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)
55 static int btrfs_scan_done = 0;
57 static char argv0_buf[ARGV0_BUF_SIZE] = "btrfs";
59 void fixup_argv0(char **argv, const char *token)
61 int len = strlen(argv0_buf);
63 snprintf(argv0_buf + len, sizeof(argv0_buf) - len, " %s", token);
67 void set_argv0(char **argv)
69 sprintf(argv0_buf, "%s", argv[0]);
72 int check_argc_exact(int nargs, int expected)
75 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
77 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
79 return nargs != expected;
82 int check_argc_min(int nargs, int expected)
84 if (nargs < expected) {
85 fprintf(stderr, "%s: too few arguments\n", argv0_buf);
92 int check_argc_max(int nargs, int expected)
94 if (nargs > expected) {
95 fprintf(stderr, "%s: too many arguments\n", argv0_buf);
104 * Discard the given range in one go
106 static int discard_range(int fd, u64 start, u64 len)
108 u64 range[2] = { start, len };
110 if (ioctl(fd, BLKDISCARD, &range) < 0)
116 * Discard blocks in the given range in 1G chunks, the process is interruptible
118 static int discard_blocks(int fd, u64 start, u64 len)
122 u64 chunk_size = min_t(u64, len, 1*1024*1024*1024);
125 ret = discard_range(fd, start, chunk_size);
135 static u64 reference_root_table[] = {
136 [1] = BTRFS_ROOT_TREE_OBJECTID,
137 [2] = BTRFS_EXTENT_TREE_OBJECTID,
138 [3] = BTRFS_CHUNK_TREE_OBJECTID,
139 [4] = BTRFS_DEV_TREE_OBJECTID,
140 [5] = BTRFS_FS_TREE_OBJECTID,
141 [6] = BTRFS_CSUM_TREE_OBJECTID,
144 int test_uuid_unique(char *fs_uuid)
147 blkid_dev_iterate iter = NULL;
148 blkid_dev dev = NULL;
149 blkid_cache cache = NULL;
151 if (blkid_get_cache(&cache, 0) < 0) {
152 printf("ERROR: lblkid cache get failed\n");
155 blkid_probe_all(cache);
156 iter = blkid_dev_iterate_begin(cache);
157 blkid_dev_set_search(iter, "UUID", fs_uuid);
159 while (blkid_dev_next(iter, &dev) == 0) {
160 dev = blkid_verify(cache, dev);
167 blkid_dev_iterate_end(iter);
168 blkid_put_cache(cache);
173 int make_btrfs(int fd, const char *device, const char *label, char *fs_uuid,
174 u64 blocks[7], u64 num_bytes, u32 nodesize,
175 u32 leafsize, u32 sectorsize, u32 stripesize, u64 features)
177 struct btrfs_super_block super;
178 struct extent_buffer *buf = NULL;
179 struct btrfs_root_item root_item;
180 struct btrfs_disk_key disk_key;
181 struct btrfs_extent_item *extent_item;
182 struct btrfs_inode_item *inode_item;
183 struct btrfs_chunk *chunk;
184 struct btrfs_dev_item *dev_item;
185 struct btrfs_dev_extent *dev_extent;
186 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
196 int skinny_metadata = !!(features &
197 BTRFS_FEATURE_INCOMPAT_SKINNY_METADATA);
199 first_free = BTRFS_SUPER_INFO_OFFSET + sectorsize * 2 - 1;
200 first_free &= ~((u64)sectorsize - 1);
202 memset(&super, 0, sizeof(super));
204 num_bytes = (num_bytes / sectorsize) * sectorsize;
206 if (uuid_parse(fs_uuid, super.fsid) != 0) {
207 fprintf(stderr, "could not parse UUID: %s\n", fs_uuid);
211 if (!test_uuid_unique(fs_uuid)) {
212 fprintf(stderr, "non-unique UUID: %s\n", fs_uuid);
217 uuid_generate(super.fsid);
219 uuid_generate(super.dev_item.uuid);
220 uuid_generate(chunk_tree_uuid);
222 btrfs_set_super_bytenr(&super, blocks[0]);
223 btrfs_set_super_num_devices(&super, 1);
224 btrfs_set_super_magic(&super, BTRFS_MAGIC);
225 btrfs_set_super_generation(&super, 1);
226 btrfs_set_super_root(&super, blocks[1]);
227 btrfs_set_super_chunk_root(&super, blocks[3]);
228 btrfs_set_super_total_bytes(&super, num_bytes);
229 btrfs_set_super_bytes_used(&super, 6 * leafsize);
230 btrfs_set_super_sectorsize(&super, sectorsize);
231 btrfs_set_super_leafsize(&super, leafsize);
232 btrfs_set_super_nodesize(&super, nodesize);
233 btrfs_set_super_stripesize(&super, stripesize);
234 btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
235 btrfs_set_super_chunk_root_generation(&super, 1);
236 btrfs_set_super_cache_generation(&super, -1);
237 btrfs_set_super_incompat_flags(&super, features);
239 strncpy(super.label, label, BTRFS_LABEL_SIZE - 1);
241 buf = malloc(sizeof(*buf) + max(sectorsize, leafsize));
243 /* create the tree of root objects */
244 memset(buf->data, 0, leafsize);
246 btrfs_set_header_bytenr(buf, blocks[1]);
247 btrfs_set_header_nritems(buf, 4);
248 btrfs_set_header_generation(buf, 1);
249 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
250 btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
251 write_extent_buffer(buf, super.fsid, btrfs_header_fsid(),
254 write_extent_buffer(buf, chunk_tree_uuid,
255 btrfs_header_chunk_tree_uuid(buf),
258 /* create the items for the root tree */
259 memset(&root_item, 0, sizeof(root_item));
260 inode_item = &root_item.inode;
261 btrfs_set_stack_inode_generation(inode_item, 1);
262 btrfs_set_stack_inode_size(inode_item, 3);
263 btrfs_set_stack_inode_nlink(inode_item, 1);
264 btrfs_set_stack_inode_nbytes(inode_item, leafsize);
265 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
266 btrfs_set_root_refs(&root_item, 1);
267 btrfs_set_root_used(&root_item, leafsize);
268 btrfs_set_root_generation(&root_item, 1);
270 memset(&disk_key, 0, sizeof(disk_key));
271 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
272 btrfs_set_disk_key_offset(&disk_key, 0);
275 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize) - sizeof(root_item);
276 btrfs_set_root_bytenr(&root_item, blocks[2]);
277 btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
278 btrfs_set_item_key(buf, &disk_key, nritems);
279 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
280 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
282 write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
283 nritems), sizeof(root_item));
286 itemoff = itemoff - sizeof(root_item);
287 btrfs_set_root_bytenr(&root_item, blocks[4]);
288 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
289 btrfs_set_item_key(buf, &disk_key, nritems);
290 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
291 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
293 write_extent_buffer(buf, &root_item,
294 btrfs_item_ptr_offset(buf, nritems),
298 itemoff = itemoff - sizeof(root_item);
299 btrfs_set_root_bytenr(&root_item, blocks[5]);
300 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
301 btrfs_set_item_key(buf, &disk_key, nritems);
302 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
303 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
305 write_extent_buffer(buf, &root_item,
306 btrfs_item_ptr_offset(buf, nritems),
310 itemoff = itemoff - sizeof(root_item);
311 btrfs_set_root_bytenr(&root_item, blocks[6]);
312 btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_TREE_OBJECTID);
313 btrfs_set_item_key(buf, &disk_key, nritems);
314 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
315 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
317 write_extent_buffer(buf, &root_item,
318 btrfs_item_ptr_offset(buf, nritems),
323 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
324 ret = pwrite(fd, buf->data, leafsize, blocks[1]);
325 if (ret != leafsize) {
326 ret = (ret < 0 ? -errno : -EIO);
330 /* create the items for the extent tree */
331 memset(buf->data+sizeof(struct btrfs_header), 0,
332 leafsize-sizeof(struct btrfs_header));
334 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize);
335 for (i = 1; i < 7; i++) {
336 item_size = sizeof(struct btrfs_extent_item);
337 if (!skinny_metadata)
338 item_size += sizeof(struct btrfs_tree_block_info);
340 BUG_ON(blocks[i] < first_free);
341 BUG_ON(blocks[i] < blocks[i - 1]);
343 /* create extent item */
344 itemoff -= item_size;
345 btrfs_set_disk_key_objectid(&disk_key, blocks[i]);
346 if (skinny_metadata) {
347 btrfs_set_disk_key_type(&disk_key,
348 BTRFS_METADATA_ITEM_KEY);
349 btrfs_set_disk_key_offset(&disk_key, 0);
351 btrfs_set_disk_key_type(&disk_key,
352 BTRFS_EXTENT_ITEM_KEY);
353 btrfs_set_disk_key_offset(&disk_key, leafsize);
355 btrfs_set_item_key(buf, &disk_key, nritems);
356 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
358 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
360 extent_item = btrfs_item_ptr(buf, nritems,
361 struct btrfs_extent_item);
362 btrfs_set_extent_refs(buf, extent_item, 1);
363 btrfs_set_extent_generation(buf, extent_item, 1);
364 btrfs_set_extent_flags(buf, extent_item,
365 BTRFS_EXTENT_FLAG_TREE_BLOCK);
368 /* create extent ref */
369 ref_root = reference_root_table[i];
370 btrfs_set_disk_key_objectid(&disk_key, blocks[i]);
371 btrfs_set_disk_key_offset(&disk_key, ref_root);
372 btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
373 btrfs_set_item_key(buf, &disk_key, nritems);
374 btrfs_set_item_offset(buf, btrfs_item_nr(nritems),
376 btrfs_set_item_size(buf, btrfs_item_nr(nritems), 0);
379 btrfs_set_header_bytenr(buf, blocks[2]);
380 btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
381 btrfs_set_header_nritems(buf, nritems);
382 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
383 ret = pwrite(fd, buf->data, leafsize, blocks[2]);
384 if (ret != leafsize) {
385 ret = (ret < 0 ? -errno : -EIO);
389 /* create the chunk tree */
390 memset(buf->data+sizeof(struct btrfs_header), 0,
391 leafsize-sizeof(struct btrfs_header));
393 item_size = sizeof(*dev_item);
394 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize) - item_size;
396 /* first device 1 (there is no device 0) */
397 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
398 btrfs_set_disk_key_offset(&disk_key, 1);
399 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
400 btrfs_set_item_key(buf, &disk_key, nritems);
401 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
402 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
404 dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
405 btrfs_set_device_id(buf, dev_item, 1);
406 btrfs_set_device_generation(buf, dev_item, 0);
407 btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
408 btrfs_set_device_bytes_used(buf, dev_item,
409 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
410 btrfs_set_device_io_align(buf, dev_item, sectorsize);
411 btrfs_set_device_io_width(buf, dev_item, sectorsize);
412 btrfs_set_device_sector_size(buf, dev_item, sectorsize);
413 btrfs_set_device_type(buf, dev_item, 0);
415 write_extent_buffer(buf, super.dev_item.uuid,
416 (unsigned long)btrfs_device_uuid(dev_item),
418 write_extent_buffer(buf, super.fsid,
419 (unsigned long)btrfs_device_fsid(dev_item),
421 read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
425 item_size = btrfs_chunk_item_size(1);
426 itemoff = itemoff - item_size;
428 /* then we have chunk 0 */
429 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
430 btrfs_set_disk_key_offset(&disk_key, 0);
431 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
432 btrfs_set_item_key(buf, &disk_key, nritems);
433 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
434 btrfs_set_item_size(buf, btrfs_item_nr(nritems), item_size);
436 chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
437 btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
438 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
439 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
440 btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
441 btrfs_set_chunk_io_align(buf, chunk, sectorsize);
442 btrfs_set_chunk_io_width(buf, chunk, sectorsize);
443 btrfs_set_chunk_sector_size(buf, chunk, sectorsize);
444 btrfs_set_chunk_num_stripes(buf, chunk, 1);
445 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
446 btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
449 write_extent_buffer(buf, super.dev_item.uuid,
450 (unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
453 /* copy the key for the chunk to the system array */
454 ptr = super.sys_chunk_array;
455 array_size = sizeof(disk_key);
457 memcpy(ptr, &disk_key, sizeof(disk_key));
458 ptr += sizeof(disk_key);
460 /* copy the chunk to the system array */
461 read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
462 array_size += item_size;
464 btrfs_set_super_sys_array_size(&super, array_size);
466 btrfs_set_header_bytenr(buf, blocks[3]);
467 btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
468 btrfs_set_header_nritems(buf, nritems);
469 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
470 ret = pwrite(fd, buf->data, leafsize, blocks[3]);
471 if (ret != leafsize) {
472 ret = (ret < 0 ? -errno : -EIO);
476 /* create the device tree */
477 memset(buf->data+sizeof(struct btrfs_header), 0,
478 leafsize-sizeof(struct btrfs_header));
480 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize) -
481 sizeof(struct btrfs_dev_extent);
483 btrfs_set_disk_key_objectid(&disk_key, 1);
484 btrfs_set_disk_key_offset(&disk_key, 0);
485 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
486 btrfs_set_item_key(buf, &disk_key, nritems);
487 btrfs_set_item_offset(buf, btrfs_item_nr(nritems), itemoff);
488 btrfs_set_item_size(buf, btrfs_item_nr(nritems),
489 sizeof(struct btrfs_dev_extent));
490 dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
491 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
492 BTRFS_CHUNK_TREE_OBJECTID);
493 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
494 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
495 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
497 write_extent_buffer(buf, chunk_tree_uuid,
498 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
501 btrfs_set_dev_extent_length(buf, dev_extent,
502 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
505 btrfs_set_header_bytenr(buf, blocks[4]);
506 btrfs_set_header_owner(buf, BTRFS_DEV_TREE_OBJECTID);
507 btrfs_set_header_nritems(buf, nritems);
508 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
509 ret = pwrite(fd, buf->data, leafsize, blocks[4]);
510 if (ret != leafsize) {
511 ret = (ret < 0 ? -errno : -EIO);
515 /* create the FS root */
516 memset(buf->data+sizeof(struct btrfs_header), 0,
517 leafsize-sizeof(struct btrfs_header));
518 btrfs_set_header_bytenr(buf, blocks[5]);
519 btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
520 btrfs_set_header_nritems(buf, 0);
521 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
522 ret = pwrite(fd, buf->data, leafsize, blocks[5]);
523 if (ret != leafsize) {
524 ret = (ret < 0 ? -errno : -EIO);
527 /* finally create the csum root */
528 memset(buf->data+sizeof(struct btrfs_header), 0,
529 leafsize-sizeof(struct btrfs_header));
530 btrfs_set_header_bytenr(buf, blocks[6]);
531 btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
532 btrfs_set_header_nritems(buf, 0);
533 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
534 ret = pwrite(fd, buf->data, leafsize, blocks[6]);
535 if (ret != leafsize) {
536 ret = (ret < 0 ? -errno : -EIO);
540 /* and write out the super block */
541 BUG_ON(sizeof(super) > sectorsize);
542 memset(buf->data, 0, sectorsize);
543 memcpy(buf->data, &super, sizeof(super));
544 buf->len = sectorsize;
545 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
546 ret = pwrite(fd, buf->data, sectorsize, blocks[0]);
547 if (ret != sectorsize) {
548 ret = (ret < 0 ? -errno : -EIO);
559 u64 btrfs_device_size(int fd, struct stat *st)
562 if (S_ISREG(st->st_mode)) {
565 if (!S_ISBLK(st->st_mode)) {
568 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
574 static int zero_blocks(int fd, off_t start, size_t len)
576 char *buf = malloc(len);
583 written = pwrite(fd, buf, len, start);
590 #define ZERO_DEV_BYTES (2 * 1024 * 1024)
592 /* don't write outside the device by clamping the region to the device size */
593 static int zero_dev_clamped(int fd, off_t start, ssize_t len, u64 dev_size)
595 off_t end = max(start, start + len);
598 /* and don't overwrite the disk labels on sparc */
599 start = max(start, 1024);
600 end = max(end, 1024);
603 start = min_t(u64, start, dev_size);
604 end = min_t(u64, end, dev_size);
606 return zero_blocks(fd, start, end - start);
609 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
610 struct btrfs_root *root, int fd, char *path,
611 u64 block_count, u32 io_width, u32 io_align,
614 struct btrfs_super_block *disk_super;
615 struct btrfs_super_block *super = root->fs_info->super_copy;
616 struct btrfs_device *device;
617 struct btrfs_dev_item *dev_item;
623 device = kzalloc(sizeof(*device), GFP_NOFS);
626 buf = kmalloc(sectorsize, GFP_NOFS);
631 BUG_ON(sizeof(*disk_super) > sectorsize);
632 memset(buf, 0, sectorsize);
634 disk_super = (struct btrfs_super_block *)buf;
635 dev_item = &disk_super->dev_item;
637 uuid_generate(device->uuid);
640 device->io_width = io_width;
641 device->io_align = io_align;
642 device->sector_size = sectorsize;
644 device->writeable = 1;
645 device->total_bytes = block_count;
646 device->bytes_used = 0;
647 device->total_ios = 0;
648 device->dev_root = root->fs_info->dev_root;
650 ret = btrfs_add_device(trans, root, device);
653 total_bytes = btrfs_super_total_bytes(super) + block_count;
654 btrfs_set_super_total_bytes(super, total_bytes);
656 num_devs = btrfs_super_num_devices(super) + 1;
657 btrfs_set_super_num_devices(super, num_devs);
659 memcpy(disk_super, super, sizeof(*disk_super));
661 printf("adding device %s id %llu\n", path,
662 (unsigned long long)device->devid);
664 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
665 btrfs_set_stack_device_id(dev_item, device->devid);
666 btrfs_set_stack_device_type(dev_item, device->type);
667 btrfs_set_stack_device_io_align(dev_item, device->io_align);
668 btrfs_set_stack_device_io_width(dev_item, device->io_width);
669 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
670 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
671 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
672 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
674 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
675 BUG_ON(ret != sectorsize);
678 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
679 device->fs_devices = root->fs_info->fs_devices;
683 static void btrfs_wipe_existing_sb(int fd)
685 const char *off = NULL;
690 blkid_probe pr = NULL;
692 pr = blkid_new_probe();
696 if (blkid_probe_set_device(pr, fd, 0, 0))
699 rc = blkid_probe_lookup_value(pr, "SBMAGIC_OFFSET", &off, NULL);
701 rc = blkid_probe_lookup_value(pr, "SBMAGIC", NULL, &len);
703 if (rc || len == 0 || off == NULL)
706 offset = strtoll(off, NULL, 10);
707 if (len > sizeof(buf))
711 rc = pwrite(fd, buf, len, offset);
715 blkid_free_probe(pr);
719 int btrfs_prepare_device(int fd, char *file, int zero_end, u64 *block_count_ret,
720 u64 max_block_count, int *mixed, int discard)
726 ret = fstat(fd, &st);
728 fprintf(stderr, "unable to stat %s\n", file);
732 block_count = btrfs_device_size(fd, &st);
733 if (block_count == 0) {
734 fprintf(stderr, "unable to find %s size\n", file);
738 block_count = min(block_count, max_block_count);
740 if (block_count < BTRFS_MKFS_SMALL_VOLUME_SIZE && !(*mixed))
745 * We intentionally ignore errors from the discard ioctl. It
746 * is not necessary for the mkfs functionality but just an
749 if (discard_range(fd, 0, 0) == 0) {
750 fprintf(stderr, "Performing full device TRIM (%s) ...\n",
751 pretty_size(block_count));
752 discard_blocks(fd, 0, block_count);
756 ret = zero_dev_clamped(fd, 0, ZERO_DEV_BYTES, block_count);
757 for (i = 0 ; !ret && i < BTRFS_SUPER_MIRROR_MAX; i++)
758 ret = zero_dev_clamped(fd, btrfs_sb_offset(i),
759 BTRFS_SUPER_INFO_SIZE, block_count);
760 if (!ret && zero_end)
761 ret = zero_dev_clamped(fd, block_count - ZERO_DEV_BYTES,
762 ZERO_DEV_BYTES, block_count);
765 fprintf(stderr, "ERROR: failed to zero device '%s' - %s\n",
766 file, strerror(-ret));
770 btrfs_wipe_existing_sb(fd);
772 *block_count_ret = block_count;
776 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
777 struct btrfs_root *root, u64 objectid)
780 struct btrfs_inode_item inode_item;
781 time_t now = time(NULL);
783 memset(&inode_item, 0, sizeof(inode_item));
784 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
785 btrfs_set_stack_inode_size(&inode_item, 0);
786 btrfs_set_stack_inode_nlink(&inode_item, 1);
787 btrfs_set_stack_inode_nbytes(&inode_item, root->leafsize);
788 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
789 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
790 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
791 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
792 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
793 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
794 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
795 btrfs_set_stack_timespec_sec(&inode_item.otime, 0);
796 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
798 if (root->fs_info->tree_root == root)
799 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
801 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
805 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
809 btrfs_set_root_dirid(&root->root_item, objectid);
816 * checks if a path is a block device node
817 * Returns negative errno on failure, otherwise
818 * returns 1 for blockdev, 0 for not-blockdev
820 int is_block_device(const char *path)
824 if (stat(path, &statbuf) < 0)
827 return S_ISBLK(statbuf.st_mode);
831 * check if given path is a mount point
832 * return 1 if yes. 0 if no. -1 for error
834 int is_mount_point(const char *path)
840 f = setmntent("/proc/self/mounts", "r");
844 while ((mnt = getmntent(f)) != NULL) {
845 if (strcmp(mnt->mnt_dir, path))
855 * Find the mount point for a mounted device.
856 * On success, returns 0 with mountpoint in *mp.
857 * On failure, returns -errno (not mounted yields -EINVAL)
858 * Is noisy on failures, expects to be given a mounted device.
860 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size)
865 ret = is_block_device(dev);
868 fprintf(stderr, "%s is not a block device\n", dev);
871 fprintf(stderr, "Could not check %s: %s\n",
872 dev, strerror(-ret));
877 fd = open(dev, O_RDONLY);
880 fprintf(stderr, "Could not open %s: %s\n", dev, strerror(errno));
884 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
887 } else { /* mounted, all good */
897 * Given a pathname, return a filehandle to:
898 * the original pathname or,
899 * if the pathname is a mounted btrfs device, to its mountpoint.
901 * On error, return -1, errno should be set.
903 int open_path_or_dev_mnt(const char *path, DIR **dirstream)
905 char mp[BTRFS_PATH_NAME_MAX + 1];
908 if (is_block_device(path)) {
911 ret = get_btrfs_mount(path, mp, sizeof(mp));
913 /* not a mounted btrfs dev */
917 fdmnt = open_file_or_dir(mp, dirstream);
919 fdmnt = open_file_or_dir(path, dirstream);
925 /* checks if a device is a loop device */
926 static int is_loop_device (const char* device) {
929 if(stat(device, &statbuf) < 0)
932 return (S_ISBLK(statbuf.st_mode) &&
933 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
937 /* Takes a loop device path (e.g. /dev/loop0) and returns
938 * the associated file (e.g. /images/my_btrfs.img) */
939 static int resolve_loop_device(const char* loop_dev, char* loop_file,
946 char real_loop_dev[PATH_MAX];
948 if (!realpath(loop_dev, real_loop_dev))
950 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
951 if (!(f = fopen(p, "r")))
954 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
955 ret = fscanf(f, fmt, loop_file);
963 /* Checks whether a and b are identical or device
964 * files associated with the same block device
966 static int is_same_blk_file(const char* a, const char* b)
968 struct stat st_buf_a, st_buf_b;
969 char real_a[PATH_MAX];
970 char real_b[PATH_MAX];
972 if(!realpath(a, real_a))
975 if (!realpath(b, real_b))
978 /* Identical path? */
979 if(strcmp(real_a, real_b) == 0)
982 if(stat(a, &st_buf_a) < 0 ||
983 stat(b, &st_buf_b) < 0)
990 /* Same blockdevice? */
991 if(S_ISBLK(st_buf_a.st_mode) &&
992 S_ISBLK(st_buf_b.st_mode) &&
993 st_buf_a.st_rdev == st_buf_b.st_rdev)
999 if (st_buf_a.st_dev == st_buf_b.st_dev &&
1000 st_buf_a.st_ino == st_buf_b.st_ino)
1008 /* checks if a and b are identical or device
1009 * files associated with the same block device or
1010 * if one file is a loop device that uses the other
1013 static int is_same_loop_file(const char* a, const char* b)
1015 char res_a[PATH_MAX];
1016 char res_b[PATH_MAX];
1017 const char* final_a = NULL;
1018 const char* final_b = NULL;
1021 /* Resolve a if it is a loop device */
1022 if((ret = is_loop_device(a)) < 0) {
1027 ret = resolve_loop_device(a, res_a, sizeof(res_a));
1038 /* Resolve b if it is a loop device */
1039 if ((ret = is_loop_device(b)) < 0) {
1044 ret = resolve_loop_device(b, res_b, sizeof(res_b));
1055 return is_same_blk_file(final_a, final_b);
1058 /* Checks if a file exists and is a block or regular file*/
1059 static int is_existing_blk_or_reg_file(const char* filename)
1063 if(stat(filename, &st_buf) < 0) {
1070 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
1073 /* Checks if a file is used (directly or indirectly via a loop device)
1074 * by a device in fs_devices
1076 static int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices,
1080 struct list_head *head;
1081 struct list_head *cur;
1082 struct btrfs_device *device;
1084 head = &fs_devices->devices;
1085 list_for_each(cur, head) {
1086 device = list_entry(cur, struct btrfs_device, dev_list);
1088 if((ret = is_same_loop_file(device->name, file)))
1096 * Resolve a pathname to a device mapper node to /dev/mapper/<name>
1097 * Returns NULL on invalid input or malloc failure; Other failures
1098 * will be handled by the caller using the input pathame.
1100 char *canonicalize_dm_name(const char *ptname)
1104 char path[PATH_MAX], name[PATH_MAX], *res = NULL;
1106 if (!ptname || !*ptname)
1109 snprintf(path, sizeof(path), "/sys/block/%s/dm/name", ptname);
1110 if (!(f = fopen(path, "r")))
1113 /* read <name>\n from sysfs */
1114 if (fgets(name, sizeof(name), f) && (sz = strlen(name)) > 1) {
1115 name[sz - 1] = '\0';
1116 snprintf(path, sizeof(path), "/dev/mapper/%s", name);
1118 if (access(path, F_OK) == 0)
1126 * Resolve a pathname to a canonical device node, e.g. /dev/sda1 or
1127 * to a device mapper pathname.
1128 * Returns NULL on invalid input or malloc failure; Other failures
1129 * will be handled by the caller using the input pathame.
1131 char *canonicalize_path(const char *path)
1133 char *canonical, *p;
1135 if (!path || !*path)
1138 canonical = realpath(path, NULL);
1140 return strdup(path);
1141 p = strrchr(canonical, '/');
1142 if (p && strncmp(p, "/dm-", 4) == 0 && isdigit(*(p + 4))) {
1143 char *dm = canonicalize_dm_name(p + 1);
1154 * returns 1 if the device was mounted, < 0 on error or 0 if everything
1155 * is safe to continue.
1157 int check_mounted(const char* file)
1162 fd = open(file, O_RDONLY);
1164 fprintf (stderr, "check_mounted(): Could not open %s\n", file);
1168 ret = check_mounted_where(fd, file, NULL, 0, NULL);
1174 int check_mounted_where(int fd, const char *file, char *where, int size,
1175 struct btrfs_fs_devices **fs_dev_ret)
1180 struct btrfs_fs_devices *fs_devices_mnt = NULL;
1184 /* scan the initial device */
1185 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
1186 &total_devs, BTRFS_SUPER_INFO_OFFSET, 0);
1187 is_btrfs = (ret >= 0);
1189 /* scan other devices */
1190 if (is_btrfs && total_devs > 1) {
1191 ret = btrfs_scan_lblkid();
1196 /* iterate over the list of currently mountes filesystems */
1197 if ((f = setmntent ("/proc/self/mounts", "r")) == NULL)
1200 while ((mnt = getmntent (f)) != NULL) {
1202 if(strcmp(mnt->mnt_type, "btrfs") != 0)
1205 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
1207 /* ignore entries in the mount table that are not
1208 associated with a file*/
1209 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
1210 goto out_mntloop_err;
1214 ret = is_same_loop_file(file, mnt->mnt_fsname);
1218 goto out_mntloop_err;
1223 /* Did we find an entry in mnt table? */
1224 if (mnt && size && where) {
1225 strncpy(where, mnt->mnt_dir, size);
1229 *fs_dev_ret = fs_devices_mnt;
1231 ret = (mnt != NULL);
1239 struct pending_dir {
1240 struct list_head list;
1241 char name[PATH_MAX];
1244 int btrfs_register_one_device(const char *fname)
1246 struct btrfs_ioctl_vol_args args;
1251 fd = open("/dev/btrfs-control", O_RDWR);
1253 fprintf(stderr, "failed to open /dev/btrfs-control "
1254 "skipping device registration: %s\n",
1258 strncpy(args.name, fname, BTRFS_PATH_NAME_MAX);
1259 args.name[BTRFS_PATH_NAME_MAX-1] = 0;
1260 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
1263 fprintf(stderr, "ERROR: device scan failed '%s' - %s\n",
1264 fname, strerror(e));
1272 * Register all devices in the fs_uuid list created in the user
1273 * space. Ensure btrfs_scan_lblkid() is called before this func.
1275 int btrfs_register_all_devices(void)
1278 struct btrfs_fs_devices *fs_devices;
1279 struct btrfs_device *device;
1280 struct list_head *all_uuids;
1282 all_uuids = btrfs_scanned_uuids();
1284 list_for_each_entry(fs_devices, all_uuids, list) {
1285 list_for_each_entry(device, &fs_devices->devices, dev_list) {
1286 if (strlen(device->name) != 0) {
1287 err = btrfs_register_one_device(device->name);
1298 int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
1301 struct btrfs_super_block *disk_super;
1305 buf = malloc(BTRFS_SUPER_INFO_SIZE);
1310 ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
1311 if (ret != BTRFS_SUPER_INFO_SIZE)
1315 disk_super = (struct btrfs_super_block *)buf;
1316 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC)
1319 if (!memcmp(disk_super->fsid, root->fs_info->super_copy->fsid,
1328 static const char* unit_suffix_binary[] =
1329 { "B", "KiB", "MiB", "GiB", "TiB", "PiB", "EiB"};
1330 static const char* unit_suffix_decimal[] =
1331 { "B", "kB", "mB", "gB", "tB", "pB", "eB"};
1333 int pretty_size_snprintf(u64 size, char *str, size_t str_size, unsigned unit_mode)
1339 const char** suffix = NULL;
1345 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_RAW) {
1346 snprintf(str, str_size, "%llu", size);
1350 if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_BINARY) {
1353 suffix = unit_suffix_binary;
1354 } else if ((unit_mode & ~UNITS_MODE_MASK) == UNITS_DECIMAL) {
1357 suffix = unit_suffix_decimal;
1362 fprintf(stderr, "INTERNAL ERROR: unknown unit base, mode %d\n",
1370 switch (unit_mode & UNITS_MODE_MASK) {
1371 case UNITS_TBYTES: base *= mult; num_divs++;
1372 case UNITS_GBYTES: base *= mult; num_divs++;
1373 case UNITS_MBYTES: base *= mult; num_divs++;
1374 case UNITS_KBYTES: num_divs++;
1381 while (size >= mult) {
1388 if (num_divs >= ARRAY_SIZE(unit_suffix_binary)) {
1390 printf("INTERNAL ERROR: unsupported unit suffix, index %d\n",
1395 fraction = (float)last_size / base;
1397 return snprintf(str, str_size, "%.2f%s", fraction, suffix[num_divs]);
1401 * __strncpy__null - strncpy with null termination
1402 * @dest: the target array
1403 * @src: the source string
1404 * @n: maximum bytes to copy (size of *dest)
1406 * Like strncpy, but ensures destination is null-terminated.
1408 * Copies the string pointed to by src, including the terminating null
1409 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
1410 * of n bytes. Then ensure that dest is null-terminated.
1412 char *__strncpy__null(char *dest, const char *src, size_t n)
1414 strncpy(dest, src, n);
1421 * Checks to make sure that the label matches our requirements.
1423 0 if everything is safe and usable
1424 -1 if the label is too long
1426 static int check_label(const char *input)
1428 int len = strlen(input);
1430 if (len > BTRFS_LABEL_SIZE - 1) {
1431 fprintf(stderr, "ERROR: Label %s is too long (max %d)\n",
1432 input, BTRFS_LABEL_SIZE - 1);
1439 static int set_label_unmounted(const char *dev, const char *label)
1441 struct btrfs_trans_handle *trans;
1442 struct btrfs_root *root;
1445 ret = check_mounted(dev);
1447 fprintf(stderr, "FATAL: error checking %s mount status\n", dev);
1451 fprintf(stderr, "ERROR: dev %s is mounted, use mount point\n",
1456 /* Open the super_block at the default location
1457 * and as read-write.
1459 root = open_ctree(dev, 0, OPEN_CTREE_WRITES);
1460 if (!root) /* errors are printed by open_ctree() */
1463 trans = btrfs_start_transaction(root, 1);
1464 snprintf(root->fs_info->super_copy->label, BTRFS_LABEL_SIZE, "%s",
1466 btrfs_commit_transaction(trans, root);
1468 /* Now we close it since we are done. */
1473 static int set_label_mounted(const char *mount_path, const char *label)
1477 fd = open(mount_path, O_RDONLY | O_NOATIME);
1479 fprintf(stderr, "ERROR: unable to access '%s'\n", mount_path);
1483 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
1484 fprintf(stderr, "ERROR: unable to set label %s\n",
1494 static int get_label_unmounted(const char *dev, char *label)
1496 struct btrfs_root *root;
1499 ret = check_mounted(dev);
1501 fprintf(stderr, "FATAL: error checking %s mount status\n", dev);
1505 fprintf(stderr, "ERROR: dev %s is mounted, use mount point\n",
1510 /* Open the super_block at the default location
1513 root = open_ctree(dev, 0, 0);
1517 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
1519 /* Now we close it since we are done. */
1525 * If a partition is mounted, try to get the filesystem label via its
1526 * mounted path rather than device. Return the corresponding error
1527 * the user specified the device path.
1529 int get_label_mounted(const char *mount_path, char *labelp)
1531 char label[BTRFS_LABEL_SIZE];
1534 fd = open(mount_path, O_RDONLY | O_NOATIME);
1536 fprintf(stderr, "ERROR: unable to access '%s'\n", mount_path);
1540 memset(label, '\0', sizeof(label));
1541 if (ioctl(fd, BTRFS_IOC_GET_FSLABEL, label) < 0) {
1542 fprintf(stderr, "ERROR: unable get label %s\n", strerror(errno));
1547 strncpy(labelp, label, sizeof(label));
1552 int get_label(const char *btrfs_dev, char *label)
1556 ret = is_existing_blk_or_reg_file(btrfs_dev);
1558 ret = get_label_mounted(btrfs_dev, label);
1560 ret = get_label_unmounted(btrfs_dev, label);
1565 int set_label(const char *btrfs_dev, const char *label)
1569 if (check_label(label))
1572 ret = is_existing_blk_or_reg_file(btrfs_dev);
1574 ret = set_label_mounted(btrfs_dev, label);
1576 ret = set_label_unmounted(btrfs_dev, label);
1581 int btrfs_scan_block_devices(int run_ioctl)
1587 struct btrfs_fs_devices *tmp_devices;
1589 FILE *proc_partitions;
1597 proc_partitions = fopen("/proc/partitions","r");
1598 if (!proc_partitions) {
1599 fprintf(stderr, "Unable to open '/proc/partitions' for scanning\n");
1602 /* skip the header */
1603 for (i = 0; i < 2; i++)
1604 if (!fgets(buf, 1023, proc_partitions)) {
1606 "Unable to read '/proc/partitions' for scanning\n");
1607 fclose(proc_partitions);
1611 strcpy(fullpath,"/dev/");
1612 while(fgets(buf, 1023, proc_partitions)) {
1613 ret = sscanf(buf," %*d %*d %*d %99s", fullpath + 5);
1616 "failed to scan device name from /proc/partitions\n");
1621 * multipath and MD devices may register as a btrfs filesystem
1622 * both through the original block device and through
1623 * the special (/dev/mapper or /dev/mdX) entry.
1624 * This scans the special entries last
1626 special = strncmp(fullpath, "/dev/dm-", strlen("/dev/dm-")) == 0;
1628 special = strncmp(fullpath, "/dev/md", strlen("/dev/md")) == 0;
1630 if (scans == 0 && special)
1632 if (scans > 0 && !special)
1635 ret = lstat(fullpath, &st);
1637 fprintf(stderr, "failed to stat %s\n", fullpath);
1640 if (!S_ISBLK(st.st_mode)) {
1644 fd = open(fullpath, O_RDONLY);
1646 if (errno != ENOMEDIUM)
1647 fprintf(stderr, "failed to open %s: %s\n",
1648 fullpath, strerror(errno));
1651 ret = btrfs_scan_one_device(fd, fullpath, &tmp_devices,
1653 BTRFS_SUPER_INFO_OFFSET, 0);
1654 if (ret == 0 && run_ioctl > 0) {
1655 btrfs_register_one_device(fullpath);
1660 fclose(proc_partitions);
1670 * A not-so-good version fls64. No fascinating optimization since
1671 * no one except parse_size use it
1673 static int fls64(u64 x)
1677 for (i = 0; i <64; i++)
1678 if (x << i & (1ULL << 63))
1683 u64 parse_size(char *s)
1691 fprintf(stderr, "ERROR: Size value is empty\n");
1696 "ERROR: Size value '%s' is less equal than 0\n", s);
1699 ret = strtoull(s, &endptr, 10);
1701 fprintf(stderr, "ERROR: Size value '%s' is invalid\n", s);
1704 if (endptr[0] && endptr[1]) {
1705 fprintf(stderr, "ERROR: Illegal suffix contains character '%c' in wrong position\n",
1710 * strtoll returns LLONG_MAX when overflow, if this happens,
1711 * need to call strtoull to get the real size
1713 if (errno == ERANGE && ret == ULLONG_MAX) {
1715 "ERROR: Size value '%s' is too large for u64\n", s);
1719 c = tolower(endptr[0]);
1742 fprintf(stderr, "ERROR: Unknown size descriptor '%c'\n",
1747 /* Check whether ret * mult overflow */
1748 if (fls64(ret) + fls64(mult) - 1 > 64) {
1750 "ERROR: Size value '%s' is too large for u64\n", s);
1757 int open_file_or_dir3(const char *fname, DIR **dirstream, int open_flags)
1763 ret = stat(fname, &st);
1767 if (S_ISDIR(st.st_mode)) {
1768 *dirstream = opendir(fname);
1771 fd = dirfd(*dirstream);
1772 } else if (S_ISREG(st.st_mode) || S_ISLNK(st.st_mode)) {
1773 fd = open(fname, open_flags);
1776 * we set this on purpose, in case the caller output
1777 * strerror(errno) as success
1785 closedir(*dirstream);
1790 int open_file_or_dir(const char *fname, DIR **dirstream)
1792 return open_file_or_dir3(fname, dirstream, O_RDWR);
1795 void close_file_or_dir(int fd, DIR *dirstream)
1798 closedir(dirstream);
1803 int get_device_info(int fd, u64 devid,
1804 struct btrfs_ioctl_dev_info_args *di_args)
1808 di_args->devid = devid;
1809 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
1811 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
1812 return ret ? -errno : 0;
1816 * For a given path, fill in the ioctl fs_ and info_ args.
1817 * If the path is a btrfs mountpoint, fill info for all devices.
1818 * If the path is a btrfs device, fill in only that device.
1820 * The path provided must be either on a mounted btrfs fs,
1821 * or be a mounted btrfs device.
1823 * Returns 0 on success, or a negative errno.
1825 int get_fs_info(char *path, struct btrfs_ioctl_fs_info_args *fi_args,
1826 struct btrfs_ioctl_dev_info_args **di_ret)
1832 struct btrfs_fs_devices *fs_devices_mnt = NULL;
1833 struct btrfs_ioctl_dev_info_args *di_args;
1834 char mp[BTRFS_PATH_NAME_MAX + 1];
1835 DIR *dirstream = NULL;
1837 memset(fi_args, 0, sizeof(*fi_args));
1839 if (is_block_device(path)) {
1840 struct btrfs_super_block *disk_super;
1841 char buf[BTRFS_SUPER_INFO_SIZE];
1844 /* Ensure it's mounted, then set path to the mountpoint */
1845 fd = open(path, O_RDONLY);
1848 fprintf(stderr, "Couldn't open %s: %s\n",
1849 path, strerror(errno));
1852 ret = check_mounted_where(fd, path, mp, sizeof(mp),
1861 /* Only fill in this one device */
1862 fi_args->num_devices = 1;
1864 disk_super = (struct btrfs_super_block *)buf;
1865 ret = btrfs_read_dev_super(fd, disk_super,
1866 BTRFS_SUPER_INFO_OFFSET, 0);
1871 devid = btrfs_stack_device_id(&disk_super->dev_item);
1873 fi_args->max_id = devid;
1876 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
1880 /* at this point path must not be for a block device */
1881 fd = open_file_or_dir(path, &dirstream);
1887 /* fill in fi_args if not just a single device */
1888 if (fi_args->num_devices != 1) {
1889 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
1896 if (!fi_args->num_devices)
1899 di_args = *di_ret = malloc((fi_args->num_devices) * sizeof(*di_args));
1905 for (; i <= fi_args->max_id; ++i) {
1906 BUG_ON(ndevs >= fi_args->num_devices);
1907 ret = get_device_info(fd, i, &di_args[ndevs]);
1916 * only when the only dev we wanted to find is not there then
1917 * let any error be returned
1919 if (fi_args->num_devices != 1) {
1925 close_file_or_dir(fd, dirstream);
1929 #define isoctal(c) (((c) & ~7) == '0')
1931 static inline void translate(char *f, char *t)
1933 while (*f != '\0') {
1935 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
1936 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
1946 * Checks if the swap device.
1947 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
1949 static int is_swap_device(const char *file)
1960 if (stat(file, &st_buf) < 0)
1962 if (S_ISBLK(st_buf.st_mode))
1963 dev = st_buf.st_rdev;
1964 else if (S_ISREG(st_buf.st_mode)) {
1965 dev = st_buf.st_dev;
1966 ino = st_buf.st_ino;
1970 if ((f = fopen("/proc/swaps", "r")) == NULL)
1973 /* skip the first line */
1974 if (fgets(tmp, sizeof(tmp), f) == NULL)
1977 while (fgets(tmp, sizeof(tmp), f) != NULL) {
1978 if ((cp = strchr(tmp, ' ')) != NULL)
1980 if ((cp = strchr(tmp, '\t')) != NULL)
1982 translate(tmp, buf);
1983 if (stat(buf, &st_buf) != 0)
1985 if (S_ISBLK(st_buf.st_mode)) {
1986 if (dev == st_buf.st_rdev) {
1990 } else if (S_ISREG(st_buf.st_mode)) {
1991 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {
2005 * Check for existing filesystem or partition table on device.
2007 * 1 for existing fs or partition
2008 * 0 for nothing found
2009 * -1 for internal error
2016 blkid_probe pr = NULL;
2020 if (!device || !*device)
2023 ret = -1; /* will reset on success of all setup calls */
2025 pr = blkid_new_probe_from_filename(device);
2029 size = blkid_probe_get_size(pr);
2033 /* nothing to overwrite on a 0-length device */
2039 ret = blkid_probe_enable_partitions(pr, 1);
2043 ret = blkid_do_fullprobe(pr);
2048 * Blkid returns 1 for nothing found and 0 when it finds a signature,
2049 * but we want the exact opposite, so reverse the return value here.
2051 * In addition print some useful diagnostics about what actually is
2059 if (!blkid_probe_lookup_value(pr, "TYPE", &type, NULL)) {
2061 "%s appears to contain an existing "
2062 "filesystem (%s).\n", device, type);
2063 } else if (!blkid_probe_lookup_value(pr, "PTTYPE", &type, NULL)) {
2065 "%s appears to contain a partition "
2066 "table (%s).\n", device, type);
2069 "%s appears to contain something weird "
2070 "according to blkid\n", device);
2076 blkid_free_probe(pr);
2079 "probe of %s failed, cannot detect "
2080 "existing filesystem.\n", device);
2084 int test_num_disk_vs_raid(u64 metadata_profile, u64 data_profile,
2085 u64 dev_cnt, int mixed, char *estr)
2093 allowed |= BTRFS_BLOCK_GROUP_RAID10;
2095 allowed |= BTRFS_BLOCK_GROUP_RAID6;
2097 allowed |= BTRFS_BLOCK_GROUP_RAID0 | BTRFS_BLOCK_GROUP_RAID1 |
2098 BTRFS_BLOCK_GROUP_RAID5;
2101 allowed |= BTRFS_BLOCK_GROUP_DUP;
2104 if (metadata_profile & ~allowed) {
2105 snprintf(estr, sz, "unable to create FS with metadata "
2106 "profile %llu (have %llu devices)\n",
2107 metadata_profile, dev_cnt);
2110 if (data_profile & ~allowed) {
2111 snprintf(estr, sz, "unable to create FS with data "
2112 "profile %llu (have %llu devices)\n",
2113 metadata_profile, dev_cnt);
2117 if (!mixed && (data_profile & BTRFS_BLOCK_GROUP_DUP)) {
2119 "dup for data is allowed only in mixed mode");
2125 /* Check if disk is suitable for btrfs
2127 * 1: something is wrong, estr provides the error
2130 int test_dev_for_mkfs(char *file, int force_overwrite, char *estr)
2136 ret = is_swap_device(file);
2138 snprintf(estr, sz, "error checking %s status: %s\n", file,
2143 snprintf(estr, sz, "%s is a swap device\n", file);
2146 if (!force_overwrite) {
2147 if (check_overwrite(file)) {
2148 snprintf(estr, sz, "Use the -f option to force overwrite.\n");
2152 ret = check_mounted(file);
2154 snprintf(estr, sz, "error checking %s mount status\n",
2159 snprintf(estr, sz, "%s is mounted\n", file);
2162 /* check if the device is busy */
2163 fd = open(file, O_RDWR|O_EXCL);
2165 snprintf(estr, sz, "unable to open %s: %s\n", file,
2169 if (fstat(fd, &st)) {
2170 snprintf(estr, sz, "unable to stat %s: %s\n", file,
2175 if (!S_ISBLK(st.st_mode)) {
2176 fprintf(stderr, "'%s' is not a block device\n", file);
2184 int btrfs_scan_lblkid()
2189 struct btrfs_fs_devices *tmp_devices;
2190 blkid_dev_iterate iter = NULL;
2191 blkid_dev dev = NULL;
2192 blkid_cache cache = NULL;
2193 char path[PATH_MAX];
2195 if (btrfs_scan_done)
2198 if (blkid_get_cache(&cache, 0) < 0) {
2199 printf("ERROR: lblkid cache get failed\n");
2202 blkid_probe_all(cache);
2203 iter = blkid_dev_iterate_begin(cache);
2204 blkid_dev_set_search(iter, "TYPE", "btrfs");
2205 while (blkid_dev_next(iter, &dev) == 0) {
2206 dev = blkid_verify(cache, dev);
2209 /* if we are here its definitely a btrfs disk*/
2210 strncpy(path, blkid_dev_devname(dev), PATH_MAX);
2212 fd = open(path, O_RDONLY);
2214 printf("ERROR: could not open %s\n", path);
2217 ret = btrfs_scan_one_device(fd, path, &tmp_devices,
2218 &num_devices, BTRFS_SUPER_INFO_OFFSET, 0);
2220 printf("ERROR: could not scan %s\n", path);
2227 blkid_dev_iterate_end(iter);
2228 blkid_put_cache(cache);
2230 btrfs_scan_done = 1;
2235 int is_vol_small(char *file)
2242 fd = open(file, O_RDONLY);
2245 if (fstat(fd, &st) < 0) {
2250 size = btrfs_device_size(fd, &st);
2255 if (size < BTRFS_MKFS_SMALL_VOLUME_SIZE) {
2265 * This reads a line from the stdin and only returns non-zero if the
2266 * first whitespace delimited token is a case insensitive match with yes
2269 int ask_user(char *question)
2271 char buf[30] = {0,};
2272 char *saveptr = NULL;
2275 printf("%s [y/N]: ", question);
2277 return fgets(buf, sizeof(buf) - 1, stdin) &&
2278 (answer = strtok_r(buf, " \t\n\r", &saveptr)) &&
2279 (!strcasecmp(answer, "yes") || !strcasecmp(answer, "y"));
2284 * - file or directory return the containing tree root id
2285 * - subvolume return its own tree id
2286 * - BTRFS_EMPTY_SUBVOL_DIR_OBJECTID (directory with ino == 2) the result is
2287 * undefined and function returns -1
2289 int lookup_ino_rootid(int fd, u64 *rootid)
2291 struct btrfs_ioctl_ino_lookup_args args;
2295 memset(&args, 0, sizeof(args));
2297 args.objectid = BTRFS_FIRST_FREE_OBJECTID;
2299 ret = ioctl(fd, BTRFS_IOC_INO_LOOKUP, &args);
2302 fprintf(stderr, "ERROR: Failed to lookup root id - %s\n",
2307 *rootid = args.treeid;
2313 * return 0 if a btrfs mount point is found
2314 * return 1 if a mount point is found but not btrfs
2315 * return <0 if something goes wrong
2317 int find_mount_root(const char *path, char **mount_root)
2325 int longest_matchlen = 0;
2326 char *longest_match = NULL;
2328 fd = open(path, O_RDONLY | O_NOATIME);
2333 mnttab = setmntent("/proc/self/mounts", "r");
2337 while ((ent = getmntent(mnttab))) {
2338 len = strlen(ent->mnt_dir);
2339 if (strncmp(ent->mnt_dir, path, len) == 0) {
2340 /* match found and use the latest match */
2341 if (longest_matchlen <= len) {
2342 free(longest_match);
2343 longest_matchlen = len;
2344 longest_match = strdup(ent->mnt_dir);
2345 not_btrfs = strcmp(ent->mnt_type, "btrfs");
2354 free(longest_match);
2359 *mount_root = realpath(longest_match, NULL);
2363 free(longest_match);
2367 int test_minimum_size(const char *file, u32 leafsize)
2370 struct stat statbuf;
2372 fd = open(file, O_RDONLY);
2375 if (stat(file, &statbuf) < 0) {
2379 if (btrfs_device_size(fd, &statbuf) < btrfs_min_dev_size(leafsize)) {
2388 * test if name is a correct subvolume name
2389 * this function return
2390 * 0-> name is not a correct subvolume name
2391 * 1-> name is a correct subvolume name
2393 int test_issubvolname(const char *name)
2395 return name[0] != '\0' && !strchr(name, '/') &&
2396 strcmp(name, ".") && strcmp(name, "..");
2400 * test if path is a directory
2401 * this function return
2402 * 0-> path exists but it is not a directory
2403 * 1-> path exists and it is a directory
2404 * -1 -> path is unaccessible
2406 int test_isdir(const char *path)
2411 ret = stat(path, &st);
2415 return S_ISDIR(st.st_mode);
2418 void units_set_mode(unsigned *units, unsigned mode)
2420 unsigned base = *units & UNITS_MODE_MASK;
2422 *units = base | mode;
2425 void units_set_base(unsigned *units, unsigned base)
2427 unsigned mode = *units & ~UNITS_MODE_MASK;
2429 *units = base | mode;
2432 int find_next_key(struct btrfs_path *path, struct btrfs_key *key)
2436 for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
2437 if (!path->nodes[level])
2439 if (path->slots[level] + 1 >=
2440 btrfs_header_nritems(path->nodes[level]))
2443 btrfs_item_key_to_cpu(path->nodes[level], key,
2444 path->slots[level] + 1);
2446 btrfs_node_key_to_cpu(path->nodes[level], key,
2447 path->slots[level] + 1);