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>
38 #include <sys/statfs.h>
39 #include <linux/magic.h>
42 #include "kerncompat.h"
43 #include "radix-tree.h"
46 #include "transaction.h"
54 #define BLKDISCARD _IO(0x12,119)
57 static int btrfs_scan_done = 0;
59 static char argv0_buf[ARGV0_BUF_SIZE] = "btrfs";
61 static int rand_seed_initlized = 0;
62 static unsigned short rand_seed[3];
64 const char *get_argv0_buf(void)
69 void fixup_argv0(char **argv, const char *token)
71 int len = strlen(argv0_buf);
73 snprintf(argv0_buf + len, sizeof(argv0_buf) - len, " %s", token);
77 void set_argv0(char **argv)
79 strncpy(argv0_buf, argv[0], sizeof(argv0_buf));
80 argv0_buf[sizeof(argv0_buf) - 1] = 0;
83 int check_argc_exact(int nargs, int expected)
86 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
88 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
90 return nargs != expected;
93 int check_argc_min(int nargs, int expected)
95 if (nargs < expected) {
96 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
103 int check_argc_max(int nargs, int expected)
105 if (nargs > expected) {
106 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
115 * Discard the given range in one go
117 static int discard_range(int fd, u64 start, u64 len)
119 u64 range[2] = { start, len };
121 if (ioctl(fd, BLKDISCARD, &range) < 0)
127 * Discard blocks in the given range in 1G chunks, the process is interruptible
129 static int discard_blocks(int fd, u64 start, u64 len)
133 u64 chunk_size = min_t(u64, len, 1*1024*1024*1024);
136 ret = discard_range(fd, start, chunk_size);
146 static u64 reference_root_table[] = {
147 [1] = BTRFS_ROOT_TREE_OBJECTID,
148 [2] = BTRFS_EXTENT_TREE_OBJECTID,
149 [3] = BTRFS_CHUNK_TREE_OBJECTID,
150 [4] = BTRFS_DEV_TREE_OBJECTID,
151 [5] = BTRFS_FS_TREE_OBJECTID,
152 [6] = BTRFS_CSUM_TREE_OBJECTID,
155 int test_uuid_unique(char *fs_uuid)
158 blkid_dev_iterate iter = NULL;
159 blkid_dev dev = NULL;
160 blkid_cache cache = NULL;
162 if (blkid_get_cache(&cache, NULL) < 0) {
163 printf("ERROR: lblkid cache get failed\n");
166 blkid_probe_all(cache);
167 iter = blkid_dev_iterate_begin(cache);
168 blkid_dev_set_search(iter, "UUID", fs_uuid);
170 while (blkid_dev_next(iter, &dev) == 0) {
171 dev = blkid_verify(cache, dev);
178 blkid_dev_iterate_end(iter);
179 blkid_put_cache(cache);
185 * Reserve space from free_tree.
186 * The algorithm is very simple, find the first cache_extent with enough space
187 * and allocate from its beginning.
189 static int reserve_free_space(struct cache_tree *free_tree, u64 len,
192 struct cache_extent *cache;
196 cache = first_cache_extent(free_tree);
198 if (cache->size > len) {
200 *ret_start = cache->start;
203 if (cache->size == 0) {
204 remove_cache_extent(free_tree, cache);
211 cache = next_cache_extent(cache);
219 * Improved version of make_btrfs().
222 * 1) Do chunk allocation to avoid used data
223 * And after this function, extent type matches chunk type
224 * 2) Better structured code
225 * No super long hand written codes to initialized all tree blocks
226 * Split into small blocks and reuse codes.
227 * TODO: Reuse tree operation facilities by introducing new flags
229 static int make_convert_btrfs(int fd, struct btrfs_mkfs_config *cfg,
230 struct btrfs_convert_context *cctx)
232 struct cache_tree *free = &cctx->free;
233 struct cache_tree *used = &cctx->used;
235 u64 meta_chunk_start;
238 /* Shouldn't happen */
239 BUG_ON(cache_tree_empty(used));
242 * reserve space for temporary superblock first
243 * Here we allocate a little larger space, to keep later
244 * free space will be STRIPE_LEN aligned
246 ret = reserve_free_space(free, BTRFS_STRIPE_LEN,
252 * Then reserve system chunk space
253 * TODO: Change system group size depending on cctx->total_bytes.
254 * If using current 4M, it can only handle less than one TB for
255 * worst case and then run out of sys space.
257 ret = reserve_free_space(free, BTRFS_MKFS_SYSTEM_GROUP_SIZE,
261 ret = reserve_free_space(free, BTRFS_CONVERT_META_GROUP_SIZE,
271 * @fs_uuid - if NULL, generates a UUID, returns back the new filesystem UUID
273 int make_btrfs(int fd, struct btrfs_mkfs_config *cfg,
274 struct btrfs_convert_context *cctx)
276 struct btrfs_super_block super;
277 struct extent_buffer *buf;
278 struct btrfs_root_item root_item;
279 struct btrfs_disk_key disk_key;
280 struct btrfs_extent_item *extent_item;
281 struct btrfs_inode_item *inode_item;
282 struct btrfs_chunk *chunk;
283 struct btrfs_dev_item *dev_item;
284 struct btrfs_dev_extent *dev_extent;
285 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
295 int skinny_metadata = !!(cfg->features &
296 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA);
300 return make_convert_btrfs(fd, cfg, cctx);
301 buf = malloc(sizeof(*buf) + max(cfg->sectorsize, cfg->nodesize));
305 first_free = BTRFS_SUPER_INFO_OFFSET + cfg->sectorsize * 2 - 1;
306 first_free &= ~((u64)cfg->sectorsize - 1);
308 memset(&super, 0, sizeof(super));
310 num_bytes = (cfg->num_bytes / cfg->sectorsize) * cfg->sectorsize;
311 if (cfg->fs_uuid && *cfg->fs_uuid) {
312 if (uuid_parse(cfg->fs_uuid, super.fsid) != 0) {
313 error("cannot not parse UUID: %s", cfg->fs_uuid);
317 if (!test_uuid_unique(cfg->fs_uuid)) {
318 error("non-unique UUID: %s", cfg->fs_uuid);
323 uuid_generate(super.fsid);
325 uuid_unparse(super.fsid, cfg->fs_uuid);
327 uuid_generate(super.dev_item.uuid);
328 uuid_generate(chunk_tree_uuid);
330 btrfs_set_super_bytenr(&super, cfg->blocks[0]);
331 btrfs_set_super_num_devices(&super, 1);
332 btrfs_set_super_magic(&super, BTRFS_MAGIC);
333 btrfs_set_super_generation(&super, 1);
334 btrfs_set_super_root(&super, cfg->blocks[1]);
335 btrfs_set_super_chunk_root(&super, cfg->blocks[3]);
336 btrfs_set_super_total_bytes(&super, num_bytes);
337 btrfs_set_super_bytes_used(&super, 6 * cfg->nodesize);
338 btrfs_set_super_sectorsize(&super, cfg->sectorsize);
339 btrfs_set_super_leafsize(&super, cfg->nodesize);
340 btrfs_set_super_nodesize(&super, cfg->nodesize);
341 btrfs_set_super_stripesize(&super, cfg->stripesize);
342 btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
343 btrfs_set_super_chunk_root_generation(&super, 1);
344 btrfs_set_super_cache_generation(&super, -1);
345 btrfs_set_super_incompat_flags(&super, cfg->features);
347 __strncpy_null(super.label, cfg->label, BTRFS_LABEL_SIZE - 1);
349 /* create the tree of root objects */
350 memset(buf->data, 0, cfg->nodesize);
351 buf->len = cfg->nodesize;
352 btrfs_set_header_bytenr(buf, cfg->blocks[1]);
353 btrfs_set_header_nritems(buf, 4);
354 btrfs_set_header_generation(buf, 1);
355 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
356 btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
357 write_extent_buffer(buf, super.fsid, btrfs_header_fsid(),
360 write_extent_buffer(buf, chunk_tree_uuid,
361 btrfs_header_chunk_tree_uuid(buf),
364 /* create the items for the root tree */
365 memset(&root_item, 0, sizeof(root_item));
366 inode_item = &root_item.inode;
367 btrfs_set_stack_inode_generation(inode_item, 1);
368 btrfs_set_stack_inode_size(inode_item, 3);
369 btrfs_set_stack_inode_nlink(inode_item, 1);
370 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
371 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
372 btrfs_set_root_refs(&root_item, 1);
373 btrfs_set_root_used(&root_item, cfg->nodesize);
374 btrfs_set_root_generation(&root_item, 1);
376 memset(&disk_key, 0, sizeof(disk_key));
377 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
378 btrfs_set_disk_key_offset(&disk_key, 0);
381 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - sizeof(root_item);
382 btrfs_set_root_bytenr(&root_item, cfg->blocks[2]);
383 btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
384 btrfs_set_item_key(buf, &disk_key, nritems);
385 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
386 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
388 write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
389 nritems), sizeof(root_item));
392 itemoff = itemoff - sizeof(root_item);
393 btrfs_set_root_bytenr(&root_item, cfg->blocks[4]);
394 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
395 btrfs_set_item_key(buf, &disk_key, nritems);
396 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
397 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
399 write_extent_buffer(buf, &root_item,
400 btrfs_item_ptr_offset(buf, nritems),
404 itemoff = itemoff - sizeof(root_item);
405 btrfs_set_root_bytenr(&root_item, cfg->blocks[5]);
406 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
407 btrfs_set_item_key(buf, &disk_key, nritems);
408 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
409 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
411 write_extent_buffer(buf, &root_item,
412 btrfs_item_ptr_offset(buf, nritems),
416 itemoff = itemoff - sizeof(root_item);
417 btrfs_set_root_bytenr(&root_item, cfg->blocks[6]);
418 btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_TREE_OBJECTID);
419 btrfs_set_item_key(buf, &disk_key, nritems);
420 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
421 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
423 write_extent_buffer(buf, &root_item,
424 btrfs_item_ptr_offset(buf, nritems),
429 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
430 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[1]);
431 if (ret != cfg->nodesize) {
432 ret = (ret < 0 ? -errno : -EIO);
436 /* create the items for the extent tree */
437 memset(buf->data + sizeof(struct btrfs_header), 0,
438 cfg->nodesize - sizeof(struct btrfs_header));
440 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
441 for (i = 1; i < 7; i++) {
442 item_size = sizeof(struct btrfs_extent_item);
443 if (!skinny_metadata)
444 item_size += sizeof(struct btrfs_tree_block_info);
446 BUG_ON(cfg->blocks[i] < first_free);
447 BUG_ON(cfg->blocks[i] < cfg->blocks[i - 1]);
449 /* create extent item */
450 itemoff -= item_size;
451 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
452 if (skinny_metadata) {
453 btrfs_set_disk_key_type(&disk_key,
454 BTRFS_METADATA_ITEM_KEY);
455 btrfs_set_disk_key_offset(&disk_key, 0);
457 btrfs_set_disk_key_type(&disk_key,
458 BTRFS_EXTENT_ITEM_KEY);
459 btrfs_set_disk_key_offset(&disk_key, cfg->nodesize);
461 btrfs_set_item_key(buf, &disk_key, nritems);
462 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
464 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
466 extent_item = btrfs_item_ptr(buf, nritems,
467 struct btrfs_extent_item);
468 btrfs_set_extent_refs(buf, extent_item, 1);
469 btrfs_set_extent_generation(buf, extent_item, 1);
470 btrfs_set_extent_flags(buf, extent_item,
471 BTRFS_EXTENT_FLAG_TREE_BLOCK);
474 /* create extent ref */
475 ref_root = reference_root_table[i];
476 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
477 btrfs_set_disk_key_offset(&disk_key, ref_root);
478 btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
479 btrfs_set_item_key(buf, &disk_key, nritems);
480 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
482 btrfs_set_item_size(buf, btrfs_item_nr(nritems), 0);
485 btrfs_set_header_bytenr(buf, cfg->blocks[2]);
486 btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
487 btrfs_set_header_nritems(buf, nritems);
488 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
489 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[2]);
490 if (ret != cfg->nodesize) {
491 ret = (ret < 0 ? -errno : -EIO);
495 /* create the chunk tree */
496 memset(buf->data + sizeof(struct btrfs_header), 0,
497 cfg->nodesize - sizeof(struct btrfs_header));
499 item_size = sizeof(*dev_item);
500 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - item_size;
502 /* first device 1 (there is no device 0) */
503 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
504 btrfs_set_disk_key_offset(&disk_key, 1);
505 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
506 btrfs_set_item_key(buf, &disk_key, nritems);
507 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
508 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
510 dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
511 btrfs_set_device_id(buf, dev_item, 1);
512 btrfs_set_device_generation(buf, dev_item, 0);
513 btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
514 btrfs_set_device_bytes_used(buf, dev_item,
515 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
516 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
517 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
518 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
519 btrfs_set_device_type(buf, dev_item, 0);
521 write_extent_buffer(buf, super.dev_item.uuid,
522 (unsigned long)btrfs_device_uuid(dev_item),
524 write_extent_buffer(buf, super.fsid,
525 (unsigned long)btrfs_device_fsid(dev_item),
527 read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
531 item_size = btrfs_chunk_item_size(1);
532 itemoff = itemoff - item_size;
534 /* then we have chunk 0 */
535 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
536 btrfs_set_disk_key_offset(&disk_key, 0);
537 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
538 btrfs_set_item_key(buf, &disk_key, nritems);
539 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
540 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
542 chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
543 btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
544 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
545 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
546 btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
547 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
548 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
549 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
550 btrfs_set_chunk_num_stripes(buf, chunk, 1);
551 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
552 btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
555 write_extent_buffer(buf, super.dev_item.uuid,
556 (unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
559 /* copy the key for the chunk to the system array */
560 ptr = super.sys_chunk_array;
561 array_size = sizeof(disk_key);
563 memcpy(ptr, &disk_key, sizeof(disk_key));
564 ptr += sizeof(disk_key);
566 /* copy the chunk to the system array */
567 read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
568 array_size += item_size;
570 btrfs_set_super_sys_array_size(&super, array_size);
572 btrfs_set_header_bytenr(buf, cfg->blocks[3]);
573 btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
574 btrfs_set_header_nritems(buf, nritems);
575 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
576 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[3]);
577 if (ret != cfg->nodesize) {
578 ret = (ret < 0 ? -errno : -EIO);
582 /* create the device tree */
583 memset(buf->data + sizeof(struct btrfs_header), 0,
584 cfg->nodesize - sizeof(struct btrfs_header));
586 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) -
587 sizeof(struct btrfs_dev_extent);
589 btrfs_set_disk_key_objectid(&disk_key, 1);
590 btrfs_set_disk_key_offset(&disk_key, 0);
591 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
592 btrfs_set_item_key(buf, &disk_key, nritems);
593 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
594 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
595 sizeof(struct btrfs_dev_extent));
596 dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
597 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
598 BTRFS_CHUNK_TREE_OBJECTID);
599 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
600 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
601 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
603 write_extent_buffer(buf, chunk_tree_uuid,
604 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
607 btrfs_set_dev_extent_length(buf, dev_extent,
608 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
611 btrfs_set_header_bytenr(buf, cfg->blocks[4]);
612 btrfs_set_header_owner(buf, BTRFS_DEV_TREE_OBJECTID);
613 btrfs_set_header_nritems(buf, nritems);
614 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
615 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[4]);
616 if (ret != cfg->nodesize) {
617 ret = (ret < 0 ? -errno : -EIO);
621 /* create the FS root */
622 memset(buf->data + sizeof(struct btrfs_header), 0,
623 cfg->nodesize - sizeof(struct btrfs_header));
624 btrfs_set_header_bytenr(buf, cfg->blocks[5]);
625 btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
626 btrfs_set_header_nritems(buf, 0);
627 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
628 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[5]);
629 if (ret != cfg->nodesize) {
630 ret = (ret < 0 ? -errno : -EIO);
633 /* finally create the csum root */
634 memset(buf->data + sizeof(struct btrfs_header), 0,
635 cfg->nodesize - sizeof(struct btrfs_header));
636 btrfs_set_header_bytenr(buf, cfg->blocks[6]);
637 btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
638 btrfs_set_header_nritems(buf, 0);
639 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
640 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[6]);
641 if (ret != cfg->nodesize) {
642 ret = (ret < 0 ? -errno : -EIO);
646 /* and write out the super block */
647 BUG_ON(sizeof(super) > cfg->sectorsize);
648 memset(buf->data, 0, BTRFS_SUPER_INFO_SIZE);
649 memcpy(buf->data, &super, sizeof(super));
650 buf->len = BTRFS_SUPER_INFO_SIZE;
651 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
652 ret = pwrite(fd, buf->data, BTRFS_SUPER_INFO_SIZE, cfg->blocks[0]);
653 if (ret != BTRFS_SUPER_INFO_SIZE) {
654 ret = (ret < 0 ? -errno : -EIO);
665 static const struct btrfs_fs_feature {
669 } mkfs_features[] = {
670 { "mixed-bg", BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS,
671 "mixed data and metadata block groups" },
672 { "extref", BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF,
673 "increased hardlink limit per file to 65536" },
674 { "raid56", BTRFS_FEATURE_INCOMPAT_RAID56,
675 "raid56 extended format" },
676 { "skinny-metadata", BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA,
677 "reduced-size metadata extent refs" },
678 { "no-holes", BTRFS_FEATURE_INCOMPAT_NO_HOLES,
679 "no explicit hole extents for files" },
680 /* Keep this one last */
681 { "list-all", BTRFS_FEATURE_LIST_ALL, NULL }
684 static int parse_one_fs_feature(const char *name, u64 *flags)
689 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
690 if (name[0] == '^' &&
691 !strcmp(mkfs_features[i].name, name + 1)) {
692 *flags &= ~ mkfs_features[i].flag;
694 } else if (!strcmp(mkfs_features[i].name, name)) {
695 *flags |= mkfs_features[i].flag;
703 void btrfs_parse_features_to_string(char *buf, u64 flags)
709 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
710 if (flags & mkfs_features[i].flag) {
713 strcat(buf, mkfs_features[i].name);
718 void btrfs_process_fs_features(u64 flags)
722 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
723 if (flags & mkfs_features[i].flag) {
724 printf("Turning ON incompat feature '%s': %s\n",
725 mkfs_features[i].name,
726 mkfs_features[i].desc);
731 void btrfs_list_all_fs_features(u64 mask_disallowed)
735 fprintf(stderr, "Filesystem features available:\n");
736 for (i = 0; i < ARRAY_SIZE(mkfs_features) - 1; i++) {
737 char *is_default = "";
739 if (mkfs_features[i].flag & mask_disallowed)
741 if (mkfs_features[i].flag & BTRFS_MKFS_DEFAULT_FEATURES)
742 is_default = ", default";
743 fprintf(stderr, "%-20s- %s (0x%llx%s)\n",
744 mkfs_features[i].name,
745 mkfs_features[i].desc,
746 mkfs_features[i].flag,
752 * Return NULL if all features were parsed fine, otherwise return the name of
753 * the first unparsed.
755 char* btrfs_parse_fs_features(char *namelist, u64 *flags)
758 char *save_ptr = NULL; /* Satisfy static checkers */
760 for (this_char = strtok_r(namelist, ",", &save_ptr);
762 this_char = strtok_r(NULL, ",", &save_ptr)) {
763 if (parse_one_fs_feature(this_char, flags))
770 u64 btrfs_device_size(int fd, struct stat *st)
773 if (S_ISREG(st->st_mode)) {
776 if (!S_ISBLK(st->st_mode)) {
779 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
785 static int zero_blocks(int fd, off_t start, size_t len)
787 char *buf = malloc(len);
794 written = pwrite(fd, buf, len, start);
801 #define ZERO_DEV_BYTES (2 * 1024 * 1024)
803 /* don't write outside the device by clamping the region to the device size */
804 static int zero_dev_clamped(int fd, off_t start, ssize_t len, u64 dev_size)
806 off_t end = max(start, start + len);
809 /* and don't overwrite the disk labels on sparc */
810 start = max(start, 1024);
811 end = max(end, 1024);
814 start = min_t(u64, start, dev_size);
815 end = min_t(u64, end, dev_size);
817 return zero_blocks(fd, start, end - start);
820 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
821 struct btrfs_root *root, int fd, char *path,
822 u64 device_total_bytes, u32 io_width, u32 io_align,
825 struct btrfs_super_block *disk_super;
826 struct btrfs_super_block *super = root->fs_info->super_copy;
827 struct btrfs_device *device;
828 struct btrfs_dev_item *dev_item;
834 device_total_bytes = (device_total_bytes / sectorsize) * sectorsize;
836 device = kzalloc(sizeof(*device), GFP_NOFS);
839 buf = kzalloc(sectorsize, GFP_NOFS);
842 BUG_ON(sizeof(*disk_super) > sectorsize);
844 disk_super = (struct btrfs_super_block *)buf;
845 dev_item = &disk_super->dev_item;
847 uuid_generate(device->uuid);
850 device->io_width = io_width;
851 device->io_align = io_align;
852 device->sector_size = sectorsize;
854 device->writeable = 1;
855 device->total_bytes = device_total_bytes;
856 device->bytes_used = 0;
857 device->total_ios = 0;
858 device->dev_root = root->fs_info->dev_root;
859 device->name = strdup(path);
863 INIT_LIST_HEAD(&device->dev_list);
864 ret = btrfs_add_device(trans, root, device);
867 fs_total_bytes = btrfs_super_total_bytes(super) + device_total_bytes;
868 btrfs_set_super_total_bytes(super, fs_total_bytes);
870 num_devs = btrfs_super_num_devices(super) + 1;
871 btrfs_set_super_num_devices(super, num_devs);
873 memcpy(disk_super, super, sizeof(*disk_super));
875 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
876 btrfs_set_stack_device_id(dev_item, device->devid);
877 btrfs_set_stack_device_type(dev_item, device->type);
878 btrfs_set_stack_device_io_align(dev_item, device->io_align);
879 btrfs_set_stack_device_io_width(dev_item, device->io_width);
880 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
881 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
882 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
883 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
885 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
886 BUG_ON(ret != sectorsize);
889 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
890 device->fs_devices = root->fs_info->fs_devices;
899 static int btrfs_wipe_existing_sb(int fd)
901 const char *off = NULL;
906 blkid_probe pr = NULL;
908 pr = blkid_new_probe();
912 if (blkid_probe_set_device(pr, fd, 0, 0)) {
917 ret = blkid_probe_lookup_value(pr, "SBMAGIC_OFFSET", &off, NULL);
919 ret = blkid_probe_lookup_value(pr, "SBMAGIC", NULL, &len);
921 if (ret || len == 0 || off == NULL) {
923 * If lookup fails, the probe did not find any values, eg. for
924 * a file image or a loop device. Soft error.
930 offset = strtoll(off, NULL, 10);
931 if (len > sizeof(buf))
935 ret = pwrite(fd, buf, len, offset);
937 error("cannot wipe existing superblock: %s", strerror(errno));
939 } else if (ret != len) {
940 error("cannot wipe existing superblock: wrote %d of %zd", ret, len);
946 blkid_free_probe(pr);
950 int btrfs_prepare_device(int fd, const char *file, int zero_end,
951 u64 *block_count_ret, u64 max_block_count, int discard)
957 ret = fstat(fd, &st);
959 error("unable to stat %s: %s", file, strerror(errno));
963 block_count = btrfs_device_size(fd, &st);
964 if (block_count == 0) {
965 error("unable to determine size of %s", file);
969 block_count = min(block_count, max_block_count);
973 * We intentionally ignore errors from the discard ioctl. It
974 * is not necessary for the mkfs functionality but just an
977 if (discard_range(fd, 0, 0) == 0) {
978 printf("Performing full device TRIM (%s) ...\n",
979 pretty_size(block_count));
980 discard_blocks(fd, 0, block_count);
984 ret = zero_dev_clamped(fd, 0, ZERO_DEV_BYTES, block_count);
985 for (i = 0 ; !ret && i < BTRFS_SUPER_MIRROR_MAX; i++)
986 ret = zero_dev_clamped(fd, btrfs_sb_offset(i),
987 BTRFS_SUPER_INFO_SIZE, block_count);
988 if (!ret && zero_end)
989 ret = zero_dev_clamped(fd, block_count - ZERO_DEV_BYTES,
990 ZERO_DEV_BYTES, block_count);
993 error("failed to zero device '%s': %s", file, strerror(-ret));
997 ret = btrfs_wipe_existing_sb(fd);
999 error("cannot wipe superblocks on %s", file);
1003 *block_count_ret = block_count;
1007 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
1008 struct btrfs_root *root, u64 objectid)
1011 struct btrfs_inode_item inode_item;
1012 time_t now = time(NULL);
1014 memset(&inode_item, 0, sizeof(inode_item));
1015 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
1016 btrfs_set_stack_inode_size(&inode_item, 0);
1017 btrfs_set_stack_inode_nlink(&inode_item, 1);
1018 btrfs_set_stack_inode_nbytes(&inode_item, root->nodesize);
1019 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
1020 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
1021 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
1022 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
1023 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
1024 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
1025 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
1026 btrfs_set_stack_timespec_sec(&inode_item.otime, 0);
1027 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
1029 if (root->fs_info->tree_root == root)
1030 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
1032 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
1036 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
1040 btrfs_set_root_dirid(&root->root_item, objectid);
1047 * checks if a path is a block device node
1048 * Returns negative errno on failure, otherwise
1049 * returns 1 for blockdev, 0 for not-blockdev
1051 int is_block_device(const char *path)
1053 struct stat statbuf;
1055 if (stat(path, &statbuf) < 0)
1058 return !!S_ISBLK(statbuf.st_mode);
1062 * check if given path is a mount point
1063 * return 1 if yes. 0 if no. -1 for error
1065 int is_mount_point(const char *path)
1071 f = setmntent("/proc/self/mounts", "r");
1075 while ((mnt = getmntent(f)) != NULL) {
1076 if (strcmp(mnt->mnt_dir, path))
1085 static int is_reg_file(const char *path)
1087 struct stat statbuf;
1089 if (stat(path, &statbuf) < 0)
1091 return S_ISREG(statbuf.st_mode);
1095 * This function checks if the given input parameter is
1097 * return <0 : some error in the given input
1098 * return BTRFS_ARG_UNKNOWN: unknown input
1099 * return BTRFS_ARG_UUID: given input is uuid
1100 * return BTRFS_ARG_MNTPOINT: given input is path
1101 * return BTRFS_ARG_REG: given input is regular file
1102 * return BTRFS_ARG_BLKDEV: given input is block device
1104 int check_arg_type(const char *input)
1107 char path[PATH_MAX];
1112 if (realpath(input, path)) {
1113 if (is_block_device(path) == 1)
1114 return BTRFS_ARG_BLKDEV;
1116 if (is_mount_point(path) == 1)
1117 return BTRFS_ARG_MNTPOINT;
1119 if (is_reg_file(path))
1120 return BTRFS_ARG_REG;
1122 return BTRFS_ARG_UNKNOWN;
1125 if (strlen(input) == (BTRFS_UUID_UNPARSED_SIZE - 1) &&
1126 !uuid_parse(input, uuid))
1127 return BTRFS_ARG_UUID;
1129 return BTRFS_ARG_UNKNOWN;
1133 * Find the mount point for a mounted device.
1134 * On success, returns 0 with mountpoint in *mp.
1135 * On failure, returns -errno (not mounted yields -EINVAL)
1136 * Is noisy on failures, expects to be given a mounted device.
1138 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size)
1143 ret = is_block_device(dev);
1146 error("not a block device: %s", dev);
1149 error("cannot check %s: %s", dev, strerror(-ret));
1154 fd = open(dev, O_RDONLY);
1157 error("cannot open %s: %s", dev, strerror(errno));
1161 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
1164 } else { /* mounted, all good */
1174 * Given a pathname, return a filehandle to:
1175 * the original pathname or,
1176 * if the pathname is a mounted btrfs device, to its mountpoint.
1178 * On error, return -1, errno should be set.
1180 int open_path_or_dev_mnt(const char *path, DIR **dirstream, int verbose)
1185 if (is_block_device(path)) {
1186 ret = get_btrfs_mount(path, mp, sizeof(mp));
1188 /* not a mounted btrfs dev */
1189 error_on(verbose, "'%s' is not a mounted btrfs device",
1194 ret = open_file_or_dir(mp, dirstream);
1195 error_on(verbose && ret < 0, "can't access '%s': %s",
1196 path, strerror(errno));
1198 ret = btrfs_open_dir(path, dirstream, 1);
1205 * Do the following checks before calling open_file_or_dir():
1206 * 1: path is in a btrfs filesystem
1207 * 2: path is a directory
1209 int btrfs_open_dir(const char *path, DIR **dirstream, int verbose)
1215 if (statfs(path, &stfs) != 0) {
1216 error_on(verbose, "cannot access '%s': %s", path,
1221 if (stfs.f_type != BTRFS_SUPER_MAGIC) {
1222 error_on(verbose, "not a btrfs filesystem: %s", path);
1226 if (stat(path, &st) != 0) {
1227 error_on(verbose, "cannot access '%s': %s", path,
1232 if (!S_ISDIR(st.st_mode)) {
1233 error_on(verbose, "not a directory: %s", path);
1237 ret = open_file_or_dir(path, dirstream);
1239 error_on(verbose, "cannot access '%s': %s", path,
1246 /* checks if a device is a loop device */
1247 static int is_loop_device (const char* device) {
1248 struct stat statbuf;
1250 if(stat(device, &statbuf) < 0)
1253 return (S_ISBLK(statbuf.st_mode) &&
1254 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
1258 * Takes a loop device path (e.g. /dev/loop0) and returns
1259 * the associated file (e.g. /images/my_btrfs.img) using
1262 static int resolve_loop_device_with_loopdev(const char* loop_dev, char* loop_file)
1266 struct loop_info64 lo64;
1268 fd = open(loop_dev, O_RDONLY | O_NONBLOCK);
1271 ret = ioctl(fd, LOOP_GET_STATUS64, &lo64);
1277 memcpy(loop_file, lo64.lo_file_name, sizeof(lo64.lo_file_name));
1278 loop_file[sizeof(lo64.lo_file_name)] = 0;
1286 /* Takes a loop device path (e.g. /dev/loop0) and returns
1287 * the associated file (e.g. /images/my_btrfs.img) */
1288 static int resolve_loop_device(const char* loop_dev, char* loop_file,
1295 char real_loop_dev[PATH_MAX];
1297 if (!realpath(loop_dev, real_loop_dev))
1299 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
1300 if (!(f = fopen(p, "r"))) {
1301 if (errno == ENOENT)
1303 * It's possibly a partitioned loop device, which is
1304 * resolvable with loopdev API.
1306 return resolve_loop_device_with_loopdev(loop_dev, loop_file);
1310 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
1311 ret = fscanf(f, fmt, loop_file);
1320 * Checks whether a and b are identical or device
1321 * files associated with the same block device
1323 static int is_same_blk_file(const char* a, const char* b)
1325 struct stat st_buf_a, st_buf_b;
1326 char real_a[PATH_MAX];
1327 char real_b[PATH_MAX];
1329 if (!realpath(a, real_a))
1330 strncpy_null(real_a, a);
1332 if (!realpath(b, real_b))
1333 strncpy_null(real_b, b);
1335 /* Identical path? */
1336 if (strcmp(real_a, real_b) == 0)
1339 if (stat(a, &st_buf_a) < 0 || stat(b, &st_buf_b) < 0) {
1340 if (errno == ENOENT)
1345 /* Same blockdevice? */
1346 if (S_ISBLK(st_buf_a.st_mode) && S_ISBLK(st_buf_b.st_mode) &&
1347 st_buf_a.st_rdev == st_buf_b.st_rdev) {
1352 if (st_buf_a.st_dev == st_buf_b.st_dev &&
1353 st_buf_a.st_ino == st_buf_b.st_ino) {
1360 /* checks if a and b are identical or device
1361 * files associated with the same block device or
1362 * if one file is a loop device that uses the other
1365 static int is_same_loop_file(const char* a, const char* b)
1367 char res_a[PATH_MAX];
1368 char res_b[PATH_MAX];
1369 const char* final_a = NULL;
1370 const char* final_b = NULL;
1373 /* Resolve a if it is a loop device */
1374 if((ret = is_loop_device(a)) < 0) {
1379 ret = resolve_loop_device(a, res_a, sizeof(res_a));
1390 /* Resolve b if it is a loop device */
1391 if ((ret = is_loop_device(b)) < 0) {
1396 ret = resolve_loop_device(b, res_b, sizeof(res_b));
1407 return is_same_blk_file(final_a, final_b);
1410 /* Checks if a file exists and is a block or regular file*/
1411 static int is_existing_blk_or_reg_file(const char* filename)
1415 if(stat(filename, &st_buf) < 0) {
1422 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
1425 /* Checks if a file is used (directly or indirectly via a loop device)
1426 * by a device in fs_devices
1428 static int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices,
1432 struct list_head *head;
1433 struct list_head *cur;
1434 struct btrfs_device *device;
1436 head = &fs_devices->devices;
1437 list_for_each(cur, head) {
1438 device = list_entry(cur, struct btrfs_device, dev_list);
1440 if((ret = is_same_loop_file(device->name, file)))
1448 * Resolve a pathname to a device mapper node to /dev/mapper/<name>
1449 * Returns NULL on invalid input or malloc failure; Other failures
1450 * will be handled by the caller using the input pathame.
1452 char *canonicalize_dm_name(const char *ptname)
1456 char path[PATH_MAX], name[PATH_MAX], *res = NULL;
1458 if (!ptname || !*ptname)
1461 snprintf(path, sizeof(path), "/sys/block/%s/dm/name", ptname);
1462 if (!(f = fopen(path, "r")))
1465 /* read <name>\n from sysfs */
1466 if (fgets(name, sizeof(name), f) && (sz = strlen(name)) > 1) {
1467 name[sz - 1] = '\0';
1468 snprintf(path, sizeof(path), "/dev/mapper/%s", name);
1470 if (access(path, F_OK) == 0)
1478 * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or
1479 * to a device mapper pathname.
1480 * Returns NULL on invalid input or malloc failure; Other failures
1481 * will be handled by the caller using the input pathame.
1483 char *canonicalize_path(const char *path)
1485 char *canonical, *p;
1487 if (!path || !*path)
1490 canonical = realpath(path, NULL);
1492 return strdup(path);
1493 p = strrchr(canonical, '/');
1494 if (p && strncmp(p, "/dm-", 4) == 0 && isdigit(*(p + 4))) {
1495 char *dm = canonicalize_dm_name(p + 1);
1506 * returns 1 if the device was mounted, < 0 on error or 0 if everything
1507 * is safe to continue.
1509 int check_mounted(const char* file)
1514 fd = open(file, O_RDONLY);
1516 error("mount check: cannot open %s: %s", file,
1521 ret = check_mounted_where(fd, file, NULL, 0, NULL);
1527 int check_mounted_where(int fd, const char *file, char *where, int size,
1528 struct btrfs_fs_devices **fs_dev_ret)
1533 struct btrfs_fs_devices *fs_devices_mnt = NULL;
1537 /* scan the initial device */
1538 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
1539 &total_devs, BTRFS_SUPER_INFO_OFFSET, 0);
1540 is_btrfs = (ret >= 0);
1542 /* scan other devices */
1543 if (is_btrfs && total_devs > 1) {
1544 ret = btrfs_scan_lblkid();
1549 /* iterate over the list of currently mounted filesystems */
1550 if ((f = setmntent ("/proc/self/mounts", "r")) == NULL)
1553 while ((mnt = getmntent (f)) != NULL) {
1555 if(strcmp(mnt->mnt_type, "btrfs") != 0)
1558 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
1560 /* ignore entries in the mount table that are not
1561 associated with a file*/
1562 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
1563 goto out_mntloop_err;
1567 ret = is_same_loop_file(file, mnt->mnt_fsname);
1571 goto out_mntloop_err;
1576 /* Did we find an entry in mnt table? */
1577 if (mnt && size && where) {
1578 strncpy(where, mnt->mnt_dir, size);
1582 *fs_dev_ret = fs_devices_mnt;
1584 ret = (mnt != NULL);
1592 struct pending_dir {
1593 struct list_head list;
1594 char name[PATH_MAX];
1597 int btrfs_register_one_device(const char *fname)
1599 struct btrfs_ioctl_vol_args args;
1603 fd = open("/dev/btrfs-control", O_RDWR);
1606 "failed to open /dev/btrfs-control, skipping device registration: %s",
1610 memset(&args, 0, sizeof(args));
1611 strncpy_null(args.name, fname);
1612 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
1614 error("device scan failed on '%s': %s", fname,
1623 * Register all devices in the fs_uuid list created in the user
1624 * space. Ensure btrfs_scan_lblkid() is called before this func.
1626 int btrfs_register_all_devices(void)
1630 struct btrfs_fs_devices *fs_devices;
1631 struct btrfs_device *device;
1632 struct list_head *all_uuids;
1634 all_uuids = btrfs_scanned_uuids();
1636 list_for_each_entry(fs_devices, all_uuids, list) {
1637 list_for_each_entry(device, &fs_devices->devices, dev_list) {
1639 err = btrfs_register_one_device(device->name);
1649 int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
1652 struct btrfs_super_block *disk_super;
1656 buf = malloc(BTRFS_SUPER_INFO_SIZE);
1661 ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
1662 if (ret != BTRFS_SUPER_INFO_SIZE)
1666 disk_super = (struct btrfs_super_block *)buf;
1667 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC)
1670 if (!memcmp(disk_super->fsid, root->fs_info->super_copy->fsid,
1680 * Note: this function uses a static per-thread buffer. Do not call this
1681 * function more than 10 times within one argument list!
1683 const char *pretty_size_mode(u64 size, unsigned mode)
1685 static __thread int ps_index = 0;
1686 static __thread char ps_array[10][32];
1689 ret = ps_array[ps_index];
1692 (void)pretty_size_snprintf(size, ret, 32, mode);
1697 static const char* unit_suffix_binary[] =
1698 { "B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB"};
1699 static const char* unit_suffix_decimal[] =
1700 { "B", "kB", "MB", "GB", "TB", "PB", "EB"};
1702 int pretty_size_snprintf(u64 size, char *str, size_t str_size, unsigned unit_mode)
1708 const char** suffix = NULL;
1714 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_RAW) {
1715 snprintf(str, str_size, "%llu", size);
1719 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_BINARY) {
1722 suffix = unit_suffix_binary;
1723 } else if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_DECIMAL) {
1726 suffix = unit_suffix_decimal;
1731 fprintf(stderr, "INTERNAL ERROR: unknown unit base, mode %d\n",
1739 switch (unit_mode & UNITS_MODE_MASK) {
1740 case UNITS_TBYTES: base *= mult; num_divs++;
1741 case UNITS_GBYTES: base *= mult; num_divs++;
1742 case UNITS_MBYTES: base *= mult; num_divs++;
1743 case UNITS_KBYTES: num_divs++;
1750 while (size >= mult) {
1756 * If the value is smaller than base, we didn't do any
1757 * division, in that case, base should be 1, not original
1758 * base, or the unit will be wrong
1764 if (num_divs >= ARRAY_SIZE(unit_suffix_binary)) {
1766 printf("INTERNAL ERROR: unsupported unit suffix, index %d\n",
1771 fraction = (float)last_size / base;
1773 return snprintf(str, str_size, "%.2f%s", fraction, suffix[num_divs]);
1777 * __strncpy_null - strncpy with null termination
1778 * @dest: the target array
1779 * @src: the source string
1780 * @n: maximum bytes to copy (size of *dest)
1782 * Like strncpy, but ensures destination is null-terminated.
1784 * Copies the string pointed to by src, including the terminating null
1785 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
1786 * of n bytes. Then ensure that dest is null-terminated.
1788 char *__strncpy_null(char *dest, const char *src, size_t n)
1790 strncpy(dest, src, n);
1797 * Checks to make sure that the label matches our requirements.
1799 0 if everything is safe and usable
1800 -1 if the label is too long
1802 static int check_label(const char *input)
1804 int len = strlen(input);
1806 if (len > BTRFS_LABEL_SIZE - 1) {
1807 error("label %s is too long (max %d)", input,
1808 BTRFS_LABEL_SIZE - 1);
1815 static int set_label_unmounted(const char *dev, const char *label)
1817 struct btrfs_trans_handle *trans;
1818 struct btrfs_root *root;
1821 ret = check_mounted(dev);
1823 error("checking mount status of %s failed: %d", dev, ret);
1827 error("device %s is mounted, use mount point", dev);
1831 /* Open the super_block at the default location
1832 * and as read-write.
1834 root = open_ctree(dev, 0, OPEN_CTREE_WRITES);
1835 if (!root) /* errors are printed by open_ctree() */
1838 trans = btrfs_start_transaction(root, 1);
1839 __strncpy_null(root->fs_info->super_copy->label, label, BTRFS_LABEL_SIZE - 1);
1841 btrfs_commit_transaction(trans, root);
1843 /* Now we close it since we are done. */
1848 static int set_label_mounted(const char *mount_path, const char *labelp)
1851 char label[BTRFS_LABEL_SIZE];
1853 fd = open(mount_path, O_RDONLY | O_NOATIME);
1855 error("unable to access %s: %s", mount_path, strerror(errno));
1859 memset(label, 0, sizeof(label));
1860 __strncpy_null(label, labelp, BTRFS_LABEL_SIZE - 1);
1861 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
1862 error("unable to set label of %s: %s", mount_path,
1872 int get_label_unmounted(const char *dev, char *label)
1874 struct btrfs_root *root;
1877 ret = check_mounted(dev);
1879 error("checking mount status of %s failed: %d", dev, ret);
1883 /* Open the super_block at the default location
1886 root = open_ctree(dev, 0, 0);
1890 __strncpy_null(label, root->fs_info->super_copy->label,
1891 BTRFS_LABEL_SIZE - 1);
1893 /* Now we close it since we are done. */
1899 * If a partition is mounted, try to get the filesystem label via its
1900 * mounted path rather than device. Return the corresponding error
1901 * the user specified the device path.
1903 int get_label_mounted(const char *mount_path, char *labelp)
1905 char label[BTRFS_LABEL_SIZE];
1909 fd = open(mount_path, O_RDONLY | O_NOATIME);
1911 error("unable to access %s: %s", mount_path, strerror(errno));
1915 memset(label, '\0', sizeof(label));
1916 ret = ioctl(fd, BTRFS_IOC_GET_FSLABEL, label);
1918 if (errno != ENOTTY)
1919 error("unable to get label of %s: %s", mount_path,
1926 __strncpy_null(labelp, label, BTRFS_LABEL_SIZE - 1);
1931 int get_label(const char *btrfs_dev, char *label)
1935 ret = is_existing_blk_or_reg_file(btrfs_dev);
1937 ret = get_label_mounted(btrfs_dev, label);
1939 ret = get_label_unmounted(btrfs_dev, label);
1944 int set_label(const char *btrfs_dev, const char *label)
1948 if (check_label(label))
1951 ret = is_existing_blk_or_reg_file(btrfs_dev);
1953 ret = set_label_mounted(btrfs_dev, label);
1955 ret = set_label_unmounted(btrfs_dev, label);
1961 * A not-so-good version fls64. No fascinating optimization since
1962 * no one except parse_size use it
1964 static int fls64(u64 x)
1968 for (i = 0; i <64; i++)
1969 if (x << i & (1ULL << 63))
1974 u64 parse_size(char *s)
1982 error("size value is empty");
1986 error("size value '%s' is less equal than 0", s);
1989 ret = strtoull(s, &endptr, 10);
1991 error("size value '%s' is invalid", s);
1994 if (endptr[0] && endptr[1]) {
1995 error("illegal suffix contains character '%c' in wrong position",
2000 * strtoll returns LLONG_MAX when overflow, if this happens,
2001 * need to call strtoull to get the real size
2003 if (errno == ERANGE && ret == ULLONG_MAX) {
2004 error("size value '%s' is too large for u64", s);
2008 c = tolower(endptr[0]);
2031 error("unknown size descriptor '%c'", c);
2035 /* Check whether ret * mult overflow */
2036 if (fls64(ret) + fls64(mult) - 1 > 64) {
2037 error("size value '%s' is too large for u64", s);
2044 u64 parse_qgroupid(const char *p)
2046 char *s = strchr(p, '/');
2047 const char *ptr_src_end = p + strlen(p);
2048 char *ptr_parse_end = NULL;
2057 /* Numeric format like '0/257' is the primary case */
2059 id = strtoull(p, &ptr_parse_end, 10);
2060 if (ptr_parse_end != ptr_src_end)
2064 level = strtoull(p, &ptr_parse_end, 10);
2065 if (ptr_parse_end != s)
2068 id = strtoull(s + 1, &ptr_parse_end, 10);
2069 if (ptr_parse_end != ptr_src_end)
2072 return (level << BTRFS_QGROUP_LEVEL_SHIFT) | id;
2075 /* Path format like subv at 'my_subvol' is the fallback case */
2076 ret = test_issubvolume(p);
2077 if (ret < 0 || !ret)
2079 fd = open(p, O_RDONLY);
2082 ret = lookup_ino_rootid(fd, &id);
2089 error("invalid qgroupid or subvolume path: %s", p);
2093 int open_file_or_dir3(const char *fname, DIR **dirstream, int open_flags)
2099 ret = stat(fname, &st);
2103 if (S_ISDIR(st.st_mode)) {
2104 *dirstream = opendir(fname);
2107 fd = dirfd(*dirstream);
2108 } else if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) {
2109 fd = open(fname, open_flags);
2112 * we set this on purpose, in case the caller output
2113 * strerror(errno) as success
2121 closedir(*dirstream);
2128 int open_file_or_dir(const char *fname, DIR **dirstream)
2130 return open_file_or_dir3(fname, dirstream, O_RDWR);
2133 void close_file_or_dir(int fd, DIR *dirstream)
2136 closedir(dirstream);
2141 int get_device_info(int fd, u64 devid,
2142 struct btrfs_ioctl_dev_info_args *di_args)
2146 di_args->devid = devid;
2147 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
2149 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
2150 return ret < 0 ? -errno : 0;
2153 static u64 find_max_device_id(struct btrfs_ioctl_search_args *search_args,
2156 struct btrfs_dev_item *dev_item;
2157 char *buf = search_args->buf;
2159 buf += (nr_items - 1) * (sizeof(struct btrfs_ioctl_search_header)
2160 + sizeof(struct btrfs_dev_item));
2161 buf += sizeof(struct btrfs_ioctl_search_header);
2163 dev_item = (struct btrfs_dev_item *)buf;
2165 return btrfs_stack_device_id(dev_item);
2168 static int search_chunk_tree_for_fs_info(int fd,
2169 struct btrfs_ioctl_fs_info_args *fi_args)
2173 u64 start_devid = 1;
2174 struct btrfs_ioctl_search_args search_args;
2175 struct btrfs_ioctl_search_key *search_key = &search_args.key;
2177 fi_args->num_devices = 0;
2179 max_items = BTRFS_SEARCH_ARGS_BUFSIZE
2180 / (sizeof(struct btrfs_ioctl_search_header)
2181 + sizeof(struct btrfs_dev_item));
2183 search_key->tree_id = BTRFS_CHUNK_TREE_OBJECTID;
2184 search_key->min_objectid = BTRFS_DEV_ITEMS_OBJECTID;
2185 search_key->max_objectid = BTRFS_DEV_ITEMS_OBJECTID;
2186 search_key->min_type = BTRFS_DEV_ITEM_KEY;
2187 search_key->max_type = BTRFS_DEV_ITEM_KEY;
2188 search_key->min_transid = 0;
2189 search_key->max_transid = (u64)-1;
2190 search_key->nr_items = max_items;
2191 search_key->max_offset = (u64)-1;
2194 search_key->min_offset = start_devid;
2196 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH, &search_args);
2200 fi_args->num_devices += (u64)search_key->nr_items;
2202 if (search_key->nr_items == max_items) {
2203 start_devid = find_max_device_id(&search_args,
2204 search_key->nr_items) + 1;
2208 /* get the lastest max_id to stay consistent with the num_devices */
2209 if (search_key->nr_items == 0)
2211 * last tree_search returns an empty buf, use the devid of
2212 * the last dev_item of the previous tree_search
2214 fi_args->max_id = start_devid - 1;
2216 fi_args->max_id = find_max_device_id(&search_args,
2217 search_key->nr_items);
2223 * For a given path, fill in the ioctl fs_ and info_ args.
2224 * If the path is a btrfs mountpoint, fill info for all devices.
2225 * If the path is a btrfs device, fill in only that device.
2227 * The path provided must be either on a mounted btrfs fs,
2228 * or be a mounted btrfs device.
2230 * Returns 0 on success, or a negative errno.
2232 int get_fs_info(char *path, struct btrfs_ioctl_fs_info_args *fi_args,
2233 struct btrfs_ioctl_dev_info_args **di_ret)
2240 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2241 struct btrfs_ioctl_dev_info_args *di_args;
2242 struct btrfs_ioctl_dev_info_args tmp;
2244 DIR *dirstream = NULL;
2246 memset(fi_args, 0, sizeof(*fi_args));
2248 if (is_block_device(path) == 1) {
2249 struct btrfs_super_block *disk_super;
2250 char buf[BTRFS_SUPER_INFO_SIZE];
2253 /* Ensure it's mounted, then set path to the mountpoint */
2254 fd = open(path, O_RDONLY);
2257 error("cannot open %s: %s", path, strerror(errno));
2260 ret = check_mounted_where(fd, path, mp, sizeof(mp),
2269 /* Only fill in this one device */
2270 fi_args->num_devices = 1;
2272 disk_super = (struct btrfs_super_block *)buf;
2273 ret = btrfs_read_dev_super(fd, disk_super,
2274 BTRFS_SUPER_INFO_OFFSET, 0);
2279 devid = btrfs_stack_device_id(&disk_super->dev_item);
2281 fi_args->max_id = devid;
2284 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
2288 /* at this point path must not be for a block device */
2289 fd = open_file_or_dir(path, &dirstream);
2295 /* fill in fi_args if not just a single device */
2296 if (fi_args->num_devices != 1) {
2297 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
2304 * The fs_args->num_devices does not include seed devices
2306 ret = search_chunk_tree_for_fs_info(fd, fi_args);
2311 * search_chunk_tree_for_fs_info() will lacks the devid 0
2312 * so manual probe for it here.
2314 ret = get_device_info(fd, 0, &tmp);
2316 fi_args->num_devices++;
2324 if (!fi_args->num_devices)
2327 di_args = *di_ret = malloc((fi_args->num_devices) * sizeof(*di_args));
2334 memcpy(di_args, &tmp, sizeof(tmp));
2335 for (; i <= fi_args->max_id; ++i) {
2336 ret = get_device_info(fd, i, &di_args[ndevs]);
2345 * only when the only dev we wanted to find is not there then
2346 * let any error be returned
2348 if (fi_args->num_devices != 1) {
2354 close_file_or_dir(fd, dirstream);
2358 #define isoctal(c) (((c) & ~7) == '0')
2360 static inline void translate(char *f, char *t)
2362 while (*f != '\0') {
2364 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
2365 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
2375 * Checks if the swap device.
2376 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
2378 static int is_swap_device(const char *file)
2389 if (stat(file, &st_buf) < 0)
2391 if (S_ISBLK(st_buf.st_mode))
2392 dev = st_buf.st_rdev;
2393 else if (S_ISREG(st_buf.st_mode)) {
2394 dev = st_buf.st_dev;
2395 ino = st_buf.st_ino;
2399 if ((f = fopen("/proc/swaps", "r")) == NULL)
2402 /* skip the first line */
2403 if (fgets(tmp, sizeof(tmp), f) == NULL)
2406 while (fgets(tmp, sizeof(tmp), f) != NULL) {
2407 if ((cp = strchr(tmp, ' ')) != NULL)
2409 if ((cp = strchr(tmp, '\t')) != NULL)
2411 translate(tmp, buf);
2412 if (stat(buf, &st_buf) != 0)
2414 if (S_ISBLK(st_buf.st_mode)) {
2415 if (dev == st_buf.st_rdev) {
2419 } else if (S_ISREG(st_buf.st_mode)) {
2420 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {
2434 * Check for existing filesystem or partition table on device.
2436 * 1 for existing fs or partition
2437 * 0 for nothing found
2438 * -1 for internal error
2440 static int check_overwrite(const char *device)
2443 blkid_probe pr = NULL;
2447 if (!device || !*device)
2450 ret = -1; /* will reset on success of all setup calls */
2452 pr = blkid_new_probe_from_filename(device);
2456 size = blkid_probe_get_size(pr);
2460 /* nothing to overwrite on a 0-length device */
2466 ret = blkid_probe_enable_partitions(pr, 1);
2470 ret = blkid_do_fullprobe(pr);
2475 * Blkid returns 1 for nothing found and 0 when it finds a signature,
2476 * but we want the exact opposite, so reverse the return value here.
2478 * In addition print some useful diagnostics about what actually is
2486 if (!blkid_probe_lookup_value(pr, "TYPE", &type, NULL)) {
2488 "%s appears to contain an existing "
2489 "filesystem (%s).\n", device, type);
2490 } else if (!blkid_probe_lookup_value(pr, "PTTYPE", &type, NULL)) {
2492 "%s appears to contain a partition "
2493 "table (%s).\n", device, type);
2496 "%s appears to contain something weird "
2497 "according to blkid\n", device);
2503 blkid_free_probe(pr);
2506 "probe of %s failed, cannot detect "
2507 "existing filesystem.\n", device);
2511 static int group_profile_devs_min(u64 flag)
2513 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
2514 case 0: /* single */
2515 case BTRFS_BLOCK_GROUP_DUP:
2517 case BTRFS_BLOCK_GROUP_RAID0:
2518 case BTRFS_BLOCK_GROUP_RAID1:
2519 case BTRFS_BLOCK_GROUP_RAID5:
2521 case BTRFS_BLOCK_GROUP_RAID6:
2523 case BTRFS_BLOCK_GROUP_RAID10:
2530 int test_num_disk_vs_raid(u64 metadata_profile, u64 data_profile,
2531 u64 dev_cnt, int mixed, int ssd)
2538 allowed |= BTRFS_BLOCK_GROUP_RAID10;
2540 allowed |= BTRFS_BLOCK_GROUP_RAID6;
2542 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
2543 BTRFS_BLOCK_GROUP_RAID5;
2545 allowed |= BTRFS_BLOCK_GROUP_DUP;
2549 ((metadata_profile | data_profile) & BTRFS_BLOCK_GROUP_DUP)) {
2550 warning("DUP is not recommended on filesystem with multiple devices");
2552 if (metadata_profile & ~allowed) {
2554 "ERROR: unable to create FS with metadata profile %s "
2555 "(have %llu devices but %d devices are required)\n",
2556 btrfs_group_profile_str(metadata_profile), dev_cnt,
2557 group_profile_devs_min(metadata_profile));
2560 if (data_profile & ~allowed) {
2562 "ERROR: unable to create FS with data profile %s "
2563 "(have %llu devices but %d devices are required)\n",
2564 btrfs_group_profile_str(data_profile), dev_cnt,
2565 group_profile_devs_min(data_profile));
2569 warning_on(!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP) && ssd,
2570 "DUP may not actually lead to 2 copies on the device, see manual page");
2575 int group_profile_max_safe_loss(u64 flags)
2577 switch (flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
2578 case 0: /* single */
2579 case BTRFS_BLOCK_GROUP_DUP:
2580 case BTRFS_BLOCK_GROUP_RAID0:
2582 case BTRFS_BLOCK_GROUP_RAID1:
2583 case BTRFS_BLOCK_GROUP_RAID5:
2584 case BTRFS_BLOCK_GROUP_RAID10:
2586 case BTRFS_BLOCK_GROUP_RAID6:
2594 * Check if a device is suitable for btrfs
2596 * 1: something is wrong, an error is printed
2599 int test_dev_for_mkfs(const char *file, int force_overwrite)
2604 ret = is_swap_device(file);
2606 error("checking status of %s: %s", file, strerror(-ret));
2610 error("%s is a swap device", file);
2613 if (!force_overwrite) {
2614 if (check_overwrite(file)) {
2615 error("use the -f option to force overwrite of %s",
2620 ret = check_mounted(file);
2622 error("cannot check mount status of %s: %s", file,
2627 error("%s is mounted", file);
2630 /* check if the device is busy */
2631 fd = open(file, O_RDWR|O_EXCL);
2633 error("unable to open %s: %s", file, strerror(errno));
2636 if (fstat(fd, &st)) {
2637 error("unable to stat %s: %s", file, strerror(errno));
2641 if (!S_ISBLK(st.st_mode)) {
2642 error("%s is not a block device", file);
2650 int btrfs_scan_lblkid(void)
2655 struct btrfs_fs_devices *tmp_devices;
2656 blkid_dev_iterate iter = NULL;
2657 blkid_dev dev = NULL;
2658 blkid_cache cache = NULL;
2659 char path[PATH_MAX];
2661 if (btrfs_scan_done)
2664 if (blkid_get_cache(&cache, NULL) < 0) {
2665 error("blkid cache get failed");
2668 blkid_probe_all(cache);
2669 iter = blkid_dev_iterate_begin(cache);
2670 blkid_dev_set_search(iter, "TYPE", "btrfs");
2671 while (blkid_dev_next(iter, &dev) == 0) {
2672 dev = blkid_verify(cache, dev);
2675 /* if we are here its definitely a btrfs disk*/
2676 strncpy_null(path, blkid_dev_devname(dev));
2678 fd = open(path, O_RDONLY);
2680 error("cannot open %s: %s", path, strerror(errno));
2683 ret = btrfs_scan_one_device(fd, path, &tmp_devices,
2684 &num_devices, BTRFS_SUPER_INFO_OFFSET, 0);
2686 error("cannot scan %s: %s", path, strerror(-ret));
2693 blkid_dev_iterate_end(iter);
2694 blkid_put_cache(cache);
2696 btrfs_scan_done = 1;
2701 int is_vol_small(const char *file)
2708 fd = open(file, O_RDONLY);
2711 if (fstat(fd, &st) < 0) {
2716 size = btrfs_device_size(fd, &st);
2721 if (size < BTRFS_MKFS_SMALL_VOLUME_SIZE) {
2731 * This reads a line from the stdin and only returns non-zero if the
2732 * first whitespace delimited token is a case insensitive match with yes
2735 int ask_user(const char *question)
2737 char buf[30] = {0,};
2738 char *saveptr = NULL;
2741 printf("%s [y/N]: ", question);
2743 return fgets(buf, sizeof(buf) - 1, stdin) &&
2744 (answer = strtok_r(buf, " \t\n\r", &saveptr)) &&
2745 (!strcasecmp(answer, "yes") || !strcasecmp(answer, "y"));
2750 * - file or directory return the containing tree root id
2751 * - subvolume return its own tree id
2752 * - BTRFS_EMPTY_SUBVOL_DIR_OBJECTID (directory with ino == 2) the result is
2753 * undefined and function returns -1
2755 int lookup_ino_rootid(int fd, u64 *rootid)
2757 struct btrfs_ioctl_ino_lookup_args args;
2760 memset(&args, 0, sizeof(args));
2762 args.objectid = BTRFS_FIRST_FREE_OBJECTID;
2764 ret = ioctl(fd, BTRFS_IOC_INO_LOOKUP, &args);
2766 error("failed to lookup root id: %s", strerror(errno));
2770 *rootid = args.treeid;
2776 * return 0 if a btrfs mount point is found
2777 * return 1 if a mount point is found but not btrfs
2778 * return <0 if something goes wrong
2780 int find_mount_root(const char *path, char **mount_root)
2788 int longest_matchlen = 0;
2789 char *longest_match = NULL;
2791 fd = open(path, O_RDONLY | O_NOATIME);
2796 mnttab = setmntent("/proc/self/mounts", "r");
2800 while ((ent = getmntent(mnttab))) {
2801 len = strlen(ent->mnt_dir);
2802 if (strncmp(ent->mnt_dir, path, len) == 0) {
2803 /* match found and use the latest match */
2804 if (longest_matchlen <= len) {
2805 free(longest_match);
2806 longest_matchlen = len;
2807 longest_match = strdup(ent->mnt_dir);
2808 not_btrfs = strcmp(ent->mnt_type, "btrfs");
2817 free(longest_match);
2822 *mount_root = realpath(longest_match, NULL);
2826 free(longest_match);
2830 int test_minimum_size(const char *file, u32 nodesize)
2833 struct stat statbuf;
2835 fd = open(file, O_RDONLY);
2838 if (stat(file, &statbuf) < 0) {
2842 if (btrfs_device_size(fd, &statbuf) < btrfs_min_dev_size(nodesize)) {
2852 * Test if path is a directory
2854 * 0 - path exists but it is not a directory
2855 * 1 - path exists and it is a directory
2858 int test_isdir(const char *path)
2863 ret = stat(path, &st);
2867 return !!S_ISDIR(st.st_mode);
2870 void units_set_mode(unsigned *units, unsigned mode)
2872 unsigned base = *units & UNITS_MODE_MASK;
2874 *units = base | mode;
2877 void units_set_base(unsigned *units, unsigned base)
2879 unsigned mode = *units & ~UNITS_MODE_MASK;
2881 *units = base | mode;
2884 int find_next_key(struct btrfs_path *path, struct btrfs_key *key)
2888 for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
2889 if (!path->nodes[level])
2891 if (path->slots[level] + 1 >=
2892 btrfs_header_nritems(path->nodes[level]))
2895 btrfs_item_key_to_cpu(path->nodes[level], key,
2896 path->slots[level] + 1);
2898 btrfs_node_key_to_cpu(path->nodes[level], key,
2899 path->slots[level] + 1);
2905 const char* btrfs_group_type_str(u64 flag)
2907 u64 mask = BTRFS_BLOCK_GROUP_TYPE_MASK |
2908 BTRFS_SPACE_INFO_GLOBAL_RSV;
2910 switch (flag & mask) {
2911 case BTRFS_BLOCK_GROUP_DATA:
2913 case BTRFS_BLOCK_GROUP_SYSTEM:
2915 case BTRFS_BLOCK_GROUP_METADATA:
2917 case BTRFS_BLOCK_GROUP_DATA|BTRFS_BLOCK_GROUP_METADATA:
2918 return "Data+Metadata";
2919 case BTRFS_SPACE_INFO_GLOBAL_RSV:
2920 return "GlobalReserve";
2926 const char* btrfs_group_profile_str(u64 flag)
2928 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
2931 case BTRFS_BLOCK_GROUP_RAID0:
2933 case BTRFS_BLOCK_GROUP_RAID1:
2935 case BTRFS_BLOCK_GROUP_RAID5:
2937 case BTRFS_BLOCK_GROUP_RAID6:
2939 case BTRFS_BLOCK_GROUP_DUP:
2941 case BTRFS_BLOCK_GROUP_RAID10:
2948 u64 disk_size(const char *path)
2952 if (statfs(path, &sfs) < 0)
2955 return sfs.f_bsize * sfs.f_blocks;
2958 u64 get_partition_size(const char *dev)
2961 int fd = open(dev, O_RDONLY);
2965 if (ioctl(fd, BLKGETSIZE64, &result) < 0) {
2974 int btrfs_tree_search2_ioctl_supported(int fd)
2976 struct btrfs_ioctl_search_args_v2 *args2;
2977 struct btrfs_ioctl_search_key *sk;
2978 int args2_size = 1024;
2979 char args2_buf[args2_size];
2981 static int v2_supported = -1;
2983 if (v2_supported != -1)
2984 return v2_supported;
2986 args2 = (struct btrfs_ioctl_search_args_v2 *)args2_buf;
2990 * Search for the extent tree item in the root tree.
2992 sk->tree_id = BTRFS_ROOT_TREE_OBJECTID;
2993 sk->min_objectid = BTRFS_EXTENT_TREE_OBJECTID;
2994 sk->max_objectid = BTRFS_EXTENT_TREE_OBJECTID;
2995 sk->min_type = BTRFS_ROOT_ITEM_KEY;
2996 sk->max_type = BTRFS_ROOT_ITEM_KEY;
2998 sk->max_offset = (u64)-1;
2999 sk->min_transid = 0;
3000 sk->max_transid = (u64)-1;
3002 args2->buf_size = args2_size - sizeof(struct btrfs_ioctl_search_args_v2);
3003 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH_V2, args2);
3004 if (ret == -EOPNOTSUPP)
3011 return v2_supported;
3014 int btrfs_check_nodesize(u32 nodesize, u32 sectorsize, u64 features)
3016 if (nodesize < sectorsize) {
3017 error("illegal nodesize %u (smaller than %u)",
3018 nodesize, sectorsize);
3020 } else if (nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) {
3021 error("illegal nodesize %u (larger than %u)",
3022 nodesize, BTRFS_MAX_METADATA_BLOCKSIZE);
3024 } else if (nodesize & (sectorsize - 1)) {
3025 error("illegal nodesize %u (not aligned to %u)",
3026 nodesize, sectorsize);
3028 } else if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS &&
3029 nodesize != sectorsize) {
3030 error("illegal nodesize %u (not equal to %u for mixed block group)",
3031 nodesize, sectorsize);
3038 * Copy a path argument from SRC to DEST and check the SRC length if it's at
3039 * most PATH_MAX and fits into DEST. DESTLEN is supposed to be exact size of
3041 * The destination buffer is zero terminated.
3042 * Return < 0 for error, 0 otherwise.
3044 int arg_copy_path(char *dest, const char *src, int destlen)
3046 size_t len = strlen(src);
3048 if (len >= PATH_MAX || len >= destlen)
3049 return -ENAMETOOLONG;
3051 __strncpy_null(dest, src, destlen);
3056 unsigned int get_unit_mode_from_arg(int *argc, char *argv[], int df_mode)
3058 unsigned int unit_mode = UNITS_DEFAULT;
3062 for (arg_i = 0; arg_i < *argc; arg_i++) {
3063 if (!strcmp(argv[arg_i], "--"))
3066 if (!strcmp(argv[arg_i], "--raw")) {
3067 unit_mode = UNITS_RAW;
3071 if (!strcmp(argv[arg_i], "--human-readable")) {
3072 unit_mode = UNITS_HUMAN_BINARY;
3077 if (!strcmp(argv[arg_i], "--iec")) {
3078 units_set_mode(&unit_mode, UNITS_BINARY);
3082 if (!strcmp(argv[arg_i], "--si")) {
3083 units_set_mode(&unit_mode, UNITS_DECIMAL);
3088 if (!strcmp(argv[arg_i], "--kbytes")) {
3089 units_set_base(&unit_mode, UNITS_KBYTES);
3093 if (!strcmp(argv[arg_i], "--mbytes")) {
3094 units_set_base(&unit_mode, UNITS_MBYTES);
3098 if (!strcmp(argv[arg_i], "--gbytes")) {
3099 units_set_base(&unit_mode, UNITS_GBYTES);
3103 if (!strcmp(argv[arg_i], "--tbytes")) {
3104 units_set_base(&unit_mode, UNITS_TBYTES);
3112 if (!strcmp(argv[arg_i], "-b")) {
3113 unit_mode = UNITS_RAW;
3117 if (!strcmp(argv[arg_i], "-h")) {
3118 unit_mode = UNITS_HUMAN_BINARY;
3122 if (!strcmp(argv[arg_i], "-H")) {
3123 unit_mode = UNITS_HUMAN_DECIMAL;
3127 if (!strcmp(argv[arg_i], "-k")) {
3128 units_set_base(&unit_mode, UNITS_KBYTES);
3132 if (!strcmp(argv[arg_i], "-m")) {
3133 units_set_base(&unit_mode, UNITS_MBYTES);
3137 if (!strcmp(argv[arg_i], "-g")) {
3138 units_set_base(&unit_mode, UNITS_GBYTES);
3142 if (!strcmp(argv[arg_i], "-t")) {
3143 units_set_base(&unit_mode, UNITS_TBYTES);
3149 for (arg_i = 0, arg_end = 0; arg_i < *argc; arg_i++) {
3152 argv[arg_end] = argv[arg_i];
3161 int string_is_numerical(const char *str)
3163 if (!(*str >= '0' && *str <= '9'))
3165 while (*str >= '0' && *str <= '9')
3173 * Preprocess @argv with getopt_long to reorder options and consume the "--"
3175 * Unknown short and long options are reported, optionally the @usage is printed
3178 void clean_args_no_options(int argc, char *argv[], const char * const *usagestr)
3180 static const struct option long_options[] = {
3185 int c = getopt_long(argc, argv, "", long_options, NULL);
3198 /* Subvolume helper functions */
3200 * test if name is a correct subvolume name
3201 * this function return
3202 * 0-> name is not a correct subvolume name
3203 * 1-> name is a correct subvolume name
3205 int test_issubvolname(const char *name)
3207 return name[0] != '\0' && !strchr(name, '/') &&
3208 strcmp(name, ".") && strcmp(name, "..");
3212 * Test if path is a subvolume
3214 * 0 - path exists but it is not a subvolume
3215 * 1 - path exists and it is a subvolume
3218 int test_issubvolume(const char *path)
3224 res = stat(path, &st);
3228 if (st.st_ino != BTRFS_FIRST_FREE_OBJECTID || !S_ISDIR(st.st_mode))
3231 res = statfs(path, &stfs);
3235 return (int)stfs.f_type == BTRFS_SUPER_MAGIC;
3238 const char *subvol_strip_mountpoint(const char *mnt, const char *full_path)
3240 int len = strlen(mnt);
3244 if (mnt[len - 1] != '/')
3247 return full_path + len;
3254 * 1: Error; and error info printed to the terminal. Fixme.
3255 * 2: If the fullpath is root tree instead of subvol tree
3257 int get_subvol_info(const char *fullpath, struct root_info *get_ri)
3264 const char *svpath = NULL;
3265 DIR *dirstream1 = NULL;
3266 DIR *dirstream2 = NULL;
3268 ret = test_issubvolume(fullpath);
3272 error("not a subvolume: %s", fullpath);
3276 ret = find_mount_root(fullpath, &mnt);
3280 error("%s doesn't belong to btrfs mount point", fullpath);
3284 svpath = subvol_strip_mountpoint(mnt, fullpath);
3286 fd = btrfs_open_dir(fullpath, &dirstream1, 1);
3290 ret = btrfs_list_get_path_rootid(fd, &sv_id);
3292 error("can't get rootid for '%s'", fullpath);
3296 mntfd = btrfs_open_dir(mnt, &dirstream2, 1);
3300 if (sv_id == BTRFS_FS_TREE_OBJECTID) {
3303 * So that caller may decide if thats an error or just fine.
3308 memset(get_ri, 0, sizeof(*get_ri));
3309 get_ri->root_id = sv_id;
3311 ret = btrfs_get_subvol(mntfd, get_ri);
3313 error("can't find '%s': %d", svpath, ret);
3316 close_file_or_dir(mntfd, dirstream2);
3317 close_file_or_dir(fd, dirstream1);
3323 void init_rand_seed(u64 seed)
3327 /* only use the last 48 bits */
3328 for (i = 0; i < 3; i++) {
3329 rand_seed[i] = (unsigned short)(seed ^ (unsigned short)(-1));
3332 rand_seed_initlized = 1;
3335 static void __init_seed(void)
3341 if(rand_seed_initlized)
3343 /* Use urandom as primary seed source. */
3344 fd = open("/dev/urandom", O_RDONLY);
3346 ret = read(fd, rand_seed, sizeof(rand_seed));
3348 if (ret < sizeof(rand_seed))
3352 /* Use time and pid as fallback seed */
3353 warning("failed to read /dev/urandom, use time and pid as random seed");
3354 gettimeofday(&tv, 0);
3355 rand_seed[0] = getpid() ^ (tv.tv_sec & 0xFFFF);
3356 rand_seed[1] = getppid() ^ (tv.tv_usec & 0xFFFF);
3357 rand_seed[2] = (tv.tv_sec ^ tv.tv_usec) >> 16;
3359 rand_seed_initlized = 1;
3366 * Don't use nrand48, its range is [0,2^31) The highest bit will alwasy
3367 * be 0. Use jrand48 to include the highest bit.
3369 return (u32)jrand48(rand_seed);
3372 unsigned int rand_range(unsigned int upper)
3376 * Use the full 48bits to mod, which would be more uniformly
3379 return (unsigned int)(jrand48(rand_seed) % upper);