
6956 spin_unlock(&info->block_group_cache_lock);
6958 btrfs_remove_free_space_cache(block_group);
6959 down_write(&block_group->space_info->groups_sem);
6960 list_del(&block_group->list);
6961 up_write(&block_group->space_info->groups_sem);
6963 WARN_ON(atomic_read(&block_group->count) != 1);
6966 spin_lock(&info->block_group_cache_lock);
6968 spin_unlock(&info->block_group_cache_lock);
6970 /* now that all the block groups are freed, go through and
6971 * free all the space_info structs. This is only called during
6972 * the final stages of unmount, and so we know nobody is
6973 * using them. We call synchronize_rcu() once before we start,
6974 * just to be on the safe side.
6978 while(!list_empty(&info->space_info)) {
6979 space_info = list_entry(info->space_info.next,
6980 struct btrfs_space_info,
6983 list_del(&space_info->list);
6989 int btrfs_read_block_groups(struct btrfs_root *root)
6991 struct btrfs_path *path;
6993 struct btrfs_block_group_cache *cache;
6994 struct btrfs_fs_info *info = root->fs_info;
6995 struct btrfs_space_info *space_info;
6996 struct btrfs_key key;
6997 struct btrfs_key found_key;
6998 struct extent_buffer *leaf;
7000 root = info->extent_root;
7003 btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
7004 path = btrfs_alloc_path();
7009 ret = find_first_block_group(root, path, &key);
7017 leaf = path->nodes[0];
7018 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
7019 cache = kzalloc(sizeof(*cache), GFP_NOFS);
7025 atomic_set(&cache->count, 1);
7026 spin_lock_init(&cache->lock);
7027 spin_lock_init(&cache->tree_lock);
7028 mutex_init(&cache->cache_mutex);
7029 INIT_LIST_HEAD(&cache->list);
7030 INIT_LIST_HEAD(&cache->cluster_list);
7031 read_extent_buffer(leaf, &cache->item,
7032 btrfs_item_ptr_offset(leaf, path->slots[0]),
7033 sizeof(cache->item));
7034 memcpy(&cache->key, &found_key, sizeof(found_key));
7036 key.objectid = found_key.objectid + found_key.offset;
7037 btrfs_release_path(root, path);
7038 cache->flags = btrfs_block_group_flags(&cache->item);
7040 ret = update_space_info(info, cache->flags, found_key.offset,
7041 btrfs_block_group_used(&cache->item),
7044 cache->space_info = space_info;
7045 down_write(&space_info->groups_sem);
7046 list_add_tail(&cache->list, &space_info->block_groups);
7047 up_write(&space_info->groups_sem);
7049 ret = btrfs_add_block_group_cache(root->fs_info, cache);
7052 set_avail_alloc_bits(root->fs_info, cache->flags);
7053 if (btrfs_chunk_readonly(root, cache->key.objectid))
7054 set_block_group_readonly(cache);
7058 btrfs_free_path(path);
7062 int btrfs_make_block_group(struct btrfs_trans_handle *trans,
7063 struct btrfs_root *root, u64 bytes_used,
7064 u64 type, u64 chunk_objectid, u64 chunk_offset,
7068 struct btrfs_root *extent_root;
7069 struct btrfs_block_group_cache *cache;
7071 extent_root = root->fs_info->extent_root;
7073 root->fs_info->last_trans_log_full_commit = trans->transid;
7075 cache = kzalloc(sizeof(*cache), GFP_NOFS);
7079 cache->key.objectid = chunk_offset;
7080 cache->key.offset = size;
7081 cache->key.type = BTRFS_BLOCK_GROUP_ITEM_KEY;
7082 atomic_set(&cache->count, 1);
7083 spin_lock_init(&cache->lock);
7084 spin_lock_init(&cache->tree_lock);
7085 mutex_init(&cache->cache_mutex);
7086 INIT_LIST_HEAD(&cache->list);
7087 INIT_LIST_HEAD(&cache->cluster_list);
7089 btrfs_set_block_group_used(&cache->item, bytes_used);
7090 btrfs_set_block_group_chunk_objectid(&cache->item, chunk_objectid);
7091 cache->flags = type;
7092 btrfs_set_block_group_flags(&cache->item, type);
7094 ret = update_space_info(root->fs_info, cache->flags, size, bytes_used,
7095 &cache->space_info);
7097 down_write(&cache->space_info->groups_sem);
7098 list_add_tail(&cache->list, &cache->space_info->block_groups);
7099 up_write(&cache->space_info->groups_sem);
7101 ret = btrfs_add_block_group_cache(root->fs_info, cache);
7104 ret = btrfs_insert_item(trans, extent_root, &cache->key, &cache->item,
7105 sizeof(cache->item));
7108 set_avail_alloc_bits(extent_root->fs_info, type);
7113 int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
7114 struct btrfs_root *root, u64 group_start)
7116 struct btrfs_path *path;
7117 struct btrfs_block_group_cache *block_group;
7118 struct btrfs_free_cluster *cluster;
7119 struct btrfs_key key;
7122 root = root->fs_info->extent_root;
7124 block_group = btrfs_lookup_block_group(root->fs_info, group_start);
7125 BUG_ON(!block_group);
7126 BUG_ON(!block_group->ro);
7128 memcpy(&key, &block_group->key, sizeof(key));
7130 /* make sure this block group isn't part of an allocation cluster */
7131 cluster = &root->fs_info->data_alloc_cluster;
7132 spin_lock(&cluster->refill_lock);
7133 btrfs_return_cluster_to_free_space(block_group, cluster);
7134 spin_unlock(&cluster->refill_lock);
7137 * make sure this block group isn't part of a metadata
7138 * allocation cluster
7140 cluster = &root->fs_info->meta_alloc_cluster;
7141 spin_lock(&cluster->refill_lock);
7142 btrfs_return_cluster_to_free_space(block_group, cluster);
7143 spin_unlock(&cluster->refill_lock);
7145 path = btrfs_alloc_path();
7148 spin_lock(&root->fs_info->block_group_cache_lock);
7149 rb_erase(&block_group->cache_node,
7150 &root->fs_info->block_group_cache_tree);
7151 spin_unlock(&root->fs_info->block_group_cache_lock);
7152 btrfs_remove_free_space_cache(block_group);
7153 down_write(&block_group->space_info->groups_sem);
7155 * we must use list_del_init so people can check to see if they
7156 * are still on the list after taking the semaphore
7158 list_del_init(&block_group->list);
7159 up_write(&block_group->space_info->groups_sem);
7161 spin_lock(&block_group->space_info->lock);
7162 block_group->space_info->total_bytes -= block_group->key.offset;
7163 block_group->space_info->bytes_readonly -= block_group->key.offset;
7164 spin_unlock(&block_group->space_info->lock);
7165 block_group->space_info->full = 0;
7167 btrfs_put_block_group(block_group);
7168 btrfs_put_block_group(block_group);
7170 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
7176 ret = btrfs_del_item(trans, root, path);
7178 btrfs_free_path(path);