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 const char *get_argv0_buf(void)
62 void fixup_argv0(char **argv, const char *token)
64 int len = strlen(argv0_buf);
66 snprintf(argv0_buf + len, sizeof(argv0_buf) - len, " %s", token);
70 void set_argv0(char **argv)
72 strncpy(argv0_buf, argv[0], sizeof(argv0_buf));
73 argv0_buf[sizeof(argv0_buf) - 1] = 0;
76 int check_argc_exact(int nargs, int expected)
79 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
81 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
83 return nargs != expected;
86 int check_argc_min(int nargs, int expected)
88 if (nargs < expected) {
89 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
96 int check_argc_max(int nargs, int expected)
98 if (nargs > expected) {
99 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
108 * Discard the given range in one go
110 static int discard_range(int fd, u64 start, u64 len)
112 u64 range[2] = { start, len };
114 if (ioctl(fd, BLKDISCARD, &range) < 0)
120 * Discard blocks in the given range in 1G chunks, the process is interruptible
122 static int discard_blocks(int fd, u64 start, u64 len)
126 u64 chunk_size = min_t(u64, len, 1*1024*1024*1024);
129 ret = discard_range(fd, start, chunk_size);
139 static u64 reference_root_table[] = {
140 [1] = BTRFS_ROOT_TREE_OBJECTID,
141 [2] = BTRFS_EXTENT_TREE_OBJECTID,
142 [3] = BTRFS_CHUNK_TREE_OBJECTID,
143 [4] = BTRFS_DEV_TREE_OBJECTID,
144 [5] = BTRFS_FS_TREE_OBJECTID,
145 [6] = BTRFS_CSUM_TREE_OBJECTID,
148 int test_uuid_unique(char *fs_uuid)
151 blkid_dev_iterate iter = NULL;
152 blkid_dev dev = NULL;
153 blkid_cache cache = NULL;
155 if (blkid_get_cache(&cache, 0) < 0) {
156 printf("ERROR: lblkid cache get failed\n");
159 blkid_probe_all(cache);
160 iter = blkid_dev_iterate_begin(cache);
161 blkid_dev_set_search(iter, "UUID", fs_uuid);
163 while (blkid_dev_next(iter, &dev) == 0) {
164 dev = blkid_verify(cache, dev);
171 blkid_dev_iterate_end(iter);
172 blkid_put_cache(cache);
178 * @fs_uuid - if NULL, generates a UUID, returns back the new filesystem UUID
180 int make_btrfs(int fd, const char *device, const char *label, char *fs_uuid,
181 u64 blocks[7], u64 num_bytes, u32 nodesize,
182 u32 sectorsize, u32 stripesize, u64 features)
184 struct btrfs_super_block super;
185 struct extent_buffer *buf = NULL;
186 struct btrfs_root_item root_item;
187 struct btrfs_disk_key disk_key;
188 struct btrfs_extent_item *extent_item;
189 struct btrfs_inode_item *inode_item;
190 struct btrfs_chunk *chunk;
191 struct btrfs_dev_item *dev_item;
192 struct btrfs_dev_extent *dev_extent;
193 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
203 int skinny_metadata = !!(features &
204 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA);
206 first_free = BTRFS_SUPER_INFO_OFFSET + sectorsize * 2 - 1;
207 first_free &= ~((u64)sectorsize - 1);
209 memset(&super, 0, sizeof(super));
211 num_bytes = (num_bytes / sectorsize) * sectorsize;
212 if (fs_uuid && *fs_uuid) {
213 if (uuid_parse(fs_uuid, super.fsid) != 0) {
214 fprintf(stderr, "could not parse UUID: %s\n", fs_uuid);
218 if (!test_uuid_unique(fs_uuid)) {
219 fprintf(stderr, "non-unique UUID: %s\n", fs_uuid);
224 uuid_generate(super.fsid);
226 uuid_unparse(super.fsid, fs_uuid);
228 uuid_generate(super.dev_item.uuid);
229 uuid_generate(chunk_tree_uuid);
231 btrfs_set_super_bytenr(&super, blocks[0]);
232 btrfs_set_super_num_devices(&super, 1);
233 btrfs_set_super_magic(&super, BTRFS_MAGIC);
234 btrfs_set_super_generation(&super, 1);
235 btrfs_set_super_root(&super, blocks[1]);
236 btrfs_set_super_chunk_root(&super, blocks[3]);
237 btrfs_set_super_total_bytes(&super, num_bytes);
238 btrfs_set_super_bytes_used(&super, 6 * nodesize);
239 btrfs_set_super_sectorsize(&super, sectorsize);
240 btrfs_set_super_leafsize(&super, nodesize);
241 btrfs_set_super_nodesize(&super, nodesize);
242 btrfs_set_super_stripesize(&super, stripesize);
243 btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
244 btrfs_set_super_chunk_root_generation(&super, 1);
245 btrfs_set_super_cache_generation(&super, -1);
246 btrfs_set_super_incompat_flags(&super, features);
248 strncpy(super.label, label, BTRFS_LABEL_SIZE - 1);
250 buf = malloc(sizeof(*buf) + max(sectorsize, nodesize));
252 /* create the tree of root objects */
253 memset(buf->data, 0, nodesize);
255 btrfs_set_header_bytenr(buf, blocks[1]);
256 btrfs_set_header_nritems(buf, 4);
257 btrfs_set_header_generation(buf, 1);
258 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
259 btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
260 write_extent_buffer(buf, super.fsid, btrfs_header_fsid(),
263 write_extent_buffer(buf, chunk_tree_uuid,
264 btrfs_header_chunk_tree_uuid(buf),
267 /* create the items for the root tree */
268 memset(&root_item, 0, sizeof(root_item));
269 inode_item = &root_item.inode;
270 btrfs_set_stack_inode_generation(inode_item, 1);
271 btrfs_set_stack_inode_size(inode_item, 3);
272 btrfs_set_stack_inode_nlink(inode_item, 1);
273 btrfs_set_stack_inode_nbytes(inode_item, nodesize);
274 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
275 btrfs_set_root_refs(&root_item, 1);
276 btrfs_set_root_used(&root_item, nodesize);
277 btrfs_set_root_generation(&root_item, 1);
279 memset(&disk_key, 0, sizeof(disk_key));
280 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
281 btrfs_set_disk_key_offset(&disk_key, 0);
284 itemoff = __BTRFS_LEAF_DATA_SIZE(nodesize) - sizeof(root_item);
285 btrfs_set_root_bytenr(&root_item, blocks[2]);
286 btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
287 btrfs_set_item_key(buf, &disk_key, nritems);
288 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
289 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
291 write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
292 nritems), sizeof(root_item));
295 itemoff = itemoff - sizeof(root_item);
296 btrfs_set_root_bytenr(&root_item, blocks[4]);
297 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
298 btrfs_set_item_key(buf, &disk_key, nritems);
299 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
300 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
302 write_extent_buffer(buf, &root_item,
303 btrfs_item_ptr_offset(buf, nritems),
307 itemoff = itemoff - sizeof(root_item);
308 btrfs_set_root_bytenr(&root_item, blocks[5]);
309 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
310 btrfs_set_item_key(buf, &disk_key, nritems);
311 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
312 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
314 write_extent_buffer(buf, &root_item,
315 btrfs_item_ptr_offset(buf, nritems),
319 itemoff = itemoff - sizeof(root_item);
320 btrfs_set_root_bytenr(&root_item, blocks[6]);
321 btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_TREE_OBJECTID);
322 btrfs_set_item_key(buf, &disk_key, nritems);
323 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
324 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
326 write_extent_buffer(buf, &root_item,
327 btrfs_item_ptr_offset(buf, nritems),
332 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
333 ret = pwrite(fd, buf->data, nodesize, blocks[1]);
334 if (ret != nodesize) {
335 ret = (ret < 0 ? -errno : -EIO);
339 /* create the items for the extent tree */
340 memset(buf->data + sizeof(struct btrfs_header), 0,
341 nodesize - sizeof(struct btrfs_header));
343 itemoff = __BTRFS_LEAF_DATA_SIZE(nodesize);
344 for (i = 1; i < 7; i++) {
345 item_size = sizeof(struct btrfs_extent_item);
346 if (!skinny_metadata)
347 item_size += sizeof(struct btrfs_tree_block_info);
349 BUG_ON(blocks[i] < first_free);
350 BUG_ON(blocks[i] < blocks[i - 1]);
352 /* create extent item */
353 itemoff -= item_size;
354 btrfs_set_disk_key_objectid(&disk_key, blocks[i]);
355 if (skinny_metadata) {
356 btrfs_set_disk_key_type(&disk_key,
357 BTRFS_METADATA_ITEM_KEY);
358 btrfs_set_disk_key_offset(&disk_key, 0);
360 btrfs_set_disk_key_type(&disk_key,
361 BTRFS_EXTENT_ITEM_KEY);
362 btrfs_set_disk_key_offset(&disk_key, nodesize);
364 btrfs_set_item_key(buf, &disk_key, nritems);
365 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
367 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
369 extent_item = btrfs_item_ptr(buf, nritems,
370 struct btrfs_extent_item);
371 btrfs_set_extent_refs(buf, extent_item, 1);
372 btrfs_set_extent_generation(buf, extent_item, 1);
373 btrfs_set_extent_flags(buf, extent_item,
374 BTRFS_EXTENT_FLAG_TREE_BLOCK);
377 /* create extent ref */
378 ref_root = reference_root_table[i];
379 btrfs_set_disk_key_objectid(&disk_key, blocks[i]);
380 btrfs_set_disk_key_offset(&disk_key, ref_root);
381 btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
382 btrfs_set_item_key(buf, &disk_key, nritems);
383 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
385 btrfs_set_item_size(buf, btrfs_item_nr(nritems), 0);
388 btrfs_set_header_bytenr(buf, blocks[2]);
389 btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
390 btrfs_set_header_nritems(buf, nritems);
391 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
392 ret = pwrite(fd, buf->data, nodesize, blocks[2]);
393 if (ret != nodesize) {
394 ret = (ret < 0 ? -errno : -EIO);
398 /* create the chunk tree */
399 memset(buf->data + sizeof(struct btrfs_header), 0,
400 nodesize - sizeof(struct btrfs_header));
402 item_size = sizeof(*dev_item);
403 itemoff = __BTRFS_LEAF_DATA_SIZE(nodesize) - item_size;
405 /* first device 1 (there is no device 0) */
406 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
407 btrfs_set_disk_key_offset(&disk_key, 1);
408 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
409 btrfs_set_item_key(buf, &disk_key, nritems);
410 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
411 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
413 dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
414 btrfs_set_device_id(buf, dev_item, 1);
415 btrfs_set_device_generation(buf, dev_item, 0);
416 btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
417 btrfs_set_device_bytes_used(buf, dev_item,
418 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
419 btrfs_set_device_io_align(buf, dev_item, sectorsize);
420 btrfs_set_device_io_width(buf, dev_item, sectorsize);
421 btrfs_set_device_sector_size(buf, dev_item, sectorsize);
422 btrfs_set_device_type(buf, dev_item, 0);
424 write_extent_buffer(buf, super.dev_item.uuid,
425 (unsigned long)btrfs_device_uuid(dev_item),
427 write_extent_buffer(buf, super.fsid,
428 (unsigned long)btrfs_device_fsid(dev_item),
430 read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
434 item_size = btrfs_chunk_item_size(1);
435 itemoff = itemoff - item_size;
437 /* then we have chunk 0 */
438 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
439 btrfs_set_disk_key_offset(&disk_key, 0);
440 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
441 btrfs_set_item_key(buf, &disk_key, nritems);
442 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
443 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
445 chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
446 btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
447 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
448 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
449 btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
450 btrfs_set_chunk_io_align(buf, chunk, sectorsize);
451 btrfs_set_chunk_io_width(buf, chunk, sectorsize);
452 btrfs_set_chunk_sector_size(buf, chunk, sectorsize);
453 btrfs_set_chunk_num_stripes(buf, chunk, 1);
454 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
455 btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
458 write_extent_buffer(buf, super.dev_item.uuid,
459 (unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
462 /* copy the key for the chunk to the system array */
463 ptr = super.sys_chunk_array;
464 array_size = sizeof(disk_key);
466 memcpy(ptr, &disk_key, sizeof(disk_key));
467 ptr += sizeof(disk_key);
469 /* copy the chunk to the system array */
470 read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
471 array_size += item_size;
473 btrfs_set_super_sys_array_size(&super, array_size);
475 btrfs_set_header_bytenr(buf, blocks[3]);
476 btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
477 btrfs_set_header_nritems(buf, nritems);
478 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
479 ret = pwrite(fd, buf->data, nodesize, blocks[3]);
480 if (ret != nodesize) {
481 ret = (ret < 0 ? -errno : -EIO);
485 /* create the device tree */
486 memset(buf->data + sizeof(struct btrfs_header), 0,
487 nodesize - sizeof(struct btrfs_header));
489 itemoff = __BTRFS_LEAF_DATA_SIZE(nodesize) -
490 sizeof(struct btrfs_dev_extent);
492 btrfs_set_disk_key_objectid(&disk_key, 1);
493 btrfs_set_disk_key_offset(&disk_key, 0);
494 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
495 btrfs_set_item_key(buf, &disk_key, nritems);
496 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
497 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
498 sizeof(struct btrfs_dev_extent));
499 dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
500 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
501 BTRFS_CHUNK_TREE_OBJECTID);
502 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
503 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
504 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
506 write_extent_buffer(buf, chunk_tree_uuid,
507 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
510 btrfs_set_dev_extent_length(buf, dev_extent,
511 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
514 btrfs_set_header_bytenr(buf, blocks[4]);
515 btrfs_set_header_owner(buf, BTRFS_DEV_TREE_OBJECTID);
516 btrfs_set_header_nritems(buf, nritems);
517 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
518 ret = pwrite(fd, buf->data, nodesize, blocks[4]);
519 if (ret != nodesize) {
520 ret = (ret < 0 ? -errno : -EIO);
524 /* create the FS root */
525 memset(buf->data + sizeof(struct btrfs_header), 0,
526 nodesize - sizeof(struct btrfs_header));
527 btrfs_set_header_bytenr(buf, blocks[5]);
528 btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
529 btrfs_set_header_nritems(buf, 0);
530 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
531 ret = pwrite(fd, buf->data, nodesize, blocks[5]);
532 if (ret != nodesize) {
533 ret = (ret < 0 ? -errno : -EIO);
536 /* finally create the csum root */
537 memset(buf->data + sizeof(struct btrfs_header), 0,
538 nodesize - sizeof(struct btrfs_header));
539 btrfs_set_header_bytenr(buf, blocks[6]);
540 btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
541 btrfs_set_header_nritems(buf, 0);
542 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
543 ret = pwrite(fd, buf->data, nodesize, blocks[6]);
544 if (ret != nodesize) {
545 ret = (ret < 0 ? -errno : -EIO);
549 /* and write out the super block */
550 BUG_ON(sizeof(super) > sectorsize);
551 memset(buf->data, 0, sectorsize);
552 memcpy(buf->data, &super, sizeof(super));
553 buf->len = sectorsize;
554 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
555 ret = pwrite(fd, buf->data, sectorsize, blocks[0]);
556 if (ret != sectorsize) {
557 ret = (ret < 0 ? -errno : -EIO);
568 static const struct btrfs_fs_feature {
572 } mkfs_features[] = {
573 { "mixed-bg", BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS,
574 "mixed data and metadata block groups" },
575 { "extref", BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF,
576 "increased hardlink limit per file to 65536" },
577 { "raid56", BTRFS_FEATURE_INCOMPAT_RAID56,
578 "raid56 extended format" },
579 { "skinny-metadata", BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA,
580 "reduced-size metadata extent refs" },
581 { "no-holes", BTRFS_FEATURE_INCOMPAT_NO_HOLES,
582 "no explicit hole extents for files" },
583 /* Keep this one last */
584 { "list-all", BTRFS_FEATURE_LIST_ALL, NULL }
587 static int parse_one_fs_feature(const char *name, u64 *flags)
592 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
593 if (name[0] == '^' &&
594 !strcmp(mkfs_features[i].name, name + 1)) {
595 *flags &= ~ mkfs_features[i].flag;
597 } else if (!strcmp(mkfs_features[i].name, name)) {
598 *flags |= mkfs_features[i].flag;
606 void btrfs_parse_features_to_string(char *buf, u64 flags)
612 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
613 if (flags & mkfs_features[i].flag) {
616 strcat(buf, mkfs_features[i].name);
621 void btrfs_process_fs_features(u64 flags)
625 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
626 if (flags & mkfs_features[i].flag) {
627 printf("Turning ON incompat feature '%s': %s\n",
628 mkfs_features[i].name,
629 mkfs_features[i].desc);
634 void btrfs_list_all_fs_features(u64 mask_disallowed)
638 fprintf(stderr, "Filesystem features available:\n");
639 for (i = 0; i < ARRAY_SIZE(mkfs_features) - 1; i++) {
640 char *is_default = "";
642 if (mkfs_features[i].flag & mask_disallowed)
644 if (mkfs_features[i].flag & BTRFS_MKFS_DEFAULT_FEATURES)
645 is_default = ", default";
646 fprintf(stderr, "%-20s- %s (0x%llx%s)\n",
647 mkfs_features[i].name,
648 mkfs_features[i].desc,
649 mkfs_features[i].flag,
655 * Return NULL if all features were parsed fine, otherwise return the name of
656 * the first unparsed.
658 char* btrfs_parse_fs_features(char *namelist, u64 *flags)
661 char *save_ptr = NULL; /* Satisfy static checkers */
663 for (this_char = strtok_r(namelist, ",", &save_ptr);
665 this_char = strtok_r(NULL, ",", &save_ptr)) {
666 if (parse_one_fs_feature(this_char, flags))
673 u64 btrfs_device_size(int fd, struct stat *st)
676 if (S_ISREG(st->st_mode)) {
679 if (!S_ISBLK(st->st_mode)) {
682 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
688 static int zero_blocks(int fd, off_t start, size_t len)
690 char *buf = malloc(len);
697 written = pwrite(fd, buf, len, start);
704 #define ZERO_DEV_BYTES (2 * 1024 * 1024)
706 /* don't write outside the device by clamping the region to the device size */
707 static int zero_dev_clamped(int fd, off_t start, ssize_t len, u64 dev_size)
709 off_t end = max(start, start + len);
712 /* and don't overwrite the disk labels on sparc */
713 start = max(start, 1024);
714 end = max(end, 1024);
717 start = min_t(u64, start, dev_size);
718 end = min_t(u64, end, dev_size);
720 return zero_blocks(fd, start, end - start);
723 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
724 struct btrfs_root *root, int fd, char *path,
725 u64 block_count, u32 io_width, u32 io_align,
728 struct btrfs_super_block *disk_super;
729 struct btrfs_super_block *super = root->fs_info->super_copy;
730 struct btrfs_device *device;
731 struct btrfs_dev_item *dev_item;
737 device = kzalloc(sizeof(*device), GFP_NOFS);
740 buf = kmalloc(sectorsize, GFP_NOFS);
745 BUG_ON(sizeof(*disk_super) > sectorsize);
746 memset(buf, 0, sectorsize);
748 disk_super = (struct btrfs_super_block *)buf;
749 dev_item = &disk_super->dev_item;
751 uuid_generate(device->uuid);
754 device->io_width = io_width;
755 device->io_align = io_align;
756 device->sector_size = sectorsize;
758 device->writeable = 1;
759 device->total_bytes = block_count;
760 device->bytes_used = 0;
761 device->total_ios = 0;
762 device->dev_root = root->fs_info->dev_root;
763 device->name = strdup(path);
765 ret = btrfs_add_device(trans, root, device);
768 total_bytes = btrfs_super_total_bytes(super) + block_count;
769 btrfs_set_super_total_bytes(super, total_bytes);
771 num_devs = btrfs_super_num_devices(super) + 1;
772 btrfs_set_super_num_devices(super, num_devs);
774 memcpy(disk_super, super, sizeof(*disk_super));
776 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
777 btrfs_set_stack_device_id(dev_item, device->devid);
778 btrfs_set_stack_device_type(dev_item, device->type);
779 btrfs_set_stack_device_io_align(dev_item, device->io_align);
780 btrfs_set_stack_device_io_width(dev_item, device->io_width);
781 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
782 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
783 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
784 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
786 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
787 BUG_ON(ret != sectorsize);
790 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
791 device->fs_devices = root->fs_info->fs_devices;
795 static void btrfs_wipe_existing_sb(int fd)
797 const char *off = NULL;
802 blkid_probe pr = NULL;
804 pr = blkid_new_probe();
808 if (blkid_probe_set_device(pr, fd, 0, 0))
811 rc = blkid_probe_lookup_value(pr, "SBMAGIC_OFFSET", &off, NULL);
813 rc = blkid_probe_lookup_value(pr, "SBMAGIC", NULL, &len);
815 if (rc || len == 0 || off == NULL)
818 offset = strtoll(off, NULL, 10);
819 if (len > sizeof(buf))
823 rc = pwrite(fd, buf, len, offset);
827 blkid_free_probe(pr);
831 int btrfs_prepare_device(int fd, char *file, int zero_end, u64 *block_count_ret,
832 u64 max_block_count, int *mixed, int discard)
838 ret = fstat(fd, &st);
840 fprintf(stderr, "unable to stat %s\n", file);
844 block_count = btrfs_device_size(fd, &st);
845 if (block_count == 0) {
846 fprintf(stderr, "unable to find %s size\n", file);
850 block_count = min(block_count, max_block_count);
852 if (block_count < BTRFS_MKFS_SMALL_VOLUME_SIZE && !(*mixed))
857 * We intentionally ignore errors from the discard ioctl. It
858 * is not necessary for the mkfs functionality but just an
861 if (discard_range(fd, 0, 0) == 0) {
862 printf("Performing full device TRIM (%s) ...\n",
863 pretty_size(block_count));
864 discard_blocks(fd, 0, block_count);
868 ret = zero_dev_clamped(fd, 0, ZERO_DEV_BYTES, block_count);
869 for (i = 0 ; !ret && i < BTRFS_SUPER_MIRROR_MAX; i++)
870 ret = zero_dev_clamped(fd, btrfs_sb_offset(i),
871 BTRFS_SUPER_INFO_SIZE, block_count);
872 if (!ret && zero_end)
873 ret = zero_dev_clamped(fd, block_count - ZERO_DEV_BYTES,
874 ZERO_DEV_BYTES, block_count);
877 fprintf(stderr, "ERROR: failed to zero device '%s' - %s\n",
878 file, strerror(-ret));
882 btrfs_wipe_existing_sb(fd);
884 *block_count_ret = block_count;
888 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
889 struct btrfs_root *root, u64 objectid)
892 struct btrfs_inode_item inode_item;
893 time_t now = time(NULL);
895 memset(&inode_item, 0, sizeof(inode_item));
896 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
897 btrfs_set_stack_inode_size(&inode_item, 0);
898 btrfs_set_stack_inode_nlink(&inode_item, 1);
899 btrfs_set_stack_inode_nbytes(&inode_item, root->nodesize);
900 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
901 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
902 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
903 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
904 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
905 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
906 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
907 btrfs_set_stack_timespec_sec(&inode_item.otime, 0);
908 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
910 if (root->fs_info->tree_root == root)
911 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
913 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
917 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
921 btrfs_set_root_dirid(&root->root_item, objectid);
928 * checks if a path is a block device node
929 * Returns negative errno on failure, otherwise
930 * returns 1 for blockdev, 0 for not-blockdev
932 int is_block_device(const char *path)
936 if (stat(path, &statbuf) < 0)
939 return S_ISBLK(statbuf.st_mode);
943 * check if given path is a mount point
944 * return 1 if yes. 0 if no. -1 for error
946 int is_mount_point(const char *path)
952 f = setmntent("/proc/self/mounts", "r");
956 while ((mnt = getmntent(f)) != NULL) {
957 if (strcmp(mnt->mnt_dir, path))
966 static int is_reg_file(const char *path)
970 if (stat(path, &statbuf) < 0)
972 return S_ISREG(statbuf.st_mode);
976 * This function checks if the given input parameter is
978 * return <0 : some error in the given input
979 * return BTRFS_ARG_UNKNOWN: unknown input
980 * return BTRFS_ARG_UUID: given input is uuid
981 * return BTRFS_ARG_MNTPOINT: given input is path
982 * return BTRFS_ARG_REG: given input is regular file
984 int check_arg_type(const char *input)
992 if (realpath(input, path)) {
993 if (is_block_device(path) == 1)
994 return BTRFS_ARG_BLKDEV;
996 if (is_mount_point(path) == 1)
997 return BTRFS_ARG_MNTPOINT;
999 if (is_reg_file(path))
1000 return BTRFS_ARG_REG;
1002 return BTRFS_ARG_UNKNOWN;
1005 if (strlen(input) == (BTRFS_UUID_UNPARSED_SIZE - 1) &&
1006 !uuid_parse(input, uuid))
1007 return BTRFS_ARG_UUID;
1009 return BTRFS_ARG_UNKNOWN;
1013 * Find the mount point for a mounted device.
1014 * On success, returns 0 with mountpoint in *mp.
1015 * On failure, returns -errno (not mounted yields -EINVAL)
1016 * Is noisy on failures, expects to be given a mounted device.
1018 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size)
1023 ret = is_block_device(dev);
1026 fprintf(stderr, "%s is not a block device\n", dev);
1029 fprintf(stderr, "Could not check %s: %s\n",
1030 dev, strerror(-ret));
1035 fd = open(dev, O_RDONLY);
1038 fprintf(stderr, "Could not open %s: %s\n", dev, strerror(errno));
1042 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
1045 } else { /* mounted, all good */
1055 * Given a pathname, return a filehandle to:
1056 * the original pathname or,
1057 * if the pathname is a mounted btrfs device, to its mountpoint.
1059 * On error, return -1, errno should be set.
1061 int open_path_or_dev_mnt(const char *path, DIR **dirstream)
1063 char mp[BTRFS_PATH_NAME_MAX + 1];
1066 if (is_block_device(path)) {
1069 ret = get_btrfs_mount(path, mp, sizeof(mp));
1071 /* not a mounted btrfs dev */
1075 fdmnt = open_file_or_dir(mp, dirstream);
1077 fdmnt = open_file_or_dir(path, dirstream);
1083 /* checks if a device is a loop device */
1084 static int is_loop_device (const char* device) {
1085 struct stat statbuf;
1087 if(stat(device, &statbuf) < 0)
1090 return (S_ISBLK(statbuf.st_mode) &&
1091 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
1095 /* Takes a loop device path (e.g. /dev/loop0) and returns
1096 * the associated file (e.g. /images/my_btrfs.img) */
1097 static int resolve_loop_device(const char* loop_dev, char* loop_file,
1104 char real_loop_dev[PATH_MAX];
1106 if (!realpath(loop_dev, real_loop_dev))
1108 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
1109 if (!(f = fopen(p, "r")))
1112 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
1113 ret = fscanf(f, fmt, loop_file);
1122 * Checks whether a and b are identical or device
1123 * files associated with the same block device
1125 static int is_same_blk_file(const char* a, const char* b)
1127 struct stat st_buf_a, st_buf_b;
1128 char real_a[PATH_MAX];
1129 char real_b[PATH_MAX];
1131 if (!realpath(a, real_a))
1132 strncpy_null(real_a, a);
1134 if (!realpath(b, real_b))
1135 strncpy_null(real_b, b);
1137 /* Identical path? */
1138 if (strcmp(real_a, real_b) == 0)
1141 if (stat(a, &st_buf_a) < 0 || stat(b, &st_buf_b) < 0) {
1142 if (errno == ENOENT)
1147 /* Same blockdevice? */
1148 if (S_ISBLK(st_buf_a.st_mode) && S_ISBLK(st_buf_b.st_mode) &&
1149 st_buf_a.st_rdev == st_buf_b.st_rdev) {
1154 if (st_buf_a.st_dev == st_buf_b.st_dev &&
1155 st_buf_a.st_ino == st_buf_b.st_ino) {
1162 /* checks if a and b are identical or device
1163 * files associated with the same block device or
1164 * if one file is a loop device that uses the other
1167 static int is_same_loop_file(const char* a, const char* b)
1169 char res_a[PATH_MAX];
1170 char res_b[PATH_MAX];
1171 const char* final_a = NULL;
1172 const char* final_b = NULL;
1175 /* Resolve a if it is a loop device */
1176 if((ret = is_loop_device(a)) < 0) {
1181 ret = resolve_loop_device(a, res_a, sizeof(res_a));
1192 /* Resolve b if it is a loop device */
1193 if ((ret = is_loop_device(b)) < 0) {
1198 ret = resolve_loop_device(b, res_b, sizeof(res_b));
1209 return is_same_blk_file(final_a, final_b);
1212 /* Checks if a file exists and is a block or regular file*/
1213 static int is_existing_blk_or_reg_file(const char* filename)
1217 if(stat(filename, &st_buf) < 0) {
1224 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
1227 /* Checks if a file is used (directly or indirectly via a loop device)
1228 * by a device in fs_devices
1230 static int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices,
1234 struct list_head *head;
1235 struct list_head *cur;
1236 struct btrfs_device *device;
1238 head = &fs_devices->devices;
1239 list_for_each(cur, head) {
1240 device = list_entry(cur, struct btrfs_device, dev_list);
1242 if((ret = is_same_loop_file(device->name, file)))
1250 * Resolve a pathname to a device mapper node to /dev/mapper/<name>
1251 * Returns NULL on invalid input or malloc failure; Other failures
1252 * will be handled by the caller using the input pathame.
1254 char *canonicalize_dm_name(const char *ptname)
1258 char path[PATH_MAX], name[PATH_MAX], *res = NULL;
1260 if (!ptname || !*ptname)
1263 snprintf(path, sizeof(path), "/sys/block/%s/dm/name", ptname);
1264 if (!(f = fopen(path, "r")))
1267 /* read <name>\n from sysfs */
1268 if (fgets(name, sizeof(name), f) && (sz = strlen(name)) > 1) {
1269 name[sz - 1] = '\0';
1270 snprintf(path, sizeof(path), "/dev/mapper/%s", name);
1272 if (access(path, F_OK) == 0)
1280 * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or
1281 * to a device mapper pathname.
1282 * Returns NULL on invalid input or malloc failure; Other failures
1283 * will be handled by the caller using the input pathame.
1285 char *canonicalize_path(const char *path)
1287 char *canonical, *p;
1289 if (!path || !*path)
1292 canonical = realpath(path, NULL);
1294 return strdup(path);
1295 p = strrchr(canonical, '/');
1296 if (p && strncmp(p, "/dm-", 4) == 0 && isdigit(*(p + 4))) {
1297 char *dm = canonicalize_dm_name(p + 1);
1308 * returns 1 if the device was mounted, < 0 on error or 0 if everything
1309 * is safe to continue.
1311 int check_mounted(const char* file)
1316 fd = open(file, O_RDONLY);
1318 fprintf (stderr, "check_mounted(): Could not open %s\n", file);
1322 ret = check_mounted_where(fd, file, NULL, 0, NULL);
1328 int check_mounted_where(int fd, const char *file, char *where, int size,
1329 struct btrfs_fs_devices **fs_dev_ret)
1334 struct btrfs_fs_devices *fs_devices_mnt = NULL;
1338 /* scan the initial device */
1339 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
1340 &total_devs, BTRFS_SUPER_INFO_OFFSET, 0);
1341 is_btrfs = (ret >= 0);
1343 /* scan other devices */
1344 if (is_btrfs && total_devs > 1) {
1345 ret = btrfs_scan_lblkid();
1350 /* iterate over the list of currently mountes filesystems */
1351 if ((f = setmntent ("/proc/self/mounts", "r")) == NULL)
1354 while ((mnt = getmntent (f)) != NULL) {
1356 if(strcmp(mnt->mnt_type, "btrfs") != 0)
1359 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
1361 /* ignore entries in the mount table that are not
1362 associated with a file*/
1363 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
1364 goto out_mntloop_err;
1368 ret = is_same_loop_file(file, mnt->mnt_fsname);
1372 goto out_mntloop_err;
1377 /* Did we find an entry in mnt table? */
1378 if (mnt && size && where) {
1379 strncpy(where, mnt->mnt_dir, size);
1383 *fs_dev_ret = fs_devices_mnt;
1385 ret = (mnt != NULL);
1393 struct pending_dir {
1394 struct list_head list;
1395 char name[PATH_MAX];
1398 int btrfs_register_one_device(const char *fname)
1400 struct btrfs_ioctl_vol_args args;
1405 fd = open("/dev/btrfs-control", O_RDWR);
1407 fprintf(stderr, "failed to open /dev/btrfs-control "
1408 "skipping device registration: %s\n",
1412 strncpy(args.name, fname, BTRFS_PATH_NAME_MAX);
1413 args.name[BTRFS_PATH_NAME_MAX-1] = 0;
1414 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
1417 fprintf(stderr, "ERROR: device scan failed '%s' - %s\n",
1418 fname, strerror(e));
1426 * Register all devices in the fs_uuid list created in the user
1427 * space. Ensure btrfs_scan_lblkid() is called before this func.
1429 int btrfs_register_all_devices(void)
1432 struct btrfs_fs_devices *fs_devices;
1433 struct btrfs_device *device;
1434 struct list_head *all_uuids;
1436 all_uuids = btrfs_scanned_uuids();
1438 list_for_each_entry(fs_devices, all_uuids, list) {
1439 list_for_each_entry(device, &fs_devices->devices, dev_list) {
1440 if (strlen(device->name) != 0) {
1441 err = btrfs_register_one_device(device->name);
1452 int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
1455 struct btrfs_super_block *disk_super;
1459 buf = malloc(BTRFS_SUPER_INFO_SIZE);
1464 ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
1465 if (ret != BTRFS_SUPER_INFO_SIZE)
1469 disk_super = (struct btrfs_super_block *)buf;
1470 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC)
1473 if (!memcmp(disk_super->fsid, root->fs_info->super_copy->fsid,
1482 static const char* unit_suffix_binary[] =
1483 { "B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB"};
1484 static const char* unit_suffix_decimal[] =
1485 { "B", "kB", "MB", "GB", "TB", "PB", "EB"};
1487 int pretty_size_snprintf(u64 size, char *str, size_t str_size, unsigned unit_mode)
1493 const char** suffix = NULL;
1499 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_RAW) {
1500 snprintf(str, str_size, "%llu", size);
1504 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_BINARY) {
1507 suffix = unit_suffix_binary;
1508 } else if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_DECIMAL) {
1511 suffix = unit_suffix_decimal;
1516 fprintf(stderr, "INTERNAL ERROR: unknown unit base, mode %d\n",
1524 switch (unit_mode & UNITS_MODE_MASK) {
1525 case UNITS_TBYTES: base *= mult; num_divs++;
1526 case UNITS_GBYTES: base *= mult; num_divs++;
1527 case UNITS_MBYTES: base *= mult; num_divs++;
1528 case UNITS_KBYTES: num_divs++;
1535 while (size >= mult) {
1542 if (num_divs >= ARRAY_SIZE(unit_suffix_binary)) {
1544 printf("INTERNAL ERROR: unsupported unit suffix, index %d\n",
1549 fraction = (float)last_size / base;
1551 return snprintf(str, str_size, "%.2f%s", fraction, suffix[num_divs]);
1555 * __strncpy__null - strncpy with null termination
1556 * @dest: the target array
1557 * @src: the source string
1558 * @n: maximum bytes to copy (size of *dest)
1560 * Like strncpy, but ensures destination is null-terminated.
1562 * Copies the string pointed to by src, including the terminating null
1563 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
1564 * of n bytes. Then ensure that dest is null-terminated.
1566 char *__strncpy__null(char *dest, const char *src, size_t n)
1568 strncpy(dest, src, n);
1575 * Checks to make sure that the label matches our requirements.
1577 0 if everything is safe and usable
1578 -1 if the label is too long
1580 static int check_label(const char *input)
1582 int len = strlen(input);
1584 if (len > BTRFS_LABEL_SIZE - 1) {
1585 fprintf(stderr, "ERROR: Label %s is too long (max %d)\n",
1586 input, BTRFS_LABEL_SIZE - 1);
1593 static int set_label_unmounted(const char *dev, const char *label)
1595 struct btrfs_trans_handle *trans;
1596 struct btrfs_root *root;
1599 ret = check_mounted(dev);
1601 fprintf(stderr, "FATAL: error checking %s mount status\n", dev);
1605 fprintf(stderr, "ERROR: dev %s is mounted, use mount point\n",
1610 /* Open the super_block at the default location
1611 * and as read-write.
1613 root = open_ctree(dev, 0, OPEN_CTREE_WRITES);
1614 if (!root) /* errors are printed by open_ctree() */
1617 trans = btrfs_start_transaction(root, 1);
1618 snprintf(root->fs_info->super_copy->label, BTRFS_LABEL_SIZE, "%s",
1620 btrfs_commit_transaction(trans, root);
1622 /* Now we close it since we are done. */
1627 static int set_label_mounted(const char *mount_path, const char *label)
1631 fd = open(mount_path, O_RDONLY | O_NOATIME);
1633 fprintf(stderr, "ERROR: unable to access '%s'\n", mount_path);
1637 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
1638 fprintf(stderr, "ERROR: unable to set label %s\n",
1648 static int get_label_unmounted(const char *dev, char *label)
1650 struct btrfs_root *root;
1653 ret = check_mounted(dev);
1655 fprintf(stderr, "FATAL: error checking %s mount status\n", dev);
1659 fprintf(stderr, "ERROR: dev %s is mounted, use mount point\n",
1664 /* Open the super_block at the default location
1667 root = open_ctree(dev, 0, 0);
1671 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
1673 /* Now we close it since we are done. */
1679 * If a partition is mounted, try to get the filesystem label via its
1680 * mounted path rather than device. Return the corresponding error
1681 * the user specified the device path.
1683 int get_label_mounted(const char *mount_path, char *labelp)
1685 char label[BTRFS_LABEL_SIZE];
1688 fd = open(mount_path, O_RDONLY | O_NOATIME);
1690 fprintf(stderr, "ERROR: unable to access '%s'\n", mount_path);
1694 memset(label, '\0', sizeof(label));
1695 if (ioctl(fd, BTRFS_IOC_GET_FSLABEL, label) < 0) {
1696 fprintf(stderr, "ERROR: unable get label %s\n", strerror(errno));
1701 strncpy(labelp, label, sizeof(label));
1706 int get_label(const char *btrfs_dev, char *label)
1710 ret = is_existing_blk_or_reg_file(btrfs_dev);
1712 ret = get_label_mounted(btrfs_dev, label);
1714 ret = get_label_unmounted(btrfs_dev, label);
1719 int set_label(const char *btrfs_dev, const char *label)
1723 if (check_label(label))
1726 ret = is_existing_blk_or_reg_file(btrfs_dev);
1728 ret = set_label_mounted(btrfs_dev, label);
1730 ret = set_label_unmounted(btrfs_dev, label);
1736 * Unsafe subvolume check.
1738 * This only checks ino == BTRFS_FIRST_FREE_OBJECTID, even it is not in a
1739 * btrfs mount point.
1740 * Must use together with other reliable method like btrfs ioctl.
1742 static int __is_subvol(const char *path)
1747 ret = lstat(path, &st);
1751 return st.st_ino == BTRFS_FIRST_FREE_OBJECTID;
1755 * A not-so-good version fls64. No fascinating optimization since
1756 * no one except parse_size use it
1758 static int fls64(u64 x)
1762 for (i = 0; i <64; i++)
1763 if (x << i & (1ULL << 63))
1768 u64 parse_size(char *s)
1776 fprintf(stderr, "ERROR: Size value is empty\n");
1781 "ERROR: Size value '%s' is less equal than 0\n", s);
1784 ret = strtoull(s, &endptr, 10);
1786 fprintf(stderr, "ERROR: Size value '%s' is invalid\n", s);
1789 if (endptr[0] && endptr[1]) {
1790 fprintf(stderr, "ERROR: Illegal suffix contains character '%c' in wrong position\n",
1795 * strtoll returns LLONG_MAX when overflow, if this happens,
1796 * need to call strtoull to get the real size
1798 if (errno == ERANGE && ret == ULLONG_MAX) {
1800 "ERROR: Size value '%s' is too large for u64\n", s);
1804 c = tolower(endptr[0]);
1827 fprintf(stderr, "ERROR: Unknown size descriptor '%c'\n",
1832 /* Check whether ret * mult overflow */
1833 if (fls64(ret) + fls64(mult) - 1 > 64) {
1835 "ERROR: Size value '%s' is too large for u64\n", s);
1842 u64 parse_qgroupid(const char *p)
1844 char *s = strchr(p, '/');
1845 const char *ptr_src_end = p + strlen(p);
1846 char *ptr_parse_end = NULL;
1855 /* Numeric format like '0/257' is the primary case */
1857 id = strtoull(p, &ptr_parse_end, 10);
1858 if (ptr_parse_end != ptr_src_end)
1862 level = strtoull(p, &ptr_parse_end, 10);
1863 if (ptr_parse_end != s)
1866 id = strtoull(s + 1, &ptr_parse_end, 10);
1867 if (ptr_parse_end != ptr_src_end)
1870 return (level << BTRFS_QGROUP_LEVEL_SHIFT) | id;
1873 /* Path format like subv at 'my_subvol' is the fallback case */
1874 ret = __is_subvol(p);
1875 if (ret < 0 || !ret)
1877 fd = open(p, O_RDONLY);
1880 ret = lookup_ino_rootid(fd, &id);
1887 fprintf(stderr, "ERROR: invalid qgroupid or subvolume path: %s\n", p);
1891 int open_file_or_dir3(const char *fname, DIR **dirstream, int open_flags)
1897 ret = stat(fname, &st);
1901 if (S_ISDIR(st.st_mode)) {
1902 *dirstream = opendir(fname);
1905 fd = dirfd(*dirstream);
1906 } else if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) {
1907 fd = open(fname, open_flags);
1910 * we set this on purpose, in case the caller output
1911 * strerror(errno) as success
1919 closedir(*dirstream);
1926 int open_file_or_dir(const char *fname, DIR **dirstream)
1928 return open_file_or_dir3(fname, dirstream, O_RDWR);
1931 void close_file_or_dir(int fd, DIR *dirstream)
1934 closedir(dirstream);
1939 int get_device_info(int fd, u64 devid,
1940 struct btrfs_ioctl_dev_info_args *di_args)
1944 di_args->devid = devid;
1945 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
1947 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
1948 return ret ? -errno : 0;
1951 static u64 find_max_device_id(struct btrfs_ioctl_search_args *search_args,
1954 struct btrfs_dev_item *dev_item;
1955 char *buf = search_args->buf;
1957 buf += (nr_items - 1) * (sizeof(struct btrfs_ioctl_search_header)
1958 + sizeof(struct btrfs_dev_item));
1959 buf += sizeof(struct btrfs_ioctl_search_header);
1961 dev_item = (struct btrfs_dev_item *)buf;
1963 return btrfs_stack_device_id(dev_item);
1966 static int search_chunk_tree_for_fs_info(int fd,
1967 struct btrfs_ioctl_fs_info_args *fi_args)
1971 u64 start_devid = 1;
1972 struct btrfs_ioctl_search_args search_args;
1973 struct btrfs_ioctl_search_key *search_key = &search_args.key;
1975 fi_args->num_devices = 0;
1977 max_items = BTRFS_SEARCH_ARGS_BUFSIZE
1978 / (sizeof(struct btrfs_ioctl_search_header)
1979 + sizeof(struct btrfs_dev_item));
1981 search_key->tree_id = BTRFS_CHUNK_TREE_OBJECTID;
1982 search_key->min_objectid = BTRFS_DEV_ITEMS_OBJECTID;
1983 search_key->max_objectid = BTRFS_DEV_ITEMS_OBJECTID;
1984 search_key->min_type = BTRFS_DEV_ITEM_KEY;
1985 search_key->max_type = BTRFS_DEV_ITEM_KEY;
1986 search_key->min_transid = 0;
1987 search_key->max_transid = (u64)-1;
1988 search_key->nr_items = max_items;
1989 search_key->max_offset = (u64)-1;
1992 search_key->min_offset = start_devid;
1994 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH, &search_args);
1998 fi_args->num_devices += (u64)search_key->nr_items;
2000 if (search_key->nr_items == max_items) {
2001 start_devid = find_max_device_id(&search_args,
2002 search_key->nr_items) + 1;
2006 /* get the lastest max_id to stay consistent with the num_devices */
2007 if (search_key->nr_items == 0)
2009 * last tree_search returns an empty buf, use the devid of
2010 * the last dev_item of the previous tree_search
2012 fi_args->max_id = start_devid - 1;
2014 fi_args->max_id = find_max_device_id(&search_args,
2015 search_key->nr_items);
2021 * For a given path, fill in the ioctl fs_ and info_ args.
2022 * If the path is a btrfs mountpoint, fill info for all devices.
2023 * If the path is a btrfs device, fill in only that device.
2025 * The path provided must be either on a mounted btrfs fs,
2026 * or be a mounted btrfs device.
2028 * Returns 0 on success, or a negative errno.
2030 int get_fs_info(char *path, struct btrfs_ioctl_fs_info_args *fi_args,
2031 struct btrfs_ioctl_dev_info_args **di_ret)
2038 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2039 struct btrfs_ioctl_dev_info_args *di_args;
2040 struct btrfs_ioctl_dev_info_args tmp;
2041 char mp[BTRFS_PATH_NAME_MAX + 1];
2042 DIR *dirstream = NULL;
2044 memset(fi_args, 0, sizeof(*fi_args));
2046 if (is_block_device(path)) {
2047 struct btrfs_super_block *disk_super;
2048 char buf[BTRFS_SUPER_INFO_SIZE];
2051 /* Ensure it's mounted, then set path to the mountpoint */
2052 fd = open(path, O_RDONLY);
2055 fprintf(stderr, "Couldn't open %s: %s\n",
2056 path, strerror(errno));
2059 ret = check_mounted_where(fd, path, mp, sizeof(mp),
2068 /* Only fill in this one device */
2069 fi_args->num_devices = 1;
2071 disk_super = (struct btrfs_super_block *)buf;
2072 ret = btrfs_read_dev_super(fd, disk_super,
2073 BTRFS_SUPER_INFO_OFFSET, 0);
2078 devid = btrfs_stack_device_id(&disk_super->dev_item);
2080 fi_args->max_id = devid;
2083 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
2087 /* at this point path must not be for a block device */
2088 fd = open_file_or_dir(path, &dirstream);
2094 /* fill in fi_args if not just a single device */
2095 if (fi_args->num_devices != 1) {
2096 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
2103 * The fs_args->num_devices does not include seed devices
2105 ret = search_chunk_tree_for_fs_info(fd, fi_args);
2110 * search_chunk_tree_for_fs_info() will lacks the devid 0
2111 * so manual probe for it here.
2113 ret = get_device_info(fd, 0, &tmp);
2115 fi_args->num_devices++;
2123 if (!fi_args->num_devices)
2126 di_args = *di_ret = malloc((fi_args->num_devices) * sizeof(*di_args));
2133 memcpy(di_args, &tmp, sizeof(tmp));
2134 for (; i <= fi_args->max_id; ++i) {
2135 ret = get_device_info(fd, i, &di_args[ndevs]);
2144 * only when the only dev we wanted to find is not there then
2145 * let any error be returned
2147 if (fi_args->num_devices != 1) {
2153 close_file_or_dir(fd, dirstream);
2157 #define isoctal(c) (((c) & ~7) == '0')
2159 static inline void translate(char *f, char *t)
2161 while (*f != '\0') {
2163 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
2164 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
2174 * Checks if the swap device.
2175 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
2177 static int is_swap_device(const char *file)
2188 if (stat(file, &st_buf) < 0)
2190 if (S_ISBLK(st_buf.st_mode))
2191 dev = st_buf.st_rdev;
2192 else if (S_ISREG(st_buf.st_mode)) {
2193 dev = st_buf.st_dev;
2194 ino = st_buf.st_ino;
2198 if ((f = fopen("/proc/swaps", "r")) == NULL)
2201 /* skip the first line */
2202 if (fgets(tmp, sizeof(tmp), f) == NULL)
2205 while (fgets(tmp, sizeof(tmp), f) != NULL) {
2206 if ((cp = strchr(tmp, ' ')) != NULL)
2208 if ((cp = strchr(tmp, '\t')) != NULL)
2210 translate(tmp, buf);
2211 if (stat(buf, &st_buf) != 0)
2213 if (S_ISBLK(st_buf.st_mode)) {
2214 if (dev == st_buf.st_rdev) {
2218 } else if (S_ISREG(st_buf.st_mode)) {
2219 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {
2233 * Check for existing filesystem or partition table on device.
2235 * 1 for existing fs or partition
2236 * 0 for nothing found
2237 * -1 for internal error
2244 blkid_probe pr = NULL;
2248 if (!device || !*device)
2251 ret = -1; /* will reset on success of all setup calls */
2253 pr = blkid_new_probe_from_filename(device);
2257 size = blkid_probe_get_size(pr);
2261 /* nothing to overwrite on a 0-length device */
2267 ret = blkid_probe_enable_partitions(pr, 1);
2271 ret = blkid_do_fullprobe(pr);
2276 * Blkid returns 1 for nothing found and 0 when it finds a signature,
2277 * but we want the exact opposite, so reverse the return value here.
2279 * In addition print some useful diagnostics about what actually is
2287 if (!blkid_probe_lookup_value(pr, "TYPE", &type, NULL)) {
2289 "%s appears to contain an existing "
2290 "filesystem (%s).\n", device, type);
2291 } else if (!blkid_probe_lookup_value(pr, "PTTYPE", &type, NULL)) {
2293 "%s appears to contain a partition "
2294 "table (%s).\n", device, type);
2297 "%s appears to contain something weird "
2298 "according to blkid\n", device);
2304 blkid_free_probe(pr);
2307 "probe of %s failed, cannot detect "
2308 "existing filesystem.\n", device);
2312 static int group_profile_devs_min(u64 flag)
2314 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
2315 case 0: /* single */
2316 case BTRFS_BLOCK_GROUP_DUP:
2318 case BTRFS_BLOCK_GROUP_RAID0:
2319 case BTRFS_BLOCK_GROUP_RAID1:
2320 case BTRFS_BLOCK_GROUP_RAID5:
2322 case BTRFS_BLOCK_GROUP_RAID6:
2324 case BTRFS_BLOCK_GROUP_RAID10:
2331 int test_num_disk_vs_raid(u64 metadata_profile, u64 data_profile,
2332 u64 dev_cnt, int mixed, char *estr)
2340 allowed |= BTRFS_BLOCK_GROUP_RAID10;
2342 allowed |= BTRFS_BLOCK_GROUP_RAID6;
2344 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
2345 BTRFS_BLOCK_GROUP_RAID5;
2348 allowed |= BTRFS_BLOCK_GROUP_DUP;
2352 ((metadata_profile | data_profile) & BTRFS_BLOCK_GROUP_DUP)) {
2354 "DUP is not allowed when FS has multiple devices\n");
2357 if (metadata_profile & ~allowed) {
2359 "unable to create FS with metadata profile %s "
2360 "(have %llu devices but %d devices are required)\n",
2361 btrfs_group_profile_str(metadata_profile), dev_cnt,
2362 group_profile_devs_min(metadata_profile));
2365 if (data_profile & ~allowed) {
2367 "unable to create FS with data profile %s "
2368 "(have %llu devices but %d devices are required)\n",
2369 btrfs_group_profile_str(data_profile), dev_cnt,
2370 group_profile_devs_min(data_profile));
2374 if (!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP)) {
2376 "dup for data is allowed only in mixed mode");
2382 int group_profile_max_safe_loss(u64 flags)
2384 switch (flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
2385 case 0: /* single */
2386 case BTRFS_BLOCK_GROUP_DUP:
2387 case BTRFS_BLOCK_GROUP_RAID0:
2389 case BTRFS_BLOCK_GROUP_RAID1:
2390 case BTRFS_BLOCK_GROUP_RAID5:
2391 case BTRFS_BLOCK_GROUP_RAID10:
2393 case BTRFS_BLOCK_GROUP_RAID6:
2401 * Check if a device is suitable for btrfs
2403 * 1: something is wrong, an error is printed
2406 int test_dev_for_mkfs(char *file, int force_overwrite)
2411 ret = is_swap_device(file);
2413 fprintf(stderr, "ERROR: checking status of %s: %s\n", file,
2418 fprintf(stderr, "ERROR: %s is a swap device\n", file);
2421 if (!force_overwrite) {
2422 if (check_overwrite(file)) {
2423 fprintf(stderr, "Use the -f option to force overwrite.\n");
2427 ret = check_mounted(file);
2429 fprintf(stderr, "ERROR: checking mount status of %s: %s\n",
2430 file, strerror(-ret));
2434 fprintf(stderr, "ERROR: %s is mounted\n", file);
2437 /* check if the device is busy */
2438 fd = open(file, O_RDWR|O_EXCL);
2440 fprintf(stderr, "ERROR: unable to open %s: %s\n", file,
2444 if (fstat(fd, &st)) {
2445 fprintf(stderr, "ERROR: unable to stat %s: %s\n", file,
2450 if (!S_ISBLK(st.st_mode)) {
2451 fprintf(stderr, "ERROR: %s is not a block device\n", file);
2459 int btrfs_scan_lblkid()
2464 struct btrfs_fs_devices *tmp_devices;
2465 blkid_dev_iterate iter = NULL;
2466 blkid_dev dev = NULL;
2467 blkid_cache cache = NULL;
2468 char path[PATH_MAX];
2470 if (btrfs_scan_done)
2473 if (blkid_get_cache(&cache, 0) < 0) {
2474 printf("ERROR: lblkid cache get failed\n");
2477 blkid_probe_all(cache);
2478 iter = blkid_dev_iterate_begin(cache);
2479 blkid_dev_set_search(iter, "TYPE", "btrfs");
2480 while (blkid_dev_next(iter, &dev) == 0) {
2481 dev = blkid_verify(cache, dev);
2484 /* if we are here its definitely a btrfs disk*/
2485 strncpy_null(path, blkid_dev_devname(dev));
2487 fd = open(path, O_RDONLY);
2489 printf("ERROR: could not open %s\n", path);
2492 ret = btrfs_scan_one_device(fd, path, &tmp_devices,
2493 &num_devices, BTRFS_SUPER_INFO_OFFSET, 0);
2495 printf("ERROR: could not scan %s\n", path);
2502 blkid_dev_iterate_end(iter);
2503 blkid_put_cache(cache);
2505 btrfs_scan_done = 1;
2510 int is_vol_small(char *file)
2517 fd = open(file, O_RDONLY);
2520 if (fstat(fd, &st) < 0) {
2525 size = btrfs_device_size(fd, &st);
2530 if (size < BTRFS_MKFS_SMALL_VOLUME_SIZE) {
2540 * This reads a line from the stdin and only returns non-zero if the
2541 * first whitespace delimited token is a case insensitive match with yes
2544 int ask_user(char *question)
2546 char buf[30] = {0,};
2547 char *saveptr = NULL;
2550 printf("%s [y/N]: ", question);
2552 return fgets(buf, sizeof(buf) - 1, stdin) &&
2553 (answer = strtok_r(buf, " \t\n\r", &saveptr)) &&
2554 (!strcasecmp(answer, "yes") || !strcasecmp(answer, "y"));
2559 * - file or directory return the containing tree root id
2560 * - subvolume return its own tree id
2561 * - BTRFS_EMPTY_SUBVOL_DIR_OBJECTID (directory with ino == 2) the result is
2562 * undefined and function returns -1
2564 int lookup_ino_rootid(int fd, u64 *rootid)
2566 struct btrfs_ioctl_ino_lookup_args args;
2570 memset(&args, 0, sizeof(args));
2572 args.objectid = BTRFS_FIRST_FREE_OBJECTID;
2574 ret = ioctl(fd, BTRFS_IOC_INO_LOOKUP, &args);
2577 fprintf(stderr, "ERROR: Failed to lookup root id - %s\n",
2582 *rootid = args.treeid;
2588 * return 0 if a btrfs mount point is found
2589 * return 1 if a mount point is found but not btrfs
2590 * return <0 if something goes wrong
2592 int find_mount_root(const char *path, char **mount_root)
2600 int longest_matchlen = 0;
2601 char *longest_match = NULL;
2603 fd = open(path, O_RDONLY | O_NOATIME);
2608 mnttab = setmntent("/proc/self/mounts", "r");
2612 while ((ent = getmntent(mnttab))) {
2613 len = strlen(ent->mnt_dir);
2614 if (strncmp(ent->mnt_dir, path, len) == 0) {
2615 /* match found and use the latest match */
2616 if (longest_matchlen <= len) {
2617 free(longest_match);
2618 longest_matchlen = len;
2619 longest_match = strdup(ent->mnt_dir);
2620 not_btrfs = strcmp(ent->mnt_type, "btrfs");
2629 free(longest_match);
2634 *mount_root = realpath(longest_match, NULL);
2638 free(longest_match);
2642 int test_minimum_size(const char *file, u32 nodesize)
2645 struct stat statbuf;
2647 fd = open(file, O_RDONLY);
2650 if (stat(file, &statbuf) < 0) {
2654 if (btrfs_device_size(fd, &statbuf) < btrfs_min_dev_size(nodesize)) {
2663 * test if name is a correct subvolume name
2664 * this function return
2665 * 0-> name is not a correct subvolume name
2666 * 1-> name is a correct subvolume name
2668 int test_issubvolname(const char *name)
2670 return name[0] != '\0' && !strchr(name, '/') &&
2671 strcmp(name, ".") && strcmp(name, "..");
2675 * test if path is a directory
2676 * this function return
2677 * 0-> path exists but it is not a directory
2678 * 1-> path exists and it is a directory
2679 * -1 -> path is unaccessible
2681 int test_isdir(const char *path)
2686 ret = stat(path, &st);
2690 return S_ISDIR(st.st_mode);
2693 void units_set_mode(unsigned *units, unsigned mode)
2695 unsigned base = *units & UNITS_MODE_MASK;
2697 *units = base | mode;
2700 void units_set_base(unsigned *units, unsigned base)
2702 unsigned mode = *units & ~UNITS_MODE_MASK;
2704 *units = base | mode;
2707 int find_next_key(struct btrfs_path *path, struct btrfs_key *key)
2711 for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
2712 if (!path->nodes[level])
2714 if (path->slots[level] + 1 >=
2715 btrfs_header_nritems(path->nodes[level]))
2718 btrfs_item_key_to_cpu(path->nodes[level], key,
2719 path->slots[level] + 1);
2721 btrfs_node_key_to_cpu(path->nodes[level], key,
2722 path->slots[level] + 1);
2728 char* btrfs_group_type_str(u64 flag)
2730 u64 mask = BTRFS_BLOCK_GROUP_TYPE_MASK |
2731 BTRFS_SPACE_INFO_GLOBAL_RSV;
2733 switch (flag & mask) {
2734 case BTRFS_BLOCK_GROUP_DATA:
2736 case BTRFS_BLOCK_GROUP_SYSTEM:
2738 case BTRFS_BLOCK_GROUP_METADATA:
2740 case BTRFS_BLOCK_GROUP_DATA|BTRFS_BLOCK_GROUP_METADATA:
2741 return "Data+Metadata";
2742 case BTRFS_SPACE_INFO_GLOBAL_RSV:
2743 return "GlobalReserve";
2749 char* btrfs_group_profile_str(u64 flag)
2751 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
2754 case BTRFS_BLOCK_GROUP_RAID0:
2756 case BTRFS_BLOCK_GROUP_RAID1:
2758 case BTRFS_BLOCK_GROUP_RAID5:
2760 case BTRFS_BLOCK_GROUP_RAID6:
2762 case BTRFS_BLOCK_GROUP_DUP:
2764 case BTRFS_BLOCK_GROUP_RAID10:
2771 u64 disk_size(char *path)
2775 if (statfs(path, &sfs) < 0)
2778 return sfs.f_bsize * sfs.f_blocks;
2781 u64 get_partition_size(char *dev)
2784 int fd = open(dev, O_RDONLY);
2788 if (ioctl(fd, BLKGETSIZE64, &result) < 0) {
2797 int btrfs_tree_search2_ioctl_supported(int fd)
2799 struct btrfs_ioctl_search_args_v2 *args2;
2800 struct btrfs_ioctl_search_key *sk;
2801 int args2_size = 1024;
2802 char args2_buf[args2_size];
2804 static int v2_supported = -1;
2806 if (v2_supported != -1)
2807 return v2_supported;
2809 args2 = (struct btrfs_ioctl_search_args_v2 *)args2_buf;
2813 * Search for the extent tree item in the root tree.
2815 sk->tree_id = BTRFS_ROOT_TREE_OBJECTID;
2816 sk->min_objectid = BTRFS_EXTENT_TREE_OBJECTID;
2817 sk->max_objectid = BTRFS_EXTENT_TREE_OBJECTID;
2818 sk->min_type = BTRFS_ROOT_ITEM_KEY;
2819 sk->max_type = BTRFS_ROOT_ITEM_KEY;
2821 sk->max_offset = (u64)-1;
2822 sk->min_transid = 0;
2823 sk->max_transid = (u64)-1;
2825 args2->buf_size = args2_size - sizeof(struct btrfs_ioctl_search_args_v2);
2826 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH_V2, args2);
2827 if (ret == -EOPNOTSUPP)
2834 return v2_supported;
2837 int btrfs_check_nodesize(u32 nodesize, u32 sectorsize)
2839 if (nodesize < sectorsize) {
2841 "ERROR: Illegal nodesize %u (smaller than %u)\n",
2842 nodesize, sectorsize);
2844 } else if (nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) {
2846 "ERROR: Illegal nodesize %u (larger than %u)\n",
2847 nodesize, BTRFS_MAX_METADATA_BLOCKSIZE);
2849 } else if (nodesize & (sectorsize - 1)) {
2851 "ERROR: Illegal nodesize %u (not aligned to %u)\n",
2852 nodesize, sectorsize);