2 * Copyright (C) 2007 Oracle. All rights reserved.
3 * Copyright (C) 2008 Morey Roof. All rights reserved.
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.
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.
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.
23 #include <sys/ioctl.h>
24 #include <sys/mount.h>
25 #include <sys/types.h>
27 #include <uuid/uuid.h>
32 #include <linux/loop.h>
33 #include <linux/major.h>
34 #include <linux/kdev_t.h>
36 #include <blkid/blkid.h>
39 #include "kerncompat.h"
40 #include "radix-tree.h"
43 #include "transaction.h"
50 #define BLKDISCARD _IO(0x12,119)
53 static int btrfs_scan_done = 0;
55 static char argv0_buf[ARGV0_BUF_SIZE] = "btrfs";
57 void fixup_argv0(char **argv, const char *token)
59 int len = strlen(argv0_buf);
61 snprintf(argv0_buf + len, sizeof(argv0_buf) - len, " %s", token);
65 void set_argv0(char **argv)
67 strncpy(argv0_buf, argv[0], sizeof(argv0_buf));
68 argv0_buf[sizeof(argv0_buf) - 1] = 0;
71 int check_argc_exact(int nargs, int expected)
74 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
76 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
78 return nargs != expected;
81 int check_argc_min(int nargs, int expected)
83 if (nargs < expected) {
84 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
91 int check_argc_max(int nargs, int expected)
93 if (nargs > expected) {
94 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
103 * Discard the given range in one go
105 static int discard_range(int fd, u64 start, u64 len)
107 u64 range[2] = { start, len };
109 if (ioctl(fd, BLKDISCARD, &range) < 0)
115 * Discard blocks in the given range in 1G chunks, the process is interruptible
117 static int discard_blocks(int fd, u64 start, u64 len)
121 u64 chunk_size = min_t(u64, len, 1*1024*1024*1024);
124 ret = discard_range(fd, start, chunk_size);
134 static u64 reference_root_table[] = {
135 [1] = BTRFS_ROOT_TREE_OBJECTID,
136 [2] = BTRFS_EXTENT_TREE_OBJECTID,
137 [3] = BTRFS_CHUNK_TREE_OBJECTID,
138 [4] = BTRFS_DEV_TREE_OBJECTID,
139 [5] = BTRFS_FS_TREE_OBJECTID,
140 [6] = BTRFS_CSUM_TREE_OBJECTID,
143 int test_uuid_unique(char *fs_uuid)
146 blkid_dev_iterate iter = NULL;
147 blkid_dev dev = NULL;
148 blkid_cache cache = NULL;
150 if (blkid_get_cache(&cache, 0) < 0) {
151 printf("ERROR: lblkid cache get failed\n");
154 blkid_probe_all(cache);
155 iter = blkid_dev_iterate_begin(cache);
156 blkid_dev_set_search(iter, "UUID", fs_uuid);
158 while (blkid_dev_next(iter, &dev) == 0) {
159 dev = blkid_verify(cache, dev);
166 blkid_dev_iterate_end(iter);
167 blkid_put_cache(cache);
172 int make_btrfs(int fd, const char *device, const char *label, char *fs_uuid,
173 u64 blocks[7], u64 num_bytes, u32 nodesize,
174 u32 sectorsize, u32 stripesize, u64 features)
176 struct btrfs_super_block super;
177 struct extent_buffer *buf = NULL;
178 struct btrfs_root_item root_item;
179 struct btrfs_disk_key disk_key;
180 struct btrfs_extent_item *extent_item;
181 struct btrfs_inode_item *inode_item;
182 struct btrfs_chunk *chunk;
183 struct btrfs_dev_item *dev_item;
184 struct btrfs_dev_extent *dev_extent;
185 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
195 int skinny_metadata = !!(features &
196 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA);
198 first_free = BTRFS_SUPER_INFO_OFFSET + sectorsize * 2 - 1;
199 first_free &= ~((u64)sectorsize - 1);
201 memset(&super, 0, sizeof(super));
203 num_bytes = (num_bytes / sectorsize) * sectorsize;
205 if (uuid_parse(fs_uuid, super.fsid) != 0) {
206 fprintf(stderr, "could not parse UUID: %s\n", fs_uuid);
210 if (!test_uuid_unique(fs_uuid)) {
211 fprintf(stderr, "non-unique UUID: %s\n", fs_uuid);
216 uuid_generate(super.fsid);
218 uuid_generate(super.dev_item.uuid);
219 uuid_generate(chunk_tree_uuid);
221 btrfs_set_super_bytenr(&super, blocks[0]);
222 btrfs_set_super_num_devices(&super, 1);
223 btrfs_set_super_magic(&super, BTRFS_MAGIC);
224 btrfs_set_super_generation(&super, 1);
225 btrfs_set_super_root(&super, blocks[1]);
226 btrfs_set_super_chunk_root(&super, blocks[3]);
227 btrfs_set_super_total_bytes(&super, num_bytes);
228 btrfs_set_super_bytes_used(&super, 6 * nodesize);
229 btrfs_set_super_sectorsize(&super, sectorsize);
230 btrfs_set_super_leafsize(&super, nodesize);
231 btrfs_set_super_nodesize(&super, nodesize);
232 btrfs_set_super_stripesize(&super, stripesize);
233 btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
234 btrfs_set_super_chunk_root_generation(&super, 1);
235 btrfs_set_super_cache_generation(&super, -1);
236 btrfs_set_super_incompat_flags(&super, features);
238 strncpy(super.label, label, BTRFS_LABEL_SIZE - 1);
240 buf = malloc(sizeof(*buf) + max(sectorsize, nodesize));
242 /* create the tree of root objects */
243 memset(buf->data, 0, nodesize);
245 btrfs_set_header_bytenr(buf, blocks[1]);
246 btrfs_set_header_nritems(buf, 4);
247 btrfs_set_header_generation(buf, 1);
248 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
249 btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
250 write_extent_buffer(buf, super.fsid, btrfs_header_fsid(),
253 write_extent_buffer(buf, chunk_tree_uuid,
254 btrfs_header_chunk_tree_uuid(buf),
257 /* create the items for the root tree */
258 memset(&root_item, 0, sizeof(root_item));
259 inode_item = &root_item.inode;
260 btrfs_set_stack_inode_generation(inode_item, 1);
261 btrfs_set_stack_inode_size(inode_item, 3);
262 btrfs_set_stack_inode_nlink(inode_item, 1);
263 btrfs_set_stack_inode_nbytes(inode_item, nodesize);
264 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
265 btrfs_set_root_refs(&root_item, 1);
266 btrfs_set_root_used(&root_item, nodesize);
267 btrfs_set_root_generation(&root_item, 1);
269 memset(&disk_key, 0, sizeof(disk_key));
270 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
271 btrfs_set_disk_key_offset(&disk_key, 0);
274 itemoff = __BTRFS_LEAF_DATA_SIZE(nodesize) - sizeof(root_item);
275 btrfs_set_root_bytenr(&root_item, blocks[2]);
276 btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
277 btrfs_set_item_key(buf, &disk_key, nritems);
278 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
279 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
281 write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
282 nritems), sizeof(root_item));
285 itemoff = itemoff - sizeof(root_item);
286 btrfs_set_root_bytenr(&root_item, blocks[4]);
287 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
288 btrfs_set_item_key(buf, &disk_key, nritems);
289 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
290 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
292 write_extent_buffer(buf, &root_item,
293 btrfs_item_ptr_offset(buf, nritems),
297 itemoff = itemoff - sizeof(root_item);
298 btrfs_set_root_bytenr(&root_item, blocks[5]);
299 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
300 btrfs_set_item_key(buf, &disk_key, nritems);
301 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
302 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
304 write_extent_buffer(buf, &root_item,
305 btrfs_item_ptr_offset(buf, nritems),
309 itemoff = itemoff - sizeof(root_item);
310 btrfs_set_root_bytenr(&root_item, blocks[6]);
311 btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_TREE_OBJECTID);
312 btrfs_set_item_key(buf, &disk_key, nritems);
313 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
314 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
316 write_extent_buffer(buf, &root_item,
317 btrfs_item_ptr_offset(buf, nritems),
322 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
323 ret = pwrite(fd, buf->data, nodesize, blocks[1]);
324 if (ret != nodesize) {
325 ret = (ret < 0 ? -errno : -EIO);
329 /* create the items for the extent tree */
330 memset(buf->data + sizeof(struct btrfs_header), 0,
331 nodesize - sizeof(struct btrfs_header));
333 itemoff = __BTRFS_LEAF_DATA_SIZE(nodesize);
334 for (i = 1; i < 7; i++) {
335 item_size = sizeof(struct btrfs_extent_item);
336 if (!skinny_metadata)
337 item_size += sizeof(struct btrfs_tree_block_info);
339 BUG_ON(blocks[i] < first_free);
340 BUG_ON(blocks[i] < blocks[i - 1]);
342 /* create extent item */
343 itemoff -= item_size;
344 btrfs_set_disk_key_objectid(&disk_key, blocks[i]);
345 if (skinny_metadata) {
346 btrfs_set_disk_key_type(&disk_key,
347 BTRFS_METADATA_ITEM_KEY);
348 btrfs_set_disk_key_offset(&disk_key, 0);
350 btrfs_set_disk_key_type(&disk_key,
351 BTRFS_EXTENT_ITEM_KEY);
352 btrfs_set_disk_key_offset(&disk_key, nodesize);
354 btrfs_set_item_key(buf, &disk_key, nritems);
355 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
357 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
359 extent_item = btrfs_item_ptr(buf, nritems,
360 struct btrfs_extent_item);
361 btrfs_set_extent_refs(buf, extent_item, 1);
362 btrfs_set_extent_generation(buf, extent_item, 1);
363 btrfs_set_extent_flags(buf, extent_item,
364 BTRFS_EXTENT_FLAG_TREE_BLOCK);
367 /* create extent ref */
368 ref_root = reference_root_table[i];
369 btrfs_set_disk_key_objectid(&disk_key, blocks[i]);
370 btrfs_set_disk_key_offset(&disk_key, ref_root);
371 btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
372 btrfs_set_item_key(buf, &disk_key, nritems);
373 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
375 btrfs_set_item_size(buf, btrfs_item_nr(nritems), 0);
378 btrfs_set_header_bytenr(buf, blocks[2]);
379 btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
380 btrfs_set_header_nritems(buf, nritems);
381 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
382 ret = pwrite(fd, buf->data, nodesize, blocks[2]);
383 if (ret != nodesize) {
384 ret = (ret < 0 ? -errno : -EIO);
388 /* create the chunk tree */
389 memset(buf->data + sizeof(struct btrfs_header), 0,
390 nodesize - sizeof(struct btrfs_header));
392 item_size = sizeof(*dev_item);
393 itemoff = __BTRFS_LEAF_DATA_SIZE(nodesize) - item_size;
395 /* first device 1 (there is no device 0) */
396 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
397 btrfs_set_disk_key_offset(&disk_key, 1);
398 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
399 btrfs_set_item_key(buf, &disk_key, nritems);
400 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
401 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
403 dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
404 btrfs_set_device_id(buf, dev_item, 1);
405 btrfs_set_device_generation(buf, dev_item, 0);
406 btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
407 btrfs_set_device_bytes_used(buf, dev_item,
408 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
409 btrfs_set_device_io_align(buf, dev_item, sectorsize);
410 btrfs_set_device_io_width(buf, dev_item, sectorsize);
411 btrfs_set_device_sector_size(buf, dev_item, sectorsize);
412 btrfs_set_device_type(buf, dev_item, 0);
414 write_extent_buffer(buf, super.dev_item.uuid,
415 (unsigned long)btrfs_device_uuid(dev_item),
417 write_extent_buffer(buf, super.fsid,
418 (unsigned long)btrfs_device_fsid(dev_item),
420 read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
424 item_size = btrfs_chunk_item_size(1);
425 itemoff = itemoff - item_size;
427 /* then we have chunk 0 */
428 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
429 btrfs_set_disk_key_offset(&disk_key, 0);
430 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
431 btrfs_set_item_key(buf, &disk_key, nritems);
432 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
433 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
435 chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
436 btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
437 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
438 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
439 btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
440 btrfs_set_chunk_io_align(buf, chunk, sectorsize);
441 btrfs_set_chunk_io_width(buf, chunk, sectorsize);
442 btrfs_set_chunk_sector_size(buf, chunk, sectorsize);
443 btrfs_set_chunk_num_stripes(buf, chunk, 1);
444 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
445 btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
448 write_extent_buffer(buf, super.dev_item.uuid,
449 (unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
452 /* copy the key for the chunk to the system array */
453 ptr = super.sys_chunk_array;
454 array_size = sizeof(disk_key);
456 memcpy(ptr, &disk_key, sizeof(disk_key));
457 ptr += sizeof(disk_key);
459 /* copy the chunk to the system array */
460 read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
461 array_size += item_size;
463 btrfs_set_super_sys_array_size(&super, array_size);
465 btrfs_set_header_bytenr(buf, blocks[3]);
466 btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
467 btrfs_set_header_nritems(buf, nritems);
468 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
469 ret = pwrite(fd, buf->data, nodesize, blocks[3]);
470 if (ret != nodesize) {
471 ret = (ret < 0 ? -errno : -EIO);
475 /* create the device tree */
476 memset(buf->data + sizeof(struct btrfs_header), 0,
477 nodesize - sizeof(struct btrfs_header));
479 itemoff = __BTRFS_LEAF_DATA_SIZE(nodesize) -
480 sizeof(struct btrfs_dev_extent);
482 btrfs_set_disk_key_objectid(&disk_key, 1);
483 btrfs_set_disk_key_offset(&disk_key, 0);
484 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
485 btrfs_set_item_key(buf, &disk_key, nritems);
486 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
487 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
488 sizeof(struct btrfs_dev_extent));
489 dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
490 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
491 BTRFS_CHUNK_TREE_OBJECTID);
492 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
493 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
494 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
496 write_extent_buffer(buf, chunk_tree_uuid,
497 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
500 btrfs_set_dev_extent_length(buf, dev_extent,
501 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
504 btrfs_set_header_bytenr(buf, blocks[4]);
505 btrfs_set_header_owner(buf, BTRFS_DEV_TREE_OBJECTID);
506 btrfs_set_header_nritems(buf, nritems);
507 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
508 ret = pwrite(fd, buf->data, nodesize, blocks[4]);
509 if (ret != nodesize) {
510 ret = (ret < 0 ? -errno : -EIO);
514 /* create the FS root */
515 memset(buf->data + sizeof(struct btrfs_header), 0,
516 nodesize - sizeof(struct btrfs_header));
517 btrfs_set_header_bytenr(buf, blocks[5]);
518 btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
519 btrfs_set_header_nritems(buf, 0);
520 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
521 ret = pwrite(fd, buf->data, nodesize, blocks[5]);
522 if (ret != nodesize) {
523 ret = (ret < 0 ? -errno : -EIO);
526 /* finally create the csum root */
527 memset(buf->data + sizeof(struct btrfs_header), 0,
528 nodesize - sizeof(struct btrfs_header));
529 btrfs_set_header_bytenr(buf, blocks[6]);
530 btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
531 btrfs_set_header_nritems(buf, 0);
532 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
533 ret = pwrite(fd, buf->data, nodesize, blocks[6]);
534 if (ret != nodesize) {
535 ret = (ret < 0 ? -errno : -EIO);
539 /* and write out the super block */
540 BUG_ON(sizeof(super) > sectorsize);
541 memset(buf->data, 0, sectorsize);
542 memcpy(buf->data, &super, sizeof(super));
543 buf->len = sectorsize;
544 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
545 ret = pwrite(fd, buf->data, sectorsize, blocks[0]);
546 if (ret != sectorsize) {
547 ret = (ret < 0 ? -errno : -EIO);
558 static const struct btrfs_fs_feature {
562 } mkfs_features[] = {
563 { "mixed-bg", BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS,
564 "mixed data and metadata block groups" },
565 { "extref", BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF,
566 "increased hardlink limit per file to 65536" },
567 { "raid56", BTRFS_FEATURE_INCOMPAT_RAID56,
568 "raid56 extended format" },
569 { "skinny-metadata", BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA,
570 "reduced-size metadata extent refs" },
571 { "no-holes", BTRFS_FEATURE_INCOMPAT_NO_HOLES,
572 "no explicit hole extents for files" },
573 /* Keep this one last */
574 { "list-all", BTRFS_FEATURE_LIST_ALL, NULL }
577 static int parse_one_fs_feature(const char *name, u64 *flags)
582 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
583 if (name[0] == '^' &&
584 !strcmp(mkfs_features[i].name, name + 1)) {
585 *flags &= ~ mkfs_features[i].flag;
587 } else if (!strcmp(mkfs_features[i].name, name)) {
588 *flags |= mkfs_features[i].flag;
596 void btrfs_process_fs_features(u64 flags)
600 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
601 if (flags & mkfs_features[i].flag) {
602 printf("Turning ON incompat feature '%s': %s\n",
603 mkfs_features[i].name,
604 mkfs_features[i].desc);
609 void btrfs_list_all_fs_features(u64 mask_disallowed)
613 fprintf(stderr, "Filesystem features available:\n");
614 for (i = 0; i < ARRAY_SIZE(mkfs_features) - 1; i++) {
615 char *is_default = "";
617 if (mkfs_features[i].flag & mask_disallowed)
619 if (mkfs_features[i].flag & BTRFS_MKFS_DEFAULT_FEATURES)
620 is_default = ", default";
621 fprintf(stderr, "%-20s- %s (0x%llx%s)\n",
622 mkfs_features[i].name,
623 mkfs_features[i].desc,
624 mkfs_features[i].flag,
630 * Return NULL if all features were parsed fine, otherwise return the name of
631 * the first unparsed.
633 char* btrfs_parse_fs_features(char *namelist, u64 *flags)
636 char *save_ptr = NULL; /* Satisfy static checkers */
638 for (this_char = strtok_r(namelist, ",", &save_ptr);
640 this_char = strtok_r(NULL, ",", &save_ptr)) {
641 if (parse_one_fs_feature(this_char, flags))
648 u64 btrfs_device_size(int fd, struct stat *st)
651 if (S_ISREG(st->st_mode)) {
654 if (!S_ISBLK(st->st_mode)) {
657 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
663 static int zero_blocks(int fd, off_t start, size_t len)
665 char *buf = malloc(len);
672 written = pwrite(fd, buf, len, start);
679 #define ZERO_DEV_BYTES (2 * 1024 * 1024)
681 /* don't write outside the device by clamping the region to the device size */
682 static int zero_dev_clamped(int fd, off_t start, ssize_t len, u64 dev_size)
684 off_t end = max(start, start + len);
687 /* and don't overwrite the disk labels on sparc */
688 start = max(start, 1024);
689 end = max(end, 1024);
692 start = min_t(u64, start, dev_size);
693 end = min_t(u64, end, dev_size);
695 return zero_blocks(fd, start, end - start);
698 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
699 struct btrfs_root *root, int fd, char *path,
700 u64 block_count, u32 io_width, u32 io_align,
703 struct btrfs_super_block *disk_super;
704 struct btrfs_super_block *super = root->fs_info->super_copy;
705 struct btrfs_device *device;
706 struct btrfs_dev_item *dev_item;
712 device = kzalloc(sizeof(*device), GFP_NOFS);
715 buf = kmalloc(sectorsize, GFP_NOFS);
720 BUG_ON(sizeof(*disk_super) > sectorsize);
721 memset(buf, 0, sectorsize);
723 disk_super = (struct btrfs_super_block *)buf;
724 dev_item = &disk_super->dev_item;
726 uuid_generate(device->uuid);
729 device->io_width = io_width;
730 device->io_align = io_align;
731 device->sector_size = sectorsize;
733 device->writeable = 1;
734 device->total_bytes = block_count;
735 device->bytes_used = 0;
736 device->total_ios = 0;
737 device->dev_root = root->fs_info->dev_root;
739 ret = btrfs_add_device(trans, root, device);
742 total_bytes = btrfs_super_total_bytes(super) + block_count;
743 btrfs_set_super_total_bytes(super, total_bytes);
745 num_devs = btrfs_super_num_devices(super) + 1;
746 btrfs_set_super_num_devices(super, num_devs);
748 memcpy(disk_super, super, sizeof(*disk_super));
750 printf("adding device %s id %llu\n", path,
751 (unsigned long long)device->devid);
753 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
754 btrfs_set_stack_device_id(dev_item, device->devid);
755 btrfs_set_stack_device_type(dev_item, device->type);
756 btrfs_set_stack_device_io_align(dev_item, device->io_align);
757 btrfs_set_stack_device_io_width(dev_item, device->io_width);
758 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
759 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
760 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
761 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
763 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
764 BUG_ON(ret != sectorsize);
767 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
768 device->fs_devices = root->fs_info->fs_devices;
772 static void btrfs_wipe_existing_sb(int fd)
774 const char *off = NULL;
779 blkid_probe pr = NULL;
781 pr = blkid_new_probe();
785 if (blkid_probe_set_device(pr, fd, 0, 0))
788 rc = blkid_probe_lookup_value(pr, "SBMAGIC_OFFSET", &off, NULL);
790 rc = blkid_probe_lookup_value(pr, "SBMAGIC", NULL, &len);
792 if (rc || len == 0 || off == NULL)
795 offset = strtoll(off, NULL, 10);
796 if (len > sizeof(buf))
800 rc = pwrite(fd, buf, len, offset);
804 blkid_free_probe(pr);
808 int btrfs_prepare_device(int fd, char *file, int zero_end, u64 *block_count_ret,
809 u64 max_block_count, int *mixed, int discard)
815 ret = fstat(fd, &st);
817 fprintf(stderr, "unable to stat %s\n", file);
821 block_count = btrfs_device_size(fd, &st);
822 if (block_count == 0) {
823 fprintf(stderr, "unable to find %s size\n", file);
827 block_count = min(block_count, max_block_count);
829 if (block_count < BTRFS_MKFS_SMALL_VOLUME_SIZE && !(*mixed))
834 * We intentionally ignore errors from the discard ioctl. It
835 * is not necessary for the mkfs functionality but just an
838 if (discard_range(fd, 0, 0) == 0) {
839 printf("Performing full device TRIM (%s) ...\n",
840 pretty_size(block_count));
841 discard_blocks(fd, 0, block_count);
845 ret = zero_dev_clamped(fd, 0, ZERO_DEV_BYTES, block_count);
846 for (i = 0 ; !ret && i < BTRFS_SUPER_MIRROR_MAX; i++)
847 ret = zero_dev_clamped(fd, btrfs_sb_offset(i),
848 BTRFS_SUPER_INFO_SIZE, block_count);
849 if (!ret && zero_end)
850 ret = zero_dev_clamped(fd, block_count - ZERO_DEV_BYTES,
851 ZERO_DEV_BYTES, block_count);
854 fprintf(stderr, "ERROR: failed to zero device '%s' - %s\n",
855 file, strerror(-ret));
859 btrfs_wipe_existing_sb(fd);
861 *block_count_ret = block_count;
865 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
866 struct btrfs_root *root, u64 objectid)
869 struct btrfs_inode_item inode_item;
870 time_t now = time(NULL);
872 memset(&inode_item, 0, sizeof(inode_item));
873 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
874 btrfs_set_stack_inode_size(&inode_item, 0);
875 btrfs_set_stack_inode_nlink(&inode_item, 1);
876 btrfs_set_stack_inode_nbytes(&inode_item, root->nodesize);
877 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
878 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
879 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
880 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
881 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
882 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
883 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
884 btrfs_set_stack_timespec_sec(&inode_item.otime, 0);
885 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
887 if (root->fs_info->tree_root == root)
888 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
890 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
894 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
898 btrfs_set_root_dirid(&root->root_item, objectid);
905 * checks if a path is a block device node
906 * Returns negative errno on failure, otherwise
907 * returns 1 for blockdev, 0 for not-blockdev
909 int is_block_device(const char *path)
913 if (stat(path, &statbuf) < 0)
916 return S_ISBLK(statbuf.st_mode);
920 * check if given path is a mount point
921 * return 1 if yes. 0 if no. -1 for error
923 int is_mount_point(const char *path)
929 f = setmntent("/proc/self/mounts", "r");
933 while ((mnt = getmntent(f)) != NULL) {
934 if (strcmp(mnt->mnt_dir, path))
943 static int is_reg_file(const char *path)
947 if (stat(path, &statbuf) < 0)
949 return S_ISREG(statbuf.st_mode);
953 * This function checks if the given input parameter is
955 * return <0 : some error in the given input
956 * return BTRFS_ARG_UNKNOWN: unknown input
957 * return BTRFS_ARG_UUID: given input is uuid
958 * return BTRFS_ARG_MNTPOINT: given input is path
959 * return BTRFS_ARG_REG: given input is regular file
961 int check_arg_type(const char *input)
969 if (realpath(input, path)) {
970 if (is_block_device(path) == 1)
971 return BTRFS_ARG_BLKDEV;
973 if (is_mount_point(path) == 1)
974 return BTRFS_ARG_MNTPOINT;
976 if (is_reg_file(path))
977 return BTRFS_ARG_REG;
979 return BTRFS_ARG_UNKNOWN;
982 if (strlen(input) == (BTRFS_UUID_UNPARSED_SIZE - 1) &&
983 !uuid_parse(input, uuid))
984 return BTRFS_ARG_UUID;
986 return BTRFS_ARG_UNKNOWN;
990 * Find the mount point for a mounted device.
991 * On success, returns 0 with mountpoint in *mp.
992 * On failure, returns -errno (not mounted yields -EINVAL)
993 * Is noisy on failures, expects to be given a mounted device.
995 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size)
1000 ret = is_block_device(dev);
1003 fprintf(stderr, "%s is not a block device\n", dev);
1006 fprintf(stderr, "Could not check %s: %s\n",
1007 dev, strerror(-ret));
1012 fd = open(dev, O_RDONLY);
1015 fprintf(stderr, "Could not open %s: %s\n", dev, strerror(errno));
1019 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
1022 } else { /* mounted, all good */
1032 * Given a pathname, return a filehandle to:
1033 * the original pathname or,
1034 * if the pathname is a mounted btrfs device, to its mountpoint.
1036 * On error, return -1, errno should be set.
1038 int open_path_or_dev_mnt(const char *path, DIR **dirstream)
1040 char mp[BTRFS_PATH_NAME_MAX + 1];
1043 if (is_block_device(path)) {
1046 ret = get_btrfs_mount(path, mp, sizeof(mp));
1048 /* not a mounted btrfs dev */
1052 fdmnt = open_file_or_dir(mp, dirstream);
1054 fdmnt = open_file_or_dir(path, dirstream);
1060 /* checks if a device is a loop device */
1061 static int is_loop_device (const char* device) {
1062 struct stat statbuf;
1064 if(stat(device, &statbuf) < 0)
1067 return (S_ISBLK(statbuf.st_mode) &&
1068 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
1072 /* Takes a loop device path (e.g. /dev/loop0) and returns
1073 * the associated file (e.g. /images/my_btrfs.img) */
1074 static int resolve_loop_device(const char* loop_dev, char* loop_file,
1081 char real_loop_dev[PATH_MAX];
1083 if (!realpath(loop_dev, real_loop_dev))
1085 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
1086 if (!(f = fopen(p, "r")))
1089 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
1090 ret = fscanf(f, fmt, loop_file);
1099 * Checks whether a and b are identical or device
1100 * files associated with the same block device
1102 static int is_same_blk_file(const char* a, const char* b)
1104 struct stat st_buf_a, st_buf_b;
1105 char real_a[PATH_MAX];
1106 char real_b[PATH_MAX];
1108 if (!realpath(a, real_a))
1109 strncpy_null(real_a, a);
1111 if (!realpath(b, real_b))
1112 strncpy_null(real_b, b);
1114 /* Identical path? */
1115 if (strcmp(real_a, real_b) == 0)
1118 if (stat(a, &st_buf_a) < 0 || stat(b, &st_buf_b) < 0) {
1119 if (errno == ENOENT)
1124 /* Same blockdevice? */
1125 if (S_ISBLK(st_buf_a.st_mode) && S_ISBLK(st_buf_b.st_mode) &&
1126 st_buf_a.st_rdev == st_buf_b.st_rdev) {
1131 if (st_buf_a.st_dev == st_buf_b.st_dev &&
1132 st_buf_a.st_ino == st_buf_b.st_ino) {
1139 /* checks if a and b are identical or device
1140 * files associated with the same block device or
1141 * if one file is a loop device that uses the other
1144 static int is_same_loop_file(const char* a, const char* b)
1146 char res_a[PATH_MAX];
1147 char res_b[PATH_MAX];
1148 const char* final_a = NULL;
1149 const char* final_b = NULL;
1152 /* Resolve a if it is a loop device */
1153 if((ret = is_loop_device(a)) < 0) {
1158 ret = resolve_loop_device(a, res_a, sizeof(res_a));
1169 /* Resolve b if it is a loop device */
1170 if ((ret = is_loop_device(b)) < 0) {
1175 ret = resolve_loop_device(b, res_b, sizeof(res_b));
1186 return is_same_blk_file(final_a, final_b);
1189 /* Checks if a file exists and is a block or regular file*/
1190 static int is_existing_blk_or_reg_file(const char* filename)
1194 if(stat(filename, &st_buf) < 0) {
1201 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
1204 /* Checks if a file is used (directly or indirectly via a loop device)
1205 * by a device in fs_devices
1207 static int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices,
1211 struct list_head *head;
1212 struct list_head *cur;
1213 struct btrfs_device *device;
1215 head = &fs_devices->devices;
1216 list_for_each(cur, head) {
1217 device = list_entry(cur, struct btrfs_device, dev_list);
1219 if((ret = is_same_loop_file(device->name, file)))
1227 * Resolve a pathname to a device mapper node to /dev/mapper/<name>
1228 * Returns NULL on invalid input or malloc failure; Other failures
1229 * will be handled by the caller using the input pathame.
1231 char *canonicalize_dm_name(const char *ptname)
1235 char path[PATH_MAX], name[PATH_MAX], *res = NULL;
1237 if (!ptname || !*ptname)
1240 snprintf(path, sizeof(path), "/sys/block/%s/dm/name", ptname);
1241 if (!(f = fopen(path, "r")))
1244 /* read <name>\n from sysfs */
1245 if (fgets(name, sizeof(name), f) && (sz = strlen(name)) > 1) {
1246 name[sz - 1] = '\0';
1247 snprintf(path, sizeof(path), "/dev/mapper/%s", name);
1249 if (access(path, F_OK) == 0)
1257 * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or
1258 * to a device mapper pathname.
1259 * Returns NULL on invalid input or malloc failure; Other failures
1260 * will be handled by the caller using the input pathame.
1262 char *canonicalize_path(const char *path)
1264 char *canonical, *p;
1266 if (!path || !*path)
1269 canonical = realpath(path, NULL);
1271 return strdup(path);
1272 p = strrchr(canonical, '/');
1273 if (p && strncmp(p, "/dm-", 4) == 0 && isdigit(*(p + 4))) {
1274 char *dm = canonicalize_dm_name(p + 1);
1285 * returns 1 if the device was mounted, < 0 on error or 0 if everything
1286 * is safe to continue.
1288 int check_mounted(const char* file)
1293 fd = open(file, O_RDONLY);
1295 fprintf (stderr, "check_mounted(): Could not open %s\n", file);
1299 ret = check_mounted_where(fd, file, NULL, 0, NULL);
1305 int check_mounted_where(int fd, const char *file, char *where, int size,
1306 struct btrfs_fs_devices **fs_dev_ret)
1311 struct btrfs_fs_devices *fs_devices_mnt = NULL;
1315 /* scan the initial device */
1316 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
1317 &total_devs, BTRFS_SUPER_INFO_OFFSET, 0);
1318 is_btrfs = (ret >= 0);
1320 /* scan other devices */
1321 if (is_btrfs && total_devs > 1) {
1322 ret = btrfs_scan_lblkid();
1327 /* iterate over the list of currently mountes filesystems */
1328 if ((f = setmntent ("/proc/self/mounts", "r")) == NULL)
1331 while ((mnt = getmntent (f)) != NULL) {
1333 if(strcmp(mnt->mnt_type, "btrfs") != 0)
1336 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
1338 /* ignore entries in the mount table that are not
1339 associated with a file*/
1340 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
1341 goto out_mntloop_err;
1345 ret = is_same_loop_file(file, mnt->mnt_fsname);
1349 goto out_mntloop_err;
1354 /* Did we find an entry in mnt table? */
1355 if (mnt && size && where) {
1356 strncpy(where, mnt->mnt_dir, size);
1360 *fs_dev_ret = fs_devices_mnt;
1362 ret = (mnt != NULL);
1370 struct pending_dir {
1371 struct list_head list;
1372 char name[PATH_MAX];
1375 int btrfs_register_one_device(const char *fname)
1377 struct btrfs_ioctl_vol_args args;
1382 fd = open("/dev/btrfs-control", O_RDWR);
1384 fprintf(stderr, "failed to open /dev/btrfs-control "
1385 "skipping device registration: %s\n",
1389 strncpy(args.name, fname, BTRFS_PATH_NAME_MAX);
1390 args.name[BTRFS_PATH_NAME_MAX-1] = 0;
1391 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
1394 fprintf(stderr, "ERROR: device scan failed '%s' - %s\n",
1395 fname, strerror(e));
1403 * Register all devices in the fs_uuid list created in the user
1404 * space. Ensure btrfs_scan_lblkid() is called before this func.
1406 int btrfs_register_all_devices(void)
1409 struct btrfs_fs_devices *fs_devices;
1410 struct btrfs_device *device;
1411 struct list_head *all_uuids;
1413 all_uuids = btrfs_scanned_uuids();
1415 list_for_each_entry(fs_devices, all_uuids, list) {
1416 list_for_each_entry(device, &fs_devices->devices, dev_list) {
1417 if (strlen(device->name) != 0) {
1418 err = btrfs_register_one_device(device->name);
1429 int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
1432 struct btrfs_super_block *disk_super;
1436 buf = malloc(BTRFS_SUPER_INFO_SIZE);
1441 ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
1442 if (ret != BTRFS_SUPER_INFO_SIZE)
1446 disk_super = (struct btrfs_super_block *)buf;
1447 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC)
1450 if (!memcmp(disk_super->fsid, root->fs_info->super_copy->fsid,
1459 static const char* unit_suffix_binary[] =
1460 { "B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB"};
1461 static const char* unit_suffix_decimal[] =
1462 { "B", "kB", "MB", "GB", "TB", "PB", "EB"};
1464 int pretty_size_snprintf(u64 size, char *str, size_t str_size, unsigned unit_mode)
1470 const char** suffix = NULL;
1476 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_RAW) {
1477 snprintf(str, str_size, "%llu", size);
1481 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_BINARY) {
1484 suffix = unit_suffix_binary;
1485 } else if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_DECIMAL) {
1488 suffix = unit_suffix_decimal;
1493 fprintf(stderr, "INTERNAL ERROR: unknown unit base, mode %d\n",
1501 switch (unit_mode & UNITS_MODE_MASK) {
1502 case UNITS_TBYTES: base *= mult; num_divs++;
1503 case UNITS_GBYTES: base *= mult; num_divs++;
1504 case UNITS_MBYTES: base *= mult; num_divs++;
1505 case UNITS_KBYTES: num_divs++;
1512 while (size >= mult) {
1519 if (num_divs >= ARRAY_SIZE(unit_suffix_binary)) {
1521 printf("INTERNAL ERROR: unsupported unit suffix, index %d\n",
1526 fraction = (float)last_size / base;
1528 return snprintf(str, str_size, "%.2f%s", fraction, suffix[num_divs]);
1532 * __strncpy__null - strncpy with null termination
1533 * @dest: the target array
1534 * @src: the source string
1535 * @n: maximum bytes to copy (size of *dest)
1537 * Like strncpy, but ensures destination is null-terminated.
1539 * Copies the string pointed to by src, including the terminating null
1540 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
1541 * of n bytes. Then ensure that dest is null-terminated.
1543 char *__strncpy__null(char *dest, const char *src, size_t n)
1545 strncpy(dest, src, n);
1552 * Checks to make sure that the label matches our requirements.
1554 0 if everything is safe and usable
1555 -1 if the label is too long
1557 static int check_label(const char *input)
1559 int len = strlen(input);
1561 if (len > BTRFS_LABEL_SIZE - 1) {
1562 fprintf(stderr, "ERROR: Label %s is too long (max %d)\n",
1563 input, BTRFS_LABEL_SIZE - 1);
1570 static int set_label_unmounted(const char *dev, const char *label)
1572 struct btrfs_trans_handle *trans;
1573 struct btrfs_root *root;
1576 ret = check_mounted(dev);
1578 fprintf(stderr, "FATAL: error checking %s mount status\n", dev);
1582 fprintf(stderr, "ERROR: dev %s is mounted, use mount point\n",
1587 /* Open the super_block at the default location
1588 * and as read-write.
1590 root = open_ctree(dev, 0, OPEN_CTREE_WRITES);
1591 if (!root) /* errors are printed by open_ctree() */
1594 trans = btrfs_start_transaction(root, 1);
1595 snprintf(root->fs_info->super_copy->label, BTRFS_LABEL_SIZE, "%s",
1597 btrfs_commit_transaction(trans, root);
1599 /* Now we close it since we are done. */
1604 static int set_label_mounted(const char *mount_path, const char *label)
1608 fd = open(mount_path, O_RDONLY | O_NOATIME);
1610 fprintf(stderr, "ERROR: unable to access '%s'\n", mount_path);
1614 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
1615 fprintf(stderr, "ERROR: unable to set label %s\n",
1625 static int get_label_unmounted(const char *dev, char *label)
1627 struct btrfs_root *root;
1630 ret = check_mounted(dev);
1632 fprintf(stderr, "FATAL: error checking %s mount status\n", dev);
1636 fprintf(stderr, "ERROR: dev %s is mounted, use mount point\n",
1641 /* Open the super_block at the default location
1644 root = open_ctree(dev, 0, 0);
1648 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
1650 /* Now we close it since we are done. */
1656 * If a partition is mounted, try to get the filesystem label via its
1657 * mounted path rather than device. Return the corresponding error
1658 * the user specified the device path.
1660 int get_label_mounted(const char *mount_path, char *labelp)
1662 char label[BTRFS_LABEL_SIZE];
1665 fd = open(mount_path, O_RDONLY | O_NOATIME);
1667 fprintf(stderr, "ERROR: unable to access '%s'\n", mount_path);
1671 memset(label, '\0', sizeof(label));
1672 if (ioctl(fd, BTRFS_IOC_GET_FSLABEL, label) < 0) {
1673 fprintf(stderr, "ERROR: unable get label %s\n", strerror(errno));
1678 strncpy(labelp, label, sizeof(label));
1683 int get_label(const char *btrfs_dev, char *label)
1687 ret = is_existing_blk_or_reg_file(btrfs_dev);
1689 ret = get_label_mounted(btrfs_dev, label);
1691 ret = get_label_unmounted(btrfs_dev, label);
1696 int set_label(const char *btrfs_dev, const char *label)
1700 if (check_label(label))
1703 ret = is_existing_blk_or_reg_file(btrfs_dev);
1705 ret = set_label_mounted(btrfs_dev, label);
1707 ret = set_label_unmounted(btrfs_dev, label);
1713 * Unsafe subvolume check.
1715 * This only checks ino == BTRFS_FIRST_FREE_OBJECTID, even it is not in a
1716 * btrfs mount point.
1717 * Must use together with other reliable method like btrfs ioctl.
1719 static int __is_subvol(const char *path)
1724 ret = lstat(path, &st);
1728 return st.st_ino == BTRFS_FIRST_FREE_OBJECTID;
1732 * A not-so-good version fls64. No fascinating optimization since
1733 * no one except parse_size use it
1735 static int fls64(u64 x)
1739 for (i = 0; i <64; i++)
1740 if (x << i & (1ULL << 63))
1745 u64 parse_size(char *s)
1753 fprintf(stderr, "ERROR: Size value is empty\n");
1758 "ERROR: Size value '%s' is less equal than 0\n", s);
1761 ret = strtoull(s, &endptr, 10);
1763 fprintf(stderr, "ERROR: Size value '%s' is invalid\n", s);
1766 if (endptr[0] && endptr[1]) {
1767 fprintf(stderr, "ERROR: Illegal suffix contains character '%c' in wrong position\n",
1772 * strtoll returns LLONG_MAX when overflow, if this happens,
1773 * need to call strtoull to get the real size
1775 if (errno == ERANGE && ret == ULLONG_MAX) {
1777 "ERROR: Size value '%s' is too large for u64\n", s);
1781 c = tolower(endptr[0]);
1804 fprintf(stderr, "ERROR: Unknown size descriptor '%c'\n",
1809 /* Check whether ret * mult overflow */
1810 if (fls64(ret) + fls64(mult) - 1 > 64) {
1812 "ERROR: Size value '%s' is too large for u64\n", s);
1819 u64 parse_qgroupid(const char *p)
1821 char *s = strchr(p, '/');
1822 const char *ptr_src_end = p + strlen(p);
1823 char *ptr_parse_end = NULL;
1832 /* Numeric format like '0/257' is the primary case */
1834 id = strtoull(p, &ptr_parse_end, 10);
1835 if (ptr_parse_end != ptr_src_end)
1839 level = strtoull(p, &ptr_parse_end, 10);
1840 if (ptr_parse_end != s)
1843 id = strtoull(s + 1, &ptr_parse_end, 10);
1844 if (ptr_parse_end != ptr_src_end)
1847 return (level << BTRFS_QGROUP_LEVEL_SHIFT) | id;
1850 /* Path format like subv at 'my_subvol' is the fallback case */
1851 ret = __is_subvol(p);
1852 if (ret < 0 || !ret)
1854 fd = open(p, O_RDONLY);
1857 ret = lookup_ino_rootid(fd, &id);
1864 fprintf(stderr, "ERROR: invalid qgroupid or subvolume path: %s\n", p);
1868 int open_file_or_dir3(const char *fname, DIR **dirstream, int open_flags)
1874 ret = stat(fname, &st);
1878 if (S_ISDIR(st.st_mode)) {
1879 *dirstream = opendir(fname);
1882 fd = dirfd(*dirstream);
1883 } else if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) {
1884 fd = open(fname, open_flags);
1887 * we set this on purpose, in case the caller output
1888 * strerror(errno) as success
1896 closedir(*dirstream);
1901 int open_file_or_dir(const char *fname, DIR **dirstream)
1903 return open_file_or_dir3(fname, dirstream, O_RDWR);
1906 void close_file_or_dir(int fd, DIR *dirstream)
1909 closedir(dirstream);
1914 int get_device_info(int fd, u64 devid,
1915 struct btrfs_ioctl_dev_info_args *di_args)
1919 di_args->devid = devid;
1920 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
1922 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
1923 return ret ? -errno : 0;
1926 static u64 find_max_device_id(struct btrfs_ioctl_search_args *search_args,
1929 struct btrfs_dev_item *dev_item;
1930 char *buf = search_args->buf;
1932 buf += (nr_items - 1) * (sizeof(struct btrfs_ioctl_search_header)
1933 + sizeof(struct btrfs_dev_item));
1934 buf += sizeof(struct btrfs_ioctl_search_header);
1936 dev_item = (struct btrfs_dev_item *)buf;
1938 return btrfs_stack_device_id(dev_item);
1941 static int search_chunk_tree_for_fs_info(int fd,
1942 struct btrfs_ioctl_fs_info_args *fi_args)
1946 u64 start_devid = 1;
1947 struct btrfs_ioctl_search_args search_args;
1948 struct btrfs_ioctl_search_key *search_key = &search_args.key;
1950 fi_args->num_devices = 0;
1952 max_items = BTRFS_SEARCH_ARGS_BUFSIZE
1953 / (sizeof(struct btrfs_ioctl_search_header)
1954 + sizeof(struct btrfs_dev_item));
1956 search_key->tree_id = BTRFS_CHUNK_TREE_OBJECTID;
1957 search_key->min_objectid = BTRFS_DEV_ITEMS_OBJECTID;
1958 search_key->max_objectid = BTRFS_DEV_ITEMS_OBJECTID;
1959 search_key->min_type = BTRFS_DEV_ITEM_KEY;
1960 search_key->max_type = BTRFS_DEV_ITEM_KEY;
1961 search_key->min_transid = 0;
1962 search_key->max_transid = (u64)-1;
1963 search_key->nr_items = max_items;
1964 search_key->max_offset = (u64)-1;
1967 search_key->min_offset = start_devid;
1969 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH, &search_args);
1973 fi_args->num_devices += (u64)search_key->nr_items;
1975 if (search_key->nr_items == max_items) {
1976 start_devid = find_max_device_id(&search_args,
1977 search_key->nr_items) + 1;
1981 /* get the lastest max_id to stay consistent with the num_devices */
1982 if (search_key->nr_items == 0)
1984 * last tree_search returns an empty buf, use the devid of
1985 * the last dev_item of the previous tree_search
1987 fi_args->max_id = start_devid - 1;
1989 fi_args->max_id = find_max_device_id(&search_args,
1990 search_key->nr_items);
1996 * For a given path, fill in the ioctl fs_ and info_ args.
1997 * If the path is a btrfs mountpoint, fill info for all devices.
1998 * If the path is a btrfs device, fill in only that device.
2000 * The path provided must be either on a mounted btrfs fs,
2001 * or be a mounted btrfs device.
2003 * Returns 0 on success, or a negative errno.
2005 int get_fs_info(char *path, struct btrfs_ioctl_fs_info_args *fi_args,
2006 struct btrfs_ioctl_dev_info_args **di_ret)
2013 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2014 struct btrfs_ioctl_dev_info_args *di_args;
2015 struct btrfs_ioctl_dev_info_args tmp;
2016 char mp[BTRFS_PATH_NAME_MAX + 1];
2017 DIR *dirstream = NULL;
2019 memset(fi_args, 0, sizeof(*fi_args));
2021 if (is_block_device(path)) {
2022 struct btrfs_super_block *disk_super;
2023 char buf[BTRFS_SUPER_INFO_SIZE];
2026 /* Ensure it's mounted, then set path to the mountpoint */
2027 fd = open(path, O_RDONLY);
2030 fprintf(stderr, "Couldn't open %s: %s\n",
2031 path, strerror(errno));
2034 ret = check_mounted_where(fd, path, mp, sizeof(mp),
2043 /* Only fill in this one device */
2044 fi_args->num_devices = 1;
2046 disk_super = (struct btrfs_super_block *)buf;
2047 ret = btrfs_read_dev_super(fd, disk_super,
2048 BTRFS_SUPER_INFO_OFFSET, 0);
2053 devid = btrfs_stack_device_id(&disk_super->dev_item);
2055 fi_args->max_id = devid;
2058 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
2062 /* at this point path must not be for a block device */
2063 fd = open_file_or_dir(path, &dirstream);
2069 /* fill in fi_args if not just a single device */
2070 if (fi_args->num_devices != 1) {
2071 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
2078 * The fs_args->num_devices does not include seed devices
2080 ret = search_chunk_tree_for_fs_info(fd, fi_args);
2085 * search_chunk_tree_for_fs_info() will lacks the devid 0
2086 * so manual probe for it here.
2088 ret = get_device_info(fd, 0, &tmp);
2090 fi_args->num_devices++;
2098 if (!fi_args->num_devices)
2101 di_args = *di_ret = malloc((fi_args->num_devices) * sizeof(*di_args));
2108 memcpy(di_args, &tmp, sizeof(tmp));
2109 for (; i <= fi_args->max_id; ++i) {
2110 ret = get_device_info(fd, i, &di_args[ndevs]);
2119 * only when the only dev we wanted to find is not there then
2120 * let any error be returned
2122 if (fi_args->num_devices != 1) {
2128 close_file_or_dir(fd, dirstream);
2132 #define isoctal(c) (((c) & ~7) == '0')
2134 static inline void translate(char *f, char *t)
2136 while (*f != '\0') {
2138 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
2139 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
2149 * Checks if the swap device.
2150 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
2152 static int is_swap_device(const char *file)
2163 if (stat(file, &st_buf) < 0)
2165 if (S_ISBLK(st_buf.st_mode))
2166 dev = st_buf.st_rdev;
2167 else if (S_ISREG(st_buf.st_mode)) {
2168 dev = st_buf.st_dev;
2169 ino = st_buf.st_ino;
2173 if ((f = fopen("/proc/swaps", "r")) == NULL)
2176 /* skip the first line */
2177 if (fgets(tmp, sizeof(tmp), f) == NULL)
2180 while (fgets(tmp, sizeof(tmp), f) != NULL) {
2181 if ((cp = strchr(tmp, ' ')) != NULL)
2183 if ((cp = strchr(tmp, '\t')) != NULL)
2185 translate(tmp, buf);
2186 if (stat(buf, &st_buf) != 0)
2188 if (S_ISBLK(st_buf.st_mode)) {
2189 if (dev == st_buf.st_rdev) {
2193 } else if (S_ISREG(st_buf.st_mode)) {
2194 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {
2208 * Check for existing filesystem or partition table on device.
2210 * 1 for existing fs or partition
2211 * 0 for nothing found
2212 * -1 for internal error
2219 blkid_probe pr = NULL;
2223 if (!device || !*device)
2226 ret = -1; /* will reset on success of all setup calls */
2228 pr = blkid_new_probe_from_filename(device);
2232 size = blkid_probe_get_size(pr);
2236 /* nothing to overwrite on a 0-length device */
2242 ret = blkid_probe_enable_partitions(pr, 1);
2246 ret = blkid_do_fullprobe(pr);
2251 * Blkid returns 1 for nothing found and 0 when it finds a signature,
2252 * but we want the exact opposite, so reverse the return value here.
2254 * In addition print some useful diagnostics about what actually is
2262 if (!blkid_probe_lookup_value(pr, "TYPE", &type, NULL)) {
2264 "%s appears to contain an existing "
2265 "filesystem (%s).\n", device, type);
2266 } else if (!blkid_probe_lookup_value(pr, "PTTYPE", &type, NULL)) {
2268 "%s appears to contain a partition "
2269 "table (%s).\n", device, type);
2272 "%s appears to contain something weird "
2273 "according to blkid\n", device);
2279 blkid_free_probe(pr);
2282 "probe of %s failed, cannot detect "
2283 "existing filesystem.\n", device);
2287 static int group_profile_devs_min(u64 flag)
2289 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
2290 case 0: /* single */
2291 case BTRFS_BLOCK_GROUP_DUP:
2293 case BTRFS_BLOCK_GROUP_RAID0:
2294 case BTRFS_BLOCK_GROUP_RAID1:
2295 case BTRFS_BLOCK_GROUP_RAID5:
2297 case BTRFS_BLOCK_GROUP_RAID6:
2299 case BTRFS_BLOCK_GROUP_RAID10:
2306 int test_num_disk_vs_raid(u64 metadata_profile, u64 data_profile,
2307 u64 dev_cnt, int mixed, char *estr)
2315 allowed |= BTRFS_BLOCK_GROUP_RAID10;
2317 allowed |= BTRFS_BLOCK_GROUP_RAID6;
2319 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
2320 BTRFS_BLOCK_GROUP_RAID5;
2323 allowed |= BTRFS_BLOCK_GROUP_DUP;
2327 ((metadata_profile | data_profile) & BTRFS_BLOCK_GROUP_DUP)) {
2329 "DUP is not allowed when FS has multiple devices\n");
2332 if (metadata_profile & ~allowed) {
2334 "unable to create FS with metadata profile %s "
2335 "(have %llu devices but %d devices are required)\n",
2336 btrfs_group_profile_str(metadata_profile), dev_cnt,
2337 group_profile_devs_min(metadata_profile));
2340 if (data_profile & ~allowed) {
2342 "unable to create FS with data profile %s "
2343 "(have %llu devices but %d devices are required)\n",
2344 btrfs_group_profile_str(data_profile), dev_cnt,
2345 group_profile_devs_min(data_profile));
2349 if (!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP)) {
2351 "dup for data is allowed only in mixed mode");
2357 int group_profile_max_safe_loss(u64 flags)
2359 switch (flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
2360 case 0: /* single */
2361 case BTRFS_BLOCK_GROUP_DUP:
2362 case BTRFS_BLOCK_GROUP_RAID0:
2364 case BTRFS_BLOCK_GROUP_RAID1:
2365 case BTRFS_BLOCK_GROUP_RAID5:
2366 case BTRFS_BLOCK_GROUP_RAID10:
2368 case BTRFS_BLOCK_GROUP_RAID6:
2375 /* Check if disk is suitable for btrfs
2377 * 1: something is wrong, estr provides the error
2380 int test_dev_for_mkfs(char *file, int force_overwrite, char *estr)
2386 ret = is_swap_device(file);
2388 snprintf(estr, sz, "error checking %s status: %s\n", file,
2393 snprintf(estr, sz, "%s is a swap device\n", file);
2396 if (!force_overwrite) {
2397 if (check_overwrite(file)) {
2398 snprintf(estr, sz, "Use the -f option to force overwrite.\n");
2402 ret = check_mounted(file);
2404 snprintf(estr, sz, "error checking %s mount status\n",
2409 snprintf(estr, sz, "%s is mounted\n", file);
2412 /* check if the device is busy */
2413 fd = open(file, O_RDWR|O_EXCL);
2415 snprintf(estr, sz, "unable to open %s: %s\n", file,
2419 if (fstat(fd, &st)) {
2420 snprintf(estr, sz, "unable to stat %s: %s\n", file,
2425 if (!S_ISBLK(st.st_mode)) {
2426 fprintf(stderr, "'%s' is not a block device\n", file);
2434 int btrfs_scan_lblkid()
2439 struct btrfs_fs_devices *tmp_devices;
2440 blkid_dev_iterate iter = NULL;
2441 blkid_dev dev = NULL;
2442 blkid_cache cache = NULL;
2443 char path[PATH_MAX];
2445 if (btrfs_scan_done)
2448 if (blkid_get_cache(&cache, 0) < 0) {
2449 printf("ERROR: lblkid cache get failed\n");
2452 blkid_probe_all(cache);
2453 iter = blkid_dev_iterate_begin(cache);
2454 blkid_dev_set_search(iter, "TYPE", "btrfs");
2455 while (blkid_dev_next(iter, &dev) == 0) {
2456 dev = blkid_verify(cache, dev);
2459 /* if we are here its definitely a btrfs disk*/
2460 strncpy_null(path, blkid_dev_devname(dev));
2462 fd = open(path, O_RDONLY);
2464 printf("ERROR: could not open %s\n", path);
2467 ret = btrfs_scan_one_device(fd, path, &tmp_devices,
2468 &num_devices, BTRFS_SUPER_INFO_OFFSET, 0);
2470 printf("ERROR: could not scan %s\n", path);
2477 blkid_dev_iterate_end(iter);
2478 blkid_put_cache(cache);
2480 btrfs_scan_done = 1;
2485 int is_vol_small(char *file)
2492 fd = open(file, O_RDONLY);
2495 if (fstat(fd, &st) < 0) {
2500 size = btrfs_device_size(fd, &st);
2505 if (size < BTRFS_MKFS_SMALL_VOLUME_SIZE) {
2515 * This reads a line from the stdin and only returns non-zero if the
2516 * first whitespace delimited token is a case insensitive match with yes
2519 int ask_user(char *question)
2521 char buf[30] = {0,};
2522 char *saveptr = NULL;
2525 printf("%s [y/N]: ", question);
2527 return fgets(buf, sizeof(buf) - 1, stdin) &&
2528 (answer = strtok_r(buf, " \t\n\r", &saveptr)) &&
2529 (!strcasecmp(answer, "yes") || !strcasecmp(answer, "y"));
2534 * - file or directory return the containing tree root id
2535 * - subvolume return its own tree id
2536 * - BTRFS_EMPTY_SUBVOL_DIR_OBJECTID (directory with ino == 2) the result is
2537 * undefined and function returns -1
2539 int lookup_ino_rootid(int fd, u64 *rootid)
2541 struct btrfs_ioctl_ino_lookup_args args;
2545 memset(&args, 0, sizeof(args));
2547 args.objectid = BTRFS_FIRST_FREE_OBJECTID;
2549 ret = ioctl(fd, BTRFS_IOC_INO_LOOKUP, &args);
2552 fprintf(stderr, "ERROR: Failed to lookup root id - %s\n",
2557 *rootid = args.treeid;
2563 * return 0 if a btrfs mount point is found
2564 * return 1 if a mount point is found but not btrfs
2565 * return <0 if something goes wrong
2567 int find_mount_root(const char *path, char **mount_root)
2575 int longest_matchlen = 0;
2576 char *longest_match = NULL;
2578 fd = open(path, O_RDONLY | O_NOATIME);
2583 mnttab = setmntent("/proc/self/mounts", "r");
2587 while ((ent = getmntent(mnttab))) {
2588 len = strlen(ent->mnt_dir);
2589 if (strncmp(ent->mnt_dir, path, len) == 0) {
2590 /* match found and use the latest match */
2591 if (longest_matchlen <= len) {
2592 free(longest_match);
2593 longest_matchlen = len;
2594 longest_match = strdup(ent->mnt_dir);
2595 not_btrfs = strcmp(ent->mnt_type, "btrfs");
2604 free(longest_match);
2609 *mount_root = realpath(longest_match, NULL);
2613 free(longest_match);
2617 int test_minimum_size(const char *file, u32 nodesize)
2620 struct stat statbuf;
2622 fd = open(file, O_RDONLY);
2625 if (stat(file, &statbuf) < 0) {
2629 if (btrfs_device_size(fd, &statbuf) < btrfs_min_dev_size(nodesize)) {
2638 * test if name is a correct subvolume name
2639 * this function return
2640 * 0-> name is not a correct subvolume name
2641 * 1-> name is a correct subvolume name
2643 int test_issubvolname(const char *name)
2645 return name[0] != '\0' && !strchr(name, '/') &&
2646 strcmp(name, ".") && strcmp(name, "..");
2650 * test if path is a directory
2651 * this function return
2652 * 0-> path exists but it is not a directory
2653 * 1-> path exists and it is a directory
2654 * -1 -> path is unaccessible
2656 int test_isdir(const char *path)
2661 ret = stat(path, &st);
2665 return S_ISDIR(st.st_mode);
2668 void units_set_mode(unsigned *units, unsigned mode)
2670 unsigned base = *units & UNITS_MODE_MASK;
2672 *units = base | mode;
2675 void units_set_base(unsigned *units, unsigned base)
2677 unsigned mode = *units & ~UNITS_MODE_MASK;
2679 *units = base | mode;
2682 int find_next_key(struct btrfs_path *path, struct btrfs_key *key)
2686 for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
2687 if (!path->nodes[level])
2689 if (path->slots[level] + 1 >=
2690 btrfs_header_nritems(path->nodes[level]))
2693 btrfs_item_key_to_cpu(path->nodes[level], key,
2694 path->slots[level] + 1);
2696 btrfs_node_key_to_cpu(path->nodes[level], key,
2697 path->slots[level] + 1);
2703 char* btrfs_group_type_str(u64 flag)
2705 u64 mask = BTRFS_BLOCK_GROUP_TYPE_MASK |
2706 BTRFS_SPACE_INFO_GLOBAL_RSV;
2708 switch (flag & mask) {
2709 case BTRFS_BLOCK_GROUP_DATA:
2711 case BTRFS_BLOCK_GROUP_SYSTEM:
2713 case BTRFS_BLOCK_GROUP_METADATA:
2715 case BTRFS_BLOCK_GROUP_DATA|BTRFS_BLOCK_GROUP_METADATA:
2716 return "Data+Metadata";
2717 case BTRFS_SPACE_INFO_GLOBAL_RSV:
2718 return "GlobalReserve";
2724 char* btrfs_group_profile_str(u64 flag)
2726 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
2729 case BTRFS_BLOCK_GROUP_RAID0:
2731 case BTRFS_BLOCK_GROUP_RAID1:
2733 case BTRFS_BLOCK_GROUP_RAID5:
2735 case BTRFS_BLOCK_GROUP_RAID6:
2737 case BTRFS_BLOCK_GROUP_DUP:
2739 case BTRFS_BLOCK_GROUP_RAID10:
2746 u64 disk_size(char *path)
2750 if (statfs(path, &sfs) < 0)
2753 return sfs.f_bsize * sfs.f_blocks;
2756 u64 get_partition_size(char *dev)
2759 int fd = open(dev, O_RDONLY);
2763 if (ioctl(fd, BLKGETSIZE64, &result) < 0) {
2772 int btrfs_tree_search2_ioctl_supported(int fd)
2774 struct btrfs_ioctl_search_args_v2 *args2;
2775 struct btrfs_ioctl_search_key *sk;
2776 int args2_size = 1024;
2777 char args2_buf[args2_size];
2779 static int v2_supported = -1;
2781 if (v2_supported != -1)
2782 return v2_supported;
2784 args2 = (struct btrfs_ioctl_search_args_v2 *)args2_buf;
2788 * Search for the extent tree item in the root tree.
2790 sk->tree_id = BTRFS_ROOT_TREE_OBJECTID;
2791 sk->min_objectid = BTRFS_EXTENT_TREE_OBJECTID;
2792 sk->max_objectid = BTRFS_EXTENT_TREE_OBJECTID;
2793 sk->min_type = BTRFS_ROOT_ITEM_KEY;
2794 sk->max_type = BTRFS_ROOT_ITEM_KEY;
2796 sk->max_offset = (u64)-1;
2797 sk->min_transid = 0;
2798 sk->max_transid = (u64)-1;
2800 args2->buf_size = args2_size - sizeof(struct btrfs_ioctl_search_args_v2);
2801 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH_V2, args2);
2802 if (ret == -EOPNOTSUPP)
2809 return v2_supported;
2812 int btrfs_check_nodesize(u32 nodesize, u32 sectorsize)
2814 if (nodesize < sectorsize) {
2816 "ERROR: Illegal nodesize %u (smaller than %u)\n",
2817 nodesize, sectorsize);
2819 } else if (nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) {
2821 "ERROR: Illegal nodesize %u (larger than %u)\n",
2822 nodesize, BTRFS_MAX_METADATA_BLOCKSIZE);
2824 } else if (nodesize & (sectorsize - 1)) {
2826 "ERROR: Illegal nodesize %u (not aligned to %u)\n",
2827 nodesize, sectorsize);