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);
218 static inline int write_temp_super(int fd, struct btrfs_super_block *sb,
224 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
225 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
226 btrfs_csum_final(crc, (char *)&sb->csum[0]);
227 ret = pwrite(fd, sb, BTRFS_SUPER_INFO_SIZE, sb_bytenr);
228 if (ret < BTRFS_SUPER_INFO_SIZE)
229 ret = (ret < 0 ? -errno : -EIO);
236 * Setup temporary superblock at cfg->super_bynter
237 * Needed info are extracted from cfg, and root_bytenr, chunk_bytenr
239 * For now sys chunk array will be empty and dev_item is empty too.
240 * They will be re-initialized at temp chunk tree setup.
242 static int setup_temp_super(int fd, struct btrfs_mkfs_config *cfg,
243 u64 root_bytenr, u64 chunk_bytenr)
245 unsigned char chunk_uuid[BTRFS_UUID_SIZE];
246 char super_buf[BTRFS_SUPER_INFO_SIZE];
247 struct btrfs_super_block *super = (struct btrfs_super_block *)super_buf;
251 * We rely on cfg->chunk_uuid and cfg->fs_uuid to pass uuid
252 * for other functions.
253 * Caller must allocate space for them
255 BUG_ON(!cfg->chunk_uuid || !cfg->fs_uuid);
256 memset(super_buf, 0, BTRFS_SUPER_INFO_SIZE);
257 cfg->num_bytes = round_down(cfg->num_bytes, cfg->sectorsize);
259 if (cfg->fs_uuid && *cfg->fs_uuid) {
260 if (uuid_parse(cfg->fs_uuid, super->fsid) != 0) {
261 error("cound not parse UUID: %s", cfg->fs_uuid);
265 if (!test_uuid_unique(cfg->fs_uuid)) {
266 error("non-unique UUID: %s", cfg->fs_uuid);
271 uuid_generate(super->fsid);
272 uuid_unparse(super->fsid, cfg->fs_uuid);
274 uuid_generate(chunk_uuid);
275 uuid_unparse(chunk_uuid, cfg->chunk_uuid);
277 btrfs_set_super_bytenr(super, cfg->super_bytenr);
278 btrfs_set_super_num_devices(super, 1);
279 btrfs_set_super_magic(super, BTRFS_MAGIC);
280 btrfs_set_super_generation(super, 1);
281 btrfs_set_super_root(super, root_bytenr);
282 btrfs_set_super_chunk_root(super, chunk_bytenr);
283 btrfs_set_super_total_bytes(super, cfg->num_bytes);
285 * Temporary filesystem will only have 6 tree roots:
286 * chunk tree, root tree, extent_tree, device tree, fs tree
289 btrfs_set_super_bytes_used(super, 6 * cfg->nodesize);
290 btrfs_set_super_sectorsize(super, cfg->sectorsize);
291 btrfs_set_super_leafsize(super, cfg->nodesize);
292 btrfs_set_super_nodesize(super, cfg->nodesize);
293 btrfs_set_super_stripesize(super, cfg->stripesize);
294 btrfs_set_super_csum_type(super, BTRFS_CSUM_TYPE_CRC32);
295 btrfs_set_super_chunk_root(super, chunk_bytenr);
296 btrfs_set_super_cache_generation(super, -1);
297 btrfs_set_super_incompat_flags(super, cfg->features);
299 __strncpy_null(super->label, cfg->label, BTRFS_LABEL_SIZE - 1);
301 /* Sys chunk array will be re-initialized at chunk tree init time */
302 super->sys_chunk_array_size = 0;
304 ret = write_temp_super(fd, super, cfg->super_bytenr);
310 * Setup an extent buffer for tree block.
312 static int setup_temp_extent_buffer(struct extent_buffer *buf,
313 struct btrfs_mkfs_config *cfg,
314 u64 bytenr, u64 owner)
316 unsigned char fsid[BTRFS_FSID_SIZE];
317 unsigned char chunk_uuid[BTRFS_UUID_SIZE];
320 /* We rely on cfg->fs_uuid and chunk_uuid to fsid and chunk uuid */
321 BUG_ON(!cfg->fs_uuid || !cfg->chunk_uuid);
322 ret = uuid_parse(cfg->fs_uuid, fsid);
325 ret = uuid_parse(cfg->chunk_uuid, chunk_uuid);
329 memset(buf->data, 0, cfg->nodesize);
330 buf->len = cfg->nodesize;
331 btrfs_set_header_bytenr(buf, bytenr);
332 btrfs_set_header_generation(buf, 1);
333 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
334 btrfs_set_header_owner(buf, owner);
335 btrfs_set_header_flags(buf, BTRFS_HEADER_FLAG_WRITTEN);
336 write_extent_buffer(buf, chunk_uuid, btrfs_header_chunk_tree_uuid(buf),
338 write_extent_buffer(buf, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
342 static inline int write_temp_extent_buffer(int fd, struct extent_buffer *buf,
347 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
349 /* Temporary extent buffer is always mapped 1:1 on disk */
350 ret = pwrite(fd, buf->data, buf->len, bytenr);
352 ret = (ret < 0 ? ret : -EIO);
359 * Insert a root item for temporary tree root
361 * Only used in make_btrfs_v2().
363 static void insert_temp_root_item(struct extent_buffer *buf,
364 struct btrfs_mkfs_config *cfg,
365 int *slot, u32 *itemoff, u64 objectid,
368 struct btrfs_root_item root_item;
369 struct btrfs_inode_item *inode_item;
370 struct btrfs_disk_key disk_key;
372 btrfs_set_header_nritems(buf, *slot + 1);
373 (*itemoff) -= sizeof(root_item);
374 memset(&root_item, 0, sizeof(root_item));
375 inode_item = &root_item.inode;
376 btrfs_set_stack_inode_generation(inode_item, 1);
377 btrfs_set_stack_inode_size(inode_item, 3);
378 btrfs_set_stack_inode_nlink(inode_item, 1);
379 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
380 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
381 btrfs_set_root_refs(&root_item, 1);
382 btrfs_set_root_used(&root_item, cfg->nodesize);
383 btrfs_set_root_generation(&root_item, 1);
384 btrfs_set_root_bytenr(&root_item, bytenr);
386 memset(&disk_key, 0, sizeof(disk_key));
387 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
388 btrfs_set_disk_key_objectid(&disk_key, objectid);
389 btrfs_set_disk_key_offset(&disk_key, 0);
391 btrfs_set_item_key(buf, &disk_key, *slot);
392 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
393 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(root_item));
394 write_extent_buffer(buf, &root_item,
395 btrfs_item_ptr_offset(buf, *slot),
400 static int setup_temp_root_tree(int fd, struct btrfs_mkfs_config *cfg,
401 u64 root_bytenr, u64 extent_bytenr,
402 u64 dev_bytenr, u64 fs_bytenr, u64 csum_bytenr)
404 struct extent_buffer *buf = NULL;
405 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
410 * Provided bytenr must in ascending order, or tree root will have a
413 BUG_ON(!(root_bytenr < extent_bytenr && extent_bytenr < dev_bytenr &&
414 dev_bytenr < fs_bytenr && fs_bytenr < csum_bytenr));
415 buf = malloc(sizeof(*buf) + cfg->nodesize);
419 ret = setup_temp_extent_buffer(buf, cfg, root_bytenr,
420 BTRFS_ROOT_TREE_OBJECTID);
424 insert_temp_root_item(buf, cfg, &slot, &itemoff,
425 BTRFS_EXTENT_TREE_OBJECTID, extent_bytenr);
426 insert_temp_root_item(buf, cfg, &slot, &itemoff,
427 BTRFS_DEV_TREE_OBJECTID, dev_bytenr);
428 insert_temp_root_item(buf, cfg, &slot, &itemoff,
429 BTRFS_FS_TREE_OBJECTID, fs_bytenr);
430 insert_temp_root_item(buf, cfg, &slot, &itemoff,
431 BTRFS_CSUM_TREE_OBJECTID, csum_bytenr);
433 ret = write_temp_extent_buffer(fd, buf, root_bytenr);
440 * Improved version of make_btrfs().
443 * 1) Do chunk allocation to avoid used data
444 * And after this function, extent type matches chunk type
445 * 2) Better structured code
446 * No super long hand written codes to initialized all tree blocks
447 * Split into small blocks and reuse codes.
448 * TODO: Reuse tree operation facilities by introducing new flags
450 static int make_convert_btrfs(int fd, struct btrfs_mkfs_config *cfg,
451 struct btrfs_convert_context *cctx)
453 struct cache_tree *free = &cctx->free;
454 struct cache_tree *used = &cctx->used;
456 u64 meta_chunk_start;
457 /* chunk tree bytenr, in system chunk */
459 /* metadata trees bytenr, in metadata chunk */
467 /* Shouldn't happen */
468 BUG_ON(cache_tree_empty(used));
471 * reserve space for temporary superblock first
472 * Here we allocate a little larger space, to keep later
473 * free space will be STRIPE_LEN aligned
475 ret = reserve_free_space(free, BTRFS_STRIPE_LEN,
481 * Then reserve system chunk space
482 * TODO: Change system group size depending on cctx->total_bytes.
483 * If using current 4M, it can only handle less than one TB for
484 * worst case and then run out of sys space.
486 ret = reserve_free_space(free, BTRFS_MKFS_SYSTEM_GROUP_SIZE,
490 ret = reserve_free_space(free, BTRFS_CONVERT_META_GROUP_SIZE,
496 * Allocated meta/sys chunks will be mapped 1:1 with device offset.
498 * Inside the allocated metadata chunk, the layout will be:
499 * | offset | contents |
500 * -------------------------------------
502 * | +nodesize | extent root |
503 * | +nodesize * 2 | device root |
504 * | +nodesize * 3 | fs tree |
505 * | +nodesize * 4 | csum tree |
506 * -------------------------------------
507 * Inside the allocated system chunk, the layout will be:
508 * | offset | contents |
509 * -------------------------------------
510 * | +0 | chunk root |
511 * -------------------------------------
513 chunk_bytenr = sys_chunk_start;
514 root_bytenr = meta_chunk_start;
515 extent_bytenr = meta_chunk_start + cfg->nodesize;
516 dev_bytenr = meta_chunk_start + cfg->nodesize * 2;
517 fs_bytenr = meta_chunk_start + cfg->nodesize * 3;
518 csum_bytenr = meta_chunk_start + cfg->nodesize * 4;
520 ret = setup_temp_super(fd, cfg, root_bytenr, chunk_bytenr);
524 ret = setup_temp_root_tree(fd, cfg, root_bytenr, extent_bytenr,
525 dev_bytenr, fs_bytenr, csum_bytenr);
534 * @fs_uuid - if NULL, generates a UUID, returns back the new filesystem UUID
536 int make_btrfs(int fd, struct btrfs_mkfs_config *cfg,
537 struct btrfs_convert_context *cctx)
539 struct btrfs_super_block super;
540 struct extent_buffer *buf;
541 struct btrfs_root_item root_item;
542 struct btrfs_disk_key disk_key;
543 struct btrfs_extent_item *extent_item;
544 struct btrfs_inode_item *inode_item;
545 struct btrfs_chunk *chunk;
546 struct btrfs_dev_item *dev_item;
547 struct btrfs_dev_extent *dev_extent;
548 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
558 int skinny_metadata = !!(cfg->features &
559 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA);
563 return make_convert_btrfs(fd, cfg, cctx);
564 buf = malloc(sizeof(*buf) + max(cfg->sectorsize, cfg->nodesize));
568 first_free = BTRFS_SUPER_INFO_OFFSET + cfg->sectorsize * 2 - 1;
569 first_free &= ~((u64)cfg->sectorsize - 1);
571 memset(&super, 0, sizeof(super));
573 num_bytes = (cfg->num_bytes / cfg->sectorsize) * cfg->sectorsize;
574 if (cfg->fs_uuid && *cfg->fs_uuid) {
575 if (uuid_parse(cfg->fs_uuid, super.fsid) != 0) {
576 error("cannot not parse UUID: %s", cfg->fs_uuid);
580 if (!test_uuid_unique(cfg->fs_uuid)) {
581 error("non-unique UUID: %s", cfg->fs_uuid);
586 uuid_generate(super.fsid);
588 uuid_unparse(super.fsid, cfg->fs_uuid);
590 uuid_generate(super.dev_item.uuid);
591 uuid_generate(chunk_tree_uuid);
593 btrfs_set_super_bytenr(&super, cfg->blocks[0]);
594 btrfs_set_super_num_devices(&super, 1);
595 btrfs_set_super_magic(&super, BTRFS_MAGIC);
596 btrfs_set_super_generation(&super, 1);
597 btrfs_set_super_root(&super, cfg->blocks[1]);
598 btrfs_set_super_chunk_root(&super, cfg->blocks[3]);
599 btrfs_set_super_total_bytes(&super, num_bytes);
600 btrfs_set_super_bytes_used(&super, 6 * cfg->nodesize);
601 btrfs_set_super_sectorsize(&super, cfg->sectorsize);
602 btrfs_set_super_leafsize(&super, cfg->nodesize);
603 btrfs_set_super_nodesize(&super, cfg->nodesize);
604 btrfs_set_super_stripesize(&super, cfg->stripesize);
605 btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
606 btrfs_set_super_chunk_root_generation(&super, 1);
607 btrfs_set_super_cache_generation(&super, -1);
608 btrfs_set_super_incompat_flags(&super, cfg->features);
610 __strncpy_null(super.label, cfg->label, BTRFS_LABEL_SIZE - 1);
612 /* create the tree of root objects */
613 memset(buf->data, 0, cfg->nodesize);
614 buf->len = cfg->nodesize;
615 btrfs_set_header_bytenr(buf, cfg->blocks[1]);
616 btrfs_set_header_nritems(buf, 4);
617 btrfs_set_header_generation(buf, 1);
618 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
619 btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
620 write_extent_buffer(buf, super.fsid, btrfs_header_fsid(),
623 write_extent_buffer(buf, chunk_tree_uuid,
624 btrfs_header_chunk_tree_uuid(buf),
627 /* create the items for the root tree */
628 memset(&root_item, 0, sizeof(root_item));
629 inode_item = &root_item.inode;
630 btrfs_set_stack_inode_generation(inode_item, 1);
631 btrfs_set_stack_inode_size(inode_item, 3);
632 btrfs_set_stack_inode_nlink(inode_item, 1);
633 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
634 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
635 btrfs_set_root_refs(&root_item, 1);
636 btrfs_set_root_used(&root_item, cfg->nodesize);
637 btrfs_set_root_generation(&root_item, 1);
639 memset(&disk_key, 0, sizeof(disk_key));
640 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
641 btrfs_set_disk_key_offset(&disk_key, 0);
644 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - sizeof(root_item);
645 btrfs_set_root_bytenr(&root_item, cfg->blocks[2]);
646 btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
647 btrfs_set_item_key(buf, &disk_key, nritems);
648 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
649 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
651 write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
652 nritems), sizeof(root_item));
655 itemoff = itemoff - sizeof(root_item);
656 btrfs_set_root_bytenr(&root_item, cfg->blocks[4]);
657 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
658 btrfs_set_item_key(buf, &disk_key, nritems);
659 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
660 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
662 write_extent_buffer(buf, &root_item,
663 btrfs_item_ptr_offset(buf, nritems),
667 itemoff = itemoff - sizeof(root_item);
668 btrfs_set_root_bytenr(&root_item, cfg->blocks[5]);
669 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
670 btrfs_set_item_key(buf, &disk_key, nritems);
671 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
672 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
674 write_extent_buffer(buf, &root_item,
675 btrfs_item_ptr_offset(buf, nritems),
679 itemoff = itemoff - sizeof(root_item);
680 btrfs_set_root_bytenr(&root_item, cfg->blocks[6]);
681 btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_TREE_OBJECTID);
682 btrfs_set_item_key(buf, &disk_key, nritems);
683 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
684 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
686 write_extent_buffer(buf, &root_item,
687 btrfs_item_ptr_offset(buf, nritems),
692 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
693 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[1]);
694 if (ret != cfg->nodesize) {
695 ret = (ret < 0 ? -errno : -EIO);
699 /* create the items for the extent tree */
700 memset(buf->data + sizeof(struct btrfs_header), 0,
701 cfg->nodesize - sizeof(struct btrfs_header));
703 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
704 for (i = 1; i < 7; i++) {
705 item_size = sizeof(struct btrfs_extent_item);
706 if (!skinny_metadata)
707 item_size += sizeof(struct btrfs_tree_block_info);
709 BUG_ON(cfg->blocks[i] < first_free);
710 BUG_ON(cfg->blocks[i] < cfg->blocks[i - 1]);
712 /* create extent item */
713 itemoff -= item_size;
714 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
715 if (skinny_metadata) {
716 btrfs_set_disk_key_type(&disk_key,
717 BTRFS_METADATA_ITEM_KEY);
718 btrfs_set_disk_key_offset(&disk_key, 0);
720 btrfs_set_disk_key_type(&disk_key,
721 BTRFS_EXTENT_ITEM_KEY);
722 btrfs_set_disk_key_offset(&disk_key, cfg->nodesize);
724 btrfs_set_item_key(buf, &disk_key, nritems);
725 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
727 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
729 extent_item = btrfs_item_ptr(buf, nritems,
730 struct btrfs_extent_item);
731 btrfs_set_extent_refs(buf, extent_item, 1);
732 btrfs_set_extent_generation(buf, extent_item, 1);
733 btrfs_set_extent_flags(buf, extent_item,
734 BTRFS_EXTENT_FLAG_TREE_BLOCK);
737 /* create extent ref */
738 ref_root = reference_root_table[i];
739 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
740 btrfs_set_disk_key_offset(&disk_key, ref_root);
741 btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
742 btrfs_set_item_key(buf, &disk_key, nritems);
743 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
745 btrfs_set_item_size(buf, btrfs_item_nr(nritems), 0);
748 btrfs_set_header_bytenr(buf, cfg->blocks[2]);
749 btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
750 btrfs_set_header_nritems(buf, nritems);
751 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
752 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[2]);
753 if (ret != cfg->nodesize) {
754 ret = (ret < 0 ? -errno : -EIO);
758 /* create the chunk tree */
759 memset(buf->data + sizeof(struct btrfs_header), 0,
760 cfg->nodesize - sizeof(struct btrfs_header));
762 item_size = sizeof(*dev_item);
763 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - item_size;
765 /* first device 1 (there is no device 0) */
766 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
767 btrfs_set_disk_key_offset(&disk_key, 1);
768 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
769 btrfs_set_item_key(buf, &disk_key, nritems);
770 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
771 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
773 dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
774 btrfs_set_device_id(buf, dev_item, 1);
775 btrfs_set_device_generation(buf, dev_item, 0);
776 btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
777 btrfs_set_device_bytes_used(buf, dev_item,
778 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
779 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
780 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
781 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
782 btrfs_set_device_type(buf, dev_item, 0);
784 write_extent_buffer(buf, super.dev_item.uuid,
785 (unsigned long)btrfs_device_uuid(dev_item),
787 write_extent_buffer(buf, super.fsid,
788 (unsigned long)btrfs_device_fsid(dev_item),
790 read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
794 item_size = btrfs_chunk_item_size(1);
795 itemoff = itemoff - item_size;
797 /* then we have chunk 0 */
798 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
799 btrfs_set_disk_key_offset(&disk_key, 0);
800 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
801 btrfs_set_item_key(buf, &disk_key, nritems);
802 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
803 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
805 chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
806 btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
807 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
808 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
809 btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
810 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
811 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
812 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
813 btrfs_set_chunk_num_stripes(buf, chunk, 1);
814 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
815 btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
818 write_extent_buffer(buf, super.dev_item.uuid,
819 (unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
822 /* copy the key for the chunk to the system array */
823 ptr = super.sys_chunk_array;
824 array_size = sizeof(disk_key);
826 memcpy(ptr, &disk_key, sizeof(disk_key));
827 ptr += sizeof(disk_key);
829 /* copy the chunk to the system array */
830 read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
831 array_size += item_size;
833 btrfs_set_super_sys_array_size(&super, array_size);
835 btrfs_set_header_bytenr(buf, cfg->blocks[3]);
836 btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
837 btrfs_set_header_nritems(buf, nritems);
838 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
839 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[3]);
840 if (ret != cfg->nodesize) {
841 ret = (ret < 0 ? -errno : -EIO);
845 /* create the device tree */
846 memset(buf->data + sizeof(struct btrfs_header), 0,
847 cfg->nodesize - sizeof(struct btrfs_header));
849 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) -
850 sizeof(struct btrfs_dev_extent);
852 btrfs_set_disk_key_objectid(&disk_key, 1);
853 btrfs_set_disk_key_offset(&disk_key, 0);
854 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
855 btrfs_set_item_key(buf, &disk_key, nritems);
856 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
857 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
858 sizeof(struct btrfs_dev_extent));
859 dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
860 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
861 BTRFS_CHUNK_TREE_OBJECTID);
862 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
863 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
864 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
866 write_extent_buffer(buf, chunk_tree_uuid,
867 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
870 btrfs_set_dev_extent_length(buf, dev_extent,
871 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
874 btrfs_set_header_bytenr(buf, cfg->blocks[4]);
875 btrfs_set_header_owner(buf, BTRFS_DEV_TREE_OBJECTID);
876 btrfs_set_header_nritems(buf, nritems);
877 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
878 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[4]);
879 if (ret != cfg->nodesize) {
880 ret = (ret < 0 ? -errno : -EIO);
884 /* create the FS root */
885 memset(buf->data + sizeof(struct btrfs_header), 0,
886 cfg->nodesize - sizeof(struct btrfs_header));
887 btrfs_set_header_bytenr(buf, cfg->blocks[5]);
888 btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
889 btrfs_set_header_nritems(buf, 0);
890 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
891 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[5]);
892 if (ret != cfg->nodesize) {
893 ret = (ret < 0 ? -errno : -EIO);
896 /* finally create the csum root */
897 memset(buf->data + sizeof(struct btrfs_header), 0,
898 cfg->nodesize - sizeof(struct btrfs_header));
899 btrfs_set_header_bytenr(buf, cfg->blocks[6]);
900 btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
901 btrfs_set_header_nritems(buf, 0);
902 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
903 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[6]);
904 if (ret != cfg->nodesize) {
905 ret = (ret < 0 ? -errno : -EIO);
909 /* and write out the super block */
910 BUG_ON(sizeof(super) > cfg->sectorsize);
911 memset(buf->data, 0, BTRFS_SUPER_INFO_SIZE);
912 memcpy(buf->data, &super, sizeof(super));
913 buf->len = BTRFS_SUPER_INFO_SIZE;
914 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
915 ret = pwrite(fd, buf->data, BTRFS_SUPER_INFO_SIZE, cfg->blocks[0]);
916 if (ret != BTRFS_SUPER_INFO_SIZE) {
917 ret = (ret < 0 ? -errno : -EIO);
928 static const struct btrfs_fs_feature {
932 } mkfs_features[] = {
933 { "mixed-bg", BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS,
934 "mixed data and metadata block groups" },
935 { "extref", BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF,
936 "increased hardlink limit per file to 65536" },
937 { "raid56", BTRFS_FEATURE_INCOMPAT_RAID56,
938 "raid56 extended format" },
939 { "skinny-metadata", BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA,
940 "reduced-size metadata extent refs" },
941 { "no-holes", BTRFS_FEATURE_INCOMPAT_NO_HOLES,
942 "no explicit hole extents for files" },
943 /* Keep this one last */
944 { "list-all", BTRFS_FEATURE_LIST_ALL, NULL }
947 static int parse_one_fs_feature(const char *name, u64 *flags)
952 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
953 if (name[0] == '^' &&
954 !strcmp(mkfs_features[i].name, name + 1)) {
955 *flags &= ~ mkfs_features[i].flag;
957 } else if (!strcmp(mkfs_features[i].name, name)) {
958 *flags |= mkfs_features[i].flag;
966 void btrfs_parse_features_to_string(char *buf, u64 flags)
972 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
973 if (flags & mkfs_features[i].flag) {
976 strcat(buf, mkfs_features[i].name);
981 void btrfs_process_fs_features(u64 flags)
985 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
986 if (flags & mkfs_features[i].flag) {
987 printf("Turning ON incompat feature '%s': %s\n",
988 mkfs_features[i].name,
989 mkfs_features[i].desc);
994 void btrfs_list_all_fs_features(u64 mask_disallowed)
998 fprintf(stderr, "Filesystem features available:\n");
999 for (i = 0; i < ARRAY_SIZE(mkfs_features) - 1; i++) {
1000 char *is_default = "";
1002 if (mkfs_features[i].flag & mask_disallowed)
1004 if (mkfs_features[i].flag & BTRFS_MKFS_DEFAULT_FEATURES)
1005 is_default = ", default";
1006 fprintf(stderr, "%-20s- %s (0x%llx%s)\n",
1007 mkfs_features[i].name,
1008 mkfs_features[i].desc,
1009 mkfs_features[i].flag,
1015 * Return NULL if all features were parsed fine, otherwise return the name of
1016 * the first unparsed.
1018 char* btrfs_parse_fs_features(char *namelist, u64 *flags)
1021 char *save_ptr = NULL; /* Satisfy static checkers */
1023 for (this_char = strtok_r(namelist, ",", &save_ptr);
1025 this_char = strtok_r(NULL, ",", &save_ptr)) {
1026 if (parse_one_fs_feature(this_char, flags))
1033 u64 btrfs_device_size(int fd, struct stat *st)
1036 if (S_ISREG(st->st_mode)) {
1039 if (!S_ISBLK(st->st_mode)) {
1042 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
1048 static int zero_blocks(int fd, off_t start, size_t len)
1050 char *buf = malloc(len);
1056 memset(buf, 0, len);
1057 written = pwrite(fd, buf, len, start);
1064 #define ZERO_DEV_BYTES (2 * 1024 * 1024)
1066 /* don't write outside the device by clamping the region to the device size */
1067 static int zero_dev_clamped(int fd, off_t start, ssize_t len, u64 dev_size)
1069 off_t end = max(start, start + len);
1072 /* and don't overwrite the disk labels on sparc */
1073 start = max(start, 1024);
1074 end = max(end, 1024);
1077 start = min_t(u64, start, dev_size);
1078 end = min_t(u64, end, dev_size);
1080 return zero_blocks(fd, start, end - start);
1083 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
1084 struct btrfs_root *root, int fd, char *path,
1085 u64 device_total_bytes, u32 io_width, u32 io_align,
1088 struct btrfs_super_block *disk_super;
1089 struct btrfs_super_block *super = root->fs_info->super_copy;
1090 struct btrfs_device *device;
1091 struct btrfs_dev_item *dev_item;
1097 device_total_bytes = (device_total_bytes / sectorsize) * sectorsize;
1099 device = kzalloc(sizeof(*device), GFP_NOFS);
1102 buf = kzalloc(sectorsize, GFP_NOFS);
1105 BUG_ON(sizeof(*disk_super) > sectorsize);
1107 disk_super = (struct btrfs_super_block *)buf;
1108 dev_item = &disk_super->dev_item;
1110 uuid_generate(device->uuid);
1113 device->io_width = io_width;
1114 device->io_align = io_align;
1115 device->sector_size = sectorsize;
1117 device->writeable = 1;
1118 device->total_bytes = device_total_bytes;
1119 device->bytes_used = 0;
1120 device->total_ios = 0;
1121 device->dev_root = root->fs_info->dev_root;
1122 device->name = strdup(path);
1126 INIT_LIST_HEAD(&device->dev_list);
1127 ret = btrfs_add_device(trans, root, device);
1130 fs_total_bytes = btrfs_super_total_bytes(super) + device_total_bytes;
1131 btrfs_set_super_total_bytes(super, fs_total_bytes);
1133 num_devs = btrfs_super_num_devices(super) + 1;
1134 btrfs_set_super_num_devices(super, num_devs);
1136 memcpy(disk_super, super, sizeof(*disk_super));
1138 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
1139 btrfs_set_stack_device_id(dev_item, device->devid);
1140 btrfs_set_stack_device_type(dev_item, device->type);
1141 btrfs_set_stack_device_io_align(dev_item, device->io_align);
1142 btrfs_set_stack_device_io_width(dev_item, device->io_width);
1143 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
1144 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
1145 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
1146 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
1148 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
1149 BUG_ON(ret != sectorsize);
1152 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
1153 device->fs_devices = root->fs_info->fs_devices;
1162 static int btrfs_wipe_existing_sb(int fd)
1164 const char *off = NULL;
1169 blkid_probe pr = NULL;
1171 pr = blkid_new_probe();
1175 if (blkid_probe_set_device(pr, fd, 0, 0)) {
1180 ret = blkid_probe_lookup_value(pr, "SBMAGIC_OFFSET", &off, NULL);
1182 ret = blkid_probe_lookup_value(pr, "SBMAGIC", NULL, &len);
1184 if (ret || len == 0 || off == NULL) {
1186 * If lookup fails, the probe did not find any values, eg. for
1187 * a file image or a loop device. Soft error.
1193 offset = strtoll(off, NULL, 10);
1194 if (len > sizeof(buf))
1197 memset(buf, 0, len);
1198 ret = pwrite(fd, buf, len, offset);
1200 error("cannot wipe existing superblock: %s", strerror(errno));
1202 } else if (ret != len) {
1203 error("cannot wipe existing superblock: wrote %d of %zd", ret, len);
1209 blkid_free_probe(pr);
1213 int btrfs_prepare_device(int fd, const char *file, int zero_end,
1214 u64 *block_count_ret, u64 max_block_count, int discard)
1220 ret = fstat(fd, &st);
1222 error("unable to stat %s: %s", file, strerror(errno));
1226 block_count = btrfs_device_size(fd, &st);
1227 if (block_count == 0) {
1228 error("unable to determine size of %s", file);
1231 if (max_block_count)
1232 block_count = min(block_count, max_block_count);
1236 * We intentionally ignore errors from the discard ioctl. It
1237 * is not necessary for the mkfs functionality but just an
1240 if (discard_range(fd, 0, 0) == 0) {
1241 printf("Performing full device TRIM (%s) ...\n",
1242 pretty_size(block_count));
1243 discard_blocks(fd, 0, block_count);
1247 ret = zero_dev_clamped(fd, 0, ZERO_DEV_BYTES, block_count);
1248 for (i = 0 ; !ret && i < BTRFS_SUPER_MIRROR_MAX; i++)
1249 ret = zero_dev_clamped(fd, btrfs_sb_offset(i),
1250 BTRFS_SUPER_INFO_SIZE, block_count);
1251 if (!ret && zero_end)
1252 ret = zero_dev_clamped(fd, block_count - ZERO_DEV_BYTES,
1253 ZERO_DEV_BYTES, block_count);
1256 error("failed to zero device '%s': %s", file, strerror(-ret));
1260 ret = btrfs_wipe_existing_sb(fd);
1262 error("cannot wipe superblocks on %s", file);
1266 *block_count_ret = block_count;
1270 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
1271 struct btrfs_root *root, u64 objectid)
1274 struct btrfs_inode_item inode_item;
1275 time_t now = time(NULL);
1277 memset(&inode_item, 0, sizeof(inode_item));
1278 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
1279 btrfs_set_stack_inode_size(&inode_item, 0);
1280 btrfs_set_stack_inode_nlink(&inode_item, 1);
1281 btrfs_set_stack_inode_nbytes(&inode_item, root->nodesize);
1282 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
1283 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
1284 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
1285 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
1286 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
1287 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
1288 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
1289 btrfs_set_stack_timespec_sec(&inode_item.otime, 0);
1290 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
1292 if (root->fs_info->tree_root == root)
1293 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
1295 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
1299 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
1303 btrfs_set_root_dirid(&root->root_item, objectid);
1310 * checks if a path is a block device node
1311 * Returns negative errno on failure, otherwise
1312 * returns 1 for blockdev, 0 for not-blockdev
1314 int is_block_device(const char *path)
1316 struct stat statbuf;
1318 if (stat(path, &statbuf) < 0)
1321 return !!S_ISBLK(statbuf.st_mode);
1325 * check if given path is a mount point
1326 * return 1 if yes. 0 if no. -1 for error
1328 int is_mount_point(const char *path)
1334 f = setmntent("/proc/self/mounts", "r");
1338 while ((mnt = getmntent(f)) != NULL) {
1339 if (strcmp(mnt->mnt_dir, path))
1348 static int is_reg_file(const char *path)
1350 struct stat statbuf;
1352 if (stat(path, &statbuf) < 0)
1354 return S_ISREG(statbuf.st_mode);
1358 * This function checks if the given input parameter is
1360 * return <0 : some error in the given input
1361 * return BTRFS_ARG_UNKNOWN: unknown input
1362 * return BTRFS_ARG_UUID: given input is uuid
1363 * return BTRFS_ARG_MNTPOINT: given input is path
1364 * return BTRFS_ARG_REG: given input is regular file
1365 * return BTRFS_ARG_BLKDEV: given input is block device
1367 int check_arg_type(const char *input)
1370 char path[PATH_MAX];
1375 if (realpath(input, path)) {
1376 if (is_block_device(path) == 1)
1377 return BTRFS_ARG_BLKDEV;
1379 if (is_mount_point(path) == 1)
1380 return BTRFS_ARG_MNTPOINT;
1382 if (is_reg_file(path))
1383 return BTRFS_ARG_REG;
1385 return BTRFS_ARG_UNKNOWN;
1388 if (strlen(input) == (BTRFS_UUID_UNPARSED_SIZE - 1) &&
1389 !uuid_parse(input, uuid))
1390 return BTRFS_ARG_UUID;
1392 return BTRFS_ARG_UNKNOWN;
1396 * Find the mount point for a mounted device.
1397 * On success, returns 0 with mountpoint in *mp.
1398 * On failure, returns -errno (not mounted yields -EINVAL)
1399 * Is noisy on failures, expects to be given a mounted device.
1401 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size)
1406 ret = is_block_device(dev);
1409 error("not a block device: %s", dev);
1412 error("cannot check %s: %s", dev, strerror(-ret));
1417 fd = open(dev, O_RDONLY);
1420 error("cannot open %s: %s", dev, strerror(errno));
1424 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
1427 } else { /* mounted, all good */
1437 * Given a pathname, return a filehandle to:
1438 * the original pathname or,
1439 * if the pathname is a mounted btrfs device, to its mountpoint.
1441 * On error, return -1, errno should be set.
1443 int open_path_or_dev_mnt(const char *path, DIR **dirstream, int verbose)
1448 if (is_block_device(path)) {
1449 ret = get_btrfs_mount(path, mp, sizeof(mp));
1451 /* not a mounted btrfs dev */
1452 error_on(verbose, "'%s' is not a mounted btrfs device",
1457 ret = open_file_or_dir(mp, dirstream);
1458 error_on(verbose && ret < 0, "can't access '%s': %s",
1459 path, strerror(errno));
1461 ret = btrfs_open_dir(path, dirstream, 1);
1468 * Do the following checks before calling open_file_or_dir():
1469 * 1: path is in a btrfs filesystem
1470 * 2: path is a directory
1472 int btrfs_open_dir(const char *path, DIR **dirstream, int verbose)
1478 if (statfs(path, &stfs) != 0) {
1479 error_on(verbose, "cannot access '%s': %s", path,
1484 if (stfs.f_type != BTRFS_SUPER_MAGIC) {
1485 error_on(verbose, "not a btrfs filesystem: %s", path);
1489 if (stat(path, &st) != 0) {
1490 error_on(verbose, "cannot access '%s': %s", path,
1495 if (!S_ISDIR(st.st_mode)) {
1496 error_on(verbose, "not a directory: %s", path);
1500 ret = open_file_or_dir(path, dirstream);
1502 error_on(verbose, "cannot access '%s': %s", path,
1509 /* checks if a device is a loop device */
1510 static int is_loop_device (const char* device) {
1511 struct stat statbuf;
1513 if(stat(device, &statbuf) < 0)
1516 return (S_ISBLK(statbuf.st_mode) &&
1517 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
1521 * Takes a loop device path (e.g. /dev/loop0) and returns
1522 * the associated file (e.g. /images/my_btrfs.img) using
1525 static int resolve_loop_device_with_loopdev(const char* loop_dev, char* loop_file)
1529 struct loop_info64 lo64;
1531 fd = open(loop_dev, O_RDONLY | O_NONBLOCK);
1534 ret = ioctl(fd, LOOP_GET_STATUS64, &lo64);
1540 memcpy(loop_file, lo64.lo_file_name, sizeof(lo64.lo_file_name));
1541 loop_file[sizeof(lo64.lo_file_name)] = 0;
1549 /* Takes a loop device path (e.g. /dev/loop0) and returns
1550 * the associated file (e.g. /images/my_btrfs.img) */
1551 static int resolve_loop_device(const char* loop_dev, char* loop_file,
1558 char real_loop_dev[PATH_MAX];
1560 if (!realpath(loop_dev, real_loop_dev))
1562 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
1563 if (!(f = fopen(p, "r"))) {
1564 if (errno == ENOENT)
1566 * It's possibly a partitioned loop device, which is
1567 * resolvable with loopdev API.
1569 return resolve_loop_device_with_loopdev(loop_dev, loop_file);
1573 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
1574 ret = fscanf(f, fmt, loop_file);
1583 * Checks whether a and b are identical or device
1584 * files associated with the same block device
1586 static int is_same_blk_file(const char* a, const char* b)
1588 struct stat st_buf_a, st_buf_b;
1589 char real_a[PATH_MAX];
1590 char real_b[PATH_MAX];
1592 if (!realpath(a, real_a))
1593 strncpy_null(real_a, a);
1595 if (!realpath(b, real_b))
1596 strncpy_null(real_b, b);
1598 /* Identical path? */
1599 if (strcmp(real_a, real_b) == 0)
1602 if (stat(a, &st_buf_a) < 0 || stat(b, &st_buf_b) < 0) {
1603 if (errno == ENOENT)
1608 /* Same blockdevice? */
1609 if (S_ISBLK(st_buf_a.st_mode) && S_ISBLK(st_buf_b.st_mode) &&
1610 st_buf_a.st_rdev == st_buf_b.st_rdev) {
1615 if (st_buf_a.st_dev == st_buf_b.st_dev &&
1616 st_buf_a.st_ino == st_buf_b.st_ino) {
1623 /* checks if a and b are identical or device
1624 * files associated with the same block device or
1625 * if one file is a loop device that uses the other
1628 static int is_same_loop_file(const char* a, const char* b)
1630 char res_a[PATH_MAX];
1631 char res_b[PATH_MAX];
1632 const char* final_a = NULL;
1633 const char* final_b = NULL;
1636 /* Resolve a if it is a loop device */
1637 if((ret = is_loop_device(a)) < 0) {
1642 ret = resolve_loop_device(a, res_a, sizeof(res_a));
1653 /* Resolve b if it is a loop device */
1654 if ((ret = is_loop_device(b)) < 0) {
1659 ret = resolve_loop_device(b, res_b, sizeof(res_b));
1670 return is_same_blk_file(final_a, final_b);
1673 /* Checks if a file exists and is a block or regular file*/
1674 static int is_existing_blk_or_reg_file(const char* filename)
1678 if(stat(filename, &st_buf) < 0) {
1685 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
1688 /* Checks if a file is used (directly or indirectly via a loop device)
1689 * by a device in fs_devices
1691 static int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices,
1695 struct list_head *head;
1696 struct list_head *cur;
1697 struct btrfs_device *device;
1699 head = &fs_devices->devices;
1700 list_for_each(cur, head) {
1701 device = list_entry(cur, struct btrfs_device, dev_list);
1703 if((ret = is_same_loop_file(device->name, file)))
1711 * Resolve a pathname to a device mapper node to /dev/mapper/<name>
1712 * Returns NULL on invalid input or malloc failure; Other failures
1713 * will be handled by the caller using the input pathame.
1715 char *canonicalize_dm_name(const char *ptname)
1719 char path[PATH_MAX], name[PATH_MAX], *res = NULL;
1721 if (!ptname || !*ptname)
1724 snprintf(path, sizeof(path), "/sys/block/%s/dm/name", ptname);
1725 if (!(f = fopen(path, "r")))
1728 /* read <name>\n from sysfs */
1729 if (fgets(name, sizeof(name), f) && (sz = strlen(name)) > 1) {
1730 name[sz - 1] = '\0';
1731 snprintf(path, sizeof(path), "/dev/mapper/%s", name);
1733 if (access(path, F_OK) == 0)
1741 * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or
1742 * to a device mapper pathname.
1743 * Returns NULL on invalid input or malloc failure; Other failures
1744 * will be handled by the caller using the input pathame.
1746 char *canonicalize_path(const char *path)
1748 char *canonical, *p;
1750 if (!path || !*path)
1753 canonical = realpath(path, NULL);
1755 return strdup(path);
1756 p = strrchr(canonical, '/');
1757 if (p && strncmp(p, "/dm-", 4) == 0 && isdigit(*(p + 4))) {
1758 char *dm = canonicalize_dm_name(p + 1);
1769 * returns 1 if the device was mounted, < 0 on error or 0 if everything
1770 * is safe to continue.
1772 int check_mounted(const char* file)
1777 fd = open(file, O_RDONLY);
1779 error("mount check: cannot open %s: %s", file,
1784 ret = check_mounted_where(fd, file, NULL, 0, NULL);
1790 int check_mounted_where(int fd, const char *file, char *where, int size,
1791 struct btrfs_fs_devices **fs_dev_ret)
1796 struct btrfs_fs_devices *fs_devices_mnt = NULL;
1800 /* scan the initial device */
1801 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
1802 &total_devs, BTRFS_SUPER_INFO_OFFSET, 0);
1803 is_btrfs = (ret >= 0);
1805 /* scan other devices */
1806 if (is_btrfs && total_devs > 1) {
1807 ret = btrfs_scan_lblkid();
1812 /* iterate over the list of currently mounted filesystems */
1813 if ((f = setmntent ("/proc/self/mounts", "r")) == NULL)
1816 while ((mnt = getmntent (f)) != NULL) {
1818 if(strcmp(mnt->mnt_type, "btrfs") != 0)
1821 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
1823 /* ignore entries in the mount table that are not
1824 associated with a file*/
1825 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
1826 goto out_mntloop_err;
1830 ret = is_same_loop_file(file, mnt->mnt_fsname);
1834 goto out_mntloop_err;
1839 /* Did we find an entry in mnt table? */
1840 if (mnt && size && where) {
1841 strncpy(where, mnt->mnt_dir, size);
1845 *fs_dev_ret = fs_devices_mnt;
1847 ret = (mnt != NULL);
1855 struct pending_dir {
1856 struct list_head list;
1857 char name[PATH_MAX];
1860 int btrfs_register_one_device(const char *fname)
1862 struct btrfs_ioctl_vol_args args;
1866 fd = open("/dev/btrfs-control", O_RDWR);
1869 "failed to open /dev/btrfs-control, skipping device registration: %s",
1873 memset(&args, 0, sizeof(args));
1874 strncpy_null(args.name, fname);
1875 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
1877 error("device scan failed on '%s': %s", fname,
1886 * Register all devices in the fs_uuid list created in the user
1887 * space. Ensure btrfs_scan_lblkid() is called before this func.
1889 int btrfs_register_all_devices(void)
1893 struct btrfs_fs_devices *fs_devices;
1894 struct btrfs_device *device;
1895 struct list_head *all_uuids;
1897 all_uuids = btrfs_scanned_uuids();
1899 list_for_each_entry(fs_devices, all_uuids, list) {
1900 list_for_each_entry(device, &fs_devices->devices, dev_list) {
1902 err = btrfs_register_one_device(device->name);
1912 int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
1915 struct btrfs_super_block *disk_super;
1919 buf = malloc(BTRFS_SUPER_INFO_SIZE);
1924 ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
1925 if (ret != BTRFS_SUPER_INFO_SIZE)
1929 disk_super = (struct btrfs_super_block *)buf;
1930 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC)
1933 if (!memcmp(disk_super->fsid, root->fs_info->super_copy->fsid,
1943 * Note: this function uses a static per-thread buffer. Do not call this
1944 * function more than 10 times within one argument list!
1946 const char *pretty_size_mode(u64 size, unsigned mode)
1948 static __thread int ps_index = 0;
1949 static __thread char ps_array[10][32];
1952 ret = ps_array[ps_index];
1955 (void)pretty_size_snprintf(size, ret, 32, mode);
1960 static const char* unit_suffix_binary[] =
1961 { "B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB"};
1962 static const char* unit_suffix_decimal[] =
1963 { "B", "kB", "MB", "GB", "TB", "PB", "EB"};
1965 int pretty_size_snprintf(u64 size, char *str, size_t str_size, unsigned unit_mode)
1971 const char** suffix = NULL;
1977 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_RAW) {
1978 snprintf(str, str_size, "%llu", size);
1982 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_BINARY) {
1985 suffix = unit_suffix_binary;
1986 } else if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_DECIMAL) {
1989 suffix = unit_suffix_decimal;
1994 fprintf(stderr, "INTERNAL ERROR: unknown unit base, mode %d\n",
2002 switch (unit_mode & UNITS_MODE_MASK) {
2003 case UNITS_TBYTES: base *= mult; num_divs++;
2004 case UNITS_GBYTES: base *= mult; num_divs++;
2005 case UNITS_MBYTES: base *= mult; num_divs++;
2006 case UNITS_KBYTES: num_divs++;
2013 while (size >= mult) {
2019 * If the value is smaller than base, we didn't do any
2020 * division, in that case, base should be 1, not original
2021 * base, or the unit will be wrong
2027 if (num_divs >= ARRAY_SIZE(unit_suffix_binary)) {
2029 printf("INTERNAL ERROR: unsupported unit suffix, index %d\n",
2034 fraction = (float)last_size / base;
2036 return snprintf(str, str_size, "%.2f%s", fraction, suffix[num_divs]);
2040 * __strncpy_null - strncpy with null termination
2041 * @dest: the target array
2042 * @src: the source string
2043 * @n: maximum bytes to copy (size of *dest)
2045 * Like strncpy, but ensures destination is null-terminated.
2047 * Copies the string pointed to by src, including the terminating null
2048 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
2049 * of n bytes. Then ensure that dest is null-terminated.
2051 char *__strncpy_null(char *dest, const char *src, size_t n)
2053 strncpy(dest, src, n);
2060 * Checks to make sure that the label matches our requirements.
2062 0 if everything is safe and usable
2063 -1 if the label is too long
2065 static int check_label(const char *input)
2067 int len = strlen(input);
2069 if (len > BTRFS_LABEL_SIZE - 1) {
2070 error("label %s is too long (max %d)", input,
2071 BTRFS_LABEL_SIZE - 1);
2078 static int set_label_unmounted(const char *dev, const char *label)
2080 struct btrfs_trans_handle *trans;
2081 struct btrfs_root *root;
2084 ret = check_mounted(dev);
2086 error("checking mount status of %s failed: %d", dev, ret);
2090 error("device %s is mounted, use mount point", dev);
2094 /* Open the super_block at the default location
2095 * and as read-write.
2097 root = open_ctree(dev, 0, OPEN_CTREE_WRITES);
2098 if (!root) /* errors are printed by open_ctree() */
2101 trans = btrfs_start_transaction(root, 1);
2102 __strncpy_null(root->fs_info->super_copy->label, label, BTRFS_LABEL_SIZE - 1);
2104 btrfs_commit_transaction(trans, root);
2106 /* Now we close it since we are done. */
2111 static int set_label_mounted(const char *mount_path, const char *labelp)
2114 char label[BTRFS_LABEL_SIZE];
2116 fd = open(mount_path, O_RDONLY | O_NOATIME);
2118 error("unable to access %s: %s", mount_path, strerror(errno));
2122 memset(label, 0, sizeof(label));
2123 __strncpy_null(label, labelp, BTRFS_LABEL_SIZE - 1);
2124 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
2125 error("unable to set label of %s: %s", mount_path,
2135 int get_label_unmounted(const char *dev, char *label)
2137 struct btrfs_root *root;
2140 ret = check_mounted(dev);
2142 error("checking mount status of %s failed: %d", dev, ret);
2146 /* Open the super_block at the default location
2149 root = open_ctree(dev, 0, 0);
2153 __strncpy_null(label, root->fs_info->super_copy->label,
2154 BTRFS_LABEL_SIZE - 1);
2156 /* Now we close it since we are done. */
2162 * If a partition is mounted, try to get the filesystem label via its
2163 * mounted path rather than device. Return the corresponding error
2164 * the user specified the device path.
2166 int get_label_mounted(const char *mount_path, char *labelp)
2168 char label[BTRFS_LABEL_SIZE];
2172 fd = open(mount_path, O_RDONLY | O_NOATIME);
2174 error("unable to access %s: %s", mount_path, strerror(errno));
2178 memset(label, '\0', sizeof(label));
2179 ret = ioctl(fd, BTRFS_IOC_GET_FSLABEL, label);
2181 if (errno != ENOTTY)
2182 error("unable to get label of %s: %s", mount_path,
2189 __strncpy_null(labelp, label, BTRFS_LABEL_SIZE - 1);
2194 int get_label(const char *btrfs_dev, char *label)
2198 ret = is_existing_blk_or_reg_file(btrfs_dev);
2200 ret = get_label_mounted(btrfs_dev, label);
2202 ret = get_label_unmounted(btrfs_dev, label);
2207 int set_label(const char *btrfs_dev, const char *label)
2211 if (check_label(label))
2214 ret = is_existing_blk_or_reg_file(btrfs_dev);
2216 ret = set_label_mounted(btrfs_dev, label);
2218 ret = set_label_unmounted(btrfs_dev, label);
2224 * A not-so-good version fls64. No fascinating optimization since
2225 * no one except parse_size use it
2227 static int fls64(u64 x)
2231 for (i = 0; i <64; i++)
2232 if (x << i & (1ULL << 63))
2237 u64 parse_size(char *s)
2245 error("size value is empty");
2249 error("size value '%s' is less equal than 0", s);
2252 ret = strtoull(s, &endptr, 10);
2254 error("size value '%s' is invalid", s);
2257 if (endptr[0] && endptr[1]) {
2258 error("illegal suffix contains character '%c' in wrong position",
2263 * strtoll returns LLONG_MAX when overflow, if this happens,
2264 * need to call strtoull to get the real size
2266 if (errno == ERANGE && ret == ULLONG_MAX) {
2267 error("size value '%s' is too large for u64", s);
2271 c = tolower(endptr[0]);
2294 error("unknown size descriptor '%c'", c);
2298 /* Check whether ret * mult overflow */
2299 if (fls64(ret) + fls64(mult) - 1 > 64) {
2300 error("size value '%s' is too large for u64", s);
2307 u64 parse_qgroupid(const char *p)
2309 char *s = strchr(p, '/');
2310 const char *ptr_src_end = p + strlen(p);
2311 char *ptr_parse_end = NULL;
2320 /* Numeric format like '0/257' is the primary case */
2322 id = strtoull(p, &ptr_parse_end, 10);
2323 if (ptr_parse_end != ptr_src_end)
2327 level = strtoull(p, &ptr_parse_end, 10);
2328 if (ptr_parse_end != s)
2331 id = strtoull(s + 1, &ptr_parse_end, 10);
2332 if (ptr_parse_end != ptr_src_end)
2335 return (level << BTRFS_QGROUP_LEVEL_SHIFT) | id;
2338 /* Path format like subv at 'my_subvol' is the fallback case */
2339 ret = test_issubvolume(p);
2340 if (ret < 0 || !ret)
2342 fd = open(p, O_RDONLY);
2345 ret = lookup_ino_rootid(fd, &id);
2352 error("invalid qgroupid or subvolume path: %s", p);
2356 int open_file_or_dir3(const char *fname, DIR **dirstream, int open_flags)
2362 ret = stat(fname, &st);
2366 if (S_ISDIR(st.st_mode)) {
2367 *dirstream = opendir(fname);
2370 fd = dirfd(*dirstream);
2371 } else if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) {
2372 fd = open(fname, open_flags);
2375 * we set this on purpose, in case the caller output
2376 * strerror(errno) as success
2384 closedir(*dirstream);
2391 int open_file_or_dir(const char *fname, DIR **dirstream)
2393 return open_file_or_dir3(fname, dirstream, O_RDWR);
2396 void close_file_or_dir(int fd, DIR *dirstream)
2399 closedir(dirstream);
2404 int get_device_info(int fd, u64 devid,
2405 struct btrfs_ioctl_dev_info_args *di_args)
2409 di_args->devid = devid;
2410 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
2412 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
2413 return ret < 0 ? -errno : 0;
2416 static u64 find_max_device_id(struct btrfs_ioctl_search_args *search_args,
2419 struct btrfs_dev_item *dev_item;
2420 char *buf = search_args->buf;
2422 buf += (nr_items - 1) * (sizeof(struct btrfs_ioctl_search_header)
2423 + sizeof(struct btrfs_dev_item));
2424 buf += sizeof(struct btrfs_ioctl_search_header);
2426 dev_item = (struct btrfs_dev_item *)buf;
2428 return btrfs_stack_device_id(dev_item);
2431 static int search_chunk_tree_for_fs_info(int fd,
2432 struct btrfs_ioctl_fs_info_args *fi_args)
2436 u64 start_devid = 1;
2437 struct btrfs_ioctl_search_args search_args;
2438 struct btrfs_ioctl_search_key *search_key = &search_args.key;
2440 fi_args->num_devices = 0;
2442 max_items = BTRFS_SEARCH_ARGS_BUFSIZE
2443 / (sizeof(struct btrfs_ioctl_search_header)
2444 + sizeof(struct btrfs_dev_item));
2446 search_key->tree_id = BTRFS_CHUNK_TREE_OBJECTID;
2447 search_key->min_objectid = BTRFS_DEV_ITEMS_OBJECTID;
2448 search_key->max_objectid = BTRFS_DEV_ITEMS_OBJECTID;
2449 search_key->min_type = BTRFS_DEV_ITEM_KEY;
2450 search_key->max_type = BTRFS_DEV_ITEM_KEY;
2451 search_key->min_transid = 0;
2452 search_key->max_transid = (u64)-1;
2453 search_key->nr_items = max_items;
2454 search_key->max_offset = (u64)-1;
2457 search_key->min_offset = start_devid;
2459 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH, &search_args);
2463 fi_args->num_devices += (u64)search_key->nr_items;
2465 if (search_key->nr_items == max_items) {
2466 start_devid = find_max_device_id(&search_args,
2467 search_key->nr_items) + 1;
2471 /* get the lastest max_id to stay consistent with the num_devices */
2472 if (search_key->nr_items == 0)
2474 * last tree_search returns an empty buf, use the devid of
2475 * the last dev_item of the previous tree_search
2477 fi_args->max_id = start_devid - 1;
2479 fi_args->max_id = find_max_device_id(&search_args,
2480 search_key->nr_items);
2486 * For a given path, fill in the ioctl fs_ and info_ args.
2487 * If the path is a btrfs mountpoint, fill info for all devices.
2488 * If the path is a btrfs device, fill in only that device.
2490 * The path provided must be either on a mounted btrfs fs,
2491 * or be a mounted btrfs device.
2493 * Returns 0 on success, or a negative errno.
2495 int get_fs_info(char *path, struct btrfs_ioctl_fs_info_args *fi_args,
2496 struct btrfs_ioctl_dev_info_args **di_ret)
2503 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2504 struct btrfs_ioctl_dev_info_args *di_args;
2505 struct btrfs_ioctl_dev_info_args tmp;
2507 DIR *dirstream = NULL;
2509 memset(fi_args, 0, sizeof(*fi_args));
2511 if (is_block_device(path) == 1) {
2512 struct btrfs_super_block *disk_super;
2513 char buf[BTRFS_SUPER_INFO_SIZE];
2516 /* Ensure it's mounted, then set path to the mountpoint */
2517 fd = open(path, O_RDONLY);
2520 error("cannot open %s: %s", path, strerror(errno));
2523 ret = check_mounted_where(fd, path, mp, sizeof(mp),
2532 /* Only fill in this one device */
2533 fi_args->num_devices = 1;
2535 disk_super = (struct btrfs_super_block *)buf;
2536 ret = btrfs_read_dev_super(fd, disk_super,
2537 BTRFS_SUPER_INFO_OFFSET, 0);
2542 devid = btrfs_stack_device_id(&disk_super->dev_item);
2544 fi_args->max_id = devid;
2547 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
2551 /* at this point path must not be for a block device */
2552 fd = open_file_or_dir(path, &dirstream);
2558 /* fill in fi_args if not just a single device */
2559 if (fi_args->num_devices != 1) {
2560 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
2567 * The fs_args->num_devices does not include seed devices
2569 ret = search_chunk_tree_for_fs_info(fd, fi_args);
2574 * search_chunk_tree_for_fs_info() will lacks the devid 0
2575 * so manual probe for it here.
2577 ret = get_device_info(fd, 0, &tmp);
2579 fi_args->num_devices++;
2587 if (!fi_args->num_devices)
2590 di_args = *di_ret = malloc((fi_args->num_devices) * sizeof(*di_args));
2597 memcpy(di_args, &tmp, sizeof(tmp));
2598 for (; i <= fi_args->max_id; ++i) {
2599 ret = get_device_info(fd, i, &di_args[ndevs]);
2608 * only when the only dev we wanted to find is not there then
2609 * let any error be returned
2611 if (fi_args->num_devices != 1) {
2617 close_file_or_dir(fd, dirstream);
2621 #define isoctal(c) (((c) & ~7) == '0')
2623 static inline void translate(char *f, char *t)
2625 while (*f != '\0') {
2627 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
2628 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
2638 * Checks if the swap device.
2639 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
2641 static int is_swap_device(const char *file)
2652 if (stat(file, &st_buf) < 0)
2654 if (S_ISBLK(st_buf.st_mode))
2655 dev = st_buf.st_rdev;
2656 else if (S_ISREG(st_buf.st_mode)) {
2657 dev = st_buf.st_dev;
2658 ino = st_buf.st_ino;
2662 if ((f = fopen("/proc/swaps", "r")) == NULL)
2665 /* skip the first line */
2666 if (fgets(tmp, sizeof(tmp), f) == NULL)
2669 while (fgets(tmp, sizeof(tmp), f) != NULL) {
2670 if ((cp = strchr(tmp, ' ')) != NULL)
2672 if ((cp = strchr(tmp, '\t')) != NULL)
2674 translate(tmp, buf);
2675 if (stat(buf, &st_buf) != 0)
2677 if (S_ISBLK(st_buf.st_mode)) {
2678 if (dev == st_buf.st_rdev) {
2682 } else if (S_ISREG(st_buf.st_mode)) {
2683 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {
2697 * Check for existing filesystem or partition table on device.
2699 * 1 for existing fs or partition
2700 * 0 for nothing found
2701 * -1 for internal error
2703 static int check_overwrite(const char *device)
2706 blkid_probe pr = NULL;
2710 if (!device || !*device)
2713 ret = -1; /* will reset on success of all setup calls */
2715 pr = blkid_new_probe_from_filename(device);
2719 size = blkid_probe_get_size(pr);
2723 /* nothing to overwrite on a 0-length device */
2729 ret = blkid_probe_enable_partitions(pr, 1);
2733 ret = blkid_do_fullprobe(pr);
2738 * Blkid returns 1 for nothing found and 0 when it finds a signature,
2739 * but we want the exact opposite, so reverse the return value here.
2741 * In addition print some useful diagnostics about what actually is
2749 if (!blkid_probe_lookup_value(pr, "TYPE", &type, NULL)) {
2751 "%s appears to contain an existing "
2752 "filesystem (%s).\n", device, type);
2753 } else if (!blkid_probe_lookup_value(pr, "PTTYPE", &type, NULL)) {
2755 "%s appears to contain a partition "
2756 "table (%s).\n", device, type);
2759 "%s appears to contain something weird "
2760 "according to blkid\n", device);
2766 blkid_free_probe(pr);
2769 "probe of %s failed, cannot detect "
2770 "existing filesystem.\n", device);
2774 static int group_profile_devs_min(u64 flag)
2776 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
2777 case 0: /* single */
2778 case BTRFS_BLOCK_GROUP_DUP:
2780 case BTRFS_BLOCK_GROUP_RAID0:
2781 case BTRFS_BLOCK_GROUP_RAID1:
2782 case BTRFS_BLOCK_GROUP_RAID5:
2784 case BTRFS_BLOCK_GROUP_RAID6:
2786 case BTRFS_BLOCK_GROUP_RAID10:
2793 int test_num_disk_vs_raid(u64 metadata_profile, u64 data_profile,
2794 u64 dev_cnt, int mixed, int ssd)
2801 allowed |= BTRFS_BLOCK_GROUP_RAID10;
2803 allowed |= BTRFS_BLOCK_GROUP_RAID6;
2805 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
2806 BTRFS_BLOCK_GROUP_RAID5;
2808 allowed |= BTRFS_BLOCK_GROUP_DUP;
2812 ((metadata_profile | data_profile) & BTRFS_BLOCK_GROUP_DUP)) {
2813 warning("DUP is not recommended on filesystem with multiple devices");
2815 if (metadata_profile & ~allowed) {
2817 "ERROR: unable to create FS with metadata profile %s "
2818 "(have %llu devices but %d devices are required)\n",
2819 btrfs_group_profile_str(metadata_profile), dev_cnt,
2820 group_profile_devs_min(metadata_profile));
2823 if (data_profile & ~allowed) {
2825 "ERROR: unable to create FS with data profile %s "
2826 "(have %llu devices but %d devices are required)\n",
2827 btrfs_group_profile_str(data_profile), dev_cnt,
2828 group_profile_devs_min(data_profile));
2832 warning_on(!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP) && ssd,
2833 "DUP may not actually lead to 2 copies on the device, see manual page");
2838 int group_profile_max_safe_loss(u64 flags)
2840 switch (flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
2841 case 0: /* single */
2842 case BTRFS_BLOCK_GROUP_DUP:
2843 case BTRFS_BLOCK_GROUP_RAID0:
2845 case BTRFS_BLOCK_GROUP_RAID1:
2846 case BTRFS_BLOCK_GROUP_RAID5:
2847 case BTRFS_BLOCK_GROUP_RAID10:
2849 case BTRFS_BLOCK_GROUP_RAID6:
2857 * Check if a device is suitable for btrfs
2859 * 1: something is wrong, an error is printed
2862 int test_dev_for_mkfs(const char *file, int force_overwrite)
2867 ret = is_swap_device(file);
2869 error("checking status of %s: %s", file, strerror(-ret));
2873 error("%s is a swap device", file);
2876 if (!force_overwrite) {
2877 if (check_overwrite(file)) {
2878 error("use the -f option to force overwrite of %s",
2883 ret = check_mounted(file);
2885 error("cannot check mount status of %s: %s", file,
2890 error("%s is mounted", file);
2893 /* check if the device is busy */
2894 fd = open(file, O_RDWR|O_EXCL);
2896 error("unable to open %s: %s", file, strerror(errno));
2899 if (fstat(fd, &st)) {
2900 error("unable to stat %s: %s", file, strerror(errno));
2904 if (!S_ISBLK(st.st_mode)) {
2905 error("%s is not a block device", file);
2913 int btrfs_scan_lblkid(void)
2918 struct btrfs_fs_devices *tmp_devices;
2919 blkid_dev_iterate iter = NULL;
2920 blkid_dev dev = NULL;
2921 blkid_cache cache = NULL;
2922 char path[PATH_MAX];
2924 if (btrfs_scan_done)
2927 if (blkid_get_cache(&cache, NULL) < 0) {
2928 error("blkid cache get failed");
2931 blkid_probe_all(cache);
2932 iter = blkid_dev_iterate_begin(cache);
2933 blkid_dev_set_search(iter, "TYPE", "btrfs");
2934 while (blkid_dev_next(iter, &dev) == 0) {
2935 dev = blkid_verify(cache, dev);
2938 /* if we are here its definitely a btrfs disk*/
2939 strncpy_null(path, blkid_dev_devname(dev));
2941 fd = open(path, O_RDONLY);
2943 error("cannot open %s: %s", path, strerror(errno));
2946 ret = btrfs_scan_one_device(fd, path, &tmp_devices,
2947 &num_devices, BTRFS_SUPER_INFO_OFFSET, 0);
2949 error("cannot scan %s: %s", path, strerror(-ret));
2956 blkid_dev_iterate_end(iter);
2957 blkid_put_cache(cache);
2959 btrfs_scan_done = 1;
2964 int is_vol_small(const char *file)
2971 fd = open(file, O_RDONLY);
2974 if (fstat(fd, &st) < 0) {
2979 size = btrfs_device_size(fd, &st);
2984 if (size < BTRFS_MKFS_SMALL_VOLUME_SIZE) {
2994 * This reads a line from the stdin and only returns non-zero if the
2995 * first whitespace delimited token is a case insensitive match with yes
2998 int ask_user(const char *question)
3000 char buf[30] = {0,};
3001 char *saveptr = NULL;
3004 printf("%s [y/N]: ", question);
3006 return fgets(buf, sizeof(buf) - 1, stdin) &&
3007 (answer = strtok_r(buf, " \t\n\r", &saveptr)) &&
3008 (!strcasecmp(answer, "yes") || !strcasecmp(answer, "y"));
3013 * - file or directory return the containing tree root id
3014 * - subvolume return its own tree id
3015 * - BTRFS_EMPTY_SUBVOL_DIR_OBJECTID (directory with ino == 2) the result is
3016 * undefined and function returns -1
3018 int lookup_ino_rootid(int fd, u64 *rootid)
3020 struct btrfs_ioctl_ino_lookup_args args;
3023 memset(&args, 0, sizeof(args));
3025 args.objectid = BTRFS_FIRST_FREE_OBJECTID;
3027 ret = ioctl(fd, BTRFS_IOC_INO_LOOKUP, &args);
3029 error("failed to lookup root id: %s", strerror(errno));
3033 *rootid = args.treeid;
3039 * return 0 if a btrfs mount point is found
3040 * return 1 if a mount point is found but not btrfs
3041 * return <0 if something goes wrong
3043 int find_mount_root(const char *path, char **mount_root)
3051 int longest_matchlen = 0;
3052 char *longest_match = NULL;
3054 fd = open(path, O_RDONLY | O_NOATIME);
3059 mnttab = setmntent("/proc/self/mounts", "r");
3063 while ((ent = getmntent(mnttab))) {
3064 len = strlen(ent->mnt_dir);
3065 if (strncmp(ent->mnt_dir, path, len) == 0) {
3066 /* match found and use the latest match */
3067 if (longest_matchlen <= len) {
3068 free(longest_match);
3069 longest_matchlen = len;
3070 longest_match = strdup(ent->mnt_dir);
3071 not_btrfs = strcmp(ent->mnt_type, "btrfs");
3080 free(longest_match);
3085 *mount_root = realpath(longest_match, NULL);
3089 free(longest_match);
3093 int test_minimum_size(const char *file, u32 nodesize)
3096 struct stat statbuf;
3098 fd = open(file, O_RDONLY);
3101 if (stat(file, &statbuf) < 0) {
3105 if (btrfs_device_size(fd, &statbuf) < btrfs_min_dev_size(nodesize)) {
3115 * Test if path is a directory
3117 * 0 - path exists but it is not a directory
3118 * 1 - path exists and it is a directory
3121 int test_isdir(const char *path)
3126 ret = stat(path, &st);
3130 return !!S_ISDIR(st.st_mode);
3133 void units_set_mode(unsigned *units, unsigned mode)
3135 unsigned base = *units & UNITS_MODE_MASK;
3137 *units = base | mode;
3140 void units_set_base(unsigned *units, unsigned base)
3142 unsigned mode = *units & ~UNITS_MODE_MASK;
3144 *units = base | mode;
3147 int find_next_key(struct btrfs_path *path, struct btrfs_key *key)
3151 for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
3152 if (!path->nodes[level])
3154 if (path->slots[level] + 1 >=
3155 btrfs_header_nritems(path->nodes[level]))
3158 btrfs_item_key_to_cpu(path->nodes[level], key,
3159 path->slots[level] + 1);
3161 btrfs_node_key_to_cpu(path->nodes[level], key,
3162 path->slots[level] + 1);
3168 const char* btrfs_group_type_str(u64 flag)
3170 u64 mask = BTRFS_BLOCK_GROUP_TYPE_MASK |
3171 BTRFS_SPACE_INFO_GLOBAL_RSV;
3173 switch (flag & mask) {
3174 case BTRFS_BLOCK_GROUP_DATA:
3176 case BTRFS_BLOCK_GROUP_SYSTEM:
3178 case BTRFS_BLOCK_GROUP_METADATA:
3180 case BTRFS_BLOCK_GROUP_DATA|BTRFS_BLOCK_GROUP_METADATA:
3181 return "Data+Metadata";
3182 case BTRFS_SPACE_INFO_GLOBAL_RSV:
3183 return "GlobalReserve";
3189 const char* btrfs_group_profile_str(u64 flag)
3191 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3194 case BTRFS_BLOCK_GROUP_RAID0:
3196 case BTRFS_BLOCK_GROUP_RAID1:
3198 case BTRFS_BLOCK_GROUP_RAID5:
3200 case BTRFS_BLOCK_GROUP_RAID6:
3202 case BTRFS_BLOCK_GROUP_DUP:
3204 case BTRFS_BLOCK_GROUP_RAID10:
3211 u64 disk_size(const char *path)
3215 if (statfs(path, &sfs) < 0)
3218 return sfs.f_bsize * sfs.f_blocks;
3221 u64 get_partition_size(const char *dev)
3224 int fd = open(dev, O_RDONLY);
3228 if (ioctl(fd, BLKGETSIZE64, &result) < 0) {
3237 int btrfs_tree_search2_ioctl_supported(int fd)
3239 struct btrfs_ioctl_search_args_v2 *args2;
3240 struct btrfs_ioctl_search_key *sk;
3241 int args2_size = 1024;
3242 char args2_buf[args2_size];
3244 static int v2_supported = -1;
3246 if (v2_supported != -1)
3247 return v2_supported;
3249 args2 = (struct btrfs_ioctl_search_args_v2 *)args2_buf;
3253 * Search for the extent tree item in the root tree.
3255 sk->tree_id = BTRFS_ROOT_TREE_OBJECTID;
3256 sk->min_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3257 sk->max_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3258 sk->min_type = BTRFS_ROOT_ITEM_KEY;
3259 sk->max_type = BTRFS_ROOT_ITEM_KEY;
3261 sk->max_offset = (u64)-1;
3262 sk->min_transid = 0;
3263 sk->max_transid = (u64)-1;
3265 args2->buf_size = args2_size - sizeof(struct btrfs_ioctl_search_args_v2);
3266 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH_V2, args2);
3267 if (ret == -EOPNOTSUPP)
3274 return v2_supported;
3277 int btrfs_check_nodesize(u32 nodesize, u32 sectorsize, u64 features)
3279 if (nodesize < sectorsize) {
3280 error("illegal nodesize %u (smaller than %u)",
3281 nodesize, sectorsize);
3283 } else if (nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) {
3284 error("illegal nodesize %u (larger than %u)",
3285 nodesize, BTRFS_MAX_METADATA_BLOCKSIZE);
3287 } else if (nodesize & (sectorsize - 1)) {
3288 error("illegal nodesize %u (not aligned to %u)",
3289 nodesize, sectorsize);
3291 } else if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS &&
3292 nodesize != sectorsize) {
3293 error("illegal nodesize %u (not equal to %u for mixed block group)",
3294 nodesize, sectorsize);
3301 * Copy a path argument from SRC to DEST and check the SRC length if it's at
3302 * most PATH_MAX and fits into DEST. DESTLEN is supposed to be exact size of
3304 * The destination buffer is zero terminated.
3305 * Return < 0 for error, 0 otherwise.
3307 int arg_copy_path(char *dest, const char *src, int destlen)
3309 size_t len = strlen(src);
3311 if (len >= PATH_MAX || len >= destlen)
3312 return -ENAMETOOLONG;
3314 __strncpy_null(dest, src, destlen);
3319 unsigned int get_unit_mode_from_arg(int *argc, char *argv[], int df_mode)
3321 unsigned int unit_mode = UNITS_DEFAULT;
3325 for (arg_i = 0; arg_i < *argc; arg_i++) {
3326 if (!strcmp(argv[arg_i], "--"))
3329 if (!strcmp(argv[arg_i], "--raw")) {
3330 unit_mode = UNITS_RAW;
3334 if (!strcmp(argv[arg_i], "--human-readable")) {
3335 unit_mode = UNITS_HUMAN_BINARY;
3340 if (!strcmp(argv[arg_i], "--iec")) {
3341 units_set_mode(&unit_mode, UNITS_BINARY);
3345 if (!strcmp(argv[arg_i], "--si")) {
3346 units_set_mode(&unit_mode, UNITS_DECIMAL);
3351 if (!strcmp(argv[arg_i], "--kbytes")) {
3352 units_set_base(&unit_mode, UNITS_KBYTES);
3356 if (!strcmp(argv[arg_i], "--mbytes")) {
3357 units_set_base(&unit_mode, UNITS_MBYTES);
3361 if (!strcmp(argv[arg_i], "--gbytes")) {
3362 units_set_base(&unit_mode, UNITS_GBYTES);
3366 if (!strcmp(argv[arg_i], "--tbytes")) {
3367 units_set_base(&unit_mode, UNITS_TBYTES);
3375 if (!strcmp(argv[arg_i], "-b")) {
3376 unit_mode = UNITS_RAW;
3380 if (!strcmp(argv[arg_i], "-h")) {
3381 unit_mode = UNITS_HUMAN_BINARY;
3385 if (!strcmp(argv[arg_i], "-H")) {
3386 unit_mode = UNITS_HUMAN_DECIMAL;
3390 if (!strcmp(argv[arg_i], "-k")) {
3391 units_set_base(&unit_mode, UNITS_KBYTES);
3395 if (!strcmp(argv[arg_i], "-m")) {
3396 units_set_base(&unit_mode, UNITS_MBYTES);
3400 if (!strcmp(argv[arg_i], "-g")) {
3401 units_set_base(&unit_mode, UNITS_GBYTES);
3405 if (!strcmp(argv[arg_i], "-t")) {
3406 units_set_base(&unit_mode, UNITS_TBYTES);
3412 for (arg_i = 0, arg_end = 0; arg_i < *argc; arg_i++) {
3415 argv[arg_end] = argv[arg_i];
3424 int string_is_numerical(const char *str)
3426 if (!(*str >= '0' && *str <= '9'))
3428 while (*str >= '0' && *str <= '9')
3436 * Preprocess @argv with getopt_long to reorder options and consume the "--"
3438 * Unknown short and long options are reported, optionally the @usage is printed
3441 void clean_args_no_options(int argc, char *argv[], const char * const *usagestr)
3443 static const struct option long_options[] = {
3448 int c = getopt_long(argc, argv, "", long_options, NULL);
3461 /* Subvolume helper functions */
3463 * test if name is a correct subvolume name
3464 * this function return
3465 * 0-> name is not a correct subvolume name
3466 * 1-> name is a correct subvolume name
3468 int test_issubvolname(const char *name)
3470 return name[0] != '\0' && !strchr(name, '/') &&
3471 strcmp(name, ".") && strcmp(name, "..");
3475 * Test if path is a subvolume
3477 * 0 - path exists but it is not a subvolume
3478 * 1 - path exists and it is a subvolume
3481 int test_issubvolume(const char *path)
3487 res = stat(path, &st);
3491 if (st.st_ino != BTRFS_FIRST_FREE_OBJECTID || !S_ISDIR(st.st_mode))
3494 res = statfs(path, &stfs);
3498 return (int)stfs.f_type == BTRFS_SUPER_MAGIC;
3501 const char *subvol_strip_mountpoint(const char *mnt, const char *full_path)
3503 int len = strlen(mnt);
3507 if (mnt[len - 1] != '/')
3510 return full_path + len;
3517 * 1: Error; and error info printed to the terminal. Fixme.
3518 * 2: If the fullpath is root tree instead of subvol tree
3520 int get_subvol_info(const char *fullpath, struct root_info *get_ri)
3527 const char *svpath = NULL;
3528 DIR *dirstream1 = NULL;
3529 DIR *dirstream2 = NULL;
3531 ret = test_issubvolume(fullpath);
3535 error("not a subvolume: %s", fullpath);
3539 ret = find_mount_root(fullpath, &mnt);
3543 error("%s doesn't belong to btrfs mount point", fullpath);
3547 svpath = subvol_strip_mountpoint(mnt, fullpath);
3549 fd = btrfs_open_dir(fullpath, &dirstream1, 1);
3553 ret = btrfs_list_get_path_rootid(fd, &sv_id);
3555 error("can't get rootid for '%s'", fullpath);
3559 mntfd = btrfs_open_dir(mnt, &dirstream2, 1);
3563 if (sv_id == BTRFS_FS_TREE_OBJECTID) {
3566 * So that caller may decide if thats an error or just fine.
3571 memset(get_ri, 0, sizeof(*get_ri));
3572 get_ri->root_id = sv_id;
3574 ret = btrfs_get_subvol(mntfd, get_ri);
3576 error("can't find '%s': %d", svpath, ret);
3579 close_file_or_dir(mntfd, dirstream2);
3580 close_file_or_dir(fd, dirstream1);
3586 void init_rand_seed(u64 seed)
3590 /* only use the last 48 bits */
3591 for (i = 0; i < 3; i++) {
3592 rand_seed[i] = (unsigned short)(seed ^ (unsigned short)(-1));
3595 rand_seed_initlized = 1;
3598 static void __init_seed(void)
3604 if(rand_seed_initlized)
3606 /* Use urandom as primary seed source. */
3607 fd = open("/dev/urandom", O_RDONLY);
3609 ret = read(fd, rand_seed, sizeof(rand_seed));
3611 if (ret < sizeof(rand_seed))
3615 /* Use time and pid as fallback seed */
3616 warning("failed to read /dev/urandom, use time and pid as random seed");
3617 gettimeofday(&tv, 0);
3618 rand_seed[0] = getpid() ^ (tv.tv_sec & 0xFFFF);
3619 rand_seed[1] = getppid() ^ (tv.tv_usec & 0xFFFF);
3620 rand_seed[2] = (tv.tv_sec ^ tv.tv_usec) >> 16;
3622 rand_seed_initlized = 1;
3629 * Don't use nrand48, its range is [0,2^31) The highest bit will alwasy
3630 * be 0. Use jrand48 to include the highest bit.
3632 return (u32)jrand48(rand_seed);
3635 unsigned int rand_range(unsigned int upper)
3639 * Use the full 48bits to mod, which would be more uniformly
3642 return (unsigned int)(jrand48(rand_seed) % upper);