?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
btrfs-progs: update CHANGES for 4.7.3
[btrfs-progs-unstable/devel.git] / disk-io.c
blob2fd3330f2c0e6e3656f8df3626a65a0f9e175be6
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_fs_info *fs_info,
55 struct extent_buffer *buf)
56 {
57
58 struct btrfs_fs_devices *fs_devices;
59 u32 nodesize = btrfs_super_nodesize(fs_info->super_copy);
60 int ret = BTRFS_BAD_FSID;
61
62 if (buf->start != btrfs_header_bytenr(buf))
63 return BTRFS_BAD_BYTENR;
64 if (btrfs_header_level(buf) >= BTRFS_MAX_LEVEL)
65 return BTRFS_BAD_LEVEL;
66 if (btrfs_header_nritems(buf) > max_nritems(btrfs_header_level(buf),
67 nodesize))
68 return BTRFS_BAD_NRITEMS;
69
70 /* Only leaf can be empty */
71 if (btrfs_header_nritems(buf) == 0 &&
72 btrfs_header_level(buf) != 0)
73 return BTRFS_BAD_NRITEMS;
74
75 fs_devices = fs_info->fs_devices;
76 while (fs_devices) {
77 if (fs_info->ignore_fsid_mismatch ||
78 !memcmp_extent_buffer(buf, fs_devices->fsid,
79 btrfs_header_fsid(),
80 BTRFS_FSID_SIZE)) {
81 ret = 0;
82 break;
83 }
84 fs_devices = fs_devices->seed;
85 }
86 return ret;
87 }
88
89 static void print_tree_block_error(struct btrfs_fs_info *fs_info,
90 struct extent_buffer *eb,
91 int err)
92 {
93 char fs_uuid[BTRFS_UUID_UNPARSED_SIZE] = {'\0'};
94 char found_uuid[BTRFS_UUID_UNPARSED_SIZE] = {'\0'};
95 u8 buf[BTRFS_UUID_SIZE];
96
97 switch (err) {
98 case BTRFS_BAD_FSID:
99 read_extent_buffer(eb, buf, btrfs_header_fsid(),
100 BTRFS_UUID_SIZE);
101 uuid_unparse(buf, found_uuid);
102 uuid_unparse(fs_info->fsid, fs_uuid);
103 fprintf(stderr, "fsid mismatch, want=%s, have=%s\n",
104 fs_uuid, found_uuid);
105 break;
106 case BTRFS_BAD_BYTENR:
107 fprintf(stderr, "bytenr mismatch, want=%llu, have=%llu\n",
108 eb->start, btrfs_header_bytenr(eb));
109 break;
110 case BTRFS_BAD_LEVEL:
111 fprintf(stderr, "bad level, %u > %u\n",
112 btrfs_header_level(eb), BTRFS_MAX_LEVEL);
113 break;
114 case BTRFS_BAD_NRITEMS:
115 fprintf(stderr, "invalid nr_items: %u\n",
116 btrfs_header_nritems(eb));
117 break;
118 }
119 }
120
121 u32 btrfs_csum_data(struct btrfs_root *root, char *data, u32 seed, size_t len)
122 {
123 return crc32c(seed, data, len);
124 }
125
126 void btrfs_csum_final(u32 crc, char *result)
127 {
128 put_unaligned_le32(~crc, result);
129 }
130
131 static int __csum_tree_block_size(struct extent_buffer *buf, u16 csum_size,
132 int verify, int silent)
133 {
134 char result[BTRFS_CSUM_SIZE];
135 u32 len;
136 u32 crc = ~(u32)0;
137
138 len = buf->len - BTRFS_CSUM_SIZE;
139 crc = crc32c(crc, buf->data + BTRFS_CSUM_SIZE, len);
140 btrfs_csum_final(crc, result);
141
142 if (verify) {
143 if (memcmp_extent_buffer(buf, result, 0, csum_size)) {
144 if (!silent)
145 printk("checksum verify failed on %llu found %08X wanted %08X\n",
146 (unsigned long long)buf->start,
147 *((u32 *)result),
148 *((u32*)(char *)buf->data));
149 return 1;
150 }
151 } else {
152 write_extent_buffer(buf, result, 0, csum_size);
153 }
154 return 0;
155 }
156
157 int csum_tree_block_size(struct extent_buffer *buf, u16 csum_size, int verify)
158 {
159 return __csum_tree_block_size(buf, csum_size, verify, 0);
160 }
161
162 int verify_tree_block_csum_silent(struct extent_buffer *buf, u16 csum_size)
163 {
164 return __csum_tree_block_size(buf, csum_size, 1, 1);
165 }
166
167 static int csum_tree_block_fs_info(struct btrfs_fs_info *fs_info,
168 struct extent_buffer *buf, int verify)
169 {
170 u16 csum_size =
171 btrfs_super_csum_size(fs_info->super_copy);
172 if (verify && fs_info->suppress_check_block_errors)
173 return verify_tree_block_csum_silent(buf, csum_size);
174 return csum_tree_block_size(buf, csum_size, verify);
175 }
176
177 int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
178 int verify)
179 {
180 return csum_tree_block_fs_info(root->fs_info, buf, verify);
181 }
182
183 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
184 u64 bytenr, u32 blocksize)
185 {
186 return find_extent_buffer(&root->fs_info->extent_cache,
187 bytenr, blocksize);
188 }
189
190 struct extent_buffer* btrfs_find_create_tree_block(
191 struct btrfs_fs_info *fs_info, u64 bytenr, u32 blocksize)
192 {
193 return alloc_extent_buffer(&fs_info->extent_cache, bytenr, blocksize);
194 }
195
196 void readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
197 u64 parent_transid)
198 {
199 struct extent_buffer *eb;
200 u64 length;
201 struct btrfs_multi_bio *multi = NULL;
202 struct btrfs_device *device;
203
204 eb = btrfs_find_tree_block(root, bytenr, blocksize);
205 if (!(eb && btrfs_buffer_uptodate(eb, parent_transid)) &&
206 !btrfs_map_block(&root->fs_info->mapping_tree, READ,
207 bytenr, &length, &multi, 0, NULL)) {
208 device = multi->stripes[0].dev;
209 device->total_ios++;
210 blocksize = min(blocksize, (u32)(64 * 1024));
211 readahead(device->fd, multi->stripes[0].physical, blocksize);
212 }
213
214 free_extent_buffer(eb);
215 kfree(multi);
216 }
217
218 static int verify_parent_transid(struct extent_io_tree *io_tree,
219 struct extent_buffer *eb, u64 parent_transid,
220 int ignore)
221 {
222 int ret;
223
224 if (!parent_transid || btrfs_header_generation(eb) == parent_transid)
225 return 0;
226
227 if (extent_buffer_uptodate(eb) &&
228 btrfs_header_generation(eb) == parent_transid) {
229 ret = 0;
230 goto out;
231 }
232 printk("parent transid verify failed on %llu wanted %llu found %llu\n",
233 (unsigned long long)eb->start,
234 (unsigned long long)parent_transid,
235 (unsigned long long)btrfs_header_generation(eb));
236 if (ignore) {
237 eb->flags |= EXTENT_BAD_TRANSID;
238 printk("Ignoring transid failure\n");
239 return 0;
240 }
241
242 ret = 1;
243 out:
244 clear_extent_buffer_uptodate(io_tree, eb);
245 return ret;
246
247 }
248
249
250 int read_whole_eb(struct btrfs_fs_info *info, struct extent_buffer *eb, int mirror)
251 {
252 unsigned long offset = 0;
253 struct btrfs_multi_bio *multi = NULL;
254 struct btrfs_device *device;
255 int ret = 0;
256 u64 read_len;
257 unsigned long bytes_left = eb->len;
258
259 while (bytes_left) {
260 read_len = bytes_left;
261 device = NULL;
262
263 if (!info->on_restoring &&
264 eb->start != BTRFS_SUPER_INFO_OFFSET) {
265 ret = btrfs_map_block(&info->mapping_tree, READ,
266 eb->start + offset, &read_len, &multi,
267 mirror, NULL);
268 if (ret) {
269 printk("Couldn't map the block %Lu\n", eb->start + offset);
270 kfree(multi);
271 return -EIO;
272 }
273 device = multi->stripes[0].dev;
274
275 if (device->fd <= 0) {
276 kfree(multi);
277 return -EIO;
278 }
279
280 eb->fd = device->fd;
281 device->total_ios++;
282 eb->dev_bytenr = multi->stripes[0].physical;
283 kfree(multi);
284 multi = NULL;
285 } else {
286 /* special case for restore metadump */
287 list_for_each_entry(device, &info->fs_devices->devices, dev_list) {
288 if (device->devid == 1)
289 break;
290 }
291
292 eb->fd = device->fd;
293 eb->dev_bytenr = eb->start;
294 device->total_ios++;
295 }
296
297 if (read_len > bytes_left)
298 read_len = bytes_left;
299
300 ret = read_extent_from_disk(eb, offset, read_len);
301 if (ret)
302 return -EIO;
303 offset += read_len;
304 bytes_left -= read_len;
305 }
306 return 0;
307 }
308
309 struct extent_buffer* read_tree_block_fs_info(
310 struct btrfs_fs_info *fs_info, u64 bytenr, u32 blocksize,
311 u64 parent_transid)
312 {
313 int ret;
314 struct extent_buffer *eb;
315 u64 best_transid = 0;
316 u32 sectorsize = btrfs_super_sectorsize(fs_info->super_copy);
317 u32 nodesize = btrfs_super_nodesize(fs_info->super_copy);
318 int mirror_num = 0;
319 int good_mirror = 0;
320 int num_copies;
321 int ignore = 0;
322
323 /*
324 * Don't even try to create tree block for unaligned tree block
325 * bytenr.
326 * Such unaligned tree block will free overlapping extent buffer,
327 * causing use-after-free bugs for fuzzed images.
328 */
329 if (!IS_ALIGNED(bytenr, sectorsize)) {
330 error("tree block bytenr %llu is not aligned to sectorsize %u",
331 bytenr, sectorsize);
332 return ERR_PTR(-EIO);
333 }
334 if (!IS_ALIGNED(blocksize, nodesize)) {
335 error("tree block size %u is not aligned to nodesize %u",
336 blocksize, nodesize);
337 return ERR_PTR(-EIO);
338 }
339 eb = btrfs_find_create_tree_block(fs_info, bytenr, blocksize);
340 if (!eb)
341 return ERR_PTR(-ENOMEM);
342
343 if (btrfs_buffer_uptodate(eb, parent_transid))
344 return eb;
345
346 while (1) {
347 ret = read_whole_eb(fs_info, eb, mirror_num);
348 if (ret == 0 && csum_tree_block_fs_info(fs_info, eb, 1) == 0 &&
349 check_tree_block(fs_info, eb) == 0 &&
350 verify_parent_transid(eb->tree, eb, parent_transid, ignore)
351 == 0) {
352 if (eb->flags & EXTENT_BAD_TRANSID &&
353 list_empty(&eb->recow)) {
354 list_add_tail(&eb->recow,
355 &fs_info->recow_ebs);
356 eb->refs++;
357 }
358 btrfs_set_buffer_uptodate(eb);
359 return eb;
360 }
361 if (ignore) {
362 if (check_tree_block(fs_info, eb)) {
363 if (!fs_info->suppress_check_block_errors)
364 print_tree_block_error(fs_info, eb,
365 check_tree_block(fs_info, eb));
366 } else {
367 if (!fs_info->suppress_check_block_errors)
368 fprintf(stderr, "Csum didn't match\n");
369 }
370 ret = -EIO;
371 break;
372 }
373 num_copies = btrfs_num_copies(&fs_info->mapping_tree,
374 eb->start, eb->len);
375 if (num_copies == 1) {
376 ignore = 1;
377 continue;
378 }
379 if (btrfs_header_generation(eb) > best_transid && mirror_num) {
380 best_transid = btrfs_header_generation(eb);
381 good_mirror = mirror_num;
382 }
383 mirror_num++;
384 if (mirror_num > num_copies) {
385 mirror_num = good_mirror;
386 ignore = 1;
387 continue;
388 }
389 }
390 free_extent_buffer(eb);
391 return ERR_PTR(ret);
392 }
393
394 int read_extent_data(struct btrfs_root *root, char *data,
395 u64 logical, u64 *len, int mirror)
396 {
397 u64 offset = 0;
398 struct btrfs_multi_bio *multi = NULL;
399 struct btrfs_fs_info *info = root->fs_info;
400 struct btrfs_device *device;
401 int ret = 0;
402 u64 max_len = *len;
403
404 ret = btrfs_map_block(&info->mapping_tree, READ, logical, len,
405 &multi, mirror, NULL);
406 if (ret) {
407 fprintf(stderr, "Couldn't map the block %llu\n",
408 logical + offset);
409 goto err;
410 }
411 device = multi->stripes[0].dev;
412
413 if (device->fd <= 0)
414 goto err;
415 if (*len > max_len)
416 *len = max_len;
417
418 ret = pread64(device->fd, data, *len, multi->stripes[0].physical);
419 if (ret != *len)
420 ret = -EIO;
421 else
422 ret = 0;
423 err:
424 kfree(multi);
425 return ret;
426 }
427
428 int write_and_map_eb(struct btrfs_trans_handle *trans,
429 struct btrfs_root *root,
430 struct extent_buffer *eb)
431 {
432 int ret;
433 int dev_nr;
434 u64 length;
435 u64 *raid_map = NULL;
436 struct btrfs_multi_bio *multi = NULL;
437
438 dev_nr = 0;
439 length = eb->len;
440 ret = btrfs_map_block(&root->fs_info->mapping_tree, WRITE,
441 eb->start, &length, &multi, 0, &raid_map);
442
443 if (raid_map) {
444 ret = write_raid56_with_parity(root->fs_info, eb, multi,
445 length, raid_map);
446 BUG_ON(ret);
447 } else while (dev_nr < multi->num_stripes) {
448 BUG_ON(ret);
449 eb->fd = multi->stripes[dev_nr].dev->fd;
450 eb->dev_bytenr = multi->stripes[dev_nr].physical;
451 multi->stripes[dev_nr].dev->total_ios++;
452 dev_nr++;
453 ret = write_extent_to_disk(eb);
454 BUG_ON(ret);
455 }
456 kfree(raid_map);
457 kfree(multi);
458 return 0;
459 }
460
461 int write_tree_block(struct btrfs_trans_handle *trans,
462 struct btrfs_root *root,
463 struct extent_buffer *eb)
464 {
465 if (check_tree_block(root->fs_info, eb)) {
466 print_tree_block_error(root->fs_info, eb,
467 check_tree_block(root->fs_info, eb));
468 BUG();
469 }
470
471 if (trans && !btrfs_buffer_uptodate(eb, trans->transid))
472 BUG();
473
474 btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
475 csum_tree_block(root, eb, 0);
476
477 return write_and_map_eb(trans, root, eb);
478 }
479
480 int __setup_root(u32 nodesize, u32 leafsize, u32 sectorsize,
481 u32 stripesize, struct btrfs_root *root,
482 struct btrfs_fs_info *fs_info, u64 objectid)
483 {
484 root->node = NULL;
485 root->commit_root = NULL;
486 root->sectorsize = sectorsize;
487 root->nodesize = nodesize;
488 root->leafsize = leafsize;
489 root->stripesize = stripesize;
490 root->ref_cows = 0;
491 root->track_dirty = 0;
492
493 root->fs_info = fs_info;
494 root->objectid = objectid;
495 root->last_trans = 0;
496 root->highest_inode = 0;
497 root->last_inode_alloc = 0;
498
499 INIT_LIST_HEAD(&root->dirty_list);
500 INIT_LIST_HEAD(&root->orphan_data_extents);
501 memset(&root->root_key, 0, sizeof(root->root_key));
502 memset(&root->root_item, 0, sizeof(root->root_item));
503 root->root_key.objectid = objectid;
504 return 0;
505 }
506
507 static int update_cowonly_root(struct btrfs_trans_handle *trans,
508 struct btrfs_root *root)
509 {
510 int ret;
511 u64 old_root_bytenr;
512 struct btrfs_root *tree_root = root->fs_info->tree_root;
513
514 btrfs_write_dirty_block_groups(trans, root);
515 while(1) {
516 old_root_bytenr = btrfs_root_bytenr(&root->root_item);
517 if (old_root_bytenr == root->node->start)
518 break;
519 btrfs_set_root_bytenr(&root->root_item,
520 root->node->start);
521 btrfs_set_root_generation(&root->root_item,
522 trans->transid);
523 root->root_item.level = btrfs_header_level(root->node);
524 ret = btrfs_update_root(trans, tree_root,
525 &root->root_key,
526 &root->root_item);
527 BUG_ON(ret);
528 btrfs_write_dirty_block_groups(trans, root);
529 }
530 return 0;
531 }
532
533 static int commit_tree_roots(struct btrfs_trans_handle *trans,
534 struct btrfs_fs_info *fs_info)
535 {
536 struct btrfs_root *root;
537 struct list_head *next;
538 struct extent_buffer *eb;
539 int ret;
540
541 if (fs_info->readonly)
542 return 0;
543
544 eb = fs_info->tree_root->node;
545 extent_buffer_get(eb);
546 ret = btrfs_cow_block(trans, fs_info->tree_root, eb, NULL, 0, &eb);
547 free_extent_buffer(eb);
548 if (ret)
549 return ret;
550
551 while(!list_empty(&fs_info->dirty_cowonly_roots)) {
552 next = fs_info->dirty_cowonly_roots.next;
553 list_del_init(next);
554 root = list_entry(next, struct btrfs_root, dirty_list);
555 update_cowonly_root(trans, root);
556 free_extent_buffer(root->commit_root);
557 root->commit_root = NULL;
558 }
559
560 return 0;
561 }
562
563 static int __commit_transaction(struct btrfs_trans_handle *trans,
564 struct btrfs_root *root)
565 {
566 u64 start;
567 u64 end;
568 struct extent_buffer *eb;
569 struct extent_io_tree *tree = &root->fs_info->extent_cache;
570 int ret;
571
572 while(1) {
573 ret = find_first_extent_bit(tree, 0, &start, &end,
574 EXTENT_DIRTY);
575 if (ret)
576 break;
577 while(start <= end) {
578 eb = find_first_extent_buffer(tree, start);
579 BUG_ON(!eb || eb->start != start);
580 ret = write_tree_block(trans, root, eb);
581 BUG_ON(ret);
582 start += eb->len;
583 clear_extent_buffer_dirty(eb);
584 free_extent_buffer(eb);
585 }
586 }
587 return 0;
588 }
589
590 int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
591 struct btrfs_root *root)
592 {
593 u64 transid = trans->transid;
594 int ret = 0;
595 struct btrfs_fs_info *fs_info = root->fs_info;
596
597 if (root->commit_root == root->node)
598 goto commit_tree;
599 if (root == root->fs_info->tree_root)
600 goto commit_tree;
601 if (root == root->fs_info->chunk_root)
602 goto commit_tree;
603
604 free_extent_buffer(root->commit_root);
605 root->commit_root = NULL;
606
607 btrfs_set_root_bytenr(&root->root_item, root->node->start);
608 btrfs_set_root_generation(&root->root_item, trans->transid);
609 root->root_item.level = btrfs_header_level(root->node);
610 ret = btrfs_update_root(trans, root->fs_info->tree_root,
611 &root->root_key, &root->root_item);
612 BUG_ON(ret);
613 commit_tree:
614 ret = commit_tree_roots(trans, fs_info);
615 BUG_ON(ret);
616 ret = __commit_transaction(trans, root);
617 BUG_ON(ret);
618 write_ctree_super(trans, root);
619 btrfs_finish_extent_commit(trans, fs_info->extent_root,
620 &fs_info->pinned_extents);
621 btrfs_free_transaction(root, trans);
622 free_extent_buffer(root->commit_root);
623 root->commit_root = NULL;
624 fs_info->running_transaction = NULL;
625 fs_info->last_trans_committed = transid;
626 return 0;
627 }
628
629 static int find_and_setup_root(struct btrfs_root *tree_root,
630 struct btrfs_fs_info *fs_info,
631 u64 objectid, struct btrfs_root *root)
632 {
633 int ret;
634 u32 blocksize;
635 u64 generation;
636
637 __setup_root(tree_root->nodesize, tree_root->leafsize,
638 tree_root->sectorsize, tree_root->stripesize,
639 root, fs_info, objectid);
640 ret = btrfs_find_last_root(tree_root, objectid,
641 &root->root_item, &root->root_key);
642 if (ret)
643 return ret;
644
645 blocksize = root->nodesize;
646 generation = btrfs_root_generation(&root->root_item);
647 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
648 blocksize, generation);
649 if (!extent_buffer_uptodate(root->node))
650 return -EIO;
651
652 return 0;
653 }
654
655 static int find_and_setup_log_root(struct btrfs_root *tree_root,
656 struct btrfs_fs_info *fs_info,
657 struct btrfs_super_block *disk_super)
658 {
659 u32 blocksize;
660 u64 blocknr = btrfs_super_log_root(disk_super);
661 struct btrfs_root *log_root = malloc(sizeof(struct btrfs_root));
662
663 if (!log_root)
664 return -ENOMEM;
665
666 if (blocknr == 0) {
667 free(log_root);
668 return 0;
669 }
670
671 blocksize = tree_root->nodesize;
672
673 __setup_root(tree_root->nodesize, tree_root->leafsize,
674 tree_root->sectorsize, tree_root->stripesize,
675 log_root, fs_info, BTRFS_TREE_LOG_OBJECTID);
676
677 log_root->node = read_tree_block(tree_root, blocknr,
678 blocksize,
679 btrfs_super_generation(disk_super) + 1);
680
681 fs_info->log_root_tree = log_root;
682
683 if (!extent_buffer_uptodate(log_root->node)) {
684 free_extent_buffer(log_root->node);
685 free(log_root);
686 fs_info->log_root_tree = NULL;
687 return -EIO;
688 }
689
690 return 0;
691 }
692
693 int btrfs_free_fs_root(struct btrfs_root *root)
694 {
695 if (root->node)
696 free_extent_buffer(root->node);
697 if (root->commit_root)
698 free_extent_buffer(root->commit_root);
699 kfree(root);
700 return 0;
701 }
702
703 static void __free_fs_root(struct rb_node *node)
704 {
705 struct btrfs_root *root;
706
707 root = container_of(node, struct btrfs_root, rb_node);
708 btrfs_free_fs_root(root);
709 }
710
711 FREE_RB_BASED_TREE(fs_roots, __free_fs_root);
712
713 struct btrfs_root *btrfs_read_fs_root_no_cache(struct btrfs_fs_info *fs_info,
714 struct btrfs_key *location)
715 {
716 struct btrfs_root *root;
717 struct btrfs_root *tree_root = fs_info->tree_root;
718 struct btrfs_path *path;
719 struct extent_buffer *l;
720 u64 generation;
721 u32 blocksize;
722 int ret = 0;
723
724 root = calloc(1, sizeof(*root));
725 if (!root)
726 return ERR_PTR(-ENOMEM);
727 if (location->offset == (u64)-1) {
728 ret = find_and_setup_root(tree_root, fs_info,
729 location->objectid, root);
730 if (ret) {
731 free(root);
732 return ERR_PTR(ret);
733 }
734 goto insert;
735 }
736
737 __setup_root(tree_root->nodesize, tree_root->leafsize,
738 tree_root->sectorsize, tree_root->stripesize,
739 root, fs_info, location->objectid);
740
741 path = btrfs_alloc_path();
742 BUG_ON(!path);
743 ret = btrfs_search_slot(NULL, tree_root, location, path, 0, 0);
744 if (ret != 0) {
745 if (ret > 0)
746 ret = -ENOENT;
747 goto out;
748 }
749 l = path->nodes[0];
750 read_extent_buffer(l, &root->root_item,
751 btrfs_item_ptr_offset(l, path->slots[0]),
752 sizeof(root->root_item));
753 memcpy(&root->root_key, location, sizeof(*location));
754 ret = 0;
755 out:
756 btrfs_free_path(path);
757 if (ret) {
758 free(root);
759 return ERR_PTR(ret);
760 }
761 generation = btrfs_root_generation(&root->root_item);
762 blocksize = root->nodesize;
763 root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
764 blocksize, generation);
765 if (!extent_buffer_uptodate(root->node)) {
766 free(root);
767 return ERR_PTR(-EIO);
768 }
769 insert:
770 root->ref_cows = 1;
771 return root;
772 }
773
774 static int btrfs_fs_roots_compare_objectids(struct rb_node *node,
775 void *data)
776 {
777 u64 objectid = *((u64 *)data);
778 struct btrfs_root *root;
779
780 root = rb_entry(node, struct btrfs_root, rb_node);
781 if (objectid > root->objectid)
782 return 1;
783 else if (objectid < root->objectid)
784 return -1;
785 else
786 return 0;
787 }
788
789 static int btrfs_fs_roots_compare_roots(struct rb_node *node1,
790 struct rb_node *node2)
791 {
792 struct btrfs_root *root;
793
794 root = rb_entry(node2, struct btrfs_root, rb_node);
795 return btrfs_fs_roots_compare_objectids(node1, (void *)&root->objectid);
796 }
797
798 struct btrfs_root *btrfs_read_fs_root(struct btrfs_fs_info *fs_info,
799 struct btrfs_key *location)
800 {
801 struct btrfs_root *root;
802 struct rb_node *node;
803 int ret;
804 u64 objectid = location->objectid;
805
806 if (location->objectid == BTRFS_ROOT_TREE_OBJECTID)
807 return fs_info->tree_root;
808 if (location->objectid == BTRFS_EXTENT_TREE_OBJECTID)
809 return fs_info->extent_root;
810 if (location->objectid == BTRFS_CHUNK_TREE_OBJECTID)
811 return fs_info->chunk_root;
812 if (location->objectid == BTRFS_DEV_TREE_OBJECTID)
813 return fs_info->dev_root;
814 if (location->objectid == BTRFS_CSUM_TREE_OBJECTID)
815 return fs_info->csum_root;
816 if (location->objectid == BTRFS_QUOTA_TREE_OBJECTID)
817 return fs_info->quota_root;
818
819 BUG_ON(location->objectid == BTRFS_TREE_RELOC_OBJECTID ||
820 location->offset != (u64)-1);
821
822 node = rb_search(&fs_info->fs_root_tree, (void *)&objectid,
823 btrfs_fs_roots_compare_objectids, NULL);
824 if (node)
825 return container_of(node, struct btrfs_root, rb_node);
826
827 root = btrfs_read_fs_root_no_cache(fs_info, location);
828 if (IS_ERR(root))
829 return root;
830
831 ret = rb_insert(&fs_info->fs_root_tree, &root->rb_node,
832 btrfs_fs_roots_compare_roots);
833 BUG_ON(ret);
834 return root;
835 }
836
837 void btrfs_free_fs_info(struct btrfs_fs_info *fs_info)
838 {
839 free(fs_info->tree_root);
840 free(fs_info->extent_root);
841 free(fs_info->chunk_root);
842 free(fs_info->dev_root);
843 free(fs_info->csum_root);
844 free(fs_info->quota_root);
845 free(fs_info->free_space_root);
846 free(fs_info->super_copy);
847 free(fs_info->log_root_tree);
848 free(fs_info);
849 }
850
851 struct btrfs_fs_info *btrfs_new_fs_info(int writable, u64 sb_bytenr)
852 {
853 struct btrfs_fs_info *fs_info;
854
855 fs_info = calloc(1, sizeof(struct btrfs_fs_info));
856 if (!fs_info)
857 return NULL;
858
859 fs_info->tree_root = calloc(1, sizeof(struct btrfs_root));
860 fs_info->extent_root = calloc(1, sizeof(struct btrfs_root));
861 fs_info->chunk_root = calloc(1, sizeof(struct btrfs_root));
862 fs_info->dev_root = calloc(1, sizeof(struct btrfs_root));
863 fs_info->csum_root = calloc(1, sizeof(struct btrfs_root));
864 fs_info->quota_root = calloc(1, sizeof(struct btrfs_root));
865 fs_info->free_space_root = calloc(1, sizeof(struct btrfs_root));
866 fs_info->super_copy = calloc(1, BTRFS_SUPER_INFO_SIZE);
867
868 if (!fs_info->tree_root || !fs_info->extent_root ||
869 !fs_info->chunk_root || !fs_info->dev_root ||
870 !fs_info->csum_root || !fs_info->quota_root ||
871 !fs_info->free_space_root || !fs_info->super_copy)
872 goto free_all;
873
874 extent_io_tree_init(&fs_info->extent_cache);
875 extent_io_tree_init(&fs_info->free_space_cache);
876 extent_io_tree_init(&fs_info->block_group_cache);
877 extent_io_tree_init(&fs_info->pinned_extents);
878 extent_io_tree_init(&fs_info->pending_del);
879 extent_io_tree_init(&fs_info->extent_ins);
880 fs_info->excluded_extents = NULL;
881
882 fs_info->fs_root_tree = RB_ROOT;
883 cache_tree_init(&fs_info->mapping_tree.cache_tree);
884
885 mutex_init(&fs_info->fs_mutex);
886 INIT_LIST_HEAD(&fs_info->dirty_cowonly_roots);
887 INIT_LIST_HEAD(&fs_info->space_info);
888 INIT_LIST_HEAD(&fs_info->recow_ebs);
889
890 if (!writable)
891 fs_info->readonly = 1;
892
893 fs_info->super_bytenr = sb_bytenr;
894 fs_info->data_alloc_profile = (u64)-1;
895 fs_info->metadata_alloc_profile = (u64)-1;
896 fs_info->system_alloc_profile = fs_info->metadata_alloc_profile;
897 return fs_info;
898 free_all:
899 btrfs_free_fs_info(fs_info);
900 return NULL;
901 }
902
903 int btrfs_check_fs_compatibility(struct btrfs_super_block *sb, int writable)
904 {
905 u64 features;
906
907 features = btrfs_super_incompat_flags(sb) &
908 ~BTRFS_FEATURE_INCOMPAT_SUPP;
909 if (features) {
910 printk("couldn't open because of unsupported "
911 "option features (%Lx).\n",
912 (unsigned long long)features);
913 return -ENOTSUP;
914 }
915
916 features = btrfs_super_incompat_flags(sb);
917 if (!(features & BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF)) {
918 features |= BTRFS_FEATURE_INCOMPAT_MIXED_BACKREF;
919 btrfs_set_super_incompat_flags(sb, features);
920 }
921
922 features = btrfs_super_compat_ro_flags(sb) &
923 ~BTRFS_FEATURE_COMPAT_RO_SUPP;
924 if (writable && features) {
925 printk("couldn't open RDWR because of unsupported "
926 "option features (%Lx).\n",
927 (unsigned long long)features);
928 return -ENOTSUP;
929 }
930 return 0;
931 }
932
933 static int find_best_backup_root(struct btrfs_super_block *super)
934 {
935 struct btrfs_root_backup *backup;
936 u64 orig_gen = btrfs_super_generation(super);
937 u64 gen = 0;
938 int best_index = 0;
939 int i;
940
941 for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
942 backup = super->super_roots + i;
943 if (btrfs_backup_tree_root_gen(backup) != orig_gen &&
944 btrfs_backup_tree_root_gen(backup) > gen) {
945 best_index = i;
946 gen = btrfs_backup_tree_root_gen(backup);
947 }
948 }
949 return best_index;
950 }
951
952 static int setup_root_or_create_block(struct btrfs_fs_info *fs_info,
953 unsigned flags,
954 struct btrfs_root *info_root,
955 u64 objectid, char *str)
956 {
957 struct btrfs_super_block *sb = fs_info->super_copy;
958 struct btrfs_root *root = fs_info->tree_root;
959 u32 nodesize = btrfs_super_nodesize(sb);
960 int ret;
961
962 ret = find_and_setup_root(root, fs_info, objectid, info_root);
963 if (ret) {
964 printk("Couldn't setup %s tree\n", str);
965 if (!(flags & OPEN_CTREE_PARTIAL))
966 return -EIO;
967 /*
968 * Need a blank node here just so we don't screw up in the
969 * million of places that assume a root has a valid ->node
970 */
971 info_root->node =
972 btrfs_find_create_tree_block(fs_info, 0, nodesize);
973 if (!info_root->node)
974 return -ENOMEM;
975 clear_extent_buffer_uptodate(NULL, info_root->node);
976 }
977
978 return 0;
979 }
980
981 int btrfs_setup_all_roots(struct btrfs_fs_info *fs_info, u64 root_tree_bytenr,
982 unsigned flags)
983 {
984 struct btrfs_super_block *sb = fs_info->super_copy;
985 struct btrfs_root *root;
986 struct btrfs_key key;
987 u32 sectorsize;
988 u32 nodesize;
989 u32 leafsize;
990 u32 stripesize;
991 u64 generation;
992 u32 blocksize;
993 int ret;
994
995 nodesize = btrfs_super_nodesize(sb);
996 leafsize = btrfs_super_leafsize(sb);
997 sectorsize = btrfs_super_sectorsize(sb);
998 stripesize = btrfs_super_stripesize(sb);
999
1000 root = fs_info->tree_root;
1001 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1002 root, fs_info, BTRFS_ROOT_TREE_OBJECTID);
1003 blocksize = root->nodesize;
1004 generation = btrfs_super_generation(sb);
1005
1006 if (!root_tree_bytenr && !(flags & OPEN_CTREE_BACKUP_ROOT)) {
1007 root_tree_bytenr = btrfs_super_root(sb);
1008 } else if (flags & OPEN_CTREE_BACKUP_ROOT) {
1009 struct btrfs_root_backup *backup;
1010 int index = find_best_backup_root(sb);
1011 if (index >= BTRFS_NUM_BACKUP_ROOTS) {
1012 fprintf(stderr, "Invalid backup root number\n");
1013 return -EIO;
1014 }
1015 backup = fs_info->super_copy->super_roots + index;
1016 root_tree_bytenr = btrfs_backup_tree_root(backup);
1017 generation = btrfs_backup_tree_root_gen(backup);
1018 }
1019
1020 root->node = read_tree_block(root, root_tree_bytenr, blocksize,
1021 generation);
1022 if (!extent_buffer_uptodate(root->node)) {
1023 fprintf(stderr, "Couldn't read tree root\n");
1024 return -EIO;
1025 }
1026
1027 ret = setup_root_or_create_block(fs_info, flags, fs_info->extent_root,
1028 BTRFS_EXTENT_TREE_OBJECTID, "extent");
1029 if (ret)
1030 return ret;
1031 fs_info->extent_root->track_dirty = 1;
1032
1033 ret = find_and_setup_root(root, fs_info, BTRFS_DEV_TREE_OBJECTID,
1034 fs_info->dev_root);
1035 if (ret) {
1036 printk("Couldn't setup device tree\n");
1037 return -EIO;
1038 }
1039 fs_info->dev_root->track_dirty = 1;
1040
1041 ret = setup_root_or_create_block(fs_info, flags, fs_info->csum_root,
1042 BTRFS_CSUM_TREE_OBJECTID, "csum");
1043 if (ret)
1044 return ret;
1045 fs_info->csum_root->track_dirty = 1;
1046
1047 ret = find_and_setup_root(root, fs_info, BTRFS_QUOTA_TREE_OBJECTID,
1048 fs_info->quota_root);
1049 if (ret == 0)
1050 fs_info->quota_enabled = 1;
1051
1052 if (btrfs_fs_compat_ro(fs_info, BTRFS_FEATURE_COMPAT_RO_FREE_SPACE_TREE)) {
1053 ret = find_and_setup_root(root, fs_info, BTRFS_FREE_SPACE_TREE_OBJECTID,
1054 fs_info->free_space_root);
1055 if (ret) {
1056 printk("Couldn't read free space tree\n");
1057 return -EIO;
1058 }
1059 fs_info->free_space_root->track_dirty = 1;
1060 }
1061
1062 ret = find_and_setup_log_root(root, fs_info, sb);
1063 if (ret) {
1064 printk("Couldn't setup log root tree\n");
1065 if (!(flags & OPEN_CTREE_PARTIAL))
1066 return -EIO;
1067 }
1068
1069 fs_info->generation = generation;
1070 fs_info->last_trans_committed = generation;
1071 if (extent_buffer_uptodate(fs_info->extent_root->node) &&
1072 !(flags & OPEN_CTREE_NO_BLOCK_GROUPS))
1073 btrfs_read_block_groups(fs_info->tree_root);
1074
1075 key.objectid = BTRFS_FS_TREE_OBJECTID;
1076 key.type = BTRFS_ROOT_ITEM_KEY;
1077 key.offset = (u64)-1;
1078 fs_info->fs_root = btrfs_read_fs_root(fs_info, &key);
1079
1080 if (IS_ERR(fs_info->fs_root))
1081 return -EIO;
1082 return 0;
1083 }
1084
1085 void btrfs_release_all_roots(struct btrfs_fs_info *fs_info)
1086 {
1087 if (fs_info->free_space_root)
1088 free_extent_buffer(fs_info->free_space_root->node);
1089 if (fs_info->quota_root)
1090 free_extent_buffer(fs_info->quota_root->node);
1091 if (fs_info->csum_root)
1092 free_extent_buffer(fs_info->csum_root->node);
1093 if (fs_info->dev_root)
1094 free_extent_buffer(fs_info->dev_root->node);
1095 if (fs_info->extent_root)
1096 free_extent_buffer(fs_info->extent_root->node);
1097 if (fs_info->tree_root)
1098 free_extent_buffer(fs_info->tree_root->node);
1099 if (fs_info->log_root_tree)
1100 free_extent_buffer(fs_info->log_root_tree->node);
1101 if (fs_info->chunk_root)
1102 free_extent_buffer(fs_info->chunk_root->node);
1103 }
1104
1105 static void free_map_lookup(struct cache_extent *ce)
1106 {
1107 struct map_lookup *map;
1108
1109 map = container_of(ce, struct map_lookup, ce);
1110 kfree(map);
1111 }
1112
1113 FREE_EXTENT_CACHE_BASED_TREE(mapping_cache, free_map_lookup);
1114
1115 void btrfs_cleanup_all_caches(struct btrfs_fs_info *fs_info)
1116 {
1117 while (!list_empty(&fs_info->recow_ebs)) {
1118 struct extent_buffer *eb;
1119 eb = list_first_entry(&fs_info->recow_ebs,
1120 struct extent_buffer, recow);
1121 list_del_init(&eb->recow);
1122 free_extent_buffer(eb);
1123 }
1124 free_mapping_cache_tree(&fs_info->mapping_tree.cache_tree);
1125 extent_io_tree_cleanup(&fs_info->extent_cache);
1126 extent_io_tree_cleanup(&fs_info->free_space_cache);
1127 extent_io_tree_cleanup(&fs_info->block_group_cache);
1128 extent_io_tree_cleanup(&fs_info->pinned_extents);
1129 extent_io_tree_cleanup(&fs_info->pending_del);
1130 extent_io_tree_cleanup(&fs_info->extent_ins);
1131 }
1132
1133 int btrfs_scan_fs_devices(int fd, const char *path,
1134 struct btrfs_fs_devices **fs_devices,
1135 u64 sb_bytenr, unsigned sbflags,
1136 int skip_devices)
1137 {
1138 u64 total_devs;
1139 u64 dev_size;
1140 off_t seek_ret;
1141 int ret;
1142 if (!sb_bytenr)
1143 sb_bytenr = BTRFS_SUPER_INFO_OFFSET;
1144
1145 seek_ret = lseek(fd, 0, SEEK_END);
1146 if (seek_ret < 0)
1147 return -errno;
1148
1149 dev_size = seek_ret;
1150 lseek(fd, 0, SEEK_SET);
1151 if (sb_bytenr > dev_size) {
1152 error("superblock bytenr %llu is larger than device size %llu",
1153 (unsigned long long)sb_bytenr,
1154 (unsigned long long)dev_size);
1155 return -EINVAL;
1156 }
1157
1158 ret = btrfs_scan_one_device(fd, path, fs_devices,
1159 &total_devs, sb_bytenr, sbflags);
1160 if (ret) {
1161 fprintf(stderr, "No valid Btrfs found on %s\n", path);
1162 return ret;
1163 }
1164
1165 if (!skip_devices && total_devs != 1) {
1166 ret = btrfs_scan_lblkid();
1167 if (ret)
1168 return ret;
1169 }
1170 return 0;
1171 }
1172
1173 int btrfs_setup_chunk_tree_and_device_map(struct btrfs_fs_info *fs_info,
1174 u64 chunk_root_bytenr)
1175 {
1176 struct btrfs_super_block *sb = fs_info->super_copy;
1177 u32 sectorsize;
1178 u32 nodesize;
1179 u32 leafsize;
1180 u32 blocksize;
1181 u32 stripesize;
1182 u64 generation;
1183 int ret;
1184
1185 nodesize = btrfs_super_nodesize(sb);
1186 leafsize = btrfs_super_leafsize(sb);
1187 sectorsize = btrfs_super_sectorsize(sb);
1188 stripesize = btrfs_super_stripesize(sb);
1189
1190 __setup_root(nodesize, leafsize, sectorsize, stripesize,
1191 fs_info->chunk_root, fs_info, BTRFS_CHUNK_TREE_OBJECTID);
1192
1193 ret = btrfs_read_sys_array(fs_info->chunk_root);
1194 if (ret)
1195 return ret;
1196
1197 blocksize = fs_info->chunk_root->nodesize;
1198 generation = btrfs_super_chunk_root_generation(sb);
1199
1200 if (chunk_root_bytenr && !IS_ALIGNED(chunk_root_bytenr,
1201 btrfs_super_sectorsize(sb))) {
1202 warning("chunk_root_bytenr %llu is unaligned to %u, ignore it",
1203 chunk_root_bytenr, btrfs_super_sectorsize(sb));
1204 chunk_root_bytenr = 0;
1205 }
1206
1207 if (!chunk_root_bytenr)
1208 chunk_root_bytenr = btrfs_super_chunk_root(sb);
1209 else
1210 generation = 0;
1211
1212 fs_info->chunk_root->node = read_tree_block(fs_info->chunk_root,
1213 chunk_root_bytenr,
1214 blocksize, generation);
1215 if (!extent_buffer_uptodate(fs_info->chunk_root->node)) {
1216 if (fs_info->ignore_chunk_tree_error) {
1217 warning("cannot read chunk root, continue anyway");
1218 fs_info->chunk_root = NULL;
1219 return 0;
1220 } else {
1221 error("cannot read chunk root");
1222 return -EIO;
1223 }
1224 }
1225
1226 if (!(btrfs_super_flags(sb) & BTRFS_SUPER_FLAG_METADUMP)) {
1227 ret = btrfs_read_chunk_tree(fs_info->chunk_root);
1228 if (ret) {
1229 fprintf(stderr, "Couldn't read chunk tree\n");
1230 return ret;
1231 }
1232 }
1233 return 0;
1234 }
1235
1236 static struct btrfs_fs_info *__open_ctree_fd(int fp, const char *path,
1237 u64 sb_bytenr,
1238 u64 root_tree_bytenr,
1239 u64 chunk_root_bytenr,
1240 unsigned flags)
1241 {
1242 struct btrfs_fs_info *fs_info;
1243 struct btrfs_super_block *disk_super;
1244 struct btrfs_fs_devices *fs_devices = NULL;
1245 struct extent_buffer *eb;
1246 int ret;
1247 int oflags;
1248 unsigned sbflags = SBREAD_DEFAULT;
1249
1250 if (sb_bytenr == 0)
1251 sb_bytenr = BTRFS_SUPER_INFO_OFFSET;
1252
1253 /* try to drop all the caches */
1254 if (posix_fadvise(fp, 0, 0, POSIX_FADV_DONTNEED))
1255 fprintf(stderr, "Warning, could not drop caches\n");
1256
1257 fs_info = btrfs_new_fs_info(flags & OPEN_CTREE_WRITES, sb_bytenr);
1258 if (!fs_info) {
1259 fprintf(stderr, "Failed to allocate memory for fs_info\n");
1260 return NULL;
1261 }
1262 if (flags & OPEN_CTREE_RESTORE)
1263 fs_info->on_restoring = 1;
1264 if (flags & OPEN_CTREE_SUPPRESS_CHECK_BLOCK_ERRORS)
1265 fs_info->suppress_check_block_errors = 1;
1266 if (flags & OPEN_CTREE_IGNORE_FSID_MISMATCH)
1267 fs_info->ignore_fsid_mismatch = 1;
1268 if (flags & OPEN_CTREE_IGNORE_CHUNK_TREE_ERROR)
1269 fs_info->ignore_chunk_tree_error = 1;
1270
1271 if ((flags & OPEN_CTREE_RECOVER_SUPER)
1272 && (flags & OPEN_CTREE_FS_PARTIAL)) {
1273 fprintf(stderr,
1274 "cannot open a partially created filesystem for recovery");
1275 goto out;
1276 }
1277
1278 if (flags & OPEN_CTREE_FS_PARTIAL)
1279 sbflags = SBREAD_PARTIAL;
1280
1281 ret = btrfs_scan_fs_devices(fp, path, &fs_devices, sb_bytenr, sbflags,
1282 (flags & OPEN_CTREE_NO_DEVICES));
1283 if (ret)
1284 goto out;
1285
1286 fs_info->fs_devices = fs_devices;
1287 if (flags & OPEN_CTREE_WRITES)
1288 oflags = O_RDWR;
1289 else
1290 oflags = O_RDONLY;
1291
1292 if (flags & OPEN_CTREE_EXCLUSIVE)
1293 oflags |= O_EXCL;
1294
1295 ret = btrfs_open_devices(fs_devices, oflags);
1296 if (ret)
1297 goto out;
1298
1299 disk_super = fs_info->super_copy;
1300 if (flags & OPEN_CTREE_RECOVER_SUPER)
1301 ret = btrfs_read_dev_super(fs_devices->latest_bdev, disk_super,
1302 sb_bytenr, SBREAD_RECOVER);
1303 else
1304 ret = btrfs_read_dev_super(fp, disk_super, sb_bytenr,
1305 sbflags);
1306 if (ret) {
1307 printk("No valid btrfs found\n");
1308 goto out_devices;
1309 }
1310
1311 if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_CHANGING_FSID &&
1312 !fs_info->ignore_fsid_mismatch) {
1313 fprintf(stderr, "ERROR: Filesystem UUID change in progress\n");
1314 goto out_devices;
1315 }
1316
1317 memcpy(fs_info->fsid, &disk_super->fsid, BTRFS_FSID_SIZE);
1318
1319 ret = btrfs_check_fs_compatibility(fs_info->super_copy,
1320 flags & OPEN_CTREE_WRITES);
1321 if (ret)
1322 goto out_devices;
1323
1324 ret = btrfs_setup_chunk_tree_and_device_map(fs_info, chunk_root_bytenr);
1325 if (ret)
1326 goto out_chunk;
1327
1328 /* Chunk tree root is unable to read, return directly */
1329 if (!fs_info->chunk_root)
1330 return fs_info;
1331
1332 eb = fs_info->chunk_root->node;
1333 read_extent_buffer(eb, fs_info->chunk_tree_uuid,
1334 btrfs_header_chunk_tree_uuid(eb),
1335 BTRFS_UUID_SIZE);
1336
1337 ret = btrfs_setup_all_roots(fs_info, root_tree_bytenr, flags);
1338 if (ret && !(flags & __OPEN_CTREE_RETURN_CHUNK_ROOT) &&
1339 !fs_info->ignore_chunk_tree_error)
1340 goto out_chunk;
1341
1342 return fs_info;
1343
1344 out_chunk:
1345 btrfs_release_all_roots(fs_info);
1346 btrfs_cleanup_all_caches(fs_info);
1347 out_devices:
1348 btrfs_close_devices(fs_devices);
1349 out:
1350 btrfs_free_fs_info(fs_info);
1351 return NULL;
1352 }
1353
1354 struct btrfs_fs_info *open_ctree_fs_info(const char *filename,
1355 u64 sb_bytenr, u64 root_tree_bytenr,
1356 u64 chunk_root_bytenr,
1357 unsigned flags)
1358 {
1359 int fp;
1360 int ret;
1361 struct btrfs_fs_info *info;
1362 int oflags = O_RDWR;
1363 struct stat st;
1364
1365 ret = stat(filename, &st);
1366 if (ret < 0) {
1367 error("cannot stat '%s': %s", filename, strerror(errno));
1368 return NULL;
1369 }
1370 if (!(((st.st_mode & S_IFMT) == S_IFREG) || ((st.st_mode & S_IFMT) == S_IFBLK))) {
1371 error("not a regular file or block device: %s", filename);
1372 return NULL;
1373 }
1374
1375 if (!(flags & OPEN_CTREE_WRITES))
1376 oflags = O_RDONLY;
1377
1378 fp = open(filename, oflags);
1379 if (fp < 0) {
1380 error("cannot open '%s': %s", filename, strerror(errno));
1381 return NULL;
1382 }
1383 info = __open_ctree_fd(fp, filename, sb_bytenr, root_tree_bytenr,
1384 chunk_root_bytenr, flags);
1385 close(fp);
1386 return info;
1387 }
1388
1389 struct btrfs_root *open_ctree(const char *filename, u64 sb_bytenr,
1390 unsigned flags)
1391 {
1392 struct btrfs_fs_info *info;
1393
1394 /* This flags may not return fs_info with any valid root */
1395 BUG_ON(flags & OPEN_CTREE_IGNORE_CHUNK_TREE_ERROR);
1396 info = open_ctree_fs_info(filename, sb_bytenr, 0, 0, flags);
1397 if (!info)
1398 return NULL;
1399 if (flags & __OPEN_CTREE_RETURN_CHUNK_ROOT)
1400 return info->chunk_root;
1401 return info->fs_root;
1402 }
1403
1404 struct btrfs_root *open_ctree_fd(int fp, const char *path, u64 sb_bytenr,
1405 unsigned flags)
1406 {
1407 struct btrfs_fs_info *info;
1408
1409 /* This flags may not return fs_info with any valid root */
1410 BUG_ON(flags & OPEN_CTREE_IGNORE_CHUNK_TREE_ERROR);
1411 info = __open_ctree_fd(fp, path, sb_bytenr, 0, 0, flags);
1412 if (!info)
1413 return NULL;
1414 if (flags & __OPEN_CTREE_RETURN_CHUNK_ROOT)
1415 return info->chunk_root;
1416 return info->fs_root;
1417 }
1418
1419 /*
1420 * Check if the super is valid:
1421 * - nodesize/sectorsize - minimum, maximum, alignment
1422 * - tree block starts - alignment
1423 * - number of devices - something sane
1424 * - sys array size - maximum
1425 */
1426 static int check_super(struct btrfs_super_block *sb, unsigned sbflags)
1427 {
1428 char result[BTRFS_CSUM_SIZE];
1429 u32 crc;
1430 u16 csum_type;
1431 int csum_size;
1432
1433 if (btrfs_super_magic(sb) != BTRFS_MAGIC) {
1434 if (btrfs_super_magic(sb) == BTRFS_MAGIC_PARTIAL) {
1435 if (!(sbflags & SBREAD_PARTIAL)) {
1436 error("superblock magic doesn't match");
1437 return -EIO;
1438 }
1439 }
1440 }
1441
1442 csum_type = btrfs_super_csum_type(sb);
1443 if (csum_type >= ARRAY_SIZE(btrfs_csum_sizes)) {
1444 error("unsupported checksum algorithm %u", csum_type);
1445 return -EIO;
1446 }
1447 csum_size = btrfs_csum_sizes[csum_type];
1448
1449 crc = ~(u32)0;
1450 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
1451 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
1452 btrfs_csum_final(crc, result);
1453
1454 if (memcmp(result, sb->csum, csum_size)) {
1455 error("superblock checksum mismatch");
1456 return -EIO;
1457 }
1458 if (btrfs_super_root_level(sb) >= BTRFS_MAX_LEVEL) {
1459 error("tree_root level too big: %d >= %d",
1460 btrfs_super_root_level(sb), BTRFS_MAX_LEVEL);
1461 goto error_out;
1462 }
1463 if (btrfs_super_chunk_root_level(sb) >= BTRFS_MAX_LEVEL) {
1464 error("chunk_root level too big: %d >= %d",
1465 btrfs_super_chunk_root_level(sb), BTRFS_MAX_LEVEL);
1466 goto error_out;
1467 }
1468 if (btrfs_super_log_root_level(sb) >= BTRFS_MAX_LEVEL) {
1469 error("log_root level too big: %d >= %d",
1470 btrfs_super_log_root_level(sb), BTRFS_MAX_LEVEL);
1471 goto error_out;
1472 }
1473
1474 if (!IS_ALIGNED(btrfs_super_root(sb), 4096)) {
1475 error("tree_root block unaligned: %llu", btrfs_super_root(sb));
1476 goto error_out;
1477 }
1478 if (!IS_ALIGNED(btrfs_super_chunk_root(sb), 4096)) {
1479 error("chunk_root block unaligned: %llu",
1480 btrfs_super_chunk_root(sb));
1481 goto error_out;
1482 }
1483 if (!IS_ALIGNED(btrfs_super_log_root(sb), 4096)) {
1484 error("log_root block unaligned: %llu",
1485 btrfs_super_log_root(sb));
1486 goto error_out;
1487 }
1488 if (btrfs_super_nodesize(sb) < 4096) {
1489 error("nodesize too small: %u < 4096",
1490 btrfs_super_nodesize(sb));
1491 goto error_out;
1492 }
1493 if (!IS_ALIGNED(btrfs_super_nodesize(sb), 4096)) {
1494 error("nodesize unaligned: %u", btrfs_super_nodesize(sb));
1495 goto error_out;
1496 }
1497 if (btrfs_super_sectorsize(sb) < 4096) {
1498 error("sectorsize too small: %u < 4096",
1499 btrfs_super_sectorsize(sb));
1500 goto error_out;
1501 }
1502 if (!IS_ALIGNED(btrfs_super_sectorsize(sb), 4096)) {
1503 error("sectorsize unaligned: %u", btrfs_super_sectorsize(sb));
1504 goto error_out;
1505 }
1506 if (btrfs_super_total_bytes(sb) == 0) {
1507 error("invalid total_bytes 0");
1508 goto error_out;
1509 }
1510 if (btrfs_super_bytes_used(sb) < 6 * btrfs_super_nodesize(sb)) {
1511 error("invalid bytes_used %llu", btrfs_super_bytes_used(sb));
1512 goto error_out;
1513 }
1514 if ((btrfs_super_stripesize(sb) != 4096)
1515 && (btrfs_super_stripesize(sb) != btrfs_super_sectorsize(sb))) {
1516 error("invalid stripesize %u", btrfs_super_stripesize(sb));
1517 goto error_out;
1518 }
1519
1520 if (memcmp(sb->fsid, sb->dev_item.fsid, BTRFS_UUID_SIZE) != 0) {
1521 char fsid[BTRFS_UUID_UNPARSED_SIZE];
1522 char dev_fsid[BTRFS_UUID_UNPARSED_SIZE];
1523
1524 uuid_unparse(sb->fsid, fsid);
1525 uuid_unparse(sb->dev_item.fsid, dev_fsid);
1526 error("dev_item UUID does not match fsid: %s != %s",
1527 dev_fsid, fsid);
1528 goto error_out;
1529 }
1530
1531 /*
1532 * Hint to catch really bogus numbers, bitflips or so
1533 */
1534 if (btrfs_super_num_devices(sb) > (1UL << 31)) {
1535 warning("suspicious number of devices: %llu",
1536 btrfs_super_num_devices(sb));
1537 }
1538
1539 if (btrfs_super_num_devices(sb) == 0) {
1540 error("number of devices is 0");
1541 goto error_out;
1542 }
1543
1544 /*
1545 * Obvious sys_chunk_array corruptions, it must hold at least one key
1546 * and one chunk
1547 */
1548 if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) {
1549 error("system chunk array too big %u > %u",
1550 btrfs_super_sys_array_size(sb),
1551 BTRFS_SYSTEM_CHUNK_ARRAY_SIZE);
1552 goto error_out;
1553 }
1554 if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key)
1555 + sizeof(struct btrfs_chunk)) {
1556 error("system chunk array too small %u < %lu",
1557 btrfs_super_sys_array_size(sb),
1558 sizeof(struct btrfs_disk_key) +
1559 sizeof(struct btrfs_chunk));
1560 goto error_out;
1561 }
1562
1563 return 0;
1564
1565 error_out:
1566 error("superblock checksum matches but it has invalid members");
1567 return -EIO;
1568 }
1569
1570 int btrfs_read_dev_super(int fd, struct btrfs_super_block *sb, u64 sb_bytenr,
1571 unsigned sbflags)
1572 {
1573 u8 fsid[BTRFS_FSID_SIZE];
1574 int fsid_is_initialized = 0;
1575 char tmp[BTRFS_SUPER_INFO_SIZE];
1576 struct btrfs_super_block *buf = (struct btrfs_super_block *)tmp;
1577 int i;
1578 int ret;
1579 int max_super = sbflags & SBREAD_RECOVER ? BTRFS_SUPER_MIRROR_MAX : 1;
1580 u64 transid = 0;
1581 u64 bytenr;
1582
1583 if (sb_bytenr != BTRFS_SUPER_INFO_OFFSET) {
1584 ret = pread64(fd, buf, BTRFS_SUPER_INFO_SIZE, sb_bytenr);
1585 if (ret < BTRFS_SUPER_INFO_SIZE)
1586 return -1;
1587
1588 if (btrfs_super_bytenr(buf) != sb_bytenr)
1589 return -1;
1590
1591 if (check_super(buf, sbflags))
1592 return -1;
1593 memcpy(sb, buf, BTRFS_SUPER_INFO_SIZE);
1594 return 0;
1595 }
1596
1597 /*
1598 * we would like to check all the supers, but that would make
1599 * a btrfs mount succeed after a mkfs from a different FS.
1600 * So, we need to add a special mount option to scan for
1601 * later supers, using BTRFS_SUPER_MIRROR_MAX instead
1602 */
1603
1604 for (i = 0; i < max_super; i++) {
1605 bytenr = btrfs_sb_offset(i);
1606 ret = pread64(fd, buf, BTRFS_SUPER_INFO_SIZE, bytenr);
1607 if (ret < BTRFS_SUPER_INFO_SIZE)
1608 break;
1609
1610 if (btrfs_super_bytenr(buf) != bytenr )
1611 continue;
1612 /* if magic is NULL, the device was removed */
1613 if (btrfs_super_magic(buf) == 0 && i == 0)
1614 break;
1615 if (check_super(buf, sbflags))
1616 continue;
1617
1618 if (!fsid_is_initialized) {
1619 memcpy(fsid, buf->fsid, sizeof(fsid));
1620 fsid_is_initialized = 1;
1621 } else if (memcmp(fsid, buf->fsid, sizeof(fsid))) {
1622 /*
1623 * the superblocks (the original one and
1624 * its backups) contain data of different
1625 * filesystems -> the super cannot be trusted
1626 */
1627 continue;
1628 }
1629
1630 if (btrfs_super_generation(buf) > transid) {
1631 memcpy(sb, buf, BTRFS_SUPER_INFO_SIZE);
1632 transid = btrfs_super_generation(buf);
1633 }
1634 }
1635
1636 return transid > 0 ? 0 : -1;
1637 }
1638
1639 static int write_dev_supers(struct btrfs_root *root,
1640 struct btrfs_super_block *sb,
1641 struct btrfs_device *device)
1642 {
1643 u64 bytenr;
1644 u32 crc;
1645 int i, ret;
1646
1647 if (root->fs_info->super_bytenr != BTRFS_SUPER_INFO_OFFSET) {
1648 btrfs_set_super_bytenr(sb, root->fs_info->super_bytenr);
1649 crc = ~(u32)0;
1650 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
1651 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
1652 btrfs_csum_final(crc, (char *)&sb->csum[0]);
1653
1654 /*
1655 * super_copy is BTRFS_SUPER_INFO_SIZE bytes and is
1656 * zero filled, we can use it directly
1657 */
1658 ret = pwrite64(device->fd, root->fs_info->super_copy,
1659 BTRFS_SUPER_INFO_SIZE,
1660 root->fs_info->super_bytenr);
1661 if (ret != BTRFS_SUPER_INFO_SIZE)
1662 goto write_err;
1663 return 0;
1664 }
1665
1666 for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
1667 bytenr = btrfs_sb_offset(i);
1668 if (bytenr + BTRFS_SUPER_INFO_SIZE > device->total_bytes)
1669 break;
1670
1671 btrfs_set_super_bytenr(sb, bytenr);
1672
1673 crc = ~(u32)0;
1674 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
1675 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
1676 btrfs_csum_final(crc, (char *)&sb->csum[0]);
1677
1678 /*
1679 * super_copy is BTRFS_SUPER_INFO_SIZE bytes and is
1680 * zero filled, we can use it directly
1681 */
1682 ret = pwrite64(device->fd, root->fs_info->super_copy,
1683 BTRFS_SUPER_INFO_SIZE, bytenr);
1684 if (ret != BTRFS_SUPER_INFO_SIZE)
1685 goto write_err;
1686 }
1687
1688 return 0;
1689
1690 write_err:
1691 if (ret > 0)
1692 fprintf(stderr, "WARNING: failed to write all sb data\n");
1693 else
1694 fprintf(stderr, "WARNING: failed to write sb: %s\n",
1695 strerror(errno));
1696 return ret;
1697 }
1698
1699 int write_all_supers(struct btrfs_root *root)
1700 {
1701 struct list_head *cur;
1702 struct list_head *head = &root->fs_info->fs_devices->devices;
1703 struct btrfs_device *dev;
1704 struct btrfs_super_block *sb;
1705 struct btrfs_dev_item *dev_item;
1706 int ret;
1707 u64 flags;
1708
1709 sb = root->fs_info->super_copy;
1710 dev_item = &sb->dev_item;
1711 list_for_each(cur, head) {
1712 dev = list_entry(cur, struct btrfs_device, dev_list);
1713 if (!dev->writeable)
1714 continue;
1715
1716 btrfs_set_stack_device_generation(dev_item, 0);
1717 btrfs_set_stack_device_type(dev_item, dev->type);
1718 btrfs_set_stack_device_id(dev_item, dev->devid);
1719 btrfs_set_stack_device_total_bytes(dev_item, dev->total_bytes);
1720 btrfs_set_stack_device_bytes_used(dev_item, dev->bytes_used);
1721 btrfs_set_stack_device_io_align(dev_item, dev->io_align);
1722 btrfs_set_stack_device_io_width(dev_item, dev->io_width);
1723 btrfs_set_stack_device_sector_size(dev_item, dev->sector_size);
1724 memcpy(dev_item->uuid, dev->uuid, BTRFS_UUID_SIZE);
1725 memcpy(dev_item->fsid, dev->fs_devices->fsid, BTRFS_UUID_SIZE);
1726
1727 flags = btrfs_super_flags(sb);
1728 btrfs_set_super_flags(sb, flags | BTRFS_HEADER_FLAG_WRITTEN);
1729
1730 ret = write_dev_supers(root, sb, dev);
1731 BUG_ON(ret);
1732 }
1733 return 0;
1734 }
1735
1736 int write_ctree_super(struct btrfs_trans_handle *trans,
1737 struct btrfs_root *root)
1738 {
1739 int ret;
1740 struct btrfs_root *tree_root = root->fs_info->tree_root;
1741 struct btrfs_root *chunk_root = root->fs_info->chunk_root;
1742
1743 if (root->fs_info->readonly)
1744 return 0;
1745
1746 btrfs_set_super_generation(root->fs_info->super_copy,
1747 trans->transid);
1748 btrfs_set_super_root(root->fs_info->super_copy,
1749 tree_root->node->start);
1750 btrfs_set_super_root_level(root->fs_info->super_copy,
1751 btrfs_header_level(tree_root->node));
1752 btrfs_set_super_chunk_root(root->fs_info->super_copy,
1753 chunk_root->node->start);
1754 btrfs_set_super_chunk_root_level(root->fs_info->super_copy,
1755 btrfs_header_level(chunk_root->node));
1756 btrfs_set_super_chunk_root_generation(root->fs_info->super_copy,
1757 btrfs_header_generation(chunk_root->node));
1758
1759 ret = write_all_supers(root);
1760 if (ret)
1761 fprintf(stderr, "failed to write new super block err %d\n", ret);
1762 return ret;
1763 }
1764
1765 int close_ctree_fs_info(struct btrfs_fs_info *fs_info)
1766 {
1767 int ret;
1768 struct btrfs_trans_handle *trans;
1769 struct btrfs_root *root = fs_info->tree_root;
1770
1771 if (fs_info->last_trans_committed !=
1772 fs_info->generation) {
1773 BUG_ON(!root);
1774 trans = btrfs_start_transaction(root, 1);
1775 btrfs_commit_transaction(trans, root);
1776 trans = btrfs_start_transaction(root, 1);
1777 ret = commit_tree_roots(trans, fs_info);
1778 BUG_ON(ret);
1779 ret = __commit_transaction(trans, root);
1780 BUG_ON(ret);
1781 write_ctree_super(trans, root);
1782 btrfs_free_transaction(root, trans);
1783 }
1784
1785 if (fs_info->finalize_on_close) {
1786 btrfs_set_super_magic(fs_info->super_copy, BTRFS_MAGIC);
1787 root->fs_info->finalize_on_close = 0;
1788 ret = write_all_supers(root);
1789 if (ret)
1790 fprintf(stderr,
1791 "failed to write new super block err %d\n", ret);
1792 }
1793 btrfs_free_block_groups(fs_info);
1794
1795 free_fs_roots_tree(&fs_info->fs_root_tree);
1796
1797 btrfs_release_all_roots(fs_info);
1798 btrfs_close_devices(fs_info->fs_devices);
1799 btrfs_cleanup_all_caches(fs_info);
1800 btrfs_free_fs_info(fs_info);
1801 return 0;
1802 }
1803
1804 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
1805 struct extent_buffer *eb)
1806 {
1807 return clear_extent_buffer_dirty(eb);
1808 }
1809
1810 int wait_on_tree_block_writeback(struct btrfs_root *root,
1811 struct extent_buffer *eb)
1812 {
1813 return 0;
1814 }
1815
1816 void btrfs_mark_buffer_dirty(struct extent_buffer *eb)
1817 {
1818 set_extent_buffer_dirty(eb);
1819 }
1820
1821 int btrfs_buffer_uptodate(struct extent_buffer *buf, u64 parent_transid)
1822 {
1823 int ret;
1824
1825 ret = extent_buffer_uptodate(buf);
1826 if (!ret)
1827 return ret;
1828
1829 ret = verify_parent_transid(buf->tree, buf, parent_transid, 1);
1830 return !ret;
1831 }
1832
1833 int btrfs_set_buffer_uptodate(struct extent_buffer *eb)
1834 {
1835 return set_extent_buffer_uptodate(eb);
1836 }
(deprecated) Btrfs progs developments and patch integration