?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
x86/amd-iommu: Add per IOMMU reference counting
[linux/fpc-iii.git] / fs / btrfs / extent-tree.c
blob94627c4cc193343aacc82b3aca1a6e9eaaefd8dd
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 for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
199 bytenr = btrfs_sb_offset(i);
200 ret = btrfs_rmap_block(&root->fs_info->mapping_tree,
201 cache->key.objectid, bytenr,
202 0, &logical, &nr, &stripe_len);
203 BUG_ON(ret);
204
205 while (nr--) {
206 cache->bytes_super += stripe_len;
207 ret = add_excluded_extent(root, logical[nr],
208 stripe_len);
209 BUG_ON(ret);
210 }
211
212 kfree(logical);
213 }
214 return 0;
215 }
216
217 static struct btrfs_caching_control *
218 get_caching_control(struct btrfs_block_group_cache *cache)
219 {
220 struct btrfs_caching_control *ctl;
221
222 spin_lock(&cache->lock);
223 if (cache->cached != BTRFS_CACHE_STARTED) {
224 spin_unlock(&cache->lock);
225 return NULL;
226 }
227
228 ctl = cache->caching_ctl;
229 atomic_inc(&ctl->count);
230 spin_unlock(&cache->lock);
231 return ctl;
232 }
233
234 static void put_caching_control(struct btrfs_caching_control *ctl)
235 {
236 if (atomic_dec_and_test(&ctl->count))
237 kfree(ctl);
238 }
239
240 /*
241 * this is only called by cache_block_group, since we could have freed extents
242 * we need to check the pinned_extents for any extents that can't be used yet
243 * since their free space will be released as soon as the transaction commits.
244 */
245 static u64 add_new_free_space(struct btrfs_block_group_cache *block_group,
246 struct btrfs_fs_info *info, u64 start, u64 end)
247 {
248 u64 extent_start, extent_end, size, total_added = 0;
249 int ret;
250
251 while (start < end) {
252 ret = find_first_extent_bit(info->pinned_extents, start,
253 &extent_start, &extent_end,
254 EXTENT_DIRTY | EXTENT_UPTODATE);
255 if (ret)
256 break;
257
258 if (extent_start == start) {
259 start = extent_end + 1;
260 } else if (extent_start > start && extent_start < end) {
261 size = extent_start - start;
262 total_added += size;
263 ret = btrfs_add_free_space(block_group, start,
264 size);
265 BUG_ON(ret);
266 start = extent_end + 1;
267 } else {
268 break;
269 }
270 }
271
272 if (start < end) {
273 size = end - start;
274 total_added += size;
275 ret = btrfs_add_free_space(block_group, start, size);
276 BUG_ON(ret);
277 }
278
279 return total_added;
280 }
281
282 static int caching_kthread(void *data)
283 {
284 struct btrfs_block_group_cache *block_group = data;
285 struct btrfs_fs_info *fs_info = block_group->fs_info;
286 struct btrfs_caching_control *caching_ctl = block_group->caching_ctl;
287 struct btrfs_root *extent_root = fs_info->extent_root;
288 struct btrfs_path *path;
289 struct extent_buffer *leaf;
290 struct btrfs_key key;
291 u64 total_found = 0;
292 u64 last = 0;
293 u32 nritems;
294 int ret = 0;
295
296 path = btrfs_alloc_path();
297 if (!path)
298 return -ENOMEM;
299
300 exclude_super_stripes(extent_root, block_group);
301 spin_lock(&block_group->space_info->lock);
302 block_group->space_info->bytes_super += block_group->bytes_super;
303 spin_unlock(&block_group->space_info->lock);
304
305 last = max_t(u64, block_group->key.objectid, BTRFS_SUPER_INFO_OFFSET);
306
307 /*
308 * We don't want to deadlock with somebody trying to allocate a new
309 * extent for the extent root while also trying to search the extent
310 * root to add free space. So we skip locking and search the commit
311 * root, since its read-only
312 */
313 path->skip_locking = 1;
314 path->search_commit_root = 1;
315 path->reada = 2;
316
317 key.objectid = last;
318 key.offset = 0;
319 key.type = BTRFS_EXTENT_ITEM_KEY;
320 again:
321 mutex_lock(&caching_ctl->mutex);
322 /* need to make sure the commit_root doesn't disappear */
323 down_read(&fs_info->extent_commit_sem);
324
325 ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
326 if (ret < 0)
327 goto err;
328
329 leaf = path->nodes[0];
330 nritems = btrfs_header_nritems(leaf);
331
332 while (1) {
333 smp_mb();
334 if (fs_info->closing > 1) {
335 last = (u64)-1;
336 break;
337 }
338
339 if (path->slots[0] < nritems) {
340 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
341 } else {
342 ret = find_next_key(path, 0, &key);
343 if (ret)
344 break;
345
346 caching_ctl->progress = last;
347 btrfs_release_path(extent_root, path);
348 up_read(&fs_info->extent_commit_sem);
349 mutex_unlock(&caching_ctl->mutex);
350 if (btrfs_transaction_in_commit(fs_info))
351 schedule_timeout(1);
352 else
353 cond_resched();
354 goto again;
355 }
356
357 if (key.objectid < block_group->key.objectid) {
358 path->slots[0]++;
359 continue;
360 }
361
362 if (key.objectid >= block_group->key.objectid +
363 block_group->key.offset)
364 break;
365
366 if (key.type == BTRFS_EXTENT_ITEM_KEY) {
367 total_found += add_new_free_space(block_group,
368 fs_info, last,
369 key.objectid);
370 last = key.objectid + key.offset;
371
372 if (total_found > (1024 * 1024 * 2)) {
373 total_found = 0;
374 wake_up(&caching_ctl->wait);
375 }
376 }
377 path->slots[0]++;
378 }
379 ret = 0;
380
381 total_found += add_new_free_space(block_group, fs_info, last,
382 block_group->key.objectid +
383 block_group->key.offset);
384 caching_ctl->progress = (u64)-1;
385
386 spin_lock(&block_group->lock);
387 block_group->caching_ctl = NULL;
388 block_group->cached = BTRFS_CACHE_FINISHED;
389 spin_unlock(&block_group->lock);
390
391 err:
392 btrfs_free_path(path);
393 up_read(&fs_info->extent_commit_sem);
394
395 free_excluded_extents(extent_root, block_group);
396
397 mutex_unlock(&caching_ctl->mutex);
398 wake_up(&caching_ctl->wait);
399
400 put_caching_control(caching_ctl);
401 atomic_dec(&block_group->space_info->caching_threads);
402 return 0;
403 }
404
405 static int cache_block_group(struct btrfs_block_group_cache *cache)
406 {
407 struct btrfs_fs_info *fs_info = cache->fs_info;
408 struct btrfs_caching_control *caching_ctl;
409 struct task_struct *tsk;
410 int ret = 0;
411
412 smp_mb();
413 if (cache->cached != BTRFS_CACHE_NO)
414 return 0;
415
416 caching_ctl = kzalloc(sizeof(*caching_ctl), GFP_KERNEL);
417 BUG_ON(!caching_ctl);
418
419 INIT_LIST_HEAD(&caching_ctl->list);
420 mutex_init(&caching_ctl->mutex);
421 init_waitqueue_head(&caching_ctl->wait);
422 caching_ctl->block_group = cache;
423 caching_ctl->progress = cache->key.objectid;
424 /* one for caching kthread, one for caching block group list */
425 atomic_set(&caching_ctl->count, 2);
426
427 spin_lock(&cache->lock);
428 if (cache->cached != BTRFS_CACHE_NO) {
429 spin_unlock(&cache->lock);
430 kfree(caching_ctl);
431 return 0;
432 }
433 cache->caching_ctl = caching_ctl;
434 cache->cached = BTRFS_CACHE_STARTED;
435 spin_unlock(&cache->lock);
436
437 down_write(&fs_info->extent_commit_sem);
438 list_add_tail(&caching_ctl->list, &fs_info->caching_block_groups);
439 up_write(&fs_info->extent_commit_sem);
440
441 atomic_inc(&cache->space_info->caching_threads);
442
443 tsk = kthread_run(caching_kthread, cache, "btrfs-cache-%llu\n",
444 cache->key.objectid);
445 if (IS_ERR(tsk)) {
446 ret = PTR_ERR(tsk);
447 printk(KERN_ERR "error running thread %d\n", ret);
448 BUG();
449 }
450
451 return ret;
452 }
453
454 /*
455 * return the block group that starts at or after bytenr
456 */
457 static struct btrfs_block_group_cache *
458 btrfs_lookup_first_block_group(struct btrfs_fs_info *info, u64 bytenr)
459 {
460 struct btrfs_block_group_cache *cache;
461
462 cache = block_group_cache_tree_search(info, bytenr, 0);
463
464 return cache;
465 }
466
467 /*
468 * return the block group that contains the given bytenr
469 */
470 struct btrfs_block_group_cache *btrfs_lookup_block_group(
471 struct btrfs_fs_info *info,
472 u64 bytenr)
473 {
474 struct btrfs_block_group_cache *cache;
475
476 cache = block_group_cache_tree_search(info, bytenr, 1);
477
478 return cache;
479 }
480
481 void btrfs_put_block_group(struct btrfs_block_group_cache *cache)
482 {
483 if (atomic_dec_and_test(&cache->count))
484 kfree(cache);
485 }
486
487 static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info,
488 u64 flags)
489 {
490 struct list_head *head = &info->space_info;
491 struct btrfs_space_info *found;
492
493 rcu_read_lock();
494 list_for_each_entry_rcu(found, head, list) {
495 if (found->flags == flags) {
496 rcu_read_unlock();
497 return found;
498 }
499 }
500 rcu_read_unlock();
501 return NULL;
502 }
503
504 /*
505 * after adding space to the filesystem, we need to clear the full flags
506 * on all the space infos.
507 */
508 void btrfs_clear_space_info_full(struct btrfs_fs_info *info)
509 {
510 struct list_head *head = &info->space_info;
511 struct btrfs_space_info *found;
512
513 rcu_read_lock();
514 list_for_each_entry_rcu(found, head, list)
515 found->full = 0;
516 rcu_read_unlock();
517 }
518
519 static u64 div_factor(u64 num, int factor)
520 {
521 if (factor == 10)
522 return num;
523 num *= factor;
524 do_div(num, 10);
525 return num;
526 }
527
528 u64 btrfs_find_block_group(struct btrfs_root *root,
529 u64 search_start, u64 search_hint, int owner)
530 {
531 struct btrfs_block_group_cache *cache;
532 u64 used;
533 u64 last = max(search_hint, search_start);
534 u64 group_start = 0;
535 int full_search = 0;
536 int factor = 9;
537 int wrapped = 0;
538 again:
539 while (1) {
540 cache = btrfs_lookup_first_block_group(root->fs_info, last);
541 if (!cache)
542 break;
543
544 spin_lock(&cache->lock);
545 last = cache->key.objectid + cache->key.offset;
546 used = btrfs_block_group_used(&cache->item);
547
548 if ((full_search || !cache->ro) &&
549 block_group_bits(cache, BTRFS_BLOCK_GROUP_METADATA)) {
550 if (used + cache->pinned + cache->reserved <
551 div_factor(cache->key.offset, factor)) {
552 group_start = cache->key.objectid;
553 spin_unlock(&cache->lock);
554 btrfs_put_block_group(cache);
555 goto found;
556 }
557 }
558 spin_unlock(&cache->lock);
559 btrfs_put_block_group(cache);
560 cond_resched();
561 }
562 if (!wrapped) {
563 last = search_start;
564 wrapped = 1;
565 goto again;
566 }
567 if (!full_search && factor < 10) {
568 last = search_start;
569 full_search = 1;
570 factor = 10;
571 goto again;
572 }
573 found:
574 return group_start;
575 }
576
577 /* simple helper to search for an existing extent at a given offset */
578 int btrfs_lookup_extent(struct btrfs_root *root, u64 start, u64 len)
579 {
580 int ret;
581 struct btrfs_key key;
582 struct btrfs_path *path;
583
584 path = btrfs_alloc_path();
585 BUG_ON(!path);
586 key.objectid = start;
587 key.offset = len;
588 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
589 ret = btrfs_search_slot(NULL, root->fs_info->extent_root, &key, path,
590 0, 0);
591 btrfs_free_path(path);
592 return ret;
593 }
594
595 /*
596 * Back reference rules. Back refs have three main goals:
597 *
598 * 1) differentiate between all holders of references to an extent so that
599 * when a reference is dropped we can make sure it was a valid reference
600 * before freeing the extent.
601 *
602 * 2) Provide enough information to quickly find the holders of an extent
603 * if we notice a given block is corrupted or bad.
604 *
605 * 3) Make it easy to migrate blocks for FS shrinking or storage pool
606 * maintenance. This is actually the same as #2, but with a slightly
607 * different use case.
608 *
609 * There are two kinds of back refs. The implicit back refs is optimized
610 * for pointers in non-shared tree blocks. For a given pointer in a block,
611 * back refs of this kind provide information about the block's owner tree
612 * and the pointer's key. These information allow us to find the block by
613 * b-tree searching. The full back refs is for pointers in tree blocks not
614 * referenced by their owner trees. The location of tree block is recorded
615 * in the back refs. Actually the full back refs is generic, and can be
616 * used in all cases the implicit back refs is used. The major shortcoming
617 * of the full back refs is its overhead. Every time a tree block gets
618 * COWed, we have to update back refs entry for all pointers in it.
619 *
620 * For a newly allocated tree block, we use implicit back refs for
621 * pointers in it. This means most tree related operations only involve
622 * implicit back refs. For a tree block created in old transaction, the
623 * only way to drop a reference to it is COW it. So we can detect the
624 * event that tree block loses its owner tree's reference and do the
625 * back refs conversion.
626 *
627 * When a tree block is COW'd through a tree, there are four cases:
628 *
629 * The reference count of the block is one and the tree is the block's
630 * owner tree. Nothing to do in this case.
631 *
632 * The reference count of the block is one and the tree is not the
633 * block's owner tree. In this case, full back refs is used for pointers
634 * in the block. Remove these full back refs, add implicit back refs for
635 * every pointers in the new block.
636 *
637 * The reference count of the block is greater than one and the tree is
638 * the block's owner tree. In this case, implicit back refs is used for
639 * pointers in the block. Add full back refs for every pointers in the
640 * block, increase lower level extents' reference counts. The original
641 * implicit back refs are entailed to the new block.
642 *
643 * The reference count of the block is greater than one and the tree is
644 * not the block's owner tree. Add implicit back refs for every pointer in
645 * the new block, increase lower level extents' reference count.
646 *
647 * Back Reference Key composing:
648 *
649 * The key objectid corresponds to the first byte in the extent,
650 * The key type is used to differentiate between types of back refs.
651 * There are different meanings of the key offset for different types
652 * of back refs.
653 *
654 * File extents can be referenced by:
655 *
656 * - multiple snapshots, subvolumes, or different generations in one subvol
657 * - different files inside a single subvolume
658 * - different offsets inside a file (bookend extents in file.c)
659 *
660 * The extent ref structure for the implicit back refs has fields for:
661 *
662 * - Objectid of the subvolume root
663 * - objectid of the file holding the reference
664 * - original offset in the file
665 * - how many bookend extents
666 *
667 * The key offset for the implicit back refs is hash of the first
668 * three fields.
669 *
670 * The extent ref structure for the full back refs has field for:
671 *
672 * - number of pointers in the tree leaf
673 *
674 * The key offset for the implicit back refs is the first byte of
675 * the tree leaf
676 *
677 * When a file extent is allocated, The implicit back refs is used.
678 * the fields are filled in:
679 *
680 * (root_key.objectid, inode objectid, offset in file, 1)
681 *
682 * When a file extent is removed file truncation, we find the
683 * corresponding implicit back refs and check the following fields:
684 *
685 * (btrfs_header_owner(leaf), inode objectid, offset in file)
686 *
687 * Btree extents can be referenced by:
688 *
689 * - Different subvolumes
690 *
691 * Both the implicit back refs and the full back refs for tree blocks
692 * only consist of key. The key offset for the implicit back refs is
693 * objectid of block's owner tree. The key offset for the full back refs
694 * is the first byte of parent block.
695 *
696 * When implicit back refs is used, information about the lowest key and
697 * level of the tree block are required. These information are stored in
698 * tree block info structure.
699 */
700
701 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
702 static int convert_extent_item_v0(struct btrfs_trans_handle *trans,
703 struct btrfs_root *root,
704 struct btrfs_path *path,
705 u64 owner, u32 extra_size)
706 {
707 struct btrfs_extent_item *item;
708 struct btrfs_extent_item_v0 *ei0;
709 struct btrfs_extent_ref_v0 *ref0;
710 struct btrfs_tree_block_info *bi;
711 struct extent_buffer *leaf;
712 struct btrfs_key key;
713 struct btrfs_key found_key;
714 u32 new_size = sizeof(*item);
715 u64 refs;
716 int ret;
717
718 leaf = path->nodes[0];
719 BUG_ON(btrfs_item_size_nr(leaf, path->slots[0]) != sizeof(*ei0));
720
721 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
722 ei0 = btrfs_item_ptr(leaf, path->slots[0],
723 struct btrfs_extent_item_v0);
724 refs = btrfs_extent_refs_v0(leaf, ei0);
725
726 if (owner == (u64)-1) {
727 while (1) {
728 if (path->slots[0] >= btrfs_header_nritems(leaf)) {
729 ret = btrfs_next_leaf(root, path);
730 if (ret < 0)
731 return ret;
732 BUG_ON(ret > 0);
733 leaf = path->nodes[0];
734 }
735 btrfs_item_key_to_cpu(leaf, &found_key,
736 path->slots[0]);
737 BUG_ON(key.objectid != found_key.objectid);
738 if (found_key.type != BTRFS_EXTENT_REF_V0_KEY) {
739 path->slots[0]++;
740 continue;
741 }
742 ref0 = btrfs_item_ptr(leaf, path->slots[0],
743 struct btrfs_extent_ref_v0);
744 owner = btrfs_ref_objectid_v0(leaf, ref0);
745 break;
746 }
747 }
748 btrfs_release_path(root, path);
749
750 if (owner < BTRFS_FIRST_FREE_OBJECTID)
751 new_size += sizeof(*bi);
752
753 new_size -= sizeof(*ei0);
754 ret = btrfs_search_slot(trans, root, &key, path,
755 new_size + extra_size, 1);
756 if (ret < 0)
757 return ret;
758 BUG_ON(ret);
759
760 ret = btrfs_extend_item(trans, root, path, new_size);
761 BUG_ON(ret);
762
763 leaf = path->nodes[0];
764 item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
765 btrfs_set_extent_refs(leaf, item, refs);
766 /* FIXME: get real generation */
767 btrfs_set_extent_generation(leaf, item, 0);
768 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
769 btrfs_set_extent_flags(leaf, item,
770 BTRFS_EXTENT_FLAG_TREE_BLOCK |
771 BTRFS_BLOCK_FLAG_FULL_BACKREF);
772 bi = (struct btrfs_tree_block_info *)(item + 1);
773 /* FIXME: get first key of the block */
774 memset_extent_buffer(leaf, 0, (unsigned long)bi, sizeof(*bi));
775 btrfs_set_tree_block_level(leaf, bi, (int)owner);
776 } else {
777 btrfs_set_extent_flags(leaf, item, BTRFS_EXTENT_FLAG_DATA);
778 }
779 btrfs_mark_buffer_dirty(leaf);
780 return 0;
781 }
782 #endif
783
784 static u64 hash_extent_data_ref(u64 root_objectid, u64 owner, u64 offset)
785 {
786 u32 high_crc = ~(u32)0;
787 u32 low_crc = ~(u32)0;
788 __le64 lenum;
789
790 lenum = cpu_to_le64(root_objectid);
791 high_crc = crc32c(high_crc, &lenum, sizeof(lenum));
792 lenum = cpu_to_le64(owner);
793 low_crc = crc32c(low_crc, &lenum, sizeof(lenum));
794 lenum = cpu_to_le64(offset);
795 low_crc = crc32c(low_crc, &lenum, sizeof(lenum));
796
797 return ((u64)high_crc << 31) ^ (u64)low_crc;
798 }
799
800 static u64 hash_extent_data_ref_item(struct extent_buffer *leaf,
801 struct btrfs_extent_data_ref *ref)
802 {
803 return hash_extent_data_ref(btrfs_extent_data_ref_root(leaf, ref),
804 btrfs_extent_data_ref_objectid(leaf, ref),
805 btrfs_extent_data_ref_offset(leaf, ref));
806 }
807
808 static int match_extent_data_ref(struct extent_buffer *leaf,
809 struct btrfs_extent_data_ref *ref,
810 u64 root_objectid, u64 owner, u64 offset)
811 {
812 if (btrfs_extent_data_ref_root(leaf, ref) != root_objectid ||
813 btrfs_extent_data_ref_objectid(leaf, ref) != owner ||
814 btrfs_extent_data_ref_offset(leaf, ref) != offset)
815 return 0;
816 return 1;
817 }
818
819 static noinline int lookup_extent_data_ref(struct btrfs_trans_handle *trans,
820 struct btrfs_root *root,
821 struct btrfs_path *path,
822 u64 bytenr, u64 parent,
823 u64 root_objectid,
824 u64 owner, u64 offset)
825 {
826 struct btrfs_key key;
827 struct btrfs_extent_data_ref *ref;
828 struct extent_buffer *leaf;
829 u32 nritems;
830 int ret;
831 int recow;
832 int err = -ENOENT;
833
834 key.objectid = bytenr;
835 if (parent) {
836 key.type = BTRFS_SHARED_DATA_REF_KEY;
837 key.offset = parent;
838 } else {
839 key.type = BTRFS_EXTENT_DATA_REF_KEY;
840 key.offset = hash_extent_data_ref(root_objectid,
841 owner, offset);
842 }
843 again:
844 recow = 0;
845 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
846 if (ret < 0) {
847 err = ret;
848 goto fail;
849 }
850
851 if (parent) {
852 if (!ret)
853 return 0;
854 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
855 key.type = BTRFS_EXTENT_REF_V0_KEY;
856 btrfs_release_path(root, path);
857 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
858 if (ret < 0) {
859 err = ret;
860 goto fail;
861 }
862 if (!ret)
863 return 0;
864 #endif
865 goto fail;
866 }
867
868 leaf = path->nodes[0];
869 nritems = btrfs_header_nritems(leaf);
870 while (1) {
871 if (path->slots[0] >= nritems) {
872 ret = btrfs_next_leaf(root, path);
873 if (ret < 0)
874 err = ret;
875 if (ret)
876 goto fail;
877
878 leaf = path->nodes[0];
879 nritems = btrfs_header_nritems(leaf);
880 recow = 1;
881 }
882
883 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
884 if (key.objectid != bytenr ||
885 key.type != BTRFS_EXTENT_DATA_REF_KEY)
886 goto fail;
887
888 ref = btrfs_item_ptr(leaf, path->slots[0],
889 struct btrfs_extent_data_ref);
890
891 if (match_extent_data_ref(leaf, ref, root_objectid,
892 owner, offset)) {
893 if (recow) {
894 btrfs_release_path(root, path);
895 goto again;
896 }
897 err = 0;
898 break;
899 }
900 path->slots[0]++;
901 }
902 fail:
903 return err;
904 }
905
906 static noinline int insert_extent_data_ref(struct btrfs_trans_handle *trans,
907 struct btrfs_root *root,
908 struct btrfs_path *path,
909 u64 bytenr, u64 parent,
910 u64 root_objectid, u64 owner,
911 u64 offset, int refs_to_add)
912 {
913 struct btrfs_key key;
914 struct extent_buffer *leaf;
915 u32 size;
916 u32 num_refs;
917 int ret;
918
919 key.objectid = bytenr;
920 if (parent) {
921 key.type = BTRFS_SHARED_DATA_REF_KEY;
922 key.offset = parent;
923 size = sizeof(struct btrfs_shared_data_ref);
924 } else {
925 key.type = BTRFS_EXTENT_DATA_REF_KEY;
926 key.offset = hash_extent_data_ref(root_objectid,
927 owner, offset);
928 size = sizeof(struct btrfs_extent_data_ref);
929 }
930
931 ret = btrfs_insert_empty_item(trans, root, path, &key, size);
932 if (ret && ret != -EEXIST)
933 goto fail;
934
935 leaf = path->nodes[0];
936 if (parent) {
937 struct btrfs_shared_data_ref *ref;
938 ref = btrfs_item_ptr(leaf, path->slots[0],
939 struct btrfs_shared_data_ref);
940 if (ret == 0) {
941 btrfs_set_shared_data_ref_count(leaf, ref, refs_to_add);
942 } else {
943 num_refs = btrfs_shared_data_ref_count(leaf, ref);
944 num_refs += refs_to_add;
945 btrfs_set_shared_data_ref_count(leaf, ref, num_refs);
946 }
947 } else {
948 struct btrfs_extent_data_ref *ref;
949 while (ret == -EEXIST) {
950 ref = btrfs_item_ptr(leaf, path->slots[0],
951 struct btrfs_extent_data_ref);
952 if (match_extent_data_ref(leaf, ref, root_objectid,
953 owner, offset))
954 break;
955 btrfs_release_path(root, path);
956 key.offset++;
957 ret = btrfs_insert_empty_item(trans, root, path, &key,
958 size);
959 if (ret && ret != -EEXIST)
960 goto fail;
961
962 leaf = path->nodes[0];
963 }
964 ref = btrfs_item_ptr(leaf, path->slots[0],
965 struct btrfs_extent_data_ref);
966 if (ret == 0) {
967 btrfs_set_extent_data_ref_root(leaf, ref,
968 root_objectid);
969 btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
970 btrfs_set_extent_data_ref_offset(leaf, ref, offset);
971 btrfs_set_extent_data_ref_count(leaf, ref, refs_to_add);
972 } else {
973 num_refs = btrfs_extent_data_ref_count(leaf, ref);
974 num_refs += refs_to_add;
975 btrfs_set_extent_data_ref_count(leaf, ref, num_refs);
976 }
977 }
978 btrfs_mark_buffer_dirty(leaf);
979 ret = 0;
980 fail:
981 btrfs_release_path(root, path);
982 return ret;
983 }
984
985 static noinline int remove_extent_data_ref(struct btrfs_trans_handle *trans,
986 struct btrfs_root *root,
987 struct btrfs_path *path,
988 int refs_to_drop)
989 {
990 struct btrfs_key key;
991 struct btrfs_extent_data_ref *ref1 = NULL;
992 struct btrfs_shared_data_ref *ref2 = NULL;
993 struct extent_buffer *leaf;
994 u32 num_refs = 0;
995 int ret = 0;
996
997 leaf = path->nodes[0];
998 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
999
1000 if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
1001 ref1 = btrfs_item_ptr(leaf, path->slots[0],
1002 struct btrfs_extent_data_ref);
1003 num_refs = btrfs_extent_data_ref_count(leaf, ref1);
1004 } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
1005 ref2 = btrfs_item_ptr(leaf, path->slots[0],
1006 struct btrfs_shared_data_ref);
1007 num_refs = btrfs_shared_data_ref_count(leaf, ref2);
1008 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1009 } else if (key.type == BTRFS_EXTENT_REF_V0_KEY) {
1010 struct btrfs_extent_ref_v0 *ref0;
1011 ref0 = btrfs_item_ptr(leaf, path->slots[0],
1012 struct btrfs_extent_ref_v0);
1013 num_refs = btrfs_ref_count_v0(leaf, ref0);
1014 #endif
1015 } else {
1016 BUG();
1017 }
1018
1019 BUG_ON(num_refs < refs_to_drop);
1020 num_refs -= refs_to_drop;
1021
1022 if (num_refs == 0) {
1023 ret = btrfs_del_item(trans, root, path);
1024 } else {
1025 if (key.type == BTRFS_EXTENT_DATA_REF_KEY)
1026 btrfs_set_extent_data_ref_count(leaf, ref1, num_refs);
1027 else if (key.type == BTRFS_SHARED_DATA_REF_KEY)
1028 btrfs_set_shared_data_ref_count(leaf, ref2, num_refs);
1029 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1030 else {
1031 struct btrfs_extent_ref_v0 *ref0;
1032 ref0 = btrfs_item_ptr(leaf, path->slots[0],
1033 struct btrfs_extent_ref_v0);
1034 btrfs_set_ref_count_v0(leaf, ref0, num_refs);
1035 }
1036 #endif
1037 btrfs_mark_buffer_dirty(leaf);
1038 }
1039 return ret;
1040 }
1041
1042 static noinline u32 extent_data_ref_count(struct btrfs_root *root,
1043 struct btrfs_path *path,
1044 struct btrfs_extent_inline_ref *iref)
1045 {
1046 struct btrfs_key key;
1047 struct extent_buffer *leaf;
1048 struct btrfs_extent_data_ref *ref1;
1049 struct btrfs_shared_data_ref *ref2;
1050 u32 num_refs = 0;
1051
1052 leaf = path->nodes[0];
1053 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
1054 if (iref) {
1055 if (btrfs_extent_inline_ref_type(leaf, iref) ==
1056 BTRFS_EXTENT_DATA_REF_KEY) {
1057 ref1 = (struct btrfs_extent_data_ref *)(&iref->offset);
1058 num_refs = btrfs_extent_data_ref_count(leaf, ref1);
1059 } else {
1060 ref2 = (struct btrfs_shared_data_ref *)(iref + 1);
1061 num_refs = btrfs_shared_data_ref_count(leaf, ref2);
1062 }
1063 } else if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
1064 ref1 = btrfs_item_ptr(leaf, path->slots[0],
1065 struct btrfs_extent_data_ref);
1066 num_refs = btrfs_extent_data_ref_count(leaf, ref1);
1067 } else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
1068 ref2 = btrfs_item_ptr(leaf, path->slots[0],
1069 struct btrfs_shared_data_ref);
1070 num_refs = btrfs_shared_data_ref_count(leaf, ref2);
1071 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1072 } else if (key.type == BTRFS_EXTENT_REF_V0_KEY) {
1073 struct btrfs_extent_ref_v0 *ref0;
1074 ref0 = btrfs_item_ptr(leaf, path->slots[0],
1075 struct btrfs_extent_ref_v0);
1076 num_refs = btrfs_ref_count_v0(leaf, ref0);
1077 #endif
1078 } else {
1079 WARN_ON(1);
1080 }
1081 return num_refs;
1082 }
1083
1084 static noinline int lookup_tree_block_ref(struct btrfs_trans_handle *trans,
1085 struct btrfs_root *root,
1086 struct btrfs_path *path,
1087 u64 bytenr, u64 parent,
1088 u64 root_objectid)
1089 {
1090 struct btrfs_key key;
1091 int ret;
1092
1093 key.objectid = bytenr;
1094 if (parent) {
1095 key.type = BTRFS_SHARED_BLOCK_REF_KEY;
1096 key.offset = parent;
1097 } else {
1098 key.type = BTRFS_TREE_BLOCK_REF_KEY;
1099 key.offset = root_objectid;
1100 }
1101
1102 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1103 if (ret > 0)
1104 ret = -ENOENT;
1105 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1106 if (ret == -ENOENT && parent) {
1107 btrfs_release_path(root, path);
1108 key.type = BTRFS_EXTENT_REF_V0_KEY;
1109 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
1110 if (ret > 0)
1111 ret = -ENOENT;
1112 }
1113 #endif
1114 return ret;
1115 }
1116
1117 static noinline int insert_tree_block_ref(struct btrfs_trans_handle *trans,
1118 struct btrfs_root *root,
1119 struct btrfs_path *path,
1120 u64 bytenr, u64 parent,
1121 u64 root_objectid)
1122 {
1123 struct btrfs_key key;
1124 int ret;
1125
1126 key.objectid = bytenr;
1127 if (parent) {
1128 key.type = BTRFS_SHARED_BLOCK_REF_KEY;
1129 key.offset = parent;
1130 } else {
1131 key.type = BTRFS_TREE_BLOCK_REF_KEY;
1132 key.offset = root_objectid;
1133 }
1134
1135 ret = btrfs_insert_empty_item(trans, root, path, &key, 0);
1136 btrfs_release_path(root, path);
1137 return ret;
1138 }
1139
1140 static inline int extent_ref_type(u64 parent, u64 owner)
1141 {
1142 int type;
1143 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1144 if (parent > 0)
1145 type = BTRFS_SHARED_BLOCK_REF_KEY;
1146 else
1147 type = BTRFS_TREE_BLOCK_REF_KEY;
1148 } else {
1149 if (parent > 0)
1150 type = BTRFS_SHARED_DATA_REF_KEY;
1151 else
1152 type = BTRFS_EXTENT_DATA_REF_KEY;
1153 }
1154 return type;
1155 }
1156
1157 static int find_next_key(struct btrfs_path *path, int level,
1158 struct btrfs_key *key)
1159
1160 {
1161 for (; level < BTRFS_MAX_LEVEL; level++) {
1162 if (!path->nodes[level])
1163 break;
1164 if (path->slots[level] + 1 >=
1165 btrfs_header_nritems(path->nodes[level]))
1166 continue;
1167 if (level == 0)
1168 btrfs_item_key_to_cpu(path->nodes[level], key,
1169 path->slots[level] + 1);
1170 else
1171 btrfs_node_key_to_cpu(path->nodes[level], key,
1172 path->slots[level] + 1);
1173 return 0;
1174 }
1175 return 1;
1176 }
1177
1178 /*
1179 * look for inline back ref. if back ref is found, *ref_ret is set
1180 * to the address of inline back ref, and 0 is returned.
1181 *
1182 * if back ref isn't found, *ref_ret is set to the address where it
1183 * should be inserted, and -ENOENT is returned.
1184 *
1185 * if insert is true and there are too many inline back refs, the path
1186 * points to the extent item, and -EAGAIN is returned.
1187 *
1188 * NOTE: inline back refs are ordered in the same way that back ref
1189 * items in the tree are ordered.
1190 */
1191 static noinline_for_stack
1192 int lookup_inline_extent_backref(struct btrfs_trans_handle *trans,
1193 struct btrfs_root *root,
1194 struct btrfs_path *path,
1195 struct btrfs_extent_inline_ref **ref_ret,
1196 u64 bytenr, u64 num_bytes,
1197 u64 parent, u64 root_objectid,
1198 u64 owner, u64 offset, int insert)
1199 {
1200 struct btrfs_key key;
1201 struct extent_buffer *leaf;
1202 struct btrfs_extent_item *ei;
1203 struct btrfs_extent_inline_ref *iref;
1204 u64 flags;
1205 u64 item_size;
1206 unsigned long ptr;
1207 unsigned long end;
1208 int extra_size;
1209 int type;
1210 int want;
1211 int ret;
1212 int err = 0;
1213
1214 key.objectid = bytenr;
1215 key.type = BTRFS_EXTENT_ITEM_KEY;
1216 key.offset = num_bytes;
1217
1218 want = extent_ref_type(parent, owner);
1219 if (insert) {
1220 extra_size = btrfs_extent_inline_ref_size(want);
1221 path->keep_locks = 1;
1222 } else
1223 extra_size = -1;
1224 ret = btrfs_search_slot(trans, root, &key, path, extra_size, 1);
1225 if (ret < 0) {
1226 err = ret;
1227 goto out;
1228 }
1229 BUG_ON(ret);
1230
1231 leaf = path->nodes[0];
1232 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1233 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1234 if (item_size < sizeof(*ei)) {
1235 if (!insert) {
1236 err = -ENOENT;
1237 goto out;
1238 }
1239 ret = convert_extent_item_v0(trans, root, path, owner,
1240 extra_size);
1241 if (ret < 0) {
1242 err = ret;
1243 goto out;
1244 }
1245 leaf = path->nodes[0];
1246 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1247 }
1248 #endif
1249 BUG_ON(item_size < sizeof(*ei));
1250
1251 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1252 flags = btrfs_extent_flags(leaf, ei);
1253
1254 ptr = (unsigned long)(ei + 1);
1255 end = (unsigned long)ei + item_size;
1256
1257 if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
1258 ptr += sizeof(struct btrfs_tree_block_info);
1259 BUG_ON(ptr > end);
1260 } else {
1261 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_DATA));
1262 }
1263
1264 err = -ENOENT;
1265 while (1) {
1266 if (ptr >= end) {
1267 WARN_ON(ptr > end);
1268 break;
1269 }
1270 iref = (struct btrfs_extent_inline_ref *)ptr;
1271 type = btrfs_extent_inline_ref_type(leaf, iref);
1272 if (want < type)
1273 break;
1274 if (want > type) {
1275 ptr += btrfs_extent_inline_ref_size(type);
1276 continue;
1277 }
1278
1279 if (type == BTRFS_EXTENT_DATA_REF_KEY) {
1280 struct btrfs_extent_data_ref *dref;
1281 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1282 if (match_extent_data_ref(leaf, dref, root_objectid,
1283 owner, offset)) {
1284 err = 0;
1285 break;
1286 }
1287 if (hash_extent_data_ref_item(leaf, dref) <
1288 hash_extent_data_ref(root_objectid, owner, offset))
1289 break;
1290 } else {
1291 u64 ref_offset;
1292 ref_offset = btrfs_extent_inline_ref_offset(leaf, iref);
1293 if (parent > 0) {
1294 if (parent == ref_offset) {
1295 err = 0;
1296 break;
1297 }
1298 if (ref_offset < parent)
1299 break;
1300 } else {
1301 if (root_objectid == ref_offset) {
1302 err = 0;
1303 break;
1304 }
1305 if (ref_offset < root_objectid)
1306 break;
1307 }
1308 }
1309 ptr += btrfs_extent_inline_ref_size(type);
1310 }
1311 if (err == -ENOENT && insert) {
1312 if (item_size + extra_size >=
1313 BTRFS_MAX_EXTENT_ITEM_SIZE(root)) {
1314 err = -EAGAIN;
1315 goto out;
1316 }
1317 /*
1318 * To add new inline back ref, we have to make sure
1319 * there is no corresponding back ref item.
1320 * For simplicity, we just do not add new inline back
1321 * ref if there is any kind of item for this block
1322 */
1323 if (find_next_key(path, 0, &key) == 0 &&
1324 key.objectid == bytenr &&
1325 key.type < BTRFS_BLOCK_GROUP_ITEM_KEY) {
1326 err = -EAGAIN;
1327 goto out;
1328 }
1329 }
1330 *ref_ret = (struct btrfs_extent_inline_ref *)ptr;
1331 out:
1332 if (insert) {
1333 path->keep_locks = 0;
1334 btrfs_unlock_up_safe(path, 1);
1335 }
1336 return err;
1337 }
1338
1339 /*
1340 * helper to add new inline back ref
1341 */
1342 static noinline_for_stack
1343 int setup_inline_extent_backref(struct btrfs_trans_handle *trans,
1344 struct btrfs_root *root,
1345 struct btrfs_path *path,
1346 struct btrfs_extent_inline_ref *iref,
1347 u64 parent, u64 root_objectid,
1348 u64 owner, u64 offset, int refs_to_add,
1349 struct btrfs_delayed_extent_op *extent_op)
1350 {
1351 struct extent_buffer *leaf;
1352 struct btrfs_extent_item *ei;
1353 unsigned long ptr;
1354 unsigned long end;
1355 unsigned long item_offset;
1356 u64 refs;
1357 int size;
1358 int type;
1359 int ret;
1360
1361 leaf = path->nodes[0];
1362 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1363 item_offset = (unsigned long)iref - (unsigned long)ei;
1364
1365 type = extent_ref_type(parent, owner);
1366 size = btrfs_extent_inline_ref_size(type);
1367
1368 ret = btrfs_extend_item(trans, root, path, size);
1369 BUG_ON(ret);
1370
1371 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1372 refs = btrfs_extent_refs(leaf, ei);
1373 refs += refs_to_add;
1374 btrfs_set_extent_refs(leaf, ei, refs);
1375 if (extent_op)
1376 __run_delayed_extent_op(extent_op, leaf, ei);
1377
1378 ptr = (unsigned long)ei + item_offset;
1379 end = (unsigned long)ei + btrfs_item_size_nr(leaf, path->slots[0]);
1380 if (ptr < end - size)
1381 memmove_extent_buffer(leaf, ptr + size, ptr,
1382 end - size - ptr);
1383
1384 iref = (struct btrfs_extent_inline_ref *)ptr;
1385 btrfs_set_extent_inline_ref_type(leaf, iref, type);
1386 if (type == BTRFS_EXTENT_DATA_REF_KEY) {
1387 struct btrfs_extent_data_ref *dref;
1388 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1389 btrfs_set_extent_data_ref_root(leaf, dref, root_objectid);
1390 btrfs_set_extent_data_ref_objectid(leaf, dref, owner);
1391 btrfs_set_extent_data_ref_offset(leaf, dref, offset);
1392 btrfs_set_extent_data_ref_count(leaf, dref, refs_to_add);
1393 } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
1394 struct btrfs_shared_data_ref *sref;
1395 sref = (struct btrfs_shared_data_ref *)(iref + 1);
1396 btrfs_set_shared_data_ref_count(leaf, sref, refs_to_add);
1397 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
1398 } else if (type == BTRFS_SHARED_BLOCK_REF_KEY) {
1399 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
1400 } else {
1401 btrfs_set_extent_inline_ref_offset(leaf, iref, root_objectid);
1402 }
1403 btrfs_mark_buffer_dirty(leaf);
1404 return 0;
1405 }
1406
1407 static int lookup_extent_backref(struct btrfs_trans_handle *trans,
1408 struct btrfs_root *root,
1409 struct btrfs_path *path,
1410 struct btrfs_extent_inline_ref **ref_ret,
1411 u64 bytenr, u64 num_bytes, u64 parent,
1412 u64 root_objectid, u64 owner, u64 offset)
1413 {
1414 int ret;
1415
1416 ret = lookup_inline_extent_backref(trans, root, path, ref_ret,
1417 bytenr, num_bytes, parent,
1418 root_objectid, owner, offset, 0);
1419 if (ret != -ENOENT)
1420 return ret;
1421
1422 btrfs_release_path(root, path);
1423 *ref_ret = NULL;
1424
1425 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1426 ret = lookup_tree_block_ref(trans, root, path, bytenr, parent,
1427 root_objectid);
1428 } else {
1429 ret = lookup_extent_data_ref(trans, root, path, bytenr, parent,
1430 root_objectid, owner, offset);
1431 }
1432 return ret;
1433 }
1434
1435 /*
1436 * helper to update/remove inline back ref
1437 */
1438 static noinline_for_stack
1439 int update_inline_extent_backref(struct btrfs_trans_handle *trans,
1440 struct btrfs_root *root,
1441 struct btrfs_path *path,
1442 struct btrfs_extent_inline_ref *iref,
1443 int refs_to_mod,
1444 struct btrfs_delayed_extent_op *extent_op)
1445 {
1446 struct extent_buffer *leaf;
1447 struct btrfs_extent_item *ei;
1448 struct btrfs_extent_data_ref *dref = NULL;
1449 struct btrfs_shared_data_ref *sref = NULL;
1450 unsigned long ptr;
1451 unsigned long end;
1452 u32 item_size;
1453 int size;
1454 int type;
1455 int ret;
1456 u64 refs;
1457
1458 leaf = path->nodes[0];
1459 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1460 refs = btrfs_extent_refs(leaf, ei);
1461 WARN_ON(refs_to_mod < 0 && refs + refs_to_mod <= 0);
1462 refs += refs_to_mod;
1463 btrfs_set_extent_refs(leaf, ei, refs);
1464 if (extent_op)
1465 __run_delayed_extent_op(extent_op, leaf, ei);
1466
1467 type = btrfs_extent_inline_ref_type(leaf, iref);
1468
1469 if (type == BTRFS_EXTENT_DATA_REF_KEY) {
1470 dref = (struct btrfs_extent_data_ref *)(&iref->offset);
1471 refs = btrfs_extent_data_ref_count(leaf, dref);
1472 } else if (type == BTRFS_SHARED_DATA_REF_KEY) {
1473 sref = (struct btrfs_shared_data_ref *)(iref + 1);
1474 refs = btrfs_shared_data_ref_count(leaf, sref);
1475 } else {
1476 refs = 1;
1477 BUG_ON(refs_to_mod != -1);
1478 }
1479
1480 BUG_ON(refs_to_mod < 0 && refs < -refs_to_mod);
1481 refs += refs_to_mod;
1482
1483 if (refs > 0) {
1484 if (type == BTRFS_EXTENT_DATA_REF_KEY)
1485 btrfs_set_extent_data_ref_count(leaf, dref, refs);
1486 else
1487 btrfs_set_shared_data_ref_count(leaf, sref, refs);
1488 } else {
1489 size = btrfs_extent_inline_ref_size(type);
1490 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1491 ptr = (unsigned long)iref;
1492 end = (unsigned long)ei + item_size;
1493 if (ptr + size < end)
1494 memmove_extent_buffer(leaf, ptr, ptr + size,
1495 end - ptr - size);
1496 item_size -= size;
1497 ret = btrfs_truncate_item(trans, root, path, item_size, 1);
1498 BUG_ON(ret);
1499 }
1500 btrfs_mark_buffer_dirty(leaf);
1501 return 0;
1502 }
1503
1504 static noinline_for_stack
1505 int insert_inline_extent_backref(struct btrfs_trans_handle *trans,
1506 struct btrfs_root *root,
1507 struct btrfs_path *path,
1508 u64 bytenr, u64 num_bytes, u64 parent,
1509 u64 root_objectid, u64 owner,
1510 u64 offset, int refs_to_add,
1511 struct btrfs_delayed_extent_op *extent_op)
1512 {
1513 struct btrfs_extent_inline_ref *iref;
1514 int ret;
1515
1516 ret = lookup_inline_extent_backref(trans, root, path, &iref,
1517 bytenr, num_bytes, parent,
1518 root_objectid, owner, offset, 1);
1519 if (ret == 0) {
1520 BUG_ON(owner < BTRFS_FIRST_FREE_OBJECTID);
1521 ret = update_inline_extent_backref(trans, root, path, iref,
1522 refs_to_add, extent_op);
1523 } else if (ret == -ENOENT) {
1524 ret = setup_inline_extent_backref(trans, root, path, iref,
1525 parent, root_objectid,
1526 owner, offset, refs_to_add,
1527 extent_op);
1528 }
1529 return ret;
1530 }
1531
1532 static int insert_extent_backref(struct btrfs_trans_handle *trans,
1533 struct btrfs_root *root,
1534 struct btrfs_path *path,
1535 u64 bytenr, u64 parent, u64 root_objectid,
1536 u64 owner, u64 offset, int refs_to_add)
1537 {
1538 int ret;
1539 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1540 BUG_ON(refs_to_add != 1);
1541 ret = insert_tree_block_ref(trans, root, path, bytenr,
1542 parent, root_objectid);
1543 } else {
1544 ret = insert_extent_data_ref(trans, root, path, bytenr,
1545 parent, root_objectid,
1546 owner, offset, refs_to_add);
1547 }
1548 return ret;
1549 }
1550
1551 static int remove_extent_backref(struct btrfs_trans_handle *trans,
1552 struct btrfs_root *root,
1553 struct btrfs_path *path,
1554 struct btrfs_extent_inline_ref *iref,
1555 int refs_to_drop, int is_data)
1556 {
1557 int ret;
1558
1559 BUG_ON(!is_data && refs_to_drop != 1);
1560 if (iref) {
1561 ret = update_inline_extent_backref(trans, root, path, iref,
1562 -refs_to_drop, NULL);
1563 } else if (is_data) {
1564 ret = remove_extent_data_ref(trans, root, path, refs_to_drop);
1565 } else {
1566 ret = btrfs_del_item(trans, root, path);
1567 }
1568 return ret;
1569 }
1570
1571 static void btrfs_issue_discard(struct block_device *bdev,
1572 u64 start, u64 len)
1573 {
1574 blkdev_issue_discard(bdev, start >> 9, len >> 9, GFP_KERNEL,
1575 DISCARD_FL_BARRIER);
1576 }
1577
1578 static int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr,
1579 u64 num_bytes)
1580 {
1581 int ret;
1582 u64 map_length = num_bytes;
1583 struct btrfs_multi_bio *multi = NULL;
1584
1585 if (!btrfs_test_opt(root, DISCARD))
1586 return 0;
1587
1588 /* Tell the block device(s) that the sectors can be discarded */
1589 ret = btrfs_map_block(&root->fs_info->mapping_tree, READ,
1590 bytenr, &map_length, &multi, 0);
1591 if (!ret) {
1592 struct btrfs_bio_stripe *stripe = multi->stripes;
1593 int i;
1594
1595 if (map_length > num_bytes)
1596 map_length = num_bytes;
1597
1598 for (i = 0; i < multi->num_stripes; i++, stripe++) {
1599 btrfs_issue_discard(stripe->dev->bdev,
1600 stripe->physical,
1601 map_length);
1602 }
1603 kfree(multi);
1604 }
1605
1606 return ret;
1607 }
1608
1609 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
1610 struct btrfs_root *root,
1611 u64 bytenr, u64 num_bytes, u64 parent,
1612 u64 root_objectid, u64 owner, u64 offset)
1613 {
1614 int ret;
1615 BUG_ON(owner < BTRFS_FIRST_FREE_OBJECTID &&
1616 root_objectid == BTRFS_TREE_LOG_OBJECTID);
1617
1618 if (owner < BTRFS_FIRST_FREE_OBJECTID) {
1619 ret = btrfs_add_delayed_tree_ref(trans, bytenr, num_bytes,
1620 parent, root_objectid, (int)owner,
1621 BTRFS_ADD_DELAYED_REF, NULL);
1622 } else {
1623 ret = btrfs_add_delayed_data_ref(trans, bytenr, num_bytes,
1624 parent, root_objectid, owner, offset,
1625 BTRFS_ADD_DELAYED_REF, NULL);
1626 }
1627 return ret;
1628 }
1629
1630 static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
1631 struct btrfs_root *root,
1632 u64 bytenr, u64 num_bytes,
1633 u64 parent, u64 root_objectid,
1634 u64 owner, u64 offset, int refs_to_add,
1635 struct btrfs_delayed_extent_op *extent_op)
1636 {
1637 struct btrfs_path *path;
1638 struct extent_buffer *leaf;
1639 struct btrfs_extent_item *item;
1640 u64 refs;
1641 int ret;
1642 int err = 0;
1643
1644 path = btrfs_alloc_path();
1645 if (!path)
1646 return -ENOMEM;
1647
1648 path->reada = 1;
1649 path->leave_spinning = 1;
1650 /* this will setup the path even if it fails to insert the back ref */
1651 ret = insert_inline_extent_backref(trans, root->fs_info->extent_root,
1652 path, bytenr, num_bytes, parent,
1653 root_objectid, owner, offset,
1654 refs_to_add, extent_op);
1655 if (ret == 0)
1656 goto out;
1657
1658 if (ret != -EAGAIN) {
1659 err = ret;
1660 goto out;
1661 }
1662
1663 leaf = path->nodes[0];
1664 item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1665 refs = btrfs_extent_refs(leaf, item);
1666 btrfs_set_extent_refs(leaf, item, refs + refs_to_add);
1667 if (extent_op)
1668 __run_delayed_extent_op(extent_op, leaf, item);
1669
1670 btrfs_mark_buffer_dirty(leaf);
1671 btrfs_release_path(root->fs_info->extent_root, path);
1672
1673 path->reada = 1;
1674 path->leave_spinning = 1;
1675
1676 /* now insert the actual backref */
1677 ret = insert_extent_backref(trans, root->fs_info->extent_root,
1678 path, bytenr, parent, root_objectid,
1679 owner, offset, refs_to_add);
1680 BUG_ON(ret);
1681 out:
1682 btrfs_free_path(path);
1683 return err;
1684 }
1685
1686 static int run_delayed_data_ref(struct btrfs_trans_handle *trans,
1687 struct btrfs_root *root,
1688 struct btrfs_delayed_ref_node *node,
1689 struct btrfs_delayed_extent_op *extent_op,
1690 int insert_reserved)
1691 {
1692 int ret = 0;
1693 struct btrfs_delayed_data_ref *ref;
1694 struct btrfs_key ins;
1695 u64 parent = 0;
1696 u64 ref_root = 0;
1697 u64 flags = 0;
1698
1699 ins.objectid = node->bytenr;
1700 ins.offset = node->num_bytes;
1701 ins.type = BTRFS_EXTENT_ITEM_KEY;
1702
1703 ref = btrfs_delayed_node_to_data_ref(node);
1704 if (node->type == BTRFS_SHARED_DATA_REF_KEY)
1705 parent = ref->parent;
1706 else
1707 ref_root = ref->root;
1708
1709 if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
1710 if (extent_op) {
1711 BUG_ON(extent_op->update_key);
1712 flags |= extent_op->flags_to_set;
1713 }
1714 ret = alloc_reserved_file_extent(trans, root,
1715 parent, ref_root, flags,
1716 ref->objectid, ref->offset,
1717 &ins, node->ref_mod);
1718 } else if (node->action == BTRFS_ADD_DELAYED_REF) {
1719 ret = __btrfs_inc_extent_ref(trans, root, node->bytenr,
1720 node->num_bytes, parent,
1721 ref_root, ref->objectid,
1722 ref->offset, node->ref_mod,
1723 extent_op);
1724 } else if (node->action == BTRFS_DROP_DELAYED_REF) {
1725 ret = __btrfs_free_extent(trans, root, node->bytenr,
1726 node->num_bytes, parent,
1727 ref_root, ref->objectid,
1728 ref->offset, node->ref_mod,
1729 extent_op);
1730 } else {
1731 BUG();
1732 }
1733 return ret;
1734 }
1735
1736 static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
1737 struct extent_buffer *leaf,
1738 struct btrfs_extent_item *ei)
1739 {
1740 u64 flags = btrfs_extent_flags(leaf, ei);
1741 if (extent_op->update_flags) {
1742 flags |= extent_op->flags_to_set;
1743 btrfs_set_extent_flags(leaf, ei, flags);
1744 }
1745
1746 if (extent_op->update_key) {
1747 struct btrfs_tree_block_info *bi;
1748 BUG_ON(!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK));
1749 bi = (struct btrfs_tree_block_info *)(ei + 1);
1750 btrfs_set_tree_block_key(leaf, bi, &extent_op->key);
1751 }
1752 }
1753
1754 static int run_delayed_extent_op(struct btrfs_trans_handle *trans,
1755 struct btrfs_root *root,
1756 struct btrfs_delayed_ref_node *node,
1757 struct btrfs_delayed_extent_op *extent_op)
1758 {
1759 struct btrfs_key key;
1760 struct btrfs_path *path;
1761 struct btrfs_extent_item *ei;
1762 struct extent_buffer *leaf;
1763 u32 item_size;
1764 int ret;
1765 int err = 0;
1766
1767 path = btrfs_alloc_path();
1768 if (!path)
1769 return -ENOMEM;
1770
1771 key.objectid = node->bytenr;
1772 key.type = BTRFS_EXTENT_ITEM_KEY;
1773 key.offset = node->num_bytes;
1774
1775 path->reada = 1;
1776 path->leave_spinning = 1;
1777 ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key,
1778 path, 0, 1);
1779 if (ret < 0) {
1780 err = ret;
1781 goto out;
1782 }
1783 if (ret > 0) {
1784 err = -EIO;
1785 goto out;
1786 }
1787
1788 leaf = path->nodes[0];
1789 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1790 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
1791 if (item_size < sizeof(*ei)) {
1792 ret = convert_extent_item_v0(trans, root->fs_info->extent_root,
1793 path, (u64)-1, 0);
1794 if (ret < 0) {
1795 err = ret;
1796 goto out;
1797 }
1798 leaf = path->nodes[0];
1799 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
1800 }
1801 #endif
1802 BUG_ON(item_size < sizeof(*ei));
1803 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
1804 __run_delayed_extent_op(extent_op, leaf, ei);
1805
1806 btrfs_mark_buffer_dirty(leaf);
1807 out:
1808 btrfs_free_path(path);
1809 return err;
1810 }
1811
1812 static int run_delayed_tree_ref(struct btrfs_trans_handle *trans,
1813 struct btrfs_root *root,
1814 struct btrfs_delayed_ref_node *node,
1815 struct btrfs_delayed_extent_op *extent_op,
1816 int insert_reserved)
1817 {
1818 int ret = 0;
1819 struct btrfs_delayed_tree_ref *ref;
1820 struct btrfs_key ins;
1821 u64 parent = 0;
1822 u64 ref_root = 0;
1823
1824 ins.objectid = node->bytenr;
1825 ins.offset = node->num_bytes;
1826 ins.type = BTRFS_EXTENT_ITEM_KEY;
1827
1828 ref = btrfs_delayed_node_to_tree_ref(node);
1829 if (node->type == BTRFS_SHARED_BLOCK_REF_KEY)
1830 parent = ref->parent;
1831 else
1832 ref_root = ref->root;
1833
1834 BUG_ON(node->ref_mod != 1);
1835 if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
1836 BUG_ON(!extent_op || !extent_op->update_flags ||
1837 !extent_op->update_key);
1838 ret = alloc_reserved_tree_block(trans, root,
1839 parent, ref_root,
1840 extent_op->flags_to_set,
1841 &extent_op->key,
1842 ref->level, &ins);
1843 } else if (node->action == BTRFS_ADD_DELAYED_REF) {
1844 ret = __btrfs_inc_extent_ref(trans, root, node->bytenr,
1845 node->num_bytes, parent, ref_root,
1846 ref->level, 0, 1, extent_op);
1847 } else if (node->action == BTRFS_DROP_DELAYED_REF) {
1848 ret = __btrfs_free_extent(trans, root, node->bytenr,
1849 node->num_bytes, parent, ref_root,
1850 ref->level, 0, 1, extent_op);
1851 } else {
1852 BUG();
1853 }
1854 return ret;
1855 }
1856
1857
1858 /* helper function to actually process a single delayed ref entry */
1859 static int run_one_delayed_ref(struct btrfs_trans_handle *trans,
1860 struct btrfs_root *root,
1861 struct btrfs_delayed_ref_node *node,
1862 struct btrfs_delayed_extent_op *extent_op,
1863 int insert_reserved)
1864 {
1865 int ret;
1866 if (btrfs_delayed_ref_is_head(node)) {
1867 struct btrfs_delayed_ref_head *head;
1868 /*
1869 * we've hit the end of the chain and we were supposed
1870 * to insert this extent into the tree. But, it got
1871 * deleted before we ever needed to insert it, so all
1872 * we have to do is clean up the accounting
1873 */
1874 BUG_ON(extent_op);
1875 head = btrfs_delayed_node_to_head(node);
1876 if (insert_reserved) {
1877 int mark_free = 0;
1878 struct extent_buffer *must_clean = NULL;
1879
1880 ret = pin_down_bytes(trans, root, NULL,
1881 node->bytenr, node->num_bytes,
1882 head->is_data, 1, &must_clean);
1883 if (ret > 0)
1884 mark_free = 1;
1885
1886 if (must_clean) {
1887 clean_tree_block(NULL, root, must_clean);
1888 btrfs_tree_unlock(must_clean);
1889 free_extent_buffer(must_clean);
1890 }
1891 if (head->is_data) {
1892 ret = btrfs_del_csums(trans, root,
1893 node->bytenr,
1894 node->num_bytes);
1895 BUG_ON(ret);
1896 }
1897 if (mark_free) {
1898 ret = btrfs_free_reserved_extent(root,
1899 node->bytenr,
1900 node->num_bytes);
1901 BUG_ON(ret);
1902 }
1903 }
1904 mutex_unlock(&head->mutex);
1905 return 0;
1906 }
1907
1908 if (node->type == BTRFS_TREE_BLOCK_REF_KEY ||
1909 node->type == BTRFS_SHARED_BLOCK_REF_KEY)
1910 ret = run_delayed_tree_ref(trans, root, node, extent_op,
1911 insert_reserved);
1912 else if (node->type == BTRFS_EXTENT_DATA_REF_KEY ||
1913 node->type == BTRFS_SHARED_DATA_REF_KEY)
1914 ret = run_delayed_data_ref(trans, root, node, extent_op,
1915 insert_reserved);
1916 else
1917 BUG();
1918 return ret;
1919 }
1920
1921 static noinline struct btrfs_delayed_ref_node *
1922 select_delayed_ref(struct btrfs_delayed_ref_head *head)
1923 {
1924 struct rb_node *node;
1925 struct btrfs_delayed_ref_node *ref;
1926 int action = BTRFS_ADD_DELAYED_REF;
1927 again:
1928 /*
1929 * select delayed ref of type BTRFS_ADD_DELAYED_REF first.
1930 * this prevents ref count from going down to zero when
1931 * there still are pending delayed ref.
1932 */
1933 node = rb_prev(&head->node.rb_node);
1934 while (1) {
1935 if (!node)
1936 break;
1937 ref = rb_entry(node, struct btrfs_delayed_ref_node,
1938 rb_node);
1939 if (ref->bytenr != head->node.bytenr)
1940 break;
1941 if (ref->action == action)
1942 return ref;
1943 node = rb_prev(node);
1944 }
1945 if (action == BTRFS_ADD_DELAYED_REF) {
1946 action = BTRFS_DROP_DELAYED_REF;
1947 goto again;
1948 }
1949 return NULL;
1950 }
1951
1952 static noinline int run_clustered_refs(struct btrfs_trans_handle *trans,
1953 struct btrfs_root *root,
1954 struct list_head *cluster)
1955 {
1956 struct btrfs_delayed_ref_root *delayed_refs;
1957 struct btrfs_delayed_ref_node *ref;
1958 struct btrfs_delayed_ref_head *locked_ref = NULL;
1959 struct btrfs_delayed_extent_op *extent_op;
1960 int ret;
1961 int count = 0;
1962 int must_insert_reserved = 0;
1963
1964 delayed_refs = &trans->transaction->delayed_refs;
1965 while (1) {
1966 if (!locked_ref) {
1967 /* pick a new head ref from the cluster list */
1968 if (list_empty(cluster))
1969 break;
1970
1971 locked_ref = list_entry(cluster->next,
1972 struct btrfs_delayed_ref_head, cluster);
1973
1974 /* grab the lock that says we are going to process
1975 * all the refs for this head */
1976 ret = btrfs_delayed_ref_lock(trans, locked_ref);
1977
1978 /*
1979 * we may have dropped the spin lock to get the head
1980 * mutex lock, and that might have given someone else
1981 * time to free the head. If that's true, it has been
1982 * removed from our list and we can move on.
1983 */
1984 if (ret == -EAGAIN) {
1985 locked_ref = NULL;
1986 count++;
1987 continue;
1988 }
1989 }
1990
1991 /*
1992 * record the must insert reserved flag before we
1993 * drop the spin lock.
1994 */
1995 must_insert_reserved = locked_ref->must_insert_reserved;
1996 locked_ref->must_insert_reserved = 0;
1997
1998 extent_op = locked_ref->extent_op;
1999 locked_ref->extent_op = NULL;
2000
2001 /*
2002 * locked_ref is the head node, so we have to go one
2003 * node back for any delayed ref updates
2004 */
2005 ref = select_delayed_ref(locked_ref);
2006 if (!ref) {
2007 /* All delayed refs have been processed, Go ahead
2008 * and send the head node to run_one_delayed_ref,
2009 * so that any accounting fixes can happen
2010 */
2011 ref = &locked_ref->node;
2012
2013 if (extent_op && must_insert_reserved) {
2014 kfree(extent_op);
2015 extent_op = NULL;
2016 }
2017
2018 if (extent_op) {
2019 spin_unlock(&delayed_refs->lock);
2020
2021 ret = run_delayed_extent_op(trans, root,
2022 ref, extent_op);
2023 BUG_ON(ret);
2024 kfree(extent_op);
2025
2026 cond_resched();
2027 spin_lock(&delayed_refs->lock);
2028 continue;
2029 }
2030
2031 list_del_init(&locked_ref->cluster);
2032 locked_ref = NULL;
2033 }
2034
2035 ref->in_tree = 0;
2036 rb_erase(&ref->rb_node, &delayed_refs->root);
2037 delayed_refs->num_entries--;
2038
2039 spin_unlock(&delayed_refs->lock);
2040
2041 ret = run_one_delayed_ref(trans, root, ref, extent_op,
2042 must_insert_reserved);
2043 BUG_ON(ret);
2044
2045 btrfs_put_delayed_ref(ref);
2046 kfree(extent_op);
2047 count++;
2048
2049 cond_resched();
2050 spin_lock(&delayed_refs->lock);
2051 }
2052 return count;
2053 }
2054
2055 /*
2056 * this starts processing the delayed reference count updates and
2057 * extent insertions we have queued up so far. count can be
2058 * 0, which means to process everything in the tree at the start
2059 * of the run (but not newly added entries), or it can be some target
2060 * number you'd like to process.
2061 */
2062 int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
2063 struct btrfs_root *root, unsigned long count)
2064 {
2065 struct rb_node *node;
2066 struct btrfs_delayed_ref_root *delayed_refs;
2067 struct btrfs_delayed_ref_node *ref;
2068 struct list_head cluster;
2069 int ret;
2070 int run_all = count == (unsigned long)-1;
2071 int run_most = 0;
2072
2073 if (root == root->fs_info->extent_root)
2074 root = root->fs_info->tree_root;
2075
2076 delayed_refs = &trans->transaction->delayed_refs;
2077 INIT_LIST_HEAD(&cluster);
2078 again:
2079 spin_lock(&delayed_refs->lock);
2080 if (count == 0) {
2081 count = delayed_refs->num_entries * 2;
2082 run_most = 1;
2083 }
2084 while (1) {
2085 if (!(run_all || run_most) &&
2086 delayed_refs->num_heads_ready < 64)
2087 break;
2088
2089 /*
2090 * go find something we can process in the rbtree. We start at
2091 * the beginning of the tree, and then build a cluster
2092 * of refs to process starting at the first one we are able to
2093 * lock
2094 */
2095 ret = btrfs_find_ref_cluster(trans, &cluster,
2096 delayed_refs->run_delayed_start);
2097 if (ret)
2098 break;
2099
2100 ret = run_clustered_refs(trans, root, &cluster);
2101 BUG_ON(ret < 0);
2102
2103 count -= min_t(unsigned long, ret, count);
2104
2105 if (count == 0)
2106 break;
2107 }
2108
2109 if (run_all) {
2110 node = rb_first(&delayed_refs->root);
2111 if (!node)
2112 goto out;
2113 count = (unsigned long)-1;
2114
2115 while (node) {
2116 ref = rb_entry(node, struct btrfs_delayed_ref_node,
2117 rb_node);
2118 if (btrfs_delayed_ref_is_head(ref)) {
2119 struct btrfs_delayed_ref_head *head;
2120
2121 head = btrfs_delayed_node_to_head(ref);
2122 atomic_inc(&ref->refs);
2123
2124 spin_unlock(&delayed_refs->lock);
2125 mutex_lock(&head->mutex);
2126 mutex_unlock(&head->mutex);
2127
2128 btrfs_put_delayed_ref(ref);
2129 cond_resched();
2130 goto again;
2131 }
2132 node = rb_next(node);
2133 }
2134 spin_unlock(&delayed_refs->lock);
2135 schedule_timeout(1);
2136 goto again;
2137 }
2138 out:
2139 spin_unlock(&delayed_refs->lock);
2140 return 0;
2141 }
2142
2143 int btrfs_set_disk_extent_flags(struct btrfs_trans_handle *trans,
2144 struct btrfs_root *root,
2145 u64 bytenr, u64 num_bytes, u64 flags,
2146 int is_data)
2147 {
2148 struct btrfs_delayed_extent_op *extent_op;
2149 int ret;
2150
2151 extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
2152 if (!extent_op)
2153 return -ENOMEM;
2154
2155 extent_op->flags_to_set = flags;
2156 extent_op->update_flags = 1;
2157 extent_op->update_key = 0;
2158 extent_op->is_data = is_data ? 1 : 0;
2159
2160 ret = btrfs_add_delayed_extent_op(trans, bytenr, num_bytes, extent_op);
2161 if (ret)
2162 kfree(extent_op);
2163 return ret;
2164 }
2165
2166 static noinline int check_delayed_ref(struct btrfs_trans_handle *trans,
2167 struct btrfs_root *root,
2168 struct btrfs_path *path,
2169 u64 objectid, u64 offset, u64 bytenr)
2170 {
2171 struct btrfs_delayed_ref_head *head;
2172 struct btrfs_delayed_ref_node *ref;
2173 struct btrfs_delayed_data_ref *data_ref;
2174 struct btrfs_delayed_ref_root *delayed_refs;
2175 struct rb_node *node;
2176 int ret = 0;
2177
2178 ret = -ENOENT;
2179 delayed_refs = &trans->transaction->delayed_refs;
2180 spin_lock(&delayed_refs->lock);
2181 head = btrfs_find_delayed_ref_head(trans, bytenr);
2182 if (!head)
2183 goto out;
2184
2185 if (!mutex_trylock(&head->mutex)) {
2186 atomic_inc(&head->node.refs);
2187 spin_unlock(&delayed_refs->lock);
2188
2189 btrfs_release_path(root->fs_info->extent_root, path);
2190
2191 mutex_lock(&head->mutex);
2192 mutex_unlock(&head->mutex);
2193 btrfs_put_delayed_ref(&head->node);
2194 return -EAGAIN;
2195 }
2196
2197 node = rb_prev(&head->node.rb_node);
2198 if (!node)
2199 goto out_unlock;
2200
2201 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
2202
2203 if (ref->bytenr != bytenr)
2204 goto out_unlock;
2205
2206 ret = 1;
2207 if (ref->type != BTRFS_EXTENT_DATA_REF_KEY)
2208 goto out_unlock;
2209
2210 data_ref = btrfs_delayed_node_to_data_ref(ref);
2211
2212 node = rb_prev(node);
2213 if (node) {
2214 ref = rb_entry(node, struct btrfs_delayed_ref_node, rb_node);
2215 if (ref->bytenr == bytenr)
2216 goto out_unlock;
2217 }
2218
2219 if (data_ref->root != root->root_key.objectid ||
2220 data_ref->objectid != objectid || data_ref->offset != offset)
2221 goto out_unlock;
2222
2223 ret = 0;
2224 out_unlock:
2225 mutex_unlock(&head->mutex);
2226 out:
2227 spin_unlock(&delayed_refs->lock);
2228 return ret;
2229 }
2230
2231 static noinline int check_committed_ref(struct btrfs_trans_handle *trans,
2232 struct btrfs_root *root,
2233 struct btrfs_path *path,
2234 u64 objectid, u64 offset, u64 bytenr)
2235 {
2236 struct btrfs_root *extent_root = root->fs_info->extent_root;
2237 struct extent_buffer *leaf;
2238 struct btrfs_extent_data_ref *ref;
2239 struct btrfs_extent_inline_ref *iref;
2240 struct btrfs_extent_item *ei;
2241 struct btrfs_key key;
2242 u32 item_size;
2243 int ret;
2244
2245 key.objectid = bytenr;
2246 key.offset = (u64)-1;
2247 key.type = BTRFS_EXTENT_ITEM_KEY;
2248
2249 ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
2250 if (ret < 0)
2251 goto out;
2252 BUG_ON(ret == 0);
2253
2254 ret = -ENOENT;
2255 if (path->slots[0] == 0)
2256 goto out;
2257
2258 path->slots[0]--;
2259 leaf = path->nodes[0];
2260 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
2261
2262 if (key.objectid != bytenr || key.type != BTRFS_EXTENT_ITEM_KEY)
2263 goto out;
2264
2265 ret = 1;
2266 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
2267 #ifdef BTRFS_COMPAT_EXTENT_TREE_V0
2268 if (item_size < sizeof(*ei)) {
2269 WARN_ON(item_size != sizeof(struct btrfs_extent_item_v0));
2270 goto out;
2271 }
2272 #endif
2273 ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
2274
2275 if (item_size != sizeof(*ei) +
2276 btrfs_extent_inline_ref_size(BTRFS_EXTENT_DATA_REF_KEY))
2277 goto out;
2278
2279 if (btrfs_extent_generation(leaf, ei) <=
2280 btrfs_root_last_snapshot(&root->root_item))
2281 goto out;
2282
2283 iref = (struct btrfs_extent_inline_ref *)(ei + 1);
2284 if (btrfs_extent_inline_ref_type(leaf, iref) !=
2285 BTRFS_EXTENT_DATA_REF_KEY)
2286 goto out;
2287
2288 ref = (struct btrfs_extent_data_ref *)(&iref->offset);
2289 if (btrfs_extent_refs(leaf, ei) !=
2290 btrfs_extent_data_ref_count(leaf, ref) ||
2291 btrfs_extent_data_ref_root(leaf, ref) !=
2292 root->root_key.objectid ||
2293 btrfs_extent_data_ref_objectid(leaf, ref) != objectid ||
2294 btrfs_extent_data_ref_offset(leaf, ref) != offset)
2295 goto out;
2296
2297 ret = 0;
2298 out:
2299 return ret;
2300 }
2301
2302 int btrfs_cross_ref_exist(struct btrfs_trans_handle *trans,
2303 struct btrfs_root *root,
2304 u64 objectid, u64 offset, u64 bytenr)
2305 {
2306 struct btrfs_path *path;
2307 int ret;
2308 int ret2;
2309
2310 path = btrfs_alloc_path();
2311 if (!path)
2312 return -ENOENT;
2313
2314 do {
2315 ret = check_committed_ref(trans, root, path, objectid,
2316 offset, bytenr);
2317 if (ret && ret != -ENOENT)
2318 goto out;
2319
2320 ret2 = check_delayed_ref(trans, root, path, objectid,
2321 offset, bytenr);
2322 } while (ret2 == -EAGAIN);
2323
2324 if (ret2 && ret2 != -ENOENT) {
2325 ret = ret2;
2326 goto out;
2327 }
2328
2329 if (ret != -ENOENT || ret2 != -ENOENT)
2330 ret = 0;
2331 out:
2332 btrfs_free_path(path);
2333 return ret;
2334 }
2335
2336 #if 0
2337 int btrfs_cache_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2338 struct extent_buffer *buf, u32 nr_extents)
2339 {
2340 struct btrfs_key key;
2341 struct btrfs_file_extent_item *fi;
2342 u64 root_gen;
2343 u32 nritems;
2344 int i;
2345 int level;
2346 int ret = 0;
2347 int shared = 0;
2348
2349 if (!root->ref_cows)
2350 return 0;
2351
2352 if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
2353 shared = 0;
2354 root_gen = root->root_key.offset;
2355 } else {
2356 shared = 1;
2357 root_gen = trans->transid - 1;
2358 }
2359
2360 level = btrfs_header_level(buf);
2361 nritems = btrfs_header_nritems(buf);
2362
2363 if (level == 0) {
2364 struct btrfs_leaf_ref *ref;
2365 struct btrfs_extent_info *info;
2366
2367 ref = btrfs_alloc_leaf_ref(root, nr_extents);
2368 if (!ref) {
2369 ret = -ENOMEM;
2370 goto out;
2371 }
2372
2373 ref->root_gen = root_gen;
2374 ref->bytenr = buf->start;
2375 ref->owner = btrfs_header_owner(buf);
2376 ref->generation = btrfs_header_generation(buf);
2377 ref->nritems = nr_extents;
2378 info = ref->extents;
2379
2380 for (i = 0; nr_extents > 0 && i < nritems; i++) {
2381 u64 disk_bytenr;
2382 btrfs_item_key_to_cpu(buf, &key, i);
2383 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
2384 continue;
2385 fi = btrfs_item_ptr(buf, i,
2386 struct btrfs_file_extent_item);
2387 if (btrfs_file_extent_type(buf, fi) ==
2388 BTRFS_FILE_EXTENT_INLINE)
2389 continue;
2390 disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
2391 if (disk_bytenr == 0)
2392 continue;
2393
2394 info->bytenr = disk_bytenr;
2395 info->num_bytes =
2396 btrfs_file_extent_disk_num_bytes(buf, fi);
2397 info->objectid = key.objectid;
2398 info->offset = key.offset;
2399 info++;
2400 }
2401
2402 ret = btrfs_add_leaf_ref(root, ref, shared);
2403 if (ret == -EEXIST && shared) {
2404 struct btrfs_leaf_ref *old;
2405 old = btrfs_lookup_leaf_ref(root, ref->bytenr);
2406 BUG_ON(!old);
2407 btrfs_remove_leaf_ref(root, old);
2408 btrfs_free_leaf_ref(root, old);
2409 ret = btrfs_add_leaf_ref(root, ref, shared);
2410 }
2411 WARN_ON(ret);
2412 btrfs_free_leaf_ref(root, ref);
2413 }
2414 out:
2415 return ret;
2416 }
2417
2418 /* when a block goes through cow, we update the reference counts of
2419 * everything that block points to. The internal pointers of the block
2420 * can be in just about any order, and it is likely to have clusters of
2421 * things that are close together and clusters of things that are not.
2422 *
2423 * To help reduce the seeks that come with updating all of these reference
2424 * counts, sort them by byte number before actual updates are done.
2425 *
2426 * struct refsort is used to match byte number to slot in the btree block.
2427 * we sort based on the byte number and then use the slot to actually
2428 * find the item.
2429 *
2430 * struct refsort is smaller than strcut btrfs_item and smaller than
2431 * struct btrfs_key_ptr. Since we're currently limited to the page size
2432 * for a btree block, there's no way for a kmalloc of refsorts for a
2433 * single node to be bigger than a page.
2434 */
2435 struct refsort {
2436 u64 bytenr;
2437 u32 slot;
2438 };
2439
2440 /*
2441 * for passing into sort()
2442 */
2443 static int refsort_cmp(const void *a_void, const void *b_void)
2444 {
2445 const struct refsort *a = a_void;
2446 const struct refsort *b = b_void;
2447
2448 if (a->bytenr < b->bytenr)
2449 return -1;
2450 if (a->bytenr > b->bytenr)
2451 return 1;
2452 return 0;
2453 }
2454 #endif
2455
2456 static int __btrfs_mod_ref(struct btrfs_trans_handle *trans,
2457 struct btrfs_root *root,
2458 struct extent_buffer *buf,
2459 int full_backref, int inc)
2460 {
2461 u64 bytenr;
2462 u64 num_bytes;
2463 u64 parent;
2464 u64 ref_root;
2465 u32 nritems;
2466 struct btrfs_key key;
2467 struct btrfs_file_extent_item *fi;
2468 int i;
2469 int level;
2470 int ret = 0;
2471 int (*process_func)(struct btrfs_trans_handle *, struct btrfs_root *,
2472 u64, u64, u64, u64, u64, u64);
2473
2474 ref_root = btrfs_header_owner(buf);
2475 nritems = btrfs_header_nritems(buf);
2476 level = btrfs_header_level(buf);
2477
2478 if (!root->ref_cows && level == 0)
2479 return 0;
2480
2481 if (inc)
2482 process_func = btrfs_inc_extent_ref;
2483 else
2484 process_func = btrfs_free_extent;
2485
2486 if (full_backref)
2487 parent = buf->start;
2488 else
2489 parent = 0;
2490
2491 for (i = 0; i < nritems; i++) {
2492 if (level == 0) {
2493 btrfs_item_key_to_cpu(buf, &key, i);
2494 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
2495 continue;
2496 fi = btrfs_item_ptr(buf, i,
2497 struct btrfs_file_extent_item);
2498 if (btrfs_file_extent_type(buf, fi) ==
2499 BTRFS_FILE_EXTENT_INLINE)
2500 continue;
2501 bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
2502 if (bytenr == 0)
2503 continue;
2504
2505 num_bytes = btrfs_file_extent_disk_num_bytes(buf, fi);
2506 key.offset -= btrfs_file_extent_offset(buf, fi);
2507 ret = process_func(trans, root, bytenr, num_bytes,
2508 parent, ref_root, key.objectid,
2509 key.offset);
2510 if (ret)
2511 goto fail;
2512 } else {
2513 bytenr = btrfs_node_blockptr(buf, i);
2514 num_bytes = btrfs_level_size(root, level - 1);
2515 ret = process_func(trans, root, bytenr, num_bytes,
2516 parent, ref_root, level - 1, 0);
2517 if (ret)
2518 goto fail;
2519 }
2520 }
2521 return 0;
2522 fail:
2523 BUG();
2524 return ret;
2525 }
2526
2527 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2528 struct extent_buffer *buf, int full_backref)
2529 {
2530 return __btrfs_mod_ref(trans, root, buf, full_backref, 1);
2531 }
2532
2533 int btrfs_dec_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2534 struct extent_buffer *buf, int full_backref)
2535 {
2536 return __btrfs_mod_ref(trans, root, buf, full_backref, 0);
2537 }
2538
2539 static int write_one_cache_group(struct btrfs_trans_handle *trans,
2540 struct btrfs_root *root,
2541 struct btrfs_path *path,
2542 struct btrfs_block_group_cache *cache)
2543 {
2544 int ret;
2545 struct btrfs_root *extent_root = root->fs_info->extent_root;
2546 unsigned long bi;
2547 struct extent_buffer *leaf;
2548
2549 ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1);
2550 if (ret < 0)
2551 goto fail;
2552 BUG_ON(ret);
2553
2554 leaf = path->nodes[0];
2555 bi = btrfs_item_ptr_offset(leaf, path->slots[0]);
2556 write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item));
2557 btrfs_mark_buffer_dirty(leaf);
2558 btrfs_release_path(extent_root, path);
2559 fail:
2560 if (ret)
2561 return ret;
2562 return 0;
2563
2564 }
2565
2566 static struct btrfs_block_group_cache *
2567 next_block_group(struct btrfs_root *root,
2568 struct btrfs_block_group_cache *cache)
2569 {
2570 struct rb_node *node;
2571 spin_lock(&root->fs_info->block_group_cache_lock);
2572 node = rb_next(&cache->cache_node);
2573 btrfs_put_block_group(cache);
2574 if (node) {
2575 cache = rb_entry(node, struct btrfs_block_group_cache,
2576 cache_node);
2577 atomic_inc(&cache->count);
2578 } else
2579 cache = NULL;
2580 spin_unlock(&root->fs_info->block_group_cache_lock);
2581 return cache;
2582 }
2583
2584 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
2585 struct btrfs_root *root)
2586 {
2587 struct btrfs_block_group_cache *cache;
2588 int err = 0;
2589 struct btrfs_path *path;
2590 u64 last = 0;
2591
2592 path = btrfs_alloc_path();
2593 if (!path)
2594 return -ENOMEM;
2595
2596 while (1) {
2597 if (last == 0) {
2598 err = btrfs_run_delayed_refs(trans, root,
2599 (unsigned long)-1);
2600 BUG_ON(err);
2601 }
2602
2603 cache = btrfs_lookup_first_block_group(root->fs_info, last);
2604 while (cache) {
2605 if (cache->dirty)
2606 break;
2607 cache = next_block_group(root, cache);
2608 }
2609 if (!cache) {
2610 if (last == 0)
2611 break;
2612 last = 0;
2613 continue;
2614 }
2615
2616 cache->dirty = 0;
2617 last = cache->key.objectid + cache->key.offset;
2618
2619 err = write_one_cache_group(trans, root, path, cache);
2620 BUG_ON(err);
2621 btrfs_put_block_group(cache);
2622 }
2623
2624 btrfs_free_path(path);
2625 return 0;
2626 }
2627
2628 int btrfs_extent_readonly(struct btrfs_root *root, u64 bytenr)
2629 {
2630 struct btrfs_block_group_cache *block_group;
2631 int readonly = 0;
2632
2633 block_group = btrfs_lookup_block_group(root->fs_info, bytenr);
2634 if (!block_group || block_group->ro)
2635 readonly = 1;
2636 if (block_group)
2637 btrfs_put_block_group(block_group);
2638 return readonly;
2639 }
2640
2641 static int update_space_info(struct btrfs_fs_info *info, u64 flags,
2642 u64 total_bytes, u64 bytes_used,
2643 struct btrfs_space_info **space_info)
2644 {
2645 struct btrfs_space_info *found;
2646
2647 found = __find_space_info(info, flags);
2648 if (found) {
2649 spin_lock(&found->lock);
2650 found->total_bytes += total_bytes;
2651 found->bytes_used += bytes_used;
2652 found->full = 0;
2653 spin_unlock(&found->lock);
2654 *space_info = found;
2655 return 0;
2656 }
2657 found = kzalloc(sizeof(*found), GFP_NOFS);
2658 if (!found)
2659 return -ENOMEM;
2660
2661 INIT_LIST_HEAD(&found->block_groups);
2662 init_rwsem(&found->groups_sem);
2663 spin_lock_init(&found->lock);
2664 found->flags = flags;
2665 found->total_bytes = total_bytes;
2666 found->bytes_used = bytes_used;
2667 found->bytes_pinned = 0;
2668 found->bytes_reserved = 0;
2669 found->bytes_readonly = 0;
2670 found->bytes_delalloc = 0;
2671 found->full = 0;
2672 found->force_alloc = 0;
2673 *space_info = found;
2674 list_add_rcu(&found->list, &info->space_info);
2675 atomic_set(&found->caching_threads, 0);
2676 return 0;
2677 }
2678
2679 static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags)
2680 {
2681 u64 extra_flags = flags & (BTRFS_BLOCK_GROUP_RAID0 |
2682 BTRFS_BLOCK_GROUP_RAID1 |
2683 BTRFS_BLOCK_GROUP_RAID10 |
2684 BTRFS_BLOCK_GROUP_DUP);
2685 if (extra_flags) {
2686 if (flags & BTRFS_BLOCK_GROUP_DATA)
2687 fs_info->avail_data_alloc_bits |= extra_flags;
2688 if (flags & BTRFS_BLOCK_GROUP_METADATA)
2689 fs_info->avail_metadata_alloc_bits |= extra_flags;
2690 if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
2691 fs_info->avail_system_alloc_bits |= extra_flags;
2692 }
2693 }
2694
2695 static void set_block_group_readonly(struct btrfs_block_group_cache *cache)
2696 {
2697 spin_lock(&cache->space_info->lock);
2698 spin_lock(&cache->lock);
2699 if (!cache->ro) {
2700 cache->space_info->bytes_readonly += cache->key.offset -
2701 btrfs_block_group_used(&cache->item);
2702 cache->ro = 1;
2703 }
2704 spin_unlock(&cache->lock);
2705 spin_unlock(&cache->space_info->lock);
2706 }
2707
2708 u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags)
2709 {
2710 u64 num_devices = root->fs_info->fs_devices->rw_devices;
2711
2712 if (num_devices == 1)
2713 flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0);
2714 if (num_devices < 4)
2715 flags &= ~BTRFS_BLOCK_GROUP_RAID10;
2716
2717 if ((flags & BTRFS_BLOCK_GROUP_DUP) &&
2718 (flags & (BTRFS_BLOCK_GROUP_RAID1 |
2719 BTRFS_BLOCK_GROUP_RAID10))) {
2720 flags &= ~BTRFS_BLOCK_GROUP_DUP;
2721 }
2722
2723 if ((flags & BTRFS_BLOCK_GROUP_RAID1) &&
2724 (flags & BTRFS_BLOCK_GROUP_RAID10)) {
2725 flags &= ~BTRFS_BLOCK_GROUP_RAID1;
2726 }
2727
2728 if ((flags & BTRFS_BLOCK_GROUP_RAID0) &&
2729 ((flags & BTRFS_BLOCK_GROUP_RAID1) |
2730 (flags & BTRFS_BLOCK_GROUP_RAID10) |
2731 (flags & BTRFS_BLOCK_GROUP_DUP)))
2732 flags &= ~BTRFS_BLOCK_GROUP_RAID0;
2733 return flags;
2734 }
2735
2736 static u64 btrfs_get_alloc_profile(struct btrfs_root *root, u64 data)
2737 {
2738 struct btrfs_fs_info *info = root->fs_info;
2739 u64 alloc_profile;
2740
2741 if (data) {
2742 alloc_profile = info->avail_data_alloc_bits &
2743 info->data_alloc_profile;
2744 data = BTRFS_BLOCK_GROUP_DATA | alloc_profile;
2745 } else if (root == root->fs_info->chunk_root) {
2746 alloc_profile = info->avail_system_alloc_bits &
2747 info->system_alloc_profile;
2748 data = BTRFS_BLOCK_GROUP_SYSTEM | alloc_profile;
2749 } else {
2750 alloc_profile = info->avail_metadata_alloc_bits &
2751 info->metadata_alloc_profile;
2752 data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile;
2753 }
2754
2755 return btrfs_reduce_alloc_profile(root, data);
2756 }
2757
2758 void btrfs_set_inode_space_info(struct btrfs_root *root, struct inode *inode)
2759 {
2760 u64 alloc_target;
2761
2762 alloc_target = btrfs_get_alloc_profile(root, 1);
2763 BTRFS_I(inode)->space_info = __find_space_info(root->fs_info,
2764 alloc_target);
2765 }
2766
2767 static u64 calculate_bytes_needed(struct btrfs_root *root, int num_items)
2768 {
2769 u64 num_bytes;
2770 int level;
2771
2772 level = BTRFS_MAX_LEVEL - 2;
2773 /*
2774 * NOTE: these calculations are absolutely the worst possible case.
2775 * This assumes that _every_ item we insert will require a new leaf, and
2776 * that the tree has grown to its maximum level size.
2777 */
2778
2779 /*
2780 * for every item we insert we could insert both an extent item and a
2781 * extent ref item. Then for ever item we insert, we will need to cow
2782 * both the original leaf, plus the leaf to the left and right of it.
2783 *
2784 * Unless we are talking about the extent root, then we just want the
2785 * number of items * 2, since we just need the extent item plus its ref.
2786 */
2787 if (root == root->fs_info->extent_root)
2788 num_bytes = num_items * 2;
2789 else
2790 num_bytes = (num_items + (2 * num_items)) * 3;
2791
2792 /*
2793 * num_bytes is total number of leaves we could need times the leaf
2794 * size, and then for every leaf we could end up cow'ing 2 nodes per
2795 * level, down to the leaf level.
2796 */
2797 num_bytes = (num_bytes * root->leafsize) +
2798 (num_bytes * (level * 2)) * root->nodesize;
2799
2800 return num_bytes;
2801 }
2802
2803 /*
2804 * Unreserve metadata space for delalloc. If we have less reserved credits than
2805 * we have extents, this function does nothing.
2806 */
2807 int btrfs_unreserve_metadata_for_delalloc(struct btrfs_root *root,
2808 struct inode *inode, int num_items)
2809 {
2810 struct btrfs_fs_info *info = root->fs_info;
2811 struct btrfs_space_info *meta_sinfo;
2812 u64 num_bytes;
2813 u64 alloc_target;
2814 bool bug = false;
2815
2816 /* get the space info for where the metadata will live */
2817 alloc_target = btrfs_get_alloc_profile(root, 0);
2818 meta_sinfo = __find_space_info(info, alloc_target);
2819
2820 num_bytes = calculate_bytes_needed(root->fs_info->extent_root,
2821 num_items);
2822
2823 spin_lock(&meta_sinfo->lock);
2824 spin_lock(&BTRFS_I(inode)->accounting_lock);
2825 if (BTRFS_I(inode)->reserved_extents <=
2826 BTRFS_I(inode)->outstanding_extents) {
2827 spin_unlock(&BTRFS_I(inode)->accounting_lock);
2828 spin_unlock(&meta_sinfo->lock);
2829 return 0;
2830 }
2831 spin_unlock(&BTRFS_I(inode)->accounting_lock);
2832
2833 BTRFS_I(inode)->reserved_extents--;
2834 BUG_ON(BTRFS_I(inode)->reserved_extents < 0);
2835
2836 if (meta_sinfo->bytes_delalloc < num_bytes) {
2837 bug = true;
2838 meta_sinfo->bytes_delalloc = 0;
2839 } else {
2840 meta_sinfo->bytes_delalloc -= num_bytes;
2841 }
2842 spin_unlock(&meta_sinfo->lock);
2843
2844 BUG_ON(bug);
2845
2846 return 0;
2847 }
2848
2849 static void check_force_delalloc(struct btrfs_space_info *meta_sinfo)
2850 {
2851 u64 thresh;
2852
2853 thresh = meta_sinfo->bytes_used + meta_sinfo->bytes_reserved +
2854 meta_sinfo->bytes_pinned + meta_sinfo->bytes_readonly +
2855 meta_sinfo->bytes_super + meta_sinfo->bytes_root +
2856 meta_sinfo->bytes_may_use;
2857
2858 thresh = meta_sinfo->total_bytes - thresh;
2859 thresh *= 80;
2860 do_div(thresh, 100);
2861 if (thresh <= meta_sinfo->bytes_delalloc)
2862 meta_sinfo->force_delalloc = 1;
2863 else
2864 meta_sinfo->force_delalloc = 0;
2865 }
2866
2867 struct async_flush {
2868 struct btrfs_root *root;
2869 struct btrfs_space_info *info;
2870 struct btrfs_work work;
2871 };
2872
2873 static noinline void flush_delalloc_async(struct btrfs_work *work)
2874 {
2875 struct async_flush *async;
2876 struct btrfs_root *root;
2877 struct btrfs_space_info *info;
2878
2879 async = container_of(work, struct async_flush, work);
2880 root = async->root;
2881 info = async->info;
2882
2883 btrfs_start_delalloc_inodes(root);
2884 wake_up(&info->flush_wait);
2885 btrfs_wait_ordered_extents(root, 0);
2886
2887 spin_lock(&info->lock);
2888 info->flushing = 0;
2889 spin_unlock(&info->lock);
2890 wake_up(&info->flush_wait);
2891
2892 kfree(async);
2893 }
2894
2895 static void wait_on_flush(struct btrfs_space_info *info)
2896 {
2897 DEFINE_WAIT(wait);
2898 u64 used;
2899
2900 while (1) {
2901 prepare_to_wait(&info->flush_wait, &wait,
2902 TASK_UNINTERRUPTIBLE);
2903 spin_lock(&info->lock);
2904 if (!info->flushing) {
2905 spin_unlock(&info->lock);
2906 break;
2907 }
2908
2909 used = info->bytes_used + info->bytes_reserved +
2910 info->bytes_pinned + info->bytes_readonly +
2911 info->bytes_super + info->bytes_root +
2912 info->bytes_may_use + info->bytes_delalloc;
2913 if (used < info->total_bytes) {
2914 spin_unlock(&info->lock);
2915 break;
2916 }
2917 spin_unlock(&info->lock);
2918 schedule();
2919 }
2920 finish_wait(&info->flush_wait, &wait);
2921 }
2922
2923 static void flush_delalloc(struct btrfs_root *root,
2924 struct btrfs_space_info *info)
2925 {
2926 struct async_flush *async;
2927 bool wait = false;
2928
2929 spin_lock(&info->lock);
2930
2931 if (!info->flushing) {
2932 info->flushing = 1;
2933 init_waitqueue_head(&info->flush_wait);
2934 } else {
2935 wait = true;
2936 }
2937
2938 spin_unlock(&info->lock);
2939
2940 if (wait) {
2941 wait_on_flush(info);
2942 return;
2943 }
2944
2945 async = kzalloc(sizeof(*async), GFP_NOFS);
2946 if (!async)
2947 goto flush;
2948
2949 async->root = root;
2950 async->info = info;
2951 async->work.func = flush_delalloc_async;
2952
2953 btrfs_queue_worker(&root->fs_info->enospc_workers,
2954 &async->work);
2955 wait_on_flush(info);
2956 return;
2957
2958 flush:
2959 btrfs_start_delalloc_inodes(root);
2960 btrfs_wait_ordered_extents(root, 0);
2961
2962 spin_lock(&info->lock);
2963 info->flushing = 0;
2964 spin_unlock(&info->lock);
2965 wake_up(&info->flush_wait);
2966 }
2967
2968 static int maybe_allocate_chunk(struct btrfs_root *root,
2969 struct btrfs_space_info *info)
2970 {
2971 struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
2972 struct btrfs_trans_handle *trans;
2973 bool wait = false;
2974 int ret = 0;
2975 u64 min_metadata;
2976 u64 free_space;
2977
2978 free_space = btrfs_super_total_bytes(disk_super);
2979 /*
2980 * we allow the metadata to grow to a max of either 10gb or 5% of the
2981 * space in the volume.
2982 */
2983 min_metadata = min((u64)10 * 1024 * 1024 * 1024,
2984 div64_u64(free_space * 5, 100));
2985 if (info->total_bytes >= min_metadata) {
2986 spin_unlock(&info->lock);
2987 return 0;
2988 }
2989
2990 if (info->full) {
2991 spin_unlock(&info->lock);
2992 return 0;
2993 }
2994
2995 if (!info->allocating_chunk) {
2996 info->force_alloc = 1;
2997 info->allocating_chunk = 1;
2998 init_waitqueue_head(&info->allocate_wait);
2999 } else {
3000 wait = true;
3001 }
3002
3003 spin_unlock(&info->lock);
3004
3005 if (wait) {
3006 wait_event(info->allocate_wait,
3007 !info->allocating_chunk);
3008 return 1;
3009 }
3010
3011 trans = btrfs_start_transaction(root, 1);
3012 if (!trans) {
3013 ret = -ENOMEM;
3014 goto out;
3015 }
3016
3017 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
3018 4096 + 2 * 1024 * 1024,
3019 info->flags, 0);
3020 btrfs_end_transaction(trans, root);
3021 if (ret)
3022 goto out;
3023 out:
3024 spin_lock(&info->lock);
3025 info->allocating_chunk = 0;
3026 spin_unlock(&info->lock);
3027 wake_up(&info->allocate_wait);
3028
3029 if (ret)
3030 return 0;
3031 return 1;
3032 }
3033
3034 /*
3035 * Reserve metadata space for delalloc.
3036 */
3037 int btrfs_reserve_metadata_for_delalloc(struct btrfs_root *root,
3038 struct inode *inode, int num_items)
3039 {
3040 struct btrfs_fs_info *info = root->fs_info;
3041 struct btrfs_space_info *meta_sinfo;
3042 u64 num_bytes;
3043 u64 used;
3044 u64 alloc_target;
3045 int flushed = 0;
3046 int force_delalloc;
3047
3048 /* get the space info for where the metadata will live */
3049 alloc_target = btrfs_get_alloc_profile(root, 0);
3050 meta_sinfo = __find_space_info(info, alloc_target);
3051
3052 num_bytes = calculate_bytes_needed(root->fs_info->extent_root,
3053 num_items);
3054 again:
3055 spin_lock(&meta_sinfo->lock);
3056
3057 force_delalloc = meta_sinfo->force_delalloc;
3058
3059 if (unlikely(!meta_sinfo->bytes_root))
3060 meta_sinfo->bytes_root = calculate_bytes_needed(root, 6);
3061
3062 if (!flushed)
3063 meta_sinfo->bytes_delalloc += num_bytes;
3064
3065 used = meta_sinfo->bytes_used + meta_sinfo->bytes_reserved +
3066 meta_sinfo->bytes_pinned + meta_sinfo->bytes_readonly +
3067 meta_sinfo->bytes_super + meta_sinfo->bytes_root +
3068 meta_sinfo->bytes_may_use + meta_sinfo->bytes_delalloc;
3069
3070 if (used > meta_sinfo->total_bytes) {
3071 flushed++;
3072
3073 if (flushed == 1) {
3074 if (maybe_allocate_chunk(root, meta_sinfo))
3075 goto again;
3076 flushed++;
3077 } else {
3078 spin_unlock(&meta_sinfo->lock);
3079 }
3080
3081 if (flushed == 2) {
3082 filemap_flush(inode->i_mapping);
3083 goto again;
3084 } else if (flushed == 3) {
3085 flush_delalloc(root, meta_sinfo);
3086 goto again;
3087 }
3088 spin_lock(&meta_sinfo->lock);
3089 meta_sinfo->bytes_delalloc -= num_bytes;
3090 spin_unlock(&meta_sinfo->lock);
3091 printk(KERN_ERR "enospc, has %d, reserved %d\n",
3092 BTRFS_I(inode)->outstanding_extents,
3093 BTRFS_I(inode)->reserved_extents);
3094 dump_space_info(meta_sinfo, 0, 0);
3095 return -ENOSPC;
3096 }
3097
3098 BTRFS_I(inode)->reserved_extents++;
3099 check_force_delalloc(meta_sinfo);
3100 spin_unlock(&meta_sinfo->lock);
3101
3102 if (!flushed && force_delalloc)
3103 filemap_flush(inode->i_mapping);
3104
3105 return 0;
3106 }
3107
3108 /*
3109 * unreserve num_items number of items worth of metadata space. This needs to
3110 * be paired with btrfs_reserve_metadata_space.
3111 *
3112 * NOTE: if you have the option, run this _AFTER_ you do a
3113 * btrfs_end_transaction, since btrfs_end_transaction will run delayed ref
3114 * oprations which will result in more used metadata, so we want to make sure we
3115 * can do that without issue.
3116 */
3117 int btrfs_unreserve_metadata_space(struct btrfs_root *root, int num_items)
3118 {
3119 struct btrfs_fs_info *info = root->fs_info;
3120 struct btrfs_space_info *meta_sinfo;
3121 u64 num_bytes;
3122 u64 alloc_target;
3123 bool bug = false;
3124
3125 /* get the space info for where the metadata will live */
3126 alloc_target = btrfs_get_alloc_profile(root, 0);
3127 meta_sinfo = __find_space_info(info, alloc_target);
3128
3129 num_bytes = calculate_bytes_needed(root, num_items);
3130
3131 spin_lock(&meta_sinfo->lock);
3132 if (meta_sinfo->bytes_may_use < num_bytes) {
3133 bug = true;
3134 meta_sinfo->bytes_may_use = 0;
3135 } else {
3136 meta_sinfo->bytes_may_use -= num_bytes;
3137 }
3138 spin_unlock(&meta_sinfo->lock);
3139
3140 BUG_ON(bug);
3141
3142 return 0;
3143 }
3144
3145 /*
3146 * Reserve some metadata space for use. We'll calculate the worste case number
3147 * of bytes that would be needed to modify num_items number of items. If we
3148 * have space, fantastic, if not, you get -ENOSPC. Please call
3149 * btrfs_unreserve_metadata_space when you are done for the _SAME_ number of
3150 * items you reserved, since whatever metadata you needed should have already
3151 * been allocated.
3152 *
3153 * This will commit the transaction to make more space if we don't have enough
3154 * metadata space. THe only time we don't do this is if we're reserving space
3155 * inside of a transaction, then we will just return -ENOSPC and it is the
3156 * callers responsibility to handle it properly.
3157 */
3158 int btrfs_reserve_metadata_space(struct btrfs_root *root, int num_items)
3159 {
3160 struct btrfs_fs_info *info = root->fs_info;
3161 struct btrfs_space_info *meta_sinfo;
3162 u64 num_bytes;
3163 u64 used;
3164 u64 alloc_target;
3165 int retries = 0;
3166
3167 /* get the space info for where the metadata will live */
3168 alloc_target = btrfs_get_alloc_profile(root, 0);
3169 meta_sinfo = __find_space_info(info, alloc_target);
3170
3171 num_bytes = calculate_bytes_needed(root, num_items);
3172 again:
3173 spin_lock(&meta_sinfo->lock);
3174
3175 if (unlikely(!meta_sinfo->bytes_root))
3176 meta_sinfo->bytes_root = calculate_bytes_needed(root, 6);
3177
3178 if (!retries)
3179 meta_sinfo->bytes_may_use += num_bytes;
3180
3181 used = meta_sinfo->bytes_used + meta_sinfo->bytes_reserved +
3182 meta_sinfo->bytes_pinned + meta_sinfo->bytes_readonly +
3183 meta_sinfo->bytes_super + meta_sinfo->bytes_root +
3184 meta_sinfo->bytes_may_use + meta_sinfo->bytes_delalloc;
3185
3186 if (used > meta_sinfo->total_bytes) {
3187 retries++;
3188 if (retries == 1) {
3189 if (maybe_allocate_chunk(root, meta_sinfo))
3190 goto again;
3191 retries++;
3192 } else {
3193 spin_unlock(&meta_sinfo->lock);
3194 }
3195
3196 if (retries == 2) {
3197 flush_delalloc(root, meta_sinfo);
3198 goto again;
3199 }
3200 spin_lock(&meta_sinfo->lock);
3201 meta_sinfo->bytes_may_use -= num_bytes;
3202 spin_unlock(&meta_sinfo->lock);
3203
3204 dump_space_info(meta_sinfo, 0, 0);
3205 return -ENOSPC;
3206 }
3207
3208 check_force_delalloc(meta_sinfo);
3209 spin_unlock(&meta_sinfo->lock);
3210
3211 return 0;
3212 }
3213
3214 /*
3215 * This will check the space that the inode allocates from to make sure we have
3216 * enough space for bytes.
3217 */
3218 int btrfs_check_data_free_space(struct btrfs_root *root, struct inode *inode,
3219 u64 bytes)
3220 {
3221 struct btrfs_space_info *data_sinfo;
3222 int ret = 0, committed = 0;
3223
3224 /* make sure bytes are sectorsize aligned */
3225 bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
3226
3227 data_sinfo = BTRFS_I(inode)->space_info;
3228 if (!data_sinfo)
3229 goto alloc;
3230
3231 again:
3232 /* make sure we have enough space to handle the data first */
3233 spin_lock(&data_sinfo->lock);
3234 if (data_sinfo->total_bytes - data_sinfo->bytes_used -
3235 data_sinfo->bytes_delalloc - data_sinfo->bytes_reserved -
3236 data_sinfo->bytes_pinned - data_sinfo->bytes_readonly -
3237 data_sinfo->bytes_may_use - data_sinfo->bytes_super < bytes) {
3238 struct btrfs_trans_handle *trans;
3239
3240 /*
3241 * if we don't have enough free bytes in this space then we need
3242 * to alloc a new chunk.
3243 */
3244 if (!data_sinfo->full) {
3245 u64 alloc_target;
3246
3247 data_sinfo->force_alloc = 1;
3248 spin_unlock(&data_sinfo->lock);
3249 alloc:
3250 alloc_target = btrfs_get_alloc_profile(root, 1);
3251 trans = btrfs_start_transaction(root, 1);
3252 if (!trans)
3253 return -ENOMEM;
3254
3255 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
3256 bytes + 2 * 1024 * 1024,
3257 alloc_target, 0);
3258 btrfs_end_transaction(trans, root);
3259 if (ret)
3260 return ret;
3261
3262 if (!data_sinfo) {
3263 btrfs_set_inode_space_info(root, inode);
3264 data_sinfo = BTRFS_I(inode)->space_info;
3265 }
3266 goto again;
3267 }
3268 spin_unlock(&data_sinfo->lock);
3269
3270 /* commit the current transaction and try again */
3271 if (!committed && !root->fs_info->open_ioctl_trans) {
3272 committed = 1;
3273 trans = btrfs_join_transaction(root, 1);
3274 if (!trans)
3275 return -ENOMEM;
3276 ret = btrfs_commit_transaction(trans, root);
3277 if (ret)
3278 return ret;
3279 goto again;
3280 }
3281
3282 printk(KERN_ERR "no space left, need %llu, %llu delalloc bytes"
3283 ", %llu bytes_used, %llu bytes_reserved, "
3284 "%llu bytes_pinned, %llu bytes_readonly, %llu may use "
3285 "%llu total\n", (unsigned long long)bytes,
3286 (unsigned long long)data_sinfo->bytes_delalloc,
3287 (unsigned long long)data_sinfo->bytes_used,
3288 (unsigned long long)data_sinfo->bytes_reserved,
3289 (unsigned long long)data_sinfo->bytes_pinned,
3290 (unsigned long long)data_sinfo->bytes_readonly,
3291 (unsigned long long)data_sinfo->bytes_may_use,
3292 (unsigned long long)data_sinfo->total_bytes);
3293 return -ENOSPC;
3294 }
3295 data_sinfo->bytes_may_use += bytes;
3296 BTRFS_I(inode)->reserved_bytes += bytes;
3297 spin_unlock(&data_sinfo->lock);
3298
3299 return 0;
3300 }
3301
3302 /*
3303 * if there was an error for whatever reason after calling
3304 * btrfs_check_data_free_space, call this so we can cleanup the counters.
3305 */
3306 void btrfs_free_reserved_data_space(struct btrfs_root *root,
3307 struct inode *inode, u64 bytes)
3308 {
3309 struct btrfs_space_info *data_sinfo;
3310
3311 /* make sure bytes are sectorsize aligned */
3312 bytes = (bytes + root->sectorsize - 1) & ~((u64)root->sectorsize - 1);
3313
3314 data_sinfo = BTRFS_I(inode)->space_info;
3315 spin_lock(&data_sinfo->lock);
3316 data_sinfo->bytes_may_use -= bytes;
3317 BTRFS_I(inode)->reserved_bytes -= bytes;
3318 spin_unlock(&data_sinfo->lock);
3319 }
3320
3321 /* called when we are adding a delalloc extent to the inode's io_tree */
3322 void btrfs_delalloc_reserve_space(struct btrfs_root *root, struct inode *inode,
3323 u64 bytes)
3324 {
3325 struct btrfs_space_info *data_sinfo;
3326
3327 /* get the space info for where this inode will be storing its data */
3328 data_sinfo = BTRFS_I(inode)->space_info;
3329
3330 /* make sure we have enough space to handle the data first */
3331 spin_lock(&data_sinfo->lock);
3332 data_sinfo->bytes_delalloc += bytes;
3333
3334 /*
3335 * we are adding a delalloc extent without calling
3336 * btrfs_check_data_free_space first. This happens on a weird
3337 * writepage condition, but shouldn't hurt our accounting
3338 */
3339 if (unlikely(bytes > BTRFS_I(inode)->reserved_bytes)) {
3340 data_sinfo->bytes_may_use -= BTRFS_I(inode)->reserved_bytes;
3341 BTRFS_I(inode)->reserved_bytes = 0;
3342 } else {
3343 data_sinfo->bytes_may_use -= bytes;
3344 BTRFS_I(inode)->reserved_bytes -= bytes;
3345 }
3346
3347 spin_unlock(&data_sinfo->lock);
3348 }
3349
3350 /* called when we are clearing an delalloc extent from the inode's io_tree */
3351 void btrfs_delalloc_free_space(struct btrfs_root *root, struct inode *inode,
3352 u64 bytes)
3353 {
3354 struct btrfs_space_info *info;
3355
3356 info = BTRFS_I(inode)->space_info;
3357
3358 spin_lock(&info->lock);
3359 info->bytes_delalloc -= bytes;
3360 spin_unlock(&info->lock);
3361 }
3362
3363 static void force_metadata_allocation(struct btrfs_fs_info *info)
3364 {
3365 struct list_head *head = &info->space_info;
3366 struct btrfs_space_info *found;
3367
3368 rcu_read_lock();
3369 list_for_each_entry_rcu(found, head, list) {
3370 if (found->flags & BTRFS_BLOCK_GROUP_METADATA)
3371 found->force_alloc = 1;
3372 }
3373 rcu_read_unlock();
3374 }
3375
3376 static int do_chunk_alloc(struct btrfs_trans_handle *trans,
3377 struct btrfs_root *extent_root, u64 alloc_bytes,
3378 u64 flags, int force)
3379 {
3380 struct btrfs_space_info *space_info;
3381 struct btrfs_fs_info *fs_info = extent_root->fs_info;
3382 u64 thresh;
3383 int ret = 0;
3384
3385 mutex_lock(&fs_info->chunk_mutex);
3386
3387 flags = btrfs_reduce_alloc_profile(extent_root, flags);
3388
3389 space_info = __find_space_info(extent_root->fs_info, flags);
3390 if (!space_info) {
3391 ret = update_space_info(extent_root->fs_info, flags,
3392 0, 0, &space_info);
3393 BUG_ON(ret);
3394 }
3395 BUG_ON(!space_info);
3396
3397 spin_lock(&space_info->lock);
3398 if (space_info->force_alloc)
3399 force = 1;
3400 if (space_info->full) {
3401 spin_unlock(&space_info->lock);
3402 goto out;
3403 }
3404
3405 thresh = space_info->total_bytes - space_info->bytes_readonly;
3406 thresh = div_factor(thresh, 8);
3407 if (!force &&
3408 (space_info->bytes_used + space_info->bytes_pinned +
3409 space_info->bytes_reserved + alloc_bytes) < thresh) {
3410 spin_unlock(&space_info->lock);
3411 goto out;
3412 }
3413 spin_unlock(&space_info->lock);
3414
3415 /*
3416 * if we're doing a data chunk, go ahead and make sure that
3417 * we keep a reasonable number of metadata chunks allocated in the
3418 * FS as well.
3419 */
3420 if (flags & BTRFS_BLOCK_GROUP_DATA && fs_info->metadata_ratio) {
3421 fs_info->data_chunk_allocations++;
3422 if (!(fs_info->data_chunk_allocations %
3423 fs_info->metadata_ratio))
3424 force_metadata_allocation(fs_info);
3425 }
3426
3427 ret = btrfs_alloc_chunk(trans, extent_root, flags);
3428 spin_lock(&space_info->lock);
3429 if (ret)
3430 space_info->full = 1;
3431 space_info->force_alloc = 0;
3432 spin_unlock(&space_info->lock);
3433 out:
3434 mutex_unlock(&extent_root->fs_info->chunk_mutex);
3435 return ret;
3436 }
3437
3438 static int update_block_group(struct btrfs_trans_handle *trans,
3439 struct btrfs_root *root,
3440 u64 bytenr, u64 num_bytes, int alloc,
3441 int mark_free)
3442 {
3443 struct btrfs_block_group_cache *cache;
3444 struct btrfs_fs_info *info = root->fs_info;
3445 u64 total = num_bytes;
3446 u64 old_val;
3447 u64 byte_in_group;
3448
3449 /* block accounting for super block */
3450 spin_lock(&info->delalloc_lock);
3451 old_val = btrfs_super_bytes_used(&info->super_copy);
3452 if (alloc)
3453 old_val += num_bytes;
3454 else
3455 old_val -= num_bytes;
3456 btrfs_set_super_bytes_used(&info->super_copy, old_val);
3457
3458 /* block accounting for root item */
3459 old_val = btrfs_root_used(&root->root_item);
3460 if (alloc)
3461 old_val += num_bytes;
3462 else
3463 old_val -= num_bytes;
3464 btrfs_set_root_used(&root->root_item, 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 static u64 stripe_align(struct btrfs_root *root, u64 val)
4053 {
4054 u64 mask = ((u64)root->stripesize - 1);
4055 u64 ret = (val + mask) & ~mask;
4056 return ret;
4057 }
4058
4059 /*
4060 * when we wait for progress in the block group caching, its because
4061 * our allocation attempt failed at least once. So, we must sleep
4062 * and let some progress happen before we try again.
4063 *
4064 * This function will sleep at least once waiting for new free space to
4065 * show up, and then it will check the block group free space numbers
4066 * for our min num_bytes. Another option is to have it go ahead
4067 * and look in the rbtree for a free extent of a given size, but this
4068 * is a good start.
4069 */
4070 static noinline int
4071 wait_block_group_cache_progress(struct btrfs_block_group_cache *cache,
4072 u64 num_bytes)
4073 {
4074 struct btrfs_caching_control *caching_ctl;
4075 DEFINE_WAIT(wait);
4076
4077 caching_ctl = get_caching_control(cache);
4078 if (!caching_ctl)
4079 return 0;
4080
4081 wait_event(caching_ctl->wait, block_group_cache_done(cache) ||
4082 (cache->free_space >= num_bytes));
4083
4084 put_caching_control(caching_ctl);
4085 return 0;
4086 }
4087
4088 static noinline int
4089 wait_block_group_cache_done(struct btrfs_block_group_cache *cache)
4090 {
4091 struct btrfs_caching_control *caching_ctl;
4092 DEFINE_WAIT(wait);
4093
4094 caching_ctl = get_caching_control(cache);
4095 if (!caching_ctl)
4096 return 0;
4097
4098 wait_event(caching_ctl->wait, block_group_cache_done(cache));
4099
4100 put_caching_control(caching_ctl);
4101 return 0;
4102 }
4103
4104 enum btrfs_loop_type {
4105 LOOP_FIND_IDEAL = 0,
4106 LOOP_CACHING_NOWAIT = 1,
4107 LOOP_CACHING_WAIT = 2,
4108 LOOP_ALLOC_CHUNK = 3,
4109 LOOP_NO_EMPTY_SIZE = 4,
4110 };
4111
4112 /*
4113 * walks the btree of allocated extents and find a hole of a given size.
4114 * The key ins is changed to record the hole:
4115 * ins->objectid == block start
4116 * ins->flags = BTRFS_EXTENT_ITEM_KEY
4117 * ins->offset == number of blocks
4118 * Any available blocks before search_start are skipped.
4119 */
4120 static noinline int find_free_extent(struct btrfs_trans_handle *trans,
4121 struct btrfs_root *orig_root,
4122 u64 num_bytes, u64 empty_size,
4123 u64 search_start, u64 search_end,
4124 u64 hint_byte, struct btrfs_key *ins,
4125 u64 exclude_start, u64 exclude_nr,
4126 int data)
4127 {
4128 int ret = 0;
4129 struct btrfs_root *root = orig_root->fs_info->extent_root;
4130 struct btrfs_free_cluster *last_ptr = NULL;
4131 struct btrfs_block_group_cache *block_group = NULL;
4132 int empty_cluster = 2 * 1024 * 1024;
4133 int allowed_chunk_alloc = 0;
4134 int done_chunk_alloc = 0;
4135 struct btrfs_space_info *space_info;
4136 int last_ptr_loop = 0;
4137 int loop = 0;
4138 bool found_uncached_bg = false;
4139 bool failed_cluster_refill = false;
4140 bool failed_alloc = false;
4141 u64 ideal_cache_percent = 0;
4142 u64 ideal_cache_offset = 0;
4143
4144 WARN_ON(num_bytes < root->sectorsize);
4145 btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
4146 ins->objectid = 0;
4147 ins->offset = 0;
4148
4149 space_info = __find_space_info(root->fs_info, data);
4150
4151 if (orig_root->ref_cows || empty_size)
4152 allowed_chunk_alloc = 1;
4153
4154 if (data & BTRFS_BLOCK_GROUP_METADATA) {
4155 last_ptr = &root->fs_info->meta_alloc_cluster;
4156 if (!btrfs_test_opt(root, SSD))
4157 empty_cluster = 64 * 1024;
4158 }
4159
4160 if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD)) {
4161 last_ptr = &root->fs_info->data_alloc_cluster;
4162 }
4163
4164 if (last_ptr) {
4165 spin_lock(&last_ptr->lock);
4166 if (last_ptr->block_group)
4167 hint_byte = last_ptr->window_start;
4168 spin_unlock(&last_ptr->lock);
4169 }
4170
4171 search_start = max(search_start, first_logical_byte(root, 0));
4172 search_start = max(search_start, hint_byte);
4173
4174 if (!last_ptr)
4175 empty_cluster = 0;
4176
4177 if (search_start == hint_byte) {
4178 ideal_cache:
4179 block_group = btrfs_lookup_block_group(root->fs_info,
4180 search_start);
4181 /*
4182 * we don't want to use the block group if it doesn't match our
4183 * allocation bits, or if its not cached.
4184 *
4185 * However if we are re-searching with an ideal block group
4186 * picked out then we don't care that the block group is cached.
4187 */
4188 if (block_group && block_group_bits(block_group, data) &&
4189 (block_group->cached != BTRFS_CACHE_NO ||
4190 search_start == ideal_cache_offset)) {
4191 down_read(&space_info->groups_sem);
4192 if (list_empty(&block_group->list) ||
4193 block_group->ro) {
4194 /*
4195 * someone is removing this block group,
4196 * we can't jump into the have_block_group
4197 * target because our list pointers are not
4198 * valid
4199 */
4200 btrfs_put_block_group(block_group);
4201 up_read(&space_info->groups_sem);
4202 } else {
4203 goto have_block_group;
4204 }
4205 } else if (block_group) {
4206 btrfs_put_block_group(block_group);
4207 }
4208 }
4209 search:
4210 down_read(&space_info->groups_sem);
4211 list_for_each_entry(block_group, &space_info->block_groups, list) {
4212 u64 offset;
4213 int cached;
4214
4215 atomic_inc(&block_group->count);
4216 search_start = block_group->key.objectid;
4217
4218 have_block_group:
4219 if (unlikely(block_group->cached == BTRFS_CACHE_NO)) {
4220 u64 free_percent;
4221
4222 free_percent = btrfs_block_group_used(&block_group->item);
4223 free_percent *= 100;
4224 free_percent = div64_u64(free_percent,
4225 block_group->key.offset);
4226 free_percent = 100 - free_percent;
4227 if (free_percent > ideal_cache_percent &&
4228 likely(!block_group->ro)) {
4229 ideal_cache_offset = block_group->key.objectid;
4230 ideal_cache_percent = free_percent;
4231 }
4232
4233 /*
4234 * We only want to start kthread caching if we are at
4235 * the point where we will wait for caching to make
4236 * progress, or if our ideal search is over and we've
4237 * found somebody to start caching.
4238 */
4239 if (loop > LOOP_CACHING_NOWAIT ||
4240 (loop > LOOP_FIND_IDEAL &&
4241 atomic_read(&space_info->caching_threads) < 2)) {
4242 ret = cache_block_group(block_group);
4243 BUG_ON(ret);
4244 }
4245 found_uncached_bg = true;
4246
4247 /*
4248 * If loop is set for cached only, try the next block
4249 * group.
4250 */
4251 if (loop == LOOP_FIND_IDEAL)
4252 goto loop;
4253 }
4254
4255 cached = block_group_cache_done(block_group);
4256 if (unlikely(!cached))
4257 found_uncached_bg = true;
4258
4259 if (unlikely(block_group->ro))
4260 goto loop;
4261
4262 /*
4263 * Ok we want to try and use the cluster allocator, so lets look
4264 * there, unless we are on LOOP_NO_EMPTY_SIZE, since we will
4265 * have tried the cluster allocator plenty of times at this
4266 * point and not have found anything, so we are likely way too
4267 * fragmented for the clustering stuff to find anything, so lets
4268 * just skip it and let the allocator find whatever block it can
4269 * find
4270 */
4271 if (last_ptr && loop < LOOP_NO_EMPTY_SIZE) {
4272 /*
4273 * the refill lock keeps out other
4274 * people trying to start a new cluster
4275 */
4276 spin_lock(&last_ptr->refill_lock);
4277 if (last_ptr->block_group &&
4278 (last_ptr->block_group->ro ||
4279 !block_group_bits(last_ptr->block_group, data))) {
4280 offset = 0;
4281 goto refill_cluster;
4282 }
4283
4284 offset = btrfs_alloc_from_cluster(block_group, last_ptr,
4285 num_bytes, search_start);
4286 if (offset) {
4287 /* we have a block, we're done */
4288 spin_unlock(&last_ptr->refill_lock);
4289 goto checks;
4290 }
4291
4292 spin_lock(&last_ptr->lock);
4293 /*
4294 * whoops, this cluster doesn't actually point to
4295 * this block group. Get a ref on the block
4296 * group is does point to and try again
4297 */
4298 if (!last_ptr_loop && last_ptr->block_group &&
4299 last_ptr->block_group != block_group) {
4300
4301 btrfs_put_block_group(block_group);
4302 block_group = last_ptr->block_group;
4303 atomic_inc(&block_group->count);
4304 spin_unlock(&last_ptr->lock);
4305 spin_unlock(&last_ptr->refill_lock);
4306
4307 last_ptr_loop = 1;
4308 search_start = block_group->key.objectid;
4309 /*
4310 * we know this block group is properly
4311 * in the list because
4312 * btrfs_remove_block_group, drops the
4313 * cluster before it removes the block
4314 * group from the list
4315 */
4316 goto have_block_group;
4317 }
4318 spin_unlock(&last_ptr->lock);
4319 refill_cluster:
4320 /*
4321 * this cluster didn't work out, free it and
4322 * start over
4323 */
4324 btrfs_return_cluster_to_free_space(NULL, last_ptr);
4325
4326 last_ptr_loop = 0;
4327
4328 /* allocate a cluster in this block group */
4329 ret = btrfs_find_space_cluster(trans, root,
4330 block_group, last_ptr,
4331 offset, num_bytes,
4332 empty_cluster + empty_size);
4333 if (ret == 0) {
4334 /*
4335 * now pull our allocation out of this
4336 * cluster
4337 */
4338 offset = btrfs_alloc_from_cluster(block_group,
4339 last_ptr, num_bytes,
4340 search_start);
4341 if (offset) {
4342 /* we found one, proceed */
4343 spin_unlock(&last_ptr->refill_lock);
4344 goto checks;
4345 }
4346 } else if (!cached && loop > LOOP_CACHING_NOWAIT
4347 && !failed_cluster_refill) {
4348 spin_unlock(&last_ptr->refill_lock);
4349
4350 failed_cluster_refill = true;
4351 wait_block_group_cache_progress(block_group,
4352 num_bytes + empty_cluster + empty_size);
4353 goto have_block_group;
4354 }
4355
4356 /*
4357 * at this point we either didn't find a cluster
4358 * or we weren't able to allocate a block from our
4359 * cluster. Free the cluster we've been trying
4360 * to use, and go to the next block group
4361 */
4362 btrfs_return_cluster_to_free_space(NULL, last_ptr);
4363 spin_unlock(&last_ptr->refill_lock);
4364 goto loop;
4365 }
4366
4367 offset = btrfs_find_space_for_alloc(block_group, search_start,
4368 num_bytes, empty_size);
4369 /*
4370 * If we didn't find a chunk, and we haven't failed on this
4371 * block group before, and this block group is in the middle of
4372 * caching and we are ok with waiting, then go ahead and wait
4373 * for progress to be made, and set failed_alloc to true.
4374 *
4375 * If failed_alloc is true then we've already waited on this
4376 * block group once and should move on to the next block group.
4377 */
4378 if (!offset && !failed_alloc && !cached &&
4379 loop > LOOP_CACHING_NOWAIT) {
4380 wait_block_group_cache_progress(block_group,
4381 num_bytes + empty_size);
4382 failed_alloc = true;
4383 goto have_block_group;
4384 } else if (!offset) {
4385 goto loop;
4386 }
4387 checks:
4388 search_start = stripe_align(root, offset);
4389 /* move on to the next group */
4390 if (search_start + num_bytes >= search_end) {
4391 btrfs_add_free_space(block_group, offset, num_bytes);
4392 goto loop;
4393 }
4394
4395 /* move on to the next group */
4396 if (search_start + num_bytes >
4397 block_group->key.objectid + block_group->key.offset) {
4398 btrfs_add_free_space(block_group, offset, num_bytes);
4399 goto loop;
4400 }
4401
4402 if (exclude_nr > 0 &&
4403 (search_start + num_bytes > exclude_start &&
4404 search_start < exclude_start + exclude_nr)) {
4405 search_start = exclude_start + exclude_nr;
4406
4407 btrfs_add_free_space(block_group, offset, num_bytes);
4408 /*
4409 * if search_start is still in this block group
4410 * then we just re-search this block group
4411 */
4412 if (search_start >= block_group->key.objectid &&
4413 search_start < (block_group->key.objectid +
4414 block_group->key.offset))
4415 goto have_block_group;
4416 goto loop;
4417 }
4418
4419 ins->objectid = search_start;
4420 ins->offset = num_bytes;
4421
4422 if (offset < search_start)
4423 btrfs_add_free_space(block_group, offset,
4424 search_start - offset);
4425 BUG_ON(offset > search_start);
4426
4427 update_reserved_extents(block_group, num_bytes, 1);
4428
4429 /* we are all good, lets return */
4430 break;
4431 loop:
4432 failed_cluster_refill = false;
4433 failed_alloc = false;
4434 btrfs_put_block_group(block_group);
4435 }
4436 up_read(&space_info->groups_sem);
4437
4438 /* LOOP_FIND_IDEAL, only search caching/cached bg's, and don't wait for
4439 * for them to make caching progress. Also
4440 * determine the best possible bg to cache
4441 * LOOP_CACHING_NOWAIT, search partially cached block groups, kicking
4442 * caching kthreads as we move along
4443 * LOOP_CACHING_WAIT, search everything, and wait if our bg is caching
4444 * LOOP_ALLOC_CHUNK, force a chunk allocation and try again
4445 * LOOP_NO_EMPTY_SIZE, set empty_size and empty_cluster to 0 and try
4446 * again
4447 */
4448 if (!ins->objectid && loop < LOOP_NO_EMPTY_SIZE &&
4449 (found_uncached_bg || empty_size || empty_cluster ||
4450 allowed_chunk_alloc)) {
4451 if (loop == LOOP_FIND_IDEAL && found_uncached_bg) {
4452 found_uncached_bg = false;
4453 loop++;
4454 if (!ideal_cache_percent &&
4455 atomic_read(&space_info->caching_threads))
4456 goto search;
4457
4458 /*
4459 * 1 of the following 2 things have happened so far
4460 *
4461 * 1) We found an ideal block group for caching that
4462 * is mostly full and will cache quickly, so we might
4463 * as well wait for it.
4464 *
4465 * 2) We searched for cached only and we didn't find
4466 * anything, and we didn't start any caching kthreads
4467 * either, so chances are we will loop through and
4468 * start a couple caching kthreads, and then come back
4469 * around and just wait for them. This will be slower
4470 * because we will have 2 caching kthreads reading at
4471 * the same time when we could have just started one
4472 * and waited for it to get far enough to give us an
4473 * allocation, so go ahead and go to the wait caching
4474 * loop.
4475 */
4476 loop = LOOP_CACHING_WAIT;
4477 search_start = ideal_cache_offset;
4478 ideal_cache_percent = 0;
4479 goto ideal_cache;
4480 } else if (loop == LOOP_FIND_IDEAL) {
4481 /*
4482 * Didn't find a uncached bg, wait on anything we find
4483 * next.
4484 */
4485 loop = LOOP_CACHING_WAIT;
4486 goto search;
4487 }
4488
4489 if (loop < LOOP_CACHING_WAIT) {
4490 loop++;
4491 goto search;
4492 }
4493
4494 if (loop == LOOP_ALLOC_CHUNK) {
4495 empty_size = 0;
4496 empty_cluster = 0;
4497 }
4498
4499 if (allowed_chunk_alloc) {
4500 ret = do_chunk_alloc(trans, root, num_bytes +
4501 2 * 1024 * 1024, data, 1);
4502 allowed_chunk_alloc = 0;
4503 done_chunk_alloc = 1;
4504 } else if (!done_chunk_alloc) {
4505 space_info->force_alloc = 1;
4506 }
4507
4508 if (loop < LOOP_NO_EMPTY_SIZE) {
4509 loop++;
4510 goto search;
4511 }
4512 ret = -ENOSPC;
4513 } else if (!ins->objectid) {
4514 ret = -ENOSPC;
4515 }
4516
4517 /* we found what we needed */
4518 if (ins->objectid) {
4519 if (!(data & BTRFS_BLOCK_GROUP_DATA))
4520 trans->block_group = block_group->key.objectid;
4521
4522 btrfs_put_block_group(block_group);
4523 ret = 0;
4524 }
4525
4526 return ret;
4527 }
4528
4529 static void dump_space_info(struct btrfs_space_info *info, u64 bytes,
4530 int dump_block_groups)
4531 {
4532 struct btrfs_block_group_cache *cache;
4533
4534 spin_lock(&info->lock);
4535 printk(KERN_INFO "space_info has %llu free, is %sfull\n",
4536 (unsigned long long)(info->total_bytes - info->bytes_used -
4537 info->bytes_pinned - info->bytes_reserved -
4538 info->bytes_super),
4539 (info->full) ? "" : "not ");
4540 printk(KERN_INFO "space_info total=%llu, pinned=%llu, delalloc=%llu,"
4541 " may_use=%llu, used=%llu, root=%llu, super=%llu, reserved=%llu"
4542 "\n",
4543 (unsigned long long)info->total_bytes,
4544 (unsigned long long)info->bytes_pinned,
4545 (unsigned long long)info->bytes_delalloc,
4546 (unsigned long long)info->bytes_may_use,
4547 (unsigned long long)info->bytes_used,
4548 (unsigned long long)info->bytes_root,
4549 (unsigned long long)info->bytes_super,
4550 (unsigned long long)info->bytes_reserved);
4551 spin_unlock(&info->lock);
4552
4553 if (!dump_block_groups)
4554 return;
4555
4556 down_read(&info->groups_sem);
4557 list_for_each_entry(cache, &info->block_groups, list) {
4558 spin_lock(&cache->lock);
4559 printk(KERN_INFO "block group %llu has %llu bytes, %llu used "
4560 "%llu pinned %llu reserved\n",
4561 (unsigned long long)cache->key.objectid,
4562 (unsigned long long)cache->key.offset,
4563 (unsigned long long)btrfs_block_group_used(&cache->item),
4564 (unsigned long long)cache->pinned,
4565 (unsigned long long)cache->reserved);
4566 btrfs_dump_free_space(cache, bytes);
4567 spin_unlock(&cache->lock);
4568 }
4569 up_read(&info->groups_sem);
4570 }
4571
4572 int btrfs_reserve_extent(struct btrfs_trans_handle *trans,
4573 struct btrfs_root *root,
4574 u64 num_bytes, u64 min_alloc_size,
4575 u64 empty_size, u64 hint_byte,
4576 u64 search_end, struct btrfs_key *ins,
4577 u64 data)
4578 {
4579 int ret;
4580 u64 search_start = 0;
4581 struct btrfs_fs_info *info = root->fs_info;
4582
4583 data = btrfs_get_alloc_profile(root, data);
4584 again:
4585 /*
4586 * the only place that sets empty_size is btrfs_realloc_node, which
4587 * is not called recursively on allocations
4588 */
4589 if (empty_size || root->ref_cows) {
4590 if (!(data & BTRFS_BLOCK_GROUP_METADATA)) {
4591 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
4592 2 * 1024 * 1024,
4593 BTRFS_BLOCK_GROUP_METADATA |
4594 (info->metadata_alloc_profile &
4595 info->avail_metadata_alloc_bits), 0);
4596 }
4597 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
4598 num_bytes + 2 * 1024 * 1024, data, 0);
4599 }
4600
4601 WARN_ON(num_bytes < root->sectorsize);
4602 ret = find_free_extent(trans, root, num_bytes, empty_size,
4603 search_start, search_end, hint_byte, ins,
4604 trans->alloc_exclude_start,
4605 trans->alloc_exclude_nr, data);
4606
4607 if (ret == -ENOSPC && num_bytes > min_alloc_size) {
4608 num_bytes = num_bytes >> 1;
4609 num_bytes = num_bytes & ~(root->sectorsize - 1);
4610 num_bytes = max(num_bytes, min_alloc_size);
4611 do_chunk_alloc(trans, root->fs_info->extent_root,
4612 num_bytes, data, 1);
4613 goto again;
4614 }
4615 if (ret == -ENOSPC) {
4616 struct btrfs_space_info *sinfo;
4617
4618 sinfo = __find_space_info(root->fs_info, data);
4619 printk(KERN_ERR "btrfs allocation failed flags %llu, "
4620 "wanted %llu\n", (unsigned long long)data,
4621 (unsigned long long)num_bytes);
4622 dump_space_info(sinfo, num_bytes, 1);
4623 }
4624
4625 return ret;
4626 }
4627
4628 int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len)
4629 {
4630 struct btrfs_block_group_cache *cache;
4631 int ret = 0;
4632
4633 cache = btrfs_lookup_block_group(root->fs_info, start);
4634 if (!cache) {
4635 printk(KERN_ERR "Unable to find block group for %llu\n",
4636 (unsigned long long)start);
4637 return -ENOSPC;
4638 }
4639
4640 ret = btrfs_discard_extent(root, start, len);
4641
4642 btrfs_add_free_space(cache, start, len);
4643 update_reserved_extents(cache, len, 0);
4644 btrfs_put_block_group(cache);
4645
4646 return ret;
4647 }
4648
4649 static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
4650 struct btrfs_root *root,
4651 u64 parent, u64 root_objectid,
4652 u64 flags, u64 owner, u64 offset,
4653 struct btrfs_key *ins, int ref_mod)
4654 {
4655 int ret;
4656 struct btrfs_fs_info *fs_info = root->fs_info;
4657 struct btrfs_extent_item *extent_item;
4658 struct btrfs_extent_inline_ref *iref;
4659 struct btrfs_path *path;
4660 struct extent_buffer *leaf;
4661 int type;
4662 u32 size;
4663
4664 if (parent > 0)
4665 type = BTRFS_SHARED_DATA_REF_KEY;
4666 else
4667 type = BTRFS_EXTENT_DATA_REF_KEY;
4668
4669 size = sizeof(*extent_item) + btrfs_extent_inline_ref_size(type);
4670
4671 path = btrfs_alloc_path();
4672 BUG_ON(!path);
4673
4674 path->leave_spinning = 1;
4675 ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
4676 ins, size);
4677 BUG_ON(ret);
4678
4679 leaf = path->nodes[0];
4680 extent_item = btrfs_item_ptr(leaf, path->slots[0],
4681 struct btrfs_extent_item);
4682 btrfs_set_extent_refs(leaf, extent_item, ref_mod);
4683 btrfs_set_extent_generation(leaf, extent_item, trans->transid);
4684 btrfs_set_extent_flags(leaf, extent_item,
4685 flags | BTRFS_EXTENT_FLAG_DATA);
4686
4687 iref = (struct btrfs_extent_inline_ref *)(extent_item + 1);
4688 btrfs_set_extent_inline_ref_type(leaf, iref, type);
4689 if (parent > 0) {
4690 struct btrfs_shared_data_ref *ref;
4691 ref = (struct btrfs_shared_data_ref *)(iref + 1);
4692 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
4693 btrfs_set_shared_data_ref_count(leaf, ref, ref_mod);
4694 } else {
4695 struct btrfs_extent_data_ref *ref;
4696 ref = (struct btrfs_extent_data_ref *)(&iref->offset);
4697 btrfs_set_extent_data_ref_root(leaf, ref, root_objectid);
4698 btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
4699 btrfs_set_extent_data_ref_offset(leaf, ref, offset);
4700 btrfs_set_extent_data_ref_count(leaf, ref, ref_mod);
4701 }
4702
4703 btrfs_mark_buffer_dirty(path->nodes[0]);
4704 btrfs_free_path(path);
4705
4706 ret = update_block_group(trans, root, ins->objectid, ins->offset,
4707 1, 0);
4708 if (ret) {
4709 printk(KERN_ERR "btrfs update block group failed for %llu "
4710 "%llu\n", (unsigned long long)ins->objectid,
4711 (unsigned long long)ins->offset);
4712 BUG();
4713 }
4714 return ret;
4715 }
4716
4717 static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
4718 struct btrfs_root *root,
4719 u64 parent, u64 root_objectid,
4720 u64 flags, struct btrfs_disk_key *key,
4721 int level, struct btrfs_key *ins)
4722 {
4723 int ret;
4724 struct btrfs_fs_info *fs_info = root->fs_info;
4725 struct btrfs_extent_item *extent_item;
4726 struct btrfs_tree_block_info *block_info;
4727 struct btrfs_extent_inline_ref *iref;
4728 struct btrfs_path *path;
4729 struct extent_buffer *leaf;
4730 u32 size = sizeof(*extent_item) + sizeof(*block_info) + sizeof(*iref);
4731
4732 path = btrfs_alloc_path();
4733 BUG_ON(!path);
4734
4735 path->leave_spinning = 1;
4736 ret = btrfs_insert_empty_item(trans, fs_info->extent_root, path,
4737 ins, size);
4738 BUG_ON(ret);
4739
4740 leaf = path->nodes[0];
4741 extent_item = btrfs_item_ptr(leaf, path->slots[0],
4742 struct btrfs_extent_item);
4743 btrfs_set_extent_refs(leaf, extent_item, 1);
4744 btrfs_set_extent_generation(leaf, extent_item, trans->transid);
4745 btrfs_set_extent_flags(leaf, extent_item,
4746 flags | BTRFS_EXTENT_FLAG_TREE_BLOCK);
4747 block_info = (struct btrfs_tree_block_info *)(extent_item + 1);
4748
4749 btrfs_set_tree_block_key(leaf, block_info, key);
4750 btrfs_set_tree_block_level(leaf, block_info, level);
4751
4752 iref = (struct btrfs_extent_inline_ref *)(block_info + 1);
4753 if (parent > 0) {
4754 BUG_ON(!(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF));
4755 btrfs_set_extent_inline_ref_type(leaf, iref,
4756 BTRFS_SHARED_BLOCK_REF_KEY);
4757 btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
4758 } else {
4759 btrfs_set_extent_inline_ref_type(leaf, iref,
4760 BTRFS_TREE_BLOCK_REF_KEY);
4761 btrfs_set_extent_inline_ref_offset(leaf, iref, root_objectid);
4762 }
4763
4764 btrfs_mark_buffer_dirty(leaf);
4765 btrfs_free_path(path);
4766
4767 ret = update_block_group(trans, root, ins->objectid, ins->offset,
4768 1, 0);
4769 if (ret) {
4770 printk(KERN_ERR "btrfs update block group failed for %llu "
4771 "%llu\n", (unsigned long long)ins->objectid,
4772 (unsigned long long)ins->offset);
4773 BUG();
4774 }
4775 return ret;
4776 }
4777
4778 int btrfs_alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
4779 struct btrfs_root *root,
4780 u64 root_objectid, u64 owner,
4781 u64 offset, struct btrfs_key *ins)
4782 {
4783 int ret;
4784
4785 BUG_ON(root_objectid == BTRFS_TREE_LOG_OBJECTID);
4786
4787 ret = btrfs_add_delayed_data_ref(trans, ins->objectid, ins->offset,
4788 0, root_objectid, owner, offset,
4789 BTRFS_ADD_DELAYED_EXTENT, NULL);
4790 return ret;
4791 }
4792
4793 /*
4794 * this is used by the tree logging recovery code. It records that
4795 * an extent has been allocated and makes sure to clear the free
4796 * space cache bits as well
4797 */
4798 int btrfs_alloc_logged_file_extent(struct btrfs_trans_handle *trans,
4799 struct btrfs_root *root,
4800 u64 root_objectid, u64 owner, u64 offset,
4801 struct btrfs_key *ins)
4802 {
4803 int ret;
4804 struct btrfs_block_group_cache *block_group;
4805 struct btrfs_caching_control *caching_ctl;
4806 u64 start = ins->objectid;
4807 u64 num_bytes = ins->offset;
4808
4809 block_group = btrfs_lookup_block_group(root->fs_info, ins->objectid);
4810 cache_block_group(block_group);
4811 caching_ctl = get_caching_control(block_group);
4812
4813 if (!caching_ctl) {
4814 BUG_ON(!block_group_cache_done(block_group));
4815 ret = btrfs_remove_free_space(block_group, start, num_bytes);
4816 BUG_ON(ret);
4817 } else {
4818 mutex_lock(&caching_ctl->mutex);
4819
4820 if (start >= caching_ctl->progress) {
4821 ret = add_excluded_extent(root, start, num_bytes);
4822 BUG_ON(ret);
4823 } else if (start + num_bytes <= caching_ctl->progress) {
4824 ret = btrfs_remove_free_space(block_group,
4825 start, num_bytes);
4826 BUG_ON(ret);
4827 } else {
4828 num_bytes = caching_ctl->progress - start;
4829 ret = btrfs_remove_free_space(block_group,
4830 start, num_bytes);
4831 BUG_ON(ret);
4832
4833 start = caching_ctl->progress;
4834 num_bytes = ins->objectid + ins->offset -
4835 caching_ctl->progress;
4836 ret = add_excluded_extent(root, start, num_bytes);
4837 BUG_ON(ret);
4838 }
4839
4840 mutex_unlock(&caching_ctl->mutex);
4841 put_caching_control(caching_ctl);
4842 }
4843
4844 update_reserved_extents(block_group, ins->offset, 1);
4845 btrfs_put_block_group(block_group);
4846 ret = alloc_reserved_file_extent(trans, root, 0, root_objectid,
4847 0, owner, offset, ins, 1);
4848 return ret;
4849 }
4850
4851 /*
4852 * finds a free extent and does all the dirty work required for allocation
4853 * returns the key for the extent through ins, and a tree buffer for
4854 * the first block of the extent through buf.
4855 *
4856 * returns 0 if everything worked, non-zero otherwise.
4857 */
4858 static int alloc_tree_block(struct btrfs_trans_handle *trans,
4859 struct btrfs_root *root,
4860 u64 num_bytes, u64 parent, u64 root_objectid,
4861 struct btrfs_disk_key *key, int level,
4862 u64 empty_size, u64 hint_byte, u64 search_end,
4863 struct btrfs_key *ins)
4864 {
4865 int ret;
4866 u64 flags = 0;
4867
4868 ret = btrfs_reserve_extent(trans, root, num_bytes, num_bytes,
4869 empty_size, hint_byte, search_end,
4870 ins, 0);
4871 if (ret)
4872 return ret;
4873
4874 if (root_objectid == BTRFS_TREE_RELOC_OBJECTID) {
4875 if (parent == 0)
4876 parent = ins->objectid;
4877 flags |= BTRFS_BLOCK_FLAG_FULL_BACKREF;
4878 } else
4879 BUG_ON(parent > 0);
4880
4881 if (root_objectid != BTRFS_TREE_LOG_OBJECTID) {
4882 struct btrfs_delayed_extent_op *extent_op;
4883 extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
4884 BUG_ON(!extent_op);
4885 if (key)
4886 memcpy(&extent_op->key, key, sizeof(extent_op->key));
4887 else
4888 memset(&extent_op->key, 0, sizeof(extent_op->key));
4889 extent_op->flags_to_set = flags;
4890 extent_op->update_key = 1;
4891 extent_op->update_flags = 1;
4892 extent_op->is_data = 0;
4893
4894 ret = btrfs_add_delayed_tree_ref(trans, ins->objectid,
4895 ins->offset, parent, root_objectid,
4896 level, BTRFS_ADD_DELAYED_EXTENT,
4897 extent_op);
4898 BUG_ON(ret);
4899 }
4900 return ret;
4901 }
4902
4903 struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans,
4904 struct btrfs_root *root,
4905 u64 bytenr, u32 blocksize,
4906 int level)
4907 {
4908 struct extent_buffer *buf;
4909
4910 buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
4911 if (!buf)
4912 return ERR_PTR(-ENOMEM);
4913 btrfs_set_header_generation(buf, trans->transid);
4914 btrfs_set_buffer_lockdep_class(buf, level);
4915 btrfs_tree_lock(buf);
4916 clean_tree_block(trans, root, buf);
4917
4918 btrfs_set_lock_blocking(buf);
4919 btrfs_set_buffer_uptodate(buf);
4920
4921 if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) {
4922 set_extent_dirty(&root->dirty_log_pages, buf->start,
4923 buf->start + buf->len - 1, GFP_NOFS);
4924 } else {
4925 set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
4926 buf->start + buf->len - 1, GFP_NOFS);
4927 }
4928 trans->blocks_used++;
4929 /* this returns a buffer locked for blocking */
4930 return buf;
4931 }
4932
4933 /*
4934 * helper function to allocate a block for a given tree
4935 * returns the tree buffer or NULL.
4936 */
4937 struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
4938 struct btrfs_root *root, u32 blocksize,
4939 u64 parent, u64 root_objectid,
4940 struct btrfs_disk_key *key, int level,
4941 u64 hint, u64 empty_size)
4942 {
4943 struct btrfs_key ins;
4944 int ret;
4945 struct extent_buffer *buf;
4946
4947 ret = alloc_tree_block(trans, root, blocksize, parent, root_objectid,
4948 key, level, empty_size, hint, (u64)-1, &ins);
4949 if (ret) {
4950 BUG_ON(ret > 0);
4951 return ERR_PTR(ret);
4952 }
4953
4954 buf = btrfs_init_new_buffer(trans, root, ins.objectid,
4955 blocksize, level);
4956 return buf;
4957 }
4958
4959 struct walk_control {
4960 u64 refs[BTRFS_MAX_LEVEL];
4961 u64 flags[BTRFS_MAX_LEVEL];
4962 struct btrfs_key update_progress;
4963 int stage;
4964 int level;
4965 int shared_level;
4966 int update_ref;
4967 int keep_locks;
4968 int reada_slot;
4969 int reada_count;
4970 };
4971
4972 #define DROP_REFERENCE 1
4973 #define UPDATE_BACKREF 2
4974
4975 static noinline void reada_walk_down(struct btrfs_trans_handle *trans,
4976 struct btrfs_root *root,
4977 struct walk_control *wc,
4978 struct btrfs_path *path)
4979 {
4980 u64 bytenr;
4981 u64 generation;
4982 u64 refs;
4983 u64 flags;
4984 u64 last = 0;
4985 u32 nritems;
4986 u32 blocksize;
4987 struct btrfs_key key;
4988 struct extent_buffer *eb;
4989 int ret;
4990 int slot;
4991 int nread = 0;
4992
4993 if (path->slots[wc->level] < wc->reada_slot) {
4994 wc->reada_count = wc->reada_count * 2 / 3;
4995 wc->reada_count = max(wc->reada_count, 2);
4996 } else {
4997 wc->reada_count = wc->reada_count * 3 / 2;
4998 wc->reada_count = min_t(int, wc->reada_count,
4999 BTRFS_NODEPTRS_PER_BLOCK(root));
5000 }
5001
5002 eb = path->nodes[wc->level];
5003 nritems = btrfs_header_nritems(eb);
5004 blocksize = btrfs_level_size(root, wc->level - 1);
5005
5006 for (slot = path->slots[wc->level]; slot < nritems; slot++) {
5007 if (nread >= wc->reada_count)
5008 break;
5009
5010 cond_resched();
5011 bytenr = btrfs_node_blockptr(eb, slot);
5012 generation = btrfs_node_ptr_generation(eb, slot);
5013
5014 if (slot == path->slots[wc->level])
5015 goto reada;
5016
5017 if (wc->stage == UPDATE_BACKREF &&
5018 generation <= root->root_key.offset)
5019 continue;
5020
5021 /* We don't lock the tree block, it's OK to be racy here */
5022 ret = btrfs_lookup_extent_info(trans, root, bytenr, blocksize,
5023 &refs, &flags);
5024 BUG_ON(ret);
5025 BUG_ON(refs == 0);
5026
5027 if (wc->stage == DROP_REFERENCE) {
5028 if (refs == 1)
5029 goto reada;
5030
5031 if (wc->level == 1 &&
5032 (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
5033 continue;
5034 if (!wc->update_ref ||
5035 generation <= root->root_key.offset)
5036 continue;
5037 btrfs_node_key_to_cpu(eb, &key, slot);
5038 ret = btrfs_comp_cpu_keys(&key,
5039 &wc->update_progress);
5040 if (ret < 0)
5041 continue;
5042 } else {
5043 if (wc->level == 1 &&
5044 (flags & BTRFS_BLOCK_FLAG_FULL_BACKREF))
5045 continue;
5046 }
5047 reada:
5048 ret = readahead_tree_block(root, bytenr, blocksize,
5049 generation);
5050 if (ret)
5051 break;
5052 last = bytenr + blocksize;
5053 nread++;
5054 }
5055 wc->reada_slot = slot;
5056 }
5057
5058 /*
5059 * hepler to process tree block while walking down the tree.
5060 *
5061 * when wc->stage == UPDATE_BACKREF, this function updates
5062 * back refs for pointers in the block.
5063 *
5064 * NOTE: return value 1 means we should stop walking down.
5065 */
5066 static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
5067 struct btrfs_root *root,
5068 struct btrfs_path *path,
5069 struct walk_control *wc, int lookup_info)
5070 {
5071 int level = wc->level;
5072 struct extent_buffer *eb = path->nodes[level];
5073 u64 flag = BTRFS_BLOCK_FLAG_FULL_BACKREF;
5074 int ret;
5075
5076 if (wc->stage == UPDATE_BACKREF &&
5077 btrfs_header_owner(eb) != root->root_key.objectid)
5078 return 1;
5079
5080 /*
5081 * when reference count of tree block is 1, it won't increase
5082 * again. once full backref flag is set, we never clear it.
5083 */
5084 if (lookup_info &&
5085 ((wc->stage == DROP_REFERENCE && wc->refs[level] != 1) ||
5086 (wc->stage == UPDATE_BACKREF && !(wc->flags[level] & flag)))) {
5087 BUG_ON(!path->locks[level]);
5088 ret = btrfs_lookup_extent_info(trans, root,
5089 eb->start, eb->len,
5090 &wc->refs[level],
5091 &wc->flags[level]);
5092 BUG_ON(ret);
5093 BUG_ON(wc->refs[level] == 0);
5094 }
5095
5096 if (wc->stage == DROP_REFERENCE) {
5097 if (wc->refs[level] > 1)
5098 return 1;
5099
5100 if (path->locks[level] && !wc->keep_locks) {
5101 btrfs_tree_unlock(eb);
5102 path->locks[level] = 0;
5103 }
5104 return 0;
5105 }
5106
5107 /* wc->stage == UPDATE_BACKREF */
5108 if (!(wc->flags[level] & flag)) {
5109 BUG_ON(!path->locks[level]);
5110 ret = btrfs_inc_ref(trans, root, eb, 1);
5111 BUG_ON(ret);
5112 ret = btrfs_dec_ref(trans, root, eb, 0);
5113 BUG_ON(ret);
5114 ret = btrfs_set_disk_extent_flags(trans, root, eb->start,
5115 eb->len, flag, 0);
5116 BUG_ON(ret);
5117 wc->flags[level] |= flag;
5118 }
5119
5120 /*
5121 * the block is shared by multiple trees, so it's not good to
5122 * keep the tree lock
5123 */
5124 if (path->locks[level] && level > 0) {
5125 btrfs_tree_unlock(eb);
5126 path->locks[level] = 0;
5127 }
5128 return 0;
5129 }
5130
5131 /*
5132 * hepler to process tree block pointer.
5133 *
5134 * when wc->stage == DROP_REFERENCE, this function checks
5135 * reference count of the block pointed to. if the block
5136 * is shared and we need update back refs for the subtree
5137 * rooted at the block, this function changes wc->stage to
5138 * UPDATE_BACKREF. if the block is shared and there is no
5139 * need to update back, this function drops the reference
5140 * to the block.
5141 *
5142 * NOTE: return value 1 means we should stop walking down.
5143 */
5144 static noinline int do_walk_down(struct btrfs_trans_handle *trans,
5145 struct btrfs_root *root,
5146 struct btrfs_path *path,
5147 struct walk_control *wc, int *lookup_info)
5148 {
5149 u64 bytenr;
5150 u64 generation;
5151 u64 parent;
5152 u32 blocksize;
5153 struct btrfs_key key;
5154 struct extent_buffer *next;
5155 int level = wc->level;
5156 int reada = 0;
5157 int ret = 0;
5158
5159 generation = btrfs_node_ptr_generation(path->nodes[level],
5160 path->slots[level]);
5161 /*
5162 * if the lower level block was created before the snapshot
5163 * was created, we know there is no need to update back refs
5164 * for the subtree
5165 */
5166 if (wc->stage == UPDATE_BACKREF &&
5167 generation <= root->root_key.offset) {
5168 *lookup_info = 1;
5169 return 1;
5170 }
5171
5172 bytenr = btrfs_node_blockptr(path->nodes[level], path->slots[level]);
5173 blocksize = btrfs_level_size(root, level - 1);
5174
5175 next = btrfs_find_tree_block(root, bytenr, blocksize);
5176 if (!next) {
5177 next = btrfs_find_create_tree_block(root, bytenr, blocksize);
5178 reada = 1;
5179 }
5180 btrfs_tree_lock(next);
5181 btrfs_set_lock_blocking(next);
5182
5183 ret = btrfs_lookup_extent_info(trans, root, bytenr, blocksize,
5184 &wc->refs[level - 1],
5185 &wc->flags[level - 1]);
5186 BUG_ON(ret);
5187 BUG_ON(wc->refs[level - 1] == 0);
5188 *lookup_info = 0;
5189
5190 if (wc->stage == DROP_REFERENCE) {
5191 if (wc->refs[level - 1] > 1) {
5192 if (level == 1 &&
5193 (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
5194 goto skip;
5195
5196 if (!wc->update_ref ||
5197 generation <= root->root_key.offset)
5198 goto skip;
5199
5200 btrfs_node_key_to_cpu(path->nodes[level], &key,
5201 path->slots[level]);
5202 ret = btrfs_comp_cpu_keys(&key, &wc->update_progress);
5203 if (ret < 0)
5204 goto skip;
5205
5206 wc->stage = UPDATE_BACKREF;
5207 wc->shared_level = level - 1;
5208 }
5209 } else {
5210 if (level == 1 &&
5211 (wc->flags[0] & BTRFS_BLOCK_FLAG_FULL_BACKREF))
5212 goto skip;
5213 }
5214
5215 if (!btrfs_buffer_uptodate(next, generation)) {
5216 btrfs_tree_unlock(next);
5217 free_extent_buffer(next);
5218 next = NULL;
5219 *lookup_info = 1;
5220 }
5221
5222 if (!next) {
5223 if (reada && level == 1)
5224 reada_walk_down(trans, root, wc, path);
5225 next = read_tree_block(root, bytenr, blocksize, generation);
5226 btrfs_tree_lock(next);
5227 btrfs_set_lock_blocking(next);
5228 }
5229
5230 level--;
5231 BUG_ON(level != btrfs_header_level(next));
5232 path->nodes[level] = next;
5233 path->slots[level] = 0;
5234 path->locks[level] = 1;
5235 wc->level = level;
5236 if (wc->level == 1)
5237 wc->reada_slot = 0;
5238 return 0;
5239 skip:
5240 wc->refs[level - 1] = 0;
5241 wc->flags[level - 1] = 0;
5242 if (wc->stage == DROP_REFERENCE) {
5243 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF) {
5244 parent = path->nodes[level]->start;
5245 } else {
5246 BUG_ON(root->root_key.objectid !=
5247 btrfs_header_owner(path->nodes[level]));
5248 parent = 0;
5249 }
5250
5251 ret = btrfs_free_extent(trans, root, bytenr, blocksize, parent,
5252 root->root_key.objectid, level - 1, 0);
5253 BUG_ON(ret);
5254 }
5255 btrfs_tree_unlock(next);
5256 free_extent_buffer(next);
5257 *lookup_info = 1;
5258 return 1;
5259 }
5260
5261 /*
5262 * hepler to process tree block while walking up the tree.
5263 *
5264 * when wc->stage == DROP_REFERENCE, this function drops
5265 * reference count on the block.
5266 *
5267 * when wc->stage == UPDATE_BACKREF, this function changes
5268 * wc->stage back to DROP_REFERENCE if we changed wc->stage
5269 * to UPDATE_BACKREF previously while processing the block.
5270 *
5271 * NOTE: return value 1 means we should stop walking up.
5272 */
5273 static noinline int walk_up_proc(struct btrfs_trans_handle *trans,
5274 struct btrfs_root *root,
5275 struct btrfs_path *path,
5276 struct walk_control *wc)
5277 {
5278 int ret = 0;
5279 int level = wc->level;
5280 struct extent_buffer *eb = path->nodes[level];
5281 u64 parent = 0;
5282
5283 if (wc->stage == UPDATE_BACKREF) {
5284 BUG_ON(wc->shared_level < level);
5285 if (level < wc->shared_level)
5286 goto out;
5287
5288 ret = find_next_key(path, level + 1, &wc->update_progress);
5289 if (ret > 0)
5290 wc->update_ref = 0;
5291
5292 wc->stage = DROP_REFERENCE;
5293 wc->shared_level = -1;
5294 path->slots[level] = 0;
5295
5296 /*
5297 * check reference count again if the block isn't locked.
5298 * we should start walking down the tree again if reference
5299 * count is one.
5300 */
5301 if (!path->locks[level]) {
5302 BUG_ON(level == 0);
5303 btrfs_tree_lock(eb);
5304 btrfs_set_lock_blocking(eb);
5305 path->locks[level] = 1;
5306
5307 ret = btrfs_lookup_extent_info(trans, root,
5308 eb->start, eb->len,
5309 &wc->refs[level],
5310 &wc->flags[level]);
5311 BUG_ON(ret);
5312 BUG_ON(wc->refs[level] == 0);
5313 if (wc->refs[level] == 1) {
5314 btrfs_tree_unlock(eb);
5315 path->locks[level] = 0;
5316 return 1;
5317 }
5318 }
5319 }
5320
5321 /* wc->stage == DROP_REFERENCE */
5322 BUG_ON(wc->refs[level] > 1 && !path->locks[level]);
5323
5324 if (wc->refs[level] == 1) {
5325 if (level == 0) {
5326 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
5327 ret = btrfs_dec_ref(trans, root, eb, 1);
5328 else
5329 ret = btrfs_dec_ref(trans, root, eb, 0);
5330 BUG_ON(ret);
5331 }
5332 /* make block locked assertion in clean_tree_block happy */
5333 if (!path->locks[level] &&
5334 btrfs_header_generation(eb) == trans->transid) {
5335 btrfs_tree_lock(eb);
5336 btrfs_set_lock_blocking(eb);
5337 path->locks[level] = 1;
5338 }
5339 clean_tree_block(trans, root, eb);
5340 }
5341
5342 if (eb == root->node) {
5343 if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
5344 parent = eb->start;
5345 else
5346 BUG_ON(root->root_key.objectid !=
5347 btrfs_header_owner(eb));
5348 } else {
5349 if (wc->flags[level + 1] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
5350 parent = path->nodes[level + 1]->start;
5351 else
5352 BUG_ON(root->root_key.objectid !=
5353 btrfs_header_owner(path->nodes[level + 1]));
5354 }
5355
5356 ret = btrfs_free_extent(trans, root, eb->start, eb->len, parent,
5357 root->root_key.objectid, level, 0);
5358 BUG_ON(ret);
5359 out:
5360 wc->refs[level] = 0;
5361 wc->flags[level] = 0;
5362 return ret;
5363 }
5364
5365 static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
5366 struct btrfs_root *root,
5367 struct btrfs_path *path,
5368 struct walk_control *wc)
5369 {
5370 int level = wc->level;
5371 int lookup_info = 1;
5372 int ret;
5373
5374 while (level >= 0) {
5375 if (path->slots[level] >=
5376 btrfs_header_nritems(path->nodes[level]))
5377 break;
5378
5379 ret = walk_down_proc(trans, root, path, wc, lookup_info);
5380 if (ret > 0)
5381 break;
5382
5383 if (level == 0)
5384 break;
5385
5386 ret = do_walk_down(trans, root, path, wc, &lookup_info);
5387 if (ret > 0) {
5388 path->slots[level]++;
5389 continue;
5390 }
5391 level = wc->level;
5392 }
5393 return 0;
5394 }
5395
5396 static noinline int walk_up_tree(struct btrfs_trans_handle *trans,
5397 struct btrfs_root *root,
5398 struct btrfs_path *path,
5399 struct walk_control *wc, int max_level)
5400 {
5401 int level = wc->level;
5402 int ret;
5403
5404 path->slots[level] = btrfs_header_nritems(path->nodes[level]);
5405 while (level < max_level && path->nodes[level]) {
5406 wc->level = level;
5407 if (path->slots[level] + 1 <
5408 btrfs_header_nritems(path->nodes[level])) {
5409 path->slots[level]++;
5410 return 0;
5411 } else {
5412 ret = walk_up_proc(trans, root, path, wc);
5413 if (ret > 0)
5414 return 0;
5415
5416 if (path->locks[level]) {
5417 btrfs_tree_unlock(path->nodes[level]);
5418 path->locks[level] = 0;
5419 }
5420 free_extent_buffer(path->nodes[level]);
5421 path->nodes[level] = NULL;
5422 level++;
5423 }
5424 }
5425 return 1;
5426 }
5427
5428 /*
5429 * drop a subvolume tree.
5430 *
5431 * this function traverses the tree freeing any blocks that only
5432 * referenced by the tree.
5433 *
5434 * when a shared tree block is found. this function decreases its
5435 * reference count by one. if update_ref is true, this function
5436 * also make sure backrefs for the shared block and all lower level
5437 * blocks are properly updated.
5438 */
5439 int btrfs_drop_snapshot(struct btrfs_root *root, int update_ref)
5440 {
5441 struct btrfs_path *path;
5442 struct btrfs_trans_handle *trans;
5443 struct btrfs_root *tree_root = root->fs_info->tree_root;
5444 struct btrfs_root_item *root_item = &root->root_item;
5445 struct walk_control *wc;
5446 struct btrfs_key key;
5447 int err = 0;
5448 int ret;
5449 int level;
5450
5451 path = btrfs_alloc_path();
5452 BUG_ON(!path);
5453
5454 wc = kzalloc(sizeof(*wc), GFP_NOFS);
5455 BUG_ON(!wc);
5456
5457 trans = btrfs_start_transaction(tree_root, 1);
5458
5459 if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
5460 level = btrfs_header_level(root->node);
5461 path->nodes[level] = btrfs_lock_root_node(root);
5462 btrfs_set_lock_blocking(path->nodes[level]);
5463 path->slots[level] = 0;
5464 path->locks[level] = 1;
5465 memset(&wc->update_progress, 0,
5466 sizeof(wc->update_progress));
5467 } else {
5468 btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
5469 memcpy(&wc->update_progress, &key,
5470 sizeof(wc->update_progress));
5471
5472 level = root_item->drop_level;
5473 BUG_ON(level == 0);
5474 path->lowest_level = level;
5475 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
5476 path->lowest_level = 0;
5477 if (ret < 0) {
5478 err = ret;
5479 goto out;
5480 }
5481 WARN_ON(ret > 0);
5482
5483 /*
5484 * unlock our path, this is safe because only this
5485 * function is allowed to delete this snapshot
5486 */
5487 btrfs_unlock_up_safe(path, 0);
5488
5489 level = btrfs_header_level(root->node);
5490 while (1) {
5491 btrfs_tree_lock(path->nodes[level]);
5492 btrfs_set_lock_blocking(path->nodes[level]);
5493
5494 ret = btrfs_lookup_extent_info(trans, root,
5495 path->nodes[level]->start,
5496 path->nodes[level]->len,
5497 &wc->refs[level],
5498 &wc->flags[level]);
5499 BUG_ON(ret);
5500 BUG_ON(wc->refs[level] == 0);
5501
5502 if (level == root_item->drop_level)
5503 break;
5504
5505 btrfs_tree_unlock(path->nodes[level]);
5506 WARN_ON(wc->refs[level] != 1);
5507 level--;
5508 }
5509 }
5510
5511 wc->level = level;
5512 wc->shared_level = -1;
5513 wc->stage = DROP_REFERENCE;
5514 wc->update_ref = update_ref;
5515 wc->keep_locks = 0;
5516 wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(root);
5517
5518 while (1) {
5519 ret = walk_down_tree(trans, root, path, wc);
5520 if (ret < 0) {
5521 err = ret;
5522 break;
5523 }
5524
5525 ret = walk_up_tree(trans, root, path, wc, BTRFS_MAX_LEVEL);
5526 if (ret < 0) {
5527 err = ret;
5528 break;
5529 }
5530
5531 if (ret > 0) {
5532 BUG_ON(wc->stage != DROP_REFERENCE);
5533 break;
5534 }
5535
5536 if (wc->stage == DROP_REFERENCE) {
5537 level = wc->level;
5538 btrfs_node_key(path->nodes[level],
5539 &root_item->drop_progress,
5540 path->slots[level]);
5541 root_item->drop_level = level;
5542 }
5543
5544 BUG_ON(wc->level == 0);
5545 if (trans->transaction->in_commit ||
5546 trans->transaction->delayed_refs.flushing) {
5547 ret = btrfs_update_root(trans, tree_root,
5548 &root->root_key,
5549 root_item);
5550 BUG_ON(ret);
5551
5552 btrfs_end_transaction(trans, tree_root);
5553 trans = btrfs_start_transaction(tree_root, 1);
5554 } else {
5555 unsigned long update;
5556 update = trans->delayed_ref_updates;
5557 trans->delayed_ref_updates = 0;
5558 if (update)
5559 btrfs_run_delayed_refs(trans, tree_root,
5560 update);
5561 }
5562 }
5563 btrfs_release_path(root, path);
5564 BUG_ON(err);
5565
5566 ret = btrfs_del_root(trans, tree_root, &root->root_key);
5567 BUG_ON(ret);
5568
5569 if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
5570 ret = btrfs_find_last_root(tree_root, root->root_key.objectid,
5571 NULL, NULL);
5572 BUG_ON(ret < 0);
5573 if (ret > 0) {
5574 ret = btrfs_del_orphan_item(trans, tree_root,
5575 root->root_key.objectid);
5576 BUG_ON(ret);
5577 }
5578 }
5579
5580 if (root->in_radix) {
5581 btrfs_free_fs_root(tree_root->fs_info, root);
5582 } else {
5583 free_extent_buffer(root->node);
5584 free_extent_buffer(root->commit_root);
5585 kfree(root);
5586 }
5587 out:
5588 btrfs_end_transaction(trans, tree_root);
5589 kfree(wc);
5590 btrfs_free_path(path);
5591 return err;
5592 }
5593
5594 /*
5595 * drop subtree rooted at tree block 'node'.
5596 *
5597 * NOTE: this function will unlock and release tree block 'node'
5598 */
5599 int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
5600 struct btrfs_root *root,
5601 struct extent_buffer *node,
5602 struct extent_buffer *parent)
5603 {
5604 struct btrfs_path *path;
5605 struct walk_control *wc;
5606 int level;
5607 int parent_level;
5608 int ret = 0;
5609 int wret;
5610
5611 BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
5612
5613 path = btrfs_alloc_path();
5614 BUG_ON(!path);
5615
5616 wc = kzalloc(sizeof(*wc), GFP_NOFS);
5617 BUG_ON(!wc);
5618
5619 btrfs_assert_tree_locked(parent);
5620 parent_level = btrfs_header_level(parent);
5621 extent_buffer_get(parent);
5622 path->nodes[parent_level] = parent;
5623 path->slots[parent_level] = btrfs_header_nritems(parent);
5624
5625 btrfs_assert_tree_locked(node);
5626 level = btrfs_header_level(node);
5627 path->nodes[level] = node;
5628 path->slots[level] = 0;
5629 path->locks[level] = 1;
5630
5631 wc->refs[parent_level] = 1;
5632 wc->flags[parent_level] = BTRFS_BLOCK_FLAG_FULL_BACKREF;
5633 wc->level = level;
5634 wc->shared_level = -1;
5635 wc->stage = DROP_REFERENCE;
5636 wc->update_ref = 0;
5637 wc->keep_locks = 1;
5638 wc->reada_count = BTRFS_NODEPTRS_PER_BLOCK(root);
5639
5640 while (1) {
5641 wret = walk_down_tree(trans, root, path, wc);
5642 if (wret < 0) {
5643 ret = wret;
5644 break;
5645 }
5646
5647 wret = walk_up_tree(trans, root, path, wc, parent_level);
5648 if (wret < 0)
5649 ret = wret;
5650 if (wret != 0)
5651 break;
5652 }
5653
5654 kfree(wc);
5655 btrfs_free_path(path);
5656 return ret;
5657 }
5658
5659 #if 0
5660 static unsigned long calc_ra(unsigned long start, unsigned long last,
5661 unsigned long nr)
5662 {
5663 return min(last, start + nr - 1);
5664 }
5665
5666 static noinline int relocate_inode_pages(struct inode *inode, u64 start,
5667 u64 len)
5668 {
5669 u64 page_start;
5670 u64 page_end;
5671 unsigned long first_index;
5672 unsigned long last_index;
5673 unsigned long i;
5674 struct page *page;
5675 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
5676 struct file_ra_state *ra;
5677 struct btrfs_ordered_extent *ordered;
5678 unsigned int total_read = 0;
5679 unsigned int total_dirty = 0;
5680 int ret = 0;
5681
5682 ra = kzalloc(sizeof(*ra), GFP_NOFS);
5683
5684 mutex_lock(&inode->i_mutex);
5685 first_index = start >> PAGE_CACHE_SHIFT;
5686 last_index = (start + len - 1) >> PAGE_CACHE_SHIFT;
5687
5688 /* make sure the dirty trick played by the caller work */
5689 ret = invalidate_inode_pages2_range(inode->i_mapping,
5690 first_index, last_index);
5691 if (ret)
5692 goto out_unlock;
5693
5694 file_ra_state_init(ra, inode->i_mapping);
5695
5696 for (i = first_index ; i <= last_index; i++) {
5697 if (total_read % ra->ra_pages == 0) {
5698 btrfs_force_ra(inode->i_mapping, ra, NULL, i,
5699 calc_ra(i, last_index, ra->ra_pages));
5700 }
5701 total_read++;
5702 again:
5703 if (((u64)i << PAGE_CACHE_SHIFT) > i_size_read(inode))
5704 BUG_ON(1);
5705 page = grab_cache_page(inode->i_mapping, i);
5706 if (!page) {
5707 ret = -ENOMEM;
5708 goto out_unlock;
5709 }
5710 if (!PageUptodate(page)) {
5711 btrfs_readpage(NULL, page);
5712 lock_page(page);
5713 if (!PageUptodate(page)) {
5714 unlock_page(page);
5715 page_cache_release(page);
5716 ret = -EIO;
5717 goto out_unlock;
5718 }
5719 }
5720 wait_on_page_writeback(page);
5721
5722 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
5723 page_end = page_start + PAGE_CACHE_SIZE - 1;
5724 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
5725
5726 ordered = btrfs_lookup_ordered_extent(inode, page_start);
5727 if (ordered) {
5728 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
5729 unlock_page(page);
5730 page_cache_release(page);
5731 btrfs_start_ordered_extent(inode, ordered, 1);
5732 btrfs_put_ordered_extent(ordered);
5733 goto again;
5734 }
5735 set_page_extent_mapped(page);
5736
5737 if (i == first_index)
5738 set_extent_bits(io_tree, page_start, page_end,
5739 EXTENT_BOUNDARY, GFP_NOFS);
5740 btrfs_set_extent_delalloc(inode, page_start, page_end);
5741
5742 set_page_dirty(page);
5743 total_dirty++;
5744
5745 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
5746 unlock_page(page);
5747 page_cache_release(page);
5748 }
5749
5750 out_unlock:
5751 kfree(ra);
5752 mutex_unlock(&inode->i_mutex);
5753 balance_dirty_pages_ratelimited_nr(inode->i_mapping, total_dirty);
5754 return ret;
5755 }
5756
5757 static noinline int relocate_data_extent(struct inode *reloc_inode,
5758 struct btrfs_key *extent_key,
5759 u64 offset)
5760 {
5761 struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
5762 struct extent_map_tree *em_tree = &BTRFS_I(reloc_inode)->extent_tree;
5763 struct extent_map *em;
5764 u64 start = extent_key->objectid - offset;
5765 u64 end = start + extent_key->offset - 1;
5766
5767 em = alloc_extent_map(GFP_NOFS);
5768 BUG_ON(!em || IS_ERR(em));
5769
5770 em->start = start;
5771 em->len = extent_key->offset;
5772 em->block_len = extent_key->offset;
5773 em->block_start = extent_key->objectid;
5774 em->bdev = root->fs_info->fs_devices->latest_bdev;
5775 set_bit(EXTENT_FLAG_PINNED, &em->flags);
5776
5777 /* setup extent map to cheat btrfs_readpage */
5778 lock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS);
5779 while (1) {
5780 int ret;
5781 write_lock(&em_tree->lock);
5782 ret = add_extent_mapping(em_tree, em);
5783 write_unlock(&em_tree->lock);
5784 if (ret != -EEXIST) {
5785 free_extent_map(em);
5786 break;
5787 }
5788 btrfs_drop_extent_cache(reloc_inode, start, end, 0);
5789 }
5790 unlock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS);
5791
5792 return relocate_inode_pages(reloc_inode, start, extent_key->offset);
5793 }
5794
5795 struct btrfs_ref_path {
5796 u64 extent_start;
5797 u64 nodes[BTRFS_MAX_LEVEL];
5798 u64 root_objectid;
5799 u64 root_generation;
5800 u64 owner_objectid;
5801 u32 num_refs;
5802 int lowest_level;
5803 int current_level;
5804 int shared_level;
5805
5806 struct btrfs_key node_keys[BTRFS_MAX_LEVEL];
5807 u64 new_nodes[BTRFS_MAX_LEVEL];
5808 };
5809
5810 struct disk_extent {
5811 u64 ram_bytes;
5812 u64 disk_bytenr;
5813 u64 disk_num_bytes;
5814 u64 offset;
5815 u64 num_bytes;
5816 u8 compression;
5817 u8 encryption;
5818 u16 other_encoding;
5819 };
5820
5821 static int is_cowonly_root(u64 root_objectid)
5822 {
5823 if (root_objectid == BTRFS_ROOT_TREE_OBJECTID ||
5824 root_objectid == BTRFS_EXTENT_TREE_OBJECTID ||
5825 root_objectid == BTRFS_CHUNK_TREE_OBJECTID ||
5826 root_objectid == BTRFS_DEV_TREE_OBJECTID ||
5827 root_objectid == BTRFS_TREE_LOG_OBJECTID ||
5828 root_objectid == BTRFS_CSUM_TREE_OBJECTID)
5829 return 1;
5830 return 0;
5831 }
5832
5833 static noinline int __next_ref_path(struct btrfs_trans_handle *trans,
5834 struct btrfs_root *extent_root,
5835 struct btrfs_ref_path *ref_path,
5836 int first_time)
5837 {
5838 struct extent_buffer *leaf;
5839 struct btrfs_path *path;
5840 struct btrfs_extent_ref *ref;
5841 struct btrfs_key key;
5842 struct btrfs_key found_key;
5843 u64 bytenr;
5844 u32 nritems;
5845 int level;
5846 int ret = 1;
5847
5848 path = btrfs_alloc_path();
5849 if (!path)
5850 return -ENOMEM;
5851
5852 if (first_time) {
5853 ref_path->lowest_level = -1;
5854 ref_path->current_level = -1;
5855 ref_path->shared_level = -1;
5856 goto walk_up;
5857 }
5858 walk_down:
5859 level = ref_path->current_level - 1;
5860 while (level >= -1) {
5861 u64 parent;
5862 if (level < ref_path->lowest_level)
5863 break;
5864
5865 if (level >= 0)
5866 bytenr = ref_path->nodes[level];
5867 else
5868 bytenr = ref_path->extent_start;
5869 BUG_ON(bytenr == 0);
5870
5871 parent = ref_path->nodes[level + 1];
5872 ref_path->nodes[level + 1] = 0;
5873 ref_path->current_level = level;
5874 BUG_ON(parent == 0);
5875
5876 key.objectid = bytenr;
5877 key.offset = parent + 1;
5878 key.type = BTRFS_EXTENT_REF_KEY;
5879
5880 ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 0);
5881 if (ret < 0)
5882 goto out;
5883 BUG_ON(ret == 0);
5884
5885 leaf = path->nodes[0];
5886 nritems = btrfs_header_nritems(leaf);
5887 if (path->slots[0] >= nritems) {
5888 ret = btrfs_next_leaf(extent_root, path);
5889 if (ret < 0)
5890 goto out;
5891 if (ret > 0)
5892 goto next;
5893 leaf = path->nodes[0];
5894 }
5895
5896 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
5897 if (found_key.objectid == bytenr &&
5898 found_key.type == BTRFS_EXTENT_REF_KEY) {
5899 if (level < ref_path->shared_level)
5900 ref_path->shared_level = level;
5901 goto found;
5902 }
5903 next:
5904 level--;
5905 btrfs_release_path(extent_root, path);
5906 cond_resched();
5907 }
5908 /* reached lowest level */
5909 ret = 1;
5910 goto out;
5911 walk_up:
5912 level = ref_path->current_level;
5913 while (level < BTRFS_MAX_LEVEL - 1) {
5914 u64 ref_objectid;
5915
5916 if (level >= 0)
5917 bytenr = ref_path->nodes[level];
5918 else
5919 bytenr = ref_path->extent_start;
5920
5921 BUG_ON(bytenr == 0);
5922
5923 key.objectid = bytenr;
5924 key.offset = 0;
5925 key.type = BTRFS_EXTENT_REF_KEY;
5926
5927 ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 0);
5928 if (ret < 0)
5929 goto out;
5930
5931 leaf = path->nodes[0];
5932 nritems = btrfs_header_nritems(leaf);
5933 if (path->slots[0] >= nritems) {
5934 ret = btrfs_next_leaf(extent_root, path);
5935 if (ret < 0)
5936 goto out;
5937 if (ret > 0) {
5938 /* the extent was freed by someone */
5939 if (ref_path->lowest_level == level)
5940 goto out;
5941 btrfs_release_path(extent_root, path);
5942 goto walk_down;
5943 }
5944 leaf = path->nodes[0];
5945 }
5946
5947 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
5948 if (found_key.objectid != bytenr ||
5949 found_key.type != BTRFS_EXTENT_REF_KEY) {
5950 /* the extent was freed by someone */
5951 if (ref_path->lowest_level == level) {
5952 ret = 1;
5953 goto out;
5954 }
5955 btrfs_release_path(extent_root, path);
5956 goto walk_down;
5957 }
5958 found:
5959 ref = btrfs_item_ptr(leaf, path->slots[0],
5960 struct btrfs_extent_ref);
5961 ref_objectid = btrfs_ref_objectid(leaf, ref);
5962 if (ref_objectid < BTRFS_FIRST_FREE_OBJECTID) {
5963 if (first_time) {
5964 level = (int)ref_objectid;
5965 BUG_ON(level >= BTRFS_MAX_LEVEL);
5966 ref_path->lowest_level = level;
5967 ref_path->current_level = level;
5968 ref_path->nodes[level] = bytenr;
5969 } else {
5970 WARN_ON(ref_objectid != level);
5971 }
5972 } else {
5973 WARN_ON(level != -1);
5974 }
5975 first_time = 0;
5976
5977 if (ref_path->lowest_level == level) {
5978 ref_path->owner_objectid = ref_objectid;
5979 ref_path->num_refs = btrfs_ref_num_refs(leaf, ref);
5980 }
5981
5982 /*
5983 * the block is tree root or the block isn't in reference
5984 * counted tree.
5985 */
5986 if (found_key.objectid == found_key.offset ||
5987 is_cowonly_root(btrfs_ref_root(leaf, ref))) {
5988 ref_path->root_objectid = btrfs_ref_root(leaf, ref);
5989 ref_path->root_generation =
5990 btrfs_ref_generation(leaf, ref);
5991 if (level < 0) {
5992 /* special reference from the tree log */
5993 ref_path->nodes[0] = found_key.offset;
5994 ref_path->current_level = 0;
5995 }
5996 ret = 0;
5997 goto out;
5998 }
5999
6000 level++;
6001 BUG_ON(ref_path->nodes[level] != 0);
6002 ref_path->nodes[level] = found_key.offset;
6003 ref_path->current_level = level;
6004
6005 /*
6006 * the reference was created in the running transaction,
6007 * no need to continue walking up.
6008 */
6009 if (btrfs_ref_generation(leaf, ref) == trans->transid) {
6010 ref_path->root_objectid = btrfs_ref_root(leaf, ref);
6011 ref_path->root_generation =
6012 btrfs_ref_generation(leaf, ref);
6013 ret = 0;
6014 goto out;
6015 }
6016
6017 btrfs_release_path(extent_root, path);
6018 cond_resched();
6019 }
6020 /* reached max tree level, but no tree root found. */
6021 BUG();
6022 out:
6023 btrfs_free_path(path);
6024 return ret;
6025 }
6026
6027 static int btrfs_first_ref_path(struct btrfs_trans_handle *trans,
6028 struct btrfs_root *extent_root,
6029 struct btrfs_ref_path *ref_path,
6030 u64 extent_start)
6031 {
6032 memset(ref_path, 0, sizeof(*ref_path));
6033 ref_path->extent_start = extent_start;
6034
6035 return __next_ref_path(trans, extent_root, ref_path, 1);
6036 }
6037
6038 static int btrfs_next_ref_path(struct btrfs_trans_handle *trans,
6039 struct btrfs_root *extent_root,
6040 struct btrfs_ref_path *ref_path)
6041 {
6042 return __next_ref_path(trans, extent_root, ref_path, 0);
6043 }
6044
6045 static noinline int get_new_locations(struct inode *reloc_inode,
6046 struct btrfs_key *extent_key,
6047 u64 offset, int no_fragment,
6048 struct disk_extent **extents,
6049 int *nr_extents)
6050 {
6051 struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
6052 struct btrfs_path *path;
6053 struct btrfs_file_extent_item *fi;
6054 struct extent_buffer *leaf;
6055 struct disk_extent *exts = *extents;
6056 struct btrfs_key found_key;
6057 u64 cur_pos;
6058 u64 last_byte;
6059 u32 nritems;
6060 int nr = 0;
6061 int max = *nr_extents;
6062 int ret;
6063
6064 WARN_ON(!no_fragment && *extents);
6065 if (!exts) {
6066 max = 1;
6067 exts = kmalloc(sizeof(*exts) * max, GFP_NOFS);
6068 if (!exts)
6069 return -ENOMEM;
6070 }
6071
6072 path = btrfs_alloc_path();
6073 BUG_ON(!path);
6074
6075 cur_pos = extent_key->objectid - offset;
6076 last_byte = extent_key->objectid + extent_key->offset;
6077 ret = btrfs_lookup_file_extent(NULL, root, path, reloc_inode->i_ino,
6078 cur_pos, 0);
6079 if (ret < 0)
6080 goto out;
6081 if (ret > 0) {
6082 ret = -ENOENT;
6083 goto out;
6084 }
6085
6086 while (1) {
6087 leaf = path->nodes[0];
6088 nritems = btrfs_header_nritems(leaf);
6089 if (path->slots[0] >= nritems) {
6090 ret = btrfs_next_leaf(root, path);
6091 if (ret < 0)
6092 goto out;
6093 if (ret > 0)
6094 break;
6095 leaf = path->nodes[0];
6096 }
6097
6098 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
6099 if (found_key.offset != cur_pos ||
6100 found_key.type != BTRFS_EXTENT_DATA_KEY ||
6101 found_key.objectid != reloc_inode->i_ino)
6102 break;
6103
6104 fi = btrfs_item_ptr(leaf, path->slots[0],
6105 struct btrfs_file_extent_item);
6106 if (btrfs_file_extent_type(leaf, fi) !=
6107 BTRFS_FILE_EXTENT_REG ||
6108 btrfs_file_extent_disk_bytenr(leaf, fi) == 0)
6109 break;
6110
6111 if (nr == max) {
6112 struct disk_extent *old = exts;
6113 max *= 2;
6114 exts = kzalloc(sizeof(*exts) * max, GFP_NOFS);
6115 memcpy(exts, old, sizeof(*exts) * nr);
6116 if (old != *extents)
6117 kfree(old);
6118 }
6119
6120 exts[nr].disk_bytenr =
6121 btrfs_file_extent_disk_bytenr(leaf, fi);
6122 exts[nr].disk_num_bytes =
6123 btrfs_file_extent_disk_num_bytes(leaf, fi);
6124 exts[nr].offset = btrfs_file_extent_offset(leaf, fi);
6125 exts[nr].num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
6126 exts[nr].ram_bytes = btrfs_file_extent_ram_bytes(leaf, fi);
6127 exts[nr].compression = btrfs_file_extent_compression(leaf, fi);
6128 exts[nr].encryption = btrfs_file_extent_encryption(leaf, fi);
6129 exts[nr].other_encoding = btrfs_file_extent_other_encoding(leaf,
6130 fi);
6131 BUG_ON(exts[nr].offset > 0);
6132 BUG_ON(exts[nr].compression || exts[nr].encryption);
6133 BUG_ON(exts[nr].num_bytes != exts[nr].disk_num_bytes);
6134
6135 cur_pos += exts[nr].num_bytes;
6136 nr++;
6137
6138 if (cur_pos + offset >= last_byte)
6139 break;
6140
6141 if (no_fragment) {
6142 ret = 1;
6143 goto out;
6144 }
6145 path->slots[0]++;
6146 }
6147
6148 BUG_ON(cur_pos + offset > last_byte);
6149 if (cur_pos + offset < last_byte) {
6150 ret = -ENOENT;
6151 goto out;
6152 }
6153 ret = 0;
6154 out:
6155 btrfs_free_path(path);
6156 if (ret) {
6157 if (exts != *extents)
6158 kfree(exts);
6159 } else {
6160 *extents = exts;
6161 *nr_extents = nr;
6162 }
6163 return ret;
6164 }
6165
6166 static noinline int replace_one_extent(struct btrfs_trans_handle *trans,
6167 struct btrfs_root *root,
6168 struct btrfs_path *path,
6169 struct btrfs_key *extent_key,
6170 struct btrfs_key *leaf_key,
6171 struct btrfs_ref_path *ref_path,
6172 struct disk_extent *new_extents,
6173 int nr_extents)
6174 {
6175 struct extent_buffer *leaf;
6176 struct btrfs_file_extent_item *fi;
6177 struct inode *inode = NULL;
6178 struct btrfs_key key;
6179 u64 lock_start = 0;
6180 u64 lock_end = 0;
6181 u64 num_bytes;
6182 u64 ext_offset;
6183 u64 search_end = (u64)-1;
6184 u32 nritems;
6185 int nr_scaned = 0;
6186 int extent_locked = 0;
6187 int extent_type;
6188 int ret;
6189
6190 memcpy(&key, leaf_key, sizeof(key));
6191 if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS) {
6192 if (key.objectid < ref_path->owner_objectid ||
6193 (key.objectid == ref_path->owner_objectid &&
6194 key.type < BTRFS_EXTENT_DATA_KEY)) {
6195 key.objectid = ref_path->owner_objectid;
6196 key.type = BTRFS_EXTENT_DATA_KEY;
6197 key.offset = 0;
6198 }
6199 }
6200
6201 while (1) {
6202 ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
6203 if (ret < 0)
6204 goto out;
6205
6206 leaf = path->nodes[0];
6207 nritems = btrfs_header_nritems(leaf);
6208 next:
6209 if (extent_locked && ret > 0) {
6210 /*
6211 * the file extent item was modified by someone
6212 * before the extent got locked.
6213 */
6214 unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
6215 lock_end, GFP_NOFS);
6216 extent_locked = 0;
6217 }
6218
6219 if (path->slots[0] >= nritems) {
6220 if (++nr_scaned > 2)
6221 break;
6222
6223 BUG_ON(extent_locked);
6224 ret = btrfs_next_leaf(root, path);
6225 if (ret < 0)
6226 goto out;
6227 if (ret > 0)
6228 break;
6229 leaf = path->nodes[0];
6230 nritems = btrfs_header_nritems(leaf);
6231 }
6232
6233 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
6234
6235 if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS) {
6236 if ((key.objectid > ref_path->owner_objectid) ||
6237 (key.objectid == ref_path->owner_objectid &&
6238 key.type > BTRFS_EXTENT_DATA_KEY) ||
6239 key.offset >= search_end)
6240 break;
6241 }
6242
6243 if (inode && key.objectid != inode->i_ino) {
6244 BUG_ON(extent_locked);
6245 btrfs_release_path(root, path);
6246 mutex_unlock(&inode->i_mutex);
6247 iput(inode);
6248 inode = NULL;
6249 continue;
6250 }
6251
6252 if (key.type != BTRFS_EXTENT_DATA_KEY) {
6253 path->slots[0]++;
6254 ret = 1;
6255 goto next;
6256 }
6257 fi = btrfs_item_ptr(leaf, path->slots[0],
6258 struct btrfs_file_extent_item);
6259 extent_type = btrfs_file_extent_type(leaf, fi);
6260 if ((extent_type != BTRFS_FILE_EXTENT_REG &&
6261 extent_type != BTRFS_FILE_EXTENT_PREALLOC) ||
6262 (btrfs_file_extent_disk_bytenr(leaf, fi) !=
6263 extent_key->objectid)) {
6264 path->slots[0]++;
6265 ret = 1;
6266 goto next;
6267 }
6268
6269 num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
6270 ext_offset = btrfs_file_extent_offset(leaf, fi);
6271
6272 if (search_end == (u64)-1) {
6273 search_end = key.offset - ext_offset +
6274 btrfs_file_extent_ram_bytes(leaf, fi);
6275 }
6276
6277 if (!extent_locked) {
6278 lock_start = key.offset;
6279 lock_end = lock_start + num_bytes - 1;
6280 } else {
6281 if (lock_start > key.offset ||
6282 lock_end + 1 < key.offset + num_bytes) {
6283 unlock_extent(&BTRFS_I(inode)->io_tree,
6284 lock_start, lock_end, GFP_NOFS);
6285 extent_locked = 0;
6286 }
6287 }
6288
6289 if (!inode) {
6290 btrfs_release_path(root, path);
6291
6292 inode = btrfs_iget_locked(root->fs_info->sb,
6293 key.objectid, root);
6294 if (inode->i_state & I_NEW) {
6295 BTRFS_I(inode)->root = root;
6296 BTRFS_I(inode)->location.objectid =
6297 key.objectid;
6298 BTRFS_I(inode)->location.type =
6299 BTRFS_INODE_ITEM_KEY;
6300 BTRFS_I(inode)->location.offset = 0;
6301 btrfs_read_locked_inode(inode);
6302 unlock_new_inode(inode);
6303 }
6304 /*
6305 * some code call btrfs_commit_transaction while
6306 * holding the i_mutex, so we can't use mutex_lock
6307 * here.
6308 */
6309 if (is_bad_inode(inode) ||
6310 !mutex_trylock(&inode->i_mutex)) {
6311 iput(inode);
6312 inode = NULL;
6313 key.offset = (u64)-1;
6314 goto skip;
6315 }
6316 }
6317
6318 if (!extent_locked) {
6319 struct btrfs_ordered_extent *ordered;
6320
6321 btrfs_release_path(root, path);
6322
6323 lock_extent(&BTRFS_I(inode)->io_tree, lock_start,
6324 lock_end, GFP_NOFS);
6325 ordered = btrfs_lookup_first_ordered_extent(inode,
6326 lock_end);
6327 if (ordered &&
6328 ordered->file_offset <= lock_end &&
6329 ordered->file_offset + ordered->len > lock_start) {
6330 unlock_extent(&BTRFS_I(inode)->io_tree,
6331 lock_start, lock_end, GFP_NOFS);
6332 btrfs_start_ordered_extent(inode, ordered, 1);
6333 btrfs_put_ordered_extent(ordered);
6334 key.offset += num_bytes;
6335 goto skip;
6336 }
6337 if (ordered)
6338 btrfs_put_ordered_extent(ordered);
6339
6340 extent_locked = 1;
6341 continue;
6342 }
6343
6344 if (nr_extents == 1) {
6345 /* update extent pointer in place */
6346 btrfs_set_file_extent_disk_bytenr(leaf, fi,
6347 new_extents[0].disk_bytenr);
6348 btrfs_set_file_extent_disk_num_bytes(leaf, fi,
6349 new_extents[0].disk_num_bytes);
6350 btrfs_mark_buffer_dirty(leaf);
6351
6352 btrfs_drop_extent_cache(inode, key.offset,
6353 key.offset + num_bytes - 1, 0);
6354
6355 ret = btrfs_inc_extent_ref(trans, root,
6356 new_extents[0].disk_bytenr,
6357 new_extents[0].disk_num_bytes,
6358 leaf->start,
6359 root->root_key.objectid,
6360 trans->transid,
6361 key.objectid);
6362 BUG_ON(ret);
6363
6364 ret = btrfs_free_extent(trans, root,
6365 extent_key->objectid,
6366 extent_key->offset,
6367 leaf->start,
6368 btrfs_header_owner(leaf),
6369 btrfs_header_generation(leaf),
6370 key.objectid, 0);
6371 BUG_ON(ret);
6372
6373 btrfs_release_path(root, path);
6374 key.offset += num_bytes;
6375 } else {
6376 BUG_ON(1);
6377 #if 0
6378 u64 alloc_hint;
6379 u64 extent_len;
6380 int i;
6381 /*
6382 * drop old extent pointer at first, then insert the
6383 * new pointers one bye one
6384 */
6385 btrfs_release_path(root, path);
6386 ret = btrfs_drop_extents(trans, root, inode, key.offset,
6387 key.offset + num_bytes,
6388 key.offset, &alloc_hint);
6389 BUG_ON(ret);
6390
6391 for (i = 0; i < nr_extents; i++) {
6392 if (ext_offset >= new_extents[i].num_bytes) {
6393 ext_offset -= new_extents[i].num_bytes;
6394 continue;
6395 }
6396 extent_len = min(new_extents[i].num_bytes -
6397 ext_offset, num_bytes);
6398
6399 ret = btrfs_insert_empty_item(trans, root,
6400 path, &key,
6401 sizeof(*fi));
6402 BUG_ON(ret);
6403
6404 leaf = path->nodes[0];
6405 fi = btrfs_item_ptr(leaf, path->slots[0],
6406 struct btrfs_file_extent_item);
6407 btrfs_set_file_extent_generation(leaf, fi,
6408 trans->transid);
6409 btrfs_set_file_extent_type(leaf, fi,
6410 BTRFS_FILE_EXTENT_REG);
6411 btrfs_set_file_extent_disk_bytenr(leaf, fi,
6412 new_extents[i].disk_bytenr);
6413 btrfs_set_file_extent_disk_num_bytes(leaf, fi,
6414 new_extents[i].disk_num_bytes);
6415 btrfs_set_file_extent_ram_bytes(leaf, fi,
6416 new_extents[i].ram_bytes);
6417
6418 btrfs_set_file_extent_compression(leaf, fi,
6419 new_extents[i].compression);
6420 btrfs_set_file_extent_encryption(leaf, fi,
6421 new_extents[i].encryption);
6422 btrfs_set_file_extent_other_encoding(leaf, fi,
6423 new_extents[i].other_encoding);
6424
6425 btrfs_set_file_extent_num_bytes(leaf, fi,
6426 extent_len);
6427 ext_offset += new_extents[i].offset;
6428 btrfs_set_file_extent_offset(leaf, fi,
6429 ext_offset);
6430 btrfs_mark_buffer_dirty(leaf);
6431
6432 btrfs_drop_extent_cache(inode, key.offset,
6433 key.offset + extent_len - 1, 0);
6434
6435 ret = btrfs_inc_extent_ref(trans, root,
6436 new_extents[i].disk_bytenr,
6437 new_extents[i].disk_num_bytes,
6438 leaf->start,
6439 root->root_key.objectid,
6440 trans->transid, key.objectid);
6441 BUG_ON(ret);
6442 btrfs_release_path(root, path);
6443
6444 inode_add_bytes(inode, extent_len);
6445
6446 ext_offset = 0;
6447 num_bytes -= extent_len;
6448 key.offset += extent_len;
6449
6450 if (num_bytes == 0)
6451 break;
6452 }
6453 BUG_ON(i >= nr_extents);
6454 #endif
6455 }
6456
6457 if (extent_locked) {
6458 unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
6459 lock_end, GFP_NOFS);
6460 extent_locked = 0;
6461 }
6462 skip:
6463 if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS &&
6464 key.offset >= search_end)
6465 break;
6466
6467 cond_resched();
6468 }
6469 ret = 0;
6470 out:
6471 btrfs_release_path(root, path);
6472 if (inode) {
6473 mutex_unlock(&inode->i_mutex);
6474 if (extent_locked) {
6475 unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
6476 lock_end, GFP_NOFS);
6477 }
6478 iput(inode);
6479 }
6480 return ret;
6481 }
6482
6483 int btrfs_reloc_tree_cache_ref(struct btrfs_trans_handle *trans,
6484 struct btrfs_root *root,
6485 struct extent_buffer *buf, u64 orig_start)
6486 {
6487 int level;
6488 int ret;
6489
6490 BUG_ON(btrfs_header_generation(buf) != trans->transid);
6491 BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
6492
6493 level = btrfs_header_level(buf);
6494 if (level == 0) {
6495 struct btrfs_leaf_ref *ref;
6496 struct btrfs_leaf_ref *orig_ref;
6497
6498 orig_ref = btrfs_lookup_leaf_ref(root, orig_start);
6499 if (!orig_ref)
6500 return -ENOENT;
6501
6502 ref = btrfs_alloc_leaf_ref(root, orig_ref->nritems);
6503 if (!ref) {
6504 btrfs_free_leaf_ref(root, orig_ref);
6505 return -ENOMEM;
6506 }
6507
6508 ref->nritems = orig_ref->nritems;
6509 memcpy(ref->extents, orig_ref->extents,
6510 sizeof(ref->extents[0]) * ref->nritems);
6511
6512 btrfs_free_leaf_ref(root, orig_ref);
6513
6514 ref->root_gen = trans->transid;
6515 ref->bytenr = buf->start;
6516 ref->owner = btrfs_header_owner(buf);
6517 ref->generation = btrfs_header_generation(buf);
6518
6519 ret = btrfs_add_leaf_ref(root, ref, 0);
6520 WARN_ON(ret);
6521 btrfs_free_leaf_ref(root, ref);
6522 }
6523 return 0;
6524 }
6525
6526 static noinline int invalidate_extent_cache(struct btrfs_root *root,
6527 struct extent_buffer *leaf,
6528 struct btrfs_block_group_cache *group,
6529 struct btrfs_root *target_root)
6530 {
6531 struct btrfs_key key;
6532 struct inode *inode = NULL;
6533 struct btrfs_file_extent_item *fi;
6534 u64 num_bytes;
6535 u64 skip_objectid = 0;
6536 u32 nritems;
6537 u32 i;
6538
6539 nritems = btrfs_header_nritems(leaf);
6540 for (i = 0; i < nritems; i++) {
6541 btrfs_item_key_to_cpu(leaf, &key, i);
6542 if (key.objectid == skip_objectid ||
6543 key.type != BTRFS_EXTENT_DATA_KEY)
6544 continue;
6545 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
6546 if (btrfs_file_extent_type(leaf, fi) ==
6547 BTRFS_FILE_EXTENT_INLINE)
6548 continue;
6549 if (btrfs_file_extent_disk_bytenr(leaf, fi) == 0)
6550 continue;
6551 if (!inode || inode->i_ino != key.objectid) {
6552 iput(inode);
6553 inode = btrfs_ilookup(target_root->fs_info->sb,
6554 key.objectid, target_root, 1);
6555 }
6556 if (!inode) {
6557 skip_objectid = key.objectid;
6558 continue;
6559 }
6560 num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
6561
6562 lock_extent(&BTRFS_I(inode)->io_tree, key.offset,
6563 key.offset + num_bytes - 1, GFP_NOFS);
6564 btrfs_drop_extent_cache(inode, key.offset,
6565 key.offset + num_bytes - 1, 1);
6566 unlock_extent(&BTRFS_I(inode)->io_tree, key.offset,
6567 key.offset + num_bytes - 1, GFP_NOFS);
6568 cond_resched();
6569 }
6570 iput(inode);
6571 return 0;
6572 }
6573
6574 static noinline int replace_extents_in_leaf(struct btrfs_trans_handle *trans,
6575 struct btrfs_root *root,
6576 struct extent_buffer *leaf,
6577 struct btrfs_block_group_cache *group,
6578 struct inode *reloc_inode)
6579 {
6580 struct btrfs_key key;
6581 struct btrfs_key extent_key;
6582 struct btrfs_file_extent_item *fi;
6583 struct btrfs_leaf_ref *ref;
6584 struct disk_extent *new_extent;
6585 u64 bytenr;
6586 u64 num_bytes;
6587 u32 nritems;
6588 u32 i;
6589 int ext_index;
6590 int nr_extent;
6591 int ret;
6592
6593 new_extent = kmalloc(sizeof(*new_extent), GFP_NOFS);
6594 BUG_ON(!new_extent);
6595
6596 ref = btrfs_lookup_leaf_ref(root, leaf->start);
6597 BUG_ON(!ref);
6598
6599 ext_index = -1;
6600 nritems = btrfs_header_nritems(leaf);
6601 for (i = 0; i < nritems; i++) {
6602 btrfs_item_key_to_cpu(leaf, &key, i);
6603 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
6604 continue;
6605 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
6606 if (btrfs_file_extent_type(leaf, fi) ==
6607 BTRFS_FILE_EXTENT_INLINE)
6608 continue;
6609 bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
6610 num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
6611 if (bytenr == 0)
6612 continue;
6613
6614 ext_index++;
6615 if (bytenr >= group->key.objectid + group->key.offset ||
6616 bytenr + num_bytes <= group->key.objectid)
6617 continue;
6618
6619 extent_key.objectid = bytenr;
6620 extent_key.offset = num_bytes;
6621 extent_key.type = BTRFS_EXTENT_ITEM_KEY;
6622 nr_extent = 1;
6623 ret = get_new_locations(reloc_inode, &extent_key,
6624 group->key.objectid, 1,
6625 &new_extent, &nr_extent);
6626 if (ret > 0)
6627 continue;
6628 BUG_ON(ret < 0);
6629
6630 BUG_ON(ref->extents[ext_index].bytenr != bytenr);
6631 BUG_ON(ref->extents[ext_index].num_bytes != num_bytes);
6632 ref->extents[ext_index].bytenr = new_extent->disk_bytenr;
6633 ref->extents[ext_index].num_bytes = new_extent->disk_num_bytes;
6634
6635 btrfs_set_file_extent_disk_bytenr(leaf, fi,
6636 new_extent->disk_bytenr);
6637 btrfs_set_file_extent_disk_num_bytes(leaf, fi,
6638 new_extent->disk_num_bytes);
6639 btrfs_mark_buffer_dirty(leaf);
6640
6641 ret = btrfs_inc_extent_ref(trans, root,
6642 new_extent->disk_bytenr,
6643 new_extent->disk_num_bytes,
6644 leaf->start,
6645 root->root_key.objectid,
6646 trans->transid, key.objectid);
6647 BUG_ON(ret);
6648
6649 ret = btrfs_free_extent(trans, root,
6650 bytenr, num_bytes, leaf->start,
6651 btrfs_header_owner(leaf),
6652 btrfs_header_generation(leaf),
6653 key.objectid, 0);
6654 BUG_ON(ret);
6655 cond_resched();
6656 }
6657 kfree(new_extent);
6658 BUG_ON(ext_index + 1 != ref->nritems);
6659 btrfs_free_leaf_ref(root, ref);
6660 return 0;
6661 }
6662
6663 int btrfs_free_reloc_root(struct btrfs_trans_handle *trans,
6664 struct btrfs_root *root)
6665 {
6666 struct btrfs_root *reloc_root;
6667 int ret;
6668
6669 if (root->reloc_root) {
6670 reloc_root = root->reloc_root;
6671 root->reloc_root = NULL;
6672 list_add(&reloc_root->dead_list,
6673 &root->fs_info->dead_reloc_roots);
6674
6675 btrfs_set_root_bytenr(&reloc_root->root_item,
6676 reloc_root->node->start);
6677 btrfs_set_root_level(&root->root_item,
6678 btrfs_header_level(reloc_root->node));
6679 memset(&reloc_root->root_item.drop_progress, 0,
6680 sizeof(struct btrfs_disk_key));
6681 reloc_root->root_item.drop_level = 0;
6682
6683 ret = btrfs_update_root(trans, root->fs_info->tree_root,
6684 &reloc_root->root_key,
6685 &reloc_root->root_item);
6686 BUG_ON(ret);
6687 }
6688 return 0;
6689 }
6690
6691 int btrfs_drop_dead_reloc_roots(struct btrfs_root *root)
6692 {
6693 struct btrfs_trans_handle *trans;
6694 struct btrfs_root *reloc_root;
6695 struct btrfs_root *prev_root = NULL;
6696 struct list_head dead_roots;
6697 int ret;
6698 unsigned long nr;
6699
6700 INIT_LIST_HEAD(&dead_roots);
6701 list_splice_init(&root->fs_info->dead_reloc_roots, &dead_roots);
6702
6703 while (!list_empty(&dead_roots)) {
6704 reloc_root = list_entry(dead_roots.prev,
6705 struct btrfs_root, dead_list);
6706 list_del_init(&reloc_root->dead_list);
6707
6708 BUG_ON(reloc_root->commit_root != NULL);
6709 while (1) {
6710 trans = btrfs_join_transaction(root, 1);
6711 BUG_ON(!trans);
6712
6713 mutex_lock(&root->fs_info->drop_mutex);
6714 ret = btrfs_drop_snapshot(trans, reloc_root);
6715 if (ret != -EAGAIN)
6716 break;
6717 mutex_unlock(&root->fs_info->drop_mutex);
6718
6719 nr = trans->blocks_used;
6720 ret = btrfs_end_transaction(trans, root);
6721 BUG_ON(ret);
6722 btrfs_btree_balance_dirty(root, nr);
6723 }
6724
6725 free_extent_buffer(reloc_root->node);
6726
6727 ret = btrfs_del_root(trans, root->fs_info->tree_root,
6728 &reloc_root->root_key);
6729 BUG_ON(ret);
6730 mutex_unlock(&root->fs_info->drop_mutex);
6731
6732 nr = trans->blocks_used;
6733 ret = btrfs_end_transaction(trans, root);
6734 BUG_ON(ret);
6735 btrfs_btree_balance_dirty(root, nr);
6736
6737 kfree(prev_root);
6738 prev_root = reloc_root;
6739 }
6740 if (prev_root) {
6741 btrfs_remove_leaf_refs(prev_root, (u64)-1, 0);
6742 kfree(prev_root);
6743 }
6744 return 0;
6745 }
6746
6747 int btrfs_add_dead_reloc_root(struct btrfs_root *root)
6748 {
6749 list_add(&root->dead_list, &root->fs_info->dead_reloc_roots);
6750 return 0;
6751 }
6752
6753 int btrfs_cleanup_reloc_trees(struct btrfs_root *root)
6754 {
6755 struct btrfs_root *reloc_root;
6756 struct btrfs_trans_handle *trans;
6757 struct btrfs_key location;
6758 int found;
6759 int ret;
6760
6761 mutex_lock(&root->fs_info->tree_reloc_mutex);
6762 ret = btrfs_find_dead_roots(root, BTRFS_TREE_RELOC_OBJECTID, NULL);
6763 BUG_ON(ret);
6764 found = !list_empty(&root->fs_info->dead_reloc_roots);
6765 mutex_unlock(&root->fs_info->tree_reloc_mutex);
6766
6767 if (found) {
6768 trans = btrfs_start_transaction(root, 1);
6769 BUG_ON(!trans);
6770 ret = btrfs_commit_transaction(trans, root);
6771 BUG_ON(ret);
6772 }
6773
6774 location.objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
6775 location.offset = (u64)-1;
6776 location.type = BTRFS_ROOT_ITEM_KEY;
6777
6778 reloc_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
6779 BUG_ON(!reloc_root);
6780 btrfs_orphan_cleanup(reloc_root);
6781 return 0;
6782 }
6783
6784 static noinline int init_reloc_tree(struct btrfs_trans_handle *trans,
6785 struct btrfs_root *root)
6786 {
6787 struct btrfs_root *reloc_root;
6788 struct extent_buffer *eb;
6789 struct btrfs_root_item *root_item;
6790 struct btrfs_key root_key;
6791 int ret;
6792
6793 BUG_ON(!root->ref_cows);
6794 if (root->reloc_root)
6795 return 0;
6796
6797 root_item = kmalloc(sizeof(*root_item), GFP_NOFS);
6798 BUG_ON(!root_item);
6799
6800 ret = btrfs_copy_root(trans, root, root->commit_root,
6801 &eb, BTRFS_TREE_RELOC_OBJECTID);
6802 BUG_ON(ret);
6803
6804 root_key.objectid = BTRFS_TREE_RELOC_OBJECTID;
6805 root_key.offset = root->root_key.objectid;
6806 root_key.type = BTRFS_ROOT_ITEM_KEY;
6807
6808 memcpy(root_item, &root->root_item, sizeof(root_item));
6809 btrfs_set_root_refs(root_item, 0);
6810 btrfs_set_root_bytenr(root_item, eb->start);
6811 btrfs_set_root_level(root_item, btrfs_header_level(eb));
6812 btrfs_set_root_generation(root_item, trans->transid);
6813
6814 btrfs_tree_unlock(eb);
6815 free_extent_buffer(eb);
6816
6817 ret = btrfs_insert_root(trans, root->fs_info->tree_root,
6818 &root_key, root_item);
6819 BUG_ON(ret);
6820 kfree(root_item);
6821
6822 reloc_root = btrfs_read_fs_root_no_radix(root->fs_info->tree_root,
6823 &root_key);
6824 BUG_ON(!reloc_root);
6825 reloc_root->last_trans = trans->transid;
6826 reloc_root->commit_root = NULL;
6827 reloc_root->ref_tree = &root->fs_info->reloc_ref_tree;
6828
6829 root->reloc_root = reloc_root;
6830 return 0;
6831 }
6832
6833 /*
6834 * Core function of space balance.
6835 *
6836 * The idea is using reloc trees to relocate tree blocks in reference
6837 * counted roots. There is one reloc tree for each subvol, and all
6838 * reloc trees share same root key objectid. Reloc trees are snapshots
6839 * of the latest committed roots of subvols (root->commit_root).
6840 *
6841 * To relocate a tree block referenced by a subvol, there are two steps.
6842 * COW the block through subvol's reloc tree, then update block pointer
6843 * in the subvol to point to the new block. Since all reloc trees share
6844 * same root key objectid, doing special handing for tree blocks owned
6845 * by them is easy. Once a tree block has been COWed in one reloc tree,
6846 * we can use the resulting new block directly when the same block is
6847 * required to COW again through other reloc trees. By this way, relocated
6848 * tree blocks are shared between reloc trees, so they are also shared
6849 * between subvols.
6850 */
6851 static noinline int relocate_one_path(struct btrfs_trans_handle *trans,
6852 struct btrfs_root *root,
6853 struct btrfs_path *path,
6854 struct btrfs_key *first_key,
6855 struct btrfs_ref_path *ref_path,
6856 struct btrfs_block_group_cache *group,
6857 struct inode *reloc_inode)
6858 {
6859 struct btrfs_root *reloc_root;
6860 struct extent_buffer *eb = NULL;
6861 struct btrfs_key *keys;
6862 u64 *nodes;
6863 int level;
6864 int shared_level;
6865 int lowest_level = 0;
6866 int ret;
6867
6868 if (ref_path->owner_objectid < BTRFS_FIRST_FREE_OBJECTID)
6869 lowest_level = ref_path->owner_objectid;
6870
6871 if (!root->ref_cows) {
6872 path->lowest_level = lowest_level;
6873 ret = btrfs_search_slot(trans, root, first_key, path, 0, 1);
6874 BUG_ON(ret < 0);
6875 path->lowest_level = 0;
6876 btrfs_release_path(root, path);
6877 return 0;
6878 }
6879
6880 mutex_lock(&root->fs_info->tree_reloc_mutex);
6881 ret = init_reloc_tree(trans, root);
6882 BUG_ON(ret);
6883 reloc_root = root->reloc_root;
6884
6885 shared_level = ref_path->shared_level;
6886 ref_path->shared_level = BTRFS_MAX_LEVEL - 1;
6887
6888 keys = ref_path->node_keys;
6889 nodes = ref_path->new_nodes;
6890 memset(&keys[shared_level + 1], 0,
6891 sizeof(*keys) * (BTRFS_MAX_LEVEL - shared_level - 1));
6892 memset(&nodes[shared_level + 1], 0,
6893 sizeof(*nodes) * (BTRFS_MAX_LEVEL - shared_level - 1));
6894
6895 if (nodes[lowest_level] == 0) {
6896 path->lowest_level = lowest_level;
6897 ret = btrfs_search_slot(trans, reloc_root, first_key, path,
6898 0, 1);
6899 BUG_ON(ret);
6900 for (level = lowest_level; level < BTRFS_MAX_LEVEL; level++) {
6901 eb = path->nodes[level];
6902 if (!eb || eb == reloc_root->node)
6903 break;
6904 nodes[level] = eb->start;
6905 if (level == 0)
6906 btrfs_item_key_to_cpu(eb, &keys[level], 0);
6907 else
6908 btrfs_node_key_to_cpu(eb, &keys[level], 0);
6909 }
6910 if (nodes[0] &&
6911 ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
6912 eb = path->nodes[0];
6913 ret = replace_extents_in_leaf(trans, reloc_root, eb,
6914 group, reloc_inode);
6915 BUG_ON(ret);
6916 }
6917 btrfs_release_path(reloc_root, path);
6918 } else {
6919 ret = btrfs_merge_path(trans, reloc_root, keys, nodes,
6920 lowest_level);
6921 BUG_ON(ret);
6922 }
6923
6924 /*
6925 * replace tree blocks in the fs tree with tree blocks in
6926 * the reloc tree.
6927 */
6928 ret = btrfs_merge_path(trans, root, keys, nodes, lowest_level);
6929 BUG_ON(ret < 0);
6930
6931 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
6932 ret = btrfs_search_slot(trans, reloc_root, first_key, path,
6933 0, 0);
6934 BUG_ON(ret);
6935 extent_buffer_get(path->nodes[0]);
6936 eb = path->nodes[0];
6937 btrfs_release_path(reloc_root, path);
6938 ret = invalidate_extent_cache(reloc_root, eb, group, root);
6939 BUG_ON(ret);
6940 free_extent_buffer(eb);
6941 }
6942
6943 mutex_unlock(&root->fs_info->tree_reloc_mutex);
6944 path->lowest_level = 0;
6945 return 0;
6946 }
6947
6948 static noinline int relocate_tree_block(struct btrfs_trans_handle *trans,
6949 struct btrfs_root *root,
6950 struct btrfs_path *path,
6951 struct btrfs_key *first_key,
6952 struct btrfs_ref_path *ref_path)
6953 {
6954 int ret;
6955
6956 ret = relocate_one_path(trans, root, path, first_key,
6957 ref_path, NULL, NULL);
6958 BUG_ON(ret);
6959
6960 return 0;
6961 }
6962
6963 static noinline int del_extent_zero(struct btrfs_trans_handle *trans,
6964 struct btrfs_root *extent_root,
6965 struct btrfs_path *path,
6966 struct btrfs_key *extent_key)
6967 {
6968 int ret;
6969
6970 ret = btrfs_search_slot(trans, extent_root, extent_key, path, -1, 1);
6971 if (ret)
6972 goto out;
6973 ret = btrfs_del_item(trans, extent_root, path);
6974 out:
6975 btrfs_release_path(extent_root, path);
6976 return ret;
6977 }
6978
6979 static noinline struct btrfs_root *read_ref_root(struct btrfs_fs_info *fs_info,
6980 struct btrfs_ref_path *ref_path)
6981 {
6982 struct btrfs_key root_key;
6983
6984 root_key.objectid = ref_path->root_objectid;
6985 root_key.type = BTRFS_ROOT_ITEM_KEY;
6986 if (is_cowonly_root(ref_path->root_objectid))
6987 root_key.offset = 0;
6988 else
6989 root_key.offset = (u64)-1;
6990
6991 return btrfs_read_fs_root_no_name(fs_info, &root_key);
6992 }
6993
6994 static noinline int relocate_one_extent(struct btrfs_root *extent_root,
6995 struct btrfs_path *path,
6996 struct btrfs_key *extent_key,
6997 struct btrfs_block_group_cache *group,
6998 struct inode *reloc_inode, int pass)
6999 {
7000 struct btrfs_trans_handle *trans;
7001 struct btrfs_root *found_root;
7002 struct btrfs_ref_path *ref_path = NULL;
7003 struct disk_extent *new_extents = NULL;
7004 int nr_extents = 0;
7005 int loops;
7006 int ret;
7007 int level;
7008 struct btrfs_key first_key;
7009 u64 prev_block = 0;
7010
7011
7012 trans = btrfs_start_transaction(extent_root, 1);
7013 BUG_ON(!trans);
7014
7015 if (extent_key->objectid == 0) {
7016 ret = del_extent_zero(trans, extent_root, path, extent_key);
7017 goto out;
7018 }
7019
7020 ref_path = kmalloc(sizeof(*ref_path), GFP_NOFS);
7021 if (!ref_path) {
7022 ret = -ENOMEM;
7023 goto out;
7024 }
7025
7026 for (loops = 0; ; loops++) {
7027 if (loops == 0) {
7028 ret = btrfs_first_ref_path(trans, extent_root, ref_path,
7029 extent_key->objectid);
7030 } else {
7031 ret = btrfs_next_ref_path(trans, extent_root, ref_path);
7032 }
7033 if (ret < 0)
7034 goto out;
7035 if (ret > 0)
7036 break;
7037
7038 if (ref_path->root_objectid == BTRFS_TREE_LOG_OBJECTID ||
7039 ref_path->root_objectid == BTRFS_TREE_RELOC_OBJECTID)
7040 continue;
7041
7042 found_root = read_ref_root(extent_root->fs_info, ref_path);
7043 BUG_ON(!found_root);
7044 /*
7045 * for reference counted tree, only process reference paths
7046 * rooted at the latest committed root.
7047 */
7048 if (found_root->ref_cows &&
7049 ref_path->root_generation != found_root->root_key.offset)
7050 continue;
7051
7052 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
7053 if (pass == 0) {
7054 /*
7055 * copy data extents to new locations
7056 */
7057 u64 group_start = group->key.objectid;
7058 ret = relocate_data_extent(reloc_inode,
7059 extent_key,
7060 group_start);
7061 if (ret < 0)
7062 goto out;
7063 break;
7064 }
7065 level = 0;
7066 } else {
7067 level = ref_path->owner_objectid;
7068 }
7069
7070 if (prev_block != ref_path->nodes[level]) {
7071 struct extent_buffer *eb;
7072 u64 block_start = ref_path->nodes[level];
7073 u64 block_size = btrfs_level_size(found_root, level);
7074
7075 eb = read_tree_block(found_root, block_start,
7076 block_size, 0);
7077 btrfs_tree_lock(eb);
7078 BUG_ON(level != btrfs_header_level(eb));
7079
7080 if (level == 0)
7081 btrfs_item_key_to_cpu(eb, &first_key, 0);
7082 else
7083 btrfs_node_key_to_cpu(eb, &first_key, 0);
7084
7085 btrfs_tree_unlock(eb);
7086 free_extent_buffer(eb);
7087 prev_block = block_start;
7088 }
7089
7090 mutex_lock(&extent_root->fs_info->trans_mutex);
7091 btrfs_record_root_in_trans(found_root);
7092 mutex_unlock(&extent_root->fs_info->trans_mutex);
7093 if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
7094 /*
7095 * try to update data extent references while
7096 * keeping metadata shared between snapshots.
7097 */
7098 if (pass == 1) {
7099 ret = relocate_one_path(trans, found_root,
7100 path, &first_key, ref_path,
7101 group, reloc_inode);
7102 if (ret < 0)
7103 goto out;
7104 continue;
7105 }
7106 /*
7107 * use fallback method to process the remaining
7108 * references.
7109 */
7110 if (!new_extents) {
7111 u64 group_start = group->key.objectid;
7112 new_extents = kmalloc(sizeof(*new_extents),
7113 GFP_NOFS);
7114 nr_extents = 1;
7115 ret = get_new_locations(reloc_inode,
7116 extent_key,
7117 group_start, 1,
7118 &new_extents,
7119 &nr_extents);
7120 if (ret)
7121 goto out;
7122 }
7123 ret = replace_one_extent(trans, found_root,
7124 path, extent_key,
7125 &first_key, ref_path,
7126 new_extents, nr_extents);
7127 } else {
7128 ret = relocate_tree_block(trans, found_root, path,
7129 &first_key, ref_path);
7130 }
7131 if (ret < 0)
7132 goto out;
7133 }
7134 ret = 0;
7135 out:
7136 btrfs_end_transaction(trans, extent_root);
7137 kfree(new_extents);
7138 kfree(ref_path);
7139 return ret;
7140 }
7141 #endif
7142
7143 static u64 update_block_group_flags(struct btrfs_root *root, u64 flags)
7144 {
7145 u64 num_devices;
7146 u64 stripped = BTRFS_BLOCK_GROUP_RAID0 |
7147 BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10;
7148
7149 num_devices = root->fs_info->fs_devices->rw_devices;
7150 if (num_devices == 1) {
7151 stripped |= BTRFS_BLOCK_GROUP_DUP;
7152 stripped = flags & ~stripped;
7153
7154 /* turn raid0 into single device chunks */
7155 if (flags & BTRFS_BLOCK_GROUP_RAID0)
7156 return stripped;
7157
7158 /* turn mirroring into duplication */
7159 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
7160 BTRFS_BLOCK_GROUP_RAID10))
7161 return stripped | BTRFS_BLOCK_GROUP_DUP;
7162 return flags;
7163 } else {
7164 /* they already had raid on here, just return */
7165 if (flags & stripped)
7166 return flags;
7167
7168 stripped |= BTRFS_BLOCK_GROUP_DUP;
7169 stripped = flags & ~stripped;
7170
7171 /* switch duplicated blocks with raid1 */
7172 if (flags & BTRFS_BLOCK_GROUP_DUP)
7173 return stripped | BTRFS_BLOCK_GROUP_RAID1;
7174
7175 /* turn single device chunks into raid0 */
7176 return stripped | BTRFS_BLOCK_GROUP_RAID0;
7177 }
7178 return flags;
7179 }
7180
7181 static int __alloc_chunk_for_shrink(struct btrfs_root *root,
7182 struct btrfs_block_group_cache *shrink_block_group,
7183 int force)
7184 {
7185 struct btrfs_trans_handle *trans;
7186 u64 new_alloc_flags;
7187 u64 calc;
7188
7189 spin_lock(&shrink_block_group->lock);
7190 if (btrfs_block_group_used(&shrink_block_group->item) +
7191 shrink_block_group->reserved > 0) {
7192 spin_unlock(&shrink_block_group->lock);
7193
7194 trans = btrfs_start_transaction(root, 1);
7195 spin_lock(&shrink_block_group->lock);
7196
7197 new_alloc_flags = update_block_group_flags(root,
7198 shrink_block_group->flags);
7199 if (new_alloc_flags != shrink_block_group->flags) {
7200 calc =
7201 btrfs_block_group_used(&shrink_block_group->item);
7202 } else {
7203 calc = shrink_block_group->key.offset;
7204 }
7205 spin_unlock(&shrink_block_group->lock);
7206
7207 do_chunk_alloc(trans, root->fs_info->extent_root,
7208 calc + 2 * 1024 * 1024, new_alloc_flags, force);
7209
7210 btrfs_end_transaction(trans, root);
7211 } else
7212 spin_unlock(&shrink_block_group->lock);
7213 return 0;
7214 }
7215
7216
7217 int btrfs_prepare_block_group_relocation(struct btrfs_root *root,
7218 struct btrfs_block_group_cache *group)
7219
7220 {
7221 __alloc_chunk_for_shrink(root, group, 1);
7222 set_block_group_readonly(group);
7223 return 0;
7224 }
7225
7226 /*
7227 * checks to see if its even possible to relocate this block group.
7228 *
7229 * @return - -1 if it's not a good idea to relocate this block group, 0 if its
7230 * ok to go ahead and try.
7231 */
7232 int btrfs_can_relocate(struct btrfs_root *root, u64 bytenr)
7233 {
7234 struct btrfs_block_group_cache *block_group;
7235 struct btrfs_space_info *space_info;
7236 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
7237 struct btrfs_device *device;
7238 int full = 0;
7239 int ret = 0;
7240
7241 block_group = btrfs_lookup_block_group(root->fs_info, bytenr);
7242
7243 /* odd, couldn't find the block group, leave it alone */
7244 if (!block_group)
7245 return -1;
7246
7247 /* no bytes used, we're good */
7248 if (!btrfs_block_group_used(&block_group->item))
7249 goto out;
7250
7251 space_info = block_group->space_info;
7252 spin_lock(&space_info->lock);
7253
7254 full = space_info->full;
7255
7256 /*
7257 * if this is the last block group we have in this space, we can't
7258 * relocate it unless we're able to allocate a new chunk below.
7259 *
7260 * Otherwise, we need to make sure we have room in the space to handle
7261 * all of the extents from this block group. If we can, we're good
7262 */
7263 if ((space_info->total_bytes != block_group->key.offset) &&
7264 (space_info->bytes_used + space_info->bytes_reserved +
7265 space_info->bytes_pinned + space_info->bytes_readonly +
7266 btrfs_block_group_used(&block_group->item) <
7267 space_info->total_bytes)) {
7268 spin_unlock(&space_info->lock);
7269 goto out;
7270 }
7271 spin_unlock(&space_info->lock);
7272
7273 /*
7274 * ok we don't have enough space, but maybe we have free space on our
7275 * devices to allocate new chunks for relocation, so loop through our
7276 * alloc devices and guess if we have enough space. However, if we
7277 * were marked as full, then we know there aren't enough chunks, and we
7278 * can just return.
7279 */
7280 ret = -1;
7281 if (full)
7282 goto out;
7283
7284 mutex_lock(&root->fs_info->chunk_mutex);
7285 list_for_each_entry(device, &fs_devices->alloc_list, dev_alloc_list) {
7286 u64 min_free = btrfs_block_group_used(&block_group->item);
7287 u64 dev_offset, max_avail;
7288
7289 /*
7290 * check to make sure we can actually find a chunk with enough
7291 * space to fit our block group in.
7292 */
7293 if (device->total_bytes > device->bytes_used + min_free) {
7294 ret = find_free_dev_extent(NULL, device, min_free,
7295 &dev_offset, &max_avail);
7296 if (!ret)
7297 break;
7298 ret = -1;
7299 }
7300 }
7301 mutex_unlock(&root->fs_info->chunk_mutex);
7302 out:
7303 btrfs_put_block_group(block_group);
7304 return ret;
7305 }
7306
7307 static int find_first_block_group(struct btrfs_root *root,
7308 struct btrfs_path *path, struct btrfs_key *key)
7309 {
7310 int ret = 0;
7311 struct btrfs_key found_key;
7312 struct extent_buffer *leaf;
7313 int slot;
7314
7315 ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
7316 if (ret < 0)
7317 goto out;
7318
7319 while (1) {
7320 slot = path->slots[0];
7321 leaf = path->nodes[0];
7322 if (slot >= btrfs_header_nritems(leaf)) {
7323 ret = btrfs_next_leaf(root, path);
7324 if (ret == 0)
7325 continue;
7326 if (ret < 0)
7327 goto out;
7328 break;
7329 }
7330 btrfs_item_key_to_cpu(leaf, &found_key, slot);
7331
7332 if (found_key.objectid >= key->objectid &&
7333 found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) {
7334 ret = 0;
7335 goto out;
7336 }
7337 path->slots[0]++;
7338 }
7339 ret = -ENOENT;
7340 out:
7341 return ret;
7342 }
7343
7344 int btrfs_free_block_groups(struct btrfs_fs_info *info)
7345 {
7346 struct btrfs_block_group_cache *block_group;
7347 struct btrfs_space_info *space_info;
7348 struct btrfs_caching_control *caching_ctl;
7349 struct rb_node *n;
7350
7351 down_write(&info->extent_commit_sem);
7352 while (!list_empty(&info->caching_block_groups)) {
7353 caching_ctl = list_entry(info->caching_block_groups.next,
7354 struct btrfs_caching_control, list);
7355 list_del(&caching_ctl->list);
7356 put_caching_control(caching_ctl);
7357 }
7358 up_write(&info->extent_commit_sem);
7359
7360 spin_lock(&info->block_group_cache_lock);
7361 while ((n = rb_last(&info->block_group_cache_tree)) != NULL) {
7362 block_group = rb_entry(n, struct btrfs_block_group_cache,
7363 cache_node);
7364 rb_erase(&block_group->cache_node,
7365 &info->block_group_cache_tree);
7366 spin_unlock(&info->block_group_cache_lock);
7367
7368 down_write(&block_group->space_info->groups_sem);
7369 list_del(&block_group->list);
7370 up_write(&block_group->space_info->groups_sem);
7371
7372 if (block_group->cached == BTRFS_CACHE_STARTED)
7373 wait_block_group_cache_done(block_group);
7374
7375 btrfs_remove_free_space_cache(block_group);
7376
7377 WARN_ON(atomic_read(&block_group->count) != 1);
7378 kfree(block_group);
7379
7380 spin_lock(&info->block_group_cache_lock);
7381 }
7382 spin_unlock(&info->block_group_cache_lock);
7383
7384 /* now that all the block groups are freed, go through and
7385 * free all the space_info structs. This is only called during
7386 * the final stages of unmount, and so we know nobody is
7387 * using them. We call synchronize_rcu() once before we start,
7388 * just to be on the safe side.
7389 */
7390 synchronize_rcu();
7391
7392 while(!list_empty(&info->space_info)) {
7393 space_info = list_entry(info->space_info.next,
7394 struct btrfs_space_info,
7395 list);
7396
7397 list_del(&space_info->list);
7398 kfree(space_info);
7399 }
7400 return 0;
7401 }
7402
7403 int btrfs_read_block_groups(struct btrfs_root *root)
7404 {
7405 struct btrfs_path *path;
7406 int ret;
7407 struct btrfs_block_group_cache *cache;
7408 struct btrfs_fs_info *info = root->fs_info;
7409 struct btrfs_space_info *space_info;
7410 struct btrfs_key key;
7411 struct btrfs_key found_key;
7412 struct extent_buffer *leaf;
7413
7414 root = info->extent_root;
7415 key.objectid = 0;
7416 key.offset = 0;
7417 btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
7418 path = btrfs_alloc_path();
7419 if (!path)
7420 return -ENOMEM;
7421
7422 while (1) {
7423 ret = find_first_block_group(root, path, &key);
7424 if (ret > 0) {
7425 ret = 0;
7426 goto error;
7427 }
7428 if (ret != 0)
7429 goto error;
7430
7431 leaf = path->nodes[0];
7432 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
7433 cache = kzalloc(sizeof(*cache), GFP_NOFS);
7434 if (!cache) {
7435 ret = -ENOMEM;
7436 break;
7437 }
7438
7439 atomic_set(&cache->count, 1);
7440 spin_lock_init(&cache->lock);
7441 spin_lock_init(&cache->tree_lock);
7442 cache->fs_info = info;
7443 INIT_LIST_HEAD(&cache->list);
7444 INIT_LIST_HEAD(&cache->cluster_list);
7445
7446 /*
7447 * we only want to have 32k of ram per block group for keeping
7448 * track of free space, and if we pass 1/2 of that we want to
7449 * start converting things over to using bitmaps
7450 */
7451 cache->extents_thresh = ((1024 * 32) / 2) /
7452 sizeof(struct btrfs_free_space);
7453
7454 read_extent_buffer(leaf, &cache->item,
7455 btrfs_item_ptr_offset(leaf, path->slots[0]),
7456 sizeof(cache->item));
7457 memcpy(&cache->key, &found_key, sizeof(found_key));
7458
7459 key.objectid = found_key.objectid + found_key.offset;
7460 btrfs_release_path(root, path);
7461 cache->flags = btrfs_block_group_flags(&cache->item);
7462 cache->sectorsize = root->sectorsize;
7463
7464 /*
7465 * check for two cases, either we are full, and therefore
7466 * don't need to bother with the caching work since we won't
7467 * find any space, or we are empty, and we can just add all
7468 * the space in and be done with it. This saves us _alot_ of
7469 * time, particularly in the full case.
7470 */
7471 if (found_key.offset == btrfs_block_group_used(&cache->item)) {
7472 exclude_super_stripes(root, cache);
7473 cache->last_byte_to_unpin = (u64)-1;
7474 cache->cached = BTRFS_CACHE_FINISHED;
7475 free_excluded_extents(root, cache);
7476 } else if (btrfs_block_group_used(&cache->item) == 0) {
7477 exclude_super_stripes(root, cache);
7478 cache->last_byte_to_unpin = (u64)-1;
7479 cache->cached = BTRFS_CACHE_FINISHED;
7480 add_new_free_space(cache, root->fs_info,
7481 found_key.objectid,
7482 found_key.objectid +
7483 found_key.offset);
7484 free_excluded_extents(root, cache);
7485 }
7486
7487 ret = update_space_info(info, cache->flags, found_key.offset,
7488 btrfs_block_group_used(&cache->item),
7489 &space_info);
7490 BUG_ON(ret);
7491 cache->space_info = space_info;
7492 spin_lock(&cache->space_info->lock);
7493 cache->space_info->bytes_super += cache->bytes_super;
7494 spin_unlock(&cache->space_info->lock);
7495
7496 down_write(&space_info->groups_sem);
7497 list_add_tail(&cache->list, &space_info->block_groups);
7498 up_write(&space_info->groups_sem);
7499
7500 ret = btrfs_add_block_group_cache(root->fs_info, cache);
7501 BUG_ON(ret);
7502
7503 set_avail_alloc_bits(root->fs_info, cache->flags);
7504 if (btrfs_chunk_readonly(root, cache->key.objectid))
7505 set_block_group_readonly(cache);
7506 }
7507 ret = 0;
7508 error:
7509 btrfs_free_path(path);
7510 return ret;
7511 }
7512
7513 int btrfs_make_block_group(struct btrfs_trans_handle *trans,
7514 struct btrfs_root *root, u64 bytes_used,
7515 u64 type, u64 chunk_objectid, u64 chunk_offset,
7516 u64 size)
7517 {
7518 int ret;
7519 struct btrfs_root *extent_root;
7520 struct btrfs_block_group_cache *cache;
7521
7522 extent_root = root->fs_info->extent_root;
7523
7524 root->fs_info->last_trans_log_full_commit = trans->transid;
7525
7526 cache = kzalloc(sizeof(*cache), GFP_NOFS);
7527 if (!cache)
7528 return -ENOMEM;
7529
7530 cache->key.objectid = chunk_offset;
7531 cache->key.offset = size;
7532 cache->key.type = BTRFS_BLOCK_GROUP_ITEM_KEY;
7533 cache->sectorsize = root->sectorsize;
7534
7535 /*
7536 * we only want to have 32k of ram per block group for keeping track
7537 * of free space, and if we pass 1/2 of that we want to start
7538 * converting things over to using bitmaps
7539 */
7540 cache->extents_thresh = ((1024 * 32) / 2) /
7541 sizeof(struct btrfs_free_space);
7542 atomic_set(&cache->count, 1);
7543 spin_lock_init(&cache->lock);
7544 spin_lock_init(&cache->tree_lock);
7545 INIT_LIST_HEAD(&cache->list);
7546 INIT_LIST_HEAD(&cache->cluster_list);
7547
7548 btrfs_set_block_group_used(&cache->item, bytes_used);
7549 btrfs_set_block_group_chunk_objectid(&cache->item, chunk_objectid);
7550 cache->flags = type;
7551 btrfs_set_block_group_flags(&cache->item, type);
7552
7553 cache->last_byte_to_unpin = (u64)-1;
7554 cache->cached = BTRFS_CACHE_FINISHED;
7555 exclude_super_stripes(root, cache);
7556
7557 add_new_free_space(cache, root->fs_info, chunk_offset,
7558 chunk_offset + size);
7559
7560 free_excluded_extents(root, cache);
7561
7562 ret = update_space_info(root->fs_info, cache->flags, size, bytes_used,
7563 &cache->space_info);
7564 BUG_ON(ret);
7565
7566 spin_lock(&cache->space_info->lock);
7567 cache->space_info->bytes_super += cache->bytes_super;
7568 spin_unlock(&cache->space_info->lock);
7569
7570 down_write(&cache->space_info->groups_sem);
7571 list_add_tail(&cache->list, &cache->space_info->block_groups);
7572 up_write(&cache->space_info->groups_sem);
7573
7574 ret = btrfs_add_block_group_cache(root->fs_info, cache);
7575 BUG_ON(ret);
7576
7577 ret = btrfs_insert_item(trans, extent_root, &cache->key, &cache->item,
7578 sizeof(cache->item));
7579 BUG_ON(ret);
7580
7581 set_avail_alloc_bits(extent_root->fs_info, type);
7582
7583 return 0;
7584 }
7585
7586 int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
7587 struct btrfs_root *root, u64 group_start)
7588 {
7589 struct btrfs_path *path;
7590 struct btrfs_block_group_cache *block_group;
7591 struct btrfs_free_cluster *cluster;
7592 struct btrfs_key key;
7593 int ret;
7594
7595 root = root->fs_info->extent_root;
7596
7597 block_group = btrfs_lookup_block_group(root->fs_info, group_start);
7598 BUG_ON(!block_group);
7599 BUG_ON(!block_group->ro);
7600
7601 memcpy(&key, &block_group->key, sizeof(key));
7602
7603 /* make sure this block group isn't part of an allocation cluster */
7604 cluster = &root->fs_info->data_alloc_cluster;
7605 spin_lock(&cluster->refill_lock);
7606 btrfs_return_cluster_to_free_space(block_group, cluster);
7607 spin_unlock(&cluster->refill_lock);
7608
7609 /*
7610 * make sure this block group isn't part of a metadata
7611 * allocation cluster
7612 */
7613 cluster = &root->fs_info->meta_alloc_cluster;
7614 spin_lock(&cluster->refill_lock);
7615 btrfs_return_cluster_to_free_space(block_group, cluster);
7616 spin_unlock(&cluster->refill_lock);
7617
7618 path = btrfs_alloc_path();
7619 BUG_ON(!path);
7620
7621 spin_lock(&root->fs_info->block_group_cache_lock);
7622 rb_erase(&block_group->cache_node,
7623 &root->fs_info->block_group_cache_tree);
7624 spin_unlock(&root->fs_info->block_group_cache_lock);
7625
7626 down_write(&block_group->space_info->groups_sem);
7627 /*
7628 * we must use list_del_init so people can check to see if they
7629 * are still on the list after taking the semaphore
7630 */
7631 list_del_init(&block_group->list);
7632 up_write(&block_group->space_info->groups_sem);
7633
7634 if (block_group->cached == BTRFS_CACHE_STARTED)
7635 wait_block_group_cache_done(block_group);
7636
7637 btrfs_remove_free_space_cache(block_group);
7638
7639 spin_lock(&block_group->space_info->lock);
7640 block_group->space_info->total_bytes -= block_group->key.offset;
7641 block_group->space_info->bytes_readonly -= block_group->key.offset;
7642 spin_unlock(&block_group->space_info->lock);
7643
7644 btrfs_clear_space_info_full(root->fs_info);
7645
7646 btrfs_put_block_group(block_group);
7647 btrfs_put_block_group(block_group);
7648
7649 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
7650 if (ret > 0)
7651 ret = -EIO;
7652 if (ret < 0)
7653 goto out;
7654
7655 ret = btrfs_del_item(trans, root, path);
7656 out:
7657 btrfs_free_path(path);
7658 return ret;
7659 }
Linux with added FPC-III support