2 * Copyright (C) 2007 Oracle. All rights reserved.
3 * Copyright (C) 2008 Morey Roof. All rights reserved.
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public
7 * License v2 as published by the Free Software Foundation.
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * General Public License for more details.
14 * You should have received a copy of the GNU General Public
15 * License along with this program; if not, write to the
16 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
17 * Boston, MA 021110-1307, USA.
23 #include <sys/ioctl.h>
24 #include <sys/mount.h>
25 #include <sys/types.h>
27 #include <uuid/uuid.h>
32 #include <linux/loop.h>
33 #include <linux/major.h>
34 #include <linux/kdev_t.h>
36 #include <blkid/blkid.h>
38 #include <sys/statfs.h>
39 #include <linux/magic.h>
41 #include <sys/utsname.h>
43 #include "kerncompat.h"
44 #include "radix-tree.h"
47 #include "transaction.h"
55 #define BLKDISCARD _IO(0x12,119)
58 static int btrfs_scan_done = 0;
60 static char argv0_buf[ARGV0_BUF_SIZE] = "btrfs";
62 static int rand_seed_initlized = 0;
63 static unsigned short rand_seed[3];
65 const char *get_argv0_buf(void)
70 void fixup_argv0(char **argv, const char *token)
72 int len = strlen(argv0_buf);
74 snprintf(argv0_buf + len, sizeof(argv0_buf) - len, " %s", token);
78 void set_argv0(char **argv)
80 strncpy(argv0_buf, argv[0], sizeof(argv0_buf));
81 argv0_buf[sizeof(argv0_buf) - 1] = 0;
84 int check_argc_exact(int nargs, int expected)
87 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
89 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
91 return nargs != expected;
94 int check_argc_min(int nargs, int expected)
96 if (nargs < expected) {
97 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
104 int check_argc_max(int nargs, int expected)
106 if (nargs > expected) {
107 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
116 * Discard the given range in one go
118 static int discard_range(int fd, u64 start, u64 len)
120 u64 range[2] = { start, len };
122 if (ioctl(fd, BLKDISCARD, &range) < 0)
128 * Discard blocks in the given range in 1G chunks, the process is interruptible
130 static int discard_blocks(int fd, u64 start, u64 len)
134 u64 chunk_size = min_t(u64, len, 1*1024*1024*1024);
137 ret = discard_range(fd, start, chunk_size);
147 static u64 reference_root_table[] = {
148 [1] = BTRFS_ROOT_TREE_OBJECTID,
149 [2] = BTRFS_EXTENT_TREE_OBJECTID,
150 [3] = BTRFS_CHUNK_TREE_OBJECTID,
151 [4] = BTRFS_DEV_TREE_OBJECTID,
152 [5] = BTRFS_FS_TREE_OBJECTID,
153 [6] = BTRFS_CSUM_TREE_OBJECTID,
156 int test_uuid_unique(char *fs_uuid)
159 blkid_dev_iterate iter = NULL;
160 blkid_dev dev = NULL;
161 blkid_cache cache = NULL;
163 if (blkid_get_cache(&cache, NULL) < 0) {
164 printf("ERROR: lblkid cache get failed\n");
167 blkid_probe_all(cache);
168 iter = blkid_dev_iterate_begin(cache);
169 blkid_dev_set_search(iter, "UUID", fs_uuid);
171 while (blkid_dev_next(iter, &dev) == 0) {
172 dev = blkid_verify(cache, dev);
179 blkid_dev_iterate_end(iter);
180 blkid_put_cache(cache);
186 * Reserve space from free_tree.
187 * The algorithm is very simple, find the first cache_extent with enough space
188 * and allocate from its beginning.
190 static int reserve_free_space(struct cache_tree *free_tree, u64 len,
193 struct cache_extent *cache;
196 ASSERT(ret_start != NULL);
197 cache = first_cache_extent(free_tree);
199 if (cache->size > len) {
201 *ret_start = cache->start;
204 if (cache->size == 0) {
205 remove_cache_extent(free_tree, cache);
212 cache = next_cache_extent(cache);
219 static inline int write_temp_super(int fd, struct btrfs_super_block *sb,
225 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
226 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
227 btrfs_csum_final(crc, &sb->csum[0]);
228 ret = pwrite(fd, sb, BTRFS_SUPER_INFO_SIZE, sb_bytenr);
229 if (ret < BTRFS_SUPER_INFO_SIZE)
230 ret = (ret < 0 ? -errno : -EIO);
237 * Setup temporary superblock at cfg->super_bynter
238 * Needed info are extracted from cfg, and root_bytenr, chunk_bytenr
240 * For now sys chunk array will be empty and dev_item is empty too.
241 * They will be re-initialized at temp chunk tree setup.
243 * The superblock signature is not valid, denotes a partially created
244 * filesystem, needs to be finalized.
246 static int setup_temp_super(int fd, struct btrfs_mkfs_config *cfg,
247 u64 root_bytenr, u64 chunk_bytenr)
249 unsigned char chunk_uuid[BTRFS_UUID_SIZE];
250 char super_buf[BTRFS_SUPER_INFO_SIZE];
251 struct btrfs_super_block *super = (struct btrfs_super_block *)super_buf;
254 memset(super_buf, 0, BTRFS_SUPER_INFO_SIZE);
255 cfg->num_bytes = round_down(cfg->num_bytes, cfg->sectorsize);
258 if (uuid_parse(cfg->fs_uuid, super->fsid) != 0) {
259 error("cound not parse UUID: %s", cfg->fs_uuid);
263 if (!test_uuid_unique(cfg->fs_uuid)) {
264 error("non-unique UUID: %s", cfg->fs_uuid);
269 uuid_generate(super->fsid);
270 uuid_unparse(super->fsid, cfg->fs_uuid);
272 uuid_generate(chunk_uuid);
273 uuid_unparse(chunk_uuid, cfg->chunk_uuid);
275 btrfs_set_super_bytenr(super, cfg->super_bytenr);
276 btrfs_set_super_num_devices(super, 1);
277 btrfs_set_super_magic(super, BTRFS_MAGIC_PARTIAL);
278 btrfs_set_super_generation(super, 1);
279 btrfs_set_super_root(super, root_bytenr);
280 btrfs_set_super_chunk_root(super, chunk_bytenr);
281 btrfs_set_super_total_bytes(super, cfg->num_bytes);
283 * Temporary filesystem will only have 6 tree roots:
284 * chunk tree, root tree, extent_tree, device tree, fs tree
287 btrfs_set_super_bytes_used(super, 6 * cfg->nodesize);
288 btrfs_set_super_sectorsize(super, cfg->sectorsize);
289 btrfs_set_super_leafsize(super, cfg->nodesize);
290 btrfs_set_super_nodesize(super, cfg->nodesize);
291 btrfs_set_super_stripesize(super, cfg->stripesize);
292 btrfs_set_super_csum_type(super, BTRFS_CSUM_TYPE_CRC32);
293 btrfs_set_super_chunk_root(super, chunk_bytenr);
294 btrfs_set_super_cache_generation(super, -1);
295 btrfs_set_super_incompat_flags(super, cfg->features);
297 __strncpy_null(super->label, cfg->label, BTRFS_LABEL_SIZE - 1);
299 /* Sys chunk array will be re-initialized at chunk tree init time */
300 super->sys_chunk_array_size = 0;
302 ret = write_temp_super(fd, super, cfg->super_bytenr);
308 * Setup an extent buffer for tree block.
310 static int setup_temp_extent_buffer(struct extent_buffer *buf,
311 struct btrfs_mkfs_config *cfg,
312 u64 bytenr, u64 owner)
314 unsigned char fsid[BTRFS_FSID_SIZE];
315 unsigned char chunk_uuid[BTRFS_UUID_SIZE];
318 ret = uuid_parse(cfg->fs_uuid, fsid);
321 ret = uuid_parse(cfg->chunk_uuid, chunk_uuid);
325 memset(buf->data, 0, cfg->nodesize);
326 buf->len = cfg->nodesize;
327 btrfs_set_header_bytenr(buf, bytenr);
328 btrfs_set_header_generation(buf, 1);
329 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
330 btrfs_set_header_owner(buf, owner);
331 btrfs_set_header_flags(buf, BTRFS_HEADER_FLAG_WRITTEN);
332 write_extent_buffer(buf, chunk_uuid, btrfs_header_chunk_tree_uuid(buf),
334 write_extent_buffer(buf, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
338 static inline int write_temp_extent_buffer(int fd, struct extent_buffer *buf,
343 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
345 /* Temporary extent buffer is always mapped 1:1 on disk */
346 ret = pwrite(fd, buf->data, buf->len, bytenr);
348 ret = (ret < 0 ? ret : -EIO);
355 * Insert a root item for temporary tree root
357 * Only used in make_btrfs_v2().
359 static void insert_temp_root_item(struct extent_buffer *buf,
360 struct btrfs_mkfs_config *cfg,
361 int *slot, u32 *itemoff, u64 objectid,
364 struct btrfs_root_item root_item;
365 struct btrfs_inode_item *inode_item;
366 struct btrfs_disk_key disk_key;
368 btrfs_set_header_nritems(buf, *slot + 1);
369 (*itemoff) -= sizeof(root_item);
370 memset(&root_item, 0, sizeof(root_item));
371 inode_item = &root_item.inode;
372 btrfs_set_stack_inode_generation(inode_item, 1);
373 btrfs_set_stack_inode_size(inode_item, 3);
374 btrfs_set_stack_inode_nlink(inode_item, 1);
375 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
376 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
377 btrfs_set_root_refs(&root_item, 1);
378 btrfs_set_root_used(&root_item, cfg->nodesize);
379 btrfs_set_root_generation(&root_item, 1);
380 btrfs_set_root_bytenr(&root_item, bytenr);
382 memset(&disk_key, 0, sizeof(disk_key));
383 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
384 btrfs_set_disk_key_objectid(&disk_key, objectid);
385 btrfs_set_disk_key_offset(&disk_key, 0);
387 btrfs_set_item_key(buf, &disk_key, *slot);
388 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
389 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(root_item));
390 write_extent_buffer(buf, &root_item,
391 btrfs_item_ptr_offset(buf, *slot),
396 static int setup_temp_root_tree(int fd, struct btrfs_mkfs_config *cfg,
397 u64 root_bytenr, u64 extent_bytenr,
398 u64 dev_bytenr, u64 fs_bytenr, u64 csum_bytenr)
400 struct extent_buffer *buf = NULL;
401 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
406 * Provided bytenr must in ascending order, or tree root will have a
409 if (!(root_bytenr < extent_bytenr && extent_bytenr < dev_bytenr &&
410 dev_bytenr < fs_bytenr && fs_bytenr < csum_bytenr)) {
411 error("bad tree bytenr order: "
412 "root < extent %llu < %llu, "
413 "extent < dev %llu < %llu, "
414 "dev < fs %llu < %llu, "
415 "fs < csum %llu < %llu",
416 (unsigned long long)root_bytenr,
417 (unsigned long long)extent_bytenr,
418 (unsigned long long)extent_bytenr,
419 (unsigned long long)dev_bytenr,
420 (unsigned long long)dev_bytenr,
421 (unsigned long long)fs_bytenr,
422 (unsigned long long)fs_bytenr,
423 (unsigned long long)csum_bytenr);
426 buf = malloc(sizeof(*buf) + cfg->nodesize);
430 ret = setup_temp_extent_buffer(buf, cfg, root_bytenr,
431 BTRFS_ROOT_TREE_OBJECTID);
435 insert_temp_root_item(buf, cfg, &slot, &itemoff,
436 BTRFS_EXTENT_TREE_OBJECTID, extent_bytenr);
437 insert_temp_root_item(buf, cfg, &slot, &itemoff,
438 BTRFS_DEV_TREE_OBJECTID, dev_bytenr);
439 insert_temp_root_item(buf, cfg, &slot, &itemoff,
440 BTRFS_FS_TREE_OBJECTID, fs_bytenr);
441 insert_temp_root_item(buf, cfg, &slot, &itemoff,
442 BTRFS_CSUM_TREE_OBJECTID, csum_bytenr);
444 ret = write_temp_extent_buffer(fd, buf, root_bytenr);
450 static int insert_temp_dev_item(int fd, struct extent_buffer *buf,
451 struct btrfs_mkfs_config *cfg,
452 int *slot, u32 *itemoff)
454 struct btrfs_disk_key disk_key;
455 struct btrfs_dev_item *dev_item;
456 char super_buf[BTRFS_SUPER_INFO_SIZE];
457 unsigned char dev_uuid[BTRFS_UUID_SIZE];
458 unsigned char fsid[BTRFS_FSID_SIZE];
459 struct btrfs_super_block *super = (struct btrfs_super_block *)super_buf;
462 ret = pread(fd, super_buf, BTRFS_SUPER_INFO_SIZE, cfg->super_bytenr);
463 if (ret < BTRFS_SUPER_INFO_SIZE) {
464 ret = (ret < 0 ? -errno : -EIO);
468 btrfs_set_header_nritems(buf, *slot + 1);
469 (*itemoff) -= sizeof(*dev_item);
470 /* setup device item 1, 0 is for replace case */
471 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
472 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
473 btrfs_set_disk_key_offset(&disk_key, 1);
474 btrfs_set_item_key(buf, &disk_key, *slot);
475 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
476 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(*dev_item));
478 dev_item = btrfs_item_ptr(buf, *slot, struct btrfs_dev_item);
479 /* Generate device uuid */
480 uuid_generate(dev_uuid);
481 write_extent_buffer(buf, dev_uuid,
482 (unsigned long)btrfs_device_uuid(dev_item),
484 uuid_parse(cfg->fs_uuid, fsid);
485 write_extent_buffer(buf, fsid,
486 (unsigned long)btrfs_device_fsid(dev_item),
488 btrfs_set_device_id(buf, dev_item, 1);
489 btrfs_set_device_generation(buf, dev_item, 0);
490 btrfs_set_device_total_bytes(buf, dev_item, cfg->num_bytes);
492 * The number must match the initial SYSTEM and META chunk size
494 btrfs_set_device_bytes_used(buf, dev_item,
495 BTRFS_MKFS_SYSTEM_GROUP_SIZE +
496 BTRFS_CONVERT_META_GROUP_SIZE);
497 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
498 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
499 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
500 btrfs_set_device_type(buf, dev_item, 0);
502 /* Super dev_item is not complete, copy the complete one to sb */
503 read_extent_buffer(buf, &super->dev_item, (unsigned long)dev_item,
505 ret = write_temp_super(fd, super, cfg->super_bytenr);
511 static int insert_temp_chunk_item(int fd, struct extent_buffer *buf,
512 struct btrfs_mkfs_config *cfg,
513 int *slot, u32 *itemoff, u64 start, u64 len,
516 struct btrfs_chunk *chunk;
517 struct btrfs_disk_key disk_key;
518 char super_buf[BTRFS_SUPER_INFO_SIZE];
519 struct btrfs_super_block *sb = (struct btrfs_super_block *)super_buf;
522 ret = pread(fd, super_buf, BTRFS_SUPER_INFO_SIZE,
524 if (ret < BTRFS_SUPER_INFO_SIZE) {
525 ret = (ret < 0 ? ret : -EIO);
529 btrfs_set_header_nritems(buf, *slot + 1);
530 (*itemoff) -= btrfs_chunk_item_size(1);
531 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
532 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
533 btrfs_set_disk_key_offset(&disk_key, start);
534 btrfs_set_item_key(buf, &disk_key, *slot);
535 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
536 btrfs_set_item_size(buf, btrfs_item_nr(*slot),
537 btrfs_chunk_item_size(1));
539 chunk = btrfs_item_ptr(buf, *slot, struct btrfs_chunk);
540 btrfs_set_chunk_length(buf, chunk, len);
541 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
542 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
543 btrfs_set_chunk_type(buf, chunk, type);
544 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
545 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
546 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
547 btrfs_set_chunk_num_stripes(buf, chunk, 1);
548 /* TODO: Support DUP profile for system chunk */
549 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
550 /* We are doing 1:1 mapping, so start is its dev offset */
551 btrfs_set_stripe_offset_nr(buf, chunk, 0, start);
552 write_extent_buffer(buf, &sb->dev_item.uuid,
553 (unsigned long)btrfs_stripe_dev_uuid_nr(chunk, 0),
558 * If it's system chunk, also copy it to super block.
560 if (type & BTRFS_BLOCK_GROUP_SYSTEM) {
563 cur = (char *)sb->sys_chunk_array + sb->sys_chunk_array_size;
564 memcpy(cur, &disk_key, sizeof(disk_key));
565 cur += sizeof(disk_key);
566 read_extent_buffer(buf, cur, (unsigned long int)chunk,
567 btrfs_chunk_item_size(1));
568 sb->sys_chunk_array_size += btrfs_chunk_item_size(1) +
571 ret = write_temp_super(fd, sb, cfg->super_bytenr);
576 static int setup_temp_chunk_tree(int fd, struct btrfs_mkfs_config *cfg,
577 u64 sys_chunk_start, u64 meta_chunk_start,
580 struct extent_buffer *buf = NULL;
581 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
585 /* Must ensure SYS chunk starts before META chunk */
586 if (meta_chunk_start < sys_chunk_start) {
587 error("wrong chunk order: meta < system %llu < %llu",
588 (unsigned long long)meta_chunk_start,
589 (unsigned long long)sys_chunk_start);
592 buf = malloc(sizeof(*buf) + cfg->nodesize);
595 ret = setup_temp_extent_buffer(buf, cfg, chunk_bytenr,
596 BTRFS_CHUNK_TREE_OBJECTID);
600 ret = insert_temp_dev_item(fd, buf, cfg, &slot, &itemoff);
603 ret = insert_temp_chunk_item(fd, buf, cfg, &slot, &itemoff,
605 BTRFS_MKFS_SYSTEM_GROUP_SIZE,
606 BTRFS_BLOCK_GROUP_SYSTEM);
609 ret = insert_temp_chunk_item(fd, buf, cfg, &slot, &itemoff,
611 BTRFS_CONVERT_META_GROUP_SIZE,
612 BTRFS_BLOCK_GROUP_METADATA);
615 ret = write_temp_extent_buffer(fd, buf, chunk_bytenr);
622 static void insert_temp_dev_extent(struct extent_buffer *buf,
623 int *slot, u32 *itemoff, u64 start, u64 len)
625 struct btrfs_dev_extent *dev_extent;
626 struct btrfs_disk_key disk_key;
628 btrfs_set_header_nritems(buf, *slot + 1);
629 (*itemoff) -= sizeof(*dev_extent);
630 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
631 btrfs_set_disk_key_objectid(&disk_key, 1);
632 btrfs_set_disk_key_offset(&disk_key, start);
633 btrfs_set_item_key(buf, &disk_key, *slot);
634 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
635 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(*dev_extent));
637 dev_extent = btrfs_item_ptr(buf, *slot, struct btrfs_dev_extent);
638 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
639 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
640 btrfs_set_dev_extent_length(buf, dev_extent, len);
641 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, start);
642 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
643 BTRFS_CHUNK_TREE_OBJECTID);
647 static int setup_temp_dev_tree(int fd, struct btrfs_mkfs_config *cfg,
648 u64 sys_chunk_start, u64 meta_chunk_start,
651 struct extent_buffer *buf = NULL;
652 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
656 /* Must ensure SYS chunk starts before META chunk */
657 if (meta_chunk_start < sys_chunk_start) {
658 error("wrong chunk order: meta < system %llu < %llu",
659 (unsigned long long)meta_chunk_start,
660 (unsigned long long)sys_chunk_start);
663 buf = malloc(sizeof(*buf) + cfg->nodesize);
666 ret = setup_temp_extent_buffer(buf, cfg, dev_bytenr,
667 BTRFS_DEV_TREE_OBJECTID);
670 insert_temp_dev_extent(buf, &slot, &itemoff, sys_chunk_start,
671 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
672 insert_temp_dev_extent(buf, &slot, &itemoff, meta_chunk_start,
673 BTRFS_CONVERT_META_GROUP_SIZE);
674 ret = write_temp_extent_buffer(fd, buf, dev_bytenr);
680 static int setup_temp_fs_tree(int fd, struct btrfs_mkfs_config *cfg,
683 struct extent_buffer *buf = NULL;
686 buf = malloc(sizeof(*buf) + cfg->nodesize);
689 ret = setup_temp_extent_buffer(buf, cfg, fs_bytenr,
690 BTRFS_FS_TREE_OBJECTID);
694 * Temporary fs tree is completely empty.
696 ret = write_temp_extent_buffer(fd, buf, fs_bytenr);
702 static int setup_temp_csum_tree(int fd, struct btrfs_mkfs_config *cfg,
705 struct extent_buffer *buf = NULL;
708 buf = malloc(sizeof(*buf) + cfg->nodesize);
711 ret = setup_temp_extent_buffer(buf, cfg, csum_bytenr,
712 BTRFS_CSUM_TREE_OBJECTID);
716 * Temporary csum tree is completely empty.
718 ret = write_temp_extent_buffer(fd, buf, csum_bytenr);
725 * Insert one temporary extent item.
727 * NOTE: if skinny_metadata is not enabled, this function must be called
728 * after all other trees are initialized.
729 * Or fs without skinny-metadata will be screwed up.
731 static int insert_temp_extent_item(int fd, struct extent_buffer *buf,
732 struct btrfs_mkfs_config *cfg,
733 int *slot, u32 *itemoff, u64 bytenr,
736 struct extent_buffer *tmp;
737 struct btrfs_extent_item *ei;
738 struct btrfs_extent_inline_ref *iref;
739 struct btrfs_disk_key disk_key;
740 struct btrfs_disk_key tree_info_key;
741 struct btrfs_tree_block_info *info;
743 int skinny_metadata = cfg->features &
744 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA;
748 itemsize = sizeof(*ei) + sizeof(*iref);
750 itemsize = sizeof(*ei) + sizeof(*iref) +
751 sizeof(struct btrfs_tree_block_info);
753 btrfs_set_header_nritems(buf, *slot + 1);
754 *(itemoff) -= itemsize;
756 if (skinny_metadata) {
757 btrfs_set_disk_key_type(&disk_key, BTRFS_METADATA_ITEM_KEY);
758 btrfs_set_disk_key_offset(&disk_key, 0);
760 btrfs_set_disk_key_type(&disk_key, BTRFS_EXTENT_ITEM_KEY);
761 btrfs_set_disk_key_offset(&disk_key, cfg->nodesize);
763 btrfs_set_disk_key_objectid(&disk_key, bytenr);
765 btrfs_set_item_key(buf, &disk_key, *slot);
766 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
767 btrfs_set_item_size(buf, btrfs_item_nr(*slot), itemsize);
769 ei = btrfs_item_ptr(buf, *slot, struct btrfs_extent_item);
770 btrfs_set_extent_refs(buf, ei, 1);
771 btrfs_set_extent_generation(buf, ei, 1);
772 btrfs_set_extent_flags(buf, ei, BTRFS_EXTENT_FLAG_TREE_BLOCK);
774 if (skinny_metadata) {
775 iref = (struct btrfs_extent_inline_ref *)(ei + 1);
777 info = (struct btrfs_tree_block_info *)(ei + 1);
778 iref = (struct btrfs_extent_inline_ref *)(info + 1);
780 btrfs_set_extent_inline_ref_type(buf, iref,
781 BTRFS_TREE_BLOCK_REF_KEY);
782 btrfs_set_extent_inline_ref_offset(buf, iref, ref_root);
789 * Lastly, check the tree block key by read the tree block
790 * Since we do 1:1 mapping for convert case, we can directly
791 * read the bytenr from disk
793 tmp = malloc(sizeof(*tmp) + cfg->nodesize);
796 ret = setup_temp_extent_buffer(tmp, cfg, bytenr, ref_root);
799 ret = pread(fd, tmp->data, cfg->nodesize, bytenr);
800 if (ret < cfg->nodesize) {
801 ret = (ret < 0 ? -errno : -EIO);
804 if (btrfs_header_nritems(tmp) == 0) {
805 btrfs_set_disk_key_type(&tree_info_key, 0);
806 btrfs_set_disk_key_objectid(&tree_info_key, 0);
807 btrfs_set_disk_key_offset(&tree_info_key, 0);
809 btrfs_item_key(tmp, &tree_info_key, 0);
811 btrfs_set_tree_block_key(buf, info, &tree_info_key);
818 static void insert_temp_block_group(struct extent_buffer *buf,
819 struct btrfs_mkfs_config *cfg,
820 int *slot, u32 *itemoff,
821 u64 bytenr, u64 len, u64 used, u64 flag)
823 struct btrfs_block_group_item bgi;
824 struct btrfs_disk_key disk_key;
826 btrfs_set_header_nritems(buf, *slot + 1);
827 (*itemoff) -= sizeof(bgi);
828 btrfs_set_disk_key_type(&disk_key, BTRFS_BLOCK_GROUP_ITEM_KEY);
829 btrfs_set_disk_key_objectid(&disk_key, bytenr);
830 btrfs_set_disk_key_offset(&disk_key, len);
831 btrfs_set_item_key(buf, &disk_key, *slot);
832 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
833 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(bgi));
835 btrfs_set_block_group_flags(&bgi, flag);
836 btrfs_set_block_group_used(&bgi, used);
837 btrfs_set_block_group_chunk_objectid(&bgi,
838 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
839 write_extent_buffer(buf, &bgi, btrfs_item_ptr_offset(buf, *slot),
844 static int setup_temp_extent_tree(int fd, struct btrfs_mkfs_config *cfg,
845 u64 chunk_bytenr, u64 root_bytenr,
846 u64 extent_bytenr, u64 dev_bytenr,
847 u64 fs_bytenr, u64 csum_bytenr)
849 struct extent_buffer *buf = NULL;
850 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
855 * We must ensure provided bytenr are in ascending order,
856 * or extent tree key order will be broken.
858 if (!(chunk_bytenr < root_bytenr && root_bytenr < extent_bytenr &&
859 extent_bytenr < dev_bytenr && dev_bytenr < fs_bytenr &&
860 fs_bytenr < csum_bytenr)) {
861 error("bad tree bytenr order: "
862 "chunk < root %llu < %llu, "
863 "root < extent %llu < %llu, "
864 "extent < dev %llu < %llu, "
865 "dev < fs %llu < %llu, "
866 "fs < csum %llu < %llu",
867 (unsigned long long)chunk_bytenr,
868 (unsigned long long)root_bytenr,
869 (unsigned long long)root_bytenr,
870 (unsigned long long)extent_bytenr,
871 (unsigned long long)extent_bytenr,
872 (unsigned long long)dev_bytenr,
873 (unsigned long long)dev_bytenr,
874 (unsigned long long)fs_bytenr,
875 (unsigned long long)fs_bytenr,
876 (unsigned long long)csum_bytenr);
879 buf = malloc(sizeof(*buf) + cfg->nodesize);
883 ret = setup_temp_extent_buffer(buf, cfg, extent_bytenr,
884 BTRFS_EXTENT_TREE_OBJECTID);
888 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
889 chunk_bytenr, BTRFS_CHUNK_TREE_OBJECTID);
893 insert_temp_block_group(buf, cfg, &slot, &itemoff, chunk_bytenr,
894 BTRFS_MKFS_SYSTEM_GROUP_SIZE, cfg->nodesize,
895 BTRFS_BLOCK_GROUP_SYSTEM);
897 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
898 root_bytenr, BTRFS_ROOT_TREE_OBJECTID);
902 /* 5 tree block used, root, extent, dev, fs and csum*/
903 insert_temp_block_group(buf, cfg, &slot, &itemoff, root_bytenr,
904 BTRFS_CONVERT_META_GROUP_SIZE, cfg->nodesize * 5,
905 BTRFS_BLOCK_GROUP_METADATA);
907 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
908 extent_bytenr, BTRFS_EXTENT_TREE_OBJECTID);
911 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
912 dev_bytenr, BTRFS_DEV_TREE_OBJECTID);
915 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
916 fs_bytenr, BTRFS_FS_TREE_OBJECTID);
919 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
920 csum_bytenr, BTRFS_CSUM_TREE_OBJECTID);
924 ret = write_temp_extent_buffer(fd, buf, extent_bytenr);
931 * Improved version of make_btrfs().
934 * 1) Do chunk allocation to avoid used data
935 * And after this function, extent type matches chunk type
936 * 2) Better structured code
937 * No super long hand written codes to initialized all tree blocks
938 * Split into small blocks and reuse codes.
939 * TODO: Reuse tree operation facilities by introducing new flags
941 static int make_convert_btrfs(int fd, struct btrfs_mkfs_config *cfg,
942 struct btrfs_convert_context *cctx)
944 struct cache_tree *free = &cctx->free;
945 struct cache_tree *used = &cctx->used;
947 u64 meta_chunk_start;
948 /* chunk tree bytenr, in system chunk */
950 /* metadata trees bytenr, in metadata chunk */
958 /* Shouldn't happen */
959 BUG_ON(cache_tree_empty(used));
962 * reserve space for temporary superblock first
963 * Here we allocate a little larger space, to keep later
964 * free space will be STRIPE_LEN aligned
966 ret = reserve_free_space(free, BTRFS_STRIPE_LEN,
972 * Then reserve system chunk space
973 * TODO: Change system group size depending on cctx->total_bytes.
974 * If using current 4M, it can only handle less than one TB for
975 * worst case and then run out of sys space.
977 ret = reserve_free_space(free, BTRFS_MKFS_SYSTEM_GROUP_SIZE,
981 ret = reserve_free_space(free, BTRFS_CONVERT_META_GROUP_SIZE,
987 * Allocated meta/sys chunks will be mapped 1:1 with device offset.
989 * Inside the allocated metadata chunk, the layout will be:
990 * | offset | contents |
991 * -------------------------------------
993 * | +nodesize | extent root |
994 * | +nodesize * 2 | device root |
995 * | +nodesize * 3 | fs tree |
996 * | +nodesize * 4 | csum tree |
997 * -------------------------------------
998 * Inside the allocated system chunk, the layout will be:
999 * | offset | contents |
1000 * -------------------------------------
1001 * | +0 | chunk root |
1002 * -------------------------------------
1004 chunk_bytenr = sys_chunk_start;
1005 root_bytenr = meta_chunk_start;
1006 extent_bytenr = meta_chunk_start + cfg->nodesize;
1007 dev_bytenr = meta_chunk_start + cfg->nodesize * 2;
1008 fs_bytenr = meta_chunk_start + cfg->nodesize * 3;
1009 csum_bytenr = meta_chunk_start + cfg->nodesize * 4;
1011 ret = setup_temp_super(fd, cfg, root_bytenr, chunk_bytenr);
1015 ret = setup_temp_root_tree(fd, cfg, root_bytenr, extent_bytenr,
1016 dev_bytenr, fs_bytenr, csum_bytenr);
1019 ret = setup_temp_chunk_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
1023 ret = setup_temp_dev_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
1027 ret = setup_temp_fs_tree(fd, cfg, fs_bytenr);
1030 ret = setup_temp_csum_tree(fd, cfg, csum_bytenr);
1034 * Setup extent tree last, since it may need to read tree block key
1035 * for non-skinny metadata case.
1037 ret = setup_temp_extent_tree(fd, cfg, chunk_bytenr, root_bytenr,
1038 extent_bytenr, dev_bytenr, fs_bytenr,
1045 * @fs_uuid - if NULL, generates a UUID, returns back the new filesystem UUID
1047 * The superblock signature is not valid, denotes a partially created
1048 * filesystem, needs to be finalized.
1050 int make_btrfs(int fd, struct btrfs_mkfs_config *cfg,
1051 struct btrfs_convert_context *cctx)
1053 struct btrfs_super_block super;
1054 struct extent_buffer *buf;
1055 struct btrfs_root_item root_item;
1056 struct btrfs_disk_key disk_key;
1057 struct btrfs_extent_item *extent_item;
1058 struct btrfs_inode_item *inode_item;
1059 struct btrfs_chunk *chunk;
1060 struct btrfs_dev_item *dev_item;
1061 struct btrfs_dev_extent *dev_extent;
1062 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
1072 int skinny_metadata = !!(cfg->features &
1073 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA);
1077 return make_convert_btrfs(fd, cfg, cctx);
1078 buf = malloc(sizeof(*buf) + max(cfg->sectorsize, cfg->nodesize));
1082 first_free = BTRFS_SUPER_INFO_OFFSET + cfg->sectorsize * 2 - 1;
1083 first_free &= ~((u64)cfg->sectorsize - 1);
1085 memset(&super, 0, sizeof(super));
1087 num_bytes = (cfg->num_bytes / cfg->sectorsize) * cfg->sectorsize;
1088 if (*cfg->fs_uuid) {
1089 if (uuid_parse(cfg->fs_uuid, super.fsid) != 0) {
1090 error("cannot not parse UUID: %s", cfg->fs_uuid);
1094 if (!test_uuid_unique(cfg->fs_uuid)) {
1095 error("non-unique UUID: %s", cfg->fs_uuid);
1100 uuid_generate(super.fsid);
1102 uuid_unparse(super.fsid, cfg->fs_uuid);
1104 uuid_generate(super.dev_item.uuid);
1105 uuid_generate(chunk_tree_uuid);
1107 btrfs_set_super_bytenr(&super, cfg->blocks[0]);
1108 btrfs_set_super_num_devices(&super, 1);
1109 btrfs_set_super_magic(&super, BTRFS_MAGIC_PARTIAL);
1110 btrfs_set_super_generation(&super, 1);
1111 btrfs_set_super_root(&super, cfg->blocks[1]);
1112 btrfs_set_super_chunk_root(&super, cfg->blocks[3]);
1113 btrfs_set_super_total_bytes(&super, num_bytes);
1114 btrfs_set_super_bytes_used(&super, 6 * cfg->nodesize);
1115 btrfs_set_super_sectorsize(&super, cfg->sectorsize);
1116 btrfs_set_super_leafsize(&super, cfg->nodesize);
1117 btrfs_set_super_nodesize(&super, cfg->nodesize);
1118 btrfs_set_super_stripesize(&super, cfg->stripesize);
1119 btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
1120 btrfs_set_super_chunk_root_generation(&super, 1);
1121 btrfs_set_super_cache_generation(&super, -1);
1122 btrfs_set_super_incompat_flags(&super, cfg->features);
1124 __strncpy_null(super.label, cfg->label, BTRFS_LABEL_SIZE - 1);
1126 /* create the tree of root objects */
1127 memset(buf->data, 0, cfg->nodesize);
1128 buf->len = cfg->nodesize;
1129 btrfs_set_header_bytenr(buf, cfg->blocks[1]);
1130 btrfs_set_header_nritems(buf, 4);
1131 btrfs_set_header_generation(buf, 1);
1132 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
1133 btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
1134 write_extent_buffer(buf, super.fsid, btrfs_header_fsid(),
1137 write_extent_buffer(buf, chunk_tree_uuid,
1138 btrfs_header_chunk_tree_uuid(buf),
1141 /* create the items for the root tree */
1142 memset(&root_item, 0, sizeof(root_item));
1143 inode_item = &root_item.inode;
1144 btrfs_set_stack_inode_generation(inode_item, 1);
1145 btrfs_set_stack_inode_size(inode_item, 3);
1146 btrfs_set_stack_inode_nlink(inode_item, 1);
1147 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
1148 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
1149 btrfs_set_root_refs(&root_item, 1);
1150 btrfs_set_root_used(&root_item, cfg->nodesize);
1151 btrfs_set_root_generation(&root_item, 1);
1153 memset(&disk_key, 0, sizeof(disk_key));
1154 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
1155 btrfs_set_disk_key_offset(&disk_key, 0);
1158 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - sizeof(root_item);
1159 btrfs_set_root_bytenr(&root_item, cfg->blocks[2]);
1160 btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
1161 btrfs_set_item_key(buf, &disk_key, nritems);
1162 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1163 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1165 write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
1166 nritems), sizeof(root_item));
1169 itemoff = itemoff - sizeof(root_item);
1170 btrfs_set_root_bytenr(&root_item, cfg->blocks[4]);
1171 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
1172 btrfs_set_item_key(buf, &disk_key, nritems);
1173 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1174 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1176 write_extent_buffer(buf, &root_item,
1177 btrfs_item_ptr_offset(buf, nritems),
1181 itemoff = itemoff - sizeof(root_item);
1182 btrfs_set_root_bytenr(&root_item, cfg->blocks[5]);
1183 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
1184 btrfs_set_item_key(buf, &disk_key, nritems);
1185 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1186 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1188 write_extent_buffer(buf, &root_item,
1189 btrfs_item_ptr_offset(buf, nritems),
1193 itemoff = itemoff - sizeof(root_item);
1194 btrfs_set_root_bytenr(&root_item, cfg->blocks[6]);
1195 btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_TREE_OBJECTID);
1196 btrfs_set_item_key(buf, &disk_key, nritems);
1197 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1198 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1200 write_extent_buffer(buf, &root_item,
1201 btrfs_item_ptr_offset(buf, nritems),
1206 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1207 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[1]);
1208 if (ret != cfg->nodesize) {
1209 ret = (ret < 0 ? -errno : -EIO);
1213 /* create the items for the extent tree */
1214 memset(buf->data + sizeof(struct btrfs_header), 0,
1215 cfg->nodesize - sizeof(struct btrfs_header));
1217 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
1218 for (i = 1; i < 7; i++) {
1219 item_size = sizeof(struct btrfs_extent_item);
1220 if (!skinny_metadata)
1221 item_size += sizeof(struct btrfs_tree_block_info);
1223 if (cfg->blocks[i] < first_free) {
1224 error("block[%d] below first free: %llu < %llu",
1225 i, (unsigned long long)cfg->blocks[i],
1226 (unsigned long long)first_free);
1230 if (cfg->blocks[i] < cfg->blocks[i - 1]) {
1231 error("blocks %d and %d in reverse order: %llu < %llu",
1233 (unsigned long long)cfg->blocks[i],
1234 (unsigned long long)cfg->blocks[i - 1]);
1239 /* create extent item */
1240 itemoff -= item_size;
1241 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
1242 if (skinny_metadata) {
1243 btrfs_set_disk_key_type(&disk_key,
1244 BTRFS_METADATA_ITEM_KEY);
1245 btrfs_set_disk_key_offset(&disk_key, 0);
1247 btrfs_set_disk_key_type(&disk_key,
1248 BTRFS_EXTENT_ITEM_KEY);
1249 btrfs_set_disk_key_offset(&disk_key, cfg->nodesize);
1251 btrfs_set_item_key(buf, &disk_key, nritems);
1252 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
1254 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1256 extent_item = btrfs_item_ptr(buf, nritems,
1257 struct btrfs_extent_item);
1258 btrfs_set_extent_refs(buf, extent_item, 1);
1259 btrfs_set_extent_generation(buf, extent_item, 1);
1260 btrfs_set_extent_flags(buf, extent_item,
1261 BTRFS_EXTENT_FLAG_TREE_BLOCK);
1264 /* create extent ref */
1265 ref_root = reference_root_table[i];
1266 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
1267 btrfs_set_disk_key_offset(&disk_key, ref_root);
1268 btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
1269 btrfs_set_item_key(buf, &disk_key, nritems);
1270 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
1272 btrfs_set_item_size(buf, btrfs_item_nr(nritems), 0);
1275 btrfs_set_header_bytenr(buf, cfg->blocks[2]);
1276 btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
1277 btrfs_set_header_nritems(buf, nritems);
1278 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1279 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[2]);
1280 if (ret != cfg->nodesize) {
1281 ret = (ret < 0 ? -errno : -EIO);
1285 /* create the chunk tree */
1286 memset(buf->data + sizeof(struct btrfs_header), 0,
1287 cfg->nodesize - sizeof(struct btrfs_header));
1289 item_size = sizeof(*dev_item);
1290 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - item_size;
1292 /* first device 1 (there is no device 0) */
1293 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
1294 btrfs_set_disk_key_offset(&disk_key, 1);
1295 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
1296 btrfs_set_item_key(buf, &disk_key, nritems);
1297 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1298 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
1300 dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
1301 btrfs_set_device_id(buf, dev_item, 1);
1302 btrfs_set_device_generation(buf, dev_item, 0);
1303 btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
1304 btrfs_set_device_bytes_used(buf, dev_item,
1305 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1306 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
1307 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
1308 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
1309 btrfs_set_device_type(buf, dev_item, 0);
1311 write_extent_buffer(buf, super.dev_item.uuid,
1312 (unsigned long)btrfs_device_uuid(dev_item),
1314 write_extent_buffer(buf, super.fsid,
1315 (unsigned long)btrfs_device_fsid(dev_item),
1317 read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
1321 item_size = btrfs_chunk_item_size(1);
1322 itemoff = itemoff - item_size;
1324 /* then we have chunk 0 */
1325 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1326 btrfs_set_disk_key_offset(&disk_key, 0);
1327 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
1328 btrfs_set_item_key(buf, &disk_key, nritems);
1329 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1330 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
1332 chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
1333 btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1334 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
1335 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
1336 btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
1337 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
1338 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
1339 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
1340 btrfs_set_chunk_num_stripes(buf, chunk, 1);
1341 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
1342 btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
1345 write_extent_buffer(buf, super.dev_item.uuid,
1346 (unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
1349 /* copy the key for the chunk to the system array */
1350 ptr = super.sys_chunk_array;
1351 array_size = sizeof(disk_key);
1353 memcpy(ptr, &disk_key, sizeof(disk_key));
1354 ptr += sizeof(disk_key);
1356 /* copy the chunk to the system array */
1357 read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
1358 array_size += item_size;
1360 btrfs_set_super_sys_array_size(&super, array_size);
1362 btrfs_set_header_bytenr(buf, cfg->blocks[3]);
1363 btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
1364 btrfs_set_header_nritems(buf, nritems);
1365 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1366 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[3]);
1367 if (ret != cfg->nodesize) {
1368 ret = (ret < 0 ? -errno : -EIO);
1372 /* create the device tree */
1373 memset(buf->data + sizeof(struct btrfs_header), 0,
1374 cfg->nodesize - sizeof(struct btrfs_header));
1376 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) -
1377 sizeof(struct btrfs_dev_extent);
1379 btrfs_set_disk_key_objectid(&disk_key, 1);
1380 btrfs_set_disk_key_offset(&disk_key, 0);
1381 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
1382 btrfs_set_item_key(buf, &disk_key, nritems);
1383 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1384 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1385 sizeof(struct btrfs_dev_extent));
1386 dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
1387 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
1388 BTRFS_CHUNK_TREE_OBJECTID);
1389 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
1390 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1391 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
1393 write_extent_buffer(buf, chunk_tree_uuid,
1394 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
1397 btrfs_set_dev_extent_length(buf, dev_extent,
1398 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1401 btrfs_set_header_bytenr(buf, cfg->blocks[4]);
1402 btrfs_set_header_owner(buf, BTRFS_DEV_TREE_OBJECTID);
1403 btrfs_set_header_nritems(buf, nritems);
1404 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1405 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[4]);
1406 if (ret != cfg->nodesize) {
1407 ret = (ret < 0 ? -errno : -EIO);
1411 /* create the FS root */
1412 memset(buf->data + sizeof(struct btrfs_header), 0,
1413 cfg->nodesize - sizeof(struct btrfs_header));
1414 btrfs_set_header_bytenr(buf, cfg->blocks[5]);
1415 btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
1416 btrfs_set_header_nritems(buf, 0);
1417 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1418 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[5]);
1419 if (ret != cfg->nodesize) {
1420 ret = (ret < 0 ? -errno : -EIO);
1423 /* finally create the csum root */
1424 memset(buf->data + sizeof(struct btrfs_header), 0,
1425 cfg->nodesize - sizeof(struct btrfs_header));
1426 btrfs_set_header_bytenr(buf, cfg->blocks[6]);
1427 btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
1428 btrfs_set_header_nritems(buf, 0);
1429 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1430 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[6]);
1431 if (ret != cfg->nodesize) {
1432 ret = (ret < 0 ? -errno : -EIO);
1436 /* and write out the super block */
1437 memset(buf->data, 0, BTRFS_SUPER_INFO_SIZE);
1438 memcpy(buf->data, &super, sizeof(super));
1439 buf->len = BTRFS_SUPER_INFO_SIZE;
1440 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1441 ret = pwrite(fd, buf->data, BTRFS_SUPER_INFO_SIZE, cfg->blocks[0]);
1442 if (ret != BTRFS_SUPER_INFO_SIZE) {
1443 ret = (ret < 0 ? -errno : -EIO);
1454 static const struct btrfs_fs_feature {
1458 } mkfs_features[] = {
1459 { "mixed-bg", BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS,
1460 "mixed data and metadata block groups" },
1461 { "extref", BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF,
1462 "increased hardlink limit per file to 65536" },
1463 { "raid56", BTRFS_FEATURE_INCOMPAT_RAID56,
1464 "raid56 extended format" },
1465 { "skinny-metadata", BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA,
1466 "reduced-size metadata extent refs" },
1467 { "no-holes", BTRFS_FEATURE_INCOMPAT_NO_HOLES,
1468 "no explicit hole extents for files" },
1469 /* Keep this one last */
1470 { "list-all", BTRFS_FEATURE_LIST_ALL, NULL }
1473 static int parse_one_fs_feature(const char *name, u64 *flags)
1478 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1479 if (name[0] == '^' &&
1480 !strcmp(mkfs_features[i].name, name + 1)) {
1481 *flags &= ~ mkfs_features[i].flag;
1483 } else if (!strcmp(mkfs_features[i].name, name)) {
1484 *flags |= mkfs_features[i].flag;
1492 void btrfs_parse_features_to_string(char *buf, u64 flags)
1498 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1499 if (flags & mkfs_features[i].flag) {
1502 strcat(buf, mkfs_features[i].name);
1507 void btrfs_process_fs_features(u64 flags)
1511 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1512 if (flags & mkfs_features[i].flag) {
1513 printf("Turning ON incompat feature '%s': %s\n",
1514 mkfs_features[i].name,
1515 mkfs_features[i].desc);
1520 void btrfs_list_all_fs_features(u64 mask_disallowed)
1524 fprintf(stderr, "Filesystem features available:\n");
1525 for (i = 0; i < ARRAY_SIZE(mkfs_features) - 1; i++) {
1526 char *is_default = "";
1528 if (mkfs_features[i].flag & mask_disallowed)
1530 if (mkfs_features[i].flag & BTRFS_MKFS_DEFAULT_FEATURES)
1531 is_default = ", default";
1532 fprintf(stderr, "%-20s- %s (0x%llx%s)\n",
1533 mkfs_features[i].name,
1534 mkfs_features[i].desc,
1535 mkfs_features[i].flag,
1541 * Return NULL if all features were parsed fine, otherwise return the name of
1542 * the first unparsed.
1544 char* btrfs_parse_fs_features(char *namelist, u64 *flags)
1547 char *save_ptr = NULL; /* Satisfy static checkers */
1549 for (this_char = strtok_r(namelist, ",", &save_ptr);
1551 this_char = strtok_r(NULL, ",", &save_ptr)) {
1552 if (parse_one_fs_feature(this_char, flags))
1559 void print_kernel_version(FILE *stream, u32 version)
1563 v[0] = version & 0xFF;
1564 v[1] = (version >> 8) & 0xFF;
1565 v[2] = version >> 16;
1566 fprintf(stream, "%u.%u", v[2], v[1]);
1568 fprintf(stream, ".%u", v[0]);
1571 u32 get_running_kernel_version(void)
1573 struct utsname utsbuf;
1575 char *saveptr = NULL;
1579 if (strcmp(utsbuf.sysname, "Linux") != 0) {
1580 error("unsupported system: %s", utsbuf.sysname);
1584 tmp = strchr(utsbuf.release, '-');
1588 tmp = strtok_r(utsbuf.release, ".", &saveptr);
1589 if (!string_is_numerical(tmp))
1591 version = atoi(tmp) << 16;
1592 tmp = strtok_r(NULL, ".", &saveptr);
1593 if (!string_is_numerical(tmp))
1595 version |= atoi(tmp) << 8;
1596 tmp = strtok_r(NULL, ".", &saveptr);
1598 if (!string_is_numerical(tmp))
1600 version |= atoi(tmp);
1606 u64 btrfs_device_size(int fd, struct stat *st)
1609 if (S_ISREG(st->st_mode)) {
1612 if (!S_ISBLK(st->st_mode)) {
1615 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
1621 static int zero_blocks(int fd, off_t start, size_t len)
1623 char *buf = malloc(len);
1629 memset(buf, 0, len);
1630 written = pwrite(fd, buf, len, start);
1637 #define ZERO_DEV_BYTES (2 * 1024 * 1024)
1639 /* don't write outside the device by clamping the region to the device size */
1640 static int zero_dev_clamped(int fd, off_t start, ssize_t len, u64 dev_size)
1642 off_t end = max(start, start + len);
1645 /* and don't overwrite the disk labels on sparc */
1646 start = max(start, 1024);
1647 end = max(end, 1024);
1650 start = min_t(u64, start, dev_size);
1651 end = min_t(u64, end, dev_size);
1653 return zero_blocks(fd, start, end - start);
1656 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
1657 struct btrfs_root *root, int fd, const char *path,
1658 u64 device_total_bytes, u32 io_width, u32 io_align,
1661 struct btrfs_super_block *disk_super;
1662 struct btrfs_super_block *super = root->fs_info->super_copy;
1663 struct btrfs_device *device;
1664 struct btrfs_dev_item *dev_item;
1670 device_total_bytes = (device_total_bytes / sectorsize) * sectorsize;
1672 device = calloc(1, sizeof(*device));
1677 buf = calloc(1, sectorsize);
1683 disk_super = (struct btrfs_super_block *)buf;
1684 dev_item = &disk_super->dev_item;
1686 uuid_generate(device->uuid);
1689 device->io_width = io_width;
1690 device->io_align = io_align;
1691 device->sector_size = sectorsize;
1693 device->writeable = 1;
1694 device->total_bytes = device_total_bytes;
1695 device->bytes_used = 0;
1696 device->total_ios = 0;
1697 device->dev_root = root->fs_info->dev_root;
1698 device->name = strdup(path);
1699 if (!device->name) {
1704 INIT_LIST_HEAD(&device->dev_list);
1705 ret = btrfs_add_device(trans, root, device);
1709 fs_total_bytes = btrfs_super_total_bytes(super) + device_total_bytes;
1710 btrfs_set_super_total_bytes(super, fs_total_bytes);
1712 num_devs = btrfs_super_num_devices(super) + 1;
1713 btrfs_set_super_num_devices(super, num_devs);
1715 memcpy(disk_super, super, sizeof(*disk_super));
1717 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
1718 btrfs_set_stack_device_id(dev_item, device->devid);
1719 btrfs_set_stack_device_type(dev_item, device->type);
1720 btrfs_set_stack_device_io_align(dev_item, device->io_align);
1721 btrfs_set_stack_device_io_width(dev_item, device->io_width);
1722 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
1723 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
1724 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
1725 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
1727 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
1728 BUG_ON(ret != sectorsize);
1731 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
1732 device->fs_devices = root->fs_info->fs_devices;
1741 static int btrfs_wipe_existing_sb(int fd)
1743 const char *off = NULL;
1748 blkid_probe pr = NULL;
1750 pr = blkid_new_probe();
1754 if (blkid_probe_set_device(pr, fd, 0, 0)) {
1759 ret = blkid_probe_lookup_value(pr, "SBMAGIC_OFFSET", &off, NULL);
1761 ret = blkid_probe_lookup_value(pr, "SBMAGIC", NULL, &len);
1763 if (ret || len == 0 || off == NULL) {
1765 * If lookup fails, the probe did not find any values, eg. for
1766 * a file image or a loop device. Soft error.
1772 offset = strtoll(off, NULL, 10);
1773 if (len > sizeof(buf))
1776 memset(buf, 0, len);
1777 ret = pwrite(fd, buf, len, offset);
1779 error("cannot wipe existing superblock: %s", strerror(errno));
1781 } else if (ret != len) {
1782 error("cannot wipe existing superblock: wrote %d of %zd", ret, len);
1788 blkid_free_probe(pr);
1792 int btrfs_prepare_device(int fd, const char *file, u64 *block_count_ret,
1793 u64 max_block_count, unsigned opflags)
1799 ret = fstat(fd, &st);
1801 error("unable to stat %s: %s", file, strerror(errno));
1805 block_count = btrfs_device_size(fd, &st);
1806 if (block_count == 0) {
1807 error("unable to determine size of %s", file);
1810 if (max_block_count)
1811 block_count = min(block_count, max_block_count);
1813 if (opflags & PREP_DEVICE_DISCARD) {
1815 * We intentionally ignore errors from the discard ioctl. It
1816 * is not necessary for the mkfs functionality but just an
1819 if (discard_range(fd, 0, 0) == 0) {
1820 if (opflags & PREP_DEVICE_VERBOSE)
1821 printf("Performing full device TRIM (%s) ...\n",
1822 pretty_size(block_count));
1823 discard_blocks(fd, 0, block_count);
1827 ret = zero_dev_clamped(fd, 0, ZERO_DEV_BYTES, block_count);
1828 for (i = 0 ; !ret && i < BTRFS_SUPER_MIRROR_MAX; i++)
1829 ret = zero_dev_clamped(fd, btrfs_sb_offset(i),
1830 BTRFS_SUPER_INFO_SIZE, block_count);
1831 if (!ret && (opflags & PREP_DEVICE_ZERO_END))
1832 ret = zero_dev_clamped(fd, block_count - ZERO_DEV_BYTES,
1833 ZERO_DEV_BYTES, block_count);
1836 error("failed to zero device '%s': %s", file, strerror(-ret));
1840 ret = btrfs_wipe_existing_sb(fd);
1842 error("cannot wipe superblocks on %s", file);
1846 *block_count_ret = block_count;
1850 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
1851 struct btrfs_root *root, u64 objectid)
1854 struct btrfs_inode_item inode_item;
1855 time_t now = time(NULL);
1857 memset(&inode_item, 0, sizeof(inode_item));
1858 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
1859 btrfs_set_stack_inode_size(&inode_item, 0);
1860 btrfs_set_stack_inode_nlink(&inode_item, 1);
1861 btrfs_set_stack_inode_nbytes(&inode_item, root->nodesize);
1862 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
1863 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
1864 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
1865 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
1866 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
1867 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
1868 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
1869 btrfs_set_stack_timespec_sec(&inode_item.otime, 0);
1870 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
1872 if (root->fs_info->tree_root == root)
1873 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
1875 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
1879 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
1883 btrfs_set_root_dirid(&root->root_item, objectid);
1890 * checks if a path is a block device node
1891 * Returns negative errno on failure, otherwise
1892 * returns 1 for blockdev, 0 for not-blockdev
1894 int is_block_device(const char *path)
1896 struct stat statbuf;
1898 if (stat(path, &statbuf) < 0)
1901 return !!S_ISBLK(statbuf.st_mode);
1905 * check if given path is a mount point
1906 * return 1 if yes. 0 if no. -1 for error
1908 int is_mount_point(const char *path)
1914 f = setmntent("/proc/self/mounts", "r");
1918 while ((mnt = getmntent(f)) != NULL) {
1919 if (strcmp(mnt->mnt_dir, path))
1928 static int is_reg_file(const char *path)
1930 struct stat statbuf;
1932 if (stat(path, &statbuf) < 0)
1934 return S_ISREG(statbuf.st_mode);
1938 * This function checks if the given input parameter is
1940 * return <0 : some error in the given input
1941 * return BTRFS_ARG_UNKNOWN: unknown input
1942 * return BTRFS_ARG_UUID: given input is uuid
1943 * return BTRFS_ARG_MNTPOINT: given input is path
1944 * return BTRFS_ARG_REG: given input is regular file
1945 * return BTRFS_ARG_BLKDEV: given input is block device
1947 int check_arg_type(const char *input)
1950 char path[PATH_MAX];
1955 if (realpath(input, path)) {
1956 if (is_block_device(path) == 1)
1957 return BTRFS_ARG_BLKDEV;
1959 if (is_mount_point(path) == 1)
1960 return BTRFS_ARG_MNTPOINT;
1962 if (is_reg_file(path))
1963 return BTRFS_ARG_REG;
1965 return BTRFS_ARG_UNKNOWN;
1968 if (strlen(input) == (BTRFS_UUID_UNPARSED_SIZE - 1) &&
1969 !uuid_parse(input, uuid))
1970 return BTRFS_ARG_UUID;
1972 return BTRFS_ARG_UNKNOWN;
1976 * Find the mount point for a mounted device.
1977 * On success, returns 0 with mountpoint in *mp.
1978 * On failure, returns -errno (not mounted yields -EINVAL)
1979 * Is noisy on failures, expects to be given a mounted device.
1981 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size)
1986 ret = is_block_device(dev);
1989 error("not a block device: %s", dev);
1992 error("cannot check %s: %s", dev, strerror(-ret));
1997 fd = open(dev, O_RDONLY);
2000 error("cannot open %s: %s", dev, strerror(errno));
2004 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
2007 } else { /* mounted, all good */
2017 * Given a pathname, return a filehandle to:
2018 * the original pathname or,
2019 * if the pathname is a mounted btrfs device, to its mountpoint.
2021 * On error, return -1, errno should be set.
2023 int open_path_or_dev_mnt(const char *path, DIR **dirstream, int verbose)
2028 if (is_block_device(path)) {
2029 ret = get_btrfs_mount(path, mp, sizeof(mp));
2031 /* not a mounted btrfs dev */
2032 error_on(verbose, "'%s' is not a mounted btrfs device",
2037 ret = open_file_or_dir(mp, dirstream);
2038 error_on(verbose && ret < 0, "can't access '%s': %s",
2039 path, strerror(errno));
2041 ret = btrfs_open_dir(path, dirstream, 1);
2048 * Do the following checks before calling open_file_or_dir():
2049 * 1: path is in a btrfs filesystem
2050 * 2: path is a directory
2052 int btrfs_open_dir(const char *path, DIR **dirstream, int verbose)
2058 if (statfs(path, &stfs) != 0) {
2059 error_on(verbose, "cannot access '%s': %s", path,
2064 if (stfs.f_type != BTRFS_SUPER_MAGIC) {
2065 error_on(verbose, "not a btrfs filesystem: %s", path);
2069 if (stat(path, &st) != 0) {
2070 error_on(verbose, "cannot access '%s': %s", path,
2075 if (!S_ISDIR(st.st_mode)) {
2076 error_on(verbose, "not a directory: %s", path);
2080 ret = open_file_or_dir(path, dirstream);
2082 error_on(verbose, "cannot access '%s': %s", path,
2089 /* checks if a device is a loop device */
2090 static int is_loop_device (const char* device) {
2091 struct stat statbuf;
2093 if(stat(device, &statbuf) < 0)
2096 return (S_ISBLK(statbuf.st_mode) &&
2097 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
2101 * Takes a loop device path (e.g. /dev/loop0) and returns
2102 * the associated file (e.g. /images/my_btrfs.img) using
2105 static int resolve_loop_device_with_loopdev(const char* loop_dev, char* loop_file)
2109 struct loop_info64 lo64;
2111 fd = open(loop_dev, O_RDONLY | O_NONBLOCK);
2114 ret = ioctl(fd, LOOP_GET_STATUS64, &lo64);
2120 memcpy(loop_file, lo64.lo_file_name, sizeof(lo64.lo_file_name));
2121 loop_file[sizeof(lo64.lo_file_name)] = 0;
2129 /* Takes a loop device path (e.g. /dev/loop0) and returns
2130 * the associated file (e.g. /images/my_btrfs.img) */
2131 static int resolve_loop_device(const char* loop_dev, char* loop_file,
2138 char real_loop_dev[PATH_MAX];
2140 if (!realpath(loop_dev, real_loop_dev))
2142 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
2143 if (!(f = fopen(p, "r"))) {
2144 if (errno == ENOENT)
2146 * It's possibly a partitioned loop device, which is
2147 * resolvable with loopdev API.
2149 return resolve_loop_device_with_loopdev(loop_dev, loop_file);
2153 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
2154 ret = fscanf(f, fmt, loop_file);
2163 * Checks whether a and b are identical or device
2164 * files associated with the same block device
2166 static int is_same_blk_file(const char* a, const char* b)
2168 struct stat st_buf_a, st_buf_b;
2169 char real_a[PATH_MAX];
2170 char real_b[PATH_MAX];
2172 if (!realpath(a, real_a))
2173 strncpy_null(real_a, a);
2175 if (!realpath(b, real_b))
2176 strncpy_null(real_b, b);
2178 /* Identical path? */
2179 if (strcmp(real_a, real_b) == 0)
2182 if (stat(a, &st_buf_a) < 0 || stat(b, &st_buf_b) < 0) {
2183 if (errno == ENOENT)
2188 /* Same blockdevice? */
2189 if (S_ISBLK(st_buf_a.st_mode) && S_ISBLK(st_buf_b.st_mode) &&
2190 st_buf_a.st_rdev == st_buf_b.st_rdev) {
2195 if (st_buf_a.st_dev == st_buf_b.st_dev &&
2196 st_buf_a.st_ino == st_buf_b.st_ino) {
2203 /* checks if a and b are identical or device
2204 * files associated with the same block device or
2205 * if one file is a loop device that uses the other
2208 static int is_same_loop_file(const char* a, const char* b)
2210 char res_a[PATH_MAX];
2211 char res_b[PATH_MAX];
2212 const char* final_a = NULL;
2213 const char* final_b = NULL;
2216 /* Resolve a if it is a loop device */
2217 if((ret = is_loop_device(a)) < 0) {
2222 ret = resolve_loop_device(a, res_a, sizeof(res_a));
2233 /* Resolve b if it is a loop device */
2234 if ((ret = is_loop_device(b)) < 0) {
2239 ret = resolve_loop_device(b, res_b, sizeof(res_b));
2250 return is_same_blk_file(final_a, final_b);
2253 /* Checks if a file exists and is a block or regular file*/
2254 static int is_existing_blk_or_reg_file(const char* filename)
2258 if(stat(filename, &st_buf) < 0) {
2265 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
2268 /* Checks if a file is used (directly or indirectly via a loop device)
2269 * by a device in fs_devices
2271 static int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices,
2275 struct list_head *head;
2276 struct list_head *cur;
2277 struct btrfs_device *device;
2279 head = &fs_devices->devices;
2280 list_for_each(cur, head) {
2281 device = list_entry(cur, struct btrfs_device, dev_list);
2283 if((ret = is_same_loop_file(device->name, file)))
2291 * Resolve a pathname to a device mapper node to /dev/mapper/<name>
2292 * Returns NULL on invalid input or malloc failure; Other failures
2293 * will be handled by the caller using the input pathame.
2295 char *canonicalize_dm_name(const char *ptname)
2299 char path[PATH_MAX], name[PATH_MAX], *res = NULL;
2301 if (!ptname || !*ptname)
2304 snprintf(path, sizeof(path), "/sys/block/%s/dm/name", ptname);
2305 if (!(f = fopen(path, "r")))
2308 /* read <name>\n from sysfs */
2309 if (fgets(name, sizeof(name), f) && (sz = strlen(name)) > 1) {
2310 name[sz - 1] = '\0';
2311 snprintf(path, sizeof(path), "/dev/mapper/%s", name);
2313 if (access(path, F_OK) == 0)
2321 * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or
2322 * to a device mapper pathname.
2323 * Returns NULL on invalid input or malloc failure; Other failures
2324 * will be handled by the caller using the input pathame.
2326 char *canonicalize_path(const char *path)
2328 char *canonical, *p;
2330 if (!path || !*path)
2333 canonical = realpath(path, NULL);
2335 return strdup(path);
2336 p = strrchr(canonical, '/');
2337 if (p && strncmp(p, "/dm-", 4) == 0 && isdigit(*(p + 4))) {
2338 char *dm = canonicalize_dm_name(p + 1);
2349 * returns 1 if the device was mounted, < 0 on error or 0 if everything
2350 * is safe to continue.
2352 int check_mounted(const char* file)
2357 fd = open(file, O_RDONLY);
2359 error("mount check: cannot open %s: %s", file,
2364 ret = check_mounted_where(fd, file, NULL, 0, NULL);
2370 int check_mounted_where(int fd, const char *file, char *where, int size,
2371 struct btrfs_fs_devices **fs_dev_ret)
2376 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2380 /* scan the initial device */
2381 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
2382 &total_devs, BTRFS_SUPER_INFO_OFFSET, SBREAD_DEFAULT);
2383 is_btrfs = (ret >= 0);
2385 /* scan other devices */
2386 if (is_btrfs && total_devs > 1) {
2387 ret = btrfs_scan_devices();
2392 /* iterate over the list of currently mounted filesystems */
2393 if ((f = setmntent ("/proc/self/mounts", "r")) == NULL)
2396 while ((mnt = getmntent (f)) != NULL) {
2398 if(strcmp(mnt->mnt_type, "btrfs") != 0)
2401 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
2403 /* ignore entries in the mount table that are not
2404 associated with a file*/
2405 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
2406 goto out_mntloop_err;
2410 ret = is_same_loop_file(file, mnt->mnt_fsname);
2414 goto out_mntloop_err;
2419 /* Did we find an entry in mnt table? */
2420 if (mnt && size && where) {
2421 strncpy(where, mnt->mnt_dir, size);
2425 *fs_dev_ret = fs_devices_mnt;
2427 ret = (mnt != NULL);
2435 struct pending_dir {
2436 struct list_head list;
2437 char name[PATH_MAX];
2440 int btrfs_register_one_device(const char *fname)
2442 struct btrfs_ioctl_vol_args args;
2446 fd = open("/dev/btrfs-control", O_RDWR);
2449 "failed to open /dev/btrfs-control, skipping device registration: %s",
2453 memset(&args, 0, sizeof(args));
2454 strncpy_null(args.name, fname);
2455 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
2457 error("device scan failed on '%s': %s", fname,
2466 * Register all devices in the fs_uuid list created in the user
2467 * space. Ensure btrfs_scan_devices() is called before this func.
2469 int btrfs_register_all_devices(void)
2473 struct btrfs_fs_devices *fs_devices;
2474 struct btrfs_device *device;
2475 struct list_head *all_uuids;
2477 all_uuids = btrfs_scanned_uuids();
2479 list_for_each_entry(fs_devices, all_uuids, list) {
2480 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2482 err = btrfs_register_one_device(device->name);
2492 int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
2495 struct btrfs_super_block *disk_super;
2499 buf = malloc(BTRFS_SUPER_INFO_SIZE);
2504 ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
2505 if (ret != BTRFS_SUPER_INFO_SIZE)
2509 disk_super = (struct btrfs_super_block *)buf;
2511 * Accept devices from the same filesystem, allow partially created
2514 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC &&
2515 btrfs_super_magic(disk_super) != BTRFS_MAGIC_PARTIAL)
2518 if (!memcmp(disk_super->fsid, root->fs_info->super_copy->fsid,
2528 * Note: this function uses a static per-thread buffer. Do not call this
2529 * function more than 10 times within one argument list!
2531 const char *pretty_size_mode(u64 size, unsigned mode)
2533 static __thread int ps_index = 0;
2534 static __thread char ps_array[10][32];
2537 ret = ps_array[ps_index];
2540 (void)pretty_size_snprintf(size, ret, 32, mode);
2545 static const char* unit_suffix_binary[] =
2546 { "B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB"};
2547 static const char* unit_suffix_decimal[] =
2548 { "B", "kB", "MB", "GB", "TB", "PB", "EB"};
2550 int pretty_size_snprintf(u64 size, char *str, size_t str_size, unsigned unit_mode)
2556 const char** suffix = NULL;
2562 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_RAW) {
2563 snprintf(str, str_size, "%llu", size);
2567 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_BINARY) {
2570 suffix = unit_suffix_binary;
2571 } else if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_DECIMAL) {
2574 suffix = unit_suffix_decimal;
2579 fprintf(stderr, "INTERNAL ERROR: unknown unit base, mode %d\n",
2587 switch (unit_mode & UNITS_MODE_MASK) {
2588 case UNITS_TBYTES: base *= mult; num_divs++;
2589 case UNITS_GBYTES: base *= mult; num_divs++;
2590 case UNITS_MBYTES: base *= mult; num_divs++;
2591 case UNITS_KBYTES: num_divs++;
2598 while (size >= mult) {
2604 * If the value is smaller than base, we didn't do any
2605 * division, in that case, base should be 1, not original
2606 * base, or the unit will be wrong
2612 if (num_divs >= ARRAY_SIZE(unit_suffix_binary)) {
2614 printf("INTERNAL ERROR: unsupported unit suffix, index %d\n",
2619 fraction = (float)last_size / base;
2621 return snprintf(str, str_size, "%.2f%s", fraction, suffix[num_divs]);
2625 * __strncpy_null - strncpy with null termination
2626 * @dest: the target array
2627 * @src: the source string
2628 * @n: maximum bytes to copy (size of *dest)
2630 * Like strncpy, but ensures destination is null-terminated.
2632 * Copies the string pointed to by src, including the terminating null
2633 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
2634 * of n bytes. Then ensure that dest is null-terminated.
2636 char *__strncpy_null(char *dest, const char *src, size_t n)
2638 strncpy(dest, src, n);
2645 * Checks to make sure that the label matches our requirements.
2647 0 if everything is safe and usable
2648 -1 if the label is too long
2650 static int check_label(const char *input)
2652 int len = strlen(input);
2654 if (len > BTRFS_LABEL_SIZE - 1) {
2655 error("label %s is too long (max %d)", input,
2656 BTRFS_LABEL_SIZE - 1);
2663 static int set_label_unmounted(const char *dev, const char *label)
2665 struct btrfs_trans_handle *trans;
2666 struct btrfs_root *root;
2669 ret = check_mounted(dev);
2671 error("checking mount status of %s failed: %d", dev, ret);
2675 error("device %s is mounted, use mount point", dev);
2679 /* Open the super_block at the default location
2680 * and as read-write.
2682 root = open_ctree(dev, 0, OPEN_CTREE_WRITES);
2683 if (!root) /* errors are printed by open_ctree() */
2686 trans = btrfs_start_transaction(root, 1);
2687 __strncpy_null(root->fs_info->super_copy->label, label, BTRFS_LABEL_SIZE - 1);
2689 btrfs_commit_transaction(trans, root);
2691 /* Now we close it since we are done. */
2696 static int set_label_mounted(const char *mount_path, const char *labelp)
2699 char label[BTRFS_LABEL_SIZE];
2701 fd = open(mount_path, O_RDONLY | O_NOATIME);
2703 error("unable to access %s: %s", mount_path, strerror(errno));
2707 memset(label, 0, sizeof(label));
2708 __strncpy_null(label, labelp, BTRFS_LABEL_SIZE - 1);
2709 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
2710 error("unable to set label of %s: %s", mount_path,
2720 int get_label_unmounted(const char *dev, char *label)
2722 struct btrfs_root *root;
2725 ret = check_mounted(dev);
2727 error("checking mount status of %s failed: %d", dev, ret);
2731 /* Open the super_block at the default location
2734 root = open_ctree(dev, 0, 0);
2738 __strncpy_null(label, root->fs_info->super_copy->label,
2739 BTRFS_LABEL_SIZE - 1);
2741 /* Now we close it since we are done. */
2747 * If a partition is mounted, try to get the filesystem label via its
2748 * mounted path rather than device. Return the corresponding error
2749 * the user specified the device path.
2751 int get_label_mounted(const char *mount_path, char *labelp)
2753 char label[BTRFS_LABEL_SIZE];
2757 fd = open(mount_path, O_RDONLY | O_NOATIME);
2759 error("unable to access %s: %s", mount_path, strerror(errno));
2763 memset(label, '\0', sizeof(label));
2764 ret = ioctl(fd, BTRFS_IOC_GET_FSLABEL, label);
2766 if (errno != ENOTTY)
2767 error("unable to get label of %s: %s", mount_path,
2774 __strncpy_null(labelp, label, BTRFS_LABEL_SIZE - 1);
2779 int get_label(const char *btrfs_dev, char *label)
2783 ret = is_existing_blk_or_reg_file(btrfs_dev);
2785 ret = get_label_mounted(btrfs_dev, label);
2787 ret = get_label_unmounted(btrfs_dev, label);
2792 int set_label(const char *btrfs_dev, const char *label)
2796 if (check_label(label))
2799 ret = is_existing_blk_or_reg_file(btrfs_dev);
2801 ret = set_label_mounted(btrfs_dev, label);
2803 ret = set_label_unmounted(btrfs_dev, label);
2809 * A not-so-good version fls64. No fascinating optimization since
2810 * no one except parse_size use it
2812 static int fls64(u64 x)
2816 for (i = 0; i <64; i++)
2817 if (x << i & (1ULL << 63))
2822 u64 parse_size(char *s)
2830 error("size value is empty");
2834 error("size value '%s' is less equal than 0", s);
2837 ret = strtoull(s, &endptr, 10);
2839 error("size value '%s' is invalid", s);
2842 if (endptr[0] && endptr[1]) {
2843 error("illegal suffix contains character '%c' in wrong position",
2848 * strtoll returns LLONG_MAX when overflow, if this happens,
2849 * need to call strtoull to get the real size
2851 if (errno == ERANGE && ret == ULLONG_MAX) {
2852 error("size value '%s' is too large for u64", s);
2856 c = tolower(endptr[0]);
2879 error("unknown size descriptor '%c'", c);
2883 /* Check whether ret * mult overflow */
2884 if (fls64(ret) + fls64(mult) - 1 > 64) {
2885 error("size value '%s' is too large for u64", s);
2892 u64 parse_qgroupid(const char *p)
2894 char *s = strchr(p, '/');
2895 const char *ptr_src_end = p + strlen(p);
2896 char *ptr_parse_end = NULL;
2905 /* Numeric format like '0/257' is the primary case */
2907 id = strtoull(p, &ptr_parse_end, 10);
2908 if (ptr_parse_end != ptr_src_end)
2912 level = strtoull(p, &ptr_parse_end, 10);
2913 if (ptr_parse_end != s)
2916 id = strtoull(s + 1, &ptr_parse_end, 10);
2917 if (ptr_parse_end != ptr_src_end)
2920 return (level << BTRFS_QGROUP_LEVEL_SHIFT) | id;
2923 /* Path format like subv at 'my_subvol' is the fallback case */
2924 ret = test_issubvolume(p);
2925 if (ret < 0 || !ret)
2927 fd = open(p, O_RDONLY);
2930 ret = lookup_ino_rootid(fd, &id);
2932 error("failed to lookup root id: %s", strerror(-ret));
2939 error("invalid qgroupid or subvolume path: %s", p);
2943 int open_file_or_dir3(const char *fname, DIR **dirstream, int open_flags)
2949 ret = stat(fname, &st);
2953 if (S_ISDIR(st.st_mode)) {
2954 *dirstream = opendir(fname);
2957 fd = dirfd(*dirstream);
2958 } else if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) {
2959 fd = open(fname, open_flags);
2962 * we set this on purpose, in case the caller output
2963 * strerror(errno) as success
2971 closedir(*dirstream);
2978 int open_file_or_dir(const char *fname, DIR **dirstream)
2980 return open_file_or_dir3(fname, dirstream, O_RDWR);
2983 void close_file_or_dir(int fd, DIR *dirstream)
2986 closedir(dirstream);
2991 int get_device_info(int fd, u64 devid,
2992 struct btrfs_ioctl_dev_info_args *di_args)
2996 di_args->devid = devid;
2997 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
2999 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
3000 return ret < 0 ? -errno : 0;
3003 static u64 find_max_device_id(struct btrfs_ioctl_search_args *search_args,
3006 struct btrfs_dev_item *dev_item;
3007 char *buf = search_args->buf;
3009 buf += (nr_items - 1) * (sizeof(struct btrfs_ioctl_search_header)
3010 + sizeof(struct btrfs_dev_item));
3011 buf += sizeof(struct btrfs_ioctl_search_header);
3013 dev_item = (struct btrfs_dev_item *)buf;
3015 return btrfs_stack_device_id(dev_item);
3018 static int search_chunk_tree_for_fs_info(int fd,
3019 struct btrfs_ioctl_fs_info_args *fi_args)
3023 u64 start_devid = 1;
3024 struct btrfs_ioctl_search_args search_args;
3025 struct btrfs_ioctl_search_key *search_key = &search_args.key;
3027 fi_args->num_devices = 0;
3029 max_items = BTRFS_SEARCH_ARGS_BUFSIZE
3030 / (sizeof(struct btrfs_ioctl_search_header)
3031 + sizeof(struct btrfs_dev_item));
3033 search_key->tree_id = BTRFS_CHUNK_TREE_OBJECTID;
3034 search_key->min_objectid = BTRFS_DEV_ITEMS_OBJECTID;
3035 search_key->max_objectid = BTRFS_DEV_ITEMS_OBJECTID;
3036 search_key->min_type = BTRFS_DEV_ITEM_KEY;
3037 search_key->max_type = BTRFS_DEV_ITEM_KEY;
3038 search_key->min_transid = 0;
3039 search_key->max_transid = (u64)-1;
3040 search_key->nr_items = max_items;
3041 search_key->max_offset = (u64)-1;
3044 search_key->min_offset = start_devid;
3046 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH, &search_args);
3050 fi_args->num_devices += (u64)search_key->nr_items;
3052 if (search_key->nr_items == max_items) {
3053 start_devid = find_max_device_id(&search_args,
3054 search_key->nr_items) + 1;
3058 /* get the lastest max_id to stay consistent with the num_devices */
3059 if (search_key->nr_items == 0)
3061 * last tree_search returns an empty buf, use the devid of
3062 * the last dev_item of the previous tree_search
3064 fi_args->max_id = start_devid - 1;
3066 fi_args->max_id = find_max_device_id(&search_args,
3067 search_key->nr_items);
3073 * For a given path, fill in the ioctl fs_ and info_ args.
3074 * If the path is a btrfs mountpoint, fill info for all devices.
3075 * If the path is a btrfs device, fill in only that device.
3077 * The path provided must be either on a mounted btrfs fs,
3078 * or be a mounted btrfs device.
3080 * Returns 0 on success, or a negative errno.
3082 int get_fs_info(const char *path, struct btrfs_ioctl_fs_info_args *fi_args,
3083 struct btrfs_ioctl_dev_info_args **di_ret)
3090 struct btrfs_fs_devices *fs_devices_mnt = NULL;
3091 struct btrfs_ioctl_dev_info_args *di_args;
3092 struct btrfs_ioctl_dev_info_args tmp;
3094 DIR *dirstream = NULL;
3096 memset(fi_args, 0, sizeof(*fi_args));
3098 if (is_block_device(path) == 1) {
3099 struct btrfs_super_block *disk_super;
3100 char buf[BTRFS_SUPER_INFO_SIZE];
3103 /* Ensure it's mounted, then set path to the mountpoint */
3104 fd = open(path, O_RDONLY);
3107 error("cannot open %s: %s", path, strerror(errno));
3110 ret = check_mounted_where(fd, path, mp, sizeof(mp),
3119 /* Only fill in this one device */
3120 fi_args->num_devices = 1;
3122 disk_super = (struct btrfs_super_block *)buf;
3123 ret = btrfs_read_dev_super(fd, disk_super,
3124 BTRFS_SUPER_INFO_OFFSET, 0);
3129 devid = btrfs_stack_device_id(&disk_super->dev_item);
3131 fi_args->max_id = devid;
3134 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
3138 /* at this point path must not be for a block device */
3139 fd = open_file_or_dir(path, &dirstream);
3145 /* fill in fi_args if not just a single device */
3146 if (fi_args->num_devices != 1) {
3147 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
3154 * The fs_args->num_devices does not include seed devices
3156 ret = search_chunk_tree_for_fs_info(fd, fi_args);
3161 * search_chunk_tree_for_fs_info() will lacks the devid 0
3162 * so manual probe for it here.
3164 ret = get_device_info(fd, 0, &tmp);
3166 fi_args->num_devices++;
3174 if (!fi_args->num_devices)
3177 di_args = *di_ret = malloc((fi_args->num_devices) * sizeof(*di_args));
3184 memcpy(di_args, &tmp, sizeof(tmp));
3185 for (; i <= fi_args->max_id; ++i) {
3186 ret = get_device_info(fd, i, &di_args[ndevs]);
3195 * only when the only dev we wanted to find is not there then
3196 * let any error be returned
3198 if (fi_args->num_devices != 1) {
3204 close_file_or_dir(fd, dirstream);
3208 #define isoctal(c) (((c) & ~7) == '0')
3210 static inline void translate(char *f, char *t)
3212 while (*f != '\0') {
3214 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
3215 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
3225 * Checks if the swap device.
3226 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
3228 static int is_swap_device(const char *file)
3239 if (stat(file, &st_buf) < 0)
3241 if (S_ISBLK(st_buf.st_mode))
3242 dev = st_buf.st_rdev;
3243 else if (S_ISREG(st_buf.st_mode)) {
3244 dev = st_buf.st_dev;
3245 ino = st_buf.st_ino;
3249 if ((f = fopen("/proc/swaps", "r")) == NULL)
3252 /* skip the first line */
3253 if (fgets(tmp, sizeof(tmp), f) == NULL)
3256 while (fgets(tmp, sizeof(tmp), f) != NULL) {
3257 if ((cp = strchr(tmp, ' ')) != NULL)
3259 if ((cp = strchr(tmp, '\t')) != NULL)
3261 translate(tmp, buf);
3262 if (stat(buf, &st_buf) != 0)
3264 if (S_ISBLK(st_buf.st_mode)) {
3265 if (dev == st_buf.st_rdev) {
3269 } else if (S_ISREG(st_buf.st_mode)) {
3270 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {
3284 * Check for existing filesystem or partition table on device.
3286 * 1 for existing fs or partition
3287 * 0 for nothing found
3288 * -1 for internal error
3290 static int check_overwrite(const char *device)
3293 blkid_probe pr = NULL;
3297 if (!device || !*device)
3300 ret = -1; /* will reset on success of all setup calls */
3302 pr = blkid_new_probe_from_filename(device);
3306 size = blkid_probe_get_size(pr);
3310 /* nothing to overwrite on a 0-length device */
3316 ret = blkid_probe_enable_partitions(pr, 1);
3320 ret = blkid_do_fullprobe(pr);
3325 * Blkid returns 1 for nothing found and 0 when it finds a signature,
3326 * but we want the exact opposite, so reverse the return value here.
3328 * In addition print some useful diagnostics about what actually is
3336 if (!blkid_probe_lookup_value(pr, "TYPE", &type, NULL)) {
3338 "%s appears to contain an existing "
3339 "filesystem (%s).\n", device, type);
3340 } else if (!blkid_probe_lookup_value(pr, "PTTYPE", &type, NULL)) {
3342 "%s appears to contain a partition "
3343 "table (%s).\n", device, type);
3346 "%s appears to contain something weird "
3347 "according to blkid\n", device);
3353 blkid_free_probe(pr);
3356 "probe of %s failed, cannot detect "
3357 "existing filesystem.\n", device);
3361 static int group_profile_devs_min(u64 flag)
3363 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3364 case 0: /* single */
3365 case BTRFS_BLOCK_GROUP_DUP:
3367 case BTRFS_BLOCK_GROUP_RAID0:
3368 case BTRFS_BLOCK_GROUP_RAID1:
3369 case BTRFS_BLOCK_GROUP_RAID5:
3371 case BTRFS_BLOCK_GROUP_RAID6:
3373 case BTRFS_BLOCK_GROUP_RAID10:
3380 int test_num_disk_vs_raid(u64 metadata_profile, u64 data_profile,
3381 u64 dev_cnt, int mixed, int ssd)
3384 u64 profile = metadata_profile | data_profile;
3389 allowed |= BTRFS_BLOCK_GROUP_RAID10;
3391 allowed |= BTRFS_BLOCK_GROUP_RAID6;
3393 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
3394 BTRFS_BLOCK_GROUP_RAID5;
3396 allowed |= BTRFS_BLOCK_GROUP_DUP;
3399 if (dev_cnt > 1 && profile & BTRFS_BLOCK_GROUP_DUP) {
3400 warning("DUP is not recommended on filesystem with multiple devices");
3402 if (metadata_profile & ~allowed) {
3404 "ERROR: unable to create FS with metadata profile %s "
3405 "(have %llu devices but %d devices are required)\n",
3406 btrfs_group_profile_str(metadata_profile), dev_cnt,
3407 group_profile_devs_min(metadata_profile));
3410 if (data_profile & ~allowed) {
3412 "ERROR: unable to create FS with data profile %s "
3413 "(have %llu devices but %d devices are required)\n",
3414 btrfs_group_profile_str(data_profile), dev_cnt,
3415 group_profile_devs_min(data_profile));
3419 if (dev_cnt == 3 && profile & BTRFS_BLOCK_GROUP_RAID6) {
3420 warning("RAID6 is not recommended on filesystem with 3 devices only");
3422 if (dev_cnt == 2 && profile & BTRFS_BLOCK_GROUP_RAID5) {
3423 warning("RAID5 is not recommended on filesystem with 2 devices only");
3425 warning_on(!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP) && ssd,
3426 "DUP may not actually lead to 2 copies on the device, see manual page");
3431 int group_profile_max_safe_loss(u64 flags)
3433 switch (flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3434 case 0: /* single */
3435 case BTRFS_BLOCK_GROUP_DUP:
3436 case BTRFS_BLOCK_GROUP_RAID0:
3438 case BTRFS_BLOCK_GROUP_RAID1:
3439 case BTRFS_BLOCK_GROUP_RAID5:
3440 case BTRFS_BLOCK_GROUP_RAID10:
3442 case BTRFS_BLOCK_GROUP_RAID6:
3450 * Check if a device is suitable for btrfs
3452 * 1: something is wrong, an error is printed
3455 int test_dev_for_mkfs(const char *file, int force_overwrite)
3460 ret = is_swap_device(file);
3462 error("checking status of %s: %s", file, strerror(-ret));
3466 error("%s is a swap device", file);
3469 if (!force_overwrite) {
3470 if (check_overwrite(file)) {
3471 error("use the -f option to force overwrite of %s",
3476 ret = check_mounted(file);
3478 error("cannot check mount status of %s: %s", file,
3483 error("%s is mounted", file);
3486 /* check if the device is busy */
3487 fd = open(file, O_RDWR|O_EXCL);
3489 error("unable to open %s: %s", file, strerror(errno));
3492 if (fstat(fd, &st)) {
3493 error("unable to stat %s: %s", file, strerror(errno));
3497 if (!S_ISBLK(st.st_mode)) {
3498 error("%s is not a block device", file);
3506 int btrfs_scan_devices(void)
3511 struct btrfs_fs_devices *tmp_devices;
3512 blkid_dev_iterate iter = NULL;
3513 blkid_dev dev = NULL;
3514 blkid_cache cache = NULL;
3515 char path[PATH_MAX];
3517 if (btrfs_scan_done)
3520 if (blkid_get_cache(&cache, NULL) < 0) {
3521 error("blkid cache get failed");
3524 blkid_probe_all(cache);
3525 iter = blkid_dev_iterate_begin(cache);
3526 blkid_dev_set_search(iter, "TYPE", "btrfs");
3527 while (blkid_dev_next(iter, &dev) == 0) {
3528 dev = blkid_verify(cache, dev);
3531 /* if we are here its definitely a btrfs disk*/
3532 strncpy_null(path, blkid_dev_devname(dev));
3534 fd = open(path, O_RDONLY);
3536 error("cannot open %s: %s", path, strerror(errno));
3539 ret = btrfs_scan_one_device(fd, path, &tmp_devices,
3540 &num_devices, BTRFS_SUPER_INFO_OFFSET,
3543 error("cannot scan %s: %s", path, strerror(-ret));
3550 blkid_dev_iterate_end(iter);
3551 blkid_put_cache(cache);
3553 btrfs_scan_done = 1;
3558 int is_vol_small(const char *file)
3565 fd = open(file, O_RDONLY);
3568 if (fstat(fd, &st) < 0) {
3573 size = btrfs_device_size(fd, &st);
3578 if (size < BTRFS_MKFS_SMALL_VOLUME_SIZE) {
3588 * This reads a line from the stdin and only returns non-zero if the
3589 * first whitespace delimited token is a case insensitive match with yes
3592 int ask_user(const char *question)
3594 char buf[30] = {0,};
3595 char *saveptr = NULL;
3598 printf("%s [y/N]: ", question);
3600 return fgets(buf, sizeof(buf) - 1, stdin) &&
3601 (answer = strtok_r(buf, " \t\n\r", &saveptr)) &&
3602 (!strcasecmp(answer, "yes") || !strcasecmp(answer, "y"));
3607 * - file or directory return the containing tree root id
3608 * - subvolume return its own tree id
3609 * - BTRFS_EMPTY_SUBVOL_DIR_OBJECTID (directory with ino == 2) the result is
3610 * undefined and function returns -1
3612 int lookup_ino_rootid(int fd, u64 *rootid)
3614 struct btrfs_ioctl_ino_lookup_args args;
3617 memset(&args, 0, sizeof(args));
3619 args.objectid = BTRFS_FIRST_FREE_OBJECTID;
3621 ret = ioctl(fd, BTRFS_IOC_INO_LOOKUP, &args);
3625 *rootid = args.treeid;
3631 * return 0 if a btrfs mount point is found
3632 * return 1 if a mount point is found but not btrfs
3633 * return <0 if something goes wrong
3635 int find_mount_root(const char *path, char **mount_root)
3643 int longest_matchlen = 0;
3644 char *longest_match = NULL;
3646 fd = open(path, O_RDONLY | O_NOATIME);
3651 mnttab = setmntent("/proc/self/mounts", "r");
3655 while ((ent = getmntent(mnttab))) {
3656 len = strlen(ent->mnt_dir);
3657 if (strncmp(ent->mnt_dir, path, len) == 0) {
3658 /* match found and use the latest match */
3659 if (longest_matchlen <= len) {
3660 free(longest_match);
3661 longest_matchlen = len;
3662 longest_match = strdup(ent->mnt_dir);
3663 not_btrfs = strcmp(ent->mnt_type, "btrfs");
3672 free(longest_match);
3677 *mount_root = realpath(longest_match, NULL);
3681 free(longest_match);
3685 int test_minimum_size(const char *file, u32 nodesize)
3688 struct stat statbuf;
3690 fd = open(file, O_RDONLY);
3693 if (stat(file, &statbuf) < 0) {
3697 if (btrfs_device_size(fd, &statbuf) < btrfs_min_dev_size(nodesize)) {
3707 * Test if path is a directory
3709 * 0 - path exists but it is not a directory
3710 * 1 - path exists and it is a directory
3713 int test_isdir(const char *path)
3718 ret = stat(path, &st);
3722 return !!S_ISDIR(st.st_mode);
3725 void units_set_mode(unsigned *units, unsigned mode)
3727 unsigned base = *units & UNITS_MODE_MASK;
3729 *units = base | mode;
3732 void units_set_base(unsigned *units, unsigned base)
3734 unsigned mode = *units & ~UNITS_MODE_MASK;
3736 *units = base | mode;
3739 int find_next_key(struct btrfs_path *path, struct btrfs_key *key)
3743 for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
3744 if (!path->nodes[level])
3746 if (path->slots[level] + 1 >=
3747 btrfs_header_nritems(path->nodes[level]))
3750 btrfs_item_key_to_cpu(path->nodes[level], key,
3751 path->slots[level] + 1);
3753 btrfs_node_key_to_cpu(path->nodes[level], key,
3754 path->slots[level] + 1);
3760 const char* btrfs_group_type_str(u64 flag)
3762 u64 mask = BTRFS_BLOCK_GROUP_TYPE_MASK |
3763 BTRFS_SPACE_INFO_GLOBAL_RSV;
3765 switch (flag & mask) {
3766 case BTRFS_BLOCK_GROUP_DATA:
3768 case BTRFS_BLOCK_GROUP_SYSTEM:
3770 case BTRFS_BLOCK_GROUP_METADATA:
3772 case BTRFS_BLOCK_GROUP_DATA|BTRFS_BLOCK_GROUP_METADATA:
3773 return "Data+Metadata";
3774 case BTRFS_SPACE_INFO_GLOBAL_RSV:
3775 return "GlobalReserve";
3781 const char* btrfs_group_profile_str(u64 flag)
3783 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3786 case BTRFS_BLOCK_GROUP_RAID0:
3788 case BTRFS_BLOCK_GROUP_RAID1:
3790 case BTRFS_BLOCK_GROUP_RAID5:
3792 case BTRFS_BLOCK_GROUP_RAID6:
3794 case BTRFS_BLOCK_GROUP_DUP:
3796 case BTRFS_BLOCK_GROUP_RAID10:
3803 u64 disk_size(const char *path)
3807 if (statfs(path, &sfs) < 0)
3810 return sfs.f_bsize * sfs.f_blocks;
3813 u64 get_partition_size(const char *dev)
3816 int fd = open(dev, O_RDONLY);
3820 if (ioctl(fd, BLKGETSIZE64, &result) < 0) {
3829 int btrfs_tree_search2_ioctl_supported(int fd)
3831 struct btrfs_ioctl_search_args_v2 *args2;
3832 struct btrfs_ioctl_search_key *sk;
3833 int args2_size = 1024;
3834 char args2_buf[args2_size];
3836 static int v2_supported = -1;
3838 if (v2_supported != -1)
3839 return v2_supported;
3841 args2 = (struct btrfs_ioctl_search_args_v2 *)args2_buf;
3845 * Search for the extent tree item in the root tree.
3847 sk->tree_id = BTRFS_ROOT_TREE_OBJECTID;
3848 sk->min_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3849 sk->max_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3850 sk->min_type = BTRFS_ROOT_ITEM_KEY;
3851 sk->max_type = BTRFS_ROOT_ITEM_KEY;
3853 sk->max_offset = (u64)-1;
3854 sk->min_transid = 0;
3855 sk->max_transid = (u64)-1;
3857 args2->buf_size = args2_size - sizeof(struct btrfs_ioctl_search_args_v2);
3858 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH_V2, args2);
3859 if (ret == -EOPNOTSUPP)
3866 return v2_supported;
3869 int btrfs_check_nodesize(u32 nodesize, u32 sectorsize, u64 features)
3871 if (nodesize < sectorsize) {
3872 error("illegal nodesize %u (smaller than %u)",
3873 nodesize, sectorsize);
3875 } else if (nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) {
3876 error("illegal nodesize %u (larger than %u)",
3877 nodesize, BTRFS_MAX_METADATA_BLOCKSIZE);
3879 } else if (nodesize & (sectorsize - 1)) {
3880 error("illegal nodesize %u (not aligned to %u)",
3881 nodesize, sectorsize);
3883 } else if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS &&
3884 nodesize != sectorsize) {
3885 error("illegal nodesize %u (not equal to %u for mixed block group)",
3886 nodesize, sectorsize);
3893 * Copy a path argument from SRC to DEST and check the SRC length if it's at
3894 * most PATH_MAX and fits into DEST. DESTLEN is supposed to be exact size of
3896 * The destination buffer is zero terminated.
3897 * Return < 0 for error, 0 otherwise.
3899 int arg_copy_path(char *dest, const char *src, int destlen)
3901 size_t len = strlen(src);
3903 if (len >= PATH_MAX || len >= destlen)
3904 return -ENAMETOOLONG;
3906 __strncpy_null(dest, src, destlen);
3911 unsigned int get_unit_mode_from_arg(int *argc, char *argv[], int df_mode)
3913 unsigned int unit_mode = UNITS_DEFAULT;
3917 for (arg_i = 0; arg_i < *argc; arg_i++) {
3918 if (!strcmp(argv[arg_i], "--"))
3921 if (!strcmp(argv[arg_i], "--raw")) {
3922 unit_mode = UNITS_RAW;
3926 if (!strcmp(argv[arg_i], "--human-readable")) {
3927 unit_mode = UNITS_HUMAN_BINARY;
3932 if (!strcmp(argv[arg_i], "--iec")) {
3933 units_set_mode(&unit_mode, UNITS_BINARY);
3937 if (!strcmp(argv[arg_i], "--si")) {
3938 units_set_mode(&unit_mode, UNITS_DECIMAL);
3943 if (!strcmp(argv[arg_i], "--kbytes")) {
3944 units_set_base(&unit_mode, UNITS_KBYTES);
3948 if (!strcmp(argv[arg_i], "--mbytes")) {
3949 units_set_base(&unit_mode, UNITS_MBYTES);
3953 if (!strcmp(argv[arg_i], "--gbytes")) {
3954 units_set_base(&unit_mode, UNITS_GBYTES);
3958 if (!strcmp(argv[arg_i], "--tbytes")) {
3959 units_set_base(&unit_mode, UNITS_TBYTES);
3967 if (!strcmp(argv[arg_i], "-b")) {
3968 unit_mode = UNITS_RAW;
3972 if (!strcmp(argv[arg_i], "-h")) {
3973 unit_mode = UNITS_HUMAN_BINARY;
3977 if (!strcmp(argv[arg_i], "-H")) {
3978 unit_mode = UNITS_HUMAN_DECIMAL;
3982 if (!strcmp(argv[arg_i], "-k")) {
3983 units_set_base(&unit_mode, UNITS_KBYTES);
3987 if (!strcmp(argv[arg_i], "-m")) {
3988 units_set_base(&unit_mode, UNITS_MBYTES);
3992 if (!strcmp(argv[arg_i], "-g")) {
3993 units_set_base(&unit_mode, UNITS_GBYTES);
3997 if (!strcmp(argv[arg_i], "-t")) {
3998 units_set_base(&unit_mode, UNITS_TBYTES);
4004 for (arg_i = 0, arg_end = 0; arg_i < *argc; arg_i++) {
4007 argv[arg_end] = argv[arg_i];
4016 int string_is_numerical(const char *str)
4018 if (!(*str >= '0' && *str <= '9'))
4020 while (*str >= '0' && *str <= '9')
4028 * Preprocess @argv with getopt_long to reorder options and consume the "--"
4030 * Unknown short and long options are reported, optionally the @usage is printed
4033 void clean_args_no_options(int argc, char *argv[], const char * const *usagestr)
4035 static const struct option long_options[] = {
4040 int c = getopt_long(argc, argv, "", long_options, NULL);
4054 * Same as clean_args_no_options but pass through arguments that could look
4055 * like short options. Eg. reisze which takes a negative resize argument like
4058 * This accepts only two forms:
4059 * - "-- option1 option2 ..."
4060 * - "option1 option2 ..."
4062 void clean_args_no_options_relaxed(int argc, char *argv[], const char * const *usagestr)
4067 if (strcmp(argv[1], "--") == 0)
4071 /* Subvolume helper functions */
4073 * test if name is a correct subvolume name
4074 * this function return
4075 * 0-> name is not a correct subvolume name
4076 * 1-> name is a correct subvolume name
4078 int test_issubvolname(const char *name)
4080 return name[0] != '\0' && !strchr(name, '/') &&
4081 strcmp(name, ".") && strcmp(name, "..");
4085 * Test if path is a subvolume
4087 * 0 - path exists but it is not a subvolume
4088 * 1 - path exists and it is a subvolume
4091 int test_issubvolume(const char *path)
4097 res = stat(path, &st);
4101 if (st.st_ino != BTRFS_FIRST_FREE_OBJECTID || !S_ISDIR(st.st_mode))
4104 res = statfs(path, &stfs);
4108 return (int)stfs.f_type == BTRFS_SUPER_MAGIC;
4111 const char *subvol_strip_mountpoint(const char *mnt, const char *full_path)
4113 int len = strlen(mnt);
4117 if (mnt[len - 1] != '/')
4120 return full_path + len;
4127 * 1: Error; and error info printed to the terminal. Fixme.
4128 * 2: If the fullpath is root tree instead of subvol tree
4130 int get_subvol_info(const char *fullpath, struct root_info *get_ri)
4137 const char *svpath = NULL;
4138 DIR *dirstream1 = NULL;
4139 DIR *dirstream2 = NULL;
4141 ret = test_issubvolume(fullpath);
4145 error("not a subvolume: %s", fullpath);
4149 ret = find_mount_root(fullpath, &mnt);
4153 error("%s doesn't belong to btrfs mount point", fullpath);
4157 svpath = subvol_strip_mountpoint(mnt, fullpath);
4159 fd = btrfs_open_dir(fullpath, &dirstream1, 1);
4163 ret = btrfs_list_get_path_rootid(fd, &sv_id);
4165 error("can't get rootid for '%s'", fullpath);
4169 mntfd = btrfs_open_dir(mnt, &dirstream2, 1);
4173 if (sv_id == BTRFS_FS_TREE_OBJECTID) {
4176 * So that caller may decide if thats an error or just fine.
4181 memset(get_ri, 0, sizeof(*get_ri));
4182 get_ri->root_id = sv_id;
4184 ret = btrfs_get_subvol(mntfd, get_ri);
4186 error("can't find '%s': %d", svpath, ret);
4189 close_file_or_dir(mntfd, dirstream2);
4190 close_file_or_dir(fd, dirstream1);
4196 void init_rand_seed(u64 seed)
4200 /* only use the last 48 bits */
4201 for (i = 0; i < 3; i++) {
4202 rand_seed[i] = (unsigned short)(seed ^ (unsigned short)(-1));
4205 rand_seed_initlized = 1;
4208 static void __init_seed(void)
4214 if(rand_seed_initlized)
4216 /* Use urandom as primary seed source. */
4217 fd = open("/dev/urandom", O_RDONLY);
4219 ret = read(fd, rand_seed, sizeof(rand_seed));
4221 if (ret < sizeof(rand_seed))
4225 /* Use time and pid as fallback seed */
4226 warning("failed to read /dev/urandom, use time and pid as random seed");
4227 gettimeofday(&tv, 0);
4228 rand_seed[0] = getpid() ^ (tv.tv_sec & 0xFFFF);
4229 rand_seed[1] = getppid() ^ (tv.tv_usec & 0xFFFF);
4230 rand_seed[2] = (tv.tv_sec ^ tv.tv_usec) >> 16;
4232 rand_seed_initlized = 1;
4239 * Don't use nrand48, its range is [0,2^31) The highest bit will alwasy
4240 * be 0. Use jrand48 to include the highest bit.
4242 return (u32)jrand48(rand_seed);
4245 unsigned int rand_range(unsigned int upper)
4249 * Use the full 48bits to mod, which would be more uniformly
4252 return (unsigned int)(jrand48(rand_seed) % upper);