1 /*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
4 This file is part of systemd.
6 Copyright 2014 Lennart Poettering
8 systemd is free software; you can redistribute it and/or modify it
9 under the terms of the GNU Lesser General Public License as published by
10 the Free Software Foundation; either version 2.1 of the License, or
11 (at your option) any later version.
13 systemd is distributed in the hope that it will be useful, but
14 WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 Lesser General Public License for more details.
18 You should have received a copy of the GNU Lesser General Public License
19 along with systemd; If not, see <http://www.gnu.org/licenses/>.
26 #ifdef HAVE_LINUX_BTRFS_H
27 #include <linux/btrfs.h>
32 #include "path-util.h"
36 #include "selinux-util.h"
37 #include "smack-util.h"
38 #include "btrfs-ctree.h"
39 #include "btrfs-util.h"
41 static int validate_subvolume_name(const char *name) {
43 if (!filename_is_valid(name))
46 if (strlen(name) > BTRFS_SUBVOL_NAME_MAX)
52 static int open_parent(const char *path, int flags) {
53 _cleanup_free_ char *parent = NULL;
58 r = path_get_parent(path, &parent);
62 fd = open(parent, flags);
69 static int extract_subvolume_name(const char *path, const char **subvolume) {
78 r = validate_subvolume_name(fn);
86 int btrfs_is_snapshot(int fd) {
90 /* On btrfs subvolumes always have the inode 256 */
92 if (fstat(fd, &st) < 0)
95 if (!S_ISDIR(st.st_mode) || st.st_ino != 256)
98 if (fstatfs(fd, &sfs) < 0)
101 return F_TYPE_EQUAL(sfs.f_type, BTRFS_SUPER_MAGIC);
104 int btrfs_subvol_snapshot(const char *old_path, const char *new_path, bool read_only, bool fallback_copy) {
105 struct btrfs_ioctl_vol_args_v2 args = {
106 .flags = read_only ? BTRFS_SUBVOL_RDONLY : 0,
108 _cleanup_close_ int old_fd = -1, new_fd = -1;
109 const char *subvolume;
114 old_fd = open(old_path, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY);
118 r = btrfs_is_snapshot(old_fd);
124 r = btrfs_subvol_make(new_path);
128 r = copy_directory_fd(old_fd, new_path, true);
130 btrfs_subvol_remove(new_path);
135 r = btrfs_subvol_set_read_only(new_path, true);
137 btrfs_subvol_remove(new_path);
148 r = extract_subvolume_name(new_path, &subvolume);
152 new_fd = open_parent(new_path, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY);
156 strncpy(args.name, subvolume, sizeof(args.name)-1);
159 if (ioctl(new_fd, BTRFS_IOC_SNAP_CREATE_V2, &args) < 0)
165 int btrfs_subvol_make(const char *path) {
166 struct btrfs_ioctl_vol_args args = {};
167 _cleanup_close_ int fd = -1;
168 const char *subvolume;
173 r = extract_subvolume_name(path, &subvolume);
177 fd = open_parent(path, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY);
181 strncpy(args.name, subvolume, sizeof(args.name)-1);
183 if (ioctl(fd, BTRFS_IOC_SUBVOL_CREATE, &args) < 0)
189 int btrfs_subvol_make_label(const char *path) {
194 r = mac_selinux_create_file_prepare(path, S_IFDIR);
198 r = btrfs_subvol_make(path);
199 mac_selinux_create_file_clear();
204 return mac_smack_fix(path, false, false);
207 int btrfs_subvol_remove(const char *path) {
208 struct btrfs_ioctl_vol_args args = {};
209 _cleanup_close_ int fd = -1;
210 const char *subvolume;
215 r = extract_subvolume_name(path, &subvolume);
219 fd = open_parent(path, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY);
223 strncpy(args.name, subvolume, sizeof(args.name)-1);
225 if (ioctl(fd, BTRFS_IOC_SNAP_DESTROY, &args) < 0)
231 int btrfs_subvol_set_read_only_fd(int fd, bool b) {
232 uint64_t flags, nflags;
237 if (fstat(fd, &st) < 0)
240 if (!S_ISDIR(st.st_mode) || st.st_ino != 256)
243 if (ioctl(fd, BTRFS_IOC_SUBVOL_GETFLAGS, &flags) < 0)
247 nflags = flags | BTRFS_SUBVOL_RDONLY;
249 nflags = flags & ~BTRFS_SUBVOL_RDONLY;
254 if (ioctl(fd, BTRFS_IOC_SUBVOL_SETFLAGS, &nflags) < 0)
260 int btrfs_subvol_set_read_only(const char *path, bool b) {
261 _cleanup_close_ int fd = -1;
263 fd = open(path, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY);
267 return btrfs_subvol_set_read_only_fd(fd, b);
270 int btrfs_subvol_get_read_only_fd(int fd) {
273 if (ioctl(fd, BTRFS_IOC_SUBVOL_GETFLAGS, &flags) < 0)
276 return !!(flags & BTRFS_SUBVOL_RDONLY);
279 int btrfs_reflink(int infd, int outfd) {
285 r = ioctl(outfd, BTRFS_IOC_CLONE, infd);
292 int btrfs_clone_range(int infd, uint64_t in_offset, int outfd, uint64_t out_offset, uint64_t sz) {
293 struct btrfs_ioctl_clone_range_args args = {
295 .src_offset = in_offset,
297 .dest_offset = out_offset,
305 r = ioctl(outfd, BTRFS_IOC_CLONE_RANGE, &args);
312 int btrfs_get_block_device(const char *path, dev_t *dev) {
313 struct btrfs_ioctl_fs_info_args fsi = {};
314 _cleanup_close_ int fd = -1;
320 fd = open(path, O_RDONLY|O_NOCTTY|O_CLOEXEC|O_DIRECTORY);
324 if (ioctl(fd, BTRFS_IOC_FS_INFO, &fsi) < 0)
327 /* We won't do this for btrfs RAID */
328 if (fsi.num_devices != 1)
331 for (id = 1; id <= fsi.max_id; id++) {
332 struct btrfs_ioctl_dev_info_args di = {
337 if (ioctl(fd, BTRFS_IOC_DEV_INFO, &di) < 0) {
344 if (stat((char*) di.path, &st) < 0)
347 if (!S_ISBLK(st.st_mode))
350 if (major(st.st_rdev) == 0)
360 int btrfs_subvol_get_id_fd(int fd, uint64_t *ret) {
361 struct btrfs_ioctl_ino_lookup_args args = {
362 .objectid = BTRFS_FIRST_FREE_OBJECTID
368 if (ioctl(fd, BTRFS_IOC_INO_LOOKUP, &args) < 0)
375 static bool btrfs_ioctl_search_args_inc(struct btrfs_ioctl_search_args *args) {
378 /* the objectid, type, offset together make up the btrfs key,
379 * which is considered a single 136byte integer when
380 * comparing. This call increases the counter by one, dealing
381 * with the overflow between the overflows */
383 if (args->key.min_offset < (uint64_t) -1) {
384 args->key.min_offset++;
388 if (args->key.min_type < (uint8_t) -1) {
389 args->key.min_type++;
390 args->key.min_offset = 0;
394 if (args->key.min_objectid < (uint64_t) -1) {
395 args->key.min_objectid++;
396 args->key.min_offset = 0;
397 args->key.min_type = 0;
404 static void btrfs_ioctl_search_args_set(struct btrfs_ioctl_search_args *args, const struct btrfs_ioctl_search_header *h) {
408 args->key.min_objectid = h->objectid;
409 args->key.min_type = h->type;
410 args->key.min_offset = h->offset;
413 static int btrfs_ioctl_search_args_compare(const struct btrfs_ioctl_search_args *args) {
416 /* Compare min and max */
418 if (args->key.min_objectid < args->key.max_objectid)
420 if (args->key.min_objectid > args->key.max_objectid)
423 if (args->key.min_type < args->key.max_type)
425 if (args->key.min_type > args->key.max_type)
428 if (args->key.min_offset < args->key.max_offset)
430 if (args->key.min_offset > args->key.max_offset)
436 #define FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) \
438 (sh) = (const struct btrfs_ioctl_search_header*) (args).buf; \
439 (i) < (args).key.nr_items; \
441 (sh) = (const struct btrfs_ioctl_search_header*) ((uint8_t*) (sh) + sizeof(struct btrfs_ioctl_search_header) + (sh)->len))
443 #define BTRFS_IOCTL_SEARCH_HEADER_BODY(sh) \
444 ((void*) ((uint8_t*) sh + sizeof(struct btrfs_ioctl_search_header)))
446 int btrfs_subvol_get_info_fd(int fd, BtrfsSubvolInfo *ret) {
447 struct btrfs_ioctl_search_args args = {
448 /* Tree of tree roots */
449 .key.tree_id = BTRFS_ROOT_TREE_OBJECTID,
451 /* Look precisely for the subvolume items */
452 .key.min_type = BTRFS_ROOT_ITEM_KEY,
453 .key.max_type = BTRFS_ROOT_ITEM_KEY,
456 .key.max_offset = (uint64_t) -1,
458 /* No restrictions on the other components */
459 .key.min_transid = 0,
460 .key.max_transid = (uint64_t) -1,
470 r = btrfs_subvol_get_id_fd(fd, &subvol_id);
474 args.key.min_objectid = args.key.max_objectid = subvol_id;
476 while (btrfs_ioctl_search_args_compare(&args) <= 0) {
477 const struct btrfs_ioctl_search_header *sh;
480 args.key.nr_items = 256;
481 if (ioctl(fd, BTRFS_IOC_TREE_SEARCH, &args) < 0)
484 if (args.key.nr_items <= 0)
487 FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) {
489 const struct btrfs_root_item *ri;
491 /* Make sure we start the next search at least from this entry */
492 btrfs_ioctl_search_args_set(&args, sh);
494 if (sh->objectid != subvol_id)
496 if (sh->type != BTRFS_ROOT_ITEM_KEY)
499 /* Older versions of the struct lacked the otime setting */
500 if (sh->len < offsetof(struct btrfs_root_item, otime) + sizeof(struct btrfs_timespec))
503 ri = BTRFS_IOCTL_SEARCH_HEADER_BODY(sh);
505 ret->otime = (usec_t) le64toh(ri->otime.sec) * USEC_PER_SEC +
506 (usec_t) le32toh(ri->otime.nsec) / NSEC_PER_USEC;
508 ret->subvol_id = subvol_id;
509 ret->read_only = !!(le64toh(ri->flags) & BTRFS_ROOT_SUBVOL_RDONLY);
511 assert_cc(sizeof(ri->uuid) == sizeof(ret->uuid));
512 memcpy(&ret->uuid, ri->uuid, sizeof(ret->uuid));
513 memcpy(&ret->parent_uuid, ri->parent_uuid, sizeof(ret->parent_uuid));
519 /* Increase search key by one, to read the next item, if we can. */
520 if (!btrfs_ioctl_search_args_inc(&args))
531 int btrfs_subvol_get_quota_fd(int fd, BtrfsQuotaInfo *ret) {
533 struct btrfs_ioctl_search_args args = {
534 /* Tree of quota items */
535 .key.tree_id = BTRFS_QUOTA_TREE_OBJECTID,
537 /* The object ID is always 0 */
538 .key.min_objectid = 0,
539 .key.max_objectid = 0,
541 /* Look precisely for the quota items */
542 .key.min_type = BTRFS_QGROUP_STATUS_KEY,
543 .key.max_type = BTRFS_QGROUP_LIMIT_KEY,
545 /* No restrictions on the other components */
546 .key.min_transid = 0,
547 .key.max_transid = (uint64_t) -1,
551 bool found_info = false, found_limit = false;
557 r = btrfs_subvol_get_id_fd(fd, &subvol_id);
561 args.key.min_offset = args.key.max_offset = subvol_id;
563 while (btrfs_ioctl_search_args_compare(&args) <= 0) {
564 const struct btrfs_ioctl_search_header *sh;
567 args.key.nr_items = 256;
568 if (ioctl(fd, BTRFS_IOC_TREE_SEARCH, &args) < 0)
571 if (args.key.nr_items <= 0)
574 FOREACH_BTRFS_IOCTL_SEARCH_HEADER(i, sh, args) {
576 /* Make sure we start the next search at least from this entry */
577 btrfs_ioctl_search_args_set(&args, sh);
579 if (sh->objectid != 0)
581 if (sh->offset != subvol_id)
584 if (sh->type == BTRFS_QGROUP_INFO_KEY) {
585 const struct btrfs_qgroup_info_item *qii = BTRFS_IOCTL_SEARCH_HEADER_BODY(sh);
587 ret->referred = le64toh(qii->rfer);
588 ret->exclusive = le64toh(qii->excl);
592 } else if (sh->type == BTRFS_QGROUP_LIMIT_KEY) {
593 const struct btrfs_qgroup_limit_item *qli = BTRFS_IOCTL_SEARCH_HEADER_BODY(sh);
595 ret->referred_max = le64toh(qli->max_rfer);
596 ret->exclusive_max = le64toh(qli->max_excl);
598 if (ret->referred_max == 0)
599 ret->referred_max = (uint64_t) -1;
600 if (ret->exclusive_max == 0)
601 ret->exclusive_max = (uint64_t) -1;
606 if (found_info && found_limit)
610 /* Increase search key by one, to read the next item, if we can. */
611 if (!btrfs_ioctl_search_args_inc(&args))
616 if (!found_limit && !found_info)
620 ret->referred = (uint64_t) -1;
621 ret->exclusive = (uint64_t) -1;
625 ret->referred_max = (uint64_t) -1;
626 ret->exclusive_max = (uint64_t) -1;
632 int btrfs_defrag_fd(int fd) {
635 if (ioctl(fd, BTRFS_IOC_DEFRAG, NULL) < 0)
641 int btrfs_defrag(const char *p) {
642 _cleanup_close_ int fd = -1;
644 fd = open(p, O_RDWR|O_CLOEXEC|O_NOCTTY|O_NOFOLLOW);
648 return btrfs_defrag_fd(fd);