2 * Copyright (C) 2007 Oracle. All rights reserved.
3 * Copyright (C) 2008 Morey Roof. All rights reserved.
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public
7 * License v2 as published by the Free Software Foundation.
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * General Public License for more details.
14 * You should have received a copy of the GNU General Public
15 * License along with this program; if not, write to the
16 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
17 * Boston, MA 021110-1307, USA.
23 #include <sys/ioctl.h>
24 #include <sys/mount.h>
25 #include <sys/types.h>
27 #include <uuid/uuid.h>
32 #include <linux/loop.h>
33 #include <linux/major.h>
34 #include <linux/kdev_t.h>
36 #include <blkid/blkid.h>
38 #include <sys/statfs.h>
39 #include <linux/magic.h>
41 #include <sys/utsname.h>
42 #include <linux/version.h>
44 #include "kerncompat.h"
45 #include "radix-tree.h"
48 #include "transaction.h"
56 #define BLKDISCARD _IO(0x12,119)
59 static int btrfs_scan_done = 0;
61 static char argv0_buf[ARGV0_BUF_SIZE] = "btrfs";
63 static int rand_seed_initlized = 0;
64 static unsigned short rand_seed[3];
66 struct btrfs_config bconf;
68 const char *get_argv0_buf(void)
73 void fixup_argv0(char **argv, const char *token)
75 int len = strlen(argv0_buf);
77 snprintf(argv0_buf + len, sizeof(argv0_buf) - len, " %s", token);
81 void set_argv0(char **argv)
83 strncpy(argv0_buf, argv[0], sizeof(argv0_buf));
84 argv0_buf[sizeof(argv0_buf) - 1] = 0;
87 int check_argc_exact(int nargs, int expected)
90 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
92 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
94 return nargs != expected;
97 int check_argc_min(int nargs, int expected)
99 if (nargs < expected) {
100 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
107 int check_argc_max(int nargs, int expected)
109 if (nargs > expected) {
110 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
119 * Discard the given range in one go
121 static int discard_range(int fd, u64 start, u64 len)
123 u64 range[2] = { start, len };
125 if (ioctl(fd, BLKDISCARD, &range) < 0)
131 * Discard blocks in the given range in 1G chunks, the process is interruptible
133 static int discard_blocks(int fd, u64 start, u64 len)
137 u64 chunk_size = min_t(u64, len, SZ_1G);
140 ret = discard_range(fd, start, chunk_size);
150 static u64 reference_root_table[] = {
151 [1] = BTRFS_ROOT_TREE_OBJECTID,
152 [2] = BTRFS_EXTENT_TREE_OBJECTID,
153 [3] = BTRFS_CHUNK_TREE_OBJECTID,
154 [4] = BTRFS_DEV_TREE_OBJECTID,
155 [5] = BTRFS_FS_TREE_OBJECTID,
156 [6] = BTRFS_CSUM_TREE_OBJECTID,
159 int test_uuid_unique(char *fs_uuid)
162 blkid_dev_iterate iter = NULL;
163 blkid_dev dev = NULL;
164 blkid_cache cache = NULL;
166 if (blkid_get_cache(&cache, NULL) < 0) {
167 printf("ERROR: lblkid cache get failed\n");
170 blkid_probe_all(cache);
171 iter = blkid_dev_iterate_begin(cache);
172 blkid_dev_set_search(iter, "UUID", fs_uuid);
174 while (blkid_dev_next(iter, &dev) == 0) {
175 dev = blkid_verify(cache, dev);
182 blkid_dev_iterate_end(iter);
183 blkid_put_cache(cache);
189 * Reserve space from free_tree.
190 * The algorithm is very simple, find the first cache_extent with enough space
191 * and allocate from its beginning.
193 static int reserve_free_space(struct cache_tree *free_tree, u64 len,
196 struct cache_extent *cache;
199 ASSERT(ret_start != NULL);
200 cache = first_cache_extent(free_tree);
202 if (cache->size > len) {
204 *ret_start = cache->start;
207 if (cache->size == 0) {
208 remove_cache_extent(free_tree, cache);
215 cache = next_cache_extent(cache);
222 static inline int write_temp_super(int fd, struct btrfs_super_block *sb,
228 crc = btrfs_csum_data(NULL, (char *)sb + BTRFS_CSUM_SIZE, crc,
229 BTRFS_SUPER_INFO_SIZE - BTRFS_CSUM_SIZE);
230 btrfs_csum_final(crc, &sb->csum[0]);
231 ret = pwrite(fd, sb, BTRFS_SUPER_INFO_SIZE, sb_bytenr);
232 if (ret < BTRFS_SUPER_INFO_SIZE)
233 ret = (ret < 0 ? -errno : -EIO);
240 * Setup temporary superblock at cfg->super_bynter
241 * Needed info are extracted from cfg, and root_bytenr, chunk_bytenr
243 * For now sys chunk array will be empty and dev_item is empty too.
244 * They will be re-initialized at temp chunk tree setup.
246 * The superblock signature is not valid, denotes a partially created
247 * filesystem, needs to be finalized.
249 static int setup_temp_super(int fd, struct btrfs_mkfs_config *cfg,
250 u64 root_bytenr, u64 chunk_bytenr)
252 unsigned char chunk_uuid[BTRFS_UUID_SIZE];
253 char super_buf[BTRFS_SUPER_INFO_SIZE];
254 struct btrfs_super_block *super = (struct btrfs_super_block *)super_buf;
257 memset(super_buf, 0, BTRFS_SUPER_INFO_SIZE);
258 cfg->num_bytes = round_down(cfg->num_bytes, cfg->sectorsize);
261 if (uuid_parse(cfg->fs_uuid, super->fsid) != 0) {
262 error("cound not parse UUID: %s", cfg->fs_uuid);
266 if (!test_uuid_unique(cfg->fs_uuid)) {
267 error("non-unique UUID: %s", cfg->fs_uuid);
272 uuid_generate(super->fsid);
273 uuid_unparse(super->fsid, cfg->fs_uuid);
275 uuid_generate(chunk_uuid);
276 uuid_unparse(chunk_uuid, cfg->chunk_uuid);
278 btrfs_set_super_bytenr(super, cfg->super_bytenr);
279 btrfs_set_super_num_devices(super, 1);
280 btrfs_set_super_magic(super, BTRFS_MAGIC_PARTIAL);
281 btrfs_set_super_generation(super, 1);
282 btrfs_set_super_root(super, root_bytenr);
283 btrfs_set_super_chunk_root(super, chunk_bytenr);
284 btrfs_set_super_total_bytes(super, cfg->num_bytes);
286 * Temporary filesystem will only have 6 tree roots:
287 * chunk tree, root tree, extent_tree, device tree, fs tree
290 btrfs_set_super_bytes_used(super, 6 * cfg->nodesize);
291 btrfs_set_super_sectorsize(super, cfg->sectorsize);
292 btrfs_set_super_leafsize(super, cfg->nodesize);
293 btrfs_set_super_nodesize(super, cfg->nodesize);
294 btrfs_set_super_stripesize(super, cfg->stripesize);
295 btrfs_set_super_csum_type(super, BTRFS_CSUM_TYPE_CRC32);
296 btrfs_set_super_chunk_root(super, chunk_bytenr);
297 btrfs_set_super_cache_generation(super, -1);
298 btrfs_set_super_incompat_flags(super, cfg->features);
300 __strncpy_null(super->label, cfg->label, BTRFS_LABEL_SIZE - 1);
302 /* Sys chunk array will be re-initialized at chunk tree init time */
303 super->sys_chunk_array_size = 0;
305 ret = write_temp_super(fd, super, cfg->super_bytenr);
311 * Setup an extent buffer for tree block.
313 static int setup_temp_extent_buffer(struct extent_buffer *buf,
314 struct btrfs_mkfs_config *cfg,
315 u64 bytenr, u64 owner)
317 unsigned char fsid[BTRFS_FSID_SIZE];
318 unsigned char chunk_uuid[BTRFS_UUID_SIZE];
321 ret = uuid_parse(cfg->fs_uuid, fsid);
324 ret = uuid_parse(cfg->chunk_uuid, chunk_uuid);
328 memset(buf->data, 0, cfg->nodesize);
329 buf->len = cfg->nodesize;
330 btrfs_set_header_bytenr(buf, bytenr);
331 btrfs_set_header_generation(buf, 1);
332 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
333 btrfs_set_header_owner(buf, owner);
334 btrfs_set_header_flags(buf, BTRFS_HEADER_FLAG_WRITTEN);
335 write_extent_buffer(buf, chunk_uuid, btrfs_header_chunk_tree_uuid(buf),
337 write_extent_buffer(buf, fsid, btrfs_header_fsid(), BTRFS_FSID_SIZE);
341 static inline int write_temp_extent_buffer(int fd, struct extent_buffer *buf,
346 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
348 /* Temporary extent buffer is always mapped 1:1 on disk */
349 ret = pwrite(fd, buf->data, buf->len, bytenr);
351 ret = (ret < 0 ? ret : -EIO);
358 * Insert a root item for temporary tree root
360 * Only used in make_btrfs_v2().
362 static void insert_temp_root_item(struct extent_buffer *buf,
363 struct btrfs_mkfs_config *cfg,
364 int *slot, u32 *itemoff, u64 objectid,
367 struct btrfs_root_item root_item;
368 struct btrfs_inode_item *inode_item;
369 struct btrfs_disk_key disk_key;
371 btrfs_set_header_nritems(buf, *slot + 1);
372 (*itemoff) -= sizeof(root_item);
373 memset(&root_item, 0, sizeof(root_item));
374 inode_item = &root_item.inode;
375 btrfs_set_stack_inode_generation(inode_item, 1);
376 btrfs_set_stack_inode_size(inode_item, 3);
377 btrfs_set_stack_inode_nlink(inode_item, 1);
378 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
379 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
380 btrfs_set_root_refs(&root_item, 1);
381 btrfs_set_root_used(&root_item, cfg->nodesize);
382 btrfs_set_root_generation(&root_item, 1);
383 btrfs_set_root_bytenr(&root_item, bytenr);
385 memset(&disk_key, 0, sizeof(disk_key));
386 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
387 btrfs_set_disk_key_objectid(&disk_key, objectid);
388 btrfs_set_disk_key_offset(&disk_key, 0);
390 btrfs_set_item_key(buf, &disk_key, *slot);
391 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
392 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(root_item));
393 write_extent_buffer(buf, &root_item,
394 btrfs_item_ptr_offset(buf, *slot),
399 static int setup_temp_root_tree(int fd, struct btrfs_mkfs_config *cfg,
400 u64 root_bytenr, u64 extent_bytenr,
401 u64 dev_bytenr, u64 fs_bytenr, u64 csum_bytenr)
403 struct extent_buffer *buf = NULL;
404 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
409 * Provided bytenr must in ascending order, or tree root will have a
412 if (!(root_bytenr < extent_bytenr && extent_bytenr < dev_bytenr &&
413 dev_bytenr < fs_bytenr && fs_bytenr < csum_bytenr)) {
414 error("bad tree bytenr order: "
415 "root < extent %llu < %llu, "
416 "extent < dev %llu < %llu, "
417 "dev < fs %llu < %llu, "
418 "fs < csum %llu < %llu",
419 (unsigned long long)root_bytenr,
420 (unsigned long long)extent_bytenr,
421 (unsigned long long)extent_bytenr,
422 (unsigned long long)dev_bytenr,
423 (unsigned long long)dev_bytenr,
424 (unsigned long long)fs_bytenr,
425 (unsigned long long)fs_bytenr,
426 (unsigned long long)csum_bytenr);
429 buf = malloc(sizeof(*buf) + cfg->nodesize);
433 ret = setup_temp_extent_buffer(buf, cfg, root_bytenr,
434 BTRFS_ROOT_TREE_OBJECTID);
438 insert_temp_root_item(buf, cfg, &slot, &itemoff,
439 BTRFS_EXTENT_TREE_OBJECTID, extent_bytenr);
440 insert_temp_root_item(buf, cfg, &slot, &itemoff,
441 BTRFS_DEV_TREE_OBJECTID, dev_bytenr);
442 insert_temp_root_item(buf, cfg, &slot, &itemoff,
443 BTRFS_FS_TREE_OBJECTID, fs_bytenr);
444 insert_temp_root_item(buf, cfg, &slot, &itemoff,
445 BTRFS_CSUM_TREE_OBJECTID, csum_bytenr);
447 ret = write_temp_extent_buffer(fd, buf, root_bytenr);
453 static int insert_temp_dev_item(int fd, struct extent_buffer *buf,
454 struct btrfs_mkfs_config *cfg,
455 int *slot, u32 *itemoff)
457 struct btrfs_disk_key disk_key;
458 struct btrfs_dev_item *dev_item;
459 char super_buf[BTRFS_SUPER_INFO_SIZE];
460 unsigned char dev_uuid[BTRFS_UUID_SIZE];
461 unsigned char fsid[BTRFS_FSID_SIZE];
462 struct btrfs_super_block *super = (struct btrfs_super_block *)super_buf;
465 ret = pread(fd, super_buf, BTRFS_SUPER_INFO_SIZE, cfg->super_bytenr);
466 if (ret < BTRFS_SUPER_INFO_SIZE) {
467 ret = (ret < 0 ? -errno : -EIO);
471 btrfs_set_header_nritems(buf, *slot + 1);
472 (*itemoff) -= sizeof(*dev_item);
473 /* setup device item 1, 0 is for replace case */
474 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
475 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
476 btrfs_set_disk_key_offset(&disk_key, 1);
477 btrfs_set_item_key(buf, &disk_key, *slot);
478 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
479 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(*dev_item));
481 dev_item = btrfs_item_ptr(buf, *slot, struct btrfs_dev_item);
482 /* Generate device uuid */
483 uuid_generate(dev_uuid);
484 write_extent_buffer(buf, dev_uuid,
485 (unsigned long)btrfs_device_uuid(dev_item),
487 uuid_parse(cfg->fs_uuid, fsid);
488 write_extent_buffer(buf, fsid,
489 (unsigned long)btrfs_device_fsid(dev_item),
491 btrfs_set_device_id(buf, dev_item, 1);
492 btrfs_set_device_generation(buf, dev_item, 0);
493 btrfs_set_device_total_bytes(buf, dev_item, cfg->num_bytes);
495 * The number must match the initial SYSTEM and META chunk size
497 btrfs_set_device_bytes_used(buf, dev_item,
498 BTRFS_MKFS_SYSTEM_GROUP_SIZE +
499 BTRFS_CONVERT_META_GROUP_SIZE);
500 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
501 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
502 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
503 btrfs_set_device_type(buf, dev_item, 0);
505 /* Super dev_item is not complete, copy the complete one to sb */
506 read_extent_buffer(buf, &super->dev_item, (unsigned long)dev_item,
508 ret = write_temp_super(fd, super, cfg->super_bytenr);
514 static int insert_temp_chunk_item(int fd, struct extent_buffer *buf,
515 struct btrfs_mkfs_config *cfg,
516 int *slot, u32 *itemoff, u64 start, u64 len,
519 struct btrfs_chunk *chunk;
520 struct btrfs_disk_key disk_key;
521 char super_buf[BTRFS_SUPER_INFO_SIZE];
522 struct btrfs_super_block *sb = (struct btrfs_super_block *)super_buf;
525 ret = pread(fd, super_buf, BTRFS_SUPER_INFO_SIZE,
527 if (ret < BTRFS_SUPER_INFO_SIZE) {
528 ret = (ret < 0 ? ret : -EIO);
532 btrfs_set_header_nritems(buf, *slot + 1);
533 (*itemoff) -= btrfs_chunk_item_size(1);
534 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
535 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
536 btrfs_set_disk_key_offset(&disk_key, start);
537 btrfs_set_item_key(buf, &disk_key, *slot);
538 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
539 btrfs_set_item_size(buf, btrfs_item_nr(*slot),
540 btrfs_chunk_item_size(1));
542 chunk = btrfs_item_ptr(buf, *slot, struct btrfs_chunk);
543 btrfs_set_chunk_length(buf, chunk, len);
544 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
545 btrfs_set_chunk_stripe_len(buf, chunk, BTRFS_STRIPE_LEN);
546 btrfs_set_chunk_type(buf, chunk, type);
547 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
548 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
549 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
550 btrfs_set_chunk_num_stripes(buf, chunk, 1);
551 /* TODO: Support DUP profile for system chunk */
552 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
553 /* We are doing 1:1 mapping, so start is its dev offset */
554 btrfs_set_stripe_offset_nr(buf, chunk, 0, start);
555 write_extent_buffer(buf, &sb->dev_item.uuid,
556 (unsigned long)btrfs_stripe_dev_uuid_nr(chunk, 0),
561 * If it's system chunk, also copy it to super block.
563 if (type & BTRFS_BLOCK_GROUP_SYSTEM) {
567 cur = (char *)sb->sys_chunk_array
568 + btrfs_super_sys_array_size(sb);
569 memcpy(cur, &disk_key, sizeof(disk_key));
570 cur += sizeof(disk_key);
571 read_extent_buffer(buf, cur, (unsigned long int)chunk,
572 btrfs_chunk_item_size(1));
573 array_size = btrfs_super_sys_array_size(sb);
574 array_size += btrfs_chunk_item_size(1) +
576 btrfs_set_super_sys_array_size(sb, array_size);
578 ret = write_temp_super(fd, sb, cfg->super_bytenr);
583 static int setup_temp_chunk_tree(int fd, struct btrfs_mkfs_config *cfg,
584 u64 sys_chunk_start, u64 meta_chunk_start,
587 struct extent_buffer *buf = NULL;
588 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
592 /* Must ensure SYS chunk starts before META chunk */
593 if (meta_chunk_start < sys_chunk_start) {
594 error("wrong chunk order: meta < system %llu < %llu",
595 (unsigned long long)meta_chunk_start,
596 (unsigned long long)sys_chunk_start);
599 buf = malloc(sizeof(*buf) + cfg->nodesize);
602 ret = setup_temp_extent_buffer(buf, cfg, chunk_bytenr,
603 BTRFS_CHUNK_TREE_OBJECTID);
607 ret = insert_temp_dev_item(fd, buf, cfg, &slot, &itemoff);
610 ret = insert_temp_chunk_item(fd, buf, cfg, &slot, &itemoff,
612 BTRFS_MKFS_SYSTEM_GROUP_SIZE,
613 BTRFS_BLOCK_GROUP_SYSTEM);
616 ret = insert_temp_chunk_item(fd, buf, cfg, &slot, &itemoff,
618 BTRFS_CONVERT_META_GROUP_SIZE,
619 BTRFS_BLOCK_GROUP_METADATA);
622 ret = write_temp_extent_buffer(fd, buf, chunk_bytenr);
629 static void insert_temp_dev_extent(struct extent_buffer *buf,
630 int *slot, u32 *itemoff, u64 start, u64 len)
632 struct btrfs_dev_extent *dev_extent;
633 struct btrfs_disk_key disk_key;
635 btrfs_set_header_nritems(buf, *slot + 1);
636 (*itemoff) -= sizeof(*dev_extent);
637 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
638 btrfs_set_disk_key_objectid(&disk_key, 1);
639 btrfs_set_disk_key_offset(&disk_key, start);
640 btrfs_set_item_key(buf, &disk_key, *slot);
641 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
642 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(*dev_extent));
644 dev_extent = btrfs_item_ptr(buf, *slot, struct btrfs_dev_extent);
645 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
646 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
647 btrfs_set_dev_extent_length(buf, dev_extent, len);
648 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, start);
649 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
650 BTRFS_CHUNK_TREE_OBJECTID);
654 static int setup_temp_dev_tree(int fd, struct btrfs_mkfs_config *cfg,
655 u64 sys_chunk_start, u64 meta_chunk_start,
658 struct extent_buffer *buf = NULL;
659 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
663 /* Must ensure SYS chunk starts before META chunk */
664 if (meta_chunk_start < sys_chunk_start) {
665 error("wrong chunk order: meta < system %llu < %llu",
666 (unsigned long long)meta_chunk_start,
667 (unsigned long long)sys_chunk_start);
670 buf = malloc(sizeof(*buf) + cfg->nodesize);
673 ret = setup_temp_extent_buffer(buf, cfg, dev_bytenr,
674 BTRFS_DEV_TREE_OBJECTID);
677 insert_temp_dev_extent(buf, &slot, &itemoff, sys_chunk_start,
678 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
679 insert_temp_dev_extent(buf, &slot, &itemoff, meta_chunk_start,
680 BTRFS_CONVERT_META_GROUP_SIZE);
681 ret = write_temp_extent_buffer(fd, buf, dev_bytenr);
687 static int setup_temp_fs_tree(int fd, struct btrfs_mkfs_config *cfg,
690 struct extent_buffer *buf = NULL;
693 buf = malloc(sizeof(*buf) + cfg->nodesize);
696 ret = setup_temp_extent_buffer(buf, cfg, fs_bytenr,
697 BTRFS_FS_TREE_OBJECTID);
701 * Temporary fs tree is completely empty.
703 ret = write_temp_extent_buffer(fd, buf, fs_bytenr);
709 static int setup_temp_csum_tree(int fd, struct btrfs_mkfs_config *cfg,
712 struct extent_buffer *buf = NULL;
715 buf = malloc(sizeof(*buf) + cfg->nodesize);
718 ret = setup_temp_extent_buffer(buf, cfg, csum_bytenr,
719 BTRFS_CSUM_TREE_OBJECTID);
723 * Temporary csum tree is completely empty.
725 ret = write_temp_extent_buffer(fd, buf, csum_bytenr);
732 * Insert one temporary extent item.
734 * NOTE: if skinny_metadata is not enabled, this function must be called
735 * after all other trees are initialized.
736 * Or fs without skinny-metadata will be screwed up.
738 static int insert_temp_extent_item(int fd, struct extent_buffer *buf,
739 struct btrfs_mkfs_config *cfg,
740 int *slot, u32 *itemoff, u64 bytenr,
743 struct extent_buffer *tmp;
744 struct btrfs_extent_item *ei;
745 struct btrfs_extent_inline_ref *iref;
746 struct btrfs_disk_key disk_key;
747 struct btrfs_disk_key tree_info_key;
748 struct btrfs_tree_block_info *info;
750 int skinny_metadata = cfg->features &
751 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA;
755 itemsize = sizeof(*ei) + sizeof(*iref);
757 itemsize = sizeof(*ei) + sizeof(*iref) +
758 sizeof(struct btrfs_tree_block_info);
760 btrfs_set_header_nritems(buf, *slot + 1);
761 *(itemoff) -= itemsize;
763 if (skinny_metadata) {
764 btrfs_set_disk_key_type(&disk_key, BTRFS_METADATA_ITEM_KEY);
765 btrfs_set_disk_key_offset(&disk_key, 0);
767 btrfs_set_disk_key_type(&disk_key, BTRFS_EXTENT_ITEM_KEY);
768 btrfs_set_disk_key_offset(&disk_key, cfg->nodesize);
770 btrfs_set_disk_key_objectid(&disk_key, bytenr);
772 btrfs_set_item_key(buf, &disk_key, *slot);
773 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
774 btrfs_set_item_size(buf, btrfs_item_nr(*slot), itemsize);
776 ei = btrfs_item_ptr(buf, *slot, struct btrfs_extent_item);
777 btrfs_set_extent_refs(buf, ei, 1);
778 btrfs_set_extent_generation(buf, ei, 1);
779 btrfs_set_extent_flags(buf, ei, BTRFS_EXTENT_FLAG_TREE_BLOCK);
781 if (skinny_metadata) {
782 iref = (struct btrfs_extent_inline_ref *)(ei + 1);
784 info = (struct btrfs_tree_block_info *)(ei + 1);
785 iref = (struct btrfs_extent_inline_ref *)(info + 1);
787 btrfs_set_extent_inline_ref_type(buf, iref,
788 BTRFS_TREE_BLOCK_REF_KEY);
789 btrfs_set_extent_inline_ref_offset(buf, iref, ref_root);
796 * Lastly, check the tree block key by read the tree block
797 * Since we do 1:1 mapping for convert case, we can directly
798 * read the bytenr from disk
800 tmp = malloc(sizeof(*tmp) + cfg->nodesize);
803 ret = setup_temp_extent_buffer(tmp, cfg, bytenr, ref_root);
806 ret = pread(fd, tmp->data, cfg->nodesize, bytenr);
807 if (ret < cfg->nodesize) {
808 ret = (ret < 0 ? -errno : -EIO);
811 if (btrfs_header_nritems(tmp) == 0) {
812 btrfs_set_disk_key_type(&tree_info_key, 0);
813 btrfs_set_disk_key_objectid(&tree_info_key, 0);
814 btrfs_set_disk_key_offset(&tree_info_key, 0);
816 btrfs_item_key(tmp, &tree_info_key, 0);
818 btrfs_set_tree_block_key(buf, info, &tree_info_key);
825 static void insert_temp_block_group(struct extent_buffer *buf,
826 struct btrfs_mkfs_config *cfg,
827 int *slot, u32 *itemoff,
828 u64 bytenr, u64 len, u64 used, u64 flag)
830 struct btrfs_block_group_item bgi;
831 struct btrfs_disk_key disk_key;
833 btrfs_set_header_nritems(buf, *slot + 1);
834 (*itemoff) -= sizeof(bgi);
835 btrfs_set_disk_key_type(&disk_key, BTRFS_BLOCK_GROUP_ITEM_KEY);
836 btrfs_set_disk_key_objectid(&disk_key, bytenr);
837 btrfs_set_disk_key_offset(&disk_key, len);
838 btrfs_set_item_key(buf, &disk_key, *slot);
839 btrfs_set_item_offset(buf, btrfs_item_nr(*slot), *itemoff);
840 btrfs_set_item_size(buf, btrfs_item_nr(*slot), sizeof(bgi));
842 btrfs_set_block_group_flags(&bgi, flag);
843 btrfs_set_block_group_used(&bgi, used);
844 btrfs_set_block_group_chunk_objectid(&bgi,
845 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
846 write_extent_buffer(buf, &bgi, btrfs_item_ptr_offset(buf, *slot),
851 static int setup_temp_extent_tree(int fd, struct btrfs_mkfs_config *cfg,
852 u64 chunk_bytenr, u64 root_bytenr,
853 u64 extent_bytenr, u64 dev_bytenr,
854 u64 fs_bytenr, u64 csum_bytenr)
856 struct extent_buffer *buf = NULL;
857 u32 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
862 * We must ensure provided bytenr are in ascending order,
863 * or extent tree key order will be broken.
865 if (!(chunk_bytenr < root_bytenr && root_bytenr < extent_bytenr &&
866 extent_bytenr < dev_bytenr && dev_bytenr < fs_bytenr &&
867 fs_bytenr < csum_bytenr)) {
868 error("bad tree bytenr order: "
869 "chunk < root %llu < %llu, "
870 "root < extent %llu < %llu, "
871 "extent < dev %llu < %llu, "
872 "dev < fs %llu < %llu, "
873 "fs < csum %llu < %llu",
874 (unsigned long long)chunk_bytenr,
875 (unsigned long long)root_bytenr,
876 (unsigned long long)root_bytenr,
877 (unsigned long long)extent_bytenr,
878 (unsigned long long)extent_bytenr,
879 (unsigned long long)dev_bytenr,
880 (unsigned long long)dev_bytenr,
881 (unsigned long long)fs_bytenr,
882 (unsigned long long)fs_bytenr,
883 (unsigned long long)csum_bytenr);
886 buf = malloc(sizeof(*buf) + cfg->nodesize);
890 ret = setup_temp_extent_buffer(buf, cfg, extent_bytenr,
891 BTRFS_EXTENT_TREE_OBJECTID);
895 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
896 chunk_bytenr, BTRFS_CHUNK_TREE_OBJECTID);
900 insert_temp_block_group(buf, cfg, &slot, &itemoff, chunk_bytenr,
901 BTRFS_MKFS_SYSTEM_GROUP_SIZE, cfg->nodesize,
902 BTRFS_BLOCK_GROUP_SYSTEM);
904 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
905 root_bytenr, BTRFS_ROOT_TREE_OBJECTID);
909 /* 5 tree block used, root, extent, dev, fs and csum*/
910 insert_temp_block_group(buf, cfg, &slot, &itemoff, root_bytenr,
911 BTRFS_CONVERT_META_GROUP_SIZE, cfg->nodesize * 5,
912 BTRFS_BLOCK_GROUP_METADATA);
914 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
915 extent_bytenr, BTRFS_EXTENT_TREE_OBJECTID);
918 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
919 dev_bytenr, BTRFS_DEV_TREE_OBJECTID);
922 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
923 fs_bytenr, BTRFS_FS_TREE_OBJECTID);
926 ret = insert_temp_extent_item(fd, buf, cfg, &slot, &itemoff,
927 csum_bytenr, BTRFS_CSUM_TREE_OBJECTID);
931 ret = write_temp_extent_buffer(fd, buf, extent_bytenr);
938 * Improved version of make_btrfs().
941 * 1) Do chunk allocation to avoid used data
942 * And after this function, extent type matches chunk type
943 * 2) Better structured code
944 * No super long hand written codes to initialized all tree blocks
945 * Split into small blocks and reuse codes.
946 * TODO: Reuse tree operation facilities by introducing new flags
948 int make_convert_btrfs(int fd, struct btrfs_mkfs_config *cfg,
949 struct btrfs_convert_context *cctx)
951 struct cache_tree *free = &cctx->free;
952 struct cache_tree *used = &cctx->used;
954 u64 meta_chunk_start;
955 /* chunk tree bytenr, in system chunk */
957 /* metadata trees bytenr, in metadata chunk */
965 /* Shouldn't happen */
966 BUG_ON(cache_tree_empty(used));
969 * reserve space for temporary superblock first
970 * Here we allocate a little larger space, to keep later
971 * free space will be STRIPE_LEN aligned
973 ret = reserve_free_space(free, BTRFS_STRIPE_LEN,
979 * Then reserve system chunk space
980 * TODO: Change system group size depending on cctx->total_bytes.
981 * If using current 4M, it can only handle less than one TB for
982 * worst case and then run out of sys space.
984 ret = reserve_free_space(free, BTRFS_MKFS_SYSTEM_GROUP_SIZE,
988 ret = reserve_free_space(free, BTRFS_CONVERT_META_GROUP_SIZE,
994 * Allocated meta/sys chunks will be mapped 1:1 with device offset.
996 * Inside the allocated metadata chunk, the layout will be:
997 * | offset | contents |
998 * -------------------------------------
1000 * | +nodesize | extent root |
1001 * | +nodesize * 2 | device root |
1002 * | +nodesize * 3 | fs tree |
1003 * | +nodesize * 4 | csum tree |
1004 * -------------------------------------
1005 * Inside the allocated system chunk, the layout will be:
1006 * | offset | contents |
1007 * -------------------------------------
1008 * | +0 | chunk root |
1009 * -------------------------------------
1011 chunk_bytenr = sys_chunk_start;
1012 root_bytenr = meta_chunk_start;
1013 extent_bytenr = meta_chunk_start + cfg->nodesize;
1014 dev_bytenr = meta_chunk_start + cfg->nodesize * 2;
1015 fs_bytenr = meta_chunk_start + cfg->nodesize * 3;
1016 csum_bytenr = meta_chunk_start + cfg->nodesize * 4;
1018 ret = setup_temp_super(fd, cfg, root_bytenr, chunk_bytenr);
1022 ret = setup_temp_root_tree(fd, cfg, root_bytenr, extent_bytenr,
1023 dev_bytenr, fs_bytenr, csum_bytenr);
1026 ret = setup_temp_chunk_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
1030 ret = setup_temp_dev_tree(fd, cfg, sys_chunk_start, meta_chunk_start,
1034 ret = setup_temp_fs_tree(fd, cfg, fs_bytenr);
1037 ret = setup_temp_csum_tree(fd, cfg, csum_bytenr);
1041 * Setup extent tree last, since it may need to read tree block key
1042 * for non-skinny metadata case.
1044 ret = setup_temp_extent_tree(fd, cfg, chunk_bytenr, root_bytenr,
1045 extent_bytenr, dev_bytenr, fs_bytenr,
1052 * @fs_uuid - if NULL, generates a UUID, returns back the new filesystem UUID
1054 * The superblock signature is not valid, denotes a partially created
1055 * filesystem, needs to be finalized.
1057 int make_btrfs(int fd, struct btrfs_mkfs_config *cfg)
1059 struct btrfs_super_block super;
1060 struct extent_buffer *buf;
1061 struct btrfs_root_item root_item;
1062 struct btrfs_disk_key disk_key;
1063 struct btrfs_extent_item *extent_item;
1064 struct btrfs_inode_item *inode_item;
1065 struct btrfs_chunk *chunk;
1066 struct btrfs_dev_item *dev_item;
1067 struct btrfs_dev_extent *dev_extent;
1068 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
1078 int skinny_metadata = !!(cfg->features &
1079 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA);
1082 buf = malloc(sizeof(*buf) + max(cfg->sectorsize, cfg->nodesize));
1086 first_free = BTRFS_SUPER_INFO_OFFSET + cfg->sectorsize * 2 - 1;
1087 first_free &= ~((u64)cfg->sectorsize - 1);
1089 memset(&super, 0, sizeof(super));
1091 num_bytes = (cfg->num_bytes / cfg->sectorsize) * cfg->sectorsize;
1092 if (*cfg->fs_uuid) {
1093 if (uuid_parse(cfg->fs_uuid, super.fsid) != 0) {
1094 error("cannot not parse UUID: %s", cfg->fs_uuid);
1098 if (!test_uuid_unique(cfg->fs_uuid)) {
1099 error("non-unique UUID: %s", cfg->fs_uuid);
1104 uuid_generate(super.fsid);
1105 uuid_unparse(super.fsid, cfg->fs_uuid);
1107 uuid_generate(super.dev_item.uuid);
1108 uuid_generate(chunk_tree_uuid);
1110 btrfs_set_super_bytenr(&super, cfg->blocks[0]);
1111 btrfs_set_super_num_devices(&super, 1);
1112 btrfs_set_super_magic(&super, BTRFS_MAGIC_PARTIAL);
1113 btrfs_set_super_generation(&super, 1);
1114 btrfs_set_super_root(&super, cfg->blocks[1]);
1115 btrfs_set_super_chunk_root(&super, cfg->blocks[3]);
1116 btrfs_set_super_total_bytes(&super, num_bytes);
1117 btrfs_set_super_bytes_used(&super, 6 * cfg->nodesize);
1118 btrfs_set_super_sectorsize(&super, cfg->sectorsize);
1119 btrfs_set_super_leafsize(&super, cfg->nodesize);
1120 btrfs_set_super_nodesize(&super, cfg->nodesize);
1121 btrfs_set_super_stripesize(&super, cfg->stripesize);
1122 btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
1123 btrfs_set_super_chunk_root_generation(&super, 1);
1124 btrfs_set_super_cache_generation(&super, -1);
1125 btrfs_set_super_incompat_flags(&super, cfg->features);
1127 __strncpy_null(super.label, cfg->label, BTRFS_LABEL_SIZE - 1);
1129 /* create the tree of root objects */
1130 memset(buf->data, 0, cfg->nodesize);
1131 buf->len = cfg->nodesize;
1132 btrfs_set_header_bytenr(buf, cfg->blocks[1]);
1133 btrfs_set_header_nritems(buf, 4);
1134 btrfs_set_header_generation(buf, 1);
1135 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
1136 btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
1137 write_extent_buffer(buf, super.fsid, btrfs_header_fsid(),
1140 write_extent_buffer(buf, chunk_tree_uuid,
1141 btrfs_header_chunk_tree_uuid(buf),
1144 /* create the items for the root tree */
1145 memset(&root_item, 0, sizeof(root_item));
1146 inode_item = &root_item.inode;
1147 btrfs_set_stack_inode_generation(inode_item, 1);
1148 btrfs_set_stack_inode_size(inode_item, 3);
1149 btrfs_set_stack_inode_nlink(inode_item, 1);
1150 btrfs_set_stack_inode_nbytes(inode_item, cfg->nodesize);
1151 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
1152 btrfs_set_root_refs(&root_item, 1);
1153 btrfs_set_root_used(&root_item, cfg->nodesize);
1154 btrfs_set_root_generation(&root_item, 1);
1156 memset(&disk_key, 0, sizeof(disk_key));
1157 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
1158 btrfs_set_disk_key_offset(&disk_key, 0);
1161 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - sizeof(root_item);
1162 btrfs_set_root_bytenr(&root_item, cfg->blocks[2]);
1163 btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
1164 btrfs_set_item_key(buf, &disk_key, nritems);
1165 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1166 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1168 write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
1169 nritems), sizeof(root_item));
1172 itemoff = itemoff - sizeof(root_item);
1173 btrfs_set_root_bytenr(&root_item, cfg->blocks[4]);
1174 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
1175 btrfs_set_item_key(buf, &disk_key, nritems);
1176 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1177 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1179 write_extent_buffer(buf, &root_item,
1180 btrfs_item_ptr_offset(buf, nritems),
1184 itemoff = itemoff - sizeof(root_item);
1185 btrfs_set_root_bytenr(&root_item, cfg->blocks[5]);
1186 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
1187 btrfs_set_item_key(buf, &disk_key, nritems);
1188 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1189 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1191 write_extent_buffer(buf, &root_item,
1192 btrfs_item_ptr_offset(buf, nritems),
1196 itemoff = itemoff - sizeof(root_item);
1197 btrfs_set_root_bytenr(&root_item, cfg->blocks[6]);
1198 btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_TREE_OBJECTID);
1199 btrfs_set_item_key(buf, &disk_key, nritems);
1200 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1201 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1203 write_extent_buffer(buf, &root_item,
1204 btrfs_item_ptr_offset(buf, nritems),
1209 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1210 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[1]);
1211 if (ret != cfg->nodesize) {
1212 ret = (ret < 0 ? -errno : -EIO);
1216 /* create the items for the extent tree */
1217 memset(buf->data + sizeof(struct btrfs_header), 0,
1218 cfg->nodesize - sizeof(struct btrfs_header));
1220 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize);
1221 for (i = 1; i < 7; i++) {
1222 item_size = sizeof(struct btrfs_extent_item);
1223 if (!skinny_metadata)
1224 item_size += sizeof(struct btrfs_tree_block_info);
1226 if (cfg->blocks[i] < first_free) {
1227 error("block[%d] below first free: %llu < %llu",
1228 i, (unsigned long long)cfg->blocks[i],
1229 (unsigned long long)first_free);
1233 if (cfg->blocks[i] < cfg->blocks[i - 1]) {
1234 error("blocks %d and %d in reverse order: %llu < %llu",
1236 (unsigned long long)cfg->blocks[i],
1237 (unsigned long long)cfg->blocks[i - 1]);
1242 /* create extent item */
1243 itemoff -= item_size;
1244 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
1245 if (skinny_metadata) {
1246 btrfs_set_disk_key_type(&disk_key,
1247 BTRFS_METADATA_ITEM_KEY);
1248 btrfs_set_disk_key_offset(&disk_key, 0);
1250 btrfs_set_disk_key_type(&disk_key,
1251 BTRFS_EXTENT_ITEM_KEY);
1252 btrfs_set_disk_key_offset(&disk_key, cfg->nodesize);
1254 btrfs_set_item_key(buf, &disk_key, nritems);
1255 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
1257 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1259 extent_item = btrfs_item_ptr(buf, nritems,
1260 struct btrfs_extent_item);
1261 btrfs_set_extent_refs(buf, extent_item, 1);
1262 btrfs_set_extent_generation(buf, extent_item, 1);
1263 btrfs_set_extent_flags(buf, extent_item,
1264 BTRFS_EXTENT_FLAG_TREE_BLOCK);
1267 /* create extent ref */
1268 ref_root = reference_root_table[i];
1269 btrfs_set_disk_key_objectid(&disk_key, cfg->blocks[i]);
1270 btrfs_set_disk_key_offset(&disk_key, ref_root);
1271 btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
1272 btrfs_set_item_key(buf, &disk_key, nritems);
1273 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
1275 btrfs_set_item_size(buf, btrfs_item_nr(nritems), 0);
1278 btrfs_set_header_bytenr(buf, cfg->blocks[2]);
1279 btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
1280 btrfs_set_header_nritems(buf, nritems);
1281 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1282 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[2]);
1283 if (ret != cfg->nodesize) {
1284 ret = (ret < 0 ? -errno : -EIO);
1288 /* create the chunk tree */
1289 memset(buf->data + sizeof(struct btrfs_header), 0,
1290 cfg->nodesize - sizeof(struct btrfs_header));
1292 item_size = sizeof(*dev_item);
1293 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) - item_size;
1295 /* first device 1 (there is no device 0) */
1296 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
1297 btrfs_set_disk_key_offset(&disk_key, 1);
1298 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
1299 btrfs_set_item_key(buf, &disk_key, nritems);
1300 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1301 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
1303 dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
1304 btrfs_set_device_id(buf, dev_item, 1);
1305 btrfs_set_device_generation(buf, dev_item, 0);
1306 btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
1307 btrfs_set_device_bytes_used(buf, dev_item,
1308 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1309 btrfs_set_device_io_align(buf, dev_item, cfg->sectorsize);
1310 btrfs_set_device_io_width(buf, dev_item, cfg->sectorsize);
1311 btrfs_set_device_sector_size(buf, dev_item, cfg->sectorsize);
1312 btrfs_set_device_type(buf, dev_item, 0);
1314 write_extent_buffer(buf, super.dev_item.uuid,
1315 (unsigned long)btrfs_device_uuid(dev_item),
1317 write_extent_buffer(buf, super.fsid,
1318 (unsigned long)btrfs_device_fsid(dev_item),
1320 read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
1324 item_size = btrfs_chunk_item_size(1);
1325 itemoff = itemoff - item_size;
1327 /* then we have chunk 0 */
1328 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1329 btrfs_set_disk_key_offset(&disk_key, 0);
1330 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
1331 btrfs_set_item_key(buf, &disk_key, nritems);
1332 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1333 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
1335 chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
1336 btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1337 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
1338 btrfs_set_chunk_stripe_len(buf, chunk, BTRFS_STRIPE_LEN);
1339 btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
1340 btrfs_set_chunk_io_align(buf, chunk, cfg->sectorsize);
1341 btrfs_set_chunk_io_width(buf, chunk, cfg->sectorsize);
1342 btrfs_set_chunk_sector_size(buf, chunk, cfg->sectorsize);
1343 btrfs_set_chunk_num_stripes(buf, chunk, 1);
1344 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
1345 btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
1348 write_extent_buffer(buf, super.dev_item.uuid,
1349 (unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
1352 /* copy the key for the chunk to the system array */
1353 ptr = super.sys_chunk_array;
1354 array_size = sizeof(disk_key);
1356 memcpy(ptr, &disk_key, sizeof(disk_key));
1357 ptr += sizeof(disk_key);
1359 /* copy the chunk to the system array */
1360 read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
1361 array_size += item_size;
1363 btrfs_set_super_sys_array_size(&super, array_size);
1365 btrfs_set_header_bytenr(buf, cfg->blocks[3]);
1366 btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
1367 btrfs_set_header_nritems(buf, nritems);
1368 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1369 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[3]);
1370 if (ret != cfg->nodesize) {
1371 ret = (ret < 0 ? -errno : -EIO);
1375 /* create the device tree */
1376 memset(buf->data + sizeof(struct btrfs_header), 0,
1377 cfg->nodesize - sizeof(struct btrfs_header));
1379 itemoff = __BTRFS_LEAF_DATA_SIZE(cfg->nodesize) -
1380 sizeof(struct btrfs_dev_extent);
1382 btrfs_set_disk_key_objectid(&disk_key, 1);
1383 btrfs_set_disk_key_offset(&disk_key, 0);
1384 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
1385 btrfs_set_item_key(buf, &disk_key, nritems);
1386 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
1387 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
1388 sizeof(struct btrfs_dev_extent));
1389 dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
1390 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
1391 BTRFS_CHUNK_TREE_OBJECTID);
1392 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
1393 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
1394 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
1396 write_extent_buffer(buf, chunk_tree_uuid,
1397 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
1400 btrfs_set_dev_extent_length(buf, dev_extent,
1401 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
1404 btrfs_set_header_bytenr(buf, cfg->blocks[4]);
1405 btrfs_set_header_owner(buf, BTRFS_DEV_TREE_OBJECTID);
1406 btrfs_set_header_nritems(buf, nritems);
1407 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1408 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[4]);
1409 if (ret != cfg->nodesize) {
1410 ret = (ret < 0 ? -errno : -EIO);
1414 /* create the FS root */
1415 memset(buf->data + sizeof(struct btrfs_header), 0,
1416 cfg->nodesize - sizeof(struct btrfs_header));
1417 btrfs_set_header_bytenr(buf, cfg->blocks[5]);
1418 btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
1419 btrfs_set_header_nritems(buf, 0);
1420 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1421 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[5]);
1422 if (ret != cfg->nodesize) {
1423 ret = (ret < 0 ? -errno : -EIO);
1426 /* finally create the csum root */
1427 memset(buf->data + sizeof(struct btrfs_header), 0,
1428 cfg->nodesize - sizeof(struct btrfs_header));
1429 btrfs_set_header_bytenr(buf, cfg->blocks[6]);
1430 btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
1431 btrfs_set_header_nritems(buf, 0);
1432 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1433 ret = pwrite(fd, buf->data, cfg->nodesize, cfg->blocks[6]);
1434 if (ret != cfg->nodesize) {
1435 ret = (ret < 0 ? -errno : -EIO);
1439 /* and write out the super block */
1440 memset(buf->data, 0, BTRFS_SUPER_INFO_SIZE);
1441 memcpy(buf->data, &super, sizeof(super));
1442 buf->len = BTRFS_SUPER_INFO_SIZE;
1443 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
1444 ret = pwrite(fd, buf->data, BTRFS_SUPER_INFO_SIZE, cfg->blocks[0]);
1445 if (ret != BTRFS_SUPER_INFO_SIZE) {
1446 ret = (ret < 0 ? -errno : -EIO);
1457 #define VERSION_TO_STRING3(a,b,c) #a "." #b "." #c, KERNEL_VERSION(a,b,c)
1458 #define VERSION_TO_STRING2(a,b) #a "." #b, KERNEL_VERSION(a,b,0)
1461 * Feature stability status and versions: compat <= safe <= default
1463 static const struct btrfs_fs_feature {
1466 const char *sysfs_name;
1468 * Compatibility with kernel of given version. Filesystem can be
1471 const char *compat_str;
1474 * Considered safe for use, but is not on by default, even if the
1475 * kernel supports the feature.
1477 const char *safe_str;
1480 * Considered safe for use and will be turned on by default if
1481 * supported by the running kernel.
1483 const char *default_str;
1486 } mkfs_features[] = {
1487 { "mixed-bg", BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS,
1489 VERSION_TO_STRING3(2,6,37),
1490 VERSION_TO_STRING3(2,6,37),
1492 "mixed data and metadata block groups" },
1493 { "extref", BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF,
1495 VERSION_TO_STRING2(3,7),
1496 VERSION_TO_STRING2(3,12),
1497 VERSION_TO_STRING2(3,12),
1498 "increased hardlink limit per file to 65536" },
1499 { "raid56", BTRFS_FEATURE_INCOMPAT_RAID56,
1501 VERSION_TO_STRING2(3,9),
1504 "raid56 extended format" },
1505 { "skinny-metadata", BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA,
1507 VERSION_TO_STRING2(3,10),
1508 VERSION_TO_STRING2(3,18),
1509 VERSION_TO_STRING2(3,18),
1510 "reduced-size metadata extent refs" },
1511 { "no-holes", BTRFS_FEATURE_INCOMPAT_NO_HOLES,
1513 VERSION_TO_STRING2(3,14),
1514 VERSION_TO_STRING2(4,0),
1516 "no explicit hole extents for files" },
1517 /* Keep this one last */
1518 { "list-all", BTRFS_FEATURE_LIST_ALL, NULL }
1521 static int parse_one_fs_feature(const char *name, u64 *flags)
1526 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1527 if (name[0] == '^' &&
1528 !strcmp(mkfs_features[i].name, name + 1)) {
1529 *flags &= ~ mkfs_features[i].flag;
1531 } else if (!strcmp(mkfs_features[i].name, name)) {
1532 *flags |= mkfs_features[i].flag;
1540 void btrfs_parse_features_to_string(char *buf, u64 flags)
1546 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1547 if (flags & mkfs_features[i].flag) {
1550 strcat(buf, mkfs_features[i].name);
1555 void btrfs_process_fs_features(u64 flags)
1559 for (i = 0; i < ARRAY_SIZE(mkfs_features); i++) {
1560 if (flags & mkfs_features[i].flag) {
1561 printf("Turning ON incompat feature '%s': %s\n",
1562 mkfs_features[i].name,
1563 mkfs_features[i].desc);
1568 void btrfs_list_all_fs_features(u64 mask_disallowed)
1572 fprintf(stderr, "Filesystem features available:\n");
1573 for (i = 0; i < ARRAY_SIZE(mkfs_features) - 1; i++) {
1574 const struct btrfs_fs_feature *feat = &mkfs_features[i];
1576 if (feat->flag & mask_disallowed)
1578 fprintf(stderr, "%-20s- %s (0x%llx", feat->name, feat->desc,
1580 if (feat->compat_ver)
1581 fprintf(stderr, ", compat=%s", feat->compat_str);
1583 fprintf(stderr, ", safe=%s", feat->safe_str);
1584 if (feat->default_ver)
1585 fprintf(stderr, ", default=%s", feat->default_str);
1586 fprintf(stderr, ")\n");
1591 * Return NULL if all features were parsed fine, otherwise return the name of
1592 * the first unparsed.
1594 char* btrfs_parse_fs_features(char *namelist, u64 *flags)
1597 char *save_ptr = NULL; /* Satisfy static checkers */
1599 for (this_char = strtok_r(namelist, ",", &save_ptr);
1601 this_char = strtok_r(NULL, ",", &save_ptr)) {
1602 if (parse_one_fs_feature(this_char, flags))
1609 void print_kernel_version(FILE *stream, u32 version)
1613 v[0] = version & 0xFF;
1614 v[1] = (version >> 8) & 0xFF;
1615 v[2] = version >> 16;
1616 fprintf(stream, "%u.%u", v[2], v[1]);
1618 fprintf(stream, ".%u", v[0]);
1621 u32 get_running_kernel_version(void)
1623 struct utsname utsbuf;
1625 char *saveptr = NULL;
1629 if (strcmp(utsbuf.sysname, "Linux") != 0) {
1630 error("unsupported system: %s", utsbuf.sysname);
1634 tmp = strchr(utsbuf.release, '-');
1638 tmp = strtok_r(utsbuf.release, ".", &saveptr);
1639 if (!string_is_numerical(tmp))
1641 version = atoi(tmp) << 16;
1642 tmp = strtok_r(NULL, ".", &saveptr);
1643 if (!string_is_numerical(tmp))
1645 version |= atoi(tmp) << 8;
1646 tmp = strtok_r(NULL, ".", &saveptr);
1648 if (!string_is_numerical(tmp))
1650 version |= atoi(tmp);
1656 u64 btrfs_device_size(int fd, struct stat *st)
1659 if (S_ISREG(st->st_mode)) {
1662 if (!S_ISBLK(st->st_mode)) {
1665 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
1671 static int zero_blocks(int fd, off_t start, size_t len)
1673 char *buf = malloc(len);
1679 memset(buf, 0, len);
1680 written = pwrite(fd, buf, len, start);
1687 #define ZERO_DEV_BYTES SZ_2M
1689 /* don't write outside the device by clamping the region to the device size */
1690 static int zero_dev_clamped(int fd, off_t start, ssize_t len, u64 dev_size)
1692 off_t end = max(start, start + len);
1695 /* and don't overwrite the disk labels on sparc */
1696 start = max(start, 1024);
1697 end = max(end, 1024);
1700 start = min_t(u64, start, dev_size);
1701 end = min_t(u64, end, dev_size);
1703 return zero_blocks(fd, start, end - start);
1706 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
1707 struct btrfs_root *root, int fd, const char *path,
1708 u64 device_total_bytes, u32 io_width, u32 io_align,
1711 struct btrfs_super_block *disk_super;
1712 struct btrfs_super_block *super = root->fs_info->super_copy;
1713 struct btrfs_device *device;
1714 struct btrfs_dev_item *dev_item;
1720 device_total_bytes = (device_total_bytes / sectorsize) * sectorsize;
1722 device = calloc(1, sizeof(*device));
1727 buf = calloc(1, sectorsize);
1733 disk_super = (struct btrfs_super_block *)buf;
1734 dev_item = &disk_super->dev_item;
1736 uuid_generate(device->uuid);
1739 device->io_width = io_width;
1740 device->io_align = io_align;
1741 device->sector_size = sectorsize;
1743 device->writeable = 1;
1744 device->total_bytes = device_total_bytes;
1745 device->bytes_used = 0;
1746 device->total_ios = 0;
1747 device->dev_root = root->fs_info->dev_root;
1748 device->name = strdup(path);
1749 if (!device->name) {
1754 INIT_LIST_HEAD(&device->dev_list);
1755 ret = btrfs_add_device(trans, root, device);
1759 fs_total_bytes = btrfs_super_total_bytes(super) + device_total_bytes;
1760 btrfs_set_super_total_bytes(super, fs_total_bytes);
1762 num_devs = btrfs_super_num_devices(super) + 1;
1763 btrfs_set_super_num_devices(super, num_devs);
1765 memcpy(disk_super, super, sizeof(*disk_super));
1767 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
1768 btrfs_set_stack_device_id(dev_item, device->devid);
1769 btrfs_set_stack_device_type(dev_item, device->type);
1770 btrfs_set_stack_device_io_align(dev_item, device->io_align);
1771 btrfs_set_stack_device_io_width(dev_item, device->io_width);
1772 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
1773 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
1774 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
1775 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
1777 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
1778 BUG_ON(ret != sectorsize);
1781 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
1782 device->fs_devices = root->fs_info->fs_devices;
1791 static int btrfs_wipe_existing_sb(int fd)
1793 const char *off = NULL;
1798 blkid_probe pr = NULL;
1800 pr = blkid_new_probe();
1804 if (blkid_probe_set_device(pr, fd, 0, 0)) {
1809 ret = blkid_probe_lookup_value(pr, "SBMAGIC_OFFSET", &off, NULL);
1811 ret = blkid_probe_lookup_value(pr, "SBMAGIC", NULL, &len);
1813 if (ret || len == 0 || off == NULL) {
1815 * If lookup fails, the probe did not find any values, eg. for
1816 * a file image or a loop device. Soft error.
1822 offset = strtoll(off, NULL, 10);
1823 if (len > sizeof(buf))
1826 memset(buf, 0, len);
1827 ret = pwrite(fd, buf, len, offset);
1829 error("cannot wipe existing superblock: %s", strerror(errno));
1831 } else if (ret != len) {
1832 error("cannot wipe existing superblock: wrote %d of %zd", ret, len);
1838 blkid_free_probe(pr);
1842 int btrfs_prepare_device(int fd, const char *file, u64 *block_count_ret,
1843 u64 max_block_count, unsigned opflags)
1849 ret = fstat(fd, &st);
1851 error("unable to stat %s: %s", file, strerror(errno));
1855 block_count = btrfs_device_size(fd, &st);
1856 if (block_count == 0) {
1857 error("unable to determine size of %s", file);
1860 if (max_block_count)
1861 block_count = min(block_count, max_block_count);
1863 if (opflags & PREP_DEVICE_DISCARD) {
1865 * We intentionally ignore errors from the discard ioctl. It
1866 * is not necessary for the mkfs functionality but just an
1869 if (discard_range(fd, 0, 0) == 0) {
1870 if (opflags & PREP_DEVICE_VERBOSE)
1871 printf("Performing full device TRIM %s (%s) ...\n",
1872 file, pretty_size(block_count));
1873 discard_blocks(fd, 0, block_count);
1877 ret = zero_dev_clamped(fd, 0, ZERO_DEV_BYTES, block_count);
1878 for (i = 0 ; !ret && i < BTRFS_SUPER_MIRROR_MAX; i++)
1879 ret = zero_dev_clamped(fd, btrfs_sb_offset(i),
1880 BTRFS_SUPER_INFO_SIZE, block_count);
1881 if (!ret && (opflags & PREP_DEVICE_ZERO_END))
1882 ret = zero_dev_clamped(fd, block_count - ZERO_DEV_BYTES,
1883 ZERO_DEV_BYTES, block_count);
1886 error("failed to zero device '%s': %s", file, strerror(-ret));
1890 ret = btrfs_wipe_existing_sb(fd);
1892 error("cannot wipe superblocks on %s", file);
1896 *block_count_ret = block_count;
1900 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
1901 struct btrfs_root *root, u64 objectid)
1904 struct btrfs_inode_item inode_item;
1905 time_t now = time(NULL);
1907 memset(&inode_item, 0, sizeof(inode_item));
1908 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
1909 btrfs_set_stack_inode_size(&inode_item, 0);
1910 btrfs_set_stack_inode_nlink(&inode_item, 1);
1911 btrfs_set_stack_inode_nbytes(&inode_item, root->nodesize);
1912 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
1913 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
1914 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
1915 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
1916 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
1917 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
1918 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
1919 btrfs_set_stack_timespec_sec(&inode_item.otime, now);
1920 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
1922 if (root->fs_info->tree_root == root)
1923 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
1925 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
1929 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
1933 btrfs_set_root_dirid(&root->root_item, objectid);
1940 * checks if a path is a block device node
1941 * Returns negative errno on failure, otherwise
1942 * returns 1 for blockdev, 0 for not-blockdev
1944 int is_block_device(const char *path)
1946 struct stat statbuf;
1948 if (stat(path, &statbuf) < 0)
1951 return !!S_ISBLK(statbuf.st_mode);
1955 * check if given path is a mount point
1956 * return 1 if yes. 0 if no. -1 for error
1958 int is_mount_point(const char *path)
1964 f = setmntent("/proc/self/mounts", "r");
1968 while ((mnt = getmntent(f)) != NULL) {
1969 if (strcmp(mnt->mnt_dir, path))
1978 static int is_reg_file(const char *path)
1980 struct stat statbuf;
1982 if (stat(path, &statbuf) < 0)
1984 return S_ISREG(statbuf.st_mode);
1988 * This function checks if the given input parameter is
1990 * return <0 : some error in the given input
1991 * return BTRFS_ARG_UNKNOWN: unknown input
1992 * return BTRFS_ARG_UUID: given input is uuid
1993 * return BTRFS_ARG_MNTPOINT: given input is path
1994 * return BTRFS_ARG_REG: given input is regular file
1995 * return BTRFS_ARG_BLKDEV: given input is block device
1997 int check_arg_type(const char *input)
2000 char path[PATH_MAX];
2005 if (realpath(input, path)) {
2006 if (is_block_device(path) == 1)
2007 return BTRFS_ARG_BLKDEV;
2009 if (is_mount_point(path) == 1)
2010 return BTRFS_ARG_MNTPOINT;
2012 if (is_reg_file(path))
2013 return BTRFS_ARG_REG;
2015 return BTRFS_ARG_UNKNOWN;
2018 if (strlen(input) == (BTRFS_UUID_UNPARSED_SIZE - 1) &&
2019 !uuid_parse(input, uuid))
2020 return BTRFS_ARG_UUID;
2022 return BTRFS_ARG_UNKNOWN;
2026 * Find the mount point for a mounted device.
2027 * On success, returns 0 with mountpoint in *mp.
2028 * On failure, returns -errno (not mounted yields -EINVAL)
2029 * Is noisy on failures, expects to be given a mounted device.
2031 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size)
2036 ret = is_block_device(dev);
2039 error("not a block device: %s", dev);
2042 error("cannot check %s: %s", dev, strerror(-ret));
2047 fd = open(dev, O_RDONLY);
2050 error("cannot open %s: %s", dev, strerror(errno));
2054 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
2057 } else { /* mounted, all good */
2067 * Given a pathname, return a filehandle to:
2068 * the original pathname or,
2069 * if the pathname is a mounted btrfs device, to its mountpoint.
2071 * On error, return -1, errno should be set.
2073 int open_path_or_dev_mnt(const char *path, DIR **dirstream, int verbose)
2078 if (is_block_device(path)) {
2079 ret = get_btrfs_mount(path, mp, sizeof(mp));
2081 /* not a mounted btrfs dev */
2082 error_on(verbose, "'%s' is not a mounted btrfs device",
2087 ret = open_file_or_dir(mp, dirstream);
2088 error_on(verbose && ret < 0, "can't access '%s': %s",
2089 path, strerror(errno));
2091 ret = btrfs_open_dir(path, dirstream, 1);
2098 * Do the following checks before calling open_file_or_dir():
2099 * 1: path is in a btrfs filesystem
2100 * 2: path is a directory
2102 int btrfs_open_dir(const char *path, DIR **dirstream, int verbose)
2108 if (statfs(path, &stfs) != 0) {
2109 error_on(verbose, "cannot access '%s': %s", path,
2114 if (stfs.f_type != BTRFS_SUPER_MAGIC) {
2115 error_on(verbose, "not a btrfs filesystem: %s", path);
2119 if (stat(path, &st) != 0) {
2120 error_on(verbose, "cannot access '%s': %s", path,
2125 if (!S_ISDIR(st.st_mode)) {
2126 error_on(verbose, "not a directory: %s", path);
2130 ret = open_file_or_dir(path, dirstream);
2132 error_on(verbose, "cannot access '%s': %s", path,
2139 /* checks if a device is a loop device */
2140 static int is_loop_device (const char* device) {
2141 struct stat statbuf;
2143 if(stat(device, &statbuf) < 0)
2146 return (S_ISBLK(statbuf.st_mode) &&
2147 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
2151 * Takes a loop device path (e.g. /dev/loop0) and returns
2152 * the associated file (e.g. /images/my_btrfs.img) using
2155 static int resolve_loop_device_with_loopdev(const char* loop_dev, char* loop_file)
2159 struct loop_info64 lo64;
2161 fd = open(loop_dev, O_RDONLY | O_NONBLOCK);
2164 ret = ioctl(fd, LOOP_GET_STATUS64, &lo64);
2170 memcpy(loop_file, lo64.lo_file_name, sizeof(lo64.lo_file_name));
2171 loop_file[sizeof(lo64.lo_file_name)] = 0;
2179 /* Takes a loop device path (e.g. /dev/loop0) and returns
2180 * the associated file (e.g. /images/my_btrfs.img) */
2181 static int resolve_loop_device(const char* loop_dev, char* loop_file,
2188 char real_loop_dev[PATH_MAX];
2190 if (!realpath(loop_dev, real_loop_dev))
2192 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
2193 if (!(f = fopen(p, "r"))) {
2194 if (errno == ENOENT)
2196 * It's possibly a partitioned loop device, which is
2197 * resolvable with loopdev API.
2199 return resolve_loop_device_with_loopdev(loop_dev, loop_file);
2203 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
2204 ret = fscanf(f, fmt, loop_file);
2213 * Checks whether a and b are identical or device
2214 * files associated with the same block device
2216 static int is_same_blk_file(const char* a, const char* b)
2218 struct stat st_buf_a, st_buf_b;
2219 char real_a[PATH_MAX];
2220 char real_b[PATH_MAX];
2222 if (!realpath(a, real_a))
2223 strncpy_null(real_a, a);
2225 if (!realpath(b, real_b))
2226 strncpy_null(real_b, b);
2228 /* Identical path? */
2229 if (strcmp(real_a, real_b) == 0)
2232 if (stat(a, &st_buf_a) < 0 || stat(b, &st_buf_b) < 0) {
2233 if (errno == ENOENT)
2238 /* Same blockdevice? */
2239 if (S_ISBLK(st_buf_a.st_mode) && S_ISBLK(st_buf_b.st_mode) &&
2240 st_buf_a.st_rdev == st_buf_b.st_rdev) {
2245 if (st_buf_a.st_dev == st_buf_b.st_dev &&
2246 st_buf_a.st_ino == st_buf_b.st_ino) {
2253 /* checks if a and b are identical or device
2254 * files associated with the same block device or
2255 * if one file is a loop device that uses the other
2258 static int is_same_loop_file(const char* a, const char* b)
2260 char res_a[PATH_MAX];
2261 char res_b[PATH_MAX];
2262 const char* final_a = NULL;
2263 const char* final_b = NULL;
2266 /* Resolve a if it is a loop device */
2267 if((ret = is_loop_device(a)) < 0) {
2272 ret = resolve_loop_device(a, res_a, sizeof(res_a));
2283 /* Resolve b if it is a loop device */
2284 if ((ret = is_loop_device(b)) < 0) {
2289 ret = resolve_loop_device(b, res_b, sizeof(res_b));
2300 return is_same_blk_file(final_a, final_b);
2303 /* Checks if a file exists and is a block or regular file*/
2304 static int is_existing_blk_or_reg_file(const char* filename)
2308 if(stat(filename, &st_buf) < 0) {
2315 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
2318 /* Checks if a file is used (directly or indirectly via a loop device)
2319 * by a device in fs_devices
2321 static int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices,
2325 struct list_head *head;
2326 struct list_head *cur;
2327 struct btrfs_device *device;
2329 head = &fs_devices->devices;
2330 list_for_each(cur, head) {
2331 device = list_entry(cur, struct btrfs_device, dev_list);
2333 if((ret = is_same_loop_file(device->name, file)))
2341 * Resolve a pathname to a device mapper node to /dev/mapper/<name>
2342 * Returns NULL on invalid input or malloc failure; Other failures
2343 * will be handled by the caller using the input pathame.
2345 char *canonicalize_dm_name(const char *ptname)
2349 char path[PATH_MAX], name[PATH_MAX], *res = NULL;
2351 if (!ptname || !*ptname)
2354 snprintf(path, sizeof(path), "/sys/block/%s/dm/name", ptname);
2355 if (!(f = fopen(path, "r")))
2358 /* read <name>\n from sysfs */
2359 if (fgets(name, sizeof(name), f) && (sz = strlen(name)) > 1) {
2360 name[sz - 1] = '\0';
2361 snprintf(path, sizeof(path), "/dev/mapper/%s", name);
2363 if (access(path, F_OK) == 0)
2371 * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or
2372 * to a device mapper pathname.
2373 * Returns NULL on invalid input or malloc failure; Other failures
2374 * will be handled by the caller using the input pathame.
2376 char *canonicalize_path(const char *path)
2378 char *canonical, *p;
2380 if (!path || !*path)
2383 canonical = realpath(path, NULL);
2385 return strdup(path);
2386 p = strrchr(canonical, '/');
2387 if (p && strncmp(p, "/dm-", 4) == 0 && isdigit(*(p + 4))) {
2388 char *dm = canonicalize_dm_name(p + 1);
2399 * returns 1 if the device was mounted, < 0 on error or 0 if everything
2400 * is safe to continue.
2402 int check_mounted(const char* file)
2407 fd = open(file, O_RDONLY);
2409 error("mount check: cannot open %s: %s", file,
2414 ret = check_mounted_where(fd, file, NULL, 0, NULL);
2420 int check_mounted_where(int fd, const char *file, char *where, int size,
2421 struct btrfs_fs_devices **fs_dev_ret)
2426 struct btrfs_fs_devices *fs_devices_mnt = NULL;
2430 /* scan the initial device */
2431 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
2432 &total_devs, BTRFS_SUPER_INFO_OFFSET, SBREAD_DEFAULT);
2433 is_btrfs = (ret >= 0);
2435 /* scan other devices */
2436 if (is_btrfs && total_devs > 1) {
2437 ret = btrfs_scan_devices();
2442 /* iterate over the list of currently mounted filesystems */
2443 if ((f = setmntent ("/proc/self/mounts", "r")) == NULL)
2446 while ((mnt = getmntent (f)) != NULL) {
2448 if(strcmp(mnt->mnt_type, "btrfs") != 0)
2451 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
2453 /* ignore entries in the mount table that are not
2454 associated with a file*/
2455 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
2456 goto out_mntloop_err;
2460 ret = is_same_loop_file(file, mnt->mnt_fsname);
2464 goto out_mntloop_err;
2469 /* Did we find an entry in mnt table? */
2470 if (mnt && size && where) {
2471 strncpy(where, mnt->mnt_dir, size);
2475 *fs_dev_ret = fs_devices_mnt;
2477 ret = (mnt != NULL);
2485 struct pending_dir {
2486 struct list_head list;
2487 char name[PATH_MAX];
2490 int btrfs_register_one_device(const char *fname)
2492 struct btrfs_ioctl_vol_args args;
2496 fd = open("/dev/btrfs-control", O_RDWR);
2499 "failed to open /dev/btrfs-control, skipping device registration: %s",
2503 memset(&args, 0, sizeof(args));
2504 strncpy_null(args.name, fname);
2505 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
2507 error("device scan failed on '%s': %s", fname,
2516 * Register all devices in the fs_uuid list created in the user
2517 * space. Ensure btrfs_scan_devices() is called before this func.
2519 int btrfs_register_all_devices(void)
2523 struct btrfs_fs_devices *fs_devices;
2524 struct btrfs_device *device;
2525 struct list_head *all_uuids;
2527 all_uuids = btrfs_scanned_uuids();
2529 list_for_each_entry(fs_devices, all_uuids, list) {
2530 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2532 err = btrfs_register_one_device(device->name);
2542 int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
2545 struct btrfs_super_block *disk_super;
2549 buf = malloc(BTRFS_SUPER_INFO_SIZE);
2554 ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
2555 if (ret != BTRFS_SUPER_INFO_SIZE)
2559 disk_super = (struct btrfs_super_block *)buf;
2561 * Accept devices from the same filesystem, allow partially created
2564 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC &&
2565 btrfs_super_magic(disk_super) != BTRFS_MAGIC_PARTIAL)
2568 if (!memcmp(disk_super->fsid, root->fs_info->super_copy->fsid,
2578 * Note: this function uses a static per-thread buffer. Do not call this
2579 * function more than 10 times within one argument list!
2581 const char *pretty_size_mode(u64 size, unsigned mode)
2583 static __thread int ps_index = 0;
2584 static __thread char ps_array[10][32];
2587 ret = ps_array[ps_index];
2590 (void)pretty_size_snprintf(size, ret, 32, mode);
2595 static const char* unit_suffix_binary[] =
2596 { "B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB"};
2597 static const char* unit_suffix_decimal[] =
2598 { "B", "kB", "MB", "GB", "TB", "PB", "EB"};
2600 int pretty_size_snprintf(u64 size, char *str, size_t str_size, unsigned unit_mode)
2606 const char** suffix = NULL;
2613 negative = !!(unit_mode & UNITS_NEGATIVE);
2614 unit_mode &= ~UNITS_NEGATIVE;
2616 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_RAW) {
2618 snprintf(str, str_size, "%lld", size);
2620 snprintf(str, str_size, "%llu", size);
2624 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_BINARY) {
2627 suffix = unit_suffix_binary;
2628 } else if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_DECIMAL) {
2631 suffix = unit_suffix_decimal;
2636 fprintf(stderr, "INTERNAL ERROR: unknown unit base, mode %d\n",
2644 switch (unit_mode & UNITS_MODE_MASK) {
2645 case UNITS_TBYTES: base *= mult; num_divs++;
2646 case UNITS_GBYTES: base *= mult; num_divs++;
2647 case UNITS_MBYTES: base *= mult; num_divs++;
2648 case UNITS_KBYTES: num_divs++;
2656 s64 ssize = (s64)size;
2657 s64 last_ssize = ssize;
2659 while ((ssize < 0 ? -ssize : ssize) >= mult) {
2664 last_size = (u64)last_ssize;
2666 while (size >= mult) {
2673 * If the value is smaller than base, we didn't do any
2674 * division, in that case, base should be 1, not original
2675 * base, or the unit will be wrong
2681 if (num_divs >= ARRAY_SIZE(unit_suffix_binary)) {
2683 printf("INTERNAL ERROR: unsupported unit suffix, index %d\n",
2690 fraction = (float)(s64)last_size / base;
2692 fraction = (float)last_size / base;
2695 return snprintf(str, str_size, "%.2f%s", fraction, suffix[num_divs]);
2699 * __strncpy_null - strncpy with null termination
2700 * @dest: the target array
2701 * @src: the source string
2702 * @n: maximum bytes to copy (size of *dest)
2704 * Like strncpy, but ensures destination is null-terminated.
2706 * Copies the string pointed to by src, including the terminating null
2707 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
2708 * of n bytes. Then ensure that dest is null-terminated.
2710 char *__strncpy_null(char *dest, const char *src, size_t n)
2712 strncpy(dest, src, n);
2719 * Checks to make sure that the label matches our requirements.
2721 0 if everything is safe and usable
2722 -1 if the label is too long
2724 static int check_label(const char *input)
2726 int len = strlen(input);
2728 if (len > BTRFS_LABEL_SIZE - 1) {
2729 error("label %s is too long (max %d)", input,
2730 BTRFS_LABEL_SIZE - 1);
2737 static int set_label_unmounted(const char *dev, const char *label)
2739 struct btrfs_trans_handle *trans;
2740 struct btrfs_root *root;
2743 ret = check_mounted(dev);
2745 error("checking mount status of %s failed: %d", dev, ret);
2749 error("device %s is mounted, use mount point", dev);
2753 /* Open the super_block at the default location
2754 * and as read-write.
2756 root = open_ctree(dev, 0, OPEN_CTREE_WRITES);
2757 if (!root) /* errors are printed by open_ctree() */
2760 trans = btrfs_start_transaction(root, 1);
2761 __strncpy_null(root->fs_info->super_copy->label, label, BTRFS_LABEL_SIZE - 1);
2763 btrfs_commit_transaction(trans, root);
2765 /* Now we close it since we are done. */
2770 static int set_label_mounted(const char *mount_path, const char *labelp)
2773 char label[BTRFS_LABEL_SIZE];
2775 fd = open(mount_path, O_RDONLY | O_NOATIME);
2777 error("unable to access %s: %s", mount_path, strerror(errno));
2781 memset(label, 0, sizeof(label));
2782 __strncpy_null(label, labelp, BTRFS_LABEL_SIZE - 1);
2783 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
2784 error("unable to set label of %s: %s", mount_path,
2794 int get_label_unmounted(const char *dev, char *label)
2796 struct btrfs_root *root;
2799 ret = check_mounted(dev);
2801 error("checking mount status of %s failed: %d", dev, ret);
2805 /* Open the super_block at the default location
2808 root = open_ctree(dev, 0, 0);
2812 __strncpy_null(label, root->fs_info->super_copy->label,
2813 BTRFS_LABEL_SIZE - 1);
2815 /* Now we close it since we are done. */
2821 * If a partition is mounted, try to get the filesystem label via its
2822 * mounted path rather than device. Return the corresponding error
2823 * the user specified the device path.
2825 int get_label_mounted(const char *mount_path, char *labelp)
2827 char label[BTRFS_LABEL_SIZE];
2831 fd = open(mount_path, O_RDONLY | O_NOATIME);
2833 error("unable to access %s: %s", mount_path, strerror(errno));
2837 memset(label, '\0', sizeof(label));
2838 ret = ioctl(fd, BTRFS_IOC_GET_FSLABEL, label);
2840 if (errno != ENOTTY)
2841 error("unable to get label of %s: %s", mount_path,
2848 __strncpy_null(labelp, label, BTRFS_LABEL_SIZE - 1);
2853 int get_label(const char *btrfs_dev, char *label)
2857 ret = is_existing_blk_or_reg_file(btrfs_dev);
2859 ret = get_label_mounted(btrfs_dev, label);
2861 ret = get_label_unmounted(btrfs_dev, label);
2866 int set_label(const char *btrfs_dev, const char *label)
2870 if (check_label(label))
2873 ret = is_existing_blk_or_reg_file(btrfs_dev);
2875 ret = set_label_mounted(btrfs_dev, label);
2877 ret = set_label_unmounted(btrfs_dev, label);
2883 * A not-so-good version fls64. No fascinating optimization since
2884 * no one except parse_size use it
2886 static int fls64(u64 x)
2890 for (i = 0; i <64; i++)
2891 if (x << i & (1ULL << 63))
2896 u64 parse_size(char *s)
2904 error("size value is empty");
2908 error("size value '%s' is less equal than 0", s);
2911 ret = strtoull(s, &endptr, 10);
2913 error("size value '%s' is invalid", s);
2916 if (endptr[0] && endptr[1]) {
2917 error("illegal suffix contains character '%c' in wrong position",
2922 * strtoll returns LLONG_MAX when overflow, if this happens,
2923 * need to call strtoull to get the real size
2925 if (errno == ERANGE && ret == ULLONG_MAX) {
2926 error("size value '%s' is too large for u64", s);
2930 c = tolower(endptr[0]);
2953 error("unknown size descriptor '%c'", c);
2957 /* Check whether ret * mult overflow */
2958 if (fls64(ret) + fls64(mult) - 1 > 64) {
2959 error("size value '%s' is too large for u64", s);
2966 u64 parse_qgroupid(const char *p)
2968 char *s = strchr(p, '/');
2969 const char *ptr_src_end = p + strlen(p);
2970 char *ptr_parse_end = NULL;
2979 /* Numeric format like '0/257' is the primary case */
2981 id = strtoull(p, &ptr_parse_end, 10);
2982 if (ptr_parse_end != ptr_src_end)
2986 level = strtoull(p, &ptr_parse_end, 10);
2987 if (ptr_parse_end != s)
2990 id = strtoull(s + 1, &ptr_parse_end, 10);
2991 if (ptr_parse_end != ptr_src_end)
2994 return (level << BTRFS_QGROUP_LEVEL_SHIFT) | id;
2997 /* Path format like subv at 'my_subvol' is the fallback case */
2998 ret = test_issubvolume(p);
2999 if (ret < 0 || !ret)
3001 fd = open(p, O_RDONLY);
3004 ret = lookup_path_rootid(fd, &id);
3006 error("failed to lookup root id: %s", strerror(-ret));
3013 error("invalid qgroupid or subvolume path: %s", p);
3017 int open_file_or_dir3(const char *fname, DIR **dirstream, int open_flags)
3023 ret = stat(fname, &st);
3027 if (S_ISDIR(st.st_mode)) {
3028 *dirstream = opendir(fname);
3031 fd = dirfd(*dirstream);
3032 } else if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) {
3033 fd = open(fname, open_flags);
3036 * we set this on purpose, in case the caller output
3037 * strerror(errno) as success
3045 closedir(*dirstream);
3052 int open_file_or_dir(const char *fname, DIR **dirstream)
3054 return open_file_or_dir3(fname, dirstream, O_RDWR);
3057 void close_file_or_dir(int fd, DIR *dirstream)
3060 closedir(dirstream);
3065 int get_device_info(int fd, u64 devid,
3066 struct btrfs_ioctl_dev_info_args *di_args)
3070 di_args->devid = devid;
3071 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
3073 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
3074 return ret < 0 ? -errno : 0;
3077 static u64 find_max_device_id(struct btrfs_ioctl_search_args *search_args,
3080 struct btrfs_dev_item *dev_item;
3081 char *buf = search_args->buf;
3083 buf += (nr_items - 1) * (sizeof(struct btrfs_ioctl_search_header)
3084 + sizeof(struct btrfs_dev_item));
3085 buf += sizeof(struct btrfs_ioctl_search_header);
3087 dev_item = (struct btrfs_dev_item *)buf;
3089 return btrfs_stack_device_id(dev_item);
3092 static int search_chunk_tree_for_fs_info(int fd,
3093 struct btrfs_ioctl_fs_info_args *fi_args)
3097 u64 start_devid = 1;
3098 struct btrfs_ioctl_search_args search_args;
3099 struct btrfs_ioctl_search_key *search_key = &search_args.key;
3101 fi_args->num_devices = 0;
3103 max_items = BTRFS_SEARCH_ARGS_BUFSIZE
3104 / (sizeof(struct btrfs_ioctl_search_header)
3105 + sizeof(struct btrfs_dev_item));
3107 search_key->tree_id = BTRFS_CHUNK_TREE_OBJECTID;
3108 search_key->min_objectid = BTRFS_DEV_ITEMS_OBJECTID;
3109 search_key->max_objectid = BTRFS_DEV_ITEMS_OBJECTID;
3110 search_key->min_type = BTRFS_DEV_ITEM_KEY;
3111 search_key->max_type = BTRFS_DEV_ITEM_KEY;
3112 search_key->min_transid = 0;
3113 search_key->max_transid = (u64)-1;
3114 search_key->nr_items = max_items;
3115 search_key->max_offset = (u64)-1;
3118 search_key->min_offset = start_devid;
3120 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH, &search_args);
3124 fi_args->num_devices += (u64)search_key->nr_items;
3126 if (search_key->nr_items == max_items) {
3127 start_devid = find_max_device_id(&search_args,
3128 search_key->nr_items) + 1;
3132 /* get the lastest max_id to stay consistent with the num_devices */
3133 if (search_key->nr_items == 0)
3135 * last tree_search returns an empty buf, use the devid of
3136 * the last dev_item of the previous tree_search
3138 fi_args->max_id = start_devid - 1;
3140 fi_args->max_id = find_max_device_id(&search_args,
3141 search_key->nr_items);
3147 * For a given path, fill in the ioctl fs_ and info_ args.
3148 * If the path is a btrfs mountpoint, fill info for all devices.
3149 * If the path is a btrfs device, fill in only that device.
3151 * The path provided must be either on a mounted btrfs fs,
3152 * or be a mounted btrfs device.
3154 * Returns 0 on success, or a negative errno.
3156 int get_fs_info(const char *path, struct btrfs_ioctl_fs_info_args *fi_args,
3157 struct btrfs_ioctl_dev_info_args **di_ret)
3164 struct btrfs_fs_devices *fs_devices_mnt = NULL;
3165 struct btrfs_ioctl_dev_info_args *di_args;
3166 struct btrfs_ioctl_dev_info_args tmp;
3168 DIR *dirstream = NULL;
3170 memset(fi_args, 0, sizeof(*fi_args));
3172 if (is_block_device(path) == 1) {
3173 struct btrfs_super_block *disk_super;
3174 char buf[BTRFS_SUPER_INFO_SIZE];
3176 /* Ensure it's mounted, then set path to the mountpoint */
3177 fd = open(path, O_RDONLY);
3180 error("cannot open %s: %s", path, strerror(errno));
3183 ret = check_mounted_where(fd, path, mp, sizeof(mp),
3192 /* Only fill in this one device */
3193 fi_args->num_devices = 1;
3195 disk_super = (struct btrfs_super_block *)buf;
3196 ret = btrfs_read_dev_super(fd, disk_super,
3197 BTRFS_SUPER_INFO_OFFSET, 0);
3202 last_devid = btrfs_stack_device_id(&disk_super->dev_item);
3203 fi_args->max_id = last_devid;
3205 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
3209 /* at this point path must not be for a block device */
3210 fd = open_file_or_dir(path, &dirstream);
3216 /* fill in fi_args if not just a single device */
3217 if (fi_args->num_devices != 1) {
3218 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
3225 * The fs_args->num_devices does not include seed devices
3227 ret = search_chunk_tree_for_fs_info(fd, fi_args);
3232 * search_chunk_tree_for_fs_info() will lacks the devid 0
3233 * so manual probe for it here.
3235 ret = get_device_info(fd, 0, &tmp);
3237 fi_args->num_devices++;
3240 if (last_devid == 0)
3245 if (!fi_args->num_devices)
3248 di_args = *di_ret = malloc((fi_args->num_devices) * sizeof(*di_args));
3255 memcpy(di_args, &tmp, sizeof(tmp));
3256 for (; last_devid <= fi_args->max_id; last_devid++) {
3257 ret = get_device_info(fd, last_devid, &di_args[ndevs]);
3266 * only when the only dev we wanted to find is not there then
3267 * let any error be returned
3269 if (fi_args->num_devices != 1) {
3275 close_file_or_dir(fd, dirstream);
3279 #define isoctal(c) (((c) & ~7) == '0')
3281 static inline void translate(char *f, char *t)
3283 while (*f != '\0') {
3285 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
3286 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
3296 * Checks if the swap device.
3297 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
3299 static int is_swap_device(const char *file)
3310 if (stat(file, &st_buf) < 0)
3312 if (S_ISBLK(st_buf.st_mode))
3313 dev = st_buf.st_rdev;
3314 else if (S_ISREG(st_buf.st_mode)) {
3315 dev = st_buf.st_dev;
3316 ino = st_buf.st_ino;
3320 if ((f = fopen("/proc/swaps", "r")) == NULL)
3323 /* skip the first line */
3324 if (fgets(tmp, sizeof(tmp), f) == NULL)
3327 while (fgets(tmp, sizeof(tmp), f) != NULL) {
3328 if ((cp = strchr(tmp, ' ')) != NULL)
3330 if ((cp = strchr(tmp, '\t')) != NULL)
3332 translate(tmp, buf);
3333 if (stat(buf, &st_buf) != 0)
3335 if (S_ISBLK(st_buf.st_mode)) {
3336 if (dev == st_buf.st_rdev) {
3340 } else if (S_ISREG(st_buf.st_mode)) {
3341 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {
3355 * Check for existing filesystem or partition table on device.
3357 * 1 for existing fs or partition
3358 * 0 for nothing found
3359 * -1 for internal error
3361 static int check_overwrite(const char *device)
3364 blkid_probe pr = NULL;
3368 if (!device || !*device)
3371 ret = -1; /* will reset on success of all setup calls */
3373 pr = blkid_new_probe_from_filename(device);
3377 size = blkid_probe_get_size(pr);
3381 /* nothing to overwrite on a 0-length device */
3387 ret = blkid_probe_enable_partitions(pr, 1);
3391 ret = blkid_do_fullprobe(pr);
3396 * Blkid returns 1 for nothing found and 0 when it finds a signature,
3397 * but we want the exact opposite, so reverse the return value here.
3399 * In addition print some useful diagnostics about what actually is
3407 if (!blkid_probe_lookup_value(pr, "TYPE", &type, NULL)) {
3409 "%s appears to contain an existing "
3410 "filesystem (%s).\n", device, type);
3411 } else if (!blkid_probe_lookup_value(pr, "PTTYPE", &type, NULL)) {
3413 "%s appears to contain a partition "
3414 "table (%s).\n", device, type);
3417 "%s appears to contain something weird "
3418 "according to blkid\n", device);
3424 blkid_free_probe(pr);
3427 "probe of %s failed, cannot detect "
3428 "existing filesystem.\n", device);
3432 static int group_profile_devs_min(u64 flag)
3434 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3435 case 0: /* single */
3436 case BTRFS_BLOCK_GROUP_DUP:
3438 case BTRFS_BLOCK_GROUP_RAID0:
3439 case BTRFS_BLOCK_GROUP_RAID1:
3440 case BTRFS_BLOCK_GROUP_RAID5:
3442 case BTRFS_BLOCK_GROUP_RAID6:
3444 case BTRFS_BLOCK_GROUP_RAID10:
3451 int test_num_disk_vs_raid(u64 metadata_profile, u64 data_profile,
3452 u64 dev_cnt, int mixed, int ssd)
3455 u64 profile = metadata_profile | data_profile;
3460 allowed |= BTRFS_BLOCK_GROUP_RAID10;
3462 allowed |= BTRFS_BLOCK_GROUP_RAID6;
3464 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
3465 BTRFS_BLOCK_GROUP_RAID5;
3467 allowed |= BTRFS_BLOCK_GROUP_DUP;
3470 if (dev_cnt > 1 && profile & BTRFS_BLOCK_GROUP_DUP) {
3471 warning("DUP is not recommended on filesystem with multiple devices");
3473 if (metadata_profile & ~allowed) {
3475 "ERROR: unable to create FS with metadata profile %s "
3476 "(have %llu devices but %d devices are required)\n",
3477 btrfs_group_profile_str(metadata_profile), dev_cnt,
3478 group_profile_devs_min(metadata_profile));
3481 if (data_profile & ~allowed) {
3483 "ERROR: unable to create FS with data profile %s "
3484 "(have %llu devices but %d devices are required)\n",
3485 btrfs_group_profile_str(data_profile), dev_cnt,
3486 group_profile_devs_min(data_profile));
3490 if (dev_cnt == 3 && profile & BTRFS_BLOCK_GROUP_RAID6) {
3491 warning("RAID6 is not recommended on filesystem with 3 devices only");
3493 if (dev_cnt == 2 && profile & BTRFS_BLOCK_GROUP_RAID5) {
3494 warning("RAID5 is not recommended on filesystem with 2 devices only");
3496 warning_on(!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP) && ssd,
3497 "DUP may not actually lead to 2 copies on the device, see manual page");
3502 int group_profile_max_safe_loss(u64 flags)
3504 switch (flags & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3505 case 0: /* single */
3506 case BTRFS_BLOCK_GROUP_DUP:
3507 case BTRFS_BLOCK_GROUP_RAID0:
3509 case BTRFS_BLOCK_GROUP_RAID1:
3510 case BTRFS_BLOCK_GROUP_RAID5:
3511 case BTRFS_BLOCK_GROUP_RAID10:
3513 case BTRFS_BLOCK_GROUP_RAID6:
3521 * Check if a device is suitable for btrfs
3523 * 1: something is wrong, an error is printed
3526 int test_dev_for_mkfs(const char *file, int force_overwrite)
3531 ret = is_swap_device(file);
3533 error("checking status of %s: %s", file, strerror(-ret));
3537 error("%s is a swap device", file);
3540 if (!force_overwrite) {
3541 if (check_overwrite(file)) {
3542 error("use the -f option to force overwrite of %s",
3547 ret = check_mounted(file);
3549 error("cannot check mount status of %s: %s", file,
3554 error("%s is mounted", file);
3557 /* check if the device is busy */
3558 fd = open(file, O_RDWR|O_EXCL);
3560 error("unable to open %s: %s", file, strerror(errno));
3563 if (fstat(fd, &st)) {
3564 error("unable to stat %s: %s", file, strerror(errno));
3568 if (!S_ISBLK(st.st_mode)) {
3569 error("%s is not a block device", file);
3577 int btrfs_scan_devices(void)
3582 struct btrfs_fs_devices *tmp_devices;
3583 blkid_dev_iterate iter = NULL;
3584 blkid_dev dev = NULL;
3585 blkid_cache cache = NULL;
3586 char path[PATH_MAX];
3588 if (btrfs_scan_done)
3591 if (blkid_get_cache(&cache, NULL) < 0) {
3592 error("blkid cache get failed");
3595 blkid_probe_all(cache);
3596 iter = blkid_dev_iterate_begin(cache);
3597 blkid_dev_set_search(iter, "TYPE", "btrfs");
3598 while (blkid_dev_next(iter, &dev) == 0) {
3599 dev = blkid_verify(cache, dev);
3602 /* if we are here its definitely a btrfs disk*/
3603 strncpy_null(path, blkid_dev_devname(dev));
3605 fd = open(path, O_RDONLY);
3607 error("cannot open %s: %s", path, strerror(errno));
3610 ret = btrfs_scan_one_device(fd, path, &tmp_devices,
3611 &num_devices, BTRFS_SUPER_INFO_OFFSET,
3614 error("cannot scan %s: %s", path, strerror(-ret));
3621 blkid_dev_iterate_end(iter);
3622 blkid_put_cache(cache);
3624 btrfs_scan_done = 1;
3629 int is_vol_small(const char *file)
3636 fd = open(file, O_RDONLY);
3639 if (fstat(fd, &st) < 0) {
3644 size = btrfs_device_size(fd, &st);
3649 if (size < BTRFS_MKFS_SMALL_VOLUME_SIZE) {
3659 * This reads a line from the stdin and only returns non-zero if the
3660 * first whitespace delimited token is a case insensitive match with yes
3663 int ask_user(const char *question)
3665 char buf[30] = {0,};
3666 char *saveptr = NULL;
3669 printf("%s [y/N]: ", question);
3671 return fgets(buf, sizeof(buf) - 1, stdin) &&
3672 (answer = strtok_r(buf, " \t\n\r", &saveptr)) &&
3673 (!strcasecmp(answer, "yes") || !strcasecmp(answer, "y"));
3677 * return 0 if a btrfs mount point is found
3678 * return 1 if a mount point is found but not btrfs
3679 * return <0 if something goes wrong
3681 int find_mount_root(const char *path, char **mount_root)
3689 int longest_matchlen = 0;
3690 char *longest_match = NULL;
3692 fd = open(path, O_RDONLY | O_NOATIME);
3697 mnttab = setmntent("/proc/self/mounts", "r");
3701 while ((ent = getmntent(mnttab))) {
3702 len = strlen(ent->mnt_dir);
3703 if (strncmp(ent->mnt_dir, path, len) == 0) {
3704 /* match found and use the latest match */
3705 if (longest_matchlen <= len) {
3706 free(longest_match);
3707 longest_matchlen = len;
3708 longest_match = strdup(ent->mnt_dir);
3709 not_btrfs = strcmp(ent->mnt_type, "btrfs");
3718 free(longest_match);
3723 *mount_root = realpath(longest_match, NULL);
3727 free(longest_match);
3731 int test_minimum_size(const char *file, u32 nodesize)
3734 struct stat statbuf;
3736 fd = open(file, O_RDONLY);
3739 if (stat(file, &statbuf) < 0) {
3743 if (btrfs_device_size(fd, &statbuf) < btrfs_min_dev_size(nodesize)) {
3753 * Test if path is a directory
3755 * 0 - path exists but it is not a directory
3756 * 1 - path exists and it is a directory
3759 int test_isdir(const char *path)
3764 ret = stat(path, &st);
3768 return !!S_ISDIR(st.st_mode);
3771 void units_set_mode(unsigned *units, unsigned mode)
3773 unsigned base = *units & UNITS_MODE_MASK;
3775 *units = base | mode;
3778 void units_set_base(unsigned *units, unsigned base)
3780 unsigned mode = *units & ~UNITS_MODE_MASK;
3782 *units = base | mode;
3785 int find_next_key(struct btrfs_path *path, struct btrfs_key *key)
3789 for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
3790 if (!path->nodes[level])
3792 if (path->slots[level] + 1 >=
3793 btrfs_header_nritems(path->nodes[level]))
3796 btrfs_item_key_to_cpu(path->nodes[level], key,
3797 path->slots[level] + 1);
3799 btrfs_node_key_to_cpu(path->nodes[level], key,
3800 path->slots[level] + 1);
3806 const char* btrfs_group_type_str(u64 flag)
3808 u64 mask = BTRFS_BLOCK_GROUP_TYPE_MASK |
3809 BTRFS_SPACE_INFO_GLOBAL_RSV;
3811 switch (flag & mask) {
3812 case BTRFS_BLOCK_GROUP_DATA:
3814 case BTRFS_BLOCK_GROUP_SYSTEM:
3816 case BTRFS_BLOCK_GROUP_METADATA:
3818 case BTRFS_BLOCK_GROUP_DATA|BTRFS_BLOCK_GROUP_METADATA:
3819 return "Data+Metadata";
3820 case BTRFS_SPACE_INFO_GLOBAL_RSV:
3821 return "GlobalReserve";
3827 const char* btrfs_group_profile_str(u64 flag)
3829 switch (flag & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
3832 case BTRFS_BLOCK_GROUP_RAID0:
3834 case BTRFS_BLOCK_GROUP_RAID1:
3836 case BTRFS_BLOCK_GROUP_RAID5:
3838 case BTRFS_BLOCK_GROUP_RAID6:
3840 case BTRFS_BLOCK_GROUP_DUP:
3842 case BTRFS_BLOCK_GROUP_RAID10:
3849 u64 disk_size(const char *path)
3853 if (statfs(path, &sfs) < 0)
3856 return sfs.f_bsize * sfs.f_blocks;
3859 u64 get_partition_size(const char *dev)
3862 int fd = open(dev, O_RDONLY);
3866 if (ioctl(fd, BLKGETSIZE64, &result) < 0) {
3876 * Check if the BTRFS_IOC_TREE_SEARCH_V2 ioctl is supported on a given
3877 * filesystem, opened at fd
3879 int btrfs_tree_search2_ioctl_supported(int fd)
3881 struct btrfs_ioctl_search_args_v2 *args2;
3882 struct btrfs_ioctl_search_key *sk;
3883 int args2_size = 1024;
3884 char args2_buf[args2_size];
3887 args2 = (struct btrfs_ioctl_search_args_v2 *)args2_buf;
3891 * Search for the extent tree item in the root tree.
3893 sk->tree_id = BTRFS_ROOT_TREE_OBJECTID;
3894 sk->min_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3895 sk->max_objectid = BTRFS_EXTENT_TREE_OBJECTID;
3896 sk->min_type = BTRFS_ROOT_ITEM_KEY;
3897 sk->max_type = BTRFS_ROOT_ITEM_KEY;
3899 sk->max_offset = (u64)-1;
3900 sk->min_transid = 0;
3901 sk->max_transid = (u64)-1;
3903 args2->buf_size = args2_size - sizeof(struct btrfs_ioctl_search_args_v2);
3904 ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH_V2, args2);
3905 if (ret == -EOPNOTSUPP)
3912 int btrfs_check_nodesize(u32 nodesize, u32 sectorsize, u64 features)
3914 if (nodesize < sectorsize) {
3915 error("illegal nodesize %u (smaller than %u)",
3916 nodesize, sectorsize);
3918 } else if (nodesize > BTRFS_MAX_METADATA_BLOCKSIZE) {
3919 error("illegal nodesize %u (larger than %u)",
3920 nodesize, BTRFS_MAX_METADATA_BLOCKSIZE);
3922 } else if (nodesize & (sectorsize - 1)) {
3923 error("illegal nodesize %u (not aligned to %u)",
3924 nodesize, sectorsize);
3926 } else if (features & BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS &&
3927 nodesize != sectorsize) {
3928 error("illegal nodesize %u (not equal to %u for mixed block group)",
3929 nodesize, sectorsize);
3936 * Copy a path argument from SRC to DEST and check the SRC length if it's at
3937 * most PATH_MAX and fits into DEST. DESTLEN is supposed to be exact size of
3939 * The destination buffer is zero terminated.
3940 * Return < 0 for error, 0 otherwise.
3942 int arg_copy_path(char *dest, const char *src, int destlen)
3944 size_t len = strlen(src);
3946 if (len >= PATH_MAX || len >= destlen)
3947 return -ENAMETOOLONG;
3949 __strncpy_null(dest, src, destlen);
3954 unsigned int get_unit_mode_from_arg(int *argc, char *argv[], int df_mode)
3956 unsigned int unit_mode = UNITS_DEFAULT;
3960 for (arg_i = 0; arg_i < *argc; arg_i++) {
3961 if (!strcmp(argv[arg_i], "--"))
3964 if (!strcmp(argv[arg_i], "--raw")) {
3965 unit_mode = UNITS_RAW;
3969 if (!strcmp(argv[arg_i], "--human-readable")) {
3970 unit_mode = UNITS_HUMAN_BINARY;
3975 if (!strcmp(argv[arg_i], "--iec")) {
3976 units_set_mode(&unit_mode, UNITS_BINARY);
3980 if (!strcmp(argv[arg_i], "--si")) {
3981 units_set_mode(&unit_mode, UNITS_DECIMAL);
3986 if (!strcmp(argv[arg_i], "--kbytes")) {
3987 units_set_base(&unit_mode, UNITS_KBYTES);
3991 if (!strcmp(argv[arg_i], "--mbytes")) {
3992 units_set_base(&unit_mode, UNITS_MBYTES);
3996 if (!strcmp(argv[arg_i], "--gbytes")) {
3997 units_set_base(&unit_mode, UNITS_GBYTES);
4001 if (!strcmp(argv[arg_i], "--tbytes")) {
4002 units_set_base(&unit_mode, UNITS_TBYTES);
4010 if (!strcmp(argv[arg_i], "-b")) {
4011 unit_mode = UNITS_RAW;
4015 if (!strcmp(argv[arg_i], "-h")) {
4016 unit_mode = UNITS_HUMAN_BINARY;
4020 if (!strcmp(argv[arg_i], "-H")) {
4021 unit_mode = UNITS_HUMAN_DECIMAL;
4025 if (!strcmp(argv[arg_i], "-k")) {
4026 units_set_base(&unit_mode, UNITS_KBYTES);
4030 if (!strcmp(argv[arg_i], "-m")) {
4031 units_set_base(&unit_mode, UNITS_MBYTES);
4035 if (!strcmp(argv[arg_i], "-g")) {
4036 units_set_base(&unit_mode, UNITS_GBYTES);
4040 if (!strcmp(argv[arg_i], "-t")) {
4041 units_set_base(&unit_mode, UNITS_TBYTES);
4047 for (arg_i = 0, arg_end = 0; arg_i < *argc; arg_i++) {
4050 argv[arg_end] = argv[arg_i];
4059 int string_is_numerical(const char *str)
4063 if (!(*str >= '0' && *str <= '9'))
4065 while (*str >= '0' && *str <= '9')
4073 * Preprocess @argv with getopt_long to reorder options and consume the "--"
4075 * Unknown short and long options are reported, optionally the @usage is printed
4078 void clean_args_no_options(int argc, char *argv[], const char * const *usagestr)
4080 static const struct option long_options[] = {
4085 int c = getopt_long(argc, argv, "", long_options, NULL);
4099 * Same as clean_args_no_options but pass through arguments that could look
4100 * like short options. Eg. reisze which takes a negative resize argument like
4103 * This accepts only two forms:
4104 * - "-- option1 option2 ..."
4105 * - "option1 option2 ..."
4107 void clean_args_no_options_relaxed(int argc, char *argv[], const char * const *usagestr)
4112 if (strcmp(argv[1], "--") == 0)
4116 /* Subvolume helper functions */
4118 * test if name is a correct subvolume name
4119 * this function return
4120 * 0-> name is not a correct subvolume name
4121 * 1-> name is a correct subvolume name
4123 int test_issubvolname(const char *name)
4125 return name[0] != '\0' && !strchr(name, '/') &&
4126 strcmp(name, ".") && strcmp(name, "..");
4130 * Test if path is a subvolume
4132 * 0 - path exists but it is not a subvolume
4133 * 1 - path exists and it is a subvolume
4136 int test_issubvolume(const char *path)
4142 res = stat(path, &st);
4146 if (st.st_ino != BTRFS_FIRST_FREE_OBJECTID || !S_ISDIR(st.st_mode))
4149 res = statfs(path, &stfs);
4153 return (int)stfs.f_type == BTRFS_SUPER_MAGIC;
4156 const char *subvol_strip_mountpoint(const char *mnt, const char *full_path)
4158 int len = strlen(mnt);
4162 if (mnt[len - 1] != '/')
4165 return full_path + len;
4172 * 1: Error; and error info printed to the terminal. Fixme.
4173 * 2: If the fullpath is root tree instead of subvol tree
4175 int get_subvol_info(const char *fullpath, struct root_info *get_ri)
4182 const char *svpath = NULL;
4183 DIR *dirstream1 = NULL;
4184 DIR *dirstream2 = NULL;
4186 ret = test_issubvolume(fullpath);
4190 error("not a subvolume: %s", fullpath);
4194 ret = find_mount_root(fullpath, &mnt);
4198 error("%s doesn't belong to btrfs mount point", fullpath);
4202 svpath = subvol_strip_mountpoint(mnt, fullpath);
4204 fd = btrfs_open_dir(fullpath, &dirstream1, 1);
4208 ret = btrfs_list_get_path_rootid(fd, &sv_id);
4212 mntfd = btrfs_open_dir(mnt, &dirstream2, 1);
4216 memset(get_ri, 0, sizeof(*get_ri));
4217 get_ri->root_id = sv_id;
4219 if (sv_id == BTRFS_FS_TREE_OBJECTID)
4220 ret = btrfs_get_toplevel_subvol(mntfd, get_ri);
4222 ret = btrfs_get_subvol(mntfd, get_ri);
4224 error("can't find '%s': %d", svpath, ret);
4227 close_file_or_dir(mntfd, dirstream2);
4228 close_file_or_dir(fd, dirstream1);
4234 void init_rand_seed(u64 seed)
4238 /* only use the last 48 bits */
4239 for (i = 0; i < 3; i++) {
4240 rand_seed[i] = (unsigned short)(seed ^ (unsigned short)(-1));
4243 rand_seed_initlized = 1;
4246 static void __init_seed(void)
4252 if(rand_seed_initlized)
4254 /* Use urandom as primary seed source. */
4255 fd = open("/dev/urandom", O_RDONLY);
4257 ret = read(fd, rand_seed, sizeof(rand_seed));
4259 if (ret < sizeof(rand_seed))
4263 /* Use time and pid as fallback seed */
4264 warning("failed to read /dev/urandom, use time and pid as random seed");
4265 gettimeofday(&tv, 0);
4266 rand_seed[0] = getpid() ^ (tv.tv_sec & 0xFFFF);
4267 rand_seed[1] = getppid() ^ (tv.tv_usec & 0xFFFF);
4268 rand_seed[2] = (tv.tv_sec ^ tv.tv_usec) >> 16;
4270 rand_seed_initlized = 1;
4277 * Don't use nrand48, its range is [0,2^31) The highest bit will alwasy
4278 * be 0. Use jrand48 to include the highest bit.
4280 return (u32)jrand48(rand_seed);
4283 unsigned int rand_range(unsigned int upper)
4287 * Use the full 48bits to mod, which would be more uniformly
4290 return (unsigned int)(jrand48(rand_seed) % upper);
4293 void btrfs_config_init(void)