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);
681 static int setup_temp_csum_tree(int fd, struct btrfs_mkfs_config *cfg,
684 struct extent_buffer *buf = NULL;
687 buf = malloc(sizeof(*buf) + cfg->nodesize);
690 ret = setup_temp_extent_buffer(buf, cfg, csum_bytenr,
691 BTRFS_CSUM_TREE_OBJECTID);
695 * Temporary csum tree is completely empty.
697 ret = write_temp_extent_buffer(fd, buf, csum_bytenr);
704 * Improved version of make_btrfs().
707 * 1) Do chunk allocation to avoid used data
708 * And after this function, extent type matches chunk type
709 * 2) Better structured code
710 * No super long hand written codes to initialized all tree blocks
711 * Split into small blocks and reuse codes.
712 * TODO: Reuse tree operation facilities by introducing new flags
714 static int make_convert_btrfs(int fd, struct btrfs_mkfs_config *cfg,
715 struct btrfs_convert_context *cctx)
717 struct cache_tree *free = &cctx->free;
718 struct cache_tree *used = &cctx->used;
720 u64 meta_chunk_start;
721 /* chunk tree bytenr, in system chunk */
723 /* metadata trees bytenr, in metadata chunk */
731 /* Shouldn't happen */
732 BUG_ON(cache_tree_empty(used));
735 * reserve space for temporary superblock first
736 * Here we allocate a little larger space, to keep later
737 * free space will be STRIPE_LEN aligned
739 ret = reserve_free_space(free, BTRFS_STRIPE_LEN,
745 * Then reserve system chunk space
746 * TODO: Change system group size depending on cctx->total_bytes.
747 * If using current 4M, it can only handle less than one TB for
748 * worst case and then run out of sys space.
750 ret = reserve_free_space(free, BTRFS_MKFS_SYSTEM_GROUP_SIZE,
754 ret = reserve_free_space(free, BTRFS_CONVERT_META_GROUP_SIZE,
760 * Allocated meta/sys chunks will be mapped 1:1 with device offset.
762 * Inside the allocated metadata chunk, the layout will be:
763 * | offset | contents |
764 * -------------------------------------
766 * | +nodesize | extent root |
767 * | +nodesize * 2 | device root |
768 * | +nodesize * 3 | fs tree |
769 * | +nodesize * 4 | csum tree |
770 * -------------------------------------
771 * Inside the allocated system chunk, the layout will be:
772 * | offset | contents |
773 * -------------------------------------
774 * | +0 | chunk root |
775 * -------------------------------------
777 chunk_bytenr = sys_chunk_start;
778 root_bytenr = meta_chunk_start;
779 extent_bytenr = meta_chunk_start + cfg->nodesize;
780 dev_bytenr = meta_chunk_start + cfg->nodesize * 2;
781 fs_bytenr = meta_chunk_start + cfg->nodesize * 3;
782 csum_bytenr = meta_chunk_start + cfg->nodesize * 4;
784 ret = setup_temp_super(fd, cfg, root_bytenr, chunk_bytenr);
788 ret = setup_temp_root_tree(fd, cfg, root_bytenr, extent_bytenr,
789 dev_bytenr, fs_bytenr, csum_bytenr);
792 ret = setup_temp_chunk_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
796 ret = setup_temp_dev_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
800 ret = setup_temp_fs_tree(fd, cfg, fs_bytenr);
803 ret = setup_temp_csum_tree(fd, cfg, csum_bytenr);
810 * @fs_uuid - if NULL, generates a UUID, returns back the new filesystem UUID
812 int make_btrfs(int fd, struct btrfs_mkfs_config *cfg,
813 struct btrfs_convert_context *cctx)
815 struct btrfs_super_block super;
816 struct extent_buffer *buf;
817 struct btrfs_root_item root_item;
818 struct btrfs_disk_key disk_key;
819 struct btrfs_extent_item *extent_item;
820 struct btrfs_inode_item *inode_item;
821 struct btrfs_chunk *chunk;
822 struct btrfs_dev_item *dev_item;
823 struct btrfs_dev_extent *dev_extent;
824 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
834 int skinny_metadata = !!(cfg->features &
835 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA);
839 return make_convert_btrfs(fd, cfg, cctx);
840 buf = malloc(sizeof(*buf) + max(cfg->sectorsize, cfg->nodesize));
844 first_free = BTRFS_SUPER_INFO_OFFSET + cfg->sectorsize * 2 - 1;
845 first_free &= ~((u64)cfg->sectorsize - 1);
847 memset(&super, 0, sizeof(super));
849 num_bytes = (cfg->num_bytes / cfg->sectorsize) * cfg->sectorsize;
850 if (cfg->fs_uuid && *cfg->fs_uuid) {
851 if (uuid_parse(cfg->fs_uuid, super.fsid) != 0) {
852 error("cannot not parse UUID: %s", cfg->fs_uuid);
856 if (!test_uuid_unique(cfg->fs_uuid)) {
857 error("non-unique UUID: %s", cfg->fs_uuid);
862 uuid_generate(super.fsid);
864 uuid_unparse(super.fsid, cfg->fs_uuid);
866 uuid_generate(super.dev_item.uuid);
867 uuid_generate(chunk_tree_uuid);
869 btrfs_set_super_bytenr(&super, cfg->blocks[0]);
870 btrfs_set_super_num_devices(&super, 1);
871 btrfs_set_super_magic(&super, BTRFS_MAGIC);
872 btrfs_set_super_generation(&super, 1);
873 btrfs_set_super_root(&super, cfg->blocks[1]);
874 btrfs_set_super_chunk_root(&super, cfg->blocks[3]);
875 btrfs_set_super_total_bytes(&super, num_bytes);
876 btrfs_set_super_bytes_used(&super, 6 * cfg->nodesize);
877 btrfs_set_super_sectorsize(&super, cfg->sectorsize);
878 btrfs_set_super_leafsize(&super, cfg->nodesize);
879 btrfs_set_super_nodesize(&super, cfg->nodesize);
880 btrfs_set_super_stripesize(&super, cfg->stripesize);
881 btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
882 btrfs_set_super_chunk_root_generation(&super, 1);
883 btrfs_set_super_cache_generation(&super, -1);
884 btrfs_set_super_incompat_flags(&super, cfg->features);
886 __strncpy_null(super.label, cfg->label, BTRFS_LABEL_SIZE - 1);
888 /* create the tree of root objects */
889 memset(buf->data, 0, cfg->nodesize);
890 buf->len = cfg->nodesize;
891 btrfs_set_header_bytenr(buf, cfg->blocks[1]);
892 btrfs_set_header_nritems(buf, 4);
893 btrfs_set_header_generation(buf, 1);
894 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
895 btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
896 write_extent_buffer(buf, super.fsid, btrfs_header_fsid(),
899 write_extent_buffer(buf, chunk_tree_uuid,
900 btrfs_header_chunk_tree_uuid(buf),
903 /* create the items for the root tree */
904 memset(&root_item, 0, sizeof(root_item));
905 inode_item = &root_item.inode;
906 btrfs_set_stack_inode_generation(inode_item, 1);
907 btrfs_set_stack_inode_size(inode_item, 3);
908 btrfs_set_stack_inode_nlink(inode_item, 1);
909 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
910 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
911 btrfs_set_root_refs(&root_item, 1);
912 btrfs_set_root_used(&root_item, cfg->nodesize);
913 btrfs_set_root_generation(&root_item, 1);
915 memset(&disk_key, 0, sizeof(disk_key));
916 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
917 btrfs_set_disk_key_offset(&disk_key, 0);
920 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - sizeof(root_item);
921 btrfs_set_root_bytenr(&root_item, cfg->blocks[2]);
922 btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
923 btrfs_set_item_key(buf, &disk_key, nritems);
924 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
925 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
927 write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
928 nritems), sizeof(root_item));
931 itemoff = itemoff - sizeof(root_item);
932 btrfs_set_root_bytenr(&root_item, cfg->blocks[4]);
933 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
934 btrfs_set_item_key(buf, &disk_key, nritems);
935 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
936 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
938 write_extent_buffer(buf, &root_item,
939 btrfs_item_ptr_offset(buf, nritems),
943 itemoff = itemoff - sizeof(root_item);
944 btrfs_set_root_bytenr(&root_item, cfg->blocks[5]);
945 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
946 btrfs_set_item_key(buf, &disk_key, nritems);
947 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
948 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
950 write_extent_buffer(buf, &root_item,
951 btrfs_item_ptr_offset(buf, nritems),
955 itemoff = itemoff - sizeof(root_item);
956 btrfs_set_root_bytenr(&root_item, cfg->blocks[6]);
957 btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_TREE_OBJECTID);
958 btrfs_set_item_key(buf, &disk_key, nritems);
959 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
960 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
962 write_extent_buffer(buf, &root_item,
963 btrfs_item_ptr_offset(buf, nritems),
968 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
969 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[1]);
970 if (ret != cfg->nodesize) {
971 ret = (ret < 0 ? -errno : -EIO);
975 /* create the items for the extent tree */
976 memset(buf->data + sizeof(struct btrfs_header), 0,
977 cfg->nodesize - sizeof(struct btrfs_header));
979 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
980 for (i = 1; i < 7; i++) {
981 item_size = sizeof(struct btrfs_extent_item);
982 if (!skinny_metadata)
983 item_size += sizeof(struct btrfs_tree_block_info);
985 BUG_ON(cfg->blocks[i] < first_free);
986 BUG_ON(cfg->blocks[i] < cfg->blocks[i - 1]);
988 /* create extent item */
989 itemoff -= item_size;
990 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
991 if (skinny_metadata) {
992 btrfs_set_disk_key_type(&disk_key,
993 BTRFS_METADATA_ITEM_KEY);
994 btrfs_set_disk_key_offset(&disk_key, 0);
996 btrfs_set_disk_key_type(&disk_key,
997 BTRFS_EXTENT_ITEM_KEY);
998 btrfs_set_disk_key_offset(&disk_key, cfg->nodesize);
1000 btrfs_set_item_key(buf, &disk_key, nritems);
1001 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
1003 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1005 extent_item = btrfs_item_ptr(buf, nritems,
1006 struct btrfs_extent_item);
1007 btrfs_set_extent_refs(buf, extent_item, 1);
1008 btrfs_set_extent_generation(buf, extent_item, 1);
1009 btrfs_set_extent_flags(buf, extent_item,
1010 BTRFS_EXTENT_FLAG_TREE_BLOCK);
1013 /* create extent ref */
1014 ref_root = reference_root_table[i];
1015 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
1016 btrfs_set_disk_key_offset(&disk_key, ref_root);
1017 btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
1018 btrfs_set_item_key(buf, &disk_key, nritems);
1019 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
1021 btrfs_set_item_size(buf, btrfs_item_nr(nritems), 0);
1024 btrfs_set_header_bytenr(buf, cfg->blocks[2]);
1025 btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
1026 btrfs_set_header_nritems(buf, nritems);
1027 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1028 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[2]);
1029 if (ret != cfg->nodesize) {
1030 ret = (ret < 0 ? -errno : -EIO);
1034 /* create the chunk tree */
1035 memset(buf->data + sizeof(struct btrfs_header), 0,
1036 cfg->nodesize - sizeof(struct btrfs_header));
1038 item_size = sizeof(*dev_item);
1039 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - item_size;
1041 /* first device 1 (there is no device 0) */
1042 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
1043 btrfs_set_disk_key_offset(&disk_key, 1);
1044 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
1045 btrfs_set_item_key(buf, &disk_key, nritems);
1046 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1047 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
1049 dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
1050 btrfs_set_device_id(buf, dev_item, 1);
1051 btrfs_set_device_generation(buf, dev_item, 0);
1052 btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
1053 btrfs_set_device_bytes_used(buf, dev_item,
1054 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1055 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
1056 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
1057 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
1058 btrfs_set_device_type(buf, dev_item, 0);
1060 write_extent_buffer(buf, super.dev_item.uuid,
1061 (unsigned long)btrfs_device_uuid(dev_item),
1063 write_extent_buffer(buf, super.fsid,
1064 (unsigned long)btrfs_device_fsid(dev_item),
1066 read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
1070 item_size = btrfs_chunk_item_size(1);
1071 itemoff = itemoff - item_size;
1073 /* then we have chunk 0 */
1074 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1075 btrfs_set_disk_key_offset(&disk_key, 0);
1076 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
1077 btrfs_set_item_key(buf, &disk_key, nritems);
1078 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1079 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
1081 chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
1082 btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1083 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
1084 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
1085 btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
1086 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
1087 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
1088 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
1089 btrfs_set_chunk_num_stripes(buf, chunk, 1);
1090 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
1091 btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
1094 write_extent_buffer(buf, super.dev_item.uuid,
1095 (unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
1098 /* copy the key for the chunk to the system array */
1099 ptr = super.sys_chunk_array;
1100 array_size = sizeof(disk_key);
1102 memcpy(ptr, &disk_key, sizeof(disk_key));
1103 ptr += sizeof(disk_key);
1105 /* copy the chunk to the system array */
1106 read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
1107 array_size += item_size;
1109 btrfs_set_super_sys_array_size(&super, array_size);
1111 btrfs_set_header_bytenr(buf, cfg->blocks[3]);
1112 btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
1113 btrfs_set_header_nritems(buf, nritems);
1114 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1115 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[3]);
1116 if (ret != cfg->nodesize) {
1117 ret = (ret < 0 ? -errno : -EIO);
1121 /* create the device tree */
1122 memset(buf->data + sizeof(struct btrfs_header), 0,
1123 cfg->nodesize - sizeof(struct btrfs_header));
1125 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) -
1126 sizeof(struct btrfs_dev_extent);
1128 btrfs_set_disk_key_objectid(&disk_key, 1);
1129 btrfs_set_disk_key_offset(&disk_key, 0);
1130 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
1131 btrfs_set_item_key(buf, &disk_key, nritems);
1132 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1133 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1134 sizeof(struct btrfs_dev_extent));
1135 dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
1136 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
1137 BTRFS_CHUNK_TREE_OBJECTID);
1138 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
1139 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1140 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
1142 write_extent_buffer(buf, chunk_tree_uuid,
1143 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
1146 btrfs_set_dev_extent_length(buf, dev_extent,
1147 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1150 btrfs_set_header_bytenr(buf, cfg->blocks[4]);
1151 btrfs_set_header_owner(buf, BTRFS_DEV_TREE_OBJECTID);
1152 btrfs_set_header_nritems(buf, nritems);
1153 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1154 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[4]);
1155 if (ret != cfg->nodesize) {
1156 ret = (ret < 0 ? -errno : -EIO);
1160 /* create the FS root */
1161 memset(buf->data + sizeof(struct btrfs_header), 0,
1162 cfg->nodesize - sizeof(struct btrfs_header));
1163 btrfs_set_header_bytenr(buf, cfg->blocks[5]);
1164 btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
1165 btrfs_set_header_nritems(buf, 0);
1166 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1167 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[5]);
1168 if (ret != cfg->nodesize) {
1169 ret = (ret < 0 ? -errno : -EIO);
1172 /* finally create the csum root */
1173 memset(buf->data + sizeof(struct btrfs_header), 0,
1174 cfg->nodesize - sizeof(struct btrfs_header));
1175 btrfs_set_header_bytenr(buf, cfg->blocks[6]);
1176 btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
1177 btrfs_set_header_nritems(buf, 0);
1178 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1179 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[6]);
1180 if (ret != cfg->nodesize) {
1181 ret = (ret < 0 ? -errno : -EIO);
1185 /* and write out the super block */
1186 BUG_ON(sizeof(super) > cfg->sectorsize);
1187 memset(buf->data, 0, BTRFS_SUPER_INFO_SIZE);
1188 memcpy(buf->data, &super, sizeof(super));
1189 buf->len = BTRFS_SUPER_INFO_SIZE;
1190 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1191 ret = pwrite(fd, buf->data, BTRFS_SUPER_INFO_SIZE, cfg->blocks[0]);
1192 if (ret != BTRFS_SUPER_INFO_SIZE) {
1193 ret = (ret < 0 ? -errno : -EIO);
1204 static const struct btrfs_fs_feature {
1208 } mkfs_features[] = {
1209 { "mixed-bg", BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS,
1210 "mixed data and metadata block groups" },
1211 { "extref", BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF,
1212 "increased hardlink limit per file to 65536" },
1213 { "raid56", BTRFS_FEATURE_INCOMPAT_RAID56,
1214 "raid56 extended format" },
1215 { "skinny-metadata", BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA,
1216 "reduced-size metadata extent refs" },
1217 { "no-holes", BTRFS_FEATURE_INCOMPAT_NO_HOLES,
1218 "no explicit hole extents for files" },
1219 /* Keep this one last */
1220 { "list-all", BTRFS_FEATURE_LIST_ALL, NULL }
1223 static int parse_one_fs_feature(const char *name, u64 *flags)
1228 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1229 if (name[0] == '^' &&
1230 !strcmp(mkfs_features[i].name, name + 1)) {
1231 *flags &= ~ mkfs_features[i].flag;
1233 } else if (!strcmp(mkfs_features[i].name, name)) {
1234 *flags |= mkfs_features[i].flag;
1242 void btrfs_parse_features_to_string(char *buf, u64 flags)
1248 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1249 if (flags & mkfs_features[i].flag) {
1252 strcat(buf, mkfs_features[i].name);
1257 void btrfs_process_fs_features(u64 flags)
1261 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1262 if (flags & mkfs_features[i].flag) {
1263 printf("Turning ON incompat feature '%s': %s\n",
1264 mkfs_features[i].name,
1265 mkfs_features[i].desc);
1270 void btrfs_list_all_fs_features(u64 mask_disallowed)
1274 fprintf(stderr, "Filesystem features available:\n");
1275 for (i = 0; i < ARRAY_SIZE(mkfs_features) - 1; i++) {
1276 char *is_default = "";
1278 if (mkfs_features[i].flag & mask_disallowed)
1280 if (mkfs_features[i].flag & BTRFS_MKFS_DEFAULT_FEATURES)
1281 is_default = ", default";
1282 fprintf(stderr, "%-20s- %s (0x%llx%s)\n",
1283 mkfs_features[i].name,
1284 mkfs_features[i].desc,
1285 mkfs_features[i].flag,
1291 * Return NULL if all features were parsed fine, otherwise return the name of
1292 * the first unparsed.
1294 char* btrfs_parse_fs_features(char *namelist, u64 *flags)
1297 char *save_ptr = NULL; /* Satisfy static checkers */
1299 for (this_char = strtok_r(namelist, ",", &save_ptr);
1301 this_char = strtok_r(NULL, ",", &save_ptr)) {
1302 if (parse_one_fs_feature(this_char, flags))
1309 u64 btrfs_device_size(int fd, struct stat *st)
1312 if (S_ISREG(st->st_mode)) {
1315 if (!S_ISBLK(st->st_mode)) {
1318 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
1324 static int zero_blocks(int fd, off_t start, size_t len)
1326 char *buf = malloc(len);
1332 memset(buf, 0, len);
1333 written = pwrite(fd, buf, len, start);
1340 #define ZERO_DEV_BYTES (2 * 1024 * 1024)
1342 /* don't write outside the device by clamping the region to the device size */
1343 static int zero_dev_clamped(int fd, off_t start, ssize_t len, u64 dev_size)
1345 off_t end = max(start, start + len);
1348 /* and don't overwrite the disk labels on sparc */
1349 start = max(start, 1024);
1350 end = max(end, 1024);
1353 start = min_t(u64, start, dev_size);
1354 end = min_t(u64, end, dev_size);
1356 return zero_blocks(fd, start, end - start);
1359 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
1360 struct btrfs_root *root, int fd, char *path,
1361 u64 device_total_bytes, u32 io_width, u32 io_align,
1364 struct btrfs_super_block *disk_super;
1365 struct btrfs_super_block *super = root->fs_info->super_copy;
1366 struct btrfs_device *device;
1367 struct btrfs_dev_item *dev_item;
1373 device_total_bytes = (device_total_bytes / sectorsize) * sectorsize;
1375 device = kzalloc(sizeof(*device), GFP_NOFS);
1378 buf = kzalloc(sectorsize, GFP_NOFS);
1381 BUG_ON(sizeof(*disk_super) > sectorsize);
1383 disk_super = (struct btrfs_super_block *)buf;
1384 dev_item = &disk_super->dev_item;
1386 uuid_generate(device->uuid);
1389 device->io_width = io_width;
1390 device->io_align = io_align;
1391 device->sector_size = sectorsize;
1393 device->writeable = 1;
1394 device->total_bytes = device_total_bytes;
1395 device->bytes_used = 0;
1396 device->total_ios = 0;
1397 device->dev_root = root->fs_info->dev_root;
1398 device->name = strdup(path);
1402 INIT_LIST_HEAD(&device->dev_list);
1403 ret = btrfs_add_device(trans, root, device);
1406 fs_total_bytes = btrfs_super_total_bytes(super) + device_total_bytes;
1407 btrfs_set_super_total_bytes(super, fs_total_bytes);
1409 num_devs = btrfs_super_num_devices(super) + 1;
1410 btrfs_set_super_num_devices(super, num_devs);
1412 memcpy(disk_super, super, sizeof(*disk_super));
1414 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
1415 btrfs_set_stack_device_id(dev_item, device->devid);
1416 btrfs_set_stack_device_type(dev_item, device->type);
1417 btrfs_set_stack_device_io_align(dev_item, device->io_align);
1418 btrfs_set_stack_device_io_width(dev_item, device->io_width);
1419 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
1420 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
1421 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
1422 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
1424 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
1425 BUG_ON(ret != sectorsize);
1428 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
1429 device->fs_devices = root->fs_info->fs_devices;
1438 static int btrfs_wipe_existing_sb(int fd)
1440 const char *off = NULL;
1445 blkid_probe pr = NULL;
1447 pr = blkid_new_probe();
1451 if (blkid_probe_set_device(pr, fd, 0, 0)) {
1456 ret = blkid_probe_lookup_value(pr, "SBMAGIC_OFFSET", &off, NULL);
1458 ret = blkid_probe_lookup_value(pr, "SBMAGIC", NULL, &len);
1460 if (ret || len == 0 || off == NULL) {
1462 * If lookup fails, the probe did not find any values, eg. for
1463 * a file image or a loop device. Soft error.
1469 offset = strtoll(off, NULL, 10);
1470 if (len > sizeof(buf))
1473 memset(buf, 0, len);
1474 ret = pwrite(fd, buf, len, offset);
1476 error("cannot wipe existing superblock: %s", strerror(errno));
1478 } else if (ret != len) {
1479 error("cannot wipe existing superblock: wrote %d of %zd", ret, len);
1485 blkid_free_probe(pr);
1489 int btrfs_prepare_device(int fd, const char *file, int zero_end,
1490 u64 *block_count_ret, u64 max_block_count, int discard)
1496 ret = fstat(fd, &st);
1498 error("unable to stat %s: %s", file, strerror(errno));
1502 block_count = btrfs_device_size(fd, &st);
1503 if (block_count == 0) {
1504 error("unable to determine size of %s", file);
1507 if (max_block_count)
1508 block_count = min(block_count, max_block_count);
1512 * We intentionally ignore errors from the discard ioctl. It
1513 * is not necessary for the mkfs functionality but just an
1516 if (discard_range(fd, 0, 0) == 0) {
1517 printf("Performing full device TRIM (%s) ...\n",
1518 pretty_size(block_count));
1519 discard_blocks(fd, 0, block_count);
1523 ret = zero_dev_clamped(fd, 0, ZERO_DEV_BYTES, block_count);
1524 for (i = 0 ; !ret && i < BTRFS_SUPER_MIRROR_MAX; i++)
1525 ret = zero_dev_clamped(fd, btrfs_sb_offset(i),
1526 BTRFS_SUPER_INFO_SIZE, block_count);
1527 if (!ret && zero_end)
1528 ret = zero_dev_clamped(fd, block_count - ZERO_DEV_BYTES,
1529 ZERO_DEV_BYTES, block_count);
1532 error("failed to zero device '%s': %s", file, strerror(-ret));
1536 ret = btrfs_wipe_existing_sb(fd);
1538 error("cannot wipe superblocks on %s", file);
1542 *block_count_ret = block_count;
1546 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
1547 struct btrfs_root *root, u64 objectid)
1550 struct btrfs_inode_item inode_item;
1551 time_t now = time(NULL);
1553 memset(&inode_item, 0, sizeof(inode_item));
1554 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
1555 btrfs_set_stack_inode_size(&inode_item, 0);
1556 btrfs_set_stack_inode_nlink(&inode_item, 1);
1557 btrfs_set_stack_inode_nbytes(&inode_item, root->nodesize);
1558 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
1559 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
1560 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
1561 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
1562 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
1563 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
1564 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
1565 btrfs_set_stack_timespec_sec(&inode_item.otime, 0);
1566 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
1568 if (root->fs_info->tree_root == root)
1569 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
1571 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
1575 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
1579 btrfs_set_root_dirid(&root->root_item, objectid);
1586 * checks if a path is a block device node
1587 * Returns negative errno on failure, otherwise
1588 * returns 1 for blockdev, 0 for not-blockdev
1590 int is_block_device(const char *path)
1592 struct stat statbuf;
1594 if (stat(path, &statbuf) < 0)
1597 return !!S_ISBLK(statbuf.st_mode);
1601 * check if given path is a mount point
1602 * return 1 if yes. 0 if no. -1 for error
1604 int is_mount_point(const char *path)
1610 f = setmntent("/proc/self/mounts", "r");
1614 while ((mnt = getmntent(f)) != NULL) {
1615 if (strcmp(mnt->mnt_dir, path))
1624 static int is_reg_file(const char *path)
1626 struct stat statbuf;
1628 if (stat(path, &statbuf) < 0)
1630 return S_ISREG(statbuf.st_mode);
1634 * This function checks if the given input parameter is
1636 * return <0 : some error in the given input
1637 * return BTRFS_ARG_UNKNOWN: unknown input
1638 * return BTRFS_ARG_UUID: given input is uuid
1639 * return BTRFS_ARG_MNTPOINT: given input is path
1640 * return BTRFS_ARG_REG: given input is regular file
1641 * return BTRFS_ARG_BLKDEV: given input is block device
1643 int check_arg_type(const char *input)
1646 char path[PATH_MAX];
1651 if (realpath(input, path)) {
1652 if (is_block_device(path) == 1)
1653 return BTRFS_ARG_BLKDEV;
1655 if (is_mount_point(path) == 1)
1656 return BTRFS_ARG_MNTPOINT;
1658 if (is_reg_file(path))
1659 return BTRFS_ARG_REG;
1661 return BTRFS_ARG_UNKNOWN;
1664 if (strlen(input) == (BTRFS_UUID_UNPARSED_SIZE - 1) &&
1665 !uuid_parse(input, uuid))
1666 return BTRFS_ARG_UUID;
1668 return BTRFS_ARG_UNKNOWN;
1672 * Find the mount point for a mounted device.
1673 * On success, returns 0 with mountpoint in *mp.
1674 * On failure, returns -errno (not mounted yields -EINVAL)
1675 * Is noisy on failures, expects to be given a mounted device.
1677 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size)
1682 ret = is_block_device(dev);
1685 error("not a block device: %s", dev);
1688 error("cannot check %s: %s", dev, strerror(-ret));
1693 fd = open(dev, O_RDONLY);
1696 error("cannot open %s: %s", dev, strerror(errno));
1700 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
1703 } else { /* mounted, all good */
1713 * Given a pathname, return a filehandle to:
1714 * the original pathname or,
1715 * if the pathname is a mounted btrfs device, to its mountpoint.
1717 * On error, return -1, errno should be set.
1719 int open_path_or_dev_mnt(const char *path, DIR **dirstream, int verbose)
1724 if (is_block_device(path)) {
1725 ret = get_btrfs_mount(path, mp, sizeof(mp));
1727 /* not a mounted btrfs dev */
1728 error_on(verbose, "'%s' is not a mounted btrfs device",
1733 ret = open_file_or_dir(mp, dirstream);
1734 error_on(verbose && ret < 0, "can't access '%s': %s",
1735 path, strerror(errno));
1737 ret = btrfs_open_dir(path, dirstream, 1);
1744 * Do the following checks before calling open_file_or_dir():
1745 * 1: path is in a btrfs filesystem
1746 * 2: path is a directory
1748 int btrfs_open_dir(const char *path, DIR **dirstream, int verbose)
1754 if (statfs(path, &stfs) != 0) {
1755 error_on(verbose, "cannot access '%s': %s", path,
1760 if (stfs.f_type != BTRFS_SUPER_MAGIC) {
1761 error_on(verbose, "not a btrfs filesystem: %s", path);
1765 if (stat(path, &st) != 0) {
1766 error_on(verbose, "cannot access '%s': %s", path,
1771 if (!S_ISDIR(st.st_mode)) {
1772 error_on(verbose, "not a directory: %s", path);
1776 ret = open_file_or_dir(path, dirstream);
1778 error_on(verbose, "cannot access '%s': %s", path,
1785 /* checks if a device is a loop device */
1786 static int is_loop_device (const char* device) {
1787 struct stat statbuf;
1789 if(stat(device, &statbuf) < 0)
1792 return (S_ISBLK(statbuf.st_mode) &&
1793 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
1797 * Takes a loop device path (e.g. /dev/loop0) and returns
1798 * the associated file (e.g. /images/my_btrfs.img) using
1801 static int resolve_loop_device_with_loopdev(const char* loop_dev, char* loop_file)
1805 struct loop_info64 lo64;
1807 fd = open(loop_dev, O_RDONLY | O_NONBLOCK);
1810 ret = ioctl(fd, LOOP_GET_STATUS64, &lo64);
1816 memcpy(loop_file, lo64.lo_file_name, sizeof(lo64.lo_file_name));
1817 loop_file[sizeof(lo64.lo_file_name)] = 0;
1825 /* Takes a loop device path (e.g. /dev/loop0) and returns
1826 * the associated file (e.g. /images/my_btrfs.img) */
1827 static int resolve_loop_device(const char* loop_dev, char* loop_file,
1834 char real_loop_dev[PATH_MAX];
1836 if (!realpath(loop_dev, real_loop_dev))
1838 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
1839 if (!(f = fopen(p, "r"))) {
1840 if (errno == ENOENT)
1842 * It's possibly a partitioned loop device, which is
1843 * resolvable with loopdev API.
1845 return resolve_loop_device_with_loopdev(loop_dev, loop_file);
1849 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
1850 ret = fscanf(f, fmt, loop_file);
1859 * Checks whether a and b are identical or device
1860 * files associated with the same block device
1862 static int is_same_blk_file(const char* a, const char* b)
1864 struct stat st_buf_a, st_buf_b;
1865 char real_a[PATH_MAX];
1866 char real_b[PATH_MAX];
1868 if (!realpath(a, real_a))
1869 strncpy_null(real_a, a);
1871 if (!realpath(b, real_b))
1872 strncpy_null(real_b, b);
1874 /* Identical path? */
1875 if (strcmp(real_a, real_b) == 0)
1878 if (stat(a, &st_buf_a) < 0 || stat(b, &st_buf_b) < 0) {
1879 if (errno == ENOENT)
1884 /* Same blockdevice? */
1885 if (S_ISBLK(st_buf_a.st_mode) && S_ISBLK(st_buf_b.st_mode) &&
1886 st_buf_a.st_rdev == st_buf_b.st_rdev) {
1891 if (st_buf_a.st_dev == st_buf_b.st_dev &&
1892 st_buf_a.st_ino == st_buf_b.st_ino) {
1899 /* checks if a and b are identical or device
1900 * files associated with the same block device or
1901 * if one file is a loop device that uses the other
1904 static int is_same_loop_file(const char* a, const char* b)
1906 char res_a[PATH_MAX];
1907 char res_b[PATH_MAX];
1908 const char* final_a = NULL;
1909 const char* final_b = NULL;
1912 /* Resolve a if it is a loop device */
1913 if((ret = is_loop_device(a)) < 0) {
1918 ret = resolve_loop_device(a, res_a, sizeof(res_a));
1929 /* Resolve b if it is a loop device */
1930 if ((ret = is_loop_device(b)) < 0) {
1935 ret = resolve_loop_device(b, res_b, sizeof(res_b));
1946 return is_same_blk_file(final_a, final_b);
1949 /* Checks if a file exists and is a block or regular file*/
1950 static int is_existing_blk_or_reg_file(const char* filename)
1954 if(stat(filename, &st_buf) < 0) {
1961 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
1964 /* Checks if a file is used (directly or indirectly via a loop device)
1965 * by a device in fs_devices
1967 static int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices,
1971 struct list_head *head;
1972 struct list_head *cur;
1973 struct btrfs_device *device;
1975 head = &fs_devices->devices;
1976 list_for_each(cur, head) {
1977 device = list_entry(cur, struct btrfs_device, dev_list);
1979 if((ret = is_same_loop_file(device->name, file)))
1987 * Resolve a pathname to a device mapper node to /dev/mapper/<name>
1988 * Returns NULL on invalid input or malloc failure; Other failures
1989 * will be handled by the caller using the input pathame.
1991 char *canonicalize_dm_name(const char *ptname)
1995 char path[PATH_MAX], name[PATH_MAX], *res = NULL;
1997 if (!ptname || !*ptname)
2000 snprintf(path, sizeof(path), "/sys/block/%s/dm/name", ptname);
2001 if (!(f = fopen(path, "r")))
2004 /* read <name>\n from sysfs */
2005 if (fgets(name, sizeof(name), f) && (sz = strlen(name)) > 1) {
2006 name[sz - 1] = '\0';
2007 snprintf(path, sizeof(path), "/dev/mapper/%s", name);
2009 if (access(path, F_OK) == 0)
2017 * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or
2018 * to a device mapper pathname.
2019 * Returns NULL on invalid input or malloc failure; Other failures
2020 * will be handled by the caller using the input pathame.
2022 char *canonicalize_path(const char *path)
2024 char *canonical, *p;
2026 if (!path || !*path)
2029 canonical = realpath(path, NULL);
2031 return strdup(path);
2032 p = strrchr(canonical, '/');
2033 if (p && strncmp(p, "/dm-", 4) == 0 && isdigit(*(p + 4))) {
2034 char *dm = canonicalize_dm_name(p + 1);
2045 * returns 1 if the device was mounted, < 0 on error or 0 if everything
2046 * is safe to continue.
2048 int check_mounted(const char* file)
2053 fd = open(file, O_RDONLY);
2055 error("mount check: cannot open %s: %s", file,
2060 ret = check_mounted_where(fd, file, NULL, 0, NULL);
2066 int check_mounted_where(int fd, const char *file, char *where, int size,
2067 struct btrfs_fs_devices **fs_dev_ret)
2072 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2076 /* scan the initial device */
2077 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
2078 &total_devs, BTRFS_SUPER_INFO_OFFSET, 0);
2079 is_btrfs = (ret >= 0);
2081 /* scan other devices */
2082 if (is_btrfs && total_devs > 1) {
2083 ret = btrfs_scan_lblkid();
2088 /* iterate over the list of currently mounted filesystems */
2089 if ((f = setmntent ("/proc/self/mounts", "r")) == NULL)
2092 while ((mnt = getmntent (f)) != NULL) {
2094 if(strcmp(mnt->mnt_type, "btrfs") != 0)
2097 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
2099 /* ignore entries in the mount table that are not
2100 associated with a file*/
2101 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
2102 goto out_mntloop_err;
2106 ret = is_same_loop_file(file, mnt->mnt_fsname);
2110 goto out_mntloop_err;
2115 /* Did we find an entry in mnt table? */
2116 if (mnt && size && where) {
2117 strncpy(where, mnt->mnt_dir, size);
2121 *fs_dev_ret = fs_devices_mnt;
2123 ret = (mnt != NULL);
2131 struct pending_dir {
2132 struct list_head list;
2133 char name[PATH_MAX];
2136 int btrfs_register_one_device(const char *fname)
2138 struct btrfs_ioctl_vol_args args;
2142 fd = open("/dev/btrfs-control", O_RDWR);
2145 "failed to open /dev/btrfs-control, skipping device registration: %s",
2149 memset(&args, 0, sizeof(args));
2150 strncpy_null(args.name, fname);
2151 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
2153 error("device scan failed on '%s': %s", fname,
2162 * Register all devices in the fs_uuid list created in the user
2163 * space. Ensure btrfs_scan_lblkid() is called before this func.
2165 int btrfs_register_all_devices(void)
2169 struct btrfs_fs_devices *fs_devices;
2170 struct btrfs_device *device;
2171 struct list_head *all_uuids;
2173 all_uuids = btrfs_scanned_uuids();
2175 list_for_each_entry(fs_devices, all_uuids, list) {
2176 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2178 err = btrfs_register_one_device(device->name);
2188 int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
2191 struct btrfs_super_block *disk_super;
2195 buf = malloc(BTRFS_SUPER_INFO_SIZE);
2200 ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
2201 if (ret != BTRFS_SUPER_INFO_SIZE)
2205 disk_super = (struct btrfs_super_block *)buf;
2206 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC)
2209 if (!memcmp(disk_super->fsid, root->fs_info->super_copy->fsid,
2219 * Note: this function uses a static per-thread buffer. Do not call this
2220 * function more than 10 times within one argument list!
2222 const char *pretty_size_mode(u64 size, unsigned mode)
2224 static __thread int ps_index = 0;
2225 static __thread char ps_array[10][32];
2228 ret = ps_array[ps_index];
2231 (void)pretty_size_snprintf(size, ret, 32, mode);
2236 static const char* unit_suffix_binary[] =
2237 { "B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB"};
2238 static const char* unit_suffix_decimal[] =
2239 { "B", "kB", "MB", "GB", "TB", "PB", "EB"};
2241 int pretty_size_snprintf(u64 size, char *str, size_t str_size, unsigned unit_mode)
2247 const char** suffix = NULL;
2253 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_RAW) {
2254 snprintf(str, str_size, "%llu", size);
2258 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_BINARY) {
2261 suffix = unit_suffix_binary;
2262 } else if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_DECIMAL) {
2265 suffix = unit_suffix_decimal;
2270 fprintf(stderr, "INTERNAL ERROR: unknown unit base, mode %d\n",
2278 switch (unit_mode & UNITS_MODE_MASK) {
2279 case UNITS_TBYTES: base *= mult; num_divs++;
2280 case UNITS_GBYTES: base *= mult; num_divs++;
2281 case UNITS_MBYTES: base *= mult; num_divs++;
2282 case UNITS_KBYTES: num_divs++;
2289 while (size >= mult) {
2295 * If the value is smaller than base, we didn't do any
2296 * division, in that case, base should be 1, not original
2297 * base, or the unit will be wrong
2303 if (num_divs >= ARRAY_SIZE(unit_suffix_binary)) {
2305 printf("INTERNAL ERROR: unsupported unit suffix, index %d\n",
2310 fraction = (float)last_size / base;
2312 return snprintf(str, str_size, "%.2f%s", fraction, suffix[num_divs]);
2316 * __strncpy_null - strncpy with null termination
2317 * @dest: the target array
2318 * @src: the source string
2319 * @n: maximum bytes to copy (size of *dest)
2321 * Like strncpy, but ensures destination is null-terminated.
2323 * Copies the string pointed to by src, including the terminating null
2324 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
2325 * of n bytes. Then ensure that dest is null-terminated.
2327 char *__strncpy_null(char *dest, const char *src, size_t n)
2329 strncpy(dest, src, n);
2336 * Checks to make sure that the label matches our requirements.
2338 0 if everything is safe and usable
2339 -1 if the label is too long
2341 static int check_label(const char *input)
2343 int len = strlen(input);
2345 if (len > BTRFS_LABEL_SIZE - 1) {
2346 error("label %s is too long (max %d)", input,
2347 BTRFS_LABEL_SIZE - 1);
2354 static int set_label_unmounted(const char *dev, const char *label)
2356 struct btrfs_trans_handle *trans;
2357 struct btrfs_root *root;
2360 ret = check_mounted(dev);
2362 error("checking mount status of %s failed: %d", dev, ret);
2366 error("device %s is mounted, use mount point", dev);
2370 /* Open the super_block at the default location
2371 * and as read-write.
2373 root = open_ctree(dev, 0, OPEN_CTREE_WRITES);
2374 if (!root) /* errors are printed by open_ctree() */
2377 trans = btrfs_start_transaction(root, 1);
2378 __strncpy_null(root->fs_info->super_copy->label, label, BTRFS_LABEL_SIZE - 1);
2380 btrfs_commit_transaction(trans, root);
2382 /* Now we close it since we are done. */
2387 static int set_label_mounted(const char *mount_path, const char *labelp)
2390 char label[BTRFS_LABEL_SIZE];
2392 fd = open(mount_path, O_RDONLY | O_NOATIME);
2394 error("unable to access %s: %s", mount_path, strerror(errno));
2398 memset(label, 0, sizeof(label));
2399 __strncpy_null(label, labelp, BTRFS_LABEL_SIZE - 1);
2400 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
2401 error("unable to set label of %s: %s", mount_path,
2411 int get_label_unmounted(const char *dev, char *label)
2413 struct btrfs_root *root;
2416 ret = check_mounted(dev);
2418 error("checking mount status of %s failed: %d", dev, ret);
2422 /* Open the super_block at the default location
2425 root = open_ctree(dev, 0, 0);
2429 __strncpy_null(label, root->fs_info->super_copy->label,
2430 BTRFS_LABEL_SIZE - 1);
2432 /* Now we close it since we are done. */
2438 * If a partition is mounted, try to get the filesystem label via its
2439 * mounted path rather than device. Return the corresponding error
2440 * the user specified the device path.
2442 int get_label_mounted(const char *mount_path, char *labelp)
2444 char label[BTRFS_LABEL_SIZE];
2448 fd = open(mount_path, O_RDONLY | O_NOATIME);
2450 error("unable to access %s: %s", mount_path, strerror(errno));
2454 memset(label, '\0', sizeof(label));
2455 ret = ioctl(fd, BTRFS_IOC_GET_FSLABEL, label);
2457 if (errno != ENOTTY)
2458 error("unable to get label of %s: %s", mount_path,
2465 __strncpy_null(labelp, label, BTRFS_LABEL_SIZE - 1);
2470 int get_label(const char *btrfs_dev, char *label)
2474 ret = is_existing_blk_or_reg_file(btrfs_dev);
2476 ret = get_label_mounted(btrfs_dev, label);
2478 ret = get_label_unmounted(btrfs_dev, label);
2483 int set_label(const char *btrfs_dev, const char *label)
2487 if (check_label(label))
2490 ret = is_existing_blk_or_reg_file(btrfs_dev);
2492 ret = set_label_mounted(btrfs_dev, label);
2494 ret = set_label_unmounted(btrfs_dev, label);
2500 * A not-so-good version fls64. No fascinating optimization since
2501 * no one except parse_size use it
2503 static int fls64(u64 x)
2507 for (i = 0; i <64; i++)
2508 if (x << i & (1ULL << 63))
2513 u64 parse_size(char *s)
2521 error("size value is empty");
2525 error("size value '%s' is less equal than 0", s);
2528 ret = strtoull(s, &endptr, 10);
2530 error("size value '%s' is invalid", s);
2533 if (endptr[0] && endptr[1]) {
2534 error("illegal suffix contains character '%c' in wrong position",
2539 * strtoll returns LLONG_MAX when overflow, if this happens,
2540 * need to call strtoull to get the real size
2542 if (errno == ERANGE && ret == ULLONG_MAX) {
2543 error("size value '%s' is too large for u64", s);
2547 c = tolower(endptr[0]);
2570 error("unknown size descriptor '%c'", c);
2574 /* Check whether ret * mult overflow */
2575 if (fls64(ret) + fls64(mult) - 1 > 64) {
2576 error("size value '%s' is too large for u64", s);
2583 u64 parse_qgroupid(const char *p)
2585 char *s = strchr(p, '/');
2586 const char *ptr_src_end = p + strlen(p);
2587 char *ptr_parse_end = NULL;
2596 /* Numeric format like '0/257' is the primary case */
2598 id = strtoull(p, &ptr_parse_end, 10);
2599 if (ptr_parse_end != ptr_src_end)
2603 level = strtoull(p, &ptr_parse_end, 10);
2604 if (ptr_parse_end != s)
2607 id = strtoull(s + 1, &ptr_parse_end, 10);
2608 if (ptr_parse_end != ptr_src_end)
2611 return (level << BTRFS_QGROUP_LEVEL_SHIFT) | id;
2614 /* Path format like subv at 'my_subvol' is the fallback case */
2615 ret = test_issubvolume(p);
2616 if (ret < 0 || !ret)
2618 fd = open(p, O_RDONLY);
2621 ret = lookup_ino_rootid(fd, &id);
2628 error("invalid qgroupid or subvolume path: %s", p);
2632 int open_file_or_dir3(const char *fname, DIR **dirstream, int open_flags)
2638 ret = stat(fname, &st);
2642 if (S_ISDIR(st.st_mode)) {
2643 *dirstream = opendir(fname);
2646 fd = dirfd(*dirstream);
2647 } else if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) {
2648 fd = open(fname, open_flags);
2651 * we set this on purpose, in case the caller output
2652 * strerror(errno) as success
2660 closedir(*dirstream);
2667 int open_file_or_dir(const char *fname, DIR **dirstream)
2669 return open_file_or_dir3(fname, dirstream, O_RDWR);
2672 void close_file_or_dir(int fd, DIR *dirstream)
2675 closedir(dirstream);
2680 int get_device_info(int fd, u64 devid,
2681 struct btrfs_ioctl_dev_info_args *di_args)
2685 di_args->devid = devid;
2686 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
2688 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
2689 return ret < 0 ? -errno : 0;
2692 static u64 find_max_device_id(struct btrfs_ioctl_search_args *search_args,
2695 struct btrfs_dev_item *dev_item;
2696 char *buf = search_args->buf;
2698 buf += (nr_items - 1) * (sizeof(struct btrfs_ioctl_search_header)
2699 + sizeof(struct btrfs_dev_item));
2700 buf += sizeof(struct btrfs_ioctl_search_header);
2702 dev_item = (struct btrfs_dev_item *)buf;
2704 return btrfs_stack_device_id(dev_item);
2707 static int search_chunk_tree_for_fs_info(int fd,
2708 struct btrfs_ioctl_fs_info_args *fi_args)
2712 u64 start_devid = 1;
2713 struct btrfs_ioctl_search_args search_args;
2714 struct btrfs_ioctl_search_key *search_key = &search_args.key;
2716 fi_args->num_devices = 0;
2718 max_items = BTRFS_SEARCH_ARGS_BUFSIZE
2719 / (sizeof(struct btrfs_ioctl_search_header)
2720 + sizeof(struct btrfs_dev_item));
2722 search_key->tree_id = BTRFS_CHUNK_TREE_OBJECTID;
2723 search_key->min_objectid = BTRFS_DEV_ITEMS_OBJECTID;
2724 search_key->max_objectid = BTRFS_DEV_ITEMS_OBJECTID;
2725 search_key->min_type = BTRFS_DEV_ITEM_KEY;
2726 search_key->max_type = BTRFS_DEV_ITEM_KEY;
2727 search_key->min_transid = 0;
2728 search_key->max_transid = (u64)-1;
2729 search_key->nr_items = max_items;
2730 search_key->max_offset = (u64)-1;
2733 search_key->min_offset = start_devid;
2735 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH, &search_args);
2739 fi_args->num_devices += (u64)search_key->nr_items;
2741 if (search_key->nr_items == max_items) {
2742 start_devid = find_max_device_id(&search_args,
2743 search_key->nr_items) + 1;
2747 /* get the lastest max_id to stay consistent with the num_devices */
2748 if (search_key->nr_items == 0)
2750 * last tree_search returns an empty buf, use the devid of
2751 * the last dev_item of the previous tree_search
2753 fi_args->max_id = start_devid - 1;
2755 fi_args->max_id = find_max_device_id(&search_args,
2756 search_key->nr_items);
2762 * For a given path, fill in the ioctl fs_ and info_ args.
2763 * If the path is a btrfs mountpoint, fill info for all devices.
2764 * If the path is a btrfs device, fill in only that device.
2766 * The path provided must be either on a mounted btrfs fs,
2767 * or be a mounted btrfs device.
2769 * Returns 0 on success, or a negative errno.
2771 int get_fs_info(char *path, struct btrfs_ioctl_fs_info_args *fi_args,
2772 struct btrfs_ioctl_dev_info_args **di_ret)
2779 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2780 struct btrfs_ioctl_dev_info_args *di_args;
2781 struct btrfs_ioctl_dev_info_args tmp;
2783 DIR *dirstream = NULL;
2785 memset(fi_args, 0, sizeof(*fi_args));
2787 if (is_block_device(path) == 1) {
2788 struct btrfs_super_block *disk_super;
2789 char buf[BTRFS_SUPER_INFO_SIZE];
2792 /* Ensure it's mounted, then set path to the mountpoint */
2793 fd = open(path, O_RDONLY);
2796 error("cannot open %s: %s", path, strerror(errno));
2799 ret = check_mounted_where(fd, path, mp, sizeof(mp),
2808 /* Only fill in this one device */
2809 fi_args->num_devices = 1;
2811 disk_super = (struct btrfs_super_block *)buf;
2812 ret = btrfs_read_dev_super(fd, disk_super,
2813 BTRFS_SUPER_INFO_OFFSET, 0);
2818 devid = btrfs_stack_device_id(&disk_super->dev_item);
2820 fi_args->max_id = devid;
2823 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
2827 /* at this point path must not be for a block device */
2828 fd = open_file_or_dir(path, &dirstream);
2834 /* fill in fi_args if not just a single device */
2835 if (fi_args->num_devices != 1) {
2836 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
2843 * The fs_args->num_devices does not include seed devices
2845 ret = search_chunk_tree_for_fs_info(fd, fi_args);
2850 * search_chunk_tree_for_fs_info() will lacks the devid 0
2851 * so manual probe for it here.
2853 ret = get_device_info(fd, 0, &tmp);
2855 fi_args->num_devices++;
2863 if (!fi_args->num_devices)
2866 di_args = *di_ret = malloc((fi_args->num_devices) * sizeof(*di_args));
2873 memcpy(di_args, &tmp, sizeof(tmp));
2874 for (; i <= fi_args->max_id; ++i) {
2875 ret = get_device_info(fd, i, &di_args[ndevs]);
2884 * only when the only dev we wanted to find is not there then
2885 * let any error be returned
2887 if (fi_args->num_devices != 1) {
2893 close_file_or_dir(fd, dirstream);
2897 #define isoctal(c) (((c) & ~7) == '0')
2899 static inline void translate(char *f, char *t)
2901 while (*f != '\0') {
2903 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
2904 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
2914 * Checks if the swap device.
2915 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
2917 static int is_swap_device(const char *file)
2928 if (stat(file, &st_buf) < 0)
2930 if (S_ISBLK(st_buf.st_mode))
2931 dev = st_buf.st_rdev;
2932 else if (S_ISREG(st_buf.st_mode)) {
2933 dev = st_buf.st_dev;
2934 ino = st_buf.st_ino;
2938 if ((f = fopen("/proc/swaps", "r")) == NULL)
2941 /* skip the first line */
2942 if (fgets(tmp, sizeof(tmp), f) == NULL)
2945 while (fgets(tmp, sizeof(tmp), f) != NULL) {
2946 if ((cp = strchr(tmp, ' ')) != NULL)
2948 if ((cp = strchr(tmp, '\t')) != NULL)
2950 translate(tmp, buf);
2951 if (stat(buf, &st_buf) != 0)
2953 if (S_ISBLK(st_buf.st_mode)) {
2954 if (dev == st_buf.st_rdev) {
2958 } else if (S_ISREG(st_buf.st_mode)) {
2959 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {
2973 * Check for existing filesystem or partition table on device.
2975 * 1 for existing fs or partition
2976 * 0 for nothing found
2977 * -1 for internal error
2979 static int check_overwrite(const char *device)
2982 blkid_probe pr = NULL;
2986 if (!device || !*device)
2989 ret = -1; /* will reset on success of all setup calls */
2991 pr = blkid_new_probe_from_filename(device);
2995 size = blkid_probe_get_size(pr);
2999 /* nothing to overwrite on a 0-length device */
3005 ret = blkid_probe_enable_partitions(pr, 1);
3009 ret = blkid_do_fullprobe(pr);
3014 * Blkid returns 1 for nothing found and 0 when it finds a signature,
3015 * but we want the exact opposite, so reverse the return value here.
3017 * In addition print some useful diagnostics about what actually is
3025 if (!blkid_probe_lookup_value(pr, "TYPE", &type, NULL)) {
3027 "%s appears to contain an existing "
3028 "filesystem (%s).\n", device, type);
3029 } else if (!blkid_probe_lookup_value(pr, "PTTYPE", &type, NULL)) {
3031 "%s appears to contain a partition "
3032 "table (%s).\n", device, type);
3035 "%s appears to contain something weird "
3036 "according to blkid\n", device);
3042 blkid_free_probe(pr);
3045 "probe of %s failed, cannot detect "
3046 "existing filesystem.\n", device);
3050 static int group_profile_devs_min(u64 flag)
3052 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3053 case 0: /* single */
3054 case BTRFS_BLOCK_GROUP_DUP:
3056 case BTRFS_BLOCK_GROUP_RAID0:
3057 case BTRFS_BLOCK_GROUP_RAID1:
3058 case BTRFS_BLOCK_GROUP_RAID5:
3060 case BTRFS_BLOCK_GROUP_RAID6:
3062 case BTRFS_BLOCK_GROUP_RAID10:
3069 int test_num_disk_vs_raid(u64 metadata_profile, u64 data_profile,
3070 u64 dev_cnt, int mixed, int ssd)
3077 allowed |= BTRFS_BLOCK_GROUP_RAID10;
3079 allowed |= BTRFS_BLOCK_GROUP_RAID6;
3081 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
3082 BTRFS_BLOCK_GROUP_RAID5;
3084 allowed |= BTRFS_BLOCK_GROUP_DUP;
3088 ((metadata_profile | data_profile) & BTRFS_BLOCK_GROUP_DUP)) {
3089 warning("DUP is not recommended on filesystem with multiple devices");
3091 if (metadata_profile & ~allowed) {
3093 "ERROR: unable to create FS with metadata profile %s "
3094 "(have %llu devices but %d devices are required)\n",
3095 btrfs_group_profile_str(metadata_profile), dev_cnt,
3096 group_profile_devs_min(metadata_profile));
3099 if (data_profile & ~allowed) {
3101 "ERROR: unable to create FS with data profile %s "
3102 "(have %llu devices but %d devices are required)\n",
3103 btrfs_group_profile_str(data_profile), dev_cnt,
3104 group_profile_devs_min(data_profile));
3108 warning_on(!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP) && ssd,
3109 "DUP may not actually lead to 2 copies on the device, see manual page");
3114 int group_profile_max_safe_loss(u64 flags)
3116 switch (flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3117 case 0: /* single */
3118 case BTRFS_BLOCK_GROUP_DUP:
3119 case BTRFS_BLOCK_GROUP_RAID0:
3121 case BTRFS_BLOCK_GROUP_RAID1:
3122 case BTRFS_BLOCK_GROUP_RAID5:
3123 case BTRFS_BLOCK_GROUP_RAID10:
3125 case BTRFS_BLOCK_GROUP_RAID6:
3133 * Check if a device is suitable for btrfs
3135 * 1: something is wrong, an error is printed
3138 int test_dev_for_mkfs(const char *file, int force_overwrite)
3143 ret = is_swap_device(file);
3145 error("checking status of %s: %s", file, strerror(-ret));
3149 error("%s is a swap device", file);
3152 if (!force_overwrite) {
3153 if (check_overwrite(file)) {
3154 error("use the -f option to force overwrite of %s",
3159 ret = check_mounted(file);
3161 error("cannot check mount status of %s: %s", file,
3166 error("%s is mounted", file);
3169 /* check if the device is busy */
3170 fd = open(file, O_RDWR|O_EXCL);
3172 error("unable to open %s: %s", file, strerror(errno));
3175 if (fstat(fd, &st)) {
3176 error("unable to stat %s: %s", file, strerror(errno));
3180 if (!S_ISBLK(st.st_mode)) {
3181 error("%s is not a block device", file);
3189 int btrfs_scan_lblkid(void)
3194 struct btrfs_fs_devices *tmp_devices;
3195 blkid_dev_iterate iter = NULL;
3196 blkid_dev dev = NULL;
3197 blkid_cache cache = NULL;
3198 char path[PATH_MAX];
3200 if (btrfs_scan_done)
3203 if (blkid_get_cache(&cache, NULL) < 0) {
3204 error("blkid cache get failed");
3207 blkid_probe_all(cache);
3208 iter = blkid_dev_iterate_begin(cache);
3209 blkid_dev_set_search(iter, "TYPE", "btrfs");
3210 while (blkid_dev_next(iter, &dev) == 0) {
3211 dev = blkid_verify(cache, dev);
3214 /* if we are here its definitely a btrfs disk*/
3215 strncpy_null(path, blkid_dev_devname(dev));
3217 fd = open(path, O_RDONLY);
3219 error("cannot open %s: %s", path, strerror(errno));
3222 ret = btrfs_scan_one_device(fd, path, &tmp_devices,
3223 &num_devices, BTRFS_SUPER_INFO_OFFSET, 0);
3225 error("cannot scan %s: %s", path, strerror(-ret));
3232 blkid_dev_iterate_end(iter);
3233 blkid_put_cache(cache);
3235 btrfs_scan_done = 1;
3240 int is_vol_small(const char *file)
3247 fd = open(file, O_RDONLY);
3250 if (fstat(fd, &st) < 0) {
3255 size = btrfs_device_size(fd, &st);
3260 if (size < BTRFS_MKFS_SMALL_VOLUME_SIZE) {
3270 * This reads a line from the stdin and only returns non-zero if the
3271 * first whitespace delimited token is a case insensitive match with yes
3274 int ask_user(const char *question)
3276 char buf[30] = {0,};
3277 char *saveptr = NULL;
3280 printf("%s [y/N]: ", question);
3282 return fgets(buf, sizeof(buf) - 1, stdin) &&
3283 (answer = strtok_r(buf, " \t\n\r", &saveptr)) &&
3284 (!strcasecmp(answer, "yes") || !strcasecmp(answer, "y"));
3289 * - file or directory return the containing tree root id
3290 * - subvolume return its own tree id
3291 * - BTRFS_EMPTY_SUBVOL_DIR_OBJECTID (directory with ino == 2) the result is
3292 * undefined and function returns -1
3294 int lookup_ino_rootid(int fd, u64 *rootid)
3296 struct btrfs_ioctl_ino_lookup_args args;
3299 memset(&args, 0, sizeof(args));
3301 args.objectid = BTRFS_FIRST_FREE_OBJECTID;
3303 ret = ioctl(fd, BTRFS_IOC_INO_LOOKUP, &args);
3305 error("failed to lookup root id: %s", strerror(errno));
3309 *rootid = args.treeid;
3315 * return 0 if a btrfs mount point is found
3316 * return 1 if a mount point is found but not btrfs
3317 * return <0 if something goes wrong
3319 int find_mount_root(const char *path, char **mount_root)
3327 int longest_matchlen = 0;
3328 char *longest_match = NULL;
3330 fd = open(path, O_RDONLY | O_NOATIME);
3335 mnttab = setmntent("/proc/self/mounts", "r");
3339 while ((ent = getmntent(mnttab))) {
3340 len = strlen(ent->mnt_dir);
3341 if (strncmp(ent->mnt_dir, path, len) == 0) {
3342 /* match found and use the latest match */
3343 if (longest_matchlen <= len) {
3344 free(longest_match);
3345 longest_matchlen = len;
3346 longest_match = strdup(ent->mnt_dir);
3347 not_btrfs = strcmp(ent->mnt_type, "btrfs");
3356 free(longest_match);
3361 *mount_root = realpath(longest_match, NULL);
3365 free(longest_match);
3369 int test_minimum_size(const char *file, u32 nodesize)
3372 struct stat statbuf;
3374 fd = open(file, O_RDONLY);
3377 if (stat(file, &statbuf) < 0) {
3381 if (btrfs_device_size(fd, &statbuf) < btrfs_min_dev_size(nodesize)) {
3391 * Test if path is a directory
3393 * 0 - path exists but it is not a directory
3394 * 1 - path exists and it is a directory
3397 int test_isdir(const char *path)
3402 ret = stat(path, &st);
3406 return !!S_ISDIR(st.st_mode);
3409 void units_set_mode(unsigned *units, unsigned mode)
3411 unsigned base = *units & UNITS_MODE_MASK;
3413 *units = base | mode;
3416 void units_set_base(unsigned *units, unsigned base)
3418 unsigned mode = *units & ~UNITS_MODE_MASK;
3420 *units = base | mode;
3423 int find_next_key(struct btrfs_path *path, struct btrfs_key *key)
3427 for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
3428 if (!path->nodes[level])
3430 if (path->slots[level] + 1 >=
3431 btrfs_header_nritems(path->nodes[level]))
3434 btrfs_item_key_to_cpu(path->nodes[level], key,
3435 path->slots[level] + 1);
3437 btrfs_node_key_to_cpu(path->nodes[level], key,
3438 path->slots[level] + 1);
3444 const char* btrfs_group_type_str(u64 flag)
3446 u64 mask = BTRFS_BLOCK_GROUP_TYPE_MASK |
3447 BTRFS_SPACE_INFO_GLOBAL_RSV;
3449 switch (flag & mask) {
3450 case BTRFS_BLOCK_GROUP_DATA:
3452 case BTRFS_BLOCK_GROUP_SYSTEM:
3454 case BTRFS_BLOCK_GROUP_METADATA:
3456 case BTRFS_BLOCK_GROUP_DATA|BTRFS_BLOCK_GROUP_METADATA:
3457 return "Data+Metadata";
3458 case BTRFS_SPACE_INFO_GLOBAL_RSV:
3459 return "GlobalReserve";
3465 const char* btrfs_group_profile_str(u64 flag)
3467 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3470 case BTRFS_BLOCK_GROUP_RAID0:
3472 case BTRFS_BLOCK_GROUP_RAID1:
3474 case BTRFS_BLOCK_GROUP_RAID5:
3476 case BTRFS_BLOCK_GROUP_RAID6:
3478 case BTRFS_BLOCK_GROUP_DUP:
3480 case BTRFS_BLOCK_GROUP_RAID10:
3487 u64 disk_size(const char *path)
3491 if (statfs(path, &sfs) < 0)
3494 return sfs.f_bsize * sfs.f_blocks;
3497 u64 get_partition_size(const char *dev)
3500 int fd = open(dev, O_RDONLY);
3504 if (ioctl(fd, BLKGETSIZE64, &result) < 0) {
3513 int btrfs_tree_search2_ioctl_supported(int fd)
3515 struct btrfs_ioctl_search_args_v2 *args2;
3516 struct btrfs_ioctl_search_key *sk;
3517 int args2_size = 1024;
3518 char args2_buf[args2_size];
3520 static int v2_supported = -1;
3522 if (v2_supported != -1)
3523 return v2_supported;
3525 args2 = (struct btrfs_ioctl_search_args_v2 *)args2_buf;
3529 * Search for the extent tree item in the root tree.
3531 sk->tree_id = BTRFS_ROOT_TREE_OBJECTID;
3532 sk->min_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3533 sk->max_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3534 sk->min_type = BTRFS_ROOT_ITEM_KEY;
3535 sk->max_type = BTRFS_ROOT_ITEM_KEY;
3537 sk->max_offset = (u64)-1;
3538 sk->min_transid = 0;
3539 sk->max_transid = (u64)-1;
3541 args2->buf_size = args2_size - sizeof(struct btrfs_ioctl_search_args_v2);
3542 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH_V2, args2);
3543 if (ret == -EOPNOTSUPP)
3550 return v2_supported;
3553 int btrfs_check_nodesize(u32 nodesize, u32 sectorsize, u64 features)
3555 if (nodesize < sectorsize) {
3556 error("illegal nodesize %u (smaller than %u)",
3557 nodesize, sectorsize);
3559 } else if (nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) {
3560 error("illegal nodesize %u (larger than %u)",
3561 nodesize, BTRFS_MAX_METADATA_BLOCKSIZE);
3563 } else if (nodesize & (sectorsize - 1)) {
3564 error("illegal nodesize %u (not aligned to %u)",
3565 nodesize, sectorsize);
3567 } else if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS &&
3568 nodesize != sectorsize) {
3569 error("illegal nodesize %u (not equal to %u for mixed block group)",
3570 nodesize, sectorsize);
3577 * Copy a path argument from SRC to DEST and check the SRC length if it's at
3578 * most PATH_MAX and fits into DEST. DESTLEN is supposed to be exact size of
3580 * The destination buffer is zero terminated.
3581 * Return < 0 for error, 0 otherwise.
3583 int arg_copy_path(char *dest, const char *src, int destlen)
3585 size_t len = strlen(src);
3587 if (len >= PATH_MAX || len >= destlen)
3588 return -ENAMETOOLONG;
3590 __strncpy_null(dest, src, destlen);
3595 unsigned int get_unit_mode_from_arg(int *argc, char *argv[], int df_mode)
3597 unsigned int unit_mode = UNITS_DEFAULT;
3601 for (arg_i = 0; arg_i < *argc; arg_i++) {
3602 if (!strcmp(argv[arg_i], "--"))
3605 if (!strcmp(argv[arg_i], "--raw")) {
3606 unit_mode = UNITS_RAW;
3610 if (!strcmp(argv[arg_i], "--human-readable")) {
3611 unit_mode = UNITS_HUMAN_BINARY;
3616 if (!strcmp(argv[arg_i], "--iec")) {
3617 units_set_mode(&unit_mode, UNITS_BINARY);
3621 if (!strcmp(argv[arg_i], "--si")) {
3622 units_set_mode(&unit_mode, UNITS_DECIMAL);
3627 if (!strcmp(argv[arg_i], "--kbytes")) {
3628 units_set_base(&unit_mode, UNITS_KBYTES);
3632 if (!strcmp(argv[arg_i], "--mbytes")) {
3633 units_set_base(&unit_mode, UNITS_MBYTES);
3637 if (!strcmp(argv[arg_i], "--gbytes")) {
3638 units_set_base(&unit_mode, UNITS_GBYTES);
3642 if (!strcmp(argv[arg_i], "--tbytes")) {
3643 units_set_base(&unit_mode, UNITS_TBYTES);
3651 if (!strcmp(argv[arg_i], "-b")) {
3652 unit_mode = UNITS_RAW;
3656 if (!strcmp(argv[arg_i], "-h")) {
3657 unit_mode = UNITS_HUMAN_BINARY;
3661 if (!strcmp(argv[arg_i], "-H")) {
3662 unit_mode = UNITS_HUMAN_DECIMAL;
3666 if (!strcmp(argv[arg_i], "-k")) {
3667 units_set_base(&unit_mode, UNITS_KBYTES);
3671 if (!strcmp(argv[arg_i], "-m")) {
3672 units_set_base(&unit_mode, UNITS_MBYTES);
3676 if (!strcmp(argv[arg_i], "-g")) {
3677 units_set_base(&unit_mode, UNITS_GBYTES);
3681 if (!strcmp(argv[arg_i], "-t")) {
3682 units_set_base(&unit_mode, UNITS_TBYTES);
3688 for (arg_i = 0, arg_end = 0; arg_i < *argc; arg_i++) {
3691 argv[arg_end] = argv[arg_i];
3700 int string_is_numerical(const char *str)
3702 if (!(*str >= '0' && *str <= '9'))
3704 while (*str >= '0' && *str <= '9')
3712 * Preprocess @argv with getopt_long to reorder options and consume the "--"
3714 * Unknown short and long options are reported, optionally the @usage is printed
3717 void clean_args_no_options(int argc, char *argv[], const char * const *usagestr)
3719 static const struct option long_options[] = {
3724 int c = getopt_long(argc, argv, "", long_options, NULL);
3737 /* Subvolume helper functions */
3739 * test if name is a correct subvolume name
3740 * this function return
3741 * 0-> name is not a correct subvolume name
3742 * 1-> name is a correct subvolume name
3744 int test_issubvolname(const char *name)
3746 return name[0] != '\0' && !strchr(name, '/') &&
3747 strcmp(name, ".") && strcmp(name, "..");
3751 * Test if path is a subvolume
3753 * 0 - path exists but it is not a subvolume
3754 * 1 - path exists and it is a subvolume
3757 int test_issubvolume(const char *path)
3763 res = stat(path, &st);
3767 if (st.st_ino != BTRFS_FIRST_FREE_OBJECTID || !S_ISDIR(st.st_mode))
3770 res = statfs(path, &stfs);
3774 return (int)stfs.f_type == BTRFS_SUPER_MAGIC;
3777 const char *subvol_strip_mountpoint(const char *mnt, const char *full_path)
3779 int len = strlen(mnt);
3783 if (mnt[len - 1] != '/')
3786 return full_path + len;
3793 * 1: Error; and error info printed to the terminal. Fixme.
3794 * 2: If the fullpath is root tree instead of subvol tree
3796 int get_subvol_info(const char *fullpath, struct root_info *get_ri)
3803 const char *svpath = NULL;
3804 DIR *dirstream1 = NULL;
3805 DIR *dirstream2 = NULL;
3807 ret = test_issubvolume(fullpath);
3811 error("not a subvolume: %s", fullpath);
3815 ret = find_mount_root(fullpath, &mnt);
3819 error("%s doesn't belong to btrfs mount point", fullpath);
3823 svpath = subvol_strip_mountpoint(mnt, fullpath);
3825 fd = btrfs_open_dir(fullpath, &dirstream1, 1);
3829 ret = btrfs_list_get_path_rootid(fd, &sv_id);
3831 error("can't get rootid for '%s'", fullpath);
3835 mntfd = btrfs_open_dir(mnt, &dirstream2, 1);
3839 if (sv_id == BTRFS_FS_TREE_OBJECTID) {
3842 * So that caller may decide if thats an error or just fine.
3847 memset(get_ri, 0, sizeof(*get_ri));
3848 get_ri->root_id = sv_id;
3850 ret = btrfs_get_subvol(mntfd, get_ri);
3852 error("can't find '%s': %d", svpath, ret);
3855 close_file_or_dir(mntfd, dirstream2);
3856 close_file_or_dir(fd, dirstream1);
3862 void init_rand_seed(u64 seed)
3866 /* only use the last 48 bits */
3867 for (i = 0; i < 3; i++) {
3868 rand_seed[i] = (unsigned short)(seed ^ (unsigned short)(-1));
3871 rand_seed_initlized = 1;
3874 static void __init_seed(void)
3880 if(rand_seed_initlized)
3882 /* Use urandom as primary seed source. */
3883 fd = open("/dev/urandom", O_RDONLY);
3885 ret = read(fd, rand_seed, sizeof(rand_seed));
3887 if (ret < sizeof(rand_seed))
3891 /* Use time and pid as fallback seed */
3892 warning("failed to read /dev/urandom, use time and pid as random seed");
3893 gettimeofday(&tv, 0);
3894 rand_seed[0] = getpid() ^ (tv.tv_sec & 0xFFFF);
3895 rand_seed[1] = getppid() ^ (tv.tv_usec & 0xFFFF);
3896 rand_seed[2] = (tv.tv_sec ^ tv.tv_usec) >> 16;
3898 rand_seed_initlized = 1;
3905 * Don't use nrand48, its range is [0,2^31) The highest bit will alwasy
3906 * be 0. Use jrand48 to include the highest bit.
3908 return (u32)jrand48(rand_seed);
3911 unsigned int rand_range(unsigned int upper)
3915 * Use the full 48bits to mod, which would be more uniformly
3918 return (unsigned int)(jrand48(rand_seed) % upper);