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.
20 #define _XOPEN_SOURCE 700
21 #define __USE_XOPEN2K8
22 #define __XOPEN2K8 /* due to an error in dirent.h, to get dirfd() */
23 #define _GNU_SOURCE /* O_NOATIME */
27 #include <sys/ioctl.h>
28 #include <sys/mount.h>
29 #include <sys/types.h>
31 #include <uuid/uuid.h>
36 #include <linux/loop.h>
37 #include <linux/major.h>
38 #include <linux/kdev_t.h>
40 #include <blkid/blkid.h>
41 #include "kerncompat.h"
42 #include "radix-tree.h"
45 #include "transaction.h"
52 #define BLKDISCARD _IO(0x12,119)
56 * Discard the given range in one go
58 static int discard_range(int fd
, u64 start
, u64 len
)
60 u64 range
[2] = { start
, len
};
62 if (ioctl(fd
, BLKDISCARD
, &range
) < 0)
68 * Discard blocks in the given range in 1G chunks, the process is interruptible
70 static int discard_blocks(int fd
, u64 start
, u64 len
)
74 u64 chunk_size
= min_t(u64
, len
, 1*1024*1024*1024);
77 ret
= discard_range(fd
, start
, chunk_size
);
87 static u64 reference_root_table
[] = {
88 [1] = BTRFS_ROOT_TREE_OBJECTID
,
89 [2] = BTRFS_EXTENT_TREE_OBJECTID
,
90 [3] = BTRFS_CHUNK_TREE_OBJECTID
,
91 [4] = BTRFS_DEV_TREE_OBJECTID
,
92 [5] = BTRFS_FS_TREE_OBJECTID
,
93 [6] = BTRFS_CSUM_TREE_OBJECTID
,
96 int test_uuid_unique(char *fs_uuid
)
99 blkid_dev_iterate iter
= NULL
;
100 blkid_dev dev
= NULL
;
101 blkid_cache cache
= NULL
;
103 if (blkid_get_cache(&cache
, 0) < 0) {
104 printf("ERROR: lblkid cache get failed\n");
107 blkid_probe_all(cache
);
108 iter
= blkid_dev_iterate_begin(cache
);
109 blkid_dev_set_search(iter
, "UUID", fs_uuid
);
111 while (blkid_dev_next(iter
, &dev
) == 0) {
112 dev
= blkid_verify(cache
, dev
);
119 blkid_dev_iterate_end(iter
);
120 blkid_put_cache(cache
);
125 int make_btrfs(int fd
, const char *device
, const char *label
, char *fs_uuid
,
126 u64 blocks
[7], u64 num_bytes
, u32 nodesize
,
127 u32 leafsize
, u32 sectorsize
, u32 stripesize
, u64 features
)
129 struct btrfs_super_block super
;
130 struct extent_buffer
*buf
= NULL
;
131 struct btrfs_root_item root_item
;
132 struct btrfs_disk_key disk_key
;
133 struct btrfs_extent_item
*extent_item
;
134 struct btrfs_inode_item
*inode_item
;
135 struct btrfs_chunk
*chunk
;
136 struct btrfs_dev_item
*dev_item
;
137 struct btrfs_dev_extent
*dev_extent
;
138 u8 chunk_tree_uuid
[BTRFS_UUID_SIZE
];
148 int skinny_metadata
= !!(features
&
149 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA
);
151 first_free
= BTRFS_SUPER_INFO_OFFSET
+ sectorsize
* 2 - 1;
152 first_free
&= ~((u64
)sectorsize
- 1);
154 memset(&super
, 0, sizeof(super
));
156 num_bytes
= (num_bytes
/ sectorsize
) * sectorsize
;
158 if (uuid_parse(fs_uuid
, super
.fsid
) != 0) {
159 fprintf(stderr
, "could not parse UUID: %s\n", fs_uuid
);
163 if (!test_uuid_unique(fs_uuid
)) {
164 fprintf(stderr
, "non-unique UUID: %s\n", fs_uuid
);
169 uuid_generate(super
.fsid
);
171 uuid_generate(super
.dev_item
.uuid
);
172 uuid_generate(chunk_tree_uuid
);
174 btrfs_set_super_bytenr(&super
, blocks
[0]);
175 btrfs_set_super_num_devices(&super
, 1);
176 btrfs_set_super_magic(&super
, BTRFS_MAGIC
);
177 btrfs_set_super_generation(&super
, 1);
178 btrfs_set_super_root(&super
, blocks
[1]);
179 btrfs_set_super_chunk_root(&super
, blocks
[3]);
180 btrfs_set_super_total_bytes(&super
, num_bytes
);
181 btrfs_set_super_bytes_used(&super
, 6 * leafsize
);
182 btrfs_set_super_sectorsize(&super
, sectorsize
);
183 btrfs_set_super_leafsize(&super
, leafsize
);
184 btrfs_set_super_nodesize(&super
, nodesize
);
185 btrfs_set_super_stripesize(&super
, stripesize
);
186 btrfs_set_super_csum_type(&super
, BTRFS_CSUM_TYPE_CRC32
);
187 btrfs_set_super_chunk_root_generation(&super
, 1);
188 btrfs_set_super_cache_generation(&super
, -1);
189 btrfs_set_super_incompat_flags(&super
, features
);
191 strncpy(super
.label
, label
, BTRFS_LABEL_SIZE
- 1);
193 buf
= malloc(sizeof(*buf
) + max(sectorsize
, leafsize
));
195 /* create the tree of root objects */
196 memset(buf
->data
, 0, leafsize
);
198 btrfs_set_header_bytenr(buf
, blocks
[1]);
199 btrfs_set_header_nritems(buf
, 4);
200 btrfs_set_header_generation(buf
, 1);
201 btrfs_set_header_backref_rev(buf
, BTRFS_MIXED_BACKREF_REV
);
202 btrfs_set_header_owner(buf
, BTRFS_ROOT_TREE_OBJECTID
);
203 write_extent_buffer(buf
, super
.fsid
, btrfs_header_fsid(),
206 write_extent_buffer(buf
, chunk_tree_uuid
,
207 btrfs_header_chunk_tree_uuid(buf
),
210 /* create the items for the root tree */
211 memset(&root_item
, 0, sizeof(root_item
));
212 inode_item
= &root_item
.inode
;
213 btrfs_set_stack_inode_generation(inode_item
, 1);
214 btrfs_set_stack_inode_size(inode_item
, 3);
215 btrfs_set_stack_inode_nlink(inode_item
, 1);
216 btrfs_set_stack_inode_nbytes(inode_item
, leafsize
);
217 btrfs_set_stack_inode_mode(inode_item
, S_IFDIR
| 0755);
218 btrfs_set_root_refs(&root_item
, 1);
219 btrfs_set_root_used(&root_item
, leafsize
);
220 btrfs_set_root_generation(&root_item
, 1);
222 memset(&disk_key
, 0, sizeof(disk_key
));
223 btrfs_set_disk_key_type(&disk_key
, BTRFS_ROOT_ITEM_KEY
);
224 btrfs_set_disk_key_offset(&disk_key
, 0);
227 itemoff
= __BTRFS_LEAF_DATA_SIZE(leafsize
) - sizeof(root_item
);
228 btrfs_set_root_bytenr(&root_item
, blocks
[2]);
229 btrfs_set_disk_key_objectid(&disk_key
, BTRFS_EXTENT_TREE_OBJECTID
);
230 btrfs_set_item_key(buf
, &disk_key
, nritems
);
231 btrfs_set_item_offset(buf
, btrfs_item_nr(nritems
), itemoff
);
232 btrfs_set_item_size(buf
, btrfs_item_nr(nritems
),
234 write_extent_buffer(buf
, &root_item
, btrfs_item_ptr_offset(buf
,
235 nritems
), sizeof(root_item
));
238 itemoff
= itemoff
- sizeof(root_item
);
239 btrfs_set_root_bytenr(&root_item
, blocks
[4]);
240 btrfs_set_disk_key_objectid(&disk_key
, BTRFS_DEV_TREE_OBJECTID
);
241 btrfs_set_item_key(buf
, &disk_key
, nritems
);
242 btrfs_set_item_offset(buf
, btrfs_item_nr(nritems
), itemoff
);
243 btrfs_set_item_size(buf
, btrfs_item_nr(nritems
),
245 write_extent_buffer(buf
, &root_item
,
246 btrfs_item_ptr_offset(buf
, nritems
),
250 itemoff
= itemoff
- sizeof(root_item
);
251 btrfs_set_root_bytenr(&root_item
, blocks
[5]);
252 btrfs_set_disk_key_objectid(&disk_key
, BTRFS_FS_TREE_OBJECTID
);
253 btrfs_set_item_key(buf
, &disk_key
, nritems
);
254 btrfs_set_item_offset(buf
, btrfs_item_nr(nritems
), itemoff
);
255 btrfs_set_item_size(buf
, btrfs_item_nr(nritems
),
257 write_extent_buffer(buf
, &root_item
,
258 btrfs_item_ptr_offset(buf
, nritems
),
262 itemoff
= itemoff
- sizeof(root_item
);
263 btrfs_set_root_bytenr(&root_item
, blocks
[6]);
264 btrfs_set_disk_key_objectid(&disk_key
, BTRFS_CSUM_TREE_OBJECTID
);
265 btrfs_set_item_key(buf
, &disk_key
, nritems
);
266 btrfs_set_item_offset(buf
, btrfs_item_nr(nritems
), itemoff
);
267 btrfs_set_item_size(buf
, btrfs_item_nr(nritems
),
269 write_extent_buffer(buf
, &root_item
,
270 btrfs_item_ptr_offset(buf
, nritems
),
275 csum_tree_block_size(buf
, BTRFS_CRC32_SIZE
, 0);
276 ret
= pwrite(fd
, buf
->data
, leafsize
, blocks
[1]);
277 if (ret
!= leafsize
) {
278 ret
= (ret
< 0 ? -errno
: -EIO
);
282 /* create the items for the extent tree */
283 memset(buf
->data
+sizeof(struct btrfs_header
), 0,
284 leafsize
-sizeof(struct btrfs_header
));
286 itemoff
= __BTRFS_LEAF_DATA_SIZE(leafsize
);
287 for (i
= 1; i
< 7; i
++) {
288 item_size
= sizeof(struct btrfs_extent_item
);
289 if (!skinny_metadata
)
290 item_size
+= sizeof(struct btrfs_tree_block_info
);
292 BUG_ON(blocks
[i
] < first_free
);
293 BUG_ON(blocks
[i
] < blocks
[i
- 1]);
295 /* create extent item */
296 itemoff
-= item_size
;
297 btrfs_set_disk_key_objectid(&disk_key
, blocks
[i
]);
298 if (skinny_metadata
) {
299 btrfs_set_disk_key_type(&disk_key
,
300 BTRFS_METADATA_ITEM_KEY
);
301 btrfs_set_disk_key_offset(&disk_key
, 0);
303 btrfs_set_disk_key_type(&disk_key
,
304 BTRFS_EXTENT_ITEM_KEY
);
305 btrfs_set_disk_key_offset(&disk_key
, leafsize
);
307 btrfs_set_item_key(buf
, &disk_key
, nritems
);
308 btrfs_set_item_offset(buf
, btrfs_item_nr(nritems
),
310 btrfs_set_item_size(buf
, btrfs_item_nr(nritems
),
312 extent_item
= btrfs_item_ptr(buf
, nritems
,
313 struct btrfs_extent_item
);
314 btrfs_set_extent_refs(buf
, extent_item
, 1);
315 btrfs_set_extent_generation(buf
, extent_item
, 1);
316 btrfs_set_extent_flags(buf
, extent_item
,
317 BTRFS_EXTENT_FLAG_TREE_BLOCK
);
320 /* create extent ref */
321 ref_root
= reference_root_table
[i
];
322 btrfs_set_disk_key_objectid(&disk_key
, blocks
[i
]);
323 btrfs_set_disk_key_offset(&disk_key
, ref_root
);
324 btrfs_set_disk_key_type(&disk_key
, BTRFS_TREE_BLOCK_REF_KEY
);
325 btrfs_set_item_key(buf
, &disk_key
, nritems
);
326 btrfs_set_item_offset(buf
, btrfs_item_nr(nritems
),
328 btrfs_set_item_size(buf
, btrfs_item_nr(nritems
), 0);
331 btrfs_set_header_bytenr(buf
, blocks
[2]);
332 btrfs_set_header_owner(buf
, BTRFS_EXTENT_TREE_OBJECTID
);
333 btrfs_set_header_nritems(buf
, nritems
);
334 csum_tree_block_size(buf
, BTRFS_CRC32_SIZE
, 0);
335 ret
= pwrite(fd
, buf
->data
, leafsize
, blocks
[2]);
336 if (ret
!= leafsize
) {
337 ret
= (ret
< 0 ? -errno
: -EIO
);
341 /* create the chunk tree */
342 memset(buf
->data
+sizeof(struct btrfs_header
), 0,
343 leafsize
-sizeof(struct btrfs_header
));
345 item_size
= sizeof(*dev_item
);
346 itemoff
= __BTRFS_LEAF_DATA_SIZE(leafsize
) - item_size
;
348 /* first device 1 (there is no device 0) */
349 btrfs_set_disk_key_objectid(&disk_key
, BTRFS_DEV_ITEMS_OBJECTID
);
350 btrfs_set_disk_key_offset(&disk_key
, 1);
351 btrfs_set_disk_key_type(&disk_key
, BTRFS_DEV_ITEM_KEY
);
352 btrfs_set_item_key(buf
, &disk_key
, nritems
);
353 btrfs_set_item_offset(buf
, btrfs_item_nr(nritems
), itemoff
);
354 btrfs_set_item_size(buf
, btrfs_item_nr(nritems
), item_size
);
356 dev_item
= btrfs_item_ptr(buf
, nritems
, struct btrfs_dev_item
);
357 btrfs_set_device_id(buf
, dev_item
, 1);
358 btrfs_set_device_generation(buf
, dev_item
, 0);
359 btrfs_set_device_total_bytes(buf
, dev_item
, num_bytes
);
360 btrfs_set_device_bytes_used(buf
, dev_item
,
361 BTRFS_MKFS_SYSTEM_GROUP_SIZE
);
362 btrfs_set_device_io_align(buf
, dev_item
, sectorsize
);
363 btrfs_set_device_io_width(buf
, dev_item
, sectorsize
);
364 btrfs_set_device_sector_size(buf
, dev_item
, sectorsize
);
365 btrfs_set_device_type(buf
, dev_item
, 0);
367 write_extent_buffer(buf
, super
.dev_item
.uuid
,
368 (unsigned long)btrfs_device_uuid(dev_item
),
370 write_extent_buffer(buf
, super
.fsid
,
371 (unsigned long)btrfs_device_fsid(dev_item
),
373 read_extent_buffer(buf
, &super
.dev_item
, (unsigned long)dev_item
,
377 item_size
= btrfs_chunk_item_size(1);
378 itemoff
= itemoff
- item_size
;
380 /* then we have chunk 0 */
381 btrfs_set_disk_key_objectid(&disk_key
, BTRFS_FIRST_CHUNK_TREE_OBJECTID
);
382 btrfs_set_disk_key_offset(&disk_key
, 0);
383 btrfs_set_disk_key_type(&disk_key
, BTRFS_CHUNK_ITEM_KEY
);
384 btrfs_set_item_key(buf
, &disk_key
, nritems
);
385 btrfs_set_item_offset(buf
, btrfs_item_nr(nritems
), itemoff
);
386 btrfs_set_item_size(buf
, btrfs_item_nr(nritems
), item_size
);
388 chunk
= btrfs_item_ptr(buf
, nritems
, struct btrfs_chunk
);
389 btrfs_set_chunk_length(buf
, chunk
, BTRFS_MKFS_SYSTEM_GROUP_SIZE
);
390 btrfs_set_chunk_owner(buf
, chunk
, BTRFS_EXTENT_TREE_OBJECTID
);
391 btrfs_set_chunk_stripe_len(buf
, chunk
, 64 * 1024);
392 btrfs_set_chunk_type(buf
, chunk
, BTRFS_BLOCK_GROUP_SYSTEM
);
393 btrfs_set_chunk_io_align(buf
, chunk
, sectorsize
);
394 btrfs_set_chunk_io_width(buf
, chunk
, sectorsize
);
395 btrfs_set_chunk_sector_size(buf
, chunk
, sectorsize
);
396 btrfs_set_chunk_num_stripes(buf
, chunk
, 1);
397 btrfs_set_stripe_devid_nr(buf
, chunk
, 0, 1);
398 btrfs_set_stripe_offset_nr(buf
, chunk
, 0, 0);
401 write_extent_buffer(buf
, super
.dev_item
.uuid
,
402 (unsigned long)btrfs_stripe_dev_uuid(&chunk
->stripe
),
405 /* copy the key for the chunk to the system array */
406 ptr
= super
.sys_chunk_array
;
407 array_size
= sizeof(disk_key
);
409 memcpy(ptr
, &disk_key
, sizeof(disk_key
));
410 ptr
+= sizeof(disk_key
);
412 /* copy the chunk to the system array */
413 read_extent_buffer(buf
, ptr
, (unsigned long)chunk
, item_size
);
414 array_size
+= item_size
;
416 btrfs_set_super_sys_array_size(&super
, array_size
);
418 btrfs_set_header_bytenr(buf
, blocks
[3]);
419 btrfs_set_header_owner(buf
, BTRFS_CHUNK_TREE_OBJECTID
);
420 btrfs_set_header_nritems(buf
, nritems
);
421 csum_tree_block_size(buf
, BTRFS_CRC32_SIZE
, 0);
422 ret
= pwrite(fd
, buf
->data
, leafsize
, blocks
[3]);
423 if (ret
!= leafsize
) {
424 ret
= (ret
< 0 ? -errno
: -EIO
);
428 /* create the device tree */
429 memset(buf
->data
+sizeof(struct btrfs_header
), 0,
430 leafsize
-sizeof(struct btrfs_header
));
432 itemoff
= __BTRFS_LEAF_DATA_SIZE(leafsize
) -
433 sizeof(struct btrfs_dev_extent
);
435 btrfs_set_disk_key_objectid(&disk_key
, 1);
436 btrfs_set_disk_key_offset(&disk_key
, 0);
437 btrfs_set_disk_key_type(&disk_key
, BTRFS_DEV_EXTENT_KEY
);
438 btrfs_set_item_key(buf
, &disk_key
, nritems
);
439 btrfs_set_item_offset(buf
, btrfs_item_nr(nritems
), itemoff
);
440 btrfs_set_item_size(buf
, btrfs_item_nr(nritems
),
441 sizeof(struct btrfs_dev_extent
));
442 dev_extent
= btrfs_item_ptr(buf
, nritems
, struct btrfs_dev_extent
);
443 btrfs_set_dev_extent_chunk_tree(buf
, dev_extent
,
444 BTRFS_CHUNK_TREE_OBJECTID
);
445 btrfs_set_dev_extent_chunk_objectid(buf
, dev_extent
,
446 BTRFS_FIRST_CHUNK_TREE_OBJECTID
);
447 btrfs_set_dev_extent_chunk_offset(buf
, dev_extent
, 0);
449 write_extent_buffer(buf
, chunk_tree_uuid
,
450 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent
),
453 btrfs_set_dev_extent_length(buf
, dev_extent
,
454 BTRFS_MKFS_SYSTEM_GROUP_SIZE
);
457 btrfs_set_header_bytenr(buf
, blocks
[4]);
458 btrfs_set_header_owner(buf
, BTRFS_DEV_TREE_OBJECTID
);
459 btrfs_set_header_nritems(buf
, nritems
);
460 csum_tree_block_size(buf
, BTRFS_CRC32_SIZE
, 0);
461 ret
= pwrite(fd
, buf
->data
, leafsize
, blocks
[4]);
462 if (ret
!= leafsize
) {
463 ret
= (ret
< 0 ? -errno
: -EIO
);
467 /* create the FS root */
468 memset(buf
->data
+sizeof(struct btrfs_header
), 0,
469 leafsize
-sizeof(struct btrfs_header
));
470 btrfs_set_header_bytenr(buf
, blocks
[5]);
471 btrfs_set_header_owner(buf
, BTRFS_FS_TREE_OBJECTID
);
472 btrfs_set_header_nritems(buf
, 0);
473 csum_tree_block_size(buf
, BTRFS_CRC32_SIZE
, 0);
474 ret
= pwrite(fd
, buf
->data
, leafsize
, blocks
[5]);
475 if (ret
!= leafsize
) {
476 ret
= (ret
< 0 ? -errno
: -EIO
);
479 /* finally create the csum root */
480 memset(buf
->data
+sizeof(struct btrfs_header
), 0,
481 leafsize
-sizeof(struct btrfs_header
));
482 btrfs_set_header_bytenr(buf
, blocks
[6]);
483 btrfs_set_header_owner(buf
, BTRFS_CSUM_TREE_OBJECTID
);
484 btrfs_set_header_nritems(buf
, 0);
485 csum_tree_block_size(buf
, BTRFS_CRC32_SIZE
, 0);
486 ret
= pwrite(fd
, buf
->data
, leafsize
, blocks
[6]);
487 if (ret
!= leafsize
) {
488 ret
= (ret
< 0 ? -errno
: -EIO
);
492 /* and write out the super block */
493 BUG_ON(sizeof(super
) > sectorsize
);
494 memset(buf
->data
, 0, sectorsize
);
495 memcpy(buf
->data
, &super
, sizeof(super
));
496 buf
->len
= sectorsize
;
497 csum_tree_block_size(buf
, BTRFS_CRC32_SIZE
, 0);
498 ret
= pwrite(fd
, buf
->data
, sectorsize
, blocks
[0]);
499 if (ret
!= sectorsize
) {
500 ret
= (ret
< 0 ? -errno
: -EIO
);
511 u64
btrfs_device_size(int fd
, struct stat
*st
)
514 if (S_ISREG(st
->st_mode
)) {
517 if (!S_ISBLK(st
->st_mode
)) {
520 if (ioctl(fd
, BLKGETSIZE64
, &size
) >= 0) {
526 static int zero_blocks(int fd
, off_t start
, size_t len
)
528 char *buf
= malloc(len
);
535 written
= pwrite(fd
, buf
, len
, start
);
542 static int zero_dev_start(int fd
)
545 size_t len
= 2 * 1024 * 1024;
548 /* don't overwrite the disk labels on sparc */
552 return zero_blocks(fd
, start
, len
);
555 static int zero_dev_end(int fd
, u64 dev_size
)
557 size_t len
= 2 * 1024 * 1024;
558 off_t start
= dev_size
- len
;
560 return zero_blocks(fd
, start
, len
);
563 int btrfs_add_to_fsid(struct btrfs_trans_handle
*trans
,
564 struct btrfs_root
*root
, int fd
, char *path
,
565 u64 block_count
, u32 io_width
, u32 io_align
,
568 struct btrfs_super_block
*disk_super
;
569 struct btrfs_super_block
*super
= root
->fs_info
->super_copy
;
570 struct btrfs_device
*device
;
571 struct btrfs_dev_item
*dev_item
;
577 device
= kzalloc(sizeof(*device
), GFP_NOFS
);
580 buf
= kmalloc(sectorsize
, GFP_NOFS
);
585 BUG_ON(sizeof(*disk_super
) > sectorsize
);
586 memset(buf
, 0, sectorsize
);
588 disk_super
= (struct btrfs_super_block
*)buf
;
589 dev_item
= &disk_super
->dev_item
;
591 uuid_generate(device
->uuid
);
594 device
->io_width
= io_width
;
595 device
->io_align
= io_align
;
596 device
->sector_size
= sectorsize
;
598 device
->writeable
= 1;
599 device
->total_bytes
= block_count
;
600 device
->bytes_used
= 0;
601 device
->total_ios
= 0;
602 device
->dev_root
= root
->fs_info
->dev_root
;
604 ret
= btrfs_add_device(trans
, root
, device
);
607 total_bytes
= btrfs_super_total_bytes(super
) + block_count
;
608 btrfs_set_super_total_bytes(super
, total_bytes
);
610 num_devs
= btrfs_super_num_devices(super
) + 1;
611 btrfs_set_super_num_devices(super
, num_devs
);
613 memcpy(disk_super
, super
, sizeof(*disk_super
));
615 printf("adding device %s id %llu\n", path
,
616 (unsigned long long)device
->devid
);
618 btrfs_set_super_bytenr(disk_super
, BTRFS_SUPER_INFO_OFFSET
);
619 btrfs_set_stack_device_id(dev_item
, device
->devid
);
620 btrfs_set_stack_device_type(dev_item
, device
->type
);
621 btrfs_set_stack_device_io_align(dev_item
, device
->io_align
);
622 btrfs_set_stack_device_io_width(dev_item
, device
->io_width
);
623 btrfs_set_stack_device_sector_size(dev_item
, device
->sector_size
);
624 btrfs_set_stack_device_total_bytes(dev_item
, device
->total_bytes
);
625 btrfs_set_stack_device_bytes_used(dev_item
, device
->bytes_used
);
626 memcpy(&dev_item
->uuid
, device
->uuid
, BTRFS_UUID_SIZE
);
628 ret
= pwrite(fd
, buf
, sectorsize
, BTRFS_SUPER_INFO_OFFSET
);
629 BUG_ON(ret
!= sectorsize
);
632 list_add(&device
->dev_list
, &root
->fs_info
->fs_devices
->devices
);
633 device
->fs_devices
= root
->fs_info
->fs_devices
;
637 int btrfs_prepare_device(int fd
, char *file
, int zero_end
, u64
*block_count_ret
,
638 u64 max_block_count
, int *mixed
, int discard
)
645 ret
= fstat(fd
, &st
);
647 fprintf(stderr
, "unable to stat %s\n", file
);
651 block_count
= btrfs_device_size(fd
, &st
);
652 if (block_count
== 0) {
653 fprintf(stderr
, "unable to find %s size\n", file
);
657 block_count
= min(block_count
, max_block_count
);
659 if (block_count
< 1024 * 1024 * 1024 && !(*mixed
)) {
660 printf("SMALL VOLUME: forcing mixed metadata/data groups\n");
666 * We intentionally ignore errors from the discard ioctl. It
667 * is not necessary for the mkfs functionality but just an
670 if (discard_range(fd
, 0, 0) == 0) {
671 fprintf(stderr
, "Performing full device TRIM (%s) ...\n",
672 pretty_size(block_count
));
673 discard_blocks(fd
, 0, block_count
);
677 ret
= zero_dev_start(fd
);
681 for (i
= 0 ; i
< BTRFS_SUPER_MIRROR_MAX
; i
++) {
682 bytenr
= btrfs_sb_offset(i
);
683 if (bytenr
>= block_count
)
685 ret
= zero_blocks(fd
, bytenr
, BTRFS_SUPER_INFO_SIZE
);
691 ret
= zero_dev_end(fd
, block_count
);
695 *block_count_ret
= block_count
;
699 fprintf(stderr
, "ERROR: failed to zero device '%s' - %s\n",
700 file
, strerror(-ret
));
702 } else if (ret
> 0) {
703 fprintf(stderr
, "ERROR: failed to zero device '%s' - %d\n",
710 int btrfs_make_root_dir(struct btrfs_trans_handle
*trans
,
711 struct btrfs_root
*root
, u64 objectid
)
714 struct btrfs_inode_item inode_item
;
715 time_t now
= time(NULL
);
717 memset(&inode_item
, 0, sizeof(inode_item
));
718 btrfs_set_stack_inode_generation(&inode_item
, trans
->transid
);
719 btrfs_set_stack_inode_size(&inode_item
, 0);
720 btrfs_set_stack_inode_nlink(&inode_item
, 1);
721 btrfs_set_stack_inode_nbytes(&inode_item
, root
->leafsize
);
722 btrfs_set_stack_inode_mode(&inode_item
, S_IFDIR
| 0755);
723 btrfs_set_stack_timespec_sec(&inode_item
.atime
, now
);
724 btrfs_set_stack_timespec_nsec(&inode_item
.atime
, 0);
725 btrfs_set_stack_timespec_sec(&inode_item
.ctime
, now
);
726 btrfs_set_stack_timespec_nsec(&inode_item
.ctime
, 0);
727 btrfs_set_stack_timespec_sec(&inode_item
.mtime
, now
);
728 btrfs_set_stack_timespec_nsec(&inode_item
.mtime
, 0);
729 btrfs_set_stack_timespec_sec(&inode_item
.otime
, 0);
730 btrfs_set_stack_timespec_nsec(&inode_item
.otime
, 0);
732 if (root
->fs_info
->tree_root
== root
)
733 btrfs_set_super_root_dir(root
->fs_info
->super_copy
, objectid
);
735 ret
= btrfs_insert_inode(trans
, root
, objectid
, &inode_item
);
739 ret
= btrfs_insert_inode_ref(trans
, root
, "..", 2, objectid
, objectid
, 0);
743 btrfs_set_root_dirid(&root
->root_item
, objectid
);
750 * checks if a path is a block device node
751 * Returns negative errno on failure, otherwise
752 * returns 1 for blockdev, 0 for not-blockdev
754 int is_block_device(const char *path
)
758 if (stat(path
, &statbuf
) < 0)
761 return S_ISBLK(statbuf
.st_mode
);
765 * check if given path is a mount point
766 * return 1 if yes. 0 if no. -1 for error
768 int is_mount_point(const char *path
)
774 f
= setmntent("/proc/self/mounts", "r");
778 while ((mnt
= getmntent(f
)) != NULL
) {
779 if (strcmp(mnt
->mnt_dir
, path
))
789 * Find the mount point for a mounted device.
790 * On success, returns 0 with mountpoint in *mp.
791 * On failure, returns -errno (not mounted yields -EINVAL)
792 * Is noisy on failures, expects to be given a mounted device.
794 int get_btrfs_mount(const char *dev
, char *mp
, size_t mp_size
)
799 ret
= is_block_device(dev
);
802 fprintf(stderr
, "%s is not a block device\n", dev
);
805 fprintf(stderr
, "Could not check %s: %s\n",
806 dev
, strerror(-ret
));
811 fd
= open(dev
, O_RDONLY
);
814 fprintf(stderr
, "Could not open %s: %s\n", dev
, strerror(errno
));
818 ret
= check_mounted_where(fd
, dev
, mp
, mp_size
, NULL
);
821 } else { /* mounted, all good */
831 * Given a pathname, return a filehandle to:
832 * the original pathname or,
833 * if the pathname is a mounted btrfs device, to its mountpoint.
835 * On error, return -1, errno should be set.
837 int open_path_or_dev_mnt(const char *path
, DIR **dirstream
)
839 char mp
[BTRFS_PATH_NAME_MAX
+ 1];
842 if (is_block_device(path
)) {
845 ret
= get_btrfs_mount(path
, mp
, sizeof(mp
));
847 /* not a mounted btrfs dev */
851 fdmnt
= open_file_or_dir(mp
, dirstream
);
853 fdmnt
= open_file_or_dir(path
, dirstream
);
859 /* checks if a device is a loop device */
860 static int is_loop_device (const char* device
) {
863 if(stat(device
, &statbuf
) < 0)
866 return (S_ISBLK(statbuf
.st_mode
) &&
867 MAJOR(statbuf
.st_rdev
) == LOOP_MAJOR
);
871 /* Takes a loop device path (e.g. /dev/loop0) and returns
872 * the associated file (e.g. /images/my_btrfs.img) */
873 static int resolve_loop_device(const char* loop_dev
, char* loop_file
,
880 char real_loop_dev
[PATH_MAX
];
882 if (!realpath(loop_dev
, real_loop_dev
))
884 snprintf(p
, PATH_MAX
, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev
, '/'));
885 if (!(f
= fopen(p
, "r")))
888 snprintf(fmt
, 20, "%%%i[^\n]", max_len
-1);
889 ret
= fscanf(f
, fmt
, loop_file
);
897 /* Checks whether a and b are identical or device
898 * files associated with the same block device
900 static int is_same_blk_file(const char* a
, const char* b
)
902 struct stat st_buf_a
, st_buf_b
;
903 char real_a
[PATH_MAX
];
904 char real_b
[PATH_MAX
];
906 if(!realpath(a
, real_a
))
909 if (!realpath(b
, real_b
))
912 /* Identical path? */
913 if(strcmp(real_a
, real_b
) == 0)
916 if(stat(a
, &st_buf_a
) < 0 ||
917 stat(b
, &st_buf_b
) < 0)
924 /* Same blockdevice? */
925 if(S_ISBLK(st_buf_a
.st_mode
) &&
926 S_ISBLK(st_buf_b
.st_mode
) &&
927 st_buf_a
.st_rdev
== st_buf_b
.st_rdev
)
933 if (st_buf_a
.st_dev
== st_buf_b
.st_dev
&&
934 st_buf_a
.st_ino
== st_buf_b
.st_ino
)
942 /* checks if a and b are identical or device
943 * files associated with the same block device or
944 * if one file is a loop device that uses the other
947 static int is_same_loop_file(const char* a
, const char* b
)
949 char res_a
[PATH_MAX
];
950 char res_b
[PATH_MAX
];
951 const char* final_a
= NULL
;
952 const char* final_b
= NULL
;
955 /* Resolve a if it is a loop device */
956 if((ret
= is_loop_device(a
)) < 0) {
961 ret
= resolve_loop_device(a
, res_a
, sizeof(res_a
));
972 /* Resolve b if it is a loop device */
973 if ((ret
= is_loop_device(b
)) < 0) {
978 ret
= resolve_loop_device(b
, res_b
, sizeof(res_b
));
989 return is_same_blk_file(final_a
, final_b
);
992 /* Checks if a file exists and is a block or regular file*/
993 static int is_existing_blk_or_reg_file(const char* filename
)
997 if(stat(filename
, &st_buf
) < 0) {
1004 return (S_ISBLK(st_buf
.st_mode
) || S_ISREG(st_buf
.st_mode
));
1007 /* Checks if a file is used (directly or indirectly via a loop device)
1008 * by a device in fs_devices
1010 static int blk_file_in_dev_list(struct btrfs_fs_devices
* fs_devices
,
1014 struct list_head
*head
;
1015 struct list_head
*cur
;
1016 struct btrfs_device
*device
;
1018 head
= &fs_devices
->devices
;
1019 list_for_each(cur
, head
) {
1020 device
= list_entry(cur
, struct btrfs_device
, dev_list
);
1022 if((ret
= is_same_loop_file(device
->name
, file
)))
1030 * Resolve a pathname to a device mapper node to /dev/mapper/<name>
1031 * Returns NULL on invalid input or malloc failure; Other failures
1032 * will be handled by the caller using the input pathame.
1034 char *canonicalize_dm_name(const char *ptname
)
1038 char path
[PATH_MAX
], name
[PATH_MAX
], *res
= NULL
;
1040 if (!ptname
|| !*ptname
)
1043 snprintf(path
, sizeof(path
), "/sys/block/%s/dm/name", ptname
);
1044 if (!(f
= fopen(path
, "r")))
1047 /* read <name>\n from sysfs */
1048 if (fgets(name
, sizeof(name
), f
) && (sz
= strlen(name
)) > 1) {
1049 name
[sz
- 1] = '\0';
1050 snprintf(path
, sizeof(path
), "/dev/mapper/%s", name
);
1052 if (access(path
, F_OK
) == 0)
1060 * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or
1061 * to a device mapper pathname.
1062 * Returns NULL on invalid input or malloc failure; Other failures
1063 * will be handled by the caller using the input pathame.
1065 char *canonicalize_path(const char *path
)
1067 char *canonical
, *p
;
1069 if (!path
|| !*path
)
1072 canonical
= realpath(path
, NULL
);
1074 return strdup(path
);
1075 p
= strrchr(canonical
, '/');
1076 if (p
&& strncmp(p
, "/dm-", 4) == 0 && isdigit(*(p
+ 4))) {
1077 char *dm
= canonicalize_dm_name(p
+ 1);
1088 * returns 1 if the device was mounted, < 0 on error or 0 if everything
1089 * is safe to continue.
1091 int check_mounted(const char* file
)
1096 fd
= open(file
, O_RDONLY
);
1098 fprintf (stderr
, "check_mounted(): Could not open %s\n", file
);
1102 ret
= check_mounted_where(fd
, file
, NULL
, 0, NULL
);
1108 int check_mounted_where(int fd
, const char *file
, char *where
, int size
,
1109 struct btrfs_fs_devices
**fs_dev_ret
)
1114 struct btrfs_fs_devices
*fs_devices_mnt
= NULL
;
1118 /* scan the initial device */
1119 ret
= btrfs_scan_one_device(fd
, file
, &fs_devices_mnt
,
1120 &total_devs
, BTRFS_SUPER_INFO_OFFSET
);
1121 is_btrfs
= (ret
>= 0);
1123 /* scan other devices */
1124 if (is_btrfs
&& total_devs
> 1) {
1125 if ((ret
= btrfs_scan_for_fsid(!BTRFS_UPDATE_KERNEL
)))
1129 /* iterate over the list of currently mountes filesystems */
1130 if ((f
= setmntent ("/proc/self/mounts", "r")) == NULL
)
1133 while ((mnt
= getmntent (f
)) != NULL
) {
1135 if(strcmp(mnt
->mnt_type
, "btrfs") != 0)
1138 ret
= blk_file_in_dev_list(fs_devices_mnt
, mnt
->mnt_fsname
);
1140 /* ignore entries in the mount table that are not
1141 associated with a file*/
1142 if((ret
= is_existing_blk_or_reg_file(mnt
->mnt_fsname
)) < 0)
1143 goto out_mntloop_err
;
1147 ret
= is_same_loop_file(file
, mnt
->mnt_fsname
);
1151 goto out_mntloop_err
;
1156 /* Did we find an entry in mnt table? */
1157 if (mnt
&& size
&& where
) {
1158 strncpy(where
, mnt
->mnt_dir
, size
);
1162 *fs_dev_ret
= fs_devices_mnt
;
1164 ret
= (mnt
!= NULL
);
1172 struct pending_dir
{
1173 struct list_head list
;
1174 char name
[PATH_MAX
];
1177 void btrfs_register_one_device(char *fname
)
1179 struct btrfs_ioctl_vol_args args
;
1184 fd
= open("/dev/btrfs-control", O_RDONLY
);
1186 fprintf(stderr
, "failed to open /dev/btrfs-control "
1187 "skipping device registration: %s\n",
1191 strncpy(args
.name
, fname
, BTRFS_PATH_NAME_MAX
);
1192 args
.name
[BTRFS_PATH_NAME_MAX
-1] = 0;
1193 ret
= ioctl(fd
, BTRFS_IOC_SCAN_DEV
, &args
);
1196 fprintf(stderr
, "ERROR: device scan failed '%s' - %s\n",
1197 fname
, strerror(e
));
1202 int btrfs_scan_one_dir(char *dirname
, int run_ioctl
)
1205 struct dirent
*dirent
;
1206 struct pending_dir
*pending
;
1212 struct list_head pending_list
;
1213 struct btrfs_fs_devices
*tmp_devices
;
1216 INIT_LIST_HEAD(&pending_list
);
1218 pending
= malloc(sizeof(*pending
));
1221 strcpy(pending
->name
, dirname
);
1224 dirname_len
= strlen(pending
->name
);
1225 fullpath
= malloc(PATH_MAX
);
1226 dirname
= pending
->name
;
1232 dirp
= opendir(dirname
);
1234 fprintf(stderr
, "Unable to open %s for scanning\n", dirname
);
1239 dirent
= readdir(dirp
);
1242 if (dirent
->d_name
[0] == '.')
1244 if (dirname_len
+ strlen(dirent
->d_name
) + 2 > PATH_MAX
) {
1248 snprintf(fullpath
, PATH_MAX
, "%s/%s", dirname
, dirent
->d_name
);
1249 ret
= lstat(fullpath
, &st
);
1251 fprintf(stderr
, "failed to stat %s\n", fullpath
);
1254 if (S_ISLNK(st
.st_mode
))
1256 if (S_ISDIR(st
.st_mode
)) {
1257 struct pending_dir
*next
= malloc(sizeof(*next
));
1262 strcpy(next
->name
, fullpath
);
1263 list_add_tail(&next
->list
, &pending_list
);
1265 if (!S_ISBLK(st
.st_mode
)) {
1268 fd
= open(fullpath
, O_RDONLY
);
1270 /* ignore the following errors:
1271 ENXIO (device don't exists)
1272 ENOMEDIUM (No medium found ->
1273 like a cd tray empty)
1275 if(errno
!= ENXIO
&& errno
!= ENOMEDIUM
)
1276 fprintf(stderr
, "failed to read %s: %s\n",
1277 fullpath
, strerror(errno
));
1280 ret
= btrfs_scan_one_device(fd
, fullpath
, &tmp_devices
,
1282 BTRFS_SUPER_INFO_OFFSET
);
1283 if (ret
== 0 && run_ioctl
> 0) {
1284 btrfs_register_one_device(fullpath
);
1288 if (!list_empty(&pending_list
)) {
1290 pending
= list_entry(pending_list
.next
, struct pending_dir
,
1293 list_del(&pending
->list
);
1302 while (!list_empty(&pending_list
)) {
1303 pending
= list_entry(pending_list
.next
, struct pending_dir
,
1305 list_del(&pending
->list
);
1313 int btrfs_scan_for_fsid(int run_ioctls
)
1317 ret
= scan_for_btrfs(BTRFS_SCAN_PROC
, run_ioctls
);
1319 ret
= scan_for_btrfs(BTRFS_SCAN_DEV
, run_ioctls
);
1323 int btrfs_device_already_in_root(struct btrfs_root
*root
, int fd
,
1326 struct btrfs_super_block
*disk_super
;
1330 buf
= malloc(BTRFS_SUPER_INFO_SIZE
);
1335 ret
= pread(fd
, buf
, BTRFS_SUPER_INFO_SIZE
, super_offset
);
1336 if (ret
!= BTRFS_SUPER_INFO_SIZE
)
1340 disk_super
= (struct btrfs_super_block
*)buf
;
1341 if (btrfs_super_magic(disk_super
) != BTRFS_MAGIC
)
1344 if (!memcmp(disk_super
->fsid
, root
->fs_info
->super_copy
->fsid
,
1353 static char *size_strs
[] = { "", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB"};
1354 int pretty_size_snprintf(u64 size
, char *str
, size_t str_bytes
)
1366 u64 last_size
= size
;
1368 while(size
>= 1024){
1374 if (num_divs
>= ARRAY_SIZE(size_strs
)) {
1378 fraction
= (float)last_size
/ 1024;
1380 return snprintf(str
, str_bytes
, "%.2f%s", fraction
,
1381 size_strs
[num_divs
]);
1385 * __strncpy__null - strncpy with null termination
1386 * @dest: the target array
1387 * @src: the source string
1388 * @n: maximum bytes to copy (size of *dest)
1390 * Like strncpy, but ensures destination is null-terminated.
1392 * Copies the string pointed to by src, including the terminating null
1393 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
1394 * of n bytes. Then ensure that dest is null-terminated.
1396 char *__strncpy__null(char *dest
, const char *src
, size_t n
)
1398 strncpy(dest
, src
, n
);
1405 * Checks to make sure that the label matches our requirements.
1407 0 if everything is safe and usable
1408 -1 if the label is too long
1410 static int check_label(const char *input
)
1412 int len
= strlen(input
);
1414 if (len
> BTRFS_LABEL_SIZE
- 1) {
1415 fprintf(stderr
, "ERROR: Label %s is too long (max %d)\n",
1416 input
, BTRFS_LABEL_SIZE
- 1);
1423 static int set_label_unmounted(const char *dev
, const char *label
)
1425 struct btrfs_trans_handle
*trans
;
1426 struct btrfs_root
*root
;
1429 ret
= check_mounted(dev
);
1431 fprintf(stderr
, "FATAL: error checking %s mount status\n", dev
);
1435 fprintf(stderr
, "ERROR: dev %s is mounted, use mount point\n",
1440 /* Open the super_block at the default location
1441 * and as read-write.
1443 root
= open_ctree(dev
, 0, OPEN_CTREE_WRITES
);
1444 if (!root
) /* errors are printed by open_ctree() */
1447 trans
= btrfs_start_transaction(root
, 1);
1448 snprintf(root
->fs_info
->super_copy
->label
, BTRFS_LABEL_SIZE
, "%s",
1450 btrfs_commit_transaction(trans
, root
);
1452 /* Now we close it since we are done. */
1457 static int set_label_mounted(const char *mount_path
, const char *label
)
1461 fd
= open(mount_path
, O_RDONLY
| O_NOATIME
);
1463 fprintf(stderr
, "ERROR: unable to access '%s'\n", mount_path
);
1467 if (ioctl(fd
, BTRFS_IOC_SET_FSLABEL
, label
) < 0) {
1468 fprintf(stderr
, "ERROR: unable to set label %s\n",
1478 static int get_label_unmounted(const char *dev
, char *label
)
1480 struct btrfs_root
*root
;
1483 ret
= check_mounted(dev
);
1485 fprintf(stderr
, "FATAL: error checking %s mount status\n", dev
);
1489 fprintf(stderr
, "ERROR: dev %s is mounted, use mount point\n",
1494 /* Open the super_block at the default location
1497 root
= open_ctree(dev
, 0, 0);
1501 memcpy(label
, root
->fs_info
->super_copy
->label
, BTRFS_LABEL_SIZE
);
1503 /* Now we close it since we are done. */
1509 * If a partition is mounted, try to get the filesystem label via its
1510 * mounted path rather than device. Return the corresponding error
1511 * the user specified the device path.
1513 int get_label_mounted(const char *mount_path
, char *labelp
)
1515 char label
[BTRFS_LABEL_SIZE
];
1518 fd
= open(mount_path
, O_RDONLY
| O_NOATIME
);
1520 fprintf(stderr
, "ERROR: unable to access '%s'\n", mount_path
);
1524 memset(label
, '\0', sizeof(label
));
1525 if (ioctl(fd
, BTRFS_IOC_GET_FSLABEL
, label
) < 0) {
1526 fprintf(stderr
, "ERROR: unable get label %s\n", strerror(errno
));
1531 strncpy(labelp
, label
, sizeof(label
));
1536 int get_label(const char *btrfs_dev
, char *label
)
1540 if (is_existing_blk_or_reg_file(btrfs_dev
))
1541 ret
= get_label_unmounted(btrfs_dev
, label
);
1543 ret
= get_label_mounted(btrfs_dev
, label
);
1548 int set_label(const char *btrfs_dev
, const char *label
)
1550 if (check_label(label
))
1553 return is_existing_blk_or_reg_file(btrfs_dev
) ?
1554 set_label_unmounted(btrfs_dev
, label
) :
1555 set_label_mounted(btrfs_dev
, label
);
1558 int btrfs_scan_block_devices(int run_ioctl
)
1564 struct btrfs_fs_devices
*tmp_devices
;
1566 FILE *proc_partitions
;
1574 proc_partitions
= fopen("/proc/partitions","r");
1575 if (!proc_partitions
) {
1576 fprintf(stderr
, "Unable to open '/proc/partitions' for scanning\n");
1579 /* skip the header */
1580 for (i
= 0; i
< 2; i
++)
1581 if (!fgets(buf
, 1023, proc_partitions
)) {
1583 "Unable to read '/proc/partitions' for scanning\n");
1584 fclose(proc_partitions
);
1588 strcpy(fullpath
,"/dev/");
1589 while(fgets(buf
, 1023, proc_partitions
)) {
1590 i
= sscanf(buf
," %*d %*d %*d %99s", fullpath
+5);
1593 * multipath and MD devices may register as a btrfs filesystem
1594 * both through the original block device and through
1595 * the special (/dev/mapper or /dev/mdX) entry.
1596 * This scans the special entries last
1598 special
= strncmp(fullpath
, "/dev/dm-", strlen("/dev/dm-")) == 0;
1600 special
= strncmp(fullpath
, "/dev/md", strlen("/dev/md")) == 0;
1602 if (scans
== 0 && special
)
1604 if (scans
> 0 && !special
)
1607 ret
= lstat(fullpath
, &st
);
1609 fprintf(stderr
, "failed to stat %s\n", fullpath
);
1612 if (!S_ISBLK(st
.st_mode
)) {
1616 fd
= open(fullpath
, O_RDONLY
);
1618 if (errno
!= ENOMEDIUM
)
1619 fprintf(stderr
, "failed to open %s: %s\n",
1620 fullpath
, strerror(errno
));
1623 ret
= btrfs_scan_one_device(fd
, fullpath
, &tmp_devices
,
1625 BTRFS_SUPER_INFO_OFFSET
);
1626 if (ret
== 0 && run_ioctl
> 0) {
1627 btrfs_register_one_device(fullpath
);
1632 fclose(proc_partitions
);
1642 * A not-so-good version fls64. No fascinating optimization since
1643 * no one except parse_size use it
1645 static int fls64(u64 x
)
1649 for (i
= 0; i
<64; i
++)
1650 if (x
<< i
& (1UL << 63))
1655 u64
parse_size(char *s
)
1663 fprintf(stderr
, "ERROR: Size value is empty\n");
1668 "ERROR: Size value '%s' is less equal than 0\n", s
);
1671 ret
= strtoull(s
, &endptr
, 10);
1673 fprintf(stderr
, "ERROR: Size value '%s' is invalid\n", s
);
1676 if (endptr
[0] && endptr
[1]) {
1677 fprintf(stderr
, "ERROR: Illegal suffix contains character '%c' in wrong position\n",
1682 * strtoll returns LLONG_MAX when overflow, if this happens,
1683 * need to call strtoull to get the real size
1685 if (errno
== ERANGE
&& ret
== ULLONG_MAX
) {
1687 "ERROR: Size value '%s' is too large for u64\n", s
);
1691 c
= tolower(endptr
[0]);
1714 fprintf(stderr
, "ERROR: Unknown size descriptor '%c'\n",
1719 /* Check whether ret * mult overflow */
1720 if (fls64(ret
) + fls64(mult
) - 1 > 64) {
1722 "ERROR: Size value '%s' is too large for u64\n", s
);
1729 int open_file_or_dir3(const char *fname
, DIR **dirstream
, int open_flags
)
1735 ret
= stat(fname
, &st
);
1739 if (S_ISDIR(st
.st_mode
)) {
1740 *dirstream
= opendir(fname
);
1743 fd
= dirfd(*dirstream
);
1744 } else if (S_ISREG(st
.st_mode
) || S_ISLNK(st
.st_mode
)) {
1745 fd
= open(fname
, open_flags
);
1748 * we set this on purpose, in case the caller output
1749 * strerror(errno) as success
1757 closedir(*dirstream
);
1762 int open_file_or_dir(const char *fname
, DIR **dirstream
)
1764 return open_file_or_dir3(fname
, dirstream
, O_RDWR
);
1767 void close_file_or_dir(int fd
, DIR *dirstream
)
1770 closedir(dirstream
);
1775 int get_device_info(int fd
, u64 devid
,
1776 struct btrfs_ioctl_dev_info_args
*di_args
)
1780 di_args
->devid
= devid
;
1781 memset(&di_args
->uuid
, '\0', sizeof(di_args
->uuid
));
1783 ret
= ioctl(fd
, BTRFS_IOC_DEV_INFO
, di_args
);
1784 return ret
? -errno
: 0;
1788 * For a given path, fill in the ioctl fs_ and info_ args.
1789 * If the path is a btrfs mountpoint, fill info for all devices.
1790 * If the path is a btrfs device, fill in only that device.
1792 * The path provided must be either on a mounted btrfs fs,
1793 * or be a mounted btrfs device.
1795 * Returns 0 on success, or a negative errno.
1797 int get_fs_info(char *path
, struct btrfs_ioctl_fs_info_args
*fi_args
,
1798 struct btrfs_ioctl_dev_info_args
**di_ret
)
1804 struct btrfs_fs_devices
*fs_devices_mnt
= NULL
;
1805 struct btrfs_ioctl_dev_info_args
*di_args
;
1806 char mp
[BTRFS_PATH_NAME_MAX
+ 1];
1807 DIR *dirstream
= NULL
;
1809 memset(fi_args
, 0, sizeof(*fi_args
));
1811 if (is_block_device(path
)) {
1812 struct btrfs_super_block
*disk_super
;
1813 char buf
[BTRFS_SUPER_INFO_SIZE
];
1816 /* Ensure it's mounted, then set path to the mountpoint */
1817 fd
= open(path
, O_RDONLY
);
1820 fprintf(stderr
, "Couldn't open %s: %s\n",
1821 path
, strerror(errno
));
1824 ret
= check_mounted_where(fd
, path
, mp
, sizeof(mp
),
1833 /* Only fill in this one device */
1834 fi_args
->num_devices
= 1;
1836 disk_super
= (struct btrfs_super_block
*)buf
;
1837 ret
= btrfs_read_dev_super(fd
, disk_super
, BTRFS_SUPER_INFO_OFFSET
);
1842 devid
= btrfs_stack_device_id(&disk_super
->dev_item
);
1844 fi_args
->max_id
= devid
;
1847 memcpy(fi_args
->fsid
, fs_devices_mnt
->fsid
, BTRFS_FSID_SIZE
);
1851 /* at this point path must not be for a block device */
1852 fd
= open_file_or_dir(path
, &dirstream
);
1858 /* fill in fi_args if not just a single device */
1859 if (fi_args
->num_devices
!= 1) {
1860 ret
= ioctl(fd
, BTRFS_IOC_FS_INFO
, fi_args
);
1867 if (!fi_args
->num_devices
)
1870 di_args
= *di_ret
= malloc(fi_args
->num_devices
* sizeof(*di_args
));
1876 for (; i
<= fi_args
->max_id
; ++i
) {
1877 BUG_ON(ndevs
>= fi_args
->num_devices
);
1878 ret
= get_device_info(fd
, i
, &di_args
[ndevs
]);
1887 * only when the only dev we wanted to find is not there then
1888 * let any error be returned
1890 if (fi_args
->num_devices
!= 1) {
1896 close_file_or_dir(fd
, dirstream
);
1900 #define isoctal(c) (((c) & ~7) == '0')
1902 static inline void translate(char *f
, char *t
)
1904 while (*f
!= '\0') {
1906 isoctal(f
[1]) && isoctal(f
[2]) && isoctal(f
[3])) {
1907 *t
++ = 64*(f
[1] & 7) + 8*(f
[2] & 7) + (f
[3] & 7);
1917 * Checks if the swap device.
1918 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
1920 static int is_swap_device(const char *file
)
1931 if (stat(file
, &st_buf
) < 0)
1933 if (S_ISBLK(st_buf
.st_mode
))
1934 dev
= st_buf
.st_rdev
;
1935 else if (S_ISREG(st_buf
.st_mode
)) {
1936 dev
= st_buf
.st_dev
;
1937 ino
= st_buf
.st_ino
;
1941 if ((f
= fopen("/proc/swaps", "r")) == NULL
)
1944 /* skip the first line */
1945 if (fgets(tmp
, sizeof(tmp
), f
) == NULL
)
1948 while (fgets(tmp
, sizeof(tmp
), f
) != NULL
) {
1949 if ((cp
= strchr(tmp
, ' ')) != NULL
)
1951 if ((cp
= strchr(tmp
, '\t')) != NULL
)
1953 translate(tmp
, buf
);
1954 if (stat(buf
, &st_buf
) != 0)
1956 if (S_ISBLK(st_buf
.st_mode
)) {
1957 if (dev
== st_buf
.st_rdev
) {
1961 } else if (S_ISREG(st_buf
.st_mode
)) {
1962 if (dev
== st_buf
.st_dev
&& ino
== st_buf
.st_ino
) {
1976 * Check for existing filesystem or partition table on device.
1978 * 1 for existing fs or partition
1979 * 0 for nothing found
1980 * -1 for internal error
1987 blkid_probe pr
= NULL
;
1991 if (!device
|| !*device
)
1994 ret
= -1; /* will reset on success of all setup calls */
1996 pr
= blkid_new_probe_from_filename(device
);
2000 size
= blkid_probe_get_size(pr
);
2004 /* nothing to overwrite on a 0-length device */
2010 ret
= blkid_probe_enable_partitions(pr
, 1);
2014 ret
= blkid_do_fullprobe(pr
);
2019 * Blkid returns 1 for nothing found and 0 when it finds a signature,
2020 * but we want the exact opposite, so reverse the return value here.
2022 * In addition print some useful diagnostics about what actually is
2030 if (!blkid_probe_lookup_value(pr
, "TYPE", &type
, NULL
)) {
2032 "%s appears to contain an existing "
2033 "filesystem (%s).\n", device
, type
);
2034 } else if (!blkid_probe_lookup_value(pr
, "PTTYPE", &type
, NULL
)) {
2036 "%s appears to contain a partition "
2037 "table (%s).\n", device
, type
);
2040 "%s appears to contain something weird "
2041 "according to blkid\n", device
);
2047 blkid_free_probe(pr
);
2050 "probe of %s failed, cannot detect "
2051 "existing filesystem.\n", device
);
2055 int test_num_disk_vs_raid(u64 metadata_profile
, u64 data_profile
,
2056 u64 dev_cnt
, int mixed
, char *estr
)
2064 allowed
|= BTRFS_BLOCK_GROUP_RAID10
;
2066 allowed
|= BTRFS_BLOCK_GROUP_RAID6
;
2068 allowed
|= BTRFS_BLOCK_GROUP_RAID0
| BTRFS_BLOCK_GROUP_RAID1
|
2069 BTRFS_BLOCK_GROUP_RAID5
;
2072 allowed
|= BTRFS_BLOCK_GROUP_DUP
;
2075 if (metadata_profile
& ~allowed
) {
2076 snprintf(estr
, sz
, "unable to create FS with metadata "
2077 "profile %llu (have %llu devices)\n",
2078 metadata_profile
, dev_cnt
);
2081 if (data_profile
& ~allowed
) {
2082 snprintf(estr
, sz
, "unable to create FS with data "
2083 "profile %llu (have %llu devices)\n",
2084 metadata_profile
, dev_cnt
);
2088 if (!mixed
&& (data_profile
& BTRFS_BLOCK_GROUP_DUP
)) {
2090 "dup for data is allowed only in mixed mode");
2096 /* Check if disk is suitable for btrfs
2098 * 1: something is wrong, estr provides the error
2101 int test_dev_for_mkfs(char *file
, int force_overwrite
, char *estr
)
2107 ret
= is_swap_device(file
);
2109 snprintf(estr
, sz
, "error checking %s status: %s\n", file
,
2114 snprintf(estr
, sz
, "%s is a swap device\n", file
);
2117 if (!force_overwrite
) {
2118 if (check_overwrite(file
)) {
2119 snprintf(estr
, sz
, "Use the -f option to force overwrite.\n");
2123 ret
= check_mounted(file
);
2125 snprintf(estr
, sz
, "error checking %s mount status\n",
2130 snprintf(estr
, sz
, "%s is mounted\n", file
);
2133 /* check if the device is busy */
2134 fd
= open(file
, O_RDWR
|O_EXCL
);
2136 snprintf(estr
, sz
, "unable to open %s: %s\n", file
,
2140 if (fstat(fd
, &st
)) {
2141 snprintf(estr
, sz
, "unable to stat %s: %s\n", file
,
2146 if (!S_ISBLK(st
.st_mode
)) {
2147 fprintf(stderr
, "'%s' is not a block device\n", file
);
2155 int btrfs_scan_lblkid(int update_kernel
)
2160 struct btrfs_fs_devices
*tmp_devices
;
2161 blkid_dev_iterate iter
= NULL
;
2162 blkid_dev dev
= NULL
;
2163 blkid_cache cache
= NULL
;
2164 char path
[PATH_MAX
];
2166 if (blkid_get_cache(&cache
, 0) < 0) {
2167 printf("ERROR: lblkid cache get failed\n");
2170 blkid_probe_all(cache
);
2171 iter
= blkid_dev_iterate_begin(cache
);
2172 blkid_dev_set_search(iter
, "TYPE", "btrfs");
2173 while (blkid_dev_next(iter
, &dev
) == 0) {
2174 dev
= blkid_verify(cache
, dev
);
2177 /* if we are here its definitely a btrfs disk*/
2178 strncpy(path
, blkid_dev_devname(dev
), PATH_MAX
);
2180 fd
= open(path
, O_RDONLY
);
2182 printf("ERROR: could not open %s\n", path
);
2185 ret
= btrfs_scan_one_device(fd
, path
, &tmp_devices
,
2186 &num_devices
, BTRFS_SUPER_INFO_OFFSET
);
2188 printf("ERROR: could not scan %s\n", path
);
2195 btrfs_register_one_device(path
);
2197 blkid_dev_iterate_end(iter
);
2198 blkid_put_cache(cache
);
2203 * scans devs for the btrfs
2205 int scan_for_btrfs(int where
, int update_kernel
)
2210 case BTRFS_SCAN_PROC
:
2211 ret
= btrfs_scan_block_devices(update_kernel
);
2213 case BTRFS_SCAN_DEV
:
2214 ret
= btrfs_scan_one_dir("/dev", update_kernel
);
2216 case BTRFS_SCAN_LBLKID
:
2217 ret
= btrfs_scan_lblkid(update_kernel
);
2223 int is_vol_small(char *file
)
2230 fd
= open(file
, O_RDONLY
);
2233 if (fstat(fd
, &st
) < 0) {
2238 size
= btrfs_device_size(fd
, &st
);
2243 if (size
< 1024 * 1024 * 1024) {
2253 * This reads a line from the stdin and only returns non-zero if the
2254 * first whitespace delimited token is a case insensitive match with yes
2257 int ask_user(char *question
)
2259 char buf
[30] = {0,};
2260 char *saveptr
= NULL
;
2263 printf("%s [y/N]: ", question
);
2265 return fgets(buf
, sizeof(buf
) - 1, stdin
) &&
2266 (answer
= strtok_r(buf
, " \t\n\r", &saveptr
)) &&
2267 (!strcasecmp(answer
, "yes") || !strcasecmp(answer
, "y"));
2272 * - file or directory return the containing tree root id
2273 * - subvolume return it's own tree id
2274 * - BTRFS_EMPTY_SUBVOL_DIR_OBJECTID (directory with ino == 2) the result is
2275 * undefined and function returns -1
2277 int lookup_ino_rootid(int fd
, u64
*rootid
)
2279 struct btrfs_ioctl_ino_lookup_args args
;
2283 memset(&args
, 0, sizeof(args
));
2285 args
.objectid
= BTRFS_FIRST_FREE_OBJECTID
;
2287 ret
= ioctl(fd
, BTRFS_IOC_INO_LOOKUP
, &args
);
2290 fprintf(stderr
, "ERROR: Failed to lookup root id - %s\n",
2295 *rootid
= args
.treeid
;
2300 int find_mount_root(const char *path
, char **mount_root
)
2307 int longest_matchlen
= 0;
2308 char *longest_match
= NULL
;
2310 fd
= open(path
, O_RDONLY
| O_NOATIME
);
2315 mnttab
= setmntent("/proc/self/mounts", "r");
2319 while ((ent
= getmntent(mnttab
))) {
2320 len
= strlen(ent
->mnt_dir
);
2321 if (strncmp(ent
->mnt_dir
, path
, len
) == 0) {
2323 if (longest_matchlen
< len
) {
2324 free(longest_match
);
2325 longest_matchlen
= len
;
2326 longest_match
= strdup(ent
->mnt_dir
);
2332 if (!longest_match
) {
2334 "ERROR: Failed to find mount root for path %s.\n",
2340 *mount_root
= realpath(longest_match
, NULL
);
2344 free(longest_match
);