2 * Copyright (C) 2007 Oracle. All rights reserved.
3 * Copyright (C) 2008 Morey Roof. All rights reserved.
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public
7 * License v2 as published by the Free Software Foundation.
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * General Public License for more details.
14 * You should have received a copy of the GNU General Public
15 * License along with this program; if not, write to the
16 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
17 * Boston, MA 021110-1307, USA.
23 #include <sys/ioctl.h>
24 #include <sys/mount.h>
25 #include <sys/types.h>
27 #include <uuid/uuid.h>
32 #include <linux/loop.h>
33 #include <linux/major.h>
34 #include <linux/kdev_t.h>
36 #include <blkid/blkid.h>
38 #include <sys/statfs.h>
39 #include <linux/magic.h>
42 #include "kerncompat.h"
43 #include "radix-tree.h"
46 #include "transaction.h"
54 #define BLKDISCARD _IO(0x12,119)
57 static int btrfs_scan_done = 0;
59 static char argv0_buf[ARGV0_BUF_SIZE] = "btrfs";
61 static int rand_seed_initlized = 0;
62 static unsigned short rand_seed[3];
64 const char *get_argv0_buf(void)
69 void fixup_argv0(char **argv, const char *token)
71 int len = strlen(argv0_buf);
73 snprintf(argv0_buf + len, sizeof(argv0_buf) - len, " %s", token);
77 void set_argv0(char **argv)
79 strncpy(argv0_buf, argv[0], sizeof(argv0_buf));
80 argv0_buf[sizeof(argv0_buf) - 1] = 0;
83 int check_argc_exact(int nargs, int expected)
86 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
88 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
90 return nargs != expected;
93 int check_argc_min(int nargs, int expected)
95 if (nargs < expected) {
96 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
103 int check_argc_max(int nargs, int expected)
105 if (nargs > expected) {
106 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
115 * Discard the given range in one go
117 static int discard_range(int fd, u64 start, u64 len)
119 u64 range[2] = { start, len };
121 if (ioctl(fd, BLKDISCARD, &range) < 0)
127 * Discard blocks in the given range in 1G chunks, the process is interruptible
129 static int discard_blocks(int fd, u64 start, u64 len)
133 u64 chunk_size = min_t(u64, len, 1*1024*1024*1024);
136 ret = discard_range(fd, start, chunk_size);
146 static u64 reference_root_table[] = {
147 [1] = BTRFS_ROOT_TREE_OBJECTID,
148 [2] = BTRFS_EXTENT_TREE_OBJECTID,
149 [3] = BTRFS_CHUNK_TREE_OBJECTID,
150 [4] = BTRFS_DEV_TREE_OBJECTID,
151 [5] = BTRFS_FS_TREE_OBJECTID,
152 [6] = BTRFS_CSUM_TREE_OBJECTID,
155 int test_uuid_unique(char *fs_uuid)
158 blkid_dev_iterate iter = NULL;
159 blkid_dev dev = NULL;
160 blkid_cache cache = NULL;
162 if (blkid_get_cache(&cache, NULL) < 0) {
163 printf("ERROR: lblkid cache get failed\n");
166 blkid_probe_all(cache);
167 iter = blkid_dev_iterate_begin(cache);
168 blkid_dev_set_search(iter, "UUID", fs_uuid);
170 while (blkid_dev_next(iter, &dev) == 0) {
171 dev = blkid_verify(cache, dev);
178 blkid_dev_iterate_end(iter);
179 blkid_put_cache(cache);
185 * Reserve space from free_tree.
186 * The algorithm is very simple, find the first cache_extent with enough space
187 * and allocate from its beginning.
189 static int reserve_free_space(struct cache_tree *free_tree, u64 len,
192 struct cache_extent *cache;
195 ASSERT(ret_start != NULL);
196 cache = first_cache_extent(free_tree);
198 if (cache->size > len) {
200 *ret_start = cache->start;
203 if (cache->size == 0) {
204 remove_cache_extent(free_tree, cache);
211 cache = next_cache_extent(cache);
218 static inline int write_temp_super(int fd, struct btrfs_super_block *sb,
224 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
225 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
226 btrfs_csum_final(crc, (char *)&sb->csum[0]);
227 ret = pwrite(fd, sb, BTRFS_SUPER_INFO_SIZE, sb_bytenr);
228 if (ret < BTRFS_SUPER_INFO_SIZE)
229 ret = (ret < 0 ? -errno : -EIO);
236 * Setup temporary superblock at cfg->super_bynter
237 * Needed info are extracted from cfg, and root_bytenr, chunk_bytenr
239 * For now sys chunk array will be empty and dev_item is empty too.
240 * They will be re-initialized at temp chunk tree setup.
242 * The superblock signature is not valid, denotes a partially created
243 * filesystem, needs to be finalized.
245 static int setup_temp_super(int fd, struct btrfs_mkfs_config *cfg,
246 u64 root_bytenr, u64 chunk_bytenr)
248 unsigned char chunk_uuid[BTRFS_UUID_SIZE];
249 char super_buf[BTRFS_SUPER_INFO_SIZE];
250 struct btrfs_super_block *super = (struct btrfs_super_block *)super_buf;
253 memset(super_buf, 0, BTRFS_SUPER_INFO_SIZE);
254 cfg->num_bytes = round_down(cfg->num_bytes, cfg->sectorsize);
257 if (uuid_parse(cfg->fs_uuid, super->fsid) != 0) {
258 error("cound not parse UUID: %s", cfg->fs_uuid);
262 if (!test_uuid_unique(cfg->fs_uuid)) {
263 error("non-unique UUID: %s", cfg->fs_uuid);
268 uuid_generate(super->fsid);
269 uuid_unparse(super->fsid, cfg->fs_uuid);
271 uuid_generate(chunk_uuid);
272 uuid_unparse(chunk_uuid, cfg->chunk_uuid);
274 btrfs_set_super_bytenr(super, cfg->super_bytenr);
275 btrfs_set_super_num_devices(super, 1);
276 btrfs_set_super_magic(super, BTRFS_MAGIC_PARTIAL);
277 btrfs_set_super_generation(super, 1);
278 btrfs_set_super_root(super, root_bytenr);
279 btrfs_set_super_chunk_root(super, chunk_bytenr);
280 btrfs_set_super_total_bytes(super, cfg->num_bytes);
282 * Temporary filesystem will only have 6 tree roots:
283 * chunk tree, root tree, extent_tree, device tree, fs tree
286 btrfs_set_super_bytes_used(super, 6 * cfg->nodesize);
287 btrfs_set_super_sectorsize(super, cfg->sectorsize);
288 btrfs_set_super_leafsize(super, cfg->nodesize);
289 btrfs_set_super_nodesize(super, cfg->nodesize);
290 btrfs_set_super_stripesize(super, cfg->stripesize);
291 btrfs_set_super_csum_type(super, BTRFS_CSUM_TYPE_CRC32);
292 btrfs_set_super_chunk_root(super, chunk_bytenr);
293 btrfs_set_super_cache_generation(super, -1);
294 btrfs_set_super_incompat_flags(super, cfg->features);
296 __strncpy_null(super->label, cfg->label, BTRFS_LABEL_SIZE - 1);
298 /* Sys chunk array will be re-initialized at chunk tree init time */
299 super->sys_chunk_array_size = 0;
301 ret = write_temp_super(fd, super, cfg->super_bytenr);
307 * Setup an extent buffer for tree block.
309 static int setup_temp_extent_buffer(struct extent_buffer *buf,
310 struct btrfs_mkfs_config *cfg,
311 u64 bytenr, u64 owner)
313 unsigned char fsid[BTRFS_FSID_SIZE];
314 unsigned char chunk_uuid[BTRFS_UUID_SIZE];
317 ret = uuid_parse(cfg->fs_uuid, fsid);
320 ret = uuid_parse(cfg->chunk_uuid, chunk_uuid);
324 memset(buf->data, 0, cfg->nodesize);
325 buf->len = cfg->nodesize;
326 btrfs_set_header_bytenr(buf, bytenr);
327 btrfs_set_header_generation(buf, 1);
328 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
329 btrfs_set_header_owner(buf, owner);
330 btrfs_set_header_flags(buf, BTRFS_HEADER_FLAG_WRITTEN);
331 write_extent_buffer(buf, chunk_uuid, btrfs_header_chunk_tree_uuid(buf),
333 write_extent_buffer(buf, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
337 static inline int write_temp_extent_buffer(int fd, struct extent_buffer *buf,
342 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
344 /* Temporary extent buffer is always mapped 1:1 on disk */
345 ret = pwrite(fd, buf->data, buf->len, bytenr);
347 ret = (ret < 0 ? ret : -EIO);
354 * Insert a root item for temporary tree root
356 * Only used in make_btrfs_v2().
358 static void insert_temp_root_item(struct extent_buffer *buf,
359 struct btrfs_mkfs_config *cfg,
360 int *slot, u32 *itemoff, u64 objectid,
363 struct btrfs_root_item root_item;
364 struct btrfs_inode_item *inode_item;
365 struct btrfs_disk_key disk_key;
367 btrfs_set_header_nritems(buf, *slot + 1);
368 (*itemoff) -= sizeof(root_item);
369 memset(&root_item, 0, sizeof(root_item));
370 inode_item = &root_item.inode;
371 btrfs_set_stack_inode_generation(inode_item, 1);
372 btrfs_set_stack_inode_size(inode_item, 3);
373 btrfs_set_stack_inode_nlink(inode_item, 1);
374 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
375 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
376 btrfs_set_root_refs(&root_item, 1);
377 btrfs_set_root_used(&root_item, cfg->nodesize);
378 btrfs_set_root_generation(&root_item, 1);
379 btrfs_set_root_bytenr(&root_item, bytenr);
381 memset(&disk_key, 0, sizeof(disk_key));
382 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
383 btrfs_set_disk_key_objectid(&disk_key, objectid);
384 btrfs_set_disk_key_offset(&disk_key, 0);
386 btrfs_set_item_key(buf, &disk_key, *slot);
387 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
388 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(root_item));
389 write_extent_buffer(buf, &root_item,
390 btrfs_item_ptr_offset(buf, *slot),
395 static int setup_temp_root_tree(int fd, struct btrfs_mkfs_config *cfg,
396 u64 root_bytenr, u64 extent_bytenr,
397 u64 dev_bytenr, u64 fs_bytenr, u64 csum_bytenr)
399 struct extent_buffer *buf = NULL;
400 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
405 * Provided bytenr must in ascending order, or tree root will have a
408 if (!(root_bytenr < extent_bytenr && extent_bytenr < dev_bytenr &&
409 dev_bytenr < fs_bytenr && fs_bytenr < csum_bytenr)) {
410 error("bad tree bytenr order: "
411 "root < extent %llu < %llu, "
412 "extent < dev %llu < %llu, "
413 "dev < fs %llu < %llu, "
414 "fs < csum %llu < %llu",
415 (unsigned long long)root_bytenr,
416 (unsigned long long)extent_bytenr,
417 (unsigned long long)extent_bytenr,
418 (unsigned long long)dev_bytenr,
419 (unsigned long long)dev_bytenr,
420 (unsigned long long)fs_bytenr,
421 (unsigned long long)fs_bytenr,
422 (unsigned long long)csum_bytenr);
425 buf = malloc(sizeof(*buf) + cfg->nodesize);
429 ret = setup_temp_extent_buffer(buf, cfg, root_bytenr,
430 BTRFS_ROOT_TREE_OBJECTID);
434 insert_temp_root_item(buf, cfg, &slot, &itemoff,
435 BTRFS_EXTENT_TREE_OBJECTID, extent_bytenr);
436 insert_temp_root_item(buf, cfg, &slot, &itemoff,
437 BTRFS_DEV_TREE_OBJECTID, dev_bytenr);
438 insert_temp_root_item(buf, cfg, &slot, &itemoff,
439 BTRFS_FS_TREE_OBJECTID, fs_bytenr);
440 insert_temp_root_item(buf, cfg, &slot, &itemoff,
441 BTRFS_CSUM_TREE_OBJECTID, csum_bytenr);
443 ret = write_temp_extent_buffer(fd, buf, root_bytenr);
449 static int insert_temp_dev_item(int fd, struct extent_buffer *buf,
450 struct btrfs_mkfs_config *cfg,
451 int *slot, u32 *itemoff)
453 struct btrfs_disk_key disk_key;
454 struct btrfs_dev_item *dev_item;
455 char super_buf[BTRFS_SUPER_INFO_SIZE];
456 unsigned char dev_uuid[BTRFS_UUID_SIZE];
457 unsigned char fsid[BTRFS_FSID_SIZE];
458 struct btrfs_super_block *super = (struct btrfs_super_block *)super_buf;
461 ret = pread(fd, super_buf, BTRFS_SUPER_INFO_SIZE, cfg->super_bytenr);
462 if (ret < BTRFS_SUPER_INFO_SIZE) {
463 ret = (ret < 0 ? -errno : -EIO);
467 btrfs_set_header_nritems(buf, *slot + 1);
468 (*itemoff) -= sizeof(*dev_item);
469 /* setup device item 1, 0 is for replace case */
470 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
471 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
472 btrfs_set_disk_key_offset(&disk_key, 1);
473 btrfs_set_item_key(buf, &disk_key, *slot);
474 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
475 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(*dev_item));
477 dev_item = btrfs_item_ptr(buf, *slot, struct btrfs_dev_item);
478 /* Generate device uuid */
479 uuid_generate(dev_uuid);
480 write_extent_buffer(buf, dev_uuid,
481 (unsigned long)btrfs_device_uuid(dev_item),
483 uuid_parse(cfg->fs_uuid, fsid);
484 write_extent_buffer(buf, fsid,
485 (unsigned long)btrfs_device_fsid(dev_item),
487 btrfs_set_device_id(buf, dev_item, 1);
488 btrfs_set_device_generation(buf, dev_item, 0);
489 btrfs_set_device_total_bytes(buf, dev_item, cfg->num_bytes);
491 * The number must match the initial SYSTEM and META chunk size
493 btrfs_set_device_bytes_used(buf, dev_item,
494 BTRFS_MKFS_SYSTEM_GROUP_SIZE +
495 BTRFS_CONVERT_META_GROUP_SIZE);
496 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
497 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
498 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
499 btrfs_set_device_type(buf, dev_item, 0);
501 /* Super dev_item is not complete, copy the complete one to sb */
502 read_extent_buffer(buf, &super->dev_item, (unsigned long)dev_item,
504 ret = write_temp_super(fd, super, cfg->super_bytenr);
510 static int insert_temp_chunk_item(int fd, struct extent_buffer *buf,
511 struct btrfs_mkfs_config *cfg,
512 int *slot, u32 *itemoff, u64 start, u64 len,
515 struct btrfs_chunk *chunk;
516 struct btrfs_disk_key disk_key;
517 char super_buf[BTRFS_SUPER_INFO_SIZE];
518 struct btrfs_super_block *sb = (struct btrfs_super_block *)super_buf;
521 ret = pread(fd, super_buf, BTRFS_SUPER_INFO_SIZE,
523 if (ret < BTRFS_SUPER_INFO_SIZE) {
524 ret = (ret < 0 ? ret : -EIO);
528 btrfs_set_header_nritems(buf, *slot + 1);
529 (*itemoff) -= btrfs_chunk_item_size(1);
530 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
531 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
532 btrfs_set_disk_key_offset(&disk_key, start);
533 btrfs_set_item_key(buf, &disk_key, *slot);
534 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
535 btrfs_set_item_size(buf, btrfs_item_nr(*slot),
536 btrfs_chunk_item_size(1));
538 chunk = btrfs_item_ptr(buf, *slot, struct btrfs_chunk);
539 btrfs_set_chunk_length(buf, chunk, len);
540 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
541 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
542 btrfs_set_chunk_type(buf, chunk, type);
543 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
544 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
545 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
546 btrfs_set_chunk_num_stripes(buf, chunk, 1);
547 /* TODO: Support DUP profile for system chunk */
548 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
549 /* We are doing 1:1 mapping, so start is its dev offset */
550 btrfs_set_stripe_offset_nr(buf, chunk, 0, start);
551 write_extent_buffer(buf, &sb->dev_item.uuid,
552 (unsigned long)btrfs_stripe_dev_uuid_nr(chunk, 0),
557 * If it's system chunk, also copy it to super block.
559 if (type & BTRFS_BLOCK_GROUP_SYSTEM) {
562 cur = (char *)sb->sys_chunk_array + sb->sys_chunk_array_size;
563 memcpy(cur, &disk_key, sizeof(disk_key));
564 cur += sizeof(disk_key);
565 read_extent_buffer(buf, cur, (unsigned long int)chunk,
566 btrfs_chunk_item_size(1));
567 sb->sys_chunk_array_size += btrfs_chunk_item_size(1) +
570 ret = write_temp_super(fd, sb, cfg->super_bytenr);
575 static int setup_temp_chunk_tree(int fd, struct btrfs_mkfs_config *cfg,
576 u64 sys_chunk_start, u64 meta_chunk_start,
579 struct extent_buffer *buf = NULL;
580 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
584 /* Must ensure SYS chunk starts before META chunk */
585 if (meta_chunk_start < sys_chunk_start) {
586 error("wrong chunk order: meta < system %llu < %llu",
587 (unsigned long long)meta_chunk_start,
588 (unsigned long long)sys_chunk_start);
591 buf = malloc(sizeof(*buf) + cfg->nodesize);
594 ret = setup_temp_extent_buffer(buf, cfg, chunk_bytenr,
595 BTRFS_CHUNK_TREE_OBJECTID);
599 ret = insert_temp_dev_item(fd, buf, cfg, &slot, &itemoff);
602 ret = insert_temp_chunk_item(fd, buf, cfg, &slot, &itemoff,
604 BTRFS_MKFS_SYSTEM_GROUP_SIZE,
605 BTRFS_BLOCK_GROUP_SYSTEM);
608 ret = insert_temp_chunk_item(fd, buf, cfg, &slot, &itemoff,
610 BTRFS_CONVERT_META_GROUP_SIZE,
611 BTRFS_BLOCK_GROUP_METADATA);
614 ret = write_temp_extent_buffer(fd, buf, chunk_bytenr);
621 static void insert_temp_dev_extent(struct extent_buffer *buf,
622 int *slot, u32 *itemoff, u64 start, u64 len)
624 struct btrfs_dev_extent *dev_extent;
625 struct btrfs_disk_key disk_key;
627 btrfs_set_header_nritems(buf, *slot + 1);
628 (*itemoff) -= sizeof(*dev_extent);
629 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
630 btrfs_set_disk_key_objectid(&disk_key, 1);
631 btrfs_set_disk_key_offset(&disk_key, start);
632 btrfs_set_item_key(buf, &disk_key, *slot);
633 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
634 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(*dev_extent));
636 dev_extent = btrfs_item_ptr(buf, *slot, struct btrfs_dev_extent);
637 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
638 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
639 btrfs_set_dev_extent_length(buf, dev_extent, len);
640 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, start);
641 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
642 BTRFS_CHUNK_TREE_OBJECTID);
646 static int setup_temp_dev_tree(int fd, struct btrfs_mkfs_config *cfg,
647 u64 sys_chunk_start, u64 meta_chunk_start,
650 struct extent_buffer *buf = NULL;
651 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
655 /* Must ensure SYS chunk starts before META chunk */
656 if (meta_chunk_start < sys_chunk_start) {
657 error("wrong chunk order: meta < system %llu < %llu",
658 (unsigned long long)meta_chunk_start,
659 (unsigned long long)sys_chunk_start);
662 buf = malloc(sizeof(*buf) + cfg->nodesize);
665 ret = setup_temp_extent_buffer(buf, cfg, dev_bytenr,
666 BTRFS_DEV_TREE_OBJECTID);
669 insert_temp_dev_extent(buf, &slot, &itemoff, sys_chunk_start,
670 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
671 insert_temp_dev_extent(buf, &slot, &itemoff, meta_chunk_start,
672 BTRFS_CONVERT_META_GROUP_SIZE);
673 ret = write_temp_extent_buffer(fd, buf, dev_bytenr);
679 static int setup_temp_fs_tree(int fd, struct btrfs_mkfs_config *cfg,
682 struct extent_buffer *buf = NULL;
685 buf = malloc(sizeof(*buf) + cfg->nodesize);
688 ret = setup_temp_extent_buffer(buf, cfg, fs_bytenr,
689 BTRFS_FS_TREE_OBJECTID);
693 * Temporary fs tree is completely empty.
695 ret = write_temp_extent_buffer(fd, buf, fs_bytenr);
701 static int setup_temp_csum_tree(int fd, struct btrfs_mkfs_config *cfg,
704 struct extent_buffer *buf = NULL;
707 buf = malloc(sizeof(*buf) + cfg->nodesize);
710 ret = setup_temp_extent_buffer(buf, cfg, csum_bytenr,
711 BTRFS_CSUM_TREE_OBJECTID);
715 * Temporary csum tree is completely empty.
717 ret = write_temp_extent_buffer(fd, buf, csum_bytenr);
724 * Insert one temporary extent item.
726 * NOTE: if skinny_metadata is not enabled, this function must be called
727 * after all other trees are initialized.
728 * Or fs without skinny-metadata will be screwed up.
730 static int insert_temp_extent_item(int fd, struct extent_buffer *buf,
731 struct btrfs_mkfs_config *cfg,
732 int *slot, u32 *itemoff, u64 bytenr,
735 struct extent_buffer *tmp;
736 struct btrfs_extent_item *ei;
737 struct btrfs_extent_inline_ref *iref;
738 struct btrfs_disk_key disk_key;
739 struct btrfs_disk_key tree_info_key;
740 struct btrfs_tree_block_info *info;
742 int skinny_metadata = cfg->features &
743 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA;
747 itemsize = sizeof(*ei) + sizeof(*iref);
749 itemsize = sizeof(*ei) + sizeof(*iref) +
750 sizeof(struct btrfs_tree_block_info);
752 btrfs_set_header_nritems(buf, *slot + 1);
753 *(itemoff) -= itemsize;
755 if (skinny_metadata) {
756 btrfs_set_disk_key_type(&disk_key, BTRFS_METADATA_ITEM_KEY);
757 btrfs_set_disk_key_offset(&disk_key, 0);
759 btrfs_set_disk_key_type(&disk_key, BTRFS_EXTENT_ITEM_KEY);
760 btrfs_set_disk_key_offset(&disk_key, cfg->nodesize);
762 btrfs_set_disk_key_objectid(&disk_key, bytenr);
764 btrfs_set_item_key(buf, &disk_key, *slot);
765 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
766 btrfs_set_item_size(buf, btrfs_item_nr(*slot), itemsize);
768 ei = btrfs_item_ptr(buf, *slot, struct btrfs_extent_item);
769 btrfs_set_extent_refs(buf, ei, 1);
770 btrfs_set_extent_generation(buf, ei, 1);
771 btrfs_set_extent_flags(buf, ei, BTRFS_EXTENT_FLAG_TREE_BLOCK);
773 if (skinny_metadata) {
774 iref = (struct btrfs_extent_inline_ref *)(ei + 1);
776 info = (struct btrfs_tree_block_info *)(ei + 1);
777 iref = (struct btrfs_extent_inline_ref *)(info + 1);
779 btrfs_set_extent_inline_ref_type(buf, iref,
780 BTRFS_TREE_BLOCK_REF_KEY);
781 btrfs_set_extent_inline_ref_offset(buf, iref, ref_root);
788 * Lastly, check the tree block key by read the tree block
789 * Since we do 1:1 mapping for convert case, we can directly
790 * read the bytenr from disk
792 tmp = malloc(sizeof(*tmp) + cfg->nodesize);
795 ret = setup_temp_extent_buffer(tmp, cfg, bytenr, ref_root);
798 ret = pread(fd, tmp->data, cfg->nodesize, bytenr);
799 if (ret < cfg->nodesize) {
800 ret = (ret < 0 ? -errno : -EIO);
803 if (btrfs_header_nritems(tmp) == 0) {
804 btrfs_set_disk_key_type(&tree_info_key, 0);
805 btrfs_set_disk_key_objectid(&tree_info_key, 0);
806 btrfs_set_disk_key_offset(&tree_info_key, 0);
808 btrfs_item_key(tmp, &tree_info_key, 0);
810 btrfs_set_tree_block_key(buf, info, &tree_info_key);
817 static void insert_temp_block_group(struct extent_buffer *buf,
818 struct btrfs_mkfs_config *cfg,
819 int *slot, u32 *itemoff,
820 u64 bytenr, u64 len, u64 used, u64 flag)
822 struct btrfs_block_group_item bgi;
823 struct btrfs_disk_key disk_key;
825 btrfs_set_header_nritems(buf, *slot + 1);
826 (*itemoff) -= sizeof(bgi);
827 btrfs_set_disk_key_type(&disk_key, BTRFS_BLOCK_GROUP_ITEM_KEY);
828 btrfs_set_disk_key_objectid(&disk_key, bytenr);
829 btrfs_set_disk_key_offset(&disk_key, len);
830 btrfs_set_item_key(buf, &disk_key, *slot);
831 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
832 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(bgi));
834 btrfs_set_block_group_flags(&bgi, flag);
835 btrfs_set_block_group_used(&bgi, used);
836 btrfs_set_block_group_chunk_objectid(&bgi,
837 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
838 write_extent_buffer(buf, &bgi, btrfs_item_ptr_offset(buf, *slot),
843 static int setup_temp_extent_tree(int fd, struct btrfs_mkfs_config *cfg,
844 u64 chunk_bytenr, u64 root_bytenr,
845 u64 extent_bytenr, u64 dev_bytenr,
846 u64 fs_bytenr, u64 csum_bytenr)
848 struct extent_buffer *buf = NULL;
849 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
854 * We must ensure provided bytenr are in ascending order,
855 * or extent tree key order will be broken.
857 if (!(chunk_bytenr < root_bytenr && root_bytenr < extent_bytenr &&
858 extent_bytenr < dev_bytenr && dev_bytenr < fs_bytenr &&
859 fs_bytenr < csum_bytenr)) {
860 error("bad tree bytenr order: "
861 "chunk < root %llu < %llu, "
862 "root < extent %llu < %llu, "
863 "extent < dev %llu < %llu, "
864 "dev < fs %llu < %llu, "
865 "fs < csum %llu < %llu",
866 (unsigned long long)chunk_bytenr,
867 (unsigned long long)root_bytenr,
868 (unsigned long long)root_bytenr,
869 (unsigned long long)extent_bytenr,
870 (unsigned long long)extent_bytenr,
871 (unsigned long long)dev_bytenr,
872 (unsigned long long)dev_bytenr,
873 (unsigned long long)fs_bytenr,
874 (unsigned long long)fs_bytenr,
875 (unsigned long long)csum_bytenr);
878 buf = malloc(sizeof(*buf) + cfg->nodesize);
882 ret = setup_temp_extent_buffer(buf, cfg, extent_bytenr,
883 BTRFS_EXTENT_TREE_OBJECTID);
887 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
888 chunk_bytenr, BTRFS_CHUNK_TREE_OBJECTID);
892 insert_temp_block_group(buf, cfg, &slot, &itemoff, chunk_bytenr,
893 BTRFS_MKFS_SYSTEM_GROUP_SIZE, cfg->nodesize,
894 BTRFS_BLOCK_GROUP_SYSTEM);
896 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
897 root_bytenr, BTRFS_ROOT_TREE_OBJECTID);
901 /* 5 tree block used, root, extent, dev, fs and csum*/
902 insert_temp_block_group(buf, cfg, &slot, &itemoff, root_bytenr,
903 BTRFS_CONVERT_META_GROUP_SIZE, cfg->nodesize * 5,
904 BTRFS_BLOCK_GROUP_METADATA);
906 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
907 extent_bytenr, BTRFS_EXTENT_TREE_OBJECTID);
910 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
911 dev_bytenr, BTRFS_DEV_TREE_OBJECTID);
914 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
915 fs_bytenr, BTRFS_FS_TREE_OBJECTID);
918 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
919 csum_bytenr, BTRFS_CSUM_TREE_OBJECTID);
923 ret = write_temp_extent_buffer(fd, buf, extent_bytenr);
930 * Improved version of make_btrfs().
933 * 1) Do chunk allocation to avoid used data
934 * And after this function, extent type matches chunk type
935 * 2) Better structured code
936 * No super long hand written codes to initialized all tree blocks
937 * Split into small blocks and reuse codes.
938 * TODO: Reuse tree operation facilities by introducing new flags
940 static int make_convert_btrfs(int fd, struct btrfs_mkfs_config *cfg,
941 struct btrfs_convert_context *cctx)
943 struct cache_tree *free = &cctx->free;
944 struct cache_tree *used = &cctx->used;
946 u64 meta_chunk_start;
947 /* chunk tree bytenr, in system chunk */
949 /* metadata trees bytenr, in metadata chunk */
957 /* Shouldn't happen */
958 BUG_ON(cache_tree_empty(used));
961 * reserve space for temporary superblock first
962 * Here we allocate a little larger space, to keep later
963 * free space will be STRIPE_LEN aligned
965 ret = reserve_free_space(free, BTRFS_STRIPE_LEN,
971 * Then reserve system chunk space
972 * TODO: Change system group size depending on cctx->total_bytes.
973 * If using current 4M, it can only handle less than one TB for
974 * worst case and then run out of sys space.
976 ret = reserve_free_space(free, BTRFS_MKFS_SYSTEM_GROUP_SIZE,
980 ret = reserve_free_space(free, BTRFS_CONVERT_META_GROUP_SIZE,
986 * Allocated meta/sys chunks will be mapped 1:1 with device offset.
988 * Inside the allocated metadata chunk, the layout will be:
989 * | offset | contents |
990 * -------------------------------------
992 * | +nodesize | extent root |
993 * | +nodesize * 2 | device root |
994 * | +nodesize * 3 | fs tree |
995 * | +nodesize * 4 | csum tree |
996 * -------------------------------------
997 * Inside the allocated system chunk, the layout will be:
998 * | offset | contents |
999 * -------------------------------------
1000 * | +0 | chunk root |
1001 * -------------------------------------
1003 chunk_bytenr = sys_chunk_start;
1004 root_bytenr = meta_chunk_start;
1005 extent_bytenr = meta_chunk_start + cfg->nodesize;
1006 dev_bytenr = meta_chunk_start + cfg->nodesize * 2;
1007 fs_bytenr = meta_chunk_start + cfg->nodesize * 3;
1008 csum_bytenr = meta_chunk_start + cfg->nodesize * 4;
1010 ret = setup_temp_super(fd, cfg, root_bytenr, chunk_bytenr);
1014 ret = setup_temp_root_tree(fd, cfg, root_bytenr, extent_bytenr,
1015 dev_bytenr, fs_bytenr, csum_bytenr);
1018 ret = setup_temp_chunk_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
1022 ret = setup_temp_dev_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
1026 ret = setup_temp_fs_tree(fd, cfg, fs_bytenr);
1029 ret = setup_temp_csum_tree(fd, cfg, csum_bytenr);
1033 * Setup extent tree last, since it may need to read tree block key
1034 * for non-skinny metadata case.
1036 ret = setup_temp_extent_tree(fd, cfg, chunk_bytenr, root_bytenr,
1037 extent_bytenr, dev_bytenr, fs_bytenr,
1044 * @fs_uuid - if NULL, generates a UUID, returns back the new filesystem UUID
1046 * The superblock signature is not valid, denotes a partially created
1047 * filesystem, needs to be finalized.
1049 int make_btrfs(int fd, struct btrfs_mkfs_config *cfg,
1050 struct btrfs_convert_context *cctx)
1052 struct btrfs_super_block super;
1053 struct extent_buffer *buf;
1054 struct btrfs_root_item root_item;
1055 struct btrfs_disk_key disk_key;
1056 struct btrfs_extent_item *extent_item;
1057 struct btrfs_inode_item *inode_item;
1058 struct btrfs_chunk *chunk;
1059 struct btrfs_dev_item *dev_item;
1060 struct btrfs_dev_extent *dev_extent;
1061 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
1071 int skinny_metadata = !!(cfg->features &
1072 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA);
1076 return make_convert_btrfs(fd, cfg, cctx);
1077 buf = malloc(sizeof(*buf) + max(cfg->sectorsize, cfg->nodesize));
1081 first_free = BTRFS_SUPER_INFO_OFFSET + cfg->sectorsize * 2 - 1;
1082 first_free &= ~((u64)cfg->sectorsize - 1);
1084 memset(&super, 0, sizeof(super));
1086 num_bytes = (cfg->num_bytes / cfg->sectorsize) * cfg->sectorsize;
1087 if (*cfg->fs_uuid) {
1088 if (uuid_parse(cfg->fs_uuid, super.fsid) != 0) {
1089 error("cannot not parse UUID: %s", cfg->fs_uuid);
1093 if (!test_uuid_unique(cfg->fs_uuid)) {
1094 error("non-unique UUID: %s", cfg->fs_uuid);
1099 uuid_generate(super.fsid);
1101 uuid_unparse(super.fsid, cfg->fs_uuid);
1103 uuid_generate(super.dev_item.uuid);
1104 uuid_generate(chunk_tree_uuid);
1106 btrfs_set_super_bytenr(&super, cfg->blocks[0]);
1107 btrfs_set_super_num_devices(&super, 1);
1108 btrfs_set_super_magic(&super, BTRFS_MAGIC_PARTIAL);
1109 btrfs_set_super_generation(&super, 1);
1110 btrfs_set_super_root(&super, cfg->blocks[1]);
1111 btrfs_set_super_chunk_root(&super, cfg->blocks[3]);
1112 btrfs_set_super_total_bytes(&super, num_bytes);
1113 btrfs_set_super_bytes_used(&super, 6 * cfg->nodesize);
1114 btrfs_set_super_sectorsize(&super, cfg->sectorsize);
1115 btrfs_set_super_leafsize(&super, cfg->nodesize);
1116 btrfs_set_super_nodesize(&super, cfg->nodesize);
1117 btrfs_set_super_stripesize(&super, cfg->stripesize);
1118 btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
1119 btrfs_set_super_chunk_root_generation(&super, 1);
1120 btrfs_set_super_cache_generation(&super, -1);
1121 btrfs_set_super_incompat_flags(&super, cfg->features);
1123 __strncpy_null(super.label, cfg->label, BTRFS_LABEL_SIZE - 1);
1125 /* create the tree of root objects */
1126 memset(buf->data, 0, cfg->nodesize);
1127 buf->len = cfg->nodesize;
1128 btrfs_set_header_bytenr(buf, cfg->blocks[1]);
1129 btrfs_set_header_nritems(buf, 4);
1130 btrfs_set_header_generation(buf, 1);
1131 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
1132 btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
1133 write_extent_buffer(buf, super.fsid, btrfs_header_fsid(),
1136 write_extent_buffer(buf, chunk_tree_uuid,
1137 btrfs_header_chunk_tree_uuid(buf),
1140 /* create the items for the root tree */
1141 memset(&root_item, 0, sizeof(root_item));
1142 inode_item = &root_item.inode;
1143 btrfs_set_stack_inode_generation(inode_item, 1);
1144 btrfs_set_stack_inode_size(inode_item, 3);
1145 btrfs_set_stack_inode_nlink(inode_item, 1);
1146 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
1147 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
1148 btrfs_set_root_refs(&root_item, 1);
1149 btrfs_set_root_used(&root_item, cfg->nodesize);
1150 btrfs_set_root_generation(&root_item, 1);
1152 memset(&disk_key, 0, sizeof(disk_key));
1153 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
1154 btrfs_set_disk_key_offset(&disk_key, 0);
1157 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - sizeof(root_item);
1158 btrfs_set_root_bytenr(&root_item, cfg->blocks[2]);
1159 btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
1160 btrfs_set_item_key(buf, &disk_key, nritems);
1161 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1162 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1164 write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
1165 nritems), sizeof(root_item));
1168 itemoff = itemoff - sizeof(root_item);
1169 btrfs_set_root_bytenr(&root_item, cfg->blocks[4]);
1170 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
1171 btrfs_set_item_key(buf, &disk_key, nritems);
1172 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1173 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1175 write_extent_buffer(buf, &root_item,
1176 btrfs_item_ptr_offset(buf, nritems),
1180 itemoff = itemoff - sizeof(root_item);
1181 btrfs_set_root_bytenr(&root_item, cfg->blocks[5]);
1182 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
1183 btrfs_set_item_key(buf, &disk_key, nritems);
1184 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1185 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1187 write_extent_buffer(buf, &root_item,
1188 btrfs_item_ptr_offset(buf, nritems),
1192 itemoff = itemoff - sizeof(root_item);
1193 btrfs_set_root_bytenr(&root_item, cfg->blocks[6]);
1194 btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_TREE_OBJECTID);
1195 btrfs_set_item_key(buf, &disk_key, nritems);
1196 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1197 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1199 write_extent_buffer(buf, &root_item,
1200 btrfs_item_ptr_offset(buf, nritems),
1205 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1206 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[1]);
1207 if (ret != cfg->nodesize) {
1208 ret = (ret < 0 ? -errno : -EIO);
1212 /* create the items for the extent tree */
1213 memset(buf->data + sizeof(struct btrfs_header), 0,
1214 cfg->nodesize - sizeof(struct btrfs_header));
1216 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
1217 for (i = 1; i < 7; i++) {
1218 item_size = sizeof(struct btrfs_extent_item);
1219 if (!skinny_metadata)
1220 item_size += sizeof(struct btrfs_tree_block_info);
1222 BUG_ON(cfg->blocks[i] < first_free);
1223 BUG_ON(cfg->blocks[i] < cfg->blocks[i - 1]);
1225 /* create extent item */
1226 itemoff -= item_size;
1227 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
1228 if (skinny_metadata) {
1229 btrfs_set_disk_key_type(&disk_key,
1230 BTRFS_METADATA_ITEM_KEY);
1231 btrfs_set_disk_key_offset(&disk_key, 0);
1233 btrfs_set_disk_key_type(&disk_key,
1234 BTRFS_EXTENT_ITEM_KEY);
1235 btrfs_set_disk_key_offset(&disk_key, cfg->nodesize);
1237 btrfs_set_item_key(buf, &disk_key, nritems);
1238 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
1240 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1242 extent_item = btrfs_item_ptr(buf, nritems,
1243 struct btrfs_extent_item);
1244 btrfs_set_extent_refs(buf, extent_item, 1);
1245 btrfs_set_extent_generation(buf, extent_item, 1);
1246 btrfs_set_extent_flags(buf, extent_item,
1247 BTRFS_EXTENT_FLAG_TREE_BLOCK);
1250 /* create extent ref */
1251 ref_root = reference_root_table[i];
1252 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
1253 btrfs_set_disk_key_offset(&disk_key, ref_root);
1254 btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
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), 0);
1261 btrfs_set_header_bytenr(buf, cfg->blocks[2]);
1262 btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
1263 btrfs_set_header_nritems(buf, nritems);
1264 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1265 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[2]);
1266 if (ret != cfg->nodesize) {
1267 ret = (ret < 0 ? -errno : -EIO);
1271 /* create the chunk tree */
1272 memset(buf->data + sizeof(struct btrfs_header), 0,
1273 cfg->nodesize - sizeof(struct btrfs_header));
1275 item_size = sizeof(*dev_item);
1276 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - item_size;
1278 /* first device 1 (there is no device 0) */
1279 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
1280 btrfs_set_disk_key_offset(&disk_key, 1);
1281 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
1282 btrfs_set_item_key(buf, &disk_key, nritems);
1283 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1284 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
1286 dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
1287 btrfs_set_device_id(buf, dev_item, 1);
1288 btrfs_set_device_generation(buf, dev_item, 0);
1289 btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
1290 btrfs_set_device_bytes_used(buf, dev_item,
1291 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1292 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
1293 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
1294 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
1295 btrfs_set_device_type(buf, dev_item, 0);
1297 write_extent_buffer(buf, super.dev_item.uuid,
1298 (unsigned long)btrfs_device_uuid(dev_item),
1300 write_extent_buffer(buf, super.fsid,
1301 (unsigned long)btrfs_device_fsid(dev_item),
1303 read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
1307 item_size = btrfs_chunk_item_size(1);
1308 itemoff = itemoff - item_size;
1310 /* then we have chunk 0 */
1311 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1312 btrfs_set_disk_key_offset(&disk_key, 0);
1313 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
1314 btrfs_set_item_key(buf, &disk_key, nritems);
1315 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1316 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
1318 chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
1319 btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1320 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
1321 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
1322 btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
1323 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
1324 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
1325 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
1326 btrfs_set_chunk_num_stripes(buf, chunk, 1);
1327 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
1328 btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
1331 write_extent_buffer(buf, super.dev_item.uuid,
1332 (unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
1335 /* copy the key for the chunk to the system array */
1336 ptr = super.sys_chunk_array;
1337 array_size = sizeof(disk_key);
1339 memcpy(ptr, &disk_key, sizeof(disk_key));
1340 ptr += sizeof(disk_key);
1342 /* copy the chunk to the system array */
1343 read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
1344 array_size += item_size;
1346 btrfs_set_super_sys_array_size(&super, array_size);
1348 btrfs_set_header_bytenr(buf, cfg->blocks[3]);
1349 btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
1350 btrfs_set_header_nritems(buf, nritems);
1351 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1352 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[3]);
1353 if (ret != cfg->nodesize) {
1354 ret = (ret < 0 ? -errno : -EIO);
1358 /* create the device tree */
1359 memset(buf->data + sizeof(struct btrfs_header), 0,
1360 cfg->nodesize - sizeof(struct btrfs_header));
1362 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) -
1363 sizeof(struct btrfs_dev_extent);
1365 btrfs_set_disk_key_objectid(&disk_key, 1);
1366 btrfs_set_disk_key_offset(&disk_key, 0);
1367 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
1368 btrfs_set_item_key(buf, &disk_key, nritems);
1369 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1370 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1371 sizeof(struct btrfs_dev_extent));
1372 dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
1373 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
1374 BTRFS_CHUNK_TREE_OBJECTID);
1375 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
1376 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1377 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
1379 write_extent_buffer(buf, chunk_tree_uuid,
1380 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
1383 btrfs_set_dev_extent_length(buf, dev_extent,
1384 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1387 btrfs_set_header_bytenr(buf, cfg->blocks[4]);
1388 btrfs_set_header_owner(buf, BTRFS_DEV_TREE_OBJECTID);
1389 btrfs_set_header_nritems(buf, nritems);
1390 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1391 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[4]);
1392 if (ret != cfg->nodesize) {
1393 ret = (ret < 0 ? -errno : -EIO);
1397 /* create the FS root */
1398 memset(buf->data + sizeof(struct btrfs_header), 0,
1399 cfg->nodesize - sizeof(struct btrfs_header));
1400 btrfs_set_header_bytenr(buf, cfg->blocks[5]);
1401 btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
1402 btrfs_set_header_nritems(buf, 0);
1403 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1404 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[5]);
1405 if (ret != cfg->nodesize) {
1406 ret = (ret < 0 ? -errno : -EIO);
1409 /* finally create the csum root */
1410 memset(buf->data + sizeof(struct btrfs_header), 0,
1411 cfg->nodesize - sizeof(struct btrfs_header));
1412 btrfs_set_header_bytenr(buf, cfg->blocks[6]);
1413 btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
1414 btrfs_set_header_nritems(buf, 0);
1415 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1416 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[6]);
1417 if (ret != cfg->nodesize) {
1418 ret = (ret < 0 ? -errno : -EIO);
1422 /* and write out the super block */
1423 BUG_ON(sizeof(super) > cfg->sectorsize);
1424 memset(buf->data, 0, BTRFS_SUPER_INFO_SIZE);
1425 memcpy(buf->data, &super, sizeof(super));
1426 buf->len = BTRFS_SUPER_INFO_SIZE;
1427 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1428 ret = pwrite(fd, buf->data, BTRFS_SUPER_INFO_SIZE, cfg->blocks[0]);
1429 if (ret != BTRFS_SUPER_INFO_SIZE) {
1430 ret = (ret < 0 ? -errno : -EIO);
1441 static const struct btrfs_fs_feature {
1445 } mkfs_features[] = {
1446 { "mixed-bg", BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS,
1447 "mixed data and metadata block groups" },
1448 { "extref", BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF,
1449 "increased hardlink limit per file to 65536" },
1450 { "raid56", BTRFS_FEATURE_INCOMPAT_RAID56,
1451 "raid56 extended format" },
1452 { "skinny-metadata", BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA,
1453 "reduced-size metadata extent refs" },
1454 { "no-holes", BTRFS_FEATURE_INCOMPAT_NO_HOLES,
1455 "no explicit hole extents for files" },
1456 /* Keep this one last */
1457 { "list-all", BTRFS_FEATURE_LIST_ALL, NULL }
1460 static int parse_one_fs_feature(const char *name, u64 *flags)
1465 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1466 if (name[0] == '^' &&
1467 !strcmp(mkfs_features[i].name, name + 1)) {
1468 *flags &= ~ mkfs_features[i].flag;
1470 } else if (!strcmp(mkfs_features[i].name, name)) {
1471 *flags |= mkfs_features[i].flag;
1479 void btrfs_parse_features_to_string(char *buf, u64 flags)
1485 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1486 if (flags & mkfs_features[i].flag) {
1489 strcat(buf, mkfs_features[i].name);
1494 void btrfs_process_fs_features(u64 flags)
1498 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1499 if (flags & mkfs_features[i].flag) {
1500 printf("Turning ON incompat feature '%s': %s\n",
1501 mkfs_features[i].name,
1502 mkfs_features[i].desc);
1507 void btrfs_list_all_fs_features(u64 mask_disallowed)
1511 fprintf(stderr, "Filesystem features available:\n");
1512 for (i = 0; i < ARRAY_SIZE(mkfs_features) - 1; i++) {
1513 char *is_default = "";
1515 if (mkfs_features[i].flag & mask_disallowed)
1517 if (mkfs_features[i].flag & BTRFS_MKFS_DEFAULT_FEATURES)
1518 is_default = ", default";
1519 fprintf(stderr, "%-20s- %s (0x%llx%s)\n",
1520 mkfs_features[i].name,
1521 mkfs_features[i].desc,
1522 mkfs_features[i].flag,
1528 * Return NULL if all features were parsed fine, otherwise return the name of
1529 * the first unparsed.
1531 char* btrfs_parse_fs_features(char *namelist, u64 *flags)
1534 char *save_ptr = NULL; /* Satisfy static checkers */
1536 for (this_char = strtok_r(namelist, ",", &save_ptr);
1538 this_char = strtok_r(NULL, ",", &save_ptr)) {
1539 if (parse_one_fs_feature(this_char, flags))
1546 u64 btrfs_device_size(int fd, struct stat *st)
1549 if (S_ISREG(st->st_mode)) {
1552 if (!S_ISBLK(st->st_mode)) {
1555 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
1561 static int zero_blocks(int fd, off_t start, size_t len)
1563 char *buf = malloc(len);
1569 memset(buf, 0, len);
1570 written = pwrite(fd, buf, len, start);
1577 #define ZERO_DEV_BYTES (2 * 1024 * 1024)
1579 /* don't write outside the device by clamping the region to the device size */
1580 static int zero_dev_clamped(int fd, off_t start, ssize_t len, u64 dev_size)
1582 off_t end = max(start, start + len);
1585 /* and don't overwrite the disk labels on sparc */
1586 start = max(start, 1024);
1587 end = max(end, 1024);
1590 start = min_t(u64, start, dev_size);
1591 end = min_t(u64, end, dev_size);
1593 return zero_blocks(fd, start, end - start);
1596 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
1597 struct btrfs_root *root, int fd, char *path,
1598 u64 device_total_bytes, u32 io_width, u32 io_align,
1601 struct btrfs_super_block *disk_super;
1602 struct btrfs_super_block *super = root->fs_info->super_copy;
1603 struct btrfs_device *device;
1604 struct btrfs_dev_item *dev_item;
1610 device_total_bytes = (device_total_bytes / sectorsize) * sectorsize;
1612 device = kzalloc(sizeof(*device), GFP_NOFS);
1615 buf = kzalloc(sectorsize, GFP_NOFS);
1618 BUG_ON(sizeof(*disk_super) > sectorsize);
1620 disk_super = (struct btrfs_super_block *)buf;
1621 dev_item = &disk_super->dev_item;
1623 uuid_generate(device->uuid);
1626 device->io_width = io_width;
1627 device->io_align = io_align;
1628 device->sector_size = sectorsize;
1630 device->writeable = 1;
1631 device->total_bytes = device_total_bytes;
1632 device->bytes_used = 0;
1633 device->total_ios = 0;
1634 device->dev_root = root->fs_info->dev_root;
1635 device->name = strdup(path);
1639 INIT_LIST_HEAD(&device->dev_list);
1640 ret = btrfs_add_device(trans, root, device);
1643 fs_total_bytes = btrfs_super_total_bytes(super) + device_total_bytes;
1644 btrfs_set_super_total_bytes(super, fs_total_bytes);
1646 num_devs = btrfs_super_num_devices(super) + 1;
1647 btrfs_set_super_num_devices(super, num_devs);
1649 memcpy(disk_super, super, sizeof(*disk_super));
1651 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
1652 btrfs_set_stack_device_id(dev_item, device->devid);
1653 btrfs_set_stack_device_type(dev_item, device->type);
1654 btrfs_set_stack_device_io_align(dev_item, device->io_align);
1655 btrfs_set_stack_device_io_width(dev_item, device->io_width);
1656 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
1657 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
1658 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
1659 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
1661 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
1662 BUG_ON(ret != sectorsize);
1665 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
1666 device->fs_devices = root->fs_info->fs_devices;
1675 static int btrfs_wipe_existing_sb(int fd)
1677 const char *off = NULL;
1682 blkid_probe pr = NULL;
1684 pr = blkid_new_probe();
1688 if (blkid_probe_set_device(pr, fd, 0, 0)) {
1693 ret = blkid_probe_lookup_value(pr, "SBMAGIC_OFFSET", &off, NULL);
1695 ret = blkid_probe_lookup_value(pr, "SBMAGIC", NULL, &len);
1697 if (ret || len == 0 || off == NULL) {
1699 * If lookup fails, the probe did not find any values, eg. for
1700 * a file image or a loop device. Soft error.
1706 offset = strtoll(off, NULL, 10);
1707 if (len > sizeof(buf))
1710 memset(buf, 0, len);
1711 ret = pwrite(fd, buf, len, offset);
1713 error("cannot wipe existing superblock: %s", strerror(errno));
1715 } else if (ret != len) {
1716 error("cannot wipe existing superblock: wrote %d of %zd", ret, len);
1722 blkid_free_probe(pr);
1726 int btrfs_prepare_device(int fd, const char *file, u64 *block_count_ret,
1727 u64 max_block_count, unsigned opflags)
1733 ret = fstat(fd, &st);
1735 error("unable to stat %s: %s", file, strerror(errno));
1739 block_count = btrfs_device_size(fd, &st);
1740 if (block_count == 0) {
1741 error("unable to determine size of %s", file);
1744 if (max_block_count)
1745 block_count = min(block_count, max_block_count);
1747 if (opflags & PREP_DEVICE_DISCARD) {
1749 * We intentionally ignore errors from the discard ioctl. It
1750 * is not necessary for the mkfs functionality but just an
1753 if (discard_range(fd, 0, 0) == 0) {
1754 if (opflags & PREP_DEVICE_VERBOSE)
1755 printf("Performing full device TRIM (%s) ...\n",
1756 pretty_size(block_count));
1757 discard_blocks(fd, 0, block_count);
1761 ret = zero_dev_clamped(fd, 0, ZERO_DEV_BYTES, block_count);
1762 for (i = 0 ; !ret && i < BTRFS_SUPER_MIRROR_MAX; i++)
1763 ret = zero_dev_clamped(fd, btrfs_sb_offset(i),
1764 BTRFS_SUPER_INFO_SIZE, block_count);
1765 if (!ret && (opflags & PREP_DEVICE_ZERO_END))
1766 ret = zero_dev_clamped(fd, block_count - ZERO_DEV_BYTES,
1767 ZERO_DEV_BYTES, block_count);
1770 error("failed to zero device '%s': %s", file, strerror(-ret));
1774 ret = btrfs_wipe_existing_sb(fd);
1776 error("cannot wipe superblocks on %s", file);
1780 *block_count_ret = block_count;
1784 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
1785 struct btrfs_root *root, u64 objectid)
1788 struct btrfs_inode_item inode_item;
1789 time_t now = time(NULL);
1791 memset(&inode_item, 0, sizeof(inode_item));
1792 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
1793 btrfs_set_stack_inode_size(&inode_item, 0);
1794 btrfs_set_stack_inode_nlink(&inode_item, 1);
1795 btrfs_set_stack_inode_nbytes(&inode_item, root->nodesize);
1796 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
1797 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
1798 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
1799 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
1800 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
1801 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
1802 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
1803 btrfs_set_stack_timespec_sec(&inode_item.otime, 0);
1804 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
1806 if (root->fs_info->tree_root == root)
1807 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
1809 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
1813 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
1817 btrfs_set_root_dirid(&root->root_item, objectid);
1824 * checks if a path is a block device node
1825 * Returns negative errno on failure, otherwise
1826 * returns 1 for blockdev, 0 for not-blockdev
1828 int is_block_device(const char *path)
1830 struct stat statbuf;
1832 if (stat(path, &statbuf) < 0)
1835 return !!S_ISBLK(statbuf.st_mode);
1839 * check if given path is a mount point
1840 * return 1 if yes. 0 if no. -1 for error
1842 int is_mount_point(const char *path)
1848 f = setmntent("/proc/self/mounts", "r");
1852 while ((mnt = getmntent(f)) != NULL) {
1853 if (strcmp(mnt->mnt_dir, path))
1862 static int is_reg_file(const char *path)
1864 struct stat statbuf;
1866 if (stat(path, &statbuf) < 0)
1868 return S_ISREG(statbuf.st_mode);
1872 * This function checks if the given input parameter is
1874 * return <0 : some error in the given input
1875 * return BTRFS_ARG_UNKNOWN: unknown input
1876 * return BTRFS_ARG_UUID: given input is uuid
1877 * return BTRFS_ARG_MNTPOINT: given input is path
1878 * return BTRFS_ARG_REG: given input is regular file
1879 * return BTRFS_ARG_BLKDEV: given input is block device
1881 int check_arg_type(const char *input)
1884 char path[PATH_MAX];
1889 if (realpath(input, path)) {
1890 if (is_block_device(path) == 1)
1891 return BTRFS_ARG_BLKDEV;
1893 if (is_mount_point(path) == 1)
1894 return BTRFS_ARG_MNTPOINT;
1896 if (is_reg_file(path))
1897 return BTRFS_ARG_REG;
1899 return BTRFS_ARG_UNKNOWN;
1902 if (strlen(input) == (BTRFS_UUID_UNPARSED_SIZE - 1) &&
1903 !uuid_parse(input, uuid))
1904 return BTRFS_ARG_UUID;
1906 return BTRFS_ARG_UNKNOWN;
1910 * Find the mount point for a mounted device.
1911 * On success, returns 0 with mountpoint in *mp.
1912 * On failure, returns -errno (not mounted yields -EINVAL)
1913 * Is noisy on failures, expects to be given a mounted device.
1915 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size)
1920 ret = is_block_device(dev);
1923 error("not a block device: %s", dev);
1926 error("cannot check %s: %s", dev, strerror(-ret));
1931 fd = open(dev, O_RDONLY);
1934 error("cannot open %s: %s", dev, strerror(errno));
1938 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
1941 } else { /* mounted, all good */
1951 * Given a pathname, return a filehandle to:
1952 * the original pathname or,
1953 * if the pathname is a mounted btrfs device, to its mountpoint.
1955 * On error, return -1, errno should be set.
1957 int open_path_or_dev_mnt(const char *path, DIR **dirstream, int verbose)
1962 if (is_block_device(path)) {
1963 ret = get_btrfs_mount(path, mp, sizeof(mp));
1965 /* not a mounted btrfs dev */
1966 error_on(verbose, "'%s' is not a mounted btrfs device",
1971 ret = open_file_or_dir(mp, dirstream);
1972 error_on(verbose && ret < 0, "can't access '%s': %s",
1973 path, strerror(errno));
1975 ret = btrfs_open_dir(path, dirstream, 1);
1982 * Do the following checks before calling open_file_or_dir():
1983 * 1: path is in a btrfs filesystem
1984 * 2: path is a directory
1986 int btrfs_open_dir(const char *path, DIR **dirstream, int verbose)
1992 if (statfs(path, &stfs) != 0) {
1993 error_on(verbose, "cannot access '%s': %s", path,
1998 if (stfs.f_type != BTRFS_SUPER_MAGIC) {
1999 error_on(verbose, "not a btrfs filesystem: %s", path);
2003 if (stat(path, &st) != 0) {
2004 error_on(verbose, "cannot access '%s': %s", path,
2009 if (!S_ISDIR(st.st_mode)) {
2010 error_on(verbose, "not a directory: %s", path);
2014 ret = open_file_or_dir(path, dirstream);
2016 error_on(verbose, "cannot access '%s': %s", path,
2023 /* checks if a device is a loop device */
2024 static int is_loop_device (const char* device) {
2025 struct stat statbuf;
2027 if(stat(device, &statbuf) < 0)
2030 return (S_ISBLK(statbuf.st_mode) &&
2031 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
2035 * Takes a loop device path (e.g. /dev/loop0) and returns
2036 * the associated file (e.g. /images/my_btrfs.img) using
2039 static int resolve_loop_device_with_loopdev(const char* loop_dev, char* loop_file)
2043 struct loop_info64 lo64;
2045 fd = open(loop_dev, O_RDONLY | O_NONBLOCK);
2048 ret = ioctl(fd, LOOP_GET_STATUS64, &lo64);
2054 memcpy(loop_file, lo64.lo_file_name, sizeof(lo64.lo_file_name));
2055 loop_file[sizeof(lo64.lo_file_name)] = 0;
2063 /* Takes a loop device path (e.g. /dev/loop0) and returns
2064 * the associated file (e.g. /images/my_btrfs.img) */
2065 static int resolve_loop_device(const char* loop_dev, char* loop_file,
2072 char real_loop_dev[PATH_MAX];
2074 if (!realpath(loop_dev, real_loop_dev))
2076 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
2077 if (!(f = fopen(p, "r"))) {
2078 if (errno == ENOENT)
2080 * It's possibly a partitioned loop device, which is
2081 * resolvable with loopdev API.
2083 return resolve_loop_device_with_loopdev(loop_dev, loop_file);
2087 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
2088 ret = fscanf(f, fmt, loop_file);
2097 * Checks whether a and b are identical or device
2098 * files associated with the same block device
2100 static int is_same_blk_file(const char* a, const char* b)
2102 struct stat st_buf_a, st_buf_b;
2103 char real_a[PATH_MAX];
2104 char real_b[PATH_MAX];
2106 if (!realpath(a, real_a))
2107 strncpy_null(real_a, a);
2109 if (!realpath(b, real_b))
2110 strncpy_null(real_b, b);
2112 /* Identical path? */
2113 if (strcmp(real_a, real_b) == 0)
2116 if (stat(a, &st_buf_a) < 0 || stat(b, &st_buf_b) < 0) {
2117 if (errno == ENOENT)
2122 /* Same blockdevice? */
2123 if (S_ISBLK(st_buf_a.st_mode) && S_ISBLK(st_buf_b.st_mode) &&
2124 st_buf_a.st_rdev == st_buf_b.st_rdev) {
2129 if (st_buf_a.st_dev == st_buf_b.st_dev &&
2130 st_buf_a.st_ino == st_buf_b.st_ino) {
2137 /* checks if a and b are identical or device
2138 * files associated with the same block device or
2139 * if one file is a loop device that uses the other
2142 static int is_same_loop_file(const char* a, const char* b)
2144 char res_a[PATH_MAX];
2145 char res_b[PATH_MAX];
2146 const char* final_a = NULL;
2147 const char* final_b = NULL;
2150 /* Resolve a if it is a loop device */
2151 if((ret = is_loop_device(a)) < 0) {
2156 ret = resolve_loop_device(a, res_a, sizeof(res_a));
2167 /* Resolve b if it is a loop device */
2168 if ((ret = is_loop_device(b)) < 0) {
2173 ret = resolve_loop_device(b, res_b, sizeof(res_b));
2184 return is_same_blk_file(final_a, final_b);
2187 /* Checks if a file exists and is a block or regular file*/
2188 static int is_existing_blk_or_reg_file(const char* filename)
2192 if(stat(filename, &st_buf) < 0) {
2199 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
2202 /* Checks if a file is used (directly or indirectly via a loop device)
2203 * by a device in fs_devices
2205 static int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices,
2209 struct list_head *head;
2210 struct list_head *cur;
2211 struct btrfs_device *device;
2213 head = &fs_devices->devices;
2214 list_for_each(cur, head) {
2215 device = list_entry(cur, struct btrfs_device, dev_list);
2217 if((ret = is_same_loop_file(device->name, file)))
2225 * Resolve a pathname to a device mapper node to /dev/mapper/<name>
2226 * Returns NULL on invalid input or malloc failure; Other failures
2227 * will be handled by the caller using the input pathame.
2229 char *canonicalize_dm_name(const char *ptname)
2233 char path[PATH_MAX], name[PATH_MAX], *res = NULL;
2235 if (!ptname || !*ptname)
2238 snprintf(path, sizeof(path), "/sys/block/%s/dm/name", ptname);
2239 if (!(f = fopen(path, "r")))
2242 /* read <name>\n from sysfs */
2243 if (fgets(name, sizeof(name), f) && (sz = strlen(name)) > 1) {
2244 name[sz - 1] = '\0';
2245 snprintf(path, sizeof(path), "/dev/mapper/%s", name);
2247 if (access(path, F_OK) == 0)
2255 * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or
2256 * to a device mapper pathname.
2257 * Returns NULL on invalid input or malloc failure; Other failures
2258 * will be handled by the caller using the input pathame.
2260 char *canonicalize_path(const char *path)
2262 char *canonical, *p;
2264 if (!path || !*path)
2267 canonical = realpath(path, NULL);
2269 return strdup(path);
2270 p = strrchr(canonical, '/');
2271 if (p && strncmp(p, "/dm-", 4) == 0 && isdigit(*(p + 4))) {
2272 char *dm = canonicalize_dm_name(p + 1);
2283 * returns 1 if the device was mounted, < 0 on error or 0 if everything
2284 * is safe to continue.
2286 int check_mounted(const char* file)
2291 fd = open(file, O_RDONLY);
2293 error("mount check: cannot open %s: %s", file,
2298 ret = check_mounted_where(fd, file, NULL, 0, NULL);
2304 int check_mounted_where(int fd, const char *file, char *where, int size,
2305 struct btrfs_fs_devices **fs_dev_ret)
2310 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2314 /* scan the initial device */
2315 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
2316 &total_devs, BTRFS_SUPER_INFO_OFFSET, SBREAD_DEFAULT);
2317 is_btrfs = (ret >= 0);
2319 /* scan other devices */
2320 if (is_btrfs && total_devs > 1) {
2321 ret = btrfs_scan_lblkid();
2326 /* iterate over the list of currently mounted filesystems */
2327 if ((f = setmntent ("/proc/self/mounts", "r")) == NULL)
2330 while ((mnt = getmntent (f)) != NULL) {
2332 if(strcmp(mnt->mnt_type, "btrfs") != 0)
2335 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
2337 /* ignore entries in the mount table that are not
2338 associated with a file*/
2339 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
2340 goto out_mntloop_err;
2344 ret = is_same_loop_file(file, mnt->mnt_fsname);
2348 goto out_mntloop_err;
2353 /* Did we find an entry in mnt table? */
2354 if (mnt && size && where) {
2355 strncpy(where, mnt->mnt_dir, size);
2359 *fs_dev_ret = fs_devices_mnt;
2361 ret = (mnt != NULL);
2369 struct pending_dir {
2370 struct list_head list;
2371 char name[PATH_MAX];
2374 int btrfs_register_one_device(const char *fname)
2376 struct btrfs_ioctl_vol_args args;
2380 fd = open("/dev/btrfs-control", O_RDWR);
2383 "failed to open /dev/btrfs-control, skipping device registration: %s",
2387 memset(&args, 0, sizeof(args));
2388 strncpy_null(args.name, fname);
2389 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
2391 error("device scan failed on '%s': %s", fname,
2400 * Register all devices in the fs_uuid list created in the user
2401 * space. Ensure btrfs_scan_lblkid() is called before this func.
2403 int btrfs_register_all_devices(void)
2407 struct btrfs_fs_devices *fs_devices;
2408 struct btrfs_device *device;
2409 struct list_head *all_uuids;
2411 all_uuids = btrfs_scanned_uuids();
2413 list_for_each_entry(fs_devices, all_uuids, list) {
2414 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2416 err = btrfs_register_one_device(device->name);
2426 int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
2429 struct btrfs_super_block *disk_super;
2433 buf = malloc(BTRFS_SUPER_INFO_SIZE);
2438 ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
2439 if (ret != BTRFS_SUPER_INFO_SIZE)
2443 disk_super = (struct btrfs_super_block *)buf;
2445 * Accept devices from the same filesystem, allow partially created
2448 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC &&
2449 btrfs_super_magic(disk_super) != BTRFS_MAGIC_PARTIAL)
2452 if (!memcmp(disk_super->fsid, root->fs_info->super_copy->fsid,
2462 * Note: this function uses a static per-thread buffer. Do not call this
2463 * function more than 10 times within one argument list!
2465 const char *pretty_size_mode(u64 size, unsigned mode)
2467 static __thread int ps_index = 0;
2468 static __thread char ps_array[10][32];
2471 ret = ps_array[ps_index];
2474 (void)pretty_size_snprintf(size, ret, 32, mode);
2479 static const char* unit_suffix_binary[] =
2480 { "B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB"};
2481 static const char* unit_suffix_decimal[] =
2482 { "B", "kB", "MB", "GB", "TB", "PB", "EB"};
2484 int pretty_size_snprintf(u64 size, char *str, size_t str_size, unsigned unit_mode)
2490 const char** suffix = NULL;
2496 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_RAW) {
2497 snprintf(str, str_size, "%llu", size);
2501 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_BINARY) {
2504 suffix = unit_suffix_binary;
2505 } else if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_DECIMAL) {
2508 suffix = unit_suffix_decimal;
2513 fprintf(stderr, "INTERNAL ERROR: unknown unit base, mode %d\n",
2521 switch (unit_mode & UNITS_MODE_MASK) {
2522 case UNITS_TBYTES: base *= mult; num_divs++;
2523 case UNITS_GBYTES: base *= mult; num_divs++;
2524 case UNITS_MBYTES: base *= mult; num_divs++;
2525 case UNITS_KBYTES: num_divs++;
2532 while (size >= mult) {
2538 * If the value is smaller than base, we didn't do any
2539 * division, in that case, base should be 1, not original
2540 * base, or the unit will be wrong
2546 if (num_divs >= ARRAY_SIZE(unit_suffix_binary)) {
2548 printf("INTERNAL ERROR: unsupported unit suffix, index %d\n",
2553 fraction = (float)last_size / base;
2555 return snprintf(str, str_size, "%.2f%s", fraction, suffix[num_divs]);
2559 * __strncpy_null - strncpy with null termination
2560 * @dest: the target array
2561 * @src: the source string
2562 * @n: maximum bytes to copy (size of *dest)
2564 * Like strncpy, but ensures destination is null-terminated.
2566 * Copies the string pointed to by src, including the terminating null
2567 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
2568 * of n bytes. Then ensure that dest is null-terminated.
2570 char *__strncpy_null(char *dest, const char *src, size_t n)
2572 strncpy(dest, src, n);
2579 * Checks to make sure that the label matches our requirements.
2581 0 if everything is safe and usable
2582 -1 if the label is too long
2584 static int check_label(const char *input)
2586 int len = strlen(input);
2588 if (len > BTRFS_LABEL_SIZE - 1) {
2589 error("label %s is too long (max %d)", input,
2590 BTRFS_LABEL_SIZE - 1);
2597 static int set_label_unmounted(const char *dev, const char *label)
2599 struct btrfs_trans_handle *trans;
2600 struct btrfs_root *root;
2603 ret = check_mounted(dev);
2605 error("checking mount status of %s failed: %d", dev, ret);
2609 error("device %s is mounted, use mount point", dev);
2613 /* Open the super_block at the default location
2614 * and as read-write.
2616 root = open_ctree(dev, 0, OPEN_CTREE_WRITES);
2617 if (!root) /* errors are printed by open_ctree() */
2620 trans = btrfs_start_transaction(root, 1);
2621 __strncpy_null(root->fs_info->super_copy->label, label, BTRFS_LABEL_SIZE - 1);
2623 btrfs_commit_transaction(trans, root);
2625 /* Now we close it since we are done. */
2630 static int set_label_mounted(const char *mount_path, const char *labelp)
2633 char label[BTRFS_LABEL_SIZE];
2635 fd = open(mount_path, O_RDONLY | O_NOATIME);
2637 error("unable to access %s: %s", mount_path, strerror(errno));
2641 memset(label, 0, sizeof(label));
2642 __strncpy_null(label, labelp, BTRFS_LABEL_SIZE - 1);
2643 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
2644 error("unable to set label of %s: %s", mount_path,
2654 int get_label_unmounted(const char *dev, char *label)
2656 struct btrfs_root *root;
2659 ret = check_mounted(dev);
2661 error("checking mount status of %s failed: %d", dev, ret);
2665 /* Open the super_block at the default location
2668 root = open_ctree(dev, 0, 0);
2672 __strncpy_null(label, root->fs_info->super_copy->label,
2673 BTRFS_LABEL_SIZE - 1);
2675 /* Now we close it since we are done. */
2681 * If a partition is mounted, try to get the filesystem label via its
2682 * mounted path rather than device. Return the corresponding error
2683 * the user specified the device path.
2685 int get_label_mounted(const char *mount_path, char *labelp)
2687 char label[BTRFS_LABEL_SIZE];
2691 fd = open(mount_path, O_RDONLY | O_NOATIME);
2693 error("unable to access %s: %s", mount_path, strerror(errno));
2697 memset(label, '\0', sizeof(label));
2698 ret = ioctl(fd, BTRFS_IOC_GET_FSLABEL, label);
2700 if (errno != ENOTTY)
2701 error("unable to get label of %s: %s", mount_path,
2708 __strncpy_null(labelp, label, BTRFS_LABEL_SIZE - 1);
2713 int get_label(const char *btrfs_dev, char *label)
2717 ret = is_existing_blk_or_reg_file(btrfs_dev);
2719 ret = get_label_mounted(btrfs_dev, label);
2721 ret = get_label_unmounted(btrfs_dev, label);
2726 int set_label(const char *btrfs_dev, const char *label)
2730 if (check_label(label))
2733 ret = is_existing_blk_or_reg_file(btrfs_dev);
2735 ret = set_label_mounted(btrfs_dev, label);
2737 ret = set_label_unmounted(btrfs_dev, label);
2743 * A not-so-good version fls64. No fascinating optimization since
2744 * no one except parse_size use it
2746 static int fls64(u64 x)
2750 for (i = 0; i <64; i++)
2751 if (x << i & (1ULL << 63))
2756 u64 parse_size(char *s)
2764 error("size value is empty");
2768 error("size value '%s' is less equal than 0", s);
2771 ret = strtoull(s, &endptr, 10);
2773 error("size value '%s' is invalid", s);
2776 if (endptr[0] && endptr[1]) {
2777 error("illegal suffix contains character '%c' in wrong position",
2782 * strtoll returns LLONG_MAX when overflow, if this happens,
2783 * need to call strtoull to get the real size
2785 if (errno == ERANGE && ret == ULLONG_MAX) {
2786 error("size value '%s' is too large for u64", s);
2790 c = tolower(endptr[0]);
2813 error("unknown size descriptor '%c'", c);
2817 /* Check whether ret * mult overflow */
2818 if (fls64(ret) + fls64(mult) - 1 > 64) {
2819 error("size value '%s' is too large for u64", s);
2826 u64 parse_qgroupid(const char *p)
2828 char *s = strchr(p, '/');
2829 const char *ptr_src_end = p + strlen(p);
2830 char *ptr_parse_end = NULL;
2839 /* Numeric format like '0/257' is the primary case */
2841 id = strtoull(p, &ptr_parse_end, 10);
2842 if (ptr_parse_end != ptr_src_end)
2846 level = strtoull(p, &ptr_parse_end, 10);
2847 if (ptr_parse_end != s)
2850 id = strtoull(s + 1, &ptr_parse_end, 10);
2851 if (ptr_parse_end != ptr_src_end)
2854 return (level << BTRFS_QGROUP_LEVEL_SHIFT) | id;
2857 /* Path format like subv at 'my_subvol' is the fallback case */
2858 ret = test_issubvolume(p);
2859 if (ret < 0 || !ret)
2861 fd = open(p, O_RDONLY);
2864 ret = lookup_ino_rootid(fd, &id);
2866 error("failed to lookup root id: %s", strerror(-ret));
2873 error("invalid qgroupid or subvolume path: %s", p);
2877 int open_file_or_dir3(const char *fname, DIR **dirstream, int open_flags)
2883 ret = stat(fname, &st);
2887 if (S_ISDIR(st.st_mode)) {
2888 *dirstream = opendir(fname);
2891 fd = dirfd(*dirstream);
2892 } else if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) {
2893 fd = open(fname, open_flags);
2896 * we set this on purpose, in case the caller output
2897 * strerror(errno) as success
2905 closedir(*dirstream);
2912 int open_file_or_dir(const char *fname, DIR **dirstream)
2914 return open_file_or_dir3(fname, dirstream, O_RDWR);
2917 void close_file_or_dir(int fd, DIR *dirstream)
2920 closedir(dirstream);
2925 int get_device_info(int fd, u64 devid,
2926 struct btrfs_ioctl_dev_info_args *di_args)
2930 di_args->devid = devid;
2931 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
2933 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
2934 return ret < 0 ? -errno : 0;
2937 static u64 find_max_device_id(struct btrfs_ioctl_search_args *search_args,
2940 struct btrfs_dev_item *dev_item;
2941 char *buf = search_args->buf;
2943 buf += (nr_items - 1) * (sizeof(struct btrfs_ioctl_search_header)
2944 + sizeof(struct btrfs_dev_item));
2945 buf += sizeof(struct btrfs_ioctl_search_header);
2947 dev_item = (struct btrfs_dev_item *)buf;
2949 return btrfs_stack_device_id(dev_item);
2952 static int search_chunk_tree_for_fs_info(int fd,
2953 struct btrfs_ioctl_fs_info_args *fi_args)
2957 u64 start_devid = 1;
2958 struct btrfs_ioctl_search_args search_args;
2959 struct btrfs_ioctl_search_key *search_key = &search_args.key;
2961 fi_args->num_devices = 0;
2963 max_items = BTRFS_SEARCH_ARGS_BUFSIZE
2964 / (sizeof(struct btrfs_ioctl_search_header)
2965 + sizeof(struct btrfs_dev_item));
2967 search_key->tree_id = BTRFS_CHUNK_TREE_OBJECTID;
2968 search_key->min_objectid = BTRFS_DEV_ITEMS_OBJECTID;
2969 search_key->max_objectid = BTRFS_DEV_ITEMS_OBJECTID;
2970 search_key->min_type = BTRFS_DEV_ITEM_KEY;
2971 search_key->max_type = BTRFS_DEV_ITEM_KEY;
2972 search_key->min_transid = 0;
2973 search_key->max_transid = (u64)-1;
2974 search_key->nr_items = max_items;
2975 search_key->max_offset = (u64)-1;
2978 search_key->min_offset = start_devid;
2980 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH, &search_args);
2984 fi_args->num_devices += (u64)search_key->nr_items;
2986 if (search_key->nr_items == max_items) {
2987 start_devid = find_max_device_id(&search_args,
2988 search_key->nr_items) + 1;
2992 /* get the lastest max_id to stay consistent with the num_devices */
2993 if (search_key->nr_items == 0)
2995 * last tree_search returns an empty buf, use the devid of
2996 * the last dev_item of the previous tree_search
2998 fi_args->max_id = start_devid - 1;
3000 fi_args->max_id = find_max_device_id(&search_args,
3001 search_key->nr_items);
3007 * For a given path, fill in the ioctl fs_ and info_ args.
3008 * If the path is a btrfs mountpoint, fill info for all devices.
3009 * If the path is a btrfs device, fill in only that device.
3011 * The path provided must be either on a mounted btrfs fs,
3012 * or be a mounted btrfs device.
3014 * Returns 0 on success, or a negative errno.
3016 int get_fs_info(char *path, struct btrfs_ioctl_fs_info_args *fi_args,
3017 struct btrfs_ioctl_dev_info_args **di_ret)
3024 struct btrfs_fs_devices *fs_devices_mnt = NULL;
3025 struct btrfs_ioctl_dev_info_args *di_args;
3026 struct btrfs_ioctl_dev_info_args tmp;
3028 DIR *dirstream = NULL;
3030 memset(fi_args, 0, sizeof(*fi_args));
3032 if (is_block_device(path) == 1) {
3033 struct btrfs_super_block *disk_super;
3034 char buf[BTRFS_SUPER_INFO_SIZE];
3037 /* Ensure it's mounted, then set path to the mountpoint */
3038 fd = open(path, O_RDONLY);
3041 error("cannot open %s: %s", path, strerror(errno));
3044 ret = check_mounted_where(fd, path, mp, sizeof(mp),
3053 /* Only fill in this one device */
3054 fi_args->num_devices = 1;
3056 disk_super = (struct btrfs_super_block *)buf;
3057 ret = btrfs_read_dev_super(fd, disk_super,
3058 BTRFS_SUPER_INFO_OFFSET, 0);
3063 devid = btrfs_stack_device_id(&disk_super->dev_item);
3065 fi_args->max_id = devid;
3068 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
3072 /* at this point path must not be for a block device */
3073 fd = open_file_or_dir(path, &dirstream);
3079 /* fill in fi_args if not just a single device */
3080 if (fi_args->num_devices != 1) {
3081 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
3088 * The fs_args->num_devices does not include seed devices
3090 ret = search_chunk_tree_for_fs_info(fd, fi_args);
3095 * search_chunk_tree_for_fs_info() will lacks the devid 0
3096 * so manual probe for it here.
3098 ret = get_device_info(fd, 0, &tmp);
3100 fi_args->num_devices++;
3108 if (!fi_args->num_devices)
3111 di_args = *di_ret = malloc((fi_args->num_devices) * sizeof(*di_args));
3118 memcpy(di_args, &tmp, sizeof(tmp));
3119 for (; i <= fi_args->max_id; ++i) {
3120 ret = get_device_info(fd, i, &di_args[ndevs]);
3129 * only when the only dev we wanted to find is not there then
3130 * let any error be returned
3132 if (fi_args->num_devices != 1) {
3138 close_file_or_dir(fd, dirstream);
3142 #define isoctal(c) (((c) & ~7) == '0')
3144 static inline void translate(char *f, char *t)
3146 while (*f != '\0') {
3148 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
3149 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
3159 * Checks if the swap device.
3160 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
3162 static int is_swap_device(const char *file)
3173 if (stat(file, &st_buf) < 0)
3175 if (S_ISBLK(st_buf.st_mode))
3176 dev = st_buf.st_rdev;
3177 else if (S_ISREG(st_buf.st_mode)) {
3178 dev = st_buf.st_dev;
3179 ino = st_buf.st_ino;
3183 if ((f = fopen("/proc/swaps", "r")) == NULL)
3186 /* skip the first line */
3187 if (fgets(tmp, sizeof(tmp), f) == NULL)
3190 while (fgets(tmp, sizeof(tmp), f) != NULL) {
3191 if ((cp = strchr(tmp, ' ')) != NULL)
3193 if ((cp = strchr(tmp, '\t')) != NULL)
3195 translate(tmp, buf);
3196 if (stat(buf, &st_buf) != 0)
3198 if (S_ISBLK(st_buf.st_mode)) {
3199 if (dev == st_buf.st_rdev) {
3203 } else if (S_ISREG(st_buf.st_mode)) {
3204 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {
3218 * Check for existing filesystem or partition table on device.
3220 * 1 for existing fs or partition
3221 * 0 for nothing found
3222 * -1 for internal error
3224 static int check_overwrite(const char *device)
3227 blkid_probe pr = NULL;
3231 if (!device || !*device)
3234 ret = -1; /* will reset on success of all setup calls */
3236 pr = blkid_new_probe_from_filename(device);
3240 size = blkid_probe_get_size(pr);
3244 /* nothing to overwrite on a 0-length device */
3250 ret = blkid_probe_enable_partitions(pr, 1);
3254 ret = blkid_do_fullprobe(pr);
3259 * Blkid returns 1 for nothing found and 0 when it finds a signature,
3260 * but we want the exact opposite, so reverse the return value here.
3262 * In addition print some useful diagnostics about what actually is
3270 if (!blkid_probe_lookup_value(pr, "TYPE", &type, NULL)) {
3272 "%s appears to contain an existing "
3273 "filesystem (%s).\n", device, type);
3274 } else if (!blkid_probe_lookup_value(pr, "PTTYPE", &type, NULL)) {
3276 "%s appears to contain a partition "
3277 "table (%s).\n", device, type);
3280 "%s appears to contain something weird "
3281 "according to blkid\n", device);
3287 blkid_free_probe(pr);
3290 "probe of %s failed, cannot detect "
3291 "existing filesystem.\n", device);
3295 static int group_profile_devs_min(u64 flag)
3297 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3298 case 0: /* single */
3299 case BTRFS_BLOCK_GROUP_DUP:
3301 case BTRFS_BLOCK_GROUP_RAID0:
3302 case BTRFS_BLOCK_GROUP_RAID1:
3303 case BTRFS_BLOCK_GROUP_RAID5:
3305 case BTRFS_BLOCK_GROUP_RAID6:
3307 case BTRFS_BLOCK_GROUP_RAID10:
3314 int test_num_disk_vs_raid(u64 metadata_profile, u64 data_profile,
3315 u64 dev_cnt, int mixed, int ssd)
3318 u64 profile = metadata_profile | data_profile;
3323 allowed |= BTRFS_BLOCK_GROUP_RAID10;
3325 allowed |= BTRFS_BLOCK_GROUP_RAID6;
3327 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
3328 BTRFS_BLOCK_GROUP_RAID5;
3330 allowed |= BTRFS_BLOCK_GROUP_DUP;
3333 if (dev_cnt > 1 && profile & BTRFS_BLOCK_GROUP_DUP) {
3334 warning("DUP is not recommended on filesystem with multiple devices");
3336 if (metadata_profile & ~allowed) {
3338 "ERROR: unable to create FS with metadata profile %s "
3339 "(have %llu devices but %d devices are required)\n",
3340 btrfs_group_profile_str(metadata_profile), dev_cnt,
3341 group_profile_devs_min(metadata_profile));
3344 if (data_profile & ~allowed) {
3346 "ERROR: unable to create FS with data profile %s "
3347 "(have %llu devices but %d devices are required)\n",
3348 btrfs_group_profile_str(data_profile), dev_cnt,
3349 group_profile_devs_min(data_profile));
3353 if (dev_cnt == 3 && profile & BTRFS_BLOCK_GROUP_RAID6) {
3354 warning("RAID6 is not recommended on filesystem with 3 devices only");
3356 if (dev_cnt == 2 && profile & BTRFS_BLOCK_GROUP_RAID5) {
3357 warning("RAID5 is not recommended on filesystem with 2 devices only");
3359 warning_on(!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP) && ssd,
3360 "DUP may not actually lead to 2 copies on the device, see manual page");
3365 int group_profile_max_safe_loss(u64 flags)
3367 switch (flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3368 case 0: /* single */
3369 case BTRFS_BLOCK_GROUP_DUP:
3370 case BTRFS_BLOCK_GROUP_RAID0:
3372 case BTRFS_BLOCK_GROUP_RAID1:
3373 case BTRFS_BLOCK_GROUP_RAID5:
3374 case BTRFS_BLOCK_GROUP_RAID10:
3376 case BTRFS_BLOCK_GROUP_RAID6:
3384 * Check if a device is suitable for btrfs
3386 * 1: something is wrong, an error is printed
3389 int test_dev_for_mkfs(const char *file, int force_overwrite)
3394 ret = is_swap_device(file);
3396 error("checking status of %s: %s", file, strerror(-ret));
3400 error("%s is a swap device", file);
3403 if (!force_overwrite) {
3404 if (check_overwrite(file)) {
3405 error("use the -f option to force overwrite of %s",
3410 ret = check_mounted(file);
3412 error("cannot check mount status of %s: %s", file,
3417 error("%s is mounted", file);
3420 /* check if the device is busy */
3421 fd = open(file, O_RDWR|O_EXCL);
3423 error("unable to open %s: %s", file, strerror(errno));
3426 if (fstat(fd, &st)) {
3427 error("unable to stat %s: %s", file, strerror(errno));
3431 if (!S_ISBLK(st.st_mode)) {
3432 error("%s is not a block device", file);
3440 int btrfs_scan_lblkid(void)
3445 struct btrfs_fs_devices *tmp_devices;
3446 blkid_dev_iterate iter = NULL;
3447 blkid_dev dev = NULL;
3448 blkid_cache cache = NULL;
3449 char path[PATH_MAX];
3451 if (btrfs_scan_done)
3454 if (blkid_get_cache(&cache, NULL) < 0) {
3455 error("blkid cache get failed");
3458 blkid_probe_all(cache);
3459 iter = blkid_dev_iterate_begin(cache);
3460 blkid_dev_set_search(iter, "TYPE", "btrfs");
3461 while (blkid_dev_next(iter, &dev) == 0) {
3462 dev = blkid_verify(cache, dev);
3465 /* if we are here its definitely a btrfs disk*/
3466 strncpy_null(path, blkid_dev_devname(dev));
3468 fd = open(path, O_RDONLY);
3470 error("cannot open %s: %s", path, strerror(errno));
3473 ret = btrfs_scan_one_device(fd, path, &tmp_devices,
3474 &num_devices, BTRFS_SUPER_INFO_OFFSET,
3477 error("cannot scan %s: %s", path, strerror(-ret));
3484 blkid_dev_iterate_end(iter);
3485 blkid_put_cache(cache);
3487 btrfs_scan_done = 1;
3492 int is_vol_small(const char *file)
3499 fd = open(file, O_RDONLY);
3502 if (fstat(fd, &st) < 0) {
3507 size = btrfs_device_size(fd, &st);
3512 if (size < BTRFS_MKFS_SMALL_VOLUME_SIZE) {
3522 * This reads a line from the stdin and only returns non-zero if the
3523 * first whitespace delimited token is a case insensitive match with yes
3526 int ask_user(const char *question)
3528 char buf[30] = {0,};
3529 char *saveptr = NULL;
3532 printf("%s [y/N]: ", question);
3534 return fgets(buf, sizeof(buf) - 1, stdin) &&
3535 (answer = strtok_r(buf, " \t\n\r", &saveptr)) &&
3536 (!strcasecmp(answer, "yes") || !strcasecmp(answer, "y"));
3541 * - file or directory return the containing tree root id
3542 * - subvolume return its own tree id
3543 * - BTRFS_EMPTY_SUBVOL_DIR_OBJECTID (directory with ino == 2) the result is
3544 * undefined and function returns -1
3546 int lookup_ino_rootid(int fd, u64 *rootid)
3548 struct btrfs_ioctl_ino_lookup_args args;
3551 memset(&args, 0, sizeof(args));
3553 args.objectid = BTRFS_FIRST_FREE_OBJECTID;
3555 ret = ioctl(fd, BTRFS_IOC_INO_LOOKUP, &args);
3559 *rootid = args.treeid;
3565 * return 0 if a btrfs mount point is found
3566 * return 1 if a mount point is found but not btrfs
3567 * return <0 if something goes wrong
3569 int find_mount_root(const char *path, char **mount_root)
3577 int longest_matchlen = 0;
3578 char *longest_match = NULL;
3580 fd = open(path, O_RDONLY | O_NOATIME);
3585 mnttab = setmntent("/proc/self/mounts", "r");
3589 while ((ent = getmntent(mnttab))) {
3590 len = strlen(ent->mnt_dir);
3591 if (strncmp(ent->mnt_dir, path, len) == 0) {
3592 /* match found and use the latest match */
3593 if (longest_matchlen <= len) {
3594 free(longest_match);
3595 longest_matchlen = len;
3596 longest_match = strdup(ent->mnt_dir);
3597 not_btrfs = strcmp(ent->mnt_type, "btrfs");
3606 free(longest_match);
3611 *mount_root = realpath(longest_match, NULL);
3615 free(longest_match);
3619 int test_minimum_size(const char *file, u32 nodesize)
3622 struct stat statbuf;
3624 fd = open(file, O_RDONLY);
3627 if (stat(file, &statbuf) < 0) {
3631 if (btrfs_device_size(fd, &statbuf) < btrfs_min_dev_size(nodesize)) {
3641 * Test if path is a directory
3643 * 0 - path exists but it is not a directory
3644 * 1 - path exists and it is a directory
3647 int test_isdir(const char *path)
3652 ret = stat(path, &st);
3656 return !!S_ISDIR(st.st_mode);
3659 void units_set_mode(unsigned *units, unsigned mode)
3661 unsigned base = *units & UNITS_MODE_MASK;
3663 *units = base | mode;
3666 void units_set_base(unsigned *units, unsigned base)
3668 unsigned mode = *units & ~UNITS_MODE_MASK;
3670 *units = base | mode;
3673 int find_next_key(struct btrfs_path *path, struct btrfs_key *key)
3677 for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
3678 if (!path->nodes[level])
3680 if (path->slots[level] + 1 >=
3681 btrfs_header_nritems(path->nodes[level]))
3684 btrfs_item_key_to_cpu(path->nodes[level], key,
3685 path->slots[level] + 1);
3687 btrfs_node_key_to_cpu(path->nodes[level], key,
3688 path->slots[level] + 1);
3694 const char* btrfs_group_type_str(u64 flag)
3696 u64 mask = BTRFS_BLOCK_GROUP_TYPE_MASK |
3697 BTRFS_SPACE_INFO_GLOBAL_RSV;
3699 switch (flag & mask) {
3700 case BTRFS_BLOCK_GROUP_DATA:
3702 case BTRFS_BLOCK_GROUP_SYSTEM:
3704 case BTRFS_BLOCK_GROUP_METADATA:
3706 case BTRFS_BLOCK_GROUP_DATA|BTRFS_BLOCK_GROUP_METADATA:
3707 return "Data+Metadata";
3708 case BTRFS_SPACE_INFO_GLOBAL_RSV:
3709 return "GlobalReserve";
3715 const char* btrfs_group_profile_str(u64 flag)
3717 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3720 case BTRFS_BLOCK_GROUP_RAID0:
3722 case BTRFS_BLOCK_GROUP_RAID1:
3724 case BTRFS_BLOCK_GROUP_RAID5:
3726 case BTRFS_BLOCK_GROUP_RAID6:
3728 case BTRFS_BLOCK_GROUP_DUP:
3730 case BTRFS_BLOCK_GROUP_RAID10:
3737 u64 disk_size(const char *path)
3741 if (statfs(path, &sfs) < 0)
3744 return sfs.f_bsize * sfs.f_blocks;
3747 u64 get_partition_size(const char *dev)
3750 int fd = open(dev, O_RDONLY);
3754 if (ioctl(fd, BLKGETSIZE64, &result) < 0) {
3763 int btrfs_tree_search2_ioctl_supported(int fd)
3765 struct btrfs_ioctl_search_args_v2 *args2;
3766 struct btrfs_ioctl_search_key *sk;
3767 int args2_size = 1024;
3768 char args2_buf[args2_size];
3770 static int v2_supported = -1;
3772 if (v2_supported != -1)
3773 return v2_supported;
3775 args2 = (struct btrfs_ioctl_search_args_v2 *)args2_buf;
3779 * Search for the extent tree item in the root tree.
3781 sk->tree_id = BTRFS_ROOT_TREE_OBJECTID;
3782 sk->min_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3783 sk->max_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3784 sk->min_type = BTRFS_ROOT_ITEM_KEY;
3785 sk->max_type = BTRFS_ROOT_ITEM_KEY;
3787 sk->max_offset = (u64)-1;
3788 sk->min_transid = 0;
3789 sk->max_transid = (u64)-1;
3791 args2->buf_size = args2_size - sizeof(struct btrfs_ioctl_search_args_v2);
3792 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH_V2, args2);
3793 if (ret == -EOPNOTSUPP)
3800 return v2_supported;
3803 int btrfs_check_nodesize(u32 nodesize, u32 sectorsize, u64 features)
3805 if (nodesize < sectorsize) {
3806 error("illegal nodesize %u (smaller than %u)",
3807 nodesize, sectorsize);
3809 } else if (nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) {
3810 error("illegal nodesize %u (larger than %u)",
3811 nodesize, BTRFS_MAX_METADATA_BLOCKSIZE);
3813 } else if (nodesize & (sectorsize - 1)) {
3814 error("illegal nodesize %u (not aligned to %u)",
3815 nodesize, sectorsize);
3817 } else if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS &&
3818 nodesize != sectorsize) {
3819 error("illegal nodesize %u (not equal to %u for mixed block group)",
3820 nodesize, sectorsize);
3827 * Copy a path argument from SRC to DEST and check the SRC length if it's at
3828 * most PATH_MAX and fits into DEST. DESTLEN is supposed to be exact size of
3830 * The destination buffer is zero terminated.
3831 * Return < 0 for error, 0 otherwise.
3833 int arg_copy_path(char *dest, const char *src, int destlen)
3835 size_t len = strlen(src);
3837 if (len >= PATH_MAX || len >= destlen)
3838 return -ENAMETOOLONG;
3840 __strncpy_null(dest, src, destlen);
3845 unsigned int get_unit_mode_from_arg(int *argc, char *argv[], int df_mode)
3847 unsigned int unit_mode = UNITS_DEFAULT;
3851 for (arg_i = 0; arg_i < *argc; arg_i++) {
3852 if (!strcmp(argv[arg_i], "--"))
3855 if (!strcmp(argv[arg_i], "--raw")) {
3856 unit_mode = UNITS_RAW;
3860 if (!strcmp(argv[arg_i], "--human-readable")) {
3861 unit_mode = UNITS_HUMAN_BINARY;
3866 if (!strcmp(argv[arg_i], "--iec")) {
3867 units_set_mode(&unit_mode, UNITS_BINARY);
3871 if (!strcmp(argv[arg_i], "--si")) {
3872 units_set_mode(&unit_mode, UNITS_DECIMAL);
3877 if (!strcmp(argv[arg_i], "--kbytes")) {
3878 units_set_base(&unit_mode, UNITS_KBYTES);
3882 if (!strcmp(argv[arg_i], "--mbytes")) {
3883 units_set_base(&unit_mode, UNITS_MBYTES);
3887 if (!strcmp(argv[arg_i], "--gbytes")) {
3888 units_set_base(&unit_mode, UNITS_GBYTES);
3892 if (!strcmp(argv[arg_i], "--tbytes")) {
3893 units_set_base(&unit_mode, UNITS_TBYTES);
3901 if (!strcmp(argv[arg_i], "-b")) {
3902 unit_mode = UNITS_RAW;
3906 if (!strcmp(argv[arg_i], "-h")) {
3907 unit_mode = UNITS_HUMAN_BINARY;
3911 if (!strcmp(argv[arg_i], "-H")) {
3912 unit_mode = UNITS_HUMAN_DECIMAL;
3916 if (!strcmp(argv[arg_i], "-k")) {
3917 units_set_base(&unit_mode, UNITS_KBYTES);
3921 if (!strcmp(argv[arg_i], "-m")) {
3922 units_set_base(&unit_mode, UNITS_MBYTES);
3926 if (!strcmp(argv[arg_i], "-g")) {
3927 units_set_base(&unit_mode, UNITS_GBYTES);
3931 if (!strcmp(argv[arg_i], "-t")) {
3932 units_set_base(&unit_mode, UNITS_TBYTES);
3938 for (arg_i = 0, arg_end = 0; arg_i < *argc; arg_i++) {
3941 argv[arg_end] = argv[arg_i];
3950 int string_is_numerical(const char *str)
3952 if (!(*str >= '0' && *str <= '9'))
3954 while (*str >= '0' && *str <= '9')
3962 * Preprocess @argv with getopt_long to reorder options and consume the "--"
3964 * Unknown short and long options are reported, optionally the @usage is printed
3967 void clean_args_no_options(int argc, char *argv[], const char * const *usagestr)
3969 static const struct option long_options[] = {
3974 int c = getopt_long(argc, argv, "", long_options, NULL);
3988 * Same as clean_args_no_options but pass through arguments that could look
3989 * like short options. Eg. reisze which takes a negative resize argument like
3992 * This accepts only two forms:
3993 * - "-- option1 option2 ..."
3994 * - "option1 option2 ..."
3996 void clean_args_no_options_relaxed(int argc, char *argv[], const char * const *usagestr)
4001 if (strcmp(argv[1], "--") == 0)
4005 /* Subvolume helper functions */
4007 * test if name is a correct subvolume name
4008 * this function return
4009 * 0-> name is not a correct subvolume name
4010 * 1-> name is a correct subvolume name
4012 int test_issubvolname(const char *name)
4014 return name[0] != '\0' && !strchr(name, '/') &&
4015 strcmp(name, ".") && strcmp(name, "..");
4019 * Test if path is a subvolume
4021 * 0 - path exists but it is not a subvolume
4022 * 1 - path exists and it is a subvolume
4025 int test_issubvolume(const char *path)
4031 res = stat(path, &st);
4035 if (st.st_ino != BTRFS_FIRST_FREE_OBJECTID || !S_ISDIR(st.st_mode))
4038 res = statfs(path, &stfs);
4042 return (int)stfs.f_type == BTRFS_SUPER_MAGIC;
4045 const char *subvol_strip_mountpoint(const char *mnt, const char *full_path)
4047 int len = strlen(mnt);
4051 if (mnt[len - 1] != '/')
4054 return full_path + len;
4061 * 1: Error; and error info printed to the terminal. Fixme.
4062 * 2: If the fullpath is root tree instead of subvol tree
4064 int get_subvol_info(const char *fullpath, struct root_info *get_ri)
4071 const char *svpath = NULL;
4072 DIR *dirstream1 = NULL;
4073 DIR *dirstream2 = NULL;
4075 ret = test_issubvolume(fullpath);
4079 error("not a subvolume: %s", fullpath);
4083 ret = find_mount_root(fullpath, &mnt);
4087 error("%s doesn't belong to btrfs mount point", fullpath);
4091 svpath = subvol_strip_mountpoint(mnt, fullpath);
4093 fd = btrfs_open_dir(fullpath, &dirstream1, 1);
4097 ret = btrfs_list_get_path_rootid(fd, &sv_id);
4099 error("can't get rootid for '%s'", fullpath);
4103 mntfd = btrfs_open_dir(mnt, &dirstream2, 1);
4107 if (sv_id == BTRFS_FS_TREE_OBJECTID) {
4110 * So that caller may decide if thats an error or just fine.
4115 memset(get_ri, 0, sizeof(*get_ri));
4116 get_ri->root_id = sv_id;
4118 ret = btrfs_get_subvol(mntfd, get_ri);
4120 error("can't find '%s': %d", svpath, ret);
4123 close_file_or_dir(mntfd, dirstream2);
4124 close_file_or_dir(fd, dirstream1);
4130 void init_rand_seed(u64 seed)
4134 /* only use the last 48 bits */
4135 for (i = 0; i < 3; i++) {
4136 rand_seed[i] = (unsigned short)(seed ^ (unsigned short)(-1));
4139 rand_seed_initlized = 1;
4142 static void __init_seed(void)
4148 if(rand_seed_initlized)
4150 /* Use urandom as primary seed source. */
4151 fd = open("/dev/urandom", O_RDONLY);
4153 ret = read(fd, rand_seed, sizeof(rand_seed));
4155 if (ret < sizeof(rand_seed))
4159 /* Use time and pid as fallback seed */
4160 warning("failed to read /dev/urandom, use time and pid as random seed");
4161 gettimeofday(&tv, 0);
4162 rand_seed[0] = getpid() ^ (tv.tv_sec & 0xFFFF);
4163 rand_seed[1] = getppid() ^ (tv.tv_usec & 0xFFFF);
4164 rand_seed[2] = (tv.tv_sec ^ tv.tv_usec) >> 16;
4166 rand_seed_initlized = 1;
4173 * Don't use nrand48, its range is [0,2^31) The highest bit will alwasy
4174 * be 0. Use jrand48 to include the highest bit.
4176 return (u32)jrand48(rand_seed);
4179 unsigned int rand_range(unsigned int upper)
4183 * Use the full 48bits to mod, which would be more uniformly
4186 return (unsigned int)(jrand48(rand_seed) % upper);