?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
sh: PVR detection for 2nd cut SH7786.
[linux-2.6.git] / fs / btrfs / extent-tree.c
blob56e50137d0e6d066266c074b1e07e10d10cb1b5f
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 "compat.h"
26 #include "hash.h"
27 #include "ctree.h"
28 #include "disk-io.h"
29 #include "print-tree.h"
30 #include "transaction.h"
31 #include "volumes.h"
32 #include "locking.h"
33 #include "free-space-cache.h"
34
35 static int update_block_group(struct btrfs_trans_handle *trans,
36 struct btrfs_root *root,
37 u64 bytenr, u64 num_bytes, int alloc,
38 int mark_free);
39 static int update_reserved_extents(struct btrfs_block_group_cache *cache,
40 u64 num_bytes, int reserve);
41 static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
42 struct btrfs_root *root,
43 u64 bytenr, u64 num_bytes, u64 parent,
44 u64 root_objectid, u64 owner_objectid,
45 u64 owner_offset, int refs_to_drop,
46 struct btrfs_delayed_extent_op *extra_op);
47 static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
48 struct extent_buffer *leaf,
49 struct btrfs_extent_item *ei);
50 static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
51 struct btrfs_root *root,
52 u64 parent, u64 root_objectid,
53 u64 flags, u64 owner, u64 offset,
54 struct btrfs_key *ins, int ref_mod);
55 static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
56 struct btrfs_root *root,
57 u64 parent, u64 root_objectid,
58 u64 flags, struct btrfs_disk_key *key,
59 int level, struct btrfs_key *ins);
60 static int do_chunk_alloc(struct btrfs_trans_handle *trans,
61 struct btrfs_root *extent_root, u64 alloc_bytes,
62 u64 flags, int force);
63 static int pin_down_bytes(struct btrfs_trans_handle *trans,
64 struct btrfs_root *root,
65 struct btrfs_path *path,
66 u64 bytenr, u64 num_bytes,
67 int is_data, int reserved,
68 struct extent_buffer **must_clean);
69 static int find_next_key(struct btrfs_path *path, int level,
70 struct btrfs_key *key);
71 static void dump_space_info(struct btrfs_space_info *info, u64 bytes,
72 int dump_block_groups);
73
74 static noinline int
75 block_group_cache_done(struct btrfs_block_group_cache *cache)
76 {
77 smp_mb();
78 return cache->cached == BTRFS_CACHE_FINISHED;
79 }
80
81 static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits)
82 {
83 return (cache->flags & bits) == bits;
84 }
85
86 /*
87 * this adds the block group to the fs_info rb tree for the block group
88 * cache
89 */
90 static int btrfs_add_block_group_cache(struct btrfs_fs_info *info,
91 struct btrfs_block_group_cache *block_group)
92 {
93 struct rb_node **p;
94 struct rb_node *parent = NULL;
95 struct btrfs_block_group_cache *cache;
96
97 spin_lock(&info->block_group_cache_lock);
98 p = &info->block_group_cache_tree.rb_node;
99
100 while (*p) {
101 parent = *p;
102 cache = rb_entry(parent, struct btrfs_block_group_cache,
103 cache_node);
104 if (block_group->key.objectid < cache->key.objectid) {
105 p = &(*p)->rb_left;
106 } else if (block_group->key.objectid > cache->key.objectid) {
107 p = &(*p)->rb_right;
108 } else {
109 spin_unlock(&info->block_group_cache_lock);
110 return -EEXIST;
111 }
112 }
113
114 rb_link_node(&block_group->cache_node, parent, p);
115 rb_insert_color(&block_group->cache_node,
116 &info->block_group_cache_tree);
117 spin_unlock(&info->block_group_cache_lock);
118
119 return 0;
120 }
121
122 /*
123 * This will return the block group at or after bytenr if contains is 0, else
124 * it will return the block group that contains the bytenr
125 */
126 static struct btrfs_block_group_cache *
127 block_group_cache_tree_search(struct btrfs_fs_info *info, u64 bytenr,
128 int contains)
129 {
130 struct btrfs_block_group_cache *cache, *ret = NULL;
131 struct rb_node *n;
132 u64 end, start;
133
134 spin_lock(&info->block_group_cache_lock);
135 n = info->block_group_cache_tree.rb_node;
136
137 while (n) {
138 cache = rb_entry(n, struct btrfs_block_group_cache,
139 cache_node);
140 end = cache->key.objectid + cache->key.offset - 1;
141 start = cache->key.objectid;
142
143 if (bytenr < start) {
144 if (!contains && (!ret || start < ret->key.objectid))
145 ret = cache;
146 n = n->rb_left;
147 } else if (bytenr > start) {
148 if (contains && bytenr <= end) {
149 ret = cache;
150 break;
151 }
152 n = n->rb_right;
153 } else {
154 ret = cache;
155 break;
156 }
157 }
158 if (ret)
159 atomic_inc(&ret->count);
160 spin_unlock(&info->block_group_cache_lock);
161
162 return ret;
163 }
164
165 static int add_excluded_extent(struct btrfs_root *root,
166 u64 start, u64 num_bytes)
167 {
168 u64 end = start + num_bytes - 1;
169 set_extent_bits(&root->fs_info->freed_extents[0],
170 start, end, EXTENT_UPTODATE, GFP_NOFS);
171 set_extent_bits(&root->fs_info->freed_extents[1],
172 start, end, EXTENT_UPTODATE, GFP_NOFS);
173 return 0;
174 }
175
176 static void free_excluded_extents(struct btrfs_root *root,
177 struct btrfs_block_group_cache *cache)
178 {
179 u64 start, end;
180
181 start = cache->key.objectid;
182 end = start + cache->key.offset - 1;
183
184 clear_extent_bits(&root->fs_info->freed_extents[0],
185 start, end, EXTENT_UPTODATE, GFP_NOFS);
186 clear_extent_bits(&root->fs_info->freed_extents[1],
187 start, end, EXTENT_UPTODATE, GFP_NOFS);
188 }
189
190 static int exclude_super_stripes(struct btrfs_root *root,
191 struct btrfs_block_group_cache *cache)
192 {
193 u64 bytenr;
194 u64 *logical;
195 int stripe_len;
196 int i, nr, ret;
197
198 if (cache->key.objectid < BTRFS_SUPER_INFO_OFFSET) {
199 stripe_len = BTRFS_SUPER_INFO_OFFSET - cache->key.objectid;
200 cache->bytes_super += stripe_len;
201 ret = add_excluded_extent(root, cache->key.objectid,
202 stripe_len);
203 BUG_ON(ret);
204 }
205
206 for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
207 bytenr = btrfs_sb_offset(i);
208 ret = btrfs_rmap_block(&root->fs_info->mapping_tree,
209 cache->key.objectid, bytenr,
210 0, &logical, &nr, &stripe_len);
211 BUG_ON(ret);
212
213 while (nr--) {
214 cache->bytes_super += stripe_len;
215 ret = add_excluded_extent(root, logical[nr],
216 stripe_len);
217 BUG_ON(ret);
218 }
219
220 kfree(logical);
221 }
222 return 0;
223 }
224
225 static struct btrfs_caching_control *
226 get_caching_control(struct btrfs_block_group_cache *cache)
227 {
228 struct btrfs_caching_control *ctl;
229
230 spin_lock(&cache->lock);
231 if (cache->cached != BTRFS_CACHE_STARTED) {
232 spin_unlock(&cache->lock);
233 return NULL;
234 }
235
236 ctl = cache->caching_ctl;
237 atomic_inc(&ctl->count);
238 spin_unlock(&cache->lock);
239 return ctl;
240 }
241
242 static void put_caching_control(struct btrfs_caching_control *ctl)
243 {
244 if (atomic_dec_and_test(&ctl->count))
245 kfree(ctl);
246 }
247
248 /*
249 * this is only called by cache_block_group, since we could have freed extents
250 * we need to check the pinned_extents for any extents that can't be used yet
251 * since their free space will be released as soon as the transaction commits.
252 */
253 static u64 add_new_free_space(struct btrfs_block_group_cache *block_group,
254 struct btrfs_fs_info *info, u64 start, u64 end)
255 {
256 u64 extent_start, extent_end, size, total_added = 0;
257 int ret;
258
259 while (start < end) {
260 ret = find_first_extent_bit(info->pinned_extents, start,
261 &extent_start, &extent_end,
262 EXTENT_DIRTY | EXTENT_UPTODATE);
263 if (ret)
264 break;
265
266 if (extent_start <= start) {
267 start = extent_end + 1;
268 } else if (extent_start > start && extent_start < end) {
269 size = extent_start - start;
270 total_added += size;
271 ret = btrfs_add_free_space(block_group, start,
272 size);
273 BUG_ON(ret);
274 start = extent_end + 1;
275 } else {
276 break;
277 }
278 }
279
280 if (start < end) {
281 size = end - start;
282 total_added += size;
283 ret = btrfs_add_free_space(block_group, start, size);
284 BUG_ON(ret);
285 }
286
287 return total_added;
288 }
289
290 static int caching_kthread(void *data)
291 {
292 struct btrfs_block_group_cache *block_group = data;
293 struct btrfs_fs_info *fs_info = block_group->fs_info;
294 struct btrfs_caching_control *caching_ctl = block_group->caching_ctl;
295 struct btrfs_root *extent_root = fs_info->extent_root;
296 struct btrfs_path *path;
297 struct extent_buffer *leaf;
298 struct btrfs_key key;
299 u64 total_found = 0;
300 u64 last = 0;
301 u32 nritems;
302 int ret = 0;
303
304 path = btrfs_alloc_path();
305 if (!path)
306 return -ENOMEM;
307
308 exclude_super_stripes(extent_root, block_group);
309 spin_lock(&block_group->space_info->lock);
310 block_group->space_info->bytes_super += block_group->bytes_super;
311 spin_unlock(&block_group->space_info->lock);
312
313 last = max_t(u64, block_group->key.objectid, BTRFS_SUPER_INFO_OFFSET);
314
315 /*
316 * We don't want to deadlock with somebody trying to allocate a new
317 * extent for the extent root while also trying to search the extent
318 * root to add free space. So we skip locking and search the commit
319 * root, since its read-only
320 */
321 path->skip_locking = 1;
322 path->search_commit_root = 1;
323 path->reada = 2;
324
325 key.objectid = last;
326 key.offset = 0;
327 key.type = BTRFS_EXTENT_ITEM_KEY;
328 again:
329 mutex_lock(&caching_ctl->mutex);
330 /* need to make sure the commit_root doesn't disappear */
331 down_read(&fs_info->extent_commit_sem);
332
333 ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
334 if (ret < 0)
335 goto err;
336
337 leaf = path->nodes[0];
338 nritems = btrfs_header_nritems(leaf);
339
340 while (1) {
341 smp_mb();
342 if (fs_info->closing > 1) {
343 last = (u64)-1;
344 break;
345 }
346
347 if (path->slots[0] < nritems) {
348 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
349 } else {
350 ret = find_next_key(path, 0, &key);
351 if (ret)
352 break;
353
354 caching_ctl->progress = last;
355 btrfs_release_path(extent_root, path);
356 up_read(&fs_info->extent_commit_sem);
357 mutex_unlock(&caching_ctl->mutex);
358 if (btrfs_transaction_in_commit(fs_info))
359 schedule_timeout(1);
360 else
361 cond_resched();
362 goto again;
363 }
364
365 if (key.objectid < block_group->key.objectid) {
366 path->slots[0]++;
367 continue;
368 }
369
370 if (key.objectid >= block_group->key.objectid +
371 block_group->key.offset)
372 break;
373
374 if (key.type == BTRFS_EXTENT_ITEM_KEY) {
375 total_found += add_new_free_space(block_group,
376 fs_info, last,
377 key.objectid);
378 last = key.objectid + key.offset;
379
380 if (total_found > (1024 * 1024 * 2)) {
381 total_found = 0;
382 wake_up(&caching_ctl->wait);
383 }
384 }
385 path->slots[0]++;
386 }
387 ret = 0;
388
389 total_found += add_new_free_space(block_group, fs_info, last,
390 block_group->key.objectid +
391 block_group->key.offset);
392 caching_ctl->progress = (u64)-1;
393
394 spin_lock(&block_group->lock);
395 block_group->caching_ctl = NULL;
396 block_group->cached = BTRFS_CACHE_FINISHED;
397 spin_unlock(&block_group->lock);
398
399 err:
400 btrfs_free_path(path);
401 up_read(&fs_info->extent_commit_sem);
402
403 free_excluded_extents(extent_root, block_group);
404
405 mutex_unlock(&caching_ctl->mutex);
406 wake_up(&caching_ctl->wait);
407
408 put_caching_control(caching_ctl);
409 atomic_dec(&block_group->space_info->caching_threads);
410 return 0;
411 }
412
413 static int cache_block_group(struct btrfs_block_group_cache *cache)
414 {
415 struct btrfs_fs_info *fs_info = cache->fs_info;
416 struct btrfs_caching_control *caching_ctl;
417 struct task_struct *tsk;
418 int ret = 0;
419
420 smp_mb();
421 if (cache->cached != BTRFS_CACHE_NO)
422 return 0;
423
424 caching_ctl = kzalloc(sizeof(*caching_ctl), GFP_KERNEL);
425 BUG_ON(!caching_ctl);
426
427 INIT_LIST_HEAD(&caching_ctl->list);
428 mutex_init(&caching_ctl->mutex);
429 init_waitqueue_head(&caching_ctl->wait);
430 caching_ctl->block_group = cache;
431 caching_ctl->progress = cache->key.objectid;
432 /* one for caching kthread, one for caching block group list */
433 atomic_set(&caching_ctl->count, 2);
434
435 spin_lock(&cache->lock);
436 if (cache->cached != BTRFS_CACHE_NO) {
437 spin_unlock(&cache->lock);
438 kfree(caching_ctl);
439 return 0;
440 }
441 cache->caching_ctl = caching_ctl;
442 cache->cached = BTRFS_CACHE_STARTED;
443 spin_unlock(&cache->lock);
444
445 down_write(&fs_info->extent_commit_sem);
446 list_add_tail(&caching_ctl->list, &fs_info->caching_block_groups);
447 up_write(&fs_info->extent_commit_sem);
448
449 atomic_inc(&cache->space_info->caching_threads);
450
451 tsk = kthread_run(caching_kthread, cache, "btrfs-cache-%llu\n",
452 cache->key.objectid);
453 if (IS_ERR(tsk)) {
454 ret = PTR_ERR(tsk);
455 printk(KERN_ERR "error running thread %d\n", ret);
456 BUG();
457 }
458
459 return ret;
460 }
461
462 /*
463 * return the block group that starts at or after bytenr
464 */
465 static struct btrfs_block_group_cache *
466 btrfs_lookup_first_block_group(struct btrfs_fs_info *info, u64 bytenr)
467 {
468 struct btrfs_block_group_cache *cache;
469
470 cache = block_group_cache_tree_search(info, bytenr, 0);
471
472 return cache;
473 }
474
475 /*
476 * return the block group that contains the given bytenr
477 */
478 struct btrfs_block_group_cache *btrfs_lookup_block_group(
479 struct btrfs_fs_info *info,
480 u64 bytenr)
481 {
482 struct btrfs_block_group_cache *cache;
483
484 cache = block_group_cache_tree_search(info, bytenr, 1);
485
486 return cache;
487 }
488
489 void btrfs_put_block_group(struct btrfs_block_group_cache *cache)
490 {
491 if (atomic_dec_and_test(&cache->count))
492 kfree(cache);
493 }
494
495 static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info,
496 u64 flags)
497 {
498 struct list_head *head = &info->space_info;
499 struct btrfs_space_info *found;
500
501 rcu_read_lock();
502 list_for_each_entry_rcu(found, head, list) {
503 if (found->flags == flags) {
504 rcu_read_unlock();
505 return found;
506 }
507 }
508 rcu_read_unlock();
509 return NULL;
510 }
511
512 /*
513 * after adding space to the filesystem, we need to clear the full flags
514 * on all the space infos.
515 */
516 void btrfs_clear_space_info_full(struct btrfs_fs_info *info)
517 {
518 struct list_head *head = &info->space_info;
519 struct btrfs_space_info *found;
520
521 rcu_read_lock();
522 list_for_each_entry_rcu(found, head, list)
523 found->full = 0;
524 rcu_read_unlock();
525 }
526
527 static u64 div_factor(u64 num, int factor)
528 {
529 if (factor == 10)
530 return num;
531 num *= factor;
532 do_div(num, 10);
533 return num;
534 }
535
536 u64 btrfs_find_block_group(struct btrfs_root *root,
537 u64 search_start, u64 search_hint, int owner)
538 {
539 struct btrfs_block_group_cache *cache;
540 u64 used;
541 u64 last = max(search_hint, search_start);
542 u64 group_start = 0;
543 int full_search = 0;
544 int factor = 9;
545 int wrapped = 0;
546 again:
547 while (1) {
548 cache = btrfs_lookup_first_block_group(root->fs_info, last);
549 if (!cache)
550 break;
551
552 spin_lock(&cache->lock);
553 last = cache->key.objectid + cache->key.offset;
554 used = btrfs_block_group_used(&cache->item);
555
556 if ((full_search || !cache->ro) &&
557 block_group_bits(cache, BTRFS_BLOCK_GROUP_METADATA)) {
558 if (used + cache->pinned + cache->reserved <
559 div_factor(cache->key.offset, factor)) {
560 group_start = cache->key.objectid;
561 spin_unlock(&cache->lock);
562 btrfs_put_block_group(cache);
563 goto found;
564 }
565 }
566 spin_unlock(&cache->lock);
567 btrfs_put_block_group(cache);
568 cond_resched();
569 }
570 if (!wrapped) {
571 last = search_start;
572 wrapped = 1;
573 goto again;
574 }
575 if (!full_search && factor < 10) {
576 last = search_start;
577 full_search = 1;
578 factor = 10;
579 goto again;
580 }
581 found:
582 return group_start;
583 }
584
585 /* simple helper to search for an existing extent at a given offset */
586 int btrfs_lookup_extent(struct btrfs_root *root, u64 start, u64 len)
587 {
588 int ret;
589 struct btrfs_key key;
590 struct btrfs_path *path;
591
592 path = btrfs_alloc_path();
593 BUG_ON(!path);
594 key.objectid = start;
595 key.offset = len;
596 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
597 ret = btrfs_search_slot(NULL, root->fs_info->extent_root, &key, path,
598 0, 0);
599 btrfs_free_path(path);
600 return ret;
601 }
602
603 /*
604 * Back reference rules. Back refs have three main goals:
605 *
606 * 1) differentiate between all holders of references to an extent so that
607 * when a reference is dropped we can make sure it was a valid reference
608 * before freeing the extent.
609 *
610 * 2) Provide enough information to quickly find the holders of an extent
611 * if we notice a given block is corrupted or bad.
612 *
613 * 3) Make it easy to migrate blocks for FS shrinking or storage pool
614 * maintenance. This is actually the same as #2, but with a slightly
615 * different use case.
616 *
617 * There are two kinds of back refs. The implicit back refs is optimized
618 * for pointers in non-shared tree blocks. For a given pointer in a block,
619 * back refs of this kind provide information about the block's owner tree
620 * and the pointer's key. These information allow us to find the block by
621 * b-tree searching. The full back refs is for pointers in tree blocks not
622 * referenced by their owner trees. The location of tree block is recorded
623 * in the back refs. Actually the full back refs is generic, and can be
624 * used in all cases the implicit back refs is used. The major shortcoming
625 * of the full back refs is its overhead. Every time a tree block gets
626 * COWed, we have to update back refs entry for all pointers in it.
627 *
628 * For a newly allocated tree block, we use implicit back refs for
629 * pointers in it. This means most tree related operations only involve
630 * implicit back refs. For a tree block created in old transaction, the
631 * only way to drop a reference to it is COW it. So we can detect the
632 * event that tree block loses its owner tree's reference and do the
633 * back refs conversion.
634 *
635 * When a tree block is COW'd through a tree, there are four cases:
636 *
637 * The reference count of the block is one and the tree is the block's
638 * owner tree. Nothing to do in this case.
639 *
640 * The reference count of the block is one and the tree is not the
641 * block's owner tree. In this case, full back refs is used for pointers
642 * in the block. Remove these full back refs, add implicit back refs for
643 * every pointers in the new block.
644 *
645 * The reference count of the block is greater than one and the tree is
646 * the block's owner tree. In this case, implicit back refs is used for
647 * pointers in the block. Add full back refs for every pointers in the
648 * block, increase lower level extents' reference counts. The original
649 * implicit back refs are entailed to the new block.
650 *
651 * The reference count of the block is greater than one and the tree is
652 * not the block's owner tree. Add implicit back refs for every pointer in
653 * the new block, increase lower level extents' reference count.
654 *
655 * Back Reference Key composing:
656 *
657 * The key objectid corresponds to the first byte in the extent,
658 * The key type is used to differentiate between types of back refs.
659 * There are different meanings of the key offset for different types
660 * of back refs.
661 *
662 * File extents can be referenced by:
663 *
664 * - multiple snapshots, subvolumes, or different generations in one subvol
665 * - different files inside a single subvolume
666 * - different offsets inside a file (bookend extents in file.c)
667 *
668 * The extent ref structure for the implicit back refs has fields for:
669 *
670 * - Objectid of the subvolume root
671 * - objectid of the file holding the reference
672 * - original offset in the file
673 * - how many bookend extents
674 *
675 * The key offset for the implicit back refs is hash of the first
676 * three fields.
677 *
678 * The extent ref structure for the full back refs has field for:
679 *
680 * - number of pointers in the tree leaf
681 *
682 * The key offset for the implicit back refs is the first byte of
683 * the tree leaf
684 *
685 * When a file extent is allocated, The implicit back refs is used.
686 * the fields are filled in:
687 *
688 * (root_key.objectid, inode objectid, offset in file, 1)
689 *
690 * When a file extent is removed file truncation, we find the
691 * corresponding implicit back refs and check the following fields:
692 *
693 * (btrfs_header_owner(leaf), inode objectid, offset in file)
694 *
695 * Btree extents can be referenced by:
696 *
697 * - Different subvolumes
698 *
699 * Both the implicit back refs and the full back refs for tree blocks
700 * only consist of key. The key offset for the implicit back refs is
701 * objectid of block's owner tree. The key offset for the full back refs
702 * is the first byte of parent block.
703 *
704 * When implicit back refs is used, information about the lowest key and
705 * level of the tree block are required. These information are stored in
706 * tree block info structure.
707 */
708
709 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
710 static int convert_extent_item_v0(struct btrfs_trans_handle *trans,
711 struct btrfs_root *root,
712 struct btrfs_path *path,
713 u64 owner, u32 extra_size)
714 {
715 struct btrfs_extent_item *item;
716 struct btrfs_extent_item_v0 *ei0;
717 struct btrfs_extent_ref_v0 *ref0;
718 struct btrfs_tree_block_info *bi;
719 struct extent_buffer *leaf;
720 struct btrfs_key key;
721 struct btrfs_key found_key;
722 u32 new_size = sizeof(*item);
723 u64 refs;
724 int ret;
725
726 leaf = path->nodes[0];
727 BUG_ON(btrfs_item_size_nr(leaf, path->slots[0]) != sizeof(*ei0));
728
729 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
730 ei0 = btrfs_item_ptr(leaf, path->slots[0],
731 struct btrfs_extent_item_v0);
732 refs = btrfs_extent_refs_v0(leaf, ei0);
733
734 if (owner == (u64)-1) {
735 while (1) {
736 if (path->slots[0] >= btrfs_header_nritems(leaf)) {
737 ret = btrfs_next_leaf(root, path);
738 if (ret < 0)
739 return ret;
740 BUG_ON(ret > 0);
741 leaf = path->nodes[0];
742 }
743 btrfs_item_key_to_cpu(leaf, &found_key,
744 path->slots[0]);
745 BUG_ON(key.objectid != found_key.objectid);
746 if (found_key.type != BTRFS_EXTENT_REF_V0_KEY) {
747 path->slots[0]++;
748 continue;
749 }
750 ref0 = btrfs_item_ptr(leaf, path->slots[0],
751 struct btrfs_extent_ref_v0);
752 owner = btrfs_ref_objectid_v0(leaf, ref0);
753 break;
754 }
755 }
756 btrfs_release_path(root, path);
757
758 if (owner < BTRFS_FIRST_FREE_OBJECTID)
759 new_size += sizeof(*bi);
760
761 new_size -= sizeof(*ei0);
762 ret = btrfs_search_slot(trans, root, &key, path,
763 new_size + extra_size, 1);
764 if (ret < 0)
765 return ret;
766 BUG_ON(ret);
767
768 ret = btrfs_extend_item(trans, root, path, new_size);
769 BUG_ON(ret);
770
771 leaf = path->nodes[0];
772 item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
773 btrfs_set_extent_refs(leaf, item, refs);
774 /* FIXME: get real generation */
775 btrfs_set_extent_generation(leaf, item, 0);
776 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
777 btrfs_set_extent_flags(leaf, item,
778 BTRFS_EXTENT_FLAG_TREE_BLOCK |
779 BTRFS_BLOCK_FLAG_FULL_BACKREF);
780 bi = (struct btrfs_tree_block_info *)(item + 1);
781 /* FIXME: get first key of the block */
782 memset_extent_buffer(leaf, 0, (unsigned long)bi, sizeof(*bi));
783 btrfs_set_tree_block_level(leaf, bi, (int)owner);
784 } else {
785 btrfs_set_extent_flags(leaf, item, BTRFS_EXTENT_FLAG_DATA);
786 }
787 btrfs_mark_buffer_dirty(leaf);
788 return 0;
789 }
790 #endif
791
792 static u64 hash_extent_data_ref(u64 root_objectid, u64 owner, u64 offset)
793 {
794 u32 high_crc = ~(u32)0;
795 u32 low_crc = ~(u32)0;
796 __le64 lenum;
797
798 lenum = cpu_to_le64(root_objectid);
799 high_crc = crc32c(high_crc, &lenum, sizeof(lenum));
800 lenum = cpu_to_le64(owner);
801 low_crc = crc32c(low_crc, &lenum, sizeof(lenum));
802 lenum = cpu_to_le64(offset);
803 low_crc = crc32c(low_crc, &lenum, sizeof(lenum));
804
805 return ((u64)high_crc << 31) ^ (u64)low_crc;
806 }
807
808 static u64 hash_extent_data_ref_item(struct extent_buffer *leaf,
809 struct btrfs_extent_data_ref *ref)
810 {
811 return hash_extent_data_ref(btrfs_extent_data_ref_root(leaf, ref),
812 btrfs_extent_data_ref_objectid(leaf, ref),
813 btrfs_extent_data_ref_offset(leaf, ref));
814 }
815
816 static int match_extent_data_ref(struct extent_buffer *leaf,
817 struct btrfs_extent_data_ref *ref,
818 u64 root_objectid, u64 owner, u64 offset)
819 {
820 if (btrfs_extent_data_ref_root(leaf, ref) != root_objectid ||
821 btrfs_extent_data_ref_objectid(leaf, ref) != owner ||
822 btrfs_extent_data_ref_offset(leaf, ref) != offset)
823 return 0;
824 return 1;
825 }
826
827 static noinline int lookup_extent_data_ref(struct btrfs_trans_handle *trans,
828 struct btrfs_root *root,
829 struct btrfs_path *path,
830 u64 bytenr, u64 parent,
831 u64 root_objectid,
832 u64 owner, u64 offset)
833 {
834 struct btrfs_key key;
835 struct btrfs_extent_data_ref *ref;
836 struct extent_buffer *leaf;
837 u32 nritems;
838 int ret;
839 int recow;
840 int err = -ENOENT;
841
842 key.objectid = bytenr;
843 if (parent) {
844 key.type = BTRFS_SHARED_DATA_REF_KEY;
845 key.offset = parent;
846 } else {
847 key.type = BTRFS_EXTENT_DATA_REF_KEY;
848 key.offset = hash_extent_data_ref(root_objectid,
849 owner, offset);
850 }
851 again:
852 recow = 0;
853 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
854 if (ret < 0) {
855 err = ret;
856 goto fail;
857 }
858
859 if (parent) {
860 if (!ret)
861 return 0;
862 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
863 key.type = BTRFS_EXTENT_REF_V0_KEY;
864 btrfs_release_path(root, path);
865 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
866 if (ret < 0) {
867 err = ret;
868 goto fail;
869 }
870 if (!ret)
871 return 0;
872 #endif
873 goto fail;
874 }
875
876 leaf = path->nodes[0];
877 nritems = btrfs_header_nritems(leaf);
878 while (1) {
879 if (path->slots[0] >= nritems) {
880 ret = btrfs_next_leaf(root, path);
881 if (ret < 0)
882 err = ret;
883 if (ret)
884 goto fail;
885
886 leaf = path->nodes[0];
887 nritems = btrfs_header_nritems(leaf);
888 recow = 1;
889 }
890
891 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
892 if (key.objectid != bytenr ||
893 key.type != BTRFS_EXTENT_DATA_REF_KEY)
894 goto fail;
895
896 ref = btrfs_item_ptr(leaf, path->slots[0],
897 struct btrfs_extent_data_ref);
898
899 if (match_extent_data_ref(leaf, ref, root_objectid,
900 owner, offset)) {
901 if (recow) {
902 btrfs_release_path(root, path);
903 goto again;
904 }
905 err = 0;
906 break;
907 }
908 path->slots[0]++;
909 }
910 fail:
911 return err;
912 }
913
914 static noinline int insert_extent_data_ref(struct btrfs_trans_handle *trans,
915 struct btrfs_root *root,
916 struct btrfs_path *path,
917 u64 bytenr, u64 parent,
918 u64 root_objectid, u64 owner,
919 u64 offset, int refs_to_add)
920 {
921 struct btrfs_key key;
922 struct extent_buffer *leaf;
923 u32 size;
924 u32 num_refs;
925 int ret;
926
927 key.objectid = bytenr;
928 if (parent) {
929 key.type = BTRFS_SHARED_DATA_REF_KEY;
930 key.offset = parent;
931 size = sizeof(struct btrfs_shared_data_ref);
932 } else {
933 key.type = BTRFS_EXTENT_DATA_REF_KEY;
934 key.offset = hash_extent_data_ref(root_objectid,
935 owner, offset);
936 size = sizeof(struct btrfs_extent_data_ref);
937 }
938
939 ret = btrfs_insert_empty_item(trans, root, path, &key, size);
940 if (ret && ret != -EEXIST)
941 goto fail;
942
943 leaf = path->nodes[0];
944 if (parent) {
945 struct btrfs_shared_data_ref *ref;
946 ref = btrfs_item_ptr(leaf, path->slots[0],
947 struct btrfs_shared_data_ref);
948 if (ret == 0) {
949 btrfs_set_shared_data_ref_count(leaf, ref, refs_to_add);
950 } else {
951 num_refs = btrfs_shared_data_ref_count(leaf, ref);
952 num_refs += refs_to_add;
953 btrfs_set_shared_data_ref_count(leaf, ref, num_refs);
954 }
955 } else {
956 struct btrfs_extent_data_ref *ref;
957 while (ret == -EEXIST) {
958 ref = btrfs_item_ptr(leaf, path->slots[0],
959 struct btrfs_extent_data_ref);
960 if (match_extent_data_ref(leaf, ref, root_objectid,
961 owner, offset))
962 break;
963 btrfs_release_path(root, path);
964 key.offset++;
965 ret = btrfs_insert_empty_item(trans, root, path, &key,
966 size);
967 if (ret && ret != -EEXIST)
968 goto fail;
969
970 leaf = path->nodes[0];
971 }
972 ref = btrfs_item_ptr(leaf, path->slots[0],
973 struct btrfs_extent_data_ref);
974 if (ret == 0) {
975 btrfs_set_extent_data_ref_root(leaf, ref,
976 root_objectid);
977 btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
978 btrfs_set_extent_data_ref_offset(leaf, ref, offset);
979 btrfs_set_extent_data_ref_count(leaf, ref, refs_to_add);
980 } else {
981 num_refs = btrfs_extent_data_ref_count(leaf, ref);
982 num_refs += refs_to_add;
983 btrfs_set_extent_data_ref_count(leaf, ref, num_refs);
984 }
985 }
986 btrfs_mark_buffer_dirty(leaf);
987 ret = 0;
988 fail:
989 btrfs_release_path(root, path);
990 return ret;
991 }
992
993 static noinline int remove_extent_data_ref(struct btrfs_trans_handle *trans,
994 struct btrfs_root *root,
995 struct btrfs_path *path,
996 int refs_to_drop)
997 {
998 struct btrfs_key key;
999 struct btrfs_extent_data_ref *ref1 = NULL;
1000 struct btrfs_shared_data_ref *ref2 = NULL;
1001 struct extent_buffer *leaf;
1002 u32 num_refs = 0;
1003 int ret = 0;
1004
1005 leaf = path->nodes[0];
1006 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
1007
1008 if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
1009 ref1 = btrfs_item_ptr(leaf, path->slots[0],
1010 struct btrfs_extent_data_ref);
1011 num_refs = btrfs_extent_data_ref_count(leaf, ref1);
1012 } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
1013 ref2 = btrfs_item_ptr(leaf, path->slots[0],
1014 struct btrfs_shared_data_ref);
1015 num_refs = btrfs_shared_data_ref_count(leaf, ref2);
1016 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1017 } else if (key.type == BTRFS_EXTENT_REF_V0_KEY) {
1018 struct btrfs_extent_ref_v0 *ref0;
1019 ref0 = btrfs_item_ptr(leaf, path->slots[0],
1020 struct btrfs_extent_ref_v0);
1021 num_refs = btrfs_ref_count_v0(leaf, ref0);
1022 #endif
1023 } else {
1024 BUG();
1025 }
1026
1027 BUG_ON(num_refs < refs_to_drop);
1028 num_refs -= refs_to_drop;
1029
1030 if (num_refs == 0) {
1031 ret = btrfs_del_item(trans, root, path);
1032 } else {
1033 if (key.type == BTRFS_EXTENT_DATA_REF_KEY)
1034 btrfs_set_extent_data_ref_count(leaf, ref1, num_refs);
1035 else if (key.type == BTRFS_SHARED_DATA_REF_KEY)
1036 btrfs_set_shared_data_ref_count(leaf, ref2, num_refs);
1037 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1038 else {
1039 struct btrfs_extent_ref_v0 *ref0;
1040 ref0 = btrfs_item_ptr(leaf, path->slots[0],
1041 struct btrfs_extent_ref_v0);
1042 btrfs_set_ref_count_v0(leaf, ref0, num_refs);
1043 }
1044 #endif
1045 btrfs_mark_buffer_dirty(leaf);
1046 }
1047 return ret;
1048 }
1049
1050 static noinline u32 extent_data_ref_count(struct btrfs_root *root,
1051 struct btrfs_path *path,
1052 struct btrfs_extent_inline_ref *iref)
1053 {
1054 struct btrfs_key key;
1055 struct extent_buffer *leaf;
1056 struct btrfs_extent_data_ref *ref1;
1057 struct btrfs_shared_data_ref *ref2;
1058 u32 num_refs = 0;
1059
1060 leaf = path->nodes[0];
1061 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
1062 if (iref) {
1063 if (btrfs_extent_inline_ref_type(leaf, iref) ==
1064 BTRFS_EXTENT_DATA_REF_KEY) {
1065 ref1 = (struct btrfs_extent_data_ref *)(&iref->offset);
1066 num_refs = btrfs_extent_data_ref_count(leaf, ref1);
1067 } else {
1068 ref2 = (struct btrfs_shared_data_ref *)(iref + 1);
1069 num_refs = btrfs_shared_data_ref_count(leaf, ref2);
1070 }
1071 } else if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
1072 ref1 = btrfs_item_ptr(leaf, path->slots[0],
1073 struct btrfs_extent_data_ref);
1074 num_refs = btrfs_extent_data_ref_count(leaf, ref1);
1075 } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
1076 ref2 = btrfs_item_ptr(leaf, path->slots[0],
1077 struct btrfs_shared_data_ref);
1078 num_refs = btrfs_shared_data_ref_count(leaf, ref2);
1079 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1080 } else if (key.type == BTRFS_EXTENT_REF_V0_KEY) {
1081 struct btrfs_extent_ref_v0 *ref0;
1082 ref0 = btrfs_item_ptr(leaf, path->slots[0],
1083 struct btrfs_extent_ref_v0);
1084 num_refs = btrfs_ref_count_v0(leaf, ref0);
1085 #endif
1086 } else {
1087 WARN_ON(1);
1088 }
1089 return num_refs;
1090 }
1091
1092 static noinline int lookup_tree_block_ref(struct btrfs_trans_handle *trans,
1093 struct btrfs_root *root,
1094 struct btrfs_path *path,
1095 u64 bytenr, u64 parent,
1096 u64 root_objectid)
1097 {
1098 struct btrfs_key key;
1099 int ret;
1100
1101 key.objectid = bytenr;
1102 if (parent) {
1103 key.type = BTRFS_SHARED_BLOCK_REF_KEY;
1104 key.offset = parent;
1105 } else {
1106 key.type = BTRFS_TREE_BLOCK_REF_KEY;
1107 key.offset = root_objectid;
1108 }
1109
1110 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1111 if (ret > 0)
1112 ret = -ENOENT;
1113 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1114 if (ret == -ENOENT && parent) {
1115 btrfs_release_path(root, path);
1116 key.type = BTRFS_EXTENT_REF_V0_KEY;
1117 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1118 if (ret > 0)
1119 ret = -ENOENT;
1120 }
1121 #endif
1122 return ret;
1123 }
1124
1125 static noinline int insert_tree_block_ref(struct btrfs_trans_handle *trans,
1126 struct btrfs_root *root,
1127 struct btrfs_path *path,
1128 u64 bytenr, u64 parent,
1129 u64 root_objectid)
1130 {
1131 struct btrfs_key key;
1132 int ret;
1133
1134 key.objectid = bytenr;
1135 if (parent) {
1136 key.type = BTRFS_SHARED_BLOCK_REF_KEY;
1137 key.offset = parent;
1138 } else {
1139 key.type = BTRFS_TREE_BLOCK_REF_KEY;
1140 key.offset = root_objectid;
1141 }
1142
1143 ret = btrfs_insert_empty_item(trans, root, path, &key, 0);
1144 btrfs_release_path(root, path);
1145 return ret;
1146 }
1147
1148 static inline int extent_ref_type(u64 parent, u64 owner)
1149 {
1150 int type;
1151 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1152 if (parent > 0)
1153 type = BTRFS_SHARED_BLOCK_REF_KEY;
1154 else
1155 type = BTRFS_TREE_BLOCK_REF_KEY;
1156 } else {
1157 if (parent > 0)
1158 type = BTRFS_SHARED_DATA_REF_KEY;
1159 else
1160 type = BTRFS_EXTENT_DATA_REF_KEY;
1161 }
1162 return type;
1163 }
1164
1165 static int find_next_key(struct btrfs_path *path, int level,
1166 struct btrfs_key *key)
1167
1168 {
1169 for (; level < BTRFS_MAX_LEVEL; level++) {
1170 if (!path->nodes[level])
1171 break;
1172 if (path->slots[level] + 1 >=
1173 btrfs_header_nritems(path->nodes[level]))
1174 continue;
1175 if (level == 0)
1176 btrfs_item_key_to_cpu(path->nodes[level], key,
1177 path->slots[level] + 1);
1178 else
1179 btrfs_node_key_to_cpu(path->nodes[level], key,
1180 path->slots[level] + 1);
1181 return 0;
1182 }
1183 return 1;
1184 }
1185
1186 /*
1187 * look for inline back ref. if back ref is found, *ref_ret is set
1188 * to the address of inline back ref, and 0 is returned.
1189 *
1190 * if back ref isn't found, *ref_ret is set to the address where it
1191 * should be inserted, and -ENOENT is returned.
1192 *
1193 * if insert is true and there are too many inline back refs, the path
1194 * points to the extent item, and -EAGAIN is returned.
1195 *
1196 * NOTE: inline back refs are ordered in the same way that back ref
1197 * items in the tree are ordered.
1198 */
1199 static noinline_for_stack
1200 int lookup_inline_extent_backref(struct btrfs_trans_handle *trans,
1201 struct btrfs_root *root,
1202 struct btrfs_path *path,
1203 struct btrfs_extent_inline_ref **ref_ret,
1204 u64 bytenr, u64 num_bytes,
1205 u64 parent, u64 root_objectid,
1206 u64 owner, u64 offset, int insert)
1207 {
1208 struct btrfs_key key;
1209 struct extent_buffer *leaf;
1210 struct btrfs_extent_item *ei;
1211 struct btrfs_extent_inline_ref *iref;
1212 u64 flags;
1213 u64 item_size;
1214 unsigned long ptr;
1215 unsigned long end;
1216 int extra_size;
1217 int type;
1218 int want;
1219 int ret;
1220 int err = 0;
1221
1222 key.objectid = bytenr;
1223 key.type = BTRFS_EXTENT_ITEM_KEY;
1224 key.offset = num_bytes;
1225
1226 want = extent_ref_type(parent, owner);
1227 if (insert) {
1228 extra_size = btrfs_extent_inline_ref_size(want);
1229 path->keep_locks = 1;
1230 } else
1231 extra_size = -1;
1232 ret = btrfs_search_slot(trans, root, &key, path, extra_size, 1);
1233 if (ret < 0) {
1234 err = ret;
1235 goto out;
1236 }
1237 BUG_ON(ret);
1238
1239 leaf = path->nodes[0];
1240 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1241 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1242 if (item_size < sizeof(*ei)) {
1243 if (!insert) {
1244 err = -ENOENT;
1245 goto out;
1246 }
1247 ret = convert_extent_item_v0(trans, root, path, owner,
1248 extra_size);
1249 if (ret < 0) {
1250 err = ret;
1251 goto out;
1252 }
1253 leaf = path->nodes[0];
1254 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1255 }
1256 #endif
1257 BUG_ON(item_size < sizeof(*ei));
1258
1259 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1260 flags = btrfs_extent_flags(leaf, ei);
1261
1262 ptr = (unsigned long)(ei + 1);
1263 end = (unsigned long)ei + item_size;
1264
1265 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
1266 ptr += sizeof(struct btrfs_tree_block_info);
1267 BUG_ON(ptr > end);
1268 } else {
1269 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
1270 }
1271
1272 err = -ENOENT;
1273 while (1) {
1274 if (ptr >= end) {
1275 WARN_ON(ptr > end);
1276 break;
1277 }
1278 iref = (struct btrfs_extent_inline_ref *)ptr;
1279 type = btrfs_extent_inline_ref_type(leaf, iref);
1280 if (want < type)
1281 break;
1282 if (want > type) {
1283 ptr += btrfs_extent_inline_ref_size(type);
1284 continue;
1285 }
1286
1287 if (type == BTRFS_EXTENT_DATA_REF_KEY) {
1288 struct btrfs_extent_data_ref *dref;
1289 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1290 if (match_extent_data_ref(leaf, dref, root_objectid,
1291 owner, offset)) {
1292 err = 0;
1293 break;
1294 }
1295 if (hash_extent_data_ref_item(leaf, dref) <
1296 hash_extent_data_ref(root_objectid, owner, offset))
1297 break;
1298 } else {
1299 u64 ref_offset;
1300 ref_offset = btrfs_extent_inline_ref_offset(leaf, iref);
1301 if (parent > 0) {
1302 if (parent == ref_offset) {
1303 err = 0;
1304 break;
1305 }
1306 if (ref_offset < parent)
1307 break;
1308 } else {
1309 if (root_objectid == ref_offset) {
1310 err = 0;
1311 break;
1312 }
1313 if (ref_offset < root_objectid)
1314 break;
1315 }
1316 }
1317 ptr += btrfs_extent_inline_ref_size(type);
1318 }
1319 if (err == -ENOENT && insert) {
1320 if (item_size + extra_size >=
1321 BTRFS_MAX_EXTENT_ITEM_SIZE(root)) {
1322 err = -EAGAIN;
1323 goto out;
1324 }
1325 /*
1326 * To add new inline back ref, we have to make sure
1327 * there is no corresponding back ref item.
1328 * For simplicity, we just do not add new inline back
1329 * ref if there is any kind of item for this block
1330 */
1331 if (find_next_key(path, 0, &key) == 0 &&
1332 key.objectid == bytenr &&
1333 key.type < BTRFS_BLOCK_GROUP_ITEM_KEY) {
1334 err = -EAGAIN;
1335 goto out;
1336 }
1337 }
1338 *ref_ret = (struct btrfs_extent_inline_ref *)ptr;
1339 out:
1340 if (insert) {
1341 path->keep_locks = 0;
1342 btrfs_unlock_up_safe(path, 1);
1343 }
1344 return err;
1345 }
1346
1347 /*
1348 * helper to add new inline back ref
1349 */
1350 static noinline_for_stack
1351 int setup_inline_extent_backref(struct btrfs_trans_handle *trans,
1352 struct btrfs_root *root,
1353 struct btrfs_path *path,
1354 struct btrfs_extent_inline_ref *iref,
1355 u64 parent, u64 root_objectid,
1356 u64 owner, u64 offset, int refs_to_add,
1357 struct btrfs_delayed_extent_op *extent_op)
1358 {
1359 struct extent_buffer *leaf;
1360 struct btrfs_extent_item *ei;
1361 unsigned long ptr;
1362 unsigned long end;
1363 unsigned long item_offset;
1364 u64 refs;
1365 int size;
1366 int type;
1367 int ret;
1368
1369 leaf = path->nodes[0];
1370 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1371 item_offset = (unsigned long)iref - (unsigned long)ei;
1372
1373 type = extent_ref_type(parent, owner);
1374 size = btrfs_extent_inline_ref_size(type);
1375
1376 ret = btrfs_extend_item(trans, root, path, size);
1377 BUG_ON(ret);
1378
1379 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1380 refs = btrfs_extent_refs(leaf, ei);
1381 refs += refs_to_add;
1382 btrfs_set_extent_refs(leaf, ei, refs);
1383 if (extent_op)
1384 __run_delayed_extent_op(extent_op, leaf, ei);
1385
1386 ptr = (unsigned long)ei + item_offset;
1387 end = (unsigned long)ei + btrfs_item_size_nr(leaf, path->slots[0]);
1388 if (ptr < end - size)
1389 memmove_extent_buffer(leaf, ptr + size, ptr,
1390 end - size - ptr);
1391
1392 iref = (struct btrfs_extent_inline_ref *)ptr;
1393 btrfs_set_extent_inline_ref_type(leaf, iref, type);
1394 if (type == BTRFS_EXTENT_DATA_REF_KEY) {
1395 struct btrfs_extent_data_ref *dref;
1396 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1397 btrfs_set_extent_data_ref_root(leaf, dref, root_objectid);
1398 btrfs_set_extent_data_ref_objectid(leaf, dref, owner);
1399 btrfs_set_extent_data_ref_offset(leaf, dref, offset);
1400 btrfs_set_extent_data_ref_count(leaf, dref, refs_to_add);
1401 } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
1402 struct btrfs_shared_data_ref *sref;
1403 sref = (struct btrfs_shared_data_ref *)(iref + 1);
1404 btrfs_set_shared_data_ref_count(leaf, sref, refs_to_add);
1405 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
1406 } else if (type == BTRFS_SHARED_BLOCK_REF_KEY) {
1407 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
1408 } else {
1409 btrfs_set_extent_inline_ref_offset(leaf, iref, root_objectid);
1410 }
1411 btrfs_mark_buffer_dirty(leaf);
1412 return 0;
1413 }
1414
1415 static int lookup_extent_backref(struct btrfs_trans_handle *trans,
1416 struct btrfs_root *root,
1417 struct btrfs_path *path,
1418 struct btrfs_extent_inline_ref **ref_ret,
1419 u64 bytenr, u64 num_bytes, u64 parent,
1420 u64 root_objectid, u64 owner, u64 offset)
1421 {
1422 int ret;
1423
1424 ret = lookup_inline_extent_backref(trans, root, path, ref_ret,
1425 bytenr, num_bytes, parent,
1426 root_objectid, owner, offset, 0);
1427 if (ret != -ENOENT)
1428 return ret;
1429
1430 btrfs_release_path(root, path);
1431 *ref_ret = NULL;
1432
1433 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1434 ret = lookup_tree_block_ref(trans, root, path, bytenr, parent,
1435 root_objectid);
1436 } else {
1437 ret = lookup_extent_data_ref(trans, root, path, bytenr, parent,
1438 root_objectid, owner, offset);
1439 }
1440 return ret;
1441 }
1442
1443 /*
1444 * helper to update/remove inline back ref
1445 */
1446 static noinline_for_stack
1447 int update_inline_extent_backref(struct btrfs_trans_handle *trans,
1448 struct btrfs_root *root,
1449 struct btrfs_path *path,
1450 struct btrfs_extent_inline_ref *iref,
1451 int refs_to_mod,
1452 struct btrfs_delayed_extent_op *extent_op)
1453 {
1454 struct extent_buffer *leaf;
1455 struct btrfs_extent_item *ei;
1456 struct btrfs_extent_data_ref *dref = NULL;
1457 struct btrfs_shared_data_ref *sref = NULL;
1458 unsigned long ptr;
1459 unsigned long end;
1460 u32 item_size;
1461 int size;
1462 int type;
1463 int ret;
1464 u64 refs;
1465
1466 leaf = path->nodes[0];
1467 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1468 refs = btrfs_extent_refs(leaf, ei);
1469 WARN_ON(refs_to_mod < 0 && refs + refs_to_mod <= 0);
1470 refs += refs_to_mod;
1471 btrfs_set_extent_refs(leaf, ei, refs);
1472 if (extent_op)
1473 __run_delayed_extent_op(extent_op, leaf, ei);
1474
1475 type = btrfs_extent_inline_ref_type(leaf, iref);
1476
1477 if (type == BTRFS_EXTENT_DATA_REF_KEY) {
1478 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1479 refs = btrfs_extent_data_ref_count(leaf, dref);
1480 } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
1481 sref = (struct btrfs_shared_data_ref *)(iref + 1);
1482 refs = btrfs_shared_data_ref_count(leaf, sref);
1483 } else {
1484 refs = 1;
1485 BUG_ON(refs_to_mod != -1);
1486 }
1487
1488 BUG_ON(refs_to_mod < 0 && refs < -refs_to_mod);
1489 refs += refs_to_mod;
1490
1491 if (refs > 0) {
1492 if (type == BTRFS_EXTENT_DATA_REF_KEY)
1493 btrfs_set_extent_data_ref_count(leaf, dref, refs);
1494 else
1495 btrfs_set_shared_data_ref_count(leaf, sref, refs);
1496 } else {
1497 size = btrfs_extent_inline_ref_size(type);
1498 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1499 ptr = (unsigned long)iref;
1500 end = (unsigned long)ei + item_size;
1501 if (ptr + size < end)
1502 memmove_extent_buffer(leaf, ptr, ptr + size,
1503 end - ptr - size);
1504 item_size -= size;
1505 ret = btrfs_truncate_item(trans, root, path, item_size, 1);
1506 BUG_ON(ret);
1507 }
1508 btrfs_mark_buffer_dirty(leaf);
1509 return 0;
1510 }
1511
1512 static noinline_for_stack
1513 int insert_inline_extent_backref(struct btrfs_trans_handle *trans,
1514 struct btrfs_root *root,
1515 struct btrfs_path *path,
1516 u64 bytenr, u64 num_bytes, u64 parent,
1517 u64 root_objectid, u64 owner,
1518 u64 offset, int refs_to_add,
1519 struct btrfs_delayed_extent_op *extent_op)
1520 {
1521 struct btrfs_extent_inline_ref *iref;
1522 int ret;
1523
1524 ret = lookup_inline_extent_backref(trans, root, path, &iref,
1525 bytenr, num_bytes, parent,
1526 root_objectid, owner, offset, 1);
1527 if (ret == 0) {
1528 BUG_ON(owner < BTRFS_FIRST_FREE_OBJECTID);
1529 ret = update_inline_extent_backref(trans, root, path, iref,
1530 refs_to_add, extent_op);
1531 } else if (ret == -ENOENT) {
1532 ret = setup_inline_extent_backref(trans, root, path, iref,
1533 parent, root_objectid,
1534 owner, offset, refs_to_add,
1535 extent_op);
1536 }
1537 return ret;
1538 }
1539
1540 static int insert_extent_backref(struct btrfs_trans_handle *trans,
1541 struct btrfs_root *root,
1542 struct btrfs_path *path,
1543 u64 bytenr, u64 parent, u64 root_objectid,
1544 u64 owner, u64 offset, int refs_to_add)
1545 {
1546 int ret;
1547 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1548 BUG_ON(refs_to_add != 1);
1549 ret = insert_tree_block_ref(trans, root, path, bytenr,
1550 parent, root_objectid);
1551 } else {
1552 ret = insert_extent_data_ref(trans, root, path, bytenr,
1553 parent, root_objectid,
1554 owner, offset, refs_to_add);
1555 }
1556 return ret;
1557 }
1558
1559 static int remove_extent_backref(struct btrfs_trans_handle *trans,
1560 struct btrfs_root *root,
1561 struct btrfs_path *path,
1562 struct btrfs_extent_inline_ref *iref,
1563 int refs_to_drop, int is_data)
1564 {
1565 int ret;
1566
1567 BUG_ON(!is_data && refs_to_drop != 1);
1568 if (iref) {
1569 ret = update_inline_extent_backref(trans, root, path, iref,
1570 -refs_to_drop, NULL);
1571 } else if (is_data) {
1572 ret = remove_extent_data_ref(trans, root, path, refs_to_drop);
1573 } else {
1574 ret = btrfs_del_item(trans, root, path);
1575 }
1576 return ret;
1577 }
1578
1579 static void btrfs_issue_discard(struct block_device *bdev,
1580 u64 start, u64 len)
1581 {
1582 blkdev_issue_discard(bdev, start >> 9, len >> 9, GFP_KERNEL,
1583 DISCARD_FL_BARRIER);
1584 }
1585
1586 static int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr,
1587 u64 num_bytes)
1588 {
1589 int ret;
1590 u64 map_length = num_bytes;
1591 struct btrfs_multi_bio *multi = NULL;
1592
1593 if (!btrfs_test_opt(root, DISCARD))
1594 return 0;
1595
1596 /* Tell the block device(s) that the sectors can be discarded */
1597 ret = btrfs_map_block(&root->fs_info->mapping_tree, READ,
1598 bytenr, &map_length, &multi, 0);
1599 if (!ret) {
1600 struct btrfs_bio_stripe *stripe = multi->stripes;
1601 int i;
1602
1603 if (map_length > num_bytes)
1604 map_length = num_bytes;
1605
1606 for (i = 0; i < multi->num_stripes; i++, stripe++) {
1607 btrfs_issue_discard(stripe->dev->bdev,
1608 stripe->physical,
1609 map_length);
1610 }
1611 kfree(multi);
1612 }
1613
1614 return ret;
1615 }
1616
1617 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
1618 struct btrfs_root *root,
1619 u64 bytenr, u64 num_bytes, u64 parent,
1620 u64 root_objectid, u64 owner, u64 offset)
1621 {
1622 int ret;
1623 BUG_ON(owner < BTRFS_FIRST_FREE_OBJECTID &&
1624 root_objectid == BTRFS_TREE_LOG_OBJECTID);
1625
1626 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1627 ret = btrfs_add_delayed_tree_ref(trans, bytenr, num_bytes,
1628 parent, root_objectid, (int)owner,
1629 BTRFS_ADD_DELAYED_REF, NULL);
1630 } else {
1631 ret = btrfs_add_delayed_data_ref(trans, bytenr, num_bytes,
1632 parent, root_objectid, owner, offset,
1633 BTRFS_ADD_DELAYED_REF, NULL);
1634 }
1635 return ret;
1636 }
1637
1638 static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
1639 struct btrfs_root *root,
1640 u64 bytenr, u64 num_bytes,
1641 u64 parent, u64 root_objectid,
1642 u64 owner, u64 offset, int refs_to_add,
1643 struct btrfs_delayed_extent_op *extent_op)
1644 {
1645 struct btrfs_path *path;
1646 struct extent_buffer *leaf;
1647 struct btrfs_extent_item *item;
1648 u64 refs;
1649 int ret;
1650 int err = 0;
1651
1652 path = btrfs_alloc_path();
1653 if (!path)
1654 return -ENOMEM;
1655
1656 path->reada = 1;
1657 path->leave_spinning = 1;
1658 /* this will setup the path even if it fails to insert the back ref */
1659 ret = insert_inline_extent_backref(trans, root->fs_info->extent_root,
1660 path, bytenr, num_bytes, parent,
1661 root_objectid, owner, offset,
1662 refs_to_add, extent_op);
1663 if (ret == 0)
1664 goto out;
1665
1666 if (ret != -EAGAIN) {
1667 err = ret;
1668 goto out;
1669 }
1670
1671 leaf = path->nodes[0];
1672 item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1673 refs = btrfs_extent_refs(leaf, item);
1674 btrfs_set_extent_refs(leaf, item, refs + refs_to_add);
1675 if (extent_op)
1676 __run_delayed_extent_op(extent_op, leaf, item);
1677
1678 btrfs_mark_buffer_dirty(leaf);
1679 btrfs_release_path(root->fs_info->extent_root, path);
1680
1681 path->reada = 1;
1682 path->leave_spinning = 1;
1683
1684 /* now insert the actual backref */
1685 ret = insert_extent_backref(trans, root->fs_info->extent_root,
1686 path, bytenr, parent, root_objectid,
1687 owner, offset, refs_to_add);
1688 BUG_ON(ret);
1689 out:
1690 btrfs_free_path(path);
1691 return err;
1692 }
1693
1694 static int run_delayed_data_ref(struct btrfs_trans_handle *trans,
1695 struct btrfs_root *root,
1696 struct btrfs_delayed_ref_node *node,
1697 struct btrfs_delayed_extent_op *extent_op,
1698 int insert_reserved)
1699 {
1700 int ret = 0;
1701 struct btrfs_delayed_data_ref *ref;
1702 struct btrfs_key ins;
1703 u64 parent = 0;
1704 u64 ref_root = 0;
1705 u64 flags = 0;
1706
1707 ins.objectid = node->bytenr;
1708 ins.offset = node->num_bytes;
1709 ins.type = BTRFS_EXTENT_ITEM_KEY;
1710
1711 ref = btrfs_delayed_node_to_data_ref(node);
1712 if (node->type == BTRFS_SHARED_DATA_REF_KEY)
1713 parent = ref->parent;
1714 else
1715 ref_root = ref->root;
1716
1717 if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
1718 if (extent_op) {
1719 BUG_ON(extent_op->update_key);
1720 flags |= extent_op->flags_to_set;
1721 }
1722 ret = alloc_reserved_file_extent(trans, root,
1723 parent, ref_root, flags,
1724 ref->objectid, ref->offset,
1725 &ins, node->ref_mod);
1726 } else if (node->action == BTRFS_ADD_DELAYED_REF) {
1727 ret = __btrfs_inc_extent_ref(trans, root, node->bytenr,
1728 node->num_bytes, parent,
1729 ref_root, ref->objectid,
1730 ref->offset, node->ref_mod,
1731 extent_op);
1732 } else if (node->action == BTRFS_DROP_DELAYED_REF) {
1733 ret = __btrfs_free_extent(trans, root, node->bytenr,
1734 node->num_bytes, parent,
1735 ref_root, ref->objectid,
1736 ref->offset, node->ref_mod,
1737 extent_op);
1738 } else {
1739 BUG();
1740 }
1741 return ret;
1742 }
1743
1744 static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
1745 struct extent_buffer *leaf,
1746 struct btrfs_extent_item *ei)
1747 {
1748 u64 flags = btrfs_extent_flags(leaf, ei);
1749 if (extent_op->update_flags) {
1750 flags |= extent_op->flags_to_set;
1751 btrfs_set_extent_flags(leaf, ei, flags);
1752 }
1753
1754 if (extent_op->update_key) {
1755 struct btrfs_tree_block_info *bi;
1756 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK));
1757 bi = (struct btrfs_tree_block_info *)(ei + 1);
1758 btrfs_set_tree_block_key(leaf, bi, &extent_op->key);
1759 }
1760 }
1761
1762 static int run_delayed_extent_op(struct btrfs_trans_handle *trans,
1763 struct btrfs_root *root,
1764 struct btrfs_delayed_ref_node *node,
1765 struct btrfs_delayed_extent_op *extent_op)
1766 {
1767 struct btrfs_key key;
1768 struct btrfs_path *path;
1769 struct btrfs_extent_item *ei;
1770 struct extent_buffer *leaf;
1771 u32 item_size;
1772 int ret;
1773 int err = 0;
1774
1775 path = btrfs_alloc_path();
1776 if (!path)
1777 return -ENOMEM;
1778
1779 key.objectid = node->bytenr;
1780 key.type = BTRFS_EXTENT_ITEM_KEY;
1781 key.offset = node->num_bytes;
1782
1783 path->reada = 1;
1784 path->leave_spinning = 1;
1785 ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key,
1786 path, 0, 1);
1787 if (ret < 0) {
1788 err = ret;
1789 goto out;
1790 }
1791 if (ret > 0) {
1792 err = -EIO;
1793 goto out;
1794 }
1795
1796 leaf = path->nodes[0];
1797 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1798 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1799 if (item_size < sizeof(*ei)) {
1800 ret = convert_extent_item_v0(trans, root->fs_info->extent_root,
1801 path, (u64)-1, 0);
1802 if (ret < 0) {
1803 err = ret;
1804 goto out;
1805 }
1806 leaf = path->nodes[0];
1807 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1808 }
1809 #endif
1810 BUG_ON(item_size < sizeof(*ei));
1811 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1812 __run_delayed_extent_op(extent_op, leaf, ei);
1813
1814 btrfs_mark_buffer_dirty(leaf);
1815 out:
1816 btrfs_free_path(path);
1817 return err;
1818 }
1819
1820 static int run_delayed_tree_ref(struct btrfs_trans_handle *trans,
1821 struct btrfs_root *root,
1822 struct btrfs_delayed_ref_node *node,
1823 struct btrfs_delayed_extent_op *extent_op,
1824 int insert_reserved)
1825 {
1826 int ret = 0;
1827 struct btrfs_delayed_tree_ref *ref;
1828 struct btrfs_key ins;
1829 u64 parent = 0;
1830 u64 ref_root = 0;
1831
1832 ins.objectid = node->bytenr;
1833 ins.offset = node->num_bytes;
1834 ins.type = BTRFS_EXTENT_ITEM_KEY;
1835
1836 ref = btrfs_delayed_node_to_tree_ref(node);
1837 if (node->type == BTRFS_SHARED_BLOCK_REF_KEY)
1838 parent = ref->parent;
1839 else
1840 ref_root = ref->root;
1841
1842 BUG_ON(node->ref_mod != 1);
1843 if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
1844 BUG_ON(!extent_op || !extent_op->update_flags ||
1845 !extent_op->update_key);
1846 ret = alloc_reserved_tree_block(trans, root,
1847 parent, ref_root,
1848 extent_op->flags_to_set,
1849 &extent_op->key,
1850 ref->level, &ins);
1851 } else if (node->action == BTRFS_ADD_DELAYED_REF) {
1852 ret = __btrfs_inc_extent_ref(trans, root, node->bytenr,
1853 node->num_bytes, parent, ref_root,
1854 ref->level, 0, 1, extent_op);
1855 } else if (node->action == BTRFS_DROP_DELAYED_REF) {
1856 ret = __btrfs_free_extent(trans, root, node->bytenr,
1857 node->num_bytes, parent, ref_root,
1858 ref->level, 0, 1, extent_op);
1859 } else {
1860 BUG();
1861 }
1862 return ret;
1863 }
1864
1865
1866 /* helper function to actually process a single delayed ref entry */
1867 static int run_one_delayed_ref(struct btrfs_trans_handle *trans,
1868 struct btrfs_root *root,
1869 struct btrfs_delayed_ref_node *node,
1870 struct btrfs_delayed_extent_op *extent_op,
1871 int insert_reserved)
1872 {
1873 int ret;
1874 if (btrfs_delayed_ref_is_head(node)) {
1875 struct btrfs_delayed_ref_head *head;
1876 /*
1877 * we've hit the end of the chain and we were supposed
1878 * to insert this extent into the tree. But, it got
1879 * deleted before we ever needed to insert it, so all
1880 * we have to do is clean up the accounting
1881 */
1882 BUG_ON(extent_op);
1883 head = btrfs_delayed_node_to_head(node);
1884 if (insert_reserved) {
1885 int mark_free = 0;
1886 struct extent_buffer *must_clean = NULL;
1887
1888 ret = pin_down_bytes(trans, root, NULL,
1889 node->bytenr, node->num_bytes,
1890 head->is_data, 1, &must_clean);
1891 if (ret > 0)
1892 mark_free = 1;
1893
1894 if (must_clean) {
1895 clean_tree_block(NULL, root, must_clean);
1896 btrfs_tree_unlock(must_clean);
1897 free_extent_buffer(must_clean);
1898 }
1899 if (head->is_data) {
1900 ret = btrfs_del_csums(trans, root,
1901 node->bytenr,
1902 node->num_bytes);
1903 BUG_ON(ret);
1904 }
1905 if (mark_free) {
1906 ret = btrfs_free_reserved_extent(root,
1907 node->bytenr,
1908 node->num_bytes);
1909 BUG_ON(ret);
1910 }
1911 }
1912 mutex_unlock(&head->mutex);
1913 return 0;
1914 }
1915
1916 if (node->type == BTRFS_TREE_BLOCK_REF_KEY ||
1917 node->type == BTRFS_SHARED_BLOCK_REF_KEY)
1918 ret = run_delayed_tree_ref(trans, root, node, extent_op,
1919 insert_reserved);
1920 else if (node->type == BTRFS_EXTENT_DATA_REF_KEY ||
1921 node->type == BTRFS_SHARED_DATA_REF_KEY)
1922 ret = run_delayed_data_ref(trans, root, node, extent_op,
1923 insert_reserved);
1924 else
1925 BUG();
1926 return ret;
1927 }
1928
1929 static noinline struct btrfs_delayed_ref_node *
1930 select_delayed_ref(struct btrfs_delayed_ref_head *head)
1931 {
1932 struct rb_node *node;
1933 struct btrfs_delayed_ref_node *ref;
1934 int action = BTRFS_ADD_DELAYED_REF;
1935 again:
1936 /*
1937 * select delayed ref of type BTRFS_ADD_DELAYED_REF first.
1938 * this prevents ref count from going down to zero when
1939 * there still are pending delayed ref.
1940 */
1941 node = rb_prev(&head->node.rb_node);
1942 while (1) {
1943 if (!node)
1944 break;
1945 ref = rb_entry(node, struct btrfs_delayed_ref_node,
1946 rb_node);
1947 if (ref->bytenr != head->node.bytenr)
1948 break;
1949 if (ref->action == action)
1950 return ref;
1951 node = rb_prev(node);
1952 }
1953 if (action == BTRFS_ADD_DELAYED_REF) {
1954 action = BTRFS_DROP_DELAYED_REF;
1955 goto again;
1956 }
1957 return NULL;
1958 }
1959
1960 static noinline int run_clustered_refs(struct btrfs_trans_handle *trans,
1961 struct btrfs_root *root,
1962 struct list_head *cluster)
1963 {
1964 struct btrfs_delayed_ref_root *delayed_refs;
1965 struct btrfs_delayed_ref_node *ref;
1966 struct btrfs_delayed_ref_head *locked_ref = NULL;
1967 struct btrfs_delayed_extent_op *extent_op;
1968 int ret;
1969 int count = 0;
1970 int must_insert_reserved = 0;
1971
1972 delayed_refs = &trans->transaction->delayed_refs;
1973 while (1) {
1974 if (!locked_ref) {
1975 /* pick a new head ref from the cluster list */
1976 if (list_empty(cluster))
1977 break;
1978
1979 locked_ref = list_entry(cluster->next,
1980 struct btrfs_delayed_ref_head, cluster);
1981
1982 /* grab the lock that says we are going to process
1983 * all the refs for this head */
1984 ret = btrfs_delayed_ref_lock(trans, locked_ref);
1985
1986 /*
1987 * we may have dropped the spin lock to get the head
1988 * mutex lock, and that might have given someone else
1989 * time to free the head. If that's true, it has been
1990 * removed from our list and we can move on.
1991 */
1992 if (ret == -EAGAIN) {
1993 locked_ref = NULL;
1994 count++;
1995 continue;
1996 }
1997 }
1998
1999 /*
2000 * record the must insert reserved flag before we
2001 * drop the spin lock.
2002 */
2003 must_insert_reserved = locked_ref->must_insert_reserved;
2004 locked_ref->must_insert_reserved = 0;
2005
2006 extent_op = locked_ref->extent_op;
2007 locked_ref->extent_op = NULL;
2008
2009 /*
2010 * locked_ref is the head node, so we have to go one
2011 * node back for any delayed ref updates
2012 */
2013 ref = select_delayed_ref(locked_ref);
2014 if (!ref) {
2015 /* All delayed refs have been processed, Go ahead
2016 * and send the head node to run_one_delayed_ref,
2017 * so that any accounting fixes can happen
2018 */
2019 ref = &locked_ref->node;
2020
2021 if (extent_op && must_insert_reserved) {
2022 kfree(extent_op);
2023 extent_op = NULL;
2024 }
2025
2026 if (extent_op) {
2027 spin_unlock(&delayed_refs->lock);
2028
2029 ret = run_delayed_extent_op(trans, root,
2030 ref, extent_op);
2031 BUG_ON(ret);
2032 kfree(extent_op);
2033
2034 cond_resched();
2035 spin_lock(&delayed_refs->lock);
2036 continue;
2037 }
2038
2039 list_del_init(&locked_ref->cluster);
2040 locked_ref = NULL;
2041 }
2042
2043 ref->in_tree = 0;
2044 rb_erase(&ref->rb_node, &delayed_refs->root);
2045 delayed_refs->num_entries--;
2046
2047 spin_unlock(&delayed_refs->lock);
2048
2049 ret = run_one_delayed_ref(trans, root, ref, extent_op,
2050 must_insert_reserved);
2051 BUG_ON(ret);
2052
2053 btrfs_put_delayed_ref(ref);
2054 kfree(extent_op);
2055 count++;
2056
2057 cond_resched();
2058 spin_lock(&delayed_refs->lock);
2059 }
2060 return count;
2061 }
2062
2063 /*
2064 * this starts processing the delayed reference count updates and
2065 * extent insertions we have queued up so far. count can be
2066 * 0, which means to process everything in the tree at the start
2067 * of the run (but not newly added entries), or it can be some target
2068 * number you'd like to process.
2069 */
2070 int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
2071 struct btrfs_root *root, unsigned long count)
2072 {
2073 struct rb_node *node;
2074 struct btrfs_delayed_ref_root *delayed_refs;
2075 struct btrfs_delayed_ref_node *ref;
2076 struct list_head cluster;
2077 int ret;
2078 int run_all = count == (unsigned long)-1;
2079 int run_most = 0;
2080
2081 if (root == root->fs_info->extent_root)
2082 root = root->fs_info->tree_root;
2083
2084 delayed_refs = &trans->transaction->delayed_refs;
2085 INIT_LIST_HEAD(&cluster);
2086 again:
2087 spin_lock(&delayed_refs->lock);
2088 if (count == 0) {
2089 count = delayed_refs->num_entries * 2;
2090 run_most = 1;
2091 }
2092 while (1) {
2093 if (!(run_all || run_most) &&
2094 delayed_refs->num_heads_ready < 64)
2095 break;
2096
2097 /*
2098 * go find something we can process in the rbtree. We start at
2099 * the beginning of the tree, and then build a cluster
2100 * of refs to process starting at the first one we are able to
2101 * lock
2102 */
2103 ret = btrfs_find_ref_cluster(trans, &cluster,
2104 delayed_refs->run_delayed_start);
2105 if (ret)
2106 break;
2107
2108 ret = run_clustered_refs(trans, root, &cluster);
2109 BUG_ON(ret < 0);
2110
2111 count -= min_t(unsigned long, ret, count);
2112
2113 if (count == 0)
2114 break;
2115 }
2116
2117 if (run_all) {
2118 node = rb_first(&delayed_refs->root);
2119 if (!node)
2120 goto out;
2121 count = (unsigned long)-1;
2122
2123 while (node) {
2124 ref = rb_entry(node, struct btrfs_delayed_ref_node,
2125 rb_node);
2126 if (btrfs_delayed_ref_is_head(ref)) {
2127 struct btrfs_delayed_ref_head *head;
2128
2129 head = btrfs_delayed_node_to_head(ref);
2130 atomic_inc(&ref->refs);
2131
2132 spin_unlock(&delayed_refs->lock);
2133 mutex_lock(&head->mutex);
2134 mutex_unlock(&head->mutex);
2135
2136 btrfs_put_delayed_ref(ref);
2137 cond_resched();
2138 goto again;
2139 }
2140 node = rb_next(node);
2141 }
2142 spin_unlock(&delayed_refs->lock);
2143 schedule_timeout(1);
2144 goto again;
2145 }
2146 out:
2147 spin_unlock(&delayed_refs->lock);
2148 return 0;
2149 }
2150
2151 int btrfs_set_disk_extent_flags(struct btrfs_trans_handle *trans,
2152 struct btrfs_root *root,
2153 u64 bytenr, u64 num_bytes, u64 flags,
2154 int is_data)
2155 {
2156 struct btrfs_delayed_extent_op *extent_op;
2157 int ret;
2158
2159 extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
2160 if (!extent_op)
2161 return -ENOMEM;
2162
2163 extent_op->flags_to_set = flags;
2164 extent_op->update_flags = 1;
2165 extent_op->update_key = 0;
2166 extent_op->is_data = is_data ? 1 : 0;
2167
2168 ret = btrfs_add_delayed_extent_op(trans, bytenr, num_bytes, extent_op);
2169 if (ret)
2170 kfree(extent_op);
2171 return ret;
2172 }
2173
2174 static noinline int check_delayed_ref(struct btrfs_trans_handle *trans,
2175 struct btrfs_root *root,
2176 struct btrfs_path *path,
2177 u64 objectid, u64 offset, u64 bytenr)
2178 {
2179 struct btrfs_delayed_ref_head *head;
2180 struct btrfs_delayed_ref_node *ref;
2181 struct btrfs_delayed_data_ref *data_ref;
2182 struct btrfs_delayed_ref_root *delayed_refs;
2183 struct rb_node *node;
2184 int ret = 0;
2185
2186 ret = -ENOENT;
2187 delayed_refs = &trans->transaction->delayed_refs;
2188 spin_lock(&delayed_refs->lock);
2189 head = btrfs_find_delayed_ref_head(trans, bytenr);
2190 if (!head)
2191 goto out;
2192
2193 if (!mutex_trylock(&head->mutex)) {
2194 atomic_inc(&head->node.refs);
2195 spin_unlock(&delayed_refs->lock);
2196
2197 btrfs_release_path(root->fs_info->extent_root, path);
2198
2199 mutex_lock(&head->mutex);
2200 mutex_unlock(&head->mutex);
2201 btrfs_put_delayed_ref(&head->node);
2202 return -EAGAIN;
2203 }
2204
2205 node = rb_prev(&head->node.rb_node);
2206 if (!node)
2207 goto out_unlock;
2208
2209 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
2210
2211 if (ref->bytenr != bytenr)
2212 goto out_unlock;
2213
2214 ret = 1;
2215 if (ref->type != BTRFS_EXTENT_DATA_REF_KEY)
2216 goto out_unlock;
2217
2218 data_ref = btrfs_delayed_node_to_data_ref(ref);
2219
2220 node = rb_prev(node);
2221 if (node) {
2222 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
2223 if (ref->bytenr == bytenr)
2224 goto out_unlock;
2225 }
2226
2227 if (data_ref->root != root->root_key.objectid ||
2228 data_ref->objectid != objectid || data_ref->offset != offset)
2229 goto out_unlock;
2230
2231 ret = 0;
2232 out_unlock:
2233 mutex_unlock(&head->mutex);
2234 out:
2235 spin_unlock(&delayed_refs->lock);
2236 return ret;
2237 }
2238
2239 static noinline int check_committed_ref(struct btrfs_trans_handle *trans,
2240 struct btrfs_root *root,
2241 struct btrfs_path *path,
2242 u64 objectid, u64 offset, u64 bytenr)
2243 {
2244 struct btrfs_root *extent_root = root->fs_info->extent_root;
2245 struct extent_buffer *leaf;
2246 struct btrfs_extent_data_ref *ref;
2247 struct btrfs_extent_inline_ref *iref;
2248 struct btrfs_extent_item *ei;
2249 struct btrfs_key key;
2250 u32 item_size;
2251 int ret;
2252
2253 key.objectid = bytenr;
2254 key.offset = (u64)-1;
2255 key.type = BTRFS_EXTENT_ITEM_KEY;
2256
2257 ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
2258 if (ret < 0)
2259 goto out;
2260 BUG_ON(ret == 0);
2261
2262 ret = -ENOENT;
2263 if (path->slots[0] == 0)
2264 goto out;
2265
2266 path->slots[0]--;
2267 leaf = path->nodes[0];
2268 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
2269
2270 if (key.objectid != bytenr || key.type != BTRFS_EXTENT_ITEM_KEY)
2271 goto out;
2272
2273 ret = 1;
2274 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
2275 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
2276 if (item_size < sizeof(*ei)) {
2277 WARN_ON(item_size != sizeof(struct btrfs_extent_item_v0));
2278 goto out;
2279 }
2280 #endif
2281 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
2282
2283 if (item_size != sizeof(*ei) +
2284 btrfs_extent_inline_ref_size(BTRFS_EXTENT_DATA_REF_KEY))
2285 goto out;
2286
2287 if (btrfs_extent_generation(leaf, ei) <=
2288 btrfs_root_last_snapshot(&root->root_item))
2289 goto out;
2290
2291 iref = (struct btrfs_extent_inline_ref *)(ei + 1);
2292 if (btrfs_extent_inline_ref_type(leaf, iref) !=
2293 BTRFS_EXTENT_DATA_REF_KEY)
2294 goto out;
2295
2296 ref = (struct btrfs_extent_data_ref *)(&iref->offset);
2297 if (btrfs_extent_refs(leaf, ei) !=
2298 btrfs_extent_data_ref_count(leaf, ref) ||
2299 btrfs_extent_data_ref_root(leaf, ref) !=
2300 root->root_key.objectid ||
2301 btrfs_extent_data_ref_objectid(leaf, ref) != objectid ||
2302 btrfs_extent_data_ref_offset(leaf, ref) != offset)
2303 goto out;
2304
2305 ret = 0;
2306 out:
2307 return ret;
2308 }
2309
2310 int btrfs_cross_ref_exist(struct btrfs_trans_handle *trans,
2311 struct btrfs_root *root,
2312 u64 objectid, u64 offset, u64 bytenr)
2313 {
2314 struct btrfs_path *path;
2315 int ret;
2316 int ret2;
2317
2318 path = btrfs_alloc_path();
2319 if (!path)
2320 return -ENOENT;
2321
2322 do {
2323 ret = check_committed_ref(trans, root, path, objectid,
2324 offset, bytenr);
2325 if (ret && ret != -ENOENT)
2326 goto out;
2327
2328 ret2 = check_delayed_ref(trans, root, path, objectid,
2329 offset, bytenr);
2330 } while (ret2 == -EAGAIN);
2331
2332 if (ret2 && ret2 != -ENOENT) {
2333 ret = ret2;
2334 goto out;
2335 }
2336
2337 if (ret != -ENOENT || ret2 != -ENOENT)
2338 ret = 0;
2339 out:
2340 btrfs_free_path(path);
2341 return ret;
2342 }
2343
2344 #if 0
2345 int btrfs_cache_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2346 struct extent_buffer *buf, u32 nr_extents)
2347 {
2348 struct btrfs_key key;
2349 struct btrfs_file_extent_item *fi;
2350 u64 root_gen;
2351 u32 nritems;
2352 int i;
2353 int level;
2354 int ret = 0;
2355 int shared = 0;
2356
2357 if (!root->ref_cows)
2358 return 0;
2359
2360 if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
2361 shared = 0;
2362 root_gen = root->root_key.offset;
2363 } else {
2364 shared = 1;
2365 root_gen = trans->transid - 1;
2366 }
2367
2368 level = btrfs_header_level(buf);
2369 nritems = btrfs_header_nritems(buf);
2370
2371 if (level == 0) {
2372 struct btrfs_leaf_ref *ref;
2373 struct btrfs_extent_info *info;
2374
2375 ref = btrfs_alloc_leaf_ref(root, nr_extents);
2376 if (!ref) {
2377 ret = -ENOMEM;
2378 goto out;
2379 }
2380
2381 ref->root_gen = root_gen;
2382 ref->bytenr = buf->start;
2383 ref->owner = btrfs_header_owner(buf);
2384 ref->generation = btrfs_header_generation(buf);
2385 ref->nritems = nr_extents;
2386 info = ref->extents;
2387
2388 for (i = 0; nr_extents > 0 && i < nritems; i++) {
2389 u64 disk_bytenr;
2390 btrfs_item_key_to_cpu(buf, &key, i);
2391 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
2392 continue;
2393 fi = btrfs_item_ptr(buf, i,
2394 struct btrfs_file_extent_item);
2395 if (btrfs_file_extent_type(buf, fi) ==
2396 BTRFS_FILE_EXTENT_INLINE)
2397 continue;
2398 disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
2399 if (disk_bytenr == 0)
2400 continue;
2401
2402 info->bytenr = disk_bytenr;
2403 info->num_bytes =
2404 btrfs_file_extent_disk_num_bytes(buf, fi);
2405 info->objectid = key.objectid;
2406 info->offset = key.offset;
2407 info++;
2408 }
2409
2410 ret = btrfs_add_leaf_ref(root, ref, shared);
2411 if (ret == -EEXIST && shared) {
2412 struct btrfs_leaf_ref *old;
2413 old = btrfs_lookup_leaf_ref(root, ref->bytenr);
2414 BUG_ON(!old);
2415 btrfs_remove_leaf_ref(root, old);
2416 btrfs_free_leaf_ref(root, old);
2417 ret = btrfs_add_leaf_ref(root, ref, shared);
2418 }
2419 WARN_ON(ret);
2420 btrfs_free_leaf_ref(root, ref);
2421 }
2422 out:
2423 return ret;
2424 }
2425
2426 /* when a block goes through cow, we update the reference counts of
2427 * everything that block points to. The internal pointers of the block
2428 * can be in just about any order, and it is likely to have clusters of
2429 * things that are close together and clusters of things that are not.
2430 *
2431 * To help reduce the seeks that come with updating all of these reference
2432 * counts, sort them by byte number before actual updates are done.
2433 *
2434 * struct refsort is used to match byte number to slot in the btree block.
2435 * we sort based on the byte number and then use the slot to actually
2436 * find the item.
2437 *
2438 * struct refsort is smaller than strcut btrfs_item and smaller than
2439 * struct btrfs_key_ptr. Since we're currently limited to the page size
2440 * for a btree block, there's no way for a kmalloc of refsorts for a
2441 * single node to be bigger than a page.
2442 */
2443 struct refsort {
2444 u64 bytenr;
2445 u32 slot;
2446 };
2447
2448 /*
2449 * for passing into sort()
2450 */
2451 static int refsort_cmp(const void *a_void, const void *b_void)
2452 {
2453 const struct refsort *a = a_void;
2454 const struct refsort *b = b_void;
2455
2456 if (a->bytenr < b->bytenr)
2457 return -1;
2458 if (a->bytenr > b->bytenr)
2459 return 1;
2460 return 0;
2461 }
2462 #endif
2463
2464 static int __btrfs_mod_ref(struct btrfs_trans_handle *trans,
2465 struct btrfs_root *root,
2466 struct extent_buffer *buf,
2467 int full_backref, int inc)
2468 {
2469 u64 bytenr;
2470 u64 num_bytes;
2471 u64 parent;
2472 u64 ref_root;
2473 u32 nritems;
2474 struct btrfs_key key;
2475 struct btrfs_file_extent_item *fi;
2476 int i;
2477 int level;
2478 int ret = 0;
2479 int (*process_func)(struct btrfs_trans_handle *, struct btrfs_root *,
2480 u64, u64, u64, u64, u64, u64);
2481
2482 ref_root = btrfs_header_owner(buf);
2483 nritems = btrfs_header_nritems(buf);
2484 level = btrfs_header_level(buf);
2485
2486 if (!root->ref_cows && level == 0)
2487 return 0;
2488
2489 if (inc)
2490 process_func = btrfs_inc_extent_ref;
2491 else
2492 process_func = btrfs_free_extent;
2493
2494 if (full_backref)
2495 parent = buf->start;
2496 else
2497 parent = 0;
2498
2499 for (i = 0; i < nritems; i++) {
2500 if (level == 0) {
2501 btrfs_item_key_to_cpu(buf, &key, i);
2502 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
2503 continue;
2504 fi = btrfs_item_ptr(buf, i,
2505 struct btrfs_file_extent_item);
2506 if (btrfs_file_extent_type(buf, fi) ==
2507 BTRFS_FILE_EXTENT_INLINE)
2508 continue;
2509 bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
2510 if (bytenr == 0)
2511 continue;
2512
2513 num_bytes = btrfs_file_extent_disk_num_bytes(buf, fi);
2514 key.offset -= btrfs_file_extent_offset(buf, fi);
2515 ret = process_func(trans, root, bytenr, num_bytes,
2516 parent, ref_root, key.objectid,
2517 key.offset);
2518 if (ret)
2519 goto fail;
2520 } else {
2521 bytenr = btrfs_node_blockptr(buf, i);
2522 num_bytes = btrfs_level_size(root, level - 1);
2523 ret = process_func(trans, root, bytenr, num_bytes,
2524 parent, ref_root, level - 1, 0);
2525 if (ret)
2526 goto fail;
2527 }
2528 }
2529 return 0;
2530 fail:
2531 BUG();
2532 return ret;
2533 }
2534
2535 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2536 struct extent_buffer *buf, int full_backref)
2537 {
2538 return __btrfs_mod_ref(trans, root, buf, full_backref, 1);
2539 }
2540
2541 int btrfs_dec_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2542 struct extent_buffer *buf, int full_backref)
2543 {
2544 return __btrfs_mod_ref(trans, root, buf, full_backref, 0);
2545 }
2546
2547 static int write_one_cache_group(struct btrfs_trans_handle *trans,
2548 struct btrfs_root *root,
2549 struct btrfs_path *path,
2550 struct btrfs_block_group_cache *cache)
2551 {
2552 int ret;
2553 struct btrfs_root *extent_root = root->fs_info->extent_root;
2554 unsigned long bi;
2555 struct extent_buffer *leaf;
2556
2557 ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1);
2558 if (ret < 0)
2559 goto fail;
2560 BUG_ON(ret);
2561
2562 leaf = path->nodes[0];
2563 bi = btrfs_item_ptr_offset(leaf, path->slots[0]);
2564 write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item));
2565 btrfs_mark_buffer_dirty(leaf);
2566 btrfs_release_path(extent_root, path);
2567 fail:
2568 if (ret)
2569 return ret;
2570 return 0;
2571
2572 }
2573
2574 static struct btrfs_block_group_cache *
2575 next_block_group(struct btrfs_root *root,
2576 struct btrfs_block_group_cache *cache)
2577 {
2578 struct rb_node *node;
2579 spin_lock(&root->fs_info->block_group_cache_lock);
2580 node = rb_next(&cache->cache_node);
2581 btrfs_put_block_group(cache);
2582 if (node) {
2583 cache = rb_entry(node, struct btrfs_block_group_cache,
2584 cache_node);
2585 atomic_inc(&cache->count);
2586 } else
2587 cache = NULL;
2588 spin_unlock(&root->fs_info->block_group_cache_lock);
2589 return cache;
2590 }
2591
2592 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
2593 struct btrfs_root *root)
2594 {
2595 struct btrfs_block_group_cache *cache;
2596 int err = 0;
2597 struct btrfs_path *path;
2598 u64 last = 0;
2599
2600 path = btrfs_alloc_path();
2601 if (!path)
2602 return -ENOMEM;
2603
2604 while (1) {
2605 if (last == 0) {
2606 err = btrfs_run_delayed_refs(trans, root,
2607 (unsigned long)-1);
2608 BUG_ON(err);
2609 }
2610
2611 cache = btrfs_lookup_first_block_group(root->fs_info, last);
2612 while (cache) {
2613 if (cache->dirty)
2614 break;
2615 cache = next_block_group(root, cache);
2616 }
2617 if (!cache) {
2618 if (last == 0)
2619 break;
2620 last = 0;
2621 continue;
2622 }
2623
2624 cache->dirty = 0;
2625 last = cache->key.objectid + cache->key.offset;
2626
2627 err = write_one_cache_group(trans, root, path, cache);
2628 BUG_ON(err);
2629 btrfs_put_block_group(cache);
2630 }
2631
2632 btrfs_free_path(path);
2633 return 0;
2634 }
2635
2636 int btrfs_extent_readonly(struct btrfs_root *root, u64 bytenr)
2637 {
2638 struct btrfs_block_group_cache *block_group;
2639 int readonly = 0;
2640
2641 block_group = btrfs_lookup_block_group(root->fs_info, bytenr);
2642 if (!block_group || block_group->ro)
2643 readonly = 1;
2644 if (block_group)
2645 btrfs_put_block_group(block_group);
2646 return readonly;
2647 }
2648
2649 static int update_space_info(struct btrfs_fs_info *info, u64 flags,
2650 u64 total_bytes, u64 bytes_used,
2651 struct btrfs_space_info **space_info)
2652 {
2653 struct btrfs_space_info *found;
2654
2655 found = __find_space_info(info, flags);
2656 if (found) {
2657 spin_lock(&found->lock);
2658 found->total_bytes += total_bytes;
2659 found->bytes_used += bytes_used;
2660 found->full = 0;
2661 spin_unlock(&found->lock);
2662 *space_info = found;
2663 return 0;
2664 }
2665 found = kzalloc(sizeof(*found), GFP_NOFS);
2666 if (!found)
2667 return -ENOMEM;
2668
2669 INIT_LIST_HEAD(&found->block_groups);
2670 init_rwsem(&found->groups_sem);
2671 spin_lock_init(&found->lock);
2672 found->flags = flags;
2673 found->total_bytes = total_bytes;
2674 found->bytes_used = bytes_used;
2675 found->bytes_pinned = 0;
2676 found->bytes_reserved = 0;
2677 found->bytes_readonly = 0;
2678 found->bytes_delalloc = 0;
2679 found->full = 0;
2680 found->force_alloc = 0;
2681 *space_info = found;
2682 list_add_rcu(&found->list, &info->space_info);
2683 atomic_set(&found->caching_threads, 0);
2684 return 0;
2685 }
2686
2687 static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags)
2688 {
2689 u64 extra_flags = flags & (BTRFS_BLOCK_GROUP_RAID0 |
2690 BTRFS_BLOCK_GROUP_RAID1 |
2691 BTRFS_BLOCK_GROUP_RAID10 |
2692 BTRFS_BLOCK_GROUP_DUP);
2693 if (extra_flags) {
2694 if (flags & BTRFS_BLOCK_GROUP_DATA)
2695 fs_info->avail_data_alloc_bits |= extra_flags;
2696 if (flags & BTRFS_BLOCK_GROUP_METADATA)
2697 fs_info->avail_metadata_alloc_bits |= extra_flags;
2698 if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
2699 fs_info->avail_system_alloc_bits |= extra_flags;
2700 }
2701 }
2702
2703 static void set_block_group_readonly(struct btrfs_block_group_cache *cache)
2704 {
2705 spin_lock(&cache->space_info->lock);
2706 spin_lock(&cache->lock);
2707 if (!cache->ro) {
2708 cache->space_info->bytes_readonly += cache->key.offset -
2709 btrfs_block_group_used(&cache->item);
2710 cache->ro = 1;
2711 }
2712 spin_unlock(&cache->lock);
2713 spin_unlock(&cache->space_info->lock);
2714 }
2715
2716 u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags)
2717 {
2718 u64 num_devices = root->fs_info->fs_devices->rw_devices;
2719
2720 if (num_devices == 1)
2721 flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0);
2722 if (num_devices < 4)
2723 flags &= ~BTRFS_BLOCK_GROUP_RAID10;
2724
2725 if ((flags & BTRFS_BLOCK_GROUP_DUP) &&
2726 (flags & (BTRFS_BLOCK_GROUP_RAID1 |
2727 BTRFS_BLOCK_GROUP_RAID10))) {
2728 flags &= ~BTRFS_BLOCK_GROUP_DUP;
2729 }
2730
2731 if ((flags & BTRFS_BLOCK_GROUP_RAID1) &&
2732 (flags & BTRFS_BLOCK_GROUP_RAID10)) {
2733 flags &= ~BTRFS_BLOCK_GROUP_RAID1;
2734 }
2735
2736 if ((flags & BTRFS_BLOCK_GROUP_RAID0) &&
2737 ((flags & BTRFS_BLOCK_GROUP_RAID1) |
2738 (flags & BTRFS_BLOCK_GROUP_RAID10) |
2739 (flags & BTRFS_BLOCK_GROUP_DUP)))
2740 flags &= ~BTRFS_BLOCK_GROUP_RAID0;
2741 return flags;
2742 }
2743
2744 static u64 btrfs_get_alloc_profile(struct btrfs_root *root, u64 data)
2745 {
2746 struct btrfs_fs_info *info = root->fs_info;
2747 u64 alloc_profile;
2748
2749 if (data) {
2750 alloc_profile = info->avail_data_alloc_bits &
2751 info->data_alloc_profile;
2752 data = BTRFS_BLOCK_GROUP_DATA | alloc_profile;
2753 } else if (root == root->fs_info->chunk_root) {
2754 alloc_profile = info->avail_system_alloc_bits &
2755 info->system_alloc_profile;
2756 data = BTRFS_BLOCK_GROUP_SYSTEM | alloc_profile;
2757 } else {
2758 alloc_profile = info->avail_metadata_alloc_bits &
2759 info->metadata_alloc_profile;
2760 data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile;
2761 }
2762
2763 return btrfs_reduce_alloc_profile(root, data);
2764 }
2765
2766 void btrfs_set_inode_space_info(struct btrfs_root *root, struct inode *inode)
2767 {
2768 u64 alloc_target;
2769
2770 alloc_target = btrfs_get_alloc_profile(root, 1);
2771 BTRFS_I(inode)->space_info = __find_space_info(root->fs_info,
2772 alloc_target);
2773 }
2774
2775 static u64 calculate_bytes_needed(struct btrfs_root *root, int num_items)
2776 {
2777 u64 num_bytes;
2778 int level;
2779
2780 level = BTRFS_MAX_LEVEL - 2;
2781 /*
2782 * NOTE: these calculations are absolutely the worst possible case.
2783 * This assumes that _every_ item we insert will require a new leaf, and
2784 * that the tree has grown to its maximum level size.
2785 */
2786
2787 /*
2788 * for every item we insert we could insert both an extent item and a
2789 * extent ref item. Then for ever item we insert, we will need to cow
2790 * both the original leaf, plus the leaf to the left and right of it.
2791 *
2792 * Unless we are talking about the extent root, then we just want the
2793 * number of items * 2, since we just need the extent item plus its ref.
2794 */
2795 if (root == root->fs_info->extent_root)
2796 num_bytes = num_items * 2;
2797 else
2798 num_bytes = (num_items + (2 * num_items)) * 3;
2799
2800 /*
2801 * num_bytes is total number of leaves we could need times the leaf
2802 * size, and then for every leaf we could end up cow'ing 2 nodes per
2803 * level, down to the leaf level.
2804 */
2805 num_bytes = (num_bytes * root->leafsize) +
2806 (num_bytes * (level * 2)) * root->nodesize;
2807
2808 return num_bytes;
2809 }
2810
2811 /*
2812 * Unreserve metadata space for delalloc. If we have less reserved credits than
2813 * we have extents, this function does nothing.
2814 */
2815 int btrfs_unreserve_metadata_for_delalloc(struct btrfs_root *root,
2816 struct inode *inode, int num_items)
2817 {
2818 struct btrfs_fs_info *info = root->fs_info;
2819 struct btrfs_space_info *meta_sinfo;
2820 u64 num_bytes;
2821 u64 alloc_target;
2822 bool bug = false;
2823
2824 /* get the space info for where the metadata will live */
2825 alloc_target = btrfs_get_alloc_profile(root, 0);
2826 meta_sinfo = __find_space_info(info, alloc_target);
2827
2828 num_bytes = calculate_bytes_needed(root->fs_info->extent_root,
2829 num_items);
2830
2831 spin_lock(&meta_sinfo->lock);
2832 spin_lock(&BTRFS_I(inode)->accounting_lock);
2833 if (BTRFS_I(inode)->reserved_extents <=
2834 BTRFS_I(inode)->outstanding_extents) {
2835 spin_unlock(&BTRFS_I(inode)->accounting_lock);
2836 spin_unlock(&meta_sinfo->lock);
2837 return 0;
2838 }
2839 spin_unlock(&BTRFS_I(inode)->accounting_lock);
2840
2841 BTRFS_I(inode)->reserved_extents--;
2842 BUG_ON(BTRFS_I(inode)->reserved_extents < 0);
2843
2844 if (meta_sinfo->bytes_delalloc < num_bytes) {
2845 bug = true;
2846 meta_sinfo->bytes_delalloc = 0;
2847 } else {
2848 meta_sinfo->bytes_delalloc -= num_bytes;
2849 }
2850 spin_unlock(&meta_sinfo->lock);
2851
2852 BUG_ON(bug);
2853
2854 return 0;
2855 }
2856
2857 static void check_force_delalloc(struct btrfs_space_info *meta_sinfo)
2858 {
2859 u64 thresh;
2860
2861 thresh = meta_sinfo->bytes_used + meta_sinfo->bytes_reserved +
2862 meta_sinfo->bytes_pinned + meta_sinfo->bytes_readonly +
2863 meta_sinfo->bytes_super + meta_sinfo->bytes_root +
2864 meta_sinfo->bytes_may_use;
2865
2866 thresh = meta_sinfo->total_bytes - thresh;
2867 thresh *= 80;
2868 do_div(thresh, 100);
2869 if (thresh <= meta_sinfo->bytes_delalloc)
2870 meta_sinfo->force_delalloc = 1;
2871 else
2872 meta_sinfo->force_delalloc = 0;
2873 }
2874
2875 struct async_flush {
2876 struct btrfs_root *root;
2877 struct btrfs_space_info *info;
2878 struct btrfs_work work;
2879 };
2880
2881 static noinline void flush_delalloc_async(struct btrfs_work *work)
2882 {
2883 struct async_flush *async;
2884 struct btrfs_root *root;
2885 struct btrfs_space_info *info;
2886
2887 async = container_of(work, struct async_flush, work);
2888 root = async->root;
2889 info = async->info;
2890
2891 btrfs_start_delalloc_inodes(root, 0);
2892 wake_up(&info->flush_wait);
2893 btrfs_wait_ordered_extents(root, 0, 0);
2894
2895 spin_lock(&info->lock);
2896 info->flushing = 0;
2897 spin_unlock(&info->lock);
2898 wake_up(&info->flush_wait);
2899
2900 kfree(async);
2901 }
2902
2903 static void wait_on_flush(struct btrfs_space_info *info)
2904 {
2905 DEFINE_WAIT(wait);
2906 u64 used;
2907
2908 while (1) {
2909 prepare_to_wait(&info->flush_wait, &wait,
2910 TASK_UNINTERRUPTIBLE);
2911 spin_lock(&info->lock);
2912 if (!info->flushing) {
2913 spin_unlock(&info->lock);
2914 break;
2915 }
2916
2917 used = info->bytes_used + info->bytes_reserved +
2918 info->bytes_pinned + info->bytes_readonly +
2919 info->bytes_super + info->bytes_root +
2920 info->bytes_may_use + info->bytes_delalloc;
2921 if (used < info->total_bytes) {
2922 spin_unlock(&info->lock);
2923 break;
2924 }
2925 spin_unlock(&info->lock);
2926 schedule();
2927 }
2928 finish_wait(&info->flush_wait, &wait);
2929 }
2930
2931 static void flush_delalloc(struct btrfs_root *root,
2932 struct btrfs_space_info *info)
2933 {
2934 struct async_flush *async;
2935 bool wait = false;
2936
2937 spin_lock(&info->lock);
2938
2939 if (!info->flushing) {
2940 info->flushing = 1;
2941 init_waitqueue_head(&info->flush_wait);
2942 } else {
2943 wait = true;
2944 }
2945
2946 spin_unlock(&info->lock);
2947
2948 if (wait) {
2949 wait_on_flush(info);
2950 return;
2951 }
2952
2953 async = kzalloc(sizeof(*async), GFP_NOFS);
2954 if (!async)
2955 goto flush;
2956
2957 async->root = root;
2958 async->info = info;
2959 async->work.func = flush_delalloc_async;
2960
2961 btrfs_queue_worker(&root->fs_info->enospc_workers,
2962 &async->work);
2963 wait_on_flush(info);
2964 return;
2965
2966 flush:
2967 btrfs_start_delalloc_inodes(root, 0);
2968 btrfs_wait_ordered_extents(root, 0, 0);
2969
2970 spin_lock(&info->lock);
2971 info->flushing = 0;
2972 spin_unlock(&info->lock);
2973 wake_up(&info->flush_wait);
2974 }
2975
2976 static int maybe_allocate_chunk(struct btrfs_root *root,
2977 struct btrfs_space_info *info)
2978 {
2979 struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
2980 struct btrfs_trans_handle *trans;
2981 bool wait = false;
2982 int ret = 0;
2983 u64 min_metadata;
2984 u64 free_space;
2985
2986 free_space = btrfs_super_total_bytes(disk_super);
2987 /*
2988 * we allow the metadata to grow to a max of either 10gb or 5% of the
2989 * space in the volume.
2990 */
2991 min_metadata = min((u64)10 * 1024 * 1024 * 1024,
2992 div64_u64(free_space * 5, 100));
2993 if (info->total_bytes >= min_metadata) {
2994 spin_unlock(&info->lock);
2995 return 0;
2996 }
2997
2998 if (info->full) {
2999 spin_unlock(&info->lock);
3000 return 0;
3001 }
3002
3003 if (!info->allocating_chunk) {
3004 info->force_alloc = 1;
3005 info->allocating_chunk = 1;
3006 init_waitqueue_head(&info->allocate_wait);
3007 } else {
3008 wait = true;
3009 }
3010
3011 spin_unlock(&info->lock);
3012
3013 if (wait) {
3014 wait_event(info->allocate_wait,
3015 !info->allocating_chunk);
3016 return 1;
3017 }
3018
3019 trans = btrfs_start_transaction(root, 1);
3020 if (!trans) {
3021 ret = -ENOMEM;
3022 goto out;
3023 }
3024
3025 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
3026 4096 + 2 * 1024 * 1024,
3027 info->flags, 0);
3028 btrfs_end_transaction(trans, root);
3029 if (ret)
3030 goto out;
3031 out:
3032 spin_lock(&info->lock);
3033 info->allocating_chunk = 0;
3034 spin_unlock(&info->lock);
3035 wake_up(&info->allocate_wait);
3036
3037 if (ret)
3038 return 0;
3039 return 1;
3040 }
3041
3042 /*
3043 * Reserve metadata space for delalloc.
3044 */
3045 int btrfs_reserve_metadata_for_delalloc(struct btrfs_root *root,
3046 struct inode *inode, int num_items)
3047 {
3048 struct btrfs_fs_info *info = root->fs_info;
3049 struct btrfs_space_info *meta_sinfo;
3050 u64 num_bytes;
3051 u64 used;
3052 u64 alloc_target;
3053 int flushed = 0;
3054 int force_delalloc;
3055
3056 /* get the space info for where the metadata will live */
3057 alloc_target = btrfs_get_alloc_profile(root, 0);
3058 meta_sinfo = __find_space_info(info, alloc_target);
3059
3060 num_bytes = calculate_bytes_needed(root->fs_info->extent_root,
3061 num_items);
3062 again:
3063 spin_lock(&meta_sinfo->lock);
3064
3065 force_delalloc = meta_sinfo->force_delalloc;
3066
3067 if (unlikely(!meta_sinfo->bytes_root))
3068 meta_sinfo->bytes_root = calculate_bytes_needed(root, 6);
3069
3070 if (!flushed)
3071 meta_sinfo->bytes_delalloc += num_bytes;
3072
3073 used = meta_sinfo->bytes_used + meta_sinfo->bytes_reserved +
3074 meta_sinfo->bytes_pinned + meta_sinfo->bytes_readonly +
3075 meta_sinfo->bytes_super + meta_sinfo->bytes_root +
3076 meta_sinfo->bytes_may_use + meta_sinfo->bytes_delalloc;
3077
3078 if (used > meta_sinfo->total_bytes) {
3079 flushed++;
3080
3081 if (flushed == 1) {
3082 if (maybe_allocate_chunk(root, meta_sinfo))
3083 goto again;
3084 flushed++;
3085 } else {
3086 spin_unlock(&meta_sinfo->lock);
3087 }
3088
3089 if (flushed == 2) {
3090 filemap_flush(inode->i_mapping);
3091 goto again;
3092 } else if (flushed == 3) {
3093 flush_delalloc(root, meta_sinfo);
3094 goto again;
3095 }
3096 spin_lock(&meta_sinfo->lock);
3097 meta_sinfo->bytes_delalloc -= num_bytes;
3098 spin_unlock(&meta_sinfo->lock);
3099 printk(KERN_ERR "enospc, has %d, reserved %d\n",
3100 BTRFS_I(inode)->outstanding_extents,
3101 BTRFS_I(inode)->reserved_extents);
3102 dump_space_info(meta_sinfo, 0, 0);
3103 return -ENOSPC;
3104 }
3105
3106 BTRFS_I(inode)->reserved_extents++;
3107 check_force_delalloc(meta_sinfo);
3108 spin_unlock(&meta_sinfo->lock);
3109
3110 if (!flushed && force_delalloc)
3111 filemap_flush(inode->i_mapping);
3112
3113 return 0;
3114 }
3115
3116 /*
3117 * unreserve num_items number of items worth of metadata space. This needs to
3118 * be paired with btrfs_reserve_metadata_space.
3119 *
3120 * NOTE: if you have the option, run this _AFTER_ you do a
3121 * btrfs_end_transaction, since btrfs_end_transaction will run delayed ref
3122 * oprations which will result in more used metadata, so we want to make sure we
3123 * can do that without issue.
3124 */
3125 int btrfs_unreserve_metadata_space(struct btrfs_root *root, int num_items)
3126 {
3127 struct btrfs_fs_info *info = root->fs_info;
3128 struct btrfs_space_info *meta_sinfo;
3129 u64 num_bytes;
3130 u64 alloc_target;
3131 bool bug = false;
3132
3133 /* get the space info for where the metadata will live */
3134 alloc_target = btrfs_get_alloc_profile(root, 0);
3135 meta_sinfo = __find_space_info(info, alloc_target);
3136
3137 num_bytes = calculate_bytes_needed(root, num_items);
3138
3139 spin_lock(&meta_sinfo->lock);
3140 if (meta_sinfo->bytes_may_use < num_bytes) {
3141 bug = true;
3142 meta_sinfo->bytes_may_use = 0;
3143 } else {
3144 meta_sinfo->bytes_may_use -= num_bytes;
3145 }
3146 spin_unlock(&meta_sinfo->lock);
3147
3148 BUG_ON(bug);
3149
3150 return 0;
3151 }
3152
3153 /*
3154 * Reserve some metadata space for use. We'll calculate the worste case number
3155 * of bytes that would be needed to modify num_items number of items. If we
3156 * have space, fantastic, if not, you get -ENOSPC. Please call
3157 * btrfs_unreserve_metadata_space when you are done for the _SAME_ number of
3158 * items you reserved, since whatever metadata you needed should have already
3159 * been allocated.
3160 *
3161 * This will commit the transaction to make more space if we don't have enough
3162 * metadata space. THe only time we don't do this is if we're reserving space
3163 * inside of a transaction, then we will just return -ENOSPC and it is the
3164 * callers responsibility to handle it properly.
3165 */
3166 int btrfs_reserve_metadata_space(struct btrfs_root *root, int num_items)
3167 {
3168 struct btrfs_fs_info *info = root->fs_info;
3169 struct btrfs_space_info *meta_sinfo;
3170 u64 num_bytes;
3171 u64 used;
3172 u64 alloc_target;
3173 int retries = 0;
3174
3175 /* get the space info for where the metadata will live */
3176 alloc_target = btrfs_get_alloc_profile(root, 0);
3177 meta_sinfo = __find_space_info(info, alloc_target);
3178
3179 num_bytes = calculate_bytes_needed(root, num_items);
3180 again:
3181 spin_lock(&meta_sinfo->lock);
3182
3183 if (unlikely(!meta_sinfo->bytes_root))
3184 meta_sinfo->bytes_root = calculate_bytes_needed(root, 6);
3185
3186 if (!retries)
3187 meta_sinfo->bytes_may_use += num_bytes;
3188
3189 used = meta_sinfo->bytes_used + meta_sinfo->bytes_reserved +
3190 meta_sinfo->bytes_pinned + meta_sinfo->bytes_readonly +
3191 meta_sinfo->bytes_super + meta_sinfo->bytes_root +
3192 meta_sinfo->bytes_may_use + meta_sinfo->bytes_delalloc;
3193
3194 if (used > meta_sinfo->total_bytes) {
3195 retries++;
3196 if (retries == 1) {
3197 if (maybe_allocate_chunk(root, meta_sinfo))
3198 goto again;
3199 retries++;
3200 } else {
3201 spin_unlock(&meta_sinfo->lock);
3202 }
3203
3204 if (retries == 2) {
3205 flush_delalloc(root, meta_sinfo);
3206 goto again;
3207 }
3208 spin_lock(&meta_sinfo->lock);
3209 meta_sinfo->bytes_may_use -= num_bytes;
3210 spin_unlock(&meta_sinfo->lock);
3211
3212 dump_space_info(meta_sinfo, 0, 0);
3213 return -ENOSPC;
3214 }
3215
3216 check_force_delalloc(meta_sinfo);
3217 spin_unlock(&meta_sinfo->lock);
3218
3219 return 0;
3220 }
3221
3222 /*
3223 * This will check the space that the inode allocates from to make sure we have
3224 * enough space for bytes.
3225 */
3226 int btrfs_check_data_free_space(struct btrfs_root *root, struct inode *inode,
3227 u64 bytes)
3228 {
3229 struct btrfs_space_info *data_sinfo;
3230 int ret = 0, committed = 0;
3231
3232 /* make sure bytes are sectorsize aligned */
3233 bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
3234
3235 data_sinfo = BTRFS_I(inode)->space_info;
3236 if (!data_sinfo)
3237 goto alloc;
3238
3239 again:
3240 /* make sure we have enough space to handle the data first */
3241 spin_lock(&data_sinfo->lock);
3242 if (data_sinfo->total_bytes - data_sinfo->bytes_used -
3243 data_sinfo->bytes_delalloc - data_sinfo->bytes_reserved -
3244 data_sinfo->bytes_pinned - data_sinfo->bytes_readonly -
3245 data_sinfo->bytes_may_use - data_sinfo->bytes_super < bytes) {
3246 struct btrfs_trans_handle *trans;
3247
3248 /*
3249 * if we don't have enough free bytes in this space then we need
3250 * to alloc a new chunk.
3251 */
3252 if (!data_sinfo->full) {
3253 u64 alloc_target;
3254
3255 data_sinfo->force_alloc = 1;
3256 spin_unlock(&data_sinfo->lock);
3257 alloc:
3258 alloc_target = btrfs_get_alloc_profile(root, 1);
3259 trans = btrfs_start_transaction(root, 1);
3260 if (!trans)
3261 return -ENOMEM;
3262
3263 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
3264 bytes + 2 * 1024 * 1024,
3265 alloc_target, 0);
3266 btrfs_end_transaction(trans, root);
3267 if (ret)
3268 return ret;
3269
3270 if (!data_sinfo) {
3271 btrfs_set_inode_space_info(root, inode);
3272 data_sinfo = BTRFS_I(inode)->space_info;
3273 }
3274 goto again;
3275 }
3276 spin_unlock(&data_sinfo->lock);
3277
3278 /* commit the current transaction and try again */
3279 if (!committed && !root->fs_info->open_ioctl_trans) {
3280 committed = 1;
3281 trans = btrfs_join_transaction(root, 1);
3282 if (!trans)
3283 return -ENOMEM;
3284 ret = btrfs_commit_transaction(trans, root);
3285 if (ret)
3286 return ret;
3287 goto again;
3288 }
3289
3290 printk(KERN_ERR "no space left, need %llu, %llu delalloc bytes"
3291 ", %llu bytes_used, %llu bytes_reserved, "
3292 "%llu bytes_pinned, %llu bytes_readonly, %llu may use "
3293 "%llu total\n", (unsigned long long)bytes,
3294 (unsigned long long)data_sinfo->bytes_delalloc,
3295 (unsigned long long)data_sinfo->bytes_used,
3296 (unsigned long long)data_sinfo->bytes_reserved,
3297 (unsigned long long)data_sinfo->bytes_pinned,
3298 (unsigned long long)data_sinfo->bytes_readonly,
3299 (unsigned long long)data_sinfo->bytes_may_use,
3300 (unsigned long long)data_sinfo->total_bytes);
3301 return -ENOSPC;
3302 }
3303 data_sinfo->bytes_may_use += bytes;
3304 BTRFS_I(inode)->reserved_bytes += bytes;
3305 spin_unlock(&data_sinfo->lock);
3306
3307 return 0;
3308 }
3309
3310 /*
3311 * if there was an error for whatever reason after calling
3312 * btrfs_check_data_free_space, call this so we can cleanup the counters.
3313 */
3314 void btrfs_free_reserved_data_space(struct btrfs_root *root,
3315 struct inode *inode, u64 bytes)
3316 {
3317 struct btrfs_space_info *data_sinfo;
3318
3319 /* make sure bytes are sectorsize aligned */
3320 bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
3321
3322 data_sinfo = BTRFS_I(inode)->space_info;
3323 spin_lock(&data_sinfo->lock);
3324 data_sinfo->bytes_may_use -= bytes;
3325 BTRFS_I(inode)->reserved_bytes -= bytes;
3326 spin_unlock(&data_sinfo->lock);
3327 }
3328
3329 /* called when we are adding a delalloc extent to the inode's io_tree */
3330 void btrfs_delalloc_reserve_space(struct btrfs_root *root, struct inode *inode,
3331 u64 bytes)
3332 {
3333 struct btrfs_space_info *data_sinfo;
3334
3335 /* get the space info for where this inode will be storing its data */
3336 data_sinfo = BTRFS_I(inode)->space_info;
3337
3338 /* make sure we have enough space to handle the data first */
3339 spin_lock(&data_sinfo->lock);
3340 data_sinfo->bytes_delalloc += bytes;
3341
3342 /*
3343 * we are adding a delalloc extent without calling
3344 * btrfs_check_data_free_space first. This happens on a weird
3345 * writepage condition, but shouldn't hurt our accounting
3346 */
3347 if (unlikely(bytes > BTRFS_I(inode)->reserved_bytes)) {
3348 data_sinfo->bytes_may_use -= BTRFS_I(inode)->reserved_bytes;
3349 BTRFS_I(inode)->reserved_bytes = 0;
3350 } else {
3351 data_sinfo->bytes_may_use -= bytes;
3352 BTRFS_I(inode)->reserved_bytes -= bytes;
3353 }
3354
3355 spin_unlock(&data_sinfo->lock);
3356 }
3357
3358 /* called when we are clearing an delalloc extent from the inode's io_tree */
3359 void btrfs_delalloc_free_space(struct btrfs_root *root, struct inode *inode,
3360 u64 bytes)
3361 {
3362 struct btrfs_space_info *info;
3363
3364 info = BTRFS_I(inode)->space_info;
3365
3366 spin_lock(&info->lock);
3367 info->bytes_delalloc -= bytes;
3368 spin_unlock(&info->lock);
3369 }
3370
3371 static void force_metadata_allocation(struct btrfs_fs_info *info)
3372 {
3373 struct list_head *head = &info->space_info;
3374 struct btrfs_space_info *found;
3375
3376 rcu_read_lock();
3377 list_for_each_entry_rcu(found, head, list) {
3378 if (found->flags & BTRFS_BLOCK_GROUP_METADATA)
3379 found->force_alloc = 1;
3380 }
3381 rcu_read_unlock();
3382 }
3383
3384 static int do_chunk_alloc(struct btrfs_trans_handle *trans,
3385 struct btrfs_root *extent_root, u64 alloc_bytes,
3386 u64 flags, int force)
3387 {
3388 struct btrfs_space_info *space_info;
3389 struct btrfs_fs_info *fs_info = extent_root->fs_info;
3390 u64 thresh;
3391 int ret = 0;
3392
3393 mutex_lock(&fs_info->chunk_mutex);
3394
3395 flags = btrfs_reduce_alloc_profile(extent_root, flags);
3396
3397 space_info = __find_space_info(extent_root->fs_info, flags);
3398 if (!space_info) {
3399 ret = update_space_info(extent_root->fs_info, flags,
3400 0, 0, &space_info);
3401 BUG_ON(ret);
3402 }
3403 BUG_ON(!space_info);
3404
3405 spin_lock(&space_info->lock);
3406 if (space_info->force_alloc)
3407 force = 1;
3408 if (space_info->full) {
3409 spin_unlock(&space_info->lock);
3410 goto out;
3411 }
3412
3413 thresh = space_info->total_bytes - space_info->bytes_readonly;
3414 thresh = div_factor(thresh, 8);
3415 if (!force &&
3416 (space_info->bytes_used + space_info->bytes_pinned +
3417 space_info->bytes_reserved + alloc_bytes) < thresh) {
3418 spin_unlock(&space_info->lock);
3419 goto out;
3420 }
3421 spin_unlock(&space_info->lock);
3422
3423 /*
3424 * if we're doing a data chunk, go ahead and make sure that
3425 * we keep a reasonable number of metadata chunks allocated in the
3426 * FS as well.
3427 */
3428 if (flags & BTRFS_BLOCK_GROUP_DATA && fs_info->metadata_ratio) {
3429 fs_info->data_chunk_allocations++;
3430 if (!(fs_info->data_chunk_allocations %
3431 fs_info->metadata_ratio))
3432 force_metadata_allocation(fs_info);
3433 }
3434
3435 ret = btrfs_alloc_chunk(trans, extent_root, flags);
3436 spin_lock(&space_info->lock);
3437 if (ret)
3438 space_info->full = 1;
3439 space_info->force_alloc = 0;
3440 spin_unlock(&space_info->lock);
3441 out:
3442 mutex_unlock(&extent_root->fs_info->chunk_mutex);
3443 return ret;
3444 }
3445
3446 static int update_block_group(struct btrfs_trans_handle *trans,
3447 struct btrfs_root *root,
3448 u64 bytenr, u64 num_bytes, int alloc,
3449 int mark_free)
3450 {
3451 struct btrfs_block_group_cache *cache;
3452 struct btrfs_fs_info *info = root->fs_info;
3453 u64 total = num_bytes;
3454 u64 old_val;
3455 u64 byte_in_group;
3456
3457 /* block accounting for super block */
3458 spin_lock(&info->delalloc_lock);
3459 old_val = btrfs_super_bytes_used(&info->super_copy);
3460 if (alloc)
3461 old_val += num_bytes;
3462 else
3463 old_val -= num_bytes;
3464 btrfs_set_super_bytes_used(&info->super_copy, old_val);
3465 spin_unlock(&info->delalloc_lock);
3466
3467 while (total) {
3468 cache = btrfs_lookup_block_group(info, bytenr);
3469 if (!cache)
3470 return -1;
3471 byte_in_group = bytenr - cache->key.objectid;
3472 WARN_ON(byte_in_group > cache->key.offset);
3473
3474 spin_lock(&cache->space_info->lock);
3475 spin_lock(&cache->lock);
3476 cache->dirty = 1;
3477 old_val = btrfs_block_group_used(&cache->item);
3478 num_bytes = min(total, cache->key.offset - byte_in_group);
3479 if (alloc) {
3480 old_val += num_bytes;
3481 btrfs_set_block_group_used(&cache->item, old_val);
3482 cache->reserved -= num_bytes;
3483 cache->space_info->bytes_used += num_bytes;
3484 cache->space_info->bytes_reserved -= num_bytes;
3485 if (cache->ro)
3486 cache->space_info->bytes_readonly -= num_bytes;
3487 spin_unlock(&cache->lock);
3488 spin_unlock(&cache->space_info->lock);
3489 } else {
3490 old_val -= num_bytes;
3491 cache->space_info->bytes_used -= num_bytes;
3492 if (cache->ro)
3493 cache->space_info->bytes_readonly += num_bytes;
3494 btrfs_set_block_group_used(&cache->item, old_val);
3495 spin_unlock(&cache->lock);
3496 spin_unlock(&cache->space_info->lock);
3497 if (mark_free) {
3498 int ret;
3499
3500 ret = btrfs_discard_extent(root, bytenr,
3501 num_bytes);
3502 WARN_ON(ret);
3503
3504 ret = btrfs_add_free_space(cache, bytenr,
3505 num_bytes);
3506 WARN_ON(ret);
3507 }
3508 }
3509 btrfs_put_block_group(cache);
3510 total -= num_bytes;
3511 bytenr += num_bytes;
3512 }
3513 return 0;
3514 }
3515
3516 static u64 first_logical_byte(struct btrfs_root *root, u64 search_start)
3517 {
3518 struct btrfs_block_group_cache *cache;
3519 u64 bytenr;
3520
3521 cache = btrfs_lookup_first_block_group(root->fs_info, search_start);
3522 if (!cache)
3523 return 0;
3524
3525 bytenr = cache->key.objectid;
3526 btrfs_put_block_group(cache);
3527
3528 return bytenr;
3529 }
3530
3531 /*
3532 * this function must be called within transaction
3533 */
3534 int btrfs_pin_extent(struct btrfs_root *root,
3535 u64 bytenr, u64 num_bytes, int reserved)
3536 {
3537 struct btrfs_fs_info *fs_info = root->fs_info;
3538 struct btrfs_block_group_cache *cache;
3539
3540 cache = btrfs_lookup_block_group(fs_info, bytenr);
3541 BUG_ON(!cache);
3542
3543 spin_lock(&cache->space_info->lock);
3544 spin_lock(&cache->lock);
3545 cache->pinned += num_bytes;
3546 cache->space_info->bytes_pinned += num_bytes;
3547 if (reserved) {
3548 cache->reserved -= num_bytes;
3549 cache->space_info->bytes_reserved -= num_bytes;
3550 }
3551 spin_unlock(&cache->lock);
3552 spin_unlock(&cache->space_info->lock);
3553
3554 btrfs_put_block_group(cache);
3555
3556 set_extent_dirty(fs_info->pinned_extents,
3557 bytenr, bytenr + num_bytes - 1, GFP_NOFS);
3558 return 0;
3559 }
3560
3561 static int update_reserved_extents(struct btrfs_block_group_cache *cache,
3562 u64 num_bytes, int reserve)
3563 {
3564 spin_lock(&cache->space_info->lock);
3565 spin_lock(&cache->lock);
3566 if (reserve) {
3567 cache->reserved += num_bytes;
3568 cache->space_info->bytes_reserved += num_bytes;
3569 } else {
3570 cache->reserved -= num_bytes;
3571 cache->space_info->bytes_reserved -= num_bytes;
3572 }
3573 spin_unlock(&cache->lock);
3574 spin_unlock(&cache->space_info->lock);
3575 return 0;
3576 }
3577
3578 int btrfs_prepare_extent_commit(struct btrfs_trans_handle *trans,
3579 struct btrfs_root *root)
3580 {
3581 struct btrfs_fs_info *fs_info = root->fs_info;
3582 struct btrfs_caching_control *next;
3583 struct btrfs_caching_control *caching_ctl;
3584 struct btrfs_block_group_cache *cache;
3585
3586 down_write(&fs_info->extent_commit_sem);
3587
3588 list_for_each_entry_safe(caching_ctl, next,
3589 &fs_info->caching_block_groups, list) {
3590 cache = caching_ctl->block_group;
3591 if (block_group_cache_done(cache)) {
3592 cache->last_byte_to_unpin = (u64)-1;
3593 list_del_init(&caching_ctl->list);
3594 put_caching_control(caching_ctl);
3595 } else {
3596 cache->last_byte_to_unpin = caching_ctl->progress;
3597 }
3598 }
3599
3600 if (fs_info->pinned_extents == &fs_info->freed_extents[0])
3601 fs_info->pinned_extents = &fs_info->freed_extents[1];
3602 else
3603 fs_info->pinned_extents = &fs_info->freed_extents[0];
3604
3605 up_write(&fs_info->extent_commit_sem);
3606 return 0;
3607 }
3608
3609 static int unpin_extent_range(struct btrfs_root *root, u64 start, u64 end)
3610 {
3611 struct btrfs_fs_info *fs_info = root->fs_info;
3612 struct btrfs_block_group_cache *cache = NULL;
3613 u64 len;
3614
3615 while (start <= end) {
3616 if (!cache ||
3617 start >= cache->key.objectid + cache->key.offset) {
3618 if (cache)
3619 btrfs_put_block_group(cache);
3620 cache = btrfs_lookup_block_group(fs_info, start);
3621 BUG_ON(!cache);
3622 }
3623
3624 len = cache->key.objectid + cache->key.offset - start;
3625 len = min(len, end + 1 - start);
3626
3627 if (start < cache->last_byte_to_unpin) {
3628 len = min(len, cache->last_byte_to_unpin - start);
3629 btrfs_add_free_space(cache, start, len);
3630 }
3631
3632 spin_lock(&cache->space_info->lock);
3633 spin_lock(&cache->lock);
3634 cache->pinned -= len;
3635 cache->space_info->bytes_pinned -= len;
3636 spin_unlock(&cache->lock);
3637 spin_unlock(&cache->space_info->lock);
3638
3639 start += len;
3640 }
3641
3642 if (cache)
3643 btrfs_put_block_group(cache);
3644 return 0;
3645 }
3646
3647 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
3648 struct btrfs_root *root)
3649 {
3650 struct btrfs_fs_info *fs_info = root->fs_info;
3651 struct extent_io_tree *unpin;
3652 u64 start;
3653 u64 end;
3654 int ret;
3655
3656 if (fs_info->pinned_extents == &fs_info->freed_extents[0])
3657 unpin = &fs_info->freed_extents[1];
3658 else
3659 unpin = &fs_info->freed_extents[0];
3660
3661 while (1) {
3662 ret = find_first_extent_bit(unpin, 0, &start, &end,
3663 EXTENT_DIRTY);
3664 if (ret)
3665 break;
3666
3667 ret = btrfs_discard_extent(root, start, end + 1 - start);
3668
3669 clear_extent_dirty(unpin, start, end, GFP_NOFS);
3670 unpin_extent_range(root, start, end);
3671 cond_resched();
3672 }
3673
3674 return ret;
3675 }
3676
3677 static int pin_down_bytes(struct btrfs_trans_handle *trans,
3678 struct btrfs_root *root,
3679 struct btrfs_path *path,
3680 u64 bytenr, u64 num_bytes,
3681 int is_data, int reserved,
3682 struct extent_buffer **must_clean)
3683 {
3684 int err = 0;
3685 struct extent_buffer *buf;
3686
3687 if (is_data)
3688 goto pinit;
3689
3690 /*
3691 * discard is sloooow, and so triggering discards on
3692 * individual btree blocks isn't a good plan. Just
3693 * pin everything in discard mode.
3694 */
3695 if (btrfs_test_opt(root, DISCARD))
3696 goto pinit;
3697
3698 buf = btrfs_find_tree_block(root, bytenr, num_bytes);
3699 if (!buf)
3700 goto pinit;
3701
3702 /* we can reuse a block if it hasn't been written
3703 * and it is from this transaction. We can't
3704 * reuse anything from the tree log root because
3705 * it has tiny sub-transactions.
3706 */
3707 if (btrfs_buffer_uptodate(buf, 0) &&
3708 btrfs_try_tree_lock(buf)) {
3709 u64 header_owner = btrfs_header_owner(buf);
3710 u64 header_transid = btrfs_header_generation(buf);
3711 if (header_owner != BTRFS_TREE_LOG_OBJECTID &&
3712 header_transid == trans->transid &&
3713 !btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
3714 *must_clean = buf;
3715 return 1;
3716 }
3717 btrfs_tree_unlock(buf);
3718 }
3719 free_extent_buffer(buf);
3720 pinit:
3721 if (path)
3722 btrfs_set_path_blocking(path);
3723 /* unlocks the pinned mutex */
3724 btrfs_pin_extent(root, bytenr, num_bytes, reserved);
3725
3726 BUG_ON(err < 0);
3727 return 0;
3728 }
3729
3730 static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
3731 struct btrfs_root *root,
3732 u64 bytenr, u64 num_bytes, u64 parent,
3733 u64 root_objectid, u64 owner_objectid,
3734 u64 owner_offset, int refs_to_drop,
3735 struct btrfs_delayed_extent_op *extent_op)
3736 {
3737 struct btrfs_key key;
3738 struct btrfs_path *path;
3739 struct btrfs_fs_info *info = root->fs_info;
3740 struct btrfs_root *extent_root = info->extent_root;
3741 struct extent_buffer *leaf;
3742 struct btrfs_extent_item *ei;
3743 struct btrfs_extent_inline_ref *iref;
3744 int ret;
3745 int is_data;
3746 int extent_slot = 0;
3747 int found_extent = 0;
3748 int num_to_del = 1;
3749 u32 item_size;
3750 u64 refs;
3751
3752 path = btrfs_alloc_path();
3753 if (!path)
3754 return -ENOMEM;
3755
3756 path->reada = 1;
3757 path->leave_spinning = 1;
3758
3759 is_data = owner_objectid >= BTRFS_FIRST_FREE_OBJECTID;
3760 BUG_ON(!is_data && refs_to_drop != 1);
3761
3762 ret = lookup_extent_backref(trans, extent_root, path, &iref,
3763 bytenr, num_bytes, parent,
3764 root_objectid, owner_objectid,
3765 owner_offset);
3766 if (ret == 0) {
3767 extent_slot = path->slots[0];
3768 while (extent_slot >= 0) {
3769 btrfs_item_key_to_cpu(path->nodes[0], &key,
3770 extent_slot);
3771 if (key.objectid != bytenr)
3772 break;
3773 if (key.type == BTRFS_EXTENT_ITEM_KEY &&
3774 key.offset == num_bytes) {
3775 found_extent = 1;
3776 break;
3777 }
3778 if (path->slots[0] - extent_slot > 5)
3779 break;
3780 extent_slot--;
3781 }
3782 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
3783 item_size = btrfs_item_size_nr(path->nodes[0], extent_slot);
3784 if (found_extent && item_size < sizeof(*ei))
3785 found_extent = 0;
3786 #endif
3787 if (!found_extent) {
3788 BUG_ON(iref);
3789 ret = remove_extent_backref(trans, extent_root, path,
3790 NULL, refs_to_drop,
3791 is_data);
3792 BUG_ON(ret);
3793 btrfs_release_path(extent_root, path);
3794 path->leave_spinning = 1;
3795
3796 key.objectid = bytenr;
3797 key.type = BTRFS_EXTENT_ITEM_KEY;
3798 key.offset = num_bytes;
3799
3800 ret = btrfs_search_slot(trans, extent_root,
3801 &key, path, -1, 1);
3802 if (ret) {
3803 printk(KERN_ERR "umm, got %d back from search"
3804 ", was looking for %llu\n", ret,
3805 (unsigned long long)bytenr);
3806 btrfs_print_leaf(extent_root, path->nodes[0]);
3807 }
3808 BUG_ON(ret);
3809 extent_slot = path->slots[0];
3810 }
3811 } else {
3812 btrfs_print_leaf(extent_root, path->nodes[0]);
3813 WARN_ON(1);
3814 printk(KERN_ERR "btrfs unable to find ref byte nr %llu "
3815 "parent %llu root %llu owner %llu offset %llu\n",
3816 (unsigned long long)bytenr,
3817 (unsigned long long)parent,
3818 (unsigned long long)root_objectid,
3819 (unsigned long long)owner_objectid,
3820 (unsigned long long)owner_offset);
3821 }
3822
3823 leaf = path->nodes[0];
3824 item_size = btrfs_item_size_nr(leaf, extent_slot);
3825 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
3826 if (item_size < sizeof(*ei)) {
3827 BUG_ON(found_extent || extent_slot != path->slots[0]);
3828 ret = convert_extent_item_v0(trans, extent_root, path,
3829 owner_objectid, 0);
3830 BUG_ON(ret < 0);
3831
3832 btrfs_release_path(extent_root, path);
3833 path->leave_spinning = 1;
3834
3835 key.objectid = bytenr;
3836 key.type = BTRFS_EXTENT_ITEM_KEY;
3837 key.offset = num_bytes;
3838
3839 ret = btrfs_search_slot(trans, extent_root, &key, path,
3840 -1, 1);
3841 if (ret) {
3842 printk(KERN_ERR "umm, got %d back from search"
3843 ", was looking for %llu\n", ret,
3844 (unsigned long long)bytenr);
3845 btrfs_print_leaf(extent_root, path->nodes[0]);
3846 }
3847 BUG_ON(ret);
3848 extent_slot = path->slots[0];
3849 leaf = path->nodes[0];
3850 item_size = btrfs_item_size_nr(leaf, extent_slot);
3851 }
3852 #endif
3853 BUG_ON(item_size < sizeof(*ei));
3854 ei = btrfs_item_ptr(leaf, extent_slot,
3855 struct btrfs_extent_item);
3856 if (owner_objectid < BTRFS_FIRST_FREE_OBJECTID) {
3857 struct btrfs_tree_block_info *bi;
3858 BUG_ON(item_size < sizeof(*ei) + sizeof(*bi));
3859 bi = (struct btrfs_tree_block_info *)(ei + 1);
3860 WARN_ON(owner_objectid != btrfs_tree_block_level(leaf, bi));
3861 }
3862
3863 refs = btrfs_extent_refs(leaf, ei);
3864 BUG_ON(refs < refs_to_drop);
3865 refs -= refs_to_drop;
3866
3867 if (refs > 0) {
3868 if (extent_op)
3869 __run_delayed_extent_op(extent_op, leaf, ei);
3870 /*
3871 * In the case of inline back ref, reference count will
3872 * be updated by remove_extent_backref
3873 */
3874 if (iref) {
3875 BUG_ON(!found_extent);
3876 } else {
3877 btrfs_set_extent_refs(leaf, ei, refs);
3878 btrfs_mark_buffer_dirty(leaf);
3879 }
3880 if (found_extent) {
3881 ret = remove_extent_backref(trans, extent_root, path,
3882 iref, refs_to_drop,
3883 is_data);
3884 BUG_ON(ret);
3885 }
3886 } else {
3887 int mark_free = 0;
3888 struct extent_buffer *must_clean = NULL;
3889
3890 if (found_extent) {
3891 BUG_ON(is_data && refs_to_drop !=
3892 extent_data_ref_count(root, path, iref));
3893 if (iref) {
3894 BUG_ON(path->slots[0] != extent_slot);
3895 } else {
3896 BUG_ON(path->slots[0] != extent_slot + 1);
3897 path->slots[0] = extent_slot;
3898 num_to_del = 2;
3899 }
3900 }
3901
3902 ret = pin_down_bytes(trans, root, path, bytenr,
3903 num_bytes, is_data, 0, &must_clean);
3904 if (ret > 0)
3905 mark_free = 1;
3906 BUG_ON(ret < 0);
3907 /*
3908 * it is going to be very rare for someone to be waiting
3909 * on the block we're freeing. del_items might need to
3910 * schedule, so rather than get fancy, just force it
3911 * to blocking here
3912 */
3913 if (must_clean)
3914 btrfs_set_lock_blocking(must_clean);
3915
3916 ret = btrfs_del_items(trans, extent_root, path, path->slots[0],
3917 num_to_del);
3918 BUG_ON(ret);
3919 btrfs_release_path(extent_root, path);
3920
3921 if (must_clean) {
3922 clean_tree_block(NULL, root, must_clean);
3923 btrfs_tree_unlock(must_clean);
3924 free_extent_buffer(must_clean);
3925 }
3926
3927 if (is_data) {
3928 ret = btrfs_del_csums(trans, root, bytenr, num_bytes);
3929 BUG_ON(ret);
3930 } else {
3931 invalidate_mapping_pages(info->btree_inode->i_mapping,
3932 bytenr >> PAGE_CACHE_SHIFT,
3933 (bytenr + num_bytes - 1) >> PAGE_CACHE_SHIFT);
3934 }
3935
3936 ret = update_block_group(trans, root, bytenr, num_bytes, 0,
3937 mark_free);
3938 BUG_ON(ret);
3939 }
3940 btrfs_free_path(path);
3941 return ret;
3942 }
3943
3944 /*
3945 * when we free an extent, it is possible (and likely) that we free the last
3946 * delayed ref for that extent as well. This searches the delayed ref tree for
3947 * a given extent, and if there are no other delayed refs to be processed, it
3948 * removes it from the tree.
3949 */
3950 static noinline int check_ref_cleanup(struct btrfs_trans_handle *trans,
3951 struct btrfs_root *root, u64 bytenr)
3952 {
3953 struct btrfs_delayed_ref_head *head;
3954 struct btrfs_delayed_ref_root *delayed_refs;
3955 struct btrfs_delayed_ref_node *ref;
3956 struct rb_node *node;
3957 int ret;
3958
3959 delayed_refs = &trans->transaction->delayed_refs;
3960 spin_lock(&delayed_refs->lock);
3961 head = btrfs_find_delayed_ref_head(trans, bytenr);
3962 if (!head)
3963 goto out;
3964
3965 node = rb_prev(&head->node.rb_node);
3966 if (!node)
3967 goto out;
3968
3969 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
3970
3971 /* there are still entries for this ref, we can't drop it */
3972 if (ref->bytenr == bytenr)
3973 goto out;
3974
3975 if (head->extent_op) {
3976 if (!head->must_insert_reserved)
3977 goto out;
3978 kfree(head->extent_op);
3979 head->extent_op = NULL;
3980 }
3981
3982 /*
3983 * waiting for the lock here would deadlock. If someone else has it
3984 * locked they are already in the process of dropping it anyway
3985 */
3986 if (!mutex_trylock(&head->mutex))
3987 goto out;
3988
3989 /*
3990 * at this point we have a head with no other entries. Go
3991 * ahead and process it.
3992 */
3993 head->node.in_tree = 0;
3994 rb_erase(&head->node.rb_node, &delayed_refs->root);
3995
3996 delayed_refs->num_entries--;
3997
3998 /*
3999 * we don't take a ref on the node because we're removing it from the
4000 * tree, so we just steal the ref the tree was holding.
4001 */
4002 delayed_refs->num_heads--;
4003 if (list_empty(&head->cluster))
4004 delayed_refs->num_heads_ready--;
4005
4006 list_del_init(&head->cluster);
4007 spin_unlock(&delayed_refs->lock);
4008
4009 ret = run_one_delayed_ref(trans, root->fs_info->tree_root,
4010 &head->node, head->extent_op,
4011 head->must_insert_reserved);
4012 BUG_ON(ret);
4013 btrfs_put_delayed_ref(&head->node);
4014 return 0;
4015 out:
4016 spin_unlock(&delayed_refs->lock);
4017 return 0;
4018 }
4019
4020 int btrfs_free_extent(struct btrfs_trans_handle *trans,
4021 struct btrfs_root *root,
4022 u64 bytenr, u64 num_bytes, u64 parent,
4023 u64 root_objectid, u64 owner, u64 offset)
4024 {
4025 int ret;
4026
4027 /*
4028 * tree log blocks never actually go into the extent allocation
4029 * tree, just update pinning info and exit early.
4030 */
4031 if (root_objectid == BTRFS_TREE_LOG_OBJECTID) {
4032 WARN_ON(owner >= BTRFS_FIRST_FREE_OBJECTID);
4033 /* unlocks the pinned mutex */
4034 btrfs_pin_extent(root, bytenr, num_bytes, 1);
4035 ret = 0;
4036 } else if (owner < BTRFS_FIRST_FREE_OBJECTID) {
4037 ret = btrfs_add_delayed_tree_ref(trans, bytenr, num_bytes,
4038 parent, root_objectid, (int)owner,
4039 BTRFS_DROP_DELAYED_REF, NULL);
4040 BUG_ON(ret);
4041 ret = check_ref_cleanup(trans, root, bytenr);
4042 BUG_ON(ret);
4043 } else {
4044 ret = btrfs_add_delayed_data_ref(trans, bytenr, num_bytes,
4045 parent, root_objectid, owner,
4046 offset, BTRFS_DROP_DELAYED_REF, NULL);
4047 BUG_ON(ret);
4048 }
4049 return ret;
4050 }
4051
4052 int btrfs_free_tree_block(struct btrfs_trans_handle *trans,
4053 struct btrfs_root *root,
4054 u64 bytenr, u32 blocksize,
4055 u64 parent, u64 root_objectid, int level)
4056 {
4057 u64 used;
4058 spin_lock(&root->node_lock);
4059 used = btrfs_root_used(&root->root_item) - blocksize;
4060 btrfs_set_root_used(&root->root_item, used);
4061 spin_unlock(&root->node_lock);
4062
4063 return btrfs_free_extent(trans, root, bytenr, blocksize,
4064 parent, root_objectid, level, 0);
4065 }
4066
4067 static u64 stripe_align(struct btrfs_root *root, u64 val)
4068 {
4069 u64 mask = ((u64)root->stripesize - 1);
4070 u64 ret = (val + mask) & ~mask;
4071 return ret;
4072 }
4073
4074 /*
4075 * when we wait for progress in the block group caching, its because
4076 * our allocation attempt failed at least once. So, we must sleep
4077 * and let some progress happen before we try again.
4078 *
4079 * This function will sleep at least once waiting for new free space to
4080 * show up, and then it will check the block group free space numbers
4081 * for our min num_bytes. Another option is to have it go ahead
4082 * and look in the rbtree for a free extent of a given size, but this
4083 * is a good start.
4084 */
4085 static noinline int
4086 wait_block_group_cache_progress(struct btrfs_block_group_cache *cache,
4087 u64 num_bytes)
4088 {
4089 struct btrfs_caching_control *caching_ctl;
4090 DEFINE_WAIT(wait);
4091
4092 caching_ctl = get_caching_control(cache);
4093 if (!caching_ctl)
4094 return 0;
4095
4096 wait_event(caching_ctl->wait, block_group_cache_done(cache) ||
4097 (cache->free_space >= num_bytes));
4098
4099 put_caching_control(caching_ctl);
4100 return 0;
4101 }
4102
4103 static noinline int
4104 wait_block_group_cache_done(struct btrfs_block_group_cache *cache)
4105 {
4106 struct btrfs_caching_control *caching_ctl;
4107 DEFINE_WAIT(wait);
4108
4109 caching_ctl = get_caching_control(cache);
4110 if (!caching_ctl)
4111 return 0;
4112
4113 wait_event(caching_ctl->wait, block_group_cache_done(cache));
4114
4115 put_caching_control(caching_ctl);
4116 return 0;
4117 }
4118
4119 enum btrfs_loop_type {
4120 LOOP_FIND_IDEAL = 0,
4121 LOOP_CACHING_NOWAIT = 1,
4122 LOOP_CACHING_WAIT = 2,
4123 LOOP_ALLOC_CHUNK = 3,
4124 LOOP_NO_EMPTY_SIZE = 4,
4125 };
4126
4127 /*
4128 * walks the btree of allocated extents and find a hole of a given size.
4129 * The key ins is changed to record the hole:
4130 * ins->objectid == block start
4131 * ins->flags = BTRFS_EXTENT_ITEM_KEY
4132 * ins->offset == number of blocks
4133 * Any available blocks before search_start are skipped.
4134 */
4135 static noinline int find_free_extent(struct btrfs_trans_handle *trans,
4136 struct btrfs_root *orig_root,
4137 u64 num_bytes, u64 empty_size,
4138 u64 search_start, u64 search_end,
4139 u64 hint_byte, struct btrfs_key *ins,
4140 u64 exclude_start, u64 exclude_nr,
4141 int data)
4142 {
4143 int ret = 0;
4144 struct btrfs_root *root = orig_root->fs_info->extent_root;
4145 struct btrfs_free_cluster *last_ptr = NULL;
4146 struct btrfs_block_group_cache *block_group = NULL;
4147 int empty_cluster = 2 * 1024 * 1024;
4148 int allowed_chunk_alloc = 0;
4149 int done_chunk_alloc = 0;
4150 struct btrfs_space_info *space_info;
4151 int last_ptr_loop = 0;
4152 int loop = 0;
4153 bool found_uncached_bg = false;
4154 bool failed_cluster_refill = false;
4155 bool failed_alloc = false;
4156 u64 ideal_cache_percent = 0;
4157 u64 ideal_cache_offset = 0;
4158
4159 WARN_ON(num_bytes < root->sectorsize);
4160 btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
4161 ins->objectid = 0;
4162 ins->offset = 0;
4163
4164 space_info = __find_space_info(root->fs_info, data);
4165
4166 if (orig_root->ref_cows || empty_size)
4167 allowed_chunk_alloc = 1;
4168
4169 if (data & BTRFS_BLOCK_GROUP_METADATA) {
4170 last_ptr = &root->fs_info->meta_alloc_cluster;
4171 if (!btrfs_test_opt(root, SSD))
4172 empty_cluster = 64 * 1024;
4173 }
4174
4175 if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD)) {
4176 last_ptr = &root->fs_info->data_alloc_cluster;
4177 }
4178
4179 if (last_ptr) {
4180 spin_lock(&last_ptr->lock);
4181 if (last_ptr->block_group)
4182 hint_byte = last_ptr->window_start;
4183 spin_unlock(&last_ptr->lock);
4184 }
4185
4186 search_start = max(search_start, first_logical_byte(root, 0));
4187 search_start = max(search_start, hint_byte);
4188
4189 if (!last_ptr)
4190 empty_cluster = 0;
4191
4192 if (search_start == hint_byte) {
4193 ideal_cache:
4194 block_group = btrfs_lookup_block_group(root->fs_info,
4195 search_start);
4196 /*
4197 * we don't want to use the block group if it doesn't match our
4198 * allocation bits, or if its not cached.
4199 *
4200 * However if we are re-searching with an ideal block group
4201 * picked out then we don't care that the block group is cached.
4202 */
4203 if (block_group && block_group_bits(block_group, data) &&
4204 (block_group->cached != BTRFS_CACHE_NO ||
4205 search_start == ideal_cache_offset)) {
4206 down_read(&space_info->groups_sem);
4207 if (list_empty(&block_group->list) ||
4208 block_group->ro) {
4209 /*
4210 * someone is removing this block group,
4211 * we can't jump into the have_block_group
4212 * target because our list pointers are not
4213 * valid
4214 */
4215 btrfs_put_block_group(block_group);
4216 up_read(&space_info->groups_sem);
4217 } else {
4218 goto have_block_group;
4219 }
4220 } else if (block_group) {
4221 btrfs_put_block_group(block_group);
4222 }
4223 }
4224 search:
4225 down_read(&space_info->groups_sem);
4226 list_for_each_entry(block_group, &space_info->block_groups, list) {
4227 u64 offset;
4228 int cached;
4229
4230 atomic_inc(&block_group->count);
4231 search_start = block_group->key.objectid;
4232
4233 have_block_group:
4234 if (unlikely(block_group->cached == BTRFS_CACHE_NO)) {
4235 u64 free_percent;
4236
4237 free_percent = btrfs_block_group_used(&block_group->item);
4238 free_percent *= 100;
4239 free_percent = div64_u64(free_percent,
4240 block_group->key.offset);
4241 free_percent = 100 - free_percent;
4242 if (free_percent > ideal_cache_percent &&
4243 likely(!block_group->ro)) {
4244 ideal_cache_offset = block_group->key.objectid;
4245 ideal_cache_percent = free_percent;
4246 }
4247
4248 /*
4249 * We only want to start kthread caching if we are at
4250 * the point where we will wait for caching to make
4251 * progress, or if our ideal search is over and we've
4252 * found somebody to start caching.
4253 */
4254 if (loop > LOOP_CACHING_NOWAIT ||
4255 (loop > LOOP_FIND_IDEAL &&
4256 atomic_read(&space_info->caching_threads) < 2)) {
4257 ret = cache_block_group(block_group);
4258 BUG_ON(ret);
4259 }
4260 found_uncached_bg = true;
4261
4262 /*
4263 * If loop is set for cached only, try the next block
4264 * group.
4265 */
4266 if (loop == LOOP_FIND_IDEAL)
4267 goto loop;
4268 }
4269
4270 cached = block_group_cache_done(block_group);
4271 if (unlikely(!cached))
4272 found_uncached_bg = true;
4273
4274 if (unlikely(block_group->ro))
4275 goto loop;
4276
4277 /*
4278 * Ok we want to try and use the cluster allocator, so lets look
4279 * there, unless we are on LOOP_NO_EMPTY_SIZE, since we will
4280 * have tried the cluster allocator plenty of times at this
4281 * point and not have found anything, so we are likely way too
4282 * fragmented for the clustering stuff to find anything, so lets
4283 * just skip it and let the allocator find whatever block it can
4284 * find
4285 */
4286 if (last_ptr && loop < LOOP_NO_EMPTY_SIZE) {
4287 /*
4288 * the refill lock keeps out other
4289 * people trying to start a new cluster
4290 */
4291 spin_lock(&last_ptr->refill_lock);
4292 if (last_ptr->block_group &&
4293 (last_ptr->block_group->ro ||
4294 !block_group_bits(last_ptr->block_group, data))) {
4295 offset = 0;
4296 goto refill_cluster;
4297 }
4298
4299 offset = btrfs_alloc_from_cluster(block_group, last_ptr,
4300 num_bytes, search_start);
4301 if (offset) {
4302 /* we have a block, we're done */
4303 spin_unlock(&last_ptr->refill_lock);
4304 goto checks;
4305 }
4306
4307 spin_lock(&last_ptr->lock);
4308 /*
4309 * whoops, this cluster doesn't actually point to
4310 * this block group. Get a ref on the block
4311 * group is does point to and try again
4312 */
4313 if (!last_ptr_loop && last_ptr->block_group &&
4314 last_ptr->block_group != block_group) {
4315
4316 btrfs_put_block_group(block_group);
4317 block_group = last_ptr->block_group;
4318 atomic_inc(&block_group->count);
4319 spin_unlock(&last_ptr->lock);
4320 spin_unlock(&last_ptr->refill_lock);
4321
4322 last_ptr_loop = 1;
4323 search_start = block_group->key.objectid;
4324 /*
4325 * we know this block group is properly
4326 * in the list because
4327 * btrfs_remove_block_group, drops the
4328 * cluster before it removes the block
4329 * group from the list
4330 */
4331 goto have_block_group;
4332 }
4333 spin_unlock(&last_ptr->lock);
4334 refill_cluster:
4335 /*
4336 * this cluster didn't work out, free it and
4337 * start over
4338 */
4339 btrfs_return_cluster_to_free_space(NULL, last_ptr);
4340
4341 last_ptr_loop = 0;
4342
4343 /* allocate a cluster in this block group */
4344 ret = btrfs_find_space_cluster(trans, root,
4345 block_group, last_ptr,
4346 offset, num_bytes,
4347 empty_cluster + empty_size);
4348 if (ret == 0) {
4349 /*
4350 * now pull our allocation out of this
4351 * cluster
4352 */
4353 offset = btrfs_alloc_from_cluster(block_group,
4354 last_ptr, num_bytes,
4355 search_start);
4356 if (offset) {
4357 /* we found one, proceed */
4358 spin_unlock(&last_ptr->refill_lock);
4359 goto checks;
4360 }
4361 } else if (!cached && loop > LOOP_CACHING_NOWAIT
4362 && !failed_cluster_refill) {
4363 spin_unlock(&last_ptr->refill_lock);
4364
4365 failed_cluster_refill = true;
4366 wait_block_group_cache_progress(block_group,
4367 num_bytes + empty_cluster + empty_size);
4368 goto have_block_group;
4369 }
4370
4371 /*
4372 * at this point we either didn't find a cluster
4373 * or we weren't able to allocate a block from our
4374 * cluster. Free the cluster we've been trying
4375 * to use, and go to the next block group
4376 */
4377 btrfs_return_cluster_to_free_space(NULL, last_ptr);
4378 spin_unlock(&last_ptr->refill_lock);
4379 goto loop;
4380 }
4381
4382 offset = btrfs_find_space_for_alloc(block_group, search_start,
4383 num_bytes, empty_size);
4384 /*
4385 * If we didn't find a chunk, and we haven't failed on this
4386 * block group before, and this block group is in the middle of
4387 * caching and we are ok with waiting, then go ahead and wait
4388 * for progress to be made, and set failed_alloc to true.
4389 *
4390 * If failed_alloc is true then we've already waited on this
4391 * block group once and should move on to the next block group.
4392 */
4393 if (!offset && !failed_alloc && !cached &&
4394 loop > LOOP_CACHING_NOWAIT) {
4395 wait_block_group_cache_progress(block_group,
4396 num_bytes + empty_size);
4397 failed_alloc = true;
4398 goto have_block_group;
4399 } else if (!offset) {
4400 goto loop;
4401 }
4402 checks:
4403 search_start = stripe_align(root, offset);
4404 /* move on to the next group */
4405 if (search_start + num_bytes >= search_end) {
4406 btrfs_add_free_space(block_group, offset, num_bytes);
4407 goto loop;
4408 }
4409
4410 /* move on to the next group */
4411 if (search_start + num_bytes >
4412 block_group->key.objectid + block_group->key.offset) {
4413 btrfs_add_free_space(block_group, offset, num_bytes);
4414 goto loop;
4415 }
4416
4417 if (exclude_nr > 0 &&
4418 (search_start + num_bytes > exclude_start &&
4419 search_start < exclude_start + exclude_nr)) {
4420 search_start = exclude_start + exclude_nr;
4421
4422 btrfs_add_free_space(block_group, offset, num_bytes);
4423 /*
4424 * if search_start is still in this block group
4425 * then we just re-search this block group
4426 */
4427 if (search_start >= block_group->key.objectid &&
4428 search_start < (block_group->key.objectid +
4429 block_group->key.offset))
4430 goto have_block_group;
4431 goto loop;
4432 }
4433
4434 ins->objectid = search_start;
4435 ins->offset = num_bytes;
4436
4437 if (offset < search_start)
4438 btrfs_add_free_space(block_group, offset,
4439 search_start - offset);
4440 BUG_ON(offset > search_start);
4441
4442 update_reserved_extents(block_group, num_bytes, 1);
4443
4444 /* we are all good, lets return */
4445 break;
4446 loop:
4447 failed_cluster_refill = false;
4448 failed_alloc = false;
4449 btrfs_put_block_group(block_group);
4450 }
4451 up_read(&space_info->groups_sem);
4452
4453 /* LOOP_FIND_IDEAL, only search caching/cached bg's, and don't wait for
4454 * for them to make caching progress. Also
4455 * determine the best possible bg to cache
4456 * LOOP_CACHING_NOWAIT, search partially cached block groups, kicking
4457 * caching kthreads as we move along
4458 * LOOP_CACHING_WAIT, search everything, and wait if our bg is caching
4459 * LOOP_ALLOC_CHUNK, force a chunk allocation and try again
4460 * LOOP_NO_EMPTY_SIZE, set empty_size and empty_cluster to 0 and try
4461 * again
4462 */
4463 if (!ins->objectid && loop < LOOP_NO_EMPTY_SIZE &&
4464 (found_uncached_bg || empty_size || empty_cluster ||
4465 allowed_chunk_alloc)) {
4466 if (loop == LOOP_FIND_IDEAL && found_uncached_bg) {
4467 found_uncached_bg = false;
4468 loop++;
4469 if (!ideal_cache_percent &&
4470 atomic_read(&space_info->caching_threads))
4471 goto search;
4472
4473 /*
4474 * 1 of the following 2 things have happened so far
4475 *
4476 * 1) We found an ideal block group for caching that
4477 * is mostly full and will cache quickly, so we might
4478 * as well wait for it.
4479 *
4480 * 2) We searched for cached only and we didn't find
4481 * anything, and we didn't start any caching kthreads
4482 * either, so chances are we will loop through and
4483 * start a couple caching kthreads, and then come back
4484 * around and just wait for them. This will be slower
4485 * because we will have 2 caching kthreads reading at
4486 * the same time when we could have just started one
4487 * and waited for it to get far enough to give us an
4488 * allocation, so go ahead and go to the wait caching
4489 * loop.
4490 */
4491 loop = LOOP_CACHING_WAIT;
4492 search_start = ideal_cache_offset;
4493 ideal_cache_percent = 0;
4494 goto ideal_cache;
4495 } else if (loop == LOOP_FIND_IDEAL) {
4496 /*
4497 * Didn't find a uncached bg, wait on anything we find
4498 * next.
4499 */
4500 loop = LOOP_CACHING_WAIT;
4501 goto search;
4502 }
4503
4504 if (loop < LOOP_CACHING_WAIT) {
4505 loop++;
4506 goto search;
4507 }
4508
4509 if (loop == LOOP_ALLOC_CHUNK) {
4510 empty_size = 0;
4511 empty_cluster = 0;
4512 }
4513
4514 if (allowed_chunk_alloc) {
4515 ret = do_chunk_alloc(trans, root, num_bytes +
4516 2 * 1024 * 1024, data, 1);
4517 allowed_chunk_alloc = 0;
4518 done_chunk_alloc = 1;
4519 } else if (!done_chunk_alloc) {
4520 space_info->force_alloc = 1;
4521 }
4522
4523 if (loop < LOOP_NO_EMPTY_SIZE) {
4524 loop++;
4525 goto search;
4526 }
4527 ret = -ENOSPC;
4528 } else if (!ins->objectid) {
4529 ret = -ENOSPC;
4530 }
4531
4532 /* we found what we needed */
4533 if (ins->objectid) {
4534 if (!(data & BTRFS_BLOCK_GROUP_DATA))
4535 trans->block_group = block_group->key.objectid;
4536
4537 btrfs_put_block_group(block_group);
4538 ret = 0;
4539 }
4540
4541 return ret;
4542 }
4543
4544 static void dump_space_info(struct btrfs_space_info *info, u64 bytes,
4545 int dump_block_groups)
4546 {
4547 struct btrfs_block_group_cache *cache;
4548
4549 spin_lock(&info->lock);
4550 printk(KERN_INFO "space_info has %llu free, is %sfull\n",
4551 (unsigned long long)(info->total_bytes - info->bytes_used -
4552 info->bytes_pinned - info->bytes_reserved -
4553 info->bytes_super),
4554 (info->full) ? "" : "not ");
4555 printk(KERN_INFO "space_info total=%llu, pinned=%llu, delalloc=%llu,"
4556 " may_use=%llu, used=%llu, root=%llu, super=%llu, reserved=%llu"
4557 "\n",
4558 (unsigned long long)info->total_bytes,
4559 (unsigned long long)info->bytes_pinned,
4560 (unsigned long long)info->bytes_delalloc,
4561 (unsigned long long)info->bytes_may_use,
4562 (unsigned long long)info->bytes_used,
4563 (unsigned long long)info->bytes_root,
4564 (unsigned long long)info->bytes_super,
4565 (unsigned long long)info->bytes_reserved);
4566 spin_unlock(&info->lock);
4567
4568 if (!dump_block_groups)
4569 return;
4570
4571 down_read(&info->groups_sem);
4572 list_for_each_entry(cache, &info->block_groups, list) {
4573 spin_lock(&cache->lock);
4574 printk(KERN_INFO "block group %llu has %llu bytes, %llu used "
4575 "%llu pinned %llu reserved\n",
4576 (unsigned long long)cache->key.objectid,
4577 (unsigned long long)cache->key.offset,
4578 (unsigned long long)btrfs_block_group_used(&cache->item),
4579 (unsigned long long)cache->pinned,
4580 (unsigned long long)cache->reserved);
4581 btrfs_dump_free_space(cache, bytes);
4582 spin_unlock(&cache->lock);
4583 }
4584 up_read(&info->groups_sem);
4585 }
4586
4587 int btrfs_reserve_extent(struct btrfs_trans_handle *trans,
4588 struct btrfs_root *root,
4589 u64 num_bytes, u64 min_alloc_size,
4590 u64 empty_size, u64 hint_byte,
4591 u64 search_end, struct btrfs_key *ins,
4592 u64 data)
4593 {
4594 int ret;
4595 u64 search_start = 0;
4596
4597 data = btrfs_get_alloc_profile(root, data);
4598 again:
4599 /*
4600 * the only place that sets empty_size is btrfs_realloc_node, which
4601 * is not called recursively on allocations
4602 */
4603 if (empty_size || root->ref_cows)
4604 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
4605 num_bytes + 2 * 1024 * 1024, data, 0);
4606
4607 WARN_ON(num_bytes < root->sectorsize);
4608 ret = find_free_extent(trans, root, num_bytes, empty_size,
4609 search_start, search_end, hint_byte, ins,
4610 trans->alloc_exclude_start,
4611 trans->alloc_exclude_nr, data);
4612
4613 if (ret == -ENOSPC && num_bytes > min_alloc_size) {
4614 num_bytes = num_bytes >> 1;
4615 num_bytes = num_bytes & ~(root->sectorsize - 1);
4616 num_bytes = max(num_bytes, min_alloc_size);
4617 do_chunk_alloc(trans, root->fs_info->extent_root,
4618 num_bytes, data, 1);
4619 goto again;
4620 }
4621 if (ret == -ENOSPC) {
4622 struct btrfs_space_info *sinfo;
4623
4624 sinfo = __find_space_info(root->fs_info, data);
4625 printk(KERN_ERR "btrfs allocation failed flags %llu, "
4626 "wanted %llu\n", (unsigned long long)data,
4627 (unsigned long long)num_bytes);
4628 dump_space_info(sinfo, num_bytes, 1);
4629 }
4630
4631 return ret;
4632 }
4633
4634 int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len)
4635 {
4636 struct btrfs_block_group_cache *cache;
4637 int ret = 0;
4638
4639 cache = btrfs_lookup_block_group(root->fs_info, start);
4640 if (!cache) {
4641 printk(KERN_ERR "Unable to find block group for %llu\n",
4642 (unsigned long long)start);
4643 return -ENOSPC;
4644 }
4645
4646 ret = btrfs_discard_extent(root, start, len);
4647
4648 btrfs_add_free_space(cache, start, len);
4649 update_reserved_extents(cache, len, 0);
4650 btrfs_put_block_group(cache);
4651
4652 return ret;
4653 }
4654
4655 static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
4656 struct btrfs_root *root,
4657 u64 parent, u64 root_objectid,
4658 u64 flags, u64 owner, u64 offset,
4659 struct btrfs_key *ins, int ref_mod)
4660 {
4661 int ret;
4662 struct btrfs_fs_info *fs_info = root->fs_info;
4663 struct btrfs_extent_item *extent_item;
4664 struct btrfs_extent_inline_ref *iref;
4665 struct btrfs_path *path;
4666 struct extent_buffer *leaf;
4667 int type;
4668 u32 size;
4669
4670 if (parent > 0)
4671 type = BTRFS_SHARED_DATA_REF_KEY;
4672 else
4673 type = BTRFS_EXTENT_DATA_REF_KEY;
4674
4675 size = sizeof(*extent_item) + btrfs_extent_inline_ref_size(type);
4676
4677 path = btrfs_alloc_path();
4678 BUG_ON(!path);
4679
4680 path->leave_spinning = 1;
4681 ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
4682 ins, size);
4683 BUG_ON(ret);
4684
4685 leaf = path->nodes[0];
4686 extent_item = btrfs_item_ptr(leaf, path->slots[0],
4687 struct btrfs_extent_item);
4688 btrfs_set_extent_refs(leaf, extent_item, ref_mod);
4689 btrfs_set_extent_generation(leaf, extent_item, trans->transid);
4690 btrfs_set_extent_flags(leaf, extent_item,
4691 flags | BTRFS_EXTENT_FLAG_DATA);
4692
4693 iref = (struct btrfs_extent_inline_ref *)(extent_item + 1);
4694 btrfs_set_extent_inline_ref_type(leaf, iref, type);
4695 if (parent > 0) {
4696 struct btrfs_shared_data_ref *ref;
4697 ref = (struct btrfs_shared_data_ref *)(iref + 1);
4698 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
4699 btrfs_set_shared_data_ref_count(leaf, ref, ref_mod);
4700 } else {
4701 struct btrfs_extent_data_ref *ref;
4702 ref = (struct btrfs_extent_data_ref *)(&iref->offset);
4703 btrfs_set_extent_data_ref_root(leaf, ref, root_objectid);
4704 btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
4705 btrfs_set_extent_data_ref_offset(leaf, ref, offset);
4706 btrfs_set_extent_data_ref_count(leaf, ref, ref_mod);
4707 }
4708
4709 btrfs_mark_buffer_dirty(path->nodes[0]);
4710 btrfs_free_path(path);
4711
4712 ret = update_block_group(trans, root, ins->objectid, ins->offset,
4713 1, 0);
4714 if (ret) {
4715 printk(KERN_ERR "btrfs update block group failed for %llu "
4716 "%llu\n", (unsigned long long)ins->objectid,
4717 (unsigned long long)ins->offset);
4718 BUG();
4719 }
4720 return ret;
4721 }
4722
4723 static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
4724 struct btrfs_root *root,
4725 u64 parent, u64 root_objectid,
4726 u64 flags, struct btrfs_disk_key *key,
4727 int level, struct btrfs_key *ins)
4728 {
4729 int ret;
4730 struct btrfs_fs_info *fs_info = root->fs_info;
4731 struct btrfs_extent_item *extent_item;
4732 struct btrfs_tree_block_info *block_info;
4733 struct btrfs_extent_inline_ref *iref;
4734 struct btrfs_path *path;
4735 struct extent_buffer *leaf;
4736 u32 size = sizeof(*extent_item) + sizeof(*block_info) + sizeof(*iref);
4737
4738 path = btrfs_alloc_path();
4739 BUG_ON(!path);
4740
4741 path->leave_spinning = 1;
4742 ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
4743 ins, size);
4744 BUG_ON(ret);
4745
4746 leaf = path->nodes[0];
4747 extent_item = btrfs_item_ptr(leaf, path->slots[0],
4748 struct btrfs_extent_item);
4749 btrfs_set_extent_refs(leaf, extent_item, 1);
4750 btrfs_set_extent_generation(leaf, extent_item, trans->transid);
4751 btrfs_set_extent_flags(leaf, extent_item,
4752 flags | BTRFS_EXTENT_FLAG_TREE_BLOCK);
4753 block_info = (struct btrfs_tree_block_info *)(extent_item + 1);
4754
4755 btrfs_set_tree_block_key(leaf, block_info, key);
4756 btrfs_set_tree_block_level(leaf, block_info, level);
4757
4758 iref = (struct btrfs_extent_inline_ref *)(block_info + 1);
4759 if (parent > 0) {
4760 BUG_ON(!(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF));
4761 btrfs_set_extent_inline_ref_type(leaf, iref,
4762 BTRFS_SHARED_BLOCK_REF_KEY);
4763 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
4764 } else {
4765 btrfs_set_extent_inline_ref_type(leaf, iref,
4766 BTRFS_TREE_BLOCK_REF_KEY);
4767 btrfs_set_extent_inline_ref_offset(leaf, iref, root_objectid);
4768 }
4769
4770 btrfs_mark_buffer_dirty(leaf);
4771 btrfs_free_path(path);
4772
4773 ret = update_block_group(trans, root, ins->objectid, ins->offset,
4774 1, 0);
4775 if (ret) {
4776 printk(KERN_ERR "btrfs update block group failed for %llu "
4777 "%llu\n", (unsigned long long)ins->objectid,
4778 (unsigned long long)ins->offset);
4779 BUG();
4780 }
4781 return ret;
4782 }
4783
4784 int btrfs_alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
4785 struct btrfs_root *root,
4786 u64 root_objectid, u64 owner,
4787 u64 offset, struct btrfs_key *ins)
4788 {
4789 int ret;
4790
4791 BUG_ON(root_objectid == BTRFS_TREE_LOG_OBJECTID);
4792
4793 ret = btrfs_add_delayed_data_ref(trans, ins->objectid, ins->offset,
4794 0, root_objectid, owner, offset,
4795 BTRFS_ADD_DELAYED_EXTENT, NULL);
4796 return ret;
4797 }
4798
4799 /*
4800 * this is used by the tree logging recovery code. It records that
4801 * an extent has been allocated and makes sure to clear the free
4802 * space cache bits as well
4803 */
4804 int btrfs_alloc_logged_file_extent(struct btrfs_trans_handle *trans,
4805 struct btrfs_root *root,
4806 u64 root_objectid, u64 owner, u64 offset,
4807 struct btrfs_key *ins)
4808 {
4809 int ret;
4810 struct btrfs_block_group_cache *block_group;
4811 struct btrfs_caching_control *caching_ctl;
4812 u64 start = ins->objectid;
4813 u64 num_bytes = ins->offset;
4814
4815 block_group = btrfs_lookup_block_group(root->fs_info, ins->objectid);
4816 cache_block_group(block_group);
4817 caching_ctl = get_caching_control(block_group);
4818
4819 if (!caching_ctl) {
4820 BUG_ON(!block_group_cache_done(block_group));
4821 ret = btrfs_remove_free_space(block_group, start, num_bytes);
4822 BUG_ON(ret);
4823 } else {
4824 mutex_lock(&caching_ctl->mutex);
4825
4826 if (start >= caching_ctl->progress) {
4827 ret = add_excluded_extent(root, start, num_bytes);
4828 BUG_ON(ret);
4829 } else if (start + num_bytes <= caching_ctl->progress) {
4830 ret = btrfs_remove_free_space(block_group,
4831 start, num_bytes);
4832 BUG_ON(ret);
4833 } else {
4834 num_bytes = caching_ctl->progress - start;
4835 ret = btrfs_remove_free_space(block_group,
4836 start, num_bytes);
4837 BUG_ON(ret);
4838
4839 start = caching_ctl->progress;
4840 num_bytes = ins->objectid + ins->offset -
4841 caching_ctl->progress;
4842 ret = add_excluded_extent(root, start, num_bytes);
4843 BUG_ON(ret);
4844 }
4845
4846 mutex_unlock(&caching_ctl->mutex);
4847 put_caching_control(caching_ctl);
4848 }
4849
4850 update_reserved_extents(block_group, ins->offset, 1);
4851 btrfs_put_block_group(block_group);
4852 ret = alloc_reserved_file_extent(trans, root, 0, root_objectid,
4853 0, owner, offset, ins, 1);
4854 return ret;
4855 }
4856
4857 /*
4858 * finds a free extent and does all the dirty work required for allocation
4859 * returns the key for the extent through ins, and a tree buffer for
4860 * the first block of the extent through buf.
4861 *
4862 * returns 0 if everything worked, non-zero otherwise.
4863 */
4864 static int alloc_tree_block(struct btrfs_trans_handle *trans,
4865 struct btrfs_root *root,
4866 u64 num_bytes, u64 parent, u64 root_objectid,
4867 struct btrfs_disk_key *key, int level,
4868 u64 empty_size, u64 hint_byte, u64 search_end,
4869 struct btrfs_key *ins)
4870 {
4871 int ret;
4872 u64 flags = 0;
4873
4874 ret = btrfs_reserve_extent(trans, root, num_bytes, num_bytes,
4875 empty_size, hint_byte, search_end,
4876 ins, 0);
4877 if (ret)
4878 return ret;
4879
4880 if (root_objectid == BTRFS_TREE_RELOC_OBJECTID) {
4881 if (parent == 0)
4882 parent = ins->objectid;
4883 flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF;
4884 } else
4885 BUG_ON(parent > 0);
4886
4887 if (root_objectid != BTRFS_TREE_LOG_OBJECTID) {
4888 struct btrfs_delayed_extent_op *extent_op;
4889 extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
4890 BUG_ON(!extent_op);
4891 if (key)
4892 memcpy(&extent_op->key, key, sizeof(extent_op->key));
4893 else
4894 memset(&extent_op->key, 0, sizeof(extent_op->key));
4895 extent_op->flags_to_set = flags;
4896 extent_op->update_key = 1;
4897 extent_op->update_flags = 1;
4898 extent_op->is_data = 0;
4899
4900 ret = btrfs_add_delayed_tree_ref(trans, ins->objectid,
4901 ins->offset, parent, root_objectid,
4902 level, BTRFS_ADD_DELAYED_EXTENT,
4903 extent_op);
4904 BUG_ON(ret);
4905 }
4906
4907 if (root_objectid == root->root_key.objectid) {
4908 u64 used;
4909 spin_lock(&root->node_lock);
4910 used = btrfs_root_used(&root->root_item) + num_bytes;
4911 btrfs_set_root_used(&root->root_item, used);
4912 spin_unlock(&root->node_lock);
4913 }
4914 return ret;
4915 }
4916
4917 struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans,
4918 struct btrfs_root *root,
4919 u64 bytenr, u32 blocksize,
4920 int level)
4921 {
4922 struct extent_buffer *buf;
4923
4924 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
4925 if (!buf)
4926 return ERR_PTR(-ENOMEM);
4927 btrfs_set_header_generation(buf, trans->transid);
4928 btrfs_set_buffer_lockdep_class(buf, level);
4929 btrfs_tree_lock(buf);
4930 clean_tree_block(trans, root, buf);
4931
4932 btrfs_set_lock_blocking(buf);
4933 btrfs_set_buffer_uptodate(buf);
4934
4935 if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) {
4936 /*
4937 * we allow two log transactions at a time, use different
4938 * EXENT bit to differentiate dirty pages.
4939 */
4940 if (root->log_transid % 2 == 0)
4941 set_extent_dirty(&root->dirty_log_pages, buf->start,
4942 buf->start + buf->len - 1, GFP_NOFS);
4943 else
4944 set_extent_new(&root->dirty_log_pages, buf->start,
4945 buf->start + buf->len - 1, GFP_NOFS);
4946 } else {
4947 set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
4948 buf->start + buf->len - 1, GFP_NOFS);
4949 }
4950 trans->blocks_used++;
4951 /* this returns a buffer locked for blocking */
4952 return buf;
4953 }
4954
4955 /*
4956 * helper function to allocate a block for a given tree
4957 * returns the tree buffer or NULL.
4958 */
4959 struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
4960 struct btrfs_root *root, u32 blocksize,
4961 u64 parent, u64 root_objectid,
4962 struct btrfs_disk_key *key, int level,
4963 u64 hint, u64 empty_size)
4964 {
4965 struct btrfs_key ins;
4966 int ret;
4967 struct extent_buffer *buf;
4968
4969 ret = alloc_tree_block(trans, root, blocksize, parent, root_objectid,
4970 key, level, empty_size, hint, (u64)-1, &ins);
4971 if (ret) {
4972 BUG_ON(ret > 0);
4973 return ERR_PTR(ret);
4974 }
4975
4976 buf = btrfs_init_new_buffer(trans, root, ins.objectid,
4977 blocksize, level);
4978 return buf;
4979 }
4980
4981 struct walk_control {
4982 u64 refs[BTRFS_MAX_LEVEL];
4983 u64 flags[BTRFS_MAX_LEVEL];
4984 struct btrfs_key update_progress;
4985 int stage;
4986 int level;
4987 int shared_level;
4988 int update_ref;
4989 int keep_locks;
4990 int reada_slot;
4991 int reada_count;
4992 };
4993
4994 #define DROP_REFERENCE 1
4995 #define UPDATE_BACKREF 2
4996
4997 static noinline void reada_walk_down(struct btrfs_trans_handle *trans,
4998 struct btrfs_root *root,
4999 struct walk_control *wc,
5000 struct btrfs_path *path)
5001 {
5002 u64 bytenr;
5003 u64 generation;
5004 u64 refs;
5005 u64 flags;
5006 u64 last = 0;
5007 u32 nritems;
5008 u32 blocksize;
5009 struct btrfs_key key;
5010 struct extent_buffer *eb;
5011 int ret;
5012 int slot;
5013 int nread = 0;
5014
5015 if (path->slots[wc->level] < wc->reada_slot) {
5016 wc->reada_count = wc->reada_count * 2 / 3;
5017 wc->reada_count = max(wc->reada_count, 2);
5018 } else {
5019 wc->reada_count = wc->reada_count * 3 / 2;
5020 wc->reada_count = min_t(int, wc->reada_count,
5021 BTRFS_NODEPTRS_PER_BLOCK(root));
5022 }
5023
5024 eb = path->nodes[wc->level];
5025 nritems = btrfs_header_nritems(eb);
5026 blocksize = btrfs_level_size(root, wc->level - 1);
5027
5028 for (slot = path->slots[wc->level]; slot < nritems; slot++) {
5029 if (nread >= wc->reada_count)
5030 break;
5031
5032 cond_resched();
5033 bytenr = btrfs_node_blockptr(eb, slot);
5034 generation = btrfs_node_ptr_generation(eb, slot);
5035
5036 if (slot == path->slots[wc->level])
5037 goto reada;
5038
5039 if (wc->stage == UPDATE_BACKREF &&
5040 generation <= root->root_key.offset)
5041 continue;
5042
5043 /* We don't lock the tree block, it's OK to be racy here */
5044 ret = btrfs_lookup_extent_info(trans, root, bytenr, blocksize,
5045 &refs, &flags);
5046 BUG_ON(ret);
5047 BUG_ON(refs == 0);
5048
5049 if (wc->stage == DROP_REFERENCE) {
5050 if (refs == 1)
5051 goto reada;
5052
5053 if (wc->level == 1 &&
5054 (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
5055 continue;
5056 if (!wc->update_ref ||
5057 generation <= root->root_key.offset)
5058 continue;
5059 btrfs_node_key_to_cpu(eb, &key, slot);
5060 ret = btrfs_comp_cpu_keys(&key,
5061 &wc->update_progress);
5062 if (ret < 0)
5063 continue;
5064 } else {
5065 if (wc->level == 1 &&
5066 (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
5067 continue;
5068 }
5069 reada:
5070 ret = readahead_tree_block(root, bytenr, blocksize,
5071 generation);
5072 if (ret)
5073 break;
5074 last = bytenr + blocksize;
5075 nread++;
5076 }
5077 wc->reada_slot = slot;
5078 }
5079
5080 /*
5081 * hepler to process tree block while walking down the tree.
5082 *
5083 * when wc->stage == UPDATE_BACKREF, this function updates
5084 * back refs for pointers in the block.
5085 *
5086 * NOTE: return value 1 means we should stop walking down.
5087 */
5088 static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
5089 struct btrfs_root *root,
5090 struct btrfs_path *path,
5091 struct walk_control *wc, int lookup_info)
5092 {
5093 int level = wc->level;
5094 struct extent_buffer *eb = path->nodes[level];
5095 u64 flag = BTRFS_BLOCK_FLAG_FULL_BACKREF;
5096 int ret;
5097
5098 if (wc->stage == UPDATE_BACKREF &&
5099 btrfs_header_owner(eb) != root->root_key.objectid)
5100 return 1;
5101
5102 /*
5103 * when reference count of tree block is 1, it won't increase
5104 * again. once full backref flag is set, we never clear it.
5105 */
5106 if (lookup_info &&
5107 ((wc->stage == DROP_REFERENCE && wc->refs[level] != 1) ||
5108 (wc->stage == UPDATE_BACKREF && !(wc->flags[level] & flag)))) {
5109 BUG_ON(!path->locks[level]);
5110 ret = btrfs_lookup_extent_info(trans, root,
5111 eb->start, eb->len,
5112 &wc->refs[level],
5113 &wc->flags[level]);
5114 BUG_ON(ret);
5115 BUG_ON(wc->refs[level] == 0);
5116 }
5117
5118 if (wc->stage == DROP_REFERENCE) {
5119 if (wc->refs[level] > 1)
5120 return 1;
5121
5122 if (path->locks[level] && !wc->keep_locks) {
5123 btrfs_tree_unlock(eb);
5124 path->locks[level] = 0;
5125 }
5126 return 0;
5127 }
5128
5129 /* wc->stage == UPDATE_BACKREF */
5130 if (!(wc->flags[level] & flag)) {
5131 BUG_ON(!path->locks[level]);
5132 ret = btrfs_inc_ref(trans, root, eb, 1);
5133 BUG_ON(ret);
5134 ret = btrfs_dec_ref(trans, root, eb, 0);
5135 BUG_ON(ret);
5136 ret = btrfs_set_disk_extent_flags(trans, root, eb->start,
5137 eb->len, flag, 0);
5138 BUG_ON(ret);
5139 wc->flags[level] |= flag;
5140 }
5141
5142 /*
5143 * the block is shared by multiple trees, so it's not good to
5144 * keep the tree lock
5145 */
5146 if (path->locks[level] && level > 0) {
5147 btrfs_tree_unlock(eb);
5148 path->locks[level] = 0;
5149 }
5150 return 0;
5151 }
5152
5153 /*
5154 * hepler to process tree block pointer.
5155 *
5156 * when wc->stage == DROP_REFERENCE, this function checks
5157 * reference count of the block pointed to. if the block
5158 * is shared and we need update back refs for the subtree
5159 * rooted at the block, this function changes wc->stage to
5160 * UPDATE_BACKREF. if the block is shared and there is no
5161 * need to update back, this function drops the reference
5162 * to the block.
5163 *
5164 * NOTE: return value 1 means we should stop walking down.
5165 */
5166 static noinline int do_walk_down(struct btrfs_trans_handle *trans,
5167 struct btrfs_root *root,
5168 struct btrfs_path *path,
5169 struct walk_control *wc, int *lookup_info)
5170 {
5171 u64 bytenr;
5172 u64 generation;
5173 u64 parent;
5174 u32 blocksize;
5175 struct btrfs_key key;
5176 struct extent_buffer *next;
5177 int level = wc->level;
5178 int reada = 0;
5179 int ret = 0;
5180
5181 generation = btrfs_node_ptr_generation(path->nodes[level],
5182 path->slots[level]);
5183 /*
5184 * if the lower level block was created before the snapshot
5185 * was created, we know there is no need to update back refs
5186 * for the subtree
5187 */
5188 if (wc->stage == UPDATE_BACKREF &&
5189 generation <= root->root_key.offset) {
5190 *lookup_info = 1;
5191 return 1;
5192 }
5193
5194 bytenr = btrfs_node_blockptr(path->nodes[level], path->slots[level]);
5195 blocksize = btrfs_level_size(root, level - 1);
5196
5197 next = btrfs_find_tree_block(root, bytenr, blocksize);
5198 if (!next) {
5199 next = btrfs_find_create_tree_block(root, bytenr, blocksize);
5200 reada = 1;
5201 }
5202 btrfs_tree_lock(next);
5203 btrfs_set_lock_blocking(next);
5204
5205 ret = btrfs_lookup_extent_info(trans, root, bytenr, blocksize,
5206 &wc->refs[level - 1],
5207 &wc->flags[level - 1]);
5208 BUG_ON(ret);
5209 BUG_ON(wc->refs[level - 1] == 0);
5210 *lookup_info = 0;
5211
5212 if (wc->stage == DROP_REFERENCE) {
5213 if (wc->refs[level - 1] > 1) {
5214 if (level == 1 &&
5215 (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
5216 goto skip;
5217
5218 if (!wc->update_ref ||
5219 generation <= root->root_key.offset)
5220 goto skip;
5221
5222 btrfs_node_key_to_cpu(path->nodes[level], &key,
5223 path->slots[level]);
5224 ret = btrfs_comp_cpu_keys(&key, &wc->update_progress);
5225 if (ret < 0)
5226 goto skip;
5227
5228 wc->stage = UPDATE_BACKREF;
5229 wc->shared_level = level - 1;
5230 }
5231 } else {
5232 if (level == 1 &&
5233 (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
5234 goto skip;
5235 }
5236
5237 if (!btrfs_buffer_uptodate(next, generation)) {
5238 btrfs_tree_unlock(next);
5239 free_extent_buffer(next);
5240 next = NULL;
5241 *lookup_info = 1;
5242 }
5243
5244 if (!next) {
5245 if (reada && level == 1)
5246 reada_walk_down(trans, root, wc, path);
5247 next = read_tree_block(root, bytenr, blocksize, generation);
5248 btrfs_tree_lock(next);
5249 btrfs_set_lock_blocking(next);
5250 }
5251
5252 level--;
5253 BUG_ON(level != btrfs_header_level(next));
5254 path->nodes[level] = next;
5255 path->slots[level] = 0;
5256 path->locks[level] = 1;
5257 wc->level = level;
5258 if (wc->level == 1)
5259 wc->reada_slot = 0;
5260 return 0;
5261 skip:
5262 wc->refs[level - 1] = 0;
5263 wc->flags[level - 1] = 0;
5264 if (wc->stage == DROP_REFERENCE) {
5265 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF) {
5266 parent = path->nodes[level]->start;
5267 } else {
5268 BUG_ON(root->root_key.objectid !=
5269 btrfs_header_owner(path->nodes[level]));
5270 parent = 0;
5271 }
5272
5273 ret = btrfs_free_extent(trans, root, bytenr, blocksize, parent,
5274 root->root_key.objectid, level - 1, 0);
5275 BUG_ON(ret);
5276 }
5277 btrfs_tree_unlock(next);
5278 free_extent_buffer(next);
5279 *lookup_info = 1;
5280 return 1;
5281 }
5282
5283 /*
5284 * hepler to process tree block while walking up the tree.
5285 *
5286 * when wc->stage == DROP_REFERENCE, this function drops
5287 * reference count on the block.
5288 *
5289 * when wc->stage == UPDATE_BACKREF, this function changes
5290 * wc->stage back to DROP_REFERENCE if we changed wc->stage
5291 * to UPDATE_BACKREF previously while processing the block.
5292 *
5293 * NOTE: return value 1 means we should stop walking up.
5294 */
5295 static noinline int walk_up_proc(struct btrfs_trans_handle *trans,
5296 struct btrfs_root *root,
5297 struct btrfs_path *path,
5298 struct walk_control *wc)
5299 {
5300 int ret = 0;
5301 int level = wc->level;
5302 struct extent_buffer *eb = path->nodes[level];
5303 u64 parent = 0;
5304
5305 if (wc->stage == UPDATE_BACKREF) {
5306 BUG_ON(wc->shared_level < level);
5307 if (level < wc->shared_level)
5308 goto out;
5309
5310 ret = find_next_key(path, level + 1, &wc->update_progress);
5311 if (ret > 0)
5312 wc->update_ref = 0;
5313
5314 wc->stage = DROP_REFERENCE;
5315 wc->shared_level = -1;
5316 path->slots[level] = 0;
5317
5318 /*
5319 * check reference count again if the block isn't locked.
5320 * we should start walking down the tree again if reference
5321 * count is one.
5322 */
5323 if (!path->locks[level]) {
5324 BUG_ON(level == 0);
5325 btrfs_tree_lock(eb);
5326 btrfs_set_lock_blocking(eb);
5327 path->locks[level] = 1;
5328
5329 ret = btrfs_lookup_extent_info(trans, root,
5330 eb->start, eb->len,
5331 &wc->refs[level],
5332 &wc->flags[level]);
5333 BUG_ON(ret);
5334 BUG_ON(wc->refs[level] == 0);
5335 if (wc->refs[level] == 1) {
5336 btrfs_tree_unlock(eb);
5337 path->locks[level] = 0;
5338 return 1;
5339 }
5340 }
5341 }
5342
5343 /* wc->stage == DROP_REFERENCE */
5344 BUG_ON(wc->refs[level] > 1 && !path->locks[level]);
5345
5346 if (wc->refs[level] == 1) {
5347 if (level == 0) {
5348 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
5349 ret = btrfs_dec_ref(trans, root, eb, 1);
5350 else
5351 ret = btrfs_dec_ref(trans, root, eb, 0);
5352 BUG_ON(ret);
5353 }
5354 /* make block locked assertion in clean_tree_block happy */
5355 if (!path->locks[level] &&
5356 btrfs_header_generation(eb) == trans->transid) {
5357 btrfs_tree_lock(eb);
5358 btrfs_set_lock_blocking(eb);
5359 path->locks[level] = 1;
5360 }
5361 clean_tree_block(trans, root, eb);
5362 }
5363
5364 if (eb == root->node) {
5365 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
5366 parent = eb->start;
5367 else
5368 BUG_ON(root->root_key.objectid !=
5369 btrfs_header_owner(eb));
5370 } else {
5371 if (wc->flags[level + 1] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
5372 parent = path->nodes[level + 1]->start;
5373 else
5374 BUG_ON(root->root_key.objectid !=
5375 btrfs_header_owner(path->nodes[level + 1]));
5376 }
5377
5378 ret = btrfs_free_extent(trans, root, eb->start, eb->len, parent,
5379 root->root_key.objectid, level, 0);
5380 BUG_ON(ret);
5381 out:
5382 wc->refs[level] = 0;
5383 wc->flags[level] = 0;
5384 return ret;
5385 }
5386
5387 static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
5388 struct btrfs_root *root,
5389 struct btrfs_path *path,
5390 struct walk_control *wc)
5391 {
5392 int level = wc->level;
5393 int lookup_info = 1;
5394 int ret;
5395
5396 while (level >= 0) {
5397 if (path->slots[level] >=
5398 btrfs_header_nritems(path->nodes[level]))
5399 break;
5400
5401 ret = walk_down_proc(trans, root, path, wc, lookup_info);
5402 if (ret > 0)
5403 break;
5404
5405 if (level == 0)
5406 break;
5407
5408 ret = do_walk_down(trans, root, path, wc, &lookup_info);
5409 if (ret > 0) {
5410 path->slots[level]++;
5411 continue;
5412 }
5413 level = wc->level;
5414 }
5415 return 0;
5416 }
5417
5418 static noinline int walk_up_tree(struct btrfs_trans_handle *trans,
5419 struct btrfs_root *root,
5420 struct btrfs_path *path,
5421 struct walk_control *wc, int max_level)
5422 {
5423 int level = wc->level;
5424 int ret;
5425
5426 path->slots[level] = btrfs_header_nritems(path->nodes[level]);
5427 while (level < max_level && path->nodes[level]) {
5428 wc->level = level;
5429 if (path->slots[level] + 1 <
5430 btrfs_header_nritems(path->nodes[level])) {
5431 path->slots[level]++;
5432 return 0;
5433 } else {
5434 ret = walk_up_proc(trans, root, path, wc);
5435 if (ret > 0)
5436 return 0;
5437
5438 if (path->locks[level]) {
5439 btrfs_tree_unlock(path->nodes[level]);
5440 path->locks[level] = 0;
5441 }
5442 free_extent_buffer(path->nodes[level]);
5443 path->nodes[level] = NULL;
5444 level++;
5445 }
5446 }
5447 return 1;
5448 }
5449
5450 /*
5451 * drop a subvolume tree.
5452 *
5453 * this function traverses the tree freeing any blocks that only
5454 * referenced by the tree.
5455 *
5456 * when a shared tree block is found. this function decreases its
5457 * reference count by one. if update_ref is true, this function
5458 * also make sure backrefs for the shared block and all lower level
5459 * blocks are properly updated.
5460 */
5461 int btrfs_drop_snapshot(struct btrfs_root *root, int update_ref)
5462 {
5463 struct btrfs_path *path;
5464 struct btrfs_trans_handle *trans;
5465 struct btrfs_root *tree_root = root->fs_info->tree_root;
5466 struct btrfs_root_item *root_item = &root->root_item;
5467 struct walk_control *wc;
5468 struct btrfs_key key;
5469 int err = 0;
5470 int ret;
5471 int level;
5472
5473 path = btrfs_alloc_path();
5474 BUG_ON(!path);
5475
5476 wc = kzalloc(sizeof(*wc), GFP_NOFS);
5477 BUG_ON(!wc);
5478
5479 trans = btrfs_start_transaction(tree_root, 1);
5480
5481 if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
5482 level = btrfs_header_level(root->node);
5483 path->nodes[level] = btrfs_lock_root_node(root);
5484 btrfs_set_lock_blocking(path->nodes[level]);
5485 path->slots[level] = 0;
5486 path->locks[level] = 1;
5487 memset(&wc->update_progress, 0,
5488 sizeof(wc->update_progress));
5489 } else {
5490 btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
5491 memcpy(&wc->update_progress, &key,
5492 sizeof(wc->update_progress));
5493
5494 level = root_item->drop_level;
5495 BUG_ON(level == 0);
5496 path->lowest_level = level;
5497 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
5498 path->lowest_level = 0;
5499 if (ret < 0) {
5500 err = ret;
5501 goto out;
5502 }
5503 WARN_ON(ret > 0);
5504
5505 /*
5506 * unlock our path, this is safe because only this
5507 * function is allowed to delete this snapshot
5508 */
5509 btrfs_unlock_up_safe(path, 0);
5510
5511 level = btrfs_header_level(root->node);
5512 while (1) {
5513 btrfs_tree_lock(path->nodes[level]);
5514 btrfs_set_lock_blocking(path->nodes[level]);
5515
5516 ret = btrfs_lookup_extent_info(trans, root,
5517 path->nodes[level]->start,
5518 path->nodes[level]->len,
5519 &wc->refs[level],
5520 &wc->flags[level]);
5521 BUG_ON(ret);
5522 BUG_ON(wc->refs[level] == 0);
5523
5524 if (level == root_item->drop_level)
5525 break;
5526
5527 btrfs_tree_unlock(path->nodes[level]);
5528 WARN_ON(wc->refs[level] != 1);
5529 level--;
5530 }
5531 }
5532
5533 wc->level = level;
5534 wc->shared_level = -1;
5535 wc->stage = DROP_REFERENCE;
5536 wc->update_ref = update_ref;
5537 wc->keep_locks = 0;
5538 wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(root);
5539
5540 while (1) {
5541 ret = walk_down_tree(trans, root, path, wc);
5542 if (ret < 0) {
5543 err = ret;
5544 break;
5545 }
5546
5547 ret = walk_up_tree(trans, root, path, wc, BTRFS_MAX_LEVEL);
5548 if (ret < 0) {
5549 err = ret;
5550 break;
5551 }
5552
5553 if (ret > 0) {
5554 BUG_ON(wc->stage != DROP_REFERENCE);
5555 break;
5556 }
5557
5558 if (wc->stage == DROP_REFERENCE) {
5559 level = wc->level;
5560 btrfs_node_key(path->nodes[level],
5561 &root_item->drop_progress,
5562 path->slots[level]);
5563 root_item->drop_level = level;
5564 }
5565
5566 BUG_ON(wc->level == 0);
5567 if (trans->transaction->in_commit ||
5568 trans->transaction->delayed_refs.flushing) {
5569 ret = btrfs_update_root(trans, tree_root,
5570 &root->root_key,
5571 root_item);
5572 BUG_ON(ret);
5573
5574 btrfs_end_transaction(trans, tree_root);
5575 trans = btrfs_start_transaction(tree_root, 1);
5576 } else {
5577 unsigned long update;
5578 update = trans->delayed_ref_updates;
5579 trans->delayed_ref_updates = 0;
5580 if (update)
5581 btrfs_run_delayed_refs(trans, tree_root,
5582 update);
5583 }
5584 }
5585 btrfs_release_path(root, path);
5586 BUG_ON(err);
5587
5588 ret = btrfs_del_root(trans, tree_root, &root->root_key);
5589 BUG_ON(ret);
5590
5591 if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
5592 ret = btrfs_find_last_root(tree_root, root->root_key.objectid,
5593 NULL, NULL);
5594 BUG_ON(ret < 0);
5595 if (ret > 0) {
5596 ret = btrfs_del_orphan_item(trans, tree_root,
5597 root->root_key.objectid);
5598 BUG_ON(ret);
5599 }
5600 }
5601
5602 if (root->in_radix) {
5603 btrfs_free_fs_root(tree_root->fs_info, root);
5604 } else {
5605 free_extent_buffer(root->node);
5606 free_extent_buffer(root->commit_root);
5607 kfree(root);
5608 }
5609 out:
5610 btrfs_end_transaction(trans, tree_root);
5611 kfree(wc);
5612 btrfs_free_path(path);
5613 return err;
5614 }
5615
5616 /*
5617 * drop subtree rooted at tree block 'node'.
5618 *
5619 * NOTE: this function will unlock and release tree block 'node'
5620 */
5621 int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
5622 struct btrfs_root *root,
5623 struct extent_buffer *node,
5624 struct extent_buffer *parent)
5625 {
5626 struct btrfs_path *path;
5627 struct walk_control *wc;
5628 int level;
5629 int parent_level;
5630 int ret = 0;
5631 int wret;
5632
5633 BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
5634
5635 path = btrfs_alloc_path();
5636 BUG_ON(!path);
5637
5638 wc = kzalloc(sizeof(*wc), GFP_NOFS);
5639 BUG_ON(!wc);
5640
5641 btrfs_assert_tree_locked(parent);
5642 parent_level = btrfs_header_level(parent);
5643 extent_buffer_get(parent);
5644 path->nodes[parent_level] = parent;
5645 path->slots[parent_level] = btrfs_header_nritems(parent);
5646
5647 btrfs_assert_tree_locked(node);
5648 level = btrfs_header_level(node);
5649 path->nodes[level] = node;
5650 path->slots[level] = 0;
5651 path->locks[level] = 1;
5652
5653 wc->refs[parent_level] = 1;
5654 wc->flags[parent_level] = BTRFS_BLOCK_FLAG_FULL_BACKREF;
5655 wc->level = level;
5656 wc->shared_level = -1;
5657 wc->stage = DROP_REFERENCE;
5658 wc->update_ref = 0;
5659 wc->keep_locks = 1;
5660 wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(root);
5661
5662 while (1) {
5663 wret = walk_down_tree(trans, root, path, wc);
5664 if (wret < 0) {
5665 ret = wret;
5666 break;
5667 }
5668
5669 wret = walk_up_tree(trans, root, path, wc, parent_level);
5670 if (wret < 0)
5671 ret = wret;
5672 if (wret != 0)
5673 break;
5674 }
5675
5676 kfree(wc);
5677 btrfs_free_path(path);
5678 return ret;
5679 }
5680
5681 #if 0
5682 static unsigned long calc_ra(unsigned long start, unsigned long last,
5683 unsigned long nr)
5684 {
5685 return min(last, start + nr - 1);
5686 }
5687
5688 static noinline int relocate_inode_pages(struct inode *inode, u64 start,
5689 u64 len)
5690 {
5691 u64 page_start;
5692 u64 page_end;
5693 unsigned long first_index;
5694 unsigned long last_index;
5695 unsigned long i;
5696 struct page *page;
5697 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
5698 struct file_ra_state *ra;
5699 struct btrfs_ordered_extent *ordered;
5700 unsigned int total_read = 0;
5701 unsigned int total_dirty = 0;
5702 int ret = 0;
5703
5704 ra = kzalloc(sizeof(*ra), GFP_NOFS);
5705
5706 mutex_lock(&inode->i_mutex);
5707 first_index = start >> PAGE_CACHE_SHIFT;
5708 last_index = (start + len - 1) >> PAGE_CACHE_SHIFT;
5709
5710 /* make sure the dirty trick played by the caller work */
5711 ret = invalidate_inode_pages2_range(inode->i_mapping,
5712 first_index, last_index);
5713 if (ret)
5714 goto out_unlock;
5715
5716 file_ra_state_init(ra, inode->i_mapping);
5717
5718 for (i = first_index ; i <= last_index; i++) {
5719 if (total_read % ra->ra_pages == 0) {
5720 btrfs_force_ra(inode->i_mapping, ra, NULL, i,
5721 calc_ra(i, last_index, ra->ra_pages));
5722 }
5723 total_read++;
5724 again:
5725 if (((u64)i << PAGE_CACHE_SHIFT) > i_size_read(inode))
5726 BUG_ON(1);
5727 page = grab_cache_page(inode->i_mapping, i);
5728 if (!page) {
5729 ret = -ENOMEM;
5730 goto out_unlock;
5731 }
5732 if (!PageUptodate(page)) {
5733 btrfs_readpage(NULL, page);
5734 lock_page(page);
5735 if (!PageUptodate(page)) {
5736 unlock_page(page);
5737 page_cache_release(page);
5738 ret = -EIO;
5739 goto out_unlock;
5740 }
5741 }
5742 wait_on_page_writeback(page);
5743
5744 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
5745 page_end = page_start + PAGE_CACHE_SIZE - 1;
5746 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
5747
5748 ordered = btrfs_lookup_ordered_extent(inode, page_start);
5749 if (ordered) {
5750 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
5751 unlock_page(page);
5752 page_cache_release(page);
5753 btrfs_start_ordered_extent(inode, ordered, 1);
5754 btrfs_put_ordered_extent(ordered);
5755 goto again;
5756 }
5757 set_page_extent_mapped(page);
5758
5759 if (i == first_index)
5760 set_extent_bits(io_tree, page_start, page_end,
5761 EXTENT_BOUNDARY, GFP_NOFS);
5762 btrfs_set_extent_delalloc(inode, page_start, page_end);
5763
5764 set_page_dirty(page);
5765 total_dirty++;
5766
5767 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
5768 unlock_page(page);
5769 page_cache_release(page);
5770 }
5771
5772 out_unlock:
5773 kfree(ra);
5774 mutex_unlock(&inode->i_mutex);
5775 balance_dirty_pages_ratelimited_nr(inode->i_mapping, total_dirty);
5776 return ret;
5777 }
5778
5779 static noinline int relocate_data_extent(struct inode *reloc_inode,
5780 struct btrfs_key *extent_key,
5781 u64 offset)
5782 {
5783 struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
5784 struct extent_map_tree *em_tree = &BTRFS_I(reloc_inode)->extent_tree;
5785 struct extent_map *em;
5786 u64 start = extent_key->objectid - offset;
5787 u64 end = start + extent_key->offset - 1;
5788
5789 em = alloc_extent_map(GFP_NOFS);
5790 BUG_ON(!em || IS_ERR(em));
5791
5792 em->start = start;
5793 em->len = extent_key->offset;
5794 em->block_len = extent_key->offset;
5795 em->block_start = extent_key->objectid;
5796 em->bdev = root->fs_info->fs_devices->latest_bdev;
5797 set_bit(EXTENT_FLAG_PINNED, &em->flags);
5798
5799 /* setup extent map to cheat btrfs_readpage */
5800 lock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS);
5801 while (1) {
5802 int ret;
5803 write_lock(&em_tree->lock);
5804 ret = add_extent_mapping(em_tree, em);
5805 write_unlock(&em_tree->lock);
5806 if (ret != -EEXIST) {
5807 free_extent_map(em);
5808 break;
5809 }
5810 btrfs_drop_extent_cache(reloc_inode, start, end, 0);
5811 }
5812 unlock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS);
5813
5814 return relocate_inode_pages(reloc_inode, start, extent_key->offset);
5815 }
5816
5817 struct btrfs_ref_path {
5818 u64 extent_start;
5819 u64 nodes[BTRFS_MAX_LEVEL];
5820 u64 root_objectid;
5821 u64 root_generation;
5822 u64 owner_objectid;
5823 u32 num_refs;
5824 int lowest_level;
5825 int current_level;
5826 int shared_level;
5827
5828 struct btrfs_key node_keys[BTRFS_MAX_LEVEL];
5829 u64 new_nodes[BTRFS_MAX_LEVEL];
5830 };
5831
5832 struct disk_extent {
5833 u64 ram_bytes;
5834 u64 disk_bytenr;
5835 u64 disk_num_bytes;
5836 u64 offset;
5837 u64 num_bytes;
5838 u8 compression;
5839 u8 encryption;
5840 u16 other_encoding;
5841 };
5842
5843 static int is_cowonly_root(u64 root_objectid)
5844 {
5845 if (root_objectid == BTRFS_ROOT_TREE_OBJECTID ||
5846 root_objectid == BTRFS_EXTENT_TREE_OBJECTID ||
5847 root_objectid == BTRFS_CHUNK_TREE_OBJECTID ||
5848 root_objectid == BTRFS_DEV_TREE_OBJECTID ||
5849 root_objectid == BTRFS_TREE_LOG_OBJECTID ||
5850 root_objectid == BTRFS_CSUM_TREE_OBJECTID)
5851 return 1;
5852 return 0;
5853 }
5854
5855 static noinline int __next_ref_path(struct btrfs_trans_handle *trans,
5856 struct btrfs_root *extent_root,
5857 struct btrfs_ref_path *ref_path,
5858 int first_time)
5859 {
5860 struct extent_buffer *leaf;
5861 struct btrfs_path *path;
5862 struct btrfs_extent_ref *ref;
5863 struct btrfs_key key;
5864 struct btrfs_key found_key;
5865 u64 bytenr;
5866 u32 nritems;
5867 int level;
5868 int ret = 1;
5869
5870 path = btrfs_alloc_path();
5871 if (!path)
5872 return -ENOMEM;
5873
5874 if (first_time) {
5875 ref_path->lowest_level = -1;
5876 ref_path->current_level = -1;
5877 ref_path->shared_level = -1;
5878 goto walk_up;
5879 }
5880 walk_down:
5881 level = ref_path->current_level - 1;
5882 while (level >= -1) {
5883 u64 parent;
5884 if (level < ref_path->lowest_level)
5885 break;
5886
5887 if (level >= 0)
5888 bytenr = ref_path->nodes[level];
5889 else
5890 bytenr = ref_path->extent_start;
5891 BUG_ON(bytenr == 0);
5892
5893 parent = ref_path->nodes[level + 1];
5894 ref_path->nodes[level + 1] = 0;
5895 ref_path->current_level = level;
5896 BUG_ON(parent == 0);
5897
5898 key.objectid = bytenr;
5899 key.offset = parent + 1;
5900 key.type = BTRFS_EXTENT_REF_KEY;
5901
5902 ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 0);
5903 if (ret < 0)
5904 goto out;
5905 BUG_ON(ret == 0);
5906
5907 leaf = path->nodes[0];
5908 nritems = btrfs_header_nritems(leaf);
5909 if (path->slots[0] >= nritems) {
5910 ret = btrfs_next_leaf(extent_root, path);
5911 if (ret < 0)
5912 goto out;
5913 if (ret > 0)
5914 goto next;
5915 leaf = path->nodes[0];
5916 }
5917
5918 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
5919 if (found_key.objectid == bytenr &&
5920 found_key.type == BTRFS_EXTENT_REF_KEY) {
5921 if (level < ref_path->shared_level)
5922 ref_path->shared_level = level;
5923 goto found;
5924 }
5925 next:
5926 level--;
5927 btrfs_release_path(extent_root, path);
5928 cond_resched();
5929 }
5930 /* reached lowest level */
5931 ret = 1;
5932 goto out;
5933 walk_up:
5934 level = ref_path->current_level;
5935 while (level < BTRFS_MAX_LEVEL - 1) {
5936 u64 ref_objectid;
5937
5938 if (level >= 0)
5939 bytenr = ref_path->nodes[level];
5940 else
5941 bytenr = ref_path->extent_start;
5942
5943 BUG_ON(bytenr == 0);
5944
5945 key.objectid = bytenr;
5946 key.offset = 0;
5947 key.type = BTRFS_EXTENT_REF_KEY;
5948
5949 ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 0);
5950 if (ret < 0)
5951 goto out;
5952
5953 leaf = path->nodes[0];
5954 nritems = btrfs_header_nritems(leaf);
5955 if (path->slots[0] >= nritems) {
5956 ret = btrfs_next_leaf(extent_root, path);
5957 if (ret < 0)
5958 goto out;
5959 if (ret > 0) {
5960 /* the extent was freed by someone */
5961 if (ref_path->lowest_level == level)
5962 goto out;
5963 btrfs_release_path(extent_root, path);
5964 goto walk_down;
5965 }
5966 leaf = path->nodes[0];
5967 }
5968
5969 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
5970 if (found_key.objectid != bytenr ||
5971 found_key.type != BTRFS_EXTENT_REF_KEY) {
5972 /* the extent was freed by someone */
5973 if (ref_path->lowest_level == level) {
5974 ret = 1;
5975 goto out;
5976 }
5977 btrfs_release_path(extent_root, path);
5978 goto walk_down;
5979 }
5980 found:
5981 ref = btrfs_item_ptr(leaf, path->slots[0],
5982 struct btrfs_extent_ref);
5983 ref_objectid = btrfs_ref_objectid(leaf, ref);
5984 if (ref_objectid < BTRFS_FIRST_FREE_OBJECTID) {
5985 if (first_time) {
5986 level = (int)ref_objectid;
5987 BUG_ON(level >= BTRFS_MAX_LEVEL);
5988 ref_path->lowest_level = level;
5989 ref_path->current_level = level;
5990 ref_path->nodes[level] = bytenr;
5991 } else {
5992 WARN_ON(ref_objectid != level);
5993 }
5994 } else {
5995 WARN_ON(level != -1);
5996 }
5997 first_time = 0;
5998
5999 if (ref_path->lowest_level == level) {
6000 ref_path->owner_objectid = ref_objectid;
6001 ref_path->num_refs = btrfs_ref_num_refs(leaf, ref);
6002 }
6003
6004 /*
6005 * the block is tree root or the block isn't in reference
6006 * counted tree.
6007 */
6008 if (found_key.objectid == found_key.offset ||
6009 is_cowonly_root(btrfs_ref_root(leaf, ref))) {
6010 ref_path->root_objectid = btrfs_ref_root(leaf, ref);
6011 ref_path->root_generation =
6012 btrfs_ref_generation(leaf, ref);
6013 if (level < 0) {
6014 /* special reference from the tree log */
6015 ref_path->nodes[0] = found_key.offset;
6016 ref_path->current_level = 0;
6017 }
6018 ret = 0;
6019 goto out;
6020 }
6021
6022 level++;
6023 BUG_ON(ref_path->nodes[level] != 0);
6024 ref_path->nodes[level] = found_key.offset;
6025 ref_path->current_level = level;
6026
6027 /*
6028 * the reference was created in the running transaction,
6029 * no need to continue walking up.
6030 */
6031 if (btrfs_ref_generation(leaf, ref) == trans->transid) {
6032 ref_path->root_objectid = btrfs_ref_root(leaf, ref);
6033 ref_path->root_generation =
6034 btrfs_ref_generation(leaf, ref);
6035 ret = 0;
6036 goto out;
6037 }
6038
6039 btrfs_release_path(extent_root, path);
6040 cond_resched();
6041 }
6042 /* reached max tree level, but no tree root found. */
6043 BUG();
6044 out:
6045 btrfs_free_path(path);
6046 return ret;
6047 }
6048
6049 static int btrfs_first_ref_path(struct btrfs_trans_handle *trans,
6050 struct btrfs_root *extent_root,
6051 struct btrfs_ref_path *ref_path,
6052 u64 extent_start)
6053 {
6054 memset(ref_path, 0, sizeof(*ref_path));
6055 ref_path->extent_start = extent_start;
6056
6057 return __next_ref_path(trans, extent_root, ref_path, 1);
6058 }
6059
6060 static int btrfs_next_ref_path(struct btrfs_trans_handle *trans,
6061 struct btrfs_root *extent_root,
6062 struct btrfs_ref_path *ref_path)
6063 {
6064 return __next_ref_path(trans, extent_root, ref_path, 0);
6065 }
6066
6067 static noinline int get_new_locations(struct inode *reloc_inode,
6068 struct btrfs_key *extent_key,
6069 u64 offset, int no_fragment,
6070 struct disk_extent **extents,
6071 int *nr_extents)
6072 {
6073 struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
6074 struct btrfs_path *path;
6075 struct btrfs_file_extent_item *fi;
6076 struct extent_buffer *leaf;
6077 struct disk_extent *exts = *extents;
6078 struct btrfs_key found_key;
6079 u64 cur_pos;
6080 u64 last_byte;
6081 u32 nritems;
6082 int nr = 0;
6083 int max = *nr_extents;
6084 int ret;
6085
6086 WARN_ON(!no_fragment && *extents);
6087 if (!exts) {
6088 max = 1;
6089 exts = kmalloc(sizeof(*exts) * max, GFP_NOFS);
6090 if (!exts)
6091 return -ENOMEM;
6092 }
6093
6094 path = btrfs_alloc_path();
6095 BUG_ON(!path);
6096
6097 cur_pos = extent_key->objectid - offset;
6098 last_byte = extent_key->objectid + extent_key->offset;
6099 ret = btrfs_lookup_file_extent(NULL, root, path, reloc_inode->i_ino,
6100 cur_pos, 0);
6101 if (ret < 0)
6102 goto out;
6103 if (ret > 0) {
6104 ret = -ENOENT;
6105 goto out;
6106 }
6107
6108 while (1) {
6109 leaf = path->nodes[0];
6110 nritems = btrfs_header_nritems(leaf);
6111 if (path->slots[0] >= nritems) {
6112 ret = btrfs_next_leaf(root, path);
6113 if (ret < 0)
6114 goto out;
6115 if (ret > 0)
6116 break;
6117 leaf = path->nodes[0];
6118 }
6119
6120 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
6121 if (found_key.offset != cur_pos ||
6122 found_key.type != BTRFS_EXTENT_DATA_KEY ||
6123 found_key.objectid != reloc_inode->i_ino)
6124 break;
6125
6126 fi = btrfs_item_ptr(leaf, path->slots[0],
6127 struct btrfs_file_extent_item);
6128 if (btrfs_file_extent_type(leaf, fi) !=
6129 BTRFS_FILE_EXTENT_REG ||
6130 btrfs_file_extent_disk_bytenr(leaf, fi) == 0)
6131 break;
6132
6133 if (nr == max) {
6134 struct disk_extent *old = exts;
6135 max *= 2;
6136 exts = kzalloc(sizeof(*exts) * max, GFP_NOFS);
6137 memcpy(exts, old, sizeof(*exts) * nr);
6138 if (old != *extents)
6139 kfree(old);
6140 }
6141
6142 exts[nr].disk_bytenr =
6143 btrfs_file_extent_disk_bytenr(leaf, fi);
6144 exts[nr].disk_num_bytes =
6145 btrfs_file_extent_disk_num_bytes(leaf, fi);
6146 exts[nr].offset = btrfs_file_extent_offset(leaf, fi);
6147 exts[nr].num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
6148 exts[nr].ram_bytes = btrfs_file_extent_ram_bytes(leaf, fi);
6149 exts[nr].compression = btrfs_file_extent_compression(leaf, fi);
6150 exts[nr].encryption = btrfs_file_extent_encryption(leaf, fi);
6151 exts[nr].other_encoding = btrfs_file_extent_other_encoding(leaf,
6152 fi);
6153 BUG_ON(exts[nr].offset > 0);
6154 BUG_ON(exts[nr].compression || exts[nr].encryption);
6155 BUG_ON(exts[nr].num_bytes != exts[nr].disk_num_bytes);
6156
6157 cur_pos += exts[nr].num_bytes;
6158 nr++;
6159
6160 if (cur_pos + offset >= last_byte)
6161 break;
6162
6163 if (no_fragment) {
6164 ret = 1;
6165 goto out;
6166 }
6167 path->slots[0]++;
6168 }
6169
6170 BUG_ON(cur_pos + offset > last_byte);
6171 if (cur_pos + offset < last_byte) {
6172 ret = -ENOENT;
6173 goto out;
6174 }
6175 ret = 0;
6176 out:
6177 btrfs_free_path(path);
6178 if (ret) {
6179 if (exts != *extents)
6180 kfree(exts);
6181 } else {
6182 *extents = exts;
6183 *nr_extents = nr;
6184 }
6185 return ret;
6186 }
6187
6188 static noinline int replace_one_extent(struct btrfs_trans_handle *trans,
6189 struct btrfs_root *root,
6190 struct btrfs_path *path,
6191 struct btrfs_key *extent_key,
6192 struct btrfs_key *leaf_key,
6193 struct btrfs_ref_path *ref_path,
6194 struct disk_extent *new_extents,
6195 int nr_extents)
6196 {
6197 struct extent_buffer *leaf;
6198 struct btrfs_file_extent_item *fi;
6199 struct inode *inode = NULL;
6200 struct btrfs_key key;
6201 u64 lock_start = 0;
6202 u64 lock_end = 0;
6203 u64 num_bytes;
6204 u64 ext_offset;
6205 u64 search_end = (u64)-1;
6206 u32 nritems;
6207 int nr_scaned = 0;
6208 int extent_locked = 0;
6209 int extent_type;
6210 int ret;
6211
6212 memcpy(&key, leaf_key, sizeof(key));
6213 if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS) {
6214 if (key.objectid < ref_path->owner_objectid ||
6215 (key.objectid == ref_path->owner_objectid &&
6216 key.type < BTRFS_EXTENT_DATA_KEY)) {
6217 key.objectid = ref_path->owner_objectid;
6218 key.type = BTRFS_EXTENT_DATA_KEY;
6219 key.offset = 0;
6220 }
6221 }
6222
6223 while (1) {
6224 ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
6225 if (ret < 0)
6226 goto out;
6227
6228 leaf = path->nodes[0];
6229 nritems = btrfs_header_nritems(leaf);
6230 next:
6231 if (extent_locked && ret > 0) {
6232 /*
6233 * the file extent item was modified by someone
6234 * before the extent got locked.
6235 */
6236 unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
6237 lock_end, GFP_NOFS);
6238 extent_locked = 0;
6239 }
6240
6241 if (path->slots[0] >= nritems) {
6242 if (++nr_scaned > 2)
6243 break;
6244
6245 BUG_ON(extent_locked);
6246 ret = btrfs_next_leaf(root, path);
6247 if (ret < 0)
6248 goto out;
6249 if (ret > 0)
6250 break;
6251 leaf = path->nodes[0];
6252 nritems = btrfs_header_nritems(leaf);
6253 }
6254
6255 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
6256
6257 if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS) {
6258 if ((key.objectid > ref_path->owner_objectid) ||
6259 (key.objectid == ref_path->owner_objectid &&
6260 key.type > BTRFS_EXTENT_DATA_KEY) ||
6261 key.offset >= search_end)
6262 break;
6263 }
6264
6265 if (inode && key.objectid != inode->i_ino) {
6266 BUG_ON(extent_locked);
6267 btrfs_release_path(root, path);
6268 mutex_unlock(&inode->i_mutex);
6269 iput(inode);
6270 inode = NULL;
6271 continue;
6272 }
6273
6274 if (key.type != BTRFS_EXTENT_DATA_KEY) {
6275 path->slots[0]++;
6276 ret = 1;
6277 goto next;
6278 }
6279 fi = btrfs_item_ptr(leaf, path->slots[0],
6280 struct btrfs_file_extent_item);
6281 extent_type = btrfs_file_extent_type(leaf, fi);
6282 if ((extent_type != BTRFS_FILE_EXTENT_REG &&
6283 extent_type != BTRFS_FILE_EXTENT_PREALLOC) ||
6284 (btrfs_file_extent_disk_bytenr(leaf, fi) !=
6285 extent_key->objectid)) {
6286 path->slots[0]++;
6287 ret = 1;
6288 goto next;
6289 }
6290
6291 num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
6292 ext_offset = btrfs_file_extent_offset(leaf, fi);
6293
6294 if (search_end == (u64)-1) {
6295 search_end = key.offset - ext_offset +
6296 btrfs_file_extent_ram_bytes(leaf, fi);
6297 }
6298
6299 if (!extent_locked) {
6300 lock_start = key.offset;
6301 lock_end = lock_start + num_bytes - 1;
6302 } else {
6303 if (lock_start > key.offset ||
6304 lock_end + 1 < key.offset + num_bytes) {
6305 unlock_extent(&BTRFS_I(inode)->io_tree,
6306 lock_start, lock_end, GFP_NOFS);
6307 extent_locked = 0;
6308 }
6309 }
6310
6311 if (!inode) {
6312 btrfs_release_path(root, path);
6313
6314 inode = btrfs_iget_locked(root->fs_info->sb,
6315 key.objectid, root);
6316 if (inode->i_state & I_NEW) {
6317 BTRFS_I(inode)->root = root;
6318 BTRFS_I(inode)->location.objectid =
6319 key.objectid;
6320 BTRFS_I(inode)->location.type =
6321 BTRFS_INODE_ITEM_KEY;
6322 BTRFS_I(inode)->location.offset = 0;
6323 btrfs_read_locked_inode(inode);
6324 unlock_new_inode(inode);
6325 }
6326 /*
6327 * some code call btrfs_commit_transaction while
6328 * holding the i_mutex, so we can't use mutex_lock
6329 * here.
6330 */
6331 if (is_bad_inode(inode) ||
6332 !mutex_trylock(&inode->i_mutex)) {
6333 iput(inode);
6334 inode = NULL;
6335 key.offset = (u64)-1;
6336 goto skip;
6337 }
6338 }
6339
6340 if (!extent_locked) {
6341 struct btrfs_ordered_extent *ordered;
6342
6343 btrfs_release_path(root, path);
6344
6345 lock_extent(&BTRFS_I(inode)->io_tree, lock_start,
6346 lock_end, GFP_NOFS);
6347 ordered = btrfs_lookup_first_ordered_extent(inode,
6348 lock_end);
6349 if (ordered &&
6350 ordered->file_offset <= lock_end &&
6351 ordered->file_offset + ordered->len > lock_start) {
6352 unlock_extent(&BTRFS_I(inode)->io_tree,
6353 lock_start, lock_end, GFP_NOFS);
6354 btrfs_start_ordered_extent(inode, ordered, 1);
6355 btrfs_put_ordered_extent(ordered);
6356 key.offset += num_bytes;
6357 goto skip;
6358 }
6359 if (ordered)
6360 btrfs_put_ordered_extent(ordered);
6361
6362 extent_locked = 1;
6363 continue;
6364 }
6365
6366 if (nr_extents == 1) {
6367 /* update extent pointer in place */
6368 btrfs_set_file_extent_disk_bytenr(leaf, fi,
6369 new_extents[0].disk_bytenr);
6370 btrfs_set_file_extent_disk_num_bytes(leaf, fi,
6371 new_extents[0].disk_num_bytes);
6372 btrfs_mark_buffer_dirty(leaf);
6373
6374 btrfs_drop_extent_cache(inode, key.offset,
6375 key.offset + num_bytes - 1, 0);
6376
6377 ret = btrfs_inc_extent_ref(trans, root,
6378 new_extents[0].disk_bytenr,
6379 new_extents[0].disk_num_bytes,
6380 leaf->start,
6381 root->root_key.objectid,
6382 trans->transid,
6383 key.objectid);
6384 BUG_ON(ret);
6385
6386 ret = btrfs_free_extent(trans, root,
6387 extent_key->objectid,
6388 extent_key->offset,
6389 leaf->start,
6390 btrfs_header_owner(leaf),
6391 btrfs_header_generation(leaf),
6392 key.objectid, 0);
6393 BUG_ON(ret);
6394
6395 btrfs_release_path(root, path);
6396 key.offset += num_bytes;
6397 } else {
6398 BUG_ON(1);
6399 #if 0
6400 u64 alloc_hint;
6401 u64 extent_len;
6402 int i;
6403 /*
6404 * drop old extent pointer at first, then insert the
6405 * new pointers one bye one
6406 */
6407 btrfs_release_path(root, path);
6408 ret = btrfs_drop_extents(trans, root, inode, key.offset,
6409 key.offset + num_bytes,
6410 key.offset, &alloc_hint);
6411 BUG_ON(ret);
6412
6413 for (i = 0; i < nr_extents; i++) {
6414 if (ext_offset >= new_extents[i].num_bytes) {
6415 ext_offset -= new_extents[i].num_bytes;
6416 continue;
6417 }
6418 extent_len = min(new_extents[i].num_bytes -
6419 ext_offset, num_bytes);
6420
6421 ret = btrfs_insert_empty_item(trans, root,
6422 path, &key,
6423 sizeof(*fi));
6424 BUG_ON(ret);
6425
6426 leaf = path->nodes[0];
6427 fi = btrfs_item_ptr(leaf, path->slots[0],
6428 struct btrfs_file_extent_item);
6429 btrfs_set_file_extent_generation(leaf, fi,
6430 trans->transid);
6431 btrfs_set_file_extent_type(leaf, fi,
6432 BTRFS_FILE_EXTENT_REG);
6433 btrfs_set_file_extent_disk_bytenr(leaf, fi,
6434 new_extents[i].disk_bytenr);
6435 btrfs_set_file_extent_disk_num_bytes(leaf, fi,
6436 new_extents[i].disk_num_bytes);
6437 btrfs_set_file_extent_ram_bytes(leaf, fi,
6438 new_extents[i].ram_bytes);
6439
6440 btrfs_set_file_extent_compression(leaf, fi,
6441 new_extents[i].compression);
6442 btrfs_set_file_extent_encryption(leaf, fi,
6443 new_extents[i].encryption);
6444 btrfs_set_file_extent_other_encoding(leaf, fi,
6445 new_extents[i].other_encoding);
6446
6447 btrfs_set_file_extent_num_bytes(leaf, fi,
6448 extent_len);
6449 ext_offset += new_extents[i].offset;
6450 btrfs_set_file_extent_offset(leaf, fi,
6451 ext_offset);
6452 btrfs_mark_buffer_dirty(leaf);
6453
6454 btrfs_drop_extent_cache(inode, key.offset,
6455 key.offset + extent_len - 1, 0);
6456
6457 ret = btrfs_inc_extent_ref(trans, root,
6458 new_extents[i].disk_bytenr,
6459 new_extents[i].disk_num_bytes,
6460 leaf->start,
6461 root->root_key.objectid,
6462 trans->transid, key.objectid);
6463 BUG_ON(ret);
6464 btrfs_release_path(root, path);
6465
6466 inode_add_bytes(inode, extent_len);
6467
6468 ext_offset = 0;
6469 num_bytes -= extent_len;
6470 key.offset += extent_len;
6471
6472 if (num_bytes == 0)
6473 break;
6474 }
6475 BUG_ON(i >= nr_extents);
6476 #endif
6477 }
6478
6479 if (extent_locked) {
6480 unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
6481 lock_end, GFP_NOFS);
6482 extent_locked = 0;
6483 }
6484 skip:
6485 if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS &&
6486 key.offset >= search_end)
6487 break;
6488
6489 cond_resched();
6490 }
6491 ret = 0;
6492 out:
6493 btrfs_release_path(root, path);
6494 if (inode) {
6495 mutex_unlock(&inode->i_mutex);
6496 if (extent_locked) {
6497 unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
6498 lock_end, GFP_NOFS);
6499 }
6500 iput(inode);
6501 }
6502 return ret;
6503 }
6504
6505 int btrfs_reloc_tree_cache_ref(struct btrfs_trans_handle *trans,
6506 struct btrfs_root *root,
6507 struct extent_buffer *buf, u64 orig_start)
6508 {
6509 int level;
6510 int ret;
6511
6512 BUG_ON(btrfs_header_generation(buf) != trans->transid);
6513 BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
6514
6515 level = btrfs_header_level(buf);
6516 if (level == 0) {
6517 struct btrfs_leaf_ref *ref;
6518 struct btrfs_leaf_ref *orig_ref;
6519
6520 orig_ref = btrfs_lookup_leaf_ref(root, orig_start);
6521 if (!orig_ref)
6522 return -ENOENT;
6523
6524 ref = btrfs_alloc_leaf_ref(root, orig_ref->nritems);
6525 if (!ref) {
6526 btrfs_free_leaf_ref(root, orig_ref);
6527 return -ENOMEM;
6528 }
6529
6530 ref->nritems = orig_ref->nritems;
6531 memcpy(ref->extents, orig_ref->extents,
6532 sizeof(ref->extents[0]) * ref->nritems);
6533
6534 btrfs_free_leaf_ref(root, orig_ref);
6535
6536 ref->root_gen = trans->transid;
6537 ref->bytenr = buf->start;
6538 ref->owner = btrfs_header_owner(buf);
6539 ref->generation = btrfs_header_generation(buf);
6540
6541 ret = btrfs_add_leaf_ref(root, ref, 0);
6542 WARN_ON(ret);
6543 btrfs_free_leaf_ref(root, ref);
6544 }
6545 return 0;
6546 }
6547
6548 static noinline int invalidate_extent_cache(struct btrfs_root *root,
6549 struct extent_buffer *leaf,
6550 struct btrfs_block_group_cache *group,
6551 struct btrfs_root *target_root)
6552 {
6553 struct btrfs_key key;
6554 struct inode *inode = NULL;
6555 struct btrfs_file_extent_item *fi;
6556 u64 num_bytes;
6557 u64 skip_objectid = 0;
6558 u32 nritems;
6559 u32 i;
6560
6561 nritems = btrfs_header_nritems(leaf);
6562 for (i = 0; i < nritems; i++) {
6563 btrfs_item_key_to_cpu(leaf, &key, i);
6564 if (key.objectid == skip_objectid ||
6565 key.type != BTRFS_EXTENT_DATA_KEY)
6566 continue;
6567 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
6568 if (btrfs_file_extent_type(leaf, fi) ==
6569 BTRFS_FILE_EXTENT_INLINE)
6570 continue;
6571 if (btrfs_file_extent_disk_bytenr(leaf, fi) == 0)
6572 continue;
6573 if (!inode || inode->i_ino != key.objectid) {
6574 iput(inode);
6575 inode = btrfs_ilookup(target_root->fs_info->sb,
6576 key.objectid, target_root, 1);
6577 }
6578 if (!inode) {
6579 skip_objectid = key.objectid;
6580 continue;
6581 }
6582 num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
6583
6584 lock_extent(&BTRFS_I(inode)->io_tree, key.offset,
6585 key.offset + num_bytes - 1, GFP_NOFS);
6586 btrfs_drop_extent_cache(inode, key.offset,
6587 key.offset + num_bytes - 1, 1);
6588 unlock_extent(&BTRFS_I(inode)->io_tree, key.offset,
6589 key.offset + num_bytes - 1, GFP_NOFS);
6590 cond_resched();
6591 }
6592 iput(inode);
6593 return 0;
6594 }
6595
6596 static noinline int replace_extents_in_leaf(struct btrfs_trans_handle *trans,
6597 struct btrfs_root *root,
6598 struct extent_buffer *leaf,
6599 struct btrfs_block_group_cache *group,
6600 struct inode *reloc_inode)
6601 {
6602 struct btrfs_key key;
6603 struct btrfs_key extent_key;
6604 struct btrfs_file_extent_item *fi;
6605 struct btrfs_leaf_ref *ref;
6606 struct disk_extent *new_extent;
6607 u64 bytenr;
6608 u64 num_bytes;
6609 u32 nritems;
6610 u32 i;
6611 int ext_index;
6612 int nr_extent;
6613 int ret;
6614
6615 new_extent = kmalloc(sizeof(*new_extent), GFP_NOFS);
6616 BUG_ON(!new_extent);
6617
6618 ref = btrfs_lookup_leaf_ref(root, leaf->start);
6619 BUG_ON(!ref);
6620
6621 ext_index = -1;
6622 nritems = btrfs_header_nritems(leaf);
6623 for (i = 0; i < nritems; i++) {
6624 btrfs_item_key_to_cpu(leaf, &key, i);
6625 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
6626 continue;
6627 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
6628 if (btrfs_file_extent_type(leaf, fi) ==
6629 BTRFS_FILE_EXTENT_INLINE)
6630 continue;
6631 bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
6632 num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
6633 if (bytenr == 0)
6634 continue;
6635
6636 ext_index++;
6637 if (bytenr >= group->key.objectid + group->key.offset ||
6638 bytenr + num_bytes <= group->key.objectid)
6639 continue;
6640
6641 extent_key.objectid = bytenr;
6642 extent_key.offset = num_bytes;
6643 extent_key.type = BTRFS_EXTENT_ITEM_KEY;
6644 nr_extent = 1;
6645 ret = get_new_locations(reloc_inode, &extent_key,
6646 group->key.objectid, 1,
6647 &new_extent, &nr_extent);
6648 if (ret > 0)
6649 continue;
6650 BUG_ON(ret < 0);
6651
6652 BUG_ON(ref->extents[ext_index].bytenr != bytenr);
6653 BUG_ON(ref->extents[ext_index].num_bytes != num_bytes);
6654 ref->extents[ext_index].bytenr = new_extent->disk_bytenr;
6655 ref->extents[ext_index].num_bytes = new_extent->disk_num_bytes;
6656
6657 btrfs_set_file_extent_disk_bytenr(leaf, fi,
6658 new_extent->disk_bytenr);
6659 btrfs_set_file_extent_disk_num_bytes(leaf, fi,
6660 new_extent->disk_num_bytes);
6661 btrfs_mark_buffer_dirty(leaf);
6662
6663 ret = btrfs_inc_extent_ref(trans, root,
6664 new_extent->disk_bytenr,
6665 new_extent->disk_num_bytes,
6666 leaf->start,
6667 root->root_key.objectid,
6668 trans->transid, key.objectid);
6669 BUG_ON(ret);
6670
6671 ret = btrfs_free_extent(trans, root,
6672 bytenr, num_bytes, leaf->start,
6673 btrfs_header_owner(leaf),
6674 btrfs_header_generation(leaf),
6675 key.objectid, 0);
6676 BUG_ON(ret);
6677 cond_resched();
6678 }
6679 kfree(new_extent);
6680 BUG_ON(ext_index + 1 != ref->nritems);
6681 btrfs_free_leaf_ref(root, ref);
6682 return 0;
6683 }
6684
6685 int btrfs_free_reloc_root(struct btrfs_trans_handle *trans,
6686 struct btrfs_root *root)
6687 {
6688 struct btrfs_root *reloc_root;
6689 int ret;
6690
6691 if (root->reloc_root) {
6692 reloc_root = root->reloc_root;
6693 root->reloc_root = NULL;
6694 list_add(&reloc_root->dead_list,
6695 &root->fs_info->dead_reloc_roots);
6696
6697 btrfs_set_root_bytenr(&reloc_root->root_item,
6698 reloc_root->node->start);
6699 btrfs_set_root_level(&root->root_item,
6700 btrfs_header_level(reloc_root->node));
6701 memset(&reloc_root->root_item.drop_progress, 0,
6702 sizeof(struct btrfs_disk_key));
6703 reloc_root->root_item.drop_level = 0;
6704
6705 ret = btrfs_update_root(trans, root->fs_info->tree_root,
6706 &reloc_root->root_key,
6707 &reloc_root->root_item);
6708 BUG_ON(ret);
6709 }
6710 return 0;
6711 }
6712
6713 int btrfs_drop_dead_reloc_roots(struct btrfs_root *root)
6714 {
6715 struct btrfs_trans_handle *trans;
6716 struct btrfs_root *reloc_root;
6717 struct btrfs_root *prev_root = NULL;
6718 struct list_head dead_roots;
6719 int ret;
6720 unsigned long nr;
6721
6722 INIT_LIST_HEAD(&dead_roots);
6723 list_splice_init(&root->fs_info->dead_reloc_roots, &dead_roots);
6724
6725 while (!list_empty(&dead_roots)) {
6726 reloc_root = list_entry(dead_roots.prev,
6727 struct btrfs_root, dead_list);
6728 list_del_init(&reloc_root->dead_list);
6729
6730 BUG_ON(reloc_root->commit_root != NULL);
6731 while (1) {
6732 trans = btrfs_join_transaction(root, 1);
6733 BUG_ON(!trans);
6734
6735 mutex_lock(&root->fs_info->drop_mutex);
6736 ret = btrfs_drop_snapshot(trans, reloc_root);
6737 if (ret != -EAGAIN)
6738 break;
6739 mutex_unlock(&root->fs_info->drop_mutex);
6740
6741 nr = trans->blocks_used;
6742 ret = btrfs_end_transaction(trans, root);
6743 BUG_ON(ret);
6744 btrfs_btree_balance_dirty(root, nr);
6745 }
6746
6747 free_extent_buffer(reloc_root->node);
6748
6749 ret = btrfs_del_root(trans, root->fs_info->tree_root,
6750 &reloc_root->root_key);
6751 BUG_ON(ret);
6752 mutex_unlock(&root->fs_info->drop_mutex);
6753
6754 nr = trans->blocks_used;
6755 ret = btrfs_end_transaction(trans, root);
6756 BUG_ON(ret);
6757 btrfs_btree_balance_dirty(root, nr);
6758
6759 kfree(prev_root);
6760 prev_root = reloc_root;
6761 }
6762 if (prev_root) {
6763 btrfs_remove_leaf_refs(prev_root, (u64)-1, 0);
6764 kfree(prev_root);
6765 }
6766 return 0;
6767 }
6768
6769 int btrfs_add_dead_reloc_root(struct btrfs_root *root)
6770 {
6771 list_add(&root->dead_list, &root->fs_info->dead_reloc_roots);
6772 return 0;
6773 }
6774
6775 int btrfs_cleanup_reloc_trees(struct btrfs_root *root)
6776 {
6777 struct btrfs_root *reloc_root;
6778 struct btrfs_trans_handle *trans;
6779 struct btrfs_key location;
6780 int found;
6781 int ret;
6782
6783 mutex_lock(&root->fs_info->tree_reloc_mutex);
6784 ret = btrfs_find_dead_roots(root, BTRFS_TREE_RELOC_OBJECTID, NULL);
6785 BUG_ON(ret);
6786 found = !list_empty(&root->fs_info->dead_reloc_roots);
6787 mutex_unlock(&root->fs_info->tree_reloc_mutex);
6788
6789 if (found) {
6790 trans = btrfs_start_transaction(root, 1);
6791 BUG_ON(!trans);
6792 ret = btrfs_commit_transaction(trans, root);
6793 BUG_ON(ret);
6794 }
6795
6796 location.objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
6797 location.offset = (u64)-1;
6798 location.type = BTRFS_ROOT_ITEM_KEY;
6799
6800 reloc_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
6801 BUG_ON(!reloc_root);
6802 btrfs_orphan_cleanup(reloc_root);
6803 return 0;
6804 }
6805
6806 static noinline int init_reloc_tree(struct btrfs_trans_handle *trans,
6807 struct btrfs_root *root)
6808 {
6809 struct btrfs_root *reloc_root;
6810 struct extent_buffer *eb;
6811 struct btrfs_root_item *root_item;
6812 struct btrfs_key root_key;
6813 int ret;
6814
6815 BUG_ON(!root->ref_cows);
6816 if (root->reloc_root)
6817 return 0;
6818
6819 root_item = kmalloc(sizeof(*root_item), GFP_NOFS);
6820 BUG_ON(!root_item);
6821
6822 ret = btrfs_copy_root(trans, root, root->commit_root,
6823 &eb, BTRFS_TREE_RELOC_OBJECTID);
6824 BUG_ON(ret);
6825
6826 root_key.objectid = BTRFS_TREE_RELOC_OBJECTID;
6827 root_key.offset = root->root_key.objectid;
6828 root_key.type = BTRFS_ROOT_ITEM_KEY;
6829
6830 memcpy(root_item, &root->root_item, sizeof(root_item));
6831 btrfs_set_root_refs(root_item, 0);
6832 btrfs_set_root_bytenr(root_item, eb->start);
6833 btrfs_set_root_level(root_item, btrfs_header_level(eb));
6834 btrfs_set_root_generation(root_item, trans->transid);
6835
6836 btrfs_tree_unlock(eb);
6837 free_extent_buffer(eb);
6838
6839 ret = btrfs_insert_root(trans, root->fs_info->tree_root,
6840 &root_key, root_item);
6841 BUG_ON(ret);
6842 kfree(root_item);
6843
6844 reloc_root = btrfs_read_fs_root_no_radix(root->fs_info->tree_root,
6845 &root_key);
6846 BUG_ON(!reloc_root);
6847 reloc_root->last_trans = trans->transid;
6848 reloc_root->commit_root = NULL;
6849 reloc_root->ref_tree = &root->fs_info->reloc_ref_tree;
6850
6851 root->reloc_root = reloc_root;
6852 return 0;
6853 }
6854
6855 /*
6856 * Core function of space balance.
6857 *
6858 * The idea is using reloc trees to relocate tree blocks in reference
6859 * counted roots. There is one reloc tree for each subvol, and all
6860 * reloc trees share same root key objectid. Reloc trees are snapshots
6861 * of the latest committed roots of subvols (root->commit_root).
6862 *
6863 * To relocate a tree block referenced by a subvol, there are two steps.
6864 * COW the block through subvol's reloc tree, then update block pointer
6865 * in the subvol to point to the new block. Since all reloc trees share
6866 * same root key objectid, doing special handing for tree blocks owned
6867 * by them is easy. Once a tree block has been COWed in one reloc tree,
6868 * we can use the resulting new block directly when the same block is
6869 * required to COW again through other reloc trees. By this way, relocated
6870 * tree blocks are shared between reloc trees, so they are also shared
6871 * between subvols.
6872 */
6873 static noinline int relocate_one_path(struct btrfs_trans_handle *trans,
6874 struct btrfs_root *root,
6875 struct btrfs_path *path,
6876 struct btrfs_key *first_key,
6877 struct btrfs_ref_path *ref_path,
6878 struct btrfs_block_group_cache *group,
6879 struct inode *reloc_inode)
6880 {
6881 struct btrfs_root *reloc_root;
6882 struct extent_buffer *eb = NULL;
6883 struct btrfs_key *keys;
6884 u64 *nodes;
6885 int level;
6886 int shared_level;
6887 int lowest_level = 0;
6888 int ret;
6889
6890 if (ref_path->owner_objectid < BTRFS_FIRST_FREE_OBJECTID)
6891 lowest_level = ref_path->owner_objectid;
6892
6893 if (!root->ref_cows) {
6894 path->lowest_level = lowest_level;
6895 ret = btrfs_search_slot(trans, root, first_key, path, 0, 1);
6896 BUG_ON(ret < 0);
6897 path->lowest_level = 0;
6898 btrfs_release_path(root, path);
6899 return 0;
6900 }
6901
6902 mutex_lock(&root->fs_info->tree_reloc_mutex);
6903 ret = init_reloc_tree(trans, root);
6904 BUG_ON(ret);
6905 reloc_root = root->reloc_root;
6906
6907 shared_level = ref_path->shared_level;
6908 ref_path->shared_level = BTRFS_MAX_LEVEL - 1;
6909
6910 keys = ref_path->node_keys;
6911 nodes = ref_path->new_nodes;
6912 memset(&keys[shared_level + 1], 0,
6913 sizeof(*keys) * (BTRFS_MAX_LEVEL - shared_level - 1));
6914 memset(&nodes[shared_level + 1], 0,
6915 sizeof(*nodes) * (BTRFS_MAX_LEVEL - shared_level - 1));
6916
6917 if (nodes[lowest_level] == 0) {
6918 path->lowest_level = lowest_level;
6919 ret = btrfs_search_slot(trans, reloc_root, first_key, path,
6920 0, 1);
6921 BUG_ON(ret);
6922 for (level = lowest_level; level < BTRFS_MAX_LEVEL; level++) {
6923 eb = path->nodes[level];
6924 if (!eb || eb == reloc_root->node)
6925 break;
6926 nodes[level] = eb->start;
6927 if (level == 0)
6928 btrfs_item_key_to_cpu(eb, &keys[level], 0);
6929 else
6930 btrfs_node_key_to_cpu(eb, &keys[level], 0);
6931 }
6932 if (nodes[0] &&
6933 ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
6934 eb = path->nodes[0];
6935 ret = replace_extents_in_leaf(trans, reloc_root, eb,
6936 group, reloc_inode);
6937 BUG_ON(ret);
6938 }
6939 btrfs_release_path(reloc_root, path);
6940 } else {
6941 ret = btrfs_merge_path(trans, reloc_root, keys, nodes,
6942 lowest_level);
6943 BUG_ON(ret);
6944 }
6945
6946 /*
6947 * replace tree blocks in the fs tree with tree blocks in
6948 * the reloc tree.
6949 */
6950 ret = btrfs_merge_path(trans, root, keys, nodes, lowest_level);
6951 BUG_ON(ret < 0);
6952
6953 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
6954 ret = btrfs_search_slot(trans, reloc_root, first_key, path,
6955 0, 0);
6956 BUG_ON(ret);
6957 extent_buffer_get(path->nodes[0]);
6958 eb = path->nodes[0];
6959 btrfs_release_path(reloc_root, path);
6960 ret = invalidate_extent_cache(reloc_root, eb, group, root);
6961 BUG_ON(ret);
6962 free_extent_buffer(eb);
6963 }
6964
6965 mutex_unlock(&root->fs_info->tree_reloc_mutex);
6966 path->lowest_level = 0;
6967 return 0;
6968 }
6969
6970 static noinline int relocate_tree_block(struct btrfs_trans_handle *trans,
6971 struct btrfs_root *root,
6972 struct btrfs_path *path,
6973 struct btrfs_key *first_key,
6974 struct btrfs_ref_path *ref_path)
6975 {
6976 int ret;
6977
6978 ret = relocate_one_path(trans, root, path, first_key,
6979 ref_path, NULL, NULL);
6980 BUG_ON(ret);
6981
6982 return 0;
6983 }
6984
6985 static noinline int del_extent_zero(struct btrfs_trans_handle *trans,
6986 struct btrfs_root *extent_root,
6987 struct btrfs_path *path,
6988 struct btrfs_key *extent_key)
6989 {
6990 int ret;
6991
6992 ret = btrfs_search_slot(trans, extent_root, extent_key, path, -1, 1);
6993 if (ret)
6994 goto out;
6995 ret = btrfs_del_item(trans, extent_root, path);
6996 out:
6997 btrfs_release_path(extent_root, path);
6998 return ret;
6999 }
7000
7001 static noinline struct btrfs_root *read_ref_root(struct btrfs_fs_info *fs_info,
7002 struct btrfs_ref_path *ref_path)
7003 {
7004 struct btrfs_key root_key;
7005
7006 root_key.objectid = ref_path->root_objectid;
7007 root_key.type = BTRFS_ROOT_ITEM_KEY;
7008 if (is_cowonly_root(ref_path->root_objectid))
7009 root_key.offset = 0;
7010 else
7011 root_key.offset = (u64)-1;
7012
7013 return btrfs_read_fs_root_no_name(fs_info, &root_key);
7014 }
7015
7016 static noinline int relocate_one_extent(struct btrfs_root *extent_root,
7017 struct btrfs_path *path,
7018 struct btrfs_key *extent_key,
7019 struct btrfs_block_group_cache *group,
7020 struct inode *reloc_inode, int pass)
7021 {
7022 struct btrfs_trans_handle *trans;
7023 struct btrfs_root *found_root;
7024 struct btrfs_ref_path *ref_path = NULL;
7025 struct disk_extent *new_extents = NULL;
7026 int nr_extents = 0;
7027 int loops;
7028 int ret;
7029 int level;
7030 struct btrfs_key first_key;
7031 u64 prev_block = 0;
7032
7033
7034 trans = btrfs_start_transaction(extent_root, 1);
7035 BUG_ON(!trans);
7036
7037 if (extent_key->objectid == 0) {
7038 ret = del_extent_zero(trans, extent_root, path, extent_key);
7039 goto out;
7040 }
7041
7042 ref_path = kmalloc(sizeof(*ref_path), GFP_NOFS);
7043 if (!ref_path) {
7044 ret = -ENOMEM;
7045 goto out;
7046 }
7047
7048 for (loops = 0; ; loops++) {
7049 if (loops == 0) {
7050 ret = btrfs_first_ref_path(trans, extent_root, ref_path,
7051 extent_key->objectid);
7052 } else {
7053 ret = btrfs_next_ref_path(trans, extent_root, ref_path);
7054 }
7055 if (ret < 0)
7056 goto out;
7057 if (ret > 0)
7058 break;
7059
7060 if (ref_path->root_objectid == BTRFS_TREE_LOG_OBJECTID ||
7061 ref_path->root_objectid == BTRFS_TREE_RELOC_OBJECTID)
7062 continue;
7063
7064 found_root = read_ref_root(extent_root->fs_info, ref_path);
7065 BUG_ON(!found_root);
7066 /*
7067 * for reference counted tree, only process reference paths
7068 * rooted at the latest committed root.
7069 */
7070 if (found_root->ref_cows &&
7071 ref_path->root_generation != found_root->root_key.offset)
7072 continue;
7073
7074 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
7075 if (pass == 0) {
7076 /*
7077 * copy data extents to new locations
7078 */
7079 u64 group_start = group->key.objectid;
7080 ret = relocate_data_extent(reloc_inode,
7081 extent_key,
7082 group_start);
7083 if (ret < 0)
7084 goto out;
7085 break;
7086 }
7087 level = 0;
7088 } else {
7089 level = ref_path->owner_objectid;
7090 }
7091
7092 if (prev_block != ref_path->nodes[level]) {
7093 struct extent_buffer *eb;
7094 u64 block_start = ref_path->nodes[level];
7095 u64 block_size = btrfs_level_size(found_root, level);
7096
7097 eb = read_tree_block(found_root, block_start,
7098 block_size, 0);
7099 btrfs_tree_lock(eb);
7100 BUG_ON(level != btrfs_header_level(eb));
7101
7102 if (level == 0)
7103 btrfs_item_key_to_cpu(eb, &first_key, 0);
7104 else
7105 btrfs_node_key_to_cpu(eb, &first_key, 0);
7106
7107 btrfs_tree_unlock(eb);
7108 free_extent_buffer(eb);
7109 prev_block = block_start;
7110 }
7111
7112 mutex_lock(&extent_root->fs_info->trans_mutex);
7113 btrfs_record_root_in_trans(found_root);
7114 mutex_unlock(&extent_root->fs_info->trans_mutex);
7115 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
7116 /*
7117 * try to update data extent references while
7118 * keeping metadata shared between snapshots.
7119 */
7120 if (pass == 1) {
7121 ret = relocate_one_path(trans, found_root,
7122 path, &first_key, ref_path,
7123 group, reloc_inode);
7124 if (ret < 0)
7125 goto out;
7126 continue;
7127 }
7128 /*
7129 * use fallback method to process the remaining
7130 * references.
7131 */
7132 if (!new_extents) {
7133 u64 group_start = group->key.objectid;
7134 new_extents = kmalloc(sizeof(*new_extents),
7135 GFP_NOFS);
7136 nr_extents = 1;
7137 ret = get_new_locations(reloc_inode,
7138 extent_key,
7139 group_start, 1,
7140 &new_extents,
7141 &nr_extents);
7142 if (ret)
7143 goto out;
7144 }
7145 ret = replace_one_extent(trans, found_root,
7146 path, extent_key,
7147 &first_key, ref_path,
7148 new_extents, nr_extents);
7149 } else {
7150 ret = relocate_tree_block(trans, found_root, path,
7151 &first_key, ref_path);
7152 }
7153 if (ret < 0)
7154 goto out;
7155 }
7156 ret = 0;
7157 out:
7158 btrfs_end_transaction(trans, extent_root);
7159 kfree(new_extents);
7160 kfree(ref_path);
7161 return ret;
7162 }
7163 #endif
7164
7165 static u64 update_block_group_flags(struct btrfs_root *root, u64 flags)
7166 {
7167 u64 num_devices;
7168 u64 stripped = BTRFS_BLOCK_GROUP_RAID0 |
7169 BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10;
7170
7171 num_devices = root->fs_info->fs_devices->rw_devices;
7172 if (num_devices == 1) {
7173 stripped |= BTRFS_BLOCK_GROUP_DUP;
7174 stripped = flags & ~stripped;
7175
7176 /* turn raid0 into single device chunks */
7177 if (flags & BTRFS_BLOCK_GROUP_RAID0)
7178 return stripped;
7179
7180 /* turn mirroring into duplication */
7181 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
7182 BTRFS_BLOCK_GROUP_RAID10))
7183 return stripped | BTRFS_BLOCK_GROUP_DUP;
7184 return flags;
7185 } else {
7186 /* they already had raid on here, just return */
7187 if (flags & stripped)
7188 return flags;
7189
7190 stripped |= BTRFS_BLOCK_GROUP_DUP;
7191 stripped = flags & ~stripped;
7192
7193 /* switch duplicated blocks with raid1 */
7194 if (flags & BTRFS_BLOCK_GROUP_DUP)
7195 return stripped | BTRFS_BLOCK_GROUP_RAID1;
7196
7197 /* turn single device chunks into raid0 */
7198 return stripped | BTRFS_BLOCK_GROUP_RAID0;
7199 }
7200 return flags;
7201 }
7202
7203 static int __alloc_chunk_for_shrink(struct btrfs_root *root,
7204 struct btrfs_block_group_cache *shrink_block_group,
7205 int force)
7206 {
7207 struct btrfs_trans_handle *trans;
7208 u64 new_alloc_flags;
7209 u64 calc;
7210
7211 spin_lock(&shrink_block_group->lock);
7212 if (btrfs_block_group_used(&shrink_block_group->item) +
7213 shrink_block_group->reserved > 0) {
7214 spin_unlock(&shrink_block_group->lock);
7215
7216 trans = btrfs_start_transaction(root, 1);
7217 spin_lock(&shrink_block_group->lock);
7218
7219 new_alloc_flags = update_block_group_flags(root,
7220 shrink_block_group->flags);
7221 if (new_alloc_flags != shrink_block_group->flags) {
7222 calc =
7223 btrfs_block_group_used(&shrink_block_group->item);
7224 } else {
7225 calc = shrink_block_group->key.offset;
7226 }
7227 spin_unlock(&shrink_block_group->lock);
7228
7229 do_chunk_alloc(trans, root->fs_info->extent_root,
7230 calc + 2 * 1024 * 1024, new_alloc_flags, force);
7231
7232 btrfs_end_transaction(trans, root);
7233 } else
7234 spin_unlock(&shrink_block_group->lock);
7235 return 0;
7236 }
7237
7238
7239 int btrfs_prepare_block_group_relocation(struct btrfs_root *root,
7240 struct btrfs_block_group_cache *group)
7241
7242 {
7243 __alloc_chunk_for_shrink(root, group, 1);
7244 set_block_group_readonly(group);
7245 return 0;
7246 }
7247
7248 /*
7249 * checks to see if its even possible to relocate this block group.
7250 *
7251 * @return - -1 if it's not a good idea to relocate this block group, 0 if its
7252 * ok to go ahead and try.
7253 */
7254 int btrfs_can_relocate(struct btrfs_root *root, u64 bytenr)
7255 {
7256 struct btrfs_block_group_cache *block_group;
7257 struct btrfs_space_info *space_info;
7258 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
7259 struct btrfs_device *device;
7260 int full = 0;
7261 int ret = 0;
7262
7263 block_group = btrfs_lookup_block_group(root->fs_info, bytenr);
7264
7265 /* odd, couldn't find the block group, leave it alone */
7266 if (!block_group)
7267 return -1;
7268
7269 /* no bytes used, we're good */
7270 if (!btrfs_block_group_used(&block_group->item))
7271 goto out;
7272
7273 space_info = block_group->space_info;
7274 spin_lock(&space_info->lock);
7275
7276 full = space_info->full;
7277
7278 /*
7279 * if this is the last block group we have in this space, we can't
7280 * relocate it unless we're able to allocate a new chunk below.
7281 *
7282 * Otherwise, we need to make sure we have room in the space to handle
7283 * all of the extents from this block group. If we can, we're good
7284 */
7285 if ((space_info->total_bytes != block_group->key.offset) &&
7286 (space_info->bytes_used + space_info->bytes_reserved +
7287 space_info->bytes_pinned + space_info->bytes_readonly +
7288 btrfs_block_group_used(&block_group->item) <
7289 space_info->total_bytes)) {
7290 spin_unlock(&space_info->lock);
7291 goto out;
7292 }
7293 spin_unlock(&space_info->lock);
7294
7295 /*
7296 * ok we don't have enough space, but maybe we have free space on our
7297 * devices to allocate new chunks for relocation, so loop through our
7298 * alloc devices and guess if we have enough space. However, if we
7299 * were marked as full, then we know there aren't enough chunks, and we
7300 * can just return.
7301 */
7302 ret = -1;
7303 if (full)
7304 goto out;
7305
7306 mutex_lock(&root->fs_info->chunk_mutex);
7307 list_for_each_entry(device, &fs_devices->alloc_list, dev_alloc_list) {
7308 u64 min_free = btrfs_block_group_used(&block_group->item);
7309 u64 dev_offset, max_avail;
7310
7311 /*
7312 * check to make sure we can actually find a chunk with enough
7313 * space to fit our block group in.
7314 */
7315 if (device->total_bytes > device->bytes_used + min_free) {
7316 ret = find_free_dev_extent(NULL, device, min_free,
7317 &dev_offset, &max_avail);
7318 if (!ret)
7319 break;
7320 ret = -1;
7321 }
7322 }
7323 mutex_unlock(&root->fs_info->chunk_mutex);
7324 out:
7325 btrfs_put_block_group(block_group);
7326 return ret;
7327 }
7328
7329 static int find_first_block_group(struct btrfs_root *root,
7330 struct btrfs_path *path, struct btrfs_key *key)
7331 {
7332 int ret = 0;
7333 struct btrfs_key found_key;
7334 struct extent_buffer *leaf;
7335 int slot;
7336
7337 ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
7338 if (ret < 0)
7339 goto out;
7340
7341 while (1) {
7342 slot = path->slots[0];
7343 leaf = path->nodes[0];
7344 if (slot >= btrfs_header_nritems(leaf)) {
7345 ret = btrfs_next_leaf(root, path);
7346 if (ret == 0)
7347 continue;
7348 if (ret < 0)
7349 goto out;
7350 break;
7351 }
7352 btrfs_item_key_to_cpu(leaf, &found_key, slot);
7353
7354 if (found_key.objectid >= key->objectid &&
7355 found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) {
7356 ret = 0;
7357 goto out;
7358 }
7359 path->slots[0]++;
7360 }
7361 ret = -ENOENT;
7362 out:
7363 return ret;
7364 }
7365
7366 int btrfs_free_block_groups(struct btrfs_fs_info *info)
7367 {
7368 struct btrfs_block_group_cache *block_group;
7369 struct btrfs_space_info *space_info;
7370 struct btrfs_caching_control *caching_ctl;
7371 struct rb_node *n;
7372
7373 down_write(&info->extent_commit_sem);
7374 while (!list_empty(&info->caching_block_groups)) {
7375 caching_ctl = list_entry(info->caching_block_groups.next,
7376 struct btrfs_caching_control, list);
7377 list_del(&caching_ctl->list);
7378 put_caching_control(caching_ctl);
7379 }
7380 up_write(&info->extent_commit_sem);
7381
7382 spin_lock(&info->block_group_cache_lock);
7383 while ((n = rb_last(&info->block_group_cache_tree)) != NULL) {
7384 block_group = rb_entry(n, struct btrfs_block_group_cache,
7385 cache_node);
7386 rb_erase(&block_group->cache_node,
7387 &info->block_group_cache_tree);
7388 spin_unlock(&info->block_group_cache_lock);
7389
7390 down_write(&block_group->space_info->groups_sem);
7391 list_del(&block_group->list);
7392 up_write(&block_group->space_info->groups_sem);
7393
7394 if (block_group->cached == BTRFS_CACHE_STARTED)
7395 wait_block_group_cache_done(block_group);
7396
7397 btrfs_remove_free_space_cache(block_group);
7398
7399 WARN_ON(atomic_read(&block_group->count) != 1);
7400 kfree(block_group);
7401
7402 spin_lock(&info->block_group_cache_lock);
7403 }
7404 spin_unlock(&info->block_group_cache_lock);
7405
7406 /* now that all the block groups are freed, go through and
7407 * free all the space_info structs. This is only called during
7408 * the final stages of unmount, and so we know nobody is
7409 * using them. We call synchronize_rcu() once before we start,
7410 * just to be on the safe side.
7411 */
7412 synchronize_rcu();
7413
7414 while(!list_empty(&info->space_info)) {
7415 space_info = list_entry(info->space_info.next,
7416 struct btrfs_space_info,
7417 list);
7418
7419 list_del(&space_info->list);
7420 kfree(space_info);
7421 }
7422 return 0;
7423 }
7424
7425 int btrfs_read_block_groups(struct btrfs_root *root)
7426 {
7427 struct btrfs_path *path;
7428 int ret;
7429 struct btrfs_block_group_cache *cache;
7430 struct btrfs_fs_info *info = root->fs_info;
7431 struct btrfs_space_info *space_info;
7432 struct btrfs_key key;
7433 struct btrfs_key found_key;
7434 struct extent_buffer *leaf;
7435
7436 root = info->extent_root;
7437 key.objectid = 0;
7438 key.offset = 0;
7439 btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
7440 path = btrfs_alloc_path();
7441 if (!path)
7442 return -ENOMEM;
7443
7444 while (1) {
7445 ret = find_first_block_group(root, path, &key);
7446 if (ret > 0) {
7447 ret = 0;
7448 goto error;
7449 }
7450 if (ret != 0)
7451 goto error;
7452
7453 leaf = path->nodes[0];
7454 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
7455 cache = kzalloc(sizeof(*cache), GFP_NOFS);
7456 if (!cache) {
7457 ret = -ENOMEM;
7458 break;
7459 }
7460
7461 atomic_set(&cache->count, 1);
7462 spin_lock_init(&cache->lock);
7463 spin_lock_init(&cache->tree_lock);
7464 cache->fs_info = info;
7465 INIT_LIST_HEAD(&cache->list);
7466 INIT_LIST_HEAD(&cache->cluster_list);
7467
7468 /*
7469 * we only want to have 32k of ram per block group for keeping
7470 * track of free space, and if we pass 1/2 of that we want to
7471 * start converting things over to using bitmaps
7472 */
7473 cache->extents_thresh = ((1024 * 32) / 2) /
7474 sizeof(struct btrfs_free_space);
7475
7476 read_extent_buffer(leaf, &cache->item,
7477 btrfs_item_ptr_offset(leaf, path->slots[0]),
7478 sizeof(cache->item));
7479 memcpy(&cache->key, &found_key, sizeof(found_key));
7480
7481 key.objectid = found_key.objectid + found_key.offset;
7482 btrfs_release_path(root, path);
7483 cache->flags = btrfs_block_group_flags(&cache->item);
7484 cache->sectorsize = root->sectorsize;
7485
7486 /*
7487 * check for two cases, either we are full, and therefore
7488 * don't need to bother with the caching work since we won't
7489 * find any space, or we are empty, and we can just add all
7490 * the space in and be done with it. This saves us _alot_ of
7491 * time, particularly in the full case.
7492 */
7493 if (found_key.offset == btrfs_block_group_used(&cache->item)) {
7494 exclude_super_stripes(root, cache);
7495 cache->last_byte_to_unpin = (u64)-1;
7496 cache->cached = BTRFS_CACHE_FINISHED;
7497 free_excluded_extents(root, cache);
7498 } else if (btrfs_block_group_used(&cache->item) == 0) {
7499 exclude_super_stripes(root, cache);
7500 cache->last_byte_to_unpin = (u64)-1;
7501 cache->cached = BTRFS_CACHE_FINISHED;
7502 add_new_free_space(cache, root->fs_info,
7503 found_key.objectid,
7504 found_key.objectid +
7505 found_key.offset);
7506 free_excluded_extents(root, cache);
7507 }
7508
7509 ret = update_space_info(info, cache->flags, found_key.offset,
7510 btrfs_block_group_used(&cache->item),
7511 &space_info);
7512 BUG_ON(ret);
7513 cache->space_info = space_info;
7514 spin_lock(&cache->space_info->lock);
7515 cache->space_info->bytes_super += cache->bytes_super;
7516 spin_unlock(&cache->space_info->lock);
7517
7518 down_write(&space_info->groups_sem);
7519 list_add_tail(&cache->list, &space_info->block_groups);
7520 up_write(&space_info->groups_sem);
7521
7522 ret = btrfs_add_block_group_cache(root->fs_info, cache);
7523 BUG_ON(ret);
7524
7525 set_avail_alloc_bits(root->fs_info, cache->flags);
7526 if (btrfs_chunk_readonly(root, cache->key.objectid))
7527 set_block_group_readonly(cache);
7528 }
7529 ret = 0;
7530 error:
7531 btrfs_free_path(path);
7532 return ret;
7533 }
7534
7535 int btrfs_make_block_group(struct btrfs_trans_handle *trans,
7536 struct btrfs_root *root, u64 bytes_used,
7537 u64 type, u64 chunk_objectid, u64 chunk_offset,
7538 u64 size)
7539 {
7540 int ret;
7541 struct btrfs_root *extent_root;
7542 struct btrfs_block_group_cache *cache;
7543
7544 extent_root = root->fs_info->extent_root;
7545
7546 root->fs_info->last_trans_log_full_commit = trans->transid;
7547
7548 cache = kzalloc(sizeof(*cache), GFP_NOFS);
7549 if (!cache)
7550 return -ENOMEM;
7551
7552 cache->key.objectid = chunk_offset;
7553 cache->key.offset = size;
7554 cache->key.type = BTRFS_BLOCK_GROUP_ITEM_KEY;
7555 cache->sectorsize = root->sectorsize;
7556
7557 /*
7558 * we only want to have 32k of ram per block group for keeping track
7559 * of free space, and if we pass 1/2 of that we want to start
7560 * converting things over to using bitmaps
7561 */
7562 cache->extents_thresh = ((1024 * 32) / 2) /
7563 sizeof(struct btrfs_free_space);
7564 atomic_set(&cache->count, 1);
7565 spin_lock_init(&cache->lock);
7566 spin_lock_init(&cache->tree_lock);
7567 INIT_LIST_HEAD(&cache->list);
7568 INIT_LIST_HEAD(&cache->cluster_list);
7569
7570 btrfs_set_block_group_used(&cache->item, bytes_used);
7571 btrfs_set_block_group_chunk_objectid(&cache->item, chunk_objectid);
7572 cache->flags = type;
7573 btrfs_set_block_group_flags(&cache->item, type);
7574
7575 cache->last_byte_to_unpin = (u64)-1;
7576 cache->cached = BTRFS_CACHE_FINISHED;
7577 exclude_super_stripes(root, cache);
7578
7579 add_new_free_space(cache, root->fs_info, chunk_offset,
7580 chunk_offset + size);
7581
7582 free_excluded_extents(root, cache);
7583
7584 ret = update_space_info(root->fs_info, cache->flags, size, bytes_used,
7585 &cache->space_info);
7586 BUG_ON(ret);
7587
7588 spin_lock(&cache->space_info->lock);
7589 cache->space_info->bytes_super += cache->bytes_super;
7590 spin_unlock(&cache->space_info->lock);
7591
7592 down_write(&cache->space_info->groups_sem);
7593 list_add_tail(&cache->list, &cache->space_info->block_groups);
7594 up_write(&cache->space_info->groups_sem);
7595
7596 ret = btrfs_add_block_group_cache(root->fs_info, cache);
7597 BUG_ON(ret);
7598
7599 ret = btrfs_insert_item(trans, extent_root, &cache->key, &cache->item,
7600 sizeof(cache->item));
7601 BUG_ON(ret);
7602
7603 set_avail_alloc_bits(extent_root->fs_info, type);
7604
7605 return 0;
7606 }
7607
7608 int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
7609 struct btrfs_root *root, u64 group_start)
7610 {
7611 struct btrfs_path *path;
7612 struct btrfs_block_group_cache *block_group;
7613 struct btrfs_free_cluster *cluster;
7614 struct btrfs_key key;
7615 int ret;
7616
7617 root = root->fs_info->extent_root;
7618
7619 block_group = btrfs_lookup_block_group(root->fs_info, group_start);
7620 BUG_ON(!block_group);
7621 BUG_ON(!block_group->ro);
7622
7623 memcpy(&key, &block_group->key, sizeof(key));
7624
7625 /* make sure this block group isn't part of an allocation cluster */
7626 cluster = &root->fs_info->data_alloc_cluster;
7627 spin_lock(&cluster->refill_lock);
7628 btrfs_return_cluster_to_free_space(block_group, cluster);
7629 spin_unlock(&cluster->refill_lock);
7630
7631 /*
7632 * make sure this block group isn't part of a metadata
7633 * allocation cluster
7634 */
7635 cluster = &root->fs_info->meta_alloc_cluster;
7636 spin_lock(&cluster->refill_lock);
7637 btrfs_return_cluster_to_free_space(block_group, cluster);
7638 spin_unlock(&cluster->refill_lock);
7639
7640 path = btrfs_alloc_path();
7641 BUG_ON(!path);
7642
7643 spin_lock(&root->fs_info->block_group_cache_lock);
7644 rb_erase(&block_group->cache_node,
7645 &root->fs_info->block_group_cache_tree);
7646 spin_unlock(&root->fs_info->block_group_cache_lock);
7647
7648 down_write(&block_group->space_info->groups_sem);
7649 /*
7650 * we must use list_del_init so people can check to see if they
7651 * are still on the list after taking the semaphore
7652 */
7653 list_del_init(&block_group->list);
7654 up_write(&block_group->space_info->groups_sem);
7655
7656 if (block_group->cached == BTRFS_CACHE_STARTED)
7657 wait_block_group_cache_done(block_group);
7658
7659 btrfs_remove_free_space_cache(block_group);
7660
7661 spin_lock(&block_group->space_info->lock);
7662 block_group->space_info->total_bytes -= block_group->key.offset;
7663 block_group->space_info->bytes_readonly -= block_group->key.offset;
7664 spin_unlock(&block_group->space_info->lock);
7665
7666 btrfs_clear_space_info_full(root->fs_info);
7667
7668 btrfs_put_block_group(block_group);
7669 btrfs_put_block_group(block_group);
7670
7671 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
7672 if (ret > 0)
7673 ret = -EIO;
7674 if (ret < 0)
7675 goto out;
7676
7677 ret = btrfs_del_item(trans, root, path);
7678 out:
7679 btrfs_free_path(path);
7680 return ret;
7681 }
Linus' kernel tree