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);
660 * Improved version of make_btrfs().
663 * 1) Do chunk allocation to avoid used data
664 * And after this function, extent type matches chunk type
665 * 2) Better structured code
666 * No super long hand written codes to initialized all tree blocks
667 * Split into small blocks and reuse codes.
668 * TODO: Reuse tree operation facilities by introducing new flags
670 static int make_convert_btrfs(int fd, struct btrfs_mkfs_config *cfg,
671 struct btrfs_convert_context *cctx)
673 struct cache_tree *free = &cctx->free;
674 struct cache_tree *used = &cctx->used;
676 u64 meta_chunk_start;
677 /* chunk tree bytenr, in system chunk */
679 /* metadata trees bytenr, in metadata chunk */
687 /* Shouldn't happen */
688 BUG_ON(cache_tree_empty(used));
691 * reserve space for temporary superblock first
692 * Here we allocate a little larger space, to keep later
693 * free space will be STRIPE_LEN aligned
695 ret = reserve_free_space(free, BTRFS_STRIPE_LEN,
701 * Then reserve system chunk space
702 * TODO: Change system group size depending on cctx->total_bytes.
703 * If using current 4M, it can only handle less than one TB for
704 * worst case and then run out of sys space.
706 ret = reserve_free_space(free, BTRFS_MKFS_SYSTEM_GROUP_SIZE,
710 ret = reserve_free_space(free, BTRFS_CONVERT_META_GROUP_SIZE,
716 * Allocated meta/sys chunks will be mapped 1:1 with device offset.
718 * Inside the allocated metadata chunk, the layout will be:
719 * | offset | contents |
720 * -------------------------------------
722 * | +nodesize | extent root |
723 * | +nodesize * 2 | device root |
724 * | +nodesize * 3 | fs tree |
725 * | +nodesize * 4 | csum tree |
726 * -------------------------------------
727 * Inside the allocated system chunk, the layout will be:
728 * | offset | contents |
729 * -------------------------------------
730 * | +0 | chunk root |
731 * -------------------------------------
733 chunk_bytenr = sys_chunk_start;
734 root_bytenr = meta_chunk_start;
735 extent_bytenr = meta_chunk_start + cfg->nodesize;
736 dev_bytenr = meta_chunk_start + cfg->nodesize * 2;
737 fs_bytenr = meta_chunk_start + cfg->nodesize * 3;
738 csum_bytenr = meta_chunk_start + cfg->nodesize * 4;
740 ret = setup_temp_super(fd, cfg, root_bytenr, chunk_bytenr);
744 ret = setup_temp_root_tree(fd, cfg, root_bytenr, extent_bytenr,
745 dev_bytenr, fs_bytenr, csum_bytenr);
748 ret = setup_temp_chunk_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
752 ret = setup_temp_dev_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
759 * @fs_uuid - if NULL, generates a UUID, returns back the new filesystem UUID
761 int make_btrfs(int fd, struct btrfs_mkfs_config *cfg,
762 struct btrfs_convert_context *cctx)
764 struct btrfs_super_block super;
765 struct extent_buffer *buf;
766 struct btrfs_root_item root_item;
767 struct btrfs_disk_key disk_key;
768 struct btrfs_extent_item *extent_item;
769 struct btrfs_inode_item *inode_item;
770 struct btrfs_chunk *chunk;
771 struct btrfs_dev_item *dev_item;
772 struct btrfs_dev_extent *dev_extent;
773 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
783 int skinny_metadata = !!(cfg->features &
784 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA);
788 return make_convert_btrfs(fd, cfg, cctx);
789 buf = malloc(sizeof(*buf) + max(cfg->sectorsize, cfg->nodesize));
793 first_free = BTRFS_SUPER_INFO_OFFSET + cfg->sectorsize * 2 - 1;
794 first_free &= ~((u64)cfg->sectorsize - 1);
796 memset(&super, 0, sizeof(super));
798 num_bytes = (cfg->num_bytes / cfg->sectorsize) * cfg->sectorsize;
799 if (cfg->fs_uuid && *cfg->fs_uuid) {
800 if (uuid_parse(cfg->fs_uuid, super.fsid) != 0) {
801 error("cannot not parse UUID: %s", cfg->fs_uuid);
805 if (!test_uuid_unique(cfg->fs_uuid)) {
806 error("non-unique UUID: %s", cfg->fs_uuid);
811 uuid_generate(super.fsid);
813 uuid_unparse(super.fsid, cfg->fs_uuid);
815 uuid_generate(super.dev_item.uuid);
816 uuid_generate(chunk_tree_uuid);
818 btrfs_set_super_bytenr(&super, cfg->blocks[0]);
819 btrfs_set_super_num_devices(&super, 1);
820 btrfs_set_super_magic(&super, BTRFS_MAGIC);
821 btrfs_set_super_generation(&super, 1);
822 btrfs_set_super_root(&super, cfg->blocks[1]);
823 btrfs_set_super_chunk_root(&super, cfg->blocks[3]);
824 btrfs_set_super_total_bytes(&super, num_bytes);
825 btrfs_set_super_bytes_used(&super, 6 * cfg->nodesize);
826 btrfs_set_super_sectorsize(&super, cfg->sectorsize);
827 btrfs_set_super_leafsize(&super, cfg->nodesize);
828 btrfs_set_super_nodesize(&super, cfg->nodesize);
829 btrfs_set_super_stripesize(&super, cfg->stripesize);
830 btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
831 btrfs_set_super_chunk_root_generation(&super, 1);
832 btrfs_set_super_cache_generation(&super, -1);
833 btrfs_set_super_incompat_flags(&super, cfg->features);
835 __strncpy_null(super.label, cfg->label, BTRFS_LABEL_SIZE - 1);
837 /* create the tree of root objects */
838 memset(buf->data, 0, cfg->nodesize);
839 buf->len = cfg->nodesize;
840 btrfs_set_header_bytenr(buf, cfg->blocks[1]);
841 btrfs_set_header_nritems(buf, 4);
842 btrfs_set_header_generation(buf, 1);
843 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
844 btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
845 write_extent_buffer(buf, super.fsid, btrfs_header_fsid(),
848 write_extent_buffer(buf, chunk_tree_uuid,
849 btrfs_header_chunk_tree_uuid(buf),
852 /* create the items for the root tree */
853 memset(&root_item, 0, sizeof(root_item));
854 inode_item = &root_item.inode;
855 btrfs_set_stack_inode_generation(inode_item, 1);
856 btrfs_set_stack_inode_size(inode_item, 3);
857 btrfs_set_stack_inode_nlink(inode_item, 1);
858 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
859 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
860 btrfs_set_root_refs(&root_item, 1);
861 btrfs_set_root_used(&root_item, cfg->nodesize);
862 btrfs_set_root_generation(&root_item, 1);
864 memset(&disk_key, 0, sizeof(disk_key));
865 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
866 btrfs_set_disk_key_offset(&disk_key, 0);
869 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - sizeof(root_item);
870 btrfs_set_root_bytenr(&root_item, cfg->blocks[2]);
871 btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
872 btrfs_set_item_key(buf, &disk_key, nritems);
873 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
874 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
876 write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
877 nritems), sizeof(root_item));
880 itemoff = itemoff - sizeof(root_item);
881 btrfs_set_root_bytenr(&root_item, cfg->blocks[4]);
882 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
883 btrfs_set_item_key(buf, &disk_key, nritems);
884 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
885 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
887 write_extent_buffer(buf, &root_item,
888 btrfs_item_ptr_offset(buf, nritems),
892 itemoff = itemoff - sizeof(root_item);
893 btrfs_set_root_bytenr(&root_item, cfg->blocks[5]);
894 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
895 btrfs_set_item_key(buf, &disk_key, nritems);
896 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
897 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
899 write_extent_buffer(buf, &root_item,
900 btrfs_item_ptr_offset(buf, nritems),
904 itemoff = itemoff - sizeof(root_item);
905 btrfs_set_root_bytenr(&root_item, cfg->blocks[6]);
906 btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_TREE_OBJECTID);
907 btrfs_set_item_key(buf, &disk_key, nritems);
908 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
909 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
911 write_extent_buffer(buf, &root_item,
912 btrfs_item_ptr_offset(buf, nritems),
917 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
918 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[1]);
919 if (ret != cfg->nodesize) {
920 ret = (ret < 0 ? -errno : -EIO);
924 /* create the items for the extent tree */
925 memset(buf->data + sizeof(struct btrfs_header), 0,
926 cfg->nodesize - sizeof(struct btrfs_header));
928 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
929 for (i = 1; i < 7; i++) {
930 item_size = sizeof(struct btrfs_extent_item);
931 if (!skinny_metadata)
932 item_size += sizeof(struct btrfs_tree_block_info);
934 BUG_ON(cfg->blocks[i] < first_free);
935 BUG_ON(cfg->blocks[i] < cfg->blocks[i - 1]);
937 /* create extent item */
938 itemoff -= item_size;
939 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
940 if (skinny_metadata) {
941 btrfs_set_disk_key_type(&disk_key,
942 BTRFS_METADATA_ITEM_KEY);
943 btrfs_set_disk_key_offset(&disk_key, 0);
945 btrfs_set_disk_key_type(&disk_key,
946 BTRFS_EXTENT_ITEM_KEY);
947 btrfs_set_disk_key_offset(&disk_key, cfg->nodesize);
949 btrfs_set_item_key(buf, &disk_key, nritems);
950 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
952 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
954 extent_item = btrfs_item_ptr(buf, nritems,
955 struct btrfs_extent_item);
956 btrfs_set_extent_refs(buf, extent_item, 1);
957 btrfs_set_extent_generation(buf, extent_item, 1);
958 btrfs_set_extent_flags(buf, extent_item,
959 BTRFS_EXTENT_FLAG_TREE_BLOCK);
962 /* create extent ref */
963 ref_root = reference_root_table[i];
964 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
965 btrfs_set_disk_key_offset(&disk_key, ref_root);
966 btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
967 btrfs_set_item_key(buf, &disk_key, nritems);
968 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
970 btrfs_set_item_size(buf, btrfs_item_nr(nritems), 0);
973 btrfs_set_header_bytenr(buf, cfg->blocks[2]);
974 btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
975 btrfs_set_header_nritems(buf, nritems);
976 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
977 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[2]);
978 if (ret != cfg->nodesize) {
979 ret = (ret < 0 ? -errno : -EIO);
983 /* create the chunk tree */
984 memset(buf->data + sizeof(struct btrfs_header), 0,
985 cfg->nodesize - sizeof(struct btrfs_header));
987 item_size = sizeof(*dev_item);
988 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - item_size;
990 /* first device 1 (there is no device 0) */
991 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
992 btrfs_set_disk_key_offset(&disk_key, 1);
993 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
994 btrfs_set_item_key(buf, &disk_key, nritems);
995 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
996 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
998 dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
999 btrfs_set_device_id(buf, dev_item, 1);
1000 btrfs_set_device_generation(buf, dev_item, 0);
1001 btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
1002 btrfs_set_device_bytes_used(buf, dev_item,
1003 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1004 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
1005 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
1006 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
1007 btrfs_set_device_type(buf, dev_item, 0);
1009 write_extent_buffer(buf, super.dev_item.uuid,
1010 (unsigned long)btrfs_device_uuid(dev_item),
1012 write_extent_buffer(buf, super.fsid,
1013 (unsigned long)btrfs_device_fsid(dev_item),
1015 read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
1019 item_size = btrfs_chunk_item_size(1);
1020 itemoff = itemoff - item_size;
1022 /* then we have chunk 0 */
1023 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1024 btrfs_set_disk_key_offset(&disk_key, 0);
1025 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
1026 btrfs_set_item_key(buf, &disk_key, nritems);
1027 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1028 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
1030 chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
1031 btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1032 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
1033 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
1034 btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
1035 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
1036 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
1037 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
1038 btrfs_set_chunk_num_stripes(buf, chunk, 1);
1039 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
1040 btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
1043 write_extent_buffer(buf, super.dev_item.uuid,
1044 (unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
1047 /* copy the key for the chunk to the system array */
1048 ptr = super.sys_chunk_array;
1049 array_size = sizeof(disk_key);
1051 memcpy(ptr, &disk_key, sizeof(disk_key));
1052 ptr += sizeof(disk_key);
1054 /* copy the chunk to the system array */
1055 read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
1056 array_size += item_size;
1058 btrfs_set_super_sys_array_size(&super, array_size);
1060 btrfs_set_header_bytenr(buf, cfg->blocks[3]);
1061 btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
1062 btrfs_set_header_nritems(buf, nritems);
1063 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1064 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[3]);
1065 if (ret != cfg->nodesize) {
1066 ret = (ret < 0 ? -errno : -EIO);
1070 /* create the device tree */
1071 memset(buf->data + sizeof(struct btrfs_header), 0,
1072 cfg->nodesize - sizeof(struct btrfs_header));
1074 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) -
1075 sizeof(struct btrfs_dev_extent);
1077 btrfs_set_disk_key_objectid(&disk_key, 1);
1078 btrfs_set_disk_key_offset(&disk_key, 0);
1079 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
1080 btrfs_set_item_key(buf, &disk_key, nritems);
1081 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1082 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1083 sizeof(struct btrfs_dev_extent));
1084 dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
1085 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
1086 BTRFS_CHUNK_TREE_OBJECTID);
1087 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
1088 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1089 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
1091 write_extent_buffer(buf, chunk_tree_uuid,
1092 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
1095 btrfs_set_dev_extent_length(buf, dev_extent,
1096 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1099 btrfs_set_header_bytenr(buf, cfg->blocks[4]);
1100 btrfs_set_header_owner(buf, BTRFS_DEV_TREE_OBJECTID);
1101 btrfs_set_header_nritems(buf, nritems);
1102 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1103 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[4]);
1104 if (ret != cfg->nodesize) {
1105 ret = (ret < 0 ? -errno : -EIO);
1109 /* create the FS root */
1110 memset(buf->data + sizeof(struct btrfs_header), 0,
1111 cfg->nodesize - sizeof(struct btrfs_header));
1112 btrfs_set_header_bytenr(buf, cfg->blocks[5]);
1113 btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
1114 btrfs_set_header_nritems(buf, 0);
1115 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1116 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[5]);
1117 if (ret != cfg->nodesize) {
1118 ret = (ret < 0 ? -errno : -EIO);
1121 /* finally create the csum root */
1122 memset(buf->data + sizeof(struct btrfs_header), 0,
1123 cfg->nodesize - sizeof(struct btrfs_header));
1124 btrfs_set_header_bytenr(buf, cfg->blocks[6]);
1125 btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
1126 btrfs_set_header_nritems(buf, 0);
1127 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1128 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[6]);
1129 if (ret != cfg->nodesize) {
1130 ret = (ret < 0 ? -errno : -EIO);
1134 /* and write out the super block */
1135 BUG_ON(sizeof(super) > cfg->sectorsize);
1136 memset(buf->data, 0, BTRFS_SUPER_INFO_SIZE);
1137 memcpy(buf->data, &super, sizeof(super));
1138 buf->len = BTRFS_SUPER_INFO_SIZE;
1139 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1140 ret = pwrite(fd, buf->data, BTRFS_SUPER_INFO_SIZE, cfg->blocks[0]);
1141 if (ret != BTRFS_SUPER_INFO_SIZE) {
1142 ret = (ret < 0 ? -errno : -EIO);
1153 static const struct btrfs_fs_feature {
1157 } mkfs_features[] = {
1158 { "mixed-bg", BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS,
1159 "mixed data and metadata block groups" },
1160 { "extref", BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF,
1161 "increased hardlink limit per file to 65536" },
1162 { "raid56", BTRFS_FEATURE_INCOMPAT_RAID56,
1163 "raid56 extended format" },
1164 { "skinny-metadata", BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA,
1165 "reduced-size metadata extent refs" },
1166 { "no-holes", BTRFS_FEATURE_INCOMPAT_NO_HOLES,
1167 "no explicit hole extents for files" },
1168 /* Keep this one last */
1169 { "list-all", BTRFS_FEATURE_LIST_ALL, NULL }
1172 static int parse_one_fs_feature(const char *name, u64 *flags)
1177 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1178 if (name[0] == '^' &&
1179 !strcmp(mkfs_features[i].name, name + 1)) {
1180 *flags &= ~ mkfs_features[i].flag;
1182 } else if (!strcmp(mkfs_features[i].name, name)) {
1183 *flags |= mkfs_features[i].flag;
1191 void btrfs_parse_features_to_string(char *buf, u64 flags)
1197 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1198 if (flags & mkfs_features[i].flag) {
1201 strcat(buf, mkfs_features[i].name);
1206 void btrfs_process_fs_features(u64 flags)
1210 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1211 if (flags & mkfs_features[i].flag) {
1212 printf("Turning ON incompat feature '%s': %s\n",
1213 mkfs_features[i].name,
1214 mkfs_features[i].desc);
1219 void btrfs_list_all_fs_features(u64 mask_disallowed)
1223 fprintf(stderr, "Filesystem features available:\n");
1224 for (i = 0; i < ARRAY_SIZE(mkfs_features) - 1; i++) {
1225 char *is_default = "";
1227 if (mkfs_features[i].flag & mask_disallowed)
1229 if (mkfs_features[i].flag & BTRFS_MKFS_DEFAULT_FEATURES)
1230 is_default = ", default";
1231 fprintf(stderr, "%-20s- %s (0x%llx%s)\n",
1232 mkfs_features[i].name,
1233 mkfs_features[i].desc,
1234 mkfs_features[i].flag,
1240 * Return NULL if all features were parsed fine, otherwise return the name of
1241 * the first unparsed.
1243 char* btrfs_parse_fs_features(char *namelist, u64 *flags)
1246 char *save_ptr = NULL; /* Satisfy static checkers */
1248 for (this_char = strtok_r(namelist, ",", &save_ptr);
1250 this_char = strtok_r(NULL, ",", &save_ptr)) {
1251 if (parse_one_fs_feature(this_char, flags))
1258 u64 btrfs_device_size(int fd, struct stat *st)
1261 if (S_ISREG(st->st_mode)) {
1264 if (!S_ISBLK(st->st_mode)) {
1267 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
1273 static int zero_blocks(int fd, off_t start, size_t len)
1275 char *buf = malloc(len);
1281 memset(buf, 0, len);
1282 written = pwrite(fd, buf, len, start);
1289 #define ZERO_DEV_BYTES (2 * 1024 * 1024)
1291 /* don't write outside the device by clamping the region to the device size */
1292 static int zero_dev_clamped(int fd, off_t start, ssize_t len, u64 dev_size)
1294 off_t end = max(start, start + len);
1297 /* and don't overwrite the disk labels on sparc */
1298 start = max(start, 1024);
1299 end = max(end, 1024);
1302 start = min_t(u64, start, dev_size);
1303 end = min_t(u64, end, dev_size);
1305 return zero_blocks(fd, start, end - start);
1308 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
1309 struct btrfs_root *root, int fd, char *path,
1310 u64 device_total_bytes, u32 io_width, u32 io_align,
1313 struct btrfs_super_block *disk_super;
1314 struct btrfs_super_block *super = root->fs_info->super_copy;
1315 struct btrfs_device *device;
1316 struct btrfs_dev_item *dev_item;
1322 device_total_bytes = (device_total_bytes / sectorsize) * sectorsize;
1324 device = kzalloc(sizeof(*device), GFP_NOFS);
1327 buf = kzalloc(sectorsize, GFP_NOFS);
1330 BUG_ON(sizeof(*disk_super) > sectorsize);
1332 disk_super = (struct btrfs_super_block *)buf;
1333 dev_item = &disk_super->dev_item;
1335 uuid_generate(device->uuid);
1338 device->io_width = io_width;
1339 device->io_align = io_align;
1340 device->sector_size = sectorsize;
1342 device->writeable = 1;
1343 device->total_bytes = device_total_bytes;
1344 device->bytes_used = 0;
1345 device->total_ios = 0;
1346 device->dev_root = root->fs_info->dev_root;
1347 device->name = strdup(path);
1351 INIT_LIST_HEAD(&device->dev_list);
1352 ret = btrfs_add_device(trans, root, device);
1355 fs_total_bytes = btrfs_super_total_bytes(super) + device_total_bytes;
1356 btrfs_set_super_total_bytes(super, fs_total_bytes);
1358 num_devs = btrfs_super_num_devices(super) + 1;
1359 btrfs_set_super_num_devices(super, num_devs);
1361 memcpy(disk_super, super, sizeof(*disk_super));
1363 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
1364 btrfs_set_stack_device_id(dev_item, device->devid);
1365 btrfs_set_stack_device_type(dev_item, device->type);
1366 btrfs_set_stack_device_io_align(dev_item, device->io_align);
1367 btrfs_set_stack_device_io_width(dev_item, device->io_width);
1368 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
1369 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
1370 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
1371 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
1373 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
1374 BUG_ON(ret != sectorsize);
1377 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
1378 device->fs_devices = root->fs_info->fs_devices;
1387 static int btrfs_wipe_existing_sb(int fd)
1389 const char *off = NULL;
1394 blkid_probe pr = NULL;
1396 pr = blkid_new_probe();
1400 if (blkid_probe_set_device(pr, fd, 0, 0)) {
1405 ret = blkid_probe_lookup_value(pr, "SBMAGIC_OFFSET", &off, NULL);
1407 ret = blkid_probe_lookup_value(pr, "SBMAGIC", NULL, &len);
1409 if (ret || len == 0 || off == NULL) {
1411 * If lookup fails, the probe did not find any values, eg. for
1412 * a file image or a loop device. Soft error.
1418 offset = strtoll(off, NULL, 10);
1419 if (len > sizeof(buf))
1422 memset(buf, 0, len);
1423 ret = pwrite(fd, buf, len, offset);
1425 error("cannot wipe existing superblock: %s", strerror(errno));
1427 } else if (ret != len) {
1428 error("cannot wipe existing superblock: wrote %d of %zd", ret, len);
1434 blkid_free_probe(pr);
1438 int btrfs_prepare_device(int fd, const char *file, int zero_end,
1439 u64 *block_count_ret, u64 max_block_count, int discard)
1445 ret = fstat(fd, &st);
1447 error("unable to stat %s: %s", file, strerror(errno));
1451 block_count = btrfs_device_size(fd, &st);
1452 if (block_count == 0) {
1453 error("unable to determine size of %s", file);
1456 if (max_block_count)
1457 block_count = min(block_count, max_block_count);
1461 * We intentionally ignore errors from the discard ioctl. It
1462 * is not necessary for the mkfs functionality but just an
1465 if (discard_range(fd, 0, 0) == 0) {
1466 printf("Performing full device TRIM (%s) ...\n",
1467 pretty_size(block_count));
1468 discard_blocks(fd, 0, block_count);
1472 ret = zero_dev_clamped(fd, 0, ZERO_DEV_BYTES, block_count);
1473 for (i = 0 ; !ret && i < BTRFS_SUPER_MIRROR_MAX; i++)
1474 ret = zero_dev_clamped(fd, btrfs_sb_offset(i),
1475 BTRFS_SUPER_INFO_SIZE, block_count);
1476 if (!ret && zero_end)
1477 ret = zero_dev_clamped(fd, block_count - ZERO_DEV_BYTES,
1478 ZERO_DEV_BYTES, block_count);
1481 error("failed to zero device '%s': %s", file, strerror(-ret));
1485 ret = btrfs_wipe_existing_sb(fd);
1487 error("cannot wipe superblocks on %s", file);
1491 *block_count_ret = block_count;
1495 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
1496 struct btrfs_root *root, u64 objectid)
1499 struct btrfs_inode_item inode_item;
1500 time_t now = time(NULL);
1502 memset(&inode_item, 0, sizeof(inode_item));
1503 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
1504 btrfs_set_stack_inode_size(&inode_item, 0);
1505 btrfs_set_stack_inode_nlink(&inode_item, 1);
1506 btrfs_set_stack_inode_nbytes(&inode_item, root->nodesize);
1507 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
1508 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
1509 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
1510 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
1511 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
1512 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
1513 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
1514 btrfs_set_stack_timespec_sec(&inode_item.otime, 0);
1515 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
1517 if (root->fs_info->tree_root == root)
1518 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
1520 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
1524 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
1528 btrfs_set_root_dirid(&root->root_item, objectid);
1535 * checks if a path is a block device node
1536 * Returns negative errno on failure, otherwise
1537 * returns 1 for blockdev, 0 for not-blockdev
1539 int is_block_device(const char *path)
1541 struct stat statbuf;
1543 if (stat(path, &statbuf) < 0)
1546 return !!S_ISBLK(statbuf.st_mode);
1550 * check if given path is a mount point
1551 * return 1 if yes. 0 if no. -1 for error
1553 int is_mount_point(const char *path)
1559 f = setmntent("/proc/self/mounts", "r");
1563 while ((mnt = getmntent(f)) != NULL) {
1564 if (strcmp(mnt->mnt_dir, path))
1573 static int is_reg_file(const char *path)
1575 struct stat statbuf;
1577 if (stat(path, &statbuf) < 0)
1579 return S_ISREG(statbuf.st_mode);
1583 * This function checks if the given input parameter is
1585 * return <0 : some error in the given input
1586 * return BTRFS_ARG_UNKNOWN: unknown input
1587 * return BTRFS_ARG_UUID: given input is uuid
1588 * return BTRFS_ARG_MNTPOINT: given input is path
1589 * return BTRFS_ARG_REG: given input is regular file
1590 * return BTRFS_ARG_BLKDEV: given input is block device
1592 int check_arg_type(const char *input)
1595 char path[PATH_MAX];
1600 if (realpath(input, path)) {
1601 if (is_block_device(path) == 1)
1602 return BTRFS_ARG_BLKDEV;
1604 if (is_mount_point(path) == 1)
1605 return BTRFS_ARG_MNTPOINT;
1607 if (is_reg_file(path))
1608 return BTRFS_ARG_REG;
1610 return BTRFS_ARG_UNKNOWN;
1613 if (strlen(input) == (BTRFS_UUID_UNPARSED_SIZE - 1) &&
1614 !uuid_parse(input, uuid))
1615 return BTRFS_ARG_UUID;
1617 return BTRFS_ARG_UNKNOWN;
1621 * Find the mount point for a mounted device.
1622 * On success, returns 0 with mountpoint in *mp.
1623 * On failure, returns -errno (not mounted yields -EINVAL)
1624 * Is noisy on failures, expects to be given a mounted device.
1626 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size)
1631 ret = is_block_device(dev);
1634 error("not a block device: %s", dev);
1637 error("cannot check %s: %s", dev, strerror(-ret));
1642 fd = open(dev, O_RDONLY);
1645 error("cannot open %s: %s", dev, strerror(errno));
1649 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
1652 } else { /* mounted, all good */
1662 * Given a pathname, return a filehandle to:
1663 * the original pathname or,
1664 * if the pathname is a mounted btrfs device, to its mountpoint.
1666 * On error, return -1, errno should be set.
1668 int open_path_or_dev_mnt(const char *path, DIR **dirstream, int verbose)
1673 if (is_block_device(path)) {
1674 ret = get_btrfs_mount(path, mp, sizeof(mp));
1676 /* not a mounted btrfs dev */
1677 error_on(verbose, "'%s' is not a mounted btrfs device",
1682 ret = open_file_or_dir(mp, dirstream);
1683 error_on(verbose && ret < 0, "can't access '%s': %s",
1684 path, strerror(errno));
1686 ret = btrfs_open_dir(path, dirstream, 1);
1693 * Do the following checks before calling open_file_or_dir():
1694 * 1: path is in a btrfs filesystem
1695 * 2: path is a directory
1697 int btrfs_open_dir(const char *path, DIR **dirstream, int verbose)
1703 if (statfs(path, &stfs) != 0) {
1704 error_on(verbose, "cannot access '%s': %s", path,
1709 if (stfs.f_type != BTRFS_SUPER_MAGIC) {
1710 error_on(verbose, "not a btrfs filesystem: %s", path);
1714 if (stat(path, &st) != 0) {
1715 error_on(verbose, "cannot access '%s': %s", path,
1720 if (!S_ISDIR(st.st_mode)) {
1721 error_on(verbose, "not a directory: %s", path);
1725 ret = open_file_or_dir(path, dirstream);
1727 error_on(verbose, "cannot access '%s': %s", path,
1734 /* checks if a device is a loop device */
1735 static int is_loop_device (const char* device) {
1736 struct stat statbuf;
1738 if(stat(device, &statbuf) < 0)
1741 return (S_ISBLK(statbuf.st_mode) &&
1742 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
1746 * Takes a loop device path (e.g. /dev/loop0) and returns
1747 * the associated file (e.g. /images/my_btrfs.img) using
1750 static int resolve_loop_device_with_loopdev(const char* loop_dev, char* loop_file)
1754 struct loop_info64 lo64;
1756 fd = open(loop_dev, O_RDONLY | O_NONBLOCK);
1759 ret = ioctl(fd, LOOP_GET_STATUS64, &lo64);
1765 memcpy(loop_file, lo64.lo_file_name, sizeof(lo64.lo_file_name));
1766 loop_file[sizeof(lo64.lo_file_name)] = 0;
1774 /* Takes a loop device path (e.g. /dev/loop0) and returns
1775 * the associated file (e.g. /images/my_btrfs.img) */
1776 static int resolve_loop_device(const char* loop_dev, char* loop_file,
1783 char real_loop_dev[PATH_MAX];
1785 if (!realpath(loop_dev, real_loop_dev))
1787 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
1788 if (!(f = fopen(p, "r"))) {
1789 if (errno == ENOENT)
1791 * It's possibly a partitioned loop device, which is
1792 * resolvable with loopdev API.
1794 return resolve_loop_device_with_loopdev(loop_dev, loop_file);
1798 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
1799 ret = fscanf(f, fmt, loop_file);
1808 * Checks whether a and b are identical or device
1809 * files associated with the same block device
1811 static int is_same_blk_file(const char* a, const char* b)
1813 struct stat st_buf_a, st_buf_b;
1814 char real_a[PATH_MAX];
1815 char real_b[PATH_MAX];
1817 if (!realpath(a, real_a))
1818 strncpy_null(real_a, a);
1820 if (!realpath(b, real_b))
1821 strncpy_null(real_b, b);
1823 /* Identical path? */
1824 if (strcmp(real_a, real_b) == 0)
1827 if (stat(a, &st_buf_a) < 0 || stat(b, &st_buf_b) < 0) {
1828 if (errno == ENOENT)
1833 /* Same blockdevice? */
1834 if (S_ISBLK(st_buf_a.st_mode) && S_ISBLK(st_buf_b.st_mode) &&
1835 st_buf_a.st_rdev == st_buf_b.st_rdev) {
1840 if (st_buf_a.st_dev == st_buf_b.st_dev &&
1841 st_buf_a.st_ino == st_buf_b.st_ino) {
1848 /* checks if a and b are identical or device
1849 * files associated with the same block device or
1850 * if one file is a loop device that uses the other
1853 static int is_same_loop_file(const char* a, const char* b)
1855 char res_a[PATH_MAX];
1856 char res_b[PATH_MAX];
1857 const char* final_a = NULL;
1858 const char* final_b = NULL;
1861 /* Resolve a if it is a loop device */
1862 if((ret = is_loop_device(a)) < 0) {
1867 ret = resolve_loop_device(a, res_a, sizeof(res_a));
1878 /* Resolve b if it is a loop device */
1879 if ((ret = is_loop_device(b)) < 0) {
1884 ret = resolve_loop_device(b, res_b, sizeof(res_b));
1895 return is_same_blk_file(final_a, final_b);
1898 /* Checks if a file exists and is a block or regular file*/
1899 static int is_existing_blk_or_reg_file(const char* filename)
1903 if(stat(filename, &st_buf) < 0) {
1910 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
1913 /* Checks if a file is used (directly or indirectly via a loop device)
1914 * by a device in fs_devices
1916 static int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices,
1920 struct list_head *head;
1921 struct list_head *cur;
1922 struct btrfs_device *device;
1924 head = &fs_devices->devices;
1925 list_for_each(cur, head) {
1926 device = list_entry(cur, struct btrfs_device, dev_list);
1928 if((ret = is_same_loop_file(device->name, file)))
1936 * Resolve a pathname to a device mapper node to /dev/mapper/<name>
1937 * Returns NULL on invalid input or malloc failure; Other failures
1938 * will be handled by the caller using the input pathame.
1940 char *canonicalize_dm_name(const char *ptname)
1944 char path[PATH_MAX], name[PATH_MAX], *res = NULL;
1946 if (!ptname || !*ptname)
1949 snprintf(path, sizeof(path), "/sys/block/%s/dm/name", ptname);
1950 if (!(f = fopen(path, "r")))
1953 /* read <name>\n from sysfs */
1954 if (fgets(name, sizeof(name), f) && (sz = strlen(name)) > 1) {
1955 name[sz - 1] = '\0';
1956 snprintf(path, sizeof(path), "/dev/mapper/%s", name);
1958 if (access(path, F_OK) == 0)
1966 * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or
1967 * to a device mapper pathname.
1968 * Returns NULL on invalid input or malloc failure; Other failures
1969 * will be handled by the caller using the input pathame.
1971 char *canonicalize_path(const char *path)
1973 char *canonical, *p;
1975 if (!path || !*path)
1978 canonical = realpath(path, NULL);
1980 return strdup(path);
1981 p = strrchr(canonical, '/');
1982 if (p && strncmp(p, "/dm-", 4) == 0 && isdigit(*(p + 4))) {
1983 char *dm = canonicalize_dm_name(p + 1);
1994 * returns 1 if the device was mounted, < 0 on error or 0 if everything
1995 * is safe to continue.
1997 int check_mounted(const char* file)
2002 fd = open(file, O_RDONLY);
2004 error("mount check: cannot open %s: %s", file,
2009 ret = check_mounted_where(fd, file, NULL, 0, NULL);
2015 int check_mounted_where(int fd, const char *file, char *where, int size,
2016 struct btrfs_fs_devices **fs_dev_ret)
2021 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2025 /* scan the initial device */
2026 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
2027 &total_devs, BTRFS_SUPER_INFO_OFFSET, 0);
2028 is_btrfs = (ret >= 0);
2030 /* scan other devices */
2031 if (is_btrfs && total_devs > 1) {
2032 ret = btrfs_scan_lblkid();
2037 /* iterate over the list of currently mounted filesystems */
2038 if ((f = setmntent ("/proc/self/mounts", "r")) == NULL)
2041 while ((mnt = getmntent (f)) != NULL) {
2043 if(strcmp(mnt->mnt_type, "btrfs") != 0)
2046 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
2048 /* ignore entries in the mount table that are not
2049 associated with a file*/
2050 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
2051 goto out_mntloop_err;
2055 ret = is_same_loop_file(file, mnt->mnt_fsname);
2059 goto out_mntloop_err;
2064 /* Did we find an entry in mnt table? */
2065 if (mnt && size && where) {
2066 strncpy(where, mnt->mnt_dir, size);
2070 *fs_dev_ret = fs_devices_mnt;
2072 ret = (mnt != NULL);
2080 struct pending_dir {
2081 struct list_head list;
2082 char name[PATH_MAX];
2085 int btrfs_register_one_device(const char *fname)
2087 struct btrfs_ioctl_vol_args args;
2091 fd = open("/dev/btrfs-control", O_RDWR);
2094 "failed to open /dev/btrfs-control, skipping device registration: %s",
2098 memset(&args, 0, sizeof(args));
2099 strncpy_null(args.name, fname);
2100 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
2102 error("device scan failed on '%s': %s", fname,
2111 * Register all devices in the fs_uuid list created in the user
2112 * space. Ensure btrfs_scan_lblkid() is called before this func.
2114 int btrfs_register_all_devices(void)
2118 struct btrfs_fs_devices *fs_devices;
2119 struct btrfs_device *device;
2120 struct list_head *all_uuids;
2122 all_uuids = btrfs_scanned_uuids();
2124 list_for_each_entry(fs_devices, all_uuids, list) {
2125 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2127 err = btrfs_register_one_device(device->name);
2137 int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
2140 struct btrfs_super_block *disk_super;
2144 buf = malloc(BTRFS_SUPER_INFO_SIZE);
2149 ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
2150 if (ret != BTRFS_SUPER_INFO_SIZE)
2154 disk_super = (struct btrfs_super_block *)buf;
2155 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC)
2158 if (!memcmp(disk_super->fsid, root->fs_info->super_copy->fsid,
2168 * Note: this function uses a static per-thread buffer. Do not call this
2169 * function more than 10 times within one argument list!
2171 const char *pretty_size_mode(u64 size, unsigned mode)
2173 static __thread int ps_index = 0;
2174 static __thread char ps_array[10][32];
2177 ret = ps_array[ps_index];
2180 (void)pretty_size_snprintf(size, ret, 32, mode);
2185 static const char* unit_suffix_binary[] =
2186 { "B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB"};
2187 static const char* unit_suffix_decimal[] =
2188 { "B", "kB", "MB", "GB", "TB", "PB", "EB"};
2190 int pretty_size_snprintf(u64 size, char *str, size_t str_size, unsigned unit_mode)
2196 const char** suffix = NULL;
2202 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_RAW) {
2203 snprintf(str, str_size, "%llu", size);
2207 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_BINARY) {
2210 suffix = unit_suffix_binary;
2211 } else if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_DECIMAL) {
2214 suffix = unit_suffix_decimal;
2219 fprintf(stderr, "INTERNAL ERROR: unknown unit base, mode %d\n",
2227 switch (unit_mode & UNITS_MODE_MASK) {
2228 case UNITS_TBYTES: base *= mult; num_divs++;
2229 case UNITS_GBYTES: base *= mult; num_divs++;
2230 case UNITS_MBYTES: base *= mult; num_divs++;
2231 case UNITS_KBYTES: num_divs++;
2238 while (size >= mult) {
2244 * If the value is smaller than base, we didn't do any
2245 * division, in that case, base should be 1, not original
2246 * base, or the unit will be wrong
2252 if (num_divs >= ARRAY_SIZE(unit_suffix_binary)) {
2254 printf("INTERNAL ERROR: unsupported unit suffix, index %d\n",
2259 fraction = (float)last_size / base;
2261 return snprintf(str, str_size, "%.2f%s", fraction, suffix[num_divs]);
2265 * __strncpy_null - strncpy with null termination
2266 * @dest: the target array
2267 * @src: the source string
2268 * @n: maximum bytes to copy (size of *dest)
2270 * Like strncpy, but ensures destination is null-terminated.
2272 * Copies the string pointed to by src, including the terminating null
2273 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
2274 * of n bytes. Then ensure that dest is null-terminated.
2276 char *__strncpy_null(char *dest, const char *src, size_t n)
2278 strncpy(dest, src, n);
2285 * Checks to make sure that the label matches our requirements.
2287 0 if everything is safe and usable
2288 -1 if the label is too long
2290 static int check_label(const char *input)
2292 int len = strlen(input);
2294 if (len > BTRFS_LABEL_SIZE - 1) {
2295 error("label %s is too long (max %d)", input,
2296 BTRFS_LABEL_SIZE - 1);
2303 static int set_label_unmounted(const char *dev, const char *label)
2305 struct btrfs_trans_handle *trans;
2306 struct btrfs_root *root;
2309 ret = check_mounted(dev);
2311 error("checking mount status of %s failed: %d", dev, ret);
2315 error("device %s is mounted, use mount point", dev);
2319 /* Open the super_block at the default location
2320 * and as read-write.
2322 root = open_ctree(dev, 0, OPEN_CTREE_WRITES);
2323 if (!root) /* errors are printed by open_ctree() */
2326 trans = btrfs_start_transaction(root, 1);
2327 __strncpy_null(root->fs_info->super_copy->label, label, BTRFS_LABEL_SIZE - 1);
2329 btrfs_commit_transaction(trans, root);
2331 /* Now we close it since we are done. */
2336 static int set_label_mounted(const char *mount_path, const char *labelp)
2339 char label[BTRFS_LABEL_SIZE];
2341 fd = open(mount_path, O_RDONLY | O_NOATIME);
2343 error("unable to access %s: %s", mount_path, strerror(errno));
2347 memset(label, 0, sizeof(label));
2348 __strncpy_null(label, labelp, BTRFS_LABEL_SIZE - 1);
2349 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
2350 error("unable to set label of %s: %s", mount_path,
2360 int get_label_unmounted(const char *dev, char *label)
2362 struct btrfs_root *root;
2365 ret = check_mounted(dev);
2367 error("checking mount status of %s failed: %d", dev, ret);
2371 /* Open the super_block at the default location
2374 root = open_ctree(dev, 0, 0);
2378 __strncpy_null(label, root->fs_info->super_copy->label,
2379 BTRFS_LABEL_SIZE - 1);
2381 /* Now we close it since we are done. */
2387 * If a partition is mounted, try to get the filesystem label via its
2388 * mounted path rather than device. Return the corresponding error
2389 * the user specified the device path.
2391 int get_label_mounted(const char *mount_path, char *labelp)
2393 char label[BTRFS_LABEL_SIZE];
2397 fd = open(mount_path, O_RDONLY | O_NOATIME);
2399 error("unable to access %s: %s", mount_path, strerror(errno));
2403 memset(label, '\0', sizeof(label));
2404 ret = ioctl(fd, BTRFS_IOC_GET_FSLABEL, label);
2406 if (errno != ENOTTY)
2407 error("unable to get label of %s: %s", mount_path,
2414 __strncpy_null(labelp, label, BTRFS_LABEL_SIZE - 1);
2419 int get_label(const char *btrfs_dev, char *label)
2423 ret = is_existing_blk_or_reg_file(btrfs_dev);
2425 ret = get_label_mounted(btrfs_dev, label);
2427 ret = get_label_unmounted(btrfs_dev, label);
2432 int set_label(const char *btrfs_dev, const char *label)
2436 if (check_label(label))
2439 ret = is_existing_blk_or_reg_file(btrfs_dev);
2441 ret = set_label_mounted(btrfs_dev, label);
2443 ret = set_label_unmounted(btrfs_dev, label);
2449 * A not-so-good version fls64. No fascinating optimization since
2450 * no one except parse_size use it
2452 static int fls64(u64 x)
2456 for (i = 0; i <64; i++)
2457 if (x << i & (1ULL << 63))
2462 u64 parse_size(char *s)
2470 error("size value is empty");
2474 error("size value '%s' is less equal than 0", s);
2477 ret = strtoull(s, &endptr, 10);
2479 error("size value '%s' is invalid", s);
2482 if (endptr[0] && endptr[1]) {
2483 error("illegal suffix contains character '%c' in wrong position",
2488 * strtoll returns LLONG_MAX when overflow, if this happens,
2489 * need to call strtoull to get the real size
2491 if (errno == ERANGE && ret == ULLONG_MAX) {
2492 error("size value '%s' is too large for u64", s);
2496 c = tolower(endptr[0]);
2519 error("unknown size descriptor '%c'", c);
2523 /* Check whether ret * mult overflow */
2524 if (fls64(ret) + fls64(mult) - 1 > 64) {
2525 error("size value '%s' is too large for u64", s);
2532 u64 parse_qgroupid(const char *p)
2534 char *s = strchr(p, '/');
2535 const char *ptr_src_end = p + strlen(p);
2536 char *ptr_parse_end = NULL;
2545 /* Numeric format like '0/257' is the primary case */
2547 id = strtoull(p, &ptr_parse_end, 10);
2548 if (ptr_parse_end != ptr_src_end)
2552 level = strtoull(p, &ptr_parse_end, 10);
2553 if (ptr_parse_end != s)
2556 id = strtoull(s + 1, &ptr_parse_end, 10);
2557 if (ptr_parse_end != ptr_src_end)
2560 return (level << BTRFS_QGROUP_LEVEL_SHIFT) | id;
2563 /* Path format like subv at 'my_subvol' is the fallback case */
2564 ret = test_issubvolume(p);
2565 if (ret < 0 || !ret)
2567 fd = open(p, O_RDONLY);
2570 ret = lookup_ino_rootid(fd, &id);
2577 error("invalid qgroupid or subvolume path: %s", p);
2581 int open_file_or_dir3(const char *fname, DIR **dirstream, int open_flags)
2587 ret = stat(fname, &st);
2591 if (S_ISDIR(st.st_mode)) {
2592 *dirstream = opendir(fname);
2595 fd = dirfd(*dirstream);
2596 } else if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) {
2597 fd = open(fname, open_flags);
2600 * we set this on purpose, in case the caller output
2601 * strerror(errno) as success
2609 closedir(*dirstream);
2616 int open_file_or_dir(const char *fname, DIR **dirstream)
2618 return open_file_or_dir3(fname, dirstream, O_RDWR);
2621 void close_file_or_dir(int fd, DIR *dirstream)
2624 closedir(dirstream);
2629 int get_device_info(int fd, u64 devid,
2630 struct btrfs_ioctl_dev_info_args *di_args)
2634 di_args->devid = devid;
2635 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
2637 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
2638 return ret < 0 ? -errno : 0;
2641 static u64 find_max_device_id(struct btrfs_ioctl_search_args *search_args,
2644 struct btrfs_dev_item *dev_item;
2645 char *buf = search_args->buf;
2647 buf += (nr_items - 1) * (sizeof(struct btrfs_ioctl_search_header)
2648 + sizeof(struct btrfs_dev_item));
2649 buf += sizeof(struct btrfs_ioctl_search_header);
2651 dev_item = (struct btrfs_dev_item *)buf;
2653 return btrfs_stack_device_id(dev_item);
2656 static int search_chunk_tree_for_fs_info(int fd,
2657 struct btrfs_ioctl_fs_info_args *fi_args)
2661 u64 start_devid = 1;
2662 struct btrfs_ioctl_search_args search_args;
2663 struct btrfs_ioctl_search_key *search_key = &search_args.key;
2665 fi_args->num_devices = 0;
2667 max_items = BTRFS_SEARCH_ARGS_BUFSIZE
2668 / (sizeof(struct btrfs_ioctl_search_header)
2669 + sizeof(struct btrfs_dev_item));
2671 search_key->tree_id = BTRFS_CHUNK_TREE_OBJECTID;
2672 search_key->min_objectid = BTRFS_DEV_ITEMS_OBJECTID;
2673 search_key->max_objectid = BTRFS_DEV_ITEMS_OBJECTID;
2674 search_key->min_type = BTRFS_DEV_ITEM_KEY;
2675 search_key->max_type = BTRFS_DEV_ITEM_KEY;
2676 search_key->min_transid = 0;
2677 search_key->max_transid = (u64)-1;
2678 search_key->nr_items = max_items;
2679 search_key->max_offset = (u64)-1;
2682 search_key->min_offset = start_devid;
2684 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH, &search_args);
2688 fi_args->num_devices += (u64)search_key->nr_items;
2690 if (search_key->nr_items == max_items) {
2691 start_devid = find_max_device_id(&search_args,
2692 search_key->nr_items) + 1;
2696 /* get the lastest max_id to stay consistent with the num_devices */
2697 if (search_key->nr_items == 0)
2699 * last tree_search returns an empty buf, use the devid of
2700 * the last dev_item of the previous tree_search
2702 fi_args->max_id = start_devid - 1;
2704 fi_args->max_id = find_max_device_id(&search_args,
2705 search_key->nr_items);
2711 * For a given path, fill in the ioctl fs_ and info_ args.
2712 * If the path is a btrfs mountpoint, fill info for all devices.
2713 * If the path is a btrfs device, fill in only that device.
2715 * The path provided must be either on a mounted btrfs fs,
2716 * or be a mounted btrfs device.
2718 * Returns 0 on success, or a negative errno.
2720 int get_fs_info(char *path, struct btrfs_ioctl_fs_info_args *fi_args,
2721 struct btrfs_ioctl_dev_info_args **di_ret)
2728 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2729 struct btrfs_ioctl_dev_info_args *di_args;
2730 struct btrfs_ioctl_dev_info_args tmp;
2732 DIR *dirstream = NULL;
2734 memset(fi_args, 0, sizeof(*fi_args));
2736 if (is_block_device(path) == 1) {
2737 struct btrfs_super_block *disk_super;
2738 char buf[BTRFS_SUPER_INFO_SIZE];
2741 /* Ensure it's mounted, then set path to the mountpoint */
2742 fd = open(path, O_RDONLY);
2745 error("cannot open %s: %s", path, strerror(errno));
2748 ret = check_mounted_where(fd, path, mp, sizeof(mp),
2757 /* Only fill in this one device */
2758 fi_args->num_devices = 1;
2760 disk_super = (struct btrfs_super_block *)buf;
2761 ret = btrfs_read_dev_super(fd, disk_super,
2762 BTRFS_SUPER_INFO_OFFSET, 0);
2767 devid = btrfs_stack_device_id(&disk_super->dev_item);
2769 fi_args->max_id = devid;
2772 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
2776 /* at this point path must not be for a block device */
2777 fd = open_file_or_dir(path, &dirstream);
2783 /* fill in fi_args if not just a single device */
2784 if (fi_args->num_devices != 1) {
2785 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
2792 * The fs_args->num_devices does not include seed devices
2794 ret = search_chunk_tree_for_fs_info(fd, fi_args);
2799 * search_chunk_tree_for_fs_info() will lacks the devid 0
2800 * so manual probe for it here.
2802 ret = get_device_info(fd, 0, &tmp);
2804 fi_args->num_devices++;
2812 if (!fi_args->num_devices)
2815 di_args = *di_ret = malloc((fi_args->num_devices) * sizeof(*di_args));
2822 memcpy(di_args, &tmp, sizeof(tmp));
2823 for (; i <= fi_args->max_id; ++i) {
2824 ret = get_device_info(fd, i, &di_args[ndevs]);
2833 * only when the only dev we wanted to find is not there then
2834 * let any error be returned
2836 if (fi_args->num_devices != 1) {
2842 close_file_or_dir(fd, dirstream);
2846 #define isoctal(c) (((c) & ~7) == '0')
2848 static inline void translate(char *f, char *t)
2850 while (*f != '\0') {
2852 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
2853 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
2863 * Checks if the swap device.
2864 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
2866 static int is_swap_device(const char *file)
2877 if (stat(file, &st_buf) < 0)
2879 if (S_ISBLK(st_buf.st_mode))
2880 dev = st_buf.st_rdev;
2881 else if (S_ISREG(st_buf.st_mode)) {
2882 dev = st_buf.st_dev;
2883 ino = st_buf.st_ino;
2887 if ((f = fopen("/proc/swaps", "r")) == NULL)
2890 /* skip the first line */
2891 if (fgets(tmp, sizeof(tmp), f) == NULL)
2894 while (fgets(tmp, sizeof(tmp), f) != NULL) {
2895 if ((cp = strchr(tmp, ' ')) != NULL)
2897 if ((cp = strchr(tmp, '\t')) != NULL)
2899 translate(tmp, buf);
2900 if (stat(buf, &st_buf) != 0)
2902 if (S_ISBLK(st_buf.st_mode)) {
2903 if (dev == st_buf.st_rdev) {
2907 } else if (S_ISREG(st_buf.st_mode)) {
2908 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {
2922 * Check for existing filesystem or partition table on device.
2924 * 1 for existing fs or partition
2925 * 0 for nothing found
2926 * -1 for internal error
2928 static int check_overwrite(const char *device)
2931 blkid_probe pr = NULL;
2935 if (!device || !*device)
2938 ret = -1; /* will reset on success of all setup calls */
2940 pr = blkid_new_probe_from_filename(device);
2944 size = blkid_probe_get_size(pr);
2948 /* nothing to overwrite on a 0-length device */
2954 ret = blkid_probe_enable_partitions(pr, 1);
2958 ret = blkid_do_fullprobe(pr);
2963 * Blkid returns 1 for nothing found and 0 when it finds a signature,
2964 * but we want the exact opposite, so reverse the return value here.
2966 * In addition print some useful diagnostics about what actually is
2974 if (!blkid_probe_lookup_value(pr, "TYPE", &type, NULL)) {
2976 "%s appears to contain an existing "
2977 "filesystem (%s).\n", device, type);
2978 } else if (!blkid_probe_lookup_value(pr, "PTTYPE", &type, NULL)) {
2980 "%s appears to contain a partition "
2981 "table (%s).\n", device, type);
2984 "%s appears to contain something weird "
2985 "according to blkid\n", device);
2991 blkid_free_probe(pr);
2994 "probe of %s failed, cannot detect "
2995 "existing filesystem.\n", device);
2999 static int group_profile_devs_min(u64 flag)
3001 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3002 case 0: /* single */
3003 case BTRFS_BLOCK_GROUP_DUP:
3005 case BTRFS_BLOCK_GROUP_RAID0:
3006 case BTRFS_BLOCK_GROUP_RAID1:
3007 case BTRFS_BLOCK_GROUP_RAID5:
3009 case BTRFS_BLOCK_GROUP_RAID6:
3011 case BTRFS_BLOCK_GROUP_RAID10:
3018 int test_num_disk_vs_raid(u64 metadata_profile, u64 data_profile,
3019 u64 dev_cnt, int mixed, int ssd)
3026 allowed |= BTRFS_BLOCK_GROUP_RAID10;
3028 allowed |= BTRFS_BLOCK_GROUP_RAID6;
3030 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
3031 BTRFS_BLOCK_GROUP_RAID5;
3033 allowed |= BTRFS_BLOCK_GROUP_DUP;
3037 ((metadata_profile | data_profile) & BTRFS_BLOCK_GROUP_DUP)) {
3038 warning("DUP is not recommended on filesystem with multiple devices");
3040 if (metadata_profile & ~allowed) {
3042 "ERROR: unable to create FS with metadata profile %s "
3043 "(have %llu devices but %d devices are required)\n",
3044 btrfs_group_profile_str(metadata_profile), dev_cnt,
3045 group_profile_devs_min(metadata_profile));
3048 if (data_profile & ~allowed) {
3050 "ERROR: unable to create FS with data profile %s "
3051 "(have %llu devices but %d devices are required)\n",
3052 btrfs_group_profile_str(data_profile), dev_cnt,
3053 group_profile_devs_min(data_profile));
3057 warning_on(!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP) && ssd,
3058 "DUP may not actually lead to 2 copies on the device, see manual page");
3063 int group_profile_max_safe_loss(u64 flags)
3065 switch (flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3066 case 0: /* single */
3067 case BTRFS_BLOCK_GROUP_DUP:
3068 case BTRFS_BLOCK_GROUP_RAID0:
3070 case BTRFS_BLOCK_GROUP_RAID1:
3071 case BTRFS_BLOCK_GROUP_RAID5:
3072 case BTRFS_BLOCK_GROUP_RAID10:
3074 case BTRFS_BLOCK_GROUP_RAID6:
3082 * Check if a device is suitable for btrfs
3084 * 1: something is wrong, an error is printed
3087 int test_dev_for_mkfs(const char *file, int force_overwrite)
3092 ret = is_swap_device(file);
3094 error("checking status of %s: %s", file, strerror(-ret));
3098 error("%s is a swap device", file);
3101 if (!force_overwrite) {
3102 if (check_overwrite(file)) {
3103 error("use the -f option to force overwrite of %s",
3108 ret = check_mounted(file);
3110 error("cannot check mount status of %s: %s", file,
3115 error("%s is mounted", file);
3118 /* check if the device is busy */
3119 fd = open(file, O_RDWR|O_EXCL);
3121 error("unable to open %s: %s", file, strerror(errno));
3124 if (fstat(fd, &st)) {
3125 error("unable to stat %s: %s", file, strerror(errno));
3129 if (!S_ISBLK(st.st_mode)) {
3130 error("%s is not a block device", file);
3138 int btrfs_scan_lblkid(void)
3143 struct btrfs_fs_devices *tmp_devices;
3144 blkid_dev_iterate iter = NULL;
3145 blkid_dev dev = NULL;
3146 blkid_cache cache = NULL;
3147 char path[PATH_MAX];
3149 if (btrfs_scan_done)
3152 if (blkid_get_cache(&cache, NULL) < 0) {
3153 error("blkid cache get failed");
3156 blkid_probe_all(cache);
3157 iter = blkid_dev_iterate_begin(cache);
3158 blkid_dev_set_search(iter, "TYPE", "btrfs");
3159 while (blkid_dev_next(iter, &dev) == 0) {
3160 dev = blkid_verify(cache, dev);
3163 /* if we are here its definitely a btrfs disk*/
3164 strncpy_null(path, blkid_dev_devname(dev));
3166 fd = open(path, O_RDONLY);
3168 error("cannot open %s: %s", path, strerror(errno));
3171 ret = btrfs_scan_one_device(fd, path, &tmp_devices,
3172 &num_devices, BTRFS_SUPER_INFO_OFFSET, 0);
3174 error("cannot scan %s: %s", path, strerror(-ret));
3181 blkid_dev_iterate_end(iter);
3182 blkid_put_cache(cache);
3184 btrfs_scan_done = 1;
3189 int is_vol_small(const char *file)
3196 fd = open(file, O_RDONLY);
3199 if (fstat(fd, &st) < 0) {
3204 size = btrfs_device_size(fd, &st);
3209 if (size < BTRFS_MKFS_SMALL_VOLUME_SIZE) {
3219 * This reads a line from the stdin and only returns non-zero if the
3220 * first whitespace delimited token is a case insensitive match with yes
3223 int ask_user(const char *question)
3225 char buf[30] = {0,};
3226 char *saveptr = NULL;
3229 printf("%s [y/N]: ", question);
3231 return fgets(buf, sizeof(buf) - 1, stdin) &&
3232 (answer = strtok_r(buf, " \t\n\r", &saveptr)) &&
3233 (!strcasecmp(answer, "yes") || !strcasecmp(answer, "y"));
3238 * - file or directory return the containing tree root id
3239 * - subvolume return its own tree id
3240 * - BTRFS_EMPTY_SUBVOL_DIR_OBJECTID (directory with ino == 2) the result is
3241 * undefined and function returns -1
3243 int lookup_ino_rootid(int fd, u64 *rootid)
3245 struct btrfs_ioctl_ino_lookup_args args;
3248 memset(&args, 0, sizeof(args));
3250 args.objectid = BTRFS_FIRST_FREE_OBJECTID;
3252 ret = ioctl(fd, BTRFS_IOC_INO_LOOKUP, &args);
3254 error("failed to lookup root id: %s", strerror(errno));
3258 *rootid = args.treeid;
3264 * return 0 if a btrfs mount point is found
3265 * return 1 if a mount point is found but not btrfs
3266 * return <0 if something goes wrong
3268 int find_mount_root(const char *path, char **mount_root)
3276 int longest_matchlen = 0;
3277 char *longest_match = NULL;
3279 fd = open(path, O_RDONLY | O_NOATIME);
3284 mnttab = setmntent("/proc/self/mounts", "r");
3288 while ((ent = getmntent(mnttab))) {
3289 len = strlen(ent->mnt_dir);
3290 if (strncmp(ent->mnt_dir, path, len) == 0) {
3291 /* match found and use the latest match */
3292 if (longest_matchlen <= len) {
3293 free(longest_match);
3294 longest_matchlen = len;
3295 longest_match = strdup(ent->mnt_dir);
3296 not_btrfs = strcmp(ent->mnt_type, "btrfs");
3305 free(longest_match);
3310 *mount_root = realpath(longest_match, NULL);
3314 free(longest_match);
3318 int test_minimum_size(const char *file, u32 nodesize)
3321 struct stat statbuf;
3323 fd = open(file, O_RDONLY);
3326 if (stat(file, &statbuf) < 0) {
3330 if (btrfs_device_size(fd, &statbuf) < btrfs_min_dev_size(nodesize)) {
3340 * Test if path is a directory
3342 * 0 - path exists but it is not a directory
3343 * 1 - path exists and it is a directory
3346 int test_isdir(const char *path)
3351 ret = stat(path, &st);
3355 return !!S_ISDIR(st.st_mode);
3358 void units_set_mode(unsigned *units, unsigned mode)
3360 unsigned base = *units & UNITS_MODE_MASK;
3362 *units = base | mode;
3365 void units_set_base(unsigned *units, unsigned base)
3367 unsigned mode = *units & ~UNITS_MODE_MASK;
3369 *units = base | mode;
3372 int find_next_key(struct btrfs_path *path, struct btrfs_key *key)
3376 for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
3377 if (!path->nodes[level])
3379 if (path->slots[level] + 1 >=
3380 btrfs_header_nritems(path->nodes[level]))
3383 btrfs_item_key_to_cpu(path->nodes[level], key,
3384 path->slots[level] + 1);
3386 btrfs_node_key_to_cpu(path->nodes[level], key,
3387 path->slots[level] + 1);
3393 const char* btrfs_group_type_str(u64 flag)
3395 u64 mask = BTRFS_BLOCK_GROUP_TYPE_MASK |
3396 BTRFS_SPACE_INFO_GLOBAL_RSV;
3398 switch (flag & mask) {
3399 case BTRFS_BLOCK_GROUP_DATA:
3401 case BTRFS_BLOCK_GROUP_SYSTEM:
3403 case BTRFS_BLOCK_GROUP_METADATA:
3405 case BTRFS_BLOCK_GROUP_DATA|BTRFS_BLOCK_GROUP_METADATA:
3406 return "Data+Metadata";
3407 case BTRFS_SPACE_INFO_GLOBAL_RSV:
3408 return "GlobalReserve";
3414 const char* btrfs_group_profile_str(u64 flag)
3416 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3419 case BTRFS_BLOCK_GROUP_RAID0:
3421 case BTRFS_BLOCK_GROUP_RAID1:
3423 case BTRFS_BLOCK_GROUP_RAID5:
3425 case BTRFS_BLOCK_GROUP_RAID6:
3427 case BTRFS_BLOCK_GROUP_DUP:
3429 case BTRFS_BLOCK_GROUP_RAID10:
3436 u64 disk_size(const char *path)
3440 if (statfs(path, &sfs) < 0)
3443 return sfs.f_bsize * sfs.f_blocks;
3446 u64 get_partition_size(const char *dev)
3449 int fd = open(dev, O_RDONLY);
3453 if (ioctl(fd, BLKGETSIZE64, &result) < 0) {
3462 int btrfs_tree_search2_ioctl_supported(int fd)
3464 struct btrfs_ioctl_search_args_v2 *args2;
3465 struct btrfs_ioctl_search_key *sk;
3466 int args2_size = 1024;
3467 char args2_buf[args2_size];
3469 static int v2_supported = -1;
3471 if (v2_supported != -1)
3472 return v2_supported;
3474 args2 = (struct btrfs_ioctl_search_args_v2 *)args2_buf;
3478 * Search for the extent tree item in the root tree.
3480 sk->tree_id = BTRFS_ROOT_TREE_OBJECTID;
3481 sk->min_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3482 sk->max_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3483 sk->min_type = BTRFS_ROOT_ITEM_KEY;
3484 sk->max_type = BTRFS_ROOT_ITEM_KEY;
3486 sk->max_offset = (u64)-1;
3487 sk->min_transid = 0;
3488 sk->max_transid = (u64)-1;
3490 args2->buf_size = args2_size - sizeof(struct btrfs_ioctl_search_args_v2);
3491 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH_V2, args2);
3492 if (ret == -EOPNOTSUPP)
3499 return v2_supported;
3502 int btrfs_check_nodesize(u32 nodesize, u32 sectorsize, u64 features)
3504 if (nodesize < sectorsize) {
3505 error("illegal nodesize %u (smaller than %u)",
3506 nodesize, sectorsize);
3508 } else if (nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) {
3509 error("illegal nodesize %u (larger than %u)",
3510 nodesize, BTRFS_MAX_METADATA_BLOCKSIZE);
3512 } else if (nodesize & (sectorsize - 1)) {
3513 error("illegal nodesize %u (not aligned to %u)",
3514 nodesize, sectorsize);
3516 } else if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS &&
3517 nodesize != sectorsize) {
3518 error("illegal nodesize %u (not equal to %u for mixed block group)",
3519 nodesize, sectorsize);
3526 * Copy a path argument from SRC to DEST and check the SRC length if it's at
3527 * most PATH_MAX and fits into DEST. DESTLEN is supposed to be exact size of
3529 * The destination buffer is zero terminated.
3530 * Return < 0 for error, 0 otherwise.
3532 int arg_copy_path(char *dest, const char *src, int destlen)
3534 size_t len = strlen(src);
3536 if (len >= PATH_MAX || len >= destlen)
3537 return -ENAMETOOLONG;
3539 __strncpy_null(dest, src, destlen);
3544 unsigned int get_unit_mode_from_arg(int *argc, char *argv[], int df_mode)
3546 unsigned int unit_mode = UNITS_DEFAULT;
3550 for (arg_i = 0; arg_i < *argc; arg_i++) {
3551 if (!strcmp(argv[arg_i], "--"))
3554 if (!strcmp(argv[arg_i], "--raw")) {
3555 unit_mode = UNITS_RAW;
3559 if (!strcmp(argv[arg_i], "--human-readable")) {
3560 unit_mode = UNITS_HUMAN_BINARY;
3565 if (!strcmp(argv[arg_i], "--iec")) {
3566 units_set_mode(&unit_mode, UNITS_BINARY);
3570 if (!strcmp(argv[arg_i], "--si")) {
3571 units_set_mode(&unit_mode, UNITS_DECIMAL);
3576 if (!strcmp(argv[arg_i], "--kbytes")) {
3577 units_set_base(&unit_mode, UNITS_KBYTES);
3581 if (!strcmp(argv[arg_i], "--mbytes")) {
3582 units_set_base(&unit_mode, UNITS_MBYTES);
3586 if (!strcmp(argv[arg_i], "--gbytes")) {
3587 units_set_base(&unit_mode, UNITS_GBYTES);
3591 if (!strcmp(argv[arg_i], "--tbytes")) {
3592 units_set_base(&unit_mode, UNITS_TBYTES);
3600 if (!strcmp(argv[arg_i], "-b")) {
3601 unit_mode = UNITS_RAW;
3605 if (!strcmp(argv[arg_i], "-h")) {
3606 unit_mode = UNITS_HUMAN_BINARY;
3610 if (!strcmp(argv[arg_i], "-H")) {
3611 unit_mode = UNITS_HUMAN_DECIMAL;
3615 if (!strcmp(argv[arg_i], "-k")) {
3616 units_set_base(&unit_mode, UNITS_KBYTES);
3620 if (!strcmp(argv[arg_i], "-m")) {
3621 units_set_base(&unit_mode, UNITS_MBYTES);
3625 if (!strcmp(argv[arg_i], "-g")) {
3626 units_set_base(&unit_mode, UNITS_GBYTES);
3630 if (!strcmp(argv[arg_i], "-t")) {
3631 units_set_base(&unit_mode, UNITS_TBYTES);
3637 for (arg_i = 0, arg_end = 0; arg_i < *argc; arg_i++) {
3640 argv[arg_end] = argv[arg_i];
3649 int string_is_numerical(const char *str)
3651 if (!(*str >= '0' && *str <= '9'))
3653 while (*str >= '0' && *str <= '9')
3661 * Preprocess @argv with getopt_long to reorder options and consume the "--"
3663 * Unknown short and long options are reported, optionally the @usage is printed
3666 void clean_args_no_options(int argc, char *argv[], const char * const *usagestr)
3668 static const struct option long_options[] = {
3673 int c = getopt_long(argc, argv, "", long_options, NULL);
3686 /* Subvolume helper functions */
3688 * test if name is a correct subvolume name
3689 * this function return
3690 * 0-> name is not a correct subvolume name
3691 * 1-> name is a correct subvolume name
3693 int test_issubvolname(const char *name)
3695 return name[0] != '\0' && !strchr(name, '/') &&
3696 strcmp(name, ".") && strcmp(name, "..");
3700 * Test if path is a subvolume
3702 * 0 - path exists but it is not a subvolume
3703 * 1 - path exists and it is a subvolume
3706 int test_issubvolume(const char *path)
3712 res = stat(path, &st);
3716 if (st.st_ino != BTRFS_FIRST_FREE_OBJECTID || !S_ISDIR(st.st_mode))
3719 res = statfs(path, &stfs);
3723 return (int)stfs.f_type == BTRFS_SUPER_MAGIC;
3726 const char *subvol_strip_mountpoint(const char *mnt, const char *full_path)
3728 int len = strlen(mnt);
3732 if (mnt[len - 1] != '/')
3735 return full_path + len;
3742 * 1: Error; and error info printed to the terminal. Fixme.
3743 * 2: If the fullpath is root tree instead of subvol tree
3745 int get_subvol_info(const char *fullpath, struct root_info *get_ri)
3752 const char *svpath = NULL;
3753 DIR *dirstream1 = NULL;
3754 DIR *dirstream2 = NULL;
3756 ret = test_issubvolume(fullpath);
3760 error("not a subvolume: %s", fullpath);
3764 ret = find_mount_root(fullpath, &mnt);
3768 error("%s doesn't belong to btrfs mount point", fullpath);
3772 svpath = subvol_strip_mountpoint(mnt, fullpath);
3774 fd = btrfs_open_dir(fullpath, &dirstream1, 1);
3778 ret = btrfs_list_get_path_rootid(fd, &sv_id);
3780 error("can't get rootid for '%s'", fullpath);
3784 mntfd = btrfs_open_dir(mnt, &dirstream2, 1);
3788 if (sv_id == BTRFS_FS_TREE_OBJECTID) {
3791 * So that caller may decide if thats an error or just fine.
3796 memset(get_ri, 0, sizeof(*get_ri));
3797 get_ri->root_id = sv_id;
3799 ret = btrfs_get_subvol(mntfd, get_ri);
3801 error("can't find '%s': %d", svpath, ret);
3804 close_file_or_dir(mntfd, dirstream2);
3805 close_file_or_dir(fd, dirstream1);
3811 void init_rand_seed(u64 seed)
3815 /* only use the last 48 bits */
3816 for (i = 0; i < 3; i++) {
3817 rand_seed[i] = (unsigned short)(seed ^ (unsigned short)(-1));
3820 rand_seed_initlized = 1;
3823 static void __init_seed(void)
3829 if(rand_seed_initlized)
3831 /* Use urandom as primary seed source. */
3832 fd = open("/dev/urandom", O_RDONLY);
3834 ret = read(fd, rand_seed, sizeof(rand_seed));
3836 if (ret < sizeof(rand_seed))
3840 /* Use time and pid as fallback seed */
3841 warning("failed to read /dev/urandom, use time and pid as random seed");
3842 gettimeofday(&tv, 0);
3843 rand_seed[0] = getpid() ^ (tv.tv_sec & 0xFFFF);
3844 rand_seed[1] = getppid() ^ (tv.tv_usec & 0xFFFF);
3845 rand_seed[2] = (tv.tv_sec ^ tv.tv_usec) >> 16;
3847 rand_seed_initlized = 1;
3854 * Don't use nrand48, its range is [0,2^31) The highest bit will alwasy
3855 * be 0. Use jrand48 to include the highest bit.
3857 return (u32)jrand48(rand_seed);
3860 unsigned int rand_range(unsigned int upper)
3864 * Use the full 48bits to mod, which would be more uniformly
3867 return (unsigned int)(jrand48(rand_seed) % upper);