?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
sata_mv: workaround errata SATA#13
[linux-2.6/kvm.git] / fs / btrfs / extent-tree.c
blob178df4c67de447e87514e90284a7ccc294dc8df1
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 #include "free-space-cache.h"
35
36 #define PENDING_EXTENT_INSERT 0
37 #define PENDING_EXTENT_DELETE 1
38 #define PENDING_BACKREF_UPDATE 2
39
40 struct pending_extent_op {
41 int type;
42 u64 bytenr;
43 u64 num_bytes;
44 u64 parent;
45 u64 orig_parent;
46 u64 generation;
47 u64 orig_generation;
48 int level;
49 struct list_head list;
50 int del;
51 };
52
53 static int __btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans,
54 struct btrfs_root *root, u64 parent,
55 u64 root_objectid, u64 ref_generation,
56 u64 owner, struct btrfs_key *ins,
57 int ref_mod);
58 static int update_reserved_extents(struct btrfs_root *root,
59 u64 bytenr, u64 num, int reserve);
60 static int update_block_group(struct btrfs_trans_handle *trans,
61 struct btrfs_root *root,
62 u64 bytenr, u64 num_bytes, int alloc,
63 int mark_free);
64 static noinline int __btrfs_free_extent(struct btrfs_trans_handle *trans,
65 struct btrfs_root *root,
66 u64 bytenr, u64 num_bytes, u64 parent,
67 u64 root_objectid, u64 ref_generation,
68 u64 owner_objectid, int pin,
69 int ref_to_drop);
70
71 static int do_chunk_alloc(struct btrfs_trans_handle *trans,
72 struct btrfs_root *extent_root, u64 alloc_bytes,
73 u64 flags, int force);
74
75 static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits)
76 {
77 return (cache->flags & bits) == bits;
78 }
79
80 /*
81 * this adds the block group to the fs_info rb tree for the block group
82 * cache
83 */
84 static int btrfs_add_block_group_cache(struct btrfs_fs_info *info,
85 struct btrfs_block_group_cache *block_group)
86 {
87 struct rb_node **p;
88 struct rb_node *parent = NULL;
89 struct btrfs_block_group_cache *cache;
90
91 spin_lock(&info->block_group_cache_lock);
92 p = &info->block_group_cache_tree.rb_node;
93
94 while (*p) {
95 parent = *p;
96 cache = rb_entry(parent, struct btrfs_block_group_cache,
97 cache_node);
98 if (block_group->key.objectid < cache->key.objectid) {
99 p = &(*p)->rb_left;
100 } else if (block_group->key.objectid > cache->key.objectid) {
101 p = &(*p)->rb_right;
102 } else {
103 spin_unlock(&info->block_group_cache_lock);
104 return -EEXIST;
105 }
106 }
107
108 rb_link_node(&block_group->cache_node, parent, p);
109 rb_insert_color(&block_group->cache_node,
110 &info->block_group_cache_tree);
111 spin_unlock(&info->block_group_cache_lock);
112
113 return 0;
114 }
115
116 /*
117 * This will return the block group at or after bytenr if contains is 0, else
118 * it will return the block group that contains the bytenr
119 */
120 static struct btrfs_block_group_cache *
121 block_group_cache_tree_search(struct btrfs_fs_info *info, u64 bytenr,
122 int contains)
123 {
124 struct btrfs_block_group_cache *cache, *ret = NULL;
125 struct rb_node *n;
126 u64 end, start;
127
128 spin_lock(&info->block_group_cache_lock);
129 n = info->block_group_cache_tree.rb_node;
130
131 while (n) {
132 cache = rb_entry(n, struct btrfs_block_group_cache,
133 cache_node);
134 end = cache->key.objectid + cache->key.offset - 1;
135 start = cache->key.objectid;
136
137 if (bytenr < start) {
138 if (!contains && (!ret || start < ret->key.objectid))
139 ret = cache;
140 n = n->rb_left;
141 } else if (bytenr > start) {
142 if (contains && bytenr <= end) {
143 ret = cache;
144 break;
145 }
146 n = n->rb_right;
147 } else {
148 ret = cache;
149 break;
150 }
151 }
152 if (ret)
153 atomic_inc(&ret->count);
154 spin_unlock(&info->block_group_cache_lock);
155
156 return ret;
157 }
158
159 /*
160 * this is only called by cache_block_group, since we could have freed extents
161 * we need to check the pinned_extents for any extents that can't be used yet
162 * since their free space will be released as soon as the transaction commits.
163 */
164 static int add_new_free_space(struct btrfs_block_group_cache *block_group,
165 struct btrfs_fs_info *info, u64 start, u64 end)
166 {
167 u64 extent_start, extent_end, size;
168 int ret;
169
170 while (start < end) {
171 ret = find_first_extent_bit(&info->pinned_extents, start,
172 &extent_start, &extent_end,
173 EXTENT_DIRTY);
174 if (ret)
175 break;
176
177 if (extent_start == start) {
178 start = extent_end + 1;
179 } else if (extent_start > start && extent_start < end) {
180 size = extent_start - start;
181 ret = btrfs_add_free_space(block_group, start,
182 size);
183 BUG_ON(ret);
184 start = extent_end + 1;
185 } else {
186 break;
187 }
188 }
189
190 if (start < end) {
191 size = end - start;
192 ret = btrfs_add_free_space(block_group, start, size);
193 BUG_ON(ret);
194 }
195
196 return 0;
197 }
198
199 static int remove_sb_from_cache(struct btrfs_root *root,
200 struct btrfs_block_group_cache *cache)
201 {
202 u64 bytenr;
203 u64 *logical;
204 int stripe_len;
205 int i, nr, ret;
206
207 for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
208 bytenr = btrfs_sb_offset(i);
209 ret = btrfs_rmap_block(&root->fs_info->mapping_tree,
210 cache->key.objectid, bytenr, 0,
211 &logical, &nr, &stripe_len);
212 BUG_ON(ret);
213 while (nr--) {
214 btrfs_remove_free_space(cache, logical[nr],
215 stripe_len);
216 }
217 kfree(logical);
218 }
219 return 0;
220 }
221
222 static int cache_block_group(struct btrfs_root *root,
223 struct btrfs_block_group_cache *block_group)
224 {
225 struct btrfs_path *path;
226 int ret = 0;
227 struct btrfs_key key;
228 struct extent_buffer *leaf;
229 int slot;
230 u64 last;
231
232 if (!block_group)
233 return 0;
234
235 root = root->fs_info->extent_root;
236
237 if (block_group->cached)
238 return 0;
239
240 path = btrfs_alloc_path();
241 if (!path)
242 return -ENOMEM;
243
244 path->reada = 2;
245 /*
246 * we get into deadlocks with paths held by callers of this function.
247 * since the alloc_mutex is protecting things right now, just
248 * skip the locking here
249 */
250 path->skip_locking = 1;
251 last = max_t(u64, block_group->key.objectid, BTRFS_SUPER_INFO_OFFSET);
252 key.objectid = last;
253 key.offset = 0;
254 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
255 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
256 if (ret < 0)
257 goto err;
258
259 while (1) {
260 leaf = path->nodes[0];
261 slot = path->slots[0];
262 if (slot >= btrfs_header_nritems(leaf)) {
263 ret = btrfs_next_leaf(root, path);
264 if (ret < 0)
265 goto err;
266 if (ret == 0)
267 continue;
268 else
269 break;
270 }
271 btrfs_item_key_to_cpu(leaf, &key, slot);
272 if (key.objectid < block_group->key.objectid)
273 goto next;
274
275 if (key.objectid >= block_group->key.objectid +
276 block_group->key.offset)
277 break;
278
279 if (btrfs_key_type(&key) == BTRFS_EXTENT_ITEM_KEY) {
280 add_new_free_space(block_group, root->fs_info, last,
281 key.objectid);
282
283 last = key.objectid + key.offset;
284 }
285 next:
286 path->slots[0]++;
287 }
288
289 add_new_free_space(block_group, root->fs_info, last,
290 block_group->key.objectid +
291 block_group->key.offset);
292
293 block_group->cached = 1;
294 remove_sb_from_cache(root, block_group);
295 ret = 0;
296 err:
297 btrfs_free_path(path);
298 return ret;
299 }
300
301 /*
302 * return the block group that starts at or after bytenr
303 */
304 static struct btrfs_block_group_cache *
305 btrfs_lookup_first_block_group(struct btrfs_fs_info *info, u64 bytenr)
306 {
307 struct btrfs_block_group_cache *cache;
308
309 cache = block_group_cache_tree_search(info, bytenr, 0);
310
311 return cache;
312 }
313
314 /*
315 * return the block group that contains teh given bytenr
316 */
317 struct btrfs_block_group_cache *btrfs_lookup_block_group(
318 struct btrfs_fs_info *info,
319 u64 bytenr)
320 {
321 struct btrfs_block_group_cache *cache;
322
323 cache = block_group_cache_tree_search(info, bytenr, 1);
324
325 return cache;
326 }
327
328 void btrfs_put_block_group(struct btrfs_block_group_cache *cache)
329 {
330 if (atomic_dec_and_test(&cache->count))
331 kfree(cache);
332 }
333
334 static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info,
335 u64 flags)
336 {
337 struct list_head *head = &info->space_info;
338 struct btrfs_space_info *found;
339
340 rcu_read_lock();
341 list_for_each_entry_rcu(found, head, list) {
342 if (found->flags == flags) {
343 rcu_read_unlock();
344 return found;
345 }
346 }
347 rcu_read_unlock();
348 return NULL;
349 }
350
351 /*
352 * after adding space to the filesystem, we need to clear the full flags
353 * on all the space infos.
354 */
355 void btrfs_clear_space_info_full(struct btrfs_fs_info *info)
356 {
357 struct list_head *head = &info->space_info;
358 struct btrfs_space_info *found;
359
360 rcu_read_lock();
361 list_for_each_entry_rcu(found, head, list)
362 found->full = 0;
363 rcu_read_unlock();
364 }
365
366 static u64 div_factor(u64 num, int factor)
367 {
368 if (factor == 10)
369 return num;
370 num *= factor;
371 do_div(num, 10);
372 return num;
373 }
374
375 u64 btrfs_find_block_group(struct btrfs_root *root,
376 u64 search_start, u64 search_hint, int owner)
377 {
378 struct btrfs_block_group_cache *cache;
379 u64 used;
380 u64 last = max(search_hint, search_start);
381 u64 group_start = 0;
382 int full_search = 0;
383 int factor = 9;
384 int wrapped = 0;
385 again:
386 while (1) {
387 cache = btrfs_lookup_first_block_group(root->fs_info, last);
388 if (!cache)
389 break;
390
391 spin_lock(&cache->lock);
392 last = cache->key.objectid + cache->key.offset;
393 used = btrfs_block_group_used(&cache->item);
394
395 if ((full_search || !cache->ro) &&
396 block_group_bits(cache, BTRFS_BLOCK_GROUP_METADATA)) {
397 if (used + cache->pinned + cache->reserved <
398 div_factor(cache->key.offset, factor)) {
399 group_start = cache->key.objectid;
400 spin_unlock(&cache->lock);
401 btrfs_put_block_group(cache);
402 goto found;
403 }
404 }
405 spin_unlock(&cache->lock);
406 btrfs_put_block_group(cache);
407 cond_resched();
408 }
409 if (!wrapped) {
410 last = search_start;
411 wrapped = 1;
412 goto again;
413 }
414 if (!full_search && factor < 10) {
415 last = search_start;
416 full_search = 1;
417 factor = 10;
418 goto again;
419 }
420 found:
421 return group_start;
422 }
423
424 /* simple helper to search for an existing extent at a given offset */
425 int btrfs_lookup_extent(struct btrfs_root *root, u64 start, u64 len)
426 {
427 int ret;
428 struct btrfs_key key;
429 struct btrfs_path *path;
430
431 path = btrfs_alloc_path();
432 BUG_ON(!path);
433 key.objectid = start;
434 key.offset = len;
435 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
436 ret = btrfs_search_slot(NULL, root->fs_info->extent_root, &key, path,
437 0, 0);
438 btrfs_free_path(path);
439 return ret;
440 }
441
442 /*
443 * Back reference rules. Back refs have three main goals:
444 *
445 * 1) differentiate between all holders of references to an extent so that
446 * when a reference is dropped we can make sure it was a valid reference
447 * before freeing the extent.
448 *
449 * 2) Provide enough information to quickly find the holders of an extent
450 * if we notice a given block is corrupted or bad.
451 *
452 * 3) Make it easy to migrate blocks for FS shrinking or storage pool
453 * maintenance. This is actually the same as #2, but with a slightly
454 * different use case.
455 *
456 * File extents can be referenced by:
457 *
458 * - multiple snapshots, subvolumes, or different generations in one subvol
459 * - different files inside a single subvolume
460 * - different offsets inside a file (bookend extents in file.c)
461 *
462 * The extent ref structure has fields for:
463 *
464 * - Objectid of the subvolume root
465 * - Generation number of the tree holding the reference
466 * - objectid of the file holding the reference
467 * - number of references holding by parent node (alway 1 for tree blocks)
468 *
469 * Btree leaf may hold multiple references to a file extent. In most cases,
470 * these references are from same file and the corresponding offsets inside
471 * the file are close together.
472 *
473 * When a file extent is allocated the fields are filled in:
474 * (root_key.objectid, trans->transid, inode objectid, 1)
475 *
476 * When a leaf is cow'd new references are added for every file extent found
477 * in the leaf. It looks similar to the create case, but trans->transid will
478 * be different when the block is cow'd.
479 *
480 * (root_key.objectid, trans->transid, inode objectid,
481 * number of references in the leaf)
482 *
483 * When a file extent is removed either during snapshot deletion or
484 * file truncation, we find the corresponding back reference and check
485 * the following fields:
486 *
487 * (btrfs_header_owner(leaf), btrfs_header_generation(leaf),
488 * inode objectid)
489 *
490 * Btree extents can be referenced by:
491 *
492 * - Different subvolumes
493 * - Different generations of the same subvolume
494 *
495 * When a tree block is created, back references are inserted:
496 *
497 * (root->root_key.objectid, trans->transid, level, 1)
498 *
499 * When a tree block is cow'd, new back references are added for all the
500 * blocks it points to. If the tree block isn't in reference counted root,
501 * the old back references are removed. These new back references are of
502 * the form (trans->transid will have increased since creation):
503 *
504 * (root->root_key.objectid, trans->transid, level, 1)
505 *
506 * When a backref is in deleting, the following fields are checked:
507 *
508 * if backref was for a tree root:
509 * (btrfs_header_owner(itself), btrfs_header_generation(itself), level)
510 * else
511 * (btrfs_header_owner(parent), btrfs_header_generation(parent), level)
512 *
513 * Back Reference Key composing:
514 *
515 * The key objectid corresponds to the first byte in the extent, the key
516 * type is set to BTRFS_EXTENT_REF_KEY, and the key offset is the first
517 * byte of parent extent. If a extent is tree root, the key offset is set
518 * to the key objectid.
519 */
520
521 static noinline int lookup_extent_backref(struct btrfs_trans_handle *trans,
522 struct btrfs_root *root,
523 struct btrfs_path *path,
524 u64 bytenr, u64 parent,
525 u64 ref_root, u64 ref_generation,
526 u64 owner_objectid, int del)
527 {
528 struct btrfs_key key;
529 struct btrfs_extent_ref *ref;
530 struct extent_buffer *leaf;
531 u64 ref_objectid;
532 int ret;
533
534 key.objectid = bytenr;
535 key.type = BTRFS_EXTENT_REF_KEY;
536 key.offset = parent;
537
538 ret = btrfs_search_slot(trans, root, &key, path, del ? -1 : 0, 1);
539 if (ret < 0)
540 goto out;
541 if (ret > 0) {
542 ret = -ENOENT;
543 goto out;
544 }
545
546 leaf = path->nodes[0];
547 ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_ref);
548 ref_objectid = btrfs_ref_objectid(leaf, ref);
549 if (btrfs_ref_root(leaf, ref) != ref_root ||
550 btrfs_ref_generation(leaf, ref) != ref_generation ||
551 (ref_objectid != owner_objectid &&
552 ref_objectid != BTRFS_MULTIPLE_OBJECTIDS)) {
553 ret = -EIO;
554 WARN_ON(1);
555 goto out;
556 }
557 ret = 0;
558 out:
559 return ret;
560 }
561
562 static noinline int insert_extent_backref(struct btrfs_trans_handle *trans,
563 struct btrfs_root *root,
564 struct btrfs_path *path,
565 u64 bytenr, u64 parent,
566 u64 ref_root, u64 ref_generation,
567 u64 owner_objectid,
568 int refs_to_add)
569 {
570 struct btrfs_key key;
571 struct extent_buffer *leaf;
572 struct btrfs_extent_ref *ref;
573 u32 num_refs;
574 int ret;
575
576 key.objectid = bytenr;
577 key.type = BTRFS_EXTENT_REF_KEY;
578 key.offset = parent;
579
580 ret = btrfs_insert_empty_item(trans, root, path, &key, sizeof(*ref));
581 if (ret == 0) {
582 leaf = path->nodes[0];
583 ref = btrfs_item_ptr(leaf, path->slots[0],
584 struct btrfs_extent_ref);
585 btrfs_set_ref_root(leaf, ref, ref_root);
586 btrfs_set_ref_generation(leaf, ref, ref_generation);
587 btrfs_set_ref_objectid(leaf, ref, owner_objectid);
588 btrfs_set_ref_num_refs(leaf, ref, refs_to_add);
589 } else if (ret == -EEXIST) {
590 u64 existing_owner;
591
592 BUG_ON(owner_objectid < BTRFS_FIRST_FREE_OBJECTID);
593 leaf = path->nodes[0];
594 ref = btrfs_item_ptr(leaf, path->slots[0],
595 struct btrfs_extent_ref);
596 if (btrfs_ref_root(leaf, ref) != ref_root ||
597 btrfs_ref_generation(leaf, ref) != ref_generation) {
598 ret = -EIO;
599 WARN_ON(1);
600 goto out;
601 }
602
603 num_refs = btrfs_ref_num_refs(leaf, ref);
604 BUG_ON(num_refs == 0);
605 btrfs_set_ref_num_refs(leaf, ref, num_refs + refs_to_add);
606
607 existing_owner = btrfs_ref_objectid(leaf, ref);
608 if (existing_owner != owner_objectid &&
609 existing_owner != BTRFS_MULTIPLE_OBJECTIDS) {
610 btrfs_set_ref_objectid(leaf, ref,
611 BTRFS_MULTIPLE_OBJECTIDS);
612 }
613 ret = 0;
614 } else {
615 goto out;
616 }
617 btrfs_unlock_up_safe(path, 1);
618 btrfs_mark_buffer_dirty(path->nodes[0]);
619 out:
620 btrfs_release_path(root, path);
621 return ret;
622 }
623
624 static noinline int remove_extent_backref(struct btrfs_trans_handle *trans,
625 struct btrfs_root *root,
626 struct btrfs_path *path,
627 int refs_to_drop)
628 {
629 struct extent_buffer *leaf;
630 struct btrfs_extent_ref *ref;
631 u32 num_refs;
632 int ret = 0;
633
634 leaf = path->nodes[0];
635 ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_ref);
636 num_refs = btrfs_ref_num_refs(leaf, ref);
637 BUG_ON(num_refs < refs_to_drop);
638 num_refs -= refs_to_drop;
639 if (num_refs == 0) {
640 ret = btrfs_del_item(trans, root, path);
641 } else {
642 btrfs_set_ref_num_refs(leaf, ref, num_refs);
643 btrfs_mark_buffer_dirty(leaf);
644 }
645 btrfs_release_path(root, path);
646 return ret;
647 }
648
649 #ifdef BIO_RW_DISCARD
650 static void btrfs_issue_discard(struct block_device *bdev,
651 u64 start, u64 len)
652 {
653 blkdev_issue_discard(bdev, start >> 9, len >> 9, GFP_KERNEL);
654 }
655 #endif
656
657 static int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr,
658 u64 num_bytes)
659 {
660 #ifdef BIO_RW_DISCARD
661 int ret;
662 u64 map_length = num_bytes;
663 struct btrfs_multi_bio *multi = NULL;
664
665 /* Tell the block device(s) that the sectors can be discarded */
666 ret = btrfs_map_block(&root->fs_info->mapping_tree, READ,
667 bytenr, &map_length, &multi, 0);
668 if (!ret) {
669 struct btrfs_bio_stripe *stripe = multi->stripes;
670 int i;
671
672 if (map_length > num_bytes)
673 map_length = num_bytes;
674
675 for (i = 0; i < multi->num_stripes; i++, stripe++) {
676 btrfs_issue_discard(stripe->dev->bdev,
677 stripe->physical,
678 map_length);
679 }
680 kfree(multi);
681 }
682
683 return ret;
684 #else
685 return 0;
686 #endif
687 }
688
689 static int __btrfs_update_extent_ref(struct btrfs_trans_handle *trans,
690 struct btrfs_root *root, u64 bytenr,
691 u64 num_bytes,
692 u64 orig_parent, u64 parent,
693 u64 orig_root, u64 ref_root,
694 u64 orig_generation, u64 ref_generation,
695 u64 owner_objectid)
696 {
697 int ret;
698 int pin = owner_objectid < BTRFS_FIRST_FREE_OBJECTID;
699
700 ret = btrfs_update_delayed_ref(trans, bytenr, num_bytes,
701 orig_parent, parent, orig_root,
702 ref_root, orig_generation,
703 ref_generation, owner_objectid, pin);
704 BUG_ON(ret);
705 return ret;
706 }
707
708 int btrfs_update_extent_ref(struct btrfs_trans_handle *trans,
709 struct btrfs_root *root, u64 bytenr,
710 u64 num_bytes, u64 orig_parent, u64 parent,
711 u64 ref_root, u64 ref_generation,
712 u64 owner_objectid)
713 {
714 int ret;
715 if (ref_root == BTRFS_TREE_LOG_OBJECTID &&
716 owner_objectid < BTRFS_FIRST_FREE_OBJECTID)
717 return 0;
718
719 ret = __btrfs_update_extent_ref(trans, root, bytenr, num_bytes,
720 orig_parent, parent, ref_root,
721 ref_root, ref_generation,
722 ref_generation, owner_objectid);
723 return ret;
724 }
725 static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
726 struct btrfs_root *root, u64 bytenr,
727 u64 num_bytes,
728 u64 orig_parent, u64 parent,
729 u64 orig_root, u64 ref_root,
730 u64 orig_generation, u64 ref_generation,
731 u64 owner_objectid)
732 {
733 int ret;
734
735 ret = btrfs_add_delayed_ref(trans, bytenr, num_bytes, parent, ref_root,
736 ref_generation, owner_objectid,
737 BTRFS_ADD_DELAYED_REF, 0);
738 BUG_ON(ret);
739 return ret;
740 }
741
742 static noinline_for_stack int add_extent_ref(struct btrfs_trans_handle *trans,
743 struct btrfs_root *root, u64 bytenr,
744 u64 num_bytes, u64 parent, u64 ref_root,
745 u64 ref_generation, u64 owner_objectid,
746 int refs_to_add)
747 {
748 struct btrfs_path *path;
749 int ret;
750 struct btrfs_key key;
751 struct extent_buffer *l;
752 struct btrfs_extent_item *item;
753 u32 refs;
754
755 path = btrfs_alloc_path();
756 if (!path)
757 return -ENOMEM;
758
759 path->reada = 1;
760 path->leave_spinning = 1;
761 key.objectid = bytenr;
762 key.type = BTRFS_EXTENT_ITEM_KEY;
763 key.offset = num_bytes;
764
765 /* first find the extent item and update its reference count */
766 ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key,
767 path, 0, 1);
768 if (ret < 0) {
769 btrfs_set_path_blocking(path);
770 return ret;
771 }
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_unlock_up_safe(path, 1);
795
796 btrfs_mark_buffer_dirty(path->nodes[0]);
797
798 btrfs_release_path(root->fs_info->extent_root, path);
799
800 path->reada = 1;
801 path->leave_spinning = 1;
802
803 /* now insert the actual backref */
804 ret = insert_extent_backref(trans, root->fs_info->extent_root,
805 path, bytenr, parent,
806 ref_root, ref_generation,
807 owner_objectid, refs_to_add);
808 BUG_ON(ret);
809 btrfs_free_path(path);
810 return 0;
811 }
812
813 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
814 struct btrfs_root *root,
815 u64 bytenr, u64 num_bytes, u64 parent,
816 u64 ref_root, u64 ref_generation,
817 u64 owner_objectid)
818 {
819 int ret;
820 if (ref_root == BTRFS_TREE_LOG_OBJECTID &&
821 owner_objectid < BTRFS_FIRST_FREE_OBJECTID)
822 return 0;
823
824 ret = __btrfs_inc_extent_ref(trans, root, bytenr, num_bytes, 0, parent,
825 0, ref_root, 0, ref_generation,
826 owner_objectid);
827 return ret;
828 }
829
830 static int drop_delayed_ref(struct btrfs_trans_handle *trans,
831 struct btrfs_root *root,
832 struct btrfs_delayed_ref_node *node)
833 {
834 int ret = 0;
835 struct btrfs_delayed_ref *ref = btrfs_delayed_node_to_ref(node);
836
837 BUG_ON(node->ref_mod == 0);
838 ret = __btrfs_free_extent(trans, root, node->bytenr, node->num_bytes,
839 node->parent, ref->root, ref->generation,
840 ref->owner_objectid, ref->pin, node->ref_mod);
841
842 return ret;
843 }
844
845 /* helper function to actually process a single delayed ref entry */
846 static noinline int run_one_delayed_ref(struct btrfs_trans_handle *trans,
847 struct btrfs_root *root,
848 struct btrfs_delayed_ref_node *node,
849 int insert_reserved)
850 {
851 int ret;
852 struct btrfs_delayed_ref *ref;
853
854 if (node->parent == (u64)-1) {
855 struct btrfs_delayed_ref_head *head;
856 /*
857 * we've hit the end of the chain and we were supposed
858 * to insert this extent into the tree. But, it got
859 * deleted before we ever needed to insert it, so all
860 * we have to do is clean up the accounting
861 */
862 if (insert_reserved) {
863 update_reserved_extents(root, node->bytenr,
864 node->num_bytes, 0);
865 }
866 head = btrfs_delayed_node_to_head(node);
867 mutex_unlock(&head->mutex);
868 return 0;
869 }
870
871 ref = btrfs_delayed_node_to_ref(node);
872 if (ref->action == BTRFS_ADD_DELAYED_REF) {
873 if (insert_reserved) {
874 struct btrfs_key ins;
875
876 ins.objectid = node->bytenr;
877 ins.offset = node->num_bytes;
878 ins.type = BTRFS_EXTENT_ITEM_KEY;
879
880 /* record the full extent allocation */
881 ret = __btrfs_alloc_reserved_extent(trans, root,
882 node->parent, ref->root,
883 ref->generation, ref->owner_objectid,
884 &ins, node->ref_mod);
885 update_reserved_extents(root, node->bytenr,
886 node->num_bytes, 0);
887 } else {
888 /* just add one backref */
889 ret = add_extent_ref(trans, root, node->bytenr,
890 node->num_bytes,
891 node->parent, ref->root, ref->generation,
892 ref->owner_objectid, node->ref_mod);
893 }
894 BUG_ON(ret);
895 } else if (ref->action == BTRFS_DROP_DELAYED_REF) {
896 WARN_ON(insert_reserved);
897 ret = drop_delayed_ref(trans, root, node);
898 }
899 return 0;
900 }
901
902 static noinline struct btrfs_delayed_ref_node *
903 select_delayed_ref(struct btrfs_delayed_ref_head *head)
904 {
905 struct rb_node *node;
906 struct btrfs_delayed_ref_node *ref;
907 int action = BTRFS_ADD_DELAYED_REF;
908 again:
909 /*
910 * select delayed ref of type BTRFS_ADD_DELAYED_REF first.
911 * this prevents ref count from going down to zero when
912 * there still are pending delayed ref.
913 */
914 node = rb_prev(&head->node.rb_node);
915 while (1) {
916 if (!node)
917 break;
918 ref = rb_entry(node, struct btrfs_delayed_ref_node,
919 rb_node);
920 if (ref->bytenr != head->node.bytenr)
921 break;
922 if (btrfs_delayed_node_to_ref(ref)->action == action)
923 return ref;
924 node = rb_prev(node);
925 }
926 if (action == BTRFS_ADD_DELAYED_REF) {
927 action = BTRFS_DROP_DELAYED_REF;
928 goto again;
929 }
930 return NULL;
931 }
932
933 static noinline int run_clustered_refs(struct btrfs_trans_handle *trans,
934 struct btrfs_root *root,
935 struct list_head *cluster)
936 {
937 struct btrfs_delayed_ref_root *delayed_refs;
938 struct btrfs_delayed_ref_node *ref;
939 struct btrfs_delayed_ref_head *locked_ref = NULL;
940 int ret;
941 int count = 0;
942 int must_insert_reserved = 0;
943
944 delayed_refs = &trans->transaction->delayed_refs;
945 while (1) {
946 if (!locked_ref) {
947 /* pick a new head ref from the cluster list */
948 if (list_empty(cluster))
949 break;
950
951 locked_ref = list_entry(cluster->next,
952 struct btrfs_delayed_ref_head, cluster);
953
954 /* grab the lock that says we are going to process
955 * all the refs for this head */
956 ret = btrfs_delayed_ref_lock(trans, locked_ref);
957
958 /*
959 * we may have dropped the spin lock to get the head
960 * mutex lock, and that might have given someone else
961 * time to free the head. If that's true, it has been
962 * removed from our list and we can move on.
963 */
964 if (ret == -EAGAIN) {
965 locked_ref = NULL;
966 count++;
967 continue;
968 }
969 }
970
971 /*
972 * record the must insert reserved flag before we
973 * drop the spin lock.
974 */
975 must_insert_reserved = locked_ref->must_insert_reserved;
976 locked_ref->must_insert_reserved = 0;
977
978 /*
979 * locked_ref is the head node, so we have to go one
980 * node back for any delayed ref updates
981 */
982 ref = select_delayed_ref(locked_ref);
983 if (!ref) {
984 /* All delayed refs have been processed, Go ahead
985 * and send the head node to run_one_delayed_ref,
986 * so that any accounting fixes can happen
987 */
988 ref = &locked_ref->node;
989 list_del_init(&locked_ref->cluster);
990 locked_ref = NULL;
991 }
992
993 ref->in_tree = 0;
994 rb_erase(&ref->rb_node, &delayed_refs->root);
995 delayed_refs->num_entries--;
996 spin_unlock(&delayed_refs->lock);
997
998 ret = run_one_delayed_ref(trans, root, ref,
999 must_insert_reserved);
1000 BUG_ON(ret);
1001 btrfs_put_delayed_ref(ref);
1002
1003 count++;
1004 cond_resched();
1005 spin_lock(&delayed_refs->lock);
1006 }
1007 return count;
1008 }
1009
1010 /*
1011 * this starts processing the delayed reference count updates and
1012 * extent insertions we have queued up so far. count can be
1013 * 0, which means to process everything in the tree at the start
1014 * of the run (but not newly added entries), or it can be some target
1015 * number you'd like to process.
1016 */
1017 int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
1018 struct btrfs_root *root, unsigned long count)
1019 {
1020 struct rb_node *node;
1021 struct btrfs_delayed_ref_root *delayed_refs;
1022 struct btrfs_delayed_ref_node *ref;
1023 struct list_head cluster;
1024 int ret;
1025 int run_all = count == (unsigned long)-1;
1026 int run_most = 0;
1027
1028 if (root == root->fs_info->extent_root)
1029 root = root->fs_info->tree_root;
1030
1031 delayed_refs = &trans->transaction->delayed_refs;
1032 INIT_LIST_HEAD(&cluster);
1033 again:
1034 spin_lock(&delayed_refs->lock);
1035 if (count == 0) {
1036 count = delayed_refs->num_entries * 2;
1037 run_most = 1;
1038 }
1039 while (1) {
1040 if (!(run_all || run_most) &&
1041 delayed_refs->num_heads_ready < 64)
1042 break;
1043
1044 /*
1045 * go find something we can process in the rbtree. We start at
1046 * the beginning of the tree, and then build a cluster
1047 * of refs to process starting at the first one we are able to
1048 * lock
1049 */
1050 ret = btrfs_find_ref_cluster(trans, &cluster,
1051 delayed_refs->run_delayed_start);
1052 if (ret)
1053 break;
1054
1055 ret = run_clustered_refs(trans, root, &cluster);
1056 BUG_ON(ret < 0);
1057
1058 count -= min_t(unsigned long, ret, count);
1059
1060 if (count == 0)
1061 break;
1062 }
1063
1064 if (run_all) {
1065 node = rb_first(&delayed_refs->root);
1066 if (!node)
1067 goto out;
1068 count = (unsigned long)-1;
1069
1070 while (node) {
1071 ref = rb_entry(node, struct btrfs_delayed_ref_node,
1072 rb_node);
1073 if (btrfs_delayed_ref_is_head(ref)) {
1074 struct btrfs_delayed_ref_head *head;
1075
1076 head = btrfs_delayed_node_to_head(ref);
1077 atomic_inc(&ref->refs);
1078
1079 spin_unlock(&delayed_refs->lock);
1080 mutex_lock(&head->mutex);
1081 mutex_unlock(&head->mutex);
1082
1083 btrfs_put_delayed_ref(ref);
1084 cond_resched();
1085 goto again;
1086 }
1087 node = rb_next(node);
1088 }
1089 spin_unlock(&delayed_refs->lock);
1090 schedule_timeout(1);
1091 goto again;
1092 }
1093 out:
1094 spin_unlock(&delayed_refs->lock);
1095 return 0;
1096 }
1097
1098 int btrfs_cross_ref_exist(struct btrfs_trans_handle *trans,
1099 struct btrfs_root *root, u64 objectid, u64 bytenr)
1100 {
1101 struct btrfs_root *extent_root = root->fs_info->extent_root;
1102 struct btrfs_path *path;
1103 struct extent_buffer *leaf;
1104 struct btrfs_extent_ref *ref_item;
1105 struct btrfs_key key;
1106 struct btrfs_key found_key;
1107 u64 ref_root;
1108 u64 last_snapshot;
1109 u32 nritems;
1110 int ret;
1111
1112 key.objectid = bytenr;
1113 key.offset = (u64)-1;
1114 key.type = BTRFS_EXTENT_ITEM_KEY;
1115
1116 path = btrfs_alloc_path();
1117 ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
1118 if (ret < 0)
1119 goto out;
1120 BUG_ON(ret == 0);
1121
1122 ret = -ENOENT;
1123 if (path->slots[0] == 0)
1124 goto out;
1125
1126 path->slots[0]--;
1127 leaf = path->nodes[0];
1128 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1129
1130 if (found_key.objectid != bytenr ||
1131 found_key.type != BTRFS_EXTENT_ITEM_KEY)
1132 goto out;
1133
1134 last_snapshot = btrfs_root_last_snapshot(&root->root_item);
1135 while (1) {
1136 leaf = path->nodes[0];
1137 nritems = btrfs_header_nritems(leaf);
1138 if (path->slots[0] >= nritems) {
1139 ret = btrfs_next_leaf(extent_root, path);
1140 if (ret < 0)
1141 goto out;
1142 if (ret == 0)
1143 continue;
1144 break;
1145 }
1146 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
1147 if (found_key.objectid != bytenr)
1148 break;
1149
1150 if (found_key.type != BTRFS_EXTENT_REF_KEY) {
1151 path->slots[0]++;
1152 continue;
1153 }
1154
1155 ref_item = btrfs_item_ptr(leaf, path->slots[0],
1156 struct btrfs_extent_ref);
1157 ref_root = btrfs_ref_root(leaf, ref_item);
1158 if ((ref_root != root->root_key.objectid &&
1159 ref_root != BTRFS_TREE_LOG_OBJECTID) ||
1160 objectid != btrfs_ref_objectid(leaf, ref_item)) {
1161 ret = 1;
1162 goto out;
1163 }
1164 if (btrfs_ref_generation(leaf, ref_item) <= last_snapshot) {
1165 ret = 1;
1166 goto out;
1167 }
1168
1169 path->slots[0]++;
1170 }
1171 ret = 0;
1172 out:
1173 btrfs_free_path(path);
1174 return ret;
1175 }
1176
1177 int btrfs_cache_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1178 struct extent_buffer *buf, u32 nr_extents)
1179 {
1180 struct btrfs_key key;
1181 struct btrfs_file_extent_item *fi;
1182 u64 root_gen;
1183 u32 nritems;
1184 int i;
1185 int level;
1186 int ret = 0;
1187 int shared = 0;
1188
1189 if (!root->ref_cows)
1190 return 0;
1191
1192 if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
1193 shared = 0;
1194 root_gen = root->root_key.offset;
1195 } else {
1196 shared = 1;
1197 root_gen = trans->transid - 1;
1198 }
1199
1200 level = btrfs_header_level(buf);
1201 nritems = btrfs_header_nritems(buf);
1202
1203 if (level == 0) {
1204 struct btrfs_leaf_ref *ref;
1205 struct btrfs_extent_info *info;
1206
1207 ref = btrfs_alloc_leaf_ref(root, nr_extents);
1208 if (!ref) {
1209 ret = -ENOMEM;
1210 goto out;
1211 }
1212
1213 ref->root_gen = root_gen;
1214 ref->bytenr = buf->start;
1215 ref->owner = btrfs_header_owner(buf);
1216 ref->generation = btrfs_header_generation(buf);
1217 ref->nritems = nr_extents;
1218 info = ref->extents;
1219
1220 for (i = 0; nr_extents > 0 && i < nritems; i++) {
1221 u64 disk_bytenr;
1222 btrfs_item_key_to_cpu(buf, &key, i);
1223 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
1224 continue;
1225 fi = btrfs_item_ptr(buf, i,
1226 struct btrfs_file_extent_item);
1227 if (btrfs_file_extent_type(buf, fi) ==
1228 BTRFS_FILE_EXTENT_INLINE)
1229 continue;
1230 disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
1231 if (disk_bytenr == 0)
1232 continue;
1233
1234 info->bytenr = disk_bytenr;
1235 info->num_bytes =
1236 btrfs_file_extent_disk_num_bytes(buf, fi);
1237 info->objectid = key.objectid;
1238 info->offset = key.offset;
1239 info++;
1240 }
1241
1242 ret = btrfs_add_leaf_ref(root, ref, shared);
1243 if (ret == -EEXIST && shared) {
1244 struct btrfs_leaf_ref *old;
1245 old = btrfs_lookup_leaf_ref(root, ref->bytenr);
1246 BUG_ON(!old);
1247 btrfs_remove_leaf_ref(root, old);
1248 btrfs_free_leaf_ref(root, old);
1249 ret = btrfs_add_leaf_ref(root, ref, shared);
1250 }
1251 WARN_ON(ret);
1252 btrfs_free_leaf_ref(root, ref);
1253 }
1254 out:
1255 return ret;
1256 }
1257
1258 /* when a block goes through cow, we update the reference counts of
1259 * everything that block points to. The internal pointers of the block
1260 * can be in just about any order, and it is likely to have clusters of
1261 * things that are close together and clusters of things that are not.
1262 *
1263 * To help reduce the seeks that come with updating all of these reference
1264 * counts, sort them by byte number before actual updates are done.
1265 *
1266 * struct refsort is used to match byte number to slot in the btree block.
1267 * we sort based on the byte number and then use the slot to actually
1268 * find the item.
1269 *
1270 * struct refsort is smaller than strcut btrfs_item and smaller than
1271 * struct btrfs_key_ptr. Since we're currently limited to the page size
1272 * for a btree block, there's no way for a kmalloc of refsorts for a
1273 * single node to be bigger than a page.
1274 */
1275 struct refsort {
1276 u64 bytenr;
1277 u32 slot;
1278 };
1279
1280 /*
1281 * for passing into sort()
1282 */
1283 static int refsort_cmp(const void *a_void, const void *b_void)
1284 {
1285 const struct refsort *a = a_void;
1286 const struct refsort *b = b_void;
1287
1288 if (a->bytenr < b->bytenr)
1289 return -1;
1290 if (a->bytenr > b->bytenr)
1291 return 1;
1292 return 0;
1293 }
1294
1295
1296 noinline int btrfs_inc_ref(struct btrfs_trans_handle *trans,
1297 struct btrfs_root *root,
1298 struct extent_buffer *orig_buf,
1299 struct extent_buffer *buf, u32 *nr_extents)
1300 {
1301 u64 bytenr;
1302 u64 ref_root;
1303 u64 orig_root;
1304 u64 ref_generation;
1305 u64 orig_generation;
1306 struct refsort *sorted;
1307 u32 nritems;
1308 u32 nr_file_extents = 0;
1309 struct btrfs_key key;
1310 struct btrfs_file_extent_item *fi;
1311 int i;
1312 int level;
1313 int ret = 0;
1314 int faili = 0;
1315 int refi = 0;
1316 int slot;
1317 int (*process_func)(struct btrfs_trans_handle *, struct btrfs_root *,
1318 u64, u64, u64, u64, u64, u64, u64, u64, u64);
1319
1320 ref_root = btrfs_header_owner(buf);
1321 ref_generation = btrfs_header_generation(buf);
1322 orig_root = btrfs_header_owner(orig_buf);
1323 orig_generation = btrfs_header_generation(orig_buf);
1324
1325 nritems = btrfs_header_nritems(buf);
1326 level = btrfs_header_level(buf);
1327
1328 sorted = kmalloc(sizeof(struct refsort) * nritems, GFP_NOFS);
1329 BUG_ON(!sorted);
1330
1331 if (root->ref_cows) {
1332 process_func = __btrfs_inc_extent_ref;
1333 } else {
1334 if (level == 0 &&
1335 root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID)
1336 goto out;
1337 if (level != 0 &&
1338 root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID)
1339 goto out;
1340 process_func = __btrfs_update_extent_ref;
1341 }
1342
1343 /*
1344 * we make two passes through the items. In the first pass we
1345 * only record the byte number and slot. Then we sort based on
1346 * byte number and do the actual work based on the sorted results
1347 */
1348 for (i = 0; i < nritems; i++) {
1349 cond_resched();
1350 if (level == 0) {
1351 btrfs_item_key_to_cpu(buf, &key, i);
1352 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
1353 continue;
1354 fi = btrfs_item_ptr(buf, i,
1355 struct btrfs_file_extent_item);
1356 if (btrfs_file_extent_type(buf, fi) ==
1357 BTRFS_FILE_EXTENT_INLINE)
1358 continue;
1359 bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
1360 if (bytenr == 0)
1361 continue;
1362
1363 nr_file_extents++;
1364 sorted[refi].bytenr = bytenr;
1365 sorted[refi].slot = i;
1366 refi++;
1367 } else {
1368 bytenr = btrfs_node_blockptr(buf, i);
1369 sorted[refi].bytenr = bytenr;
1370 sorted[refi].slot = i;
1371 refi++;
1372 }
1373 }
1374 /*
1375 * if refi == 0, we didn't actually put anything into the sorted
1376 * array and we're done
1377 */
1378 if (refi == 0)
1379 goto out;
1380
1381 sort(sorted, refi, sizeof(struct refsort), refsort_cmp, NULL);
1382
1383 for (i = 0; i < refi; i++) {
1384 cond_resched();
1385 slot = sorted[i].slot;
1386 bytenr = sorted[i].bytenr;
1387
1388 if (level == 0) {
1389 btrfs_item_key_to_cpu(buf, &key, slot);
1390 fi = btrfs_item_ptr(buf, slot,
1391 struct btrfs_file_extent_item);
1392
1393 bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
1394 if (bytenr == 0)
1395 continue;
1396
1397 ret = process_func(trans, root, bytenr,
1398 btrfs_file_extent_disk_num_bytes(buf, fi),
1399 orig_buf->start, buf->start,
1400 orig_root, ref_root,
1401 orig_generation, ref_generation,
1402 key.objectid);
1403
1404 if (ret) {
1405 faili = slot;
1406 WARN_ON(1);
1407 goto fail;
1408 }
1409 } else {
1410 ret = process_func(trans, root, bytenr, buf->len,
1411 orig_buf->start, buf->start,
1412 orig_root, ref_root,
1413 orig_generation, ref_generation,
1414 level - 1);
1415 if (ret) {
1416 faili = slot;
1417 WARN_ON(1);
1418 goto fail;
1419 }
1420 }
1421 }
1422 out:
1423 kfree(sorted);
1424 if (nr_extents) {
1425 if (level == 0)
1426 *nr_extents = nr_file_extents;
1427 else
1428 *nr_extents = nritems;
1429 }
1430 return 0;
1431 fail:
1432 kfree(sorted);
1433 WARN_ON(1);
1434 return ret;
1435 }
1436
1437 int btrfs_update_ref(struct btrfs_trans_handle *trans,
1438 struct btrfs_root *root, struct extent_buffer *orig_buf,
1439 struct extent_buffer *buf, int start_slot, int nr)
1440
1441 {
1442 u64 bytenr;
1443 u64 ref_root;
1444 u64 orig_root;
1445 u64 ref_generation;
1446 u64 orig_generation;
1447 struct btrfs_key key;
1448 struct btrfs_file_extent_item *fi;
1449 int i;
1450 int ret;
1451 int slot;
1452 int level;
1453
1454 BUG_ON(start_slot < 0);
1455 BUG_ON(start_slot + nr > btrfs_header_nritems(buf));
1456
1457 ref_root = btrfs_header_owner(buf);
1458 ref_generation = btrfs_header_generation(buf);
1459 orig_root = btrfs_header_owner(orig_buf);
1460 orig_generation = btrfs_header_generation(orig_buf);
1461 level = btrfs_header_level(buf);
1462
1463 if (!root->ref_cows) {
1464 if (level == 0 &&
1465 root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID)
1466 return 0;
1467 if (level != 0 &&
1468 root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID)
1469 return 0;
1470 }
1471
1472 for (i = 0, slot = start_slot; i < nr; i++, slot++) {
1473 cond_resched();
1474 if (level == 0) {
1475 btrfs_item_key_to_cpu(buf, &key, slot);
1476 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
1477 continue;
1478 fi = btrfs_item_ptr(buf, slot,
1479 struct btrfs_file_extent_item);
1480 if (btrfs_file_extent_type(buf, fi) ==
1481 BTRFS_FILE_EXTENT_INLINE)
1482 continue;
1483 bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
1484 if (bytenr == 0)
1485 continue;
1486 ret = __btrfs_update_extent_ref(trans, root, bytenr,
1487 btrfs_file_extent_disk_num_bytes(buf, fi),
1488 orig_buf->start, buf->start,
1489 orig_root, ref_root, orig_generation,
1490 ref_generation, key.objectid);
1491 if (ret)
1492 goto fail;
1493 } else {
1494 bytenr = btrfs_node_blockptr(buf, slot);
1495 ret = __btrfs_update_extent_ref(trans, root, bytenr,
1496 buf->len, orig_buf->start,
1497 buf->start, orig_root, ref_root,
1498 orig_generation, ref_generation,
1499 level - 1);
1500 if (ret)
1501 goto fail;
1502 }
1503 }
1504 return 0;
1505 fail:
1506 WARN_ON(1);
1507 return -1;
1508 }
1509
1510 static int write_one_cache_group(struct btrfs_trans_handle *trans,
1511 struct btrfs_root *root,
1512 struct btrfs_path *path,
1513 struct btrfs_block_group_cache *cache)
1514 {
1515 int ret;
1516 struct btrfs_root *extent_root = root->fs_info->extent_root;
1517 unsigned long bi;
1518 struct extent_buffer *leaf;
1519
1520 ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1);
1521 if (ret < 0)
1522 goto fail;
1523 BUG_ON(ret);
1524
1525 leaf = path->nodes[0];
1526 bi = btrfs_item_ptr_offset(leaf, path->slots[0]);
1527 write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item));
1528 btrfs_mark_buffer_dirty(leaf);
1529 btrfs_release_path(extent_root, path);
1530 fail:
1531 if (ret)
1532 return ret;
1533 return 0;
1534
1535 }
1536
1537 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
1538 struct btrfs_root *root)
1539 {
1540 struct btrfs_block_group_cache *cache, *entry;
1541 struct rb_node *n;
1542 int err = 0;
1543 int werr = 0;
1544 struct btrfs_path *path;
1545 u64 last = 0;
1546
1547 path = btrfs_alloc_path();
1548 if (!path)
1549 return -ENOMEM;
1550
1551 while (1) {
1552 cache = NULL;
1553 spin_lock(&root->fs_info->block_group_cache_lock);
1554 for (n = rb_first(&root->fs_info->block_group_cache_tree);
1555 n; n = rb_next(n)) {
1556 entry = rb_entry(n, struct btrfs_block_group_cache,
1557 cache_node);
1558 if (entry->dirty) {
1559 cache = entry;
1560 break;
1561 }
1562 }
1563 spin_unlock(&root->fs_info->block_group_cache_lock);
1564
1565 if (!cache)
1566 break;
1567
1568 cache->dirty = 0;
1569 last += cache->key.offset;
1570
1571 err = write_one_cache_group(trans, root,
1572 path, cache);
1573 /*
1574 * if we fail to write the cache group, we want
1575 * to keep it marked dirty in hopes that a later
1576 * write will work
1577 */
1578 if (err) {
1579 werr = err;
1580 continue;
1581 }
1582 }
1583 btrfs_free_path(path);
1584 return werr;
1585 }
1586
1587 int btrfs_extent_readonly(struct btrfs_root *root, u64 bytenr)
1588 {
1589 struct btrfs_block_group_cache *block_group;
1590 int readonly = 0;
1591
1592 block_group = btrfs_lookup_block_group(root->fs_info, bytenr);
1593 if (!block_group || block_group->ro)
1594 readonly = 1;
1595 if (block_group)
1596 btrfs_put_block_group(block_group);
1597 return readonly;
1598 }
1599
1600 static int update_space_info(struct btrfs_fs_info *info, u64 flags,
1601 u64 total_bytes, u64 bytes_used,
1602 struct btrfs_space_info **space_info)
1603 {
1604 struct btrfs_space_info *found;
1605
1606 found = __find_space_info(info, flags);
1607 if (found) {
1608 spin_lock(&found->lock);
1609 found->total_bytes += total_bytes;
1610 found->bytes_used += bytes_used;
1611 found->full = 0;
1612 spin_unlock(&found->lock);
1613 *space_info = found;
1614 return 0;
1615 }
1616 found = kzalloc(sizeof(*found), GFP_NOFS);
1617 if (!found)
1618 return -ENOMEM;
1619
1620 INIT_LIST_HEAD(&found->block_groups);
1621 init_rwsem(&found->groups_sem);
1622 spin_lock_init(&found->lock);
1623 found->flags = flags;
1624 found->total_bytes = total_bytes;
1625 found->bytes_used = bytes_used;
1626 found->bytes_pinned = 0;
1627 found->bytes_reserved = 0;
1628 found->bytes_readonly = 0;
1629 found->bytes_delalloc = 0;
1630 found->full = 0;
1631 found->force_alloc = 0;
1632 *space_info = found;
1633 list_add_rcu(&found->list, &info->space_info);
1634 return 0;
1635 }
1636
1637 static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags)
1638 {
1639 u64 extra_flags = flags & (BTRFS_BLOCK_GROUP_RAID0 |
1640 BTRFS_BLOCK_GROUP_RAID1 |
1641 BTRFS_BLOCK_GROUP_RAID10 |
1642 BTRFS_BLOCK_GROUP_DUP);
1643 if (extra_flags) {
1644 if (flags & BTRFS_BLOCK_GROUP_DATA)
1645 fs_info->avail_data_alloc_bits |= extra_flags;
1646 if (flags & BTRFS_BLOCK_GROUP_METADATA)
1647 fs_info->avail_metadata_alloc_bits |= extra_flags;
1648 if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
1649 fs_info->avail_system_alloc_bits |= extra_flags;
1650 }
1651 }
1652
1653 static void set_block_group_readonly(struct btrfs_block_group_cache *cache)
1654 {
1655 spin_lock(&cache->space_info->lock);
1656 spin_lock(&cache->lock);
1657 if (!cache->ro) {
1658 cache->space_info->bytes_readonly += cache->key.offset -
1659 btrfs_block_group_used(&cache->item);
1660 cache->ro = 1;
1661 }
1662 spin_unlock(&cache->lock);
1663 spin_unlock(&cache->space_info->lock);
1664 }
1665
1666 u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags)
1667 {
1668 u64 num_devices = root->fs_info->fs_devices->rw_devices;
1669
1670 if (num_devices == 1)
1671 flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0);
1672 if (num_devices < 4)
1673 flags &= ~BTRFS_BLOCK_GROUP_RAID10;
1674
1675 if ((flags & BTRFS_BLOCK_GROUP_DUP) &&
1676 (flags & (BTRFS_BLOCK_GROUP_RAID1 |
1677 BTRFS_BLOCK_GROUP_RAID10))) {
1678 flags &= ~BTRFS_BLOCK_GROUP_DUP;
1679 }
1680
1681 if ((flags & BTRFS_BLOCK_GROUP_RAID1) &&
1682 (flags & BTRFS_BLOCK_GROUP_RAID10)) {
1683 flags &= ~BTRFS_BLOCK_GROUP_RAID1;
1684 }
1685
1686 if ((flags & BTRFS_BLOCK_GROUP_RAID0) &&
1687 ((flags & BTRFS_BLOCK_GROUP_RAID1) |
1688 (flags & BTRFS_BLOCK_GROUP_RAID10) |
1689 (flags & BTRFS_BLOCK_GROUP_DUP)))
1690 flags &= ~BTRFS_BLOCK_GROUP_RAID0;
1691 return flags;
1692 }
1693
1694 static u64 btrfs_get_alloc_profile(struct btrfs_root *root, u64 data)
1695 {
1696 struct btrfs_fs_info *info = root->fs_info;
1697 u64 alloc_profile;
1698
1699 if (data) {
1700 alloc_profile = info->avail_data_alloc_bits &
1701 info->data_alloc_profile;
1702 data = BTRFS_BLOCK_GROUP_DATA | alloc_profile;
1703 } else if (root == root->fs_info->chunk_root) {
1704 alloc_profile = info->avail_system_alloc_bits &
1705 info->system_alloc_profile;
1706 data = BTRFS_BLOCK_GROUP_SYSTEM | alloc_profile;
1707 } else {
1708 alloc_profile = info->avail_metadata_alloc_bits &
1709 info->metadata_alloc_profile;
1710 data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile;
1711 }
1712
1713 return btrfs_reduce_alloc_profile(root, data);
1714 }
1715
1716 void btrfs_set_inode_space_info(struct btrfs_root *root, struct inode *inode)
1717 {
1718 u64 alloc_target;
1719
1720 alloc_target = btrfs_get_alloc_profile(root, 1);
1721 BTRFS_I(inode)->space_info = __find_space_info(root->fs_info,
1722 alloc_target);
1723 }
1724
1725 /*
1726 * for now this just makes sure we have at least 5% of our metadata space free
1727 * for use.
1728 */
1729 int btrfs_check_metadata_free_space(struct btrfs_root *root)
1730 {
1731 struct btrfs_fs_info *info = root->fs_info;
1732 struct btrfs_space_info *meta_sinfo;
1733 u64 alloc_target, thresh;
1734 int committed = 0, ret;
1735
1736 /* get the space info for where the metadata will live */
1737 alloc_target = btrfs_get_alloc_profile(root, 0);
1738 meta_sinfo = __find_space_info(info, alloc_target);
1739
1740 again:
1741 spin_lock(&meta_sinfo->lock);
1742 if (!meta_sinfo->full)
1743 thresh = meta_sinfo->total_bytes * 80;
1744 else
1745 thresh = meta_sinfo->total_bytes * 95;
1746
1747 do_div(thresh, 100);
1748
1749 if (meta_sinfo->bytes_used + meta_sinfo->bytes_reserved +
1750 meta_sinfo->bytes_pinned + meta_sinfo->bytes_readonly > thresh) {
1751 struct btrfs_trans_handle *trans;
1752 if (!meta_sinfo->full) {
1753 meta_sinfo->force_alloc = 1;
1754 spin_unlock(&meta_sinfo->lock);
1755
1756 trans = btrfs_start_transaction(root, 1);
1757 if (!trans)
1758 return -ENOMEM;
1759
1760 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
1761 2 * 1024 * 1024, alloc_target, 0);
1762 btrfs_end_transaction(trans, root);
1763 goto again;
1764 }
1765 spin_unlock(&meta_sinfo->lock);
1766
1767 if (!committed) {
1768 committed = 1;
1769 trans = btrfs_join_transaction(root, 1);
1770 if (!trans)
1771 return -ENOMEM;
1772 ret = btrfs_commit_transaction(trans, root);
1773 if (ret)
1774 return ret;
1775 goto again;
1776 }
1777 return -ENOSPC;
1778 }
1779 spin_unlock(&meta_sinfo->lock);
1780
1781 return 0;
1782 }
1783
1784 /*
1785 * This will check the space that the inode allocates from to make sure we have
1786 * enough space for bytes.
1787 */
1788 int btrfs_check_data_free_space(struct btrfs_root *root, struct inode *inode,
1789 u64 bytes)
1790 {
1791 struct btrfs_space_info *data_sinfo;
1792 int ret = 0, committed = 0;
1793
1794 /* make sure bytes are sectorsize aligned */
1795 bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
1796
1797 data_sinfo = BTRFS_I(inode)->space_info;
1798 again:
1799 /* make sure we have enough space to handle the data first */
1800 spin_lock(&data_sinfo->lock);
1801 if (data_sinfo->total_bytes - data_sinfo->bytes_used -
1802 data_sinfo->bytes_delalloc - data_sinfo->bytes_reserved -
1803 data_sinfo->bytes_pinned - data_sinfo->bytes_readonly -
1804 data_sinfo->bytes_may_use < bytes) {
1805 struct btrfs_trans_handle *trans;
1806
1807 /*
1808 * if we don't have enough free bytes in this space then we need
1809 * to alloc a new chunk.
1810 */
1811 if (!data_sinfo->full) {
1812 u64 alloc_target;
1813
1814 data_sinfo->force_alloc = 1;
1815 spin_unlock(&data_sinfo->lock);
1816
1817 alloc_target = btrfs_get_alloc_profile(root, 1);
1818 trans = btrfs_start_transaction(root, 1);
1819 if (!trans)
1820 return -ENOMEM;
1821
1822 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
1823 bytes + 2 * 1024 * 1024,
1824 alloc_target, 0);
1825 btrfs_end_transaction(trans, root);
1826 if (ret)
1827 return ret;
1828 goto again;
1829 }
1830 spin_unlock(&data_sinfo->lock);
1831
1832 /* commit the current transaction and try again */
1833 if (!committed) {
1834 committed = 1;
1835 trans = btrfs_join_transaction(root, 1);
1836 if (!trans)
1837 return -ENOMEM;
1838 ret = btrfs_commit_transaction(trans, root);
1839 if (ret)
1840 return ret;
1841 goto again;
1842 }
1843
1844 printk(KERN_ERR "no space left, need %llu, %llu delalloc bytes"
1845 ", %llu bytes_used, %llu bytes_reserved, "
1846 "%llu bytes_pinned, %llu bytes_readonly, %llu may use"
1847 "%llu total\n", bytes, data_sinfo->bytes_delalloc,
1848 data_sinfo->bytes_used, data_sinfo->bytes_reserved,
1849 data_sinfo->bytes_pinned, data_sinfo->bytes_readonly,
1850 data_sinfo->bytes_may_use, data_sinfo->total_bytes);
1851 return -ENOSPC;
1852 }
1853 data_sinfo->bytes_may_use += bytes;
1854 BTRFS_I(inode)->reserved_bytes += bytes;
1855 spin_unlock(&data_sinfo->lock);
1856
1857 return btrfs_check_metadata_free_space(root);
1858 }
1859
1860 /*
1861 * if there was an error for whatever reason after calling
1862 * btrfs_check_data_free_space, call this so we can cleanup the counters.
1863 */
1864 void btrfs_free_reserved_data_space(struct btrfs_root *root,
1865 struct inode *inode, u64 bytes)
1866 {
1867 struct btrfs_space_info *data_sinfo;
1868
1869 /* make sure bytes are sectorsize aligned */
1870 bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
1871
1872 data_sinfo = BTRFS_I(inode)->space_info;
1873 spin_lock(&data_sinfo->lock);
1874 data_sinfo->bytes_may_use -= bytes;
1875 BTRFS_I(inode)->reserved_bytes -= bytes;
1876 spin_unlock(&data_sinfo->lock);
1877 }
1878
1879 /* called when we are adding a delalloc extent to the inode's io_tree */
1880 void btrfs_delalloc_reserve_space(struct btrfs_root *root, struct inode *inode,
1881 u64 bytes)
1882 {
1883 struct btrfs_space_info *data_sinfo;
1884
1885 /* get the space info for where this inode will be storing its data */
1886 data_sinfo = BTRFS_I(inode)->space_info;
1887
1888 /* make sure we have enough space to handle the data first */
1889 spin_lock(&data_sinfo->lock);
1890 data_sinfo->bytes_delalloc += bytes;
1891
1892 /*
1893 * we are adding a delalloc extent without calling
1894 * btrfs_check_data_free_space first. This happens on a weird
1895 * writepage condition, but shouldn't hurt our accounting
1896 */
1897 if (unlikely(bytes > BTRFS_I(inode)->reserved_bytes)) {
1898 data_sinfo->bytes_may_use -= BTRFS_I(inode)->reserved_bytes;
1899 BTRFS_I(inode)->reserved_bytes = 0;
1900 } else {
1901 data_sinfo->bytes_may_use -= bytes;
1902 BTRFS_I(inode)->reserved_bytes -= bytes;
1903 }
1904
1905 spin_unlock(&data_sinfo->lock);
1906 }
1907
1908 /* called when we are clearing an delalloc extent from the inode's io_tree */
1909 void btrfs_delalloc_free_space(struct btrfs_root *root, struct inode *inode,
1910 u64 bytes)
1911 {
1912 struct btrfs_space_info *info;
1913
1914 info = BTRFS_I(inode)->space_info;
1915
1916 spin_lock(&info->lock);
1917 info->bytes_delalloc -= bytes;
1918 spin_unlock(&info->lock);
1919 }
1920
1921 static int do_chunk_alloc(struct btrfs_trans_handle *trans,
1922 struct btrfs_root *extent_root, u64 alloc_bytes,
1923 u64 flags, int force)
1924 {
1925 struct btrfs_space_info *space_info;
1926 u64 thresh;
1927 int ret = 0;
1928
1929 mutex_lock(&extent_root->fs_info->chunk_mutex);
1930
1931 flags = btrfs_reduce_alloc_profile(extent_root, flags);
1932
1933 space_info = __find_space_info(extent_root->fs_info, flags);
1934 if (!space_info) {
1935 ret = update_space_info(extent_root->fs_info, flags,
1936 0, 0, &space_info);
1937 BUG_ON(ret);
1938 }
1939 BUG_ON(!space_info);
1940
1941 spin_lock(&space_info->lock);
1942 if (space_info->force_alloc) {
1943 force = 1;
1944 space_info->force_alloc = 0;
1945 }
1946 if (space_info->full) {
1947 spin_unlock(&space_info->lock);
1948 goto out;
1949 }
1950
1951 thresh = space_info->total_bytes - space_info->bytes_readonly;
1952 thresh = div_factor(thresh, 6);
1953 if (!force &&
1954 (space_info->bytes_used + space_info->bytes_pinned +
1955 space_info->bytes_reserved + alloc_bytes) < thresh) {
1956 spin_unlock(&space_info->lock);
1957 goto out;
1958 }
1959 spin_unlock(&space_info->lock);
1960
1961 ret = btrfs_alloc_chunk(trans, extent_root, flags);
1962 if (ret)
1963 space_info->full = 1;
1964 out:
1965 mutex_unlock(&extent_root->fs_info->chunk_mutex);
1966 return ret;
1967 }
1968
1969 static int update_block_group(struct btrfs_trans_handle *trans,
1970 struct btrfs_root *root,
1971 u64 bytenr, u64 num_bytes, int alloc,
1972 int mark_free)
1973 {
1974 struct btrfs_block_group_cache *cache;
1975 struct btrfs_fs_info *info = root->fs_info;
1976 u64 total = num_bytes;
1977 u64 old_val;
1978 u64 byte_in_group;
1979
1980 while (total) {
1981 cache = btrfs_lookup_block_group(info, bytenr);
1982 if (!cache)
1983 return -1;
1984 byte_in_group = bytenr - cache->key.objectid;
1985 WARN_ON(byte_in_group > cache->key.offset);
1986
1987 spin_lock(&cache->space_info->lock);
1988 spin_lock(&cache->lock);
1989 cache->dirty = 1;
1990 old_val = btrfs_block_group_used(&cache->item);
1991 num_bytes = min(total, cache->key.offset - byte_in_group);
1992 if (alloc) {
1993 old_val += num_bytes;
1994 cache->space_info->bytes_used += num_bytes;
1995 if (cache->ro)
1996 cache->space_info->bytes_readonly -= num_bytes;
1997 btrfs_set_block_group_used(&cache->item, old_val);
1998 spin_unlock(&cache->lock);
1999 spin_unlock(&cache->space_info->lock);
2000 } else {
2001 old_val -= num_bytes;
2002 cache->space_info->bytes_used -= num_bytes;
2003 if (cache->ro)
2004 cache->space_info->bytes_readonly += num_bytes;
2005 btrfs_set_block_group_used(&cache->item, old_val);
2006 spin_unlock(&cache->lock);
2007 spin_unlock(&cache->space_info->lock);
2008 if (mark_free) {
2009 int ret;
2010
2011 ret = btrfs_discard_extent(root, bytenr,
2012 num_bytes);
2013 WARN_ON(ret);
2014
2015 ret = btrfs_add_free_space(cache, bytenr,
2016 num_bytes);
2017 WARN_ON(ret);
2018 }
2019 }
2020 btrfs_put_block_group(cache);
2021 total -= num_bytes;
2022 bytenr += num_bytes;
2023 }
2024 return 0;
2025 }
2026
2027 static u64 first_logical_byte(struct btrfs_root *root, u64 search_start)
2028 {
2029 struct btrfs_block_group_cache *cache;
2030 u64 bytenr;
2031
2032 cache = btrfs_lookup_first_block_group(root->fs_info, search_start);
2033 if (!cache)
2034 return 0;
2035
2036 bytenr = cache->key.objectid;
2037 btrfs_put_block_group(cache);
2038
2039 return bytenr;
2040 }
2041
2042 int btrfs_update_pinned_extents(struct btrfs_root *root,
2043 u64 bytenr, u64 num, int pin)
2044 {
2045 u64 len;
2046 struct btrfs_block_group_cache *cache;
2047 struct btrfs_fs_info *fs_info = root->fs_info;
2048
2049 if (pin) {
2050 set_extent_dirty(&fs_info->pinned_extents,
2051 bytenr, bytenr + num - 1, GFP_NOFS);
2052 } else {
2053 clear_extent_dirty(&fs_info->pinned_extents,
2054 bytenr, bytenr + num - 1, GFP_NOFS);
2055 }
2056
2057 while (num > 0) {
2058 cache = btrfs_lookup_block_group(fs_info, bytenr);
2059 BUG_ON(!cache);
2060 len = min(num, cache->key.offset -
2061 (bytenr - cache->key.objectid));
2062 if (pin) {
2063 spin_lock(&cache->space_info->lock);
2064 spin_lock(&cache->lock);
2065 cache->pinned += len;
2066 cache->space_info->bytes_pinned += len;
2067 spin_unlock(&cache->lock);
2068 spin_unlock(&cache->space_info->lock);
2069 fs_info->total_pinned += len;
2070 } else {
2071 spin_lock(&cache->space_info->lock);
2072 spin_lock(&cache->lock);
2073 cache->pinned -= len;
2074 cache->space_info->bytes_pinned -= len;
2075 spin_unlock(&cache->lock);
2076 spin_unlock(&cache->space_info->lock);
2077 fs_info->total_pinned -= len;
2078 if (cache->cached)
2079 btrfs_add_free_space(cache, bytenr, len);
2080 }
2081 btrfs_put_block_group(cache);
2082 bytenr += len;
2083 num -= len;
2084 }
2085 return 0;
2086 }
2087
2088 static int update_reserved_extents(struct btrfs_root *root,
2089 u64 bytenr, u64 num, int reserve)
2090 {
2091 u64 len;
2092 struct btrfs_block_group_cache *cache;
2093 struct btrfs_fs_info *fs_info = root->fs_info;
2094
2095 while (num > 0) {
2096 cache = btrfs_lookup_block_group(fs_info, bytenr);
2097 BUG_ON(!cache);
2098 len = min(num, cache->key.offset -
2099 (bytenr - cache->key.objectid));
2100
2101 spin_lock(&cache->space_info->lock);
2102 spin_lock(&cache->lock);
2103 if (reserve) {
2104 cache->reserved += len;
2105 cache->space_info->bytes_reserved += len;
2106 } else {
2107 cache->reserved -= len;
2108 cache->space_info->bytes_reserved -= len;
2109 }
2110 spin_unlock(&cache->lock);
2111 spin_unlock(&cache->space_info->lock);
2112 btrfs_put_block_group(cache);
2113 bytenr += len;
2114 num -= len;
2115 }
2116 return 0;
2117 }
2118
2119 int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy)
2120 {
2121 u64 last = 0;
2122 u64 start;
2123 u64 end;
2124 struct extent_io_tree *pinned_extents = &root->fs_info->pinned_extents;
2125 int ret;
2126
2127 while (1) {
2128 ret = find_first_extent_bit(pinned_extents, last,
2129 &start, &end, EXTENT_DIRTY);
2130 if (ret)
2131 break;
2132 set_extent_dirty(copy, start, end, GFP_NOFS);
2133 last = end + 1;
2134 }
2135 return 0;
2136 }
2137
2138 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
2139 struct btrfs_root *root,
2140 struct extent_io_tree *unpin)
2141 {
2142 u64 start;
2143 u64 end;
2144 int ret;
2145
2146 while (1) {
2147 ret = find_first_extent_bit(unpin, 0, &start, &end,
2148 EXTENT_DIRTY);
2149 if (ret)
2150 break;
2151
2152 ret = btrfs_discard_extent(root, start, end + 1 - start);
2153
2154 /* unlocks the pinned mutex */
2155 btrfs_update_pinned_extents(root, start, end + 1 - start, 0);
2156 clear_extent_dirty(unpin, start, end, GFP_NOFS);
2157
2158 cond_resched();
2159 }
2160 return ret;
2161 }
2162
2163 static int pin_down_bytes(struct btrfs_trans_handle *trans,
2164 struct btrfs_root *root,
2165 struct btrfs_path *path,
2166 u64 bytenr, u64 num_bytes, int is_data,
2167 struct extent_buffer **must_clean)
2168 {
2169 int err = 0;
2170 struct extent_buffer *buf;
2171
2172 if (is_data)
2173 goto pinit;
2174
2175 buf = btrfs_find_tree_block(root, bytenr, num_bytes);
2176 if (!buf)
2177 goto pinit;
2178
2179 /* we can reuse a block if it hasn't been written
2180 * and it is from this transaction. We can't
2181 * reuse anything from the tree log root because
2182 * it has tiny sub-transactions.
2183 */
2184 if (btrfs_buffer_uptodate(buf, 0) &&
2185 btrfs_try_tree_lock(buf)) {
2186 u64 header_owner = btrfs_header_owner(buf);
2187 u64 header_transid = btrfs_header_generation(buf);
2188 if (header_owner != BTRFS_TREE_LOG_OBJECTID &&
2189 header_owner != BTRFS_TREE_RELOC_OBJECTID &&
2190 header_owner != BTRFS_DATA_RELOC_TREE_OBJECTID &&
2191 header_transid == trans->transid &&
2192 !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
2193 *must_clean = buf;
2194 return 1;
2195 }
2196 btrfs_tree_unlock(buf);
2197 }
2198 free_extent_buffer(buf);
2199 pinit:
2200 btrfs_set_path_blocking(path);
2201 /* unlocks the pinned mutex */
2202 btrfs_update_pinned_extents(root, bytenr, num_bytes, 1);
2203
2204 BUG_ON(err < 0);
2205 return 0;
2206 }
2207
2208 /*
2209 * remove an extent from the root, returns 0 on success
2210 */
2211 static int __free_extent(struct btrfs_trans_handle *trans,
2212 struct btrfs_root *root,
2213 u64 bytenr, u64 num_bytes, u64 parent,
2214 u64 root_objectid, u64 ref_generation,
2215 u64 owner_objectid, int pin, int mark_free,
2216 int refs_to_drop)
2217 {
2218 struct btrfs_path *path;
2219 struct btrfs_key key;
2220 struct btrfs_fs_info *info = root->fs_info;
2221 struct btrfs_root *extent_root = info->extent_root;
2222 struct extent_buffer *leaf;
2223 int ret;
2224 int extent_slot = 0;
2225 int found_extent = 0;
2226 int num_to_del = 1;
2227 struct btrfs_extent_item *ei;
2228 u32 refs;
2229
2230 key.objectid = bytenr;
2231 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
2232 key.offset = num_bytes;
2233 path = btrfs_alloc_path();
2234 if (!path)
2235 return -ENOMEM;
2236
2237 path->reada = 1;
2238 path->leave_spinning = 1;
2239 ret = lookup_extent_backref(trans, extent_root, path,
2240 bytenr, parent, root_objectid,
2241 ref_generation, owner_objectid, 1);
2242 if (ret == 0) {
2243 struct btrfs_key found_key;
2244 extent_slot = path->slots[0];
2245 while (extent_slot > 0) {
2246 extent_slot--;
2247 btrfs_item_key_to_cpu(path->nodes[0], &found_key,
2248 extent_slot);
2249 if (found_key.objectid != bytenr)
2250 break;
2251 if (found_key.type == BTRFS_EXTENT_ITEM_KEY &&
2252 found_key.offset == num_bytes) {
2253 found_extent = 1;
2254 break;
2255 }
2256 if (path->slots[0] - extent_slot > 5)
2257 break;
2258 }
2259 if (!found_extent) {
2260 ret = remove_extent_backref(trans, extent_root, path,
2261 refs_to_drop);
2262 BUG_ON(ret);
2263 btrfs_release_path(extent_root, path);
2264 path->leave_spinning = 1;
2265 ret = btrfs_search_slot(trans, extent_root,
2266 &key, path, -1, 1);
2267 if (ret) {
2268 printk(KERN_ERR "umm, got %d back from search"
2269 ", was looking for %llu\n", ret,
2270 (unsigned long long)bytenr);
2271 btrfs_print_leaf(extent_root, path->nodes[0]);
2272 }
2273 BUG_ON(ret);
2274 extent_slot = path->slots[0];
2275 }
2276 } else {
2277 btrfs_print_leaf(extent_root, path->nodes[0]);
2278 WARN_ON(1);
2279 printk(KERN_ERR "btrfs unable to find ref byte nr %llu "
2280 "parent %llu root %llu gen %llu owner %llu\n",
2281 (unsigned long long)bytenr,
2282 (unsigned long long)parent,
2283 (unsigned long long)root_objectid,
2284 (unsigned long long)ref_generation,
2285 (unsigned long long)owner_objectid);
2286 }
2287
2288 leaf = path->nodes[0];
2289 ei = btrfs_item_ptr(leaf, extent_slot,
2290 struct btrfs_extent_item);
2291 refs = btrfs_extent_refs(leaf, ei);
2292
2293 /*
2294 * we're not allowed to delete the extent item if there
2295 * are other delayed ref updates pending
2296 */
2297
2298 BUG_ON(refs < refs_to_drop);
2299 refs -= refs_to_drop;
2300 btrfs_set_extent_refs(leaf, ei, refs);
2301 btrfs_mark_buffer_dirty(leaf);
2302
2303 if (refs == 0 && found_extent &&
2304 path->slots[0] == extent_slot + 1) {
2305 struct btrfs_extent_ref *ref;
2306 ref = btrfs_item_ptr(leaf, path->slots[0],
2307 struct btrfs_extent_ref);
2308 BUG_ON(btrfs_ref_num_refs(leaf, ref) != refs_to_drop);
2309 /* if the back ref and the extent are next to each other
2310 * they get deleted below in one shot
2311 */
2312 path->slots[0] = extent_slot;
2313 num_to_del = 2;
2314 } else if (found_extent) {
2315 /* otherwise delete the extent back ref */
2316 ret = remove_extent_backref(trans, extent_root, path,
2317 refs_to_drop);
2318 BUG_ON(ret);
2319 /* if refs are 0, we need to setup the path for deletion */
2320 if (refs == 0) {
2321 btrfs_release_path(extent_root, path);
2322 path->leave_spinning = 1;
2323 ret = btrfs_search_slot(trans, extent_root, &key, path,
2324 -1, 1);
2325 BUG_ON(ret);
2326 }
2327 }
2328
2329 if (refs == 0) {
2330 u64 super_used;
2331 u64 root_used;
2332 struct extent_buffer *must_clean = NULL;
2333
2334 if (pin) {
2335 ret = pin_down_bytes(trans, root, path,
2336 bytenr, num_bytes,
2337 owner_objectid >= BTRFS_FIRST_FREE_OBJECTID,
2338 &must_clean);
2339 if (ret > 0)
2340 mark_free = 1;
2341 BUG_ON(ret < 0);
2342 }
2343
2344 /* block accounting for super block */
2345 spin_lock(&info->delalloc_lock);
2346 super_used = btrfs_super_bytes_used(&info->super_copy);
2347 btrfs_set_super_bytes_used(&info->super_copy,
2348 super_used - num_bytes);
2349
2350 /* block accounting for root item */
2351 root_used = btrfs_root_used(&root->root_item);
2352 btrfs_set_root_used(&root->root_item,
2353 root_used - num_bytes);
2354 spin_unlock(&info->delalloc_lock);
2355
2356 /*
2357 * it is going to be very rare for someone to be waiting
2358 * on the block we're freeing. del_items might need to
2359 * schedule, so rather than get fancy, just force it
2360 * to blocking here
2361 */
2362 if (must_clean)
2363 btrfs_set_lock_blocking(must_clean);
2364
2365 ret = btrfs_del_items(trans, extent_root, path, path->slots[0],
2366 num_to_del);
2367 BUG_ON(ret);
2368 btrfs_release_path(extent_root, path);
2369
2370 if (must_clean) {
2371 clean_tree_block(NULL, root, must_clean);
2372 btrfs_tree_unlock(must_clean);
2373 free_extent_buffer(must_clean);
2374 }
2375
2376 if (owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
2377 ret = btrfs_del_csums(trans, root, bytenr, num_bytes);
2378 BUG_ON(ret);
2379 } else {
2380 invalidate_mapping_pages(info->btree_inode->i_mapping,
2381 bytenr >> PAGE_CACHE_SHIFT,
2382 (bytenr + num_bytes - 1) >> PAGE_CACHE_SHIFT);
2383 }
2384
2385 ret = update_block_group(trans, root, bytenr, num_bytes, 0,
2386 mark_free);
2387 BUG_ON(ret);
2388 }
2389 btrfs_free_path(path);
2390 return ret;
2391 }
2392
2393 /*
2394 * remove an extent from the root, returns 0 on success
2395 */
2396 static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
2397 struct btrfs_root *root,
2398 u64 bytenr, u64 num_bytes, u64 parent,
2399 u64 root_objectid, u64 ref_generation,
2400 u64 owner_objectid, int pin,
2401 int refs_to_drop)
2402 {
2403 WARN_ON(num_bytes < root->sectorsize);
2404
2405 /*
2406 * if metadata always pin
2407 * if data pin when any transaction has committed this
2408 */
2409 if (owner_objectid < BTRFS_FIRST_FREE_OBJECTID ||
2410 ref_generation != trans->transid)
2411 pin = 1;
2412
2413 if (ref_generation != trans->transid)
2414 pin = 1;
2415
2416 return __free_extent(trans, root, bytenr, num_bytes, parent,
2417 root_objectid, ref_generation,
2418 owner_objectid, pin, pin == 0, refs_to_drop);
2419 }
2420
2421 /*
2422 * when we free an extent, it is possible (and likely) that we free the last
2423 * delayed ref for that extent as well. This searches the delayed ref tree for
2424 * a given extent, and if there are no other delayed refs to be processed, it
2425 * removes it from the tree.
2426 */
2427 static noinline int check_ref_cleanup(struct btrfs_trans_handle *trans,
2428 struct btrfs_root *root, u64 bytenr)
2429 {
2430 struct btrfs_delayed_ref_head *head;
2431 struct btrfs_delayed_ref_root *delayed_refs;
2432 struct btrfs_delayed_ref_node *ref;
2433 struct rb_node *node;
2434 int ret;
2435
2436 delayed_refs = &trans->transaction->delayed_refs;
2437 spin_lock(&delayed_refs->lock);
2438 head = btrfs_find_delayed_ref_head(trans, bytenr);
2439 if (!head)
2440 goto out;
2441
2442 node = rb_prev(&head->node.rb_node);
2443 if (!node)
2444 goto out;
2445
2446 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
2447
2448 /* there are still entries for this ref, we can't drop it */
2449 if (ref->bytenr == bytenr)
2450 goto out;
2451
2452 /*
2453 * waiting for the lock here would deadlock. If someone else has it
2454 * locked they are already in the process of dropping it anyway
2455 */
2456 if (!mutex_trylock(&head->mutex))
2457 goto out;
2458
2459 /*
2460 * at this point we have a head with no other entries. Go
2461 * ahead and process it.
2462 */
2463 head->node.in_tree = 0;
2464 rb_erase(&head->node.rb_node, &delayed_refs->root);
2465
2466 delayed_refs->num_entries--;
2467
2468 /*
2469 * we don't take a ref on the node because we're removing it from the
2470 * tree, so we just steal the ref the tree was holding.
2471 */
2472 delayed_refs->num_heads--;
2473 if (list_empty(&head->cluster))
2474 delayed_refs->num_heads_ready--;
2475
2476 list_del_init(&head->cluster);
2477 spin_unlock(&delayed_refs->lock);
2478
2479 ret = run_one_delayed_ref(trans, root->fs_info->tree_root,
2480 &head->node, head->must_insert_reserved);
2481 BUG_ON(ret);
2482 btrfs_put_delayed_ref(&head->node);
2483 return 0;
2484 out:
2485 spin_unlock(&delayed_refs->lock);
2486 return 0;
2487 }
2488
2489 int btrfs_free_extent(struct btrfs_trans_handle *trans,
2490 struct btrfs_root *root,
2491 u64 bytenr, u64 num_bytes, u64 parent,
2492 u64 root_objectid, u64 ref_generation,
2493 u64 owner_objectid, int pin)
2494 {
2495 int ret;
2496
2497 /*
2498 * tree log blocks never actually go into the extent allocation
2499 * tree, just update pinning info and exit early.
2500 *
2501 * data extents referenced by the tree log do need to have
2502 * their reference counts bumped.
2503 */
2504 if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID &&
2505 owner_objectid < BTRFS_FIRST_FREE_OBJECTID) {
2506 /* unlocks the pinned mutex */
2507 btrfs_update_pinned_extents(root, bytenr, num_bytes, 1);
2508 update_reserved_extents(root, bytenr, num_bytes, 0);
2509 ret = 0;
2510 } else {
2511 ret = btrfs_add_delayed_ref(trans, bytenr, num_bytes, parent,
2512 root_objectid, ref_generation,
2513 owner_objectid,
2514 BTRFS_DROP_DELAYED_REF, 1);
2515 BUG_ON(ret);
2516 ret = check_ref_cleanup(trans, root, bytenr);
2517 BUG_ON(ret);
2518 }
2519 return ret;
2520 }
2521
2522 static u64 stripe_align(struct btrfs_root *root, u64 val)
2523 {
2524 u64 mask = ((u64)root->stripesize - 1);
2525 u64 ret = (val + mask) & ~mask;
2526 return ret;
2527 }
2528
2529 /*
2530 * walks the btree of allocated extents and find a hole of a given size.
2531 * The key ins is changed to record the hole:
2532 * ins->objectid == block start
2533 * ins->flags = BTRFS_EXTENT_ITEM_KEY
2534 * ins->offset == number of blocks
2535 * Any available blocks before search_start are skipped.
2536 */
2537 static noinline int find_free_extent(struct btrfs_trans_handle *trans,
2538 struct btrfs_root *orig_root,
2539 u64 num_bytes, u64 empty_size,
2540 u64 search_start, u64 search_end,
2541 u64 hint_byte, struct btrfs_key *ins,
2542 u64 exclude_start, u64 exclude_nr,
2543 int data)
2544 {
2545 int ret = 0;
2546 struct btrfs_root *root = orig_root->fs_info->extent_root;
2547 struct btrfs_free_cluster *last_ptr = NULL;
2548 struct btrfs_block_group_cache *block_group = NULL;
2549 int empty_cluster = 2 * 1024 * 1024;
2550 int allowed_chunk_alloc = 0;
2551 struct btrfs_space_info *space_info;
2552 int last_ptr_loop = 0;
2553 int loop = 0;
2554
2555 WARN_ON(num_bytes < root->sectorsize);
2556 btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
2557 ins->objectid = 0;
2558 ins->offset = 0;
2559
2560 space_info = __find_space_info(root->fs_info, data);
2561
2562 if (orig_root->ref_cows || empty_size)
2563 allowed_chunk_alloc = 1;
2564
2565 if (data & BTRFS_BLOCK_GROUP_METADATA) {
2566 last_ptr = &root->fs_info->meta_alloc_cluster;
2567 if (!btrfs_test_opt(root, SSD))
2568 empty_cluster = 64 * 1024;
2569 }
2570
2571 if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD)) {
2572 last_ptr = &root->fs_info->data_alloc_cluster;
2573 }
2574
2575 if (last_ptr) {
2576 spin_lock(&last_ptr->lock);
2577 if (last_ptr->block_group)
2578 hint_byte = last_ptr->window_start;
2579 spin_unlock(&last_ptr->lock);
2580 }
2581
2582 search_start = max(search_start, first_logical_byte(root, 0));
2583 search_start = max(search_start, hint_byte);
2584
2585 if (!last_ptr) {
2586 empty_cluster = 0;
2587 loop = 1;
2588 }
2589
2590 if (search_start == hint_byte) {
2591 block_group = btrfs_lookup_block_group(root->fs_info,
2592 search_start);
2593 if (block_group && block_group_bits(block_group, data)) {
2594 down_read(&space_info->groups_sem);
2595 goto have_block_group;
2596 } else if (block_group) {
2597 btrfs_put_block_group(block_group);
2598 }
2599 }
2600
2601 search:
2602 down_read(&space_info->groups_sem);
2603 list_for_each_entry(block_group, &space_info->block_groups, list) {
2604 u64 offset;
2605
2606 atomic_inc(&block_group->count);
2607 search_start = block_group->key.objectid;
2608
2609 have_block_group:
2610 if (unlikely(!block_group->cached)) {
2611 mutex_lock(&block_group->cache_mutex);
2612 ret = cache_block_group(root, block_group);
2613 mutex_unlock(&block_group->cache_mutex);
2614 if (ret) {
2615 btrfs_put_block_group(block_group);
2616 break;
2617 }
2618 }
2619
2620 if (unlikely(block_group->ro))
2621 goto loop;
2622
2623 if (last_ptr) {
2624 /*
2625 * the refill lock keeps out other
2626 * people trying to start a new cluster
2627 */
2628 spin_lock(&last_ptr->refill_lock);
2629 offset = btrfs_alloc_from_cluster(block_group, last_ptr,
2630 num_bytes, search_start);
2631 if (offset) {
2632 /* we have a block, we're done */
2633 spin_unlock(&last_ptr->refill_lock);
2634 goto checks;
2635 }
2636
2637 spin_lock(&last_ptr->lock);
2638 /*
2639 * whoops, this cluster doesn't actually point to
2640 * this block group. Get a ref on the block
2641 * group is does point to and try again
2642 */
2643 if (!last_ptr_loop && last_ptr->block_group &&
2644 last_ptr->block_group != block_group) {
2645
2646 btrfs_put_block_group(block_group);
2647 block_group = last_ptr->block_group;
2648 atomic_inc(&block_group->count);
2649 spin_unlock(&last_ptr->lock);
2650 spin_unlock(&last_ptr->refill_lock);
2651
2652 last_ptr_loop = 1;
2653 search_start = block_group->key.objectid;
2654 goto have_block_group;
2655 }
2656 spin_unlock(&last_ptr->lock);
2657
2658 /*
2659 * this cluster didn't work out, free it and
2660 * start over
2661 */
2662 btrfs_return_cluster_to_free_space(NULL, last_ptr);
2663
2664 last_ptr_loop = 0;
2665
2666 /* allocate a cluster in this block group */
2667 ret = btrfs_find_space_cluster(trans,
2668 block_group, last_ptr,
2669 offset, num_bytes,
2670 empty_cluster + empty_size);
2671 if (ret == 0) {
2672 /*
2673 * now pull our allocation out of this
2674 * cluster
2675 */
2676 offset = btrfs_alloc_from_cluster(block_group,
2677 last_ptr, num_bytes,
2678 search_start);
2679 if (offset) {
2680 /* we found one, proceed */
2681 spin_unlock(&last_ptr->refill_lock);
2682 goto checks;
2683 }
2684 }
2685 /*
2686 * at this point we either didn't find a cluster
2687 * or we weren't able to allocate a block from our
2688 * cluster. Free the cluster we've been trying
2689 * to use, and go to the next block group
2690 */
2691 if (loop < 2) {
2692 btrfs_return_cluster_to_free_space(NULL,
2693 last_ptr);
2694 spin_unlock(&last_ptr->refill_lock);
2695 goto loop;
2696 }
2697 spin_unlock(&last_ptr->refill_lock);
2698 }
2699
2700 offset = btrfs_find_space_for_alloc(block_group, search_start,
2701 num_bytes, empty_size);
2702 if (!offset)
2703 goto loop;
2704 checks:
2705 search_start = stripe_align(root, offset);
2706
2707 /* move on to the next group */
2708 if (search_start + num_bytes >= search_end) {
2709 btrfs_add_free_space(block_group, offset, num_bytes);
2710 goto loop;
2711 }
2712
2713 /* move on to the next group */
2714 if (search_start + num_bytes >
2715 block_group->key.objectid + block_group->key.offset) {
2716 btrfs_add_free_space(block_group, offset, num_bytes);
2717 goto loop;
2718 }
2719
2720 if (exclude_nr > 0 &&
2721 (search_start + num_bytes > exclude_start &&
2722 search_start < exclude_start + exclude_nr)) {
2723 search_start = exclude_start + exclude_nr;
2724
2725 btrfs_add_free_space(block_group, offset, num_bytes);
2726 /*
2727 * if search_start is still in this block group
2728 * then we just re-search this block group
2729 */
2730 if (search_start >= block_group->key.objectid &&
2731 search_start < (block_group->key.objectid +
2732 block_group->key.offset))
2733 goto have_block_group;
2734 goto loop;
2735 }
2736
2737 ins->objectid = search_start;
2738 ins->offset = num_bytes;
2739
2740 if (offset < search_start)
2741 btrfs_add_free_space(block_group, offset,
2742 search_start - offset);
2743 BUG_ON(offset > search_start);
2744
2745 /* we are all good, lets return */
2746 break;
2747 loop:
2748 btrfs_put_block_group(block_group);
2749 }
2750 up_read(&space_info->groups_sem);
2751
2752 /* loop == 0, try to find a clustered alloc in every block group
2753 * loop == 1, try again after forcing a chunk allocation
2754 * loop == 2, set empty_size and empty_cluster to 0 and try again
2755 */
2756 if (!ins->objectid && loop < 3 &&
2757 (empty_size || empty_cluster || allowed_chunk_alloc)) {
2758 if (loop >= 2) {
2759 empty_size = 0;
2760 empty_cluster = 0;
2761 }
2762
2763 if (allowed_chunk_alloc) {
2764 ret = do_chunk_alloc(trans, root, num_bytes +
2765 2 * 1024 * 1024, data, 1);
2766 allowed_chunk_alloc = 0;
2767 } else {
2768 space_info->force_alloc = 1;
2769 }
2770
2771 if (loop < 3) {
2772 loop++;
2773 goto search;
2774 }
2775 ret = -ENOSPC;
2776 } else if (!ins->objectid) {
2777 ret = -ENOSPC;
2778 }
2779
2780 /* we found what we needed */
2781 if (ins->objectid) {
2782 if (!(data & BTRFS_BLOCK_GROUP_DATA))
2783 trans->block_group = block_group->key.objectid;
2784
2785 btrfs_put_block_group(block_group);
2786 ret = 0;
2787 }
2788
2789 return ret;
2790 }
2791
2792 static void dump_space_info(struct btrfs_space_info *info, u64 bytes)
2793 {
2794 struct btrfs_block_group_cache *cache;
2795
2796 printk(KERN_INFO "space_info has %llu free, is %sfull\n",
2797 (unsigned long long)(info->total_bytes - info->bytes_used -
2798 info->bytes_pinned - info->bytes_reserved),
2799 (info->full) ? "" : "not ");
2800 printk(KERN_INFO "space_info total=%llu, pinned=%llu, delalloc=%llu,"
2801 " may_use=%llu, used=%llu\n", info->total_bytes,
2802 info->bytes_pinned, info->bytes_delalloc, info->bytes_may_use,
2803 info->bytes_used);
2804
2805 down_read(&info->groups_sem);
2806 list_for_each_entry(cache, &info->block_groups, list) {
2807 spin_lock(&cache->lock);
2808 printk(KERN_INFO "block group %llu has %llu bytes, %llu used "
2809 "%llu pinned %llu reserved\n",
2810 (unsigned long long)cache->key.objectid,
2811 (unsigned long long)cache->key.offset,
2812 (unsigned long long)btrfs_block_group_used(&cache->item),
2813 (unsigned long long)cache->pinned,
2814 (unsigned long long)cache->reserved);
2815 btrfs_dump_free_space(cache, bytes);
2816 spin_unlock(&cache->lock);
2817 }
2818 up_read(&info->groups_sem);
2819 }
2820
2821 static int __btrfs_reserve_extent(struct btrfs_trans_handle *trans,
2822 struct btrfs_root *root,
2823 u64 num_bytes, u64 min_alloc_size,
2824 u64 empty_size, u64 hint_byte,
2825 u64 search_end, struct btrfs_key *ins,
2826 u64 data)
2827 {
2828 int ret;
2829 u64 search_start = 0;
2830 struct btrfs_fs_info *info = root->fs_info;
2831
2832 data = btrfs_get_alloc_profile(root, data);
2833 again:
2834 /*
2835 * the only place that sets empty_size is btrfs_realloc_node, which
2836 * is not called recursively on allocations
2837 */
2838 if (empty_size || root->ref_cows) {
2839 if (!(data & BTRFS_BLOCK_GROUP_METADATA)) {
2840 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
2841 2 * 1024 * 1024,
2842 BTRFS_BLOCK_GROUP_METADATA |
2843 (info->metadata_alloc_profile &
2844 info->avail_metadata_alloc_bits), 0);
2845 }
2846 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
2847 num_bytes + 2 * 1024 * 1024, data, 0);
2848 }
2849
2850 WARN_ON(num_bytes < root->sectorsize);
2851 ret = find_free_extent(trans, root, num_bytes, empty_size,
2852 search_start, search_end, hint_byte, ins,
2853 trans->alloc_exclude_start,
2854 trans->alloc_exclude_nr, data);
2855
2856 if (ret == -ENOSPC && num_bytes > min_alloc_size) {
2857 num_bytes = num_bytes >> 1;
2858 num_bytes = num_bytes & ~(root->sectorsize - 1);
2859 num_bytes = max(num_bytes, min_alloc_size);
2860 do_chunk_alloc(trans, root->fs_info->extent_root,
2861 num_bytes, data, 1);
2862 goto again;
2863 }
2864 if (ret) {
2865 struct btrfs_space_info *sinfo;
2866
2867 sinfo = __find_space_info(root->fs_info, data);
2868 printk(KERN_ERR "btrfs allocation failed flags %llu, "
2869 "wanted %llu\n", (unsigned long long)data,
2870 (unsigned long long)num_bytes);
2871 dump_space_info(sinfo, num_bytes);
2872 BUG();
2873 }
2874
2875 return ret;
2876 }
2877
2878 int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len)
2879 {
2880 struct btrfs_block_group_cache *cache;
2881 int ret = 0;
2882
2883 cache = btrfs_lookup_block_group(root->fs_info, start);
2884 if (!cache) {
2885 printk(KERN_ERR "Unable to find block group for %llu\n",
2886 (unsigned long long)start);
2887 return -ENOSPC;
2888 }
2889
2890 ret = btrfs_discard_extent(root, start, len);
2891
2892 btrfs_add_free_space(cache, start, len);
2893 btrfs_put_block_group(cache);
2894 update_reserved_extents(root, start, len, 0);
2895
2896 return ret;
2897 }
2898
2899 int btrfs_reserve_extent(struct btrfs_trans_handle *trans,
2900 struct btrfs_root *root,
2901 u64 num_bytes, u64 min_alloc_size,
2902 u64 empty_size, u64 hint_byte,
2903 u64 search_end, struct btrfs_key *ins,
2904 u64 data)
2905 {
2906 int ret;
2907 ret = __btrfs_reserve_extent(trans, root, num_bytes, min_alloc_size,
2908 empty_size, hint_byte, search_end, ins,
2909 data);
2910 update_reserved_extents(root, ins->objectid, ins->offset, 1);
2911 return ret;
2912 }
2913
2914 static int __btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans,
2915 struct btrfs_root *root, u64 parent,
2916 u64 root_objectid, u64 ref_generation,
2917 u64 owner, struct btrfs_key *ins,
2918 int ref_mod)
2919 {
2920 int ret;
2921 u64 super_used;
2922 u64 root_used;
2923 u64 num_bytes = ins->offset;
2924 u32 sizes[2];
2925 struct btrfs_fs_info *info = root->fs_info;
2926 struct btrfs_root *extent_root = info->extent_root;
2927 struct btrfs_extent_item *extent_item;
2928 struct btrfs_extent_ref *ref;
2929 struct btrfs_path *path;
2930 struct btrfs_key keys[2];
2931
2932 if (parent == 0)
2933 parent = ins->objectid;
2934
2935 /* block accounting for super block */
2936 spin_lock(&info->delalloc_lock);
2937 super_used = btrfs_super_bytes_used(&info->super_copy);
2938 btrfs_set_super_bytes_used(&info->super_copy, super_used + num_bytes);
2939
2940 /* block accounting for root item */
2941 root_used = btrfs_root_used(&root->root_item);
2942 btrfs_set_root_used(&root->root_item, root_used + num_bytes);
2943 spin_unlock(&info->delalloc_lock);
2944
2945 memcpy(&keys[0], ins, sizeof(*ins));
2946 keys[1].objectid = ins->objectid;
2947 keys[1].type = BTRFS_EXTENT_REF_KEY;
2948 keys[1].offset = parent;
2949 sizes[0] = sizeof(*extent_item);
2950 sizes[1] = sizeof(*ref);
2951
2952 path = btrfs_alloc_path();
2953 BUG_ON(!path);
2954
2955 path->leave_spinning = 1;
2956 ret = btrfs_insert_empty_items(trans, extent_root, path, keys,
2957 sizes, 2);
2958 BUG_ON(ret);
2959
2960 extent_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
2961 struct btrfs_extent_item);
2962 btrfs_set_extent_refs(path->nodes[0], extent_item, ref_mod);
2963 ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
2964 struct btrfs_extent_ref);
2965
2966 btrfs_set_ref_root(path->nodes[0], ref, root_objectid);
2967 btrfs_set_ref_generation(path->nodes[0], ref, ref_generation);
2968 btrfs_set_ref_objectid(path->nodes[0], ref, owner);
2969 btrfs_set_ref_num_refs(path->nodes[0], ref, ref_mod);
2970
2971 btrfs_mark_buffer_dirty(path->nodes[0]);
2972
2973 trans->alloc_exclude_start = 0;
2974 trans->alloc_exclude_nr = 0;
2975 btrfs_free_path(path);
2976
2977 if (ret)
2978 goto out;
2979
2980 ret = update_block_group(trans, root, ins->objectid,
2981 ins->offset, 1, 0);
2982 if (ret) {
2983 printk(KERN_ERR "btrfs update block group failed for %llu "
2984 "%llu\n", (unsigned long long)ins->objectid,
2985 (unsigned long long)ins->offset);
2986 BUG();
2987 }
2988 out:
2989 return ret;
2990 }
2991
2992 int btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans,
2993 struct btrfs_root *root, u64 parent,
2994 u64 root_objectid, u64 ref_generation,
2995 u64 owner, struct btrfs_key *ins)
2996 {
2997 int ret;
2998
2999 if (root_objectid == BTRFS_TREE_LOG_OBJECTID)
3000 return 0;
3001
3002 ret = btrfs_add_delayed_ref(trans, ins->objectid,
3003 ins->offset, parent, root_objectid,
3004 ref_generation, owner,
3005 BTRFS_ADD_DELAYED_EXTENT, 0);
3006 BUG_ON(ret);
3007 return ret;
3008 }
3009
3010 /*
3011 * this is used by the tree logging recovery code. It records that
3012 * an extent has been allocated and makes sure to clear the free
3013 * space cache bits as well
3014 */
3015 int btrfs_alloc_logged_extent(struct btrfs_trans_handle *trans,
3016 struct btrfs_root *root, u64 parent,
3017 u64 root_objectid, u64 ref_generation,
3018 u64 owner, struct btrfs_key *ins)
3019 {
3020 int ret;
3021 struct btrfs_block_group_cache *block_group;
3022
3023 block_group = btrfs_lookup_block_group(root->fs_info, ins->objectid);
3024 mutex_lock(&block_group->cache_mutex);
3025 cache_block_group(root, block_group);
3026 mutex_unlock(&block_group->cache_mutex);
3027
3028 ret = btrfs_remove_free_space(block_group, ins->objectid,
3029 ins->offset);
3030 BUG_ON(ret);
3031 btrfs_put_block_group(block_group);
3032 ret = __btrfs_alloc_reserved_extent(trans, root, parent, root_objectid,
3033 ref_generation, owner, ins, 1);
3034 return ret;
3035 }
3036
3037 /*
3038 * finds a free extent and does all the dirty work required for allocation
3039 * returns the key for the extent through ins, and a tree buffer for
3040 * the first block of the extent through buf.
3041 *
3042 * returns 0 if everything worked, non-zero otherwise.
3043 */
3044 int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
3045 struct btrfs_root *root,
3046 u64 num_bytes, u64 parent, u64 min_alloc_size,
3047 u64 root_objectid, u64 ref_generation,
3048 u64 owner_objectid, u64 empty_size, u64 hint_byte,
3049 u64 search_end, struct btrfs_key *ins, u64 data)
3050 {
3051 int ret;
3052 ret = __btrfs_reserve_extent(trans, root, num_bytes,
3053 min_alloc_size, empty_size, hint_byte,
3054 search_end, ins, data);
3055 BUG_ON(ret);
3056 if (root_objectid != BTRFS_TREE_LOG_OBJECTID) {
3057 ret = btrfs_add_delayed_ref(trans, ins->objectid,
3058 ins->offset, parent, root_objectid,
3059 ref_generation, owner_objectid,
3060 BTRFS_ADD_DELAYED_EXTENT, 0);
3061 BUG_ON(ret);
3062 }
3063 update_reserved_extents(root, ins->objectid, ins->offset, 1);
3064 return ret;
3065 }
3066
3067 struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans,
3068 struct btrfs_root *root,
3069 u64 bytenr, u32 blocksize,
3070 int level)
3071 {
3072 struct extent_buffer *buf;
3073
3074 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
3075 if (!buf)
3076 return ERR_PTR(-ENOMEM);
3077 btrfs_set_header_generation(buf, trans->transid);
3078 btrfs_set_buffer_lockdep_class(buf, level);
3079 btrfs_tree_lock(buf);
3080 clean_tree_block(trans, root, buf);
3081
3082 btrfs_set_lock_blocking(buf);
3083 btrfs_set_buffer_uptodate(buf);
3084
3085 if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) {
3086 set_extent_dirty(&root->dirty_log_pages, buf->start,
3087 buf->start + buf->len - 1, GFP_NOFS);
3088 } else {
3089 set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
3090 buf->start + buf->len - 1, GFP_NOFS);
3091 }
3092 trans->blocks_used++;
3093 /* this returns a buffer locked for blocking */
3094 return buf;
3095 }
3096
3097 /*
3098 * helper function to allocate a block for a given tree
3099 * returns the tree buffer or NULL.
3100 */
3101 struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
3102 struct btrfs_root *root,
3103 u32 blocksize, u64 parent,
3104 u64 root_objectid,
3105 u64 ref_generation,
3106 int level,
3107 u64 hint,
3108 u64 empty_size)
3109 {
3110 struct btrfs_key ins;
3111 int ret;
3112 struct extent_buffer *buf;
3113
3114 ret = btrfs_alloc_extent(trans, root, blocksize, parent, blocksize,
3115 root_objectid, ref_generation, level,
3116 empty_size, hint, (u64)-1, &ins, 0);
3117 if (ret) {
3118 BUG_ON(ret > 0);
3119 return ERR_PTR(ret);
3120 }
3121
3122 buf = btrfs_init_new_buffer(trans, root, ins.objectid,
3123 blocksize, level);
3124 return buf;
3125 }
3126
3127 int btrfs_drop_leaf_ref(struct btrfs_trans_handle *trans,
3128 struct btrfs_root *root, struct extent_buffer *leaf)
3129 {
3130 u64 leaf_owner;
3131 u64 leaf_generation;
3132 struct refsort *sorted;
3133 struct btrfs_key key;
3134 struct btrfs_file_extent_item *fi;
3135 int i;
3136 int nritems;
3137 int ret;
3138 int refi = 0;
3139 int slot;
3140
3141 BUG_ON(!btrfs_is_leaf(leaf));
3142 nritems = btrfs_header_nritems(leaf);
3143 leaf_owner = btrfs_header_owner(leaf);
3144 leaf_generation = btrfs_header_generation(leaf);
3145
3146 sorted = kmalloc(sizeof(*sorted) * nritems, GFP_NOFS);
3147 /* we do this loop twice. The first time we build a list
3148 * of the extents we have a reference on, then we sort the list
3149 * by bytenr. The second time around we actually do the
3150 * extent freeing.
3151 */
3152 for (i = 0; i < nritems; i++) {
3153 u64 disk_bytenr;
3154 cond_resched();
3155
3156 btrfs_item_key_to_cpu(leaf, &key, i);
3157
3158 /* only extents have references, skip everything else */
3159 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
3160 continue;
3161
3162 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
3163
3164 /* inline extents live in the btree, they don't have refs */
3165 if (btrfs_file_extent_type(leaf, fi) ==
3166 BTRFS_FILE_EXTENT_INLINE)
3167 continue;
3168
3169 disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
3170
3171 /* holes don't have refs */
3172 if (disk_bytenr == 0)
3173 continue;
3174
3175 sorted[refi].bytenr = disk_bytenr;
3176 sorted[refi].slot = i;
3177 refi++;
3178 }
3179
3180 if (refi == 0)
3181 goto out;
3182
3183 sort(sorted, refi, sizeof(struct refsort), refsort_cmp, NULL);
3184
3185 for (i = 0; i < refi; i++) {
3186 u64 disk_bytenr;
3187
3188 disk_bytenr = sorted[i].bytenr;
3189 slot = sorted[i].slot;
3190
3191 cond_resched();
3192
3193 btrfs_item_key_to_cpu(leaf, &key, slot);
3194 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
3195 continue;
3196
3197 fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
3198
3199 ret = btrfs_free_extent(trans, root, disk_bytenr,
3200 btrfs_file_extent_disk_num_bytes(leaf, fi),
3201 leaf->start, leaf_owner, leaf_generation,
3202 key.objectid, 0);
3203 BUG_ON(ret);
3204
3205 atomic_inc(&root->fs_info->throttle_gen);
3206 wake_up(&root->fs_info->transaction_throttle);
3207 cond_resched();
3208 }
3209 out:
3210 kfree(sorted);
3211 return 0;
3212 }
3213
3214 static noinline int cache_drop_leaf_ref(struct btrfs_trans_handle *trans,
3215 struct btrfs_root *root,
3216 struct btrfs_leaf_ref *ref)
3217 {
3218 int i;
3219 int ret;
3220 struct btrfs_extent_info *info;
3221 struct refsort *sorted;
3222
3223 if (ref->nritems == 0)
3224 return 0;
3225
3226 sorted = kmalloc(sizeof(*sorted) * ref->nritems, GFP_NOFS);
3227 for (i = 0; i < ref->nritems; i++) {
3228 sorted[i].bytenr = ref->extents[i].bytenr;
3229 sorted[i].slot = i;
3230 }
3231 sort(sorted, ref->nritems, sizeof(struct refsort), refsort_cmp, NULL);
3232
3233 /*
3234 * the items in the ref were sorted when the ref was inserted
3235 * into the ref cache, so this is already in order
3236 */
3237 for (i = 0; i < ref->nritems; i++) {
3238 info = ref->extents + sorted[i].slot;
3239 ret = btrfs_free_extent(trans, root, info->bytenr,
3240 info->num_bytes, ref->bytenr,
3241 ref->owner, ref->generation,
3242 info->objectid, 0);
3243
3244 atomic_inc(&root->fs_info->throttle_gen);
3245 wake_up(&root->fs_info->transaction_throttle);
3246 cond_resched();
3247
3248 BUG_ON(ret);
3249 info++;
3250 }
3251
3252 kfree(sorted);
3253 return 0;
3254 }
3255
3256 static int drop_snap_lookup_refcount(struct btrfs_trans_handle *trans,
3257 struct btrfs_root *root, u64 start,
3258 u64 len, u32 *refs)
3259 {
3260 int ret;
3261
3262 ret = btrfs_lookup_extent_ref(trans, root, start, len, refs);
3263 BUG_ON(ret);
3264
3265 #if 0 /* some debugging code in case we see problems here */
3266 /* if the refs count is one, it won't get increased again. But
3267 * if the ref count is > 1, someone may be decreasing it at
3268 * the same time we are.
3269 */
3270 if (*refs != 1) {
3271 struct extent_buffer *eb = NULL;
3272 eb = btrfs_find_create_tree_block(root, start, len);
3273 if (eb)
3274 btrfs_tree_lock(eb);
3275
3276 mutex_lock(&root->fs_info->alloc_mutex);
3277 ret = lookup_extent_ref(NULL, root, start, len, refs);
3278 BUG_ON(ret);
3279 mutex_unlock(&root->fs_info->alloc_mutex);
3280
3281 if (eb) {
3282 btrfs_tree_unlock(eb);
3283 free_extent_buffer(eb);
3284 }
3285 if (*refs == 1) {
3286 printk(KERN_ERR "btrfs block %llu went down to one "
3287 "during drop_snap\n", (unsigned long long)start);
3288 }
3289
3290 }
3291 #endif
3292
3293 cond_resched();
3294 return ret;
3295 }
3296
3297 /*
3298 * this is used while deleting old snapshots, and it drops the refs
3299 * on a whole subtree starting from a level 1 node.
3300 *
3301 * The idea is to sort all the leaf pointers, and then drop the
3302 * ref on all the leaves in order. Most of the time the leaves
3303 * will have ref cache entries, so no leaf IOs will be required to
3304 * find the extents they have references on.
3305 *
3306 * For each leaf, any references it has are also dropped in order
3307 *
3308 * This ends up dropping the references in something close to optimal
3309 * order for reading and modifying the extent allocation tree.
3310 */
3311 static noinline int drop_level_one_refs(struct btrfs_trans_handle *trans,
3312 struct btrfs_root *root,
3313 struct btrfs_path *path)
3314 {
3315 u64 bytenr;
3316 u64 root_owner;
3317 u64 root_gen;
3318 struct extent_buffer *eb = path->nodes[1];
3319 struct extent_buffer *leaf;
3320 struct btrfs_leaf_ref *ref;
3321 struct refsort *sorted = NULL;
3322 int nritems = btrfs_header_nritems(eb);
3323 int ret;
3324 int i;
3325 int refi = 0;
3326 int slot = path->slots[1];
3327 u32 blocksize = btrfs_level_size(root, 0);
3328 u32 refs;
3329
3330 if (nritems == 0)
3331 goto out;
3332
3333 root_owner = btrfs_header_owner(eb);
3334 root_gen = btrfs_header_generation(eb);
3335 sorted = kmalloc(sizeof(*sorted) * nritems, GFP_NOFS);
3336
3337 /*
3338 * step one, sort all the leaf pointers so we don't scribble
3339 * randomly into the extent allocation tree
3340 */
3341 for (i = slot; i < nritems; i++) {
3342 sorted[refi].bytenr = btrfs_node_blockptr(eb, i);
3343 sorted[refi].slot = i;
3344 refi++;
3345 }
3346
3347 /*
3348 * nritems won't be zero, but if we're picking up drop_snapshot
3349 * after a crash, slot might be > 0, so double check things
3350 * just in case.
3351 */
3352 if (refi == 0)
3353 goto out;
3354
3355 sort(sorted, refi, sizeof(struct refsort), refsort_cmp, NULL);
3356
3357 /*
3358 * the first loop frees everything the leaves point to
3359 */
3360 for (i = 0; i < refi; i++) {
3361 u64 ptr_gen;
3362
3363 bytenr = sorted[i].bytenr;
3364
3365 /*
3366 * check the reference count on this leaf. If it is > 1
3367 * we just decrement it below and don't update any
3368 * of the refs the leaf points to.
3369 */
3370 ret = drop_snap_lookup_refcount(trans, root, bytenr,
3371 blocksize, &refs);
3372 BUG_ON(ret);
3373 if (refs != 1)
3374 continue;
3375
3376 ptr_gen = btrfs_node_ptr_generation(eb, sorted[i].slot);
3377
3378 /*
3379 * the leaf only had one reference, which means the
3380 * only thing pointing to this leaf is the snapshot
3381 * we're deleting. It isn't possible for the reference
3382 * count to increase again later
3383 *
3384 * The reference cache is checked for the leaf,
3385 * and if found we'll be able to drop any refs held by
3386 * the leaf without needing to read it in.
3387 */
3388 ref = btrfs_lookup_leaf_ref(root, bytenr);
3389 if (ref && ref->generation != ptr_gen) {
3390 btrfs_free_leaf_ref(root, ref);
3391 ref = NULL;
3392 }
3393 if (ref) {
3394 ret = cache_drop_leaf_ref(trans, root, ref);
3395 BUG_ON(ret);
3396 btrfs_remove_leaf_ref(root, ref);
3397 btrfs_free_leaf_ref(root, ref);
3398 } else {
3399 /*
3400 * the leaf wasn't in the reference cache, so
3401 * we have to read it.
3402 */
3403 leaf = read_tree_block(root, bytenr, blocksize,
3404 ptr_gen);
3405 ret = btrfs_drop_leaf_ref(trans, root, leaf);
3406 BUG_ON(ret);
3407 free_extent_buffer(leaf);
3408 }
3409 atomic_inc(&root->fs_info->throttle_gen);
3410 wake_up(&root->fs_info->transaction_throttle);
3411 cond_resched();
3412 }
3413
3414 /*
3415 * run through the loop again to free the refs on the leaves.
3416 * This is faster than doing it in the loop above because
3417 * the leaves are likely to be clustered together. We end up
3418 * working in nice chunks on the extent allocation tree.
3419 */
3420 for (i = 0; i < refi; i++) {
3421 bytenr = sorted[i].bytenr;
3422 ret = btrfs_free_extent(trans, root, bytenr,
3423 blocksize, eb->start,
3424 root_owner, root_gen, 0, 1);
3425 BUG_ON(ret);
3426
3427 atomic_inc(&root->fs_info->throttle_gen);
3428 wake_up(&root->fs_info->transaction_throttle);
3429 cond_resched();
3430 }
3431 out:
3432 kfree(sorted);
3433
3434 /*
3435 * update the path to show we've processed the entire level 1
3436 * node. This will get saved into the root's drop_snapshot_progress
3437 * field so these drops are not repeated again if this transaction
3438 * commits.
3439 */
3440 path->slots[1] = nritems;
3441 return 0;
3442 }
3443
3444 /*
3445 * helper function for drop_snapshot, this walks down the tree dropping ref
3446 * counts as it goes.
3447 */
3448 static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
3449 struct btrfs_root *root,
3450 struct btrfs_path *path, int *level)
3451 {
3452 u64 root_owner;
3453 u64 root_gen;
3454 u64 bytenr;
3455 u64 ptr_gen;
3456 struct extent_buffer *next;
3457 struct extent_buffer *cur;
3458 struct extent_buffer *parent;
3459 u32 blocksize;
3460 int ret;
3461 u32 refs;
3462
3463 WARN_ON(*level < 0);
3464 WARN_ON(*level >= BTRFS_MAX_LEVEL);
3465 ret = drop_snap_lookup_refcount(trans, root, path->nodes[*level]->start,
3466 path->nodes[*level]->len, &refs);
3467 BUG_ON(ret);
3468 if (refs > 1)
3469 goto out;
3470
3471 /*
3472 * walk down to the last node level and free all the leaves
3473 */
3474 while (*level >= 0) {
3475 WARN_ON(*level < 0);
3476 WARN_ON(*level >= BTRFS_MAX_LEVEL);
3477 cur = path->nodes[*level];
3478
3479 if (btrfs_header_level(cur) != *level)
3480 WARN_ON(1);
3481
3482 if (path->slots[*level] >=
3483 btrfs_header_nritems(cur))
3484 break;
3485
3486 /* the new code goes down to level 1 and does all the
3487 * leaves pointed to that node in bulk. So, this check
3488 * for level 0 will always be false.
3489 *
3490 * But, the disk format allows the drop_snapshot_progress
3491 * field in the root to leave things in a state where
3492 * a leaf will need cleaning up here. If someone crashes
3493 * with the old code and then boots with the new code,
3494 * we might find a leaf here.
3495 */
3496 if (*level == 0) {
3497 ret = btrfs_drop_leaf_ref(trans, root, cur);
3498 BUG_ON(ret);
3499 break;
3500 }
3501
3502 /*
3503 * once we get to level one, process the whole node
3504 * at once, including everything below it.
3505 */
3506 if (*level == 1) {
3507 ret = drop_level_one_refs(trans, root, path);
3508 BUG_ON(ret);
3509 break;
3510 }
3511
3512 bytenr = btrfs_node_blockptr(cur, path->slots[*level]);
3513 ptr_gen = btrfs_node_ptr_generation(cur, path->slots[*level]);
3514 blocksize = btrfs_level_size(root, *level - 1);
3515
3516 ret = drop_snap_lookup_refcount(trans, root, bytenr,
3517 blocksize, &refs);
3518 BUG_ON(ret);
3519
3520 /*
3521 * if there is more than one reference, we don't need
3522 * to read that node to drop any references it has. We
3523 * just drop the ref we hold on that node and move on to the
3524 * next slot in this level.
3525 */
3526 if (refs != 1) {
3527 parent = path->nodes[*level];
3528 root_owner = btrfs_header_owner(parent);
3529 root_gen = btrfs_header_generation(parent);
3530 path->slots[*level]++;
3531
3532 ret = btrfs_free_extent(trans, root, bytenr,
3533 blocksize, parent->start,
3534 root_owner, root_gen,
3535 *level - 1, 1);
3536 BUG_ON(ret);
3537
3538 atomic_inc(&root->fs_info->throttle_gen);
3539 wake_up(&root->fs_info->transaction_throttle);
3540 cond_resched();
3541
3542 continue;
3543 }
3544
3545 /*
3546 * we need to keep freeing things in the next level down.
3547 * read the block and loop around to process it
3548 */
3549 next = read_tree_block(root, bytenr, blocksize, ptr_gen);
3550 WARN_ON(*level <= 0);
3551 if (path->nodes[*level-1])
3552 free_extent_buffer(path->nodes[*level-1]);
3553 path->nodes[*level-1] = next;
3554 *level = btrfs_header_level(next);
3555 path->slots[*level] = 0;
3556 cond_resched();
3557 }
3558 out:
3559 WARN_ON(*level < 0);
3560 WARN_ON(*level >= BTRFS_MAX_LEVEL);
3561
3562 if (path->nodes[*level] == root->node) {
3563 parent = path->nodes[*level];
3564 bytenr = path->nodes[*level]->start;
3565 } else {
3566 parent = path->nodes[*level + 1];
3567 bytenr = btrfs_node_blockptr(parent, path->slots[*level + 1]);
3568 }
3569
3570 blocksize = btrfs_level_size(root, *level);
3571 root_owner = btrfs_header_owner(parent);
3572 root_gen = btrfs_header_generation(parent);
3573
3574 /*
3575 * cleanup and free the reference on the last node
3576 * we processed
3577 */
3578 ret = btrfs_free_extent(trans, root, bytenr, blocksize,
3579 parent->start, root_owner, root_gen,
3580 *level, 1);
3581 free_extent_buffer(path->nodes[*level]);
3582 path->nodes[*level] = NULL;
3583
3584 *level += 1;
3585 BUG_ON(ret);
3586
3587 cond_resched();
3588 return 0;
3589 }
3590
3591 /*
3592 * helper function for drop_subtree, this function is similar to
3593 * walk_down_tree. The main difference is that it checks reference
3594 * counts while tree blocks are locked.
3595 */
3596 static noinline int walk_down_subtree(struct btrfs_trans_handle *trans,
3597 struct btrfs_root *root,
3598 struct btrfs_path *path, int *level)
3599 {
3600 struct extent_buffer *next;
3601 struct extent_buffer *cur;
3602 struct extent_buffer *parent;
3603 u64 bytenr;
3604 u64 ptr_gen;
3605 u32 blocksize;
3606 u32 refs;
3607 int ret;
3608
3609 cur = path->nodes[*level];
3610 ret = btrfs_lookup_extent_ref(trans, root, cur->start, cur->len,
3611 &refs);
3612 BUG_ON(ret);
3613 if (refs > 1)
3614 goto out;
3615
3616 while (*level >= 0) {
3617 cur = path->nodes[*level];
3618 if (*level == 0) {
3619 ret = btrfs_drop_leaf_ref(trans, root, cur);
3620 BUG_ON(ret);
3621 clean_tree_block(trans, root, cur);
3622 break;
3623 }
3624 if (path->slots[*level] >= btrfs_header_nritems(cur)) {
3625 clean_tree_block(trans, root, cur);
3626 break;
3627 }
3628
3629 bytenr = btrfs_node_blockptr(cur, path->slots[*level]);
3630 blocksize = btrfs_level_size(root, *level - 1);
3631 ptr_gen = btrfs_node_ptr_generation(cur, path->slots[*level]);
3632
3633 next = read_tree_block(root, bytenr, blocksize, ptr_gen);
3634 btrfs_tree_lock(next);
3635 btrfs_set_lock_blocking(next);
3636
3637 ret = btrfs_lookup_extent_ref(trans, root, bytenr, blocksize,
3638 &refs);
3639 BUG_ON(ret);
3640 if (refs > 1) {
3641 parent = path->nodes[*level];
3642 ret = btrfs_free_extent(trans, root, bytenr,
3643 blocksize, parent->start,
3644 btrfs_header_owner(parent),
3645 btrfs_header_generation(parent),
3646 *level - 1, 1);
3647 BUG_ON(ret);
3648 path->slots[*level]++;
3649 btrfs_tree_unlock(next);
3650 free_extent_buffer(next);
3651 continue;
3652 }
3653
3654 *level = btrfs_header_level(next);
3655 path->nodes[*level] = next;
3656 path->slots[*level] = 0;
3657 path->locks[*level] = 1;
3658 cond_resched();
3659 }
3660 out:
3661 parent = path->nodes[*level + 1];
3662 bytenr = path->nodes[*level]->start;
3663 blocksize = path->nodes[*level]->len;
3664
3665 ret = btrfs_free_extent(trans, root, bytenr, blocksize,
3666 parent->start, btrfs_header_owner(parent),
3667 btrfs_header_generation(parent), *level, 1);
3668 BUG_ON(ret);
3669
3670 if (path->locks[*level]) {
3671 btrfs_tree_unlock(path->nodes[*level]);
3672 path->locks[*level] = 0;
3673 }
3674 free_extent_buffer(path->nodes[*level]);
3675 path->nodes[*level] = NULL;
3676 *level += 1;
3677 cond_resched();
3678 return 0;
3679 }
3680
3681 /*
3682 * helper for dropping snapshots. This walks back up the tree in the path
3683 * to find the first node higher up where we haven't yet gone through
3684 * all the slots
3685 */
3686 static noinline int walk_up_tree(struct btrfs_trans_handle *trans,
3687 struct btrfs_root *root,
3688 struct btrfs_path *path,
3689 int *level, int max_level)
3690 {
3691 u64 root_owner;
3692 u64 root_gen;
3693 struct btrfs_root_item *root_item = &root->root_item;
3694 int i;
3695 int slot;
3696 int ret;
3697
3698 for (i = *level; i < max_level && path->nodes[i]; i++) {
3699 slot = path->slots[i];
3700 if (slot < btrfs_header_nritems(path->nodes[i]) - 1) {
3701 struct extent_buffer *node;
3702 struct btrfs_disk_key disk_key;
3703
3704 /*
3705 * there is more work to do in this level.
3706 * Update the drop_progress marker to reflect
3707 * the work we've done so far, and then bump
3708 * the slot number
3709 */
3710 node = path->nodes[i];
3711 path->slots[i]++;
3712 *level = i;
3713 WARN_ON(*level == 0);
3714 btrfs_node_key(node, &disk_key, path->slots[i]);
3715 memcpy(&root_item->drop_progress,
3716 &disk_key, sizeof(disk_key));
3717 root_item->drop_level = i;
3718 return 0;
3719 } else {
3720 struct extent_buffer *parent;
3721
3722 /*
3723 * this whole node is done, free our reference
3724 * on it and go up one level
3725 */
3726 if (path->nodes[*level] == root->node)
3727 parent = path->nodes[*level];
3728 else
3729 parent = path->nodes[*level + 1];
3730
3731 root_owner = btrfs_header_owner(parent);
3732 root_gen = btrfs_header_generation(parent);
3733
3734 clean_tree_block(trans, root, path->nodes[*level]);
3735 ret = btrfs_free_extent(trans, root,
3736 path->nodes[*level]->start,
3737 path->nodes[*level]->len,
3738 parent->start, root_owner,
3739 root_gen, *level, 1);
3740 BUG_ON(ret);
3741 if (path->locks[*level]) {
3742 btrfs_tree_unlock(path->nodes[*level]);
3743 path->locks[*level] = 0;
3744 }
3745 free_extent_buffer(path->nodes[*level]);
3746 path->nodes[*level] = NULL;
3747 *level = i + 1;
3748 }
3749 }
3750 return 1;
3751 }
3752
3753 /*
3754 * drop the reference count on the tree rooted at 'snap'. This traverses
3755 * the tree freeing any blocks that have a ref count of zero after being
3756 * decremented.
3757 */
3758 int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
3759 *root)
3760 {
3761 int ret = 0;
3762 int wret;
3763 int level;
3764 struct btrfs_path *path;
3765 int i;
3766 int orig_level;
3767 int update_count;
3768 struct btrfs_root_item *root_item = &root->root_item;
3769
3770 WARN_ON(!mutex_is_locked(&root->fs_info->drop_mutex));
3771 path = btrfs_alloc_path();
3772 BUG_ON(!path);
3773
3774 level = btrfs_header_level(root->node);
3775 orig_level = level;
3776 if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
3777 path->nodes[level] = root->node;
3778 extent_buffer_get(root->node);
3779 path->slots[level] = 0;
3780 } else {
3781 struct btrfs_key key;
3782 struct btrfs_disk_key found_key;
3783 struct extent_buffer *node;
3784
3785 btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
3786 level = root_item->drop_level;
3787 path->lowest_level = level;
3788 wret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3789 if (wret < 0) {
3790 ret = wret;
3791 goto out;
3792 }
3793 node = path->nodes[level];
3794 btrfs_node_key(node, &found_key, path->slots[level]);
3795 WARN_ON(memcmp(&found_key, &root_item->drop_progress,
3796 sizeof(found_key)));
3797 /*
3798 * unlock our path, this is safe because only this
3799 * function is allowed to delete this snapshot
3800 */
3801 for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
3802 if (path->nodes[i] && path->locks[i]) {
3803 path->locks[i] = 0;
3804 btrfs_tree_unlock(path->nodes[i]);
3805 }
3806 }
3807 }
3808 while (1) {
3809 unsigned long update;
3810 wret = walk_down_tree(trans, root, path, &level);
3811 if (wret > 0)
3812 break;
3813 if (wret < 0)
3814 ret = wret;
3815
3816 wret = walk_up_tree(trans, root, path, &level,
3817 BTRFS_MAX_LEVEL);
3818 if (wret > 0)
3819 break;
3820 if (wret < 0)
3821 ret = wret;
3822 if (trans->transaction->in_commit ||
3823 trans->transaction->delayed_refs.flushing) {
3824 ret = -EAGAIN;
3825 break;
3826 }
3827 atomic_inc(&root->fs_info->throttle_gen);
3828 wake_up(&root->fs_info->transaction_throttle);
3829 for (update_count = 0; update_count < 16; update_count++) {
3830 update = trans->delayed_ref_updates;
3831 trans->delayed_ref_updates = 0;
3832 if (update)
3833 btrfs_run_delayed_refs(trans, root, update);
3834 else
3835 break;
3836 }
3837 }
3838 for (i = 0; i <= orig_level; i++) {
3839 if (path->nodes[i]) {
3840 free_extent_buffer(path->nodes[i]);
3841 path->nodes[i] = NULL;
3842 }
3843 }
3844 out:
3845 btrfs_free_path(path);
3846 return ret;
3847 }
3848
3849 int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
3850 struct btrfs_root *root,
3851 struct extent_buffer *node,
3852 struct extent_buffer *parent)
3853 {
3854 struct btrfs_path *path;
3855 int level;
3856 int parent_level;
3857 int ret = 0;
3858 int wret;
3859
3860 path = btrfs_alloc_path();
3861 BUG_ON(!path);
3862
3863 btrfs_assert_tree_locked(parent);
3864 parent_level = btrfs_header_level(parent);
3865 extent_buffer_get(parent);
3866 path->nodes[parent_level] = parent;
3867 path->slots[parent_level] = btrfs_header_nritems(parent);
3868
3869 btrfs_assert_tree_locked(node);
3870 level = btrfs_header_level(node);
3871 extent_buffer_get(node);
3872 path->nodes[level] = node;
3873 path->slots[level] = 0;
3874
3875 while (1) {
3876 wret = walk_down_subtree(trans, root, path, &level);
3877 if (wret < 0)
3878 ret = wret;
3879 if (wret != 0)
3880 break;
3881
3882 wret = walk_up_tree(trans, root, path, &level, parent_level);
3883 if (wret < 0)
3884 ret = wret;
3885 if (wret != 0)
3886 break;
3887 }
3888
3889 btrfs_free_path(path);
3890 return ret;
3891 }
3892
3893 static unsigned long calc_ra(unsigned long start, unsigned long last,
3894 unsigned long nr)
3895 {
3896 return min(last, start + nr - 1);
3897 }
3898
3899 static noinline int relocate_inode_pages(struct inode *inode, u64 start,
3900 u64 len)
3901 {
3902 u64 page_start;
3903 u64 page_end;
3904 unsigned long first_index;
3905 unsigned long last_index;
3906 unsigned long i;
3907 struct page *page;
3908 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
3909 struct file_ra_state *ra;
3910 struct btrfs_ordered_extent *ordered;
3911 unsigned int total_read = 0;
3912 unsigned int total_dirty = 0;
3913 int ret = 0;
3914
3915 ra = kzalloc(sizeof(*ra), GFP_NOFS);
3916
3917 mutex_lock(&inode->i_mutex);
3918 first_index = start >> PAGE_CACHE_SHIFT;
3919 last_index = (start + len - 1) >> PAGE_CACHE_SHIFT;
3920
3921 /* make sure the dirty trick played by the caller work */
3922 ret = invalidate_inode_pages2_range(inode->i_mapping,
3923 first_index, last_index);
3924 if (ret)
3925 goto out_unlock;
3926
3927 file_ra_state_init(ra, inode->i_mapping);
3928
3929 for (i = first_index ; i <= last_index; i++) {
3930 if (total_read % ra->ra_pages == 0) {
3931 btrfs_force_ra(inode->i_mapping, ra, NULL, i,
3932 calc_ra(i, last_index, ra->ra_pages));
3933 }
3934 total_read++;
3935 again:
3936 if (((u64)i << PAGE_CACHE_SHIFT) > i_size_read(inode))
3937 BUG_ON(1);
3938 page = grab_cache_page(inode->i_mapping, i);
3939 if (!page) {
3940 ret = -ENOMEM;
3941 goto out_unlock;
3942 }
3943 if (!PageUptodate(page)) {
3944 btrfs_readpage(NULL, page);
3945 lock_page(page);
3946 if (!PageUptodate(page)) {
3947 unlock_page(page);
3948 page_cache_release(page);
3949 ret = -EIO;
3950 goto out_unlock;
3951 }
3952 }
3953 wait_on_page_writeback(page);
3954
3955 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
3956 page_end = page_start + PAGE_CACHE_SIZE - 1;
3957 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
3958
3959 ordered = btrfs_lookup_ordered_extent(inode, page_start);
3960 if (ordered) {
3961 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
3962 unlock_page(page);
3963 page_cache_release(page);
3964 btrfs_start_ordered_extent(inode, ordered, 1);
3965 btrfs_put_ordered_extent(ordered);
3966 goto again;
3967 }
3968 set_page_extent_mapped(page);
3969
3970 if (i == first_index)
3971 set_extent_bits(io_tree, page_start, page_end,
3972 EXTENT_BOUNDARY, GFP_NOFS);
3973 btrfs_set_extent_delalloc(inode, page_start, page_end);
3974
3975 set_page_dirty(page);
3976 total_dirty++;
3977
3978 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
3979 unlock_page(page);
3980 page_cache_release(page);
3981 }
3982
3983 out_unlock:
3984 kfree(ra);
3985 mutex_unlock(&inode->i_mutex);
3986 balance_dirty_pages_ratelimited_nr(inode->i_mapping, total_dirty);
3987 return ret;
3988 }
3989
3990 static noinline int relocate_data_extent(struct inode *reloc_inode,
3991 struct btrfs_key *extent_key,
3992 u64 offset)
3993 {
3994 struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
3995 struct extent_map_tree *em_tree = &BTRFS_I(reloc_inode)->extent_tree;
3996 struct extent_map *em;
3997 u64 start = extent_key->objectid - offset;
3998 u64 end = start + extent_key->offset - 1;
3999
4000 em = alloc_extent_map(GFP_NOFS);
4001 BUG_ON(!em || IS_ERR(em));
4002
4003 em->start = start;
4004 em->len = extent_key->offset;
4005 em->block_len = extent_key->offset;
4006 em->block_start = extent_key->objectid;
4007 em->bdev = root->fs_info->fs_devices->latest_bdev;
4008 set_bit(EXTENT_FLAG_PINNED, &em->flags);
4009
4010 /* setup extent map to cheat btrfs_readpage */
4011 lock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS);
4012 while (1) {
4013 int ret;
4014 spin_lock(&em_tree->lock);
4015 ret = add_extent_mapping(em_tree, em);
4016 spin_unlock(&em_tree->lock);
4017 if (ret != -EEXIST) {
4018 free_extent_map(em);
4019 break;
4020 }
4021 btrfs_drop_extent_cache(reloc_inode, start, end, 0);
4022 }
4023 unlock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS);
4024
4025 return relocate_inode_pages(reloc_inode, start, extent_key->offset);
4026 }
4027
4028 struct btrfs_ref_path {
4029 u64 extent_start;
4030 u64 nodes[BTRFS_MAX_LEVEL];
4031 u64 root_objectid;
4032 u64 root_generation;
4033 u64 owner_objectid;
4034 u32 num_refs;
4035 int lowest_level;
4036 int current_level;
4037 int shared_level;
4038
4039 struct btrfs_key node_keys[BTRFS_MAX_LEVEL];
4040 u64 new_nodes[BTRFS_MAX_LEVEL];
4041 };
4042
4043 struct disk_extent {
4044 u64 ram_bytes;
4045 u64 disk_bytenr;
4046 u64 disk_num_bytes;
4047 u64 offset;
4048 u64 num_bytes;
4049 u8 compression;
4050 u8 encryption;
4051 u16 other_encoding;
4052 };
4053
4054 static int is_cowonly_root(u64 root_objectid)
4055 {
4056 if (root_objectid == BTRFS_ROOT_TREE_OBJECTID ||
4057 root_objectid == BTRFS_EXTENT_TREE_OBJECTID ||
4058 root_objectid == BTRFS_CHUNK_TREE_OBJECTID ||
4059 root_objectid == BTRFS_DEV_TREE_OBJECTID ||
4060 root_objectid == BTRFS_TREE_LOG_OBJECTID ||
4061 root_objectid == BTRFS_CSUM_TREE_OBJECTID)
4062 return 1;
4063 return 0;
4064 }
4065
4066 static noinline int __next_ref_path(struct btrfs_trans_handle *trans,
4067 struct btrfs_root *extent_root,
4068 struct btrfs_ref_path *ref_path,
4069 int first_time)
4070 {
4071 struct extent_buffer *leaf;
4072 struct btrfs_path *path;
4073 struct btrfs_extent_ref *ref;
4074 struct btrfs_key key;
4075 struct btrfs_key found_key;
4076 u64 bytenr;
4077 u32 nritems;
4078 int level;
4079 int ret = 1;
4080
4081 path = btrfs_alloc_path();
4082 if (!path)
4083 return -ENOMEM;
4084
4085 if (first_time) {
4086 ref_path->lowest_level = -1;
4087 ref_path->current_level = -1;
4088 ref_path->shared_level = -1;
4089 goto walk_up;
4090 }
4091 walk_down:
4092 level = ref_path->current_level - 1;
4093 while (level >= -1) {
4094 u64 parent;
4095 if (level < ref_path->lowest_level)
4096 break;
4097
4098 if (level >= 0)
4099 bytenr = ref_path->nodes[level];
4100 else
4101 bytenr = ref_path->extent_start;
4102 BUG_ON(bytenr == 0);
4103
4104 parent = ref_path->nodes[level + 1];
4105 ref_path->nodes[level + 1] = 0;
4106 ref_path->current_level = level;
4107 BUG_ON(parent == 0);
4108
4109 key.objectid = bytenr;
4110 key.offset = parent + 1;
4111 key.type = BTRFS_EXTENT_REF_KEY;
4112
4113 ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 0);
4114 if (ret < 0)
4115 goto out;
4116 BUG_ON(ret == 0);
4117
4118 leaf = path->nodes[0];
4119 nritems = btrfs_header_nritems(leaf);
4120 if (path->slots[0] >= nritems) {
4121 ret = btrfs_next_leaf(extent_root, path);
4122 if (ret < 0)
4123 goto out;
4124 if (ret > 0)
4125 goto next;
4126 leaf = path->nodes[0];
4127 }
4128
4129 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
4130 if (found_key.objectid == bytenr &&
4131 found_key.type == BTRFS_EXTENT_REF_KEY) {
4132 if (level < ref_path->shared_level)
4133 ref_path->shared_level = level;
4134 goto found;
4135 }
4136 next:
4137 level--;
4138 btrfs_release_path(extent_root, path);
4139 cond_resched();
4140 }
4141 /* reached lowest level */
4142 ret = 1;
4143 goto out;
4144 walk_up:
4145 level = ref_path->current_level;
4146 while (level < BTRFS_MAX_LEVEL - 1) {
4147 u64 ref_objectid;
4148
4149 if (level >= 0)
4150 bytenr = ref_path->nodes[level];
4151 else
4152 bytenr = ref_path->extent_start;
4153
4154 BUG_ON(bytenr == 0);
4155
4156 key.objectid = bytenr;
4157 key.offset = 0;
4158 key.type = BTRFS_EXTENT_REF_KEY;
4159
4160 ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 0);
4161 if (ret < 0)
4162 goto out;
4163
4164 leaf = path->nodes[0];
4165 nritems = btrfs_header_nritems(leaf);
4166 if (path->slots[0] >= nritems) {
4167 ret = btrfs_next_leaf(extent_root, path);
4168 if (ret < 0)
4169 goto out;
4170 if (ret > 0) {
4171 /* the extent was freed by someone */
4172 if (ref_path->lowest_level == level)
4173 goto out;
4174 btrfs_release_path(extent_root, path);
4175 goto walk_down;
4176 }
4177 leaf = path->nodes[0];
4178 }
4179
4180 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
4181 if (found_key.objectid != bytenr ||
4182 found_key.type != BTRFS_EXTENT_REF_KEY) {
4183 /* the extent was freed by someone */
4184 if (ref_path->lowest_level == level) {
4185 ret = 1;
4186 goto out;
4187 }
4188 btrfs_release_path(extent_root, path);
4189 goto walk_down;
4190 }
4191 found:
4192 ref = btrfs_item_ptr(leaf, path->slots[0],
4193 struct btrfs_extent_ref);
4194 ref_objectid = btrfs_ref_objectid(leaf, ref);
4195 if (ref_objectid < BTRFS_FIRST_FREE_OBJECTID) {
4196 if (first_time) {
4197 level = (int)ref_objectid;
4198 BUG_ON(level >= BTRFS_MAX_LEVEL);
4199 ref_path->lowest_level = level;
4200 ref_path->current_level = level;
4201 ref_path->nodes[level] = bytenr;
4202 } else {
4203 WARN_ON(ref_objectid != level);
4204 }
4205 } else {
4206 WARN_ON(level != -1);
4207 }
4208 first_time = 0;
4209
4210 if (ref_path->lowest_level == level) {
4211 ref_path->owner_objectid = ref_objectid;
4212 ref_path->num_refs = btrfs_ref_num_refs(leaf, ref);
4213 }
4214
4215 /*
4216 * the block is tree root or the block isn't in reference
4217 * counted tree.
4218 */
4219 if (found_key.objectid == found_key.offset ||
4220 is_cowonly_root(btrfs_ref_root(leaf, ref))) {
4221 ref_path->root_objectid = btrfs_ref_root(leaf, ref);
4222 ref_path->root_generation =
4223 btrfs_ref_generation(leaf, ref);
4224 if (level < 0) {
4225 /* special reference from the tree log */
4226 ref_path->nodes[0] = found_key.offset;
4227 ref_path->current_level = 0;
4228 }
4229 ret = 0;
4230 goto out;
4231 }
4232
4233 level++;
4234 BUG_ON(ref_path->nodes[level] != 0);
4235 ref_path->nodes[level] = found_key.offset;
4236 ref_path->current_level = level;
4237
4238 /*
4239 * the reference was created in the running transaction,
4240 * no need to continue walking up.
4241 */
4242 if (btrfs_ref_generation(leaf, ref) == trans->transid) {
4243 ref_path->root_objectid = btrfs_ref_root(leaf, ref);
4244 ref_path->root_generation =
4245 btrfs_ref_generation(leaf, ref);
4246 ret = 0;
4247 goto out;
4248 }
4249
4250 btrfs_release_path(extent_root, path);
4251 cond_resched();
4252 }
4253 /* reached max tree level, but no tree root found. */
4254 BUG();
4255 out:
4256 btrfs_free_path(path);
4257 return ret;
4258 }
4259
4260 static int btrfs_first_ref_path(struct btrfs_trans_handle *trans,
4261 struct btrfs_root *extent_root,
4262 struct btrfs_ref_path *ref_path,
4263 u64 extent_start)
4264 {
4265 memset(ref_path, 0, sizeof(*ref_path));
4266 ref_path->extent_start = extent_start;
4267
4268 return __next_ref_path(trans, extent_root, ref_path, 1);
4269 }
4270
4271 static int btrfs_next_ref_path(struct btrfs_trans_handle *trans,
4272 struct btrfs_root *extent_root,
4273 struct btrfs_ref_path *ref_path)
4274 {
4275 return __next_ref_path(trans, extent_root, ref_path, 0);
4276 }
4277
4278 static noinline int get_new_locations(struct inode *reloc_inode,
4279 struct btrfs_key *extent_key,
4280 u64 offset, int no_fragment,
4281 struct disk_extent **extents,
4282 int *nr_extents)
4283 {
4284 struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
4285 struct btrfs_path *path;
4286 struct btrfs_file_extent_item *fi;
4287 struct extent_buffer *leaf;
4288 struct disk_extent *exts = *extents;
4289 struct btrfs_key found_key;
4290 u64 cur_pos;
4291 u64 last_byte;
4292 u32 nritems;
4293 int nr = 0;
4294 int max = *nr_extents;
4295 int ret;
4296
4297 WARN_ON(!no_fragment && *extents);
4298 if (!exts) {
4299 max = 1;
4300 exts = kmalloc(sizeof(*exts) * max, GFP_NOFS);
4301 if (!exts)
4302 return -ENOMEM;
4303 }
4304
4305 path = btrfs_alloc_path();
4306 BUG_ON(!path);
4307
4308 cur_pos = extent_key->objectid - offset;
4309 last_byte = extent_key->objectid + extent_key->offset;
4310 ret = btrfs_lookup_file_extent(NULL, root, path, reloc_inode->i_ino,
4311 cur_pos, 0);
4312 if (ret < 0)
4313 goto out;
4314 if (ret > 0) {
4315 ret = -ENOENT;
4316 goto out;
4317 }
4318
4319 while (1) {
4320 leaf = path->nodes[0];
4321 nritems = btrfs_header_nritems(leaf);
4322 if (path->slots[0] >= nritems) {
4323 ret = btrfs_next_leaf(root, path);
4324 if (ret < 0)
4325 goto out;
4326 if (ret > 0)
4327 break;
4328 leaf = path->nodes[0];
4329 }
4330
4331 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
4332 if (found_key.offset != cur_pos ||
4333 found_key.type != BTRFS_EXTENT_DATA_KEY ||
4334 found_key.objectid != reloc_inode->i_ino)
4335 break;
4336
4337 fi = btrfs_item_ptr(leaf, path->slots[0],
4338 struct btrfs_file_extent_item);
4339 if (btrfs_file_extent_type(leaf, fi) !=
4340 BTRFS_FILE_EXTENT_REG ||
4341 btrfs_file_extent_disk_bytenr(leaf, fi) == 0)
4342 break;
4343
4344 if (nr == max) {
4345 struct disk_extent *old = exts;
4346 max *= 2;
4347 exts = kzalloc(sizeof(*exts) * max, GFP_NOFS);
4348 memcpy(exts, old, sizeof(*exts) * nr);
4349 if (old != *extents)
4350 kfree(old);
4351 }
4352
4353 exts[nr].disk_bytenr =
4354 btrfs_file_extent_disk_bytenr(leaf, fi);
4355 exts[nr].disk_num_bytes =
4356 btrfs_file_extent_disk_num_bytes(leaf, fi);
4357 exts[nr].offset = btrfs_file_extent_offset(leaf, fi);
4358 exts[nr].num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
4359 exts[nr].ram_bytes = btrfs_file_extent_ram_bytes(leaf, fi);
4360 exts[nr].compression = btrfs_file_extent_compression(leaf, fi);
4361 exts[nr].encryption = btrfs_file_extent_encryption(leaf, fi);
4362 exts[nr].other_encoding = btrfs_file_extent_other_encoding(leaf,
4363 fi);
4364 BUG_ON(exts[nr].offset > 0);
4365 BUG_ON(exts[nr].compression || exts[nr].encryption);
4366 BUG_ON(exts[nr].num_bytes != exts[nr].disk_num_bytes);
4367
4368 cur_pos += exts[nr].num_bytes;
4369 nr++;
4370
4371 if (cur_pos + offset >= last_byte)
4372 break;
4373
4374 if (no_fragment) {
4375 ret = 1;
4376 goto out;
4377 }
4378 path->slots[0]++;
4379 }
4380
4381 BUG_ON(cur_pos + offset > last_byte);
4382 if (cur_pos + offset < last_byte) {
4383 ret = -ENOENT;
4384 goto out;
4385 }
4386 ret = 0;
4387 out:
4388 btrfs_free_path(path);
4389 if (ret) {
4390 if (exts != *extents)
4391 kfree(exts);
4392 } else {
4393 *extents = exts;
4394 *nr_extents = nr;
4395 }
4396 return ret;
4397 }
4398
4399 static noinline int replace_one_extent(struct btrfs_trans_handle *trans,
4400 struct btrfs_root *root,
4401 struct btrfs_path *path,
4402 struct btrfs_key *extent_key,
4403 struct btrfs_key *leaf_key,
4404 struct btrfs_ref_path *ref_path,
4405 struct disk_extent *new_extents,
4406 int nr_extents)
4407 {
4408 struct extent_buffer *leaf;
4409 struct btrfs_file_extent_item *fi;
4410 struct inode *inode = NULL;
4411 struct btrfs_key key;
4412 u64 lock_start = 0;
4413 u64 lock_end = 0;
4414 u64 num_bytes;
4415 u64 ext_offset;
4416 u64 search_end = (u64)-1;
4417 u32 nritems;
4418 int nr_scaned = 0;
4419 int extent_locked = 0;
4420 int extent_type;
4421 int ret;
4422
4423 memcpy(&key, leaf_key, sizeof(key));
4424 if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS) {
4425 if (key.objectid < ref_path->owner_objectid ||
4426 (key.objectid == ref_path->owner_objectid &&
4427 key.type < BTRFS_EXTENT_DATA_KEY)) {
4428 key.objectid = ref_path->owner_objectid;
4429 key.type = BTRFS_EXTENT_DATA_KEY;
4430 key.offset = 0;
4431 }
4432 }
4433
4434 while (1) {
4435 ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
4436 if (ret < 0)
4437 goto out;
4438
4439 leaf = path->nodes[0];
4440 nritems = btrfs_header_nritems(leaf);
4441 next:
4442 if (extent_locked && ret > 0) {
4443 /*
4444 * the file extent item was modified by someone
4445 * before the extent got locked.
4446 */
4447 unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
4448 lock_end, GFP_NOFS);
4449 extent_locked = 0;
4450 }
4451
4452 if (path->slots[0] >= nritems) {
4453 if (++nr_scaned > 2)
4454 break;
4455
4456 BUG_ON(extent_locked);
4457 ret = btrfs_next_leaf(root, path);
4458 if (ret < 0)
4459 goto out;
4460 if (ret > 0)
4461 break;
4462 leaf = path->nodes[0];
4463 nritems = btrfs_header_nritems(leaf);
4464 }
4465
4466 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
4467
4468 if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS) {
4469 if ((key.objectid > ref_path->owner_objectid) ||
4470 (key.objectid == ref_path->owner_objectid &&
4471 key.type > BTRFS_EXTENT_DATA_KEY) ||
4472 key.offset >= search_end)
4473 break;
4474 }
4475
4476 if (inode && key.objectid != inode->i_ino) {
4477 BUG_ON(extent_locked);
4478 btrfs_release_path(root, path);
4479 mutex_unlock(&inode->i_mutex);
4480 iput(inode);
4481 inode = NULL;
4482 continue;
4483 }
4484
4485 if (key.type != BTRFS_EXTENT_DATA_KEY) {
4486 path->slots[0]++;
4487 ret = 1;
4488 goto next;
4489 }
4490 fi = btrfs_item_ptr(leaf, path->slots[0],
4491 struct btrfs_file_extent_item);
4492 extent_type = btrfs_file_extent_type(leaf, fi);
4493 if ((extent_type != BTRFS_FILE_EXTENT_REG &&
4494 extent_type != BTRFS_FILE_EXTENT_PREALLOC) ||
4495 (btrfs_file_extent_disk_bytenr(leaf, fi) !=
4496 extent_key->objectid)) {
4497 path->slots[0]++;
4498 ret = 1;
4499 goto next;
4500 }
4501
4502 num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
4503 ext_offset = btrfs_file_extent_offset(leaf, fi);
4504
4505 if (search_end == (u64)-1) {
4506 search_end = key.offset - ext_offset +
4507 btrfs_file_extent_ram_bytes(leaf, fi);
4508 }
4509
4510 if (!extent_locked) {
4511 lock_start = key.offset;
4512 lock_end = lock_start + num_bytes - 1;
4513 } else {
4514 if (lock_start > key.offset ||
4515 lock_end + 1 < key.offset + num_bytes) {
4516 unlock_extent(&BTRFS_I(inode)->io_tree,
4517 lock_start, lock_end, GFP_NOFS);
4518 extent_locked = 0;
4519 }
4520 }
4521
4522 if (!inode) {
4523 btrfs_release_path(root, path);
4524
4525 inode = btrfs_iget_locked(root->fs_info->sb,
4526 key.objectid, root);
4527 if (inode->i_state & I_NEW) {
4528 BTRFS_I(inode)->root = root;
4529 BTRFS_I(inode)->location.objectid =
4530 key.objectid;
4531 BTRFS_I(inode)->location.type =
4532 BTRFS_INODE_ITEM_KEY;
4533 BTRFS_I(inode)->location.offset = 0;
4534 btrfs_read_locked_inode(inode);
4535 unlock_new_inode(inode);
4536 }
4537 /*
4538 * some code call btrfs_commit_transaction while
4539 * holding the i_mutex, so we can't use mutex_lock
4540 * here.
4541 */
4542 if (is_bad_inode(inode) ||
4543 !mutex_trylock(&inode->i_mutex)) {
4544 iput(inode);
4545 inode = NULL;
4546 key.offset = (u64)-1;
4547 goto skip;
4548 }
4549 }
4550
4551 if (!extent_locked) {
4552 struct btrfs_ordered_extent *ordered;
4553
4554 btrfs_release_path(root, path);
4555
4556 lock_extent(&BTRFS_I(inode)->io_tree, lock_start,
4557 lock_end, GFP_NOFS);
4558 ordered = btrfs_lookup_first_ordered_extent(inode,
4559 lock_end);
4560 if (ordered &&
4561 ordered->file_offset <= lock_end &&
4562 ordered->file_offset + ordered->len > lock_start) {
4563 unlock_extent(&BTRFS_I(inode)->io_tree,
4564 lock_start, lock_end, GFP_NOFS);
4565 btrfs_start_ordered_extent(inode, ordered, 1);
4566 btrfs_put_ordered_extent(ordered);
4567 key.offset += num_bytes;
4568 goto skip;
4569 }
4570 if (ordered)
4571 btrfs_put_ordered_extent(ordered);
4572
4573 extent_locked = 1;
4574 continue;
4575 }
4576
4577 if (nr_extents == 1) {
4578 /* update extent pointer in place */
4579 btrfs_set_file_extent_disk_bytenr(leaf, fi,
4580 new_extents[0].disk_bytenr);
4581 btrfs_set_file_extent_disk_num_bytes(leaf, fi,
4582 new_extents[0].disk_num_bytes);
4583 btrfs_mark_buffer_dirty(leaf);
4584
4585 btrfs_drop_extent_cache(inode, key.offset,
4586 key.offset + num_bytes - 1, 0);
4587
4588 ret = btrfs_inc_extent_ref(trans, root,
4589 new_extents[0].disk_bytenr,
4590 new_extents[0].disk_num_bytes,
4591 leaf->start,
4592 root->root_key.objectid,
4593 trans->transid,
4594 key.objectid);
4595 BUG_ON(ret);
4596
4597 ret = btrfs_free_extent(trans, root,
4598 extent_key->objectid,
4599 extent_key->offset,
4600 leaf->start,
4601 btrfs_header_owner(leaf),
4602 btrfs_header_generation(leaf),
4603 key.objectid, 0);
4604 BUG_ON(ret);
4605
4606 btrfs_release_path(root, path);
4607 key.offset += num_bytes;
4608 } else {
4609 BUG_ON(1);
4610 #if 0
4611 u64 alloc_hint;
4612 u64 extent_len;
4613 int i;
4614 /*
4615 * drop old extent pointer at first, then insert the
4616 * new pointers one bye one
4617 */
4618 btrfs_release_path(root, path);
4619 ret = btrfs_drop_extents(trans, root, inode, key.offset,
4620 key.offset + num_bytes,
4621 key.offset, &alloc_hint);
4622 BUG_ON(ret);
4623
4624 for (i = 0; i < nr_extents; i++) {
4625 if (ext_offset >= new_extents[i].num_bytes) {
4626 ext_offset -= new_extents[i].num_bytes;
4627 continue;
4628 }
4629 extent_len = min(new_extents[i].num_bytes -
4630 ext_offset, num_bytes);
4631
4632 ret = btrfs_insert_empty_item(trans, root,
4633 path, &key,
4634 sizeof(*fi));
4635 BUG_ON(ret);
4636
4637 leaf = path->nodes[0];
4638 fi = btrfs_item_ptr(leaf, path->slots[0],
4639 struct btrfs_file_extent_item);
4640 btrfs_set_file_extent_generation(leaf, fi,
4641 trans->transid);
4642 btrfs_set_file_extent_type(leaf, fi,
4643 BTRFS_FILE_EXTENT_REG);
4644 btrfs_set_file_extent_disk_bytenr(leaf, fi,
4645 new_extents[i].disk_bytenr);
4646 btrfs_set_file_extent_disk_num_bytes(leaf, fi,
4647 new_extents[i].disk_num_bytes);
4648 btrfs_set_file_extent_ram_bytes(leaf, fi,
4649 new_extents[i].ram_bytes);
4650
4651 btrfs_set_file_extent_compression(leaf, fi,
4652 new_extents[i].compression);
4653 btrfs_set_file_extent_encryption(leaf, fi,
4654 new_extents[i].encryption);
4655 btrfs_set_file_extent_other_encoding(leaf, fi,
4656 new_extents[i].other_encoding);
4657
4658 btrfs_set_file_extent_num_bytes(leaf, fi,
4659 extent_len);
4660 ext_offset += new_extents[i].offset;
4661 btrfs_set_file_extent_offset(leaf, fi,
4662 ext_offset);
4663 btrfs_mark_buffer_dirty(leaf);
4664
4665 btrfs_drop_extent_cache(inode, key.offset,
4666 key.offset + extent_len - 1, 0);
4667
4668 ret = btrfs_inc_extent_ref(trans, root,
4669 new_extents[i].disk_bytenr,
4670 new_extents[i].disk_num_bytes,
4671 leaf->start,
4672 root->root_key.objectid,
4673 trans->transid, key.objectid);
4674 BUG_ON(ret);
4675 btrfs_release_path(root, path);
4676
4677 inode_add_bytes(inode, extent_len);
4678
4679 ext_offset = 0;
4680 num_bytes -= extent_len;
4681 key.offset += extent_len;
4682
4683 if (num_bytes == 0)
4684 break;
4685 }
4686 BUG_ON(i >= nr_extents);
4687 #endif
4688 }
4689
4690 if (extent_locked) {
4691 unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
4692 lock_end, GFP_NOFS);
4693 extent_locked = 0;
4694 }
4695 skip:
4696 if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS &&
4697 key.offset >= search_end)
4698 break;
4699
4700 cond_resched();
4701 }
4702 ret = 0;
4703 out:
4704 btrfs_release_path(root, path);
4705 if (inode) {
4706 mutex_unlock(&inode->i_mutex);
4707 if (extent_locked) {
4708 unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
4709 lock_end, GFP_NOFS);
4710 }
4711 iput(inode);
4712 }
4713 return ret;
4714 }
4715
4716 int btrfs_reloc_tree_cache_ref(struct btrfs_trans_handle *trans,
4717 struct btrfs_root *root,
4718 struct extent_buffer *buf, u64 orig_start)
4719 {
4720 int level;
4721 int ret;
4722
4723 BUG_ON(btrfs_header_generation(buf) != trans->transid);
4724 BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
4725
4726 level = btrfs_header_level(buf);
4727 if (level == 0) {
4728 struct btrfs_leaf_ref *ref;
4729 struct btrfs_leaf_ref *orig_ref;
4730
4731 orig_ref = btrfs_lookup_leaf_ref(root, orig_start);
4732 if (!orig_ref)
4733 return -ENOENT;
4734
4735 ref = btrfs_alloc_leaf_ref(root, orig_ref->nritems);
4736 if (!ref) {
4737 btrfs_free_leaf_ref(root, orig_ref);
4738 return -ENOMEM;
4739 }
4740
4741 ref->nritems = orig_ref->nritems;
4742 memcpy(ref->extents, orig_ref->extents,
4743 sizeof(ref->extents[0]) * ref->nritems);
4744
4745 btrfs_free_leaf_ref(root, orig_ref);
4746
4747 ref->root_gen = trans->transid;
4748 ref->bytenr = buf->start;
4749 ref->owner = btrfs_header_owner(buf);
4750 ref->generation = btrfs_header_generation(buf);
4751
4752 ret = btrfs_add_leaf_ref(root, ref, 0);
4753 WARN_ON(ret);
4754 btrfs_free_leaf_ref(root, ref);
4755 }
4756 return 0;
4757 }
4758
4759 static noinline int invalidate_extent_cache(struct btrfs_root *root,
4760 struct extent_buffer *leaf,
4761 struct btrfs_block_group_cache *group,
4762 struct btrfs_root *target_root)
4763 {
4764 struct btrfs_key key;
4765 struct inode *inode = NULL;
4766 struct btrfs_file_extent_item *fi;
4767 u64 num_bytes;
4768 u64 skip_objectid = 0;
4769 u32 nritems;
4770 u32 i;
4771
4772 nritems = btrfs_header_nritems(leaf);
4773 for (i = 0; i < nritems; i++) {
4774 btrfs_item_key_to_cpu(leaf, &key, i);
4775 if (key.objectid == skip_objectid ||
4776 key.type != BTRFS_EXTENT_DATA_KEY)
4777 continue;
4778 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
4779 if (btrfs_file_extent_type(leaf, fi) ==
4780 BTRFS_FILE_EXTENT_INLINE)
4781 continue;
4782 if (btrfs_file_extent_disk_bytenr(leaf, fi) == 0)
4783 continue;
4784 if (!inode || inode->i_ino != key.objectid) {
4785 iput(inode);
4786 inode = btrfs_ilookup(target_root->fs_info->sb,
4787 key.objectid, target_root, 1);
4788 }
4789 if (!inode) {
4790 skip_objectid = key.objectid;
4791 continue;
4792 }
4793 num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
4794
4795 lock_extent(&BTRFS_I(inode)->io_tree, key.offset,
4796 key.offset + num_bytes - 1, GFP_NOFS);
4797 btrfs_drop_extent_cache(inode, key.offset,
4798 key.offset + num_bytes - 1, 1);
4799 unlock_extent(&BTRFS_I(inode)->io_tree, key.offset,
4800 key.offset + num_bytes - 1, GFP_NOFS);
4801 cond_resched();
4802 }
4803 iput(inode);
4804 return 0;
4805 }
4806
4807 static noinline int replace_extents_in_leaf(struct btrfs_trans_handle *trans,
4808 struct btrfs_root *root,
4809 struct extent_buffer *leaf,
4810 struct btrfs_block_group_cache *group,
4811 struct inode *reloc_inode)
4812 {
4813 struct btrfs_key key;
4814 struct btrfs_key extent_key;
4815 struct btrfs_file_extent_item *fi;
4816 struct btrfs_leaf_ref *ref;
4817 struct disk_extent *new_extent;
4818 u64 bytenr;
4819 u64 num_bytes;
4820 u32 nritems;
4821 u32 i;
4822 int ext_index;
4823 int nr_extent;
4824 int ret;
4825
4826 new_extent = kmalloc(sizeof(*new_extent), GFP_NOFS);
4827 BUG_ON(!new_extent);
4828
4829 ref = btrfs_lookup_leaf_ref(root, leaf->start);
4830 BUG_ON(!ref);
4831
4832 ext_index = -1;
4833 nritems = btrfs_header_nritems(leaf);
4834 for (i = 0; i < nritems; i++) {
4835 btrfs_item_key_to_cpu(leaf, &key, i);
4836 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
4837 continue;
4838 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
4839 if (btrfs_file_extent_type(leaf, fi) ==
4840 BTRFS_FILE_EXTENT_INLINE)
4841 continue;
4842 bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
4843 num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
4844 if (bytenr == 0)
4845 continue;
4846
4847 ext_index++;
4848 if (bytenr >= group->key.objectid + group->key.offset ||
4849 bytenr + num_bytes <= group->key.objectid)
4850 continue;
4851
4852 extent_key.objectid = bytenr;
4853 extent_key.offset = num_bytes;
4854 extent_key.type = BTRFS_EXTENT_ITEM_KEY;
4855 nr_extent = 1;
4856 ret = get_new_locations(reloc_inode, &extent_key,
4857 group->key.objectid, 1,
4858 &new_extent, &nr_extent);
4859 if (ret > 0)
4860 continue;
4861 BUG_ON(ret < 0);
4862
4863 BUG_ON(ref->extents[ext_index].bytenr != bytenr);
4864 BUG_ON(ref->extents[ext_index].num_bytes != num_bytes);
4865 ref->extents[ext_index].bytenr = new_extent->disk_bytenr;
4866 ref->extents[ext_index].num_bytes = new_extent->disk_num_bytes;
4867
4868 btrfs_set_file_extent_disk_bytenr(leaf, fi,
4869 new_extent->disk_bytenr);
4870 btrfs_set_file_extent_disk_num_bytes(leaf, fi,
4871 new_extent->disk_num_bytes);
4872 btrfs_mark_buffer_dirty(leaf);
4873
4874 ret = btrfs_inc_extent_ref(trans, root,
4875 new_extent->disk_bytenr,
4876 new_extent->disk_num_bytes,
4877 leaf->start,
4878 root->root_key.objectid,
4879 trans->transid, key.objectid);
4880 BUG_ON(ret);
4881
4882 ret = btrfs_free_extent(trans, root,
4883 bytenr, num_bytes, leaf->start,
4884 btrfs_header_owner(leaf),
4885 btrfs_header_generation(leaf),
4886 key.objectid, 0);
4887 BUG_ON(ret);
4888 cond_resched();
4889 }
4890 kfree(new_extent);
4891 BUG_ON(ext_index + 1 != ref->nritems);
4892 btrfs_free_leaf_ref(root, ref);
4893 return 0;
4894 }
4895
4896 int btrfs_free_reloc_root(struct btrfs_trans_handle *trans,
4897 struct btrfs_root *root)
4898 {
4899 struct btrfs_root *reloc_root;
4900 int ret;
4901
4902 if (root->reloc_root) {
4903 reloc_root = root->reloc_root;
4904 root->reloc_root = NULL;
4905 list_add(&reloc_root->dead_list,
4906 &root->fs_info->dead_reloc_roots);
4907
4908 btrfs_set_root_bytenr(&reloc_root->root_item,
4909 reloc_root->node->start);
4910 btrfs_set_root_level(&root->root_item,
4911 btrfs_header_level(reloc_root->node));
4912 memset(&reloc_root->root_item.drop_progress, 0,
4913 sizeof(struct btrfs_disk_key));
4914 reloc_root->root_item.drop_level = 0;
4915
4916 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4917 &reloc_root->root_key,
4918 &reloc_root->root_item);
4919 BUG_ON(ret);
4920 }
4921 return 0;
4922 }
4923
4924 int btrfs_drop_dead_reloc_roots(struct btrfs_root *root)
4925 {
4926 struct btrfs_trans_handle *trans;
4927 struct btrfs_root *reloc_root;
4928 struct btrfs_root *prev_root = NULL;
4929 struct list_head dead_roots;
4930 int ret;
4931 unsigned long nr;
4932
4933 INIT_LIST_HEAD(&dead_roots);
4934 list_splice_init(&root->fs_info->dead_reloc_roots, &dead_roots);
4935
4936 while (!list_empty(&dead_roots)) {
4937 reloc_root = list_entry(dead_roots.prev,
4938 struct btrfs_root, dead_list);
4939 list_del_init(&reloc_root->dead_list);
4940
4941 BUG_ON(reloc_root->commit_root != NULL);
4942 while (1) {
4943 trans = btrfs_join_transaction(root, 1);
4944 BUG_ON(!trans);
4945
4946 mutex_lock(&root->fs_info->drop_mutex);
4947 ret = btrfs_drop_snapshot(trans, reloc_root);
4948 if (ret != -EAGAIN)
4949 break;
4950 mutex_unlock(&root->fs_info->drop_mutex);
4951
4952 nr = trans->blocks_used;
4953 ret = btrfs_end_transaction(trans, root);
4954 BUG_ON(ret);
4955 btrfs_btree_balance_dirty(root, nr);
4956 }
4957
4958 free_extent_buffer(reloc_root->node);
4959
4960 ret = btrfs_del_root(trans, root->fs_info->tree_root,
4961 &reloc_root->root_key);
4962 BUG_ON(ret);
4963 mutex_unlock(&root->fs_info->drop_mutex);
4964
4965 nr = trans->blocks_used;
4966 ret = btrfs_end_transaction(trans, root);
4967 BUG_ON(ret);
4968 btrfs_btree_balance_dirty(root, nr);
4969
4970 kfree(prev_root);
4971 prev_root = reloc_root;
4972 }
4973 if (prev_root) {
4974 btrfs_remove_leaf_refs(prev_root, (u64)-1, 0);
4975 kfree(prev_root);
4976 }
4977 return 0;
4978 }
4979
4980 int btrfs_add_dead_reloc_root(struct btrfs_root *root)
4981 {
4982 list_add(&root->dead_list, &root->fs_info->dead_reloc_roots);
4983 return 0;
4984 }
4985
4986 int btrfs_cleanup_reloc_trees(struct btrfs_root *root)
4987 {
4988 struct btrfs_root *reloc_root;
4989 struct btrfs_trans_handle *trans;
4990 struct btrfs_key location;
4991 int found;
4992 int ret;
4993
4994 mutex_lock(&root->fs_info->tree_reloc_mutex);
4995 ret = btrfs_find_dead_roots(root, BTRFS_TREE_RELOC_OBJECTID, NULL);
4996 BUG_ON(ret);
4997 found = !list_empty(&root->fs_info->dead_reloc_roots);
4998 mutex_unlock(&root->fs_info->tree_reloc_mutex);
4999
5000 if (found) {
5001 trans = btrfs_start_transaction(root, 1);
5002 BUG_ON(!trans);
5003 ret = btrfs_commit_transaction(trans, root);
5004 BUG_ON(ret);
5005 }
5006
5007 location.objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
5008 location.offset = (u64)-1;
5009 location.type = BTRFS_ROOT_ITEM_KEY;
5010
5011 reloc_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
5012 BUG_ON(!reloc_root);
5013 btrfs_orphan_cleanup(reloc_root);
5014 return 0;
5015 }
5016
5017 static noinline int init_reloc_tree(struct btrfs_trans_handle *trans,
5018 struct btrfs_root *root)
5019 {
5020 struct btrfs_root *reloc_root;
5021 struct extent_buffer *eb;
5022 struct btrfs_root_item *root_item;
5023 struct btrfs_key root_key;
5024 int ret;
5025
5026 BUG_ON(!root->ref_cows);
5027 if (root->reloc_root)
5028 return 0;
5029
5030 root_item = kmalloc(sizeof(*root_item), GFP_NOFS);
5031 BUG_ON(!root_item);
5032
5033 ret = btrfs_copy_root(trans, root, root->commit_root,
5034 &eb, BTRFS_TREE_RELOC_OBJECTID);
5035 BUG_ON(ret);
5036
5037 root_key.objectid = BTRFS_TREE_RELOC_OBJECTID;
5038 root_key.offset = root->root_key.objectid;
5039 root_key.type = BTRFS_ROOT_ITEM_KEY;
5040
5041 memcpy(root_item, &root->root_item, sizeof(root_item));
5042 btrfs_set_root_refs(root_item, 0);
5043 btrfs_set_root_bytenr(root_item, eb->start);
5044 btrfs_set_root_level(root_item, btrfs_header_level(eb));
5045 btrfs_set_root_generation(root_item, trans->transid);
5046
5047 btrfs_tree_unlock(eb);
5048 free_extent_buffer(eb);
5049
5050 ret = btrfs_insert_root(trans, root->fs_info->tree_root,
5051 &root_key, root_item);
5052 BUG_ON(ret);
5053 kfree(root_item);
5054
5055 reloc_root = btrfs_read_fs_root_no_radix(root->fs_info->tree_root,
5056 &root_key);
5057 BUG_ON(!reloc_root);
5058 reloc_root->last_trans = trans->transid;
5059 reloc_root->commit_root = NULL;
5060 reloc_root->ref_tree = &root->fs_info->reloc_ref_tree;
5061
5062 root->reloc_root = reloc_root;
5063 return 0;
5064 }
5065
5066 /*
5067 * Core function of space balance.
5068 *
5069 * The idea is using reloc trees to relocate tree blocks in reference
5070 * counted roots. There is one reloc tree for each subvol, and all
5071 * reloc trees share same root key objectid. Reloc trees are snapshots
5072 * of the latest committed roots of subvols (root->commit_root).
5073 *
5074 * To relocate a tree block referenced by a subvol, there are two steps.
5075 * COW the block through subvol's reloc tree, then update block pointer
5076 * in the subvol to point to the new block. Since all reloc trees share
5077 * same root key objectid, doing special handing for tree blocks owned
5078 * by them is easy. Once a tree block has been COWed in one reloc tree,
5079 * we can use the resulting new block directly when the same block is
5080 * required to COW again through other reloc trees. By this way, relocated
5081 * tree blocks are shared between reloc trees, so they are also shared
5082 * between subvols.
5083 */
5084 static noinline int relocate_one_path(struct btrfs_trans_handle *trans,
5085 struct btrfs_root *root,
5086 struct btrfs_path *path,
5087 struct btrfs_key *first_key,
5088 struct btrfs_ref_path *ref_path,
5089 struct btrfs_block_group_cache *group,
5090 struct inode *reloc_inode)
5091 {
5092 struct btrfs_root *reloc_root;
5093 struct extent_buffer *eb = NULL;
5094 struct btrfs_key *keys;
5095 u64 *nodes;
5096 int level;
5097 int shared_level;
5098 int lowest_level = 0;
5099 int ret;
5100
5101 if (ref_path->owner_objectid < BTRFS_FIRST_FREE_OBJECTID)
5102 lowest_level = ref_path->owner_objectid;
5103
5104 if (!root->ref_cows) {
5105 path->lowest_level = lowest_level;
5106 ret = btrfs_search_slot(trans, root, first_key, path, 0, 1);
5107 BUG_ON(ret < 0);
5108 path->lowest_level = 0;
5109 btrfs_release_path(root, path);
5110 return 0;
5111 }
5112
5113 mutex_lock(&root->fs_info->tree_reloc_mutex);
5114 ret = init_reloc_tree(trans, root);
5115 BUG_ON(ret);
5116 reloc_root = root->reloc_root;
5117
5118 shared_level = ref_path->shared_level;
5119 ref_path->shared_level = BTRFS_MAX_LEVEL - 1;
5120
5121 keys = ref_path->node_keys;
5122 nodes = ref_path->new_nodes;
5123 memset(&keys[shared_level + 1], 0,
5124 sizeof(*keys) * (BTRFS_MAX_LEVEL - shared_level - 1));
5125 memset(&nodes[shared_level + 1], 0,
5126 sizeof(*nodes) * (BTRFS_MAX_LEVEL - shared_level - 1));
5127
5128 if (nodes[lowest_level] == 0) {
5129 path->lowest_level = lowest_level;
5130 ret = btrfs_search_slot(trans, reloc_root, first_key, path,
5131 0, 1);
5132 BUG_ON(ret);
5133 for (level = lowest_level; level < BTRFS_MAX_LEVEL; level++) {
5134 eb = path->nodes[level];
5135 if (!eb || eb == reloc_root->node)
5136 break;
5137 nodes[level] = eb->start;
5138 if (level == 0)
5139 btrfs_item_key_to_cpu(eb, &keys[level], 0);
5140 else
5141 btrfs_node_key_to_cpu(eb, &keys[level], 0);
5142 }
5143 if (nodes[0] &&
5144 ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
5145 eb = path->nodes[0];
5146 ret = replace_extents_in_leaf(trans, reloc_root, eb,
5147 group, reloc_inode);
5148 BUG_ON(ret);
5149 }
5150 btrfs_release_path(reloc_root, path);
5151 } else {
5152 ret = btrfs_merge_path(trans, reloc_root, keys, nodes,
5153 lowest_level);
5154 BUG_ON(ret);
5155 }
5156
5157 /*
5158 * replace tree blocks in the fs tree with tree blocks in
5159 * the reloc tree.
5160 */
5161 ret = btrfs_merge_path(trans, root, keys, nodes, lowest_level);
5162 BUG_ON(ret < 0);
5163
5164 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
5165 ret = btrfs_search_slot(trans, reloc_root, first_key, path,
5166 0, 0);
5167 BUG_ON(ret);
5168 extent_buffer_get(path->nodes[0]);
5169 eb = path->nodes[0];
5170 btrfs_release_path(reloc_root, path);
5171 ret = invalidate_extent_cache(reloc_root, eb, group, root);
5172 BUG_ON(ret);
5173 free_extent_buffer(eb);
5174 }
5175
5176 mutex_unlock(&root->fs_info->tree_reloc_mutex);
5177 path->lowest_level = 0;
5178 return 0;
5179 }
5180
5181 static noinline int relocate_tree_block(struct btrfs_trans_handle *trans,
5182 struct btrfs_root *root,
5183 struct btrfs_path *path,
5184 struct btrfs_key *first_key,
5185 struct btrfs_ref_path *ref_path)
5186 {
5187 int ret;
5188
5189 ret = relocate_one_path(trans, root, path, first_key,
5190 ref_path, NULL, NULL);
5191 BUG_ON(ret);
5192
5193 return 0;
5194 }
5195
5196 static noinline int del_extent_zero(struct btrfs_trans_handle *trans,
5197 struct btrfs_root *extent_root,
5198 struct btrfs_path *path,
5199 struct btrfs_key *extent_key)
5200 {
5201 int ret;
5202
5203 ret = btrfs_search_slot(trans, extent_root, extent_key, path, -1, 1);
5204 if (ret)
5205 goto out;
5206 ret = btrfs_del_item(trans, extent_root, path);
5207 out:
5208 btrfs_release_path(extent_root, path);
5209 return ret;
5210 }
5211
5212 static noinline struct btrfs_root *read_ref_root(struct btrfs_fs_info *fs_info,
5213 struct btrfs_ref_path *ref_path)
5214 {
5215 struct btrfs_key root_key;
5216
5217 root_key.objectid = ref_path->root_objectid;
5218 root_key.type = BTRFS_ROOT_ITEM_KEY;
5219 if (is_cowonly_root(ref_path->root_objectid))
5220 root_key.offset = 0;
5221 else
5222 root_key.offset = (u64)-1;
5223
5224 return btrfs_read_fs_root_no_name(fs_info, &root_key);
5225 }
5226
5227 static noinline int relocate_one_extent(struct btrfs_root *extent_root,
5228 struct btrfs_path *path,
5229 struct btrfs_key *extent_key,
5230 struct btrfs_block_group_cache *group,
5231 struct inode *reloc_inode, int pass)
5232 {
5233 struct btrfs_trans_handle *trans;
5234 struct btrfs_root *found_root;
5235 struct btrfs_ref_path *ref_path = NULL;
5236 struct disk_extent *new_extents = NULL;
5237 int nr_extents = 0;
5238 int loops;
5239 int ret;
5240 int level;
5241 struct btrfs_key first_key;
5242 u64 prev_block = 0;
5243
5244
5245 trans = btrfs_start_transaction(extent_root, 1);
5246 BUG_ON(!trans);
5247
5248 if (extent_key->objectid == 0) {
5249 ret = del_extent_zero(trans, extent_root, path, extent_key);
5250 goto out;
5251 }
5252
5253 ref_path = kmalloc(sizeof(*ref_path), GFP_NOFS);
5254 if (!ref_path) {
5255 ret = -ENOMEM;
5256 goto out;
5257 }
5258
5259 for (loops = 0; ; loops++) {
5260 if (loops == 0) {
5261 ret = btrfs_first_ref_path(trans, extent_root, ref_path,
5262 extent_key->objectid);
5263 } else {
5264 ret = btrfs_next_ref_path(trans, extent_root, ref_path);
5265 }
5266 if (ret < 0)
5267 goto out;
5268 if (ret > 0)
5269 break;
5270
5271 if (ref_path->root_objectid == BTRFS_TREE_LOG_OBJECTID ||
5272 ref_path->root_objectid == BTRFS_TREE_RELOC_OBJECTID)
5273 continue;
5274
5275 found_root = read_ref_root(extent_root->fs_info, ref_path);
5276 BUG_ON(!found_root);
5277 /*
5278 * for reference counted tree, only process reference paths
5279 * rooted at the latest committed root.
5280 */
5281 if (found_root->ref_cows &&
5282 ref_path->root_generation != found_root->root_key.offset)
5283 continue;
5284
5285 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
5286 if (pass == 0) {
5287 /*
5288 * copy data extents to new locations
5289 */
5290 u64 group_start = group->key.objectid;
5291 ret = relocate_data_extent(reloc_inode,
5292 extent_key,
5293 group_start);
5294 if (ret < 0)
5295 goto out;
5296 break;
5297 }
5298 level = 0;
5299 } else {
5300 level = ref_path->owner_objectid;
5301 }
5302
5303 if (prev_block != ref_path->nodes[level]) {
5304 struct extent_buffer *eb;
5305 u64 block_start = ref_path->nodes[level];
5306 u64 block_size = btrfs_level_size(found_root, level);
5307
5308 eb = read_tree_block(found_root, block_start,
5309 block_size, 0);
5310 btrfs_tree_lock(eb);
5311 BUG_ON(level != btrfs_header_level(eb));
5312
5313 if (level == 0)
5314 btrfs_item_key_to_cpu(eb, &first_key, 0);
5315 else
5316 btrfs_node_key_to_cpu(eb, &first_key, 0);
5317
5318 btrfs_tree_unlock(eb);
5319 free_extent_buffer(eb);
5320 prev_block = block_start;
5321 }
5322
5323 mutex_lock(&extent_root->fs_info->trans_mutex);
5324 btrfs_record_root_in_trans(found_root);
5325 mutex_unlock(&extent_root->fs_info->trans_mutex);
5326 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
5327 /*
5328 * try to update data extent references while
5329 * keeping metadata shared between snapshots.
5330 */
5331 if (pass == 1) {
5332 ret = relocate_one_path(trans, found_root,
5333 path, &first_key, ref_path,
5334 group, reloc_inode);
5335 if (ret < 0)
5336 goto out;
5337 continue;
5338 }
5339 /*
5340 * use fallback method to process the remaining
5341 * references.
5342 */
5343 if (!new_extents) {
5344 u64 group_start = group->key.objectid;
5345 new_extents = kmalloc(sizeof(*new_extents),
5346 GFP_NOFS);
5347 nr_extents = 1;
5348 ret = get_new_locations(reloc_inode,
5349 extent_key,
5350 group_start, 1,
5351 &new_extents,
5352 &nr_extents);
5353 if (ret)
5354 goto out;
5355 }
5356 ret = replace_one_extent(trans, found_root,
5357 path, extent_key,
5358 &first_key, ref_path,
5359 new_extents, nr_extents);
5360 } else {
5361 ret = relocate_tree_block(trans, found_root, path,
5362 &first_key, ref_path);
5363 }
5364 if (ret < 0)
5365 goto out;
5366 }
5367 ret = 0;
5368 out:
5369 btrfs_end_transaction(trans, extent_root);
5370 kfree(new_extents);
5371 kfree(ref_path);
5372 return ret;
5373 }
5374
5375 static u64 update_block_group_flags(struct btrfs_root *root, u64 flags)
5376 {
5377 u64 num_devices;
5378 u64 stripped = BTRFS_BLOCK_GROUP_RAID0 |
5379 BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10;
5380
5381 num_devices = root->fs_info->fs_devices->rw_devices;
5382 if (num_devices == 1) {
5383 stripped |= BTRFS_BLOCK_GROUP_DUP;
5384 stripped = flags & ~stripped;
5385
5386 /* turn raid0 into single device chunks */
5387 if (flags & BTRFS_BLOCK_GROUP_RAID0)
5388 return stripped;
5389
5390 /* turn mirroring into duplication */
5391 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
5392 BTRFS_BLOCK_GROUP_RAID10))
5393 return stripped | BTRFS_BLOCK_GROUP_DUP;
5394 return flags;
5395 } else {
5396 /* they already had raid on here, just return */
5397 if (flags & stripped)
5398 return flags;
5399
5400 stripped |= BTRFS_BLOCK_GROUP_DUP;
5401 stripped = flags & ~stripped;
5402
5403 /* switch duplicated blocks with raid1 */
5404 if (flags & BTRFS_BLOCK_GROUP_DUP)
5405 return stripped | BTRFS_BLOCK_GROUP_RAID1;
5406
5407 /* turn single device chunks into raid0 */
5408 return stripped | BTRFS_BLOCK_GROUP_RAID0;
5409 }
5410 return flags;
5411 }
5412
5413 static int __alloc_chunk_for_shrink(struct btrfs_root *root,
5414 struct btrfs_block_group_cache *shrink_block_group,
5415 int force)
5416 {
5417 struct btrfs_trans_handle *trans;
5418 u64 new_alloc_flags;
5419 u64 calc;
5420
5421 spin_lock(&shrink_block_group->lock);
5422 if (btrfs_block_group_used(&shrink_block_group->item) > 0) {
5423 spin_unlock(&shrink_block_group->lock);
5424
5425 trans = btrfs_start_transaction(root, 1);
5426 spin_lock(&shrink_block_group->lock);
5427
5428 new_alloc_flags = update_block_group_flags(root,
5429 shrink_block_group->flags);
5430 if (new_alloc_flags != shrink_block_group->flags) {
5431 calc =
5432 btrfs_block_group_used(&shrink_block_group->item);
5433 } else {
5434 calc = shrink_block_group->key.offset;
5435 }
5436 spin_unlock(&shrink_block_group->lock);
5437
5438 do_chunk_alloc(trans, root->fs_info->extent_root,
5439 calc + 2 * 1024 * 1024, new_alloc_flags, force);
5440
5441 btrfs_end_transaction(trans, root);
5442 } else
5443 spin_unlock(&shrink_block_group->lock);
5444 return 0;
5445 }
5446
5447 static int __insert_orphan_inode(struct btrfs_trans_handle *trans,
5448 struct btrfs_root *root,
5449 u64 objectid, u64 size)
5450 {
5451 struct btrfs_path *path;
5452 struct btrfs_inode_item *item;
5453 struct extent_buffer *leaf;
5454 int ret;
5455
5456 path = btrfs_alloc_path();
5457 if (!path)
5458 return -ENOMEM;
5459
5460 path->leave_spinning = 1;
5461 ret = btrfs_insert_empty_inode(trans, root, path, objectid);
5462 if (ret)
5463 goto out;
5464
5465 leaf = path->nodes[0];
5466 item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_inode_item);
5467 memset_extent_buffer(leaf, 0, (unsigned long)item, sizeof(*item));
5468 btrfs_set_inode_generation(leaf, item, 1);
5469 btrfs_set_inode_size(leaf, item, size);
5470 btrfs_set_inode_mode(leaf, item, S_IFREG | 0600);
5471 btrfs_set_inode_flags(leaf, item, BTRFS_INODE_NOCOMPRESS);
5472 btrfs_mark_buffer_dirty(leaf);
5473 btrfs_release_path(root, path);
5474 out:
5475 btrfs_free_path(path);
5476 return ret;
5477 }
5478
5479 static noinline struct inode *create_reloc_inode(struct btrfs_fs_info *fs_info,
5480 struct btrfs_block_group_cache *group)
5481 {
5482 struct inode *inode = NULL;
5483 struct btrfs_trans_handle *trans;
5484 struct btrfs_root *root;
5485 struct btrfs_key root_key;
5486 u64 objectid = BTRFS_FIRST_FREE_OBJECTID;
5487 int err = 0;
5488
5489 root_key.objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
5490 root_key.type = BTRFS_ROOT_ITEM_KEY;
5491 root_key.offset = (u64)-1;
5492 root = btrfs_read_fs_root_no_name(fs_info, &root_key);
5493 if (IS_ERR(root))
5494 return ERR_CAST(root);
5495
5496 trans = btrfs_start_transaction(root, 1);
5497 BUG_ON(!trans);
5498
5499 err = btrfs_find_free_objectid(trans, root, objectid, &objectid);
5500 if (err)
5501 goto out;
5502
5503 err = __insert_orphan_inode(trans, root, objectid, group->key.offset);
5504 BUG_ON(err);
5505
5506 err = btrfs_insert_file_extent(trans, root, objectid, 0, 0, 0,
5507 group->key.offset, 0, group->key.offset,
5508 0, 0, 0);
5509 BUG_ON(err);
5510
5511 inode = btrfs_iget_locked(root->fs_info->sb, objectid, root);
5512 if (inode->i_state & I_NEW) {
5513 BTRFS_I(inode)->root = root;
5514 BTRFS_I(inode)->location.objectid = objectid;
5515 BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY;
5516 BTRFS_I(inode)->location.offset = 0;
5517 btrfs_read_locked_inode(inode);
5518 unlock_new_inode(inode);
5519 BUG_ON(is_bad_inode(inode));
5520 } else {
5521 BUG_ON(1);
5522 }
5523 BTRFS_I(inode)->index_cnt = group->key.objectid;
5524
5525 err = btrfs_orphan_add(trans, inode);
5526 out:
5527 btrfs_end_transaction(trans, root);
5528 if (err) {
5529 if (inode)
5530 iput(inode);
5531 inode = ERR_PTR(err);
5532 }
5533 return inode;
5534 }
5535
5536 int btrfs_reloc_clone_csums(struct inode *inode, u64 file_pos, u64 len)
5537 {
5538
5539 struct btrfs_ordered_sum *sums;
5540 struct btrfs_sector_sum *sector_sum;
5541 struct btrfs_ordered_extent *ordered;
5542 struct btrfs_root *root = BTRFS_I(inode)->root;
5543 struct list_head list;
5544 size_t offset;
5545 int ret;
5546 u64 disk_bytenr;
5547
5548 INIT_LIST_HEAD(&list);
5549
5550 ordered = btrfs_lookup_ordered_extent(inode, file_pos);
5551 BUG_ON(ordered->file_offset != file_pos || ordered->len != len);
5552
5553 disk_bytenr = file_pos + BTRFS_I(inode)->index_cnt;
5554 ret = btrfs_lookup_csums_range(root->fs_info->csum_root, disk_bytenr,
5555 disk_bytenr + len - 1, &list);
5556
5557 while (!list_empty(&list)) {
5558 sums = list_entry(list.next, struct btrfs_ordered_sum, list);
5559 list_del_init(&sums->list);
5560
5561 sector_sum = sums->sums;
5562 sums->bytenr = ordered->start;
5563
5564 offset = 0;
5565 while (offset < sums->len) {
5566 sector_sum->bytenr += ordered->start - disk_bytenr;
5567 sector_sum++;
5568 offset += root->sectorsize;
5569 }
5570
5571 btrfs_add_ordered_sum(inode, ordered, sums);
5572 }
5573 btrfs_put_ordered_extent(ordered);
5574 return 0;
5575 }
5576
5577 int btrfs_relocate_block_group(struct btrfs_root *root, u64 group_start)
5578 {
5579 struct btrfs_trans_handle *trans;
5580 struct btrfs_path *path;
5581 struct btrfs_fs_info *info = root->fs_info;
5582 struct extent_buffer *leaf;
5583 struct inode *reloc_inode;
5584 struct btrfs_block_group_cache *block_group;
5585 struct btrfs_key key;
5586 u64 skipped;
5587 u64 cur_byte;
5588 u64 total_found;
5589 u32 nritems;
5590 int ret;
5591 int progress;
5592 int pass = 0;
5593
5594 root = root->fs_info->extent_root;
5595
5596 block_group = btrfs_lookup_block_group(info, group_start);
5597 BUG_ON(!block_group);
5598
5599 printk(KERN_INFO "btrfs relocating block group %llu flags %llu\n",
5600 (unsigned long long)block_group->key.objectid,
5601 (unsigned long long)block_group->flags);
5602
5603 path = btrfs_alloc_path();
5604 BUG_ON(!path);
5605
5606 reloc_inode = create_reloc_inode(info, block_group);
5607 BUG_ON(IS_ERR(reloc_inode));
5608
5609 __alloc_chunk_for_shrink(root, block_group, 1);
5610 set_block_group_readonly(block_group);
5611
5612 btrfs_start_delalloc_inodes(info->tree_root);
5613 btrfs_wait_ordered_extents(info->tree_root, 0);
5614 again:
5615 skipped = 0;
5616 total_found = 0;
5617 progress = 0;
5618 key.objectid = block_group->key.objectid;
5619 key.offset = 0;
5620 key.type = 0;
5621 cur_byte = key.objectid;
5622
5623 trans = btrfs_start_transaction(info->tree_root, 1);
5624 btrfs_commit_transaction(trans, info->tree_root);
5625
5626 mutex_lock(&root->fs_info->cleaner_mutex);
5627 btrfs_clean_old_snapshots(info->tree_root);
5628 btrfs_remove_leaf_refs(info->tree_root, (u64)-1, 1);
5629 mutex_unlock(&root->fs_info->cleaner_mutex);
5630
5631 trans = btrfs_start_transaction(info->tree_root, 1);
5632 btrfs_commit_transaction(trans, info->tree_root);
5633
5634 while (1) {
5635 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
5636 if (ret < 0)
5637 goto out;
5638 next:
5639 leaf = path->nodes[0];
5640 nritems = btrfs_header_nritems(leaf);
5641 if (path->slots[0] >= nritems) {
5642 ret = btrfs_next_leaf(root, path);
5643 if (ret < 0)
5644 goto out;
5645 if (ret == 1) {
5646 ret = 0;
5647 break;
5648 }
5649 leaf = path->nodes[0];
5650 nritems = btrfs_header_nritems(leaf);
5651 }
5652
5653 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
5654
5655 if (key.objectid >= block_group->key.objectid +
5656 block_group->key.offset)
5657 break;
5658
5659 if (progress && need_resched()) {
5660 btrfs_release_path(root, path);
5661 cond_resched();
5662 progress = 0;
5663 continue;
5664 }
5665 progress = 1;
5666
5667 if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY ||
5668 key.objectid + key.offset <= cur_byte) {
5669 path->slots[0]++;
5670 goto next;
5671 }
5672
5673 total_found++;
5674 cur_byte = key.objectid + key.offset;
5675 btrfs_release_path(root, path);
5676
5677 __alloc_chunk_for_shrink(root, block_group, 0);
5678 ret = relocate_one_extent(root, path, &key, block_group,
5679 reloc_inode, pass);
5680 BUG_ON(ret < 0);
5681 if (ret > 0)
5682 skipped++;
5683
5684 key.objectid = cur_byte;
5685 key.type = 0;
5686 key.offset = 0;
5687 }
5688
5689 btrfs_release_path(root, path);
5690
5691 if (pass == 0) {
5692 btrfs_wait_ordered_range(reloc_inode, 0, (u64)-1);
5693 invalidate_mapping_pages(reloc_inode->i_mapping, 0, -1);
5694 }
5695
5696 if (total_found > 0) {
5697 printk(KERN_INFO "btrfs found %llu extents in pass %d\n",
5698 (unsigned long long)total_found, pass);
5699 pass++;
5700 if (total_found == skipped && pass > 2) {
5701 iput(reloc_inode);
5702 reloc_inode = create_reloc_inode(info, block_group);
5703 pass = 0;
5704 }
5705 goto again;
5706 }
5707
5708 /* delete reloc_inode */
5709 iput(reloc_inode);
5710
5711 /* unpin extents in this range */
5712 trans = btrfs_start_transaction(info->tree_root, 1);
5713 btrfs_commit_transaction(trans, info->tree_root);
5714
5715 spin_lock(&block_group->lock);
5716 WARN_ON(block_group->pinned > 0);
5717 WARN_ON(block_group->reserved > 0);
5718 WARN_ON(btrfs_block_group_used(&block_group->item) > 0);
5719 spin_unlock(&block_group->lock);
5720 btrfs_put_block_group(block_group);
5721 ret = 0;
5722 out:
5723 btrfs_free_path(path);
5724 return ret;
5725 }
5726
5727 static int find_first_block_group(struct btrfs_root *root,
5728 struct btrfs_path *path, struct btrfs_key *key)
5729 {
5730 int ret = 0;
5731 struct btrfs_key found_key;
5732 struct extent_buffer *leaf;
5733 int slot;
5734
5735 ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
5736 if (ret < 0)
5737 goto out;
5738
5739 while (1) {
5740 slot = path->slots[0];
5741 leaf = path->nodes[0];
5742 if (slot >= btrfs_header_nritems(leaf)) {
5743 ret = btrfs_next_leaf(root, path);
5744 if (ret == 0)
5745 continue;
5746 if (ret < 0)
5747 goto out;
5748 break;
5749 }
5750 btrfs_item_key_to_cpu(leaf, &found_key, slot);
5751
5752 if (found_key.objectid >= key->objectid &&
5753 found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) {
5754 ret = 0;
5755 goto out;
5756 }
5757 path->slots[0]++;
5758 }
5759 ret = -ENOENT;
5760 out:
5761 return ret;
5762 }
5763
5764 int btrfs_free_block_groups(struct btrfs_fs_info *info)
5765 {
5766 struct btrfs_block_group_cache *block_group;
5767 struct btrfs_space_info *space_info;
5768 struct rb_node *n;
5769
5770 spin_lock(&info->block_group_cache_lock);
5771 while ((n = rb_last(&info->block_group_cache_tree)) != NULL) {
5772 block_group = rb_entry(n, struct btrfs_block_group_cache,
5773 cache_node);
5774 rb_erase(&block_group->cache_node,
5775 &info->block_group_cache_tree);
5776 spin_unlock(&info->block_group_cache_lock);
5777
5778 btrfs_remove_free_space_cache(block_group);
5779 down_write(&block_group->space_info->groups_sem);
5780 list_del(&block_group->list);
5781 up_write(&block_group->space_info->groups_sem);
5782
5783 WARN_ON(atomic_read(&block_group->count) != 1);
5784 kfree(block_group);
5785
5786 spin_lock(&info->block_group_cache_lock);
5787 }
5788 spin_unlock(&info->block_group_cache_lock);
5789
5790 /* now that all the block groups are freed, go through and
5791 * free all the space_info structs. This is only called during
5792 * the final stages of unmount, and so we know nobody is
5793 * using them. We call synchronize_rcu() once before we start,
5794 * just to be on the safe side.
5795 */
5796 synchronize_rcu();
5797
5798 while(!list_empty(&info->space_info)) {
5799 space_info = list_entry(info->space_info.next,
5800 struct btrfs_space_info,
5801 list);
5802
5803 list_del(&space_info->list);
5804 kfree(space_info);
5805 }
5806 return 0;
5807 }
5808
5809 int btrfs_read_block_groups(struct btrfs_root *root)
5810 {
5811 struct btrfs_path *path;
5812 int ret;
5813 struct btrfs_block_group_cache *cache;
5814 struct btrfs_fs_info *info = root->fs_info;
5815 struct btrfs_space_info *space_info;
5816 struct btrfs_key key;
5817 struct btrfs_key found_key;
5818 struct extent_buffer *leaf;
5819
5820 root = info->extent_root;
5821 key.objectid = 0;
5822 key.offset = 0;
5823 btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
5824 path = btrfs_alloc_path();
5825 if (!path)
5826 return -ENOMEM;
5827
5828 while (1) {
5829 ret = find_first_block_group(root, path, &key);
5830 if (ret > 0) {
5831 ret = 0;
5832 goto error;
5833 }
5834 if (ret != 0)
5835 goto error;
5836
5837 leaf = path->nodes[0];
5838 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
5839 cache = kzalloc(sizeof(*cache), GFP_NOFS);
5840 if (!cache) {
5841 ret = -ENOMEM;
5842 break;
5843 }
5844
5845 atomic_set(&cache->count, 1);
5846 spin_lock_init(&cache->lock);
5847 spin_lock_init(&cache->tree_lock);
5848 mutex_init(&cache->cache_mutex);
5849 INIT_LIST_HEAD(&cache->list);
5850 INIT_LIST_HEAD(&cache->cluster_list);
5851 read_extent_buffer(leaf, &cache->item,
5852 btrfs_item_ptr_offset(leaf, path->slots[0]),
5853 sizeof(cache->item));
5854 memcpy(&cache->key, &found_key, sizeof(found_key));
5855
5856 key.objectid = found_key.objectid + found_key.offset;
5857 btrfs_release_path(root, path);
5858 cache->flags = btrfs_block_group_flags(&cache->item);
5859
5860 ret = update_space_info(info, cache->flags, found_key.offset,
5861 btrfs_block_group_used(&cache->item),
5862 &space_info);
5863 BUG_ON(ret);
5864 cache->space_info = space_info;
5865 down_write(&space_info->groups_sem);
5866 list_add_tail(&cache->list, &space_info->block_groups);
5867 up_write(&space_info->groups_sem);
5868
5869 ret = btrfs_add_block_group_cache(root->fs_info, cache);
5870 BUG_ON(ret);
5871
5872 set_avail_alloc_bits(root->fs_info, cache->flags);
5873 if (btrfs_chunk_readonly(root, cache->key.objectid))
5874 set_block_group_readonly(cache);
5875 }
5876 ret = 0;
5877 error:
5878 btrfs_free_path(path);
5879 return ret;
5880 }
5881
5882 int btrfs_make_block_group(struct btrfs_trans_handle *trans,
5883 struct btrfs_root *root, u64 bytes_used,
5884 u64 type, u64 chunk_objectid, u64 chunk_offset,
5885 u64 size)
5886 {
5887 int ret;
5888 struct btrfs_root *extent_root;
5889 struct btrfs_block_group_cache *cache;
5890
5891 extent_root = root->fs_info->extent_root;
5892
5893 root->fs_info->last_trans_log_full_commit = trans->transid;
5894
5895 cache = kzalloc(sizeof(*cache), GFP_NOFS);
5896 if (!cache)
5897 return -ENOMEM;
5898
5899 cache->key.objectid = chunk_offset;
5900 cache->key.offset = size;
5901 cache->key.type = BTRFS_BLOCK_GROUP_ITEM_KEY;
5902 atomic_set(&cache->count, 1);
5903 spin_lock_init(&cache->lock);
5904 spin_lock_init(&cache->tree_lock);
5905 mutex_init(&cache->cache_mutex);
5906 INIT_LIST_HEAD(&cache->list);
5907 INIT_LIST_HEAD(&cache->cluster_list);
5908
5909 btrfs_set_block_group_used(&cache->item, bytes_used);
5910 btrfs_set_block_group_chunk_objectid(&cache->item, chunk_objectid);
5911 cache->flags = type;
5912 btrfs_set_block_group_flags(&cache->item, type);
5913
5914 ret = update_space_info(root->fs_info, cache->flags, size, bytes_used,
5915 &cache->space_info);
5916 BUG_ON(ret);
5917 down_write(&cache->space_info->groups_sem);
5918 list_add_tail(&cache->list, &cache->space_info->block_groups);
5919 up_write(&cache->space_info->groups_sem);
5920
5921 ret = btrfs_add_block_group_cache(root->fs_info, cache);
5922 BUG_ON(ret);
5923
5924 ret = btrfs_insert_item(trans, extent_root, &cache->key, &cache->item,
5925 sizeof(cache->item));
5926 BUG_ON(ret);
5927
5928 set_avail_alloc_bits(extent_root->fs_info, type);
5929
5930 return 0;
5931 }
5932
5933 int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
5934 struct btrfs_root *root, u64 group_start)
5935 {
5936 struct btrfs_path *path;
5937 struct btrfs_block_group_cache *block_group;
5938 struct btrfs_key key;
5939 int ret;
5940
5941 root = root->fs_info->extent_root;
5942
5943 block_group = btrfs_lookup_block_group(root->fs_info, group_start);
5944 BUG_ON(!block_group);
5945 BUG_ON(!block_group->ro);
5946
5947 memcpy(&key, &block_group->key, sizeof(key));
5948
5949 path = btrfs_alloc_path();
5950 BUG_ON(!path);
5951
5952 spin_lock(&root->fs_info->block_group_cache_lock);
5953 rb_erase(&block_group->cache_node,
5954 &root->fs_info->block_group_cache_tree);
5955 spin_unlock(&root->fs_info->block_group_cache_lock);
5956 btrfs_remove_free_space_cache(block_group);
5957 down_write(&block_group->space_info->groups_sem);
5958 list_del(&block_group->list);
5959 up_write(&block_group->space_info->groups_sem);
5960
5961 spin_lock(&block_group->space_info->lock);
5962 block_group->space_info->total_bytes -= block_group->key.offset;
5963 block_group->space_info->bytes_readonly -= block_group->key.offset;
5964 spin_unlock(&block_group->space_info->lock);
5965 block_group->space_info->full = 0;
5966
5967 btrfs_put_block_group(block_group);
5968 btrfs_put_block_group(block_group);
5969
5970 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
5971 if (ret > 0)
5972 ret = -EIO;
5973 if (ret < 0)
5974 goto out;
5975
5976 ret = btrfs_del_item(trans, root, path);
5977 out:
5978 btrfs_free_path(path);
5979 return ret;
5980 }
kernel-based virtual machine