2 * Copyright (C) 2007 Oracle. All rights reserved.
3 * Copyright (C) 2008 Morey Roof. All rights reserved.
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public
7 * License v2 as published by the Free Software Foundation.
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * General Public License for more details.
14 * You should have received a copy of the GNU General Public
15 * License along with this program; if not, write to the
16 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
17 * Boston, MA 021110-1307, USA.
23 #include <sys/ioctl.h>
24 #include <sys/mount.h>
25 #include <sys/types.h>
27 #include <uuid/uuid.h>
32 #include <linux/loop.h>
33 #include <linux/major.h>
34 #include <linux/kdev_t.h>
36 #include <blkid/blkid.h>
38 #include <sys/statfs.h>
39 #include <linux/magic.h>
42 #include "kerncompat.h"
43 #include "radix-tree.h"
46 #include "transaction.h"
54 #define BLKDISCARD _IO(0x12,119)
57 static int btrfs_scan_done
= 0;
59 static char argv0_buf
[ARGV0_BUF_SIZE
] = "btrfs";
61 static int rand_seed_initlized
= 0;
62 static unsigned short rand_seed
[3];
64 const char *get_argv0_buf(void)
69 void fixup_argv0(char **argv
, const char *token
)
71 int len
= strlen(argv0_buf
);
73 snprintf(argv0_buf
+ len
, sizeof(argv0_buf
) - len
, " %s", token
);
77 void set_argv0(char **argv
)
79 strncpy(argv0_buf
, argv
[0], sizeof(argv0_buf
));
80 argv0_buf
[sizeof(argv0_buf
) - 1] = 0;
83 int check_argc_exact(int nargs
, int expected
)
86 fprintf(stderr
, "%s: too few arguments\n", argv0_buf
);
88 fprintf(stderr
, "%s: too many arguments\n", argv0_buf
);
90 return nargs
!= expected
;
93 int check_argc_min(int nargs
, int expected
)
95 if (nargs
< expected
) {
96 fprintf(stderr
, "%s: too few arguments\n", argv0_buf
);
103 int check_argc_max(int nargs
, int expected
)
105 if (nargs
> expected
) {
106 fprintf(stderr
, "%s: too many arguments\n", argv0_buf
);
115 * Discard the given range in one go
117 static int discard_range(int fd
, u64 start
, u64 len
)
119 u64 range
[2] = { start
, len
};
121 if (ioctl(fd
, BLKDISCARD
, &range
) < 0)
127 * Discard blocks in the given range in 1G chunks, the process is interruptible
129 static int discard_blocks(int fd
, u64 start
, u64 len
)
133 u64 chunk_size
= min_t(u64
, len
, 1*1024*1024*1024);
136 ret
= discard_range(fd
, start
, chunk_size
);
146 static u64 reference_root_table
[] = {
147 [1] = BTRFS_ROOT_TREE_OBJECTID
,
148 [2] = BTRFS_EXTENT_TREE_OBJECTID
,
149 [3] = BTRFS_CHUNK_TREE_OBJECTID
,
150 [4] = BTRFS_DEV_TREE_OBJECTID
,
151 [5] = BTRFS_FS_TREE_OBJECTID
,
152 [6] = BTRFS_CSUM_TREE_OBJECTID
,
155 int test_uuid_unique(char *fs_uuid
)
158 blkid_dev_iterate iter
= NULL
;
159 blkid_dev dev
= NULL
;
160 blkid_cache cache
= NULL
;
162 if (blkid_get_cache(&cache
, NULL
) < 0) {
163 printf("ERROR: lblkid cache get failed\n");
166 blkid_probe_all(cache
);
167 iter
= blkid_dev_iterate_begin(cache
);
168 blkid_dev_set_search(iter
, "UUID", fs_uuid
);
170 while (blkid_dev_next(iter
, &dev
) == 0) {
171 dev
= blkid_verify(cache
, dev
);
178 blkid_dev_iterate_end(iter
);
179 blkid_put_cache(cache
);
185 * Reserve space from free_tree.
186 * The algorithm is very simple, find the first cache_extent with enough space
187 * and allocate from its beginning.
189 static int reserve_free_space(struct cache_tree
*free_tree
, u64 len
,
192 struct cache_extent
*cache
;
196 cache
= first_cache_extent(free_tree
);
198 if (cache
->size
> len
) {
200 *ret_start
= cache
->start
;
203 if (cache
->size
== 0) {
204 remove_cache_extent(free_tree
, cache
);
211 cache
= next_cache_extent(cache
);
218 static inline int write_temp_super(int fd
, struct btrfs_super_block
*sb
,
224 crc
= btrfs_csum_data(NULL
, (char *)sb
+ BTRFS_CSUM_SIZE
, crc
,
225 BTRFS_SUPER_INFO_SIZE
- BTRFS_CSUM_SIZE
);
226 btrfs_csum_final(crc
, (char *)&sb
->csum
[0]);
227 ret
= pwrite(fd
, sb
, BTRFS_SUPER_INFO_SIZE
, sb_bytenr
);
228 if (ret
< BTRFS_SUPER_INFO_SIZE
)
229 ret
= (ret
< 0 ? -errno
: -EIO
);
236 * Setup temporary superblock at cfg->super_bynter
237 * Needed info are extracted from cfg, and root_bytenr, chunk_bytenr
239 * For now sys chunk array will be empty and dev_item is empty too.
240 * They will be re-initialized at temp chunk tree setup.
242 static int setup_temp_super(int fd
, struct btrfs_mkfs_config
*cfg
,
243 u64 root_bytenr
, u64 chunk_bytenr
)
245 unsigned char chunk_uuid
[BTRFS_UUID_SIZE
];
246 char super_buf
[BTRFS_SUPER_INFO_SIZE
];
247 struct btrfs_super_block
*super
= (struct btrfs_super_block
*)super_buf
;
251 * We rely on cfg->chunk_uuid and cfg->fs_uuid to pass uuid
252 * for other functions.
253 * Caller must allocate space for them
255 BUG_ON(!cfg
->chunk_uuid
|| !cfg
->fs_uuid
);
256 memset(super_buf
, 0, BTRFS_SUPER_INFO_SIZE
);
257 cfg
->num_bytes
= round_down(cfg
->num_bytes
, cfg
->sectorsize
);
259 if (cfg
->fs_uuid
&& *cfg
->fs_uuid
) {
260 if (uuid_parse(cfg
->fs_uuid
, super
->fsid
) != 0) {
261 error("cound not parse UUID: %s", cfg
->fs_uuid
);
265 if (!test_uuid_unique(cfg
->fs_uuid
)) {
266 error("non-unique UUID: %s", cfg
->fs_uuid
);
271 uuid_generate(super
->fsid
);
272 uuid_unparse(super
->fsid
, cfg
->fs_uuid
);
274 uuid_generate(chunk_uuid
);
275 uuid_unparse(chunk_uuid
, cfg
->chunk_uuid
);
277 btrfs_set_super_bytenr(super
, cfg
->super_bytenr
);
278 btrfs_set_super_num_devices(super
, 1);
279 btrfs_set_super_magic(super
, BTRFS_MAGIC
);
280 btrfs_set_super_generation(super
, 1);
281 btrfs_set_super_root(super
, root_bytenr
);
282 btrfs_set_super_chunk_root(super
, chunk_bytenr
);
283 btrfs_set_super_total_bytes(super
, cfg
->num_bytes
);
285 * Temporary filesystem will only have 6 tree roots:
286 * chunk tree, root tree, extent_tree, device tree, fs tree
289 btrfs_set_super_bytes_used(super
, 6 * cfg
->nodesize
);
290 btrfs_set_super_sectorsize(super
, cfg
->sectorsize
);
291 btrfs_set_super_leafsize(super
, cfg
->nodesize
);
292 btrfs_set_super_nodesize(super
, cfg
->nodesize
);
293 btrfs_set_super_stripesize(super
, cfg
->stripesize
);
294 btrfs_set_super_csum_type(super
, BTRFS_CSUM_TYPE_CRC32
);
295 btrfs_set_super_chunk_root(super
, chunk_bytenr
);
296 btrfs_set_super_cache_generation(super
, -1);
297 btrfs_set_super_incompat_flags(super
, cfg
->features
);
299 __strncpy_null(super
->label
, cfg
->label
, BTRFS_LABEL_SIZE
- 1);
301 /* Sys chunk array will be re-initialized at chunk tree init time */
302 super
->sys_chunk_array_size
= 0;
304 ret
= write_temp_super(fd
, super
, cfg
->super_bytenr
);
310 * Setup an extent buffer for tree block.
312 static int setup_temp_extent_buffer(struct extent_buffer
*buf
,
313 struct btrfs_mkfs_config
*cfg
,
314 u64 bytenr
, u64 owner
)
316 unsigned char fsid
[BTRFS_FSID_SIZE
];
317 unsigned char chunk_uuid
[BTRFS_UUID_SIZE
];
320 /* We rely on cfg->fs_uuid and chunk_uuid to fsid and chunk uuid */
321 BUG_ON(!cfg
->fs_uuid
|| !cfg
->chunk_uuid
);
322 ret
= uuid_parse(cfg
->fs_uuid
, fsid
);
325 ret
= uuid_parse(cfg
->chunk_uuid
, chunk_uuid
);
329 memset(buf
->data
, 0, cfg
->nodesize
);
330 buf
->len
= cfg
->nodesize
;
331 btrfs_set_header_bytenr(buf
, bytenr
);
332 btrfs_set_header_generation(buf
, 1);
333 btrfs_set_header_backref_rev(buf
, BTRFS_MIXED_BACKREF_REV
);
334 btrfs_set_header_owner(buf
, owner
);
335 btrfs_set_header_flags(buf
, BTRFS_HEADER_FLAG_WRITTEN
);
336 write_extent_buffer(buf
, chunk_uuid
, btrfs_header_chunk_tree_uuid(buf
),
338 write_extent_buffer(buf
, fsid
, btrfs_header_fsid(), BTRFS_FSID_SIZE
);
342 static inline int write_temp_extent_buffer(int fd
, struct extent_buffer
*buf
,
347 csum_tree_block_size(buf
, BTRFS_CRC32_SIZE
, 0);
349 /* Temporary extent buffer is always mapped 1:1 on disk */
350 ret
= pwrite(fd
, buf
->data
, buf
->len
, bytenr
);
352 ret
= (ret
< 0 ? ret
: -EIO
);
359 * Insert a root item for temporary tree root
361 * Only used in make_btrfs_v2().
363 static void insert_temp_root_item(struct extent_buffer
*buf
,
364 struct btrfs_mkfs_config
*cfg
,
365 int *slot
, u32
*itemoff
, u64 objectid
,
368 struct btrfs_root_item root_item
;
369 struct btrfs_inode_item
*inode_item
;
370 struct btrfs_disk_key disk_key
;
372 btrfs_set_header_nritems(buf
, *slot
+ 1);
373 (*itemoff
) -= sizeof(root_item
);
374 memset(&root_item
, 0, sizeof(root_item
));
375 inode_item
= &root_item
.inode
;
376 btrfs_set_stack_inode_generation(inode_item
, 1);
377 btrfs_set_stack_inode_size(inode_item
, 3);
378 btrfs_set_stack_inode_nlink(inode_item
, 1);
379 btrfs_set_stack_inode_nbytes(inode_item
, cfg
->nodesize
);
380 btrfs_set_stack_inode_mode(inode_item
, S_IFDIR
| 0755);
381 btrfs_set_root_refs(&root_item
, 1);
382 btrfs_set_root_used(&root_item
, cfg
->nodesize
);
383 btrfs_set_root_generation(&root_item
, 1);
384 btrfs_set_root_bytenr(&root_item
, bytenr
);
386 memset(&disk_key
, 0, sizeof(disk_key
));
387 btrfs_set_disk_key_type(&disk_key
, BTRFS_ROOT_ITEM_KEY
);
388 btrfs_set_disk_key_objectid(&disk_key
, objectid
);
389 btrfs_set_disk_key_offset(&disk_key
, 0);
391 btrfs_set_item_key(buf
, &disk_key
, *slot
);
392 btrfs_set_item_offset(buf
, btrfs_item_nr(*slot
), *itemoff
);
393 btrfs_set_item_size(buf
, btrfs_item_nr(*slot
), sizeof(root_item
));
394 write_extent_buffer(buf
, &root_item
,
395 btrfs_item_ptr_offset(buf
, *slot
),
400 static int setup_temp_root_tree(int fd
, struct btrfs_mkfs_config
*cfg
,
401 u64 root_bytenr
, u64 extent_bytenr
,
402 u64 dev_bytenr
, u64 fs_bytenr
, u64 csum_bytenr
)
404 struct extent_buffer
*buf
= NULL
;
405 u32 itemoff
= __BTRFS_LEAF_DATA_SIZE(cfg
->nodesize
);
410 * Provided bytenr must in ascending order, or tree root will have a
413 BUG_ON(!(root_bytenr
< extent_bytenr
&& extent_bytenr
< dev_bytenr
&&
414 dev_bytenr
< fs_bytenr
&& fs_bytenr
< csum_bytenr
));
415 buf
= malloc(sizeof(*buf
) + cfg
->nodesize
);
419 ret
= setup_temp_extent_buffer(buf
, cfg
, root_bytenr
,
420 BTRFS_ROOT_TREE_OBJECTID
);
424 insert_temp_root_item(buf
, cfg
, &slot
, &itemoff
,
425 BTRFS_EXTENT_TREE_OBJECTID
, extent_bytenr
);
426 insert_temp_root_item(buf
, cfg
, &slot
, &itemoff
,
427 BTRFS_DEV_TREE_OBJECTID
, dev_bytenr
);
428 insert_temp_root_item(buf
, cfg
, &slot
, &itemoff
,
429 BTRFS_FS_TREE_OBJECTID
, fs_bytenr
);
430 insert_temp_root_item(buf
, cfg
, &slot
, &itemoff
,
431 BTRFS_CSUM_TREE_OBJECTID
, csum_bytenr
);
433 ret
= write_temp_extent_buffer(fd
, buf
, root_bytenr
);
439 static int insert_temp_dev_item(int fd
, struct extent_buffer
*buf
,
440 struct btrfs_mkfs_config
*cfg
,
441 int *slot
, u32
*itemoff
)
443 struct btrfs_disk_key disk_key
;
444 struct btrfs_dev_item
*dev_item
;
445 char super_buf
[BTRFS_SUPER_INFO_SIZE
];
446 unsigned char dev_uuid
[BTRFS_UUID_SIZE
];
447 unsigned char fsid
[BTRFS_FSID_SIZE
];
448 struct btrfs_super_block
*super
= (struct btrfs_super_block
*)super_buf
;
451 ret
= pread(fd
, super_buf
, BTRFS_SUPER_INFO_SIZE
, cfg
->super_bytenr
);
452 if (ret
< BTRFS_SUPER_INFO_SIZE
) {
453 ret
= (ret
< 0 ? -errno
: -EIO
);
457 btrfs_set_header_nritems(buf
, *slot
+ 1);
458 (*itemoff
) -= sizeof(*dev_item
);
459 /* setup device item 1, 0 is for replace case */
460 btrfs_set_disk_key_type(&disk_key
, BTRFS_DEV_ITEM_KEY
);
461 btrfs_set_disk_key_objectid(&disk_key
, BTRFS_DEV_ITEMS_OBJECTID
);
462 btrfs_set_disk_key_offset(&disk_key
, 1);
463 btrfs_set_item_key(buf
, &disk_key
, *slot
);
464 btrfs_set_item_offset(buf
, btrfs_item_nr(*slot
), *itemoff
);
465 btrfs_set_item_size(buf
, btrfs_item_nr(*slot
), sizeof(*dev_item
));
467 dev_item
= btrfs_item_ptr(buf
, *slot
, struct btrfs_dev_item
);
468 /* Generate device uuid */
469 uuid_generate(dev_uuid
);
470 write_extent_buffer(buf
, dev_uuid
,
471 (unsigned long)btrfs_device_uuid(dev_item
),
473 uuid_parse(cfg
->fs_uuid
, fsid
);
474 write_extent_buffer(buf
, fsid
,
475 (unsigned long)btrfs_device_fsid(dev_item
),
477 btrfs_set_device_id(buf
, dev_item
, 1);
478 btrfs_set_device_generation(buf
, dev_item
, 0);
479 btrfs_set_device_total_bytes(buf
, dev_item
, cfg
->num_bytes
);
481 * The number must match the initial SYSTEM and META chunk size
483 btrfs_set_device_bytes_used(buf
, dev_item
,
484 BTRFS_MKFS_SYSTEM_GROUP_SIZE
+
485 BTRFS_CONVERT_META_GROUP_SIZE
);
486 btrfs_set_device_io_align(buf
, dev_item
, cfg
->sectorsize
);
487 btrfs_set_device_io_width(buf
, dev_item
, cfg
->sectorsize
);
488 btrfs_set_device_sector_size(buf
, dev_item
, cfg
->sectorsize
);
489 btrfs_set_device_type(buf
, dev_item
, 0);
491 /* Super dev_item is not complete, copy the complete one to sb */
492 read_extent_buffer(buf
, &super
->dev_item
, (unsigned long)dev_item
,
494 ret
= write_temp_super(fd
, super
, cfg
->super_bytenr
);
500 static int insert_temp_chunk_item(int fd
, struct extent_buffer
*buf
,
501 struct btrfs_mkfs_config
*cfg
,
502 int *slot
, u32
*itemoff
, u64 start
, u64 len
,
505 struct btrfs_chunk
*chunk
;
506 struct btrfs_disk_key disk_key
;
507 char super_buf
[BTRFS_SUPER_INFO_SIZE
];
508 struct btrfs_super_block
*sb
= (struct btrfs_super_block
*)super_buf
;
511 ret
= pread(fd
, super_buf
, BTRFS_SUPER_INFO_SIZE
,
513 if (ret
< BTRFS_SUPER_INFO_SIZE
) {
514 ret
= (ret
< 0 ? ret
: -EIO
);
518 btrfs_set_header_nritems(buf
, *slot
+ 1);
519 (*itemoff
) -= btrfs_chunk_item_size(1);
520 btrfs_set_disk_key_type(&disk_key
, BTRFS_CHUNK_ITEM_KEY
);
521 btrfs_set_disk_key_objectid(&disk_key
, BTRFS_FIRST_CHUNK_TREE_OBJECTID
);
522 btrfs_set_disk_key_offset(&disk_key
, start
);
523 btrfs_set_item_key(buf
, &disk_key
, *slot
);
524 btrfs_set_item_offset(buf
, btrfs_item_nr(*slot
), *itemoff
);
525 btrfs_set_item_size(buf
, btrfs_item_nr(*slot
),
526 btrfs_chunk_item_size(1));
528 chunk
= btrfs_item_ptr(buf
, *slot
, struct btrfs_chunk
);
529 btrfs_set_chunk_length(buf
, chunk
, len
);
530 btrfs_set_chunk_owner(buf
, chunk
, BTRFS_EXTENT_TREE_OBJECTID
);
531 btrfs_set_chunk_stripe_len(buf
, chunk
, 64 * 1024);
532 btrfs_set_chunk_type(buf
, chunk
, type
);
533 btrfs_set_chunk_io_align(buf
, chunk
, cfg
->sectorsize
);
534 btrfs_set_chunk_io_width(buf
, chunk
, cfg
->sectorsize
);
535 btrfs_set_chunk_sector_size(buf
, chunk
, cfg
->sectorsize
);
536 btrfs_set_chunk_num_stripes(buf
, chunk
, 1);
537 /* TODO: Support DUP profile for system chunk */
538 btrfs_set_stripe_devid_nr(buf
, chunk
, 0, 1);
539 /* We are doing 1:1 mapping, so start is its dev offset */
540 btrfs_set_stripe_offset_nr(buf
, chunk
, 0, start
);
541 write_extent_buffer(buf
, &sb
->dev_item
.uuid
,
542 (unsigned long)btrfs_stripe_dev_uuid_nr(chunk
, 0),
547 * If it's system chunk, also copy it to super block.
549 if (type
& BTRFS_BLOCK_GROUP_SYSTEM
) {
552 cur
= (char *)sb
->sys_chunk_array
+ sb
->sys_chunk_array_size
;
553 memcpy(cur
, &disk_key
, sizeof(disk_key
));
554 cur
+= sizeof(disk_key
);
555 read_extent_buffer(buf
, cur
, (unsigned long int)chunk
,
556 btrfs_chunk_item_size(1));
557 sb
->sys_chunk_array_size
+= btrfs_chunk_item_size(1) +
560 ret
= write_temp_super(fd
, sb
, cfg
->super_bytenr
);
565 static int setup_temp_chunk_tree(int fd
, struct btrfs_mkfs_config
*cfg
,
566 u64 sys_chunk_start
, u64 meta_chunk_start
,
569 struct extent_buffer
*buf
= NULL
;
570 u32 itemoff
= __BTRFS_LEAF_DATA_SIZE(cfg
->nodesize
);
574 /* Must ensure SYS chunk starts before META chunk */
575 BUG_ON(meta_chunk_start
< sys_chunk_start
);
576 buf
= malloc(sizeof(*buf
) + cfg
->nodesize
);
579 ret
= setup_temp_extent_buffer(buf
, cfg
, chunk_bytenr
,
580 BTRFS_CHUNK_TREE_OBJECTID
);
584 ret
= insert_temp_dev_item(fd
, buf
, cfg
, &slot
, &itemoff
);
587 ret
= insert_temp_chunk_item(fd
, buf
, cfg
, &slot
, &itemoff
,
589 BTRFS_MKFS_SYSTEM_GROUP_SIZE
,
590 BTRFS_BLOCK_GROUP_SYSTEM
);
593 ret
= insert_temp_chunk_item(fd
, buf
, cfg
, &slot
, &itemoff
,
595 BTRFS_CONVERT_META_GROUP_SIZE
,
596 BTRFS_BLOCK_GROUP_METADATA
);
599 ret
= write_temp_extent_buffer(fd
, buf
, chunk_bytenr
);
607 * Improved version of make_btrfs().
610 * 1) Do chunk allocation to avoid used data
611 * And after this function, extent type matches chunk type
612 * 2) Better structured code
613 * No super long hand written codes to initialized all tree blocks
614 * Split into small blocks and reuse codes.
615 * TODO: Reuse tree operation facilities by introducing new flags
617 static int make_convert_btrfs(int fd
, struct btrfs_mkfs_config
*cfg
,
618 struct btrfs_convert_context
*cctx
)
620 struct cache_tree
*free
= &cctx
->free
;
621 struct cache_tree
*used
= &cctx
->used
;
623 u64 meta_chunk_start
;
624 /* chunk tree bytenr, in system chunk */
626 /* metadata trees bytenr, in metadata chunk */
634 /* Shouldn't happen */
635 BUG_ON(cache_tree_empty(used
));
638 * reserve space for temporary superblock first
639 * Here we allocate a little larger space, to keep later
640 * free space will be STRIPE_LEN aligned
642 ret
= reserve_free_space(free
, BTRFS_STRIPE_LEN
,
648 * Then reserve system chunk space
649 * TODO: Change system group size depending on cctx->total_bytes.
650 * If using current 4M, it can only handle less than one TB for
651 * worst case and then run out of sys space.
653 ret
= reserve_free_space(free
, BTRFS_MKFS_SYSTEM_GROUP_SIZE
,
657 ret
= reserve_free_space(free
, BTRFS_CONVERT_META_GROUP_SIZE
,
663 * Allocated meta/sys chunks will be mapped 1:1 with device offset.
665 * Inside the allocated metadata chunk, the layout will be:
666 * | offset | contents |
667 * -------------------------------------
669 * | +nodesize | extent root |
670 * | +nodesize * 2 | device root |
671 * | +nodesize * 3 | fs tree |
672 * | +nodesize * 4 | csum tree |
673 * -------------------------------------
674 * Inside the allocated system chunk, the layout will be:
675 * | offset | contents |
676 * -------------------------------------
677 * | +0 | chunk root |
678 * -------------------------------------
680 chunk_bytenr
= sys_chunk_start
;
681 root_bytenr
= meta_chunk_start
;
682 extent_bytenr
= meta_chunk_start
+ cfg
->nodesize
;
683 dev_bytenr
= meta_chunk_start
+ cfg
->nodesize
* 2;
684 fs_bytenr
= meta_chunk_start
+ cfg
->nodesize
* 3;
685 csum_bytenr
= meta_chunk_start
+ cfg
->nodesize
* 4;
687 ret
= setup_temp_super(fd
, cfg
, root_bytenr
, chunk_bytenr
);
691 ret
= setup_temp_root_tree(fd
, cfg
, root_bytenr
, extent_bytenr
,
692 dev_bytenr
, fs_bytenr
, csum_bytenr
);
695 ret
= setup_temp_chunk_tree(fd
, cfg
, sys_chunk_start
, meta_chunk_start
,
705 * @fs_uuid - if NULL, generates a UUID, returns back the new filesystem UUID
707 int make_btrfs(int fd
, struct btrfs_mkfs_config
*cfg
,
708 struct btrfs_convert_context
*cctx
)
710 struct btrfs_super_block super
;
711 struct extent_buffer
*buf
;
712 struct btrfs_root_item root_item
;
713 struct btrfs_disk_key disk_key
;
714 struct btrfs_extent_item
*extent_item
;
715 struct btrfs_inode_item
*inode_item
;
716 struct btrfs_chunk
*chunk
;
717 struct btrfs_dev_item
*dev_item
;
718 struct btrfs_dev_extent
*dev_extent
;
719 u8 chunk_tree_uuid
[BTRFS_UUID_SIZE
];
729 int skinny_metadata
= !!(cfg
->features
&
730 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA
);
734 return make_convert_btrfs(fd
, cfg
, cctx
);
735 buf
= malloc(sizeof(*buf
) + max(cfg
->sectorsize
, cfg
->nodesize
));
739 first_free
= BTRFS_SUPER_INFO_OFFSET
+ cfg
->sectorsize
* 2 - 1;
740 first_free
&= ~((u64
)cfg
->sectorsize
- 1);
742 memset(&super
, 0, sizeof(super
));
744 num_bytes
= (cfg
->num_bytes
/ cfg
->sectorsize
) * cfg
->sectorsize
;
745 if (cfg
->fs_uuid
&& *cfg
->fs_uuid
) {
746 if (uuid_parse(cfg
->fs_uuid
, super
.fsid
) != 0) {
747 error("cannot not parse UUID: %s", cfg
->fs_uuid
);
751 if (!test_uuid_unique(cfg
->fs_uuid
)) {
752 error("non-unique UUID: %s", cfg
->fs_uuid
);
757 uuid_generate(super
.fsid
);
759 uuid_unparse(super
.fsid
, cfg
->fs_uuid
);
761 uuid_generate(super
.dev_item
.uuid
);
762 uuid_generate(chunk_tree_uuid
);
764 btrfs_set_super_bytenr(&super
, cfg
->blocks
[0]);
765 btrfs_set_super_num_devices(&super
, 1);
766 btrfs_set_super_magic(&super
, BTRFS_MAGIC
);
767 btrfs_set_super_generation(&super
, 1);
768 btrfs_set_super_root(&super
, cfg
->blocks
[1]);
769 btrfs_set_super_chunk_root(&super
, cfg
->blocks
[3]);
770 btrfs_set_super_total_bytes(&super
, num_bytes
);
771 btrfs_set_super_bytes_used(&super
, 6 * cfg
->nodesize
);
772 btrfs_set_super_sectorsize(&super
, cfg
->sectorsize
);
773 btrfs_set_super_leafsize(&super
, cfg
->nodesize
);
774 btrfs_set_super_nodesize(&super
, cfg
->nodesize
);
775 btrfs_set_super_stripesize(&super
, cfg
->stripesize
);
776 btrfs_set_super_csum_type(&super
, BTRFS_CSUM_TYPE_CRC32
);
777 btrfs_set_super_chunk_root_generation(&super
, 1);
778 btrfs_set_super_cache_generation(&super
, -1);
779 btrfs_set_super_incompat_flags(&super
, cfg
->features
);
781 __strncpy_null(super
.label
, cfg
->label
, BTRFS_LABEL_SIZE
- 1);
783 /* create the tree of root objects */
784 memset(buf
->data
, 0, cfg
->nodesize
);
785 buf
->len
= cfg
->nodesize
;
786 btrfs_set_header_bytenr(buf
, cfg
->blocks
[1]);
787 btrfs_set_header_nritems(buf
, 4);
788 btrfs_set_header_generation(buf
, 1);
789 btrfs_set_header_backref_rev(buf
, BTRFS_MIXED_BACKREF_REV
);
790 btrfs_set_header_owner(buf
, BTRFS_ROOT_TREE_OBJECTID
);
791 write_extent_buffer(buf
, super
.fsid
, btrfs_header_fsid(),
794 write_extent_buffer(buf
, chunk_tree_uuid
,
795 btrfs_header_chunk_tree_uuid(buf
),
798 /* create the items for the root tree */
799 memset(&root_item
, 0, sizeof(root_item
));
800 inode_item
= &root_item
.inode
;
801 btrfs_set_stack_inode_generation(inode_item
, 1);
802 btrfs_set_stack_inode_size(inode_item
, 3);
803 btrfs_set_stack_inode_nlink(inode_item
, 1);
804 btrfs_set_stack_inode_nbytes(inode_item
, cfg
->nodesize
);
805 btrfs_set_stack_inode_mode(inode_item
, S_IFDIR
| 0755);
806 btrfs_set_root_refs(&root_item
, 1);
807 btrfs_set_root_used(&root_item
, cfg
->nodesize
);
808 btrfs_set_root_generation(&root_item
, 1);
810 memset(&disk_key
, 0, sizeof(disk_key
));
811 btrfs_set_disk_key_type(&disk_key
, BTRFS_ROOT_ITEM_KEY
);
812 btrfs_set_disk_key_offset(&disk_key
, 0);
815 itemoff
= __BTRFS_LEAF_DATA_SIZE(cfg
->nodesize
) - sizeof(root_item
);
816 btrfs_set_root_bytenr(&root_item
, cfg
->blocks
[2]);
817 btrfs_set_disk_key_objectid(&disk_key
, BTRFS_EXTENT_TREE_OBJECTID
);
818 btrfs_set_item_key(buf
, &disk_key
, nritems
);
819 btrfs_set_item_offset(buf
, btrfs_item_nr(nritems
), itemoff
);
820 btrfs_set_item_size(buf
, btrfs_item_nr(nritems
),
822 write_extent_buffer(buf
, &root_item
, btrfs_item_ptr_offset(buf
,
823 nritems
), sizeof(root_item
));
826 itemoff
= itemoff
- sizeof(root_item
);
827 btrfs_set_root_bytenr(&root_item
, cfg
->blocks
[4]);
828 btrfs_set_disk_key_objectid(&disk_key
, BTRFS_DEV_TREE_OBJECTID
);
829 btrfs_set_item_key(buf
, &disk_key
, nritems
);
830 btrfs_set_item_offset(buf
, btrfs_item_nr(nritems
), itemoff
);
831 btrfs_set_item_size(buf
, btrfs_item_nr(nritems
),
833 write_extent_buffer(buf
, &root_item
,
834 btrfs_item_ptr_offset(buf
, nritems
),
838 itemoff
= itemoff
- sizeof(root_item
);
839 btrfs_set_root_bytenr(&root_item
, cfg
->blocks
[5]);
840 btrfs_set_disk_key_objectid(&disk_key
, BTRFS_FS_TREE_OBJECTID
);
841 btrfs_set_item_key(buf
, &disk_key
, nritems
);
842 btrfs_set_item_offset(buf
, btrfs_item_nr(nritems
), itemoff
);
843 btrfs_set_item_size(buf
, btrfs_item_nr(nritems
),
845 write_extent_buffer(buf
, &root_item
,
846 btrfs_item_ptr_offset(buf
, nritems
),
850 itemoff
= itemoff
- sizeof(root_item
);
851 btrfs_set_root_bytenr(&root_item
, cfg
->blocks
[6]);
852 btrfs_set_disk_key_objectid(&disk_key
, BTRFS_CSUM_TREE_OBJECTID
);
853 btrfs_set_item_key(buf
, &disk_key
, nritems
);
854 btrfs_set_item_offset(buf
, btrfs_item_nr(nritems
), itemoff
);
855 btrfs_set_item_size(buf
, btrfs_item_nr(nritems
),
857 write_extent_buffer(buf
, &root_item
,
858 btrfs_item_ptr_offset(buf
, nritems
),
863 csum_tree_block_size(buf
, BTRFS_CRC32_SIZE
, 0);
864 ret
= pwrite(fd
, buf
->data
, cfg
->nodesize
, cfg
->blocks
[1]);
865 if (ret
!= cfg
->nodesize
) {
866 ret
= (ret
< 0 ? -errno
: -EIO
);
870 /* create the items for the extent tree */
871 memset(buf
->data
+ sizeof(struct btrfs_header
), 0,
872 cfg
->nodesize
- sizeof(struct btrfs_header
));
874 itemoff
= __BTRFS_LEAF_DATA_SIZE(cfg
->nodesize
);
875 for (i
= 1; i
< 7; i
++) {
876 item_size
= sizeof(struct btrfs_extent_item
);
877 if (!skinny_metadata
)
878 item_size
+= sizeof(struct btrfs_tree_block_info
);
880 BUG_ON(cfg
->blocks
[i
] < first_free
);
881 BUG_ON(cfg
->blocks
[i
] < cfg
->blocks
[i
- 1]);
883 /* create extent item */
884 itemoff
-= item_size
;
885 btrfs_set_disk_key_objectid(&disk_key
, cfg
->blocks
[i
]);
886 if (skinny_metadata
) {
887 btrfs_set_disk_key_type(&disk_key
,
888 BTRFS_METADATA_ITEM_KEY
);
889 btrfs_set_disk_key_offset(&disk_key
, 0);
891 btrfs_set_disk_key_type(&disk_key
,
892 BTRFS_EXTENT_ITEM_KEY
);
893 btrfs_set_disk_key_offset(&disk_key
, cfg
->nodesize
);
895 btrfs_set_item_key(buf
, &disk_key
, nritems
);
896 btrfs_set_item_offset(buf
, btrfs_item_nr(nritems
),
898 btrfs_set_item_size(buf
, btrfs_item_nr(nritems
),
900 extent_item
= btrfs_item_ptr(buf
, nritems
,
901 struct btrfs_extent_item
);
902 btrfs_set_extent_refs(buf
, extent_item
, 1);
903 btrfs_set_extent_generation(buf
, extent_item
, 1);
904 btrfs_set_extent_flags(buf
, extent_item
,
905 BTRFS_EXTENT_FLAG_TREE_BLOCK
);
908 /* create extent ref */
909 ref_root
= reference_root_table
[i
];
910 btrfs_set_disk_key_objectid(&disk_key
, cfg
->blocks
[i
]);
911 btrfs_set_disk_key_offset(&disk_key
, ref_root
);
912 btrfs_set_disk_key_type(&disk_key
, BTRFS_TREE_BLOCK_REF_KEY
);
913 btrfs_set_item_key(buf
, &disk_key
, nritems
);
914 btrfs_set_item_offset(buf
, btrfs_item_nr(nritems
),
916 btrfs_set_item_size(buf
, btrfs_item_nr(nritems
), 0);
919 btrfs_set_header_bytenr(buf
, cfg
->blocks
[2]);
920 btrfs_set_header_owner(buf
, BTRFS_EXTENT_TREE_OBJECTID
);
921 btrfs_set_header_nritems(buf
, nritems
);
922 csum_tree_block_size(buf
, BTRFS_CRC32_SIZE
, 0);
923 ret
= pwrite(fd
, buf
->data
, cfg
->nodesize
, cfg
->blocks
[2]);
924 if (ret
!= cfg
->nodesize
) {
925 ret
= (ret
< 0 ? -errno
: -EIO
);
929 /* create the chunk tree */
930 memset(buf
->data
+ sizeof(struct btrfs_header
), 0,
931 cfg
->nodesize
- sizeof(struct btrfs_header
));
933 item_size
= sizeof(*dev_item
);
934 itemoff
= __BTRFS_LEAF_DATA_SIZE(cfg
->nodesize
) - item_size
;
936 /* first device 1 (there is no device 0) */
937 btrfs_set_disk_key_objectid(&disk_key
, BTRFS_DEV_ITEMS_OBJECTID
);
938 btrfs_set_disk_key_offset(&disk_key
, 1);
939 btrfs_set_disk_key_type(&disk_key
, BTRFS_DEV_ITEM_KEY
);
940 btrfs_set_item_key(buf
, &disk_key
, nritems
);
941 btrfs_set_item_offset(buf
, btrfs_item_nr(nritems
), itemoff
);
942 btrfs_set_item_size(buf
, btrfs_item_nr(nritems
), item_size
);
944 dev_item
= btrfs_item_ptr(buf
, nritems
, struct btrfs_dev_item
);
945 btrfs_set_device_id(buf
, dev_item
, 1);
946 btrfs_set_device_generation(buf
, dev_item
, 0);
947 btrfs_set_device_total_bytes(buf
, dev_item
, num_bytes
);
948 btrfs_set_device_bytes_used(buf
, dev_item
,
949 BTRFS_MKFS_SYSTEM_GROUP_SIZE
);
950 btrfs_set_device_io_align(buf
, dev_item
, cfg
->sectorsize
);
951 btrfs_set_device_io_width(buf
, dev_item
, cfg
->sectorsize
);
952 btrfs_set_device_sector_size(buf
, dev_item
, cfg
->sectorsize
);
953 btrfs_set_device_type(buf
, dev_item
, 0);
955 write_extent_buffer(buf
, super
.dev_item
.uuid
,
956 (unsigned long)btrfs_device_uuid(dev_item
),
958 write_extent_buffer(buf
, super
.fsid
,
959 (unsigned long)btrfs_device_fsid(dev_item
),
961 read_extent_buffer(buf
, &super
.dev_item
, (unsigned long)dev_item
,
965 item_size
= btrfs_chunk_item_size(1);
966 itemoff
= itemoff
- item_size
;
968 /* then we have chunk 0 */
969 btrfs_set_disk_key_objectid(&disk_key
, BTRFS_FIRST_CHUNK_TREE_OBJECTID
);
970 btrfs_set_disk_key_offset(&disk_key
, 0);
971 btrfs_set_disk_key_type(&disk_key
, BTRFS_CHUNK_ITEM_KEY
);
972 btrfs_set_item_key(buf
, &disk_key
, nritems
);
973 btrfs_set_item_offset(buf
, btrfs_item_nr(nritems
), itemoff
);
974 btrfs_set_item_size(buf
, btrfs_item_nr(nritems
), item_size
);
976 chunk
= btrfs_item_ptr(buf
, nritems
, struct btrfs_chunk
);
977 btrfs_set_chunk_length(buf
, chunk
, BTRFS_MKFS_SYSTEM_GROUP_SIZE
);
978 btrfs_set_chunk_owner(buf
, chunk
, BTRFS_EXTENT_TREE_OBJECTID
);
979 btrfs_set_chunk_stripe_len(buf
, chunk
, 64 * 1024);
980 btrfs_set_chunk_type(buf
, chunk
, BTRFS_BLOCK_GROUP_SYSTEM
);
981 btrfs_set_chunk_io_align(buf
, chunk
, cfg
->sectorsize
);
982 btrfs_set_chunk_io_width(buf
, chunk
, cfg
->sectorsize
);
983 btrfs_set_chunk_sector_size(buf
, chunk
, cfg
->sectorsize
);
984 btrfs_set_chunk_num_stripes(buf
, chunk
, 1);
985 btrfs_set_stripe_devid_nr(buf
, chunk
, 0, 1);
986 btrfs_set_stripe_offset_nr(buf
, chunk
, 0, 0);
989 write_extent_buffer(buf
, super
.dev_item
.uuid
,
990 (unsigned long)btrfs_stripe_dev_uuid(&chunk
->stripe
),
993 /* copy the key for the chunk to the system array */
994 ptr
= super
.sys_chunk_array
;
995 array_size
= sizeof(disk_key
);
997 memcpy(ptr
, &disk_key
, sizeof(disk_key
));
998 ptr
+= sizeof(disk_key
);
1000 /* copy the chunk to the system array */
1001 read_extent_buffer(buf
, ptr
, (unsigned long)chunk
, item_size
);
1002 array_size
+= item_size
;
1004 btrfs_set_super_sys_array_size(&super
, array_size
);
1006 btrfs_set_header_bytenr(buf
, cfg
->blocks
[3]);
1007 btrfs_set_header_owner(buf
, BTRFS_CHUNK_TREE_OBJECTID
);
1008 btrfs_set_header_nritems(buf
, nritems
);
1009 csum_tree_block_size(buf
, BTRFS_CRC32_SIZE
, 0);
1010 ret
= pwrite(fd
, buf
->data
, cfg
->nodesize
, cfg
->blocks
[3]);
1011 if (ret
!= cfg
->nodesize
) {
1012 ret
= (ret
< 0 ? -errno
: -EIO
);
1016 /* create the device tree */
1017 memset(buf
->data
+ sizeof(struct btrfs_header
), 0,
1018 cfg
->nodesize
- sizeof(struct btrfs_header
));
1020 itemoff
= __BTRFS_LEAF_DATA_SIZE(cfg
->nodesize
) -
1021 sizeof(struct btrfs_dev_extent
);
1023 btrfs_set_disk_key_objectid(&disk_key
, 1);
1024 btrfs_set_disk_key_offset(&disk_key
, 0);
1025 btrfs_set_disk_key_type(&disk_key
, BTRFS_DEV_EXTENT_KEY
);
1026 btrfs_set_item_key(buf
, &disk_key
, nritems
);
1027 btrfs_set_item_offset(buf
, btrfs_item_nr(nritems
), itemoff
);
1028 btrfs_set_item_size(buf
, btrfs_item_nr(nritems
),
1029 sizeof(struct btrfs_dev_extent
));
1030 dev_extent
= btrfs_item_ptr(buf
, nritems
, struct btrfs_dev_extent
);
1031 btrfs_set_dev_extent_chunk_tree(buf
, dev_extent
,
1032 BTRFS_CHUNK_TREE_OBJECTID
);
1033 btrfs_set_dev_extent_chunk_objectid(buf
, dev_extent
,
1034 BTRFS_FIRST_CHUNK_TREE_OBJECTID
);
1035 btrfs_set_dev_extent_chunk_offset(buf
, dev_extent
, 0);
1037 write_extent_buffer(buf
, chunk_tree_uuid
,
1038 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent
),
1041 btrfs_set_dev_extent_length(buf
, dev_extent
,
1042 BTRFS_MKFS_SYSTEM_GROUP_SIZE
);
1045 btrfs_set_header_bytenr(buf
, cfg
->blocks
[4]);
1046 btrfs_set_header_owner(buf
, BTRFS_DEV_TREE_OBJECTID
);
1047 btrfs_set_header_nritems(buf
, nritems
);
1048 csum_tree_block_size(buf
, BTRFS_CRC32_SIZE
, 0);
1049 ret
= pwrite(fd
, buf
->data
, cfg
->nodesize
, cfg
->blocks
[4]);
1050 if (ret
!= cfg
->nodesize
) {
1051 ret
= (ret
< 0 ? -errno
: -EIO
);
1055 /* create the FS root */
1056 memset(buf
->data
+ sizeof(struct btrfs_header
), 0,
1057 cfg
->nodesize
- sizeof(struct btrfs_header
));
1058 btrfs_set_header_bytenr(buf
, cfg
->blocks
[5]);
1059 btrfs_set_header_owner(buf
, BTRFS_FS_TREE_OBJECTID
);
1060 btrfs_set_header_nritems(buf
, 0);
1061 csum_tree_block_size(buf
, BTRFS_CRC32_SIZE
, 0);
1062 ret
= pwrite(fd
, buf
->data
, cfg
->nodesize
, cfg
->blocks
[5]);
1063 if (ret
!= cfg
->nodesize
) {
1064 ret
= (ret
< 0 ? -errno
: -EIO
);
1067 /* finally create the csum root */
1068 memset(buf
->data
+ sizeof(struct btrfs_header
), 0,
1069 cfg
->nodesize
- sizeof(struct btrfs_header
));
1070 btrfs_set_header_bytenr(buf
, cfg
->blocks
[6]);
1071 btrfs_set_header_owner(buf
, BTRFS_CSUM_TREE_OBJECTID
);
1072 btrfs_set_header_nritems(buf
, 0);
1073 csum_tree_block_size(buf
, BTRFS_CRC32_SIZE
, 0);
1074 ret
= pwrite(fd
, buf
->data
, cfg
->nodesize
, cfg
->blocks
[6]);
1075 if (ret
!= cfg
->nodesize
) {
1076 ret
= (ret
< 0 ? -errno
: -EIO
);
1080 /* and write out the super block */
1081 BUG_ON(sizeof(super
) > cfg
->sectorsize
);
1082 memset(buf
->data
, 0, BTRFS_SUPER_INFO_SIZE
);
1083 memcpy(buf
->data
, &super
, sizeof(super
));
1084 buf
->len
= BTRFS_SUPER_INFO_SIZE
;
1085 csum_tree_block_size(buf
, BTRFS_CRC32_SIZE
, 0);
1086 ret
= pwrite(fd
, buf
->data
, BTRFS_SUPER_INFO_SIZE
, cfg
->blocks
[0]);
1087 if (ret
!= BTRFS_SUPER_INFO_SIZE
) {
1088 ret
= (ret
< 0 ? -errno
: -EIO
);
1099 static const struct btrfs_fs_feature
{
1103 } mkfs_features
[] = {
1104 { "mixed-bg", BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS
,
1105 "mixed data and metadata block groups" },
1106 { "extref", BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF
,
1107 "increased hardlink limit per file to 65536" },
1108 { "raid56", BTRFS_FEATURE_INCOMPAT_RAID56
,
1109 "raid56 extended format" },
1110 { "skinny-metadata", BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA
,
1111 "reduced-size metadata extent refs" },
1112 { "no-holes", BTRFS_FEATURE_INCOMPAT_NO_HOLES
,
1113 "no explicit hole extents for files" },
1114 /* Keep this one last */
1115 { "list-all", BTRFS_FEATURE_LIST_ALL
, NULL
}
1118 static int parse_one_fs_feature(const char *name
, u64
*flags
)
1123 for (i
= 0; i
< ARRAY_SIZE(mkfs_features
); i
++) {
1124 if (name
[0] == '^' &&
1125 !strcmp(mkfs_features
[i
].name
, name
+ 1)) {
1126 *flags
&= ~ mkfs_features
[i
].flag
;
1128 } else if (!strcmp(mkfs_features
[i
].name
, name
)) {
1129 *flags
|= mkfs_features
[i
].flag
;
1137 void btrfs_parse_features_to_string(char *buf
, u64 flags
)
1143 for (i
= 0; i
< ARRAY_SIZE(mkfs_features
); i
++) {
1144 if (flags
& mkfs_features
[i
].flag
) {
1147 strcat(buf
, mkfs_features
[i
].name
);
1152 void btrfs_process_fs_features(u64 flags
)
1156 for (i
= 0; i
< ARRAY_SIZE(mkfs_features
); i
++) {
1157 if (flags
& mkfs_features
[i
].flag
) {
1158 printf("Turning ON incompat feature '%s': %s\n",
1159 mkfs_features
[i
].name
,
1160 mkfs_features
[i
].desc
);
1165 void btrfs_list_all_fs_features(u64 mask_disallowed
)
1169 fprintf(stderr
, "Filesystem features available:\n");
1170 for (i
= 0; i
< ARRAY_SIZE(mkfs_features
) - 1; i
++) {
1171 char *is_default
= "";
1173 if (mkfs_features
[i
].flag
& mask_disallowed
)
1175 if (mkfs_features
[i
].flag
& BTRFS_MKFS_DEFAULT_FEATURES
)
1176 is_default
= ", default";
1177 fprintf(stderr
, "%-20s- %s (0x%llx%s)\n",
1178 mkfs_features
[i
].name
,
1179 mkfs_features
[i
].desc
,
1180 mkfs_features
[i
].flag
,
1186 * Return NULL if all features were parsed fine, otherwise return the name of
1187 * the first unparsed.
1189 char* btrfs_parse_fs_features(char *namelist
, u64
*flags
)
1192 char *save_ptr
= NULL
; /* Satisfy static checkers */
1194 for (this_char
= strtok_r(namelist
, ",", &save_ptr
);
1196 this_char
= strtok_r(NULL
, ",", &save_ptr
)) {
1197 if (parse_one_fs_feature(this_char
, flags
))
1204 u64
btrfs_device_size(int fd
, struct stat
*st
)
1207 if (S_ISREG(st
->st_mode
)) {
1210 if (!S_ISBLK(st
->st_mode
)) {
1213 if (ioctl(fd
, BLKGETSIZE64
, &size
) >= 0) {
1219 static int zero_blocks(int fd
, off_t start
, size_t len
)
1221 char *buf
= malloc(len
);
1227 memset(buf
, 0, len
);
1228 written
= pwrite(fd
, buf
, len
, start
);
1235 #define ZERO_DEV_BYTES (2 * 1024 * 1024)
1237 /* don't write outside the device by clamping the region to the device size */
1238 static int zero_dev_clamped(int fd
, off_t start
, ssize_t len
, u64 dev_size
)
1240 off_t end
= max(start
, start
+ len
);
1243 /* and don't overwrite the disk labels on sparc */
1244 start
= max(start
, 1024);
1245 end
= max(end
, 1024);
1248 start
= min_t(u64
, start
, dev_size
);
1249 end
= min_t(u64
, end
, dev_size
);
1251 return zero_blocks(fd
, start
, end
- start
);
1254 int btrfs_add_to_fsid(struct btrfs_trans_handle
*trans
,
1255 struct btrfs_root
*root
, int fd
, char *path
,
1256 u64 device_total_bytes
, u32 io_width
, u32 io_align
,
1259 struct btrfs_super_block
*disk_super
;
1260 struct btrfs_super_block
*super
= root
->fs_info
->super_copy
;
1261 struct btrfs_device
*device
;
1262 struct btrfs_dev_item
*dev_item
;
1268 device_total_bytes
= (device_total_bytes
/ sectorsize
) * sectorsize
;
1270 device
= kzalloc(sizeof(*device
), GFP_NOFS
);
1273 buf
= kzalloc(sectorsize
, GFP_NOFS
);
1276 BUG_ON(sizeof(*disk_super
) > sectorsize
);
1278 disk_super
= (struct btrfs_super_block
*)buf
;
1279 dev_item
= &disk_super
->dev_item
;
1281 uuid_generate(device
->uuid
);
1284 device
->io_width
= io_width
;
1285 device
->io_align
= io_align
;
1286 device
->sector_size
= sectorsize
;
1288 device
->writeable
= 1;
1289 device
->total_bytes
= device_total_bytes
;
1290 device
->bytes_used
= 0;
1291 device
->total_ios
= 0;
1292 device
->dev_root
= root
->fs_info
->dev_root
;
1293 device
->name
= strdup(path
);
1297 INIT_LIST_HEAD(&device
->dev_list
);
1298 ret
= btrfs_add_device(trans
, root
, device
);
1301 fs_total_bytes
= btrfs_super_total_bytes(super
) + device_total_bytes
;
1302 btrfs_set_super_total_bytes(super
, fs_total_bytes
);
1304 num_devs
= btrfs_super_num_devices(super
) + 1;
1305 btrfs_set_super_num_devices(super
, num_devs
);
1307 memcpy(disk_super
, super
, sizeof(*disk_super
));
1309 btrfs_set_super_bytenr(disk_super
, BTRFS_SUPER_INFO_OFFSET
);
1310 btrfs_set_stack_device_id(dev_item
, device
->devid
);
1311 btrfs_set_stack_device_type(dev_item
, device
->type
);
1312 btrfs_set_stack_device_io_align(dev_item
, device
->io_align
);
1313 btrfs_set_stack_device_io_width(dev_item
, device
->io_width
);
1314 btrfs_set_stack_device_sector_size(dev_item
, device
->sector_size
);
1315 btrfs_set_stack_device_total_bytes(dev_item
, device
->total_bytes
);
1316 btrfs_set_stack_device_bytes_used(dev_item
, device
->bytes_used
);
1317 memcpy(&dev_item
->uuid
, device
->uuid
, BTRFS_UUID_SIZE
);
1319 ret
= pwrite(fd
, buf
, sectorsize
, BTRFS_SUPER_INFO_OFFSET
);
1320 BUG_ON(ret
!= sectorsize
);
1323 list_add(&device
->dev_list
, &root
->fs_info
->fs_devices
->devices
);
1324 device
->fs_devices
= root
->fs_info
->fs_devices
;
1333 static int btrfs_wipe_existing_sb(int fd
)
1335 const char *off
= NULL
;
1340 blkid_probe pr
= NULL
;
1342 pr
= blkid_new_probe();
1346 if (blkid_probe_set_device(pr
, fd
, 0, 0)) {
1351 ret
= blkid_probe_lookup_value(pr
, "SBMAGIC_OFFSET", &off
, NULL
);
1353 ret
= blkid_probe_lookup_value(pr
, "SBMAGIC", NULL
, &len
);
1355 if (ret
|| len
== 0 || off
== NULL
) {
1357 * If lookup fails, the probe did not find any values, eg. for
1358 * a file image or a loop device. Soft error.
1364 offset
= strtoll(off
, NULL
, 10);
1365 if (len
> sizeof(buf
))
1368 memset(buf
, 0, len
);
1369 ret
= pwrite(fd
, buf
, len
, offset
);
1371 error("cannot wipe existing superblock: %s", strerror(errno
));
1373 } else if (ret
!= len
) {
1374 error("cannot wipe existing superblock: wrote %d of %zd", ret
, len
);
1380 blkid_free_probe(pr
);
1384 int btrfs_prepare_device(int fd
, const char *file
, int zero_end
,
1385 u64
*block_count_ret
, u64 max_block_count
, int discard
)
1391 ret
= fstat(fd
, &st
);
1393 error("unable to stat %s: %s", file
, strerror(errno
));
1397 block_count
= btrfs_device_size(fd
, &st
);
1398 if (block_count
== 0) {
1399 error("unable to determine size of %s", file
);
1402 if (max_block_count
)
1403 block_count
= min(block_count
, max_block_count
);
1407 * We intentionally ignore errors from the discard ioctl. It
1408 * is not necessary for the mkfs functionality but just an
1411 if (discard_range(fd
, 0, 0) == 0) {
1412 printf("Performing full device TRIM (%s) ...\n",
1413 pretty_size(block_count
));
1414 discard_blocks(fd
, 0, block_count
);
1418 ret
= zero_dev_clamped(fd
, 0, ZERO_DEV_BYTES
, block_count
);
1419 for (i
= 0 ; !ret
&& i
< BTRFS_SUPER_MIRROR_MAX
; i
++)
1420 ret
= zero_dev_clamped(fd
, btrfs_sb_offset(i
),
1421 BTRFS_SUPER_INFO_SIZE
, block_count
);
1422 if (!ret
&& zero_end
)
1423 ret
= zero_dev_clamped(fd
, block_count
- ZERO_DEV_BYTES
,
1424 ZERO_DEV_BYTES
, block_count
);
1427 error("failed to zero device '%s': %s", file
, strerror(-ret
));
1431 ret
= btrfs_wipe_existing_sb(fd
);
1433 error("cannot wipe superblocks on %s", file
);
1437 *block_count_ret
= block_count
;
1441 int btrfs_make_root_dir(struct btrfs_trans_handle
*trans
,
1442 struct btrfs_root
*root
, u64 objectid
)
1445 struct btrfs_inode_item inode_item
;
1446 time_t now
= time(NULL
);
1448 memset(&inode_item
, 0, sizeof(inode_item
));
1449 btrfs_set_stack_inode_generation(&inode_item
, trans
->transid
);
1450 btrfs_set_stack_inode_size(&inode_item
, 0);
1451 btrfs_set_stack_inode_nlink(&inode_item
, 1);
1452 btrfs_set_stack_inode_nbytes(&inode_item
, root
->nodesize
);
1453 btrfs_set_stack_inode_mode(&inode_item
, S_IFDIR
| 0755);
1454 btrfs_set_stack_timespec_sec(&inode_item
.atime
, now
);
1455 btrfs_set_stack_timespec_nsec(&inode_item
.atime
, 0);
1456 btrfs_set_stack_timespec_sec(&inode_item
.ctime
, now
);
1457 btrfs_set_stack_timespec_nsec(&inode_item
.ctime
, 0);
1458 btrfs_set_stack_timespec_sec(&inode_item
.mtime
, now
);
1459 btrfs_set_stack_timespec_nsec(&inode_item
.mtime
, 0);
1460 btrfs_set_stack_timespec_sec(&inode_item
.otime
, 0);
1461 btrfs_set_stack_timespec_nsec(&inode_item
.otime
, 0);
1463 if (root
->fs_info
->tree_root
== root
)
1464 btrfs_set_super_root_dir(root
->fs_info
->super_copy
, objectid
);
1466 ret
= btrfs_insert_inode(trans
, root
, objectid
, &inode_item
);
1470 ret
= btrfs_insert_inode_ref(trans
, root
, "..", 2, objectid
, objectid
, 0);
1474 btrfs_set_root_dirid(&root
->root_item
, objectid
);
1481 * checks if a path is a block device node
1482 * Returns negative errno on failure, otherwise
1483 * returns 1 for blockdev, 0 for not-blockdev
1485 int is_block_device(const char *path
)
1487 struct stat statbuf
;
1489 if (stat(path
, &statbuf
) < 0)
1492 return !!S_ISBLK(statbuf
.st_mode
);
1496 * check if given path is a mount point
1497 * return 1 if yes. 0 if no. -1 for error
1499 int is_mount_point(const char *path
)
1505 f
= setmntent("/proc/self/mounts", "r");
1509 while ((mnt
= getmntent(f
)) != NULL
) {
1510 if (strcmp(mnt
->mnt_dir
, path
))
1519 static int is_reg_file(const char *path
)
1521 struct stat statbuf
;
1523 if (stat(path
, &statbuf
) < 0)
1525 return S_ISREG(statbuf
.st_mode
);
1529 * This function checks if the given input parameter is
1531 * return <0 : some error in the given input
1532 * return BTRFS_ARG_UNKNOWN: unknown input
1533 * return BTRFS_ARG_UUID: given input is uuid
1534 * return BTRFS_ARG_MNTPOINT: given input is path
1535 * return BTRFS_ARG_REG: given input is regular file
1536 * return BTRFS_ARG_BLKDEV: given input is block device
1538 int check_arg_type(const char *input
)
1541 char path
[PATH_MAX
];
1546 if (realpath(input
, path
)) {
1547 if (is_block_device(path
) == 1)
1548 return BTRFS_ARG_BLKDEV
;
1550 if (is_mount_point(path
) == 1)
1551 return BTRFS_ARG_MNTPOINT
;
1553 if (is_reg_file(path
))
1554 return BTRFS_ARG_REG
;
1556 return BTRFS_ARG_UNKNOWN
;
1559 if (strlen(input
) == (BTRFS_UUID_UNPARSED_SIZE
- 1) &&
1560 !uuid_parse(input
, uuid
))
1561 return BTRFS_ARG_UUID
;
1563 return BTRFS_ARG_UNKNOWN
;
1567 * Find the mount point for a mounted device.
1568 * On success, returns 0 with mountpoint in *mp.
1569 * On failure, returns -errno (not mounted yields -EINVAL)
1570 * Is noisy on failures, expects to be given a mounted device.
1572 int get_btrfs_mount(const char *dev
, char *mp
, size_t mp_size
)
1577 ret
= is_block_device(dev
);
1580 error("not a block device: %s", dev
);
1583 error("cannot check %s: %s", dev
, strerror(-ret
));
1588 fd
= open(dev
, O_RDONLY
);
1591 error("cannot open %s: %s", dev
, strerror(errno
));
1595 ret
= check_mounted_where(fd
, dev
, mp
, mp_size
, NULL
);
1598 } else { /* mounted, all good */
1608 * Given a pathname, return a filehandle to:
1609 * the original pathname or,
1610 * if the pathname is a mounted btrfs device, to its mountpoint.
1612 * On error, return -1, errno should be set.
1614 int open_path_or_dev_mnt(const char *path
, DIR **dirstream
, int verbose
)
1619 if (is_block_device(path
)) {
1620 ret
= get_btrfs_mount(path
, mp
, sizeof(mp
));
1622 /* not a mounted btrfs dev */
1623 error_on(verbose
, "'%s' is not a mounted btrfs device",
1628 ret
= open_file_or_dir(mp
, dirstream
);
1629 error_on(verbose
&& ret
< 0, "can't access '%s': %s",
1630 path
, strerror(errno
));
1632 ret
= btrfs_open_dir(path
, dirstream
, 1);
1639 * Do the following checks before calling open_file_or_dir():
1640 * 1: path is in a btrfs filesystem
1641 * 2: path is a directory
1643 int btrfs_open_dir(const char *path
, DIR **dirstream
, int verbose
)
1649 if (statfs(path
, &stfs
) != 0) {
1650 error_on(verbose
, "cannot access '%s': %s", path
,
1655 if (stfs
.f_type
!= BTRFS_SUPER_MAGIC
) {
1656 error_on(verbose
, "not a btrfs filesystem: %s", path
);
1660 if (stat(path
, &st
) != 0) {
1661 error_on(verbose
, "cannot access '%s': %s", path
,
1666 if (!S_ISDIR(st
.st_mode
)) {
1667 error_on(verbose
, "not a directory: %s", path
);
1671 ret
= open_file_or_dir(path
, dirstream
);
1673 error_on(verbose
, "cannot access '%s': %s", path
,
1680 /* checks if a device is a loop device */
1681 static int is_loop_device (const char* device
) {
1682 struct stat statbuf
;
1684 if(stat(device
, &statbuf
) < 0)
1687 return (S_ISBLK(statbuf
.st_mode
) &&
1688 MAJOR(statbuf
.st_rdev
) == LOOP_MAJOR
);
1692 * Takes a loop device path (e.g. /dev/loop0) and returns
1693 * the associated file (e.g. /images/my_btrfs.img) using
1696 static int resolve_loop_device_with_loopdev(const char* loop_dev
, char* loop_file
)
1700 struct loop_info64 lo64
;
1702 fd
= open(loop_dev
, O_RDONLY
| O_NONBLOCK
);
1705 ret
= ioctl(fd
, LOOP_GET_STATUS64
, &lo64
);
1711 memcpy(loop_file
, lo64
.lo_file_name
, sizeof(lo64
.lo_file_name
));
1712 loop_file
[sizeof(lo64
.lo_file_name
)] = 0;
1720 /* Takes a loop device path (e.g. /dev/loop0) and returns
1721 * the associated file (e.g. /images/my_btrfs.img) */
1722 static int resolve_loop_device(const char* loop_dev
, char* loop_file
,
1729 char real_loop_dev
[PATH_MAX
];
1731 if (!realpath(loop_dev
, real_loop_dev
))
1733 snprintf(p
, PATH_MAX
, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev
, '/'));
1734 if (!(f
= fopen(p
, "r"))) {
1735 if (errno
== ENOENT
)
1737 * It's possibly a partitioned loop device, which is
1738 * resolvable with loopdev API.
1740 return resolve_loop_device_with_loopdev(loop_dev
, loop_file
);
1744 snprintf(fmt
, 20, "%%%i[^\n]", max_len
-1);
1745 ret
= fscanf(f
, fmt
, loop_file
);
1754 * Checks whether a and b are identical or device
1755 * files associated with the same block device
1757 static int is_same_blk_file(const char* a
, const char* b
)
1759 struct stat st_buf_a
, st_buf_b
;
1760 char real_a
[PATH_MAX
];
1761 char real_b
[PATH_MAX
];
1763 if (!realpath(a
, real_a
))
1764 strncpy_null(real_a
, a
);
1766 if (!realpath(b
, real_b
))
1767 strncpy_null(real_b
, b
);
1769 /* Identical path? */
1770 if (strcmp(real_a
, real_b
) == 0)
1773 if (stat(a
, &st_buf_a
) < 0 || stat(b
, &st_buf_b
) < 0) {
1774 if (errno
== ENOENT
)
1779 /* Same blockdevice? */
1780 if (S_ISBLK(st_buf_a
.st_mode
) && S_ISBLK(st_buf_b
.st_mode
) &&
1781 st_buf_a
.st_rdev
== st_buf_b
.st_rdev
) {
1786 if (st_buf_a
.st_dev
== st_buf_b
.st_dev
&&
1787 st_buf_a
.st_ino
== st_buf_b
.st_ino
) {
1794 /* checks if a and b are identical or device
1795 * files associated with the same block device or
1796 * if one file is a loop device that uses the other
1799 static int is_same_loop_file(const char* a
, const char* b
)
1801 char res_a
[PATH_MAX
];
1802 char res_b
[PATH_MAX
];
1803 const char* final_a
= NULL
;
1804 const char* final_b
= NULL
;
1807 /* Resolve a if it is a loop device */
1808 if((ret
= is_loop_device(a
)) < 0) {
1813 ret
= resolve_loop_device(a
, res_a
, sizeof(res_a
));
1824 /* Resolve b if it is a loop device */
1825 if ((ret
= is_loop_device(b
)) < 0) {
1830 ret
= resolve_loop_device(b
, res_b
, sizeof(res_b
));
1841 return is_same_blk_file(final_a
, final_b
);
1844 /* Checks if a file exists and is a block or regular file*/
1845 static int is_existing_blk_or_reg_file(const char* filename
)
1849 if(stat(filename
, &st_buf
) < 0) {
1856 return (S_ISBLK(st_buf
.st_mode
) || S_ISREG(st_buf
.st_mode
));
1859 /* Checks if a file is used (directly or indirectly via a loop device)
1860 * by a device in fs_devices
1862 static int blk_file_in_dev_list(struct btrfs_fs_devices
* fs_devices
,
1866 struct list_head
*head
;
1867 struct list_head
*cur
;
1868 struct btrfs_device
*device
;
1870 head
= &fs_devices
->devices
;
1871 list_for_each(cur
, head
) {
1872 device
= list_entry(cur
, struct btrfs_device
, dev_list
);
1874 if((ret
= is_same_loop_file(device
->name
, file
)))
1882 * Resolve a pathname to a device mapper node to /dev/mapper/<name>
1883 * Returns NULL on invalid input or malloc failure; Other failures
1884 * will be handled by the caller using the input pathame.
1886 char *canonicalize_dm_name(const char *ptname
)
1890 char path
[PATH_MAX
], name
[PATH_MAX
], *res
= NULL
;
1892 if (!ptname
|| !*ptname
)
1895 snprintf(path
, sizeof(path
), "/sys/block/%s/dm/name", ptname
);
1896 if (!(f
= fopen(path
, "r")))
1899 /* read <name>\n from sysfs */
1900 if (fgets(name
, sizeof(name
), f
) && (sz
= strlen(name
)) > 1) {
1901 name
[sz
- 1] = '\0';
1902 snprintf(path
, sizeof(path
), "/dev/mapper/%s", name
);
1904 if (access(path
, F_OK
) == 0)
1912 * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or
1913 * to a device mapper pathname.
1914 * Returns NULL on invalid input or malloc failure; Other failures
1915 * will be handled by the caller using the input pathame.
1917 char *canonicalize_path(const char *path
)
1919 char *canonical
, *p
;
1921 if (!path
|| !*path
)
1924 canonical
= realpath(path
, NULL
);
1926 return strdup(path
);
1927 p
= strrchr(canonical
, '/');
1928 if (p
&& strncmp(p
, "/dm-", 4) == 0 && isdigit(*(p
+ 4))) {
1929 char *dm
= canonicalize_dm_name(p
+ 1);
1940 * returns 1 if the device was mounted, < 0 on error or 0 if everything
1941 * is safe to continue.
1943 int check_mounted(const char* file
)
1948 fd
= open(file
, O_RDONLY
);
1950 error("mount check: cannot open %s: %s", file
,
1955 ret
= check_mounted_where(fd
, file
, NULL
, 0, NULL
);
1961 int check_mounted_where(int fd
, const char *file
, char *where
, int size
,
1962 struct btrfs_fs_devices
**fs_dev_ret
)
1967 struct btrfs_fs_devices
*fs_devices_mnt
= NULL
;
1971 /* scan the initial device */
1972 ret
= btrfs_scan_one_device(fd
, file
, &fs_devices_mnt
,
1973 &total_devs
, BTRFS_SUPER_INFO_OFFSET
, 0);
1974 is_btrfs
= (ret
>= 0);
1976 /* scan other devices */
1977 if (is_btrfs
&& total_devs
> 1) {
1978 ret
= btrfs_scan_lblkid();
1983 /* iterate over the list of currently mounted filesystems */
1984 if ((f
= setmntent ("/proc/self/mounts", "r")) == NULL
)
1987 while ((mnt
= getmntent (f
)) != NULL
) {
1989 if(strcmp(mnt
->mnt_type
, "btrfs") != 0)
1992 ret
= blk_file_in_dev_list(fs_devices_mnt
, mnt
->mnt_fsname
);
1994 /* ignore entries in the mount table that are not
1995 associated with a file*/
1996 if((ret
= is_existing_blk_or_reg_file(mnt
->mnt_fsname
)) < 0)
1997 goto out_mntloop_err
;
2001 ret
= is_same_loop_file(file
, mnt
->mnt_fsname
);
2005 goto out_mntloop_err
;
2010 /* Did we find an entry in mnt table? */
2011 if (mnt
&& size
&& where
) {
2012 strncpy(where
, mnt
->mnt_dir
, size
);
2016 *fs_dev_ret
= fs_devices_mnt
;
2018 ret
= (mnt
!= NULL
);
2026 struct pending_dir
{
2027 struct list_head list
;
2028 char name
[PATH_MAX
];
2031 int btrfs_register_one_device(const char *fname
)
2033 struct btrfs_ioctl_vol_args args
;
2037 fd
= open("/dev/btrfs-control", O_RDWR
);
2040 "failed to open /dev/btrfs-control, skipping device registration: %s",
2044 memset(&args
, 0, sizeof(args
));
2045 strncpy_null(args
.name
, fname
);
2046 ret
= ioctl(fd
, BTRFS_IOC_SCAN_DEV
, &args
);
2048 error("device scan failed on '%s': %s", fname
,
2057 * Register all devices in the fs_uuid list created in the user
2058 * space. Ensure btrfs_scan_lblkid() is called before this func.
2060 int btrfs_register_all_devices(void)
2064 struct btrfs_fs_devices
*fs_devices
;
2065 struct btrfs_device
*device
;
2066 struct list_head
*all_uuids
;
2068 all_uuids
= btrfs_scanned_uuids();
2070 list_for_each_entry(fs_devices
, all_uuids
, list
) {
2071 list_for_each_entry(device
, &fs_devices
->devices
, dev_list
) {
2073 err
= btrfs_register_one_device(device
->name
);
2083 int btrfs_device_already_in_root(struct btrfs_root
*root
, int fd
,
2086 struct btrfs_super_block
*disk_super
;
2090 buf
= malloc(BTRFS_SUPER_INFO_SIZE
);
2095 ret
= pread(fd
, buf
, BTRFS_SUPER_INFO_SIZE
, super_offset
);
2096 if (ret
!= BTRFS_SUPER_INFO_SIZE
)
2100 disk_super
= (struct btrfs_super_block
*)buf
;
2101 if (btrfs_super_magic(disk_super
) != BTRFS_MAGIC
)
2104 if (!memcmp(disk_super
->fsid
, root
->fs_info
->super_copy
->fsid
,
2114 * Note: this function uses a static per-thread buffer. Do not call this
2115 * function more than 10 times within one argument list!
2117 const char *pretty_size_mode(u64 size
, unsigned mode
)
2119 static __thread
int ps_index
= 0;
2120 static __thread
char ps_array
[10][32];
2123 ret
= ps_array
[ps_index
];
2126 (void)pretty_size_snprintf(size
, ret
, 32, mode
);
2131 static const char* unit_suffix_binary
[] =
2132 { "B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB"};
2133 static const char* unit_suffix_decimal
[] =
2134 { "B", "kB", "MB", "GB", "TB", "PB", "EB"};
2136 int pretty_size_snprintf(u64 size
, char *str
, size_t str_size
, unsigned unit_mode
)
2142 const char** suffix
= NULL
;
2148 if ((unit_mode
& ~UNITS_MODE_MASK
) == UNITS_RAW
) {
2149 snprintf(str
, str_size
, "%llu", size
);
2153 if ((unit_mode
& ~UNITS_MODE_MASK
) == UNITS_BINARY
) {
2156 suffix
= unit_suffix_binary
;
2157 } else if ((unit_mode
& ~UNITS_MODE_MASK
) == UNITS_DECIMAL
) {
2160 suffix
= unit_suffix_decimal
;
2165 fprintf(stderr
, "INTERNAL ERROR: unknown unit base, mode %d\n",
2173 switch (unit_mode
& UNITS_MODE_MASK
) {
2174 case UNITS_TBYTES
: base
*= mult
; num_divs
++;
2175 case UNITS_GBYTES
: base
*= mult
; num_divs
++;
2176 case UNITS_MBYTES
: base
*= mult
; num_divs
++;
2177 case UNITS_KBYTES
: num_divs
++;
2184 while (size
>= mult
) {
2190 * If the value is smaller than base, we didn't do any
2191 * division, in that case, base should be 1, not original
2192 * base, or the unit will be wrong
2198 if (num_divs
>= ARRAY_SIZE(unit_suffix_binary
)) {
2200 printf("INTERNAL ERROR: unsupported unit suffix, index %d\n",
2205 fraction
= (float)last_size
/ base
;
2207 return snprintf(str
, str_size
, "%.2f%s", fraction
, suffix
[num_divs
]);
2211 * __strncpy_null - strncpy with null termination
2212 * @dest: the target array
2213 * @src: the source string
2214 * @n: maximum bytes to copy (size of *dest)
2216 * Like strncpy, but ensures destination is null-terminated.
2218 * Copies the string pointed to by src, including the terminating null
2219 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
2220 * of n bytes. Then ensure that dest is null-terminated.
2222 char *__strncpy_null(char *dest
, const char *src
, size_t n
)
2224 strncpy(dest
, src
, n
);
2231 * Checks to make sure that the label matches our requirements.
2233 0 if everything is safe and usable
2234 -1 if the label is too long
2236 static int check_label(const char *input
)
2238 int len
= strlen(input
);
2240 if (len
> BTRFS_LABEL_SIZE
- 1) {
2241 error("label %s is too long (max %d)", input
,
2242 BTRFS_LABEL_SIZE
- 1);
2249 static int set_label_unmounted(const char *dev
, const char *label
)
2251 struct btrfs_trans_handle
*trans
;
2252 struct btrfs_root
*root
;
2255 ret
= check_mounted(dev
);
2257 error("checking mount status of %s failed: %d", dev
, ret
);
2261 error("device %s is mounted, use mount point", dev
);
2265 /* Open the super_block at the default location
2266 * and as read-write.
2268 root
= open_ctree(dev
, 0, OPEN_CTREE_WRITES
);
2269 if (!root
) /* errors are printed by open_ctree() */
2272 trans
= btrfs_start_transaction(root
, 1);
2273 __strncpy_null(root
->fs_info
->super_copy
->label
, label
, BTRFS_LABEL_SIZE
- 1);
2275 btrfs_commit_transaction(trans
, root
);
2277 /* Now we close it since we are done. */
2282 static int set_label_mounted(const char *mount_path
, const char *labelp
)
2285 char label
[BTRFS_LABEL_SIZE
];
2287 fd
= open(mount_path
, O_RDONLY
| O_NOATIME
);
2289 error("unable to access %s: %s", mount_path
, strerror(errno
));
2293 memset(label
, 0, sizeof(label
));
2294 __strncpy_null(label
, labelp
, BTRFS_LABEL_SIZE
- 1);
2295 if (ioctl(fd
, BTRFS_IOC_SET_FSLABEL
, label
) < 0) {
2296 error("unable to set label of %s: %s", mount_path
,
2306 int get_label_unmounted(const char *dev
, char *label
)
2308 struct btrfs_root
*root
;
2311 ret
= check_mounted(dev
);
2313 error("checking mount status of %s failed: %d", dev
, ret
);
2317 /* Open the super_block at the default location
2320 root
= open_ctree(dev
, 0, 0);
2324 __strncpy_null(label
, root
->fs_info
->super_copy
->label
,
2325 BTRFS_LABEL_SIZE
- 1);
2327 /* Now we close it since we are done. */
2333 * If a partition is mounted, try to get the filesystem label via its
2334 * mounted path rather than device. Return the corresponding error
2335 * the user specified the device path.
2337 int get_label_mounted(const char *mount_path
, char *labelp
)
2339 char label
[BTRFS_LABEL_SIZE
];
2343 fd
= open(mount_path
, O_RDONLY
| O_NOATIME
);
2345 error("unable to access %s: %s", mount_path
, strerror(errno
));
2349 memset(label
, '\0', sizeof(label
));
2350 ret
= ioctl(fd
, BTRFS_IOC_GET_FSLABEL
, label
);
2352 if (errno
!= ENOTTY
)
2353 error("unable to get label of %s: %s", mount_path
,
2360 __strncpy_null(labelp
, label
, BTRFS_LABEL_SIZE
- 1);
2365 int get_label(const char *btrfs_dev
, char *label
)
2369 ret
= is_existing_blk_or_reg_file(btrfs_dev
);
2371 ret
= get_label_mounted(btrfs_dev
, label
);
2373 ret
= get_label_unmounted(btrfs_dev
, label
);
2378 int set_label(const char *btrfs_dev
, const char *label
)
2382 if (check_label(label
))
2385 ret
= is_existing_blk_or_reg_file(btrfs_dev
);
2387 ret
= set_label_mounted(btrfs_dev
, label
);
2389 ret
= set_label_unmounted(btrfs_dev
, label
);
2395 * A not-so-good version fls64. No fascinating optimization since
2396 * no one except parse_size use it
2398 static int fls64(u64 x
)
2402 for (i
= 0; i
<64; i
++)
2403 if (x
<< i
& (1ULL << 63))
2408 u64
parse_size(char *s
)
2416 error("size value is empty");
2420 error("size value '%s' is less equal than 0", s
);
2423 ret
= strtoull(s
, &endptr
, 10);
2425 error("size value '%s' is invalid", s
);
2428 if (endptr
[0] && endptr
[1]) {
2429 error("illegal suffix contains character '%c' in wrong position",
2434 * strtoll returns LLONG_MAX when overflow, if this happens,
2435 * need to call strtoull to get the real size
2437 if (errno
== ERANGE
&& ret
== ULLONG_MAX
) {
2438 error("size value '%s' is too large for u64", s
);
2442 c
= tolower(endptr
[0]);
2465 error("unknown size descriptor '%c'", c
);
2469 /* Check whether ret * mult overflow */
2470 if (fls64(ret
) + fls64(mult
) - 1 > 64) {
2471 error("size value '%s' is too large for u64", s
);
2478 u64
parse_qgroupid(const char *p
)
2480 char *s
= strchr(p
, '/');
2481 const char *ptr_src_end
= p
+ strlen(p
);
2482 char *ptr_parse_end
= NULL
;
2491 /* Numeric format like '0/257' is the primary case */
2493 id
= strtoull(p
, &ptr_parse_end
, 10);
2494 if (ptr_parse_end
!= ptr_src_end
)
2498 level
= strtoull(p
, &ptr_parse_end
, 10);
2499 if (ptr_parse_end
!= s
)
2502 id
= strtoull(s
+ 1, &ptr_parse_end
, 10);
2503 if (ptr_parse_end
!= ptr_src_end
)
2506 return (level
<< BTRFS_QGROUP_LEVEL_SHIFT
) | id
;
2509 /* Path format like subv at 'my_subvol' is the fallback case */
2510 ret
= test_issubvolume(p
);
2511 if (ret
< 0 || !ret
)
2513 fd
= open(p
, O_RDONLY
);
2516 ret
= lookup_ino_rootid(fd
, &id
);
2523 error("invalid qgroupid or subvolume path: %s", p
);
2527 int open_file_or_dir3(const char *fname
, DIR **dirstream
, int open_flags
)
2533 ret
= stat(fname
, &st
);
2537 if (S_ISDIR(st
.st_mode
)) {
2538 *dirstream
= opendir(fname
);
2541 fd
= dirfd(*dirstream
);
2542 } else if (S_ISREG(st
.st_mode
) || S_ISLNK(st
.st_mode
)) {
2543 fd
= open(fname
, open_flags
);
2546 * we set this on purpose, in case the caller output
2547 * strerror(errno) as success
2555 closedir(*dirstream
);
2562 int open_file_or_dir(const char *fname
, DIR **dirstream
)
2564 return open_file_or_dir3(fname
, dirstream
, O_RDWR
);
2567 void close_file_or_dir(int fd
, DIR *dirstream
)
2570 closedir(dirstream
);
2575 int get_device_info(int fd
, u64 devid
,
2576 struct btrfs_ioctl_dev_info_args
*di_args
)
2580 di_args
->devid
= devid
;
2581 memset(&di_args
->uuid
, '\0', sizeof(di_args
->uuid
));
2583 ret
= ioctl(fd
, BTRFS_IOC_DEV_INFO
, di_args
);
2584 return ret
< 0 ? -errno
: 0;
2587 static u64
find_max_device_id(struct btrfs_ioctl_search_args
*search_args
,
2590 struct btrfs_dev_item
*dev_item
;
2591 char *buf
= search_args
->buf
;
2593 buf
+= (nr_items
- 1) * (sizeof(struct btrfs_ioctl_search_header
)
2594 + sizeof(struct btrfs_dev_item
));
2595 buf
+= sizeof(struct btrfs_ioctl_search_header
);
2597 dev_item
= (struct btrfs_dev_item
*)buf
;
2599 return btrfs_stack_device_id(dev_item
);
2602 static int search_chunk_tree_for_fs_info(int fd
,
2603 struct btrfs_ioctl_fs_info_args
*fi_args
)
2607 u64 start_devid
= 1;
2608 struct btrfs_ioctl_search_args search_args
;
2609 struct btrfs_ioctl_search_key
*search_key
= &search_args
.key
;
2611 fi_args
->num_devices
= 0;
2613 max_items
= BTRFS_SEARCH_ARGS_BUFSIZE
2614 / (sizeof(struct btrfs_ioctl_search_header
)
2615 + sizeof(struct btrfs_dev_item
));
2617 search_key
->tree_id
= BTRFS_CHUNK_TREE_OBJECTID
;
2618 search_key
->min_objectid
= BTRFS_DEV_ITEMS_OBJECTID
;
2619 search_key
->max_objectid
= BTRFS_DEV_ITEMS_OBJECTID
;
2620 search_key
->min_type
= BTRFS_DEV_ITEM_KEY
;
2621 search_key
->max_type
= BTRFS_DEV_ITEM_KEY
;
2622 search_key
->min_transid
= 0;
2623 search_key
->max_transid
= (u64
)-1;
2624 search_key
->nr_items
= max_items
;
2625 search_key
->max_offset
= (u64
)-1;
2628 search_key
->min_offset
= start_devid
;
2630 ret
= ioctl(fd
, BTRFS_IOC_TREE_SEARCH
, &search_args
);
2634 fi_args
->num_devices
+= (u64
)search_key
->nr_items
;
2636 if (search_key
->nr_items
== max_items
) {
2637 start_devid
= find_max_device_id(&search_args
,
2638 search_key
->nr_items
) + 1;
2642 /* get the lastest max_id to stay consistent with the num_devices */
2643 if (search_key
->nr_items
== 0)
2645 * last tree_search returns an empty buf, use the devid of
2646 * the last dev_item of the previous tree_search
2648 fi_args
->max_id
= start_devid
- 1;
2650 fi_args
->max_id
= find_max_device_id(&search_args
,
2651 search_key
->nr_items
);
2657 * For a given path, fill in the ioctl fs_ and info_ args.
2658 * If the path is a btrfs mountpoint, fill info for all devices.
2659 * If the path is a btrfs device, fill in only that device.
2661 * The path provided must be either on a mounted btrfs fs,
2662 * or be a mounted btrfs device.
2664 * Returns 0 on success, or a negative errno.
2666 int get_fs_info(char *path
, struct btrfs_ioctl_fs_info_args
*fi_args
,
2667 struct btrfs_ioctl_dev_info_args
**di_ret
)
2674 struct btrfs_fs_devices
*fs_devices_mnt
= NULL
;
2675 struct btrfs_ioctl_dev_info_args
*di_args
;
2676 struct btrfs_ioctl_dev_info_args tmp
;
2678 DIR *dirstream
= NULL
;
2680 memset(fi_args
, 0, sizeof(*fi_args
));
2682 if (is_block_device(path
) == 1) {
2683 struct btrfs_super_block
*disk_super
;
2684 char buf
[BTRFS_SUPER_INFO_SIZE
];
2687 /* Ensure it's mounted, then set path to the mountpoint */
2688 fd
= open(path
, O_RDONLY
);
2691 error("cannot open %s: %s", path
, strerror(errno
));
2694 ret
= check_mounted_where(fd
, path
, mp
, sizeof(mp
),
2703 /* Only fill in this one device */
2704 fi_args
->num_devices
= 1;
2706 disk_super
= (struct btrfs_super_block
*)buf
;
2707 ret
= btrfs_read_dev_super(fd
, disk_super
,
2708 BTRFS_SUPER_INFO_OFFSET
, 0);
2713 devid
= btrfs_stack_device_id(&disk_super
->dev_item
);
2715 fi_args
->max_id
= devid
;
2718 memcpy(fi_args
->fsid
, fs_devices_mnt
->fsid
, BTRFS_FSID_SIZE
);
2722 /* at this point path must not be for a block device */
2723 fd
= open_file_or_dir(path
, &dirstream
);
2729 /* fill in fi_args if not just a single device */
2730 if (fi_args
->num_devices
!= 1) {
2731 ret
= ioctl(fd
, BTRFS_IOC_FS_INFO
, fi_args
);
2738 * The fs_args->num_devices does not include seed devices
2740 ret
= search_chunk_tree_for_fs_info(fd
, fi_args
);
2745 * search_chunk_tree_for_fs_info() will lacks the devid 0
2746 * so manual probe for it here.
2748 ret
= get_device_info(fd
, 0, &tmp
);
2750 fi_args
->num_devices
++;
2758 if (!fi_args
->num_devices
)
2761 di_args
= *di_ret
= malloc((fi_args
->num_devices
) * sizeof(*di_args
));
2768 memcpy(di_args
, &tmp
, sizeof(tmp
));
2769 for (; i
<= fi_args
->max_id
; ++i
) {
2770 ret
= get_device_info(fd
, i
, &di_args
[ndevs
]);
2779 * only when the only dev we wanted to find is not there then
2780 * let any error be returned
2782 if (fi_args
->num_devices
!= 1) {
2788 close_file_or_dir(fd
, dirstream
);
2792 #define isoctal(c) (((c) & ~7) == '0')
2794 static inline void translate(char *f
, char *t
)
2796 while (*f
!= '\0') {
2798 isoctal(f
[1]) && isoctal(f
[2]) && isoctal(f
[3])) {
2799 *t
++ = 64*(f
[1] & 7) + 8*(f
[2] & 7) + (f
[3] & 7);
2809 * Checks if the swap device.
2810 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
2812 static int is_swap_device(const char *file
)
2823 if (stat(file
, &st_buf
) < 0)
2825 if (S_ISBLK(st_buf
.st_mode
))
2826 dev
= st_buf
.st_rdev
;
2827 else if (S_ISREG(st_buf
.st_mode
)) {
2828 dev
= st_buf
.st_dev
;
2829 ino
= st_buf
.st_ino
;
2833 if ((f
= fopen("/proc/swaps", "r")) == NULL
)
2836 /* skip the first line */
2837 if (fgets(tmp
, sizeof(tmp
), f
) == NULL
)
2840 while (fgets(tmp
, sizeof(tmp
), f
) != NULL
) {
2841 if ((cp
= strchr(tmp
, ' ')) != NULL
)
2843 if ((cp
= strchr(tmp
, '\t')) != NULL
)
2845 translate(tmp
, buf
);
2846 if (stat(buf
, &st_buf
) != 0)
2848 if (S_ISBLK(st_buf
.st_mode
)) {
2849 if (dev
== st_buf
.st_rdev
) {
2853 } else if (S_ISREG(st_buf
.st_mode
)) {
2854 if (dev
== st_buf
.st_dev
&& ino
== st_buf
.st_ino
) {
2868 * Check for existing filesystem or partition table on device.
2870 * 1 for existing fs or partition
2871 * 0 for nothing found
2872 * -1 for internal error
2874 static int check_overwrite(const char *device
)
2877 blkid_probe pr
= NULL
;
2881 if (!device
|| !*device
)
2884 ret
= -1; /* will reset on success of all setup calls */
2886 pr
= blkid_new_probe_from_filename(device
);
2890 size
= blkid_probe_get_size(pr
);
2894 /* nothing to overwrite on a 0-length device */
2900 ret
= blkid_probe_enable_partitions(pr
, 1);
2904 ret
= blkid_do_fullprobe(pr
);
2909 * Blkid returns 1 for nothing found and 0 when it finds a signature,
2910 * but we want the exact opposite, so reverse the return value here.
2912 * In addition print some useful diagnostics about what actually is
2920 if (!blkid_probe_lookup_value(pr
, "TYPE", &type
, NULL
)) {
2922 "%s appears to contain an existing "
2923 "filesystem (%s).\n", device
, type
);
2924 } else if (!blkid_probe_lookup_value(pr
, "PTTYPE", &type
, NULL
)) {
2926 "%s appears to contain a partition "
2927 "table (%s).\n", device
, type
);
2930 "%s appears to contain something weird "
2931 "according to blkid\n", device
);
2937 blkid_free_probe(pr
);
2940 "probe of %s failed, cannot detect "
2941 "existing filesystem.\n", device
);
2945 static int group_profile_devs_min(u64 flag
)
2947 switch (flag
& BTRFS_BLOCK_GROUP_PROFILE_MASK
) {
2948 case 0: /* single */
2949 case BTRFS_BLOCK_GROUP_DUP
:
2951 case BTRFS_BLOCK_GROUP_RAID0
:
2952 case BTRFS_BLOCK_GROUP_RAID1
:
2953 case BTRFS_BLOCK_GROUP_RAID5
:
2955 case BTRFS_BLOCK_GROUP_RAID6
:
2957 case BTRFS_BLOCK_GROUP_RAID10
:
2964 int test_num_disk_vs_raid(u64 metadata_profile
, u64 data_profile
,
2965 u64 dev_cnt
, int mixed
, int ssd
)
2972 allowed
|= BTRFS_BLOCK_GROUP_RAID10
;
2974 allowed
|= BTRFS_BLOCK_GROUP_RAID6
;
2976 allowed
|= BTRFS_BLOCK_GROUP_RAID0
| BTRFS_BLOCK_GROUP_RAID1
|
2977 BTRFS_BLOCK_GROUP_RAID5
;
2979 allowed
|= BTRFS_BLOCK_GROUP_DUP
;
2983 ((metadata_profile
| data_profile
) & BTRFS_BLOCK_GROUP_DUP
)) {
2984 warning("DUP is not recommended on filesystem with multiple devices");
2986 if (metadata_profile
& ~allowed
) {
2988 "ERROR: unable to create FS with metadata profile %s "
2989 "(have %llu devices but %d devices are required)\n",
2990 btrfs_group_profile_str(metadata_profile
), dev_cnt
,
2991 group_profile_devs_min(metadata_profile
));
2994 if (data_profile
& ~allowed
) {
2996 "ERROR: unable to create FS with data profile %s "
2997 "(have %llu devices but %d devices are required)\n",
2998 btrfs_group_profile_str(data_profile
), dev_cnt
,
2999 group_profile_devs_min(data_profile
));
3003 warning_on(!mixed
&& (data_profile
& BTRFS_BLOCK_GROUP_DUP
) && ssd
,
3004 "DUP may not actually lead to 2 copies on the device, see manual page");
3009 int group_profile_max_safe_loss(u64 flags
)
3011 switch (flags
& BTRFS_BLOCK_GROUP_PROFILE_MASK
) {
3012 case 0: /* single */
3013 case BTRFS_BLOCK_GROUP_DUP
:
3014 case BTRFS_BLOCK_GROUP_RAID0
:
3016 case BTRFS_BLOCK_GROUP_RAID1
:
3017 case BTRFS_BLOCK_GROUP_RAID5
:
3018 case BTRFS_BLOCK_GROUP_RAID10
:
3020 case BTRFS_BLOCK_GROUP_RAID6
:
3028 * Check if a device is suitable for btrfs
3030 * 1: something is wrong, an error is printed
3033 int test_dev_for_mkfs(const char *file
, int force_overwrite
)
3038 ret
= is_swap_device(file
);
3040 error("checking status of %s: %s", file
, strerror(-ret
));
3044 error("%s is a swap device", file
);
3047 if (!force_overwrite
) {
3048 if (check_overwrite(file
)) {
3049 error("use the -f option to force overwrite of %s",
3054 ret
= check_mounted(file
);
3056 error("cannot check mount status of %s: %s", file
,
3061 error("%s is mounted", file
);
3064 /* check if the device is busy */
3065 fd
= open(file
, O_RDWR
|O_EXCL
);
3067 error("unable to open %s: %s", file
, strerror(errno
));
3070 if (fstat(fd
, &st
)) {
3071 error("unable to stat %s: %s", file
, strerror(errno
));
3075 if (!S_ISBLK(st
.st_mode
)) {
3076 error("%s is not a block device", file
);
3084 int btrfs_scan_lblkid(void)
3089 struct btrfs_fs_devices
*tmp_devices
;
3090 blkid_dev_iterate iter
= NULL
;
3091 blkid_dev dev
= NULL
;
3092 blkid_cache cache
= NULL
;
3093 char path
[PATH_MAX
];
3095 if (btrfs_scan_done
)
3098 if (blkid_get_cache(&cache
, NULL
) < 0) {
3099 error("blkid cache get failed");
3102 blkid_probe_all(cache
);
3103 iter
= blkid_dev_iterate_begin(cache
);
3104 blkid_dev_set_search(iter
, "TYPE", "btrfs");
3105 while (blkid_dev_next(iter
, &dev
) == 0) {
3106 dev
= blkid_verify(cache
, dev
);
3109 /* if we are here its definitely a btrfs disk*/
3110 strncpy_null(path
, blkid_dev_devname(dev
));
3112 fd
= open(path
, O_RDONLY
);
3114 error("cannot open %s: %s", path
, strerror(errno
));
3117 ret
= btrfs_scan_one_device(fd
, path
, &tmp_devices
,
3118 &num_devices
, BTRFS_SUPER_INFO_OFFSET
, 0);
3120 error("cannot scan %s: %s", path
, strerror(-ret
));
3127 blkid_dev_iterate_end(iter
);
3128 blkid_put_cache(cache
);
3130 btrfs_scan_done
= 1;
3135 int is_vol_small(const char *file
)
3142 fd
= open(file
, O_RDONLY
);
3145 if (fstat(fd
, &st
) < 0) {
3150 size
= btrfs_device_size(fd
, &st
);
3155 if (size
< BTRFS_MKFS_SMALL_VOLUME_SIZE
) {
3165 * This reads a line from the stdin and only returns non-zero if the
3166 * first whitespace delimited token is a case insensitive match with yes
3169 int ask_user(const char *question
)
3171 char buf
[30] = {0,};
3172 char *saveptr
= NULL
;
3175 printf("%s [y/N]: ", question
);
3177 return fgets(buf
, sizeof(buf
) - 1, stdin
) &&
3178 (answer
= strtok_r(buf
, " \t\n\r", &saveptr
)) &&
3179 (!strcasecmp(answer
, "yes") || !strcasecmp(answer
, "y"));
3184 * - file or directory return the containing tree root id
3185 * - subvolume return its own tree id
3186 * - BTRFS_EMPTY_SUBVOL_DIR_OBJECTID (directory with ino == 2) the result is
3187 * undefined and function returns -1
3189 int lookup_ino_rootid(int fd
, u64
*rootid
)
3191 struct btrfs_ioctl_ino_lookup_args args
;
3194 memset(&args
, 0, sizeof(args
));
3196 args
.objectid
= BTRFS_FIRST_FREE_OBJECTID
;
3198 ret
= ioctl(fd
, BTRFS_IOC_INO_LOOKUP
, &args
);
3200 error("failed to lookup root id: %s", strerror(errno
));
3204 *rootid
= args
.treeid
;
3210 * return 0 if a btrfs mount point is found
3211 * return 1 if a mount point is found but not btrfs
3212 * return <0 if something goes wrong
3214 int find_mount_root(const char *path
, char **mount_root
)
3222 int longest_matchlen
= 0;
3223 char *longest_match
= NULL
;
3225 fd
= open(path
, O_RDONLY
| O_NOATIME
);
3230 mnttab
= setmntent("/proc/self/mounts", "r");
3234 while ((ent
= getmntent(mnttab
))) {
3235 len
= strlen(ent
->mnt_dir
);
3236 if (strncmp(ent
->mnt_dir
, path
, len
) == 0) {
3237 /* match found and use the latest match */
3238 if (longest_matchlen
<= len
) {
3239 free(longest_match
);
3240 longest_matchlen
= len
;
3241 longest_match
= strdup(ent
->mnt_dir
);
3242 not_btrfs
= strcmp(ent
->mnt_type
, "btrfs");
3251 free(longest_match
);
3256 *mount_root
= realpath(longest_match
, NULL
);
3260 free(longest_match
);
3264 int test_minimum_size(const char *file
, u32 nodesize
)
3267 struct stat statbuf
;
3269 fd
= open(file
, O_RDONLY
);
3272 if (stat(file
, &statbuf
) < 0) {
3276 if (btrfs_device_size(fd
, &statbuf
) < btrfs_min_dev_size(nodesize
)) {
3286 * Test if path is a directory
3288 * 0 - path exists but it is not a directory
3289 * 1 - path exists and it is a directory
3292 int test_isdir(const char *path
)
3297 ret
= stat(path
, &st
);
3301 return !!S_ISDIR(st
.st_mode
);
3304 void units_set_mode(unsigned *units
, unsigned mode
)
3306 unsigned base
= *units
& UNITS_MODE_MASK
;
3308 *units
= base
| mode
;
3311 void units_set_base(unsigned *units
, unsigned base
)
3313 unsigned mode
= *units
& ~UNITS_MODE_MASK
;
3315 *units
= base
| mode
;
3318 int find_next_key(struct btrfs_path
*path
, struct btrfs_key
*key
)
3322 for (level
= 0; level
< BTRFS_MAX_LEVEL
; level
++) {
3323 if (!path
->nodes
[level
])
3325 if (path
->slots
[level
] + 1 >=
3326 btrfs_header_nritems(path
->nodes
[level
]))
3329 btrfs_item_key_to_cpu(path
->nodes
[level
], key
,
3330 path
->slots
[level
] + 1);
3332 btrfs_node_key_to_cpu(path
->nodes
[level
], key
,
3333 path
->slots
[level
] + 1);
3339 const char* btrfs_group_type_str(u64 flag
)
3341 u64 mask
= BTRFS_BLOCK_GROUP_TYPE_MASK
|
3342 BTRFS_SPACE_INFO_GLOBAL_RSV
;
3344 switch (flag
& mask
) {
3345 case BTRFS_BLOCK_GROUP_DATA
:
3347 case BTRFS_BLOCK_GROUP_SYSTEM
:
3349 case BTRFS_BLOCK_GROUP_METADATA
:
3351 case BTRFS_BLOCK_GROUP_DATA
|BTRFS_BLOCK_GROUP_METADATA
:
3352 return "Data+Metadata";
3353 case BTRFS_SPACE_INFO_GLOBAL_RSV
:
3354 return "GlobalReserve";
3360 const char* btrfs_group_profile_str(u64 flag
)
3362 switch (flag
& BTRFS_BLOCK_GROUP_PROFILE_MASK
) {
3365 case BTRFS_BLOCK_GROUP_RAID0
:
3367 case BTRFS_BLOCK_GROUP_RAID1
:
3369 case BTRFS_BLOCK_GROUP_RAID5
:
3371 case BTRFS_BLOCK_GROUP_RAID6
:
3373 case BTRFS_BLOCK_GROUP_DUP
:
3375 case BTRFS_BLOCK_GROUP_RAID10
:
3382 u64
disk_size(const char *path
)
3386 if (statfs(path
, &sfs
) < 0)
3389 return sfs
.f_bsize
* sfs
.f_blocks
;
3392 u64
get_partition_size(const char *dev
)
3395 int fd
= open(dev
, O_RDONLY
);
3399 if (ioctl(fd
, BLKGETSIZE64
, &result
) < 0) {
3408 int btrfs_tree_search2_ioctl_supported(int fd
)
3410 struct btrfs_ioctl_search_args_v2
*args2
;
3411 struct btrfs_ioctl_search_key
*sk
;
3412 int args2_size
= 1024;
3413 char args2_buf
[args2_size
];
3415 static int v2_supported
= -1;
3417 if (v2_supported
!= -1)
3418 return v2_supported
;
3420 args2
= (struct btrfs_ioctl_search_args_v2
*)args2_buf
;
3424 * Search for the extent tree item in the root tree.
3426 sk
->tree_id
= BTRFS_ROOT_TREE_OBJECTID
;
3427 sk
->min_objectid
= BTRFS_EXTENT_TREE_OBJECTID
;
3428 sk
->max_objectid
= BTRFS_EXTENT_TREE_OBJECTID
;
3429 sk
->min_type
= BTRFS_ROOT_ITEM_KEY
;
3430 sk
->max_type
= BTRFS_ROOT_ITEM_KEY
;
3432 sk
->max_offset
= (u64
)-1;
3433 sk
->min_transid
= 0;
3434 sk
->max_transid
= (u64
)-1;
3436 args2
->buf_size
= args2_size
- sizeof(struct btrfs_ioctl_search_args_v2
);
3437 ret
= ioctl(fd
, BTRFS_IOC_TREE_SEARCH_V2
, args2
);
3438 if (ret
== -EOPNOTSUPP
)
3445 return v2_supported
;
3448 int btrfs_check_nodesize(u32 nodesize
, u32 sectorsize
, u64 features
)
3450 if (nodesize
< sectorsize
) {
3451 error("illegal nodesize %u (smaller than %u)",
3452 nodesize
, sectorsize
);
3454 } else if (nodesize
> BTRFS_MAX_METADATA_BLOCKSIZE
) {
3455 error("illegal nodesize %u (larger than %u)",
3456 nodesize
, BTRFS_MAX_METADATA_BLOCKSIZE
);
3458 } else if (nodesize
& (sectorsize
- 1)) {
3459 error("illegal nodesize %u (not aligned to %u)",
3460 nodesize
, sectorsize
);
3462 } else if (features
& BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS
&&
3463 nodesize
!= sectorsize
) {
3464 error("illegal nodesize %u (not equal to %u for mixed block group)",
3465 nodesize
, sectorsize
);
3472 * Copy a path argument from SRC to DEST and check the SRC length if it's at
3473 * most PATH_MAX and fits into DEST. DESTLEN is supposed to be exact size of
3475 * The destination buffer is zero terminated.
3476 * Return < 0 for error, 0 otherwise.
3478 int arg_copy_path(char *dest
, const char *src
, int destlen
)
3480 size_t len
= strlen(src
);
3482 if (len
>= PATH_MAX
|| len
>= destlen
)
3483 return -ENAMETOOLONG
;
3485 __strncpy_null(dest
, src
, destlen
);
3490 unsigned int get_unit_mode_from_arg(int *argc
, char *argv
[], int df_mode
)
3492 unsigned int unit_mode
= UNITS_DEFAULT
;
3496 for (arg_i
= 0; arg_i
< *argc
; arg_i
++) {
3497 if (!strcmp(argv
[arg_i
], "--"))
3500 if (!strcmp(argv
[arg_i
], "--raw")) {
3501 unit_mode
= UNITS_RAW
;
3505 if (!strcmp(argv
[arg_i
], "--human-readable")) {
3506 unit_mode
= UNITS_HUMAN_BINARY
;
3511 if (!strcmp(argv
[arg_i
], "--iec")) {
3512 units_set_mode(&unit_mode
, UNITS_BINARY
);
3516 if (!strcmp(argv
[arg_i
], "--si")) {
3517 units_set_mode(&unit_mode
, UNITS_DECIMAL
);
3522 if (!strcmp(argv
[arg_i
], "--kbytes")) {
3523 units_set_base(&unit_mode
, UNITS_KBYTES
);
3527 if (!strcmp(argv
[arg_i
], "--mbytes")) {
3528 units_set_base(&unit_mode
, UNITS_MBYTES
);
3532 if (!strcmp(argv
[arg_i
], "--gbytes")) {
3533 units_set_base(&unit_mode
, UNITS_GBYTES
);
3537 if (!strcmp(argv
[arg_i
], "--tbytes")) {
3538 units_set_base(&unit_mode
, UNITS_TBYTES
);
3546 if (!strcmp(argv
[arg_i
], "-b")) {
3547 unit_mode
= UNITS_RAW
;
3551 if (!strcmp(argv
[arg_i
], "-h")) {
3552 unit_mode
= UNITS_HUMAN_BINARY
;
3556 if (!strcmp(argv
[arg_i
], "-H")) {
3557 unit_mode
= UNITS_HUMAN_DECIMAL
;
3561 if (!strcmp(argv
[arg_i
], "-k")) {
3562 units_set_base(&unit_mode
, UNITS_KBYTES
);
3566 if (!strcmp(argv
[arg_i
], "-m")) {
3567 units_set_base(&unit_mode
, UNITS_MBYTES
);
3571 if (!strcmp(argv
[arg_i
], "-g")) {
3572 units_set_base(&unit_mode
, UNITS_GBYTES
);
3576 if (!strcmp(argv
[arg_i
], "-t")) {
3577 units_set_base(&unit_mode
, UNITS_TBYTES
);
3583 for (arg_i
= 0, arg_end
= 0; arg_i
< *argc
; arg_i
++) {
3586 argv
[arg_end
] = argv
[arg_i
];
3595 int string_is_numerical(const char *str
)
3597 if (!(*str
>= '0' && *str
<= '9'))
3599 while (*str
>= '0' && *str
<= '9')
3607 * Preprocess @argv with getopt_long to reorder options and consume the "--"
3609 * Unknown short and long options are reported, optionally the @usage is printed
3612 void clean_args_no_options(int argc
, char *argv
[], const char * const *usagestr
)
3614 static const struct option long_options
[] = {
3619 int c
= getopt_long(argc
, argv
, "", long_options
, NULL
);
3632 /* Subvolume helper functions */
3634 * test if name is a correct subvolume name
3635 * this function return
3636 * 0-> name is not a correct subvolume name
3637 * 1-> name is a correct subvolume name
3639 int test_issubvolname(const char *name
)
3641 return name
[0] != '\0' && !strchr(name
, '/') &&
3642 strcmp(name
, ".") && strcmp(name
, "..");
3646 * Test if path is a subvolume
3648 * 0 - path exists but it is not a subvolume
3649 * 1 - path exists and it is a subvolume
3652 int test_issubvolume(const char *path
)
3658 res
= stat(path
, &st
);
3662 if (st
.st_ino
!= BTRFS_FIRST_FREE_OBJECTID
|| !S_ISDIR(st
.st_mode
))
3665 res
= statfs(path
, &stfs
);
3669 return (int)stfs
.f_type
== BTRFS_SUPER_MAGIC
;
3672 const char *subvol_strip_mountpoint(const char *mnt
, const char *full_path
)
3674 int len
= strlen(mnt
);
3678 if (mnt
[len
- 1] != '/')
3681 return full_path
+ len
;
3688 * 1: Error; and error info printed to the terminal. Fixme.
3689 * 2: If the fullpath is root tree instead of subvol tree
3691 int get_subvol_info(const char *fullpath
, struct root_info
*get_ri
)
3698 const char *svpath
= NULL
;
3699 DIR *dirstream1
= NULL
;
3700 DIR *dirstream2
= NULL
;
3702 ret
= test_issubvolume(fullpath
);
3706 error("not a subvolume: %s", fullpath
);
3710 ret
= find_mount_root(fullpath
, &mnt
);
3714 error("%s doesn't belong to btrfs mount point", fullpath
);
3718 svpath
= subvol_strip_mountpoint(mnt
, fullpath
);
3720 fd
= btrfs_open_dir(fullpath
, &dirstream1
, 1);
3724 ret
= btrfs_list_get_path_rootid(fd
, &sv_id
);
3726 error("can't get rootid for '%s'", fullpath
);
3730 mntfd
= btrfs_open_dir(mnt
, &dirstream2
, 1);
3734 if (sv_id
== BTRFS_FS_TREE_OBJECTID
) {
3737 * So that caller may decide if thats an error or just fine.
3742 memset(get_ri
, 0, sizeof(*get_ri
));
3743 get_ri
->root_id
= sv_id
;
3745 ret
= btrfs_get_subvol(mntfd
, get_ri
);
3747 error("can't find '%s': %d", svpath
, ret
);
3750 close_file_or_dir(mntfd
, dirstream2
);
3751 close_file_or_dir(fd
, dirstream1
);
3757 void init_rand_seed(u64 seed
)
3761 /* only use the last 48 bits */
3762 for (i
= 0; i
< 3; i
++) {
3763 rand_seed
[i
] = (unsigned short)(seed
^ (unsigned short)(-1));
3766 rand_seed_initlized
= 1;
3769 static void __init_seed(void)
3775 if(rand_seed_initlized
)
3777 /* Use urandom as primary seed source. */
3778 fd
= open("/dev/urandom", O_RDONLY
);
3780 ret
= read(fd
, rand_seed
, sizeof(rand_seed
));
3782 if (ret
< sizeof(rand_seed
))
3786 /* Use time and pid as fallback seed */
3787 warning("failed to read /dev/urandom, use time and pid as random seed");
3788 gettimeofday(&tv
, 0);
3789 rand_seed
[0] = getpid() ^ (tv
.tv_sec
& 0xFFFF);
3790 rand_seed
[1] = getppid() ^ (tv
.tv_usec
& 0xFFFF);
3791 rand_seed
[2] = (tv
.tv_sec
^ tv
.tv_usec
) >> 16;
3793 rand_seed_initlized
= 1;
3800 * Don't use nrand48, its range is [0,2^31) The highest bit will alwasy
3801 * be 0. Use jrand48 to include the highest bit.
3803 return (u32
)jrand48(rand_seed
);
3806 unsigned int rand_range(unsigned int upper
)
3810 * Use the full 48bits to mod, which would be more uniformly
3813 return (unsigned int)(jrand48(rand_seed
) % upper
);