projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Btrfs: Further reduce the concurrency penalty of defrag and drop_snapshot
[deliverable/linux.git] / fs / btrfs / extent-tree.c
1 /*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19 #include <linux/sched.h>
20 #include "ctree.h"
21 #include "disk-io.h"
22 #include "print-tree.h"
23 #include "transaction.h"
24
25 static int finish_current_insert(struct btrfs_trans_handle *trans, struct
26 btrfs_root *extent_root);
27 static int del_pending_extents(struct btrfs_trans_handle *trans, struct
28 btrfs_root *extent_root);
29
30 static int cache_block_group(struct btrfs_root *root,
31 struct btrfs_block_group_cache *block_group)
32 {
33 struct btrfs_path *path;
34 int ret;
35 struct btrfs_key key;
36 struct btrfs_leaf *leaf;
37 struct radix_tree_root *extent_radix;
38 int slot;
39 u64 i;
40 u64 last = 0;
41 u64 hole_size;
42 u64 limit;
43 int found = 0;
44
45 root = root->fs_info->extent_root;
46 extent_radix = &root->fs_info->extent_map_radix;
47
48 if (block_group->cached)
49 return 0;
50 if (block_group->data)
51 return 0;
52 path = btrfs_alloc_path();
53 if (!path)
54 return -ENOMEM;
55 path->reada = 1;
56 key.objectid = block_group->key.objectid;
57 key.flags = 0;
58 key.offset = 0;
59 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
60 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
61 if (ret < 0)
62 return ret;
63 if (ret && path->slots[0] > 0)
64 path->slots[0]--;
65 limit = block_group->key.objectid + block_group->key.offset;
66 while(1) {
67 leaf = btrfs_buffer_leaf(path->nodes[0]);
68 slot = path->slots[0];
69 if (slot >= btrfs_header_nritems(&leaf->header)) {
70 ret = btrfs_next_leaf(root, path);
71 if (ret < 0)
72 goto err;
73 if (ret == 0) {
74 continue;
75 } else {
76 if (found) {
77 hole_size = block_group->key.objectid +
78 block_group->key.offset - last;
79 } else {
80 last = block_group->key.objectid;
81 hole_size = block_group->key.offset;
82 }
83 for (i = 0; i < hole_size; i++) {
84 set_radix_bit(extent_radix,
85 last + i);
86 }
87 break;
88 }
89 }
90 btrfs_disk_key_to_cpu(&key, &leaf->items[slot].key);
91 if (key.objectid >= block_group->key.objectid +
92 block_group->key.offset) {
93 if (found) {
94 hole_size = block_group->key.objectid +
95 block_group->key.offset - last;
96 } else {
97 last = block_group->key.objectid;
98 hole_size = block_group->key.offset;
99 }
100 for (i = 0; i < hole_size; i++) {
101 set_radix_bit(extent_radix, last + i);
102 }
103 break;
104 }
105 if (btrfs_key_type(&key) == BTRFS_EXTENT_ITEM_KEY) {
106 if (!found) {
107 last = key.objectid + key.offset;
108 found = 1;
109 } else {
110 hole_size = key.objectid - last;
111 for (i = 0; i < hole_size; i++) {
112 set_radix_bit(extent_radix, last + i);
113 }
114 last = key.objectid + key.offset;
115 }
116 }
117 path->slots[0]++;
118 }
119
120 block_group->cached = 1;
121 err:
122 btrfs_free_path(path);
123 return 0;
124 }
125
126 struct btrfs_block_group_cache *btrfs_lookup_block_group(struct
127 btrfs_fs_info *info,
128 u64 blocknr)
129 {
130 struct btrfs_block_group_cache *block_group;
131 int ret;
132
133 ret = radix_tree_gang_lookup(&info->block_group_radix,
134 (void **)&block_group,
135 blocknr, 1);
136 if (ret) {
137 if (block_group->key.objectid <= blocknr && blocknr <=
138 block_group->key.objectid + block_group->key.offset)
139 return block_group;
140 }
141 ret = radix_tree_gang_lookup(&info->block_group_data_radix,
142 (void **)&block_group,
143 blocknr, 1);
144 if (ret) {
145 if (block_group->key.objectid <= blocknr && blocknr <=
146 block_group->key.objectid + block_group->key.offset)
147 return block_group;
148 }
149 return NULL;
150 }
151
152 static u64 leaf_range(struct btrfs_root *root)
153 {
154 u64 size = BTRFS_LEAF_DATA_SIZE(root);
155 do_div(size, sizeof(struct btrfs_extent_item) +
156 sizeof(struct btrfs_item));
157 return size;
158 }
159
160 static u64 find_search_start(struct btrfs_root *root,
161 struct btrfs_block_group_cache **cache_ret,
162 u64 search_start, int num)
163 {
164 unsigned long gang[8];
165 int ret;
166 struct btrfs_block_group_cache *cache = *cache_ret;
167 u64 last = max(search_start, cache->key.objectid);
168
169 if (cache->data)
170 goto out;
171 again:
172 ret = cache_block_group(root, cache);
173 if (ret)
174 goto out;
175 while(1) {
176 ret = find_first_radix_bit(&root->fs_info->extent_map_radix,
177 gang, last, ARRAY_SIZE(gang));
178 if (!ret)
179 goto out;
180 last = gang[ret-1] + 1;
181 if (num > 1) {
182 if (ret != ARRAY_SIZE(gang)) {
183 goto new_group;
184 }
185 if (gang[ret-1] - gang[0] > leaf_range(root)) {
186 continue;
187 }
188 }
189 if (gang[0] >= cache->key.objectid + cache->key.offset) {
190 goto new_group;
191 }
192 return gang[0];
193 }
194 out:
195 return max(cache->last_alloc, search_start);
196
197 new_group:
198 cache = btrfs_lookup_block_group(root->fs_info,
199 last + cache->key.offset - 1);
200 if (!cache) {
201 return max((*cache_ret)->last_alloc, search_start);
202 }
203 cache = btrfs_find_block_group(root, cache,
204 last + cache->key.offset - 1, 0, 0);
205 *cache_ret = cache;
206 goto again;
207 }
208
209 static u64 div_factor(u64 num, int factor)
210 {
211 num *= factor;
212 do_div(num, 10);
213 return num;
214 }
215
216 struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root,
217 struct btrfs_block_group_cache
218 *hint, u64 search_start,
219 int data, int owner)
220 {
221 struct btrfs_block_group_cache *cache[8];
222 struct btrfs_block_group_cache *found_group = NULL;
223 struct btrfs_fs_info *info = root->fs_info;
224 struct radix_tree_root *radix;
225 struct radix_tree_root *swap_radix;
226 u64 used;
227 u64 last = 0;
228 u64 hint_last;
229 int i;
230 int ret;
231 int full_search = 0;
232 int factor = 8;
233 int data_swap = 0;
234
235 if (!owner)
236 factor = 5;
237
238 if (data) {
239 radix = &info->block_group_data_radix;
240 swap_radix = &info->block_group_radix;
241 } else {
242 radix = &info->block_group_radix;
243 swap_radix = &info->block_group_data_radix;
244 }
245
246 if (search_start) {
247 struct btrfs_block_group_cache *shint;
248 shint = btrfs_lookup_block_group(info, search_start);
249 if (shint->data == data) {
250 used = btrfs_block_group_used(&shint->item);
251 if (used + shint->pinned <
252 div_factor(shint->key.offset, factor)) {
253 return shint;
254 }
255 }
256 }
257 if (hint && hint->data == data) {
258 used = btrfs_block_group_used(&hint->item);
259 if (used + hint->pinned <
260 div_factor(hint->key.offset, factor)) {
261 return hint;
262 }
263 if (used >= div_factor(hint->key.offset, 8)) {
264 radix_tree_tag_clear(radix,
265 hint->key.objectid +
266 hint->key.offset - 1,
267 BTRFS_BLOCK_GROUP_AVAIL);
268 }
269 last = hint->key.offset * 3;
270 if (hint->key.objectid >= last)
271 last = max(search_start + hint->key.offset - 1,
272 hint->key.objectid - last);
273 else
274 last = hint->key.objectid + hint->key.offset;
275 hint_last = last;
276 } else {
277 if (hint)
278 hint_last = max(hint->key.objectid, search_start);
279 else
280 hint_last = search_start;
281
282 last = hint_last;
283 }
284 while(1) {
285 ret = radix_tree_gang_lookup_tag(radix, (void **)cache,
286 last, ARRAY_SIZE(cache),
287 BTRFS_BLOCK_GROUP_AVAIL);
288 if (!ret)
289 break;
290 for (i = 0; i < ret; i++) {
291 last = cache[i]->key.objectid +
292 cache[i]->key.offset;
293 used = btrfs_block_group_used(&cache[i]->item);
294 if (used + cache[i]->pinned <
295 div_factor(cache[i]->key.offset, factor)) {
296 found_group = cache[i];
297 goto found;
298 }
299 if (used >= div_factor(cache[i]->key.offset, 8)) {
300 radix_tree_tag_clear(radix,
301 cache[i]->key.objectid +
302 cache[i]->key.offset - 1,
303 BTRFS_BLOCK_GROUP_AVAIL);
304 }
305 }
306 cond_resched();
307 }
308 last = hint_last;
309 again:
310 while(1) {
311 ret = radix_tree_gang_lookup(radix, (void **)cache,
312 last, ARRAY_SIZE(cache));
313 if (!ret)
314 break;
315 for (i = 0; i < ret; i++) {
316 last = cache[i]->key.objectid +
317 cache[i]->key.offset;
318 used = btrfs_block_group_used(&cache[i]->item);
319 if (used + cache[i]->pinned < cache[i]->key.offset) {
320 found_group = cache[i];
321 goto found;
322 }
323 if (used >= cache[i]->key.offset) {
324 radix_tree_tag_clear(radix,
325 cache[i]->key.objectid +
326 cache[i]->key.offset - 1,
327 BTRFS_BLOCK_GROUP_AVAIL);
328 }
329 }
330 cond_resched();
331 }
332 if (!full_search) {
333 last = search_start;
334 full_search = 1;
335 goto again;
336 }
337 if (!data_swap) {
338 struct radix_tree_root *tmp = radix;
339 data_swap = 1;
340 radix = swap_radix;
341 swap_radix = tmp;
342 last = search_start;
343 goto again;
344 }
345 if (!found_group) {
346 ret = radix_tree_gang_lookup(radix,
347 (void **)&found_group, 0, 1);
348 if (ret == 0) {
349 ret = radix_tree_gang_lookup(swap_radix,
350 (void **)&found_group,
351 0, 1);
352 }
353 BUG_ON(ret != 1);
354 }
355 found:
356 return found_group;
357 }
358
359 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
360 struct btrfs_root *root,
361 u64 blocknr, u64 num_blocks)
362 {
363 struct btrfs_path *path;
364 int ret;
365 struct btrfs_key key;
366 struct btrfs_leaf *l;
367 struct btrfs_extent_item *item;
368 u32 refs;
369
370 path = btrfs_alloc_path();
371 if (!path)
372 return -ENOMEM;
373
374 key.objectid = blocknr;
375 key.flags = 0;
376 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
377 key.offset = num_blocks;
378 ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
379 0, 1);
380 if (ret < 0)
381 return ret;
382 if (ret != 0) {
383 BUG();
384 }
385 BUG_ON(ret != 0);
386 l = btrfs_buffer_leaf(path->nodes[0]);
387 item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
388 refs = btrfs_extent_refs(item);
389 btrfs_set_extent_refs(item, refs + 1);
390 btrfs_mark_buffer_dirty(path->nodes[0]);
391
392 btrfs_release_path(root->fs_info->extent_root, path);
393 btrfs_free_path(path);
394 finish_current_insert(trans, root->fs_info->extent_root);
395 del_pending_extents(trans, root->fs_info->extent_root);
396 return 0;
397 }
398
399 static int lookup_extent_ref(struct btrfs_trans_handle *trans,
400 struct btrfs_root *root, u64 blocknr,
401 u64 num_blocks, u32 *refs)
402 {
403 struct btrfs_path *path;
404 int ret;
405 struct btrfs_key key;
406 struct btrfs_leaf *l;
407 struct btrfs_extent_item *item;
408
409 path = btrfs_alloc_path();
410 key.objectid = blocknr;
411 key.offset = num_blocks;
412 key.flags = 0;
413 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
414 ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
415 0, 0);
416 if (ret < 0)
417 goto out;
418 if (ret != 0)
419 BUG();
420 l = btrfs_buffer_leaf(path->nodes[0]);
421 item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
422 *refs = btrfs_extent_refs(item);
423 out:
424 btrfs_free_path(path);
425 return 0;
426 }
427
428 int btrfs_inc_root_ref(struct btrfs_trans_handle *trans,
429 struct btrfs_root *root)
430 {
431 return btrfs_inc_extent_ref(trans, root, bh_blocknr(root->node), 1);
432 }
433
434 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
435 struct buffer_head *buf)
436 {
437 u64 blocknr;
438 struct btrfs_node *buf_node;
439 struct btrfs_leaf *buf_leaf;
440 struct btrfs_disk_key *key;
441 struct btrfs_file_extent_item *fi;
442 int i;
443 int leaf;
444 int ret;
445 int faili;
446 int err;
447
448 if (!root->ref_cows)
449 return 0;
450 buf_node = btrfs_buffer_node(buf);
451 leaf = btrfs_is_leaf(buf_node);
452 buf_leaf = btrfs_buffer_leaf(buf);
453 for (i = 0; i < btrfs_header_nritems(&buf_node->header); i++) {
454 if (leaf) {
455 u64 disk_blocknr;
456 key = &buf_leaf->items[i].key;
457 if (btrfs_disk_key_type(key) != BTRFS_EXTENT_DATA_KEY)
458 continue;
459 fi = btrfs_item_ptr(buf_leaf, i,
460 struct btrfs_file_extent_item);
461 if (btrfs_file_extent_type(fi) ==
462 BTRFS_FILE_EXTENT_INLINE)
463 continue;
464 disk_blocknr = btrfs_file_extent_disk_blocknr(fi);
465 if (disk_blocknr == 0)
466 continue;
467 ret = btrfs_inc_extent_ref(trans, root, disk_blocknr,
468 btrfs_file_extent_disk_num_blocks(fi));
469 if (ret) {
470 faili = i;
471 goto fail;
472 }
473 } else {
474 blocknr = btrfs_node_blockptr(buf_node, i);
475 ret = btrfs_inc_extent_ref(trans, root, blocknr, 1);
476 if (ret) {
477 faili = i;
478 goto fail;
479 }
480 }
481 }
482 return 0;
483 fail:
484 WARN_ON(1);
485 for (i =0; i < faili; i++) {
486 if (leaf) {
487 u64 disk_blocknr;
488 key = &buf_leaf->items[i].key;
489 if (btrfs_disk_key_type(key) != BTRFS_EXTENT_DATA_KEY)
490 continue;
491 fi = btrfs_item_ptr(buf_leaf, i,
492 struct btrfs_file_extent_item);
493 if (btrfs_file_extent_type(fi) ==
494 BTRFS_FILE_EXTENT_INLINE)
495 continue;
496 disk_blocknr = btrfs_file_extent_disk_blocknr(fi);
497 if (disk_blocknr == 0)
498 continue;
499 err = btrfs_free_extent(trans, root, disk_blocknr,
500 btrfs_file_extent_disk_num_blocks(fi), 0);
501 BUG_ON(err);
502 } else {
503 blocknr = btrfs_node_blockptr(buf_node, i);
504 err = btrfs_free_extent(trans, root, blocknr, 1, 0);
505 BUG_ON(err);
506 }
507 }
508 return ret;
509 }
510
511 static int write_one_cache_group(struct btrfs_trans_handle *trans,
512 struct btrfs_root *root,
513 struct btrfs_path *path,
514 struct btrfs_block_group_cache *cache)
515 {
516 int ret;
517 int pending_ret;
518 struct btrfs_root *extent_root = root->fs_info->extent_root;
519 struct btrfs_block_group_item *bi;
520
521 ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1);
522 if (ret < 0)
523 goto fail;
524 BUG_ON(ret);
525 bi = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
526 struct btrfs_block_group_item);
527 memcpy(bi, &cache->item, sizeof(*bi));
528 btrfs_mark_buffer_dirty(path->nodes[0]);
529 btrfs_release_path(extent_root, path);
530 fail:
531 finish_current_insert(trans, extent_root);
532 pending_ret = del_pending_extents(trans, extent_root);
533 if (ret)
534 return ret;
535 if (pending_ret)
536 return pending_ret;
537 if (cache->data)
538 cache->last_alloc = cache->first_free;
539 return 0;
540
541 }
542
543 static int write_dirty_block_radix(struct btrfs_trans_handle *trans,
544 struct btrfs_root *root,
545 struct radix_tree_root *radix)
546 {
547 struct btrfs_block_group_cache *cache[8];
548 int ret;
549 int err = 0;
550 int werr = 0;
551 int i;
552 struct btrfs_path *path;
553 unsigned long off = 0;
554
555 path = btrfs_alloc_path();
556 if (!path)
557 return -ENOMEM;
558
559 while(1) {
560 ret = radix_tree_gang_lookup_tag(radix, (void **)cache,
561 off, ARRAY_SIZE(cache),
562 BTRFS_BLOCK_GROUP_DIRTY);
563 if (!ret)
564 break;
565 for (i = 0; i < ret; i++) {
566 err = write_one_cache_group(trans, root,
567 path, cache[i]);
568 /*
569 * if we fail to write the cache group, we want
570 * to keep it marked dirty in hopes that a later
571 * write will work
572 */
573 if (err) {
574 werr = err;
575 off = cache[i]->key.objectid +
576 cache[i]->key.offset;
577 continue;
578 }
579
580 radix_tree_tag_clear(radix, cache[i]->key.objectid +
581 cache[i]->key.offset - 1,
582 BTRFS_BLOCK_GROUP_DIRTY);
583 }
584 }
585 btrfs_free_path(path);
586 return werr;
587 }
588
589 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
590 struct btrfs_root *root)
591 {
592 int ret;
593 int ret2;
594 ret = write_dirty_block_radix(trans, root,
595 &root->fs_info->block_group_radix);
596 ret2 = write_dirty_block_radix(trans, root,
597 &root->fs_info->block_group_data_radix);
598 if (ret)
599 return ret;
600 if (ret2)
601 return ret2;
602 return 0;
603 }
604
605 static int update_block_group(struct btrfs_trans_handle *trans,
606 struct btrfs_root *root,
607 u64 blocknr, u64 num, int alloc, int mark_free,
608 int data)
609 {
610 struct btrfs_block_group_cache *cache;
611 struct btrfs_fs_info *info = root->fs_info;
612 u64 total = num;
613 u64 old_val;
614 u64 block_in_group;
615 u64 i;
616 int ret;
617
618 while(total) {
619 cache = btrfs_lookup_block_group(info, blocknr);
620 if (!cache) {
621 return -1;
622 }
623 block_in_group = blocknr - cache->key.objectid;
624 WARN_ON(block_in_group > cache->key.offset);
625 radix_tree_tag_set(cache->radix, cache->key.objectid +
626 cache->key.offset - 1,
627 BTRFS_BLOCK_GROUP_DIRTY);
628
629 old_val = btrfs_block_group_used(&cache->item);
630 num = min(total, cache->key.offset - block_in_group);
631 if (alloc) {
632 if (blocknr > cache->last_alloc)
633 cache->last_alloc = blocknr;
634 if (!cache->data) {
635 for (i = 0; i < num; i++) {
636 clear_radix_bit(&info->extent_map_radix,
637 blocknr + i);
638 }
639 }
640 if (cache->data != data &&
641 old_val < (cache->key.offset >> 1)) {
642 cache->data = data;
643 radix_tree_delete(cache->radix,
644 cache->key.objectid +
645 cache->key.offset - 1);
646
647 if (data) {
648 cache->radix =
649 &info->block_group_data_radix;
650 cache->item.flags |=
651 BTRFS_BLOCK_GROUP_DATA;
652 } else {
653 cache->radix = &info->block_group_radix;
654 cache->item.flags &=
655 ~BTRFS_BLOCK_GROUP_DATA;
656 }
657 ret = radix_tree_insert(cache->radix,
658 cache->key.objectid +
659 cache->key.offset - 1,
660 (void *)cache);
661 }
662 old_val += num;
663 } else {
664 old_val -= num;
665 if (blocknr < cache->first_free)
666 cache->first_free = blocknr;
667 if (!cache->data && mark_free) {
668 for (i = 0; i < num; i++) {
669 set_radix_bit(&info->extent_map_radix,
670 blocknr + i);
671 }
672 }
673 if (old_val < (cache->key.offset >> 1) &&
674 old_val + num >= (cache->key.offset >> 1)) {
675 radix_tree_tag_set(cache->radix,
676 cache->key.objectid +
677 cache->key.offset - 1,
678 BTRFS_BLOCK_GROUP_AVAIL);
679 }
680 }
681 btrfs_set_block_group_used(&cache->item, old_val);
682 total -= num;
683 blocknr += num;
684 }
685 return 0;
686 }
687
688 int btrfs_copy_pinned(struct btrfs_root *root, struct radix_tree_root *copy)
689 {
690 unsigned long gang[8];
691 u64 last = 0;
692 struct radix_tree_root *pinned_radix = &root->fs_info->pinned_radix;
693 int ret;
694 int i;
695
696 while(1) {
697 ret = find_first_radix_bit(pinned_radix, gang, last,
698 ARRAY_SIZE(gang));
699 if (!ret)
700 break;
701 for (i = 0 ; i < ret; i++) {
702 set_radix_bit(copy, gang[i]);
703 last = gang[i] + 1;
704 }
705 }
706 ret = find_first_radix_bit(&root->fs_info->extent_ins_radix, gang, 0,
707 ARRAY_SIZE(gang));
708 WARN_ON(ret);
709 return 0;
710 }
711
712 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
713 struct btrfs_root *root,
714 struct radix_tree_root *unpin_radix)
715 {
716 unsigned long gang[8];
717 struct btrfs_block_group_cache *block_group;
718 u64 first = 0;
719 int ret;
720 int i;
721 struct radix_tree_root *pinned_radix = &root->fs_info->pinned_radix;
722 struct radix_tree_root *extent_radix = &root->fs_info->extent_map_radix;
723
724 while(1) {
725 ret = find_first_radix_bit(unpin_radix, gang, 0,
726 ARRAY_SIZE(gang));
727 if (!ret)
728 break;
729 if (!first)
730 first = gang[0];
731 for (i = 0; i < ret; i++) {
732 clear_radix_bit(pinned_radix, gang[i]);
733 clear_radix_bit(unpin_radix, gang[i]);
734 block_group = btrfs_lookup_block_group(root->fs_info,
735 gang[i]);
736 if (block_group) {
737 WARN_ON(block_group->pinned == 0);
738 block_group->pinned--;
739 if (gang[i] < block_group->last_alloc)
740 block_group->last_alloc = gang[i];
741 if (!block_group->data)
742 set_radix_bit(extent_radix, gang[i]);
743 }
744 }
745 }
746 return 0;
747 }
748
749 static int finish_current_insert(struct btrfs_trans_handle *trans, struct
750 btrfs_root *extent_root)
751 {
752 struct btrfs_key ins;
753 struct btrfs_extent_item extent_item;
754 int i;
755 int ret;
756 int err;
757 unsigned long gang[8];
758 struct btrfs_fs_info *info = extent_root->fs_info;
759
760 btrfs_set_extent_refs(&extent_item, 1);
761 ins.offset = 1;
762 ins.flags = 0;
763 btrfs_set_key_type(&ins, BTRFS_EXTENT_ITEM_KEY);
764 btrfs_set_extent_owner(&extent_item, extent_root->root_key.objectid);
765
766 while(1) {
767 ret = find_first_radix_bit(&info->extent_ins_radix, gang, 0,
768 ARRAY_SIZE(gang));
769 if (!ret)
770 break;
771
772 for (i = 0; i < ret; i++) {
773 ins.objectid = gang[i];
774 err = btrfs_insert_item(trans, extent_root, &ins,
775 &extent_item,
776 sizeof(extent_item));
777 clear_radix_bit(&info->extent_ins_radix, gang[i]);
778 WARN_ON(err);
779 }
780 }
781 return 0;
782 }
783
784 static int pin_down_block(struct btrfs_root *root, u64 blocknr, int pending)
785 {
786 int err;
787 struct btrfs_header *header;
788 struct buffer_head *bh;
789
790 if (!pending) {
791 bh = btrfs_find_tree_block(root, blocknr);
792 if (bh) {
793 if (buffer_uptodate(bh)) {
794 u64 transid =
795 root->fs_info->running_transaction->transid;
796 header = btrfs_buffer_header(bh);
797 if (btrfs_header_generation(header) ==
798 transid) {
799 btrfs_block_release(root, bh);
800 return 0;
801 }
802 }
803 btrfs_block_release(root, bh);
804 }
805 err = set_radix_bit(&root->fs_info->pinned_radix, blocknr);
806 if (!err) {
807 struct btrfs_block_group_cache *cache;
808 cache = btrfs_lookup_block_group(root->fs_info,
809 blocknr);
810 if (cache)
811 cache->pinned++;
812 }
813 } else {
814 err = set_radix_bit(&root->fs_info->pending_del_radix, blocknr);
815 }
816 BUG_ON(err < 0);
817 return 0;
818 }
819
820 /*
821 * remove an extent from the root, returns 0 on success
822 */
823 static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
824 *root, u64 blocknr, u64 num_blocks, int pin,
825 int mark_free)
826 {
827 struct btrfs_path *path;
828 struct btrfs_key key;
829 struct btrfs_fs_info *info = root->fs_info;
830 struct btrfs_root *extent_root = info->extent_root;
831 int ret;
832 struct btrfs_extent_item *ei;
833 u32 refs;
834
835 key.objectid = blocknr;
836 key.flags = 0;
837 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
838 key.offset = num_blocks;
839
840 path = btrfs_alloc_path();
841 if (!path)
842 return -ENOMEM;
843
844 ret = btrfs_search_slot(trans, extent_root, &key, path, -1, 1);
845 if (ret < 0)
846 return ret;
847 BUG_ON(ret);
848 ei = btrfs_item_ptr(btrfs_buffer_leaf(path->nodes[0]), path->slots[0],
849 struct btrfs_extent_item);
850 BUG_ON(ei->refs == 0);
851 refs = btrfs_extent_refs(ei) - 1;
852 btrfs_set_extent_refs(ei, refs);
853 btrfs_mark_buffer_dirty(path->nodes[0]);
854 if (refs == 0) {
855 u64 super_blocks_used;
856
857 if (pin) {
858 ret = pin_down_block(root, blocknr, 0);
859 BUG_ON(ret);
860 }
861
862 super_blocks_used = btrfs_super_blocks_used(&info->super_copy);
863 btrfs_set_super_blocks_used(&info->super_copy,
864 super_blocks_used - num_blocks);
865 ret = btrfs_del_item(trans, extent_root, path);
866 if (ret) {
867 return ret;
868 }
869 ret = update_block_group(trans, root, blocknr, num_blocks, 0,
870 mark_free, 0);
871 BUG_ON(ret);
872 }
873 btrfs_free_path(path);
874 finish_current_insert(trans, extent_root);
875 return ret;
876 }
877
878 /*
879 * find all the blocks marked as pending in the radix tree and remove
880 * them from the extent map
881 */
882 static int del_pending_extents(struct btrfs_trans_handle *trans, struct
883 btrfs_root *extent_root)
884 {
885 int ret;
886 int wret;
887 int err = 0;
888 unsigned long gang[4];
889 int i;
890 struct radix_tree_root *pending_radix;
891 struct radix_tree_root *pinned_radix;
892 struct btrfs_block_group_cache *cache;
893
894 pending_radix = &extent_root->fs_info->pending_del_radix;
895 pinned_radix = &extent_root->fs_info->pinned_radix;
896
897 while(1) {
898 ret = find_first_radix_bit(pending_radix, gang, 0,
899 ARRAY_SIZE(gang));
900 if (!ret)
901 break;
902 for (i = 0; i < ret; i++) {
903 wret = set_radix_bit(pinned_radix, gang[i]);
904 if (wret == 0) {
905 cache =
906 btrfs_lookup_block_group(extent_root->fs_info,
907 gang[i]);
908 if (cache)
909 cache->pinned++;
910 }
911 if (wret < 0) {
912 printk(KERN_CRIT "set_radix_bit, err %d\n",
913 wret);
914 BUG_ON(wret < 0);
915 }
916 wret = clear_radix_bit(pending_radix, gang[i]);
917 BUG_ON(wret);
918 wret = __free_extent(trans, extent_root,
919 gang[i], 1, 0, 0);
920 if (wret)
921 err = wret;
922 }
923 }
924 return err;
925 }
926
927 /*
928 * remove an extent from the root, returns 0 on success
929 */
930 int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
931 *root, u64 blocknr, u64 num_blocks, int pin)
932 {
933 struct btrfs_root *extent_root = root->fs_info->extent_root;
934 int pending_ret;
935 int ret;
936
937 if (root == extent_root) {
938 pin_down_block(root, blocknr, 1);
939 return 0;
940 }
941 ret = __free_extent(trans, root, blocknr, num_blocks, pin, pin == 0);
942 pending_ret = del_pending_extents(trans, root->fs_info->extent_root);
943 return ret ? ret : pending_ret;
944 }
945
946 /*
947 * walks the btree of allocated extents and find a hole of a given size.
948 * The key ins is changed to record the hole:
949 * ins->objectid == block start
950 * ins->flags = BTRFS_EXTENT_ITEM_KEY
951 * ins->offset == number of blocks
952 * Any available blocks before search_start are skipped.
953 */
954 static int find_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
955 *orig_root, u64 num_blocks, u64 empty_size,
956 u64 search_start, u64 search_end, u64 hint_block,
957 struct btrfs_key *ins, u64 exclude_start,
958 u64 exclude_nr, int data)
959 {
960 struct btrfs_path *path;
961 struct btrfs_key key;
962 int ret;
963 u64 hole_size = 0;
964 int slot = 0;
965 u64 last_block = 0;
966 u64 test_block;
967 u64 orig_search_start = search_start;
968 int start_found;
969 struct btrfs_leaf *l;
970 struct btrfs_root * root = orig_root->fs_info->extent_root;
971 struct btrfs_fs_info *info = root->fs_info;
972 int total_needed = num_blocks;
973 int level;
974 struct btrfs_block_group_cache *block_group;
975 int full_scan = 0;
976 int wrapped = 0;
977 u64 limit;
978
979 WARN_ON(num_blocks < 1);
980 ins->flags = 0;
981 btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
982
983 level = btrfs_header_level(btrfs_buffer_header(root->node));
984 if (search_end == (u64)-1)
985 search_end = btrfs_super_total_blocks(&info->super_copy);
986 if (hint_block) {
987 block_group = btrfs_lookup_block_group(info, hint_block);
988 block_group = btrfs_find_block_group(root, block_group,
989 hint_block, data, 1);
990 } else {
991 block_group = btrfs_find_block_group(root,
992 trans->block_group, 0,
993 data, 1);
994 }
995
996 total_needed += empty_size;
997 path = btrfs_alloc_path();
998
999 check_failed:
1000 if (!block_group->data)
1001 search_start = find_search_start(root, &block_group,
1002 search_start, total_needed);
1003 else if (!full_scan)
1004 search_start = max(block_group->last_alloc, search_start);
1005
1006 btrfs_init_path(path);
1007 ins->objectid = search_start;
1008 ins->offset = 0;
1009 start_found = 0;
1010
1011 ret = btrfs_search_slot(trans, root, ins, path, 0, 0);
1012 if (ret < 0)
1013 goto error;
1014
1015 if (path->slots[0] > 0) {
1016 path->slots[0]--;
1017 }
1018
1019 l = btrfs_buffer_leaf(path->nodes[0]);
1020 btrfs_disk_key_to_cpu(&key, &l->items[path->slots[0]].key);
1021 /*
1022 * a rare case, go back one key if we hit a block group item
1023 * instead of an extent item
1024 */
1025 if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY &&
1026 key.objectid + key.offset >= search_start) {
1027 ins->objectid = key.objectid;
1028 ins->offset = key.offset - 1;
1029 btrfs_release_path(root, path);
1030 ret = btrfs_search_slot(trans, root, ins, path, 0, 0);
1031 if (ret < 0)
1032 goto error;
1033
1034 if (path->slots[0] > 0) {
1035 path->slots[0]--;
1036 }
1037 }
1038
1039 while (1) {
1040 l = btrfs_buffer_leaf(path->nodes[0]);
1041 slot = path->slots[0];
1042 if (slot >= btrfs_header_nritems(&l->header)) {
1043 if (start_found)
1044 limit = last_block +
1045 (block_group->key.offset >> 1);
1046 else
1047 limit = search_start +
1048 (block_group->key.offset >> 1);
1049 ret = btrfs_next_leaf(root, path);
1050 if (ret == 0)
1051 continue;
1052 if (ret < 0)
1053 goto error;
1054 if (!start_found) {
1055 ins->objectid = search_start;
1056 ins->offset = search_end - search_start;
1057 start_found = 1;
1058 goto check_pending;
1059 }
1060 ins->objectid = last_block > search_start ?
1061 last_block : search_start;
1062 ins->offset = search_end - ins->objectid;
1063 goto check_pending;
1064 }
1065
1066 btrfs_disk_key_to_cpu(&key, &l->items[slot].key);
1067 if (key.objectid >= search_start && key.objectid > last_block &&
1068 start_found) {
1069 if (last_block < search_start)
1070 last_block = search_start;
1071 hole_size = key.objectid - last_block;
1072 if (hole_size >= num_blocks) {
1073 ins->objectid = last_block;
1074 ins->offset = hole_size;
1075 goto check_pending;
1076 }
1077 }
1078
1079 if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY)
1080 goto next;
1081
1082 start_found = 1;
1083 last_block = key.objectid + key.offset;
1084 if (!full_scan && last_block >= block_group->key.objectid +
1085 block_group->key.offset) {
1086 btrfs_release_path(root, path);
1087 search_start = block_group->key.objectid +
1088 block_group->key.offset * 2;
1089 goto new_group;
1090 }
1091 next:
1092 path->slots[0]++;
1093 cond_resched();
1094 }
1095 check_pending:
1096 /* we have to make sure we didn't find an extent that has already
1097 * been allocated by the map tree or the original allocation
1098 */
1099 btrfs_release_path(root, path);
1100 BUG_ON(ins->objectid < search_start);
1101
1102 if (ins->objectid + num_blocks >= search_end) {
1103 if (full_scan) {
1104 ret = -ENOSPC;
1105 goto error;
1106 }
1107 search_start = orig_search_start;
1108 if (wrapped) {
1109 if (!full_scan)
1110 total_needed -= empty_size;
1111 full_scan = 1;
1112 } else
1113 wrapped = 1;
1114 goto new_group;
1115 }
1116 for (test_block = ins->objectid;
1117 test_block < ins->objectid + num_blocks; test_block++) {
1118 if (test_radix_bit(&info->pinned_radix, test_block) ||
1119 test_radix_bit(&info->extent_ins_radix, test_block)) {
1120 search_start = test_block + 1;
1121 goto new_group;
1122 }
1123 }
1124 if (exclude_nr > 0 && (ins->objectid + num_blocks > exclude_start &&
1125 ins->objectid < exclude_start + exclude_nr)) {
1126 search_start = exclude_start + exclude_nr;
1127 goto new_group;
1128 }
1129 if (!data) {
1130 block_group = btrfs_lookup_block_group(info, ins->objectid);
1131 if (block_group)
1132 trans->block_group = block_group;
1133 }
1134 ins->offset = num_blocks;
1135 btrfs_free_path(path);
1136 return 0;
1137
1138 new_group:
1139 if (search_start + num_blocks >= search_end) {
1140 search_start = orig_search_start;
1141 if (full_scan) {
1142 ret = -ENOSPC;
1143 goto error;
1144 }
1145 if (wrapped) {
1146 if (!full_scan)
1147 total_needed -= empty_size;
1148 full_scan = 1;
1149 } else
1150 wrapped = 1;
1151 }
1152 block_group = btrfs_lookup_block_group(info, search_start);
1153 cond_resched();
1154 if (!full_scan)
1155 block_group = btrfs_find_block_group(root, block_group,
1156 search_start, data, 0);
1157 goto check_failed;
1158
1159 error:
1160 btrfs_release_path(root, path);
1161 btrfs_free_path(path);
1162 return ret;
1163 }
1164 /*
1165 * finds a free extent and does all the dirty work required for allocation
1166 * returns the key for the extent through ins, and a tree buffer for
1167 * the first block of the extent through buf.
1168 *
1169 * returns 0 if everything worked, non-zero otherwise.
1170 */
1171 int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
1172 struct btrfs_root *root, u64 owner,
1173 u64 num_blocks, u64 empty_size, u64 hint_block,
1174 u64 search_end, struct btrfs_key *ins, int data)
1175 {
1176 int ret;
1177 int pending_ret;
1178 u64 super_blocks_used;
1179 u64 search_start = 0;
1180 struct btrfs_fs_info *info = root->fs_info;
1181 struct btrfs_root *extent_root = info->extent_root;
1182 struct btrfs_extent_item extent_item;
1183
1184 btrfs_set_extent_refs(&extent_item, 1);
1185 btrfs_set_extent_owner(&extent_item, owner);
1186
1187 WARN_ON(num_blocks < 1);
1188 ret = find_free_extent(trans, root, num_blocks, empty_size,
1189 search_start, search_end, hint_block, ins,
1190 trans->alloc_exclude_start,
1191 trans->alloc_exclude_nr, data);
1192 BUG_ON(ret);
1193 if (ret)
1194 return ret;
1195
1196 super_blocks_used = btrfs_super_blocks_used(&info->super_copy);
1197 btrfs_set_super_blocks_used(&info->super_copy, super_blocks_used +
1198 num_blocks);
1199
1200 if (root == extent_root) {
1201 BUG_ON(num_blocks != 1);
1202 set_radix_bit(&root->fs_info->extent_ins_radix, ins->objectid);
1203 goto update_block;
1204 }
1205
1206 WARN_ON(trans->alloc_exclude_nr);
1207 trans->alloc_exclude_start = ins->objectid;
1208 trans->alloc_exclude_nr = ins->offset;
1209 ret = btrfs_insert_item(trans, extent_root, ins, &extent_item,
1210 sizeof(extent_item));
1211
1212 trans->alloc_exclude_start = 0;
1213 trans->alloc_exclude_nr = 0;
1214
1215 BUG_ON(ret);
1216 finish_current_insert(trans, extent_root);
1217 pending_ret = del_pending_extents(trans, extent_root);
1218 if (ret) {
1219 return ret;
1220 }
1221 if (pending_ret) {
1222 return pending_ret;
1223 }
1224
1225 update_block:
1226 ret = update_block_group(trans, root, ins->objectid, ins->offset, 1, 0,
1227 data);
1228 BUG_ON(ret);
1229 return 0;
1230 }
1231
1232 /*
1233 * helper function to allocate a block for a given tree
1234 * returns the tree buffer or NULL.
1235 */
1236 struct buffer_head *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
1237 struct btrfs_root *root, u64 hint,
1238 u64 empty_size)
1239 {
1240 struct btrfs_key ins;
1241 int ret;
1242 struct buffer_head *buf;
1243
1244 ret = btrfs_alloc_extent(trans, root, root->root_key.objectid,
1245 1, empty_size, hint,
1246 (unsigned long)-1, &ins, 0);
1247 if (ret) {
1248 BUG_ON(ret > 0);
1249 return ERR_PTR(ret);
1250 }
1251 buf = btrfs_find_create_tree_block(root, ins.objectid);
1252 if (!buf) {
1253 btrfs_free_extent(trans, root, ins.objectid, 1, 0);
1254 return ERR_PTR(-ENOMEM);
1255 }
1256 WARN_ON(buffer_dirty(buf));
1257 set_buffer_uptodate(buf);
1258 set_buffer_checked(buf);
1259 set_radix_bit(&trans->transaction->dirty_pages, buf->b_page->index);
1260 return buf;
1261 }
1262
1263 static int drop_leaf_ref(struct btrfs_trans_handle *trans,
1264 struct btrfs_root *root, struct buffer_head *cur)
1265 {
1266 struct btrfs_disk_key *key;
1267 struct btrfs_leaf *leaf;
1268 struct btrfs_file_extent_item *fi;
1269 int i;
1270 int nritems;
1271 int ret;
1272
1273 BUG_ON(!btrfs_is_leaf(btrfs_buffer_node(cur)));
1274 leaf = btrfs_buffer_leaf(cur);
1275 nritems = btrfs_header_nritems(&leaf->header);
1276 for (i = 0; i < nritems; i++) {
1277 u64 disk_blocknr;
1278 key = &leaf->items[i].key;
1279 if (btrfs_disk_key_type(key) != BTRFS_EXTENT_DATA_KEY)
1280 continue;
1281 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
1282 if (btrfs_file_extent_type(fi) == BTRFS_FILE_EXTENT_INLINE)
1283 continue;
1284 /*
1285 * FIXME make sure to insert a trans record that
1286 * repeats the snapshot del on crash
1287 */
1288 disk_blocknr = btrfs_file_extent_disk_blocknr(fi);
1289 if (disk_blocknr == 0)
1290 continue;
1291 ret = btrfs_free_extent(trans, root, disk_blocknr,
1292 btrfs_file_extent_disk_num_blocks(fi),
1293 0);
1294 BUG_ON(ret);
1295 }
1296 return 0;
1297 }
1298
1299 static void reada_walk_down(struct btrfs_root *root,
1300 struct btrfs_node *node)
1301 {
1302 int i;
1303 u32 nritems;
1304 u64 blocknr;
1305 int ret;
1306 u32 refs;
1307
1308 nritems = btrfs_header_nritems(&node->header);
1309 for (i = 0; i < nritems; i++) {
1310 blocknr = btrfs_node_blockptr(node, i);
1311 ret = lookup_extent_ref(NULL, root, blocknr, 1, &refs);
1312 BUG_ON(ret);
1313 if (refs != 1)
1314 continue;
1315 mutex_unlock(&root->fs_info->fs_mutex);
1316 ret = readahead_tree_block(root, blocknr);
1317 cond_resched();
1318 mutex_lock(&root->fs_info->fs_mutex);
1319 if (ret)
1320 break;
1321 }
1322 }
1323
1324 /*
1325 * helper function for drop_snapshot, this walks down the tree dropping ref
1326 * counts as it goes.
1327 */
1328 static int walk_down_tree(struct btrfs_trans_handle *trans, struct btrfs_root
1329 *root, struct btrfs_path *path, int *level)
1330 {
1331 struct buffer_head *next;
1332 struct buffer_head *cur;
1333 u64 blocknr;
1334 int ret;
1335 u32 refs;
1336
1337 WARN_ON(*level < 0);
1338 WARN_ON(*level >= BTRFS_MAX_LEVEL);
1339 ret = lookup_extent_ref(trans, root, bh_blocknr(path->nodes[*level]),
1340 1, &refs);
1341 BUG_ON(ret);
1342 if (refs > 1)
1343 goto out;
1344
1345 /*
1346 * walk down to the last node level and free all the leaves
1347 */
1348 while(*level >= 0) {
1349 WARN_ON(*level < 0);
1350 WARN_ON(*level >= BTRFS_MAX_LEVEL);
1351 cur = path->nodes[*level];
1352
1353 if (*level > 0 && path->slots[*level] == 0)
1354 reada_walk_down(root, btrfs_buffer_node(cur));
1355
1356 if (btrfs_header_level(btrfs_buffer_header(cur)) != *level)
1357 WARN_ON(1);
1358
1359 if (path->slots[*level] >=
1360 btrfs_header_nritems(btrfs_buffer_header(cur)))
1361 break;
1362 if (*level == 0) {
1363 ret = drop_leaf_ref(trans, root, cur);
1364 BUG_ON(ret);
1365 break;
1366 }
1367 blocknr = btrfs_node_blockptr(btrfs_buffer_node(cur),
1368 path->slots[*level]);
1369 ret = lookup_extent_ref(trans, root, blocknr, 1, &refs);
1370 BUG_ON(ret);
1371 if (refs != 1) {
1372 path->slots[*level]++;
1373 ret = btrfs_free_extent(trans, root, blocknr, 1, 1);
1374 BUG_ON(ret);
1375 continue;
1376 }
1377 next = read_tree_block(root, blocknr);
1378 WARN_ON(*level <= 0);
1379 if (path->nodes[*level-1])
1380 btrfs_block_release(root, path->nodes[*level-1]);
1381 path->nodes[*level-1] = next;
1382 *level = btrfs_header_level(btrfs_buffer_header(next));
1383 path->slots[*level] = 0;
1384 }
1385 out:
1386 WARN_ON(*level < 0);
1387 WARN_ON(*level >= BTRFS_MAX_LEVEL);
1388 ret = btrfs_free_extent(trans, root,
1389 bh_blocknr(path->nodes[*level]), 1, 1);
1390 btrfs_block_release(root, path->nodes[*level]);
1391 path->nodes[*level] = NULL;
1392 *level += 1;
1393 BUG_ON(ret);
1394 return 0;
1395 }
1396
1397 /*
1398 * helper for dropping snapshots. This walks back up the tree in the path
1399 * to find the first node higher up where we haven't yet gone through
1400 * all the slots
1401 */
1402 static int walk_up_tree(struct btrfs_trans_handle *trans, struct btrfs_root
1403 *root, struct btrfs_path *path, int *level)
1404 {
1405 int i;
1406 int slot;
1407 int ret;
1408 struct btrfs_root_item *root_item = &root->root_item;
1409
1410 for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) {
1411 slot = path->slots[i];
1412 if (slot < btrfs_header_nritems(
1413 btrfs_buffer_header(path->nodes[i])) - 1) {
1414 struct btrfs_node *node;
1415 node = btrfs_buffer_node(path->nodes[i]);
1416 path->slots[i]++;
1417 *level = i;
1418 WARN_ON(*level == 0);
1419 memcpy(&root_item->drop_progress,
1420 &node->ptrs[path->slots[i]].key,
1421 sizeof(root_item->drop_progress));
1422 root_item->drop_level = i;
1423 return 0;
1424 } else {
1425 ret = btrfs_free_extent(trans, root,
1426 bh_blocknr(path->nodes[*level]),
1427 1, 1);
1428 BUG_ON(ret);
1429 btrfs_block_release(root, path->nodes[*level]);
1430 path->nodes[*level] = NULL;
1431 *level = i + 1;
1432 }
1433 }
1434 return 1;
1435 }
1436
1437 /*
1438 * drop the reference count on the tree rooted at 'snap'. This traverses
1439 * the tree freeing any blocks that have a ref count of zero after being
1440 * decremented.
1441 */
1442 int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
1443 *root)
1444 {
1445 int ret = 0;
1446 int wret;
1447 int level;
1448 struct btrfs_path *path;
1449 int i;
1450 int orig_level;
1451 struct btrfs_root_item *root_item = &root->root_item;
1452
1453 path = btrfs_alloc_path();
1454 BUG_ON(!path);
1455
1456 level = btrfs_header_level(btrfs_buffer_header(root->node));
1457 orig_level = level;
1458 if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
1459 path->nodes[level] = root->node;
1460 path->slots[level] = 0;
1461 } else {
1462 struct btrfs_key key;
1463 struct btrfs_disk_key *found_key;
1464 struct btrfs_node *node;
1465
1466 btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
1467 level = root_item->drop_level;
1468 path->lowest_level = level;
1469 wret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1470 if (wret < 0) {
1471 ret = wret;
1472 goto out;
1473 }
1474 node = btrfs_buffer_node(path->nodes[level]);
1475 found_key = &node->ptrs[path->slots[level]].key;
1476 WARN_ON(memcmp(found_key, &root_item->drop_progress,
1477 sizeof(*found_key)));
1478 }
1479 while(1) {
1480 wret = walk_down_tree(trans, root, path, &level);
1481 if (wret > 0)
1482 break;
1483 if (wret < 0)
1484 ret = wret;
1485
1486 wret = walk_up_tree(trans, root, path, &level);
1487 if (wret > 0)
1488 break;
1489 if (wret < 0)
1490 ret = wret;
1491 ret = -EAGAIN;
1492 get_bh(root->node);
1493 break;
1494 }
1495 for (i = 0; i <= orig_level; i++) {
1496 if (path->nodes[i]) {
1497 btrfs_block_release(root, path->nodes[i]);
1498 path->nodes[i] = 0;
1499 }
1500 }
1501 out:
1502 btrfs_free_path(path);
1503 return ret;
1504 }
1505
1506 static int free_block_group_radix(struct radix_tree_root *radix)
1507 {
1508 int ret;
1509 struct btrfs_block_group_cache *cache[8];
1510 int i;
1511
1512 while(1) {
1513 ret = radix_tree_gang_lookup(radix, (void **)cache, 0,
1514 ARRAY_SIZE(cache));
1515 if (!ret)
1516 break;
1517 for (i = 0; i < ret; i++) {
1518 radix_tree_delete(radix, cache[i]->key.objectid +
1519 cache[i]->key.offset - 1);
1520 kfree(cache[i]);
1521 }
1522 }
1523 return 0;
1524 }
1525
1526 int btrfs_free_block_groups(struct btrfs_fs_info *info)
1527 {
1528 int ret;
1529 int ret2;
1530 unsigned long gang[16];
1531 int i;
1532
1533 ret = free_block_group_radix(&info->block_group_radix);
1534 ret2 = free_block_group_radix(&info->block_group_data_radix);
1535 if (ret)
1536 return ret;
1537 if (ret2)
1538 return ret2;
1539
1540 while(1) {
1541 ret = find_first_radix_bit(&info->extent_map_radix,
1542 gang, 0, ARRAY_SIZE(gang));
1543 if (!ret)
1544 break;
1545 for (i = 0; i < ret; i++) {
1546 clear_radix_bit(&info->extent_map_radix, gang[i]);
1547 }
1548 }
1549 return 0;
1550 }
1551
1552 int btrfs_read_block_groups(struct btrfs_root *root)
1553 {
1554 struct btrfs_path *path;
1555 int ret;
1556 int err = 0;
1557 struct btrfs_block_group_item *bi;
1558 struct btrfs_block_group_cache *cache;
1559 struct btrfs_fs_info *info = root->fs_info;
1560 struct radix_tree_root *radix;
1561 struct btrfs_key key;
1562 struct btrfs_key found_key;
1563 struct btrfs_leaf *leaf;
1564 u64 group_size_blocks;
1565 u64 used;
1566
1567 group_size_blocks = BTRFS_BLOCK_GROUP_SIZE >>
1568 root->fs_info->sb->s_blocksize_bits;
1569 root = info->extent_root;
1570 key.objectid = 0;
1571 key.offset = group_size_blocks;
1572 key.flags = 0;
1573 btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
1574
1575 path = btrfs_alloc_path();
1576 if (!path)
1577 return -ENOMEM;
1578
1579 while(1) {
1580 ret = btrfs_search_slot(NULL, info->extent_root,
1581 &key, path, 0, 0);
1582 if (ret != 0) {
1583 err = ret;
1584 break;
1585 }
1586 leaf = btrfs_buffer_leaf(path->nodes[0]);
1587 btrfs_disk_key_to_cpu(&found_key,
1588 &leaf->items[path->slots[0]].key);
1589 cache = kmalloc(sizeof(*cache), GFP_NOFS);
1590 if (!cache) {
1591 err = -1;
1592 break;
1593 }
1594
1595 bi = btrfs_item_ptr(leaf, path->slots[0],
1596 struct btrfs_block_group_item);
1597 if (bi->flags & BTRFS_BLOCK_GROUP_DATA) {
1598 radix = &info->block_group_data_radix;
1599 cache->data = 1;
1600 } else {
1601 radix = &info->block_group_radix;
1602 cache->data = 0;
1603 }
1604
1605 memcpy(&cache->item, bi, sizeof(*bi));
1606 memcpy(&cache->key, &found_key, sizeof(found_key));
1607 cache->last_alloc = cache->key.objectid;
1608 cache->first_free = cache->key.objectid;
1609 cache->pinned = 0;
1610 cache->cached = 0;
1611
1612 cache->radix = radix;
1613
1614 key.objectid = found_key.objectid + found_key.offset;
1615 btrfs_release_path(root, path);
1616 ret = radix_tree_insert(radix, found_key.objectid +
1617 found_key.offset - 1,
1618 (void *)cache);
1619 BUG_ON(ret);
1620 used = btrfs_block_group_used(bi);
1621 if (used < div_factor(key.offset, 8)) {
1622 radix_tree_tag_set(radix, found_key.objectid +
1623 found_key.offset - 1,
1624 BTRFS_BLOCK_GROUP_AVAIL);
1625 }
1626 if (key.objectid >=
1627 btrfs_super_total_blocks(&info->super_copy))
1628 break;
1629 }
1630
1631 btrfs_free_path(path);
1632 return 0;
1633 }
Linux kernel - Deliverables
This page took 0.062326 seconds and 6 git commands to generate.