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;
253 memset(super_buf, 0, BTRFS_SUPER_INFO_SIZE);
254 cfg->num_bytes = round_down(cfg->num_bytes, cfg->sectorsize);
257 if (uuid_parse(cfg->fs_uuid, super->fsid) != 0) {
258 error("cound not parse UUID: %s", cfg->fs_uuid);
262 if (!test_uuid_unique(cfg->fs_uuid)) {
263 error("non-unique UUID: %s", cfg->fs_uuid);
268 uuid_generate(super->fsid);
269 uuid_unparse(super->fsid, cfg->fs_uuid);
271 uuid_generate(chunk_uuid);
272 uuid_unparse(chunk_uuid, cfg->chunk_uuid);
274 btrfs_set_super_bytenr(super, cfg->super_bytenr);
275 btrfs_set_super_num_devices(super, 1);
276 btrfs_set_super_magic(super, BTRFS_MAGIC_PARTIAL);
277 btrfs_set_super_generation(super, 1);
278 btrfs_set_super_root(super, root_bytenr);
279 btrfs_set_super_chunk_root(super, chunk_bytenr);
280 btrfs_set_super_total_bytes(super, cfg->num_bytes);
282 * Temporary filesystem will only have 6 tree roots:
283 * chunk tree, root tree, extent_tree, device tree, fs tree
286 btrfs_set_super_bytes_used(super, 6 * cfg->nodesize);
287 btrfs_set_super_sectorsize(super, cfg->sectorsize);
288 btrfs_set_super_leafsize(super, cfg->nodesize);
289 btrfs_set_super_nodesize(super, cfg->nodesize);
290 btrfs_set_super_stripesize(super, cfg->stripesize);
291 btrfs_set_super_csum_type(super, BTRFS_CSUM_TYPE_CRC32);
292 btrfs_set_super_chunk_root(super, chunk_bytenr);
293 btrfs_set_super_cache_generation(super, -1);
294 btrfs_set_super_incompat_flags(super, cfg->features);
296 __strncpy_null(super->label, cfg->label, BTRFS_LABEL_SIZE - 1);
298 /* Sys chunk array will be re-initialized at chunk tree init time */
299 super->sys_chunk_array_size = 0;
301 ret = write_temp_super(fd, super, cfg->super_bytenr);
307 * Setup an extent buffer for tree block.
309 static int setup_temp_extent_buffer(struct extent_buffer *buf,
310 struct btrfs_mkfs_config *cfg,
311 u64 bytenr, u64 owner)
313 unsigned char fsid[BTRFS_FSID_SIZE];
314 unsigned char chunk_uuid[BTRFS_UUID_SIZE];
317 ret = uuid_parse(cfg->fs_uuid, fsid);
320 ret = uuid_parse(cfg->chunk_uuid, chunk_uuid);
324 memset(buf->data, 0, cfg->nodesize);
325 buf->len = cfg->nodesize;
326 btrfs_set_header_bytenr(buf, bytenr);
327 btrfs_set_header_generation(buf, 1);
328 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
329 btrfs_set_header_owner(buf, owner);
330 btrfs_set_header_flags(buf, BTRFS_HEADER_FLAG_WRITTEN);
331 write_extent_buffer(buf, chunk_uuid, btrfs_header_chunk_tree_uuid(buf),
333 write_extent_buffer(buf, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
337 static inline int write_temp_extent_buffer(int fd, struct extent_buffer *buf,
342 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
344 /* Temporary extent buffer is always mapped 1:1 on disk */
345 ret = pwrite(fd, buf->data, buf->len, bytenr);
347 ret = (ret < 0 ? ret : -EIO);
354 * Insert a root item for temporary tree root
356 * Only used in make_btrfs_v2().
358 static void insert_temp_root_item(struct extent_buffer *buf,
359 struct btrfs_mkfs_config *cfg,
360 int *slot, u32 *itemoff, u64 objectid,
363 struct btrfs_root_item root_item;
364 struct btrfs_inode_item *inode_item;
365 struct btrfs_disk_key disk_key;
367 btrfs_set_header_nritems(buf, *slot + 1);
368 (*itemoff) -= sizeof(root_item);
369 memset(&root_item, 0, sizeof(root_item));
370 inode_item = &root_item.inode;
371 btrfs_set_stack_inode_generation(inode_item, 1);
372 btrfs_set_stack_inode_size(inode_item, 3);
373 btrfs_set_stack_inode_nlink(inode_item, 1);
374 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
375 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
376 btrfs_set_root_refs(&root_item, 1);
377 btrfs_set_root_used(&root_item, cfg->nodesize);
378 btrfs_set_root_generation(&root_item, 1);
379 btrfs_set_root_bytenr(&root_item, bytenr);
381 memset(&disk_key, 0, sizeof(disk_key));
382 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
383 btrfs_set_disk_key_objectid(&disk_key, objectid);
384 btrfs_set_disk_key_offset(&disk_key, 0);
386 btrfs_set_item_key(buf, &disk_key, *slot);
387 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
388 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(root_item));
389 write_extent_buffer(buf, &root_item,
390 btrfs_item_ptr_offset(buf, *slot),
395 static int setup_temp_root_tree(int fd, struct btrfs_mkfs_config *cfg,
396 u64 root_bytenr, u64 extent_bytenr,
397 u64 dev_bytenr, u64 fs_bytenr, u64 csum_bytenr)
399 struct extent_buffer *buf = NULL;
400 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
405 * Provided bytenr must in ascending order, or tree root will have a
408 BUG_ON(!(root_bytenr < extent_bytenr && extent_bytenr < dev_bytenr &&
409 dev_bytenr < fs_bytenr && fs_bytenr < csum_bytenr));
410 buf = malloc(sizeof(*buf) + cfg->nodesize);
414 ret = setup_temp_extent_buffer(buf, cfg, root_bytenr,
415 BTRFS_ROOT_TREE_OBJECTID);
419 insert_temp_root_item(buf, cfg, &slot, &itemoff,
420 BTRFS_EXTENT_TREE_OBJECTID, extent_bytenr);
421 insert_temp_root_item(buf, cfg, &slot, &itemoff,
422 BTRFS_DEV_TREE_OBJECTID, dev_bytenr);
423 insert_temp_root_item(buf, cfg, &slot, &itemoff,
424 BTRFS_FS_TREE_OBJECTID, fs_bytenr);
425 insert_temp_root_item(buf, cfg, &slot, &itemoff,
426 BTRFS_CSUM_TREE_OBJECTID, csum_bytenr);
428 ret = write_temp_extent_buffer(fd, buf, root_bytenr);
434 static int insert_temp_dev_item(int fd, struct extent_buffer *buf,
435 struct btrfs_mkfs_config *cfg,
436 int *slot, u32 *itemoff)
438 struct btrfs_disk_key disk_key;
439 struct btrfs_dev_item *dev_item;
440 char super_buf[BTRFS_SUPER_INFO_SIZE];
441 unsigned char dev_uuid[BTRFS_UUID_SIZE];
442 unsigned char fsid[BTRFS_FSID_SIZE];
443 struct btrfs_super_block *super = (struct btrfs_super_block *)super_buf;
446 ret = pread(fd, super_buf, BTRFS_SUPER_INFO_SIZE, cfg->super_bytenr);
447 if (ret < BTRFS_SUPER_INFO_SIZE) {
448 ret = (ret < 0 ? -errno : -EIO);
452 btrfs_set_header_nritems(buf, *slot + 1);
453 (*itemoff) -= sizeof(*dev_item);
454 /* setup device item 1, 0 is for replace case */
455 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
456 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
457 btrfs_set_disk_key_offset(&disk_key, 1);
458 btrfs_set_item_key(buf, &disk_key, *slot);
459 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
460 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(*dev_item));
462 dev_item = btrfs_item_ptr(buf, *slot, struct btrfs_dev_item);
463 /* Generate device uuid */
464 uuid_generate(dev_uuid);
465 write_extent_buffer(buf, dev_uuid,
466 (unsigned long)btrfs_device_uuid(dev_item),
468 uuid_parse(cfg->fs_uuid, fsid);
469 write_extent_buffer(buf, fsid,
470 (unsigned long)btrfs_device_fsid(dev_item),
472 btrfs_set_device_id(buf, dev_item, 1);
473 btrfs_set_device_generation(buf, dev_item, 0);
474 btrfs_set_device_total_bytes(buf, dev_item, cfg->num_bytes);
476 * The number must match the initial SYSTEM and META chunk size
478 btrfs_set_device_bytes_used(buf, dev_item,
479 BTRFS_MKFS_SYSTEM_GROUP_SIZE +
480 BTRFS_CONVERT_META_GROUP_SIZE);
481 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
482 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
483 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
484 btrfs_set_device_type(buf, dev_item, 0);
486 /* Super dev_item is not complete, copy the complete one to sb */
487 read_extent_buffer(buf, &super->dev_item, (unsigned long)dev_item,
489 ret = write_temp_super(fd, super, cfg->super_bytenr);
495 static int insert_temp_chunk_item(int fd, struct extent_buffer *buf,
496 struct btrfs_mkfs_config *cfg,
497 int *slot, u32 *itemoff, u64 start, u64 len,
500 struct btrfs_chunk *chunk;
501 struct btrfs_disk_key disk_key;
502 char super_buf[BTRFS_SUPER_INFO_SIZE];
503 struct btrfs_super_block *sb = (struct btrfs_super_block *)super_buf;
506 ret = pread(fd, super_buf, BTRFS_SUPER_INFO_SIZE,
508 if (ret < BTRFS_SUPER_INFO_SIZE) {
509 ret = (ret < 0 ? ret : -EIO);
513 btrfs_set_header_nritems(buf, *slot + 1);
514 (*itemoff) -= btrfs_chunk_item_size(1);
515 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
516 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
517 btrfs_set_disk_key_offset(&disk_key, start);
518 btrfs_set_item_key(buf, &disk_key, *slot);
519 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
520 btrfs_set_item_size(buf, btrfs_item_nr(*slot),
521 btrfs_chunk_item_size(1));
523 chunk = btrfs_item_ptr(buf, *slot, struct btrfs_chunk);
524 btrfs_set_chunk_length(buf, chunk, len);
525 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
526 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
527 btrfs_set_chunk_type(buf, chunk, type);
528 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
529 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
530 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
531 btrfs_set_chunk_num_stripes(buf, chunk, 1);
532 /* TODO: Support DUP profile for system chunk */
533 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
534 /* We are doing 1:1 mapping, so start is its dev offset */
535 btrfs_set_stripe_offset_nr(buf, chunk, 0, start);
536 write_extent_buffer(buf, &sb->dev_item.uuid,
537 (unsigned long)btrfs_stripe_dev_uuid_nr(chunk, 0),
542 * If it's system chunk, also copy it to super block.
544 if (type & BTRFS_BLOCK_GROUP_SYSTEM) {
547 cur = (char *)sb->sys_chunk_array + sb->sys_chunk_array_size;
548 memcpy(cur, &disk_key, sizeof(disk_key));
549 cur += sizeof(disk_key);
550 read_extent_buffer(buf, cur, (unsigned long int)chunk,
551 btrfs_chunk_item_size(1));
552 sb->sys_chunk_array_size += btrfs_chunk_item_size(1) +
555 ret = write_temp_super(fd, sb, cfg->super_bytenr);
560 static int setup_temp_chunk_tree(int fd, struct btrfs_mkfs_config *cfg,
561 u64 sys_chunk_start, u64 meta_chunk_start,
564 struct extent_buffer *buf = NULL;
565 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
569 /* Must ensure SYS chunk starts before META chunk */
570 BUG_ON(meta_chunk_start < sys_chunk_start);
571 buf = malloc(sizeof(*buf) + cfg->nodesize);
574 ret = setup_temp_extent_buffer(buf, cfg, chunk_bytenr,
575 BTRFS_CHUNK_TREE_OBJECTID);
579 ret = insert_temp_dev_item(fd, buf, cfg, &slot, &itemoff);
582 ret = insert_temp_chunk_item(fd, buf, cfg, &slot, &itemoff,
584 BTRFS_MKFS_SYSTEM_GROUP_SIZE,
585 BTRFS_BLOCK_GROUP_SYSTEM);
588 ret = insert_temp_chunk_item(fd, buf, cfg, &slot, &itemoff,
590 BTRFS_CONVERT_META_GROUP_SIZE,
591 BTRFS_BLOCK_GROUP_METADATA);
594 ret = write_temp_extent_buffer(fd, buf, chunk_bytenr);
601 static void insert_temp_dev_extent(struct extent_buffer *buf,
602 int *slot, u32 *itemoff, u64 start, u64 len)
604 struct btrfs_dev_extent *dev_extent;
605 struct btrfs_disk_key disk_key;
607 btrfs_set_header_nritems(buf, *slot + 1);
608 (*itemoff) -= sizeof(*dev_extent);
609 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
610 btrfs_set_disk_key_objectid(&disk_key, 1);
611 btrfs_set_disk_key_offset(&disk_key, start);
612 btrfs_set_item_key(buf, &disk_key, *slot);
613 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
614 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(*dev_extent));
616 dev_extent = btrfs_item_ptr(buf, *slot, struct btrfs_dev_extent);
617 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
618 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
619 btrfs_set_dev_extent_length(buf, dev_extent, len);
620 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, start);
621 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
622 BTRFS_CHUNK_TREE_OBJECTID);
626 static int setup_temp_dev_tree(int fd, struct btrfs_mkfs_config *cfg,
627 u64 sys_chunk_start, u64 meta_chunk_start,
630 struct extent_buffer *buf = NULL;
631 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
635 /* Must ensure SYS chunk starts before META chunk */
636 BUG_ON(meta_chunk_start < sys_chunk_start);
637 buf = malloc(sizeof(*buf) + cfg->nodesize);
640 ret = setup_temp_extent_buffer(buf, cfg, dev_bytenr,
641 BTRFS_DEV_TREE_OBJECTID);
644 insert_temp_dev_extent(buf, &slot, &itemoff, sys_chunk_start,
645 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
646 insert_temp_dev_extent(buf, &slot, &itemoff, meta_chunk_start,
647 BTRFS_CONVERT_META_GROUP_SIZE);
648 ret = write_temp_extent_buffer(fd, buf, dev_bytenr);
654 static int setup_temp_fs_tree(int fd, struct btrfs_mkfs_config *cfg,
657 struct extent_buffer *buf = NULL;
660 buf = malloc(sizeof(*buf) + cfg->nodesize);
663 ret = setup_temp_extent_buffer(buf, cfg, fs_bytenr,
664 BTRFS_FS_TREE_OBJECTID);
668 * Temporary fs tree is completely empty.
670 ret = write_temp_extent_buffer(fd, buf, fs_bytenr);
676 static int setup_temp_csum_tree(int fd, struct btrfs_mkfs_config *cfg,
679 struct extent_buffer *buf = NULL;
682 buf = malloc(sizeof(*buf) + cfg->nodesize);
685 ret = setup_temp_extent_buffer(buf, cfg, csum_bytenr,
686 BTRFS_CSUM_TREE_OBJECTID);
690 * Temporary csum tree is completely empty.
692 ret = write_temp_extent_buffer(fd, buf, csum_bytenr);
699 * Insert one temporary extent item.
701 * NOTE: if skinny_metadata is not enabled, this function must be called
702 * after all other trees are initialized.
703 * Or fs without skinny-metadata will be screwed up.
705 static int insert_temp_extent_item(int fd, struct extent_buffer *buf,
706 struct btrfs_mkfs_config *cfg,
707 int *slot, u32 *itemoff, u64 bytenr,
710 struct extent_buffer *tmp;
711 struct btrfs_extent_item *ei;
712 struct btrfs_extent_inline_ref *iref;
713 struct btrfs_disk_key disk_key;
714 struct btrfs_disk_key tree_info_key;
715 struct btrfs_tree_block_info *info;
717 int skinny_metadata = cfg->features &
718 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA;
722 itemsize = sizeof(*ei) + sizeof(*iref);
724 itemsize = sizeof(*ei) + sizeof(*iref) +
725 sizeof(struct btrfs_tree_block_info);
727 btrfs_set_header_nritems(buf, *slot + 1);
728 *(itemoff) -= itemsize;
730 if (skinny_metadata) {
731 btrfs_set_disk_key_type(&disk_key, BTRFS_METADATA_ITEM_KEY);
732 btrfs_set_disk_key_offset(&disk_key, 0);
734 btrfs_set_disk_key_type(&disk_key, BTRFS_EXTENT_ITEM_KEY);
735 btrfs_set_disk_key_offset(&disk_key, cfg->nodesize);
737 btrfs_set_disk_key_objectid(&disk_key, bytenr);
739 btrfs_set_item_key(buf, &disk_key, *slot);
740 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
741 btrfs_set_item_size(buf, btrfs_item_nr(*slot), itemsize);
743 ei = btrfs_item_ptr(buf, *slot, struct btrfs_extent_item);
744 btrfs_set_extent_refs(buf, ei, 1);
745 btrfs_set_extent_generation(buf, ei, 1);
746 btrfs_set_extent_flags(buf, ei, BTRFS_EXTENT_FLAG_TREE_BLOCK);
748 if (skinny_metadata) {
749 iref = (struct btrfs_extent_inline_ref *)(ei + 1);
751 info = (struct btrfs_tree_block_info *)(ei + 1);
752 iref = (struct btrfs_extent_inline_ref *)(info + 1);
754 btrfs_set_extent_inline_ref_type(buf, iref,
755 BTRFS_TREE_BLOCK_REF_KEY);
756 btrfs_set_extent_inline_ref_offset(buf, iref, ref_root);
763 * Lastly, check the tree block key by read the tree block
764 * Since we do 1:1 mapping for convert case, we can directly
765 * read the bytenr from disk
767 tmp = malloc(sizeof(*tmp) + cfg->nodesize);
770 ret = setup_temp_extent_buffer(tmp, cfg, bytenr, ref_root);
773 ret = pread(fd, tmp->data, cfg->nodesize, bytenr);
774 if (ret < cfg->nodesize) {
775 ret = (ret < 0 ? -errno : -EIO);
778 if (btrfs_header_nritems(tmp) == 0) {
779 btrfs_set_disk_key_type(&tree_info_key, 0);
780 btrfs_set_disk_key_objectid(&tree_info_key, 0);
781 btrfs_set_disk_key_offset(&tree_info_key, 0);
783 btrfs_item_key(tmp, &tree_info_key, 0);
785 btrfs_set_tree_block_key(buf, info, &tree_info_key);
792 static void insert_temp_block_group(struct extent_buffer *buf,
793 struct btrfs_mkfs_config *cfg,
794 int *slot, u32 *itemoff,
795 u64 bytenr, u64 len, u64 used, u64 flag)
797 struct btrfs_block_group_item bgi;
798 struct btrfs_disk_key disk_key;
800 btrfs_set_header_nritems(buf, *slot + 1);
801 (*itemoff) -= sizeof(bgi);
802 btrfs_set_disk_key_type(&disk_key, BTRFS_BLOCK_GROUP_ITEM_KEY);
803 btrfs_set_disk_key_objectid(&disk_key, bytenr);
804 btrfs_set_disk_key_offset(&disk_key, len);
805 btrfs_set_item_key(buf, &disk_key, *slot);
806 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
807 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(bgi));
809 btrfs_set_block_group_flags(&bgi, flag);
810 btrfs_set_block_group_used(&bgi, used);
811 btrfs_set_block_group_chunk_objectid(&bgi,
812 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
813 write_extent_buffer(buf, &bgi, btrfs_item_ptr_offset(buf, *slot),
818 static int setup_temp_extent_tree(int fd, struct btrfs_mkfs_config *cfg,
819 u64 chunk_bytenr, u64 root_bytenr,
820 u64 extent_bytenr, u64 dev_bytenr,
821 u64 fs_bytenr, u64 csum_bytenr)
823 struct extent_buffer *buf = NULL;
824 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
829 * We must ensure provided bytenr are in ascending order,
830 * or extent tree key order will be broken.
832 BUG_ON(!(chunk_bytenr < root_bytenr && root_bytenr < extent_bytenr &&
833 extent_bytenr < dev_bytenr && dev_bytenr < fs_bytenr &&
834 fs_bytenr < csum_bytenr));
835 buf = malloc(sizeof(*buf) + cfg->nodesize);
839 ret = setup_temp_extent_buffer(buf, cfg, extent_bytenr,
840 BTRFS_EXTENT_TREE_OBJECTID);
844 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
845 chunk_bytenr, BTRFS_CHUNK_TREE_OBJECTID);
849 insert_temp_block_group(buf, cfg, &slot, &itemoff, chunk_bytenr,
850 BTRFS_MKFS_SYSTEM_GROUP_SIZE, cfg->nodesize,
851 BTRFS_BLOCK_GROUP_SYSTEM);
853 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
854 root_bytenr, BTRFS_ROOT_TREE_OBJECTID);
858 /* 5 tree block used, root, extent, dev, fs and csum*/
859 insert_temp_block_group(buf, cfg, &slot, &itemoff, root_bytenr,
860 BTRFS_CONVERT_META_GROUP_SIZE, cfg->nodesize * 5,
861 BTRFS_BLOCK_GROUP_METADATA);
863 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
864 extent_bytenr, BTRFS_EXTENT_TREE_OBJECTID);
867 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
868 dev_bytenr, BTRFS_DEV_TREE_OBJECTID);
871 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
872 fs_bytenr, BTRFS_FS_TREE_OBJECTID);
875 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
876 csum_bytenr, BTRFS_CSUM_TREE_OBJECTID);
880 ret = write_temp_extent_buffer(fd, buf, extent_bytenr);
887 * Improved version of make_btrfs().
890 * 1) Do chunk allocation to avoid used data
891 * And after this function, extent type matches chunk type
892 * 2) Better structured code
893 * No super long hand written codes to initialized all tree blocks
894 * Split into small blocks and reuse codes.
895 * TODO: Reuse tree operation facilities by introducing new flags
897 static int make_convert_btrfs(int fd, struct btrfs_mkfs_config *cfg,
898 struct btrfs_convert_context *cctx)
900 struct cache_tree *free = &cctx->free;
901 struct cache_tree *used = &cctx->used;
903 u64 meta_chunk_start;
904 /* chunk tree bytenr, in system chunk */
906 /* metadata trees bytenr, in metadata chunk */
914 /* Shouldn't happen */
915 BUG_ON(cache_tree_empty(used));
918 * reserve space for temporary superblock first
919 * Here we allocate a little larger space, to keep later
920 * free space will be STRIPE_LEN aligned
922 ret = reserve_free_space(free, BTRFS_STRIPE_LEN,
928 * Then reserve system chunk space
929 * TODO: Change system group size depending on cctx->total_bytes.
930 * If using current 4M, it can only handle less than one TB for
931 * worst case and then run out of sys space.
933 ret = reserve_free_space(free, BTRFS_MKFS_SYSTEM_GROUP_SIZE,
937 ret = reserve_free_space(free, BTRFS_CONVERT_META_GROUP_SIZE,
943 * Allocated meta/sys chunks will be mapped 1:1 with device offset.
945 * Inside the allocated metadata chunk, the layout will be:
946 * | offset | contents |
947 * -------------------------------------
949 * | +nodesize | extent root |
950 * | +nodesize * 2 | device root |
951 * | +nodesize * 3 | fs tree |
952 * | +nodesize * 4 | csum tree |
953 * -------------------------------------
954 * Inside the allocated system chunk, the layout will be:
955 * | offset | contents |
956 * -------------------------------------
957 * | +0 | chunk root |
958 * -------------------------------------
960 chunk_bytenr = sys_chunk_start;
961 root_bytenr = meta_chunk_start;
962 extent_bytenr = meta_chunk_start + cfg->nodesize;
963 dev_bytenr = meta_chunk_start + cfg->nodesize * 2;
964 fs_bytenr = meta_chunk_start + cfg->nodesize * 3;
965 csum_bytenr = meta_chunk_start + cfg->nodesize * 4;
967 ret = setup_temp_super(fd, cfg, root_bytenr, chunk_bytenr);
971 ret = setup_temp_root_tree(fd, cfg, root_bytenr, extent_bytenr,
972 dev_bytenr, fs_bytenr, csum_bytenr);
975 ret = setup_temp_chunk_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
979 ret = setup_temp_dev_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
983 ret = setup_temp_fs_tree(fd, cfg, fs_bytenr);
986 ret = setup_temp_csum_tree(fd, cfg, csum_bytenr);
990 * Setup extent tree last, since it may need to read tree block key
991 * for non-skinny metadata case.
993 ret = setup_temp_extent_tree(fd, cfg, chunk_bytenr, root_bytenr,
994 extent_bytenr, dev_bytenr, fs_bytenr,
1001 * @fs_uuid - if NULL, generates a UUID, returns back the new filesystem UUID
1003 * The superblock signature is not valid, denotes a partially created
1004 * filesystem, needs to be finalized.
1006 int make_btrfs(int fd, struct btrfs_mkfs_config *cfg,
1007 struct btrfs_convert_context *cctx)
1009 struct btrfs_super_block super;
1010 struct extent_buffer *buf;
1011 struct btrfs_root_item root_item;
1012 struct btrfs_disk_key disk_key;
1013 struct btrfs_extent_item *extent_item;
1014 struct btrfs_inode_item *inode_item;
1015 struct btrfs_chunk *chunk;
1016 struct btrfs_dev_item *dev_item;
1017 struct btrfs_dev_extent *dev_extent;
1018 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
1028 int skinny_metadata = !!(cfg->features &
1029 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA);
1033 return make_convert_btrfs(fd, cfg, cctx);
1034 buf = malloc(sizeof(*buf) + max(cfg->sectorsize, cfg->nodesize));
1038 first_free = BTRFS_SUPER_INFO_OFFSET + cfg->sectorsize * 2 - 1;
1039 first_free &= ~((u64)cfg->sectorsize - 1);
1041 memset(&super, 0, sizeof(super));
1043 num_bytes = (cfg->num_bytes / cfg->sectorsize) * cfg->sectorsize;
1044 if (*cfg->fs_uuid) {
1045 if (uuid_parse(cfg->fs_uuid, super.fsid) != 0) {
1046 error("cannot not parse UUID: %s", cfg->fs_uuid);
1050 if (!test_uuid_unique(cfg->fs_uuid)) {
1051 error("non-unique UUID: %s", cfg->fs_uuid);
1056 uuid_generate(super.fsid);
1058 uuid_unparse(super.fsid, cfg->fs_uuid);
1060 uuid_generate(super.dev_item.uuid);
1061 uuid_generate(chunk_tree_uuid);
1063 btrfs_set_super_bytenr(&super, cfg->blocks[0]);
1064 btrfs_set_super_num_devices(&super, 1);
1065 btrfs_set_super_magic(&super, BTRFS_MAGIC_PARTIAL);
1066 btrfs_set_super_generation(&super, 1);
1067 btrfs_set_super_root(&super, cfg->blocks[1]);
1068 btrfs_set_super_chunk_root(&super, cfg->blocks[3]);
1069 btrfs_set_super_total_bytes(&super, num_bytes);
1070 btrfs_set_super_bytes_used(&super, 6 * cfg->nodesize);
1071 btrfs_set_super_sectorsize(&super, cfg->sectorsize);
1072 btrfs_set_super_leafsize(&super, cfg->nodesize);
1073 btrfs_set_super_nodesize(&super, cfg->nodesize);
1074 btrfs_set_super_stripesize(&super, cfg->stripesize);
1075 btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
1076 btrfs_set_super_chunk_root_generation(&super, 1);
1077 btrfs_set_super_cache_generation(&super, -1);
1078 btrfs_set_super_incompat_flags(&super, cfg->features);
1080 __strncpy_null(super.label, cfg->label, BTRFS_LABEL_SIZE - 1);
1082 /* create the tree of root objects */
1083 memset(buf->data, 0, cfg->nodesize);
1084 buf->len = cfg->nodesize;
1085 btrfs_set_header_bytenr(buf, cfg->blocks[1]);
1086 btrfs_set_header_nritems(buf, 4);
1087 btrfs_set_header_generation(buf, 1);
1088 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
1089 btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
1090 write_extent_buffer(buf, super.fsid, btrfs_header_fsid(),
1093 write_extent_buffer(buf, chunk_tree_uuid,
1094 btrfs_header_chunk_tree_uuid(buf),
1097 /* create the items for the root tree */
1098 memset(&root_item, 0, sizeof(root_item));
1099 inode_item = &root_item.inode;
1100 btrfs_set_stack_inode_generation(inode_item, 1);
1101 btrfs_set_stack_inode_size(inode_item, 3);
1102 btrfs_set_stack_inode_nlink(inode_item, 1);
1103 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
1104 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
1105 btrfs_set_root_refs(&root_item, 1);
1106 btrfs_set_root_used(&root_item, cfg->nodesize);
1107 btrfs_set_root_generation(&root_item, 1);
1109 memset(&disk_key, 0, sizeof(disk_key));
1110 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
1111 btrfs_set_disk_key_offset(&disk_key, 0);
1114 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - sizeof(root_item);
1115 btrfs_set_root_bytenr(&root_item, cfg->blocks[2]);
1116 btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
1117 btrfs_set_item_key(buf, &disk_key, nritems);
1118 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1119 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1121 write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
1122 nritems), sizeof(root_item));
1125 itemoff = itemoff - sizeof(root_item);
1126 btrfs_set_root_bytenr(&root_item, cfg->blocks[4]);
1127 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
1128 btrfs_set_item_key(buf, &disk_key, nritems);
1129 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1130 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1132 write_extent_buffer(buf, &root_item,
1133 btrfs_item_ptr_offset(buf, nritems),
1137 itemoff = itemoff - sizeof(root_item);
1138 btrfs_set_root_bytenr(&root_item, cfg->blocks[5]);
1139 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
1140 btrfs_set_item_key(buf, &disk_key, nritems);
1141 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1142 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1144 write_extent_buffer(buf, &root_item,
1145 btrfs_item_ptr_offset(buf, nritems),
1149 itemoff = itemoff - sizeof(root_item);
1150 btrfs_set_root_bytenr(&root_item, cfg->blocks[6]);
1151 btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_TREE_OBJECTID);
1152 btrfs_set_item_key(buf, &disk_key, nritems);
1153 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1154 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1156 write_extent_buffer(buf, &root_item,
1157 btrfs_item_ptr_offset(buf, nritems),
1162 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1163 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[1]);
1164 if (ret != cfg->nodesize) {
1165 ret = (ret < 0 ? -errno : -EIO);
1169 /* create the items for the extent tree */
1170 memset(buf->data + sizeof(struct btrfs_header), 0,
1171 cfg->nodesize - sizeof(struct btrfs_header));
1173 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
1174 for (i = 1; i < 7; i++) {
1175 item_size = sizeof(struct btrfs_extent_item);
1176 if (!skinny_metadata)
1177 item_size += sizeof(struct btrfs_tree_block_info);
1179 BUG_ON(cfg->blocks[i] < first_free);
1180 BUG_ON(cfg->blocks[i] < cfg->blocks[i - 1]);
1182 /* create extent item */
1183 itemoff -= item_size;
1184 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
1185 if (skinny_metadata) {
1186 btrfs_set_disk_key_type(&disk_key,
1187 BTRFS_METADATA_ITEM_KEY);
1188 btrfs_set_disk_key_offset(&disk_key, 0);
1190 btrfs_set_disk_key_type(&disk_key,
1191 BTRFS_EXTENT_ITEM_KEY);
1192 btrfs_set_disk_key_offset(&disk_key, cfg->nodesize);
1194 btrfs_set_item_key(buf, &disk_key, nritems);
1195 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
1197 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1199 extent_item = btrfs_item_ptr(buf, nritems,
1200 struct btrfs_extent_item);
1201 btrfs_set_extent_refs(buf, extent_item, 1);
1202 btrfs_set_extent_generation(buf, extent_item, 1);
1203 btrfs_set_extent_flags(buf, extent_item,
1204 BTRFS_EXTENT_FLAG_TREE_BLOCK);
1207 /* create extent ref */
1208 ref_root = reference_root_table[i];
1209 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
1210 btrfs_set_disk_key_offset(&disk_key, ref_root);
1211 btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
1212 btrfs_set_item_key(buf, &disk_key, nritems);
1213 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
1215 btrfs_set_item_size(buf, btrfs_item_nr(nritems), 0);
1218 btrfs_set_header_bytenr(buf, cfg->blocks[2]);
1219 btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
1220 btrfs_set_header_nritems(buf, nritems);
1221 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1222 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[2]);
1223 if (ret != cfg->nodesize) {
1224 ret = (ret < 0 ? -errno : -EIO);
1228 /* create the chunk tree */
1229 memset(buf->data + sizeof(struct btrfs_header), 0,
1230 cfg->nodesize - sizeof(struct btrfs_header));
1232 item_size = sizeof(*dev_item);
1233 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - item_size;
1235 /* first device 1 (there is no device 0) */
1236 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
1237 btrfs_set_disk_key_offset(&disk_key, 1);
1238 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
1239 btrfs_set_item_key(buf, &disk_key, nritems);
1240 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1241 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
1243 dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
1244 btrfs_set_device_id(buf, dev_item, 1);
1245 btrfs_set_device_generation(buf, dev_item, 0);
1246 btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
1247 btrfs_set_device_bytes_used(buf, dev_item,
1248 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1249 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
1250 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
1251 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
1252 btrfs_set_device_type(buf, dev_item, 0);
1254 write_extent_buffer(buf, super.dev_item.uuid,
1255 (unsigned long)btrfs_device_uuid(dev_item),
1257 write_extent_buffer(buf, super.fsid,
1258 (unsigned long)btrfs_device_fsid(dev_item),
1260 read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
1264 item_size = btrfs_chunk_item_size(1);
1265 itemoff = itemoff - item_size;
1267 /* then we have chunk 0 */
1268 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1269 btrfs_set_disk_key_offset(&disk_key, 0);
1270 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
1271 btrfs_set_item_key(buf, &disk_key, nritems);
1272 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1273 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
1275 chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
1276 btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1277 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
1278 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
1279 btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
1280 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
1281 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
1282 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
1283 btrfs_set_chunk_num_stripes(buf, chunk, 1);
1284 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
1285 btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
1288 write_extent_buffer(buf, super.dev_item.uuid,
1289 (unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
1292 /* copy the key for the chunk to the system array */
1293 ptr = super.sys_chunk_array;
1294 array_size = sizeof(disk_key);
1296 memcpy(ptr, &disk_key, sizeof(disk_key));
1297 ptr += sizeof(disk_key);
1299 /* copy the chunk to the system array */
1300 read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
1301 array_size += item_size;
1303 btrfs_set_super_sys_array_size(&super, array_size);
1305 btrfs_set_header_bytenr(buf, cfg->blocks[3]);
1306 btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
1307 btrfs_set_header_nritems(buf, nritems);
1308 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1309 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[3]);
1310 if (ret != cfg->nodesize) {
1311 ret = (ret < 0 ? -errno : -EIO);
1315 /* create the device tree */
1316 memset(buf->data + sizeof(struct btrfs_header), 0,
1317 cfg->nodesize - sizeof(struct btrfs_header));
1319 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) -
1320 sizeof(struct btrfs_dev_extent);
1322 btrfs_set_disk_key_objectid(&disk_key, 1);
1323 btrfs_set_disk_key_offset(&disk_key, 0);
1324 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
1325 btrfs_set_item_key(buf, &disk_key, nritems);
1326 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1327 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1328 sizeof(struct btrfs_dev_extent));
1329 dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
1330 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
1331 BTRFS_CHUNK_TREE_OBJECTID);
1332 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
1333 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1334 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
1336 write_extent_buffer(buf, chunk_tree_uuid,
1337 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
1340 btrfs_set_dev_extent_length(buf, dev_extent,
1341 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1344 btrfs_set_header_bytenr(buf, cfg->blocks[4]);
1345 btrfs_set_header_owner(buf, BTRFS_DEV_TREE_OBJECTID);
1346 btrfs_set_header_nritems(buf, nritems);
1347 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1348 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[4]);
1349 if (ret != cfg->nodesize) {
1350 ret = (ret < 0 ? -errno : -EIO);
1354 /* create the FS root */
1355 memset(buf->data + sizeof(struct btrfs_header), 0,
1356 cfg->nodesize - sizeof(struct btrfs_header));
1357 btrfs_set_header_bytenr(buf, cfg->blocks[5]);
1358 btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
1359 btrfs_set_header_nritems(buf, 0);
1360 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1361 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[5]);
1362 if (ret != cfg->nodesize) {
1363 ret = (ret < 0 ? -errno : -EIO);
1366 /* finally create the csum root */
1367 memset(buf->data + sizeof(struct btrfs_header), 0,
1368 cfg->nodesize - sizeof(struct btrfs_header));
1369 btrfs_set_header_bytenr(buf, cfg->blocks[6]);
1370 btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
1371 btrfs_set_header_nritems(buf, 0);
1372 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1373 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[6]);
1374 if (ret != cfg->nodesize) {
1375 ret = (ret < 0 ? -errno : -EIO);
1379 /* and write out the super block */
1380 BUG_ON(sizeof(super) > cfg->sectorsize);
1381 memset(buf->data, 0, BTRFS_SUPER_INFO_SIZE);
1382 memcpy(buf->data, &super, sizeof(super));
1383 buf->len = BTRFS_SUPER_INFO_SIZE;
1384 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1385 ret = pwrite(fd, buf->data, BTRFS_SUPER_INFO_SIZE, cfg->blocks[0]);
1386 if (ret != BTRFS_SUPER_INFO_SIZE) {
1387 ret = (ret < 0 ? -errno : -EIO);
1398 static const struct btrfs_fs_feature {
1402 } mkfs_features[] = {
1403 { "mixed-bg", BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS,
1404 "mixed data and metadata block groups" },
1405 { "extref", BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF,
1406 "increased hardlink limit per file to 65536" },
1407 { "raid56", BTRFS_FEATURE_INCOMPAT_RAID56,
1408 "raid56 extended format" },
1409 { "skinny-metadata", BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA,
1410 "reduced-size metadata extent refs" },
1411 { "no-holes", BTRFS_FEATURE_INCOMPAT_NO_HOLES,
1412 "no explicit hole extents for files" },
1413 /* Keep this one last */
1414 { "list-all", BTRFS_FEATURE_LIST_ALL, NULL }
1417 static int parse_one_fs_feature(const char *name, u64 *flags)
1422 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1423 if (name[0] == '^' &&
1424 !strcmp(mkfs_features[i].name, name + 1)) {
1425 *flags &= ~ mkfs_features[i].flag;
1427 } else if (!strcmp(mkfs_features[i].name, name)) {
1428 *flags |= mkfs_features[i].flag;
1436 void btrfs_parse_features_to_string(char *buf, u64 flags)
1442 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1443 if (flags & mkfs_features[i].flag) {
1446 strcat(buf, mkfs_features[i].name);
1451 void btrfs_process_fs_features(u64 flags)
1455 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1456 if (flags & mkfs_features[i].flag) {
1457 printf("Turning ON incompat feature '%s': %s\n",
1458 mkfs_features[i].name,
1459 mkfs_features[i].desc);
1464 void btrfs_list_all_fs_features(u64 mask_disallowed)
1468 fprintf(stderr, "Filesystem features available:\n");
1469 for (i = 0; i < ARRAY_SIZE(mkfs_features) - 1; i++) {
1470 char *is_default = "";
1472 if (mkfs_features[i].flag & mask_disallowed)
1474 if (mkfs_features[i].flag & BTRFS_MKFS_DEFAULT_FEATURES)
1475 is_default = ", default";
1476 fprintf(stderr, "%-20s- %s (0x%llx%s)\n",
1477 mkfs_features[i].name,
1478 mkfs_features[i].desc,
1479 mkfs_features[i].flag,
1485 * Return NULL if all features were parsed fine, otherwise return the name of
1486 * the first unparsed.
1488 char* btrfs_parse_fs_features(char *namelist, u64 *flags)
1491 char *save_ptr = NULL; /* Satisfy static checkers */
1493 for (this_char = strtok_r(namelist, ",", &save_ptr);
1495 this_char = strtok_r(NULL, ",", &save_ptr)) {
1496 if (parse_one_fs_feature(this_char, flags))
1503 u64 btrfs_device_size(int fd, struct stat *st)
1506 if (S_ISREG(st->st_mode)) {
1509 if (!S_ISBLK(st->st_mode)) {
1512 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
1518 static int zero_blocks(int fd, off_t start, size_t len)
1520 char *buf = malloc(len);
1526 memset(buf, 0, len);
1527 written = pwrite(fd, buf, len, start);
1534 #define ZERO_DEV_BYTES (2 * 1024 * 1024)
1536 /* don't write outside the device by clamping the region to the device size */
1537 static int zero_dev_clamped(int fd, off_t start, ssize_t len, u64 dev_size)
1539 off_t end = max(start, start + len);
1542 /* and don't overwrite the disk labels on sparc */
1543 start = max(start, 1024);
1544 end = max(end, 1024);
1547 start = min_t(u64, start, dev_size);
1548 end = min_t(u64, end, dev_size);
1550 return zero_blocks(fd, start, end - start);
1553 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
1554 struct btrfs_root *root, int fd, char *path,
1555 u64 device_total_bytes, u32 io_width, u32 io_align,
1558 struct btrfs_super_block *disk_super;
1559 struct btrfs_super_block *super = root->fs_info->super_copy;
1560 struct btrfs_device *device;
1561 struct btrfs_dev_item *dev_item;
1567 device_total_bytes = (device_total_bytes / sectorsize) * sectorsize;
1569 device = kzalloc(sizeof(*device), GFP_NOFS);
1572 buf = kzalloc(sectorsize, GFP_NOFS);
1575 BUG_ON(sizeof(*disk_super) > sectorsize);
1577 disk_super = (struct btrfs_super_block *)buf;
1578 dev_item = &disk_super->dev_item;
1580 uuid_generate(device->uuid);
1583 device->io_width = io_width;
1584 device->io_align = io_align;
1585 device->sector_size = sectorsize;
1587 device->writeable = 1;
1588 device->total_bytes = device_total_bytes;
1589 device->bytes_used = 0;
1590 device->total_ios = 0;
1591 device->dev_root = root->fs_info->dev_root;
1592 device->name = strdup(path);
1596 INIT_LIST_HEAD(&device->dev_list);
1597 ret = btrfs_add_device(trans, root, device);
1600 fs_total_bytes = btrfs_super_total_bytes(super) + device_total_bytes;
1601 btrfs_set_super_total_bytes(super, fs_total_bytes);
1603 num_devs = btrfs_super_num_devices(super) + 1;
1604 btrfs_set_super_num_devices(super, num_devs);
1606 memcpy(disk_super, super, sizeof(*disk_super));
1608 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
1609 btrfs_set_stack_device_id(dev_item, device->devid);
1610 btrfs_set_stack_device_type(dev_item, device->type);
1611 btrfs_set_stack_device_io_align(dev_item, device->io_align);
1612 btrfs_set_stack_device_io_width(dev_item, device->io_width);
1613 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
1614 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
1615 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
1616 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
1618 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
1619 BUG_ON(ret != sectorsize);
1622 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
1623 device->fs_devices = root->fs_info->fs_devices;
1632 static int btrfs_wipe_existing_sb(int fd)
1634 const char *off = NULL;
1639 blkid_probe pr = NULL;
1641 pr = blkid_new_probe();
1645 if (blkid_probe_set_device(pr, fd, 0, 0)) {
1650 ret = blkid_probe_lookup_value(pr, "SBMAGIC_OFFSET", &off, NULL);
1652 ret = blkid_probe_lookup_value(pr, "SBMAGIC", NULL, &len);
1654 if (ret || len == 0 || off == NULL) {
1656 * If lookup fails, the probe did not find any values, eg. for
1657 * a file image or a loop device. Soft error.
1663 offset = strtoll(off, NULL, 10);
1664 if (len > sizeof(buf))
1667 memset(buf, 0, len);
1668 ret = pwrite(fd, buf, len, offset);
1670 error("cannot wipe existing superblock: %s", strerror(errno));
1672 } else if (ret != len) {
1673 error("cannot wipe existing superblock: wrote %d of %zd", ret, len);
1679 blkid_free_probe(pr);
1683 int btrfs_prepare_device(int fd, const char *file, u64 *block_count_ret,
1684 u64 max_block_count, unsigned opflags)
1690 ret = fstat(fd, &st);
1692 error("unable to stat %s: %s", file, strerror(errno));
1696 block_count = btrfs_device_size(fd, &st);
1697 if (block_count == 0) {
1698 error("unable to determine size of %s", file);
1701 if (max_block_count)
1702 block_count = min(block_count, max_block_count);
1704 if (opflags & PREP_DEVICE_DISCARD) {
1706 * We intentionally ignore errors from the discard ioctl. It
1707 * is not necessary for the mkfs functionality but just an
1710 if (discard_range(fd, 0, 0) == 0) {
1711 if (opflags & PREP_DEVICE_VERBOSE)
1712 printf("Performing full device TRIM (%s) ...\n",
1713 pretty_size(block_count));
1714 discard_blocks(fd, 0, block_count);
1718 ret = zero_dev_clamped(fd, 0, ZERO_DEV_BYTES, block_count);
1719 for (i = 0 ; !ret && i < BTRFS_SUPER_MIRROR_MAX; i++)
1720 ret = zero_dev_clamped(fd, btrfs_sb_offset(i),
1721 BTRFS_SUPER_INFO_SIZE, block_count);
1722 if (!ret && (opflags & PREP_DEVICE_ZERO_END))
1723 ret = zero_dev_clamped(fd, block_count - ZERO_DEV_BYTES,
1724 ZERO_DEV_BYTES, block_count);
1727 error("failed to zero device '%s': %s", file, strerror(-ret));
1731 ret = btrfs_wipe_existing_sb(fd);
1733 error("cannot wipe superblocks on %s", file);
1737 *block_count_ret = block_count;
1741 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
1742 struct btrfs_root *root, u64 objectid)
1745 struct btrfs_inode_item inode_item;
1746 time_t now = time(NULL);
1748 memset(&inode_item, 0, sizeof(inode_item));
1749 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
1750 btrfs_set_stack_inode_size(&inode_item, 0);
1751 btrfs_set_stack_inode_nlink(&inode_item, 1);
1752 btrfs_set_stack_inode_nbytes(&inode_item, root->nodesize);
1753 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
1754 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
1755 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
1756 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
1757 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
1758 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
1759 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
1760 btrfs_set_stack_timespec_sec(&inode_item.otime, 0);
1761 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
1763 if (root->fs_info->tree_root == root)
1764 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
1766 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
1770 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
1774 btrfs_set_root_dirid(&root->root_item, objectid);
1781 * checks if a path is a block device node
1782 * Returns negative errno on failure, otherwise
1783 * returns 1 for blockdev, 0 for not-blockdev
1785 int is_block_device(const char *path)
1787 struct stat statbuf;
1789 if (stat(path, &statbuf) < 0)
1792 return !!S_ISBLK(statbuf.st_mode);
1796 * check if given path is a mount point
1797 * return 1 if yes. 0 if no. -1 for error
1799 int is_mount_point(const char *path)
1805 f = setmntent("/proc/self/mounts", "r");
1809 while ((mnt = getmntent(f)) != NULL) {
1810 if (strcmp(mnt->mnt_dir, path))
1819 static int is_reg_file(const char *path)
1821 struct stat statbuf;
1823 if (stat(path, &statbuf) < 0)
1825 return S_ISREG(statbuf.st_mode);
1829 * This function checks if the given input parameter is
1831 * return <0 : some error in the given input
1832 * return BTRFS_ARG_UNKNOWN: unknown input
1833 * return BTRFS_ARG_UUID: given input is uuid
1834 * return BTRFS_ARG_MNTPOINT: given input is path
1835 * return BTRFS_ARG_REG: given input is regular file
1836 * return BTRFS_ARG_BLKDEV: given input is block device
1838 int check_arg_type(const char *input)
1841 char path[PATH_MAX];
1846 if (realpath(input, path)) {
1847 if (is_block_device(path) == 1)
1848 return BTRFS_ARG_BLKDEV;
1850 if (is_mount_point(path) == 1)
1851 return BTRFS_ARG_MNTPOINT;
1853 if (is_reg_file(path))
1854 return BTRFS_ARG_REG;
1856 return BTRFS_ARG_UNKNOWN;
1859 if (strlen(input) == (BTRFS_UUID_UNPARSED_SIZE - 1) &&
1860 !uuid_parse(input, uuid))
1861 return BTRFS_ARG_UUID;
1863 return BTRFS_ARG_UNKNOWN;
1867 * Find the mount point for a mounted device.
1868 * On success, returns 0 with mountpoint in *mp.
1869 * On failure, returns -errno (not mounted yields -EINVAL)
1870 * Is noisy on failures, expects to be given a mounted device.
1872 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size)
1877 ret = is_block_device(dev);
1880 error("not a block device: %s", dev);
1883 error("cannot check %s: %s", dev, strerror(-ret));
1888 fd = open(dev, O_RDONLY);
1891 error("cannot open %s: %s", dev, strerror(errno));
1895 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
1898 } else { /* mounted, all good */
1908 * Given a pathname, return a filehandle to:
1909 * the original pathname or,
1910 * if the pathname is a mounted btrfs device, to its mountpoint.
1912 * On error, return -1, errno should be set.
1914 int open_path_or_dev_mnt(const char *path, DIR **dirstream, int verbose)
1919 if (is_block_device(path)) {
1920 ret = get_btrfs_mount(path, mp, sizeof(mp));
1922 /* not a mounted btrfs dev */
1923 error_on(verbose, "'%s' is not a mounted btrfs device",
1928 ret = open_file_or_dir(mp, dirstream);
1929 error_on(verbose && ret < 0, "can't access '%s': %s",
1930 path, strerror(errno));
1932 ret = btrfs_open_dir(path, dirstream, 1);
1939 * Do the following checks before calling open_file_or_dir():
1940 * 1: path is in a btrfs filesystem
1941 * 2: path is a directory
1943 int btrfs_open_dir(const char *path, DIR **dirstream, int verbose)
1949 if (statfs(path, &stfs) != 0) {
1950 error_on(verbose, "cannot access '%s': %s", path,
1955 if (stfs.f_type != BTRFS_SUPER_MAGIC) {
1956 error_on(verbose, "not a btrfs filesystem: %s", path);
1960 if (stat(path, &st) != 0) {
1961 error_on(verbose, "cannot access '%s': %s", path,
1966 if (!S_ISDIR(st.st_mode)) {
1967 error_on(verbose, "not a directory: %s", path);
1971 ret = open_file_or_dir(path, dirstream);
1973 error_on(verbose, "cannot access '%s': %s", path,
1980 /* checks if a device is a loop device */
1981 static int is_loop_device (const char* device) {
1982 struct stat statbuf;
1984 if(stat(device, &statbuf) < 0)
1987 return (S_ISBLK(statbuf.st_mode) &&
1988 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
1992 * Takes a loop device path (e.g. /dev/loop0) and returns
1993 * the associated file (e.g. /images/my_btrfs.img) using
1996 static int resolve_loop_device_with_loopdev(const char* loop_dev, char* loop_file)
2000 struct loop_info64 lo64;
2002 fd = open(loop_dev, O_RDONLY | O_NONBLOCK);
2005 ret = ioctl(fd, LOOP_GET_STATUS64, &lo64);
2011 memcpy(loop_file, lo64.lo_file_name, sizeof(lo64.lo_file_name));
2012 loop_file[sizeof(lo64.lo_file_name)] = 0;
2020 /* Takes a loop device path (e.g. /dev/loop0) and returns
2021 * the associated file (e.g. /images/my_btrfs.img) */
2022 static int resolve_loop_device(const char* loop_dev, char* loop_file,
2029 char real_loop_dev[PATH_MAX];
2031 if (!realpath(loop_dev, real_loop_dev))
2033 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
2034 if (!(f = fopen(p, "r"))) {
2035 if (errno == ENOENT)
2037 * It's possibly a partitioned loop device, which is
2038 * resolvable with loopdev API.
2040 return resolve_loop_device_with_loopdev(loop_dev, loop_file);
2044 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
2045 ret = fscanf(f, fmt, loop_file);
2054 * Checks whether a and b are identical or device
2055 * files associated with the same block device
2057 static int is_same_blk_file(const char* a, const char* b)
2059 struct stat st_buf_a, st_buf_b;
2060 char real_a[PATH_MAX];
2061 char real_b[PATH_MAX];
2063 if (!realpath(a, real_a))
2064 strncpy_null(real_a, a);
2066 if (!realpath(b, real_b))
2067 strncpy_null(real_b, b);
2069 /* Identical path? */
2070 if (strcmp(real_a, real_b) == 0)
2073 if (stat(a, &st_buf_a) < 0 || stat(b, &st_buf_b) < 0) {
2074 if (errno == ENOENT)
2079 /* Same blockdevice? */
2080 if (S_ISBLK(st_buf_a.st_mode) && S_ISBLK(st_buf_b.st_mode) &&
2081 st_buf_a.st_rdev == st_buf_b.st_rdev) {
2086 if (st_buf_a.st_dev == st_buf_b.st_dev &&
2087 st_buf_a.st_ino == st_buf_b.st_ino) {
2094 /* checks if a and b are identical or device
2095 * files associated with the same block device or
2096 * if one file is a loop device that uses the other
2099 static int is_same_loop_file(const char* a, const char* b)
2101 char res_a[PATH_MAX];
2102 char res_b[PATH_MAX];
2103 const char* final_a = NULL;
2104 const char* final_b = NULL;
2107 /* Resolve a if it is a loop device */
2108 if((ret = is_loop_device(a)) < 0) {
2113 ret = resolve_loop_device(a, res_a, sizeof(res_a));
2124 /* Resolve b if it is a loop device */
2125 if ((ret = is_loop_device(b)) < 0) {
2130 ret = resolve_loop_device(b, res_b, sizeof(res_b));
2141 return is_same_blk_file(final_a, final_b);
2144 /* Checks if a file exists and is a block or regular file*/
2145 static int is_existing_blk_or_reg_file(const char* filename)
2149 if(stat(filename, &st_buf) < 0) {
2156 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
2159 /* Checks if a file is used (directly or indirectly via a loop device)
2160 * by a device in fs_devices
2162 static int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices,
2166 struct list_head *head;
2167 struct list_head *cur;
2168 struct btrfs_device *device;
2170 head = &fs_devices->devices;
2171 list_for_each(cur, head) {
2172 device = list_entry(cur, struct btrfs_device, dev_list);
2174 if((ret = is_same_loop_file(device->name, file)))
2182 * Resolve a pathname to a device mapper node to /dev/mapper/<name>
2183 * Returns NULL on invalid input or malloc failure; Other failures
2184 * will be handled by the caller using the input pathame.
2186 char *canonicalize_dm_name(const char *ptname)
2190 char path[PATH_MAX], name[PATH_MAX], *res = NULL;
2192 if (!ptname || !*ptname)
2195 snprintf(path, sizeof(path), "/sys/block/%s/dm/name", ptname);
2196 if (!(f = fopen(path, "r")))
2199 /* read <name>\n from sysfs */
2200 if (fgets(name, sizeof(name), f) && (sz = strlen(name)) > 1) {
2201 name[sz - 1] = '\0';
2202 snprintf(path, sizeof(path), "/dev/mapper/%s", name);
2204 if (access(path, F_OK) == 0)
2212 * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or
2213 * to a device mapper pathname.
2214 * Returns NULL on invalid input or malloc failure; Other failures
2215 * will be handled by the caller using the input pathame.
2217 char *canonicalize_path(const char *path)
2219 char *canonical, *p;
2221 if (!path || !*path)
2224 canonical = realpath(path, NULL);
2226 return strdup(path);
2227 p = strrchr(canonical, '/');
2228 if (p && strncmp(p, "/dm-", 4) == 0 && isdigit(*(p + 4))) {
2229 char *dm = canonicalize_dm_name(p + 1);
2240 * returns 1 if the device was mounted, < 0 on error or 0 if everything
2241 * is safe to continue.
2243 int check_mounted(const char* file)
2248 fd = open(file, O_RDONLY);
2250 error("mount check: cannot open %s: %s", file,
2255 ret = check_mounted_where(fd, file, NULL, 0, NULL);
2261 int check_mounted_where(int fd, const char *file, char *where, int size,
2262 struct btrfs_fs_devices **fs_dev_ret)
2267 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2271 /* scan the initial device */
2272 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
2273 &total_devs, BTRFS_SUPER_INFO_OFFSET, SBREAD_DEFAULT);
2274 is_btrfs = (ret >= 0);
2276 /* scan other devices */
2277 if (is_btrfs && total_devs > 1) {
2278 ret = btrfs_scan_lblkid();
2283 /* iterate over the list of currently mounted filesystems */
2284 if ((f = setmntent ("/proc/self/mounts", "r")) == NULL)
2287 while ((mnt = getmntent (f)) != NULL) {
2289 if(strcmp(mnt->mnt_type, "btrfs") != 0)
2292 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
2294 /* ignore entries in the mount table that are not
2295 associated with a file*/
2296 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
2297 goto out_mntloop_err;
2301 ret = is_same_loop_file(file, mnt->mnt_fsname);
2305 goto out_mntloop_err;
2310 /* Did we find an entry in mnt table? */
2311 if (mnt && size && where) {
2312 strncpy(where, mnt->mnt_dir, size);
2316 *fs_dev_ret = fs_devices_mnt;
2318 ret = (mnt != NULL);
2326 struct pending_dir {
2327 struct list_head list;
2328 char name[PATH_MAX];
2331 int btrfs_register_one_device(const char *fname)
2333 struct btrfs_ioctl_vol_args args;
2337 fd = open("/dev/btrfs-control", O_RDWR);
2340 "failed to open /dev/btrfs-control, skipping device registration: %s",
2344 memset(&args, 0, sizeof(args));
2345 strncpy_null(args.name, fname);
2346 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
2348 error("device scan failed on '%s': %s", fname,
2357 * Register all devices in the fs_uuid list created in the user
2358 * space. Ensure btrfs_scan_lblkid() is called before this func.
2360 int btrfs_register_all_devices(void)
2364 struct btrfs_fs_devices *fs_devices;
2365 struct btrfs_device *device;
2366 struct list_head *all_uuids;
2368 all_uuids = btrfs_scanned_uuids();
2370 list_for_each_entry(fs_devices, all_uuids, list) {
2371 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2373 err = btrfs_register_one_device(device->name);
2383 int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
2386 struct btrfs_super_block *disk_super;
2390 buf = malloc(BTRFS_SUPER_INFO_SIZE);
2395 ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
2396 if (ret != BTRFS_SUPER_INFO_SIZE)
2400 disk_super = (struct btrfs_super_block *)buf;
2402 * Accept devices from the same filesystem, allow partially created
2405 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC &&
2406 btrfs_super_magic(disk_super) != BTRFS_MAGIC_PARTIAL)
2409 if (!memcmp(disk_super->fsid, root->fs_info->super_copy->fsid,
2419 * Note: this function uses a static per-thread buffer. Do not call this
2420 * function more than 10 times within one argument list!
2422 const char *pretty_size_mode(u64 size, unsigned mode)
2424 static __thread int ps_index = 0;
2425 static __thread char ps_array[10][32];
2428 ret = ps_array[ps_index];
2431 (void)pretty_size_snprintf(size, ret, 32, mode);
2436 static const char* unit_suffix_binary[] =
2437 { "B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB"};
2438 static const char* unit_suffix_decimal[] =
2439 { "B", "kB", "MB", "GB", "TB", "PB", "EB"};
2441 int pretty_size_snprintf(u64 size, char *str, size_t str_size, unsigned unit_mode)
2447 const char** suffix = NULL;
2453 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_RAW) {
2454 snprintf(str, str_size, "%llu", size);
2458 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_BINARY) {
2461 suffix = unit_suffix_binary;
2462 } else if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_DECIMAL) {
2465 suffix = unit_suffix_decimal;
2470 fprintf(stderr, "INTERNAL ERROR: unknown unit base, mode %d\n",
2478 switch (unit_mode & UNITS_MODE_MASK) {
2479 case UNITS_TBYTES: base *= mult; num_divs++;
2480 case UNITS_GBYTES: base *= mult; num_divs++;
2481 case UNITS_MBYTES: base *= mult; num_divs++;
2482 case UNITS_KBYTES: num_divs++;
2489 while (size >= mult) {
2495 * If the value is smaller than base, we didn't do any
2496 * division, in that case, base should be 1, not original
2497 * base, or the unit will be wrong
2503 if (num_divs >= ARRAY_SIZE(unit_suffix_binary)) {
2505 printf("INTERNAL ERROR: unsupported unit suffix, index %d\n",
2510 fraction = (float)last_size / base;
2512 return snprintf(str, str_size, "%.2f%s", fraction, suffix[num_divs]);
2516 * __strncpy_null - strncpy with null termination
2517 * @dest: the target array
2518 * @src: the source string
2519 * @n: maximum bytes to copy (size of *dest)
2521 * Like strncpy, but ensures destination is null-terminated.
2523 * Copies the string pointed to by src, including the terminating null
2524 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
2525 * of n bytes. Then ensure that dest is null-terminated.
2527 char *__strncpy_null(char *dest, const char *src, size_t n)
2529 strncpy(dest, src, n);
2536 * Checks to make sure that the label matches our requirements.
2538 0 if everything is safe and usable
2539 -1 if the label is too long
2541 static int check_label(const char *input)
2543 int len = strlen(input);
2545 if (len > BTRFS_LABEL_SIZE - 1) {
2546 error("label %s is too long (max %d)", input,
2547 BTRFS_LABEL_SIZE - 1);
2554 static int set_label_unmounted(const char *dev, const char *label)
2556 struct btrfs_trans_handle *trans;
2557 struct btrfs_root *root;
2560 ret = check_mounted(dev);
2562 error("checking mount status of %s failed: %d", dev, ret);
2566 error("device %s is mounted, use mount point", dev);
2570 /* Open the super_block at the default location
2571 * and as read-write.
2573 root = open_ctree(dev, 0, OPEN_CTREE_WRITES);
2574 if (!root) /* errors are printed by open_ctree() */
2577 trans = btrfs_start_transaction(root, 1);
2578 __strncpy_null(root->fs_info->super_copy->label, label, BTRFS_LABEL_SIZE - 1);
2580 btrfs_commit_transaction(trans, root);
2582 /* Now we close it since we are done. */
2587 static int set_label_mounted(const char *mount_path, const char *labelp)
2590 char label[BTRFS_LABEL_SIZE];
2592 fd = open(mount_path, O_RDONLY | O_NOATIME);
2594 error("unable to access %s: %s", mount_path, strerror(errno));
2598 memset(label, 0, sizeof(label));
2599 __strncpy_null(label, labelp, BTRFS_LABEL_SIZE - 1);
2600 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
2601 error("unable to set label of %s: %s", mount_path,
2611 int get_label_unmounted(const char *dev, char *label)
2613 struct btrfs_root *root;
2616 ret = check_mounted(dev);
2618 error("checking mount status of %s failed: %d", dev, ret);
2622 /* Open the super_block at the default location
2625 root = open_ctree(dev, 0, 0);
2629 __strncpy_null(label, root->fs_info->super_copy->label,
2630 BTRFS_LABEL_SIZE - 1);
2632 /* Now we close it since we are done. */
2638 * If a partition is mounted, try to get the filesystem label via its
2639 * mounted path rather than device. Return the corresponding error
2640 * the user specified the device path.
2642 int get_label_mounted(const char *mount_path, char *labelp)
2644 char label[BTRFS_LABEL_SIZE];
2648 fd = open(mount_path, O_RDONLY | O_NOATIME);
2650 error("unable to access %s: %s", mount_path, strerror(errno));
2654 memset(label, '\0', sizeof(label));
2655 ret = ioctl(fd, BTRFS_IOC_GET_FSLABEL, label);
2657 if (errno != ENOTTY)
2658 error("unable to get label of %s: %s", mount_path,
2665 __strncpy_null(labelp, label, BTRFS_LABEL_SIZE - 1);
2670 int get_label(const char *btrfs_dev, char *label)
2674 ret = is_existing_blk_or_reg_file(btrfs_dev);
2676 ret = get_label_mounted(btrfs_dev, label);
2678 ret = get_label_unmounted(btrfs_dev, label);
2683 int set_label(const char *btrfs_dev, const char *label)
2687 if (check_label(label))
2690 ret = is_existing_blk_or_reg_file(btrfs_dev);
2692 ret = set_label_mounted(btrfs_dev, label);
2694 ret = set_label_unmounted(btrfs_dev, label);
2700 * A not-so-good version fls64. No fascinating optimization since
2701 * no one except parse_size use it
2703 static int fls64(u64 x)
2707 for (i = 0; i <64; i++)
2708 if (x << i & (1ULL << 63))
2713 u64 parse_size(char *s)
2721 error("size value is empty");
2725 error("size value '%s' is less equal than 0", s);
2728 ret = strtoull(s, &endptr, 10);
2730 error("size value '%s' is invalid", s);
2733 if (endptr[0] && endptr[1]) {
2734 error("illegal suffix contains character '%c' in wrong position",
2739 * strtoll returns LLONG_MAX when overflow, if this happens,
2740 * need to call strtoull to get the real size
2742 if (errno == ERANGE && ret == ULLONG_MAX) {
2743 error("size value '%s' is too large for u64", s);
2747 c = tolower(endptr[0]);
2770 error("unknown size descriptor '%c'", c);
2774 /* Check whether ret * mult overflow */
2775 if (fls64(ret) + fls64(mult) - 1 > 64) {
2776 error("size value '%s' is too large for u64", s);
2783 u64 parse_qgroupid(const char *p)
2785 char *s = strchr(p, '/');
2786 const char *ptr_src_end = p + strlen(p);
2787 char *ptr_parse_end = NULL;
2796 /* Numeric format like '0/257' is the primary case */
2798 id = strtoull(p, &ptr_parse_end, 10);
2799 if (ptr_parse_end != ptr_src_end)
2803 level = strtoull(p, &ptr_parse_end, 10);
2804 if (ptr_parse_end != s)
2807 id = strtoull(s + 1, &ptr_parse_end, 10);
2808 if (ptr_parse_end != ptr_src_end)
2811 return (level << BTRFS_QGROUP_LEVEL_SHIFT) | id;
2814 /* Path format like subv at 'my_subvol' is the fallback case */
2815 ret = test_issubvolume(p);
2816 if (ret < 0 || !ret)
2818 fd = open(p, O_RDONLY);
2821 ret = lookup_ino_rootid(fd, &id);
2823 error("failed to lookup root id: %s", strerror(-ret));
2830 error("invalid qgroupid or subvolume path: %s", p);
2834 int open_file_or_dir3(const char *fname, DIR **dirstream, int open_flags)
2840 ret = stat(fname, &st);
2844 if (S_ISDIR(st.st_mode)) {
2845 *dirstream = opendir(fname);
2848 fd = dirfd(*dirstream);
2849 } else if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) {
2850 fd = open(fname, open_flags);
2853 * we set this on purpose, in case the caller output
2854 * strerror(errno) as success
2862 closedir(*dirstream);
2869 int open_file_or_dir(const char *fname, DIR **dirstream)
2871 return open_file_or_dir3(fname, dirstream, O_RDWR);
2874 void close_file_or_dir(int fd, DIR *dirstream)
2877 closedir(dirstream);
2882 int get_device_info(int fd, u64 devid,
2883 struct btrfs_ioctl_dev_info_args *di_args)
2887 di_args->devid = devid;
2888 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
2890 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
2891 return ret < 0 ? -errno : 0;
2894 static u64 find_max_device_id(struct btrfs_ioctl_search_args *search_args,
2897 struct btrfs_dev_item *dev_item;
2898 char *buf = search_args->buf;
2900 buf += (nr_items - 1) * (sizeof(struct btrfs_ioctl_search_header)
2901 + sizeof(struct btrfs_dev_item));
2902 buf += sizeof(struct btrfs_ioctl_search_header);
2904 dev_item = (struct btrfs_dev_item *)buf;
2906 return btrfs_stack_device_id(dev_item);
2909 static int search_chunk_tree_for_fs_info(int fd,
2910 struct btrfs_ioctl_fs_info_args *fi_args)
2914 u64 start_devid = 1;
2915 struct btrfs_ioctl_search_args search_args;
2916 struct btrfs_ioctl_search_key *search_key = &search_args.key;
2918 fi_args->num_devices = 0;
2920 max_items = BTRFS_SEARCH_ARGS_BUFSIZE
2921 / (sizeof(struct btrfs_ioctl_search_header)
2922 + sizeof(struct btrfs_dev_item));
2924 search_key->tree_id = BTRFS_CHUNK_TREE_OBJECTID;
2925 search_key->min_objectid = BTRFS_DEV_ITEMS_OBJECTID;
2926 search_key->max_objectid = BTRFS_DEV_ITEMS_OBJECTID;
2927 search_key->min_type = BTRFS_DEV_ITEM_KEY;
2928 search_key->max_type = BTRFS_DEV_ITEM_KEY;
2929 search_key->min_transid = 0;
2930 search_key->max_transid = (u64)-1;
2931 search_key->nr_items = max_items;
2932 search_key->max_offset = (u64)-1;
2935 search_key->min_offset = start_devid;
2937 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH, &search_args);
2941 fi_args->num_devices += (u64)search_key->nr_items;
2943 if (search_key->nr_items == max_items) {
2944 start_devid = find_max_device_id(&search_args,
2945 search_key->nr_items) + 1;
2949 /* get the lastest max_id to stay consistent with the num_devices */
2950 if (search_key->nr_items == 0)
2952 * last tree_search returns an empty buf, use the devid of
2953 * the last dev_item of the previous tree_search
2955 fi_args->max_id = start_devid - 1;
2957 fi_args->max_id = find_max_device_id(&search_args,
2958 search_key->nr_items);
2964 * For a given path, fill in the ioctl fs_ and info_ args.
2965 * If the path is a btrfs mountpoint, fill info for all devices.
2966 * If the path is a btrfs device, fill in only that device.
2968 * The path provided must be either on a mounted btrfs fs,
2969 * or be a mounted btrfs device.
2971 * Returns 0 on success, or a negative errno.
2973 int get_fs_info(char *path, struct btrfs_ioctl_fs_info_args *fi_args,
2974 struct btrfs_ioctl_dev_info_args **di_ret)
2981 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2982 struct btrfs_ioctl_dev_info_args *di_args;
2983 struct btrfs_ioctl_dev_info_args tmp;
2985 DIR *dirstream = NULL;
2987 memset(fi_args, 0, sizeof(*fi_args));
2989 if (is_block_device(path) == 1) {
2990 struct btrfs_super_block *disk_super;
2991 char buf[BTRFS_SUPER_INFO_SIZE];
2994 /* Ensure it's mounted, then set path to the mountpoint */
2995 fd = open(path, O_RDONLY);
2998 error("cannot open %s: %s", path, strerror(errno));
3001 ret = check_mounted_where(fd, path, mp, sizeof(mp),
3010 /* Only fill in this one device */
3011 fi_args->num_devices = 1;
3013 disk_super = (struct btrfs_super_block *)buf;
3014 ret = btrfs_read_dev_super(fd, disk_super,
3015 BTRFS_SUPER_INFO_OFFSET, 0);
3020 devid = btrfs_stack_device_id(&disk_super->dev_item);
3022 fi_args->max_id = devid;
3025 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
3029 /* at this point path must not be for a block device */
3030 fd = open_file_or_dir(path, &dirstream);
3036 /* fill in fi_args if not just a single device */
3037 if (fi_args->num_devices != 1) {
3038 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
3045 * The fs_args->num_devices does not include seed devices
3047 ret = search_chunk_tree_for_fs_info(fd, fi_args);
3052 * search_chunk_tree_for_fs_info() will lacks the devid 0
3053 * so manual probe for it here.
3055 ret = get_device_info(fd, 0, &tmp);
3057 fi_args->num_devices++;
3065 if (!fi_args->num_devices)
3068 di_args = *di_ret = malloc((fi_args->num_devices) * sizeof(*di_args));
3075 memcpy(di_args, &tmp, sizeof(tmp));
3076 for (; i <= fi_args->max_id; ++i) {
3077 ret = get_device_info(fd, i, &di_args[ndevs]);
3086 * only when the only dev we wanted to find is not there then
3087 * let any error be returned
3089 if (fi_args->num_devices != 1) {
3095 close_file_or_dir(fd, dirstream);
3099 #define isoctal(c) (((c) & ~7) == '0')
3101 static inline void translate(char *f, char *t)
3103 while (*f != '\0') {
3105 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
3106 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
3116 * Checks if the swap device.
3117 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
3119 static int is_swap_device(const char *file)
3130 if (stat(file, &st_buf) < 0)
3132 if (S_ISBLK(st_buf.st_mode))
3133 dev = st_buf.st_rdev;
3134 else if (S_ISREG(st_buf.st_mode)) {
3135 dev = st_buf.st_dev;
3136 ino = st_buf.st_ino;
3140 if ((f = fopen("/proc/swaps", "r")) == NULL)
3143 /* skip the first line */
3144 if (fgets(tmp, sizeof(tmp), f) == NULL)
3147 while (fgets(tmp, sizeof(tmp), f) != NULL) {
3148 if ((cp = strchr(tmp, ' ')) != NULL)
3150 if ((cp = strchr(tmp, '\t')) != NULL)
3152 translate(tmp, buf);
3153 if (stat(buf, &st_buf) != 0)
3155 if (S_ISBLK(st_buf.st_mode)) {
3156 if (dev == st_buf.st_rdev) {
3160 } else if (S_ISREG(st_buf.st_mode)) {
3161 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {
3175 * Check for existing filesystem or partition table on device.
3177 * 1 for existing fs or partition
3178 * 0 for nothing found
3179 * -1 for internal error
3181 static int check_overwrite(const char *device)
3184 blkid_probe pr = NULL;
3188 if (!device || !*device)
3191 ret = -1; /* will reset on success of all setup calls */
3193 pr = blkid_new_probe_from_filename(device);
3197 size = blkid_probe_get_size(pr);
3201 /* nothing to overwrite on a 0-length device */
3207 ret = blkid_probe_enable_partitions(pr, 1);
3211 ret = blkid_do_fullprobe(pr);
3216 * Blkid returns 1 for nothing found and 0 when it finds a signature,
3217 * but we want the exact opposite, so reverse the return value here.
3219 * In addition print some useful diagnostics about what actually is
3227 if (!blkid_probe_lookup_value(pr, "TYPE", &type, NULL)) {
3229 "%s appears to contain an existing "
3230 "filesystem (%s).\n", device, type);
3231 } else if (!blkid_probe_lookup_value(pr, "PTTYPE", &type, NULL)) {
3233 "%s appears to contain a partition "
3234 "table (%s).\n", device, type);
3237 "%s appears to contain something weird "
3238 "according to blkid\n", device);
3244 blkid_free_probe(pr);
3247 "probe of %s failed, cannot detect "
3248 "existing filesystem.\n", device);
3252 static int group_profile_devs_min(u64 flag)
3254 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3255 case 0: /* single */
3256 case BTRFS_BLOCK_GROUP_DUP:
3258 case BTRFS_BLOCK_GROUP_RAID0:
3259 case BTRFS_BLOCK_GROUP_RAID1:
3260 case BTRFS_BLOCK_GROUP_RAID5:
3262 case BTRFS_BLOCK_GROUP_RAID6:
3264 case BTRFS_BLOCK_GROUP_RAID10:
3271 int test_num_disk_vs_raid(u64 metadata_profile, u64 data_profile,
3272 u64 dev_cnt, int mixed, int ssd)
3275 u64 profile = metadata_profile | data_profile;
3280 allowed |= BTRFS_BLOCK_GROUP_RAID10;
3282 allowed |= BTRFS_BLOCK_GROUP_RAID6;
3284 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
3285 BTRFS_BLOCK_GROUP_RAID5;
3287 allowed |= BTRFS_BLOCK_GROUP_DUP;
3290 if (dev_cnt > 1 && profile & BTRFS_BLOCK_GROUP_DUP) {
3291 warning("DUP is not recommended on filesystem with multiple devices");
3293 if (metadata_profile & ~allowed) {
3295 "ERROR: unable to create FS with metadata profile %s "
3296 "(have %llu devices but %d devices are required)\n",
3297 btrfs_group_profile_str(metadata_profile), dev_cnt,
3298 group_profile_devs_min(metadata_profile));
3301 if (data_profile & ~allowed) {
3303 "ERROR: unable to create FS with data profile %s "
3304 "(have %llu devices but %d devices are required)\n",
3305 btrfs_group_profile_str(data_profile), dev_cnt,
3306 group_profile_devs_min(data_profile));
3310 if (dev_cnt == 3 && profile & BTRFS_BLOCK_GROUP_RAID6) {
3311 warning("RAID6 is not recommended on filesystem with 3 devices only");
3313 if (dev_cnt == 2 && profile & BTRFS_BLOCK_GROUP_RAID5) {
3314 warning("RAID5 is not recommended on filesystem with 2 devices only");
3316 warning_on(!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP) && ssd,
3317 "DUP may not actually lead to 2 copies on the device, see manual page");
3322 int group_profile_max_safe_loss(u64 flags)
3324 switch (flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3325 case 0: /* single */
3326 case BTRFS_BLOCK_GROUP_DUP:
3327 case BTRFS_BLOCK_GROUP_RAID0:
3329 case BTRFS_BLOCK_GROUP_RAID1:
3330 case BTRFS_BLOCK_GROUP_RAID5:
3331 case BTRFS_BLOCK_GROUP_RAID10:
3333 case BTRFS_BLOCK_GROUP_RAID6:
3341 * Check if a device is suitable for btrfs
3343 * 1: something is wrong, an error is printed
3346 int test_dev_for_mkfs(const char *file, int force_overwrite)
3351 ret = is_swap_device(file);
3353 error("checking status of %s: %s", file, strerror(-ret));
3357 error("%s is a swap device", file);
3360 if (!force_overwrite) {
3361 if (check_overwrite(file)) {
3362 error("use the -f option to force overwrite of %s",
3367 ret = check_mounted(file);
3369 error("cannot check mount status of %s: %s", file,
3374 error("%s is mounted", file);
3377 /* check if the device is busy */
3378 fd = open(file, O_RDWR|O_EXCL);
3380 error("unable to open %s: %s", file, strerror(errno));
3383 if (fstat(fd, &st)) {
3384 error("unable to stat %s: %s", file, strerror(errno));
3388 if (!S_ISBLK(st.st_mode)) {
3389 error("%s is not a block device", file);
3397 int btrfs_scan_lblkid(void)
3402 struct btrfs_fs_devices *tmp_devices;
3403 blkid_dev_iterate iter = NULL;
3404 blkid_dev dev = NULL;
3405 blkid_cache cache = NULL;
3406 char path[PATH_MAX];
3408 if (btrfs_scan_done)
3411 if (blkid_get_cache(&cache, NULL) < 0) {
3412 error("blkid cache get failed");
3415 blkid_probe_all(cache);
3416 iter = blkid_dev_iterate_begin(cache);
3417 blkid_dev_set_search(iter, "TYPE", "btrfs");
3418 while (blkid_dev_next(iter, &dev) == 0) {
3419 dev = blkid_verify(cache, dev);
3422 /* if we are here its definitely a btrfs disk*/
3423 strncpy_null(path, blkid_dev_devname(dev));
3425 fd = open(path, O_RDONLY);
3427 error("cannot open %s: %s", path, strerror(errno));
3430 ret = btrfs_scan_one_device(fd, path, &tmp_devices,
3431 &num_devices, BTRFS_SUPER_INFO_OFFSET,
3434 error("cannot scan %s: %s", path, strerror(-ret));
3441 blkid_dev_iterate_end(iter);
3442 blkid_put_cache(cache);
3444 btrfs_scan_done = 1;
3449 int is_vol_small(const char *file)
3456 fd = open(file, O_RDONLY);
3459 if (fstat(fd, &st) < 0) {
3464 size = btrfs_device_size(fd, &st);
3469 if (size < BTRFS_MKFS_SMALL_VOLUME_SIZE) {
3479 * This reads a line from the stdin and only returns non-zero if the
3480 * first whitespace delimited token is a case insensitive match with yes
3483 int ask_user(const char *question)
3485 char buf[30] = {0,};
3486 char *saveptr = NULL;
3489 printf("%s [y/N]: ", question);
3491 return fgets(buf, sizeof(buf) - 1, stdin) &&
3492 (answer = strtok_r(buf, " \t\n\r", &saveptr)) &&
3493 (!strcasecmp(answer, "yes") || !strcasecmp(answer, "y"));
3498 * - file or directory return the containing tree root id
3499 * - subvolume return its own tree id
3500 * - BTRFS_EMPTY_SUBVOL_DIR_OBJECTID (directory with ino == 2) the result is
3501 * undefined and function returns -1
3503 int lookup_ino_rootid(int fd, u64 *rootid)
3505 struct btrfs_ioctl_ino_lookup_args args;
3508 memset(&args, 0, sizeof(args));
3510 args.objectid = BTRFS_FIRST_FREE_OBJECTID;
3512 ret = ioctl(fd, BTRFS_IOC_INO_LOOKUP, &args);
3516 *rootid = args.treeid;
3522 * return 0 if a btrfs mount point is found
3523 * return 1 if a mount point is found but not btrfs
3524 * return <0 if something goes wrong
3526 int find_mount_root(const char *path, char **mount_root)
3534 int longest_matchlen = 0;
3535 char *longest_match = NULL;
3537 fd = open(path, O_RDONLY | O_NOATIME);
3542 mnttab = setmntent("/proc/self/mounts", "r");
3546 while ((ent = getmntent(mnttab))) {
3547 len = strlen(ent->mnt_dir);
3548 if (strncmp(ent->mnt_dir, path, len) == 0) {
3549 /* match found and use the latest match */
3550 if (longest_matchlen <= len) {
3551 free(longest_match);
3552 longest_matchlen = len;
3553 longest_match = strdup(ent->mnt_dir);
3554 not_btrfs = strcmp(ent->mnt_type, "btrfs");
3563 free(longest_match);
3568 *mount_root = realpath(longest_match, NULL);
3572 free(longest_match);
3576 int test_minimum_size(const char *file, u32 nodesize)
3579 struct stat statbuf;
3581 fd = open(file, O_RDONLY);
3584 if (stat(file, &statbuf) < 0) {
3588 if (btrfs_device_size(fd, &statbuf) < btrfs_min_dev_size(nodesize)) {
3598 * Test if path is a directory
3600 * 0 - path exists but it is not a directory
3601 * 1 - path exists and it is a directory
3604 int test_isdir(const char *path)
3609 ret = stat(path, &st);
3613 return !!S_ISDIR(st.st_mode);
3616 void units_set_mode(unsigned *units, unsigned mode)
3618 unsigned base = *units & UNITS_MODE_MASK;
3620 *units = base | mode;
3623 void units_set_base(unsigned *units, unsigned base)
3625 unsigned mode = *units & ~UNITS_MODE_MASK;
3627 *units = base | mode;
3630 int find_next_key(struct btrfs_path *path, struct btrfs_key *key)
3634 for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
3635 if (!path->nodes[level])
3637 if (path->slots[level] + 1 >=
3638 btrfs_header_nritems(path->nodes[level]))
3641 btrfs_item_key_to_cpu(path->nodes[level], key,
3642 path->slots[level] + 1);
3644 btrfs_node_key_to_cpu(path->nodes[level], key,
3645 path->slots[level] + 1);
3651 const char* btrfs_group_type_str(u64 flag)
3653 u64 mask = BTRFS_BLOCK_GROUP_TYPE_MASK |
3654 BTRFS_SPACE_INFO_GLOBAL_RSV;
3656 switch (flag & mask) {
3657 case BTRFS_BLOCK_GROUP_DATA:
3659 case BTRFS_BLOCK_GROUP_SYSTEM:
3661 case BTRFS_BLOCK_GROUP_METADATA:
3663 case BTRFS_BLOCK_GROUP_DATA|BTRFS_BLOCK_GROUP_METADATA:
3664 return "Data+Metadata";
3665 case BTRFS_SPACE_INFO_GLOBAL_RSV:
3666 return "GlobalReserve";
3672 const char* btrfs_group_profile_str(u64 flag)
3674 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3677 case BTRFS_BLOCK_GROUP_RAID0:
3679 case BTRFS_BLOCK_GROUP_RAID1:
3681 case BTRFS_BLOCK_GROUP_RAID5:
3683 case BTRFS_BLOCK_GROUP_RAID6:
3685 case BTRFS_BLOCK_GROUP_DUP:
3687 case BTRFS_BLOCK_GROUP_RAID10:
3694 u64 disk_size(const char *path)
3698 if (statfs(path, &sfs) < 0)
3701 return sfs.f_bsize * sfs.f_blocks;
3704 u64 get_partition_size(const char *dev)
3707 int fd = open(dev, O_RDONLY);
3711 if (ioctl(fd, BLKGETSIZE64, &result) < 0) {
3720 int btrfs_tree_search2_ioctl_supported(int fd)
3722 struct btrfs_ioctl_search_args_v2 *args2;
3723 struct btrfs_ioctl_search_key *sk;
3724 int args2_size = 1024;
3725 char args2_buf[args2_size];
3727 static int v2_supported = -1;
3729 if (v2_supported != -1)
3730 return v2_supported;
3732 args2 = (struct btrfs_ioctl_search_args_v2 *)args2_buf;
3736 * Search for the extent tree item in the root tree.
3738 sk->tree_id = BTRFS_ROOT_TREE_OBJECTID;
3739 sk->min_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3740 sk->max_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3741 sk->min_type = BTRFS_ROOT_ITEM_KEY;
3742 sk->max_type = BTRFS_ROOT_ITEM_KEY;
3744 sk->max_offset = (u64)-1;
3745 sk->min_transid = 0;
3746 sk->max_transid = (u64)-1;
3748 args2->buf_size = args2_size - sizeof(struct btrfs_ioctl_search_args_v2);
3749 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH_V2, args2);
3750 if (ret == -EOPNOTSUPP)
3757 return v2_supported;
3760 int btrfs_check_nodesize(u32 nodesize, u32 sectorsize, u64 features)
3762 if (nodesize < sectorsize) {
3763 error("illegal nodesize %u (smaller than %u)",
3764 nodesize, sectorsize);
3766 } else if (nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) {
3767 error("illegal nodesize %u (larger than %u)",
3768 nodesize, BTRFS_MAX_METADATA_BLOCKSIZE);
3770 } else if (nodesize & (sectorsize - 1)) {
3771 error("illegal nodesize %u (not aligned to %u)",
3772 nodesize, sectorsize);
3774 } else if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS &&
3775 nodesize != sectorsize) {
3776 error("illegal nodesize %u (not equal to %u for mixed block group)",
3777 nodesize, sectorsize);
3784 * Copy a path argument from SRC to DEST and check the SRC length if it's at
3785 * most PATH_MAX and fits into DEST. DESTLEN is supposed to be exact size of
3787 * The destination buffer is zero terminated.
3788 * Return < 0 for error, 0 otherwise.
3790 int arg_copy_path(char *dest, const char *src, int destlen)
3792 size_t len = strlen(src);
3794 if (len >= PATH_MAX || len >= destlen)
3795 return -ENAMETOOLONG;
3797 __strncpy_null(dest, src, destlen);
3802 unsigned int get_unit_mode_from_arg(int *argc, char *argv[], int df_mode)
3804 unsigned int unit_mode = UNITS_DEFAULT;
3808 for (arg_i = 0; arg_i < *argc; arg_i++) {
3809 if (!strcmp(argv[arg_i], "--"))
3812 if (!strcmp(argv[arg_i], "--raw")) {
3813 unit_mode = UNITS_RAW;
3817 if (!strcmp(argv[arg_i], "--human-readable")) {
3818 unit_mode = UNITS_HUMAN_BINARY;
3823 if (!strcmp(argv[arg_i], "--iec")) {
3824 units_set_mode(&unit_mode, UNITS_BINARY);
3828 if (!strcmp(argv[arg_i], "--si")) {
3829 units_set_mode(&unit_mode, UNITS_DECIMAL);
3834 if (!strcmp(argv[arg_i], "--kbytes")) {
3835 units_set_base(&unit_mode, UNITS_KBYTES);
3839 if (!strcmp(argv[arg_i], "--mbytes")) {
3840 units_set_base(&unit_mode, UNITS_MBYTES);
3844 if (!strcmp(argv[arg_i], "--gbytes")) {
3845 units_set_base(&unit_mode, UNITS_GBYTES);
3849 if (!strcmp(argv[arg_i], "--tbytes")) {
3850 units_set_base(&unit_mode, UNITS_TBYTES);
3858 if (!strcmp(argv[arg_i], "-b")) {
3859 unit_mode = UNITS_RAW;
3863 if (!strcmp(argv[arg_i], "-h")) {
3864 unit_mode = UNITS_HUMAN_BINARY;
3868 if (!strcmp(argv[arg_i], "-H")) {
3869 unit_mode = UNITS_HUMAN_DECIMAL;
3873 if (!strcmp(argv[arg_i], "-k")) {
3874 units_set_base(&unit_mode, UNITS_KBYTES);
3878 if (!strcmp(argv[arg_i], "-m")) {
3879 units_set_base(&unit_mode, UNITS_MBYTES);
3883 if (!strcmp(argv[arg_i], "-g")) {
3884 units_set_base(&unit_mode, UNITS_GBYTES);
3888 if (!strcmp(argv[arg_i], "-t")) {
3889 units_set_base(&unit_mode, UNITS_TBYTES);
3895 for (arg_i = 0, arg_end = 0; arg_i < *argc; arg_i++) {
3898 argv[arg_end] = argv[arg_i];
3907 int string_is_numerical(const char *str)
3909 if (!(*str >= '0' && *str <= '9'))
3911 while (*str >= '0' && *str <= '9')
3919 * Preprocess @argv with getopt_long to reorder options and consume the "--"
3921 * Unknown short and long options are reported, optionally the @usage is printed
3924 void clean_args_no_options(int argc, char *argv[], const char * const *usagestr)
3926 static const struct option long_options[] = {
3931 int c = getopt_long(argc, argv, "", long_options, NULL);
3945 * Same as clean_args_no_options but pass through arguments that could look
3946 * like short options. Eg. reisze which takes a negative resize argument like
3949 * This accepts only two forms:
3950 * - "-- option1 option2 ..."
3951 * - "option1 option2 ..."
3953 void clean_args_no_options_relaxed(int argc, char *argv[], const char * const *usagestr)
3958 if (strcmp(argv[1], "--") == 0)
3962 /* Subvolume helper functions */
3964 * test if name is a correct subvolume name
3965 * this function return
3966 * 0-> name is not a correct subvolume name
3967 * 1-> name is a correct subvolume name
3969 int test_issubvolname(const char *name)
3971 return name[0] != '\0' && !strchr(name, '/') &&
3972 strcmp(name, ".") && strcmp(name, "..");
3976 * Test if path is a subvolume
3978 * 0 - path exists but it is not a subvolume
3979 * 1 - path exists and it is a subvolume
3982 int test_issubvolume(const char *path)
3988 res = stat(path, &st);
3992 if (st.st_ino != BTRFS_FIRST_FREE_OBJECTID || !S_ISDIR(st.st_mode))
3995 res = statfs(path, &stfs);
3999 return (int)stfs.f_type == BTRFS_SUPER_MAGIC;
4002 const char *subvol_strip_mountpoint(const char *mnt, const char *full_path)
4004 int len = strlen(mnt);
4008 if (mnt[len - 1] != '/')
4011 return full_path + len;
4018 * 1: Error; and error info printed to the terminal. Fixme.
4019 * 2: If the fullpath is root tree instead of subvol tree
4021 int get_subvol_info(const char *fullpath, struct root_info *get_ri)
4028 const char *svpath = NULL;
4029 DIR *dirstream1 = NULL;
4030 DIR *dirstream2 = NULL;
4032 ret = test_issubvolume(fullpath);
4036 error("not a subvolume: %s", fullpath);
4040 ret = find_mount_root(fullpath, &mnt);
4044 error("%s doesn't belong to btrfs mount point", fullpath);
4048 svpath = subvol_strip_mountpoint(mnt, fullpath);
4050 fd = btrfs_open_dir(fullpath, &dirstream1, 1);
4054 ret = btrfs_list_get_path_rootid(fd, &sv_id);
4056 error("can't get rootid for '%s'", fullpath);
4060 mntfd = btrfs_open_dir(mnt, &dirstream2, 1);
4064 if (sv_id == BTRFS_FS_TREE_OBJECTID) {
4067 * So that caller may decide if thats an error or just fine.
4072 memset(get_ri, 0, sizeof(*get_ri));
4073 get_ri->root_id = sv_id;
4075 ret = btrfs_get_subvol(mntfd, get_ri);
4077 error("can't find '%s': %d", svpath, ret);
4080 close_file_or_dir(mntfd, dirstream2);
4081 close_file_or_dir(fd, dirstream1);
4087 void init_rand_seed(u64 seed)
4091 /* only use the last 48 bits */
4092 for (i = 0; i < 3; i++) {
4093 rand_seed[i] = (unsigned short)(seed ^ (unsigned short)(-1));
4096 rand_seed_initlized = 1;
4099 static void __init_seed(void)
4105 if(rand_seed_initlized)
4107 /* Use urandom as primary seed source. */
4108 fd = open("/dev/urandom", O_RDONLY);
4110 ret = read(fd, rand_seed, sizeof(rand_seed));
4112 if (ret < sizeof(rand_seed))
4116 /* Use time and pid as fallback seed */
4117 warning("failed to read /dev/urandom, use time and pid as random seed");
4118 gettimeofday(&tv, 0);
4119 rand_seed[0] = getpid() ^ (tv.tv_sec & 0xFFFF);
4120 rand_seed[1] = getppid() ^ (tv.tv_usec & 0xFFFF);
4121 rand_seed[2] = (tv.tv_sec ^ tv.tv_usec) >> 16;
4123 rand_seed_initlized = 1;
4130 * Don't use nrand48, its range is [0,2^31) The highest bit will alwasy
4131 * be 0. Use jrand48 to include the highest bit.
4133 return (u32)jrand48(rand_seed);
4136 unsigned int rand_range(unsigned int upper)
4140 * Use the full 48bits to mod, which would be more uniformly
4143 return (unsigned int)(jrand48(rand_seed) % upper);