projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Btrfs: do extent allocation and reference count updates in the background
[deliverable/linux.git] / fs / btrfs / extent-tree.c
1 /*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
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.
7 *
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.
12 *
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.
17 */
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>
24 #include "compat.h"
25 #include "hash.h"
26 #include "crc32c.h"
27 #include "ctree.h"
28 #include "disk-io.h"
29 #include "print-tree.h"
30 #include "transaction.h"
31 #include "volumes.h"
32 #include "locking.h"
33 #include "ref-cache.h"
34
35 #define PENDING_EXTENT_INSERT 0
36 #define PENDING_EXTENT_DELETE 1
37 #define PENDING_BACKREF_UPDATE 2
38
39 struct pending_extent_op {
40 int type;
41 u64 bytenr;
42 u64 num_bytes;
43 u64 parent;
44 u64 orig_parent;
45 u64 generation;
46 u64 orig_generation;
47 int level;
48 struct list_head list;
49 int del;
50 };
51
52 static int __btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans,
53 struct btrfs_root *root, u64 parent,
54 u64 root_objectid, u64 ref_generation,
55 u64 owner, struct btrfs_key *ins,
56 int ref_mod);
57 static int update_reserved_extents(struct btrfs_root *root,
58 u64 bytenr, u64 num, int reserve);
59 static int pin_down_bytes(struct btrfs_trans_handle *trans,
60 struct btrfs_root *root,
61 u64 bytenr, u64 num_bytes, int is_data);
62 static int update_block_group(struct btrfs_trans_handle *trans,
63 struct btrfs_root *root,
64 u64 bytenr, u64 num_bytes, int alloc,
65 int mark_free);
66 static noinline int __btrfs_free_extent(struct btrfs_trans_handle *trans,
67 struct btrfs_root *root,
68 u64 bytenr, u64 num_bytes, u64 parent,
69 u64 root_objectid, u64 ref_generation,
70 u64 owner_objectid, int pin,
71 int ref_to_drop);
72
73 static int do_chunk_alloc(struct btrfs_trans_handle *trans,
74 struct btrfs_root *extent_root, u64 alloc_bytes,
75 u64 flags, int force);
76
77 static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits)
78 {
79 return (cache->flags & bits) == bits;
80 }
81
82 /*
83 * this adds the block group to the fs_info rb tree for the block group
84 * cache
85 */
86 static int btrfs_add_block_group_cache(struct btrfs_fs_info *info,
87 struct btrfs_block_group_cache *block_group)
88 {
89 struct rb_node **p;
90 struct rb_node *parent = NULL;
91 struct btrfs_block_group_cache *cache;
92
93 spin_lock(&info->block_group_cache_lock);
94 p = &info->block_group_cache_tree.rb_node;
95
96 while (*p) {
97 parent = *p;
98 cache = rb_entry(parent, struct btrfs_block_group_cache,
99 cache_node);
100 if (block_group->key.objectid < cache->key.objectid) {
101 p = &(*p)->rb_left;
102 } else if (block_group->key.objectid > cache->key.objectid) {
103 p = &(*p)->rb_right;
104 } else {
105 spin_unlock(&info->block_group_cache_lock);
106 return -EEXIST;
107 }
108 }
109
110 rb_link_node(&block_group->cache_node, parent, p);
111 rb_insert_color(&block_group->cache_node,
112 &info->block_group_cache_tree);
113 spin_unlock(&info->block_group_cache_lock);
114
115 return 0;
116 }
117
118 /*
119 * This will return the block group at or after bytenr if contains is 0, else
120 * it will return the block group that contains the bytenr
121 */
122 static struct btrfs_block_group_cache *
123 block_group_cache_tree_search(struct btrfs_fs_info *info, u64 bytenr,
124 int contains)
125 {
126 struct btrfs_block_group_cache *cache, *ret = NULL;
127 struct rb_node *n;
128 u64 end, start;
129
130 spin_lock(&info->block_group_cache_lock);
131 n = info->block_group_cache_tree.rb_node;
132
133 while (n) {
134 cache = rb_entry(n, struct btrfs_block_group_cache,
135 cache_node);
136 end = cache->key.objectid + cache->key.offset - 1;
137 start = cache->key.objectid;
138
139 if (bytenr < start) {
140 if (!contains && (!ret || start < ret->key.objectid))
141 ret = cache;
142 n = n->rb_left;
143 } else if (bytenr > start) {
144 if (contains && bytenr <= end) {
145 ret = cache;
146 break;
147 }
148 n = n->rb_right;
149 } else {
150 ret = cache;
151 break;
152 }
153 }
154 if (ret)
155 atomic_inc(&ret->count);
156 spin_unlock(&info->block_group_cache_lock);
157
158 return ret;
159 }
160
161 /*
162 * this is only called by cache_block_group, since we could have freed extents
163 * we need to check the pinned_extents for any extents that can't be used yet
164 * since their free space will be released as soon as the transaction commits.
165 */
166 static int add_new_free_space(struct btrfs_block_group_cache *block_group,
167 struct btrfs_fs_info *info, u64 start, u64 end)
168 {
169 u64 extent_start, extent_end, size;
170 int ret;
171
172 mutex_lock(&info->pinned_mutex);
173 while (start < end) {
174 ret = find_first_extent_bit(&info->pinned_extents, start,
175 &extent_start, &extent_end,
176 EXTENT_DIRTY);
177 if (ret)
178 break;
179
180 if (extent_start == start) {
181 start = extent_end + 1;
182 } else if (extent_start > start && extent_start < end) {
183 size = extent_start - start;
184 ret = btrfs_add_free_space(block_group, start,
185 size);
186 BUG_ON(ret);
187 start = extent_end + 1;
188 } else {
189 break;
190 }
191 }
192
193 if (start < end) {
194 size = end - start;
195 ret = btrfs_add_free_space(block_group, start, size);
196 BUG_ON(ret);
197 }
198 mutex_unlock(&info->pinned_mutex);
199
200 return 0;
201 }
202
203 static int remove_sb_from_cache(struct btrfs_root *root,
204 struct btrfs_block_group_cache *cache)
205 {
206 u64 bytenr;
207 u64 *logical;
208 int stripe_len;
209 int i, nr, ret;
210
211 for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
212 bytenr = btrfs_sb_offset(i);
213 ret = btrfs_rmap_block(&root->fs_info->mapping_tree,
214 cache->key.objectid, bytenr, 0,
215 &logical, &nr, &stripe_len);
216 BUG_ON(ret);
217 while (nr--) {
218 btrfs_remove_free_space(cache, logical[nr],
219 stripe_len);
220 }
221 kfree(logical);
222 }
223 return 0;
224 }
225
226 static int cache_block_group(struct btrfs_root *root,
227 struct btrfs_block_group_cache *block_group)
228 {
229 struct btrfs_path *path;
230 int ret = 0;
231 struct btrfs_key key;
232 struct extent_buffer *leaf;
233 int slot;
234 u64 last;
235
236 if (!block_group)
237 return 0;
238
239 root = root->fs_info->extent_root;
240
241 if (block_group->cached)
242 return 0;
243
244 path = btrfs_alloc_path();
245 if (!path)
246 return -ENOMEM;
247
248 path->reada = 2;
249 /*
250 * we get into deadlocks with paths held by callers of this function.
251 * since the alloc_mutex is protecting things right now, just
252 * skip the locking here
253 */
254 path->skip_locking = 1;
255 last = max_t(u64, block_group->key.objectid, BTRFS_SUPER_INFO_OFFSET);
256 key.objectid = last;
257 key.offset = 0;
258 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
259 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
260 if (ret < 0)
261 goto err;
262
263 while (1) {
264 leaf = path->nodes[0];
265 slot = path->slots[0];
266 if (slot >= btrfs_header_nritems(leaf)) {
267 ret = btrfs_next_leaf(root, path);
268 if (ret < 0)
269 goto err;
270 if (ret == 0)
271 continue;
272 else
273 break;
274 }
275 btrfs_item_key_to_cpu(leaf, &key, slot);
276 if (key.objectid < block_group->key.objectid)
277 goto next;
278
279 if (key.objectid >= block_group->key.objectid +
280 block_group->key.offset)
281 break;
282
283 if (btrfs_key_type(&key) == BTRFS_EXTENT_ITEM_KEY) {
284 add_new_free_space(block_group, root->fs_info, last,
285 key.objectid);
286
287 last = key.objectid + key.offset;
288 }
289 next:
290 path->slots[0]++;
291 }
292
293 add_new_free_space(block_group, root->fs_info, last,
294 block_group->key.objectid +
295 block_group->key.offset);
296
297 remove_sb_from_cache(root, block_group);
298 block_group->cached = 1;
299 ret = 0;
300 err:
301 btrfs_free_path(path);
302 return ret;
303 }
304
305 /*
306 * return the block group that starts at or after bytenr
307 */
308 static struct btrfs_block_group_cache *
309 btrfs_lookup_first_block_group(struct btrfs_fs_info *info, u64 bytenr)
310 {
311 struct btrfs_block_group_cache *cache;
312
313 cache = block_group_cache_tree_search(info, bytenr, 0);
314
315 return cache;
316 }
317
318 /*
319 * return the block group that contains teh given bytenr
320 */
321 struct btrfs_block_group_cache *btrfs_lookup_block_group(
322 struct btrfs_fs_info *info,
323 u64 bytenr)
324 {
325 struct btrfs_block_group_cache *cache;
326
327 cache = block_group_cache_tree_search(info, bytenr, 1);
328
329 return cache;
330 }
331
332 static inline void put_block_group(struct btrfs_block_group_cache *cache)
333 {
334 if (atomic_dec_and_test(&cache->count))
335 kfree(cache);
336 }
337
338 static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info,
339 u64 flags)
340 {
341 struct list_head *head = &info->space_info;
342 struct btrfs_space_info *found;
343
344 rcu_read_lock();
345 list_for_each_entry_rcu(found, head, list) {
346 if (found->flags == flags) {
347 rcu_read_unlock();
348 return found;
349 }
350 }
351 rcu_read_unlock();
352 return NULL;
353 }
354
355 /*
356 * after adding space to the filesystem, we need to clear the full flags
357 * on all the space infos.
358 */
359 void btrfs_clear_space_info_full(struct btrfs_fs_info *info)
360 {
361 struct list_head *head = &info->space_info;
362 struct btrfs_space_info *found;
363
364 rcu_read_lock();
365 list_for_each_entry_rcu(found, head, list)
366 found->full = 0;
367 rcu_read_unlock();
368 }
369
370 static u64 div_factor(u64 num, int factor)
371 {
372 if (factor == 10)
373 return num;
374 num *= factor;
375 do_div(num, 10);
376 return num;
377 }
378
379 u64 btrfs_find_block_group(struct btrfs_root *root,
380 u64 search_start, u64 search_hint, int owner)
381 {
382 struct btrfs_block_group_cache *cache;
383 u64 used;
384 u64 last = max(search_hint, search_start);
385 u64 group_start = 0;
386 int full_search = 0;
387 int factor = 9;
388 int wrapped = 0;
389 again:
390 while (1) {
391 cache = btrfs_lookup_first_block_group(root->fs_info, last);
392 if (!cache)
393 break;
394
395 spin_lock(&cache->lock);
396 last = cache->key.objectid + cache->key.offset;
397 used = btrfs_block_group_used(&cache->item);
398
399 if ((full_search || !cache->ro) &&
400 block_group_bits(cache, BTRFS_BLOCK_GROUP_METADATA)) {
401 if (used + cache->pinned + cache->reserved <
402 div_factor(cache->key.offset, factor)) {
403 group_start = cache->key.objectid;
404 spin_unlock(&cache->lock);
405 put_block_group(cache);
406 goto found;
407 }
408 }
409 spin_unlock(&cache->lock);
410 put_block_group(cache);
411 cond_resched();
412 }
413 if (!wrapped) {
414 last = search_start;
415 wrapped = 1;
416 goto again;
417 }
418 if (!full_search && factor < 10) {
419 last = search_start;
420 full_search = 1;
421 factor = 10;
422 goto again;
423 }
424 found:
425 return group_start;
426 }
427
428 /* simple helper to search for an existing extent at a given offset */
429 int btrfs_lookup_extent(struct btrfs_root *root, u64 start, u64 len)
430 {
431 int ret;
432 struct btrfs_key key;
433 struct btrfs_path *path;
434
435 path = btrfs_alloc_path();
436 BUG_ON(!path);
437 key.objectid = start;
438 key.offset = len;
439 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
440 ret = btrfs_search_slot(NULL, root->fs_info->extent_root, &key, path,
441 0, 0);
442 btrfs_free_path(path);
443 return ret;
444 }
445
446 /*
447 * Back reference rules. Back refs have three main goals:
448 *
449 * 1) differentiate between all holders of references to an extent so that
450 * when a reference is dropped we can make sure it was a valid reference
451 * before freeing the extent.
452 *
453 * 2) Provide enough information to quickly find the holders of an extent
454 * if we notice a given block is corrupted or bad.
455 *
456 * 3) Make it easy to migrate blocks for FS shrinking or storage pool
457 * maintenance. This is actually the same as #2, but with a slightly
458 * different use case.
459 *
460 * File extents can be referenced by:
461 *
462 * - multiple snapshots, subvolumes, or different generations in one subvol
463 * - different files inside a single subvolume
464 * - different offsets inside a file (bookend extents in file.c)
465 *
466 * The extent ref structure has fields for:
467 *
468 * - Objectid of the subvolume root
469 * - Generation number of the tree holding the reference
470 * - objectid of the file holding the reference
471 * - number of references holding by parent node (alway 1 for tree blocks)
472 *
473 * Btree leaf may hold multiple references to a file extent. In most cases,
474 * these references are from same file and the corresponding offsets inside
475 * the file are close together.
476 *
477 * When a file extent is allocated the fields are filled in:
478 * (root_key.objectid, trans->transid, inode objectid, 1)
479 *
480 * When a leaf is cow'd new references are added for every file extent found
481 * in the leaf. It looks similar to the create case, but trans->transid will
482 * be different when the block is cow'd.
483 *
484 * (root_key.objectid, trans->transid, inode objectid,
485 * number of references in the leaf)
486 *
487 * When a file extent is removed either during snapshot deletion or
488 * file truncation, we find the corresponding back reference and check
489 * the following fields:
490 *
491 * (btrfs_header_owner(leaf), btrfs_header_generation(leaf),
492 * inode objectid)
493 *
494 * Btree extents can be referenced by:
495 *
496 * - Different subvolumes
497 * - Different generations of the same subvolume
498 *
499 * When a tree block is created, back references are inserted:
500 *
501 * (root->root_key.objectid, trans->transid, level, 1)
502 *
503 * When a tree block is cow'd, new back references are added for all the
504 * blocks it points to. If the tree block isn't in reference counted root,
505 * the old back references are removed. These new back references are of
506 * the form (trans->transid will have increased since creation):
507 *
508 * (root->root_key.objectid, trans->transid, level, 1)
509 *
510 * When a backref is in deleting, the following fields are checked:
511 *
512 * if backref was for a tree root:
513 * (btrfs_header_owner(itself), btrfs_header_generation(itself), level)
514 * else
515 * (btrfs_header_owner(parent), btrfs_header_generation(parent), level)
516 *
517 * Back Reference Key composing:
518 *
519 * The key objectid corresponds to the first byte in the extent, the key
520 * type is set to BTRFS_EXTENT_REF_KEY, and the key offset is the first
521 * byte of parent extent. If a extent is tree root, the key offset is set
522 * to the key objectid.
523 */
524
525 static noinline int lookup_extent_backref(struct btrfs_trans_handle *trans,
526 struct btrfs_root *root,
527 struct btrfs_path *path,
528 u64 bytenr, u64 parent,
529 u64 ref_root, u64 ref_generation,
530 u64 owner_objectid, int del)
531 {
532 struct btrfs_key key;
533 struct btrfs_extent_ref *ref;
534 struct extent_buffer *leaf;
535 u64 ref_objectid;
536 int ret;
537
538 key.objectid = bytenr;
539 key.type = BTRFS_EXTENT_REF_KEY;
540 key.offset = parent;
541
542 ret = btrfs_search_slot(trans, root, &key, path, del ? -1 : 0, 1);
543 if (ret < 0)
544 goto out;
545 if (ret > 0) {
546 ret = -ENOENT;
547 goto out;
548 }
549
550 leaf = path->nodes[0];
551 ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_ref);
552 ref_objectid = btrfs_ref_objectid(leaf, ref);
553 if (btrfs_ref_root(leaf, ref) != ref_root ||
554 btrfs_ref_generation(leaf, ref) != ref_generation ||
555 (ref_objectid != owner_objectid &&
556 ref_objectid != BTRFS_MULTIPLE_OBJECTIDS)) {
557 ret = -EIO;
558 WARN_ON(1);
559 goto out;
560 }
561 ret = 0;
562 out:
563 return ret;
564 }
565
566 static noinline int insert_extent_backref(struct btrfs_trans_handle *trans,
567 struct btrfs_root *root,
568 struct btrfs_path *path,
569 u64 bytenr, u64 parent,
570 u64 ref_root, u64 ref_generation,
571 u64 owner_objectid,
572 int refs_to_add)
573 {
574 struct btrfs_key key;
575 struct extent_buffer *leaf;
576 struct btrfs_extent_ref *ref;
577 u32 num_refs;
578 int ret;
579
580 key.objectid = bytenr;
581 key.type = BTRFS_EXTENT_REF_KEY;
582 key.offset = parent;
583
584 ret = btrfs_insert_empty_item(trans, root, path, &key, sizeof(*ref));
585 if (ret == 0) {
586 leaf = path->nodes[0];
587 ref = btrfs_item_ptr(leaf, path->slots[0],
588 struct btrfs_extent_ref);
589 btrfs_set_ref_root(leaf, ref, ref_root);
590 btrfs_set_ref_generation(leaf, ref, ref_generation);
591 btrfs_set_ref_objectid(leaf, ref, owner_objectid);
592 btrfs_set_ref_num_refs(leaf, ref, refs_to_add);
593 } else if (ret == -EEXIST) {
594 u64 existing_owner;
595
596 BUG_ON(owner_objectid < BTRFS_FIRST_FREE_OBJECTID);
597 leaf = path->nodes[0];
598 ref = btrfs_item_ptr(leaf, path->slots[0],
599 struct btrfs_extent_ref);
600 if (btrfs_ref_root(leaf, ref) != ref_root ||
601 btrfs_ref_generation(leaf, ref) != ref_generation) {
602 ret = -EIO;
603 WARN_ON(1);
604 goto out;
605 }
606
607 num_refs = btrfs_ref_num_refs(leaf, ref);
608 BUG_ON(num_refs == 0);
609 btrfs_set_ref_num_refs(leaf, ref, num_refs + refs_to_add);
610
611 existing_owner = btrfs_ref_objectid(leaf, ref);
612 if (existing_owner != owner_objectid &&
613 existing_owner != BTRFS_MULTIPLE_OBJECTIDS) {
614 btrfs_set_ref_objectid(leaf, ref,
615 BTRFS_MULTIPLE_OBJECTIDS);
616 }
617 ret = 0;
618 } else {
619 goto out;
620 }
621 btrfs_mark_buffer_dirty(path->nodes[0]);
622 out:
623 btrfs_release_path(root, path);
624 return ret;
625 }
626
627 static noinline int remove_extent_backref(struct btrfs_trans_handle *trans,
628 struct btrfs_root *root,
629 struct btrfs_path *path,
630 int refs_to_drop)
631 {
632 struct extent_buffer *leaf;
633 struct btrfs_extent_ref *ref;
634 u32 num_refs;
635 int ret = 0;
636
637 leaf = path->nodes[0];
638 ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_ref);
639 num_refs = btrfs_ref_num_refs(leaf, ref);
640 BUG_ON(num_refs < refs_to_drop);
641 num_refs -= refs_to_drop;
642 if (num_refs == 0) {
643 ret = btrfs_del_item(trans, root, path);
644 } else {
645 btrfs_set_ref_num_refs(leaf, ref, num_refs);
646 btrfs_mark_buffer_dirty(leaf);
647 }
648 btrfs_release_path(root, path);
649 return ret;
650 }
651
652 #ifdef BIO_RW_DISCARD
653 static void btrfs_issue_discard(struct block_device *bdev,
654 u64 start, u64 len)
655 {
656 blkdev_issue_discard(bdev, start >> 9, len >> 9, GFP_KERNEL);
657 }
658 #endif
659
660 static int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr,
661 u64 num_bytes)
662 {
663 #ifdef BIO_RW_DISCARD
664 int ret;
665 u64 map_length = num_bytes;
666 struct btrfs_multi_bio *multi = NULL;
667
668 /* Tell the block device(s) that the sectors can be discarded */
669 ret = btrfs_map_block(&root->fs_info->mapping_tree, READ,
670 bytenr, &map_length, &multi, 0);
671 if (!ret) {
672 struct btrfs_bio_stripe *stripe = multi->stripes;
673 int i;
674
675 if (map_length > num_bytes)
676 map_length = num_bytes;
677
678 for (i = 0; i < multi->num_stripes; i++, stripe++) {
679 btrfs_issue_discard(stripe->dev->bdev,
680 stripe->physical,
681 map_length);
682 }
683 kfree(multi);
684 }
685
686 return ret;
687 #else
688 return 0;
689 #endif
690 }
691
692 static int __btrfs_update_extent_ref(struct btrfs_trans_handle *trans,
693 struct btrfs_root *root, u64 bytenr,
694 u64 num_bytes,
695 u64 orig_parent, u64 parent,
696 u64 orig_root, u64 ref_root,
697 u64 orig_generation, u64 ref_generation,
698 u64 owner_objectid)
699 {
700 int ret;
701 int pin = owner_objectid < BTRFS_FIRST_FREE_OBJECTID;
702
703 ret = btrfs_update_delayed_ref(trans, bytenr, num_bytes,
704 orig_parent, parent, orig_root,
705 ref_root, orig_generation,
706 ref_generation, owner_objectid, pin);
707 BUG_ON(ret);
708 return ret;
709 }
710
711 int btrfs_update_extent_ref(struct btrfs_trans_handle *trans,
712 struct btrfs_root *root, u64 bytenr,
713 u64 num_bytes, u64 orig_parent, u64 parent,
714 u64 ref_root, u64 ref_generation,
715 u64 owner_objectid)
716 {
717 int ret;
718 if (ref_root == BTRFS_TREE_LOG_OBJECTID &&
719 owner_objectid < BTRFS_FIRST_FREE_OBJECTID)
720 return 0;
721
722 ret = __btrfs_update_extent_ref(trans, root, bytenr, num_bytes,
723 orig_parent, parent, ref_root,
724 ref_root, ref_generation,
725 ref_generation, owner_objectid);
726 return ret;
727 }
728 static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
729 struct btrfs_root *root, u64 bytenr,
730 u64 num_bytes,
731 u64 orig_parent, u64 parent,
732 u64 orig_root, u64 ref_root,
733 u64 orig_generation, u64 ref_generation,
734 u64 owner_objectid)
735 {
736 int ret;
737
738 ret = btrfs_add_delayed_ref(trans, bytenr, num_bytes, parent, ref_root,
739 ref_generation, owner_objectid,
740 BTRFS_ADD_DELAYED_REF, 0);
741 BUG_ON(ret);
742 return ret;
743 }
744
745 static noinline_for_stack int add_extent_ref(struct btrfs_trans_handle *trans,
746 struct btrfs_root *root, u64 bytenr,
747 u64 num_bytes, u64 parent, u64 ref_root,
748 u64 ref_generation, u64 owner_objectid,
749 int refs_to_add)
750 {
751 struct btrfs_path *path;
752 int ret;
753 struct btrfs_key key;
754 struct extent_buffer *l;
755 struct btrfs_extent_item *item;
756 u32 refs;
757
758 path = btrfs_alloc_path();
759 if (!path)
760 return -ENOMEM;
761
762 path->reada = 1;
763 key.objectid = bytenr;
764 key.type = BTRFS_EXTENT_ITEM_KEY;
765 key.offset = num_bytes;
766
767 /* first find the extent item and update its reference count */
768 ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key,
769 path, 0, 1);
770 if (ret < 0)
771 return ret;
772
773 if (ret > 0) {
774 WARN_ON(1);
775 btrfs_free_path(path);
776 return -EIO;
777 }
778 l = path->nodes[0];
779
780 btrfs_item_key_to_cpu(l, &key, path->slots[0]);
781 if (key.objectid != bytenr) {
782 btrfs_print_leaf(root->fs_info->extent_root, path->nodes[0]);
783 printk(KERN_ERR "btrfs wanted %llu found %llu\n",
784 (unsigned long long)bytenr,
785 (unsigned long long)key.objectid);
786 BUG();
787 }
788 BUG_ON(key.type != BTRFS_EXTENT_ITEM_KEY);
789
790 item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
791
792 refs = btrfs_extent_refs(l, item);
793 btrfs_set_extent_refs(l, item, refs + refs_to_add);
794 btrfs_mark_buffer_dirty(path->nodes[0]);
795
796 btrfs_release_path(root->fs_info->extent_root, path);
797
798 path->reada = 1;
799 /* now insert the actual backref */
800 ret = insert_extent_backref(trans, root->fs_info->extent_root,
801 path, bytenr, parent,
802 ref_root, ref_generation,
803 owner_objectid, refs_to_add);
804 BUG_ON(ret);
805 btrfs_free_path(path);
806 return 0;
807 }
808
809 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
810 struct btrfs_root *root,
811 u64 bytenr, u64 num_bytes, u64 parent,
812 u64 ref_root, u64 ref_generation,
813 u64 owner_objectid)
814 {
815 int ret;
816 if (ref_root == BTRFS_TREE_LOG_OBJECTID &&
817 owner_objectid < BTRFS_FIRST_FREE_OBJECTID)
818 return 0;
819
820 ret = __btrfs_inc_extent_ref(trans, root, bytenr, num_bytes, 0, parent,
821 0, ref_root, 0, ref_generation,
822 owner_objectid);
823 return ret;
824 }
825
826 static int drop_delayed_ref(struct btrfs_trans_handle *trans,
827 struct btrfs_root *root,
828 struct btrfs_delayed_ref_node *node)
829 {
830 int ret = 0;
831 struct btrfs_delayed_ref *ref = btrfs_delayed_node_to_ref(node);
832
833 BUG_ON(node->ref_mod == 0);
834 ret = __btrfs_free_extent(trans, root, node->bytenr, node->num_bytes,
835 node->parent, ref->root, ref->generation,
836 ref->owner_objectid, ref->pin, node->ref_mod);
837
838 return ret;
839 }
840
841 /* helper function to actually process a single delayed ref entry */
842 static noinline int run_one_delayed_ref(struct btrfs_trans_handle *trans,
843 struct btrfs_root *root,
844 struct btrfs_delayed_ref_node *node,
845 int insert_reserved)
846 {
847 int ret;
848 struct btrfs_delayed_ref *ref;
849
850 if (node->parent == (u64)-1) {
851 struct btrfs_delayed_ref_head *head;
852 /*
853 * we've hit the end of the chain and we were supposed
854 * to insert this extent into the tree. But, it got
855 * deleted before we ever needed to insert it, so all
856 * we have to do is clean up the accounting
857 */
858 if (insert_reserved) {
859 update_reserved_extents(root, node->bytenr,
860 node->num_bytes, 0);
861 }
862 head = btrfs_delayed_node_to_head(node);
863 mutex_unlock(&head->mutex);
864 return 0;
865 }
866
867 ref = btrfs_delayed_node_to_ref(node);
868 if (ref->action == BTRFS_ADD_DELAYED_REF) {
869 if (insert_reserved) {
870 struct btrfs_key ins;
871
872 ins.objectid = node->bytenr;
873 ins.offset = node->num_bytes;
874 ins.type = BTRFS_EXTENT_ITEM_KEY;
875
876 /* record the full extent allocation */
877 ret = __btrfs_alloc_reserved_extent(trans, root,
878 node->parent, ref->root,
879 ref->generation, ref->owner_objectid,
880 &ins, node->ref_mod);
881 update_reserved_extents(root, node->bytenr,
882 node->num_bytes, 0);
883 } else {
884 /* just add one backref */
885 ret = add_extent_ref(trans, root, node->bytenr,
886 node->num_bytes,
887 node->parent, ref->root, ref->generation,
888 ref->owner_objectid, node->ref_mod);
889 }
890 BUG_ON(ret);
891 } else if (ref->action == BTRFS_DROP_DELAYED_REF) {
892 WARN_ON(insert_reserved);
893 ret = drop_delayed_ref(trans, root, node);
894 }
895 return 0;
896 }
897
898 static noinline struct btrfs_delayed_ref_node *
899 select_delayed_ref(struct btrfs_delayed_ref_head *head)
900 {
901 struct rb_node *node;
902 struct btrfs_delayed_ref_node *ref;
903 int action = BTRFS_ADD_DELAYED_REF;
904 again:
905 /*
906 * select delayed ref of type BTRFS_ADD_DELAYED_REF first.
907 * this prevents ref count from going down to zero when
908 * there still are pending delayed ref.
909 */
910 node = rb_prev(&head->node.rb_node);
911 while (1) {
912 if (!node)
913 break;
914 ref = rb_entry(node, struct btrfs_delayed_ref_node,
915 rb_node);
916 if (ref->bytenr != head->node.bytenr)
917 break;
918 if (btrfs_delayed_node_to_ref(ref)->action == action)
919 return ref;
920 node = rb_prev(node);
921 }
922 if (action == BTRFS_ADD_DELAYED_REF) {
923 action = BTRFS_DROP_DELAYED_REF;
924 goto again;
925 }
926 return NULL;
927 }
928
929 /*
930 * this starts processing the delayed reference count updates and
931 * extent insertions we have queued up so far. count can be
932 * 0, which means to process everything in the tree at the start
933 * of the run (but not newly added entries), or it can be some target
934 * number you'd like to process.
935 */
936 int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
937 struct btrfs_root *root, unsigned long count)
938 {
939 struct rb_node *node;
940 struct btrfs_delayed_ref_root *delayed_refs;
941 struct btrfs_delayed_ref_node *ref;
942 struct btrfs_delayed_ref_head *locked_ref = NULL;
943 int ret;
944 int must_insert_reserved = 0;
945 int run_all = count == (unsigned long)-1;
946
947 if (root == root->fs_info->extent_root)
948 root = root->fs_info->tree_root;
949
950 delayed_refs = &trans->transaction->delayed_refs;
951 again:
952 spin_lock(&delayed_refs->lock);
953 if (count == 0)
954 count = delayed_refs->num_entries;
955 while (1) {
956 if (!locked_ref) {
957 /*
958 * no locked ref, go find something we can
959 * process in the rbtree. We start at
960 * the beginning of the tree, there may be less
961 * lock contention if we do something smarter here.
962 */
963 node = rb_first(&delayed_refs->root);
964 if (!node) {
965 spin_unlock(&delayed_refs->lock);
966 break;
967 }
968
969 ref = rb_entry(node, struct btrfs_delayed_ref_node,
970 rb_node);
971 ret = btrfs_lock_delayed_ref(trans, ref, &locked_ref);
972 if (ret) {
973 spin_unlock(&delayed_refs->lock);
974 break;
975 }
976 }
977
978 /*
979 * record the must insert reserved flag before we
980 * drop the spin lock.
981 */
982 must_insert_reserved = locked_ref->must_insert_reserved;
983 locked_ref->must_insert_reserved = 0;
984
985 /*
986 * locked_ref is the head node, so we have to go one
987 * node back for any delayed ref updates
988 */
989
990 ref = select_delayed_ref(locked_ref);
991 if (!ref) {
992 /* All delayed refs have been processed, Go ahead
993 * and send the head node to run_one_delayed_ref,
994 * so that any accounting fixes can happen
995 */
996 ref = &locked_ref->node;
997 locked_ref = NULL;
998 }
999
1000 ref->in_tree = 0;
1001 rb_erase(&ref->rb_node, &delayed_refs->root);
1002 delayed_refs->num_entries--;
1003 spin_unlock(&delayed_refs->lock);
1004
1005 ret = run_one_delayed_ref(trans, root, ref,
1006 must_insert_reserved);
1007 BUG_ON(ret);
1008 btrfs_put_delayed_ref(ref);
1009
1010 /* once we lock the head ref, we have to process all the
1011 * entries for it. So, we might end up doing more entries
1012 * that count was asking us to do.
1013 */
1014 if (count > 0)
1015 count--;
1016
1017 /*
1018 * we set locked_ref to null above if we're all done
1019 * with this bytenr
1020 */
1021 if (!locked_ref && count == 0)
1022 break;
1023
1024 spin_lock(&delayed_refs->lock);
1025 }
1026 if (run_all) {
1027 spin_lock(&delayed_refs->lock);
1028 node = rb_first(&delayed_refs->root);
1029 if (!node) {
1030 spin_unlock(&delayed_refs->lock);
1031 goto out;
1032 }
1033
1034 while (node) {
1035 ref = rb_entry(node, struct btrfs_delayed_ref_node,
1036 rb_node);
1037 if (btrfs_delayed_ref_is_head(ref)) {
1038 struct btrfs_delayed_ref_head *head;
1039
1040 head = btrfs_delayed_node_to_head(ref);
1041 atomic_inc(&ref->refs);
1042
1043 spin_unlock(&delayed_refs->lock);
1044 mutex_lock(&head->mutex);
1045 mutex_unlock(&head->mutex);
1046
1047 btrfs_put_delayed_ref(ref);
1048 goto again;
1049 }
1050 node = rb_next(node);
1051 }
1052 spin_unlock(&delayed_refs->lock);
1053 count = (unsigned long)-1;
1054 schedule_timeout(1);
1055 goto again;
1056 }
1057 out:
1058 return 0;
1059 }
1060
1061 int btrfs_cross_ref_exist(struct btrfs_trans_handle *trans,
1062 struct btrfs_root *root, u64 objectid, u64 bytenr)
1063 {
1064 struct btrfs_root *extent_root = root->fs_info->extent_root;
1065 struct btrfs_path *path;
1066 struct extent_buffer *leaf;
1067 struct btrfs_extent_ref *ref_item;
1068 struct btrfs_key key;
1069 struct btrfs_key found_key;
1070 u64 ref_root;
1071 u64 last_snapshot;
1072 u32 nritems;
1073 int ret;
1074
1075 key.objectid = bytenr;
1076 key.offset = (u64)-1;
1077 key.type = BTRFS_EXTENT_ITEM_KEY;
1078
1079 path = btrfs_alloc_path();
1080 ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
1081 if (ret < 0)
1082 goto out;
1083 BUG_ON(ret == 0);
1084
1085 ret = -ENOENT;
1086 if (path->slots[0] == 0)
1087 goto out;
1088
1089 path->slots[0]--;
1090 leaf = path->nodes[0];
1091 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1092
1093 if (found_key.objectid != bytenr ||
1094 found_key.type != BTRFS_EXTENT_ITEM_KEY)
1095 goto out;
1096
1097 last_snapshot = btrfs_root_last_snapshot(&root->root_item);
1098 while (1) {
1099 leaf = path->nodes[0];
1100 nritems = btrfs_header_nritems(leaf);
1101 if (path->slots[0] >= nritems) {
1102 ret = btrfs_next_leaf(extent_root, path);
1103 if (ret < 0)
1104 goto out;
1105 if (ret == 0)
1106 continue;
1107 break;
1108 }
1109 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1110 if (found_key.objectid != bytenr)
1111 break;
1112
1113 if (found_key.type != BTRFS_EXTENT_REF_KEY) {
1114 path->slots[0]++;
1115 continue;
1116 }
1117
1118 ref_item = btrfs_item_ptr(leaf, path->slots[0],
1119 struct btrfs_extent_ref);
1120 ref_root = btrfs_ref_root(leaf, ref_item);
1121 if ((ref_root != root->root_key.objectid &&
1122 ref_root != BTRFS_TREE_LOG_OBJECTID) ||
1123 objectid != btrfs_ref_objectid(leaf, ref_item)) {
1124 ret = 1;
1125 goto out;
1126 }
1127 if (btrfs_ref_generation(leaf, ref_item) <= last_snapshot) {
1128 ret = 1;
1129 goto out;
1130 }
1131
1132 path->slots[0]++;
1133 }
1134 ret = 0;
1135 out:
1136 btrfs_free_path(path);
1137 return ret;
1138 }
1139
1140 int btrfs_cache_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1141 struct extent_buffer *buf, u32 nr_extents)
1142 {
1143 struct btrfs_key key;
1144 struct btrfs_file_extent_item *fi;
1145 u64 root_gen;
1146 u32 nritems;
1147 int i;
1148 int level;
1149 int ret = 0;
1150 int shared = 0;
1151
1152 if (!root->ref_cows)
1153 return 0;
1154
1155 if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
1156 shared = 0;
1157 root_gen = root->root_key.offset;
1158 } else {
1159 shared = 1;
1160 root_gen = trans->transid - 1;
1161 }
1162
1163 level = btrfs_header_level(buf);
1164 nritems = btrfs_header_nritems(buf);
1165
1166 if (level == 0) {
1167 struct btrfs_leaf_ref *ref;
1168 struct btrfs_extent_info *info;
1169
1170 ref = btrfs_alloc_leaf_ref(root, nr_extents);
1171 if (!ref) {
1172 ret = -ENOMEM;
1173 goto out;
1174 }
1175
1176 ref->root_gen = root_gen;
1177 ref->bytenr = buf->start;
1178 ref->owner = btrfs_header_owner(buf);
1179 ref->generation = btrfs_header_generation(buf);
1180 ref->nritems = nr_extents;
1181 info = ref->extents;
1182
1183 for (i = 0; nr_extents > 0 && i < nritems; i++) {
1184 u64 disk_bytenr;
1185 btrfs_item_key_to_cpu(buf, &key, i);
1186 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
1187 continue;
1188 fi = btrfs_item_ptr(buf, i,
1189 struct btrfs_file_extent_item);
1190 if (btrfs_file_extent_type(buf, fi) ==
1191 BTRFS_FILE_EXTENT_INLINE)
1192 continue;
1193 disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
1194 if (disk_bytenr == 0)
1195 continue;
1196
1197 info->bytenr = disk_bytenr;
1198 info->num_bytes =
1199 btrfs_file_extent_disk_num_bytes(buf, fi);
1200 info->objectid = key.objectid;
1201 info->offset = key.offset;
1202 info++;
1203 }
1204
1205 ret = btrfs_add_leaf_ref(root, ref, shared);
1206 if (ret == -EEXIST && shared) {
1207 struct btrfs_leaf_ref *old;
1208 old = btrfs_lookup_leaf_ref(root, ref->bytenr);
1209 BUG_ON(!old);
1210 btrfs_remove_leaf_ref(root, old);
1211 btrfs_free_leaf_ref(root, old);
1212 ret = btrfs_add_leaf_ref(root, ref, shared);
1213 }
1214 WARN_ON(ret);
1215 btrfs_free_leaf_ref(root, ref);
1216 }
1217 out:
1218 return ret;
1219 }
1220
1221 /* when a block goes through cow, we update the reference counts of
1222 * everything that block points to. The internal pointers of the block
1223 * can be in just about any order, and it is likely to have clusters of
1224 * things that are close together and clusters of things that are not.
1225 *
1226 * To help reduce the seeks that come with updating all of these reference
1227 * counts, sort them by byte number before actual updates are done.
1228 *
1229 * struct refsort is used to match byte number to slot in the btree block.
1230 * we sort based on the byte number and then use the slot to actually
1231 * find the item.
1232 *
1233 * struct refsort is smaller than strcut btrfs_item and smaller than
1234 * struct btrfs_key_ptr. Since we're currently limited to the page size
1235 * for a btree block, there's no way for a kmalloc of refsorts for a
1236 * single node to be bigger than a page.
1237 */
1238 struct refsort {
1239 u64 bytenr;
1240 u32 slot;
1241 };
1242
1243 /*
1244 * for passing into sort()
1245 */
1246 static int refsort_cmp(const void *a_void, const void *b_void)
1247 {
1248 const struct refsort *a = a_void;
1249 const struct refsort *b = b_void;
1250
1251 if (a->bytenr < b->bytenr)
1252 return -1;
1253 if (a->bytenr > b->bytenr)
1254 return 1;
1255 return 0;
1256 }
1257
1258
1259 noinline int btrfs_inc_ref(struct btrfs_trans_handle *trans,
1260 struct btrfs_root *root,
1261 struct extent_buffer *orig_buf,
1262 struct extent_buffer *buf, u32 *nr_extents)
1263 {
1264 u64 bytenr;
1265 u64 ref_root;
1266 u64 orig_root;
1267 u64 ref_generation;
1268 u64 orig_generation;
1269 struct refsort *sorted;
1270 u32 nritems;
1271 u32 nr_file_extents = 0;
1272 struct btrfs_key key;
1273 struct btrfs_file_extent_item *fi;
1274 int i;
1275 int level;
1276 int ret = 0;
1277 int faili = 0;
1278 int refi = 0;
1279 int slot;
1280 int (*process_func)(struct btrfs_trans_handle *, struct btrfs_root *,
1281 u64, u64, u64, u64, u64, u64, u64, u64, u64);
1282
1283 ref_root = btrfs_header_owner(buf);
1284 ref_generation = btrfs_header_generation(buf);
1285 orig_root = btrfs_header_owner(orig_buf);
1286 orig_generation = btrfs_header_generation(orig_buf);
1287
1288 nritems = btrfs_header_nritems(buf);
1289 level = btrfs_header_level(buf);
1290
1291 sorted = kmalloc(sizeof(struct refsort) * nritems, GFP_NOFS);
1292 BUG_ON(!sorted);
1293
1294 if (root->ref_cows) {
1295 process_func = __btrfs_inc_extent_ref;
1296 } else {
1297 if (level == 0 &&
1298 root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID)
1299 goto out;
1300 if (level != 0 &&
1301 root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID)
1302 goto out;
1303 process_func = __btrfs_update_extent_ref;
1304 }
1305
1306 /*
1307 * we make two passes through the items. In the first pass we
1308 * only record the byte number and slot. Then we sort based on
1309 * byte number and do the actual work based on the sorted results
1310 */
1311 for (i = 0; i < nritems; i++) {
1312 cond_resched();
1313 if (level == 0) {
1314 btrfs_item_key_to_cpu(buf, &key, i);
1315 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
1316 continue;
1317 fi = btrfs_item_ptr(buf, i,
1318 struct btrfs_file_extent_item);
1319 if (btrfs_file_extent_type(buf, fi) ==
1320 BTRFS_FILE_EXTENT_INLINE)
1321 continue;
1322 bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
1323 if (bytenr == 0)
1324 continue;
1325
1326 nr_file_extents++;
1327 sorted[refi].bytenr = bytenr;
1328 sorted[refi].slot = i;
1329 refi++;
1330 } else {
1331 bytenr = btrfs_node_blockptr(buf, i);
1332 sorted[refi].bytenr = bytenr;
1333 sorted[refi].slot = i;
1334 refi++;
1335 }
1336 }
1337 /*
1338 * if refi == 0, we didn't actually put anything into the sorted
1339 * array and we're done
1340 */
1341 if (refi == 0)
1342 goto out;
1343
1344 sort(sorted, refi, sizeof(struct refsort), refsort_cmp, NULL);
1345
1346 for (i = 0; i < refi; i++) {
1347 cond_resched();
1348 slot = sorted[i].slot;
1349 bytenr = sorted[i].bytenr;
1350
1351 if (level == 0) {
1352 btrfs_item_key_to_cpu(buf, &key, slot);
1353 fi = btrfs_item_ptr(buf, slot,
1354 struct btrfs_file_extent_item);
1355
1356 bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
1357 if (bytenr == 0)
1358 continue;
1359
1360 ret = process_func(trans, root, bytenr,
1361 btrfs_file_extent_disk_num_bytes(buf, fi),
1362 orig_buf->start, buf->start,
1363 orig_root, ref_root,
1364 orig_generation, ref_generation,
1365 key.objectid);
1366
1367 if (ret) {
1368 faili = slot;
1369 WARN_ON(1);
1370 goto fail;
1371 }
1372 } else {
1373 ret = process_func(trans, root, bytenr, buf->len,
1374 orig_buf->start, buf->start,
1375 orig_root, ref_root,
1376 orig_generation, ref_generation,
1377 level - 1);
1378 if (ret) {
1379 faili = slot;
1380 WARN_ON(1);
1381 goto fail;
1382 }
1383 }
1384 }
1385 out:
1386 kfree(sorted);
1387 if (nr_extents) {
1388 if (level == 0)
1389 *nr_extents = nr_file_extents;
1390 else
1391 *nr_extents = nritems;
1392 }
1393 return 0;
1394 fail:
1395 kfree(sorted);
1396 WARN_ON(1);
1397 return ret;
1398 }
1399
1400 int btrfs_update_ref(struct btrfs_trans_handle *trans,
1401 struct btrfs_root *root, struct extent_buffer *orig_buf,
1402 struct extent_buffer *buf, int start_slot, int nr)
1403
1404 {
1405 u64 bytenr;
1406 u64 ref_root;
1407 u64 orig_root;
1408 u64 ref_generation;
1409 u64 orig_generation;
1410 struct btrfs_key key;
1411 struct btrfs_file_extent_item *fi;
1412 int i;
1413 int ret;
1414 int slot;
1415 int level;
1416
1417 BUG_ON(start_slot < 0);
1418 BUG_ON(start_slot + nr > btrfs_header_nritems(buf));
1419
1420 ref_root = btrfs_header_owner(buf);
1421 ref_generation = btrfs_header_generation(buf);
1422 orig_root = btrfs_header_owner(orig_buf);
1423 orig_generation = btrfs_header_generation(orig_buf);
1424 level = btrfs_header_level(buf);
1425
1426 if (!root->ref_cows) {
1427 if (level == 0 &&
1428 root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID)
1429 return 0;
1430 if (level != 0 &&
1431 root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID)
1432 return 0;
1433 }
1434
1435 for (i = 0, slot = start_slot; i < nr; i++, slot++) {
1436 cond_resched();
1437 if (level == 0) {
1438 btrfs_item_key_to_cpu(buf, &key, slot);
1439 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
1440 continue;
1441 fi = btrfs_item_ptr(buf, slot,
1442 struct btrfs_file_extent_item);
1443 if (btrfs_file_extent_type(buf, fi) ==
1444 BTRFS_FILE_EXTENT_INLINE)
1445 continue;
1446 bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
1447 if (bytenr == 0)
1448 continue;
1449 ret = __btrfs_update_extent_ref(trans, root, bytenr,
1450 btrfs_file_extent_disk_num_bytes(buf, fi),
1451 orig_buf->start, buf->start,
1452 orig_root, ref_root, orig_generation,
1453 ref_generation, key.objectid);
1454 if (ret)
1455 goto fail;
1456 } else {
1457 bytenr = btrfs_node_blockptr(buf, slot);
1458 ret = __btrfs_update_extent_ref(trans, root, bytenr,
1459 buf->len, orig_buf->start,
1460 buf->start, orig_root, ref_root,
1461 orig_generation, ref_generation,
1462 level - 1);
1463 if (ret)
1464 goto fail;
1465 }
1466 }
1467 return 0;
1468 fail:
1469 WARN_ON(1);
1470 return -1;
1471 }
1472
1473 static int write_one_cache_group(struct btrfs_trans_handle *trans,
1474 struct btrfs_root *root,
1475 struct btrfs_path *path,
1476 struct btrfs_block_group_cache *cache)
1477 {
1478 int ret;
1479 struct btrfs_root *extent_root = root->fs_info->extent_root;
1480 unsigned long bi;
1481 struct extent_buffer *leaf;
1482
1483 ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1);
1484 if (ret < 0)
1485 goto fail;
1486 BUG_ON(ret);
1487
1488 leaf = path->nodes[0];
1489 bi = btrfs_item_ptr_offset(leaf, path->slots[0]);
1490 write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item));
1491 btrfs_mark_buffer_dirty(leaf);
1492 btrfs_release_path(extent_root, path);
1493 fail:
1494 if (ret)
1495 return ret;
1496 return 0;
1497
1498 }
1499
1500 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
1501 struct btrfs_root *root)
1502 {
1503 struct btrfs_block_group_cache *cache, *entry;
1504 struct rb_node *n;
1505 int err = 0;
1506 int werr = 0;
1507 struct btrfs_path *path;
1508 u64 last = 0;
1509
1510 path = btrfs_alloc_path();
1511 if (!path)
1512 return -ENOMEM;
1513
1514 while (1) {
1515 cache = NULL;
1516 spin_lock(&root->fs_info->block_group_cache_lock);
1517 for (n = rb_first(&root->fs_info->block_group_cache_tree);
1518 n; n = rb_next(n)) {
1519 entry = rb_entry(n, struct btrfs_block_group_cache,
1520 cache_node);
1521 if (entry->dirty) {
1522 cache = entry;
1523 break;
1524 }
1525 }
1526 spin_unlock(&root->fs_info->block_group_cache_lock);
1527
1528 if (!cache)
1529 break;
1530
1531 cache->dirty = 0;
1532 last += cache->key.offset;
1533
1534 err = write_one_cache_group(trans, root,
1535 path, cache);
1536 /*
1537 * if we fail to write the cache group, we want
1538 * to keep it marked dirty in hopes that a later
1539 * write will work
1540 */
1541 if (err) {
1542 werr = err;
1543 continue;
1544 }
1545 }
1546 btrfs_free_path(path);
1547 return werr;
1548 }
1549
1550 int btrfs_extent_readonly(struct btrfs_root *root, u64 bytenr)
1551 {
1552 struct btrfs_block_group_cache *block_group;
1553 int readonly = 0;
1554
1555 block_group = btrfs_lookup_block_group(root->fs_info, bytenr);
1556 if (!block_group || block_group->ro)
1557 readonly = 1;
1558 if (block_group)
1559 put_block_group(block_group);
1560 return readonly;
1561 }
1562
1563 static int update_space_info(struct btrfs_fs_info *info, u64 flags,
1564 u64 total_bytes, u64 bytes_used,
1565 struct btrfs_space_info **space_info)
1566 {
1567 struct btrfs_space_info *found;
1568
1569 found = __find_space_info(info, flags);
1570 if (found) {
1571 spin_lock(&found->lock);
1572 found->total_bytes += total_bytes;
1573 found->bytes_used += bytes_used;
1574 found->full = 0;
1575 spin_unlock(&found->lock);
1576 *space_info = found;
1577 return 0;
1578 }
1579 found = kzalloc(sizeof(*found), GFP_NOFS);
1580 if (!found)
1581 return -ENOMEM;
1582
1583 INIT_LIST_HEAD(&found->block_groups);
1584 init_rwsem(&found->groups_sem);
1585 spin_lock_init(&found->lock);
1586 found->flags = flags;
1587 found->total_bytes = total_bytes;
1588 found->bytes_used = bytes_used;
1589 found->bytes_pinned = 0;
1590 found->bytes_reserved = 0;
1591 found->bytes_readonly = 0;
1592 found->bytes_delalloc = 0;
1593 found->full = 0;
1594 found->force_alloc = 0;
1595 *space_info = found;
1596 list_add_rcu(&found->list, &info->space_info);
1597 return 0;
1598 }
1599
1600 static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags)
1601 {
1602 u64 extra_flags = flags & (BTRFS_BLOCK_GROUP_RAID0 |
1603 BTRFS_BLOCK_GROUP_RAID1 |
1604 BTRFS_BLOCK_GROUP_RAID10 |
1605 BTRFS_BLOCK_GROUP_DUP);
1606 if (extra_flags) {
1607 if (flags & BTRFS_BLOCK_GROUP_DATA)
1608 fs_info->avail_data_alloc_bits |= extra_flags;
1609 if (flags & BTRFS_BLOCK_GROUP_METADATA)
1610 fs_info->avail_metadata_alloc_bits |= extra_flags;
1611 if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
1612 fs_info->avail_system_alloc_bits |= extra_flags;
1613 }
1614 }
1615
1616 static void set_block_group_readonly(struct btrfs_block_group_cache *cache)
1617 {
1618 spin_lock(&cache->space_info->lock);
1619 spin_lock(&cache->lock);
1620 if (!cache->ro) {
1621 cache->space_info->bytes_readonly += cache->key.offset -
1622 btrfs_block_group_used(&cache->item);
1623 cache->ro = 1;
1624 }
1625 spin_unlock(&cache->lock);
1626 spin_unlock(&cache->space_info->lock);
1627 }
1628
1629 u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags)
1630 {
1631 u64 num_devices = root->fs_info->fs_devices->rw_devices;
1632
1633 if (num_devices == 1)
1634 flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0);
1635 if (num_devices < 4)
1636 flags &= ~BTRFS_BLOCK_GROUP_RAID10;
1637
1638 if ((flags & BTRFS_BLOCK_GROUP_DUP) &&
1639 (flags & (BTRFS_BLOCK_GROUP_RAID1 |
1640 BTRFS_BLOCK_GROUP_RAID10))) {
1641 flags &= ~BTRFS_BLOCK_GROUP_DUP;
1642 }
1643
1644 if ((flags & BTRFS_BLOCK_GROUP_RAID1) &&
1645 (flags & BTRFS_BLOCK_GROUP_RAID10)) {
1646 flags &= ~BTRFS_BLOCK_GROUP_RAID1;
1647 }
1648
1649 if ((flags & BTRFS_BLOCK_GROUP_RAID0) &&
1650 ((flags & BTRFS_BLOCK_GROUP_RAID1) |
1651 (flags & BTRFS_BLOCK_GROUP_RAID10) |
1652 (flags & BTRFS_BLOCK_GROUP_DUP)))
1653 flags &= ~BTRFS_BLOCK_GROUP_RAID0;
1654 return flags;
1655 }
1656
1657 static u64 btrfs_get_alloc_profile(struct btrfs_root *root, u64 data)
1658 {
1659 struct btrfs_fs_info *info = root->fs_info;
1660 u64 alloc_profile;
1661
1662 if (data) {
1663 alloc_profile = info->avail_data_alloc_bits &
1664 info->data_alloc_profile;
1665 data = BTRFS_BLOCK_GROUP_DATA | alloc_profile;
1666 } else if (root == root->fs_info->chunk_root) {
1667 alloc_profile = info->avail_system_alloc_bits &
1668 info->system_alloc_profile;
1669 data = BTRFS_BLOCK_GROUP_SYSTEM | alloc_profile;
1670 } else {
1671 alloc_profile = info->avail_metadata_alloc_bits &
1672 info->metadata_alloc_profile;
1673 data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile;
1674 }
1675
1676 return btrfs_reduce_alloc_profile(root, data);
1677 }
1678
1679 void btrfs_set_inode_space_info(struct btrfs_root *root, struct inode *inode)
1680 {
1681 u64 alloc_target;
1682
1683 alloc_target = btrfs_get_alloc_profile(root, 1);
1684 BTRFS_I(inode)->space_info = __find_space_info(root->fs_info,
1685 alloc_target);
1686 }
1687
1688 /*
1689 * for now this just makes sure we have at least 5% of our metadata space free
1690 * for use.
1691 */
1692 int btrfs_check_metadata_free_space(struct btrfs_root *root)
1693 {
1694 struct btrfs_fs_info *info = root->fs_info;
1695 struct btrfs_space_info *meta_sinfo;
1696 u64 alloc_target, thresh;
1697 int committed = 0, ret;
1698
1699 /* get the space info for where the metadata will live */
1700 alloc_target = btrfs_get_alloc_profile(root, 0);
1701 meta_sinfo = __find_space_info(info, alloc_target);
1702
1703 again:
1704 spin_lock(&meta_sinfo->lock);
1705 if (!meta_sinfo->full)
1706 thresh = meta_sinfo->total_bytes * 80;
1707 else
1708 thresh = meta_sinfo->total_bytes * 95;
1709
1710 do_div(thresh, 100);
1711
1712 if (meta_sinfo->bytes_used + meta_sinfo->bytes_reserved +
1713 meta_sinfo->bytes_pinned + meta_sinfo->bytes_readonly > thresh) {
1714 struct btrfs_trans_handle *trans;
1715 if (!meta_sinfo->full) {
1716 meta_sinfo->force_alloc = 1;
1717 spin_unlock(&meta_sinfo->lock);
1718
1719 trans = btrfs_start_transaction(root, 1);
1720 if (!trans)
1721 return -ENOMEM;
1722
1723 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
1724 2 * 1024 * 1024, alloc_target, 0);
1725 btrfs_end_transaction(trans, root);
1726 goto again;
1727 }
1728 spin_unlock(&meta_sinfo->lock);
1729
1730 if (!committed) {
1731 committed = 1;
1732 trans = btrfs_join_transaction(root, 1);
1733 if (!trans)
1734 return -ENOMEM;
1735 ret = btrfs_commit_transaction(trans, root);
1736 if (ret)
1737 return ret;
1738 goto again;
1739 }
1740 return -ENOSPC;
1741 }
1742 spin_unlock(&meta_sinfo->lock);
1743
1744 return 0;
1745 }
1746
1747 /*
1748 * This will check the space that the inode allocates from to make sure we have
1749 * enough space for bytes.
1750 */
1751 int btrfs_check_data_free_space(struct btrfs_root *root, struct inode *inode,
1752 u64 bytes)
1753 {
1754 struct btrfs_space_info *data_sinfo;
1755 int ret = 0, committed = 0;
1756
1757 /* make sure bytes are sectorsize aligned */
1758 bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
1759
1760 data_sinfo = BTRFS_I(inode)->space_info;
1761 again:
1762 /* make sure we have enough space to handle the data first */
1763 spin_lock(&data_sinfo->lock);
1764 if (data_sinfo->total_bytes - data_sinfo->bytes_used -
1765 data_sinfo->bytes_delalloc - data_sinfo->bytes_reserved -
1766 data_sinfo->bytes_pinned - data_sinfo->bytes_readonly -
1767 data_sinfo->bytes_may_use < bytes) {
1768 struct btrfs_trans_handle *trans;
1769
1770 /*
1771 * if we don't have enough free bytes in this space then we need
1772 * to alloc a new chunk.
1773 */
1774 if (!data_sinfo->full) {
1775 u64 alloc_target;
1776
1777 data_sinfo->force_alloc = 1;
1778 spin_unlock(&data_sinfo->lock);
1779
1780 alloc_target = btrfs_get_alloc_profile(root, 1);
1781 trans = btrfs_start_transaction(root, 1);
1782 if (!trans)
1783 return -ENOMEM;
1784
1785 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
1786 bytes + 2 * 1024 * 1024,
1787 alloc_target, 0);
1788 btrfs_end_transaction(trans, root);
1789 if (ret)
1790 return ret;
1791 goto again;
1792 }
1793 spin_unlock(&data_sinfo->lock);
1794
1795 /* commit the current transaction and try again */
1796 if (!committed) {
1797 committed = 1;
1798 trans = btrfs_join_transaction(root, 1);
1799 if (!trans)
1800 return -ENOMEM;
1801 ret = btrfs_commit_transaction(trans, root);
1802 if (ret)
1803 return ret;
1804 goto again;
1805 }
1806
1807 printk(KERN_ERR "no space left, need %llu, %llu delalloc bytes"
1808 ", %llu bytes_used, %llu bytes_reserved, "
1809 "%llu bytes_pinned, %llu bytes_readonly, %llu may use"
1810 "%llu total\n", bytes, data_sinfo->bytes_delalloc,
1811 data_sinfo->bytes_used, data_sinfo->bytes_reserved,
1812 data_sinfo->bytes_pinned, data_sinfo->bytes_readonly,
1813 data_sinfo->bytes_may_use, data_sinfo->total_bytes);
1814 return -ENOSPC;
1815 }
1816 data_sinfo->bytes_may_use += bytes;
1817 BTRFS_I(inode)->reserved_bytes += bytes;
1818 spin_unlock(&data_sinfo->lock);
1819
1820 return btrfs_check_metadata_free_space(root);
1821 }
1822
1823 /*
1824 * if there was an error for whatever reason after calling
1825 * btrfs_check_data_free_space, call this so we can cleanup the counters.
1826 */
1827 void btrfs_free_reserved_data_space(struct btrfs_root *root,
1828 struct inode *inode, u64 bytes)
1829 {
1830 struct btrfs_space_info *data_sinfo;
1831
1832 /* make sure bytes are sectorsize aligned */
1833 bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
1834
1835 data_sinfo = BTRFS_I(inode)->space_info;
1836 spin_lock(&data_sinfo->lock);
1837 data_sinfo->bytes_may_use -= bytes;
1838 BTRFS_I(inode)->reserved_bytes -= bytes;
1839 spin_unlock(&data_sinfo->lock);
1840 }
1841
1842 /* called when we are adding a delalloc extent to the inode's io_tree */
1843 void btrfs_delalloc_reserve_space(struct btrfs_root *root, struct inode *inode,
1844 u64 bytes)
1845 {
1846 struct btrfs_space_info *data_sinfo;
1847
1848 /* get the space info for where this inode will be storing its data */
1849 data_sinfo = BTRFS_I(inode)->space_info;
1850
1851 /* make sure we have enough space to handle the data first */
1852 spin_lock(&data_sinfo->lock);
1853 data_sinfo->bytes_delalloc += bytes;
1854
1855 /*
1856 * we are adding a delalloc extent without calling
1857 * btrfs_check_data_free_space first. This happens on a weird
1858 * writepage condition, but shouldn't hurt our accounting
1859 */
1860 if (unlikely(bytes > BTRFS_I(inode)->reserved_bytes)) {
1861 data_sinfo->bytes_may_use -= BTRFS_I(inode)->reserved_bytes;
1862 BTRFS_I(inode)->reserved_bytes = 0;
1863 } else {
1864 data_sinfo->bytes_may_use -= bytes;
1865 BTRFS_I(inode)->reserved_bytes -= bytes;
1866 }
1867
1868 spin_unlock(&data_sinfo->lock);
1869 }
1870
1871 /* called when we are clearing an delalloc extent from the inode's io_tree */
1872 void btrfs_delalloc_free_space(struct btrfs_root *root, struct inode *inode,
1873 u64 bytes)
1874 {
1875 struct btrfs_space_info *info;
1876
1877 info = BTRFS_I(inode)->space_info;
1878
1879 spin_lock(&info->lock);
1880 info->bytes_delalloc -= bytes;
1881 spin_unlock(&info->lock);
1882 }
1883
1884 static int do_chunk_alloc(struct btrfs_trans_handle *trans,
1885 struct btrfs_root *extent_root, u64 alloc_bytes,
1886 u64 flags, int force)
1887 {
1888 struct btrfs_space_info *space_info;
1889 u64 thresh;
1890 int ret = 0;
1891
1892 mutex_lock(&extent_root->fs_info->chunk_mutex);
1893
1894 flags = btrfs_reduce_alloc_profile(extent_root, flags);
1895
1896 space_info = __find_space_info(extent_root->fs_info, flags);
1897 if (!space_info) {
1898 ret = update_space_info(extent_root->fs_info, flags,
1899 0, 0, &space_info);
1900 BUG_ON(ret);
1901 }
1902 BUG_ON(!space_info);
1903
1904 spin_lock(&space_info->lock);
1905 if (space_info->force_alloc) {
1906 force = 1;
1907 space_info->force_alloc = 0;
1908 }
1909 if (space_info->full) {
1910 spin_unlock(&space_info->lock);
1911 goto out;
1912 }
1913
1914 thresh = space_info->total_bytes - space_info->bytes_readonly;
1915 thresh = div_factor(thresh, 6);
1916 if (!force &&
1917 (space_info->bytes_used + space_info->bytes_pinned +
1918 space_info->bytes_reserved + alloc_bytes) < thresh) {
1919 spin_unlock(&space_info->lock);
1920 goto out;
1921 }
1922 spin_unlock(&space_info->lock);
1923
1924 ret = btrfs_alloc_chunk(trans, extent_root, flags);
1925 if (ret)
1926 space_info->full = 1;
1927 out:
1928 mutex_unlock(&extent_root->fs_info->chunk_mutex);
1929 return ret;
1930 }
1931
1932 static int update_block_group(struct btrfs_trans_handle *trans,
1933 struct btrfs_root *root,
1934 u64 bytenr, u64 num_bytes, int alloc,
1935 int mark_free)
1936 {
1937 struct btrfs_block_group_cache *cache;
1938 struct btrfs_fs_info *info = root->fs_info;
1939 u64 total = num_bytes;
1940 u64 old_val;
1941 u64 byte_in_group;
1942
1943 while (total) {
1944 cache = btrfs_lookup_block_group(info, bytenr);
1945 if (!cache)
1946 return -1;
1947 byte_in_group = bytenr - cache->key.objectid;
1948 WARN_ON(byte_in_group > cache->key.offset);
1949
1950 spin_lock(&cache->space_info->lock);
1951 spin_lock(&cache->lock);
1952 cache->dirty = 1;
1953 old_val = btrfs_block_group_used(&cache->item);
1954 num_bytes = min(total, cache->key.offset - byte_in_group);
1955 if (alloc) {
1956 old_val += num_bytes;
1957 cache->space_info->bytes_used += num_bytes;
1958 if (cache->ro)
1959 cache->space_info->bytes_readonly -= num_bytes;
1960 btrfs_set_block_group_used(&cache->item, old_val);
1961 spin_unlock(&cache->lock);
1962 spin_unlock(&cache->space_info->lock);
1963 } else {
1964 old_val -= num_bytes;
1965 cache->space_info->bytes_used -= num_bytes;
1966 if (cache->ro)
1967 cache->space_info->bytes_readonly += num_bytes;
1968 btrfs_set_block_group_used(&cache->item, old_val);
1969 spin_unlock(&cache->lock);
1970 spin_unlock(&cache->space_info->lock);
1971 if (mark_free) {
1972 int ret;
1973
1974 ret = btrfs_discard_extent(root, bytenr,
1975 num_bytes);
1976 WARN_ON(ret);
1977
1978 ret = btrfs_add_free_space(cache, bytenr,
1979 num_bytes);
1980 WARN_ON(ret);
1981 }
1982 }
1983 put_block_group(cache);
1984 total -= num_bytes;
1985 bytenr += num_bytes;
1986 }
1987 return 0;
1988 }
1989
1990 static u64 first_logical_byte(struct btrfs_root *root, u64 search_start)
1991 {
1992 struct btrfs_block_group_cache *cache;
1993 u64 bytenr;
1994
1995 cache = btrfs_lookup_first_block_group(root->fs_info, search_start);
1996 if (!cache)
1997 return 0;
1998
1999 bytenr = cache->key.objectid;
2000 put_block_group(cache);
2001
2002 return bytenr;
2003 }
2004
2005 int btrfs_update_pinned_extents(struct btrfs_root *root,
2006 u64 bytenr, u64 num, int pin)
2007 {
2008 u64 len;
2009 struct btrfs_block_group_cache *cache;
2010 struct btrfs_fs_info *fs_info = root->fs_info;
2011
2012 WARN_ON(!mutex_is_locked(&root->fs_info->pinned_mutex));
2013 if (pin) {
2014 set_extent_dirty(&fs_info->pinned_extents,
2015 bytenr, bytenr + num - 1, GFP_NOFS);
2016 } else {
2017 clear_extent_dirty(&fs_info->pinned_extents,
2018 bytenr, bytenr + num - 1, GFP_NOFS);
2019 }
2020 while (num > 0) {
2021 cache = btrfs_lookup_block_group(fs_info, bytenr);
2022 BUG_ON(!cache);
2023 len = min(num, cache->key.offset -
2024 (bytenr - cache->key.objectid));
2025 if (pin) {
2026 spin_lock(&cache->space_info->lock);
2027 spin_lock(&cache->lock);
2028 cache->pinned += len;
2029 cache->space_info->bytes_pinned += len;
2030 spin_unlock(&cache->lock);
2031 spin_unlock(&cache->space_info->lock);
2032 fs_info->total_pinned += len;
2033 } else {
2034 spin_lock(&cache->space_info->lock);
2035 spin_lock(&cache->lock);
2036 cache->pinned -= len;
2037 cache->space_info->bytes_pinned -= len;
2038 spin_unlock(&cache->lock);
2039 spin_unlock(&cache->space_info->lock);
2040 fs_info->total_pinned -= len;
2041 if (cache->cached)
2042 btrfs_add_free_space(cache, bytenr, len);
2043 }
2044 put_block_group(cache);
2045 bytenr += len;
2046 num -= len;
2047 }
2048 return 0;
2049 }
2050
2051 static int update_reserved_extents(struct btrfs_root *root,
2052 u64 bytenr, u64 num, int reserve)
2053 {
2054 u64 len;
2055 struct btrfs_block_group_cache *cache;
2056 struct btrfs_fs_info *fs_info = root->fs_info;
2057
2058 while (num > 0) {
2059 cache = btrfs_lookup_block_group(fs_info, bytenr);
2060 BUG_ON(!cache);
2061 len = min(num, cache->key.offset -
2062 (bytenr - cache->key.objectid));
2063
2064 spin_lock(&cache->space_info->lock);
2065 spin_lock(&cache->lock);
2066 if (reserve) {
2067 cache->reserved += len;
2068 cache->space_info->bytes_reserved += len;
2069 } else {
2070 cache->reserved -= len;
2071 cache->space_info->bytes_reserved -= len;
2072 }
2073 spin_unlock(&cache->lock);
2074 spin_unlock(&cache->space_info->lock);
2075 put_block_group(cache);
2076 bytenr += len;
2077 num -= len;
2078 }
2079 return 0;
2080 }
2081
2082 int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy)
2083 {
2084 u64 last = 0;
2085 u64 start;
2086 u64 end;
2087 struct extent_io_tree *pinned_extents = &root->fs_info->pinned_extents;
2088 int ret;
2089
2090 mutex_lock(&root->fs_info->pinned_mutex);
2091 while (1) {
2092 ret = find_first_extent_bit(pinned_extents, last,
2093 &start, &end, EXTENT_DIRTY);
2094 if (ret)
2095 break;
2096 set_extent_dirty(copy, start, end, GFP_NOFS);
2097 last = end + 1;
2098 }
2099 mutex_unlock(&root->fs_info->pinned_mutex);
2100 return 0;
2101 }
2102
2103 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
2104 struct btrfs_root *root,
2105 struct extent_io_tree *unpin)
2106 {
2107 u64 start;
2108 u64 end;
2109 int ret;
2110
2111 mutex_lock(&root->fs_info->pinned_mutex);
2112 while (1) {
2113 ret = find_first_extent_bit(unpin, 0, &start, &end,
2114 EXTENT_DIRTY);
2115 if (ret)
2116 break;
2117
2118 ret = btrfs_discard_extent(root, start, end + 1 - start);
2119
2120 btrfs_update_pinned_extents(root, start, end + 1 - start, 0);
2121 clear_extent_dirty(unpin, start, end, GFP_NOFS);
2122
2123 if (need_resched()) {
2124 mutex_unlock(&root->fs_info->pinned_mutex);
2125 cond_resched();
2126 mutex_lock(&root->fs_info->pinned_mutex);
2127 }
2128 }
2129 mutex_unlock(&root->fs_info->pinned_mutex);
2130 return ret;
2131 }
2132
2133 static int pin_down_bytes(struct btrfs_trans_handle *trans,
2134 struct btrfs_root *root,
2135 u64 bytenr, u64 num_bytes, int is_data)
2136 {
2137 int err = 0;
2138 struct extent_buffer *buf;
2139
2140 if (is_data)
2141 goto pinit;
2142
2143 buf = btrfs_find_tree_block(root, bytenr, num_bytes);
2144 if (!buf)
2145 goto pinit;
2146
2147 /* we can reuse a block if it hasn't been written
2148 * and it is from this transaction. We can't
2149 * reuse anything from the tree log root because
2150 * it has tiny sub-transactions.
2151 */
2152 if (btrfs_buffer_uptodate(buf, 0) &&
2153 btrfs_try_tree_lock(buf)) {
2154 u64 header_owner = btrfs_header_owner(buf);
2155 u64 header_transid = btrfs_header_generation(buf);
2156 if (header_owner != BTRFS_TREE_LOG_OBJECTID &&
2157 header_owner != BTRFS_TREE_RELOC_OBJECTID &&
2158 header_owner != BTRFS_DATA_RELOC_TREE_OBJECTID &&
2159 header_transid == trans->transid &&
2160 !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
2161 clean_tree_block(NULL, root, buf);
2162 btrfs_tree_unlock(buf);
2163 free_extent_buffer(buf);
2164 return 1;
2165 }
2166 btrfs_tree_unlock(buf);
2167 }
2168 free_extent_buffer(buf);
2169 pinit:
2170 btrfs_update_pinned_extents(root, bytenr, num_bytes, 1);
2171
2172 BUG_ON(err < 0);
2173 return 0;
2174 }
2175
2176 /*
2177 * remove an extent from the root, returns 0 on success
2178 */
2179 static int __free_extent(struct btrfs_trans_handle *trans,
2180 struct btrfs_root *root,
2181 u64 bytenr, u64 num_bytes, u64 parent,
2182 u64 root_objectid, u64 ref_generation,
2183 u64 owner_objectid, int pin, int mark_free,
2184 int refs_to_drop)
2185 {
2186 struct btrfs_path *path;
2187 struct btrfs_key key;
2188 struct btrfs_fs_info *info = root->fs_info;
2189 struct btrfs_root *extent_root = info->extent_root;
2190 struct extent_buffer *leaf;
2191 int ret;
2192 int extent_slot = 0;
2193 int found_extent = 0;
2194 int num_to_del = 1;
2195 struct btrfs_extent_item *ei;
2196 u32 refs;
2197
2198 key.objectid = bytenr;
2199 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
2200 key.offset = num_bytes;
2201 path = btrfs_alloc_path();
2202 if (!path)
2203 return -ENOMEM;
2204
2205 path->reada = 1;
2206 ret = lookup_extent_backref(trans, extent_root, path,
2207 bytenr, parent, root_objectid,
2208 ref_generation, owner_objectid, 1);
2209 if (ret == 0) {
2210 struct btrfs_key found_key;
2211 extent_slot = path->slots[0];
2212 while (extent_slot > 0) {
2213 extent_slot--;
2214 btrfs_item_key_to_cpu(path->nodes[0], &found_key,
2215 extent_slot);
2216 if (found_key.objectid != bytenr)
2217 break;
2218 if (found_key.type == BTRFS_EXTENT_ITEM_KEY &&
2219 found_key.offset == num_bytes) {
2220 found_extent = 1;
2221 break;
2222 }
2223 if (path->slots[0] - extent_slot > 5)
2224 break;
2225 }
2226 if (!found_extent) {
2227 ret = remove_extent_backref(trans, extent_root, path,
2228 refs_to_drop);
2229 BUG_ON(ret);
2230 btrfs_release_path(extent_root, path);
2231 ret = btrfs_search_slot(trans, extent_root,
2232 &key, path, -1, 1);
2233 if (ret) {
2234 printk(KERN_ERR "umm, got %d back from search"
2235 ", was looking for %llu\n", ret,
2236 (unsigned long long)bytenr);
2237 btrfs_print_leaf(extent_root, path->nodes[0]);
2238 }
2239 BUG_ON(ret);
2240 extent_slot = path->slots[0];
2241 }
2242 } else {
2243 btrfs_print_leaf(extent_root, path->nodes[0]);
2244 WARN_ON(1);
2245 printk(KERN_ERR "btrfs unable to find ref byte nr %llu "
2246 "parent %llu root %llu gen %llu owner %llu\n",
2247 (unsigned long long)bytenr,
2248 (unsigned long long)parent,
2249 (unsigned long long)root_objectid,
2250 (unsigned long long)ref_generation,
2251 (unsigned long long)owner_objectid);
2252 }
2253
2254 leaf = path->nodes[0];
2255 ei = btrfs_item_ptr(leaf, extent_slot,
2256 struct btrfs_extent_item);
2257 refs = btrfs_extent_refs(leaf, ei);
2258
2259 /*
2260 * we're not allowed to delete the extent item if there
2261 * are other delayed ref updates pending
2262 */
2263
2264 BUG_ON(refs < refs_to_drop);
2265 refs -= refs_to_drop;
2266 btrfs_set_extent_refs(leaf, ei, refs);
2267 btrfs_mark_buffer_dirty(leaf);
2268
2269 if (refs == 0 && found_extent &&
2270 path->slots[0] == extent_slot + 1) {
2271 struct btrfs_extent_ref *ref;
2272 ref = btrfs_item_ptr(leaf, path->slots[0],
2273 struct btrfs_extent_ref);
2274 BUG_ON(btrfs_ref_num_refs(leaf, ref) != refs_to_drop);
2275 /* if the back ref and the extent are next to each other
2276 * they get deleted below in one shot
2277 */
2278 path->slots[0] = extent_slot;
2279 num_to_del = 2;
2280 } else if (found_extent) {
2281 /* otherwise delete the extent back ref */
2282 ret = remove_extent_backref(trans, extent_root, path,
2283 refs_to_drop);
2284 BUG_ON(ret);
2285 /* if refs are 0, we need to setup the path for deletion */
2286 if (refs == 0) {
2287 btrfs_release_path(extent_root, path);
2288 ret = btrfs_search_slot(trans, extent_root, &key, path,
2289 -1, 1);
2290 BUG_ON(ret);
2291 }
2292 }
2293
2294 if (refs == 0) {
2295 u64 super_used;
2296 u64 root_used;
2297
2298 if (pin) {
2299 mutex_lock(&root->fs_info->pinned_mutex);
2300 ret = pin_down_bytes(trans, root, bytenr, num_bytes,
2301 owner_objectid >= BTRFS_FIRST_FREE_OBJECTID);
2302 mutex_unlock(&root->fs_info->pinned_mutex);
2303 if (ret > 0)
2304 mark_free = 1;
2305 BUG_ON(ret < 0);
2306 }
2307 /* block accounting for super block */
2308 spin_lock(&info->delalloc_lock);
2309 super_used = btrfs_super_bytes_used(&info->super_copy);
2310 btrfs_set_super_bytes_used(&info->super_copy,
2311 super_used - num_bytes);
2312
2313 /* block accounting for root item */
2314 root_used = btrfs_root_used(&root->root_item);
2315 btrfs_set_root_used(&root->root_item,
2316 root_used - num_bytes);
2317 spin_unlock(&info->delalloc_lock);
2318 ret = btrfs_del_items(trans, extent_root, path, path->slots[0],
2319 num_to_del);
2320 BUG_ON(ret);
2321 btrfs_release_path(extent_root, path);
2322
2323 if (owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
2324 ret = btrfs_del_csums(trans, root, bytenr, num_bytes);
2325 BUG_ON(ret);
2326 }
2327
2328 ret = update_block_group(trans, root, bytenr, num_bytes, 0,
2329 mark_free);
2330 BUG_ON(ret);
2331 }
2332 btrfs_free_path(path);
2333 return ret;
2334 }
2335
2336 /*
2337 * remove an extent from the root, returns 0 on success
2338 */
2339 static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
2340 struct btrfs_root *root,
2341 u64 bytenr, u64 num_bytes, u64 parent,
2342 u64 root_objectid, u64 ref_generation,
2343 u64 owner_objectid, int pin,
2344 int refs_to_drop)
2345 {
2346 WARN_ON(num_bytes < root->sectorsize);
2347
2348 /*
2349 * if metadata always pin
2350 * if data pin when any transaction has committed this
2351 */
2352 if (owner_objectid < BTRFS_FIRST_FREE_OBJECTID ||
2353 ref_generation != trans->transid)
2354 pin = 1;
2355
2356 if (ref_generation != trans->transid)
2357 pin = 1;
2358
2359 return __free_extent(trans, root, bytenr, num_bytes, parent,
2360 root_objectid, ref_generation,
2361 owner_objectid, pin, pin == 0, refs_to_drop);
2362 }
2363
2364 int btrfs_free_extent(struct btrfs_trans_handle *trans,
2365 struct btrfs_root *root,
2366 u64 bytenr, u64 num_bytes, u64 parent,
2367 u64 root_objectid, u64 ref_generation,
2368 u64 owner_objectid, int pin)
2369 {
2370 int ret;
2371
2372 /*
2373 * tree log blocks never actually go into the extent allocation
2374 * tree, just update pinning info and exit early.
2375 *
2376 * data extents referenced by the tree log do need to have
2377 * their reference counts bumped.
2378 */
2379 if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID &&
2380 owner_objectid < BTRFS_FIRST_FREE_OBJECTID) {
2381 mutex_lock(&root->fs_info->pinned_mutex);
2382 btrfs_update_pinned_extents(root, bytenr, num_bytes, 1);
2383 mutex_unlock(&root->fs_info->pinned_mutex);
2384 update_reserved_extents(root, bytenr, num_bytes, 0);
2385 ret = 0;
2386 } else {
2387 ret = btrfs_add_delayed_ref(trans, bytenr, num_bytes, parent,
2388 root_objectid, ref_generation,
2389 owner_objectid,
2390 BTRFS_DROP_DELAYED_REF, 1);
2391 }
2392 return ret;
2393 }
2394
2395 static u64 stripe_align(struct btrfs_root *root, u64 val)
2396 {
2397 u64 mask = ((u64)root->stripesize - 1);
2398 u64 ret = (val + mask) & ~mask;
2399 return ret;
2400 }
2401
2402 /*
2403 * walks the btree of allocated extents and find a hole of a given size.
2404 * The key ins is changed to record the hole:
2405 * ins->objectid == block start
2406 * ins->flags = BTRFS_EXTENT_ITEM_KEY
2407 * ins->offset == number of blocks
2408 * Any available blocks before search_start are skipped.
2409 */
2410 static noinline int find_free_extent(struct btrfs_trans_handle *trans,
2411 struct btrfs_root *orig_root,
2412 u64 num_bytes, u64 empty_size,
2413 u64 search_start, u64 search_end,
2414 u64 hint_byte, struct btrfs_key *ins,
2415 u64 exclude_start, u64 exclude_nr,
2416 int data)
2417 {
2418 int ret = 0;
2419 struct btrfs_root *root = orig_root->fs_info->extent_root;
2420 u64 total_needed = num_bytes;
2421 u64 *last_ptr = NULL;
2422 u64 last_wanted = 0;
2423 struct btrfs_block_group_cache *block_group = NULL;
2424 int chunk_alloc_done = 0;
2425 int empty_cluster = 2 * 1024 * 1024;
2426 int allowed_chunk_alloc = 0;
2427 struct list_head *head = NULL, *cur = NULL;
2428 int loop = 0;
2429 int extra_loop = 0;
2430 struct btrfs_space_info *space_info;
2431
2432 WARN_ON(num_bytes < root->sectorsize);
2433 btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
2434 ins->objectid = 0;
2435 ins->offset = 0;
2436
2437 if (orig_root->ref_cows || empty_size)
2438 allowed_chunk_alloc = 1;
2439
2440 if (data & BTRFS_BLOCK_GROUP_METADATA) {
2441 last_ptr = &root->fs_info->last_alloc;
2442 if (!btrfs_test_opt(root, SSD))
2443 empty_cluster = 64 * 1024;
2444 }
2445
2446 if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD))
2447 last_ptr = &root->fs_info->last_data_alloc;
2448
2449 if (last_ptr) {
2450 if (*last_ptr) {
2451 hint_byte = *last_ptr;
2452 last_wanted = *last_ptr;
2453 } else
2454 empty_size += empty_cluster;
2455 } else {
2456 empty_cluster = 0;
2457 }
2458 search_start = max(search_start, first_logical_byte(root, 0));
2459 search_start = max(search_start, hint_byte);
2460
2461 if (last_wanted && search_start != last_wanted) {
2462 last_wanted = 0;
2463 empty_size += empty_cluster;
2464 }
2465
2466 total_needed += empty_size;
2467 block_group = btrfs_lookup_block_group(root->fs_info, search_start);
2468 if (!block_group)
2469 block_group = btrfs_lookup_first_block_group(root->fs_info,
2470 search_start);
2471 space_info = __find_space_info(root->fs_info, data);
2472
2473 down_read(&space_info->groups_sem);
2474 while (1) {
2475 struct btrfs_free_space *free_space;
2476 /*
2477 * the only way this happens if our hint points to a block
2478 * group thats not of the proper type, while looping this
2479 * should never happen
2480 */
2481 if (empty_size)
2482 extra_loop = 1;
2483
2484 if (!block_group)
2485 goto new_group_no_lock;
2486
2487 if (unlikely(!block_group->cached)) {
2488 mutex_lock(&block_group->cache_mutex);
2489 ret = cache_block_group(root, block_group);
2490 mutex_unlock(&block_group->cache_mutex);
2491 if (ret)
2492 break;
2493 }
2494
2495 mutex_lock(&block_group->alloc_mutex);
2496 if (unlikely(!block_group_bits(block_group, data)))
2497 goto new_group;
2498
2499 if (unlikely(block_group->ro))
2500 goto new_group;
2501
2502 free_space = btrfs_find_free_space(block_group, search_start,
2503 total_needed);
2504 if (free_space) {
2505 u64 start = block_group->key.objectid;
2506 u64 end = block_group->key.objectid +
2507 block_group->key.offset;
2508
2509 search_start = stripe_align(root, free_space->offset);
2510
2511 /* move on to the next group */
2512 if (search_start + num_bytes >= search_end)
2513 goto new_group;
2514
2515 /* move on to the next group */
2516 if (search_start + num_bytes > end)
2517 goto new_group;
2518
2519 if (last_wanted && search_start != last_wanted) {
2520 total_needed += empty_cluster;
2521 empty_size += empty_cluster;
2522 last_wanted = 0;
2523 /*
2524 * if search_start is still in this block group
2525 * then we just re-search this block group
2526 */
2527 if (search_start >= start &&
2528 search_start < end) {
2529 mutex_unlock(&block_group->alloc_mutex);
2530 continue;
2531 }
2532
2533 /* else we go to the next block group */
2534 goto new_group;
2535 }
2536
2537 if (exclude_nr > 0 &&
2538 (search_start + num_bytes > exclude_start &&
2539 search_start < exclude_start + exclude_nr)) {
2540 search_start = exclude_start + exclude_nr;
2541 /*
2542 * if search_start is still in this block group
2543 * then we just re-search this block group
2544 */
2545 if (search_start >= start &&
2546 search_start < end) {
2547 mutex_unlock(&block_group->alloc_mutex);
2548 last_wanted = 0;
2549 continue;
2550 }
2551
2552 /* else we go to the next block group */
2553 goto new_group;
2554 }
2555
2556 ins->objectid = search_start;
2557 ins->offset = num_bytes;
2558
2559 btrfs_remove_free_space_lock(block_group, search_start,
2560 num_bytes);
2561 /* we are all good, lets return */
2562 mutex_unlock(&block_group->alloc_mutex);
2563 break;
2564 }
2565 new_group:
2566 mutex_unlock(&block_group->alloc_mutex);
2567 put_block_group(block_group);
2568 block_group = NULL;
2569 new_group_no_lock:
2570 /* don't try to compare new allocations against the
2571 * last allocation any more
2572 */
2573 last_wanted = 0;
2574
2575 /*
2576 * Here's how this works.
2577 * loop == 0: we were searching a block group via a hint
2578 * and didn't find anything, so we start at
2579 * the head of the block groups and keep searching
2580 * loop == 1: we're searching through all of the block groups
2581 * if we hit the head again we have searched
2582 * all of the block groups for this space and we
2583 * need to try and allocate, if we cant error out.
2584 * loop == 2: we allocated more space and are looping through
2585 * all of the block groups again.
2586 */
2587 if (loop == 0) {
2588 head = &space_info->block_groups;
2589 cur = head->next;
2590 loop++;
2591 } else if (loop == 1 && cur == head) {
2592 int keep_going;
2593
2594 /* at this point we give up on the empty_size
2595 * allocations and just try to allocate the min
2596 * space.
2597 *
2598 * The extra_loop field was set if an empty_size
2599 * allocation was attempted above, and if this
2600 * is try we need to try the loop again without
2601 * the additional empty_size.
2602 */
2603 total_needed -= empty_size;
2604 empty_size = 0;
2605 keep_going = extra_loop;
2606 loop++;
2607
2608 if (allowed_chunk_alloc && !chunk_alloc_done) {
2609 up_read(&space_info->groups_sem);
2610 ret = do_chunk_alloc(trans, root, num_bytes +
2611 2 * 1024 * 1024, data, 1);
2612 down_read(&space_info->groups_sem);
2613 if (ret < 0)
2614 goto loop_check;
2615 head = &space_info->block_groups;
2616 /*
2617 * we've allocated a new chunk, keep
2618 * trying
2619 */
2620 keep_going = 1;
2621 chunk_alloc_done = 1;
2622 } else if (!allowed_chunk_alloc) {
2623 space_info->force_alloc = 1;
2624 }
2625 loop_check:
2626 if (keep_going) {
2627 cur = head->next;
2628 extra_loop = 0;
2629 } else {
2630 break;
2631 }
2632 } else if (cur == head) {
2633 break;
2634 }
2635
2636 block_group = list_entry(cur, struct btrfs_block_group_cache,
2637 list);
2638 atomic_inc(&block_group->count);
2639
2640 search_start = block_group->key.objectid;
2641 cur = cur->next;
2642 }
2643
2644 /* we found what we needed */
2645 if (ins->objectid) {
2646 if (!(data & BTRFS_BLOCK_GROUP_DATA))
2647 trans->block_group = block_group->key.objectid;
2648
2649 if (last_ptr)
2650 *last_ptr = ins->objectid + ins->offset;
2651 ret = 0;
2652 } else if (!ret) {
2653 printk(KERN_ERR "btrfs searching for %llu bytes, "
2654 "num_bytes %llu, loop %d, allowed_alloc %d\n",
2655 (unsigned long long)total_needed,
2656 (unsigned long long)num_bytes,
2657 loop, allowed_chunk_alloc);
2658 ret = -ENOSPC;
2659 }
2660 if (block_group)
2661 put_block_group(block_group);
2662
2663 up_read(&space_info->groups_sem);
2664 return ret;
2665 }
2666
2667 static void dump_space_info(struct btrfs_space_info *info, u64 bytes)
2668 {
2669 struct btrfs_block_group_cache *cache;
2670
2671 printk(KERN_INFO "space_info has %llu free, is %sfull\n",
2672 (unsigned long long)(info->total_bytes - info->bytes_used -
2673 info->bytes_pinned - info->bytes_reserved),
2674 (info->full) ? "" : "not ");
2675 printk(KERN_INFO "space_info total=%llu, pinned=%llu, delalloc=%llu,"
2676 " may_use=%llu, used=%llu\n", info->total_bytes,
2677 info->bytes_pinned, info->bytes_delalloc, info->bytes_may_use,
2678 info->bytes_used);
2679
2680 down_read(&info->groups_sem);
2681 list_for_each_entry(cache, &info->block_groups, list) {
2682 spin_lock(&cache->lock);
2683 printk(KERN_INFO "block group %llu has %llu bytes, %llu used "
2684 "%llu pinned %llu reserved\n",
2685 (unsigned long long)cache->key.objectid,
2686 (unsigned long long)cache->key.offset,
2687 (unsigned long long)btrfs_block_group_used(&cache->item),
2688 (unsigned long long)cache->pinned,
2689 (unsigned long long)cache->reserved);
2690 btrfs_dump_free_space(cache, bytes);
2691 spin_unlock(&cache->lock);
2692 }
2693 up_read(&info->groups_sem);
2694 }
2695
2696 static int __btrfs_reserve_extent(struct btrfs_trans_handle *trans,
2697 struct btrfs_root *root,
2698 u64 num_bytes, u64 min_alloc_size,
2699 u64 empty_size, u64 hint_byte,
2700 u64 search_end, struct btrfs_key *ins,
2701 u64 data)
2702 {
2703 int ret;
2704 u64 search_start = 0;
2705 struct btrfs_fs_info *info = root->fs_info;
2706
2707 data = btrfs_get_alloc_profile(root, data);
2708 again:
2709 /*
2710 * the only place that sets empty_size is btrfs_realloc_node, which
2711 * is not called recursively on allocations
2712 */
2713 if (empty_size || root->ref_cows) {
2714 if (!(data & BTRFS_BLOCK_GROUP_METADATA)) {
2715 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
2716 2 * 1024 * 1024,
2717 BTRFS_BLOCK_GROUP_METADATA |
2718 (info->metadata_alloc_profile &
2719 info->avail_metadata_alloc_bits), 0);
2720 }
2721 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
2722 num_bytes + 2 * 1024 * 1024, data, 0);
2723 }
2724
2725 WARN_ON(num_bytes < root->sectorsize);
2726 ret = find_free_extent(trans, root, num_bytes, empty_size,
2727 search_start, search_end, hint_byte, ins,
2728 trans->alloc_exclude_start,
2729 trans->alloc_exclude_nr, data);
2730
2731 if (ret == -ENOSPC && num_bytes > min_alloc_size) {
2732 num_bytes = num_bytes >> 1;
2733 num_bytes = num_bytes & ~(root->sectorsize - 1);
2734 num_bytes = max(num_bytes, min_alloc_size);
2735 do_chunk_alloc(trans, root->fs_info->extent_root,
2736 num_bytes, data, 1);
2737 goto again;
2738 }
2739 if (ret) {
2740 struct btrfs_space_info *sinfo;
2741
2742 sinfo = __find_space_info(root->fs_info, data);
2743 printk(KERN_ERR "btrfs allocation failed flags %llu, "
2744 "wanted %llu\n", (unsigned long long)data,
2745 (unsigned long long)num_bytes);
2746 dump_space_info(sinfo, num_bytes);
2747 BUG();
2748 }
2749
2750 return ret;
2751 }
2752
2753 int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len)
2754 {
2755 struct btrfs_block_group_cache *cache;
2756 int ret = 0;
2757
2758 cache = btrfs_lookup_block_group(root->fs_info, start);
2759 if (!cache) {
2760 printk(KERN_ERR "Unable to find block group for %llu\n",
2761 (unsigned long long)start);
2762 return -ENOSPC;
2763 }
2764
2765 ret = btrfs_discard_extent(root, start, len);
2766
2767 btrfs_add_free_space(cache, start, len);
2768 put_block_group(cache);
2769 update_reserved_extents(root, start, len, 0);
2770
2771 return ret;
2772 }
2773
2774 int btrfs_reserve_extent(struct btrfs_trans_handle *trans,
2775 struct btrfs_root *root,
2776 u64 num_bytes, u64 min_alloc_size,
2777 u64 empty_size, u64 hint_byte,
2778 u64 search_end, struct btrfs_key *ins,
2779 u64 data)
2780 {
2781 int ret;
2782 ret = __btrfs_reserve_extent(trans, root, num_bytes, min_alloc_size,
2783 empty_size, hint_byte, search_end, ins,
2784 data);
2785 update_reserved_extents(root, ins->objectid, ins->offset, 1);
2786 return ret;
2787 }
2788
2789 static int __btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans,
2790 struct btrfs_root *root, u64 parent,
2791 u64 root_objectid, u64 ref_generation,
2792 u64 owner, struct btrfs_key *ins,
2793 int ref_mod)
2794 {
2795 int ret;
2796 u64 super_used;
2797 u64 root_used;
2798 u64 num_bytes = ins->offset;
2799 u32 sizes[2];
2800 struct btrfs_fs_info *info = root->fs_info;
2801 struct btrfs_root *extent_root = info->extent_root;
2802 struct btrfs_extent_item *extent_item;
2803 struct btrfs_extent_ref *ref;
2804 struct btrfs_path *path;
2805 struct btrfs_key keys[2];
2806
2807 if (parent == 0)
2808 parent = ins->objectid;
2809
2810 /* block accounting for super block */
2811 spin_lock(&info->delalloc_lock);
2812 super_used = btrfs_super_bytes_used(&info->super_copy);
2813 btrfs_set_super_bytes_used(&info->super_copy, super_used + num_bytes);
2814
2815 /* block accounting for root item */
2816 root_used = btrfs_root_used(&root->root_item);
2817 btrfs_set_root_used(&root->root_item, root_used + num_bytes);
2818 spin_unlock(&info->delalloc_lock);
2819
2820 memcpy(&keys[0], ins, sizeof(*ins));
2821 keys[1].objectid = ins->objectid;
2822 keys[1].type = BTRFS_EXTENT_REF_KEY;
2823 keys[1].offset = parent;
2824 sizes[0] = sizeof(*extent_item);
2825 sizes[1] = sizeof(*ref);
2826
2827 path = btrfs_alloc_path();
2828 BUG_ON(!path);
2829
2830 ret = btrfs_insert_empty_items(trans, extent_root, path, keys,
2831 sizes, 2);
2832 BUG_ON(ret);
2833
2834 extent_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
2835 struct btrfs_extent_item);
2836 btrfs_set_extent_refs(path->nodes[0], extent_item, ref_mod);
2837 ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
2838 struct btrfs_extent_ref);
2839
2840 btrfs_set_ref_root(path->nodes[0], ref, root_objectid);
2841 btrfs_set_ref_generation(path->nodes[0], ref, ref_generation);
2842 btrfs_set_ref_objectid(path->nodes[0], ref, owner);
2843 btrfs_set_ref_num_refs(path->nodes[0], ref, ref_mod);
2844
2845 btrfs_mark_buffer_dirty(path->nodes[0]);
2846
2847 trans->alloc_exclude_start = 0;
2848 trans->alloc_exclude_nr = 0;
2849 btrfs_free_path(path);
2850
2851 if (ret)
2852 goto out;
2853
2854 ret = update_block_group(trans, root, ins->objectid,
2855 ins->offset, 1, 0);
2856 if (ret) {
2857 printk(KERN_ERR "btrfs update block group failed for %llu "
2858 "%llu\n", (unsigned long long)ins->objectid,
2859 (unsigned long long)ins->offset);
2860 BUG();
2861 }
2862 out:
2863 return ret;
2864 }
2865
2866 int btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans,
2867 struct btrfs_root *root, u64 parent,
2868 u64 root_objectid, u64 ref_generation,
2869 u64 owner, struct btrfs_key *ins)
2870 {
2871 int ret;
2872
2873 if (root_objectid == BTRFS_TREE_LOG_OBJECTID)
2874 return 0;
2875
2876 ret = btrfs_add_delayed_ref(trans, ins->objectid,
2877 ins->offset, parent, root_objectid,
2878 ref_generation, owner,
2879 BTRFS_ADD_DELAYED_EXTENT, 0);
2880 BUG_ON(ret);
2881 return ret;
2882 }
2883
2884 /*
2885 * this is used by the tree logging recovery code. It records that
2886 * an extent has been allocated and makes sure to clear the free
2887 * space cache bits as well
2888 */
2889 int btrfs_alloc_logged_extent(struct btrfs_trans_handle *trans,
2890 struct btrfs_root *root, u64 parent,
2891 u64 root_objectid, u64 ref_generation,
2892 u64 owner, struct btrfs_key *ins)
2893 {
2894 int ret;
2895 struct btrfs_block_group_cache *block_group;
2896
2897 block_group = btrfs_lookup_block_group(root->fs_info, ins->objectid);
2898 mutex_lock(&block_group->cache_mutex);
2899 cache_block_group(root, block_group);
2900 mutex_unlock(&block_group->cache_mutex);
2901
2902 ret = btrfs_remove_free_space(block_group, ins->objectid,
2903 ins->offset);
2904 BUG_ON(ret);
2905 put_block_group(block_group);
2906 ret = __btrfs_alloc_reserved_extent(trans, root, parent, root_objectid,
2907 ref_generation, owner, ins, 1);
2908 return ret;
2909 }
2910
2911 /*
2912 * finds a free extent and does all the dirty work required for allocation
2913 * returns the key for the extent through ins, and a tree buffer for
2914 * the first block of the extent through buf.
2915 *
2916 * returns 0 if everything worked, non-zero otherwise.
2917 */
2918 int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
2919 struct btrfs_root *root,
2920 u64 num_bytes, u64 parent, u64 min_alloc_size,
2921 u64 root_objectid, u64 ref_generation,
2922 u64 owner_objectid, u64 empty_size, u64 hint_byte,
2923 u64 search_end, struct btrfs_key *ins, u64 data)
2924 {
2925 int ret;
2926 ret = __btrfs_reserve_extent(trans, root, num_bytes,
2927 min_alloc_size, empty_size, hint_byte,
2928 search_end, ins, data);
2929 BUG_ON(ret);
2930 if (root_objectid != BTRFS_TREE_LOG_OBJECTID) {
2931 ret = btrfs_add_delayed_ref(trans, ins->objectid,
2932 ins->offset, parent, root_objectid,
2933 ref_generation, owner_objectid,
2934 BTRFS_ADD_DELAYED_EXTENT, 0);
2935 BUG_ON(ret);
2936 }
2937 update_reserved_extents(root, ins->objectid, ins->offset, 1);
2938 return ret;
2939 }
2940
2941 struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans,
2942 struct btrfs_root *root,
2943 u64 bytenr, u32 blocksize,
2944 int level)
2945 {
2946 struct extent_buffer *buf;
2947
2948 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
2949 if (!buf)
2950 return ERR_PTR(-ENOMEM);
2951 btrfs_set_header_generation(buf, trans->transid);
2952 btrfs_set_buffer_lockdep_class(buf, level);
2953 btrfs_tree_lock(buf);
2954 clean_tree_block(trans, root, buf);
2955
2956 btrfs_set_lock_blocking(buf);
2957 btrfs_set_buffer_uptodate(buf);
2958
2959 if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) {
2960 set_extent_dirty(&root->dirty_log_pages, buf->start,
2961 buf->start + buf->len - 1, GFP_NOFS);
2962 } else {
2963 set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
2964 buf->start + buf->len - 1, GFP_NOFS);
2965 }
2966 trans->blocks_used++;
2967 /* this returns a buffer locked for blocking */
2968 return buf;
2969 }
2970
2971 /*
2972 * helper function to allocate a block for a given tree
2973 * returns the tree buffer or NULL.
2974 */
2975 struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
2976 struct btrfs_root *root,
2977 u32 blocksize, u64 parent,
2978 u64 root_objectid,
2979 u64 ref_generation,
2980 int level,
2981 u64 hint,
2982 u64 empty_size)
2983 {
2984 struct btrfs_key ins;
2985 int ret;
2986 struct extent_buffer *buf;
2987
2988 ret = btrfs_alloc_extent(trans, root, blocksize, parent, blocksize,
2989 root_objectid, ref_generation, level,
2990 empty_size, hint, (u64)-1, &ins, 0);
2991 if (ret) {
2992 BUG_ON(ret > 0);
2993 return ERR_PTR(ret);
2994 }
2995
2996 buf = btrfs_init_new_buffer(trans, root, ins.objectid,
2997 blocksize, level);
2998 return buf;
2999 }
3000
3001 int btrfs_drop_leaf_ref(struct btrfs_trans_handle *trans,
3002 struct btrfs_root *root, struct extent_buffer *leaf)
3003 {
3004 u64 leaf_owner;
3005 u64 leaf_generation;
3006 struct refsort *sorted;
3007 struct btrfs_key key;
3008 struct btrfs_file_extent_item *fi;
3009 int i;
3010 int nritems;
3011 int ret;
3012 int refi = 0;
3013 int slot;
3014
3015 BUG_ON(!btrfs_is_leaf(leaf));
3016 nritems = btrfs_header_nritems(leaf);
3017 leaf_owner = btrfs_header_owner(leaf);
3018 leaf_generation = btrfs_header_generation(leaf);
3019
3020 sorted = kmalloc(sizeof(*sorted) * nritems, GFP_NOFS);
3021 /* we do this loop twice. The first time we build a list
3022 * of the extents we have a reference on, then we sort the list
3023 * by bytenr. The second time around we actually do the
3024 * extent freeing.
3025 */
3026 for (i = 0; i < nritems; i++) {
3027 u64 disk_bytenr;
3028 cond_resched();
3029
3030 btrfs_item_key_to_cpu(leaf, &key, i);
3031
3032 /* only extents have references, skip everything else */
3033 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
3034 continue;
3035
3036 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
3037
3038 /* inline extents live in the btree, they don't have refs */
3039 if (btrfs_file_extent_type(leaf, fi) ==
3040 BTRFS_FILE_EXTENT_INLINE)
3041 continue;
3042
3043 disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
3044
3045 /* holes don't have refs */
3046 if (disk_bytenr == 0)
3047 continue;
3048
3049 sorted[refi].bytenr = disk_bytenr;
3050 sorted[refi].slot = i;
3051 refi++;
3052 }
3053
3054 if (refi == 0)
3055 goto out;
3056
3057 sort(sorted, refi, sizeof(struct refsort), refsort_cmp, NULL);
3058
3059 for (i = 0; i < refi; i++) {
3060 u64 disk_bytenr;
3061
3062 disk_bytenr = sorted[i].bytenr;
3063 slot = sorted[i].slot;
3064
3065 cond_resched();
3066
3067 btrfs_item_key_to_cpu(leaf, &key, slot);
3068 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
3069 continue;
3070
3071 fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
3072
3073 ret = btrfs_free_extent(trans, root, disk_bytenr,
3074 btrfs_file_extent_disk_num_bytes(leaf, fi),
3075 leaf->start, leaf_owner, leaf_generation,
3076 key.objectid, 0);
3077 BUG_ON(ret);
3078
3079 atomic_inc(&root->fs_info->throttle_gen);
3080 wake_up(&root->fs_info->transaction_throttle);
3081 cond_resched();
3082 }
3083 out:
3084 kfree(sorted);
3085 return 0;
3086 }
3087
3088 static noinline int cache_drop_leaf_ref(struct btrfs_trans_handle *trans,
3089 struct btrfs_root *root,
3090 struct btrfs_leaf_ref *ref)
3091 {
3092 int i;
3093 int ret;
3094 struct btrfs_extent_info *info;
3095 struct refsort *sorted;
3096
3097 if (ref->nritems == 0)
3098 return 0;
3099
3100 sorted = kmalloc(sizeof(*sorted) * ref->nritems, GFP_NOFS);
3101 for (i = 0; i < ref->nritems; i++) {
3102 sorted[i].bytenr = ref->extents[i].bytenr;
3103 sorted[i].slot = i;
3104 }
3105 sort(sorted, ref->nritems, sizeof(struct refsort), refsort_cmp, NULL);
3106
3107 /*
3108 * the items in the ref were sorted when the ref was inserted
3109 * into the ref cache, so this is already in order
3110 */
3111 for (i = 0; i < ref->nritems; i++) {
3112 info = ref->extents + sorted[i].slot;
3113 ret = btrfs_free_extent(trans, root, info->bytenr,
3114 info->num_bytes, ref->bytenr,
3115 ref->owner, ref->generation,
3116 info->objectid, 0);
3117
3118 atomic_inc(&root->fs_info->throttle_gen);
3119 wake_up(&root->fs_info->transaction_throttle);
3120 cond_resched();
3121
3122 BUG_ON(ret);
3123 info++;
3124 }
3125
3126 kfree(sorted);
3127 return 0;
3128 }
3129
3130 static int drop_snap_lookup_refcount(struct btrfs_trans_handle *trans,
3131 struct btrfs_root *root, u64 start,
3132 u64 len, u32 *refs)
3133 {
3134 int ret;
3135
3136 ret = btrfs_lookup_extent_ref(trans, root, start, len, refs);
3137 BUG_ON(ret);
3138
3139 #if 0 /* some debugging code in case we see problems here */
3140 /* if the refs count is one, it won't get increased again. But
3141 * if the ref count is > 1, someone may be decreasing it at
3142 * the same time we are.
3143 */
3144 if (*refs != 1) {
3145 struct extent_buffer *eb = NULL;
3146 eb = btrfs_find_create_tree_block(root, start, len);
3147 if (eb)
3148 btrfs_tree_lock(eb);
3149
3150 mutex_lock(&root->fs_info->alloc_mutex);
3151 ret = lookup_extent_ref(NULL, root, start, len, refs);
3152 BUG_ON(ret);
3153 mutex_unlock(&root->fs_info->alloc_mutex);
3154
3155 if (eb) {
3156 btrfs_tree_unlock(eb);
3157 free_extent_buffer(eb);
3158 }
3159 if (*refs == 1) {
3160 printk(KERN_ERR "btrfs block %llu went down to one "
3161 "during drop_snap\n", (unsigned long long)start);
3162 }
3163
3164 }
3165 #endif
3166
3167 cond_resched();
3168 return ret;
3169 }
3170
3171 /*
3172 * this is used while deleting old snapshots, and it drops the refs
3173 * on a whole subtree starting from a level 1 node.
3174 *
3175 * The idea is to sort all the leaf pointers, and then drop the
3176 * ref on all the leaves in order. Most of the time the leaves
3177 * will have ref cache entries, so no leaf IOs will be required to
3178 * find the extents they have references on.
3179 *
3180 * For each leaf, any references it has are also dropped in order
3181 *
3182 * This ends up dropping the references in something close to optimal
3183 * order for reading and modifying the extent allocation tree.
3184 */
3185 static noinline int drop_level_one_refs(struct btrfs_trans_handle *trans,
3186 struct btrfs_root *root,
3187 struct btrfs_path *path)
3188 {
3189 u64 bytenr;
3190 u64 root_owner;
3191 u64 root_gen;
3192 struct extent_buffer *eb = path->nodes[1];
3193 struct extent_buffer *leaf;
3194 struct btrfs_leaf_ref *ref;
3195 struct refsort *sorted = NULL;
3196 int nritems = btrfs_header_nritems(eb);
3197 int ret;
3198 int i;
3199 int refi = 0;
3200 int slot = path->slots[1];
3201 u32 blocksize = btrfs_level_size(root, 0);
3202 u32 refs;
3203
3204 if (nritems == 0)
3205 goto out;
3206
3207 root_owner = btrfs_header_owner(eb);
3208 root_gen = btrfs_header_generation(eb);
3209 sorted = kmalloc(sizeof(*sorted) * nritems, GFP_NOFS);
3210
3211 /*
3212 * step one, sort all the leaf pointers so we don't scribble
3213 * randomly into the extent allocation tree
3214 */
3215 for (i = slot; i < nritems; i++) {
3216 sorted[refi].bytenr = btrfs_node_blockptr(eb, i);
3217 sorted[refi].slot = i;
3218 refi++;
3219 }
3220
3221 /*
3222 * nritems won't be zero, but if we're picking up drop_snapshot
3223 * after a crash, slot might be > 0, so double check things
3224 * just in case.
3225 */
3226 if (refi == 0)
3227 goto out;
3228
3229 sort(sorted, refi, sizeof(struct refsort), refsort_cmp, NULL);
3230
3231 /*
3232 * the first loop frees everything the leaves point to
3233 */
3234 for (i = 0; i < refi; i++) {
3235 u64 ptr_gen;
3236
3237 bytenr = sorted[i].bytenr;
3238
3239 /*
3240 * check the reference count on this leaf. If it is > 1
3241 * we just decrement it below and don't update any
3242 * of the refs the leaf points to.
3243 */
3244 ret = drop_snap_lookup_refcount(trans, root, bytenr,
3245 blocksize, &refs);
3246 BUG_ON(ret);
3247 if (refs != 1)
3248 continue;
3249
3250 ptr_gen = btrfs_node_ptr_generation(eb, sorted[i].slot);
3251
3252 /*
3253 * the leaf only had one reference, which means the
3254 * only thing pointing to this leaf is the snapshot
3255 * we're deleting. It isn't possible for the reference
3256 * count to increase again later
3257 *
3258 * The reference cache is checked for the leaf,
3259 * and if found we'll be able to drop any refs held by
3260 * the leaf without needing to read it in.
3261 */
3262 ref = btrfs_lookup_leaf_ref(root, bytenr);
3263 if (ref && ref->generation != ptr_gen) {
3264 btrfs_free_leaf_ref(root, ref);
3265 ref = NULL;
3266 }
3267 if (ref) {
3268 ret = cache_drop_leaf_ref(trans, root, ref);
3269 BUG_ON(ret);
3270 btrfs_remove_leaf_ref(root, ref);
3271 btrfs_free_leaf_ref(root, ref);
3272 } else {
3273 /*
3274 * the leaf wasn't in the reference cache, so
3275 * we have to read it.
3276 */
3277 leaf = read_tree_block(root, bytenr, blocksize,
3278 ptr_gen);
3279 ret = btrfs_drop_leaf_ref(trans, root, leaf);
3280 BUG_ON(ret);
3281 free_extent_buffer(leaf);
3282 }
3283 atomic_inc(&root->fs_info->throttle_gen);
3284 wake_up(&root->fs_info->transaction_throttle);
3285 cond_resched();
3286 }
3287
3288 /*
3289 * run through the loop again to free the refs on the leaves.
3290 * This is faster than doing it in the loop above because
3291 * the leaves are likely to be clustered together. We end up
3292 * working in nice chunks on the extent allocation tree.
3293 */
3294 for (i = 0; i < refi; i++) {
3295 bytenr = sorted[i].bytenr;
3296 ret = btrfs_free_extent(trans, root, bytenr,
3297 blocksize, eb->start,
3298 root_owner, root_gen, 0, 1);
3299 BUG_ON(ret);
3300
3301 atomic_inc(&root->fs_info->throttle_gen);
3302 wake_up(&root->fs_info->transaction_throttle);
3303 cond_resched();
3304 }
3305 out:
3306 kfree(sorted);
3307
3308 /*
3309 * update the path to show we've processed the entire level 1
3310 * node. This will get saved into the root's drop_snapshot_progress
3311 * field so these drops are not repeated again if this transaction
3312 * commits.
3313 */
3314 path->slots[1] = nritems;
3315 return 0;
3316 }
3317
3318 /*
3319 * helper function for drop_snapshot, this walks down the tree dropping ref
3320 * counts as it goes.
3321 */
3322 static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
3323 struct btrfs_root *root,
3324 struct btrfs_path *path, int *level)
3325 {
3326 u64 root_owner;
3327 u64 root_gen;
3328 u64 bytenr;
3329 u64 ptr_gen;
3330 struct extent_buffer *next;
3331 struct extent_buffer *cur;
3332 struct extent_buffer *parent;
3333 u32 blocksize;
3334 int ret;
3335 u32 refs;
3336
3337 WARN_ON(*level < 0);
3338 WARN_ON(*level >= BTRFS_MAX_LEVEL);
3339 ret = drop_snap_lookup_refcount(trans, root, path->nodes[*level]->start,
3340 path->nodes[*level]->len, &refs);
3341 BUG_ON(ret);
3342 if (refs > 1)
3343 goto out;
3344
3345 /*
3346 * walk down to the last node level and free all the leaves
3347 */
3348 while (*level >= 0) {
3349 WARN_ON(*level < 0);
3350 WARN_ON(*level >= BTRFS_MAX_LEVEL);
3351 cur = path->nodes[*level];
3352
3353 if (btrfs_header_level(cur) != *level)
3354 WARN_ON(1);
3355
3356 if (path->slots[*level] >=
3357 btrfs_header_nritems(cur))
3358 break;
3359
3360 /* the new code goes down to level 1 and does all the
3361 * leaves pointed to that node in bulk. So, this check
3362 * for level 0 will always be false.
3363 *
3364 * But, the disk format allows the drop_snapshot_progress
3365 * field in the root to leave things in a state where
3366 * a leaf will need cleaning up here. If someone crashes
3367 * with the old code and then boots with the new code,
3368 * we might find a leaf here.
3369 */
3370 if (*level == 0) {
3371 ret = btrfs_drop_leaf_ref(trans, root, cur);
3372 BUG_ON(ret);
3373 break;
3374 }
3375
3376 /*
3377 * once we get to level one, process the whole node
3378 * at once, including everything below it.
3379 */
3380 if (*level == 1) {
3381 ret = drop_level_one_refs(trans, root, path);
3382 BUG_ON(ret);
3383 break;
3384 }
3385
3386 bytenr = btrfs_node_blockptr(cur, path->slots[*level]);
3387 ptr_gen = btrfs_node_ptr_generation(cur, path->slots[*level]);
3388 blocksize = btrfs_level_size(root, *level - 1);
3389
3390 ret = drop_snap_lookup_refcount(trans, root, bytenr,
3391 blocksize, &refs);
3392 BUG_ON(ret);
3393
3394 /*
3395 * if there is more than one reference, we don't need
3396 * to read that node to drop any references it has. We
3397 * just drop the ref we hold on that node and move on to the
3398 * next slot in this level.
3399 */
3400 if (refs != 1) {
3401 parent = path->nodes[*level];
3402 root_owner = btrfs_header_owner(parent);
3403 root_gen = btrfs_header_generation(parent);
3404 path->slots[*level]++;
3405
3406 ret = btrfs_free_extent(trans, root, bytenr,
3407 blocksize, parent->start,
3408 root_owner, root_gen,
3409 *level - 1, 1);
3410 BUG_ON(ret);
3411
3412 atomic_inc(&root->fs_info->throttle_gen);
3413 wake_up(&root->fs_info->transaction_throttle);
3414 cond_resched();
3415
3416 continue;
3417 }
3418
3419 /*
3420 * we need to keep freeing things in the next level down.
3421 * read the block and loop around to process it
3422 */
3423 next = read_tree_block(root, bytenr, blocksize, ptr_gen);
3424 WARN_ON(*level <= 0);
3425 if (path->nodes[*level-1])
3426 free_extent_buffer(path->nodes[*level-1]);
3427 path->nodes[*level-1] = next;
3428 *level = btrfs_header_level(next);
3429 path->slots[*level] = 0;
3430 cond_resched();
3431 }
3432 out:
3433 WARN_ON(*level < 0);
3434 WARN_ON(*level >= BTRFS_MAX_LEVEL);
3435
3436 if (path->nodes[*level] == root->node) {
3437 parent = path->nodes[*level];
3438 bytenr = path->nodes[*level]->start;
3439 } else {
3440 parent = path->nodes[*level + 1];
3441 bytenr = btrfs_node_blockptr(parent, path->slots[*level + 1]);
3442 }
3443
3444 blocksize = btrfs_level_size(root, *level);
3445 root_owner = btrfs_header_owner(parent);
3446 root_gen = btrfs_header_generation(parent);
3447
3448 /*
3449 * cleanup and free the reference on the last node
3450 * we processed
3451 */
3452 ret = btrfs_free_extent(trans, root, bytenr, blocksize,
3453 parent->start, root_owner, root_gen,
3454 *level, 1);
3455 free_extent_buffer(path->nodes[*level]);
3456 path->nodes[*level] = NULL;
3457
3458 *level += 1;
3459 BUG_ON(ret);
3460
3461 cond_resched();
3462 return 0;
3463 }
3464
3465 /*
3466 * helper function for drop_subtree, this function is similar to
3467 * walk_down_tree. The main difference is that it checks reference
3468 * counts while tree blocks are locked.
3469 */
3470 static noinline int walk_down_subtree(struct btrfs_trans_handle *trans,
3471 struct btrfs_root *root,
3472 struct btrfs_path *path, int *level)
3473 {
3474 struct extent_buffer *next;
3475 struct extent_buffer *cur;
3476 struct extent_buffer *parent;
3477 u64 bytenr;
3478 u64 ptr_gen;
3479 u32 blocksize;
3480 u32 refs;
3481 int ret;
3482
3483 cur = path->nodes[*level];
3484 ret = btrfs_lookup_extent_ref(trans, root, cur->start, cur->len,
3485 &refs);
3486 BUG_ON(ret);
3487 if (refs > 1)
3488 goto out;
3489
3490 while (*level >= 0) {
3491 cur = path->nodes[*level];
3492 if (*level == 0) {
3493 ret = btrfs_drop_leaf_ref(trans, root, cur);
3494 BUG_ON(ret);
3495 clean_tree_block(trans, root, cur);
3496 break;
3497 }
3498 if (path->slots[*level] >= btrfs_header_nritems(cur)) {
3499 clean_tree_block(trans, root, cur);
3500 break;
3501 }
3502
3503 bytenr = btrfs_node_blockptr(cur, path->slots[*level]);
3504 blocksize = btrfs_level_size(root, *level - 1);
3505 ptr_gen = btrfs_node_ptr_generation(cur, path->slots[*level]);
3506
3507 next = read_tree_block(root, bytenr, blocksize, ptr_gen);
3508 btrfs_tree_lock(next);
3509 btrfs_set_lock_blocking(next);
3510
3511 ret = btrfs_lookup_extent_ref(trans, root, bytenr, blocksize,
3512 &refs);
3513 BUG_ON(ret);
3514 if (refs > 1) {
3515 parent = path->nodes[*level];
3516 ret = btrfs_free_extent(trans, root, bytenr,
3517 blocksize, parent->start,
3518 btrfs_header_owner(parent),
3519 btrfs_header_generation(parent),
3520 *level - 1, 1);
3521 BUG_ON(ret);
3522 path->slots[*level]++;
3523 btrfs_tree_unlock(next);
3524 free_extent_buffer(next);
3525 continue;
3526 }
3527
3528 *level = btrfs_header_level(next);
3529 path->nodes[*level] = next;
3530 path->slots[*level] = 0;
3531 path->locks[*level] = 1;
3532 cond_resched();
3533 }
3534 out:
3535 parent = path->nodes[*level + 1];
3536 bytenr = path->nodes[*level]->start;
3537 blocksize = path->nodes[*level]->len;
3538
3539 ret = btrfs_free_extent(trans, root, bytenr, blocksize,
3540 parent->start, btrfs_header_owner(parent),
3541 btrfs_header_generation(parent), *level, 1);
3542 BUG_ON(ret);
3543
3544 if (path->locks[*level]) {
3545 btrfs_tree_unlock(path->nodes[*level]);
3546 path->locks[*level] = 0;
3547 }
3548 free_extent_buffer(path->nodes[*level]);
3549 path->nodes[*level] = NULL;
3550 *level += 1;
3551 cond_resched();
3552 return 0;
3553 }
3554
3555 /*
3556 * helper for dropping snapshots. This walks back up the tree in the path
3557 * to find the first node higher up where we haven't yet gone through
3558 * all the slots
3559 */
3560 static noinline int walk_up_tree(struct btrfs_trans_handle *trans,
3561 struct btrfs_root *root,
3562 struct btrfs_path *path,
3563 int *level, int max_level)
3564 {
3565 u64 root_owner;
3566 u64 root_gen;
3567 struct btrfs_root_item *root_item = &root->root_item;
3568 int i;
3569 int slot;
3570 int ret;
3571
3572 for (i = *level; i < max_level && path->nodes[i]; i++) {
3573 slot = path->slots[i];
3574 if (slot < btrfs_header_nritems(path->nodes[i]) - 1) {
3575 struct extent_buffer *node;
3576 struct btrfs_disk_key disk_key;
3577
3578 /*
3579 * there is more work to do in this level.
3580 * Update the drop_progress marker to reflect
3581 * the work we've done so far, and then bump
3582 * the slot number
3583 */
3584 node = path->nodes[i];
3585 path->slots[i]++;
3586 *level = i;
3587 WARN_ON(*level == 0);
3588 btrfs_node_key(node, &disk_key, path->slots[i]);
3589 memcpy(&root_item->drop_progress,
3590 &disk_key, sizeof(disk_key));
3591 root_item->drop_level = i;
3592 return 0;
3593 } else {
3594 struct extent_buffer *parent;
3595
3596 /*
3597 * this whole node is done, free our reference
3598 * on it and go up one level
3599 */
3600 if (path->nodes[*level] == root->node)
3601 parent = path->nodes[*level];
3602 else
3603 parent = path->nodes[*level + 1];
3604
3605 root_owner = btrfs_header_owner(parent);
3606 root_gen = btrfs_header_generation(parent);
3607
3608 clean_tree_block(trans, root, path->nodes[*level]);
3609 ret = btrfs_free_extent(trans, root,
3610 path->nodes[*level]->start,
3611 path->nodes[*level]->len,
3612 parent->start, root_owner,
3613 root_gen, *level, 1);
3614 BUG_ON(ret);
3615 if (path->locks[*level]) {
3616 btrfs_tree_unlock(path->nodes[*level]);
3617 path->locks[*level] = 0;
3618 }
3619 free_extent_buffer(path->nodes[*level]);
3620 path->nodes[*level] = NULL;
3621 *level = i + 1;
3622 }
3623 }
3624 return 1;
3625 }
3626
3627 /*
3628 * drop the reference count on the tree rooted at 'snap'. This traverses
3629 * the tree freeing any blocks that have a ref count of zero after being
3630 * decremented.
3631 */
3632 int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
3633 *root)
3634 {
3635 int ret = 0;
3636 int wret;
3637 int level;
3638 struct btrfs_path *path;
3639 int i;
3640 int orig_level;
3641 struct btrfs_root_item *root_item = &root->root_item;
3642
3643 WARN_ON(!mutex_is_locked(&root->fs_info->drop_mutex));
3644 path = btrfs_alloc_path();
3645 BUG_ON(!path);
3646
3647 level = btrfs_header_level(root->node);
3648 orig_level = level;
3649 if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
3650 path->nodes[level] = root->node;
3651 extent_buffer_get(root->node);
3652 path->slots[level] = 0;
3653 } else {
3654 struct btrfs_key key;
3655 struct btrfs_disk_key found_key;
3656 struct extent_buffer *node;
3657
3658 btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
3659 level = root_item->drop_level;
3660 path->lowest_level = level;
3661 wret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3662 if (wret < 0) {
3663 ret = wret;
3664 goto out;
3665 }
3666 node = path->nodes[level];
3667 btrfs_node_key(node, &found_key, path->slots[level]);
3668 WARN_ON(memcmp(&found_key, &root_item->drop_progress,
3669 sizeof(found_key)));
3670 /*
3671 * unlock our path, this is safe because only this
3672 * function is allowed to delete this snapshot
3673 */
3674 for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
3675 if (path->nodes[i] && path->locks[i]) {
3676 path->locks[i] = 0;
3677 btrfs_tree_unlock(path->nodes[i]);
3678 }
3679 }
3680 }
3681 while (1) {
3682 wret = walk_down_tree(trans, root, path, &level);
3683 if (wret > 0)
3684 break;
3685 if (wret < 0)
3686 ret = wret;
3687
3688 wret = walk_up_tree(trans, root, path, &level,
3689 BTRFS_MAX_LEVEL);
3690 if (wret > 0)
3691 break;
3692 if (wret < 0)
3693 ret = wret;
3694 if (trans->transaction->in_commit) {
3695 ret = -EAGAIN;
3696 break;
3697 }
3698 atomic_inc(&root->fs_info->throttle_gen);
3699 wake_up(&root->fs_info->transaction_throttle);
3700 }
3701 for (i = 0; i <= orig_level; i++) {
3702 if (path->nodes[i]) {
3703 free_extent_buffer(path->nodes[i]);
3704 path->nodes[i] = NULL;
3705 }
3706 }
3707 out:
3708 btrfs_free_path(path);
3709 return ret;
3710 }
3711
3712 int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
3713 struct btrfs_root *root,
3714 struct extent_buffer *node,
3715 struct extent_buffer *parent)
3716 {
3717 struct btrfs_path *path;
3718 int level;
3719 int parent_level;
3720 int ret = 0;
3721 int wret;
3722
3723 path = btrfs_alloc_path();
3724 BUG_ON(!path);
3725
3726 btrfs_assert_tree_locked(parent);
3727 parent_level = btrfs_header_level(parent);
3728 extent_buffer_get(parent);
3729 path->nodes[parent_level] = parent;
3730 path->slots[parent_level] = btrfs_header_nritems(parent);
3731
3732 btrfs_assert_tree_locked(node);
3733 level = btrfs_header_level(node);
3734 extent_buffer_get(node);
3735 path->nodes[level] = node;
3736 path->slots[level] = 0;
3737
3738 while (1) {
3739 wret = walk_down_subtree(trans, root, path, &level);
3740 if (wret < 0)
3741 ret = wret;
3742 if (wret != 0)
3743 break;
3744
3745 wret = walk_up_tree(trans, root, path, &level, parent_level);
3746 if (wret < 0)
3747 ret = wret;
3748 if (wret != 0)
3749 break;
3750 }
3751
3752 btrfs_free_path(path);
3753 return ret;
3754 }
3755
3756 static unsigned long calc_ra(unsigned long start, unsigned long last,
3757 unsigned long nr)
3758 {
3759 return min(last, start + nr - 1);
3760 }
3761
3762 static noinline int relocate_inode_pages(struct inode *inode, u64 start,
3763 u64 len)
3764 {
3765 u64 page_start;
3766 u64 page_end;
3767 unsigned long first_index;
3768 unsigned long last_index;
3769 unsigned long i;
3770 struct page *page;
3771 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
3772 struct file_ra_state *ra;
3773 struct btrfs_ordered_extent *ordered;
3774 unsigned int total_read = 0;
3775 unsigned int total_dirty = 0;
3776 int ret = 0;
3777
3778 ra = kzalloc(sizeof(*ra), GFP_NOFS);
3779
3780 mutex_lock(&inode->i_mutex);
3781 first_index = start >> PAGE_CACHE_SHIFT;
3782 last_index = (start + len - 1) >> PAGE_CACHE_SHIFT;
3783
3784 /* make sure the dirty trick played by the caller work */
3785 ret = invalidate_inode_pages2_range(inode->i_mapping,
3786 first_index, last_index);
3787 if (ret)
3788 goto out_unlock;
3789
3790 file_ra_state_init(ra, inode->i_mapping);
3791
3792 for (i = first_index ; i <= last_index; i++) {
3793 if (total_read % ra->ra_pages == 0) {
3794 btrfs_force_ra(inode->i_mapping, ra, NULL, i,
3795 calc_ra(i, last_index, ra->ra_pages));
3796 }
3797 total_read++;
3798 again:
3799 if (((u64)i << PAGE_CACHE_SHIFT) > i_size_read(inode))
3800 BUG_ON(1);
3801 page = grab_cache_page(inode->i_mapping, i);
3802 if (!page) {
3803 ret = -ENOMEM;
3804 goto out_unlock;
3805 }
3806 if (!PageUptodate(page)) {
3807 btrfs_readpage(NULL, page);
3808 lock_page(page);
3809 if (!PageUptodate(page)) {
3810 unlock_page(page);
3811 page_cache_release(page);
3812 ret = -EIO;
3813 goto out_unlock;
3814 }
3815 }
3816 wait_on_page_writeback(page);
3817
3818 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
3819 page_end = page_start + PAGE_CACHE_SIZE - 1;
3820 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
3821
3822 ordered = btrfs_lookup_ordered_extent(inode, page_start);
3823 if (ordered) {
3824 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
3825 unlock_page(page);
3826 page_cache_release(page);
3827 btrfs_start_ordered_extent(inode, ordered, 1);
3828 btrfs_put_ordered_extent(ordered);
3829 goto again;
3830 }
3831 set_page_extent_mapped(page);
3832
3833 if (i == first_index)
3834 set_extent_bits(io_tree, page_start, page_end,
3835 EXTENT_BOUNDARY, GFP_NOFS);
3836 btrfs_set_extent_delalloc(inode, page_start, page_end);
3837
3838 set_page_dirty(page);
3839 total_dirty++;
3840
3841 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
3842 unlock_page(page);
3843 page_cache_release(page);
3844 }
3845
3846 out_unlock:
3847 kfree(ra);
3848 mutex_unlock(&inode->i_mutex);
3849 balance_dirty_pages_ratelimited_nr(inode->i_mapping, total_dirty);
3850 return ret;
3851 }
3852
3853 static noinline int relocate_data_extent(struct inode *reloc_inode,
3854 struct btrfs_key *extent_key,
3855 u64 offset)
3856 {
3857 struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
3858 struct extent_map_tree *em_tree = &BTRFS_I(reloc_inode)->extent_tree;
3859 struct extent_map *em;
3860 u64 start = extent_key->objectid - offset;
3861 u64 end = start + extent_key->offset - 1;
3862
3863 em = alloc_extent_map(GFP_NOFS);
3864 BUG_ON(!em || IS_ERR(em));
3865
3866 em->start = start;
3867 em->len = extent_key->offset;
3868 em->block_len = extent_key->offset;
3869 em->block_start = extent_key->objectid;
3870 em->bdev = root->fs_info->fs_devices->latest_bdev;
3871 set_bit(EXTENT_FLAG_PINNED, &em->flags);
3872
3873 /* setup extent map to cheat btrfs_readpage */
3874 lock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS);
3875 while (1) {
3876 int ret;
3877 spin_lock(&em_tree->lock);
3878 ret = add_extent_mapping(em_tree, em);
3879 spin_unlock(&em_tree->lock);
3880 if (ret != -EEXIST) {
3881 free_extent_map(em);
3882 break;
3883 }
3884 btrfs_drop_extent_cache(reloc_inode, start, end, 0);
3885 }
3886 unlock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS);
3887
3888 return relocate_inode_pages(reloc_inode, start, extent_key->offset);
3889 }
3890
3891 struct btrfs_ref_path {
3892 u64 extent_start;
3893 u64 nodes[BTRFS_MAX_LEVEL];
3894 u64 root_objectid;
3895 u64 root_generation;
3896 u64 owner_objectid;
3897 u32 num_refs;
3898 int lowest_level;
3899 int current_level;
3900 int shared_level;
3901
3902 struct btrfs_key node_keys[BTRFS_MAX_LEVEL];
3903 u64 new_nodes[BTRFS_MAX_LEVEL];
3904 };
3905
3906 struct disk_extent {
3907 u64 ram_bytes;
3908 u64 disk_bytenr;
3909 u64 disk_num_bytes;
3910 u64 offset;
3911 u64 num_bytes;
3912 u8 compression;
3913 u8 encryption;
3914 u16 other_encoding;
3915 };
3916
3917 static int is_cowonly_root(u64 root_objectid)
3918 {
3919 if (root_objectid == BTRFS_ROOT_TREE_OBJECTID ||
3920 root_objectid == BTRFS_EXTENT_TREE_OBJECTID ||
3921 root_objectid == BTRFS_CHUNK_TREE_OBJECTID ||
3922 root_objectid == BTRFS_DEV_TREE_OBJECTID ||
3923 root_objectid == BTRFS_TREE_LOG_OBJECTID ||
3924 root_objectid == BTRFS_CSUM_TREE_OBJECTID)
3925 return 1;
3926 return 0;
3927 }
3928
3929 static noinline int __next_ref_path(struct btrfs_trans_handle *trans,
3930 struct btrfs_root *extent_root,
3931 struct btrfs_ref_path *ref_path,
3932 int first_time)
3933 {
3934 struct extent_buffer *leaf;
3935 struct btrfs_path *path;
3936 struct btrfs_extent_ref *ref;
3937 struct btrfs_key key;
3938 struct btrfs_key found_key;
3939 u64 bytenr;
3940 u32 nritems;
3941 int level;
3942 int ret = 1;
3943
3944 path = btrfs_alloc_path();
3945 if (!path)
3946 return -ENOMEM;
3947
3948 if (first_time) {
3949 ref_path->lowest_level = -1;
3950 ref_path->current_level = -1;
3951 ref_path->shared_level = -1;
3952 goto walk_up;
3953 }
3954 walk_down:
3955 level = ref_path->current_level - 1;
3956 while (level >= -1) {
3957 u64 parent;
3958 if (level < ref_path->lowest_level)
3959 break;
3960
3961 if (level >= 0)
3962 bytenr = ref_path->nodes[level];
3963 else
3964 bytenr = ref_path->extent_start;
3965 BUG_ON(bytenr == 0);
3966
3967 parent = ref_path->nodes[level + 1];
3968 ref_path->nodes[level + 1] = 0;
3969 ref_path->current_level = level;
3970 BUG_ON(parent == 0);
3971
3972 key.objectid = bytenr;
3973 key.offset = parent + 1;
3974 key.type = BTRFS_EXTENT_REF_KEY;
3975
3976 ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 0);
3977 if (ret < 0)
3978 goto out;
3979 BUG_ON(ret == 0);
3980
3981 leaf = path->nodes[0];
3982 nritems = btrfs_header_nritems(leaf);
3983 if (path->slots[0] >= nritems) {
3984 ret = btrfs_next_leaf(extent_root, path);
3985 if (ret < 0)
3986 goto out;
3987 if (ret > 0)
3988 goto next;
3989 leaf = path->nodes[0];
3990 }
3991
3992 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
3993 if (found_key.objectid == bytenr &&
3994 found_key.type == BTRFS_EXTENT_REF_KEY) {
3995 if (level < ref_path->shared_level)
3996 ref_path->shared_level = level;
3997 goto found;
3998 }
3999 next:
4000 level--;
4001 btrfs_release_path(extent_root, path);
4002 cond_resched();
4003 }
4004 /* reached lowest level */
4005 ret = 1;
4006 goto out;
4007 walk_up:
4008 level = ref_path->current_level;
4009 while (level < BTRFS_MAX_LEVEL - 1) {
4010 u64 ref_objectid;
4011
4012 if (level >= 0)
4013 bytenr = ref_path->nodes[level];
4014 else
4015 bytenr = ref_path->extent_start;
4016
4017 BUG_ON(bytenr == 0);
4018
4019 key.objectid = bytenr;
4020 key.offset = 0;
4021 key.type = BTRFS_EXTENT_REF_KEY;
4022
4023 ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 0);
4024 if (ret < 0)
4025 goto out;
4026
4027 leaf = path->nodes[0];
4028 nritems = btrfs_header_nritems(leaf);
4029 if (path->slots[0] >= nritems) {
4030 ret = btrfs_next_leaf(extent_root, path);
4031 if (ret < 0)
4032 goto out;
4033 if (ret > 0) {
4034 /* the extent was freed by someone */
4035 if (ref_path->lowest_level == level)
4036 goto out;
4037 btrfs_release_path(extent_root, path);
4038 goto walk_down;
4039 }
4040 leaf = path->nodes[0];
4041 }
4042
4043 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
4044 if (found_key.objectid != bytenr ||
4045 found_key.type != BTRFS_EXTENT_REF_KEY) {
4046 /* the extent was freed by someone */
4047 if (ref_path->lowest_level == level) {
4048 ret = 1;
4049 goto out;
4050 }
4051 btrfs_release_path(extent_root, path);
4052 goto walk_down;
4053 }
4054 found:
4055 ref = btrfs_item_ptr(leaf, path->slots[0],
4056 struct btrfs_extent_ref);
4057 ref_objectid = btrfs_ref_objectid(leaf, ref);
4058 if (ref_objectid < BTRFS_FIRST_FREE_OBJECTID) {
4059 if (first_time) {
4060 level = (int)ref_objectid;
4061 BUG_ON(level >= BTRFS_MAX_LEVEL);
4062 ref_path->lowest_level = level;
4063 ref_path->current_level = level;
4064 ref_path->nodes[level] = bytenr;
4065 } else {
4066 WARN_ON(ref_objectid != level);
4067 }
4068 } else {
4069 WARN_ON(level != -1);
4070 }
4071 first_time = 0;
4072
4073 if (ref_path->lowest_level == level) {
4074 ref_path->owner_objectid = ref_objectid;
4075 ref_path->num_refs = btrfs_ref_num_refs(leaf, ref);
4076 }
4077
4078 /*
4079 * the block is tree root or the block isn't in reference
4080 * counted tree.
4081 */
4082 if (found_key.objectid == found_key.offset ||
4083 is_cowonly_root(btrfs_ref_root(leaf, ref))) {
4084 ref_path->root_objectid = btrfs_ref_root(leaf, ref);
4085 ref_path->root_generation =
4086 btrfs_ref_generation(leaf, ref);
4087 if (level < 0) {
4088 /* special reference from the tree log */
4089 ref_path->nodes[0] = found_key.offset;
4090 ref_path->current_level = 0;
4091 }
4092 ret = 0;
4093 goto out;
4094 }
4095
4096 level++;
4097 BUG_ON(ref_path->nodes[level] != 0);
4098 ref_path->nodes[level] = found_key.offset;
4099 ref_path->current_level = level;
4100
4101 /*
4102 * the reference was created in the running transaction,
4103 * no need to continue walking up.
4104 */
4105 if (btrfs_ref_generation(leaf, ref) == trans->transid) {
4106 ref_path->root_objectid = btrfs_ref_root(leaf, ref);
4107 ref_path->root_generation =
4108 btrfs_ref_generation(leaf, ref);
4109 ret = 0;
4110 goto out;
4111 }
4112
4113 btrfs_release_path(extent_root, path);
4114 cond_resched();
4115 }
4116 /* reached max tree level, but no tree root found. */
4117 BUG();
4118 out:
4119 btrfs_free_path(path);
4120 return ret;
4121 }
4122
4123 static int btrfs_first_ref_path(struct btrfs_trans_handle *trans,
4124 struct btrfs_root *extent_root,
4125 struct btrfs_ref_path *ref_path,
4126 u64 extent_start)
4127 {
4128 memset(ref_path, 0, sizeof(*ref_path));
4129 ref_path->extent_start = extent_start;
4130
4131 return __next_ref_path(trans, extent_root, ref_path, 1);
4132 }
4133
4134 static int btrfs_next_ref_path(struct btrfs_trans_handle *trans,
4135 struct btrfs_root *extent_root,
4136 struct btrfs_ref_path *ref_path)
4137 {
4138 return __next_ref_path(trans, extent_root, ref_path, 0);
4139 }
4140
4141 static noinline int get_new_locations(struct inode *reloc_inode,
4142 struct btrfs_key *extent_key,
4143 u64 offset, int no_fragment,
4144 struct disk_extent **extents,
4145 int *nr_extents)
4146 {
4147 struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
4148 struct btrfs_path *path;
4149 struct btrfs_file_extent_item *fi;
4150 struct extent_buffer *leaf;
4151 struct disk_extent *exts = *extents;
4152 struct btrfs_key found_key;
4153 u64 cur_pos;
4154 u64 last_byte;
4155 u32 nritems;
4156 int nr = 0;
4157 int max = *nr_extents;
4158 int ret;
4159
4160 WARN_ON(!no_fragment && *extents);
4161 if (!exts) {
4162 max = 1;
4163 exts = kmalloc(sizeof(*exts) * max, GFP_NOFS);
4164 if (!exts)
4165 return -ENOMEM;
4166 }
4167
4168 path = btrfs_alloc_path();
4169 BUG_ON(!path);
4170
4171 cur_pos = extent_key->objectid - offset;
4172 last_byte = extent_key->objectid + extent_key->offset;
4173 ret = btrfs_lookup_file_extent(NULL, root, path, reloc_inode->i_ino,
4174 cur_pos, 0);
4175 if (ret < 0)
4176 goto out;
4177 if (ret > 0) {
4178 ret = -ENOENT;
4179 goto out;
4180 }
4181
4182 while (1) {
4183 leaf = path->nodes[0];
4184 nritems = btrfs_header_nritems(leaf);
4185 if (path->slots[0] >= nritems) {
4186 ret = btrfs_next_leaf(root, path);
4187 if (ret < 0)
4188 goto out;
4189 if (ret > 0)
4190 break;
4191 leaf = path->nodes[0];
4192 }
4193
4194 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
4195 if (found_key.offset != cur_pos ||
4196 found_key.type != BTRFS_EXTENT_DATA_KEY ||
4197 found_key.objectid != reloc_inode->i_ino)
4198 break;
4199
4200 fi = btrfs_item_ptr(leaf, path->slots[0],
4201 struct btrfs_file_extent_item);
4202 if (btrfs_file_extent_type(leaf, fi) !=
4203 BTRFS_FILE_EXTENT_REG ||
4204 btrfs_file_extent_disk_bytenr(leaf, fi) == 0)
4205 break;
4206
4207 if (nr == max) {
4208 struct disk_extent *old = exts;
4209 max *= 2;
4210 exts = kzalloc(sizeof(*exts) * max, GFP_NOFS);
4211 memcpy(exts, old, sizeof(*exts) * nr);
4212 if (old != *extents)
4213 kfree(old);
4214 }
4215
4216 exts[nr].disk_bytenr =
4217 btrfs_file_extent_disk_bytenr(leaf, fi);
4218 exts[nr].disk_num_bytes =
4219 btrfs_file_extent_disk_num_bytes(leaf, fi);
4220 exts[nr].offset = btrfs_file_extent_offset(leaf, fi);
4221 exts[nr].num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
4222 exts[nr].ram_bytes = btrfs_file_extent_ram_bytes(leaf, fi);
4223 exts[nr].compression = btrfs_file_extent_compression(leaf, fi);
4224 exts[nr].encryption = btrfs_file_extent_encryption(leaf, fi);
4225 exts[nr].other_encoding = btrfs_file_extent_other_encoding(leaf,
4226 fi);
4227 BUG_ON(exts[nr].offset > 0);
4228 BUG_ON(exts[nr].compression || exts[nr].encryption);
4229 BUG_ON(exts[nr].num_bytes != exts[nr].disk_num_bytes);
4230
4231 cur_pos += exts[nr].num_bytes;
4232 nr++;
4233
4234 if (cur_pos + offset >= last_byte)
4235 break;
4236
4237 if (no_fragment) {
4238 ret = 1;
4239 goto out;
4240 }
4241 path->slots[0]++;
4242 }
4243
4244 BUG_ON(cur_pos + offset > last_byte);
4245 if (cur_pos + offset < last_byte) {
4246 ret = -ENOENT;
4247 goto out;
4248 }
4249 ret = 0;
4250 out:
4251 btrfs_free_path(path);
4252 if (ret) {
4253 if (exts != *extents)
4254 kfree(exts);
4255 } else {
4256 *extents = exts;
4257 *nr_extents = nr;
4258 }
4259 return ret;
4260 }
4261
4262 static noinline int replace_one_extent(struct btrfs_trans_handle *trans,
4263 struct btrfs_root *root,
4264 struct btrfs_path *path,
4265 struct btrfs_key *extent_key,
4266 struct btrfs_key *leaf_key,
4267 struct btrfs_ref_path *ref_path,
4268 struct disk_extent *new_extents,
4269 int nr_extents)
4270 {
4271 struct extent_buffer *leaf;
4272 struct btrfs_file_extent_item *fi;
4273 struct inode *inode = NULL;
4274 struct btrfs_key key;
4275 u64 lock_start = 0;
4276 u64 lock_end = 0;
4277 u64 num_bytes;
4278 u64 ext_offset;
4279 u64 search_end = (u64)-1;
4280 u32 nritems;
4281 int nr_scaned = 0;
4282 int extent_locked = 0;
4283 int extent_type;
4284 int ret;
4285
4286 memcpy(&key, leaf_key, sizeof(key));
4287 if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS) {
4288 if (key.objectid < ref_path->owner_objectid ||
4289 (key.objectid == ref_path->owner_objectid &&
4290 key.type < BTRFS_EXTENT_DATA_KEY)) {
4291 key.objectid = ref_path->owner_objectid;
4292 key.type = BTRFS_EXTENT_DATA_KEY;
4293 key.offset = 0;
4294 }
4295 }
4296
4297 while (1) {
4298 ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
4299 if (ret < 0)
4300 goto out;
4301
4302 leaf = path->nodes[0];
4303 nritems = btrfs_header_nritems(leaf);
4304 next:
4305 if (extent_locked && ret > 0) {
4306 /*
4307 * the file extent item was modified by someone
4308 * before the extent got locked.
4309 */
4310 unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
4311 lock_end, GFP_NOFS);
4312 extent_locked = 0;
4313 }
4314
4315 if (path->slots[0] >= nritems) {
4316 if (++nr_scaned > 2)
4317 break;
4318
4319 BUG_ON(extent_locked);
4320 ret = btrfs_next_leaf(root, path);
4321 if (ret < 0)
4322 goto out;
4323 if (ret > 0)
4324 break;
4325 leaf = path->nodes[0];
4326 nritems = btrfs_header_nritems(leaf);
4327 }
4328
4329 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
4330
4331 if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS) {
4332 if ((key.objectid > ref_path->owner_objectid) ||
4333 (key.objectid == ref_path->owner_objectid &&
4334 key.type > BTRFS_EXTENT_DATA_KEY) ||
4335 key.offset >= search_end)
4336 break;
4337 }
4338
4339 if (inode && key.objectid != inode->i_ino) {
4340 BUG_ON(extent_locked);
4341 btrfs_release_path(root, path);
4342 mutex_unlock(&inode->i_mutex);
4343 iput(inode);
4344 inode = NULL;
4345 continue;
4346 }
4347
4348 if (key.type != BTRFS_EXTENT_DATA_KEY) {
4349 path->slots[0]++;
4350 ret = 1;
4351 goto next;
4352 }
4353 fi = btrfs_item_ptr(leaf, path->slots[0],
4354 struct btrfs_file_extent_item);
4355 extent_type = btrfs_file_extent_type(leaf, fi);
4356 if ((extent_type != BTRFS_FILE_EXTENT_REG &&
4357 extent_type != BTRFS_FILE_EXTENT_PREALLOC) ||
4358 (btrfs_file_extent_disk_bytenr(leaf, fi) !=
4359 extent_key->objectid)) {
4360 path->slots[0]++;
4361 ret = 1;
4362 goto next;
4363 }
4364
4365 num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
4366 ext_offset = btrfs_file_extent_offset(leaf, fi);
4367
4368 if (search_end == (u64)-1) {
4369 search_end = key.offset - ext_offset +
4370 btrfs_file_extent_ram_bytes(leaf, fi);
4371 }
4372
4373 if (!extent_locked) {
4374 lock_start = key.offset;
4375 lock_end = lock_start + num_bytes - 1;
4376 } else {
4377 if (lock_start > key.offset ||
4378 lock_end + 1 < key.offset + num_bytes) {
4379 unlock_extent(&BTRFS_I(inode)->io_tree,
4380 lock_start, lock_end, GFP_NOFS);
4381 extent_locked = 0;
4382 }
4383 }
4384
4385 if (!inode) {
4386 btrfs_release_path(root, path);
4387
4388 inode = btrfs_iget_locked(root->fs_info->sb,
4389 key.objectid, root);
4390 if (inode->i_state & I_NEW) {
4391 BTRFS_I(inode)->root = root;
4392 BTRFS_I(inode)->location.objectid =
4393 key.objectid;
4394 BTRFS_I(inode)->location.type =
4395 BTRFS_INODE_ITEM_KEY;
4396 BTRFS_I(inode)->location.offset = 0;
4397 btrfs_read_locked_inode(inode);
4398 unlock_new_inode(inode);
4399 }
4400 /*
4401 * some code call btrfs_commit_transaction while
4402 * holding the i_mutex, so we can't use mutex_lock
4403 * here.
4404 */
4405 if (is_bad_inode(inode) ||
4406 !mutex_trylock(&inode->i_mutex)) {
4407 iput(inode);
4408 inode = NULL;
4409 key.offset = (u64)-1;
4410 goto skip;
4411 }
4412 }
4413
4414 if (!extent_locked) {
4415 struct btrfs_ordered_extent *ordered;
4416
4417 btrfs_release_path(root, path);
4418
4419 lock_extent(&BTRFS_I(inode)->io_tree, lock_start,
4420 lock_end, GFP_NOFS);
4421 ordered = btrfs_lookup_first_ordered_extent(inode,
4422 lock_end);
4423 if (ordered &&
4424 ordered->file_offset <= lock_end &&
4425 ordered->file_offset + ordered->len > lock_start) {
4426 unlock_extent(&BTRFS_I(inode)->io_tree,
4427 lock_start, lock_end, GFP_NOFS);
4428 btrfs_start_ordered_extent(inode, ordered, 1);
4429 btrfs_put_ordered_extent(ordered);
4430 key.offset += num_bytes;
4431 goto skip;
4432 }
4433 if (ordered)
4434 btrfs_put_ordered_extent(ordered);
4435
4436 extent_locked = 1;
4437 continue;
4438 }
4439
4440 if (nr_extents == 1) {
4441 /* update extent pointer in place */
4442 btrfs_set_file_extent_disk_bytenr(leaf, fi,
4443 new_extents[0].disk_bytenr);
4444 btrfs_set_file_extent_disk_num_bytes(leaf, fi,
4445 new_extents[0].disk_num_bytes);
4446 btrfs_mark_buffer_dirty(leaf);
4447
4448 btrfs_drop_extent_cache(inode, key.offset,
4449 key.offset + num_bytes - 1, 0);
4450
4451 ret = btrfs_inc_extent_ref(trans, root,
4452 new_extents[0].disk_bytenr,
4453 new_extents[0].disk_num_bytes,
4454 leaf->start,
4455 root->root_key.objectid,
4456 trans->transid,
4457 key.objectid);
4458 BUG_ON(ret);
4459
4460 ret = btrfs_free_extent(trans, root,
4461 extent_key->objectid,
4462 extent_key->offset,
4463 leaf->start,
4464 btrfs_header_owner(leaf),
4465 btrfs_header_generation(leaf),
4466 key.objectid, 0);
4467 BUG_ON(ret);
4468
4469 btrfs_release_path(root, path);
4470 key.offset += num_bytes;
4471 } else {
4472 BUG_ON(1);
4473 #if 0
4474 u64 alloc_hint;
4475 u64 extent_len;
4476 int i;
4477 /*
4478 * drop old extent pointer at first, then insert the
4479 * new pointers one bye one
4480 */
4481 btrfs_release_path(root, path);
4482 ret = btrfs_drop_extents(trans, root, inode, key.offset,
4483 key.offset + num_bytes,
4484 key.offset, &alloc_hint);
4485 BUG_ON(ret);
4486
4487 for (i = 0; i < nr_extents; i++) {
4488 if (ext_offset >= new_extents[i].num_bytes) {
4489 ext_offset -= new_extents[i].num_bytes;
4490 continue;
4491 }
4492 extent_len = min(new_extents[i].num_bytes -
4493 ext_offset, num_bytes);
4494
4495 ret = btrfs_insert_empty_item(trans, root,
4496 path, &key,
4497 sizeof(*fi));
4498 BUG_ON(ret);
4499
4500 leaf = path->nodes[0];
4501 fi = btrfs_item_ptr(leaf, path->slots[0],
4502 struct btrfs_file_extent_item);
4503 btrfs_set_file_extent_generation(leaf, fi,
4504 trans->transid);
4505 btrfs_set_file_extent_type(leaf, fi,
4506 BTRFS_FILE_EXTENT_REG);
4507 btrfs_set_file_extent_disk_bytenr(leaf, fi,
4508 new_extents[i].disk_bytenr);
4509 btrfs_set_file_extent_disk_num_bytes(leaf, fi,
4510 new_extents[i].disk_num_bytes);
4511 btrfs_set_file_extent_ram_bytes(leaf, fi,
4512 new_extents[i].ram_bytes);
4513
4514 btrfs_set_file_extent_compression(leaf, fi,
4515 new_extents[i].compression);
4516 btrfs_set_file_extent_encryption(leaf, fi,
4517 new_extents[i].encryption);
4518 btrfs_set_file_extent_other_encoding(leaf, fi,
4519 new_extents[i].other_encoding);
4520
4521 btrfs_set_file_extent_num_bytes(leaf, fi,
4522 extent_len);
4523 ext_offset += new_extents[i].offset;
4524 btrfs_set_file_extent_offset(leaf, fi,
4525 ext_offset);
4526 btrfs_mark_buffer_dirty(leaf);
4527
4528 btrfs_drop_extent_cache(inode, key.offset,
4529 key.offset + extent_len - 1, 0);
4530
4531 ret = btrfs_inc_extent_ref(trans, root,
4532 new_extents[i].disk_bytenr,
4533 new_extents[i].disk_num_bytes,
4534 leaf->start,
4535 root->root_key.objectid,
4536 trans->transid, key.objectid);
4537 BUG_ON(ret);
4538 btrfs_release_path(root, path);
4539
4540 inode_add_bytes(inode, extent_len);
4541
4542 ext_offset = 0;
4543 num_bytes -= extent_len;
4544 key.offset += extent_len;
4545
4546 if (num_bytes == 0)
4547 break;
4548 }
4549 BUG_ON(i >= nr_extents);
4550 #endif
4551 }
4552
4553 if (extent_locked) {
4554 unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
4555 lock_end, GFP_NOFS);
4556 extent_locked = 0;
4557 }
4558 skip:
4559 if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS &&
4560 key.offset >= search_end)
4561 break;
4562
4563 cond_resched();
4564 }
4565 ret = 0;
4566 out:
4567 btrfs_release_path(root, path);
4568 if (inode) {
4569 mutex_unlock(&inode->i_mutex);
4570 if (extent_locked) {
4571 unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
4572 lock_end, GFP_NOFS);
4573 }
4574 iput(inode);
4575 }
4576 return ret;
4577 }
4578
4579 int btrfs_reloc_tree_cache_ref(struct btrfs_trans_handle *trans,
4580 struct btrfs_root *root,
4581 struct extent_buffer *buf, u64 orig_start)
4582 {
4583 int level;
4584 int ret;
4585
4586 BUG_ON(btrfs_header_generation(buf) != trans->transid);
4587 BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
4588
4589 level = btrfs_header_level(buf);
4590 if (level == 0) {
4591 struct btrfs_leaf_ref *ref;
4592 struct btrfs_leaf_ref *orig_ref;
4593
4594 orig_ref = btrfs_lookup_leaf_ref(root, orig_start);
4595 if (!orig_ref)
4596 return -ENOENT;
4597
4598 ref = btrfs_alloc_leaf_ref(root, orig_ref->nritems);
4599 if (!ref) {
4600 btrfs_free_leaf_ref(root, orig_ref);
4601 return -ENOMEM;
4602 }
4603
4604 ref->nritems = orig_ref->nritems;
4605 memcpy(ref->extents, orig_ref->extents,
4606 sizeof(ref->extents[0]) * ref->nritems);
4607
4608 btrfs_free_leaf_ref(root, orig_ref);
4609
4610 ref->root_gen = trans->transid;
4611 ref->bytenr = buf->start;
4612 ref->owner = btrfs_header_owner(buf);
4613 ref->generation = btrfs_header_generation(buf);
4614
4615 ret = btrfs_add_leaf_ref(root, ref, 0);
4616 WARN_ON(ret);
4617 btrfs_free_leaf_ref(root, ref);
4618 }
4619 return 0;
4620 }
4621
4622 static noinline int invalidate_extent_cache(struct btrfs_root *root,
4623 struct extent_buffer *leaf,
4624 struct btrfs_block_group_cache *group,
4625 struct btrfs_root *target_root)
4626 {
4627 struct btrfs_key key;
4628 struct inode *inode = NULL;
4629 struct btrfs_file_extent_item *fi;
4630 u64 num_bytes;
4631 u64 skip_objectid = 0;
4632 u32 nritems;
4633 u32 i;
4634
4635 nritems = btrfs_header_nritems(leaf);
4636 for (i = 0; i < nritems; i++) {
4637 btrfs_item_key_to_cpu(leaf, &key, i);
4638 if (key.objectid == skip_objectid ||
4639 key.type != BTRFS_EXTENT_DATA_KEY)
4640 continue;
4641 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
4642 if (btrfs_file_extent_type(leaf, fi) ==
4643 BTRFS_FILE_EXTENT_INLINE)
4644 continue;
4645 if (btrfs_file_extent_disk_bytenr(leaf, fi) == 0)
4646 continue;
4647 if (!inode || inode->i_ino != key.objectid) {
4648 iput(inode);
4649 inode = btrfs_ilookup(target_root->fs_info->sb,
4650 key.objectid, target_root, 1);
4651 }
4652 if (!inode) {
4653 skip_objectid = key.objectid;
4654 continue;
4655 }
4656 num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
4657
4658 lock_extent(&BTRFS_I(inode)->io_tree, key.offset,
4659 key.offset + num_bytes - 1, GFP_NOFS);
4660 btrfs_drop_extent_cache(inode, key.offset,
4661 key.offset + num_bytes - 1, 1);
4662 unlock_extent(&BTRFS_I(inode)->io_tree, key.offset,
4663 key.offset + num_bytes - 1, GFP_NOFS);
4664 cond_resched();
4665 }
4666 iput(inode);
4667 return 0;
4668 }
4669
4670 static noinline int replace_extents_in_leaf(struct btrfs_trans_handle *trans,
4671 struct btrfs_root *root,
4672 struct extent_buffer *leaf,
4673 struct btrfs_block_group_cache *group,
4674 struct inode *reloc_inode)
4675 {
4676 struct btrfs_key key;
4677 struct btrfs_key extent_key;
4678 struct btrfs_file_extent_item *fi;
4679 struct btrfs_leaf_ref *ref;
4680 struct disk_extent *new_extent;
4681 u64 bytenr;
4682 u64 num_bytes;
4683 u32 nritems;
4684 u32 i;
4685 int ext_index;
4686 int nr_extent;
4687 int ret;
4688
4689 new_extent = kmalloc(sizeof(*new_extent), GFP_NOFS);
4690 BUG_ON(!new_extent);
4691
4692 ref = btrfs_lookup_leaf_ref(root, leaf->start);
4693 BUG_ON(!ref);
4694
4695 ext_index = -1;
4696 nritems = btrfs_header_nritems(leaf);
4697 for (i = 0; i < nritems; i++) {
4698 btrfs_item_key_to_cpu(leaf, &key, i);
4699 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
4700 continue;
4701 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
4702 if (btrfs_file_extent_type(leaf, fi) ==
4703 BTRFS_FILE_EXTENT_INLINE)
4704 continue;
4705 bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
4706 num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
4707 if (bytenr == 0)
4708 continue;
4709
4710 ext_index++;
4711 if (bytenr >= group->key.objectid + group->key.offset ||
4712 bytenr + num_bytes <= group->key.objectid)
4713 continue;
4714
4715 extent_key.objectid = bytenr;
4716 extent_key.offset = num_bytes;
4717 extent_key.type = BTRFS_EXTENT_ITEM_KEY;
4718 nr_extent = 1;
4719 ret = get_new_locations(reloc_inode, &extent_key,
4720 group->key.objectid, 1,
4721 &new_extent, &nr_extent);
4722 if (ret > 0)
4723 continue;
4724 BUG_ON(ret < 0);
4725
4726 BUG_ON(ref->extents[ext_index].bytenr != bytenr);
4727 BUG_ON(ref->extents[ext_index].num_bytes != num_bytes);
4728 ref->extents[ext_index].bytenr = new_extent->disk_bytenr;
4729 ref->extents[ext_index].num_bytes = new_extent->disk_num_bytes;
4730
4731 btrfs_set_file_extent_disk_bytenr(leaf, fi,
4732 new_extent->disk_bytenr);
4733 btrfs_set_file_extent_disk_num_bytes(leaf, fi,
4734 new_extent->disk_num_bytes);
4735 btrfs_mark_buffer_dirty(leaf);
4736
4737 ret = btrfs_inc_extent_ref(trans, root,
4738 new_extent->disk_bytenr,
4739 new_extent->disk_num_bytes,
4740 leaf->start,
4741 root->root_key.objectid,
4742 trans->transid, key.objectid);
4743 BUG_ON(ret);
4744
4745 ret = btrfs_free_extent(trans, root,
4746 bytenr, num_bytes, leaf->start,
4747 btrfs_header_owner(leaf),
4748 btrfs_header_generation(leaf),
4749 key.objectid, 0);
4750 BUG_ON(ret);
4751 cond_resched();
4752 }
4753 kfree(new_extent);
4754 BUG_ON(ext_index + 1 != ref->nritems);
4755 btrfs_free_leaf_ref(root, ref);
4756 return 0;
4757 }
4758
4759 int btrfs_free_reloc_root(struct btrfs_trans_handle *trans,
4760 struct btrfs_root *root)
4761 {
4762 struct btrfs_root *reloc_root;
4763 int ret;
4764
4765 if (root->reloc_root) {
4766 reloc_root = root->reloc_root;
4767 root->reloc_root = NULL;
4768 list_add(&reloc_root->dead_list,
4769 &root->fs_info->dead_reloc_roots);
4770
4771 btrfs_set_root_bytenr(&reloc_root->root_item,
4772 reloc_root->node->start);
4773 btrfs_set_root_level(&root->root_item,
4774 btrfs_header_level(reloc_root->node));
4775 memset(&reloc_root->root_item.drop_progress, 0,
4776 sizeof(struct btrfs_disk_key));
4777 reloc_root->root_item.drop_level = 0;
4778
4779 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4780 &reloc_root->root_key,
4781 &reloc_root->root_item);
4782 BUG_ON(ret);
4783 }
4784 return 0;
4785 }
4786
4787 int btrfs_drop_dead_reloc_roots(struct btrfs_root *root)
4788 {
4789 struct btrfs_trans_handle *trans;
4790 struct btrfs_root *reloc_root;
4791 struct btrfs_root *prev_root = NULL;
4792 struct list_head dead_roots;
4793 int ret;
4794 unsigned long nr;
4795
4796 INIT_LIST_HEAD(&dead_roots);
4797 list_splice_init(&root->fs_info->dead_reloc_roots, &dead_roots);
4798
4799 while (!list_empty(&dead_roots)) {
4800 reloc_root = list_entry(dead_roots.prev,
4801 struct btrfs_root, dead_list);
4802 list_del_init(&reloc_root->dead_list);
4803
4804 BUG_ON(reloc_root->commit_root != NULL);
4805 while (1) {
4806 trans = btrfs_join_transaction(root, 1);
4807 BUG_ON(!trans);
4808
4809 mutex_lock(&root->fs_info->drop_mutex);
4810 ret = btrfs_drop_snapshot(trans, reloc_root);
4811 if (ret != -EAGAIN)
4812 break;
4813 mutex_unlock(&root->fs_info->drop_mutex);
4814
4815 nr = trans->blocks_used;
4816 ret = btrfs_end_transaction(trans, root);
4817 BUG_ON(ret);
4818 btrfs_btree_balance_dirty(root, nr);
4819 }
4820
4821 free_extent_buffer(reloc_root->node);
4822
4823 ret = btrfs_del_root(trans, root->fs_info->tree_root,
4824 &reloc_root->root_key);
4825 BUG_ON(ret);
4826 mutex_unlock(&root->fs_info->drop_mutex);
4827
4828 nr = trans->blocks_used;
4829 ret = btrfs_end_transaction(trans, root);
4830 BUG_ON(ret);
4831 btrfs_btree_balance_dirty(root, nr);
4832
4833 kfree(prev_root);
4834 prev_root = reloc_root;
4835 }
4836 if (prev_root) {
4837 btrfs_remove_leaf_refs(prev_root, (u64)-1, 0);
4838 kfree(prev_root);
4839 }
4840 return 0;
4841 }
4842
4843 int btrfs_add_dead_reloc_root(struct btrfs_root *root)
4844 {
4845 list_add(&root->dead_list, &root->fs_info->dead_reloc_roots);
4846 return 0;
4847 }
4848
4849 int btrfs_cleanup_reloc_trees(struct btrfs_root *root)
4850 {
4851 struct btrfs_root *reloc_root;
4852 struct btrfs_trans_handle *trans;
4853 struct btrfs_key location;
4854 int found;
4855 int ret;
4856
4857 mutex_lock(&root->fs_info->tree_reloc_mutex);
4858 ret = btrfs_find_dead_roots(root, BTRFS_TREE_RELOC_OBJECTID, NULL);
4859 BUG_ON(ret);
4860 found = !list_empty(&root->fs_info->dead_reloc_roots);
4861 mutex_unlock(&root->fs_info->tree_reloc_mutex);
4862
4863 if (found) {
4864 trans = btrfs_start_transaction(root, 1);
4865 BUG_ON(!trans);
4866 ret = btrfs_commit_transaction(trans, root);
4867 BUG_ON(ret);
4868 }
4869
4870 location.objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
4871 location.offset = (u64)-1;
4872 location.type = BTRFS_ROOT_ITEM_KEY;
4873
4874 reloc_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
4875 BUG_ON(!reloc_root);
4876 btrfs_orphan_cleanup(reloc_root);
4877 return 0;
4878 }
4879
4880 static noinline int init_reloc_tree(struct btrfs_trans_handle *trans,
4881 struct btrfs_root *root)
4882 {
4883 struct btrfs_root *reloc_root;
4884 struct extent_buffer *eb;
4885 struct btrfs_root_item *root_item;
4886 struct btrfs_key root_key;
4887 int ret;
4888
4889 BUG_ON(!root->ref_cows);
4890 if (root->reloc_root)
4891 return 0;
4892
4893 root_item = kmalloc(sizeof(*root_item), GFP_NOFS);
4894 BUG_ON(!root_item);
4895
4896 ret = btrfs_copy_root(trans, root, root->commit_root,
4897 &eb, BTRFS_TREE_RELOC_OBJECTID);
4898 BUG_ON(ret);
4899
4900 root_key.objectid = BTRFS_TREE_RELOC_OBJECTID;
4901 root_key.offset = root->root_key.objectid;
4902 root_key.type = BTRFS_ROOT_ITEM_KEY;
4903
4904 memcpy(root_item, &root->root_item, sizeof(root_item));
4905 btrfs_set_root_refs(root_item, 0);
4906 btrfs_set_root_bytenr(root_item, eb->start);
4907 btrfs_set_root_level(root_item, btrfs_header_level(eb));
4908 btrfs_set_root_generation(root_item, trans->transid);
4909
4910 btrfs_tree_unlock(eb);
4911 free_extent_buffer(eb);
4912
4913 ret = btrfs_insert_root(trans, root->fs_info->tree_root,
4914 &root_key, root_item);
4915 BUG_ON(ret);
4916 kfree(root_item);
4917
4918 reloc_root = btrfs_read_fs_root_no_radix(root->fs_info->tree_root,
4919 &root_key);
4920 BUG_ON(!reloc_root);
4921 reloc_root->last_trans = trans->transid;
4922 reloc_root->commit_root = NULL;
4923 reloc_root->ref_tree = &root->fs_info->reloc_ref_tree;
4924
4925 root->reloc_root = reloc_root;
4926 return 0;
4927 }
4928
4929 /*
4930 * Core function of space balance.
4931 *
4932 * The idea is using reloc trees to relocate tree blocks in reference
4933 * counted roots. There is one reloc tree for each subvol, and all
4934 * reloc trees share same root key objectid. Reloc trees are snapshots
4935 * of the latest committed roots of subvols (root->commit_root).
4936 *
4937 * To relocate a tree block referenced by a subvol, there are two steps.
4938 * COW the block through subvol's reloc tree, then update block pointer
4939 * in the subvol to point to the new block. Since all reloc trees share
4940 * same root key objectid, doing special handing for tree blocks owned
4941 * by them is easy. Once a tree block has been COWed in one reloc tree,
4942 * we can use the resulting new block directly when the same block is
4943 * required to COW again through other reloc trees. By this way, relocated
4944 * tree blocks are shared between reloc trees, so they are also shared
4945 * between subvols.
4946 */
4947 static noinline int relocate_one_path(struct btrfs_trans_handle *trans,
4948 struct btrfs_root *root,
4949 struct btrfs_path *path,
4950 struct btrfs_key *first_key,
4951 struct btrfs_ref_path *ref_path,
4952 struct btrfs_block_group_cache *group,
4953 struct inode *reloc_inode)
4954 {
4955 struct btrfs_root *reloc_root;
4956 struct extent_buffer *eb = NULL;
4957 struct btrfs_key *keys;
4958 u64 *nodes;
4959 int level;
4960 int shared_level;
4961 int lowest_level = 0;
4962 int ret;
4963
4964 if (ref_path->owner_objectid < BTRFS_FIRST_FREE_OBJECTID)
4965 lowest_level = ref_path->owner_objectid;
4966
4967 if (!root->ref_cows) {
4968 path->lowest_level = lowest_level;
4969 ret = btrfs_search_slot(trans, root, first_key, path, 0, 1);
4970 BUG_ON(ret < 0);
4971 path->lowest_level = 0;
4972 btrfs_release_path(root, path);
4973 return 0;
4974 }
4975
4976 mutex_lock(&root->fs_info->tree_reloc_mutex);
4977 ret = init_reloc_tree(trans, root);
4978 BUG_ON(ret);
4979 reloc_root = root->reloc_root;
4980
4981 shared_level = ref_path->shared_level;
4982 ref_path->shared_level = BTRFS_MAX_LEVEL - 1;
4983
4984 keys = ref_path->node_keys;
4985 nodes = ref_path->new_nodes;
4986 memset(&keys[shared_level + 1], 0,
4987 sizeof(*keys) * (BTRFS_MAX_LEVEL - shared_level - 1));
4988 memset(&nodes[shared_level + 1], 0,
4989 sizeof(*nodes) * (BTRFS_MAX_LEVEL - shared_level - 1));
4990
4991 if (nodes[lowest_level] == 0) {
4992 path->lowest_level = lowest_level;
4993 ret = btrfs_search_slot(trans, reloc_root, first_key, path,
4994 0, 1);
4995 BUG_ON(ret);
4996 for (level = lowest_level; level < BTRFS_MAX_LEVEL; level++) {
4997 eb = path->nodes[level];
4998 if (!eb || eb == reloc_root->node)
4999 break;
5000 nodes[level] = eb->start;
5001 if (level == 0)
5002 btrfs_item_key_to_cpu(eb, &keys[level], 0);
5003 else
5004 btrfs_node_key_to_cpu(eb, &keys[level], 0);
5005 }
5006 if (nodes[0] &&
5007 ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
5008 eb = path->nodes[0];
5009 ret = replace_extents_in_leaf(trans, reloc_root, eb,
5010 group, reloc_inode);
5011 BUG_ON(ret);
5012 }
5013 btrfs_release_path(reloc_root, path);
5014 } else {
5015 ret = btrfs_merge_path(trans, reloc_root, keys, nodes,
5016 lowest_level);
5017 BUG_ON(ret);
5018 }
5019
5020 /*
5021 * replace tree blocks in the fs tree with tree blocks in
5022 * the reloc tree.
5023 */
5024 ret = btrfs_merge_path(trans, root, keys, nodes, lowest_level);
5025 BUG_ON(ret < 0);
5026
5027 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
5028 ret = btrfs_search_slot(trans, reloc_root, first_key, path,
5029 0, 0);
5030 BUG_ON(ret);
5031 extent_buffer_get(path->nodes[0]);
5032 eb = path->nodes[0];
5033 btrfs_release_path(reloc_root, path);
5034 ret = invalidate_extent_cache(reloc_root, eb, group, root);
5035 BUG_ON(ret);
5036 free_extent_buffer(eb);
5037 }
5038
5039 mutex_unlock(&root->fs_info->tree_reloc_mutex);
5040 path->lowest_level = 0;
5041 return 0;
5042 }
5043
5044 static noinline int relocate_tree_block(struct btrfs_trans_handle *trans,
5045 struct btrfs_root *root,
5046 struct btrfs_path *path,
5047 struct btrfs_key *first_key,
5048 struct btrfs_ref_path *ref_path)
5049 {
5050 int ret;
5051
5052 ret = relocate_one_path(trans, root, path, first_key,
5053 ref_path, NULL, NULL);
5054 BUG_ON(ret);
5055
5056 return 0;
5057 }
5058
5059 static noinline int del_extent_zero(struct btrfs_trans_handle *trans,
5060 struct btrfs_root *extent_root,
5061 struct btrfs_path *path,
5062 struct btrfs_key *extent_key)
5063 {
5064 int ret;
5065
5066 ret = btrfs_search_slot(trans, extent_root, extent_key, path, -1, 1);
5067 if (ret)
5068 goto out;
5069 ret = btrfs_del_item(trans, extent_root, path);
5070 out:
5071 btrfs_release_path(extent_root, path);
5072 return ret;
5073 }
5074
5075 static noinline struct btrfs_root *read_ref_root(struct btrfs_fs_info *fs_info,
5076 struct btrfs_ref_path *ref_path)
5077 {
5078 struct btrfs_key root_key;
5079
5080 root_key.objectid = ref_path->root_objectid;
5081 root_key.type = BTRFS_ROOT_ITEM_KEY;
5082 if (is_cowonly_root(ref_path->root_objectid))
5083 root_key.offset = 0;
5084 else
5085 root_key.offset = (u64)-1;
5086
5087 return btrfs_read_fs_root_no_name(fs_info, &root_key);
5088 }
5089
5090 static noinline int relocate_one_extent(struct btrfs_root *extent_root,
5091 struct btrfs_path *path,
5092 struct btrfs_key *extent_key,
5093 struct btrfs_block_group_cache *group,
5094 struct inode *reloc_inode, int pass)
5095 {
5096 struct btrfs_trans_handle *trans;
5097 struct btrfs_root *found_root;
5098 struct btrfs_ref_path *ref_path = NULL;
5099 struct disk_extent *new_extents = NULL;
5100 int nr_extents = 0;
5101 int loops;
5102 int ret;
5103 int level;
5104 struct btrfs_key first_key;
5105 u64 prev_block = 0;
5106
5107
5108 trans = btrfs_start_transaction(extent_root, 1);
5109 BUG_ON(!trans);
5110
5111 if (extent_key->objectid == 0) {
5112 ret = del_extent_zero(trans, extent_root, path, extent_key);
5113 goto out;
5114 }
5115
5116 ref_path = kmalloc(sizeof(*ref_path), GFP_NOFS);
5117 if (!ref_path) {
5118 ret = -ENOMEM;
5119 goto out;
5120 }
5121
5122 for (loops = 0; ; loops++) {
5123 if (loops == 0) {
5124 ret = btrfs_first_ref_path(trans, extent_root, ref_path,
5125 extent_key->objectid);
5126 } else {
5127 ret = btrfs_next_ref_path(trans, extent_root, ref_path);
5128 }
5129 if (ret < 0)
5130 goto out;
5131 if (ret > 0)
5132 break;
5133
5134 if (ref_path->root_objectid == BTRFS_TREE_LOG_OBJECTID ||
5135 ref_path->root_objectid == BTRFS_TREE_RELOC_OBJECTID)
5136 continue;
5137
5138 found_root = read_ref_root(extent_root->fs_info, ref_path);
5139 BUG_ON(!found_root);
5140 /*
5141 * for reference counted tree, only process reference paths
5142 * rooted at the latest committed root.
5143 */
5144 if (found_root->ref_cows &&
5145 ref_path->root_generation != found_root->root_key.offset)
5146 continue;
5147
5148 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
5149 if (pass == 0) {
5150 /*
5151 * copy data extents to new locations
5152 */
5153 u64 group_start = group->key.objectid;
5154 ret = relocate_data_extent(reloc_inode,
5155 extent_key,
5156 group_start);
5157 if (ret < 0)
5158 goto out;
5159 break;
5160 }
5161 level = 0;
5162 } else {
5163 level = ref_path->owner_objectid;
5164 }
5165
5166 if (prev_block != ref_path->nodes[level]) {
5167 struct extent_buffer *eb;
5168 u64 block_start = ref_path->nodes[level];
5169 u64 block_size = btrfs_level_size(found_root, level);
5170
5171 eb = read_tree_block(found_root, block_start,
5172 block_size, 0);
5173 btrfs_tree_lock(eb);
5174 BUG_ON(level != btrfs_header_level(eb));
5175
5176 if (level == 0)
5177 btrfs_item_key_to_cpu(eb, &first_key, 0);
5178 else
5179 btrfs_node_key_to_cpu(eb, &first_key, 0);
5180
5181 btrfs_tree_unlock(eb);
5182 free_extent_buffer(eb);
5183 prev_block = block_start;
5184 }
5185
5186 mutex_lock(&extent_root->fs_info->trans_mutex);
5187 btrfs_record_root_in_trans(found_root);
5188 mutex_unlock(&extent_root->fs_info->trans_mutex);
5189 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
5190 /*
5191 * try to update data extent references while
5192 * keeping metadata shared between snapshots.
5193 */
5194 if (pass == 1) {
5195 ret = relocate_one_path(trans, found_root,
5196 path, &first_key, ref_path,
5197 group, reloc_inode);
5198 if (ret < 0)
5199 goto out;
5200 continue;
5201 }
5202 /*
5203 * use fallback method to process the remaining
5204 * references.
5205 */
5206 if (!new_extents) {
5207 u64 group_start = group->key.objectid;
5208 new_extents = kmalloc(sizeof(*new_extents),
5209 GFP_NOFS);
5210 nr_extents = 1;
5211 ret = get_new_locations(reloc_inode,
5212 extent_key,
5213 group_start, 1,
5214 &new_extents,
5215 &nr_extents);
5216 if (ret)
5217 goto out;
5218 }
5219 ret = replace_one_extent(trans, found_root,
5220 path, extent_key,
5221 &first_key, ref_path,
5222 new_extents, nr_extents);
5223 } else {
5224 ret = relocate_tree_block(trans, found_root, path,
5225 &first_key, ref_path);
5226 }
5227 if (ret < 0)
5228 goto out;
5229 }
5230 ret = 0;
5231 out:
5232 btrfs_end_transaction(trans, extent_root);
5233 kfree(new_extents);
5234 kfree(ref_path);
5235 return ret;
5236 }
5237
5238 static u64 update_block_group_flags(struct btrfs_root *root, u64 flags)
5239 {
5240 u64 num_devices;
5241 u64 stripped = BTRFS_BLOCK_GROUP_RAID0 |
5242 BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10;
5243
5244 num_devices = root->fs_info->fs_devices->rw_devices;
5245 if (num_devices == 1) {
5246 stripped |= BTRFS_BLOCK_GROUP_DUP;
5247 stripped = flags & ~stripped;
5248
5249 /* turn raid0 into single device chunks */
5250 if (flags & BTRFS_BLOCK_GROUP_RAID0)
5251 return stripped;
5252
5253 /* turn mirroring into duplication */
5254 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
5255 BTRFS_BLOCK_GROUP_RAID10))
5256 return stripped | BTRFS_BLOCK_GROUP_DUP;
5257 return flags;
5258 } else {
5259 /* they already had raid on here, just return */
5260 if (flags & stripped)
5261 return flags;
5262
5263 stripped |= BTRFS_BLOCK_GROUP_DUP;
5264 stripped = flags & ~stripped;
5265
5266 /* switch duplicated blocks with raid1 */
5267 if (flags & BTRFS_BLOCK_GROUP_DUP)
5268 return stripped | BTRFS_BLOCK_GROUP_RAID1;
5269
5270 /* turn single device chunks into raid0 */
5271 return stripped | BTRFS_BLOCK_GROUP_RAID0;
5272 }
5273 return flags;
5274 }
5275
5276 static int __alloc_chunk_for_shrink(struct btrfs_root *root,
5277 struct btrfs_block_group_cache *shrink_block_group,
5278 int force)
5279 {
5280 struct btrfs_trans_handle *trans;
5281 u64 new_alloc_flags;
5282 u64 calc;
5283
5284 spin_lock(&shrink_block_group->lock);
5285 if (btrfs_block_group_used(&shrink_block_group->item) > 0) {
5286 spin_unlock(&shrink_block_group->lock);
5287
5288 trans = btrfs_start_transaction(root, 1);
5289 spin_lock(&shrink_block_group->lock);
5290
5291 new_alloc_flags = update_block_group_flags(root,
5292 shrink_block_group->flags);
5293 if (new_alloc_flags != shrink_block_group->flags) {
5294 calc =
5295 btrfs_block_group_used(&shrink_block_group->item);
5296 } else {
5297 calc = shrink_block_group->key.offset;
5298 }
5299 spin_unlock(&shrink_block_group->lock);
5300
5301 do_chunk_alloc(trans, root->fs_info->extent_root,
5302 calc + 2 * 1024 * 1024, new_alloc_flags, force);
5303
5304 btrfs_end_transaction(trans, root);
5305 } else
5306 spin_unlock(&shrink_block_group->lock);
5307 return 0;
5308 }
5309
5310 static int __insert_orphan_inode(struct btrfs_trans_handle *trans,
5311 struct btrfs_root *root,
5312 u64 objectid, u64 size)
5313 {
5314 struct btrfs_path *path;
5315 struct btrfs_inode_item *item;
5316 struct extent_buffer *leaf;
5317 int ret;
5318
5319 path = btrfs_alloc_path();
5320 if (!path)
5321 return -ENOMEM;
5322
5323 ret = btrfs_insert_empty_inode(trans, root, path, objectid);
5324 if (ret)
5325 goto out;
5326
5327 leaf = path->nodes[0];
5328 item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_inode_item);
5329 memset_extent_buffer(leaf, 0, (unsigned long)item, sizeof(*item));
5330 btrfs_set_inode_generation(leaf, item, 1);
5331 btrfs_set_inode_size(leaf, item, size);
5332 btrfs_set_inode_mode(leaf, item, S_IFREG | 0600);
5333 btrfs_set_inode_flags(leaf, item, BTRFS_INODE_NOCOMPRESS);
5334 btrfs_mark_buffer_dirty(leaf);
5335 btrfs_release_path(root, path);
5336 out:
5337 btrfs_free_path(path);
5338 return ret;
5339 }
5340
5341 static noinline struct inode *create_reloc_inode(struct btrfs_fs_info *fs_info,
5342 struct btrfs_block_group_cache *group)
5343 {
5344 struct inode *inode = NULL;
5345 struct btrfs_trans_handle *trans;
5346 struct btrfs_root *root;
5347 struct btrfs_key root_key;
5348 u64 objectid = BTRFS_FIRST_FREE_OBJECTID;
5349 int err = 0;
5350
5351 root_key.objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
5352 root_key.type = BTRFS_ROOT_ITEM_KEY;
5353 root_key.offset = (u64)-1;
5354 root = btrfs_read_fs_root_no_name(fs_info, &root_key);
5355 if (IS_ERR(root))
5356 return ERR_CAST(root);
5357
5358 trans = btrfs_start_transaction(root, 1);
5359 BUG_ON(!trans);
5360
5361 err = btrfs_find_free_objectid(trans, root, objectid, &objectid);
5362 if (err)
5363 goto out;
5364
5365 err = __insert_orphan_inode(trans, root, objectid, group->key.offset);
5366 BUG_ON(err);
5367
5368 err = btrfs_insert_file_extent(trans, root, objectid, 0, 0, 0,
5369 group->key.offset, 0, group->key.offset,
5370 0, 0, 0);
5371 BUG_ON(err);
5372
5373 inode = btrfs_iget_locked(root->fs_info->sb, objectid, root);
5374 if (inode->i_state & I_NEW) {
5375 BTRFS_I(inode)->root = root;
5376 BTRFS_I(inode)->location.objectid = objectid;
5377 BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY;
5378 BTRFS_I(inode)->location.offset = 0;
5379 btrfs_read_locked_inode(inode);
5380 unlock_new_inode(inode);
5381 BUG_ON(is_bad_inode(inode));
5382 } else {
5383 BUG_ON(1);
5384 }
5385 BTRFS_I(inode)->index_cnt = group->key.objectid;
5386
5387 err = btrfs_orphan_add(trans, inode);
5388 out:
5389 btrfs_end_transaction(trans, root);
5390 if (err) {
5391 if (inode)
5392 iput(inode);
5393 inode = ERR_PTR(err);
5394 }
5395 return inode;
5396 }
5397
5398 int btrfs_reloc_clone_csums(struct inode *inode, u64 file_pos, u64 len)
5399 {
5400
5401 struct btrfs_ordered_sum *sums;
5402 struct btrfs_sector_sum *sector_sum;
5403 struct btrfs_ordered_extent *ordered;
5404 struct btrfs_root *root = BTRFS_I(inode)->root;
5405 struct list_head list;
5406 size_t offset;
5407 int ret;
5408 u64 disk_bytenr;
5409
5410 INIT_LIST_HEAD(&list);
5411
5412 ordered = btrfs_lookup_ordered_extent(inode, file_pos);
5413 BUG_ON(ordered->file_offset != file_pos || ordered->len != len);
5414
5415 disk_bytenr = file_pos + BTRFS_I(inode)->index_cnt;
5416 ret = btrfs_lookup_csums_range(root->fs_info->csum_root, disk_bytenr,
5417 disk_bytenr + len - 1, &list);
5418
5419 while (!list_empty(&list)) {
5420 sums = list_entry(list.next, struct btrfs_ordered_sum, list);
5421 list_del_init(&sums->list);
5422
5423 sector_sum = sums->sums;
5424 sums->bytenr = ordered->start;
5425
5426 offset = 0;
5427 while (offset < sums->len) {
5428 sector_sum->bytenr += ordered->start - disk_bytenr;
5429 sector_sum++;
5430 offset += root->sectorsize;
5431 }
5432
5433 btrfs_add_ordered_sum(inode, ordered, sums);
5434 }
5435 btrfs_put_ordered_extent(ordered);
5436 return 0;
5437 }
5438
5439 int btrfs_relocate_block_group(struct btrfs_root *root, u64 group_start)
5440 {
5441 struct btrfs_trans_handle *trans;
5442 struct btrfs_path *path;
5443 struct btrfs_fs_info *info = root->fs_info;
5444 struct extent_buffer *leaf;
5445 struct inode *reloc_inode;
5446 struct btrfs_block_group_cache *block_group;
5447 struct btrfs_key key;
5448 u64 skipped;
5449 u64 cur_byte;
5450 u64 total_found;
5451 u32 nritems;
5452 int ret;
5453 int progress;
5454 int pass = 0;
5455
5456 root = root->fs_info->extent_root;
5457
5458 block_group = btrfs_lookup_block_group(info, group_start);
5459 BUG_ON(!block_group);
5460
5461 printk(KERN_INFO "btrfs relocating block group %llu flags %llu\n",
5462 (unsigned long long)block_group->key.objectid,
5463 (unsigned long long)block_group->flags);
5464
5465 path = btrfs_alloc_path();
5466 BUG_ON(!path);
5467
5468 reloc_inode = create_reloc_inode(info, block_group);
5469 BUG_ON(IS_ERR(reloc_inode));
5470
5471 __alloc_chunk_for_shrink(root, block_group, 1);
5472 set_block_group_readonly(block_group);
5473
5474 btrfs_start_delalloc_inodes(info->tree_root);
5475 btrfs_wait_ordered_extents(info->tree_root, 0);
5476 again:
5477 skipped = 0;
5478 total_found = 0;
5479 progress = 0;
5480 key.objectid = block_group->key.objectid;
5481 key.offset = 0;
5482 key.type = 0;
5483 cur_byte = key.objectid;
5484
5485 trans = btrfs_start_transaction(info->tree_root, 1);
5486 btrfs_commit_transaction(trans, info->tree_root);
5487
5488 mutex_lock(&root->fs_info->cleaner_mutex);
5489 btrfs_clean_old_snapshots(info->tree_root);
5490 btrfs_remove_leaf_refs(info->tree_root, (u64)-1, 1);
5491 mutex_unlock(&root->fs_info->cleaner_mutex);
5492
5493 trans = btrfs_start_transaction(info->tree_root, 1);
5494 btrfs_commit_transaction(trans, info->tree_root);
5495
5496 while (1) {
5497 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
5498 if (ret < 0)
5499 goto out;
5500 next:
5501 leaf = path->nodes[0];
5502 nritems = btrfs_header_nritems(leaf);
5503 if (path->slots[0] >= nritems) {
5504 ret = btrfs_next_leaf(root, path);
5505 if (ret < 0)
5506 goto out;
5507 if (ret == 1) {
5508 ret = 0;
5509 break;
5510 }
5511 leaf = path->nodes[0];
5512 nritems = btrfs_header_nritems(leaf);
5513 }
5514
5515 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
5516
5517 if (key.objectid >= block_group->key.objectid +
5518 block_group->key.offset)
5519 break;
5520
5521 if (progress && need_resched()) {
5522 btrfs_release_path(root, path);
5523 cond_resched();
5524 progress = 0;
5525 continue;
5526 }
5527 progress = 1;
5528
5529 if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY ||
5530 key.objectid + key.offset <= cur_byte) {
5531 path->slots[0]++;
5532 goto next;
5533 }
5534
5535 total_found++;
5536 cur_byte = key.objectid + key.offset;
5537 btrfs_release_path(root, path);
5538
5539 __alloc_chunk_for_shrink(root, block_group, 0);
5540 ret = relocate_one_extent(root, path, &key, block_group,
5541 reloc_inode, pass);
5542 BUG_ON(ret < 0);
5543 if (ret > 0)
5544 skipped++;
5545
5546 key.objectid = cur_byte;
5547 key.type = 0;
5548 key.offset = 0;
5549 }
5550
5551 btrfs_release_path(root, path);
5552
5553 if (pass == 0) {
5554 btrfs_wait_ordered_range(reloc_inode, 0, (u64)-1);
5555 invalidate_mapping_pages(reloc_inode->i_mapping, 0, -1);
5556 }
5557
5558 if (total_found > 0) {
5559 printk(KERN_INFO "btrfs found %llu extents in pass %d\n",
5560 (unsigned long long)total_found, pass);
5561 pass++;
5562 if (total_found == skipped && pass > 2) {
5563 iput(reloc_inode);
5564 reloc_inode = create_reloc_inode(info, block_group);
5565 pass = 0;
5566 }
5567 goto again;
5568 }
5569
5570 /* delete reloc_inode */
5571 iput(reloc_inode);
5572
5573 /* unpin extents in this range */
5574 trans = btrfs_start_transaction(info->tree_root, 1);
5575 btrfs_commit_transaction(trans, info->tree_root);
5576
5577 spin_lock(&block_group->lock);
5578 WARN_ON(block_group->pinned > 0);
5579 WARN_ON(block_group->reserved > 0);
5580 WARN_ON(btrfs_block_group_used(&block_group->item) > 0);
5581 spin_unlock(&block_group->lock);
5582 put_block_group(block_group);
5583 ret = 0;
5584 out:
5585 btrfs_free_path(path);
5586 return ret;
5587 }
5588
5589 static int find_first_block_group(struct btrfs_root *root,
5590 struct btrfs_path *path, struct btrfs_key *key)
5591 {
5592 int ret = 0;
5593 struct btrfs_key found_key;
5594 struct extent_buffer *leaf;
5595 int slot;
5596
5597 ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
5598 if (ret < 0)
5599 goto out;
5600
5601 while (1) {
5602 slot = path->slots[0];
5603 leaf = path->nodes[0];
5604 if (slot >= btrfs_header_nritems(leaf)) {
5605 ret = btrfs_next_leaf(root, path);
5606 if (ret == 0)
5607 continue;
5608 if (ret < 0)
5609 goto out;
5610 break;
5611 }
5612 btrfs_item_key_to_cpu(leaf, &found_key, slot);
5613
5614 if (found_key.objectid >= key->objectid &&
5615 found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) {
5616 ret = 0;
5617 goto out;
5618 }
5619 path->slots[0]++;
5620 }
5621 ret = -ENOENT;
5622 out:
5623 return ret;
5624 }
5625
5626 int btrfs_free_block_groups(struct btrfs_fs_info *info)
5627 {
5628 struct btrfs_block_group_cache *block_group;
5629 struct btrfs_space_info *space_info;
5630 struct rb_node *n;
5631
5632 spin_lock(&info->block_group_cache_lock);
5633 while ((n = rb_last(&info->block_group_cache_tree)) != NULL) {
5634 block_group = rb_entry(n, struct btrfs_block_group_cache,
5635 cache_node);
5636 rb_erase(&block_group->cache_node,
5637 &info->block_group_cache_tree);
5638 spin_unlock(&info->block_group_cache_lock);
5639
5640 btrfs_remove_free_space_cache(block_group);
5641 down_write(&block_group->space_info->groups_sem);
5642 list_del(&block_group->list);
5643 up_write(&block_group->space_info->groups_sem);
5644
5645 WARN_ON(atomic_read(&block_group->count) != 1);
5646 kfree(block_group);
5647
5648 spin_lock(&info->block_group_cache_lock);
5649 }
5650 spin_unlock(&info->block_group_cache_lock);
5651
5652 /* now that all the block groups are freed, go through and
5653 * free all the space_info structs. This is only called during
5654 * the final stages of unmount, and so we know nobody is
5655 * using them. We call synchronize_rcu() once before we start,
5656 * just to be on the safe side.
5657 */
5658 synchronize_rcu();
5659
5660 while(!list_empty(&info->space_info)) {
5661 space_info = list_entry(info->space_info.next,
5662 struct btrfs_space_info,
5663 list);
5664
5665 list_del(&space_info->list);
5666 kfree(space_info);
5667 }
5668 return 0;
5669 }
5670
5671 int btrfs_read_block_groups(struct btrfs_root *root)
5672 {
5673 struct btrfs_path *path;
5674 int ret;
5675 struct btrfs_block_group_cache *cache;
5676 struct btrfs_fs_info *info = root->fs_info;
5677 struct btrfs_space_info *space_info;
5678 struct btrfs_key key;
5679 struct btrfs_key found_key;
5680 struct extent_buffer *leaf;
5681
5682 root = info->extent_root;
5683 key.objectid = 0;
5684 key.offset = 0;
5685 btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
5686 path = btrfs_alloc_path();
5687 if (!path)
5688 return -ENOMEM;
5689
5690 while (1) {
5691 ret = find_first_block_group(root, path, &key);
5692 if (ret > 0) {
5693 ret = 0;
5694 goto error;
5695 }
5696 if (ret != 0)
5697 goto error;
5698
5699 leaf = path->nodes[0];
5700 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
5701 cache = kzalloc(sizeof(*cache), GFP_NOFS);
5702 if (!cache) {
5703 ret = -ENOMEM;
5704 break;
5705 }
5706
5707 atomic_set(&cache->count, 1);
5708 spin_lock_init(&cache->lock);
5709 mutex_init(&cache->alloc_mutex);
5710 mutex_init(&cache->cache_mutex);
5711 INIT_LIST_HEAD(&cache->list);
5712 read_extent_buffer(leaf, &cache->item,
5713 btrfs_item_ptr_offset(leaf, path->slots[0]),
5714 sizeof(cache->item));
5715 memcpy(&cache->key, &found_key, sizeof(found_key));
5716
5717 key.objectid = found_key.objectid + found_key.offset;
5718 btrfs_release_path(root, path);
5719 cache->flags = btrfs_block_group_flags(&cache->item);
5720
5721 ret = update_space_info(info, cache->flags, found_key.offset,
5722 btrfs_block_group_used(&cache->item),
5723 &space_info);
5724 BUG_ON(ret);
5725 cache->space_info = space_info;
5726 down_write(&space_info->groups_sem);
5727 list_add_tail(&cache->list, &space_info->block_groups);
5728 up_write(&space_info->groups_sem);
5729
5730 ret = btrfs_add_block_group_cache(root->fs_info, cache);
5731 BUG_ON(ret);
5732
5733 set_avail_alloc_bits(root->fs_info, cache->flags);
5734 if (btrfs_chunk_readonly(root, cache->key.objectid))
5735 set_block_group_readonly(cache);
5736 }
5737 ret = 0;
5738 error:
5739 btrfs_free_path(path);
5740 return ret;
5741 }
5742
5743 int btrfs_make_block_group(struct btrfs_trans_handle *trans,
5744 struct btrfs_root *root, u64 bytes_used,
5745 u64 type, u64 chunk_objectid, u64 chunk_offset,
5746 u64 size)
5747 {
5748 int ret;
5749 struct btrfs_root *extent_root;
5750 struct btrfs_block_group_cache *cache;
5751
5752 extent_root = root->fs_info->extent_root;
5753
5754 root->fs_info->last_trans_new_blockgroup = trans->transid;
5755
5756 cache = kzalloc(sizeof(*cache), GFP_NOFS);
5757 if (!cache)
5758 return -ENOMEM;
5759
5760 cache->key.objectid = chunk_offset;
5761 cache->key.offset = size;
5762 cache->key.type = BTRFS_BLOCK_GROUP_ITEM_KEY;
5763 atomic_set(&cache->count, 1);
5764 spin_lock_init(&cache->lock);
5765 mutex_init(&cache->alloc_mutex);
5766 mutex_init(&cache->cache_mutex);
5767 INIT_LIST_HEAD(&cache->list);
5768
5769 btrfs_set_block_group_used(&cache->item, bytes_used);
5770 btrfs_set_block_group_chunk_objectid(&cache->item, chunk_objectid);
5771 cache->flags = type;
5772 btrfs_set_block_group_flags(&cache->item, type);
5773
5774 ret = update_space_info(root->fs_info, cache->flags, size, bytes_used,
5775 &cache->space_info);
5776 BUG_ON(ret);
5777 down_write(&cache->space_info->groups_sem);
5778 list_add_tail(&cache->list, &cache->space_info->block_groups);
5779 up_write(&cache->space_info->groups_sem);
5780
5781 ret = btrfs_add_block_group_cache(root->fs_info, cache);
5782 BUG_ON(ret);
5783
5784 ret = btrfs_insert_item(trans, extent_root, &cache->key, &cache->item,
5785 sizeof(cache->item));
5786 BUG_ON(ret);
5787
5788 set_avail_alloc_bits(extent_root->fs_info, type);
5789
5790 return 0;
5791 }
5792
5793 int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
5794 struct btrfs_root *root, u64 group_start)
5795 {
5796 struct btrfs_path *path;
5797 struct btrfs_block_group_cache *block_group;
5798 struct btrfs_key key;
5799 int ret;
5800
5801 root = root->fs_info->extent_root;
5802
5803 block_group = btrfs_lookup_block_group(root->fs_info, group_start);
5804 BUG_ON(!block_group);
5805 BUG_ON(!block_group->ro);
5806
5807 memcpy(&key, &block_group->key, sizeof(key));
5808
5809 path = btrfs_alloc_path();
5810 BUG_ON(!path);
5811
5812 spin_lock(&root->fs_info->block_group_cache_lock);
5813 rb_erase(&block_group->cache_node,
5814 &root->fs_info->block_group_cache_tree);
5815 spin_unlock(&root->fs_info->block_group_cache_lock);
5816 btrfs_remove_free_space_cache(block_group);
5817 down_write(&block_group->space_info->groups_sem);
5818 list_del(&block_group->list);
5819 up_write(&block_group->space_info->groups_sem);
5820
5821 spin_lock(&block_group->space_info->lock);
5822 block_group->space_info->total_bytes -= block_group->key.offset;
5823 block_group->space_info->bytes_readonly -= block_group->key.offset;
5824 spin_unlock(&block_group->space_info->lock);
5825 block_group->space_info->full = 0;
5826
5827 put_block_group(block_group);
5828 put_block_group(block_group);
5829
5830 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
5831 if (ret > 0)
5832 ret = -EIO;
5833 if (ret < 0)
5834 goto out;
5835
5836 ret = btrfs_del_item(trans, root, path);
5837 out:
5838 btrfs_free_path(path);
5839 return ret;
5840 }
Linux kernel - Deliverables
This page took 0.160698 seconds and 6 git commands to generate.