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>
40 #include <sys/utsname.h>
41 #include <linux/version.h>
43 #include "kerncompat.h"
44 #include "radix-tree.h"
47 #include "transaction.h"
53 #include "mkfs/common.h"
56 #define BLKDISCARD _IO(0x12,119)
59 static int btrfs_scan_done = 0;
61 static int rand_seed_initlized = 0;
62 static unsigned short rand_seed[3];
64 struct btrfs_config bconf;
67 * Discard the given range in one go
69 static int discard_range(int fd, u64 start, u64 len)
71 u64 range[2] = { start, len };
73 if (ioctl(fd, BLKDISCARD, &range) < 0)
79 * Discard blocks in the given range in 1G chunks, the process is interruptible
81 static int discard_blocks(int fd, u64 start, u64 len)
85 u64 chunk_size = min_t(u64, len, SZ_1G);
88 ret = discard_range(fd, start, chunk_size);
98 static u64 reference_root_table[] = {
99 [1] = BTRFS_ROOT_TREE_OBJECTID,
100 [2] = BTRFS_EXTENT_TREE_OBJECTID,
101 [3] = BTRFS_CHUNK_TREE_OBJECTID,
102 [4] = BTRFS_DEV_TREE_OBJECTID,
103 [5] = BTRFS_FS_TREE_OBJECTID,
104 [6] = BTRFS_CSUM_TREE_OBJECTID,
107 int test_uuid_unique(char *fs_uuid)
110 blkid_dev_iterate iter = NULL;
111 blkid_dev dev = NULL;
112 blkid_cache cache = NULL;
114 if (blkid_get_cache(&cache, NULL) < 0) {
115 printf("ERROR: lblkid cache get failed\n");
118 blkid_probe_all(cache);
119 iter = blkid_dev_iterate_begin(cache);
120 blkid_dev_set_search(iter, "UUID", fs_uuid);
122 while (blkid_dev_next(iter, &dev) == 0) {
123 dev = blkid_verify(cache, dev);
130 blkid_dev_iterate_end(iter);
131 blkid_put_cache(cache);
137 * Reserve space from free_tree.
138 * The algorithm is very simple, find the first cache_extent with enough space
139 * and allocate from its beginning.
141 static int reserve_free_space(struct cache_tree *free_tree, u64 len,
144 struct cache_extent *cache;
147 ASSERT(ret_start != NULL);
148 cache = first_cache_extent(free_tree);
150 if (cache->size > len) {
152 *ret_start = cache->start;
155 if (cache->size == 0) {
156 remove_cache_extent(free_tree, cache);
163 cache = next_cache_extent(cache);
170 static inline int write_temp_super(int fd, struct btrfs_super_block *sb,
176 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
177 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
178 btrfs_csum_final(crc, &sb->csum[0]);
179 ret = pwrite(fd, sb, BTRFS_SUPER_INFO_SIZE, sb_bytenr);
180 if (ret < BTRFS_SUPER_INFO_SIZE)
181 ret = (ret < 0 ? -errno : -EIO);
188 * Setup temporary superblock at cfg->super_bynter
189 * Needed info are extracted from cfg, and root_bytenr, chunk_bytenr
191 * For now sys chunk array will be empty and dev_item is empty too.
192 * They will be re-initialized at temp chunk tree setup.
194 * The superblock signature is not valid, denotes a partially created
195 * filesystem, needs to be finalized.
197 static int setup_temp_super(int fd, struct btrfs_mkfs_config *cfg,
198 u64 root_bytenr, u64 chunk_bytenr)
200 unsigned char chunk_uuid[BTRFS_UUID_SIZE];
201 char super_buf[BTRFS_SUPER_INFO_SIZE];
202 struct btrfs_super_block *super = (struct btrfs_super_block *)super_buf;
205 memset(super_buf, 0, BTRFS_SUPER_INFO_SIZE);
206 cfg->num_bytes = round_down(cfg->num_bytes, cfg->sectorsize);
209 if (uuid_parse(cfg->fs_uuid, super->fsid) != 0) {
210 error("cound not parse UUID: %s", cfg->fs_uuid);
214 if (!test_uuid_unique(cfg->fs_uuid)) {
215 error("non-unique UUID: %s", cfg->fs_uuid);
220 uuid_generate(super->fsid);
221 uuid_unparse(super->fsid, cfg->fs_uuid);
223 uuid_generate(chunk_uuid);
224 uuid_unparse(chunk_uuid, cfg->chunk_uuid);
226 btrfs_set_super_bytenr(super, cfg->super_bytenr);
227 btrfs_set_super_num_devices(super, 1);
228 btrfs_set_super_magic(super, BTRFS_MAGIC_PARTIAL);
229 btrfs_set_super_generation(super, 1);
230 btrfs_set_super_root(super, root_bytenr);
231 btrfs_set_super_chunk_root(super, chunk_bytenr);
232 btrfs_set_super_total_bytes(super, cfg->num_bytes);
234 * Temporary filesystem will only have 6 tree roots:
235 * chunk tree, root tree, extent_tree, device tree, fs tree
238 btrfs_set_super_bytes_used(super, 6 * cfg->nodesize);
239 btrfs_set_super_sectorsize(super, cfg->sectorsize);
240 btrfs_set_super_leafsize(super, cfg->nodesize);
241 btrfs_set_super_nodesize(super, cfg->nodesize);
242 btrfs_set_super_stripesize(super, cfg->stripesize);
243 btrfs_set_super_csum_type(super, BTRFS_CSUM_TYPE_CRC32);
244 btrfs_set_super_chunk_root(super, chunk_bytenr);
245 btrfs_set_super_cache_generation(super, -1);
246 btrfs_set_super_incompat_flags(super, cfg->features);
248 __strncpy_null(super->label, cfg->label, BTRFS_LABEL_SIZE - 1);
250 /* Sys chunk array will be re-initialized at chunk tree init time */
251 super->sys_chunk_array_size = 0;
253 ret = write_temp_super(fd, super, cfg->super_bytenr);
259 * Setup an extent buffer for tree block.
261 static int setup_temp_extent_buffer(struct extent_buffer *buf,
262 struct btrfs_mkfs_config *cfg,
263 u64 bytenr, u64 owner)
265 unsigned char fsid[BTRFS_FSID_SIZE];
266 unsigned char chunk_uuid[BTRFS_UUID_SIZE];
269 ret = uuid_parse(cfg->fs_uuid, fsid);
272 ret = uuid_parse(cfg->chunk_uuid, chunk_uuid);
276 memset(buf->data, 0, cfg->nodesize);
277 buf->len = cfg->nodesize;
278 btrfs_set_header_bytenr(buf, bytenr);
279 btrfs_set_header_generation(buf, 1);
280 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
281 btrfs_set_header_owner(buf, owner);
282 btrfs_set_header_flags(buf, BTRFS_HEADER_FLAG_WRITTEN);
283 write_extent_buffer(buf, chunk_uuid, btrfs_header_chunk_tree_uuid(buf),
285 write_extent_buffer(buf, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
289 static inline int write_temp_extent_buffer(int fd, struct extent_buffer *buf,
294 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
296 /* Temporary extent buffer is always mapped 1:1 on disk */
297 ret = pwrite(fd, buf->data, buf->len, bytenr);
299 ret = (ret < 0 ? ret : -EIO);
306 * Insert a root item for temporary tree root
308 * Only used in make_btrfs_v2().
310 static void insert_temp_root_item(struct extent_buffer *buf,
311 struct btrfs_mkfs_config *cfg,
312 int *slot, u32 *itemoff, u64 objectid,
315 struct btrfs_root_item root_item;
316 struct btrfs_inode_item *inode_item;
317 struct btrfs_disk_key disk_key;
319 btrfs_set_header_nritems(buf, *slot + 1);
320 (*itemoff) -= sizeof(root_item);
321 memset(&root_item, 0, sizeof(root_item));
322 inode_item = &root_item.inode;
323 btrfs_set_stack_inode_generation(inode_item, 1);
324 btrfs_set_stack_inode_size(inode_item, 3);
325 btrfs_set_stack_inode_nlink(inode_item, 1);
326 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
327 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
328 btrfs_set_root_refs(&root_item, 1);
329 btrfs_set_root_used(&root_item, cfg->nodesize);
330 btrfs_set_root_generation(&root_item, 1);
331 btrfs_set_root_bytenr(&root_item, bytenr);
333 memset(&disk_key, 0, sizeof(disk_key));
334 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
335 btrfs_set_disk_key_objectid(&disk_key, objectid);
336 btrfs_set_disk_key_offset(&disk_key, 0);
338 btrfs_set_item_key(buf, &disk_key, *slot);
339 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
340 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(root_item));
341 write_extent_buffer(buf, &root_item,
342 btrfs_item_ptr_offset(buf, *slot),
347 static int setup_temp_root_tree(int fd, struct btrfs_mkfs_config *cfg,
348 u64 root_bytenr, u64 extent_bytenr,
349 u64 dev_bytenr, u64 fs_bytenr, u64 csum_bytenr)
351 struct extent_buffer *buf = NULL;
352 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
357 * Provided bytenr must in ascending order, or tree root will have a
360 if (!(root_bytenr < extent_bytenr && extent_bytenr < dev_bytenr &&
361 dev_bytenr < fs_bytenr && fs_bytenr < csum_bytenr)) {
362 error("bad tree bytenr order: "
363 "root < extent %llu < %llu, "
364 "extent < dev %llu < %llu, "
365 "dev < fs %llu < %llu, "
366 "fs < csum %llu < %llu",
367 (unsigned long long)root_bytenr,
368 (unsigned long long)extent_bytenr,
369 (unsigned long long)extent_bytenr,
370 (unsigned long long)dev_bytenr,
371 (unsigned long long)dev_bytenr,
372 (unsigned long long)fs_bytenr,
373 (unsigned long long)fs_bytenr,
374 (unsigned long long)csum_bytenr);
377 buf = malloc(sizeof(*buf) + cfg->nodesize);
381 ret = setup_temp_extent_buffer(buf, cfg, root_bytenr,
382 BTRFS_ROOT_TREE_OBJECTID);
386 insert_temp_root_item(buf, cfg, &slot, &itemoff,
387 BTRFS_EXTENT_TREE_OBJECTID, extent_bytenr);
388 insert_temp_root_item(buf, cfg, &slot, &itemoff,
389 BTRFS_DEV_TREE_OBJECTID, dev_bytenr);
390 insert_temp_root_item(buf, cfg, &slot, &itemoff,
391 BTRFS_FS_TREE_OBJECTID, fs_bytenr);
392 insert_temp_root_item(buf, cfg, &slot, &itemoff,
393 BTRFS_CSUM_TREE_OBJECTID, csum_bytenr);
395 ret = write_temp_extent_buffer(fd, buf, root_bytenr);
401 static int insert_temp_dev_item(int fd, struct extent_buffer *buf,
402 struct btrfs_mkfs_config *cfg,
403 int *slot, u32 *itemoff)
405 struct btrfs_disk_key disk_key;
406 struct btrfs_dev_item *dev_item;
407 char super_buf[BTRFS_SUPER_INFO_SIZE];
408 unsigned char dev_uuid[BTRFS_UUID_SIZE];
409 unsigned char fsid[BTRFS_FSID_SIZE];
410 struct btrfs_super_block *super = (struct btrfs_super_block *)super_buf;
413 ret = pread(fd, super_buf, BTRFS_SUPER_INFO_SIZE, cfg->super_bytenr);
414 if (ret < BTRFS_SUPER_INFO_SIZE) {
415 ret = (ret < 0 ? -errno : -EIO);
419 btrfs_set_header_nritems(buf, *slot + 1);
420 (*itemoff) -= sizeof(*dev_item);
421 /* setup device item 1, 0 is for replace case */
422 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
423 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
424 btrfs_set_disk_key_offset(&disk_key, 1);
425 btrfs_set_item_key(buf, &disk_key, *slot);
426 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
427 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(*dev_item));
429 dev_item = btrfs_item_ptr(buf, *slot, struct btrfs_dev_item);
430 /* Generate device uuid */
431 uuid_generate(dev_uuid);
432 write_extent_buffer(buf, dev_uuid,
433 (unsigned long)btrfs_device_uuid(dev_item),
435 uuid_parse(cfg->fs_uuid, fsid);
436 write_extent_buffer(buf, fsid,
437 (unsigned long)btrfs_device_fsid(dev_item),
439 btrfs_set_device_id(buf, dev_item, 1);
440 btrfs_set_device_generation(buf, dev_item, 0);
441 btrfs_set_device_total_bytes(buf, dev_item, cfg->num_bytes);
443 * The number must match the initial SYSTEM and META chunk size
445 btrfs_set_device_bytes_used(buf, dev_item,
446 BTRFS_MKFS_SYSTEM_GROUP_SIZE +
447 BTRFS_CONVERT_META_GROUP_SIZE);
448 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
449 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
450 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
451 btrfs_set_device_type(buf, dev_item, 0);
453 /* Super dev_item is not complete, copy the complete one to sb */
454 read_extent_buffer(buf, &super->dev_item, (unsigned long)dev_item,
456 ret = write_temp_super(fd, super, cfg->super_bytenr);
462 static int insert_temp_chunk_item(int fd, struct extent_buffer *buf,
463 struct btrfs_mkfs_config *cfg,
464 int *slot, u32 *itemoff, u64 start, u64 len,
467 struct btrfs_chunk *chunk;
468 struct btrfs_disk_key disk_key;
469 char super_buf[BTRFS_SUPER_INFO_SIZE];
470 struct btrfs_super_block *sb = (struct btrfs_super_block *)super_buf;
473 ret = pread(fd, super_buf, BTRFS_SUPER_INFO_SIZE,
475 if (ret < BTRFS_SUPER_INFO_SIZE) {
476 ret = (ret < 0 ? ret : -EIO);
480 btrfs_set_header_nritems(buf, *slot + 1);
481 (*itemoff) -= btrfs_chunk_item_size(1);
482 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
483 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
484 btrfs_set_disk_key_offset(&disk_key, start);
485 btrfs_set_item_key(buf, &disk_key, *slot);
486 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
487 btrfs_set_item_size(buf, btrfs_item_nr(*slot),
488 btrfs_chunk_item_size(1));
490 chunk = btrfs_item_ptr(buf, *slot, struct btrfs_chunk);
491 btrfs_set_chunk_length(buf, chunk, len);
492 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
493 btrfs_set_chunk_stripe_len(buf, chunk, BTRFS_STRIPE_LEN);
494 btrfs_set_chunk_type(buf, chunk, type);
495 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
496 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
497 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
498 btrfs_set_chunk_num_stripes(buf, chunk, 1);
499 /* TODO: Support DUP profile for system chunk */
500 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
501 /* We are doing 1:1 mapping, so start is its dev offset */
502 btrfs_set_stripe_offset_nr(buf, chunk, 0, start);
503 write_extent_buffer(buf, &sb->dev_item.uuid,
504 (unsigned long)btrfs_stripe_dev_uuid_nr(chunk, 0),
509 * If it's system chunk, also copy it to super block.
511 if (type & BTRFS_BLOCK_GROUP_SYSTEM) {
515 cur = (char *)sb->sys_chunk_array
516 + btrfs_super_sys_array_size(sb);
517 memcpy(cur, &disk_key, sizeof(disk_key));
518 cur += sizeof(disk_key);
519 read_extent_buffer(buf, cur, (unsigned long int)chunk,
520 btrfs_chunk_item_size(1));
521 array_size = btrfs_super_sys_array_size(sb);
522 array_size += btrfs_chunk_item_size(1) +
524 btrfs_set_super_sys_array_size(sb, array_size);
526 ret = write_temp_super(fd, sb, cfg->super_bytenr);
531 static int setup_temp_chunk_tree(int fd, struct btrfs_mkfs_config *cfg,
532 u64 sys_chunk_start, u64 meta_chunk_start,
535 struct extent_buffer *buf = NULL;
536 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
540 /* Must ensure SYS chunk starts before META chunk */
541 if (meta_chunk_start < sys_chunk_start) {
542 error("wrong chunk order: meta < system %llu < %llu",
543 (unsigned long long)meta_chunk_start,
544 (unsigned long long)sys_chunk_start);
547 buf = malloc(sizeof(*buf) + cfg->nodesize);
550 ret = setup_temp_extent_buffer(buf, cfg, chunk_bytenr,
551 BTRFS_CHUNK_TREE_OBJECTID);
555 ret = insert_temp_dev_item(fd, buf, cfg, &slot, &itemoff);
558 ret = insert_temp_chunk_item(fd, buf, cfg, &slot, &itemoff,
560 BTRFS_MKFS_SYSTEM_GROUP_SIZE,
561 BTRFS_BLOCK_GROUP_SYSTEM);
564 ret = insert_temp_chunk_item(fd, buf, cfg, &slot, &itemoff,
566 BTRFS_CONVERT_META_GROUP_SIZE,
567 BTRFS_BLOCK_GROUP_METADATA);
570 ret = write_temp_extent_buffer(fd, buf, chunk_bytenr);
577 static void insert_temp_dev_extent(struct extent_buffer *buf,
578 int *slot, u32 *itemoff, u64 start, u64 len)
580 struct btrfs_dev_extent *dev_extent;
581 struct btrfs_disk_key disk_key;
583 btrfs_set_header_nritems(buf, *slot + 1);
584 (*itemoff) -= sizeof(*dev_extent);
585 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
586 btrfs_set_disk_key_objectid(&disk_key, 1);
587 btrfs_set_disk_key_offset(&disk_key, start);
588 btrfs_set_item_key(buf, &disk_key, *slot);
589 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
590 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(*dev_extent));
592 dev_extent = btrfs_item_ptr(buf, *slot, struct btrfs_dev_extent);
593 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
594 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
595 btrfs_set_dev_extent_length(buf, dev_extent, len);
596 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, start);
597 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
598 BTRFS_CHUNK_TREE_OBJECTID);
602 static int setup_temp_dev_tree(int fd, struct btrfs_mkfs_config *cfg,
603 u64 sys_chunk_start, u64 meta_chunk_start,
606 struct extent_buffer *buf = NULL;
607 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
611 /* Must ensure SYS chunk starts before META chunk */
612 if (meta_chunk_start < sys_chunk_start) {
613 error("wrong chunk order: meta < system %llu < %llu",
614 (unsigned long long)meta_chunk_start,
615 (unsigned long long)sys_chunk_start);
618 buf = malloc(sizeof(*buf) + cfg->nodesize);
621 ret = setup_temp_extent_buffer(buf, cfg, dev_bytenr,
622 BTRFS_DEV_TREE_OBJECTID);
625 insert_temp_dev_extent(buf, &slot, &itemoff, sys_chunk_start,
626 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
627 insert_temp_dev_extent(buf, &slot, &itemoff, meta_chunk_start,
628 BTRFS_CONVERT_META_GROUP_SIZE);
629 ret = write_temp_extent_buffer(fd, buf, dev_bytenr);
635 static int setup_temp_fs_tree(int fd, struct btrfs_mkfs_config *cfg,
638 struct extent_buffer *buf = NULL;
641 buf = malloc(sizeof(*buf) + cfg->nodesize);
644 ret = setup_temp_extent_buffer(buf, cfg, fs_bytenr,
645 BTRFS_FS_TREE_OBJECTID);
649 * Temporary fs tree is completely empty.
651 ret = write_temp_extent_buffer(fd, buf, fs_bytenr);
657 static int setup_temp_csum_tree(int fd, struct btrfs_mkfs_config *cfg,
660 struct extent_buffer *buf = NULL;
663 buf = malloc(sizeof(*buf) + cfg->nodesize);
666 ret = setup_temp_extent_buffer(buf, cfg, csum_bytenr,
667 BTRFS_CSUM_TREE_OBJECTID);
671 * Temporary csum tree is completely empty.
673 ret = write_temp_extent_buffer(fd, buf, csum_bytenr);
680 * Insert one temporary extent item.
682 * NOTE: if skinny_metadata is not enabled, this function must be called
683 * after all other trees are initialized.
684 * Or fs without skinny-metadata will be screwed up.
686 static int insert_temp_extent_item(int fd, struct extent_buffer *buf,
687 struct btrfs_mkfs_config *cfg,
688 int *slot, u32 *itemoff, u64 bytenr,
691 struct extent_buffer *tmp;
692 struct btrfs_extent_item *ei;
693 struct btrfs_extent_inline_ref *iref;
694 struct btrfs_disk_key disk_key;
695 struct btrfs_disk_key tree_info_key;
696 struct btrfs_tree_block_info *info;
698 int skinny_metadata = cfg->features &
699 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA;
703 itemsize = sizeof(*ei) + sizeof(*iref);
705 itemsize = sizeof(*ei) + sizeof(*iref) +
706 sizeof(struct btrfs_tree_block_info);
708 btrfs_set_header_nritems(buf, *slot + 1);
709 *(itemoff) -= itemsize;
711 if (skinny_metadata) {
712 btrfs_set_disk_key_type(&disk_key, BTRFS_METADATA_ITEM_KEY);
713 btrfs_set_disk_key_offset(&disk_key, 0);
715 btrfs_set_disk_key_type(&disk_key, BTRFS_EXTENT_ITEM_KEY);
716 btrfs_set_disk_key_offset(&disk_key, cfg->nodesize);
718 btrfs_set_disk_key_objectid(&disk_key, bytenr);
720 btrfs_set_item_key(buf, &disk_key, *slot);
721 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
722 btrfs_set_item_size(buf, btrfs_item_nr(*slot), itemsize);
724 ei = btrfs_item_ptr(buf, *slot, struct btrfs_extent_item);
725 btrfs_set_extent_refs(buf, ei, 1);
726 btrfs_set_extent_generation(buf, ei, 1);
727 btrfs_set_extent_flags(buf, ei, BTRFS_EXTENT_FLAG_TREE_BLOCK);
729 if (skinny_metadata) {
730 iref = (struct btrfs_extent_inline_ref *)(ei + 1);
732 info = (struct btrfs_tree_block_info *)(ei + 1);
733 iref = (struct btrfs_extent_inline_ref *)(info + 1);
735 btrfs_set_extent_inline_ref_type(buf, iref,
736 BTRFS_TREE_BLOCK_REF_KEY);
737 btrfs_set_extent_inline_ref_offset(buf, iref, ref_root);
744 * Lastly, check the tree block key by read the tree block
745 * Since we do 1:1 mapping for convert case, we can directly
746 * read the bytenr from disk
748 tmp = malloc(sizeof(*tmp) + cfg->nodesize);
751 ret = setup_temp_extent_buffer(tmp, cfg, bytenr, ref_root);
754 ret = pread(fd, tmp->data, cfg->nodesize, bytenr);
755 if (ret < cfg->nodesize) {
756 ret = (ret < 0 ? -errno : -EIO);
759 if (btrfs_header_nritems(tmp) == 0) {
760 btrfs_set_disk_key_type(&tree_info_key, 0);
761 btrfs_set_disk_key_objectid(&tree_info_key, 0);
762 btrfs_set_disk_key_offset(&tree_info_key, 0);
764 btrfs_item_key(tmp, &tree_info_key, 0);
766 btrfs_set_tree_block_key(buf, info, &tree_info_key);
773 static void insert_temp_block_group(struct extent_buffer *buf,
774 struct btrfs_mkfs_config *cfg,
775 int *slot, u32 *itemoff,
776 u64 bytenr, u64 len, u64 used, u64 flag)
778 struct btrfs_block_group_item bgi;
779 struct btrfs_disk_key disk_key;
781 btrfs_set_header_nritems(buf, *slot + 1);
782 (*itemoff) -= sizeof(bgi);
783 btrfs_set_disk_key_type(&disk_key, BTRFS_BLOCK_GROUP_ITEM_KEY);
784 btrfs_set_disk_key_objectid(&disk_key, bytenr);
785 btrfs_set_disk_key_offset(&disk_key, len);
786 btrfs_set_item_key(buf, &disk_key, *slot);
787 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
788 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(bgi));
790 btrfs_set_block_group_flags(&bgi, flag);
791 btrfs_set_block_group_used(&bgi, used);
792 btrfs_set_block_group_chunk_objectid(&bgi,
793 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
794 write_extent_buffer(buf, &bgi, btrfs_item_ptr_offset(buf, *slot),
799 static int setup_temp_extent_tree(int fd, struct btrfs_mkfs_config *cfg,
800 u64 chunk_bytenr, u64 root_bytenr,
801 u64 extent_bytenr, u64 dev_bytenr,
802 u64 fs_bytenr, u64 csum_bytenr)
804 struct extent_buffer *buf = NULL;
805 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
810 * We must ensure provided bytenr are in ascending order,
811 * or extent tree key order will be broken.
813 if (!(chunk_bytenr < root_bytenr && root_bytenr < extent_bytenr &&
814 extent_bytenr < dev_bytenr && dev_bytenr < fs_bytenr &&
815 fs_bytenr < csum_bytenr)) {
816 error("bad tree bytenr order: "
817 "chunk < root %llu < %llu, "
818 "root < extent %llu < %llu, "
819 "extent < dev %llu < %llu, "
820 "dev < fs %llu < %llu, "
821 "fs < csum %llu < %llu",
822 (unsigned long long)chunk_bytenr,
823 (unsigned long long)root_bytenr,
824 (unsigned long long)root_bytenr,
825 (unsigned long long)extent_bytenr,
826 (unsigned long long)extent_bytenr,
827 (unsigned long long)dev_bytenr,
828 (unsigned long long)dev_bytenr,
829 (unsigned long long)fs_bytenr,
830 (unsigned long long)fs_bytenr,
831 (unsigned long long)csum_bytenr);
834 buf = malloc(sizeof(*buf) + cfg->nodesize);
838 ret = setup_temp_extent_buffer(buf, cfg, extent_bytenr,
839 BTRFS_EXTENT_TREE_OBJECTID);
843 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
844 chunk_bytenr, BTRFS_CHUNK_TREE_OBJECTID);
848 insert_temp_block_group(buf, cfg, &slot, &itemoff, chunk_bytenr,
849 BTRFS_MKFS_SYSTEM_GROUP_SIZE, cfg->nodesize,
850 BTRFS_BLOCK_GROUP_SYSTEM);
852 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
853 root_bytenr, BTRFS_ROOT_TREE_OBJECTID);
857 /* 5 tree block used, root, extent, dev, fs and csum*/
858 insert_temp_block_group(buf, cfg, &slot, &itemoff, root_bytenr,
859 BTRFS_CONVERT_META_GROUP_SIZE, cfg->nodesize * 5,
860 BTRFS_BLOCK_GROUP_METADATA);
862 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
863 extent_bytenr, BTRFS_EXTENT_TREE_OBJECTID);
866 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
867 dev_bytenr, BTRFS_DEV_TREE_OBJECTID);
870 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
871 fs_bytenr, BTRFS_FS_TREE_OBJECTID);
874 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
875 csum_bytenr, BTRFS_CSUM_TREE_OBJECTID);
879 ret = write_temp_extent_buffer(fd, buf, extent_bytenr);
886 * Improved version of make_btrfs().
889 * 1) Do chunk allocation to avoid used data
890 * And after this function, extent type matches chunk type
891 * 2) Better structured code
892 * No super long hand written codes to initialized all tree blocks
893 * Split into small blocks and reuse codes.
894 * TODO: Reuse tree operation facilities by introducing new flags
896 int make_convert_btrfs(int fd, struct btrfs_mkfs_config *cfg,
897 struct btrfs_convert_context *cctx)
899 struct cache_tree *free = &cctx->free;
900 struct cache_tree *used = &cctx->used;
902 u64 meta_chunk_start;
903 /* chunk tree bytenr, in system chunk */
905 /* metadata trees bytenr, in metadata chunk */
913 /* Shouldn't happen */
914 BUG_ON(cache_tree_empty(used));
917 * reserve space for temporary superblock first
918 * Here we allocate a little larger space, to keep later
919 * free space will be STRIPE_LEN aligned
921 ret = reserve_free_space(free, BTRFS_STRIPE_LEN,
927 * Then reserve system chunk space
928 * TODO: Change system group size depending on cctx->total_bytes.
929 * If using current 4M, it can only handle less than one TB for
930 * worst case and then run out of sys space.
932 ret = reserve_free_space(free, BTRFS_MKFS_SYSTEM_GROUP_SIZE,
936 ret = reserve_free_space(free, BTRFS_CONVERT_META_GROUP_SIZE,
942 * Allocated meta/sys chunks will be mapped 1:1 with device offset.
944 * Inside the allocated metadata chunk, the layout will be:
945 * | offset | contents |
946 * -------------------------------------
948 * | +nodesize | extent root |
949 * | +nodesize * 2 | device root |
950 * | +nodesize * 3 | fs tree |
951 * | +nodesize * 4 | csum tree |
952 * -------------------------------------
953 * Inside the allocated system chunk, the layout will be:
954 * | offset | contents |
955 * -------------------------------------
956 * | +0 | chunk root |
957 * -------------------------------------
959 chunk_bytenr = sys_chunk_start;
960 root_bytenr = meta_chunk_start;
961 extent_bytenr = meta_chunk_start + cfg->nodesize;
962 dev_bytenr = meta_chunk_start + cfg->nodesize * 2;
963 fs_bytenr = meta_chunk_start + cfg->nodesize * 3;
964 csum_bytenr = meta_chunk_start + cfg->nodesize * 4;
966 ret = setup_temp_super(fd, cfg, root_bytenr, chunk_bytenr);
970 ret = setup_temp_root_tree(fd, cfg, root_bytenr, extent_bytenr,
971 dev_bytenr, fs_bytenr, csum_bytenr);
974 ret = setup_temp_chunk_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
978 ret = setup_temp_dev_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
982 ret = setup_temp_fs_tree(fd, cfg, fs_bytenr);
985 ret = setup_temp_csum_tree(fd, cfg, csum_bytenr);
989 * Setup extent tree last, since it may need to read tree block key
990 * for non-skinny metadata case.
992 ret = setup_temp_extent_tree(fd, cfg, chunk_bytenr, root_bytenr,
993 extent_bytenr, dev_bytenr, fs_bytenr,
1000 * @fs_uuid - if NULL, generates a UUID, returns back the new filesystem UUID
1002 * The superblock signature is not valid, denotes a partially created
1003 * filesystem, needs to be finalized.
1005 int make_btrfs(int fd, struct btrfs_mkfs_config *cfg)
1007 struct btrfs_super_block super;
1008 struct extent_buffer *buf;
1009 struct btrfs_root_item root_item;
1010 struct btrfs_disk_key disk_key;
1011 struct btrfs_extent_item *extent_item;
1012 struct btrfs_inode_item *inode_item;
1013 struct btrfs_chunk *chunk;
1014 struct btrfs_dev_item *dev_item;
1015 struct btrfs_dev_extent *dev_extent;
1016 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
1026 int skinny_metadata = !!(cfg->features &
1027 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA);
1030 buf = malloc(sizeof(*buf) + max(cfg->sectorsize, cfg->nodesize));
1034 first_free = BTRFS_SUPER_INFO_OFFSET + cfg->sectorsize * 2 - 1;
1035 first_free &= ~((u64)cfg->sectorsize - 1);
1037 memset(&super, 0, sizeof(super));
1039 num_bytes = (cfg->num_bytes / cfg->sectorsize) * cfg->sectorsize;
1040 if (*cfg->fs_uuid) {
1041 if (uuid_parse(cfg->fs_uuid, super.fsid) != 0) {
1042 error("cannot not parse UUID: %s", cfg->fs_uuid);
1046 if (!test_uuid_unique(cfg->fs_uuid)) {
1047 error("non-unique UUID: %s", cfg->fs_uuid);
1052 uuid_generate(super.fsid);
1053 uuid_unparse(super.fsid, cfg->fs_uuid);
1055 uuid_generate(super.dev_item.uuid);
1056 uuid_generate(chunk_tree_uuid);
1058 btrfs_set_super_bytenr(&super, cfg->blocks[0]);
1059 btrfs_set_super_num_devices(&super, 1);
1060 btrfs_set_super_magic(&super, BTRFS_MAGIC_PARTIAL);
1061 btrfs_set_super_generation(&super, 1);
1062 btrfs_set_super_root(&super, cfg->blocks[1]);
1063 btrfs_set_super_chunk_root(&super, cfg->blocks[3]);
1064 btrfs_set_super_total_bytes(&super, num_bytes);
1065 btrfs_set_super_bytes_used(&super, 6 * cfg->nodesize);
1066 btrfs_set_super_sectorsize(&super, cfg->sectorsize);
1067 btrfs_set_super_leafsize(&super, cfg->nodesize);
1068 btrfs_set_super_nodesize(&super, cfg->nodesize);
1069 btrfs_set_super_stripesize(&super, cfg->stripesize);
1070 btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
1071 btrfs_set_super_chunk_root_generation(&super, 1);
1072 btrfs_set_super_cache_generation(&super, -1);
1073 btrfs_set_super_incompat_flags(&super, cfg->features);
1075 __strncpy_null(super.label, cfg->label, BTRFS_LABEL_SIZE - 1);
1077 /* create the tree of root objects */
1078 memset(buf->data, 0, cfg->nodesize);
1079 buf->len = cfg->nodesize;
1080 btrfs_set_header_bytenr(buf, cfg->blocks[1]);
1081 btrfs_set_header_nritems(buf, 4);
1082 btrfs_set_header_generation(buf, 1);
1083 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
1084 btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
1085 write_extent_buffer(buf, super.fsid, btrfs_header_fsid(),
1088 write_extent_buffer(buf, chunk_tree_uuid,
1089 btrfs_header_chunk_tree_uuid(buf),
1092 /* create the items for the root tree */
1093 memset(&root_item, 0, sizeof(root_item));
1094 inode_item = &root_item.inode;
1095 btrfs_set_stack_inode_generation(inode_item, 1);
1096 btrfs_set_stack_inode_size(inode_item, 3);
1097 btrfs_set_stack_inode_nlink(inode_item, 1);
1098 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
1099 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
1100 btrfs_set_root_refs(&root_item, 1);
1101 btrfs_set_root_used(&root_item, cfg->nodesize);
1102 btrfs_set_root_generation(&root_item, 1);
1104 memset(&disk_key, 0, sizeof(disk_key));
1105 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
1106 btrfs_set_disk_key_offset(&disk_key, 0);
1109 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - sizeof(root_item);
1110 btrfs_set_root_bytenr(&root_item, cfg->blocks[2]);
1111 btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
1112 btrfs_set_item_key(buf, &disk_key, nritems);
1113 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1114 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1116 write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
1117 nritems), sizeof(root_item));
1120 itemoff = itemoff - sizeof(root_item);
1121 btrfs_set_root_bytenr(&root_item, cfg->blocks[4]);
1122 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
1123 btrfs_set_item_key(buf, &disk_key, nritems);
1124 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1125 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1127 write_extent_buffer(buf, &root_item,
1128 btrfs_item_ptr_offset(buf, nritems),
1132 itemoff = itemoff - sizeof(root_item);
1133 btrfs_set_root_bytenr(&root_item, cfg->blocks[5]);
1134 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
1135 btrfs_set_item_key(buf, &disk_key, nritems);
1136 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1137 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1139 write_extent_buffer(buf, &root_item,
1140 btrfs_item_ptr_offset(buf, nritems),
1144 itemoff = itemoff - sizeof(root_item);
1145 btrfs_set_root_bytenr(&root_item, cfg->blocks[6]);
1146 btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_TREE_OBJECTID);
1147 btrfs_set_item_key(buf, &disk_key, nritems);
1148 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1149 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1151 write_extent_buffer(buf, &root_item,
1152 btrfs_item_ptr_offset(buf, nritems),
1157 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1158 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[1]);
1159 if (ret != cfg->nodesize) {
1160 ret = (ret < 0 ? -errno : -EIO);
1164 /* create the items for the extent tree */
1165 memset(buf->data + sizeof(struct btrfs_header), 0,
1166 cfg->nodesize - sizeof(struct btrfs_header));
1168 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
1169 for (i = 1; i < 7; i++) {
1170 item_size = sizeof(struct btrfs_extent_item);
1171 if (!skinny_metadata)
1172 item_size += sizeof(struct btrfs_tree_block_info);
1174 if (cfg->blocks[i] < first_free) {
1175 error("block[%d] below first free: %llu < %llu",
1176 i, (unsigned long long)cfg->blocks[i],
1177 (unsigned long long)first_free);
1181 if (cfg->blocks[i] < cfg->blocks[i - 1]) {
1182 error("blocks %d and %d in reverse order: %llu < %llu",
1184 (unsigned long long)cfg->blocks[i],
1185 (unsigned long long)cfg->blocks[i - 1]);
1190 /* create extent item */
1191 itemoff -= item_size;
1192 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
1193 if (skinny_metadata) {
1194 btrfs_set_disk_key_type(&disk_key,
1195 BTRFS_METADATA_ITEM_KEY);
1196 btrfs_set_disk_key_offset(&disk_key, 0);
1198 btrfs_set_disk_key_type(&disk_key,
1199 BTRFS_EXTENT_ITEM_KEY);
1200 btrfs_set_disk_key_offset(&disk_key, cfg->nodesize);
1202 btrfs_set_item_key(buf, &disk_key, nritems);
1203 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
1205 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1207 extent_item = btrfs_item_ptr(buf, nritems,
1208 struct btrfs_extent_item);
1209 btrfs_set_extent_refs(buf, extent_item, 1);
1210 btrfs_set_extent_generation(buf, extent_item, 1);
1211 btrfs_set_extent_flags(buf, extent_item,
1212 BTRFS_EXTENT_FLAG_TREE_BLOCK);
1215 /* create extent ref */
1216 ref_root = reference_root_table[i];
1217 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
1218 btrfs_set_disk_key_offset(&disk_key, ref_root);
1219 btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
1220 btrfs_set_item_key(buf, &disk_key, nritems);
1221 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
1223 btrfs_set_item_size(buf, btrfs_item_nr(nritems), 0);
1226 btrfs_set_header_bytenr(buf, cfg->blocks[2]);
1227 btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
1228 btrfs_set_header_nritems(buf, nritems);
1229 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1230 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[2]);
1231 if (ret != cfg->nodesize) {
1232 ret = (ret < 0 ? -errno : -EIO);
1236 /* create the chunk tree */
1237 memset(buf->data + sizeof(struct btrfs_header), 0,
1238 cfg->nodesize - sizeof(struct btrfs_header));
1240 item_size = sizeof(*dev_item);
1241 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - item_size;
1243 /* first device 1 (there is no device 0) */
1244 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
1245 btrfs_set_disk_key_offset(&disk_key, 1);
1246 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
1247 btrfs_set_item_key(buf, &disk_key, nritems);
1248 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1249 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
1251 dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
1252 btrfs_set_device_id(buf, dev_item, 1);
1253 btrfs_set_device_generation(buf, dev_item, 0);
1254 btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
1255 btrfs_set_device_bytes_used(buf, dev_item,
1256 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1257 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
1258 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
1259 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
1260 btrfs_set_device_type(buf, dev_item, 0);
1262 write_extent_buffer(buf, super.dev_item.uuid,
1263 (unsigned long)btrfs_device_uuid(dev_item),
1265 write_extent_buffer(buf, super.fsid,
1266 (unsigned long)btrfs_device_fsid(dev_item),
1268 read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
1272 item_size = btrfs_chunk_item_size(1);
1273 itemoff = itemoff - item_size;
1275 /* then we have chunk 0 */
1276 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1277 btrfs_set_disk_key_offset(&disk_key, 0);
1278 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
1279 btrfs_set_item_key(buf, &disk_key, nritems);
1280 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1281 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
1283 chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
1284 btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1285 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
1286 btrfs_set_chunk_stripe_len(buf, chunk, BTRFS_STRIPE_LEN);
1287 btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
1288 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
1289 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
1290 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
1291 btrfs_set_chunk_num_stripes(buf, chunk, 1);
1292 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
1293 btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
1296 write_extent_buffer(buf, super.dev_item.uuid,
1297 (unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
1300 /* copy the key for the chunk to the system array */
1301 ptr = super.sys_chunk_array;
1302 array_size = sizeof(disk_key);
1304 memcpy(ptr, &disk_key, sizeof(disk_key));
1305 ptr += sizeof(disk_key);
1307 /* copy the chunk to the system array */
1308 read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
1309 array_size += item_size;
1311 btrfs_set_super_sys_array_size(&super, array_size);
1313 btrfs_set_header_bytenr(buf, cfg->blocks[3]);
1314 btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
1315 btrfs_set_header_nritems(buf, nritems);
1316 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1317 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[3]);
1318 if (ret != cfg->nodesize) {
1319 ret = (ret < 0 ? -errno : -EIO);
1323 /* create the device tree */
1324 memset(buf->data + sizeof(struct btrfs_header), 0,
1325 cfg->nodesize - sizeof(struct btrfs_header));
1327 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) -
1328 sizeof(struct btrfs_dev_extent);
1330 btrfs_set_disk_key_objectid(&disk_key, 1);
1331 btrfs_set_disk_key_offset(&disk_key, 0);
1332 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
1333 btrfs_set_item_key(buf, &disk_key, nritems);
1334 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1335 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1336 sizeof(struct btrfs_dev_extent));
1337 dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
1338 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
1339 BTRFS_CHUNK_TREE_OBJECTID);
1340 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
1341 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1342 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
1344 write_extent_buffer(buf, chunk_tree_uuid,
1345 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
1348 btrfs_set_dev_extent_length(buf, dev_extent,
1349 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1352 btrfs_set_header_bytenr(buf, cfg->blocks[4]);
1353 btrfs_set_header_owner(buf, BTRFS_DEV_TREE_OBJECTID);
1354 btrfs_set_header_nritems(buf, nritems);
1355 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1356 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[4]);
1357 if (ret != cfg->nodesize) {
1358 ret = (ret < 0 ? -errno : -EIO);
1362 /* create the FS root */
1363 memset(buf->data + sizeof(struct btrfs_header), 0,
1364 cfg->nodesize - sizeof(struct btrfs_header));
1365 btrfs_set_header_bytenr(buf, cfg->blocks[5]);
1366 btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
1367 btrfs_set_header_nritems(buf, 0);
1368 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1369 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[5]);
1370 if (ret != cfg->nodesize) {
1371 ret = (ret < 0 ? -errno : -EIO);
1374 /* finally create the csum root */
1375 memset(buf->data + sizeof(struct btrfs_header), 0,
1376 cfg->nodesize - sizeof(struct btrfs_header));
1377 btrfs_set_header_bytenr(buf, cfg->blocks[6]);
1378 btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
1379 btrfs_set_header_nritems(buf, 0);
1380 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1381 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[6]);
1382 if (ret != cfg->nodesize) {
1383 ret = (ret < 0 ? -errno : -EIO);
1387 /* and write out the super block */
1388 memset(buf->data, 0, BTRFS_SUPER_INFO_SIZE);
1389 memcpy(buf->data, &super, sizeof(super));
1390 buf->len = BTRFS_SUPER_INFO_SIZE;
1391 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1392 ret = pwrite(fd, buf->data, BTRFS_SUPER_INFO_SIZE, cfg->blocks[0]);
1393 if (ret != BTRFS_SUPER_INFO_SIZE) {
1394 ret = (ret < 0 ? -errno : -EIO);
1405 #define VERSION_TO_STRING3(a,b,c) #a "." #b "." #c, KERNEL_VERSION(a,b,c)
1406 #define VERSION_TO_STRING2(a,b) #a "." #b, KERNEL_VERSION(a,b,0)
1409 * Feature stability status and versions: compat <= safe <= default
1411 static const struct btrfs_fs_feature {
1414 const char *sysfs_name;
1416 * Compatibility with kernel of given version. Filesystem can be
1419 const char *compat_str;
1422 * Considered safe for use, but is not on by default, even if the
1423 * kernel supports the feature.
1425 const char *safe_str;
1428 * Considered safe for use and will be turned on by default if
1429 * supported by the running kernel.
1431 const char *default_str;
1434 } mkfs_features[] = {
1435 { "mixed-bg", BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS,
1437 VERSION_TO_STRING3(2,6,37),
1438 VERSION_TO_STRING3(2,6,37),
1440 "mixed data and metadata block groups" },
1441 { "extref", BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF,
1443 VERSION_TO_STRING2(3,7),
1444 VERSION_TO_STRING2(3,12),
1445 VERSION_TO_STRING2(3,12),
1446 "increased hardlink limit per file to 65536" },
1447 { "raid56", BTRFS_FEATURE_INCOMPAT_RAID56,
1449 VERSION_TO_STRING2(3,9),
1452 "raid56 extended format" },
1453 { "skinny-metadata", BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA,
1455 VERSION_TO_STRING2(3,10),
1456 VERSION_TO_STRING2(3,18),
1457 VERSION_TO_STRING2(3,18),
1458 "reduced-size metadata extent refs" },
1459 { "no-holes", BTRFS_FEATURE_INCOMPAT_NO_HOLES,
1461 VERSION_TO_STRING2(3,14),
1462 VERSION_TO_STRING2(4,0),
1464 "no explicit hole extents for files" },
1465 /* Keep this one last */
1466 { "list-all", BTRFS_FEATURE_LIST_ALL, NULL }
1469 static int parse_one_fs_feature(const char *name, u64 *flags)
1474 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1475 if (name[0] == '^' &&
1476 !strcmp(mkfs_features[i].name, name + 1)) {
1477 *flags &= ~ mkfs_features[i].flag;
1479 } else if (!strcmp(mkfs_features[i].name, name)) {
1480 *flags |= mkfs_features[i].flag;
1488 void btrfs_parse_features_to_string(char *buf, u64 flags)
1494 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1495 if (flags & mkfs_features[i].flag) {
1498 strcat(buf, mkfs_features[i].name);
1503 void btrfs_process_fs_features(u64 flags)
1507 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1508 if (flags & mkfs_features[i].flag) {
1509 printf("Turning ON incompat feature '%s': %s\n",
1510 mkfs_features[i].name,
1511 mkfs_features[i].desc);
1516 void btrfs_list_all_fs_features(u64 mask_disallowed)
1520 fprintf(stderr, "Filesystem features available:\n");
1521 for (i = 0; i < ARRAY_SIZE(mkfs_features) - 1; i++) {
1522 const struct btrfs_fs_feature *feat = &mkfs_features[i];
1524 if (feat->flag & mask_disallowed)
1526 fprintf(stderr, "%-20s- %s (0x%llx", feat->name, feat->desc,
1528 if (feat->compat_ver)
1529 fprintf(stderr, ", compat=%s", feat->compat_str);
1531 fprintf(stderr, ", safe=%s", feat->safe_str);
1532 if (feat->default_ver)
1533 fprintf(stderr, ", default=%s", feat->default_str);
1534 fprintf(stderr, ")\n");
1539 * Return NULL if all features were parsed fine, otherwise return the name of
1540 * the first unparsed.
1542 char* btrfs_parse_fs_features(char *namelist, u64 *flags)
1545 char *save_ptr = NULL; /* Satisfy static checkers */
1547 for (this_char = strtok_r(namelist, ",", &save_ptr);
1549 this_char = strtok_r(NULL, ",", &save_ptr)) {
1550 if (parse_one_fs_feature(this_char, flags))
1557 void print_kernel_version(FILE *stream, u32 version)
1561 v[0] = version & 0xFF;
1562 v[1] = (version >> 8) & 0xFF;
1563 v[2] = version >> 16;
1564 fprintf(stream, "%u.%u", v[2], v[1]);
1566 fprintf(stream, ".%u", v[0]);
1569 u32 get_running_kernel_version(void)
1571 struct utsname utsbuf;
1573 char *saveptr = NULL;
1577 if (strcmp(utsbuf.sysname, "Linux") != 0) {
1578 error("unsupported system: %s", utsbuf.sysname);
1582 tmp = strchr(utsbuf.release, '-');
1586 tmp = strtok_r(utsbuf.release, ".", &saveptr);
1587 if (!string_is_numerical(tmp))
1589 version = atoi(tmp) << 16;
1590 tmp = strtok_r(NULL, ".", &saveptr);
1591 if (!string_is_numerical(tmp))
1593 version |= atoi(tmp) << 8;
1594 tmp = strtok_r(NULL, ".", &saveptr);
1596 if (!string_is_numerical(tmp))
1598 version |= atoi(tmp);
1604 u64 btrfs_device_size(int fd, struct stat *st)
1607 if (S_ISREG(st->st_mode)) {
1610 if (!S_ISBLK(st->st_mode)) {
1613 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
1619 static int zero_blocks(int fd, off_t start, size_t len)
1621 char *buf = malloc(len);
1627 memset(buf, 0, len);
1628 written = pwrite(fd, buf, len, start);
1635 #define ZERO_DEV_BYTES SZ_2M
1637 /* don't write outside the device by clamping the region to the device size */
1638 static int zero_dev_clamped(int fd, off_t start, ssize_t len, u64 dev_size)
1640 off_t end = max(start, start + len);
1643 /* and don't overwrite the disk labels on sparc */
1644 start = max(start, 1024);
1645 end = max(end, 1024);
1648 start = min_t(u64, start, dev_size);
1649 end = min_t(u64, end, dev_size);
1651 return zero_blocks(fd, start, end - start);
1654 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
1655 struct btrfs_root *root, int fd, const char *path,
1656 u64 device_total_bytes, u32 io_width, u32 io_align,
1659 struct btrfs_super_block *disk_super;
1660 struct btrfs_super_block *super = root->fs_info->super_copy;
1661 struct btrfs_device *device;
1662 struct btrfs_dev_item *dev_item;
1668 device_total_bytes = (device_total_bytes / sectorsize) * sectorsize;
1670 device = calloc(1, sizeof(*device));
1675 buf = calloc(1, sectorsize);
1681 disk_super = (struct btrfs_super_block *)buf;
1682 dev_item = &disk_super->dev_item;
1684 uuid_generate(device->uuid);
1687 device->io_width = io_width;
1688 device->io_align = io_align;
1689 device->sector_size = sectorsize;
1691 device->writeable = 1;
1692 device->total_bytes = device_total_bytes;
1693 device->bytes_used = 0;
1694 device->total_ios = 0;
1695 device->dev_root = root->fs_info->dev_root;
1696 device->name = strdup(path);
1697 if (!device->name) {
1702 INIT_LIST_HEAD(&device->dev_list);
1703 ret = btrfs_add_device(trans, root, device);
1707 fs_total_bytes = btrfs_super_total_bytes(super) + device_total_bytes;
1708 btrfs_set_super_total_bytes(super, fs_total_bytes);
1710 num_devs = btrfs_super_num_devices(super) + 1;
1711 btrfs_set_super_num_devices(super, num_devs);
1713 memcpy(disk_super, super, sizeof(*disk_super));
1715 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
1716 btrfs_set_stack_device_id(dev_item, device->devid);
1717 btrfs_set_stack_device_type(dev_item, device->type);
1718 btrfs_set_stack_device_io_align(dev_item, device->io_align);
1719 btrfs_set_stack_device_io_width(dev_item, device->io_width);
1720 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
1721 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
1722 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
1723 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
1725 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
1726 BUG_ON(ret != sectorsize);
1729 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
1730 device->fs_devices = root->fs_info->fs_devices;
1739 static int btrfs_wipe_existing_sb(int fd)
1741 const char *off = NULL;
1746 blkid_probe pr = NULL;
1748 pr = blkid_new_probe();
1752 if (blkid_probe_set_device(pr, fd, 0, 0)) {
1757 ret = blkid_probe_lookup_value(pr, "SBMAGIC_OFFSET", &off, NULL);
1759 ret = blkid_probe_lookup_value(pr, "SBMAGIC", NULL, &len);
1761 if (ret || len == 0 || off == NULL) {
1763 * If lookup fails, the probe did not find any values, eg. for
1764 * a file image or a loop device. Soft error.
1770 offset = strtoll(off, NULL, 10);
1771 if (len > sizeof(buf))
1774 memset(buf, 0, len);
1775 ret = pwrite(fd, buf, len, offset);
1777 error("cannot wipe existing superblock: %s", strerror(errno));
1779 } else if (ret != len) {
1780 error("cannot wipe existing superblock: wrote %d of %zd", ret, len);
1786 blkid_free_probe(pr);
1790 int btrfs_prepare_device(int fd, const char *file, u64 *block_count_ret,
1791 u64 max_block_count, unsigned opflags)
1797 ret = fstat(fd, &st);
1799 error("unable to stat %s: %s", file, strerror(errno));
1803 block_count = btrfs_device_size(fd, &st);
1804 if (block_count == 0) {
1805 error("unable to determine size of %s", file);
1808 if (max_block_count)
1809 block_count = min(block_count, max_block_count);
1811 if (opflags & PREP_DEVICE_DISCARD) {
1813 * We intentionally ignore errors from the discard ioctl. It
1814 * is not necessary for the mkfs functionality but just an
1817 if (discard_range(fd, 0, 0) == 0) {
1818 if (opflags & PREP_DEVICE_VERBOSE)
1819 printf("Performing full device TRIM %s (%s) ...\n",
1820 file, pretty_size(block_count));
1821 discard_blocks(fd, 0, block_count);
1825 ret = zero_dev_clamped(fd, 0, ZERO_DEV_BYTES, block_count);
1826 for (i = 0 ; !ret && i < BTRFS_SUPER_MIRROR_MAX; i++)
1827 ret = zero_dev_clamped(fd, btrfs_sb_offset(i),
1828 BTRFS_SUPER_INFO_SIZE, block_count);
1829 if (!ret && (opflags & PREP_DEVICE_ZERO_END))
1830 ret = zero_dev_clamped(fd, block_count - ZERO_DEV_BYTES,
1831 ZERO_DEV_BYTES, block_count);
1834 error("failed to zero device '%s': %s", file, strerror(-ret));
1838 ret = btrfs_wipe_existing_sb(fd);
1840 error("cannot wipe superblocks on %s", file);
1844 *block_count_ret = block_count;
1848 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
1849 struct btrfs_root *root, u64 objectid)
1852 struct btrfs_inode_item inode_item;
1853 time_t now = time(NULL);
1855 memset(&inode_item, 0, sizeof(inode_item));
1856 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
1857 btrfs_set_stack_inode_size(&inode_item, 0);
1858 btrfs_set_stack_inode_nlink(&inode_item, 1);
1859 btrfs_set_stack_inode_nbytes(&inode_item, root->nodesize);
1860 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
1861 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
1862 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
1863 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
1864 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
1865 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
1866 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
1867 btrfs_set_stack_timespec_sec(&inode_item.otime, now);
1868 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
1870 if (root->fs_info->tree_root == root)
1871 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
1873 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
1877 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
1881 btrfs_set_root_dirid(&root->root_item, objectid);
1888 * checks if a path is a block device node
1889 * Returns negative errno on failure, otherwise
1890 * returns 1 for blockdev, 0 for not-blockdev
1892 int is_block_device(const char *path)
1894 struct stat statbuf;
1896 if (stat(path, &statbuf) < 0)
1899 return !!S_ISBLK(statbuf.st_mode);
1903 * check if given path is a mount point
1904 * return 1 if yes. 0 if no. -1 for error
1906 int is_mount_point(const char *path)
1912 f = setmntent("/proc/self/mounts", "r");
1916 while ((mnt = getmntent(f)) != NULL) {
1917 if (strcmp(mnt->mnt_dir, path))
1926 static int is_reg_file(const char *path)
1928 struct stat statbuf;
1930 if (stat(path, &statbuf) < 0)
1932 return S_ISREG(statbuf.st_mode);
1936 * This function checks if the given input parameter is
1938 * return <0 : some error in the given input
1939 * return BTRFS_ARG_UNKNOWN: unknown input
1940 * return BTRFS_ARG_UUID: given input is uuid
1941 * return BTRFS_ARG_MNTPOINT: given input is path
1942 * return BTRFS_ARG_REG: given input is regular file
1943 * return BTRFS_ARG_BLKDEV: given input is block device
1945 int check_arg_type(const char *input)
1948 char path[PATH_MAX];
1953 if (realpath(input, path)) {
1954 if (is_block_device(path) == 1)
1955 return BTRFS_ARG_BLKDEV;
1957 if (is_mount_point(path) == 1)
1958 return BTRFS_ARG_MNTPOINT;
1960 if (is_reg_file(path))
1961 return BTRFS_ARG_REG;
1963 return BTRFS_ARG_UNKNOWN;
1966 if (strlen(input) == (BTRFS_UUID_UNPARSED_SIZE - 1) &&
1967 !uuid_parse(input, uuid))
1968 return BTRFS_ARG_UUID;
1970 return BTRFS_ARG_UNKNOWN;
1974 * Find the mount point for a mounted device.
1975 * On success, returns 0 with mountpoint in *mp.
1976 * On failure, returns -errno (not mounted yields -EINVAL)
1977 * Is noisy on failures, expects to be given a mounted device.
1979 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size)
1984 ret = is_block_device(dev);
1987 error("not a block device: %s", dev);
1990 error("cannot check %s: %s", dev, strerror(-ret));
1995 fd = open(dev, O_RDONLY);
1998 error("cannot open %s: %s", dev, strerror(errno));
2002 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
2005 } else { /* mounted, all good */
2015 * Given a pathname, return a filehandle to:
2016 * the original pathname or,
2017 * if the pathname is a mounted btrfs device, to its mountpoint.
2019 * On error, return -1, errno should be set.
2021 int open_path_or_dev_mnt(const char *path, DIR **dirstream, int verbose)
2026 if (is_block_device(path)) {
2027 ret = get_btrfs_mount(path, mp, sizeof(mp));
2029 /* not a mounted btrfs dev */
2030 error_on(verbose, "'%s' is not a mounted btrfs device",
2035 ret = open_file_or_dir(mp, dirstream);
2036 error_on(verbose && ret < 0, "can't access '%s': %s",
2037 path, strerror(errno));
2039 ret = btrfs_open_dir(path, dirstream, 1);
2046 * Do the following checks before calling open_file_or_dir():
2047 * 1: path is in a btrfs filesystem
2048 * 2: path is a directory
2050 int btrfs_open_dir(const char *path, DIR **dirstream, int verbose)
2056 if (statfs(path, &stfs) != 0) {
2057 error_on(verbose, "cannot access '%s': %s", path,
2062 if (stfs.f_type != BTRFS_SUPER_MAGIC) {
2063 error_on(verbose, "not a btrfs filesystem: %s", path);
2067 if (stat(path, &st) != 0) {
2068 error_on(verbose, "cannot access '%s': %s", path,
2073 if (!S_ISDIR(st.st_mode)) {
2074 error_on(verbose, "not a directory: %s", path);
2078 ret = open_file_or_dir(path, dirstream);
2080 error_on(verbose, "cannot access '%s': %s", path,
2087 /* checks if a device is a loop device */
2088 static int is_loop_device (const char* device) {
2089 struct stat statbuf;
2091 if(stat(device, &statbuf) < 0)
2094 return (S_ISBLK(statbuf.st_mode) &&
2095 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
2099 * Takes a loop device path (e.g. /dev/loop0) and returns
2100 * the associated file (e.g. /images/my_btrfs.img) using
2103 static int resolve_loop_device_with_loopdev(const char* loop_dev, char* loop_file)
2107 struct loop_info64 lo64;
2109 fd = open(loop_dev, O_RDONLY | O_NONBLOCK);
2112 ret = ioctl(fd, LOOP_GET_STATUS64, &lo64);
2118 memcpy(loop_file, lo64.lo_file_name, sizeof(lo64.lo_file_name));
2119 loop_file[sizeof(lo64.lo_file_name)] = 0;
2127 /* Takes a loop device path (e.g. /dev/loop0) and returns
2128 * the associated file (e.g. /images/my_btrfs.img) */
2129 static int resolve_loop_device(const char* loop_dev, char* loop_file,
2136 char real_loop_dev[PATH_MAX];
2138 if (!realpath(loop_dev, real_loop_dev))
2140 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
2141 if (!(f = fopen(p, "r"))) {
2142 if (errno == ENOENT)
2144 * It's possibly a partitioned loop device, which is
2145 * resolvable with loopdev API.
2147 return resolve_loop_device_with_loopdev(loop_dev, loop_file);
2151 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
2152 ret = fscanf(f, fmt, loop_file);
2161 * Checks whether a and b are identical or device
2162 * files associated with the same block device
2164 static int is_same_blk_file(const char* a, const char* b)
2166 struct stat st_buf_a, st_buf_b;
2167 char real_a[PATH_MAX];
2168 char real_b[PATH_MAX];
2170 if (!realpath(a, real_a))
2171 strncpy_null(real_a, a);
2173 if (!realpath(b, real_b))
2174 strncpy_null(real_b, b);
2176 /* Identical path? */
2177 if (strcmp(real_a, real_b) == 0)
2180 if (stat(a, &st_buf_a) < 0 || stat(b, &st_buf_b) < 0) {
2181 if (errno == ENOENT)
2186 /* Same blockdevice? */
2187 if (S_ISBLK(st_buf_a.st_mode) && S_ISBLK(st_buf_b.st_mode) &&
2188 st_buf_a.st_rdev == st_buf_b.st_rdev) {
2193 if (st_buf_a.st_dev == st_buf_b.st_dev &&
2194 st_buf_a.st_ino == st_buf_b.st_ino) {
2201 /* checks if a and b are identical or device
2202 * files associated with the same block device or
2203 * if one file is a loop device that uses the other
2206 static int is_same_loop_file(const char* a, const char* b)
2208 char res_a[PATH_MAX];
2209 char res_b[PATH_MAX];
2210 const char* final_a = NULL;
2211 const char* final_b = NULL;
2214 /* Resolve a if it is a loop device */
2215 if((ret = is_loop_device(a)) < 0) {
2220 ret = resolve_loop_device(a, res_a, sizeof(res_a));
2231 /* Resolve b if it is a loop device */
2232 if ((ret = is_loop_device(b)) < 0) {
2237 ret = resolve_loop_device(b, res_b, sizeof(res_b));
2248 return is_same_blk_file(final_a, final_b);
2251 /* Checks if a file exists and is a block or regular file*/
2252 static int is_existing_blk_or_reg_file(const char* filename)
2256 if(stat(filename, &st_buf) < 0) {
2263 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
2266 /* Checks if a file is used (directly or indirectly via a loop device)
2267 * by a device in fs_devices
2269 static int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices,
2273 struct list_head *head;
2274 struct list_head *cur;
2275 struct btrfs_device *device;
2277 head = &fs_devices->devices;
2278 list_for_each(cur, head) {
2279 device = list_entry(cur, struct btrfs_device, dev_list);
2281 if((ret = is_same_loop_file(device->name, file)))
2289 * Resolve a pathname to a device mapper node to /dev/mapper/<name>
2290 * Returns NULL on invalid input or malloc failure; Other failures
2291 * will be handled by the caller using the input pathame.
2293 char *canonicalize_dm_name(const char *ptname)
2297 char path[PATH_MAX], name[PATH_MAX], *res = NULL;
2299 if (!ptname || !*ptname)
2302 snprintf(path, sizeof(path), "/sys/block/%s/dm/name", ptname);
2303 if (!(f = fopen(path, "r")))
2306 /* read <name>\n from sysfs */
2307 if (fgets(name, sizeof(name), f) && (sz = strlen(name)) > 1) {
2308 name[sz - 1] = '\0';
2309 snprintf(path, sizeof(path), "/dev/mapper/%s", name);
2311 if (access(path, F_OK) == 0)
2319 * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or
2320 * to a device mapper pathname.
2321 * Returns NULL on invalid input or malloc failure; Other failures
2322 * will be handled by the caller using the input pathame.
2324 char *canonicalize_path(const char *path)
2326 char *canonical, *p;
2328 if (!path || !*path)
2331 canonical = realpath(path, NULL);
2333 return strdup(path);
2334 p = strrchr(canonical, '/');
2335 if (p && strncmp(p, "/dm-", 4) == 0 && isdigit(*(p + 4))) {
2336 char *dm = canonicalize_dm_name(p + 1);
2347 * returns 1 if the device was mounted, < 0 on error or 0 if everything
2348 * is safe to continue.
2350 int check_mounted(const char* file)
2355 fd = open(file, O_RDONLY);
2357 error("mount check: cannot open %s: %s", file,
2362 ret = check_mounted_where(fd, file, NULL, 0, NULL);
2368 int check_mounted_where(int fd, const char *file, char *where, int size,
2369 struct btrfs_fs_devices **fs_dev_ret)
2374 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2378 /* scan the initial device */
2379 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
2380 &total_devs, BTRFS_SUPER_INFO_OFFSET, SBREAD_DEFAULT);
2381 is_btrfs = (ret >= 0);
2383 /* scan other devices */
2384 if (is_btrfs && total_devs > 1) {
2385 ret = btrfs_scan_devices();
2390 /* iterate over the list of currently mounted filesystems */
2391 if ((f = setmntent ("/proc/self/mounts", "r")) == NULL)
2394 while ((mnt = getmntent (f)) != NULL) {
2396 if(strcmp(mnt->mnt_type, "btrfs") != 0)
2399 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
2401 /* ignore entries in the mount table that are not
2402 associated with a file*/
2403 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
2404 goto out_mntloop_err;
2408 ret = is_same_loop_file(file, mnt->mnt_fsname);
2412 goto out_mntloop_err;
2417 /* Did we find an entry in mnt table? */
2418 if (mnt && size && where) {
2419 strncpy(where, mnt->mnt_dir, size);
2423 *fs_dev_ret = fs_devices_mnt;
2425 ret = (mnt != NULL);
2433 struct pending_dir {
2434 struct list_head list;
2435 char name[PATH_MAX];
2438 int btrfs_register_one_device(const char *fname)
2440 struct btrfs_ioctl_vol_args args;
2444 fd = open("/dev/btrfs-control", O_RDWR);
2447 "failed to open /dev/btrfs-control, skipping device registration: %s",
2451 memset(&args, 0, sizeof(args));
2452 strncpy_null(args.name, fname);
2453 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
2455 error("device scan failed on '%s': %s", fname,
2464 * Register all devices in the fs_uuid list created in the user
2465 * space. Ensure btrfs_scan_devices() is called before this func.
2467 int btrfs_register_all_devices(void)
2471 struct btrfs_fs_devices *fs_devices;
2472 struct btrfs_device *device;
2473 struct list_head *all_uuids;
2475 all_uuids = btrfs_scanned_uuids();
2477 list_for_each_entry(fs_devices, all_uuids, list) {
2478 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2480 err = btrfs_register_one_device(device->name);
2490 int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
2493 struct btrfs_super_block *disk_super;
2497 buf = malloc(BTRFS_SUPER_INFO_SIZE);
2502 ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
2503 if (ret != BTRFS_SUPER_INFO_SIZE)
2507 disk_super = (struct btrfs_super_block *)buf;
2509 * Accept devices from the same filesystem, allow partially created
2512 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC &&
2513 btrfs_super_magic(disk_super) != BTRFS_MAGIC_PARTIAL)
2516 if (!memcmp(disk_super->fsid, root->fs_info->super_copy->fsid,
2526 * Note: this function uses a static per-thread buffer. Do not call this
2527 * function more than 10 times within one argument list!
2529 const char *pretty_size_mode(u64 size, unsigned mode)
2531 static __thread int ps_index = 0;
2532 static __thread char ps_array[10][32];
2535 ret = ps_array[ps_index];
2538 (void)pretty_size_snprintf(size, ret, 32, mode);
2543 static const char* unit_suffix_binary[] =
2544 { "B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB"};
2545 static const char* unit_suffix_decimal[] =
2546 { "B", "kB", "MB", "GB", "TB", "PB", "EB"};
2548 int pretty_size_snprintf(u64 size, char *str, size_t str_size, unsigned unit_mode)
2554 const char** suffix = NULL;
2561 negative = !!(unit_mode & UNITS_NEGATIVE);
2562 unit_mode &= ~UNITS_NEGATIVE;
2564 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_RAW) {
2566 snprintf(str, str_size, "%lld", size);
2568 snprintf(str, str_size, "%llu", size);
2572 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_BINARY) {
2575 suffix = unit_suffix_binary;
2576 } else if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_DECIMAL) {
2579 suffix = unit_suffix_decimal;
2584 fprintf(stderr, "INTERNAL ERROR: unknown unit base, mode %d\n",
2592 switch (unit_mode & UNITS_MODE_MASK) {
2593 case UNITS_TBYTES: base *= mult; num_divs++;
2594 case UNITS_GBYTES: base *= mult; num_divs++;
2595 case UNITS_MBYTES: base *= mult; num_divs++;
2596 case UNITS_KBYTES: num_divs++;
2604 s64 ssize = (s64)size;
2605 s64 last_ssize = ssize;
2607 while ((ssize < 0 ? -ssize : ssize) >= mult) {
2612 last_size = (u64)last_ssize;
2614 while (size >= mult) {
2621 * If the value is smaller than base, we didn't do any
2622 * division, in that case, base should be 1, not original
2623 * base, or the unit will be wrong
2629 if (num_divs >= ARRAY_SIZE(unit_suffix_binary)) {
2631 printf("INTERNAL ERROR: unsupported unit suffix, index %d\n",
2638 fraction = (float)(s64)last_size / base;
2640 fraction = (float)last_size / base;
2643 return snprintf(str, str_size, "%.2f%s", fraction, suffix[num_divs]);
2647 * __strncpy_null - strncpy with null termination
2648 * @dest: the target array
2649 * @src: the source string
2650 * @n: maximum bytes to copy (size of *dest)
2652 * Like strncpy, but ensures destination is null-terminated.
2654 * Copies the string pointed to by src, including the terminating null
2655 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
2656 * of n bytes. Then ensure that dest is null-terminated.
2658 char *__strncpy_null(char *dest, const char *src, size_t n)
2660 strncpy(dest, src, n);
2667 * Checks to make sure that the label matches our requirements.
2669 0 if everything is safe and usable
2670 -1 if the label is too long
2672 static int check_label(const char *input)
2674 int len = strlen(input);
2676 if (len > BTRFS_LABEL_SIZE - 1) {
2677 error("label %s is too long (max %d)", input,
2678 BTRFS_LABEL_SIZE - 1);
2685 static int set_label_unmounted(const char *dev, const char *label)
2687 struct btrfs_trans_handle *trans;
2688 struct btrfs_root *root;
2691 ret = check_mounted(dev);
2693 error("checking mount status of %s failed: %d", dev, ret);
2697 error("device %s is mounted, use mount point", dev);
2701 /* Open the super_block at the default location
2702 * and as read-write.
2704 root = open_ctree(dev, 0, OPEN_CTREE_WRITES);
2705 if (!root) /* errors are printed by open_ctree() */
2708 trans = btrfs_start_transaction(root, 1);
2709 __strncpy_null(root->fs_info->super_copy->label, label, BTRFS_LABEL_SIZE - 1);
2711 btrfs_commit_transaction(trans, root);
2713 /* Now we close it since we are done. */
2718 static int set_label_mounted(const char *mount_path, const char *labelp)
2721 char label[BTRFS_LABEL_SIZE];
2723 fd = open(mount_path, O_RDONLY | O_NOATIME);
2725 error("unable to access %s: %s", mount_path, strerror(errno));
2729 memset(label, 0, sizeof(label));
2730 __strncpy_null(label, labelp, BTRFS_LABEL_SIZE - 1);
2731 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
2732 error("unable to set label of %s: %s", mount_path,
2742 int get_label_unmounted(const char *dev, char *label)
2744 struct btrfs_root *root;
2747 ret = check_mounted(dev);
2749 error("checking mount status of %s failed: %d", dev, ret);
2753 /* Open the super_block at the default location
2756 root = open_ctree(dev, 0, 0);
2760 __strncpy_null(label, root->fs_info->super_copy->label,
2761 BTRFS_LABEL_SIZE - 1);
2763 /* Now we close it since we are done. */
2769 * If a partition is mounted, try to get the filesystem label via its
2770 * mounted path rather than device. Return the corresponding error
2771 * the user specified the device path.
2773 int get_label_mounted(const char *mount_path, char *labelp)
2775 char label[BTRFS_LABEL_SIZE];
2779 fd = open(mount_path, O_RDONLY | O_NOATIME);
2781 error("unable to access %s: %s", mount_path, strerror(errno));
2785 memset(label, '\0', sizeof(label));
2786 ret = ioctl(fd, BTRFS_IOC_GET_FSLABEL, label);
2788 if (errno != ENOTTY)
2789 error("unable to get label of %s: %s", mount_path,
2796 __strncpy_null(labelp, label, BTRFS_LABEL_SIZE - 1);
2801 int get_label(const char *btrfs_dev, char *label)
2805 ret = is_existing_blk_or_reg_file(btrfs_dev);
2807 ret = get_label_mounted(btrfs_dev, label);
2809 ret = get_label_unmounted(btrfs_dev, label);
2814 int set_label(const char *btrfs_dev, const char *label)
2818 if (check_label(label))
2821 ret = is_existing_blk_or_reg_file(btrfs_dev);
2823 ret = set_label_mounted(btrfs_dev, label);
2825 ret = set_label_unmounted(btrfs_dev, label);
2831 * A not-so-good version fls64. No fascinating optimization since
2832 * no one except parse_size use it
2834 static int fls64(u64 x)
2838 for (i = 0; i <64; i++)
2839 if (x << i & (1ULL << 63))
2844 u64 parse_size(char *s)
2852 error("size value is empty");
2856 error("size value '%s' is less equal than 0", s);
2859 ret = strtoull(s, &endptr, 10);
2861 error("size value '%s' is invalid", s);
2864 if (endptr[0] && endptr[1]) {
2865 error("illegal suffix contains character '%c' in wrong position",
2870 * strtoll returns LLONG_MAX when overflow, if this happens,
2871 * need to call strtoull to get the real size
2873 if (errno == ERANGE && ret == ULLONG_MAX) {
2874 error("size value '%s' is too large for u64", s);
2878 c = tolower(endptr[0]);
2901 error("unknown size descriptor '%c'", c);
2905 /* Check whether ret * mult overflow */
2906 if (fls64(ret) + fls64(mult) - 1 > 64) {
2907 error("size value '%s' is too large for u64", s);
2914 u64 parse_qgroupid(const char *p)
2916 char *s = strchr(p, '/');
2917 const char *ptr_src_end = p + strlen(p);
2918 char *ptr_parse_end = NULL;
2927 /* Numeric format like '0/257' is the primary case */
2929 id = strtoull(p, &ptr_parse_end, 10);
2930 if (ptr_parse_end != ptr_src_end)
2934 level = strtoull(p, &ptr_parse_end, 10);
2935 if (ptr_parse_end != s)
2938 id = strtoull(s + 1, &ptr_parse_end, 10);
2939 if (ptr_parse_end != ptr_src_end)
2942 return (level << BTRFS_QGROUP_LEVEL_SHIFT) | id;
2945 /* Path format like subv at 'my_subvol' is the fallback case */
2946 ret = test_issubvolume(p);
2947 if (ret < 0 || !ret)
2949 fd = open(p, O_RDONLY);
2952 ret = lookup_path_rootid(fd, &id);
2954 error("failed to lookup root id: %s", strerror(-ret));
2961 error("invalid qgroupid or subvolume path: %s", p);
2965 int open_file_or_dir3(const char *fname, DIR **dirstream, int open_flags)
2971 ret = stat(fname, &st);
2975 if (S_ISDIR(st.st_mode)) {
2976 *dirstream = opendir(fname);
2979 fd = dirfd(*dirstream);
2980 } else if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) {
2981 fd = open(fname, open_flags);
2984 * we set this on purpose, in case the caller output
2985 * strerror(errno) as success
2993 closedir(*dirstream);
3000 int open_file_or_dir(const char *fname, DIR **dirstream)
3002 return open_file_or_dir3(fname, dirstream, O_RDWR);
3005 void close_file_or_dir(int fd, DIR *dirstream)
3008 closedir(dirstream);
3013 int get_device_info(int fd, u64 devid,
3014 struct btrfs_ioctl_dev_info_args *di_args)
3018 di_args->devid = devid;
3019 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
3021 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
3022 return ret < 0 ? -errno : 0;
3025 static u64 find_max_device_id(struct btrfs_ioctl_search_args *search_args,
3028 struct btrfs_dev_item *dev_item;
3029 char *buf = search_args->buf;
3031 buf += (nr_items - 1) * (sizeof(struct btrfs_ioctl_search_header)
3032 + sizeof(struct btrfs_dev_item));
3033 buf += sizeof(struct btrfs_ioctl_search_header);
3035 dev_item = (struct btrfs_dev_item *)buf;
3037 return btrfs_stack_device_id(dev_item);
3040 static int search_chunk_tree_for_fs_info(int fd,
3041 struct btrfs_ioctl_fs_info_args *fi_args)
3045 u64 start_devid = 1;
3046 struct btrfs_ioctl_search_args search_args;
3047 struct btrfs_ioctl_search_key *search_key = &search_args.key;
3049 fi_args->num_devices = 0;
3051 max_items = BTRFS_SEARCH_ARGS_BUFSIZE
3052 / (sizeof(struct btrfs_ioctl_search_header)
3053 + sizeof(struct btrfs_dev_item));
3055 search_key->tree_id = BTRFS_CHUNK_TREE_OBJECTID;
3056 search_key->min_objectid = BTRFS_DEV_ITEMS_OBJECTID;
3057 search_key->max_objectid = BTRFS_DEV_ITEMS_OBJECTID;
3058 search_key->min_type = BTRFS_DEV_ITEM_KEY;
3059 search_key->max_type = BTRFS_DEV_ITEM_KEY;
3060 search_key->min_transid = 0;
3061 search_key->max_transid = (u64)-1;
3062 search_key->nr_items = max_items;
3063 search_key->max_offset = (u64)-1;
3066 search_key->min_offset = start_devid;
3068 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH, &search_args);
3072 fi_args->num_devices += (u64)search_key->nr_items;
3074 if (search_key->nr_items == max_items) {
3075 start_devid = find_max_device_id(&search_args,
3076 search_key->nr_items) + 1;
3080 /* get the lastest max_id to stay consistent with the num_devices */
3081 if (search_key->nr_items == 0)
3083 * last tree_search returns an empty buf, use the devid of
3084 * the last dev_item of the previous tree_search
3086 fi_args->max_id = start_devid - 1;
3088 fi_args->max_id = find_max_device_id(&search_args,
3089 search_key->nr_items);
3095 * For a given path, fill in the ioctl fs_ and info_ args.
3096 * If the path is a btrfs mountpoint, fill info for all devices.
3097 * If the path is a btrfs device, fill in only that device.
3099 * The path provided must be either on a mounted btrfs fs,
3100 * or be a mounted btrfs device.
3102 * Returns 0 on success, or a negative errno.
3104 int get_fs_info(const char *path, struct btrfs_ioctl_fs_info_args *fi_args,
3105 struct btrfs_ioctl_dev_info_args **di_ret)
3112 struct btrfs_fs_devices *fs_devices_mnt = NULL;
3113 struct btrfs_ioctl_dev_info_args *di_args;
3114 struct btrfs_ioctl_dev_info_args tmp;
3116 DIR *dirstream = NULL;
3118 memset(fi_args, 0, sizeof(*fi_args));
3120 if (is_block_device(path) == 1) {
3121 struct btrfs_super_block *disk_super;
3122 char buf[BTRFS_SUPER_INFO_SIZE];
3124 /* Ensure it's mounted, then set path to the mountpoint */
3125 fd = open(path, O_RDONLY);
3128 error("cannot open %s: %s", path, strerror(errno));
3131 ret = check_mounted_where(fd, path, mp, sizeof(mp),
3140 /* Only fill in this one device */
3141 fi_args->num_devices = 1;
3143 disk_super = (struct btrfs_super_block *)buf;
3144 ret = btrfs_read_dev_super(fd, disk_super,
3145 BTRFS_SUPER_INFO_OFFSET, 0);
3150 last_devid = btrfs_stack_device_id(&disk_super->dev_item);
3151 fi_args->max_id = last_devid;
3153 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
3157 /* at this point path must not be for a block device */
3158 fd = open_file_or_dir(path, &dirstream);
3164 /* fill in fi_args if not just a single device */
3165 if (fi_args->num_devices != 1) {
3166 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
3173 * The fs_args->num_devices does not include seed devices
3175 ret = search_chunk_tree_for_fs_info(fd, fi_args);
3180 * search_chunk_tree_for_fs_info() will lacks the devid 0
3181 * so manual probe for it here.
3183 ret = get_device_info(fd, 0, &tmp);
3185 fi_args->num_devices++;
3188 if (last_devid == 0)
3193 if (!fi_args->num_devices)
3196 di_args = *di_ret = malloc((fi_args->num_devices) * sizeof(*di_args));
3203 memcpy(di_args, &tmp, sizeof(tmp));
3204 for (; last_devid <= fi_args->max_id; last_devid++) {
3205 ret = get_device_info(fd, last_devid, &di_args[ndevs]);
3214 * only when the only dev we wanted to find is not there then
3215 * let any error be returned
3217 if (fi_args->num_devices != 1) {
3223 close_file_or_dir(fd, dirstream);
3227 #define isoctal(c) (((c) & ~7) == '0')
3229 static inline void translate(char *f, char *t)
3231 while (*f != '\0') {
3233 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
3234 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
3244 * Checks if the swap device.
3245 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
3247 static int is_swap_device(const char *file)
3258 if (stat(file, &st_buf) < 0)
3260 if (S_ISBLK(st_buf.st_mode))
3261 dev = st_buf.st_rdev;
3262 else if (S_ISREG(st_buf.st_mode)) {
3263 dev = st_buf.st_dev;
3264 ino = st_buf.st_ino;
3268 if ((f = fopen("/proc/swaps", "r")) == NULL)
3271 /* skip the first line */
3272 if (fgets(tmp, sizeof(tmp), f) == NULL)
3275 while (fgets(tmp, sizeof(tmp), f) != NULL) {
3276 if ((cp = strchr(tmp, ' ')) != NULL)
3278 if ((cp = strchr(tmp, '\t')) != NULL)
3280 translate(tmp, buf);
3281 if (stat(buf, &st_buf) != 0)
3283 if (S_ISBLK(st_buf.st_mode)) {
3284 if (dev == st_buf.st_rdev) {
3288 } else if (S_ISREG(st_buf.st_mode)) {
3289 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {
3303 * Check for existing filesystem or partition table on device.
3305 * 1 for existing fs or partition
3306 * 0 for nothing found
3307 * -1 for internal error
3309 static int check_overwrite(const char *device)
3312 blkid_probe pr = NULL;
3316 if (!device || !*device)
3319 ret = -1; /* will reset on success of all setup calls */
3321 pr = blkid_new_probe_from_filename(device);
3325 size = blkid_probe_get_size(pr);
3329 /* nothing to overwrite on a 0-length device */
3335 ret = blkid_probe_enable_partitions(pr, 1);
3339 ret = blkid_do_fullprobe(pr);
3344 * Blkid returns 1 for nothing found and 0 when it finds a signature,
3345 * but we want the exact opposite, so reverse the return value here.
3347 * In addition print some useful diagnostics about what actually is
3355 if (!blkid_probe_lookup_value(pr, "TYPE", &type, NULL)) {
3357 "%s appears to contain an existing "
3358 "filesystem (%s).\n", device, type);
3359 } else if (!blkid_probe_lookup_value(pr, "PTTYPE", &type, NULL)) {
3361 "%s appears to contain a partition "
3362 "table (%s).\n", device, type);
3365 "%s appears to contain something weird "
3366 "according to blkid\n", device);
3372 blkid_free_probe(pr);
3375 "probe of %s failed, cannot detect "
3376 "existing filesystem.\n", device);
3380 static int group_profile_devs_min(u64 flag)
3382 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3383 case 0: /* single */
3384 case BTRFS_BLOCK_GROUP_DUP:
3386 case BTRFS_BLOCK_GROUP_RAID0:
3387 case BTRFS_BLOCK_GROUP_RAID1:
3388 case BTRFS_BLOCK_GROUP_RAID5:
3390 case BTRFS_BLOCK_GROUP_RAID6:
3392 case BTRFS_BLOCK_GROUP_RAID10:
3399 int test_num_disk_vs_raid(u64 metadata_profile, u64 data_profile,
3400 u64 dev_cnt, int mixed, int ssd)
3403 u64 profile = metadata_profile | data_profile;
3408 allowed |= BTRFS_BLOCK_GROUP_RAID10;
3410 allowed |= BTRFS_BLOCK_GROUP_RAID6;
3412 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
3413 BTRFS_BLOCK_GROUP_RAID5;
3415 allowed |= BTRFS_BLOCK_GROUP_DUP;
3418 if (dev_cnt > 1 && profile & BTRFS_BLOCK_GROUP_DUP) {
3419 warning("DUP is not recommended on filesystem with multiple devices");
3421 if (metadata_profile & ~allowed) {
3423 "ERROR: unable to create FS with metadata profile %s "
3424 "(have %llu devices but %d devices are required)\n",
3425 btrfs_group_profile_str(metadata_profile), dev_cnt,
3426 group_profile_devs_min(metadata_profile));
3429 if (data_profile & ~allowed) {
3431 "ERROR: unable to create FS with data profile %s "
3432 "(have %llu devices but %d devices are required)\n",
3433 btrfs_group_profile_str(data_profile), dev_cnt,
3434 group_profile_devs_min(data_profile));
3438 if (dev_cnt == 3 && profile & BTRFS_BLOCK_GROUP_RAID6) {
3439 warning("RAID6 is not recommended on filesystem with 3 devices only");
3441 if (dev_cnt == 2 && profile & BTRFS_BLOCK_GROUP_RAID5) {
3442 warning("RAID5 is not recommended on filesystem with 2 devices only");
3444 warning_on(!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP) && ssd,
3445 "DUP may not actually lead to 2 copies on the device, see manual page");
3450 int group_profile_max_safe_loss(u64 flags)
3452 switch (flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3453 case 0: /* single */
3454 case BTRFS_BLOCK_GROUP_DUP:
3455 case BTRFS_BLOCK_GROUP_RAID0:
3457 case BTRFS_BLOCK_GROUP_RAID1:
3458 case BTRFS_BLOCK_GROUP_RAID5:
3459 case BTRFS_BLOCK_GROUP_RAID10:
3461 case BTRFS_BLOCK_GROUP_RAID6:
3469 * Check if a device is suitable for btrfs
3471 * 1: something is wrong, an error is printed
3474 int test_dev_for_mkfs(const char *file, int force_overwrite)
3479 ret = is_swap_device(file);
3481 error("checking status of %s: %s", file, strerror(-ret));
3485 error("%s is a swap device", file);
3488 if (!force_overwrite) {
3489 if (check_overwrite(file)) {
3490 error("use the -f option to force overwrite of %s",
3495 ret = check_mounted(file);
3497 error("cannot check mount status of %s: %s", file,
3502 error("%s is mounted", file);
3505 /* check if the device is busy */
3506 fd = open(file, O_RDWR|O_EXCL);
3508 error("unable to open %s: %s", file, strerror(errno));
3511 if (fstat(fd, &st)) {
3512 error("unable to stat %s: %s", file, strerror(errno));
3516 if (!S_ISBLK(st.st_mode)) {
3517 error("%s is not a block device", file);
3525 int btrfs_scan_devices(void)
3530 struct btrfs_fs_devices *tmp_devices;
3531 blkid_dev_iterate iter = NULL;
3532 blkid_dev dev = NULL;
3533 blkid_cache cache = NULL;
3534 char path[PATH_MAX];
3536 if (btrfs_scan_done)
3539 if (blkid_get_cache(&cache, NULL) < 0) {
3540 error("blkid cache get failed");
3543 blkid_probe_all(cache);
3544 iter = blkid_dev_iterate_begin(cache);
3545 blkid_dev_set_search(iter, "TYPE", "btrfs");
3546 while (blkid_dev_next(iter, &dev) == 0) {
3547 dev = blkid_verify(cache, dev);
3550 /* if we are here its definitely a btrfs disk*/
3551 strncpy_null(path, blkid_dev_devname(dev));
3553 fd = open(path, O_RDONLY);
3555 error("cannot open %s: %s", path, strerror(errno));
3558 ret = btrfs_scan_one_device(fd, path, &tmp_devices,
3559 &num_devices, BTRFS_SUPER_INFO_OFFSET,
3562 error("cannot scan %s: %s", path, strerror(-ret));
3569 blkid_dev_iterate_end(iter);
3570 blkid_put_cache(cache);
3572 btrfs_scan_done = 1;
3577 int is_vol_small(const char *file)
3584 fd = open(file, O_RDONLY);
3587 if (fstat(fd, &st) < 0) {
3592 size = btrfs_device_size(fd, &st);
3597 if (size < BTRFS_MKFS_SMALL_VOLUME_SIZE) {
3607 * This reads a line from the stdin and only returns non-zero if the
3608 * first whitespace delimited token is a case insensitive match with yes
3611 int ask_user(const char *question)
3613 char buf[30] = {0,};
3614 char *saveptr = NULL;
3617 printf("%s [y/N]: ", question);
3619 return fgets(buf, sizeof(buf) - 1, stdin) &&
3620 (answer = strtok_r(buf, " \t\n\r", &saveptr)) &&
3621 (!strcasecmp(answer, "yes") || !strcasecmp(answer, "y"));
3625 * return 0 if a btrfs mount point is found
3626 * return 1 if a mount point is found but not btrfs
3627 * return <0 if something goes wrong
3629 int find_mount_root(const char *path, char **mount_root)
3637 int longest_matchlen = 0;
3638 char *longest_match = NULL;
3640 fd = open(path, O_RDONLY | O_NOATIME);
3645 mnttab = setmntent("/proc/self/mounts", "r");
3649 while ((ent = getmntent(mnttab))) {
3650 len = strlen(ent->mnt_dir);
3651 if (strncmp(ent->mnt_dir, path, len) == 0) {
3652 /* match found and use the latest match */
3653 if (longest_matchlen <= len) {
3654 free(longest_match);
3655 longest_matchlen = len;
3656 longest_match = strdup(ent->mnt_dir);
3657 not_btrfs = strcmp(ent->mnt_type, "btrfs");
3666 free(longest_match);
3671 *mount_root = realpath(longest_match, NULL);
3675 free(longest_match);
3679 int test_minimum_size(const char *file, u32 nodesize)
3682 struct stat statbuf;
3684 fd = open(file, O_RDONLY);
3687 if (stat(file, &statbuf) < 0) {
3691 if (btrfs_device_size(fd, &statbuf) < btrfs_min_dev_size(nodesize)) {
3701 * Test if path is a directory
3703 * 0 - path exists but it is not a directory
3704 * 1 - path exists and it is a directory
3707 int test_isdir(const char *path)
3712 ret = stat(path, &st);
3716 return !!S_ISDIR(st.st_mode);
3719 void units_set_mode(unsigned *units, unsigned mode)
3721 unsigned base = *units & UNITS_MODE_MASK;
3723 *units = base | mode;
3726 void units_set_base(unsigned *units, unsigned base)
3728 unsigned mode = *units & ~UNITS_MODE_MASK;
3730 *units = base | mode;
3733 int find_next_key(struct btrfs_path *path, struct btrfs_key *key)
3737 for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
3738 if (!path->nodes[level])
3740 if (path->slots[level] + 1 >=
3741 btrfs_header_nritems(path->nodes[level]))
3744 btrfs_item_key_to_cpu(path->nodes[level], key,
3745 path->slots[level] + 1);
3747 btrfs_node_key_to_cpu(path->nodes[level], key,
3748 path->slots[level] + 1);
3754 const char* btrfs_group_type_str(u64 flag)
3756 u64 mask = BTRFS_BLOCK_GROUP_TYPE_MASK |
3757 BTRFS_SPACE_INFO_GLOBAL_RSV;
3759 switch (flag & mask) {
3760 case BTRFS_BLOCK_GROUP_DATA:
3762 case BTRFS_BLOCK_GROUP_SYSTEM:
3764 case BTRFS_BLOCK_GROUP_METADATA:
3766 case BTRFS_BLOCK_GROUP_DATA|BTRFS_BLOCK_GROUP_METADATA:
3767 return "Data+Metadata";
3768 case BTRFS_SPACE_INFO_GLOBAL_RSV:
3769 return "GlobalReserve";
3775 const char* btrfs_group_profile_str(u64 flag)
3777 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3780 case BTRFS_BLOCK_GROUP_RAID0:
3782 case BTRFS_BLOCK_GROUP_RAID1:
3784 case BTRFS_BLOCK_GROUP_RAID5:
3786 case BTRFS_BLOCK_GROUP_RAID6:
3788 case BTRFS_BLOCK_GROUP_DUP:
3790 case BTRFS_BLOCK_GROUP_RAID10:
3797 u64 disk_size(const char *path)
3801 if (statfs(path, &sfs) < 0)
3804 return sfs.f_bsize * sfs.f_blocks;
3807 u64 get_partition_size(const char *dev)
3810 int fd = open(dev, O_RDONLY);
3814 if (ioctl(fd, BLKGETSIZE64, &result) < 0) {
3824 * Check if the BTRFS_IOC_TREE_SEARCH_V2 ioctl is supported on a given
3825 * filesystem, opened at fd
3827 int btrfs_tree_search2_ioctl_supported(int fd)
3829 struct btrfs_ioctl_search_args_v2 *args2;
3830 struct btrfs_ioctl_search_key *sk;
3831 int args2_size = 1024;
3832 char args2_buf[args2_size];
3835 args2 = (struct btrfs_ioctl_search_args_v2 *)args2_buf;
3839 * Search for the extent tree item in the root tree.
3841 sk->tree_id = BTRFS_ROOT_TREE_OBJECTID;
3842 sk->min_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3843 sk->max_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3844 sk->min_type = BTRFS_ROOT_ITEM_KEY;
3845 sk->max_type = BTRFS_ROOT_ITEM_KEY;
3847 sk->max_offset = (u64)-1;
3848 sk->min_transid = 0;
3849 sk->max_transid = (u64)-1;
3851 args2->buf_size = args2_size - sizeof(struct btrfs_ioctl_search_args_v2);
3852 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH_V2, args2);
3853 if (ret == -EOPNOTSUPP)
3860 int btrfs_check_nodesize(u32 nodesize, u32 sectorsize, u64 features)
3862 if (nodesize < sectorsize) {
3863 error("illegal nodesize %u (smaller than %u)",
3864 nodesize, sectorsize);
3866 } else if (nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) {
3867 error("illegal nodesize %u (larger than %u)",
3868 nodesize, BTRFS_MAX_METADATA_BLOCKSIZE);
3870 } else if (nodesize & (sectorsize - 1)) {
3871 error("illegal nodesize %u (not aligned to %u)",
3872 nodesize, sectorsize);
3874 } else if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS &&
3875 nodesize != sectorsize) {
3876 error("illegal nodesize %u (not equal to %u for mixed block group)",
3877 nodesize, sectorsize);
3884 * Copy a path argument from SRC to DEST and check the SRC length if it's at
3885 * most PATH_MAX and fits into DEST. DESTLEN is supposed to be exact size of
3887 * The destination buffer is zero terminated.
3888 * Return < 0 for error, 0 otherwise.
3890 int arg_copy_path(char *dest, const char *src, int destlen)
3892 size_t len = strlen(src);
3894 if (len >= PATH_MAX || len >= destlen)
3895 return -ENAMETOOLONG;
3897 __strncpy_null(dest, src, destlen);
3902 unsigned int get_unit_mode_from_arg(int *argc, char *argv[], int df_mode)
3904 unsigned int unit_mode = UNITS_DEFAULT;
3908 for (arg_i = 0; arg_i < *argc; arg_i++) {
3909 if (!strcmp(argv[arg_i], "--"))
3912 if (!strcmp(argv[arg_i], "--raw")) {
3913 unit_mode = UNITS_RAW;
3917 if (!strcmp(argv[arg_i], "--human-readable")) {
3918 unit_mode = UNITS_HUMAN_BINARY;
3923 if (!strcmp(argv[arg_i], "--iec")) {
3924 units_set_mode(&unit_mode, UNITS_BINARY);
3928 if (!strcmp(argv[arg_i], "--si")) {
3929 units_set_mode(&unit_mode, UNITS_DECIMAL);
3934 if (!strcmp(argv[arg_i], "--kbytes")) {
3935 units_set_base(&unit_mode, UNITS_KBYTES);
3939 if (!strcmp(argv[arg_i], "--mbytes")) {
3940 units_set_base(&unit_mode, UNITS_MBYTES);
3944 if (!strcmp(argv[arg_i], "--gbytes")) {
3945 units_set_base(&unit_mode, UNITS_GBYTES);
3949 if (!strcmp(argv[arg_i], "--tbytes")) {
3950 units_set_base(&unit_mode, UNITS_TBYTES);
3958 if (!strcmp(argv[arg_i], "-b")) {
3959 unit_mode = UNITS_RAW;
3963 if (!strcmp(argv[arg_i], "-h")) {
3964 unit_mode = UNITS_HUMAN_BINARY;
3968 if (!strcmp(argv[arg_i], "-H")) {
3969 unit_mode = UNITS_HUMAN_DECIMAL;
3973 if (!strcmp(argv[arg_i], "-k")) {
3974 units_set_base(&unit_mode, UNITS_KBYTES);
3978 if (!strcmp(argv[arg_i], "-m")) {
3979 units_set_base(&unit_mode, UNITS_MBYTES);
3983 if (!strcmp(argv[arg_i], "-g")) {
3984 units_set_base(&unit_mode, UNITS_GBYTES);
3988 if (!strcmp(argv[arg_i], "-t")) {
3989 units_set_base(&unit_mode, UNITS_TBYTES);
3995 for (arg_i = 0, arg_end = 0; arg_i < *argc; arg_i++) {
3998 argv[arg_end] = argv[arg_i];
4007 int string_is_numerical(const char *str)
4011 if (!(*str >= '0' && *str <= '9'))
4013 while (*str >= '0' && *str <= '9')
4020 /* Subvolume helper functions */
4022 * test if name is a correct subvolume name
4023 * this function return
4024 * 0-> name is not a correct subvolume name
4025 * 1-> name is a correct subvolume name
4027 int test_issubvolname(const char *name)
4029 return name[0] != '\0' && !strchr(name, '/') &&
4030 strcmp(name, ".") && strcmp(name, "..");
4034 * Test if path is a subvolume
4036 * 0 - path exists but it is not a subvolume
4037 * 1 - path exists and it is a subvolume
4040 int test_issubvolume(const char *path)
4046 res = stat(path, &st);
4050 if (st.st_ino != BTRFS_FIRST_FREE_OBJECTID || !S_ISDIR(st.st_mode))
4053 res = statfs(path, &stfs);
4057 return (int)stfs.f_type == BTRFS_SUPER_MAGIC;
4060 const char *subvol_strip_mountpoint(const char *mnt, const char *full_path)
4062 int len = strlen(mnt);
4066 if (mnt[len - 1] != '/')
4069 return full_path + len;
4076 * 1: Error; and error info printed to the terminal. Fixme.
4077 * 2: If the fullpath is root tree instead of subvol tree
4079 int get_subvol_info(const char *fullpath, struct root_info *get_ri)
4086 const char *svpath = NULL;
4087 DIR *dirstream1 = NULL;
4088 DIR *dirstream2 = NULL;
4090 ret = test_issubvolume(fullpath);
4094 error("not a subvolume: %s", fullpath);
4098 ret = find_mount_root(fullpath, &mnt);
4102 error("%s doesn't belong to btrfs mount point", fullpath);
4106 svpath = subvol_strip_mountpoint(mnt, fullpath);
4108 fd = btrfs_open_dir(fullpath, &dirstream1, 1);
4112 ret = btrfs_list_get_path_rootid(fd, &sv_id);
4116 mntfd = btrfs_open_dir(mnt, &dirstream2, 1);
4120 memset(get_ri, 0, sizeof(*get_ri));
4121 get_ri->root_id = sv_id;
4123 if (sv_id == BTRFS_FS_TREE_OBJECTID)
4124 ret = btrfs_get_toplevel_subvol(mntfd, get_ri);
4126 ret = btrfs_get_subvol(mntfd, get_ri);
4128 error("can't find '%s': %d", svpath, ret);
4131 close_file_or_dir(mntfd, dirstream2);
4132 close_file_or_dir(fd, dirstream1);
4138 void init_rand_seed(u64 seed)
4142 /* only use the last 48 bits */
4143 for (i = 0; i < 3; i++) {
4144 rand_seed[i] = (unsigned short)(seed ^ (unsigned short)(-1));
4147 rand_seed_initlized = 1;
4150 static void __init_seed(void)
4156 if(rand_seed_initlized)
4158 /* Use urandom as primary seed source. */
4159 fd = open("/dev/urandom", O_RDONLY);
4161 ret = read(fd, rand_seed, sizeof(rand_seed));
4163 if (ret < sizeof(rand_seed))
4167 /* Use time and pid as fallback seed */
4168 warning("failed to read /dev/urandom, use time and pid as random seed");
4169 gettimeofday(&tv, 0);
4170 rand_seed[0] = getpid() ^ (tv.tv_sec & 0xFFFF);
4171 rand_seed[1] = getppid() ^ (tv.tv_usec & 0xFFFF);
4172 rand_seed[2] = (tv.tv_sec ^ tv.tv_usec) >> 16;
4174 rand_seed_initlized = 1;
4181 * Don't use nrand48, its range is [0,2^31) The highest bit will alwasy
4182 * be 0. Use jrand48 to include the highest bit.
4184 return (u32)jrand48(rand_seed);
4187 unsigned int rand_range(unsigned int upper)
4191 * Use the full 48bits to mod, which would be more uniformly
4194 return (unsigned int)(jrand48(rand_seed) % upper);
4197 void btrfs_config_init(void)