1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (C) 2007 Oracle. All rights reserved.
6 #include <linux/kernel.h>
8 #include <linux/file.h>
10 #include <linux/fsnotify.h>
11 #include <linux/pagemap.h>
12 #include <linux/highmem.h>
13 #include <linux/time.h>
14 #include <linux/string.h>
15 #include <linux/backing-dev.h>
16 #include <linux/mount.h>
17 #include <linux/namei.h>
18 #include <linux/writeback.h>
19 #include <linux/compat.h>
20 #include <linux/security.h>
21 #include <linux/xattr.h>
23 #include <linux/slab.h>
24 #include <linux/blkdev.h>
25 #include <linux/uuid.h>
26 #include <linux/btrfs.h>
27 #include <linux/uaccess.h>
28 #include <linux/iversion.h>
32 #include "transaction.h"
33 #include "btrfs_inode.h"
34 #include "print-tree.h"
38 #include "rcu-string.h"
40 #include "dev-replace.h"
45 #include "compression.h"
46 #include "space-info.h"
47 #include "delalloc-space.h"
48 #include "block-group.h"
51 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
52 * structures are incorrect, as the timespec structure from userspace
53 * is 4 bytes too small. We define these alternatives here to teach
54 * the kernel about the 32-bit struct packing.
56 struct btrfs_ioctl_timespec_32 {
59 } __attribute__ ((__packed__));
61 struct btrfs_ioctl_received_subvol_args_32 {
62 char uuid[BTRFS_UUID_SIZE]; /* in */
63 __u64 stransid; /* in */
64 __u64 rtransid; /* out */
65 struct btrfs_ioctl_timespec_32 stime; /* in */
66 struct btrfs_ioctl_timespec_32 rtime; /* out */
68 __u64 reserved[16]; /* in */
69 } __attribute__ ((__packed__));
71 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
72 struct btrfs_ioctl_received_subvol_args_32)
75 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
76 struct btrfs_ioctl_send_args_32 {
77 __s64 send_fd; /* in */
78 __u64 clone_sources_count; /* in */
79 compat_uptr_t clone_sources; /* in */
80 __u64 parent_root; /* in */
82 __u64 reserved[4]; /* in */
83 } __attribute__ ((__packed__));
85 #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
86 struct btrfs_ioctl_send_args_32)
89 /* Mask out flags that are inappropriate for the given type of inode. */
90 static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode,
93 if (S_ISDIR(inode->i_mode))
95 else if (S_ISREG(inode->i_mode))
96 return flags & ~FS_DIRSYNC_FL;
98 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
102 * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS
105 static unsigned int btrfs_inode_flags_to_fsflags(unsigned int flags)
107 unsigned int iflags = 0;
109 if (flags & BTRFS_INODE_SYNC)
110 iflags |= FS_SYNC_FL;
111 if (flags & BTRFS_INODE_IMMUTABLE)
112 iflags |= FS_IMMUTABLE_FL;
113 if (flags & BTRFS_INODE_APPEND)
114 iflags |= FS_APPEND_FL;
115 if (flags & BTRFS_INODE_NODUMP)
116 iflags |= FS_NODUMP_FL;
117 if (flags & BTRFS_INODE_NOATIME)
118 iflags |= FS_NOATIME_FL;
119 if (flags & BTRFS_INODE_DIRSYNC)
120 iflags |= FS_DIRSYNC_FL;
121 if (flags & BTRFS_INODE_NODATACOW)
122 iflags |= FS_NOCOW_FL;
124 if (flags & BTRFS_INODE_NOCOMPRESS)
125 iflags |= FS_NOCOMP_FL;
126 else if (flags & BTRFS_INODE_COMPRESS)
127 iflags |= FS_COMPR_FL;
133 * Update inode->i_flags based on the btrfs internal flags.
135 void btrfs_sync_inode_flags_to_i_flags(struct inode *inode)
137 struct btrfs_inode *binode = BTRFS_I(inode);
138 unsigned int new_fl = 0;
140 if (binode->flags & BTRFS_INODE_SYNC)
142 if (binode->flags & BTRFS_INODE_IMMUTABLE)
143 new_fl |= S_IMMUTABLE;
144 if (binode->flags & BTRFS_INODE_APPEND)
146 if (binode->flags & BTRFS_INODE_NOATIME)
148 if (binode->flags & BTRFS_INODE_DIRSYNC)
151 set_mask_bits(&inode->i_flags,
152 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
156 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
158 struct btrfs_inode *binode = BTRFS_I(file_inode(file));
159 unsigned int flags = btrfs_inode_flags_to_fsflags(binode->flags);
161 if (copy_to_user(arg, &flags, sizeof(flags)))
167 * Check if @flags are a supported and valid set of FS_*_FL flags and that
168 * the old and new flags are not conflicting
170 static int check_fsflags(unsigned int old_flags, unsigned int flags)
172 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
173 FS_NOATIME_FL | FS_NODUMP_FL | \
174 FS_SYNC_FL | FS_DIRSYNC_FL | \
175 FS_NOCOMP_FL | FS_COMPR_FL |
179 /* COMPR and NOCOMP on new/old are valid */
180 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
183 if ((flags & FS_COMPR_FL) && (flags & FS_NOCOW_FL))
186 /* NOCOW and compression options are mutually exclusive */
187 if ((old_flags & FS_NOCOW_FL) && (flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
189 if ((flags & FS_NOCOW_FL) && (old_flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
195 static int check_fsflags_compatible(struct btrfs_fs_info *fs_info,
198 if (btrfs_is_zoned(fs_info) && (flags & FS_NOCOW_FL))
204 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
206 struct inode *inode = file_inode(file);
207 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
208 struct btrfs_inode *binode = BTRFS_I(inode);
209 struct btrfs_root *root = binode->root;
210 struct btrfs_trans_handle *trans;
211 unsigned int fsflags, old_fsflags;
213 const char *comp = NULL;
216 if (!inode_owner_or_capable(inode))
219 if (btrfs_root_readonly(root))
222 if (copy_from_user(&fsflags, arg, sizeof(fsflags)))
225 ret = mnt_want_write_file(file);
230 fsflags = btrfs_mask_fsflags_for_type(inode, fsflags);
231 old_fsflags = btrfs_inode_flags_to_fsflags(binode->flags);
233 ret = vfs_ioc_setflags_prepare(inode, old_fsflags, fsflags);
237 ret = check_fsflags(old_fsflags, fsflags);
241 ret = check_fsflags_compatible(fs_info, fsflags);
245 binode_flags = binode->flags;
246 if (fsflags & FS_SYNC_FL)
247 binode_flags |= BTRFS_INODE_SYNC;
249 binode_flags &= ~BTRFS_INODE_SYNC;
250 if (fsflags & FS_IMMUTABLE_FL)
251 binode_flags |= BTRFS_INODE_IMMUTABLE;
253 binode_flags &= ~BTRFS_INODE_IMMUTABLE;
254 if (fsflags & FS_APPEND_FL)
255 binode_flags |= BTRFS_INODE_APPEND;
257 binode_flags &= ~BTRFS_INODE_APPEND;
258 if (fsflags & FS_NODUMP_FL)
259 binode_flags |= BTRFS_INODE_NODUMP;
261 binode_flags &= ~BTRFS_INODE_NODUMP;
262 if (fsflags & FS_NOATIME_FL)
263 binode_flags |= BTRFS_INODE_NOATIME;
265 binode_flags &= ~BTRFS_INODE_NOATIME;
266 if (fsflags & FS_DIRSYNC_FL)
267 binode_flags |= BTRFS_INODE_DIRSYNC;
269 binode_flags &= ~BTRFS_INODE_DIRSYNC;
270 if (fsflags & FS_NOCOW_FL) {
271 if (S_ISREG(inode->i_mode)) {
273 * It's safe to turn csums off here, no extents exist.
274 * Otherwise we want the flag to reflect the real COW
275 * status of the file and will not set it.
277 if (inode->i_size == 0)
278 binode_flags |= BTRFS_INODE_NODATACOW |
279 BTRFS_INODE_NODATASUM;
281 binode_flags |= BTRFS_INODE_NODATACOW;
285 * Revert back under same assumptions as above
287 if (S_ISREG(inode->i_mode)) {
288 if (inode->i_size == 0)
289 binode_flags &= ~(BTRFS_INODE_NODATACOW |
290 BTRFS_INODE_NODATASUM);
292 binode_flags &= ~BTRFS_INODE_NODATACOW;
297 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
298 * flag may be changed automatically if compression code won't make
301 if (fsflags & FS_NOCOMP_FL) {
302 binode_flags &= ~BTRFS_INODE_COMPRESS;
303 binode_flags |= BTRFS_INODE_NOCOMPRESS;
304 } else if (fsflags & FS_COMPR_FL) {
306 if (IS_SWAPFILE(inode)) {
311 binode_flags |= BTRFS_INODE_COMPRESS;
312 binode_flags &= ~BTRFS_INODE_NOCOMPRESS;
314 comp = btrfs_compress_type2str(fs_info->compress_type);
315 if (!comp || comp[0] == 0)
316 comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
318 binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
325 trans = btrfs_start_transaction(root, 3);
327 ret = PTR_ERR(trans);
332 ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp,
335 btrfs_abort_transaction(trans, ret);
339 ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL,
341 if (ret && ret != -ENODATA) {
342 btrfs_abort_transaction(trans, ret);
347 binode->flags = binode_flags;
348 btrfs_sync_inode_flags_to_i_flags(inode);
349 inode_inc_iversion(inode);
350 inode->i_ctime = current_time(inode);
351 ret = btrfs_update_inode(trans, root, BTRFS_I(inode));
354 btrfs_end_transaction(trans);
357 mnt_drop_write_file(file);
362 * Translate btrfs internal inode flags to xflags as expected by the
363 * FS_IOC_FSGETXATT ioctl. Filter only the supported ones, unknown flags are
366 static unsigned int btrfs_inode_flags_to_xflags(unsigned int flags)
368 unsigned int xflags = 0;
370 if (flags & BTRFS_INODE_APPEND)
371 xflags |= FS_XFLAG_APPEND;
372 if (flags & BTRFS_INODE_IMMUTABLE)
373 xflags |= FS_XFLAG_IMMUTABLE;
374 if (flags & BTRFS_INODE_NOATIME)
375 xflags |= FS_XFLAG_NOATIME;
376 if (flags & BTRFS_INODE_NODUMP)
377 xflags |= FS_XFLAG_NODUMP;
378 if (flags & BTRFS_INODE_SYNC)
379 xflags |= FS_XFLAG_SYNC;
384 /* Check if @flags are a supported and valid set of FS_XFLAGS_* flags */
385 static int check_xflags(unsigned int flags)
387 if (flags & ~(FS_XFLAG_APPEND | FS_XFLAG_IMMUTABLE | FS_XFLAG_NOATIME |
388 FS_XFLAG_NODUMP | FS_XFLAG_SYNC))
393 bool btrfs_exclop_start(struct btrfs_fs_info *fs_info,
394 enum btrfs_exclusive_operation type)
396 return !cmpxchg(&fs_info->exclusive_operation, BTRFS_EXCLOP_NONE, type);
399 void btrfs_exclop_finish(struct btrfs_fs_info *fs_info)
401 WRITE_ONCE(fs_info->exclusive_operation, BTRFS_EXCLOP_NONE);
402 sysfs_notify(&fs_info->fs_devices->fsid_kobj, NULL, "exclusive_operation");
406 * Set the xflags from the internal inode flags. The remaining items of fsxattr
409 static int btrfs_ioctl_fsgetxattr(struct file *file, void __user *arg)
411 struct btrfs_inode *binode = BTRFS_I(file_inode(file));
414 simple_fill_fsxattr(&fa, btrfs_inode_flags_to_xflags(binode->flags));
415 if (copy_to_user(arg, &fa, sizeof(fa)))
421 static int btrfs_ioctl_fssetxattr(struct file *file, void __user *arg)
423 struct inode *inode = file_inode(file);
424 struct btrfs_inode *binode = BTRFS_I(inode);
425 struct btrfs_root *root = binode->root;
426 struct btrfs_trans_handle *trans;
427 struct fsxattr fa, old_fa;
429 unsigned old_i_flags;
432 if (!inode_owner_or_capable(inode))
435 if (btrfs_root_readonly(root))
438 if (copy_from_user(&fa, arg, sizeof(fa)))
441 ret = check_xflags(fa.fsx_xflags);
445 if (fa.fsx_extsize != 0 || fa.fsx_projid != 0 || fa.fsx_cowextsize != 0)
448 ret = mnt_want_write_file(file);
454 old_flags = binode->flags;
455 old_i_flags = inode->i_flags;
457 simple_fill_fsxattr(&old_fa,
458 btrfs_inode_flags_to_xflags(binode->flags));
459 ret = vfs_ioc_fssetxattr_check(inode, &old_fa, &fa);
463 if (fa.fsx_xflags & FS_XFLAG_SYNC)
464 binode->flags |= BTRFS_INODE_SYNC;
466 binode->flags &= ~BTRFS_INODE_SYNC;
467 if (fa.fsx_xflags & FS_XFLAG_IMMUTABLE)
468 binode->flags |= BTRFS_INODE_IMMUTABLE;
470 binode->flags &= ~BTRFS_INODE_IMMUTABLE;
471 if (fa.fsx_xflags & FS_XFLAG_APPEND)
472 binode->flags |= BTRFS_INODE_APPEND;
474 binode->flags &= ~BTRFS_INODE_APPEND;
475 if (fa.fsx_xflags & FS_XFLAG_NODUMP)
476 binode->flags |= BTRFS_INODE_NODUMP;
478 binode->flags &= ~BTRFS_INODE_NODUMP;
479 if (fa.fsx_xflags & FS_XFLAG_NOATIME)
480 binode->flags |= BTRFS_INODE_NOATIME;
482 binode->flags &= ~BTRFS_INODE_NOATIME;
484 /* 1 item for the inode */
485 trans = btrfs_start_transaction(root, 1);
487 ret = PTR_ERR(trans);
491 btrfs_sync_inode_flags_to_i_flags(inode);
492 inode_inc_iversion(inode);
493 inode->i_ctime = current_time(inode);
494 ret = btrfs_update_inode(trans, root, BTRFS_I(inode));
496 btrfs_end_transaction(trans);
500 binode->flags = old_flags;
501 inode->i_flags = old_i_flags;
505 mnt_drop_write_file(file);
510 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
512 struct inode *inode = file_inode(file);
514 return put_user(inode->i_generation, arg);
517 static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info,
520 struct btrfs_device *device;
521 struct request_queue *q;
522 struct fstrim_range range;
523 u64 minlen = ULLONG_MAX;
527 if (!capable(CAP_SYS_ADMIN))
531 * btrfs_trim_block_group() depends on space cache, which is not
532 * available in zoned filesystem. So, disallow fitrim on a zoned
533 * filesystem for now.
535 if (btrfs_is_zoned(fs_info))
539 * If the fs is mounted with nologreplay, which requires it to be
540 * mounted in RO mode as well, we can not allow discard on free space
541 * inside block groups, because log trees refer to extents that are not
542 * pinned in a block group's free space cache (pinning the extents is
543 * precisely the first phase of replaying a log tree).
545 if (btrfs_test_opt(fs_info, NOLOGREPLAY))
549 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
553 q = bdev_get_queue(device->bdev);
554 if (blk_queue_discard(q)) {
556 minlen = min_t(u64, q->limits.discard_granularity,
564 if (copy_from_user(&range, arg, sizeof(range)))
568 * NOTE: Don't truncate the range using super->total_bytes. Bytenr of
569 * block group is in the logical address space, which can be any
570 * sectorsize aligned bytenr in the range [0, U64_MAX].
572 if (range.len < fs_info->sb->s_blocksize)
575 range.minlen = max(range.minlen, minlen);
576 ret = btrfs_trim_fs(fs_info, &range);
580 if (copy_to_user(arg, &range, sizeof(range)))
586 int __pure btrfs_is_empty_uuid(u8 *uuid)
590 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
597 static noinline int create_subvol(struct inode *dir,
598 struct dentry *dentry,
599 const char *name, int namelen,
600 struct btrfs_qgroup_inherit *inherit)
602 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
603 struct btrfs_trans_handle *trans;
604 struct btrfs_key key;
605 struct btrfs_root_item *root_item;
606 struct btrfs_inode_item *inode_item;
607 struct extent_buffer *leaf;
608 struct btrfs_root *root = BTRFS_I(dir)->root;
609 struct btrfs_root *new_root;
610 struct btrfs_block_rsv block_rsv;
611 struct timespec64 cur_time = current_time(dir);
619 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
623 ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid);
627 ret = get_anon_bdev(&anon_dev);
632 * Don't create subvolume whose level is not zero. Or qgroup will be
633 * screwed up since it assumes subvolume qgroup's level to be 0.
635 if (btrfs_qgroup_level(objectid)) {
640 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
642 * The same as the snapshot creation, please see the comment
643 * of create_snapshot().
645 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false);
649 trans = btrfs_start_transaction(root, 0);
651 ret = PTR_ERR(trans);
652 btrfs_subvolume_release_metadata(root, &block_rsv);
655 trans->block_rsv = &block_rsv;
656 trans->bytes_reserved = block_rsv.size;
658 ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit);
662 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0,
663 BTRFS_NESTING_NORMAL);
669 btrfs_mark_buffer_dirty(leaf);
671 inode_item = &root_item->inode;
672 btrfs_set_stack_inode_generation(inode_item, 1);
673 btrfs_set_stack_inode_size(inode_item, 3);
674 btrfs_set_stack_inode_nlink(inode_item, 1);
675 btrfs_set_stack_inode_nbytes(inode_item,
677 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
679 btrfs_set_root_flags(root_item, 0);
680 btrfs_set_root_limit(root_item, 0);
681 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
683 btrfs_set_root_bytenr(root_item, leaf->start);
684 btrfs_set_root_generation(root_item, trans->transid);
685 btrfs_set_root_level(root_item, 0);
686 btrfs_set_root_refs(root_item, 1);
687 btrfs_set_root_used(root_item, leaf->len);
688 btrfs_set_root_last_snapshot(root_item, 0);
690 btrfs_set_root_generation_v2(root_item,
691 btrfs_root_generation(root_item));
692 generate_random_guid(root_item->uuid);
693 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
694 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
695 root_item->ctime = root_item->otime;
696 btrfs_set_root_ctransid(root_item, trans->transid);
697 btrfs_set_root_otransid(root_item, trans->transid);
699 btrfs_tree_unlock(leaf);
700 free_extent_buffer(leaf);
703 btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID);
705 key.objectid = objectid;
707 key.type = BTRFS_ROOT_ITEM_KEY;
708 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
713 key.offset = (u64)-1;
714 new_root = btrfs_get_new_fs_root(fs_info, objectid, anon_dev);
715 if (IS_ERR(new_root)) {
716 free_anon_bdev(anon_dev);
717 ret = PTR_ERR(new_root);
718 btrfs_abort_transaction(trans, ret);
721 /* Freeing will be done in btrfs_put_root() of new_root */
724 btrfs_record_root_in_trans(trans, new_root);
726 ret = btrfs_create_subvol_root(trans, new_root, root);
727 btrfs_put_root(new_root);
729 /* We potentially lose an unused inode item here */
730 btrfs_abort_transaction(trans, ret);
735 * insert the directory item
737 ret = btrfs_set_inode_index(BTRFS_I(dir), &index);
739 btrfs_abort_transaction(trans, ret);
743 ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key,
744 BTRFS_FT_DIR, index);
746 btrfs_abort_transaction(trans, ret);
750 btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2);
751 ret = btrfs_update_inode(trans, root, BTRFS_I(dir));
753 btrfs_abort_transaction(trans, ret);
757 ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid,
758 btrfs_ino(BTRFS_I(dir)), index, name, namelen);
760 btrfs_abort_transaction(trans, ret);
764 ret = btrfs_uuid_tree_add(trans, root_item->uuid,
765 BTRFS_UUID_KEY_SUBVOL, objectid);
767 btrfs_abort_transaction(trans, ret);
771 trans->block_rsv = NULL;
772 trans->bytes_reserved = 0;
773 btrfs_subvolume_release_metadata(root, &block_rsv);
775 err = btrfs_commit_transaction(trans);
780 inode = btrfs_lookup_dentry(dir, dentry);
782 return PTR_ERR(inode);
783 d_instantiate(dentry, inode);
789 free_anon_bdev(anon_dev);
794 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
795 struct dentry *dentry, bool readonly,
796 struct btrfs_qgroup_inherit *inherit)
798 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
800 struct btrfs_pending_snapshot *pending_snapshot;
801 struct btrfs_trans_handle *trans;
804 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
807 if (atomic_read(&root->nr_swapfiles)) {
809 "cannot snapshot subvolume with active swapfile");
813 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
814 if (!pending_snapshot)
817 ret = get_anon_bdev(&pending_snapshot->anon_dev);
820 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
822 pending_snapshot->path = btrfs_alloc_path();
823 if (!pending_snapshot->root_item || !pending_snapshot->path) {
828 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
829 BTRFS_BLOCK_RSV_TEMP);
831 * 1 - parent dir inode
834 * 2 - root ref/backref
835 * 1 - root of snapshot
838 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
839 &pending_snapshot->block_rsv, 8,
844 pending_snapshot->dentry = dentry;
845 pending_snapshot->root = root;
846 pending_snapshot->readonly = readonly;
847 pending_snapshot->dir = dir;
848 pending_snapshot->inherit = inherit;
850 trans = btrfs_start_transaction(root, 0);
852 ret = PTR_ERR(trans);
856 spin_lock(&fs_info->trans_lock);
857 list_add(&pending_snapshot->list,
858 &trans->transaction->pending_snapshots);
859 spin_unlock(&fs_info->trans_lock);
861 ret = btrfs_commit_transaction(trans);
865 ret = pending_snapshot->error;
869 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
873 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
875 ret = PTR_ERR(inode);
879 d_instantiate(dentry, inode);
881 pending_snapshot->anon_dev = 0;
883 /* Prevent double freeing of anon_dev */
884 if (ret && pending_snapshot->snap)
885 pending_snapshot->snap->anon_dev = 0;
886 btrfs_put_root(pending_snapshot->snap);
887 btrfs_subvolume_release_metadata(root, &pending_snapshot->block_rsv);
889 if (pending_snapshot->anon_dev)
890 free_anon_bdev(pending_snapshot->anon_dev);
891 kfree(pending_snapshot->root_item);
892 btrfs_free_path(pending_snapshot->path);
893 kfree(pending_snapshot);
898 /* copy of may_delete in fs/namei.c()
899 * Check whether we can remove a link victim from directory dir, check
900 * whether the type of victim is right.
901 * 1. We can't do it if dir is read-only (done in permission())
902 * 2. We should have write and exec permissions on dir
903 * 3. We can't remove anything from append-only dir
904 * 4. We can't do anything with immutable dir (done in permission())
905 * 5. If the sticky bit on dir is set we should either
906 * a. be owner of dir, or
907 * b. be owner of victim, or
908 * c. have CAP_FOWNER capability
909 * 6. If the victim is append-only or immutable we can't do anything with
910 * links pointing to it.
911 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
912 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
913 * 9. We can't remove a root or mountpoint.
914 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
915 * nfs_async_unlink().
918 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
922 if (d_really_is_negative(victim))
925 BUG_ON(d_inode(victim->d_parent) != dir);
926 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
928 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
933 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
934 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
937 if (!d_is_dir(victim))
941 } else if (d_is_dir(victim))
945 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
950 /* copy of may_create in fs/namei.c() */
951 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
953 if (d_really_is_positive(child))
957 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
961 * Create a new subvolume below @parent. This is largely modeled after
962 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
963 * inside this filesystem so it's quite a bit simpler.
965 static noinline int btrfs_mksubvol(const struct path *parent,
966 const char *name, int namelen,
967 struct btrfs_root *snap_src,
969 struct btrfs_qgroup_inherit *inherit)
971 struct inode *dir = d_inode(parent->dentry);
972 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
973 struct dentry *dentry;
976 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
980 dentry = lookup_one_len(name, parent->dentry, namelen);
981 error = PTR_ERR(dentry);
985 error = btrfs_may_create(dir, dentry);
990 * even if this name doesn't exist, we may get hash collisions.
991 * check for them now when we can safely fail
993 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
999 down_read(&fs_info->subvol_sem);
1001 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
1005 error = create_snapshot(snap_src, dir, dentry, readonly, inherit);
1007 error = create_subvol(dir, dentry, name, namelen, inherit);
1010 fsnotify_mkdir(dir, dentry);
1012 up_read(&fs_info->subvol_sem);
1020 static noinline int btrfs_mksnapshot(const struct path *parent,
1021 const char *name, int namelen,
1022 struct btrfs_root *root,
1024 struct btrfs_qgroup_inherit *inherit)
1027 bool snapshot_force_cow = false;
1030 * Force new buffered writes to reserve space even when NOCOW is
1031 * possible. This is to avoid later writeback (running dealloc) to
1032 * fallback to COW mode and unexpectedly fail with ENOSPC.
1034 btrfs_drew_read_lock(&root->snapshot_lock);
1036 ret = btrfs_start_delalloc_snapshot(root);
1041 * All previous writes have started writeback in NOCOW mode, so now
1042 * we force future writes to fallback to COW mode during snapshot
1045 atomic_inc(&root->snapshot_force_cow);
1046 snapshot_force_cow = true;
1048 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
1050 ret = btrfs_mksubvol(parent, name, namelen,
1051 root, readonly, inherit);
1053 if (snapshot_force_cow)
1054 atomic_dec(&root->snapshot_force_cow);
1055 btrfs_drew_read_unlock(&root->snapshot_lock);
1060 * When we're defragging a range, we don't want to kick it off again
1061 * if it is really just waiting for delalloc to send it down.
1062 * If we find a nice big extent or delalloc range for the bytes in the
1063 * file you want to defrag, we return 0 to let you know to skip this
1066 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
1068 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1069 struct extent_map *em = NULL;
1070 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1073 read_lock(&em_tree->lock);
1074 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
1075 read_unlock(&em_tree->lock);
1078 end = extent_map_end(em);
1079 free_extent_map(em);
1080 if (end - offset > thresh)
1083 /* if we already have a nice delalloc here, just stop */
1085 end = count_range_bits(io_tree, &offset, offset + thresh,
1086 thresh, EXTENT_DELALLOC, 1);
1093 * helper function to walk through a file and find extents
1094 * newer than a specific transid, and smaller than thresh.
1096 * This is used by the defragging code to find new and small
1099 static int find_new_extents(struct btrfs_root *root,
1100 struct inode *inode, u64 newer_than,
1101 u64 *off, u32 thresh)
1103 struct btrfs_path *path;
1104 struct btrfs_key min_key;
1105 struct extent_buffer *leaf;
1106 struct btrfs_file_extent_item *extent;
1109 u64 ino = btrfs_ino(BTRFS_I(inode));
1111 path = btrfs_alloc_path();
1115 min_key.objectid = ino;
1116 min_key.type = BTRFS_EXTENT_DATA_KEY;
1117 min_key.offset = *off;
1120 ret = btrfs_search_forward(root, &min_key, path, newer_than);
1124 if (min_key.objectid != ino)
1126 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
1129 leaf = path->nodes[0];
1130 extent = btrfs_item_ptr(leaf, path->slots[0],
1131 struct btrfs_file_extent_item);
1133 type = btrfs_file_extent_type(leaf, extent);
1134 if (type == BTRFS_FILE_EXTENT_REG &&
1135 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
1136 check_defrag_in_cache(inode, min_key.offset, thresh)) {
1137 *off = min_key.offset;
1138 btrfs_free_path(path);
1143 if (path->slots[0] < btrfs_header_nritems(leaf)) {
1144 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
1148 if (min_key.offset == (u64)-1)
1152 btrfs_release_path(path);
1155 btrfs_free_path(path);
1159 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
1161 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1162 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1163 struct extent_map *em;
1164 u64 len = PAGE_SIZE;
1167 * hopefully we have this extent in the tree already, try without
1168 * the full extent lock
1170 read_lock(&em_tree->lock);
1171 em = lookup_extent_mapping(em_tree, start, len);
1172 read_unlock(&em_tree->lock);
1175 struct extent_state *cached = NULL;
1176 u64 end = start + len - 1;
1178 /* get the big lock and read metadata off disk */
1179 lock_extent_bits(io_tree, start, end, &cached);
1180 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len);
1181 unlock_extent_cached(io_tree, start, end, &cached);
1190 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1192 struct extent_map *next;
1195 /* this is the last extent */
1196 if (em->start + em->len >= i_size_read(inode))
1199 next = defrag_lookup_extent(inode, em->start + em->len);
1200 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1202 else if ((em->block_start + em->block_len == next->block_start) &&
1203 (em->block_len > SZ_128K && next->block_len > SZ_128K))
1206 free_extent_map(next);
1210 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1211 u64 *last_len, u64 *skip, u64 *defrag_end,
1214 struct extent_map *em;
1216 bool next_mergeable = true;
1217 bool prev_mergeable = true;
1220 * make sure that once we start defragging an extent, we keep on
1223 if (start < *defrag_end)
1228 em = defrag_lookup_extent(inode, start);
1232 /* this will cover holes, and inline extents */
1233 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1239 prev_mergeable = false;
1241 next_mergeable = defrag_check_next_extent(inode, em);
1243 * we hit a real extent, if it is big or the next extent is not a
1244 * real extent, don't bother defragging it
1246 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1247 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1251 * last_len ends up being a counter of how many bytes we've defragged.
1252 * every time we choose not to defrag an extent, we reset *last_len
1253 * so that the next tiny extent will force a defrag.
1255 * The end result of this is that tiny extents before a single big
1256 * extent will force at least part of that big extent to be defragged.
1259 *defrag_end = extent_map_end(em);
1262 *skip = extent_map_end(em);
1266 free_extent_map(em);
1271 * it doesn't do much good to defrag one or two pages
1272 * at a time. This pulls in a nice chunk of pages
1273 * to COW and defrag.
1275 * It also makes sure the delalloc code has enough
1276 * dirty data to avoid making new small extents as part
1279 * It's a good idea to start RA on this range
1280 * before calling this.
1282 static int cluster_pages_for_defrag(struct inode *inode,
1283 struct page **pages,
1284 unsigned long start_index,
1285 unsigned long num_pages)
1287 unsigned long file_end;
1288 u64 isize = i_size_read(inode);
1292 u64 start = (u64)start_index << PAGE_SHIFT;
1297 struct btrfs_ordered_extent *ordered;
1298 struct extent_state *cached_state = NULL;
1299 struct extent_io_tree *tree;
1300 struct extent_changeset *data_reserved = NULL;
1301 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1303 file_end = (isize - 1) >> PAGE_SHIFT;
1304 if (!isize || start_index > file_end)
1307 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1309 ret = btrfs_delalloc_reserve_space(BTRFS_I(inode), &data_reserved,
1310 start, page_cnt << PAGE_SHIFT);
1314 tree = &BTRFS_I(inode)->io_tree;
1316 /* step one, lock all the pages */
1317 for (i = 0; i < page_cnt; i++) {
1320 page = find_or_create_page(inode->i_mapping,
1321 start_index + i, mask);
1325 ret = set_page_extent_mapped(page);
1332 page_start = page_offset(page);
1333 page_end = page_start + PAGE_SIZE - 1;
1335 lock_extent_bits(tree, page_start, page_end,
1337 ordered = btrfs_lookup_ordered_extent(BTRFS_I(inode),
1339 unlock_extent_cached(tree, page_start, page_end,
1345 btrfs_start_ordered_extent(ordered, 1);
1346 btrfs_put_ordered_extent(ordered);
1349 * we unlocked the page above, so we need check if
1350 * it was released or not.
1352 if (page->mapping != inode->i_mapping) {
1359 if (!PageUptodate(page)) {
1360 btrfs_readpage(NULL, page);
1362 if (!PageUptodate(page)) {
1370 if (page->mapping != inode->i_mapping) {
1382 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1386 * so now we have a nice long stream of locked
1387 * and up to date pages, lets wait on them
1389 for (i = 0; i < i_done; i++)
1390 wait_on_page_writeback(pages[i]);
1392 page_start = page_offset(pages[0]);
1393 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
1395 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1396 page_start, page_end - 1, &cached_state);
1399 * When defragmenting we skip ranges that have holes or inline extents,
1400 * (check should_defrag_range()), to avoid unnecessary IO and wasting
1401 * space. At btrfs_defrag_file(), we check if a range should be defragged
1402 * before locking the inode and then, if it should, we trigger a sync
1403 * page cache readahead - we lock the inode only after that to avoid
1404 * blocking for too long other tasks that possibly want to operate on
1405 * other file ranges. But before we were able to get the inode lock,
1406 * some other task may have punched a hole in the range, or we may have
1407 * now an inline extent, in which case we should not defrag. So check
1408 * for that here, where we have the inode and the range locked, and bail
1409 * out if that happened.
1411 search_start = page_start;
1412 while (search_start < page_end) {
1413 struct extent_map *em;
1415 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, search_start,
1416 page_end - search_start);
1419 goto out_unlock_range;
1421 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1422 free_extent_map(em);
1423 /* Ok, 0 means we did not defrag anything */
1425 goto out_unlock_range;
1427 search_start = extent_map_end(em);
1428 free_extent_map(em);
1431 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1432 page_end - 1, EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING |
1433 EXTENT_DEFRAG, 0, 0, &cached_state);
1435 if (i_done != page_cnt) {
1436 spin_lock(&BTRFS_I(inode)->lock);
1437 btrfs_mod_outstanding_extents(BTRFS_I(inode), 1);
1438 spin_unlock(&BTRFS_I(inode)->lock);
1439 btrfs_delalloc_release_space(BTRFS_I(inode), data_reserved,
1440 start, (page_cnt - i_done) << PAGE_SHIFT, true);
1444 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1447 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1448 page_start, page_end - 1, &cached_state);
1450 for (i = 0; i < i_done; i++) {
1451 clear_page_dirty_for_io(pages[i]);
1452 ClearPageChecked(pages[i]);
1453 set_page_dirty(pages[i]);
1454 unlock_page(pages[i]);
1457 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
1458 extent_changeset_free(data_reserved);
1462 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1463 page_start, page_end - 1, &cached_state);
1465 for (i = 0; i < i_done; i++) {
1466 unlock_page(pages[i]);
1469 btrfs_delalloc_release_space(BTRFS_I(inode), data_reserved,
1470 start, page_cnt << PAGE_SHIFT, true);
1471 btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT);
1472 extent_changeset_free(data_reserved);
1477 int btrfs_defrag_file(struct inode *inode, struct file *file,
1478 struct btrfs_ioctl_defrag_range_args *range,
1479 u64 newer_than, unsigned long max_to_defrag)
1481 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1482 struct btrfs_root *root = BTRFS_I(inode)->root;
1483 struct file_ra_state *ra = NULL;
1484 unsigned long last_index;
1485 u64 isize = i_size_read(inode);
1489 u64 newer_off = range->start;
1491 unsigned long ra_index = 0;
1493 int defrag_count = 0;
1494 int compress_type = BTRFS_COMPRESS_ZLIB;
1495 u32 extent_thresh = range->extent_thresh;
1496 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
1497 unsigned long cluster = max_cluster;
1498 u64 new_align = ~((u64)SZ_128K - 1);
1499 struct page **pages = NULL;
1500 bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
1505 if (range->start >= isize)
1509 if (range->compress_type >= BTRFS_NR_COMPRESS_TYPES)
1511 if (range->compress_type)
1512 compress_type = range->compress_type;
1515 if (extent_thresh == 0)
1516 extent_thresh = SZ_256K;
1519 * If we were not given a file, allocate a readahead context. As
1520 * readahead is just an optimization, defrag will work without it so
1521 * we don't error out.
1524 ra = kzalloc(sizeof(*ra), GFP_KERNEL);
1526 file_ra_state_init(ra, inode->i_mapping);
1531 pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL);
1537 /* find the last page to defrag */
1538 if (range->start + range->len > range->start) {
1539 last_index = min_t(u64, isize - 1,
1540 range->start + range->len - 1) >> PAGE_SHIFT;
1542 last_index = (isize - 1) >> PAGE_SHIFT;
1546 ret = find_new_extents(root, inode, newer_than,
1547 &newer_off, SZ_64K);
1549 range->start = newer_off;
1551 * we always align our defrag to help keep
1552 * the extents in the file evenly spaced
1554 i = (newer_off & new_align) >> PAGE_SHIFT;
1558 i = range->start >> PAGE_SHIFT;
1561 max_to_defrag = last_index - i + 1;
1564 * make writeback starts from i, so the defrag range can be
1565 * written sequentially.
1567 if (i < inode->i_mapping->writeback_index)
1568 inode->i_mapping->writeback_index = i;
1570 while (i <= last_index && defrag_count < max_to_defrag &&
1571 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
1573 * make sure we stop running if someone unmounts
1576 if (!(inode->i_sb->s_flags & SB_ACTIVE))
1579 if (btrfs_defrag_cancelled(fs_info)) {
1580 btrfs_debug(fs_info, "defrag_file cancelled");
1585 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
1586 extent_thresh, &last_len, &skip,
1587 &defrag_end, do_compress)){
1590 * the should_defrag function tells us how much to skip
1591 * bump our counter by the suggested amount
1593 next = DIV_ROUND_UP(skip, PAGE_SIZE);
1594 i = max(i + 1, next);
1599 cluster = (PAGE_ALIGN(defrag_end) >>
1601 cluster = min(cluster, max_cluster);
1603 cluster = max_cluster;
1606 if (i + cluster > ra_index) {
1607 ra_index = max(i, ra_index);
1609 page_cache_sync_readahead(inode->i_mapping, ra,
1610 file, ra_index, cluster);
1611 ra_index += cluster;
1615 if (IS_SWAPFILE(inode)) {
1619 BTRFS_I(inode)->defrag_compress = compress_type;
1620 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1623 inode_unlock(inode);
1627 defrag_count += ret;
1628 balance_dirty_pages_ratelimited(inode->i_mapping);
1629 inode_unlock(inode);
1632 if (newer_off == (u64)-1)
1638 newer_off = max(newer_off + 1,
1639 (u64)i << PAGE_SHIFT);
1641 ret = find_new_extents(root, inode, newer_than,
1642 &newer_off, SZ_64K);
1644 range->start = newer_off;
1645 i = (newer_off & new_align) >> PAGE_SHIFT;
1652 last_len += ret << PAGE_SHIFT;
1660 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1661 filemap_flush(inode->i_mapping);
1662 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1663 &BTRFS_I(inode)->runtime_flags))
1664 filemap_flush(inode->i_mapping);
1667 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1668 btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
1669 } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) {
1670 btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
1678 BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
1679 inode_unlock(inode);
1687 static noinline int btrfs_ioctl_resize(struct file *file,
1690 struct inode *inode = file_inode(file);
1691 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1695 struct btrfs_root *root = BTRFS_I(inode)->root;
1696 struct btrfs_ioctl_vol_args *vol_args;
1697 struct btrfs_trans_handle *trans;
1698 struct btrfs_device *device = NULL;
1701 char *devstr = NULL;
1705 if (!capable(CAP_SYS_ADMIN))
1708 ret = mnt_want_write_file(file);
1712 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_RESIZE)) {
1713 mnt_drop_write_file(file);
1714 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1717 vol_args = memdup_user(arg, sizeof(*vol_args));
1718 if (IS_ERR(vol_args)) {
1719 ret = PTR_ERR(vol_args);
1723 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1725 sizestr = vol_args->name;
1726 devstr = strchr(sizestr, ':');
1728 sizestr = devstr + 1;
1730 devstr = vol_args->name;
1731 ret = kstrtoull(devstr, 10, &devid);
1738 btrfs_info(fs_info, "resizing devid %llu", devid);
1741 device = btrfs_find_device(fs_info->fs_devices, devid, NULL, NULL);
1743 btrfs_info(fs_info, "resizer unable to find device %llu",
1749 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
1751 "resizer unable to apply on readonly device %llu",
1757 if (!strcmp(sizestr, "max"))
1758 new_size = device->bdev->bd_inode->i_size;
1760 if (sizestr[0] == '-') {
1763 } else if (sizestr[0] == '+') {
1767 new_size = memparse(sizestr, &retptr);
1768 if (*retptr != '\0' || new_size == 0) {
1774 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
1779 old_size = btrfs_device_get_total_bytes(device);
1782 if (new_size > old_size) {
1786 new_size = old_size - new_size;
1787 } else if (mod > 0) {
1788 if (new_size > ULLONG_MAX - old_size) {
1792 new_size = old_size + new_size;
1795 if (new_size < SZ_256M) {
1799 if (new_size > device->bdev->bd_inode->i_size) {
1804 new_size = round_down(new_size, fs_info->sectorsize);
1806 if (new_size > old_size) {
1807 trans = btrfs_start_transaction(root, 0);
1808 if (IS_ERR(trans)) {
1809 ret = PTR_ERR(trans);
1812 ret = btrfs_grow_device(trans, device, new_size);
1813 btrfs_commit_transaction(trans);
1814 } else if (new_size < old_size) {
1815 ret = btrfs_shrink_device(device, new_size);
1816 } /* equal, nothing need to do */
1818 if (ret == 0 && new_size != old_size)
1819 btrfs_info_in_rcu(fs_info,
1820 "resize device %s (devid %llu) from %llu to %llu",
1821 rcu_str_deref(device->name), device->devid,
1822 old_size, new_size);
1826 btrfs_exclop_finish(fs_info);
1827 mnt_drop_write_file(file);
1831 static noinline int __btrfs_ioctl_snap_create(struct file *file,
1832 const char *name, unsigned long fd, int subvol,
1834 struct btrfs_qgroup_inherit *inherit)
1839 if (!S_ISDIR(file_inode(file)->i_mode))
1842 ret = mnt_want_write_file(file);
1846 namelen = strlen(name);
1847 if (strchr(name, '/')) {
1849 goto out_drop_write;
1852 if (name[0] == '.' &&
1853 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1855 goto out_drop_write;
1859 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1860 NULL, readonly, inherit);
1862 struct fd src = fdget(fd);
1863 struct inode *src_inode;
1866 goto out_drop_write;
1869 src_inode = file_inode(src.file);
1870 if (src_inode->i_sb != file_inode(file)->i_sb) {
1871 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1872 "Snapshot src from another FS");
1874 } else if (!inode_owner_or_capable(src_inode)) {
1876 * Subvolume creation is not restricted, but snapshots
1877 * are limited to own subvolumes only
1881 ret = btrfs_mksnapshot(&file->f_path, name, namelen,
1882 BTRFS_I(src_inode)->root,
1888 mnt_drop_write_file(file);
1893 static noinline int btrfs_ioctl_snap_create(struct file *file,
1894 void __user *arg, int subvol)
1896 struct btrfs_ioctl_vol_args *vol_args;
1899 if (!S_ISDIR(file_inode(file)->i_mode))
1902 vol_args = memdup_user(arg, sizeof(*vol_args));
1903 if (IS_ERR(vol_args))
1904 return PTR_ERR(vol_args);
1905 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1907 ret = __btrfs_ioctl_snap_create(file, vol_args->name, vol_args->fd,
1908 subvol, false, NULL);
1914 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1915 void __user *arg, int subvol)
1917 struct btrfs_ioctl_vol_args_v2 *vol_args;
1919 bool readonly = false;
1920 struct btrfs_qgroup_inherit *inherit = NULL;
1922 if (!S_ISDIR(file_inode(file)->i_mode))
1925 vol_args = memdup_user(arg, sizeof(*vol_args));
1926 if (IS_ERR(vol_args))
1927 return PTR_ERR(vol_args);
1928 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1930 if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) {
1935 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1937 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1940 if (vol_args->size < sizeof(*inherit) ||
1941 vol_args->size > PAGE_SIZE) {
1945 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1946 if (IS_ERR(inherit)) {
1947 ret = PTR_ERR(inherit);
1951 if (inherit->num_qgroups > PAGE_SIZE ||
1952 inherit->num_ref_copies > PAGE_SIZE ||
1953 inherit->num_excl_copies > PAGE_SIZE) {
1958 nums = inherit->num_qgroups + 2 * inherit->num_ref_copies +
1959 2 * inherit->num_excl_copies;
1960 if (vol_args->size != struct_size(inherit, qgroups, nums)) {
1966 ret = __btrfs_ioctl_snap_create(file, vol_args->name, vol_args->fd,
1967 subvol, readonly, inherit);
1977 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1980 struct inode *inode = file_inode(file);
1981 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1982 struct btrfs_root *root = BTRFS_I(inode)->root;
1986 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1989 down_read(&fs_info->subvol_sem);
1990 if (btrfs_root_readonly(root))
1991 flags |= BTRFS_SUBVOL_RDONLY;
1992 up_read(&fs_info->subvol_sem);
1994 if (copy_to_user(arg, &flags, sizeof(flags)))
2000 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
2003 struct inode *inode = file_inode(file);
2004 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2005 struct btrfs_root *root = BTRFS_I(inode)->root;
2006 struct btrfs_trans_handle *trans;
2011 if (!inode_owner_or_capable(inode))
2014 ret = mnt_want_write_file(file);
2018 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2020 goto out_drop_write;
2023 if (copy_from_user(&flags, arg, sizeof(flags))) {
2025 goto out_drop_write;
2028 if (flags & ~BTRFS_SUBVOL_RDONLY) {
2030 goto out_drop_write;
2033 down_write(&fs_info->subvol_sem);
2036 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
2039 root_flags = btrfs_root_flags(&root->root_item);
2040 if (flags & BTRFS_SUBVOL_RDONLY) {
2041 btrfs_set_root_flags(&root->root_item,
2042 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
2045 * Block RO -> RW transition if this subvolume is involved in
2048 spin_lock(&root->root_item_lock);
2049 if (root->send_in_progress == 0) {
2050 btrfs_set_root_flags(&root->root_item,
2051 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
2052 spin_unlock(&root->root_item_lock);
2054 spin_unlock(&root->root_item_lock);
2056 "Attempt to set subvolume %llu read-write during send",
2057 root->root_key.objectid);
2063 trans = btrfs_start_transaction(root, 1);
2064 if (IS_ERR(trans)) {
2065 ret = PTR_ERR(trans);
2069 ret = btrfs_update_root(trans, fs_info->tree_root,
2070 &root->root_key, &root->root_item);
2072 btrfs_end_transaction(trans);
2076 ret = btrfs_commit_transaction(trans);
2080 btrfs_set_root_flags(&root->root_item, root_flags);
2082 up_write(&fs_info->subvol_sem);
2084 mnt_drop_write_file(file);
2089 static noinline int key_in_sk(struct btrfs_key *key,
2090 struct btrfs_ioctl_search_key *sk)
2092 struct btrfs_key test;
2095 test.objectid = sk->min_objectid;
2096 test.type = sk->min_type;
2097 test.offset = sk->min_offset;
2099 ret = btrfs_comp_cpu_keys(key, &test);
2103 test.objectid = sk->max_objectid;
2104 test.type = sk->max_type;
2105 test.offset = sk->max_offset;
2107 ret = btrfs_comp_cpu_keys(key, &test);
2113 static noinline int copy_to_sk(struct btrfs_path *path,
2114 struct btrfs_key *key,
2115 struct btrfs_ioctl_search_key *sk,
2118 unsigned long *sk_offset,
2122 struct extent_buffer *leaf;
2123 struct btrfs_ioctl_search_header sh;
2124 struct btrfs_key test;
2125 unsigned long item_off;
2126 unsigned long item_len;
2132 leaf = path->nodes[0];
2133 slot = path->slots[0];
2134 nritems = btrfs_header_nritems(leaf);
2136 if (btrfs_header_generation(leaf) > sk->max_transid) {
2140 found_transid = btrfs_header_generation(leaf);
2142 for (i = slot; i < nritems; i++) {
2143 item_off = btrfs_item_ptr_offset(leaf, i);
2144 item_len = btrfs_item_size_nr(leaf, i);
2146 btrfs_item_key_to_cpu(leaf, key, i);
2147 if (!key_in_sk(key, sk))
2150 if (sizeof(sh) + item_len > *buf_size) {
2157 * return one empty item back for v1, which does not
2161 *buf_size = sizeof(sh) + item_len;
2166 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
2171 sh.objectid = key->objectid;
2172 sh.offset = key->offset;
2173 sh.type = key->type;
2175 sh.transid = found_transid;
2178 * Copy search result header. If we fault then loop again so we
2179 * can fault in the pages and -EFAULT there if there's a
2180 * problem. Otherwise we'll fault and then copy the buffer in
2181 * properly this next time through
2183 if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) {
2188 *sk_offset += sizeof(sh);
2191 char __user *up = ubuf + *sk_offset;
2193 * Copy the item, same behavior as above, but reset the
2194 * * sk_offset so we copy the full thing again.
2196 if (read_extent_buffer_to_user_nofault(leaf, up,
2197 item_off, item_len)) {
2199 *sk_offset -= sizeof(sh);
2203 *sk_offset += item_len;
2207 if (ret) /* -EOVERFLOW from above */
2210 if (*num_found >= sk->nr_items) {
2217 test.objectid = sk->max_objectid;
2218 test.type = sk->max_type;
2219 test.offset = sk->max_offset;
2220 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2222 else if (key->offset < (u64)-1)
2224 else if (key->type < (u8)-1) {
2227 } else if (key->objectid < (u64)-1) {
2235 * 0: all items from this leaf copied, continue with next
2236 * 1: * more items can be copied, but unused buffer is too small
2237 * * all items were found
2238 * Either way, it will stops the loop which iterates to the next
2240 * -EOVERFLOW: item was to large for buffer
2241 * -EFAULT: could not copy extent buffer back to userspace
2246 static noinline int search_ioctl(struct inode *inode,
2247 struct btrfs_ioctl_search_key *sk,
2251 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
2252 struct btrfs_root *root;
2253 struct btrfs_key key;
2254 struct btrfs_path *path;
2257 unsigned long sk_offset = 0;
2259 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2260 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2264 path = btrfs_alloc_path();
2268 if (sk->tree_id == 0) {
2269 /* search the root of the inode that was passed */
2270 root = btrfs_grab_root(BTRFS_I(inode)->root);
2272 root = btrfs_get_fs_root(info, sk->tree_id, true);
2274 btrfs_free_path(path);
2275 return PTR_ERR(root);
2279 key.objectid = sk->min_objectid;
2280 key.type = sk->min_type;
2281 key.offset = sk->min_offset;
2284 ret = fault_in_pages_writeable(ubuf + sk_offset,
2285 *buf_size - sk_offset);
2289 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2295 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
2296 &sk_offset, &num_found);
2297 btrfs_release_path(path);
2305 sk->nr_items = num_found;
2306 btrfs_put_root(root);
2307 btrfs_free_path(path);
2311 static noinline int btrfs_ioctl_tree_search(struct file *file,
2314 struct btrfs_ioctl_search_args __user *uargs;
2315 struct btrfs_ioctl_search_key sk;
2316 struct inode *inode;
2320 if (!capable(CAP_SYS_ADMIN))
2323 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2325 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2328 buf_size = sizeof(uargs->buf);
2330 inode = file_inode(file);
2331 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2334 * In the origin implementation an overflow is handled by returning a
2335 * search header with a len of zero, so reset ret.
2337 if (ret == -EOVERFLOW)
2340 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2345 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2348 struct btrfs_ioctl_search_args_v2 __user *uarg;
2349 struct btrfs_ioctl_search_args_v2 args;
2350 struct inode *inode;
2353 const size_t buf_limit = SZ_16M;
2355 if (!capable(CAP_SYS_ADMIN))
2358 /* copy search header and buffer size */
2359 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2360 if (copy_from_user(&args, uarg, sizeof(args)))
2363 buf_size = args.buf_size;
2365 /* limit result size to 16MB */
2366 if (buf_size > buf_limit)
2367 buf_size = buf_limit;
2369 inode = file_inode(file);
2370 ret = search_ioctl(inode, &args.key, &buf_size,
2371 (char __user *)(&uarg->buf[0]));
2372 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2374 else if (ret == -EOVERFLOW &&
2375 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2382 * Search INODE_REFs to identify path name of 'dirid' directory
2383 * in a 'tree_id' tree. and sets path name to 'name'.
2385 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2386 u64 tree_id, u64 dirid, char *name)
2388 struct btrfs_root *root;
2389 struct btrfs_key key;
2395 struct btrfs_inode_ref *iref;
2396 struct extent_buffer *l;
2397 struct btrfs_path *path;
2399 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2404 path = btrfs_alloc_path();
2408 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
2410 root = btrfs_get_fs_root(info, tree_id, true);
2412 ret = PTR_ERR(root);
2417 key.objectid = dirid;
2418 key.type = BTRFS_INODE_REF_KEY;
2419 key.offset = (u64)-1;
2422 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2426 ret = btrfs_previous_item(root, path, dirid,
2427 BTRFS_INODE_REF_KEY);
2437 slot = path->slots[0];
2438 btrfs_item_key_to_cpu(l, &key, slot);
2440 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2441 len = btrfs_inode_ref_name_len(l, iref);
2443 total_len += len + 1;
2445 ret = -ENAMETOOLONG;
2450 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2452 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2455 btrfs_release_path(path);
2456 key.objectid = key.offset;
2457 key.offset = (u64)-1;
2458 dirid = key.objectid;
2460 memmove(name, ptr, total_len);
2461 name[total_len] = '\0';
2464 btrfs_put_root(root);
2465 btrfs_free_path(path);
2469 static int btrfs_search_path_in_tree_user(struct inode *inode,
2470 struct btrfs_ioctl_ino_lookup_user_args *args)
2472 struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2473 struct super_block *sb = inode->i_sb;
2474 struct btrfs_key upper_limit = BTRFS_I(inode)->location;
2475 u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
2476 u64 dirid = args->dirid;
2477 unsigned long item_off;
2478 unsigned long item_len;
2479 struct btrfs_inode_ref *iref;
2480 struct btrfs_root_ref *rref;
2481 struct btrfs_root *root = NULL;
2482 struct btrfs_path *path;
2483 struct btrfs_key key, key2;
2484 struct extent_buffer *leaf;
2485 struct inode *temp_inode;
2492 path = btrfs_alloc_path();
2497 * If the bottom subvolume does not exist directly under upper_limit,
2498 * construct the path in from the bottom up.
2500 if (dirid != upper_limit.objectid) {
2501 ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
2503 root = btrfs_get_fs_root(fs_info, treeid, true);
2505 ret = PTR_ERR(root);
2509 key.objectid = dirid;
2510 key.type = BTRFS_INODE_REF_KEY;
2511 key.offset = (u64)-1;
2513 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2516 } else if (ret > 0) {
2517 ret = btrfs_previous_item(root, path, dirid,
2518 BTRFS_INODE_REF_KEY);
2521 } else if (ret > 0) {
2527 leaf = path->nodes[0];
2528 slot = path->slots[0];
2529 btrfs_item_key_to_cpu(leaf, &key, slot);
2531 iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
2532 len = btrfs_inode_ref_name_len(leaf, iref);
2534 total_len += len + 1;
2535 if (ptr < args->path) {
2536 ret = -ENAMETOOLONG;
2541 read_extent_buffer(leaf, ptr,
2542 (unsigned long)(iref + 1), len);
2544 /* Check the read+exec permission of this directory */
2545 ret = btrfs_previous_item(root, path, dirid,
2546 BTRFS_INODE_ITEM_KEY);
2549 } else if (ret > 0) {
2554 leaf = path->nodes[0];
2555 slot = path->slots[0];
2556 btrfs_item_key_to_cpu(leaf, &key2, slot);
2557 if (key2.objectid != dirid) {
2562 temp_inode = btrfs_iget(sb, key2.objectid, root);
2563 if (IS_ERR(temp_inode)) {
2564 ret = PTR_ERR(temp_inode);
2567 ret = inode_permission(temp_inode, MAY_READ | MAY_EXEC);
2574 if (key.offset == upper_limit.objectid)
2576 if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
2581 btrfs_release_path(path);
2582 key.objectid = key.offset;
2583 key.offset = (u64)-1;
2584 dirid = key.objectid;
2587 memmove(args->path, ptr, total_len);
2588 args->path[total_len] = '\0';
2589 btrfs_put_root(root);
2591 btrfs_release_path(path);
2594 /* Get the bottom subvolume's name from ROOT_REF */
2595 key.objectid = treeid;
2596 key.type = BTRFS_ROOT_REF_KEY;
2597 key.offset = args->treeid;
2598 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2601 } else if (ret > 0) {
2606 leaf = path->nodes[0];
2607 slot = path->slots[0];
2608 btrfs_item_key_to_cpu(leaf, &key, slot);
2610 item_off = btrfs_item_ptr_offset(leaf, slot);
2611 item_len = btrfs_item_size_nr(leaf, slot);
2612 /* Check if dirid in ROOT_REF corresponds to passed dirid */
2613 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2614 if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2619 /* Copy subvolume's name */
2620 item_off += sizeof(struct btrfs_root_ref);
2621 item_len -= sizeof(struct btrfs_root_ref);
2622 read_extent_buffer(leaf, args->name, item_off, item_len);
2623 args->name[item_len] = 0;
2626 btrfs_put_root(root);
2628 btrfs_free_path(path);
2632 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2635 struct btrfs_ioctl_ino_lookup_args *args;
2636 struct inode *inode;
2639 args = memdup_user(argp, sizeof(*args));
2641 return PTR_ERR(args);
2643 inode = file_inode(file);
2646 * Unprivileged query to obtain the containing subvolume root id. The
2647 * path is reset so it's consistent with btrfs_search_path_in_tree.
2649 if (args->treeid == 0)
2650 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2652 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2657 if (!capable(CAP_SYS_ADMIN)) {
2662 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2663 args->treeid, args->objectid,
2667 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2675 * Version of ino_lookup ioctl (unprivileged)
2677 * The main differences from ino_lookup ioctl are:
2679 * 1. Read + Exec permission will be checked using inode_permission() during
2680 * path construction. -EACCES will be returned in case of failure.
2681 * 2. Path construction will be stopped at the inode number which corresponds
2682 * to the fd with which this ioctl is called. If constructed path does not
2683 * exist under fd's inode, -EACCES will be returned.
2684 * 3. The name of bottom subvolume is also searched and filled.
2686 static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2688 struct btrfs_ioctl_ino_lookup_user_args *args;
2689 struct inode *inode;
2692 args = memdup_user(argp, sizeof(*args));
2694 return PTR_ERR(args);
2696 inode = file_inode(file);
2698 if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2699 BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2701 * The subvolume does not exist under fd with which this is
2708 ret = btrfs_search_path_in_tree_user(inode, args);
2710 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2717 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2718 static int btrfs_ioctl_get_subvol_info(struct file *file, void __user *argp)
2720 struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2721 struct btrfs_fs_info *fs_info;
2722 struct btrfs_root *root;
2723 struct btrfs_path *path;
2724 struct btrfs_key key;
2725 struct btrfs_root_item *root_item;
2726 struct btrfs_root_ref *rref;
2727 struct extent_buffer *leaf;
2728 unsigned long item_off;
2729 unsigned long item_len;
2730 struct inode *inode;
2734 path = btrfs_alloc_path();
2738 subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2740 btrfs_free_path(path);
2744 inode = file_inode(file);
2745 fs_info = BTRFS_I(inode)->root->fs_info;
2747 /* Get root_item of inode's subvolume */
2748 key.objectid = BTRFS_I(inode)->root->root_key.objectid;
2749 root = btrfs_get_fs_root(fs_info, key.objectid, true);
2751 ret = PTR_ERR(root);
2754 root_item = &root->root_item;
2756 subvol_info->treeid = key.objectid;
2758 subvol_info->generation = btrfs_root_generation(root_item);
2759 subvol_info->flags = btrfs_root_flags(root_item);
2761 memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
2762 memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
2764 memcpy(subvol_info->received_uuid, root_item->received_uuid,
2767 subvol_info->ctransid = btrfs_root_ctransid(root_item);
2768 subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
2769 subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
2771 subvol_info->otransid = btrfs_root_otransid(root_item);
2772 subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
2773 subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
2775 subvol_info->stransid = btrfs_root_stransid(root_item);
2776 subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
2777 subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
2779 subvol_info->rtransid = btrfs_root_rtransid(root_item);
2780 subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
2781 subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
2783 if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
2784 /* Search root tree for ROOT_BACKREF of this subvolume */
2785 key.type = BTRFS_ROOT_BACKREF_KEY;
2787 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2790 } else if (path->slots[0] >=
2791 btrfs_header_nritems(path->nodes[0])) {
2792 ret = btrfs_next_leaf(fs_info->tree_root, path);
2795 } else if (ret > 0) {
2801 leaf = path->nodes[0];
2802 slot = path->slots[0];
2803 btrfs_item_key_to_cpu(leaf, &key, slot);
2804 if (key.objectid == subvol_info->treeid &&
2805 key.type == BTRFS_ROOT_BACKREF_KEY) {
2806 subvol_info->parent_id = key.offset;
2808 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2809 subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
2811 item_off = btrfs_item_ptr_offset(leaf, slot)
2812 + sizeof(struct btrfs_root_ref);
2813 item_len = btrfs_item_size_nr(leaf, slot)
2814 - sizeof(struct btrfs_root_ref);
2815 read_extent_buffer(leaf, subvol_info->name,
2816 item_off, item_len);
2823 if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
2827 btrfs_put_root(root);
2829 btrfs_free_path(path);
2835 * Return ROOT_REF information of the subvolume containing this inode
2836 * except the subvolume name.
2838 static int btrfs_ioctl_get_subvol_rootref(struct file *file, void __user *argp)
2840 struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
2841 struct btrfs_root_ref *rref;
2842 struct btrfs_root *root;
2843 struct btrfs_path *path;
2844 struct btrfs_key key;
2845 struct extent_buffer *leaf;
2846 struct inode *inode;
2852 path = btrfs_alloc_path();
2856 rootrefs = memdup_user(argp, sizeof(*rootrefs));
2857 if (IS_ERR(rootrefs)) {
2858 btrfs_free_path(path);
2859 return PTR_ERR(rootrefs);
2862 inode = file_inode(file);
2863 root = BTRFS_I(inode)->root->fs_info->tree_root;
2864 objectid = BTRFS_I(inode)->root->root_key.objectid;
2866 key.objectid = objectid;
2867 key.type = BTRFS_ROOT_REF_KEY;
2868 key.offset = rootrefs->min_treeid;
2871 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2874 } else if (path->slots[0] >=
2875 btrfs_header_nritems(path->nodes[0])) {
2876 ret = btrfs_next_leaf(root, path);
2879 } else if (ret > 0) {
2885 leaf = path->nodes[0];
2886 slot = path->slots[0];
2888 btrfs_item_key_to_cpu(leaf, &key, slot);
2889 if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
2894 if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
2899 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2900 rootrefs->rootref[found].treeid = key.offset;
2901 rootrefs->rootref[found].dirid =
2902 btrfs_root_ref_dirid(leaf, rref);
2905 ret = btrfs_next_item(root, path);
2908 } else if (ret > 0) {
2915 if (!ret || ret == -EOVERFLOW) {
2916 rootrefs->num_items = found;
2917 /* update min_treeid for next search */
2919 rootrefs->min_treeid =
2920 rootrefs->rootref[found - 1].treeid + 1;
2921 if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
2926 btrfs_free_path(path);
2931 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2935 struct dentry *parent = file->f_path.dentry;
2936 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2937 struct dentry *dentry;
2938 struct inode *dir = d_inode(parent);
2939 struct inode *inode;
2940 struct btrfs_root *root = BTRFS_I(dir)->root;
2941 struct btrfs_root *dest = NULL;
2942 struct btrfs_ioctl_vol_args *vol_args = NULL;
2943 struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL;
2944 char *subvol_name, *subvol_name_ptr = NULL;
2947 bool destroy_parent = false;
2950 vol_args2 = memdup_user(arg, sizeof(*vol_args2));
2951 if (IS_ERR(vol_args2))
2952 return PTR_ERR(vol_args2);
2954 if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) {
2960 * If SPEC_BY_ID is not set, we are looking for the subvolume by
2961 * name, same as v1 currently does.
2963 if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) {
2964 vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0;
2965 subvol_name = vol_args2->name;
2967 err = mnt_want_write_file(file);
2971 if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) {
2976 err = mnt_want_write_file(file);
2980 dentry = btrfs_get_dentry(fs_info->sb,
2981 BTRFS_FIRST_FREE_OBJECTID,
2982 vol_args2->subvolid, 0, 0);
2983 if (IS_ERR(dentry)) {
2984 err = PTR_ERR(dentry);
2985 goto out_drop_write;
2989 * Change the default parent since the subvolume being
2990 * deleted can be outside of the current mount point.
2992 parent = btrfs_get_parent(dentry);
2995 * At this point dentry->d_name can point to '/' if the
2996 * subvolume we want to destroy is outsite of the
2997 * current mount point, so we need to release the
2998 * current dentry and execute the lookup to return a new
2999 * one with ->d_name pointing to the
3000 * <mount point>/subvol_name.
3003 if (IS_ERR(parent)) {
3004 err = PTR_ERR(parent);
3005 goto out_drop_write;
3007 dir = d_inode(parent);
3010 * If v2 was used with SPEC_BY_ID, a new parent was
3011 * allocated since the subvolume can be outside of the
3012 * current mount point. Later on we need to release this
3013 * new parent dentry.
3015 destroy_parent = true;
3017 subvol_name_ptr = btrfs_get_subvol_name_from_objectid(
3018 fs_info, vol_args2->subvolid);
3019 if (IS_ERR(subvol_name_ptr)) {
3020 err = PTR_ERR(subvol_name_ptr);
3023 /* subvol_name_ptr is already NULL termined */
3024 subvol_name = (char *)kbasename(subvol_name_ptr);
3027 vol_args = memdup_user(arg, sizeof(*vol_args));
3028 if (IS_ERR(vol_args))
3029 return PTR_ERR(vol_args);
3031 vol_args->name[BTRFS_PATH_NAME_MAX] = 0;
3032 subvol_name = vol_args->name;
3034 err = mnt_want_write_file(file);
3039 subvol_namelen = strlen(subvol_name);
3041 if (strchr(subvol_name, '/') ||
3042 strncmp(subvol_name, "..", subvol_namelen) == 0) {
3044 goto free_subvol_name;
3047 if (!S_ISDIR(dir->i_mode)) {
3049 goto free_subvol_name;
3052 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
3054 goto free_subvol_name;
3055 dentry = lookup_one_len(subvol_name, parent, subvol_namelen);
3056 if (IS_ERR(dentry)) {
3057 err = PTR_ERR(dentry);
3058 goto out_unlock_dir;
3061 if (d_really_is_negative(dentry)) {
3066 inode = d_inode(dentry);
3067 dest = BTRFS_I(inode)->root;
3068 if (!capable(CAP_SYS_ADMIN)) {
3070 * Regular user. Only allow this with a special mount
3071 * option, when the user has write+exec access to the
3072 * subvol root, and when rmdir(2) would have been
3075 * Note that this is _not_ check that the subvol is
3076 * empty or doesn't contain data that we wouldn't
3077 * otherwise be able to delete.
3079 * Users who want to delete empty subvols should try
3083 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
3087 * Do not allow deletion if the parent dir is the same
3088 * as the dir to be deleted. That means the ioctl
3089 * must be called on the dentry referencing the root
3090 * of the subvol, not a random directory contained
3097 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
3102 /* check if subvolume may be deleted by a user */
3103 err = btrfs_may_delete(dir, dentry, 1);
3107 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
3113 err = btrfs_delete_subvolume(dir, dentry);
3114 inode_unlock(inode);
3116 fsnotify_rmdir(dir, dentry);
3125 kfree(subvol_name_ptr);
3130 mnt_drop_write_file(file);
3137 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
3139 struct inode *inode = file_inode(file);
3140 struct btrfs_root *root = BTRFS_I(inode)->root;
3141 struct btrfs_ioctl_defrag_range_args *range;
3144 ret = mnt_want_write_file(file);
3148 if (btrfs_root_readonly(root)) {
3153 switch (inode->i_mode & S_IFMT) {
3155 if (!capable(CAP_SYS_ADMIN)) {
3159 ret = btrfs_defrag_root(root);
3163 * Note that this does not check the file descriptor for write
3164 * access. This prevents defragmenting executables that are
3165 * running and allows defrag on files open in read-only mode.
3167 if (!capable(CAP_SYS_ADMIN) &&
3168 inode_permission(inode, MAY_WRITE)) {
3173 range = kzalloc(sizeof(*range), GFP_KERNEL);
3180 if (copy_from_user(range, argp,
3186 /* compression requires us to start the IO */
3187 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
3188 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
3189 range->extent_thresh = (u32)-1;
3192 /* the rest are all set to zero by kzalloc */
3193 range->len = (u64)-1;
3195 ret = btrfs_defrag_file(file_inode(file), file,
3196 range, BTRFS_OLDEST_GENERATION, 0);
3205 mnt_drop_write_file(file);
3209 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
3211 struct btrfs_ioctl_vol_args *vol_args;
3214 if (!capable(CAP_SYS_ADMIN))
3217 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD))
3218 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3220 vol_args = memdup_user(arg, sizeof(*vol_args));
3221 if (IS_ERR(vol_args)) {
3222 ret = PTR_ERR(vol_args);
3226 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3227 ret = btrfs_init_new_device(fs_info, vol_args->name);
3230 btrfs_info(fs_info, "disk added %s", vol_args->name);
3234 btrfs_exclop_finish(fs_info);
3238 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
3240 struct inode *inode = file_inode(file);
3241 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3242 struct btrfs_ioctl_vol_args_v2 *vol_args;
3245 if (!capable(CAP_SYS_ADMIN))
3248 ret = mnt_want_write_file(file);
3252 vol_args = memdup_user(arg, sizeof(*vol_args));
3253 if (IS_ERR(vol_args)) {
3254 ret = PTR_ERR(vol_args);
3258 if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) {
3263 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REMOVE)) {
3264 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3268 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
3269 ret = btrfs_rm_device(fs_info, NULL, vol_args->devid);
3271 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
3272 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
3274 btrfs_exclop_finish(fs_info);
3277 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
3278 btrfs_info(fs_info, "device deleted: id %llu",
3281 btrfs_info(fs_info, "device deleted: %s",
3287 mnt_drop_write_file(file);
3291 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
3293 struct inode *inode = file_inode(file);
3294 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3295 struct btrfs_ioctl_vol_args *vol_args;
3298 if (!capable(CAP_SYS_ADMIN))
3301 ret = mnt_want_write_file(file);
3305 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REMOVE)) {
3306 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3307 goto out_drop_write;
3310 vol_args = memdup_user(arg, sizeof(*vol_args));
3311 if (IS_ERR(vol_args)) {
3312 ret = PTR_ERR(vol_args);
3316 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3317 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
3320 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
3323 btrfs_exclop_finish(fs_info);
3325 mnt_drop_write_file(file);
3330 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
3333 struct btrfs_ioctl_fs_info_args *fi_args;
3334 struct btrfs_device *device;
3335 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
3339 fi_args = memdup_user(arg, sizeof(*fi_args));
3340 if (IS_ERR(fi_args))
3341 return PTR_ERR(fi_args);
3343 flags_in = fi_args->flags;
3344 memset(fi_args, 0, sizeof(*fi_args));
3347 fi_args->num_devices = fs_devices->num_devices;
3349 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
3350 if (device->devid > fi_args->max_id)
3351 fi_args->max_id = device->devid;
3355 memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
3356 fi_args->nodesize = fs_info->nodesize;
3357 fi_args->sectorsize = fs_info->sectorsize;
3358 fi_args->clone_alignment = fs_info->sectorsize;
3360 if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) {
3361 fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy);
3362 fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy);
3363 fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO;
3366 if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) {
3367 fi_args->generation = fs_info->generation;
3368 fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION;
3371 if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) {
3372 memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid,
3373 sizeof(fi_args->metadata_uuid));
3374 fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID;
3377 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
3384 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
3387 struct btrfs_ioctl_dev_info_args *di_args;
3388 struct btrfs_device *dev;
3390 char *s_uuid = NULL;
3392 di_args = memdup_user(arg, sizeof(*di_args));
3393 if (IS_ERR(di_args))
3394 return PTR_ERR(di_args);
3396 if (!btrfs_is_empty_uuid(di_args->uuid))
3397 s_uuid = di_args->uuid;
3400 dev = btrfs_find_device(fs_info->fs_devices, di_args->devid, s_uuid,
3408 di_args->devid = dev->devid;
3409 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
3410 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
3411 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
3413 strncpy(di_args->path, rcu_str_deref(dev->name),
3414 sizeof(di_args->path) - 1);
3415 di_args->path[sizeof(di_args->path) - 1] = 0;
3417 di_args->path[0] = '\0';
3422 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
3429 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3431 struct inode *inode = file_inode(file);
3432 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3433 struct btrfs_root *root = BTRFS_I(inode)->root;
3434 struct btrfs_root *new_root;
3435 struct btrfs_dir_item *di;
3436 struct btrfs_trans_handle *trans;
3437 struct btrfs_path *path = NULL;
3438 struct btrfs_disk_key disk_key;
3443 if (!capable(CAP_SYS_ADMIN))
3446 ret = mnt_want_write_file(file);
3450 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3456 objectid = BTRFS_FS_TREE_OBJECTID;
3458 new_root = btrfs_get_fs_root(fs_info, objectid, true);
3459 if (IS_ERR(new_root)) {
3460 ret = PTR_ERR(new_root);
3463 if (!is_fstree(new_root->root_key.objectid)) {
3468 path = btrfs_alloc_path();
3474 trans = btrfs_start_transaction(root, 1);
3475 if (IS_ERR(trans)) {
3476 ret = PTR_ERR(trans);
3480 dir_id = btrfs_super_root_dir(fs_info->super_copy);
3481 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
3482 dir_id, "default", 7, 1);
3483 if (IS_ERR_OR_NULL(di)) {
3484 btrfs_release_path(path);
3485 btrfs_end_transaction(trans);
3487 "Umm, you don't have the default diritem, this isn't going to work");
3492 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3493 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3494 btrfs_mark_buffer_dirty(path->nodes[0]);
3495 btrfs_release_path(path);
3497 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
3498 btrfs_end_transaction(trans);
3500 btrfs_put_root(new_root);
3501 btrfs_free_path(path);
3503 mnt_drop_write_file(file);
3507 static void get_block_group_info(struct list_head *groups_list,
3508 struct btrfs_ioctl_space_info *space)
3510 struct btrfs_block_group *block_group;
3512 space->total_bytes = 0;
3513 space->used_bytes = 0;
3515 list_for_each_entry(block_group, groups_list, list) {
3516 space->flags = block_group->flags;
3517 space->total_bytes += block_group->length;
3518 space->used_bytes += block_group->used;
3522 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
3525 struct btrfs_ioctl_space_args space_args;
3526 struct btrfs_ioctl_space_info space;
3527 struct btrfs_ioctl_space_info *dest;
3528 struct btrfs_ioctl_space_info *dest_orig;
3529 struct btrfs_ioctl_space_info __user *user_dest;
3530 struct btrfs_space_info *info;
3531 static const u64 types[] = {
3532 BTRFS_BLOCK_GROUP_DATA,
3533 BTRFS_BLOCK_GROUP_SYSTEM,
3534 BTRFS_BLOCK_GROUP_METADATA,
3535 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
3543 if (copy_from_user(&space_args,
3544 (struct btrfs_ioctl_space_args __user *)arg,
3545 sizeof(space_args)))
3548 for (i = 0; i < num_types; i++) {
3549 struct btrfs_space_info *tmp;
3552 list_for_each_entry(tmp, &fs_info->space_info, list) {
3553 if (tmp->flags == types[i]) {
3562 down_read(&info->groups_sem);
3563 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3564 if (!list_empty(&info->block_groups[c]))
3567 up_read(&info->groups_sem);
3571 * Global block reserve, exported as a space_info
3575 /* space_slots == 0 means they are asking for a count */
3576 if (space_args.space_slots == 0) {
3577 space_args.total_spaces = slot_count;
3581 slot_count = min_t(u64, space_args.space_slots, slot_count);
3583 alloc_size = sizeof(*dest) * slot_count;
3585 /* we generally have at most 6 or so space infos, one for each raid
3586 * level. So, a whole page should be more than enough for everyone
3588 if (alloc_size > PAGE_SIZE)
3591 space_args.total_spaces = 0;
3592 dest = kmalloc(alloc_size, GFP_KERNEL);
3597 /* now we have a buffer to copy into */
3598 for (i = 0; i < num_types; i++) {
3599 struct btrfs_space_info *tmp;
3605 list_for_each_entry(tmp, &fs_info->space_info, list) {
3606 if (tmp->flags == types[i]) {
3614 down_read(&info->groups_sem);
3615 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3616 if (!list_empty(&info->block_groups[c])) {
3617 get_block_group_info(&info->block_groups[c],
3619 memcpy(dest, &space, sizeof(space));
3621 space_args.total_spaces++;
3627 up_read(&info->groups_sem);
3631 * Add global block reserve
3634 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
3636 spin_lock(&block_rsv->lock);
3637 space.total_bytes = block_rsv->size;
3638 space.used_bytes = block_rsv->size - block_rsv->reserved;
3639 spin_unlock(&block_rsv->lock);
3640 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3641 memcpy(dest, &space, sizeof(space));
3642 space_args.total_spaces++;
3645 user_dest = (struct btrfs_ioctl_space_info __user *)
3646 (arg + sizeof(struct btrfs_ioctl_space_args));
3648 if (copy_to_user(user_dest, dest_orig, alloc_size))
3653 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3659 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3662 struct btrfs_trans_handle *trans;
3666 trans = btrfs_attach_transaction_barrier(root);
3667 if (IS_ERR(trans)) {
3668 if (PTR_ERR(trans) != -ENOENT)
3669 return PTR_ERR(trans);
3671 /* No running transaction, don't bother */
3672 transid = root->fs_info->last_trans_committed;
3675 transid = trans->transid;
3676 ret = btrfs_commit_transaction_async(trans, 0);
3678 btrfs_end_transaction(trans);
3683 if (copy_to_user(argp, &transid, sizeof(transid)))
3688 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
3694 if (copy_from_user(&transid, argp, sizeof(transid)))
3697 transid = 0; /* current trans */
3699 return btrfs_wait_for_commit(fs_info, transid);
3702 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3704 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
3705 struct btrfs_ioctl_scrub_args *sa;
3708 if (!capable(CAP_SYS_ADMIN))
3711 sa = memdup_user(arg, sizeof(*sa));
3715 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3716 ret = mnt_want_write_file(file);
3721 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
3722 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3726 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
3727 * error. This is important as it allows user space to know how much
3728 * progress scrub has done. For example, if scrub is canceled we get
3729 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
3730 * space. Later user space can inspect the progress from the structure
3731 * btrfs_ioctl_scrub_args and resume scrub from where it left off
3732 * previously (btrfs-progs does this).
3733 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
3734 * then return -EFAULT to signal the structure was not copied or it may
3735 * be corrupt and unreliable due to a partial copy.
3737 if (copy_to_user(arg, sa, sizeof(*sa)))
3740 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3741 mnt_drop_write_file(file);
3747 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
3749 if (!capable(CAP_SYS_ADMIN))
3752 return btrfs_scrub_cancel(fs_info);
3755 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
3758 struct btrfs_ioctl_scrub_args *sa;
3761 if (!capable(CAP_SYS_ADMIN))
3764 sa = memdup_user(arg, sizeof(*sa));
3768 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
3770 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3777 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
3780 struct btrfs_ioctl_get_dev_stats *sa;
3783 sa = memdup_user(arg, sizeof(*sa));
3787 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3792 ret = btrfs_get_dev_stats(fs_info, sa);
3794 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
3801 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
3804 struct btrfs_ioctl_dev_replace_args *p;
3807 if (!capable(CAP_SYS_ADMIN))
3810 p = memdup_user(arg, sizeof(*p));
3815 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3816 if (sb_rdonly(fs_info->sb)) {
3820 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) {
3821 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3823 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
3824 btrfs_exclop_finish(fs_info);
3827 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3828 btrfs_dev_replace_status(fs_info, p);
3831 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3832 p->result = btrfs_dev_replace_cancel(fs_info);
3840 if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
3847 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3853 struct btrfs_ioctl_ino_path_args *ipa = NULL;
3854 struct inode_fs_paths *ipath = NULL;
3855 struct btrfs_path *path;
3857 if (!capable(CAP_DAC_READ_SEARCH))
3860 path = btrfs_alloc_path();
3866 ipa = memdup_user(arg, sizeof(*ipa));
3873 size = min_t(u32, ipa->size, 4096);
3874 ipath = init_ipath(size, root, path);
3875 if (IS_ERR(ipath)) {
3876 ret = PTR_ERR(ipath);
3881 ret = paths_from_inode(ipa->inum, ipath);
3885 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3886 rel_ptr = ipath->fspath->val[i] -
3887 (u64)(unsigned long)ipath->fspath->val;
3888 ipath->fspath->val[i] = rel_ptr;
3891 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
3892 ipath->fspath, size);
3899 btrfs_free_path(path);
3906 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3908 struct btrfs_data_container *inodes = ctx;
3909 const size_t c = 3 * sizeof(u64);
3911 if (inodes->bytes_left >= c) {
3912 inodes->bytes_left -= c;
3913 inodes->val[inodes->elem_cnt] = inum;
3914 inodes->val[inodes->elem_cnt + 1] = offset;
3915 inodes->val[inodes->elem_cnt + 2] = root;
3916 inodes->elem_cnt += 3;
3918 inodes->bytes_missing += c - inodes->bytes_left;
3919 inodes->bytes_left = 0;
3920 inodes->elem_missed += 3;
3926 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
3927 void __user *arg, int version)
3931 struct btrfs_ioctl_logical_ino_args *loi;
3932 struct btrfs_data_container *inodes = NULL;
3933 struct btrfs_path *path = NULL;
3936 if (!capable(CAP_SYS_ADMIN))
3939 loi = memdup_user(arg, sizeof(*loi));
3941 return PTR_ERR(loi);
3944 ignore_offset = false;
3945 size = min_t(u32, loi->size, SZ_64K);
3947 /* All reserved bits must be 0 for now */
3948 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
3952 /* Only accept flags we have defined so far */
3953 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
3957 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
3958 size = min_t(u32, loi->size, SZ_16M);
3961 path = btrfs_alloc_path();
3967 inodes = init_data_container(size);
3968 if (IS_ERR(inodes)) {
3969 ret = PTR_ERR(inodes);
3974 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
3975 build_ino_list, inodes, ignore_offset);
3981 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
3987 btrfs_free_path(path);
3995 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
3996 struct btrfs_ioctl_balance_args *bargs)
3998 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4000 bargs->flags = bctl->flags;
4002 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
4003 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4004 if (atomic_read(&fs_info->balance_pause_req))
4005 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4006 if (atomic_read(&fs_info->balance_cancel_req))
4007 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4009 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4010 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4011 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4013 spin_lock(&fs_info->balance_lock);
4014 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4015 spin_unlock(&fs_info->balance_lock);
4018 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4020 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4021 struct btrfs_fs_info *fs_info = root->fs_info;
4022 struct btrfs_ioctl_balance_args *bargs;
4023 struct btrfs_balance_control *bctl;
4024 bool need_unlock; /* for mut. excl. ops lock */
4027 if (!capable(CAP_SYS_ADMIN))
4030 ret = mnt_want_write_file(file);
4035 if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) {
4036 mutex_lock(&fs_info->balance_mutex);
4042 * mut. excl. ops lock is locked. Three possibilities:
4043 * (1) some other op is running
4044 * (2) balance is running
4045 * (3) balance is paused -- special case (think resume)
4047 mutex_lock(&fs_info->balance_mutex);
4048 if (fs_info->balance_ctl) {
4049 /* this is either (2) or (3) */
4050 if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4051 mutex_unlock(&fs_info->balance_mutex);
4053 * Lock released to allow other waiters to continue,
4054 * we'll reexamine the status again.
4056 mutex_lock(&fs_info->balance_mutex);
4058 if (fs_info->balance_ctl &&
4059 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4061 need_unlock = false;
4065 mutex_unlock(&fs_info->balance_mutex);
4069 mutex_unlock(&fs_info->balance_mutex);
4075 mutex_unlock(&fs_info->balance_mutex);
4076 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4083 bargs = memdup_user(arg, sizeof(*bargs));
4084 if (IS_ERR(bargs)) {
4085 ret = PTR_ERR(bargs);
4089 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4090 if (!fs_info->balance_ctl) {
4095 bctl = fs_info->balance_ctl;
4096 spin_lock(&fs_info->balance_lock);
4097 bctl->flags |= BTRFS_BALANCE_RESUME;
4098 spin_unlock(&fs_info->balance_lock);
4106 if (fs_info->balance_ctl) {
4111 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4118 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4119 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4120 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4122 bctl->flags = bargs->flags;
4124 /* balance everything - no filters */
4125 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4128 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4135 * Ownership of bctl and exclusive operation goes to btrfs_balance.
4136 * bctl is freed in reset_balance_state, or, if restriper was paused
4137 * all the way until unmount, in free_fs_info. The flag should be
4138 * cleared after reset_balance_state.
4140 need_unlock = false;
4142 ret = btrfs_balance(fs_info, bctl, bargs);
4145 if ((ret == 0 || ret == -ECANCELED) && arg) {
4146 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4155 mutex_unlock(&fs_info->balance_mutex);
4157 btrfs_exclop_finish(fs_info);
4159 mnt_drop_write_file(file);
4163 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
4165 if (!capable(CAP_SYS_ADMIN))
4169 case BTRFS_BALANCE_CTL_PAUSE:
4170 return btrfs_pause_balance(fs_info);
4171 case BTRFS_BALANCE_CTL_CANCEL:
4172 return btrfs_cancel_balance(fs_info);
4178 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
4181 struct btrfs_ioctl_balance_args *bargs;
4184 if (!capable(CAP_SYS_ADMIN))
4187 mutex_lock(&fs_info->balance_mutex);
4188 if (!fs_info->balance_ctl) {
4193 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4199 btrfs_update_ioctl_balance_args(fs_info, bargs);
4201 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4206 mutex_unlock(&fs_info->balance_mutex);
4210 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4212 struct inode *inode = file_inode(file);
4213 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4214 struct btrfs_ioctl_quota_ctl_args *sa;
4217 if (!capable(CAP_SYS_ADMIN))
4220 ret = mnt_want_write_file(file);
4224 sa = memdup_user(arg, sizeof(*sa));
4230 down_write(&fs_info->subvol_sem);
4233 case BTRFS_QUOTA_CTL_ENABLE:
4234 ret = btrfs_quota_enable(fs_info);
4236 case BTRFS_QUOTA_CTL_DISABLE:
4237 ret = btrfs_quota_disable(fs_info);
4245 up_write(&fs_info->subvol_sem);
4247 mnt_drop_write_file(file);
4251 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4253 struct inode *inode = file_inode(file);
4254 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4255 struct btrfs_root *root = BTRFS_I(inode)->root;
4256 struct btrfs_ioctl_qgroup_assign_args *sa;
4257 struct btrfs_trans_handle *trans;
4261 if (!capable(CAP_SYS_ADMIN))
4264 ret = mnt_want_write_file(file);
4268 sa = memdup_user(arg, sizeof(*sa));
4274 trans = btrfs_join_transaction(root);
4275 if (IS_ERR(trans)) {
4276 ret = PTR_ERR(trans);
4281 ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
4283 ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
4286 /* update qgroup status and info */
4287 err = btrfs_run_qgroups(trans);
4289 btrfs_handle_fs_error(fs_info, err,
4290 "failed to update qgroup status and info");
4291 err = btrfs_end_transaction(trans);
4298 mnt_drop_write_file(file);
4302 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4304 struct inode *inode = file_inode(file);
4305 struct btrfs_root *root = BTRFS_I(inode)->root;
4306 struct btrfs_ioctl_qgroup_create_args *sa;
4307 struct btrfs_trans_handle *trans;
4311 if (!capable(CAP_SYS_ADMIN))
4314 ret = mnt_want_write_file(file);
4318 sa = memdup_user(arg, sizeof(*sa));
4324 if (!sa->qgroupid) {
4329 trans = btrfs_join_transaction(root);
4330 if (IS_ERR(trans)) {
4331 ret = PTR_ERR(trans);
4336 ret = btrfs_create_qgroup(trans, sa->qgroupid);
4338 ret = btrfs_remove_qgroup(trans, sa->qgroupid);
4341 err = btrfs_end_transaction(trans);
4348 mnt_drop_write_file(file);
4352 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4354 struct inode *inode = file_inode(file);
4355 struct btrfs_root *root = BTRFS_I(inode)->root;
4356 struct btrfs_ioctl_qgroup_limit_args *sa;
4357 struct btrfs_trans_handle *trans;
4362 if (!capable(CAP_SYS_ADMIN))
4365 ret = mnt_want_write_file(file);
4369 sa = memdup_user(arg, sizeof(*sa));
4375 trans = btrfs_join_transaction(root);
4376 if (IS_ERR(trans)) {
4377 ret = PTR_ERR(trans);
4381 qgroupid = sa->qgroupid;
4383 /* take the current subvol as qgroup */
4384 qgroupid = root->root_key.objectid;
4387 ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
4389 err = btrfs_end_transaction(trans);
4396 mnt_drop_write_file(file);
4400 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4402 struct inode *inode = file_inode(file);
4403 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4404 struct btrfs_ioctl_quota_rescan_args *qsa;
4407 if (!capable(CAP_SYS_ADMIN))
4410 ret = mnt_want_write_file(file);
4414 qsa = memdup_user(arg, sizeof(*qsa));
4425 ret = btrfs_qgroup_rescan(fs_info);
4430 mnt_drop_write_file(file);
4434 static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
4437 struct btrfs_ioctl_quota_rescan_args *qsa;
4440 if (!capable(CAP_SYS_ADMIN))
4443 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
4447 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4449 qsa->progress = fs_info->qgroup_rescan_progress.objectid;
4452 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4459 static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info,
4462 if (!capable(CAP_SYS_ADMIN))
4465 return btrfs_qgroup_wait_for_completion(fs_info, true);
4468 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4469 struct btrfs_ioctl_received_subvol_args *sa)
4471 struct inode *inode = file_inode(file);
4472 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4473 struct btrfs_root *root = BTRFS_I(inode)->root;
4474 struct btrfs_root_item *root_item = &root->root_item;
4475 struct btrfs_trans_handle *trans;
4476 struct timespec64 ct = current_time(inode);
4478 int received_uuid_changed;
4480 if (!inode_owner_or_capable(inode))
4483 ret = mnt_want_write_file(file);
4487 down_write(&fs_info->subvol_sem);
4489 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
4494 if (btrfs_root_readonly(root)) {
4501 * 2 - uuid items (received uuid + subvol uuid)
4503 trans = btrfs_start_transaction(root, 3);
4504 if (IS_ERR(trans)) {
4505 ret = PTR_ERR(trans);
4510 sa->rtransid = trans->transid;
4511 sa->rtime.sec = ct.tv_sec;
4512 sa->rtime.nsec = ct.tv_nsec;
4514 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4516 if (received_uuid_changed &&
4517 !btrfs_is_empty_uuid(root_item->received_uuid)) {
4518 ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
4519 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4520 root->root_key.objectid);
4521 if (ret && ret != -ENOENT) {
4522 btrfs_abort_transaction(trans, ret);
4523 btrfs_end_transaction(trans);
4527 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4528 btrfs_set_root_stransid(root_item, sa->stransid);
4529 btrfs_set_root_rtransid(root_item, sa->rtransid);
4530 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4531 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4532 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4533 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4535 ret = btrfs_update_root(trans, fs_info->tree_root,
4536 &root->root_key, &root->root_item);
4538 btrfs_end_transaction(trans);
4541 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4542 ret = btrfs_uuid_tree_add(trans, sa->uuid,
4543 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4544 root->root_key.objectid);
4545 if (ret < 0 && ret != -EEXIST) {
4546 btrfs_abort_transaction(trans, ret);
4547 btrfs_end_transaction(trans);
4551 ret = btrfs_commit_transaction(trans);
4553 up_write(&fs_info->subvol_sem);
4554 mnt_drop_write_file(file);
4559 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4562 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4563 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4566 args32 = memdup_user(arg, sizeof(*args32));
4568 return PTR_ERR(args32);
4570 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
4576 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4577 args64->stransid = args32->stransid;
4578 args64->rtransid = args32->rtransid;
4579 args64->stime.sec = args32->stime.sec;
4580 args64->stime.nsec = args32->stime.nsec;
4581 args64->rtime.sec = args32->rtime.sec;
4582 args64->rtime.nsec = args32->rtime.nsec;
4583 args64->flags = args32->flags;
4585 ret = _btrfs_ioctl_set_received_subvol(file, args64);
4589 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4590 args32->stransid = args64->stransid;
4591 args32->rtransid = args64->rtransid;
4592 args32->stime.sec = args64->stime.sec;
4593 args32->stime.nsec = args64->stime.nsec;
4594 args32->rtime.sec = args64->rtime.sec;
4595 args32->rtime.nsec = args64->rtime.nsec;
4596 args32->flags = args64->flags;
4598 ret = copy_to_user(arg, args32, sizeof(*args32));
4609 static long btrfs_ioctl_set_received_subvol(struct file *file,
4612 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4615 sa = memdup_user(arg, sizeof(*sa));
4619 ret = _btrfs_ioctl_set_received_subvol(file, sa);
4624 ret = copy_to_user(arg, sa, sizeof(*sa));
4633 static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
4638 char label[BTRFS_LABEL_SIZE];
4640 spin_lock(&fs_info->super_lock);
4641 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4642 spin_unlock(&fs_info->super_lock);
4644 len = strnlen(label, BTRFS_LABEL_SIZE);
4646 if (len == BTRFS_LABEL_SIZE) {
4648 "label is too long, return the first %zu bytes",
4652 ret = copy_to_user(arg, label, len);
4654 return ret ? -EFAULT : 0;
4657 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4659 struct inode *inode = file_inode(file);
4660 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4661 struct btrfs_root *root = BTRFS_I(inode)->root;
4662 struct btrfs_super_block *super_block = fs_info->super_copy;
4663 struct btrfs_trans_handle *trans;
4664 char label[BTRFS_LABEL_SIZE];
4667 if (!capable(CAP_SYS_ADMIN))
4670 if (copy_from_user(label, arg, sizeof(label)))
4673 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4675 "unable to set label with more than %d bytes",
4676 BTRFS_LABEL_SIZE - 1);
4680 ret = mnt_want_write_file(file);
4684 trans = btrfs_start_transaction(root, 0);
4685 if (IS_ERR(trans)) {
4686 ret = PTR_ERR(trans);
4690 spin_lock(&fs_info->super_lock);
4691 strcpy(super_block->label, label);
4692 spin_unlock(&fs_info->super_lock);
4693 ret = btrfs_commit_transaction(trans);
4696 mnt_drop_write_file(file);
4700 #define INIT_FEATURE_FLAGS(suffix) \
4701 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4702 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4703 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4705 int btrfs_ioctl_get_supported_features(void __user *arg)
4707 static const struct btrfs_ioctl_feature_flags features[3] = {
4708 INIT_FEATURE_FLAGS(SUPP),
4709 INIT_FEATURE_FLAGS(SAFE_SET),
4710 INIT_FEATURE_FLAGS(SAFE_CLEAR)
4713 if (copy_to_user(arg, &features, sizeof(features)))
4719 static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
4722 struct btrfs_super_block *super_block = fs_info->super_copy;
4723 struct btrfs_ioctl_feature_flags features;
4725 features.compat_flags = btrfs_super_compat_flags(super_block);
4726 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4727 features.incompat_flags = btrfs_super_incompat_flags(super_block);
4729 if (copy_to_user(arg, &features, sizeof(features)))
4735 static int check_feature_bits(struct btrfs_fs_info *fs_info,
4736 enum btrfs_feature_set set,
4737 u64 change_mask, u64 flags, u64 supported_flags,
4738 u64 safe_set, u64 safe_clear)
4740 const char *type = btrfs_feature_set_name(set);
4742 u64 disallowed, unsupported;
4743 u64 set_mask = flags & change_mask;
4744 u64 clear_mask = ~flags & change_mask;
4746 unsupported = set_mask & ~supported_flags;
4748 names = btrfs_printable_features(set, unsupported);
4751 "this kernel does not support the %s feature bit%s",
4752 names, strchr(names, ',') ? "s" : "");
4756 "this kernel does not support %s bits 0x%llx",
4761 disallowed = set_mask & ~safe_set;
4763 names = btrfs_printable_features(set, disallowed);
4766 "can't set the %s feature bit%s while mounted",
4767 names, strchr(names, ',') ? "s" : "");
4771 "can't set %s bits 0x%llx while mounted",
4776 disallowed = clear_mask & ~safe_clear;
4778 names = btrfs_printable_features(set, disallowed);
4781 "can't clear the %s feature bit%s while mounted",
4782 names, strchr(names, ',') ? "s" : "");
4786 "can't clear %s bits 0x%llx while mounted",
4794 #define check_feature(fs_info, change_mask, flags, mask_base) \
4795 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
4796 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
4797 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
4798 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4800 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4802 struct inode *inode = file_inode(file);
4803 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4804 struct btrfs_root *root = BTRFS_I(inode)->root;
4805 struct btrfs_super_block *super_block = fs_info->super_copy;
4806 struct btrfs_ioctl_feature_flags flags[2];
4807 struct btrfs_trans_handle *trans;
4811 if (!capable(CAP_SYS_ADMIN))
4814 if (copy_from_user(flags, arg, sizeof(flags)))
4818 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4819 !flags[0].incompat_flags)
4822 ret = check_feature(fs_info, flags[0].compat_flags,
4823 flags[1].compat_flags, COMPAT);
4827 ret = check_feature(fs_info, flags[0].compat_ro_flags,
4828 flags[1].compat_ro_flags, COMPAT_RO);
4832 ret = check_feature(fs_info, flags[0].incompat_flags,
4833 flags[1].incompat_flags, INCOMPAT);
4837 ret = mnt_want_write_file(file);
4841 trans = btrfs_start_transaction(root, 0);
4842 if (IS_ERR(trans)) {
4843 ret = PTR_ERR(trans);
4844 goto out_drop_write;
4847 spin_lock(&fs_info->super_lock);
4848 newflags = btrfs_super_compat_flags(super_block);
4849 newflags |= flags[0].compat_flags & flags[1].compat_flags;
4850 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4851 btrfs_set_super_compat_flags(super_block, newflags);
4853 newflags = btrfs_super_compat_ro_flags(super_block);
4854 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4855 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4856 btrfs_set_super_compat_ro_flags(super_block, newflags);
4858 newflags = btrfs_super_incompat_flags(super_block);
4859 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4860 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4861 btrfs_set_super_incompat_flags(super_block, newflags);
4862 spin_unlock(&fs_info->super_lock);
4864 ret = btrfs_commit_transaction(trans);
4866 mnt_drop_write_file(file);
4871 static int _btrfs_ioctl_send(struct file *file, void __user *argp, bool compat)
4873 struct btrfs_ioctl_send_args *arg;
4877 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
4878 struct btrfs_ioctl_send_args_32 args32;
4880 ret = copy_from_user(&args32, argp, sizeof(args32));
4883 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
4886 arg->send_fd = args32.send_fd;
4887 arg->clone_sources_count = args32.clone_sources_count;
4888 arg->clone_sources = compat_ptr(args32.clone_sources);
4889 arg->parent_root = args32.parent_root;
4890 arg->flags = args32.flags;
4891 memcpy(arg->reserved, args32.reserved,
4892 sizeof(args32.reserved));
4897 arg = memdup_user(argp, sizeof(*arg));
4899 return PTR_ERR(arg);
4901 ret = btrfs_ioctl_send(file, arg);
4906 long btrfs_ioctl(struct file *file, unsigned int
4907 cmd, unsigned long arg)
4909 struct inode *inode = file_inode(file);
4910 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4911 struct btrfs_root *root = BTRFS_I(inode)->root;
4912 void __user *argp = (void __user *)arg;
4915 case FS_IOC_GETFLAGS:
4916 return btrfs_ioctl_getflags(file, argp);
4917 case FS_IOC_SETFLAGS:
4918 return btrfs_ioctl_setflags(file, argp);
4919 case FS_IOC_GETVERSION:
4920 return btrfs_ioctl_getversion(file, argp);
4921 case FS_IOC_GETFSLABEL:
4922 return btrfs_ioctl_get_fslabel(fs_info, argp);
4923 case FS_IOC_SETFSLABEL:
4924 return btrfs_ioctl_set_fslabel(file, argp);
4926 return btrfs_ioctl_fitrim(fs_info, argp);
4927 case BTRFS_IOC_SNAP_CREATE:
4928 return btrfs_ioctl_snap_create(file, argp, 0);
4929 case BTRFS_IOC_SNAP_CREATE_V2:
4930 return btrfs_ioctl_snap_create_v2(file, argp, 0);
4931 case BTRFS_IOC_SUBVOL_CREATE:
4932 return btrfs_ioctl_snap_create(file, argp, 1);
4933 case BTRFS_IOC_SUBVOL_CREATE_V2:
4934 return btrfs_ioctl_snap_create_v2(file, argp, 1);
4935 case BTRFS_IOC_SNAP_DESTROY:
4936 return btrfs_ioctl_snap_destroy(file, argp, false);
4937 case BTRFS_IOC_SNAP_DESTROY_V2:
4938 return btrfs_ioctl_snap_destroy(file, argp, true);
4939 case BTRFS_IOC_SUBVOL_GETFLAGS:
4940 return btrfs_ioctl_subvol_getflags(file, argp);
4941 case BTRFS_IOC_SUBVOL_SETFLAGS:
4942 return btrfs_ioctl_subvol_setflags(file, argp);
4943 case BTRFS_IOC_DEFAULT_SUBVOL:
4944 return btrfs_ioctl_default_subvol(file, argp);
4945 case BTRFS_IOC_DEFRAG:
4946 return btrfs_ioctl_defrag(file, NULL);
4947 case BTRFS_IOC_DEFRAG_RANGE:
4948 return btrfs_ioctl_defrag(file, argp);
4949 case BTRFS_IOC_RESIZE:
4950 return btrfs_ioctl_resize(file, argp);
4951 case BTRFS_IOC_ADD_DEV:
4952 return btrfs_ioctl_add_dev(fs_info, argp);
4953 case BTRFS_IOC_RM_DEV:
4954 return btrfs_ioctl_rm_dev(file, argp);
4955 case BTRFS_IOC_RM_DEV_V2:
4956 return btrfs_ioctl_rm_dev_v2(file, argp);
4957 case BTRFS_IOC_FS_INFO:
4958 return btrfs_ioctl_fs_info(fs_info, argp);
4959 case BTRFS_IOC_DEV_INFO:
4960 return btrfs_ioctl_dev_info(fs_info, argp);
4961 case BTRFS_IOC_BALANCE:
4962 return btrfs_ioctl_balance(file, NULL);
4963 case BTRFS_IOC_TREE_SEARCH:
4964 return btrfs_ioctl_tree_search(file, argp);
4965 case BTRFS_IOC_TREE_SEARCH_V2:
4966 return btrfs_ioctl_tree_search_v2(file, argp);
4967 case BTRFS_IOC_INO_LOOKUP:
4968 return btrfs_ioctl_ino_lookup(file, argp);
4969 case BTRFS_IOC_INO_PATHS:
4970 return btrfs_ioctl_ino_to_path(root, argp);
4971 case BTRFS_IOC_LOGICAL_INO:
4972 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
4973 case BTRFS_IOC_LOGICAL_INO_V2:
4974 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
4975 case BTRFS_IOC_SPACE_INFO:
4976 return btrfs_ioctl_space_info(fs_info, argp);
4977 case BTRFS_IOC_SYNC: {
4980 ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false);
4983 ret = btrfs_sync_fs(inode->i_sb, 1);
4985 * The transaction thread may want to do more work,
4986 * namely it pokes the cleaner kthread that will start
4987 * processing uncleaned subvols.
4989 wake_up_process(fs_info->transaction_kthread);
4992 case BTRFS_IOC_START_SYNC:
4993 return btrfs_ioctl_start_sync(root, argp);
4994 case BTRFS_IOC_WAIT_SYNC:
4995 return btrfs_ioctl_wait_sync(fs_info, argp);
4996 case BTRFS_IOC_SCRUB:
4997 return btrfs_ioctl_scrub(file, argp);
4998 case BTRFS_IOC_SCRUB_CANCEL:
4999 return btrfs_ioctl_scrub_cancel(fs_info);
5000 case BTRFS_IOC_SCRUB_PROGRESS:
5001 return btrfs_ioctl_scrub_progress(fs_info, argp);
5002 case BTRFS_IOC_BALANCE_V2:
5003 return btrfs_ioctl_balance(file, argp);
5004 case BTRFS_IOC_BALANCE_CTL:
5005 return btrfs_ioctl_balance_ctl(fs_info, arg);
5006 case BTRFS_IOC_BALANCE_PROGRESS:
5007 return btrfs_ioctl_balance_progress(fs_info, argp);
5008 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5009 return btrfs_ioctl_set_received_subvol(file, argp);
5011 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5012 return btrfs_ioctl_set_received_subvol_32(file, argp);
5014 case BTRFS_IOC_SEND:
5015 return _btrfs_ioctl_send(file, argp, false);
5016 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5017 case BTRFS_IOC_SEND_32:
5018 return _btrfs_ioctl_send(file, argp, true);
5020 case BTRFS_IOC_GET_DEV_STATS:
5021 return btrfs_ioctl_get_dev_stats(fs_info, argp);
5022 case BTRFS_IOC_QUOTA_CTL:
5023 return btrfs_ioctl_quota_ctl(file, argp);
5024 case BTRFS_IOC_QGROUP_ASSIGN:
5025 return btrfs_ioctl_qgroup_assign(file, argp);
5026 case BTRFS_IOC_QGROUP_CREATE:
5027 return btrfs_ioctl_qgroup_create(file, argp);
5028 case BTRFS_IOC_QGROUP_LIMIT:
5029 return btrfs_ioctl_qgroup_limit(file, argp);
5030 case BTRFS_IOC_QUOTA_RESCAN:
5031 return btrfs_ioctl_quota_rescan(file, argp);
5032 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5033 return btrfs_ioctl_quota_rescan_status(fs_info, argp);
5034 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5035 return btrfs_ioctl_quota_rescan_wait(fs_info, argp);
5036 case BTRFS_IOC_DEV_REPLACE:
5037 return btrfs_ioctl_dev_replace(fs_info, argp);
5038 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5039 return btrfs_ioctl_get_supported_features(argp);
5040 case BTRFS_IOC_GET_FEATURES:
5041 return btrfs_ioctl_get_features(fs_info, argp);
5042 case BTRFS_IOC_SET_FEATURES:
5043 return btrfs_ioctl_set_features(file, argp);
5044 case FS_IOC_FSGETXATTR:
5045 return btrfs_ioctl_fsgetxattr(file, argp);
5046 case FS_IOC_FSSETXATTR:
5047 return btrfs_ioctl_fssetxattr(file, argp);
5048 case BTRFS_IOC_GET_SUBVOL_INFO:
5049 return btrfs_ioctl_get_subvol_info(file, argp);
5050 case BTRFS_IOC_GET_SUBVOL_ROOTREF:
5051 return btrfs_ioctl_get_subvol_rootref(file, argp);
5052 case BTRFS_IOC_INO_LOOKUP_USER:
5053 return btrfs_ioctl_ino_lookup_user(file, argp);
5059 #ifdef CONFIG_COMPAT
5060 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5063 * These all access 32-bit values anyway so no further
5064 * handling is necessary.
5067 case FS_IOC32_GETFLAGS:
5068 cmd = FS_IOC_GETFLAGS;
5070 case FS_IOC32_SETFLAGS:
5071 cmd = FS_IOC_SETFLAGS;
5073 case FS_IOC32_GETVERSION:
5074 cmd = FS_IOC_GETVERSION;
5078 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));