2 * Copyright (C) 2007 Oracle. All rights reserved.
3 * Copyright (C) 2008 Morey Roof. All rights reserved.
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public
7 * License v2 as published by the Free Software Foundation.
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * General Public License for more details.
14 * You should have received a copy of the GNU General Public
15 * License along with this program; if not, write to the
16 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
17 * Boston, MA 021110-1307, USA.
23 #include <sys/ioctl.h>
24 #include <sys/mount.h>
25 #include <sys/types.h>
27 #include <uuid/uuid.h>
32 #include <linux/loop.h>
33 #include <linux/major.h>
34 #include <linux/kdev_t.h>
36 #include <blkid/blkid.h>
38 #include <sys/statfs.h>
39 #include <linux/magic.h>
40 #include <sys/utsname.h>
41 #include <linux/version.h>
43 #include "kerncompat.h"
44 #include "radix-tree.h"
47 #include "transaction.h"
55 #define BLKDISCARD _IO(0x12,119)
58 static int btrfs_scan_done = 0;
60 static int rand_seed_initlized = 0;
61 static unsigned short rand_seed[3];
63 struct btrfs_config bconf;
66 * Discard the given range in one go
68 static int discard_range(int fd, u64 start, u64 len)
70 u64 range[2] = { start, len };
72 if (ioctl(fd, BLKDISCARD, &range) < 0)
78 * Discard blocks in the given range in 1G chunks, the process is interruptible
80 static int discard_blocks(int fd, u64 start, u64 len)
84 u64 chunk_size = min_t(u64, len, SZ_1G);
87 ret = discard_range(fd, start, chunk_size);
97 static u64 reference_root_table[] = {
98 [1] = BTRFS_ROOT_TREE_OBJECTID,
99 [2] = BTRFS_EXTENT_TREE_OBJECTID,
100 [3] = BTRFS_CHUNK_TREE_OBJECTID,
101 [4] = BTRFS_DEV_TREE_OBJECTID,
102 [5] = BTRFS_FS_TREE_OBJECTID,
103 [6] = BTRFS_CSUM_TREE_OBJECTID,
106 int test_uuid_unique(char *fs_uuid)
109 blkid_dev_iterate iter = NULL;
110 blkid_dev dev = NULL;
111 blkid_cache cache = NULL;
113 if (blkid_get_cache(&cache, NULL) < 0) {
114 printf("ERROR: lblkid cache get failed\n");
117 blkid_probe_all(cache);
118 iter = blkid_dev_iterate_begin(cache);
119 blkid_dev_set_search(iter, "UUID", fs_uuid);
121 while (blkid_dev_next(iter, &dev) == 0) {
122 dev = blkid_verify(cache, dev);
129 blkid_dev_iterate_end(iter);
130 blkid_put_cache(cache);
136 * Reserve space from free_tree.
137 * The algorithm is very simple, find the first cache_extent with enough space
138 * and allocate from its beginning.
140 static int reserve_free_space(struct cache_tree *free_tree, u64 len,
143 struct cache_extent *cache;
146 ASSERT(ret_start != NULL);
147 cache = first_cache_extent(free_tree);
149 if (cache->size > len) {
151 *ret_start = cache->start;
154 if (cache->size == 0) {
155 remove_cache_extent(free_tree, cache);
162 cache = next_cache_extent(cache);
169 static inline int write_temp_super(int fd, struct btrfs_super_block *sb,
175 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
176 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
177 btrfs_csum_final(crc, &sb->csum[0]);
178 ret = pwrite(fd, sb, BTRFS_SUPER_INFO_SIZE, sb_bytenr);
179 if (ret < BTRFS_SUPER_INFO_SIZE)
180 ret = (ret < 0 ? -errno : -EIO);
187 * Setup temporary superblock at cfg->super_bynter
188 * Needed info are extracted from cfg, and root_bytenr, chunk_bytenr
190 * For now sys chunk array will be empty and dev_item is empty too.
191 * They will be re-initialized at temp chunk tree setup.
193 * The superblock signature is not valid, denotes a partially created
194 * filesystem, needs to be finalized.
196 static int setup_temp_super(int fd, struct btrfs_mkfs_config *cfg,
197 u64 root_bytenr, u64 chunk_bytenr)
199 unsigned char chunk_uuid[BTRFS_UUID_SIZE];
200 char super_buf[BTRFS_SUPER_INFO_SIZE];
201 struct btrfs_super_block *super = (struct btrfs_super_block *)super_buf;
204 memset(super_buf, 0, BTRFS_SUPER_INFO_SIZE);
205 cfg->num_bytes = round_down(cfg->num_bytes, cfg->sectorsize);
208 if (uuid_parse(cfg->fs_uuid, super->fsid) != 0) {
209 error("cound not parse UUID: %s", cfg->fs_uuid);
213 if (!test_uuid_unique(cfg->fs_uuid)) {
214 error("non-unique UUID: %s", cfg->fs_uuid);
219 uuid_generate(super->fsid);
220 uuid_unparse(super->fsid, cfg->fs_uuid);
222 uuid_generate(chunk_uuid);
223 uuid_unparse(chunk_uuid, cfg->chunk_uuid);
225 btrfs_set_super_bytenr(super, cfg->super_bytenr);
226 btrfs_set_super_num_devices(super, 1);
227 btrfs_set_super_magic(super, BTRFS_MAGIC_PARTIAL);
228 btrfs_set_super_generation(super, 1);
229 btrfs_set_super_root(super, root_bytenr);
230 btrfs_set_super_chunk_root(super, chunk_bytenr);
231 btrfs_set_super_total_bytes(super, cfg->num_bytes);
233 * Temporary filesystem will only have 6 tree roots:
234 * chunk tree, root tree, extent_tree, device tree, fs tree
237 btrfs_set_super_bytes_used(super, 6 * cfg->nodesize);
238 btrfs_set_super_sectorsize(super, cfg->sectorsize);
239 btrfs_set_super_leafsize(super, cfg->nodesize);
240 btrfs_set_super_nodesize(super, cfg->nodesize);
241 btrfs_set_super_stripesize(super, cfg->stripesize);
242 btrfs_set_super_csum_type(super, BTRFS_CSUM_TYPE_CRC32);
243 btrfs_set_super_chunk_root(super, chunk_bytenr);
244 btrfs_set_super_cache_generation(super, -1);
245 btrfs_set_super_incompat_flags(super, cfg->features);
247 __strncpy_null(super->label, cfg->label, BTRFS_LABEL_SIZE - 1);
249 /* Sys chunk array will be re-initialized at chunk tree init time */
250 super->sys_chunk_array_size = 0;
252 ret = write_temp_super(fd, super, cfg->super_bytenr);
258 * Setup an extent buffer for tree block.
260 static int setup_temp_extent_buffer(struct extent_buffer *buf,
261 struct btrfs_mkfs_config *cfg,
262 u64 bytenr, u64 owner)
264 unsigned char fsid[BTRFS_FSID_SIZE];
265 unsigned char chunk_uuid[BTRFS_UUID_SIZE];
268 ret = uuid_parse(cfg->fs_uuid, fsid);
271 ret = uuid_parse(cfg->chunk_uuid, chunk_uuid);
275 memset(buf->data, 0, cfg->nodesize);
276 buf->len = cfg->nodesize;
277 btrfs_set_header_bytenr(buf, bytenr);
278 btrfs_set_header_generation(buf, 1);
279 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
280 btrfs_set_header_owner(buf, owner);
281 btrfs_set_header_flags(buf, BTRFS_HEADER_FLAG_WRITTEN);
282 write_extent_buffer(buf, chunk_uuid, btrfs_header_chunk_tree_uuid(buf),
284 write_extent_buffer(buf, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
288 static inline int write_temp_extent_buffer(int fd, struct extent_buffer *buf,
293 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
295 /* Temporary extent buffer is always mapped 1:1 on disk */
296 ret = pwrite(fd, buf->data, buf->len, bytenr);
298 ret = (ret < 0 ? ret : -EIO);
305 * Insert a root item for temporary tree root
307 * Only used in make_btrfs_v2().
309 static void insert_temp_root_item(struct extent_buffer *buf,
310 struct btrfs_mkfs_config *cfg,
311 int *slot, u32 *itemoff, u64 objectid,
314 struct btrfs_root_item root_item;
315 struct btrfs_inode_item *inode_item;
316 struct btrfs_disk_key disk_key;
318 btrfs_set_header_nritems(buf, *slot + 1);
319 (*itemoff) -= sizeof(root_item);
320 memset(&root_item, 0, sizeof(root_item));
321 inode_item = &root_item.inode;
322 btrfs_set_stack_inode_generation(inode_item, 1);
323 btrfs_set_stack_inode_size(inode_item, 3);
324 btrfs_set_stack_inode_nlink(inode_item, 1);
325 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
326 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
327 btrfs_set_root_refs(&root_item, 1);
328 btrfs_set_root_used(&root_item, cfg->nodesize);
329 btrfs_set_root_generation(&root_item, 1);
330 btrfs_set_root_bytenr(&root_item, bytenr);
332 memset(&disk_key, 0, sizeof(disk_key));
333 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
334 btrfs_set_disk_key_objectid(&disk_key, objectid);
335 btrfs_set_disk_key_offset(&disk_key, 0);
337 btrfs_set_item_key(buf, &disk_key, *slot);
338 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
339 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(root_item));
340 write_extent_buffer(buf, &root_item,
341 btrfs_item_ptr_offset(buf, *slot),
346 static int setup_temp_root_tree(int fd, struct btrfs_mkfs_config *cfg,
347 u64 root_bytenr, u64 extent_bytenr,
348 u64 dev_bytenr, u64 fs_bytenr, u64 csum_bytenr)
350 struct extent_buffer *buf = NULL;
351 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
356 * Provided bytenr must in ascending order, or tree root will have a
359 if (!(root_bytenr < extent_bytenr && extent_bytenr < dev_bytenr &&
360 dev_bytenr < fs_bytenr && fs_bytenr < csum_bytenr)) {
361 error("bad tree bytenr order: "
362 "root < extent %llu < %llu, "
363 "extent < dev %llu < %llu, "
364 "dev < fs %llu < %llu, "
365 "fs < csum %llu < %llu",
366 (unsigned long long)root_bytenr,
367 (unsigned long long)extent_bytenr,
368 (unsigned long long)extent_bytenr,
369 (unsigned long long)dev_bytenr,
370 (unsigned long long)dev_bytenr,
371 (unsigned long long)fs_bytenr,
372 (unsigned long long)fs_bytenr,
373 (unsigned long long)csum_bytenr);
376 buf = malloc(sizeof(*buf) + cfg->nodesize);
380 ret = setup_temp_extent_buffer(buf, cfg, root_bytenr,
381 BTRFS_ROOT_TREE_OBJECTID);
385 insert_temp_root_item(buf, cfg, &slot, &itemoff,
386 BTRFS_EXTENT_TREE_OBJECTID, extent_bytenr);
387 insert_temp_root_item(buf, cfg, &slot, &itemoff,
388 BTRFS_DEV_TREE_OBJECTID, dev_bytenr);
389 insert_temp_root_item(buf, cfg, &slot, &itemoff,
390 BTRFS_FS_TREE_OBJECTID, fs_bytenr);
391 insert_temp_root_item(buf, cfg, &slot, &itemoff,
392 BTRFS_CSUM_TREE_OBJECTID, csum_bytenr);
394 ret = write_temp_extent_buffer(fd, buf, root_bytenr);
400 static int insert_temp_dev_item(int fd, struct extent_buffer *buf,
401 struct btrfs_mkfs_config *cfg,
402 int *slot, u32 *itemoff)
404 struct btrfs_disk_key disk_key;
405 struct btrfs_dev_item *dev_item;
406 char super_buf[BTRFS_SUPER_INFO_SIZE];
407 unsigned char dev_uuid[BTRFS_UUID_SIZE];
408 unsigned char fsid[BTRFS_FSID_SIZE];
409 struct btrfs_super_block *super = (struct btrfs_super_block *)super_buf;
412 ret = pread(fd, super_buf, BTRFS_SUPER_INFO_SIZE, cfg->super_bytenr);
413 if (ret < BTRFS_SUPER_INFO_SIZE) {
414 ret = (ret < 0 ? -errno : -EIO);
418 btrfs_set_header_nritems(buf, *slot + 1);
419 (*itemoff) -= sizeof(*dev_item);
420 /* setup device item 1, 0 is for replace case */
421 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
422 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
423 btrfs_set_disk_key_offset(&disk_key, 1);
424 btrfs_set_item_key(buf, &disk_key, *slot);
425 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
426 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(*dev_item));
428 dev_item = btrfs_item_ptr(buf, *slot, struct btrfs_dev_item);
429 /* Generate device uuid */
430 uuid_generate(dev_uuid);
431 write_extent_buffer(buf, dev_uuid,
432 (unsigned long)btrfs_device_uuid(dev_item),
434 uuid_parse(cfg->fs_uuid, fsid);
435 write_extent_buffer(buf, fsid,
436 (unsigned long)btrfs_device_fsid(dev_item),
438 btrfs_set_device_id(buf, dev_item, 1);
439 btrfs_set_device_generation(buf, dev_item, 0);
440 btrfs_set_device_total_bytes(buf, dev_item, cfg->num_bytes);
442 * The number must match the initial SYSTEM and META chunk size
444 btrfs_set_device_bytes_used(buf, dev_item,
445 BTRFS_MKFS_SYSTEM_GROUP_SIZE +
446 BTRFS_CONVERT_META_GROUP_SIZE);
447 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
448 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
449 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
450 btrfs_set_device_type(buf, dev_item, 0);
452 /* Super dev_item is not complete, copy the complete one to sb */
453 read_extent_buffer(buf, &super->dev_item, (unsigned long)dev_item,
455 ret = write_temp_super(fd, super, cfg->super_bytenr);
461 static int insert_temp_chunk_item(int fd, struct extent_buffer *buf,
462 struct btrfs_mkfs_config *cfg,
463 int *slot, u32 *itemoff, u64 start, u64 len,
466 struct btrfs_chunk *chunk;
467 struct btrfs_disk_key disk_key;
468 char super_buf[BTRFS_SUPER_INFO_SIZE];
469 struct btrfs_super_block *sb = (struct btrfs_super_block *)super_buf;
472 ret = pread(fd, super_buf, BTRFS_SUPER_INFO_SIZE,
474 if (ret < BTRFS_SUPER_INFO_SIZE) {
475 ret = (ret < 0 ? ret : -EIO);
479 btrfs_set_header_nritems(buf, *slot + 1);
480 (*itemoff) -= btrfs_chunk_item_size(1);
481 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
482 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
483 btrfs_set_disk_key_offset(&disk_key, start);
484 btrfs_set_item_key(buf, &disk_key, *slot);
485 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
486 btrfs_set_item_size(buf, btrfs_item_nr(*slot),
487 btrfs_chunk_item_size(1));
489 chunk = btrfs_item_ptr(buf, *slot, struct btrfs_chunk);
490 btrfs_set_chunk_length(buf, chunk, len);
491 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
492 btrfs_set_chunk_stripe_len(buf, chunk, BTRFS_STRIPE_LEN);
493 btrfs_set_chunk_type(buf, chunk, type);
494 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
495 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
496 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
497 btrfs_set_chunk_num_stripes(buf, chunk, 1);
498 /* TODO: Support DUP profile for system chunk */
499 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
500 /* We are doing 1:1 mapping, so start is its dev offset */
501 btrfs_set_stripe_offset_nr(buf, chunk, 0, start);
502 write_extent_buffer(buf, &sb->dev_item.uuid,
503 (unsigned long)btrfs_stripe_dev_uuid_nr(chunk, 0),
508 * If it's system chunk, also copy it to super block.
510 if (type & BTRFS_BLOCK_GROUP_SYSTEM) {
514 cur = (char *)sb->sys_chunk_array
515 + btrfs_super_sys_array_size(sb);
516 memcpy(cur, &disk_key, sizeof(disk_key));
517 cur += sizeof(disk_key);
518 read_extent_buffer(buf, cur, (unsigned long int)chunk,
519 btrfs_chunk_item_size(1));
520 array_size = btrfs_super_sys_array_size(sb);
521 array_size += btrfs_chunk_item_size(1) +
523 btrfs_set_super_sys_array_size(sb, array_size);
525 ret = write_temp_super(fd, sb, cfg->super_bytenr);
530 static int setup_temp_chunk_tree(int fd, struct btrfs_mkfs_config *cfg,
531 u64 sys_chunk_start, u64 meta_chunk_start,
534 struct extent_buffer *buf = NULL;
535 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
539 /* Must ensure SYS chunk starts before META chunk */
540 if (meta_chunk_start < sys_chunk_start) {
541 error("wrong chunk order: meta < system %llu < %llu",
542 (unsigned long long)meta_chunk_start,
543 (unsigned long long)sys_chunk_start);
546 buf = malloc(sizeof(*buf) + cfg->nodesize);
549 ret = setup_temp_extent_buffer(buf, cfg, chunk_bytenr,
550 BTRFS_CHUNK_TREE_OBJECTID);
554 ret = insert_temp_dev_item(fd, buf, cfg, &slot, &itemoff);
557 ret = insert_temp_chunk_item(fd, buf, cfg, &slot, &itemoff,
559 BTRFS_MKFS_SYSTEM_GROUP_SIZE,
560 BTRFS_BLOCK_GROUP_SYSTEM);
563 ret = insert_temp_chunk_item(fd, buf, cfg, &slot, &itemoff,
565 BTRFS_CONVERT_META_GROUP_SIZE,
566 BTRFS_BLOCK_GROUP_METADATA);
569 ret = write_temp_extent_buffer(fd, buf, chunk_bytenr);
576 static void insert_temp_dev_extent(struct extent_buffer *buf,
577 int *slot, u32 *itemoff, u64 start, u64 len)
579 struct btrfs_dev_extent *dev_extent;
580 struct btrfs_disk_key disk_key;
582 btrfs_set_header_nritems(buf, *slot + 1);
583 (*itemoff) -= sizeof(*dev_extent);
584 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
585 btrfs_set_disk_key_objectid(&disk_key, 1);
586 btrfs_set_disk_key_offset(&disk_key, start);
587 btrfs_set_item_key(buf, &disk_key, *slot);
588 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
589 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(*dev_extent));
591 dev_extent = btrfs_item_ptr(buf, *slot, struct btrfs_dev_extent);
592 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
593 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
594 btrfs_set_dev_extent_length(buf, dev_extent, len);
595 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, start);
596 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
597 BTRFS_CHUNK_TREE_OBJECTID);
601 static int setup_temp_dev_tree(int fd, struct btrfs_mkfs_config *cfg,
602 u64 sys_chunk_start, u64 meta_chunk_start,
605 struct extent_buffer *buf = NULL;
606 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
610 /* Must ensure SYS chunk starts before META chunk */
611 if (meta_chunk_start < sys_chunk_start) {
612 error("wrong chunk order: meta < system %llu < %llu",
613 (unsigned long long)meta_chunk_start,
614 (unsigned long long)sys_chunk_start);
617 buf = malloc(sizeof(*buf) + cfg->nodesize);
620 ret = setup_temp_extent_buffer(buf, cfg, dev_bytenr,
621 BTRFS_DEV_TREE_OBJECTID);
624 insert_temp_dev_extent(buf, &slot, &itemoff, sys_chunk_start,
625 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
626 insert_temp_dev_extent(buf, &slot, &itemoff, meta_chunk_start,
627 BTRFS_CONVERT_META_GROUP_SIZE);
628 ret = write_temp_extent_buffer(fd, buf, dev_bytenr);
634 static int setup_temp_fs_tree(int fd, struct btrfs_mkfs_config *cfg,
637 struct extent_buffer *buf = NULL;
640 buf = malloc(sizeof(*buf) + cfg->nodesize);
643 ret = setup_temp_extent_buffer(buf, cfg, fs_bytenr,
644 BTRFS_FS_TREE_OBJECTID);
648 * Temporary fs tree is completely empty.
650 ret = write_temp_extent_buffer(fd, buf, fs_bytenr);
656 static int setup_temp_csum_tree(int fd, struct btrfs_mkfs_config *cfg,
659 struct extent_buffer *buf = NULL;
662 buf = malloc(sizeof(*buf) + cfg->nodesize);
665 ret = setup_temp_extent_buffer(buf, cfg, csum_bytenr,
666 BTRFS_CSUM_TREE_OBJECTID);
670 * Temporary csum tree is completely empty.
672 ret = write_temp_extent_buffer(fd, buf, csum_bytenr);
679 * Insert one temporary extent item.
681 * NOTE: if skinny_metadata is not enabled, this function must be called
682 * after all other trees are initialized.
683 * Or fs without skinny-metadata will be screwed up.
685 static int insert_temp_extent_item(int fd, struct extent_buffer *buf,
686 struct btrfs_mkfs_config *cfg,
687 int *slot, u32 *itemoff, u64 bytenr,
690 struct extent_buffer *tmp;
691 struct btrfs_extent_item *ei;
692 struct btrfs_extent_inline_ref *iref;
693 struct btrfs_disk_key disk_key;
694 struct btrfs_disk_key tree_info_key;
695 struct btrfs_tree_block_info *info;
697 int skinny_metadata = cfg->features &
698 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA;
702 itemsize = sizeof(*ei) + sizeof(*iref);
704 itemsize = sizeof(*ei) + sizeof(*iref) +
705 sizeof(struct btrfs_tree_block_info);
707 btrfs_set_header_nritems(buf, *slot + 1);
708 *(itemoff) -= itemsize;
710 if (skinny_metadata) {
711 btrfs_set_disk_key_type(&disk_key, BTRFS_METADATA_ITEM_KEY);
712 btrfs_set_disk_key_offset(&disk_key, 0);
714 btrfs_set_disk_key_type(&disk_key, BTRFS_EXTENT_ITEM_KEY);
715 btrfs_set_disk_key_offset(&disk_key, cfg->nodesize);
717 btrfs_set_disk_key_objectid(&disk_key, bytenr);
719 btrfs_set_item_key(buf, &disk_key, *slot);
720 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
721 btrfs_set_item_size(buf, btrfs_item_nr(*slot), itemsize);
723 ei = btrfs_item_ptr(buf, *slot, struct btrfs_extent_item);
724 btrfs_set_extent_refs(buf, ei, 1);
725 btrfs_set_extent_generation(buf, ei, 1);
726 btrfs_set_extent_flags(buf, ei, BTRFS_EXTENT_FLAG_TREE_BLOCK);
728 if (skinny_metadata) {
729 iref = (struct btrfs_extent_inline_ref *)(ei + 1);
731 info = (struct btrfs_tree_block_info *)(ei + 1);
732 iref = (struct btrfs_extent_inline_ref *)(info + 1);
734 btrfs_set_extent_inline_ref_type(buf, iref,
735 BTRFS_TREE_BLOCK_REF_KEY);
736 btrfs_set_extent_inline_ref_offset(buf, iref, ref_root);
743 * Lastly, check the tree block key by read the tree block
744 * Since we do 1:1 mapping for convert case, we can directly
745 * read the bytenr from disk
747 tmp = malloc(sizeof(*tmp) + cfg->nodesize);
750 ret = setup_temp_extent_buffer(tmp, cfg, bytenr, ref_root);
753 ret = pread(fd, tmp->data, cfg->nodesize, bytenr);
754 if (ret < cfg->nodesize) {
755 ret = (ret < 0 ? -errno : -EIO);
758 if (btrfs_header_nritems(tmp) == 0) {
759 btrfs_set_disk_key_type(&tree_info_key, 0);
760 btrfs_set_disk_key_objectid(&tree_info_key, 0);
761 btrfs_set_disk_key_offset(&tree_info_key, 0);
763 btrfs_item_key(tmp, &tree_info_key, 0);
765 btrfs_set_tree_block_key(buf, info, &tree_info_key);
772 static void insert_temp_block_group(struct extent_buffer *buf,
773 struct btrfs_mkfs_config *cfg,
774 int *slot, u32 *itemoff,
775 u64 bytenr, u64 len, u64 used, u64 flag)
777 struct btrfs_block_group_item bgi;
778 struct btrfs_disk_key disk_key;
780 btrfs_set_header_nritems(buf, *slot + 1);
781 (*itemoff) -= sizeof(bgi);
782 btrfs_set_disk_key_type(&disk_key, BTRFS_BLOCK_GROUP_ITEM_KEY);
783 btrfs_set_disk_key_objectid(&disk_key, bytenr);
784 btrfs_set_disk_key_offset(&disk_key, len);
785 btrfs_set_item_key(buf, &disk_key, *slot);
786 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
787 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(bgi));
789 btrfs_set_block_group_flags(&bgi, flag);
790 btrfs_set_block_group_used(&bgi, used);
791 btrfs_set_block_group_chunk_objectid(&bgi,
792 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
793 write_extent_buffer(buf, &bgi, btrfs_item_ptr_offset(buf, *slot),
798 static int setup_temp_extent_tree(int fd, struct btrfs_mkfs_config *cfg,
799 u64 chunk_bytenr, u64 root_bytenr,
800 u64 extent_bytenr, u64 dev_bytenr,
801 u64 fs_bytenr, u64 csum_bytenr)
803 struct extent_buffer *buf = NULL;
804 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
809 * We must ensure provided bytenr are in ascending order,
810 * or extent tree key order will be broken.
812 if (!(chunk_bytenr < root_bytenr && root_bytenr < extent_bytenr &&
813 extent_bytenr < dev_bytenr && dev_bytenr < fs_bytenr &&
814 fs_bytenr < csum_bytenr)) {
815 error("bad tree bytenr order: "
816 "chunk < root %llu < %llu, "
817 "root < extent %llu < %llu, "
818 "extent < dev %llu < %llu, "
819 "dev < fs %llu < %llu, "
820 "fs < csum %llu < %llu",
821 (unsigned long long)chunk_bytenr,
822 (unsigned long long)root_bytenr,
823 (unsigned long long)root_bytenr,
824 (unsigned long long)extent_bytenr,
825 (unsigned long long)extent_bytenr,
826 (unsigned long long)dev_bytenr,
827 (unsigned long long)dev_bytenr,
828 (unsigned long long)fs_bytenr,
829 (unsigned long long)fs_bytenr,
830 (unsigned long long)csum_bytenr);
833 buf = malloc(sizeof(*buf) + cfg->nodesize);
837 ret = setup_temp_extent_buffer(buf, cfg, extent_bytenr,
838 BTRFS_EXTENT_TREE_OBJECTID);
842 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
843 chunk_bytenr, BTRFS_CHUNK_TREE_OBJECTID);
847 insert_temp_block_group(buf, cfg, &slot, &itemoff, chunk_bytenr,
848 BTRFS_MKFS_SYSTEM_GROUP_SIZE, cfg->nodesize,
849 BTRFS_BLOCK_GROUP_SYSTEM);
851 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
852 root_bytenr, BTRFS_ROOT_TREE_OBJECTID);
856 /* 5 tree block used, root, extent, dev, fs and csum*/
857 insert_temp_block_group(buf, cfg, &slot, &itemoff, root_bytenr,
858 BTRFS_CONVERT_META_GROUP_SIZE, cfg->nodesize * 5,
859 BTRFS_BLOCK_GROUP_METADATA);
861 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
862 extent_bytenr, BTRFS_EXTENT_TREE_OBJECTID);
865 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
866 dev_bytenr, BTRFS_DEV_TREE_OBJECTID);
869 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
870 fs_bytenr, BTRFS_FS_TREE_OBJECTID);
873 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
874 csum_bytenr, BTRFS_CSUM_TREE_OBJECTID);
878 ret = write_temp_extent_buffer(fd, buf, extent_bytenr);
885 * Improved version of make_btrfs().
888 * 1) Do chunk allocation to avoid used data
889 * And after this function, extent type matches chunk type
890 * 2) Better structured code
891 * No super long hand written codes to initialized all tree blocks
892 * Split into small blocks and reuse codes.
893 * TODO: Reuse tree operation facilities by introducing new flags
895 int make_convert_btrfs(int fd, struct btrfs_mkfs_config *cfg,
896 struct btrfs_convert_context *cctx)
898 struct cache_tree *free = &cctx->free;
899 struct cache_tree *used = &cctx->used;
901 u64 meta_chunk_start;
902 /* chunk tree bytenr, in system chunk */
904 /* metadata trees bytenr, in metadata chunk */
912 /* Shouldn't happen */
913 BUG_ON(cache_tree_empty(used));
916 * reserve space for temporary superblock first
917 * Here we allocate a little larger space, to keep later
918 * free space will be STRIPE_LEN aligned
920 ret = reserve_free_space(free, BTRFS_STRIPE_LEN,
926 * Then reserve system chunk space
927 * TODO: Change system group size depending on cctx->total_bytes.
928 * If using current 4M, it can only handle less than one TB for
929 * worst case and then run out of sys space.
931 ret = reserve_free_space(free, BTRFS_MKFS_SYSTEM_GROUP_SIZE,
935 ret = reserve_free_space(free, BTRFS_CONVERT_META_GROUP_SIZE,
941 * Allocated meta/sys chunks will be mapped 1:1 with device offset.
943 * Inside the allocated metadata chunk, the layout will be:
944 * | offset | contents |
945 * -------------------------------------
947 * | +nodesize | extent root |
948 * | +nodesize * 2 | device root |
949 * | +nodesize * 3 | fs tree |
950 * | +nodesize * 4 | csum tree |
951 * -------------------------------------
952 * Inside the allocated system chunk, the layout will be:
953 * | offset | contents |
954 * -------------------------------------
955 * | +0 | chunk root |
956 * -------------------------------------
958 chunk_bytenr = sys_chunk_start;
959 root_bytenr = meta_chunk_start;
960 extent_bytenr = meta_chunk_start + cfg->nodesize;
961 dev_bytenr = meta_chunk_start + cfg->nodesize * 2;
962 fs_bytenr = meta_chunk_start + cfg->nodesize * 3;
963 csum_bytenr = meta_chunk_start + cfg->nodesize * 4;
965 ret = setup_temp_super(fd, cfg, root_bytenr, chunk_bytenr);
969 ret = setup_temp_root_tree(fd, cfg, root_bytenr, extent_bytenr,
970 dev_bytenr, fs_bytenr, csum_bytenr);
973 ret = setup_temp_chunk_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
977 ret = setup_temp_dev_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
981 ret = setup_temp_fs_tree(fd, cfg, fs_bytenr);
984 ret = setup_temp_csum_tree(fd, cfg, csum_bytenr);
988 * Setup extent tree last, since it may need to read tree block key
989 * for non-skinny metadata case.
991 ret = setup_temp_extent_tree(fd, cfg, chunk_bytenr, root_bytenr,
992 extent_bytenr, dev_bytenr, fs_bytenr,
999 * @fs_uuid - if NULL, generates a UUID, returns back the new filesystem UUID
1001 * The superblock signature is not valid, denotes a partially created
1002 * filesystem, needs to be finalized.
1004 int make_btrfs(int fd, struct btrfs_mkfs_config *cfg)
1006 struct btrfs_super_block super;
1007 struct extent_buffer *buf;
1008 struct btrfs_root_item root_item;
1009 struct btrfs_disk_key disk_key;
1010 struct btrfs_extent_item *extent_item;
1011 struct btrfs_inode_item *inode_item;
1012 struct btrfs_chunk *chunk;
1013 struct btrfs_dev_item *dev_item;
1014 struct btrfs_dev_extent *dev_extent;
1015 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
1025 int skinny_metadata = !!(cfg->features &
1026 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA);
1029 buf = malloc(sizeof(*buf) + max(cfg->sectorsize, cfg->nodesize));
1033 first_free = BTRFS_SUPER_INFO_OFFSET + cfg->sectorsize * 2 - 1;
1034 first_free &= ~((u64)cfg->sectorsize - 1);
1036 memset(&super, 0, sizeof(super));
1038 num_bytes = (cfg->num_bytes / cfg->sectorsize) * cfg->sectorsize;
1039 if (*cfg->fs_uuid) {
1040 if (uuid_parse(cfg->fs_uuid, super.fsid) != 0) {
1041 error("cannot not parse UUID: %s", cfg->fs_uuid);
1045 if (!test_uuid_unique(cfg->fs_uuid)) {
1046 error("non-unique UUID: %s", cfg->fs_uuid);
1051 uuid_generate(super.fsid);
1052 uuid_unparse(super.fsid, cfg->fs_uuid);
1054 uuid_generate(super.dev_item.uuid);
1055 uuid_generate(chunk_tree_uuid);
1057 btrfs_set_super_bytenr(&super, cfg->blocks[0]);
1058 btrfs_set_super_num_devices(&super, 1);
1059 btrfs_set_super_magic(&super, BTRFS_MAGIC_PARTIAL);
1060 btrfs_set_super_generation(&super, 1);
1061 btrfs_set_super_root(&super, cfg->blocks[1]);
1062 btrfs_set_super_chunk_root(&super, cfg->blocks[3]);
1063 btrfs_set_super_total_bytes(&super, num_bytes);
1064 btrfs_set_super_bytes_used(&super, 6 * cfg->nodesize);
1065 btrfs_set_super_sectorsize(&super, cfg->sectorsize);
1066 btrfs_set_super_leafsize(&super, cfg->nodesize);
1067 btrfs_set_super_nodesize(&super, cfg->nodesize);
1068 btrfs_set_super_stripesize(&super, cfg->stripesize);
1069 btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
1070 btrfs_set_super_chunk_root_generation(&super, 1);
1071 btrfs_set_super_cache_generation(&super, -1);
1072 btrfs_set_super_incompat_flags(&super, cfg->features);
1074 __strncpy_null(super.label, cfg->label, BTRFS_LABEL_SIZE - 1);
1076 /* create the tree of root objects */
1077 memset(buf->data, 0, cfg->nodesize);
1078 buf->len = cfg->nodesize;
1079 btrfs_set_header_bytenr(buf, cfg->blocks[1]);
1080 btrfs_set_header_nritems(buf, 4);
1081 btrfs_set_header_generation(buf, 1);
1082 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
1083 btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
1084 write_extent_buffer(buf, super.fsid, btrfs_header_fsid(),
1087 write_extent_buffer(buf, chunk_tree_uuid,
1088 btrfs_header_chunk_tree_uuid(buf),
1091 /* create the items for the root tree */
1092 memset(&root_item, 0, sizeof(root_item));
1093 inode_item = &root_item.inode;
1094 btrfs_set_stack_inode_generation(inode_item, 1);
1095 btrfs_set_stack_inode_size(inode_item, 3);
1096 btrfs_set_stack_inode_nlink(inode_item, 1);
1097 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
1098 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
1099 btrfs_set_root_refs(&root_item, 1);
1100 btrfs_set_root_used(&root_item, cfg->nodesize);
1101 btrfs_set_root_generation(&root_item, 1);
1103 memset(&disk_key, 0, sizeof(disk_key));
1104 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
1105 btrfs_set_disk_key_offset(&disk_key, 0);
1108 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - sizeof(root_item);
1109 btrfs_set_root_bytenr(&root_item, cfg->blocks[2]);
1110 btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
1111 btrfs_set_item_key(buf, &disk_key, nritems);
1112 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1113 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1115 write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
1116 nritems), sizeof(root_item));
1119 itemoff = itemoff - sizeof(root_item);
1120 btrfs_set_root_bytenr(&root_item, cfg->blocks[4]);
1121 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
1122 btrfs_set_item_key(buf, &disk_key, nritems);
1123 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1124 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1126 write_extent_buffer(buf, &root_item,
1127 btrfs_item_ptr_offset(buf, nritems),
1131 itemoff = itemoff - sizeof(root_item);
1132 btrfs_set_root_bytenr(&root_item, cfg->blocks[5]);
1133 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
1134 btrfs_set_item_key(buf, &disk_key, nritems);
1135 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1136 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1138 write_extent_buffer(buf, &root_item,
1139 btrfs_item_ptr_offset(buf, nritems),
1143 itemoff = itemoff - sizeof(root_item);
1144 btrfs_set_root_bytenr(&root_item, cfg->blocks[6]);
1145 btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_TREE_OBJECTID);
1146 btrfs_set_item_key(buf, &disk_key, nritems);
1147 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1148 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1150 write_extent_buffer(buf, &root_item,
1151 btrfs_item_ptr_offset(buf, nritems),
1156 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1157 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[1]);
1158 if (ret != cfg->nodesize) {
1159 ret = (ret < 0 ? -errno : -EIO);
1163 /* create the items for the extent tree */
1164 memset(buf->data + sizeof(struct btrfs_header), 0,
1165 cfg->nodesize - sizeof(struct btrfs_header));
1167 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
1168 for (i = 1; i < 7; i++) {
1169 item_size = sizeof(struct btrfs_extent_item);
1170 if (!skinny_metadata)
1171 item_size += sizeof(struct btrfs_tree_block_info);
1173 if (cfg->blocks[i] < first_free) {
1174 error("block[%d] below first free: %llu < %llu",
1175 i, (unsigned long long)cfg->blocks[i],
1176 (unsigned long long)first_free);
1180 if (cfg->blocks[i] < cfg->blocks[i - 1]) {
1181 error("blocks %d and %d in reverse order: %llu < %llu",
1183 (unsigned long long)cfg->blocks[i],
1184 (unsigned long long)cfg->blocks[i - 1]);
1189 /* create extent item */
1190 itemoff -= item_size;
1191 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
1192 if (skinny_metadata) {
1193 btrfs_set_disk_key_type(&disk_key,
1194 BTRFS_METADATA_ITEM_KEY);
1195 btrfs_set_disk_key_offset(&disk_key, 0);
1197 btrfs_set_disk_key_type(&disk_key,
1198 BTRFS_EXTENT_ITEM_KEY);
1199 btrfs_set_disk_key_offset(&disk_key, cfg->nodesize);
1201 btrfs_set_item_key(buf, &disk_key, nritems);
1202 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
1204 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1206 extent_item = btrfs_item_ptr(buf, nritems,
1207 struct btrfs_extent_item);
1208 btrfs_set_extent_refs(buf, extent_item, 1);
1209 btrfs_set_extent_generation(buf, extent_item, 1);
1210 btrfs_set_extent_flags(buf, extent_item,
1211 BTRFS_EXTENT_FLAG_TREE_BLOCK);
1214 /* create extent ref */
1215 ref_root = reference_root_table[i];
1216 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
1217 btrfs_set_disk_key_offset(&disk_key, ref_root);
1218 btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
1219 btrfs_set_item_key(buf, &disk_key, nritems);
1220 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
1222 btrfs_set_item_size(buf, btrfs_item_nr(nritems), 0);
1225 btrfs_set_header_bytenr(buf, cfg->blocks[2]);
1226 btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
1227 btrfs_set_header_nritems(buf, nritems);
1228 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1229 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[2]);
1230 if (ret != cfg->nodesize) {
1231 ret = (ret < 0 ? -errno : -EIO);
1235 /* create the chunk tree */
1236 memset(buf->data + sizeof(struct btrfs_header), 0,
1237 cfg->nodesize - sizeof(struct btrfs_header));
1239 item_size = sizeof(*dev_item);
1240 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - item_size;
1242 /* first device 1 (there is no device 0) */
1243 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
1244 btrfs_set_disk_key_offset(&disk_key, 1);
1245 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
1246 btrfs_set_item_key(buf, &disk_key, nritems);
1247 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1248 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
1250 dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
1251 btrfs_set_device_id(buf, dev_item, 1);
1252 btrfs_set_device_generation(buf, dev_item, 0);
1253 btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
1254 btrfs_set_device_bytes_used(buf, dev_item,
1255 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1256 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
1257 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
1258 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
1259 btrfs_set_device_type(buf, dev_item, 0);
1261 write_extent_buffer(buf, super.dev_item.uuid,
1262 (unsigned long)btrfs_device_uuid(dev_item),
1264 write_extent_buffer(buf, super.fsid,
1265 (unsigned long)btrfs_device_fsid(dev_item),
1267 read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
1271 item_size = btrfs_chunk_item_size(1);
1272 itemoff = itemoff - item_size;
1274 /* then we have chunk 0 */
1275 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1276 btrfs_set_disk_key_offset(&disk_key, 0);
1277 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
1278 btrfs_set_item_key(buf, &disk_key, nritems);
1279 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1280 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
1282 chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
1283 btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1284 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
1285 btrfs_set_chunk_stripe_len(buf, chunk, BTRFS_STRIPE_LEN);
1286 btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
1287 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
1288 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
1289 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
1290 btrfs_set_chunk_num_stripes(buf, chunk, 1);
1291 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
1292 btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
1295 write_extent_buffer(buf, super.dev_item.uuid,
1296 (unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
1299 /* copy the key for the chunk to the system array */
1300 ptr = super.sys_chunk_array;
1301 array_size = sizeof(disk_key);
1303 memcpy(ptr, &disk_key, sizeof(disk_key));
1304 ptr += sizeof(disk_key);
1306 /* copy the chunk to the system array */
1307 read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
1308 array_size += item_size;
1310 btrfs_set_super_sys_array_size(&super, array_size);
1312 btrfs_set_header_bytenr(buf, cfg->blocks[3]);
1313 btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
1314 btrfs_set_header_nritems(buf, nritems);
1315 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1316 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[3]);
1317 if (ret != cfg->nodesize) {
1318 ret = (ret < 0 ? -errno : -EIO);
1322 /* create the device tree */
1323 memset(buf->data + sizeof(struct btrfs_header), 0,
1324 cfg->nodesize - sizeof(struct btrfs_header));
1326 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) -
1327 sizeof(struct btrfs_dev_extent);
1329 btrfs_set_disk_key_objectid(&disk_key, 1);
1330 btrfs_set_disk_key_offset(&disk_key, 0);
1331 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
1332 btrfs_set_item_key(buf, &disk_key, nritems);
1333 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1334 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1335 sizeof(struct btrfs_dev_extent));
1336 dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
1337 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
1338 BTRFS_CHUNK_TREE_OBJECTID);
1339 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
1340 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1341 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
1343 write_extent_buffer(buf, chunk_tree_uuid,
1344 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
1347 btrfs_set_dev_extent_length(buf, dev_extent,
1348 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1351 btrfs_set_header_bytenr(buf, cfg->blocks[4]);
1352 btrfs_set_header_owner(buf, BTRFS_DEV_TREE_OBJECTID);
1353 btrfs_set_header_nritems(buf, nritems);
1354 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1355 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[4]);
1356 if (ret != cfg->nodesize) {
1357 ret = (ret < 0 ? -errno : -EIO);
1361 /* create the FS root */
1362 memset(buf->data + sizeof(struct btrfs_header), 0,
1363 cfg->nodesize - sizeof(struct btrfs_header));
1364 btrfs_set_header_bytenr(buf, cfg->blocks[5]);
1365 btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
1366 btrfs_set_header_nritems(buf, 0);
1367 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1368 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[5]);
1369 if (ret != cfg->nodesize) {
1370 ret = (ret < 0 ? -errno : -EIO);
1373 /* finally create the csum root */
1374 memset(buf->data + sizeof(struct btrfs_header), 0,
1375 cfg->nodesize - sizeof(struct btrfs_header));
1376 btrfs_set_header_bytenr(buf, cfg->blocks[6]);
1377 btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
1378 btrfs_set_header_nritems(buf, 0);
1379 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1380 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[6]);
1381 if (ret != cfg->nodesize) {
1382 ret = (ret < 0 ? -errno : -EIO);
1386 /* and write out the super block */
1387 memset(buf->data, 0, BTRFS_SUPER_INFO_SIZE);
1388 memcpy(buf->data, &super, sizeof(super));
1389 buf->len = BTRFS_SUPER_INFO_SIZE;
1390 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1391 ret = pwrite(fd, buf->data, BTRFS_SUPER_INFO_SIZE, cfg->blocks[0]);
1392 if (ret != BTRFS_SUPER_INFO_SIZE) {
1393 ret = (ret < 0 ? -errno : -EIO);
1404 #define VERSION_TO_STRING3(a,b,c) #a "." #b "." #c, KERNEL_VERSION(a,b,c)
1405 #define VERSION_TO_STRING2(a,b) #a "." #b, KERNEL_VERSION(a,b,0)
1408 * Feature stability status and versions: compat <= safe <= default
1410 static const struct btrfs_fs_feature {
1413 const char *sysfs_name;
1415 * Compatibility with kernel of given version. Filesystem can be
1418 const char *compat_str;
1421 * Considered safe for use, but is not on by default, even if the
1422 * kernel supports the feature.
1424 const char *safe_str;
1427 * Considered safe for use and will be turned on by default if
1428 * supported by the running kernel.
1430 const char *default_str;
1433 } mkfs_features[] = {
1434 { "mixed-bg", BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS,
1436 VERSION_TO_STRING3(2,6,37),
1437 VERSION_TO_STRING3(2,6,37),
1439 "mixed data and metadata block groups" },
1440 { "extref", BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF,
1442 VERSION_TO_STRING2(3,7),
1443 VERSION_TO_STRING2(3,12),
1444 VERSION_TO_STRING2(3,12),
1445 "increased hardlink limit per file to 65536" },
1446 { "raid56", BTRFS_FEATURE_INCOMPAT_RAID56,
1448 VERSION_TO_STRING2(3,9),
1451 "raid56 extended format" },
1452 { "skinny-metadata", BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA,
1454 VERSION_TO_STRING2(3,10),
1455 VERSION_TO_STRING2(3,18),
1456 VERSION_TO_STRING2(3,18),
1457 "reduced-size metadata extent refs" },
1458 { "no-holes", BTRFS_FEATURE_INCOMPAT_NO_HOLES,
1460 VERSION_TO_STRING2(3,14),
1461 VERSION_TO_STRING2(4,0),
1463 "no explicit hole extents for files" },
1464 /* Keep this one last */
1465 { "list-all", BTRFS_FEATURE_LIST_ALL, NULL }
1468 static int parse_one_fs_feature(const char *name, u64 *flags)
1473 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1474 if (name[0] == '^' &&
1475 !strcmp(mkfs_features[i].name, name + 1)) {
1476 *flags &= ~ mkfs_features[i].flag;
1478 } else if (!strcmp(mkfs_features[i].name, name)) {
1479 *flags |= mkfs_features[i].flag;
1487 void btrfs_parse_features_to_string(char *buf, u64 flags)
1493 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1494 if (flags & mkfs_features[i].flag) {
1497 strcat(buf, mkfs_features[i].name);
1502 void btrfs_process_fs_features(u64 flags)
1506 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1507 if (flags & mkfs_features[i].flag) {
1508 printf("Turning ON incompat feature '%s': %s\n",
1509 mkfs_features[i].name,
1510 mkfs_features[i].desc);
1515 void btrfs_list_all_fs_features(u64 mask_disallowed)
1519 fprintf(stderr, "Filesystem features available:\n");
1520 for (i = 0; i < ARRAY_SIZE(mkfs_features) - 1; i++) {
1521 const struct btrfs_fs_feature *feat = &mkfs_features[i];
1523 if (feat->flag & mask_disallowed)
1525 fprintf(stderr, "%-20s- %s (0x%llx", feat->name, feat->desc,
1527 if (feat->compat_ver)
1528 fprintf(stderr, ", compat=%s", feat->compat_str);
1530 fprintf(stderr, ", safe=%s", feat->safe_str);
1531 if (feat->default_ver)
1532 fprintf(stderr, ", default=%s", feat->default_str);
1533 fprintf(stderr, ")\n");
1538 * Return NULL if all features were parsed fine, otherwise return the name of
1539 * the first unparsed.
1541 char* btrfs_parse_fs_features(char *namelist, u64 *flags)
1544 char *save_ptr = NULL; /* Satisfy static checkers */
1546 for (this_char = strtok_r(namelist, ",", &save_ptr);
1548 this_char = strtok_r(NULL, ",", &save_ptr)) {
1549 if (parse_one_fs_feature(this_char, flags))
1556 void print_kernel_version(FILE *stream, u32 version)
1560 v[0] = version & 0xFF;
1561 v[1] = (version >> 8) & 0xFF;
1562 v[2] = version >> 16;
1563 fprintf(stream, "%u.%u", v[2], v[1]);
1565 fprintf(stream, ".%u", v[0]);
1568 u32 get_running_kernel_version(void)
1570 struct utsname utsbuf;
1572 char *saveptr = NULL;
1576 if (strcmp(utsbuf.sysname, "Linux") != 0) {
1577 error("unsupported system: %s", utsbuf.sysname);
1581 tmp = strchr(utsbuf.release, '-');
1585 tmp = strtok_r(utsbuf.release, ".", &saveptr);
1586 if (!string_is_numerical(tmp))
1588 version = atoi(tmp) << 16;
1589 tmp = strtok_r(NULL, ".", &saveptr);
1590 if (!string_is_numerical(tmp))
1592 version |= atoi(tmp) << 8;
1593 tmp = strtok_r(NULL, ".", &saveptr);
1595 if (!string_is_numerical(tmp))
1597 version |= atoi(tmp);
1603 u64 btrfs_device_size(int fd, struct stat *st)
1606 if (S_ISREG(st->st_mode)) {
1609 if (!S_ISBLK(st->st_mode)) {
1612 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
1618 static int zero_blocks(int fd, off_t start, size_t len)
1620 char *buf = malloc(len);
1626 memset(buf, 0, len);
1627 written = pwrite(fd, buf, len, start);
1634 #define ZERO_DEV_BYTES SZ_2M
1636 /* don't write outside the device by clamping the region to the device size */
1637 static int zero_dev_clamped(int fd, off_t start, ssize_t len, u64 dev_size)
1639 off_t end = max(start, start + len);
1642 /* and don't overwrite the disk labels on sparc */
1643 start = max(start, 1024);
1644 end = max(end, 1024);
1647 start = min_t(u64, start, dev_size);
1648 end = min_t(u64, end, dev_size);
1650 return zero_blocks(fd, start, end - start);
1653 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
1654 struct btrfs_root *root, int fd, const char *path,
1655 u64 device_total_bytes, u32 io_width, u32 io_align,
1658 struct btrfs_super_block *disk_super;
1659 struct btrfs_super_block *super = root->fs_info->super_copy;
1660 struct btrfs_device *device;
1661 struct btrfs_dev_item *dev_item;
1667 device_total_bytes = (device_total_bytes / sectorsize) * sectorsize;
1669 device = calloc(1, sizeof(*device));
1674 buf = calloc(1, sectorsize);
1680 disk_super = (struct btrfs_super_block *)buf;
1681 dev_item = &disk_super->dev_item;
1683 uuid_generate(device->uuid);
1686 device->io_width = io_width;
1687 device->io_align = io_align;
1688 device->sector_size = sectorsize;
1690 device->writeable = 1;
1691 device->total_bytes = device_total_bytes;
1692 device->bytes_used = 0;
1693 device->total_ios = 0;
1694 device->dev_root = root->fs_info->dev_root;
1695 device->name = strdup(path);
1696 if (!device->name) {
1701 INIT_LIST_HEAD(&device->dev_list);
1702 ret = btrfs_add_device(trans, root, device);
1706 fs_total_bytes = btrfs_super_total_bytes(super) + device_total_bytes;
1707 btrfs_set_super_total_bytes(super, fs_total_bytes);
1709 num_devs = btrfs_super_num_devices(super) + 1;
1710 btrfs_set_super_num_devices(super, num_devs);
1712 memcpy(disk_super, super, sizeof(*disk_super));
1714 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
1715 btrfs_set_stack_device_id(dev_item, device->devid);
1716 btrfs_set_stack_device_type(dev_item, device->type);
1717 btrfs_set_stack_device_io_align(dev_item, device->io_align);
1718 btrfs_set_stack_device_io_width(dev_item, device->io_width);
1719 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
1720 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
1721 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
1722 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
1724 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
1725 BUG_ON(ret != sectorsize);
1728 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
1729 device->fs_devices = root->fs_info->fs_devices;
1738 static int btrfs_wipe_existing_sb(int fd)
1740 const char *off = NULL;
1745 blkid_probe pr = NULL;
1747 pr = blkid_new_probe();
1751 if (blkid_probe_set_device(pr, fd, 0, 0)) {
1756 ret = blkid_probe_lookup_value(pr, "SBMAGIC_OFFSET", &off, NULL);
1758 ret = blkid_probe_lookup_value(pr, "SBMAGIC", NULL, &len);
1760 if (ret || len == 0 || off == NULL) {
1762 * If lookup fails, the probe did not find any values, eg. for
1763 * a file image or a loop device. Soft error.
1769 offset = strtoll(off, NULL, 10);
1770 if (len > sizeof(buf))
1773 memset(buf, 0, len);
1774 ret = pwrite(fd, buf, len, offset);
1776 error("cannot wipe existing superblock: %s", strerror(errno));
1778 } else if (ret != len) {
1779 error("cannot wipe existing superblock: wrote %d of %zd", ret, len);
1785 blkid_free_probe(pr);
1789 int btrfs_prepare_device(int fd, const char *file, u64 *block_count_ret,
1790 u64 max_block_count, unsigned opflags)
1796 ret = fstat(fd, &st);
1798 error("unable to stat %s: %s", file, strerror(errno));
1802 block_count = btrfs_device_size(fd, &st);
1803 if (block_count == 0) {
1804 error("unable to determine size of %s", file);
1807 if (max_block_count)
1808 block_count = min(block_count, max_block_count);
1810 if (opflags & PREP_DEVICE_DISCARD) {
1812 * We intentionally ignore errors from the discard ioctl. It
1813 * is not necessary for the mkfs functionality but just an
1816 if (discard_range(fd, 0, 0) == 0) {
1817 if (opflags & PREP_DEVICE_VERBOSE)
1818 printf("Performing full device TRIM %s (%s) ...\n",
1819 file, pretty_size(block_count));
1820 discard_blocks(fd, 0, block_count);
1824 ret = zero_dev_clamped(fd, 0, ZERO_DEV_BYTES, block_count);
1825 for (i = 0 ; !ret && i < BTRFS_SUPER_MIRROR_MAX; i++)
1826 ret = zero_dev_clamped(fd, btrfs_sb_offset(i),
1827 BTRFS_SUPER_INFO_SIZE, block_count);
1828 if (!ret && (opflags & PREP_DEVICE_ZERO_END))
1829 ret = zero_dev_clamped(fd, block_count - ZERO_DEV_BYTES,
1830 ZERO_DEV_BYTES, block_count);
1833 error("failed to zero device '%s': %s", file, strerror(-ret));
1837 ret = btrfs_wipe_existing_sb(fd);
1839 error("cannot wipe superblocks on %s", file);
1843 *block_count_ret = block_count;
1847 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
1848 struct btrfs_root *root, u64 objectid)
1851 struct btrfs_inode_item inode_item;
1852 time_t now = time(NULL);
1854 memset(&inode_item, 0, sizeof(inode_item));
1855 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
1856 btrfs_set_stack_inode_size(&inode_item, 0);
1857 btrfs_set_stack_inode_nlink(&inode_item, 1);
1858 btrfs_set_stack_inode_nbytes(&inode_item, root->nodesize);
1859 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
1860 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
1861 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
1862 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
1863 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
1864 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
1865 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
1866 btrfs_set_stack_timespec_sec(&inode_item.otime, now);
1867 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
1869 if (root->fs_info->tree_root == root)
1870 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
1872 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
1876 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
1880 btrfs_set_root_dirid(&root->root_item, objectid);
1887 * checks if a path is a block device node
1888 * Returns negative errno on failure, otherwise
1889 * returns 1 for blockdev, 0 for not-blockdev
1891 int is_block_device(const char *path)
1893 struct stat statbuf;
1895 if (stat(path, &statbuf) < 0)
1898 return !!S_ISBLK(statbuf.st_mode);
1902 * check if given path is a mount point
1903 * return 1 if yes. 0 if no. -1 for error
1905 int is_mount_point(const char *path)
1911 f = setmntent("/proc/self/mounts", "r");
1915 while ((mnt = getmntent(f)) != NULL) {
1916 if (strcmp(mnt->mnt_dir, path))
1925 static int is_reg_file(const char *path)
1927 struct stat statbuf;
1929 if (stat(path, &statbuf) < 0)
1931 return S_ISREG(statbuf.st_mode);
1935 * This function checks if the given input parameter is
1937 * return <0 : some error in the given input
1938 * return BTRFS_ARG_UNKNOWN: unknown input
1939 * return BTRFS_ARG_UUID: given input is uuid
1940 * return BTRFS_ARG_MNTPOINT: given input is path
1941 * return BTRFS_ARG_REG: given input is regular file
1942 * return BTRFS_ARG_BLKDEV: given input is block device
1944 int check_arg_type(const char *input)
1947 char path[PATH_MAX];
1952 if (realpath(input, path)) {
1953 if (is_block_device(path) == 1)
1954 return BTRFS_ARG_BLKDEV;
1956 if (is_mount_point(path) == 1)
1957 return BTRFS_ARG_MNTPOINT;
1959 if (is_reg_file(path))
1960 return BTRFS_ARG_REG;
1962 return BTRFS_ARG_UNKNOWN;
1965 if (strlen(input) == (BTRFS_UUID_UNPARSED_SIZE - 1) &&
1966 !uuid_parse(input, uuid))
1967 return BTRFS_ARG_UUID;
1969 return BTRFS_ARG_UNKNOWN;
1973 * Find the mount point for a mounted device.
1974 * On success, returns 0 with mountpoint in *mp.
1975 * On failure, returns -errno (not mounted yields -EINVAL)
1976 * Is noisy on failures, expects to be given a mounted device.
1978 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size)
1983 ret = is_block_device(dev);
1986 error("not a block device: %s", dev);
1989 error("cannot check %s: %s", dev, strerror(-ret));
1994 fd = open(dev, O_RDONLY);
1997 error("cannot open %s: %s", dev, strerror(errno));
2001 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
2004 } else { /* mounted, all good */
2014 * Given a pathname, return a filehandle to:
2015 * the original pathname or,
2016 * if the pathname is a mounted btrfs device, to its mountpoint.
2018 * On error, return -1, errno should be set.
2020 int open_path_or_dev_mnt(const char *path, DIR **dirstream, int verbose)
2025 if (is_block_device(path)) {
2026 ret = get_btrfs_mount(path, mp, sizeof(mp));
2028 /* not a mounted btrfs dev */
2029 error_on(verbose, "'%s' is not a mounted btrfs device",
2034 ret = open_file_or_dir(mp, dirstream);
2035 error_on(verbose && ret < 0, "can't access '%s': %s",
2036 path, strerror(errno));
2038 ret = btrfs_open_dir(path, dirstream, 1);
2045 * Do the following checks before calling open_file_or_dir():
2046 * 1: path is in a btrfs filesystem
2047 * 2: path is a directory
2049 int btrfs_open_dir(const char *path, DIR **dirstream, int verbose)
2055 if (statfs(path, &stfs) != 0) {
2056 error_on(verbose, "cannot access '%s': %s", path,
2061 if (stfs.f_type != BTRFS_SUPER_MAGIC) {
2062 error_on(verbose, "not a btrfs filesystem: %s", path);
2066 if (stat(path, &st) != 0) {
2067 error_on(verbose, "cannot access '%s': %s", path,
2072 if (!S_ISDIR(st.st_mode)) {
2073 error_on(verbose, "not a directory: %s", path);
2077 ret = open_file_or_dir(path, dirstream);
2079 error_on(verbose, "cannot access '%s': %s", path,
2086 /* checks if a device is a loop device */
2087 static int is_loop_device (const char* device) {
2088 struct stat statbuf;
2090 if(stat(device, &statbuf) < 0)
2093 return (S_ISBLK(statbuf.st_mode) &&
2094 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
2098 * Takes a loop device path (e.g. /dev/loop0) and returns
2099 * the associated file (e.g. /images/my_btrfs.img) using
2102 static int resolve_loop_device_with_loopdev(const char* loop_dev, char* loop_file)
2106 struct loop_info64 lo64;
2108 fd = open(loop_dev, O_RDONLY | O_NONBLOCK);
2111 ret = ioctl(fd, LOOP_GET_STATUS64, &lo64);
2117 memcpy(loop_file, lo64.lo_file_name, sizeof(lo64.lo_file_name));
2118 loop_file[sizeof(lo64.lo_file_name)] = 0;
2126 /* Takes a loop device path (e.g. /dev/loop0) and returns
2127 * the associated file (e.g. /images/my_btrfs.img) */
2128 static int resolve_loop_device(const char* loop_dev, char* loop_file,
2135 char real_loop_dev[PATH_MAX];
2137 if (!realpath(loop_dev, real_loop_dev))
2139 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
2140 if (!(f = fopen(p, "r"))) {
2141 if (errno == ENOENT)
2143 * It's possibly a partitioned loop device, which is
2144 * resolvable with loopdev API.
2146 return resolve_loop_device_with_loopdev(loop_dev, loop_file);
2150 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
2151 ret = fscanf(f, fmt, loop_file);
2160 * Checks whether a and b are identical or device
2161 * files associated with the same block device
2163 static int is_same_blk_file(const char* a, const char* b)
2165 struct stat st_buf_a, st_buf_b;
2166 char real_a[PATH_MAX];
2167 char real_b[PATH_MAX];
2169 if (!realpath(a, real_a))
2170 strncpy_null(real_a, a);
2172 if (!realpath(b, real_b))
2173 strncpy_null(real_b, b);
2175 /* Identical path? */
2176 if (strcmp(real_a, real_b) == 0)
2179 if (stat(a, &st_buf_a) < 0 || stat(b, &st_buf_b) < 0) {
2180 if (errno == ENOENT)
2185 /* Same blockdevice? */
2186 if (S_ISBLK(st_buf_a.st_mode) && S_ISBLK(st_buf_b.st_mode) &&
2187 st_buf_a.st_rdev == st_buf_b.st_rdev) {
2192 if (st_buf_a.st_dev == st_buf_b.st_dev &&
2193 st_buf_a.st_ino == st_buf_b.st_ino) {
2200 /* checks if a and b are identical or device
2201 * files associated with the same block device or
2202 * if one file is a loop device that uses the other
2205 static int is_same_loop_file(const char* a, const char* b)
2207 char res_a[PATH_MAX];
2208 char res_b[PATH_MAX];
2209 const char* final_a = NULL;
2210 const char* final_b = NULL;
2213 /* Resolve a if it is a loop device */
2214 if((ret = is_loop_device(a)) < 0) {
2219 ret = resolve_loop_device(a, res_a, sizeof(res_a));
2230 /* Resolve b if it is a loop device */
2231 if ((ret = is_loop_device(b)) < 0) {
2236 ret = resolve_loop_device(b, res_b, sizeof(res_b));
2247 return is_same_blk_file(final_a, final_b);
2250 /* Checks if a file exists and is a block or regular file*/
2251 static int is_existing_blk_or_reg_file(const char* filename)
2255 if(stat(filename, &st_buf) < 0) {
2262 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
2265 /* Checks if a file is used (directly or indirectly via a loop device)
2266 * by a device in fs_devices
2268 static int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices,
2272 struct list_head *head;
2273 struct list_head *cur;
2274 struct btrfs_device *device;
2276 head = &fs_devices->devices;
2277 list_for_each(cur, head) {
2278 device = list_entry(cur, struct btrfs_device, dev_list);
2280 if((ret = is_same_loop_file(device->name, file)))
2288 * Resolve a pathname to a device mapper node to /dev/mapper/<name>
2289 * Returns NULL on invalid input or malloc failure; Other failures
2290 * will be handled by the caller using the input pathame.
2292 char *canonicalize_dm_name(const char *ptname)
2296 char path[PATH_MAX], name[PATH_MAX], *res = NULL;
2298 if (!ptname || !*ptname)
2301 snprintf(path, sizeof(path), "/sys/block/%s/dm/name", ptname);
2302 if (!(f = fopen(path, "r")))
2305 /* read <name>\n from sysfs */
2306 if (fgets(name, sizeof(name), f) && (sz = strlen(name)) > 1) {
2307 name[sz - 1] = '\0';
2308 snprintf(path, sizeof(path), "/dev/mapper/%s", name);
2310 if (access(path, F_OK) == 0)
2318 * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or
2319 * to a device mapper pathname.
2320 * Returns NULL on invalid input or malloc failure; Other failures
2321 * will be handled by the caller using the input pathame.
2323 char *canonicalize_path(const char *path)
2325 char *canonical, *p;
2327 if (!path || !*path)
2330 canonical = realpath(path, NULL);
2332 return strdup(path);
2333 p = strrchr(canonical, '/');
2334 if (p && strncmp(p, "/dm-", 4) == 0 && isdigit(*(p + 4))) {
2335 char *dm = canonicalize_dm_name(p + 1);
2346 * returns 1 if the device was mounted, < 0 on error or 0 if everything
2347 * is safe to continue.
2349 int check_mounted(const char* file)
2354 fd = open(file, O_RDONLY);
2356 error("mount check: cannot open %s: %s", file,
2361 ret = check_mounted_where(fd, file, NULL, 0, NULL);
2367 int check_mounted_where(int fd, const char *file, char *where, int size,
2368 struct btrfs_fs_devices **fs_dev_ret)
2373 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2377 /* scan the initial device */
2378 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
2379 &total_devs, BTRFS_SUPER_INFO_OFFSET, SBREAD_DEFAULT);
2380 is_btrfs = (ret >= 0);
2382 /* scan other devices */
2383 if (is_btrfs && total_devs > 1) {
2384 ret = btrfs_scan_devices();
2389 /* iterate over the list of currently mounted filesystems */
2390 if ((f = setmntent ("/proc/self/mounts", "r")) == NULL)
2393 while ((mnt = getmntent (f)) != NULL) {
2395 if(strcmp(mnt->mnt_type, "btrfs") != 0)
2398 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
2400 /* ignore entries in the mount table that are not
2401 associated with a file*/
2402 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
2403 goto out_mntloop_err;
2407 ret = is_same_loop_file(file, mnt->mnt_fsname);
2411 goto out_mntloop_err;
2416 /* Did we find an entry in mnt table? */
2417 if (mnt && size && where) {
2418 strncpy(where, mnt->mnt_dir, size);
2422 *fs_dev_ret = fs_devices_mnt;
2424 ret = (mnt != NULL);
2432 struct pending_dir {
2433 struct list_head list;
2434 char name[PATH_MAX];
2437 int btrfs_register_one_device(const char *fname)
2439 struct btrfs_ioctl_vol_args args;
2443 fd = open("/dev/btrfs-control", O_RDWR);
2446 "failed to open /dev/btrfs-control, skipping device registration: %s",
2450 memset(&args, 0, sizeof(args));
2451 strncpy_null(args.name, fname);
2452 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
2454 error("device scan failed on '%s': %s", fname,
2463 * Register all devices in the fs_uuid list created in the user
2464 * space. Ensure btrfs_scan_devices() is called before this func.
2466 int btrfs_register_all_devices(void)
2470 struct btrfs_fs_devices *fs_devices;
2471 struct btrfs_device *device;
2472 struct list_head *all_uuids;
2474 all_uuids = btrfs_scanned_uuids();
2476 list_for_each_entry(fs_devices, all_uuids, list) {
2477 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2479 err = btrfs_register_one_device(device->name);
2489 int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
2492 struct btrfs_super_block *disk_super;
2496 buf = malloc(BTRFS_SUPER_INFO_SIZE);
2501 ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
2502 if (ret != BTRFS_SUPER_INFO_SIZE)
2506 disk_super = (struct btrfs_super_block *)buf;
2508 * Accept devices from the same filesystem, allow partially created
2511 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC &&
2512 btrfs_super_magic(disk_super) != BTRFS_MAGIC_PARTIAL)
2515 if (!memcmp(disk_super->fsid, root->fs_info->super_copy->fsid,
2525 * Note: this function uses a static per-thread buffer. Do not call this
2526 * function more than 10 times within one argument list!
2528 const char *pretty_size_mode(u64 size, unsigned mode)
2530 static __thread int ps_index = 0;
2531 static __thread char ps_array[10][32];
2534 ret = ps_array[ps_index];
2537 (void)pretty_size_snprintf(size, ret, 32, mode);
2542 static const char* unit_suffix_binary[] =
2543 { "B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB"};
2544 static const char* unit_suffix_decimal[] =
2545 { "B", "kB", "MB", "GB", "TB", "PB", "EB"};
2547 int pretty_size_snprintf(u64 size, char *str, size_t str_size, unsigned unit_mode)
2553 const char** suffix = NULL;
2560 negative = !!(unit_mode & UNITS_NEGATIVE);
2561 unit_mode &= ~UNITS_NEGATIVE;
2563 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_RAW) {
2565 snprintf(str, str_size, "%lld", size);
2567 snprintf(str, str_size, "%llu", size);
2571 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_BINARY) {
2574 suffix = unit_suffix_binary;
2575 } else if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_DECIMAL) {
2578 suffix = unit_suffix_decimal;
2583 fprintf(stderr, "INTERNAL ERROR: unknown unit base, mode %d\n",
2591 switch (unit_mode & UNITS_MODE_MASK) {
2592 case UNITS_TBYTES: base *= mult; num_divs++;
2593 case UNITS_GBYTES: base *= mult; num_divs++;
2594 case UNITS_MBYTES: base *= mult; num_divs++;
2595 case UNITS_KBYTES: num_divs++;
2603 s64 ssize = (s64)size;
2604 s64 last_ssize = ssize;
2606 while ((ssize < 0 ? -ssize : ssize) >= mult) {
2611 last_size = (u64)last_ssize;
2613 while (size >= mult) {
2620 * If the value is smaller than base, we didn't do any
2621 * division, in that case, base should be 1, not original
2622 * base, or the unit will be wrong
2628 if (num_divs >= ARRAY_SIZE(unit_suffix_binary)) {
2630 printf("INTERNAL ERROR: unsupported unit suffix, index %d\n",
2637 fraction = (float)(s64)last_size / base;
2639 fraction = (float)last_size / base;
2642 return snprintf(str, str_size, "%.2f%s", fraction, suffix[num_divs]);
2646 * __strncpy_null - strncpy with null termination
2647 * @dest: the target array
2648 * @src: the source string
2649 * @n: maximum bytes to copy (size of *dest)
2651 * Like strncpy, but ensures destination is null-terminated.
2653 * Copies the string pointed to by src, including the terminating null
2654 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
2655 * of n bytes. Then ensure that dest is null-terminated.
2657 char *__strncpy_null(char *dest, const char *src, size_t n)
2659 strncpy(dest, src, n);
2666 * Checks to make sure that the label matches our requirements.
2668 0 if everything is safe and usable
2669 -1 if the label is too long
2671 static int check_label(const char *input)
2673 int len = strlen(input);
2675 if (len > BTRFS_LABEL_SIZE - 1) {
2676 error("label %s is too long (max %d)", input,
2677 BTRFS_LABEL_SIZE - 1);
2684 static int set_label_unmounted(const char *dev, const char *label)
2686 struct btrfs_trans_handle *trans;
2687 struct btrfs_root *root;
2690 ret = check_mounted(dev);
2692 error("checking mount status of %s failed: %d", dev, ret);
2696 error("device %s is mounted, use mount point", dev);
2700 /* Open the super_block at the default location
2701 * and as read-write.
2703 root = open_ctree(dev, 0, OPEN_CTREE_WRITES);
2704 if (!root) /* errors are printed by open_ctree() */
2707 trans = btrfs_start_transaction(root, 1);
2708 __strncpy_null(root->fs_info->super_copy->label, label, BTRFS_LABEL_SIZE - 1);
2710 btrfs_commit_transaction(trans, root);
2712 /* Now we close it since we are done. */
2717 static int set_label_mounted(const char *mount_path, const char *labelp)
2720 char label[BTRFS_LABEL_SIZE];
2722 fd = open(mount_path, O_RDONLY | O_NOATIME);
2724 error("unable to access %s: %s", mount_path, strerror(errno));
2728 memset(label, 0, sizeof(label));
2729 __strncpy_null(label, labelp, BTRFS_LABEL_SIZE - 1);
2730 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
2731 error("unable to set label of %s: %s", mount_path,
2741 int get_label_unmounted(const char *dev, char *label)
2743 struct btrfs_root *root;
2746 ret = check_mounted(dev);
2748 error("checking mount status of %s failed: %d", dev, ret);
2752 /* Open the super_block at the default location
2755 root = open_ctree(dev, 0, 0);
2759 __strncpy_null(label, root->fs_info->super_copy->label,
2760 BTRFS_LABEL_SIZE - 1);
2762 /* Now we close it since we are done. */
2768 * If a partition is mounted, try to get the filesystem label via its
2769 * mounted path rather than device. Return the corresponding error
2770 * the user specified the device path.
2772 int get_label_mounted(const char *mount_path, char *labelp)
2774 char label[BTRFS_LABEL_SIZE];
2778 fd = open(mount_path, O_RDONLY | O_NOATIME);
2780 error("unable to access %s: %s", mount_path, strerror(errno));
2784 memset(label, '\0', sizeof(label));
2785 ret = ioctl(fd, BTRFS_IOC_GET_FSLABEL, label);
2787 if (errno != ENOTTY)
2788 error("unable to get label of %s: %s", mount_path,
2795 __strncpy_null(labelp, label, BTRFS_LABEL_SIZE - 1);
2800 int get_label(const char *btrfs_dev, char *label)
2804 ret = is_existing_blk_or_reg_file(btrfs_dev);
2806 ret = get_label_mounted(btrfs_dev, label);
2808 ret = get_label_unmounted(btrfs_dev, label);
2813 int set_label(const char *btrfs_dev, const char *label)
2817 if (check_label(label))
2820 ret = is_existing_blk_or_reg_file(btrfs_dev);
2822 ret = set_label_mounted(btrfs_dev, label);
2824 ret = set_label_unmounted(btrfs_dev, label);
2830 * A not-so-good version fls64. No fascinating optimization since
2831 * no one except parse_size use it
2833 static int fls64(u64 x)
2837 for (i = 0; i <64; i++)
2838 if (x << i & (1ULL << 63))
2843 u64 parse_size(char *s)
2851 error("size value is empty");
2855 error("size value '%s' is less equal than 0", s);
2858 ret = strtoull(s, &endptr, 10);
2860 error("size value '%s' is invalid", s);
2863 if (endptr[0] && endptr[1]) {
2864 error("illegal suffix contains character '%c' in wrong position",
2869 * strtoll returns LLONG_MAX when overflow, if this happens,
2870 * need to call strtoull to get the real size
2872 if (errno == ERANGE && ret == ULLONG_MAX) {
2873 error("size value '%s' is too large for u64", s);
2877 c = tolower(endptr[0]);
2900 error("unknown size descriptor '%c'", c);
2904 /* Check whether ret * mult overflow */
2905 if (fls64(ret) + fls64(mult) - 1 > 64) {
2906 error("size value '%s' is too large for u64", s);
2913 u64 parse_qgroupid(const char *p)
2915 char *s = strchr(p, '/');
2916 const char *ptr_src_end = p + strlen(p);
2917 char *ptr_parse_end = NULL;
2926 /* Numeric format like '0/257' is the primary case */
2928 id = strtoull(p, &ptr_parse_end, 10);
2929 if (ptr_parse_end != ptr_src_end)
2933 level = strtoull(p, &ptr_parse_end, 10);
2934 if (ptr_parse_end != s)
2937 id = strtoull(s + 1, &ptr_parse_end, 10);
2938 if (ptr_parse_end != ptr_src_end)
2941 return (level << BTRFS_QGROUP_LEVEL_SHIFT) | id;
2944 /* Path format like subv at 'my_subvol' is the fallback case */
2945 ret = test_issubvolume(p);
2946 if (ret < 0 || !ret)
2948 fd = open(p, O_RDONLY);
2951 ret = lookup_path_rootid(fd, &id);
2953 error("failed to lookup root id: %s", strerror(-ret));
2960 error("invalid qgroupid or subvolume path: %s", p);
2964 int open_file_or_dir3(const char *fname, DIR **dirstream, int open_flags)
2970 ret = stat(fname, &st);
2974 if (S_ISDIR(st.st_mode)) {
2975 *dirstream = opendir(fname);
2978 fd = dirfd(*dirstream);
2979 } else if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) {
2980 fd = open(fname, open_flags);
2983 * we set this on purpose, in case the caller output
2984 * strerror(errno) as success
2992 closedir(*dirstream);
2999 int open_file_or_dir(const char *fname, DIR **dirstream)
3001 return open_file_or_dir3(fname, dirstream, O_RDWR);
3004 void close_file_or_dir(int fd, DIR *dirstream)
3007 closedir(dirstream);
3012 int get_device_info(int fd, u64 devid,
3013 struct btrfs_ioctl_dev_info_args *di_args)
3017 di_args->devid = devid;
3018 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
3020 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
3021 return ret < 0 ? -errno : 0;
3024 static u64 find_max_device_id(struct btrfs_ioctl_search_args *search_args,
3027 struct btrfs_dev_item *dev_item;
3028 char *buf = search_args->buf;
3030 buf += (nr_items - 1) * (sizeof(struct btrfs_ioctl_search_header)
3031 + sizeof(struct btrfs_dev_item));
3032 buf += sizeof(struct btrfs_ioctl_search_header);
3034 dev_item = (struct btrfs_dev_item *)buf;
3036 return btrfs_stack_device_id(dev_item);
3039 static int search_chunk_tree_for_fs_info(int fd,
3040 struct btrfs_ioctl_fs_info_args *fi_args)
3044 u64 start_devid = 1;
3045 struct btrfs_ioctl_search_args search_args;
3046 struct btrfs_ioctl_search_key *search_key = &search_args.key;
3048 fi_args->num_devices = 0;
3050 max_items = BTRFS_SEARCH_ARGS_BUFSIZE
3051 / (sizeof(struct btrfs_ioctl_search_header)
3052 + sizeof(struct btrfs_dev_item));
3054 search_key->tree_id = BTRFS_CHUNK_TREE_OBJECTID;
3055 search_key->min_objectid = BTRFS_DEV_ITEMS_OBJECTID;
3056 search_key->max_objectid = BTRFS_DEV_ITEMS_OBJECTID;
3057 search_key->min_type = BTRFS_DEV_ITEM_KEY;
3058 search_key->max_type = BTRFS_DEV_ITEM_KEY;
3059 search_key->min_transid = 0;
3060 search_key->max_transid = (u64)-1;
3061 search_key->nr_items = max_items;
3062 search_key->max_offset = (u64)-1;
3065 search_key->min_offset = start_devid;
3067 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH, &search_args);
3071 fi_args->num_devices += (u64)search_key->nr_items;
3073 if (search_key->nr_items == max_items) {
3074 start_devid = find_max_device_id(&search_args,
3075 search_key->nr_items) + 1;
3079 /* get the lastest max_id to stay consistent with the num_devices */
3080 if (search_key->nr_items == 0)
3082 * last tree_search returns an empty buf, use the devid of
3083 * the last dev_item of the previous tree_search
3085 fi_args->max_id = start_devid - 1;
3087 fi_args->max_id = find_max_device_id(&search_args,
3088 search_key->nr_items);
3094 * For a given path, fill in the ioctl fs_ and info_ args.
3095 * If the path is a btrfs mountpoint, fill info for all devices.
3096 * If the path is a btrfs device, fill in only that device.
3098 * The path provided must be either on a mounted btrfs fs,
3099 * or be a mounted btrfs device.
3101 * Returns 0 on success, or a negative errno.
3103 int get_fs_info(const char *path, struct btrfs_ioctl_fs_info_args *fi_args,
3104 struct btrfs_ioctl_dev_info_args **di_ret)
3111 struct btrfs_fs_devices *fs_devices_mnt = NULL;
3112 struct btrfs_ioctl_dev_info_args *di_args;
3113 struct btrfs_ioctl_dev_info_args tmp;
3115 DIR *dirstream = NULL;
3117 memset(fi_args, 0, sizeof(*fi_args));
3119 if (is_block_device(path) == 1) {
3120 struct btrfs_super_block *disk_super;
3121 char buf[BTRFS_SUPER_INFO_SIZE];
3123 /* Ensure it's mounted, then set path to the mountpoint */
3124 fd = open(path, O_RDONLY);
3127 error("cannot open %s: %s", path, strerror(errno));
3130 ret = check_mounted_where(fd, path, mp, sizeof(mp),
3139 /* Only fill in this one device */
3140 fi_args->num_devices = 1;
3142 disk_super = (struct btrfs_super_block *)buf;
3143 ret = btrfs_read_dev_super(fd, disk_super,
3144 BTRFS_SUPER_INFO_OFFSET, 0);
3149 last_devid = btrfs_stack_device_id(&disk_super->dev_item);
3150 fi_args->max_id = last_devid;
3152 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
3156 /* at this point path must not be for a block device */
3157 fd = open_file_or_dir(path, &dirstream);
3163 /* fill in fi_args if not just a single device */
3164 if (fi_args->num_devices != 1) {
3165 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
3172 * The fs_args->num_devices does not include seed devices
3174 ret = search_chunk_tree_for_fs_info(fd, fi_args);
3179 * search_chunk_tree_for_fs_info() will lacks the devid 0
3180 * so manual probe for it here.
3182 ret = get_device_info(fd, 0, &tmp);
3184 fi_args->num_devices++;
3187 if (last_devid == 0)
3192 if (!fi_args->num_devices)
3195 di_args = *di_ret = malloc((fi_args->num_devices) * sizeof(*di_args));
3202 memcpy(di_args, &tmp, sizeof(tmp));
3203 for (; last_devid <= fi_args->max_id; last_devid++) {
3204 ret = get_device_info(fd, last_devid, &di_args[ndevs]);
3213 * only when the only dev we wanted to find is not there then
3214 * let any error be returned
3216 if (fi_args->num_devices != 1) {
3222 close_file_or_dir(fd, dirstream);
3226 #define isoctal(c) (((c) & ~7) == '0')
3228 static inline void translate(char *f, char *t)
3230 while (*f != '\0') {
3232 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
3233 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
3243 * Checks if the swap device.
3244 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
3246 static int is_swap_device(const char *file)
3257 if (stat(file, &st_buf) < 0)
3259 if (S_ISBLK(st_buf.st_mode))
3260 dev = st_buf.st_rdev;
3261 else if (S_ISREG(st_buf.st_mode)) {
3262 dev = st_buf.st_dev;
3263 ino = st_buf.st_ino;
3267 if ((f = fopen("/proc/swaps", "r")) == NULL)
3270 /* skip the first line */
3271 if (fgets(tmp, sizeof(tmp), f) == NULL)
3274 while (fgets(tmp, sizeof(tmp), f) != NULL) {
3275 if ((cp = strchr(tmp, ' ')) != NULL)
3277 if ((cp = strchr(tmp, '\t')) != NULL)
3279 translate(tmp, buf);
3280 if (stat(buf, &st_buf) != 0)
3282 if (S_ISBLK(st_buf.st_mode)) {
3283 if (dev == st_buf.st_rdev) {
3287 } else if (S_ISREG(st_buf.st_mode)) {
3288 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {
3302 * Check for existing filesystem or partition table on device.
3304 * 1 for existing fs or partition
3305 * 0 for nothing found
3306 * -1 for internal error
3308 static int check_overwrite(const char *device)
3311 blkid_probe pr = NULL;
3315 if (!device || !*device)
3318 ret = -1; /* will reset on success of all setup calls */
3320 pr = blkid_new_probe_from_filename(device);
3324 size = blkid_probe_get_size(pr);
3328 /* nothing to overwrite on a 0-length device */
3334 ret = blkid_probe_enable_partitions(pr, 1);
3338 ret = blkid_do_fullprobe(pr);
3343 * Blkid returns 1 for nothing found and 0 when it finds a signature,
3344 * but we want the exact opposite, so reverse the return value here.
3346 * In addition print some useful diagnostics about what actually is
3354 if (!blkid_probe_lookup_value(pr, "TYPE", &type, NULL)) {
3356 "%s appears to contain an existing "
3357 "filesystem (%s).\n", device, type);
3358 } else if (!blkid_probe_lookup_value(pr, "PTTYPE", &type, NULL)) {
3360 "%s appears to contain a partition "
3361 "table (%s).\n", device, type);
3364 "%s appears to contain something weird "
3365 "according to blkid\n", device);
3371 blkid_free_probe(pr);
3374 "probe of %s failed, cannot detect "
3375 "existing filesystem.\n", device);
3379 static int group_profile_devs_min(u64 flag)
3381 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3382 case 0: /* single */
3383 case BTRFS_BLOCK_GROUP_DUP:
3385 case BTRFS_BLOCK_GROUP_RAID0:
3386 case BTRFS_BLOCK_GROUP_RAID1:
3387 case BTRFS_BLOCK_GROUP_RAID5:
3389 case BTRFS_BLOCK_GROUP_RAID6:
3391 case BTRFS_BLOCK_GROUP_RAID10:
3398 int test_num_disk_vs_raid(u64 metadata_profile, u64 data_profile,
3399 u64 dev_cnt, int mixed, int ssd)
3402 u64 profile = metadata_profile | data_profile;
3407 allowed |= BTRFS_BLOCK_GROUP_RAID10;
3409 allowed |= BTRFS_BLOCK_GROUP_RAID6;
3411 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
3412 BTRFS_BLOCK_GROUP_RAID5;
3414 allowed |= BTRFS_BLOCK_GROUP_DUP;
3417 if (dev_cnt > 1 && profile & BTRFS_BLOCK_GROUP_DUP) {
3418 warning("DUP is not recommended on filesystem with multiple devices");
3420 if (metadata_profile & ~allowed) {
3422 "ERROR: unable to create FS with metadata profile %s "
3423 "(have %llu devices but %d devices are required)\n",
3424 btrfs_group_profile_str(metadata_profile), dev_cnt,
3425 group_profile_devs_min(metadata_profile));
3428 if (data_profile & ~allowed) {
3430 "ERROR: unable to create FS with data profile %s "
3431 "(have %llu devices but %d devices are required)\n",
3432 btrfs_group_profile_str(data_profile), dev_cnt,
3433 group_profile_devs_min(data_profile));
3437 if (dev_cnt == 3 && profile & BTRFS_BLOCK_GROUP_RAID6) {
3438 warning("RAID6 is not recommended on filesystem with 3 devices only");
3440 if (dev_cnt == 2 && profile & BTRFS_BLOCK_GROUP_RAID5) {
3441 warning("RAID5 is not recommended on filesystem with 2 devices only");
3443 warning_on(!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP) && ssd,
3444 "DUP may not actually lead to 2 copies on the device, see manual page");
3449 int group_profile_max_safe_loss(u64 flags)
3451 switch (flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3452 case 0: /* single */
3453 case BTRFS_BLOCK_GROUP_DUP:
3454 case BTRFS_BLOCK_GROUP_RAID0:
3456 case BTRFS_BLOCK_GROUP_RAID1:
3457 case BTRFS_BLOCK_GROUP_RAID5:
3458 case BTRFS_BLOCK_GROUP_RAID10:
3460 case BTRFS_BLOCK_GROUP_RAID6:
3468 * Check if a device is suitable for btrfs
3470 * 1: something is wrong, an error is printed
3473 int test_dev_for_mkfs(const char *file, int force_overwrite)
3478 ret = is_swap_device(file);
3480 error("checking status of %s: %s", file, strerror(-ret));
3484 error("%s is a swap device", file);
3487 if (!force_overwrite) {
3488 if (check_overwrite(file)) {
3489 error("use the -f option to force overwrite of %s",
3494 ret = check_mounted(file);
3496 error("cannot check mount status of %s: %s", file,
3501 error("%s is mounted", file);
3504 /* check if the device is busy */
3505 fd = open(file, O_RDWR|O_EXCL);
3507 error("unable to open %s: %s", file, strerror(errno));
3510 if (fstat(fd, &st)) {
3511 error("unable to stat %s: %s", file, strerror(errno));
3515 if (!S_ISBLK(st.st_mode)) {
3516 error("%s is not a block device", file);
3524 int btrfs_scan_devices(void)
3529 struct btrfs_fs_devices *tmp_devices;
3530 blkid_dev_iterate iter = NULL;
3531 blkid_dev dev = NULL;
3532 blkid_cache cache = NULL;
3533 char path[PATH_MAX];
3535 if (btrfs_scan_done)
3538 if (blkid_get_cache(&cache, NULL) < 0) {
3539 error("blkid cache get failed");
3542 blkid_probe_all(cache);
3543 iter = blkid_dev_iterate_begin(cache);
3544 blkid_dev_set_search(iter, "TYPE", "btrfs");
3545 while (blkid_dev_next(iter, &dev) == 0) {
3546 dev = blkid_verify(cache, dev);
3549 /* if we are here its definitely a btrfs disk*/
3550 strncpy_null(path, blkid_dev_devname(dev));
3552 fd = open(path, O_RDONLY);
3554 error("cannot open %s: %s", path, strerror(errno));
3557 ret = btrfs_scan_one_device(fd, path, &tmp_devices,
3558 &num_devices, BTRFS_SUPER_INFO_OFFSET,
3561 error("cannot scan %s: %s", path, strerror(-ret));
3568 blkid_dev_iterate_end(iter);
3569 blkid_put_cache(cache);
3571 btrfs_scan_done = 1;
3576 int is_vol_small(const char *file)
3583 fd = open(file, O_RDONLY);
3586 if (fstat(fd, &st) < 0) {
3591 size = btrfs_device_size(fd, &st);
3596 if (size < BTRFS_MKFS_SMALL_VOLUME_SIZE) {
3606 * This reads a line from the stdin and only returns non-zero if the
3607 * first whitespace delimited token is a case insensitive match with yes
3610 int ask_user(const char *question)
3612 char buf[30] = {0,};
3613 char *saveptr = NULL;
3616 printf("%s [y/N]: ", question);
3618 return fgets(buf, sizeof(buf) - 1, stdin) &&
3619 (answer = strtok_r(buf, " \t\n\r", &saveptr)) &&
3620 (!strcasecmp(answer, "yes") || !strcasecmp(answer, "y"));
3624 * return 0 if a btrfs mount point is found
3625 * return 1 if a mount point is found but not btrfs
3626 * return <0 if something goes wrong
3628 int find_mount_root(const char *path, char **mount_root)
3636 int longest_matchlen = 0;
3637 char *longest_match = NULL;
3639 fd = open(path, O_RDONLY | O_NOATIME);
3644 mnttab = setmntent("/proc/self/mounts", "r");
3648 while ((ent = getmntent(mnttab))) {
3649 len = strlen(ent->mnt_dir);
3650 if (strncmp(ent->mnt_dir, path, len) == 0) {
3651 /* match found and use the latest match */
3652 if (longest_matchlen <= len) {
3653 free(longest_match);
3654 longest_matchlen = len;
3655 longest_match = strdup(ent->mnt_dir);
3656 not_btrfs = strcmp(ent->mnt_type, "btrfs");
3665 free(longest_match);
3670 *mount_root = realpath(longest_match, NULL);
3674 free(longest_match);
3678 int test_minimum_size(const char *file, u32 nodesize)
3681 struct stat statbuf;
3683 fd = open(file, O_RDONLY);
3686 if (stat(file, &statbuf) < 0) {
3690 if (btrfs_device_size(fd, &statbuf) < btrfs_min_dev_size(nodesize)) {
3700 * Test if path is a directory
3702 * 0 - path exists but it is not a directory
3703 * 1 - path exists and it is a directory
3706 int test_isdir(const char *path)
3711 ret = stat(path, &st);
3715 return !!S_ISDIR(st.st_mode);
3718 void units_set_mode(unsigned *units, unsigned mode)
3720 unsigned base = *units & UNITS_MODE_MASK;
3722 *units = base | mode;
3725 void units_set_base(unsigned *units, unsigned base)
3727 unsigned mode = *units & ~UNITS_MODE_MASK;
3729 *units = base | mode;
3732 int find_next_key(struct btrfs_path *path, struct btrfs_key *key)
3736 for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
3737 if (!path->nodes[level])
3739 if (path->slots[level] + 1 >=
3740 btrfs_header_nritems(path->nodes[level]))
3743 btrfs_item_key_to_cpu(path->nodes[level], key,
3744 path->slots[level] + 1);
3746 btrfs_node_key_to_cpu(path->nodes[level], key,
3747 path->slots[level] + 1);
3753 const char* btrfs_group_type_str(u64 flag)
3755 u64 mask = BTRFS_BLOCK_GROUP_TYPE_MASK |
3756 BTRFS_SPACE_INFO_GLOBAL_RSV;
3758 switch (flag & mask) {
3759 case BTRFS_BLOCK_GROUP_DATA:
3761 case BTRFS_BLOCK_GROUP_SYSTEM:
3763 case BTRFS_BLOCK_GROUP_METADATA:
3765 case BTRFS_BLOCK_GROUP_DATA|BTRFS_BLOCK_GROUP_METADATA:
3766 return "Data+Metadata";
3767 case BTRFS_SPACE_INFO_GLOBAL_RSV:
3768 return "GlobalReserve";
3774 const char* btrfs_group_profile_str(u64 flag)
3776 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3779 case BTRFS_BLOCK_GROUP_RAID0:
3781 case BTRFS_BLOCK_GROUP_RAID1:
3783 case BTRFS_BLOCK_GROUP_RAID5:
3785 case BTRFS_BLOCK_GROUP_RAID6:
3787 case BTRFS_BLOCK_GROUP_DUP:
3789 case BTRFS_BLOCK_GROUP_RAID10:
3796 u64 disk_size(const char *path)
3800 if (statfs(path, &sfs) < 0)
3803 return sfs.f_bsize * sfs.f_blocks;
3806 u64 get_partition_size(const char *dev)
3809 int fd = open(dev, O_RDONLY);
3813 if (ioctl(fd, BLKGETSIZE64, &result) < 0) {
3823 * Check if the BTRFS_IOC_TREE_SEARCH_V2 ioctl is supported on a given
3824 * filesystem, opened at fd
3826 int btrfs_tree_search2_ioctl_supported(int fd)
3828 struct btrfs_ioctl_search_args_v2 *args2;
3829 struct btrfs_ioctl_search_key *sk;
3830 int args2_size = 1024;
3831 char args2_buf[args2_size];
3834 args2 = (struct btrfs_ioctl_search_args_v2 *)args2_buf;
3838 * Search for the extent tree item in the root tree.
3840 sk->tree_id = BTRFS_ROOT_TREE_OBJECTID;
3841 sk->min_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3842 sk->max_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3843 sk->min_type = BTRFS_ROOT_ITEM_KEY;
3844 sk->max_type = BTRFS_ROOT_ITEM_KEY;
3846 sk->max_offset = (u64)-1;
3847 sk->min_transid = 0;
3848 sk->max_transid = (u64)-1;
3850 args2->buf_size = args2_size - sizeof(struct btrfs_ioctl_search_args_v2);
3851 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH_V2, args2);
3852 if (ret == -EOPNOTSUPP)
3859 int btrfs_check_nodesize(u32 nodesize, u32 sectorsize, u64 features)
3861 if (nodesize < sectorsize) {
3862 error("illegal nodesize %u (smaller than %u)",
3863 nodesize, sectorsize);
3865 } else if (nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) {
3866 error("illegal nodesize %u (larger than %u)",
3867 nodesize, BTRFS_MAX_METADATA_BLOCKSIZE);
3869 } else if (nodesize & (sectorsize - 1)) {
3870 error("illegal nodesize %u (not aligned to %u)",
3871 nodesize, sectorsize);
3873 } else if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS &&
3874 nodesize != sectorsize) {
3875 error("illegal nodesize %u (not equal to %u for mixed block group)",
3876 nodesize, sectorsize);
3883 * Copy a path argument from SRC to DEST and check the SRC length if it's at
3884 * most PATH_MAX and fits into DEST. DESTLEN is supposed to be exact size of
3886 * The destination buffer is zero terminated.
3887 * Return < 0 for error, 0 otherwise.
3889 int arg_copy_path(char *dest, const char *src, int destlen)
3891 size_t len = strlen(src);
3893 if (len >= PATH_MAX || len >= destlen)
3894 return -ENAMETOOLONG;
3896 __strncpy_null(dest, src, destlen);
3901 unsigned int get_unit_mode_from_arg(int *argc, char *argv[], int df_mode)
3903 unsigned int unit_mode = UNITS_DEFAULT;
3907 for (arg_i = 0; arg_i < *argc; arg_i++) {
3908 if (!strcmp(argv[arg_i], "--"))
3911 if (!strcmp(argv[arg_i], "--raw")) {
3912 unit_mode = UNITS_RAW;
3916 if (!strcmp(argv[arg_i], "--human-readable")) {
3917 unit_mode = UNITS_HUMAN_BINARY;
3922 if (!strcmp(argv[arg_i], "--iec")) {
3923 units_set_mode(&unit_mode, UNITS_BINARY);
3927 if (!strcmp(argv[arg_i], "--si")) {
3928 units_set_mode(&unit_mode, UNITS_DECIMAL);
3933 if (!strcmp(argv[arg_i], "--kbytes")) {
3934 units_set_base(&unit_mode, UNITS_KBYTES);
3938 if (!strcmp(argv[arg_i], "--mbytes")) {
3939 units_set_base(&unit_mode, UNITS_MBYTES);
3943 if (!strcmp(argv[arg_i], "--gbytes")) {
3944 units_set_base(&unit_mode, UNITS_GBYTES);
3948 if (!strcmp(argv[arg_i], "--tbytes")) {
3949 units_set_base(&unit_mode, UNITS_TBYTES);
3957 if (!strcmp(argv[arg_i], "-b")) {
3958 unit_mode = UNITS_RAW;
3962 if (!strcmp(argv[arg_i], "-h")) {
3963 unit_mode = UNITS_HUMAN_BINARY;
3967 if (!strcmp(argv[arg_i], "-H")) {
3968 unit_mode = UNITS_HUMAN_DECIMAL;
3972 if (!strcmp(argv[arg_i], "-k")) {
3973 units_set_base(&unit_mode, UNITS_KBYTES);
3977 if (!strcmp(argv[arg_i], "-m")) {
3978 units_set_base(&unit_mode, UNITS_MBYTES);
3982 if (!strcmp(argv[arg_i], "-g")) {
3983 units_set_base(&unit_mode, UNITS_GBYTES);
3987 if (!strcmp(argv[arg_i], "-t")) {
3988 units_set_base(&unit_mode, UNITS_TBYTES);
3994 for (arg_i = 0, arg_end = 0; arg_i < *argc; arg_i++) {
3997 argv[arg_end] = argv[arg_i];
4006 int string_is_numerical(const char *str)
4010 if (!(*str >= '0' && *str <= '9'))
4012 while (*str >= '0' && *str <= '9')
4019 /* Subvolume helper functions */
4021 * test if name is a correct subvolume name
4022 * this function return
4023 * 0-> name is not a correct subvolume name
4024 * 1-> name is a correct subvolume name
4026 int test_issubvolname(const char *name)
4028 return name[0] != '\0' && !strchr(name, '/') &&
4029 strcmp(name, ".") && strcmp(name, "..");
4033 * Test if path is a subvolume
4035 * 0 - path exists but it is not a subvolume
4036 * 1 - path exists and it is a subvolume
4039 int test_issubvolume(const char *path)
4045 res = stat(path, &st);
4049 if (st.st_ino != BTRFS_FIRST_FREE_OBJECTID || !S_ISDIR(st.st_mode))
4052 res = statfs(path, &stfs);
4056 return (int)stfs.f_type == BTRFS_SUPER_MAGIC;
4059 const char *subvol_strip_mountpoint(const char *mnt, const char *full_path)
4061 int len = strlen(mnt);
4065 if (mnt[len - 1] != '/')
4068 return full_path + len;
4075 * 1: Error; and error info printed to the terminal. Fixme.
4076 * 2: If the fullpath is root tree instead of subvol tree
4078 int get_subvol_info(const char *fullpath, struct root_info *get_ri)
4085 const char *svpath = NULL;
4086 DIR *dirstream1 = NULL;
4087 DIR *dirstream2 = NULL;
4089 ret = test_issubvolume(fullpath);
4093 error("not a subvolume: %s", fullpath);
4097 ret = find_mount_root(fullpath, &mnt);
4101 error("%s doesn't belong to btrfs mount point", fullpath);
4105 svpath = subvol_strip_mountpoint(mnt, fullpath);
4107 fd = btrfs_open_dir(fullpath, &dirstream1, 1);
4111 ret = btrfs_list_get_path_rootid(fd, &sv_id);
4115 mntfd = btrfs_open_dir(mnt, &dirstream2, 1);
4119 memset(get_ri, 0, sizeof(*get_ri));
4120 get_ri->root_id = sv_id;
4122 if (sv_id == BTRFS_FS_TREE_OBJECTID)
4123 ret = btrfs_get_toplevel_subvol(mntfd, get_ri);
4125 ret = btrfs_get_subvol(mntfd, get_ri);
4127 error("can't find '%s': %d", svpath, ret);
4130 close_file_or_dir(mntfd, dirstream2);
4131 close_file_or_dir(fd, dirstream1);
4137 void init_rand_seed(u64 seed)
4141 /* only use the last 48 bits */
4142 for (i = 0; i < 3; i++) {
4143 rand_seed[i] = (unsigned short)(seed ^ (unsigned short)(-1));
4146 rand_seed_initlized = 1;
4149 static void __init_seed(void)
4155 if(rand_seed_initlized)
4157 /* Use urandom as primary seed source. */
4158 fd = open("/dev/urandom", O_RDONLY);
4160 ret = read(fd, rand_seed, sizeof(rand_seed));
4162 if (ret < sizeof(rand_seed))
4166 /* Use time and pid as fallback seed */
4167 warning("failed to read /dev/urandom, use time and pid as random seed");
4168 gettimeofday(&tv, 0);
4169 rand_seed[0] = getpid() ^ (tv.tv_sec & 0xFFFF);
4170 rand_seed[1] = getppid() ^ (tv.tv_usec & 0xFFFF);
4171 rand_seed[2] = (tv.tv_sec ^ tv.tv_usec) >> 16;
4173 rand_seed_initlized = 1;
4180 * Don't use nrand48, its range is [0,2^31) The highest bit will alwasy
4181 * be 0. Use jrand48 to include the highest bit.
4183 return (u32)jrand48(rand_seed);
4186 unsigned int rand_range(unsigned int upper)
4190 * Use the full 48bits to mod, which would be more uniformly
4193 return (unsigned int)(jrand48(rand_seed) % upper);
4196 void btrfs_config_init(void)