2 * Copyright (C) 2007 Oracle. All rights reserved.
3 * Copyright (C) 2008 Morey Roof. All rights reserved.
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public
7 * License v2 as published by the Free Software Foundation.
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * General Public License for more details.
14 * You should have received a copy of the GNU General Public
15 * License along with this program; if not, write to the
16 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
17 * Boston, MA 021110-1307, USA.
23 #include <sys/ioctl.h>
24 #include <sys/mount.h>
25 #include <sys/types.h>
27 #include <uuid/uuid.h>
32 #include <linux/loop.h>
33 #include <linux/major.h>
34 #include <linux/kdev_t.h>
36 #include <blkid/blkid.h>
38 #include <sys/statfs.h>
39 #include <linux/magic.h>
42 #include "kerncompat.h"
43 #include "radix-tree.h"
46 #include "transaction.h"
54 #define BLKDISCARD _IO(0x12,119)
57 static int btrfs_scan_done = 0;
59 static char argv0_buf[ARGV0_BUF_SIZE] = "btrfs";
61 static int rand_seed_initlized = 0;
62 static unsigned short rand_seed[3];
64 const char *get_argv0_buf(void)
69 void fixup_argv0(char **argv, const char *token)
71 int len = strlen(argv0_buf);
73 snprintf(argv0_buf + len, sizeof(argv0_buf) - len, " %s", token);
77 void set_argv0(char **argv)
79 strncpy(argv0_buf, argv[0], sizeof(argv0_buf));
80 argv0_buf[sizeof(argv0_buf) - 1] = 0;
83 int check_argc_exact(int nargs, int expected)
86 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
88 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
90 return nargs != expected;
93 int check_argc_min(int nargs, int expected)
95 if (nargs < expected) {
96 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
103 int check_argc_max(int nargs, int expected)
105 if (nargs > expected) {
106 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
115 * Discard the given range in one go
117 static int discard_range(int fd, u64 start, u64 len)
119 u64 range[2] = { start, len };
121 if (ioctl(fd, BLKDISCARD, &range) < 0)
127 * Discard blocks in the given range in 1G chunks, the process is interruptible
129 static int discard_blocks(int fd, u64 start, u64 len)
133 u64 chunk_size = min_t(u64, len, 1*1024*1024*1024);
136 ret = discard_range(fd, start, chunk_size);
146 static u64 reference_root_table[] = {
147 [1] = BTRFS_ROOT_TREE_OBJECTID,
148 [2] = BTRFS_EXTENT_TREE_OBJECTID,
149 [3] = BTRFS_CHUNK_TREE_OBJECTID,
150 [4] = BTRFS_DEV_TREE_OBJECTID,
151 [5] = BTRFS_FS_TREE_OBJECTID,
152 [6] = BTRFS_CSUM_TREE_OBJECTID,
155 int test_uuid_unique(char *fs_uuid)
158 blkid_dev_iterate iter = NULL;
159 blkid_dev dev = NULL;
160 blkid_cache cache = NULL;
162 if (blkid_get_cache(&cache, NULL) < 0) {
163 printf("ERROR: lblkid cache get failed\n");
166 blkid_probe_all(cache);
167 iter = blkid_dev_iterate_begin(cache);
168 blkid_dev_set_search(iter, "UUID", fs_uuid);
170 while (blkid_dev_next(iter, &dev) == 0) {
171 dev = blkid_verify(cache, dev);
178 blkid_dev_iterate_end(iter);
179 blkid_put_cache(cache);
185 * Reserve space from free_tree.
186 * The algorithm is very simple, find the first cache_extent with enough space
187 * and allocate from its beginning.
189 static int reserve_free_space(struct cache_tree *free_tree, u64 len,
192 struct cache_extent *cache;
196 cache = first_cache_extent(free_tree);
198 if (cache->size > len) {
200 *ret_start = cache->start;
203 if (cache->size == 0) {
204 remove_cache_extent(free_tree, cache);
211 cache = next_cache_extent(cache);
218 static inline int write_temp_super(int fd, struct btrfs_super_block *sb,
224 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
225 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
226 btrfs_csum_final(crc, (char *)&sb->csum[0]);
227 ret = pwrite(fd, sb, BTRFS_SUPER_INFO_SIZE, sb_bytenr);
228 if (ret < BTRFS_SUPER_INFO_SIZE)
229 ret = (ret < 0 ? -errno : -EIO);
236 * Setup temporary superblock at cfg->super_bynter
237 * Needed info are extracted from cfg, and root_bytenr, chunk_bytenr
239 * For now sys chunk array will be empty and dev_item is empty too.
240 * They will be re-initialized at temp chunk tree setup.
242 * 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;
254 * We rely on cfg->chunk_uuid and cfg->fs_uuid to pass uuid
255 * for other functions.
256 * Caller must allocate space for them
258 BUG_ON(!cfg->chunk_uuid || !cfg->fs_uuid);
259 memset(super_buf, 0, BTRFS_SUPER_INFO_SIZE);
260 cfg->num_bytes = round_down(cfg->num_bytes, cfg->sectorsize);
262 if (cfg->fs_uuid && *cfg->fs_uuid) {
263 if (uuid_parse(cfg->fs_uuid, super->fsid) != 0) {
264 error("cound not parse UUID: %s", cfg->fs_uuid);
268 if (!test_uuid_unique(cfg->fs_uuid)) {
269 error("non-unique UUID: %s", cfg->fs_uuid);
274 uuid_generate(super->fsid);
275 uuid_unparse(super->fsid, cfg->fs_uuid);
277 uuid_generate(chunk_uuid);
278 uuid_unparse(chunk_uuid, cfg->chunk_uuid);
280 btrfs_set_super_bytenr(super, cfg->super_bytenr);
281 btrfs_set_super_num_devices(super, 1);
282 btrfs_set_super_magic(super, BTRFS_MAGIC_PARTIAL);
283 btrfs_set_super_generation(super, 1);
284 btrfs_set_super_root(super, root_bytenr);
285 btrfs_set_super_chunk_root(super, chunk_bytenr);
286 btrfs_set_super_total_bytes(super, cfg->num_bytes);
288 * Temporary filesystem will only have 6 tree roots:
289 * chunk tree, root tree, extent_tree, device tree, fs tree
292 btrfs_set_super_bytes_used(super, 6 * cfg->nodesize);
293 btrfs_set_super_sectorsize(super, cfg->sectorsize);
294 btrfs_set_super_leafsize(super, cfg->nodesize);
295 btrfs_set_super_nodesize(super, cfg->nodesize);
296 btrfs_set_super_stripesize(super, cfg->stripesize);
297 btrfs_set_super_csum_type(super, BTRFS_CSUM_TYPE_CRC32);
298 btrfs_set_super_chunk_root(super, chunk_bytenr);
299 btrfs_set_super_cache_generation(super, -1);
300 btrfs_set_super_incompat_flags(super, cfg->features);
302 __strncpy_null(super->label, cfg->label, BTRFS_LABEL_SIZE - 1);
304 /* Sys chunk array will be re-initialized at chunk tree init time */
305 super->sys_chunk_array_size = 0;
307 ret = write_temp_super(fd, super, cfg->super_bytenr);
313 * Setup an extent buffer for tree block.
315 static int setup_temp_extent_buffer(struct extent_buffer *buf,
316 struct btrfs_mkfs_config *cfg,
317 u64 bytenr, u64 owner)
319 unsigned char fsid[BTRFS_FSID_SIZE];
320 unsigned char chunk_uuid[BTRFS_UUID_SIZE];
323 /* We rely on cfg->fs_uuid and chunk_uuid to fsid and chunk uuid */
324 BUG_ON(!cfg->fs_uuid || !cfg->chunk_uuid);
325 ret = uuid_parse(cfg->fs_uuid, fsid);
328 ret = uuid_parse(cfg->chunk_uuid, chunk_uuid);
332 memset(buf->data, 0, cfg->nodesize);
333 buf->len = cfg->nodesize;
334 btrfs_set_header_bytenr(buf, bytenr);
335 btrfs_set_header_generation(buf, 1);
336 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
337 btrfs_set_header_owner(buf, owner);
338 btrfs_set_header_flags(buf, BTRFS_HEADER_FLAG_WRITTEN);
339 write_extent_buffer(buf, chunk_uuid, btrfs_header_chunk_tree_uuid(buf),
341 write_extent_buffer(buf, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
345 static inline int write_temp_extent_buffer(int fd, struct extent_buffer *buf,
350 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
352 /* Temporary extent buffer is always mapped 1:1 on disk */
353 ret = pwrite(fd, buf->data, buf->len, bytenr);
355 ret = (ret < 0 ? ret : -EIO);
362 * Insert a root item for temporary tree root
364 * Only used in make_btrfs_v2().
366 static void insert_temp_root_item(struct extent_buffer *buf,
367 struct btrfs_mkfs_config *cfg,
368 int *slot, u32 *itemoff, u64 objectid,
371 struct btrfs_root_item root_item;
372 struct btrfs_inode_item *inode_item;
373 struct btrfs_disk_key disk_key;
375 btrfs_set_header_nritems(buf, *slot + 1);
376 (*itemoff) -= sizeof(root_item);
377 memset(&root_item, 0, sizeof(root_item));
378 inode_item = &root_item.inode;
379 btrfs_set_stack_inode_generation(inode_item, 1);
380 btrfs_set_stack_inode_size(inode_item, 3);
381 btrfs_set_stack_inode_nlink(inode_item, 1);
382 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
383 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
384 btrfs_set_root_refs(&root_item, 1);
385 btrfs_set_root_used(&root_item, cfg->nodesize);
386 btrfs_set_root_generation(&root_item, 1);
387 btrfs_set_root_bytenr(&root_item, bytenr);
389 memset(&disk_key, 0, sizeof(disk_key));
390 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
391 btrfs_set_disk_key_objectid(&disk_key, objectid);
392 btrfs_set_disk_key_offset(&disk_key, 0);
394 btrfs_set_item_key(buf, &disk_key, *slot);
395 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
396 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(root_item));
397 write_extent_buffer(buf, &root_item,
398 btrfs_item_ptr_offset(buf, *slot),
403 static int setup_temp_root_tree(int fd, struct btrfs_mkfs_config *cfg,
404 u64 root_bytenr, u64 extent_bytenr,
405 u64 dev_bytenr, u64 fs_bytenr, u64 csum_bytenr)
407 struct extent_buffer *buf = NULL;
408 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
413 * Provided bytenr must in ascending order, or tree root will have a
416 BUG_ON(!(root_bytenr < extent_bytenr && extent_bytenr < dev_bytenr &&
417 dev_bytenr < fs_bytenr && fs_bytenr < csum_bytenr));
418 buf = malloc(sizeof(*buf) + cfg->nodesize);
422 ret = setup_temp_extent_buffer(buf, cfg, root_bytenr,
423 BTRFS_ROOT_TREE_OBJECTID);
427 insert_temp_root_item(buf, cfg, &slot, &itemoff,
428 BTRFS_EXTENT_TREE_OBJECTID, extent_bytenr);
429 insert_temp_root_item(buf, cfg, &slot, &itemoff,
430 BTRFS_DEV_TREE_OBJECTID, dev_bytenr);
431 insert_temp_root_item(buf, cfg, &slot, &itemoff,
432 BTRFS_FS_TREE_OBJECTID, fs_bytenr);
433 insert_temp_root_item(buf, cfg, &slot, &itemoff,
434 BTRFS_CSUM_TREE_OBJECTID, csum_bytenr);
436 ret = write_temp_extent_buffer(fd, buf, root_bytenr);
442 static int insert_temp_dev_item(int fd, struct extent_buffer *buf,
443 struct btrfs_mkfs_config *cfg,
444 int *slot, u32 *itemoff)
446 struct btrfs_disk_key disk_key;
447 struct btrfs_dev_item *dev_item;
448 char super_buf[BTRFS_SUPER_INFO_SIZE];
449 unsigned char dev_uuid[BTRFS_UUID_SIZE];
450 unsigned char fsid[BTRFS_FSID_SIZE];
451 struct btrfs_super_block *super = (struct btrfs_super_block *)super_buf;
454 ret = pread(fd, super_buf, BTRFS_SUPER_INFO_SIZE, cfg->super_bytenr);
455 if (ret < BTRFS_SUPER_INFO_SIZE) {
456 ret = (ret < 0 ? -errno : -EIO);
460 btrfs_set_header_nritems(buf, *slot + 1);
461 (*itemoff) -= sizeof(*dev_item);
462 /* setup device item 1, 0 is for replace case */
463 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
464 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
465 btrfs_set_disk_key_offset(&disk_key, 1);
466 btrfs_set_item_key(buf, &disk_key, *slot);
467 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
468 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(*dev_item));
470 dev_item = btrfs_item_ptr(buf, *slot, struct btrfs_dev_item);
471 /* Generate device uuid */
472 uuid_generate(dev_uuid);
473 write_extent_buffer(buf, dev_uuid,
474 (unsigned long)btrfs_device_uuid(dev_item),
476 uuid_parse(cfg->fs_uuid, fsid);
477 write_extent_buffer(buf, fsid,
478 (unsigned long)btrfs_device_fsid(dev_item),
480 btrfs_set_device_id(buf, dev_item, 1);
481 btrfs_set_device_generation(buf, dev_item, 0);
482 btrfs_set_device_total_bytes(buf, dev_item, cfg->num_bytes);
484 * The number must match the initial SYSTEM and META chunk size
486 btrfs_set_device_bytes_used(buf, dev_item,
487 BTRFS_MKFS_SYSTEM_GROUP_SIZE +
488 BTRFS_CONVERT_META_GROUP_SIZE);
489 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
490 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
491 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
492 btrfs_set_device_type(buf, dev_item, 0);
494 /* Super dev_item is not complete, copy the complete one to sb */
495 read_extent_buffer(buf, &super->dev_item, (unsigned long)dev_item,
497 ret = write_temp_super(fd, super, cfg->super_bytenr);
503 static int insert_temp_chunk_item(int fd, struct extent_buffer *buf,
504 struct btrfs_mkfs_config *cfg,
505 int *slot, u32 *itemoff, u64 start, u64 len,
508 struct btrfs_chunk *chunk;
509 struct btrfs_disk_key disk_key;
510 char super_buf[BTRFS_SUPER_INFO_SIZE];
511 struct btrfs_super_block *sb = (struct btrfs_super_block *)super_buf;
514 ret = pread(fd, super_buf, BTRFS_SUPER_INFO_SIZE,
516 if (ret < BTRFS_SUPER_INFO_SIZE) {
517 ret = (ret < 0 ? ret : -EIO);
521 btrfs_set_header_nritems(buf, *slot + 1);
522 (*itemoff) -= btrfs_chunk_item_size(1);
523 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
524 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
525 btrfs_set_disk_key_offset(&disk_key, start);
526 btrfs_set_item_key(buf, &disk_key, *slot);
527 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
528 btrfs_set_item_size(buf, btrfs_item_nr(*slot),
529 btrfs_chunk_item_size(1));
531 chunk = btrfs_item_ptr(buf, *slot, struct btrfs_chunk);
532 btrfs_set_chunk_length(buf, chunk, len);
533 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
534 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
535 btrfs_set_chunk_type(buf, chunk, type);
536 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
537 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
538 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
539 btrfs_set_chunk_num_stripes(buf, chunk, 1);
540 /* TODO: Support DUP profile for system chunk */
541 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
542 /* We are doing 1:1 mapping, so start is its dev offset */
543 btrfs_set_stripe_offset_nr(buf, chunk, 0, start);
544 write_extent_buffer(buf, &sb->dev_item.uuid,
545 (unsigned long)btrfs_stripe_dev_uuid_nr(chunk, 0),
550 * If it's system chunk, also copy it to super block.
552 if (type & BTRFS_BLOCK_GROUP_SYSTEM) {
555 cur = (char *)sb->sys_chunk_array + sb->sys_chunk_array_size;
556 memcpy(cur, &disk_key, sizeof(disk_key));
557 cur += sizeof(disk_key);
558 read_extent_buffer(buf, cur, (unsigned long int)chunk,
559 btrfs_chunk_item_size(1));
560 sb->sys_chunk_array_size += btrfs_chunk_item_size(1) +
563 ret = write_temp_super(fd, sb, cfg->super_bytenr);
568 static int setup_temp_chunk_tree(int fd, struct btrfs_mkfs_config *cfg,
569 u64 sys_chunk_start, u64 meta_chunk_start,
572 struct extent_buffer *buf = NULL;
573 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
577 /* Must ensure SYS chunk starts before META chunk */
578 BUG_ON(meta_chunk_start < sys_chunk_start);
579 buf = malloc(sizeof(*buf) + cfg->nodesize);
582 ret = setup_temp_extent_buffer(buf, cfg, chunk_bytenr,
583 BTRFS_CHUNK_TREE_OBJECTID);
587 ret = insert_temp_dev_item(fd, buf, cfg, &slot, &itemoff);
590 ret = insert_temp_chunk_item(fd, buf, cfg, &slot, &itemoff,
592 BTRFS_MKFS_SYSTEM_GROUP_SIZE,
593 BTRFS_BLOCK_GROUP_SYSTEM);
596 ret = insert_temp_chunk_item(fd, buf, cfg, &slot, &itemoff,
598 BTRFS_CONVERT_META_GROUP_SIZE,
599 BTRFS_BLOCK_GROUP_METADATA);
602 ret = write_temp_extent_buffer(fd, buf, chunk_bytenr);
609 static void insert_temp_dev_extent(struct extent_buffer *buf,
610 int *slot, u32 *itemoff, u64 start, u64 len)
612 struct btrfs_dev_extent *dev_extent;
613 struct btrfs_disk_key disk_key;
615 btrfs_set_header_nritems(buf, *slot + 1);
616 (*itemoff) -= sizeof(*dev_extent);
617 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
618 btrfs_set_disk_key_objectid(&disk_key, 1);
619 btrfs_set_disk_key_offset(&disk_key, start);
620 btrfs_set_item_key(buf, &disk_key, *slot);
621 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
622 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(*dev_extent));
624 dev_extent = btrfs_item_ptr(buf, *slot, struct btrfs_dev_extent);
625 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
626 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
627 btrfs_set_dev_extent_length(buf, dev_extent, len);
628 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, start);
629 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
630 BTRFS_CHUNK_TREE_OBJECTID);
634 static int setup_temp_dev_tree(int fd, struct btrfs_mkfs_config *cfg,
635 u64 sys_chunk_start, u64 meta_chunk_start,
638 struct extent_buffer *buf = NULL;
639 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
643 /* Must ensure SYS chunk starts before META chunk */
644 BUG_ON(meta_chunk_start < sys_chunk_start);
645 buf = malloc(sizeof(*buf) + cfg->nodesize);
648 ret = setup_temp_extent_buffer(buf, cfg, dev_bytenr,
649 BTRFS_DEV_TREE_OBJECTID);
652 insert_temp_dev_extent(buf, &slot, &itemoff, sys_chunk_start,
653 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
654 insert_temp_dev_extent(buf, &slot, &itemoff, meta_chunk_start,
655 BTRFS_CONVERT_META_GROUP_SIZE);
656 ret = write_temp_extent_buffer(fd, buf, dev_bytenr);
662 static int setup_temp_fs_tree(int fd, struct btrfs_mkfs_config *cfg,
665 struct extent_buffer *buf = NULL;
668 buf = malloc(sizeof(*buf) + cfg->nodesize);
671 ret = setup_temp_extent_buffer(buf, cfg, fs_bytenr,
672 BTRFS_FS_TREE_OBJECTID);
676 * Temporary fs tree is completely empty.
678 ret = write_temp_extent_buffer(fd, buf, fs_bytenr);
684 static int setup_temp_csum_tree(int fd, struct btrfs_mkfs_config *cfg,
687 struct extent_buffer *buf = NULL;
690 buf = malloc(sizeof(*buf) + cfg->nodesize);
693 ret = setup_temp_extent_buffer(buf, cfg, csum_bytenr,
694 BTRFS_CSUM_TREE_OBJECTID);
698 * Temporary csum tree is completely empty.
700 ret = write_temp_extent_buffer(fd, buf, csum_bytenr);
707 * Insert one temporary extent item.
709 * NOTE: if skinny_metadata is not enabled, this function must be called
710 * after all other trees are initialized.
711 * Or fs without skinny-metadata will be screwed up.
713 static int insert_temp_extent_item(int fd, struct extent_buffer *buf,
714 struct btrfs_mkfs_config *cfg,
715 int *slot, u32 *itemoff, u64 bytenr,
718 struct extent_buffer *tmp;
719 struct btrfs_extent_item *ei;
720 struct btrfs_extent_inline_ref *iref;
721 struct btrfs_disk_key disk_key;
722 struct btrfs_disk_key tree_info_key;
723 struct btrfs_tree_block_info *info;
725 int skinny_metadata = cfg->features &
726 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA;
730 itemsize = sizeof(*ei) + sizeof(*iref);
732 itemsize = sizeof(*ei) + sizeof(*iref) +
733 sizeof(struct btrfs_tree_block_info);
735 btrfs_set_header_nritems(buf, *slot + 1);
736 *(itemoff) -= itemsize;
738 if (skinny_metadata) {
739 btrfs_set_disk_key_type(&disk_key, BTRFS_METADATA_ITEM_KEY);
740 btrfs_set_disk_key_offset(&disk_key, 0);
742 btrfs_set_disk_key_type(&disk_key, BTRFS_EXTENT_ITEM_KEY);
743 btrfs_set_disk_key_offset(&disk_key, cfg->nodesize);
745 btrfs_set_disk_key_objectid(&disk_key, bytenr);
747 btrfs_set_item_key(buf, &disk_key, *slot);
748 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
749 btrfs_set_item_size(buf, btrfs_item_nr(*slot), itemsize);
751 ei = btrfs_item_ptr(buf, *slot, struct btrfs_extent_item);
752 btrfs_set_extent_refs(buf, ei, 1);
753 btrfs_set_extent_generation(buf, ei, 1);
754 btrfs_set_extent_flags(buf, ei, BTRFS_EXTENT_FLAG_TREE_BLOCK);
756 if (skinny_metadata) {
757 iref = (struct btrfs_extent_inline_ref *)(ei + 1);
759 info = (struct btrfs_tree_block_info *)(ei + 1);
760 iref = (struct btrfs_extent_inline_ref *)(info + 1);
762 btrfs_set_extent_inline_ref_type(buf, iref,
763 BTRFS_TREE_BLOCK_REF_KEY);
764 btrfs_set_extent_inline_ref_offset(buf, iref, ref_root);
771 * Lastly, check the tree block key by read the tree block
772 * Since we do 1:1 mapping for convert case, we can directly
773 * read the bytenr from disk
775 tmp = malloc(sizeof(*tmp) + cfg->nodesize);
778 ret = setup_temp_extent_buffer(tmp, cfg, bytenr, ref_root);
781 ret = pread(fd, tmp->data, cfg->nodesize, bytenr);
782 if (ret < cfg->nodesize) {
783 ret = (ret < 0 ? -errno : -EIO);
786 if (btrfs_header_nritems(tmp) == 0) {
787 btrfs_set_disk_key_type(&tree_info_key, 0);
788 btrfs_set_disk_key_objectid(&tree_info_key, 0);
789 btrfs_set_disk_key_offset(&tree_info_key, 0);
791 btrfs_item_key(tmp, &tree_info_key, 0);
793 btrfs_set_tree_block_key(buf, info, &tree_info_key);
800 static void insert_temp_block_group(struct extent_buffer *buf,
801 struct btrfs_mkfs_config *cfg,
802 int *slot, u32 *itemoff,
803 u64 bytenr, u64 len, u64 used, u64 flag)
805 struct btrfs_block_group_item bgi;
806 struct btrfs_disk_key disk_key;
808 btrfs_set_header_nritems(buf, *slot + 1);
809 (*itemoff) -= sizeof(bgi);
810 btrfs_set_disk_key_type(&disk_key, BTRFS_BLOCK_GROUP_ITEM_KEY);
811 btrfs_set_disk_key_objectid(&disk_key, bytenr);
812 btrfs_set_disk_key_offset(&disk_key, len);
813 btrfs_set_item_key(buf, &disk_key, *slot);
814 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
815 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(bgi));
817 btrfs_set_block_group_flags(&bgi, flag);
818 btrfs_set_block_group_used(&bgi, used);
819 btrfs_set_block_group_chunk_objectid(&bgi,
820 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
821 write_extent_buffer(buf, &bgi, btrfs_item_ptr_offset(buf, *slot),
826 static int setup_temp_extent_tree(int fd, struct btrfs_mkfs_config *cfg,
827 u64 chunk_bytenr, u64 root_bytenr,
828 u64 extent_bytenr, u64 dev_bytenr,
829 u64 fs_bytenr, u64 csum_bytenr)
831 struct extent_buffer *buf = NULL;
832 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
837 * We must ensure provided bytenr are in ascending order,
838 * or extent tree key order will be broken.
840 BUG_ON(!(chunk_bytenr < root_bytenr && root_bytenr < extent_bytenr &&
841 extent_bytenr < dev_bytenr && dev_bytenr < fs_bytenr &&
842 fs_bytenr < csum_bytenr));
843 buf = malloc(sizeof(*buf) + cfg->nodesize);
847 ret = setup_temp_extent_buffer(buf, cfg, extent_bytenr,
848 BTRFS_EXTENT_TREE_OBJECTID);
852 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
853 chunk_bytenr, BTRFS_CHUNK_TREE_OBJECTID);
857 insert_temp_block_group(buf, cfg, &slot, &itemoff, chunk_bytenr,
858 BTRFS_MKFS_SYSTEM_GROUP_SIZE, cfg->nodesize,
859 BTRFS_BLOCK_GROUP_SYSTEM);
861 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
862 root_bytenr, BTRFS_ROOT_TREE_OBJECTID);
866 /* 5 tree block used, root, extent, dev, fs and csum*/
867 insert_temp_block_group(buf, cfg, &slot, &itemoff, root_bytenr,
868 BTRFS_CONVERT_META_GROUP_SIZE, cfg->nodesize * 5,
869 BTRFS_BLOCK_GROUP_METADATA);
871 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
872 extent_bytenr, BTRFS_EXTENT_TREE_OBJECTID);
875 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
876 dev_bytenr, BTRFS_DEV_TREE_OBJECTID);
879 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
880 fs_bytenr, BTRFS_FS_TREE_OBJECTID);
883 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
884 csum_bytenr, BTRFS_CSUM_TREE_OBJECTID);
888 ret = write_temp_extent_buffer(fd, buf, extent_bytenr);
895 * Improved version of make_btrfs().
898 * 1) Do chunk allocation to avoid used data
899 * And after this function, extent type matches chunk type
900 * 2) Better structured code
901 * No super long hand written codes to initialized all tree blocks
902 * Split into small blocks and reuse codes.
903 * TODO: Reuse tree operation facilities by introducing new flags
905 static int make_convert_btrfs(int fd, struct btrfs_mkfs_config *cfg,
906 struct btrfs_convert_context *cctx)
908 struct cache_tree *free = &cctx->free;
909 struct cache_tree *used = &cctx->used;
911 u64 meta_chunk_start;
912 /* chunk tree bytenr, in system chunk */
914 /* metadata trees bytenr, in metadata chunk */
922 /* Shouldn't happen */
923 BUG_ON(cache_tree_empty(used));
926 * reserve space for temporary superblock first
927 * Here we allocate a little larger space, to keep later
928 * free space will be STRIPE_LEN aligned
930 ret = reserve_free_space(free, BTRFS_STRIPE_LEN,
936 * Then reserve system chunk space
937 * TODO: Change system group size depending on cctx->total_bytes.
938 * If using current 4M, it can only handle less than one TB for
939 * worst case and then run out of sys space.
941 ret = reserve_free_space(free, BTRFS_MKFS_SYSTEM_GROUP_SIZE,
945 ret = reserve_free_space(free, BTRFS_CONVERT_META_GROUP_SIZE,
951 * Allocated meta/sys chunks will be mapped 1:1 with device offset.
953 * Inside the allocated metadata chunk, the layout will be:
954 * | offset | contents |
955 * -------------------------------------
957 * | +nodesize | extent root |
958 * | +nodesize * 2 | device root |
959 * | +nodesize * 3 | fs tree |
960 * | +nodesize * 4 | csum tree |
961 * -------------------------------------
962 * Inside the allocated system chunk, the layout will be:
963 * | offset | contents |
964 * -------------------------------------
965 * | +0 | chunk root |
966 * -------------------------------------
968 chunk_bytenr = sys_chunk_start;
969 root_bytenr = meta_chunk_start;
970 extent_bytenr = meta_chunk_start + cfg->nodesize;
971 dev_bytenr = meta_chunk_start + cfg->nodesize * 2;
972 fs_bytenr = meta_chunk_start + cfg->nodesize * 3;
973 csum_bytenr = meta_chunk_start + cfg->nodesize * 4;
975 ret = setup_temp_super(fd, cfg, root_bytenr, chunk_bytenr);
979 ret = setup_temp_root_tree(fd, cfg, root_bytenr, extent_bytenr,
980 dev_bytenr, fs_bytenr, csum_bytenr);
983 ret = setup_temp_chunk_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
987 ret = setup_temp_dev_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
991 ret = setup_temp_fs_tree(fd, cfg, fs_bytenr);
994 ret = setup_temp_csum_tree(fd, cfg, csum_bytenr);
998 * Setup extent tree last, since it may need to read tree block key
999 * for non-skinny metadata case.
1001 ret = setup_temp_extent_tree(fd, cfg, chunk_bytenr, root_bytenr,
1002 extent_bytenr, dev_bytenr, fs_bytenr,
1009 * @fs_uuid - if NULL, generates a UUID, returns back the new filesystem UUID
1011 * The superblock signature is not valid, denotes a partially created
1012 * filesystem, needs to be finalized.
1014 int make_btrfs(int fd, struct btrfs_mkfs_config *cfg,
1015 struct btrfs_convert_context *cctx)
1017 struct btrfs_super_block super;
1018 struct extent_buffer *buf;
1019 struct btrfs_root_item root_item;
1020 struct btrfs_disk_key disk_key;
1021 struct btrfs_extent_item *extent_item;
1022 struct btrfs_inode_item *inode_item;
1023 struct btrfs_chunk *chunk;
1024 struct btrfs_dev_item *dev_item;
1025 struct btrfs_dev_extent *dev_extent;
1026 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
1036 int skinny_metadata = !!(cfg->features &
1037 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA);
1041 return make_convert_btrfs(fd, cfg, cctx);
1042 buf = malloc(sizeof(*buf) + max(cfg->sectorsize, cfg->nodesize));
1046 first_free = BTRFS_SUPER_INFO_OFFSET + cfg->sectorsize * 2 - 1;
1047 first_free &= ~((u64)cfg->sectorsize - 1);
1049 memset(&super, 0, sizeof(super));
1051 num_bytes = (cfg->num_bytes / cfg->sectorsize) * cfg->sectorsize;
1052 if (cfg->fs_uuid && *cfg->fs_uuid) {
1053 if (uuid_parse(cfg->fs_uuid, super.fsid) != 0) {
1054 error("cannot not parse UUID: %s", cfg->fs_uuid);
1058 if (!test_uuid_unique(cfg->fs_uuid)) {
1059 error("non-unique UUID: %s", cfg->fs_uuid);
1064 uuid_generate(super.fsid);
1066 uuid_unparse(super.fsid, cfg->fs_uuid);
1068 uuid_generate(super.dev_item.uuid);
1069 uuid_generate(chunk_tree_uuid);
1071 btrfs_set_super_bytenr(&super, cfg->blocks[0]);
1072 btrfs_set_super_num_devices(&super, 1);
1073 btrfs_set_super_magic(&super, BTRFS_MAGIC_PARTIAL);
1074 btrfs_set_super_generation(&super, 1);
1075 btrfs_set_super_root(&super, cfg->blocks[1]);
1076 btrfs_set_super_chunk_root(&super, cfg->blocks[3]);
1077 btrfs_set_super_total_bytes(&super, num_bytes);
1078 btrfs_set_super_bytes_used(&super, 6 * cfg->nodesize);
1079 btrfs_set_super_sectorsize(&super, cfg->sectorsize);
1080 btrfs_set_super_leafsize(&super, cfg->nodesize);
1081 btrfs_set_super_nodesize(&super, cfg->nodesize);
1082 btrfs_set_super_stripesize(&super, cfg->stripesize);
1083 btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
1084 btrfs_set_super_chunk_root_generation(&super, 1);
1085 btrfs_set_super_cache_generation(&super, -1);
1086 btrfs_set_super_incompat_flags(&super, cfg->features);
1088 __strncpy_null(super.label, cfg->label, BTRFS_LABEL_SIZE - 1);
1090 /* create the tree of root objects */
1091 memset(buf->data, 0, cfg->nodesize);
1092 buf->len = cfg->nodesize;
1093 btrfs_set_header_bytenr(buf, cfg->blocks[1]);
1094 btrfs_set_header_nritems(buf, 4);
1095 btrfs_set_header_generation(buf, 1);
1096 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
1097 btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
1098 write_extent_buffer(buf, super.fsid, btrfs_header_fsid(),
1101 write_extent_buffer(buf, chunk_tree_uuid,
1102 btrfs_header_chunk_tree_uuid(buf),
1105 /* create the items for the root tree */
1106 memset(&root_item, 0, sizeof(root_item));
1107 inode_item = &root_item.inode;
1108 btrfs_set_stack_inode_generation(inode_item, 1);
1109 btrfs_set_stack_inode_size(inode_item, 3);
1110 btrfs_set_stack_inode_nlink(inode_item, 1);
1111 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
1112 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
1113 btrfs_set_root_refs(&root_item, 1);
1114 btrfs_set_root_used(&root_item, cfg->nodesize);
1115 btrfs_set_root_generation(&root_item, 1);
1117 memset(&disk_key, 0, sizeof(disk_key));
1118 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
1119 btrfs_set_disk_key_offset(&disk_key, 0);
1122 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - sizeof(root_item);
1123 btrfs_set_root_bytenr(&root_item, cfg->blocks[2]);
1124 btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
1125 btrfs_set_item_key(buf, &disk_key, nritems);
1126 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1127 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1129 write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
1130 nritems), sizeof(root_item));
1133 itemoff = itemoff - sizeof(root_item);
1134 btrfs_set_root_bytenr(&root_item, cfg->blocks[4]);
1135 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
1136 btrfs_set_item_key(buf, &disk_key, nritems);
1137 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1138 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1140 write_extent_buffer(buf, &root_item,
1141 btrfs_item_ptr_offset(buf, nritems),
1145 itemoff = itemoff - sizeof(root_item);
1146 btrfs_set_root_bytenr(&root_item, cfg->blocks[5]);
1147 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
1148 btrfs_set_item_key(buf, &disk_key, nritems);
1149 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1150 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1152 write_extent_buffer(buf, &root_item,
1153 btrfs_item_ptr_offset(buf, nritems),
1157 itemoff = itemoff - sizeof(root_item);
1158 btrfs_set_root_bytenr(&root_item, cfg->blocks[6]);
1159 btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_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,
1165 btrfs_item_ptr_offset(buf, nritems),
1170 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1171 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[1]);
1172 if (ret != cfg->nodesize) {
1173 ret = (ret < 0 ? -errno : -EIO);
1177 /* create the items for the extent tree */
1178 memset(buf->data + sizeof(struct btrfs_header), 0,
1179 cfg->nodesize - sizeof(struct btrfs_header));
1181 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
1182 for (i = 1; i < 7; i++) {
1183 item_size = sizeof(struct btrfs_extent_item);
1184 if (!skinny_metadata)
1185 item_size += sizeof(struct btrfs_tree_block_info);
1187 BUG_ON(cfg->blocks[i] < first_free);
1188 BUG_ON(cfg->blocks[i] < 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, 64 * 1024);
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 BUG_ON(sizeof(super) > cfg->sectorsize);
1389 memset(buf->data, 0, BTRFS_SUPER_INFO_SIZE);
1390 memcpy(buf->data, &super, sizeof(super));
1391 buf->len = BTRFS_SUPER_INFO_SIZE;
1392 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1393 ret = pwrite(fd, buf->data, BTRFS_SUPER_INFO_SIZE, cfg->blocks[0]);
1394 if (ret != BTRFS_SUPER_INFO_SIZE) {
1395 ret = (ret < 0 ? -errno : -EIO);
1406 static const struct btrfs_fs_feature {
1410 } mkfs_features[] = {
1411 { "mixed-bg", BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS,
1412 "mixed data and metadata block groups" },
1413 { "extref", BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF,
1414 "increased hardlink limit per file to 65536" },
1415 { "raid56", BTRFS_FEATURE_INCOMPAT_RAID56,
1416 "raid56 extended format" },
1417 { "skinny-metadata", BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA,
1418 "reduced-size metadata extent refs" },
1419 { "no-holes", BTRFS_FEATURE_INCOMPAT_NO_HOLES,
1420 "no explicit hole extents for files" },
1421 /* Keep this one last */
1422 { "list-all", BTRFS_FEATURE_LIST_ALL, NULL }
1425 static int parse_one_fs_feature(const char *name, u64 *flags)
1430 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1431 if (name[0] == '^' &&
1432 !strcmp(mkfs_features[i].name, name + 1)) {
1433 *flags &= ~ mkfs_features[i].flag;
1435 } else if (!strcmp(mkfs_features[i].name, name)) {
1436 *flags |= mkfs_features[i].flag;
1444 void btrfs_parse_features_to_string(char *buf, u64 flags)
1450 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1451 if (flags & mkfs_features[i].flag) {
1454 strcat(buf, mkfs_features[i].name);
1459 void btrfs_process_fs_features(u64 flags)
1463 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1464 if (flags & mkfs_features[i].flag) {
1465 printf("Turning ON incompat feature '%s': %s\n",
1466 mkfs_features[i].name,
1467 mkfs_features[i].desc);
1472 void btrfs_list_all_fs_features(u64 mask_disallowed)
1476 fprintf(stderr, "Filesystem features available:\n");
1477 for (i = 0; i < ARRAY_SIZE(mkfs_features) - 1; i++) {
1478 char *is_default = "";
1480 if (mkfs_features[i].flag & mask_disallowed)
1482 if (mkfs_features[i].flag & BTRFS_MKFS_DEFAULT_FEATURES)
1483 is_default = ", default";
1484 fprintf(stderr, "%-20s- %s (0x%llx%s)\n",
1485 mkfs_features[i].name,
1486 mkfs_features[i].desc,
1487 mkfs_features[i].flag,
1493 * Return NULL if all features were parsed fine, otherwise return the name of
1494 * the first unparsed.
1496 char* btrfs_parse_fs_features(char *namelist, u64 *flags)
1499 char *save_ptr = NULL; /* Satisfy static checkers */
1501 for (this_char = strtok_r(namelist, ",", &save_ptr);
1503 this_char = strtok_r(NULL, ",", &save_ptr)) {
1504 if (parse_one_fs_feature(this_char, flags))
1511 u64 btrfs_device_size(int fd, struct stat *st)
1514 if (S_ISREG(st->st_mode)) {
1517 if (!S_ISBLK(st->st_mode)) {
1520 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
1526 static int zero_blocks(int fd, off_t start, size_t len)
1528 char *buf = malloc(len);
1534 memset(buf, 0, len);
1535 written = pwrite(fd, buf, len, start);
1542 #define ZERO_DEV_BYTES (2 * 1024 * 1024)
1544 /* don't write outside the device by clamping the region to the device size */
1545 static int zero_dev_clamped(int fd, off_t start, ssize_t len, u64 dev_size)
1547 off_t end = max(start, start + len);
1550 /* and don't overwrite the disk labels on sparc */
1551 start = max(start, 1024);
1552 end = max(end, 1024);
1555 start = min_t(u64, start, dev_size);
1556 end = min_t(u64, end, dev_size);
1558 return zero_blocks(fd, start, end - start);
1561 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
1562 struct btrfs_root *root, int fd, char *path,
1563 u64 device_total_bytes, u32 io_width, u32 io_align,
1566 struct btrfs_super_block *disk_super;
1567 struct btrfs_super_block *super = root->fs_info->super_copy;
1568 struct btrfs_device *device;
1569 struct btrfs_dev_item *dev_item;
1575 device_total_bytes = (device_total_bytes / sectorsize) * sectorsize;
1577 device = kzalloc(sizeof(*device), GFP_NOFS);
1580 buf = kzalloc(sectorsize, GFP_NOFS);
1583 BUG_ON(sizeof(*disk_super) > sectorsize);
1585 disk_super = (struct btrfs_super_block *)buf;
1586 dev_item = &disk_super->dev_item;
1588 uuid_generate(device->uuid);
1591 device->io_width = io_width;
1592 device->io_align = io_align;
1593 device->sector_size = sectorsize;
1595 device->writeable = 1;
1596 device->total_bytes = device_total_bytes;
1597 device->bytes_used = 0;
1598 device->total_ios = 0;
1599 device->dev_root = root->fs_info->dev_root;
1600 device->name = strdup(path);
1604 INIT_LIST_HEAD(&device->dev_list);
1605 ret = btrfs_add_device(trans, root, device);
1608 fs_total_bytes = btrfs_super_total_bytes(super) + device_total_bytes;
1609 btrfs_set_super_total_bytes(super, fs_total_bytes);
1611 num_devs = btrfs_super_num_devices(super) + 1;
1612 btrfs_set_super_num_devices(super, num_devs);
1614 memcpy(disk_super, super, sizeof(*disk_super));
1616 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
1617 btrfs_set_stack_device_id(dev_item, device->devid);
1618 btrfs_set_stack_device_type(dev_item, device->type);
1619 btrfs_set_stack_device_io_align(dev_item, device->io_align);
1620 btrfs_set_stack_device_io_width(dev_item, device->io_width);
1621 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
1622 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
1623 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
1624 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
1626 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
1627 BUG_ON(ret != sectorsize);
1630 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
1631 device->fs_devices = root->fs_info->fs_devices;
1640 static int btrfs_wipe_existing_sb(int fd)
1642 const char *off = NULL;
1647 blkid_probe pr = NULL;
1649 pr = blkid_new_probe();
1653 if (blkid_probe_set_device(pr, fd, 0, 0)) {
1658 ret = blkid_probe_lookup_value(pr, "SBMAGIC_OFFSET", &off, NULL);
1660 ret = blkid_probe_lookup_value(pr, "SBMAGIC", NULL, &len);
1662 if (ret || len == 0 || off == NULL) {
1664 * If lookup fails, the probe did not find any values, eg. for
1665 * a file image or a loop device. Soft error.
1671 offset = strtoll(off, NULL, 10);
1672 if (len > sizeof(buf))
1675 memset(buf, 0, len);
1676 ret = pwrite(fd, buf, len, offset);
1678 error("cannot wipe existing superblock: %s", strerror(errno));
1680 } else if (ret != len) {
1681 error("cannot wipe existing superblock: wrote %d of %zd", ret, len);
1687 blkid_free_probe(pr);
1691 int btrfs_prepare_device(int fd, const char *file, u64 *block_count_ret,
1692 u64 max_block_count, unsigned opflags)
1698 ret = fstat(fd, &st);
1700 error("unable to stat %s: %s", file, strerror(errno));
1704 block_count = btrfs_device_size(fd, &st);
1705 if (block_count == 0) {
1706 error("unable to determine size of %s", file);
1709 if (max_block_count)
1710 block_count = min(block_count, max_block_count);
1712 if (opflags & PREP_DEVICE_DISCARD) {
1714 * We intentionally ignore errors from the discard ioctl. It
1715 * is not necessary for the mkfs functionality but just an
1718 if (discard_range(fd, 0, 0) == 0) {
1719 if (opflags & PREP_DEVICE_VERBOSE)
1720 printf("Performing full device TRIM (%s) ...\n",
1721 pretty_size(block_count));
1722 discard_blocks(fd, 0, block_count);
1726 ret = zero_dev_clamped(fd, 0, ZERO_DEV_BYTES, block_count);
1727 for (i = 0 ; !ret && i < BTRFS_SUPER_MIRROR_MAX; i++)
1728 ret = zero_dev_clamped(fd, btrfs_sb_offset(i),
1729 BTRFS_SUPER_INFO_SIZE, block_count);
1730 if (!ret && (opflags & PREP_DEVICE_ZERO_END))
1731 ret = zero_dev_clamped(fd, block_count - ZERO_DEV_BYTES,
1732 ZERO_DEV_BYTES, block_count);
1735 error("failed to zero device '%s': %s", file, strerror(-ret));
1739 ret = btrfs_wipe_existing_sb(fd);
1741 error("cannot wipe superblocks on %s", file);
1745 *block_count_ret = block_count;
1749 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
1750 struct btrfs_root *root, u64 objectid)
1753 struct btrfs_inode_item inode_item;
1754 time_t now = time(NULL);
1756 memset(&inode_item, 0, sizeof(inode_item));
1757 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
1758 btrfs_set_stack_inode_size(&inode_item, 0);
1759 btrfs_set_stack_inode_nlink(&inode_item, 1);
1760 btrfs_set_stack_inode_nbytes(&inode_item, root->nodesize);
1761 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
1762 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
1763 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
1764 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
1765 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
1766 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
1767 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
1768 btrfs_set_stack_timespec_sec(&inode_item.otime, 0);
1769 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
1771 if (root->fs_info->tree_root == root)
1772 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
1774 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
1778 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
1782 btrfs_set_root_dirid(&root->root_item, objectid);
1789 * checks if a path is a block device node
1790 * Returns negative errno on failure, otherwise
1791 * returns 1 for blockdev, 0 for not-blockdev
1793 int is_block_device(const char *path)
1795 struct stat statbuf;
1797 if (stat(path, &statbuf) < 0)
1800 return !!S_ISBLK(statbuf.st_mode);
1804 * check if given path is a mount point
1805 * return 1 if yes. 0 if no. -1 for error
1807 int is_mount_point(const char *path)
1813 f = setmntent("/proc/self/mounts", "r");
1817 while ((mnt = getmntent(f)) != NULL) {
1818 if (strcmp(mnt->mnt_dir, path))
1827 static int is_reg_file(const char *path)
1829 struct stat statbuf;
1831 if (stat(path, &statbuf) < 0)
1833 return S_ISREG(statbuf.st_mode);
1837 * This function checks if the given input parameter is
1839 * return <0 : some error in the given input
1840 * return BTRFS_ARG_UNKNOWN: unknown input
1841 * return BTRFS_ARG_UUID: given input is uuid
1842 * return BTRFS_ARG_MNTPOINT: given input is path
1843 * return BTRFS_ARG_REG: given input is regular file
1844 * return BTRFS_ARG_BLKDEV: given input is block device
1846 int check_arg_type(const char *input)
1849 char path[PATH_MAX];
1854 if (realpath(input, path)) {
1855 if (is_block_device(path) == 1)
1856 return BTRFS_ARG_BLKDEV;
1858 if (is_mount_point(path) == 1)
1859 return BTRFS_ARG_MNTPOINT;
1861 if (is_reg_file(path))
1862 return BTRFS_ARG_REG;
1864 return BTRFS_ARG_UNKNOWN;
1867 if (strlen(input) == (BTRFS_UUID_UNPARSED_SIZE - 1) &&
1868 !uuid_parse(input, uuid))
1869 return BTRFS_ARG_UUID;
1871 return BTRFS_ARG_UNKNOWN;
1875 * Find the mount point for a mounted device.
1876 * On success, returns 0 with mountpoint in *mp.
1877 * On failure, returns -errno (not mounted yields -EINVAL)
1878 * Is noisy on failures, expects to be given a mounted device.
1880 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size)
1885 ret = is_block_device(dev);
1888 error("not a block device: %s", dev);
1891 error("cannot check %s: %s", dev, strerror(-ret));
1896 fd = open(dev, O_RDONLY);
1899 error("cannot open %s: %s", dev, strerror(errno));
1903 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
1906 } else { /* mounted, all good */
1916 * Given a pathname, return a filehandle to:
1917 * the original pathname or,
1918 * if the pathname is a mounted btrfs device, to its mountpoint.
1920 * On error, return -1, errno should be set.
1922 int open_path_or_dev_mnt(const char *path, DIR **dirstream, int verbose)
1927 if (is_block_device(path)) {
1928 ret = get_btrfs_mount(path, mp, sizeof(mp));
1930 /* not a mounted btrfs dev */
1931 error_on(verbose, "'%s' is not a mounted btrfs device",
1936 ret = open_file_or_dir(mp, dirstream);
1937 error_on(verbose && ret < 0, "can't access '%s': %s",
1938 path, strerror(errno));
1940 ret = btrfs_open_dir(path, dirstream, 1);
1947 * Do the following checks before calling open_file_or_dir():
1948 * 1: path is in a btrfs filesystem
1949 * 2: path is a directory
1951 int btrfs_open_dir(const char *path, DIR **dirstream, int verbose)
1957 if (statfs(path, &stfs) != 0) {
1958 error_on(verbose, "cannot access '%s': %s", path,
1963 if (stfs.f_type != BTRFS_SUPER_MAGIC) {
1964 error_on(verbose, "not a btrfs filesystem: %s", path);
1968 if (stat(path, &st) != 0) {
1969 error_on(verbose, "cannot access '%s': %s", path,
1974 if (!S_ISDIR(st.st_mode)) {
1975 error_on(verbose, "not a directory: %s", path);
1979 ret = open_file_or_dir(path, dirstream);
1981 error_on(verbose, "cannot access '%s': %s", path,
1988 /* checks if a device is a loop device */
1989 static int is_loop_device (const char* device) {
1990 struct stat statbuf;
1992 if(stat(device, &statbuf) < 0)
1995 return (S_ISBLK(statbuf.st_mode) &&
1996 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
2000 * Takes a loop device path (e.g. /dev/loop0) and returns
2001 * the associated file (e.g. /images/my_btrfs.img) using
2004 static int resolve_loop_device_with_loopdev(const char* loop_dev, char* loop_file)
2008 struct loop_info64 lo64;
2010 fd = open(loop_dev, O_RDONLY | O_NONBLOCK);
2013 ret = ioctl(fd, LOOP_GET_STATUS64, &lo64);
2019 memcpy(loop_file, lo64.lo_file_name, sizeof(lo64.lo_file_name));
2020 loop_file[sizeof(lo64.lo_file_name)] = 0;
2028 /* Takes a loop device path (e.g. /dev/loop0) and returns
2029 * the associated file (e.g. /images/my_btrfs.img) */
2030 static int resolve_loop_device(const char* loop_dev, char* loop_file,
2037 char real_loop_dev[PATH_MAX];
2039 if (!realpath(loop_dev, real_loop_dev))
2041 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
2042 if (!(f = fopen(p, "r"))) {
2043 if (errno == ENOENT)
2045 * It's possibly a partitioned loop device, which is
2046 * resolvable with loopdev API.
2048 return resolve_loop_device_with_loopdev(loop_dev, loop_file);
2052 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
2053 ret = fscanf(f, fmt, loop_file);
2062 * Checks whether a and b are identical or device
2063 * files associated with the same block device
2065 static int is_same_blk_file(const char* a, const char* b)
2067 struct stat st_buf_a, st_buf_b;
2068 char real_a[PATH_MAX];
2069 char real_b[PATH_MAX];
2071 if (!realpath(a, real_a))
2072 strncpy_null(real_a, a);
2074 if (!realpath(b, real_b))
2075 strncpy_null(real_b, b);
2077 /* Identical path? */
2078 if (strcmp(real_a, real_b) == 0)
2081 if (stat(a, &st_buf_a) < 0 || stat(b, &st_buf_b) < 0) {
2082 if (errno == ENOENT)
2087 /* Same blockdevice? */
2088 if (S_ISBLK(st_buf_a.st_mode) && S_ISBLK(st_buf_b.st_mode) &&
2089 st_buf_a.st_rdev == st_buf_b.st_rdev) {
2094 if (st_buf_a.st_dev == st_buf_b.st_dev &&
2095 st_buf_a.st_ino == st_buf_b.st_ino) {
2102 /* checks if a and b are identical or device
2103 * files associated with the same block device or
2104 * if one file is a loop device that uses the other
2107 static int is_same_loop_file(const char* a, const char* b)
2109 char res_a[PATH_MAX];
2110 char res_b[PATH_MAX];
2111 const char* final_a = NULL;
2112 const char* final_b = NULL;
2115 /* Resolve a if it is a loop device */
2116 if((ret = is_loop_device(a)) < 0) {
2121 ret = resolve_loop_device(a, res_a, sizeof(res_a));
2132 /* Resolve b if it is a loop device */
2133 if ((ret = is_loop_device(b)) < 0) {
2138 ret = resolve_loop_device(b, res_b, sizeof(res_b));
2149 return is_same_blk_file(final_a, final_b);
2152 /* Checks if a file exists and is a block or regular file*/
2153 static int is_existing_blk_or_reg_file(const char* filename)
2157 if(stat(filename, &st_buf) < 0) {
2164 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
2167 /* Checks if a file is used (directly or indirectly via a loop device)
2168 * by a device in fs_devices
2170 static int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices,
2174 struct list_head *head;
2175 struct list_head *cur;
2176 struct btrfs_device *device;
2178 head = &fs_devices->devices;
2179 list_for_each(cur, head) {
2180 device = list_entry(cur, struct btrfs_device, dev_list);
2182 if((ret = is_same_loop_file(device->name, file)))
2190 * Resolve a pathname to a device mapper node to /dev/mapper/<name>
2191 * Returns NULL on invalid input or malloc failure; Other failures
2192 * will be handled by the caller using the input pathame.
2194 char *canonicalize_dm_name(const char *ptname)
2198 char path[PATH_MAX], name[PATH_MAX], *res = NULL;
2200 if (!ptname || !*ptname)
2203 snprintf(path, sizeof(path), "/sys/block/%s/dm/name", ptname);
2204 if (!(f = fopen(path, "r")))
2207 /* read <name>\n from sysfs */
2208 if (fgets(name, sizeof(name), f) && (sz = strlen(name)) > 1) {
2209 name[sz - 1] = '\0';
2210 snprintf(path, sizeof(path), "/dev/mapper/%s", name);
2212 if (access(path, F_OK) == 0)
2220 * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or
2221 * to a device mapper pathname.
2222 * Returns NULL on invalid input or malloc failure; Other failures
2223 * will be handled by the caller using the input pathame.
2225 char *canonicalize_path(const char *path)
2227 char *canonical, *p;
2229 if (!path || !*path)
2232 canonical = realpath(path, NULL);
2234 return strdup(path);
2235 p = strrchr(canonical, '/');
2236 if (p && strncmp(p, "/dm-", 4) == 0 && isdigit(*(p + 4))) {
2237 char *dm = canonicalize_dm_name(p + 1);
2248 * returns 1 if the device was mounted, < 0 on error or 0 if everything
2249 * is safe to continue.
2251 int check_mounted(const char* file)
2256 fd = open(file, O_RDONLY);
2258 error("mount check: cannot open %s: %s", file,
2263 ret = check_mounted_where(fd, file, NULL, 0, NULL);
2269 int check_mounted_where(int fd, const char *file, char *where, int size,
2270 struct btrfs_fs_devices **fs_dev_ret)
2275 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2279 /* scan the initial device */
2280 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
2281 &total_devs, BTRFS_SUPER_INFO_OFFSET, SBREAD_DEFAULT);
2282 is_btrfs = (ret >= 0);
2284 /* scan other devices */
2285 if (is_btrfs && total_devs > 1) {
2286 ret = btrfs_scan_lblkid();
2291 /* iterate over the list of currently mounted filesystems */
2292 if ((f = setmntent ("/proc/self/mounts", "r")) == NULL)
2295 while ((mnt = getmntent (f)) != NULL) {
2297 if(strcmp(mnt->mnt_type, "btrfs") != 0)
2300 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
2302 /* ignore entries in the mount table that are not
2303 associated with a file*/
2304 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
2305 goto out_mntloop_err;
2309 ret = is_same_loop_file(file, mnt->mnt_fsname);
2313 goto out_mntloop_err;
2318 /* Did we find an entry in mnt table? */
2319 if (mnt && size && where) {
2320 strncpy(where, mnt->mnt_dir, size);
2324 *fs_dev_ret = fs_devices_mnt;
2326 ret = (mnt != NULL);
2334 struct pending_dir {
2335 struct list_head list;
2336 char name[PATH_MAX];
2339 int btrfs_register_one_device(const char *fname)
2341 struct btrfs_ioctl_vol_args args;
2345 fd = open("/dev/btrfs-control", O_RDWR);
2348 "failed to open /dev/btrfs-control, skipping device registration: %s",
2352 memset(&args, 0, sizeof(args));
2353 strncpy_null(args.name, fname);
2354 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
2356 error("device scan failed on '%s': %s", fname,
2365 * Register all devices in the fs_uuid list created in the user
2366 * space. Ensure btrfs_scan_lblkid() is called before this func.
2368 int btrfs_register_all_devices(void)
2372 struct btrfs_fs_devices *fs_devices;
2373 struct btrfs_device *device;
2374 struct list_head *all_uuids;
2376 all_uuids = btrfs_scanned_uuids();
2378 list_for_each_entry(fs_devices, all_uuids, list) {
2379 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2381 err = btrfs_register_one_device(device->name);
2391 int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
2394 struct btrfs_super_block *disk_super;
2398 buf = malloc(BTRFS_SUPER_INFO_SIZE);
2403 ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
2404 if (ret != BTRFS_SUPER_INFO_SIZE)
2408 disk_super = (struct btrfs_super_block *)buf;
2410 * Accept devices from the same filesystem, allow partially created
2413 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC &&
2414 btrfs_super_magic(disk_super) != BTRFS_MAGIC_PARTIAL)
2417 if (!memcmp(disk_super->fsid, root->fs_info->super_copy->fsid,
2427 * Note: this function uses a static per-thread buffer. Do not call this
2428 * function more than 10 times within one argument list!
2430 const char *pretty_size_mode(u64 size, unsigned mode)
2432 static __thread int ps_index = 0;
2433 static __thread char ps_array[10][32];
2436 ret = ps_array[ps_index];
2439 (void)pretty_size_snprintf(size, ret, 32, mode);
2444 static const char* unit_suffix_binary[] =
2445 { "B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB"};
2446 static const char* unit_suffix_decimal[] =
2447 { "B", "kB", "MB", "GB", "TB", "PB", "EB"};
2449 int pretty_size_snprintf(u64 size, char *str, size_t str_size, unsigned unit_mode)
2455 const char** suffix = NULL;
2461 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_RAW) {
2462 snprintf(str, str_size, "%llu", size);
2466 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_BINARY) {
2469 suffix = unit_suffix_binary;
2470 } else if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_DECIMAL) {
2473 suffix = unit_suffix_decimal;
2478 fprintf(stderr, "INTERNAL ERROR: unknown unit base, mode %d\n",
2486 switch (unit_mode & UNITS_MODE_MASK) {
2487 case UNITS_TBYTES: base *= mult; num_divs++;
2488 case UNITS_GBYTES: base *= mult; num_divs++;
2489 case UNITS_MBYTES: base *= mult; num_divs++;
2490 case UNITS_KBYTES: num_divs++;
2497 while (size >= mult) {
2503 * If the value is smaller than base, we didn't do any
2504 * division, in that case, base should be 1, not original
2505 * base, or the unit will be wrong
2511 if (num_divs >= ARRAY_SIZE(unit_suffix_binary)) {
2513 printf("INTERNAL ERROR: unsupported unit suffix, index %d\n",
2518 fraction = (float)last_size / base;
2520 return snprintf(str, str_size, "%.2f%s", fraction, suffix[num_divs]);
2524 * __strncpy_null - strncpy with null termination
2525 * @dest: the target array
2526 * @src: the source string
2527 * @n: maximum bytes to copy (size of *dest)
2529 * Like strncpy, but ensures destination is null-terminated.
2531 * Copies the string pointed to by src, including the terminating null
2532 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
2533 * of n bytes. Then ensure that dest is null-terminated.
2535 char *__strncpy_null(char *dest, const char *src, size_t n)
2537 strncpy(dest, src, n);
2544 * Checks to make sure that the label matches our requirements.
2546 0 if everything is safe and usable
2547 -1 if the label is too long
2549 static int check_label(const char *input)
2551 int len = strlen(input);
2553 if (len > BTRFS_LABEL_SIZE - 1) {
2554 error("label %s is too long (max %d)", input,
2555 BTRFS_LABEL_SIZE - 1);
2562 static int set_label_unmounted(const char *dev, const char *label)
2564 struct btrfs_trans_handle *trans;
2565 struct btrfs_root *root;
2568 ret = check_mounted(dev);
2570 error("checking mount status of %s failed: %d", dev, ret);
2574 error("device %s is mounted, use mount point", dev);
2578 /* Open the super_block at the default location
2579 * and as read-write.
2581 root = open_ctree(dev, 0, OPEN_CTREE_WRITES);
2582 if (!root) /* errors are printed by open_ctree() */
2585 trans = btrfs_start_transaction(root, 1);
2586 __strncpy_null(root->fs_info->super_copy->label, label, BTRFS_LABEL_SIZE - 1);
2588 btrfs_commit_transaction(trans, root);
2590 /* Now we close it since we are done. */
2595 static int set_label_mounted(const char *mount_path, const char *labelp)
2598 char label[BTRFS_LABEL_SIZE];
2600 fd = open(mount_path, O_RDONLY | O_NOATIME);
2602 error("unable to access %s: %s", mount_path, strerror(errno));
2606 memset(label, 0, sizeof(label));
2607 __strncpy_null(label, labelp, BTRFS_LABEL_SIZE - 1);
2608 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
2609 error("unable to set label of %s: %s", mount_path,
2619 int get_label_unmounted(const char *dev, char *label)
2621 struct btrfs_root *root;
2624 ret = check_mounted(dev);
2626 error("checking mount status of %s failed: %d", dev, ret);
2630 /* Open the super_block at the default location
2633 root = open_ctree(dev, 0, 0);
2637 __strncpy_null(label, root->fs_info->super_copy->label,
2638 BTRFS_LABEL_SIZE - 1);
2640 /* Now we close it since we are done. */
2646 * If a partition is mounted, try to get the filesystem label via its
2647 * mounted path rather than device. Return the corresponding error
2648 * the user specified the device path.
2650 int get_label_mounted(const char *mount_path, char *labelp)
2652 char label[BTRFS_LABEL_SIZE];
2656 fd = open(mount_path, O_RDONLY | O_NOATIME);
2658 error("unable to access %s: %s", mount_path, strerror(errno));
2662 memset(label, '\0', sizeof(label));
2663 ret = ioctl(fd, BTRFS_IOC_GET_FSLABEL, label);
2665 if (errno != ENOTTY)
2666 error("unable to get label of %s: %s", mount_path,
2673 __strncpy_null(labelp, label, BTRFS_LABEL_SIZE - 1);
2678 int get_label(const char *btrfs_dev, char *label)
2682 ret = is_existing_blk_or_reg_file(btrfs_dev);
2684 ret = get_label_mounted(btrfs_dev, label);
2686 ret = get_label_unmounted(btrfs_dev, label);
2691 int set_label(const char *btrfs_dev, const char *label)
2695 if (check_label(label))
2698 ret = is_existing_blk_or_reg_file(btrfs_dev);
2700 ret = set_label_mounted(btrfs_dev, label);
2702 ret = set_label_unmounted(btrfs_dev, label);
2708 * A not-so-good version fls64. No fascinating optimization since
2709 * no one except parse_size use it
2711 static int fls64(u64 x)
2715 for (i = 0; i <64; i++)
2716 if (x << i & (1ULL << 63))
2721 u64 parse_size(char *s)
2729 error("size value is empty");
2733 error("size value '%s' is less equal than 0", s);
2736 ret = strtoull(s, &endptr, 10);
2738 error("size value '%s' is invalid", s);
2741 if (endptr[0] && endptr[1]) {
2742 error("illegal suffix contains character '%c' in wrong position",
2747 * strtoll returns LLONG_MAX when overflow, if this happens,
2748 * need to call strtoull to get the real size
2750 if (errno == ERANGE && ret == ULLONG_MAX) {
2751 error("size value '%s' is too large for u64", s);
2755 c = tolower(endptr[0]);
2778 error("unknown size descriptor '%c'", c);
2782 /* Check whether ret * mult overflow */
2783 if (fls64(ret) + fls64(mult) - 1 > 64) {
2784 error("size value '%s' is too large for u64", s);
2791 u64 parse_qgroupid(const char *p)
2793 char *s = strchr(p, '/');
2794 const char *ptr_src_end = p + strlen(p);
2795 char *ptr_parse_end = NULL;
2804 /* Numeric format like '0/257' is the primary case */
2806 id = strtoull(p, &ptr_parse_end, 10);
2807 if (ptr_parse_end != ptr_src_end)
2811 level = strtoull(p, &ptr_parse_end, 10);
2812 if (ptr_parse_end != s)
2815 id = strtoull(s + 1, &ptr_parse_end, 10);
2816 if (ptr_parse_end != ptr_src_end)
2819 return (level << BTRFS_QGROUP_LEVEL_SHIFT) | id;
2822 /* Path format like subv at 'my_subvol' is the fallback case */
2823 ret = test_issubvolume(p);
2824 if (ret < 0 || !ret)
2826 fd = open(p, O_RDONLY);
2829 ret = lookup_ino_rootid(fd, &id);
2831 error("failed to lookup root id: %s", strerror(-ret));
2838 error("invalid qgroupid or subvolume path: %s", p);
2842 int open_file_or_dir3(const char *fname, DIR **dirstream, int open_flags)
2848 ret = stat(fname, &st);
2852 if (S_ISDIR(st.st_mode)) {
2853 *dirstream = opendir(fname);
2856 fd = dirfd(*dirstream);
2857 } else if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) {
2858 fd = open(fname, open_flags);
2861 * we set this on purpose, in case the caller output
2862 * strerror(errno) as success
2870 closedir(*dirstream);
2877 int open_file_or_dir(const char *fname, DIR **dirstream)
2879 return open_file_or_dir3(fname, dirstream, O_RDWR);
2882 void close_file_or_dir(int fd, DIR *dirstream)
2885 closedir(dirstream);
2890 int get_device_info(int fd, u64 devid,
2891 struct btrfs_ioctl_dev_info_args *di_args)
2895 di_args->devid = devid;
2896 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
2898 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
2899 return ret < 0 ? -errno : 0;
2902 static u64 find_max_device_id(struct btrfs_ioctl_search_args *search_args,
2905 struct btrfs_dev_item *dev_item;
2906 char *buf = search_args->buf;
2908 buf += (nr_items - 1) * (sizeof(struct btrfs_ioctl_search_header)
2909 + sizeof(struct btrfs_dev_item));
2910 buf += sizeof(struct btrfs_ioctl_search_header);
2912 dev_item = (struct btrfs_dev_item *)buf;
2914 return btrfs_stack_device_id(dev_item);
2917 static int search_chunk_tree_for_fs_info(int fd,
2918 struct btrfs_ioctl_fs_info_args *fi_args)
2922 u64 start_devid = 1;
2923 struct btrfs_ioctl_search_args search_args;
2924 struct btrfs_ioctl_search_key *search_key = &search_args.key;
2926 fi_args->num_devices = 0;
2928 max_items = BTRFS_SEARCH_ARGS_BUFSIZE
2929 / (sizeof(struct btrfs_ioctl_search_header)
2930 + sizeof(struct btrfs_dev_item));
2932 search_key->tree_id = BTRFS_CHUNK_TREE_OBJECTID;
2933 search_key->min_objectid = BTRFS_DEV_ITEMS_OBJECTID;
2934 search_key->max_objectid = BTRFS_DEV_ITEMS_OBJECTID;
2935 search_key->min_type = BTRFS_DEV_ITEM_KEY;
2936 search_key->max_type = BTRFS_DEV_ITEM_KEY;
2937 search_key->min_transid = 0;
2938 search_key->max_transid = (u64)-1;
2939 search_key->nr_items = max_items;
2940 search_key->max_offset = (u64)-1;
2943 search_key->min_offset = start_devid;
2945 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH, &search_args);
2949 fi_args->num_devices += (u64)search_key->nr_items;
2951 if (search_key->nr_items == max_items) {
2952 start_devid = find_max_device_id(&search_args,
2953 search_key->nr_items) + 1;
2957 /* get the lastest max_id to stay consistent with the num_devices */
2958 if (search_key->nr_items == 0)
2960 * last tree_search returns an empty buf, use the devid of
2961 * the last dev_item of the previous tree_search
2963 fi_args->max_id = start_devid - 1;
2965 fi_args->max_id = find_max_device_id(&search_args,
2966 search_key->nr_items);
2972 * For a given path, fill in the ioctl fs_ and info_ args.
2973 * If the path is a btrfs mountpoint, fill info for all devices.
2974 * If the path is a btrfs device, fill in only that device.
2976 * The path provided must be either on a mounted btrfs fs,
2977 * or be a mounted btrfs device.
2979 * Returns 0 on success, or a negative errno.
2981 int get_fs_info(char *path, struct btrfs_ioctl_fs_info_args *fi_args,
2982 struct btrfs_ioctl_dev_info_args **di_ret)
2989 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2990 struct btrfs_ioctl_dev_info_args *di_args;
2991 struct btrfs_ioctl_dev_info_args tmp;
2993 DIR *dirstream = NULL;
2995 memset(fi_args, 0, sizeof(*fi_args));
2997 if (is_block_device(path) == 1) {
2998 struct btrfs_super_block *disk_super;
2999 char buf[BTRFS_SUPER_INFO_SIZE];
3002 /* Ensure it's mounted, then set path to the mountpoint */
3003 fd = open(path, O_RDONLY);
3006 error("cannot open %s: %s", path, strerror(errno));
3009 ret = check_mounted_where(fd, path, mp, sizeof(mp),
3018 /* Only fill in this one device */
3019 fi_args->num_devices = 1;
3021 disk_super = (struct btrfs_super_block *)buf;
3022 ret = btrfs_read_dev_super(fd, disk_super,
3023 BTRFS_SUPER_INFO_OFFSET, 0);
3028 devid = btrfs_stack_device_id(&disk_super->dev_item);
3030 fi_args->max_id = devid;
3033 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
3037 /* at this point path must not be for a block device */
3038 fd = open_file_or_dir(path, &dirstream);
3044 /* fill in fi_args if not just a single device */
3045 if (fi_args->num_devices != 1) {
3046 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
3053 * The fs_args->num_devices does not include seed devices
3055 ret = search_chunk_tree_for_fs_info(fd, fi_args);
3060 * search_chunk_tree_for_fs_info() will lacks the devid 0
3061 * so manual probe for it here.
3063 ret = get_device_info(fd, 0, &tmp);
3065 fi_args->num_devices++;
3073 if (!fi_args->num_devices)
3076 di_args = *di_ret = malloc((fi_args->num_devices) * sizeof(*di_args));
3083 memcpy(di_args, &tmp, sizeof(tmp));
3084 for (; i <= fi_args->max_id; ++i) {
3085 ret = get_device_info(fd, i, &di_args[ndevs]);
3094 * only when the only dev we wanted to find is not there then
3095 * let any error be returned
3097 if (fi_args->num_devices != 1) {
3103 close_file_or_dir(fd, dirstream);
3107 #define isoctal(c) (((c) & ~7) == '0')
3109 static inline void translate(char *f, char *t)
3111 while (*f != '\0') {
3113 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
3114 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
3124 * Checks if the swap device.
3125 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
3127 static int is_swap_device(const char *file)
3138 if (stat(file, &st_buf) < 0)
3140 if (S_ISBLK(st_buf.st_mode))
3141 dev = st_buf.st_rdev;
3142 else if (S_ISREG(st_buf.st_mode)) {
3143 dev = st_buf.st_dev;
3144 ino = st_buf.st_ino;
3148 if ((f = fopen("/proc/swaps", "r")) == NULL)
3151 /* skip the first line */
3152 if (fgets(tmp, sizeof(tmp), f) == NULL)
3155 while (fgets(tmp, sizeof(tmp), f) != NULL) {
3156 if ((cp = strchr(tmp, ' ')) != NULL)
3158 if ((cp = strchr(tmp, '\t')) != NULL)
3160 translate(tmp, buf);
3161 if (stat(buf, &st_buf) != 0)
3163 if (S_ISBLK(st_buf.st_mode)) {
3164 if (dev == st_buf.st_rdev) {
3168 } else if (S_ISREG(st_buf.st_mode)) {
3169 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {
3183 * Check for existing filesystem or partition table on device.
3185 * 1 for existing fs or partition
3186 * 0 for nothing found
3187 * -1 for internal error
3189 static int check_overwrite(const char *device)
3192 blkid_probe pr = NULL;
3196 if (!device || !*device)
3199 ret = -1; /* will reset on success of all setup calls */
3201 pr = blkid_new_probe_from_filename(device);
3205 size = blkid_probe_get_size(pr);
3209 /* nothing to overwrite on a 0-length device */
3215 ret = blkid_probe_enable_partitions(pr, 1);
3219 ret = blkid_do_fullprobe(pr);
3224 * Blkid returns 1 for nothing found and 0 when it finds a signature,
3225 * but we want the exact opposite, so reverse the return value here.
3227 * In addition print some useful diagnostics about what actually is
3235 if (!blkid_probe_lookup_value(pr, "TYPE", &type, NULL)) {
3237 "%s appears to contain an existing "
3238 "filesystem (%s).\n", device, type);
3239 } else if (!blkid_probe_lookup_value(pr, "PTTYPE", &type, NULL)) {
3241 "%s appears to contain a partition "
3242 "table (%s).\n", device, type);
3245 "%s appears to contain something weird "
3246 "according to blkid\n", device);
3252 blkid_free_probe(pr);
3255 "probe of %s failed, cannot detect "
3256 "existing filesystem.\n", device);
3260 static int group_profile_devs_min(u64 flag)
3262 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3263 case 0: /* single */
3264 case BTRFS_BLOCK_GROUP_DUP:
3266 case BTRFS_BLOCK_GROUP_RAID0:
3267 case BTRFS_BLOCK_GROUP_RAID1:
3268 case BTRFS_BLOCK_GROUP_RAID5:
3270 case BTRFS_BLOCK_GROUP_RAID6:
3272 case BTRFS_BLOCK_GROUP_RAID10:
3279 int test_num_disk_vs_raid(u64 metadata_profile, u64 data_profile,
3280 u64 dev_cnt, int mixed, int ssd)
3287 allowed |= BTRFS_BLOCK_GROUP_RAID10;
3289 allowed |= BTRFS_BLOCK_GROUP_RAID6;
3291 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
3292 BTRFS_BLOCK_GROUP_RAID5;
3294 allowed |= BTRFS_BLOCK_GROUP_DUP;
3298 ((metadata_profile | data_profile) & BTRFS_BLOCK_GROUP_DUP)) {
3299 warning("DUP is not recommended on filesystem with multiple devices");
3301 if (metadata_profile & ~allowed) {
3303 "ERROR: unable to create FS with metadata profile %s "
3304 "(have %llu devices but %d devices are required)\n",
3305 btrfs_group_profile_str(metadata_profile), dev_cnt,
3306 group_profile_devs_min(metadata_profile));
3309 if (data_profile & ~allowed) {
3311 "ERROR: unable to create FS with data profile %s "
3312 "(have %llu devices but %d devices are required)\n",
3313 btrfs_group_profile_str(data_profile), dev_cnt,
3314 group_profile_devs_min(data_profile));
3318 warning_on(!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP) && ssd,
3319 "DUP may not actually lead to 2 copies on the device, see manual page");
3324 int group_profile_max_safe_loss(u64 flags)
3326 switch (flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3327 case 0: /* single */
3328 case BTRFS_BLOCK_GROUP_DUP:
3329 case BTRFS_BLOCK_GROUP_RAID0:
3331 case BTRFS_BLOCK_GROUP_RAID1:
3332 case BTRFS_BLOCK_GROUP_RAID5:
3333 case BTRFS_BLOCK_GROUP_RAID10:
3335 case BTRFS_BLOCK_GROUP_RAID6:
3343 * Check if a device is suitable for btrfs
3345 * 1: something is wrong, an error is printed
3348 int test_dev_for_mkfs(const char *file, int force_overwrite)
3353 ret = is_swap_device(file);
3355 error("checking status of %s: %s", file, strerror(-ret));
3359 error("%s is a swap device", file);
3362 if (!force_overwrite) {
3363 if (check_overwrite(file)) {
3364 error("use the -f option to force overwrite of %s",
3369 ret = check_mounted(file);
3371 error("cannot check mount status of %s: %s", file,
3376 error("%s is mounted", file);
3379 /* check if the device is busy */
3380 fd = open(file, O_RDWR|O_EXCL);
3382 error("unable to open %s: %s", file, strerror(errno));
3385 if (fstat(fd, &st)) {
3386 error("unable to stat %s: %s", file, strerror(errno));
3390 if (!S_ISBLK(st.st_mode)) {
3391 error("%s is not a block device", file);
3399 int btrfs_scan_lblkid(void)
3404 struct btrfs_fs_devices *tmp_devices;
3405 blkid_dev_iterate iter = NULL;
3406 blkid_dev dev = NULL;
3407 blkid_cache cache = NULL;
3408 char path[PATH_MAX];
3410 if (btrfs_scan_done)
3413 if (blkid_get_cache(&cache, NULL) < 0) {
3414 error("blkid cache get failed");
3417 blkid_probe_all(cache);
3418 iter = blkid_dev_iterate_begin(cache);
3419 blkid_dev_set_search(iter, "TYPE", "btrfs");
3420 while (blkid_dev_next(iter, &dev) == 0) {
3421 dev = blkid_verify(cache, dev);
3424 /* if we are here its definitely a btrfs disk*/
3425 strncpy_null(path, blkid_dev_devname(dev));
3427 fd = open(path, O_RDONLY);
3429 error("cannot open %s: %s", path, strerror(errno));
3432 ret = btrfs_scan_one_device(fd, path, &tmp_devices,
3433 &num_devices, BTRFS_SUPER_INFO_OFFSET,
3436 error("cannot scan %s: %s", path, strerror(-ret));
3443 blkid_dev_iterate_end(iter);
3444 blkid_put_cache(cache);
3446 btrfs_scan_done = 1;
3451 int is_vol_small(const char *file)
3458 fd = open(file, O_RDONLY);
3461 if (fstat(fd, &st) < 0) {
3466 size = btrfs_device_size(fd, &st);
3471 if (size < BTRFS_MKFS_SMALL_VOLUME_SIZE) {
3481 * This reads a line from the stdin and only returns non-zero if the
3482 * first whitespace delimited token is a case insensitive match with yes
3485 int ask_user(const char *question)
3487 char buf[30] = {0,};
3488 char *saveptr = NULL;
3491 printf("%s [y/N]: ", question);
3493 return fgets(buf, sizeof(buf) - 1, stdin) &&
3494 (answer = strtok_r(buf, " \t\n\r", &saveptr)) &&
3495 (!strcasecmp(answer, "yes") || !strcasecmp(answer, "y"));
3500 * - file or directory return the containing tree root id
3501 * - subvolume return its own tree id
3502 * - BTRFS_EMPTY_SUBVOL_DIR_OBJECTID (directory with ino == 2) the result is
3503 * undefined and function returns -1
3505 int lookup_ino_rootid(int fd, u64 *rootid)
3507 struct btrfs_ioctl_ino_lookup_args args;
3510 memset(&args, 0, sizeof(args));
3512 args.objectid = BTRFS_FIRST_FREE_OBJECTID;
3514 ret = ioctl(fd, BTRFS_IOC_INO_LOOKUP, &args);
3518 *rootid = args.treeid;
3524 * return 0 if a btrfs mount point is found
3525 * return 1 if a mount point is found but not btrfs
3526 * return <0 if something goes wrong
3528 int find_mount_root(const char *path, char **mount_root)
3536 int longest_matchlen = 0;
3537 char *longest_match = NULL;
3539 fd = open(path, O_RDONLY | O_NOATIME);
3544 mnttab = setmntent("/proc/self/mounts", "r");
3548 while ((ent = getmntent(mnttab))) {
3549 len = strlen(ent->mnt_dir);
3550 if (strncmp(ent->mnt_dir, path, len) == 0) {
3551 /* match found and use the latest match */
3552 if (longest_matchlen <= len) {
3553 free(longest_match);
3554 longest_matchlen = len;
3555 longest_match = strdup(ent->mnt_dir);
3556 not_btrfs = strcmp(ent->mnt_type, "btrfs");
3565 free(longest_match);
3570 *mount_root = realpath(longest_match, NULL);
3574 free(longest_match);
3578 int test_minimum_size(const char *file, u32 nodesize)
3581 struct stat statbuf;
3583 fd = open(file, O_RDONLY);
3586 if (stat(file, &statbuf) < 0) {
3590 if (btrfs_device_size(fd, &statbuf) < btrfs_min_dev_size(nodesize)) {
3600 * Test if path is a directory
3602 * 0 - path exists but it is not a directory
3603 * 1 - path exists and it is a directory
3606 int test_isdir(const char *path)
3611 ret = stat(path, &st);
3615 return !!S_ISDIR(st.st_mode);
3618 void units_set_mode(unsigned *units, unsigned mode)
3620 unsigned base = *units & UNITS_MODE_MASK;
3622 *units = base | mode;
3625 void units_set_base(unsigned *units, unsigned base)
3627 unsigned mode = *units & ~UNITS_MODE_MASK;
3629 *units = base | mode;
3632 int find_next_key(struct btrfs_path *path, struct btrfs_key *key)
3636 for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
3637 if (!path->nodes[level])
3639 if (path->slots[level] + 1 >=
3640 btrfs_header_nritems(path->nodes[level]))
3643 btrfs_item_key_to_cpu(path->nodes[level], key,
3644 path->slots[level] + 1);
3646 btrfs_node_key_to_cpu(path->nodes[level], key,
3647 path->slots[level] + 1);
3653 const char* btrfs_group_type_str(u64 flag)
3655 u64 mask = BTRFS_BLOCK_GROUP_TYPE_MASK |
3656 BTRFS_SPACE_INFO_GLOBAL_RSV;
3658 switch (flag & mask) {
3659 case BTRFS_BLOCK_GROUP_DATA:
3661 case BTRFS_BLOCK_GROUP_SYSTEM:
3663 case BTRFS_BLOCK_GROUP_METADATA:
3665 case BTRFS_BLOCK_GROUP_DATA|BTRFS_BLOCK_GROUP_METADATA:
3666 return "Data+Metadata";
3667 case BTRFS_SPACE_INFO_GLOBAL_RSV:
3668 return "GlobalReserve";
3674 const char* btrfs_group_profile_str(u64 flag)
3676 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3679 case BTRFS_BLOCK_GROUP_RAID0:
3681 case BTRFS_BLOCK_GROUP_RAID1:
3683 case BTRFS_BLOCK_GROUP_RAID5:
3685 case BTRFS_BLOCK_GROUP_RAID6:
3687 case BTRFS_BLOCK_GROUP_DUP:
3689 case BTRFS_BLOCK_GROUP_RAID10:
3696 u64 disk_size(const char *path)
3700 if (statfs(path, &sfs) < 0)
3703 return sfs.f_bsize * sfs.f_blocks;
3706 u64 get_partition_size(const char *dev)
3709 int fd = open(dev, O_RDONLY);
3713 if (ioctl(fd, BLKGETSIZE64, &result) < 0) {
3722 int btrfs_tree_search2_ioctl_supported(int fd)
3724 struct btrfs_ioctl_search_args_v2 *args2;
3725 struct btrfs_ioctl_search_key *sk;
3726 int args2_size = 1024;
3727 char args2_buf[args2_size];
3729 static int v2_supported = -1;
3731 if (v2_supported != -1)
3732 return v2_supported;
3734 args2 = (struct btrfs_ioctl_search_args_v2 *)args2_buf;
3738 * Search for the extent tree item in the root tree.
3740 sk->tree_id = BTRFS_ROOT_TREE_OBJECTID;
3741 sk->min_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3742 sk->max_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3743 sk->min_type = BTRFS_ROOT_ITEM_KEY;
3744 sk->max_type = BTRFS_ROOT_ITEM_KEY;
3746 sk->max_offset = (u64)-1;
3747 sk->min_transid = 0;
3748 sk->max_transid = (u64)-1;
3750 args2->buf_size = args2_size - sizeof(struct btrfs_ioctl_search_args_v2);
3751 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH_V2, args2);
3752 if (ret == -EOPNOTSUPP)
3759 return v2_supported;
3762 int btrfs_check_nodesize(u32 nodesize, u32 sectorsize, u64 features)
3764 if (nodesize < sectorsize) {
3765 error("illegal nodesize %u (smaller than %u)",
3766 nodesize, sectorsize);
3768 } else if (nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) {
3769 error("illegal nodesize %u (larger than %u)",
3770 nodesize, BTRFS_MAX_METADATA_BLOCKSIZE);
3772 } else if (nodesize & (sectorsize - 1)) {
3773 error("illegal nodesize %u (not aligned to %u)",
3774 nodesize, sectorsize);
3776 } else if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS &&
3777 nodesize != sectorsize) {
3778 error("illegal nodesize %u (not equal to %u for mixed block group)",
3779 nodesize, sectorsize);
3786 * Copy a path argument from SRC to DEST and check the SRC length if it's at
3787 * most PATH_MAX and fits into DEST. DESTLEN is supposed to be exact size of
3789 * The destination buffer is zero terminated.
3790 * Return < 0 for error, 0 otherwise.
3792 int arg_copy_path(char *dest, const char *src, int destlen)
3794 size_t len = strlen(src);
3796 if (len >= PATH_MAX || len >= destlen)
3797 return -ENAMETOOLONG;
3799 __strncpy_null(dest, src, destlen);
3804 unsigned int get_unit_mode_from_arg(int *argc, char *argv[], int df_mode)
3806 unsigned int unit_mode = UNITS_DEFAULT;
3810 for (arg_i = 0; arg_i < *argc; arg_i++) {
3811 if (!strcmp(argv[arg_i], "--"))
3814 if (!strcmp(argv[arg_i], "--raw")) {
3815 unit_mode = UNITS_RAW;
3819 if (!strcmp(argv[arg_i], "--human-readable")) {
3820 unit_mode = UNITS_HUMAN_BINARY;
3825 if (!strcmp(argv[arg_i], "--iec")) {
3826 units_set_mode(&unit_mode, UNITS_BINARY);
3830 if (!strcmp(argv[arg_i], "--si")) {
3831 units_set_mode(&unit_mode, UNITS_DECIMAL);
3836 if (!strcmp(argv[arg_i], "--kbytes")) {
3837 units_set_base(&unit_mode, UNITS_KBYTES);
3841 if (!strcmp(argv[arg_i], "--mbytes")) {
3842 units_set_base(&unit_mode, UNITS_MBYTES);
3846 if (!strcmp(argv[arg_i], "--gbytes")) {
3847 units_set_base(&unit_mode, UNITS_GBYTES);
3851 if (!strcmp(argv[arg_i], "--tbytes")) {
3852 units_set_base(&unit_mode, UNITS_TBYTES);
3860 if (!strcmp(argv[arg_i], "-b")) {
3861 unit_mode = UNITS_RAW;
3865 if (!strcmp(argv[arg_i], "-h")) {
3866 unit_mode = UNITS_HUMAN_BINARY;
3870 if (!strcmp(argv[arg_i], "-H")) {
3871 unit_mode = UNITS_HUMAN_DECIMAL;
3875 if (!strcmp(argv[arg_i], "-k")) {
3876 units_set_base(&unit_mode, UNITS_KBYTES);
3880 if (!strcmp(argv[arg_i], "-m")) {
3881 units_set_base(&unit_mode, UNITS_MBYTES);
3885 if (!strcmp(argv[arg_i], "-g")) {
3886 units_set_base(&unit_mode, UNITS_GBYTES);
3890 if (!strcmp(argv[arg_i], "-t")) {
3891 units_set_base(&unit_mode, UNITS_TBYTES);
3897 for (arg_i = 0, arg_end = 0; arg_i < *argc; arg_i++) {
3900 argv[arg_end] = argv[arg_i];
3909 int string_is_numerical(const char *str)
3911 if (!(*str >= '0' && *str <= '9'))
3913 while (*str >= '0' && *str <= '9')
3921 * Preprocess @argv with getopt_long to reorder options and consume the "--"
3923 * Unknown short and long options are reported, optionally the @usage is printed
3926 void clean_args_no_options(int argc, char *argv[], const char * const *usagestr)
3928 static const struct option long_options[] = {
3933 int c = getopt_long(argc, argv, "", long_options, NULL);
3947 * Same as clean_args_no_options but pass through arguments that could look
3948 * like short options. Eg. reisze which takes a negative resize argument like
3951 * This accepts only two forms:
3952 * - "-- option1 option2 ..."
3953 * - "option1 option2 ..."
3955 void clean_args_no_options_relaxed(int argc, char *argv[], const char * const *usagestr)
3960 if (strcmp(argv[1], "--") == 0)
3964 /* Subvolume helper functions */
3966 * test if name is a correct subvolume name
3967 * this function return
3968 * 0-> name is not a correct subvolume name
3969 * 1-> name is a correct subvolume name
3971 int test_issubvolname(const char *name)
3973 return name[0] != '\0' && !strchr(name, '/') &&
3974 strcmp(name, ".") && strcmp(name, "..");
3978 * Test if path is a subvolume
3980 * 0 - path exists but it is not a subvolume
3981 * 1 - path exists and it is a subvolume
3984 int test_issubvolume(const char *path)
3990 res = stat(path, &st);
3994 if (st.st_ino != BTRFS_FIRST_FREE_OBJECTID || !S_ISDIR(st.st_mode))
3997 res = statfs(path, &stfs);
4001 return (int)stfs.f_type == BTRFS_SUPER_MAGIC;
4004 const char *subvol_strip_mountpoint(const char *mnt, const char *full_path)
4006 int len = strlen(mnt);
4010 if (mnt[len - 1] != '/')
4013 return full_path + len;
4020 * 1: Error; and error info printed to the terminal. Fixme.
4021 * 2: If the fullpath is root tree instead of subvol tree
4023 int get_subvol_info(const char *fullpath, struct root_info *get_ri)
4030 const char *svpath = NULL;
4031 DIR *dirstream1 = NULL;
4032 DIR *dirstream2 = NULL;
4034 ret = test_issubvolume(fullpath);
4038 error("not a subvolume: %s", fullpath);
4042 ret = find_mount_root(fullpath, &mnt);
4046 error("%s doesn't belong to btrfs mount point", fullpath);
4050 svpath = subvol_strip_mountpoint(mnt, fullpath);
4052 fd = btrfs_open_dir(fullpath, &dirstream1, 1);
4056 ret = btrfs_list_get_path_rootid(fd, &sv_id);
4058 error("can't get rootid for '%s'", fullpath);
4062 mntfd = btrfs_open_dir(mnt, &dirstream2, 1);
4066 if (sv_id == BTRFS_FS_TREE_OBJECTID) {
4069 * So that caller may decide if thats an error or just fine.
4074 memset(get_ri, 0, sizeof(*get_ri));
4075 get_ri->root_id = sv_id;
4077 ret = btrfs_get_subvol(mntfd, get_ri);
4079 error("can't find '%s': %d", svpath, ret);
4082 close_file_or_dir(mntfd, dirstream2);
4083 close_file_or_dir(fd, dirstream1);
4089 void init_rand_seed(u64 seed)
4093 /* only use the last 48 bits */
4094 for (i = 0; i < 3; i++) {
4095 rand_seed[i] = (unsigned short)(seed ^ (unsigned short)(-1));
4098 rand_seed_initlized = 1;
4101 static void __init_seed(void)
4107 if(rand_seed_initlized)
4109 /* Use urandom as primary seed source. */
4110 fd = open("/dev/urandom", O_RDONLY);
4112 ret = read(fd, rand_seed, sizeof(rand_seed));
4114 if (ret < sizeof(rand_seed))
4118 /* Use time and pid as fallback seed */
4119 warning("failed to read /dev/urandom, use time and pid as random seed");
4120 gettimeofday(&tv, 0);
4121 rand_seed[0] = getpid() ^ (tv.tv_sec & 0xFFFF);
4122 rand_seed[1] = getppid() ^ (tv.tv_usec & 0xFFFF);
4123 rand_seed[2] = (tv.tv_sec ^ tv.tv_usec) >> 16;
4125 rand_seed_initlized = 1;
4132 * Don't use nrand48, its range is [0,2^31) The highest bit will alwasy
4133 * be 0. Use jrand48 to include the highest bit.
4135 return (u32)jrand48(rand_seed);
4138 unsigned int rand_range(unsigned int upper)
4142 * Use the full 48bits to mod, which would be more uniformly
4145 return (unsigned int)(jrand48(rand_seed) % upper);