?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
regulator: simplify regulator_register() error handling
[linux/fpc-iii.git] / fs / btrfs / extent-tree.c
blobc6a4f459ad76d87cf5a7d5295ad518ad2010c38f
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 <linux/kthread.h>
25 #include <linux/slab.h>
26 #include "compat.h"
27 #include "hash.h"
28 #include "ctree.h"
29 #include "disk-io.h"
30 #include "print-tree.h"
31 #include "transaction.h"
32 #include "volumes.h"
33 #include "locking.h"
34 #include "free-space-cache.h"
35
36 static int update_block_group(struct btrfs_trans_handle *trans,
37 struct btrfs_root *root,
38 u64 bytenr, u64 num_bytes, int alloc,
39 int mark_free);
40 static int update_reserved_extents(struct btrfs_block_group_cache *cache,
41 u64 num_bytes, int reserve);
42 static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
43 struct btrfs_root *root,
44 u64 bytenr, u64 num_bytes, u64 parent,
45 u64 root_objectid, u64 owner_objectid,
46 u64 owner_offset, int refs_to_drop,
47 struct btrfs_delayed_extent_op *extra_op);
48 static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
49 struct extent_buffer *leaf,
50 struct btrfs_extent_item *ei);
51 static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
52 struct btrfs_root *root,
53 u64 parent, u64 root_objectid,
54 u64 flags, u64 owner, u64 offset,
55 struct btrfs_key *ins, int ref_mod);
56 static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
57 struct btrfs_root *root,
58 u64 parent, u64 root_objectid,
59 u64 flags, struct btrfs_disk_key *key,
60 int level, struct btrfs_key *ins);
61 static int do_chunk_alloc(struct btrfs_trans_handle *trans,
62 struct btrfs_root *extent_root, u64 alloc_bytes,
63 u64 flags, int force);
64 static int pin_down_bytes(struct btrfs_trans_handle *trans,
65 struct btrfs_root *root,
66 struct btrfs_path *path,
67 u64 bytenr, u64 num_bytes,
68 int is_data, int reserved,
69 struct extent_buffer **must_clean);
70 static int find_next_key(struct btrfs_path *path, int level,
71 struct btrfs_key *key);
72 static void dump_space_info(struct btrfs_space_info *info, u64 bytes,
73 int dump_block_groups);
74
75 static noinline int
76 block_group_cache_done(struct btrfs_block_group_cache *cache)
77 {
78 smp_mb();
79 return cache->cached == BTRFS_CACHE_FINISHED;
80 }
81
82 static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits)
83 {
84 return (cache->flags & bits) == bits;
85 }
86
87 void btrfs_get_block_group(struct btrfs_block_group_cache *cache)
88 {
89 atomic_inc(&cache->count);
90 }
91
92 void btrfs_put_block_group(struct btrfs_block_group_cache *cache)
93 {
94 if (atomic_dec_and_test(&cache->count))
95 kfree(cache);
96 }
97
98 /*
99 * this adds the block group to the fs_info rb tree for the block group
100 * cache
101 */
102 static int btrfs_add_block_group_cache(struct btrfs_fs_info *info,
103 struct btrfs_block_group_cache *block_group)
104 {
105 struct rb_node **p;
106 struct rb_node *parent = NULL;
107 struct btrfs_block_group_cache *cache;
108
109 spin_lock(&info->block_group_cache_lock);
110 p = &info->block_group_cache_tree.rb_node;
111
112 while (*p) {
113 parent = *p;
114 cache = rb_entry(parent, struct btrfs_block_group_cache,
115 cache_node);
116 if (block_group->key.objectid < cache->key.objectid) {
117 p = &(*p)->rb_left;
118 } else if (block_group->key.objectid > cache->key.objectid) {
119 p = &(*p)->rb_right;
120 } else {
121 spin_unlock(&info->block_group_cache_lock);
122 return -EEXIST;
123 }
124 }
125
126 rb_link_node(&block_group->cache_node, parent, p);
127 rb_insert_color(&block_group->cache_node,
128 &info->block_group_cache_tree);
129 spin_unlock(&info->block_group_cache_lock);
130
131 return 0;
132 }
133
134 /*
135 * This will return the block group at or after bytenr if contains is 0, else
136 * it will return the block group that contains the bytenr
137 */
138 static struct btrfs_block_group_cache *
139 block_group_cache_tree_search(struct btrfs_fs_info *info, u64 bytenr,
140 int contains)
141 {
142 struct btrfs_block_group_cache *cache, *ret = NULL;
143 struct rb_node *n;
144 u64 end, start;
145
146 spin_lock(&info->block_group_cache_lock);
147 n = info->block_group_cache_tree.rb_node;
148
149 while (n) {
150 cache = rb_entry(n, struct btrfs_block_group_cache,
151 cache_node);
152 end = cache->key.objectid + cache->key.offset - 1;
153 start = cache->key.objectid;
154
155 if (bytenr < start) {
156 if (!contains && (!ret || start < ret->key.objectid))
157 ret = cache;
158 n = n->rb_left;
159 } else if (bytenr > start) {
160 if (contains && bytenr <= end) {
161 ret = cache;
162 break;
163 }
164 n = n->rb_right;
165 } else {
166 ret = cache;
167 break;
168 }
169 }
170 if (ret)
171 btrfs_get_block_group(ret);
172 spin_unlock(&info->block_group_cache_lock);
173
174 return ret;
175 }
176
177 static int add_excluded_extent(struct btrfs_root *root,
178 u64 start, u64 num_bytes)
179 {
180 u64 end = start + num_bytes - 1;
181 set_extent_bits(&root->fs_info->freed_extents[0],
182 start, end, EXTENT_UPTODATE, GFP_NOFS);
183 set_extent_bits(&root->fs_info->freed_extents[1],
184 start, end, EXTENT_UPTODATE, GFP_NOFS);
185 return 0;
186 }
187
188 static void free_excluded_extents(struct btrfs_root *root,
189 struct btrfs_block_group_cache *cache)
190 {
191 u64 start, end;
192
193 start = cache->key.objectid;
194 end = start + cache->key.offset - 1;
195
196 clear_extent_bits(&root->fs_info->freed_extents[0],
197 start, end, EXTENT_UPTODATE, GFP_NOFS);
198 clear_extent_bits(&root->fs_info->freed_extents[1],
199 start, end, EXTENT_UPTODATE, GFP_NOFS);
200 }
201
202 static int exclude_super_stripes(struct btrfs_root *root,
203 struct btrfs_block_group_cache *cache)
204 {
205 u64 bytenr;
206 u64 *logical;
207 int stripe_len;
208 int i, nr, ret;
209
210 if (cache->key.objectid < BTRFS_SUPER_INFO_OFFSET) {
211 stripe_len = BTRFS_SUPER_INFO_OFFSET - cache->key.objectid;
212 cache->bytes_super += stripe_len;
213 ret = add_excluded_extent(root, cache->key.objectid,
214 stripe_len);
215 BUG_ON(ret);
216 }
217
218 for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
219 bytenr = btrfs_sb_offset(i);
220 ret = btrfs_rmap_block(&root->fs_info->mapping_tree,
221 cache->key.objectid, bytenr,
222 0, &logical, &nr, &stripe_len);
223 BUG_ON(ret);
224
225 while (nr--) {
226 cache->bytes_super += stripe_len;
227 ret = add_excluded_extent(root, logical[nr],
228 stripe_len);
229 BUG_ON(ret);
230 }
231
232 kfree(logical);
233 }
234 return 0;
235 }
236
237 static struct btrfs_caching_control *
238 get_caching_control(struct btrfs_block_group_cache *cache)
239 {
240 struct btrfs_caching_control *ctl;
241
242 spin_lock(&cache->lock);
243 if (cache->cached != BTRFS_CACHE_STARTED) {
244 spin_unlock(&cache->lock);
245 return NULL;
246 }
247
248 ctl = cache->caching_ctl;
249 atomic_inc(&ctl->count);
250 spin_unlock(&cache->lock);
251 return ctl;
252 }
253
254 static void put_caching_control(struct btrfs_caching_control *ctl)
255 {
256 if (atomic_dec_and_test(&ctl->count))
257 kfree(ctl);
258 }
259
260 /*
261 * this is only called by cache_block_group, since we could have freed extents
262 * we need to check the pinned_extents for any extents that can't be used yet
263 * since their free space will be released as soon as the transaction commits.
264 */
265 static u64 add_new_free_space(struct btrfs_block_group_cache *block_group,
266 struct btrfs_fs_info *info, u64 start, u64 end)
267 {
268 u64 extent_start, extent_end, size, total_added = 0;
269 int ret;
270
271 while (start < end) {
272 ret = find_first_extent_bit(info->pinned_extents, start,
273 &extent_start, &extent_end,
274 EXTENT_DIRTY | EXTENT_UPTODATE);
275 if (ret)
276 break;
277
278 if (extent_start <= start) {
279 start = extent_end + 1;
280 } else if (extent_start > start && extent_start < end) {
281 size = extent_start - start;
282 total_added += size;
283 ret = btrfs_add_free_space(block_group, start,
284 size);
285 BUG_ON(ret);
286 start = extent_end + 1;
287 } else {
288 break;
289 }
290 }
291
292 if (start < end) {
293 size = end - start;
294 total_added += size;
295 ret = btrfs_add_free_space(block_group, start, size);
296 BUG_ON(ret);
297 }
298
299 return total_added;
300 }
301
302 static int caching_kthread(void *data)
303 {
304 struct btrfs_block_group_cache *block_group = data;
305 struct btrfs_fs_info *fs_info = block_group->fs_info;
306 struct btrfs_caching_control *caching_ctl = block_group->caching_ctl;
307 struct btrfs_root *extent_root = fs_info->extent_root;
308 struct btrfs_path *path;
309 struct extent_buffer *leaf;
310 struct btrfs_key key;
311 u64 total_found = 0;
312 u64 last = 0;
313 u32 nritems;
314 int ret = 0;
315
316 path = btrfs_alloc_path();
317 if (!path)
318 return -ENOMEM;
319
320 exclude_super_stripes(extent_root, block_group);
321 spin_lock(&block_group->space_info->lock);
322 block_group->space_info->bytes_super += block_group->bytes_super;
323 spin_unlock(&block_group->space_info->lock);
324
325 last = max_t(u64, block_group->key.objectid, BTRFS_SUPER_INFO_OFFSET);
326
327 /*
328 * We don't want to deadlock with somebody trying to allocate a new
329 * extent for the extent root while also trying to search the extent
330 * root to add free space. So we skip locking and search the commit
331 * root, since its read-only
332 */
333 path->skip_locking = 1;
334 path->search_commit_root = 1;
335 path->reada = 2;
336
337 key.objectid = last;
338 key.offset = 0;
339 key.type = BTRFS_EXTENT_ITEM_KEY;
340 again:
341 mutex_lock(&caching_ctl->mutex);
342 /* need to make sure the commit_root doesn't disappear */
343 down_read(&fs_info->extent_commit_sem);
344
345 ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
346 if (ret < 0)
347 goto err;
348
349 leaf = path->nodes[0];
350 nritems = btrfs_header_nritems(leaf);
351
352 while (1) {
353 smp_mb();
354 if (fs_info->closing > 1) {
355 last = (u64)-1;
356 break;
357 }
358
359 if (path->slots[0] < nritems) {
360 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
361 } else {
362 ret = find_next_key(path, 0, &key);
363 if (ret)
364 break;
365
366 caching_ctl->progress = last;
367 btrfs_release_path(extent_root, path);
368 up_read(&fs_info->extent_commit_sem);
369 mutex_unlock(&caching_ctl->mutex);
370 if (btrfs_transaction_in_commit(fs_info))
371 schedule_timeout(1);
372 else
373 cond_resched();
374 goto again;
375 }
376
377 if (key.objectid < block_group->key.objectid) {
378 path->slots[0]++;
379 continue;
380 }
381
382 if (key.objectid >= block_group->key.objectid +
383 block_group->key.offset)
384 break;
385
386 if (key.type == BTRFS_EXTENT_ITEM_KEY) {
387 total_found += add_new_free_space(block_group,
388 fs_info, last,
389 key.objectid);
390 last = key.objectid + key.offset;
391
392 if (total_found > (1024 * 1024 * 2)) {
393 total_found = 0;
394 wake_up(&caching_ctl->wait);
395 }
396 }
397 path->slots[0]++;
398 }
399 ret = 0;
400
401 total_found += add_new_free_space(block_group, fs_info, last,
402 block_group->key.objectid +
403 block_group->key.offset);
404 caching_ctl->progress = (u64)-1;
405
406 spin_lock(&block_group->lock);
407 block_group->caching_ctl = NULL;
408 block_group->cached = BTRFS_CACHE_FINISHED;
409 spin_unlock(&block_group->lock);
410
411 err:
412 btrfs_free_path(path);
413 up_read(&fs_info->extent_commit_sem);
414
415 free_excluded_extents(extent_root, block_group);
416
417 mutex_unlock(&caching_ctl->mutex);
418 wake_up(&caching_ctl->wait);
419
420 put_caching_control(caching_ctl);
421 atomic_dec(&block_group->space_info->caching_threads);
422 btrfs_put_block_group(block_group);
423
424 return 0;
425 }
426
427 static int cache_block_group(struct btrfs_block_group_cache *cache)
428 {
429 struct btrfs_fs_info *fs_info = cache->fs_info;
430 struct btrfs_caching_control *caching_ctl;
431 struct task_struct *tsk;
432 int ret = 0;
433
434 smp_mb();
435 if (cache->cached != BTRFS_CACHE_NO)
436 return 0;
437
438 caching_ctl = kzalloc(sizeof(*caching_ctl), GFP_KERNEL);
439 BUG_ON(!caching_ctl);
440
441 INIT_LIST_HEAD(&caching_ctl->list);
442 mutex_init(&caching_ctl->mutex);
443 init_waitqueue_head(&caching_ctl->wait);
444 caching_ctl->block_group = cache;
445 caching_ctl->progress = cache->key.objectid;
446 /* one for caching kthread, one for caching block group list */
447 atomic_set(&caching_ctl->count, 2);
448
449 spin_lock(&cache->lock);
450 if (cache->cached != BTRFS_CACHE_NO) {
451 spin_unlock(&cache->lock);
452 kfree(caching_ctl);
453 return 0;
454 }
455 cache->caching_ctl = caching_ctl;
456 cache->cached = BTRFS_CACHE_STARTED;
457 spin_unlock(&cache->lock);
458
459 down_write(&fs_info->extent_commit_sem);
460 list_add_tail(&caching_ctl->list, &fs_info->caching_block_groups);
461 up_write(&fs_info->extent_commit_sem);
462
463 atomic_inc(&cache->space_info->caching_threads);
464 btrfs_get_block_group(cache);
465
466 tsk = kthread_run(caching_kthread, cache, "btrfs-cache-%llu\n",
467 cache->key.objectid);
468 if (IS_ERR(tsk)) {
469 ret = PTR_ERR(tsk);
470 printk(KERN_ERR "error running thread %d\n", ret);
471 BUG();
472 }
473
474 return ret;
475 }
476
477 /*
478 * return the block group that starts at or after bytenr
479 */
480 static struct btrfs_block_group_cache *
481 btrfs_lookup_first_block_group(struct btrfs_fs_info *info, u64 bytenr)
482 {
483 struct btrfs_block_group_cache *cache;
484
485 cache = block_group_cache_tree_search(info, bytenr, 0);
486
487 return cache;
488 }
489
490 /*
491 * return the block group that contains the given bytenr
492 */
493 struct btrfs_block_group_cache *btrfs_lookup_block_group(
494 struct btrfs_fs_info *info,
495 u64 bytenr)
496 {
497 struct btrfs_block_group_cache *cache;
498
499 cache = block_group_cache_tree_search(info, bytenr, 1);
500
501 return cache;
502 }
503
504 static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info,
505 u64 flags)
506 {
507 struct list_head *head = &info->space_info;
508 struct btrfs_space_info *found;
509
510 rcu_read_lock();
511 list_for_each_entry_rcu(found, head, list) {
512 if (found->flags == flags) {
513 rcu_read_unlock();
514 return found;
515 }
516 }
517 rcu_read_unlock();
518 return NULL;
519 }
520
521 /*
522 * after adding space to the filesystem, we need to clear the full flags
523 * on all the space infos.
524 */
525 void btrfs_clear_space_info_full(struct btrfs_fs_info *info)
526 {
527 struct list_head *head = &info->space_info;
528 struct btrfs_space_info *found;
529
530 rcu_read_lock();
531 list_for_each_entry_rcu(found, head, list)
532 found->full = 0;
533 rcu_read_unlock();
534 }
535
536 static u64 div_factor(u64 num, int factor)
537 {
538 if (factor == 10)
539 return num;
540 num *= factor;
541 do_div(num, 10);
542 return num;
543 }
544
545 u64 btrfs_find_block_group(struct btrfs_root *root,
546 u64 search_start, u64 search_hint, int owner)
547 {
548 struct btrfs_block_group_cache *cache;
549 u64 used;
550 u64 last = max(search_hint, search_start);
551 u64 group_start = 0;
552 int full_search = 0;
553 int factor = 9;
554 int wrapped = 0;
555 again:
556 while (1) {
557 cache = btrfs_lookup_first_block_group(root->fs_info, last);
558 if (!cache)
559 break;
560
561 spin_lock(&cache->lock);
562 last = cache->key.objectid + cache->key.offset;
563 used = btrfs_block_group_used(&cache->item);
564
565 if ((full_search || !cache->ro) &&
566 block_group_bits(cache, BTRFS_BLOCK_GROUP_METADATA)) {
567 if (used + cache->pinned + cache->reserved <
568 div_factor(cache->key.offset, factor)) {
569 group_start = cache->key.objectid;
570 spin_unlock(&cache->lock);
571 btrfs_put_block_group(cache);
572 goto found;
573 }
574 }
575 spin_unlock(&cache->lock);
576 btrfs_put_block_group(cache);
577 cond_resched();
578 }
579 if (!wrapped) {
580 last = search_start;
581 wrapped = 1;
582 goto again;
583 }
584 if (!full_search && factor < 10) {
585 last = search_start;
586 full_search = 1;
587 factor = 10;
588 goto again;
589 }
590 found:
591 return group_start;
592 }
593
594 /* simple helper to search for an existing extent at a given offset */
595 int btrfs_lookup_extent(struct btrfs_root *root, u64 start, u64 len)
596 {
597 int ret;
598 struct btrfs_key key;
599 struct btrfs_path *path;
600
601 path = btrfs_alloc_path();
602 BUG_ON(!path);
603 key.objectid = start;
604 key.offset = len;
605 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
606 ret = btrfs_search_slot(NULL, root->fs_info->extent_root, &key, path,
607 0, 0);
608 btrfs_free_path(path);
609 return ret;
610 }
611
612 /*
613 * Back reference rules. Back refs have three main goals:
614 *
615 * 1) differentiate between all holders of references to an extent so that
616 * when a reference is dropped we can make sure it was a valid reference
617 * before freeing the extent.
618 *
619 * 2) Provide enough information to quickly find the holders of an extent
620 * if we notice a given block is corrupted or bad.
621 *
622 * 3) Make it easy to migrate blocks for FS shrinking or storage pool
623 * maintenance. This is actually the same as #2, but with a slightly
624 * different use case.
625 *
626 * There are two kinds of back refs. The implicit back refs is optimized
627 * for pointers in non-shared tree blocks. For a given pointer in a block,
628 * back refs of this kind provide information about the block's owner tree
629 * and the pointer's key. These information allow us to find the block by
630 * b-tree searching. The full back refs is for pointers in tree blocks not
631 * referenced by their owner trees. The location of tree block is recorded
632 * in the back refs. Actually the full back refs is generic, and can be
633 * used in all cases the implicit back refs is used. The major shortcoming
634 * of the full back refs is its overhead. Every time a tree block gets
635 * COWed, we have to update back refs entry for all pointers in it.
636 *
637 * For a newly allocated tree block, we use implicit back refs for
638 * pointers in it. This means most tree related operations only involve
639 * implicit back refs. For a tree block created in old transaction, the
640 * only way to drop a reference to it is COW it. So we can detect the
641 * event that tree block loses its owner tree's reference and do the
642 * back refs conversion.
643 *
644 * When a tree block is COW'd through a tree, there are four cases:
645 *
646 * The reference count of the block is one and the tree is the block's
647 * owner tree. Nothing to do in this case.
648 *
649 * The reference count of the block is one and the tree is not the
650 * block's owner tree. In this case, full back refs is used for pointers
651 * in the block. Remove these full back refs, add implicit back refs for
652 * every pointers in the new block.
653 *
654 * The reference count of the block is greater than one and the tree is
655 * the block's owner tree. In this case, implicit back refs is used for
656 * pointers in the block. Add full back refs for every pointers in the
657 * block, increase lower level extents' reference counts. The original
658 * implicit back refs are entailed to the new block.
659 *
660 * The reference count of the block is greater than one and the tree is
661 * not the block's owner tree. Add implicit back refs for every pointer in
662 * the new block, increase lower level extents' reference count.
663 *
664 * Back Reference Key composing:
665 *
666 * The key objectid corresponds to the first byte in the extent,
667 * The key type is used to differentiate between types of back refs.
668 * There are different meanings of the key offset for different types
669 * of back refs.
670 *
671 * File extents can be referenced by:
672 *
673 * - multiple snapshots, subvolumes, or different generations in one subvol
674 * - different files inside a single subvolume
675 * - different offsets inside a file (bookend extents in file.c)
676 *
677 * The extent ref structure for the implicit back refs has fields for:
678 *
679 * - Objectid of the subvolume root
680 * - objectid of the file holding the reference
681 * - original offset in the file
682 * - how many bookend extents
683 *
684 * The key offset for the implicit back refs is hash of the first
685 * three fields.
686 *
687 * The extent ref structure for the full back refs has field for:
688 *
689 * - number of pointers in the tree leaf
690 *
691 * The key offset for the implicit back refs is the first byte of
692 * the tree leaf
693 *
694 * When a file extent is allocated, The implicit back refs is used.
695 * the fields are filled in:
696 *
697 * (root_key.objectid, inode objectid, offset in file, 1)
698 *
699 * When a file extent is removed file truncation, we find the
700 * corresponding implicit back refs and check the following fields:
701 *
702 * (btrfs_header_owner(leaf), inode objectid, offset in file)
703 *
704 * Btree extents can be referenced by:
705 *
706 * - Different subvolumes
707 *
708 * Both the implicit back refs and the full back refs for tree blocks
709 * only consist of key. The key offset for the implicit back refs is
710 * objectid of block's owner tree. The key offset for the full back refs
711 * is the first byte of parent block.
712 *
713 * When implicit back refs is used, information about the lowest key and
714 * level of the tree block are required. These information are stored in
715 * tree block info structure.
716 */
717
718 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
719 static int convert_extent_item_v0(struct btrfs_trans_handle *trans,
720 struct btrfs_root *root,
721 struct btrfs_path *path,
722 u64 owner, u32 extra_size)
723 {
724 struct btrfs_extent_item *item;
725 struct btrfs_extent_item_v0 *ei0;
726 struct btrfs_extent_ref_v0 *ref0;
727 struct btrfs_tree_block_info *bi;
728 struct extent_buffer *leaf;
729 struct btrfs_key key;
730 struct btrfs_key found_key;
731 u32 new_size = sizeof(*item);
732 u64 refs;
733 int ret;
734
735 leaf = path->nodes[0];
736 BUG_ON(btrfs_item_size_nr(leaf, path->slots[0]) != sizeof(*ei0));
737
738 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
739 ei0 = btrfs_item_ptr(leaf, path->slots[0],
740 struct btrfs_extent_item_v0);
741 refs = btrfs_extent_refs_v0(leaf, ei0);
742
743 if (owner == (u64)-1) {
744 while (1) {
745 if (path->slots[0] >= btrfs_header_nritems(leaf)) {
746 ret = btrfs_next_leaf(root, path);
747 if (ret < 0)
748 return ret;
749 BUG_ON(ret > 0);
750 leaf = path->nodes[0];
751 }
752 btrfs_item_key_to_cpu(leaf, &found_key,
753 path->slots[0]);
754 BUG_ON(key.objectid != found_key.objectid);
755 if (found_key.type != BTRFS_EXTENT_REF_V0_KEY) {
756 path->slots[0]++;
757 continue;
758 }
759 ref0 = btrfs_item_ptr(leaf, path->slots[0],
760 struct btrfs_extent_ref_v0);
761 owner = btrfs_ref_objectid_v0(leaf, ref0);
762 break;
763 }
764 }
765 btrfs_release_path(root, path);
766
767 if (owner < BTRFS_FIRST_FREE_OBJECTID)
768 new_size += sizeof(*bi);
769
770 new_size -= sizeof(*ei0);
771 ret = btrfs_search_slot(trans, root, &key, path,
772 new_size + extra_size, 1);
773 if (ret < 0)
774 return ret;
775 BUG_ON(ret);
776
777 ret = btrfs_extend_item(trans, root, path, new_size);
778 BUG_ON(ret);
779
780 leaf = path->nodes[0];
781 item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
782 btrfs_set_extent_refs(leaf, item, refs);
783 /* FIXME: get real generation */
784 btrfs_set_extent_generation(leaf, item, 0);
785 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
786 btrfs_set_extent_flags(leaf, item,
787 BTRFS_EXTENT_FLAG_TREE_BLOCK |
788 BTRFS_BLOCK_FLAG_FULL_BACKREF);
789 bi = (struct btrfs_tree_block_info *)(item + 1);
790 /* FIXME: get first key of the block */
791 memset_extent_buffer(leaf, 0, (unsigned long)bi, sizeof(*bi));
792 btrfs_set_tree_block_level(leaf, bi, (int)owner);
793 } else {
794 btrfs_set_extent_flags(leaf, item, BTRFS_EXTENT_FLAG_DATA);
795 }
796 btrfs_mark_buffer_dirty(leaf);
797 return 0;
798 }
799 #endif
800
801 static u64 hash_extent_data_ref(u64 root_objectid, u64 owner, u64 offset)
802 {
803 u32 high_crc = ~(u32)0;
804 u32 low_crc = ~(u32)0;
805 __le64 lenum;
806
807 lenum = cpu_to_le64(root_objectid);
808 high_crc = crc32c(high_crc, &lenum, sizeof(lenum));
809 lenum = cpu_to_le64(owner);
810 low_crc = crc32c(low_crc, &lenum, sizeof(lenum));
811 lenum = cpu_to_le64(offset);
812 low_crc = crc32c(low_crc, &lenum, sizeof(lenum));
813
814 return ((u64)high_crc << 31) ^ (u64)low_crc;
815 }
816
817 static u64 hash_extent_data_ref_item(struct extent_buffer *leaf,
818 struct btrfs_extent_data_ref *ref)
819 {
820 return hash_extent_data_ref(btrfs_extent_data_ref_root(leaf, ref),
821 btrfs_extent_data_ref_objectid(leaf, ref),
822 btrfs_extent_data_ref_offset(leaf, ref));
823 }
824
825 static int match_extent_data_ref(struct extent_buffer *leaf,
826 struct btrfs_extent_data_ref *ref,
827 u64 root_objectid, u64 owner, u64 offset)
828 {
829 if (btrfs_extent_data_ref_root(leaf, ref) != root_objectid ||
830 btrfs_extent_data_ref_objectid(leaf, ref) != owner ||
831 btrfs_extent_data_ref_offset(leaf, ref) != offset)
832 return 0;
833 return 1;
834 }
835
836 static noinline int lookup_extent_data_ref(struct btrfs_trans_handle *trans,
837 struct btrfs_root *root,
838 struct btrfs_path *path,
839 u64 bytenr, u64 parent,
840 u64 root_objectid,
841 u64 owner, u64 offset)
842 {
843 struct btrfs_key key;
844 struct btrfs_extent_data_ref *ref;
845 struct extent_buffer *leaf;
846 u32 nritems;
847 int ret;
848 int recow;
849 int err = -ENOENT;
850
851 key.objectid = bytenr;
852 if (parent) {
853 key.type = BTRFS_SHARED_DATA_REF_KEY;
854 key.offset = parent;
855 } else {
856 key.type = BTRFS_EXTENT_DATA_REF_KEY;
857 key.offset = hash_extent_data_ref(root_objectid,
858 owner, offset);
859 }
860 again:
861 recow = 0;
862 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
863 if (ret < 0) {
864 err = ret;
865 goto fail;
866 }
867
868 if (parent) {
869 if (!ret)
870 return 0;
871 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
872 key.type = BTRFS_EXTENT_REF_V0_KEY;
873 btrfs_release_path(root, path);
874 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
875 if (ret < 0) {
876 err = ret;
877 goto fail;
878 }
879 if (!ret)
880 return 0;
881 #endif
882 goto fail;
883 }
884
885 leaf = path->nodes[0];
886 nritems = btrfs_header_nritems(leaf);
887 while (1) {
888 if (path->slots[0] >= nritems) {
889 ret = btrfs_next_leaf(root, path);
890 if (ret < 0)
891 err = ret;
892 if (ret)
893 goto fail;
894
895 leaf = path->nodes[0];
896 nritems = btrfs_header_nritems(leaf);
897 recow = 1;
898 }
899
900 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
901 if (key.objectid != bytenr ||
902 key.type != BTRFS_EXTENT_DATA_REF_KEY)
903 goto fail;
904
905 ref = btrfs_item_ptr(leaf, path->slots[0],
906 struct btrfs_extent_data_ref);
907
908 if (match_extent_data_ref(leaf, ref, root_objectid,
909 owner, offset)) {
910 if (recow) {
911 btrfs_release_path(root, path);
912 goto again;
913 }
914 err = 0;
915 break;
916 }
917 path->slots[0]++;
918 }
919 fail:
920 return err;
921 }
922
923 static noinline int insert_extent_data_ref(struct btrfs_trans_handle *trans,
924 struct btrfs_root *root,
925 struct btrfs_path *path,
926 u64 bytenr, u64 parent,
927 u64 root_objectid, u64 owner,
928 u64 offset, int refs_to_add)
929 {
930 struct btrfs_key key;
931 struct extent_buffer *leaf;
932 u32 size;
933 u32 num_refs;
934 int ret;
935
936 key.objectid = bytenr;
937 if (parent) {
938 key.type = BTRFS_SHARED_DATA_REF_KEY;
939 key.offset = parent;
940 size = sizeof(struct btrfs_shared_data_ref);
941 } else {
942 key.type = BTRFS_EXTENT_DATA_REF_KEY;
943 key.offset = hash_extent_data_ref(root_objectid,
944 owner, offset);
945 size = sizeof(struct btrfs_extent_data_ref);
946 }
947
948 ret = btrfs_insert_empty_item(trans, root, path, &key, size);
949 if (ret && ret != -EEXIST)
950 goto fail;
951
952 leaf = path->nodes[0];
953 if (parent) {
954 struct btrfs_shared_data_ref *ref;
955 ref = btrfs_item_ptr(leaf, path->slots[0],
956 struct btrfs_shared_data_ref);
957 if (ret == 0) {
958 btrfs_set_shared_data_ref_count(leaf, ref, refs_to_add);
959 } else {
960 num_refs = btrfs_shared_data_ref_count(leaf, ref);
961 num_refs += refs_to_add;
962 btrfs_set_shared_data_ref_count(leaf, ref, num_refs);
963 }
964 } else {
965 struct btrfs_extent_data_ref *ref;
966 while (ret == -EEXIST) {
967 ref = btrfs_item_ptr(leaf, path->slots[0],
968 struct btrfs_extent_data_ref);
969 if (match_extent_data_ref(leaf, ref, root_objectid,
970 owner, offset))
971 break;
972 btrfs_release_path(root, path);
973 key.offset++;
974 ret = btrfs_insert_empty_item(trans, root, path, &key,
975 size);
976 if (ret && ret != -EEXIST)
977 goto fail;
978
979 leaf = path->nodes[0];
980 }
981 ref = btrfs_item_ptr(leaf, path->slots[0],
982 struct btrfs_extent_data_ref);
983 if (ret == 0) {
984 btrfs_set_extent_data_ref_root(leaf, ref,
985 root_objectid);
986 btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
987 btrfs_set_extent_data_ref_offset(leaf, ref, offset);
988 btrfs_set_extent_data_ref_count(leaf, ref, refs_to_add);
989 } else {
990 num_refs = btrfs_extent_data_ref_count(leaf, ref);
991 num_refs += refs_to_add;
992 btrfs_set_extent_data_ref_count(leaf, ref, num_refs);
993 }
994 }
995 btrfs_mark_buffer_dirty(leaf);
996 ret = 0;
997 fail:
998 btrfs_release_path(root, path);
999 return ret;
1000 }
1001
1002 static noinline int remove_extent_data_ref(struct btrfs_trans_handle *trans,
1003 struct btrfs_root *root,
1004 struct btrfs_path *path,
1005 int refs_to_drop)
1006 {
1007 struct btrfs_key key;
1008 struct btrfs_extent_data_ref *ref1 = NULL;
1009 struct btrfs_shared_data_ref *ref2 = NULL;
1010 struct extent_buffer *leaf;
1011 u32 num_refs = 0;
1012 int ret = 0;
1013
1014 leaf = path->nodes[0];
1015 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
1016
1017 if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
1018 ref1 = btrfs_item_ptr(leaf, path->slots[0],
1019 struct btrfs_extent_data_ref);
1020 num_refs = btrfs_extent_data_ref_count(leaf, ref1);
1021 } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
1022 ref2 = btrfs_item_ptr(leaf, path->slots[0],
1023 struct btrfs_shared_data_ref);
1024 num_refs = btrfs_shared_data_ref_count(leaf, ref2);
1025 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1026 } else if (key.type == BTRFS_EXTENT_REF_V0_KEY) {
1027 struct btrfs_extent_ref_v0 *ref0;
1028 ref0 = btrfs_item_ptr(leaf, path->slots[0],
1029 struct btrfs_extent_ref_v0);
1030 num_refs = btrfs_ref_count_v0(leaf, ref0);
1031 #endif
1032 } else {
1033 BUG();
1034 }
1035
1036 BUG_ON(num_refs < refs_to_drop);
1037 num_refs -= refs_to_drop;
1038
1039 if (num_refs == 0) {
1040 ret = btrfs_del_item(trans, root, path);
1041 } else {
1042 if (key.type == BTRFS_EXTENT_DATA_REF_KEY)
1043 btrfs_set_extent_data_ref_count(leaf, ref1, num_refs);
1044 else if (key.type == BTRFS_SHARED_DATA_REF_KEY)
1045 btrfs_set_shared_data_ref_count(leaf, ref2, num_refs);
1046 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1047 else {
1048 struct btrfs_extent_ref_v0 *ref0;
1049 ref0 = btrfs_item_ptr(leaf, path->slots[0],
1050 struct btrfs_extent_ref_v0);
1051 btrfs_set_ref_count_v0(leaf, ref0, num_refs);
1052 }
1053 #endif
1054 btrfs_mark_buffer_dirty(leaf);
1055 }
1056 return ret;
1057 }
1058
1059 static noinline u32 extent_data_ref_count(struct btrfs_root *root,
1060 struct btrfs_path *path,
1061 struct btrfs_extent_inline_ref *iref)
1062 {
1063 struct btrfs_key key;
1064 struct extent_buffer *leaf;
1065 struct btrfs_extent_data_ref *ref1;
1066 struct btrfs_shared_data_ref *ref2;
1067 u32 num_refs = 0;
1068
1069 leaf = path->nodes[0];
1070 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
1071 if (iref) {
1072 if (btrfs_extent_inline_ref_type(leaf, iref) ==
1073 BTRFS_EXTENT_DATA_REF_KEY) {
1074 ref1 = (struct btrfs_extent_data_ref *)(&iref->offset);
1075 num_refs = btrfs_extent_data_ref_count(leaf, ref1);
1076 } else {
1077 ref2 = (struct btrfs_shared_data_ref *)(iref + 1);
1078 num_refs = btrfs_shared_data_ref_count(leaf, ref2);
1079 }
1080 } else if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
1081 ref1 = btrfs_item_ptr(leaf, path->slots[0],
1082 struct btrfs_extent_data_ref);
1083 num_refs = btrfs_extent_data_ref_count(leaf, ref1);
1084 } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
1085 ref2 = btrfs_item_ptr(leaf, path->slots[0],
1086 struct btrfs_shared_data_ref);
1087 num_refs = btrfs_shared_data_ref_count(leaf, ref2);
1088 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1089 } else if (key.type == BTRFS_EXTENT_REF_V0_KEY) {
1090 struct btrfs_extent_ref_v0 *ref0;
1091 ref0 = btrfs_item_ptr(leaf, path->slots[0],
1092 struct btrfs_extent_ref_v0);
1093 num_refs = btrfs_ref_count_v0(leaf, ref0);
1094 #endif
1095 } else {
1096 WARN_ON(1);
1097 }
1098 return num_refs;
1099 }
1100
1101 static noinline int lookup_tree_block_ref(struct btrfs_trans_handle *trans,
1102 struct btrfs_root *root,
1103 struct btrfs_path *path,
1104 u64 bytenr, u64 parent,
1105 u64 root_objectid)
1106 {
1107 struct btrfs_key key;
1108 int ret;
1109
1110 key.objectid = bytenr;
1111 if (parent) {
1112 key.type = BTRFS_SHARED_BLOCK_REF_KEY;
1113 key.offset = parent;
1114 } else {
1115 key.type = BTRFS_TREE_BLOCK_REF_KEY;
1116 key.offset = root_objectid;
1117 }
1118
1119 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1120 if (ret > 0)
1121 ret = -ENOENT;
1122 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1123 if (ret == -ENOENT && parent) {
1124 btrfs_release_path(root, path);
1125 key.type = BTRFS_EXTENT_REF_V0_KEY;
1126 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1127 if (ret > 0)
1128 ret = -ENOENT;
1129 }
1130 #endif
1131 return ret;
1132 }
1133
1134 static noinline int insert_tree_block_ref(struct btrfs_trans_handle *trans,
1135 struct btrfs_root *root,
1136 struct btrfs_path *path,
1137 u64 bytenr, u64 parent,
1138 u64 root_objectid)
1139 {
1140 struct btrfs_key key;
1141 int ret;
1142
1143 key.objectid = bytenr;
1144 if (parent) {
1145 key.type = BTRFS_SHARED_BLOCK_REF_KEY;
1146 key.offset = parent;
1147 } else {
1148 key.type = BTRFS_TREE_BLOCK_REF_KEY;
1149 key.offset = root_objectid;
1150 }
1151
1152 ret = btrfs_insert_empty_item(trans, root, path, &key, 0);
1153 btrfs_release_path(root, path);
1154 return ret;
1155 }
1156
1157 static inline int extent_ref_type(u64 parent, u64 owner)
1158 {
1159 int type;
1160 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1161 if (parent > 0)
1162 type = BTRFS_SHARED_BLOCK_REF_KEY;
1163 else
1164 type = BTRFS_TREE_BLOCK_REF_KEY;
1165 } else {
1166 if (parent > 0)
1167 type = BTRFS_SHARED_DATA_REF_KEY;
1168 else
1169 type = BTRFS_EXTENT_DATA_REF_KEY;
1170 }
1171 return type;
1172 }
1173
1174 static int find_next_key(struct btrfs_path *path, int level,
1175 struct btrfs_key *key)
1176
1177 {
1178 for (; level < BTRFS_MAX_LEVEL; level++) {
1179 if (!path->nodes[level])
1180 break;
1181 if (path->slots[level] + 1 >=
1182 btrfs_header_nritems(path->nodes[level]))
1183 continue;
1184 if (level == 0)
1185 btrfs_item_key_to_cpu(path->nodes[level], key,
1186 path->slots[level] + 1);
1187 else
1188 btrfs_node_key_to_cpu(path->nodes[level], key,
1189 path->slots[level] + 1);
1190 return 0;
1191 }
1192 return 1;
1193 }
1194
1195 /*
1196 * look for inline back ref. if back ref is found, *ref_ret is set
1197 * to the address of inline back ref, and 0 is returned.
1198 *
1199 * if back ref isn't found, *ref_ret is set to the address where it
1200 * should be inserted, and -ENOENT is returned.
1201 *
1202 * if insert is true and there are too many inline back refs, the path
1203 * points to the extent item, and -EAGAIN is returned.
1204 *
1205 * NOTE: inline back refs are ordered in the same way that back ref
1206 * items in the tree are ordered.
1207 */
1208 static noinline_for_stack
1209 int lookup_inline_extent_backref(struct btrfs_trans_handle *trans,
1210 struct btrfs_root *root,
1211 struct btrfs_path *path,
1212 struct btrfs_extent_inline_ref **ref_ret,
1213 u64 bytenr, u64 num_bytes,
1214 u64 parent, u64 root_objectid,
1215 u64 owner, u64 offset, int insert)
1216 {
1217 struct btrfs_key key;
1218 struct extent_buffer *leaf;
1219 struct btrfs_extent_item *ei;
1220 struct btrfs_extent_inline_ref *iref;
1221 u64 flags;
1222 u64 item_size;
1223 unsigned long ptr;
1224 unsigned long end;
1225 int extra_size;
1226 int type;
1227 int want;
1228 int ret;
1229 int err = 0;
1230
1231 key.objectid = bytenr;
1232 key.type = BTRFS_EXTENT_ITEM_KEY;
1233 key.offset = num_bytes;
1234
1235 want = extent_ref_type(parent, owner);
1236 if (insert) {
1237 extra_size = btrfs_extent_inline_ref_size(want);
1238 path->keep_locks = 1;
1239 } else
1240 extra_size = -1;
1241 ret = btrfs_search_slot(trans, root, &key, path, extra_size, 1);
1242 if (ret < 0) {
1243 err = ret;
1244 goto out;
1245 }
1246 BUG_ON(ret);
1247
1248 leaf = path->nodes[0];
1249 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1250 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1251 if (item_size < sizeof(*ei)) {
1252 if (!insert) {
1253 err = -ENOENT;
1254 goto out;
1255 }
1256 ret = convert_extent_item_v0(trans, root, path, owner,
1257 extra_size);
1258 if (ret < 0) {
1259 err = ret;
1260 goto out;
1261 }
1262 leaf = path->nodes[0];
1263 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1264 }
1265 #endif
1266 BUG_ON(item_size < sizeof(*ei));
1267
1268 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1269 flags = btrfs_extent_flags(leaf, ei);
1270
1271 ptr = (unsigned long)(ei + 1);
1272 end = (unsigned long)ei + item_size;
1273
1274 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
1275 ptr += sizeof(struct btrfs_tree_block_info);
1276 BUG_ON(ptr > end);
1277 } else {
1278 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
1279 }
1280
1281 err = -ENOENT;
1282 while (1) {
1283 if (ptr >= end) {
1284 WARN_ON(ptr > end);
1285 break;
1286 }
1287 iref = (struct btrfs_extent_inline_ref *)ptr;
1288 type = btrfs_extent_inline_ref_type(leaf, iref);
1289 if (want < type)
1290 break;
1291 if (want > type) {
1292 ptr += btrfs_extent_inline_ref_size(type);
1293 continue;
1294 }
1295
1296 if (type == BTRFS_EXTENT_DATA_REF_KEY) {
1297 struct btrfs_extent_data_ref *dref;
1298 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1299 if (match_extent_data_ref(leaf, dref, root_objectid,
1300 owner, offset)) {
1301 err = 0;
1302 break;
1303 }
1304 if (hash_extent_data_ref_item(leaf, dref) <
1305 hash_extent_data_ref(root_objectid, owner, offset))
1306 break;
1307 } else {
1308 u64 ref_offset;
1309 ref_offset = btrfs_extent_inline_ref_offset(leaf, iref);
1310 if (parent > 0) {
1311 if (parent == ref_offset) {
1312 err = 0;
1313 break;
1314 }
1315 if (ref_offset < parent)
1316 break;
1317 } else {
1318 if (root_objectid == ref_offset) {
1319 err = 0;
1320 break;
1321 }
1322 if (ref_offset < root_objectid)
1323 break;
1324 }
1325 }
1326 ptr += btrfs_extent_inline_ref_size(type);
1327 }
1328 if (err == -ENOENT && insert) {
1329 if (item_size + extra_size >=
1330 BTRFS_MAX_EXTENT_ITEM_SIZE(root)) {
1331 err = -EAGAIN;
1332 goto out;
1333 }
1334 /*
1335 * To add new inline back ref, we have to make sure
1336 * there is no corresponding back ref item.
1337 * For simplicity, we just do not add new inline back
1338 * ref if there is any kind of item for this block
1339 */
1340 if (find_next_key(path, 0, &key) == 0 &&
1341 key.objectid == bytenr &&
1342 key.type < BTRFS_BLOCK_GROUP_ITEM_KEY) {
1343 err = -EAGAIN;
1344 goto out;
1345 }
1346 }
1347 *ref_ret = (struct btrfs_extent_inline_ref *)ptr;
1348 out:
1349 if (insert) {
1350 path->keep_locks = 0;
1351 btrfs_unlock_up_safe(path, 1);
1352 }
1353 return err;
1354 }
1355
1356 /*
1357 * helper to add new inline back ref
1358 */
1359 static noinline_for_stack
1360 int setup_inline_extent_backref(struct btrfs_trans_handle *trans,
1361 struct btrfs_root *root,
1362 struct btrfs_path *path,
1363 struct btrfs_extent_inline_ref *iref,
1364 u64 parent, u64 root_objectid,
1365 u64 owner, u64 offset, int refs_to_add,
1366 struct btrfs_delayed_extent_op *extent_op)
1367 {
1368 struct extent_buffer *leaf;
1369 struct btrfs_extent_item *ei;
1370 unsigned long ptr;
1371 unsigned long end;
1372 unsigned long item_offset;
1373 u64 refs;
1374 int size;
1375 int type;
1376 int ret;
1377
1378 leaf = path->nodes[0];
1379 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1380 item_offset = (unsigned long)iref - (unsigned long)ei;
1381
1382 type = extent_ref_type(parent, owner);
1383 size = btrfs_extent_inline_ref_size(type);
1384
1385 ret = btrfs_extend_item(trans, root, path, size);
1386 BUG_ON(ret);
1387
1388 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1389 refs = btrfs_extent_refs(leaf, ei);
1390 refs += refs_to_add;
1391 btrfs_set_extent_refs(leaf, ei, refs);
1392 if (extent_op)
1393 __run_delayed_extent_op(extent_op, leaf, ei);
1394
1395 ptr = (unsigned long)ei + item_offset;
1396 end = (unsigned long)ei + btrfs_item_size_nr(leaf, path->slots[0]);
1397 if (ptr < end - size)
1398 memmove_extent_buffer(leaf, ptr + size, ptr,
1399 end - size - ptr);
1400
1401 iref = (struct btrfs_extent_inline_ref *)ptr;
1402 btrfs_set_extent_inline_ref_type(leaf, iref, type);
1403 if (type == BTRFS_EXTENT_DATA_REF_KEY) {
1404 struct btrfs_extent_data_ref *dref;
1405 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1406 btrfs_set_extent_data_ref_root(leaf, dref, root_objectid);
1407 btrfs_set_extent_data_ref_objectid(leaf, dref, owner);
1408 btrfs_set_extent_data_ref_offset(leaf, dref, offset);
1409 btrfs_set_extent_data_ref_count(leaf, dref, refs_to_add);
1410 } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
1411 struct btrfs_shared_data_ref *sref;
1412 sref = (struct btrfs_shared_data_ref *)(iref + 1);
1413 btrfs_set_shared_data_ref_count(leaf, sref, refs_to_add);
1414 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
1415 } else if (type == BTRFS_SHARED_BLOCK_REF_KEY) {
1416 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
1417 } else {
1418 btrfs_set_extent_inline_ref_offset(leaf, iref, root_objectid);
1419 }
1420 btrfs_mark_buffer_dirty(leaf);
1421 return 0;
1422 }
1423
1424 static int lookup_extent_backref(struct btrfs_trans_handle *trans,
1425 struct btrfs_root *root,
1426 struct btrfs_path *path,
1427 struct btrfs_extent_inline_ref **ref_ret,
1428 u64 bytenr, u64 num_bytes, u64 parent,
1429 u64 root_objectid, u64 owner, u64 offset)
1430 {
1431 int ret;
1432
1433 ret = lookup_inline_extent_backref(trans, root, path, ref_ret,
1434 bytenr, num_bytes, parent,
1435 root_objectid, owner, offset, 0);
1436 if (ret != -ENOENT)
1437 return ret;
1438
1439 btrfs_release_path(root, path);
1440 *ref_ret = NULL;
1441
1442 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1443 ret = lookup_tree_block_ref(trans, root, path, bytenr, parent,
1444 root_objectid);
1445 } else {
1446 ret = lookup_extent_data_ref(trans, root, path, bytenr, parent,
1447 root_objectid, owner, offset);
1448 }
1449 return ret;
1450 }
1451
1452 /*
1453 * helper to update/remove inline back ref
1454 */
1455 static noinline_for_stack
1456 int update_inline_extent_backref(struct btrfs_trans_handle *trans,
1457 struct btrfs_root *root,
1458 struct btrfs_path *path,
1459 struct btrfs_extent_inline_ref *iref,
1460 int refs_to_mod,
1461 struct btrfs_delayed_extent_op *extent_op)
1462 {
1463 struct extent_buffer *leaf;
1464 struct btrfs_extent_item *ei;
1465 struct btrfs_extent_data_ref *dref = NULL;
1466 struct btrfs_shared_data_ref *sref = NULL;
1467 unsigned long ptr;
1468 unsigned long end;
1469 u32 item_size;
1470 int size;
1471 int type;
1472 int ret;
1473 u64 refs;
1474
1475 leaf = path->nodes[0];
1476 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1477 refs = btrfs_extent_refs(leaf, ei);
1478 WARN_ON(refs_to_mod < 0 && refs + refs_to_mod <= 0);
1479 refs += refs_to_mod;
1480 btrfs_set_extent_refs(leaf, ei, refs);
1481 if (extent_op)
1482 __run_delayed_extent_op(extent_op, leaf, ei);
1483
1484 type = btrfs_extent_inline_ref_type(leaf, iref);
1485
1486 if (type == BTRFS_EXTENT_DATA_REF_KEY) {
1487 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1488 refs = btrfs_extent_data_ref_count(leaf, dref);
1489 } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
1490 sref = (struct btrfs_shared_data_ref *)(iref + 1);
1491 refs = btrfs_shared_data_ref_count(leaf, sref);
1492 } else {
1493 refs = 1;
1494 BUG_ON(refs_to_mod != -1);
1495 }
1496
1497 BUG_ON(refs_to_mod < 0 && refs < -refs_to_mod);
1498 refs += refs_to_mod;
1499
1500 if (refs > 0) {
1501 if (type == BTRFS_EXTENT_DATA_REF_KEY)
1502 btrfs_set_extent_data_ref_count(leaf, dref, refs);
1503 else
1504 btrfs_set_shared_data_ref_count(leaf, sref, refs);
1505 } else {
1506 size = btrfs_extent_inline_ref_size(type);
1507 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1508 ptr = (unsigned long)iref;
1509 end = (unsigned long)ei + item_size;
1510 if (ptr + size < end)
1511 memmove_extent_buffer(leaf, ptr, ptr + size,
1512 end - ptr - size);
1513 item_size -= size;
1514 ret = btrfs_truncate_item(trans, root, path, item_size, 1);
1515 BUG_ON(ret);
1516 }
1517 btrfs_mark_buffer_dirty(leaf);
1518 return 0;
1519 }
1520
1521 static noinline_for_stack
1522 int insert_inline_extent_backref(struct btrfs_trans_handle *trans,
1523 struct btrfs_root *root,
1524 struct btrfs_path *path,
1525 u64 bytenr, u64 num_bytes, u64 parent,
1526 u64 root_objectid, u64 owner,
1527 u64 offset, int refs_to_add,
1528 struct btrfs_delayed_extent_op *extent_op)
1529 {
1530 struct btrfs_extent_inline_ref *iref;
1531 int ret;
1532
1533 ret = lookup_inline_extent_backref(trans, root, path, &iref,
1534 bytenr, num_bytes, parent,
1535 root_objectid, owner, offset, 1);
1536 if (ret == 0) {
1537 BUG_ON(owner < BTRFS_FIRST_FREE_OBJECTID);
1538 ret = update_inline_extent_backref(trans, root, path, iref,
1539 refs_to_add, extent_op);
1540 } else if (ret == -ENOENT) {
1541 ret = setup_inline_extent_backref(trans, root, path, iref,
1542 parent, root_objectid,
1543 owner, offset, refs_to_add,
1544 extent_op);
1545 }
1546 return ret;
1547 }
1548
1549 static int insert_extent_backref(struct btrfs_trans_handle *trans,
1550 struct btrfs_root *root,
1551 struct btrfs_path *path,
1552 u64 bytenr, u64 parent, u64 root_objectid,
1553 u64 owner, u64 offset, int refs_to_add)
1554 {
1555 int ret;
1556 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1557 BUG_ON(refs_to_add != 1);
1558 ret = insert_tree_block_ref(trans, root, path, bytenr,
1559 parent, root_objectid);
1560 } else {
1561 ret = insert_extent_data_ref(trans, root, path, bytenr,
1562 parent, root_objectid,
1563 owner, offset, refs_to_add);
1564 }
1565 return ret;
1566 }
1567
1568 static int remove_extent_backref(struct btrfs_trans_handle *trans,
1569 struct btrfs_root *root,
1570 struct btrfs_path *path,
1571 struct btrfs_extent_inline_ref *iref,
1572 int refs_to_drop, int is_data)
1573 {
1574 int ret;
1575
1576 BUG_ON(!is_data && refs_to_drop != 1);
1577 if (iref) {
1578 ret = update_inline_extent_backref(trans, root, path, iref,
1579 -refs_to_drop, NULL);
1580 } else if (is_data) {
1581 ret = remove_extent_data_ref(trans, root, path, refs_to_drop);
1582 } else {
1583 ret = btrfs_del_item(trans, root, path);
1584 }
1585 return ret;
1586 }
1587
1588 static void btrfs_issue_discard(struct block_device *bdev,
1589 u64 start, u64 len)
1590 {
1591 blkdev_issue_discard(bdev, start >> 9, len >> 9, GFP_KERNEL,
1592 BLKDEV_IFL_WAIT | BLKDEV_IFL_BARRIER);
1593 }
1594
1595 static int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr,
1596 u64 num_bytes)
1597 {
1598 int ret;
1599 u64 map_length = num_bytes;
1600 struct btrfs_multi_bio *multi = NULL;
1601
1602 if (!btrfs_test_opt(root, DISCARD))
1603 return 0;
1604
1605 /* Tell the block device(s) that the sectors can be discarded */
1606 ret = btrfs_map_block(&root->fs_info->mapping_tree, READ,
1607 bytenr, &map_length, &multi, 0);
1608 if (!ret) {
1609 struct btrfs_bio_stripe *stripe = multi->stripes;
1610 int i;
1611
1612 if (map_length > num_bytes)
1613 map_length = num_bytes;
1614
1615 for (i = 0; i < multi->num_stripes; i++, stripe++) {
1616 btrfs_issue_discard(stripe->dev->bdev,
1617 stripe->physical,
1618 map_length);
1619 }
1620 kfree(multi);
1621 }
1622
1623 return ret;
1624 }
1625
1626 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
1627 struct btrfs_root *root,
1628 u64 bytenr, u64 num_bytes, u64 parent,
1629 u64 root_objectid, u64 owner, u64 offset)
1630 {
1631 int ret;
1632 BUG_ON(owner < BTRFS_FIRST_FREE_OBJECTID &&
1633 root_objectid == BTRFS_TREE_LOG_OBJECTID);
1634
1635 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1636 ret = btrfs_add_delayed_tree_ref(trans, bytenr, num_bytes,
1637 parent, root_objectid, (int)owner,
1638 BTRFS_ADD_DELAYED_REF, NULL);
1639 } else {
1640 ret = btrfs_add_delayed_data_ref(trans, bytenr, num_bytes,
1641 parent, root_objectid, owner, offset,
1642 BTRFS_ADD_DELAYED_REF, NULL);
1643 }
1644 return ret;
1645 }
1646
1647 static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
1648 struct btrfs_root *root,
1649 u64 bytenr, u64 num_bytes,
1650 u64 parent, u64 root_objectid,
1651 u64 owner, u64 offset, int refs_to_add,
1652 struct btrfs_delayed_extent_op *extent_op)
1653 {
1654 struct btrfs_path *path;
1655 struct extent_buffer *leaf;
1656 struct btrfs_extent_item *item;
1657 u64 refs;
1658 int ret;
1659 int err = 0;
1660
1661 path = btrfs_alloc_path();
1662 if (!path)
1663 return -ENOMEM;
1664
1665 path->reada = 1;
1666 path->leave_spinning = 1;
1667 /* this will setup the path even if it fails to insert the back ref */
1668 ret = insert_inline_extent_backref(trans, root->fs_info->extent_root,
1669 path, bytenr, num_bytes, parent,
1670 root_objectid, owner, offset,
1671 refs_to_add, extent_op);
1672 if (ret == 0)
1673 goto out;
1674
1675 if (ret != -EAGAIN) {
1676 err = ret;
1677 goto out;
1678 }
1679
1680 leaf = path->nodes[0];
1681 item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1682 refs = btrfs_extent_refs(leaf, item);
1683 btrfs_set_extent_refs(leaf, item, refs + refs_to_add);
1684 if (extent_op)
1685 __run_delayed_extent_op(extent_op, leaf, item);
1686
1687 btrfs_mark_buffer_dirty(leaf);
1688 btrfs_release_path(root->fs_info->extent_root, path);
1689
1690 path->reada = 1;
1691 path->leave_spinning = 1;
1692
1693 /* now insert the actual backref */
1694 ret = insert_extent_backref(trans, root->fs_info->extent_root,
1695 path, bytenr, parent, root_objectid,
1696 owner, offset, refs_to_add);
1697 BUG_ON(ret);
1698 out:
1699 btrfs_free_path(path);
1700 return err;
1701 }
1702
1703 static int run_delayed_data_ref(struct btrfs_trans_handle *trans,
1704 struct btrfs_root *root,
1705 struct btrfs_delayed_ref_node *node,
1706 struct btrfs_delayed_extent_op *extent_op,
1707 int insert_reserved)
1708 {
1709 int ret = 0;
1710 struct btrfs_delayed_data_ref *ref;
1711 struct btrfs_key ins;
1712 u64 parent = 0;
1713 u64 ref_root = 0;
1714 u64 flags = 0;
1715
1716 ins.objectid = node->bytenr;
1717 ins.offset = node->num_bytes;
1718 ins.type = BTRFS_EXTENT_ITEM_KEY;
1719
1720 ref = btrfs_delayed_node_to_data_ref(node);
1721 if (node->type == BTRFS_SHARED_DATA_REF_KEY)
1722 parent = ref->parent;
1723 else
1724 ref_root = ref->root;
1725
1726 if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
1727 if (extent_op) {
1728 BUG_ON(extent_op->update_key);
1729 flags |= extent_op->flags_to_set;
1730 }
1731 ret = alloc_reserved_file_extent(trans, root,
1732 parent, ref_root, flags,
1733 ref->objectid, ref->offset,
1734 &ins, node->ref_mod);
1735 } else if (node->action == BTRFS_ADD_DELAYED_REF) {
1736 ret = __btrfs_inc_extent_ref(trans, root, node->bytenr,
1737 node->num_bytes, parent,
1738 ref_root, ref->objectid,
1739 ref->offset, node->ref_mod,
1740 extent_op);
1741 } else if (node->action == BTRFS_DROP_DELAYED_REF) {
1742 ret = __btrfs_free_extent(trans, root, node->bytenr,
1743 node->num_bytes, parent,
1744 ref_root, ref->objectid,
1745 ref->offset, node->ref_mod,
1746 extent_op);
1747 } else {
1748 BUG();
1749 }
1750 return ret;
1751 }
1752
1753 static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
1754 struct extent_buffer *leaf,
1755 struct btrfs_extent_item *ei)
1756 {
1757 u64 flags = btrfs_extent_flags(leaf, ei);
1758 if (extent_op->update_flags) {
1759 flags |= extent_op->flags_to_set;
1760 btrfs_set_extent_flags(leaf, ei, flags);
1761 }
1762
1763 if (extent_op->update_key) {
1764 struct btrfs_tree_block_info *bi;
1765 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK));
1766 bi = (struct btrfs_tree_block_info *)(ei + 1);
1767 btrfs_set_tree_block_key(leaf, bi, &extent_op->key);
1768 }
1769 }
1770
1771 static int run_delayed_extent_op(struct btrfs_trans_handle *trans,
1772 struct btrfs_root *root,
1773 struct btrfs_delayed_ref_node *node,
1774 struct btrfs_delayed_extent_op *extent_op)
1775 {
1776 struct btrfs_key key;
1777 struct btrfs_path *path;
1778 struct btrfs_extent_item *ei;
1779 struct extent_buffer *leaf;
1780 u32 item_size;
1781 int ret;
1782 int err = 0;
1783
1784 path = btrfs_alloc_path();
1785 if (!path)
1786 return -ENOMEM;
1787
1788 key.objectid = node->bytenr;
1789 key.type = BTRFS_EXTENT_ITEM_KEY;
1790 key.offset = node->num_bytes;
1791
1792 path->reada = 1;
1793 path->leave_spinning = 1;
1794 ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key,
1795 path, 0, 1);
1796 if (ret < 0) {
1797 err = ret;
1798 goto out;
1799 }
1800 if (ret > 0) {
1801 err = -EIO;
1802 goto out;
1803 }
1804
1805 leaf = path->nodes[0];
1806 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1807 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1808 if (item_size < sizeof(*ei)) {
1809 ret = convert_extent_item_v0(trans, root->fs_info->extent_root,
1810 path, (u64)-1, 0);
1811 if (ret < 0) {
1812 err = ret;
1813 goto out;
1814 }
1815 leaf = path->nodes[0];
1816 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1817 }
1818 #endif
1819 BUG_ON(item_size < sizeof(*ei));
1820 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1821 __run_delayed_extent_op(extent_op, leaf, ei);
1822
1823 btrfs_mark_buffer_dirty(leaf);
1824 out:
1825 btrfs_free_path(path);
1826 return err;
1827 }
1828
1829 static int run_delayed_tree_ref(struct btrfs_trans_handle *trans,
1830 struct btrfs_root *root,
1831 struct btrfs_delayed_ref_node *node,
1832 struct btrfs_delayed_extent_op *extent_op,
1833 int insert_reserved)
1834 {
1835 int ret = 0;
1836 struct btrfs_delayed_tree_ref *ref;
1837 struct btrfs_key ins;
1838 u64 parent = 0;
1839 u64 ref_root = 0;
1840
1841 ins.objectid = node->bytenr;
1842 ins.offset = node->num_bytes;
1843 ins.type = BTRFS_EXTENT_ITEM_KEY;
1844
1845 ref = btrfs_delayed_node_to_tree_ref(node);
1846 if (node->type == BTRFS_SHARED_BLOCK_REF_KEY)
1847 parent = ref->parent;
1848 else
1849 ref_root = ref->root;
1850
1851 BUG_ON(node->ref_mod != 1);
1852 if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
1853 BUG_ON(!extent_op || !extent_op->update_flags ||
1854 !extent_op->update_key);
1855 ret = alloc_reserved_tree_block(trans, root,
1856 parent, ref_root,
1857 extent_op->flags_to_set,
1858 &extent_op->key,
1859 ref->level, &ins);
1860 } else if (node->action == BTRFS_ADD_DELAYED_REF) {
1861 ret = __btrfs_inc_extent_ref(trans, root, node->bytenr,
1862 node->num_bytes, parent, ref_root,
1863 ref->level, 0, 1, extent_op);
1864 } else if (node->action == BTRFS_DROP_DELAYED_REF) {
1865 ret = __btrfs_free_extent(trans, root, node->bytenr,
1866 node->num_bytes, parent, ref_root,
1867 ref->level, 0, 1, extent_op);
1868 } else {
1869 BUG();
1870 }
1871 return ret;
1872 }
1873
1874
1875 /* helper function to actually process a single delayed ref entry */
1876 static int run_one_delayed_ref(struct btrfs_trans_handle *trans,
1877 struct btrfs_root *root,
1878 struct btrfs_delayed_ref_node *node,
1879 struct btrfs_delayed_extent_op *extent_op,
1880 int insert_reserved)
1881 {
1882 int ret;
1883 if (btrfs_delayed_ref_is_head(node)) {
1884 struct btrfs_delayed_ref_head *head;
1885 /*
1886 * we've hit the end of the chain and we were supposed
1887 * to insert this extent into the tree. But, it got
1888 * deleted before we ever needed to insert it, so all
1889 * we have to do is clean up the accounting
1890 */
1891 BUG_ON(extent_op);
1892 head = btrfs_delayed_node_to_head(node);
1893 if (insert_reserved) {
1894 int mark_free = 0;
1895 struct extent_buffer *must_clean = NULL;
1896
1897 ret = pin_down_bytes(trans, root, NULL,
1898 node->bytenr, node->num_bytes,
1899 head->is_data, 1, &must_clean);
1900 if (ret > 0)
1901 mark_free = 1;
1902
1903 if (must_clean) {
1904 clean_tree_block(NULL, root, must_clean);
1905 btrfs_tree_unlock(must_clean);
1906 free_extent_buffer(must_clean);
1907 }
1908 if (head->is_data) {
1909 ret = btrfs_del_csums(trans, root,
1910 node->bytenr,
1911 node->num_bytes);
1912 BUG_ON(ret);
1913 }
1914 if (mark_free) {
1915 ret = btrfs_free_reserved_extent(root,
1916 node->bytenr,
1917 node->num_bytes);
1918 BUG_ON(ret);
1919 }
1920 }
1921 mutex_unlock(&head->mutex);
1922 return 0;
1923 }
1924
1925 if (node->type == BTRFS_TREE_BLOCK_REF_KEY ||
1926 node->type == BTRFS_SHARED_BLOCK_REF_KEY)
1927 ret = run_delayed_tree_ref(trans, root, node, extent_op,
1928 insert_reserved);
1929 else if (node->type == BTRFS_EXTENT_DATA_REF_KEY ||
1930 node->type == BTRFS_SHARED_DATA_REF_KEY)
1931 ret = run_delayed_data_ref(trans, root, node, extent_op,
1932 insert_reserved);
1933 else
1934 BUG();
1935 return ret;
1936 }
1937
1938 static noinline struct btrfs_delayed_ref_node *
1939 select_delayed_ref(struct btrfs_delayed_ref_head *head)
1940 {
1941 struct rb_node *node;
1942 struct btrfs_delayed_ref_node *ref;
1943 int action = BTRFS_ADD_DELAYED_REF;
1944 again:
1945 /*
1946 * select delayed ref of type BTRFS_ADD_DELAYED_REF first.
1947 * this prevents ref count from going down to zero when
1948 * there still are pending delayed ref.
1949 */
1950 node = rb_prev(&head->node.rb_node);
1951 while (1) {
1952 if (!node)
1953 break;
1954 ref = rb_entry(node, struct btrfs_delayed_ref_node,
1955 rb_node);
1956 if (ref->bytenr != head->node.bytenr)
1957 break;
1958 if (ref->action == action)
1959 return ref;
1960 node = rb_prev(node);
1961 }
1962 if (action == BTRFS_ADD_DELAYED_REF) {
1963 action = BTRFS_DROP_DELAYED_REF;
1964 goto again;
1965 }
1966 return NULL;
1967 }
1968
1969 static noinline int run_clustered_refs(struct btrfs_trans_handle *trans,
1970 struct btrfs_root *root,
1971 struct list_head *cluster)
1972 {
1973 struct btrfs_delayed_ref_root *delayed_refs;
1974 struct btrfs_delayed_ref_node *ref;
1975 struct btrfs_delayed_ref_head *locked_ref = NULL;
1976 struct btrfs_delayed_extent_op *extent_op;
1977 int ret;
1978 int count = 0;
1979 int must_insert_reserved = 0;
1980
1981 delayed_refs = &trans->transaction->delayed_refs;
1982 while (1) {
1983 if (!locked_ref) {
1984 /* pick a new head ref from the cluster list */
1985 if (list_empty(cluster))
1986 break;
1987
1988 locked_ref = list_entry(cluster->next,
1989 struct btrfs_delayed_ref_head, cluster);
1990
1991 /* grab the lock that says we are going to process
1992 * all the refs for this head */
1993 ret = btrfs_delayed_ref_lock(trans, locked_ref);
1994
1995 /*
1996 * we may have dropped the spin lock to get the head
1997 * mutex lock, and that might have given someone else
1998 * time to free the head. If that's true, it has been
1999 * removed from our list and we can move on.
2000 */
2001 if (ret == -EAGAIN) {
2002 locked_ref = NULL;
2003 count++;
2004 continue;
2005 }
2006 }
2007
2008 /*
2009 * record the must insert reserved flag before we
2010 * drop the spin lock.
2011 */
2012 must_insert_reserved = locked_ref->must_insert_reserved;
2013 locked_ref->must_insert_reserved = 0;
2014
2015 extent_op = locked_ref->extent_op;
2016 locked_ref->extent_op = NULL;
2017
2018 /*
2019 * locked_ref is the head node, so we have to go one
2020 * node back for any delayed ref updates
2021 */
2022 ref = select_delayed_ref(locked_ref);
2023 if (!ref) {
2024 /* All delayed refs have been processed, Go ahead
2025 * and send the head node to run_one_delayed_ref,
2026 * so that any accounting fixes can happen
2027 */
2028 ref = &locked_ref->node;
2029
2030 if (extent_op && must_insert_reserved) {
2031 kfree(extent_op);
2032 extent_op = NULL;
2033 }
2034
2035 if (extent_op) {
2036 spin_unlock(&delayed_refs->lock);
2037
2038 ret = run_delayed_extent_op(trans, root,
2039 ref, extent_op);
2040 BUG_ON(ret);
2041 kfree(extent_op);
2042
2043 cond_resched();
2044 spin_lock(&delayed_refs->lock);
2045 continue;
2046 }
2047
2048 list_del_init(&locked_ref->cluster);
2049 locked_ref = NULL;
2050 }
2051
2052 ref->in_tree = 0;
2053 rb_erase(&ref->rb_node, &delayed_refs->root);
2054 delayed_refs->num_entries--;
2055
2056 spin_unlock(&delayed_refs->lock);
2057
2058 ret = run_one_delayed_ref(trans, root, ref, extent_op,
2059 must_insert_reserved);
2060 BUG_ON(ret);
2061
2062 btrfs_put_delayed_ref(ref);
2063 kfree(extent_op);
2064 count++;
2065
2066 cond_resched();
2067 spin_lock(&delayed_refs->lock);
2068 }
2069 return count;
2070 }
2071
2072 /*
2073 * this starts processing the delayed reference count updates and
2074 * extent insertions we have queued up so far. count can be
2075 * 0, which means to process everything in the tree at the start
2076 * of the run (but not newly added entries), or it can be some target
2077 * number you'd like to process.
2078 */
2079 int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
2080 struct btrfs_root *root, unsigned long count)
2081 {
2082 struct rb_node *node;
2083 struct btrfs_delayed_ref_root *delayed_refs;
2084 struct btrfs_delayed_ref_node *ref;
2085 struct list_head cluster;
2086 int ret;
2087 int run_all = count == (unsigned long)-1;
2088 int run_most = 0;
2089
2090 if (root == root->fs_info->extent_root)
2091 root = root->fs_info->tree_root;
2092
2093 delayed_refs = &trans->transaction->delayed_refs;
2094 INIT_LIST_HEAD(&cluster);
2095 again:
2096 spin_lock(&delayed_refs->lock);
2097 if (count == 0) {
2098 count = delayed_refs->num_entries * 2;
2099 run_most = 1;
2100 }
2101 while (1) {
2102 if (!(run_all || run_most) &&
2103 delayed_refs->num_heads_ready < 64)
2104 break;
2105
2106 /*
2107 * go find something we can process in the rbtree. We start at
2108 * the beginning of the tree, and then build a cluster
2109 * of refs to process starting at the first one we are able to
2110 * lock
2111 */
2112 ret = btrfs_find_ref_cluster(trans, &cluster,
2113 delayed_refs->run_delayed_start);
2114 if (ret)
2115 break;
2116
2117 ret = run_clustered_refs(trans, root, &cluster);
2118 BUG_ON(ret < 0);
2119
2120 count -= min_t(unsigned long, ret, count);
2121
2122 if (count == 0)
2123 break;
2124 }
2125
2126 if (run_all) {
2127 node = rb_first(&delayed_refs->root);
2128 if (!node)
2129 goto out;
2130 count = (unsigned long)-1;
2131
2132 while (node) {
2133 ref = rb_entry(node, struct btrfs_delayed_ref_node,
2134 rb_node);
2135 if (btrfs_delayed_ref_is_head(ref)) {
2136 struct btrfs_delayed_ref_head *head;
2137
2138 head = btrfs_delayed_node_to_head(ref);
2139 atomic_inc(&ref->refs);
2140
2141 spin_unlock(&delayed_refs->lock);
2142 mutex_lock(&head->mutex);
2143 mutex_unlock(&head->mutex);
2144
2145 btrfs_put_delayed_ref(ref);
2146 cond_resched();
2147 goto again;
2148 }
2149 node = rb_next(node);
2150 }
2151 spin_unlock(&delayed_refs->lock);
2152 schedule_timeout(1);
2153 goto again;
2154 }
2155 out:
2156 spin_unlock(&delayed_refs->lock);
2157 return 0;
2158 }
2159
2160 int btrfs_set_disk_extent_flags(struct btrfs_trans_handle *trans,
2161 struct btrfs_root *root,
2162 u64 bytenr, u64 num_bytes, u64 flags,
2163 int is_data)
2164 {
2165 struct btrfs_delayed_extent_op *extent_op;
2166 int ret;
2167
2168 extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
2169 if (!extent_op)
2170 return -ENOMEM;
2171
2172 extent_op->flags_to_set = flags;
2173 extent_op->update_flags = 1;
2174 extent_op->update_key = 0;
2175 extent_op->is_data = is_data ? 1 : 0;
2176
2177 ret = btrfs_add_delayed_extent_op(trans, bytenr, num_bytes, extent_op);
2178 if (ret)
2179 kfree(extent_op);
2180 return ret;
2181 }
2182
2183 static noinline int check_delayed_ref(struct btrfs_trans_handle *trans,
2184 struct btrfs_root *root,
2185 struct btrfs_path *path,
2186 u64 objectid, u64 offset, u64 bytenr)
2187 {
2188 struct btrfs_delayed_ref_head *head;
2189 struct btrfs_delayed_ref_node *ref;
2190 struct btrfs_delayed_data_ref *data_ref;
2191 struct btrfs_delayed_ref_root *delayed_refs;
2192 struct rb_node *node;
2193 int ret = 0;
2194
2195 ret = -ENOENT;
2196 delayed_refs = &trans->transaction->delayed_refs;
2197 spin_lock(&delayed_refs->lock);
2198 head = btrfs_find_delayed_ref_head(trans, bytenr);
2199 if (!head)
2200 goto out;
2201
2202 if (!mutex_trylock(&head->mutex)) {
2203 atomic_inc(&head->node.refs);
2204 spin_unlock(&delayed_refs->lock);
2205
2206 btrfs_release_path(root->fs_info->extent_root, path);
2207
2208 mutex_lock(&head->mutex);
2209 mutex_unlock(&head->mutex);
2210 btrfs_put_delayed_ref(&head->node);
2211 return -EAGAIN;
2212 }
2213
2214 node = rb_prev(&head->node.rb_node);
2215 if (!node)
2216 goto out_unlock;
2217
2218 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
2219
2220 if (ref->bytenr != bytenr)
2221 goto out_unlock;
2222
2223 ret = 1;
2224 if (ref->type != BTRFS_EXTENT_DATA_REF_KEY)
2225 goto out_unlock;
2226
2227 data_ref = btrfs_delayed_node_to_data_ref(ref);
2228
2229 node = rb_prev(node);
2230 if (node) {
2231 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
2232 if (ref->bytenr == bytenr)
2233 goto out_unlock;
2234 }
2235
2236 if (data_ref->root != root->root_key.objectid ||
2237 data_ref->objectid != objectid || data_ref->offset != offset)
2238 goto out_unlock;
2239
2240 ret = 0;
2241 out_unlock:
2242 mutex_unlock(&head->mutex);
2243 out:
2244 spin_unlock(&delayed_refs->lock);
2245 return ret;
2246 }
2247
2248 static noinline int check_committed_ref(struct btrfs_trans_handle *trans,
2249 struct btrfs_root *root,
2250 struct btrfs_path *path,
2251 u64 objectid, u64 offset, u64 bytenr)
2252 {
2253 struct btrfs_root *extent_root = root->fs_info->extent_root;
2254 struct extent_buffer *leaf;
2255 struct btrfs_extent_data_ref *ref;
2256 struct btrfs_extent_inline_ref *iref;
2257 struct btrfs_extent_item *ei;
2258 struct btrfs_key key;
2259 u32 item_size;
2260 int ret;
2261
2262 key.objectid = bytenr;
2263 key.offset = (u64)-1;
2264 key.type = BTRFS_EXTENT_ITEM_KEY;
2265
2266 ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
2267 if (ret < 0)
2268 goto out;
2269 BUG_ON(ret == 0);
2270
2271 ret = -ENOENT;
2272 if (path->slots[0] == 0)
2273 goto out;
2274
2275 path->slots[0]--;
2276 leaf = path->nodes[0];
2277 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
2278
2279 if (key.objectid != bytenr || key.type != BTRFS_EXTENT_ITEM_KEY)
2280 goto out;
2281
2282 ret = 1;
2283 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
2284 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
2285 if (item_size < sizeof(*ei)) {
2286 WARN_ON(item_size != sizeof(struct btrfs_extent_item_v0));
2287 goto out;
2288 }
2289 #endif
2290 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
2291
2292 if (item_size != sizeof(*ei) +
2293 btrfs_extent_inline_ref_size(BTRFS_EXTENT_DATA_REF_KEY))
2294 goto out;
2295
2296 if (btrfs_extent_generation(leaf, ei) <=
2297 btrfs_root_last_snapshot(&root->root_item))
2298 goto out;
2299
2300 iref = (struct btrfs_extent_inline_ref *)(ei + 1);
2301 if (btrfs_extent_inline_ref_type(leaf, iref) !=
2302 BTRFS_EXTENT_DATA_REF_KEY)
2303 goto out;
2304
2305 ref = (struct btrfs_extent_data_ref *)(&iref->offset);
2306 if (btrfs_extent_refs(leaf, ei) !=
2307 btrfs_extent_data_ref_count(leaf, ref) ||
2308 btrfs_extent_data_ref_root(leaf, ref) !=
2309 root->root_key.objectid ||
2310 btrfs_extent_data_ref_objectid(leaf, ref) != objectid ||
2311 btrfs_extent_data_ref_offset(leaf, ref) != offset)
2312 goto out;
2313
2314 ret = 0;
2315 out:
2316 return ret;
2317 }
2318
2319 int btrfs_cross_ref_exist(struct btrfs_trans_handle *trans,
2320 struct btrfs_root *root,
2321 u64 objectid, u64 offset, u64 bytenr)
2322 {
2323 struct btrfs_path *path;
2324 int ret;
2325 int ret2;
2326
2327 path = btrfs_alloc_path();
2328 if (!path)
2329 return -ENOENT;
2330
2331 do {
2332 ret = check_committed_ref(trans, root, path, objectid,
2333 offset, bytenr);
2334 if (ret && ret != -ENOENT)
2335 goto out;
2336
2337 ret2 = check_delayed_ref(trans, root, path, objectid,
2338 offset, bytenr);
2339 } while (ret2 == -EAGAIN);
2340
2341 if (ret2 && ret2 != -ENOENT) {
2342 ret = ret2;
2343 goto out;
2344 }
2345
2346 if (ret != -ENOENT || ret2 != -ENOENT)
2347 ret = 0;
2348 out:
2349 btrfs_free_path(path);
2350 return ret;
2351 }
2352
2353 #if 0
2354 int btrfs_cache_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2355 struct extent_buffer *buf, u32 nr_extents)
2356 {
2357 struct btrfs_key key;
2358 struct btrfs_file_extent_item *fi;
2359 u64 root_gen;
2360 u32 nritems;
2361 int i;
2362 int level;
2363 int ret = 0;
2364 int shared = 0;
2365
2366 if (!root->ref_cows)
2367 return 0;
2368
2369 if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
2370 shared = 0;
2371 root_gen = root->root_key.offset;
2372 } else {
2373 shared = 1;
2374 root_gen = trans->transid - 1;
2375 }
2376
2377 level = btrfs_header_level(buf);
2378 nritems = btrfs_header_nritems(buf);
2379
2380 if (level == 0) {
2381 struct btrfs_leaf_ref *ref;
2382 struct btrfs_extent_info *info;
2383
2384 ref = btrfs_alloc_leaf_ref(root, nr_extents);
2385 if (!ref) {
2386 ret = -ENOMEM;
2387 goto out;
2388 }
2389
2390 ref->root_gen = root_gen;
2391 ref->bytenr = buf->start;
2392 ref->owner = btrfs_header_owner(buf);
2393 ref->generation = btrfs_header_generation(buf);
2394 ref->nritems = nr_extents;
2395 info = ref->extents;
2396
2397 for (i = 0; nr_extents > 0 && i < nritems; i++) {
2398 u64 disk_bytenr;
2399 btrfs_item_key_to_cpu(buf, &key, i);
2400 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
2401 continue;
2402 fi = btrfs_item_ptr(buf, i,
2403 struct btrfs_file_extent_item);
2404 if (btrfs_file_extent_type(buf, fi) ==
2405 BTRFS_FILE_EXTENT_INLINE)
2406 continue;
2407 disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
2408 if (disk_bytenr == 0)
2409 continue;
2410
2411 info->bytenr = disk_bytenr;
2412 info->num_bytes =
2413 btrfs_file_extent_disk_num_bytes(buf, fi);
2414 info->objectid = key.objectid;
2415 info->offset = key.offset;
2416 info++;
2417 }
2418
2419 ret = btrfs_add_leaf_ref(root, ref, shared);
2420 if (ret == -EEXIST && shared) {
2421 struct btrfs_leaf_ref *old;
2422 old = btrfs_lookup_leaf_ref(root, ref->bytenr);
2423 BUG_ON(!old);
2424 btrfs_remove_leaf_ref(root, old);
2425 btrfs_free_leaf_ref(root, old);
2426 ret = btrfs_add_leaf_ref(root, ref, shared);
2427 }
2428 WARN_ON(ret);
2429 btrfs_free_leaf_ref(root, ref);
2430 }
2431 out:
2432 return ret;
2433 }
2434
2435 /* when a block goes through cow, we update the reference counts of
2436 * everything that block points to. The internal pointers of the block
2437 * can be in just about any order, and it is likely to have clusters of
2438 * things that are close together and clusters of things that are not.
2439 *
2440 * To help reduce the seeks that come with updating all of these reference
2441 * counts, sort them by byte number before actual updates are done.
2442 *
2443 * struct refsort is used to match byte number to slot in the btree block.
2444 * we sort based on the byte number and then use the slot to actually
2445 * find the item.
2446 *
2447 * struct refsort is smaller than strcut btrfs_item and smaller than
2448 * struct btrfs_key_ptr. Since we're currently limited to the page size
2449 * for a btree block, there's no way for a kmalloc of refsorts for a
2450 * single node to be bigger than a page.
2451 */
2452 struct refsort {
2453 u64 bytenr;
2454 u32 slot;
2455 };
2456
2457 /*
2458 * for passing into sort()
2459 */
2460 static int refsort_cmp(const void *a_void, const void *b_void)
2461 {
2462 const struct refsort *a = a_void;
2463 const struct refsort *b = b_void;
2464
2465 if (a->bytenr < b->bytenr)
2466 return -1;
2467 if (a->bytenr > b->bytenr)
2468 return 1;
2469 return 0;
2470 }
2471 #endif
2472
2473 static int __btrfs_mod_ref(struct btrfs_trans_handle *trans,
2474 struct btrfs_root *root,
2475 struct extent_buffer *buf,
2476 int full_backref, int inc)
2477 {
2478 u64 bytenr;
2479 u64 num_bytes;
2480 u64 parent;
2481 u64 ref_root;
2482 u32 nritems;
2483 struct btrfs_key key;
2484 struct btrfs_file_extent_item *fi;
2485 int i;
2486 int level;
2487 int ret = 0;
2488 int (*process_func)(struct btrfs_trans_handle *, struct btrfs_root *,
2489 u64, u64, u64, u64, u64, u64);
2490
2491 ref_root = btrfs_header_owner(buf);
2492 nritems = btrfs_header_nritems(buf);
2493 level = btrfs_header_level(buf);
2494
2495 if (!root->ref_cows && level == 0)
2496 return 0;
2497
2498 if (inc)
2499 process_func = btrfs_inc_extent_ref;
2500 else
2501 process_func = btrfs_free_extent;
2502
2503 if (full_backref)
2504 parent = buf->start;
2505 else
2506 parent = 0;
2507
2508 for (i = 0; i < nritems; i++) {
2509 if (level == 0) {
2510 btrfs_item_key_to_cpu(buf, &key, i);
2511 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
2512 continue;
2513 fi = btrfs_item_ptr(buf, i,
2514 struct btrfs_file_extent_item);
2515 if (btrfs_file_extent_type(buf, fi) ==
2516 BTRFS_FILE_EXTENT_INLINE)
2517 continue;
2518 bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
2519 if (bytenr == 0)
2520 continue;
2521
2522 num_bytes = btrfs_file_extent_disk_num_bytes(buf, fi);
2523 key.offset -= btrfs_file_extent_offset(buf, fi);
2524 ret = process_func(trans, root, bytenr, num_bytes,
2525 parent, ref_root, key.objectid,
2526 key.offset);
2527 if (ret)
2528 goto fail;
2529 } else {
2530 bytenr = btrfs_node_blockptr(buf, i);
2531 num_bytes = btrfs_level_size(root, level - 1);
2532 ret = process_func(trans, root, bytenr, num_bytes,
2533 parent, ref_root, level - 1, 0);
2534 if (ret)
2535 goto fail;
2536 }
2537 }
2538 return 0;
2539 fail:
2540 BUG();
2541 return ret;
2542 }
2543
2544 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2545 struct extent_buffer *buf, int full_backref)
2546 {
2547 return __btrfs_mod_ref(trans, root, buf, full_backref, 1);
2548 }
2549
2550 int btrfs_dec_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2551 struct extent_buffer *buf, int full_backref)
2552 {
2553 return __btrfs_mod_ref(trans, root, buf, full_backref, 0);
2554 }
2555
2556 static int write_one_cache_group(struct btrfs_trans_handle *trans,
2557 struct btrfs_root *root,
2558 struct btrfs_path *path,
2559 struct btrfs_block_group_cache *cache)
2560 {
2561 int ret;
2562 struct btrfs_root *extent_root = root->fs_info->extent_root;
2563 unsigned long bi;
2564 struct extent_buffer *leaf;
2565
2566 ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1);
2567 if (ret < 0)
2568 goto fail;
2569 BUG_ON(ret);
2570
2571 leaf = path->nodes[0];
2572 bi = btrfs_item_ptr_offset(leaf, path->slots[0]);
2573 write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item));
2574 btrfs_mark_buffer_dirty(leaf);
2575 btrfs_release_path(extent_root, path);
2576 fail:
2577 if (ret)
2578 return ret;
2579 return 0;
2580
2581 }
2582
2583 static struct btrfs_block_group_cache *
2584 next_block_group(struct btrfs_root *root,
2585 struct btrfs_block_group_cache *cache)
2586 {
2587 struct rb_node *node;
2588 spin_lock(&root->fs_info->block_group_cache_lock);
2589 node = rb_next(&cache->cache_node);
2590 btrfs_put_block_group(cache);
2591 if (node) {
2592 cache = rb_entry(node, struct btrfs_block_group_cache,
2593 cache_node);
2594 btrfs_get_block_group(cache);
2595 } else
2596 cache = NULL;
2597 spin_unlock(&root->fs_info->block_group_cache_lock);
2598 return cache;
2599 }
2600
2601 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
2602 struct btrfs_root *root)
2603 {
2604 struct btrfs_block_group_cache *cache;
2605 int err = 0;
2606 struct btrfs_path *path;
2607 u64 last = 0;
2608
2609 path = btrfs_alloc_path();
2610 if (!path)
2611 return -ENOMEM;
2612
2613 while (1) {
2614 if (last == 0) {
2615 err = btrfs_run_delayed_refs(trans, root,
2616 (unsigned long)-1);
2617 BUG_ON(err);
2618 }
2619
2620 cache = btrfs_lookup_first_block_group(root->fs_info, last);
2621 while (cache) {
2622 if (cache->dirty)
2623 break;
2624 cache = next_block_group(root, cache);
2625 }
2626 if (!cache) {
2627 if (last == 0)
2628 break;
2629 last = 0;
2630 continue;
2631 }
2632
2633 cache->dirty = 0;
2634 last = cache->key.objectid + cache->key.offset;
2635
2636 err = write_one_cache_group(trans, root, path, cache);
2637 BUG_ON(err);
2638 btrfs_put_block_group(cache);
2639 }
2640
2641 btrfs_free_path(path);
2642 return 0;
2643 }
2644
2645 int btrfs_extent_readonly(struct btrfs_root *root, u64 bytenr)
2646 {
2647 struct btrfs_block_group_cache *block_group;
2648 int readonly = 0;
2649
2650 block_group = btrfs_lookup_block_group(root->fs_info, bytenr);
2651 if (!block_group || block_group->ro)
2652 readonly = 1;
2653 if (block_group)
2654 btrfs_put_block_group(block_group);
2655 return readonly;
2656 }
2657
2658 static int update_space_info(struct btrfs_fs_info *info, u64 flags,
2659 u64 total_bytes, u64 bytes_used,
2660 struct btrfs_space_info **space_info)
2661 {
2662 struct btrfs_space_info *found;
2663
2664 found = __find_space_info(info, flags);
2665 if (found) {
2666 spin_lock(&found->lock);
2667 found->total_bytes += total_bytes;
2668 found->bytes_used += bytes_used;
2669 found->full = 0;
2670 spin_unlock(&found->lock);
2671 *space_info = found;
2672 return 0;
2673 }
2674 found = kzalloc(sizeof(*found), GFP_NOFS);
2675 if (!found)
2676 return -ENOMEM;
2677
2678 INIT_LIST_HEAD(&found->block_groups);
2679 init_rwsem(&found->groups_sem);
2680 init_waitqueue_head(&found->flush_wait);
2681 init_waitqueue_head(&found->allocate_wait);
2682 spin_lock_init(&found->lock);
2683 found->flags = flags;
2684 found->total_bytes = total_bytes;
2685 found->bytes_used = bytes_used;
2686 found->bytes_pinned = 0;
2687 found->bytes_reserved = 0;
2688 found->bytes_readonly = 0;
2689 found->bytes_delalloc = 0;
2690 found->full = 0;
2691 found->force_alloc = 0;
2692 *space_info = found;
2693 list_add_rcu(&found->list, &info->space_info);
2694 atomic_set(&found->caching_threads, 0);
2695 return 0;
2696 }
2697
2698 static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags)
2699 {
2700 u64 extra_flags = flags & (BTRFS_BLOCK_GROUP_RAID0 |
2701 BTRFS_BLOCK_GROUP_RAID1 |
2702 BTRFS_BLOCK_GROUP_RAID10 |
2703 BTRFS_BLOCK_GROUP_DUP);
2704 if (extra_flags) {
2705 if (flags & BTRFS_BLOCK_GROUP_DATA)
2706 fs_info->avail_data_alloc_bits |= extra_flags;
2707 if (flags & BTRFS_BLOCK_GROUP_METADATA)
2708 fs_info->avail_metadata_alloc_bits |= extra_flags;
2709 if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
2710 fs_info->avail_system_alloc_bits |= extra_flags;
2711 }
2712 }
2713
2714 static void set_block_group_readonly(struct btrfs_block_group_cache *cache)
2715 {
2716 spin_lock(&cache->space_info->lock);
2717 spin_lock(&cache->lock);
2718 if (!cache->ro) {
2719 cache->space_info->bytes_readonly += cache->key.offset -
2720 btrfs_block_group_used(&cache->item);
2721 cache->ro = 1;
2722 }
2723 spin_unlock(&cache->lock);
2724 spin_unlock(&cache->space_info->lock);
2725 }
2726
2727 u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags)
2728 {
2729 u64 num_devices = root->fs_info->fs_devices->rw_devices;
2730
2731 if (num_devices == 1)
2732 flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0);
2733 if (num_devices < 4)
2734 flags &= ~BTRFS_BLOCK_GROUP_RAID10;
2735
2736 if ((flags & BTRFS_BLOCK_GROUP_DUP) &&
2737 (flags & (BTRFS_BLOCK_GROUP_RAID1 |
2738 BTRFS_BLOCK_GROUP_RAID10))) {
2739 flags &= ~BTRFS_BLOCK_GROUP_DUP;
2740 }
2741
2742 if ((flags & BTRFS_BLOCK_GROUP_RAID1) &&
2743 (flags & BTRFS_BLOCK_GROUP_RAID10)) {
2744 flags &= ~BTRFS_BLOCK_GROUP_RAID1;
2745 }
2746
2747 if ((flags & BTRFS_BLOCK_GROUP_RAID0) &&
2748 ((flags & BTRFS_BLOCK_GROUP_RAID1) |
2749 (flags & BTRFS_BLOCK_GROUP_RAID10) |
2750 (flags & BTRFS_BLOCK_GROUP_DUP)))
2751 flags &= ~BTRFS_BLOCK_GROUP_RAID0;
2752 return flags;
2753 }
2754
2755 static u64 btrfs_get_alloc_profile(struct btrfs_root *root, u64 data)
2756 {
2757 struct btrfs_fs_info *info = root->fs_info;
2758 u64 alloc_profile;
2759
2760 if (data) {
2761 alloc_profile = info->avail_data_alloc_bits &
2762 info->data_alloc_profile;
2763 data = BTRFS_BLOCK_GROUP_DATA | alloc_profile;
2764 } else if (root == root->fs_info->chunk_root) {
2765 alloc_profile = info->avail_system_alloc_bits &
2766 info->system_alloc_profile;
2767 data = BTRFS_BLOCK_GROUP_SYSTEM | alloc_profile;
2768 } else {
2769 alloc_profile = info->avail_metadata_alloc_bits &
2770 info->metadata_alloc_profile;
2771 data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile;
2772 }
2773
2774 return btrfs_reduce_alloc_profile(root, data);
2775 }
2776
2777 void btrfs_set_inode_space_info(struct btrfs_root *root, struct inode *inode)
2778 {
2779 u64 alloc_target;
2780
2781 alloc_target = btrfs_get_alloc_profile(root, 1);
2782 BTRFS_I(inode)->space_info = __find_space_info(root->fs_info,
2783 alloc_target);
2784 }
2785
2786 static u64 calculate_bytes_needed(struct btrfs_root *root, int num_items)
2787 {
2788 u64 num_bytes;
2789 int level;
2790
2791 level = BTRFS_MAX_LEVEL - 2;
2792 /*
2793 * NOTE: these calculations are absolutely the worst possible case.
2794 * This assumes that _every_ item we insert will require a new leaf, and
2795 * that the tree has grown to its maximum level size.
2796 */
2797
2798 /*
2799 * for every item we insert we could insert both an extent item and a
2800 * extent ref item. Then for ever item we insert, we will need to cow
2801 * both the original leaf, plus the leaf to the left and right of it.
2802 *
2803 * Unless we are talking about the extent root, then we just want the
2804 * number of items * 2, since we just need the extent item plus its ref.
2805 */
2806 if (root == root->fs_info->extent_root)
2807 num_bytes = num_items * 2;
2808 else
2809 num_bytes = (num_items + (2 * num_items)) * 3;
2810
2811 /*
2812 * num_bytes is total number of leaves we could need times the leaf
2813 * size, and then for every leaf we could end up cow'ing 2 nodes per
2814 * level, down to the leaf level.
2815 */
2816 num_bytes = (num_bytes * root->leafsize) +
2817 (num_bytes * (level * 2)) * root->nodesize;
2818
2819 return num_bytes;
2820 }
2821
2822 /*
2823 * Unreserve metadata space for delalloc. If we have less reserved credits than
2824 * we have extents, this function does nothing.
2825 */
2826 int btrfs_unreserve_metadata_for_delalloc(struct btrfs_root *root,
2827 struct inode *inode, int num_items)
2828 {
2829 struct btrfs_fs_info *info = root->fs_info;
2830 struct btrfs_space_info *meta_sinfo;
2831 u64 num_bytes;
2832 u64 alloc_target;
2833 bool bug = false;
2834
2835 /* get the space info for where the metadata will live */
2836 alloc_target = btrfs_get_alloc_profile(root, 0);
2837 meta_sinfo = __find_space_info(info, alloc_target);
2838
2839 num_bytes = calculate_bytes_needed(root->fs_info->extent_root,
2840 num_items);
2841
2842 spin_lock(&meta_sinfo->lock);
2843 spin_lock(&BTRFS_I(inode)->accounting_lock);
2844 if (BTRFS_I(inode)->reserved_extents <=
2845 BTRFS_I(inode)->outstanding_extents) {
2846 spin_unlock(&BTRFS_I(inode)->accounting_lock);
2847 spin_unlock(&meta_sinfo->lock);
2848 return 0;
2849 }
2850 spin_unlock(&BTRFS_I(inode)->accounting_lock);
2851
2852 BTRFS_I(inode)->reserved_extents -= num_items;
2853 BUG_ON(BTRFS_I(inode)->reserved_extents < 0);
2854
2855 if (meta_sinfo->bytes_delalloc < num_bytes) {
2856 bug = true;
2857 meta_sinfo->bytes_delalloc = 0;
2858 } else {
2859 meta_sinfo->bytes_delalloc -= num_bytes;
2860 }
2861 spin_unlock(&meta_sinfo->lock);
2862
2863 BUG_ON(bug);
2864
2865 return 0;
2866 }
2867
2868 static void check_force_delalloc(struct btrfs_space_info *meta_sinfo)
2869 {
2870 u64 thresh;
2871
2872 thresh = meta_sinfo->bytes_used + meta_sinfo->bytes_reserved +
2873 meta_sinfo->bytes_pinned + meta_sinfo->bytes_readonly +
2874 meta_sinfo->bytes_super + meta_sinfo->bytes_root +
2875 meta_sinfo->bytes_may_use;
2876
2877 thresh = meta_sinfo->total_bytes - thresh;
2878 thresh *= 80;
2879 do_div(thresh, 100);
2880 if (thresh <= meta_sinfo->bytes_delalloc)
2881 meta_sinfo->force_delalloc = 1;
2882 else
2883 meta_sinfo->force_delalloc = 0;
2884 }
2885
2886 struct async_flush {
2887 struct btrfs_root *root;
2888 struct btrfs_space_info *info;
2889 struct btrfs_work work;
2890 };
2891
2892 static noinline void flush_delalloc_async(struct btrfs_work *work)
2893 {
2894 struct async_flush *async;
2895 struct btrfs_root *root;
2896 struct btrfs_space_info *info;
2897
2898 async = container_of(work, struct async_flush, work);
2899 root = async->root;
2900 info = async->info;
2901
2902 btrfs_start_delalloc_inodes(root, 0);
2903 wake_up(&info->flush_wait);
2904 btrfs_wait_ordered_extents(root, 0, 0);
2905
2906 spin_lock(&info->lock);
2907 info->flushing = 0;
2908 spin_unlock(&info->lock);
2909 wake_up(&info->flush_wait);
2910
2911 kfree(async);
2912 }
2913
2914 static void wait_on_flush(struct btrfs_space_info *info)
2915 {
2916 DEFINE_WAIT(wait);
2917 u64 used;
2918
2919 while (1) {
2920 prepare_to_wait(&info->flush_wait, &wait,
2921 TASK_UNINTERRUPTIBLE);
2922 spin_lock(&info->lock);
2923 if (!info->flushing) {
2924 spin_unlock(&info->lock);
2925 break;
2926 }
2927
2928 used = info->bytes_used + info->bytes_reserved +
2929 info->bytes_pinned + info->bytes_readonly +
2930 info->bytes_super + info->bytes_root +
2931 info->bytes_may_use + info->bytes_delalloc;
2932 if (used < info->total_bytes) {
2933 spin_unlock(&info->lock);
2934 break;
2935 }
2936 spin_unlock(&info->lock);
2937 schedule();
2938 }
2939 finish_wait(&info->flush_wait, &wait);
2940 }
2941
2942 static void flush_delalloc(struct btrfs_root *root,
2943 struct btrfs_space_info *info)
2944 {
2945 struct async_flush *async;
2946 bool wait = false;
2947
2948 spin_lock(&info->lock);
2949
2950 if (!info->flushing)
2951 info->flushing = 1;
2952 else
2953 wait = true;
2954
2955 spin_unlock(&info->lock);
2956
2957 if (wait) {
2958 wait_on_flush(info);
2959 return;
2960 }
2961
2962 async = kzalloc(sizeof(*async), GFP_NOFS);
2963 if (!async)
2964 goto flush;
2965
2966 async->root = root;
2967 async->info = info;
2968 async->work.func = flush_delalloc_async;
2969
2970 btrfs_queue_worker(&root->fs_info->enospc_workers,
2971 &async->work);
2972 wait_on_flush(info);
2973 return;
2974
2975 flush:
2976 btrfs_start_delalloc_inodes(root, 0);
2977 btrfs_wait_ordered_extents(root, 0, 0);
2978
2979 spin_lock(&info->lock);
2980 info->flushing = 0;
2981 spin_unlock(&info->lock);
2982 wake_up(&info->flush_wait);
2983 }
2984
2985 static int maybe_allocate_chunk(struct btrfs_root *root,
2986 struct btrfs_space_info *info)
2987 {
2988 struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
2989 struct btrfs_trans_handle *trans;
2990 bool wait = false;
2991 int ret = 0;
2992 u64 min_metadata;
2993 u64 free_space;
2994
2995 free_space = btrfs_super_total_bytes(disk_super);
2996 /*
2997 * we allow the metadata to grow to a max of either 10gb or 5% of the
2998 * space in the volume.
2999 */
3000 min_metadata = min((u64)10 * 1024 * 1024 * 1024,
3001 div64_u64(free_space * 5, 100));
3002 if (info->total_bytes >= min_metadata) {
3003 spin_unlock(&info->lock);
3004 return 0;
3005 }
3006
3007 if (info->full) {
3008 spin_unlock(&info->lock);
3009 return 0;
3010 }
3011
3012 if (!info->allocating_chunk) {
3013 info->force_alloc = 1;
3014 info->allocating_chunk = 1;
3015 } else {
3016 wait = true;
3017 }
3018
3019 spin_unlock(&info->lock);
3020
3021 if (wait) {
3022 wait_event(info->allocate_wait,
3023 !info->allocating_chunk);
3024 return 1;
3025 }
3026
3027 trans = btrfs_start_transaction(root, 1);
3028 if (!trans) {
3029 ret = -ENOMEM;
3030 goto out;
3031 }
3032
3033 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
3034 4096 + 2 * 1024 * 1024,
3035 info->flags, 0);
3036 btrfs_end_transaction(trans, root);
3037 if (ret)
3038 goto out;
3039 out:
3040 spin_lock(&info->lock);
3041 info->allocating_chunk = 0;
3042 spin_unlock(&info->lock);
3043 wake_up(&info->allocate_wait);
3044
3045 if (ret)
3046 return 0;
3047 return 1;
3048 }
3049
3050 /*
3051 * Reserve metadata space for delalloc.
3052 */
3053 int btrfs_reserve_metadata_for_delalloc(struct btrfs_root *root,
3054 struct inode *inode, int num_items)
3055 {
3056 struct btrfs_fs_info *info = root->fs_info;
3057 struct btrfs_space_info *meta_sinfo;
3058 u64 num_bytes;
3059 u64 used;
3060 u64 alloc_target;
3061 int flushed = 0;
3062 int force_delalloc;
3063
3064 /* get the space info for where the metadata will live */
3065 alloc_target = btrfs_get_alloc_profile(root, 0);
3066 meta_sinfo = __find_space_info(info, alloc_target);
3067
3068 num_bytes = calculate_bytes_needed(root->fs_info->extent_root,
3069 num_items);
3070 again:
3071 spin_lock(&meta_sinfo->lock);
3072
3073 force_delalloc = meta_sinfo->force_delalloc;
3074
3075 if (unlikely(!meta_sinfo->bytes_root))
3076 meta_sinfo->bytes_root = calculate_bytes_needed(root, 6);
3077
3078 if (!flushed)
3079 meta_sinfo->bytes_delalloc += num_bytes;
3080
3081 used = meta_sinfo->bytes_used + meta_sinfo->bytes_reserved +
3082 meta_sinfo->bytes_pinned + meta_sinfo->bytes_readonly +
3083 meta_sinfo->bytes_super + meta_sinfo->bytes_root +
3084 meta_sinfo->bytes_may_use + meta_sinfo->bytes_delalloc;
3085
3086 if (used > meta_sinfo->total_bytes) {
3087 flushed++;
3088
3089 if (flushed == 1) {
3090 if (maybe_allocate_chunk(root, meta_sinfo))
3091 goto again;
3092 flushed++;
3093 } else {
3094 spin_unlock(&meta_sinfo->lock);
3095 }
3096
3097 if (flushed == 2) {
3098 filemap_flush(inode->i_mapping);
3099 goto again;
3100 } else if (flushed == 3) {
3101 flush_delalloc(root, meta_sinfo);
3102 goto again;
3103 }
3104 spin_lock(&meta_sinfo->lock);
3105 meta_sinfo->bytes_delalloc -= num_bytes;
3106 spin_unlock(&meta_sinfo->lock);
3107 printk(KERN_ERR "enospc, has %d, reserved %d\n",
3108 BTRFS_I(inode)->outstanding_extents,
3109 BTRFS_I(inode)->reserved_extents);
3110 dump_space_info(meta_sinfo, 0, 0);
3111 return -ENOSPC;
3112 }
3113
3114 BTRFS_I(inode)->reserved_extents += num_items;
3115 check_force_delalloc(meta_sinfo);
3116 spin_unlock(&meta_sinfo->lock);
3117
3118 if (!flushed && force_delalloc)
3119 filemap_flush(inode->i_mapping);
3120
3121 return 0;
3122 }
3123
3124 /*
3125 * unreserve num_items number of items worth of metadata space. This needs to
3126 * be paired with btrfs_reserve_metadata_space.
3127 *
3128 * NOTE: if you have the option, run this _AFTER_ you do a
3129 * btrfs_end_transaction, since btrfs_end_transaction will run delayed ref
3130 * oprations which will result in more used metadata, so we want to make sure we
3131 * can do that without issue.
3132 */
3133 int btrfs_unreserve_metadata_space(struct btrfs_root *root, int num_items)
3134 {
3135 struct btrfs_fs_info *info = root->fs_info;
3136 struct btrfs_space_info *meta_sinfo;
3137 u64 num_bytes;
3138 u64 alloc_target;
3139 bool bug = false;
3140
3141 /* get the space info for where the metadata will live */
3142 alloc_target = btrfs_get_alloc_profile(root, 0);
3143 meta_sinfo = __find_space_info(info, alloc_target);
3144
3145 num_bytes = calculate_bytes_needed(root, num_items);
3146
3147 spin_lock(&meta_sinfo->lock);
3148 if (meta_sinfo->bytes_may_use < num_bytes) {
3149 bug = true;
3150 meta_sinfo->bytes_may_use = 0;
3151 } else {
3152 meta_sinfo->bytes_may_use -= num_bytes;
3153 }
3154 spin_unlock(&meta_sinfo->lock);
3155
3156 BUG_ON(bug);
3157
3158 return 0;
3159 }
3160
3161 /*
3162 * Reserve some metadata space for use. We'll calculate the worste case number
3163 * of bytes that would be needed to modify num_items number of items. If we
3164 * have space, fantastic, if not, you get -ENOSPC. Please call
3165 * btrfs_unreserve_metadata_space when you are done for the _SAME_ number of
3166 * items you reserved, since whatever metadata you needed should have already
3167 * been allocated.
3168 *
3169 * This will commit the transaction to make more space if we don't have enough
3170 * metadata space. THe only time we don't do this is if we're reserving space
3171 * inside of a transaction, then we will just return -ENOSPC and it is the
3172 * callers responsibility to handle it properly.
3173 */
3174 int btrfs_reserve_metadata_space(struct btrfs_root *root, int num_items)
3175 {
3176 struct btrfs_fs_info *info = root->fs_info;
3177 struct btrfs_space_info *meta_sinfo;
3178 u64 num_bytes;
3179 u64 used;
3180 u64 alloc_target;
3181 int retries = 0;
3182
3183 /* get the space info for where the metadata will live */
3184 alloc_target = btrfs_get_alloc_profile(root, 0);
3185 meta_sinfo = __find_space_info(info, alloc_target);
3186
3187 num_bytes = calculate_bytes_needed(root, num_items);
3188 again:
3189 spin_lock(&meta_sinfo->lock);
3190
3191 if (unlikely(!meta_sinfo->bytes_root))
3192 meta_sinfo->bytes_root = calculate_bytes_needed(root, 6);
3193
3194 if (!retries)
3195 meta_sinfo->bytes_may_use += num_bytes;
3196
3197 used = meta_sinfo->bytes_used + meta_sinfo->bytes_reserved +
3198 meta_sinfo->bytes_pinned + meta_sinfo->bytes_readonly +
3199 meta_sinfo->bytes_super + meta_sinfo->bytes_root +
3200 meta_sinfo->bytes_may_use + meta_sinfo->bytes_delalloc;
3201
3202 if (used > meta_sinfo->total_bytes) {
3203 retries++;
3204 if (retries == 1) {
3205 if (maybe_allocate_chunk(root, meta_sinfo))
3206 goto again;
3207 retries++;
3208 } else {
3209 spin_unlock(&meta_sinfo->lock);
3210 }
3211
3212 if (retries == 2) {
3213 flush_delalloc(root, meta_sinfo);
3214 goto again;
3215 }
3216 spin_lock(&meta_sinfo->lock);
3217 meta_sinfo->bytes_may_use -= num_bytes;
3218 spin_unlock(&meta_sinfo->lock);
3219
3220 dump_space_info(meta_sinfo, 0, 0);
3221 return -ENOSPC;
3222 }
3223
3224 check_force_delalloc(meta_sinfo);
3225 spin_unlock(&meta_sinfo->lock);
3226
3227 return 0;
3228 }
3229
3230 /*
3231 * This will check the space that the inode allocates from to make sure we have
3232 * enough space for bytes.
3233 */
3234 int btrfs_check_data_free_space(struct btrfs_root *root, struct inode *inode,
3235 u64 bytes)
3236 {
3237 struct btrfs_space_info *data_sinfo;
3238 u64 used;
3239 int ret = 0, committed = 0, flushed = 0;
3240
3241 /* make sure bytes are sectorsize aligned */
3242 bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
3243
3244 data_sinfo = BTRFS_I(inode)->space_info;
3245 if (!data_sinfo)
3246 goto alloc;
3247
3248 again:
3249 /* make sure we have enough space to handle the data first */
3250 spin_lock(&data_sinfo->lock);
3251 used = data_sinfo->bytes_used + data_sinfo->bytes_delalloc +
3252 data_sinfo->bytes_reserved + data_sinfo->bytes_pinned +
3253 data_sinfo->bytes_readonly + data_sinfo->bytes_may_use +
3254 data_sinfo->bytes_super;
3255
3256 if (used + bytes > data_sinfo->total_bytes) {
3257 struct btrfs_trans_handle *trans;
3258
3259 if (!flushed) {
3260 spin_unlock(&data_sinfo->lock);
3261 flush_delalloc(root, data_sinfo);
3262 flushed = 1;
3263 goto again;
3264 }
3265
3266 /*
3267 * if we don't have enough free bytes in this space then we need
3268 * to alloc a new chunk.
3269 */
3270 if (!data_sinfo->full) {
3271 u64 alloc_target;
3272
3273 data_sinfo->force_alloc = 1;
3274 spin_unlock(&data_sinfo->lock);
3275 alloc:
3276 alloc_target = btrfs_get_alloc_profile(root, 1);
3277 trans = btrfs_start_transaction(root, 1);
3278 if (!trans)
3279 return -ENOMEM;
3280
3281 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
3282 bytes + 2 * 1024 * 1024,
3283 alloc_target, 0);
3284 btrfs_end_transaction(trans, root);
3285 if (ret)
3286 return ret;
3287
3288 if (!data_sinfo) {
3289 btrfs_set_inode_space_info(root, inode);
3290 data_sinfo = BTRFS_I(inode)->space_info;
3291 }
3292 goto again;
3293 }
3294 spin_unlock(&data_sinfo->lock);
3295
3296 /* commit the current transaction and try again */
3297 if (!committed && !root->fs_info->open_ioctl_trans) {
3298 committed = 1;
3299 trans = btrfs_join_transaction(root, 1);
3300 if (!trans)
3301 return -ENOMEM;
3302 ret = btrfs_commit_transaction(trans, root);
3303 if (ret)
3304 return ret;
3305 goto again;
3306 }
3307
3308 printk(KERN_ERR "no space left, need %llu, %llu delalloc bytes"
3309 ", %llu bytes_used, %llu bytes_reserved, "
3310 "%llu bytes_pinned, %llu bytes_readonly, %llu may use "
3311 "%llu total\n", (unsigned long long)bytes,
3312 (unsigned long long)data_sinfo->bytes_delalloc,
3313 (unsigned long long)data_sinfo->bytes_used,
3314 (unsigned long long)data_sinfo->bytes_reserved,
3315 (unsigned long long)data_sinfo->bytes_pinned,
3316 (unsigned long long)data_sinfo->bytes_readonly,
3317 (unsigned long long)data_sinfo->bytes_may_use,
3318 (unsigned long long)data_sinfo->total_bytes);
3319 return -ENOSPC;
3320 }
3321 data_sinfo->bytes_may_use += bytes;
3322 BTRFS_I(inode)->reserved_bytes += bytes;
3323 spin_unlock(&data_sinfo->lock);
3324
3325 return 0;
3326 }
3327
3328 /*
3329 * if there was an error for whatever reason after calling
3330 * btrfs_check_data_free_space, call this so we can cleanup the counters.
3331 */
3332 void btrfs_free_reserved_data_space(struct btrfs_root *root,
3333 struct inode *inode, u64 bytes)
3334 {
3335 struct btrfs_space_info *data_sinfo;
3336
3337 /* make sure bytes are sectorsize aligned */
3338 bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
3339
3340 data_sinfo = BTRFS_I(inode)->space_info;
3341 spin_lock(&data_sinfo->lock);
3342 data_sinfo->bytes_may_use -= bytes;
3343 BTRFS_I(inode)->reserved_bytes -= bytes;
3344 spin_unlock(&data_sinfo->lock);
3345 }
3346
3347 /* called when we are adding a delalloc extent to the inode's io_tree */
3348 void btrfs_delalloc_reserve_space(struct btrfs_root *root, struct inode *inode,
3349 u64 bytes)
3350 {
3351 struct btrfs_space_info *data_sinfo;
3352
3353 /* get the space info for where this inode will be storing its data */
3354 data_sinfo = BTRFS_I(inode)->space_info;
3355
3356 /* make sure we have enough space to handle the data first */
3357 spin_lock(&data_sinfo->lock);
3358 data_sinfo->bytes_delalloc += bytes;
3359
3360 /*
3361 * we are adding a delalloc extent without calling
3362 * btrfs_check_data_free_space first. This happens on a weird
3363 * writepage condition, but shouldn't hurt our accounting
3364 */
3365 if (unlikely(bytes > BTRFS_I(inode)->reserved_bytes)) {
3366 data_sinfo->bytes_may_use -= BTRFS_I(inode)->reserved_bytes;
3367 BTRFS_I(inode)->reserved_bytes = 0;
3368 } else {
3369 data_sinfo->bytes_may_use -= bytes;
3370 BTRFS_I(inode)->reserved_bytes -= bytes;
3371 }
3372
3373 spin_unlock(&data_sinfo->lock);
3374 }
3375
3376 /* called when we are clearing an delalloc extent from the inode's io_tree */
3377 void btrfs_delalloc_free_space(struct btrfs_root *root, struct inode *inode,
3378 u64 bytes)
3379 {
3380 struct btrfs_space_info *info;
3381
3382 info = BTRFS_I(inode)->space_info;
3383
3384 spin_lock(&info->lock);
3385 info->bytes_delalloc -= bytes;
3386 spin_unlock(&info->lock);
3387 }
3388
3389 static void force_metadata_allocation(struct btrfs_fs_info *info)
3390 {
3391 struct list_head *head = &info->space_info;
3392 struct btrfs_space_info *found;
3393
3394 rcu_read_lock();
3395 list_for_each_entry_rcu(found, head, list) {
3396 if (found->flags & BTRFS_BLOCK_GROUP_METADATA)
3397 found->force_alloc = 1;
3398 }
3399 rcu_read_unlock();
3400 }
3401
3402 static int do_chunk_alloc(struct btrfs_trans_handle *trans,
3403 struct btrfs_root *extent_root, u64 alloc_bytes,
3404 u64 flags, int force)
3405 {
3406 struct btrfs_space_info *space_info;
3407 struct btrfs_fs_info *fs_info = extent_root->fs_info;
3408 u64 thresh;
3409 int ret = 0;
3410
3411 mutex_lock(&fs_info->chunk_mutex);
3412
3413 flags = btrfs_reduce_alloc_profile(extent_root, flags);
3414
3415 space_info = __find_space_info(extent_root->fs_info, flags);
3416 if (!space_info) {
3417 ret = update_space_info(extent_root->fs_info, flags,
3418 0, 0, &space_info);
3419 BUG_ON(ret);
3420 }
3421 BUG_ON(!space_info);
3422
3423 spin_lock(&space_info->lock);
3424 if (space_info->force_alloc)
3425 force = 1;
3426 if (space_info->full) {
3427 spin_unlock(&space_info->lock);
3428 goto out;
3429 }
3430
3431 thresh = space_info->total_bytes - space_info->bytes_readonly;
3432 thresh = div_factor(thresh, 8);
3433 if (!force &&
3434 (space_info->bytes_used + space_info->bytes_pinned +
3435 space_info->bytes_reserved + alloc_bytes) < thresh) {
3436 spin_unlock(&space_info->lock);
3437 goto out;
3438 }
3439 spin_unlock(&space_info->lock);
3440
3441 /*
3442 * if we're doing a data chunk, go ahead and make sure that
3443 * we keep a reasonable number of metadata chunks allocated in the
3444 * FS as well.
3445 */
3446 if (flags & BTRFS_BLOCK_GROUP_DATA && fs_info->metadata_ratio) {
3447 fs_info->data_chunk_allocations++;
3448 if (!(fs_info->data_chunk_allocations %
3449 fs_info->metadata_ratio))
3450 force_metadata_allocation(fs_info);
3451 }
3452
3453 ret = btrfs_alloc_chunk(trans, extent_root, flags);
3454 spin_lock(&space_info->lock);
3455 if (ret)
3456 space_info->full = 1;
3457 space_info->force_alloc = 0;
3458 spin_unlock(&space_info->lock);
3459 out:
3460 mutex_unlock(&extent_root->fs_info->chunk_mutex);
3461 return ret;
3462 }
3463
3464 static int update_block_group(struct btrfs_trans_handle *trans,
3465 struct btrfs_root *root,
3466 u64 bytenr, u64 num_bytes, int alloc,
3467 int mark_free)
3468 {
3469 struct btrfs_block_group_cache *cache;
3470 struct btrfs_fs_info *info = root->fs_info;
3471 u64 total = num_bytes;
3472 u64 old_val;
3473 u64 byte_in_group;
3474
3475 /* block accounting for super block */
3476 spin_lock(&info->delalloc_lock);
3477 old_val = btrfs_super_bytes_used(&info->super_copy);
3478 if (alloc)
3479 old_val += num_bytes;
3480 else
3481 old_val -= num_bytes;
3482 btrfs_set_super_bytes_used(&info->super_copy, old_val);
3483 spin_unlock(&info->delalloc_lock);
3484
3485 while (total) {
3486 cache = btrfs_lookup_block_group(info, bytenr);
3487 if (!cache)
3488 return -1;
3489 byte_in_group = bytenr - cache->key.objectid;
3490 WARN_ON(byte_in_group > cache->key.offset);
3491
3492 spin_lock(&cache->space_info->lock);
3493 spin_lock(&cache->lock);
3494 cache->dirty = 1;
3495 old_val = btrfs_block_group_used(&cache->item);
3496 num_bytes = min(total, cache->key.offset - byte_in_group);
3497 if (alloc) {
3498 old_val += num_bytes;
3499 btrfs_set_block_group_used(&cache->item, old_val);
3500 cache->reserved -= num_bytes;
3501 cache->space_info->bytes_used += num_bytes;
3502 cache->space_info->bytes_reserved -= num_bytes;
3503 if (cache->ro)
3504 cache->space_info->bytes_readonly -= num_bytes;
3505 spin_unlock(&cache->lock);
3506 spin_unlock(&cache->space_info->lock);
3507 } else {
3508 old_val -= num_bytes;
3509 cache->space_info->bytes_used -= num_bytes;
3510 if (cache->ro)
3511 cache->space_info->bytes_readonly += num_bytes;
3512 btrfs_set_block_group_used(&cache->item, old_val);
3513 spin_unlock(&cache->lock);
3514 spin_unlock(&cache->space_info->lock);
3515 if (mark_free) {
3516 int ret;
3517
3518 ret = btrfs_discard_extent(root, bytenr,
3519 num_bytes);
3520 WARN_ON(ret);
3521
3522 ret = btrfs_add_free_space(cache, bytenr,
3523 num_bytes);
3524 WARN_ON(ret);
3525 }
3526 }
3527 btrfs_put_block_group(cache);
3528 total -= num_bytes;
3529 bytenr += num_bytes;
3530 }
3531 return 0;
3532 }
3533
3534 static u64 first_logical_byte(struct btrfs_root *root, u64 search_start)
3535 {
3536 struct btrfs_block_group_cache *cache;
3537 u64 bytenr;
3538
3539 cache = btrfs_lookup_first_block_group(root->fs_info, search_start);
3540 if (!cache)
3541 return 0;
3542
3543 bytenr = cache->key.objectid;
3544 btrfs_put_block_group(cache);
3545
3546 return bytenr;
3547 }
3548
3549 /*
3550 * this function must be called within transaction
3551 */
3552 int btrfs_pin_extent(struct btrfs_root *root,
3553 u64 bytenr, u64 num_bytes, int reserved)
3554 {
3555 struct btrfs_fs_info *fs_info = root->fs_info;
3556 struct btrfs_block_group_cache *cache;
3557
3558 cache = btrfs_lookup_block_group(fs_info, bytenr);
3559 BUG_ON(!cache);
3560
3561 spin_lock(&cache->space_info->lock);
3562 spin_lock(&cache->lock);
3563 cache->pinned += num_bytes;
3564 cache->space_info->bytes_pinned += num_bytes;
3565 if (reserved) {
3566 cache->reserved -= num_bytes;
3567 cache->space_info->bytes_reserved -= num_bytes;
3568 }
3569 spin_unlock(&cache->lock);
3570 spin_unlock(&cache->space_info->lock);
3571
3572 btrfs_put_block_group(cache);
3573
3574 set_extent_dirty(fs_info->pinned_extents,
3575 bytenr, bytenr + num_bytes - 1, GFP_NOFS);
3576 return 0;
3577 }
3578
3579 static int update_reserved_extents(struct btrfs_block_group_cache *cache,
3580 u64 num_bytes, int reserve)
3581 {
3582 spin_lock(&cache->space_info->lock);
3583 spin_lock(&cache->lock);
3584 if (reserve) {
3585 cache->reserved += num_bytes;
3586 cache->space_info->bytes_reserved += num_bytes;
3587 } else {
3588 cache->reserved -= num_bytes;
3589 cache->space_info->bytes_reserved -= num_bytes;
3590 }
3591 spin_unlock(&cache->lock);
3592 spin_unlock(&cache->space_info->lock);
3593 return 0;
3594 }
3595
3596 int btrfs_prepare_extent_commit(struct btrfs_trans_handle *trans,
3597 struct btrfs_root *root)
3598 {
3599 struct btrfs_fs_info *fs_info = root->fs_info;
3600 struct btrfs_caching_control *next;
3601 struct btrfs_caching_control *caching_ctl;
3602 struct btrfs_block_group_cache *cache;
3603
3604 down_write(&fs_info->extent_commit_sem);
3605
3606 list_for_each_entry_safe(caching_ctl, next,
3607 &fs_info->caching_block_groups, list) {
3608 cache = caching_ctl->block_group;
3609 if (block_group_cache_done(cache)) {
3610 cache->last_byte_to_unpin = (u64)-1;
3611 list_del_init(&caching_ctl->list);
3612 put_caching_control(caching_ctl);
3613 } else {
3614 cache->last_byte_to_unpin = caching_ctl->progress;
3615 }
3616 }
3617
3618 if (fs_info->pinned_extents == &fs_info->freed_extents[0])
3619 fs_info->pinned_extents = &fs_info->freed_extents[1];
3620 else
3621 fs_info->pinned_extents = &fs_info->freed_extents[0];
3622
3623 up_write(&fs_info->extent_commit_sem);
3624 return 0;
3625 }
3626
3627 static int unpin_extent_range(struct btrfs_root *root, u64 start, u64 end)
3628 {
3629 struct btrfs_fs_info *fs_info = root->fs_info;
3630 struct btrfs_block_group_cache *cache = NULL;
3631 u64 len;
3632
3633 while (start <= end) {
3634 if (!cache ||
3635 start >= cache->key.objectid + cache->key.offset) {
3636 if (cache)
3637 btrfs_put_block_group(cache);
3638 cache = btrfs_lookup_block_group(fs_info, start);
3639 BUG_ON(!cache);
3640 }
3641
3642 len = cache->key.objectid + cache->key.offset - start;
3643 len = min(len, end + 1 - start);
3644
3645 if (start < cache->last_byte_to_unpin) {
3646 len = min(len, cache->last_byte_to_unpin - start);
3647 btrfs_add_free_space(cache, start, len);
3648 }
3649
3650 spin_lock(&cache->space_info->lock);
3651 spin_lock(&cache->lock);
3652 cache->pinned -= len;
3653 cache->space_info->bytes_pinned -= len;
3654 spin_unlock(&cache->lock);
3655 spin_unlock(&cache->space_info->lock);
3656
3657 start += len;
3658 }
3659
3660 if (cache)
3661 btrfs_put_block_group(cache);
3662 return 0;
3663 }
3664
3665 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
3666 struct btrfs_root *root)
3667 {
3668 struct btrfs_fs_info *fs_info = root->fs_info;
3669 struct extent_io_tree *unpin;
3670 u64 start;
3671 u64 end;
3672 int ret;
3673
3674 if (fs_info->pinned_extents == &fs_info->freed_extents[0])
3675 unpin = &fs_info->freed_extents[1];
3676 else
3677 unpin = &fs_info->freed_extents[0];
3678
3679 while (1) {
3680 ret = find_first_extent_bit(unpin, 0, &start, &end,
3681 EXTENT_DIRTY);
3682 if (ret)
3683 break;
3684
3685 ret = btrfs_discard_extent(root, start, end + 1 - start);
3686
3687 clear_extent_dirty(unpin, start, end, GFP_NOFS);
3688 unpin_extent_range(root, start, end);
3689 cond_resched();
3690 }
3691
3692 return ret;
3693 }
3694
3695 static int pin_down_bytes(struct btrfs_trans_handle *trans,
3696 struct btrfs_root *root,
3697 struct btrfs_path *path,
3698 u64 bytenr, u64 num_bytes,
3699 int is_data, int reserved,
3700 struct extent_buffer **must_clean)
3701 {
3702 int err = 0;
3703 struct extent_buffer *buf;
3704
3705 if (is_data)
3706 goto pinit;
3707
3708 /*
3709 * discard is sloooow, and so triggering discards on
3710 * individual btree blocks isn't a good plan. Just
3711 * pin everything in discard mode.
3712 */
3713 if (btrfs_test_opt(root, DISCARD))
3714 goto pinit;
3715
3716 buf = btrfs_find_tree_block(root, bytenr, num_bytes);
3717 if (!buf)
3718 goto pinit;
3719
3720 /* we can reuse a block if it hasn't been written
3721 * and it is from this transaction. We can't
3722 * reuse anything from the tree log root because
3723 * it has tiny sub-transactions.
3724 */
3725 if (btrfs_buffer_uptodate(buf, 0) &&
3726 btrfs_try_tree_lock(buf)) {
3727 u64 header_owner = btrfs_header_owner(buf);
3728 u64 header_transid = btrfs_header_generation(buf);
3729 if (header_owner != BTRFS_TREE_LOG_OBJECTID &&
3730 header_transid == trans->transid &&
3731 !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
3732 *must_clean = buf;
3733 return 1;
3734 }
3735 btrfs_tree_unlock(buf);
3736 }
3737 free_extent_buffer(buf);
3738 pinit:
3739 if (path)
3740 btrfs_set_path_blocking(path);
3741 /* unlocks the pinned mutex */
3742 btrfs_pin_extent(root, bytenr, num_bytes, reserved);
3743
3744 BUG_ON(err < 0);
3745 return 0;
3746 }
3747
3748 static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
3749 struct btrfs_root *root,
3750 u64 bytenr, u64 num_bytes, u64 parent,
3751 u64 root_objectid, u64 owner_objectid,
3752 u64 owner_offset, int refs_to_drop,
3753 struct btrfs_delayed_extent_op *extent_op)
3754 {
3755 struct btrfs_key key;
3756 struct btrfs_path *path;
3757 struct btrfs_fs_info *info = root->fs_info;
3758 struct btrfs_root *extent_root = info->extent_root;
3759 struct extent_buffer *leaf;
3760 struct btrfs_extent_item *ei;
3761 struct btrfs_extent_inline_ref *iref;
3762 int ret;
3763 int is_data;
3764 int extent_slot = 0;
3765 int found_extent = 0;
3766 int num_to_del = 1;
3767 u32 item_size;
3768 u64 refs;
3769
3770 path = btrfs_alloc_path();
3771 if (!path)
3772 return -ENOMEM;
3773
3774 path->reada = 1;
3775 path->leave_spinning = 1;
3776
3777 is_data = owner_objectid >= BTRFS_FIRST_FREE_OBJECTID;
3778 BUG_ON(!is_data && refs_to_drop != 1);
3779
3780 ret = lookup_extent_backref(trans, extent_root, path, &iref,
3781 bytenr, num_bytes, parent,
3782 root_objectid, owner_objectid,
3783 owner_offset);
3784 if (ret == 0) {
3785 extent_slot = path->slots[0];
3786 while (extent_slot >= 0) {
3787 btrfs_item_key_to_cpu(path->nodes[0], &key,
3788 extent_slot);
3789 if (key.objectid != bytenr)
3790 break;
3791 if (key.type == BTRFS_EXTENT_ITEM_KEY &&
3792 key.offset == num_bytes) {
3793 found_extent = 1;
3794 break;
3795 }
3796 if (path->slots[0] - extent_slot > 5)
3797 break;
3798 extent_slot--;
3799 }
3800 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
3801 item_size = btrfs_item_size_nr(path->nodes[0], extent_slot);
3802 if (found_extent && item_size < sizeof(*ei))
3803 found_extent = 0;
3804 #endif
3805 if (!found_extent) {
3806 BUG_ON(iref);
3807 ret = remove_extent_backref(trans, extent_root, path,
3808 NULL, refs_to_drop,
3809 is_data);
3810 BUG_ON(ret);
3811 btrfs_release_path(extent_root, path);
3812 path->leave_spinning = 1;
3813
3814 key.objectid = bytenr;
3815 key.type = BTRFS_EXTENT_ITEM_KEY;
3816 key.offset = num_bytes;
3817
3818 ret = btrfs_search_slot(trans, extent_root,
3819 &key, path, -1, 1);
3820 if (ret) {
3821 printk(KERN_ERR "umm, got %d back from search"
3822 ", was looking for %llu\n", ret,
3823 (unsigned long long)bytenr);
3824 btrfs_print_leaf(extent_root, path->nodes[0]);
3825 }
3826 BUG_ON(ret);
3827 extent_slot = path->slots[0];
3828 }
3829 } else {
3830 btrfs_print_leaf(extent_root, path->nodes[0]);
3831 WARN_ON(1);
3832 printk(KERN_ERR "btrfs unable to find ref byte nr %llu "
3833 "parent %llu root %llu owner %llu offset %llu\n",
3834 (unsigned long long)bytenr,
3835 (unsigned long long)parent,
3836 (unsigned long long)root_objectid,
3837 (unsigned long long)owner_objectid,
3838 (unsigned long long)owner_offset);
3839 }
3840
3841 leaf = path->nodes[0];
3842 item_size = btrfs_item_size_nr(leaf, extent_slot);
3843 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
3844 if (item_size < sizeof(*ei)) {
3845 BUG_ON(found_extent || extent_slot != path->slots[0]);
3846 ret = convert_extent_item_v0(trans, extent_root, path,
3847 owner_objectid, 0);
3848 BUG_ON(ret < 0);
3849
3850 btrfs_release_path(extent_root, path);
3851 path->leave_spinning = 1;
3852
3853 key.objectid = bytenr;
3854 key.type = BTRFS_EXTENT_ITEM_KEY;
3855 key.offset = num_bytes;
3856
3857 ret = btrfs_search_slot(trans, extent_root, &key, path,
3858 -1, 1);
3859 if (ret) {
3860 printk(KERN_ERR "umm, got %d back from search"
3861 ", was looking for %llu\n", ret,
3862 (unsigned long long)bytenr);
3863 btrfs_print_leaf(extent_root, path->nodes[0]);
3864 }
3865 BUG_ON(ret);
3866 extent_slot = path->slots[0];
3867 leaf = path->nodes[0];
3868 item_size = btrfs_item_size_nr(leaf, extent_slot);
3869 }
3870 #endif
3871 BUG_ON(item_size < sizeof(*ei));
3872 ei = btrfs_item_ptr(leaf, extent_slot,
3873 struct btrfs_extent_item);
3874 if (owner_objectid < BTRFS_FIRST_FREE_OBJECTID) {
3875 struct btrfs_tree_block_info *bi;
3876 BUG_ON(item_size < sizeof(*ei) + sizeof(*bi));
3877 bi = (struct btrfs_tree_block_info *)(ei + 1);
3878 WARN_ON(owner_objectid != btrfs_tree_block_level(leaf, bi));
3879 }
3880
3881 refs = btrfs_extent_refs(leaf, ei);
3882 BUG_ON(refs < refs_to_drop);
3883 refs -= refs_to_drop;
3884
3885 if (refs > 0) {
3886 if (extent_op)
3887 __run_delayed_extent_op(extent_op, leaf, ei);
3888 /*
3889 * In the case of inline back ref, reference count will
3890 * be updated by remove_extent_backref
3891 */
3892 if (iref) {
3893 BUG_ON(!found_extent);
3894 } else {
3895 btrfs_set_extent_refs(leaf, ei, refs);
3896 btrfs_mark_buffer_dirty(leaf);
3897 }
3898 if (found_extent) {
3899 ret = remove_extent_backref(trans, extent_root, path,
3900 iref, refs_to_drop,
3901 is_data);
3902 BUG_ON(ret);
3903 }
3904 } else {
3905 int mark_free = 0;
3906 struct extent_buffer *must_clean = NULL;
3907
3908 if (found_extent) {
3909 BUG_ON(is_data && refs_to_drop !=
3910 extent_data_ref_count(root, path, iref));
3911 if (iref) {
3912 BUG_ON(path->slots[0] != extent_slot);
3913 } else {
3914 BUG_ON(path->slots[0] != extent_slot + 1);
3915 path->slots[0] = extent_slot;
3916 num_to_del = 2;
3917 }
3918 }
3919
3920 ret = pin_down_bytes(trans, root, path, bytenr,
3921 num_bytes, is_data, 0, &must_clean);
3922 if (ret > 0)
3923 mark_free = 1;
3924 BUG_ON(ret < 0);
3925 /*
3926 * it is going to be very rare for someone to be waiting
3927 * on the block we're freeing. del_items might need to
3928 * schedule, so rather than get fancy, just force it
3929 * to blocking here
3930 */
3931 if (must_clean)
3932 btrfs_set_lock_blocking(must_clean);
3933
3934 ret = btrfs_del_items(trans, extent_root, path, path->slots[0],
3935 num_to_del);
3936 BUG_ON(ret);
3937 btrfs_release_path(extent_root, path);
3938
3939 if (must_clean) {
3940 clean_tree_block(NULL, root, must_clean);
3941 btrfs_tree_unlock(must_clean);
3942 free_extent_buffer(must_clean);
3943 }
3944
3945 if (is_data) {
3946 ret = btrfs_del_csums(trans, root, bytenr, num_bytes);
3947 BUG_ON(ret);
3948 } else {
3949 invalidate_mapping_pages(info->btree_inode->i_mapping,
3950 bytenr >> PAGE_CACHE_SHIFT,
3951 (bytenr + num_bytes - 1) >> PAGE_CACHE_SHIFT);
3952 }
3953
3954 ret = update_block_group(trans, root, bytenr, num_bytes, 0,
3955 mark_free);
3956 BUG_ON(ret);
3957 }
3958 btrfs_free_path(path);
3959 return ret;
3960 }
3961
3962 /*
3963 * when we free an extent, it is possible (and likely) that we free the last
3964 * delayed ref for that extent as well. This searches the delayed ref tree for
3965 * a given extent, and if there are no other delayed refs to be processed, it
3966 * removes it from the tree.
3967 */
3968 static noinline int check_ref_cleanup(struct btrfs_trans_handle *trans,
3969 struct btrfs_root *root, u64 bytenr)
3970 {
3971 struct btrfs_delayed_ref_head *head;
3972 struct btrfs_delayed_ref_root *delayed_refs;
3973 struct btrfs_delayed_ref_node *ref;
3974 struct rb_node *node;
3975 int ret;
3976
3977 delayed_refs = &trans->transaction->delayed_refs;
3978 spin_lock(&delayed_refs->lock);
3979 head = btrfs_find_delayed_ref_head(trans, bytenr);
3980 if (!head)
3981 goto out;
3982
3983 node = rb_prev(&head->node.rb_node);
3984 if (!node)
3985 goto out;
3986
3987 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
3988
3989 /* there are still entries for this ref, we can't drop it */
3990 if (ref->bytenr == bytenr)
3991 goto out;
3992
3993 if (head->extent_op) {
3994 if (!head->must_insert_reserved)
3995 goto out;
3996 kfree(head->extent_op);
3997 head->extent_op = NULL;
3998 }
3999
4000 /*
4001 * waiting for the lock here would deadlock. If someone else has it
4002 * locked they are already in the process of dropping it anyway
4003 */
4004 if (!mutex_trylock(&head->mutex))
4005 goto out;
4006
4007 /*
4008 * at this point we have a head with no other entries. Go
4009 * ahead and process it.
4010 */
4011 head->node.in_tree = 0;
4012 rb_erase(&head->node.rb_node, &delayed_refs->root);
4013
4014 delayed_refs->num_entries--;
4015
4016 /*
4017 * we don't take a ref on the node because we're removing it from the
4018 * tree, so we just steal the ref the tree was holding.
4019 */
4020 delayed_refs->num_heads--;
4021 if (list_empty(&head->cluster))
4022 delayed_refs->num_heads_ready--;
4023
4024 list_del_init(&head->cluster);
4025 spin_unlock(&delayed_refs->lock);
4026
4027 ret = run_one_delayed_ref(trans, root->fs_info->tree_root,
4028 &head->node, head->extent_op,
4029 head->must_insert_reserved);
4030 BUG_ON(ret);
4031 btrfs_put_delayed_ref(&head->node);
4032 return 0;
4033 out:
4034 spin_unlock(&delayed_refs->lock);
4035 return 0;
4036 }
4037
4038 int btrfs_free_extent(struct btrfs_trans_handle *trans,
4039 struct btrfs_root *root,
4040 u64 bytenr, u64 num_bytes, u64 parent,
4041 u64 root_objectid, u64 owner, u64 offset)
4042 {
4043 int ret;
4044
4045 /*
4046 * tree log blocks never actually go into the extent allocation
4047 * tree, just update pinning info and exit early.
4048 */
4049 if (root_objectid == BTRFS_TREE_LOG_OBJECTID) {
4050 WARN_ON(owner >= BTRFS_FIRST_FREE_OBJECTID);
4051 /* unlocks the pinned mutex */
4052 btrfs_pin_extent(root, bytenr, num_bytes, 1);
4053 ret = 0;
4054 } else if (owner < BTRFS_FIRST_FREE_OBJECTID) {
4055 ret = btrfs_add_delayed_tree_ref(trans, bytenr, num_bytes,
4056 parent, root_objectid, (int)owner,
4057 BTRFS_DROP_DELAYED_REF, NULL);
4058 BUG_ON(ret);
4059 ret = check_ref_cleanup(trans, root, bytenr);
4060 BUG_ON(ret);
4061 } else {
4062 ret = btrfs_add_delayed_data_ref(trans, bytenr, num_bytes,
4063 parent, root_objectid, owner,
4064 offset, BTRFS_DROP_DELAYED_REF, NULL);
4065 BUG_ON(ret);
4066 }
4067 return ret;
4068 }
4069
4070 int btrfs_free_tree_block(struct btrfs_trans_handle *trans,
4071 struct btrfs_root *root,
4072 u64 bytenr, u32 blocksize,
4073 u64 parent, u64 root_objectid, int level)
4074 {
4075 u64 used;
4076 spin_lock(&root->node_lock);
4077 used = btrfs_root_used(&root->root_item) - blocksize;
4078 btrfs_set_root_used(&root->root_item, used);
4079 spin_unlock(&root->node_lock);
4080
4081 return btrfs_free_extent(trans, root, bytenr, blocksize,
4082 parent, root_objectid, level, 0);
4083 }
4084
4085 static u64 stripe_align(struct btrfs_root *root, u64 val)
4086 {
4087 u64 mask = ((u64)root->stripesize - 1);
4088 u64 ret = (val + mask) & ~mask;
4089 return ret;
4090 }
4091
4092 /*
4093 * when we wait for progress in the block group caching, its because
4094 * our allocation attempt failed at least once. So, we must sleep
4095 * and let some progress happen before we try again.
4096 *
4097 * This function will sleep at least once waiting for new free space to
4098 * show up, and then it will check the block group free space numbers
4099 * for our min num_bytes. Another option is to have it go ahead
4100 * and look in the rbtree for a free extent of a given size, but this
4101 * is a good start.
4102 */
4103 static noinline int
4104 wait_block_group_cache_progress(struct btrfs_block_group_cache *cache,
4105 u64 num_bytes)
4106 {
4107 struct btrfs_caching_control *caching_ctl;
4108 DEFINE_WAIT(wait);
4109
4110 caching_ctl = get_caching_control(cache);
4111 if (!caching_ctl)
4112 return 0;
4113
4114 wait_event(caching_ctl->wait, block_group_cache_done(cache) ||
4115 (cache->free_space >= num_bytes));
4116
4117 put_caching_control(caching_ctl);
4118 return 0;
4119 }
4120
4121 static noinline int
4122 wait_block_group_cache_done(struct btrfs_block_group_cache *cache)
4123 {
4124 struct btrfs_caching_control *caching_ctl;
4125 DEFINE_WAIT(wait);
4126
4127 caching_ctl = get_caching_control(cache);
4128 if (!caching_ctl)
4129 return 0;
4130
4131 wait_event(caching_ctl->wait, block_group_cache_done(cache));
4132
4133 put_caching_control(caching_ctl);
4134 return 0;
4135 }
4136
4137 enum btrfs_loop_type {
4138 LOOP_FIND_IDEAL = 0,
4139 LOOP_CACHING_NOWAIT = 1,
4140 LOOP_CACHING_WAIT = 2,
4141 LOOP_ALLOC_CHUNK = 3,
4142 LOOP_NO_EMPTY_SIZE = 4,
4143 };
4144
4145 /*
4146 * walks the btree of allocated extents and find a hole of a given size.
4147 * The key ins is changed to record the hole:
4148 * ins->objectid == block start
4149 * ins->flags = BTRFS_EXTENT_ITEM_KEY
4150 * ins->offset == number of blocks
4151 * Any available blocks before search_start are skipped.
4152 */
4153 static noinline int find_free_extent(struct btrfs_trans_handle *trans,
4154 struct btrfs_root *orig_root,
4155 u64 num_bytes, u64 empty_size,
4156 u64 search_start, u64 search_end,
4157 u64 hint_byte, struct btrfs_key *ins,
4158 u64 exclude_start, u64 exclude_nr,
4159 int data)
4160 {
4161 int ret = 0;
4162 struct btrfs_root *root = orig_root->fs_info->extent_root;
4163 struct btrfs_free_cluster *last_ptr = NULL;
4164 struct btrfs_block_group_cache *block_group = NULL;
4165 int empty_cluster = 2 * 1024 * 1024;
4166 int allowed_chunk_alloc = 0;
4167 int done_chunk_alloc = 0;
4168 struct btrfs_space_info *space_info;
4169 int last_ptr_loop = 0;
4170 int loop = 0;
4171 bool found_uncached_bg = false;
4172 bool failed_cluster_refill = false;
4173 bool failed_alloc = false;
4174 u64 ideal_cache_percent = 0;
4175 u64 ideal_cache_offset = 0;
4176
4177 WARN_ON(num_bytes < root->sectorsize);
4178 btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
4179 ins->objectid = 0;
4180 ins->offset = 0;
4181
4182 space_info = __find_space_info(root->fs_info, data);
4183 if (!space_info) {
4184 printk(KERN_ERR "No space info for %d\n", data);
4185 return -ENOSPC;
4186 }
4187
4188 if (orig_root->ref_cows || empty_size)
4189 allowed_chunk_alloc = 1;
4190
4191 if (data & BTRFS_BLOCK_GROUP_METADATA) {
4192 last_ptr = &root->fs_info->meta_alloc_cluster;
4193 if (!btrfs_test_opt(root, SSD))
4194 empty_cluster = 64 * 1024;
4195 }
4196
4197 if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD)) {
4198 last_ptr = &root->fs_info->data_alloc_cluster;
4199 }
4200
4201 if (last_ptr) {
4202 spin_lock(&last_ptr->lock);
4203 if (last_ptr->block_group)
4204 hint_byte = last_ptr->window_start;
4205 spin_unlock(&last_ptr->lock);
4206 }
4207
4208 search_start = max(search_start, first_logical_byte(root, 0));
4209 search_start = max(search_start, hint_byte);
4210
4211 if (!last_ptr)
4212 empty_cluster = 0;
4213
4214 if (search_start == hint_byte) {
4215 ideal_cache:
4216 block_group = btrfs_lookup_block_group(root->fs_info,
4217 search_start);
4218 /*
4219 * we don't want to use the block group if it doesn't match our
4220 * allocation bits, or if its not cached.
4221 *
4222 * However if we are re-searching with an ideal block group
4223 * picked out then we don't care that the block group is cached.
4224 */
4225 if (block_group && block_group_bits(block_group, data) &&
4226 (block_group->cached != BTRFS_CACHE_NO ||
4227 search_start == ideal_cache_offset)) {
4228 down_read(&space_info->groups_sem);
4229 if (list_empty(&block_group->list) ||
4230 block_group->ro) {
4231 /*
4232 * someone is removing this block group,
4233 * we can't jump into the have_block_group
4234 * target because our list pointers are not
4235 * valid
4236 */
4237 btrfs_put_block_group(block_group);
4238 up_read(&space_info->groups_sem);
4239 } else {
4240 goto have_block_group;
4241 }
4242 } else if (block_group) {
4243 btrfs_put_block_group(block_group);
4244 }
4245 }
4246 search:
4247 down_read(&space_info->groups_sem);
4248 list_for_each_entry(block_group, &space_info->block_groups, list) {
4249 u64 offset;
4250 int cached;
4251
4252 btrfs_get_block_group(block_group);
4253 search_start = block_group->key.objectid;
4254
4255 have_block_group:
4256 if (unlikely(block_group->cached == BTRFS_CACHE_NO)) {
4257 u64 free_percent;
4258
4259 free_percent = btrfs_block_group_used(&block_group->item);
4260 free_percent *= 100;
4261 free_percent = div64_u64(free_percent,
4262 block_group->key.offset);
4263 free_percent = 100 - free_percent;
4264 if (free_percent > ideal_cache_percent &&
4265 likely(!block_group->ro)) {
4266 ideal_cache_offset = block_group->key.objectid;
4267 ideal_cache_percent = free_percent;
4268 }
4269
4270 /*
4271 * We only want to start kthread caching if we are at
4272 * the point where we will wait for caching to make
4273 * progress, or if our ideal search is over and we've
4274 * found somebody to start caching.
4275 */
4276 if (loop > LOOP_CACHING_NOWAIT ||
4277 (loop > LOOP_FIND_IDEAL &&
4278 atomic_read(&space_info->caching_threads) < 2)) {
4279 ret = cache_block_group(block_group);
4280 BUG_ON(ret);
4281 }
4282 found_uncached_bg = true;
4283
4284 /*
4285 * If loop is set for cached only, try the next block
4286 * group.
4287 */
4288 if (loop == LOOP_FIND_IDEAL)
4289 goto loop;
4290 }
4291
4292 cached = block_group_cache_done(block_group);
4293 if (unlikely(!cached))
4294 found_uncached_bg = true;
4295
4296 if (unlikely(block_group->ro))
4297 goto loop;
4298
4299 /*
4300 * Ok we want to try and use the cluster allocator, so lets look
4301 * there, unless we are on LOOP_NO_EMPTY_SIZE, since we will
4302 * have tried the cluster allocator plenty of times at this
4303 * point and not have found anything, so we are likely way too
4304 * fragmented for the clustering stuff to find anything, so lets
4305 * just skip it and let the allocator find whatever block it can
4306 * find
4307 */
4308 if (last_ptr && loop < LOOP_NO_EMPTY_SIZE) {
4309 /*
4310 * the refill lock keeps out other
4311 * people trying to start a new cluster
4312 */
4313 spin_lock(&last_ptr->refill_lock);
4314 if (last_ptr->block_group &&
4315 (last_ptr->block_group->ro ||
4316 !block_group_bits(last_ptr->block_group, data))) {
4317 offset = 0;
4318 goto refill_cluster;
4319 }
4320
4321 offset = btrfs_alloc_from_cluster(block_group, last_ptr,
4322 num_bytes, search_start);
4323 if (offset) {
4324 /* we have a block, we're done */
4325 spin_unlock(&last_ptr->refill_lock);
4326 goto checks;
4327 }
4328
4329 spin_lock(&last_ptr->lock);
4330 /*
4331 * whoops, this cluster doesn't actually point to
4332 * this block group. Get a ref on the block
4333 * group is does point to and try again
4334 */
4335 if (!last_ptr_loop && last_ptr->block_group &&
4336 last_ptr->block_group != block_group) {
4337
4338 btrfs_put_block_group(block_group);
4339 block_group = last_ptr->block_group;
4340 btrfs_get_block_group(block_group);
4341 spin_unlock(&last_ptr->lock);
4342 spin_unlock(&last_ptr->refill_lock);
4343
4344 last_ptr_loop = 1;
4345 search_start = block_group->key.objectid;
4346 /*
4347 * we know this block group is properly
4348 * in the list because
4349 * btrfs_remove_block_group, drops the
4350 * cluster before it removes the block
4351 * group from the list
4352 */
4353 goto have_block_group;
4354 }
4355 spin_unlock(&last_ptr->lock);
4356 refill_cluster:
4357 /*
4358 * this cluster didn't work out, free it and
4359 * start over
4360 */
4361 btrfs_return_cluster_to_free_space(NULL, last_ptr);
4362
4363 last_ptr_loop = 0;
4364
4365 /* allocate a cluster in this block group */
4366 ret = btrfs_find_space_cluster(trans, root,
4367 block_group, last_ptr,
4368 offset, num_bytes,
4369 empty_cluster + empty_size);
4370 if (ret == 0) {
4371 /*
4372 * now pull our allocation out of this
4373 * cluster
4374 */
4375 offset = btrfs_alloc_from_cluster(block_group,
4376 last_ptr, num_bytes,
4377 search_start);
4378 if (offset) {
4379 /* we found one, proceed */
4380 spin_unlock(&last_ptr->refill_lock);
4381 goto checks;
4382 }
4383 } else if (!cached && loop > LOOP_CACHING_NOWAIT
4384 && !failed_cluster_refill) {
4385 spin_unlock(&last_ptr->refill_lock);
4386
4387 failed_cluster_refill = true;
4388 wait_block_group_cache_progress(block_group,
4389 num_bytes + empty_cluster + empty_size);
4390 goto have_block_group;
4391 }
4392
4393 /*
4394 * at this point we either didn't find a cluster
4395 * or we weren't able to allocate a block from our
4396 * cluster. Free the cluster we've been trying
4397 * to use, and go to the next block group
4398 */
4399 btrfs_return_cluster_to_free_space(NULL, last_ptr);
4400 spin_unlock(&last_ptr->refill_lock);
4401 goto loop;
4402 }
4403
4404 offset = btrfs_find_space_for_alloc(block_group, search_start,
4405 num_bytes, empty_size);
4406 /*
4407 * If we didn't find a chunk, and we haven't failed on this
4408 * block group before, and this block group is in the middle of
4409 * caching and we are ok with waiting, then go ahead and wait
4410 * for progress to be made, and set failed_alloc to true.
4411 *
4412 * If failed_alloc is true then we've already waited on this
4413 * block group once and should move on to the next block group.
4414 */
4415 if (!offset && !failed_alloc && !cached &&
4416 loop > LOOP_CACHING_NOWAIT) {
4417 wait_block_group_cache_progress(block_group,
4418 num_bytes + empty_size);
4419 failed_alloc = true;
4420 goto have_block_group;
4421 } else if (!offset) {
4422 goto loop;
4423 }
4424 checks:
4425 search_start = stripe_align(root, offset);
4426 /* move on to the next group */
4427 if (search_start + num_bytes >= search_end) {
4428 btrfs_add_free_space(block_group, offset, num_bytes);
4429 goto loop;
4430 }
4431
4432 /* move on to the next group */
4433 if (search_start + num_bytes >
4434 block_group->key.objectid + block_group->key.offset) {
4435 btrfs_add_free_space(block_group, offset, num_bytes);
4436 goto loop;
4437 }
4438
4439 if (exclude_nr > 0 &&
4440 (search_start + num_bytes > exclude_start &&
4441 search_start < exclude_start + exclude_nr)) {
4442 search_start = exclude_start + exclude_nr;
4443
4444 btrfs_add_free_space(block_group, offset, num_bytes);
4445 /*
4446 * if search_start is still in this block group
4447 * then we just re-search this block group
4448 */
4449 if (search_start >= block_group->key.objectid &&
4450 search_start < (block_group->key.objectid +
4451 block_group->key.offset))
4452 goto have_block_group;
4453 goto loop;
4454 }
4455
4456 ins->objectid = search_start;
4457 ins->offset = num_bytes;
4458
4459 if (offset < search_start)
4460 btrfs_add_free_space(block_group, offset,
4461 search_start - offset);
4462 BUG_ON(offset > search_start);
4463
4464 update_reserved_extents(block_group, num_bytes, 1);
4465
4466 /* we are all good, lets return */
4467 break;
4468 loop:
4469 failed_cluster_refill = false;
4470 failed_alloc = false;
4471 btrfs_put_block_group(block_group);
4472 }
4473 up_read(&space_info->groups_sem);
4474
4475 /* LOOP_FIND_IDEAL, only search caching/cached bg's, and don't wait for
4476 * for them to make caching progress. Also
4477 * determine the best possible bg to cache
4478 * LOOP_CACHING_NOWAIT, search partially cached block groups, kicking
4479 * caching kthreads as we move along
4480 * LOOP_CACHING_WAIT, search everything, and wait if our bg is caching
4481 * LOOP_ALLOC_CHUNK, force a chunk allocation and try again
4482 * LOOP_NO_EMPTY_SIZE, set empty_size and empty_cluster to 0 and try
4483 * again
4484 */
4485 if (!ins->objectid && loop < LOOP_NO_EMPTY_SIZE &&
4486 (found_uncached_bg || empty_size || empty_cluster ||
4487 allowed_chunk_alloc)) {
4488 if (loop == LOOP_FIND_IDEAL && found_uncached_bg) {
4489 found_uncached_bg = false;
4490 loop++;
4491 if (!ideal_cache_percent &&
4492 atomic_read(&space_info->caching_threads))
4493 goto search;
4494
4495 /*
4496 * 1 of the following 2 things have happened so far
4497 *
4498 * 1) We found an ideal block group for caching that
4499 * is mostly full and will cache quickly, so we might
4500 * as well wait for it.
4501 *
4502 * 2) We searched for cached only and we didn't find
4503 * anything, and we didn't start any caching kthreads
4504 * either, so chances are we will loop through and
4505 * start a couple caching kthreads, and then come back
4506 * around and just wait for them. This will be slower
4507 * because we will have 2 caching kthreads reading at
4508 * the same time when we could have just started one
4509 * and waited for it to get far enough to give us an
4510 * allocation, so go ahead and go to the wait caching
4511 * loop.
4512 */
4513 loop = LOOP_CACHING_WAIT;
4514 search_start = ideal_cache_offset;
4515 ideal_cache_percent = 0;
4516 goto ideal_cache;
4517 } else if (loop == LOOP_FIND_IDEAL) {
4518 /*
4519 * Didn't find a uncached bg, wait on anything we find
4520 * next.
4521 */
4522 loop = LOOP_CACHING_WAIT;
4523 goto search;
4524 }
4525
4526 if (loop < LOOP_CACHING_WAIT) {
4527 loop++;
4528 goto search;
4529 }
4530
4531 if (loop == LOOP_ALLOC_CHUNK) {
4532 empty_size = 0;
4533 empty_cluster = 0;
4534 }
4535
4536 if (allowed_chunk_alloc) {
4537 ret = do_chunk_alloc(trans, root, num_bytes +
4538 2 * 1024 * 1024, data, 1);
4539 allowed_chunk_alloc = 0;
4540 done_chunk_alloc = 1;
4541 } else if (!done_chunk_alloc) {
4542 space_info->force_alloc = 1;
4543 }
4544
4545 if (loop < LOOP_NO_EMPTY_SIZE) {
4546 loop++;
4547 goto search;
4548 }
4549 ret = -ENOSPC;
4550 } else if (!ins->objectid) {
4551 ret = -ENOSPC;
4552 }
4553
4554 /* we found what we needed */
4555 if (ins->objectid) {
4556 if (!(data & BTRFS_BLOCK_GROUP_DATA))
4557 trans->block_group = block_group->key.objectid;
4558
4559 btrfs_put_block_group(block_group);
4560 ret = 0;
4561 }
4562
4563 return ret;
4564 }
4565
4566 static void dump_space_info(struct btrfs_space_info *info, u64 bytes,
4567 int dump_block_groups)
4568 {
4569 struct btrfs_block_group_cache *cache;
4570
4571 spin_lock(&info->lock);
4572 printk(KERN_INFO "space_info has %llu free, is %sfull\n",
4573 (unsigned long long)(info->total_bytes - info->bytes_used -
4574 info->bytes_pinned - info->bytes_reserved -
4575 info->bytes_super),
4576 (info->full) ? "" : "not ");
4577 printk(KERN_INFO "space_info total=%llu, pinned=%llu, delalloc=%llu,"
4578 " may_use=%llu, used=%llu, root=%llu, super=%llu, reserved=%llu"
4579 "\n",
4580 (unsigned long long)info->total_bytes,
4581 (unsigned long long)info->bytes_pinned,
4582 (unsigned long long)info->bytes_delalloc,
4583 (unsigned long long)info->bytes_may_use,
4584 (unsigned long long)info->bytes_used,
4585 (unsigned long long)info->bytes_root,
4586 (unsigned long long)info->bytes_super,
4587 (unsigned long long)info->bytes_reserved);
4588 spin_unlock(&info->lock);
4589
4590 if (!dump_block_groups)
4591 return;
4592
4593 down_read(&info->groups_sem);
4594 list_for_each_entry(cache, &info->block_groups, list) {
4595 spin_lock(&cache->lock);
4596 printk(KERN_INFO "block group %llu has %llu bytes, %llu used "
4597 "%llu pinned %llu reserved\n",
4598 (unsigned long long)cache->key.objectid,
4599 (unsigned long long)cache->key.offset,
4600 (unsigned long long)btrfs_block_group_used(&cache->item),
4601 (unsigned long long)cache->pinned,
4602 (unsigned long long)cache->reserved);
4603 btrfs_dump_free_space(cache, bytes);
4604 spin_unlock(&cache->lock);
4605 }
4606 up_read(&info->groups_sem);
4607 }
4608
4609 int btrfs_reserve_extent(struct btrfs_trans_handle *trans,
4610 struct btrfs_root *root,
4611 u64 num_bytes, u64 min_alloc_size,
4612 u64 empty_size, u64 hint_byte,
4613 u64 search_end, struct btrfs_key *ins,
4614 u64 data)
4615 {
4616 int ret;
4617 u64 search_start = 0;
4618
4619 data = btrfs_get_alloc_profile(root, data);
4620 again:
4621 /*
4622 * the only place that sets empty_size is btrfs_realloc_node, which
4623 * is not called recursively on allocations
4624 */
4625 if (empty_size || root->ref_cows)
4626 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
4627 num_bytes + 2 * 1024 * 1024, data, 0);
4628
4629 WARN_ON(num_bytes < root->sectorsize);
4630 ret = find_free_extent(trans, root, num_bytes, empty_size,
4631 search_start, search_end, hint_byte, ins,
4632 trans->alloc_exclude_start,
4633 trans->alloc_exclude_nr, data);
4634
4635 if (ret == -ENOSPC && num_bytes > min_alloc_size) {
4636 num_bytes = num_bytes >> 1;
4637 num_bytes = num_bytes & ~(root->sectorsize - 1);
4638 num_bytes = max(num_bytes, min_alloc_size);
4639 do_chunk_alloc(trans, root->fs_info->extent_root,
4640 num_bytes, data, 1);
4641 goto again;
4642 }
4643 if (ret == -ENOSPC) {
4644 struct btrfs_space_info *sinfo;
4645
4646 sinfo = __find_space_info(root->fs_info, data);
4647 printk(KERN_ERR "btrfs allocation failed flags %llu, "
4648 "wanted %llu\n", (unsigned long long)data,
4649 (unsigned long long)num_bytes);
4650 dump_space_info(sinfo, num_bytes, 1);
4651 }
4652
4653 return ret;
4654 }
4655
4656 int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len)
4657 {
4658 struct btrfs_block_group_cache *cache;
4659 int ret = 0;
4660
4661 cache = btrfs_lookup_block_group(root->fs_info, start);
4662 if (!cache) {
4663 printk(KERN_ERR "Unable to find block group for %llu\n",
4664 (unsigned long long)start);
4665 return -ENOSPC;
4666 }
4667
4668 ret = btrfs_discard_extent(root, start, len);
4669
4670 btrfs_add_free_space(cache, start, len);
4671 update_reserved_extents(cache, len, 0);
4672 btrfs_put_block_group(cache);
4673
4674 return ret;
4675 }
4676
4677 static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
4678 struct btrfs_root *root,
4679 u64 parent, u64 root_objectid,
4680 u64 flags, u64 owner, u64 offset,
4681 struct btrfs_key *ins, int ref_mod)
4682 {
4683 int ret;
4684 struct btrfs_fs_info *fs_info = root->fs_info;
4685 struct btrfs_extent_item *extent_item;
4686 struct btrfs_extent_inline_ref *iref;
4687 struct btrfs_path *path;
4688 struct extent_buffer *leaf;
4689 int type;
4690 u32 size;
4691
4692 if (parent > 0)
4693 type = BTRFS_SHARED_DATA_REF_KEY;
4694 else
4695 type = BTRFS_EXTENT_DATA_REF_KEY;
4696
4697 size = sizeof(*extent_item) + btrfs_extent_inline_ref_size(type);
4698
4699 path = btrfs_alloc_path();
4700 BUG_ON(!path);
4701
4702 path->leave_spinning = 1;
4703 ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
4704 ins, size);
4705 BUG_ON(ret);
4706
4707 leaf = path->nodes[0];
4708 extent_item = btrfs_item_ptr(leaf, path->slots[0],
4709 struct btrfs_extent_item);
4710 btrfs_set_extent_refs(leaf, extent_item, ref_mod);
4711 btrfs_set_extent_generation(leaf, extent_item, trans->transid);
4712 btrfs_set_extent_flags(leaf, extent_item,
4713 flags | BTRFS_EXTENT_FLAG_DATA);
4714
4715 iref = (struct btrfs_extent_inline_ref *)(extent_item + 1);
4716 btrfs_set_extent_inline_ref_type(leaf, iref, type);
4717 if (parent > 0) {
4718 struct btrfs_shared_data_ref *ref;
4719 ref = (struct btrfs_shared_data_ref *)(iref + 1);
4720 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
4721 btrfs_set_shared_data_ref_count(leaf, ref, ref_mod);
4722 } else {
4723 struct btrfs_extent_data_ref *ref;
4724 ref = (struct btrfs_extent_data_ref *)(&iref->offset);
4725 btrfs_set_extent_data_ref_root(leaf, ref, root_objectid);
4726 btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
4727 btrfs_set_extent_data_ref_offset(leaf, ref, offset);
4728 btrfs_set_extent_data_ref_count(leaf, ref, ref_mod);
4729 }
4730
4731 btrfs_mark_buffer_dirty(path->nodes[0]);
4732 btrfs_free_path(path);
4733
4734 ret = update_block_group(trans, root, ins->objectid, ins->offset,
4735 1, 0);
4736 if (ret) {
4737 printk(KERN_ERR "btrfs update block group failed for %llu "
4738 "%llu\n", (unsigned long long)ins->objectid,
4739 (unsigned long long)ins->offset);
4740 BUG();
4741 }
4742 return ret;
4743 }
4744
4745 static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
4746 struct btrfs_root *root,
4747 u64 parent, u64 root_objectid,
4748 u64 flags, struct btrfs_disk_key *key,
4749 int level, struct btrfs_key *ins)
4750 {
4751 int ret;
4752 struct btrfs_fs_info *fs_info = root->fs_info;
4753 struct btrfs_extent_item *extent_item;
4754 struct btrfs_tree_block_info *block_info;
4755 struct btrfs_extent_inline_ref *iref;
4756 struct btrfs_path *path;
4757 struct extent_buffer *leaf;
4758 u32 size = sizeof(*extent_item) + sizeof(*block_info) + sizeof(*iref);
4759
4760 path = btrfs_alloc_path();
4761 BUG_ON(!path);
4762
4763 path->leave_spinning = 1;
4764 ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
4765 ins, size);
4766 BUG_ON(ret);
4767
4768 leaf = path->nodes[0];
4769 extent_item = btrfs_item_ptr(leaf, path->slots[0],
4770 struct btrfs_extent_item);
4771 btrfs_set_extent_refs(leaf, extent_item, 1);
4772 btrfs_set_extent_generation(leaf, extent_item, trans->transid);
4773 btrfs_set_extent_flags(leaf, extent_item,
4774 flags | BTRFS_EXTENT_FLAG_TREE_BLOCK);
4775 block_info = (struct btrfs_tree_block_info *)(extent_item + 1);
4776
4777 btrfs_set_tree_block_key(leaf, block_info, key);
4778 btrfs_set_tree_block_level(leaf, block_info, level);
4779
4780 iref = (struct btrfs_extent_inline_ref *)(block_info + 1);
4781 if (parent > 0) {
4782 BUG_ON(!(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF));
4783 btrfs_set_extent_inline_ref_type(leaf, iref,
4784 BTRFS_SHARED_BLOCK_REF_KEY);
4785 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
4786 } else {
4787 btrfs_set_extent_inline_ref_type(leaf, iref,
4788 BTRFS_TREE_BLOCK_REF_KEY);
4789 btrfs_set_extent_inline_ref_offset(leaf, iref, root_objectid);
4790 }
4791
4792 btrfs_mark_buffer_dirty(leaf);
4793 btrfs_free_path(path);
4794
4795 ret = update_block_group(trans, root, ins->objectid, ins->offset,
4796 1, 0);
4797 if (ret) {
4798 printk(KERN_ERR "btrfs update block group failed for %llu "
4799 "%llu\n", (unsigned long long)ins->objectid,
4800 (unsigned long long)ins->offset);
4801 BUG();
4802 }
4803 return ret;
4804 }
4805
4806 int btrfs_alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
4807 struct btrfs_root *root,
4808 u64 root_objectid, u64 owner,
4809 u64 offset, struct btrfs_key *ins)
4810 {
4811 int ret;
4812
4813 BUG_ON(root_objectid == BTRFS_TREE_LOG_OBJECTID);
4814
4815 ret = btrfs_add_delayed_data_ref(trans, ins->objectid, ins->offset,
4816 0, root_objectid, owner, offset,
4817 BTRFS_ADD_DELAYED_EXTENT, NULL);
4818 return ret;
4819 }
4820
4821 /*
4822 * this is used by the tree logging recovery code. It records that
4823 * an extent has been allocated and makes sure to clear the free
4824 * space cache bits as well
4825 */
4826 int btrfs_alloc_logged_file_extent(struct btrfs_trans_handle *trans,
4827 struct btrfs_root *root,
4828 u64 root_objectid, u64 owner, u64 offset,
4829 struct btrfs_key *ins)
4830 {
4831 int ret;
4832 struct btrfs_block_group_cache *block_group;
4833 struct btrfs_caching_control *caching_ctl;
4834 u64 start = ins->objectid;
4835 u64 num_bytes = ins->offset;
4836
4837 block_group = btrfs_lookup_block_group(root->fs_info, ins->objectid);
4838 cache_block_group(block_group);
4839 caching_ctl = get_caching_control(block_group);
4840
4841 if (!caching_ctl) {
4842 BUG_ON(!block_group_cache_done(block_group));
4843 ret = btrfs_remove_free_space(block_group, start, num_bytes);
4844 BUG_ON(ret);
4845 } else {
4846 mutex_lock(&caching_ctl->mutex);
4847
4848 if (start >= caching_ctl->progress) {
4849 ret = add_excluded_extent(root, start, num_bytes);
4850 BUG_ON(ret);
4851 } else if (start + num_bytes <= caching_ctl->progress) {
4852 ret = btrfs_remove_free_space(block_group,
4853 start, num_bytes);
4854 BUG_ON(ret);
4855 } else {
4856 num_bytes = caching_ctl->progress - start;
4857 ret = btrfs_remove_free_space(block_group,
4858 start, num_bytes);
4859 BUG_ON(ret);
4860
4861 start = caching_ctl->progress;
4862 num_bytes = ins->objectid + ins->offset -
4863 caching_ctl->progress;
4864 ret = add_excluded_extent(root, start, num_bytes);
4865 BUG_ON(ret);
4866 }
4867
4868 mutex_unlock(&caching_ctl->mutex);
4869 put_caching_control(caching_ctl);
4870 }
4871
4872 update_reserved_extents(block_group, ins->offset, 1);
4873 btrfs_put_block_group(block_group);
4874 ret = alloc_reserved_file_extent(trans, root, 0, root_objectid,
4875 0, owner, offset, ins, 1);
4876 return ret;
4877 }
4878
4879 /*
4880 * finds a free extent and does all the dirty work required for allocation
4881 * returns the key for the extent through ins, and a tree buffer for
4882 * the first block of the extent through buf.
4883 *
4884 * returns 0 if everything worked, non-zero otherwise.
4885 */
4886 static int alloc_tree_block(struct btrfs_trans_handle *trans,
4887 struct btrfs_root *root,
4888 u64 num_bytes, u64 parent, u64 root_objectid,
4889 struct btrfs_disk_key *key, int level,
4890 u64 empty_size, u64 hint_byte, u64 search_end,
4891 struct btrfs_key *ins)
4892 {
4893 int ret;
4894 u64 flags = 0;
4895
4896 ret = btrfs_reserve_extent(trans, root, num_bytes, num_bytes,
4897 empty_size, hint_byte, search_end,
4898 ins, 0);
4899 if (ret)
4900 return ret;
4901
4902 if (root_objectid == BTRFS_TREE_RELOC_OBJECTID) {
4903 if (parent == 0)
4904 parent = ins->objectid;
4905 flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF;
4906 } else
4907 BUG_ON(parent > 0);
4908
4909 if (root_objectid != BTRFS_TREE_LOG_OBJECTID) {
4910 struct btrfs_delayed_extent_op *extent_op;
4911 extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
4912 BUG_ON(!extent_op);
4913 if (key)
4914 memcpy(&extent_op->key, key, sizeof(extent_op->key));
4915 else
4916 memset(&extent_op->key, 0, sizeof(extent_op->key));
4917 extent_op->flags_to_set = flags;
4918 extent_op->update_key = 1;
4919 extent_op->update_flags = 1;
4920 extent_op->is_data = 0;
4921
4922 ret = btrfs_add_delayed_tree_ref(trans, ins->objectid,
4923 ins->offset, parent, root_objectid,
4924 level, BTRFS_ADD_DELAYED_EXTENT,
4925 extent_op);
4926 BUG_ON(ret);
4927 }
4928
4929 if (root_objectid == root->root_key.objectid) {
4930 u64 used;
4931 spin_lock(&root->node_lock);
4932 used = btrfs_root_used(&root->root_item) + num_bytes;
4933 btrfs_set_root_used(&root->root_item, used);
4934 spin_unlock(&root->node_lock);
4935 }
4936 return ret;
4937 }
4938
4939 struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans,
4940 struct btrfs_root *root,
4941 u64 bytenr, u32 blocksize,
4942 int level)
4943 {
4944 struct extent_buffer *buf;
4945
4946 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
4947 if (!buf)
4948 return ERR_PTR(-ENOMEM);
4949 btrfs_set_header_generation(buf, trans->transid);
4950 btrfs_set_buffer_lockdep_class(buf, level);
4951 btrfs_tree_lock(buf);
4952 clean_tree_block(trans, root, buf);
4953
4954 btrfs_set_lock_blocking(buf);
4955 btrfs_set_buffer_uptodate(buf);
4956
4957 if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) {
4958 /*
4959 * we allow two log transactions at a time, use different
4960 * EXENT bit to differentiate dirty pages.
4961 */
4962 if (root->log_transid % 2 == 0)
4963 set_extent_dirty(&root->dirty_log_pages, buf->start,
4964 buf->start + buf->len - 1, GFP_NOFS);
4965 else
4966 set_extent_new(&root->dirty_log_pages, buf->start,
4967 buf->start + buf->len - 1, GFP_NOFS);
4968 } else {
4969 set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
4970 buf->start + buf->len - 1, GFP_NOFS);
4971 }
4972 trans->blocks_used++;
4973 /* this returns a buffer locked for blocking */
4974 return buf;
4975 }
4976
4977 /*
4978 * helper function to allocate a block for a given tree
4979 * returns the tree buffer or NULL.
4980 */
4981 struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
4982 struct btrfs_root *root, u32 blocksize,
4983 u64 parent, u64 root_objectid,
4984 struct btrfs_disk_key *key, int level,
4985 u64 hint, u64 empty_size)
4986 {
4987 struct btrfs_key ins;
4988 int ret;
4989 struct extent_buffer *buf;
4990
4991 ret = alloc_tree_block(trans, root, blocksize, parent, root_objectid,
4992 key, level, empty_size, hint, (u64)-1, &ins);
4993 if (ret) {
4994 BUG_ON(ret > 0);
4995 return ERR_PTR(ret);
4996 }
4997
4998 buf = btrfs_init_new_buffer(trans, root, ins.objectid,
4999 blocksize, level);
5000 return buf;
5001 }
5002
5003 struct walk_control {
5004 u64 refs[BTRFS_MAX_LEVEL];
5005 u64 flags[BTRFS_MAX_LEVEL];
5006 struct btrfs_key update_progress;
5007 int stage;
5008 int level;
5009 int shared_level;
5010 int update_ref;
5011 int keep_locks;
5012 int reada_slot;
5013 int reada_count;
5014 };
5015
5016 #define DROP_REFERENCE 1
5017 #define UPDATE_BACKREF 2
5018
5019 static noinline void reada_walk_down(struct btrfs_trans_handle *trans,
5020 struct btrfs_root *root,
5021 struct walk_control *wc,
5022 struct btrfs_path *path)
5023 {
5024 u64 bytenr;
5025 u64 generation;
5026 u64 refs;
5027 u64 flags;
5028 u64 last = 0;
5029 u32 nritems;
5030 u32 blocksize;
5031 struct btrfs_key key;
5032 struct extent_buffer *eb;
5033 int ret;
5034 int slot;
5035 int nread = 0;
5036
5037 if (path->slots[wc->level] < wc->reada_slot) {
5038 wc->reada_count = wc->reada_count * 2 / 3;
5039 wc->reada_count = max(wc->reada_count, 2);
5040 } else {
5041 wc->reada_count = wc->reada_count * 3 / 2;
5042 wc->reada_count = min_t(int, wc->reada_count,
5043 BTRFS_NODEPTRS_PER_BLOCK(root));
5044 }
5045
5046 eb = path->nodes[wc->level];
5047 nritems = btrfs_header_nritems(eb);
5048 blocksize = btrfs_level_size(root, wc->level - 1);
5049
5050 for (slot = path->slots[wc->level]; slot < nritems; slot++) {
5051 if (nread >= wc->reada_count)
5052 break;
5053
5054 cond_resched();
5055 bytenr = btrfs_node_blockptr(eb, slot);
5056 generation = btrfs_node_ptr_generation(eb, slot);
5057
5058 if (slot == path->slots[wc->level])
5059 goto reada;
5060
5061 if (wc->stage == UPDATE_BACKREF &&
5062 generation <= root->root_key.offset)
5063 continue;
5064
5065 /* We don't lock the tree block, it's OK to be racy here */
5066 ret = btrfs_lookup_extent_info(trans, root, bytenr, blocksize,
5067 &refs, &flags);
5068 BUG_ON(ret);
5069 BUG_ON(refs == 0);
5070
5071 if (wc->stage == DROP_REFERENCE) {
5072 if (refs == 1)
5073 goto reada;
5074
5075 if (wc->level == 1 &&
5076 (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
5077 continue;
5078 if (!wc->update_ref ||
5079 generation <= root->root_key.offset)
5080 continue;
5081 btrfs_node_key_to_cpu(eb, &key, slot);
5082 ret = btrfs_comp_cpu_keys(&key,
5083 &wc->update_progress);
5084 if (ret < 0)
5085 continue;
5086 } else {
5087 if (wc->level == 1 &&
5088 (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
5089 continue;
5090 }
5091 reada:
5092 ret = readahead_tree_block(root, bytenr, blocksize,
5093 generation);
5094 if (ret)
5095 break;
5096 last = bytenr + blocksize;
5097 nread++;
5098 }
5099 wc->reada_slot = slot;
5100 }
5101
5102 /*
5103 * hepler to process tree block while walking down the tree.
5104 *
5105 * when wc->stage == UPDATE_BACKREF, this function updates
5106 * back refs for pointers in the block.
5107 *
5108 * NOTE: return value 1 means we should stop walking down.
5109 */
5110 static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
5111 struct btrfs_root *root,
5112 struct btrfs_path *path,
5113 struct walk_control *wc, int lookup_info)
5114 {
5115 int level = wc->level;
5116 struct extent_buffer *eb = path->nodes[level];
5117 u64 flag = BTRFS_BLOCK_FLAG_FULL_BACKREF;
5118 int ret;
5119
5120 if (wc->stage == UPDATE_BACKREF &&
5121 btrfs_header_owner(eb) != root->root_key.objectid)
5122 return 1;
5123
5124 /*
5125 * when reference count of tree block is 1, it won't increase
5126 * again. once full backref flag is set, we never clear it.
5127 */
5128 if (lookup_info &&
5129 ((wc->stage == DROP_REFERENCE && wc->refs[level] != 1) ||
5130 (wc->stage == UPDATE_BACKREF && !(wc->flags[level] & flag)))) {
5131 BUG_ON(!path->locks[level]);
5132 ret = btrfs_lookup_extent_info(trans, root,
5133 eb->start, eb->len,
5134 &wc->refs[level],
5135 &wc->flags[level]);
5136 BUG_ON(ret);
5137 BUG_ON(wc->refs[level] == 0);
5138 }
5139
5140 if (wc->stage == DROP_REFERENCE) {
5141 if (wc->refs[level] > 1)
5142 return 1;
5143
5144 if (path->locks[level] && !wc->keep_locks) {
5145 btrfs_tree_unlock(eb);
5146 path->locks[level] = 0;
5147 }
5148 return 0;
5149 }
5150
5151 /* wc->stage == UPDATE_BACKREF */
5152 if (!(wc->flags[level] & flag)) {
5153 BUG_ON(!path->locks[level]);
5154 ret = btrfs_inc_ref(trans, root, eb, 1);
5155 BUG_ON(ret);
5156 ret = btrfs_dec_ref(trans, root, eb, 0);
5157 BUG_ON(ret);
5158 ret = btrfs_set_disk_extent_flags(trans, root, eb->start,
5159 eb->len, flag, 0);
5160 BUG_ON(ret);
5161 wc->flags[level] |= flag;
5162 }
5163
5164 /*
5165 * the block is shared by multiple trees, so it's not good to
5166 * keep the tree lock
5167 */
5168 if (path->locks[level] && level > 0) {
5169 btrfs_tree_unlock(eb);
5170 path->locks[level] = 0;
5171 }
5172 return 0;
5173 }
5174
5175 /*
5176 * hepler to process tree block pointer.
5177 *
5178 * when wc->stage == DROP_REFERENCE, this function checks
5179 * reference count of the block pointed to. if the block
5180 * is shared and we need update back refs for the subtree
5181 * rooted at the block, this function changes wc->stage to
5182 * UPDATE_BACKREF. if the block is shared and there is no
5183 * need to update back, this function drops the reference
5184 * to the block.
5185 *
5186 * NOTE: return value 1 means we should stop walking down.
5187 */
5188 static noinline int do_walk_down(struct btrfs_trans_handle *trans,
5189 struct btrfs_root *root,
5190 struct btrfs_path *path,
5191 struct walk_control *wc, int *lookup_info)
5192 {
5193 u64 bytenr;
5194 u64 generation;
5195 u64 parent;
5196 u32 blocksize;
5197 struct btrfs_key key;
5198 struct extent_buffer *next;
5199 int level = wc->level;
5200 int reada = 0;
5201 int ret = 0;
5202
5203 generation = btrfs_node_ptr_generation(path->nodes[level],
5204 path->slots[level]);
5205 /*
5206 * if the lower level block was created before the snapshot
5207 * was created, we know there is no need to update back refs
5208 * for the subtree
5209 */
5210 if (wc->stage == UPDATE_BACKREF &&
5211 generation <= root->root_key.offset) {
5212 *lookup_info = 1;
5213 return 1;
5214 }
5215
5216 bytenr = btrfs_node_blockptr(path->nodes[level], path->slots[level]);
5217 blocksize = btrfs_level_size(root, level - 1);
5218
5219 next = btrfs_find_tree_block(root, bytenr, blocksize);
5220 if (!next) {
5221 next = btrfs_find_create_tree_block(root, bytenr, blocksize);
5222 if (!next)
5223 return -ENOMEM;
5224 reada = 1;
5225 }
5226 btrfs_tree_lock(next);
5227 btrfs_set_lock_blocking(next);
5228
5229 ret = btrfs_lookup_extent_info(trans, root, bytenr, blocksize,
5230 &wc->refs[level - 1],
5231 &wc->flags[level - 1]);
5232 BUG_ON(ret);
5233 BUG_ON(wc->refs[level - 1] == 0);
5234 *lookup_info = 0;
5235
5236 if (wc->stage == DROP_REFERENCE) {
5237 if (wc->refs[level - 1] > 1) {
5238 if (level == 1 &&
5239 (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
5240 goto skip;
5241
5242 if (!wc->update_ref ||
5243 generation <= root->root_key.offset)
5244 goto skip;
5245
5246 btrfs_node_key_to_cpu(path->nodes[level], &key,
5247 path->slots[level]);
5248 ret = btrfs_comp_cpu_keys(&key, &wc->update_progress);
5249 if (ret < 0)
5250 goto skip;
5251
5252 wc->stage = UPDATE_BACKREF;
5253 wc->shared_level = level - 1;
5254 }
5255 } else {
5256 if (level == 1 &&
5257 (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
5258 goto skip;
5259 }
5260
5261 if (!btrfs_buffer_uptodate(next, generation)) {
5262 btrfs_tree_unlock(next);
5263 free_extent_buffer(next);
5264 next = NULL;
5265 *lookup_info = 1;
5266 }
5267
5268 if (!next) {
5269 if (reada && level == 1)
5270 reada_walk_down(trans, root, wc, path);
5271 next = read_tree_block(root, bytenr, blocksize, generation);
5272 btrfs_tree_lock(next);
5273 btrfs_set_lock_blocking(next);
5274 }
5275
5276 level--;
5277 BUG_ON(level != btrfs_header_level(next));
5278 path->nodes[level] = next;
5279 path->slots[level] = 0;
5280 path->locks[level] = 1;
5281 wc->level = level;
5282 if (wc->level == 1)
5283 wc->reada_slot = 0;
5284 return 0;
5285 skip:
5286 wc->refs[level - 1] = 0;
5287 wc->flags[level - 1] = 0;
5288 if (wc->stage == DROP_REFERENCE) {
5289 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF) {
5290 parent = path->nodes[level]->start;
5291 } else {
5292 BUG_ON(root->root_key.objectid !=
5293 btrfs_header_owner(path->nodes[level]));
5294 parent = 0;
5295 }
5296
5297 ret = btrfs_free_extent(trans, root, bytenr, blocksize, parent,
5298 root->root_key.objectid, level - 1, 0);
5299 BUG_ON(ret);
5300 }
5301 btrfs_tree_unlock(next);
5302 free_extent_buffer(next);
5303 *lookup_info = 1;
5304 return 1;
5305 }
5306
5307 /*
5308 * hepler to process tree block while walking up the tree.
5309 *
5310 * when wc->stage == DROP_REFERENCE, this function drops
5311 * reference count on the block.
5312 *
5313 * when wc->stage == UPDATE_BACKREF, this function changes
5314 * wc->stage back to DROP_REFERENCE if we changed wc->stage
5315 * to UPDATE_BACKREF previously while processing the block.
5316 *
5317 * NOTE: return value 1 means we should stop walking up.
5318 */
5319 static noinline int walk_up_proc(struct btrfs_trans_handle *trans,
5320 struct btrfs_root *root,
5321 struct btrfs_path *path,
5322 struct walk_control *wc)
5323 {
5324 int ret = 0;
5325 int level = wc->level;
5326 struct extent_buffer *eb = path->nodes[level];
5327 u64 parent = 0;
5328
5329 if (wc->stage == UPDATE_BACKREF) {
5330 BUG_ON(wc->shared_level < level);
5331 if (level < wc->shared_level)
5332 goto out;
5333
5334 ret = find_next_key(path, level + 1, &wc->update_progress);
5335 if (ret > 0)
5336 wc->update_ref = 0;
5337
5338 wc->stage = DROP_REFERENCE;
5339 wc->shared_level = -1;
5340 path->slots[level] = 0;
5341
5342 /*
5343 * check reference count again if the block isn't locked.
5344 * we should start walking down the tree again if reference
5345 * count is one.
5346 */
5347 if (!path->locks[level]) {
5348 BUG_ON(level == 0);
5349 btrfs_tree_lock(eb);
5350 btrfs_set_lock_blocking(eb);
5351 path->locks[level] = 1;
5352
5353 ret = btrfs_lookup_extent_info(trans, root,
5354 eb->start, eb->len,
5355 &wc->refs[level],
5356 &wc->flags[level]);
5357 BUG_ON(ret);
5358 BUG_ON(wc->refs[level] == 0);
5359 if (wc->refs[level] == 1) {
5360 btrfs_tree_unlock(eb);
5361 path->locks[level] = 0;
5362 return 1;
5363 }
5364 }
5365 }
5366
5367 /* wc->stage == DROP_REFERENCE */
5368 BUG_ON(wc->refs[level] > 1 && !path->locks[level]);
5369
5370 if (wc->refs[level] == 1) {
5371 if (level == 0) {
5372 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
5373 ret = btrfs_dec_ref(trans, root, eb, 1);
5374 else
5375 ret = btrfs_dec_ref(trans, root, eb, 0);
5376 BUG_ON(ret);
5377 }
5378 /* make block locked assertion in clean_tree_block happy */
5379 if (!path->locks[level] &&
5380 btrfs_header_generation(eb) == trans->transid) {
5381 btrfs_tree_lock(eb);
5382 btrfs_set_lock_blocking(eb);
5383 path->locks[level] = 1;
5384 }
5385 clean_tree_block(trans, root, eb);
5386 }
5387
5388 if (eb == root->node) {
5389 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
5390 parent = eb->start;
5391 else
5392 BUG_ON(root->root_key.objectid !=
5393 btrfs_header_owner(eb));
5394 } else {
5395 if (wc->flags[level + 1] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
5396 parent = path->nodes[level + 1]->start;
5397 else
5398 BUG_ON(root->root_key.objectid !=
5399 btrfs_header_owner(path->nodes[level + 1]));
5400 }
5401
5402 ret = btrfs_free_extent(trans, root, eb->start, eb->len, parent,
5403 root->root_key.objectid, level, 0);
5404 BUG_ON(ret);
5405 out:
5406 wc->refs[level] = 0;
5407 wc->flags[level] = 0;
5408 return ret;
5409 }
5410
5411 static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
5412 struct btrfs_root *root,
5413 struct btrfs_path *path,
5414 struct walk_control *wc)
5415 {
5416 int level = wc->level;
5417 int lookup_info = 1;
5418 int ret;
5419
5420 while (level >= 0) {
5421 ret = walk_down_proc(trans, root, path, wc, lookup_info);
5422 if (ret > 0)
5423 break;
5424
5425 if (level == 0)
5426 break;
5427
5428 if (path->slots[level] >=
5429 btrfs_header_nritems(path->nodes[level]))
5430 break;
5431
5432 ret = do_walk_down(trans, root, path, wc, &lookup_info);
5433 if (ret > 0) {
5434 path->slots[level]++;
5435 continue;
5436 } else if (ret < 0)
5437 return ret;
5438 level = wc->level;
5439 }
5440 return 0;
5441 }
5442
5443 static noinline int walk_up_tree(struct btrfs_trans_handle *trans,
5444 struct btrfs_root *root,
5445 struct btrfs_path *path,
5446 struct walk_control *wc, int max_level)
5447 {
5448 int level = wc->level;
5449 int ret;
5450
5451 path->slots[level] = btrfs_header_nritems(path->nodes[level]);
5452 while (level < max_level && path->nodes[level]) {
5453 wc->level = level;
5454 if (path->slots[level] + 1 <
5455 btrfs_header_nritems(path->nodes[level])) {
5456 path->slots[level]++;
5457 return 0;
5458 } else {
5459 ret = walk_up_proc(trans, root, path, wc);
5460 if (ret > 0)
5461 return 0;
5462
5463 if (path->locks[level]) {
5464 btrfs_tree_unlock(path->nodes[level]);
5465 path->locks[level] = 0;
5466 }
5467 free_extent_buffer(path->nodes[level]);
5468 path->nodes[level] = NULL;
5469 level++;
5470 }
5471 }
5472 return 1;
5473 }
5474
5475 /*
5476 * drop a subvolume tree.
5477 *
5478 * this function traverses the tree freeing any blocks that only
5479 * referenced by the tree.
5480 *
5481 * when a shared tree block is found. this function decreases its
5482 * reference count by one. if update_ref is true, this function
5483 * also make sure backrefs for the shared block and all lower level
5484 * blocks are properly updated.
5485 */
5486 int btrfs_drop_snapshot(struct btrfs_root *root, int update_ref)
5487 {
5488 struct btrfs_path *path;
5489 struct btrfs_trans_handle *trans;
5490 struct btrfs_root *tree_root = root->fs_info->tree_root;
5491 struct btrfs_root_item *root_item = &root->root_item;
5492 struct walk_control *wc;
5493 struct btrfs_key key;
5494 int err = 0;
5495 int ret;
5496 int level;
5497
5498 path = btrfs_alloc_path();
5499 BUG_ON(!path);
5500
5501 wc = kzalloc(sizeof(*wc), GFP_NOFS);
5502 BUG_ON(!wc);
5503
5504 trans = btrfs_start_transaction(tree_root, 1);
5505
5506 if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
5507 level = btrfs_header_level(root->node);
5508 path->nodes[level] = btrfs_lock_root_node(root);
5509 btrfs_set_lock_blocking(path->nodes[level]);
5510 path->slots[level] = 0;
5511 path->locks[level] = 1;
5512 memset(&wc->update_progress, 0,
5513 sizeof(wc->update_progress));
5514 } else {
5515 btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
5516 memcpy(&wc->update_progress, &key,
5517 sizeof(wc->update_progress));
5518
5519 level = root_item->drop_level;
5520 BUG_ON(level == 0);
5521 path->lowest_level = level;
5522 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
5523 path->lowest_level = 0;
5524 if (ret < 0) {
5525 err = ret;
5526 goto out;
5527 }
5528 WARN_ON(ret > 0);
5529
5530 /*
5531 * unlock our path, this is safe because only this
5532 * function is allowed to delete this snapshot
5533 */
5534 btrfs_unlock_up_safe(path, 0);
5535
5536 level = btrfs_header_level(root->node);
5537 while (1) {
5538 btrfs_tree_lock(path->nodes[level]);
5539 btrfs_set_lock_blocking(path->nodes[level]);
5540
5541 ret = btrfs_lookup_extent_info(trans, root,
5542 path->nodes[level]->start,
5543 path->nodes[level]->len,
5544 &wc->refs[level],
5545 &wc->flags[level]);
5546 BUG_ON(ret);
5547 BUG_ON(wc->refs[level] == 0);
5548
5549 if (level == root_item->drop_level)
5550 break;
5551
5552 btrfs_tree_unlock(path->nodes[level]);
5553 WARN_ON(wc->refs[level] != 1);
5554 level--;
5555 }
5556 }
5557
5558 wc->level = level;
5559 wc->shared_level = -1;
5560 wc->stage = DROP_REFERENCE;
5561 wc->update_ref = update_ref;
5562 wc->keep_locks = 0;
5563 wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(root);
5564
5565 while (1) {
5566 ret = walk_down_tree(trans, root, path, wc);
5567 if (ret < 0) {
5568 err = ret;
5569 break;
5570 }
5571
5572 ret = walk_up_tree(trans, root, path, wc, BTRFS_MAX_LEVEL);
5573 if (ret < 0) {
5574 err = ret;
5575 break;
5576 }
5577
5578 if (ret > 0) {
5579 BUG_ON(wc->stage != DROP_REFERENCE);
5580 break;
5581 }
5582
5583 if (wc->stage == DROP_REFERENCE) {
5584 level = wc->level;
5585 btrfs_node_key(path->nodes[level],
5586 &root_item->drop_progress,
5587 path->slots[level]);
5588 root_item->drop_level = level;
5589 }
5590
5591 BUG_ON(wc->level == 0);
5592 if (trans->transaction->in_commit ||
5593 trans->transaction->delayed_refs.flushing) {
5594 ret = btrfs_update_root(trans, tree_root,
5595 &root->root_key,
5596 root_item);
5597 BUG_ON(ret);
5598
5599 btrfs_end_transaction(trans, tree_root);
5600 trans = btrfs_start_transaction(tree_root, 1);
5601 } else {
5602 unsigned long update;
5603 update = trans->delayed_ref_updates;
5604 trans->delayed_ref_updates = 0;
5605 if (update)
5606 btrfs_run_delayed_refs(trans, tree_root,
5607 update);
5608 }
5609 }
5610 btrfs_release_path(root, path);
5611 BUG_ON(err);
5612
5613 ret = btrfs_del_root(trans, tree_root, &root->root_key);
5614 BUG_ON(ret);
5615
5616 if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
5617 ret = btrfs_find_last_root(tree_root, root->root_key.objectid,
5618 NULL, NULL);
5619 BUG_ON(ret < 0);
5620 if (ret > 0) {
5621 ret = btrfs_del_orphan_item(trans, tree_root,
5622 root->root_key.objectid);
5623 BUG_ON(ret);
5624 }
5625 }
5626
5627 if (root->in_radix) {
5628 btrfs_free_fs_root(tree_root->fs_info, root);
5629 } else {
5630 free_extent_buffer(root->node);
5631 free_extent_buffer(root->commit_root);
5632 kfree(root);
5633 }
5634 out:
5635 btrfs_end_transaction(trans, tree_root);
5636 kfree(wc);
5637 btrfs_free_path(path);
5638 return err;
5639 }
5640
5641 /*
5642 * drop subtree rooted at tree block 'node'.
5643 *
5644 * NOTE: this function will unlock and release tree block 'node'
5645 */
5646 int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
5647 struct btrfs_root *root,
5648 struct extent_buffer *node,
5649 struct extent_buffer *parent)
5650 {
5651 struct btrfs_path *path;
5652 struct walk_control *wc;
5653 int level;
5654 int parent_level;
5655 int ret = 0;
5656 int wret;
5657
5658 BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
5659
5660 path = btrfs_alloc_path();
5661 BUG_ON(!path);
5662
5663 wc = kzalloc(sizeof(*wc), GFP_NOFS);
5664 BUG_ON(!wc);
5665
5666 btrfs_assert_tree_locked(parent);
5667 parent_level = btrfs_header_level(parent);
5668 extent_buffer_get(parent);
5669 path->nodes[parent_level] = parent;
5670 path->slots[parent_level] = btrfs_header_nritems(parent);
5671
5672 btrfs_assert_tree_locked(node);
5673 level = btrfs_header_level(node);
5674 path->nodes[level] = node;
5675 path->slots[level] = 0;
5676 path->locks[level] = 1;
5677
5678 wc->refs[parent_level] = 1;
5679 wc->flags[parent_level] = BTRFS_BLOCK_FLAG_FULL_BACKREF;
5680 wc->level = level;
5681 wc->shared_level = -1;
5682 wc->stage = DROP_REFERENCE;
5683 wc->update_ref = 0;
5684 wc->keep_locks = 1;
5685 wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(root);
5686
5687 while (1) {
5688 wret = walk_down_tree(trans, root, path, wc);
5689 if (wret < 0) {
5690 ret = wret;
5691 break;
5692 }
5693
5694 wret = walk_up_tree(trans, root, path, wc, parent_level);
5695 if (wret < 0)
5696 ret = wret;
5697 if (wret != 0)
5698 break;
5699 }
5700
5701 kfree(wc);
5702 btrfs_free_path(path);
5703 return ret;
5704 }
5705
5706 #if 0
5707 static unsigned long calc_ra(unsigned long start, unsigned long last,
5708 unsigned long nr)
5709 {
5710 return min(last, start + nr - 1);
5711 }
5712
5713 static noinline int relocate_inode_pages(struct inode *inode, u64 start,
5714 u64 len)
5715 {
5716 u64 page_start;
5717 u64 page_end;
5718 unsigned long first_index;
5719 unsigned long last_index;
5720 unsigned long i;
5721 struct page *page;
5722 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
5723 struct file_ra_state *ra;
5724 struct btrfs_ordered_extent *ordered;
5725 unsigned int total_read = 0;
5726 unsigned int total_dirty = 0;
5727 int ret = 0;
5728
5729 ra = kzalloc(sizeof(*ra), GFP_NOFS);
5730
5731 mutex_lock(&inode->i_mutex);
5732 first_index = start >> PAGE_CACHE_SHIFT;
5733 last_index = (start + len - 1) >> PAGE_CACHE_SHIFT;
5734
5735 /* make sure the dirty trick played by the caller work */
5736 ret = invalidate_inode_pages2_range(inode->i_mapping,
5737 first_index, last_index);
5738 if (ret)
5739 goto out_unlock;
5740
5741 file_ra_state_init(ra, inode->i_mapping);
5742
5743 for (i = first_index ; i <= last_index; i++) {
5744 if (total_read % ra->ra_pages == 0) {
5745 btrfs_force_ra(inode->i_mapping, ra, NULL, i,
5746 calc_ra(i, last_index, ra->ra_pages));
5747 }
5748 total_read++;
5749 again:
5750 if (((u64)i << PAGE_CACHE_SHIFT) > i_size_read(inode))
5751 BUG_ON(1);
5752 page = grab_cache_page(inode->i_mapping, i);
5753 if (!page) {
5754 ret = -ENOMEM;
5755 goto out_unlock;
5756 }
5757 if (!PageUptodate(page)) {
5758 btrfs_readpage(NULL, page);
5759 lock_page(page);
5760 if (!PageUptodate(page)) {
5761 unlock_page(page);
5762 page_cache_release(page);
5763 ret = -EIO;
5764 goto out_unlock;
5765 }
5766 }
5767 wait_on_page_writeback(page);
5768
5769 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
5770 page_end = page_start + PAGE_CACHE_SIZE - 1;
5771 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
5772
5773 ordered = btrfs_lookup_ordered_extent(inode, page_start);
5774 if (ordered) {
5775 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
5776 unlock_page(page);
5777 page_cache_release(page);
5778 btrfs_start_ordered_extent(inode, ordered, 1);
5779 btrfs_put_ordered_extent(ordered);
5780 goto again;
5781 }
5782 set_page_extent_mapped(page);
5783
5784 if (i == first_index)
5785 set_extent_bits(io_tree, page_start, page_end,
5786 EXTENT_BOUNDARY, GFP_NOFS);
5787 btrfs_set_extent_delalloc(inode, page_start, page_end);
5788
5789 set_page_dirty(page);
5790 total_dirty++;
5791
5792 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
5793 unlock_page(page);
5794 page_cache_release(page);
5795 }
5796
5797 out_unlock:
5798 kfree(ra);
5799 mutex_unlock(&inode->i_mutex);
5800 balance_dirty_pages_ratelimited_nr(inode->i_mapping, total_dirty);
5801 return ret;
5802 }
5803
5804 static noinline int relocate_data_extent(struct inode *reloc_inode,
5805 struct btrfs_key *extent_key,
5806 u64 offset)
5807 {
5808 struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
5809 struct extent_map_tree *em_tree = &BTRFS_I(reloc_inode)->extent_tree;
5810 struct extent_map *em;
5811 u64 start = extent_key->objectid - offset;
5812 u64 end = start + extent_key->offset - 1;
5813
5814 em = alloc_extent_map(GFP_NOFS);
5815 BUG_ON(!em || IS_ERR(em));
5816
5817 em->start = start;
5818 em->len = extent_key->offset;
5819 em->block_len = extent_key->offset;
5820 em->block_start = extent_key->objectid;
5821 em->bdev = root->fs_info->fs_devices->latest_bdev;
5822 set_bit(EXTENT_FLAG_PINNED, &em->flags);
5823
5824 /* setup extent map to cheat btrfs_readpage */
5825 lock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS);
5826 while (1) {
5827 int ret;
5828 write_lock(&em_tree->lock);
5829 ret = add_extent_mapping(em_tree, em);
5830 write_unlock(&em_tree->lock);
5831 if (ret != -EEXIST) {
5832 free_extent_map(em);
5833 break;
5834 }
5835 btrfs_drop_extent_cache(reloc_inode, start, end, 0);
5836 }
5837 unlock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS);
5838
5839 return relocate_inode_pages(reloc_inode, start, extent_key->offset);
5840 }
5841
5842 struct btrfs_ref_path {
5843 u64 extent_start;
5844 u64 nodes[BTRFS_MAX_LEVEL];
5845 u64 root_objectid;
5846 u64 root_generation;
5847 u64 owner_objectid;
5848 u32 num_refs;
5849 int lowest_level;
5850 int current_level;
5851 int shared_level;
5852
5853 struct btrfs_key node_keys[BTRFS_MAX_LEVEL];
5854 u64 new_nodes[BTRFS_MAX_LEVEL];
5855 };
5856
5857 struct disk_extent {
5858 u64 ram_bytes;
5859 u64 disk_bytenr;
5860 u64 disk_num_bytes;
5861 u64 offset;
5862 u64 num_bytes;
5863 u8 compression;
5864 u8 encryption;
5865 u16 other_encoding;
5866 };
5867
5868 static int is_cowonly_root(u64 root_objectid)
5869 {
5870 if (root_objectid == BTRFS_ROOT_TREE_OBJECTID ||
5871 root_objectid == BTRFS_EXTENT_TREE_OBJECTID ||
5872 root_objectid == BTRFS_CHUNK_TREE_OBJECTID ||
5873 root_objectid == BTRFS_DEV_TREE_OBJECTID ||
5874 root_objectid == BTRFS_TREE_LOG_OBJECTID ||
5875 root_objectid == BTRFS_CSUM_TREE_OBJECTID)
5876 return 1;
5877 return 0;
5878 }
5879
5880 static noinline int __next_ref_path(struct btrfs_trans_handle *trans,
5881 struct btrfs_root *extent_root,
5882 struct btrfs_ref_path *ref_path,
5883 int first_time)
5884 {
5885 struct extent_buffer *leaf;
5886 struct btrfs_path *path;
5887 struct btrfs_extent_ref *ref;
5888 struct btrfs_key key;
5889 struct btrfs_key found_key;
5890 u64 bytenr;
5891 u32 nritems;
5892 int level;
5893 int ret = 1;
5894
5895 path = btrfs_alloc_path();
5896 if (!path)
5897 return -ENOMEM;
5898
5899 if (first_time) {
5900 ref_path->lowest_level = -1;
5901 ref_path->current_level = -1;
5902 ref_path->shared_level = -1;
5903 goto walk_up;
5904 }
5905 walk_down:
5906 level = ref_path->current_level - 1;
5907 while (level >= -1) {
5908 u64 parent;
5909 if (level < ref_path->lowest_level)
5910 break;
5911
5912 if (level >= 0)
5913 bytenr = ref_path->nodes[level];
5914 else
5915 bytenr = ref_path->extent_start;
5916 BUG_ON(bytenr == 0);
5917
5918 parent = ref_path->nodes[level + 1];
5919 ref_path->nodes[level + 1] = 0;
5920 ref_path->current_level = level;
5921 BUG_ON(parent == 0);
5922
5923 key.objectid = bytenr;
5924 key.offset = parent + 1;
5925 key.type = BTRFS_EXTENT_REF_KEY;
5926
5927 ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 0);
5928 if (ret < 0)
5929 goto out;
5930 BUG_ON(ret == 0);
5931
5932 leaf = path->nodes[0];
5933 nritems = btrfs_header_nritems(leaf);
5934 if (path->slots[0] >= nritems) {
5935 ret = btrfs_next_leaf(extent_root, path);
5936 if (ret < 0)
5937 goto out;
5938 if (ret > 0)
5939 goto next;
5940 leaf = path->nodes[0];
5941 }
5942
5943 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
5944 if (found_key.objectid == bytenr &&
5945 found_key.type == BTRFS_EXTENT_REF_KEY) {
5946 if (level < ref_path->shared_level)
5947 ref_path->shared_level = level;
5948 goto found;
5949 }
5950 next:
5951 level--;
5952 btrfs_release_path(extent_root, path);
5953 cond_resched();
5954 }
5955 /* reached lowest level */
5956 ret = 1;
5957 goto out;
5958 walk_up:
5959 level = ref_path->current_level;
5960 while (level < BTRFS_MAX_LEVEL - 1) {
5961 u64 ref_objectid;
5962
5963 if (level >= 0)
5964 bytenr = ref_path->nodes[level];
5965 else
5966 bytenr = ref_path->extent_start;
5967
5968 BUG_ON(bytenr == 0);
5969
5970 key.objectid = bytenr;
5971 key.offset = 0;
5972 key.type = BTRFS_EXTENT_REF_KEY;
5973
5974 ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 0);
5975 if (ret < 0)
5976 goto out;
5977
5978 leaf = path->nodes[0];
5979 nritems = btrfs_header_nritems(leaf);
5980 if (path->slots[0] >= nritems) {
5981 ret = btrfs_next_leaf(extent_root, path);
5982 if (ret < 0)
5983 goto out;
5984 if (ret > 0) {
5985 /* the extent was freed by someone */
5986 if (ref_path->lowest_level == level)
5987 goto out;
5988 btrfs_release_path(extent_root, path);
5989 goto walk_down;
5990 }
5991 leaf = path->nodes[0];
5992 }
5993
5994 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
5995 if (found_key.objectid != bytenr ||
5996 found_key.type != BTRFS_EXTENT_REF_KEY) {
5997 /* the extent was freed by someone */
5998 if (ref_path->lowest_level == level) {
5999 ret = 1;
6000 goto out;
6001 }
6002 btrfs_release_path(extent_root, path);
6003 goto walk_down;
6004 }
6005 found:
6006 ref = btrfs_item_ptr(leaf, path->slots[0],
6007 struct btrfs_extent_ref);
6008 ref_objectid = btrfs_ref_objectid(leaf, ref);
6009 if (ref_objectid < BTRFS_FIRST_FREE_OBJECTID) {
6010 if (first_time) {
6011 level = (int)ref_objectid;
6012 BUG_ON(level >= BTRFS_MAX_LEVEL);
6013 ref_path->lowest_level = level;
6014 ref_path->current_level = level;
6015 ref_path->nodes[level] = bytenr;
6016 } else {
6017 WARN_ON(ref_objectid != level);
6018 }
6019 } else {
6020 WARN_ON(level != -1);
6021 }
6022 first_time = 0;
6023
6024 if (ref_path->lowest_level == level) {
6025 ref_path->owner_objectid = ref_objectid;
6026 ref_path->num_refs = btrfs_ref_num_refs(leaf, ref);
6027 }
6028
6029 /*
6030 * the block is tree root or the block isn't in reference
6031 * counted tree.
6032 */
6033 if (found_key.objectid == found_key.offset ||
6034 is_cowonly_root(btrfs_ref_root(leaf, ref))) {
6035 ref_path->root_objectid = btrfs_ref_root(leaf, ref);
6036 ref_path->root_generation =
6037 btrfs_ref_generation(leaf, ref);
6038 if (level < 0) {
6039 /* special reference from the tree log */
6040 ref_path->nodes[0] = found_key.offset;
6041 ref_path->current_level = 0;
6042 }
6043 ret = 0;
6044 goto out;
6045 }
6046
6047 level++;
6048 BUG_ON(ref_path->nodes[level] != 0);
6049 ref_path->nodes[level] = found_key.offset;
6050 ref_path->current_level = level;
6051
6052 /*
6053 * the reference was created in the running transaction,
6054 * no need to continue walking up.
6055 */
6056 if (btrfs_ref_generation(leaf, ref) == trans->transid) {
6057 ref_path->root_objectid = btrfs_ref_root(leaf, ref);
6058 ref_path->root_generation =
6059 btrfs_ref_generation(leaf, ref);
6060 ret = 0;
6061 goto out;
6062 }
6063
6064 btrfs_release_path(extent_root, path);
6065 cond_resched();
6066 }
6067 /* reached max tree level, but no tree root found. */
6068 BUG();
6069 out:
6070 btrfs_free_path(path);
6071 return ret;
6072 }
6073
6074 static int btrfs_first_ref_path(struct btrfs_trans_handle *trans,
6075 struct btrfs_root *extent_root,
6076 struct btrfs_ref_path *ref_path,
6077 u64 extent_start)
6078 {
6079 memset(ref_path, 0, sizeof(*ref_path));
6080 ref_path->extent_start = extent_start;
6081
6082 return __next_ref_path(trans, extent_root, ref_path, 1);
6083 }
6084
6085 static int btrfs_next_ref_path(struct btrfs_trans_handle *trans,
6086 struct btrfs_root *extent_root,
6087 struct btrfs_ref_path *ref_path)
6088 {
6089 return __next_ref_path(trans, extent_root, ref_path, 0);
6090 }
6091
6092 static noinline int get_new_locations(struct inode *reloc_inode,
6093 struct btrfs_key *extent_key,
6094 u64 offset, int no_fragment,
6095 struct disk_extent **extents,
6096 int *nr_extents)
6097 {
6098 struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
6099 struct btrfs_path *path;
6100 struct btrfs_file_extent_item *fi;
6101 struct extent_buffer *leaf;
6102 struct disk_extent *exts = *extents;
6103 struct btrfs_key found_key;
6104 u64 cur_pos;
6105 u64 last_byte;
6106 u32 nritems;
6107 int nr = 0;
6108 int max = *nr_extents;
6109 int ret;
6110
6111 WARN_ON(!no_fragment && *extents);
6112 if (!exts) {
6113 max = 1;
6114 exts = kmalloc(sizeof(*exts) * max, GFP_NOFS);
6115 if (!exts)
6116 return -ENOMEM;
6117 }
6118
6119 path = btrfs_alloc_path();
6120 BUG_ON(!path);
6121
6122 cur_pos = extent_key->objectid - offset;
6123 last_byte = extent_key->objectid + extent_key->offset;
6124 ret = btrfs_lookup_file_extent(NULL, root, path, reloc_inode->i_ino,
6125 cur_pos, 0);
6126 if (ret < 0)
6127 goto out;
6128 if (ret > 0) {
6129 ret = -ENOENT;
6130 goto out;
6131 }
6132
6133 while (1) {
6134 leaf = path->nodes[0];
6135 nritems = btrfs_header_nritems(leaf);
6136 if (path->slots[0] >= nritems) {
6137 ret = btrfs_next_leaf(root, path);
6138 if (ret < 0)
6139 goto out;
6140 if (ret > 0)
6141 break;
6142 leaf = path->nodes[0];
6143 }
6144
6145 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
6146 if (found_key.offset != cur_pos ||
6147 found_key.type != BTRFS_EXTENT_DATA_KEY ||
6148 found_key.objectid != reloc_inode->i_ino)
6149 break;
6150
6151 fi = btrfs_item_ptr(leaf, path->slots[0],
6152 struct btrfs_file_extent_item);
6153 if (btrfs_file_extent_type(leaf, fi) !=
6154 BTRFS_FILE_EXTENT_REG ||
6155 btrfs_file_extent_disk_bytenr(leaf, fi) == 0)
6156 break;
6157
6158 if (nr == max) {
6159 struct disk_extent *old = exts;
6160 max *= 2;
6161 exts = kzalloc(sizeof(*exts) * max, GFP_NOFS);
6162 memcpy(exts, old, sizeof(*exts) * nr);
6163 if (old != *extents)
6164 kfree(old);
6165 }
6166
6167 exts[nr].disk_bytenr =
6168 btrfs_file_extent_disk_bytenr(leaf, fi);
6169 exts[nr].disk_num_bytes =
6170 btrfs_file_extent_disk_num_bytes(leaf, fi);
6171 exts[nr].offset = btrfs_file_extent_offset(leaf, fi);
6172 exts[nr].num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
6173 exts[nr].ram_bytes = btrfs_file_extent_ram_bytes(leaf, fi);
6174 exts[nr].compression = btrfs_file_extent_compression(leaf, fi);
6175 exts[nr].encryption = btrfs_file_extent_encryption(leaf, fi);
6176 exts[nr].other_encoding = btrfs_file_extent_other_encoding(leaf,
6177 fi);
6178 BUG_ON(exts[nr].offset > 0);
6179 BUG_ON(exts[nr].compression || exts[nr].encryption);
6180 BUG_ON(exts[nr].num_bytes != exts[nr].disk_num_bytes);
6181
6182 cur_pos += exts[nr].num_bytes;
6183 nr++;
6184
6185 if (cur_pos + offset >= last_byte)
6186 break;
6187
6188 if (no_fragment) {
6189 ret = 1;
6190 goto out;
6191 }
6192 path->slots[0]++;
6193 }
6194
6195 BUG_ON(cur_pos + offset > last_byte);
6196 if (cur_pos + offset < last_byte) {
6197 ret = -ENOENT;
6198 goto out;
6199 }
6200 ret = 0;
6201 out:
6202 btrfs_free_path(path);
6203 if (ret) {
6204 if (exts != *extents)
6205 kfree(exts);
6206 } else {
6207 *extents = exts;
6208 *nr_extents = nr;
6209 }
6210 return ret;
6211 }
6212
6213 static noinline int replace_one_extent(struct btrfs_trans_handle *trans,
6214 struct btrfs_root *root,
6215 struct btrfs_path *path,
6216 struct btrfs_key *extent_key,
6217 struct btrfs_key *leaf_key,
6218 struct btrfs_ref_path *ref_path,
6219 struct disk_extent *new_extents,
6220 int nr_extents)
6221 {
6222 struct extent_buffer *leaf;
6223 struct btrfs_file_extent_item *fi;
6224 struct inode *inode = NULL;
6225 struct btrfs_key key;
6226 u64 lock_start = 0;
6227 u64 lock_end = 0;
6228 u64 num_bytes;
6229 u64 ext_offset;
6230 u64 search_end = (u64)-1;
6231 u32 nritems;
6232 int nr_scaned = 0;
6233 int extent_locked = 0;
6234 int extent_type;
6235 int ret;
6236
6237 memcpy(&key, leaf_key, sizeof(key));
6238 if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS) {
6239 if (key.objectid < ref_path->owner_objectid ||
6240 (key.objectid == ref_path->owner_objectid &&
6241 key.type < BTRFS_EXTENT_DATA_KEY)) {
6242 key.objectid = ref_path->owner_objectid;
6243 key.type = BTRFS_EXTENT_DATA_KEY;
6244 key.offset = 0;
6245 }
6246 }
6247
6248 while (1) {
6249 ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
6250 if (ret < 0)
6251 goto out;
6252
6253 leaf = path->nodes[0];
6254 nritems = btrfs_header_nritems(leaf);
6255 next:
6256 if (extent_locked && ret > 0) {
6257 /*
6258 * the file extent item was modified by someone
6259 * before the extent got locked.
6260 */
6261 unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
6262 lock_end, GFP_NOFS);
6263 extent_locked = 0;
6264 }
6265
6266 if (path->slots[0] >= nritems) {
6267 if (++nr_scaned > 2)
6268 break;
6269
6270 BUG_ON(extent_locked);
6271 ret = btrfs_next_leaf(root, path);
6272 if (ret < 0)
6273 goto out;
6274 if (ret > 0)
6275 break;
6276 leaf = path->nodes[0];
6277 nritems = btrfs_header_nritems(leaf);
6278 }
6279
6280 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
6281
6282 if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS) {
6283 if ((key.objectid > ref_path->owner_objectid) ||
6284 (key.objectid == ref_path->owner_objectid &&
6285 key.type > BTRFS_EXTENT_DATA_KEY) ||
6286 key.offset >= search_end)
6287 break;
6288 }
6289
6290 if (inode && key.objectid != inode->i_ino) {
6291 BUG_ON(extent_locked);
6292 btrfs_release_path(root, path);
6293 mutex_unlock(&inode->i_mutex);
6294 iput(inode);
6295 inode = NULL;
6296 continue;
6297 }
6298
6299 if (key.type != BTRFS_EXTENT_DATA_KEY) {
6300 path->slots[0]++;
6301 ret = 1;
6302 goto next;
6303 }
6304 fi = btrfs_item_ptr(leaf, path->slots[0],
6305 struct btrfs_file_extent_item);
6306 extent_type = btrfs_file_extent_type(leaf, fi);
6307 if ((extent_type != BTRFS_FILE_EXTENT_REG &&
6308 extent_type != BTRFS_FILE_EXTENT_PREALLOC) ||
6309 (btrfs_file_extent_disk_bytenr(leaf, fi) !=
6310 extent_key->objectid)) {
6311 path->slots[0]++;
6312 ret = 1;
6313 goto next;
6314 }
6315
6316 num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
6317 ext_offset = btrfs_file_extent_offset(leaf, fi);
6318
6319 if (search_end == (u64)-1) {
6320 search_end = key.offset - ext_offset +
6321 btrfs_file_extent_ram_bytes(leaf, fi);
6322 }
6323
6324 if (!extent_locked) {
6325 lock_start = key.offset;
6326 lock_end = lock_start + num_bytes - 1;
6327 } else {
6328 if (lock_start > key.offset ||
6329 lock_end + 1 < key.offset + num_bytes) {
6330 unlock_extent(&BTRFS_I(inode)->io_tree,
6331 lock_start, lock_end, GFP_NOFS);
6332 extent_locked = 0;
6333 }
6334 }
6335
6336 if (!inode) {
6337 btrfs_release_path(root, path);
6338
6339 inode = btrfs_iget_locked(root->fs_info->sb,
6340 key.objectid, root);
6341 if (inode->i_state & I_NEW) {
6342 BTRFS_I(inode)->root = root;
6343 BTRFS_I(inode)->location.objectid =
6344 key.objectid;
6345 BTRFS_I(inode)->location.type =
6346 BTRFS_INODE_ITEM_KEY;
6347 BTRFS_I(inode)->location.offset = 0;
6348 btrfs_read_locked_inode(inode);
6349 unlock_new_inode(inode);
6350 }
6351 /*
6352 * some code call btrfs_commit_transaction while
6353 * holding the i_mutex, so we can't use mutex_lock
6354 * here.
6355 */
6356 if (is_bad_inode(inode) ||
6357 !mutex_trylock(&inode->i_mutex)) {
6358 iput(inode);
6359 inode = NULL;
6360 key.offset = (u64)-1;
6361 goto skip;
6362 }
6363 }
6364
6365 if (!extent_locked) {
6366 struct btrfs_ordered_extent *ordered;
6367
6368 btrfs_release_path(root, path);
6369
6370 lock_extent(&BTRFS_I(inode)->io_tree, lock_start,
6371 lock_end, GFP_NOFS);
6372 ordered = btrfs_lookup_first_ordered_extent(inode,
6373 lock_end);
6374 if (ordered &&
6375 ordered->file_offset <= lock_end &&
6376 ordered->file_offset + ordered->len > lock_start) {
6377 unlock_extent(&BTRFS_I(inode)->io_tree,
6378 lock_start, lock_end, GFP_NOFS);
6379 btrfs_start_ordered_extent(inode, ordered, 1);
6380 btrfs_put_ordered_extent(ordered);
6381 key.offset += num_bytes;
6382 goto skip;
6383 }
6384 if (ordered)
6385 btrfs_put_ordered_extent(ordered);
6386
6387 extent_locked = 1;
6388 continue;
6389 }
6390
6391 if (nr_extents == 1) {
6392 /* update extent pointer in place */
6393 btrfs_set_file_extent_disk_bytenr(leaf, fi,
6394 new_extents[0].disk_bytenr);
6395 btrfs_set_file_extent_disk_num_bytes(leaf, fi,
6396 new_extents[0].disk_num_bytes);
6397 btrfs_mark_buffer_dirty(leaf);
6398
6399 btrfs_drop_extent_cache(inode, key.offset,
6400 key.offset + num_bytes - 1, 0);
6401
6402 ret = btrfs_inc_extent_ref(trans, root,
6403 new_extents[0].disk_bytenr,
6404 new_extents[0].disk_num_bytes,
6405 leaf->start,
6406 root->root_key.objectid,
6407 trans->transid,
6408 key.objectid);
6409 BUG_ON(ret);
6410
6411 ret = btrfs_free_extent(trans, root,
6412 extent_key->objectid,
6413 extent_key->offset,
6414 leaf->start,
6415 btrfs_header_owner(leaf),
6416 btrfs_header_generation(leaf),
6417 key.objectid, 0);
6418 BUG_ON(ret);
6419
6420 btrfs_release_path(root, path);
6421 key.offset += num_bytes;
6422 } else {
6423 BUG_ON(1);
6424 #if 0
6425 u64 alloc_hint;
6426 u64 extent_len;
6427 int i;
6428 /*
6429 * drop old extent pointer at first, then insert the
6430 * new pointers one bye one
6431 */
6432 btrfs_release_path(root, path);
6433 ret = btrfs_drop_extents(trans, root, inode, key.offset,
6434 key.offset + num_bytes,
6435 key.offset, &alloc_hint);
6436 BUG_ON(ret);
6437
6438 for (i = 0; i < nr_extents; i++) {
6439 if (ext_offset >= new_extents[i].num_bytes) {
6440 ext_offset -= new_extents[i].num_bytes;
6441 continue;
6442 }
6443 extent_len = min(new_extents[i].num_bytes -
6444 ext_offset, num_bytes);
6445
6446 ret = btrfs_insert_empty_item(trans, root,
6447 path, &key,
6448 sizeof(*fi));
6449 BUG_ON(ret);
6450
6451 leaf = path->nodes[0];
6452 fi = btrfs_item_ptr(leaf, path->slots[0],
6453 struct btrfs_file_extent_item);
6454 btrfs_set_file_extent_generation(leaf, fi,
6455 trans->transid);
6456 btrfs_set_file_extent_type(leaf, fi,
6457 BTRFS_FILE_EXTENT_REG);
6458 btrfs_set_file_extent_disk_bytenr(leaf, fi,
6459 new_extents[i].disk_bytenr);
6460 btrfs_set_file_extent_disk_num_bytes(leaf, fi,
6461 new_extents[i].disk_num_bytes);
6462 btrfs_set_file_extent_ram_bytes(leaf, fi,
6463 new_extents[i].ram_bytes);
6464
6465 btrfs_set_file_extent_compression(leaf, fi,
6466 new_extents[i].compression);
6467 btrfs_set_file_extent_encryption(leaf, fi,
6468 new_extents[i].encryption);
6469 btrfs_set_file_extent_other_encoding(leaf, fi,
6470 new_extents[i].other_encoding);
6471
6472 btrfs_set_file_extent_num_bytes(leaf, fi,
6473 extent_len);
6474 ext_offset += new_extents[i].offset;
6475 btrfs_set_file_extent_offset(leaf, fi,
6476 ext_offset);
6477 btrfs_mark_buffer_dirty(leaf);
6478
6479 btrfs_drop_extent_cache(inode, key.offset,
6480 key.offset + extent_len - 1, 0);
6481
6482 ret = btrfs_inc_extent_ref(trans, root,
6483 new_extents[i].disk_bytenr,
6484 new_extents[i].disk_num_bytes,
6485 leaf->start,
6486 root->root_key.objectid,
6487 trans->transid, key.objectid);
6488 BUG_ON(ret);
6489 btrfs_release_path(root, path);
6490
6491 inode_add_bytes(inode, extent_len);
6492
6493 ext_offset = 0;
6494 num_bytes -= extent_len;
6495 key.offset += extent_len;
6496
6497 if (num_bytes == 0)
6498 break;
6499 }
6500 BUG_ON(i >= nr_extents);
6501 #endif
6502 }
6503
6504 if (extent_locked) {
6505 unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
6506 lock_end, GFP_NOFS);
6507 extent_locked = 0;
6508 }
6509 skip:
6510 if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS &&
6511 key.offset >= search_end)
6512 break;
6513
6514 cond_resched();
6515 }
6516 ret = 0;
6517 out:
6518 btrfs_release_path(root, path);
6519 if (inode) {
6520 mutex_unlock(&inode->i_mutex);
6521 if (extent_locked) {
6522 unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
6523 lock_end, GFP_NOFS);
6524 }
6525 iput(inode);
6526 }
6527 return ret;
6528 }
6529
6530 int btrfs_reloc_tree_cache_ref(struct btrfs_trans_handle *trans,
6531 struct btrfs_root *root,
6532 struct extent_buffer *buf, u64 orig_start)
6533 {
6534 int level;
6535 int ret;
6536
6537 BUG_ON(btrfs_header_generation(buf) != trans->transid);
6538 BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
6539
6540 level = btrfs_header_level(buf);
6541 if (level == 0) {
6542 struct btrfs_leaf_ref *ref;
6543 struct btrfs_leaf_ref *orig_ref;
6544
6545 orig_ref = btrfs_lookup_leaf_ref(root, orig_start);
6546 if (!orig_ref)
6547 return -ENOENT;
6548
6549 ref = btrfs_alloc_leaf_ref(root, orig_ref->nritems);
6550 if (!ref) {
6551 btrfs_free_leaf_ref(root, orig_ref);
6552 return -ENOMEM;
6553 }
6554
6555 ref->nritems = orig_ref->nritems;
6556 memcpy(ref->extents, orig_ref->extents,
6557 sizeof(ref->extents[0]) * ref->nritems);
6558
6559 btrfs_free_leaf_ref(root, orig_ref);
6560
6561 ref->root_gen = trans->transid;
6562 ref->bytenr = buf->start;
6563 ref->owner = btrfs_header_owner(buf);
6564 ref->generation = btrfs_header_generation(buf);
6565
6566 ret = btrfs_add_leaf_ref(root, ref, 0);
6567 WARN_ON(ret);
6568 btrfs_free_leaf_ref(root, ref);
6569 }
6570 return 0;
6571 }
6572
6573 static noinline int invalidate_extent_cache(struct btrfs_root *root,
6574 struct extent_buffer *leaf,
6575 struct btrfs_block_group_cache *group,
6576 struct btrfs_root *target_root)
6577 {
6578 struct btrfs_key key;
6579 struct inode *inode = NULL;
6580 struct btrfs_file_extent_item *fi;
6581 struct extent_state *cached_state = NULL;
6582 u64 num_bytes;
6583 u64 skip_objectid = 0;
6584 u32 nritems;
6585 u32 i;
6586
6587 nritems = btrfs_header_nritems(leaf);
6588 for (i = 0; i < nritems; i++) {
6589 btrfs_item_key_to_cpu(leaf, &key, i);
6590 if (key.objectid == skip_objectid ||
6591 key.type != BTRFS_EXTENT_DATA_KEY)
6592 continue;
6593 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
6594 if (btrfs_file_extent_type(leaf, fi) ==
6595 BTRFS_FILE_EXTENT_INLINE)
6596 continue;
6597 if (btrfs_file_extent_disk_bytenr(leaf, fi) == 0)
6598 continue;
6599 if (!inode || inode->i_ino != key.objectid) {
6600 iput(inode);
6601 inode = btrfs_ilookup(target_root->fs_info->sb,
6602 key.objectid, target_root, 1);
6603 }
6604 if (!inode) {
6605 skip_objectid = key.objectid;
6606 continue;
6607 }
6608 num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
6609
6610 lock_extent_bits(&BTRFS_I(inode)->io_tree, key.offset,
6611 key.offset + num_bytes - 1, 0, &cached_state,
6612 GFP_NOFS);
6613 btrfs_drop_extent_cache(inode, key.offset,
6614 key.offset + num_bytes - 1, 1);
6615 unlock_extent_cached(&BTRFS_I(inode)->io_tree, key.offset,
6616 key.offset + num_bytes - 1, &cached_state,
6617 GFP_NOFS);
6618 cond_resched();
6619 }
6620 iput(inode);
6621 return 0;
6622 }
6623
6624 static noinline int replace_extents_in_leaf(struct btrfs_trans_handle *trans,
6625 struct btrfs_root *root,
6626 struct extent_buffer *leaf,
6627 struct btrfs_block_group_cache *group,
6628 struct inode *reloc_inode)
6629 {
6630 struct btrfs_key key;
6631 struct btrfs_key extent_key;
6632 struct btrfs_file_extent_item *fi;
6633 struct btrfs_leaf_ref *ref;
6634 struct disk_extent *new_extent;
6635 u64 bytenr;
6636 u64 num_bytes;
6637 u32 nritems;
6638 u32 i;
6639 int ext_index;
6640 int nr_extent;
6641 int ret;
6642
6643 new_extent = kmalloc(sizeof(*new_extent), GFP_NOFS);
6644 BUG_ON(!new_extent);
6645
6646 ref = btrfs_lookup_leaf_ref(root, leaf->start);
6647 BUG_ON(!ref);
6648
6649 ext_index = -1;
6650 nritems = btrfs_header_nritems(leaf);
6651 for (i = 0; i < nritems; i++) {
6652 btrfs_item_key_to_cpu(leaf, &key, i);
6653 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
6654 continue;
6655 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
6656 if (btrfs_file_extent_type(leaf, fi) ==
6657 BTRFS_FILE_EXTENT_INLINE)
6658 continue;
6659 bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
6660 num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
6661 if (bytenr == 0)
6662 continue;
6663
6664 ext_index++;
6665 if (bytenr >= group->key.objectid + group->key.offset ||
6666 bytenr + num_bytes <= group->key.objectid)
6667 continue;
6668
6669 extent_key.objectid = bytenr;
6670 extent_key.offset = num_bytes;
6671 extent_key.type = BTRFS_EXTENT_ITEM_KEY;
6672 nr_extent = 1;
6673 ret = get_new_locations(reloc_inode, &extent_key,
6674 group->key.objectid, 1,
6675 &new_extent, &nr_extent);
6676 if (ret > 0)
6677 continue;
6678 BUG_ON(ret < 0);
6679
6680 BUG_ON(ref->extents[ext_index].bytenr != bytenr);
6681 BUG_ON(ref->extents[ext_index].num_bytes != num_bytes);
6682 ref->extents[ext_index].bytenr = new_extent->disk_bytenr;
6683 ref->extents[ext_index].num_bytes = new_extent->disk_num_bytes;
6684
6685 btrfs_set_file_extent_disk_bytenr(leaf, fi,
6686 new_extent->disk_bytenr);
6687 btrfs_set_file_extent_disk_num_bytes(leaf, fi,
6688 new_extent->disk_num_bytes);
6689 btrfs_mark_buffer_dirty(leaf);
6690
6691 ret = btrfs_inc_extent_ref(trans, root,
6692 new_extent->disk_bytenr,
6693 new_extent->disk_num_bytes,
6694 leaf->start,
6695 root->root_key.objectid,
6696 trans->transid, key.objectid);
6697 BUG_ON(ret);
6698
6699 ret = btrfs_free_extent(trans, root,
6700 bytenr, num_bytes, leaf->start,
6701 btrfs_header_owner(leaf),
6702 btrfs_header_generation(leaf),
6703 key.objectid, 0);
6704 BUG_ON(ret);
6705 cond_resched();
6706 }
6707 kfree(new_extent);
6708 BUG_ON(ext_index + 1 != ref->nritems);
6709 btrfs_free_leaf_ref(root, ref);
6710 return 0;
6711 }
6712
6713 int btrfs_free_reloc_root(struct btrfs_trans_handle *trans,
6714 struct btrfs_root *root)
6715 {
6716 struct btrfs_root *reloc_root;
6717 int ret;
6718
6719 if (root->reloc_root) {
6720 reloc_root = root->reloc_root;
6721 root->reloc_root = NULL;
6722 list_add(&reloc_root->dead_list,
6723 &root->fs_info->dead_reloc_roots);
6724
6725 btrfs_set_root_bytenr(&reloc_root->root_item,
6726 reloc_root->node->start);
6727 btrfs_set_root_level(&root->root_item,
6728 btrfs_header_level(reloc_root->node));
6729 memset(&reloc_root->root_item.drop_progress, 0,
6730 sizeof(struct btrfs_disk_key));
6731 reloc_root->root_item.drop_level = 0;
6732
6733 ret = btrfs_update_root(trans, root->fs_info->tree_root,
6734 &reloc_root->root_key,
6735 &reloc_root->root_item);
6736 BUG_ON(ret);
6737 }
6738 return 0;
6739 }
6740
6741 int btrfs_drop_dead_reloc_roots(struct btrfs_root *root)
6742 {
6743 struct btrfs_trans_handle *trans;
6744 struct btrfs_root *reloc_root;
6745 struct btrfs_root *prev_root = NULL;
6746 struct list_head dead_roots;
6747 int ret;
6748 unsigned long nr;
6749
6750 INIT_LIST_HEAD(&dead_roots);
6751 list_splice_init(&root->fs_info->dead_reloc_roots, &dead_roots);
6752
6753 while (!list_empty(&dead_roots)) {
6754 reloc_root = list_entry(dead_roots.prev,
6755 struct btrfs_root, dead_list);
6756 list_del_init(&reloc_root->dead_list);
6757
6758 BUG_ON(reloc_root->commit_root != NULL);
6759 while (1) {
6760 trans = btrfs_join_transaction(root, 1);
6761 BUG_ON(!trans);
6762
6763 mutex_lock(&root->fs_info->drop_mutex);
6764 ret = btrfs_drop_snapshot(trans, reloc_root);
6765 if (ret != -EAGAIN)
6766 break;
6767 mutex_unlock(&root->fs_info->drop_mutex);
6768
6769 nr = trans->blocks_used;
6770 ret = btrfs_end_transaction(trans, root);
6771 BUG_ON(ret);
6772 btrfs_btree_balance_dirty(root, nr);
6773 }
6774
6775 free_extent_buffer(reloc_root->node);
6776
6777 ret = btrfs_del_root(trans, root->fs_info->tree_root,
6778 &reloc_root->root_key);
6779 BUG_ON(ret);
6780 mutex_unlock(&root->fs_info->drop_mutex);
6781
6782 nr = trans->blocks_used;
6783 ret = btrfs_end_transaction(trans, root);
6784 BUG_ON(ret);
6785 btrfs_btree_balance_dirty(root, nr);
6786
6787 kfree(prev_root);
6788 prev_root = reloc_root;
6789 }
6790 if (prev_root) {
6791 btrfs_remove_leaf_refs(prev_root, (u64)-1, 0);
6792 kfree(prev_root);
6793 }
6794 return 0;
6795 }
6796
6797 int btrfs_add_dead_reloc_root(struct btrfs_root *root)
6798 {
6799 list_add(&root->dead_list, &root->fs_info->dead_reloc_roots);
6800 return 0;
6801 }
6802
6803 int btrfs_cleanup_reloc_trees(struct btrfs_root *root)
6804 {
6805 struct btrfs_root *reloc_root;
6806 struct btrfs_trans_handle *trans;
6807 struct btrfs_key location;
6808 int found;
6809 int ret;
6810
6811 mutex_lock(&root->fs_info->tree_reloc_mutex);
6812 ret = btrfs_find_dead_roots(root, BTRFS_TREE_RELOC_OBJECTID, NULL);
6813 BUG_ON(ret);
6814 found = !list_empty(&root->fs_info->dead_reloc_roots);
6815 mutex_unlock(&root->fs_info->tree_reloc_mutex);
6816
6817 if (found) {
6818 trans = btrfs_start_transaction(root, 1);
6819 BUG_ON(!trans);
6820 ret = btrfs_commit_transaction(trans, root);
6821 BUG_ON(ret);
6822 }
6823
6824 location.objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
6825 location.offset = (u64)-1;
6826 location.type = BTRFS_ROOT_ITEM_KEY;
6827
6828 reloc_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
6829 BUG_ON(!reloc_root);
6830 btrfs_orphan_cleanup(reloc_root);
6831 return 0;
6832 }
6833
6834 static noinline int init_reloc_tree(struct btrfs_trans_handle *trans,
6835 struct btrfs_root *root)
6836 {
6837 struct btrfs_root *reloc_root;
6838 struct extent_buffer *eb;
6839 struct btrfs_root_item *root_item;
6840 struct btrfs_key root_key;
6841 int ret;
6842
6843 BUG_ON(!root->ref_cows);
6844 if (root->reloc_root)
6845 return 0;
6846
6847 root_item = kmalloc(sizeof(*root_item), GFP_NOFS);
6848 BUG_ON(!root_item);
6849
6850 ret = btrfs_copy_root(trans, root, root->commit_root,
6851 &eb, BTRFS_TREE_RELOC_OBJECTID);
6852 BUG_ON(ret);
6853
6854 root_key.objectid = BTRFS_TREE_RELOC_OBJECTID;
6855 root_key.offset = root->root_key.objectid;
6856 root_key.type = BTRFS_ROOT_ITEM_KEY;
6857
6858 memcpy(root_item, &root->root_item, sizeof(root_item));
6859 btrfs_set_root_refs(root_item, 0);
6860 btrfs_set_root_bytenr(root_item, eb->start);
6861 btrfs_set_root_level(root_item, btrfs_header_level(eb));
6862 btrfs_set_root_generation(root_item, trans->transid);
6863
6864 btrfs_tree_unlock(eb);
6865 free_extent_buffer(eb);
6866
6867 ret = btrfs_insert_root(trans, root->fs_info->tree_root,
6868 &root_key, root_item);
6869 BUG_ON(ret);
6870 kfree(root_item);
6871
6872 reloc_root = btrfs_read_fs_root_no_radix(root->fs_info->tree_root,
6873 &root_key);
6874 BUG_ON(!reloc_root);
6875 reloc_root->last_trans = trans->transid;
6876 reloc_root->commit_root = NULL;
6877 reloc_root->ref_tree = &root->fs_info->reloc_ref_tree;
6878
6879 root->reloc_root = reloc_root;
6880 return 0;
6881 }
6882
6883 /*
6884 * Core function of space balance.
6885 *
6886 * The idea is using reloc trees to relocate tree blocks in reference
6887 * counted roots. There is one reloc tree for each subvol, and all
6888 * reloc trees share same root key objectid. Reloc trees are snapshots
6889 * of the latest committed roots of subvols (root->commit_root).
6890 *
6891 * To relocate a tree block referenced by a subvol, there are two steps.
6892 * COW the block through subvol's reloc tree, then update block pointer
6893 * in the subvol to point to the new block. Since all reloc trees share
6894 * same root key objectid, doing special handing for tree blocks owned
6895 * by them is easy. Once a tree block has been COWed in one reloc tree,
6896 * we can use the resulting new block directly when the same block is
6897 * required to COW again through other reloc trees. By this way, relocated
6898 * tree blocks are shared between reloc trees, so they are also shared
6899 * between subvols.
6900 */
6901 static noinline int relocate_one_path(struct btrfs_trans_handle *trans,
6902 struct btrfs_root *root,
6903 struct btrfs_path *path,
6904 struct btrfs_key *first_key,
6905 struct btrfs_ref_path *ref_path,
6906 struct btrfs_block_group_cache *group,
6907 struct inode *reloc_inode)
6908 {
6909 struct btrfs_root *reloc_root;
6910 struct extent_buffer *eb = NULL;
6911 struct btrfs_key *keys;
6912 u64 *nodes;
6913 int level;
6914 int shared_level;
6915 int lowest_level = 0;
6916 int ret;
6917
6918 if (ref_path->owner_objectid < BTRFS_FIRST_FREE_OBJECTID)
6919 lowest_level = ref_path->owner_objectid;
6920
6921 if (!root->ref_cows) {
6922 path->lowest_level = lowest_level;
6923 ret = btrfs_search_slot(trans, root, first_key, path, 0, 1);
6924 BUG_ON(ret < 0);
6925 path->lowest_level = 0;
6926 btrfs_release_path(root, path);
6927 return 0;
6928 }
6929
6930 mutex_lock(&root->fs_info->tree_reloc_mutex);
6931 ret = init_reloc_tree(trans, root);
6932 BUG_ON(ret);
6933 reloc_root = root->reloc_root;
6934
6935 shared_level = ref_path->shared_level;
6936 ref_path->shared_level = BTRFS_MAX_LEVEL - 1;
6937
6938 keys = ref_path->node_keys;
6939 nodes = ref_path->new_nodes;
6940 memset(&keys[shared_level + 1], 0,
6941 sizeof(*keys) * (BTRFS_MAX_LEVEL - shared_level - 1));
6942 memset(&nodes[shared_level + 1], 0,
6943 sizeof(*nodes) * (BTRFS_MAX_LEVEL - shared_level - 1));
6944
6945 if (nodes[lowest_level] == 0) {
6946 path->lowest_level = lowest_level;
6947 ret = btrfs_search_slot(trans, reloc_root, first_key, path,
6948 0, 1);
6949 BUG_ON(ret);
6950 for (level = lowest_level; level < BTRFS_MAX_LEVEL; level++) {
6951 eb = path->nodes[level];
6952 if (!eb || eb == reloc_root->node)
6953 break;
6954 nodes[level] = eb->start;
6955 if (level == 0)
6956 btrfs_item_key_to_cpu(eb, &keys[level], 0);
6957 else
6958 btrfs_node_key_to_cpu(eb, &keys[level], 0);
6959 }
6960 if (nodes[0] &&
6961 ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
6962 eb = path->nodes[0];
6963 ret = replace_extents_in_leaf(trans, reloc_root, eb,
6964 group, reloc_inode);
6965 BUG_ON(ret);
6966 }
6967 btrfs_release_path(reloc_root, path);
6968 } else {
6969 ret = btrfs_merge_path(trans, reloc_root, keys, nodes,
6970 lowest_level);
6971 BUG_ON(ret);
6972 }
6973
6974 /*
6975 * replace tree blocks in the fs tree with tree blocks in
6976 * the reloc tree.
6977 */
6978 ret = btrfs_merge_path(trans, root, keys, nodes, lowest_level);
6979 BUG_ON(ret < 0);
6980
6981 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
6982 ret = btrfs_search_slot(trans, reloc_root, first_key, path,
6983 0, 0);
6984 BUG_ON(ret);
6985 extent_buffer_get(path->nodes[0]);
6986 eb = path->nodes[0];
6987 btrfs_release_path(reloc_root, path);
6988 ret = invalidate_extent_cache(reloc_root, eb, group, root);
6989 BUG_ON(ret);
6990 free_extent_buffer(eb);
6991 }
6992
6993 mutex_unlock(&root->fs_info->tree_reloc_mutex);
6994 path->lowest_level = 0;
6995 return 0;
6996 }
6997
6998 static noinline int relocate_tree_block(struct btrfs_trans_handle *trans,
6999 struct btrfs_root *root,
7000 struct btrfs_path *path,
7001 struct btrfs_key *first_key,
7002 struct btrfs_ref_path *ref_path)
7003 {
7004 int ret;
7005
7006 ret = relocate_one_path(trans, root, path, first_key,
7007 ref_path, NULL, NULL);
7008 BUG_ON(ret);
7009
7010 return 0;
7011 }
7012
7013 static noinline int del_extent_zero(struct btrfs_trans_handle *trans,
7014 struct btrfs_root *extent_root,
7015 struct btrfs_path *path,
7016 struct btrfs_key *extent_key)
7017 {
7018 int ret;
7019
7020 ret = btrfs_search_slot(trans, extent_root, extent_key, path, -1, 1);
7021 if (ret)
7022 goto out;
7023 ret = btrfs_del_item(trans, extent_root, path);
7024 out:
7025 btrfs_release_path(extent_root, path);
7026 return ret;
7027 }
7028
7029 static noinline struct btrfs_root *read_ref_root(struct btrfs_fs_info *fs_info,
7030 struct btrfs_ref_path *ref_path)
7031 {
7032 struct btrfs_key root_key;
7033
7034 root_key.objectid = ref_path->root_objectid;
7035 root_key.type = BTRFS_ROOT_ITEM_KEY;
7036 if (is_cowonly_root(ref_path->root_objectid))
7037 root_key.offset = 0;
7038 else
7039 root_key.offset = (u64)-1;
7040
7041 return btrfs_read_fs_root_no_name(fs_info, &root_key);
7042 }
7043
7044 static noinline int relocate_one_extent(struct btrfs_root *extent_root,
7045 struct btrfs_path *path,
7046 struct btrfs_key *extent_key,
7047 struct btrfs_block_group_cache *group,
7048 struct inode *reloc_inode, int pass)
7049 {
7050 struct btrfs_trans_handle *trans;
7051 struct btrfs_root *found_root;
7052 struct btrfs_ref_path *ref_path = NULL;
7053 struct disk_extent *new_extents = NULL;
7054 int nr_extents = 0;
7055 int loops;
7056 int ret;
7057 int level;
7058 struct btrfs_key first_key;
7059 u64 prev_block = 0;
7060
7061
7062 trans = btrfs_start_transaction(extent_root, 1);
7063 BUG_ON(!trans);
7064
7065 if (extent_key->objectid == 0) {
7066 ret = del_extent_zero(trans, extent_root, path, extent_key);
7067 goto out;
7068 }
7069
7070 ref_path = kmalloc(sizeof(*ref_path), GFP_NOFS);
7071 if (!ref_path) {
7072 ret = -ENOMEM;
7073 goto out;
7074 }
7075
7076 for (loops = 0; ; loops++) {
7077 if (loops == 0) {
7078 ret = btrfs_first_ref_path(trans, extent_root, ref_path,
7079 extent_key->objectid);
7080 } else {
7081 ret = btrfs_next_ref_path(trans, extent_root, ref_path);
7082 }
7083 if (ret < 0)
7084 goto out;
7085 if (ret > 0)
7086 break;
7087
7088 if (ref_path->root_objectid == BTRFS_TREE_LOG_OBJECTID ||
7089 ref_path->root_objectid == BTRFS_TREE_RELOC_OBJECTID)
7090 continue;
7091
7092 found_root = read_ref_root(extent_root->fs_info, ref_path);
7093 BUG_ON(!found_root);
7094 /*
7095 * for reference counted tree, only process reference paths
7096 * rooted at the latest committed root.
7097 */
7098 if (found_root->ref_cows &&
7099 ref_path->root_generation != found_root->root_key.offset)
7100 continue;
7101
7102 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
7103 if (pass == 0) {
7104 /*
7105 * copy data extents to new locations
7106 */
7107 u64 group_start = group->key.objectid;
7108 ret = relocate_data_extent(reloc_inode,
7109 extent_key,
7110 group_start);
7111 if (ret < 0)
7112 goto out;
7113 break;
7114 }
7115 level = 0;
7116 } else {
7117 level = ref_path->owner_objectid;
7118 }
7119
7120 if (prev_block != ref_path->nodes[level]) {
7121 struct extent_buffer *eb;
7122 u64 block_start = ref_path->nodes[level];
7123 u64 block_size = btrfs_level_size(found_root, level);
7124
7125 eb = read_tree_block(found_root, block_start,
7126 block_size, 0);
7127 btrfs_tree_lock(eb);
7128 BUG_ON(level != btrfs_header_level(eb));
7129
7130 if (level == 0)
7131 btrfs_item_key_to_cpu(eb, &first_key, 0);
7132 else
7133 btrfs_node_key_to_cpu(eb, &first_key, 0);
7134
7135 btrfs_tree_unlock(eb);
7136 free_extent_buffer(eb);
7137 prev_block = block_start;
7138 }
7139
7140 mutex_lock(&extent_root->fs_info->trans_mutex);
7141 btrfs_record_root_in_trans(found_root);
7142 mutex_unlock(&extent_root->fs_info->trans_mutex);
7143 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
7144 /*
7145 * try to update data extent references while
7146 * keeping metadata shared between snapshots.
7147 */
7148 if (pass == 1) {
7149 ret = relocate_one_path(trans, found_root,
7150 path, &first_key, ref_path,
7151 group, reloc_inode);
7152 if (ret < 0)
7153 goto out;
7154 continue;
7155 }
7156 /*
7157 * use fallback method to process the remaining
7158 * references.
7159 */
7160 if (!new_extents) {
7161 u64 group_start = group->key.objectid;
7162 new_extents = kmalloc(sizeof(*new_extents),
7163 GFP_NOFS);
7164 nr_extents = 1;
7165 ret = get_new_locations(reloc_inode,
7166 extent_key,
7167 group_start, 1,
7168 &new_extents,
7169 &nr_extents);
7170 if (ret)
7171 goto out;
7172 }
7173 ret = replace_one_extent(trans, found_root,
7174 path, extent_key,
7175 &first_key, ref_path,
7176 new_extents, nr_extents);
7177 } else {
7178 ret = relocate_tree_block(trans, found_root, path,
7179 &first_key, ref_path);
7180 }
7181 if (ret < 0)
7182 goto out;
7183 }
7184 ret = 0;
7185 out:
7186 btrfs_end_transaction(trans, extent_root);
7187 kfree(new_extents);
7188 kfree(ref_path);
7189 return ret;
7190 }
7191 #endif
7192
7193 static u64 update_block_group_flags(struct btrfs_root *root, u64 flags)
7194 {
7195 u64 num_devices;
7196 u64 stripped = BTRFS_BLOCK_GROUP_RAID0 |
7197 BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10;
7198
7199 num_devices = root->fs_info->fs_devices->rw_devices;
7200 if (num_devices == 1) {
7201 stripped |= BTRFS_BLOCK_GROUP_DUP;
7202 stripped = flags & ~stripped;
7203
7204 /* turn raid0 into single device chunks */
7205 if (flags & BTRFS_BLOCK_GROUP_RAID0)
7206 return stripped;
7207
7208 /* turn mirroring into duplication */
7209 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
7210 BTRFS_BLOCK_GROUP_RAID10))
7211 return stripped | BTRFS_BLOCK_GROUP_DUP;
7212 return flags;
7213 } else {
7214 /* they already had raid on here, just return */
7215 if (flags & stripped)
7216 return flags;
7217
7218 stripped |= BTRFS_BLOCK_GROUP_DUP;
7219 stripped = flags & ~stripped;
7220
7221 /* switch duplicated blocks with raid1 */
7222 if (flags & BTRFS_BLOCK_GROUP_DUP)
7223 return stripped | BTRFS_BLOCK_GROUP_RAID1;
7224
7225 /* turn single device chunks into raid0 */
7226 return stripped | BTRFS_BLOCK_GROUP_RAID0;
7227 }
7228 return flags;
7229 }
7230
7231 static int __alloc_chunk_for_shrink(struct btrfs_root *root,
7232 struct btrfs_block_group_cache *shrink_block_group,
7233 int force)
7234 {
7235 struct btrfs_trans_handle *trans;
7236 u64 new_alloc_flags;
7237 u64 calc;
7238
7239 spin_lock(&shrink_block_group->lock);
7240 if (btrfs_block_group_used(&shrink_block_group->item) +
7241 shrink_block_group->reserved > 0) {
7242 spin_unlock(&shrink_block_group->lock);
7243
7244 trans = btrfs_start_transaction(root, 1);
7245 spin_lock(&shrink_block_group->lock);
7246
7247 new_alloc_flags = update_block_group_flags(root,
7248 shrink_block_group->flags);
7249 if (new_alloc_flags != shrink_block_group->flags) {
7250 calc =
7251 btrfs_block_group_used(&shrink_block_group->item);
7252 } else {
7253 calc = shrink_block_group->key.offset;
7254 }
7255 spin_unlock(&shrink_block_group->lock);
7256
7257 do_chunk_alloc(trans, root->fs_info->extent_root,
7258 calc + 2 * 1024 * 1024, new_alloc_flags, force);
7259
7260 btrfs_end_transaction(trans, root);
7261 } else
7262 spin_unlock(&shrink_block_group->lock);
7263 return 0;
7264 }
7265
7266
7267 int btrfs_prepare_block_group_relocation(struct btrfs_root *root,
7268 struct btrfs_block_group_cache *group)
7269
7270 {
7271 __alloc_chunk_for_shrink(root, group, 1);
7272 set_block_group_readonly(group);
7273 return 0;
7274 }
7275
7276 /*
7277 * checks to see if its even possible to relocate this block group.
7278 *
7279 * @return - -1 if it's not a good idea to relocate this block group, 0 if its
7280 * ok to go ahead and try.
7281 */
7282 int btrfs_can_relocate(struct btrfs_root *root, u64 bytenr)
7283 {
7284 struct btrfs_block_group_cache *block_group;
7285 struct btrfs_space_info *space_info;
7286 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
7287 struct btrfs_device *device;
7288 int full = 0;
7289 int ret = 0;
7290
7291 block_group = btrfs_lookup_block_group(root->fs_info, bytenr);
7292
7293 /* odd, couldn't find the block group, leave it alone */
7294 if (!block_group)
7295 return -1;
7296
7297 /* no bytes used, we're good */
7298 if (!btrfs_block_group_used(&block_group->item))
7299 goto out;
7300
7301 space_info = block_group->space_info;
7302 spin_lock(&space_info->lock);
7303
7304 full = space_info->full;
7305
7306 /*
7307 * if this is the last block group we have in this space, we can't
7308 * relocate it unless we're able to allocate a new chunk below.
7309 *
7310 * Otherwise, we need to make sure we have room in the space to handle
7311 * all of the extents from this block group. If we can, we're good
7312 */
7313 if ((space_info->total_bytes != block_group->key.offset) &&
7314 (space_info->bytes_used + space_info->bytes_reserved +
7315 space_info->bytes_pinned + space_info->bytes_readonly +
7316 btrfs_block_group_used(&block_group->item) <
7317 space_info->total_bytes)) {
7318 spin_unlock(&space_info->lock);
7319 goto out;
7320 }
7321 spin_unlock(&space_info->lock);
7322
7323 /*
7324 * ok we don't have enough space, but maybe we have free space on our
7325 * devices to allocate new chunks for relocation, so loop through our
7326 * alloc devices and guess if we have enough space. However, if we
7327 * were marked as full, then we know there aren't enough chunks, and we
7328 * can just return.
7329 */
7330 ret = -1;
7331 if (full)
7332 goto out;
7333
7334 mutex_lock(&root->fs_info->chunk_mutex);
7335 list_for_each_entry(device, &fs_devices->alloc_list, dev_alloc_list) {
7336 u64 min_free = btrfs_block_group_used(&block_group->item);
7337 u64 dev_offset, max_avail;
7338
7339 /*
7340 * check to make sure we can actually find a chunk with enough
7341 * space to fit our block group in.
7342 */
7343 if (device->total_bytes > device->bytes_used + min_free) {
7344 ret = find_free_dev_extent(NULL, device, min_free,
7345 &dev_offset, &max_avail);
7346 if (!ret)
7347 break;
7348 ret = -1;
7349 }
7350 }
7351 mutex_unlock(&root->fs_info->chunk_mutex);
7352 out:
7353 btrfs_put_block_group(block_group);
7354 return ret;
7355 }
7356
7357 static int find_first_block_group(struct btrfs_root *root,
7358 struct btrfs_path *path, struct btrfs_key *key)
7359 {
7360 int ret = 0;
7361 struct btrfs_key found_key;
7362 struct extent_buffer *leaf;
7363 int slot;
7364
7365 ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
7366 if (ret < 0)
7367 goto out;
7368
7369 while (1) {
7370 slot = path->slots[0];
7371 leaf = path->nodes[0];
7372 if (slot >= btrfs_header_nritems(leaf)) {
7373 ret = btrfs_next_leaf(root, path);
7374 if (ret == 0)
7375 continue;
7376 if (ret < 0)
7377 goto out;
7378 break;
7379 }
7380 btrfs_item_key_to_cpu(leaf, &found_key, slot);
7381
7382 if (found_key.objectid >= key->objectid &&
7383 found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) {
7384 ret = 0;
7385 goto out;
7386 }
7387 path->slots[0]++;
7388 }
7389 out:
7390 return ret;
7391 }
7392
7393 int btrfs_free_block_groups(struct btrfs_fs_info *info)
7394 {
7395 struct btrfs_block_group_cache *block_group;
7396 struct btrfs_space_info *space_info;
7397 struct btrfs_caching_control *caching_ctl;
7398 struct rb_node *n;
7399
7400 down_write(&info->extent_commit_sem);
7401 while (!list_empty(&info->caching_block_groups)) {
7402 caching_ctl = list_entry(info->caching_block_groups.next,
7403 struct btrfs_caching_control, list);
7404 list_del(&caching_ctl->list);
7405 put_caching_control(caching_ctl);
7406 }
7407 up_write(&info->extent_commit_sem);
7408
7409 spin_lock(&info->block_group_cache_lock);
7410 while ((n = rb_last(&info->block_group_cache_tree)) != NULL) {
7411 block_group = rb_entry(n, struct btrfs_block_group_cache,
7412 cache_node);
7413 rb_erase(&block_group->cache_node,
7414 &info->block_group_cache_tree);
7415 spin_unlock(&info->block_group_cache_lock);
7416
7417 down_write(&block_group->space_info->groups_sem);
7418 list_del(&block_group->list);
7419 up_write(&block_group->space_info->groups_sem);
7420
7421 if (block_group->cached == BTRFS_CACHE_STARTED)
7422 wait_block_group_cache_done(block_group);
7423
7424 btrfs_remove_free_space_cache(block_group);
7425 btrfs_put_block_group(block_group);
7426
7427 spin_lock(&info->block_group_cache_lock);
7428 }
7429 spin_unlock(&info->block_group_cache_lock);
7430
7431 /* now that all the block groups are freed, go through and
7432 * free all the space_info structs. This is only called during
7433 * the final stages of unmount, and so we know nobody is
7434 * using them. We call synchronize_rcu() once before we start,
7435 * just to be on the safe side.
7436 */
7437 synchronize_rcu();
7438
7439 while(!list_empty(&info->space_info)) {
7440 space_info = list_entry(info->space_info.next,
7441 struct btrfs_space_info,
7442 list);
7443
7444 list_del(&space_info->list);
7445 kfree(space_info);
7446 }
7447 return 0;
7448 }
7449
7450 int btrfs_read_block_groups(struct btrfs_root *root)
7451 {
7452 struct btrfs_path *path;
7453 int ret;
7454 struct btrfs_block_group_cache *cache;
7455 struct btrfs_fs_info *info = root->fs_info;
7456 struct btrfs_space_info *space_info;
7457 struct btrfs_key key;
7458 struct btrfs_key found_key;
7459 struct extent_buffer *leaf;
7460
7461 root = info->extent_root;
7462 key.objectid = 0;
7463 key.offset = 0;
7464 btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
7465 path = btrfs_alloc_path();
7466 if (!path)
7467 return -ENOMEM;
7468
7469 while (1) {
7470 ret = find_first_block_group(root, path, &key);
7471 if (ret > 0) {
7472 ret = 0;
7473 goto error;
7474 }
7475 if (ret != 0)
7476 goto error;
7477
7478 leaf = path->nodes[0];
7479 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
7480 cache = kzalloc(sizeof(*cache), GFP_NOFS);
7481 if (!cache) {
7482 ret = -ENOMEM;
7483 break;
7484 }
7485
7486 atomic_set(&cache->count, 1);
7487 spin_lock_init(&cache->lock);
7488 spin_lock_init(&cache->tree_lock);
7489 cache->fs_info = info;
7490 INIT_LIST_HEAD(&cache->list);
7491 INIT_LIST_HEAD(&cache->cluster_list);
7492
7493 /*
7494 * we only want to have 32k of ram per block group for keeping
7495 * track of free space, and if we pass 1/2 of that we want to
7496 * start converting things over to using bitmaps
7497 */
7498 cache->extents_thresh = ((1024 * 32) / 2) /
7499 sizeof(struct btrfs_free_space);
7500
7501 read_extent_buffer(leaf, &cache->item,
7502 btrfs_item_ptr_offset(leaf, path->slots[0]),
7503 sizeof(cache->item));
7504 memcpy(&cache->key, &found_key, sizeof(found_key));
7505
7506 key.objectid = found_key.objectid + found_key.offset;
7507 btrfs_release_path(root, path);
7508 cache->flags = btrfs_block_group_flags(&cache->item);
7509 cache->sectorsize = root->sectorsize;
7510
7511 /*
7512 * check for two cases, either we are full, and therefore
7513 * don't need to bother with the caching work since we won't
7514 * find any space, or we are empty, and we can just add all
7515 * the space in and be done with it. This saves us _alot_ of
7516 * time, particularly in the full case.
7517 */
7518 if (found_key.offset == btrfs_block_group_used(&cache->item)) {
7519 exclude_super_stripes(root, cache);
7520 cache->last_byte_to_unpin = (u64)-1;
7521 cache->cached = BTRFS_CACHE_FINISHED;
7522 free_excluded_extents(root, cache);
7523 } else if (btrfs_block_group_used(&cache->item) == 0) {
7524 exclude_super_stripes(root, cache);
7525 cache->last_byte_to_unpin = (u64)-1;
7526 cache->cached = BTRFS_CACHE_FINISHED;
7527 add_new_free_space(cache, root->fs_info,
7528 found_key.objectid,
7529 found_key.objectid +
7530 found_key.offset);
7531 free_excluded_extents(root, cache);
7532 }
7533
7534 ret = update_space_info(info, cache->flags, found_key.offset,
7535 btrfs_block_group_used(&cache->item),
7536 &space_info);
7537 BUG_ON(ret);
7538 cache->space_info = space_info;
7539 spin_lock(&cache->space_info->lock);
7540 cache->space_info->bytes_super += cache->bytes_super;
7541 spin_unlock(&cache->space_info->lock);
7542
7543 down_write(&space_info->groups_sem);
7544 list_add_tail(&cache->list, &space_info->block_groups);
7545 up_write(&space_info->groups_sem);
7546
7547 ret = btrfs_add_block_group_cache(root->fs_info, cache);
7548 BUG_ON(ret);
7549
7550 set_avail_alloc_bits(root->fs_info, cache->flags);
7551 if (btrfs_chunk_readonly(root, cache->key.objectid))
7552 set_block_group_readonly(cache);
7553 }
7554 ret = 0;
7555 error:
7556 btrfs_free_path(path);
7557 return ret;
7558 }
7559
7560 int btrfs_make_block_group(struct btrfs_trans_handle *trans,
7561 struct btrfs_root *root, u64 bytes_used,
7562 u64 type, u64 chunk_objectid, u64 chunk_offset,
7563 u64 size)
7564 {
7565 int ret;
7566 struct btrfs_root *extent_root;
7567 struct btrfs_block_group_cache *cache;
7568
7569 extent_root = root->fs_info->extent_root;
7570
7571 root->fs_info->last_trans_log_full_commit = trans->transid;
7572
7573 cache = kzalloc(sizeof(*cache), GFP_NOFS);
7574 if (!cache)
7575 return -ENOMEM;
7576
7577 cache->key.objectid = chunk_offset;
7578 cache->key.offset = size;
7579 cache->key.type = BTRFS_BLOCK_GROUP_ITEM_KEY;
7580 cache->sectorsize = root->sectorsize;
7581
7582 /*
7583 * we only want to have 32k of ram per block group for keeping track
7584 * of free space, and if we pass 1/2 of that we want to start
7585 * converting things over to using bitmaps
7586 */
7587 cache->extents_thresh = ((1024 * 32) / 2) /
7588 sizeof(struct btrfs_free_space);
7589 atomic_set(&cache->count, 1);
7590 spin_lock_init(&cache->lock);
7591 spin_lock_init(&cache->tree_lock);
7592 INIT_LIST_HEAD(&cache->list);
7593 INIT_LIST_HEAD(&cache->cluster_list);
7594
7595 btrfs_set_block_group_used(&cache->item, bytes_used);
7596 btrfs_set_block_group_chunk_objectid(&cache->item, chunk_objectid);
7597 cache->flags = type;
7598 btrfs_set_block_group_flags(&cache->item, type);
7599
7600 cache->last_byte_to_unpin = (u64)-1;
7601 cache->cached = BTRFS_CACHE_FINISHED;
7602 exclude_super_stripes(root, cache);
7603
7604 add_new_free_space(cache, root->fs_info, chunk_offset,
7605 chunk_offset + size);
7606
7607 free_excluded_extents(root, cache);
7608
7609 ret = update_space_info(root->fs_info, cache->flags, size, bytes_used,
7610 &cache->space_info);
7611 BUG_ON(ret);
7612
7613 spin_lock(&cache->space_info->lock);
7614 cache->space_info->bytes_super += cache->bytes_super;
7615 spin_unlock(&cache->space_info->lock);
7616
7617 down_write(&cache->space_info->groups_sem);
7618 list_add_tail(&cache->list, &cache->space_info->block_groups);
7619 up_write(&cache->space_info->groups_sem);
7620
7621 ret = btrfs_add_block_group_cache(root->fs_info, cache);
7622 BUG_ON(ret);
7623
7624 ret = btrfs_insert_item(trans, extent_root, &cache->key, &cache->item,
7625 sizeof(cache->item));
7626 BUG_ON(ret);
7627
7628 set_avail_alloc_bits(extent_root->fs_info, type);
7629
7630 return 0;
7631 }
7632
7633 int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
7634 struct btrfs_root *root, u64 group_start)
7635 {
7636 struct btrfs_path *path;
7637 struct btrfs_block_group_cache *block_group;
7638 struct btrfs_free_cluster *cluster;
7639 struct btrfs_key key;
7640 int ret;
7641
7642 root = root->fs_info->extent_root;
7643
7644 block_group = btrfs_lookup_block_group(root->fs_info, group_start);
7645 BUG_ON(!block_group);
7646 BUG_ON(!block_group->ro);
7647
7648 memcpy(&key, &block_group->key, sizeof(key));
7649
7650 /* make sure this block group isn't part of an allocation cluster */
7651 cluster = &root->fs_info->data_alloc_cluster;
7652 spin_lock(&cluster->refill_lock);
7653 btrfs_return_cluster_to_free_space(block_group, cluster);
7654 spin_unlock(&cluster->refill_lock);
7655
7656 /*
7657 * make sure this block group isn't part of a metadata
7658 * allocation cluster
7659 */
7660 cluster = &root->fs_info->meta_alloc_cluster;
7661 spin_lock(&cluster->refill_lock);
7662 btrfs_return_cluster_to_free_space(block_group, cluster);
7663 spin_unlock(&cluster->refill_lock);
7664
7665 path = btrfs_alloc_path();
7666 BUG_ON(!path);
7667
7668 spin_lock(&root->fs_info->block_group_cache_lock);
7669 rb_erase(&block_group->cache_node,
7670 &root->fs_info->block_group_cache_tree);
7671 spin_unlock(&root->fs_info->block_group_cache_lock);
7672
7673 down_write(&block_group->space_info->groups_sem);
7674 /*
7675 * we must use list_del_init so people can check to see if they
7676 * are still on the list after taking the semaphore
7677 */
7678 list_del_init(&block_group->list);
7679 up_write(&block_group->space_info->groups_sem);
7680
7681 if (block_group->cached == BTRFS_CACHE_STARTED)
7682 wait_block_group_cache_done(block_group);
7683
7684 btrfs_remove_free_space_cache(block_group);
7685
7686 spin_lock(&block_group->space_info->lock);
7687 block_group->space_info->total_bytes -= block_group->key.offset;
7688 block_group->space_info->bytes_readonly -= block_group->key.offset;
7689 spin_unlock(&block_group->space_info->lock);
7690
7691 btrfs_clear_space_info_full(root->fs_info);
7692
7693 btrfs_put_block_group(block_group);
7694 btrfs_put_block_group(block_group);
7695
7696 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
7697 if (ret > 0)
7698 ret = -EIO;
7699 if (ret < 0)
7700 goto out;
7701
7702 ret = btrfs_del_item(trans, root, path);
7703 out:
7704 btrfs_free_path(path);
7705 return ret;
7706 }
Linux with added FPC-III support