2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include "kerncompat.h"
21 #include <sys/ioctl.h>
22 #include <sys/mount.h>
26 #include <sys/types.h>
32 #include <uuid/uuid.h>
34 #include <sys/xattr.h>
36 #include <linux/limits.h>
37 #include <blkid/blkid.h>
42 #include "transaction.h"
45 static u64 index_cnt = 2;
46 static int verbose = 1;
48 struct directory_name_entry {
52 struct list_head list;
55 struct mkfs_allocation {
62 static int create_metadata_block_groups(struct btrfs_root *root, int mixed,
63 struct mkfs_allocation *allocation)
65 struct btrfs_trans_handle *trans;
71 trans = btrfs_start_transaction(root, 1);
72 bytes_used = btrfs_super_bytes_used(root->fs_info->super_copy);
74 root->fs_info->system_allocs = 1;
75 ret = btrfs_make_block_group(trans, root, bytes_used,
76 BTRFS_BLOCK_GROUP_SYSTEM,
77 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
78 0, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
79 allocation->system += BTRFS_MKFS_SYSTEM_GROUP_SIZE;
83 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
84 &chunk_start, &chunk_size,
85 BTRFS_BLOCK_GROUP_METADATA |
86 BTRFS_BLOCK_GROUP_DATA);
89 "no space to alloc data/metadata chunk\n");
93 ret = btrfs_make_block_group(trans, root, 0,
94 BTRFS_BLOCK_GROUP_METADATA |
95 BTRFS_BLOCK_GROUP_DATA,
96 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
97 chunk_start, chunk_size);
99 allocation->mixed += chunk_size;
101 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
102 &chunk_start, &chunk_size,
103 BTRFS_BLOCK_GROUP_METADATA);
104 if (ret == -ENOSPC) {
105 fprintf(stderr, "no space to alloc metadata chunk\n");
109 ret = btrfs_make_block_group(trans, root, 0,
110 BTRFS_BLOCK_GROUP_METADATA,
111 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
112 chunk_start, chunk_size);
113 allocation->metadata += chunk_size;
117 root->fs_info->system_allocs = 0;
118 btrfs_commit_transaction(trans, root);
124 static int make_root_dir(struct btrfs_root *root, int mixed,
125 struct mkfs_allocation *allocation)
127 struct btrfs_trans_handle *trans;
128 struct btrfs_key location;
133 trans = btrfs_start_transaction(root, 1);
137 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
138 &chunk_start, &chunk_size,
139 BTRFS_BLOCK_GROUP_DATA);
140 if (ret == -ENOSPC) {
141 fprintf(stderr, "no space to alloc data chunk\n");
145 ret = btrfs_make_block_group(trans, root, 0,
146 BTRFS_BLOCK_GROUP_DATA,
147 BTRFS_FIRST_CHUNK_TREE_OBJECTID,
148 chunk_start, chunk_size);
149 allocation->data += chunk_size;
153 ret = btrfs_make_root_dir(trans, root->fs_info->tree_root,
154 BTRFS_ROOT_TREE_DIR_OBJECTID);
157 ret = btrfs_make_root_dir(trans, root, BTRFS_FIRST_FREE_OBJECTID);
160 memcpy(&location, &root->fs_info->fs_root->root_key, sizeof(location));
161 location.offset = (u64)-1;
162 ret = btrfs_insert_dir_item(trans, root->fs_info->tree_root,
164 btrfs_super_root_dir(root->fs_info->super_copy),
165 &location, BTRFS_FT_DIR, 0);
169 ret = btrfs_insert_inode_ref(trans, root->fs_info->tree_root,
170 "default", 7, location.objectid,
171 BTRFS_ROOT_TREE_DIR_OBJECTID, 0);
175 btrfs_commit_transaction(trans, root);
180 static void __recow_root(struct btrfs_trans_handle *trans,
181 struct btrfs_root *root)
184 struct extent_buffer *tmp;
186 if (trans->transid != btrfs_root_generation(&root->root_item)) {
187 extent_buffer_get(root->node);
188 ret = __btrfs_cow_block(trans, root, root->node,
189 NULL, 0, &tmp, 0, 0);
191 free_extent_buffer(tmp);
195 static void recow_roots(struct btrfs_trans_handle *trans,
196 struct btrfs_root *root)
198 struct btrfs_fs_info *info = root->fs_info;
200 __recow_root(trans, info->fs_root);
201 __recow_root(trans, info->tree_root);
202 __recow_root(trans, info->extent_root);
203 __recow_root(trans, info->chunk_root);
204 __recow_root(trans, info->dev_root);
205 __recow_root(trans, info->csum_root);
208 static int create_one_raid_group(struct btrfs_trans_handle *trans,
209 struct btrfs_root *root, u64 type,
210 struct mkfs_allocation *allocation)
217 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
218 &chunk_start, &chunk_size, type);
219 if (ret == -ENOSPC) {
220 fprintf(stderr, "not enough free space\n");
224 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
225 type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
226 chunk_start, chunk_size);
227 if ((type & BTRFS_BLOCK_GROUP_TYPE_MASK) == BTRFS_BLOCK_GROUP_DATA)
228 allocation->data += chunk_size;
229 else if ((type & BTRFS_BLOCK_GROUP_TYPE_MASK) == BTRFS_BLOCK_GROUP_METADATA)
230 allocation->metadata += chunk_size;
231 else if ((type & BTRFS_BLOCK_GROUP_TYPE_MASK) == BTRFS_BLOCK_GROUP_SYSTEM)
232 allocation->system += chunk_size;
233 else if ((type & BTRFS_BLOCK_GROUP_TYPE_MASK) ==
234 (BTRFS_BLOCK_GROUP_METADATA|BTRFS_BLOCK_GROUP_DATA))
235 allocation->mixed += chunk_size;
243 static int create_raid_groups(struct btrfs_trans_handle *trans,
244 struct btrfs_root *root, u64 data_profile,
245 int data_profile_opt, u64 metadata_profile,
247 struct mkfs_allocation *allocation)
249 u64 num_devices = btrfs_super_num_devices(root->fs_info->super_copy);
252 if (metadata_profile) {
253 u64 meta_flags = BTRFS_BLOCK_GROUP_METADATA;
255 ret = create_one_raid_group(trans, root,
256 BTRFS_BLOCK_GROUP_SYSTEM |
257 metadata_profile, allocation);
261 meta_flags |= BTRFS_BLOCK_GROUP_DATA;
263 ret = create_one_raid_group(trans, root, meta_flags |
264 metadata_profile, allocation);
268 if (!mixed && num_devices > 1 && data_profile) {
269 ret = create_one_raid_group(trans, root,
270 BTRFS_BLOCK_GROUP_DATA |
271 data_profile, allocation);
274 recow_roots(trans, root);
279 static int create_data_reloc_tree(struct btrfs_trans_handle *trans,
280 struct btrfs_root *root)
282 struct btrfs_key location;
283 struct btrfs_root_item root_item;
284 struct extent_buffer *tmp;
285 u64 objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
288 ret = btrfs_copy_root(trans, root, root->node, &tmp, objectid);
291 memcpy(&root_item, &root->root_item, sizeof(root_item));
292 btrfs_set_root_bytenr(&root_item, tmp->start);
293 btrfs_set_root_level(&root_item, btrfs_header_level(tmp));
294 btrfs_set_root_generation(&root_item, trans->transid);
295 free_extent_buffer(tmp);
297 location.objectid = objectid;
298 location.type = BTRFS_ROOT_ITEM_KEY;
300 ret = btrfs_insert_root(trans, root->fs_info->tree_root,
301 &location, &root_item);
306 static void print_usage(int ret)
308 fprintf(stderr, "usage: mkfs.btrfs [options] dev [ dev ... ]\n");
309 fprintf(stderr, "options:\n");
310 fprintf(stderr, "\t-A|--alloc-start START the offset to start the FS\n");
311 fprintf(stderr, "\t-b|--byte-count SIZE total number of bytes in the FS\n");
312 fprintf(stderr, "\t-d|--data PROFILE data profile, raid0, raid1, raid5, raid6, raid10, dup or single\n");
313 fprintf(stderr, "\t-f|--force force overwrite of existing filesystem\n");
314 fprintf(stderr, "\t-l|--leafsize SIZE deprecated, alias for nodesize\n");
315 fprintf(stderr, "\t-L|--label LABEL set a label\n");
316 fprintf(stderr, "\t-m|--metadata PROFILE metadata profile, values like data profile\n");
317 fprintf(stderr, "\t-M|--mixed mix metadata and data together\n");
318 fprintf(stderr, "\t-n|--nodesize SIZE size of btree nodes\n");
319 fprintf(stderr, "\t-s|--sectorsize SIZE min block allocation (may not mountable by current kernel)\n");
320 fprintf(stderr, "\t-r|--rootdir DIR the source directory\n");
321 fprintf(stderr, "\t-K|--nodiscard do not perform whole device TRIM\n");
322 fprintf(stderr, "\t-O|--features LIST comma separated list of filesystem features, use '-O list-all' to list features\n");
323 fprintf(stderr, "\t-U|--uuid UUID specify the filesystem UUID\n");
324 fprintf(stderr, "\t-q|--quiet no messages except errors\n");
325 fprintf(stderr, "\t-V|--version print the mkfs.btrfs version and exit\n");
326 fprintf(stderr, "%s\n", PACKAGE_STRING);
330 static void print_version(void) __attribute__((noreturn));
331 static void print_version(void)
333 fprintf(stderr, "mkfs.btrfs, part of %s\n", PACKAGE_STRING);
337 static u64 parse_profile(char *s)
339 if (strcmp(s, "raid0") == 0) {
340 return BTRFS_BLOCK_GROUP_RAID0;
341 } else if (strcasecmp(s, "raid1") == 0) {
342 return BTRFS_BLOCK_GROUP_RAID1;
343 } else if (strcasecmp(s, "raid5") == 0) {
344 return BTRFS_BLOCK_GROUP_RAID5;
345 } else if (strcasecmp(s, "raid6") == 0) {
346 return BTRFS_BLOCK_GROUP_RAID6;
347 } else if (strcasecmp(s, "raid10") == 0) {
348 return BTRFS_BLOCK_GROUP_RAID10;
349 } else if (strcasecmp(s, "dup") == 0) {
350 return BTRFS_BLOCK_GROUP_DUP;
351 } else if (strcasecmp(s, "single") == 0) {
354 fprintf(stderr, "Unknown profile %s\n", s);
360 static char *parse_label(char *input)
362 int len = strlen(input);
364 if (len >= BTRFS_LABEL_SIZE) {
365 fprintf(stderr, "Label %s is too long (max %d)\n", input,
366 BTRFS_LABEL_SIZE - 1);
369 return strdup(input);
372 static int add_directory_items(struct btrfs_trans_handle *trans,
373 struct btrfs_root *root, u64 objectid,
374 ino_t parent_inum, const char *name,
375 struct stat *st, int *dir_index_cnt)
379 struct btrfs_key location;
382 name_len = strlen(name);
384 location.objectid = objectid;
386 btrfs_set_key_type(&location, BTRFS_INODE_ITEM_KEY);
388 if (S_ISDIR(st->st_mode))
389 filetype = BTRFS_FT_DIR;
390 if (S_ISREG(st->st_mode))
391 filetype = BTRFS_FT_REG_FILE;
392 if (S_ISLNK(st->st_mode))
393 filetype = BTRFS_FT_SYMLINK;
395 ret = btrfs_insert_dir_item(trans, root, name, name_len,
396 parent_inum, &location,
397 filetype, index_cnt);
400 ret = btrfs_insert_inode_ref(trans, root, name, name_len,
401 objectid, parent_inum, index_cnt);
402 *dir_index_cnt = index_cnt;
408 static int fill_inode_item(struct btrfs_trans_handle *trans,
409 struct btrfs_root *root,
410 struct btrfs_inode_item *dst, struct stat *src)
413 u64 sectorsize = root->sectorsize;
416 * btrfs_inode_item has some reserved fields
417 * and represents on-disk inode entry, so
418 * zero everything to prevent information leak
420 memset(dst, 0, sizeof (*dst));
422 btrfs_set_stack_inode_generation(dst, trans->transid);
423 btrfs_set_stack_inode_size(dst, src->st_size);
424 btrfs_set_stack_inode_nbytes(dst, 0);
425 btrfs_set_stack_inode_block_group(dst, 0);
426 btrfs_set_stack_inode_nlink(dst, src->st_nlink);
427 btrfs_set_stack_inode_uid(dst, src->st_uid);
428 btrfs_set_stack_inode_gid(dst, src->st_gid);
429 btrfs_set_stack_inode_mode(dst, src->st_mode);
430 btrfs_set_stack_inode_rdev(dst, 0);
431 btrfs_set_stack_inode_flags(dst, 0);
432 btrfs_set_stack_timespec_sec(&dst->atime, src->st_atime);
433 btrfs_set_stack_timespec_nsec(&dst->atime, 0);
434 btrfs_set_stack_timespec_sec(&dst->ctime, src->st_ctime);
435 btrfs_set_stack_timespec_nsec(&dst->ctime, 0);
436 btrfs_set_stack_timespec_sec(&dst->mtime, src->st_mtime);
437 btrfs_set_stack_timespec_nsec(&dst->mtime, 0);
438 btrfs_set_stack_timespec_sec(&dst->otime, 0);
439 btrfs_set_stack_timespec_nsec(&dst->otime, 0);
441 if (S_ISDIR(src->st_mode)) {
442 btrfs_set_stack_inode_size(dst, 0);
443 btrfs_set_stack_inode_nlink(dst, 1);
445 if (S_ISREG(src->st_mode)) {
446 btrfs_set_stack_inode_size(dst, (u64)src->st_size);
447 if (src->st_size <= BTRFS_MAX_INLINE_DATA_SIZE(root))
448 btrfs_set_stack_inode_nbytes(dst, src->st_size);
450 blocks = src->st_size / sectorsize;
451 if (src->st_size % sectorsize)
453 blocks *= sectorsize;
454 btrfs_set_stack_inode_nbytes(dst, blocks);
457 if (S_ISLNK(src->st_mode))
458 btrfs_set_stack_inode_nbytes(dst, src->st_size + 1);
463 static int directory_select(const struct direct *entry)
465 if ((strncmp(entry->d_name, ".", entry->d_reclen) == 0) ||
466 (strncmp(entry->d_name, "..", entry->d_reclen) == 0))
472 static void free_namelist(struct direct **files, int count)
479 for (i = 0; i < count; ++i)
484 static u64 calculate_dir_inode_size(char *dirname)
487 struct direct **files, *cur_file;
488 u64 dir_inode_size = 0;
490 count = scandir(dirname, &files, directory_select, NULL);
492 for (i = 0; i < count; i++) {
494 dir_inode_size += strlen(cur_file->d_name);
497 free_namelist(files, count);
500 return dir_inode_size;
503 static int add_inode_items(struct btrfs_trans_handle *trans,
504 struct btrfs_root *root,
505 struct stat *st, char *name,
506 u64 self_objectid, ino_t parent_inum,
507 int dir_index_cnt, struct btrfs_inode_item *inode_ret)
510 struct btrfs_key inode_key;
511 struct btrfs_inode_item btrfs_inode;
515 fill_inode_item(trans, root, &btrfs_inode, st);
516 objectid = self_objectid;
518 if (S_ISDIR(st->st_mode)) {
519 inode_size = calculate_dir_inode_size(name);
520 btrfs_set_stack_inode_size(&btrfs_inode, inode_size);
523 inode_key.objectid = objectid;
524 inode_key.offset = 0;
525 btrfs_set_key_type(&inode_key, BTRFS_INODE_ITEM_KEY);
527 ret = btrfs_insert_inode(trans, root, objectid, &btrfs_inode);
529 *inode_ret = btrfs_inode;
533 static int add_xattr_item(struct btrfs_trans_handle *trans,
534 struct btrfs_root *root, u64 objectid,
535 const char *file_name)
539 char xattr_list[XATTR_LIST_MAX];
541 char cur_value[XATTR_SIZE_MAX];
542 char delimiter = '\0';
543 char *next_location = xattr_list;
545 ret = llistxattr(file_name, xattr_list, XATTR_LIST_MAX);
549 fprintf(stderr, "get a list of xattr failed for %s\n",
556 cur_name = strtok(xattr_list, &delimiter);
557 while (cur_name != NULL) {
558 cur_name_len = strlen(cur_name);
559 next_location += cur_name_len + 1;
561 ret = getxattr(file_name, cur_name, cur_value, XATTR_SIZE_MAX);
565 fprintf(stderr, "get a xattr value failed for %s attr %s\n",
566 file_name, cur_name);
570 ret = btrfs_insert_xattr_item(trans, root, cur_name,
571 cur_name_len, cur_value,
574 fprintf(stderr, "insert a xattr item failed for %s\n",
578 cur_name = strtok(next_location, &delimiter);
584 static int add_symbolic_link(struct btrfs_trans_handle *trans,
585 struct btrfs_root *root,
586 u64 objectid, const char *path_name)
589 u64 sectorsize = root->sectorsize;
590 char *buf = malloc(sectorsize);
592 ret = readlink(path_name, buf, sectorsize);
594 fprintf(stderr, "readlink failed for %s\n", path_name);
597 if (ret >= sectorsize) {
598 fprintf(stderr, "symlink too long for %s", path_name);
603 buf[ret] = '\0'; /* readlink does not do it for us */
604 ret = btrfs_insert_inline_extent(trans, root, objectid, 0,
611 static int add_file_items(struct btrfs_trans_handle *trans,
612 struct btrfs_root *root,
613 struct btrfs_inode_item *btrfs_inode, u64 objectid,
614 ino_t parent_inum, struct stat *st,
615 const char *path_name, int out_fd)
620 struct btrfs_key key;
622 u32 sectorsize = root->sectorsize;
627 struct extent_buffer *eb = NULL;
630 if (st->st_size == 0)
633 fd = open(path_name, O_RDONLY);
635 fprintf(stderr, "%s open failed\n", path_name);
639 blocks = st->st_size / sectorsize;
640 if (st->st_size % sectorsize)
643 if (st->st_size <= BTRFS_MAX_INLINE_DATA_SIZE(root)) {
644 char *buffer = malloc(st->st_size);
645 ret_read = pread64(fd, buffer, st->st_size, bytes_read);
646 if (ret_read == -1) {
647 fprintf(stderr, "%s read failed\n", path_name);
652 ret = btrfs_insert_inline_extent(trans, root, objectid, 0,
653 buffer, st->st_size);
658 /* round up our st_size to the FS blocksize */
659 total_bytes = (u64)blocks * sectorsize;
662 * do our IO in extent buffers so it can work
663 * against any raid type
665 eb = malloc(sizeof(*eb) + sectorsize);
670 memset(eb, 0, sizeof(*eb) + sectorsize);
675 * keep our extent size at 1MB max, this makes it easier to work inside
676 * the tiny block groups created during mkfs
678 cur_bytes = min(total_bytes, 1024ULL * 1024);
679 ret = btrfs_reserve_extent(trans, root, cur_bytes, 0, 0, (u64)-1,
684 first_block = key.objectid;
687 while (bytes_read < cur_bytes) {
689 memset(eb->data, 0, sectorsize);
691 ret_read = pread64(fd, eb->data, sectorsize, file_pos + bytes_read);
692 if (ret_read == -1) {
693 fprintf(stderr, "%s read failed\n", path_name);
697 eb->start = first_block + bytes_read;
698 eb->len = sectorsize;
701 * we're doing the csum before we record the extent, but
704 ret = btrfs_csum_file_block(trans, root->fs_info->csum_root,
705 first_block + bytes_read + sectorsize,
706 first_block + bytes_read,
707 eb->data, sectorsize);
711 ret = write_and_map_eb(trans, root, eb);
713 fprintf(stderr, "output file write failed\n");
717 bytes_read += sectorsize;
721 ret = btrfs_record_file_extent(trans, root, objectid, btrfs_inode,
722 file_pos, first_block, cur_bytes);
728 file_pos += cur_bytes;
729 total_bytes -= cur_bytes;
740 static char *make_path(char *dir, char *name)
744 path = malloc(strlen(dir) + strlen(name) + 2);
748 if (dir[strlen(dir) - 1] != '/')
754 static int traverse_directory(struct btrfs_trans_handle *trans,
755 struct btrfs_root *root, char *dir_name,
756 struct directory_name_entry *dir_head, int out_fd)
760 struct btrfs_inode_item cur_inode;
761 struct btrfs_inode_item *inode_item;
762 int count, i, dir_index_cnt;
763 struct direct **files;
765 struct directory_name_entry *dir_entry, *parent_dir_entry;
766 struct direct *cur_file;
767 ino_t parent_inum, cur_inum;
768 ino_t highest_inum = 0;
769 char *parent_dir_name;
770 char real_path[PATH_MAX];
771 struct btrfs_path path;
772 struct extent_buffer *leaf;
773 struct btrfs_key root_dir_key;
774 u64 root_dir_inode_size = 0;
776 /* Add list for source directory */
777 dir_entry = malloc(sizeof(struct directory_name_entry));
778 dir_entry->dir_name = dir_name;
779 dir_entry->path = realpath(dir_name, real_path);
780 if (!dir_entry->path) {
781 fprintf(stderr, "get directory real path error\n");
786 parent_inum = highest_inum + BTRFS_FIRST_FREE_OBJECTID;
787 dir_entry->inum = parent_inum;
788 list_add_tail(&dir_entry->list, &dir_head->list);
790 btrfs_init_path(&path);
792 root_dir_key.objectid = btrfs_root_dirid(&root->root_item);
793 root_dir_key.offset = 0;
794 btrfs_set_key_type(&root_dir_key, BTRFS_INODE_ITEM_KEY);
795 ret = btrfs_lookup_inode(trans, root, &path, &root_dir_key, 1);
797 fprintf(stderr, "root dir lookup error\n");
801 leaf = path.nodes[0];
802 inode_item = btrfs_item_ptr(leaf, path.slots[0],
803 struct btrfs_inode_item);
805 root_dir_inode_size = calculate_dir_inode_size(dir_name);
806 btrfs_set_inode_size(leaf, inode_item, root_dir_inode_size);
807 btrfs_mark_buffer_dirty(leaf);
809 btrfs_release_path(&path);
812 parent_dir_entry = list_entry(dir_head->list.next,
813 struct directory_name_entry,
815 list_del(&parent_dir_entry->list);
817 parent_inum = parent_dir_entry->inum;
818 parent_dir_name = parent_dir_entry->dir_name;
819 if (chdir(parent_dir_entry->path)) {
820 fprintf(stderr, "chdir error for %s\n",
826 count = scandir(parent_dir_entry->path, &files,
827 directory_select, NULL);
830 fprintf(stderr, "scandir for %s failed: %s\n",
831 parent_dir_name, strerror (errno));
836 for (i = 0; i < count; i++) {
839 if (lstat(cur_file->d_name, &st) == -1) {
840 fprintf(stderr, "lstat failed for file %s\n",
846 cur_inum = st.st_ino;
847 ret = add_directory_items(trans, root,
848 cur_inum, parent_inum,
850 &st, &dir_index_cnt);
852 fprintf(stderr, "add_directory_items failed\n");
856 ret = add_inode_items(trans, root, &st,
857 cur_file->d_name, cur_inum,
858 parent_inum, dir_index_cnt,
860 if (ret == -EEXIST) {
861 BUG_ON(st.st_nlink <= 1);
865 fprintf(stderr, "add_inode_items failed\n");
869 ret = add_xattr_item(trans, root,
870 cur_inum, cur_file->d_name);
872 fprintf(stderr, "add_xattr_item failed\n");
877 if (S_ISDIR(st.st_mode)) {
878 dir_entry = malloc(sizeof(struct directory_name_entry));
879 dir_entry->dir_name = cur_file->d_name;
880 dir_entry->path = make_path(parent_dir_entry->path,
882 dir_entry->inum = cur_inum;
883 list_add_tail(&dir_entry->list, &dir_head->list);
884 } else if (S_ISREG(st.st_mode)) {
885 ret = add_file_items(trans, root, &cur_inode,
886 cur_inum, parent_inum, &st,
887 cur_file->d_name, out_fd);
889 fprintf(stderr, "add_file_items failed\n");
892 } else if (S_ISLNK(st.st_mode)) {
893 ret = add_symbolic_link(trans, root,
894 cur_inum, cur_file->d_name);
896 fprintf(stderr, "add_symbolic_link failed\n");
902 free_namelist(files, count);
903 free(parent_dir_entry);
907 } while (!list_empty(&dir_head->list));
912 free_namelist(files, count);
914 free(parent_dir_entry);
921 static int open_target(char *output_name)
924 output_fd = open(output_name, O_CREAT | O_RDWR | O_TRUNC,
925 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP | S_IROTH);
930 static int create_chunks(struct btrfs_trans_handle *trans,
931 struct btrfs_root *root, u64 num_of_meta_chunks,
933 struct mkfs_allocation *allocation)
937 u64 meta_type = BTRFS_BLOCK_GROUP_METADATA;
938 u64 data_type = BTRFS_BLOCK_GROUP_DATA;
939 u64 minimum_data_chunk_size = 8 * 1024 * 1024;
943 for (i = 0; i < num_of_meta_chunks; i++) {
944 ret = btrfs_alloc_chunk(trans, root->fs_info->extent_root,
945 &chunk_start, &chunk_size, meta_type);
947 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
948 meta_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
949 chunk_start, chunk_size);
950 allocation->metadata += chunk_size;
952 set_extent_dirty(&root->fs_info->free_space_cache,
953 chunk_start, chunk_start + chunk_size - 1, 0);
956 if (size_of_data < minimum_data_chunk_size)
957 size_of_data = minimum_data_chunk_size;
959 ret = btrfs_alloc_data_chunk(trans, root->fs_info->extent_root,
960 &chunk_start, size_of_data, data_type);
962 ret = btrfs_make_block_group(trans, root->fs_info->extent_root, 0,
963 data_type, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
964 chunk_start, size_of_data);
965 allocation->data += size_of_data;
967 set_extent_dirty(&root->fs_info->free_space_cache,
968 chunk_start, chunk_start + size_of_data - 1, 0);
972 static int make_image(char *source_dir, struct btrfs_root *root, int out_fd)
975 struct btrfs_trans_handle *trans;
979 struct directory_name_entry dir_head;
981 struct directory_name_entry *dir_entry = NULL;
983 ret = lstat(source_dir, &root_st);
985 fprintf(stderr, "unable to lstat the %s\n", source_dir);
989 INIT_LIST_HEAD(&dir_head.list);
991 trans = btrfs_start_transaction(root, 1);
992 ret = traverse_directory(trans, root, source_dir, &dir_head, out_fd);
994 fprintf(stderr, "unable to traverse_directory\n");
997 btrfs_commit_transaction(trans, root);
1000 printf("Making image is completed.\n");
1003 while (!list_empty(&dir_head.list)) {
1004 dir_entry = list_entry(dir_head.list.next,
1005 struct directory_name_entry, list);
1006 list_del(&dir_entry->list);
1010 fprintf(stderr, "Making image is aborted.\n");
1015 * This ignores symlinks with unreadable targets and subdirs that can't
1016 * be read. It's a best-effort to give a rough estimate of the size of
1017 * a subdir. It doesn't guarantee that prepopulating btrfs from this
1018 * tree won't still run out of space.
1020 * The rounding up to 4096 is questionable. Previous code used du -B 4096.
1022 static u64 global_total_size;
1023 static int ftw_add_entry_size(const char *fpath, const struct stat *st,
1026 if (type == FTW_F || type == FTW_D)
1027 global_total_size += round_up(st->st_size, 4096);
1032 static u64 size_sourcedir(char *dir_name, u64 sectorsize,
1033 u64 *num_of_meta_chunks_ret, u64 *size_of_data_ret)
1038 u64 default_chunk_size = 8 * 1024 * 1024; /* 8MB */
1039 u64 allocated_meta_size = 8 * 1024 * 1024; /* 8MB */
1040 u64 allocated_total_size = 20 * 1024 * 1024; /* 20MB */
1041 u64 num_of_meta_chunks = 0;
1042 u64 num_of_data_chunks = 0;
1043 u64 num_of_allocated_meta_chunks =
1044 allocated_meta_size / default_chunk_size;
1046 global_total_size = 0;
1047 ret = ftw(dir_name, ftw_add_entry_size, 10);
1048 dir_size = global_total_size;
1050 fprintf(stderr, "ftw subdir walk of '%s' failed: %s\n",
1051 dir_name, strerror(errno));
1055 num_of_data_chunks = (dir_size + default_chunk_size - 1) /
1058 num_of_meta_chunks = (dir_size / 2) / default_chunk_size;
1059 if (((dir_size / 2) % default_chunk_size) != 0)
1060 num_of_meta_chunks++;
1061 if (num_of_meta_chunks <= num_of_allocated_meta_chunks)
1062 num_of_meta_chunks = 0;
1064 num_of_meta_chunks -= num_of_allocated_meta_chunks;
1066 total_size = allocated_total_size +
1067 (num_of_data_chunks * default_chunk_size) +
1068 (num_of_meta_chunks * default_chunk_size);
1070 *num_of_meta_chunks_ret = num_of_meta_chunks;
1071 *size_of_data_ret = num_of_data_chunks * default_chunk_size;
1075 static int zero_output_file(int out_fd, u64 size, u32 sectorsize)
1077 int len = sectorsize;
1078 int loop_num = size / sectorsize;
1080 char *buf = malloc(len);
1086 memset(buf, 0, len);
1087 for (i = 0; i < loop_num; i++) {
1088 written = pwrite64(out_fd, buf, len, location);
1091 location += sectorsize;
1097 static int is_ssd(const char *file)
1101 char sysfs_path[PATH_MAX];
1107 probe = blkid_new_probe_from_filename(file);
1111 /* Device number of this disk (possibly a partition) */
1112 devno = blkid_probe_get_devno(probe);
1114 blkid_free_probe(probe);
1118 /* Get whole disk name (not full path) for this devno */
1119 ret = blkid_devno_to_wholedisk(devno,
1120 wholedisk, sizeof(wholedisk), NULL);
1122 blkid_free_probe(probe);
1126 snprintf(sysfs_path, PATH_MAX, "/sys/block/%s/queue/rotational",
1129 blkid_free_probe(probe);
1131 fd = open(sysfs_path, O_RDONLY);
1136 if (read(fd, &rotational, sizeof(char)) < sizeof(char)) {
1142 return !atoi((const char *)&rotational);
1145 static void list_all_devices(struct btrfs_root *root)
1147 struct btrfs_fs_devices *fs_devices;
1148 struct btrfs_device *device;
1149 int number_of_devices = 0;
1150 u64 total_block_count = 0;
1152 fs_devices = root->fs_info->fs_devices;
1154 list_for_each_entry(device, &fs_devices->devices, dev_list)
1155 number_of_devices++;
1157 printf("Number of devices: %d\n", number_of_devices);
1158 /* printf("Total devices size: %10s\n", */
1159 /* pretty_size(total_block_count)); */
1160 printf("Devices:\n");
1161 printf(" ID SIZE PATH\n");
1162 list_for_each_entry_reverse(device, &fs_devices->devices, dev_list) {
1163 char dev_uuid[BTRFS_UUID_UNPARSED_SIZE];
1165 uuid_unparse(device->uuid, dev_uuid);
1166 printf(" %3llu %10s %s\n",
1168 pretty_size(device->total_bytes),
1170 total_block_count += device->total_bytes;
1176 int main(int ac, char **av)
1179 struct btrfs_root *root;
1180 struct btrfs_trans_handle *trans;
1182 u64 block_count = 0;
1183 u64 dev_block_count = 0;
1185 u64 alloc_start = 0;
1186 u64 metadata_profile = 0;
1187 u64 data_profile = 0;
1188 u32 nodesize = max_t(u32, sysconf(_SC_PAGESIZE),
1189 BTRFS_MKFS_DEFAULT_NODE_SIZE);
1190 u32 sectorsize = 4096;
1191 u32 stripesize = 4096;
1197 int nodesize_forced = 0;
1198 int data_profile_opt = 0;
1199 int metadata_profile_opt = 0;
1202 int force_overwrite = 0;
1203 char *source_dir = NULL;
1204 int source_dir_set = 0;
1205 u64 num_of_meta_chunks = 0;
1206 u64 size_of_data = 0;
1207 u64 source_dir_size = 0;
1210 char fs_uuid[BTRFS_UUID_UNPARSED_SIZE] = { 0 };
1211 u64 features = BTRFS_MKFS_DEFAULT_FEATURES;
1212 struct mkfs_allocation allocation = { 0 };
1213 struct btrfs_mkfs_config mkfs_cfg;
1217 static const struct option long_options[] = {
1218 { "alloc-start", required_argument, NULL, 'A'},
1219 { "byte-count", required_argument, NULL, 'b' },
1220 { "force", no_argument, NULL, 'f' },
1221 { "leafsize", required_argument, NULL, 'l' },
1222 { "label", required_argument, NULL, 'L'},
1223 { "metadata", required_argument, NULL, 'm' },
1224 { "mixed", no_argument, NULL, 'M' },
1225 { "nodesize", required_argument, NULL, 'n' },
1226 { "sectorsize", required_argument, NULL, 's' },
1227 { "data", required_argument, NULL, 'd' },
1228 { "version", no_argument, NULL, 'V' },
1229 { "rootdir", required_argument, NULL, 'r' },
1230 { "nodiscard", no_argument, NULL, 'K' },
1231 { "features", required_argument, NULL, 'O' },
1232 { "uuid", required_argument, NULL, 'U' },
1233 { "quiet", 0, NULL, 'q' },
1234 { "help", no_argument, NULL, GETOPT_VAL_HELP },
1238 c = getopt_long(ac, av, "A:b:fl:n:s:m:d:L:O:r:U:VMKq",
1239 long_options, NULL);
1244 alloc_start = parse_size(optarg);
1247 force_overwrite = 1;
1250 data_profile = parse_profile(optarg);
1251 data_profile_opt = 1;
1255 "WARNING: --leafsize is deprecated, use --nodesize\n");
1257 nodesize = parse_size(optarg);
1258 nodesize_forced = 1;
1261 label = parse_label(optarg);
1264 metadata_profile = parse_profile(optarg);
1265 metadata_profile_opt = 1;
1271 char *orig = strdup(optarg);
1274 tmp = btrfs_parse_fs_features(tmp, &features);
1277 "Unrecognized filesystem feature '%s'\n",
1283 if (features & BTRFS_FEATURE_LIST_ALL) {
1284 btrfs_list_all_fs_features(0);
1290 sectorsize = parse_size(optarg);
1293 block_count = parse_size(optarg);
1294 if (block_count <= BTRFS_MKFS_SMALL_VOLUME_SIZE)
1302 source_dir = optarg;
1306 strncpy(fs_uuid, optarg,
1307 BTRFS_UUID_UNPARSED_SIZE - 1);
1315 case GETOPT_VAL_HELP:
1317 print_usage(c != GETOPT_VAL_HELP);
1320 sectorsize = max(sectorsize, (u32)sysconf(_SC_PAGESIZE));
1321 if (btrfs_check_nodesize(nodesize, sectorsize))
1323 saved_optind = optind;
1324 dev_cnt = ac - optind;
1328 if (source_dir_set && dev_cnt > 1) {
1330 "The -r option is limited to a single device\n");
1337 if (uuid_parse(fs_uuid, dummy_uuid) != 0) {
1338 fprintf(stderr, "could not parse UUID: %s\n", fs_uuid);
1341 if (!test_uuid_unique(fs_uuid)) {
1342 fprintf(stderr, "non-unique UUID: %s\n", fs_uuid);
1347 while (dev_cnt-- > 0) {
1348 file = av[optind++];
1349 if (is_block_device(file))
1350 if (test_dev_for_mkfs(file, force_overwrite))
1354 optind = saved_optind;
1355 dev_cnt = ac - optind;
1357 file = av[optind++];
1360 if (is_vol_small(file) || mixed) {
1362 printf("SMALL VOLUME: forcing mixed metadata/data groups\n");
1367 * Set default profiles according to number of added devices.
1368 * For mixed groups defaults are single/single.
1371 if (!metadata_profile_opt) {
1372 if (dev_cnt == 1 && ssd && verbose)
1373 printf("Detected a SSD, turning off metadata "
1374 "duplication. Mkfs with -m dup if you want to "
1375 "force metadata duplication.\n");
1377 metadata_profile = (dev_cnt > 1) ?
1378 BTRFS_BLOCK_GROUP_RAID1 : (ssd) ?
1379 0: BTRFS_BLOCK_GROUP_DUP;
1381 if (!data_profile_opt) {
1382 data_profile = (dev_cnt > 1) ?
1383 BTRFS_BLOCK_GROUP_RAID0 : 0; /* raid0 or single */
1386 u32 best_nodesize = max_t(u32, sysconf(_SC_PAGESIZE), sectorsize);
1388 if (metadata_profile_opt || data_profile_opt) {
1389 if (metadata_profile != data_profile) {
1391 "ERROR: With mixed block groups data and metadata profiles must be the same\n");
1396 if (!nodesize_forced) {
1397 nodesize = best_nodesize;
1398 if (btrfs_check_nodesize(nodesize, sectorsize))
1401 if (nodesize != sectorsize) {
1402 fprintf(stderr, "Error: mixed metadata/data block groups "
1403 "require metadata blocksizes equal to the sectorsize\n");
1408 /* Check device/block_count after the nodesize is determined */
1409 if (block_count && block_count < btrfs_min_dev_size(nodesize)) {
1411 "Size '%llu' is too small to make a usable filesystem\n",
1414 "Minimum size for btrfs filesystem is %llu\n",
1415 btrfs_min_dev_size(nodesize));
1418 for (i = saved_optind; i < saved_optind + dev_cnt; i++) {
1422 ret = test_minimum_size(path, nodesize);
1424 fprintf(stderr, "Failed to check size for '%s': %s\n",
1425 path, strerror(-ret));
1430 "'%s' is too small to make a usable filesystem\n",
1433 "Minimum size for each btrfs device is %llu.\n",
1434 btrfs_min_dev_size(nodesize));
1438 ret = test_num_disk_vs_raid(metadata_profile, data_profile,
1443 /* if we are here that means all devs are good to btrfsify */
1445 printf("%s\n", PACKAGE_STRING);
1446 printf("See %s for more information.\n\n", PACKAGE_URL);
1451 if (!source_dir_set) {
1453 * open without O_EXCL so that the problem should not
1454 * occur by the following processing.
1455 * (btrfs_register_one_device() fails if O_EXCL is on)
1457 fd = open(file, O_RDWR);
1459 fprintf(stderr, "unable to open %s: %s\n", file,
1463 ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
1464 block_count, &mixed, discard);
1469 if (block_count && block_count > dev_block_count) {
1470 fprintf(stderr, "%s is smaller than requested size\n", file);
1474 fd = open_target(file);
1476 fprintf(stderr, "unable to open the %s\n", file);
1480 source_dir_size = size_sourcedir(source_dir, sectorsize,
1481 &num_of_meta_chunks, &size_of_data);
1482 if(block_count < source_dir_size)
1483 block_count = source_dir_size;
1484 ret = zero_output_file(fd, block_count, sectorsize);
1486 fprintf(stderr, "unable to zero the output file\n");
1489 /* our "device" is the new image file */
1490 dev_block_count = block_count;
1493 /* To create the first block group and chunk 0 in make_btrfs */
1494 if (dev_block_count < BTRFS_MKFS_SYSTEM_GROUP_SIZE) {
1495 fprintf(stderr, "device is too small to make filesystem\n");
1499 blocks[0] = BTRFS_SUPER_INFO_OFFSET;
1500 for (i = 1; i < 7; i++) {
1501 blocks[i] = BTRFS_SUPER_INFO_OFFSET + 1024 * 1024 +
1505 if (group_profile_max_safe_loss(metadata_profile) <
1506 group_profile_max_safe_loss(data_profile)){
1508 "WARNING: metatdata has lower redundancy than data!\n\n");
1512 * FS features that can be set by other means than -O
1513 * just set the bit here
1516 features |= BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS;
1518 if ((data_profile | metadata_profile) &
1519 (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6)) {
1520 features |= BTRFS_FEATURE_INCOMPAT_RAID56;
1523 mkfs_cfg.label = label;
1524 mkfs_cfg.fs_uuid = fs_uuid;
1525 memcpy(mkfs_cfg.blocks, blocks, sizeof(blocks));
1526 mkfs_cfg.num_bytes = dev_block_count;
1527 mkfs_cfg.nodesize = nodesize;
1528 mkfs_cfg.sectorsize = sectorsize;
1529 mkfs_cfg.stripesize = stripesize;
1530 mkfs_cfg.features = features;
1532 ret = make_btrfs(fd, &mkfs_cfg);
1534 fprintf(stderr, "error during mkfs: %s\n", strerror(-ret));
1538 root = open_ctree(file, 0, OPEN_CTREE_WRITES);
1540 fprintf(stderr, "Open ctree failed\n");
1544 root->fs_info->alloc_start = alloc_start;
1546 ret = create_metadata_block_groups(root, mixed, &allocation);
1548 fprintf(stderr, "failed to create default block groups\n");
1552 ret = make_root_dir(root, mixed, &allocation);
1554 fprintf(stderr, "failed to setup the root directory\n");
1558 trans = btrfs_start_transaction(root, 1);
1560 if (is_block_device(file))
1561 btrfs_register_one_device(file);
1566 while (dev_cnt-- > 0) {
1567 int old_mixed = mixed;
1569 file = av[optind++];
1572 * open without O_EXCL so that the problem should not
1573 * occur by the following processing.
1574 * (btrfs_register_one_device() fails if O_EXCL is on)
1576 fd = open(file, O_RDWR);
1578 fprintf(stderr, "unable to open %s: %s\n", file,
1582 ret = btrfs_device_already_in_root(root, fd,
1583 BTRFS_SUPER_INFO_OFFSET);
1585 fprintf(stderr, "skipping duplicate device %s in FS\n",
1590 ret = btrfs_prepare_device(fd, file, zero_end, &dev_block_count,
1591 block_count, &mixed, discard);
1598 ret = btrfs_add_to_fsid(trans, root, fd, file, dev_block_count,
1599 sectorsize, sectorsize, sectorsize);
1602 struct btrfs_device *device;
1604 device = container_of(root->fs_info->fs_devices->devices.next,
1605 struct btrfs_device, dev_list);
1606 printf("adding device %s id %llu\n", file,
1607 (unsigned long long)device->devid);
1610 if (is_block_device(file))
1611 btrfs_register_one_device(file);
1615 if (!source_dir_set) {
1616 ret = create_raid_groups(trans, root, data_profile,
1617 data_profile_opt, metadata_profile,
1618 mixed, &allocation);
1622 ret = create_data_reloc_tree(trans, root);
1625 btrfs_commit_transaction(trans, root);
1627 if (source_dir_set) {
1628 trans = btrfs_start_transaction(root, 1);
1629 ret = create_chunks(trans, root,
1630 num_of_meta_chunks, size_of_data,
1633 btrfs_commit_transaction(trans, root);
1635 ret = make_image(source_dir, root, fd);
1640 char features_buf[64];
1642 printf("Label: %s\n", label);
1643 printf("UUID: %s\n", fs_uuid);
1644 printf("Node size: %u\n", nodesize);
1645 printf("Sector size: %u\n", sectorsize);
1646 printf("Filesystem size: %s\n",
1647 pretty_size(btrfs_super_total_bytes(root->fs_info->super_copy)));
1648 printf("Block group profiles:\n");
1649 if (allocation.data)
1650 printf(" Data: %-8s %16s\n",
1651 btrfs_group_profile_str(data_profile),
1652 pretty_size(allocation.data));
1653 if (allocation.metadata)
1654 printf(" Metadata: %-8s %16s\n",
1655 btrfs_group_profile_str(metadata_profile),
1656 pretty_size(allocation.metadata));
1657 if (allocation.mixed)
1658 printf(" Data+Metadata: %-8s %16s\n",
1659 btrfs_group_profile_str(data_profile),
1660 pretty_size(allocation.mixed));
1661 printf(" System: %-8s %16s\n",
1662 btrfs_group_profile_str(metadata_profile),
1663 pretty_size(allocation.system));
1664 printf("SSD detected: %s\n", ssd ? "yes" : "no");
1665 btrfs_parse_features_to_string(features_buf, features);
1666 printf("Incompat features: %s", features_buf);
1669 list_all_devices(root);
1672 ret = close_ctree(root);