projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Btrfs: Add locking around volume management (device add/remove/balance)
[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 #include <linux/sched.h>
19 #include <linux/pagemap.h>
20 #include <linux/writeback.h>
21 #include "hash.h"
22 #include "crc32c.h"
23 #include "ctree.h"
24 #include "disk-io.h"
25 #include "print-tree.h"
26 #include "transaction.h"
27 #include "volumes.h"
28 #include "locking.h"
29
30 #define BLOCK_GROUP_DATA EXTENT_WRITEBACK
31 #define BLOCK_GROUP_METADATA EXTENT_UPTODATE
32 #define BLOCK_GROUP_SYSTEM EXTENT_NEW
33
34 #define BLOCK_GROUP_DIRTY EXTENT_DIRTY
35
36 static int finish_current_insert(struct btrfs_trans_handle *trans, struct
37 btrfs_root *extent_root);
38 static int del_pending_extents(struct btrfs_trans_handle *trans, struct
39 btrfs_root *extent_root);
40 static struct btrfs_block_group_cache *
41 __btrfs_find_block_group(struct btrfs_root *root,
42 struct btrfs_block_group_cache *hint,
43 u64 search_start, int data, int owner);
44
45 void maybe_lock_mutex(struct btrfs_root *root)
46 {
47 if (root != root->fs_info->extent_root &&
48 root != root->fs_info->chunk_root &&
49 root != root->fs_info->dev_root) {
50 mutex_lock(&root->fs_info->alloc_mutex);
51 }
52 }
53
54 void maybe_unlock_mutex(struct btrfs_root *root)
55 {
56 if (root != root->fs_info->extent_root &&
57 root != root->fs_info->chunk_root &&
58 root != root->fs_info->dev_root) {
59 mutex_unlock(&root->fs_info->alloc_mutex);
60 }
61 }
62
63 static int cache_block_group(struct btrfs_root *root,
64 struct btrfs_block_group_cache *block_group)
65 {
66 struct btrfs_path *path;
67 int ret;
68 struct btrfs_key key;
69 struct extent_buffer *leaf;
70 struct extent_io_tree *free_space_cache;
71 int slot;
72 u64 last = 0;
73 u64 hole_size;
74 u64 first_free;
75 int found = 0;
76
77 if (!block_group)
78 return 0;
79
80 root = root->fs_info->extent_root;
81 free_space_cache = &root->fs_info->free_space_cache;
82
83 if (block_group->cached)
84 return 0;
85
86 path = btrfs_alloc_path();
87 if (!path)
88 return -ENOMEM;
89
90 path->reada = 2;
91 /*
92 * we get into deadlocks with paths held by callers of this function.
93 * since the alloc_mutex is protecting things right now, just
94 * skip the locking here
95 */
96 path->skip_locking = 1;
97 first_free = block_group->key.objectid;
98 key.objectid = block_group->key.objectid;
99 key.offset = 0;
100 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
101 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
102 if (ret < 0)
103 return ret;
104 ret = btrfs_previous_item(root, path, 0, BTRFS_EXTENT_ITEM_KEY);
105 if (ret < 0)
106 return ret;
107 if (ret == 0) {
108 leaf = path->nodes[0];
109 btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
110 if (key.objectid + key.offset > first_free)
111 first_free = key.objectid + key.offset;
112 }
113 while(1) {
114 leaf = path->nodes[0];
115 slot = path->slots[0];
116 if (slot >= btrfs_header_nritems(leaf)) {
117 ret = btrfs_next_leaf(root, path);
118 if (ret < 0)
119 goto err;
120 if (ret == 0) {
121 continue;
122 } else {
123 break;
124 }
125 }
126 btrfs_item_key_to_cpu(leaf, &key, slot);
127 if (key.objectid < block_group->key.objectid) {
128 goto next;
129 }
130 if (key.objectid >= block_group->key.objectid +
131 block_group->key.offset) {
132 break;
133 }
134
135 if (btrfs_key_type(&key) == BTRFS_EXTENT_ITEM_KEY) {
136 if (!found) {
137 last = first_free;
138 found = 1;
139 }
140 if (key.objectid > last) {
141 hole_size = key.objectid - last;
142 set_extent_dirty(free_space_cache, last,
143 last + hole_size - 1,
144 GFP_NOFS);
145 }
146 last = key.objectid + key.offset;
147 }
148 next:
149 path->slots[0]++;
150 }
151
152 if (!found)
153 last = first_free;
154 if (block_group->key.objectid +
155 block_group->key.offset > last) {
156 hole_size = block_group->key.objectid +
157 block_group->key.offset - last;
158 set_extent_dirty(free_space_cache, last,
159 last + hole_size - 1, GFP_NOFS);
160 }
161 block_group->cached = 1;
162 err:
163 btrfs_free_path(path);
164 return 0;
165 }
166
167 struct btrfs_block_group_cache *btrfs_lookup_first_block_group(struct
168 btrfs_fs_info *info,
169 u64 bytenr)
170 {
171 struct extent_io_tree *block_group_cache;
172 struct btrfs_block_group_cache *block_group = NULL;
173 u64 ptr;
174 u64 start;
175 u64 end;
176 int ret;
177
178 bytenr = max_t(u64, bytenr,
179 BTRFS_SUPER_INFO_OFFSET + BTRFS_SUPER_INFO_SIZE);
180 block_group_cache = &info->block_group_cache;
181 ret = find_first_extent_bit(block_group_cache,
182 bytenr, &start, &end,
183 BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA |
184 BLOCK_GROUP_SYSTEM);
185 if (ret) {
186 return NULL;
187 }
188 ret = get_state_private(block_group_cache, start, &ptr);
189 if (ret)
190 return NULL;
191
192 block_group = (struct btrfs_block_group_cache *)(unsigned long)ptr;
193 return block_group;
194 }
195
196 struct btrfs_block_group_cache *btrfs_lookup_block_group(struct
197 btrfs_fs_info *info,
198 u64 bytenr)
199 {
200 struct extent_io_tree *block_group_cache;
201 struct btrfs_block_group_cache *block_group = NULL;
202 u64 ptr;
203 u64 start;
204 u64 end;
205 int ret;
206
207 bytenr = max_t(u64, bytenr,
208 BTRFS_SUPER_INFO_OFFSET + BTRFS_SUPER_INFO_SIZE);
209 block_group_cache = &info->block_group_cache;
210 ret = find_first_extent_bit(block_group_cache,
211 bytenr, &start, &end,
212 BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA |
213 BLOCK_GROUP_SYSTEM);
214 if (ret) {
215 return NULL;
216 }
217 ret = get_state_private(block_group_cache, start, &ptr);
218 if (ret)
219 return NULL;
220
221 block_group = (struct btrfs_block_group_cache *)(unsigned long)ptr;
222 if (block_group->key.objectid <= bytenr && bytenr <
223 block_group->key.objectid + block_group->key.offset)
224 return block_group;
225 return NULL;
226 }
227
228 static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits)
229 {
230 return (cache->flags & bits) == bits;
231 }
232
233 static int noinline find_search_start(struct btrfs_root *root,
234 struct btrfs_block_group_cache **cache_ret,
235 u64 *start_ret, u64 num, int data)
236 {
237 int ret;
238 struct btrfs_block_group_cache *cache = *cache_ret;
239 struct extent_io_tree *free_space_cache;
240 struct extent_state *state;
241 u64 last;
242 u64 start = 0;
243 u64 cache_miss = 0;
244 u64 total_fs_bytes;
245 u64 search_start = *start_ret;
246 int wrapped = 0;
247
248 WARN_ON(!mutex_is_locked(&root->fs_info->alloc_mutex));
249 total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
250 free_space_cache = &root->fs_info->free_space_cache;
251
252 if (!cache)
253 goto out;
254
255 again:
256 ret = cache_block_group(root, cache);
257 if (ret) {
258 goto out;
259 }
260
261 last = max(search_start, cache->key.objectid);
262 if (!block_group_bits(cache, data) || cache->ro)
263 goto new_group;
264
265 spin_lock_irq(&free_space_cache->lock);
266 state = find_first_extent_bit_state(free_space_cache, last, EXTENT_DIRTY);
267 while(1) {
268 if (!state) {
269 if (!cache_miss)
270 cache_miss = last;
271 spin_unlock_irq(&free_space_cache->lock);
272 goto new_group;
273 }
274
275 start = max(last, state->start);
276 last = state->end + 1;
277 if (last - start < num) {
278 do {
279 state = extent_state_next(state);
280 } while(state && !(state->state & EXTENT_DIRTY));
281 continue;
282 }
283 spin_unlock_irq(&free_space_cache->lock);
284 if (cache->ro) {
285 goto new_group;
286 }
287 if (start + num > cache->key.objectid + cache->key.offset)
288 goto new_group;
289 if (!block_group_bits(cache, data)) {
290 printk("block group bits don't match %Lu %d\n", cache->flags, data);
291 }
292 *start_ret = start;
293 return 0;
294 }
295 out:
296 cache = btrfs_lookup_block_group(root->fs_info, search_start);
297 if (!cache) {
298 printk("Unable to find block group for %Lu\n", search_start);
299 WARN_ON(1);
300 }
301 return -ENOSPC;
302
303 new_group:
304 last = cache->key.objectid + cache->key.offset;
305 wrapped:
306 cache = btrfs_lookup_first_block_group(root->fs_info, last);
307 if (!cache || cache->key.objectid >= total_fs_bytes) {
308 no_cache:
309 if (!wrapped) {
310 wrapped = 1;
311 last = search_start;
312 goto wrapped;
313 }
314 goto out;
315 }
316 if (cache_miss && !cache->cached) {
317 cache_block_group(root, cache);
318 last = cache_miss;
319 cache = btrfs_lookup_first_block_group(root->fs_info, last);
320 }
321 cache_miss = 0;
322 cache = __btrfs_find_block_group(root, cache, last, data, 0);
323 if (!cache)
324 goto no_cache;
325 *cache_ret = cache;
326 goto again;
327 }
328
329 static u64 div_factor(u64 num, int factor)
330 {
331 if (factor == 10)
332 return num;
333 num *= factor;
334 do_div(num, 10);
335 return num;
336 }
337
338 static int block_group_state_bits(u64 flags)
339 {
340 int bits = 0;
341 if (flags & BTRFS_BLOCK_GROUP_DATA)
342 bits |= BLOCK_GROUP_DATA;
343 if (flags & BTRFS_BLOCK_GROUP_METADATA)
344 bits |= BLOCK_GROUP_METADATA;
345 if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
346 bits |= BLOCK_GROUP_SYSTEM;
347 return bits;
348 }
349
350 static struct btrfs_block_group_cache *
351 __btrfs_find_block_group(struct btrfs_root *root,
352 struct btrfs_block_group_cache *hint,
353 u64 search_start, int data, int owner)
354 {
355 struct btrfs_block_group_cache *cache;
356 struct extent_io_tree *block_group_cache;
357 struct btrfs_block_group_cache *found_group = NULL;
358 struct btrfs_fs_info *info = root->fs_info;
359 u64 used;
360 u64 last = 0;
361 u64 start;
362 u64 end;
363 u64 free_check;
364 u64 ptr;
365 int bit;
366 int ret;
367 int full_search = 0;
368 int factor = 10;
369 int wrapped = 0;
370
371 block_group_cache = &info->block_group_cache;
372
373 if (data & BTRFS_BLOCK_GROUP_METADATA)
374 factor = 9;
375
376 bit = block_group_state_bits(data);
377
378 if (search_start) {
379 struct btrfs_block_group_cache *shint;
380 shint = btrfs_lookup_first_block_group(info, search_start);
381 if (shint && block_group_bits(shint, data) && !shint->ro) {
382 used = btrfs_block_group_used(&shint->item);
383 if (used + shint->pinned <
384 div_factor(shint->key.offset, factor)) {
385 return shint;
386 }
387 }
388 }
389 if (hint && !hint->ro && block_group_bits(hint, data)) {
390 used = btrfs_block_group_used(&hint->item);
391 if (used + hint->pinned <
392 div_factor(hint->key.offset, factor)) {
393 return hint;
394 }
395 last = hint->key.objectid + hint->key.offset;
396 } else {
397 if (hint)
398 last = max(hint->key.objectid, search_start);
399 else
400 last = search_start;
401 }
402 again:
403 while(1) {
404 ret = find_first_extent_bit(block_group_cache, last,
405 &start, &end, bit);
406 if (ret)
407 break;
408
409 ret = get_state_private(block_group_cache, start, &ptr);
410 if (ret) {
411 last = end + 1;
412 continue;
413 }
414
415 cache = (struct btrfs_block_group_cache *)(unsigned long)ptr;
416 last = cache->key.objectid + cache->key.offset;
417 used = btrfs_block_group_used(&cache->item);
418
419 if (!cache->ro && block_group_bits(cache, data)) {
420 free_check = div_factor(cache->key.offset, factor);
421 if (used + cache->pinned < free_check) {
422 found_group = cache;
423 goto found;
424 }
425 }
426 cond_resched();
427 }
428 if (!wrapped) {
429 last = search_start;
430 wrapped = 1;
431 goto again;
432 }
433 if (!full_search && factor < 10) {
434 last = search_start;
435 full_search = 1;
436 factor = 10;
437 goto again;
438 }
439 found:
440 return found_group;
441 }
442
443 struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root,
444 struct btrfs_block_group_cache
445 *hint, u64 search_start,
446 int data, int owner)
447 {
448
449 struct btrfs_block_group_cache *ret;
450 mutex_lock(&root->fs_info->alloc_mutex);
451 ret = __btrfs_find_block_group(root, hint, search_start, data, owner);
452 mutex_unlock(&root->fs_info->alloc_mutex);
453 return ret;
454 }
455 static u64 hash_extent_ref(u64 root_objectid, u64 ref_generation,
456 u64 owner, u64 owner_offset)
457 {
458 u32 high_crc = ~(u32)0;
459 u32 low_crc = ~(u32)0;
460 __le64 lenum;
461 lenum = cpu_to_le64(root_objectid);
462 high_crc = btrfs_crc32c(high_crc, &lenum, sizeof(lenum));
463 lenum = cpu_to_le64(ref_generation);
464 low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum));
465 if (owner >= BTRFS_FIRST_FREE_OBJECTID) {
466 lenum = cpu_to_le64(owner);
467 low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum));
468 lenum = cpu_to_le64(owner_offset);
469 low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum));
470 }
471 return ((u64)high_crc << 32) | (u64)low_crc;
472 }
473
474 static int match_extent_ref(struct extent_buffer *leaf,
475 struct btrfs_extent_ref *disk_ref,
476 struct btrfs_extent_ref *cpu_ref)
477 {
478 int ret;
479 int len;
480
481 if (cpu_ref->objectid)
482 len = sizeof(*cpu_ref);
483 else
484 len = 2 * sizeof(u64);
485 ret = memcmp_extent_buffer(leaf, cpu_ref, (unsigned long)disk_ref,
486 len);
487 return ret == 0;
488 }
489
490 static int noinline lookup_extent_backref(struct btrfs_trans_handle *trans,
491 struct btrfs_root *root,
492 struct btrfs_path *path, u64 bytenr,
493 u64 root_objectid,
494 u64 ref_generation, u64 owner,
495 u64 owner_offset, int del)
496 {
497 u64 hash;
498 struct btrfs_key key;
499 struct btrfs_key found_key;
500 struct btrfs_extent_ref ref;
501 struct extent_buffer *leaf;
502 struct btrfs_extent_ref *disk_ref;
503 int ret;
504 int ret2;
505
506 btrfs_set_stack_ref_root(&ref, root_objectid);
507 btrfs_set_stack_ref_generation(&ref, ref_generation);
508 btrfs_set_stack_ref_objectid(&ref, owner);
509 btrfs_set_stack_ref_offset(&ref, owner_offset);
510
511 hash = hash_extent_ref(root_objectid, ref_generation, owner,
512 owner_offset);
513 key.offset = hash;
514 key.objectid = bytenr;
515 key.type = BTRFS_EXTENT_REF_KEY;
516
517 while (1) {
518 ret = btrfs_search_slot(trans, root, &key, path,
519 del ? -1 : 0, del);
520 if (ret < 0)
521 goto out;
522 leaf = path->nodes[0];
523 if (ret != 0) {
524 u32 nritems = btrfs_header_nritems(leaf);
525 if (path->slots[0] >= nritems) {
526 ret2 = btrfs_next_leaf(root, path);
527 if (ret2)
528 goto out;
529 leaf = path->nodes[0];
530 }
531 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
532 if (found_key.objectid != bytenr ||
533 found_key.type != BTRFS_EXTENT_REF_KEY)
534 goto out;
535 key.offset = found_key.offset;
536 if (del) {
537 btrfs_release_path(root, path);
538 continue;
539 }
540 }
541 disk_ref = btrfs_item_ptr(path->nodes[0],
542 path->slots[0],
543 struct btrfs_extent_ref);
544 if (match_extent_ref(path->nodes[0], disk_ref, &ref)) {
545 ret = 0;
546 goto out;
547 }
548 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
549 key.offset = found_key.offset + 1;
550 btrfs_release_path(root, path);
551 }
552 out:
553 return ret;
554 }
555
556 /*
557 * Back reference rules. Back refs have three main goals:
558 *
559 * 1) differentiate between all holders of references to an extent so that
560 * when a reference is dropped we can make sure it was a valid reference
561 * before freeing the extent.
562 *
563 * 2) Provide enough information to quickly find the holders of an extent
564 * if we notice a given block is corrupted or bad.
565 *
566 * 3) Make it easy to migrate blocks for FS shrinking or storage pool
567 * maintenance. This is actually the same as #2, but with a slightly
568 * different use case.
569 *
570 * File extents can be referenced by:
571 *
572 * - multiple snapshots, subvolumes, or different generations in one subvol
573 * - different files inside a single subvolume (in theory, not implemented yet)
574 * - different offsets inside a file (bookend extents in file.c)
575 *
576 * The extent ref structure has fields for:
577 *
578 * - Objectid of the subvolume root
579 * - Generation number of the tree holding the reference
580 * - objectid of the file holding the reference
581 * - offset in the file corresponding to the key holding the reference
582 *
583 * When a file extent is allocated the fields are filled in:
584 * (root_key.objectid, trans->transid, inode objectid, offset in file)
585 *
586 * When a leaf is cow'd new references are added for every file extent found
587 * in the leaf. It looks the same as the create case, but trans->transid
588 * will be different when the block is cow'd.
589 *
590 * (root_key.objectid, trans->transid, inode objectid, offset in file)
591 *
592 * When a file extent is removed either during snapshot deletion or file
593 * truncation, the corresponding back reference is found
594 * by searching for:
595 *
596 * (btrfs_header_owner(leaf), btrfs_header_generation(leaf),
597 * inode objectid, offset in file)
598 *
599 * Btree extents can be referenced by:
600 *
601 * - Different subvolumes
602 * - Different generations of the same subvolume
603 *
604 * Storing sufficient information for a full reverse mapping of a btree
605 * block would require storing the lowest key of the block in the backref,
606 * and it would require updating that lowest key either before write out or
607 * every time it changed. Instead, the objectid of the lowest key is stored
608 * along with the level of the tree block. This provides a hint
609 * about where in the btree the block can be found. Searches through the
610 * btree only need to look for a pointer to that block, so they stop one
611 * level higher than the level recorded in the backref.
612 *
613 * Some btrees do not do reference counting on their extents. These
614 * include the extent tree and the tree of tree roots. Backrefs for these
615 * trees always have a generation of zero.
616 *
617 * When a tree block is created, back references are inserted:
618 *
619 * (root->root_key.objectid, trans->transid or zero, level, lowest_key_objectid)
620 *
621 * When a tree block is cow'd in a reference counted root,
622 * new back references are added for all the blocks it points to.
623 * These are of the form (trans->transid will have increased since creation):
624 *
625 * (root->root_key.objectid, trans->transid, level, lowest_key_objectid)
626 *
627 * Because the lowest_key_objectid and the level are just hints
628 * they are not used when backrefs are deleted. When a backref is deleted:
629 *
630 * if backref was for a tree root:
631 * root_objectid = root->root_key.objectid
632 * else
633 * root_objectid = btrfs_header_owner(parent)
634 *
635 * (root_objectid, btrfs_header_generation(parent) or zero, 0, 0)
636 *
637 * Back Reference Key hashing:
638 *
639 * Back references have four fields, each 64 bits long. Unfortunately,
640 * This is hashed into a single 64 bit number and placed into the key offset.
641 * The key objectid corresponds to the first byte in the extent, and the
642 * key type is set to BTRFS_EXTENT_REF_KEY
643 */
644 int btrfs_insert_extent_backref(struct btrfs_trans_handle *trans,
645 struct btrfs_root *root,
646 struct btrfs_path *path, u64 bytenr,
647 u64 root_objectid, u64 ref_generation,
648 u64 owner, u64 owner_offset)
649 {
650 u64 hash;
651 struct btrfs_key key;
652 struct btrfs_extent_ref ref;
653 struct btrfs_extent_ref *disk_ref;
654 int ret;
655
656 btrfs_set_stack_ref_root(&ref, root_objectid);
657 btrfs_set_stack_ref_generation(&ref, ref_generation);
658 btrfs_set_stack_ref_objectid(&ref, owner);
659 btrfs_set_stack_ref_offset(&ref, owner_offset);
660
661 hash = hash_extent_ref(root_objectid, ref_generation, owner,
662 owner_offset);
663 key.offset = hash;
664 key.objectid = bytenr;
665 key.type = BTRFS_EXTENT_REF_KEY;
666
667 ret = btrfs_insert_empty_item(trans, root, path, &key, sizeof(ref));
668 while (ret == -EEXIST) {
669 disk_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
670 struct btrfs_extent_ref);
671 if (match_extent_ref(path->nodes[0], disk_ref, &ref))
672 goto out;
673 key.offset++;
674 btrfs_release_path(root, path);
675 ret = btrfs_insert_empty_item(trans, root, path, &key,
676 sizeof(ref));
677 }
678 if (ret)
679 goto out;
680 disk_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
681 struct btrfs_extent_ref);
682 write_extent_buffer(path->nodes[0], &ref, (unsigned long)disk_ref,
683 sizeof(ref));
684 btrfs_mark_buffer_dirty(path->nodes[0]);
685 out:
686 btrfs_release_path(root, path);
687 return ret;
688 }
689
690 static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
691 struct btrfs_root *root,
692 u64 bytenr, u64 num_bytes,
693 u64 root_objectid, u64 ref_generation,
694 u64 owner, u64 owner_offset)
695 {
696 struct btrfs_path *path;
697 int ret;
698 struct btrfs_key key;
699 struct extent_buffer *l;
700 struct btrfs_extent_item *item;
701 u32 refs;
702
703 WARN_ON(num_bytes < root->sectorsize);
704 path = btrfs_alloc_path();
705 if (!path)
706 return -ENOMEM;
707
708 path->reada = 1;
709 key.objectid = bytenr;
710 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
711 key.offset = num_bytes;
712 ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
713 0, 1);
714 if (ret < 0)
715 return ret;
716 if (ret != 0) {
717 BUG();
718 }
719 BUG_ON(ret != 0);
720 l = path->nodes[0];
721 item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
722 refs = btrfs_extent_refs(l, item);
723 btrfs_set_extent_refs(l, item, refs + 1);
724 btrfs_mark_buffer_dirty(path->nodes[0]);
725
726 btrfs_release_path(root->fs_info->extent_root, path);
727
728 path->reada = 1;
729 ret = btrfs_insert_extent_backref(trans, root->fs_info->extent_root,
730 path, bytenr, root_objectid,
731 ref_generation, owner, owner_offset);
732 BUG_ON(ret);
733 finish_current_insert(trans, root->fs_info->extent_root);
734 del_pending_extents(trans, root->fs_info->extent_root);
735
736 btrfs_free_path(path);
737 return 0;
738 }
739
740 int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
741 struct btrfs_root *root,
742 u64 bytenr, u64 num_bytes,
743 u64 root_objectid, u64 ref_generation,
744 u64 owner, u64 owner_offset)
745 {
746 int ret;
747
748 mutex_lock(&root->fs_info->alloc_mutex);
749 ret = __btrfs_inc_extent_ref(trans, root, bytenr, num_bytes,
750 root_objectid, ref_generation,
751 owner, owner_offset);
752 mutex_unlock(&root->fs_info->alloc_mutex);
753 return ret;
754 }
755
756 int btrfs_extent_post_op(struct btrfs_trans_handle *trans,
757 struct btrfs_root *root)
758 {
759 finish_current_insert(trans, root->fs_info->extent_root);
760 del_pending_extents(trans, root->fs_info->extent_root);
761 return 0;
762 }
763
764 static int lookup_extent_ref(struct btrfs_trans_handle *trans,
765 struct btrfs_root *root, u64 bytenr,
766 u64 num_bytes, u32 *refs)
767 {
768 struct btrfs_path *path;
769 int ret;
770 struct btrfs_key key;
771 struct extent_buffer *l;
772 struct btrfs_extent_item *item;
773
774 WARN_ON(num_bytes < root->sectorsize);
775 path = btrfs_alloc_path();
776 path->reada = 1;
777 key.objectid = bytenr;
778 key.offset = num_bytes;
779 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
780 ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
781 0, 0);
782 if (ret < 0)
783 goto out;
784 if (ret != 0) {
785 btrfs_print_leaf(root, path->nodes[0]);
786 printk("failed to find block number %Lu\n", bytenr);
787 BUG();
788 }
789 l = path->nodes[0];
790 item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
791 *refs = btrfs_extent_refs(l, item);
792 out:
793 btrfs_free_path(path);
794 return 0;
795 }
796
797 u32 btrfs_count_snapshots_in_path(struct btrfs_root *root,
798 struct btrfs_path *count_path,
799 u64 expected_owner,
800 u64 first_extent)
801 {
802 struct btrfs_root *extent_root = root->fs_info->extent_root;
803 struct btrfs_path *path;
804 u64 bytenr;
805 u64 found_objectid;
806 u64 found_owner;
807 u64 root_objectid = root->root_key.objectid;
808 u32 total_count = 0;
809 u32 extent_refs;
810 u32 cur_count;
811 u32 nritems;
812 int ret;
813 struct btrfs_key key;
814 struct btrfs_key found_key;
815 struct extent_buffer *l;
816 struct btrfs_extent_item *item;
817 struct btrfs_extent_ref *ref_item;
818 int level = -1;
819
820 /* FIXME, needs locking */
821 BUG();
822
823 mutex_lock(&root->fs_info->alloc_mutex);
824 path = btrfs_alloc_path();
825 again:
826 if (level == -1)
827 bytenr = first_extent;
828 else
829 bytenr = count_path->nodes[level]->start;
830
831 cur_count = 0;
832 key.objectid = bytenr;
833 key.offset = 0;
834
835 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
836 ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
837 if (ret < 0)
838 goto out;
839 BUG_ON(ret == 0);
840
841 l = path->nodes[0];
842 btrfs_item_key_to_cpu(l, &found_key, path->slots[0]);
843
844 if (found_key.objectid != bytenr ||
845 found_key.type != BTRFS_EXTENT_ITEM_KEY) {
846 goto out;
847 }
848
849 item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
850 extent_refs = btrfs_extent_refs(l, item);
851 while (1) {
852 l = path->nodes[0];
853 nritems = btrfs_header_nritems(l);
854 if (path->slots[0] >= nritems) {
855 ret = btrfs_next_leaf(extent_root, path);
856 if (ret == 0)
857 continue;
858 break;
859 }
860 btrfs_item_key_to_cpu(l, &found_key, path->slots[0]);
861 if (found_key.objectid != bytenr)
862 break;
863
864 if (found_key.type != BTRFS_EXTENT_REF_KEY) {
865 path->slots[0]++;
866 continue;
867 }
868
869 cur_count++;
870 ref_item = btrfs_item_ptr(l, path->slots[0],
871 struct btrfs_extent_ref);
872 found_objectid = btrfs_ref_root(l, ref_item);
873
874 if (found_objectid != root_objectid) {
875 total_count = 2;
876 goto out;
877 }
878 if (level == -1) {
879 found_owner = btrfs_ref_objectid(l, ref_item);
880 if (found_owner != expected_owner) {
881 total_count = 2;
882 goto out;
883 }
884 /*
885 * nasty. we don't count a reference held by
886 * the running transaction. This allows nodatacow
887 * to avoid cow most of the time
888 */
889 if (found_owner >= BTRFS_FIRST_FREE_OBJECTID &&
890 btrfs_ref_generation(l, ref_item) ==
891 root->fs_info->generation) {
892 extent_refs--;
893 }
894 }
895 total_count = 1;
896 path->slots[0]++;
897 }
898 /*
899 * if there is more than one reference against a data extent,
900 * we have to assume the other ref is another snapshot
901 */
902 if (level == -1 && extent_refs > 1) {
903 total_count = 2;
904 goto out;
905 }
906 if (cur_count == 0) {
907 total_count = 0;
908 goto out;
909 }
910 if (level >= 0 && root->node == count_path->nodes[level])
911 goto out;
912 level++;
913 btrfs_release_path(root, path);
914 goto again;
915
916 out:
917 btrfs_free_path(path);
918 mutex_unlock(&root->fs_info->alloc_mutex);
919 return total_count;
920 }
921
922 int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
923 struct extent_buffer *buf)
924 {
925 u64 bytenr;
926 u32 nritems;
927 struct btrfs_key key;
928 struct btrfs_file_extent_item *fi;
929 int i;
930 int level;
931 int ret;
932 int faili;
933
934 if (!root->ref_cows)
935 return 0;
936
937 mutex_lock(&root->fs_info->alloc_mutex);
938 level = btrfs_header_level(buf);
939 nritems = btrfs_header_nritems(buf);
940 for (i = 0; i < nritems; i++) {
941 if (level == 0) {
942 u64 disk_bytenr;
943 btrfs_item_key_to_cpu(buf, &key, i);
944 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
945 continue;
946 fi = btrfs_item_ptr(buf, i,
947 struct btrfs_file_extent_item);
948 if (btrfs_file_extent_type(buf, fi) ==
949 BTRFS_FILE_EXTENT_INLINE)
950 continue;
951 disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
952 if (disk_bytenr == 0)
953 continue;
954 ret = __btrfs_inc_extent_ref(trans, root, disk_bytenr,
955 btrfs_file_extent_disk_num_bytes(buf, fi),
956 root->root_key.objectid, trans->transid,
957 key.objectid, key.offset);
958 if (ret) {
959 faili = i;
960 goto fail;
961 }
962 } else {
963 bytenr = btrfs_node_blockptr(buf, i);
964 btrfs_node_key_to_cpu(buf, &key, i);
965 ret = __btrfs_inc_extent_ref(trans, root, bytenr,
966 btrfs_level_size(root, level - 1),
967 root->root_key.objectid,
968 trans->transid,
969 level - 1, key.objectid);
970 if (ret) {
971 faili = i;
972 goto fail;
973 }
974 }
975 }
976 mutex_unlock(&root->fs_info->alloc_mutex);
977 return 0;
978 fail:
979 WARN_ON(1);
980 #if 0
981 for (i =0; i < faili; i++) {
982 if (level == 0) {
983 u64 disk_bytenr;
984 btrfs_item_key_to_cpu(buf, &key, i);
985 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
986 continue;
987 fi = btrfs_item_ptr(buf, i,
988 struct btrfs_file_extent_item);
989 if (btrfs_file_extent_type(buf, fi) ==
990 BTRFS_FILE_EXTENT_INLINE)
991 continue;
992 disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
993 if (disk_bytenr == 0)
994 continue;
995 err = btrfs_free_extent(trans, root, disk_bytenr,
996 btrfs_file_extent_disk_num_bytes(buf,
997 fi), 0);
998 BUG_ON(err);
999 } else {
1000 bytenr = btrfs_node_blockptr(buf, i);
1001 err = btrfs_free_extent(trans, root, bytenr,
1002 btrfs_level_size(root, level - 1), 0);
1003 BUG_ON(err);
1004 }
1005 }
1006 #endif
1007 mutex_unlock(&root->fs_info->alloc_mutex);
1008 return ret;
1009 }
1010
1011 static int write_one_cache_group(struct btrfs_trans_handle *trans,
1012 struct btrfs_root *root,
1013 struct btrfs_path *path,
1014 struct btrfs_block_group_cache *cache)
1015 {
1016 int ret;
1017 int pending_ret;
1018 struct btrfs_root *extent_root = root->fs_info->extent_root;
1019 unsigned long bi;
1020 struct extent_buffer *leaf;
1021
1022 ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1);
1023 if (ret < 0)
1024 goto fail;
1025 BUG_ON(ret);
1026
1027 leaf = path->nodes[0];
1028 bi = btrfs_item_ptr_offset(leaf, path->slots[0]);
1029 write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item));
1030 btrfs_mark_buffer_dirty(leaf);
1031 btrfs_release_path(extent_root, path);
1032 fail:
1033 finish_current_insert(trans, extent_root);
1034 pending_ret = del_pending_extents(trans, extent_root);
1035 if (ret)
1036 return ret;
1037 if (pending_ret)
1038 return pending_ret;
1039 return 0;
1040
1041 }
1042
1043 int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
1044 struct btrfs_root *root)
1045 {
1046 struct extent_io_tree *block_group_cache;
1047 struct btrfs_block_group_cache *cache;
1048 int ret;
1049 int err = 0;
1050 int werr = 0;
1051 struct btrfs_path *path;
1052 u64 last = 0;
1053 u64 start;
1054 u64 end;
1055 u64 ptr;
1056
1057 block_group_cache = &root->fs_info->block_group_cache;
1058 path = btrfs_alloc_path();
1059 if (!path)
1060 return -ENOMEM;
1061
1062 mutex_lock(&root->fs_info->alloc_mutex);
1063 while(1) {
1064 ret = find_first_extent_bit(block_group_cache, last,
1065 &start, &end, BLOCK_GROUP_DIRTY);
1066 if (ret)
1067 break;
1068
1069 last = end + 1;
1070 ret = get_state_private(block_group_cache, start, &ptr);
1071 if (ret)
1072 break;
1073 cache = (struct btrfs_block_group_cache *)(unsigned long)ptr;
1074 err = write_one_cache_group(trans, root,
1075 path, cache);
1076 /*
1077 * if we fail to write the cache group, we want
1078 * to keep it marked dirty in hopes that a later
1079 * write will work
1080 */
1081 if (err) {
1082 werr = err;
1083 continue;
1084 }
1085 clear_extent_bits(block_group_cache, start, end,
1086 BLOCK_GROUP_DIRTY, GFP_NOFS);
1087 }
1088 btrfs_free_path(path);
1089 mutex_unlock(&root->fs_info->alloc_mutex);
1090 return werr;
1091 }
1092
1093 static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info,
1094 u64 flags)
1095 {
1096 struct list_head *head = &info->space_info;
1097 struct list_head *cur;
1098 struct btrfs_space_info *found;
1099 list_for_each(cur, head) {
1100 found = list_entry(cur, struct btrfs_space_info, list);
1101 if (found->flags == flags)
1102 return found;
1103 }
1104 return NULL;
1105
1106 }
1107
1108 static int update_space_info(struct btrfs_fs_info *info, u64 flags,
1109 u64 total_bytes, u64 bytes_used,
1110 struct btrfs_space_info **space_info)
1111 {
1112 struct btrfs_space_info *found;
1113
1114 found = __find_space_info(info, flags);
1115 if (found) {
1116 found->total_bytes += total_bytes;
1117 found->bytes_used += bytes_used;
1118 found->full = 0;
1119 WARN_ON(found->total_bytes < found->bytes_used);
1120 *space_info = found;
1121 return 0;
1122 }
1123 found = kmalloc(sizeof(*found), GFP_NOFS);
1124 if (!found)
1125 return -ENOMEM;
1126
1127 list_add(&found->list, &info->space_info);
1128 found->flags = flags;
1129 found->total_bytes = total_bytes;
1130 found->bytes_used = bytes_used;
1131 found->bytes_pinned = 0;
1132 found->full = 0;
1133 found->force_alloc = 0;
1134 *space_info = found;
1135 return 0;
1136 }
1137
1138 static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags)
1139 {
1140 u64 extra_flags = flags & (BTRFS_BLOCK_GROUP_RAID0 |
1141 BTRFS_BLOCK_GROUP_RAID1 |
1142 BTRFS_BLOCK_GROUP_RAID10 |
1143 BTRFS_BLOCK_GROUP_DUP);
1144 if (extra_flags) {
1145 if (flags & BTRFS_BLOCK_GROUP_DATA)
1146 fs_info->avail_data_alloc_bits |= extra_flags;
1147 if (flags & BTRFS_BLOCK_GROUP_METADATA)
1148 fs_info->avail_metadata_alloc_bits |= extra_flags;
1149 if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
1150 fs_info->avail_system_alloc_bits |= extra_flags;
1151 }
1152 }
1153
1154 static u64 reduce_alloc_profile(struct btrfs_root *root, u64 flags)
1155 {
1156 u64 num_devices = root->fs_info->fs_devices->num_devices;
1157
1158 if (num_devices == 1)
1159 flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0);
1160 if (num_devices < 4)
1161 flags &= ~BTRFS_BLOCK_GROUP_RAID10;
1162
1163 if ((flags & BTRFS_BLOCK_GROUP_DUP) &&
1164 (flags & (BTRFS_BLOCK_GROUP_RAID1 |
1165 BTRFS_BLOCK_GROUP_RAID10))) {
1166 flags &= ~BTRFS_BLOCK_GROUP_DUP;
1167 }
1168
1169 if ((flags & BTRFS_BLOCK_GROUP_RAID1) &&
1170 (flags & BTRFS_BLOCK_GROUP_RAID10)) {
1171 flags &= ~BTRFS_BLOCK_GROUP_RAID1;
1172 }
1173
1174 if ((flags & BTRFS_BLOCK_GROUP_RAID0) &&
1175 ((flags & BTRFS_BLOCK_GROUP_RAID1) |
1176 (flags & BTRFS_BLOCK_GROUP_RAID10) |
1177 (flags & BTRFS_BLOCK_GROUP_DUP)))
1178 flags &= ~BTRFS_BLOCK_GROUP_RAID0;
1179 return flags;
1180 }
1181
1182 static int do_chunk_alloc(struct btrfs_trans_handle *trans,
1183 struct btrfs_root *extent_root, u64 alloc_bytes,
1184 u64 flags, int force)
1185 {
1186 struct btrfs_space_info *space_info;
1187 u64 thresh;
1188 u64 start;
1189 u64 num_bytes;
1190 int ret;
1191
1192 flags = reduce_alloc_profile(extent_root, flags);
1193
1194 space_info = __find_space_info(extent_root->fs_info, flags);
1195 if (!space_info) {
1196 ret = update_space_info(extent_root->fs_info, flags,
1197 0, 0, &space_info);
1198 BUG_ON(ret);
1199 }
1200 BUG_ON(!space_info);
1201
1202 if (space_info->force_alloc) {
1203 force = 1;
1204 space_info->force_alloc = 0;
1205 }
1206 if (space_info->full)
1207 goto out;
1208
1209 thresh = div_factor(space_info->total_bytes, 6);
1210 if (!force &&
1211 (space_info->bytes_used + space_info->bytes_pinned + alloc_bytes) <
1212 thresh)
1213 goto out;
1214
1215 mutex_lock(&extent_root->fs_info->chunk_mutex);
1216 ret = btrfs_alloc_chunk(trans, extent_root, &start, &num_bytes, flags);
1217 if (ret == -ENOSPC) {
1218 printk("space info full %Lu\n", flags);
1219 space_info->full = 1;
1220 goto out_unlock;
1221 }
1222 BUG_ON(ret);
1223
1224 ret = btrfs_make_block_group(trans, extent_root, 0, flags,
1225 BTRFS_FIRST_CHUNK_TREE_OBJECTID, start, num_bytes);
1226 BUG_ON(ret);
1227 out_unlock:
1228 mutex_unlock(&extent_root->fs_info->chunk_mutex);
1229 out:
1230 return 0;
1231 }
1232
1233 static int update_block_group(struct btrfs_trans_handle *trans,
1234 struct btrfs_root *root,
1235 u64 bytenr, u64 num_bytes, int alloc,
1236 int mark_free)
1237 {
1238 struct btrfs_block_group_cache *cache;
1239 struct btrfs_fs_info *info = root->fs_info;
1240 u64 total = num_bytes;
1241 u64 old_val;
1242 u64 byte_in_group;
1243 u64 start;
1244 u64 end;
1245
1246 WARN_ON(!mutex_is_locked(&root->fs_info->alloc_mutex));
1247 while(total) {
1248 cache = btrfs_lookup_block_group(info, bytenr);
1249 if (!cache) {
1250 return -1;
1251 }
1252 byte_in_group = bytenr - cache->key.objectid;
1253 WARN_ON(byte_in_group > cache->key.offset);
1254 start = cache->key.objectid;
1255 end = start + cache->key.offset - 1;
1256 set_extent_bits(&info->block_group_cache, start, end,
1257 BLOCK_GROUP_DIRTY, GFP_NOFS);
1258
1259 old_val = btrfs_block_group_used(&cache->item);
1260 num_bytes = min(total, cache->key.offset - byte_in_group);
1261 if (alloc) {
1262 old_val += num_bytes;
1263 cache->space_info->bytes_used += num_bytes;
1264 } else {
1265 old_val -= num_bytes;
1266 cache->space_info->bytes_used -= num_bytes;
1267 if (mark_free) {
1268 set_extent_dirty(&info->free_space_cache,
1269 bytenr, bytenr + num_bytes - 1,
1270 GFP_NOFS);
1271 }
1272 }
1273 btrfs_set_block_group_used(&cache->item, old_val);
1274 total -= num_bytes;
1275 bytenr += num_bytes;
1276 }
1277 return 0;
1278 }
1279
1280 static u64 first_logical_byte(struct btrfs_root *root, u64 search_start)
1281 {
1282 u64 start;
1283 u64 end;
1284 int ret;
1285 ret = find_first_extent_bit(&root->fs_info->block_group_cache,
1286 search_start, &start, &end,
1287 BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA |
1288 BLOCK_GROUP_SYSTEM);
1289 if (ret)
1290 return 0;
1291 return start;
1292 }
1293
1294
1295 static int update_pinned_extents(struct btrfs_root *root,
1296 u64 bytenr, u64 num, int pin)
1297 {
1298 u64 len;
1299 struct btrfs_block_group_cache *cache;
1300 struct btrfs_fs_info *fs_info = root->fs_info;
1301
1302 WARN_ON(!mutex_is_locked(&root->fs_info->alloc_mutex));
1303 if (pin) {
1304 set_extent_dirty(&fs_info->pinned_extents,
1305 bytenr, bytenr + num - 1, GFP_NOFS);
1306 } else {
1307 clear_extent_dirty(&fs_info->pinned_extents,
1308 bytenr, bytenr + num - 1, GFP_NOFS);
1309 }
1310 while (num > 0) {
1311 cache = btrfs_lookup_block_group(fs_info, bytenr);
1312 if (!cache) {
1313 u64 first = first_logical_byte(root, bytenr);
1314 WARN_ON(first < bytenr);
1315 len = min(first - bytenr, num);
1316 } else {
1317 len = min(num, cache->key.offset -
1318 (bytenr - cache->key.objectid));
1319 }
1320 if (pin) {
1321 if (cache) {
1322 cache->pinned += len;
1323 cache->space_info->bytes_pinned += len;
1324 }
1325 fs_info->total_pinned += len;
1326 } else {
1327 if (cache) {
1328 cache->pinned -= len;
1329 cache->space_info->bytes_pinned -= len;
1330 }
1331 fs_info->total_pinned -= len;
1332 }
1333 bytenr += len;
1334 num -= len;
1335 }
1336 return 0;
1337 }
1338
1339 int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy)
1340 {
1341 u64 last = 0;
1342 u64 start;
1343 u64 end;
1344 struct extent_io_tree *pinned_extents = &root->fs_info->pinned_extents;
1345 int ret;
1346
1347 while(1) {
1348 ret = find_first_extent_bit(pinned_extents, last,
1349 &start, &end, EXTENT_DIRTY);
1350 if (ret)
1351 break;
1352 set_extent_dirty(copy, start, end, GFP_NOFS);
1353 last = end + 1;
1354 }
1355 return 0;
1356 }
1357
1358 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
1359 struct btrfs_root *root,
1360 struct extent_io_tree *unpin)
1361 {
1362 u64 start;
1363 u64 end;
1364 int ret;
1365 struct extent_io_tree *free_space_cache;
1366 free_space_cache = &root->fs_info->free_space_cache;
1367
1368 mutex_lock(&root->fs_info->alloc_mutex);
1369 while(1) {
1370 ret = find_first_extent_bit(unpin, 0, &start, &end,
1371 EXTENT_DIRTY);
1372 if (ret)
1373 break;
1374 update_pinned_extents(root, start, end + 1 - start, 0);
1375 clear_extent_dirty(unpin, start, end, GFP_NOFS);
1376 set_extent_dirty(free_space_cache, start, end, GFP_NOFS);
1377 }
1378 mutex_unlock(&root->fs_info->alloc_mutex);
1379 return 0;
1380 }
1381
1382 static int finish_current_insert(struct btrfs_trans_handle *trans,
1383 struct btrfs_root *extent_root)
1384 {
1385 u64 start;
1386 u64 end;
1387 struct btrfs_fs_info *info = extent_root->fs_info;
1388 struct extent_buffer *eb;
1389 struct btrfs_path *path;
1390 struct btrfs_key ins;
1391 struct btrfs_disk_key first;
1392 struct btrfs_extent_item extent_item;
1393 int ret;
1394 int level;
1395 int err = 0;
1396
1397 WARN_ON(!mutex_is_locked(&extent_root->fs_info->alloc_mutex));
1398 btrfs_set_stack_extent_refs(&extent_item, 1);
1399 btrfs_set_key_type(&ins, BTRFS_EXTENT_ITEM_KEY);
1400 path = btrfs_alloc_path();
1401
1402 while(1) {
1403 ret = find_first_extent_bit(&info->extent_ins, 0, &start,
1404 &end, EXTENT_LOCKED);
1405 if (ret)
1406 break;
1407
1408 ins.objectid = start;
1409 ins.offset = end + 1 - start;
1410 err = btrfs_insert_item(trans, extent_root, &ins,
1411 &extent_item, sizeof(extent_item));
1412 clear_extent_bits(&info->extent_ins, start, end, EXTENT_LOCKED,
1413 GFP_NOFS);
1414 eb = read_tree_block(extent_root, ins.objectid, ins.offset,
1415 trans->transid);
1416 btrfs_tree_lock(eb);
1417 level = btrfs_header_level(eb);
1418 if (level == 0) {
1419 btrfs_item_key(eb, &first, 0);
1420 } else {
1421 btrfs_node_key(eb, &first, 0);
1422 }
1423 btrfs_tree_unlock(eb);
1424 free_extent_buffer(eb);
1425 /*
1426 * the first key is just a hint, so the race we've created
1427 * against reading it is fine
1428 */
1429 err = btrfs_insert_extent_backref(trans, extent_root, path,
1430 start, extent_root->root_key.objectid,
1431 0, level,
1432 btrfs_disk_key_objectid(&first));
1433 BUG_ON(err);
1434 }
1435 btrfs_free_path(path);
1436 return 0;
1437 }
1438
1439 static int pin_down_bytes(struct btrfs_root *root, u64 bytenr, u32 num_bytes,
1440 int pending)
1441 {
1442 int err = 0;
1443
1444 WARN_ON(!mutex_is_locked(&root->fs_info->alloc_mutex));
1445 if (!pending) {
1446 struct extent_buffer *buf;
1447 buf = btrfs_find_tree_block(root, bytenr, num_bytes);
1448 if (buf) {
1449 if (!btrfs_try_tree_lock(buf) &&
1450 btrfs_buffer_uptodate(buf, 0)) {
1451 u64 transid =
1452 root->fs_info->running_transaction->transid;
1453 u64 header_transid =
1454 btrfs_header_generation(buf);
1455 if (header_transid == transid &&
1456 !btrfs_header_flag(buf,
1457 BTRFS_HEADER_FLAG_WRITTEN)) {
1458 clean_tree_block(NULL, root, buf);
1459 btrfs_tree_unlock(buf);
1460 free_extent_buffer(buf);
1461 return 1;
1462 }
1463 btrfs_tree_unlock(buf);
1464 }
1465 free_extent_buffer(buf);
1466 }
1467 update_pinned_extents(root, bytenr, num_bytes, 1);
1468 } else {
1469 set_extent_bits(&root->fs_info->pending_del,
1470 bytenr, bytenr + num_bytes - 1,
1471 EXTENT_LOCKED, GFP_NOFS);
1472 }
1473 BUG_ON(err < 0);
1474 return 0;
1475 }
1476
1477 /*
1478 * remove an extent from the root, returns 0 on success
1479 */
1480 static int __free_extent(struct btrfs_trans_handle *trans, struct btrfs_root
1481 *root, u64 bytenr, u64 num_bytes,
1482 u64 root_objectid, u64 ref_generation,
1483 u64 owner_objectid, u64 owner_offset, int pin,
1484 int mark_free)
1485 {
1486 struct btrfs_path *path;
1487 struct btrfs_key key;
1488 struct btrfs_fs_info *info = root->fs_info;
1489 struct btrfs_root *extent_root = info->extent_root;
1490 struct extent_buffer *leaf;
1491 int ret;
1492 int extent_slot = 0;
1493 int found_extent = 0;
1494 int num_to_del = 1;
1495 struct btrfs_extent_item *ei;
1496 u32 refs;
1497
1498 WARN_ON(!mutex_is_locked(&root->fs_info->alloc_mutex));
1499 key.objectid = bytenr;
1500 btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
1501 key.offset = num_bytes;
1502 path = btrfs_alloc_path();
1503 if (!path)
1504 return -ENOMEM;
1505
1506 path->reada = 1;
1507 ret = lookup_extent_backref(trans, extent_root, path,
1508 bytenr, root_objectid,
1509 ref_generation,
1510 owner_objectid, owner_offset, 1);
1511 if (ret == 0) {
1512 struct btrfs_key found_key;
1513 extent_slot = path->slots[0];
1514 while(extent_slot > 0) {
1515 extent_slot--;
1516 btrfs_item_key_to_cpu(path->nodes[0], &found_key,
1517 extent_slot);
1518 if (found_key.objectid != bytenr)
1519 break;
1520 if (found_key.type == BTRFS_EXTENT_ITEM_KEY &&
1521 found_key.offset == num_bytes) {
1522 found_extent = 1;
1523 break;
1524 }
1525 if (path->slots[0] - extent_slot > 5)
1526 break;
1527 }
1528 if (!found_extent)
1529 ret = btrfs_del_item(trans, extent_root, path);
1530 } else {
1531 btrfs_print_leaf(extent_root, path->nodes[0]);
1532 WARN_ON(1);
1533 printk("Unable to find ref byte nr %Lu root %Lu "
1534 " gen %Lu owner %Lu offset %Lu\n", bytenr,
1535 root_objectid, ref_generation, owner_objectid,
1536 owner_offset);
1537 }
1538 if (!found_extent) {
1539 btrfs_release_path(extent_root, path);
1540 ret = btrfs_search_slot(trans, extent_root, &key, path, -1, 1);
1541 if (ret < 0)
1542 return ret;
1543 BUG_ON(ret);
1544 extent_slot = path->slots[0];
1545 }
1546
1547 leaf = path->nodes[0];
1548 ei = btrfs_item_ptr(leaf, extent_slot,
1549 struct btrfs_extent_item);
1550 refs = btrfs_extent_refs(leaf, ei);
1551 BUG_ON(refs == 0);
1552 refs -= 1;
1553 btrfs_set_extent_refs(leaf, ei, refs);
1554
1555 btrfs_mark_buffer_dirty(leaf);
1556
1557 if (refs == 0 && found_extent && path->slots[0] == extent_slot + 1) {
1558 /* if the back ref and the extent are next to each other
1559 * they get deleted below in one shot
1560 */
1561 path->slots[0] = extent_slot;
1562 num_to_del = 2;
1563 } else if (found_extent) {
1564 /* otherwise delete the extent back ref */
1565 ret = btrfs_del_item(trans, extent_root, path);
1566 BUG_ON(ret);
1567 /* if refs are 0, we need to setup the path for deletion */
1568 if (refs == 0) {
1569 btrfs_release_path(extent_root, path);
1570 ret = btrfs_search_slot(trans, extent_root, &key, path,
1571 -1, 1);
1572 if (ret < 0)
1573 return ret;
1574 BUG_ON(ret);
1575 }
1576 }
1577
1578 if (refs == 0) {
1579 u64 super_used;
1580 u64 root_used;
1581
1582 if (pin) {
1583 ret = pin_down_bytes(root, bytenr, num_bytes, 0);
1584 if (ret > 0)
1585 mark_free = 1;
1586 BUG_ON(ret < 0);
1587 }
1588
1589 /* block accounting for super block */
1590 spin_lock_irq(&info->delalloc_lock);
1591 super_used = btrfs_super_bytes_used(&info->super_copy);
1592 btrfs_set_super_bytes_used(&info->super_copy,
1593 super_used - num_bytes);
1594 spin_unlock_irq(&info->delalloc_lock);
1595
1596 /* block accounting for root item */
1597 root_used = btrfs_root_used(&root->root_item);
1598 btrfs_set_root_used(&root->root_item,
1599 root_used - num_bytes);
1600 ret = btrfs_del_items(trans, extent_root, path, path->slots[0],
1601 num_to_del);
1602 if (ret) {
1603 return ret;
1604 }
1605 ret = update_block_group(trans, root, bytenr, num_bytes, 0,
1606 mark_free);
1607 BUG_ON(ret);
1608 }
1609 btrfs_free_path(path);
1610 finish_current_insert(trans, extent_root);
1611 return ret;
1612 }
1613
1614 /*
1615 * find all the blocks marked as pending in the radix tree and remove
1616 * them from the extent map
1617 */
1618 static int del_pending_extents(struct btrfs_trans_handle *trans, struct
1619 btrfs_root *extent_root)
1620 {
1621 int ret;
1622 int err = 0;
1623 u64 start;
1624 u64 end;
1625 struct extent_io_tree *pending_del;
1626 struct extent_io_tree *pinned_extents;
1627
1628 WARN_ON(!mutex_is_locked(&extent_root->fs_info->alloc_mutex));
1629 pending_del = &extent_root->fs_info->pending_del;
1630 pinned_extents = &extent_root->fs_info->pinned_extents;
1631
1632 while(1) {
1633 ret = find_first_extent_bit(pending_del, 0, &start, &end,
1634 EXTENT_LOCKED);
1635 if (ret)
1636 break;
1637 update_pinned_extents(extent_root, start, end + 1 - start, 1);
1638 clear_extent_bits(pending_del, start, end, EXTENT_LOCKED,
1639 GFP_NOFS);
1640 ret = __free_extent(trans, extent_root,
1641 start, end + 1 - start,
1642 extent_root->root_key.objectid,
1643 0, 0, 0, 0, 0);
1644 if (ret)
1645 err = ret;
1646 }
1647 return err;
1648 }
1649
1650 /*
1651 * remove an extent from the root, returns 0 on success
1652 */
1653 static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
1654 struct btrfs_root *root, u64 bytenr,
1655 u64 num_bytes, u64 root_objectid,
1656 u64 ref_generation, u64 owner_objectid,
1657 u64 owner_offset, int pin)
1658 {
1659 struct btrfs_root *extent_root = root->fs_info->extent_root;
1660 int pending_ret;
1661 int ret;
1662
1663 WARN_ON(num_bytes < root->sectorsize);
1664 if (!root->ref_cows)
1665 ref_generation = 0;
1666
1667 if (root == extent_root) {
1668 pin_down_bytes(root, bytenr, num_bytes, 1);
1669 return 0;
1670 }
1671 ret = __free_extent(trans, root, bytenr, num_bytes, root_objectid,
1672 ref_generation, owner_objectid, owner_offset,
1673 pin, pin == 0);
1674 pending_ret = del_pending_extents(trans, root->fs_info->extent_root);
1675 return ret ? ret : pending_ret;
1676 }
1677
1678 int btrfs_free_extent(struct btrfs_trans_handle *trans,
1679 struct btrfs_root *root, u64 bytenr,
1680 u64 num_bytes, u64 root_objectid,
1681 u64 ref_generation, u64 owner_objectid,
1682 u64 owner_offset, int pin)
1683 {
1684 int ret;
1685
1686 maybe_lock_mutex(root);
1687 ret = __btrfs_free_extent(trans, root, bytenr, num_bytes,
1688 root_objectid, ref_generation,
1689 owner_objectid, owner_offset, pin);
1690 maybe_unlock_mutex(root);
1691 return ret;
1692 }
1693
1694 static u64 stripe_align(struct btrfs_root *root, u64 val)
1695 {
1696 u64 mask = ((u64)root->stripesize - 1);
1697 u64 ret = (val + mask) & ~mask;
1698 return ret;
1699 }
1700
1701 /*
1702 * walks the btree of allocated extents and find a hole of a given size.
1703 * The key ins is changed to record the hole:
1704 * ins->objectid == block start
1705 * ins->flags = BTRFS_EXTENT_ITEM_KEY
1706 * ins->offset == number of blocks
1707 * Any available blocks before search_start are skipped.
1708 */
1709 static int noinline find_free_extent(struct btrfs_trans_handle *trans,
1710 struct btrfs_root *orig_root,
1711 u64 num_bytes, u64 empty_size,
1712 u64 search_start, u64 search_end,
1713 u64 hint_byte, struct btrfs_key *ins,
1714 u64 exclude_start, u64 exclude_nr,
1715 int data)
1716 {
1717 int ret;
1718 u64 orig_search_start;
1719 struct btrfs_root * root = orig_root->fs_info->extent_root;
1720 struct btrfs_fs_info *info = root->fs_info;
1721 u64 total_needed = num_bytes;
1722 u64 *last_ptr = NULL;
1723 struct btrfs_block_group_cache *block_group;
1724 int full_scan = 0;
1725 int wrapped = 0;
1726 int chunk_alloc_done = 0;
1727 int empty_cluster = 2 * 1024 * 1024;
1728 int allowed_chunk_alloc = 0;
1729
1730 WARN_ON(num_bytes < root->sectorsize);
1731 btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
1732
1733 if (orig_root->ref_cows || empty_size)
1734 allowed_chunk_alloc = 1;
1735
1736 if (data & BTRFS_BLOCK_GROUP_METADATA) {
1737 last_ptr = &root->fs_info->last_alloc;
1738 empty_cluster = 256 * 1024;
1739 }
1740
1741 if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD)) {
1742 last_ptr = &root->fs_info->last_data_alloc;
1743 }
1744
1745 if (last_ptr) {
1746 if (*last_ptr)
1747 hint_byte = *last_ptr;
1748 else {
1749 empty_size += empty_cluster;
1750 }
1751 }
1752
1753 search_start = max(search_start, first_logical_byte(root, 0));
1754 orig_search_start = search_start;
1755
1756 if (search_end == (u64)-1)
1757 search_end = btrfs_super_total_bytes(&info->super_copy);
1758
1759 if (hint_byte) {
1760 block_group = btrfs_lookup_first_block_group(info, hint_byte);
1761 if (!block_group)
1762 hint_byte = search_start;
1763 block_group = __btrfs_find_block_group(root, block_group,
1764 hint_byte, data, 1);
1765 if (last_ptr && *last_ptr == 0 && block_group)
1766 hint_byte = block_group->key.objectid;
1767 } else {
1768 block_group = __btrfs_find_block_group(root,
1769 trans->block_group,
1770 search_start, data, 1);
1771 }
1772 search_start = max(search_start, hint_byte);
1773
1774 total_needed += empty_size;
1775
1776 check_failed:
1777 if (!block_group) {
1778 block_group = btrfs_lookup_first_block_group(info,
1779 search_start);
1780 if (!block_group)
1781 block_group = btrfs_lookup_first_block_group(info,
1782 orig_search_start);
1783 }
1784 if (full_scan && !chunk_alloc_done) {
1785 if (allowed_chunk_alloc) {
1786 do_chunk_alloc(trans, root,
1787 num_bytes + 2 * 1024 * 1024, data, 1);
1788 allowed_chunk_alloc = 0;
1789 } else if (block_group && block_group_bits(block_group, data)) {
1790 block_group->space_info->force_alloc = 1;
1791 }
1792 chunk_alloc_done = 1;
1793 }
1794 ret = find_search_start(root, &block_group, &search_start,
1795 total_needed, data);
1796 if (ret == -ENOSPC && last_ptr && *last_ptr) {
1797 *last_ptr = 0;
1798 block_group = btrfs_lookup_first_block_group(info,
1799 orig_search_start);
1800 search_start = orig_search_start;
1801 ret = find_search_start(root, &block_group, &search_start,
1802 total_needed, data);
1803 }
1804 if (ret == -ENOSPC)
1805 goto enospc;
1806 if (ret)
1807 goto error;
1808
1809 if (last_ptr && *last_ptr && search_start != *last_ptr) {
1810 *last_ptr = 0;
1811 if (!empty_size) {
1812 empty_size += empty_cluster;
1813 total_needed += empty_size;
1814 }
1815 block_group = btrfs_lookup_first_block_group(info,
1816 orig_search_start);
1817 search_start = orig_search_start;
1818 ret = find_search_start(root, &block_group,
1819 &search_start, total_needed, data);
1820 if (ret == -ENOSPC)
1821 goto enospc;
1822 if (ret)
1823 goto error;
1824 }
1825
1826 search_start = stripe_align(root, search_start);
1827 ins->objectid = search_start;
1828 ins->offset = num_bytes;
1829
1830 if (ins->objectid + num_bytes >= search_end)
1831 goto enospc;
1832
1833 if (ins->objectid + num_bytes >
1834 block_group->key.objectid + block_group->key.offset) {
1835 search_start = block_group->key.objectid +
1836 block_group->key.offset;
1837 goto new_group;
1838 }
1839
1840 if (test_range_bit(&info->extent_ins, ins->objectid,
1841 ins->objectid + num_bytes -1, EXTENT_LOCKED, 0)) {
1842 search_start = ins->objectid + num_bytes;
1843 goto new_group;
1844 }
1845
1846 if (test_range_bit(&info->pinned_extents, ins->objectid,
1847 ins->objectid + num_bytes -1, EXTENT_DIRTY, 0)) {
1848 search_start = ins->objectid + num_bytes;
1849 goto new_group;
1850 }
1851
1852 if (exclude_nr > 0 && (ins->objectid + num_bytes > exclude_start &&
1853 ins->objectid < exclude_start + exclude_nr)) {
1854 search_start = exclude_start + exclude_nr;
1855 goto new_group;
1856 }
1857
1858 if (!(data & BTRFS_BLOCK_GROUP_DATA)) {
1859 block_group = btrfs_lookup_block_group(info, ins->objectid);
1860 if (block_group)
1861 trans->block_group = block_group;
1862 }
1863 ins->offset = num_bytes;
1864 if (last_ptr) {
1865 *last_ptr = ins->objectid + ins->offset;
1866 if (*last_ptr ==
1867 btrfs_super_total_bytes(&root->fs_info->super_copy)) {
1868 *last_ptr = 0;
1869 }
1870 }
1871 return 0;
1872
1873 new_group:
1874 if (search_start + num_bytes >= search_end) {
1875 enospc:
1876 search_start = orig_search_start;
1877 if (full_scan) {
1878 ret = -ENOSPC;
1879 goto error;
1880 }
1881 if (wrapped) {
1882 if (!full_scan)
1883 total_needed -= empty_size;
1884 full_scan = 1;
1885 } else
1886 wrapped = 1;
1887 }
1888 block_group = btrfs_lookup_first_block_group(info, search_start);
1889 cond_resched();
1890 block_group = __btrfs_find_block_group(root, block_group,
1891 search_start, data, 0);
1892 goto check_failed;
1893
1894 error:
1895 return ret;
1896 }
1897
1898 /*
1899 * finds a free extent and does all the dirty work required for allocation
1900 * returns the key for the extent through ins, and a tree buffer for
1901 * the first block of the extent through buf.
1902 *
1903 * returns 0 if everything worked, non-zero otherwise.
1904 */
1905 int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
1906 struct btrfs_root *root,
1907 u64 num_bytes, u64 min_alloc_size,
1908 u64 root_objectid, u64 ref_generation,
1909 u64 owner, u64 owner_offset,
1910 u64 empty_size, u64 hint_byte,
1911 u64 search_end, struct btrfs_key *ins, u64 data)
1912 {
1913 int ret;
1914 int pending_ret;
1915 u64 super_used;
1916 u64 root_used;
1917 u64 search_start = 0;
1918 u64 alloc_profile;
1919 u32 sizes[2];
1920 struct btrfs_fs_info *info = root->fs_info;
1921 struct btrfs_root *extent_root = info->extent_root;
1922 struct btrfs_extent_item *extent_item;
1923 struct btrfs_extent_ref *ref;
1924 struct btrfs_path *path;
1925 struct btrfs_key keys[2];
1926
1927 maybe_lock_mutex(root);
1928
1929 if (data) {
1930 alloc_profile = info->avail_data_alloc_bits &
1931 info->data_alloc_profile;
1932 data = BTRFS_BLOCK_GROUP_DATA | alloc_profile;
1933 } else if (root == root->fs_info->chunk_root) {
1934 alloc_profile = info->avail_system_alloc_bits &
1935 info->system_alloc_profile;
1936 data = BTRFS_BLOCK_GROUP_SYSTEM | alloc_profile;
1937 } else {
1938 alloc_profile = info->avail_metadata_alloc_bits &
1939 info->metadata_alloc_profile;
1940 data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile;
1941 }
1942 again:
1943 data = reduce_alloc_profile(root, data);
1944 /*
1945 * the only place that sets empty_size is btrfs_realloc_node, which
1946 * is not called recursively on allocations
1947 */
1948 if (empty_size || root->ref_cows) {
1949 if (!(data & BTRFS_BLOCK_GROUP_METADATA)) {
1950 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
1951 2 * 1024 * 1024,
1952 BTRFS_BLOCK_GROUP_METADATA |
1953 (info->metadata_alloc_profile &
1954 info->avail_metadata_alloc_bits), 0);
1955 BUG_ON(ret);
1956 }
1957 ret = do_chunk_alloc(trans, root->fs_info->extent_root,
1958 num_bytes + 2 * 1024 * 1024, data, 0);
1959 BUG_ON(ret);
1960 }
1961
1962 WARN_ON(num_bytes < root->sectorsize);
1963 ret = find_free_extent(trans, root, num_bytes, empty_size,
1964 search_start, search_end, hint_byte, ins,
1965 trans->alloc_exclude_start,
1966 trans->alloc_exclude_nr, data);
1967
1968 if (ret == -ENOSPC && num_bytes > min_alloc_size) {
1969 num_bytes = num_bytes >> 1;
1970 num_bytes = max(num_bytes, min_alloc_size);
1971 do_chunk_alloc(trans, root->fs_info->extent_root,
1972 num_bytes, data, 1);
1973 goto again;
1974 }
1975 if (ret) {
1976 printk("allocation failed flags %Lu\n", data);
1977 }
1978 if (ret) {
1979 BUG();
1980 goto out;
1981 }
1982
1983 /* block accounting for super block */
1984 spin_lock_irq(&info->delalloc_lock);
1985 super_used = btrfs_super_bytes_used(&info->super_copy);
1986 btrfs_set_super_bytes_used(&info->super_copy, super_used + num_bytes);
1987 spin_unlock_irq(&info->delalloc_lock);
1988
1989 /* block accounting for root item */
1990 root_used = btrfs_root_used(&root->root_item);
1991 btrfs_set_root_used(&root->root_item, root_used + num_bytes);
1992
1993 clear_extent_dirty(&root->fs_info->free_space_cache,
1994 ins->objectid, ins->objectid + ins->offset - 1,
1995 GFP_NOFS);
1996
1997 if (root == extent_root) {
1998 set_extent_bits(&root->fs_info->extent_ins, ins->objectid,
1999 ins->objectid + ins->offset - 1,
2000 EXTENT_LOCKED, GFP_NOFS);
2001 goto update_block;
2002 }
2003
2004 WARN_ON(trans->alloc_exclude_nr);
2005 trans->alloc_exclude_start = ins->objectid;
2006 trans->alloc_exclude_nr = ins->offset;
2007
2008 memcpy(&keys[0], ins, sizeof(*ins));
2009 keys[1].offset = hash_extent_ref(root_objectid, ref_generation,
2010 owner, owner_offset);
2011 keys[1].objectid = ins->objectid;
2012 keys[1].type = BTRFS_EXTENT_REF_KEY;
2013 sizes[0] = sizeof(*extent_item);
2014 sizes[1] = sizeof(*ref);
2015
2016 path = btrfs_alloc_path();
2017 BUG_ON(!path);
2018
2019 ret = btrfs_insert_empty_items(trans, extent_root, path, keys,
2020 sizes, 2);
2021
2022 BUG_ON(ret);
2023 extent_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
2024 struct btrfs_extent_item);
2025 btrfs_set_extent_refs(path->nodes[0], extent_item, 1);
2026 ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
2027 struct btrfs_extent_ref);
2028
2029 btrfs_set_ref_root(path->nodes[0], ref, root_objectid);
2030 btrfs_set_ref_generation(path->nodes[0], ref, ref_generation);
2031 btrfs_set_ref_objectid(path->nodes[0], ref, owner);
2032 btrfs_set_ref_offset(path->nodes[0], ref, owner_offset);
2033
2034 btrfs_mark_buffer_dirty(path->nodes[0]);
2035
2036 trans->alloc_exclude_start = 0;
2037 trans->alloc_exclude_nr = 0;
2038 btrfs_free_path(path);
2039 finish_current_insert(trans, extent_root);
2040 pending_ret = del_pending_extents(trans, extent_root);
2041
2042 if (ret)
2043 goto out;
2044 if (pending_ret) {
2045 ret = pending_ret;
2046 goto out;
2047 }
2048
2049 update_block:
2050 ret = update_block_group(trans, root, ins->objectid, ins->offset, 1, 0);
2051 if (ret) {
2052 printk("update block group failed for %Lu %Lu\n",
2053 ins->objectid, ins->offset);
2054 BUG();
2055 }
2056 out:
2057 maybe_unlock_mutex(root);
2058 return ret;
2059 }
2060 /*
2061 * helper function to allocate a block for a given tree
2062 * returns the tree buffer or NULL.
2063 */
2064 struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
2065 struct btrfs_root *root,
2066 u32 blocksize,
2067 u64 root_objectid,
2068 u64 ref_generation,
2069 u64 first_objectid,
2070 int level,
2071 u64 hint,
2072 u64 empty_size)
2073 {
2074 struct btrfs_key ins;
2075 int ret;
2076 struct extent_buffer *buf;
2077
2078 ret = btrfs_alloc_extent(trans, root, blocksize, blocksize,
2079 root_objectid, ref_generation,
2080 level, first_objectid, empty_size, hint,
2081 (u64)-1, &ins, 0);
2082 if (ret) {
2083 BUG_ON(ret > 0);
2084 return ERR_PTR(ret);
2085 }
2086 buf = btrfs_find_create_tree_block(root, ins.objectid, blocksize);
2087 if (!buf) {
2088 btrfs_free_extent(trans, root, ins.objectid, blocksize,
2089 root->root_key.objectid, ref_generation,
2090 0, 0, 0);
2091 return ERR_PTR(-ENOMEM);
2092 }
2093 btrfs_set_header_generation(buf, trans->transid);
2094 btrfs_tree_lock(buf);
2095 clean_tree_block(trans, root, buf);
2096 btrfs_set_buffer_uptodate(buf);
2097
2098 if (PageDirty(buf->first_page)) {
2099 printk("page %lu dirty\n", buf->first_page->index);
2100 WARN_ON(1);
2101 }
2102
2103 set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
2104 buf->start + buf->len - 1, GFP_NOFS);
2105 trans->blocks_used++;
2106 return buf;
2107 }
2108
2109 static int noinline drop_leaf_ref(struct btrfs_trans_handle *trans,
2110 struct btrfs_root *root,
2111 struct extent_buffer *leaf)
2112 {
2113 u64 leaf_owner;
2114 u64 leaf_generation;
2115 struct btrfs_key key;
2116 struct btrfs_file_extent_item *fi;
2117 int i;
2118 int nritems;
2119 int ret;
2120
2121 BUG_ON(!btrfs_is_leaf(leaf));
2122 nritems = btrfs_header_nritems(leaf);
2123 leaf_owner = btrfs_header_owner(leaf);
2124 leaf_generation = btrfs_header_generation(leaf);
2125
2126 for (i = 0; i < nritems; i++) {
2127 u64 disk_bytenr;
2128
2129 btrfs_item_key_to_cpu(leaf, &key, i);
2130 if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
2131 continue;
2132 fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
2133 if (btrfs_file_extent_type(leaf, fi) ==
2134 BTRFS_FILE_EXTENT_INLINE)
2135 continue;
2136 /*
2137 * FIXME make sure to insert a trans record that
2138 * repeats the snapshot del on crash
2139 */
2140 disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
2141 if (disk_bytenr == 0)
2142 continue;
2143 ret = __btrfs_free_extent(trans, root, disk_bytenr,
2144 btrfs_file_extent_disk_num_bytes(leaf, fi),
2145 leaf_owner, leaf_generation,
2146 key.objectid, key.offset, 0);
2147 BUG_ON(ret);
2148 }
2149 return 0;
2150 }
2151
2152 static void noinline reada_walk_down(struct btrfs_root *root,
2153 struct extent_buffer *node,
2154 int slot)
2155 {
2156 u64 bytenr;
2157 u64 last = 0;
2158 u32 nritems;
2159 u32 refs;
2160 u32 blocksize;
2161 int ret;
2162 int i;
2163 int level;
2164 int skipped = 0;
2165
2166 nritems = btrfs_header_nritems(node);
2167 level = btrfs_header_level(node);
2168 if (level)
2169 return;
2170
2171 for (i = slot; i < nritems && skipped < 32; i++) {
2172 bytenr = btrfs_node_blockptr(node, i);
2173 if (last && ((bytenr > last && bytenr - last > 32 * 1024) ||
2174 (last > bytenr && last - bytenr > 32 * 1024))) {
2175 skipped++;
2176 continue;
2177 }
2178 blocksize = btrfs_level_size(root, level - 1);
2179 if (i != slot) {
2180 ret = lookup_extent_ref(NULL, root, bytenr,
2181 blocksize, &refs);
2182 BUG_ON(ret);
2183 if (refs != 1) {
2184 skipped++;
2185 continue;
2186 }
2187 }
2188 ret = readahead_tree_block(root, bytenr, blocksize,
2189 btrfs_node_ptr_generation(node, i));
2190 last = bytenr + blocksize;
2191 cond_resched();
2192 if (ret)
2193 break;
2194 }
2195 }
2196
2197 /*
2198 * we want to avoid as much random IO as we can with the alloc mutex
2199 * held, so drop the lock and do the lookup, then do it again with the
2200 * lock held.
2201 */
2202 int drop_snap_lookup_refcount(struct btrfs_root *root, u64 start, u64 len,
2203 u32 *refs)
2204 {
2205 mutex_unlock(&root->fs_info->alloc_mutex);
2206 lookup_extent_ref(NULL, root, start, len, refs);
2207 cond_resched();
2208 mutex_lock(&root->fs_info->alloc_mutex);
2209 return lookup_extent_ref(NULL, root, start, len, refs);
2210 }
2211
2212 /*
2213 * helper function for drop_snapshot, this walks down the tree dropping ref
2214 * counts as it goes.
2215 */
2216 static int noinline walk_down_tree(struct btrfs_trans_handle *trans,
2217 struct btrfs_root *root,
2218 struct btrfs_path *path, int *level)
2219 {
2220 u64 root_owner;
2221 u64 root_gen;
2222 u64 bytenr;
2223 u64 ptr_gen;
2224 struct extent_buffer *next;
2225 struct extent_buffer *cur;
2226 struct extent_buffer *parent;
2227 u32 blocksize;
2228 int ret;
2229 u32 refs;
2230
2231 mutex_lock(&root->fs_info->alloc_mutex);
2232
2233 WARN_ON(*level < 0);
2234 WARN_ON(*level >= BTRFS_MAX_LEVEL);
2235 ret = drop_snap_lookup_refcount(root, path->nodes[*level]->start,
2236 path->nodes[*level]->len, &refs);
2237 BUG_ON(ret);
2238 if (refs > 1)
2239 goto out;
2240
2241 /*
2242 * walk down to the last node level and free all the leaves
2243 */
2244 while(*level >= 0) {
2245 WARN_ON(*level < 0);
2246 WARN_ON(*level >= BTRFS_MAX_LEVEL);
2247 cur = path->nodes[*level];
2248
2249 if (btrfs_header_level(cur) != *level)
2250 WARN_ON(1);
2251
2252 if (path->slots[*level] >=
2253 btrfs_header_nritems(cur))
2254 break;
2255 if (*level == 0) {
2256 ret = drop_leaf_ref(trans, root, cur);
2257 BUG_ON(ret);
2258 break;
2259 }
2260 bytenr = btrfs_node_blockptr(cur, path->slots[*level]);
2261 ptr_gen = btrfs_node_ptr_generation(cur, path->slots[*level]);
2262 blocksize = btrfs_level_size(root, *level - 1);
2263
2264 ret = drop_snap_lookup_refcount(root, bytenr, blocksize, &refs);
2265 BUG_ON(ret);
2266 if (refs != 1) {
2267 parent = path->nodes[*level];
2268 root_owner = btrfs_header_owner(parent);
2269 root_gen = btrfs_header_generation(parent);
2270 path->slots[*level]++;
2271 ret = __btrfs_free_extent(trans, root, bytenr,
2272 blocksize, root_owner,
2273 root_gen, 0, 0, 1);
2274 BUG_ON(ret);
2275 continue;
2276 }
2277 next = btrfs_find_tree_block(root, bytenr, blocksize);
2278 if (!next || !btrfs_buffer_uptodate(next, ptr_gen)) {
2279 free_extent_buffer(next);
2280 mutex_unlock(&root->fs_info->alloc_mutex);
2281
2282 if (path->slots[*level] == 0)
2283 reada_walk_down(root, cur, path->slots[*level]);
2284
2285 next = read_tree_block(root, bytenr, blocksize,
2286 ptr_gen);
2287 cond_resched();
2288 mutex_lock(&root->fs_info->alloc_mutex);
2289
2290 /* we've dropped the lock, double check */
2291 ret = drop_snap_lookup_refcount(root, bytenr,
2292 blocksize, &refs);
2293 BUG_ON(ret);
2294 if (refs != 1) {
2295 parent = path->nodes[*level];
2296 root_owner = btrfs_header_owner(parent);
2297 root_gen = btrfs_header_generation(parent);
2298
2299 path->slots[*level]++;
2300 free_extent_buffer(next);
2301 ret = __btrfs_free_extent(trans, root, bytenr,
2302 blocksize,
2303 root_owner,
2304 root_gen, 0, 0, 1);
2305 BUG_ON(ret);
2306 continue;
2307 }
2308 }
2309 WARN_ON(*level <= 0);
2310 if (path->nodes[*level-1])
2311 free_extent_buffer(path->nodes[*level-1]);
2312 path->nodes[*level-1] = next;
2313 *level = btrfs_header_level(next);
2314 path->slots[*level] = 0;
2315 }
2316 out:
2317 WARN_ON(*level < 0);
2318 WARN_ON(*level >= BTRFS_MAX_LEVEL);
2319
2320 if (path->nodes[*level] == root->node) {
2321 root_owner = root->root_key.objectid;
2322 parent = path->nodes[*level];
2323 } else {
2324 parent = path->nodes[*level + 1];
2325 root_owner = btrfs_header_owner(parent);
2326 }
2327
2328 root_gen = btrfs_header_generation(parent);
2329 ret = __btrfs_free_extent(trans, root, path->nodes[*level]->start,
2330 path->nodes[*level]->len,
2331 root_owner, root_gen, 0, 0, 1);
2332 free_extent_buffer(path->nodes[*level]);
2333 path->nodes[*level] = NULL;
2334 *level += 1;
2335 BUG_ON(ret);
2336 mutex_unlock(&root->fs_info->alloc_mutex);
2337 cond_resched();
2338 return 0;
2339 }
2340
2341 /*
2342 * helper for dropping snapshots. This walks back up the tree in the path
2343 * to find the first node higher up where we haven't yet gone through
2344 * all the slots
2345 */
2346 static int noinline walk_up_tree(struct btrfs_trans_handle *trans,
2347 struct btrfs_root *root,
2348 struct btrfs_path *path, int *level)
2349 {
2350 u64 root_owner;
2351 u64 root_gen;
2352 struct btrfs_root_item *root_item = &root->root_item;
2353 int i;
2354 int slot;
2355 int ret;
2356
2357 for(i = *level; i < BTRFS_MAX_LEVEL - 1 && path->nodes[i]; i++) {
2358 slot = path->slots[i];
2359 if (slot < btrfs_header_nritems(path->nodes[i]) - 1) {
2360 struct extent_buffer *node;
2361 struct btrfs_disk_key disk_key;
2362 node = path->nodes[i];
2363 path->slots[i]++;
2364 *level = i;
2365 WARN_ON(*level == 0);
2366 btrfs_node_key(node, &disk_key, path->slots[i]);
2367 memcpy(&root_item->drop_progress,
2368 &disk_key, sizeof(disk_key));
2369 root_item->drop_level = i;
2370 return 0;
2371 } else {
2372 if (path->nodes[*level] == root->node) {
2373 root_owner = root->root_key.objectid;
2374 root_gen =
2375 btrfs_header_generation(path->nodes[*level]);
2376 } else {
2377 struct extent_buffer *node;
2378 node = path->nodes[*level + 1];
2379 root_owner = btrfs_header_owner(node);
2380 root_gen = btrfs_header_generation(node);
2381 }
2382 ret = btrfs_free_extent(trans, root,
2383 path->nodes[*level]->start,
2384 path->nodes[*level]->len,
2385 root_owner, root_gen, 0, 0, 1);
2386 BUG_ON(ret);
2387 free_extent_buffer(path->nodes[*level]);
2388 path->nodes[*level] = NULL;
2389 *level = i + 1;
2390 }
2391 }
2392 return 1;
2393 }
2394
2395 /*
2396 * drop the reference count on the tree rooted at 'snap'. This traverses
2397 * the tree freeing any blocks that have a ref count of zero after being
2398 * decremented.
2399 */
2400 int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
2401 *root)
2402 {
2403 int ret = 0;
2404 int wret;
2405 int level;
2406 struct btrfs_path *path;
2407 int i;
2408 int orig_level;
2409 struct btrfs_root_item *root_item = &root->root_item;
2410
2411 WARN_ON(!mutex_is_locked(&root->fs_info->drop_mutex));
2412 path = btrfs_alloc_path();
2413 BUG_ON(!path);
2414
2415 level = btrfs_header_level(root->node);
2416 orig_level = level;
2417 if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
2418 path->nodes[level] = root->node;
2419 extent_buffer_get(root->node);
2420 path->slots[level] = 0;
2421 } else {
2422 struct btrfs_key key;
2423 struct btrfs_disk_key found_key;
2424 struct extent_buffer *node;
2425
2426 btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
2427 level = root_item->drop_level;
2428 path->lowest_level = level;
2429 wret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2430 if (wret < 0) {
2431 ret = wret;
2432 goto out;
2433 }
2434 node = path->nodes[level];
2435 btrfs_node_key(node, &found_key, path->slots[level]);
2436 WARN_ON(memcmp(&found_key, &root_item->drop_progress,
2437 sizeof(found_key)));
2438 /*
2439 * unlock our path, this is safe because only this
2440 * function is allowed to delete this snapshot
2441 */
2442 for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
2443 if (path->nodes[i] && path->locks[i]) {
2444 path->locks[i] = 0;
2445 btrfs_tree_unlock(path->nodes[i]);
2446 }
2447 }
2448 }
2449 while(1) {
2450 wret = walk_down_tree(trans, root, path, &level);
2451 if (wret > 0)
2452 break;
2453 if (wret < 0)
2454 ret = wret;
2455
2456 wret = walk_up_tree(trans, root, path, &level);
2457 if (wret > 0)
2458 break;
2459 if (wret < 0)
2460 ret = wret;
2461 if (trans->transaction->in_commit) {
2462 ret = -EAGAIN;
2463 break;
2464 }
2465 }
2466 for (i = 0; i <= orig_level; i++) {
2467 if (path->nodes[i]) {
2468 free_extent_buffer(path->nodes[i]);
2469 path->nodes[i] = NULL;
2470 }
2471 }
2472 out:
2473 btrfs_free_path(path);
2474 return ret;
2475 }
2476
2477 int btrfs_free_block_groups(struct btrfs_fs_info *info)
2478 {
2479 u64 start;
2480 u64 end;
2481 u64 ptr;
2482 int ret;
2483
2484 mutex_lock(&info->alloc_mutex);
2485 while(1) {
2486 ret = find_first_extent_bit(&info->block_group_cache, 0,
2487 &start, &end, (unsigned int)-1);
2488 if (ret)
2489 break;
2490 ret = get_state_private(&info->block_group_cache, start, &ptr);
2491 if (!ret)
2492 kfree((void *)(unsigned long)ptr);
2493 clear_extent_bits(&info->block_group_cache, start,
2494 end, (unsigned int)-1, GFP_NOFS);
2495 }
2496 while(1) {
2497 ret = find_first_extent_bit(&info->free_space_cache, 0,
2498 &start, &end, EXTENT_DIRTY);
2499 if (ret)
2500 break;
2501 clear_extent_dirty(&info->free_space_cache, start,
2502 end, GFP_NOFS);
2503 }
2504 mutex_unlock(&info->alloc_mutex);
2505 return 0;
2506 }
2507
2508 static unsigned long calc_ra(unsigned long start, unsigned long last,
2509 unsigned long nr)
2510 {
2511 return min(last, start + nr - 1);
2512 }
2513
2514 static int noinline relocate_inode_pages(struct inode *inode, u64 start,
2515 u64 len)
2516 {
2517 u64 page_start;
2518 u64 page_end;
2519 unsigned long last_index;
2520 unsigned long i;
2521 struct page *page;
2522 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2523 struct file_ra_state *ra;
2524 unsigned long total_read = 0;
2525 unsigned long ra_pages;
2526 struct btrfs_trans_handle *trans;
2527
2528 ra = kzalloc(sizeof(*ra), GFP_NOFS);
2529
2530 mutex_lock(&inode->i_mutex);
2531 i = start >> PAGE_CACHE_SHIFT;
2532 last_index = (start + len - 1) >> PAGE_CACHE_SHIFT;
2533
2534 ra_pages = BTRFS_I(inode)->root->fs_info->bdi.ra_pages;
2535
2536 file_ra_state_init(ra, inode->i_mapping);
2537
2538 for (; i <= last_index; i++) {
2539 if (total_read % ra_pages == 0) {
2540 btrfs_force_ra(inode->i_mapping, ra, NULL, i,
2541 calc_ra(i, last_index, ra_pages));
2542 }
2543 total_read++;
2544 if (((u64)i << PAGE_CACHE_SHIFT) > inode->i_size)
2545 goto truncate_racing;
2546
2547 page = grab_cache_page(inode->i_mapping, i);
2548 if (!page) {
2549 goto out_unlock;
2550 }
2551 if (!PageUptodate(page)) {
2552 btrfs_readpage(NULL, page);
2553 lock_page(page);
2554 if (!PageUptodate(page)) {
2555 unlock_page(page);
2556 page_cache_release(page);
2557 goto out_unlock;
2558 }
2559 }
2560 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
2561 ClearPageDirty(page);
2562 #else
2563 cancel_dirty_page(page, PAGE_CACHE_SIZE);
2564 #endif
2565 wait_on_page_writeback(page);
2566 set_page_extent_mapped(page);
2567 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
2568 page_end = page_start + PAGE_CACHE_SIZE - 1;
2569
2570 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
2571
2572 set_extent_delalloc(io_tree, page_start,
2573 page_end, GFP_NOFS);
2574 set_page_dirty(page);
2575
2576 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
2577 unlock_page(page);
2578 page_cache_release(page);
2579 }
2580 balance_dirty_pages_ratelimited_nr(inode->i_mapping,
2581 total_read);
2582
2583 out_unlock:
2584 kfree(ra);
2585 trans = btrfs_start_transaction(BTRFS_I(inode)->root, 1);
2586 if (trans) {
2587 btrfs_add_ordered_inode(inode);
2588 btrfs_end_transaction(trans, BTRFS_I(inode)->root);
2589 mark_inode_dirty(inode);
2590 }
2591 mutex_unlock(&inode->i_mutex);
2592 return 0;
2593
2594 truncate_racing:
2595 vmtruncate(inode, inode->i_size);
2596 balance_dirty_pages_ratelimited_nr(inode->i_mapping,
2597 total_read);
2598 goto out_unlock;
2599 }
2600
2601 /*
2602 * The back references tell us which tree holds a ref on a block,
2603 * but it is possible for the tree root field in the reference to
2604 * reflect the original root before a snapshot was made. In this
2605 * case we should search through all the children of a given root
2606 * to find potential holders of references on a block.
2607 *
2608 * Instead, we do something a little less fancy and just search
2609 * all the roots for a given key/block combination.
2610 */
2611 static int find_root_for_ref(struct btrfs_root *root,
2612 struct btrfs_path *path,
2613 struct btrfs_key *key0,
2614 int level,
2615 int file_key,
2616 struct btrfs_root **found_root,
2617 u64 bytenr)
2618 {
2619 struct btrfs_key root_location;
2620 struct btrfs_root *cur_root = *found_root;
2621 struct btrfs_file_extent_item *file_extent;
2622 u64 root_search_start = BTRFS_FS_TREE_OBJECTID;
2623 u64 found_bytenr;
2624 int ret;
2625
2626 root_location.offset = (u64)-1;
2627 root_location.type = BTRFS_ROOT_ITEM_KEY;
2628 path->lowest_level = level;
2629 path->reada = 0;
2630 while(1) {
2631 ret = btrfs_search_slot(NULL, cur_root, key0, path, 0, 0);
2632 found_bytenr = 0;
2633 if (ret == 0 && file_key) {
2634 struct extent_buffer *leaf = path->nodes[0];
2635 file_extent = btrfs_item_ptr(leaf, path->slots[0],
2636 struct btrfs_file_extent_item);
2637 if (btrfs_file_extent_type(leaf, file_extent) ==
2638 BTRFS_FILE_EXTENT_REG) {
2639 found_bytenr =
2640 btrfs_file_extent_disk_bytenr(leaf,
2641 file_extent);
2642 }
2643 } else if (!file_key) {
2644 if (path->nodes[level])
2645 found_bytenr = path->nodes[level]->start;
2646 }
2647
2648 btrfs_release_path(cur_root, path);
2649
2650 if (found_bytenr == bytenr) {
2651 *found_root = cur_root;
2652 ret = 0;
2653 goto out;
2654 }
2655 ret = btrfs_search_root(root->fs_info->tree_root,
2656 root_search_start, &root_search_start);
2657 if (ret)
2658 break;
2659
2660 root_location.objectid = root_search_start;
2661 cur_root = btrfs_read_fs_root_no_name(root->fs_info,
2662 &root_location);
2663 if (!cur_root) {
2664 ret = 1;
2665 break;
2666 }
2667 }
2668 out:
2669 path->lowest_level = 0;
2670 return ret;
2671 }
2672
2673 /*
2674 * note, this releases the path
2675 */
2676 static int noinline relocate_one_reference(struct btrfs_root *extent_root,
2677 struct btrfs_path *path,
2678 struct btrfs_key *extent_key,
2679 u64 *last_file_objectid,
2680 u64 *last_file_offset,
2681 u64 *last_file_root,
2682 u64 last_extent)
2683 {
2684 struct inode *inode;
2685 struct btrfs_root *found_root;
2686 struct btrfs_key root_location;
2687 struct btrfs_key found_key;
2688 struct btrfs_extent_ref *ref;
2689 u64 ref_root;
2690 u64 ref_gen;
2691 u64 ref_objectid;
2692 u64 ref_offset;
2693 int ret;
2694 int level;
2695
2696 WARN_ON(!mutex_is_locked(&extent_root->fs_info->alloc_mutex));
2697
2698 ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
2699 struct btrfs_extent_ref);
2700 ref_root = btrfs_ref_root(path->nodes[0], ref);
2701 ref_gen = btrfs_ref_generation(path->nodes[0], ref);
2702 ref_objectid = btrfs_ref_objectid(path->nodes[0], ref);
2703 ref_offset = btrfs_ref_offset(path->nodes[0], ref);
2704 btrfs_release_path(extent_root, path);
2705
2706 root_location.objectid = ref_root;
2707 if (ref_gen == 0)
2708 root_location.offset = 0;
2709 else
2710 root_location.offset = (u64)-1;
2711 root_location.type = BTRFS_ROOT_ITEM_KEY;
2712
2713 found_root = btrfs_read_fs_root_no_name(extent_root->fs_info,
2714 &root_location);
2715 BUG_ON(!found_root);
2716 mutex_unlock(&extent_root->fs_info->alloc_mutex);
2717
2718 if (ref_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
2719 found_key.objectid = ref_objectid;
2720 found_key.type = BTRFS_EXTENT_DATA_KEY;
2721 found_key.offset = ref_offset;
2722 level = 0;
2723
2724 if (last_extent == extent_key->objectid &&
2725 *last_file_objectid == ref_objectid &&
2726 *last_file_offset == ref_offset &&
2727 *last_file_root == ref_root)
2728 goto out;
2729
2730 ret = find_root_for_ref(extent_root, path, &found_key,
2731 level, 1, &found_root,
2732 extent_key->objectid);
2733
2734 if (ret)
2735 goto out;
2736
2737 if (last_extent == extent_key->objectid &&
2738 *last_file_objectid == ref_objectid &&
2739 *last_file_offset == ref_offset &&
2740 *last_file_root == ref_root)
2741 goto out;
2742
2743 inode = btrfs_iget_locked(extent_root->fs_info->sb,
2744 ref_objectid, found_root);
2745 if (inode->i_state & I_NEW) {
2746 /* the inode and parent dir are two different roots */
2747 BTRFS_I(inode)->root = found_root;
2748 BTRFS_I(inode)->location.objectid = ref_objectid;
2749 BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY;
2750 BTRFS_I(inode)->location.offset = 0;
2751 btrfs_read_locked_inode(inode);
2752 unlock_new_inode(inode);
2753
2754 }
2755 /* this can happen if the reference is not against
2756 * the latest version of the tree root
2757 */
2758 if (is_bad_inode(inode))
2759 goto out;
2760
2761 *last_file_objectid = inode->i_ino;
2762 *last_file_root = found_root->root_key.objectid;
2763 *last_file_offset = ref_offset;
2764
2765 relocate_inode_pages(inode, ref_offset, extent_key->offset);
2766 iput(inode);
2767 } else {
2768 struct btrfs_trans_handle *trans;
2769 struct extent_buffer *eb;
2770 int needs_lock = 0;
2771
2772 eb = read_tree_block(found_root, extent_key->objectid,
2773 extent_key->offset, 0);
2774 btrfs_tree_lock(eb);
2775 level = btrfs_header_level(eb);
2776
2777 if (level == 0)
2778 btrfs_item_key_to_cpu(eb, &found_key, 0);
2779 else
2780 btrfs_node_key_to_cpu(eb, &found_key, 0);
2781
2782 btrfs_tree_unlock(eb);
2783 free_extent_buffer(eb);
2784
2785 ret = find_root_for_ref(extent_root, path, &found_key,
2786 level, 0, &found_root,
2787 extent_key->objectid);
2788
2789 if (ret)
2790 goto out;
2791
2792 /*
2793 * right here almost anything could happen to our key,
2794 * but that's ok. The cow below will either relocate it
2795 * or someone else will have relocated it. Either way,
2796 * it is in a different spot than it was before and
2797 * we're happy.
2798 */
2799
2800 trans = btrfs_start_transaction(found_root, 1);
2801
2802 if (found_root == extent_root->fs_info->extent_root ||
2803 found_root == extent_root->fs_info->chunk_root ||
2804 found_root == extent_root->fs_info->dev_root) {
2805 needs_lock = 1;
2806 mutex_lock(&extent_root->fs_info->alloc_mutex);
2807 }
2808
2809 path->lowest_level = level;
2810 path->reada = 2;
2811 ret = btrfs_search_slot(trans, found_root, &found_key, path,
2812 0, 1);
2813 path->lowest_level = 0;
2814 btrfs_release_path(found_root, path);
2815
2816 if (found_root == found_root->fs_info->extent_root)
2817 btrfs_extent_post_op(trans, found_root);
2818 if (needs_lock)
2819 mutex_unlock(&extent_root->fs_info->alloc_mutex);
2820
2821 btrfs_end_transaction(trans, found_root);
2822
2823 }
2824 out:
2825 mutex_lock(&extent_root->fs_info->alloc_mutex);
2826 return 0;
2827 }
2828
2829 static int noinline del_extent_zero(struct btrfs_root *extent_root,
2830 struct btrfs_path *path,
2831 struct btrfs_key *extent_key)
2832 {
2833 int ret;
2834 struct btrfs_trans_handle *trans;
2835
2836 trans = btrfs_start_transaction(extent_root, 1);
2837 ret = btrfs_search_slot(trans, extent_root, extent_key, path, -1, 1);
2838 if (ret > 0) {
2839 ret = -EIO;
2840 goto out;
2841 }
2842 if (ret < 0)
2843 goto out;
2844 ret = btrfs_del_item(trans, extent_root, path);
2845 out:
2846 btrfs_end_transaction(trans, extent_root);
2847 return ret;
2848 }
2849
2850 static int noinline relocate_one_extent(struct btrfs_root *extent_root,
2851 struct btrfs_path *path,
2852 struct btrfs_key *extent_key)
2853 {
2854 struct btrfs_key key;
2855 struct btrfs_key found_key;
2856 struct extent_buffer *leaf;
2857 u64 last_file_objectid = 0;
2858 u64 last_file_root = 0;
2859 u64 last_file_offset = (u64)-1;
2860 u64 last_extent = 0;
2861 u32 nritems;
2862 u32 item_size;
2863 int ret = 0;
2864
2865 if (extent_key->objectid == 0) {
2866 ret = del_extent_zero(extent_root, path, extent_key);
2867 goto out;
2868 }
2869 key.objectid = extent_key->objectid;
2870 key.type = BTRFS_EXTENT_REF_KEY;
2871 key.offset = 0;
2872
2873 while(1) {
2874 ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
2875
2876 if (ret < 0)
2877 goto out;
2878
2879 ret = 0;
2880 leaf = path->nodes[0];
2881 nritems = btrfs_header_nritems(leaf);
2882 if (path->slots[0] == nritems) {
2883 ret = btrfs_next_leaf(extent_root, path);
2884 if (ret > 0) {
2885 ret = 0;
2886 goto out;
2887 }
2888 if (ret < 0)
2889 goto out;
2890 leaf = path->nodes[0];
2891 }
2892
2893 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
2894 if (found_key.objectid != extent_key->objectid) {
2895 break;
2896 }
2897
2898 if (found_key.type != BTRFS_EXTENT_REF_KEY) {
2899 break;
2900 }
2901
2902 key.offset = found_key.offset + 1;
2903 item_size = btrfs_item_size_nr(leaf, path->slots[0]);
2904
2905 ret = relocate_one_reference(extent_root, path, extent_key,
2906 &last_file_objectid,
2907 &last_file_offset,
2908 &last_file_root, last_extent);
2909 if (ret)
2910 goto out;
2911 last_extent = extent_key->objectid;
2912 }
2913 ret = 0;
2914 out:
2915 btrfs_release_path(extent_root, path);
2916 return ret;
2917 }
2918
2919 static u64 update_block_group_flags(struct btrfs_root *root, u64 flags)
2920 {
2921 u64 num_devices;
2922 u64 stripped = BTRFS_BLOCK_GROUP_RAID0 |
2923 BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10;
2924
2925 num_devices = root->fs_info->fs_devices->num_devices;
2926 if (num_devices == 1) {
2927 stripped |= BTRFS_BLOCK_GROUP_DUP;
2928 stripped = flags & ~stripped;
2929
2930 /* turn raid0 into single device chunks */
2931 if (flags & BTRFS_BLOCK_GROUP_RAID0)
2932 return stripped;
2933
2934 /* turn mirroring into duplication */
2935 if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
2936 BTRFS_BLOCK_GROUP_RAID10))
2937 return stripped | BTRFS_BLOCK_GROUP_DUP;
2938 return flags;
2939 } else {
2940 /* they already had raid on here, just return */
2941 if (flags & stripped)
2942 return flags;
2943
2944 stripped |= BTRFS_BLOCK_GROUP_DUP;
2945 stripped = flags & ~stripped;
2946
2947 /* switch duplicated blocks with raid1 */
2948 if (flags & BTRFS_BLOCK_GROUP_DUP)
2949 return stripped | BTRFS_BLOCK_GROUP_RAID1;
2950
2951 /* turn single device chunks into raid0 */
2952 return stripped | BTRFS_BLOCK_GROUP_RAID0;
2953 }
2954 return flags;
2955 }
2956
2957 int __alloc_chunk_for_shrink(struct btrfs_root *root,
2958 struct btrfs_block_group_cache *shrink_block_group,
2959 int force)
2960 {
2961 struct btrfs_trans_handle *trans;
2962 u64 new_alloc_flags;
2963 u64 calc;
2964
2965 if (btrfs_block_group_used(&shrink_block_group->item) > 0) {
2966
2967 mutex_unlock(&root->fs_info->alloc_mutex);
2968 trans = btrfs_start_transaction(root, 1);
2969 mutex_lock(&root->fs_info->alloc_mutex);
2970
2971 new_alloc_flags = update_block_group_flags(root,
2972 shrink_block_group->flags);
2973 if (new_alloc_flags != shrink_block_group->flags) {
2974 calc =
2975 btrfs_block_group_used(&shrink_block_group->item);
2976 } else {
2977 calc = shrink_block_group->key.offset;
2978 }
2979 do_chunk_alloc(trans, root->fs_info->extent_root,
2980 calc + 2 * 1024 * 1024, new_alloc_flags, force);
2981
2982 mutex_unlock(&root->fs_info->alloc_mutex);
2983 btrfs_end_transaction(trans, root);
2984 mutex_lock(&root->fs_info->alloc_mutex);
2985 }
2986 return 0;
2987 }
2988
2989 int btrfs_shrink_extent_tree(struct btrfs_root *root, u64 shrink_start)
2990 {
2991 struct btrfs_trans_handle *trans;
2992 struct btrfs_root *tree_root = root->fs_info->tree_root;
2993 struct btrfs_path *path;
2994 u64 cur_byte;
2995 u64 total_found;
2996 u64 shrink_last_byte;
2997 struct btrfs_block_group_cache *shrink_block_group;
2998 struct btrfs_fs_info *info = root->fs_info;
2999 struct btrfs_key key;
3000 struct btrfs_key found_key;
3001 struct extent_buffer *leaf;
3002 u32 nritems;
3003 int ret;
3004 int progress;
3005
3006 mutex_lock(&root->fs_info->alloc_mutex);
3007 shrink_block_group = btrfs_lookup_block_group(root->fs_info,
3008 shrink_start);
3009 BUG_ON(!shrink_block_group);
3010
3011 shrink_last_byte = shrink_block_group->key.objectid +
3012 shrink_block_group->key.offset;
3013
3014 shrink_block_group->space_info->total_bytes -=
3015 shrink_block_group->key.offset;
3016 path = btrfs_alloc_path();
3017 root = root->fs_info->extent_root;
3018 path->reada = 2;
3019
3020 printk("btrfs relocating block group %llu flags %llu\n",
3021 (unsigned long long)shrink_start,
3022 (unsigned long long)shrink_block_group->flags);
3023
3024 __alloc_chunk_for_shrink(root, shrink_block_group, 1);
3025
3026 again:
3027
3028 shrink_block_group->ro = 1;
3029
3030 total_found = 0;
3031 progress = 0;
3032 key.objectid = shrink_start;
3033 key.offset = 0;
3034 key.type = 0;
3035 cur_byte = key.objectid;
3036
3037 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3038 if (ret < 0)
3039 goto out;
3040
3041 ret = btrfs_previous_item(root, path, 0, BTRFS_EXTENT_ITEM_KEY);
3042 if (ret < 0)
3043 goto out;
3044
3045 if (ret == 0) {
3046 leaf = path->nodes[0];
3047 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
3048 if (found_key.objectid + found_key.offset > shrink_start &&
3049 found_key.objectid < shrink_last_byte) {
3050 cur_byte = found_key.objectid;
3051 key.objectid = cur_byte;
3052 }
3053 }
3054 btrfs_release_path(root, path);
3055
3056 while(1) {
3057 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3058 if (ret < 0)
3059 goto out;
3060
3061 next:
3062 leaf = path->nodes[0];
3063 nritems = btrfs_header_nritems(leaf);
3064 if (path->slots[0] >= nritems) {
3065 ret = btrfs_next_leaf(root, path);
3066 if (ret < 0)
3067 goto out;
3068 if (ret == 1) {
3069 ret = 0;
3070 break;
3071 }
3072 leaf = path->nodes[0];
3073 nritems = btrfs_header_nritems(leaf);
3074 }
3075
3076 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
3077
3078 if (found_key.objectid >= shrink_last_byte)
3079 break;
3080
3081 if (progress && need_resched()) {
3082 memcpy(&key, &found_key, sizeof(key));
3083 cond_resched();
3084 btrfs_release_path(root, path);
3085 btrfs_search_slot(NULL, root, &key, path, 0, 0);
3086 progress = 0;
3087 goto next;
3088 }
3089 progress = 1;
3090
3091 if (btrfs_key_type(&found_key) != BTRFS_EXTENT_ITEM_KEY ||
3092 found_key.objectid + found_key.offset <= cur_byte) {
3093 memcpy(&key, &found_key, sizeof(key));
3094 key.offset++;
3095 path->slots[0]++;
3096 goto next;
3097 }
3098
3099 total_found++;
3100 cur_byte = found_key.objectid + found_key.offset;
3101 key.objectid = cur_byte;
3102 btrfs_release_path(root, path);
3103 ret = relocate_one_extent(root, path, &found_key);
3104 __alloc_chunk_for_shrink(root, shrink_block_group, 0);
3105 }
3106
3107 btrfs_release_path(root, path);
3108
3109 if (total_found > 0) {
3110 printk("btrfs relocate found %llu last extent was %llu\n",
3111 (unsigned long long)total_found,
3112 (unsigned long long)found_key.objectid);
3113 mutex_unlock(&root->fs_info->alloc_mutex);
3114 trans = btrfs_start_transaction(tree_root, 1);
3115 btrfs_commit_transaction(trans, tree_root);
3116
3117 btrfs_clean_old_snapshots(tree_root);
3118
3119 trans = btrfs_start_transaction(tree_root, 1);
3120 btrfs_commit_transaction(trans, tree_root);
3121 mutex_lock(&root->fs_info->alloc_mutex);
3122 goto again;
3123 }
3124
3125 /*
3126 * we've freed all the extents, now remove the block
3127 * group item from the tree
3128 */
3129 mutex_unlock(&root->fs_info->alloc_mutex);
3130
3131 trans = btrfs_start_transaction(root, 1);
3132 mutex_lock(&root->fs_info->alloc_mutex);
3133 memcpy(&key, &shrink_block_group->key, sizeof(key));
3134
3135 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
3136 if (ret > 0)
3137 ret = -EIO;
3138 if (ret < 0)
3139 goto out;
3140
3141 clear_extent_bits(&info->block_group_cache, key.objectid,
3142 key.objectid + key.offset - 1,
3143 (unsigned int)-1, GFP_NOFS);
3144
3145
3146 clear_extent_bits(&info->free_space_cache,
3147 key.objectid, key.objectid + key.offset - 1,
3148 (unsigned int)-1, GFP_NOFS);
3149
3150 memset(shrink_block_group, 0, sizeof(*shrink_block_group));
3151 kfree(shrink_block_group);
3152
3153 btrfs_del_item(trans, root, path);
3154 btrfs_release_path(root, path);
3155 mutex_unlock(&root->fs_info->alloc_mutex);
3156 btrfs_commit_transaction(trans, root);
3157
3158 mutex_lock(&root->fs_info->alloc_mutex);
3159
3160 /* the code to unpin extents might set a few bits in the free
3161 * space cache for this range again
3162 */
3163 clear_extent_bits(&info->free_space_cache,
3164 key.objectid, key.objectid + key.offset - 1,
3165 (unsigned int)-1, GFP_NOFS);
3166 out:
3167 btrfs_free_path(path);
3168 mutex_unlock(&root->fs_info->alloc_mutex);
3169 return ret;
3170 }
3171
3172 int find_first_block_group(struct btrfs_root *root, struct btrfs_path *path,
3173 struct btrfs_key *key)
3174 {
3175 int ret = 0;
3176 struct btrfs_key found_key;
3177 struct extent_buffer *leaf;
3178 int slot;
3179
3180 ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
3181 if (ret < 0)
3182 goto out;
3183
3184 while(1) {
3185 slot = path->slots[0];
3186 leaf = path->nodes[0];
3187 if (slot >= btrfs_header_nritems(leaf)) {
3188 ret = btrfs_next_leaf(root, path);
3189 if (ret == 0)
3190 continue;
3191 if (ret < 0)
3192 goto out;
3193 break;
3194 }
3195 btrfs_item_key_to_cpu(leaf, &found_key, slot);
3196
3197 if (found_key.objectid >= key->objectid &&
3198 found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) {
3199 ret = 0;
3200 goto out;
3201 }
3202 path->slots[0]++;
3203 }
3204 ret = -ENOENT;
3205 out:
3206 return ret;
3207 }
3208
3209 int btrfs_read_block_groups(struct btrfs_root *root)
3210 {
3211 struct btrfs_path *path;
3212 int ret;
3213 int bit;
3214 struct btrfs_block_group_cache *cache;
3215 struct btrfs_fs_info *info = root->fs_info;
3216 struct btrfs_space_info *space_info;
3217 struct extent_io_tree *block_group_cache;
3218 struct btrfs_key key;
3219 struct btrfs_key found_key;
3220 struct extent_buffer *leaf;
3221
3222 block_group_cache = &info->block_group_cache;
3223 root = info->extent_root;
3224 key.objectid = 0;
3225 key.offset = 0;
3226 btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
3227 path = btrfs_alloc_path();
3228 if (!path)
3229 return -ENOMEM;
3230
3231 mutex_lock(&root->fs_info->alloc_mutex);
3232 while(1) {
3233 ret = find_first_block_group(root, path, &key);
3234 if (ret > 0) {
3235 ret = 0;
3236 goto error;
3237 }
3238 if (ret != 0)
3239 goto error;
3240
3241 leaf = path->nodes[0];
3242 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
3243 cache = kzalloc(sizeof(*cache), GFP_NOFS);
3244 if (!cache) {
3245 ret = -ENOMEM;
3246 break;
3247 }
3248
3249 read_extent_buffer(leaf, &cache->item,
3250 btrfs_item_ptr_offset(leaf, path->slots[0]),
3251 sizeof(cache->item));
3252 memcpy(&cache->key, &found_key, sizeof(found_key));
3253
3254 key.objectid = found_key.objectid + found_key.offset;
3255 btrfs_release_path(root, path);
3256 cache->flags = btrfs_block_group_flags(&cache->item);
3257 bit = 0;
3258 if (cache->flags & BTRFS_BLOCK_GROUP_DATA) {
3259 bit = BLOCK_GROUP_DATA;
3260 } else if (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM) {
3261 bit = BLOCK_GROUP_SYSTEM;
3262 } else if (cache->flags & BTRFS_BLOCK_GROUP_METADATA) {
3263 bit = BLOCK_GROUP_METADATA;
3264 }
3265 set_avail_alloc_bits(info, cache->flags);
3266
3267 ret = update_space_info(info, cache->flags, found_key.offset,
3268 btrfs_block_group_used(&cache->item),
3269 &space_info);
3270 BUG_ON(ret);
3271 cache->space_info = space_info;
3272
3273 /* use EXTENT_LOCKED to prevent merging */
3274 set_extent_bits(block_group_cache, found_key.objectid,
3275 found_key.objectid + found_key.offset - 1,
3276 bit | EXTENT_LOCKED, GFP_NOFS);
3277 set_state_private(block_group_cache, found_key.objectid,
3278 (unsigned long)cache);
3279
3280 if (key.objectid >=
3281 btrfs_super_total_bytes(&info->super_copy))
3282 break;
3283 }
3284 ret = 0;
3285 error:
3286 btrfs_free_path(path);
3287 mutex_unlock(&root->fs_info->alloc_mutex);
3288 return ret;
3289 }
3290
3291 int btrfs_make_block_group(struct btrfs_trans_handle *trans,
3292 struct btrfs_root *root, u64 bytes_used,
3293 u64 type, u64 chunk_objectid, u64 chunk_offset,
3294 u64 size)
3295 {
3296 int ret;
3297 int bit = 0;
3298 struct btrfs_root *extent_root;
3299 struct btrfs_block_group_cache *cache;
3300 struct extent_io_tree *block_group_cache;
3301
3302 WARN_ON(!mutex_is_locked(&root->fs_info->alloc_mutex));
3303 extent_root = root->fs_info->extent_root;
3304 block_group_cache = &root->fs_info->block_group_cache;
3305
3306 cache = kzalloc(sizeof(*cache), GFP_NOFS);
3307 BUG_ON(!cache);
3308 cache->key.objectid = chunk_offset;
3309 cache->key.offset = size;
3310 btrfs_set_key_type(&cache->key, BTRFS_BLOCK_GROUP_ITEM_KEY);
3311
3312 btrfs_set_block_group_used(&cache->item, bytes_used);
3313 btrfs_set_block_group_chunk_objectid(&cache->item, chunk_objectid);
3314 cache->flags = type;
3315 btrfs_set_block_group_flags(&cache->item, type);
3316
3317 ret = update_space_info(root->fs_info, cache->flags, size, bytes_used,
3318 &cache->space_info);
3319 BUG_ON(ret);
3320
3321 bit = block_group_state_bits(type);
3322 set_extent_bits(block_group_cache, chunk_offset,
3323 chunk_offset + size - 1,
3324 bit | EXTENT_LOCKED, GFP_NOFS);
3325
3326 set_state_private(block_group_cache, chunk_offset,
3327 (unsigned long)cache);
3328 ret = btrfs_insert_item(trans, extent_root, &cache->key, &cache->item,
3329 sizeof(cache->item));
3330 BUG_ON(ret);
3331
3332 finish_current_insert(trans, extent_root);
3333 ret = del_pending_extents(trans, extent_root);
3334 BUG_ON(ret);
3335 set_avail_alloc_bits(extent_root->fs_info, type);
3336
3337 return 0;
3338 }
Linux kernel - Deliverables
This page took 0.101934 seconds and 5 git commands to generate.