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 * Insert one temporary extent item.
706 * NOTE: if skinny_metadata is not enabled, this function must be called
707 * after all other trees are initialized.
708 * Or fs without skinny-metadata will be screwed up.
710 static int insert_temp_extent_item(int fd, struct extent_buffer *buf,
711 struct btrfs_mkfs_config *cfg,
712 int *slot, u32 *itemoff, u64 bytenr,
715 struct extent_buffer *tmp;
716 struct btrfs_extent_item *ei;
717 struct btrfs_extent_inline_ref *iref;
718 struct btrfs_disk_key disk_key;
719 struct btrfs_disk_key tree_info_key;
720 struct btrfs_tree_block_info *info;
722 int skinny_metadata = cfg->features &
723 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA;
727 itemsize = sizeof(*ei) + sizeof(*iref);
729 itemsize = sizeof(*ei) + sizeof(*iref) +
730 sizeof(struct btrfs_tree_block_info);
732 btrfs_set_header_nritems(buf, *slot + 1);
733 *(itemoff) -= itemsize;
735 if (skinny_metadata) {
736 btrfs_set_disk_key_type(&disk_key, BTRFS_METADATA_ITEM_KEY);
737 btrfs_set_disk_key_offset(&disk_key, 0);
739 btrfs_set_disk_key_type(&disk_key, BTRFS_EXTENT_ITEM_KEY);
740 btrfs_set_disk_key_offset(&disk_key, cfg->nodesize);
742 btrfs_set_disk_key_objectid(&disk_key, bytenr);
744 btrfs_set_item_key(buf, &disk_key, *slot);
745 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
746 btrfs_set_item_size(buf, btrfs_item_nr(*slot), itemsize);
748 ei = btrfs_item_ptr(buf, *slot, struct btrfs_extent_item);
749 btrfs_set_extent_refs(buf, ei, 1);
750 btrfs_set_extent_generation(buf, ei, 1);
751 btrfs_set_extent_flags(buf, ei, BTRFS_EXTENT_FLAG_TREE_BLOCK);
753 if (skinny_metadata) {
754 iref = (struct btrfs_extent_inline_ref *)(ei + 1);
756 info = (struct btrfs_tree_block_info *)(ei + 1);
757 iref = (struct btrfs_extent_inline_ref *)(info + 1);
759 btrfs_set_extent_inline_ref_type(buf, iref,
760 BTRFS_TREE_BLOCK_REF_KEY);
761 btrfs_set_extent_inline_ref_offset(buf, iref, ref_root);
768 * Lastly, check the tree block key by read the tree block
769 * Since we do 1:1 mapping for convert case, we can directly
770 * read the bytenr from disk
772 tmp = malloc(sizeof(*tmp) + cfg->nodesize);
775 ret = setup_temp_extent_buffer(tmp, cfg, bytenr, ref_root);
778 ret = pread(fd, tmp->data, cfg->nodesize, bytenr);
779 if (ret < cfg->nodesize) {
780 ret = (ret < 0 ? -errno : -EIO);
783 if (btrfs_header_nritems(tmp) == 0) {
784 btrfs_set_disk_key_type(&tree_info_key, 0);
785 btrfs_set_disk_key_objectid(&tree_info_key, 0);
786 btrfs_set_disk_key_offset(&tree_info_key, 0);
788 btrfs_item_key(tmp, &tree_info_key, 0);
790 btrfs_set_tree_block_key(buf, info, &tree_info_key);
797 static void insert_temp_block_group(struct extent_buffer *buf,
798 struct btrfs_mkfs_config *cfg,
799 int *slot, u32 *itemoff,
800 u64 bytenr, u64 len, u64 used, u64 flag)
802 struct btrfs_block_group_item bgi;
803 struct btrfs_disk_key disk_key;
805 btrfs_set_header_nritems(buf, *slot + 1);
806 (*itemoff) -= sizeof(bgi);
807 btrfs_set_disk_key_type(&disk_key, BTRFS_BLOCK_GROUP_ITEM_KEY);
808 btrfs_set_disk_key_objectid(&disk_key, bytenr);
809 btrfs_set_disk_key_offset(&disk_key, len);
810 btrfs_set_item_key(buf, &disk_key, *slot);
811 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
812 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(bgi));
814 btrfs_set_block_group_flags(&bgi, flag);
815 btrfs_set_block_group_used(&bgi, used);
816 btrfs_set_block_group_chunk_objectid(&bgi,
817 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
818 write_extent_buffer(buf, &bgi, btrfs_item_ptr_offset(buf, *slot),
823 static int setup_temp_extent_tree(int fd, struct btrfs_mkfs_config *cfg,
824 u64 chunk_bytenr, u64 root_bytenr,
825 u64 extent_bytenr, u64 dev_bytenr,
826 u64 fs_bytenr, u64 csum_bytenr)
828 struct extent_buffer *buf = NULL;
829 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
834 * We must ensure provided bytenr are in ascending order,
835 * or extent tree key order will be broken.
837 BUG_ON(!(chunk_bytenr < root_bytenr && root_bytenr < extent_bytenr &&
838 extent_bytenr < dev_bytenr && dev_bytenr < fs_bytenr &&
839 fs_bytenr < csum_bytenr));
840 buf = malloc(sizeof(*buf) + cfg->nodesize);
844 ret = setup_temp_extent_buffer(buf, cfg, extent_bytenr,
845 BTRFS_EXTENT_TREE_OBJECTID);
849 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
850 chunk_bytenr, BTRFS_CHUNK_TREE_OBJECTID);
854 insert_temp_block_group(buf, cfg, &slot, &itemoff, chunk_bytenr,
855 BTRFS_MKFS_SYSTEM_GROUP_SIZE, cfg->nodesize,
856 BTRFS_BLOCK_GROUP_SYSTEM);
858 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
859 root_bytenr, BTRFS_ROOT_TREE_OBJECTID);
863 /* 5 tree block used, root, extent, dev, fs and csum*/
864 insert_temp_block_group(buf, cfg, &slot, &itemoff, root_bytenr,
865 BTRFS_CONVERT_META_GROUP_SIZE, cfg->nodesize * 5,
866 BTRFS_BLOCK_GROUP_METADATA);
868 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
869 extent_bytenr, BTRFS_EXTENT_TREE_OBJECTID);
872 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
873 dev_bytenr, BTRFS_DEV_TREE_OBJECTID);
876 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
877 fs_bytenr, BTRFS_FS_TREE_OBJECTID);
880 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
881 csum_bytenr, BTRFS_CSUM_TREE_OBJECTID);
885 ret = write_temp_extent_buffer(fd, buf, extent_bytenr);
892 * Improved version of make_btrfs().
895 * 1) Do chunk allocation to avoid used data
896 * And after this function, extent type matches chunk type
897 * 2) Better structured code
898 * No super long hand written codes to initialized all tree blocks
899 * Split into small blocks and reuse codes.
900 * TODO: Reuse tree operation facilities by introducing new flags
902 static int make_convert_btrfs(int fd, struct btrfs_mkfs_config *cfg,
903 struct btrfs_convert_context *cctx)
905 struct cache_tree *free = &cctx->free;
906 struct cache_tree *used = &cctx->used;
908 u64 meta_chunk_start;
909 /* chunk tree bytenr, in system chunk */
911 /* metadata trees bytenr, in metadata chunk */
919 /* Shouldn't happen */
920 BUG_ON(cache_tree_empty(used));
923 * reserve space for temporary superblock first
924 * Here we allocate a little larger space, to keep later
925 * free space will be STRIPE_LEN aligned
927 ret = reserve_free_space(free, BTRFS_STRIPE_LEN,
933 * Then reserve system chunk space
934 * TODO: Change system group size depending on cctx->total_bytes.
935 * If using current 4M, it can only handle less than one TB for
936 * worst case and then run out of sys space.
938 ret = reserve_free_space(free, BTRFS_MKFS_SYSTEM_GROUP_SIZE,
942 ret = reserve_free_space(free, BTRFS_CONVERT_META_GROUP_SIZE,
948 * Allocated meta/sys chunks will be mapped 1:1 with device offset.
950 * Inside the allocated metadata chunk, the layout will be:
951 * | offset | contents |
952 * -------------------------------------
954 * | +nodesize | extent root |
955 * | +nodesize * 2 | device root |
956 * | +nodesize * 3 | fs tree |
957 * | +nodesize * 4 | csum tree |
958 * -------------------------------------
959 * Inside the allocated system chunk, the layout will be:
960 * | offset | contents |
961 * -------------------------------------
962 * | +0 | chunk root |
963 * -------------------------------------
965 chunk_bytenr = sys_chunk_start;
966 root_bytenr = meta_chunk_start;
967 extent_bytenr = meta_chunk_start + cfg->nodesize;
968 dev_bytenr = meta_chunk_start + cfg->nodesize * 2;
969 fs_bytenr = meta_chunk_start + cfg->nodesize * 3;
970 csum_bytenr = meta_chunk_start + cfg->nodesize * 4;
972 ret = setup_temp_super(fd, cfg, root_bytenr, chunk_bytenr);
976 ret = setup_temp_root_tree(fd, cfg, root_bytenr, extent_bytenr,
977 dev_bytenr, fs_bytenr, csum_bytenr);
980 ret = setup_temp_chunk_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
984 ret = setup_temp_dev_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
988 ret = setup_temp_fs_tree(fd, cfg, fs_bytenr);
991 ret = setup_temp_csum_tree(fd, cfg, csum_bytenr);
995 * Setup extent tree last, since it may need to read tree block key
996 * for non-skinny metadata case.
998 ret = setup_temp_extent_tree(fd, cfg, chunk_bytenr, root_bytenr,
999 extent_bytenr, dev_bytenr, fs_bytenr,
1006 * @fs_uuid - if NULL, generates a UUID, returns back the new filesystem UUID
1008 int make_btrfs(int fd, struct btrfs_mkfs_config *cfg,
1009 struct btrfs_convert_context *cctx)
1011 struct btrfs_super_block super;
1012 struct extent_buffer *buf;
1013 struct btrfs_root_item root_item;
1014 struct btrfs_disk_key disk_key;
1015 struct btrfs_extent_item *extent_item;
1016 struct btrfs_inode_item *inode_item;
1017 struct btrfs_chunk *chunk;
1018 struct btrfs_dev_item *dev_item;
1019 struct btrfs_dev_extent *dev_extent;
1020 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
1030 int skinny_metadata = !!(cfg->features &
1031 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA);
1035 return make_convert_btrfs(fd, cfg, cctx);
1036 buf = malloc(sizeof(*buf) + max(cfg->sectorsize, cfg->nodesize));
1040 first_free = BTRFS_SUPER_INFO_OFFSET + cfg->sectorsize * 2 - 1;
1041 first_free &= ~((u64)cfg->sectorsize - 1);
1043 memset(&super, 0, sizeof(super));
1045 num_bytes = (cfg->num_bytes / cfg->sectorsize) * cfg->sectorsize;
1046 if (cfg->fs_uuid && *cfg->fs_uuid) {
1047 if (uuid_parse(cfg->fs_uuid, super.fsid) != 0) {
1048 error("cannot not parse UUID: %s", cfg->fs_uuid);
1052 if (!test_uuid_unique(cfg->fs_uuid)) {
1053 error("non-unique UUID: %s", cfg->fs_uuid);
1058 uuid_generate(super.fsid);
1060 uuid_unparse(super.fsid, cfg->fs_uuid);
1062 uuid_generate(super.dev_item.uuid);
1063 uuid_generate(chunk_tree_uuid);
1065 btrfs_set_super_bytenr(&super, cfg->blocks[0]);
1066 btrfs_set_super_num_devices(&super, 1);
1067 btrfs_set_super_magic(&super, BTRFS_MAGIC);
1068 btrfs_set_super_generation(&super, 1);
1069 btrfs_set_super_root(&super, cfg->blocks[1]);
1070 btrfs_set_super_chunk_root(&super, cfg->blocks[3]);
1071 btrfs_set_super_total_bytes(&super, num_bytes);
1072 btrfs_set_super_bytes_used(&super, 6 * cfg->nodesize);
1073 btrfs_set_super_sectorsize(&super, cfg->sectorsize);
1074 btrfs_set_super_leafsize(&super, cfg->nodesize);
1075 btrfs_set_super_nodesize(&super, cfg->nodesize);
1076 btrfs_set_super_stripesize(&super, cfg->stripesize);
1077 btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
1078 btrfs_set_super_chunk_root_generation(&super, 1);
1079 btrfs_set_super_cache_generation(&super, -1);
1080 btrfs_set_super_incompat_flags(&super, cfg->features);
1082 __strncpy_null(super.label, cfg->label, BTRFS_LABEL_SIZE - 1);
1084 /* create the tree of root objects */
1085 memset(buf->data, 0, cfg->nodesize);
1086 buf->len = cfg->nodesize;
1087 btrfs_set_header_bytenr(buf, cfg->blocks[1]);
1088 btrfs_set_header_nritems(buf, 4);
1089 btrfs_set_header_generation(buf, 1);
1090 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
1091 btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
1092 write_extent_buffer(buf, super.fsid, btrfs_header_fsid(),
1095 write_extent_buffer(buf, chunk_tree_uuid,
1096 btrfs_header_chunk_tree_uuid(buf),
1099 /* create the items for the root tree */
1100 memset(&root_item, 0, sizeof(root_item));
1101 inode_item = &root_item.inode;
1102 btrfs_set_stack_inode_generation(inode_item, 1);
1103 btrfs_set_stack_inode_size(inode_item, 3);
1104 btrfs_set_stack_inode_nlink(inode_item, 1);
1105 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
1106 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
1107 btrfs_set_root_refs(&root_item, 1);
1108 btrfs_set_root_used(&root_item, cfg->nodesize);
1109 btrfs_set_root_generation(&root_item, 1);
1111 memset(&disk_key, 0, sizeof(disk_key));
1112 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
1113 btrfs_set_disk_key_offset(&disk_key, 0);
1116 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - sizeof(root_item);
1117 btrfs_set_root_bytenr(&root_item, cfg->blocks[2]);
1118 btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
1119 btrfs_set_item_key(buf, &disk_key, nritems);
1120 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1121 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1123 write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
1124 nritems), sizeof(root_item));
1127 itemoff = itemoff - sizeof(root_item);
1128 btrfs_set_root_bytenr(&root_item, cfg->blocks[4]);
1129 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
1130 btrfs_set_item_key(buf, &disk_key, nritems);
1131 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1132 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1134 write_extent_buffer(buf, &root_item,
1135 btrfs_item_ptr_offset(buf, nritems),
1139 itemoff = itemoff - sizeof(root_item);
1140 btrfs_set_root_bytenr(&root_item, cfg->blocks[5]);
1141 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
1142 btrfs_set_item_key(buf, &disk_key, nritems);
1143 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1144 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1146 write_extent_buffer(buf, &root_item,
1147 btrfs_item_ptr_offset(buf, nritems),
1151 itemoff = itemoff - sizeof(root_item);
1152 btrfs_set_root_bytenr(&root_item, cfg->blocks[6]);
1153 btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_TREE_OBJECTID);
1154 btrfs_set_item_key(buf, &disk_key, nritems);
1155 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1156 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1158 write_extent_buffer(buf, &root_item,
1159 btrfs_item_ptr_offset(buf, nritems),
1164 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1165 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[1]);
1166 if (ret != cfg->nodesize) {
1167 ret = (ret < 0 ? -errno : -EIO);
1171 /* create the items for the extent tree */
1172 memset(buf->data + sizeof(struct btrfs_header), 0,
1173 cfg->nodesize - sizeof(struct btrfs_header));
1175 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
1176 for (i = 1; i < 7; i++) {
1177 item_size = sizeof(struct btrfs_extent_item);
1178 if (!skinny_metadata)
1179 item_size += sizeof(struct btrfs_tree_block_info);
1181 BUG_ON(cfg->blocks[i] < first_free);
1182 BUG_ON(cfg->blocks[i] < cfg->blocks[i - 1]);
1184 /* create extent item */
1185 itemoff -= item_size;
1186 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
1187 if (skinny_metadata) {
1188 btrfs_set_disk_key_type(&disk_key,
1189 BTRFS_METADATA_ITEM_KEY);
1190 btrfs_set_disk_key_offset(&disk_key, 0);
1192 btrfs_set_disk_key_type(&disk_key,
1193 BTRFS_EXTENT_ITEM_KEY);
1194 btrfs_set_disk_key_offset(&disk_key, cfg->nodesize);
1196 btrfs_set_item_key(buf, &disk_key, nritems);
1197 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
1199 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1201 extent_item = btrfs_item_ptr(buf, nritems,
1202 struct btrfs_extent_item);
1203 btrfs_set_extent_refs(buf, extent_item, 1);
1204 btrfs_set_extent_generation(buf, extent_item, 1);
1205 btrfs_set_extent_flags(buf, extent_item,
1206 BTRFS_EXTENT_FLAG_TREE_BLOCK);
1209 /* create extent ref */
1210 ref_root = reference_root_table[i];
1211 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
1212 btrfs_set_disk_key_offset(&disk_key, ref_root);
1213 btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
1214 btrfs_set_item_key(buf, &disk_key, nritems);
1215 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
1217 btrfs_set_item_size(buf, btrfs_item_nr(nritems), 0);
1220 btrfs_set_header_bytenr(buf, cfg->blocks[2]);
1221 btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
1222 btrfs_set_header_nritems(buf, nritems);
1223 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1224 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[2]);
1225 if (ret != cfg->nodesize) {
1226 ret = (ret < 0 ? -errno : -EIO);
1230 /* create the chunk tree */
1231 memset(buf->data + sizeof(struct btrfs_header), 0,
1232 cfg->nodesize - sizeof(struct btrfs_header));
1234 item_size = sizeof(*dev_item);
1235 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - item_size;
1237 /* first device 1 (there is no device 0) */
1238 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
1239 btrfs_set_disk_key_offset(&disk_key, 1);
1240 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
1241 btrfs_set_item_key(buf, &disk_key, nritems);
1242 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1243 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
1245 dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
1246 btrfs_set_device_id(buf, dev_item, 1);
1247 btrfs_set_device_generation(buf, dev_item, 0);
1248 btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
1249 btrfs_set_device_bytes_used(buf, dev_item,
1250 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1251 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
1252 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
1253 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
1254 btrfs_set_device_type(buf, dev_item, 0);
1256 write_extent_buffer(buf, super.dev_item.uuid,
1257 (unsigned long)btrfs_device_uuid(dev_item),
1259 write_extent_buffer(buf, super.fsid,
1260 (unsigned long)btrfs_device_fsid(dev_item),
1262 read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
1266 item_size = btrfs_chunk_item_size(1);
1267 itemoff = itemoff - item_size;
1269 /* then we have chunk 0 */
1270 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1271 btrfs_set_disk_key_offset(&disk_key, 0);
1272 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
1273 btrfs_set_item_key(buf, &disk_key, nritems);
1274 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1275 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
1277 chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
1278 btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1279 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
1280 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
1281 btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
1282 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
1283 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
1284 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
1285 btrfs_set_chunk_num_stripes(buf, chunk, 1);
1286 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
1287 btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
1290 write_extent_buffer(buf, super.dev_item.uuid,
1291 (unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
1294 /* copy the key for the chunk to the system array */
1295 ptr = super.sys_chunk_array;
1296 array_size = sizeof(disk_key);
1298 memcpy(ptr, &disk_key, sizeof(disk_key));
1299 ptr += sizeof(disk_key);
1301 /* copy the chunk to the system array */
1302 read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
1303 array_size += item_size;
1305 btrfs_set_super_sys_array_size(&super, array_size);
1307 btrfs_set_header_bytenr(buf, cfg->blocks[3]);
1308 btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
1309 btrfs_set_header_nritems(buf, nritems);
1310 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1311 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[3]);
1312 if (ret != cfg->nodesize) {
1313 ret = (ret < 0 ? -errno : -EIO);
1317 /* create the device tree */
1318 memset(buf->data + sizeof(struct btrfs_header), 0,
1319 cfg->nodesize - sizeof(struct btrfs_header));
1321 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) -
1322 sizeof(struct btrfs_dev_extent);
1324 btrfs_set_disk_key_objectid(&disk_key, 1);
1325 btrfs_set_disk_key_offset(&disk_key, 0);
1326 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
1327 btrfs_set_item_key(buf, &disk_key, nritems);
1328 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1329 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1330 sizeof(struct btrfs_dev_extent));
1331 dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
1332 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
1333 BTRFS_CHUNK_TREE_OBJECTID);
1334 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
1335 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1336 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
1338 write_extent_buffer(buf, chunk_tree_uuid,
1339 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
1342 btrfs_set_dev_extent_length(buf, dev_extent,
1343 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1346 btrfs_set_header_bytenr(buf, cfg->blocks[4]);
1347 btrfs_set_header_owner(buf, BTRFS_DEV_TREE_OBJECTID);
1348 btrfs_set_header_nritems(buf, nritems);
1349 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1350 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[4]);
1351 if (ret != cfg->nodesize) {
1352 ret = (ret < 0 ? -errno : -EIO);
1356 /* create the FS root */
1357 memset(buf->data + sizeof(struct btrfs_header), 0,
1358 cfg->nodesize - sizeof(struct btrfs_header));
1359 btrfs_set_header_bytenr(buf, cfg->blocks[5]);
1360 btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
1361 btrfs_set_header_nritems(buf, 0);
1362 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1363 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[5]);
1364 if (ret != cfg->nodesize) {
1365 ret = (ret < 0 ? -errno : -EIO);
1368 /* finally create the csum root */
1369 memset(buf->data + sizeof(struct btrfs_header), 0,
1370 cfg->nodesize - sizeof(struct btrfs_header));
1371 btrfs_set_header_bytenr(buf, cfg->blocks[6]);
1372 btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
1373 btrfs_set_header_nritems(buf, 0);
1374 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1375 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[6]);
1376 if (ret != cfg->nodesize) {
1377 ret = (ret < 0 ? -errno : -EIO);
1381 /* and write out the super block */
1382 BUG_ON(sizeof(super) > cfg->sectorsize);
1383 memset(buf->data, 0, BTRFS_SUPER_INFO_SIZE);
1384 memcpy(buf->data, &super, sizeof(super));
1385 buf->len = BTRFS_SUPER_INFO_SIZE;
1386 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1387 ret = pwrite(fd, buf->data, BTRFS_SUPER_INFO_SIZE, cfg->blocks[0]);
1388 if (ret != BTRFS_SUPER_INFO_SIZE) {
1389 ret = (ret < 0 ? -errno : -EIO);
1400 static const struct btrfs_fs_feature {
1404 } mkfs_features[] = {
1405 { "mixed-bg", BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS,
1406 "mixed data and metadata block groups" },
1407 { "extref", BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF,
1408 "increased hardlink limit per file to 65536" },
1409 { "raid56", BTRFS_FEATURE_INCOMPAT_RAID56,
1410 "raid56 extended format" },
1411 { "skinny-metadata", BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA,
1412 "reduced-size metadata extent refs" },
1413 { "no-holes", BTRFS_FEATURE_INCOMPAT_NO_HOLES,
1414 "no explicit hole extents for files" },
1415 /* Keep this one last */
1416 { "list-all", BTRFS_FEATURE_LIST_ALL, NULL }
1419 static int parse_one_fs_feature(const char *name, u64 *flags)
1424 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1425 if (name[0] == '^' &&
1426 !strcmp(mkfs_features[i].name, name + 1)) {
1427 *flags &= ~ mkfs_features[i].flag;
1429 } else if (!strcmp(mkfs_features[i].name, name)) {
1430 *flags |= mkfs_features[i].flag;
1438 void btrfs_parse_features_to_string(char *buf, u64 flags)
1444 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1445 if (flags & mkfs_features[i].flag) {
1448 strcat(buf, mkfs_features[i].name);
1453 void btrfs_process_fs_features(u64 flags)
1457 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1458 if (flags & mkfs_features[i].flag) {
1459 printf("Turning ON incompat feature '%s': %s\n",
1460 mkfs_features[i].name,
1461 mkfs_features[i].desc);
1466 void btrfs_list_all_fs_features(u64 mask_disallowed)
1470 fprintf(stderr, "Filesystem features available:\n");
1471 for (i = 0; i < ARRAY_SIZE(mkfs_features) - 1; i++) {
1472 char *is_default = "";
1474 if (mkfs_features[i].flag & mask_disallowed)
1476 if (mkfs_features[i].flag & BTRFS_MKFS_DEFAULT_FEATURES)
1477 is_default = ", default";
1478 fprintf(stderr, "%-20s- %s (0x%llx%s)\n",
1479 mkfs_features[i].name,
1480 mkfs_features[i].desc,
1481 mkfs_features[i].flag,
1487 * Return NULL if all features were parsed fine, otherwise return the name of
1488 * the first unparsed.
1490 char* btrfs_parse_fs_features(char *namelist, u64 *flags)
1493 char *save_ptr = NULL; /* Satisfy static checkers */
1495 for (this_char = strtok_r(namelist, ",", &save_ptr);
1497 this_char = strtok_r(NULL, ",", &save_ptr)) {
1498 if (parse_one_fs_feature(this_char, flags))
1505 u64 btrfs_device_size(int fd, struct stat *st)
1508 if (S_ISREG(st->st_mode)) {
1511 if (!S_ISBLK(st->st_mode)) {
1514 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
1520 static int zero_blocks(int fd, off_t start, size_t len)
1522 char *buf = malloc(len);
1528 memset(buf, 0, len);
1529 written = pwrite(fd, buf, len, start);
1536 #define ZERO_DEV_BYTES (2 * 1024 * 1024)
1538 /* don't write outside the device by clamping the region to the device size */
1539 static int zero_dev_clamped(int fd, off_t start, ssize_t len, u64 dev_size)
1541 off_t end = max(start, start + len);
1544 /* and don't overwrite the disk labels on sparc */
1545 start = max(start, 1024);
1546 end = max(end, 1024);
1549 start = min_t(u64, start, dev_size);
1550 end = min_t(u64, end, dev_size);
1552 return zero_blocks(fd, start, end - start);
1555 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
1556 struct btrfs_root *root, int fd, char *path,
1557 u64 device_total_bytes, u32 io_width, u32 io_align,
1560 struct btrfs_super_block *disk_super;
1561 struct btrfs_super_block *super = root->fs_info->super_copy;
1562 struct btrfs_device *device;
1563 struct btrfs_dev_item *dev_item;
1569 device_total_bytes = (device_total_bytes / sectorsize) * sectorsize;
1571 device = kzalloc(sizeof(*device), GFP_NOFS);
1574 buf = kzalloc(sectorsize, GFP_NOFS);
1577 BUG_ON(sizeof(*disk_super) > sectorsize);
1579 disk_super = (struct btrfs_super_block *)buf;
1580 dev_item = &disk_super->dev_item;
1582 uuid_generate(device->uuid);
1585 device->io_width = io_width;
1586 device->io_align = io_align;
1587 device->sector_size = sectorsize;
1589 device->writeable = 1;
1590 device->total_bytes = device_total_bytes;
1591 device->bytes_used = 0;
1592 device->total_ios = 0;
1593 device->dev_root = root->fs_info->dev_root;
1594 device->name = strdup(path);
1598 INIT_LIST_HEAD(&device->dev_list);
1599 ret = btrfs_add_device(trans, root, device);
1602 fs_total_bytes = btrfs_super_total_bytes(super) + device_total_bytes;
1603 btrfs_set_super_total_bytes(super, fs_total_bytes);
1605 num_devs = btrfs_super_num_devices(super) + 1;
1606 btrfs_set_super_num_devices(super, num_devs);
1608 memcpy(disk_super, super, sizeof(*disk_super));
1610 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
1611 btrfs_set_stack_device_id(dev_item, device->devid);
1612 btrfs_set_stack_device_type(dev_item, device->type);
1613 btrfs_set_stack_device_io_align(dev_item, device->io_align);
1614 btrfs_set_stack_device_io_width(dev_item, device->io_width);
1615 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
1616 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
1617 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
1618 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
1620 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
1621 BUG_ON(ret != sectorsize);
1624 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
1625 device->fs_devices = root->fs_info->fs_devices;
1634 static int btrfs_wipe_existing_sb(int fd)
1636 const char *off = NULL;
1641 blkid_probe pr = NULL;
1643 pr = blkid_new_probe();
1647 if (blkid_probe_set_device(pr, fd, 0, 0)) {
1652 ret = blkid_probe_lookup_value(pr, "SBMAGIC_OFFSET", &off, NULL);
1654 ret = blkid_probe_lookup_value(pr, "SBMAGIC", NULL, &len);
1656 if (ret || len == 0 || off == NULL) {
1658 * If lookup fails, the probe did not find any values, eg. for
1659 * a file image or a loop device. Soft error.
1665 offset = strtoll(off, NULL, 10);
1666 if (len > sizeof(buf))
1669 memset(buf, 0, len);
1670 ret = pwrite(fd, buf, len, offset);
1672 error("cannot wipe existing superblock: %s", strerror(errno));
1674 } else if (ret != len) {
1675 error("cannot wipe existing superblock: wrote %d of %zd", ret, len);
1681 blkid_free_probe(pr);
1685 int btrfs_prepare_device(int fd, const char *file, int zero_end,
1686 u64 *block_count_ret, u64 max_block_count, int discard)
1692 ret = fstat(fd, &st);
1694 error("unable to stat %s: %s", file, strerror(errno));
1698 block_count = btrfs_device_size(fd, &st);
1699 if (block_count == 0) {
1700 error("unable to determine size of %s", file);
1703 if (max_block_count)
1704 block_count = min(block_count, max_block_count);
1708 * We intentionally ignore errors from the discard ioctl. It
1709 * is not necessary for the mkfs functionality but just an
1712 if (discard_range(fd, 0, 0) == 0) {
1713 printf("Performing full device TRIM (%s) ...\n",
1714 pretty_size(block_count));
1715 discard_blocks(fd, 0, block_count);
1719 ret = zero_dev_clamped(fd, 0, ZERO_DEV_BYTES, block_count);
1720 for (i = 0 ; !ret && i < BTRFS_SUPER_MIRROR_MAX; i++)
1721 ret = zero_dev_clamped(fd, btrfs_sb_offset(i),
1722 BTRFS_SUPER_INFO_SIZE, block_count);
1723 if (!ret && zero_end)
1724 ret = zero_dev_clamped(fd, block_count - ZERO_DEV_BYTES,
1725 ZERO_DEV_BYTES, block_count);
1728 error("failed to zero device '%s': %s", file, strerror(-ret));
1732 ret = btrfs_wipe_existing_sb(fd);
1734 error("cannot wipe superblocks on %s", file);
1738 *block_count_ret = block_count;
1742 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
1743 struct btrfs_root *root, u64 objectid)
1746 struct btrfs_inode_item inode_item;
1747 time_t now = time(NULL);
1749 memset(&inode_item, 0, sizeof(inode_item));
1750 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
1751 btrfs_set_stack_inode_size(&inode_item, 0);
1752 btrfs_set_stack_inode_nlink(&inode_item, 1);
1753 btrfs_set_stack_inode_nbytes(&inode_item, root->nodesize);
1754 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
1755 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
1756 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
1757 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
1758 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
1759 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
1760 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
1761 btrfs_set_stack_timespec_sec(&inode_item.otime, 0);
1762 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
1764 if (root->fs_info->tree_root == root)
1765 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
1767 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
1771 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
1775 btrfs_set_root_dirid(&root->root_item, objectid);
1782 * checks if a path is a block device node
1783 * Returns negative errno on failure, otherwise
1784 * returns 1 for blockdev, 0 for not-blockdev
1786 int is_block_device(const char *path)
1788 struct stat statbuf;
1790 if (stat(path, &statbuf) < 0)
1793 return !!S_ISBLK(statbuf.st_mode);
1797 * check if given path is a mount point
1798 * return 1 if yes. 0 if no. -1 for error
1800 int is_mount_point(const char *path)
1806 f = setmntent("/proc/self/mounts", "r");
1810 while ((mnt = getmntent(f)) != NULL) {
1811 if (strcmp(mnt->mnt_dir, path))
1820 static int is_reg_file(const char *path)
1822 struct stat statbuf;
1824 if (stat(path, &statbuf) < 0)
1826 return S_ISREG(statbuf.st_mode);
1830 * This function checks if the given input parameter is
1832 * return <0 : some error in the given input
1833 * return BTRFS_ARG_UNKNOWN: unknown input
1834 * return BTRFS_ARG_UUID: given input is uuid
1835 * return BTRFS_ARG_MNTPOINT: given input is path
1836 * return BTRFS_ARG_REG: given input is regular file
1837 * return BTRFS_ARG_BLKDEV: given input is block device
1839 int check_arg_type(const char *input)
1842 char path[PATH_MAX];
1847 if (realpath(input, path)) {
1848 if (is_block_device(path) == 1)
1849 return BTRFS_ARG_BLKDEV;
1851 if (is_mount_point(path) == 1)
1852 return BTRFS_ARG_MNTPOINT;
1854 if (is_reg_file(path))
1855 return BTRFS_ARG_REG;
1857 return BTRFS_ARG_UNKNOWN;
1860 if (strlen(input) == (BTRFS_UUID_UNPARSED_SIZE - 1) &&
1861 !uuid_parse(input, uuid))
1862 return BTRFS_ARG_UUID;
1864 return BTRFS_ARG_UNKNOWN;
1868 * Find the mount point for a mounted device.
1869 * On success, returns 0 with mountpoint in *mp.
1870 * On failure, returns -errno (not mounted yields -EINVAL)
1871 * Is noisy on failures, expects to be given a mounted device.
1873 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size)
1878 ret = is_block_device(dev);
1881 error("not a block device: %s", dev);
1884 error("cannot check %s: %s", dev, strerror(-ret));
1889 fd = open(dev, O_RDONLY);
1892 error("cannot open %s: %s", dev, strerror(errno));
1896 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
1899 } else { /* mounted, all good */
1909 * Given a pathname, return a filehandle to:
1910 * the original pathname or,
1911 * if the pathname is a mounted btrfs device, to its mountpoint.
1913 * On error, return -1, errno should be set.
1915 int open_path_or_dev_mnt(const char *path, DIR **dirstream, int verbose)
1920 if (is_block_device(path)) {
1921 ret = get_btrfs_mount(path, mp, sizeof(mp));
1923 /* not a mounted btrfs dev */
1924 error_on(verbose, "'%s' is not a mounted btrfs device",
1929 ret = open_file_or_dir(mp, dirstream);
1930 error_on(verbose && ret < 0, "can't access '%s': %s",
1931 path, strerror(errno));
1933 ret = btrfs_open_dir(path, dirstream, 1);
1940 * Do the following checks before calling open_file_or_dir():
1941 * 1: path is in a btrfs filesystem
1942 * 2: path is a directory
1944 int btrfs_open_dir(const char *path, DIR **dirstream, int verbose)
1950 if (statfs(path, &stfs) != 0) {
1951 error_on(verbose, "cannot access '%s': %s", path,
1956 if (stfs.f_type != BTRFS_SUPER_MAGIC) {
1957 error_on(verbose, "not a btrfs filesystem: %s", path);
1961 if (stat(path, &st) != 0) {
1962 error_on(verbose, "cannot access '%s': %s", path,
1967 if (!S_ISDIR(st.st_mode)) {
1968 error_on(verbose, "not a directory: %s", path);
1972 ret = open_file_or_dir(path, dirstream);
1974 error_on(verbose, "cannot access '%s': %s", path,
1981 /* checks if a device is a loop device */
1982 static int is_loop_device (const char* device) {
1983 struct stat statbuf;
1985 if(stat(device, &statbuf) < 0)
1988 return (S_ISBLK(statbuf.st_mode) &&
1989 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
1993 * Takes a loop device path (e.g. /dev/loop0) and returns
1994 * the associated file (e.g. /images/my_btrfs.img) using
1997 static int resolve_loop_device_with_loopdev(const char* loop_dev, char* loop_file)
2001 struct loop_info64 lo64;
2003 fd = open(loop_dev, O_RDONLY | O_NONBLOCK);
2006 ret = ioctl(fd, LOOP_GET_STATUS64, &lo64);
2012 memcpy(loop_file, lo64.lo_file_name, sizeof(lo64.lo_file_name));
2013 loop_file[sizeof(lo64.lo_file_name)] = 0;
2021 /* Takes a loop device path (e.g. /dev/loop0) and returns
2022 * the associated file (e.g. /images/my_btrfs.img) */
2023 static int resolve_loop_device(const char* loop_dev, char* loop_file,
2030 char real_loop_dev[PATH_MAX];
2032 if (!realpath(loop_dev, real_loop_dev))
2034 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
2035 if (!(f = fopen(p, "r"))) {
2036 if (errno == ENOENT)
2038 * It's possibly a partitioned loop device, which is
2039 * resolvable with loopdev API.
2041 return resolve_loop_device_with_loopdev(loop_dev, loop_file);
2045 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
2046 ret = fscanf(f, fmt, loop_file);
2055 * Checks whether a and b are identical or device
2056 * files associated with the same block device
2058 static int is_same_blk_file(const char* a, const char* b)
2060 struct stat st_buf_a, st_buf_b;
2061 char real_a[PATH_MAX];
2062 char real_b[PATH_MAX];
2064 if (!realpath(a, real_a))
2065 strncpy_null(real_a, a);
2067 if (!realpath(b, real_b))
2068 strncpy_null(real_b, b);
2070 /* Identical path? */
2071 if (strcmp(real_a, real_b) == 0)
2074 if (stat(a, &st_buf_a) < 0 || stat(b, &st_buf_b) < 0) {
2075 if (errno == ENOENT)
2080 /* Same blockdevice? */
2081 if (S_ISBLK(st_buf_a.st_mode) && S_ISBLK(st_buf_b.st_mode) &&
2082 st_buf_a.st_rdev == st_buf_b.st_rdev) {
2087 if (st_buf_a.st_dev == st_buf_b.st_dev &&
2088 st_buf_a.st_ino == st_buf_b.st_ino) {
2095 /* checks if a and b are identical or device
2096 * files associated with the same block device or
2097 * if one file is a loop device that uses the other
2100 static int is_same_loop_file(const char* a, const char* b)
2102 char res_a[PATH_MAX];
2103 char res_b[PATH_MAX];
2104 const char* final_a = NULL;
2105 const char* final_b = NULL;
2108 /* Resolve a if it is a loop device */
2109 if((ret = is_loop_device(a)) < 0) {
2114 ret = resolve_loop_device(a, res_a, sizeof(res_a));
2125 /* Resolve b if it is a loop device */
2126 if ((ret = is_loop_device(b)) < 0) {
2131 ret = resolve_loop_device(b, res_b, sizeof(res_b));
2142 return is_same_blk_file(final_a, final_b);
2145 /* Checks if a file exists and is a block or regular file*/
2146 static int is_existing_blk_or_reg_file(const char* filename)
2150 if(stat(filename, &st_buf) < 0) {
2157 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
2160 /* Checks if a file is used (directly or indirectly via a loop device)
2161 * by a device in fs_devices
2163 static int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices,
2167 struct list_head *head;
2168 struct list_head *cur;
2169 struct btrfs_device *device;
2171 head = &fs_devices->devices;
2172 list_for_each(cur, head) {
2173 device = list_entry(cur, struct btrfs_device, dev_list);
2175 if((ret = is_same_loop_file(device->name, file)))
2183 * Resolve a pathname to a device mapper node to /dev/mapper/<name>
2184 * Returns NULL on invalid input or malloc failure; Other failures
2185 * will be handled by the caller using the input pathame.
2187 char *canonicalize_dm_name(const char *ptname)
2191 char path[PATH_MAX], name[PATH_MAX], *res = NULL;
2193 if (!ptname || !*ptname)
2196 snprintf(path, sizeof(path), "/sys/block/%s/dm/name", ptname);
2197 if (!(f = fopen(path, "r")))
2200 /* read <name>\n from sysfs */
2201 if (fgets(name, sizeof(name), f) && (sz = strlen(name)) > 1) {
2202 name[sz - 1] = '\0';
2203 snprintf(path, sizeof(path), "/dev/mapper/%s", name);
2205 if (access(path, F_OK) == 0)
2213 * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or
2214 * to a device mapper pathname.
2215 * Returns NULL on invalid input or malloc failure; Other failures
2216 * will be handled by the caller using the input pathame.
2218 char *canonicalize_path(const char *path)
2220 char *canonical, *p;
2222 if (!path || !*path)
2225 canonical = realpath(path, NULL);
2227 return strdup(path);
2228 p = strrchr(canonical, '/');
2229 if (p && strncmp(p, "/dm-", 4) == 0 && isdigit(*(p + 4))) {
2230 char *dm = canonicalize_dm_name(p + 1);
2241 * returns 1 if the device was mounted, < 0 on error or 0 if everything
2242 * is safe to continue.
2244 int check_mounted(const char* file)
2249 fd = open(file, O_RDONLY);
2251 error("mount check: cannot open %s: %s", file,
2256 ret = check_mounted_where(fd, file, NULL, 0, NULL);
2262 int check_mounted_where(int fd, const char *file, char *where, int size,
2263 struct btrfs_fs_devices **fs_dev_ret)
2268 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2272 /* scan the initial device */
2273 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
2274 &total_devs, BTRFS_SUPER_INFO_OFFSET, 0);
2275 is_btrfs = (ret >= 0);
2277 /* scan other devices */
2278 if (is_btrfs && total_devs > 1) {
2279 ret = btrfs_scan_lblkid();
2284 /* iterate over the list of currently mounted filesystems */
2285 if ((f = setmntent ("/proc/self/mounts", "r")) == NULL)
2288 while ((mnt = getmntent (f)) != NULL) {
2290 if(strcmp(mnt->mnt_type, "btrfs") != 0)
2293 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
2295 /* ignore entries in the mount table that are not
2296 associated with a file*/
2297 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
2298 goto out_mntloop_err;
2302 ret = is_same_loop_file(file, mnt->mnt_fsname);
2306 goto out_mntloop_err;
2311 /* Did we find an entry in mnt table? */
2312 if (mnt && size && where) {
2313 strncpy(where, mnt->mnt_dir, size);
2317 *fs_dev_ret = fs_devices_mnt;
2319 ret = (mnt != NULL);
2327 struct pending_dir {
2328 struct list_head list;
2329 char name[PATH_MAX];
2332 int btrfs_register_one_device(const char *fname)
2334 struct btrfs_ioctl_vol_args args;
2338 fd = open("/dev/btrfs-control", O_RDWR);
2341 "failed to open /dev/btrfs-control, skipping device registration: %s",
2345 memset(&args, 0, sizeof(args));
2346 strncpy_null(args.name, fname);
2347 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
2349 error("device scan failed on '%s': %s", fname,
2358 * Register all devices in the fs_uuid list created in the user
2359 * space. Ensure btrfs_scan_lblkid() is called before this func.
2361 int btrfs_register_all_devices(void)
2365 struct btrfs_fs_devices *fs_devices;
2366 struct btrfs_device *device;
2367 struct list_head *all_uuids;
2369 all_uuids = btrfs_scanned_uuids();
2371 list_for_each_entry(fs_devices, all_uuids, list) {
2372 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2374 err = btrfs_register_one_device(device->name);
2384 int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
2387 struct btrfs_super_block *disk_super;
2391 buf = malloc(BTRFS_SUPER_INFO_SIZE);
2396 ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
2397 if (ret != BTRFS_SUPER_INFO_SIZE)
2401 disk_super = (struct btrfs_super_block *)buf;
2402 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC)
2405 if (!memcmp(disk_super->fsid, root->fs_info->super_copy->fsid,
2415 * Note: this function uses a static per-thread buffer. Do not call this
2416 * function more than 10 times within one argument list!
2418 const char *pretty_size_mode(u64 size, unsigned mode)
2420 static __thread int ps_index = 0;
2421 static __thread char ps_array[10][32];
2424 ret = ps_array[ps_index];
2427 (void)pretty_size_snprintf(size, ret, 32, mode);
2432 static const char* unit_suffix_binary[] =
2433 { "B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB"};
2434 static const char* unit_suffix_decimal[] =
2435 { "B", "kB", "MB", "GB", "TB", "PB", "EB"};
2437 int pretty_size_snprintf(u64 size, char *str, size_t str_size, unsigned unit_mode)
2443 const char** suffix = NULL;
2449 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_RAW) {
2450 snprintf(str, str_size, "%llu", size);
2454 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_BINARY) {
2457 suffix = unit_suffix_binary;
2458 } else if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_DECIMAL) {
2461 suffix = unit_suffix_decimal;
2466 fprintf(stderr, "INTERNAL ERROR: unknown unit base, mode %d\n",
2474 switch (unit_mode & UNITS_MODE_MASK) {
2475 case UNITS_TBYTES: base *= mult; num_divs++;
2476 case UNITS_GBYTES: base *= mult; num_divs++;
2477 case UNITS_MBYTES: base *= mult; num_divs++;
2478 case UNITS_KBYTES: num_divs++;
2485 while (size >= mult) {
2491 * If the value is smaller than base, we didn't do any
2492 * division, in that case, base should be 1, not original
2493 * base, or the unit will be wrong
2499 if (num_divs >= ARRAY_SIZE(unit_suffix_binary)) {
2501 printf("INTERNAL ERROR: unsupported unit suffix, index %d\n",
2506 fraction = (float)last_size / base;
2508 return snprintf(str, str_size, "%.2f%s", fraction, suffix[num_divs]);
2512 * __strncpy_null - strncpy with null termination
2513 * @dest: the target array
2514 * @src: the source string
2515 * @n: maximum bytes to copy (size of *dest)
2517 * Like strncpy, but ensures destination is null-terminated.
2519 * Copies the string pointed to by src, including the terminating null
2520 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
2521 * of n bytes. Then ensure that dest is null-terminated.
2523 char *__strncpy_null(char *dest, const char *src, size_t n)
2525 strncpy(dest, src, n);
2532 * Checks to make sure that the label matches our requirements.
2534 0 if everything is safe and usable
2535 -1 if the label is too long
2537 static int check_label(const char *input)
2539 int len = strlen(input);
2541 if (len > BTRFS_LABEL_SIZE - 1) {
2542 error("label %s is too long (max %d)", input,
2543 BTRFS_LABEL_SIZE - 1);
2550 static int set_label_unmounted(const char *dev, const char *label)
2552 struct btrfs_trans_handle *trans;
2553 struct btrfs_root *root;
2556 ret = check_mounted(dev);
2558 error("checking mount status of %s failed: %d", dev, ret);
2562 error("device %s is mounted, use mount point", dev);
2566 /* Open the super_block at the default location
2567 * and as read-write.
2569 root = open_ctree(dev, 0, OPEN_CTREE_WRITES);
2570 if (!root) /* errors are printed by open_ctree() */
2573 trans = btrfs_start_transaction(root, 1);
2574 __strncpy_null(root->fs_info->super_copy->label, label, BTRFS_LABEL_SIZE - 1);
2576 btrfs_commit_transaction(trans, root);
2578 /* Now we close it since we are done. */
2583 static int set_label_mounted(const char *mount_path, const char *labelp)
2586 char label[BTRFS_LABEL_SIZE];
2588 fd = open(mount_path, O_RDONLY | O_NOATIME);
2590 error("unable to access %s: %s", mount_path, strerror(errno));
2594 memset(label, 0, sizeof(label));
2595 __strncpy_null(label, labelp, BTRFS_LABEL_SIZE - 1);
2596 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
2597 error("unable to set label of %s: %s", mount_path,
2607 int get_label_unmounted(const char *dev, char *label)
2609 struct btrfs_root *root;
2612 ret = check_mounted(dev);
2614 error("checking mount status of %s failed: %d", dev, ret);
2618 /* Open the super_block at the default location
2621 root = open_ctree(dev, 0, 0);
2625 __strncpy_null(label, root->fs_info->super_copy->label,
2626 BTRFS_LABEL_SIZE - 1);
2628 /* Now we close it since we are done. */
2634 * If a partition is mounted, try to get the filesystem label via its
2635 * mounted path rather than device. Return the corresponding error
2636 * the user specified the device path.
2638 int get_label_mounted(const char *mount_path, char *labelp)
2640 char label[BTRFS_LABEL_SIZE];
2644 fd = open(mount_path, O_RDONLY | O_NOATIME);
2646 error("unable to access %s: %s", mount_path, strerror(errno));
2650 memset(label, '\0', sizeof(label));
2651 ret = ioctl(fd, BTRFS_IOC_GET_FSLABEL, label);
2653 if (errno != ENOTTY)
2654 error("unable to get label of %s: %s", mount_path,
2661 __strncpy_null(labelp, label, BTRFS_LABEL_SIZE - 1);
2666 int get_label(const char *btrfs_dev, char *label)
2670 ret = is_existing_blk_or_reg_file(btrfs_dev);
2672 ret = get_label_mounted(btrfs_dev, label);
2674 ret = get_label_unmounted(btrfs_dev, label);
2679 int set_label(const char *btrfs_dev, const char *label)
2683 if (check_label(label))
2686 ret = is_existing_blk_or_reg_file(btrfs_dev);
2688 ret = set_label_mounted(btrfs_dev, label);
2690 ret = set_label_unmounted(btrfs_dev, label);
2696 * A not-so-good version fls64. No fascinating optimization since
2697 * no one except parse_size use it
2699 static int fls64(u64 x)
2703 for (i = 0; i <64; i++)
2704 if (x << i & (1ULL << 63))
2709 u64 parse_size(char *s)
2717 error("size value is empty");
2721 error("size value '%s' is less equal than 0", s);
2724 ret = strtoull(s, &endptr, 10);
2726 error("size value '%s' is invalid", s);
2729 if (endptr[0] && endptr[1]) {
2730 error("illegal suffix contains character '%c' in wrong position",
2735 * strtoll returns LLONG_MAX when overflow, if this happens,
2736 * need to call strtoull to get the real size
2738 if (errno == ERANGE && ret == ULLONG_MAX) {
2739 error("size value '%s' is too large for u64", s);
2743 c = tolower(endptr[0]);
2766 error("unknown size descriptor '%c'", c);
2770 /* Check whether ret * mult overflow */
2771 if (fls64(ret) + fls64(mult) - 1 > 64) {
2772 error("size value '%s' is too large for u64", s);
2779 u64 parse_qgroupid(const char *p)
2781 char *s = strchr(p, '/');
2782 const char *ptr_src_end = p + strlen(p);
2783 char *ptr_parse_end = NULL;
2792 /* Numeric format like '0/257' is the primary case */
2794 id = strtoull(p, &ptr_parse_end, 10);
2795 if (ptr_parse_end != ptr_src_end)
2799 level = strtoull(p, &ptr_parse_end, 10);
2800 if (ptr_parse_end != s)
2803 id = strtoull(s + 1, &ptr_parse_end, 10);
2804 if (ptr_parse_end != ptr_src_end)
2807 return (level << BTRFS_QGROUP_LEVEL_SHIFT) | id;
2810 /* Path format like subv at 'my_subvol' is the fallback case */
2811 ret = test_issubvolume(p);
2812 if (ret < 0 || !ret)
2814 fd = open(p, O_RDONLY);
2817 ret = lookup_ino_rootid(fd, &id);
2824 error("invalid qgroupid or subvolume path: %s", p);
2828 int open_file_or_dir3(const char *fname, DIR **dirstream, int open_flags)
2834 ret = stat(fname, &st);
2838 if (S_ISDIR(st.st_mode)) {
2839 *dirstream = opendir(fname);
2842 fd = dirfd(*dirstream);
2843 } else if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) {
2844 fd = open(fname, open_flags);
2847 * we set this on purpose, in case the caller output
2848 * strerror(errno) as success
2856 closedir(*dirstream);
2863 int open_file_or_dir(const char *fname, DIR **dirstream)
2865 return open_file_or_dir3(fname, dirstream, O_RDWR);
2868 void close_file_or_dir(int fd, DIR *dirstream)
2871 closedir(dirstream);
2876 int get_device_info(int fd, u64 devid,
2877 struct btrfs_ioctl_dev_info_args *di_args)
2881 di_args->devid = devid;
2882 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
2884 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
2885 return ret < 0 ? -errno : 0;
2888 static u64 find_max_device_id(struct btrfs_ioctl_search_args *search_args,
2891 struct btrfs_dev_item *dev_item;
2892 char *buf = search_args->buf;
2894 buf += (nr_items - 1) * (sizeof(struct btrfs_ioctl_search_header)
2895 + sizeof(struct btrfs_dev_item));
2896 buf += sizeof(struct btrfs_ioctl_search_header);
2898 dev_item = (struct btrfs_dev_item *)buf;
2900 return btrfs_stack_device_id(dev_item);
2903 static int search_chunk_tree_for_fs_info(int fd,
2904 struct btrfs_ioctl_fs_info_args *fi_args)
2908 u64 start_devid = 1;
2909 struct btrfs_ioctl_search_args search_args;
2910 struct btrfs_ioctl_search_key *search_key = &search_args.key;
2912 fi_args->num_devices = 0;
2914 max_items = BTRFS_SEARCH_ARGS_BUFSIZE
2915 / (sizeof(struct btrfs_ioctl_search_header)
2916 + sizeof(struct btrfs_dev_item));
2918 search_key->tree_id = BTRFS_CHUNK_TREE_OBJECTID;
2919 search_key->min_objectid = BTRFS_DEV_ITEMS_OBJECTID;
2920 search_key->max_objectid = BTRFS_DEV_ITEMS_OBJECTID;
2921 search_key->min_type = BTRFS_DEV_ITEM_KEY;
2922 search_key->max_type = BTRFS_DEV_ITEM_KEY;
2923 search_key->min_transid = 0;
2924 search_key->max_transid = (u64)-1;
2925 search_key->nr_items = max_items;
2926 search_key->max_offset = (u64)-1;
2929 search_key->min_offset = start_devid;
2931 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH, &search_args);
2935 fi_args->num_devices += (u64)search_key->nr_items;
2937 if (search_key->nr_items == max_items) {
2938 start_devid = find_max_device_id(&search_args,
2939 search_key->nr_items) + 1;
2943 /* get the lastest max_id to stay consistent with the num_devices */
2944 if (search_key->nr_items == 0)
2946 * last tree_search returns an empty buf, use the devid of
2947 * the last dev_item of the previous tree_search
2949 fi_args->max_id = start_devid - 1;
2951 fi_args->max_id = find_max_device_id(&search_args,
2952 search_key->nr_items);
2958 * For a given path, fill in the ioctl fs_ and info_ args.
2959 * If the path is a btrfs mountpoint, fill info for all devices.
2960 * If the path is a btrfs device, fill in only that device.
2962 * The path provided must be either on a mounted btrfs fs,
2963 * or be a mounted btrfs device.
2965 * Returns 0 on success, or a negative errno.
2967 int get_fs_info(char *path, struct btrfs_ioctl_fs_info_args *fi_args,
2968 struct btrfs_ioctl_dev_info_args **di_ret)
2975 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2976 struct btrfs_ioctl_dev_info_args *di_args;
2977 struct btrfs_ioctl_dev_info_args tmp;
2979 DIR *dirstream = NULL;
2981 memset(fi_args, 0, sizeof(*fi_args));
2983 if (is_block_device(path) == 1) {
2984 struct btrfs_super_block *disk_super;
2985 char buf[BTRFS_SUPER_INFO_SIZE];
2988 /* Ensure it's mounted, then set path to the mountpoint */
2989 fd = open(path, O_RDONLY);
2992 error("cannot open %s: %s", path, strerror(errno));
2995 ret = check_mounted_where(fd, path, mp, sizeof(mp),
3004 /* Only fill in this one device */
3005 fi_args->num_devices = 1;
3007 disk_super = (struct btrfs_super_block *)buf;
3008 ret = btrfs_read_dev_super(fd, disk_super,
3009 BTRFS_SUPER_INFO_OFFSET, 0);
3014 devid = btrfs_stack_device_id(&disk_super->dev_item);
3016 fi_args->max_id = devid;
3019 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
3023 /* at this point path must not be for a block device */
3024 fd = open_file_or_dir(path, &dirstream);
3030 /* fill in fi_args if not just a single device */
3031 if (fi_args->num_devices != 1) {
3032 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
3039 * The fs_args->num_devices does not include seed devices
3041 ret = search_chunk_tree_for_fs_info(fd, fi_args);
3046 * search_chunk_tree_for_fs_info() will lacks the devid 0
3047 * so manual probe for it here.
3049 ret = get_device_info(fd, 0, &tmp);
3051 fi_args->num_devices++;
3059 if (!fi_args->num_devices)
3062 di_args = *di_ret = malloc((fi_args->num_devices) * sizeof(*di_args));
3069 memcpy(di_args, &tmp, sizeof(tmp));
3070 for (; i <= fi_args->max_id; ++i) {
3071 ret = get_device_info(fd, i, &di_args[ndevs]);
3080 * only when the only dev we wanted to find is not there then
3081 * let any error be returned
3083 if (fi_args->num_devices != 1) {
3089 close_file_or_dir(fd, dirstream);
3093 #define isoctal(c) (((c) & ~7) == '0')
3095 static inline void translate(char *f, char *t)
3097 while (*f != '\0') {
3099 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
3100 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
3110 * Checks if the swap device.
3111 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
3113 static int is_swap_device(const char *file)
3124 if (stat(file, &st_buf) < 0)
3126 if (S_ISBLK(st_buf.st_mode))
3127 dev = st_buf.st_rdev;
3128 else if (S_ISREG(st_buf.st_mode)) {
3129 dev = st_buf.st_dev;
3130 ino = st_buf.st_ino;
3134 if ((f = fopen("/proc/swaps", "r")) == NULL)
3137 /* skip the first line */
3138 if (fgets(tmp, sizeof(tmp), f) == NULL)
3141 while (fgets(tmp, sizeof(tmp), f) != NULL) {
3142 if ((cp = strchr(tmp, ' ')) != NULL)
3144 if ((cp = strchr(tmp, '\t')) != NULL)
3146 translate(tmp, buf);
3147 if (stat(buf, &st_buf) != 0)
3149 if (S_ISBLK(st_buf.st_mode)) {
3150 if (dev == st_buf.st_rdev) {
3154 } else if (S_ISREG(st_buf.st_mode)) {
3155 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {
3169 * Check for existing filesystem or partition table on device.
3171 * 1 for existing fs or partition
3172 * 0 for nothing found
3173 * -1 for internal error
3175 static int check_overwrite(const char *device)
3178 blkid_probe pr = NULL;
3182 if (!device || !*device)
3185 ret = -1; /* will reset on success of all setup calls */
3187 pr = blkid_new_probe_from_filename(device);
3191 size = blkid_probe_get_size(pr);
3195 /* nothing to overwrite on a 0-length device */
3201 ret = blkid_probe_enable_partitions(pr, 1);
3205 ret = blkid_do_fullprobe(pr);
3210 * Blkid returns 1 for nothing found and 0 when it finds a signature,
3211 * but we want the exact opposite, so reverse the return value here.
3213 * In addition print some useful diagnostics about what actually is
3221 if (!blkid_probe_lookup_value(pr, "TYPE", &type, NULL)) {
3223 "%s appears to contain an existing "
3224 "filesystem (%s).\n", device, type);
3225 } else if (!blkid_probe_lookup_value(pr, "PTTYPE", &type, NULL)) {
3227 "%s appears to contain a partition "
3228 "table (%s).\n", device, type);
3231 "%s appears to contain something weird "
3232 "according to blkid\n", device);
3238 blkid_free_probe(pr);
3241 "probe of %s failed, cannot detect "
3242 "existing filesystem.\n", device);
3246 static int group_profile_devs_min(u64 flag)
3248 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3249 case 0: /* single */
3250 case BTRFS_BLOCK_GROUP_DUP:
3252 case BTRFS_BLOCK_GROUP_RAID0:
3253 case BTRFS_BLOCK_GROUP_RAID1:
3254 case BTRFS_BLOCK_GROUP_RAID5:
3256 case BTRFS_BLOCK_GROUP_RAID6:
3258 case BTRFS_BLOCK_GROUP_RAID10:
3265 int test_num_disk_vs_raid(u64 metadata_profile, u64 data_profile,
3266 u64 dev_cnt, int mixed, int ssd)
3273 allowed |= BTRFS_BLOCK_GROUP_RAID10;
3275 allowed |= BTRFS_BLOCK_GROUP_RAID6;
3277 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
3278 BTRFS_BLOCK_GROUP_RAID5;
3280 allowed |= BTRFS_BLOCK_GROUP_DUP;
3284 ((metadata_profile | data_profile) & BTRFS_BLOCK_GROUP_DUP)) {
3285 warning("DUP is not recommended on filesystem with multiple devices");
3287 if (metadata_profile & ~allowed) {
3289 "ERROR: unable to create FS with metadata profile %s "
3290 "(have %llu devices but %d devices are required)\n",
3291 btrfs_group_profile_str(metadata_profile), dev_cnt,
3292 group_profile_devs_min(metadata_profile));
3295 if (data_profile & ~allowed) {
3297 "ERROR: unable to create FS with data profile %s "
3298 "(have %llu devices but %d devices are required)\n",
3299 btrfs_group_profile_str(data_profile), dev_cnt,
3300 group_profile_devs_min(data_profile));
3304 warning_on(!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP) && ssd,
3305 "DUP may not actually lead to 2 copies on the device, see manual page");
3310 int group_profile_max_safe_loss(u64 flags)
3312 switch (flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3313 case 0: /* single */
3314 case BTRFS_BLOCK_GROUP_DUP:
3315 case BTRFS_BLOCK_GROUP_RAID0:
3317 case BTRFS_BLOCK_GROUP_RAID1:
3318 case BTRFS_BLOCK_GROUP_RAID5:
3319 case BTRFS_BLOCK_GROUP_RAID10:
3321 case BTRFS_BLOCK_GROUP_RAID6:
3329 * Check if a device is suitable for btrfs
3331 * 1: something is wrong, an error is printed
3334 int test_dev_for_mkfs(const char *file, int force_overwrite)
3339 ret = is_swap_device(file);
3341 error("checking status of %s: %s", file, strerror(-ret));
3345 error("%s is a swap device", file);
3348 if (!force_overwrite) {
3349 if (check_overwrite(file)) {
3350 error("use the -f option to force overwrite of %s",
3355 ret = check_mounted(file);
3357 error("cannot check mount status of %s: %s", file,
3362 error("%s is mounted", file);
3365 /* check if the device is busy */
3366 fd = open(file, O_RDWR|O_EXCL);
3368 error("unable to open %s: %s", file, strerror(errno));
3371 if (fstat(fd, &st)) {
3372 error("unable to stat %s: %s", file, strerror(errno));
3376 if (!S_ISBLK(st.st_mode)) {
3377 error("%s is not a block device", file);
3385 int btrfs_scan_lblkid(void)
3390 struct btrfs_fs_devices *tmp_devices;
3391 blkid_dev_iterate iter = NULL;
3392 blkid_dev dev = NULL;
3393 blkid_cache cache = NULL;
3394 char path[PATH_MAX];
3396 if (btrfs_scan_done)
3399 if (blkid_get_cache(&cache, NULL) < 0) {
3400 error("blkid cache get failed");
3403 blkid_probe_all(cache);
3404 iter = blkid_dev_iterate_begin(cache);
3405 blkid_dev_set_search(iter, "TYPE", "btrfs");
3406 while (blkid_dev_next(iter, &dev) == 0) {
3407 dev = blkid_verify(cache, dev);
3410 /* if we are here its definitely a btrfs disk*/
3411 strncpy_null(path, blkid_dev_devname(dev));
3413 fd = open(path, O_RDONLY);
3415 error("cannot open %s: %s", path, strerror(errno));
3418 ret = btrfs_scan_one_device(fd, path, &tmp_devices,
3419 &num_devices, BTRFS_SUPER_INFO_OFFSET, 0);
3421 error("cannot scan %s: %s", path, strerror(-ret));
3428 blkid_dev_iterate_end(iter);
3429 blkid_put_cache(cache);
3431 btrfs_scan_done = 1;
3436 int is_vol_small(const char *file)
3443 fd = open(file, O_RDONLY);
3446 if (fstat(fd, &st) < 0) {
3451 size = btrfs_device_size(fd, &st);
3456 if (size < BTRFS_MKFS_SMALL_VOLUME_SIZE) {
3466 * This reads a line from the stdin and only returns non-zero if the
3467 * first whitespace delimited token is a case insensitive match with yes
3470 int ask_user(const char *question)
3472 char buf[30] = {0,};
3473 char *saveptr = NULL;
3476 printf("%s [y/N]: ", question);
3478 return fgets(buf, sizeof(buf) - 1, stdin) &&
3479 (answer = strtok_r(buf, " \t\n\r", &saveptr)) &&
3480 (!strcasecmp(answer, "yes") || !strcasecmp(answer, "y"));
3485 * - file or directory return the containing tree root id
3486 * - subvolume return its own tree id
3487 * - BTRFS_EMPTY_SUBVOL_DIR_OBJECTID (directory with ino == 2) the result is
3488 * undefined and function returns -1
3490 int lookup_ino_rootid(int fd, u64 *rootid)
3492 struct btrfs_ioctl_ino_lookup_args args;
3495 memset(&args, 0, sizeof(args));
3497 args.objectid = BTRFS_FIRST_FREE_OBJECTID;
3499 ret = ioctl(fd, BTRFS_IOC_INO_LOOKUP, &args);
3501 error("failed to lookup root id: %s", strerror(errno));
3505 *rootid = args.treeid;
3511 * return 0 if a btrfs mount point is found
3512 * return 1 if a mount point is found but not btrfs
3513 * return <0 if something goes wrong
3515 int find_mount_root(const char *path, char **mount_root)
3523 int longest_matchlen = 0;
3524 char *longest_match = NULL;
3526 fd = open(path, O_RDONLY | O_NOATIME);
3531 mnttab = setmntent("/proc/self/mounts", "r");
3535 while ((ent = getmntent(mnttab))) {
3536 len = strlen(ent->mnt_dir);
3537 if (strncmp(ent->mnt_dir, path, len) == 0) {
3538 /* match found and use the latest match */
3539 if (longest_matchlen <= len) {
3540 free(longest_match);
3541 longest_matchlen = len;
3542 longest_match = strdup(ent->mnt_dir);
3543 not_btrfs = strcmp(ent->mnt_type, "btrfs");
3552 free(longest_match);
3557 *mount_root = realpath(longest_match, NULL);
3561 free(longest_match);
3565 int test_minimum_size(const char *file, u32 nodesize)
3568 struct stat statbuf;
3570 fd = open(file, O_RDONLY);
3573 if (stat(file, &statbuf) < 0) {
3577 if (btrfs_device_size(fd, &statbuf) < btrfs_min_dev_size(nodesize)) {
3587 * Test if path is a directory
3589 * 0 - path exists but it is not a directory
3590 * 1 - path exists and it is a directory
3593 int test_isdir(const char *path)
3598 ret = stat(path, &st);
3602 return !!S_ISDIR(st.st_mode);
3605 void units_set_mode(unsigned *units, unsigned mode)
3607 unsigned base = *units & UNITS_MODE_MASK;
3609 *units = base | mode;
3612 void units_set_base(unsigned *units, unsigned base)
3614 unsigned mode = *units & ~UNITS_MODE_MASK;
3616 *units = base | mode;
3619 int find_next_key(struct btrfs_path *path, struct btrfs_key *key)
3623 for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
3624 if (!path->nodes[level])
3626 if (path->slots[level] + 1 >=
3627 btrfs_header_nritems(path->nodes[level]))
3630 btrfs_item_key_to_cpu(path->nodes[level], key,
3631 path->slots[level] + 1);
3633 btrfs_node_key_to_cpu(path->nodes[level], key,
3634 path->slots[level] + 1);
3640 const char* btrfs_group_type_str(u64 flag)
3642 u64 mask = BTRFS_BLOCK_GROUP_TYPE_MASK |
3643 BTRFS_SPACE_INFO_GLOBAL_RSV;
3645 switch (flag & mask) {
3646 case BTRFS_BLOCK_GROUP_DATA:
3648 case BTRFS_BLOCK_GROUP_SYSTEM:
3650 case BTRFS_BLOCK_GROUP_METADATA:
3652 case BTRFS_BLOCK_GROUP_DATA|BTRFS_BLOCK_GROUP_METADATA:
3653 return "Data+Metadata";
3654 case BTRFS_SPACE_INFO_GLOBAL_RSV:
3655 return "GlobalReserve";
3661 const char* btrfs_group_profile_str(u64 flag)
3663 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3666 case BTRFS_BLOCK_GROUP_RAID0:
3668 case BTRFS_BLOCK_GROUP_RAID1:
3670 case BTRFS_BLOCK_GROUP_RAID5:
3672 case BTRFS_BLOCK_GROUP_RAID6:
3674 case BTRFS_BLOCK_GROUP_DUP:
3676 case BTRFS_BLOCK_GROUP_RAID10:
3683 u64 disk_size(const char *path)
3687 if (statfs(path, &sfs) < 0)
3690 return sfs.f_bsize * sfs.f_blocks;
3693 u64 get_partition_size(const char *dev)
3696 int fd = open(dev, O_RDONLY);
3700 if (ioctl(fd, BLKGETSIZE64, &result) < 0) {
3709 int btrfs_tree_search2_ioctl_supported(int fd)
3711 struct btrfs_ioctl_search_args_v2 *args2;
3712 struct btrfs_ioctl_search_key *sk;
3713 int args2_size = 1024;
3714 char args2_buf[args2_size];
3716 static int v2_supported = -1;
3718 if (v2_supported != -1)
3719 return v2_supported;
3721 args2 = (struct btrfs_ioctl_search_args_v2 *)args2_buf;
3725 * Search for the extent tree item in the root tree.
3727 sk->tree_id = BTRFS_ROOT_TREE_OBJECTID;
3728 sk->min_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3729 sk->max_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3730 sk->min_type = BTRFS_ROOT_ITEM_KEY;
3731 sk->max_type = BTRFS_ROOT_ITEM_KEY;
3733 sk->max_offset = (u64)-1;
3734 sk->min_transid = 0;
3735 sk->max_transid = (u64)-1;
3737 args2->buf_size = args2_size - sizeof(struct btrfs_ioctl_search_args_v2);
3738 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH_V2, args2);
3739 if (ret == -EOPNOTSUPP)
3746 return v2_supported;
3749 int btrfs_check_nodesize(u32 nodesize, u32 sectorsize, u64 features)
3751 if (nodesize < sectorsize) {
3752 error("illegal nodesize %u (smaller than %u)",
3753 nodesize, sectorsize);
3755 } else if (nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) {
3756 error("illegal nodesize %u (larger than %u)",
3757 nodesize, BTRFS_MAX_METADATA_BLOCKSIZE);
3759 } else if (nodesize & (sectorsize - 1)) {
3760 error("illegal nodesize %u (not aligned to %u)",
3761 nodesize, sectorsize);
3763 } else if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS &&
3764 nodesize != sectorsize) {
3765 error("illegal nodesize %u (not equal to %u for mixed block group)",
3766 nodesize, sectorsize);
3773 * Copy a path argument from SRC to DEST and check the SRC length if it's at
3774 * most PATH_MAX and fits into DEST. DESTLEN is supposed to be exact size of
3776 * The destination buffer is zero terminated.
3777 * Return < 0 for error, 0 otherwise.
3779 int arg_copy_path(char *dest, const char *src, int destlen)
3781 size_t len = strlen(src);
3783 if (len >= PATH_MAX || len >= destlen)
3784 return -ENAMETOOLONG;
3786 __strncpy_null(dest, src, destlen);
3791 unsigned int get_unit_mode_from_arg(int *argc, char *argv[], int df_mode)
3793 unsigned int unit_mode = UNITS_DEFAULT;
3797 for (arg_i = 0; arg_i < *argc; arg_i++) {
3798 if (!strcmp(argv[arg_i], "--"))
3801 if (!strcmp(argv[arg_i], "--raw")) {
3802 unit_mode = UNITS_RAW;
3806 if (!strcmp(argv[arg_i], "--human-readable")) {
3807 unit_mode = UNITS_HUMAN_BINARY;
3812 if (!strcmp(argv[arg_i], "--iec")) {
3813 units_set_mode(&unit_mode, UNITS_BINARY);
3817 if (!strcmp(argv[arg_i], "--si")) {
3818 units_set_mode(&unit_mode, UNITS_DECIMAL);
3823 if (!strcmp(argv[arg_i], "--kbytes")) {
3824 units_set_base(&unit_mode, UNITS_KBYTES);
3828 if (!strcmp(argv[arg_i], "--mbytes")) {
3829 units_set_base(&unit_mode, UNITS_MBYTES);
3833 if (!strcmp(argv[arg_i], "--gbytes")) {
3834 units_set_base(&unit_mode, UNITS_GBYTES);
3838 if (!strcmp(argv[arg_i], "--tbytes")) {
3839 units_set_base(&unit_mode, UNITS_TBYTES);
3847 if (!strcmp(argv[arg_i], "-b")) {
3848 unit_mode = UNITS_RAW;
3852 if (!strcmp(argv[arg_i], "-h")) {
3853 unit_mode = UNITS_HUMAN_BINARY;
3857 if (!strcmp(argv[arg_i], "-H")) {
3858 unit_mode = UNITS_HUMAN_DECIMAL;
3862 if (!strcmp(argv[arg_i], "-k")) {
3863 units_set_base(&unit_mode, UNITS_KBYTES);
3867 if (!strcmp(argv[arg_i], "-m")) {
3868 units_set_base(&unit_mode, UNITS_MBYTES);
3872 if (!strcmp(argv[arg_i], "-g")) {
3873 units_set_base(&unit_mode, UNITS_GBYTES);
3877 if (!strcmp(argv[arg_i], "-t")) {
3878 units_set_base(&unit_mode, UNITS_TBYTES);
3884 for (arg_i = 0, arg_end = 0; arg_i < *argc; arg_i++) {
3887 argv[arg_end] = argv[arg_i];
3896 int string_is_numerical(const char *str)
3898 if (!(*str >= '0' && *str <= '9'))
3900 while (*str >= '0' && *str <= '9')
3908 * Preprocess @argv with getopt_long to reorder options and consume the "--"
3910 * Unknown short and long options are reported, optionally the @usage is printed
3913 void clean_args_no_options(int argc, char *argv[], const char * const *usagestr)
3915 static const struct option long_options[] = {
3920 int c = getopt_long(argc, argv, "", long_options, NULL);
3933 /* Subvolume helper functions */
3935 * test if name is a correct subvolume name
3936 * this function return
3937 * 0-> name is not a correct subvolume name
3938 * 1-> name is a correct subvolume name
3940 int test_issubvolname(const char *name)
3942 return name[0] != '\0' && !strchr(name, '/') &&
3943 strcmp(name, ".") && strcmp(name, "..");
3947 * Test if path is a subvolume
3949 * 0 - path exists but it is not a subvolume
3950 * 1 - path exists and it is a subvolume
3953 int test_issubvolume(const char *path)
3959 res = stat(path, &st);
3963 if (st.st_ino != BTRFS_FIRST_FREE_OBJECTID || !S_ISDIR(st.st_mode))
3966 res = statfs(path, &stfs);
3970 return (int)stfs.f_type == BTRFS_SUPER_MAGIC;
3973 const char *subvol_strip_mountpoint(const char *mnt, const char *full_path)
3975 int len = strlen(mnt);
3979 if (mnt[len - 1] != '/')
3982 return full_path + len;
3989 * 1: Error; and error info printed to the terminal. Fixme.
3990 * 2: If the fullpath is root tree instead of subvol tree
3992 int get_subvol_info(const char *fullpath, struct root_info *get_ri)
3999 const char *svpath = NULL;
4000 DIR *dirstream1 = NULL;
4001 DIR *dirstream2 = NULL;
4003 ret = test_issubvolume(fullpath);
4007 error("not a subvolume: %s", fullpath);
4011 ret = find_mount_root(fullpath, &mnt);
4015 error("%s doesn't belong to btrfs mount point", fullpath);
4019 svpath = subvol_strip_mountpoint(mnt, fullpath);
4021 fd = btrfs_open_dir(fullpath, &dirstream1, 1);
4025 ret = btrfs_list_get_path_rootid(fd, &sv_id);
4027 error("can't get rootid for '%s'", fullpath);
4031 mntfd = btrfs_open_dir(mnt, &dirstream2, 1);
4035 if (sv_id == BTRFS_FS_TREE_OBJECTID) {
4038 * So that caller may decide if thats an error or just fine.
4043 memset(get_ri, 0, sizeof(*get_ri));
4044 get_ri->root_id = sv_id;
4046 ret = btrfs_get_subvol(mntfd, get_ri);
4048 error("can't find '%s': %d", svpath, ret);
4051 close_file_or_dir(mntfd, dirstream2);
4052 close_file_or_dir(fd, dirstream1);
4058 void init_rand_seed(u64 seed)
4062 /* only use the last 48 bits */
4063 for (i = 0; i < 3; i++) {
4064 rand_seed[i] = (unsigned short)(seed ^ (unsigned short)(-1));
4067 rand_seed_initlized = 1;
4070 static void __init_seed(void)
4076 if(rand_seed_initlized)
4078 /* Use urandom as primary seed source. */
4079 fd = open("/dev/urandom", O_RDONLY);
4081 ret = read(fd, rand_seed, sizeof(rand_seed));
4083 if (ret < sizeof(rand_seed))
4087 /* Use time and pid as fallback seed */
4088 warning("failed to read /dev/urandom, use time and pid as random seed");
4089 gettimeofday(&tv, 0);
4090 rand_seed[0] = getpid() ^ (tv.tv_sec & 0xFFFF);
4091 rand_seed[1] = getppid() ^ (tv.tv_usec & 0xFFFF);
4092 rand_seed[2] = (tv.tv_sec ^ tv.tv_usec) >> 16;
4094 rand_seed_initlized = 1;
4101 * Don't use nrand48, its range is [0,2^31) The highest bit will alwasy
4102 * be 0. Use jrand48 to include the highest bit.
4104 return (u32)jrand48(rand_seed);
4107 unsigned int rand_range(unsigned int upper)
4111 * Use the full 48bits to mod, which would be more uniformly
4114 return (unsigned int)(jrand48(rand_seed) % upper);