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>
41 #include <sys/utsname.h>
42 #include <linux/version.h>
44 #include "kerncompat.h"
45 #include "radix-tree.h"
48 #include "transaction.h"
56 #define BLKDISCARD _IO(0x12,119)
59 static int btrfs_scan_done = 0;
61 static char argv0_buf[ARGV0_BUF_SIZE] = "btrfs";
63 static int rand_seed_initlized = 0;
64 static unsigned short rand_seed[3];
66 const char *get_argv0_buf(void)
71 void fixup_argv0(char **argv, const char *token)
73 int len = strlen(argv0_buf);
75 snprintf(argv0_buf + len, sizeof(argv0_buf) - len, " %s", token);
79 void set_argv0(char **argv)
81 strncpy(argv0_buf, argv[0], sizeof(argv0_buf));
82 argv0_buf[sizeof(argv0_buf) - 1] = 0;
85 int check_argc_exact(int nargs, int expected)
88 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
90 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
92 return nargs != expected;
95 int check_argc_min(int nargs, int expected)
97 if (nargs < expected) {
98 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
105 int check_argc_max(int nargs, int expected)
107 if (nargs > expected) {
108 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
117 * Discard the given range in one go
119 static int discard_range(int fd, u64 start, u64 len)
121 u64 range[2] = { start, len };
123 if (ioctl(fd, BLKDISCARD, &range) < 0)
129 * Discard blocks in the given range in 1G chunks, the process is interruptible
131 static int discard_blocks(int fd, u64 start, u64 len)
135 u64 chunk_size = min_t(u64, len, 1*1024*1024*1024);
138 ret = discard_range(fd, start, chunk_size);
148 static u64 reference_root_table[] = {
149 [1] = BTRFS_ROOT_TREE_OBJECTID,
150 [2] = BTRFS_EXTENT_TREE_OBJECTID,
151 [3] = BTRFS_CHUNK_TREE_OBJECTID,
152 [4] = BTRFS_DEV_TREE_OBJECTID,
153 [5] = BTRFS_FS_TREE_OBJECTID,
154 [6] = BTRFS_CSUM_TREE_OBJECTID,
157 int test_uuid_unique(char *fs_uuid)
160 blkid_dev_iterate iter = NULL;
161 blkid_dev dev = NULL;
162 blkid_cache cache = NULL;
164 if (blkid_get_cache(&cache, NULL) < 0) {
165 printf("ERROR: lblkid cache get failed\n");
168 blkid_probe_all(cache);
169 iter = blkid_dev_iterate_begin(cache);
170 blkid_dev_set_search(iter, "UUID", fs_uuid);
172 while (blkid_dev_next(iter, &dev) == 0) {
173 dev = blkid_verify(cache, dev);
180 blkid_dev_iterate_end(iter);
181 blkid_put_cache(cache);
187 * Reserve space from free_tree.
188 * The algorithm is very simple, find the first cache_extent with enough space
189 * and allocate from its beginning.
191 static int reserve_free_space(struct cache_tree *free_tree, u64 len,
194 struct cache_extent *cache;
197 ASSERT(ret_start != NULL);
198 cache = first_cache_extent(free_tree);
200 if (cache->size > len) {
202 *ret_start = cache->start;
205 if (cache->size == 0) {
206 remove_cache_extent(free_tree, cache);
213 cache = next_cache_extent(cache);
220 static inline int write_temp_super(int fd, struct btrfs_super_block *sb,
226 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
227 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
228 btrfs_csum_final(crc, &sb->csum[0]);
229 ret = pwrite(fd, sb, BTRFS_SUPER_INFO_SIZE, sb_bytenr);
230 if (ret < BTRFS_SUPER_INFO_SIZE)
231 ret = (ret < 0 ? -errno : -EIO);
238 * Setup temporary superblock at cfg->super_bynter
239 * Needed info are extracted from cfg, and root_bytenr, chunk_bytenr
241 * For now sys chunk array will be empty and dev_item is empty too.
242 * They will be re-initialized at temp chunk tree setup.
244 * The superblock signature is not valid, denotes a partially created
245 * filesystem, needs to be finalized.
247 static int setup_temp_super(int fd, struct btrfs_mkfs_config *cfg,
248 u64 root_bytenr, u64 chunk_bytenr)
250 unsigned char chunk_uuid[BTRFS_UUID_SIZE];
251 char super_buf[BTRFS_SUPER_INFO_SIZE];
252 struct btrfs_super_block *super = (struct btrfs_super_block *)super_buf;
255 memset(super_buf, 0, BTRFS_SUPER_INFO_SIZE);
256 cfg->num_bytes = round_down(cfg->num_bytes, cfg->sectorsize);
259 if (uuid_parse(cfg->fs_uuid, super->fsid) != 0) {
260 error("cound not parse UUID: %s", cfg->fs_uuid);
264 if (!test_uuid_unique(cfg->fs_uuid)) {
265 error("non-unique UUID: %s", cfg->fs_uuid);
270 uuid_generate(super->fsid);
271 uuid_unparse(super->fsid, cfg->fs_uuid);
273 uuid_generate(chunk_uuid);
274 uuid_unparse(chunk_uuid, cfg->chunk_uuid);
276 btrfs_set_super_bytenr(super, cfg->super_bytenr);
277 btrfs_set_super_num_devices(super, 1);
278 btrfs_set_super_magic(super, BTRFS_MAGIC_PARTIAL);
279 btrfs_set_super_generation(super, 1);
280 btrfs_set_super_root(super, root_bytenr);
281 btrfs_set_super_chunk_root(super, chunk_bytenr);
282 btrfs_set_super_total_bytes(super, cfg->num_bytes);
284 * Temporary filesystem will only have 6 tree roots:
285 * chunk tree, root tree, extent_tree, device tree, fs tree
288 btrfs_set_super_bytes_used(super, 6 * cfg->nodesize);
289 btrfs_set_super_sectorsize(super, cfg->sectorsize);
290 btrfs_set_super_leafsize(super, cfg->nodesize);
291 btrfs_set_super_nodesize(super, cfg->nodesize);
292 btrfs_set_super_stripesize(super, cfg->stripesize);
293 btrfs_set_super_csum_type(super, BTRFS_CSUM_TYPE_CRC32);
294 btrfs_set_super_chunk_root(super, chunk_bytenr);
295 btrfs_set_super_cache_generation(super, -1);
296 btrfs_set_super_incompat_flags(super, cfg->features);
298 __strncpy_null(super->label, cfg->label, BTRFS_LABEL_SIZE - 1);
300 /* Sys chunk array will be re-initialized at chunk tree init time */
301 super->sys_chunk_array_size = 0;
303 ret = write_temp_super(fd, super, cfg->super_bytenr);
309 * Setup an extent buffer for tree block.
311 static int setup_temp_extent_buffer(struct extent_buffer *buf,
312 struct btrfs_mkfs_config *cfg,
313 u64 bytenr, u64 owner)
315 unsigned char fsid[BTRFS_FSID_SIZE];
316 unsigned char chunk_uuid[BTRFS_UUID_SIZE];
319 ret = uuid_parse(cfg->fs_uuid, fsid);
322 ret = uuid_parse(cfg->chunk_uuid, chunk_uuid);
326 memset(buf->data, 0, cfg->nodesize);
327 buf->len = cfg->nodesize;
328 btrfs_set_header_bytenr(buf, bytenr);
329 btrfs_set_header_generation(buf, 1);
330 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
331 btrfs_set_header_owner(buf, owner);
332 btrfs_set_header_flags(buf, BTRFS_HEADER_FLAG_WRITTEN);
333 write_extent_buffer(buf, chunk_uuid, btrfs_header_chunk_tree_uuid(buf),
335 write_extent_buffer(buf, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
339 static inline int write_temp_extent_buffer(int fd, struct extent_buffer *buf,
344 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
346 /* Temporary extent buffer is always mapped 1:1 on disk */
347 ret = pwrite(fd, buf->data, buf->len, bytenr);
349 ret = (ret < 0 ? ret : -EIO);
356 * Insert a root item for temporary tree root
358 * Only used in make_btrfs_v2().
360 static void insert_temp_root_item(struct extent_buffer *buf,
361 struct btrfs_mkfs_config *cfg,
362 int *slot, u32 *itemoff, u64 objectid,
365 struct btrfs_root_item root_item;
366 struct btrfs_inode_item *inode_item;
367 struct btrfs_disk_key disk_key;
369 btrfs_set_header_nritems(buf, *slot + 1);
370 (*itemoff) -= sizeof(root_item);
371 memset(&root_item, 0, sizeof(root_item));
372 inode_item = &root_item.inode;
373 btrfs_set_stack_inode_generation(inode_item, 1);
374 btrfs_set_stack_inode_size(inode_item, 3);
375 btrfs_set_stack_inode_nlink(inode_item, 1);
376 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
377 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
378 btrfs_set_root_refs(&root_item, 1);
379 btrfs_set_root_used(&root_item, cfg->nodesize);
380 btrfs_set_root_generation(&root_item, 1);
381 btrfs_set_root_bytenr(&root_item, bytenr);
383 memset(&disk_key, 0, sizeof(disk_key));
384 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
385 btrfs_set_disk_key_objectid(&disk_key, objectid);
386 btrfs_set_disk_key_offset(&disk_key, 0);
388 btrfs_set_item_key(buf, &disk_key, *slot);
389 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
390 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(root_item));
391 write_extent_buffer(buf, &root_item,
392 btrfs_item_ptr_offset(buf, *slot),
397 static int setup_temp_root_tree(int fd, struct btrfs_mkfs_config *cfg,
398 u64 root_bytenr, u64 extent_bytenr,
399 u64 dev_bytenr, u64 fs_bytenr, u64 csum_bytenr)
401 struct extent_buffer *buf = NULL;
402 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
407 * Provided bytenr must in ascending order, or tree root will have a
410 if (!(root_bytenr < extent_bytenr && extent_bytenr < dev_bytenr &&
411 dev_bytenr < fs_bytenr && fs_bytenr < csum_bytenr)) {
412 error("bad tree bytenr order: "
413 "root < extent %llu < %llu, "
414 "extent < dev %llu < %llu, "
415 "dev < fs %llu < %llu, "
416 "fs < csum %llu < %llu",
417 (unsigned long long)root_bytenr,
418 (unsigned long long)extent_bytenr,
419 (unsigned long long)extent_bytenr,
420 (unsigned long long)dev_bytenr,
421 (unsigned long long)dev_bytenr,
422 (unsigned long long)fs_bytenr,
423 (unsigned long long)fs_bytenr,
424 (unsigned long long)csum_bytenr);
427 buf = malloc(sizeof(*buf) + cfg->nodesize);
431 ret = setup_temp_extent_buffer(buf, cfg, root_bytenr,
432 BTRFS_ROOT_TREE_OBJECTID);
436 insert_temp_root_item(buf, cfg, &slot, &itemoff,
437 BTRFS_EXTENT_TREE_OBJECTID, extent_bytenr);
438 insert_temp_root_item(buf, cfg, &slot, &itemoff,
439 BTRFS_DEV_TREE_OBJECTID, dev_bytenr);
440 insert_temp_root_item(buf, cfg, &slot, &itemoff,
441 BTRFS_FS_TREE_OBJECTID, fs_bytenr);
442 insert_temp_root_item(buf, cfg, &slot, &itemoff,
443 BTRFS_CSUM_TREE_OBJECTID, csum_bytenr);
445 ret = write_temp_extent_buffer(fd, buf, root_bytenr);
451 static int insert_temp_dev_item(int fd, struct extent_buffer *buf,
452 struct btrfs_mkfs_config *cfg,
453 int *slot, u32 *itemoff)
455 struct btrfs_disk_key disk_key;
456 struct btrfs_dev_item *dev_item;
457 char super_buf[BTRFS_SUPER_INFO_SIZE];
458 unsigned char dev_uuid[BTRFS_UUID_SIZE];
459 unsigned char fsid[BTRFS_FSID_SIZE];
460 struct btrfs_super_block *super = (struct btrfs_super_block *)super_buf;
463 ret = pread(fd, super_buf, BTRFS_SUPER_INFO_SIZE, cfg->super_bytenr);
464 if (ret < BTRFS_SUPER_INFO_SIZE) {
465 ret = (ret < 0 ? -errno : -EIO);
469 btrfs_set_header_nritems(buf, *slot + 1);
470 (*itemoff) -= sizeof(*dev_item);
471 /* setup device item 1, 0 is for replace case */
472 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
473 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
474 btrfs_set_disk_key_offset(&disk_key, 1);
475 btrfs_set_item_key(buf, &disk_key, *slot);
476 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
477 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(*dev_item));
479 dev_item = btrfs_item_ptr(buf, *slot, struct btrfs_dev_item);
480 /* Generate device uuid */
481 uuid_generate(dev_uuid);
482 write_extent_buffer(buf, dev_uuid,
483 (unsigned long)btrfs_device_uuid(dev_item),
485 uuid_parse(cfg->fs_uuid, fsid);
486 write_extent_buffer(buf, fsid,
487 (unsigned long)btrfs_device_fsid(dev_item),
489 btrfs_set_device_id(buf, dev_item, 1);
490 btrfs_set_device_generation(buf, dev_item, 0);
491 btrfs_set_device_total_bytes(buf, dev_item, cfg->num_bytes);
493 * The number must match the initial SYSTEM and META chunk size
495 btrfs_set_device_bytes_used(buf, dev_item,
496 BTRFS_MKFS_SYSTEM_GROUP_SIZE +
497 BTRFS_CONVERT_META_GROUP_SIZE);
498 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
499 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
500 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
501 btrfs_set_device_type(buf, dev_item, 0);
503 /* Super dev_item is not complete, copy the complete one to sb */
504 read_extent_buffer(buf, &super->dev_item, (unsigned long)dev_item,
506 ret = write_temp_super(fd, super, cfg->super_bytenr);
512 static int insert_temp_chunk_item(int fd, struct extent_buffer *buf,
513 struct btrfs_mkfs_config *cfg,
514 int *slot, u32 *itemoff, u64 start, u64 len,
517 struct btrfs_chunk *chunk;
518 struct btrfs_disk_key disk_key;
519 char super_buf[BTRFS_SUPER_INFO_SIZE];
520 struct btrfs_super_block *sb = (struct btrfs_super_block *)super_buf;
523 ret = pread(fd, super_buf, BTRFS_SUPER_INFO_SIZE,
525 if (ret < BTRFS_SUPER_INFO_SIZE) {
526 ret = (ret < 0 ? ret : -EIO);
530 btrfs_set_header_nritems(buf, *slot + 1);
531 (*itemoff) -= btrfs_chunk_item_size(1);
532 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
533 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
534 btrfs_set_disk_key_offset(&disk_key, start);
535 btrfs_set_item_key(buf, &disk_key, *slot);
536 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
537 btrfs_set_item_size(buf, btrfs_item_nr(*slot),
538 btrfs_chunk_item_size(1));
540 chunk = btrfs_item_ptr(buf, *slot, struct btrfs_chunk);
541 btrfs_set_chunk_length(buf, chunk, len);
542 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
543 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
544 btrfs_set_chunk_type(buf, chunk, type);
545 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
546 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
547 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
548 btrfs_set_chunk_num_stripes(buf, chunk, 1);
549 /* TODO: Support DUP profile for system chunk */
550 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
551 /* We are doing 1:1 mapping, so start is its dev offset */
552 btrfs_set_stripe_offset_nr(buf, chunk, 0, start);
553 write_extent_buffer(buf, &sb->dev_item.uuid,
554 (unsigned long)btrfs_stripe_dev_uuid_nr(chunk, 0),
559 * If it's system chunk, also copy it to super block.
561 if (type & BTRFS_BLOCK_GROUP_SYSTEM) {
565 cur = (char *)sb->sys_chunk_array
566 + btrfs_super_sys_array_size(sb);
567 memcpy(cur, &disk_key, sizeof(disk_key));
568 cur += sizeof(disk_key);
569 read_extent_buffer(buf, cur, (unsigned long int)chunk,
570 btrfs_chunk_item_size(1));
571 array_size = btrfs_super_sys_array_size(sb);
572 array_size += btrfs_chunk_item_size(1) +
574 btrfs_set_super_sys_array_size(sb, array_size);
576 ret = write_temp_super(fd, sb, cfg->super_bytenr);
581 static int setup_temp_chunk_tree(int fd, struct btrfs_mkfs_config *cfg,
582 u64 sys_chunk_start, u64 meta_chunk_start,
585 struct extent_buffer *buf = NULL;
586 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
590 /* Must ensure SYS chunk starts before META chunk */
591 if (meta_chunk_start < sys_chunk_start) {
592 error("wrong chunk order: meta < system %llu < %llu",
593 (unsigned long long)meta_chunk_start,
594 (unsigned long long)sys_chunk_start);
597 buf = malloc(sizeof(*buf) + cfg->nodesize);
600 ret = setup_temp_extent_buffer(buf, cfg, chunk_bytenr,
601 BTRFS_CHUNK_TREE_OBJECTID);
605 ret = insert_temp_dev_item(fd, buf, cfg, &slot, &itemoff);
608 ret = insert_temp_chunk_item(fd, buf, cfg, &slot, &itemoff,
610 BTRFS_MKFS_SYSTEM_GROUP_SIZE,
611 BTRFS_BLOCK_GROUP_SYSTEM);
614 ret = insert_temp_chunk_item(fd, buf, cfg, &slot, &itemoff,
616 BTRFS_CONVERT_META_GROUP_SIZE,
617 BTRFS_BLOCK_GROUP_METADATA);
620 ret = write_temp_extent_buffer(fd, buf, chunk_bytenr);
627 static void insert_temp_dev_extent(struct extent_buffer *buf,
628 int *slot, u32 *itemoff, u64 start, u64 len)
630 struct btrfs_dev_extent *dev_extent;
631 struct btrfs_disk_key disk_key;
633 btrfs_set_header_nritems(buf, *slot + 1);
634 (*itemoff) -= sizeof(*dev_extent);
635 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
636 btrfs_set_disk_key_objectid(&disk_key, 1);
637 btrfs_set_disk_key_offset(&disk_key, start);
638 btrfs_set_item_key(buf, &disk_key, *slot);
639 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
640 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(*dev_extent));
642 dev_extent = btrfs_item_ptr(buf, *slot, struct btrfs_dev_extent);
643 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
644 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
645 btrfs_set_dev_extent_length(buf, dev_extent, len);
646 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, start);
647 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
648 BTRFS_CHUNK_TREE_OBJECTID);
652 static int setup_temp_dev_tree(int fd, struct btrfs_mkfs_config *cfg,
653 u64 sys_chunk_start, u64 meta_chunk_start,
656 struct extent_buffer *buf = NULL;
657 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
661 /* Must ensure SYS chunk starts before META chunk */
662 if (meta_chunk_start < sys_chunk_start) {
663 error("wrong chunk order: meta < system %llu < %llu",
664 (unsigned long long)meta_chunk_start,
665 (unsigned long long)sys_chunk_start);
668 buf = malloc(sizeof(*buf) + cfg->nodesize);
671 ret = setup_temp_extent_buffer(buf, cfg, dev_bytenr,
672 BTRFS_DEV_TREE_OBJECTID);
675 insert_temp_dev_extent(buf, &slot, &itemoff, sys_chunk_start,
676 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
677 insert_temp_dev_extent(buf, &slot, &itemoff, meta_chunk_start,
678 BTRFS_CONVERT_META_GROUP_SIZE);
679 ret = write_temp_extent_buffer(fd, buf, dev_bytenr);
685 static int setup_temp_fs_tree(int fd, struct btrfs_mkfs_config *cfg,
688 struct extent_buffer *buf = NULL;
691 buf = malloc(sizeof(*buf) + cfg->nodesize);
694 ret = setup_temp_extent_buffer(buf, cfg, fs_bytenr,
695 BTRFS_FS_TREE_OBJECTID);
699 * Temporary fs tree is completely empty.
701 ret = write_temp_extent_buffer(fd, buf, fs_bytenr);
707 static int setup_temp_csum_tree(int fd, struct btrfs_mkfs_config *cfg,
710 struct extent_buffer *buf = NULL;
713 buf = malloc(sizeof(*buf) + cfg->nodesize);
716 ret = setup_temp_extent_buffer(buf, cfg, csum_bytenr,
717 BTRFS_CSUM_TREE_OBJECTID);
721 * Temporary csum tree is completely empty.
723 ret = write_temp_extent_buffer(fd, buf, csum_bytenr);
730 * Insert one temporary extent item.
732 * NOTE: if skinny_metadata is not enabled, this function must be called
733 * after all other trees are initialized.
734 * Or fs without skinny-metadata will be screwed up.
736 static int insert_temp_extent_item(int fd, struct extent_buffer *buf,
737 struct btrfs_mkfs_config *cfg,
738 int *slot, u32 *itemoff, u64 bytenr,
741 struct extent_buffer *tmp;
742 struct btrfs_extent_item *ei;
743 struct btrfs_extent_inline_ref *iref;
744 struct btrfs_disk_key disk_key;
745 struct btrfs_disk_key tree_info_key;
746 struct btrfs_tree_block_info *info;
748 int skinny_metadata = cfg->features &
749 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA;
753 itemsize = sizeof(*ei) + sizeof(*iref);
755 itemsize = sizeof(*ei) + sizeof(*iref) +
756 sizeof(struct btrfs_tree_block_info);
758 btrfs_set_header_nritems(buf, *slot + 1);
759 *(itemoff) -= itemsize;
761 if (skinny_metadata) {
762 btrfs_set_disk_key_type(&disk_key, BTRFS_METADATA_ITEM_KEY);
763 btrfs_set_disk_key_offset(&disk_key, 0);
765 btrfs_set_disk_key_type(&disk_key, BTRFS_EXTENT_ITEM_KEY);
766 btrfs_set_disk_key_offset(&disk_key, cfg->nodesize);
768 btrfs_set_disk_key_objectid(&disk_key, bytenr);
770 btrfs_set_item_key(buf, &disk_key, *slot);
771 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
772 btrfs_set_item_size(buf, btrfs_item_nr(*slot), itemsize);
774 ei = btrfs_item_ptr(buf, *slot, struct btrfs_extent_item);
775 btrfs_set_extent_refs(buf, ei, 1);
776 btrfs_set_extent_generation(buf, ei, 1);
777 btrfs_set_extent_flags(buf, ei, BTRFS_EXTENT_FLAG_TREE_BLOCK);
779 if (skinny_metadata) {
780 iref = (struct btrfs_extent_inline_ref *)(ei + 1);
782 info = (struct btrfs_tree_block_info *)(ei + 1);
783 iref = (struct btrfs_extent_inline_ref *)(info + 1);
785 btrfs_set_extent_inline_ref_type(buf, iref,
786 BTRFS_TREE_BLOCK_REF_KEY);
787 btrfs_set_extent_inline_ref_offset(buf, iref, ref_root);
794 * Lastly, check the tree block key by read the tree block
795 * Since we do 1:1 mapping for convert case, we can directly
796 * read the bytenr from disk
798 tmp = malloc(sizeof(*tmp) + cfg->nodesize);
801 ret = setup_temp_extent_buffer(tmp, cfg, bytenr, ref_root);
804 ret = pread(fd, tmp->data, cfg->nodesize, bytenr);
805 if (ret < cfg->nodesize) {
806 ret = (ret < 0 ? -errno : -EIO);
809 if (btrfs_header_nritems(tmp) == 0) {
810 btrfs_set_disk_key_type(&tree_info_key, 0);
811 btrfs_set_disk_key_objectid(&tree_info_key, 0);
812 btrfs_set_disk_key_offset(&tree_info_key, 0);
814 btrfs_item_key(tmp, &tree_info_key, 0);
816 btrfs_set_tree_block_key(buf, info, &tree_info_key);
823 static void insert_temp_block_group(struct extent_buffer *buf,
824 struct btrfs_mkfs_config *cfg,
825 int *slot, u32 *itemoff,
826 u64 bytenr, u64 len, u64 used, u64 flag)
828 struct btrfs_block_group_item bgi;
829 struct btrfs_disk_key disk_key;
831 btrfs_set_header_nritems(buf, *slot + 1);
832 (*itemoff) -= sizeof(bgi);
833 btrfs_set_disk_key_type(&disk_key, BTRFS_BLOCK_GROUP_ITEM_KEY);
834 btrfs_set_disk_key_objectid(&disk_key, bytenr);
835 btrfs_set_disk_key_offset(&disk_key, len);
836 btrfs_set_item_key(buf, &disk_key, *slot);
837 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
838 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(bgi));
840 btrfs_set_block_group_flags(&bgi, flag);
841 btrfs_set_block_group_used(&bgi, used);
842 btrfs_set_block_group_chunk_objectid(&bgi,
843 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
844 write_extent_buffer(buf, &bgi, btrfs_item_ptr_offset(buf, *slot),
849 static int setup_temp_extent_tree(int fd, struct btrfs_mkfs_config *cfg,
850 u64 chunk_bytenr, u64 root_bytenr,
851 u64 extent_bytenr, u64 dev_bytenr,
852 u64 fs_bytenr, u64 csum_bytenr)
854 struct extent_buffer *buf = NULL;
855 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
860 * We must ensure provided bytenr are in ascending order,
861 * or extent tree key order will be broken.
863 if (!(chunk_bytenr < root_bytenr && root_bytenr < extent_bytenr &&
864 extent_bytenr < dev_bytenr && dev_bytenr < fs_bytenr &&
865 fs_bytenr < csum_bytenr)) {
866 error("bad tree bytenr order: "
867 "chunk < root %llu < %llu, "
868 "root < extent %llu < %llu, "
869 "extent < dev %llu < %llu, "
870 "dev < fs %llu < %llu, "
871 "fs < csum %llu < %llu",
872 (unsigned long long)chunk_bytenr,
873 (unsigned long long)root_bytenr,
874 (unsigned long long)root_bytenr,
875 (unsigned long long)extent_bytenr,
876 (unsigned long long)extent_bytenr,
877 (unsigned long long)dev_bytenr,
878 (unsigned long long)dev_bytenr,
879 (unsigned long long)fs_bytenr,
880 (unsigned long long)fs_bytenr,
881 (unsigned long long)csum_bytenr);
884 buf = malloc(sizeof(*buf) + cfg->nodesize);
888 ret = setup_temp_extent_buffer(buf, cfg, extent_bytenr,
889 BTRFS_EXTENT_TREE_OBJECTID);
893 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
894 chunk_bytenr, BTRFS_CHUNK_TREE_OBJECTID);
898 insert_temp_block_group(buf, cfg, &slot, &itemoff, chunk_bytenr,
899 BTRFS_MKFS_SYSTEM_GROUP_SIZE, cfg->nodesize,
900 BTRFS_BLOCK_GROUP_SYSTEM);
902 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
903 root_bytenr, BTRFS_ROOT_TREE_OBJECTID);
907 /* 5 tree block used, root, extent, dev, fs and csum*/
908 insert_temp_block_group(buf, cfg, &slot, &itemoff, root_bytenr,
909 BTRFS_CONVERT_META_GROUP_SIZE, cfg->nodesize * 5,
910 BTRFS_BLOCK_GROUP_METADATA);
912 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
913 extent_bytenr, BTRFS_EXTENT_TREE_OBJECTID);
916 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
917 dev_bytenr, BTRFS_DEV_TREE_OBJECTID);
920 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
921 fs_bytenr, BTRFS_FS_TREE_OBJECTID);
924 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
925 csum_bytenr, BTRFS_CSUM_TREE_OBJECTID);
929 ret = write_temp_extent_buffer(fd, buf, extent_bytenr);
936 * Improved version of make_btrfs().
939 * 1) Do chunk allocation to avoid used data
940 * And after this function, extent type matches chunk type
941 * 2) Better structured code
942 * No super long hand written codes to initialized all tree blocks
943 * Split into small blocks and reuse codes.
944 * TODO: Reuse tree operation facilities by introducing new flags
946 static int make_convert_btrfs(int fd, struct btrfs_mkfs_config *cfg,
947 struct btrfs_convert_context *cctx)
949 struct cache_tree *free = &cctx->free;
950 struct cache_tree *used = &cctx->used;
952 u64 meta_chunk_start;
953 /* chunk tree bytenr, in system chunk */
955 /* metadata trees bytenr, in metadata chunk */
963 /* Shouldn't happen */
964 BUG_ON(cache_tree_empty(used));
967 * reserve space for temporary superblock first
968 * Here we allocate a little larger space, to keep later
969 * free space will be STRIPE_LEN aligned
971 ret = reserve_free_space(free, BTRFS_STRIPE_LEN,
977 * Then reserve system chunk space
978 * TODO: Change system group size depending on cctx->total_bytes.
979 * If using current 4M, it can only handle less than one TB for
980 * worst case and then run out of sys space.
982 ret = reserve_free_space(free, BTRFS_MKFS_SYSTEM_GROUP_SIZE,
986 ret = reserve_free_space(free, BTRFS_CONVERT_META_GROUP_SIZE,
992 * Allocated meta/sys chunks will be mapped 1:1 with device offset.
994 * Inside the allocated metadata chunk, the layout will be:
995 * | offset | contents |
996 * -------------------------------------
998 * | +nodesize | extent root |
999 * | +nodesize * 2 | device root |
1000 * | +nodesize * 3 | fs tree |
1001 * | +nodesize * 4 | csum tree |
1002 * -------------------------------------
1003 * Inside the allocated system chunk, the layout will be:
1004 * | offset | contents |
1005 * -------------------------------------
1006 * | +0 | chunk root |
1007 * -------------------------------------
1009 chunk_bytenr = sys_chunk_start;
1010 root_bytenr = meta_chunk_start;
1011 extent_bytenr = meta_chunk_start + cfg->nodesize;
1012 dev_bytenr = meta_chunk_start + cfg->nodesize * 2;
1013 fs_bytenr = meta_chunk_start + cfg->nodesize * 3;
1014 csum_bytenr = meta_chunk_start + cfg->nodesize * 4;
1016 ret = setup_temp_super(fd, cfg, root_bytenr, chunk_bytenr);
1020 ret = setup_temp_root_tree(fd, cfg, root_bytenr, extent_bytenr,
1021 dev_bytenr, fs_bytenr, csum_bytenr);
1024 ret = setup_temp_chunk_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
1028 ret = setup_temp_dev_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
1032 ret = setup_temp_fs_tree(fd, cfg, fs_bytenr);
1035 ret = setup_temp_csum_tree(fd, cfg, csum_bytenr);
1039 * Setup extent tree last, since it may need to read tree block key
1040 * for non-skinny metadata case.
1042 ret = setup_temp_extent_tree(fd, cfg, chunk_bytenr, root_bytenr,
1043 extent_bytenr, dev_bytenr, fs_bytenr,
1050 * @fs_uuid - if NULL, generates a UUID, returns back the new filesystem UUID
1052 * The superblock signature is not valid, denotes a partially created
1053 * filesystem, needs to be finalized.
1055 int make_btrfs(int fd, struct btrfs_mkfs_config *cfg,
1056 struct btrfs_convert_context *cctx)
1058 struct btrfs_super_block super;
1059 struct extent_buffer *buf;
1060 struct btrfs_root_item root_item;
1061 struct btrfs_disk_key disk_key;
1062 struct btrfs_extent_item *extent_item;
1063 struct btrfs_inode_item *inode_item;
1064 struct btrfs_chunk *chunk;
1065 struct btrfs_dev_item *dev_item;
1066 struct btrfs_dev_extent *dev_extent;
1067 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
1077 int skinny_metadata = !!(cfg->features &
1078 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA);
1082 return make_convert_btrfs(fd, cfg, cctx);
1083 buf = malloc(sizeof(*buf) + max(cfg->sectorsize, cfg->nodesize));
1087 first_free = BTRFS_SUPER_INFO_OFFSET + cfg->sectorsize * 2 - 1;
1088 first_free &= ~((u64)cfg->sectorsize - 1);
1090 memset(&super, 0, sizeof(super));
1092 num_bytes = (cfg->num_bytes / cfg->sectorsize) * cfg->sectorsize;
1093 if (*cfg->fs_uuid) {
1094 if (uuid_parse(cfg->fs_uuid, super.fsid) != 0) {
1095 error("cannot not parse UUID: %s", cfg->fs_uuid);
1099 if (!test_uuid_unique(cfg->fs_uuid)) {
1100 error("non-unique UUID: %s", cfg->fs_uuid);
1105 uuid_generate(super.fsid);
1106 uuid_unparse(super.fsid, cfg->fs_uuid);
1108 uuid_generate(super.dev_item.uuid);
1109 uuid_generate(chunk_tree_uuid);
1111 btrfs_set_super_bytenr(&super, cfg->blocks[0]);
1112 btrfs_set_super_num_devices(&super, 1);
1113 btrfs_set_super_magic(&super, BTRFS_MAGIC_PARTIAL);
1114 btrfs_set_super_generation(&super, 1);
1115 btrfs_set_super_root(&super, cfg->blocks[1]);
1116 btrfs_set_super_chunk_root(&super, cfg->blocks[3]);
1117 btrfs_set_super_total_bytes(&super, num_bytes);
1118 btrfs_set_super_bytes_used(&super, 6 * cfg->nodesize);
1119 btrfs_set_super_sectorsize(&super, cfg->sectorsize);
1120 btrfs_set_super_leafsize(&super, cfg->nodesize);
1121 btrfs_set_super_nodesize(&super, cfg->nodesize);
1122 btrfs_set_super_stripesize(&super, cfg->stripesize);
1123 btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
1124 btrfs_set_super_chunk_root_generation(&super, 1);
1125 btrfs_set_super_cache_generation(&super, -1);
1126 btrfs_set_super_incompat_flags(&super, cfg->features);
1128 __strncpy_null(super.label, cfg->label, BTRFS_LABEL_SIZE - 1);
1130 /* create the tree of root objects */
1131 memset(buf->data, 0, cfg->nodesize);
1132 buf->len = cfg->nodesize;
1133 btrfs_set_header_bytenr(buf, cfg->blocks[1]);
1134 btrfs_set_header_nritems(buf, 4);
1135 btrfs_set_header_generation(buf, 1);
1136 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
1137 btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
1138 write_extent_buffer(buf, super.fsid, btrfs_header_fsid(),
1141 write_extent_buffer(buf, chunk_tree_uuid,
1142 btrfs_header_chunk_tree_uuid(buf),
1145 /* create the items for the root tree */
1146 memset(&root_item, 0, sizeof(root_item));
1147 inode_item = &root_item.inode;
1148 btrfs_set_stack_inode_generation(inode_item, 1);
1149 btrfs_set_stack_inode_size(inode_item, 3);
1150 btrfs_set_stack_inode_nlink(inode_item, 1);
1151 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
1152 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
1153 btrfs_set_root_refs(&root_item, 1);
1154 btrfs_set_root_used(&root_item, cfg->nodesize);
1155 btrfs_set_root_generation(&root_item, 1);
1157 memset(&disk_key, 0, sizeof(disk_key));
1158 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
1159 btrfs_set_disk_key_offset(&disk_key, 0);
1162 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - sizeof(root_item);
1163 btrfs_set_root_bytenr(&root_item, cfg->blocks[2]);
1164 btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
1165 btrfs_set_item_key(buf, &disk_key, nritems);
1166 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1167 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1169 write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
1170 nritems), sizeof(root_item));
1173 itemoff = itemoff - sizeof(root_item);
1174 btrfs_set_root_bytenr(&root_item, cfg->blocks[4]);
1175 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
1176 btrfs_set_item_key(buf, &disk_key, nritems);
1177 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1178 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1180 write_extent_buffer(buf, &root_item,
1181 btrfs_item_ptr_offset(buf, nritems),
1185 itemoff = itemoff - sizeof(root_item);
1186 btrfs_set_root_bytenr(&root_item, cfg->blocks[5]);
1187 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
1188 btrfs_set_item_key(buf, &disk_key, nritems);
1189 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1190 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1192 write_extent_buffer(buf, &root_item,
1193 btrfs_item_ptr_offset(buf, nritems),
1197 itemoff = itemoff - sizeof(root_item);
1198 btrfs_set_root_bytenr(&root_item, cfg->blocks[6]);
1199 btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_TREE_OBJECTID);
1200 btrfs_set_item_key(buf, &disk_key, nritems);
1201 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1202 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1204 write_extent_buffer(buf, &root_item,
1205 btrfs_item_ptr_offset(buf, nritems),
1210 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1211 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[1]);
1212 if (ret != cfg->nodesize) {
1213 ret = (ret < 0 ? -errno : -EIO);
1217 /* create the items for the extent tree */
1218 memset(buf->data + sizeof(struct btrfs_header), 0,
1219 cfg->nodesize - sizeof(struct btrfs_header));
1221 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
1222 for (i = 1; i < 7; i++) {
1223 item_size = sizeof(struct btrfs_extent_item);
1224 if (!skinny_metadata)
1225 item_size += sizeof(struct btrfs_tree_block_info);
1227 if (cfg->blocks[i] < first_free) {
1228 error("block[%d] below first free: %llu < %llu",
1229 i, (unsigned long long)cfg->blocks[i],
1230 (unsigned long long)first_free);
1234 if (cfg->blocks[i] < cfg->blocks[i - 1]) {
1235 error("blocks %d and %d in reverse order: %llu < %llu",
1237 (unsigned long long)cfg->blocks[i],
1238 (unsigned long long)cfg->blocks[i - 1]);
1243 /* create extent item */
1244 itemoff -= item_size;
1245 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
1246 if (skinny_metadata) {
1247 btrfs_set_disk_key_type(&disk_key,
1248 BTRFS_METADATA_ITEM_KEY);
1249 btrfs_set_disk_key_offset(&disk_key, 0);
1251 btrfs_set_disk_key_type(&disk_key,
1252 BTRFS_EXTENT_ITEM_KEY);
1253 btrfs_set_disk_key_offset(&disk_key, cfg->nodesize);
1255 btrfs_set_item_key(buf, &disk_key, nritems);
1256 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
1258 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1260 extent_item = btrfs_item_ptr(buf, nritems,
1261 struct btrfs_extent_item);
1262 btrfs_set_extent_refs(buf, extent_item, 1);
1263 btrfs_set_extent_generation(buf, extent_item, 1);
1264 btrfs_set_extent_flags(buf, extent_item,
1265 BTRFS_EXTENT_FLAG_TREE_BLOCK);
1268 /* create extent ref */
1269 ref_root = reference_root_table[i];
1270 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
1271 btrfs_set_disk_key_offset(&disk_key, ref_root);
1272 btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
1273 btrfs_set_item_key(buf, &disk_key, nritems);
1274 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
1276 btrfs_set_item_size(buf, btrfs_item_nr(nritems), 0);
1279 btrfs_set_header_bytenr(buf, cfg->blocks[2]);
1280 btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
1281 btrfs_set_header_nritems(buf, nritems);
1282 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1283 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[2]);
1284 if (ret != cfg->nodesize) {
1285 ret = (ret < 0 ? -errno : -EIO);
1289 /* create the chunk tree */
1290 memset(buf->data + sizeof(struct btrfs_header), 0,
1291 cfg->nodesize - sizeof(struct btrfs_header));
1293 item_size = sizeof(*dev_item);
1294 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - item_size;
1296 /* first device 1 (there is no device 0) */
1297 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
1298 btrfs_set_disk_key_offset(&disk_key, 1);
1299 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
1300 btrfs_set_item_key(buf, &disk_key, nritems);
1301 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1302 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
1304 dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
1305 btrfs_set_device_id(buf, dev_item, 1);
1306 btrfs_set_device_generation(buf, dev_item, 0);
1307 btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
1308 btrfs_set_device_bytes_used(buf, dev_item,
1309 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1310 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
1311 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
1312 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
1313 btrfs_set_device_type(buf, dev_item, 0);
1315 write_extent_buffer(buf, super.dev_item.uuid,
1316 (unsigned long)btrfs_device_uuid(dev_item),
1318 write_extent_buffer(buf, super.fsid,
1319 (unsigned long)btrfs_device_fsid(dev_item),
1321 read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
1325 item_size = btrfs_chunk_item_size(1);
1326 itemoff = itemoff - item_size;
1328 /* then we have chunk 0 */
1329 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1330 btrfs_set_disk_key_offset(&disk_key, 0);
1331 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
1332 btrfs_set_item_key(buf, &disk_key, nritems);
1333 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1334 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
1336 chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
1337 btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1338 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
1339 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
1340 btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
1341 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
1342 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
1343 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
1344 btrfs_set_chunk_num_stripes(buf, chunk, 1);
1345 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
1346 btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
1349 write_extent_buffer(buf, super.dev_item.uuid,
1350 (unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
1353 /* copy the key for the chunk to the system array */
1354 ptr = super.sys_chunk_array;
1355 array_size = sizeof(disk_key);
1357 memcpy(ptr, &disk_key, sizeof(disk_key));
1358 ptr += sizeof(disk_key);
1360 /* copy the chunk to the system array */
1361 read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
1362 array_size += item_size;
1364 btrfs_set_super_sys_array_size(&super, array_size);
1366 btrfs_set_header_bytenr(buf, cfg->blocks[3]);
1367 btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
1368 btrfs_set_header_nritems(buf, nritems);
1369 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1370 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[3]);
1371 if (ret != cfg->nodesize) {
1372 ret = (ret < 0 ? -errno : -EIO);
1376 /* create the device tree */
1377 memset(buf->data + sizeof(struct btrfs_header), 0,
1378 cfg->nodesize - sizeof(struct btrfs_header));
1380 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) -
1381 sizeof(struct btrfs_dev_extent);
1383 btrfs_set_disk_key_objectid(&disk_key, 1);
1384 btrfs_set_disk_key_offset(&disk_key, 0);
1385 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
1386 btrfs_set_item_key(buf, &disk_key, nritems);
1387 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1388 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1389 sizeof(struct btrfs_dev_extent));
1390 dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
1391 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
1392 BTRFS_CHUNK_TREE_OBJECTID);
1393 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
1394 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1395 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
1397 write_extent_buffer(buf, chunk_tree_uuid,
1398 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
1401 btrfs_set_dev_extent_length(buf, dev_extent,
1402 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1405 btrfs_set_header_bytenr(buf, cfg->blocks[4]);
1406 btrfs_set_header_owner(buf, BTRFS_DEV_TREE_OBJECTID);
1407 btrfs_set_header_nritems(buf, nritems);
1408 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1409 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[4]);
1410 if (ret != cfg->nodesize) {
1411 ret = (ret < 0 ? -errno : -EIO);
1415 /* create the FS root */
1416 memset(buf->data + sizeof(struct btrfs_header), 0,
1417 cfg->nodesize - sizeof(struct btrfs_header));
1418 btrfs_set_header_bytenr(buf, cfg->blocks[5]);
1419 btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
1420 btrfs_set_header_nritems(buf, 0);
1421 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1422 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[5]);
1423 if (ret != cfg->nodesize) {
1424 ret = (ret < 0 ? -errno : -EIO);
1427 /* finally create the csum root */
1428 memset(buf->data + sizeof(struct btrfs_header), 0,
1429 cfg->nodesize - sizeof(struct btrfs_header));
1430 btrfs_set_header_bytenr(buf, cfg->blocks[6]);
1431 btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
1432 btrfs_set_header_nritems(buf, 0);
1433 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1434 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[6]);
1435 if (ret != cfg->nodesize) {
1436 ret = (ret < 0 ? -errno : -EIO);
1440 /* and write out the super block */
1441 memset(buf->data, 0, BTRFS_SUPER_INFO_SIZE);
1442 memcpy(buf->data, &super, sizeof(super));
1443 buf->len = BTRFS_SUPER_INFO_SIZE;
1444 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1445 ret = pwrite(fd, buf->data, BTRFS_SUPER_INFO_SIZE, cfg->blocks[0]);
1446 if (ret != BTRFS_SUPER_INFO_SIZE) {
1447 ret = (ret < 0 ? -errno : -EIO);
1458 #define VERSION_TO_STRING3(a,b,c) #a "." #b "." #c, KERNEL_VERSION(a,b,c)
1459 #define VERSION_TO_STRING2(a,b) #a "." #b, KERNEL_VERSION(a,b,0)
1462 * Feature stability status and versions: compat <= safe <= default
1464 static const struct btrfs_fs_feature {
1467 const char *sysfs_name;
1469 * Compatibility with kernel of given version. Filesystem can be
1472 const char *compat_str;
1475 * Considered safe for use, but is not on by default, even if the
1476 * kernel supports the feature.
1478 const char *safe_str;
1481 * Considered safe for use and will be turned on by default if
1482 * supported by the running kernel.
1484 const char *default_str;
1487 } mkfs_features[] = {
1488 { "mixed-bg", BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS,
1490 VERSION_TO_STRING3(2,6,37),
1491 VERSION_TO_STRING3(2,6,37),
1493 "mixed data and metadata block groups" },
1494 { "extref", BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF,
1496 VERSION_TO_STRING2(3,7),
1497 VERSION_TO_STRING2(3,12),
1498 VERSION_TO_STRING2(3,12),
1499 "increased hardlink limit per file to 65536" },
1500 { "raid56", BTRFS_FEATURE_INCOMPAT_RAID56,
1502 VERSION_TO_STRING2(3,9),
1505 "raid56 extended format" },
1506 { "skinny-metadata", BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA,
1508 VERSION_TO_STRING2(3,10),
1509 VERSION_TO_STRING2(3,18),
1510 VERSION_TO_STRING2(3,18),
1511 "reduced-size metadata extent refs" },
1512 { "no-holes", BTRFS_FEATURE_INCOMPAT_NO_HOLES,
1514 VERSION_TO_STRING2(3,14),
1515 VERSION_TO_STRING2(4,0),
1517 "no explicit hole extents for files" },
1518 /* Keep this one last */
1519 { "list-all", BTRFS_FEATURE_LIST_ALL, NULL }
1522 static int parse_one_fs_feature(const char *name, u64 *flags)
1527 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1528 if (name[0] == '^' &&
1529 !strcmp(mkfs_features[i].name, name + 1)) {
1530 *flags &= ~ mkfs_features[i].flag;
1532 } else if (!strcmp(mkfs_features[i].name, name)) {
1533 *flags |= mkfs_features[i].flag;
1541 void btrfs_parse_features_to_string(char *buf, u64 flags)
1547 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1548 if (flags & mkfs_features[i].flag) {
1551 strcat(buf, mkfs_features[i].name);
1556 void btrfs_process_fs_features(u64 flags)
1560 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1561 if (flags & mkfs_features[i].flag) {
1562 printf("Turning ON incompat feature '%s': %s\n",
1563 mkfs_features[i].name,
1564 mkfs_features[i].desc);
1569 void btrfs_list_all_fs_features(u64 mask_disallowed)
1573 fprintf(stderr, "Filesystem features available:\n");
1574 for (i = 0; i < ARRAY_SIZE(mkfs_features) - 1; i++) {
1575 const struct btrfs_fs_feature *feat = &mkfs_features[i];
1577 if (feat->flag & mask_disallowed)
1579 fprintf(stderr, "%-20s- %s (0x%llx", feat->name, feat->desc,
1581 if (feat->compat_ver)
1582 fprintf(stderr, ", compat=%s", feat->compat_str);
1584 fprintf(stderr, ", safe=%s", feat->safe_str);
1585 if (feat->default_ver)
1586 fprintf(stderr, ", default=%s", feat->default_str);
1587 fprintf(stderr, ")\n");
1592 * Return NULL if all features were parsed fine, otherwise return the name of
1593 * the first unparsed.
1595 char* btrfs_parse_fs_features(char *namelist, u64 *flags)
1598 char *save_ptr = NULL; /* Satisfy static checkers */
1600 for (this_char = strtok_r(namelist, ",", &save_ptr);
1602 this_char = strtok_r(NULL, ",", &save_ptr)) {
1603 if (parse_one_fs_feature(this_char, flags))
1610 void print_kernel_version(FILE *stream, u32 version)
1614 v[0] = version & 0xFF;
1615 v[1] = (version >> 8) & 0xFF;
1616 v[2] = version >> 16;
1617 fprintf(stream, "%u.%u", v[2], v[1]);
1619 fprintf(stream, ".%u", v[0]);
1622 u32 get_running_kernel_version(void)
1624 struct utsname utsbuf;
1626 char *saveptr = NULL;
1630 if (strcmp(utsbuf.sysname, "Linux") != 0) {
1631 error("unsupported system: %s", utsbuf.sysname);
1635 tmp = strchr(utsbuf.release, '-');
1639 tmp = strtok_r(utsbuf.release, ".", &saveptr);
1640 if (!string_is_numerical(tmp))
1642 version = atoi(tmp) << 16;
1643 tmp = strtok_r(NULL, ".", &saveptr);
1644 if (!string_is_numerical(tmp))
1646 version |= atoi(tmp) << 8;
1647 tmp = strtok_r(NULL, ".", &saveptr);
1649 if (!string_is_numerical(tmp))
1651 version |= atoi(tmp);
1657 u64 btrfs_device_size(int fd, struct stat *st)
1660 if (S_ISREG(st->st_mode)) {
1663 if (!S_ISBLK(st->st_mode)) {
1666 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
1672 static int zero_blocks(int fd, off_t start, size_t len)
1674 char *buf = malloc(len);
1680 memset(buf, 0, len);
1681 written = pwrite(fd, buf, len, start);
1688 #define ZERO_DEV_BYTES (2 * 1024 * 1024)
1690 /* don't write outside the device by clamping the region to the device size */
1691 static int zero_dev_clamped(int fd, off_t start, ssize_t len, u64 dev_size)
1693 off_t end = max(start, start + len);
1696 /* and don't overwrite the disk labels on sparc */
1697 start = max(start, 1024);
1698 end = max(end, 1024);
1701 start = min_t(u64, start, dev_size);
1702 end = min_t(u64, end, dev_size);
1704 return zero_blocks(fd, start, end - start);
1707 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
1708 struct btrfs_root *root, int fd, const char *path,
1709 u64 device_total_bytes, u32 io_width, u32 io_align,
1712 struct btrfs_super_block *disk_super;
1713 struct btrfs_super_block *super = root->fs_info->super_copy;
1714 struct btrfs_device *device;
1715 struct btrfs_dev_item *dev_item;
1721 device_total_bytes = (device_total_bytes / sectorsize) * sectorsize;
1723 device = calloc(1, sizeof(*device));
1728 buf = calloc(1, sectorsize);
1734 disk_super = (struct btrfs_super_block *)buf;
1735 dev_item = &disk_super->dev_item;
1737 uuid_generate(device->uuid);
1740 device->io_width = io_width;
1741 device->io_align = io_align;
1742 device->sector_size = sectorsize;
1744 device->writeable = 1;
1745 device->total_bytes = device_total_bytes;
1746 device->bytes_used = 0;
1747 device->total_ios = 0;
1748 device->dev_root = root->fs_info->dev_root;
1749 device->name = strdup(path);
1750 if (!device->name) {
1755 INIT_LIST_HEAD(&device->dev_list);
1756 ret = btrfs_add_device(trans, root, device);
1760 fs_total_bytes = btrfs_super_total_bytes(super) + device_total_bytes;
1761 btrfs_set_super_total_bytes(super, fs_total_bytes);
1763 num_devs = btrfs_super_num_devices(super) + 1;
1764 btrfs_set_super_num_devices(super, num_devs);
1766 memcpy(disk_super, super, sizeof(*disk_super));
1768 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
1769 btrfs_set_stack_device_id(dev_item, device->devid);
1770 btrfs_set_stack_device_type(dev_item, device->type);
1771 btrfs_set_stack_device_io_align(dev_item, device->io_align);
1772 btrfs_set_stack_device_io_width(dev_item, device->io_width);
1773 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
1774 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
1775 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
1776 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
1778 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
1779 BUG_ON(ret != sectorsize);
1782 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
1783 device->fs_devices = root->fs_info->fs_devices;
1792 static int btrfs_wipe_existing_sb(int fd)
1794 const char *off = NULL;
1799 blkid_probe pr = NULL;
1801 pr = blkid_new_probe();
1805 if (blkid_probe_set_device(pr, fd, 0, 0)) {
1810 ret = blkid_probe_lookup_value(pr, "SBMAGIC_OFFSET", &off, NULL);
1812 ret = blkid_probe_lookup_value(pr, "SBMAGIC", NULL, &len);
1814 if (ret || len == 0 || off == NULL) {
1816 * If lookup fails, the probe did not find any values, eg. for
1817 * a file image or a loop device. Soft error.
1823 offset = strtoll(off, NULL, 10);
1824 if (len > sizeof(buf))
1827 memset(buf, 0, len);
1828 ret = pwrite(fd, buf, len, offset);
1830 error("cannot wipe existing superblock: %s", strerror(errno));
1832 } else if (ret != len) {
1833 error("cannot wipe existing superblock: wrote %d of %zd", ret, len);
1839 blkid_free_probe(pr);
1843 int btrfs_prepare_device(int fd, const char *file, u64 *block_count_ret,
1844 u64 max_block_count, unsigned opflags)
1850 ret = fstat(fd, &st);
1852 error("unable to stat %s: %s", file, strerror(errno));
1856 block_count = btrfs_device_size(fd, &st);
1857 if (block_count == 0) {
1858 error("unable to determine size of %s", file);
1861 if (max_block_count)
1862 block_count = min(block_count, max_block_count);
1864 if (opflags & PREP_DEVICE_DISCARD) {
1866 * We intentionally ignore errors from the discard ioctl. It
1867 * is not necessary for the mkfs functionality but just an
1870 if (discard_range(fd, 0, 0) == 0) {
1871 if (opflags & PREP_DEVICE_VERBOSE)
1872 printf("Performing full device TRIM %s (%s) ...\n",
1873 file, pretty_size(block_count));
1874 discard_blocks(fd, 0, block_count);
1878 ret = zero_dev_clamped(fd, 0, ZERO_DEV_BYTES, block_count);
1879 for (i = 0 ; !ret && i < BTRFS_SUPER_MIRROR_MAX; i++)
1880 ret = zero_dev_clamped(fd, btrfs_sb_offset(i),
1881 BTRFS_SUPER_INFO_SIZE, block_count);
1882 if (!ret && (opflags & PREP_DEVICE_ZERO_END))
1883 ret = zero_dev_clamped(fd, block_count - ZERO_DEV_BYTES,
1884 ZERO_DEV_BYTES, block_count);
1887 error("failed to zero device '%s': %s", file, strerror(-ret));
1891 ret = btrfs_wipe_existing_sb(fd);
1893 error("cannot wipe superblocks on %s", file);
1897 *block_count_ret = block_count;
1901 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
1902 struct btrfs_root *root, u64 objectid)
1905 struct btrfs_inode_item inode_item;
1906 time_t now = time(NULL);
1908 memset(&inode_item, 0, sizeof(inode_item));
1909 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
1910 btrfs_set_stack_inode_size(&inode_item, 0);
1911 btrfs_set_stack_inode_nlink(&inode_item, 1);
1912 btrfs_set_stack_inode_nbytes(&inode_item, root->nodesize);
1913 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
1914 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
1915 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
1916 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
1917 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
1918 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
1919 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
1920 btrfs_set_stack_timespec_sec(&inode_item.otime, now);
1921 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
1923 if (root->fs_info->tree_root == root)
1924 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
1926 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
1930 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
1934 btrfs_set_root_dirid(&root->root_item, objectid);
1941 * checks if a path is a block device node
1942 * Returns negative errno on failure, otherwise
1943 * returns 1 for blockdev, 0 for not-blockdev
1945 int is_block_device(const char *path)
1947 struct stat statbuf;
1949 if (stat(path, &statbuf) < 0)
1952 return !!S_ISBLK(statbuf.st_mode);
1956 * check if given path is a mount point
1957 * return 1 if yes. 0 if no. -1 for error
1959 int is_mount_point(const char *path)
1965 f = setmntent("/proc/self/mounts", "r");
1969 while ((mnt = getmntent(f)) != NULL) {
1970 if (strcmp(mnt->mnt_dir, path))
1979 static int is_reg_file(const char *path)
1981 struct stat statbuf;
1983 if (stat(path, &statbuf) < 0)
1985 return S_ISREG(statbuf.st_mode);
1989 * This function checks if the given input parameter is
1991 * return <0 : some error in the given input
1992 * return BTRFS_ARG_UNKNOWN: unknown input
1993 * return BTRFS_ARG_UUID: given input is uuid
1994 * return BTRFS_ARG_MNTPOINT: given input is path
1995 * return BTRFS_ARG_REG: given input is regular file
1996 * return BTRFS_ARG_BLKDEV: given input is block device
1998 int check_arg_type(const char *input)
2001 char path[PATH_MAX];
2006 if (realpath(input, path)) {
2007 if (is_block_device(path) == 1)
2008 return BTRFS_ARG_BLKDEV;
2010 if (is_mount_point(path) == 1)
2011 return BTRFS_ARG_MNTPOINT;
2013 if (is_reg_file(path))
2014 return BTRFS_ARG_REG;
2016 return BTRFS_ARG_UNKNOWN;
2019 if (strlen(input) == (BTRFS_UUID_UNPARSED_SIZE - 1) &&
2020 !uuid_parse(input, uuid))
2021 return BTRFS_ARG_UUID;
2023 return BTRFS_ARG_UNKNOWN;
2027 * Find the mount point for a mounted device.
2028 * On success, returns 0 with mountpoint in *mp.
2029 * On failure, returns -errno (not mounted yields -EINVAL)
2030 * Is noisy on failures, expects to be given a mounted device.
2032 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size)
2037 ret = is_block_device(dev);
2040 error("not a block device: %s", dev);
2043 error("cannot check %s: %s", dev, strerror(-ret));
2048 fd = open(dev, O_RDONLY);
2051 error("cannot open %s: %s", dev, strerror(errno));
2055 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
2058 } else { /* mounted, all good */
2068 * Given a pathname, return a filehandle to:
2069 * the original pathname or,
2070 * if the pathname is a mounted btrfs device, to its mountpoint.
2072 * On error, return -1, errno should be set.
2074 int open_path_or_dev_mnt(const char *path, DIR **dirstream, int verbose)
2079 if (is_block_device(path)) {
2080 ret = get_btrfs_mount(path, mp, sizeof(mp));
2082 /* not a mounted btrfs dev */
2083 error_on(verbose, "'%s' is not a mounted btrfs device",
2088 ret = open_file_or_dir(mp, dirstream);
2089 error_on(verbose && ret < 0, "can't access '%s': %s",
2090 path, strerror(errno));
2092 ret = btrfs_open_dir(path, dirstream, 1);
2099 * Do the following checks before calling open_file_or_dir():
2100 * 1: path is in a btrfs filesystem
2101 * 2: path is a directory
2103 int btrfs_open_dir(const char *path, DIR **dirstream, int verbose)
2109 if (statfs(path, &stfs) != 0) {
2110 error_on(verbose, "cannot access '%s': %s", path,
2115 if (stfs.f_type != BTRFS_SUPER_MAGIC) {
2116 error_on(verbose, "not a btrfs filesystem: %s", path);
2120 if (stat(path, &st) != 0) {
2121 error_on(verbose, "cannot access '%s': %s", path,
2126 if (!S_ISDIR(st.st_mode)) {
2127 error_on(verbose, "not a directory: %s", path);
2131 ret = open_file_or_dir(path, dirstream);
2133 error_on(verbose, "cannot access '%s': %s", path,
2140 /* checks if a device is a loop device */
2141 static int is_loop_device (const char* device) {
2142 struct stat statbuf;
2144 if(stat(device, &statbuf) < 0)
2147 return (S_ISBLK(statbuf.st_mode) &&
2148 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
2152 * Takes a loop device path (e.g. /dev/loop0) and returns
2153 * the associated file (e.g. /images/my_btrfs.img) using
2156 static int resolve_loop_device_with_loopdev(const char* loop_dev, char* loop_file)
2160 struct loop_info64 lo64;
2162 fd = open(loop_dev, O_RDONLY | O_NONBLOCK);
2165 ret = ioctl(fd, LOOP_GET_STATUS64, &lo64);
2171 memcpy(loop_file, lo64.lo_file_name, sizeof(lo64.lo_file_name));
2172 loop_file[sizeof(lo64.lo_file_name)] = 0;
2180 /* Takes a loop device path (e.g. /dev/loop0) and returns
2181 * the associated file (e.g. /images/my_btrfs.img) */
2182 static int resolve_loop_device(const char* loop_dev, char* loop_file,
2189 char real_loop_dev[PATH_MAX];
2191 if (!realpath(loop_dev, real_loop_dev))
2193 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
2194 if (!(f = fopen(p, "r"))) {
2195 if (errno == ENOENT)
2197 * It's possibly a partitioned loop device, which is
2198 * resolvable with loopdev API.
2200 return resolve_loop_device_with_loopdev(loop_dev, loop_file);
2204 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
2205 ret = fscanf(f, fmt, loop_file);
2214 * Checks whether a and b are identical or device
2215 * files associated with the same block device
2217 static int is_same_blk_file(const char* a, const char* b)
2219 struct stat st_buf_a, st_buf_b;
2220 char real_a[PATH_MAX];
2221 char real_b[PATH_MAX];
2223 if (!realpath(a, real_a))
2224 strncpy_null(real_a, a);
2226 if (!realpath(b, real_b))
2227 strncpy_null(real_b, b);
2229 /* Identical path? */
2230 if (strcmp(real_a, real_b) == 0)
2233 if (stat(a, &st_buf_a) < 0 || stat(b, &st_buf_b) < 0) {
2234 if (errno == ENOENT)
2239 /* Same blockdevice? */
2240 if (S_ISBLK(st_buf_a.st_mode) && S_ISBLK(st_buf_b.st_mode) &&
2241 st_buf_a.st_rdev == st_buf_b.st_rdev) {
2246 if (st_buf_a.st_dev == st_buf_b.st_dev &&
2247 st_buf_a.st_ino == st_buf_b.st_ino) {
2254 /* checks if a and b are identical or device
2255 * files associated with the same block device or
2256 * if one file is a loop device that uses the other
2259 static int is_same_loop_file(const char* a, const char* b)
2261 char res_a[PATH_MAX];
2262 char res_b[PATH_MAX];
2263 const char* final_a = NULL;
2264 const char* final_b = NULL;
2267 /* Resolve a if it is a loop device */
2268 if((ret = is_loop_device(a)) < 0) {
2273 ret = resolve_loop_device(a, res_a, sizeof(res_a));
2284 /* Resolve b if it is a loop device */
2285 if ((ret = is_loop_device(b)) < 0) {
2290 ret = resolve_loop_device(b, res_b, sizeof(res_b));
2301 return is_same_blk_file(final_a, final_b);
2304 /* Checks if a file exists and is a block or regular file*/
2305 static int is_existing_blk_or_reg_file(const char* filename)
2309 if(stat(filename, &st_buf) < 0) {
2316 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
2319 /* Checks if a file is used (directly or indirectly via a loop device)
2320 * by a device in fs_devices
2322 static int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices,
2326 struct list_head *head;
2327 struct list_head *cur;
2328 struct btrfs_device *device;
2330 head = &fs_devices->devices;
2331 list_for_each(cur, head) {
2332 device = list_entry(cur, struct btrfs_device, dev_list);
2334 if((ret = is_same_loop_file(device->name, file)))
2342 * Resolve a pathname to a device mapper node to /dev/mapper/<name>
2343 * Returns NULL on invalid input or malloc failure; Other failures
2344 * will be handled by the caller using the input pathame.
2346 char *canonicalize_dm_name(const char *ptname)
2350 char path[PATH_MAX], name[PATH_MAX], *res = NULL;
2352 if (!ptname || !*ptname)
2355 snprintf(path, sizeof(path), "/sys/block/%s/dm/name", ptname);
2356 if (!(f = fopen(path, "r")))
2359 /* read <name>\n from sysfs */
2360 if (fgets(name, sizeof(name), f) && (sz = strlen(name)) > 1) {
2361 name[sz - 1] = '\0';
2362 snprintf(path, sizeof(path), "/dev/mapper/%s", name);
2364 if (access(path, F_OK) == 0)
2372 * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or
2373 * to a device mapper pathname.
2374 * Returns NULL on invalid input or malloc failure; Other failures
2375 * will be handled by the caller using the input pathame.
2377 char *canonicalize_path(const char *path)
2379 char *canonical, *p;
2381 if (!path || !*path)
2384 canonical = realpath(path, NULL);
2386 return strdup(path);
2387 p = strrchr(canonical, '/');
2388 if (p && strncmp(p, "/dm-", 4) == 0 && isdigit(*(p + 4))) {
2389 char *dm = canonicalize_dm_name(p + 1);
2400 * returns 1 if the device was mounted, < 0 on error or 0 if everything
2401 * is safe to continue.
2403 int check_mounted(const char* file)
2408 fd = open(file, O_RDONLY);
2410 error("mount check: cannot open %s: %s", file,
2415 ret = check_mounted_where(fd, file, NULL, 0, NULL);
2421 int check_mounted_where(int fd, const char *file, char *where, int size,
2422 struct btrfs_fs_devices **fs_dev_ret)
2427 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2431 /* scan the initial device */
2432 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
2433 &total_devs, BTRFS_SUPER_INFO_OFFSET, SBREAD_DEFAULT);
2434 is_btrfs = (ret >= 0);
2436 /* scan other devices */
2437 if (is_btrfs && total_devs > 1) {
2438 ret = btrfs_scan_devices();
2443 /* iterate over the list of currently mounted filesystems */
2444 if ((f = setmntent ("/proc/self/mounts", "r")) == NULL)
2447 while ((mnt = getmntent (f)) != NULL) {
2449 if(strcmp(mnt->mnt_type, "btrfs") != 0)
2452 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
2454 /* ignore entries in the mount table that are not
2455 associated with a file*/
2456 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
2457 goto out_mntloop_err;
2461 ret = is_same_loop_file(file, mnt->mnt_fsname);
2465 goto out_mntloop_err;
2470 /* Did we find an entry in mnt table? */
2471 if (mnt && size && where) {
2472 strncpy(where, mnt->mnt_dir, size);
2476 *fs_dev_ret = fs_devices_mnt;
2478 ret = (mnt != NULL);
2486 struct pending_dir {
2487 struct list_head list;
2488 char name[PATH_MAX];
2491 int btrfs_register_one_device(const char *fname)
2493 struct btrfs_ioctl_vol_args args;
2497 fd = open("/dev/btrfs-control", O_RDWR);
2500 "failed to open /dev/btrfs-control, skipping device registration: %s",
2504 memset(&args, 0, sizeof(args));
2505 strncpy_null(args.name, fname);
2506 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
2508 error("device scan failed on '%s': %s", fname,
2517 * Register all devices in the fs_uuid list created in the user
2518 * space. Ensure btrfs_scan_devices() is called before this func.
2520 int btrfs_register_all_devices(void)
2524 struct btrfs_fs_devices *fs_devices;
2525 struct btrfs_device *device;
2526 struct list_head *all_uuids;
2528 all_uuids = btrfs_scanned_uuids();
2530 list_for_each_entry(fs_devices, all_uuids, list) {
2531 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2533 err = btrfs_register_one_device(device->name);
2543 int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
2546 struct btrfs_super_block *disk_super;
2550 buf = malloc(BTRFS_SUPER_INFO_SIZE);
2555 ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
2556 if (ret != BTRFS_SUPER_INFO_SIZE)
2560 disk_super = (struct btrfs_super_block *)buf;
2562 * Accept devices from the same filesystem, allow partially created
2565 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC &&
2566 btrfs_super_magic(disk_super) != BTRFS_MAGIC_PARTIAL)
2569 if (!memcmp(disk_super->fsid, root->fs_info->super_copy->fsid,
2579 * Note: this function uses a static per-thread buffer. Do not call this
2580 * function more than 10 times within one argument list!
2582 const char *pretty_size_mode(u64 size, unsigned mode)
2584 static __thread int ps_index = 0;
2585 static __thread char ps_array[10][32];
2588 ret = ps_array[ps_index];
2591 (void)pretty_size_snprintf(size, ret, 32, mode);
2596 static const char* unit_suffix_binary[] =
2597 { "B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB"};
2598 static const char* unit_suffix_decimal[] =
2599 { "B", "kB", "MB", "GB", "TB", "PB", "EB"};
2601 int pretty_size_snprintf(u64 size, char *str, size_t str_size, unsigned unit_mode)
2607 const char** suffix = NULL;
2614 negative = !!(unit_mode & UNITS_NEGATIVE);
2615 unit_mode &= ~UNITS_NEGATIVE;
2617 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_RAW) {
2619 snprintf(str, str_size, "%lld", size);
2621 snprintf(str, str_size, "%llu", size);
2625 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_BINARY) {
2628 suffix = unit_suffix_binary;
2629 } else if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_DECIMAL) {
2632 suffix = unit_suffix_decimal;
2637 fprintf(stderr, "INTERNAL ERROR: unknown unit base, mode %d\n",
2645 switch (unit_mode & UNITS_MODE_MASK) {
2646 case UNITS_TBYTES: base *= mult; num_divs++;
2647 case UNITS_GBYTES: base *= mult; num_divs++;
2648 case UNITS_MBYTES: base *= mult; num_divs++;
2649 case UNITS_KBYTES: num_divs++;
2657 s64 ssize = (s64)size;
2658 s64 last_ssize = ssize;
2660 while ((ssize < 0 ? -ssize : ssize) >= mult) {
2665 last_size = (u64)last_ssize;
2667 while (size >= mult) {
2674 * If the value is smaller than base, we didn't do any
2675 * division, in that case, base should be 1, not original
2676 * base, or the unit will be wrong
2682 if (num_divs >= ARRAY_SIZE(unit_suffix_binary)) {
2684 printf("INTERNAL ERROR: unsupported unit suffix, index %d\n",
2691 fraction = (float)(s64)last_size / base;
2693 fraction = (float)last_size / base;
2696 return snprintf(str, str_size, "%.2f%s", fraction, suffix[num_divs]);
2700 * __strncpy_null - strncpy with null termination
2701 * @dest: the target array
2702 * @src: the source string
2703 * @n: maximum bytes to copy (size of *dest)
2705 * Like strncpy, but ensures destination is null-terminated.
2707 * Copies the string pointed to by src, including the terminating null
2708 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
2709 * of n bytes. Then ensure that dest is null-terminated.
2711 char *__strncpy_null(char *dest, const char *src, size_t n)
2713 strncpy(dest, src, n);
2720 * Checks to make sure that the label matches our requirements.
2722 0 if everything is safe and usable
2723 -1 if the label is too long
2725 static int check_label(const char *input)
2727 int len = strlen(input);
2729 if (len > BTRFS_LABEL_SIZE - 1) {
2730 error("label %s is too long (max %d)", input,
2731 BTRFS_LABEL_SIZE - 1);
2738 static int set_label_unmounted(const char *dev, const char *label)
2740 struct btrfs_trans_handle *trans;
2741 struct btrfs_root *root;
2744 ret = check_mounted(dev);
2746 error("checking mount status of %s failed: %d", dev, ret);
2750 error("device %s is mounted, use mount point", dev);
2754 /* Open the super_block at the default location
2755 * and as read-write.
2757 root = open_ctree(dev, 0, OPEN_CTREE_WRITES);
2758 if (!root) /* errors are printed by open_ctree() */
2761 trans = btrfs_start_transaction(root, 1);
2762 __strncpy_null(root->fs_info->super_copy->label, label, BTRFS_LABEL_SIZE - 1);
2764 btrfs_commit_transaction(trans, root);
2766 /* Now we close it since we are done. */
2771 static int set_label_mounted(const char *mount_path, const char *labelp)
2774 char label[BTRFS_LABEL_SIZE];
2776 fd = open(mount_path, O_RDONLY | O_NOATIME);
2778 error("unable to access %s: %s", mount_path, strerror(errno));
2782 memset(label, 0, sizeof(label));
2783 __strncpy_null(label, labelp, BTRFS_LABEL_SIZE - 1);
2784 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
2785 error("unable to set label of %s: %s", mount_path,
2795 int get_label_unmounted(const char *dev, char *label)
2797 struct btrfs_root *root;
2800 ret = check_mounted(dev);
2802 error("checking mount status of %s failed: %d", dev, ret);
2806 /* Open the super_block at the default location
2809 root = open_ctree(dev, 0, 0);
2813 __strncpy_null(label, root->fs_info->super_copy->label,
2814 BTRFS_LABEL_SIZE - 1);
2816 /* Now we close it since we are done. */
2822 * If a partition is mounted, try to get the filesystem label via its
2823 * mounted path rather than device. Return the corresponding error
2824 * the user specified the device path.
2826 int get_label_mounted(const char *mount_path, char *labelp)
2828 char label[BTRFS_LABEL_SIZE];
2832 fd = open(mount_path, O_RDONLY | O_NOATIME);
2834 error("unable to access %s: %s", mount_path, strerror(errno));
2838 memset(label, '\0', sizeof(label));
2839 ret = ioctl(fd, BTRFS_IOC_GET_FSLABEL, label);
2841 if (errno != ENOTTY)
2842 error("unable to get label of %s: %s", mount_path,
2849 __strncpy_null(labelp, label, BTRFS_LABEL_SIZE - 1);
2854 int get_label(const char *btrfs_dev, char *label)
2858 ret = is_existing_blk_or_reg_file(btrfs_dev);
2860 ret = get_label_mounted(btrfs_dev, label);
2862 ret = get_label_unmounted(btrfs_dev, label);
2867 int set_label(const char *btrfs_dev, const char *label)
2871 if (check_label(label))
2874 ret = is_existing_blk_or_reg_file(btrfs_dev);
2876 ret = set_label_mounted(btrfs_dev, label);
2878 ret = set_label_unmounted(btrfs_dev, label);
2884 * A not-so-good version fls64. No fascinating optimization since
2885 * no one except parse_size use it
2887 static int fls64(u64 x)
2891 for (i = 0; i <64; i++)
2892 if (x << i & (1ULL << 63))
2897 u64 parse_size(char *s)
2905 error("size value is empty");
2909 error("size value '%s' is less equal than 0", s);
2912 ret = strtoull(s, &endptr, 10);
2914 error("size value '%s' is invalid", s);
2917 if (endptr[0] && endptr[1]) {
2918 error("illegal suffix contains character '%c' in wrong position",
2923 * strtoll returns LLONG_MAX when overflow, if this happens,
2924 * need to call strtoull to get the real size
2926 if (errno == ERANGE && ret == ULLONG_MAX) {
2927 error("size value '%s' is too large for u64", s);
2931 c = tolower(endptr[0]);
2954 error("unknown size descriptor '%c'", c);
2958 /* Check whether ret * mult overflow */
2959 if (fls64(ret) + fls64(mult) - 1 > 64) {
2960 error("size value '%s' is too large for u64", s);
2967 u64 parse_qgroupid(const char *p)
2969 char *s = strchr(p, '/');
2970 const char *ptr_src_end = p + strlen(p);
2971 char *ptr_parse_end = NULL;
2980 /* Numeric format like '0/257' is the primary case */
2982 id = strtoull(p, &ptr_parse_end, 10);
2983 if (ptr_parse_end != ptr_src_end)
2987 level = strtoull(p, &ptr_parse_end, 10);
2988 if (ptr_parse_end != s)
2991 id = strtoull(s + 1, &ptr_parse_end, 10);
2992 if (ptr_parse_end != ptr_src_end)
2995 return (level << BTRFS_QGROUP_LEVEL_SHIFT) | id;
2998 /* Path format like subv at 'my_subvol' is the fallback case */
2999 ret = test_issubvolume(p);
3000 if (ret < 0 || !ret)
3002 fd = open(p, O_RDONLY);
3005 ret = lookup_path_rootid(fd, &id);
3007 error("failed to lookup root id: %s", strerror(-ret));
3014 error("invalid qgroupid or subvolume path: %s", p);
3018 int open_file_or_dir3(const char *fname, DIR **dirstream, int open_flags)
3024 ret = stat(fname, &st);
3028 if (S_ISDIR(st.st_mode)) {
3029 *dirstream = opendir(fname);
3032 fd = dirfd(*dirstream);
3033 } else if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) {
3034 fd = open(fname, open_flags);
3037 * we set this on purpose, in case the caller output
3038 * strerror(errno) as success
3046 closedir(*dirstream);
3053 int open_file_or_dir(const char *fname, DIR **dirstream)
3055 return open_file_or_dir3(fname, dirstream, O_RDWR);
3058 void close_file_or_dir(int fd, DIR *dirstream)
3061 closedir(dirstream);
3066 int get_device_info(int fd, u64 devid,
3067 struct btrfs_ioctl_dev_info_args *di_args)
3071 di_args->devid = devid;
3072 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
3074 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
3075 return ret < 0 ? -errno : 0;
3078 static u64 find_max_device_id(struct btrfs_ioctl_search_args *search_args,
3081 struct btrfs_dev_item *dev_item;
3082 char *buf = search_args->buf;
3084 buf += (nr_items - 1) * (sizeof(struct btrfs_ioctl_search_header)
3085 + sizeof(struct btrfs_dev_item));
3086 buf += sizeof(struct btrfs_ioctl_search_header);
3088 dev_item = (struct btrfs_dev_item *)buf;
3090 return btrfs_stack_device_id(dev_item);
3093 static int search_chunk_tree_for_fs_info(int fd,
3094 struct btrfs_ioctl_fs_info_args *fi_args)
3098 u64 start_devid = 1;
3099 struct btrfs_ioctl_search_args search_args;
3100 struct btrfs_ioctl_search_key *search_key = &search_args.key;
3102 fi_args->num_devices = 0;
3104 max_items = BTRFS_SEARCH_ARGS_BUFSIZE
3105 / (sizeof(struct btrfs_ioctl_search_header)
3106 + sizeof(struct btrfs_dev_item));
3108 search_key->tree_id = BTRFS_CHUNK_TREE_OBJECTID;
3109 search_key->min_objectid = BTRFS_DEV_ITEMS_OBJECTID;
3110 search_key->max_objectid = BTRFS_DEV_ITEMS_OBJECTID;
3111 search_key->min_type = BTRFS_DEV_ITEM_KEY;
3112 search_key->max_type = BTRFS_DEV_ITEM_KEY;
3113 search_key->min_transid = 0;
3114 search_key->max_transid = (u64)-1;
3115 search_key->nr_items = max_items;
3116 search_key->max_offset = (u64)-1;
3119 search_key->min_offset = start_devid;
3121 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH, &search_args);
3125 fi_args->num_devices += (u64)search_key->nr_items;
3127 if (search_key->nr_items == max_items) {
3128 start_devid = find_max_device_id(&search_args,
3129 search_key->nr_items) + 1;
3133 /* get the lastest max_id to stay consistent with the num_devices */
3134 if (search_key->nr_items == 0)
3136 * last tree_search returns an empty buf, use the devid of
3137 * the last dev_item of the previous tree_search
3139 fi_args->max_id = start_devid - 1;
3141 fi_args->max_id = find_max_device_id(&search_args,
3142 search_key->nr_items);
3148 * For a given path, fill in the ioctl fs_ and info_ args.
3149 * If the path is a btrfs mountpoint, fill info for all devices.
3150 * If the path is a btrfs device, fill in only that device.
3152 * The path provided must be either on a mounted btrfs fs,
3153 * or be a mounted btrfs device.
3155 * Returns 0 on success, or a negative errno.
3157 int get_fs_info(const char *path, struct btrfs_ioctl_fs_info_args *fi_args,
3158 struct btrfs_ioctl_dev_info_args **di_ret)
3165 struct btrfs_fs_devices *fs_devices_mnt = NULL;
3166 struct btrfs_ioctl_dev_info_args *di_args;
3167 struct btrfs_ioctl_dev_info_args tmp;
3169 DIR *dirstream = NULL;
3171 memset(fi_args, 0, sizeof(*fi_args));
3173 if (is_block_device(path) == 1) {
3174 struct btrfs_super_block *disk_super;
3175 char buf[BTRFS_SUPER_INFO_SIZE];
3177 /* Ensure it's mounted, then set path to the mountpoint */
3178 fd = open(path, O_RDONLY);
3181 error("cannot open %s: %s", path, strerror(errno));
3184 ret = check_mounted_where(fd, path, mp, sizeof(mp),
3193 /* Only fill in this one device */
3194 fi_args->num_devices = 1;
3196 disk_super = (struct btrfs_super_block *)buf;
3197 ret = btrfs_read_dev_super(fd, disk_super,
3198 BTRFS_SUPER_INFO_OFFSET, 0);
3203 last_devid = btrfs_stack_device_id(&disk_super->dev_item);
3204 fi_args->max_id = last_devid;
3206 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
3210 /* at this point path must not be for a block device */
3211 fd = open_file_or_dir(path, &dirstream);
3217 /* fill in fi_args if not just a single device */
3218 if (fi_args->num_devices != 1) {
3219 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
3226 * The fs_args->num_devices does not include seed devices
3228 ret = search_chunk_tree_for_fs_info(fd, fi_args);
3233 * search_chunk_tree_for_fs_info() will lacks the devid 0
3234 * so manual probe for it here.
3236 ret = get_device_info(fd, 0, &tmp);
3238 fi_args->num_devices++;
3241 if (last_devid == 0)
3246 if (!fi_args->num_devices)
3249 di_args = *di_ret = malloc((fi_args->num_devices) * sizeof(*di_args));
3256 memcpy(di_args, &tmp, sizeof(tmp));
3257 for (; last_devid <= fi_args->max_id; last_devid++) {
3258 ret = get_device_info(fd, last_devid, &di_args[ndevs]);
3267 * only when the only dev we wanted to find is not there then
3268 * let any error be returned
3270 if (fi_args->num_devices != 1) {
3276 close_file_or_dir(fd, dirstream);
3280 #define isoctal(c) (((c) & ~7) == '0')
3282 static inline void translate(char *f, char *t)
3284 while (*f != '\0') {
3286 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
3287 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
3297 * Checks if the swap device.
3298 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
3300 static int is_swap_device(const char *file)
3311 if (stat(file, &st_buf) < 0)
3313 if (S_ISBLK(st_buf.st_mode))
3314 dev = st_buf.st_rdev;
3315 else if (S_ISREG(st_buf.st_mode)) {
3316 dev = st_buf.st_dev;
3317 ino = st_buf.st_ino;
3321 if ((f = fopen("/proc/swaps", "r")) == NULL)
3324 /* skip the first line */
3325 if (fgets(tmp, sizeof(tmp), f) == NULL)
3328 while (fgets(tmp, sizeof(tmp), f) != NULL) {
3329 if ((cp = strchr(tmp, ' ')) != NULL)
3331 if ((cp = strchr(tmp, '\t')) != NULL)
3333 translate(tmp, buf);
3334 if (stat(buf, &st_buf) != 0)
3336 if (S_ISBLK(st_buf.st_mode)) {
3337 if (dev == st_buf.st_rdev) {
3341 } else if (S_ISREG(st_buf.st_mode)) {
3342 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {
3356 * Check for existing filesystem or partition table on device.
3358 * 1 for existing fs or partition
3359 * 0 for nothing found
3360 * -1 for internal error
3362 static int check_overwrite(const char *device)
3365 blkid_probe pr = NULL;
3369 if (!device || !*device)
3372 ret = -1; /* will reset on success of all setup calls */
3374 pr = blkid_new_probe_from_filename(device);
3378 size = blkid_probe_get_size(pr);
3382 /* nothing to overwrite on a 0-length device */
3388 ret = blkid_probe_enable_partitions(pr, 1);
3392 ret = blkid_do_fullprobe(pr);
3397 * Blkid returns 1 for nothing found and 0 when it finds a signature,
3398 * but we want the exact opposite, so reverse the return value here.
3400 * In addition print some useful diagnostics about what actually is
3408 if (!blkid_probe_lookup_value(pr, "TYPE", &type, NULL)) {
3410 "%s appears to contain an existing "
3411 "filesystem (%s).\n", device, type);
3412 } else if (!blkid_probe_lookup_value(pr, "PTTYPE", &type, NULL)) {
3414 "%s appears to contain a partition "
3415 "table (%s).\n", device, type);
3418 "%s appears to contain something weird "
3419 "according to blkid\n", device);
3425 blkid_free_probe(pr);
3428 "probe of %s failed, cannot detect "
3429 "existing filesystem.\n", device);
3433 static int group_profile_devs_min(u64 flag)
3435 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3436 case 0: /* single */
3437 case BTRFS_BLOCK_GROUP_DUP:
3439 case BTRFS_BLOCK_GROUP_RAID0:
3440 case BTRFS_BLOCK_GROUP_RAID1:
3441 case BTRFS_BLOCK_GROUP_RAID5:
3443 case BTRFS_BLOCK_GROUP_RAID6:
3445 case BTRFS_BLOCK_GROUP_RAID10:
3452 int test_num_disk_vs_raid(u64 metadata_profile, u64 data_profile,
3453 u64 dev_cnt, int mixed, int ssd)
3456 u64 profile = metadata_profile | data_profile;
3461 allowed |= BTRFS_BLOCK_GROUP_RAID10;
3463 allowed |= BTRFS_BLOCK_GROUP_RAID6;
3465 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
3466 BTRFS_BLOCK_GROUP_RAID5;
3468 allowed |= BTRFS_BLOCK_GROUP_DUP;
3471 if (dev_cnt > 1 && profile & BTRFS_BLOCK_GROUP_DUP) {
3472 warning("DUP is not recommended on filesystem with multiple devices");
3474 if (metadata_profile & ~allowed) {
3476 "ERROR: unable to create FS with metadata profile %s "
3477 "(have %llu devices but %d devices are required)\n",
3478 btrfs_group_profile_str(metadata_profile), dev_cnt,
3479 group_profile_devs_min(metadata_profile));
3482 if (data_profile & ~allowed) {
3484 "ERROR: unable to create FS with data profile %s "
3485 "(have %llu devices but %d devices are required)\n",
3486 btrfs_group_profile_str(data_profile), dev_cnt,
3487 group_profile_devs_min(data_profile));
3491 if (dev_cnt == 3 && profile & BTRFS_BLOCK_GROUP_RAID6) {
3492 warning("RAID6 is not recommended on filesystem with 3 devices only");
3494 if (dev_cnt == 2 && profile & BTRFS_BLOCK_GROUP_RAID5) {
3495 warning("RAID5 is not recommended on filesystem with 2 devices only");
3497 warning_on(!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP) && ssd,
3498 "DUP may not actually lead to 2 copies on the device, see manual page");
3503 int group_profile_max_safe_loss(u64 flags)
3505 switch (flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3506 case 0: /* single */
3507 case BTRFS_BLOCK_GROUP_DUP:
3508 case BTRFS_BLOCK_GROUP_RAID0:
3510 case BTRFS_BLOCK_GROUP_RAID1:
3511 case BTRFS_BLOCK_GROUP_RAID5:
3512 case BTRFS_BLOCK_GROUP_RAID10:
3514 case BTRFS_BLOCK_GROUP_RAID6:
3522 * Check if a device is suitable for btrfs
3524 * 1: something is wrong, an error is printed
3527 int test_dev_for_mkfs(const char *file, int force_overwrite)
3532 ret = is_swap_device(file);
3534 error("checking status of %s: %s", file, strerror(-ret));
3538 error("%s is a swap device", file);
3541 if (!force_overwrite) {
3542 if (check_overwrite(file)) {
3543 error("use the -f option to force overwrite of %s",
3548 ret = check_mounted(file);
3550 error("cannot check mount status of %s: %s", file,
3555 error("%s is mounted", file);
3558 /* check if the device is busy */
3559 fd = open(file, O_RDWR|O_EXCL);
3561 error("unable to open %s: %s", file, strerror(errno));
3564 if (fstat(fd, &st)) {
3565 error("unable to stat %s: %s", file, strerror(errno));
3569 if (!S_ISBLK(st.st_mode)) {
3570 error("%s is not a block device", file);
3578 int btrfs_scan_devices(void)
3583 struct btrfs_fs_devices *tmp_devices;
3584 blkid_dev_iterate iter = NULL;
3585 blkid_dev dev = NULL;
3586 blkid_cache cache = NULL;
3587 char path[PATH_MAX];
3589 if (btrfs_scan_done)
3592 if (blkid_get_cache(&cache, NULL) < 0) {
3593 error("blkid cache get failed");
3596 blkid_probe_all(cache);
3597 iter = blkid_dev_iterate_begin(cache);
3598 blkid_dev_set_search(iter, "TYPE", "btrfs");
3599 while (blkid_dev_next(iter, &dev) == 0) {
3600 dev = blkid_verify(cache, dev);
3603 /* if we are here its definitely a btrfs disk*/
3604 strncpy_null(path, blkid_dev_devname(dev));
3606 fd = open(path, O_RDONLY);
3608 error("cannot open %s: %s", path, strerror(errno));
3611 ret = btrfs_scan_one_device(fd, path, &tmp_devices,
3612 &num_devices, BTRFS_SUPER_INFO_OFFSET,
3615 error("cannot scan %s: %s", path, strerror(-ret));
3622 blkid_dev_iterate_end(iter);
3623 blkid_put_cache(cache);
3625 btrfs_scan_done = 1;
3630 int is_vol_small(const char *file)
3637 fd = open(file, O_RDONLY);
3640 if (fstat(fd, &st) < 0) {
3645 size = btrfs_device_size(fd, &st);
3650 if (size < BTRFS_MKFS_SMALL_VOLUME_SIZE) {
3660 * This reads a line from the stdin and only returns non-zero if the
3661 * first whitespace delimited token is a case insensitive match with yes
3664 int ask_user(const char *question)
3666 char buf[30] = {0,};
3667 char *saveptr = NULL;
3670 printf("%s [y/N]: ", question);
3672 return fgets(buf, sizeof(buf) - 1, stdin) &&
3673 (answer = strtok_r(buf, " \t\n\r", &saveptr)) &&
3674 (!strcasecmp(answer, "yes") || !strcasecmp(answer, "y"));
3678 * return 0 if a btrfs mount point is found
3679 * return 1 if a mount point is found but not btrfs
3680 * return <0 if something goes wrong
3682 int find_mount_root(const char *path, char **mount_root)
3690 int longest_matchlen = 0;
3691 char *longest_match = NULL;
3693 fd = open(path, O_RDONLY | O_NOATIME);
3698 mnttab = setmntent("/proc/self/mounts", "r");
3702 while ((ent = getmntent(mnttab))) {
3703 len = strlen(ent->mnt_dir);
3704 if (strncmp(ent->mnt_dir, path, len) == 0) {
3705 /* match found and use the latest match */
3706 if (longest_matchlen <= len) {
3707 free(longest_match);
3708 longest_matchlen = len;
3709 longest_match = strdup(ent->mnt_dir);
3710 not_btrfs = strcmp(ent->mnt_type, "btrfs");
3719 free(longest_match);
3724 *mount_root = realpath(longest_match, NULL);
3728 free(longest_match);
3732 int test_minimum_size(const char *file, u32 nodesize)
3735 struct stat statbuf;
3737 fd = open(file, O_RDONLY);
3740 if (stat(file, &statbuf) < 0) {
3744 if (btrfs_device_size(fd, &statbuf) < btrfs_min_dev_size(nodesize)) {
3754 * Test if path is a directory
3756 * 0 - path exists but it is not a directory
3757 * 1 - path exists and it is a directory
3760 int test_isdir(const char *path)
3765 ret = stat(path, &st);
3769 return !!S_ISDIR(st.st_mode);
3772 void units_set_mode(unsigned *units, unsigned mode)
3774 unsigned base = *units & UNITS_MODE_MASK;
3776 *units = base | mode;
3779 void units_set_base(unsigned *units, unsigned base)
3781 unsigned mode = *units & ~UNITS_MODE_MASK;
3783 *units = base | mode;
3786 int find_next_key(struct btrfs_path *path, struct btrfs_key *key)
3790 for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
3791 if (!path->nodes[level])
3793 if (path->slots[level] + 1 >=
3794 btrfs_header_nritems(path->nodes[level]))
3797 btrfs_item_key_to_cpu(path->nodes[level], key,
3798 path->slots[level] + 1);
3800 btrfs_node_key_to_cpu(path->nodes[level], key,
3801 path->slots[level] + 1);
3807 const char* btrfs_group_type_str(u64 flag)
3809 u64 mask = BTRFS_BLOCK_GROUP_TYPE_MASK |
3810 BTRFS_SPACE_INFO_GLOBAL_RSV;
3812 switch (flag & mask) {
3813 case BTRFS_BLOCK_GROUP_DATA:
3815 case BTRFS_BLOCK_GROUP_SYSTEM:
3817 case BTRFS_BLOCK_GROUP_METADATA:
3819 case BTRFS_BLOCK_GROUP_DATA|BTRFS_BLOCK_GROUP_METADATA:
3820 return "Data+Metadata";
3821 case BTRFS_SPACE_INFO_GLOBAL_RSV:
3822 return "GlobalReserve";
3828 const char* btrfs_group_profile_str(u64 flag)
3830 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3833 case BTRFS_BLOCK_GROUP_RAID0:
3835 case BTRFS_BLOCK_GROUP_RAID1:
3837 case BTRFS_BLOCK_GROUP_RAID5:
3839 case BTRFS_BLOCK_GROUP_RAID6:
3841 case BTRFS_BLOCK_GROUP_DUP:
3843 case BTRFS_BLOCK_GROUP_RAID10:
3850 u64 disk_size(const char *path)
3854 if (statfs(path, &sfs) < 0)
3857 return sfs.f_bsize * sfs.f_blocks;
3860 u64 get_partition_size(const char *dev)
3863 int fd = open(dev, O_RDONLY);
3867 if (ioctl(fd, BLKGETSIZE64, &result) < 0) {
3877 * Check if the BTRFS_IOC_TREE_SEARCH_V2 ioctl is supported on a given
3878 * filesystem, opened at fd
3880 int btrfs_tree_search2_ioctl_supported(int fd)
3882 struct btrfs_ioctl_search_args_v2 *args2;
3883 struct btrfs_ioctl_search_key *sk;
3884 int args2_size = 1024;
3885 char args2_buf[args2_size];
3888 args2 = (struct btrfs_ioctl_search_args_v2 *)args2_buf;
3892 * Search for the extent tree item in the root tree.
3894 sk->tree_id = BTRFS_ROOT_TREE_OBJECTID;
3895 sk->min_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3896 sk->max_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3897 sk->min_type = BTRFS_ROOT_ITEM_KEY;
3898 sk->max_type = BTRFS_ROOT_ITEM_KEY;
3900 sk->max_offset = (u64)-1;
3901 sk->min_transid = 0;
3902 sk->max_transid = (u64)-1;
3904 args2->buf_size = args2_size - sizeof(struct btrfs_ioctl_search_args_v2);
3905 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH_V2, args2);
3906 if (ret == -EOPNOTSUPP)
3913 int btrfs_check_nodesize(u32 nodesize, u32 sectorsize, u64 features)
3915 if (nodesize < sectorsize) {
3916 error("illegal nodesize %u (smaller than %u)",
3917 nodesize, sectorsize);
3919 } else if (nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) {
3920 error("illegal nodesize %u (larger than %u)",
3921 nodesize, BTRFS_MAX_METADATA_BLOCKSIZE);
3923 } else if (nodesize & (sectorsize - 1)) {
3924 error("illegal nodesize %u (not aligned to %u)",
3925 nodesize, sectorsize);
3927 } else if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS &&
3928 nodesize != sectorsize) {
3929 error("illegal nodesize %u (not equal to %u for mixed block group)",
3930 nodesize, sectorsize);
3937 * Copy a path argument from SRC to DEST and check the SRC length if it's at
3938 * most PATH_MAX and fits into DEST. DESTLEN is supposed to be exact size of
3940 * The destination buffer is zero terminated.
3941 * Return < 0 for error, 0 otherwise.
3943 int arg_copy_path(char *dest, const char *src, int destlen)
3945 size_t len = strlen(src);
3947 if (len >= PATH_MAX || len >= destlen)
3948 return -ENAMETOOLONG;
3950 __strncpy_null(dest, src, destlen);
3955 unsigned int get_unit_mode_from_arg(int *argc, char *argv[], int df_mode)
3957 unsigned int unit_mode = UNITS_DEFAULT;
3961 for (arg_i = 0; arg_i < *argc; arg_i++) {
3962 if (!strcmp(argv[arg_i], "--"))
3965 if (!strcmp(argv[arg_i], "--raw")) {
3966 unit_mode = UNITS_RAW;
3970 if (!strcmp(argv[arg_i], "--human-readable")) {
3971 unit_mode = UNITS_HUMAN_BINARY;
3976 if (!strcmp(argv[arg_i], "--iec")) {
3977 units_set_mode(&unit_mode, UNITS_BINARY);
3981 if (!strcmp(argv[arg_i], "--si")) {
3982 units_set_mode(&unit_mode, UNITS_DECIMAL);
3987 if (!strcmp(argv[arg_i], "--kbytes")) {
3988 units_set_base(&unit_mode, UNITS_KBYTES);
3992 if (!strcmp(argv[arg_i], "--mbytes")) {
3993 units_set_base(&unit_mode, UNITS_MBYTES);
3997 if (!strcmp(argv[arg_i], "--gbytes")) {
3998 units_set_base(&unit_mode, UNITS_GBYTES);
4002 if (!strcmp(argv[arg_i], "--tbytes")) {
4003 units_set_base(&unit_mode, UNITS_TBYTES);
4011 if (!strcmp(argv[arg_i], "-b")) {
4012 unit_mode = UNITS_RAW;
4016 if (!strcmp(argv[arg_i], "-h")) {
4017 unit_mode = UNITS_HUMAN_BINARY;
4021 if (!strcmp(argv[arg_i], "-H")) {
4022 unit_mode = UNITS_HUMAN_DECIMAL;
4026 if (!strcmp(argv[arg_i], "-k")) {
4027 units_set_base(&unit_mode, UNITS_KBYTES);
4031 if (!strcmp(argv[arg_i], "-m")) {
4032 units_set_base(&unit_mode, UNITS_MBYTES);
4036 if (!strcmp(argv[arg_i], "-g")) {
4037 units_set_base(&unit_mode, UNITS_GBYTES);
4041 if (!strcmp(argv[arg_i], "-t")) {
4042 units_set_base(&unit_mode, UNITS_TBYTES);
4048 for (arg_i = 0, arg_end = 0; arg_i < *argc; arg_i++) {
4051 argv[arg_end] = argv[arg_i];
4060 int string_is_numerical(const char *str)
4064 if (!(*str >= '0' && *str <= '9'))
4066 while (*str >= '0' && *str <= '9')
4074 * Preprocess @argv with getopt_long to reorder options and consume the "--"
4076 * Unknown short and long options are reported, optionally the @usage is printed
4079 void clean_args_no_options(int argc, char *argv[], const char * const *usagestr)
4081 static const struct option long_options[] = {
4086 int c = getopt_long(argc, argv, "", long_options, NULL);
4100 * Same as clean_args_no_options but pass through arguments that could look
4101 * like short options. Eg. reisze which takes a negative resize argument like
4104 * This accepts only two forms:
4105 * - "-- option1 option2 ..."
4106 * - "option1 option2 ..."
4108 void clean_args_no_options_relaxed(int argc, char *argv[], const char * const *usagestr)
4113 if (strcmp(argv[1], "--") == 0)
4117 /* Subvolume helper functions */
4119 * test if name is a correct subvolume name
4120 * this function return
4121 * 0-> name is not a correct subvolume name
4122 * 1-> name is a correct subvolume name
4124 int test_issubvolname(const char *name)
4126 return name[0] != '\0' && !strchr(name, '/') &&
4127 strcmp(name, ".") && strcmp(name, "..");
4131 * Test if path is a subvolume
4133 * 0 - path exists but it is not a subvolume
4134 * 1 - path exists and it is a subvolume
4137 int test_issubvolume(const char *path)
4143 res = stat(path, &st);
4147 if (st.st_ino != BTRFS_FIRST_FREE_OBJECTID || !S_ISDIR(st.st_mode))
4150 res = statfs(path, &stfs);
4154 return (int)stfs.f_type == BTRFS_SUPER_MAGIC;
4157 const char *subvol_strip_mountpoint(const char *mnt, const char *full_path)
4159 int len = strlen(mnt);
4163 if (mnt[len - 1] != '/')
4166 return full_path + len;
4173 * 1: Error; and error info printed to the terminal. Fixme.
4174 * 2: If the fullpath is root tree instead of subvol tree
4176 int get_subvol_info(const char *fullpath, struct root_info *get_ri)
4183 const char *svpath = NULL;
4184 DIR *dirstream1 = NULL;
4185 DIR *dirstream2 = NULL;
4187 ret = test_issubvolume(fullpath);
4191 error("not a subvolume: %s", fullpath);
4195 ret = find_mount_root(fullpath, &mnt);
4199 error("%s doesn't belong to btrfs mount point", fullpath);
4203 svpath = subvol_strip_mountpoint(mnt, fullpath);
4205 fd = btrfs_open_dir(fullpath, &dirstream1, 1);
4209 ret = btrfs_list_get_path_rootid(fd, &sv_id);
4213 mntfd = btrfs_open_dir(mnt, &dirstream2, 1);
4217 memset(get_ri, 0, sizeof(*get_ri));
4218 get_ri->root_id = sv_id;
4220 if (sv_id == BTRFS_FS_TREE_OBJECTID)
4221 ret = btrfs_get_toplevel_subvol(mntfd, get_ri);
4223 ret = btrfs_get_subvol(mntfd, get_ri);
4225 error("can't find '%s': %d", svpath, ret);
4228 close_file_or_dir(mntfd, dirstream2);
4229 close_file_or_dir(fd, dirstream1);
4235 void init_rand_seed(u64 seed)
4239 /* only use the last 48 bits */
4240 for (i = 0; i < 3; i++) {
4241 rand_seed[i] = (unsigned short)(seed ^ (unsigned short)(-1));
4244 rand_seed_initlized = 1;
4247 static void __init_seed(void)
4253 if(rand_seed_initlized)
4255 /* Use urandom as primary seed source. */
4256 fd = open("/dev/urandom", O_RDONLY);
4258 ret = read(fd, rand_seed, sizeof(rand_seed));
4260 if (ret < sizeof(rand_seed))
4264 /* Use time and pid as fallback seed */
4265 warning("failed to read /dev/urandom, use time and pid as random seed");
4266 gettimeofday(&tv, 0);
4267 rand_seed[0] = getpid() ^ (tv.tv_sec & 0xFFFF);
4268 rand_seed[1] = getppid() ^ (tv.tv_usec & 0xFFFF);
4269 rand_seed[2] = (tv.tv_sec ^ tv.tv_usec) >> 16;
4271 rand_seed_initlized = 1;
4278 * Don't use nrand48, its range is [0,2^31) The highest bit will alwasy
4279 * be 0. Use jrand48 to include the highest bit.
4281 return (u32)jrand48(rand_seed);
4284 unsigned int rand_range(unsigned int upper)
4288 * Use the full 48bits to mod, which would be more uniformly
4291 return (unsigned int)(jrand48(rand_seed) % upper);