?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
btrfs-progs: Fix infinite loop of btrfs subvolumn sync
[btrfs-progs-unstable/devel.git] / disk-io.c
blob1d48893224113e53bb8c00c56c500e1e147e1e8a
1 /*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19 #include <stdio.h>
20 #include <stdlib.h>
21 #include <sys/types.h>
22 #include <sys/stat.h>
23 #include <fcntl.h>
24 #include <unistd.h>
25 #include <uuid/uuid.h>
26 #include "kerncompat.h"
27 #include "radix-tree.h"
28 #include "ctree.h"
29 #include "disk-io.h"
30 #include "volumes.h"
31 #include "transaction.h"
32 #include "crc32c.h"
33 #include "utils.h"
34 #include "print-tree.h"
35 #include "rbtree-utils.h"
36
37 /* specified errno for check_tree_block */
38 #define BTRFS_BAD_BYTENR (-1)
39 #define BTRFS_BAD_FSID (-2)
40 #define BTRFS_BAD_LEVEL (-3)
41 #define BTRFS_BAD_NRITEMS (-4)
42
43 /* Calculate max possible nritems for a leaf/node */
44 static u32 max_nritems(u8 level, u32 nodesize)
45 {
46
47 if (level == 0)
48 return ((nodesize - sizeof(struct btrfs_header)) /
49 sizeof(struct btrfs_item));
50 return ((nodesize - sizeof(struct btrfs_header)) /
51 sizeof(struct btrfs_key_ptr));
52 }
53
54 static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
55 {
56
57 struct btrfs_fs_devices *fs_devices;
58 int ret = BTRFS_BAD_FSID;
59
60 if (buf->start != btrfs_header_bytenr(buf))
61 return BTRFS_BAD_BYTENR;
62 if (btrfs_header_level(buf) >= BTRFS_MAX_LEVEL)
63 return BTRFS_BAD_LEVEL;
64 if (btrfs_header_nritems(buf) > max_nritems(btrfs_header_level(buf),
65 root->nodesize))
66 return BTRFS_BAD_NRITEMS;
67
68 fs_devices = root->fs_info->fs_devices;
69 while (fs_devices) {
70 if (root->fs_info->ignore_fsid_mismatch ||
71 !memcmp_extent_buffer(buf, fs_devices->fsid,
72 btrfs_header_fsid(),
73 BTRFS_FSID_SIZE)) {
74 ret = 0;
75 break;
76 }
77 fs_devices = fs_devices->seed;
78 }
79 return ret;
80 }
81
82 static void print_tree_block_error(struct btrfs_root *root,
83 struct extent_buffer *eb,
84 int err)
85 {
86 char fs_uuid[BTRFS_UUID_UNPARSED_SIZE] = {'\0'};
87 char found_uuid[BTRFS_UUID_UNPARSED_SIZE] = {'\0'};
88 u8 buf[BTRFS_UUID_SIZE];
89
90 switch (err) {
91 case BTRFS_BAD_FSID:
92 read_extent_buffer(eb, buf, btrfs_header_fsid(),
93 BTRFS_UUID_SIZE);
94 uuid_unparse(buf, found_uuid);
95 uuid_unparse(root->fs_info->fsid, fs_uuid);
96 fprintf(stderr, "fsid mismatch, want=%s, have=%s\n",
97 fs_uuid, found_uuid);
98 break;
99 case BTRFS_BAD_BYTENR:
100 fprintf(stderr, "bytenr mismatch, want=%llu, have=%llu\n",
101 eb->start, btrfs_header_bytenr(eb));
102 break;
103 case BTRFS_BAD_LEVEL:
104 fprintf(stderr, "bad level, %u > %u\n",
105 btrfs_header_level(eb), BTRFS_MAX_LEVEL);
106 break;
107 case BTRFS_BAD_NRITEMS:
108 fprintf(stderr, "invalid nr_items: %u\n",
109 btrfs_header_nritems(eb));
110 break;
111 }
112 }
113
114 u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
115 {
116 return crc32c(seed, data, len);
117 }
118
119 void btrfs_csum_final(u32 crc, char *result)
120 {
121 *(__le32 *)result = ~cpu_to_le32(crc);
122 }
123
124 static int __csum_tree_block_size(struct extent_buffer *buf, u16 csum_size,
125 int verify, int silent)
126 {
127 char *result;
128 u32 len;
129 u32 crc = ~(u32)0;
130
131 result = malloc(csum_size * sizeof(char));
132 if (!result)
133 return 1;
134
135 len = buf->len - BTRFS_CSUM_SIZE;
136 crc = crc32c(crc, buf->data + BTRFS_CSUM_SIZE, len);
137 btrfs_csum_final(crc, result);
138
139 if (verify) {
140 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
141 if (!silent)
142 printk("checksum verify failed on %llu found %08X wanted %08X\n",
143 (unsigned long long)buf->start,
144 *((u32 *)result),
145 *((u32*)(char *)buf->data));
146 free(result);
147 return 1;
148 }
149 } else {
150 write_extent_buffer(buf, result, 0, csum_size);
151 }
152 free(result);
153 return 0;
154 }
155
156 int csum_tree_block_size(struct extent_buffer *buf, u16 csum_size, int verify)
157 {
158 return __csum_tree_block_size(buf, csum_size, verify, 0);
159 }
160
161 int verify_tree_block_csum_silent(struct extent_buffer *buf, u16 csum_size)
162 {
163 return __csum_tree_block_size(buf, csum_size, 1, 1);
164 }
165
166 int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
167 int verify)
168 {
169 u16 csum_size =
170 btrfs_super_csum_size(root->fs_info->super_copy);
171 if (verify && root->fs_info->suppress_check_block_errors)
172 return verify_tree_block_csum_silent(buf, csum_size);
173 return csum_tree_block_size(buf, csum_size, verify);
174 }
175
176 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
177 u64 bytenr, u32 blocksize)
178 {
179 return find_extent_buffer(&root->fs_info->extent_cache,
180 bytenr, blocksize);
181 }
182
183 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
184 u64 bytenr, u32 blocksize)
185 {
186 return alloc_extent_buffer(&root->fs_info->extent_cache, bytenr,
187 blocksize);
188 }
189
190 void readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
191 u64 parent_transid)
192 {
193 struct extent_buffer *eb;
194 u64 length;
195 struct btrfs_multi_bio *multi = NULL;
196 struct btrfs_device *device;
197
198 eb = btrfs_find_tree_block(root, bytenr, blocksize);
199 if (!(eb && btrfs_buffer_uptodate(eb, parent_transid)) &&
200 !btrfs_map_block(&root->fs_info->mapping_tree, READ,
201 bytenr, &length, &multi, 0, NULL)) {
202 device = multi->stripes[0].dev;
203 device->total_ios++;
204 blocksize = min(blocksize, (u32)(64 * 1024));
205 readahead(device->fd, multi->stripes[0].physical, blocksize);
206 }
207
208 free_extent_buffer(eb);
209 kfree(multi);
210 }
211
212 static int verify_parent_transid(struct extent_io_tree *io_tree,
213 struct extent_buffer *eb, u64 parent_transid,
214 int ignore)
215 {
216 int ret;
217
218 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
219 return 0;
220
221 if (extent_buffer_uptodate(eb) &&
222 btrfs_header_generation(eb) == parent_transid) {
223 ret = 0;
224 goto out;
225 }
226 printk("parent transid verify failed on %llu wanted %llu found %llu\n",
227 (unsigned long long)eb->start,
228 (unsigned long long)parent_transid,
229 (unsigned long long)btrfs_header_generation(eb));
230 if (ignore) {
231 eb->flags |= EXTENT_BAD_TRANSID;
232 printk("Ignoring transid failure\n");
233 return 0;
234 }
235
236 ret = 1;
237 out:
238 clear_extent_buffer_uptodate(io_tree, eb);
239 return ret;
240
241 }
242
243
244 int read_whole_eb(struct btrfs_fs_info *info, struct extent_buffer *eb, int mirror)
245 {
246 unsigned long offset = 0;
247 struct btrfs_multi_bio *multi = NULL;
248 struct btrfs_device *device;
249 int ret = 0;
250 u64 read_len;
251 unsigned long bytes_left = eb->len;
252
253 while (bytes_left) {
254 read_len = bytes_left;
255 device = NULL;
256
257 if (!info->on_restoring &&
258 eb->start != BTRFS_SUPER_INFO_OFFSET) {
259 ret = btrfs_map_block(&info->mapping_tree, READ,
260 eb->start + offset, &read_len, &multi,
261 mirror, NULL);
262 if (ret) {
263 printk("Couldn't map the block %Lu\n", eb->start + offset);
264 kfree(multi);
265 return -EIO;
266 }
267 device = multi->stripes[0].dev;
268
269 if (device->fd <= 0) {
270 kfree(multi);
271 return -EIO;
272 }
273
274 eb->fd = device->fd;
275 device->total_ios++;
276 eb->dev_bytenr = multi->stripes[0].physical;
277 kfree(multi);
278 multi = NULL;
279 } else {
280 /* special case for restore metadump */
281 list_for_each_entry(device, &info->fs_devices->devices, dev_list) {
282 if (device->devid == 1)
283 break;
284 }
285
286 eb->fd = device->fd;
287 eb->dev_bytenr = eb->start;
288 device->total_ios++;
289 }
290
291 if (read_len > bytes_left)
292 read_len = bytes_left;
293
294 ret = read_extent_from_disk(eb, offset, read_len);
295 if (ret)
296 return -EIO;
297 offset += read_len;
298 bytes_left -= read_len;
299 }
300 return 0;
301 }
302
303 struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
304 u32 blocksize, u64 parent_transid)
305 {
306 int ret;
307 struct extent_buffer *eb;
308 u64 best_transid = 0;
309 int mirror_num = 0;
310 int good_mirror = 0;
311 int num_copies;
312 int ignore = 0;
313
314 eb = btrfs_find_create_tree_block(root, bytenr, blocksize);
315 if (!eb)
316 return ERR_PTR(-ENOMEM);
317
318 if (btrfs_buffer_uptodate(eb, parent_transid))
319 return eb;
320
321 while (1) {
322 ret = read_whole_eb(root->fs_info, eb, mirror_num);
323 if (ret == 0 && csum_tree_block(root, eb, 1) == 0 &&
324 check_tree_block(root, eb) == 0 &&
325 verify_parent_transid(eb->tree, eb, parent_transid, ignore)
326 == 0) {
327 if (eb->flags & EXTENT_BAD_TRANSID &&
328 list_empty(&eb->recow)) {
329 list_add_tail(&eb->recow,
330 &root->fs_info->recow_ebs);
331 eb->refs++;
332 }
333 btrfs_set_buffer_uptodate(eb);
334 return eb;
335 }
336 if (ignore) {
337 if (check_tree_block(root, eb)) {
338 if (!root->fs_info->suppress_check_block_errors)
339 print_tree_block_error(root, eb,
340 check_tree_block(root, eb));
341 } else {
342 if (!root->fs_info->suppress_check_block_errors)
343 fprintf(stderr, "Csum didn't match\n");
344 }
345 ret = -EIO;
346 break;
347 }
348 num_copies = btrfs_num_copies(&root->fs_info->mapping_tree,
349 eb->start, eb->len);
350 if (num_copies == 1) {
351 ignore = 1;
352 continue;
353 }
354 if (btrfs_header_generation(eb) > best_transid && mirror_num) {
355 best_transid = btrfs_header_generation(eb);
356 good_mirror = mirror_num;
357 }
358 mirror_num++;
359 if (mirror_num > num_copies) {
360 mirror_num = good_mirror;
361 ignore = 1;
362 continue;
363 }
364 }
365 free_extent_buffer(eb);
366 return ERR_PTR(ret);
367 }
368
369 int read_extent_data(struct btrfs_root *root, char *data,
370 u64 logical, u64 *len, int mirror)
371 {
372 u64 offset = 0;
373 struct btrfs_multi_bio *multi = NULL;
374 struct btrfs_fs_info *info = root->fs_info;
375 struct btrfs_device *device;
376 int ret = 0;
377 u64 max_len = *len;
378
379 ret = btrfs_map_block(&info->mapping_tree, READ, logical, len,
380 &multi, mirror, NULL);
381 if (ret) {
382 fprintf(stderr, "Couldn't map the block %llu\n",
383 logical + offset);
384 goto err;
385 }
386 device = multi->stripes[0].dev;
387
388 if (device->fd <= 0)
389 goto err;
390 if (*len > max_len)
391 *len = max_len;
392
393 ret = pread64(device->fd, data, *len, multi->stripes[0].physical);
394 if (ret != *len)
395 ret = -EIO;
396 else
397 ret = 0;
398 err:
399 kfree(multi);
400 return ret;
401 }
402
403 int write_and_map_eb(struct btrfs_trans_handle *trans,
404 struct btrfs_root *root,
405 struct extent_buffer *eb)
406 {
407 int ret;
408 int dev_nr;
409 u64 length;
410 u64 *raid_map = NULL;
411 struct btrfs_multi_bio *multi = NULL;
412
413 dev_nr = 0;
414 length = eb->len;
415 ret = btrfs_map_block(&root->fs_info->mapping_tree, WRITE,
416 eb->start, &length, &multi, 0, &raid_map);
417
418 if (raid_map) {
419 ret = write_raid56_with_parity(root->fs_info, eb, multi,
420 length, raid_map);
421 BUG_ON(ret);
422 } else while (dev_nr < multi->num_stripes) {
423 BUG_ON(ret);
424 eb->fd = multi->stripes[dev_nr].dev->fd;
425 eb->dev_bytenr = multi->stripes[dev_nr].physical;
426 multi->stripes[dev_nr].dev->total_ios++;
427 dev_nr++;
428 ret = write_extent_to_disk(eb);
429 BUG_ON(ret);
430 }
431 kfree(raid_map);
432 kfree(multi);
433 return 0;
434 }
435
436 int write_tree_block(struct btrfs_trans_handle *trans,
437 struct btrfs_root *root,
438 struct extent_buffer *eb)
439 {
440 if (check_tree_block(root, eb)) {
441 print_tree_block_error(root, eb, check_tree_block(root, eb));
442 BUG();
443 }
444
445 if (trans && !btrfs_buffer_uptodate(eb, trans->transid))
446 BUG();
447
448 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
449 csum_tree_block(root, eb, 0);
450
451 return write_and_map_eb(trans, root, eb);
452 }
453
454 int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
455 u32 stripesize, struct btrfs_root *root,
456 struct btrfs_fs_info *fs_info, u64 objectid)
457 {
458 root->node = NULL;
459 root->commit_root = NULL;
460 root->sectorsize = sectorsize;
461 root->nodesize = nodesize;
462 root->leafsize = leafsize;
463 root->stripesize = stripesize;
464 root->ref_cows = 0;
465 root->track_dirty = 0;
466
467 root->fs_info = fs_info;
468 root->objectid = objectid;
469 root->last_trans = 0;
470 root->highest_inode = 0;
471 root->last_inode_alloc = 0;
472
473 INIT_LIST_HEAD(&root->dirty_list);
474 INIT_LIST_HEAD(&root->orphan_data_extents);
475 memset(&root->root_key, 0, sizeof(root->root_key));
476 memset(&root->root_item, 0, sizeof(root->root_item));
477 root->root_key.objectid = objectid;
478 return 0;
479 }
480
481 static int update_cowonly_root(struct btrfs_trans_handle *trans,
482 struct btrfs_root *root)
483 {
484 int ret;
485 u64 old_root_bytenr;
486 struct btrfs_root *tree_root = root->fs_info->tree_root;
487
488 btrfs_write_dirty_block_groups(trans, root);
489 while(1) {
490 old_root_bytenr = btrfs_root_bytenr(&root->root_item);
491 if (old_root_bytenr == root->node->start)
492 break;
493 btrfs_set_root_bytenr(&root->root_item,
494 root->node->start);
495 btrfs_set_root_generation(&root->root_item,
496 trans->transid);
497 root->root_item.level = btrfs_header_level(root->node);
498 ret = btrfs_update_root(trans, tree_root,
499 &root->root_key,
500 &root->root_item);
501 BUG_ON(ret);
502 btrfs_write_dirty_block_groups(trans, root);
503 }
504 return 0;
505 }
506
507 static int commit_tree_roots(struct btrfs_trans_handle *trans,
508 struct btrfs_fs_info *fs_info)
509 {
510 struct btrfs_root *root;
511 struct list_head *next;
512 struct extent_buffer *eb;
513 int ret;
514
515 if (fs_info->readonly)
516 return 0;
517
518 eb = fs_info->tree_root->node;
519 extent_buffer_get(eb);
520 ret = btrfs_cow_block(trans, fs_info->tree_root, eb, NULL, 0, &eb);
521 free_extent_buffer(eb);
522 if (ret)
523 return ret;
524
525 while(!list_empty(&fs_info->dirty_cowonly_roots)) {
526 next = fs_info->dirty_cowonly_roots.next;
527 list_del_init(next);
528 root = list_entry(next, struct btrfs_root, dirty_list);
529 update_cowonly_root(trans, root);
530 free_extent_buffer(root->commit_root);
531 root->commit_root = NULL;
532 }
533
534 return 0;
535 }
536
537 static int __commit_transaction(struct btrfs_trans_handle *trans,
538 struct btrfs_root *root)
539 {
540 u64 start;
541 u64 end;
542 struct extent_buffer *eb;
543 struct extent_io_tree *tree = &root->fs_info->extent_cache;
544 int ret;
545
546 while(1) {
547 ret = find_first_extent_bit(tree, 0, &start, &end,
548 EXTENT_DIRTY);
549 if (ret)
550 break;
551 while(start <= end) {
552 eb = find_first_extent_buffer(tree, start);
553 BUG_ON(!eb || eb->start != start);
554 ret = write_tree_block(trans, root, eb);
555 BUG_ON(ret);
556 start += eb->len;
557 clear_extent_buffer_dirty(eb);
558 free_extent_buffer(eb);
559 }
560 }
561 return 0;
562 }
563
564 int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
565 struct btrfs_root *root)
566 {
567 u64 transid = trans->transid;
568 int ret = 0;
569 struct btrfs_fs_info *fs_info = root->fs_info;
570
571 if (root->commit_root == root->node)
572 goto commit_tree;
573 if (root == root->fs_info->tree_root)
574 goto commit_tree;
575 if (root == root->fs_info->chunk_root)
576 goto commit_tree;
577
578 free_extent_buffer(root->commit_root);
579 root->commit_root = NULL;
580
581 btrfs_set_root_bytenr(&root->root_item, root->node->start);
582 btrfs_set_root_generation(&root->root_item, trans->transid);
583 root->root_item.level = btrfs_header_level(root->node);
584 ret = btrfs_update_root(trans, root->fs_info->tree_root,
585 &root->root_key, &root->root_item);
586 BUG_ON(ret);
587 commit_tree:
588 ret = commit_tree_roots(trans, fs_info);
589 BUG_ON(ret);
590 ret = __commit_transaction(trans, root);
591 BUG_ON(ret);
592 write_ctree_super(trans, root);
593 btrfs_finish_extent_commit(trans, fs_info->extent_root,
594 &fs_info->pinned_extents);
595 btrfs_free_transaction(root, trans);
596 free_extent_buffer(root->commit_root);
597 root->commit_root = NULL;
598 fs_info->running_transaction = NULL;
599 fs_info->last_trans_committed = transid;
600 return 0;
601 }
602
603 static int find_and_setup_root(struct btrfs_root *tree_root,
604 struct btrfs_fs_info *fs_info,
605 u64 objectid, struct btrfs_root *root)
606 {
607 int ret;
608 u32 blocksize;
609 u64 generation;
610
611 __setup_root(tree_root->nodesize, tree_root->leafsize,
612 tree_root->sectorsize, tree_root->stripesize,
613 root, fs_info, objectid);
614 ret = btrfs_find_last_root(tree_root, objectid,
615 &root->root_item, &root->root_key);
616 if (ret)
617 return ret;
618
619 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
620 generation = btrfs_root_generation(&root->root_item);
621 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
622 blocksize, generation);
623 if (!extent_buffer_uptodate(root->node))
624 return -EIO;
625
626 return 0;
627 }
628
629 static int find_and_setup_log_root(struct btrfs_root *tree_root,
630 struct btrfs_fs_info *fs_info,
631 struct btrfs_super_block *disk_super)
632 {
633 u32 blocksize;
634 u64 blocknr = btrfs_super_log_root(disk_super);
635 struct btrfs_root *log_root = malloc(sizeof(struct btrfs_root));
636
637 if (!log_root)
638 return -ENOMEM;
639
640 if (blocknr == 0) {
641 free(log_root);
642 return 0;
643 }
644
645 blocksize = btrfs_level_size(tree_root,
646 btrfs_super_log_root_level(disk_super));
647
648 __setup_root(tree_root->nodesize, tree_root->leafsize,
649 tree_root->sectorsize, tree_root->stripesize,
650 log_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
651
652 log_root->node = read_tree_block(tree_root, blocknr,
653 blocksize,
654 btrfs_super_generation(disk_super) + 1);
655
656 fs_info->log_root_tree = log_root;
657
658 if (!extent_buffer_uptodate(log_root->node)) {
659 free_extent_buffer(log_root->node);
660 free(log_root);
661 fs_info->log_root_tree = NULL;
662 return -EIO;
663 }
664
665 return 0;
666 }
667
668 int btrfs_free_fs_root(struct btrfs_root *root)
669 {
670 if (root->node)
671 free_extent_buffer(root->node);
672 if (root->commit_root)
673 free_extent_buffer(root->commit_root);
674 kfree(root);
675 return 0;
676 }
677
678 static void __free_fs_root(struct rb_node *node)
679 {
680 struct btrfs_root *root;
681
682 root = container_of(node, struct btrfs_root, rb_node);
683 btrfs_free_fs_root(root);
684 }
685
686 FREE_RB_BASED_TREE(fs_roots, __free_fs_root);
687
688 struct btrfs_root *btrfs_read_fs_root_no_cache(struct btrfs_fs_info *fs_info,
689 struct btrfs_key *location)
690 {
691 struct btrfs_root *root;
692 struct btrfs_root *tree_root = fs_info->tree_root;
693 struct btrfs_path *path;
694 struct extent_buffer *l;
695 u64 generation;
696 u32 blocksize;
697 int ret = 0;
698
699 root = malloc(sizeof(*root));
700 if (!root)
701 return ERR_PTR(-ENOMEM);
702 memset(root, 0, sizeof(*root));
703 if (location->offset == (u64)-1) {
704 ret = find_and_setup_root(tree_root, fs_info,
705 location->objectid, root);
706 if (ret) {
707 free(root);
708 return ERR_PTR(ret);
709 }
710 goto insert;
711 }
712
713 __setup_root(tree_root->nodesize, tree_root->leafsize,
714 tree_root->sectorsize, tree_root->stripesize,
715 root, fs_info, location->objectid);
716
717 path = btrfs_alloc_path();
718 BUG_ON(!path);
719 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
720 if (ret != 0) {
721 if (ret > 0)
722 ret = -ENOENT;
723 goto out;
724 }
725 l = path->nodes[0];
726 read_extent_buffer(l, &root->root_item,
727 btrfs_item_ptr_offset(l, path->slots[0]),
728 sizeof(root->root_item));
729 memcpy(&root->root_key, location, sizeof(*location));
730 ret = 0;
731 out:
732 btrfs_free_path(path);
733 if (ret) {
734 free(root);
735 return ERR_PTR(ret);
736 }
737 generation = btrfs_root_generation(&root->root_item);
738 blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
739 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
740 blocksize, generation);
741 if (!extent_buffer_uptodate(root->node)) {
742 free(root);
743 return ERR_PTR(-EIO);
744 }
745 insert:
746 root->ref_cows = 1;
747 return root;
748 }
749
750 static int btrfs_fs_roots_compare_objectids(struct rb_node *node,
751 void *data)
752 {
753 u64 objectid = *((u64 *)data);
754 struct btrfs_root *root;
755
756 root = rb_entry(node, struct btrfs_root, rb_node);
757 if (objectid > root->objectid)
758 return 1;
759 else if (objectid < root->objectid)
760 return -1;
761 else
762 return 0;
763 }
764
765 static int btrfs_fs_roots_compare_roots(struct rb_node *node1,
766 struct rb_node *node2)
767 {
768 struct btrfs_root *root;
769
770 root = rb_entry(node2, struct btrfs_root, rb_node);
771 return btrfs_fs_roots_compare_objectids(node1, (void *)&root->objectid);
772 }
773
774 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
775 struct btrfs_key *location)
776 {
777 struct btrfs_root *root;
778 struct rb_node *node;
779 int ret;
780 u64 objectid = location->objectid;
781
782 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
783 return fs_info->tree_root;
784 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
785 return fs_info->extent_root;
786 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
787 return fs_info->chunk_root;
788 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
789 return fs_info->dev_root;
790 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
791 return fs_info->csum_root;
792 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
793 return fs_info->quota_root;
794
795 BUG_ON(location->objectid == BTRFS_TREE_RELOC_OBJECTID ||
796 location->offset != (u64)-1);
797
798 node = rb_search(&fs_info->fs_root_tree, (void *)&objectid,
799 btrfs_fs_roots_compare_objectids, NULL);
800 if (node)
801 return container_of(node, struct btrfs_root, rb_node);
802
803 root = btrfs_read_fs_root_no_cache(fs_info, location);
804 if (IS_ERR(root))
805 return root;
806
807 ret = rb_insert(&fs_info->fs_root_tree, &root->rb_node,
808 btrfs_fs_roots_compare_roots);
809 BUG_ON(ret);
810 return root;
811 }
812
813 void btrfs_free_fs_info(struct btrfs_fs_info *fs_info)
814 {
815 free(fs_info->tree_root);
816 free(fs_info->extent_root);
817 free(fs_info->chunk_root);
818 free(fs_info->dev_root);
819 free(fs_info->csum_root);
820 free(fs_info->quota_root);
821 free(fs_info->super_copy);
822 free(fs_info->log_root_tree);
823 free(fs_info);
824 }
825
826 struct btrfs_fs_info *btrfs_new_fs_info(int writable, u64 sb_bytenr)
827 {
828 struct btrfs_fs_info *fs_info;
829
830 fs_info = malloc(sizeof(struct btrfs_fs_info));
831 if (!fs_info)
832 return NULL;
833
834 memset(fs_info, 0, sizeof(struct btrfs_fs_info));
835
836 fs_info->tree_root = malloc(sizeof(struct btrfs_root));
837 fs_info->extent_root = malloc(sizeof(struct btrfs_root));
838 fs_info->chunk_root = malloc(sizeof(struct btrfs_root));
839 fs_info->dev_root = malloc(sizeof(struct btrfs_root));
840 fs_info->csum_root = malloc(sizeof(struct btrfs_root));
841 fs_info->quota_root = malloc(sizeof(struct btrfs_root));
842 fs_info->super_copy = malloc(BTRFS_SUPER_INFO_SIZE);
843
844 if (!fs_info->tree_root || !fs_info->extent_root ||
845 !fs_info->chunk_root || !fs_info->dev_root ||
846 !fs_info->csum_root || !fs_info->quota_root ||
847 !fs_info->super_copy)
848 goto free_all;
849
850 memset(fs_info->super_copy, 0, BTRFS_SUPER_INFO_SIZE);
851 memset(fs_info->tree_root, 0, sizeof(struct btrfs_root));
852 memset(fs_info->extent_root, 0, sizeof(struct btrfs_root));
853 memset(fs_info->chunk_root, 0, sizeof(struct btrfs_root));
854 memset(fs_info->dev_root, 0, sizeof(struct btrfs_root));
855 memset(fs_info->csum_root, 0, sizeof(struct btrfs_root));
856 memset(fs_info->quota_root, 0, sizeof(struct btrfs_root));
857
858 extent_io_tree_init(&fs_info->extent_cache);
859 extent_io_tree_init(&fs_info->free_space_cache);
860 extent_io_tree_init(&fs_info->block_group_cache);
861 extent_io_tree_init(&fs_info->pinned_extents);
862 extent_io_tree_init(&fs_info->pending_del);
863 extent_io_tree_init(&fs_info->extent_ins);
864 fs_info->excluded_extents = NULL;
865
866 fs_info->fs_root_tree = RB_ROOT;
867 cache_tree_init(&fs_info->mapping_tree.cache_tree);
868
869 mutex_init(&fs_info->fs_mutex);
870 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
871 INIT_LIST_HEAD(&fs_info->space_info);
872 INIT_LIST_HEAD(&fs_info->recow_ebs);
873
874 if (!writable)
875 fs_info->readonly = 1;
876
877 fs_info->super_bytenr = sb_bytenr;
878 fs_info->data_alloc_profile = (u64)-1;
879 fs_info->metadata_alloc_profile = (u64)-1;
880 fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
881 return fs_info;
882 free_all:
883 btrfs_free_fs_info(fs_info);
884 return NULL;
885 }
886
887 int btrfs_check_fs_compatibility(struct btrfs_super_block *sb, int writable)
888 {
889 u64 features;
890
891 features = btrfs_super_incompat_flags(sb) &
892 ~BTRFS_FEATURE_INCOMPAT_SUPP;
893 if (features) {
894 printk("couldn't open because of unsupported "
895 "option features (%Lx).\n",
896 (unsigned long long)features);
897 return -ENOTSUP;
898 }
899
900 features = btrfs_super_incompat_flags(sb);
901 if (!(features & BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF)) {
902 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
903 btrfs_set_super_incompat_flags(sb, features);
904 }
905
906 features = btrfs_super_compat_ro_flags(sb) &
907 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
908 if (writable && features) {
909 printk("couldn't open RDWR because of unsupported "
910 "option features (%Lx).\n",
911 (unsigned long long)features);
912 return -ENOTSUP;
913 }
914 return 0;
915 }
916
917 static int find_best_backup_root(struct btrfs_super_block *super)
918 {
919 struct btrfs_root_backup *backup;
920 u64 orig_gen = btrfs_super_generation(super);
921 u64 gen = 0;
922 int best_index = 0;
923 int i;
924
925 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
926 backup = super->super_roots + i;
927 if (btrfs_backup_tree_root_gen(backup) != orig_gen &&
928 btrfs_backup_tree_root_gen(backup) > gen) {
929 best_index = i;
930 gen = btrfs_backup_tree_root_gen(backup);
931 }
932 }
933 return best_index;
934 }
935
936 static int setup_root_or_create_block(struct btrfs_fs_info *fs_info,
937 enum btrfs_open_ctree_flags flags,
938 struct btrfs_root *info_root,
939 u64 objectid, char *str)
940 {
941 struct btrfs_super_block *sb = fs_info->super_copy;
942 struct btrfs_root *root = fs_info->tree_root;
943 u32 leafsize = btrfs_super_leafsize(sb);
944 int ret;
945
946 ret = find_and_setup_root(root, fs_info, objectid, info_root);
947 if (ret) {
948 printk("Couldn't setup %s tree\n", str);
949 if (!(flags & OPEN_CTREE_PARTIAL))
950 return -EIO;
951 /*
952 * Need a blank node here just so we don't screw up in the
953 * million of places that assume a root has a valid ->node
954 */
955 info_root->node =
956 btrfs_find_create_tree_block(info_root, 0, leafsize);
957 if (!info_root->node)
958 return -ENOMEM;
959 clear_extent_buffer_uptodate(NULL, info_root->node);
960 }
961
962 return 0;
963 }
964
965 int btrfs_setup_all_roots(struct btrfs_fs_info *fs_info, u64 root_tree_bytenr,
966 enum btrfs_open_ctree_flags flags)
967 {
968 struct btrfs_super_block *sb = fs_info->super_copy;
969 struct btrfs_root *root;
970 struct btrfs_key key;
971 u32 sectorsize;
972 u32 nodesize;
973 u32 leafsize;
974 u32 stripesize;
975 u64 generation;
976 u32 blocksize;
977 int ret;
978
979 nodesize = btrfs_super_nodesize(sb);
980 leafsize = btrfs_super_leafsize(sb);
981 sectorsize = btrfs_super_sectorsize(sb);
982 stripesize = btrfs_super_stripesize(sb);
983
984 root = fs_info->tree_root;
985 __setup_root(nodesize, leafsize, sectorsize, stripesize,
986 root, fs_info, BTRFS_ROOT_TREE_OBJECTID);
987 blocksize = btrfs_level_size(root, btrfs_super_root_level(sb));
988 generation = btrfs_super_generation(sb);
989
990 if (!root_tree_bytenr && !(flags & OPEN_CTREE_BACKUP_ROOT)) {
991 root_tree_bytenr = btrfs_super_root(sb);
992 } else if (flags & OPEN_CTREE_BACKUP_ROOT) {
993 struct btrfs_root_backup *backup;
994 int index = find_best_backup_root(sb);
995 if (index >= BTRFS_NUM_BACKUP_ROOTS) {
996 fprintf(stderr, "Invalid backup root number\n");
997 return -EIO;
998 }
999 backup = fs_info->super_copy->super_roots + index;
1000 root_tree_bytenr = btrfs_backup_tree_root(backup);
1001 generation = btrfs_backup_tree_root_gen(backup);
1002 }
1003
1004 root->node = read_tree_block(root, root_tree_bytenr, blocksize,
1005 generation);
1006 if (!extent_buffer_uptodate(root->node)) {
1007 fprintf(stderr, "Couldn't read tree root\n");
1008 return -EIO;
1009 }
1010
1011 ret = setup_root_or_create_block(fs_info, flags, fs_info->extent_root,
1012 BTRFS_EXTENT_TREE_OBJECTID, "extent");
1013 if (ret)
1014 return ret;
1015 fs_info->extent_root->track_dirty = 1;
1016
1017 ret = find_and_setup_root(root, fs_info, BTRFS_DEV_TREE_OBJECTID,
1018 fs_info->dev_root);
1019 if (ret) {
1020 printk("Couldn't setup device tree\n");
1021 return -EIO;
1022 }
1023 fs_info->dev_root->track_dirty = 1;
1024
1025 ret = setup_root_or_create_block(fs_info, flags, fs_info->csum_root,
1026 BTRFS_CSUM_TREE_OBJECTID, "csum");
1027 if (ret)
1028 return ret;
1029 fs_info->csum_root->track_dirty = 1;
1030
1031 ret = find_and_setup_root(root, fs_info, BTRFS_QUOTA_TREE_OBJECTID,
1032 fs_info->quota_root);
1033 if (ret == 0)
1034 fs_info->quota_enabled = 1;
1035
1036 ret = find_and_setup_log_root(root, fs_info, sb);
1037 if (ret) {
1038 printk("Couldn't setup log root tree\n");
1039 if (!(flags & OPEN_CTREE_PARTIAL))
1040 return -EIO;
1041 }
1042
1043 fs_info->generation = generation;
1044 fs_info->last_trans_committed = generation;
1045 if (extent_buffer_uptodate(fs_info->extent_root->node) &&
1046 !(flags & OPEN_CTREE_NO_BLOCK_GROUPS))
1047 btrfs_read_block_groups(fs_info->tree_root);
1048
1049 key.objectid = BTRFS_FS_TREE_OBJECTID;
1050 key.type = BTRFS_ROOT_ITEM_KEY;
1051 key.offset = (u64)-1;
1052 fs_info->fs_root = btrfs_read_fs_root(fs_info, &key);
1053
1054 if (IS_ERR(fs_info->fs_root))
1055 return -EIO;
1056 return 0;
1057 }
1058
1059 void btrfs_release_all_roots(struct btrfs_fs_info *fs_info)
1060 {
1061 if (fs_info->quota_root)
1062 free_extent_buffer(fs_info->quota_root->node);
1063 if (fs_info->csum_root)
1064 free_extent_buffer(fs_info->csum_root->node);
1065 if (fs_info->dev_root)
1066 free_extent_buffer(fs_info->dev_root->node);
1067 if (fs_info->extent_root)
1068 free_extent_buffer(fs_info->extent_root->node);
1069 if (fs_info->tree_root)
1070 free_extent_buffer(fs_info->tree_root->node);
1071 if (fs_info->log_root_tree)
1072 free_extent_buffer(fs_info->log_root_tree->node);
1073 if (fs_info->chunk_root)
1074 free_extent_buffer(fs_info->chunk_root->node);
1075 }
1076
1077 static void free_map_lookup(struct cache_extent *ce)
1078 {
1079 struct map_lookup *map;
1080
1081 map = container_of(ce, struct map_lookup, ce);
1082 kfree(map);
1083 }
1084
1085 FREE_EXTENT_CACHE_BASED_TREE(mapping_cache, free_map_lookup);
1086
1087 void btrfs_cleanup_all_caches(struct btrfs_fs_info *fs_info)
1088 {
1089 while (!list_empty(&fs_info->recow_ebs)) {
1090 struct extent_buffer *eb;
1091 eb = list_first_entry(&fs_info->recow_ebs,
1092 struct extent_buffer, recow);
1093 list_del_init(&eb->recow);
1094 free_extent_buffer(eb);
1095 }
1096 free_mapping_cache_tree(&fs_info->mapping_tree.cache_tree);
1097 extent_io_tree_cleanup(&fs_info->extent_cache);
1098 extent_io_tree_cleanup(&fs_info->free_space_cache);
1099 extent_io_tree_cleanup(&fs_info->block_group_cache);
1100 extent_io_tree_cleanup(&fs_info->pinned_extents);
1101 extent_io_tree_cleanup(&fs_info->pending_del);
1102 extent_io_tree_cleanup(&fs_info->extent_ins);
1103 }
1104
1105 int btrfs_scan_fs_devices(int fd, const char *path,
1106 struct btrfs_fs_devices **fs_devices,
1107 u64 sb_bytenr, int super_recover,
1108 int skip_devices)
1109 {
1110 u64 total_devs;
1111 u64 dev_size;
1112 off_t seek_ret;
1113 int ret;
1114 if (!sb_bytenr)
1115 sb_bytenr = BTRFS_SUPER_INFO_OFFSET;
1116
1117 seek_ret = lseek(fd, 0, SEEK_END);
1118 if (seek_ret < 0)
1119 return -errno;
1120
1121 dev_size = seek_ret;
1122 lseek(fd, 0, SEEK_SET);
1123 if (sb_bytenr > dev_size) {
1124 fprintf(stderr, "Superblock bytenr is larger than device size\n");
1125 return -EINVAL;
1126 }
1127
1128 ret = btrfs_scan_one_device(fd, path, fs_devices,
1129 &total_devs, sb_bytenr, super_recover);
1130 if (ret) {
1131 fprintf(stderr, "No valid Btrfs found on %s\n", path);
1132 return ret;
1133 }
1134
1135 if (!skip_devices && total_devs != 1) {
1136 ret = btrfs_scan_lblkid();
1137 if (ret)
1138 return ret;
1139 }
1140 return 0;
1141 }
1142
1143 int btrfs_setup_chunk_tree_and_device_map(struct btrfs_fs_info *fs_info)
1144 {
1145 struct btrfs_super_block *sb = fs_info->super_copy;
1146 u32 sectorsize;
1147 u32 nodesize;
1148 u32 leafsize;
1149 u32 blocksize;
1150 u32 stripesize;
1151 u64 generation;
1152 int ret;
1153
1154 nodesize = btrfs_super_nodesize(sb);
1155 leafsize = btrfs_super_leafsize(sb);
1156 sectorsize = btrfs_super_sectorsize(sb);
1157 stripesize = btrfs_super_stripesize(sb);
1158
1159 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1160 fs_info->chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
1161
1162 ret = btrfs_read_sys_array(fs_info->chunk_root);
1163 if (ret)
1164 return ret;
1165
1166 blocksize = btrfs_level_size(fs_info->chunk_root,
1167 btrfs_super_chunk_root_level(sb));
1168 generation = btrfs_super_chunk_root_generation(sb);
1169
1170 fs_info->chunk_root->node = read_tree_block(fs_info->chunk_root,
1171 btrfs_super_chunk_root(sb),
1172 blocksize, generation);
1173 if (!extent_buffer_uptodate(fs_info->chunk_root->node)) {
1174 fprintf(stderr, "Couldn't read chunk root\n");
1175 return -EIO;
1176 }
1177
1178 if (!(btrfs_super_flags(sb) & BTRFS_SUPER_FLAG_METADUMP)) {
1179 ret = btrfs_read_chunk_tree(fs_info->chunk_root);
1180 if (ret) {
1181 fprintf(stderr, "Couldn't read chunk tree\n");
1182 return ret;
1183 }
1184 }
1185 return 0;
1186 }
1187
1188 static struct btrfs_fs_info *__open_ctree_fd(int fp, const char *path,
1189 u64 sb_bytenr,
1190 u64 root_tree_bytenr,
1191 enum btrfs_open_ctree_flags flags)
1192 {
1193 struct btrfs_fs_info *fs_info;
1194 struct btrfs_super_block *disk_super;
1195 struct btrfs_fs_devices *fs_devices = NULL;
1196 struct extent_buffer *eb;
1197 int ret;
1198 int oflags;
1199
1200 if (sb_bytenr == 0)
1201 sb_bytenr = BTRFS_SUPER_INFO_OFFSET;
1202
1203 /* try to drop all the caches */
1204 if (posix_fadvise(fp, 0, 0, POSIX_FADV_DONTNEED))
1205 fprintf(stderr, "Warning, could not drop caches\n");
1206
1207 fs_info = btrfs_new_fs_info(flags & OPEN_CTREE_WRITES, sb_bytenr);
1208 if (!fs_info) {
1209 fprintf(stderr, "Failed to allocate memory for fs_info\n");
1210 return NULL;
1211 }
1212 if (flags & OPEN_CTREE_RESTORE)
1213 fs_info->on_restoring = 1;
1214 if (flags & OPEN_CTREE_SUPPRESS_CHECK_BLOCK_ERRORS)
1215 fs_info->suppress_check_block_errors = 1;
1216 if (flags & OPEN_CTREE_IGNORE_FSID_MISMATCH)
1217 fs_info->ignore_fsid_mismatch = 1;
1218
1219 ret = btrfs_scan_fs_devices(fp, path, &fs_devices, sb_bytenr,
1220 (flags & OPEN_CTREE_RECOVER_SUPER),
1221 (flags & OPEN_CTREE_NO_DEVICES));
1222 if (ret)
1223 goto out;
1224
1225 fs_info->fs_devices = fs_devices;
1226 if (flags & OPEN_CTREE_WRITES)
1227 oflags = O_RDWR;
1228 else
1229 oflags = O_RDONLY;
1230
1231 if (flags & OPEN_CTREE_EXCLUSIVE)
1232 oflags |= O_EXCL;
1233
1234 ret = btrfs_open_devices(fs_devices, oflags);
1235 if (ret)
1236 goto out;
1237
1238 disk_super = fs_info->super_copy;
1239 if (!(flags & OPEN_CTREE_RECOVER_SUPER))
1240 ret = btrfs_read_dev_super(fs_devices->latest_bdev,
1241 disk_super, sb_bytenr, 1);
1242 else
1243 ret = btrfs_read_dev_super(fp, disk_super, sb_bytenr, 0);
1244 if (ret) {
1245 printk("No valid btrfs found\n");
1246 goto out_devices;
1247 }
1248
1249 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_CHANGING_FSID &&
1250 !fs_info->ignore_fsid_mismatch) {
1251 fprintf(stderr, "ERROR: Filesystem UUID change in progress\n");
1252 goto out_devices;
1253 }
1254
1255 memcpy(fs_info->fsid, &disk_super->fsid, BTRFS_FSID_SIZE);
1256
1257 ret = btrfs_check_fs_compatibility(fs_info->super_copy,
1258 flags & OPEN_CTREE_WRITES);
1259 if (ret)
1260 goto out_devices;
1261
1262 ret = btrfs_setup_chunk_tree_and_device_map(fs_info);
1263 if (ret)
1264 goto out_chunk;
1265
1266 eb = fs_info->chunk_root->node;
1267 read_extent_buffer(eb, fs_info->chunk_tree_uuid,
1268 btrfs_header_chunk_tree_uuid(eb),
1269 BTRFS_UUID_SIZE);
1270
1271 ret = btrfs_setup_all_roots(fs_info, root_tree_bytenr, flags);
1272 if (ret && !(flags & __OPEN_CTREE_RETURN_CHUNK_ROOT))
1273 goto out_chunk;
1274
1275 return fs_info;
1276
1277 out_chunk:
1278 btrfs_release_all_roots(fs_info);
1279 btrfs_cleanup_all_caches(fs_info);
1280 out_devices:
1281 btrfs_close_devices(fs_devices);
1282 out:
1283 btrfs_free_fs_info(fs_info);
1284 return NULL;
1285 }
1286
1287 struct btrfs_fs_info *open_ctree_fs_info(const char *filename,
1288 u64 sb_bytenr, u64 root_tree_bytenr,
1289 enum btrfs_open_ctree_flags flags)
1290 {
1291 int fp;
1292 struct btrfs_fs_info *info;
1293 int oflags = O_CREAT | O_RDWR;
1294
1295 if (!(flags & OPEN_CTREE_WRITES))
1296 oflags = O_RDONLY;
1297
1298 fp = open(filename, oflags, 0600);
1299 if (fp < 0) {
1300 fprintf (stderr, "Could not open %s\n", filename);
1301 return NULL;
1302 }
1303 info = __open_ctree_fd(fp, filename, sb_bytenr, root_tree_bytenr,
1304 flags);
1305 close(fp);
1306 return info;
1307 }
1308
1309 struct btrfs_root *open_ctree(const char *filename, u64 sb_bytenr,
1310 enum btrfs_open_ctree_flags flags)
1311 {
1312 struct btrfs_fs_info *info;
1313
1314 info = open_ctree_fs_info(filename, sb_bytenr, 0, flags);
1315 if (!info)
1316 return NULL;
1317 if (flags & __OPEN_CTREE_RETURN_CHUNK_ROOT)
1318 return info->chunk_root;
1319 return info->fs_root;
1320 }
1321
1322 struct btrfs_root *open_ctree_fd(int fp, const char *path, u64 sb_bytenr,
1323 enum btrfs_open_ctree_flags flags)
1324 {
1325 struct btrfs_fs_info *info;
1326 info = __open_ctree_fd(fp, path, sb_bytenr, 0, flags);
1327 if (!info)
1328 return NULL;
1329 if (flags & __OPEN_CTREE_RETURN_CHUNK_ROOT)
1330 return info->chunk_root;
1331 return info->fs_root;
1332 }
1333
1334 int btrfs_read_dev_super(int fd, struct btrfs_super_block *sb, u64 sb_bytenr,
1335 int super_recover)
1336 {
1337 u8 fsid[BTRFS_FSID_SIZE];
1338 int fsid_is_initialized = 0;
1339 struct btrfs_super_block buf;
1340 int i;
1341 int ret;
1342 int max_super = super_recover ? BTRFS_SUPER_MIRROR_MAX : 1;
1343 u64 transid = 0;
1344 u64 bytenr;
1345
1346 if (sb_bytenr != BTRFS_SUPER_INFO_OFFSET) {
1347 ret = pread64(fd, &buf, sizeof(buf), sb_bytenr);
1348 if (ret < sizeof(buf))
1349 return -1;
1350
1351 if (btrfs_super_bytenr(&buf) != sb_bytenr ||
1352 btrfs_super_magic(&buf) != BTRFS_MAGIC)
1353 return -1;
1354
1355 memcpy(sb, &buf, sizeof(*sb));
1356 return 0;
1357 }
1358
1359 /*
1360 * we would like to check all the supers, but that would make
1361 * a btrfs mount succeed after a mkfs from a different FS.
1362 * So, we need to add a special mount option to scan for
1363 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
1364 */
1365
1366 for (i = 0; i < max_super; i++) {
1367 bytenr = btrfs_sb_offset(i);
1368 ret = pread64(fd, &buf, sizeof(buf), bytenr);
1369 if (ret < sizeof(buf))
1370 break;
1371
1372 if (btrfs_super_bytenr(&buf) != bytenr )
1373 continue;
1374 /* if magic is NULL, the device was removed */
1375 if (btrfs_super_magic(&buf) == 0 && i == 0)
1376 return -1;
1377 if (btrfs_super_magic(&buf) != BTRFS_MAGIC)
1378 continue;
1379
1380 if (!fsid_is_initialized) {
1381 memcpy(fsid, buf.fsid, sizeof(fsid));
1382 fsid_is_initialized = 1;
1383 } else if (memcmp(fsid, buf.fsid, sizeof(fsid))) {
1384 /*
1385 * the superblocks (the original one and
1386 * its backups) contain data of different
1387 * filesystems -> the super cannot be trusted
1388 */
1389 continue;
1390 }
1391
1392 if (btrfs_super_generation(&buf) > transid) {
1393 memcpy(sb, &buf, sizeof(*sb));
1394 transid = btrfs_super_generation(&buf);
1395 }
1396 }
1397
1398 return transid > 0 ? 0 : -1;
1399 }
1400
1401 static int write_dev_supers(struct btrfs_root *root,
1402 struct btrfs_super_block *sb,
1403 struct btrfs_device *device)
1404 {
1405 u64 bytenr;
1406 u32 crc;
1407 int i, ret;
1408
1409 if (root->fs_info->super_bytenr != BTRFS_SUPER_INFO_OFFSET) {
1410 btrfs_set_super_bytenr(sb, root->fs_info->super_bytenr);
1411 crc = ~(u32)0;
1412 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
1413 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
1414 btrfs_csum_final(crc, (char *)&sb->csum[0]);
1415
1416 /*
1417 * super_copy is BTRFS_SUPER_INFO_SIZE bytes and is
1418 * zero filled, we can use it directly
1419 */
1420 ret = pwrite64(device->fd, root->fs_info->super_copy,
1421 BTRFS_SUPER_INFO_SIZE,
1422 root->fs_info->super_bytenr);
1423 BUG_ON(ret != BTRFS_SUPER_INFO_SIZE);
1424 return 0;
1425 }
1426
1427 for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
1428 bytenr = btrfs_sb_offset(i);
1429 if (bytenr + BTRFS_SUPER_INFO_SIZE > device->total_bytes)
1430 break;
1431
1432 btrfs_set_super_bytenr(sb, bytenr);
1433
1434 crc = ~(u32)0;
1435 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
1436 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
1437 btrfs_csum_final(crc, (char *)&sb->csum[0]);
1438
1439 /*
1440 * super_copy is BTRFS_SUPER_INFO_SIZE bytes and is
1441 * zero filled, we can use it directly
1442 */
1443 ret = pwrite64(device->fd, root->fs_info->super_copy,
1444 BTRFS_SUPER_INFO_SIZE, bytenr);
1445 BUG_ON(ret != BTRFS_SUPER_INFO_SIZE);
1446 }
1447
1448 return 0;
1449 }
1450
1451 int write_all_supers(struct btrfs_root *root)
1452 {
1453 struct list_head *cur;
1454 struct list_head *head = &root->fs_info->fs_devices->devices;
1455 struct btrfs_device *dev;
1456 struct btrfs_super_block *sb;
1457 struct btrfs_dev_item *dev_item;
1458 int ret;
1459 u64 flags;
1460
1461 sb = root->fs_info->super_copy;
1462 dev_item = &sb->dev_item;
1463 list_for_each(cur, head) {
1464 dev = list_entry(cur, struct btrfs_device, dev_list);
1465 if (!dev->writeable)
1466 continue;
1467
1468 btrfs_set_stack_device_generation(dev_item, 0);
1469 btrfs_set_stack_device_type(dev_item, dev->type);
1470 btrfs_set_stack_device_id(dev_item, dev->devid);
1471 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
1472 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
1473 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
1474 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
1475 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
1476 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
1477 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
1478
1479 flags = btrfs_super_flags(sb);
1480 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
1481
1482 ret = write_dev_supers(root, sb, dev);
1483 BUG_ON(ret);
1484 }
1485 return 0;
1486 }
1487
1488 int write_ctree_super(struct btrfs_trans_handle *trans,
1489 struct btrfs_root *root)
1490 {
1491 int ret;
1492 struct btrfs_root *tree_root = root->fs_info->tree_root;
1493 struct btrfs_root *chunk_root = root->fs_info->chunk_root;
1494
1495 if (root->fs_info->readonly)
1496 return 0;
1497
1498 btrfs_set_super_generation(root->fs_info->super_copy,
1499 trans->transid);
1500 btrfs_set_super_root(root->fs_info->super_copy,
1501 tree_root->node->start);
1502 btrfs_set_super_root_level(root->fs_info->super_copy,
1503 btrfs_header_level(tree_root->node));
1504 btrfs_set_super_chunk_root(root->fs_info->super_copy,
1505 chunk_root->node->start);
1506 btrfs_set_super_chunk_root_level(root->fs_info->super_copy,
1507 btrfs_header_level(chunk_root->node));
1508 btrfs_set_super_chunk_root_generation(root->fs_info->super_copy,
1509 btrfs_header_generation(chunk_root->node));
1510
1511 ret = write_all_supers(root);
1512 if (ret)
1513 fprintf(stderr, "failed to write new super block err %d\n", ret);
1514 return ret;
1515 }
1516
1517 int close_ctree(struct btrfs_root *root)
1518 {
1519 int ret;
1520 struct btrfs_trans_handle *trans;
1521 struct btrfs_fs_info *fs_info = root->fs_info;
1522
1523 if (fs_info->last_trans_committed !=
1524 fs_info->generation) {
1525 trans = btrfs_start_transaction(root, 1);
1526 btrfs_commit_transaction(trans, root);
1527 trans = btrfs_start_transaction(root, 1);
1528 ret = commit_tree_roots(trans, fs_info);
1529 BUG_ON(ret);
1530 ret = __commit_transaction(trans, root);
1531 BUG_ON(ret);
1532 write_ctree_super(trans, root);
1533 btrfs_free_transaction(root, trans);
1534 }
1535 btrfs_free_block_groups(fs_info);
1536
1537 free_fs_roots_tree(&fs_info->fs_root_tree);
1538
1539 btrfs_release_all_roots(fs_info);
1540 btrfs_close_devices(fs_info->fs_devices);
1541 btrfs_cleanup_all_caches(fs_info);
1542 btrfs_free_fs_info(fs_info);
1543 return 0;
1544 }
1545
1546 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1547 struct extent_buffer *eb)
1548 {
1549 return clear_extent_buffer_dirty(eb);
1550 }
1551
1552 int wait_on_tree_block_writeback(struct btrfs_root *root,
1553 struct extent_buffer *eb)
1554 {
1555 return 0;
1556 }
1557
1558 void btrfs_mark_buffer_dirty(struct extent_buffer *eb)
1559 {
1560 set_extent_buffer_dirty(eb);
1561 }
1562
1563 int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
1564 {
1565 int ret;
1566
1567 ret = extent_buffer_uptodate(buf);
1568 if (!ret)
1569 return ret;
1570
1571 ret = verify_parent_transid(buf->tree, buf, parent_transid, 1);
1572 return !ret;
1573 }
1574
1575 int btrfs_set_buffer_uptodate(struct extent_buffer *eb)
1576 {
1577 return set_extent_buffer_uptodate(eb);
1578 }
(deprecated) Btrfs progs developments and patch integration