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]);
218 BUG_ON(ret != leafsize);
220 /* create the items for the extent tree */
221 memset(buf->data+sizeof(struct btrfs_header), 0,
222 leafsize-sizeof(struct btrfs_header));
224 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize);
225 for (i = 1; i < 7; i++) {
226 BUG_ON(blocks[i] < first_free);
227 BUG_ON(blocks[i] < blocks[i - 1]);
229 /* create extent item */
230 itemoff -= sizeof(struct btrfs_extent_item) +
231 sizeof(struct btrfs_tree_block_info);
232 btrfs_set_disk_key_objectid(&disk_key, blocks[i]);
233 btrfs_set_disk_key_offset(&disk_key, leafsize);
234 btrfs_set_disk_key_type(&disk_key, BTRFS_EXTENT_ITEM_KEY);
235 btrfs_set_item_key(buf, &disk_key, nritems);
236 btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems),
238 btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems),
239 sizeof(struct btrfs_extent_item) +
240 sizeof(struct btrfs_tree_block_info));
241 extent_item = btrfs_item_ptr(buf, nritems,
242 struct btrfs_extent_item);
243 btrfs_set_extent_refs(buf, extent_item, 1);
244 btrfs_set_extent_generation(buf, extent_item, 1);
245 btrfs_set_extent_flags(buf, extent_item,
246 BTRFS_EXTENT_FLAG_TREE_BLOCK);
249 /* create extent ref */
250 ref_root = reference_root_table[i];
251 btrfs_set_disk_key_objectid(&disk_key, blocks[i]);
252 btrfs_set_disk_key_offset(&disk_key, ref_root);
253 btrfs_set_disk_key_type(&disk_key, BTRFS_TREE_BLOCK_REF_KEY);
254 btrfs_set_item_key(buf, &disk_key, nritems);
255 btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems),
257 btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems), 0);
260 btrfs_set_header_bytenr(buf, blocks[2]);
261 btrfs_set_header_owner(buf, BTRFS_EXTENT_TREE_OBJECTID);
262 btrfs_set_header_nritems(buf, nritems);
263 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
264 ret = pwrite(fd, buf->data, leafsize, blocks[2]);
265 BUG_ON(ret != leafsize);
267 /* create the chunk tree */
268 memset(buf->data+sizeof(struct btrfs_header), 0,
269 leafsize-sizeof(struct btrfs_header));
271 item_size = sizeof(*dev_item);
272 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize) - item_size;
274 /* first device 1 (there is no device 0) */
275 btrfs_set_disk_key_objectid(&disk_key, BTRFS_DEV_ITEMS_OBJECTID);
276 btrfs_set_disk_key_offset(&disk_key, 1);
277 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_ITEM_KEY);
278 btrfs_set_item_key(buf, &disk_key, nritems);
279 btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
280 btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems), item_size);
282 dev_item = btrfs_item_ptr(buf, nritems, struct btrfs_dev_item);
283 btrfs_set_device_id(buf, dev_item, 1);
284 btrfs_set_device_generation(buf, dev_item, 0);
285 btrfs_set_device_total_bytes(buf, dev_item, num_bytes);
286 btrfs_set_device_bytes_used(buf, dev_item,
287 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
288 btrfs_set_device_io_align(buf, dev_item, sectorsize);
289 btrfs_set_device_io_width(buf, dev_item, sectorsize);
290 btrfs_set_device_sector_size(buf, dev_item, sectorsize);
291 btrfs_set_device_type(buf, dev_item, 0);
293 write_extent_buffer(buf, super.dev_item.uuid,
294 (unsigned long)btrfs_device_uuid(dev_item),
296 write_extent_buffer(buf, super.fsid,
297 (unsigned long)btrfs_device_fsid(dev_item),
299 read_extent_buffer(buf, &super.dev_item, (unsigned long)dev_item,
303 item_size = btrfs_chunk_item_size(1);
304 itemoff = itemoff - item_size;
306 /* then we have chunk 0 */
307 btrfs_set_disk_key_objectid(&disk_key, BTRFS_FIRST_CHUNK_TREE_OBJECTID);
308 btrfs_set_disk_key_offset(&disk_key, 0);
309 btrfs_set_disk_key_type(&disk_key, BTRFS_CHUNK_ITEM_KEY);
310 btrfs_set_item_key(buf, &disk_key, nritems);
311 btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
312 btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems), item_size);
314 chunk = btrfs_item_ptr(buf, nritems, struct btrfs_chunk);
315 btrfs_set_chunk_length(buf, chunk, BTRFS_MKFS_SYSTEM_GROUP_SIZE);
316 btrfs_set_chunk_owner(buf, chunk, BTRFS_EXTENT_TREE_OBJECTID);
317 btrfs_set_chunk_stripe_len(buf, chunk, 64 * 1024);
318 btrfs_set_chunk_type(buf, chunk, BTRFS_BLOCK_GROUP_SYSTEM);
319 btrfs_set_chunk_io_align(buf, chunk, sectorsize);
320 btrfs_set_chunk_io_width(buf, chunk, sectorsize);
321 btrfs_set_chunk_sector_size(buf, chunk, sectorsize);
322 btrfs_set_chunk_num_stripes(buf, chunk, 1);
323 btrfs_set_stripe_devid_nr(buf, chunk, 0, 1);
324 btrfs_set_stripe_offset_nr(buf, chunk, 0, 0);
327 write_extent_buffer(buf, super.dev_item.uuid,
328 (unsigned long)btrfs_stripe_dev_uuid(&chunk->stripe),
331 /* copy the key for the chunk to the system array */
332 ptr = super.sys_chunk_array;
333 array_size = sizeof(disk_key);
335 memcpy(ptr, &disk_key, sizeof(disk_key));
336 ptr += sizeof(disk_key);
338 /* copy the chunk to the system array */
339 read_extent_buffer(buf, ptr, (unsigned long)chunk, item_size);
340 array_size += item_size;
342 btrfs_set_super_sys_array_size(&super, array_size);
344 btrfs_set_header_bytenr(buf, blocks[3]);
345 btrfs_set_header_owner(buf, BTRFS_CHUNK_TREE_OBJECTID);
346 btrfs_set_header_nritems(buf, nritems);
347 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
348 ret = pwrite(fd, buf->data, leafsize, blocks[3]);
349 BUG_ON(ret != leafsize);
351 /* create the device tree */
352 memset(buf->data+sizeof(struct btrfs_header), 0,
353 leafsize-sizeof(struct btrfs_header));
355 itemoff = __BTRFS_LEAF_DATA_SIZE(leafsize) -
356 sizeof(struct btrfs_dev_extent);
358 btrfs_set_disk_key_objectid(&disk_key, 1);
359 btrfs_set_disk_key_offset(&disk_key, 0);
360 btrfs_set_disk_key_type(&disk_key, BTRFS_DEV_EXTENT_KEY);
361 btrfs_set_item_key(buf, &disk_key, nritems);
362 btrfs_set_item_offset(buf, btrfs_item_nr(buf, nritems), itemoff);
363 btrfs_set_item_size(buf, btrfs_item_nr(buf, nritems),
364 sizeof(struct btrfs_dev_extent));
365 dev_extent = btrfs_item_ptr(buf, nritems, struct btrfs_dev_extent);
366 btrfs_set_dev_extent_chunk_tree(buf, dev_extent,
367 BTRFS_CHUNK_TREE_OBJECTID);
368 btrfs_set_dev_extent_chunk_objectid(buf, dev_extent,
369 BTRFS_FIRST_CHUNK_TREE_OBJECTID);
370 btrfs_set_dev_extent_chunk_offset(buf, dev_extent, 0);
372 write_extent_buffer(buf, chunk_tree_uuid,
373 (unsigned long)btrfs_dev_extent_chunk_tree_uuid(dev_extent),
376 btrfs_set_dev_extent_length(buf, dev_extent,
377 BTRFS_MKFS_SYSTEM_GROUP_SIZE);
380 btrfs_set_header_bytenr(buf, blocks[4]);
381 btrfs_set_header_owner(buf, BTRFS_DEV_TREE_OBJECTID);
382 btrfs_set_header_nritems(buf, nritems);
383 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
384 ret = pwrite(fd, buf->data, leafsize, blocks[4]);
385 BUG_ON(ret != leafsize);
387 /* create the FS root */
388 memset(buf->data+sizeof(struct btrfs_header), 0,
389 leafsize-sizeof(struct btrfs_header));
390 btrfs_set_header_bytenr(buf, blocks[5]);
391 btrfs_set_header_owner(buf, BTRFS_FS_TREE_OBJECTID);
392 btrfs_set_header_nritems(buf, 0);
393 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
394 ret = pwrite(fd, buf->data, leafsize, blocks[5]);
395 BUG_ON(ret != leafsize);
397 /* finally create the csum root */
398 memset(buf->data+sizeof(struct btrfs_header), 0,
399 leafsize-sizeof(struct btrfs_header));
400 btrfs_set_header_bytenr(buf, blocks[6]);
401 btrfs_set_header_owner(buf, BTRFS_CSUM_TREE_OBJECTID);
402 btrfs_set_header_nritems(buf, 0);
403 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
404 ret = pwrite(fd, buf->data, leafsize, blocks[6]);
405 BUG_ON(ret != leafsize);
407 /* and write out the super block */
408 BUG_ON(sizeof(super) > sectorsize);
409 memset(buf->data, 0, sectorsize);
410 memcpy(buf->data, &super, sizeof(super));
411 buf->len = sectorsize;
412 csum_tree_block_size(buf, BTRFS_CRC32_SIZE, 0);
413 ret = pwrite(fd, buf->data, sectorsize, blocks[0]);
414 BUG_ON(ret != sectorsize);
421 u64 btrfs_device_size(int fd, struct stat *st)
424 if (S_ISREG(st->st_mode)) {
427 if (!S_ISBLK(st->st_mode)) {
430 if (ioctl(fd, BLKGETSIZE64, &size) >= 0) {
436 static int zero_blocks(int fd, off_t start, size_t len)
438 char *buf = malloc(len);
445 written = pwrite(fd, buf, len, start);
452 static int zero_dev_start(int fd)
455 size_t len = 2 * 1024 * 1024;
458 /* don't overwrite the disk labels on sparc */
462 return zero_blocks(fd, start, len);
465 static int zero_dev_end(int fd, u64 dev_size)
467 size_t len = 2 * 1024 * 1024;
468 off_t start = dev_size - len;
470 return zero_blocks(fd, start, len);
473 int btrfs_add_to_fsid(struct btrfs_trans_handle *trans,
474 struct btrfs_root *root, int fd, char *path,
475 u64 block_count, u32 io_width, u32 io_align,
478 struct btrfs_super_block *disk_super;
479 struct btrfs_super_block *super = root->fs_info->super_copy;
480 struct btrfs_device *device;
481 struct btrfs_dev_item *dev_item;
487 device = kzalloc(sizeof(*device), GFP_NOFS);
490 buf = kmalloc(sectorsize, GFP_NOFS);
495 BUG_ON(sizeof(*disk_super) > sectorsize);
496 memset(buf, 0, sectorsize);
498 disk_super = (struct btrfs_super_block *)buf;
499 dev_item = &disk_super->dev_item;
501 uuid_generate(device->uuid);
504 device->io_width = io_width;
505 device->io_align = io_align;
506 device->sector_size = sectorsize;
508 device->writeable = 1;
509 device->total_bytes = block_count;
510 device->bytes_used = 0;
511 device->total_ios = 0;
512 device->dev_root = root->fs_info->dev_root;
514 ret = btrfs_add_device(trans, root, device);
517 total_bytes = btrfs_super_total_bytes(super) + block_count;
518 btrfs_set_super_total_bytes(super, total_bytes);
520 num_devs = btrfs_super_num_devices(super) + 1;
521 btrfs_set_super_num_devices(super, num_devs);
523 memcpy(disk_super, super, sizeof(*disk_super));
525 printf("adding device %s id %llu\n", path,
526 (unsigned long long)device->devid);
528 btrfs_set_super_bytenr(disk_super, BTRFS_SUPER_INFO_OFFSET);
529 btrfs_set_stack_device_id(dev_item, device->devid);
530 btrfs_set_stack_device_type(dev_item, device->type);
531 btrfs_set_stack_device_io_align(dev_item, device->io_align);
532 btrfs_set_stack_device_io_width(dev_item, device->io_width);
533 btrfs_set_stack_device_sector_size(dev_item, device->sector_size);
534 btrfs_set_stack_device_total_bytes(dev_item, device->total_bytes);
535 btrfs_set_stack_device_bytes_used(dev_item, device->bytes_used);
536 memcpy(&dev_item->uuid, device->uuid, BTRFS_UUID_SIZE);
538 ret = pwrite(fd, buf, sectorsize, BTRFS_SUPER_INFO_OFFSET);
539 BUG_ON(ret != sectorsize);
542 list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
543 device->fs_devices = root->fs_info->fs_devices;
547 int btrfs_prepare_device(int fd, char *file, int zero_end, u64 *block_count_ret,
548 u64 max_block_count, int *mixed, int nodiscard)
555 ret = fstat(fd, &st);
557 fprintf(stderr, "unable to stat %s\n", file);
561 block_count = btrfs_device_size(fd, &st);
562 if (block_count == 0) {
563 fprintf(stderr, "unable to find %s size\n", file);
567 block_count = min(block_count, max_block_count);
570 if (block_count < 1024 * 1024 * 1024 && !(*mixed)) {
571 printf("SMALL VOLUME: forcing mixed metadata/data groups\n");
577 * We intentionally ignore errors from the discard ioctl. It is
578 * not necessary for the mkfs functionality but just an optimization.
580 discard_blocks(fd, 0, block_count);
583 ret = zero_dev_start(fd);
585 fprintf(stderr, "failed to zero device start %d\n", ret);
589 for (i = 0 ; i < BTRFS_SUPER_MIRROR_MAX; i++) {
590 bytenr = btrfs_sb_offset(i);
591 if (bytenr >= block_count)
593 zero_blocks(fd, bytenr, BTRFS_SUPER_INFO_SIZE);
597 ret = zero_dev_end(fd, block_count);
599 fprintf(stderr, "failed to zero device end %d\n", ret);
603 *block_count_ret = block_count;
607 int btrfs_make_root_dir(struct btrfs_trans_handle *trans,
608 struct btrfs_root *root, u64 objectid)
611 struct btrfs_inode_item inode_item;
612 time_t now = time(NULL);
614 memset(&inode_item, 0, sizeof(inode_item));
615 btrfs_set_stack_inode_generation(&inode_item, trans->transid);
616 btrfs_set_stack_inode_size(&inode_item, 0);
617 btrfs_set_stack_inode_nlink(&inode_item, 1);
618 btrfs_set_stack_inode_nbytes(&inode_item, root->leafsize);
619 btrfs_set_stack_inode_mode(&inode_item, S_IFDIR | 0755);
620 btrfs_set_stack_timespec_sec(&inode_item.atime, now);
621 btrfs_set_stack_timespec_nsec(&inode_item.atime, 0);
622 btrfs_set_stack_timespec_sec(&inode_item.ctime, now);
623 btrfs_set_stack_timespec_nsec(&inode_item.ctime, 0);
624 btrfs_set_stack_timespec_sec(&inode_item.mtime, now);
625 btrfs_set_stack_timespec_nsec(&inode_item.mtime, 0);
626 btrfs_set_stack_timespec_sec(&inode_item.otime, 0);
627 btrfs_set_stack_timespec_nsec(&inode_item.otime, 0);
629 if (root->fs_info->tree_root == root)
630 btrfs_set_super_root_dir(root->fs_info->super_copy, objectid);
632 ret = btrfs_insert_inode(trans, root, objectid, &inode_item);
636 ret = btrfs_insert_inode_ref(trans, root, "..", 2, objectid, objectid, 0);
640 btrfs_set_root_dirid(&root->root_item, objectid);
647 * checks if a path is a block device node
648 * Returns negative errno on failure, otherwise
649 * returns 1 for blockdev, 0 for not-blockdev
651 int is_block_device(const char *path) {
654 if (stat(path, &statbuf) < 0)
657 return S_ISBLK(statbuf.st_mode);
661 * Find the mount point for a mounted device.
662 * On success, returns 0 with mountpoint in *mp.
663 * On failure, returns -errno (not mounted yields -EINVAL)
664 * Is noisy on failures, expects to be given a mounted device.
666 int get_btrfs_mount(const char *dev, char *mp, size_t mp_size) {
670 ret = is_block_device(dev);
673 fprintf(stderr, "%s is not a block device\n", dev);
676 fprintf(stderr, "Could not check %s: %s\n",
677 dev, strerror(-ret));
682 fd = open(dev, O_RDONLY);
685 fprintf(stderr, "Could not open %s: %s\n", dev, strerror(errno));
689 ret = check_mounted_where(fd, dev, mp, mp_size, NULL);
691 fprintf(stderr, "%s is not a mounted btrfs device\n", dev);
693 } else { /* mounted, all good */
700 fprintf(stderr, "Could not get mountpoint for %s\n", dev);
705 * Given a pathname, return a filehandle to:
706 * the original pathname or,
707 * if the pathname is a mounted btrfs device, to its mountpoint.
709 * On error, return -1, errno should be set.
711 int open_path_or_dev_mnt(const char *path)
713 char mp[BTRFS_PATH_NAME_MAX + 1];
716 if (is_block_device(path)) {
719 ret = get_btrfs_mount(path, mp, sizeof(mp));
721 /* not a mounted btrfs dev */
725 fdmnt = open_file_or_dir(mp);
727 fdmnt = open_file_or_dir(path);
733 /* checks if a device is a loop device */
734 int is_loop_device (const char* device) {
737 if(stat(device, &statbuf) < 0)
740 return (S_ISBLK(statbuf.st_mode) &&
741 MAJOR(statbuf.st_rdev) == LOOP_MAJOR);
745 /* Takes a loop device path (e.g. /dev/loop0) and returns
746 * the associated file (e.g. /images/my_btrfs.img) */
747 int resolve_loop_device(const char* loop_dev, char* loop_file, int max_len)
753 char real_loop_dev[PATH_MAX];
755 if (!realpath(loop_dev, real_loop_dev))
757 snprintf(p, PATH_MAX, "/sys/block/%s/loop/backing_file", strrchr(real_loop_dev, '/'));
758 if (!(f = fopen(p, "r")))
761 snprintf(fmt, 20, "%%%i[^\n]", max_len-1);
762 ret = fscanf(f, fmt, loop_file);
770 /* Checks whether a and b are identical or device
771 * files associated with the same block device
773 int is_same_blk_file(const char* a, const char* b)
775 struct stat st_buf_a, st_buf_b;
776 char real_a[PATH_MAX];
777 char real_b[PATH_MAX];
779 if(!realpath(a, real_a) ||
780 !realpath(b, real_b))
785 /* Identical path? */
786 if(strcmp(real_a, real_b) == 0)
789 if(stat(a, &st_buf_a) < 0 ||
790 stat(b, &st_buf_b) < 0)
797 /* Same blockdevice? */
798 if(S_ISBLK(st_buf_a.st_mode) &&
799 S_ISBLK(st_buf_b.st_mode) &&
800 st_buf_a.st_rdev == st_buf_b.st_rdev)
806 if (st_buf_a.st_dev == st_buf_b.st_dev &&
807 st_buf_a.st_ino == st_buf_b.st_ino)
815 /* checks if a and b are identical or device
816 * files associated with the same block device or
817 * if one file is a loop device that uses the other
820 int is_same_loop_file(const char* a, const char* b)
822 char res_a[PATH_MAX];
823 char res_b[PATH_MAX];
828 /* Resolve a if it is a loop device */
829 if((ret = is_loop_device(a)) < 0) {
834 if ((ret = resolve_loop_device(a, res_a, sizeof(res_a))) < 0)
842 /* Resolve b if it is a loop device */
843 if ((ret = is_loop_device(b)) < 0) {
848 if((ret = resolve_loop_device(b, res_b, sizeof(res_b))) < 0)
856 return is_same_blk_file(final_a, final_b);
859 /* Checks if a file exists and is a block or regular file*/
860 int is_existing_blk_or_reg_file(const char* filename)
864 if(stat(filename, &st_buf) < 0) {
871 return (S_ISBLK(st_buf.st_mode) || S_ISREG(st_buf.st_mode));
874 /* Checks if a file is used (directly or indirectly via a loop device)
875 * by a device in fs_devices
877 int blk_file_in_dev_list(struct btrfs_fs_devices* fs_devices, const char* file)
880 struct list_head *head;
881 struct list_head *cur;
882 struct btrfs_device *device;
884 head = &fs_devices->devices;
885 list_for_each(cur, head) {
886 device = list_entry(cur, struct btrfs_device, dev_list);
888 if((ret = is_same_loop_file(device->name, file)))
896 * returns 1 if the device was mounted, < 0 on error or 0 if everything
897 * is safe to continue.
899 int check_mounted(const char* file)
904 fd = open(file, O_RDONLY);
906 fprintf (stderr, "check_mounted(): Could not open %s\n", file);
910 ret = check_mounted_where(fd, file, NULL, 0, NULL);
916 int check_mounted_where(int fd, const char *file, char *where, int size,
917 struct btrfs_fs_devices **fs_dev_ret)
922 struct btrfs_fs_devices *fs_devices_mnt = NULL;
926 /* scan the initial device */
927 ret = btrfs_scan_one_device(fd, file, &fs_devices_mnt,
928 &total_devs, BTRFS_SUPER_INFO_OFFSET);
929 is_btrfs = (ret >= 0);
931 /* scan other devices */
932 if (is_btrfs && total_devs > 1) {
933 if((ret = btrfs_scan_for_fsid(fs_devices_mnt, total_devs, 1)))
937 /* iterate over the list of currently mountes filesystems */
938 if ((f = setmntent ("/proc/mounts", "r")) == NULL)
941 while ((mnt = getmntent (f)) != NULL) {
943 if(strcmp(mnt->mnt_type, "btrfs") != 0)
946 ret = blk_file_in_dev_list(fs_devices_mnt, mnt->mnt_fsname);
948 /* ignore entries in the mount table that are not
949 associated with a file*/
950 if((ret = is_existing_blk_or_reg_file(mnt->mnt_fsname)) < 0)
951 goto out_mntloop_err;
955 ret = is_same_loop_file(file, mnt->mnt_fsname);
959 goto out_mntloop_err;
964 /* Did we find an entry in mnt table? */
965 if (mnt && size && where) {
966 strncpy(where, mnt->mnt_dir, size);
970 *fs_dev_ret = fs_devices_mnt;
981 struct list_head list;
985 void btrfs_register_one_device(char *fname)
987 struct btrfs_ioctl_vol_args args;
992 fd = open("/dev/btrfs-control", O_RDONLY);
994 fprintf(stderr, "failed to open /dev/btrfs-control "
995 "skipping device registration: %s\n",
999 strncpy(args.name, fname, BTRFS_PATH_NAME_MAX);
1000 args.name[BTRFS_PATH_NAME_MAX-1] = 0;
1001 ret = ioctl(fd, BTRFS_IOC_SCAN_DEV, &args);
1004 fprintf(stderr, "ERROR: device scan failed '%s' - %s\n",
1005 fname, strerror(e));
1010 int btrfs_scan_one_dir(char *dirname, int run_ioctl)
1013 struct dirent *dirent;
1014 struct pending_dir *pending;
1020 struct list_head pending_list;
1021 struct btrfs_fs_devices *tmp_devices;
1024 INIT_LIST_HEAD(&pending_list);
1026 pending = malloc(sizeof(*pending));
1029 strcpy(pending->name, dirname);
1032 dirname_len = strlen(pending->name);
1033 fullpath = malloc(PATH_MAX);
1034 dirname = pending->name;
1040 dirp = opendir(dirname);
1042 fprintf(stderr, "Unable to open %s for scanning\n", dirname);
1047 dirent = readdir(dirp);
1050 if (dirent->d_name[0] == '.')
1052 if (dirname_len + strlen(dirent->d_name) + 2 > PATH_MAX) {
1056 snprintf(fullpath, PATH_MAX, "%s/%s", dirname, dirent->d_name);
1057 ret = lstat(fullpath, &st);
1059 fprintf(stderr, "failed to stat %s\n", fullpath);
1062 if (S_ISLNK(st.st_mode))
1064 if (S_ISDIR(st.st_mode)) {
1065 struct pending_dir *next = malloc(sizeof(*next));
1070 strcpy(next->name, fullpath);
1071 list_add_tail(&next->list, &pending_list);
1073 if (!S_ISBLK(st.st_mode)) {
1076 fd = open(fullpath, O_RDONLY);
1078 /* ignore the following errors:
1079 ENXIO (device don't exists)
1080 ENOMEDIUM (No medium found ->
1081 like a cd tray empty)
1083 if(errno != ENXIO && errno != ENOMEDIUM)
1084 fprintf(stderr, "failed to read %s: %s\n",
1085 fullpath, strerror(errno));
1088 ret = btrfs_scan_one_device(fd, fullpath, &tmp_devices,
1090 BTRFS_SUPER_INFO_OFFSET);
1091 if (ret == 0 && run_ioctl > 0) {
1092 btrfs_register_one_device(fullpath);
1096 if (!list_empty(&pending_list)) {
1098 pending = list_entry(pending_list.next, struct pending_dir,
1101 list_del(&pending->list);
1115 int btrfs_scan_for_fsid(struct btrfs_fs_devices *fs_devices, u64 total_devs,
1120 ret = btrfs_scan_block_devices(run_ioctls);
1122 ret = btrfs_scan_one_dir("/dev", run_ioctls);
1126 int btrfs_device_already_in_root(struct btrfs_root *root, int fd,
1129 struct btrfs_super_block *disk_super;
1133 buf = malloc(BTRFS_SUPER_INFO_SIZE);
1138 ret = pread(fd, buf, BTRFS_SUPER_INFO_SIZE, super_offset);
1139 if (ret != BTRFS_SUPER_INFO_SIZE)
1143 disk_super = (struct btrfs_super_block *)buf;
1144 if (btrfs_super_magic(disk_super) != BTRFS_MAGIC)
1147 if (!memcmp(disk_super->fsid, root->fs_info->super_copy->fsid,
1156 static char *size_strs[] = { "", "KB", "MB", "GB", "TB",
1158 void pretty_size_snprintf(u64 size, char *str, size_t str_bytes)
1170 u64 last_size = size;
1172 while(size >= 1024){
1178 if (num_divs >= ARRAY_SIZE(size_strs)) {
1182 fraction = (float)last_size / 1024;
1184 snprintf(str, str_bytes, "%.2f%s", fraction, size_strs[num_divs]);
1188 * __strncpy__null - strncpy with null termination
1189 * @dest: the target array
1190 * @src: the source string
1191 * @n: maximum bytes to copy (size of *dest)
1193 * Like strncpy, but ensures destination is null-terminated.
1195 * Copies the string pointed to by src, including the terminating null
1196 * byte ('\0'), to the buffer pointed to by dest, up to a maximum
1197 * of n bytes. Then ensure that dest is null-terminated.
1199 char *__strncpy__null(char *dest, const char *src, size_t n)
1201 strncpy(dest, src, n);
1208 * Checks to make sure that the label matches our requirements.
1210 0 if everything is safe and usable
1211 -1 if the label is too long
1213 static int check_label(const char *input)
1215 int len = strlen(input);
1217 if (len > BTRFS_LABEL_SIZE - 1) {
1218 fprintf(stderr, "ERROR: Label %s is too long (max %d)\n",
1219 input, BTRFS_LABEL_SIZE - 1);
1226 static int set_label_unmounted(const char *dev, const char *label)
1228 struct btrfs_trans_handle *trans;
1229 struct btrfs_root *root;
1232 ret = check_mounted(dev);
1234 fprintf(stderr, "FATAL: error checking %s mount status\n", dev);
1238 fprintf(stderr, "ERROR: dev %s is mounted, use mount point\n",
1243 /* Open the super_block at the default location
1244 * and as read-write.
1246 root = open_ctree(dev, 0, 1);
1247 if (!root) /* errors are printed by open_ctree() */
1250 trans = btrfs_start_transaction(root, 1);
1251 snprintf(root->fs_info->super_copy->label, BTRFS_LABEL_SIZE, "%s",
1253 btrfs_commit_transaction(trans, root);
1255 /* Now we close it since we are done. */
1260 static int set_label_mounted(const char *mount_path, const char *label)
1264 fd = open(mount_path, O_RDONLY | O_NOATIME);
1266 fprintf(stderr, "ERROR: unable access to '%s'\n", mount_path);
1270 if (ioctl(fd, BTRFS_IOC_SET_FSLABEL, label) < 0) {
1271 fprintf(stderr, "ERROR: unable to set label %s\n",
1281 static int get_label_unmounted(const char *dev)
1283 struct btrfs_root *root;
1286 ret = check_mounted(dev);
1288 fprintf(stderr, "FATAL: error checking %s mount status\n", dev);
1292 fprintf(stderr, "ERROR: dev %s is mounted, use mount point\n",
1297 /* Open the super_block at the default location
1300 root = open_ctree(dev, 0, 0);
1304 fprintf(stdout, "%s\n", root->fs_info->super_copy->label);
1306 /* Now we close it since we are done. */
1312 * If a partition is mounted, try to get the filesystem label via its
1313 * mounted path rather than device. Return the corresponding error
1314 * the user specified the device path.
1316 static int get_label_mounted(const char *mount_path)
1318 char label[BTRFS_LABEL_SIZE];
1321 fd = open(mount_path, O_RDONLY | O_NOATIME);
1323 fprintf(stderr, "ERROR: unable access to '%s'\n", mount_path);
1327 memset(label, '\0', sizeof(label));
1328 if (ioctl(fd, BTRFS_IOC_GET_FSLABEL, label) < 0) {
1329 fprintf(stderr, "ERROR: unable get label %s\n", strerror(errno));
1334 fprintf(stdout, "%s\n", label);
1339 int get_label(const char *btrfs_dev)
1341 return is_existing_blk_or_reg_file(btrfs_dev) ?
1342 get_label_unmounted(btrfs_dev) :
1343 get_label_mounted(btrfs_dev);
1346 int set_label(const char *btrfs_dev, const char *label)
1348 if (check_label(label))
1351 return is_existing_blk_or_reg_file(btrfs_dev) ?
1352 set_label_unmounted(btrfs_dev, label) :
1353 set_label_mounted(btrfs_dev, label);
1356 int btrfs_scan_block_devices(int run_ioctl)
1362 struct btrfs_fs_devices *tmp_devices;
1364 FILE *proc_partitions;
1372 proc_partitions = fopen("/proc/partitions","r");
1373 if (!proc_partitions) {
1374 fprintf(stderr, "Unable to open '/proc/partitions' for scanning\n");
1377 /* skip the header */
1378 for (i = 0; i < 2; i++)
1379 if (!fgets(buf, 1023, proc_partitions)) {
1381 "Unable to read '/proc/partitions' for scanning\n");
1382 fclose(proc_partitions);
1386 strcpy(fullpath,"/dev/");
1387 while(fgets(buf, 1023, proc_partitions)) {
1388 i = sscanf(buf," %*d %*d %*d %99s", fullpath+5);
1391 * multipath and MD devices may register as a btrfs filesystem
1392 * both through the original block device and through
1393 * the special (/dev/mapper or /dev/mdX) entry.
1394 * This scans the special entries last
1396 special = strncmp(fullpath, "/dev/dm-", strlen("/dev/dm-")) == 0;
1398 special = strncmp(fullpath, "/dev/md", strlen("/dev/md")) == 0;
1400 if (scans == 0 && special)
1402 if (scans > 0 && !special)
1405 ret = lstat(fullpath, &st);
1407 fprintf(stderr, "failed to stat %s\n", fullpath);
1410 if (!S_ISBLK(st.st_mode)) {
1414 fd = open(fullpath, O_RDONLY);
1416 fprintf(stderr, "failed to open %s: %s\n",
1417 fullpath, strerror(errno));
1420 ret = btrfs_scan_one_device(fd, fullpath, &tmp_devices,
1422 BTRFS_SUPER_INFO_OFFSET);
1423 if (ret == 0 && run_ioctl > 0) {
1424 btrfs_register_one_device(fullpath);
1429 fclose(proc_partitions);
1438 u64 parse_size(char *s)
1444 for (i = 0; s && s[i] && isdigit(s[i]); i++) ;
1446 fprintf(stderr, "ERROR: size value is empty\n");
1468 fprintf(stderr, "ERROR: Unknown size descriptor "
1473 if (s[i] && s[i+1]) {
1474 fprintf(stderr, "ERROR: Illegal suffix contains "
1475 "character '%c' in wrong position\n",
1479 return strtoull(s, NULL, 10) * mult;
1482 int open_file_or_dir(const char *fname)
1489 ret = stat(fname, &st);
1493 if (S_ISDIR(st.st_mode)) {
1494 dirstream = opendir(fname);
1498 fd = dirfd(dirstream);
1500 fd = open(fname, O_RDWR);
1508 int get_device_info(int fd, u64 devid,
1509 struct btrfs_ioctl_dev_info_args *di_args)
1513 di_args->devid = devid;
1514 memset(&di_args->uuid, '\0', sizeof(di_args->uuid));
1516 ret = ioctl(fd, BTRFS_IOC_DEV_INFO, di_args);
1517 return ret ? -errno : 0;
1521 * For a given path, fill in the ioctl fs_ and info_ args.
1522 * If the path is a btrfs mountpoint, fill info for all devices.
1523 * If the path is a btrfs device, fill in only that device.
1525 * The path provided must be either on a mounted btrfs fs,
1526 * or be a mounted btrfs device.
1528 * Returns 0 on success, or a negative errno.
1530 int get_fs_info(char *path, struct btrfs_ioctl_fs_info_args *fi_args,
1531 struct btrfs_ioctl_dev_info_args **di_ret)
1537 struct btrfs_fs_devices *fs_devices_mnt = NULL;
1538 struct btrfs_ioctl_dev_info_args *di_args;
1539 char mp[BTRFS_PATH_NAME_MAX + 1];
1541 memset(fi_args, 0, sizeof(*fi_args));
1543 if (is_block_device(path)) {
1544 /* Ensure it's mounted, then set path to the mountpoint */
1545 fd = open(path, O_RDONLY);
1548 fprintf(stderr, "Couldn't open %s: %s\n",
1549 path, strerror(errno));
1552 ret = check_mounted_where(fd, path, mp, sizeof(mp),
1561 /* Only fill in this one device */
1562 fi_args->num_devices = 1;
1563 fi_args->max_id = fs_devices_mnt->latest_devid;
1564 i = fs_devices_mnt->latest_devid;
1565 memcpy(fi_args->fsid, fs_devices_mnt->fsid, BTRFS_FSID_SIZE);
1569 /* at this point path must not be for a block device */
1570 fd = open_file_or_dir(path);
1576 /* fill in fi_args if not just a single device */
1577 if (fi_args->num_devices != 1) {
1578 ret = ioctl(fd, BTRFS_IOC_FS_INFO, fi_args);
1585 if (!fi_args->num_devices)
1588 di_args = *di_ret = malloc(fi_args->num_devices * sizeof(*di_args));
1594 for (; i <= fi_args->max_id; ++i) {
1595 BUG_ON(ndevs >= fi_args->num_devices);
1596 ret = get_device_info(fd, i, &di_args[ndevs]);
1612 #define isoctal(c) (((c) & ~7) == '0')
1614 static inline void translate(char *f, char *t)
1616 while (*f != '\0') {
1618 isoctal(f[1]) && isoctal(f[2]) && isoctal(f[3])) {
1619 *t++ = 64*(f[1] & 7) + 8*(f[2] & 7) + (f[3] & 7);
1629 * Checks if the swap device.
1630 * Returns 1 if swap device, < 0 on error or 0 if not swap device.
1632 int is_swap_device(const char *file)
1643 if (stat(file, &st_buf) < 0)
1645 if (S_ISBLK(st_buf.st_mode))
1646 dev = st_buf.st_rdev;
1647 else if (S_ISREG(st_buf.st_mode)) {
1648 dev = st_buf.st_dev;
1649 ino = st_buf.st_ino;
1653 if ((f = fopen("/proc/swaps", "r")) == NULL)
1656 /* skip the first line */
1657 if (fgets(tmp, sizeof(tmp), f) == NULL)
1660 while (fgets(tmp, sizeof(tmp), f) != NULL) {
1661 if ((cp = strchr(tmp, ' ')) != NULL)
1663 if ((cp = strchr(tmp, '\t')) != NULL)
1665 translate(tmp, buf);
1666 if (stat(buf, &st_buf) != 0)
1668 if (S_ISBLK(st_buf.st_mode)) {
1669 if (dev == st_buf.st_rdev) {
1673 } else if (S_ISREG(st_buf.st_mode)) {
1674 if (dev == st_buf.st_dev && ino == st_buf.st_ino) {
1688 * Check for existing filesystem or partition table on device.
1690 * 1 for existing fs or partition
1691 * 0 for nothing found
1692 * -1 for internal error
1699 blkid_probe pr = NULL;
1703 if (!device || !*device)
1706 ret = -1; /* will reset on success of all setup calls */
1708 pr = blkid_new_probe_from_filename(device);
1712 size = blkid_probe_get_size(pr);
1716 /* nothing to overwrite on a 0-length device */
1722 ret = blkid_probe_enable_partitions(pr, 1);
1726 ret = blkid_do_fullprobe(pr);
1731 * Blkid returns 1 for nothing found and 0 when it finds a signature,
1732 * but we want the exact opposite, so reverse the return value here.
1734 * In addition print some useful diagnostics about what actually is
1742 if (!blkid_probe_lookup_value(pr, "TYPE", &type, NULL)) {
1744 "%s appears to contain an existing "
1745 "filesystem (%s).\n", device, type);
1746 } else if (!blkid_probe_lookup_value(pr, "PTTYPE", &type, NULL)) {
1748 "%s appears to contain a partition "
1749 "table (%s).\n", device, type);
1752 "%s appears to contain something weird "
1753 "according to blkid\n", device);
1759 blkid_free_probe(pr);
1762 "probe of %s failed, cannot detect "
1763 "existing filesystem.\n", device);
1767 /* Check if disk is suitable for btrfs
1769 * 1: something is wrong, estr provides the error
1772 int test_dev_for_mkfs(char *file, int force_overwrite, char *estr)
1777 ret = is_swap_device(file);
1779 snprintf(estr, sz, "error checking %s status: %s\n", file,
1784 snprintf(estr, sz, "%s is a swap device\n", file);
1787 if (!force_overwrite) {
1788 if (check_overwrite(file)) {
1789 snprintf(estr, sz, "Use the -f option to force overwrite.\n");
1793 ret = check_mounted(file);
1795 snprintf(estr, sz, "error checking %s mount status\n",
1800 snprintf(estr, sz, "%s is mounted\n", file);
1803 /* check if the device is busy */
1804 fd = open(file, O_RDWR|O_EXCL);
1806 snprintf(estr, sz, "unable to open %s: %s\n", file,