?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
btrfs-progs: Introduce function to setup temporary chunk root
[btrfs-progs-unstable/devel.git] / utils.c
blobfcabcb529c263bb4dc5e1e6d15c1877a769d394b
1 /*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 * Copyright (C) 2008 Morey Roof. All rights reserved.
4 *
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public
7 * License v2 as published by the Free Software Foundation.
8 *
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * General Public License for more details.
13 *
14 * You should have received a copy of the GNU General Public
15 * License along with this program; if not, write to the
16 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
17 * Boston, MA 021110-1307, USA.
18 */
19
20 #include <stdio.h>
21 #include <stdlib.h>
22 #include <string.h>
23 #include <sys/ioctl.h>
24 #include <sys/mount.h>
25 #include <sys/types.h>
26 #include <sys/stat.h>
27 #include <uuid/uuid.h>
28 #include <fcntl.h>
29 #include <unistd.h>
30 #include <mntent.h>
31 #include <ctype.h>
32 #include <linux/loop.h>
33 #include <linux/major.h>
34 #include <linux/kdev_t.h>
35 #include <limits.h>
36 #include <blkid/blkid.h>
37 #include <sys/vfs.h>
38 #include <sys/statfs.h>
39 #include <linux/magic.h>
40 #include <getopt.h>
41
42 #include "kerncompat.h"
43 #include "radix-tree.h"
44 #include "ctree.h"
45 #include "disk-io.h"
46 #include "transaction.h"
47 #include "crc32c.h"
48 #include "utils.h"
49 #include "volumes.h"
50 #include "ioctl.h"
51 #include "commands.h"
52
53 #ifndef BLKDISCARD
54 #define BLKDISCARD _IO(0x12,119)
55 #endif
56
57 static int btrfs_scan_done = 0;
58
59 static char argv0_buf[ARGV0_BUF_SIZE] = "btrfs";
60
61 static int rand_seed_initlized = 0;
62 static unsigned short rand_seed[3];
63
64 const char *get_argv0_buf(void)
65 {
66 return argv0_buf;
67 }
68
69 void fixup_argv0(char **argv, const char *token)
70 {
71 int len = strlen(argv0_buf);
72
73 snprintf(argv0_buf + len, sizeof(argv0_buf) - len, " %s", token);
74 argv[0] = argv0_buf;
75 }
76
77 void set_argv0(char **argv)
78 {
79 strncpy(argv0_buf, argv[0], sizeof(argv0_buf));
80 argv0_buf[sizeof(argv0_buf) - 1] = 0;
81 }
82
83 int check_argc_exact(int nargs, int expected)
84 {
85 if (nargs < expected)
86 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
87 if (nargs > expected)
88 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
89
90 return nargs != expected;
91 }
92
93 int check_argc_min(int nargs, int expected)
94 {
95 if (nargs < expected) {
96 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
97 return 1;
98 }
99
100 return 0;
101 }
102
103 int check_argc_max(int nargs, int expected)
104 {
105 if (nargs > expected) {
106 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
107 return 1;
108 }
109
110 return 0;
111 }
112
113
114 /*
115 * Discard the given range in one go
116 */
117 static int discard_range(int fd, u64 start, u64 len)
118 {
119 u64 range[2] = { start, len };
120
121 if (ioctl(fd, BLKDISCARD, &range) < 0)
122 return errno;
123 return 0;
124 }
125
126 /*
127 * Discard blocks in the given range in 1G chunks, the process is interruptible
128 */
129 static int discard_blocks(int fd, u64 start, u64 len)
130 {
131 while (len > 0) {
132 /* 1G granularity */
133 u64 chunk_size = min_t(u64, len, 1*1024*1024*1024);
134 int ret;
135
136 ret = discard_range(fd, start, chunk_size);
137 if (ret)
138 return ret;
139 len -= chunk_size;
140 start += chunk_size;
141 }
142
143 return 0;
144 }
145
146 static u64 reference_root_table[] = {
147 [1] = BTRFS_ROOT_TREE_OBJECTID,
148 [2] = BTRFS_EXTENT_TREE_OBJECTID,
149 [3] = BTRFS_CHUNK_TREE_OBJECTID,
150 [4] = BTRFS_DEV_TREE_OBJECTID,
151 [5] = BTRFS_FS_TREE_OBJECTID,
152 [6] = BTRFS_CSUM_TREE_OBJECTID,
153 };
154
155 int test_uuid_unique(char *fs_uuid)
156 {
157 int unique = 1;
158 blkid_dev_iterate iter = NULL;
159 blkid_dev dev = NULL;
160 blkid_cache cache = NULL;
161
162 if (blkid_get_cache(&cache, NULL) < 0) {
163 printf("ERROR: lblkid cache get failed\n");
164 return 1;
165 }
166 blkid_probe_all(cache);
167 iter = blkid_dev_iterate_begin(cache);
168 blkid_dev_set_search(iter, "UUID", fs_uuid);
169
170 while (blkid_dev_next(iter, &dev) == 0) {
171 dev = blkid_verify(cache, dev);
172 if (dev) {
173 unique = 0;
174 break;
175 }
176 }
177
178 blkid_dev_iterate_end(iter);
179 blkid_put_cache(cache);
180
181 return unique;
182 }
183
184 /*
185 * Reserve space from free_tree.
186 * The algorithm is very simple, find the first cache_extent with enough space
187 * and allocate from its beginning.
188 */
189 static int reserve_free_space(struct cache_tree *free_tree, u64 len,
190 u64 *ret_start)
191 {
192 struct cache_extent *cache;
193 int found = 0;
194
195 BUG_ON(!ret_start);
196 cache = first_cache_extent(free_tree);
197 while (cache) {
198 if (cache->size > len) {
199 found = 1;
200 *ret_start = cache->start;
201
202 cache->size -= len;
203 if (cache->size == 0) {
204 remove_cache_extent(free_tree, cache);
205 free(cache);
206 } else {
207 cache->start += len;
208 }
209 break;
210 }
211 cache = next_cache_extent(cache);
212 }
213 if (!found)
214 return -ENOSPC;
215 return 0;
216 }
217
218 static inline int write_temp_super(int fd, struct btrfs_super_block *sb,
219 u64 sb_bytenr)
220 {
221 u32 crc = ~(u32)0;
222 int ret;
223
224 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
225 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
226 btrfs_csum_final(crc, (char *)&sb->csum[0]);
227 ret = pwrite(fd, sb, BTRFS_SUPER_INFO_SIZE, sb_bytenr);
228 if (ret < BTRFS_SUPER_INFO_SIZE)
229 ret = (ret < 0 ? -errno : -EIO);
230 else
231 ret = 0;
232 return ret;
233 }
234
235 /*
236 * Setup temporary superblock at cfg->super_bynter
237 * Needed info are extracted from cfg, and root_bytenr, chunk_bytenr
238 *
239 * For now sys chunk array will be empty and dev_item is empty too.
240 * They will be re-initialized at temp chunk tree setup.
241 */
242 static int setup_temp_super(int fd, struct btrfs_mkfs_config *cfg,
243 u64 root_bytenr, u64 chunk_bytenr)
244 {
245 unsigned char chunk_uuid[BTRFS_UUID_SIZE];
246 char super_buf[BTRFS_SUPER_INFO_SIZE];
247 struct btrfs_super_block *super = (struct btrfs_super_block *)super_buf;
248 int ret;
249
250 /*
251 * We rely on cfg->chunk_uuid and cfg->fs_uuid to pass uuid
252 * for other functions.
253 * Caller must allocate space for them
254 */
255 BUG_ON(!cfg->chunk_uuid || !cfg->fs_uuid);
256 memset(super_buf, 0, BTRFS_SUPER_INFO_SIZE);
257 cfg->num_bytes = round_down(cfg->num_bytes, cfg->sectorsize);
258
259 if (cfg->fs_uuid && *cfg->fs_uuid) {
260 if (uuid_parse(cfg->fs_uuid, super->fsid) != 0) {
261 error("cound not parse UUID: %s", cfg->fs_uuid);
262 ret = -EINVAL;
263 goto out;
264 }
265 if (!test_uuid_unique(cfg->fs_uuid)) {
266 error("non-unique UUID: %s", cfg->fs_uuid);
267 ret = -EINVAL;
268 goto out;
269 }
270 } else {
271 uuid_generate(super->fsid);
272 uuid_unparse(super->fsid, cfg->fs_uuid);
273 }
274 uuid_generate(chunk_uuid);
275 uuid_unparse(chunk_uuid, cfg->chunk_uuid);
276
277 btrfs_set_super_bytenr(super, cfg->super_bytenr);
278 btrfs_set_super_num_devices(super, 1);
279 btrfs_set_super_magic(super, BTRFS_MAGIC);
280 btrfs_set_super_generation(super, 1);
281 btrfs_set_super_root(super, root_bytenr);
282 btrfs_set_super_chunk_root(super, chunk_bytenr);
283 btrfs_set_super_total_bytes(super, cfg->num_bytes);
284 /*
285 * Temporary filesystem will only have 6 tree roots:
286 * chunk tree, root tree, extent_tree, device tree, fs tree
287 * and csum tree.
288 */
289 btrfs_set_super_bytes_used(super, 6 * cfg->nodesize);
290 btrfs_set_super_sectorsize(super, cfg->sectorsize);
291 btrfs_set_super_leafsize(super, cfg->nodesize);
292 btrfs_set_super_nodesize(super, cfg->nodesize);
293 btrfs_set_super_stripesize(super, cfg->stripesize);
294 btrfs_set_super_csum_type(super, BTRFS_CSUM_TYPE_CRC32);
295 btrfs_set_super_chunk_root(super, chunk_bytenr);
296 btrfs_set_super_cache_generation(super, -1);
297 btrfs_set_super_incompat_flags(super, cfg->features);
298 if (cfg->label)
299 __strncpy_null(super->label, cfg->label, BTRFS_LABEL_SIZE - 1);
300
301 /* Sys chunk array will be re-initialized at chunk tree init time */
302 super->sys_chunk_array_size = 0;
303
304 ret = write_temp_super(fd, super, cfg->super_bytenr);
305 out:
306 return ret;
307 }
308
309 /*
310 * Setup an extent buffer for tree block.
311 */
312 static int setup_temp_extent_buffer(struct extent_buffer *buf,
313 struct btrfs_mkfs_config *cfg,
314 u64 bytenr, u64 owner)
315 {
316 unsigned char fsid[BTRFS_FSID_SIZE];
317 unsigned char chunk_uuid[BTRFS_UUID_SIZE];
318 int ret;
319
320 /* We rely on cfg->fs_uuid and chunk_uuid to fsid and chunk uuid */
321 BUG_ON(!cfg->fs_uuid || !cfg->chunk_uuid);
322 ret = uuid_parse(cfg->fs_uuid, fsid);
323 if (ret)
324 return -EINVAL;
325 ret = uuid_parse(cfg->chunk_uuid, chunk_uuid);
326 if (ret)
327 return -EINVAL;
328
329 memset(buf->data, 0, cfg->nodesize);
330 buf->len = cfg->nodesize;
331 btrfs_set_header_bytenr(buf, bytenr);
332 btrfs_set_header_generation(buf, 1);
333 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
334 btrfs_set_header_owner(buf, owner);
335 btrfs_set_header_flags(buf, BTRFS_HEADER_FLAG_WRITTEN);
336 write_extent_buffer(buf, chunk_uuid, btrfs_header_chunk_tree_uuid(buf),
337 BTRFS_UUID_SIZE);
338 write_extent_buffer(buf, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
339 return 0;
340 }
341
342 static inline int write_temp_extent_buffer(int fd, struct extent_buffer *buf,
343 u64 bytenr)
344 {
345 int ret;
346
347 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
348
349 /* Temporary extent buffer is always mapped 1:1 on disk */
350 ret = pwrite(fd, buf->data, buf->len, bytenr);
351 if (ret < buf->len)
352 ret = (ret < 0 ? ret : -EIO);
353 else
354 ret = 0;
355 return ret;
356 }
357
358 /*
359 * Insert a root item for temporary tree root
360 *
361 * Only used in make_btrfs_v2().
362 */
363 static void insert_temp_root_item(struct extent_buffer *buf,
364 struct btrfs_mkfs_config *cfg,
365 int *slot, u32 *itemoff, u64 objectid,
366 u64 bytenr)
367 {
368 struct btrfs_root_item root_item;
369 struct btrfs_inode_item *inode_item;
370 struct btrfs_disk_key disk_key;
371
372 btrfs_set_header_nritems(buf, *slot + 1);
373 (*itemoff) -= sizeof(root_item);
374 memset(&root_item, 0, sizeof(root_item));
375 inode_item = &root_item.inode;
376 btrfs_set_stack_inode_generation(inode_item, 1);
377 btrfs_set_stack_inode_size(inode_item, 3);
378 btrfs_set_stack_inode_nlink(inode_item, 1);
379 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
380 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
381 btrfs_set_root_refs(&root_item, 1);
382 btrfs_set_root_used(&root_item, cfg->nodesize);
383 btrfs_set_root_generation(&root_item, 1);
384 btrfs_set_root_bytenr(&root_item, bytenr);
385
386 memset(&disk_key, 0, sizeof(disk_key));
387 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
388 btrfs_set_disk_key_objectid(&disk_key, objectid);
389 btrfs_set_disk_key_offset(&disk_key, 0);
390
391 btrfs_set_item_key(buf, &disk_key, *slot);
392 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
393 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(root_item));
394 write_extent_buffer(buf, &root_item,
395 btrfs_item_ptr_offset(buf, *slot),
396 sizeof(root_item));
397 (*slot)++;
398 }
399
400 static int setup_temp_root_tree(int fd, struct btrfs_mkfs_config *cfg,
401 u64 root_bytenr, u64 extent_bytenr,
402 u64 dev_bytenr, u64 fs_bytenr, u64 csum_bytenr)
403 {
404 struct extent_buffer *buf = NULL;
405 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
406 int slot = 0;
407 int ret;
408
409 /*
410 * Provided bytenr must in ascending order, or tree root will have a
411 * bad key order.
412 */
413 BUG_ON(!(root_bytenr < extent_bytenr && extent_bytenr < dev_bytenr &&
414 dev_bytenr < fs_bytenr && fs_bytenr < csum_bytenr));
415 buf = malloc(sizeof(*buf) + cfg->nodesize);
416 if (!buf)
417 return -ENOMEM;
418
419 ret = setup_temp_extent_buffer(buf, cfg, root_bytenr,
420 BTRFS_ROOT_TREE_OBJECTID);
421 if (ret < 0)
422 goto out;
423
424 insert_temp_root_item(buf, cfg, &slot, &itemoff,
425 BTRFS_EXTENT_TREE_OBJECTID, extent_bytenr);
426 insert_temp_root_item(buf, cfg, &slot, &itemoff,
427 BTRFS_DEV_TREE_OBJECTID, dev_bytenr);
428 insert_temp_root_item(buf, cfg, &slot, &itemoff,
429 BTRFS_FS_TREE_OBJECTID, fs_bytenr);
430 insert_temp_root_item(buf, cfg, &slot, &itemoff,
431 BTRFS_CSUM_TREE_OBJECTID, csum_bytenr);
432
433 ret = write_temp_extent_buffer(fd, buf, root_bytenr);
434 out:
435 free(buf);
436 return ret;
437 }
438
439 static int insert_temp_dev_item(int fd, struct extent_buffer *buf,
440 struct btrfs_mkfs_config *cfg,
441 int *slot, u32 *itemoff)
442 {
443 struct btrfs_disk_key disk_key;
444 struct btrfs_dev_item *dev_item;
445 char super_buf[BTRFS_SUPER_INFO_SIZE];
446 unsigned char dev_uuid[BTRFS_UUID_SIZE];
447 unsigned char fsid[BTRFS_FSID_SIZE];
448 struct btrfs_super_block *super = (struct btrfs_super_block *)super_buf;
449 int ret;
450
451 ret = pread(fd, super_buf, BTRFS_SUPER_INFO_SIZE, cfg->super_bytenr);
452 if (ret < BTRFS_SUPER_INFO_SIZE) {
453 ret = (ret < 0 ? -errno : -EIO);
454 goto out;
455 }
456
457 btrfs_set_header_nritems(buf, *slot + 1);
458 (*itemoff) -= sizeof(*dev_item);
459 /* setup device item 1, 0 is for replace case */
460 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
461 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
462 btrfs_set_disk_key_offset(&disk_key, 1);
463 btrfs_set_item_key(buf, &disk_key, *slot);
464 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
465 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(*dev_item));
466
467 dev_item = btrfs_item_ptr(buf, *slot, struct btrfs_dev_item);
468 /* Generate device uuid */
469 uuid_generate(dev_uuid);
470 write_extent_buffer(buf, dev_uuid,
471 (unsigned long)btrfs_device_uuid(dev_item),
472 BTRFS_UUID_SIZE);
473 uuid_parse(cfg->fs_uuid, fsid);
474 write_extent_buffer(buf, fsid,
475 (unsigned long)btrfs_device_fsid(dev_item),
476 BTRFS_FSID_SIZE);
477 btrfs_set_device_id(buf, dev_item, 1);
478 btrfs_set_device_generation(buf, dev_item, 0);
479 btrfs_set_device_total_bytes(buf, dev_item, cfg->num_bytes);
480 /*
481 * The number must match the initial SYSTEM and META chunk size
482 */
483 btrfs_set_device_bytes_used(buf, dev_item,
484 BTRFS_MKFS_SYSTEM_GROUP_SIZE +
485 BTRFS_CONVERT_META_GROUP_SIZE);
486 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
487 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
488 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
489 btrfs_set_device_type(buf, dev_item, 0);
490
491 /* Super dev_item is not complete, copy the complete one to sb */
492 read_extent_buffer(buf, &super->dev_item, (unsigned long)dev_item,
493 sizeof(*dev_item));
494 ret = write_temp_super(fd, super, cfg->super_bytenr);
495 (*slot)++;
496 out:
497 return ret;
498 }
499
500 static int insert_temp_chunk_item(int fd, struct extent_buffer *buf,
501 struct btrfs_mkfs_config *cfg,
502 int *slot, u32 *itemoff, u64 start, u64 len,
503 u64 type)
504 {
505 struct btrfs_chunk *chunk;
506 struct btrfs_disk_key disk_key;
507 char super_buf[BTRFS_SUPER_INFO_SIZE];
508 struct btrfs_super_block *sb = (struct btrfs_super_block *)super_buf;
509 int ret = 0;
510
511 ret = pread(fd, super_buf, BTRFS_SUPER_INFO_SIZE,
512 cfg->super_bytenr);
513 if (ret < BTRFS_SUPER_INFO_SIZE) {
514 ret = (ret < 0 ? ret : -EIO);
515 return ret;
516 }
517
518 btrfs_set_header_nritems(buf, *slot + 1);
519 (*itemoff) -= btrfs_chunk_item_size(1);
520 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
521 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
522 btrfs_set_disk_key_offset(&disk_key, start);
523 btrfs_set_item_key(buf, &disk_key, *slot);
524 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
525 btrfs_set_item_size(buf, btrfs_item_nr(*slot),
526 btrfs_chunk_item_size(1));
527
528 chunk = btrfs_item_ptr(buf, *slot, struct btrfs_chunk);
529 btrfs_set_chunk_length(buf, chunk, len);
530 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
531 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
532 btrfs_set_chunk_type(buf, chunk, type);
533 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
534 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
535 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
536 btrfs_set_chunk_num_stripes(buf, chunk, 1);
537 /* TODO: Support DUP profile for system chunk */
538 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
539 /* We are doing 1:1 mapping, so start is its dev offset */
540 btrfs_set_stripe_offset_nr(buf, chunk, 0, start);
541 write_extent_buffer(buf, &sb->dev_item.uuid,
542 (unsigned long)btrfs_stripe_dev_uuid_nr(chunk, 0),
543 BTRFS_UUID_SIZE);
544 (*slot)++;
545
546 /*
547 * If it's system chunk, also copy it to super block.
548 */
549 if (type & BTRFS_BLOCK_GROUP_SYSTEM) {
550 char *cur;
551
552 cur = (char *)sb->sys_chunk_array + sb->sys_chunk_array_size;
553 memcpy(cur, &disk_key, sizeof(disk_key));
554 cur += sizeof(disk_key);
555 read_extent_buffer(buf, cur, (unsigned long int)chunk,
556 btrfs_chunk_item_size(1));
557 sb->sys_chunk_array_size += btrfs_chunk_item_size(1) +
558 sizeof(disk_key);
559
560 ret = write_temp_super(fd, sb, cfg->super_bytenr);
561 }
562 return ret;
563 }
564
565 static int setup_temp_chunk_tree(int fd, struct btrfs_mkfs_config *cfg,
566 u64 sys_chunk_start, u64 meta_chunk_start,
567 u64 chunk_bytenr)
568 {
569 struct extent_buffer *buf = NULL;
570 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
571 int slot = 0;
572 int ret;
573
574 /* Must ensure SYS chunk starts before META chunk */
575 BUG_ON(meta_chunk_start < sys_chunk_start);
576 buf = malloc(sizeof(*buf) + cfg->nodesize);
577 if (!buf)
578 return -ENOMEM;
579 ret = setup_temp_extent_buffer(buf, cfg, chunk_bytenr,
580 BTRFS_CHUNK_TREE_OBJECTID);
581 if (ret < 0)
582 goto out;
583
584 ret = insert_temp_dev_item(fd, buf, cfg, &slot, &itemoff);
585 if (ret < 0)
586 goto out;
587 ret = insert_temp_chunk_item(fd, buf, cfg, &slot, &itemoff,
588 sys_chunk_start,
589 BTRFS_MKFS_SYSTEM_GROUP_SIZE,
590 BTRFS_BLOCK_GROUP_SYSTEM);
591 if (ret < 0)
592 goto out;
593 ret = insert_temp_chunk_item(fd, buf, cfg, &slot, &itemoff,
594 meta_chunk_start,
595 BTRFS_CONVERT_META_GROUP_SIZE,
596 BTRFS_BLOCK_GROUP_METADATA);
597 if (ret < 0)
598 goto out;
599 ret = write_temp_extent_buffer(fd, buf, chunk_bytenr);
600
601 out:
602 free(buf);
603 return ret;
604 }
605
606 /*
607 * Improved version of make_btrfs().
608 *
609 * This one will
610 * 1) Do chunk allocation to avoid used data
611 * And after this function, extent type matches chunk type
612 * 2) Better structured code
613 * No super long hand written codes to initialized all tree blocks
614 * Split into small blocks and reuse codes.
615 * TODO: Reuse tree operation facilities by introducing new flags
616 */
617 static int make_convert_btrfs(int fd, struct btrfs_mkfs_config *cfg,
618 struct btrfs_convert_context *cctx)
619 {
620 struct cache_tree *free = &cctx->free;
621 struct cache_tree *used = &cctx->used;
622 u64 sys_chunk_start;
623 u64 meta_chunk_start;
624 /* chunk tree bytenr, in system chunk */
625 u64 chunk_bytenr;
626 /* metadata trees bytenr, in metadata chunk */
627 u64 root_bytenr;
628 u64 extent_bytenr;
629 u64 dev_bytenr;
630 u64 fs_bytenr;
631 u64 csum_bytenr;
632 int ret;
633
634 /* Shouldn't happen */
635 BUG_ON(cache_tree_empty(used));
636
637 /*
638 * reserve space for temporary superblock first
639 * Here we allocate a little larger space, to keep later
640 * free space will be STRIPE_LEN aligned
641 */
642 ret = reserve_free_space(free, BTRFS_STRIPE_LEN,
643 &cfg->super_bytenr);
644 if (ret < 0)
645 goto out;
646
647 /*
648 * Then reserve system chunk space
649 * TODO: Change system group size depending on cctx->total_bytes.
650 * If using current 4M, it can only handle less than one TB for
651 * worst case and then run out of sys space.
652 */
653 ret = reserve_free_space(free, BTRFS_MKFS_SYSTEM_GROUP_SIZE,
654 &sys_chunk_start);
655 if (ret < 0)
656 goto out;
657 ret = reserve_free_space(free, BTRFS_CONVERT_META_GROUP_SIZE,
658 &meta_chunk_start);
659 if (ret < 0)
660 goto out;
661
662 /*
663 * Allocated meta/sys chunks will be mapped 1:1 with device offset.
664 *
665 * Inside the allocated metadata chunk, the layout will be:
666 * | offset | contents |
667 * -------------------------------------
668 * | +0 | tree root |
669 * | +nodesize | extent root |
670 * | +nodesize * 2 | device root |
671 * | +nodesize * 3 | fs tree |
672 * | +nodesize * 4 | csum tree |
673 * -------------------------------------
674 * Inside the allocated system chunk, the layout will be:
675 * | offset | contents |
676 * -------------------------------------
677 * | +0 | chunk root |
678 * -------------------------------------
679 */
680 chunk_bytenr = sys_chunk_start;
681 root_bytenr = meta_chunk_start;
682 extent_bytenr = meta_chunk_start + cfg->nodesize;
683 dev_bytenr = meta_chunk_start + cfg->nodesize * 2;
684 fs_bytenr = meta_chunk_start + cfg->nodesize * 3;
685 csum_bytenr = meta_chunk_start + cfg->nodesize * 4;
686
687 ret = setup_temp_super(fd, cfg, root_bytenr, chunk_bytenr);
688 if (ret < 0)
689 goto out;
690
691 ret = setup_temp_root_tree(fd, cfg, root_bytenr, extent_bytenr,
692 dev_bytenr, fs_bytenr, csum_bytenr);
693 if (ret < 0)
694 goto out;
695 ret = setup_temp_chunk_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
696 chunk_bytenr);
697 if (ret < 0)
698 goto out;
699
700 out:
701 return ret;
702 }
703
704 /*
705 * @fs_uuid - if NULL, generates a UUID, returns back the new filesystem UUID
706 */
707 int make_btrfs(int fd, struct btrfs_mkfs_config *cfg,
708 struct btrfs_convert_context *cctx)
709 {
710 struct btrfs_super_block super;
711 struct extent_buffer *buf;
712 struct btrfs_root_item root_item;
713 struct btrfs_disk_key disk_key;
714 struct btrfs_extent_item *extent_item;
715 struct btrfs_inode_item *inode_item;
716 struct btrfs_chunk *chunk;
717 struct btrfs_dev_item *dev_item;
718 struct btrfs_dev_extent *dev_extent;
719 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
720 u8 *ptr;
721 int i;
722 int ret;
723 u32 itemoff;
724 u32 nritems = 0;
725 u64 first_free;
726 u64 ref_root;
727 u32 array_size;
728 u32 item_size;
729 int skinny_metadata = !!(cfg->features &
730 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA);
731 u64 num_bytes;
732
733 if (cctx)
734 return make_convert_btrfs(fd, cfg, cctx);
735 buf = malloc(sizeof(*buf) + max(cfg->sectorsize, cfg->nodesize));
736 if (!buf)
737 return -ENOMEM;
738
739 first_free = BTRFS_SUPER_INFO_OFFSET + cfg->sectorsize * 2 - 1;
740 first_free &= ~((u64)cfg->sectorsize - 1);
741
742 memset(&super, 0, sizeof(super));
743
744 num_bytes = (cfg->num_bytes / cfg->sectorsize) * cfg->sectorsize;
745 if (cfg->fs_uuid && *cfg->fs_uuid) {
746 if (uuid_parse(cfg->fs_uuid, super.fsid) != 0) {
747 error("cannot not parse UUID: %s", cfg->fs_uuid);
748 ret = -EINVAL;
749 goto out;
750 }
751 if (!test_uuid_unique(cfg->fs_uuid)) {
752 error("non-unique UUID: %s", cfg->fs_uuid);
753 ret = -EBUSY;
754 goto out;
755 }
756 } else {
757 uuid_generate(super.fsid);
758 if (cfg->fs_uuid)
759 uuid_unparse(super.fsid, cfg->fs_uuid);
760 }
761 uuid_generate(super.dev_item.uuid);
762 uuid_generate(chunk_tree_uuid);
763
764 btrfs_set_super_bytenr(&super, cfg->blocks[0]);
765 btrfs_set_super_num_devices(&super, 1);
766 btrfs_set_super_magic(&super, BTRFS_MAGIC);
767 btrfs_set_super_generation(&super, 1);
768 btrfs_set_super_root(&super, cfg->blocks[1]);
769 btrfs_set_super_chunk_root(&super, cfg->blocks[3]);
770 btrfs_set_super_total_bytes(&super, num_bytes);
771 btrfs_set_super_bytes_used(&super, 6 * cfg->nodesize);
772 btrfs_set_super_sectorsize(&super, cfg->sectorsize);
773 btrfs_set_super_leafsize(&super, cfg->nodesize);
774 btrfs_set_super_nodesize(&super, cfg->nodesize);
775 btrfs_set_super_stripesize(&super, cfg->stripesize);
776 btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
777 btrfs_set_super_chunk_root_generation(&super, 1);
778 btrfs_set_super_cache_generation(&super, -1);
779 btrfs_set_super_incompat_flags(&super, cfg->features);
780 if (cfg->label)
781 __strncpy_null(super.label, cfg->label, BTRFS_LABEL_SIZE - 1);
782
783 /* create the tree of root objects */
784 memset(buf->data, 0, cfg->nodesize);
785 buf->len = cfg->nodesize;
786 btrfs_set_header_bytenr(buf, cfg->blocks[1]);
787 btrfs_set_header_nritems(buf, 4);
788 btrfs_set_header_generation(buf, 1);
789 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
790 btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
791 write_extent_buffer(buf, super.fsid, btrfs_header_fsid(),
792 BTRFS_FSID_SIZE);
793
794 write_extent_buffer(buf, chunk_tree_uuid,
795 btrfs_header_chunk_tree_uuid(buf),
796 BTRFS_UUID_SIZE);
797
798 /* create the items for the root tree */
799 memset(&root_item, 0, sizeof(root_item));
800 inode_item = &root_item.inode;
801 btrfs_set_stack_inode_generation(inode_item, 1);
802 btrfs_set_stack_inode_size(inode_item, 3);
803 btrfs_set_stack_inode_nlink(inode_item, 1);
804 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
805 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
806 btrfs_set_root_refs(&root_item, 1);
807 btrfs_set_root_used(&root_item, cfg->nodesize);
808 btrfs_set_root_generation(&root_item, 1);
809
810 memset(&disk_key, 0, sizeof(disk_key));
811 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
812 btrfs_set_disk_key_offset(&disk_key, 0);
813 nritems = 0;
814
815 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - sizeof(root_item);
816 btrfs_set_root_bytenr(&root_item, cfg->blocks[2]);
817 btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
818 btrfs_set_item_key(buf, &disk_key, nritems);
819 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
820 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
821 sizeof(root_item));
822 write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
823 nritems), sizeof(root_item));
824 nritems++;
825
826 itemoff = itemoff - sizeof(root_item);
827 btrfs_set_root_bytenr(&root_item, cfg->blocks[4]);
828 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
829 btrfs_set_item_key(buf, &disk_key, nritems);
830 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
831 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
832 sizeof(root_item));
833 write_extent_buffer(buf, &root_item,
834 btrfs_item_ptr_offset(buf, nritems),
835 sizeof(root_item));
836 nritems++;
837
838 itemoff = itemoff - sizeof(root_item);
839 btrfs_set_root_bytenr(&root_item, cfg->blocks[5]);
840 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
841 btrfs_set_item_key(buf, &disk_key, nritems);
842 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
843 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
844 sizeof(root_item));
845 write_extent_buffer(buf, &root_item,
846 btrfs_item_ptr_offset(buf, nritems),
847 sizeof(root_item));
848 nritems++;
849
850 itemoff = itemoff - sizeof(root_item);
851 btrfs_set_root_bytenr(&root_item, cfg->blocks[6]);
852 btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_TREE_OBJECTID);
853 btrfs_set_item_key(buf, &disk_key, nritems);
854 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
855 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
856 sizeof(root_item));
857 write_extent_buffer(buf, &root_item,
858 btrfs_item_ptr_offset(buf, nritems),
859 sizeof(root_item));
860 nritems++;
861
862
863 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
864 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[1]);
865 if (ret != cfg->nodesize) {
866 ret = (ret < 0 ? -errno : -EIO);
867 goto out;
868 }
869
870 /* create the items for the extent tree */
871 memset(buf->data + sizeof(struct btrfs_header), 0,
872 cfg->nodesize - sizeof(struct btrfs_header));
873 nritems = 0;
874 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
875 for (i = 1; i < 7; i++) {
876 item_size = sizeof(struct btrfs_extent_item);
877 if (!skinny_metadata)
878 item_size += sizeof(struct btrfs_tree_block_info);
879
880 BUG_ON(cfg->blocks[i] < first_free);
881 BUG_ON(cfg->blocks[i] < cfg->blocks[i - 1]);
882
883 /* create extent item */
884 itemoff -= item_size;
885 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
886 if (skinny_metadata) {
887 btrfs_set_disk_key_type(&disk_key,
888 BTRFS_METADATA_ITEM_KEY);
889 btrfs_set_disk_key_offset(&disk_key, 0);
890 } else {
891 btrfs_set_disk_key_type(&disk_key,
892 BTRFS_EXTENT_ITEM_KEY);
893 btrfs_set_disk_key_offset(&disk_key, cfg->nodesize);
894 }
895 btrfs_set_item_key(buf, &disk_key, nritems);
896 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
897 itemoff);
898 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
899 item_size);
900 extent_item = btrfs_item_ptr(buf, nritems,
901 struct btrfs_extent_item);
902 btrfs_set_extent_refs(buf, extent_item, 1);
903 btrfs_set_extent_generation(buf, extent_item, 1);
904 btrfs_set_extent_flags(buf, extent_item,
905 BTRFS_EXTENT_FLAG_TREE_BLOCK);
906 nritems++;
907
908 /* create extent ref */
909 ref_root = reference_root_table[i];
910 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
911 btrfs_set_disk_key_offset(&disk_key, ref_root);
912 btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
913 btrfs_set_item_key(buf, &disk_key, nritems);
914 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
915 itemoff);
916 btrfs_set_item_size(buf, btrfs_item_nr(nritems), 0);
917 nritems++;
918 }
919 btrfs_set_header_bytenr(buf, cfg->blocks[2]);
920 btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
921 btrfs_set_header_nritems(buf, nritems);
922 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
923 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[2]);
924 if (ret != cfg->nodesize) {
925 ret = (ret < 0 ? -errno : -EIO);
926 goto out;
927 }
928
929 /* create the chunk tree */
930 memset(buf->data + sizeof(struct btrfs_header), 0,
931 cfg->nodesize - sizeof(struct btrfs_header));
932 nritems = 0;
933 item_size = sizeof(*dev_item);
934 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - item_size;
935
936 /* first device 1 (there is no device 0) */
937 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
938 btrfs_set_disk_key_offset(&disk_key, 1);
939 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
940 btrfs_set_item_key(buf, &disk_key, nritems);
941 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
942 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
943
944 dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
945 btrfs_set_device_id(buf, dev_item, 1);
946 btrfs_set_device_generation(buf, dev_item, 0);
947 btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
948 btrfs_set_device_bytes_used(buf, dev_item,
949 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
950 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
951 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
952 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
953 btrfs_set_device_type(buf, dev_item, 0);
954
955 write_extent_buffer(buf, super.dev_item.uuid,
956 (unsigned long)btrfs_device_uuid(dev_item),
957 BTRFS_UUID_SIZE);
958 write_extent_buffer(buf, super.fsid,
959 (unsigned long)btrfs_device_fsid(dev_item),
960 BTRFS_UUID_SIZE);
961 read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
962 sizeof(*dev_item));
963
964 nritems++;
965 item_size = btrfs_chunk_item_size(1);
966 itemoff = itemoff - item_size;
967
968 /* then we have chunk 0 */
969 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
970 btrfs_set_disk_key_offset(&disk_key, 0);
971 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
972 btrfs_set_item_key(buf, &disk_key, nritems);
973 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
974 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
975
976 chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
977 btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
978 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
979 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
980 btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
981 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
982 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
983 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
984 btrfs_set_chunk_num_stripes(buf, chunk, 1);
985 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
986 btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
987 nritems++;
988
989 write_extent_buffer(buf, super.dev_item.uuid,
990 (unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
991 BTRFS_UUID_SIZE);
992
993 /* copy the key for the chunk to the system array */
994 ptr = super.sys_chunk_array;
995 array_size = sizeof(disk_key);
996
997 memcpy(ptr, &disk_key, sizeof(disk_key));
998 ptr += sizeof(disk_key);
999
1000 /* copy the chunk to the system array */
1001 read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
1002 array_size += item_size;
1003 ptr += item_size;
1004 btrfs_set_super_sys_array_size(&super, array_size);
1005
1006 btrfs_set_header_bytenr(buf, cfg->blocks[3]);
1007 btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
1008 btrfs_set_header_nritems(buf, nritems);
1009 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1010 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[3]);
1011 if (ret != cfg->nodesize) {
1012 ret = (ret < 0 ? -errno : -EIO);
1013 goto out;
1014 }
1015
1016 /* create the device tree */
1017 memset(buf->data + sizeof(struct btrfs_header), 0,
1018 cfg->nodesize - sizeof(struct btrfs_header));
1019 nritems = 0;
1020 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) -
1021 sizeof(struct btrfs_dev_extent);
1022
1023 btrfs_set_disk_key_objectid(&disk_key, 1);
1024 btrfs_set_disk_key_offset(&disk_key, 0);
1025 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
1026 btrfs_set_item_key(buf, &disk_key, nritems);
1027 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1028 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1029 sizeof(struct btrfs_dev_extent));
1030 dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
1031 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
1032 BTRFS_CHUNK_TREE_OBJECTID);
1033 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
1034 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1035 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
1036
1037 write_extent_buffer(buf, chunk_tree_uuid,
1038 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
1039 BTRFS_UUID_SIZE);
1040
1041 btrfs_set_dev_extent_length(buf, dev_extent,
1042 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1043 nritems++;
1044
1045 btrfs_set_header_bytenr(buf, cfg->blocks[4]);
1046 btrfs_set_header_owner(buf, BTRFS_DEV_TREE_OBJECTID);
1047 btrfs_set_header_nritems(buf, nritems);
1048 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1049 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[4]);
1050 if (ret != cfg->nodesize) {
1051 ret = (ret < 0 ? -errno : -EIO);
1052 goto out;
1053 }
1054
1055 /* create the FS root */
1056 memset(buf->data + sizeof(struct btrfs_header), 0,
1057 cfg->nodesize - sizeof(struct btrfs_header));
1058 btrfs_set_header_bytenr(buf, cfg->blocks[5]);
1059 btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
1060 btrfs_set_header_nritems(buf, 0);
1061 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1062 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[5]);
1063 if (ret != cfg->nodesize) {
1064 ret = (ret < 0 ? -errno : -EIO);
1065 goto out;
1066 }
1067 /* finally create the csum root */
1068 memset(buf->data + sizeof(struct btrfs_header), 0,
1069 cfg->nodesize - sizeof(struct btrfs_header));
1070 btrfs_set_header_bytenr(buf, cfg->blocks[6]);
1071 btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
1072 btrfs_set_header_nritems(buf, 0);
1073 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1074 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[6]);
1075 if (ret != cfg->nodesize) {
1076 ret = (ret < 0 ? -errno : -EIO);
1077 goto out;
1078 }
1079
1080 /* and write out the super block */
1081 BUG_ON(sizeof(super) > cfg->sectorsize);
1082 memset(buf->data, 0, BTRFS_SUPER_INFO_SIZE);
1083 memcpy(buf->data, &super, sizeof(super));
1084 buf->len = BTRFS_SUPER_INFO_SIZE;
1085 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1086 ret = pwrite(fd, buf->data, BTRFS_SUPER_INFO_SIZE, cfg->blocks[0]);
1087 if (ret != BTRFS_SUPER_INFO_SIZE) {
1088 ret = (ret < 0 ? -errno : -EIO);
1089 goto out;
1090 }
1091
1092 ret = 0;
1093
1094 out:
1095 free(buf);
1096 return ret;
1097 }
1098
1099 static const struct btrfs_fs_feature {
1100 const char *name;
1101 u64 flag;
1102 const char *desc;
1103 } mkfs_features[] = {
1104 { "mixed-bg", BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS,
1105 "mixed data and metadata block groups" },
1106 { "extref", BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF,
1107 "increased hardlink limit per file to 65536" },
1108 { "raid56", BTRFS_FEATURE_INCOMPAT_RAID56,
1109 "raid56 extended format" },
1110 { "skinny-metadata", BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA,
1111 "reduced-size metadata extent refs" },
1112 { "no-holes", BTRFS_FEATURE_INCOMPAT_NO_HOLES,
1113 "no explicit hole extents for files" },
1114 /* Keep this one last */
1115 { "list-all", BTRFS_FEATURE_LIST_ALL, NULL }
1116 };
1117
1118 static int parse_one_fs_feature(const char *name, u64 *flags)
1119 {
1120 int i;
1121 int found = 0;
1122
1123 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1124 if (name[0] == '^' &&
1125 !strcmp(mkfs_features[i].name, name + 1)) {
1126 *flags &= ~ mkfs_features[i].flag;
1127 found = 1;
1128 } else if (!strcmp(mkfs_features[i].name, name)) {
1129 *flags |= mkfs_features[i].flag;
1130 found = 1;
1131 }
1132 }
1133
1134 return !found;
1135 }
1136
1137 void btrfs_parse_features_to_string(char *buf, u64 flags)
1138 {
1139 int i;
1140
1141 buf[0] = 0;
1142
1143 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1144 if (flags & mkfs_features[i].flag) {
1145 if (*buf)
1146 strcat(buf, ", ");
1147 strcat(buf, mkfs_features[i].name);
1148 }
1149 }
1150 }
1151
1152 void btrfs_process_fs_features(u64 flags)
1153 {
1154 int i;
1155
1156 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1157 if (flags & mkfs_features[i].flag) {
1158 printf("Turning ON incompat feature '%s': %s\n",
1159 mkfs_features[i].name,
1160 mkfs_features[i].desc);
1161 }
1162 }
1163 }
1164
1165 void btrfs_list_all_fs_features(u64 mask_disallowed)
1166 {
1167 int i;
1168
1169 fprintf(stderr, "Filesystem features available:\n");
1170 for (i = 0; i < ARRAY_SIZE(mkfs_features) - 1; i++) {
1171 char *is_default = "";
1172
1173 if (mkfs_features[i].flag & mask_disallowed)
1174 continue;
1175 if (mkfs_features[i].flag & BTRFS_MKFS_DEFAULT_FEATURES)
1176 is_default = ", default";
1177 fprintf(stderr, "%-20s- %s (0x%llx%s)\n",
1178 mkfs_features[i].name,
1179 mkfs_features[i].desc,
1180 mkfs_features[i].flag,
1181 is_default);
1182 }
1183 }
1184
1185 /*
1186 * Return NULL if all features were parsed fine, otherwise return the name of
1187 * the first unparsed.
1188 */
1189 char* btrfs_parse_fs_features(char *namelist, u64 *flags)
1190 {
1191 char *this_char;
1192 char *save_ptr = NULL; /* Satisfy static checkers */
1193
1194 for (this_char = strtok_r(namelist, ",", &save_ptr);
1195 this_char != NULL;
1196 this_char = strtok_r(NULL, ",", &save_ptr)) {
1197 if (parse_one_fs_feature(this_char, flags))
1198 return this_char;
1199 }
1200
1201 return NULL;
1202 }
1203
1204 u64 btrfs_device_size(int fd, struct stat *st)
1205 {
1206 u64 size;
1207 if (S_ISREG(st->st_mode)) {
1208 return st->st_size;
1209 }
1210 if (!S_ISBLK(st->st_mode)) {
1211 return 0;
1212 }
1213 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
1214 return size;
1215 }
1216 return 0;
1217 }
1218
1219 static int zero_blocks(int fd, off_t start, size_t len)
1220 {
1221 char *buf = malloc(len);
1222 int ret = 0;
1223 ssize_t written;
1224
1225 if (!buf)
1226 return -ENOMEM;
1227 memset(buf, 0, len);
1228 written = pwrite(fd, buf, len, start);
1229 if (written != len)
1230 ret = -EIO;
1231 free(buf);
1232 return ret;
1233 }
1234
1235 #define ZERO_DEV_BYTES (2 * 1024 * 1024)
1236
1237 /* don't write outside the device by clamping the region to the device size */
1238 static int zero_dev_clamped(int fd, off_t start, ssize_t len, u64 dev_size)
1239 {
1240 off_t end = max(start, start + len);
1241
1242 #ifdef __sparc__
1243 /* and don't overwrite the disk labels on sparc */
1244 start = max(start, 1024);
1245 end = max(end, 1024);
1246 #endif
1247
1248 start = min_t(u64, start, dev_size);
1249 end = min_t(u64, end, dev_size);
1250
1251 return zero_blocks(fd, start, end - start);
1252 }
1253
1254 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
1255 struct btrfs_root *root, int fd, char *path,
1256 u64 device_total_bytes, u32 io_width, u32 io_align,
1257 u32 sectorsize)
1258 {
1259 struct btrfs_super_block *disk_super;
1260 struct btrfs_super_block *super = root->fs_info->super_copy;
1261 struct btrfs_device *device;
1262 struct btrfs_dev_item *dev_item;
1263 char *buf = NULL;
1264 u64 fs_total_bytes;
1265 u64 num_devs;
1266 int ret;
1267
1268 device_total_bytes = (device_total_bytes / sectorsize) * sectorsize;
1269
1270 device = kzalloc(sizeof(*device), GFP_NOFS);
1271 if (!device)
1272 goto err_nomem;
1273 buf = kzalloc(sectorsize, GFP_NOFS);
1274 if (!buf)
1275 goto err_nomem;
1276 BUG_ON(sizeof(*disk_super) > sectorsize);
1277
1278 disk_super = (struct btrfs_super_block *)buf;
1279 dev_item = &disk_super->dev_item;
1280
1281 uuid_generate(device->uuid);
1282 device->devid = 0;
1283 device->type = 0;
1284 device->io_width = io_width;
1285 device->io_align = io_align;
1286 device->sector_size = sectorsize;
1287 device->fd = fd;
1288 device->writeable = 1;
1289 device->total_bytes = device_total_bytes;
1290 device->bytes_used = 0;
1291 device->total_ios = 0;
1292 device->dev_root = root->fs_info->dev_root;
1293 device->name = strdup(path);
1294 if (!device->name)
1295 goto err_nomem;
1296
1297 INIT_LIST_HEAD(&device->dev_list);
1298 ret = btrfs_add_device(trans, root, device);
1299 BUG_ON(ret);
1300
1301 fs_total_bytes = btrfs_super_total_bytes(super) + device_total_bytes;
1302 btrfs_set_super_total_bytes(super, fs_total_bytes);
1303
1304 num_devs = btrfs_super_num_devices(super) + 1;
1305 btrfs_set_super_num_devices(super, num_devs);
1306
1307 memcpy(disk_super, super, sizeof(*disk_super));
1308
1309 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
1310 btrfs_set_stack_device_id(dev_item, device->devid);
1311 btrfs_set_stack_device_type(dev_item, device->type);
1312 btrfs_set_stack_device_io_align(dev_item, device->io_align);
1313 btrfs_set_stack_device_io_width(dev_item, device->io_width);
1314 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
1315 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
1316 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
1317 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
1318
1319 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
1320 BUG_ON(ret != sectorsize);
1321
1322 kfree(buf);
1323 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
1324 device->fs_devices = root->fs_info->fs_devices;
1325 return 0;
1326
1327 err_nomem:
1328 kfree(device);
1329 kfree(buf);
1330 return -ENOMEM;
1331 }
1332
1333 static int btrfs_wipe_existing_sb(int fd)
1334 {
1335 const char *off = NULL;
1336 size_t len = 0;
1337 loff_t offset;
1338 char buf[BUFSIZ];
1339 int ret = 0;
1340 blkid_probe pr = NULL;
1341
1342 pr = blkid_new_probe();
1343 if (!pr)
1344 return -1;
1345
1346 if (blkid_probe_set_device(pr, fd, 0, 0)) {
1347 ret = -1;
1348 goto out;
1349 }
1350
1351 ret = blkid_probe_lookup_value(pr, "SBMAGIC_OFFSET", &off, NULL);
1352 if (!ret)
1353 ret = blkid_probe_lookup_value(pr, "SBMAGIC", NULL, &len);
1354
1355 if (ret || len == 0 || off == NULL) {
1356 /*
1357 * If lookup fails, the probe did not find any values, eg. for
1358 * a file image or a loop device. Soft error.
1359 */
1360 ret = 1;
1361 goto out;
1362 }
1363
1364 offset = strtoll(off, NULL, 10);
1365 if (len > sizeof(buf))
1366 len = sizeof(buf);
1367
1368 memset(buf, 0, len);
1369 ret = pwrite(fd, buf, len, offset);
1370 if (ret < 0) {
1371 error("cannot wipe existing superblock: %s", strerror(errno));
1372 ret = -1;
1373 } else if (ret != len) {
1374 error("cannot wipe existing superblock: wrote %d of %zd", ret, len);
1375 ret = -1;
1376 }
1377 fsync(fd);
1378
1379 out:
1380 blkid_free_probe(pr);
1381 return ret;
1382 }
1383
1384 int btrfs_prepare_device(int fd, const char *file, int zero_end,
1385 u64 *block_count_ret, u64 max_block_count, int discard)
1386 {
1387 u64 block_count;
1388 struct stat st;
1389 int i, ret;
1390
1391 ret = fstat(fd, &st);
1392 if (ret < 0) {
1393 error("unable to stat %s: %s", file, strerror(errno));
1394 return 1;
1395 }
1396
1397 block_count = btrfs_device_size(fd, &st);
1398 if (block_count == 0) {
1399 error("unable to determine size of %s", file);
1400 return 1;
1401 }
1402 if (max_block_count)
1403 block_count = min(block_count, max_block_count);
1404
1405 if (discard) {
1406 /*
1407 * We intentionally ignore errors from the discard ioctl. It
1408 * is not necessary for the mkfs functionality but just an
1409 * optimization.
1410 */
1411 if (discard_range(fd, 0, 0) == 0) {
1412 printf("Performing full device TRIM (%s) ...\n",
1413 pretty_size(block_count));
1414 discard_blocks(fd, 0, block_count);
1415 }
1416 }
1417
1418 ret = zero_dev_clamped(fd, 0, ZERO_DEV_BYTES, block_count);
1419 for (i = 0 ; !ret && i < BTRFS_SUPER_MIRROR_MAX; i++)
1420 ret = zero_dev_clamped(fd, btrfs_sb_offset(i),
1421 BTRFS_SUPER_INFO_SIZE, block_count);
1422 if (!ret && zero_end)
1423 ret = zero_dev_clamped(fd, block_count - ZERO_DEV_BYTES,
1424 ZERO_DEV_BYTES, block_count);
1425
1426 if (ret < 0) {
1427 error("failed to zero device '%s': %s", file, strerror(-ret));
1428 return 1;
1429 }
1430
1431 ret = btrfs_wipe_existing_sb(fd);
1432 if (ret < 0) {
1433 error("cannot wipe superblocks on %s", file);
1434 return 1;
1435 }
1436
1437 *block_count_ret = block_count;
1438 return 0;
1439 }
1440
1441 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
1442 struct btrfs_root *root, u64 objectid)
1443 {
1444 int ret;
1445 struct btrfs_inode_item inode_item;
1446 time_t now = time(NULL);
1447
1448 memset(&inode_item, 0, sizeof(inode_item));
1449 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
1450 btrfs_set_stack_inode_size(&inode_item, 0);
1451 btrfs_set_stack_inode_nlink(&inode_item, 1);
1452 btrfs_set_stack_inode_nbytes(&inode_item, root->nodesize);
1453 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
1454 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
1455 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
1456 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
1457 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
1458 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
1459 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
1460 btrfs_set_stack_timespec_sec(&inode_item.otime, 0);
1461 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
1462
1463 if (root->fs_info->tree_root == root)
1464 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
1465
1466 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
1467 if (ret)
1468 goto error;
1469
1470 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
1471 if (ret)
1472 goto error;
1473
1474 btrfs_set_root_dirid(&root->root_item, objectid);
1475 ret = 0;
1476 error:
1477 return ret;
1478 }
1479
1480 /*
1481 * checks if a path is a block device node
1482 * Returns negative errno on failure, otherwise
1483 * returns 1 for blockdev, 0 for not-blockdev
1484 */
1485 int is_block_device(const char *path)
1486 {
1487 struct stat statbuf;
1488
1489 if (stat(path, &statbuf) < 0)
1490 return -errno;
1491
1492 return !!S_ISBLK(statbuf.st_mode);
1493 }
1494
1495 /*
1496 * check if given path is a mount point
1497 * return 1 if yes. 0 if no. -1 for error
1498 */
1499 int is_mount_point(const char *path)
1500 {
1501 FILE *f;
1502 struct mntent *mnt;
1503 int ret = 0;
1504
1505 f = setmntent("/proc/self/mounts", "r");
1506 if (f == NULL)
1507 return -1;
1508
1509 while ((mnt = getmntent(f)) != NULL) {
1510 if (strcmp(mnt->mnt_dir, path))
1511 continue;
1512 ret = 1;
1513 break;
1514 }
1515 endmntent(f);
1516 return ret;
1517 }
1518
1519 static int is_reg_file(const char *path)
1520 {
1521 struct stat statbuf;
1522
1523 if (stat(path, &statbuf) < 0)
1524 return -errno;
1525 return S_ISREG(statbuf.st_mode);
1526 }
1527
1528 /*
1529 * This function checks if the given input parameter is
1530 * an uuid or a path
1531 * return <0 : some error in the given input
1532 * return BTRFS_ARG_UNKNOWN: unknown input
1533 * return BTRFS_ARG_UUID: given input is uuid
1534 * return BTRFS_ARG_MNTPOINT: given input is path
1535 * return BTRFS_ARG_REG: given input is regular file
1536 * return BTRFS_ARG_BLKDEV: given input is block device
1537 */
1538 int check_arg_type(const char *input)
1539 {
1540 uuid_t uuid;
1541 char path[PATH_MAX];
1542
1543 if (!input)
1544 return -EINVAL;
1545
1546 if (realpath(input, path)) {
1547 if (is_block_device(path) == 1)
1548 return BTRFS_ARG_BLKDEV;
1549
1550 if (is_mount_point(path) == 1)
1551 return BTRFS_ARG_MNTPOINT;
1552
1553 if (is_reg_file(path))
1554 return BTRFS_ARG_REG;
1555
1556 return BTRFS_ARG_UNKNOWN;
1557 }
1558
1559 if (strlen(input) == (BTRFS_UUID_UNPARSED_SIZE - 1) &&
1560 !uuid_parse(input, uuid))
1561 return BTRFS_ARG_UUID;
1562
1563 return BTRFS_ARG_UNKNOWN;
1564 }
1565
1566 /*
1567 * Find the mount point for a mounted device.
1568 * On success, returns 0 with mountpoint in *mp.
1569 * On failure, returns -errno (not mounted yields -EINVAL)
1570 * Is noisy on failures, expects to be given a mounted device.
1571 */
1572 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size)
1573 {
1574 int ret;
1575 int fd = -1;
1576
1577 ret = is_block_device(dev);
1578 if (ret <= 0) {
1579 if (!ret) {
1580 error("not a block device: %s", dev);
1581 ret = -EINVAL;
1582 } else {
1583 error("cannot check %s: %s", dev, strerror(-ret));
1584 }
1585 goto out;
1586 }
1587
1588 fd = open(dev, O_RDONLY);
1589 if (fd < 0) {
1590 ret = -errno;
1591 error("cannot open %s: %s", dev, strerror(errno));
1592 goto out;
1593 }
1594
1595 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
1596 if (!ret) {
1597 ret = -EINVAL;
1598 } else { /* mounted, all good */
1599 ret = 0;
1600 }
1601 out:
1602 if (fd != -1)
1603 close(fd);
1604 return ret;
1605 }
1606
1607 /*
1608 * Given a pathname, return a filehandle to:
1609 * the original pathname or,
1610 * if the pathname is a mounted btrfs device, to its mountpoint.
1611 *
1612 * On error, return -1, errno should be set.
1613 */
1614 int open_path_or_dev_mnt(const char *path, DIR **dirstream, int verbose)
1615 {
1616 char mp[PATH_MAX];
1617 int ret;
1618
1619 if (is_block_device(path)) {
1620 ret = get_btrfs_mount(path, mp, sizeof(mp));
1621 if (ret < 0) {
1622 /* not a mounted btrfs dev */
1623 error_on(verbose, "'%s' is not a mounted btrfs device",
1624 path);
1625 errno = EINVAL;
1626 return -1;
1627 }
1628 ret = open_file_or_dir(mp, dirstream);
1629 error_on(verbose && ret < 0, "can't access '%s': %s",
1630 path, strerror(errno));
1631 } else {
1632 ret = btrfs_open_dir(path, dirstream, 1);
1633 }
1634
1635 return ret;
1636 }
1637
1638 /*
1639 * Do the following checks before calling open_file_or_dir():
1640 * 1: path is in a btrfs filesystem
1641 * 2: path is a directory
1642 */
1643 int btrfs_open_dir(const char *path, DIR **dirstream, int verbose)
1644 {
1645 struct statfs stfs;
1646 struct stat st;
1647 int ret;
1648
1649 if (statfs(path, &stfs) != 0) {
1650 error_on(verbose, "cannot access '%s': %s", path,
1651 strerror(errno));
1652 return -1;
1653 }
1654
1655 if (stfs.f_type != BTRFS_SUPER_MAGIC) {
1656 error_on(verbose, "not a btrfs filesystem: %s", path);
1657 return -2;
1658 }
1659
1660 if (stat(path, &st) != 0) {
1661 error_on(verbose, "cannot access '%s': %s", path,
1662 strerror(errno));
1663 return -1;
1664 }
1665
1666 if (!S_ISDIR(st.st_mode)) {
1667 error_on(verbose, "not a directory: %s", path);
1668 return -3;
1669 }
1670
1671 ret = open_file_or_dir(path, dirstream);
1672 if (ret < 0) {
1673 error_on(verbose, "cannot access '%s': %s", path,
1674 strerror(errno));
1675 }
1676
1677 return ret;
1678 }
1679
1680 /* checks if a device is a loop device */
1681 static int is_loop_device (const char* device) {
1682 struct stat statbuf;
1683
1684 if(stat(device, &statbuf) < 0)
1685 return -errno;
1686
1687 return (S_ISBLK(statbuf.st_mode) &&
1688 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
1689 }
1690
1691 /*
1692 * Takes a loop device path (e.g. /dev/loop0) and returns
1693 * the associated file (e.g. /images/my_btrfs.img) using
1694 * loopdev API
1695 */
1696 static int resolve_loop_device_with_loopdev(const char* loop_dev, char* loop_file)
1697 {
1698 int fd;
1699 int ret;
1700 struct loop_info64 lo64;
1701
1702 fd = open(loop_dev, O_RDONLY | O_NONBLOCK);
1703 if (fd < 0)
1704 return -errno;
1705 ret = ioctl(fd, LOOP_GET_STATUS64, &lo64);
1706 if (ret < 0) {
1707 ret = -errno;
1708 goto out;
1709 }
1710
1711 memcpy(loop_file, lo64.lo_file_name, sizeof(lo64.lo_file_name));
1712 loop_file[sizeof(lo64.lo_file_name)] = 0;
1713
1714 out:
1715 close(fd);
1716
1717 return ret;
1718 }
1719
1720 /* Takes a loop device path (e.g. /dev/loop0) and returns
1721 * the associated file (e.g. /images/my_btrfs.img) */
1722 static int resolve_loop_device(const char* loop_dev, char* loop_file,
1723 int max_len)
1724 {
1725 int ret;
1726 FILE *f;
1727 char fmt[20];
1728 char p[PATH_MAX];
1729 char real_loop_dev[PATH_MAX];
1730
1731 if (!realpath(loop_dev, real_loop_dev))
1732 return -errno;
1733 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
1734 if (!(f = fopen(p, "r"))) {
1735 if (errno == ENOENT)
1736 /*
1737 * It's possibly a partitioned loop device, which is
1738 * resolvable with loopdev API.
1739 */
1740 return resolve_loop_device_with_loopdev(loop_dev, loop_file);
1741 return -errno;
1742 }
1743
1744 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
1745 ret = fscanf(f, fmt, loop_file);
1746 fclose(f);
1747 if (ret == EOF)
1748 return -errno;
1749
1750 return 0;
1751 }
1752
1753 /*
1754 * Checks whether a and b are identical or device
1755 * files associated with the same block device
1756 */
1757 static int is_same_blk_file(const char* a, const char* b)
1758 {
1759 struct stat st_buf_a, st_buf_b;
1760 char real_a[PATH_MAX];
1761 char real_b[PATH_MAX];
1762
1763 if (!realpath(a, real_a))
1764 strncpy_null(real_a, a);
1765
1766 if (!realpath(b, real_b))
1767 strncpy_null(real_b, b);
1768
1769 /* Identical path? */
1770 if (strcmp(real_a, real_b) == 0)
1771 return 1;
1772
1773 if (stat(a, &st_buf_a) < 0 || stat(b, &st_buf_b) < 0) {
1774 if (errno == ENOENT)
1775 return 0;
1776 return -errno;
1777 }
1778
1779 /* Same blockdevice? */
1780 if (S_ISBLK(st_buf_a.st_mode) && S_ISBLK(st_buf_b.st_mode) &&
1781 st_buf_a.st_rdev == st_buf_b.st_rdev) {
1782 return 1;
1783 }
1784
1785 /* Hardlink? */
1786 if (st_buf_a.st_dev == st_buf_b.st_dev &&
1787 st_buf_a.st_ino == st_buf_b.st_ino) {
1788 return 1;
1789 }
1790
1791 return 0;
1792 }
1793
1794 /* checks if a and b are identical or device
1795 * files associated with the same block device or
1796 * if one file is a loop device that uses the other
1797 * file.
1798 */
1799 static int is_same_loop_file(const char* a, const char* b)
1800 {
1801 char res_a[PATH_MAX];
1802 char res_b[PATH_MAX];
1803 const char* final_a = NULL;
1804 const char* final_b = NULL;
1805 int ret;
1806
1807 /* Resolve a if it is a loop device */
1808 if((ret = is_loop_device(a)) < 0) {
1809 if (ret == -ENOENT)
1810 return 0;
1811 return ret;
1812 } else if (ret) {
1813 ret = resolve_loop_device(a, res_a, sizeof(res_a));
1814 if (ret < 0) {
1815 if (errno != EPERM)
1816 return ret;
1817 } else {
1818 final_a = res_a;
1819 }
1820 } else {
1821 final_a = a;
1822 }
1823
1824 /* Resolve b if it is a loop device */
1825 if ((ret = is_loop_device(b)) < 0) {
1826 if (ret == -ENOENT)
1827 return 0;
1828 return ret;
1829 } else if (ret) {
1830 ret = resolve_loop_device(b, res_b, sizeof(res_b));
1831 if (ret < 0) {
1832 if (errno != EPERM)
1833 return ret;
1834 } else {
1835 final_b = res_b;
1836 }
1837 } else {
1838 final_b = b;
1839 }
1840
1841 return is_same_blk_file(final_a, final_b);
1842 }
1843
1844 /* Checks if a file exists and is a block or regular file*/
1845 static int is_existing_blk_or_reg_file(const char* filename)
1846 {
1847 struct stat st_buf;
1848
1849 if(stat(filename, &st_buf) < 0) {
1850 if(errno == ENOENT)
1851 return 0;
1852 else
1853 return -errno;
1854 }
1855
1856 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
1857 }
1858
1859 /* Checks if a file is used (directly or indirectly via a loop device)
1860 * by a device in fs_devices
1861 */
1862 static int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices,
1863 const char* file)
1864 {
1865 int ret;
1866 struct list_head *head;
1867 struct list_head *cur;
1868 struct btrfs_device *device;
1869
1870 head = &fs_devices->devices;
1871 list_for_each(cur, head) {
1872 device = list_entry(cur, struct btrfs_device, dev_list);
1873
1874 if((ret = is_same_loop_file(device->name, file)))
1875 return ret;
1876 }
1877
1878 return 0;
1879 }
1880
1881 /*
1882 * Resolve a pathname to a device mapper node to /dev/mapper/<name>
1883 * Returns NULL on invalid input or malloc failure; Other failures
1884 * will be handled by the caller using the input pathame.
1885 */
1886 char *canonicalize_dm_name(const char *ptname)
1887 {
1888 FILE *f;
1889 size_t sz;
1890 char path[PATH_MAX], name[PATH_MAX], *res = NULL;
1891
1892 if (!ptname || !*ptname)
1893 return NULL;
1894
1895 snprintf(path, sizeof(path), "/sys/block/%s/dm/name", ptname);
1896 if (!(f = fopen(path, "r")))
1897 return NULL;
1898
1899 /* read <name>\n from sysfs */
1900 if (fgets(name, sizeof(name), f) && (sz = strlen(name)) > 1) {
1901 name[sz - 1] = '\0';
1902 snprintf(path, sizeof(path), "/dev/mapper/%s", name);
1903
1904 if (access(path, F_OK) == 0)
1905 res = strdup(path);
1906 }
1907 fclose(f);
1908 return res;
1909 }
1910
1911 /*
1912 * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or
1913 * to a device mapper pathname.
1914 * Returns NULL on invalid input or malloc failure; Other failures
1915 * will be handled by the caller using the input pathame.
1916 */
1917 char *canonicalize_path(const char *path)
1918 {
1919 char *canonical, *p;
1920
1921 if (!path || !*path)
1922 return NULL;
1923
1924 canonical = realpath(path, NULL);
1925 if (!canonical)
1926 return strdup(path);
1927 p = strrchr(canonical, '/');
1928 if (p && strncmp(p, "/dm-", 4) == 0 && isdigit(*(p + 4))) {
1929 char *dm = canonicalize_dm_name(p + 1);
1930
1931 if (dm) {
1932 free(canonical);
1933 return dm;
1934 }
1935 }
1936 return canonical;
1937 }
1938
1939 /*
1940 * returns 1 if the device was mounted, < 0 on error or 0 if everything
1941 * is safe to continue.
1942 */
1943 int check_mounted(const char* file)
1944 {
1945 int fd;
1946 int ret;
1947
1948 fd = open(file, O_RDONLY);
1949 if (fd < 0) {
1950 error("mount check: cannot open %s: %s", file,
1951 strerror(errno));
1952 return -errno;
1953 }
1954
1955 ret = check_mounted_where(fd, file, NULL, 0, NULL);
1956 close(fd);
1957
1958 return ret;
1959 }
1960
1961 int check_mounted_where(int fd, const char *file, char *where, int size,
1962 struct btrfs_fs_devices **fs_dev_ret)
1963 {
1964 int ret;
1965 u64 total_devs = 1;
1966 int is_btrfs;
1967 struct btrfs_fs_devices *fs_devices_mnt = NULL;
1968 FILE *f;
1969 struct mntent *mnt;
1970
1971 /* scan the initial device */
1972 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
1973 &total_devs, BTRFS_SUPER_INFO_OFFSET, 0);
1974 is_btrfs = (ret >= 0);
1975
1976 /* scan other devices */
1977 if (is_btrfs && total_devs > 1) {
1978 ret = btrfs_scan_lblkid();
1979 if (ret)
1980 return ret;
1981 }
1982
1983 /* iterate over the list of currently mounted filesystems */
1984 if ((f = setmntent ("/proc/self/mounts", "r")) == NULL)
1985 return -errno;
1986
1987 while ((mnt = getmntent (f)) != NULL) {
1988 if(is_btrfs) {
1989 if(strcmp(mnt->mnt_type, "btrfs") != 0)
1990 continue;
1991
1992 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
1993 } else {
1994 /* ignore entries in the mount table that are not
1995 associated with a file*/
1996 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
1997 goto out_mntloop_err;
1998 else if(!ret)
1999 continue;
2000
2001 ret = is_same_loop_file(file, mnt->mnt_fsname);
2002 }
2003
2004 if(ret < 0)
2005 goto out_mntloop_err;
2006 else if(ret)
2007 break;
2008 }
2009
2010 /* Did we find an entry in mnt table? */
2011 if (mnt && size && where) {
2012 strncpy(where, mnt->mnt_dir, size);
2013 where[size-1] = 0;
2014 }
2015 if (fs_dev_ret)
2016 *fs_dev_ret = fs_devices_mnt;
2017
2018 ret = (mnt != NULL);
2019
2020 out_mntloop_err:
2021 endmntent (f);
2022
2023 return ret;
2024 }
2025
2026 struct pending_dir {
2027 struct list_head list;
2028 char name[PATH_MAX];
2029 };
2030
2031 int btrfs_register_one_device(const char *fname)
2032 {
2033 struct btrfs_ioctl_vol_args args;
2034 int fd;
2035 int ret;
2036
2037 fd = open("/dev/btrfs-control", O_RDWR);
2038 if (fd < 0) {
2039 warning(
2040 "failed to open /dev/btrfs-control, skipping device registration: %s",
2041 strerror(errno));
2042 return -errno;
2043 }
2044 memset(&args, 0, sizeof(args));
2045 strncpy_null(args.name, fname);
2046 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
2047 if (ret < 0) {
2048 error("device scan failed on '%s': %s", fname,
2049 strerror(errno));
2050 ret = -errno;
2051 }
2052 close(fd);
2053 return ret;
2054 }
2055
2056 /*
2057 * Register all devices in the fs_uuid list created in the user
2058 * space. Ensure btrfs_scan_lblkid() is called before this func.
2059 */
2060 int btrfs_register_all_devices(void)
2061 {
2062 int err = 0;
2063 int ret = 0;
2064 struct btrfs_fs_devices *fs_devices;
2065 struct btrfs_device *device;
2066 struct list_head *all_uuids;
2067
2068 all_uuids = btrfs_scanned_uuids();
2069
2070 list_for_each_entry(fs_devices, all_uuids, list) {
2071 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2072 if (*device->name)
2073 err = btrfs_register_one_device(device->name);
2074
2075 if (err)
2076 ret++;
2077 }
2078 }
2079
2080 return ret;
2081 }
2082
2083 int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
2084 int super_offset)
2085 {
2086 struct btrfs_super_block *disk_super;
2087 char *buf;
2088 int ret = 0;
2089
2090 buf = malloc(BTRFS_SUPER_INFO_SIZE);
2091 if (!buf) {
2092 ret = -ENOMEM;
2093 goto out;
2094 }
2095 ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
2096 if (ret != BTRFS_SUPER_INFO_SIZE)
2097 goto brelse;
2098
2099 ret = 0;
2100 disk_super = (struct btrfs_super_block *)buf;
2101 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC)
2102 goto brelse;
2103
2104 if (!memcmp(disk_super->fsid, root->fs_info->super_copy->fsid,
2105 BTRFS_FSID_SIZE))
2106 ret = 1;
2107 brelse:
2108 free(buf);
2109 out:
2110 return ret;
2111 }
2112
2113 /*
2114 * Note: this function uses a static per-thread buffer. Do not call this
2115 * function more than 10 times within one argument list!
2116 */
2117 const char *pretty_size_mode(u64 size, unsigned mode)
2118 {
2119 static __thread int ps_index = 0;
2120 static __thread char ps_array[10][32];
2121 char *ret;
2122
2123 ret = ps_array[ps_index];
2124 ps_index++;
2125 ps_index %= 10;
2126 (void)pretty_size_snprintf(size, ret, 32, mode);
2127
2128 return ret;
2129 }
2130
2131 static const char* unit_suffix_binary[] =
2132 { "B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB"};
2133 static const char* unit_suffix_decimal[] =
2134 { "B", "kB", "MB", "GB", "TB", "PB", "EB"};
2135
2136 int pretty_size_snprintf(u64 size, char *str, size_t str_size, unsigned unit_mode)
2137 {
2138 int num_divs;
2139 float fraction;
2140 u64 base = 0;
2141 int mult = 0;
2142 const char** suffix = NULL;
2143 u64 last_size;
2144
2145 if (str_size == 0)
2146 return 0;
2147
2148 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_RAW) {
2149 snprintf(str, str_size, "%llu", size);
2150 return 0;
2151 }
2152
2153 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_BINARY) {
2154 base = 1024;
2155 mult = 1024;
2156 suffix = unit_suffix_binary;
2157 } else if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_DECIMAL) {
2158 base = 1000;
2159 mult = 1000;
2160 suffix = unit_suffix_decimal;
2161 }
2162
2163 /* Unknown mode */
2164 if (!base) {
2165 fprintf(stderr, "INTERNAL ERROR: unknown unit base, mode %d\n",
2166 unit_mode);
2167 assert(0);
2168 return -1;
2169 }
2170
2171 num_divs = 0;
2172 last_size = size;
2173 switch (unit_mode & UNITS_MODE_MASK) {
2174 case UNITS_TBYTES: base *= mult; num_divs++;
2175 case UNITS_GBYTES: base *= mult; num_divs++;
2176 case UNITS_MBYTES: base *= mult; num_divs++;
2177 case UNITS_KBYTES: num_divs++;
2178 break;
2179 case UNITS_BYTES:
2180 base = 1;
2181 num_divs = 0;
2182 break;
2183 default:
2184 while (size >= mult) {
2185 last_size = size;
2186 size /= mult;
2187 num_divs++;
2188 }
2189 /*
2190 * If the value is smaller than base, we didn't do any
2191 * division, in that case, base should be 1, not original
2192 * base, or the unit will be wrong
2193 */
2194 if (num_divs == 0)
2195 base = 1;
2196 }
2197
2198 if (num_divs >= ARRAY_SIZE(unit_suffix_binary)) {
2199 str[0] = '\0';
2200 printf("INTERNAL ERROR: unsupported unit suffix, index %d\n",
2201 num_divs);
2202 assert(0);
2203 return -1;
2204 }
2205 fraction = (float)last_size / base;
2206
2207 return snprintf(str, str_size, "%.2f%s", fraction, suffix[num_divs]);
2208 }
2209
2210 /*
2211 * __strncpy_null - strncpy with null termination
2212 * @dest: the target array
2213 * @src: the source string
2214 * @n: maximum bytes to copy (size of *dest)
2215 *
2216 * Like strncpy, but ensures destination is null-terminated.
2217 *
2218 * Copies the string pointed to by src, including the terminating null
2219 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
2220 * of n bytes. Then ensure that dest is null-terminated.
2221 */
2222 char *__strncpy_null(char *dest, const char *src, size_t n)
2223 {
2224 strncpy(dest, src, n);
2225 if (n > 0)
2226 dest[n - 1] = '\0';
2227 return dest;
2228 }
2229
2230 /*
2231 * Checks to make sure that the label matches our requirements.
2232 * Returns:
2233 0 if everything is safe and usable
2234 -1 if the label is too long
2235 */
2236 static int check_label(const char *input)
2237 {
2238 int len = strlen(input);
2239
2240 if (len > BTRFS_LABEL_SIZE - 1) {
2241 error("label %s is too long (max %d)", input,
2242 BTRFS_LABEL_SIZE - 1);
2243 return -1;
2244 }
2245
2246 return 0;
2247 }
2248
2249 static int set_label_unmounted(const char *dev, const char *label)
2250 {
2251 struct btrfs_trans_handle *trans;
2252 struct btrfs_root *root;
2253 int ret;
2254
2255 ret = check_mounted(dev);
2256 if (ret < 0) {
2257 error("checking mount status of %s failed: %d", dev, ret);
2258 return -1;
2259 }
2260 if (ret > 0) {
2261 error("device %s is mounted, use mount point", dev);
2262 return -1;
2263 }
2264
2265 /* Open the super_block at the default location
2266 * and as read-write.
2267 */
2268 root = open_ctree(dev, 0, OPEN_CTREE_WRITES);
2269 if (!root) /* errors are printed by open_ctree() */
2270 return -1;
2271
2272 trans = btrfs_start_transaction(root, 1);
2273 __strncpy_null(root->fs_info->super_copy->label, label, BTRFS_LABEL_SIZE - 1);
2274
2275 btrfs_commit_transaction(trans, root);
2276
2277 /* Now we close it since we are done. */
2278 close_ctree(root);
2279 return 0;
2280 }
2281
2282 static int set_label_mounted(const char *mount_path, const char *labelp)
2283 {
2284 int fd;
2285 char label[BTRFS_LABEL_SIZE];
2286
2287 fd = open(mount_path, O_RDONLY | O_NOATIME);
2288 if (fd < 0) {
2289 error("unable to access %s: %s", mount_path, strerror(errno));
2290 return -1;
2291 }
2292
2293 memset(label, 0, sizeof(label));
2294 __strncpy_null(label, labelp, BTRFS_LABEL_SIZE - 1);
2295 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
2296 error("unable to set label of %s: %s", mount_path,
2297 strerror(errno));
2298 close(fd);
2299 return -1;
2300 }
2301
2302 close(fd);
2303 return 0;
2304 }
2305
2306 int get_label_unmounted(const char *dev, char *label)
2307 {
2308 struct btrfs_root *root;
2309 int ret;
2310
2311 ret = check_mounted(dev);
2312 if (ret < 0) {
2313 error("checking mount status of %s failed: %d", dev, ret);
2314 return -1;
2315 }
2316
2317 /* Open the super_block at the default location
2318 * and as read-only.
2319 */
2320 root = open_ctree(dev, 0, 0);
2321 if(!root)
2322 return -1;
2323
2324 __strncpy_null(label, root->fs_info->super_copy->label,
2325 BTRFS_LABEL_SIZE - 1);
2326
2327 /* Now we close it since we are done. */
2328 close_ctree(root);
2329 return 0;
2330 }
2331
2332 /*
2333 * If a partition is mounted, try to get the filesystem label via its
2334 * mounted path rather than device. Return the corresponding error
2335 * the user specified the device path.
2336 */
2337 int get_label_mounted(const char *mount_path, char *labelp)
2338 {
2339 char label[BTRFS_LABEL_SIZE];
2340 int fd;
2341 int ret;
2342
2343 fd = open(mount_path, O_RDONLY | O_NOATIME);
2344 if (fd < 0) {
2345 error("unable to access %s: %s", mount_path, strerror(errno));
2346 return -1;
2347 }
2348
2349 memset(label, '\0', sizeof(label));
2350 ret = ioctl(fd, BTRFS_IOC_GET_FSLABEL, label);
2351 if (ret < 0) {
2352 if (errno != ENOTTY)
2353 error("unable to get label of %s: %s", mount_path,
2354 strerror(errno));
2355 ret = -errno;
2356 close(fd);
2357 return ret;
2358 }
2359
2360 __strncpy_null(labelp, label, BTRFS_LABEL_SIZE - 1);
2361 close(fd);
2362 return 0;
2363 }
2364
2365 int get_label(const char *btrfs_dev, char *label)
2366 {
2367 int ret;
2368
2369 ret = is_existing_blk_or_reg_file(btrfs_dev);
2370 if (!ret)
2371 ret = get_label_mounted(btrfs_dev, label);
2372 else if (ret > 0)
2373 ret = get_label_unmounted(btrfs_dev, label);
2374
2375 return ret;
2376 }
2377
2378 int set_label(const char *btrfs_dev, const char *label)
2379 {
2380 int ret;
2381
2382 if (check_label(label))
2383 return -1;
2384
2385 ret = is_existing_blk_or_reg_file(btrfs_dev);
2386 if (!ret)
2387 ret = set_label_mounted(btrfs_dev, label);
2388 else if (ret > 0)
2389 ret = set_label_unmounted(btrfs_dev, label);
2390
2391 return ret;
2392 }
2393
2394 /*
2395 * A not-so-good version fls64. No fascinating optimization since
2396 * no one except parse_size use it
2397 */
2398 static int fls64(u64 x)
2399 {
2400 int i;
2401
2402 for (i = 0; i <64; i++)
2403 if (x << i & (1ULL << 63))
2404 return 64 - i;
2405 return 64 - i;
2406 }
2407
2408 u64 parse_size(char *s)
2409 {
2410 char c;
2411 char *endptr;
2412 u64 mult = 1;
2413 u64 ret;
2414
2415 if (!s) {
2416 error("size value is empty");
2417 exit(1);
2418 }
2419 if (s[0] == '-') {
2420 error("size value '%s' is less equal than 0", s);
2421 exit(1);
2422 }
2423 ret = strtoull(s, &endptr, 10);
2424 if (endptr == s) {
2425 error("size value '%s' is invalid", s);
2426 exit(1);
2427 }
2428 if (endptr[0] && endptr[1]) {
2429 error("illegal suffix contains character '%c' in wrong position",
2430 endptr[1]);
2431 exit(1);
2432 }
2433 /*
2434 * strtoll returns LLONG_MAX when overflow, if this happens,
2435 * need to call strtoull to get the real size
2436 */
2437 if (errno == ERANGE && ret == ULLONG_MAX) {
2438 error("size value '%s' is too large for u64", s);
2439 exit(1);
2440 }
2441 if (endptr[0]) {
2442 c = tolower(endptr[0]);
2443 switch (c) {
2444 case 'e':
2445 mult *= 1024;
2446 /* fallthrough */
2447 case 'p':
2448 mult *= 1024;
2449 /* fallthrough */
2450 case 't':
2451 mult *= 1024;
2452 /* fallthrough */
2453 case 'g':
2454 mult *= 1024;
2455 /* fallthrough */
2456 case 'm':
2457 mult *= 1024;
2458 /* fallthrough */
2459 case 'k':
2460 mult *= 1024;
2461 /* fallthrough */
2462 case 'b':
2463 break;
2464 default:
2465 error("unknown size descriptor '%c'", c);
2466 exit(1);
2467 }
2468 }
2469 /* Check whether ret * mult overflow */
2470 if (fls64(ret) + fls64(mult) - 1 > 64) {
2471 error("size value '%s' is too large for u64", s);
2472 exit(1);
2473 }
2474 ret *= mult;
2475 return ret;
2476 }
2477
2478 u64 parse_qgroupid(const char *p)
2479 {
2480 char *s = strchr(p, '/');
2481 const char *ptr_src_end = p + strlen(p);
2482 char *ptr_parse_end = NULL;
2483 u64 level;
2484 u64 id;
2485 int fd;
2486 int ret = 0;
2487
2488 if (p[0] == '/')
2489 goto path;
2490
2491 /* Numeric format like '0/257' is the primary case */
2492 if (!s) {
2493 id = strtoull(p, &ptr_parse_end, 10);
2494 if (ptr_parse_end != ptr_src_end)
2495 goto path;
2496 return id;
2497 }
2498 level = strtoull(p, &ptr_parse_end, 10);
2499 if (ptr_parse_end != s)
2500 goto path;
2501
2502 id = strtoull(s + 1, &ptr_parse_end, 10);
2503 if (ptr_parse_end != ptr_src_end)
2504 goto path;
2505
2506 return (level << BTRFS_QGROUP_LEVEL_SHIFT) | id;
2507
2508 path:
2509 /* Path format like subv at 'my_subvol' is the fallback case */
2510 ret = test_issubvolume(p);
2511 if (ret < 0 || !ret)
2512 goto err;
2513 fd = open(p, O_RDONLY);
2514 if (fd < 0)
2515 goto err;
2516 ret = lookup_ino_rootid(fd, &id);
2517 close(fd);
2518 if (ret < 0)
2519 goto err;
2520 return id;
2521
2522 err:
2523 error("invalid qgroupid or subvolume path: %s", p);
2524 exit(-1);
2525 }
2526
2527 int open_file_or_dir3(const char *fname, DIR **dirstream, int open_flags)
2528 {
2529 int ret;
2530 struct stat st;
2531 int fd;
2532
2533 ret = stat(fname, &st);
2534 if (ret < 0) {
2535 return -1;
2536 }
2537 if (S_ISDIR(st.st_mode)) {
2538 *dirstream = opendir(fname);
2539 if (!*dirstream)
2540 return -1;
2541 fd = dirfd(*dirstream);
2542 } else if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) {
2543 fd = open(fname, open_flags);
2544 } else {
2545 /*
2546 * we set this on purpose, in case the caller output
2547 * strerror(errno) as success
2548 */
2549 errno = EINVAL;
2550 return -1;
2551 }
2552 if (fd < 0) {
2553 fd = -1;
2554 if (*dirstream) {
2555 closedir(*dirstream);
2556 *dirstream = NULL;
2557 }
2558 }
2559 return fd;
2560 }
2561
2562 int open_file_or_dir(const char *fname, DIR **dirstream)
2563 {
2564 return open_file_or_dir3(fname, dirstream, O_RDWR);
2565 }
2566
2567 void close_file_or_dir(int fd, DIR *dirstream)
2568 {
2569 if (dirstream)
2570 closedir(dirstream);
2571 else if (fd >= 0)
2572 close(fd);
2573 }
2574
2575 int get_device_info(int fd, u64 devid,
2576 struct btrfs_ioctl_dev_info_args *di_args)
2577 {
2578 int ret;
2579
2580 di_args->devid = devid;
2581 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
2582
2583 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
2584 return ret < 0 ? -errno : 0;
2585 }
2586
2587 static u64 find_max_device_id(struct btrfs_ioctl_search_args *search_args,
2588 int nr_items)
2589 {
2590 struct btrfs_dev_item *dev_item;
2591 char *buf = search_args->buf;
2592
2593 buf += (nr_items - 1) * (sizeof(struct btrfs_ioctl_search_header)
2594 + sizeof(struct btrfs_dev_item));
2595 buf += sizeof(struct btrfs_ioctl_search_header);
2596
2597 dev_item = (struct btrfs_dev_item *)buf;
2598
2599 return btrfs_stack_device_id(dev_item);
2600 }
2601
2602 static int search_chunk_tree_for_fs_info(int fd,
2603 struct btrfs_ioctl_fs_info_args *fi_args)
2604 {
2605 int ret;
2606 int max_items;
2607 u64 start_devid = 1;
2608 struct btrfs_ioctl_search_args search_args;
2609 struct btrfs_ioctl_search_key *search_key = &search_args.key;
2610
2611 fi_args->num_devices = 0;
2612
2613 max_items = BTRFS_SEARCH_ARGS_BUFSIZE
2614 / (sizeof(struct btrfs_ioctl_search_header)
2615 + sizeof(struct btrfs_dev_item));
2616
2617 search_key->tree_id = BTRFS_CHUNK_TREE_OBJECTID;
2618 search_key->min_objectid = BTRFS_DEV_ITEMS_OBJECTID;
2619 search_key->max_objectid = BTRFS_DEV_ITEMS_OBJECTID;
2620 search_key->min_type = BTRFS_DEV_ITEM_KEY;
2621 search_key->max_type = BTRFS_DEV_ITEM_KEY;
2622 search_key->min_transid = 0;
2623 search_key->max_transid = (u64)-1;
2624 search_key->nr_items = max_items;
2625 search_key->max_offset = (u64)-1;
2626
2627 again:
2628 search_key->min_offset = start_devid;
2629
2630 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH, &search_args);
2631 if (ret < 0)
2632 return -errno;
2633
2634 fi_args->num_devices += (u64)search_key->nr_items;
2635
2636 if (search_key->nr_items == max_items) {
2637 start_devid = find_max_device_id(&search_args,
2638 search_key->nr_items) + 1;
2639 goto again;
2640 }
2641
2642 /* get the lastest max_id to stay consistent with the num_devices */
2643 if (search_key->nr_items == 0)
2644 /*
2645 * last tree_search returns an empty buf, use the devid of
2646 * the last dev_item of the previous tree_search
2647 */
2648 fi_args->max_id = start_devid - 1;
2649 else
2650 fi_args->max_id = find_max_device_id(&search_args,
2651 search_key->nr_items);
2652
2653 return 0;
2654 }
2655
2656 /*
2657 * For a given path, fill in the ioctl fs_ and info_ args.
2658 * If the path is a btrfs mountpoint, fill info for all devices.
2659 * If the path is a btrfs device, fill in only that device.
2660 *
2661 * The path provided must be either on a mounted btrfs fs,
2662 * or be a mounted btrfs device.
2663 *
2664 * Returns 0 on success, or a negative errno.
2665 */
2666 int get_fs_info(char *path, struct btrfs_ioctl_fs_info_args *fi_args,
2667 struct btrfs_ioctl_dev_info_args **di_ret)
2668 {
2669 int fd = -1;
2670 int ret = 0;
2671 int ndevs = 0;
2672 int i = 0;
2673 int replacing = 0;
2674 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2675 struct btrfs_ioctl_dev_info_args *di_args;
2676 struct btrfs_ioctl_dev_info_args tmp;
2677 char mp[PATH_MAX];
2678 DIR *dirstream = NULL;
2679
2680 memset(fi_args, 0, sizeof(*fi_args));
2681
2682 if (is_block_device(path) == 1) {
2683 struct btrfs_super_block *disk_super;
2684 char buf[BTRFS_SUPER_INFO_SIZE];
2685 u64 devid;
2686
2687 /* Ensure it's mounted, then set path to the mountpoint */
2688 fd = open(path, O_RDONLY);
2689 if (fd < 0) {
2690 ret = -errno;
2691 error("cannot open %s: %s", path, strerror(errno));
2692 goto out;
2693 }
2694 ret = check_mounted_where(fd, path, mp, sizeof(mp),
2695 &fs_devices_mnt);
2696 if (!ret) {
2697 ret = -EINVAL;
2698 goto out;
2699 }
2700 if (ret < 0)
2701 goto out;
2702 path = mp;
2703 /* Only fill in this one device */
2704 fi_args->num_devices = 1;
2705
2706 disk_super = (struct btrfs_super_block *)buf;
2707 ret = btrfs_read_dev_super(fd, disk_super,
2708 BTRFS_SUPER_INFO_OFFSET, 0);
2709 if (ret < 0) {
2710 ret = -EIO;
2711 goto out;
2712 }
2713 devid = btrfs_stack_device_id(&disk_super->dev_item);
2714
2715 fi_args->max_id = devid;
2716 i = devid;
2717
2718 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
2719 close(fd);
2720 }
2721
2722 /* at this point path must not be for a block device */
2723 fd = open_file_or_dir(path, &dirstream);
2724 if (fd < 0) {
2725 ret = -errno;
2726 goto out;
2727 }
2728
2729 /* fill in fi_args if not just a single device */
2730 if (fi_args->num_devices != 1) {
2731 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
2732 if (ret < 0) {
2733 ret = -errno;
2734 goto out;
2735 }
2736
2737 /*
2738 * The fs_args->num_devices does not include seed devices
2739 */
2740 ret = search_chunk_tree_for_fs_info(fd, fi_args);
2741 if (ret)
2742 goto out;
2743
2744 /*
2745 * search_chunk_tree_for_fs_info() will lacks the devid 0
2746 * so manual probe for it here.
2747 */
2748 ret = get_device_info(fd, 0, &tmp);
2749 if (!ret) {
2750 fi_args->num_devices++;
2751 ndevs++;
2752 replacing = 1;
2753 if (i == 0)
2754 i++;
2755 }
2756 }
2757
2758 if (!fi_args->num_devices)
2759 goto out;
2760
2761 di_args = *di_ret = malloc((fi_args->num_devices) * sizeof(*di_args));
2762 if (!di_args) {
2763 ret = -errno;
2764 goto out;
2765 }
2766
2767 if (replacing)
2768 memcpy(di_args, &tmp, sizeof(tmp));
2769 for (; i <= fi_args->max_id; ++i) {
2770 ret = get_device_info(fd, i, &di_args[ndevs]);
2771 if (ret == -ENODEV)
2772 continue;
2773 if (ret)
2774 goto out;
2775 ndevs++;
2776 }
2777
2778 /*
2779 * only when the only dev we wanted to find is not there then
2780 * let any error be returned
2781 */
2782 if (fi_args->num_devices != 1) {
2783 BUG_ON(ndevs == 0);
2784 ret = 0;
2785 }
2786
2787 out:
2788 close_file_or_dir(fd, dirstream);
2789 return ret;
2790 }
2791
2792 #define isoctal(c) (((c) & ~7) == '0')
2793
2794 static inline void translate(char *f, char *t)
2795 {
2796 while (*f != '\0') {
2797 if (*f == '\\' &&
2798 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
2799 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
2800 f += 4;
2801 } else
2802 *t++ = *f++;
2803 }
2804 *t = '\0';
2805 return;
2806 }
2807
2808 /*
2809 * Checks if the swap device.
2810 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
2811 */
2812 static int is_swap_device(const char *file)
2813 {
2814 FILE *f;
2815 struct stat st_buf;
2816 dev_t dev;
2817 ino_t ino = 0;
2818 char tmp[PATH_MAX];
2819 char buf[PATH_MAX];
2820 char *cp;
2821 int ret = 0;
2822
2823 if (stat(file, &st_buf) < 0)
2824 return -errno;
2825 if (S_ISBLK(st_buf.st_mode))
2826 dev = st_buf.st_rdev;
2827 else if (S_ISREG(st_buf.st_mode)) {
2828 dev = st_buf.st_dev;
2829 ino = st_buf.st_ino;
2830 } else
2831 return 0;
2832
2833 if ((f = fopen("/proc/swaps", "r")) == NULL)
2834 return 0;
2835
2836 /* skip the first line */
2837 if (fgets(tmp, sizeof(tmp), f) == NULL)
2838 goto out;
2839
2840 while (fgets(tmp, sizeof(tmp), f) != NULL) {
2841 if ((cp = strchr(tmp, ' ')) != NULL)
2842 *cp = '\0';
2843 if ((cp = strchr(tmp, '\t')) != NULL)
2844 *cp = '\0';
2845 translate(tmp, buf);
2846 if (stat(buf, &st_buf) != 0)
2847 continue;
2848 if (S_ISBLK(st_buf.st_mode)) {
2849 if (dev == st_buf.st_rdev) {
2850 ret = 1;
2851 break;
2852 }
2853 } else if (S_ISREG(st_buf.st_mode)) {
2854 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {
2855 ret = 1;
2856 break;
2857 }
2858 }
2859 }
2860
2861 out:
2862 fclose(f);
2863
2864 return ret;
2865 }
2866
2867 /*
2868 * Check for existing filesystem or partition table on device.
2869 * Returns:
2870 * 1 for existing fs or partition
2871 * 0 for nothing found
2872 * -1 for internal error
2873 */
2874 static int check_overwrite(const char *device)
2875 {
2876 const char *type;
2877 blkid_probe pr = NULL;
2878 int ret;
2879 blkid_loff_t size;
2880
2881 if (!device || !*device)
2882 return 0;
2883
2884 ret = -1; /* will reset on success of all setup calls */
2885
2886 pr = blkid_new_probe_from_filename(device);
2887 if (!pr)
2888 goto out;
2889
2890 size = blkid_probe_get_size(pr);
2891 if (size < 0)
2892 goto out;
2893
2894 /* nothing to overwrite on a 0-length device */
2895 if (size == 0) {
2896 ret = 0;
2897 goto out;
2898 }
2899
2900 ret = blkid_probe_enable_partitions(pr, 1);
2901 if (ret < 0)
2902 goto out;
2903
2904 ret = blkid_do_fullprobe(pr);
2905 if (ret < 0)
2906 goto out;
2907
2908 /*
2909 * Blkid returns 1 for nothing found and 0 when it finds a signature,
2910 * but we want the exact opposite, so reverse the return value here.
2911 *
2912 * In addition print some useful diagnostics about what actually is
2913 * on the device.
2914 */
2915 if (ret) {
2916 ret = 0;
2917 goto out;
2918 }
2919
2920 if (!blkid_probe_lookup_value(pr, "TYPE", &type, NULL)) {
2921 fprintf(stderr,
2922 "%s appears to contain an existing "
2923 "filesystem (%s).\n", device, type);
2924 } else if (!blkid_probe_lookup_value(pr, "PTTYPE", &type, NULL)) {
2925 fprintf(stderr,
2926 "%s appears to contain a partition "
2927 "table (%s).\n", device, type);
2928 } else {
2929 fprintf(stderr,
2930 "%s appears to contain something weird "
2931 "according to blkid\n", device);
2932 }
2933 ret = 1;
2934
2935 out:
2936 if (pr)
2937 blkid_free_probe(pr);
2938 if (ret == -1)
2939 fprintf(stderr,
2940 "probe of %s failed, cannot detect "
2941 "existing filesystem.\n", device);
2942 return ret;
2943 }
2944
2945 static int group_profile_devs_min(u64 flag)
2946 {
2947 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
2948 case 0: /* single */
2949 case BTRFS_BLOCK_GROUP_DUP:
2950 return 1;
2951 case BTRFS_BLOCK_GROUP_RAID0:
2952 case BTRFS_BLOCK_GROUP_RAID1:
2953 case BTRFS_BLOCK_GROUP_RAID5:
2954 return 2;
2955 case BTRFS_BLOCK_GROUP_RAID6:
2956 return 3;
2957 case BTRFS_BLOCK_GROUP_RAID10:
2958 return 4;
2959 default:
2960 return -1;
2961 }
2962 }
2963
2964 int test_num_disk_vs_raid(u64 metadata_profile, u64 data_profile,
2965 u64 dev_cnt, int mixed, int ssd)
2966 {
2967 u64 allowed = 0;
2968
2969 switch (dev_cnt) {
2970 default:
2971 case 4:
2972 allowed |= BTRFS_BLOCK_GROUP_RAID10;
2973 case 3:
2974 allowed |= BTRFS_BLOCK_GROUP_RAID6;
2975 case 2:
2976 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
2977 BTRFS_BLOCK_GROUP_RAID5;
2978 case 1:
2979 allowed |= BTRFS_BLOCK_GROUP_DUP;
2980 }
2981
2982 if (dev_cnt > 1 &&
2983 ((metadata_profile | data_profile) & BTRFS_BLOCK_GROUP_DUP)) {
2984 warning("DUP is not recommended on filesystem with multiple devices");
2985 }
2986 if (metadata_profile & ~allowed) {
2987 fprintf(stderr,
2988 "ERROR: unable to create FS with metadata profile %s "
2989 "(have %llu devices but %d devices are required)\n",
2990 btrfs_group_profile_str(metadata_profile), dev_cnt,
2991 group_profile_devs_min(metadata_profile));
2992 return 1;
2993 }
2994 if (data_profile & ~allowed) {
2995 fprintf(stderr,
2996 "ERROR: unable to create FS with data profile %s "
2997 "(have %llu devices but %d devices are required)\n",
2998 btrfs_group_profile_str(data_profile), dev_cnt,
2999 group_profile_devs_min(data_profile));
3000 return 1;
3001 }
3002
3003 warning_on(!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP) && ssd,
3004 "DUP may not actually lead to 2 copies on the device, see manual page");
3005
3006 return 0;
3007 }
3008
3009 int group_profile_max_safe_loss(u64 flags)
3010 {
3011 switch (flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3012 case 0: /* single */
3013 case BTRFS_BLOCK_GROUP_DUP:
3014 case BTRFS_BLOCK_GROUP_RAID0:
3015 return 0;
3016 case BTRFS_BLOCK_GROUP_RAID1:
3017 case BTRFS_BLOCK_GROUP_RAID5:
3018 case BTRFS_BLOCK_GROUP_RAID10:
3019 return 1;
3020 case BTRFS_BLOCK_GROUP_RAID6:
3021 return 2;
3022 default:
3023 return -1;
3024 }
3025 }
3026
3027 /*
3028 * Check if a device is suitable for btrfs
3029 * returns:
3030 * 1: something is wrong, an error is printed
3031 * 0: all is fine
3032 */
3033 int test_dev_for_mkfs(const char *file, int force_overwrite)
3034 {
3035 int ret, fd;
3036 struct stat st;
3037
3038 ret = is_swap_device(file);
3039 if (ret < 0) {
3040 error("checking status of %s: %s", file, strerror(-ret));
3041 return 1;
3042 }
3043 if (ret == 1) {
3044 error("%s is a swap device", file);
3045 return 1;
3046 }
3047 if (!force_overwrite) {
3048 if (check_overwrite(file)) {
3049 error("use the -f option to force overwrite of %s",
3050 file);
3051 return 1;
3052 }
3053 }
3054 ret = check_mounted(file);
3055 if (ret < 0) {
3056 error("cannot check mount status of %s: %s", file,
3057 strerror(-ret));
3058 return 1;
3059 }
3060 if (ret == 1) {
3061 error("%s is mounted", file);
3062 return 1;
3063 }
3064 /* check if the device is busy */
3065 fd = open(file, O_RDWR|O_EXCL);
3066 if (fd < 0) {
3067 error("unable to open %s: %s", file, strerror(errno));
3068 return 1;
3069 }
3070 if (fstat(fd, &st)) {
3071 error("unable to stat %s: %s", file, strerror(errno));
3072 close(fd);
3073 return 1;
3074 }
3075 if (!S_ISBLK(st.st_mode)) {
3076 error("%s is not a block device", file);
3077 close(fd);
3078 return 1;
3079 }
3080 close(fd);
3081 return 0;
3082 }
3083
3084 int btrfs_scan_lblkid(void)
3085 {
3086 int fd = -1;
3087 int ret;
3088 u64 num_devices;
3089 struct btrfs_fs_devices *tmp_devices;
3090 blkid_dev_iterate iter = NULL;
3091 blkid_dev dev = NULL;
3092 blkid_cache cache = NULL;
3093 char path[PATH_MAX];
3094
3095 if (btrfs_scan_done)
3096 return 0;
3097
3098 if (blkid_get_cache(&cache, NULL) < 0) {
3099 error("blkid cache get failed");
3100 return 1;
3101 }
3102 blkid_probe_all(cache);
3103 iter = blkid_dev_iterate_begin(cache);
3104 blkid_dev_set_search(iter, "TYPE", "btrfs");
3105 while (blkid_dev_next(iter, &dev) == 0) {
3106 dev = blkid_verify(cache, dev);
3107 if (!dev)
3108 continue;
3109 /* if we are here its definitely a btrfs disk*/
3110 strncpy_null(path, blkid_dev_devname(dev));
3111
3112 fd = open(path, O_RDONLY);
3113 if (fd < 0) {
3114 error("cannot open %s: %s", path, strerror(errno));
3115 continue;
3116 }
3117 ret = btrfs_scan_one_device(fd, path, &tmp_devices,
3118 &num_devices, BTRFS_SUPER_INFO_OFFSET, 0);
3119 if (ret) {
3120 error("cannot scan %s: %s", path, strerror(-ret));
3121 close (fd);
3122 continue;
3123 }
3124
3125 close(fd);
3126 }
3127 blkid_dev_iterate_end(iter);
3128 blkid_put_cache(cache);
3129
3130 btrfs_scan_done = 1;
3131
3132 return 0;
3133 }
3134
3135 int is_vol_small(const char *file)
3136 {
3137 int fd = -1;
3138 int e;
3139 struct stat st;
3140 u64 size;
3141
3142 fd = open(file, O_RDONLY);
3143 if (fd < 0)
3144 return -errno;
3145 if (fstat(fd, &st) < 0) {
3146 e = -errno;
3147 close(fd);
3148 return e;
3149 }
3150 size = btrfs_device_size(fd, &st);
3151 if (size == 0) {
3152 close(fd);
3153 return -1;
3154 }
3155 if (size < BTRFS_MKFS_SMALL_VOLUME_SIZE) {
3156 close(fd);
3157 return 1;
3158 } else {
3159 close(fd);
3160 return 0;
3161 }
3162 }
3163
3164 /*
3165 * This reads a line from the stdin and only returns non-zero if the
3166 * first whitespace delimited token is a case insensitive match with yes
3167 * or y.
3168 */
3169 int ask_user(const char *question)
3170 {
3171 char buf[30] = {0,};
3172 char *saveptr = NULL;
3173 char *answer;
3174
3175 printf("%s [y/N]: ", question);
3176
3177 return fgets(buf, sizeof(buf) - 1, stdin) &&
3178 (answer = strtok_r(buf, " \t\n\r", &saveptr)) &&
3179 (!strcasecmp(answer, "yes") || !strcasecmp(answer, "y"));
3180 }
3181
3182 /*
3183 * For a given:
3184 * - file or directory return the containing tree root id
3185 * - subvolume return its own tree id
3186 * - BTRFS_EMPTY_SUBVOL_DIR_OBJECTID (directory with ino == 2) the result is
3187 * undefined and function returns -1
3188 */
3189 int lookup_ino_rootid(int fd, u64 *rootid)
3190 {
3191 struct btrfs_ioctl_ino_lookup_args args;
3192 int ret;
3193
3194 memset(&args, 0, sizeof(args));
3195 args.treeid = 0;
3196 args.objectid = BTRFS_FIRST_FREE_OBJECTID;
3197
3198 ret = ioctl(fd, BTRFS_IOC_INO_LOOKUP, &args);
3199 if (ret < 0) {
3200 error("failed to lookup root id: %s", strerror(errno));
3201 return ret;
3202 }
3203
3204 *rootid = args.treeid;
3205
3206 return 0;
3207 }
3208
3209 /*
3210 * return 0 if a btrfs mount point is found
3211 * return 1 if a mount point is found but not btrfs
3212 * return <0 if something goes wrong
3213 */
3214 int find_mount_root(const char *path, char **mount_root)
3215 {
3216 FILE *mnttab;
3217 int fd;
3218 struct mntent *ent;
3219 int len;
3220 int ret;
3221 int not_btrfs = 1;
3222 int longest_matchlen = 0;
3223 char *longest_match = NULL;
3224
3225 fd = open(path, O_RDONLY | O_NOATIME);
3226 if (fd < 0)
3227 return -errno;
3228 close(fd);
3229
3230 mnttab = setmntent("/proc/self/mounts", "r");
3231 if (!mnttab)
3232 return -errno;
3233
3234 while ((ent = getmntent(mnttab))) {
3235 len = strlen(ent->mnt_dir);
3236 if (strncmp(ent->mnt_dir, path, len) == 0) {
3237 /* match found and use the latest match */
3238 if (longest_matchlen <= len) {
3239 free(longest_match);
3240 longest_matchlen = len;
3241 longest_match = strdup(ent->mnt_dir);
3242 not_btrfs = strcmp(ent->mnt_type, "btrfs");
3243 }
3244 }
3245 }
3246 endmntent(mnttab);
3247
3248 if (!longest_match)
3249 return -ENOENT;
3250 if (not_btrfs) {
3251 free(longest_match);
3252 return 1;
3253 }
3254
3255 ret = 0;
3256 *mount_root = realpath(longest_match, NULL);
3257 if (!*mount_root)
3258 ret = -errno;
3259
3260 free(longest_match);
3261 return ret;
3262 }
3263
3264 int test_minimum_size(const char *file, u32 nodesize)
3265 {
3266 int fd;
3267 struct stat statbuf;
3268
3269 fd = open(file, O_RDONLY);
3270 if (fd < 0)
3271 return -errno;
3272 if (stat(file, &statbuf) < 0) {
3273 close(fd);
3274 return -errno;
3275 }
3276 if (btrfs_device_size(fd, &statbuf) < btrfs_min_dev_size(nodesize)) {
3277 close(fd);
3278 return 1;
3279 }
3280 close(fd);
3281 return 0;
3282 }
3283
3284
3285 /*
3286 * Test if path is a directory
3287 * Returns:
3288 * 0 - path exists but it is not a directory
3289 * 1 - path exists and it is a directory
3290 * < 0 - error
3291 */
3292 int test_isdir(const char *path)
3293 {
3294 struct stat st;
3295 int ret;
3296
3297 ret = stat(path, &st);
3298 if (ret < 0)
3299 return -errno;
3300
3301 return !!S_ISDIR(st.st_mode);
3302 }
3303
3304 void units_set_mode(unsigned *units, unsigned mode)
3305 {
3306 unsigned base = *units & UNITS_MODE_MASK;
3307
3308 *units = base | mode;
3309 }
3310
3311 void units_set_base(unsigned *units, unsigned base)
3312 {
3313 unsigned mode = *units & ~UNITS_MODE_MASK;
3314
3315 *units = base | mode;
3316 }
3317
3318 int find_next_key(struct btrfs_path *path, struct btrfs_key *key)
3319 {
3320 int level;
3321
3322 for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
3323 if (!path->nodes[level])
3324 break;
3325 if (path->slots[level] + 1 >=
3326 btrfs_header_nritems(path->nodes[level]))
3327 continue;
3328 if (level == 0)
3329 btrfs_item_key_to_cpu(path->nodes[level], key,
3330 path->slots[level] + 1);
3331 else
3332 btrfs_node_key_to_cpu(path->nodes[level], key,
3333 path->slots[level] + 1);
3334 return 0;
3335 }
3336 return 1;
3337 }
3338
3339 const char* btrfs_group_type_str(u64 flag)
3340 {
3341 u64 mask = BTRFS_BLOCK_GROUP_TYPE_MASK |
3342 BTRFS_SPACE_INFO_GLOBAL_RSV;
3343
3344 switch (flag & mask) {
3345 case BTRFS_BLOCK_GROUP_DATA:
3346 return "Data";
3347 case BTRFS_BLOCK_GROUP_SYSTEM:
3348 return "System";
3349 case BTRFS_BLOCK_GROUP_METADATA:
3350 return "Metadata";
3351 case BTRFS_BLOCK_GROUP_DATA|BTRFS_BLOCK_GROUP_METADATA:
3352 return "Data+Metadata";
3353 case BTRFS_SPACE_INFO_GLOBAL_RSV:
3354 return "GlobalReserve";
3355 default:
3356 return "unknown";
3357 }
3358 }
3359
3360 const char* btrfs_group_profile_str(u64 flag)
3361 {
3362 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3363 case 0:
3364 return "single";
3365 case BTRFS_BLOCK_GROUP_RAID0:
3366 return "RAID0";
3367 case BTRFS_BLOCK_GROUP_RAID1:
3368 return "RAID1";
3369 case BTRFS_BLOCK_GROUP_RAID5:
3370 return "RAID5";
3371 case BTRFS_BLOCK_GROUP_RAID6:
3372 return "RAID6";
3373 case BTRFS_BLOCK_GROUP_DUP:
3374 return "DUP";
3375 case BTRFS_BLOCK_GROUP_RAID10:
3376 return "RAID10";
3377 default:
3378 return "unknown";
3379 }
3380 }
3381
3382 u64 disk_size(const char *path)
3383 {
3384 struct statfs sfs;
3385
3386 if (statfs(path, &sfs) < 0)
3387 return 0;
3388 else
3389 return sfs.f_bsize * sfs.f_blocks;
3390 }
3391
3392 u64 get_partition_size(const char *dev)
3393 {
3394 u64 result;
3395 int fd = open(dev, O_RDONLY);
3396
3397 if (fd < 0)
3398 return 0;
3399 if (ioctl(fd, BLKGETSIZE64, &result) < 0) {
3400 close(fd);
3401 return 0;
3402 }
3403 close(fd);
3404
3405 return result;
3406 }
3407
3408 int btrfs_tree_search2_ioctl_supported(int fd)
3409 {
3410 struct btrfs_ioctl_search_args_v2 *args2;
3411 struct btrfs_ioctl_search_key *sk;
3412 int args2_size = 1024;
3413 char args2_buf[args2_size];
3414 int ret;
3415 static int v2_supported = -1;
3416
3417 if (v2_supported != -1)
3418 return v2_supported;
3419
3420 args2 = (struct btrfs_ioctl_search_args_v2 *)args2_buf;
3421 sk = &(args2->key);
3422
3423 /*
3424 * Search for the extent tree item in the root tree.
3425 */
3426 sk->tree_id = BTRFS_ROOT_TREE_OBJECTID;
3427 sk->min_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3428 sk->max_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3429 sk->min_type = BTRFS_ROOT_ITEM_KEY;
3430 sk->max_type = BTRFS_ROOT_ITEM_KEY;
3431 sk->min_offset = 0;
3432 sk->max_offset = (u64)-1;
3433 sk->min_transid = 0;
3434 sk->max_transid = (u64)-1;
3435 sk->nr_items = 1;
3436 args2->buf_size = args2_size - sizeof(struct btrfs_ioctl_search_args_v2);
3437 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH_V2, args2);
3438 if (ret == -EOPNOTSUPP)
3439 v2_supported = 0;
3440 else if (ret == 0)
3441 v2_supported = 1;
3442 else
3443 return ret;
3444
3445 return v2_supported;
3446 }
3447
3448 int btrfs_check_nodesize(u32 nodesize, u32 sectorsize, u64 features)
3449 {
3450 if (nodesize < sectorsize) {
3451 error("illegal nodesize %u (smaller than %u)",
3452 nodesize, sectorsize);
3453 return -1;
3454 } else if (nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) {
3455 error("illegal nodesize %u (larger than %u)",
3456 nodesize, BTRFS_MAX_METADATA_BLOCKSIZE);
3457 return -1;
3458 } else if (nodesize & (sectorsize - 1)) {
3459 error("illegal nodesize %u (not aligned to %u)",
3460 nodesize, sectorsize);
3461 return -1;
3462 } else if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS &&
3463 nodesize != sectorsize) {
3464 error("illegal nodesize %u (not equal to %u for mixed block group)",
3465 nodesize, sectorsize);
3466 return -1;
3467 }
3468 return 0;
3469 }
3470
3471 /*
3472 * Copy a path argument from SRC to DEST and check the SRC length if it's at
3473 * most PATH_MAX and fits into DEST. DESTLEN is supposed to be exact size of
3474 * the buffer.
3475 * The destination buffer is zero terminated.
3476 * Return < 0 for error, 0 otherwise.
3477 */
3478 int arg_copy_path(char *dest, const char *src, int destlen)
3479 {
3480 size_t len = strlen(src);
3481
3482 if (len >= PATH_MAX || len >= destlen)
3483 return -ENAMETOOLONG;
3484
3485 __strncpy_null(dest, src, destlen);
3486
3487 return 0;
3488 }
3489
3490 unsigned int get_unit_mode_from_arg(int *argc, char *argv[], int df_mode)
3491 {
3492 unsigned int unit_mode = UNITS_DEFAULT;
3493 int arg_i;
3494 int arg_end;
3495
3496 for (arg_i = 0; arg_i < *argc; arg_i++) {
3497 if (!strcmp(argv[arg_i], "--"))
3498 break;
3499
3500 if (!strcmp(argv[arg_i], "--raw")) {
3501 unit_mode = UNITS_RAW;
3502 argv[arg_i] = NULL;
3503 continue;
3504 }
3505 if (!strcmp(argv[arg_i], "--human-readable")) {
3506 unit_mode = UNITS_HUMAN_BINARY;
3507 argv[arg_i] = NULL;
3508 continue;
3509 }
3510
3511 if (!strcmp(argv[arg_i], "--iec")) {
3512 units_set_mode(&unit_mode, UNITS_BINARY);
3513 argv[arg_i] = NULL;
3514 continue;
3515 }
3516 if (!strcmp(argv[arg_i], "--si")) {
3517 units_set_mode(&unit_mode, UNITS_DECIMAL);
3518 argv[arg_i] = NULL;
3519 continue;
3520 }
3521
3522 if (!strcmp(argv[arg_i], "--kbytes")) {
3523 units_set_base(&unit_mode, UNITS_KBYTES);
3524 argv[arg_i] = NULL;
3525 continue;
3526 }
3527 if (!strcmp(argv[arg_i], "--mbytes")) {
3528 units_set_base(&unit_mode, UNITS_MBYTES);
3529 argv[arg_i] = NULL;
3530 continue;
3531 }
3532 if (!strcmp(argv[arg_i], "--gbytes")) {
3533 units_set_base(&unit_mode, UNITS_GBYTES);
3534 argv[arg_i] = NULL;
3535 continue;
3536 }
3537 if (!strcmp(argv[arg_i], "--tbytes")) {
3538 units_set_base(&unit_mode, UNITS_TBYTES);
3539 argv[arg_i] = NULL;
3540 continue;
3541 }
3542
3543 if (!df_mode)
3544 continue;
3545
3546 if (!strcmp(argv[arg_i], "-b")) {
3547 unit_mode = UNITS_RAW;
3548 argv[arg_i] = NULL;
3549 continue;
3550 }
3551 if (!strcmp(argv[arg_i], "-h")) {
3552 unit_mode = UNITS_HUMAN_BINARY;
3553 argv[arg_i] = NULL;
3554 continue;
3555 }
3556 if (!strcmp(argv[arg_i], "-H")) {
3557 unit_mode = UNITS_HUMAN_DECIMAL;
3558 argv[arg_i] = NULL;
3559 continue;
3560 }
3561 if (!strcmp(argv[arg_i], "-k")) {
3562 units_set_base(&unit_mode, UNITS_KBYTES);
3563 argv[arg_i] = NULL;
3564 continue;
3565 }
3566 if (!strcmp(argv[arg_i], "-m")) {
3567 units_set_base(&unit_mode, UNITS_MBYTES);
3568 argv[arg_i] = NULL;
3569 continue;
3570 }
3571 if (!strcmp(argv[arg_i], "-g")) {
3572 units_set_base(&unit_mode, UNITS_GBYTES);
3573 argv[arg_i] = NULL;
3574 continue;
3575 }
3576 if (!strcmp(argv[arg_i], "-t")) {
3577 units_set_base(&unit_mode, UNITS_TBYTES);
3578 argv[arg_i] = NULL;
3579 continue;
3580 }
3581 }
3582
3583 for (arg_i = 0, arg_end = 0; arg_i < *argc; arg_i++) {
3584 if (!argv[arg_i])
3585 continue;
3586 argv[arg_end] = argv[arg_i];
3587 arg_end++;
3588 }
3589
3590 *argc = arg_end;
3591
3592 return unit_mode;
3593 }
3594
3595 int string_is_numerical(const char *str)
3596 {
3597 if (!(*str >= '0' && *str <= '9'))
3598 return 0;
3599 while (*str >= '0' && *str <= '9')
3600 str++;
3601 if (*str != '\0')
3602 return 0;
3603 return 1;
3604 }
3605
3606 /*
3607 * Preprocess @argv with getopt_long to reorder options and consume the "--"
3608 * option separator.
3609 * Unknown short and long options are reported, optionally the @usage is printed
3610 * before exit.
3611 */
3612 void clean_args_no_options(int argc, char *argv[], const char * const *usagestr)
3613 {
3614 static const struct option long_options[] = {
3615 {NULL, 0, NULL, 0}
3616 };
3617
3618 while (1) {
3619 int c = getopt_long(argc, argv, "", long_options, NULL);
3620
3621 if (c < 0)
3622 break;
3623
3624 switch (c) {
3625 default:
3626 if (usagestr)
3627 usage(usagestr);
3628 }
3629 }
3630 }
3631
3632 /* Subvolume helper functions */
3633 /*
3634 * test if name is a correct subvolume name
3635 * this function return
3636 * 0-> name is not a correct subvolume name
3637 * 1-> name is a correct subvolume name
3638 */
3639 int test_issubvolname(const char *name)
3640 {
3641 return name[0] != '\0' && !strchr(name, '/') &&
3642 strcmp(name, ".") && strcmp(name, "..");
3643 }
3644
3645 /*
3646 * Test if path is a subvolume
3647 * Returns:
3648 * 0 - path exists but it is not a subvolume
3649 * 1 - path exists and it is a subvolume
3650 * < 0 - error
3651 */
3652 int test_issubvolume(const char *path)
3653 {
3654 struct stat st;
3655 struct statfs stfs;
3656 int res;
3657
3658 res = stat(path, &st);
3659 if (res < 0)
3660 return -errno;
3661
3662 if (st.st_ino != BTRFS_FIRST_FREE_OBJECTID || !S_ISDIR(st.st_mode))
3663 return 0;
3664
3665 res = statfs(path, &stfs);
3666 if (res < 0)
3667 return -errno;
3668
3669 return (int)stfs.f_type == BTRFS_SUPER_MAGIC;
3670 }
3671
3672 const char *subvol_strip_mountpoint(const char *mnt, const char *full_path)
3673 {
3674 int len = strlen(mnt);
3675 if (!len)
3676 return full_path;
3677
3678 if (mnt[len - 1] != '/')
3679 len += 1;
3680
3681 return full_path + len;
3682 }
3683
3684 /*
3685 * Returns
3686 * <0: Std error
3687 * 0: All fine
3688 * 1: Error; and error info printed to the terminal. Fixme.
3689 * 2: If the fullpath is root tree instead of subvol tree
3690 */
3691 int get_subvol_info(const char *fullpath, struct root_info *get_ri)
3692 {
3693 u64 sv_id;
3694 int ret = 1;
3695 int fd = -1;
3696 int mntfd = -1;
3697 char *mnt = NULL;
3698 const char *svpath = NULL;
3699 DIR *dirstream1 = NULL;
3700 DIR *dirstream2 = NULL;
3701
3702 ret = test_issubvolume(fullpath);
3703 if (ret < 0)
3704 return ret;
3705 if (!ret) {
3706 error("not a subvolume: %s", fullpath);
3707 return 1;
3708 }
3709
3710 ret = find_mount_root(fullpath, &mnt);
3711 if (ret < 0)
3712 return ret;
3713 if (ret > 0) {
3714 error("%s doesn't belong to btrfs mount point", fullpath);
3715 return 1;
3716 }
3717 ret = 1;
3718 svpath = subvol_strip_mountpoint(mnt, fullpath);
3719
3720 fd = btrfs_open_dir(fullpath, &dirstream1, 1);
3721 if (fd < 0)
3722 goto out;
3723
3724 ret = btrfs_list_get_path_rootid(fd, &sv_id);
3725 if (ret) {
3726 error("can't get rootid for '%s'", fullpath);
3727 goto out;
3728 }
3729
3730 mntfd = btrfs_open_dir(mnt, &dirstream2, 1);
3731 if (mntfd < 0)
3732 goto out;
3733
3734 if (sv_id == BTRFS_FS_TREE_OBJECTID) {
3735 ret = 2;
3736 /*
3737 * So that caller may decide if thats an error or just fine.
3738 */
3739 goto out;
3740 }
3741
3742 memset(get_ri, 0, sizeof(*get_ri));
3743 get_ri->root_id = sv_id;
3744
3745 ret = btrfs_get_subvol(mntfd, get_ri);
3746 if (ret)
3747 error("can't find '%s': %d", svpath, ret);
3748
3749 out:
3750 close_file_or_dir(mntfd, dirstream2);
3751 close_file_or_dir(fd, dirstream1);
3752 free(mnt);
3753
3754 return ret;
3755 }
3756
3757 void init_rand_seed(u64 seed)
3758 {
3759 int i;
3760
3761 /* only use the last 48 bits */
3762 for (i = 0; i < 3; i++) {
3763 rand_seed[i] = (unsigned short)(seed ^ (unsigned short)(-1));
3764 seed >>= 16;
3765 }
3766 rand_seed_initlized = 1;
3767 }
3768
3769 static void __init_seed(void)
3770 {
3771 struct timeval tv;
3772 int ret;
3773 int fd;
3774
3775 if(rand_seed_initlized)
3776 return;
3777 /* Use urandom as primary seed source. */
3778 fd = open("/dev/urandom", O_RDONLY);
3779 if (fd >= 0) {
3780 ret = read(fd, rand_seed, sizeof(rand_seed));
3781 close(fd);
3782 if (ret < sizeof(rand_seed))
3783 goto fallback;
3784 } else {
3785 fallback:
3786 /* Use time and pid as fallback seed */
3787 warning("failed to read /dev/urandom, use time and pid as random seed");
3788 gettimeofday(&tv, 0);
3789 rand_seed[0] = getpid() ^ (tv.tv_sec & 0xFFFF);
3790 rand_seed[1] = getppid() ^ (tv.tv_usec & 0xFFFF);
3791 rand_seed[2] = (tv.tv_sec ^ tv.tv_usec) >> 16;
3792 }
3793 rand_seed_initlized = 1;
3794 }
3795
3796 u32 rand_u32(void)
3797 {
3798 __init_seed();
3799 /*
3800 * Don't use nrand48, its range is [0,2^31) The highest bit will alwasy
3801 * be 0. Use jrand48 to include the highest bit.
3802 */
3803 return (u32)jrand48(rand_seed);
3804 }
3805
3806 unsigned int rand_range(unsigned int upper)
3807 {
3808 __init_seed();
3809 /*
3810 * Use the full 48bits to mod, which would be more uniformly
3811 * distributed
3812 */
3813 return (unsigned int)(jrand48(rand_seed) % upper);
3814 }
(deprecated) Btrfs progs developments and patch integration