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);
439 static int insert_temp_dev_item(int fd, struct extent_buffer *buf,
440 struct btrfs_mkfs_config *cfg,
441 int *slot, u32 *itemoff)
443 struct btrfs_disk_key disk_key;
444 struct btrfs_dev_item *dev_item;
445 char super_buf[BTRFS_SUPER_INFO_SIZE];
446 unsigned char dev_uuid[BTRFS_UUID_SIZE];
447 unsigned char fsid[BTRFS_FSID_SIZE];
448 struct btrfs_super_block *super = (struct btrfs_super_block *)super_buf;
451 ret = pread(fd, super_buf, BTRFS_SUPER_INFO_SIZE, cfg->super_bytenr);
452 if (ret < BTRFS_SUPER_INFO_SIZE) {
453 ret = (ret < 0 ? -errno : -EIO);
457 btrfs_set_header_nritems(buf, *slot + 1);
458 (*itemoff) -= sizeof(*dev_item);
459 /* setup device item 1, 0 is for replace case */
460 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
461 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
462 btrfs_set_disk_key_offset(&disk_key, 1);
463 btrfs_set_item_key(buf, &disk_key, *slot);
464 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
465 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(*dev_item));
467 dev_item = btrfs_item_ptr(buf, *slot, struct btrfs_dev_item);
468 /* Generate device uuid */
469 uuid_generate(dev_uuid);
470 write_extent_buffer(buf, dev_uuid,
471 (unsigned long)btrfs_device_uuid(dev_item),
473 uuid_parse(cfg->fs_uuid, fsid);
474 write_extent_buffer(buf, fsid,
475 (unsigned long)btrfs_device_fsid(dev_item),
477 btrfs_set_device_id(buf, dev_item, 1);
478 btrfs_set_device_generation(buf, dev_item, 0);
479 btrfs_set_device_total_bytes(buf, dev_item, cfg->num_bytes);
481 * The number must match the initial SYSTEM and META chunk size
483 btrfs_set_device_bytes_used(buf, dev_item,
484 BTRFS_MKFS_SYSTEM_GROUP_SIZE +
485 BTRFS_CONVERT_META_GROUP_SIZE);
486 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
487 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
488 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
489 btrfs_set_device_type(buf, dev_item, 0);
491 /* Super dev_item is not complete, copy the complete one to sb */
492 read_extent_buffer(buf, &super->dev_item, (unsigned long)dev_item,
494 ret = write_temp_super(fd, super, cfg->super_bytenr);
500 static int insert_temp_chunk_item(int fd, struct extent_buffer *buf,
501 struct btrfs_mkfs_config *cfg,
502 int *slot, u32 *itemoff, u64 start, u64 len,
505 struct btrfs_chunk *chunk;
506 struct btrfs_disk_key disk_key;
507 char super_buf[BTRFS_SUPER_INFO_SIZE];
508 struct btrfs_super_block *sb = (struct btrfs_super_block *)super_buf;
511 ret = pread(fd, super_buf, BTRFS_SUPER_INFO_SIZE,
513 if (ret < BTRFS_SUPER_INFO_SIZE) {
514 ret = (ret < 0 ? ret : -EIO);
518 btrfs_set_header_nritems(buf, *slot + 1);
519 (*itemoff) -= btrfs_chunk_item_size(1);
520 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
521 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
522 btrfs_set_disk_key_offset(&disk_key, start);
523 btrfs_set_item_key(buf, &disk_key, *slot);
524 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
525 btrfs_set_item_size(buf, btrfs_item_nr(*slot),
526 btrfs_chunk_item_size(1));
528 chunk = btrfs_item_ptr(buf, *slot, struct btrfs_chunk);
529 btrfs_set_chunk_length(buf, chunk, len);
530 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
531 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
532 btrfs_set_chunk_type(buf, chunk, type);
533 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
534 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
535 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
536 btrfs_set_chunk_num_stripes(buf, chunk, 1);
537 /* TODO: Support DUP profile for system chunk */
538 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
539 /* We are doing 1:1 mapping, so start is its dev offset */
540 btrfs_set_stripe_offset_nr(buf, chunk, 0, start);
541 write_extent_buffer(buf, &sb->dev_item.uuid,
542 (unsigned long)btrfs_stripe_dev_uuid_nr(chunk, 0),
547 * If it's system chunk, also copy it to super block.
549 if (type & BTRFS_BLOCK_GROUP_SYSTEM) {
552 cur = (char *)sb->sys_chunk_array + sb->sys_chunk_array_size;
553 memcpy(cur, &disk_key, sizeof(disk_key));
554 cur += sizeof(disk_key);
555 read_extent_buffer(buf, cur, (unsigned long int)chunk,
556 btrfs_chunk_item_size(1));
557 sb->sys_chunk_array_size += btrfs_chunk_item_size(1) +
560 ret = write_temp_super(fd, sb, cfg->super_bytenr);
565 static int setup_temp_chunk_tree(int fd, struct btrfs_mkfs_config *cfg,
566 u64 sys_chunk_start, u64 meta_chunk_start,
569 struct extent_buffer *buf = NULL;
570 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
574 /* Must ensure SYS chunk starts before META chunk */
575 BUG_ON(meta_chunk_start < sys_chunk_start);
576 buf = malloc(sizeof(*buf) + cfg->nodesize);
579 ret = setup_temp_extent_buffer(buf, cfg, chunk_bytenr,
580 BTRFS_CHUNK_TREE_OBJECTID);
584 ret = insert_temp_dev_item(fd, buf, cfg, &slot, &itemoff);
587 ret = insert_temp_chunk_item(fd, buf, cfg, &slot, &itemoff,
589 BTRFS_MKFS_SYSTEM_GROUP_SIZE,
590 BTRFS_BLOCK_GROUP_SYSTEM);
593 ret = insert_temp_chunk_item(fd, buf, cfg, &slot, &itemoff,
595 BTRFS_CONVERT_META_GROUP_SIZE,
596 BTRFS_BLOCK_GROUP_METADATA);
599 ret = write_temp_extent_buffer(fd, buf, chunk_bytenr);
606 static void insert_temp_dev_extent(struct extent_buffer *buf,
607 int *slot, u32 *itemoff, u64 start, u64 len)
609 struct btrfs_dev_extent *dev_extent;
610 struct btrfs_disk_key disk_key;
612 btrfs_set_header_nritems(buf, *slot + 1);
613 (*itemoff) -= sizeof(*dev_extent);
614 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
615 btrfs_set_disk_key_objectid(&disk_key, 1);
616 btrfs_set_disk_key_offset(&disk_key, start);
617 btrfs_set_item_key(buf, &disk_key, *slot);
618 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
619 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(*dev_extent));
621 dev_extent = btrfs_item_ptr(buf, *slot, struct btrfs_dev_extent);
622 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
623 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
624 btrfs_set_dev_extent_length(buf, dev_extent, len);
625 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, start);
626 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
627 BTRFS_CHUNK_TREE_OBJECTID);
631 static int setup_temp_dev_tree(int fd, struct btrfs_mkfs_config *cfg,
632 u64 sys_chunk_start, u64 meta_chunk_start,
635 struct extent_buffer *buf = NULL;
636 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
640 /* Must ensure SYS chunk starts before META chunk */
641 BUG_ON(meta_chunk_start < sys_chunk_start);
642 buf = malloc(sizeof(*buf) + cfg->nodesize);
645 ret = setup_temp_extent_buffer(buf, cfg, dev_bytenr,
646 BTRFS_DEV_TREE_OBJECTID);
649 insert_temp_dev_extent(buf, &slot, &itemoff, sys_chunk_start,
650 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
651 insert_temp_dev_extent(buf, &slot, &itemoff, meta_chunk_start,
652 BTRFS_CONVERT_META_GROUP_SIZE);
653 ret = write_temp_extent_buffer(fd, buf, dev_bytenr);
659 static int setup_temp_fs_tree(int fd, struct btrfs_mkfs_config *cfg,
662 struct extent_buffer *buf = NULL;
665 buf = malloc(sizeof(*buf) + cfg->nodesize);
668 ret = setup_temp_extent_buffer(buf, cfg, fs_bytenr,
669 BTRFS_FS_TREE_OBJECTID);
673 * Temporary fs tree is completely empty.
675 ret = write_temp_extent_buffer(fd, buf, fs_bytenr);
682 * Improved version of make_btrfs().
685 * 1) Do chunk allocation to avoid used data
686 * And after this function, extent type matches chunk type
687 * 2) Better structured code
688 * No super long hand written codes to initialized all tree blocks
689 * Split into small blocks and reuse codes.
690 * TODO: Reuse tree operation facilities by introducing new flags
692 static int make_convert_btrfs(int fd, struct btrfs_mkfs_config *cfg,
693 struct btrfs_convert_context *cctx)
695 struct cache_tree *free = &cctx->free;
696 struct cache_tree *used = &cctx->used;
698 u64 meta_chunk_start;
699 /* chunk tree bytenr, in system chunk */
701 /* metadata trees bytenr, in metadata chunk */
709 /* Shouldn't happen */
710 BUG_ON(cache_tree_empty(used));
713 * reserve space for temporary superblock first
714 * Here we allocate a little larger space, to keep later
715 * free space will be STRIPE_LEN aligned
717 ret = reserve_free_space(free, BTRFS_STRIPE_LEN,
723 * Then reserve system chunk space
724 * TODO: Change system group size depending on cctx->total_bytes.
725 * If using current 4M, it can only handle less than one TB for
726 * worst case and then run out of sys space.
728 ret = reserve_free_space(free, BTRFS_MKFS_SYSTEM_GROUP_SIZE,
732 ret = reserve_free_space(free, BTRFS_CONVERT_META_GROUP_SIZE,
738 * Allocated meta/sys chunks will be mapped 1:1 with device offset.
740 * Inside the allocated metadata chunk, the layout will be:
741 * | offset | contents |
742 * -------------------------------------
744 * | +nodesize | extent root |
745 * | +nodesize * 2 | device root |
746 * | +nodesize * 3 | fs tree |
747 * | +nodesize * 4 | csum tree |
748 * -------------------------------------
749 * Inside the allocated system chunk, the layout will be:
750 * | offset | contents |
751 * -------------------------------------
752 * | +0 | chunk root |
753 * -------------------------------------
755 chunk_bytenr = sys_chunk_start;
756 root_bytenr = meta_chunk_start;
757 extent_bytenr = meta_chunk_start + cfg->nodesize;
758 dev_bytenr = meta_chunk_start + cfg->nodesize * 2;
759 fs_bytenr = meta_chunk_start + cfg->nodesize * 3;
760 csum_bytenr = meta_chunk_start + cfg->nodesize * 4;
762 ret = setup_temp_super(fd, cfg, root_bytenr, chunk_bytenr);
766 ret = setup_temp_root_tree(fd, cfg, root_bytenr, extent_bytenr,
767 dev_bytenr, fs_bytenr, csum_bytenr);
770 ret = setup_temp_chunk_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
774 ret = setup_temp_dev_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
778 ret = setup_temp_fs_tree(fd, cfg, fs_bytenr);
785 * @fs_uuid - if NULL, generates a UUID, returns back the new filesystem UUID
787 int make_btrfs(int fd, struct btrfs_mkfs_config *cfg,
788 struct btrfs_convert_context *cctx)
790 struct btrfs_super_block super;
791 struct extent_buffer *buf;
792 struct btrfs_root_item root_item;
793 struct btrfs_disk_key disk_key;
794 struct btrfs_extent_item *extent_item;
795 struct btrfs_inode_item *inode_item;
796 struct btrfs_chunk *chunk;
797 struct btrfs_dev_item *dev_item;
798 struct btrfs_dev_extent *dev_extent;
799 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
809 int skinny_metadata = !!(cfg->features &
810 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA);
814 return make_convert_btrfs(fd, cfg, cctx);
815 buf = malloc(sizeof(*buf) + max(cfg->sectorsize, cfg->nodesize));
819 first_free = BTRFS_SUPER_INFO_OFFSET + cfg->sectorsize * 2 - 1;
820 first_free &= ~((u64)cfg->sectorsize - 1);
822 memset(&super, 0, sizeof(super));
824 num_bytes = (cfg->num_bytes / cfg->sectorsize) * cfg->sectorsize;
825 if (cfg->fs_uuid && *cfg->fs_uuid) {
826 if (uuid_parse(cfg->fs_uuid, super.fsid) != 0) {
827 error("cannot not parse UUID: %s", cfg->fs_uuid);
831 if (!test_uuid_unique(cfg->fs_uuid)) {
832 error("non-unique UUID: %s", cfg->fs_uuid);
837 uuid_generate(super.fsid);
839 uuid_unparse(super.fsid, cfg->fs_uuid);
841 uuid_generate(super.dev_item.uuid);
842 uuid_generate(chunk_tree_uuid);
844 btrfs_set_super_bytenr(&super, cfg->blocks[0]);
845 btrfs_set_super_num_devices(&super, 1);
846 btrfs_set_super_magic(&super, BTRFS_MAGIC);
847 btrfs_set_super_generation(&super, 1);
848 btrfs_set_super_root(&super, cfg->blocks[1]);
849 btrfs_set_super_chunk_root(&super, cfg->blocks[3]);
850 btrfs_set_super_total_bytes(&super, num_bytes);
851 btrfs_set_super_bytes_used(&super, 6 * cfg->nodesize);
852 btrfs_set_super_sectorsize(&super, cfg->sectorsize);
853 btrfs_set_super_leafsize(&super, cfg->nodesize);
854 btrfs_set_super_nodesize(&super, cfg->nodesize);
855 btrfs_set_super_stripesize(&super, cfg->stripesize);
856 btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
857 btrfs_set_super_chunk_root_generation(&super, 1);
858 btrfs_set_super_cache_generation(&super, -1);
859 btrfs_set_super_incompat_flags(&super, cfg->features);
861 __strncpy_null(super.label, cfg->label, BTRFS_LABEL_SIZE - 1);
863 /* create the tree of root objects */
864 memset(buf->data, 0, cfg->nodesize);
865 buf->len = cfg->nodesize;
866 btrfs_set_header_bytenr(buf, cfg->blocks[1]);
867 btrfs_set_header_nritems(buf, 4);
868 btrfs_set_header_generation(buf, 1);
869 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
870 btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
871 write_extent_buffer(buf, super.fsid, btrfs_header_fsid(),
874 write_extent_buffer(buf, chunk_tree_uuid,
875 btrfs_header_chunk_tree_uuid(buf),
878 /* create the items for the root tree */
879 memset(&root_item, 0, sizeof(root_item));
880 inode_item = &root_item.inode;
881 btrfs_set_stack_inode_generation(inode_item, 1);
882 btrfs_set_stack_inode_size(inode_item, 3);
883 btrfs_set_stack_inode_nlink(inode_item, 1);
884 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
885 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
886 btrfs_set_root_refs(&root_item, 1);
887 btrfs_set_root_used(&root_item, cfg->nodesize);
888 btrfs_set_root_generation(&root_item, 1);
890 memset(&disk_key, 0, sizeof(disk_key));
891 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
892 btrfs_set_disk_key_offset(&disk_key, 0);
895 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - sizeof(root_item);
896 btrfs_set_root_bytenr(&root_item, cfg->blocks[2]);
897 btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
898 btrfs_set_item_key(buf, &disk_key, nritems);
899 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
900 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
902 write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
903 nritems), sizeof(root_item));
906 itemoff = itemoff - sizeof(root_item);
907 btrfs_set_root_bytenr(&root_item, cfg->blocks[4]);
908 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
909 btrfs_set_item_key(buf, &disk_key, nritems);
910 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
911 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
913 write_extent_buffer(buf, &root_item,
914 btrfs_item_ptr_offset(buf, nritems),
918 itemoff = itemoff - sizeof(root_item);
919 btrfs_set_root_bytenr(&root_item, cfg->blocks[5]);
920 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
921 btrfs_set_item_key(buf, &disk_key, nritems);
922 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
923 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
925 write_extent_buffer(buf, &root_item,
926 btrfs_item_ptr_offset(buf, nritems),
930 itemoff = itemoff - sizeof(root_item);
931 btrfs_set_root_bytenr(&root_item, cfg->blocks[6]);
932 btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_TREE_OBJECTID);
933 btrfs_set_item_key(buf, &disk_key, nritems);
934 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
935 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
937 write_extent_buffer(buf, &root_item,
938 btrfs_item_ptr_offset(buf, nritems),
943 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
944 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[1]);
945 if (ret != cfg->nodesize) {
946 ret = (ret < 0 ? -errno : -EIO);
950 /* create the items for the extent tree */
951 memset(buf->data + sizeof(struct btrfs_header), 0,
952 cfg->nodesize - sizeof(struct btrfs_header));
954 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
955 for (i = 1; i < 7; i++) {
956 item_size = sizeof(struct btrfs_extent_item);
957 if (!skinny_metadata)
958 item_size += sizeof(struct btrfs_tree_block_info);
960 BUG_ON(cfg->blocks[i] < first_free);
961 BUG_ON(cfg->blocks[i] < cfg->blocks[i - 1]);
963 /* create extent item */
964 itemoff -= item_size;
965 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
966 if (skinny_metadata) {
967 btrfs_set_disk_key_type(&disk_key,
968 BTRFS_METADATA_ITEM_KEY);
969 btrfs_set_disk_key_offset(&disk_key, 0);
971 btrfs_set_disk_key_type(&disk_key,
972 BTRFS_EXTENT_ITEM_KEY);
973 btrfs_set_disk_key_offset(&disk_key, cfg->nodesize);
975 btrfs_set_item_key(buf, &disk_key, nritems);
976 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
978 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
980 extent_item = btrfs_item_ptr(buf, nritems,
981 struct btrfs_extent_item);
982 btrfs_set_extent_refs(buf, extent_item, 1);
983 btrfs_set_extent_generation(buf, extent_item, 1);
984 btrfs_set_extent_flags(buf, extent_item,
985 BTRFS_EXTENT_FLAG_TREE_BLOCK);
988 /* create extent ref */
989 ref_root = reference_root_table[i];
990 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
991 btrfs_set_disk_key_offset(&disk_key, ref_root);
992 btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
993 btrfs_set_item_key(buf, &disk_key, nritems);
994 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
996 btrfs_set_item_size(buf, btrfs_item_nr(nritems), 0);
999 btrfs_set_header_bytenr(buf, cfg->blocks[2]);
1000 btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
1001 btrfs_set_header_nritems(buf, nritems);
1002 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1003 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[2]);
1004 if (ret != cfg->nodesize) {
1005 ret = (ret < 0 ? -errno : -EIO);
1009 /* create the chunk tree */
1010 memset(buf->data + sizeof(struct btrfs_header), 0,
1011 cfg->nodesize - sizeof(struct btrfs_header));
1013 item_size = sizeof(*dev_item);
1014 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - item_size;
1016 /* first device 1 (there is no device 0) */
1017 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
1018 btrfs_set_disk_key_offset(&disk_key, 1);
1019 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
1020 btrfs_set_item_key(buf, &disk_key, nritems);
1021 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1022 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
1024 dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
1025 btrfs_set_device_id(buf, dev_item, 1);
1026 btrfs_set_device_generation(buf, dev_item, 0);
1027 btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
1028 btrfs_set_device_bytes_used(buf, dev_item,
1029 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1030 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
1031 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
1032 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
1033 btrfs_set_device_type(buf, dev_item, 0);
1035 write_extent_buffer(buf, super.dev_item.uuid,
1036 (unsigned long)btrfs_device_uuid(dev_item),
1038 write_extent_buffer(buf, super.fsid,
1039 (unsigned long)btrfs_device_fsid(dev_item),
1041 read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
1045 item_size = btrfs_chunk_item_size(1);
1046 itemoff = itemoff - item_size;
1048 /* then we have chunk 0 */
1049 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1050 btrfs_set_disk_key_offset(&disk_key, 0);
1051 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
1052 btrfs_set_item_key(buf, &disk_key, nritems);
1053 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1054 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
1056 chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
1057 btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1058 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
1059 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
1060 btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
1061 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
1062 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
1063 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
1064 btrfs_set_chunk_num_stripes(buf, chunk, 1);
1065 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
1066 btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
1069 write_extent_buffer(buf, super.dev_item.uuid,
1070 (unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
1073 /* copy the key for the chunk to the system array */
1074 ptr = super.sys_chunk_array;
1075 array_size = sizeof(disk_key);
1077 memcpy(ptr, &disk_key, sizeof(disk_key));
1078 ptr += sizeof(disk_key);
1080 /* copy the chunk to the system array */
1081 read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
1082 array_size += item_size;
1084 btrfs_set_super_sys_array_size(&super, array_size);
1086 btrfs_set_header_bytenr(buf, cfg->blocks[3]);
1087 btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
1088 btrfs_set_header_nritems(buf, nritems);
1089 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1090 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[3]);
1091 if (ret != cfg->nodesize) {
1092 ret = (ret < 0 ? -errno : -EIO);
1096 /* create the device tree */
1097 memset(buf->data + sizeof(struct btrfs_header), 0,
1098 cfg->nodesize - sizeof(struct btrfs_header));
1100 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) -
1101 sizeof(struct btrfs_dev_extent);
1103 btrfs_set_disk_key_objectid(&disk_key, 1);
1104 btrfs_set_disk_key_offset(&disk_key, 0);
1105 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
1106 btrfs_set_item_key(buf, &disk_key, nritems);
1107 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1108 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1109 sizeof(struct btrfs_dev_extent));
1110 dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
1111 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
1112 BTRFS_CHUNK_TREE_OBJECTID);
1113 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
1114 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1115 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
1117 write_extent_buffer(buf, chunk_tree_uuid,
1118 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
1121 btrfs_set_dev_extent_length(buf, dev_extent,
1122 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1125 btrfs_set_header_bytenr(buf, cfg->blocks[4]);
1126 btrfs_set_header_owner(buf, BTRFS_DEV_TREE_OBJECTID);
1127 btrfs_set_header_nritems(buf, nritems);
1128 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1129 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[4]);
1130 if (ret != cfg->nodesize) {
1131 ret = (ret < 0 ? -errno : -EIO);
1135 /* create the FS root */
1136 memset(buf->data + sizeof(struct btrfs_header), 0,
1137 cfg->nodesize - sizeof(struct btrfs_header));
1138 btrfs_set_header_bytenr(buf, cfg->blocks[5]);
1139 btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
1140 btrfs_set_header_nritems(buf, 0);
1141 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1142 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[5]);
1143 if (ret != cfg->nodesize) {
1144 ret = (ret < 0 ? -errno : -EIO);
1147 /* finally create the csum root */
1148 memset(buf->data + sizeof(struct btrfs_header), 0,
1149 cfg->nodesize - sizeof(struct btrfs_header));
1150 btrfs_set_header_bytenr(buf, cfg->blocks[6]);
1151 btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
1152 btrfs_set_header_nritems(buf, 0);
1153 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1154 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[6]);
1155 if (ret != cfg->nodesize) {
1156 ret = (ret < 0 ? -errno : -EIO);
1160 /* and write out the super block */
1161 BUG_ON(sizeof(super) > cfg->sectorsize);
1162 memset(buf->data, 0, BTRFS_SUPER_INFO_SIZE);
1163 memcpy(buf->data, &super, sizeof(super));
1164 buf->len = BTRFS_SUPER_INFO_SIZE;
1165 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1166 ret = pwrite(fd, buf->data, BTRFS_SUPER_INFO_SIZE, cfg->blocks[0]);
1167 if (ret != BTRFS_SUPER_INFO_SIZE) {
1168 ret = (ret < 0 ? -errno : -EIO);
1179 static const struct btrfs_fs_feature {
1183 } mkfs_features[] = {
1184 { "mixed-bg", BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS,
1185 "mixed data and metadata block groups" },
1186 { "extref", BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF,
1187 "increased hardlink limit per file to 65536" },
1188 { "raid56", BTRFS_FEATURE_INCOMPAT_RAID56,
1189 "raid56 extended format" },
1190 { "skinny-metadata", BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA,
1191 "reduced-size metadata extent refs" },
1192 { "no-holes", BTRFS_FEATURE_INCOMPAT_NO_HOLES,
1193 "no explicit hole extents for files" },
1194 /* Keep this one last */
1195 { "list-all", BTRFS_FEATURE_LIST_ALL, NULL }
1198 static int parse_one_fs_feature(const char *name, u64 *flags)
1203 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1204 if (name[0] == '^' &&
1205 !strcmp(mkfs_features[i].name, name + 1)) {
1206 *flags &= ~ mkfs_features[i].flag;
1208 } else if (!strcmp(mkfs_features[i].name, name)) {
1209 *flags |= mkfs_features[i].flag;
1217 void btrfs_parse_features_to_string(char *buf, u64 flags)
1223 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1224 if (flags & mkfs_features[i].flag) {
1227 strcat(buf, mkfs_features[i].name);
1232 void btrfs_process_fs_features(u64 flags)
1236 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1237 if (flags & mkfs_features[i].flag) {
1238 printf("Turning ON incompat feature '%s': %s\n",
1239 mkfs_features[i].name,
1240 mkfs_features[i].desc);
1245 void btrfs_list_all_fs_features(u64 mask_disallowed)
1249 fprintf(stderr, "Filesystem features available:\n");
1250 for (i = 0; i < ARRAY_SIZE(mkfs_features) - 1; i++) {
1251 char *is_default = "";
1253 if (mkfs_features[i].flag & mask_disallowed)
1255 if (mkfs_features[i].flag & BTRFS_MKFS_DEFAULT_FEATURES)
1256 is_default = ", default";
1257 fprintf(stderr, "%-20s- %s (0x%llx%s)\n",
1258 mkfs_features[i].name,
1259 mkfs_features[i].desc,
1260 mkfs_features[i].flag,
1266 * Return NULL if all features were parsed fine, otherwise return the name of
1267 * the first unparsed.
1269 char* btrfs_parse_fs_features(char *namelist, u64 *flags)
1272 char *save_ptr = NULL; /* Satisfy static checkers */
1274 for (this_char = strtok_r(namelist, ",", &save_ptr);
1276 this_char = strtok_r(NULL, ",", &save_ptr)) {
1277 if (parse_one_fs_feature(this_char, flags))
1284 u64 btrfs_device_size(int fd, struct stat *st)
1287 if (S_ISREG(st->st_mode)) {
1290 if (!S_ISBLK(st->st_mode)) {
1293 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
1299 static int zero_blocks(int fd, off_t start, size_t len)
1301 char *buf = malloc(len);
1307 memset(buf, 0, len);
1308 written = pwrite(fd, buf, len, start);
1315 #define ZERO_DEV_BYTES (2 * 1024 * 1024)
1317 /* don't write outside the device by clamping the region to the device size */
1318 static int zero_dev_clamped(int fd, off_t start, ssize_t len, u64 dev_size)
1320 off_t end = max(start, start + len);
1323 /* and don't overwrite the disk labels on sparc */
1324 start = max(start, 1024);
1325 end = max(end, 1024);
1328 start = min_t(u64, start, dev_size);
1329 end = min_t(u64, end, dev_size);
1331 return zero_blocks(fd, start, end - start);
1334 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
1335 struct btrfs_root *root, int fd, char *path,
1336 u64 device_total_bytes, u32 io_width, u32 io_align,
1339 struct btrfs_super_block *disk_super;
1340 struct btrfs_super_block *super = root->fs_info->super_copy;
1341 struct btrfs_device *device;
1342 struct btrfs_dev_item *dev_item;
1348 device_total_bytes = (device_total_bytes / sectorsize) * sectorsize;
1350 device = kzalloc(sizeof(*device), GFP_NOFS);
1353 buf = kzalloc(sectorsize, GFP_NOFS);
1356 BUG_ON(sizeof(*disk_super) > sectorsize);
1358 disk_super = (struct btrfs_super_block *)buf;
1359 dev_item = &disk_super->dev_item;
1361 uuid_generate(device->uuid);
1364 device->io_width = io_width;
1365 device->io_align = io_align;
1366 device->sector_size = sectorsize;
1368 device->writeable = 1;
1369 device->total_bytes = device_total_bytes;
1370 device->bytes_used = 0;
1371 device->total_ios = 0;
1372 device->dev_root = root->fs_info->dev_root;
1373 device->name = strdup(path);
1377 INIT_LIST_HEAD(&device->dev_list);
1378 ret = btrfs_add_device(trans, root, device);
1381 fs_total_bytes = btrfs_super_total_bytes(super) + device_total_bytes;
1382 btrfs_set_super_total_bytes(super, fs_total_bytes);
1384 num_devs = btrfs_super_num_devices(super) + 1;
1385 btrfs_set_super_num_devices(super, num_devs);
1387 memcpy(disk_super, super, sizeof(*disk_super));
1389 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
1390 btrfs_set_stack_device_id(dev_item, device->devid);
1391 btrfs_set_stack_device_type(dev_item, device->type);
1392 btrfs_set_stack_device_io_align(dev_item, device->io_align);
1393 btrfs_set_stack_device_io_width(dev_item, device->io_width);
1394 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
1395 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
1396 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
1397 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
1399 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
1400 BUG_ON(ret != sectorsize);
1403 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
1404 device->fs_devices = root->fs_info->fs_devices;
1413 static int btrfs_wipe_existing_sb(int fd)
1415 const char *off = NULL;
1420 blkid_probe pr = NULL;
1422 pr = blkid_new_probe();
1426 if (blkid_probe_set_device(pr, fd, 0, 0)) {
1431 ret = blkid_probe_lookup_value(pr, "SBMAGIC_OFFSET", &off, NULL);
1433 ret = blkid_probe_lookup_value(pr, "SBMAGIC", NULL, &len);
1435 if (ret || len == 0 || off == NULL) {
1437 * If lookup fails, the probe did not find any values, eg. for
1438 * a file image or a loop device. Soft error.
1444 offset = strtoll(off, NULL, 10);
1445 if (len > sizeof(buf))
1448 memset(buf, 0, len);
1449 ret = pwrite(fd, buf, len, offset);
1451 error("cannot wipe existing superblock: %s", strerror(errno));
1453 } else if (ret != len) {
1454 error("cannot wipe existing superblock: wrote %d of %zd", ret, len);
1460 blkid_free_probe(pr);
1464 int btrfs_prepare_device(int fd, const char *file, int zero_end,
1465 u64 *block_count_ret, u64 max_block_count, int discard)
1471 ret = fstat(fd, &st);
1473 error("unable to stat %s: %s", file, strerror(errno));
1477 block_count = btrfs_device_size(fd, &st);
1478 if (block_count == 0) {
1479 error("unable to determine size of %s", file);
1482 if (max_block_count)
1483 block_count = min(block_count, max_block_count);
1487 * We intentionally ignore errors from the discard ioctl. It
1488 * is not necessary for the mkfs functionality but just an
1491 if (discard_range(fd, 0, 0) == 0) {
1492 printf("Performing full device TRIM (%s) ...\n",
1493 pretty_size(block_count));
1494 discard_blocks(fd, 0, block_count);
1498 ret = zero_dev_clamped(fd, 0, ZERO_DEV_BYTES, block_count);
1499 for (i = 0 ; !ret && i < BTRFS_SUPER_MIRROR_MAX; i++)
1500 ret = zero_dev_clamped(fd, btrfs_sb_offset(i),
1501 BTRFS_SUPER_INFO_SIZE, block_count);
1502 if (!ret && zero_end)
1503 ret = zero_dev_clamped(fd, block_count - ZERO_DEV_BYTES,
1504 ZERO_DEV_BYTES, block_count);
1507 error("failed to zero device '%s': %s", file, strerror(-ret));
1511 ret = btrfs_wipe_existing_sb(fd);
1513 error("cannot wipe superblocks on %s", file);
1517 *block_count_ret = block_count;
1521 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
1522 struct btrfs_root *root, u64 objectid)
1525 struct btrfs_inode_item inode_item;
1526 time_t now = time(NULL);
1528 memset(&inode_item, 0, sizeof(inode_item));
1529 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
1530 btrfs_set_stack_inode_size(&inode_item, 0);
1531 btrfs_set_stack_inode_nlink(&inode_item, 1);
1532 btrfs_set_stack_inode_nbytes(&inode_item, root->nodesize);
1533 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
1534 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
1535 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
1536 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
1537 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
1538 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
1539 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
1540 btrfs_set_stack_timespec_sec(&inode_item.otime, 0);
1541 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
1543 if (root->fs_info->tree_root == root)
1544 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
1546 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
1550 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
1554 btrfs_set_root_dirid(&root->root_item, objectid);
1561 * checks if a path is a block device node
1562 * Returns negative errno on failure, otherwise
1563 * returns 1 for blockdev, 0 for not-blockdev
1565 int is_block_device(const char *path)
1567 struct stat statbuf;
1569 if (stat(path, &statbuf) < 0)
1572 return !!S_ISBLK(statbuf.st_mode);
1576 * check if given path is a mount point
1577 * return 1 if yes. 0 if no. -1 for error
1579 int is_mount_point(const char *path)
1585 f = setmntent("/proc/self/mounts", "r");
1589 while ((mnt = getmntent(f)) != NULL) {
1590 if (strcmp(mnt->mnt_dir, path))
1599 static int is_reg_file(const char *path)
1601 struct stat statbuf;
1603 if (stat(path, &statbuf) < 0)
1605 return S_ISREG(statbuf.st_mode);
1609 * This function checks if the given input parameter is
1611 * return <0 : some error in the given input
1612 * return BTRFS_ARG_UNKNOWN: unknown input
1613 * return BTRFS_ARG_UUID: given input is uuid
1614 * return BTRFS_ARG_MNTPOINT: given input is path
1615 * return BTRFS_ARG_REG: given input is regular file
1616 * return BTRFS_ARG_BLKDEV: given input is block device
1618 int check_arg_type(const char *input)
1621 char path[PATH_MAX];
1626 if (realpath(input, path)) {
1627 if (is_block_device(path) == 1)
1628 return BTRFS_ARG_BLKDEV;
1630 if (is_mount_point(path) == 1)
1631 return BTRFS_ARG_MNTPOINT;
1633 if (is_reg_file(path))
1634 return BTRFS_ARG_REG;
1636 return BTRFS_ARG_UNKNOWN;
1639 if (strlen(input) == (BTRFS_UUID_UNPARSED_SIZE - 1) &&
1640 !uuid_parse(input, uuid))
1641 return BTRFS_ARG_UUID;
1643 return BTRFS_ARG_UNKNOWN;
1647 * Find the mount point for a mounted device.
1648 * On success, returns 0 with mountpoint in *mp.
1649 * On failure, returns -errno (not mounted yields -EINVAL)
1650 * Is noisy on failures, expects to be given a mounted device.
1652 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size)
1657 ret = is_block_device(dev);
1660 error("not a block device: %s", dev);
1663 error("cannot check %s: %s", dev, strerror(-ret));
1668 fd = open(dev, O_RDONLY);
1671 error("cannot open %s: %s", dev, strerror(errno));
1675 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
1678 } else { /* mounted, all good */
1688 * Given a pathname, return a filehandle to:
1689 * the original pathname or,
1690 * if the pathname is a mounted btrfs device, to its mountpoint.
1692 * On error, return -1, errno should be set.
1694 int open_path_or_dev_mnt(const char *path, DIR **dirstream, int verbose)
1699 if (is_block_device(path)) {
1700 ret = get_btrfs_mount(path, mp, sizeof(mp));
1702 /* not a mounted btrfs dev */
1703 error_on(verbose, "'%s' is not a mounted btrfs device",
1708 ret = open_file_or_dir(mp, dirstream);
1709 error_on(verbose && ret < 0, "can't access '%s': %s",
1710 path, strerror(errno));
1712 ret = btrfs_open_dir(path, dirstream, 1);
1719 * Do the following checks before calling open_file_or_dir():
1720 * 1: path is in a btrfs filesystem
1721 * 2: path is a directory
1723 int btrfs_open_dir(const char *path, DIR **dirstream, int verbose)
1729 if (statfs(path, &stfs) != 0) {
1730 error_on(verbose, "cannot access '%s': %s", path,
1735 if (stfs.f_type != BTRFS_SUPER_MAGIC) {
1736 error_on(verbose, "not a btrfs filesystem: %s", path);
1740 if (stat(path, &st) != 0) {
1741 error_on(verbose, "cannot access '%s': %s", path,
1746 if (!S_ISDIR(st.st_mode)) {
1747 error_on(verbose, "not a directory: %s", path);
1751 ret = open_file_or_dir(path, dirstream);
1753 error_on(verbose, "cannot access '%s': %s", path,
1760 /* checks if a device is a loop device */
1761 static int is_loop_device (const char* device) {
1762 struct stat statbuf;
1764 if(stat(device, &statbuf) < 0)
1767 return (S_ISBLK(statbuf.st_mode) &&
1768 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
1772 * Takes a loop device path (e.g. /dev/loop0) and returns
1773 * the associated file (e.g. /images/my_btrfs.img) using
1776 static int resolve_loop_device_with_loopdev(const char* loop_dev, char* loop_file)
1780 struct loop_info64 lo64;
1782 fd = open(loop_dev, O_RDONLY | O_NONBLOCK);
1785 ret = ioctl(fd, LOOP_GET_STATUS64, &lo64);
1791 memcpy(loop_file, lo64.lo_file_name, sizeof(lo64.lo_file_name));
1792 loop_file[sizeof(lo64.lo_file_name)] = 0;
1800 /* Takes a loop device path (e.g. /dev/loop0) and returns
1801 * the associated file (e.g. /images/my_btrfs.img) */
1802 static int resolve_loop_device(const char* loop_dev, char* loop_file,
1809 char real_loop_dev[PATH_MAX];
1811 if (!realpath(loop_dev, real_loop_dev))
1813 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
1814 if (!(f = fopen(p, "r"))) {
1815 if (errno == ENOENT)
1817 * It's possibly a partitioned loop device, which is
1818 * resolvable with loopdev API.
1820 return resolve_loop_device_with_loopdev(loop_dev, loop_file);
1824 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
1825 ret = fscanf(f, fmt, loop_file);
1834 * Checks whether a and b are identical or device
1835 * files associated with the same block device
1837 static int is_same_blk_file(const char* a, const char* b)
1839 struct stat st_buf_a, st_buf_b;
1840 char real_a[PATH_MAX];
1841 char real_b[PATH_MAX];
1843 if (!realpath(a, real_a))
1844 strncpy_null(real_a, a);
1846 if (!realpath(b, real_b))
1847 strncpy_null(real_b, b);
1849 /* Identical path? */
1850 if (strcmp(real_a, real_b) == 0)
1853 if (stat(a, &st_buf_a) < 0 || stat(b, &st_buf_b) < 0) {
1854 if (errno == ENOENT)
1859 /* Same blockdevice? */
1860 if (S_ISBLK(st_buf_a.st_mode) && S_ISBLK(st_buf_b.st_mode) &&
1861 st_buf_a.st_rdev == st_buf_b.st_rdev) {
1866 if (st_buf_a.st_dev == st_buf_b.st_dev &&
1867 st_buf_a.st_ino == st_buf_b.st_ino) {
1874 /* checks if a and b are identical or device
1875 * files associated with the same block device or
1876 * if one file is a loop device that uses the other
1879 static int is_same_loop_file(const char* a, const char* b)
1881 char res_a[PATH_MAX];
1882 char res_b[PATH_MAX];
1883 const char* final_a = NULL;
1884 const char* final_b = NULL;
1887 /* Resolve a if it is a loop device */
1888 if((ret = is_loop_device(a)) < 0) {
1893 ret = resolve_loop_device(a, res_a, sizeof(res_a));
1904 /* Resolve b if it is a loop device */
1905 if ((ret = is_loop_device(b)) < 0) {
1910 ret = resolve_loop_device(b, res_b, sizeof(res_b));
1921 return is_same_blk_file(final_a, final_b);
1924 /* Checks if a file exists and is a block or regular file*/
1925 static int is_existing_blk_or_reg_file(const char* filename)
1929 if(stat(filename, &st_buf) < 0) {
1936 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
1939 /* Checks if a file is used (directly or indirectly via a loop device)
1940 * by a device in fs_devices
1942 static int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices,
1946 struct list_head *head;
1947 struct list_head *cur;
1948 struct btrfs_device *device;
1950 head = &fs_devices->devices;
1951 list_for_each(cur, head) {
1952 device = list_entry(cur, struct btrfs_device, dev_list);
1954 if((ret = is_same_loop_file(device->name, file)))
1962 * Resolve a pathname to a device mapper node to /dev/mapper/<name>
1963 * Returns NULL on invalid input or malloc failure; Other failures
1964 * will be handled by the caller using the input pathame.
1966 char *canonicalize_dm_name(const char *ptname)
1970 char path[PATH_MAX], name[PATH_MAX], *res = NULL;
1972 if (!ptname || !*ptname)
1975 snprintf(path, sizeof(path), "/sys/block/%s/dm/name", ptname);
1976 if (!(f = fopen(path, "r")))
1979 /* read <name>\n from sysfs */
1980 if (fgets(name, sizeof(name), f) && (sz = strlen(name)) > 1) {
1981 name[sz - 1] = '\0';
1982 snprintf(path, sizeof(path), "/dev/mapper/%s", name);
1984 if (access(path, F_OK) == 0)
1992 * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or
1993 * to a device mapper pathname.
1994 * Returns NULL on invalid input or malloc failure; Other failures
1995 * will be handled by the caller using the input pathame.
1997 char *canonicalize_path(const char *path)
1999 char *canonical, *p;
2001 if (!path || !*path)
2004 canonical = realpath(path, NULL);
2006 return strdup(path);
2007 p = strrchr(canonical, '/');
2008 if (p && strncmp(p, "/dm-", 4) == 0 && isdigit(*(p + 4))) {
2009 char *dm = canonicalize_dm_name(p + 1);
2020 * returns 1 if the device was mounted, < 0 on error or 0 if everything
2021 * is safe to continue.
2023 int check_mounted(const char* file)
2028 fd = open(file, O_RDONLY);
2030 error("mount check: cannot open %s: %s", file,
2035 ret = check_mounted_where(fd, file, NULL, 0, NULL);
2041 int check_mounted_where(int fd, const char *file, char *where, int size,
2042 struct btrfs_fs_devices **fs_dev_ret)
2047 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2051 /* scan the initial device */
2052 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
2053 &total_devs, BTRFS_SUPER_INFO_OFFSET, 0);
2054 is_btrfs = (ret >= 0);
2056 /* scan other devices */
2057 if (is_btrfs && total_devs > 1) {
2058 ret = btrfs_scan_lblkid();
2063 /* iterate over the list of currently mounted filesystems */
2064 if ((f = setmntent ("/proc/self/mounts", "r")) == NULL)
2067 while ((mnt = getmntent (f)) != NULL) {
2069 if(strcmp(mnt->mnt_type, "btrfs") != 0)
2072 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
2074 /* ignore entries in the mount table that are not
2075 associated with a file*/
2076 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
2077 goto out_mntloop_err;
2081 ret = is_same_loop_file(file, mnt->mnt_fsname);
2085 goto out_mntloop_err;
2090 /* Did we find an entry in mnt table? */
2091 if (mnt && size && where) {
2092 strncpy(where, mnt->mnt_dir, size);
2096 *fs_dev_ret = fs_devices_mnt;
2098 ret = (mnt != NULL);
2106 struct pending_dir {
2107 struct list_head list;
2108 char name[PATH_MAX];
2111 int btrfs_register_one_device(const char *fname)
2113 struct btrfs_ioctl_vol_args args;
2117 fd = open("/dev/btrfs-control", O_RDWR);
2120 "failed to open /dev/btrfs-control, skipping device registration: %s",
2124 memset(&args, 0, sizeof(args));
2125 strncpy_null(args.name, fname);
2126 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
2128 error("device scan failed on '%s': %s", fname,
2137 * Register all devices in the fs_uuid list created in the user
2138 * space. Ensure btrfs_scan_lblkid() is called before this func.
2140 int btrfs_register_all_devices(void)
2144 struct btrfs_fs_devices *fs_devices;
2145 struct btrfs_device *device;
2146 struct list_head *all_uuids;
2148 all_uuids = btrfs_scanned_uuids();
2150 list_for_each_entry(fs_devices, all_uuids, list) {
2151 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2153 err = btrfs_register_one_device(device->name);
2163 int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
2166 struct btrfs_super_block *disk_super;
2170 buf = malloc(BTRFS_SUPER_INFO_SIZE);
2175 ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
2176 if (ret != BTRFS_SUPER_INFO_SIZE)
2180 disk_super = (struct btrfs_super_block *)buf;
2181 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC)
2184 if (!memcmp(disk_super->fsid, root->fs_info->super_copy->fsid,
2194 * Note: this function uses a static per-thread buffer. Do not call this
2195 * function more than 10 times within one argument list!
2197 const char *pretty_size_mode(u64 size, unsigned mode)
2199 static __thread int ps_index = 0;
2200 static __thread char ps_array[10][32];
2203 ret = ps_array[ps_index];
2206 (void)pretty_size_snprintf(size, ret, 32, mode);
2211 static const char* unit_suffix_binary[] =
2212 { "B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB"};
2213 static const char* unit_suffix_decimal[] =
2214 { "B", "kB", "MB", "GB", "TB", "PB", "EB"};
2216 int pretty_size_snprintf(u64 size, char *str, size_t str_size, unsigned unit_mode)
2222 const char** suffix = NULL;
2228 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_RAW) {
2229 snprintf(str, str_size, "%llu", size);
2233 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_BINARY) {
2236 suffix = unit_suffix_binary;
2237 } else if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_DECIMAL) {
2240 suffix = unit_suffix_decimal;
2245 fprintf(stderr, "INTERNAL ERROR: unknown unit base, mode %d\n",
2253 switch (unit_mode & UNITS_MODE_MASK) {
2254 case UNITS_TBYTES: base *= mult; num_divs++;
2255 case UNITS_GBYTES: base *= mult; num_divs++;
2256 case UNITS_MBYTES: base *= mult; num_divs++;
2257 case UNITS_KBYTES: num_divs++;
2264 while (size >= mult) {
2270 * If the value is smaller than base, we didn't do any
2271 * division, in that case, base should be 1, not original
2272 * base, or the unit will be wrong
2278 if (num_divs >= ARRAY_SIZE(unit_suffix_binary)) {
2280 printf("INTERNAL ERROR: unsupported unit suffix, index %d\n",
2285 fraction = (float)last_size / base;
2287 return snprintf(str, str_size, "%.2f%s", fraction, suffix[num_divs]);
2291 * __strncpy_null - strncpy with null termination
2292 * @dest: the target array
2293 * @src: the source string
2294 * @n: maximum bytes to copy (size of *dest)
2296 * Like strncpy, but ensures destination is null-terminated.
2298 * Copies the string pointed to by src, including the terminating null
2299 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
2300 * of n bytes. Then ensure that dest is null-terminated.
2302 char *__strncpy_null(char *dest, const char *src, size_t n)
2304 strncpy(dest, src, n);
2311 * Checks to make sure that the label matches our requirements.
2313 0 if everything is safe and usable
2314 -1 if the label is too long
2316 static int check_label(const char *input)
2318 int len = strlen(input);
2320 if (len > BTRFS_LABEL_SIZE - 1) {
2321 error("label %s is too long (max %d)", input,
2322 BTRFS_LABEL_SIZE - 1);
2329 static int set_label_unmounted(const char *dev, const char *label)
2331 struct btrfs_trans_handle *trans;
2332 struct btrfs_root *root;
2335 ret = check_mounted(dev);
2337 error("checking mount status of %s failed: %d", dev, ret);
2341 error("device %s is mounted, use mount point", dev);
2345 /* Open the super_block at the default location
2346 * and as read-write.
2348 root = open_ctree(dev, 0, OPEN_CTREE_WRITES);
2349 if (!root) /* errors are printed by open_ctree() */
2352 trans = btrfs_start_transaction(root, 1);
2353 __strncpy_null(root->fs_info->super_copy->label, label, BTRFS_LABEL_SIZE - 1);
2355 btrfs_commit_transaction(trans, root);
2357 /* Now we close it since we are done. */
2362 static int set_label_mounted(const char *mount_path, const char *labelp)
2365 char label[BTRFS_LABEL_SIZE];
2367 fd = open(mount_path, O_RDONLY | O_NOATIME);
2369 error("unable to access %s: %s", mount_path, strerror(errno));
2373 memset(label, 0, sizeof(label));
2374 __strncpy_null(label, labelp, BTRFS_LABEL_SIZE - 1);
2375 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
2376 error("unable to set label of %s: %s", mount_path,
2386 int get_label_unmounted(const char *dev, char *label)
2388 struct btrfs_root *root;
2391 ret = check_mounted(dev);
2393 error("checking mount status of %s failed: %d", dev, ret);
2397 /* Open the super_block at the default location
2400 root = open_ctree(dev, 0, 0);
2404 __strncpy_null(label, root->fs_info->super_copy->label,
2405 BTRFS_LABEL_SIZE - 1);
2407 /* Now we close it since we are done. */
2413 * If a partition is mounted, try to get the filesystem label via its
2414 * mounted path rather than device. Return the corresponding error
2415 * the user specified the device path.
2417 int get_label_mounted(const char *mount_path, char *labelp)
2419 char label[BTRFS_LABEL_SIZE];
2423 fd = open(mount_path, O_RDONLY | O_NOATIME);
2425 error("unable to access %s: %s", mount_path, strerror(errno));
2429 memset(label, '\0', sizeof(label));
2430 ret = ioctl(fd, BTRFS_IOC_GET_FSLABEL, label);
2432 if (errno != ENOTTY)
2433 error("unable to get label of %s: %s", mount_path,
2440 __strncpy_null(labelp, label, BTRFS_LABEL_SIZE - 1);
2445 int get_label(const char *btrfs_dev, char *label)
2449 ret = is_existing_blk_or_reg_file(btrfs_dev);
2451 ret = get_label_mounted(btrfs_dev, label);
2453 ret = get_label_unmounted(btrfs_dev, label);
2458 int set_label(const char *btrfs_dev, const char *label)
2462 if (check_label(label))
2465 ret = is_existing_blk_or_reg_file(btrfs_dev);
2467 ret = set_label_mounted(btrfs_dev, label);
2469 ret = set_label_unmounted(btrfs_dev, label);
2475 * A not-so-good version fls64. No fascinating optimization since
2476 * no one except parse_size use it
2478 static int fls64(u64 x)
2482 for (i = 0; i <64; i++)
2483 if (x << i & (1ULL << 63))
2488 u64 parse_size(char *s)
2496 error("size value is empty");
2500 error("size value '%s' is less equal than 0", s);
2503 ret = strtoull(s, &endptr, 10);
2505 error("size value '%s' is invalid", s);
2508 if (endptr[0] && endptr[1]) {
2509 error("illegal suffix contains character '%c' in wrong position",
2514 * strtoll returns LLONG_MAX when overflow, if this happens,
2515 * need to call strtoull to get the real size
2517 if (errno == ERANGE && ret == ULLONG_MAX) {
2518 error("size value '%s' is too large for u64", s);
2522 c = tolower(endptr[0]);
2545 error("unknown size descriptor '%c'", c);
2549 /* Check whether ret * mult overflow */
2550 if (fls64(ret) + fls64(mult) - 1 > 64) {
2551 error("size value '%s' is too large for u64", s);
2558 u64 parse_qgroupid(const char *p)
2560 char *s = strchr(p, '/');
2561 const char *ptr_src_end = p + strlen(p);
2562 char *ptr_parse_end = NULL;
2571 /* Numeric format like '0/257' is the primary case */
2573 id = strtoull(p, &ptr_parse_end, 10);
2574 if (ptr_parse_end != ptr_src_end)
2578 level = strtoull(p, &ptr_parse_end, 10);
2579 if (ptr_parse_end != s)
2582 id = strtoull(s + 1, &ptr_parse_end, 10);
2583 if (ptr_parse_end != ptr_src_end)
2586 return (level << BTRFS_QGROUP_LEVEL_SHIFT) | id;
2589 /* Path format like subv at 'my_subvol' is the fallback case */
2590 ret = test_issubvolume(p);
2591 if (ret < 0 || !ret)
2593 fd = open(p, O_RDONLY);
2596 ret = lookup_ino_rootid(fd, &id);
2603 error("invalid qgroupid or subvolume path: %s", p);
2607 int open_file_or_dir3(const char *fname, DIR **dirstream, int open_flags)
2613 ret = stat(fname, &st);
2617 if (S_ISDIR(st.st_mode)) {
2618 *dirstream = opendir(fname);
2621 fd = dirfd(*dirstream);
2622 } else if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) {
2623 fd = open(fname, open_flags);
2626 * we set this on purpose, in case the caller output
2627 * strerror(errno) as success
2635 closedir(*dirstream);
2642 int open_file_or_dir(const char *fname, DIR **dirstream)
2644 return open_file_or_dir3(fname, dirstream, O_RDWR);
2647 void close_file_or_dir(int fd, DIR *dirstream)
2650 closedir(dirstream);
2655 int get_device_info(int fd, u64 devid,
2656 struct btrfs_ioctl_dev_info_args *di_args)
2660 di_args->devid = devid;
2661 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
2663 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
2664 return ret < 0 ? -errno : 0;
2667 static u64 find_max_device_id(struct btrfs_ioctl_search_args *search_args,
2670 struct btrfs_dev_item *dev_item;
2671 char *buf = search_args->buf;
2673 buf += (nr_items - 1) * (sizeof(struct btrfs_ioctl_search_header)
2674 + sizeof(struct btrfs_dev_item));
2675 buf += sizeof(struct btrfs_ioctl_search_header);
2677 dev_item = (struct btrfs_dev_item *)buf;
2679 return btrfs_stack_device_id(dev_item);
2682 static int search_chunk_tree_for_fs_info(int fd,
2683 struct btrfs_ioctl_fs_info_args *fi_args)
2687 u64 start_devid = 1;
2688 struct btrfs_ioctl_search_args search_args;
2689 struct btrfs_ioctl_search_key *search_key = &search_args.key;
2691 fi_args->num_devices = 0;
2693 max_items = BTRFS_SEARCH_ARGS_BUFSIZE
2694 / (sizeof(struct btrfs_ioctl_search_header)
2695 + sizeof(struct btrfs_dev_item));
2697 search_key->tree_id = BTRFS_CHUNK_TREE_OBJECTID;
2698 search_key->min_objectid = BTRFS_DEV_ITEMS_OBJECTID;
2699 search_key->max_objectid = BTRFS_DEV_ITEMS_OBJECTID;
2700 search_key->min_type = BTRFS_DEV_ITEM_KEY;
2701 search_key->max_type = BTRFS_DEV_ITEM_KEY;
2702 search_key->min_transid = 0;
2703 search_key->max_transid = (u64)-1;
2704 search_key->nr_items = max_items;
2705 search_key->max_offset = (u64)-1;
2708 search_key->min_offset = start_devid;
2710 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH, &search_args);
2714 fi_args->num_devices += (u64)search_key->nr_items;
2716 if (search_key->nr_items == max_items) {
2717 start_devid = find_max_device_id(&search_args,
2718 search_key->nr_items) + 1;
2722 /* get the lastest max_id to stay consistent with the num_devices */
2723 if (search_key->nr_items == 0)
2725 * last tree_search returns an empty buf, use the devid of
2726 * the last dev_item of the previous tree_search
2728 fi_args->max_id = start_devid - 1;
2730 fi_args->max_id = find_max_device_id(&search_args,
2731 search_key->nr_items);
2737 * For a given path, fill in the ioctl fs_ and info_ args.
2738 * If the path is a btrfs mountpoint, fill info for all devices.
2739 * If the path is a btrfs device, fill in only that device.
2741 * The path provided must be either on a mounted btrfs fs,
2742 * or be a mounted btrfs device.
2744 * Returns 0 on success, or a negative errno.
2746 int get_fs_info(char *path, struct btrfs_ioctl_fs_info_args *fi_args,
2747 struct btrfs_ioctl_dev_info_args **di_ret)
2754 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2755 struct btrfs_ioctl_dev_info_args *di_args;
2756 struct btrfs_ioctl_dev_info_args tmp;
2758 DIR *dirstream = NULL;
2760 memset(fi_args, 0, sizeof(*fi_args));
2762 if (is_block_device(path) == 1) {
2763 struct btrfs_super_block *disk_super;
2764 char buf[BTRFS_SUPER_INFO_SIZE];
2767 /* Ensure it's mounted, then set path to the mountpoint */
2768 fd = open(path, O_RDONLY);
2771 error("cannot open %s: %s", path, strerror(errno));
2774 ret = check_mounted_where(fd, path, mp, sizeof(mp),
2783 /* Only fill in this one device */
2784 fi_args->num_devices = 1;
2786 disk_super = (struct btrfs_super_block *)buf;
2787 ret = btrfs_read_dev_super(fd, disk_super,
2788 BTRFS_SUPER_INFO_OFFSET, 0);
2793 devid = btrfs_stack_device_id(&disk_super->dev_item);
2795 fi_args->max_id = devid;
2798 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
2802 /* at this point path must not be for a block device */
2803 fd = open_file_or_dir(path, &dirstream);
2809 /* fill in fi_args if not just a single device */
2810 if (fi_args->num_devices != 1) {
2811 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
2818 * The fs_args->num_devices does not include seed devices
2820 ret = search_chunk_tree_for_fs_info(fd, fi_args);
2825 * search_chunk_tree_for_fs_info() will lacks the devid 0
2826 * so manual probe for it here.
2828 ret = get_device_info(fd, 0, &tmp);
2830 fi_args->num_devices++;
2838 if (!fi_args->num_devices)
2841 di_args = *di_ret = malloc((fi_args->num_devices) * sizeof(*di_args));
2848 memcpy(di_args, &tmp, sizeof(tmp));
2849 for (; i <= fi_args->max_id; ++i) {
2850 ret = get_device_info(fd, i, &di_args[ndevs]);
2859 * only when the only dev we wanted to find is not there then
2860 * let any error be returned
2862 if (fi_args->num_devices != 1) {
2868 close_file_or_dir(fd, dirstream);
2872 #define isoctal(c) (((c) & ~7) == '0')
2874 static inline void translate(char *f, char *t)
2876 while (*f != '\0') {
2878 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
2879 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
2889 * Checks if the swap device.
2890 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
2892 static int is_swap_device(const char *file)
2903 if (stat(file, &st_buf) < 0)
2905 if (S_ISBLK(st_buf.st_mode))
2906 dev = st_buf.st_rdev;
2907 else if (S_ISREG(st_buf.st_mode)) {
2908 dev = st_buf.st_dev;
2909 ino = st_buf.st_ino;
2913 if ((f = fopen("/proc/swaps", "r")) == NULL)
2916 /* skip the first line */
2917 if (fgets(tmp, sizeof(tmp), f) == NULL)
2920 while (fgets(tmp, sizeof(tmp), f) != NULL) {
2921 if ((cp = strchr(tmp, ' ')) != NULL)
2923 if ((cp = strchr(tmp, '\t')) != NULL)
2925 translate(tmp, buf);
2926 if (stat(buf, &st_buf) != 0)
2928 if (S_ISBLK(st_buf.st_mode)) {
2929 if (dev == st_buf.st_rdev) {
2933 } else if (S_ISREG(st_buf.st_mode)) {
2934 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {
2948 * Check for existing filesystem or partition table on device.
2950 * 1 for existing fs or partition
2951 * 0 for nothing found
2952 * -1 for internal error
2954 static int check_overwrite(const char *device)
2957 blkid_probe pr = NULL;
2961 if (!device || !*device)
2964 ret = -1; /* will reset on success of all setup calls */
2966 pr = blkid_new_probe_from_filename(device);
2970 size = blkid_probe_get_size(pr);
2974 /* nothing to overwrite on a 0-length device */
2980 ret = blkid_probe_enable_partitions(pr, 1);
2984 ret = blkid_do_fullprobe(pr);
2989 * Blkid returns 1 for nothing found and 0 when it finds a signature,
2990 * but we want the exact opposite, so reverse the return value here.
2992 * In addition print some useful diagnostics about what actually is
3000 if (!blkid_probe_lookup_value(pr, "TYPE", &type, NULL)) {
3002 "%s appears to contain an existing "
3003 "filesystem (%s).\n", device, type);
3004 } else if (!blkid_probe_lookup_value(pr, "PTTYPE", &type, NULL)) {
3006 "%s appears to contain a partition "
3007 "table (%s).\n", device, type);
3010 "%s appears to contain something weird "
3011 "according to blkid\n", device);
3017 blkid_free_probe(pr);
3020 "probe of %s failed, cannot detect "
3021 "existing filesystem.\n", device);
3025 static int group_profile_devs_min(u64 flag)
3027 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3028 case 0: /* single */
3029 case BTRFS_BLOCK_GROUP_DUP:
3031 case BTRFS_BLOCK_GROUP_RAID0:
3032 case BTRFS_BLOCK_GROUP_RAID1:
3033 case BTRFS_BLOCK_GROUP_RAID5:
3035 case BTRFS_BLOCK_GROUP_RAID6:
3037 case BTRFS_BLOCK_GROUP_RAID10:
3044 int test_num_disk_vs_raid(u64 metadata_profile, u64 data_profile,
3045 u64 dev_cnt, int mixed, int ssd)
3052 allowed |= BTRFS_BLOCK_GROUP_RAID10;
3054 allowed |= BTRFS_BLOCK_GROUP_RAID6;
3056 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
3057 BTRFS_BLOCK_GROUP_RAID5;
3059 allowed |= BTRFS_BLOCK_GROUP_DUP;
3063 ((metadata_profile | data_profile) & BTRFS_BLOCK_GROUP_DUP)) {
3064 warning("DUP is not recommended on filesystem with multiple devices");
3066 if (metadata_profile & ~allowed) {
3068 "ERROR: unable to create FS with metadata profile %s "
3069 "(have %llu devices but %d devices are required)\n",
3070 btrfs_group_profile_str(metadata_profile), dev_cnt,
3071 group_profile_devs_min(metadata_profile));
3074 if (data_profile & ~allowed) {
3076 "ERROR: unable to create FS with data profile %s "
3077 "(have %llu devices but %d devices are required)\n",
3078 btrfs_group_profile_str(data_profile), dev_cnt,
3079 group_profile_devs_min(data_profile));
3083 warning_on(!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP) && ssd,
3084 "DUP may not actually lead to 2 copies on the device, see manual page");
3089 int group_profile_max_safe_loss(u64 flags)
3091 switch (flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3092 case 0: /* single */
3093 case BTRFS_BLOCK_GROUP_DUP:
3094 case BTRFS_BLOCK_GROUP_RAID0:
3096 case BTRFS_BLOCK_GROUP_RAID1:
3097 case BTRFS_BLOCK_GROUP_RAID5:
3098 case BTRFS_BLOCK_GROUP_RAID10:
3100 case BTRFS_BLOCK_GROUP_RAID6:
3108 * Check if a device is suitable for btrfs
3110 * 1: something is wrong, an error is printed
3113 int test_dev_for_mkfs(const char *file, int force_overwrite)
3118 ret = is_swap_device(file);
3120 error("checking status of %s: %s", file, strerror(-ret));
3124 error("%s is a swap device", file);
3127 if (!force_overwrite) {
3128 if (check_overwrite(file)) {
3129 error("use the -f option to force overwrite of %s",
3134 ret = check_mounted(file);
3136 error("cannot check mount status of %s: %s", file,
3141 error("%s is mounted", file);
3144 /* check if the device is busy */
3145 fd = open(file, O_RDWR|O_EXCL);
3147 error("unable to open %s: %s", file, strerror(errno));
3150 if (fstat(fd, &st)) {
3151 error("unable to stat %s: %s", file, strerror(errno));
3155 if (!S_ISBLK(st.st_mode)) {
3156 error("%s is not a block device", file);
3164 int btrfs_scan_lblkid(void)
3169 struct btrfs_fs_devices *tmp_devices;
3170 blkid_dev_iterate iter = NULL;
3171 blkid_dev dev = NULL;
3172 blkid_cache cache = NULL;
3173 char path[PATH_MAX];
3175 if (btrfs_scan_done)
3178 if (blkid_get_cache(&cache, NULL) < 0) {
3179 error("blkid cache get failed");
3182 blkid_probe_all(cache);
3183 iter = blkid_dev_iterate_begin(cache);
3184 blkid_dev_set_search(iter, "TYPE", "btrfs");
3185 while (blkid_dev_next(iter, &dev) == 0) {
3186 dev = blkid_verify(cache, dev);
3189 /* if we are here its definitely a btrfs disk*/
3190 strncpy_null(path, blkid_dev_devname(dev));
3192 fd = open(path, O_RDONLY);
3194 error("cannot open %s: %s", path, strerror(errno));
3197 ret = btrfs_scan_one_device(fd, path, &tmp_devices,
3198 &num_devices, BTRFS_SUPER_INFO_OFFSET, 0);
3200 error("cannot scan %s: %s", path, strerror(-ret));
3207 blkid_dev_iterate_end(iter);
3208 blkid_put_cache(cache);
3210 btrfs_scan_done = 1;
3215 int is_vol_small(const char *file)
3222 fd = open(file, O_RDONLY);
3225 if (fstat(fd, &st) < 0) {
3230 size = btrfs_device_size(fd, &st);
3235 if (size < BTRFS_MKFS_SMALL_VOLUME_SIZE) {
3245 * This reads a line from the stdin and only returns non-zero if the
3246 * first whitespace delimited token is a case insensitive match with yes
3249 int ask_user(const char *question)
3251 char buf[30] = {0,};
3252 char *saveptr = NULL;
3255 printf("%s [y/N]: ", question);
3257 return fgets(buf, sizeof(buf) - 1, stdin) &&
3258 (answer = strtok_r(buf, " \t\n\r", &saveptr)) &&
3259 (!strcasecmp(answer, "yes") || !strcasecmp(answer, "y"));
3264 * - file or directory return the containing tree root id
3265 * - subvolume return its own tree id
3266 * - BTRFS_EMPTY_SUBVOL_DIR_OBJECTID (directory with ino == 2) the result is
3267 * undefined and function returns -1
3269 int lookup_ino_rootid(int fd, u64 *rootid)
3271 struct btrfs_ioctl_ino_lookup_args args;
3274 memset(&args, 0, sizeof(args));
3276 args.objectid = BTRFS_FIRST_FREE_OBJECTID;
3278 ret = ioctl(fd, BTRFS_IOC_INO_LOOKUP, &args);
3280 error("failed to lookup root id: %s", strerror(errno));
3284 *rootid = args.treeid;
3290 * return 0 if a btrfs mount point is found
3291 * return 1 if a mount point is found but not btrfs
3292 * return <0 if something goes wrong
3294 int find_mount_root(const char *path, char **mount_root)
3302 int longest_matchlen = 0;
3303 char *longest_match = NULL;
3305 fd = open(path, O_RDONLY | O_NOATIME);
3310 mnttab = setmntent("/proc/self/mounts", "r");
3314 while ((ent = getmntent(mnttab))) {
3315 len = strlen(ent->mnt_dir);
3316 if (strncmp(ent->mnt_dir, path, len) == 0) {
3317 /* match found and use the latest match */
3318 if (longest_matchlen <= len) {
3319 free(longest_match);
3320 longest_matchlen = len;
3321 longest_match = strdup(ent->mnt_dir);
3322 not_btrfs = strcmp(ent->mnt_type, "btrfs");
3331 free(longest_match);
3336 *mount_root = realpath(longest_match, NULL);
3340 free(longest_match);
3344 int test_minimum_size(const char *file, u32 nodesize)
3347 struct stat statbuf;
3349 fd = open(file, O_RDONLY);
3352 if (stat(file, &statbuf) < 0) {
3356 if (btrfs_device_size(fd, &statbuf) < btrfs_min_dev_size(nodesize)) {
3366 * Test if path is a directory
3368 * 0 - path exists but it is not a directory
3369 * 1 - path exists and it is a directory
3372 int test_isdir(const char *path)
3377 ret = stat(path, &st);
3381 return !!S_ISDIR(st.st_mode);
3384 void units_set_mode(unsigned *units, unsigned mode)
3386 unsigned base = *units & UNITS_MODE_MASK;
3388 *units = base | mode;
3391 void units_set_base(unsigned *units, unsigned base)
3393 unsigned mode = *units & ~UNITS_MODE_MASK;
3395 *units = base | mode;
3398 int find_next_key(struct btrfs_path *path, struct btrfs_key *key)
3402 for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
3403 if (!path->nodes[level])
3405 if (path->slots[level] + 1 >=
3406 btrfs_header_nritems(path->nodes[level]))
3409 btrfs_item_key_to_cpu(path->nodes[level], key,
3410 path->slots[level] + 1);
3412 btrfs_node_key_to_cpu(path->nodes[level], key,
3413 path->slots[level] + 1);
3419 const char* btrfs_group_type_str(u64 flag)
3421 u64 mask = BTRFS_BLOCK_GROUP_TYPE_MASK |
3422 BTRFS_SPACE_INFO_GLOBAL_RSV;
3424 switch (flag & mask) {
3425 case BTRFS_BLOCK_GROUP_DATA:
3427 case BTRFS_BLOCK_GROUP_SYSTEM:
3429 case BTRFS_BLOCK_GROUP_METADATA:
3431 case BTRFS_BLOCK_GROUP_DATA|BTRFS_BLOCK_GROUP_METADATA:
3432 return "Data+Metadata";
3433 case BTRFS_SPACE_INFO_GLOBAL_RSV:
3434 return "GlobalReserve";
3440 const char* btrfs_group_profile_str(u64 flag)
3442 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3445 case BTRFS_BLOCK_GROUP_RAID0:
3447 case BTRFS_BLOCK_GROUP_RAID1:
3449 case BTRFS_BLOCK_GROUP_RAID5:
3451 case BTRFS_BLOCK_GROUP_RAID6:
3453 case BTRFS_BLOCK_GROUP_DUP:
3455 case BTRFS_BLOCK_GROUP_RAID10:
3462 u64 disk_size(const char *path)
3466 if (statfs(path, &sfs) < 0)
3469 return sfs.f_bsize * sfs.f_blocks;
3472 u64 get_partition_size(const char *dev)
3475 int fd = open(dev, O_RDONLY);
3479 if (ioctl(fd, BLKGETSIZE64, &result) < 0) {
3488 int btrfs_tree_search2_ioctl_supported(int fd)
3490 struct btrfs_ioctl_search_args_v2 *args2;
3491 struct btrfs_ioctl_search_key *sk;
3492 int args2_size = 1024;
3493 char args2_buf[args2_size];
3495 static int v2_supported = -1;
3497 if (v2_supported != -1)
3498 return v2_supported;
3500 args2 = (struct btrfs_ioctl_search_args_v2 *)args2_buf;
3504 * Search for the extent tree item in the root tree.
3506 sk->tree_id = BTRFS_ROOT_TREE_OBJECTID;
3507 sk->min_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3508 sk->max_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3509 sk->min_type = BTRFS_ROOT_ITEM_KEY;
3510 sk->max_type = BTRFS_ROOT_ITEM_KEY;
3512 sk->max_offset = (u64)-1;
3513 sk->min_transid = 0;
3514 sk->max_transid = (u64)-1;
3516 args2->buf_size = args2_size - sizeof(struct btrfs_ioctl_search_args_v2);
3517 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH_V2, args2);
3518 if (ret == -EOPNOTSUPP)
3525 return v2_supported;
3528 int btrfs_check_nodesize(u32 nodesize, u32 sectorsize, u64 features)
3530 if (nodesize < sectorsize) {
3531 error("illegal nodesize %u (smaller than %u)",
3532 nodesize, sectorsize);
3534 } else if (nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) {
3535 error("illegal nodesize %u (larger than %u)",
3536 nodesize, BTRFS_MAX_METADATA_BLOCKSIZE);
3538 } else if (nodesize & (sectorsize - 1)) {
3539 error("illegal nodesize %u (not aligned to %u)",
3540 nodesize, sectorsize);
3542 } else if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS &&
3543 nodesize != sectorsize) {
3544 error("illegal nodesize %u (not equal to %u for mixed block group)",
3545 nodesize, sectorsize);
3552 * Copy a path argument from SRC to DEST and check the SRC length if it's at
3553 * most PATH_MAX and fits into DEST. DESTLEN is supposed to be exact size of
3555 * The destination buffer is zero terminated.
3556 * Return < 0 for error, 0 otherwise.
3558 int arg_copy_path(char *dest, const char *src, int destlen)
3560 size_t len = strlen(src);
3562 if (len >= PATH_MAX || len >= destlen)
3563 return -ENAMETOOLONG;
3565 __strncpy_null(dest, src, destlen);
3570 unsigned int get_unit_mode_from_arg(int *argc, char *argv[], int df_mode)
3572 unsigned int unit_mode = UNITS_DEFAULT;
3576 for (arg_i = 0; arg_i < *argc; arg_i++) {
3577 if (!strcmp(argv[arg_i], "--"))
3580 if (!strcmp(argv[arg_i], "--raw")) {
3581 unit_mode = UNITS_RAW;
3585 if (!strcmp(argv[arg_i], "--human-readable")) {
3586 unit_mode = UNITS_HUMAN_BINARY;
3591 if (!strcmp(argv[arg_i], "--iec")) {
3592 units_set_mode(&unit_mode, UNITS_BINARY);
3596 if (!strcmp(argv[arg_i], "--si")) {
3597 units_set_mode(&unit_mode, UNITS_DECIMAL);
3602 if (!strcmp(argv[arg_i], "--kbytes")) {
3603 units_set_base(&unit_mode, UNITS_KBYTES);
3607 if (!strcmp(argv[arg_i], "--mbytes")) {
3608 units_set_base(&unit_mode, UNITS_MBYTES);
3612 if (!strcmp(argv[arg_i], "--gbytes")) {
3613 units_set_base(&unit_mode, UNITS_GBYTES);
3617 if (!strcmp(argv[arg_i], "--tbytes")) {
3618 units_set_base(&unit_mode, UNITS_TBYTES);
3626 if (!strcmp(argv[arg_i], "-b")) {
3627 unit_mode = UNITS_RAW;
3631 if (!strcmp(argv[arg_i], "-h")) {
3632 unit_mode = UNITS_HUMAN_BINARY;
3636 if (!strcmp(argv[arg_i], "-H")) {
3637 unit_mode = UNITS_HUMAN_DECIMAL;
3641 if (!strcmp(argv[arg_i], "-k")) {
3642 units_set_base(&unit_mode, UNITS_KBYTES);
3646 if (!strcmp(argv[arg_i], "-m")) {
3647 units_set_base(&unit_mode, UNITS_MBYTES);
3651 if (!strcmp(argv[arg_i], "-g")) {
3652 units_set_base(&unit_mode, UNITS_GBYTES);
3656 if (!strcmp(argv[arg_i], "-t")) {
3657 units_set_base(&unit_mode, UNITS_TBYTES);
3663 for (arg_i = 0, arg_end = 0; arg_i < *argc; arg_i++) {
3666 argv[arg_end] = argv[arg_i];
3675 int string_is_numerical(const char *str)
3677 if (!(*str >= '0' && *str <= '9'))
3679 while (*str >= '0' && *str <= '9')
3687 * Preprocess @argv with getopt_long to reorder options and consume the "--"
3689 * Unknown short and long options are reported, optionally the @usage is printed
3692 void clean_args_no_options(int argc, char *argv[], const char * const *usagestr)
3694 static const struct option long_options[] = {
3699 int c = getopt_long(argc, argv, "", long_options, NULL);
3712 /* Subvolume helper functions */
3714 * test if name is a correct subvolume name
3715 * this function return
3716 * 0-> name is not a correct subvolume name
3717 * 1-> name is a correct subvolume name
3719 int test_issubvolname(const char *name)
3721 return name[0] != '\0' && !strchr(name, '/') &&
3722 strcmp(name, ".") && strcmp(name, "..");
3726 * Test if path is a subvolume
3728 * 0 - path exists but it is not a subvolume
3729 * 1 - path exists and it is a subvolume
3732 int test_issubvolume(const char *path)
3738 res = stat(path, &st);
3742 if (st.st_ino != BTRFS_FIRST_FREE_OBJECTID || !S_ISDIR(st.st_mode))
3745 res = statfs(path, &stfs);
3749 return (int)stfs.f_type == BTRFS_SUPER_MAGIC;
3752 const char *subvol_strip_mountpoint(const char *mnt, const char *full_path)
3754 int len = strlen(mnt);
3758 if (mnt[len - 1] != '/')
3761 return full_path + len;
3768 * 1: Error; and error info printed to the terminal. Fixme.
3769 * 2: If the fullpath is root tree instead of subvol tree
3771 int get_subvol_info(const char *fullpath, struct root_info *get_ri)
3778 const char *svpath = NULL;
3779 DIR *dirstream1 = NULL;
3780 DIR *dirstream2 = NULL;
3782 ret = test_issubvolume(fullpath);
3786 error("not a subvolume: %s", fullpath);
3790 ret = find_mount_root(fullpath, &mnt);
3794 error("%s doesn't belong to btrfs mount point", fullpath);
3798 svpath = subvol_strip_mountpoint(mnt, fullpath);
3800 fd = btrfs_open_dir(fullpath, &dirstream1, 1);
3804 ret = btrfs_list_get_path_rootid(fd, &sv_id);
3806 error("can't get rootid for '%s'", fullpath);
3810 mntfd = btrfs_open_dir(mnt, &dirstream2, 1);
3814 if (sv_id == BTRFS_FS_TREE_OBJECTID) {
3817 * So that caller may decide if thats an error or just fine.
3822 memset(get_ri, 0, sizeof(*get_ri));
3823 get_ri->root_id = sv_id;
3825 ret = btrfs_get_subvol(mntfd, get_ri);
3827 error("can't find '%s': %d", svpath, ret);
3830 close_file_or_dir(mntfd, dirstream2);
3831 close_file_or_dir(fd, dirstream1);
3837 void init_rand_seed(u64 seed)
3841 /* only use the last 48 bits */
3842 for (i = 0; i < 3; i++) {
3843 rand_seed[i] = (unsigned short)(seed ^ (unsigned short)(-1));
3846 rand_seed_initlized = 1;
3849 static void __init_seed(void)
3855 if(rand_seed_initlized)
3857 /* Use urandom as primary seed source. */
3858 fd = open("/dev/urandom", O_RDONLY);
3860 ret = read(fd, rand_seed, sizeof(rand_seed));
3862 if (ret < sizeof(rand_seed))
3866 /* Use time and pid as fallback seed */
3867 warning("failed to read /dev/urandom, use time and pid as random seed");
3868 gettimeofday(&tv, 0);
3869 rand_seed[0] = getpid() ^ (tv.tv_sec & 0xFFFF);
3870 rand_seed[1] = getppid() ^ (tv.tv_usec & 0xFFFF);
3871 rand_seed[2] = (tv.tv_sec ^ tv.tv_usec) >> 16;
3873 rand_seed_initlized = 1;
3880 * Don't use nrand48, its range is [0,2^31) The highest bit will alwasy
3881 * be 0. Use jrand48 to include the highest bit.
3883 return (u32)jrand48(rand_seed);
3886 unsigned int rand_range(unsigned int upper)
3890 * Use the full 48bits to mod, which would be more uniformly
3893 return (unsigned int)(jrand48(rand_seed) % upper);