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 */
28 #include <sys/ioctl.h>
29 #include <sys/mount.h>
31 #include <sys/types.h>
33 #include <uuid/uuid.h>
39 #include <linux/loop.h>
40 #include <linux/major.h>
41 #include <linux/kdev_t.h>
43 #include <blkid/blkid.h>
44 #include "kerncompat.h"
45 #include "radix-tree.h"
48 #include "transaction.h"
55 #define BLKGETSIZE64 0
56 static inline int ioctl(int fd, int define, u64 *size) { return 0; }
60 #define BLKDISCARD _IO(0x12,119)
64 discard_blocks(int fd, u64 start, u64 len)
66 u64 range[2] = { start, len };
68 if (ioctl(fd, BLKDISCARD, &range) < 0)
73 static u64 reference_root_table[] = {
74 [1] = BTRFS_ROOT_TREE_OBJECTID,
75 [2] = BTRFS_EXTENT_TREE_OBJECTID,
76 [3] = BTRFS_CHUNK_TREE_OBJECTID,
77 [4] = BTRFS_DEV_TREE_OBJECTID,
78 [5] = BTRFS_FS_TREE_OBJECTID,
79 [6] = BTRFS_CSUM_TREE_OBJECTID,
82 int make_btrfs(int fd, const char *device, const char *label,
83 u64 blocks[7], u64 num_bytes, u32 nodesize,
84 u32 leafsize, u32 sectorsize, u32 stripesize)
86 struct btrfs_super_block super;
87 struct extent_buffer *buf;
88 struct btrfs_root_item root_item;
89 struct btrfs_disk_key disk_key;
90 struct btrfs_extent_item *extent_item;
91 struct btrfs_inode_item *inode_item;
92 struct btrfs_chunk *chunk;
93 struct btrfs_dev_item *dev_item;
94 struct btrfs_dev_extent *dev_extent;
95 u8 chunk_tree_uuid[BTRFS_UUID_SIZE];
106 first_free = BTRFS_SUPER_INFO_OFFSET + sectorsize * 2 - 1;
107 first_free &= ~((u64)sectorsize - 1);
109 memset(&super, 0, sizeof(super));
111 num_bytes = (num_bytes / sectorsize) * sectorsize;
112 uuid_generate(super.fsid);
113 uuid_generate(super.dev_item.uuid);
114 uuid_generate(chunk_tree_uuid);
116 btrfs_set_super_bytenr(&super, blocks[0]);
117 btrfs_set_super_num_devices(&super, 1);
118 btrfs_set_super_magic(&super, BTRFS_MAGIC);
119 btrfs_set_super_generation(&super, 1);
120 btrfs_set_super_root(&super, blocks[1]);
121 btrfs_set_super_chunk_root(&super, blocks[3]);
122 btrfs_set_super_total_bytes(&super, num_bytes);
123 btrfs_set_super_bytes_used(&super, 6 * leafsize);
124 btrfs_set_super_sectorsize(&super, sectorsize);
125 btrfs_set_super_leafsize(&super, leafsize);
126 btrfs_set_super_nodesize(&super, nodesize);
127 btrfs_set_super_stripesize(&super, stripesize);
128 btrfs_set_super_csum_type(&super, BTRFS_CSUM_TYPE_CRC32);
129 btrfs_set_super_chunk_root_generation(&super, 1);
130 btrfs_set_super_cache_generation(&super, -1);
132 strncpy(super.label, label, BTRFS_LABEL_SIZE - 1);
134 buf = malloc(sizeof(*buf) + max(sectorsize, leafsize));
136 /* create the tree of root objects */
137 memset(buf->data, 0, leafsize);
139 btrfs_set_header_bytenr(buf, blocks[1]);
140 btrfs_set_header_nritems(buf, 4);
141 btrfs_set_header_generation(buf, 1);
142 btrfs_set_header_backref_rev(buf, BTRFS_MIXED_BACKREF_REV);
143 btrfs_set_header_owner(buf, BTRFS_ROOT_TREE_OBJECTID);
144 write_extent_buffer(buf, super.fsid, (unsigned long)
145 btrfs_header_fsid(buf), BTRFS_FSID_SIZE);
147 write_extent_buffer(buf, chunk_tree_uuid, (unsigned long)
148 btrfs_header_chunk_tree_uuid(buf),
151 /* create the items for the root tree */
152 memset(&root_item, 0, sizeof(root_item));
153 inode_item = &root_item.inode;
154 btrfs_set_stack_inode_generation(inode_item, 1);
155 btrfs_set_stack_inode_size(inode_item, 3);
156 btrfs_set_stack_inode_nlink(inode_item, 1);
157 btrfs_set_stack_inode_nbytes(inode_item, leafsize);
158 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
159 btrfs_set_root_refs(&root_item, 1);
160 btrfs_set_root_used(&root_item, leafsize);
161 btrfs_set_root_generation(&root_item, 1);
163 memset(&disk_key, 0, sizeof(disk_key));
164 btrfs_set_disk_key_type(&disk_key, BTRFS_ROOT_ITEM_KEY);
165 btrfs_set_disk_key_offset(&disk_key, 0);
168 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize) - sizeof(root_item);
169 btrfs_set_root_bytenr(&root_item, blocks[2]);
170 btrfs_set_disk_key_objectid(&disk_key, BTRFS_EXTENT_TREE_OBJECTID);
171 btrfs_set_item_key(buf, &disk_key, nritems);
172 btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
173 btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems),
175 write_extent_buffer(buf, &root_item, btrfs_item_ptr_offset(buf,
176 nritems), sizeof(root_item));
179 itemoff = itemoff - sizeof(root_item);
180 btrfs_set_root_bytenr(&root_item, blocks[4]);
181 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_TREE_OBJECTID);
182 btrfs_set_item_key(buf, &disk_key, nritems);
183 btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
184 btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems),
186 write_extent_buffer(buf, &root_item,
187 btrfs_item_ptr_offset(buf, nritems),
191 itemoff = itemoff - sizeof(root_item);
192 btrfs_set_root_bytenr(&root_item, blocks[5]);
193 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FS_TREE_OBJECTID);
194 btrfs_set_item_key(buf, &disk_key, nritems);
195 btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
196 btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems),
198 write_extent_buffer(buf, &root_item,
199 btrfs_item_ptr_offset(buf, nritems),
203 itemoff = itemoff - sizeof(root_item);
204 btrfs_set_root_bytenr(&root_item, blocks[6]);
205 btrfs_set_disk_key_objectid(&disk_key, BTRFS_CSUM_TREE_OBJECTID);
206 btrfs_set_item_key(buf, &disk_key, nritems);
207 btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
208 btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems),
210 write_extent_buffer(buf, &root_item,
211 btrfs_item_ptr_offset(buf, nritems),
216 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
217 ret = pwrite(fd, buf->data, leafsize, blocks[1]);
220 else if (ret != leafsize)
223 /* create the items for the extent tree */
224 memset(buf->data+sizeof(struct btrfs_header), 0,
225 leafsize-sizeof(struct btrfs_header));
227 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize);
228 for (i = 1; i < 7; i++) {
229 BUG_ON(blocks[i] < first_free);
230 BUG_ON(blocks[i] < blocks[i - 1]);
232 /* create extent item */
233 itemoff -= sizeof(struct btrfs_extent_item) +
234 sizeof(struct btrfs_tree_block_info);
235 btrfs_set_disk_key_objectid(&disk_key, blocks[i]);
236 btrfs_set_disk_key_offset(&disk_key, leafsize);
237 btrfs_set_disk_key_type(&disk_key, BTRFS_EXTENT_ITEM_KEY);
238 btrfs_set_item_key(buf, &disk_key, nritems);
239 btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems),
241 btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems),
242 sizeof(struct btrfs_extent_item) +
243 sizeof(struct btrfs_tree_block_info));
244 extent_item = btrfs_item_ptr(buf, nritems,
245 struct btrfs_extent_item);
246 btrfs_set_extent_refs(buf, extent_item, 1);
247 btrfs_set_extent_generation(buf, extent_item, 1);
248 btrfs_set_extent_flags(buf, extent_item,
249 BTRFS_EXTENT_FLAG_TREE_BLOCK);
252 /* create extent ref */
253 ref_root = reference_root_table[i];
254 btrfs_set_disk_key_objectid(&disk_key, blocks[i]);
255 btrfs_set_disk_key_offset(&disk_key, ref_root);
256 btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
257 btrfs_set_item_key(buf, &disk_key, nritems);
258 btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems),
260 btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems), 0);
263 btrfs_set_header_bytenr(buf, blocks[2]);
264 btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
265 btrfs_set_header_nritems(buf, nritems);
266 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
267 ret = pwrite(fd, buf->data, leafsize, blocks[2]);
270 else if (ret != leafsize)
273 /* create the chunk tree */
274 memset(buf->data+sizeof(struct btrfs_header), 0,
275 leafsize-sizeof(struct btrfs_header));
277 item_size = sizeof(*dev_item);
278 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize) - item_size;
280 /* first device 1 (there is no device 0) */
281 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
282 btrfs_set_disk_key_offset(&disk_key, 1);
283 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
284 btrfs_set_item_key(buf, &disk_key, nritems);
285 btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
286 btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems), item_size);
288 dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
289 btrfs_set_device_id(buf, dev_item, 1);
290 btrfs_set_device_generation(buf, dev_item, 0);
291 btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
292 btrfs_set_device_bytes_used(buf, dev_item,
293 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
294 btrfs_set_device_io_align(buf, dev_item, sectorsize);
295 btrfs_set_device_io_width(buf, dev_item, sectorsize);
296 btrfs_set_device_sector_size(buf, dev_item, sectorsize);
297 btrfs_set_device_type(buf, dev_item, 0);
299 write_extent_buffer(buf, super.dev_item.uuid,
300 (unsigned long)btrfs_device_uuid(dev_item),
302 write_extent_buffer(buf, super.fsid,
303 (unsigned long)btrfs_device_fsid(dev_item),
305 read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
309 item_size = btrfs_chunk_item_size(1);
310 itemoff = itemoff - item_size;
312 /* then we have chunk 0 */
313 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
314 btrfs_set_disk_key_offset(&disk_key, 0);
315 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
316 btrfs_set_item_key(buf, &disk_key, nritems);
317 btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
318 btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems), item_size);
320 chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
321 btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
322 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
323 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
324 btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
325 btrfs_set_chunk_io_align(buf, chunk, sectorsize);
326 btrfs_set_chunk_io_width(buf, chunk, sectorsize);
327 btrfs_set_chunk_sector_size(buf, chunk, sectorsize);
328 btrfs_set_chunk_num_stripes(buf, chunk, 1);
329 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
330 btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
333 write_extent_buffer(buf, super.dev_item.uuid,
334 (unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
337 /* copy the key for the chunk to the system array */
338 ptr = super.sys_chunk_array;
339 array_size = sizeof(disk_key);
341 memcpy(ptr, &disk_key, sizeof(disk_key));
342 ptr += sizeof(disk_key);
344 /* copy the chunk to the system array */
345 read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
346 array_size += item_size;
348 btrfs_set_super_sys_array_size(&super, array_size);
350 btrfs_set_header_bytenr(buf, blocks[3]);
351 btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
352 btrfs_set_header_nritems(buf, nritems);
353 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
354 ret = pwrite(fd, buf->data, leafsize, blocks[3]);
357 else if (ret != leafsize)
360 /* create the device tree */
361 memset(buf->data+sizeof(struct btrfs_header), 0,
362 leafsize-sizeof(struct btrfs_header));
364 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize) -
365 sizeof(struct btrfs_dev_extent);
367 btrfs_set_disk_key_objectid(&disk_key, 1);
368 btrfs_set_disk_key_offset(&disk_key, 0);
369 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
370 btrfs_set_item_key(buf, &disk_key, nritems);
371 btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
372 btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems),
373 sizeof(struct btrfs_dev_extent));
374 dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
375 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
376 BTRFS_CHUNK_TREE_OBJECTID);
377 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
378 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
379 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
381 write_extent_buffer(buf, chunk_tree_uuid,
382 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
385 btrfs_set_dev_extent_length(buf, dev_extent,
386 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
389 btrfs_set_header_bytenr(buf, blocks[4]);
390 btrfs_set_header_owner(buf, BTRFS_DEV_TREE_OBJECTID);
391 btrfs_set_header_nritems(buf, nritems);
392 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
393 ret = pwrite(fd, buf->data, leafsize, blocks[4]);
396 else if (ret != leafsize)
399 /* create the FS root */
400 memset(buf->data+sizeof(struct btrfs_header), 0,
401 leafsize-sizeof(struct btrfs_header));
402 btrfs_set_header_bytenr(buf, blocks[5]);
403 btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
404 btrfs_set_header_nritems(buf, 0);
405 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
406 ret = pwrite(fd, buf->data, leafsize, blocks[5]);
409 else if (ret != leafsize)
412 /* finally create the csum root */
413 memset(buf->data+sizeof(struct btrfs_header), 0,
414 leafsize-sizeof(struct btrfs_header));
415 btrfs_set_header_bytenr(buf, blocks[6]);
416 btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
417 btrfs_set_header_nritems(buf, 0);
418 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
419 ret = pwrite(fd, buf->data, leafsize, blocks[6]);
422 else if (ret != leafsize)
425 /* and write out the super block */
426 BUG_ON(sizeof(super) > sectorsize);
427 memset(buf->data, 0, sectorsize);
428 memcpy(buf->data, &super, sizeof(super));
429 buf->len = sectorsize;
430 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
431 ret = pwrite(fd, buf->data, sectorsize, blocks[0]);
434 else if (ret != leafsize)
441 u64 btrfs_device_size(int fd, struct stat *st)
444 if (S_ISREG(st->st_mode)) {
447 if (!S_ISBLK(st->st_mode)) {
450 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
456 static int zero_blocks(int fd, off_t start, size_t len)
458 char *buf = malloc(len);
465 written = pwrite(fd, buf, len, start);
472 static int zero_dev_start(int fd)
475 size_t len = 2 * 1024 * 1024;
478 /* don't overwrite the disk labels on sparc */
482 return zero_blocks(fd, start, len);
485 static int zero_dev_end(int fd, u64 dev_size)
487 size_t len = 2 * 1024 * 1024;
488 off_t start = dev_size - len;
490 return zero_blocks(fd, start, len);
493 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
494 struct btrfs_root *root, int fd, char *path,
495 u64 block_count, u32 io_width, u32 io_align,
498 struct btrfs_super_block *disk_super;
499 struct btrfs_super_block *super = root->fs_info->super_copy;
500 struct btrfs_device *device;
501 struct btrfs_dev_item *dev_item;
507 device = kzalloc(sizeof(*device), GFP_NOFS);
510 buf = kmalloc(sectorsize, GFP_NOFS);
515 BUG_ON(sizeof(*disk_super) > sectorsize);
516 memset(buf, 0, sectorsize);
518 disk_super = (struct btrfs_super_block *)buf;
519 dev_item = &disk_super->dev_item;
521 uuid_generate(device->uuid);
524 device->io_width = io_width;
525 device->io_align = io_align;
526 device->sector_size = sectorsize;
528 device->writeable = 1;
529 device->total_bytes = block_count;
530 device->bytes_used = 0;
531 device->total_ios = 0;
532 device->dev_root = root->fs_info->dev_root;
534 ret = btrfs_add_device(trans, root, device);
537 total_bytes = btrfs_super_total_bytes(super) + block_count;
538 btrfs_set_super_total_bytes(super, total_bytes);
540 num_devs = btrfs_super_num_devices(super) + 1;
541 btrfs_set_super_num_devices(super, num_devs);
543 memcpy(disk_super, super, sizeof(*disk_super));
545 printf("adding device %s id %llu\n", path,
546 (unsigned long long)device->devid);
548 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
549 btrfs_set_stack_device_id(dev_item, device->devid);
550 btrfs_set_stack_device_type(dev_item, device->type);
551 btrfs_set_stack_device_io_align(dev_item, device->io_align);
552 btrfs_set_stack_device_io_width(dev_item, device->io_width);
553 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
554 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
555 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
556 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
558 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
559 BUG_ON(ret != sectorsize);
562 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
563 device->fs_devices = root->fs_info->fs_devices;
567 int btrfs_prepare_device(int fd, char *file, int zero_end, u64 *block_count_ret,
568 u64 max_block_count, int *mixed, int nodiscard)
575 ret = fstat(fd, &st);
577 fprintf(stderr, "unable to stat %s\n", file);
581 block_count = btrfs_device_size(fd, &st);
582 if (block_count == 0) {
583 fprintf(stderr, "unable to find %s size\n", file);
587 block_count = min(block_count, max_block_count);
590 if (block_count < 1024 * 1024 * 1024 && !(*mixed)) {
591 printf("SMALL VOLUME: forcing mixed metadata/data groups\n");
597 * We intentionally ignore errors from the discard ioctl. It is
598 * not necessary for the mkfs functionality but just an optimization.
600 discard_blocks(fd, 0, block_count);
603 ret = zero_dev_start(fd);
605 fprintf(stderr, "failed to zero device start %d\n", ret);
609 for (i = 0 ; i < BTRFS_SUPER_MIRROR_MAX; i++) {
610 bytenr = btrfs_sb_offset(i);
611 if (bytenr >= block_count)
613 zero_blocks(fd, bytenr, BTRFS_SUPER_INFO_SIZE);
617 ret = zero_dev_end(fd, block_count);
619 fprintf(stderr, "failed to zero device end %d\n", ret);
623 *block_count_ret = block_count;
627 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
628 struct btrfs_root *root, u64 objectid)
631 struct btrfs_inode_item inode_item;
632 time_t now = time(NULL);
634 memset(&inode_item, 0, sizeof(inode_item));
635 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
636 btrfs_set_stack_inode_size(&inode_item, 0);
637 btrfs_set_stack_inode_nlink(&inode_item, 1);
638 btrfs_set_stack_inode_nbytes(&inode_item, root->leafsize);
639 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
640 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
641 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
642 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
643 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
644 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
645 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
646 btrfs_set_stack_timespec_sec(&inode_item.otime, 0);
647 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
649 if (root->fs_info->tree_root == root)
650 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
652 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
656 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
660 btrfs_set_root_dirid(&root->root_item, objectid);
667 * checks if a path is a block device node
668 * Returns negative errno on failure, otherwise
669 * returns 1 for blockdev, 0 for not-blockdev
671 int is_block_device(const char *path) {
674 if (stat(path, &statbuf) < 0)
677 return S_ISBLK(statbuf.st_mode);
681 * Find the mount point for a mounted device.
682 * On success, returns 0 with mountpoint in *mp.
683 * On failure, returns -errno (not mounted yields -EINVAL)
684 * Is noisy on failures, expects to be given a mounted device.
686 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size) {
690 ret = is_block_device(dev);
693 fprintf(stderr, "%s is not a block device\n", dev);
696 fprintf(stderr, "Could not check %s: %s\n",
697 dev, strerror(-ret));
702 fd = open(dev, O_RDONLY);
705 fprintf(stderr, "Could not open %s: %s\n", dev, strerror(errno));
709 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
711 fprintf(stderr, "%s is not a mounted btrfs device\n", dev);
713 } else { /* mounted, all good */
720 fprintf(stderr, "Could not get mountpoint for %s\n", dev);
725 * Given a pathname, return a filehandle to:
726 * the original pathname or,
727 * if the pathname is a mounted btrfs device, to its mountpoint.
729 * On error, return -1, errno should be set.
731 int open_path_or_dev_mnt(const char *path)
733 char mp[BTRFS_PATH_NAME_MAX + 1];
736 if (is_block_device(path)) {
739 ret = get_btrfs_mount(path, mp, sizeof(mp));
741 /* not a mounted btrfs dev */
745 fdmnt = open_file_or_dir(mp);
747 fdmnt = open_file_or_dir(path);
753 /* checks if a device is a loop device */
754 int is_loop_device (const char* device) {
757 if(stat(device, &statbuf) < 0)
760 return (S_ISBLK(statbuf.st_mode) &&
761 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
765 /* Takes a loop device path (e.g. /dev/loop0) and returns
766 * the associated file (e.g. /images/my_btrfs.img) */
767 int resolve_loop_device(const char* loop_dev, char* loop_file, int max_len)
773 char real_loop_dev[PATH_MAX];
775 if (!realpath(loop_dev, real_loop_dev))
777 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
778 if (!(f = fopen(p, "r")))
781 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
782 ret = fscanf(f, fmt, loop_file);
790 /* Checks whether a and b are identical or device
791 * files associated with the same block device
793 int is_same_blk_file(const char* a, const char* b)
795 struct stat st_buf_a, st_buf_b;
796 char real_a[PATH_MAX];
797 char real_b[PATH_MAX];
799 if(!realpath(a, real_a) ||
800 !realpath(b, real_b))
805 /* Identical path? */
806 if(strcmp(real_a, real_b) == 0)
809 if(stat(a, &st_buf_a) < 0 ||
810 stat(b, &st_buf_b) < 0)
817 /* Same blockdevice? */
818 if(S_ISBLK(st_buf_a.st_mode) &&
819 S_ISBLK(st_buf_b.st_mode) &&
820 st_buf_a.st_rdev == st_buf_b.st_rdev)
826 if (st_buf_a.st_dev == st_buf_b.st_dev &&
827 st_buf_a.st_ino == st_buf_b.st_ino)
835 /* checks if a and b are identical or device
836 * files associated with the same block device or
837 * if one file is a loop device that uses the other
840 int is_same_loop_file(const char* a, const char* b)
842 char res_a[PATH_MAX];
843 char res_b[PATH_MAX];
848 /* Resolve a if it is a loop device */
849 if((ret = is_loop_device(a)) < 0) {
854 if ((ret = resolve_loop_device(a, res_a, sizeof(res_a))) < 0)
862 /* Resolve b if it is a loop device */
863 if ((ret = is_loop_device(b)) < 0) {
868 if((ret = resolve_loop_device(b, res_b, sizeof(res_b))) < 0)
876 return is_same_blk_file(final_a, final_b);
879 /* Checks if a file exists and is a block or regular file*/
880 int is_existing_blk_or_reg_file(const char* filename)
884 if(stat(filename, &st_buf) < 0) {
891 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
894 /* Checks if a file is used (directly or indirectly via a loop device)
895 * by a device in fs_devices
897 int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices, const char* file)
900 struct list_head *head;
901 struct list_head *cur;
902 struct btrfs_device *device;
904 head = &fs_devices->devices;
905 list_for_each(cur, head) {
906 device = list_entry(cur, struct btrfs_device, dev_list);
908 if((ret = is_same_loop_file(device->name, file)))
916 * returns 1 if the device was mounted, < 0 on error or 0 if everything
917 * is safe to continue.
919 int check_mounted(const char* file)
924 fd = open(file, O_RDONLY);
926 fprintf (stderr, "check_mounted(): Could not open %s\n", file);
930 ret = check_mounted_where(fd, file, NULL, 0, NULL);
936 int check_mounted_where(int fd, const char *file, char *where, int size,
937 struct btrfs_fs_devices **fs_dev_ret)
942 struct btrfs_fs_devices *fs_devices_mnt = NULL;
946 /* scan the initial device */
947 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
948 &total_devs, BTRFS_SUPER_INFO_OFFSET);
949 is_btrfs = (ret >= 0);
951 /* scan other devices */
952 if (is_btrfs && total_devs > 1) {
953 if((ret = btrfs_scan_for_fsid(fs_devices_mnt, total_devs, 1)))
957 /* iterate over the list of currently mountes filesystems */
958 if ((f = setmntent ("/proc/mounts", "r")) == NULL)
961 while ((mnt = getmntent (f)) != NULL) {
963 if(strcmp(mnt->mnt_type, "btrfs") != 0)
966 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
968 /* ignore entries in the mount table that are not
969 associated with a file*/
970 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
971 goto out_mntloop_err;
975 ret = is_same_loop_file(file, mnt->mnt_fsname);
979 goto out_mntloop_err;
984 /* Did we find an entry in mnt table? */
985 if (mnt && size && where) {
986 strncpy(where, mnt->mnt_dir, size);
990 *fs_dev_ret = fs_devices_mnt;
1000 struct pending_dir {
1001 struct list_head list;
1002 char name[PATH_MAX];
1005 void btrfs_register_one_device(char *fname)
1007 struct btrfs_ioctl_vol_args args;
1012 fd = open("/dev/btrfs-control", O_RDONLY);
1014 fprintf(stderr, "failed to open /dev/btrfs-control "
1015 "skipping device registration: %s\n",
1019 strncpy(args.name, fname, BTRFS_PATH_NAME_MAX);
1020 args.name[BTRFS_PATH_NAME_MAX-1] = 0;
1021 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
1024 fprintf(stderr, "ERROR: device scan failed '%s' - %s\n",
1025 fname, strerror(e));
1030 int btrfs_scan_one_dir(char *dirname, int run_ioctl)
1033 struct dirent *dirent;
1034 struct pending_dir *pending;
1040 struct list_head pending_list;
1041 struct btrfs_fs_devices *tmp_devices;
1044 INIT_LIST_HEAD(&pending_list);
1046 pending = malloc(sizeof(*pending));
1049 strcpy(pending->name, dirname);
1052 dirname_len = strlen(pending->name);
1053 fullpath = malloc(PATH_MAX);
1054 dirname = pending->name;
1060 dirp = opendir(dirname);
1062 fprintf(stderr, "Unable to open %s for scanning\n", dirname);
1067 dirent = readdir(dirp);
1070 if (dirent->d_name[0] == '.')
1072 if (dirname_len + strlen(dirent->d_name) + 2 > PATH_MAX) {
1076 snprintf(fullpath, PATH_MAX, "%s/%s", dirname, dirent->d_name);
1077 ret = lstat(fullpath, &st);
1079 fprintf(stderr, "failed to stat %s\n", fullpath);
1082 if (S_ISLNK(st.st_mode))
1084 if (S_ISDIR(st.st_mode)) {
1085 struct pending_dir *next = malloc(sizeof(*next));
1090 strcpy(next->name, fullpath);
1091 list_add_tail(&next->list, &pending_list);
1093 if (!S_ISBLK(st.st_mode)) {
1096 fd = open(fullpath, O_RDONLY);
1098 /* ignore the following errors:
1099 ENXIO (device don't exists)
1100 ENOMEDIUM (No medium found ->
1101 like a cd tray empty)
1103 if(errno != ENXIO && errno != ENOMEDIUM)
1104 fprintf(stderr, "failed to read %s: %s\n",
1105 fullpath, strerror(errno));
1108 ret = btrfs_scan_one_device(fd, fullpath, &tmp_devices,
1110 BTRFS_SUPER_INFO_OFFSET);
1111 if (ret == 0 && run_ioctl > 0) {
1112 btrfs_register_one_device(fullpath);
1116 if (!list_empty(&pending_list)) {
1118 pending = list_entry(pending_list.next, struct pending_dir,
1121 list_del(&pending->list);
1135 int btrfs_scan_for_fsid(struct btrfs_fs_devices *fs_devices, u64 total_devs,
1140 ret = btrfs_scan_block_devices(run_ioctls);
1142 ret = btrfs_scan_one_dir("/dev", run_ioctls);
1146 int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
1149 struct btrfs_super_block *disk_super;
1153 buf = malloc(BTRFS_SUPER_INFO_SIZE);
1158 ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
1159 if (ret != BTRFS_SUPER_INFO_SIZE)
1163 disk_super = (struct btrfs_super_block *)buf;
1164 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC)
1167 if (!memcmp(disk_super->fsid, root->fs_info->super_copy->fsid,
1176 static char *size_strs[] = { "", "KB", "MB", "GB", "TB",
1178 void pretty_size_snprintf(u64 size, char *str, size_t str_bytes)
1190 u64 last_size = size;
1192 while(size >= 1024){
1198 if (num_divs >= ARRAY_SIZE(size_strs)) {
1202 fraction = (float)last_size / 1024;
1204 snprintf(str, str_bytes, "%.2f%s", fraction, size_strs[num_divs]);
1208 * __strncpy__null - strncpy with null termination
1209 * @dest: the target array
1210 * @src: the source string
1211 * @n: maximum bytes to copy (size of *dest)
1213 * Like strncpy, but ensures destination is null-terminated.
1215 * Copies the string pointed to by src, including the terminating null
1216 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
1217 * of n bytes. Then ensure that dest is null-terminated.
1219 char *__strncpy__null(char *dest, const char *src, size_t n)
1221 strncpy(dest, src, n);
1228 * Checks to make sure that the label matches our requirements.
1230 0 if everything is safe and usable
1231 -1 if the label is too long
1233 static int check_label(const char *input)
1235 int len = strlen(input);
1237 if (len > BTRFS_LABEL_SIZE - 1) {
1238 fprintf(stderr, "ERROR: Label %s is too long (max %d)\n",
1239 input, BTRFS_LABEL_SIZE - 1);
1246 static int set_label_unmounted(const char *dev, const char *label)
1248 struct btrfs_trans_handle *trans;
1249 struct btrfs_root *root;
1252 ret = check_mounted(dev);
1254 fprintf(stderr, "FATAL: error checking %s mount status\n", dev);
1258 fprintf(stderr, "ERROR: dev %s is mounted, use mount point\n",
1263 /* Open the super_block at the default location
1264 * and as read-write.
1266 root = open_ctree(dev, 0, 1);
1267 if (!root) /* errors are printed by open_ctree() */
1270 trans = btrfs_start_transaction(root, 1);
1271 snprintf(root->fs_info->super_copy->label, BTRFS_LABEL_SIZE, "%s",
1273 btrfs_commit_transaction(trans, root);
1275 /* Now we close it since we are done. */
1280 static int set_label_mounted(const char *mount_path, const char *label)
1284 fd = open(mount_path, O_RDONLY | O_NOATIME);
1286 fprintf(stderr, "ERROR: unable access to '%s'\n", mount_path);
1290 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
1291 fprintf(stderr, "ERROR: unable to set label %s\n",
1301 static int get_label_unmounted(const char *dev)
1303 struct btrfs_root *root;
1306 ret = check_mounted(dev);
1308 fprintf(stderr, "FATAL: error checking %s mount status\n", dev);
1312 fprintf(stderr, "ERROR: dev %s is mounted, use mount point\n",
1317 /* Open the super_block at the default location
1320 root = open_ctree(dev, 0, 0);
1324 fprintf(stdout, "%s\n", root->fs_info->super_copy->label);
1326 /* Now we close it since we are done. */
1332 * If a partition is mounted, try to get the filesystem label via its
1333 * mounted path rather than device. Return the corresponding error
1334 * the user specified the device path.
1336 static int get_label_mounted(const char *mount_path)
1338 char label[BTRFS_LABEL_SIZE];
1341 fd = open(mount_path, O_RDONLY | O_NOATIME);
1343 fprintf(stderr, "ERROR: unable access to '%s'\n", mount_path);
1347 memset(label, '\0', sizeof(label));
1348 if (ioctl(fd, BTRFS_IOC_GET_FSLABEL, label) < 0) {
1349 fprintf(stderr, "ERROR: unable get label %s\n", strerror(errno));
1354 fprintf(stdout, "%s\n", label);
1359 int get_label(const char *btrfs_dev)
1361 return is_existing_blk_or_reg_file(btrfs_dev) ?
1362 get_label_unmounted(btrfs_dev) :
1363 get_label_mounted(btrfs_dev);
1366 int set_label(const char *btrfs_dev, const char *label)
1368 if (check_label(label))
1371 return is_existing_blk_or_reg_file(btrfs_dev) ?
1372 set_label_unmounted(btrfs_dev, label) :
1373 set_label_mounted(btrfs_dev, label);
1376 int btrfs_scan_block_devices(int run_ioctl)
1382 struct btrfs_fs_devices *tmp_devices;
1384 FILE *proc_partitions;
1392 proc_partitions = fopen("/proc/partitions","r");
1393 if (!proc_partitions) {
1394 fprintf(stderr, "Unable to open '/proc/partitions' for scanning\n");
1397 /* skip the header */
1398 for (i = 0; i < 2; i++)
1399 if (!fgets(buf, 1023, proc_partitions)) {
1401 "Unable to read '/proc/partitions' for scanning\n");
1402 fclose(proc_partitions);
1406 strcpy(fullpath,"/dev/");
1407 while(fgets(buf, 1023, proc_partitions)) {
1408 i = sscanf(buf," %*d %*d %*d %99s", fullpath+5);
1411 * multipath and MD devices may register as a btrfs filesystem
1412 * both through the original block device and through
1413 * the special (/dev/mapper or /dev/mdX) entry.
1414 * This scans the special entries last
1416 special = strncmp(fullpath, "/dev/dm-", strlen("/dev/dm-")) == 0;
1418 special = strncmp(fullpath, "/dev/md", strlen("/dev/md")) == 0;
1420 if (scans == 0 && special)
1422 if (scans > 0 && !special)
1425 ret = lstat(fullpath, &st);
1427 fprintf(stderr, "failed to stat %s\n", fullpath);
1430 if (!S_ISBLK(st.st_mode)) {
1434 fd = open(fullpath, O_RDONLY);
1436 fprintf(stderr, "failed to open %s: %s\n",
1437 fullpath, strerror(errno));
1440 ret = btrfs_scan_one_device(fd, fullpath, &tmp_devices,
1442 BTRFS_SUPER_INFO_OFFSET);
1443 if (ret == 0 && run_ioctl > 0) {
1444 btrfs_register_one_device(fullpath);
1449 fclose(proc_partitions);
1458 u64 parse_size(char *s)
1464 for (i = 0; s && s[i] && isdigit(s[i]); i++) ;
1466 fprintf(stderr, "ERROR: size value is empty\n");
1488 fprintf(stderr, "ERROR: Unknown size descriptor "
1493 if (s[i] && s[i+1]) {
1494 fprintf(stderr, "ERROR: Illegal suffix contains "
1495 "character '%c' in wrong position\n",
1499 return strtoull(s, NULL, 10) * mult;
1502 int open_file_or_dir(const char *fname)
1509 ret = stat(fname, &st);
1513 if (S_ISDIR(st.st_mode)) {
1514 dirstream = opendir(fname);
1518 fd = dirfd(dirstream);
1520 fd = open(fname, O_RDWR);
1528 int get_device_info(int fd, u64 devid,
1529 struct btrfs_ioctl_dev_info_args *di_args)
1533 di_args->devid = devid;
1534 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
1536 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
1537 return ret ? -errno : 0;
1541 * For a given path, fill in the ioctl fs_ and info_ args.
1542 * If the path is a btrfs mountpoint, fill info for all devices.
1543 * If the path is a btrfs device, fill in only that device.
1545 * The path provided must be either on a mounted btrfs fs,
1546 * or be a mounted btrfs device.
1548 * Returns 0 on success, or a negative errno.
1550 int get_fs_info(char *path, struct btrfs_ioctl_fs_info_args *fi_args,
1551 struct btrfs_ioctl_dev_info_args **di_ret)
1557 struct btrfs_fs_devices *fs_devices_mnt = NULL;
1558 struct btrfs_ioctl_dev_info_args *di_args;
1559 char mp[BTRFS_PATH_NAME_MAX + 1];
1561 memset(fi_args, 0, sizeof(*fi_args));
1563 if (is_block_device(path)) {
1564 /* Ensure it's mounted, then set path to the mountpoint */
1565 fd = open(path, O_RDONLY);
1568 fprintf(stderr, "Couldn't open %s: %s\n",
1569 path, strerror(errno));
1572 ret = check_mounted_where(fd, path, mp, sizeof(mp),
1581 /* Only fill in this one device */
1582 fi_args->num_devices = 1;
1583 fi_args->max_id = fs_devices_mnt->latest_devid;
1584 i = fs_devices_mnt->latest_devid;
1585 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
1589 /* at this point path must not be for a block device */
1590 fd = open_file_or_dir(path);
1596 /* fill in fi_args if not just a single device */
1597 if (fi_args->num_devices != 1) {
1598 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
1605 if (!fi_args->num_devices)
1608 di_args = *di_ret = malloc(fi_args->num_devices * sizeof(*di_args));
1614 for (; i <= fi_args->max_id; ++i) {
1615 BUG_ON(ndevs >= fi_args->num_devices);
1616 ret = get_device_info(fd, i, &di_args[ndevs]);
1632 #define isoctal(c) (((c) & ~7) == '0')
1634 static inline void translate(char *f, char *t)
1636 while (*f != '\0') {
1638 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
1639 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
1649 * Checks if the swap device.
1650 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
1652 int is_swap_device(const char *file)
1663 if (stat(file, &st_buf) < 0)
1665 if (S_ISBLK(st_buf.st_mode))
1666 dev = st_buf.st_rdev;
1667 else if (S_ISREG(st_buf.st_mode)) {
1668 dev = st_buf.st_dev;
1669 ino = st_buf.st_ino;
1673 if ((f = fopen("/proc/swaps", "r")) == NULL)
1676 /* skip the first line */
1677 if (fgets(tmp, sizeof(tmp), f) == NULL)
1680 while (fgets(tmp, sizeof(tmp), f) != NULL) {
1681 if ((cp = strchr(tmp, ' ')) != NULL)
1683 if ((cp = strchr(tmp, '\t')) != NULL)
1685 translate(tmp, buf);
1686 if (stat(buf, &st_buf) != 0)
1688 if (S_ISBLK(st_buf.st_mode)) {
1689 if (dev == st_buf.st_rdev) {
1693 } else if (S_ISREG(st_buf.st_mode)) {
1694 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {
1708 * Check for existing filesystem or partition table on device.
1710 * 1 for existing fs or partition
1711 * 0 for nothing found
1712 * -1 for internal error
1719 blkid_probe pr = NULL;
1723 if (!device || !*device)
1726 ret = -1; /* will reset on success of all setup calls */
1728 pr = blkid_new_probe_from_filename(device);
1732 size = blkid_probe_get_size(pr);
1736 /* nothing to overwrite on a 0-length device */
1742 ret = blkid_probe_enable_partitions(pr, 1);
1746 ret = blkid_do_fullprobe(pr);
1751 * Blkid returns 1 for nothing found and 0 when it finds a signature,
1752 * but we want the exact opposite, so reverse the return value here.
1754 * In addition print some useful diagnostics about what actually is
1762 if (!blkid_probe_lookup_value(pr, "TYPE", &type, NULL)) {
1764 "%s appears to contain an existing "
1765 "filesystem (%s).\n", device, type);
1766 } else if (!blkid_probe_lookup_value(pr, "PTTYPE", &type, NULL)) {
1768 "%s appears to contain a partition "
1769 "table (%s).\n", device, type);
1772 "%s appears to contain something weird "
1773 "according to blkid\n", device);
1779 blkid_free_probe(pr);
1782 "probe of %s failed, cannot detect "
1783 "existing filesystem.\n", device);
1787 /* Check if disk is suitable for btrfs
1789 * 1: something is wrong, estr provides the error
1792 int test_dev_for_mkfs(char *file, int force_overwrite, char *estr)
1797 ret = is_swap_device(file);
1799 snprintf(estr, sz, "error checking %s status: %s\n", file,
1804 snprintf(estr, sz, "%s is a swap device\n", file);
1807 if (!force_overwrite) {
1808 if (check_overwrite(file)) {
1809 snprintf(estr, sz, "Use the -f option to force overwrite.\n");
1813 ret = check_mounted(file);
1815 snprintf(estr, sz, "error checking %s mount status\n",
1820 snprintf(estr, sz, "%s is mounted\n", file);
1823 /* check if the device is busy */
1824 fd = open(file, O_RDWR|O_EXCL);
1826 snprintf(estr, sz, "unable to open %s: %s\n", file,